1 /*
2  * net/tipc/link.c: TIPC link code
3  *
4  * Copyright (c) 1996-2007, 2012-2016, Ericsson AB
5  * Copyright (c) 2004-2007, 2010-2013, Wind River Systems
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions are met:
10  *
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the names of the copyright holders nor the names of its
17  *    contributors may be used to endorse or promote products derived from
18  *    this software without specific prior written permission.
19  *
20  * Alternatively, this software may be distributed under the terms of the
21  * GNU General Public License ("GPL") version 2 as published by the Free
22  * Software Foundation.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34  * POSSIBILITY OF SUCH DAMAGE.
35  */
36 
37 #include "core.h"
38 #include "subscr.h"
39 #include "link.h"
40 #include "bcast.h"
41 #include "socket.h"
42 #include "name_distr.h"
43 #include "discover.h"
44 #include "netlink.h"
45 #include "monitor.h"
46 #include "trace.h"
47 #include "crypto.h"
48 
49 #include <linux/pkt_sched.h>
50 
51 struct tipc_stats {
52 	u32 sent_pkts;
53 	u32 recv_pkts;
54 	u32 sent_states;
55 	u32 recv_states;
56 	u32 sent_probes;
57 	u32 recv_probes;
58 	u32 sent_nacks;
59 	u32 recv_nacks;
60 	u32 sent_acks;
61 	u32 sent_bundled;
62 	u32 sent_bundles;
63 	u32 recv_bundled;
64 	u32 recv_bundles;
65 	u32 retransmitted;
66 	u32 sent_fragmented;
67 	u32 sent_fragments;
68 	u32 recv_fragmented;
69 	u32 recv_fragments;
70 	u32 link_congs;		/* # port sends blocked by congestion */
71 	u32 deferred_recv;
72 	u32 duplicates;
73 	u32 max_queue_sz;	/* send queue size high water mark */
74 	u32 accu_queue_sz;	/* used for send queue size profiling */
75 	u32 queue_sz_counts;	/* used for send queue size profiling */
76 	u32 msg_length_counts;	/* used for message length profiling */
77 	u32 msg_lengths_total;	/* used for message length profiling */
78 	u32 msg_length_profile[7]; /* used for msg. length profiling */
79 };
80 
81 /**
82  * struct tipc_link - TIPC link data structure
83  * @addr: network address of link's peer node
84  * @name: link name character string
85  * @media_addr: media address to use when sending messages over link
86  * @timer: link timer
87  * @net: pointer to namespace struct
88  * @refcnt: reference counter for permanent references (owner node & timer)
89  * @peer_session: link session # being used by peer end of link
90  * @peer_bearer_id: bearer id used by link's peer endpoint
91  * @bearer_id: local bearer id used by link
92  * @tolerance: minimum link continuity loss needed to reset link [in ms]
93  * @abort_limit: # of unacknowledged continuity probes needed to reset link
94  * @state: current state of link FSM
95  * @peer_caps: bitmap describing capabilities of peer node
96  * @silent_intv_cnt: # of timer intervals without any reception from peer
97  * @proto_msg: template for control messages generated by link
98  * @pmsg: convenience pointer to "proto_msg" field
99  * @priority: current link priority
100  * @net_plane: current link network plane ('A' through 'H')
101  * @mon_state: cookie with information needed by link monitor
102  * @backlog_limit: backlog queue congestion thresholds (indexed by importance)
103  * @exp_msg_count: # of tunnelled messages expected during link changeover
104  * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset
105  * @mtu: current maximum packet size for this link
106  * @advertised_mtu: advertised own mtu when link is being established
107  * @transmitq: queue for sent, non-acked messages
108  * @backlogq: queue for messages waiting to be sent
109  * @snt_nxt: next sequence number to use for outbound messages
110  * @ackers: # of peers that needs to ack each packet before it can be released
111  * @acked: # last packet acked by a certain peer. Used for broadcast.
112  * @rcv_nxt: next sequence number to expect for inbound messages
113  * @deferred_queue: deferred queue saved OOS b'cast message received from node
114  * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
115  * @inputq: buffer queue for messages to be delivered upwards
116  * @namedq: buffer queue for name table messages to be delivered upwards
117  * @next_out: ptr to first unsent outbound message in queue
118  * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
119  * @long_msg_seq_no: next identifier to use for outbound fragmented messages
120  * @reasm_buf: head of partially reassembled inbound message fragments
121  * @bc_rcvr: marks that this is a broadcast receiver link
122  * @stats: collects statistics regarding link activity
123  * @session: session to be used by link
124  * @snd_nxt_state: next send seq number
125  * @rcv_nxt_state: next rcv seq number
126  * @in_session: have received ACTIVATE_MSG from peer
127  * @active: link is active
128  * @if_name: associated interface name
129  * @rst_cnt: link reset counter
130  * @drop_point: seq number for failover handling (FIXME)
131  * @failover_reasm_skb: saved failover msg ptr (FIXME)
132  * @failover_deferdq: deferred message queue for failover processing (FIXME)
133  * @transmq: the link's transmit queue
134  * @backlog: link's backlog by priority (importance)
135  * @snd_nxt: next sequence number to be used
136  * @rcv_unacked: # messages read by user, but not yet acked back to peer
137  * @deferdq: deferred receive queue
138  * @window: sliding window size for congestion handling
139  * @min_win: minimal send window to be used by link
140  * @ssthresh: slow start threshold for congestion handling
141  * @max_win: maximal send window to be used by link
142  * @cong_acks: congestion acks for congestion avoidance (FIXME)
143  * @checkpoint: seq number for congestion window size handling
144  * @reasm_tnlmsg: fragmentation/reassembly area for tunnel protocol message
145  * @last_gap: last gap ack blocks for bcast (FIXME)
146  * @last_ga: ptr to gap ack blocks
147  * @bc_rcvlink: the peer specific link used for broadcast reception
148  * @bc_sndlink: the namespace global link used for broadcast sending
149  * @nack_state: bcast nack state
150  * @bc_peer_is_up: peer has acked the bcast init msg
151  */
152 struct tipc_link {
153 	u32 addr;
154 	char name[TIPC_MAX_LINK_NAME];
155 	struct net *net;
156 
157 	/* Management and link supervision data */
158 	u16 peer_session;
159 	u16 session;
160 	u16 snd_nxt_state;
161 	u16 rcv_nxt_state;
162 	u32 peer_bearer_id;
163 	u32 bearer_id;
164 	u32 tolerance;
165 	u32 abort_limit;
166 	u32 state;
167 	u16 peer_caps;
168 	bool in_session;
169 	bool active;
170 	u32 silent_intv_cnt;
171 	char if_name[TIPC_MAX_IF_NAME];
172 	u32 priority;
173 	char net_plane;
174 	struct tipc_mon_state mon_state;
175 	u16 rst_cnt;
176 
177 	/* Failover/synch */
178 	u16 drop_point;
179 	struct sk_buff *failover_reasm_skb;
180 	struct sk_buff_head failover_deferdq;
181 
182 	/* Max packet negotiation */
183 	u16 mtu;
184 	u16 advertised_mtu;
185 
186 	/* Sending */
187 	struct sk_buff_head transmq;
188 	struct sk_buff_head backlogq;
189 	struct {
190 		u16 len;
191 		u16 limit;
192 		struct sk_buff *target_bskb;
193 	} backlog[5];
194 	u16 snd_nxt;
195 
196 	/* Reception */
197 	u16 rcv_nxt;
198 	u32 rcv_unacked;
199 	struct sk_buff_head deferdq;
200 	struct sk_buff_head *inputq;
201 	struct sk_buff_head *namedq;
202 
203 	/* Congestion handling */
204 	struct sk_buff_head wakeupq;
205 	u16 window;
206 	u16 min_win;
207 	u16 ssthresh;
208 	u16 max_win;
209 	u16 cong_acks;
210 	u16 checkpoint;
211 
212 	/* Fragmentation/reassembly */
213 	struct sk_buff *reasm_buf;
