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  * @net: pointer to namespace struct
86  * @peer_session: link session # being used by peer end of link
87  * @peer_bearer_id: bearer id used by link's peer endpoint
88  * @bearer_id: local bearer id used by link
89  * @tolerance: minimum link continuity loss needed to reset link [in ms]
90  * @abort_limit: # of unacknowledged continuity probes needed to reset link
91  * @state: current state of link FSM
92  * @peer_caps: bitmap describing capabilities of peer node
93  * @silent_intv_cnt: # of timer intervals without any reception from peer
94  * @priority: current link priority
95  * @net_plane: current link network plane ('A' through 'H')
96  * @mon_state: cookie with information needed by link monitor
97  * @mtu: current maximum packet size for this link
98  * @advertised_mtu: advertised own mtu when link is being established
99  * @backlogq: queue for messages waiting to be sent
100  * @ackers: # of peers that needs to ack each packet before it can be released
101  * @acked: # last packet acked by a certain peer. Used for broadcast.
102  * @rcv_nxt: next sequence number to expect for inbound messages
103  * @inputq: buffer queue for messages to be delivered upwards
104  * @namedq: buffer queue for name table messages to be delivered upwards
105  * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
106  * @reasm_buf: head of partially reassembled inbound message fragments
107  * @stats: collects statistics regarding link activity
108  * @session: session to be used by link
109  * @snd_nxt_state: next send seq number
110  * @rcv_nxt_state: next rcv seq number
111  * @in_session: have received ACTIVATE_MSG from peer
112  * @active: link is active
113  * @if_name: associated interface name
114  * @rst_cnt: link reset counter
115  * @drop_point: seq number for failover handling (FIXME)
116  * @failover_reasm_skb: saved failover msg ptr (FIXME)
117  * @failover_deferdq: deferred message queue for failover processing (FIXME)
118  * @transmq: the link's transmit queue
119  * @backlog: link's backlog by priority (importance)
120  * @snd_nxt: next sequence number to be used
121  * @rcv_unacked: # messages read by user, but not yet acked back to peer
122  * @deferdq: deferred receive queue
123  * @window: sliding window size for congestion handling
124  * @min_win: minimal send window to be used by link
125  * @ssthresh: slow start threshold for congestion handling
126  * @max_win: maximal send window to be used by link
127  * @cong_acks: congestion acks for congestion avoidance (FIXME)
128  * @checkpoint: seq number for congestion window size handling
129  * @reasm_tnlmsg: fragmentation/reassembly area for tunnel protocol message
130  * @last_gap: last gap ack blocks for bcast (FIXME)
131  * @last_ga: ptr to gap ack blocks
132  * @bc_rcvlink: the peer specific link used for broadcast reception
133  * @bc_sndlink: the namespace global link used for broadcast sending
134  * @nack_state: bcast nack state
135  * @bc_peer_is_up: peer has acked the bcast init msg
136  */
137 struct tipc_link {
138 	u32 addr;
139 	char name[TIPC_MAX_LINK_NAME];
140 	struct net *net;
141 
142 	/* Management and link supervision data */
143 	u16 peer_session;
144 	u16 session;
145 	u16 snd_nxt_state;
146 	u16 rcv_nxt_state;
147 	u32 peer_bearer_id;
148 	u32 bearer_id;
149 	u32 tolerance;
150 	u32 abort_limit;
151 	u32 state;
152 	u16 peer_caps;
153 	bool in_session;
154 	bool active;
155 	u32 silent_intv_cnt;
156 	char if_name[TIPC_MAX_IF_NAME];
157 	u32 priority;
158 	char net_plane;
159 	struct tipc_mon_state mon_state;
160 	u16 rst_cnt;
161 
162 	/* Failover/synch */
163 	u16 drop_point;
164 	struct sk_buff *failover_reasm_skb;
165 	struct sk_buff_head failover_deferdq;
166 
167 	/* Max packet negotiation */
168 	u16 mtu;
169 	u16 advertised_mtu;
170 
171 	/* Sending */
172 	struct sk_buff_head transmq;
173 	struct sk_buff_head backlogq;
174 	struct {
175 		u16 len;
176 		u16 limit;
177 		struct sk_buff *target_bskb;
178 	} backlog[5];
179 	u16 snd_nxt;
180 
181 	/* Reception */
182 	u16 rcv_nxt;
183 	u32 rcv_unacked;
184 	struct sk_buff_head deferdq;
185 	struct sk_buff_head *inputq;
186 	struct sk_buff_head *namedq;
187 
188 	/* Congestion handling */
189 	struct sk_buff_head wakeupq;
190 	u16 window;
191 	u16 min_win;
192 	u16 ssthresh;
193 	u16 max_win;
194 	u16 cong_acks;
195 	u16 checkpoint;
196 
197 	/* Fragmentation/reassembly */
198 	struct sk_buff *reasm_buf;
199 	struct sk_buff *reasm_tnlmsg;
200 
201 	/* Broadcast */
202 	u16 ackers;
203 	u16 acked;
204 	u16 last_gap;
205 	struct tipc_gap_ack_blks *last_ga;
206 	struct tipc_link *bc_rcvlink;
207 	struct tipc_link *bc_sndlink;
208 	u8 nack_state;
209 	bool bc_peer_is_up;
210 
211 	/* Statistics */
212 	struct tipc_stats stats;
213 };
214 
215 /*
216  * Error message prefixes
217  */
218 static const char *link_co_err = "Link tunneling error, ";
219 static const char *link_rst_msg = "Resetting link ";
220 
221 /* Send states for broadcast NACKs
222  */
223 enum {
224 	BC_NACK_SND_CONDITIONAL,
225 	BC_NACK_SND_UNCONDITIONAL,
226 	BC_NACK_SND_SUPPRESS,
227 };
228 
229 #define TIPC_BC_RETR_LIM  (jiffies + msecs_to_jiffies(10))
230 #define TIPC_UC_RETR_TIME (jiffies + msecs_to_jiffies(1))
231 
232 /* Link FSM states:
233  */
234 enum {
235 	LINK_ESTABLISHED     = 0xe,
236 	LINK_ESTABLISHING    = 0xe  << 4,
237 	LINK_RESET           = 0x1  << 8,
238 	LINK_RESETTING       = 0x2  << 12,
239 	LINK_PEER_RESET      = 0xd  << 16,
240 	LINK_FAILINGOVER     = 0xf  << 20,
241 	LINK_SYNCHING        = 0xc  << 24
242 };
243 
244 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
245 			       struct sk_buff_head *xmitq);
246 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
247 				      bool probe_reply, u16 rcvgap,
248 				      int tolerance, int priority,
249 				      struct sk_buff_head *xmitq);
250 static void link_print(struct tipc_link *l, const char *str);
251 static int tipc_link_build_nack_msg(struct tipc_link *l,
252 				    struct sk_buff_head *xmitq);
253 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
254 					struct sk_buff_head *xmitq);
255 static u8 __tipc_build_gap_ack_blks(struct tipc_gap_ack_blks *ga,
256 				    struct tipc_link *l, u8 start_index);
257 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, struct tipc_msg *hdr);
258 static int tipc_link_advance_transmq(struct tipc_link *l, struct tipc_link *r,
259 				     u16 acked, u16 gap,
260 				     struct tipc_gap_ack_blks *ga,
261 				     struct sk_buff_head *xmitq,
262 				     bool *retransmitted, int *rc);
263 static void tipc_link_update_cwin(struct tipc_link *l, int released,
264 				  bool retransmitted);
265 /*
266  *  Simple non-static link routines (i.e. referenced outside this file)
267  */
tipc_link_is_up(struct tipc_link * l)268 bool tipc_link_is_up(struct tipc_link *l)
269 {
270 	return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
271 }
272 
tipc_link_peer_is_down(struct tipc_link * l)273 bool tipc_link_peer_is_down(struct tipc_link *l)
274 {
275 	return l->state == LINK_PEER_RESET;
276 }
277 
tipc_link_is_reset(struct tipc_link * l)278 bool tipc_link_is_reset(struct tipc_link *l)
279 {
280 	return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
281 }
282 
tipc_link_is_establishing(struct tipc_link * l)283 bool tipc_link_is_establishing(struct tipc_link *l)
284 {
285 	return l->state == LINK_ESTABLISHING;
286 }
287 
tipc_link_is_synching(struct tipc_link * l)288 bool tipc_link_is_synching(struct tipc_link *l)
289 {
290 	return l->state == LINK_SYNCHING;
291 }
292 
tipc_link_is_failingover(struct tipc_link * l)293 bool tipc_link_is_failingover(struct tipc_link *l)
294 {
295 	return l->state == LINK_FAILINGOVER;
296 }
297 
tipc_link_is_blocked(struct tipc_link * l)298 bool tipc_link_is_blocked(struct tipc_link *l)
299 {
300 	return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
301 }
302 
link_is_bc_sndlink(struct tipc_link * l)303 static bool link_is_bc_sndlink(struct tipc_link *l)
304 {
305 	return !l->bc_sndlink;
306 }
307 
link_is_bc_rcvlink(struct tipc_link * l)308 static bool link_is_bc_rcvlink(struct tipc_link *l)
309 {
310 	return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
311 }
312 
tipc_link_set_active(struct tipc_link * l,bool active)313 void tipc_link_set_active(struct tipc_link *l, bool active)
314 {
315 	l->active = active;
316 }
317 
tipc_link_id(struct tipc_link * l)318 u32 tipc_link_id(struct tipc_link *l)
319 {
320 	return l->peer_bearer_id << 16 | l->bearer_id;
321 }
322 
tipc_link_min_win(struct tipc_link * l)323 int tipc_link_min_win(struct tipc_link *l)
324 {
325 	return l->min_win;
326 }
327 
tipc_link_max_win(struct tipc_link * l)328 int tipc_link_max_win(struct tipc_link *l)
329 {
330 	return l->max_win;
331 }
332 
tipc_link_prio(struct tipc_link * l)333 int tipc_link_prio(struct tipc_link *l)
334 {
335 	return l->priority;
336 }
337 
tipc_link_tolerance(struct tipc_link * l)338 unsigned long tipc_link_tolerance(struct tipc_link *l)
339 {
340 	return l->tolerance;
341 }
342 
tipc_link_inputq(struct tipc_link * l)343 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
344 {
345 	return l->inputq;
346 }
347 
tipc_link_plane(struct tipc_link * l)348 char tipc_link_plane(struct tipc_link *l)
349 {
350 	return l->net_plane;
351 }
352 
tipc_link_net(struct tipc_link * l)353 struct net *tipc_link_net(struct tipc_link *l)
354 {
355 	return l->net;
356 }
357 
tipc_link_update_caps(struct tipc_link * l,u16 capabilities)358 void tipc_link_update_caps(struct tipc_link *l, u16 capabilities)
359 {
360 	l->peer_caps = capabilities;
361 }
362 
tipc_link_add_bc_peer(struct tipc_link * snd_l,struct tipc_link * uc_l,struct sk_buff_head * xmitq)363 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
364 			   struct tipc_link *uc_l,
365 			   struct sk_buff_head *xmitq)
366 {
367 	struct tipc_link *rcv_l = uc_l->bc_rcvlink;
368 
369 	snd_l->ackers++;
370 	rcv_l->acked = snd_l->snd_nxt - 1;
371 	snd_l->state = LINK_ESTABLISHED;
372 	tipc_link_build_bc_init_msg(uc_l, xmitq);
373 }
374 
tipc_link_remove_bc_peer(struct tipc_link * snd_l,struct tipc_link * rcv_l,struct sk_buff_head * xmitq)375 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
376 			      struct tipc_link *rcv_l,
377 			      struct sk_buff_head *xmitq)
378 {
379 	u16 ack = snd_l->snd_nxt - 1;
380 
381 	snd_l->ackers--;
382 	rcv_l->bc_peer_is_up = true;
383 	rcv_l->state = LINK_ESTABLISHED;
384 	tipc_link_bc_ack_rcv(rcv_l, ack, 0, NULL, xmitq, NULL);
385 	trace_tipc_link_reset(rcv_l, TIPC_DUMP_ALL, "bclink removed!");
386 	tipc_link_reset(rcv_l);
387 	rcv_l->state = LINK_RESET;
388 	if (!snd_l->ackers) {
389 		trace_tipc_link_reset(snd_l, TIPC_DUMP_ALL, "zero ackers!");
390 		tipc_link_reset(snd_l);
391 		snd_l->state = LINK_RESET;
392 		__skb_queue_purge(xmitq);
393 	}
394 }
395 
tipc_link_bc_peers(struct tipc_link * l)396 int tipc_link_bc_peers(struct tipc_link *l)
397 {
398 	return l->ackers;
399 }
400 
link_bc_rcv_gap(struct tipc_link * l)401 static u16 link_bc_rcv_gap(struct tipc_link *l)
402 {
403 	struct sk_buff *skb = skb_peek(&l->deferdq);
404 	u16 gap = 0;
405 
406 	if (more(l->snd_nxt, l->rcv_nxt))
407 		gap = l->snd_nxt - l->rcv_nxt;
408 	if (skb)
409 		gap = buf_seqno(skb) - l->rcv_nxt;
410 	return gap;
411 }
412 
tipc_link_set_mtu(struct tipc_link * l,int mtu)413 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
414 {
415 	l->mtu = mtu;
416 }
417 
tipc_link_mtu(struct tipc_link * l)418 int tipc_link_mtu(struct tipc_link *l)
419 {
420 	return l->mtu;
421 }
422 
tipc_link_mss(struct tipc_link * l)423 int tipc_link_mss(struct tipc_link *l)
424 {
425 #ifdef CONFIG_TIPC_CRYPTO
426 	return l->mtu - INT_H_SIZE - EMSG_OVERHEAD;
427 #else
428 	return l->mtu - INT_H_SIZE;
429 #endif
430 }
431 
tipc_link_rcv_nxt(struct tipc_link * l)432 u16 tipc_link_rcv_nxt(struct tipc_link *l)
433 {
434 	return l->rcv_nxt;
435 }
436 
tipc_link_acked(struct tipc_link * l)437 u16 tipc_link_acked(struct tipc_link *l)
438 {
439 	return l->acked;
440 }
441 
tipc_link_name(struct tipc_link * l)442 char *tipc_link_name(struct tipc_link *l)
443 {
444 	return l->name;
445 }
446 
tipc_link_state(struct tipc_link * l)447 u32 tipc_link_state(struct tipc_link *l)
448 {
449 	return l->state;
450 }
451 
452 /**
453  * tipc_link_create - create a new link
454  * @net: pointer to associated network namespace
455  * @if_name: associated interface name
456  * @bearer_id: id (index) of associated bearer
457  * @tolerance: link tolerance to be used by link
458  * @net_plane: network plane (A,B,c..) this link belongs to
459  * @mtu: mtu to be advertised by link
460  * @priority: priority to be used by link
461  * @min_win: minimal send window to be used by link
462  * @max_win: maximal send window to be used by link
463  * @session: session to be used by link
464  * @peer: node id of peer node
465  * @peer_caps: bitmap describing peer node capabilities
466  * @bc_sndlink: the namespace global link used for broadcast sending
467  * @bc_rcvlink: the peer specific link used for broadcast reception
468  * @inputq: queue to put messages ready for delivery
469  * @namedq: queue to put binding table update messages ready for delivery
470  * @link: return value, pointer to put the created link
471  * @self: local unicast link id
472  * @peer_id: 128-bit ID of peer
473  *
474  * Return: true if link was created, otherwise false
475  */
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)476 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
477 		      int tolerance, char net_plane, u32 mtu, int priority,
478 		      u32 min_win, u32 max_win, u32 session, u32 self,
479 		      u32 peer, u8 *peer_id, u16 peer_caps,
480 		      struct tipc_link *bc_sndlink,
481 		      struct tipc_link *bc_rcvlink,
482 		      struct sk_buff_head *inputq,
483 		      struct sk_buff_head *namedq,
484 		      struct tipc_link **link)
485 {
486 	char peer_str[NODE_ID_STR_LEN] = {0,};
487 	char self_str[NODE_ID_STR_LEN] = {0,};
488 	struct tipc_link *l;
489 
490 	l = kzalloc(sizeof(*l), GFP_ATOMIC);
491 	if (!l)
492 		return false;
493 	*link = l;
494 	l->session = session;
495 
496 	/* Set link name for unicast links only */
497 	if (peer_id) {
498 		tipc_nodeid2string(self_str, tipc_own_id(net));
499 		if (strlen(self_str) > 16)
500 			sprintf(self_str, "%x", self);
501 		tipc_nodeid2string(peer_str, peer_id);
502 		if (strlen(peer_str) > 16)
503 			sprintf(peer_str, "%x", peer);
504 	}
505 	/* Peer i/f name will be completed by reset/activate message */
506 	snprintf(l->name, sizeof(l->name), "%s:%s-%s:unknown",
507 		 self_str, if_name, peer_str);
508 
509 	strcpy(l->if_name, if_name);
510 	l->addr = peer;
511 	l->peer_caps = peer_caps;
512 	l->net = net;
513 	l->in_session = false;
514 	l->bearer_id = bearer_id;
515 	l->tolerance = tolerance;
516 	if (bc_rcvlink)
517 		bc_rcvlink->tolerance = tolerance;
518 	l->net_plane = net_plane;
519 	l->advertised_mtu = mtu;
520 	l->mtu = mtu;
521 	l->priority = priority;
522 	tipc_link_set_queue_limits(l, min_win, max_win);
523 	l->ackers = 1;
524 	l->bc_sndlink = bc_sndlink;
525 	l->bc_rcvlink = bc_rcvlink;
526 	l->inputq = inputq;
527 	l->namedq = namedq;
528 	l->state = LINK_RESETTING;
529 	__skb_queue_head_init(&l->transmq);
530 	__skb_queue_head_init(&l->backlogq);
531 	__skb_queue_head_init(&l->deferdq);
532 	__skb_queue_head_init(&l->failover_deferdq);
533 	skb_queue_head_init(&l->wakeupq);
534 	skb_queue_head_init(l->inputq);
535 	return true;
536 }
537 
538 /**
539  * tipc_link_bc_create - create new link to be used for broadcast
540  * @net: pointer to associated network namespace
541  * @mtu: mtu to be used initially if no peers
542  * @min_win: minimal send window to be used by link
543  * @max_win: maximal send window to be used by link
544  * @inputq: queue to put messages ready for delivery
545  * @namedq: queue to put binding table update messages ready for delivery
546  * @link: return value, pointer to put the created link
547  * @ownnode: identity of own node
548  * @peer: node id of peer node
549  * @peer_id: 128-bit ID of peer
550  * @peer_caps: bitmap describing peer node capabilities
551  * @bc_sndlink: the namespace global link used for broadcast sending
552  *
553  * Return: true if link was created, otherwise false
554  */
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)555 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer, u8 *peer_id,
556 			 int mtu, u32 min_win, u32 max_win, u16 peer_caps,
557 			 struct sk_buff_head *inputq,
558 			 struct sk_buff_head *namedq,
559 			 struct tipc_link *bc_sndlink,
560 			 struct tipc_link **link)
561 {
562 	struct tipc_link *l;
563 
564 	if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, min_win,
565 			      max_win, 0, ownnode, peer, NULL, peer_caps,
566 			      bc_sndlink, NULL, inputq, namedq, link))
567 		return false;
568 
569 	l = *link;
570 	if (peer_id) {
571 		char peer_str[NODE_ID_STR_LEN] = {0,};
572 
573 		tipc_nodeid2string(peer_str, peer_id);
574 		if (strlen(peer_str) > 16)
575 			sprintf(peer_str, "%x", peer);
576 		/* Broadcast receiver link name: "broadcast-link:<peer>" */
577 		snprintf(l->name, sizeof(l->name), "%s:%s", tipc_bclink_name,
578 			 peer_str);
579 	} else {
580 		strcpy(l->name, tipc_bclink_name);
581 	}
582 	trace_tipc_link_reset(l, TIPC_DUMP_ALL, "bclink created!");
583 	tipc_link_reset(l);
584 	l->state = LINK_RESET;
585 	l->ackers = 0;
586 	l->bc_rcvlink = l;
587 
588 	/* Broadcast send link is always up */
589 	if (link_is_bc_sndlink(l))
590 		l->state = LINK_ESTABLISHED;
591 
592 	/* Disable replicast if even a single peer doesn't support it */
593 	if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
594 		tipc_bcast_toggle_rcast(net, false);
595 
596 	return true;
597 }
598 
599 /**
600  * tipc_link_fsm_evt - link finite state machine
601  * @l: pointer to link
602  * @evt: state machine event to be processed
603  */
tipc_link_fsm_evt(struct tipc_link * l,int evt)604 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
605 {
606 	int rc = 0;
607 	int old_state = l->state;
608 
609 	switch (l->state) {
610 	case LINK_RESETTING:
611 		switch (evt) {
612 		case LINK_PEER_RESET_EVT:
613 			l->state = LINK_PEER_RESET;
614 			break;
615 		case LINK_RESET_EVT:
616 			l->state = LINK_RESET;
617 			break;
618 		case LINK_FAILURE_EVT:
619 		case LINK_FAILOVER_BEGIN_EVT:
620 		case LINK_ESTABLISH_EVT:
621 		case LINK_FAILOVER_END_EVT:
622 		case LINK_SYNCH_BEGIN_EVT:
623 		case LINK_SYNCH_END_EVT:
624 		default:
625 			goto illegal_evt;
626 		}
627 		break;
628 	case LINK_RESET:
629 		switch (evt) {
630 		case LINK_PEER_RESET_EVT:
631 			l->state = LINK_ESTABLISHING;
632 			break;
633 		case LINK_FAILOVER_BEGIN_EVT:
634 			l->state = LINK_FAILINGOVER;
635 			break;
636 		case LINK_FAILURE_EVT:
637 		case LINK_RESET_EVT:
638 		case LINK_ESTABLISH_EVT:
639 		case LINK_FAILOVER_END_EVT:
640 			break;
641 		case LINK_SYNCH_BEGIN_EVT:
642 		case LINK_SYNCH_END_EVT:
643 		default:
644 			goto illegal_evt;
645 		}
646 		break;
647 	case LINK_PEER_RESET:
648 		switch (evt) {
649 		case LINK_RESET_EVT:
650 			l->state = LINK_ESTABLISHING;
651 			break;
652 		case LINK_PEER_RESET_EVT:
653 		case LINK_ESTABLISH_EVT:
654 		case LINK_FAILURE_EVT:
655 			break;
656 		case LINK_SYNCH_BEGIN_EVT:
657 		case LINK_SYNCH_END_EVT:
658 		case LINK_FAILOVER_BEGIN_EVT:
659 		case LINK_FAILOVER_END_EVT:
660 		default:
661 			goto illegal_evt;
662 		}
663 		break;
664 	case LINK_FAILINGOVER:
665 		switch (evt) {
666 		case LINK_FAILOVER_END_EVT:
667 			l->state = LINK_RESET;
668 			break;
669 		case LINK_PEER_RESET_EVT:
670 		case LINK_RESET_EVT:
671 		case LINK_ESTABLISH_EVT:
672 		case LINK_FAILURE_EVT:
673 			break;
674 		case LINK_FAILOVER_BEGIN_EVT:
675 		case LINK_SYNCH_BEGIN_EVT:
676 		case LINK_SYNCH_END_EVT:
677 		default:
678 			goto illegal_evt;
679 		}
680 		break;
681 	case LINK_ESTABLISHING:
682 		switch (evt) {
683 		case LINK_ESTABLISH_EVT:
684 			l->state = LINK_ESTABLISHED;
685 			break;
686 		case LINK_FAILOVER_BEGIN_EVT:
687 			l->state = LINK_FAILINGOVER;
688 			break;
689 		case LINK_RESET_EVT:
690 			l->state = LINK_RESET;
691 			break;
692 		case LINK_FAILURE_EVT:
693 		case LINK_PEER_RESET_EVT:
694 		case LINK_SYNCH_BEGIN_EVT:
695 		case LINK_FAILOVER_END_EVT:
696 			break;
697 		case LINK_SYNCH_END_EVT:
698 		default:
699 			goto illegal_evt;
700 		}
701 		break;
702 	case LINK_ESTABLISHED:
703 		switch (evt) {
704 		case LINK_PEER_RESET_EVT:
705 			l->state = LINK_PEER_RESET;
706 			rc |= TIPC_LINK_DOWN_EVT;
707 			break;
708 		case LINK_FAILURE_EVT:
709 			l->state = LINK_RESETTING;
710 			rc |= TIPC_LINK_DOWN_EVT;
711 			break;
712 		case LINK_RESET_EVT:
713 			l->state = LINK_RESET;
714 			break;
715 		case LINK_ESTABLISH_EVT:
716 		case LINK_SYNCH_END_EVT:
717 			break;
718 		case LINK_SYNCH_BEGIN_EVT:
719 			l->state = LINK_SYNCHING;
720 			break;
721 		case LINK_FAILOVER_BEGIN_EVT:
722 		case LINK_FAILOVER_END_EVT:
723 		default:
724 			goto illegal_evt;
725 		}
726 		break;
727 	case LINK_SYNCHING:
728 		switch (evt) {
729 		case LINK_PEER_RESET_EVT:
730 			l->state = LINK_PEER_RESET;
731 			rc |= TIPC_LINK_DOWN_EVT;
732 			break;
733 		case LINK_FAILURE_EVT:
734 			l->state = LINK_RESETTING;
735 			rc |= TIPC_LINK_DOWN_EVT;
736 			break;
737 		case LINK_RESET_EVT:
738 			l->state = LINK_RESET;
739 			break;
740 		case LINK_ESTABLISH_EVT:
741 		case LINK_SYNCH_BEGIN_EVT:
742 			break;
743 		case LINK_SYNCH_END_EVT:
744 			l->state = LINK_ESTABLISHED;
745 			break;
746 		case LINK_FAILOVER_BEGIN_EVT:
747 		case LINK_FAILOVER_END_EVT:
748 		default:
749 			goto illegal_evt;
750 		}
751 		break;
752 	default:
753 		pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
754 	}
755 	trace_tipc_link_fsm(l->name, old_state, l->state, evt);
756 	return rc;
757 illegal_evt:
758 	pr_err("Illegal FSM event %x in state %x on link %s\n",
759 	       evt, l->state, l->name);
760 	trace_tipc_link_fsm(l->name, old_state, l->state, evt);
761 	return rc;
762 }
763 
764 /* link_profile_stats - update statistical profiling of traffic
765  */
link_profile_stats(struct tipc_link * l)766 static void link_profile_stats(struct tipc_link *l)
767 {
768 	struct sk_buff *skb;
769 	struct tipc_msg *msg;
770 	int length;
771 
772 	/* Update counters used in statistical profiling of send traffic */
773 	l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
774 	l->stats.queue_sz_counts++;
775 
776 	skb = skb_peek(&l->transmq);
777 	if (!skb)
778 		return;
779 	msg = buf_msg(skb);
780 	length = msg_size(msg);
781 
782 	if (msg_user(msg) == MSG_FRAGMENTER) {
783 		if (msg_type(msg) != FIRST_FRAGMENT)
784 			return;
785 		length = msg_size(msg_inner_hdr(msg));
786 	}
787 	l->stats.msg_lengths_total += length;
788 	l->stats.msg_length_counts++;
789 	if (length <= 64)
790 		l->stats.msg_length_profile[0]++;
791 	else if (length <= 256)
792 		l->stats.msg_length_profile[1]++;
793 	else if (length <= 1024)
794 		l->stats.msg_length_profile[2]++;
795 	else if (length <= 4096)
796 		l->stats.msg_length_profile[3]++;
797 	else if (length <= 16384)
798 		l->stats.msg_length_profile[4]++;
799 	else if (length <= 32768)
800 		l->stats.msg_length_profile[5]++;
801 	else
802 		l->stats.msg_length_profile[6]++;
803 }
804 
805 /**
806  * tipc_link_too_silent - check if link is "too silent"
807  * @l: tipc link to be checked
808  *