214 	struct sk_buff *reasm_tnlmsg;
215 
216 	/* Broadcast */
217 	u16 ackers;
218 	u16 acked;
219 	u16 last_gap;
220 	struct tipc_gap_ack_blks *last_ga;
221 	struct tipc_link *bc_rcvlink;
222 	struct tipc_link *bc_sndlink;
223 	u8 nack_state;
224 	bool bc_peer_is_up;
225 
226 	/* Statistics */
227 	struct tipc_stats stats;
228 };
229 
230 /*
231  * Error message prefixes
232  */
233 static const char *link_co_err = "Link tunneling error, ";
234 static const char *link_rst_msg = "Resetting link ";
235 
236 /* Send states for broadcast NACKs
237  */
238 enum {
239 	BC_NACK_SND_CONDITIONAL,
240 	BC_NACK_SND_UNCONDITIONAL,
241 	BC_NACK_SND_SUPPRESS,
242 };
243 
244 #define TIPC_BC_RETR_LIM  (jiffies + msecs_to_jiffies(10))
245 #define TIPC_UC_RETR_TIME (jiffies + msecs_to_jiffies(1))
246 
247 /* Link FSM states:
248  */
249 enum {
250 	LINK_ESTABLISHED     = 0xe,
251 	LINK_ESTABLISHING    = 0xe  << 4,
252 	LINK_RESET           = 0x1  << 8,
253 	LINK_RESETTING       = 0x2  << 12,
254 	LINK_PEER_RESET      = 0xd  << 16,
255 	LINK_FAILINGOVER     = 0xf  << 20,
256 	LINK_SYNCHING        = 0xc  << 24
257 };
258 
259 /* Link FSM state checking routines
260  */
link_is_up(struct tipc_link * l)261 static int link_is_up(struct tipc_link *l)
262 {
263 	return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
264 }
265 
266 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
267 			       struct sk_buff_head *xmitq);
268 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
269 				      bool probe_reply, u16 rcvgap,
270 				      int tolerance, int priority,
271 				      struct sk_buff_head *xmitq);
272 static void link_print(struct tipc_link *l, const char *str);
273 static int tipc_link_build_nack_msg(struct tipc_link *l,
274 				    struct sk_buff_head *xmitq);
275 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
276 					struct sk_buff_head *xmitq);
277 static u8 __tipc_build_gap_ack_blks(struct tipc_gap_ack_blks *ga,
278 				    struct tipc_link *l, u8 start_index);
279 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, struct tipc_msg *hdr);
280 static int tipc_link_advance_transmq(struct tipc_link *l, struct tipc_link *r,
281 				     u16 acked, u16 gap,
282 				     struct tipc_gap_ack_blks *ga,
283 				     struct sk_buff_head *xmitq,
284 				     bool *retransmitted, int *rc);
285 static void tipc_link_update_cwin(struct tipc_link *l, int released,
286 				  bool retransmitted);
287 /*
288  *  Simple non-static link routines (i.e. referenced outside this file)
289  */
tipc_link_is_up(struct tipc_link * l)290 bool tipc_link_is_up(struct tipc_link *l)
291 {
292 	return link_is_up(l);
293 }
294 
tipc_link_peer_is_down(struct tipc_link * l)295 bool tipc_link_peer_is_down(struct tipc_link *l)
296 {
297 	return l->state == LINK_PEER_RESET;
298 }
299 
tipc_link_is_reset(struct tipc_link * l)300 bool tipc_link_is_reset(struct tipc_link *l)
301 {
302 	return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
303 }
304 
tipc_link_is_establishing(struct tipc_link * l)305 bool tipc_link_is_establishing(struct tipc_link *l)
306 {
307 	return l->state == LINK_ESTABLISHING;
308 }
309 
tipc_link_is_synching(struct tipc_link * l)310 bool tipc_link_is_synching(struct tipc_link *l)
311 {
312 	return l->state == LINK_SYNCHING;
313 }
314 
tipc_link_is_failingover(struct tipc_link * l)315 bool tipc_link_is_failingover(struct tipc_link *l)
316 {
317 	return l->state == LINK_FAILINGOVER;
318 }
319 
tipc_link_is_blocked(struct tipc_link * l)320 bool tipc_link_is_blocked(struct tipc_link *l)
321 {
322 	return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
323 }
324 
link_is_bc_sndlink(struct tipc_link * l)325 static bool link_is_bc_sndlink(struct tipc_link *l)
326 {
327 	return !l->bc_sndlink;
328 }
329 
link_is_bc_rcvlink(struct tipc_link * l)330 static bool link_is_bc_rcvlink(struct tipc_link *l)
331 {
332 	return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
333 }
334 
tipc_link_set_active(struct tipc_link * l,bool active)335 void tipc_link_set_active(struct tipc_link *l, bool active)
336 {
337 	l->active = active;
338 }
339 
tipc_link_id(struct tipc_link * l)340 u32 tipc_link_id(struct tipc_link *l)
341 {
342 	return l->peer_bearer_id << 16 | l->bearer_id;
343 }
344 
tipc_link_min_win(struct tipc_link * l)345 int tipc_link_min_win(struct tipc_link *l)
346 {
347 	return l->min_win;
348 }
349 
tipc_link_max_win(struct tipc_link * l)350 int tipc_link_max_win(struct tipc_link *l)
351 {
352 	return l->max_win;
353 }
354 
tipc_link_prio(struct tipc_link * l)355 int tipc_link_prio(struct tipc_link *l)
356 {
357 	return l->priority;
358 }
359 
tipc_link_tolerance(struct tipc_link * l)360 unsigned long tipc_link_tolerance(struct tipc_link *l)
361 {
362 	return l->tolerance;
363 }
364 
tipc_link_inputq(struct tipc_link * l)365 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
366 {
367 	return l->inputq;
368 }
369 
tipc_link_plane(struct tipc_link * l)370 char tipc_link_plane(struct tipc_link *l)
371 {
372 	return l->net_plane;
373 }
374 
tipc_link_net(struct tipc_link * l)375 struct net *tipc_link_net(struct tipc_link *l)
376 {
377 	return l->net;
378 }
379 
tipc_link_update_caps(struct tipc_link * l,u16 capabilities)380 void tipc_link_update_caps(struct tipc_link *l, u16 capabilities)
381 {
382 	l->peer_caps = capabilities;
383 }
384 
tipc_link_add_bc_peer(struct tipc_link * snd_l,struct tipc_link * uc_l,struct sk_buff_head * xmitq)385 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
386 			   struct tipc_link *uc_l,
387 			   struct sk_buff_head *xmitq)
388 {
389 	struct tipc_link *rcv_l = uc_l->bc_rcvlink;
390 
391 	snd_l->ackers++;
392 	rcv_l->acked = snd_l->snd_nxt - 1;
393 	snd_l->state = LINK_ESTABLISHED;
394 	tipc_link_build_bc_init_msg(uc_l, xmitq);
395 }
396 
tipc_link_remove_bc_peer(struct tipc_link * snd_l,struct tipc_link * rcv_l,struct sk_buff_head * xmitq)397 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
398 			      struct tipc_link *rcv_l,
399 			      struct sk_buff_head *xmitq)
400 {
401 	u16 ack = snd_l->snd_nxt - 1;
402 
403 	snd_l->ackers--;
404 	rcv_l->bc_peer_is_up = true;
405 	rcv_l->state = LINK_ESTABLISHED;
406 	tipc_link_bc_ack_rcv(rcv_l, ack, 0, NULL, xmitq, NULL);
407 	trace_tipc_link_reset(rcv_l, TIPC_DUMP_ALL, "bclink removed!");
408 	tipc_link_reset(rcv_l);
409 	rcv_l->state = LINK_RESET;
410 	if (!snd_l->ackers) {
411 		trace_tipc_link_reset(snd_l, TIPC_DUMP_ALL, "zero ackers!");
412 		tipc_link_reset(snd_l);
413 		snd_l->state = LINK_RESET;
414 		__skb_queue_purge(xmitq);
415 	}
416 }
417 
tipc_link_bc_peers(struct tipc_link * l)418 int tipc_link_bc_peers(struct tipc_link *l)
419 {
420 	return l->ackers;
421 }
422 
link_bc_rcv_gap(struct tipc_link * l)423 static u16 link_bc_rcv_gap(struct tipc_link *l)
424 {
425 	struct sk_buff *skb = skb_peek(&l->deferdq);
426 	u16 gap = 0;
427 
428 	if (more(l->snd_nxt, l->rcv_nxt))
429 		gap = l->snd_nxt - l->rcv_nxt;
430 	if (skb)
431 		gap = buf_seqno(skb) - l->rcv_nxt;
432 	return gap;
433 }
434 
tipc_link_set_mtu(struct tipc_link * l,int mtu)435 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
436 {
437 	l->mtu = mtu;
438 }
439 
tipc_link_mtu(struct tipc_link * l)440 int tipc_link_mtu(struct tipc_link *l)
441 {
442 	return l->mtu;
443 }
444 
tipc_link_mss(struct tipc_link * l)445 int tipc_link_mss(struct tipc_link *l)
446 {
447 #ifdef CONFIG_TIPC_CRYPTO
448 	return l->mtu - INT_H_SIZE - EMSG_OVERHEAD;
449 #else
450 	return l->mtu - INT_H_SIZE;
451 #endif
452 }
453 
tipc_link_rcv_nxt(struct tipc_link * l)454 u16 tipc_link_rcv_nxt(struct tipc_link *l)