809  * Return: true if the link 'silent_intv_cnt' is about to reach the
810  * 'abort_limit' value, otherwise false
811  */
tipc_link_too_silent(struct tipc_link * l)812 bool tipc_link_too_silent(struct tipc_link *l)
813 {
814 	return (l->silent_intv_cnt + 2 > l->abort_limit);
815 }
816 
817 /* tipc_link_timeout - perform periodic task as instructed from node timeout
818  */
tipc_link_timeout(struct tipc_link * l,struct sk_buff_head * xmitq)819 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
820 {
821 	int mtyp = 0;
822 	int rc = 0;
823 	bool state = false;
824 	bool probe = false;
825 	bool setup = false;
826 	u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
827 	u16 bc_acked = l->bc_rcvlink->acked;
828 	struct tipc_mon_state *mstate = &l->mon_state;
829 
830 	trace_tipc_link_timeout(l, TIPC_DUMP_NONE, " ");
831 	trace_tipc_link_too_silent(l, TIPC_DUMP_ALL, " ");
832 	switch (l->state) {
833 	case LINK_ESTABLISHED:
834 	case LINK_SYNCHING:
835 		mtyp = STATE_MSG;
836 		link_profile_stats(l);
837 		tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
838 		if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
839 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
840 		state = bc_acked != bc_snt;
841 		state |= l->bc_rcvlink->rcv_unacked;
842 		state |= l->rcv_unacked;
843 		state |= !skb_queue_empty(&l->transmq);
844 		probe = mstate->probing;
845 		probe |= l->silent_intv_cnt;
846 		if (probe || mstate->monitoring)
847 			l->silent_intv_cnt++;
848 		probe |= !skb_queue_empty(&l->deferdq);
849 		if (l->snd_nxt == l->checkpoint) {
850 			tipc_link_update_cwin(l, 0, 0);
851 			probe = true;
852 		}
853 		l->checkpoint = l->snd_nxt;
854 		break;
855 	case LINK_RESET:
856 		setup = l->rst_cnt++ <= 4;
857 		setup |= !(l->rst_cnt % 16);
858 		mtyp = RESET_MSG;
859 		break;
860 	case LINK_ESTABLISHING:
861 		setup = true;
862 		mtyp = ACTIVATE_MSG;
863 		break;
864 	case LINK_PEER_RESET:
865 	case LINK_RESETTING:
866 	case LINK_FAILINGOVER:
867 		break;
868 	default:
869 		break;
870 	}
871 
872 	if (state || probe || setup)
873 		tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq);
874 
875 	return rc;
876 }
877 
878 /**
879  * link_schedule_user - schedule a message sender for wakeup after congestion
880  * @l: congested link
881  * @hdr: header of message that is being sent
882  * Create pseudo msg to send back to user when congestion abates
883  */
link_schedule_user(struct tipc_link * l,struct tipc_msg * hdr)884 static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
885 {
886 	u32 dnode = tipc_own_addr(l->net);
887 	u32 dport = msg_origport(hdr);
888 	struct sk_buff *skb;
889 
890 	/* Create and schedule wakeup pseudo message */
891 	skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
892 			      dnode, l->addr, dport, 0, 0);
893 	if (!skb)
894 		return -ENOBUFS;
895 	msg_set_dest_droppable(buf_msg(skb), true);
896 	TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
897 	skb_queue_tail(&l->wakeupq, skb);
898 	l->stats.link_congs++;
899 	trace_tipc_link_conges(l, TIPC_DUMP_ALL, "wakeup scheduled!");
900 	return -ELINKCONG;
901 }
902 
903 /**
904  * link_prepare_wakeup - prepare users for wakeup after congestion
905  * @l: congested link
906  * Wake up a number of waiting users, as permitted by available space
907  * in the send queue
908  */
link_prepare_wakeup(struct tipc_link * l)909 static void link_prepare_wakeup(struct tipc_link *l)
910 {
911 	struct sk_buff_head *wakeupq = &l->wakeupq;
912 	struct sk_buff_head *inputq = l->inputq;
913 	struct sk_buff *skb, *tmp;
914 	struct sk_buff_head tmpq;
915 	int avail[5] = {0,};
916 	int imp = 0;
917 
918 	__skb_queue_head_init(&tmpq);
919 
920 	for (; imp <= TIPC_SYSTEM_IMPORTANCE; imp++)
921 		avail[imp] = l->backlog[imp].limit - l->backlog[imp].len;
922 
923 	skb_queue_walk_safe(wakeupq, skb, tmp) {
924 		imp = TIPC_SKB_CB(skb)->chain_imp;
925 		if (avail[imp] <= 0)
926 			continue;
927 		avail[imp]--;
928 		__skb_unlink(skb, wakeupq);
929 		__skb_queue_tail(&tmpq, skb);
930 	}
931 
932 	spin_lock_bh(&inputq->lock);
933 	skb_queue_splice_tail(&tmpq, inputq);
934 	spin_unlock_bh(&inputq->lock);
935 
936 }
937 
938 /**
939  * tipc_link_set_skb_retransmit_time - set the time at which retransmission of
940  *                                     the given skb should be next attempted
941  * @skb: skb to set a future retransmission time for
942  * @l: link the skb will be transmitted on
943  */
tipc_link_set_skb_retransmit_time(struct sk_buff * skb,struct tipc_link * l)944 static void tipc_link_set_skb_retransmit_time(struct sk_buff *skb,
945 					      struct tipc_link *l)
946 {
947 	if (link_is_bc_sndlink(l))
948 		TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
949 	else
950 		TIPC_SKB_CB(skb)->nxt_retr = TIPC_UC_RETR_TIME;
951 }
952 
tipc_link_reset(struct tipc_link * l)953 void tipc_link_reset(struct tipc_link *l)
954 {
955 	struct sk_buff_head list;
956 	u32 imp;
957 
958 	__skb_queue_head_init(&list);
959 
960 	l->in_session = false;
961 	/* Force re-synch of peer session number before establishing */
962 	l->peer_session--;
963 	l->session++;
964 	l->mtu = l->advertised_mtu;
965 
966 	spin_lock_bh(&l->wakeupq.lock);
967 	skb_queue_splice_init(&l->wakeupq, &list);
968 	spin_unlock_bh(&l->wakeupq.lock);
969 
970 	spin_lock_bh(&l->inputq->lock);
971 	skb_queue_splice_init(&list, l->inputq);
972 	spin_unlock_bh(&l->inputq->lock);
973 
974 	__skb_queue_purge(&l->transmq);
975 	__skb_queue_purge(&l->deferdq);
976 	__skb_queue_purge(&l->backlogq);
977 	__skb_queue_purge(&l->failover_deferdq);
978 	for (imp = 0; imp <= TIPC_SYSTEM_IMPORTANCE; imp++) {
979 		l->backlog[imp].len = 0;
980 		l->backlog[imp].target_bskb = NULL;
981 	}
982 	kfree_skb(l->reasm_buf);
983 	kfree_skb(l->reasm_tnlmsg);
984 	kfree_skb(l->failover_reasm_skb);
985 	l->reasm_buf = NULL;
986 	l->reasm_tnlmsg = NULL;
987 	l->failover_reasm_skb = NULL;
988 	l->rcv_unacked = 0;
989 	l->snd_nxt = 1;
990 	l->rcv_nxt = 1;
991 	l->snd_nxt_state = 1;
992 	l->rcv_nxt_state = 1;
993 	l->acked = 0;
994 	l->last_gap = 0;
995 	kfree(l->last_ga);
996 	l->last_ga = NULL;
997 	l->silent_intv_cnt = 0;
998 	l->rst_cnt = 0;
999 	l->bc_peer_is_up = false;
1000 	memset(&l->mon_state, 0, sizeof(l->mon_state));
1001 	tipc_link_reset_stats(l);
1002 }
1003 
1004 /**
1005  * tipc_link_xmit(): enqueue buffer list according to queue situation
1006  * @l: link to use
1007  * @list: chain of buffers containing message
1008  * @xmitq: returned list of packets to be sent by caller
1009  *
1010  * Consumes the buffer chain.
1011  * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
1012  * Return: 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
1013  */
tipc_link_xmit(struct tipc_link * l,struct sk_buff_head * list,struct sk_buff_head * xmitq)1014 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
1015 		   struct sk_buff_head *xmitq)
1016 {
1017 	struct sk_buff_head *backlogq = &l->backlogq;
1018 	struct sk_buff_head *transmq = &l->transmq;
1019 	struct sk_buff *skb, *_skb;
1020 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1021 	u16 ack = l->rcv_nxt - 1;
1022 	u16 seqno = l->snd_nxt;
1023 	int pkt_cnt = skb_queue_len(list);
1024 	unsigned int mss = tipc_link_mss(l);
1025 	unsigned int cwin = l->window;
1026 	unsigned int mtu = l->mtu;
1027 	struct tipc_msg *hdr;
1028 	bool new_bundle;
1029 	int rc = 0;
1030 	int imp;
1031 
1032 	if (pkt_cnt <= 0)
1033 		return 0;
1034 
1035 	hdr = buf_msg(skb_peek(list));
1036 	if (unlikely(msg_size(hdr) > mtu)) {
1037 		pr_warn("Too large msg, purging xmit list %d %d %d %d %d!\n",
1038 			skb_queue_len(list), msg_user(hdr),
1039 			msg_type(hdr), msg_size(hdr), mtu);
1040 		__skb_queue_purge(list);
1041 		return -EMSGSIZE;
1042 	}
1043 
1044 	imp = msg_importance(hdr);
1045 	/* Allow oversubscription of one data msg per source at congestion */
1046 	if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
1047 		if (imp == TIPC_SYSTEM_IMPORTANCE) {
1048 			pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
1049 			__skb_queue_purge(list);
1050 			return -ENOBUFS;
1051 		}
1052 		rc = link_schedule_user(l, hdr);
1053 	}
1054 
1055 	if (pkt_cnt > 1) {
1056 		l->stats.sent_fragmented++;
1057 		l->stats.sent_fragments += pkt_cnt;
1058 	}
1059 
1060 	/* Prepare each packet for sending, and add to relevant queue: */
1061 	while ((skb = __skb_dequeue(list))) {
1062 		if (likely(skb_queue_len(transmq) < cwin)) {
1063 			hdr = buf_msg(skb);
1064 			msg_set_seqno(hdr, seqno);
1065 			msg_set_ack(hdr, ack);
1066 			msg_set_bcast_ack(hdr, bc_ack);
1067 			_skb = skb_clone(skb, GFP_ATOMIC);
1068 			if (!_skb) {
1069 				kfree_skb(skb);
1070 				__skb_queue_purge(list);
1071 				return -ENOBUFS;
1072 			}
1073 			__skb_queue_tail(transmq, skb);
1074 			tipc_link_set_skb_retransmit_time(skb, l);
1075 			__skb_queue_tail(xmitq, _skb);
1076 			TIPC_SKB_CB(skb)->ackers = l->ackers;
1077 			l->rcv_unacked = 0;
1078 			l->stats.sent_pkts++;
1079 			seqno++;
1080 			continue;
1081 		}
1082 		if (tipc_msg_try_bundle(l->backlog[imp].target_bskb, &skb,
1083 					mss, l->addr, &new_bundle)) {
1084 			if (skb) {
1085 				/* Keep a ref. to the skb for next try */
1086 				l->backlog[imp].target_bskb = skb;
1087 				l->backlog[imp].len++;
1088 				__skb_queue_tail(backlogq, skb);
1089 			} else {
1090 				if (new_bundle) {
1091 					l->stats.sent_bundles++;
1092 					l->stats.sent_bundled++;
1093 				}
1094 				l->stats.sent_bundled++;
1095 			}
1096 			continue;
1097 		}
1098 		l->backlog[imp].target_bskb = NULL;
1099 		l->backlog[imp].len += (1 + skb_queue_len(list));
1100 		__skb_queue_tail(backlogq, skb);
1101 		skb_queue_splice_tail_init(list, backlogq);
1102 	}
1103 	l->snd_nxt = seqno;
1104 	return rc;
1105 }
1106 
tipc_link_update_cwin(struct tipc_link * l,int released,bool retransmitted)1107 static void tipc_link_update_cwin(struct tipc_link *l, int released,
1108 				  bool retransmitted)
1109 {
1110 	int bklog_len = skb_queue_len(&l->backlogq);
1111 	struct sk_buff_head *txq = &l->transmq;
1112 	int txq_len = skb_queue_len(txq);
1113 	u16 cwin = l->window;
1114 
1115 	/* Enter fast recovery */
1116 	if (unlikely(retransmitted)) {
1117 		l->ssthresh = max_t(u16, l->window / 2, 300);
1118 		l->window = min_t(u16, l->ssthresh, l->window);
1119 		return;
1120 	}
1121 	/* Enter slow start */
1122 	if (unlikely(!released)) {
1123 		l->ssthresh = max_t(u16, l->window / 2, 300);
1124 		l->window = l->min_win;
1125 		return;
1126 	}
1127 	/* Don't increase window if no pressure on the transmit queue */
1128 	if (txq_len + bklog_len < cwin)
1129 		return;
1130 
1131 	/* Don't increase window if there are holes the transmit queue */
1132 	if (txq_len && l->snd_nxt - buf_seqno(skb_peek(txq)) != txq_len)
1133 		return;
1134 
1135 	l->cong_acks += released;
1136 
1137 	/* Slow start  */
1138 	if (cwin <= l->ssthresh) {
1139 		l->window = min_t(u16, cwin + released, l->max_win);
1140 		return;
1141 	}
1142 	/* Congestion avoidance */
1143 	if (l->cong_acks < cwin)
1144 		return;
1145 	l->window = min_t(u16, ++cwin, l->max_win);
1146 	l->cong_acks = 0;
1147 }
1148 
tipc_link_advance_backlog(struct tipc_link * l,struct sk_buff_head * xmitq)1149 static void tipc_link_advance_backlog(struct tipc_link *l,
1150 				      struct sk_buff_head *xmitq)
1151 {
1152 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1153 	struct sk_buff_head *txq = &l->transmq;
1154 	struct sk_buff *skb, *_skb;
1155 	u16 ack = l->rcv_nxt - 1;
1156 	u16 seqno = l->snd_nxt;
1157 	struct tipc_msg *hdr;
1158 	u16 cwin = l->window;
1159 	u32 imp;
1160 
1161 	while (skb_queue_len(txq) < cwin) {
1162 		skb = skb_peek(&l->backlogq);
1163 		if (!skb)
1164 			break;
1165 		_skb = skb_clone(skb, GFP_ATOMIC);
1166 		if (!_skb)
1167 			break;
1168 		__skb_dequeue(&l->backlogq);
1169 		hdr = buf_msg(skb);
1170 		imp = msg_importance(hdr);
1171 		l->backlog[imp].len--;
1172 		if (unlikely(skb == l->backlog[imp].target_bskb))
1173 			l->backlog[imp].target_bskb = NULL;
1174 		__skb_queue_tail(&l->transmq, skb);
1175 		tipc_link_set_skb_retransmit_time(skb, l);
1176 
1177 		__skb_queue_tail(xmitq, _skb);
1178 		TIPC_SKB_CB(skb)->ackers = l->ackers;
1179 		msg_set_seqno(hdr, seqno);
1180 		msg_set_ack(hdr, ack);
1181 		msg_set_bcast_ack(hdr, bc_ack);
1182 		l->rcv_unacked = 0;
1183 		l->stats.sent_pkts++;
1184 		seqno++;
1185 	}
1186 	l->snd_nxt = seqno;
1187 }
1188 
1189 /**
1190  * link_retransmit_failure() - Detect repeated retransmit failures
1191  * @l: tipc link sender
1192  * @r: tipc link receiver (= l in case of unicast)
1193  * @rc: returned code
1194  *
1195  * Return: true if the repeated retransmit failures happens, otherwise
1196  * false
1197  */
link_retransmit_failure(struct tipc_link * l,struct tipc_link * r,int * rc)1198 static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r,
1199 				    int *rc)
1200 {
1201 	struct sk_buff *skb = skb_peek(&l->transmq);
1202 	struct tipc_msg *hdr;
1203 
1204 	if (!skb)
1205 		return false;
1206 
1207 	if (!TIPC_SKB_CB(skb)->retr_cnt)
1208 		return false;
1209 
1210 	if (!time_after(jiffies, TIPC_SKB_CB(skb)->retr_stamp +
1211 			msecs_to_jiffies(r->tolerance * 10)))
1212 		return false;
1213 
1214 	hdr = buf_msg(skb);
1215 	if (link_is_bc_sndlink(l) && !less(r->acked, msg_seqno(hdr)))
1216 		return false;
1217 
1218 	pr_warn("Retransmission failure on link <%s>\n", l->name);
1219 	link_print(l, "State of link ");
1220 	pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
1221 		msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
1222 	pr_info("sqno %u, prev: %x, dest: %x\n",
1223 		msg_seqno(hdr), msg_prevnode(hdr), msg_destnode(hdr));
1224 	pr_info("retr_stamp %d, retr_cnt %d\n",
1225 		jiffies_to_msecs(TIPC_SKB_CB(skb)->retr_stamp),
1226 		TIPC_SKB_CB(skb)->retr_cnt);
1227 
1228 	trace_tipc_list_dump(&l->transmq, true, "retrans failure!");
1229 	trace_tipc_link_dump(l, TIPC_DUMP_NONE, "retrans failure!");
1230 	trace_tipc_link_dump(r, TIPC_DUMP_NONE, "retrans failure!");
1231 
1232 	if (link_is_bc_sndlink(l)) {
1233 		r->state = LINK_RESET;
1234 		*rc |= TIPC_LINK_DOWN_EVT;
1235 	} else {
1236 		*rc |= tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1237 	}
1238 
1239 	return true;
1240 }
1241 
1242 /* tipc_data_input - deliver data and name distr msgs to upper layer
1243  *
1244  * Consumes buffer if message is of right type
1245  * Node lock must be held
1246  */
tipc_data_input(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * inputq)1247 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1248 			    struct sk_buff_head *inputq)
1249 {
1250 	struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq;
1251 	struct tipc_msg *hdr = buf_msg(skb);
1252 
1253 	switch (msg_user(hdr)) {
1254 	case TIPC_LOW_IMPORTANCE:
1255 	case TIPC_MEDIUM_IMPORTANCE:
1256 	case TIPC_HIGH_IMPORTANCE:
1257 	case TIPC_CRITICAL_IMPORTANCE:
1258 		if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) {
1259 			skb_queue_tail(mc_inputq, skb);
1260 			return true;
1261 		}
1262 		fallthrough;
1263 	case CONN_MANAGER:
1264 		skb_queue_tail(inputq, skb);
1265 		return true;
1266 	case GROUP_PROTOCOL:
1267 		skb_queue_tail(mc_inputq, skb);
1268 		return true;
1269 	case NAME_DISTRIBUTOR:
1270 		l->bc_rcvlink->state = LINK_ESTABLISHED;
1271 		skb_queue_tail(l->namedq, skb);
1272 		return true;
1273 	case MSG_BUNDLER:
1274 	case TUNNEL_PROTOCOL:
1275 	case MSG_FRAGMENTER:
1276 	case BCAST_PROTOCOL:
1277 		return false;
1278 #ifdef CONFIG_TIPC_CRYPTO
1279 	case MSG_CRYPTO:
1280 		if (sysctl_tipc_key_exchange_enabled &&
1281 		    TIPC_SKB_CB(skb)->decrypted) {
1282 			tipc_crypto_msg_rcv(l->net, skb);
1283 			return true;
1284 		}
1285 		fallthrough;
1286 #endif
1287 	default:
1288 		pr_warn("Dropping received illegal msg type\n");
1289 		kfree_skb(skb);
1290 		return true;
1291 	}
1292 }
1293 
1294 /* tipc_link_input - process packet that has passed link protocol check
1295  *
1296  * Consumes buffer
1297  */
tipc_link_input(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * inputq,struct sk_buff ** reasm_skb)1298 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1299 			   struct sk_buff_head *inputq,
1300 			   struct sk_buff **reasm_skb)
1301 {
1302 	struct tipc_msg *hdr = buf_msg(skb);
1303 	struct sk_buff *iskb;
1304 	struct sk_buff_head tmpq;
1305 	int usr = msg_user(hdr);
1306 	int pos = 0;
1307 
1308 	if (usr == MSG_BUNDLER) {
1309 		skb_queue_head_init(&tmpq);
1310 		l->stats.recv_bundles++;
1311 		l->stats.recv_bundled += msg_msgcnt(hdr);
1312 		while (tipc_msg_extract(skb, &iskb, &pos))
1313 			tipc_data_input(l, iskb, &tmpq);
1314 		tipc_skb_queue_splice_tail(&tmpq, inputq);
1315 		return 0;
1316 	} else if (usr == MSG_FRAGMENTER) {
1317 		l->stats.recv_fragments++;
1318 		if (tipc_buf_append(reasm_skb, &skb)) {
1319 			l->stats.recv_fragmented++;
1320 			tipc_data_input(l, skb, inputq);
1321 		} else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1322 			pr_warn_ratelimited("Unable to build fragment list\n");
1323 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1324 		}
1325 		return 0;
1326 	} else if (usr == BCAST_PROTOCOL) {
1327 		tipc_bcast_lock(l->net);
1328 		tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1329 		tipc_bcast_unlock(l->net);
1330 	}
1331 
1332 	kfree_skb(skb);
1333 	return 0;
1334 }
1335 
1336 /* tipc_link_tnl_rcv() - receive TUNNEL_PROTOCOL message, drop or process the
1337  *			 inner message along with the ones in the old link's
1338  *			 deferdq
1339  * @l: tunnel link
1340  * @skb: TUNNEL_PROTOCOL message
1341  * @inputq: queue to put messages ready for delivery
1342  */
tipc_link_tnl_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * inputq)1343 static int tipc_link_tnl_rcv(struct tipc_link *l, struct sk_buff *skb,
1344 			     struct sk_buff_head *inputq)
1345 {
1346 	struct sk_buff **reasm_skb = &l->failover_reasm_skb;
1347 	struct sk_buff **reasm_tnlmsg = &l->reasm_tnlmsg;
1348 	struct sk_buff_head *fdefq = &l->failover_deferdq;
1349 	struct tipc_msg *hdr = buf_msg(skb);
1350 	struct sk_buff *iskb;
1351 	int ipos = 0;
1352 	int rc = 0;
1353 	u16 seqno;
1354 
1355 	if (msg_type(hdr) == SYNCH_MSG) {
1356 		kfree_skb(skb);
1357 		return 0;
1358 	}
1359 
1360 	/* Not a fragment? */
1361 	if (likely(!msg_nof_fragms(hdr))) {
1362 		if (unlikely(!tipc_msg_extract(skb, &iskb, &ipos))) {
1363 			pr_warn_ratelimited("Unable to extract msg, defq: %d\n",
1364 					    skb_queue_len(fdefq));
1365 			return 0;
1366 		}
1367 		kfree_skb(skb);
1368 	} else {
1369 		/* Set fragment type for buf_append */
1370 		if (msg_fragm_no(hdr) == 1)
1371 			msg_set_type(hdr, FIRST_FRAGMENT);
1372 		else if (msg_fragm_no(hdr) < msg_nof_fragms(hdr))
1373 			msg_set_type(hdr, FRAGMENT);
1374 		else
1375 			msg_set_type(hdr, LAST_FRAGMENT);
1376 
1377 		if (!tipc_buf_append(reasm_tnlmsg, &skb)) {
1378 			/* Successful but non-complete reassembly? */
1379 			if (*reasm_tnlmsg || link_is_bc_rcvlink(l))
1380 				return 0;
1381 			pr_warn_ratelimited("Unable to reassemble tunnel msg\n");
1382 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1383 		}
1384 		iskb = skb;
1385 	}
1386 
1387 	do {
1388 		seqno = buf_seqno(iskb);
1389 		if (unlikely(less(seqno, l->drop_point))) {
1390 			kfree_skb(iskb);
1391 			continue;
1392 		}
1393 		if (unlikely(seqno != l->drop_point)) {
1394 			__tipc_skb_queue_sorted(fdefq, seqno, iskb);
1395 			continue;
1396 		}
1397 
1398 		l->drop_point++;
1399 		if (!tipc_data_input(l, iskb, inputq))
1400 			rc |= tipc_link_input(l, iskb, inputq, reasm_skb);
1401 		if (unlikely(rc))
1402 			break;
1403 	} while ((iskb = __tipc_skb_dequeue(fdefq, l->drop_point)));
1404 
1405 	return rc;
1406 }
1407 
1408 /**
1409  * tipc_get_gap_ack_blks - get Gap ACK blocks from PROTOCOL/STATE_MSG
1410  * @ga: returned pointer to the Gap ACK blocks if any
1411  * @l: the tipc link
1412  * @hdr: the PROTOCOL/STATE_MSG header
1413  * @uc: desired Gap ACK blocks type, i.e. unicast (= 1) or broadcast (= 0)
1414  *
1415  * Return: the total Gap ACK blocks size
1416  */
tipc_get_gap_ack_blks(struct tipc_gap_ack_blks ** ga,struct tipc_link * l,struct tipc_msg * hdr,bool uc)1417 u16 tipc_get_gap_ack_blks(struct tipc_gap_ack_blks **ga, struct tipc_link *l,
1418 			  struct tipc_msg *hdr, bool uc)
1419 {
1420 	struct tipc_gap_ack_blks *p;
1421 	u16 sz = 0;
1422 
1423 	/* Does peer support the Gap ACK blocks feature? */
1424 	if (l->peer_caps & TIPC_GAP_ACK_BLOCK) {
1425 		p = (struct tipc_gap_ack_blks *)msg_data(hdr);
1426 		sz = ntohs(p->len);
1427 		/* Sanity check */
1428 		if (sz == struct_size(p, gacks, size_add(p->ugack_cnt, p->bgack_cnt))) {
1429 			/* Good, check if the desired type exists */
1430 			if ((uc && p->ugack_cnt) || (!uc && p->bgack_cnt))
1431 				goto ok;
1432 		/* Backward compatible: peer might not support bc, but uc? */
1433 		} else if (uc && sz == struct_size(p, gacks, p->ugack_cnt)) {
1434 			if (p->ugack_cnt) {
1435 				p->bgack_cnt = 0;
1436 				goto ok;
1437 			}
1438 		}
1439 	}
1440 	/* Other cases: ignore! */
1441 	p = NULL;
1442 
1443 ok:
1444 	*ga = p;
1445 	return sz;
1446 }
1447 
__tipc_build_gap_ack_blks(struct tipc_gap_ack_blks * ga,struct tipc_link * l,u8 start_index)1448 static u8 __tipc_build_gap_ack_blks(struct tipc_gap_ack_blks *ga,
1449 				    struct tipc_link *l, u8 start_index)
1450 {
1451 	struct tipc_gap_ack *gacks = &ga->gacks[start_index];
1452 	struct sk_buff *skb = skb_peek(&l->deferdq);
1453 	u16 expect, seqno = 0;
1454 	u8 n = 0;
1455 
1456 	if (!skb)
1457 		return 0;
1458 
1459 	expect = buf_seqno(skb);
1460 	skb_queue_walk(&l->deferdq, skb) {
1461 		seqno = buf_seqno(skb);
1462 		if (unlikely(more(seqno, expect))) {
1463 			gacks[n].ack = htons(expect - 1);
1464 			gacks[n].gap = htons(seqno - expect);
1465 			if (++n >= MAX_GAP_ACK_BLKS / 2) {
1466 				pr_info_ratelimited("Gacks on %s: %d, ql: %d!\n",
1467 						    l->name, n,
1468 						    skb_queue_len(&l->deferdq));
1469 				return n;
1470 			}
1471 		} else if (unlikely(less(seqno, expect))) {
1472 			pr_warn("Unexpected skb in deferdq!\n");
1473 			continue;
1474 		}
1475 		expect = seqno + 1;
1476 	}
1477 
1478 	/* last block */
1479 	gacks[n].ack = htons(seqno);
1480 	gacks[n].gap = 0;
1481 	n++;
1482 	return n;
1483 }
1484 
1485 /* tipc_build_gap_ack_blks - build Gap ACK blocks
1486  * @l: tipc unicast link
1487  * @hdr: the tipc message buffer to store the Gap ACK blocks after built
1488  *
1489  * The function builds Gap ACK blocks for both the unicast & broadcast receiver
1490  * links of a certain peer, the buffer after built has the network data format
1491  * as found at the struct tipc_gap_ack_blks definition.
1492  *
1493  * returns the actual allocated memory size
1494  */
tipc_build_gap_ack_blks(struct tipc_link * l,struct tipc_msg * hdr)1495 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, struct tipc_msg *hdr)
1496 {
1497 	struct tipc_link *bcl = l->bc_rcvlink;
1498 	struct tipc_gap_ack_blks *ga;
1499 	u16 len;
1500 
1501 	ga = (struct tipc_gap_ack_blks *)msg_data(hdr);
1502 
1503 	/* Start with broadcast link first */
1504 	tipc_bcast_lock(bcl->net);
1505 	msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1506 	msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1507 	ga->bgack_cnt = __tipc_build_gap_ack_blks(ga, bcl, 0);
1508 	tipc_bcast_unlock(bcl->net);
1509 
1510 	/* Now for unicast link, but an explicit NACK only (???) */
1511 	ga->ugack_cnt = (msg_seq_gap(hdr)) ?