455 {
456 	return l->rcv_nxt;
457 }
458 
tipc_link_acked(struct tipc_link * l)459 u16 tipc_link_acked(struct tipc_link *l)
460 {
461 	return l->acked;
462 }
463 
tipc_link_name(struct tipc_link * l)464 char *tipc_link_name(struct tipc_link *l)
465 {
466 	return l->name;
467 }
468 
tipc_link_state(struct tipc_link * l)469 u32 tipc_link_state(struct tipc_link *l)
470 {
471 	return l->state;
472 }
473 
474 /**
475  * tipc_link_create - create a new link
476  * @net: pointer to associated network namespace
477  * @if_name: associated interface name
478  * @bearer_id: id (index) of associated bearer
479  * @tolerance: link tolerance to be used by link
480  * @net_plane: network plane (A,B,c..) this link belongs to
481  * @mtu: mtu to be advertised by link
482  * @priority: priority to be used by link
483  * @min_win: minimal send window to be used by link
484  * @max_win: maximal send window to be used by link
485  * @session: session to be used by link
486  * @peer: node id of peer node
487  * @peer_caps: bitmap describing peer node capabilities
488  * @bc_sndlink: the namespace global link used for broadcast sending
489  * @bc_rcvlink: the peer specific link used for broadcast reception
490  * @inputq: queue to put messages ready for delivery
491  * @namedq: queue to put binding table update messages ready for delivery
492  * @link: return value, pointer to put the created link
493  * @self: local unicast link id
494  * @peer_id: 128-bit ID of peer
495  *
496  * Return: true if link was created, otherwise false
497  */
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)498 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
499 		      int tolerance, char net_plane, u32 mtu, int priority,
500 		      u32 min_win, u32 max_win, u32 session, u32 self,
501 		      u32 peer, u8 *peer_id, u16 peer_caps,
502 		      struct tipc_link *bc_sndlink,
503 		      struct tipc_link *bc_rcvlink,
504 		      struct sk_buff_head *inputq,
505 		      struct sk_buff_head *namedq,
506 		      struct tipc_link **link)
507 {
508 	char peer_str[NODE_ID_STR_LEN] = {0,};
509 	char self_str[NODE_ID_STR_LEN] = {0,};
510 	struct tipc_link *l;
511 
512 	l = kzalloc(sizeof(*l), GFP_ATOMIC);
513 	if (!l)
514 		return false;
515 	*link = l;
516 	l->session = session;
517 
518 	/* Set link name for unicast links only */
519 	if (peer_id) {
520 		tipc_nodeid2string(self_str, tipc_own_id(net));
521 		if (strlen(self_str) > 16)
522 			sprintf(self_str, "%x", self);
523 		tipc_nodeid2string(peer_str, peer_id);
524 		if (strlen(peer_str) > 16)
525 			sprintf(peer_str, "%x", peer);
526 	}
527 	/* Peer i/f name will be completed by reset/activate message */
528 	snprintf(l->name, sizeof(l->name), "%s:%s-%s:unknown",
529 		 self_str, if_name, peer_str);
530 
531 	strcpy(l->if_name, if_name);
532 	l->addr = peer;
533 	l->peer_caps = peer_caps;
534 	l->net = net;
535 	l->in_session = false;
536 	l->bearer_id = bearer_id;
537 	l->tolerance = tolerance;
538 	if (bc_rcvlink)
539 		bc_rcvlink->tolerance = tolerance;
540 	l->net_plane = net_plane;
541 	l->advertised_mtu = mtu;
542 	l->mtu = mtu;
543 	l->priority = priority;
544 	tipc_link_set_queue_limits(l, min_win, max_win);
545 	l->ackers = 1;
546 	l->bc_sndlink = bc_sndlink;
547 	l->bc_rcvlink = bc_rcvlink;
548 	l->inputq = inputq;
549 	l->namedq = namedq;
550 	l->state = LINK_RESETTING;
551 	__skb_queue_head_init(&l->transmq);
552 	__skb_queue_head_init(&l->backlogq);
553 	__skb_queue_head_init(&l->deferdq);
554 	__skb_queue_head_init(&l->failover_deferdq);
555 	skb_queue_head_init(&l->wakeupq);
556 	skb_queue_head_init(l->inputq);
557 	return true;
558 }
559 
560 /**
561  * tipc_link_bc_create - create new link to be used for broadcast
562  * @net: pointer to associated network namespace
563  * @mtu: mtu to be used initially if no peers
564  * @min_win: minimal send window to be used by link
565  * @max_win: maximal send window to be used by link
566  * @inputq: queue to put messages ready for delivery
567  * @namedq: queue to put binding table update messages ready for delivery
568  * @link: return value, pointer to put the created link
569  * @ownnode: identity of own node
570  * @peer: node id of peer node
571  * @peer_id: 128-bit ID of peer
572  * @peer_caps: bitmap describing peer node capabilities
573  * @bc_sndlink: the namespace global link used for broadcast sending
574  *
575  * Return: true if link was created, otherwise false
576  */
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)577 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer, u8 *peer_id,
578 			 int mtu, u32 min_win, u32 max_win, u16 peer_caps,
579 			 struct sk_buff_head *inputq,
580 			 struct sk_buff_head *namedq,
581 			 struct tipc_link *bc_sndlink,
582 			 struct tipc_link **link)
583 {
584 	struct tipc_link *l;
585 
586 	if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, min_win,
587 			      max_win, 0, ownnode, peer, NULL, peer_caps,
588 			      bc_sndlink, NULL, inputq, namedq, link))
589 		return false;
590 
591 	l = *link;
592 	if (peer_id) {
593 		char peer_str[NODE_ID_STR_LEN] = {0,};
594 
595 		tipc_nodeid2string(peer_str, peer_id);
596 		if (strlen(peer_str) > 16)
597 			sprintf(peer_str, "%x", peer);
598 		/* Broadcast receiver link name: "broadcast-link:<peer>" */
599 		snprintf(l->name, sizeof(l->name), "%s:%s", tipc_bclink_name,
600 			 peer_str);
601 	} else {
602 		strcpy(l->name, tipc_bclink_name);
603 	}
604 	trace_tipc_link_reset(l, TIPC_DUMP_ALL, "bclink created!");
605 	tipc_link_reset(l);
606 	l->state = LINK_RESET;
607 	l->ackers = 0;
608 	l->bc_rcvlink = l;
609 
610 	/* Broadcast send link is always up */
611 	if (link_is_bc_sndlink(l))
612 		l->state = LINK_ESTABLISHED;
613 
614 	/* Disable replicast if even a single peer doesn't support it */
615 	if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
616 		tipc_bcast_toggle_rcast(net, false);
617 
618 	return true;
619 }
620 
621 /**
622  * tipc_link_fsm_evt - link finite state machine
623  * @l: pointer to link
624  * @evt: state machine event to be processed
625  */
tipc_link_fsm_evt(struct tipc_link * l,int evt)626 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
627 {
628 	int rc = 0;
629 	int old_state = l->state;
630 
631 	switch (l->state) {
632 	case LINK_RESETTING:
633 		switch (evt) {
634 		case LINK_PEER_RESET_EVT:
635 			l->state = LINK_PEER_RESET;
636 			break;
637 		case LINK_RESET_EVT:
638 			l->state = LINK_RESET;
639 			break;
640 		case LINK_FAILURE_EVT:
641 		case LINK_FAILOVER_BEGIN_EVT:
642 		case LINK_ESTABLISH_EVT:
643 		case LINK_FAILOVER_END_EVT:
644 		case LINK_SYNCH_BEGIN_EVT:
645 		case LINK_SYNCH_END_EVT:
646 		default:
647 			goto illegal_evt;
648 		}
649 		break;
650 	case LINK_RESET:
651 		switch (evt) {
652 		case LINK_PEER_RESET_EVT:
653 			l->state = LINK_ESTABLISHING;
654 			break;
655 		case LINK_FAILOVER_BEGIN_EVT:
656 			l->state = LINK_FAILINGOVER;
657 			break;
658 		case LINK_FAILURE_EVT:
659 		case LINK_RESET_EVT:
660 		case LINK_ESTABLISH_EVT:
661 		case LINK_FAILOVER_END_EVT:
662 			break;
663 		case LINK_SYNCH_BEGIN_EVT:
664 		case LINK_SYNCH_END_EVT:
665 		default:
666 			goto illegal_evt;
667 		}
668 		break;
669 	case LINK_PEER_RESET:
670 		switch (evt) {
671 		case LINK_RESET_EVT:
672 			l->state = LINK_ESTABLISHING;
673 			break;
674 		case LINK_PEER_RESET_EVT:
675 		case LINK_ESTABLISH_EVT:
676 		case LINK_FAILURE_EVT:
677 			break;
678 		case LINK_SYNCH_BEGIN_EVT:
679 		case LINK_SYNCH_END_EVT:
680 		case LINK_FAILOVER_BEGIN_EVT:
681 		case LINK_FAILOVER_END_EVT:
682 		default:
683 			goto illegal_evt;
684 		}
685 		break;
686 	case LINK_FAILINGOVER:
687 		switch (evt) {
688 		case LINK_FAILOVER_END_EVT:
689 			l->state = LINK_RESET;
690 			break;
691 		case LINK_PEER_RESET_EVT:
692 		case LINK_RESET_EVT:
693 		case LINK_ESTABLISH_EVT:
694 		case LINK_FAILURE_EVT:
695 			break;
696 		case LINK_FAILOVER_BEGIN_EVT:
697 		case LINK_SYNCH_BEGIN_EVT:
698 		case LINK_SYNCH_END_EVT:
699 		default:
700 			goto illegal_evt;
701 		}
702 		break;
703 	case LINK_ESTABLISHING:
704 		switch (evt) {
705 		case LINK_ESTABLISH_EVT:
706 			l->state = LINK_ESTABLISHED;
707 			break;
708 		case LINK_FAILOVER_BEGIN_EVT:
709 			l->state = LINK_FAILINGOVER;
710 			break;
711 		case LINK_RESET_EVT:
712 			l->state = LINK_RESET;
713 			break;
714 		case LINK_FAILURE_EVT:
715 		case LINK_PEER_RESET_EVT:
716 		case LINK_SYNCH_BEGIN_EVT:
717 		case LINK_FAILOVER_END_EVT:
718 			break;
719 		case LINK_SYNCH_END_EVT:
720 		default:
721 			goto illegal_evt;
722 		}
723 		break;
724 	case LINK_ESTABLISHED:
725 		switch (evt) {
726 		case LINK_PEER_RESET_EVT:
727 			l->state = LINK_PEER_RESET;
728 			rc |= TIPC_LINK_DOWN_EVT;
729 			break;
730 		case LINK_FAILURE_EVT:
731 			l->state = LINK_RESETTING;
732 			rc |= TIPC_LINK_DOWN_EVT;
733 			break;
734 		case LINK_RESET_EVT:
735 			l->state = LINK_RESET;
736 			break;
737 		case LINK_ESTABLISH_EVT:
738 		case LINK_SYNCH_END_EVT:
739 			break;
740 		case LINK_SYNCH_BEGIN_EVT:
741 			l->state = LINK_SYNCHING;
742 			break;
743 		case LINK_FAILOVER_BEGIN_EVT:
744 		case LINK_FAILOVER_END_EVT:
745 		default:
746 			goto illegal_evt;
747 		}
748 		break;
749 	case LINK_SYNCHING:
750 		switch (evt) {
751 		case LINK_PEER_RESET_EVT:
752 			l->state = LINK_PEER_RESET;
753 			rc |= TIPC_LINK_DOWN_EVT;
754 			break;
755 		case LINK_FAILURE_EVT:
756 			l->state = LINK_RESETTING;
757 			rc |= TIPC_LINK_DOWN_EVT;
758 			break;
759 		case LINK_RESET_EVT:
760 			l->state = LINK_RESET;
761 			break;
762 		case LINK_ESTABLISH_EVT:
763 		case LINK_SYNCH_BEGIN_EVT:
764 			break;
765 		case LINK_SYNCH_END_EVT:
766 			l->state = LINK_ESTABLISHED;
767 			break;
768 		case LINK_FAILOVER_BEGIN_EVT:
769 		case LINK_FAILOVER_END_EVT:
770 		default:
771 			goto illegal_evt;
772 		}
773 		break;
774 	default:
775 		pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
776 	}
777 	trace_tipc_link_fsm(l->name, old_state, l->state, evt);
778 	return rc;
779 illegal_evt:
780 	pr_err("Illegal FSM event %x in state %x on link %s\n",
781 	       evt, l->state, l->name);
782 	trace_tipc_link_fsm(l->name, old_state, l->state, evt);
783 	return rc;
784 }
785 
786 /* link_profile_stats - update statistical profiling of traffic
787  */
link_profile_stats(struct tipc_link * l)788 static void link_profile_stats(struct tipc_link *l)
789 {
790 	struct sk_buff *skb;
791 	struct tipc_msg *msg;
792 	int length;
793 
794 	/* Update counters used in statistical profiling of send traffic */
795 	l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
796 	l->stats.queue_sz_counts++;
797 
798 	skb = skb_peek(&l->transmq);
799 	if (!skb)
800 		return;
801 	msg = buf_msg(skb);
802 	length = msg_size(msg);
803 
804 	if (msg_user(msg) == MSG_FRAGMENTER) {
805 		if (msg_type(msg) != FIRST_FRAGMENT)
806 			return;
807 		length = msg_size(msg_inner_hdr(msg));
808 	}
809 	l->stats.msg_lengths_total += length;
810 	l->stats.msg_length_counts++;
811 	if (length <= 64)
812 		l->stats.msg_length_profile[0]++;
813 	else if (length <= 256)
814 		l->stats.msg_length_profile[1]++;
815 	else if (length <= 1024)
816 		l->stats.msg_length_profile[2]++;
817 	else if (length <= 4096)
818 		l->stats.msg_length_profile[3]++;
819 	else if (length <= 16384)
820 		l->stats.msg_length_profile[4]++;
821 	else if (length <= 32768)
822 		l->stats.msg_length_profile[5]++;
823 	else
824 		l->stats.msg_length_profile[6]++;
825 }
826 
827 /**
828  * tipc_link_too_silent - check if link is "too silent"
829  * @l: tipc link to be checked
830  *
831  * Return: true if the link 'silent_intv_cnt' is about to reach the
832  * 'abort_limit' value, otherwise false
833  */
tipc_link_too_silent(struct tipc_link * l)834 bool tipc_link_too_silent(struct tipc_link *l)
835 {
836 	return (l->silent_intv_cnt + 2 > l->abort_limit);
837 }
838 
839 /* tipc_link_timeout - perform periodic task as instructed from node timeout
840  */
tipc_link_timeout(struct tipc_link * l,struct sk_buff_head * xmitq)841 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
842 {
843 	int mtyp = 0;
844 	int rc = 0;
845 	bool state = false;
846 	bool probe = false;
847 	bool setup = false;
848 	u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
849 	u16 bc_acked = l->bc_rcvlink->acked;
850 	struct tipc_mon_state *mstate = &l->mon_state;
851 
852 	trace_tipc_link_timeout(l, TIPC_DUMP_NONE, " ");
853 	trace_tipc_link_too_silent(l, TIPC_DUMP_ALL, " ");
854 	switch (l->state) {
855 	case LINK_ESTABLISHED:
856 	case LINK_SYNCHING:
857 		mtyp = STATE_MSG;
858 		link_profile_stats(l);
859 		tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
860 		if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
861 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
862 		state = bc_acked != bc_snt;
863 		state |= l->bc_rcvlink->rcv_unacked;
864 		state |= l->rcv_unacked;
865 		state |= !skb_queue_empty(&l->transmq);
866 		probe = mstate->probing;
867 		probe |= l->silent_intv_cnt;
868 		if (probe || mstate->monitoring)
869 			l->silent_intv_cnt++;
870 		probe |= !skb_queue_empty(&l->deferdq);
871 		if (l->snd_nxt == l->checkpoint) {
872 			tipc_link_update_cwin(l, 0, 0);
873 			probe = true;
874 		}
875 		l->checkpoint = l->snd_nxt;
876 		break;
877 	case LINK_RESET:
878 		setup = l->rst_cnt++ <= 4;
879 		setup |= !(l->rst_cnt % 16);
880 		mtyp = RESET_MSG;
881 		break;
882 	case LINK_ESTABLISHING:
883 		setup = true;
884 		mtyp = ACTIVATE_MSG;
885 		break;
886 	case LINK_PEER_RESET:
887 	case LINK_RESETTING:
888 	case LINK_FAILINGOVER:
889 		break;
890 	default:
891 		break;
892 	}
893 
894 	if (state || probe || setup)
895 		tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq);
896 
897 	return rc;
898 }
899 
900 /**
901  * link_schedule_user - schedule a message sender for wakeup after congestion
902  * @l: congested link
903  * @hdr: header of message that is being sent
904  * Create pseudo msg to send back to user when congestion abates
905  */
link_schedule_user(struct tipc_link * l,struct tipc_msg * hdr)906 static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
907 {
908 	u32 dnode = tipc_own_addr(l->net);
909 	u32 dport = msg_origport(hdr);
910 	struct sk_buff *skb;
911 
912 	/* Create and schedule wakeup pseudo message */
913 	skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
914 			      dnode, l->addr, dport, 0, 0);
915 	if (!skb)
916 		return -ENOBUFS;
917 	msg_set_dest_droppable(buf_msg(skb), true);
918 	TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
919 	skb_queue_tail(&l->wakeupq, skb);
920 	l->stats.link_congs++;
921 	trace_tipc_link_conges(l, TIPC_DUMP_ALL, "wakeup scheduled!");
922 	return -ELINKCONG;
923 }
924 
925 /**
926  * link_prepare_wakeup - prepare users for wakeup after congestion
927  * @l: congested link
928  * Wake up a number of waiting users, as permitted by available space
929  * in the send queue
930  */
link_prepare_wakeup(struct tipc_link * l)931 static void link_prepare_wakeup(struct tipc_link *l)
932 {
933 	struct sk_buff_head *wakeupq = &l->wakeupq;
934 	struct sk_buff_head *inputq = l->inputq;
935 	struct sk_buff *skb, *tmp;