1512 			__tipc_build_gap_ack_blks(ga, l, ga->bgack_cnt) : 0;
1513 
1514 	/* Total len */
1515 	len = struct_size(ga, gacks, size_add(ga->bgack_cnt, ga->ugack_cnt));
1516 	ga->len = htons(len);
1517 	return len;
1518 }
1519 
1520 /* tipc_link_advance_transmq - advance TIPC link transmq queue by releasing
1521  *			       acked packets, also doing retransmissions if
1522  *			       gaps found
1523  * @l: tipc link with transmq queue to be advanced
1524  * @r: tipc link "receiver" i.e. in case of broadcast (= "l" if unicast)
1525  * @acked: seqno of last packet acked by peer without any gaps before
1526  * @gap: # of gap packets
1527  * @ga: buffer pointer to Gap ACK blocks from peer
1528  * @xmitq: queue for accumulating the retransmitted packets if any
1529  * @retransmitted: returned boolean value if a retransmission is really issued
1530  * @rc: returned code e.g. TIPC_LINK_DOWN_EVT if a repeated retransmit failures
1531  *      happens (- unlikely case)
1532  *
1533  * Return: the number of packets released from the link transmq
1534  */
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)1535 static int tipc_link_advance_transmq(struct tipc_link *l, struct tipc_link *r,
1536 				     u16 acked, u16 gap,
1537 				     struct tipc_gap_ack_blks *ga,
1538 				     struct sk_buff_head *xmitq,
1539 				     bool *retransmitted, int *rc)
1540 {
1541 	struct tipc_gap_ack_blks *last_ga = r->last_ga, *this_ga = NULL;
1542 	struct tipc_gap_ack *gacks = NULL;
1543 	struct sk_buff *skb, *_skb, *tmp;
1544 	struct tipc_msg *hdr;
1545 	u32 qlen = skb_queue_len(&l->transmq);
1546 	u16 nacked = acked, ngap = gap, gack_cnt = 0;
1547 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1548 	u16 ack = l->rcv_nxt - 1;
1549 	u16 seqno, n = 0;
1550 	u16 end = r->acked, start = end, offset = r->last_gap;
1551 	u16 si = (last_ga) ? last_ga->start_index : 0;
1552 	bool is_uc = !link_is_bc_sndlink(l);
1553 	bool bc_has_acked = false;
1554 
1555 	trace_tipc_link_retrans(r, acked + 1, acked + gap, &l->transmq);
1556 
1557 	/* Determine Gap ACK blocks if any for the particular link */
1558 	if (ga && is_uc) {
1559 		/* Get the Gap ACKs, uc part */
1560 		gack_cnt = ga->ugack_cnt;
1561 		gacks = &ga->gacks[ga->bgack_cnt];
1562 	} else if (ga) {
1563 		/* Copy the Gap ACKs, bc part, for later renewal if needed */
1564 		this_ga = kmemdup(ga, struct_size(ga, gacks, ga->bgack_cnt),
1565 				  GFP_ATOMIC);
1566 		if (likely(this_ga)) {
1567 			this_ga->start_index = 0;
1568 			/* Start with the bc Gap ACKs */
1569 			gack_cnt = this_ga->bgack_cnt;
1570 			gacks = &this_ga->gacks[0];
1571 		} else {
1572 			/* Hmm, we can get in trouble..., simply ignore it */
1573 			pr_warn_ratelimited("Ignoring bc Gap ACKs, no memory\n");
1574 		}
1575 	}
1576 
1577 	/* Advance the link transmq */
1578 	skb_queue_walk_safe(&l->transmq, skb, tmp) {
1579 		seqno = buf_seqno(skb);
1580 
1581 next_gap_ack:
1582 		if (less_eq(seqno, nacked)) {
1583 			if (is_uc)
1584 				goto release;
1585 			/* Skip packets peer has already acked */
1586 			if (!more(seqno, r->acked))
1587 				continue;
1588 			/* Get the next of last Gap ACK blocks */
1589 			while (more(seqno, end)) {
1590 				if (!last_ga || si >= last_ga->bgack_cnt)
1591 					break;
1592 				start = end + offset + 1;
1593 				end = ntohs(last_ga->gacks[si].ack);
1594 				offset = ntohs(last_ga->gacks[si].gap);
1595 				si++;
1596 				WARN_ONCE(more(start, end) ||
1597 					  (!offset &&
1598 					   si < last_ga->bgack_cnt) ||
1599 					  si > MAX_GAP_ACK_BLKS,
1600 					  "Corrupted Gap ACK: %d %d %d %d %d\n",
1601 					  start, end, offset, si,
1602 					  last_ga->bgack_cnt);
1603 			}
1604 			/* Check against the last Gap ACK block */
1605 			if (tipc_in_range(seqno, start, end))
1606 				continue;
1607 			/* Update/release the packet peer is acking */
1608 			bc_has_acked = true;
1609 			if (--TIPC_SKB_CB(skb)->ackers)
1610 				continue;
1611 release:
1612 			/* release skb */
1613 			__skb_unlink(skb, &l->transmq);
1614 			kfree_skb(skb);
1615 		} else if (less_eq(seqno, nacked + ngap)) {
1616 			/* First gap: check if repeated retrans failures? */
1617 			if (unlikely(seqno == acked + 1 &&
1618 				     link_retransmit_failure(l, r, rc))) {
1619 				/* Ignore this bc Gap ACKs if any */
1620 				kfree(this_ga);
1621 				this_ga = NULL;
1622 				break;
1623 			}
1624 			/* retransmit skb if unrestricted*/
1625 			if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1626 				continue;
1627 			tipc_link_set_skb_retransmit_time(skb, l);
1628 			_skb = pskb_copy(skb, GFP_ATOMIC);
1629 			if (!_skb)
1630 				continue;
1631 			hdr = buf_msg(_skb);
1632 			msg_set_ack(hdr, ack);
1633 			msg_set_bcast_ack(hdr, bc_ack);
1634 			_skb->priority = TC_PRIO_CONTROL;
1635 			__skb_queue_tail(xmitq, _skb);
1636 			l->stats.retransmitted++;
1637 			if (!is_uc)
1638 				r->stats.retransmitted++;
1639 			*retransmitted = true;
1640 			/* Increase actual retrans counter & mark first time */
1641 			if (!TIPC_SKB_CB(skb)->retr_cnt++)
1642 				TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1643 		} else {
1644 			/* retry with Gap ACK blocks if any */
1645 			if (n >= gack_cnt)
1646 				break;
1647 			nacked = ntohs(gacks[n].ack);
1648 			ngap = ntohs(gacks[n].gap);
1649 			n++;
1650 			goto next_gap_ack;
1651 		}
1652 	}
1653 
1654 	/* Renew last Gap ACK blocks for bc if needed */
1655 	if (bc_has_acked) {
1656 		if (this_ga) {
1657 			kfree(last_ga);
1658 			r->last_ga = this_ga;
1659 			r->last_gap = gap;
1660 		} else if (last_ga) {
1661 			if (less(acked, start)) {
1662 				si--;
1663 				offset = start - acked - 1;
1664 			} else if (less(acked, end)) {
1665 				acked = end;
1666 			}
1667 			if (si < last_ga->bgack_cnt) {
1668 				last_ga->start_index = si;
1669 				r->last_gap = offset;
1670 			} else {
1671 				kfree(last_ga);
1672 				r->last_ga = NULL;
1673 				r->last_gap = 0;
1674 			}
1675 		} else {
1676 			r->last_gap = 0;
1677 		}
1678 		r->acked = acked;
1679 	} else {
1680 		kfree(this_ga);
1681 	}
1682 
1683 	return qlen - skb_queue_len(&l->transmq);
1684 }
1685 
1686 /* tipc_link_build_state_msg: prepare link state message for transmission
1687  *
1688  * Note that sending of broadcast ack is coordinated among nodes, to reduce
1689  * risk of ack storms towards the sender
1690  */
tipc_link_build_state_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1691 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1692 {
1693 	if (!l)
1694 		return 0;
1695 
1696 	/* Broadcast ACK must be sent via a unicast link => defer to caller */
1697 	if (link_is_bc_rcvlink(l)) {
1698 		if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1699 			return 0;
1700 		l->rcv_unacked = 0;
1701 
1702 		/* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1703 		l->snd_nxt = l->rcv_nxt;
1704 		return TIPC_LINK_SND_STATE;
1705 	}
1706 	/* Unicast ACK */
1707 	l->rcv_unacked = 0;
1708 	l->stats.sent_acks++;
1709 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1710 	return 0;
1711 }
1712 
1713 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1714  */
tipc_link_build_reset_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1715 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1716 {
1717 	int mtyp = RESET_MSG;
1718 	struct sk_buff *skb;
1719 
1720 	if (l->state == LINK_ESTABLISHING)
1721 		mtyp = ACTIVATE_MSG;
1722 
1723 	tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1724 
1725 	/* Inform peer that this endpoint is going down if applicable */
1726 	skb = skb_peek_tail(xmitq);
1727 	if (skb && (l->state == LINK_RESET))
1728 		msg_set_peer_stopping(buf_msg(skb), 1);
1729 }
1730 
1731 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1732  * Note that sending of broadcast NACK is coordinated among nodes, to
1733  * reduce the risk of NACK storms towards the sender
1734  */
tipc_link_build_nack_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1735 static int tipc_link_build_nack_msg(struct tipc_link *l,
1736 				    struct sk_buff_head *xmitq)
1737 {
1738 	u32 def_cnt = ++l->stats.deferred_recv;
1739 	struct sk_buff_head *dfq = &l->deferdq;
1740 	u32 defq_len = skb_queue_len(dfq);
1741 	int match1, match2;
1742 
1743 	if (link_is_bc_rcvlink(l)) {
1744 		match1 = def_cnt & 0xf;
1745 		match2 = tipc_own_addr(l->net) & 0xf;
1746 		if (match1 == match2)
1747 			return TIPC_LINK_SND_STATE;
1748 		return 0;
1749 	}
1750 
1751 	if (defq_len >= 3 && !((defq_len - 3) % 16)) {
1752 		u16 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1753 
1754 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0,
1755 					  rcvgap, 0, 0, xmitq);
1756 	}
1757 	return 0;
1758 }
1759 
1760 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1761  * @l: the link that should handle the message
1762  * @skb: TIPC packet
1763  * @xmitq: queue to place packets to be sent after this call
1764  */
tipc_link_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * xmitq)1765 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1766 		  struct sk_buff_head *xmitq)
1767 {
1768 	struct sk_buff_head *defq = &l->deferdq;
1769 	struct tipc_msg *hdr = buf_msg(skb);
1770 	u16 seqno, rcv_nxt, win_lim;
1771 	int released = 0;
1772 	int rc = 0;
1773 
1774 	/* Verify and update link state */
1775 	if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1776 		return tipc_link_proto_rcv(l, skb, xmitq);
1777 
1778 	/* Don't send probe at next timeout expiration */
1779 	l->silent_intv_cnt = 0;
1780 
1781 	do {
1782 		hdr = buf_msg(skb);
1783 		seqno = msg_seqno(hdr);
1784 		rcv_nxt = l->rcv_nxt;
1785 		win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1786 
1787 		if (unlikely(!tipc_link_is_up(l))) {
1788 			if (l->state == LINK_ESTABLISHING)
1789 				rc = TIPC_LINK_UP_EVT;
1790 			kfree_skb(skb);
1791 			break;
1792 		}
1793 
1794 		/* Drop if outside receive window */
1795 		if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1796 			l->stats.duplicates++;
1797 			kfree_skb(skb);
1798 			break;
1799 		}
1800 		released += tipc_link_advance_transmq(l, l, msg_ack(hdr), 0,
1801 						      NULL, NULL, NULL, NULL);
1802 
1803 		/* Defer delivery if sequence gap */
1804 		if (unlikely(seqno != rcv_nxt)) {
1805 			if (!__tipc_skb_queue_sorted(defq, seqno, skb))
1806 				l->stats.duplicates++;
1807 			rc |= tipc_link_build_nack_msg(l, xmitq);
1808 			break;
1809 		}
1810 
1811 		/* Deliver packet */
1812 		l->rcv_nxt++;
1813 		l->stats.recv_pkts++;
1814 
1815 		if (unlikely(msg_user(hdr) == TUNNEL_PROTOCOL))
1816 			rc |= tipc_link_tnl_rcv(l, skb, l->inputq);
1817 		else if (!tipc_data_input(l, skb, l->inputq))
1818 			rc |= tipc_link_input(l, skb, l->inputq, &l->reasm_buf);
1819 		if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1820 			rc |= tipc_link_build_state_msg(l, xmitq);
1821 		if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1822 			break;
1823 	} while ((skb = __tipc_skb_dequeue(defq, l->rcv_nxt)));
1824 
1825 	/* Forward queues and wake up waiting users */
1826 	if (released) {
1827 		tipc_link_update_cwin(l, released, 0);
1828 		tipc_link_advance_backlog(l, xmitq);
1829 		if (unlikely(!skb_queue_empty(&l->wakeupq)))
1830 			link_prepare_wakeup(l);
1831 	}
1832 	return rc;
1833 }
1834 
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)1835 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1836 				      bool probe_reply, u16 rcvgap,
1837 				      int tolerance, int priority,
1838 				      struct sk_buff_head *xmitq)
1839 {
1840 	struct tipc_mon_state *mstate = &l->mon_state;
1841 	struct sk_buff_head *dfq = &l->deferdq;
1842 	struct tipc_link *bcl = l->bc_rcvlink;
1843 	struct tipc_msg *hdr;
1844 	struct sk_buff *skb;
1845 	bool node_up = tipc_link_is_up(bcl);
1846 	u16 glen = 0, bc_rcvgap = 0;
1847 	int dlen = 0;
1848 	void *data;
1849 
1850 	/* Don't send protocol message during reset or link failover */
1851 	if (tipc_link_is_blocked(l))
1852 		return;
1853 
1854 	if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1855 		return;
1856 
1857 	if ((probe || probe_reply) && !skb_queue_empty(dfq))
1858 		rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1859 
1860 	skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1861 			      tipc_max_domain_size + MAX_GAP_ACK_BLKS_SZ,
1862 			      l->addr, tipc_own_addr(l->net), 0, 0, 0);
1863 	if (!skb)
1864 		return;
1865 
1866 	hdr = buf_msg(skb);
1867 	data = msg_data(hdr);
1868 	msg_set_session(hdr, l->session);
1869 	msg_set_bearer_id(hdr, l->bearer_id);
1870 	msg_set_net_plane(hdr, l->net_plane);
1871 	msg_set_next_sent(hdr, l->snd_nxt);
1872 	msg_set_ack(hdr, l->rcv_nxt - 1);
1873 	msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1874 	msg_set_bc_ack_invalid(hdr, !node_up);
1875 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1876 	msg_set_link_tolerance(hdr, tolerance);
1877 	msg_set_linkprio(hdr, priority);
1878 	msg_set_redundant_link(hdr, node_up);
1879 	msg_set_seq_gap(hdr, 0);
1880 	msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1881 
1882 	if (mtyp == STATE_MSG) {
1883 		if (l->peer_caps & TIPC_LINK_PROTO_SEQNO)
1884 			msg_set_seqno(hdr, l->snd_nxt_state++);
1885 		msg_set_seq_gap(hdr, rcvgap);
1886 		bc_rcvgap = link_bc_rcv_gap(bcl);
1887 		msg_set_bc_gap(hdr, bc_rcvgap);
1888 		msg_set_probe(hdr, probe);
1889 		msg_set_is_keepalive(hdr, probe || probe_reply);
1890 		if (l->peer_caps & TIPC_GAP_ACK_BLOCK)
1891 			glen = tipc_build_gap_ack_blks(l, hdr);
1892 		tipc_mon_prep(l->net, data + glen, &dlen, mstate, l->bearer_id);
1893 		msg_set_size(hdr, INT_H_SIZE + glen + dlen);
1894 		skb_trim(skb, INT_H_SIZE + glen + dlen);
1895 		l->stats.sent_states++;
1896 		l->rcv_unacked = 0;
1897 	} else {
1898 		/* RESET_MSG or ACTIVATE_MSG */
1899 		if (mtyp == ACTIVATE_MSG) {
1900 			msg_set_dest_session_valid(hdr, 1);
1901 			msg_set_dest_session(hdr, l->peer_session);
1902 		}
1903 		msg_set_max_pkt(hdr, l->advertised_mtu);
1904 		strcpy(data, l->if_name);
1905 		msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1906 		skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1907 	}
1908 	if (probe)
1909 		l->stats.sent_probes++;
1910 	if (rcvgap)
1911 		l->stats.sent_nacks++;
1912 	if (bc_rcvgap)
1913 		bcl->stats.sent_nacks++;
1914 	skb->priority = TC_PRIO_CONTROL;
1915 	__skb_queue_tail(xmitq, skb);
1916 	trace_tipc_proto_build(skb, false, l->name);
1917 }
1918 
tipc_link_create_dummy_tnl_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1919 void tipc_link_create_dummy_tnl_msg(struct tipc_link *l,
1920 				    struct sk_buff_head *xmitq)
1921 {
1922 	u32 onode = tipc_own_addr(l->net);
1923 	struct tipc_msg *hdr, *ihdr;
1924 	struct sk_buff_head tnlq;
1925 	struct sk_buff *skb;
1926 	u32 dnode = l->addr;
1927 
1928 	__skb_queue_head_init(&tnlq);
1929 	skb = tipc_msg_create(TUNNEL_PROTOCOL, FAILOVER_MSG,
1930 			      INT_H_SIZE, BASIC_H_SIZE,
1931 			      dnode, onode, 0, 0, 0);
1932 	if (!skb) {
1933 		pr_warn("%sunable to create tunnel packet\n", link_co_err);
1934 		return;
1935 	}
1936 
1937 	hdr = buf_msg(skb);
1938 	msg_set_msgcnt(hdr, 1);
1939 	msg_set_bearer_id(hdr, l->peer_bearer_id);
1940 
1941 	ihdr = (struct tipc_msg *)msg_data(hdr);
1942 	tipc_msg_init(onode, ihdr, TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1943 		      BASIC_H_SIZE, dnode);
1944 	msg_set_errcode(ihdr, TIPC_ERR_NO_PORT);
1945 	__skb_queue_tail(&tnlq, skb);
1946 	tipc_link_xmit(l, &tnlq, xmitq);
1947 }
1948 
1949 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1950  * with contents of the link's transmit and backlog queues.