936 	struct sk_buff_head tmpq;
937 	int avail[5] = {0,};
938 	int imp = 0;
939 
940 	__skb_queue_head_init(&tmpq);
941 
942 	for (; imp <= TIPC_SYSTEM_IMPORTANCE; imp++)
943 		avail[imp] = l->backlog[imp].limit - l->backlog[imp].len;
944 
945 	skb_queue_walk_safe(wakeupq, skb, tmp) {
946 		imp = TIPC_SKB_CB(skb)->chain_imp;
947 		if (avail[imp] <= 0)
948 			continue;
949 		avail[imp]--;
950 		__skb_unlink(skb, wakeupq);
951 		__skb_queue_tail(&tmpq, skb);
952 	}
953 
954 	spin_lock_bh(&inputq->lock);
955 	skb_queue_splice_tail(&tmpq, inputq);
956 	spin_unlock_bh(&inputq->lock);
957 
958 }
959 
960 /**
961  * tipc_link_set_skb_retransmit_time - set the time at which retransmission of
962  *                                     the given skb should be next attempted
963  * @skb: skb to set a future retransmission time for
964  * @l: link the skb will be transmitted on
965  */
tipc_link_set_skb_retransmit_time(struct sk_buff * skb,struct tipc_link * l)966 static void tipc_link_set_skb_retransmit_time(struct sk_buff *skb,
967 					      struct tipc_link *l)
968 {
969 	if (link_is_bc_sndlink(l))
970 		TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
971 	else
972 		TIPC_SKB_CB(skb)->nxt_retr = TIPC_UC_RETR_TIME;
973 }
974 
tipc_link_reset(struct tipc_link * l)975 void tipc_link_reset(struct tipc_link *l)
976 {
977 	struct sk_buff_head list;
978 	u32 imp;
979 
980 	__skb_queue_head_init(&list);
981 
982 	l->in_session = false;
983 	/* Force re-synch of peer session number before establishing */
984 	l->peer_session--;
985 	l->session++;
986 	l->mtu = l->advertised_mtu;
987 
988 	spin_lock_bh(&l->wakeupq.lock);
989 	skb_queue_splice_init(&l->wakeupq, &list);
990 	spin_unlock_bh(&l->wakeupq.lock);
991 
992 	spin_lock_bh(&l->inputq->lock);
993 	skb_queue_splice_init(&list, l->inputq);
994 	spin_unlock_bh(&l->inputq->lock);
995 
996 	__skb_queue_purge(&l->transmq);
997 	__skb_queue_purge(&l->deferdq);
998 	__skb_queue_purge(&l->backlogq);
999 	__skb_queue_purge(&l->failover_deferdq);
1000 	for (imp = 0; imp <= TIPC_SYSTEM_IMPORTANCE; imp++) {
1001 		l->backlog[imp].len = 0;
1002 		l->backlog[imp].target_bskb = NULL;
1003 	}
1004 	kfree_skb(l->reasm_buf);
1005 	kfree_skb(l->reasm_tnlmsg);
1006 	kfree_skb(l->failover_reasm_skb);
1007 	l->reasm_buf = NULL;
1008 	l->reasm_tnlmsg = NULL;
1009 	l->failover_reasm_skb = NULL;
1010 	l->rcv_unacked = 0;
1011 	l->snd_nxt = 1;
1012 	l->rcv_nxt = 1;
1013 	l->snd_nxt_state = 1;
1014 	l->rcv_nxt_state = 1;
1015 	l->acked = 0;
1016 	l->last_gap = 0;
1017 	kfree(l->last_ga);
1018 	l->last_ga = NULL;
1019 	l->silent_intv_cnt = 0;
1020 	l->rst_cnt = 0;
1021 	l->bc_peer_is_up = false;
1022 	memset(&l->mon_state, 0, sizeof(l->mon_state));
1023 	tipc_link_reset_stats(l);
1024 }
1025 
1026 /**
1027  * tipc_link_xmit(): enqueue buffer list according to queue situation
1028  * @l: link to use
1029  * @list: chain of buffers containing message
1030  * @xmitq: returned list of packets to be sent by caller
1031  *
1032  * Consumes the buffer chain.
1033  * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
1034  * Return: 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
1035  */
tipc_link_xmit(struct tipc_link * l,struct sk_buff_head * list,struct sk_buff_head * xmitq)1036 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
1037 		   struct sk_buff_head *xmitq)
1038 {
1039 	struct sk_buff_head *backlogq = &l->backlogq;
1040 	struct sk_buff_head *transmq = &l->transmq;
1041 	struct sk_buff *skb, *_skb;
1042 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1043 	u16 ack = l->rcv_nxt - 1;
1044 	u16 seqno = l->snd_nxt;
1045 	int pkt_cnt = skb_queue_len(list);
1046 	unsigned int mss = tipc_link_mss(l);
1047 	unsigned int cwin = l->window;
1048 	unsigned int mtu = l->mtu;
1049 	struct tipc_msg *hdr;
1050 	bool new_bundle;
1051 	int rc = 0;
1052 	int imp;
1053 
1054 	if (pkt_cnt <= 0)
1055 		return 0;
1056 
1057 	hdr = buf_msg(skb_peek(list));
1058 	if (unlikely(msg_size(hdr) > mtu)) {
1059 		pr_warn("Too large msg, purging xmit list %d %d %d %d %d!\n",
1060 			skb_queue_len(list), msg_user(hdr),
1061 			msg_type(hdr), msg_size(hdr), mtu);
1062 		__skb_queue_purge(list);
1063 		return -EMSGSIZE;
1064 	}
1065 
1066 	imp = msg_importance(hdr);
1067 	/* Allow oversubscription of one data msg per source at congestion */
1068 	if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
1069 		if (imp == TIPC_SYSTEM_IMPORTANCE) {
1070 			pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
1071 			return -ENOBUFS;
1072 		}
1073 		rc = link_schedule_user(l, hdr);
1074 	}
1075 
1076 	if (pkt_cnt > 1) {
1077 		l->stats.sent_fragmented++;
1078 		l->stats.sent_fragments += pkt_cnt;
1079 	}
1080 
1081 	/* Prepare each packet for sending, and add to relevant queue: */
1082 	while ((skb = __skb_dequeue(list))) {
1083 		if (likely(skb_queue_len(transmq) < cwin)) {
1084 			hdr = buf_msg(skb);
1085 			msg_set_seqno(hdr, seqno);
1086 			msg_set_ack(hdr, ack);
1087 			msg_set_bcast_ack(hdr, bc_ack);
1088 			_skb = skb_clone(skb, GFP_ATOMIC);
1089 			if (!_skb) {
1090 				kfree_skb(skb);
1091 				__skb_queue_purge(list);
1092 				return -ENOBUFS;
1093 			}
1094 			__skb_queue_tail(transmq, skb);
1095 			tipc_link_set_skb_retransmit_time(skb, l);
1096 			__skb_queue_tail(xmitq, _skb);
1097 			TIPC_SKB_CB(skb)->ackers = l->ackers;
1098 			l->rcv_unacked = 0;
1099 			l->stats.sent_pkts++;
1100 			seqno++;
1101 			continue;
1102 		}
1103 		if (tipc_msg_try_bundle(l->backlog[imp].target_bskb, &skb,
1104 					mss, l->addr, &new_bundle)) {
1105 			if (skb) {
1106 				/* Keep a ref. to the skb for next try */
1107 				l->backlog[imp].target_bskb = skb;
1108 				l->backlog[imp].len++;
1109 				__skb_queue_tail(backlogq, skb);
1110 			} else {
1111 				if (new_bundle) {
1112 					l->stats.sent_bundles++;
1113 					l->stats.sent_bundled++;
1114 				}
1115 				l->stats.sent_bundled++;
1116 			}
1117 			continue;
1118 		}
1119 		l->backlog[imp].target_bskb = NULL;
1120 		l->backlog[imp].len += (1 + skb_queue_len(list));
1121 		__skb_queue_tail(backlogq, skb);
1122 		skb_queue_splice_tail_init(list, backlogq);
1123 	}
1124 	l->snd_nxt = seqno;
1125 	return rc;
1126 }
1127 
tipc_link_update_cwin(struct tipc_link * l,int released,bool retransmitted)1128 static void tipc_link_update_cwin(struct tipc_link *l, int released,
1129 				  bool retransmitted)
1130 {
1131 	int bklog_len = skb_queue_len(&l->backlogq);
1132 	struct sk_buff_head *txq = &l->transmq;
1133 	int txq_len = skb_queue_len(txq);
1134 	u16 cwin = l->window;
1135 
1136 	/* Enter fast recovery */
1137 	if (unlikely(retransmitted)) {
1138 		l->ssthresh = max_t(u16, l->window / 2, 300);
1139 		l->window = min_t(u16, l->ssthresh, l->window);
1140 		return;
1141 	}
1142 	/* Enter slow start */
1143 	if (unlikely(!released)) {
1144 		l->ssthresh = max_t(u16, l->window / 2, 300);
1145 		l->window = l->min_win;
1146 		return;
1147 	}
1148 	/* Don't increase window if no pressure on the transmit queue */
1149 	if (txq_len + bklog_len < cwin)
1150 		return;
1151 
1152 	/* Don't increase window if there are holes the transmit queue */
1153 	if (txq_len && l->snd_nxt - buf_seqno(skb_peek(txq)) != txq_len)
1154 		return;
1155 
1156 	l->cong_acks += released;
1157 
1158 	/* Slow start  */
1159 	if (cwin <= l->ssthresh) {
1160 		l->window = min_t(u16, cwin + released, l->max_win);
1161 		return;
1162 	}
1163 	/* Congestion avoidance */
1164 	if (l->cong_acks < cwin)
1165 		return;
1166 	l->window = min_t(u16, ++cwin, l->max_win);
1167 	l->cong_acks = 0;
1168 }
1169 
tipc_link_advance_backlog(struct tipc_link * l,struct sk_buff_head * xmitq)1170 static void tipc_link_advance_backlog(struct tipc_link *l,
1171 				      struct sk_buff_head *xmitq)
1172 {
1173 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1174 	struct sk_buff_head *txq = &l->transmq;
1175 	struct sk_buff *skb, *_skb;
1176 	u16 ack = l->rcv_nxt - 1;
1177 	u16 seqno = l->snd_nxt;
1178 	struct tipc_msg *hdr;
1179 	u16 cwin = l->window;
1180 	u32 imp;
1181 
1182 	while (skb_queue_len(txq) < cwin) {
1183 		skb = skb_peek(&l->backlogq);
1184 		if (!skb)
1185 			break;
1186 		_skb = skb_clone(skb, GFP_ATOMIC);
1187 		if (!_skb)
1188 			break;
1189 		__skb_dequeue(&l->backlogq);
1190 		hdr = buf_msg(skb);
1191 		imp = msg_importance(hdr);
1192 		l->backlog[imp].len--;
1193 		if (unlikely(skb == l->backlog[imp].target_bskb))
1194 			l->backlog[imp].target_bskb = NULL;
1195 		__skb_queue_tail(&l->transmq, skb);
1196 		tipc_link_set_skb_retransmit_time(skb, l);
1197 
1198 		__skb_queue_tail(xmitq, _skb);
1199 		TIPC_SKB_CB(skb)->ackers = l->ackers;
1200 		msg_set_seqno(hdr, seqno);
1201 		msg_set_ack(hdr, ack);
1202 		msg_set_bcast_ack(hdr, bc_ack);
1203 		l->rcv_unacked = 0;
1204 		l->stats.sent_pkts++;
1205 		seqno++;
1206 	}
1207 	l->snd_nxt = seqno;
1208 }
1209 
1210 /**
1211  * link_retransmit_failure() - Detect repeated retransmit failures
1212  * @l: tipc link sender
1213  * @r: tipc link receiver (= l in case of unicast)
1214  * @rc: returned code
1215  *
1216  * Return: true if the repeated retransmit failures happens, otherwise
1217  * false
1218  */
link_retransmit_failure(struct tipc_link * l,struct tipc_link * r,int * rc)1219 static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r,
1220 				    int *rc)
1221 {
1222 	struct sk_buff *skb = skb_peek(&l->transmq);
1223 	struct tipc_msg *hdr;
1224 
1225 	if (!skb)
1226 		return false;
1227 
1228 	if (!TIPC_SKB_CB(skb)->retr_cnt)
1229 		return false;
1230 
1231 	if (!time_after(jiffies, TIPC_SKB_CB(skb)->retr_stamp +
1232 			msecs_to_jiffies(r->tolerance * 10)))
1233 		return false;
1234 
1235 	hdr = buf_msg(skb);
1236 	if (link_is_bc_sndlink(l) && !less(r->acked, msg_seqno(hdr)))
1237 		return false;
1238 
1239 	pr_warn("Retransmission failure on link <%s>\n", l->name);
1240 	link_print(l, "State of link ");
1241 	pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
1242 		msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
1243 	pr_info("sqno %u, prev: %x, dest: %x\n",
1244 		msg_seqno(hdr), msg_prevnode(hdr), msg_destnode(hdr));
1245 	pr_info("retr_stamp %d, retr_cnt %d\n",
1246 		jiffies_to_msecs(TIPC_SKB_CB(skb)->retr_stamp),
1247 		TIPC_SKB_CB(skb)->retr_cnt);
1248 
1249 	trace_tipc_list_dump(&l->transmq, true, "retrans failure!");
1250 	trace_tipc_link_dump(l, TIPC_DUMP_NONE, "retrans failure!");
1251 	trace_tipc_link_dump(r, TIPC_DUMP_NONE, "retrans failure!");
1252 
1253 	if (link_is_bc_sndlink(l)) {
1254 		r->state = LINK_RESET;
1255 		*rc |= TIPC_LINK_DOWN_EVT;
1256 	} else {
1257 		*rc |= tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1258 	}
1259 
1260 	return true;
1261 }
1262 
1263 /* tipc_data_input - deliver data and name distr msgs to upper layer
1264  *
1265  * Consumes buffer if message is of right type
1266  * Node lock must be held
1267  */
tipc_data_input(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * inputq)1268 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1269 			    struct sk_buff_head *inputq)
1270 {
1271 	struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq;
1272 	struct tipc_msg *hdr = buf_msg(skb);
1273 
1274 	switch (msg_user(hdr)) {
1275 	case TIPC_LOW_IMPORTANCE:
1276 	case TIPC_MEDIUM_IMPORTANCE:
1277 	case TIPC_HIGH_IMPORTANCE:
1278 	case TIPC_CRITICAL_IMPORTANCE:
1279 		if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) {
1280 			skb_queue_tail(mc_inputq, skb);
1281 			return true;
1282 		}
1283 		fallthrough;
1284 	case CONN_MANAGER:
1285 		skb_queue_tail(inputq, skb);
1286 		return true;
1287 	case GROUP_PROTOCOL:
1288 		skb_queue_tail(mc_inputq, skb);
1289 		return true;
1290 	case NAME_DISTRIBUTOR:
1291 		l->bc_rcvlink->state = LINK_ESTABLISHED;
1292 		skb_queue_tail(l->namedq, skb);
1293 		return true;
1294 	case MSG_BUNDLER:
1295 	case TUNNEL_PROTOCOL:
1296 	case MSG_FRAGMENTER:
1297 	case BCAST_PROTOCOL:
1298 		return false;
1299 #ifdef CONFIG_TIPC_CRYPTO
1300 	case MSG_CRYPTO:
1301 		if (sysctl_tipc_key_exchange_enabled &&
1302 		    TIPC_SKB_CB(skb)->decrypted) {
1303 			tipc_crypto_msg_rcv(l->net, skb);
1304 			return true;
1305 		}
1306 		fallthrough;
1307 #endif
1308 	default:
1309 		pr_warn("Dropping received illegal msg type\n");
1310 		kfree_skb(skb);
1311 		return true;
1312 	}
1313 }
1314 
1315 /* tipc_link_input - process packet that has passed link protocol check
1316  *
1317  * Consumes buffer
1318  */
tipc_link_input(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * inputq,struct sk_buff ** reasm_skb)1319 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1320 			   struct sk_buff_head *inputq,
1321 			   struct sk_buff **reasm_skb)
1322 {
1323 	struct tipc_msg *hdr = buf_msg(skb);
1324 	struct sk_buff *iskb;
1325 	struct sk_buff_head tmpq;
1326 	int usr = msg_user(hdr);
1327 	int pos = 0;
1328 
1329 	if (usr == MSG_BUNDLER) {
1330 		skb_queue_head_init(&tmpq);
1331 		l->stats.recv_bundles++;
1332 		l->stats.recv_bundled += msg_msgcnt(hdr);
1333 		while (tipc_msg_extract(skb, &iskb, &pos))
1334 			tipc_data_input(l, iskb, &tmpq);
1335 		tipc_skb_queue_splice_tail(&tmpq, inputq);
1336 		return 0;
1337 	} else if (usr == MSG_FRAGMENTER) {
1338 		l->stats.recv_fragments++;
1339 		if (tipc_buf_append(reasm_skb, &skb)) {
1340 			l->stats.recv_fragmented++;
1341 			tipc_data_input(l, skb, inputq);
1342 		} else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1343 			pr_warn_ratelimited("Unable to build fragment list\n");
1344 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1345 		}
1346 		return 0;
1347 	} else if (usr == BCAST_PROTOCOL) {
1348 		tipc_bcast_lock(l->net);
1349 		tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1350 		tipc_bcast_unlock(l->net);
1351 	}
1352 
1353 	kfree_skb(skb);
1354 	return 0;
1355 }
1356 
1357 /* tipc_link_tnl_rcv() - receive TUNNEL_PROTOCOL message, drop or process the
1358  *			 inner message along with the ones in the old link's
1359  *			 deferdq
1360  * @l: tunnel link
1361  * @skb: TUNNEL_PROTOCOL message
1362  * @inputq: queue to put messages ready for delivery
1363  */
tipc_link_tnl_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * inputq)1364 static int tipc_link_tnl_rcv(struct tipc_link *l, struct sk_buff *skb,
1365 			     struct sk_buff_head *inputq)
1366 {
1367 	struct sk_buff **reasm_skb = &l->failover_reasm_skb;
1368 	struct sk_buff **reasm_tnlmsg = &l->reasm_tnlmsg;
1369 	struct sk_buff_head *fdefq = &l->failover_deferdq;
1370 	struct tipc_msg *hdr = buf_msg(skb);
1371 	struct sk_buff *iskb;
1372 	int ipos = 0;
1373 	int rc = 0;
1374 	u16 seqno;
1375 
1376 	if (msg_type(hdr) == SYNCH_MSG) {
1377 		kfree_skb(skb);
1378 		return 0;
1379 	}
1380 
1381 	/* Not a fragment? */
1382 	if (likely(!msg_nof_fragms(hdr))) {
1383 		if (unlikely(!tipc_msg_extract(skb, &iskb, &ipos))) {
1384 			pr_warn_ratelimited("Unable to extract msg, defq: %d\n",
1385 					    skb_queue_len(fdefq));
1386 			return 0;
1387 		}
1388 		kfree_skb(skb);
1389 	} else {
1390 		/* Set fragment type for buf_append */
1391 		if (msg_fragm_no(hdr) == 1)
1392 			msg_set_type(hdr, FIRST_FRAGMENT);
1393 		else if (msg_fragm_no(hdr) < msg_nof_fragms(hdr))