1951  */
tipc_link_tnl_prepare(struct tipc_link * l,struct tipc_link * tnl,int mtyp,struct sk_buff_head * xmitq)1952 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1953 			   int mtyp, struct sk_buff_head *xmitq)
1954 {
1955 	struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1956 	struct sk_buff *skb, *tnlskb;
1957 	struct tipc_msg *hdr, tnlhdr;
1958 	struct sk_buff_head *queue = &l->transmq;
1959 	struct sk_buff_head tmpxq, tnlq, frags;
1960 	u16 pktlen, pktcnt, seqno = l->snd_nxt;
1961 	bool pktcnt_need_update = false;
1962 	u16 syncpt;
1963 	int rc;
1964 
1965 	if (!tnl)
1966 		return;
1967 
1968 	__skb_queue_head_init(&tnlq);
1969 	/* Link Synching:
1970 	 * From now on, send only one single ("dummy") SYNCH message
1971 	 * to peer. The SYNCH message does not contain any data, just
1972 	 * a header conveying the synch point to the peer.
1973 	 */
1974 	if (mtyp == SYNCH_MSG && (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1975 		tnlskb = tipc_msg_create(TUNNEL_PROTOCOL, SYNCH_MSG,
1976 					 INT_H_SIZE, 0, l->addr,
1977 					 tipc_own_addr(l->net),
1978 					 0, 0, 0);
1979 		if (!tnlskb) {
1980 			pr_warn("%sunable to create dummy SYNCH_MSG\n",
1981 				link_co_err);
1982 			return;
1983 		}
1984 
1985 		hdr = buf_msg(tnlskb);
1986 		syncpt = l->snd_nxt + skb_queue_len(&l->backlogq) - 1;
1987 		msg_set_syncpt(hdr, syncpt);
1988 		msg_set_bearer_id(hdr, l->peer_bearer_id);
1989 		__skb_queue_tail(&tnlq, tnlskb);
1990 		tipc_link_xmit(tnl, &tnlq, xmitq);
1991 		return;
1992 	}
1993 
1994 	__skb_queue_head_init(&tmpxq);
1995 	__skb_queue_head_init(&frags);
1996 	/* At least one packet required for safe algorithm => add dummy */
1997 	skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1998 			      BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1999 			      0, 0, TIPC_ERR_NO_PORT);
2000 	if (!skb) {
2001 		pr_warn("%sunable to create tunnel packet\n", link_co_err);
2002 		return;
2003 	}
2004 	__skb_queue_tail(&tnlq, skb);
2005 	tipc_link_xmit(l, &tnlq, &tmpxq);
2006 	__skb_queue_purge(&tmpxq);
2007 
2008 	/* Initialize reusable tunnel packet header */
2009 	tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
2010 		      mtyp, INT_H_SIZE, l->addr);
2011 	if (mtyp == SYNCH_MSG)
2012 		pktcnt = l->snd_nxt - buf_seqno(skb_peek(&l->transmq));
2013 	else
2014 		pktcnt = skb_queue_len(&l->transmq);
2015 	pktcnt += skb_queue_len(&l->backlogq);
2016 	msg_set_msgcnt(&tnlhdr, pktcnt);
2017 	msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
2018 tnl:
2019 	/* Wrap each packet into a tunnel packet */
2020 	skb_queue_walk(queue, skb) {
2021 		hdr = buf_msg(skb);
2022 		if (queue == &l->backlogq)
2023 			msg_set_seqno(hdr, seqno++);
2024 		pktlen = msg_size(hdr);
2025 
2026 		/* Tunnel link MTU is not large enough? This could be
2027 		 * due to:
2028 		 * 1) Link MTU has just changed or set differently;
2029 		 * 2) Or FAILOVER on the top of a SYNCH message
2030 		 *
2031 		 * The 2nd case should not happen if peer supports
2032 		 * TIPC_TUNNEL_ENHANCED
2033 		 */
2034 		if (pktlen > tnl->mtu - INT_H_SIZE) {
2035 			if (mtyp == FAILOVER_MSG &&
2036 			    (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
2037 				rc = tipc_msg_fragment(skb, &tnlhdr, tnl->mtu,
2038 						       &frags);
2039 				if (rc) {
2040 					pr_warn("%sunable to frag msg: rc %d\n",
2041 						link_co_err, rc);
2042 					return;
2043 				}
2044 				pktcnt += skb_queue_len(&frags) - 1;
2045 				pktcnt_need_update = true;
2046 				skb_queue_splice_tail_init(&frags, &tnlq);
2047 				continue;
2048 			}
2049 			/* Unluckily, peer doesn't have TIPC_TUNNEL_ENHANCED
2050 			 * => Just warn it and return!
2051 			 */
2052 			pr_warn_ratelimited("%stoo large msg <%d, %d>: %d!\n",
2053 					    link_co_err, msg_user(hdr),
2054 					    msg_type(hdr), msg_size(hdr));
2055 			return;
2056 		}
2057 
2058 		msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
2059 		tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
2060 		if (!tnlskb) {
2061 			pr_warn("%sunable to send packet\n", link_co_err);
2062 			return;
2063 		}
2064 		skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
2065 		skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
2066 		__skb_queue_tail(&tnlq, tnlskb);
2067 	}
2068 	if (queue != &l->backlogq) {
2069 		queue = &l->backlogq;
2070 		goto tnl;
2071 	}
2072 
2073 	if (pktcnt_need_update)
2074 		skb_queue_walk(&tnlq, skb) {
2075 			hdr = buf_msg(skb);
2076 			msg_set_msgcnt(hdr, pktcnt);
2077 		}
2078 
2079 	tipc_link_xmit(tnl, &tnlq, xmitq);
2080 
2081 	if (mtyp == FAILOVER_MSG) {
2082 		tnl->drop_point = l->rcv_nxt;
2083 		tnl->failover_reasm_skb = l->reasm_buf;
2084 		l->reasm_buf = NULL;
2085 
2086 		/* Failover the link's deferdq */
2087 		if (unlikely(!skb_queue_empty(fdefq))) {
2088 			pr_warn("Link failover deferdq not empty: %d!\n",
2089 				skb_queue_len(fdefq));
2090 			__skb_queue_purge(fdefq);
2091 		}
2092 		skb_queue_splice_init(&l->deferdq, fdefq);
2093 	}
2094 }
2095 
2096 /**
2097  * tipc_link_failover_prepare() - prepare tnl for link failover
2098  *
2099  * This is a special version of the precursor - tipc_link_tnl_prepare(),
2100  * see the tipc_node_link_failover() for details
2101  *
2102  * @l: failover link
2103  * @tnl: tunnel link
2104  * @xmitq: queue for messages to be xmited
2105  */
tipc_link_failover_prepare(struct tipc_link * l,struct tipc_link * tnl,struct sk_buff_head * xmitq)2106 void tipc_link_failover_prepare(struct tipc_link *l, struct tipc_link *tnl,
2107 				struct sk_buff_head *xmitq)
2108 {
2109 	struct sk_buff_head *fdefq = &tnl->failover_deferdq;
2110 
2111 	tipc_link_create_dummy_tnl_msg(tnl, xmitq);
2112 
2113 	/* This failover link endpoint was never established before,
2114 	 * so it has not received anything from peer.
2115 	 * Otherwise, it must be a normal failover situation or the
2116 	 * node has entered SELF_DOWN_PEER_LEAVING and both peer nodes
2117 	 * would have to start over from scratch instead.
2118 	 */
2119 	tnl->drop_point = 1;
2120 	tnl->failover_reasm_skb = NULL;
2121 
2122 	/* Initiate the link's failover deferdq */
2123 	if (unlikely(!skb_queue_empty(fdefq))) {
2124 		pr_warn("Link failover deferdq not empty: %d!\n",
2125 			skb_queue_len(fdefq));
2126 		__skb_queue_purge(fdefq);
2127 	}
2128 }
2129 
2130 /* tipc_link_validate_msg(): validate message against current link state
2131  * Returns true if message should be accepted, otherwise false
2132  */
tipc_link_validate_msg(struct tipc_link * l,struct tipc_msg * hdr)2133 bool tipc_link_validate_msg(struct tipc_link *l, struct tipc_msg *hdr)
2134 {
2135 	u16 curr_session = l->peer_session;
2136 	u16 session = msg_session(hdr);
2137 	int mtyp = msg_type(hdr);
2138 
2139 	if (msg_user(hdr) != LINK_PROTOCOL)
2140 		return true;
2141 
2142 	switch (mtyp) {
2143 	case RESET_MSG:
2144 		if (!l->in_session)
2145 			return true;
2146 		/* Accept only RESET with new session number */
2147 		return more(session, curr_session);
2148 	case ACTIVATE_MSG:
2149 		if (!l->in_session)
2150 			return true;
2151 		/* Accept only ACTIVATE with new or current session number */
2152 		return !less(session, curr_session);
2153 	case STATE_MSG:
2154 		/* Accept only STATE with current session number */
2155 		if (!l->in_session)
2156 			return false;
2157 		if (session != curr_session)
2158 			return false;
2159 		/* Extra sanity check */
2160 		if (!tipc_link_is_up(l) && msg_ack(hdr))
2161 			return false;
2162 		if (!(l->peer_caps & TIPC_LINK_PROTO_SEQNO))
2163 			return true;
2164 		/* Accept only STATE with new sequence number */
2165 		return !less(msg_seqno(hdr), l->rcv_nxt_state);
2166 	default:
2167 		return false;
2168 	}
2169 }
2170 
2171 /* tipc_link_proto_rcv(): receive link level protocol message :
2172  * Note that network plane id propagates through the network, and may
2173  * change at any time. The node with lowest numerical id determines
2174  * network plane
2175  */
tipc_link_proto_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * xmitq)2176 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
2177 			       struct sk_buff_head *xmitq)
2178 {
2179 	struct tipc_msg *hdr = buf_msg(skb);
2180 	struct tipc_gap_ack_blks *ga = NULL;
2181 	bool reply = msg_probe(hdr), retransmitted = false;
2182 	u32 dlen = msg_data_sz(hdr), glen = 0, msg_max;
2183 	u16 peers_snd_nxt =  msg_next_sent(hdr);
2184 	u16 peers_tol = msg_link_tolerance(hdr);
2185 	u16 peers_prio = msg_linkprio(hdr);
2186 	u16 gap = msg_seq_gap(hdr);
2187 	u16 ack = msg_ack(hdr);
2188 	u16 rcv_nxt = l->rcv_nxt;
2189 	u16 rcvgap = 0;
2190 	int mtyp = msg_type(hdr);
2191 	int rc = 0, released;
2192 	char *if_name;
2193 	void *data;
2194 
2195 	trace_tipc_proto_rcv(skb, false, l->name);
2196 
2197 	if (dlen > U16_MAX)
2198 		goto exit;
2199 
2200 	if (tipc_link_is_blocked(l) || !xmitq)
2201 		goto exit;
2202 
2203 	if (tipc_own_addr(l->net) > msg_prevnode(hdr))
2204 		l->net_plane = msg_net_plane(hdr);
2205 
2206 	if (skb_linearize(skb))
2207 		goto exit;
2208 
2209 	hdr = buf_msg(skb);
2210 	data = msg_data(hdr);
2211 
2212 	if (!tipc_link_validate_msg(l, hdr)) {
2213 		trace_tipc_skb_dump(skb, false, "PROTO invalid (1)!");
2214 		trace_tipc_link_dump(l, TIPC_DUMP_NONE, "PROTO invalid (1)!");
2215 		goto exit;
2216 	}
2217 
2218 	switch (mtyp) {
2219 	case RESET_MSG:
2220 	case ACTIVATE_MSG:
2221 		msg_max = msg_max_pkt(hdr);
2222 		if (msg_max < tipc_bearer_min_mtu(l->net, l->bearer_id))
2223 			break;
2224 		/* Complete own link name with peer's interface name */
2225 		if_name =  strrchr(l->name, ':') + 1;
2226 		if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
2227 			break;
2228 		if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
2229 			break;
2230 		strncpy(if_name, data, TIPC_MAX_IF_NAME);
2231 
2232 		/* Update own tolerance if peer indicates a non-zero value */
2233 		if (tipc_in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2234 			l->tolerance = peers_tol;
2235 			l->bc_rcvlink->tolerance = peers_tol;
2236 		}
2237 		/* Update own priority if peer's priority is higher */
2238 		if (tipc_in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
2239 			l->priority = peers_prio;
2240 
2241 		/* If peer is going down we want full re-establish cycle */
2242 		if (msg_peer_stopping(hdr)) {
2243 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2244 			break;
2245 		}
2246 
2247 		/* If this endpoint was re-created while peer was ESTABLISHING
2248 		 * it doesn't know current session number. Force re-synch.