1394 			msg_set_type(hdr, FRAGMENT);
1395 		else
1396 			msg_set_type(hdr, LAST_FRAGMENT);
1397 
1398 		if (!tipc_buf_append(reasm_tnlmsg, &skb)) {
1399 			/* Successful but non-complete reassembly? */
1400 			if (*reasm_tnlmsg || link_is_bc_rcvlink(l))
1401 				return 0;
1402 			pr_warn_ratelimited("Unable to reassemble tunnel msg\n");
1403 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1404 		}
1405 		iskb = skb;
1406 	}
1407 
1408 	do {
1409 		seqno = buf_seqno(iskb);
1410 		if (unlikely(less(seqno, l->drop_point))) {
1411 			kfree_skb(iskb);
1412 			continue;
1413 		}
1414 		if (unlikely(seqno != l->drop_point)) {
1415 			__tipc_skb_queue_sorted(fdefq, seqno, iskb);
1416 			continue;
1417 		}
1418 
1419 		l->drop_point++;
1420 		if (!tipc_data_input(l, iskb, inputq))
1421 			rc |= tipc_link_input(l, iskb, inputq, reasm_skb);
1422 		if (unlikely(rc))
1423 			break;
1424 	} while ((iskb = __tipc_skb_dequeue(fdefq, l->drop_point)));
1425 
1426 	return rc;
1427 }
1428 
1429 /**
1430  * tipc_get_gap_ack_blks - get Gap ACK blocks from PROTOCOL/STATE_MSG
1431  * @ga: returned pointer to the Gap ACK blocks if any
1432  * @l: the tipc link
1433  * @hdr: the PROTOCOL/STATE_MSG header
1434  * @uc: desired Gap ACK blocks type, i.e. unicast (= 1) or broadcast (= 0)
1435  *
1436  * Return: the total Gap ACK blocks size
1437  */
tipc_get_gap_ack_blks(struct tipc_gap_ack_blks ** ga,struct tipc_link * l,struct tipc_msg * hdr,bool uc)1438 u16 tipc_get_gap_ack_blks(struct tipc_gap_ack_blks **ga, struct tipc_link *l,
1439 			  struct tipc_msg *hdr, bool uc)
1440 {
1441 	struct tipc_gap_ack_blks *p;
1442 	u16 sz = 0;
1443 
1444 	/* Does peer support the Gap ACK blocks feature? */
1445 	if (l->peer_caps & TIPC_GAP_ACK_BLOCK) {
1446 		p = (struct tipc_gap_ack_blks *)msg_data(hdr);
1447 		sz = ntohs(p->len);
1448 		/* Sanity check */
1449 		if (sz == struct_size(p, gacks, p->ugack_cnt + p->bgack_cnt)) {
1450 			/* Good, check if the desired type exists */
1451 			if ((uc && p->ugack_cnt) || (!uc && p->bgack_cnt))
1452 				goto ok;
1453 		/* Backward compatible: peer might not support bc, but uc? */
1454 		} else if (uc && sz == struct_size(p, gacks, p->ugack_cnt)) {
1455 			if (p->ugack_cnt) {
1456 				p->bgack_cnt = 0;
1457 				goto ok;
1458 			}
1459 		}
1460 	}
1461 	/* Other cases: ignore! */
1462 	p = NULL;
1463 
1464 ok:
1465 	*ga = p;
1466 	return sz;
1467 }
1468 
__tipc_build_gap_ack_blks(struct tipc_gap_ack_blks * ga,struct tipc_link * l,u8 start_index)1469 static u8 __tipc_build_gap_ack_blks(struct tipc_gap_ack_blks *ga,
1470 				    struct tipc_link *l, u8 start_index)
1471 {
1472 	struct tipc_gap_ack *gacks = &ga->gacks[start_index];
1473 	struct sk_buff *skb = skb_peek(&l->deferdq);
1474 	u16 expect, seqno = 0;
1475 	u8 n = 0;
1476 
1477 	if (!skb)
1478 		return 0;
1479 
1480 	expect = buf_seqno(skb);
1481 	skb_queue_walk(&l->deferdq, skb) {
1482 		seqno = buf_seqno(skb);
1483 		if (unlikely(more(seqno, expect))) {
1484 			gacks[n].ack = htons(expect - 1);
1485 			gacks[n].gap = htons(seqno - expect);
1486 			if (++n >= MAX_GAP_ACK_BLKS / 2) {
1487 				pr_info_ratelimited("Gacks on %s: %d, ql: %d!\n",
1488 						    l->name, n,
1489 						    skb_queue_len(&l->deferdq));
1490 				return n;
1491 			}
1492 		} else if (unlikely(less(seqno, expect))) {
1493 			pr_warn("Unexpected skb in deferdq!\n");
1494 			continue;
1495 		}
1496 		expect = seqno + 1;
1497 	}
1498 
1499 	/* last block */
1500 	gacks[n].ack = htons(seqno);
1501 	gacks[n].gap = 0;
1502 	n++;
1503 	return n;
1504 }
1505 
1506 /* tipc_build_gap_ack_blks - build Gap ACK blocks
1507  * @l: tipc unicast link
1508  * @hdr: the tipc message buffer to store the Gap ACK blocks after built
1509  *
1510  * The function builds Gap ACK blocks for both the unicast & broadcast receiver
1511  * links of a certain peer, the buffer after built has the network data format
1512  * as found at the struct tipc_gap_ack_blks definition.
1513  *
1514  * returns the actual allocated memory size
1515  */
tipc_build_gap_ack_blks(struct tipc_link * l,struct tipc_msg * hdr)1516 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, struct tipc_msg *hdr)
1517 {
1518 	struct tipc_link *bcl = l->bc_rcvlink;
1519 	struct tipc_gap_ack_blks *ga;
1520 	u16 len;
1521 
1522 	ga = (struct tipc_gap_ack_blks *)msg_data(hdr);
1523 
1524 	/* Start with broadcast link first */
1525 	tipc_bcast_lock(bcl->net);
1526 	msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1527 	msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1528 	ga->bgack_cnt = __tipc_build_gap_ack_blks(ga, bcl, 0);
1529 	tipc_bcast_unlock(bcl->net);
1530 
1531 	/* Now for unicast link, but an explicit NACK only (???) */
1532 	ga->ugack_cnt = (msg_seq_gap(hdr)) ?
1533 			__tipc_build_gap_ack_blks(ga, l, ga->bgack_cnt) : 0;
1534 
1535 	/* Total len */
1536 	len = struct_size(ga, gacks, ga->bgack_cnt + ga->ugack_cnt);
1537 	ga->len = htons(len);
1538 	return len;
1539 }
1540 
1541 /* tipc_link_advance_transmq - advance TIPC link transmq queue by releasing
1542  *			       acked packets, also doing retransmissions if
1543  *			       gaps found
1544  * @l: tipc link with transmq queue to be advanced
1545  * @r: tipc link "receiver" i.e. in case of broadcast (= "l" if unicast)
1546  * @acked: seqno of last packet acked by peer without any gaps before
1547  * @gap: # of gap packets
1548  * @ga: buffer pointer to Gap ACK blocks from peer
1549  * @xmitq: queue for accumulating the retransmitted packets if any
1550  * @retransmitted: returned boolean value if a retransmission is really issued
1551  * @rc: returned code e.g. TIPC_LINK_DOWN_EVT if a repeated retransmit failures
1552  *      happens (- unlikely case)
1553  *
1554  * Return: the number of packets released from the link transmq
1555  */
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)1556 static int tipc_link_advance_transmq(struct tipc_link *l, struct tipc_link *r,
1557 				     u16 acked, u16 gap,
1558 				     struct tipc_gap_ack_blks *ga,
1559 				     struct sk_buff_head *xmitq,
1560 				     bool *retransmitted, int *rc)
1561 {
1562 	struct tipc_gap_ack_blks *last_ga = r->last_ga, *this_ga = NULL;
1563 	struct tipc_gap_ack *gacks = NULL;
1564 	struct sk_buff *skb, *_skb, *tmp;
1565 	struct tipc_msg *hdr;
1566 	u32 qlen = skb_queue_len(&l->transmq);
1567 	u16 nacked = acked, ngap = gap, gack_cnt = 0;
1568 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1569 	u16 ack = l->rcv_nxt - 1;
1570 	u16 seqno, n = 0;
1571 	u16 end = r->acked, start = end, offset = r->last_gap;
1572 	u16 si = (last_ga) ? last_ga->start_index : 0;
1573 	bool is_uc = !link_is_bc_sndlink(l);
1574 	bool bc_has_acked = false;
1575 
1576 	trace_tipc_link_retrans(r, acked + 1, acked + gap, &l->transmq);
1577 
1578 	/* Determine Gap ACK blocks if any for the particular link */
1579 	if (ga && is_uc) {
1580 		/* Get the Gap ACKs, uc part */
1581 		gack_cnt = ga->ugack_cnt;
1582 		gacks = &ga->gacks[ga->bgack_cnt];
1583 	} else if (ga) {
1584 		/* Copy the Gap ACKs, bc part, for later renewal if needed */
1585 		this_ga = kmemdup(ga, struct_size(ga, gacks, ga->bgack_cnt),
1586 				  GFP_ATOMIC);
1587 		if (likely(this_ga)) {
1588 			this_ga->start_index = 0;
1589 			/* Start with the bc Gap ACKs */
1590 			gack_cnt = this_ga->bgack_cnt;
1591 			gacks = &this_ga->gacks[0];
1592 		} else {
1593 			/* Hmm, we can get in trouble..., simply ignore it */
1594 			pr_warn_ratelimited("Ignoring bc Gap ACKs, no memory\n");
1595 		}
1596 	}
1597 
1598 	/* Advance the link transmq */