2249 		 */
2250 		if (mtyp == ACTIVATE_MSG && msg_dest_session_valid(hdr) &&
2251 		    l->session != msg_dest_session(hdr)) {
2252 			if (less(l->session, msg_dest_session(hdr)))
2253 				l->session = msg_dest_session(hdr) + 1;
2254 			break;
2255 		}
2256 
2257 		/* ACTIVATE_MSG serves as PEER_RESET if link is already down */
2258 		if (mtyp == RESET_MSG || !tipc_link_is_up(l))
2259 			rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
2260 
2261 		/* ACTIVATE_MSG takes up link if it was already locally reset */
2262 		if (mtyp == ACTIVATE_MSG && l->state == LINK_ESTABLISHING)
2263 			rc = TIPC_LINK_UP_EVT;
2264 
2265 		l->peer_session = msg_session(hdr);
2266 		l->in_session = true;
2267 		l->peer_bearer_id = msg_bearer_id(hdr);
2268 		if (l->mtu > msg_max)
2269 			l->mtu = msg_max;
2270 		break;
2271 
2272 	case STATE_MSG:
2273 		/* Validate Gap ACK blocks, drop if invalid */
2274 		glen = tipc_get_gap_ack_blks(&ga, l, hdr, true);
2275 		if (glen > dlen)
2276 			break;
2277 
2278 		l->rcv_nxt_state = msg_seqno(hdr) + 1;
2279 
2280 		/* Update own tolerance if peer indicates a non-zero value */
2281 		if (tipc_in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2282 			l->tolerance = peers_tol;
2283 			l->bc_rcvlink->tolerance = peers_tol;
2284 		}
2285 		/* Update own prio if peer indicates a different value */
2286 		if ((peers_prio != l->priority) &&
2287 		    tipc_in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
2288 			l->priority = peers_prio;
2289 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2290 		}
2291 
2292 		l->silent_intv_cnt = 0;
2293 		l->stats.recv_states++;
2294 		if (msg_probe(hdr))
2295 			l->stats.recv_probes++;
2296 
2297 		if (!tipc_link_is_up(l)) {
2298 			if (l->state == LINK_ESTABLISHING)
2299 				rc = TIPC_LINK_UP_EVT;
2300 			break;
2301 		}
2302 
2303 		tipc_mon_rcv(l->net, data + glen, dlen - glen, l->addr,
2304 			     &l->mon_state, l->bearer_id);
2305 
2306 		/* Send NACK if peer has sent pkts we haven't received yet */
2307 		if ((reply || msg_is_keepalive(hdr)) &&
2308 		    more(peers_snd_nxt, rcv_nxt) &&
2309 		    !tipc_link_is_synching(l) &&
2310 		    skb_queue_empty(&l->deferdq))
2311 			rcvgap = peers_snd_nxt - l->rcv_nxt;
2312 		if (rcvgap || reply)
2313 			tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
2314 						  rcvgap, 0, 0, xmitq);
2315 
2316 		released = tipc_link_advance_transmq(l, l, ack, gap, ga, xmitq,
2317 						     &retransmitted, &rc);
2318 		if (gap)
2319 			l->stats.recv_nacks++;
2320 		if (released || retransmitted)
2321 			tipc_link_update_cwin(l, released, retransmitted);
2322 		if (released)
2323 			tipc_link_advance_backlog(l, xmitq);
2324 		if (unlikely(!skb_queue_empty(&l->wakeupq)))
2325 			link_prepare_wakeup(l);
2326 	}
2327 exit:
2328 	kfree_skb(skb);
2329 	return rc;
2330 }
2331 
2332 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
2333  */
tipc_link_build_bc_proto_msg(struct tipc_link * l,bool bcast,u16 peers_snd_nxt,struct sk_buff_head * xmitq)2334 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
2335 					 u16 peers_snd_nxt,
2336 					 struct sk_buff_head *xmitq)
2337 {
2338 	struct sk_buff *skb;
2339 	struct tipc_msg *hdr;
2340 	struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
2341 	u16 ack = l->rcv_nxt - 1;
2342 	u16 gap_to = peers_snd_nxt - 1;
2343 
2344 	skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
2345 			      0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
2346 	if (!skb)
2347 		return false;
2348 	hdr = buf_msg(skb);
2349 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
2350 	msg_set_bcast_ack(hdr, ack);
2351 	msg_set_bcgap_after(hdr, ack);
2352 	if (dfrd_skb)
2353 		gap_to = buf_seqno(dfrd_skb) - 1;
2354 	msg_set_bcgap_to(hdr, gap_to);
2355 	msg_set_non_seq(hdr, bcast);
2356 	__skb_queue_tail(xmitq, skb);
2357 	return true;
2358 }
2359 
2360 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
2361  *
2362  * Give a newly added peer node the sequence number where it should
2363  * start receiving and acking broadcast packets.
2364  */
tipc_link_build_bc_init_msg(struct tipc_link * l,struct sk_buff_head * xmitq)2365 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
2366 					struct sk_buff_head *xmitq)
2367 {
2368 	struct sk_buff_head list;
2369 
2370 	__skb_queue_head_init(&list);
2371 	if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
2372 		return;
2373 	msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
2374 	tipc_link_xmit(l, &list, xmitq);
2375 }
2376 
2377 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
2378  */
tipc_link_bc_init_rcv(struct tipc_link * l,struct tipc_msg * hdr)2379 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
2380 {
2381 	int mtyp = msg_type(hdr);
2382 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2383 
2384 	if (tipc_link_is_up(l))
2385 		return;
2386 
2387 	if (msg_user(hdr) == BCAST_PROTOCOL) {
2388 		l->rcv_nxt = peers_snd_nxt;
2389 		l->state = LINK_ESTABLISHED;
2390 		return;
2391 	}
2392 
2393 	if (l->peer_caps & TIPC_BCAST_SYNCH)
2394 		return;
2395 
2396 	if (msg_peer_node_is_up(hdr))
2397 		return;
2398 
2399 	/* Compatibility: accept older, less safe initial synch data */
2400 	if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
2401 		l->rcv_nxt = peers_snd_nxt;
2402 }
2403 
2404 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
2405  */
tipc_link_bc_sync_rcv(struct tipc_link * l,struct tipc_msg * hdr,struct sk_buff_head * xmitq)2406 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
2407 			  struct sk_buff_head *xmitq)
2408 {
2409 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2410 	int rc = 0;
2411 
2412 	if (!tipc_link_is_up(l))
2413 		return rc;
2414 
2415 	if (!msg_peer_node_is_up(hdr))
2416 		return rc;
2417 
2418 	/* Open when peer acknowledges our bcast init msg (pkt #1) */
2419 	if (msg_ack(hdr))
2420 		l->bc_peer_is_up = true;
2421 
2422 	if (!l->bc_peer_is_up)
2423 		return rc;
2424 
2425 	/* Ignore if peers_snd_nxt goes beyond receive window */
2426 	if (more(peers_snd_nxt, l->rcv_nxt + l->window))
2427 		return rc;
2428 
2429 	l->snd_nxt = peers_snd_nxt;
2430 	if (link_bc_rcv_gap(l))
2431 		rc |= TIPC_LINK_SND_STATE;
2432 
2433 	/* Return now if sender supports nack via STATE messages */
2434 	if (l->peer_caps & TIPC_BCAST_STATE_NACK)
2435 		return rc;
2436 
2437 	/* Otherwise, be backwards compatible */
2438 
2439 	if (!more(peers_snd_nxt, l->rcv_nxt)) {
2440 		l->nack_state = BC_NACK_SND_CONDITIONAL;
2441 		return 0;
2442 	}
2443 
2444 	/* Don't NACK if one was recently sent or peeked */
2445 	if (l->nack_state == BC_NACK_SND_SUPPRESS) {
2446 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2447 		return 0;
2448 	}
2449 
2450 	/* Conditionally delay NACK sending until next synch rcv */
2451 	if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
2452 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2453 		if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
2454 			return 0;
2455 	}
2456 
2457 	/* Send NACK now but suppress next one */
2458 	tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
2459 	l->nack_state = BC_NACK_SND_SUPPRESS;
2460 	return 0;
2461 }
2462 
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)2463 int tipc_link_bc_ack_rcv(struct tipc_link *r, u16 acked, u16 gap,
2464 			 struct tipc_gap_ack_blks *ga,
2465 			 struct sk_buff_head *xmitq,
2466 			 struct sk_buff_head *retrq)
2467 {
2468 	struct tipc_link *l = r->bc_sndlink;
2469 	bool unused = false;
2470 	int rc = 0;
2471 
2472 	if (!tipc_link_is_up(r) || !r->bc_peer_is_up)
2473 		return 0;
2474 
2475 	if (gap) {
2476 		l->stats.recv_nacks++;
2477 		r->stats.recv_nacks++;
2478 	}
2479 
2480 	if (less(acked, r->acked) || (acked == r->acked && !gap && !ga))
2481 		return 0;
2482 
2483 	trace_tipc_link_bc_ack(r, acked, gap, &l->transmq);
2484 	tipc_link_advance_transmq(l, r, acked, gap, ga, retrq, &unused, &rc);
2485 
2486 	tipc_link_advance_backlog(l, xmitq);
2487 	if (unlikely(!skb_queue_empty(&l->wakeupq)))
2488 		link_prepare_wakeup(l);
2489 
2490 	return rc;
2491 }
2492 
2493 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
2494  * This function is here for backwards compatibility, since
2495  * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
2496  */
tipc_link_bc_nack_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * xmitq)2497 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
2498 			  struct sk_buff_head *xmitq)
2499 {
2500 	struct tipc_msg *hdr = buf_msg(skb);
2501 	u32 dnode = msg_destnode(hdr);
2502 	int mtyp = msg_type(hdr);
2503 	u16 acked = msg_bcast_ack(hdr);
2504 	u16 from = acked + 1;
2505 	u16 to = msg_bcgap_to(hdr);
2506 	u16 peers_snd_nxt = to + 1;
2507 	int rc = 0;
2508 
2509 	kfree_skb(skb);
2510 
2511 	if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
2512 		return 0;
2513 
2514 	if (mtyp != STATE_MSG)
2515 		return 0;
2516 
2517 	if (dnode == tipc_own_addr(l->net)) {
2518 		rc = tipc_link_bc_ack_rcv(l, acked, to - acked, NULL, xmitq,
2519 					  xmitq);
2520 		l->stats.recv_nacks++;
2521 		return rc;
2522 	}
2523 
2524 	/* Msg for other node => suppress own NACK at next sync if applicable */
2525 	if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
2526 		l->nack_state = BC_NACK_SND_SUPPRESS;
2527 
2528 	return 0;
2529 }
2530 
tipc_link_set_queue_limits(struct tipc_link * l,u32 min_win,u32 max_win)2531 void tipc_link_set_queue_limits(struct tipc_link *l, u32 min_win, u32 max_win)
2532 {
2533 	int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
2534 
2535 	l->min_win = min_win;
2536 	l->ssthresh = max_win;
2537 	l->max_win = max_win;
2538 	l->window = min_win;
2539 	l->backlog[TIPC_LOW_IMPORTANCE].limit      = min_win * 2;
2540 	l->backlog[TIPC_MEDIUM_IMPORTANCE].limit   = min_win * 4;
2541 	l->backlog[TIPC_HIGH_IMPORTANCE].limit     = min_win * 6;
2542 	l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = min_win * 8;
2543 	l->backlog[TIPC_SYSTEM_IMPORTANCE].limit   = max_bulk;
2544 }
2545 
2546 /**
2547  * tipc_link_reset_stats - reset link statistics
2548  * @l: pointer to link
2549  */
tipc_link_reset_stats(struct tipc_link * l)2550 void tipc_link_reset_stats(struct tipc_link *l)
2551 {
2552 	memset(&l->stats, 0, sizeof(l->stats));
2553 }
2554 
link_print(struct tipc_link * l,const char * str)2555 static void link_print(struct tipc_link *l, const char *str)
2556 {
2557 	struct sk_buff *hskb = skb_peek(&l->transmq);
2558 	u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
2559 	u16 tail = l->snd_nxt - 1;
2560 
2561 	pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
2562 	pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
2563 		skb_queue_len(&l->transmq), head, tail,
2564 		skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
2565 }
2566 
2567 /* Parse and validate nested (link) properties valid for media, bearer and link
2568  */
tipc_nl_parse_link_prop(struct nlattr * prop,struct nlattr * props[])2569 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
2570 {
2571 	int err;
2572 
2573 	err = nla_parse_nested_deprecated(props, TIPC_NLA_PROP_MAX, prop,
2574 					  tipc_nl_prop_policy, NULL);
2575 	if (err)
2576 		return err;
2577 
2578 	if (props[TIPC_NLA_PROP_PRIO]) {
2579 		u32 prio;
2580 
2581 		prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
2582 		if (prio > TIPC_MAX_LINK_PRI)
2583 			return -EINVAL;
2584 	}
2585 
2586 	if (props[TIPC_NLA_PROP_TOL]) {
2587 		u32 tol;
2588 
2589 		tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
2590 		if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
2591 			return -EINVAL;
2592 	}
2593 
2594 	if (props[TIPC_NLA_PROP_WIN]) {
2595 		u32 max_win;
2596 
2597 		max_win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
2598 		if (max_win < TIPC_DEF_LINK_WIN || max_win > TIPC_MAX_LINK_WIN)
2599 			return -EINVAL;
2600 	}
2601 
2602 	return 0;
2603 }
2604 
__tipc_nl_add_stats(struct sk_buff * skb,struct tipc_stats * s)2605 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
2606 {
2607 	int i;
2608 	struct nlattr *stats;
2609 
2610 	struct nla_map {
2611 		u32 key;
2612 		u32 val;
2613 	};
2614 
2615 	struct nla_map map[] = {
2616 		{TIPC_NLA_STATS_RX_INFO, 0},
2617 		{TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
2618 		{TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
2619 		{TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
2620 		{TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
2621 		{TIPC_NLA_STATS_TX_INFO, 0},
2622 		{TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
2623 		{TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
2624 		{TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
2625 		{TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
2626 		{TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
2627 			s->msg_length_counts : 1},
2628 		{TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
2629 		{TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
2630 		{TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
2631 		{TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
2632 		{TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
2633 		{TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
2634 		{TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
2635 		{TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
2636 		{TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
2637 		{TIPC_NLA_STATS_RX_STATES, s->recv_states},
2638 		{TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
2639 		{TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
2640 		{TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
2641 		{TIPC_NLA_STATS_TX_STATES, s->sent_states},
2642 		{TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
2643 		{TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
2644 		{TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
2645 		{TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
2646 		{TIPC_NLA_STATS_DUPLICATES, s->duplicates},
2647 		{TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
2648 		{TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
2649 		{TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
2650 			(s->accu_queue_sz / s->queue_sz_counts) : 0}
2651 	};
2652 
2653 	stats = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2654 	if (!stats)
2655 		return -EMSGSIZE;
2656 
2657 	for (i = 0; i <  ARRAY_SIZE(map); i++)
2658 		if (nla_put_u32(skb, map[i].key, map[i].val))
2659 			goto msg_full;
2660 
2661 	nla_nest_end(skb, stats);
2662 
2663 	return 0;
2664 msg_full:
2665 	nla_nest_cancel(skb, stats);
2666 
2667 	return -EMSGSIZE;
2668 }
2669 
2670 /* 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)2671 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
2672 		       struct tipc_link *link, int nlflags)
2673 {
2674 	u32 self = tipc_own_addr(net);
2675 	struct nlattr *attrs;
2676 	struct nlattr *prop;
2677 	void *hdr;
2678 	int err;
2679 
2680 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2681 			  nlflags, TIPC_NL_LINK_GET);
2682 	if (!hdr)
2683 		return -EMSGSIZE;
2684 
2685 	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2686 	if (!attrs)
2687 		goto msg_full;
2688 
2689 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
2690 		goto attr_msg_full;
2691 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
2692 		goto attr_msg_full;
2693 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2694 		goto attr_msg_full;
2695 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
2696 		goto attr_msg_full;
2697 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
2698 		goto attr_msg_full;
2699 
2700 	if (tipc_link_is_up(link))
2701 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2702 			goto attr_msg_full;
2703 	if (link->active)
2704 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2705 			goto attr_msg_full;
2706 
2707 	prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2708 	if (!prop)
2709 		goto attr_msg_full;
2710 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2711 		goto prop_msg_full;
2712 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2713 		goto prop_msg_full;
2714 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2715 			link->window))
2716 		goto prop_msg_full;
2717 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2718 		goto prop_msg_full;
2719 	nla_nest_end(msg->skb, prop);
2720 
2721 	err = __tipc_nl_add_stats(msg->skb, &link->stats);
2722 	if (err)
2723 		goto attr_msg_full;
2724 
2725 	nla_nest_end(msg->skb, attrs);
2726 	genlmsg_end(msg->skb, hdr);
2727 
2728 	return 0;
2729 
2730 prop_msg_full:
2731 	nla_nest_cancel(msg->skb, prop);
2732 attr_msg_full:
2733 	nla_nest_cancel(msg->skb, attrs);
2734 msg_full:
2735 	genlmsg_cancel(msg->skb, hdr);
2736 
2737 	return -EMSGSIZE;
2738 }
2739 
__tipc_nl_add_bc_link_stat(struct sk_buff * skb,struct tipc_stats * stats)2740 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2741 				      struct tipc_stats *stats)
2742 {
2743 	int i;
2744 	struct nlattr *nest;
2745 
2746 	struct nla_map {
2747 		__u32 key;
2748 		__u32 val;
2749 	};
2750 
2751 	struct nla_map map[] = {
2752 		{TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2753 		{TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2754 		{TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2755 		{TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2756 		{TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2757 		{TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2758 		{TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2759 		{TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2760 		{TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2761 		{TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2762 		{TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2763 		{TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2764 		{TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2765 		{TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2766 		{TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2767 		{TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2768 		{TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2769 		{TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2770 		{TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2771 			(stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2772 	};
2773 
2774 	nest = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2775 	if (!nest)
2776 		return -EMSGSIZE;
2777 
2778 	for (i = 0; i <  ARRAY_SIZE(map); i++)
2779 		if (nla_put_u32(skb, map[i].key, map[i].val))
2780 			goto msg_full;
2781 
2782 	nla_nest_end(skb, nest);
2783 
2784 	return 0;
2785 msg_full:
2786 	nla_nest_cancel(skb, nest);
2787 
2788 	return -EMSGSIZE;
2789 }
2790 
tipc_nl_add_bc_link(struct net * net,struct tipc_nl_msg * msg,struct tipc_link * bcl)2791 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg,
2792 			struct tipc_link *bcl)
2793 {
2794 	int err;
2795 	void *hdr;
2796 	struct nlattr *attrs;
2797 	struct nlattr *prop;
2798 	u32 bc_mode = tipc_bcast_get_mode(net);
2799 	u32 bc_ratio = tipc_bcast_get_broadcast_ratio(net);
2800 
2801 	if (!bcl)
2802 		return 0;
2803 
2804 	tipc_bcast_lock(net);
2805 
2806 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2807 			  NLM_F_MULTI, TIPC_NL_LINK_GET);
2808 	if (!hdr) {
2809 		tipc_bcast_unlock(net);
2810 		return -EMSGSIZE;
2811 	}
2812 
2813 	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2814 	if (!attrs)
2815 		goto msg_full;
2816 
2817 	/* The broadcast link is always up */
2818 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2819 		goto attr_msg_full;
2820 
2821 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2822 		goto attr_msg_full;
2823 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2824 		goto attr_msg_full;
2825 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2826 		goto attr_msg_full;
2827 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2828 		goto attr_msg_full;
2829 
2830 	prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2831 	if (!prop)
2832 		goto attr_msg_full;
2833 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->max_win))
2834 		goto prop_msg_full;
2835 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST, bc_mode))
2836 		goto prop_msg_full;
2837 	if (bc_mode & BCLINK_MODE_SEL)
2838 		if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST_RATIO,
2839 				bc_ratio))
2840 			goto prop_msg_full;
2841 	nla_nest_end(msg->skb, prop);
2842 
2843 	err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2844 	if (err)
2845 		goto attr_msg_full;
2846 
2847 	tipc_bcast_unlock(net);
2848 	nla_nest_end(msg->skb, attrs);
2849 	genlmsg_end(msg->skb, hdr);
2850 
2851 	return 0;
2852 
2853 prop_msg_full:
2854 	nla_nest_cancel(msg->skb, prop);
2855 attr_msg_full:
2856 	nla_nest_cancel(msg->skb, attrs);
2857 msg_full:
2858 	tipc_bcast_unlock(net);
2859 	genlmsg_cancel(msg->skb, hdr);
2860 
2861 	return -EMSGSIZE;
2862 }
2863 
tipc_link_set_tolerance(struct tipc_link * l,u32 tol,struct sk_buff_head * xmitq)2864 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2865 			     struct sk_buff_head *xmitq)
2866 {
2867 	l->tolerance = tol;
2868 	if (l->bc_rcvlink)
2869 		l->bc_rcvlink->tolerance = tol;
2870 	if (tipc_link_is_up(l))
2871 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2872 }
2873 
tipc_link_set_prio(struct tipc_link * l,u32 prio,struct sk_buff_head * xmitq)2874 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2875 			struct sk_buff_head *xmitq)
2876 {
2877 	l->priority = prio;
2878 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2879 }
2880 
tipc_link_set_abort_limit(struct tipc_link * l,u32 limit)2881 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2882 {
2883 	l->abort_limit = limit;
2884 }
2885 
2886 /**
2887  * tipc_link_dump - dump TIPC link data
2888  * @l: tipc link to be dumped
2889  * @dqueues: bitmask to decide if any link queue to be dumped?
2890  *           - TIPC_DUMP_NONE: don't dump link queues
2891  *           - TIPC_DUMP_TRANSMQ: dump link transmq queue
2892  *           - TIPC_DUMP_BACKLOGQ: dump link backlog queue
2893  *           - TIPC_DUMP_DEFERDQ: dump link deferd queue
2894  *           - TIPC_DUMP_INPUTQ: dump link input queue
2895  *           - TIPC_DUMP_WAKEUP: dump link wakeup queue
2896  *           - TIPC_DUMP_ALL: dump all the link queues above
2897  * @buf: returned buffer of dump data in format
2898  */
tipc_link_dump(struct tipc_link * l,u16 dqueues,char * buf)2899 int tipc_link_dump(struct tipc_link *l, u16 dqueues, char *buf)
2900 {
2901 	int i = 0;
2902 	size_t sz = (dqueues) ? LINK_LMAX : LINK_LMIN;
2903 	struct sk_buff_head *list;
2904 	struct sk_buff *hskb, *tskb;
2905 	u32 len;
2906 
2907 	if (!l) {
2908 		i += scnprintf(buf, sz, "link data: (null)\n");
2909 		return i;
2910 	}
2911 
2912 	i += scnprintf(buf, sz, "link data: %x", l->addr);
2913 	i += scnprintf(buf + i, sz - i, " %x", l->state);
2914 	i += scnprintf(buf + i, sz - i, " %u", l->in_session);
2915 	i += scnprintf(buf + i, sz - i, " %u", l->session);
2916 	i += scnprintf(buf + i, sz - i, " %u", l->peer_session);
2917 	i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt);
2918 	i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt);
2919 	i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt_state);
2920 	i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt_state);
2921 	i += scnprintf(buf + i, sz - i, " %x", l->peer_caps);
2922 	i += scnprintf(buf + i, sz - i, " %u", l->silent_intv_cnt);
2923 	i += scnprintf(buf + i, sz - i, " %u", l->rst_cnt);
2924 	i += scnprintf(buf + i, sz - i, " %u", 0);
2925 	i += scnprintf(buf + i, sz - i, " %u", 0);
2926 	i += scnprintf(buf + i, sz - i, " %u", l->acked);
2927 
2928 	list = &l->transmq;
2929 	len = skb_queue_len(list);
2930 	hskb = skb_peek(list);
2931 	tskb = skb_peek_tail(list);
2932 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2933 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2934 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2935 
2936 	list = &l->deferdq;
2937 	len = skb_queue_len(list);
2938 	hskb = skb_peek(list);
2939 	tskb = skb_peek_tail(list);
2940 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2941 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2942 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2943 
2944 	list = &l->backlogq;
2945 	len = skb_queue_len(list);
2946 	hskb = skb_peek(list);
2947 	tskb = skb_peek_tail(list);
2948 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2949 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2950 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2951 
2952 	list = l->inputq;
2953 	len = skb_queue_len(list);
2954 	hskb = skb_peek(list);
2955 	tskb = skb_peek_tail(list);
2956 	i += scnprintf(buf + i, sz - i, " | %u %u %u\n", len,
2957 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2958 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2959 
2960 	if (dqueues & TIPC_DUMP_TRANSMQ) {
2961 		i += scnprintf(buf + i, sz - i, "transmq: ");
2962 		i += tipc_list_dump(&l->transmq, false, buf + i);
2963 	}
2964 	if (dqueues & TIPC_DUMP_BACKLOGQ) {
2965 		i += scnprintf(buf + i, sz - i,
2966 			       "backlogq: <%u %u %u %u %u>, ",
2967 			       l->backlog[TIPC_LOW_IMPORTANCE].len,
2968 			       l->backlog[TIPC_MEDIUM_IMPORTANCE].len,
2969 			       l->backlog[TIPC_HIGH_IMPORTANCE].len,
2970 			       l->backlog[TIPC_CRITICAL_IMPORTANCE].len,
2971 			       l->backlog[TIPC_SYSTEM_IMPORTANCE].len);
2972 		i += tipc_list_dump(&l->backlogq, false, buf + i);
2973 	}
2974 	if (dqueues & TIPC_DUMP_DEFERDQ) {
2975 		i += scnprintf(buf + i, sz - i, "deferdq: ");
2976 		i += tipc_list_dump(&l->deferdq, false, buf + i);
2977 	}
2978 	if (dqueues & TIPC_DUMP_INPUTQ) {
2979 		i += scnprintf(buf + i, sz - i, "inputq: ");
2980 		i += tipc_list_dump(l->inputq, false, buf + i);
2981 	}
2982 	if (dqueues & TIPC_DUMP_WAKEUP) {
2983 		i += scnprintf(buf + i, sz - i, "wakeup: ");
2984 		i += tipc_list_dump(&l->wakeupq, false, buf + i);
2985 	}
2986 
2987 	return i;
2988 }
2989