1599 	skb_queue_walk_safe(&l->transmq, skb, tmp) {
1600 		seqno = buf_seqno(skb);
1601 
1602 next_gap_ack:
1603 		if (less_eq(seqno, nacked)) {
1604 			if (is_uc)
1605 				goto release;
1606 			/* Skip packets peer has already acked */
1607 			if (!more(seqno, r->acked))
1608 				continue;
1609 			/* Get the next of last Gap ACK blocks */
1610 			while (more(seqno, end)) {
1611 				if (!last_ga || si >= last_ga->bgack_cnt)
1612 					break;
1613 				start = end + offset + 1;
1614 				end = ntohs(last_ga->gacks[si].ack);
1615 				offset = ntohs(last_ga->gacks[si].gap);
1616 				si++;
1617 				WARN_ONCE(more(start, end) ||
1618 					  (!offset &&
1619 					   si < last_ga->bgack_cnt) ||
1620 					  si > MAX_GAP_ACK_BLKS,
1621 					  "Corrupted Gap ACK: %d %d %d %d %d\n",
1622 					  start, end, offset, si,
1623 					  last_ga->bgack_cnt);
1624 			}
1625 			/* Check against the last Gap ACK block */
1626 			if (in_range(seqno, start, end))
1627 				continue;
1628 			/* Update/release the packet peer is acking */
1629 			bc_has_acked = true;
1630 			if (--TIPC_SKB_CB(skb)->ackers)
1631 				continue;
1632 release:
1633 			/* release skb */
1634 			__skb_unlink(skb, &l->transmq);
1635 			kfree_skb(skb);
1636 		} else if (less_eq(seqno, nacked + ngap)) {
1637 			/* First gap: check if repeated retrans failures? */
1638 			if (unlikely(seqno == acked + 1 &&
1639 				     link_retransmit_failure(l, r, rc))) {
1640 				/* Ignore this bc Gap ACKs if any */
1641 				kfree(this_ga);
1642 				this_ga = NULL;
1643 				break;
1644 			}
1645 			/* retransmit skb if unrestricted*/
1646 			if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1647 				continue;
1648 			tipc_link_set_skb_retransmit_time(skb, l);
1649 			_skb = pskb_copy(skb, GFP_ATOMIC);
1650 			if (!_skb)
1651 				continue;
1652 			hdr = buf_msg(_skb);
1653 			msg_set_ack(hdr, ack);
1654 			msg_set_bcast_ack(hdr, bc_ack);
1655 			_skb->priority = TC_PRIO_CONTROL;
1656 			__skb_queue_tail(xmitq, _skb);
1657 			l->stats.retransmitted++;
1658 			if (!is_uc)
1659 				r->stats.retransmitted++;
1660 			*retransmitted = true;
1661 			/* Increase actual retrans counter & mark first time */
1662 			if (!TIPC_SKB_CB(skb)->retr_cnt++)
1663 				TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1664 		} else {
1665 			/* retry with Gap ACK blocks if any */
1666 			if (n >= gack_cnt)
1667 				break;
1668 			nacked = ntohs(gacks[n].ack);
1669 			ngap = ntohs(gacks[n].gap);
1670 			n++;
1671 			goto next_gap_ack;
1672 		}
1673 	}
1674 
1675 	/* Renew last Gap ACK blocks for bc if needed */
1676 	if (bc_has_acked) {
1677 		if (this_ga) {
1678 			kfree(last_ga);
1679 			r->last_ga = this_ga;
1680 			r->last_gap = gap;
1681 		} else if (last_ga) {
1682 			if (less(acked, start)) {
1683 				si--;
1684 				offset = start - acked - 1;
1685 			} else if (less(acked, end)) {
1686 				acked = end;
1687 			}
1688 			if (si < last_ga->bgack_cnt) {
1689 				last_ga->start_index = si;
1690 				r->last_gap = offset;
1691 			} else {
1692 				kfree(last_ga);
1693 				r->last_ga = NULL;
1694 				r->last_gap = 0;
1695 			}
1696 		} else {
1697 			r->last_gap = 0;
1698 		}
1699 		r->acked = acked;
1700 	} else {
1701 		kfree(this_ga);
1702 	}
1703 
1704 	return qlen - skb_queue_len(&l->transmq);
1705 }
1706 
1707 /* tipc_link_build_state_msg: prepare link state message for transmission
1708  *
1709  * Note that sending of broadcast ack is coordinated among nodes, to reduce
1710  * risk of ack storms towards the sender
1711  */
tipc_link_build_state_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1712 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1713 {
1714 	if (!l)
1715 		return 0;
1716 
1717 	/* Broadcast ACK must be sent via a unicast link => defer to caller */
1718 	if (link_is_bc_rcvlink(l)) {
1719 		if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1720 			return 0;
1721 		l->rcv_unacked = 0;
1722 
1723 		/* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1724 		l->snd_nxt = l->rcv_nxt;
1725 		return TIPC_LINK_SND_STATE;
1726 	}
1727 	/* Unicast ACK */
1728 	l->rcv_unacked = 0;
1729 	l->stats.sent_acks++;
1730 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1731 	return 0;
1732 }
1733 
1734 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1735  */
tipc_link_build_reset_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1736 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1737 {
1738 	int mtyp = RESET_MSG;
1739 	struct sk_buff *skb;
1740 
1741 	if (l->state == LINK_ESTABLISHING)
1742 		mtyp = ACTIVATE_MSG;
1743 
1744 	tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1745 
1746 	/* Inform peer that this endpoint is going down if applicable */
1747 	skb = skb_peek_tail(xmitq);
1748 	if (skb && (l->state == LINK_RESET))
1749 		msg_set_peer_stopping(buf_msg(skb), 1);
1750 }
1751 
1752 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1753  * Note that sending of broadcast NACK is coordinated among nodes, to
1754  * reduce the risk of NACK storms towards the sender
1755  */
tipc_link_build_nack_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1756 static int tipc_link_build_nack_msg(struct tipc_link *l,
1757 				    struct sk_buff_head *xmitq)
1758 {
1759 	u32 def_cnt = ++l->stats.deferred_recv;
1760 	struct sk_buff_head *dfq = &l->deferdq;
1761 	u32 defq_len = skb_queue_len(dfq);
1762 	int match1, match2;
1763 
1764 	if (link_is_bc_rcvlink(l)) {
1765 		match1 = def_cnt & 0xf;
1766 		match2 = tipc_own_addr(l->net) & 0xf;
1767 		if (match1 == match2)
1768 			return TIPC_LINK_SND_STATE;
1769 		return 0;
1770 	}
1771 
1772 	if (defq_len >= 3 && !((defq_len - 3) % 16)) {
1773 		u16 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1774 
1775 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0,
1776 					  rcvgap, 0, 0, xmitq);
1777 	}
1778 	return 0;
1779 }
1780 
1781 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1782  * @l: the link that should handle the message
1783  * @skb: TIPC packet
1784  * @xmitq: queue to place packets to be sent after this call
1785  */
tipc_link_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * xmitq)1786 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1787 		  struct sk_buff_head *xmitq)
1788 {
1789 	struct sk_buff_head *defq = &l->deferdq;
1790 	struct tipc_msg *hdr = buf_msg(skb);
1791 	u16 seqno, rcv_nxt, win_lim;
1792 	int released = 0;
1793 	int rc = 0;
1794 
1795 	/* Verify and update link state */
1796 	if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1797 		return tipc_link_proto_rcv(l, skb, xmitq);
1798 
1799 	/* Don't send probe at next timeout expiration */
1800 	l->silent_intv_cnt = 0;
1801 
1802 	do {
1803 		hdr = buf_msg(skb);
1804 		seqno = msg_seqno(hdr);
1805 		rcv_nxt = l->rcv_nxt;
1806 		win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1807 
1808 		if (unlikely(!link_is_up(l))) {
1809 			if (l->state == LINK_ESTABLISHING)
1810 				rc = TIPC_LINK_UP_EVT;
1811 			kfree_skb(skb);
1812 			break;
1813 		}
1814 
1815 		/* Drop if outside receive window */
1816 		if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1817 			l->stats.duplicates++;
1818 			kfree_skb(skb);
1819 			break;
1820 		}
1821 		released += tipc_link_advance_transmq(l, l, msg_ack(hdr), 0,
1822 						      NULL, NULL, NULL, NULL);
1823 
1824 		/* Defer delivery if sequence gap */
1825 		if (unlikely(seqno != rcv_nxt)) {
1826 			if (!__tipc_skb_queue_sorted(defq, seqno, skb))
1827 				l->stats.duplicates++;
1828 			rc |= tipc_link_build_nack_msg(l, xmitq);
1829 			break;
1830 		}
1831 
1832 		/* Deliver packet */
1833 		l->rcv_nxt++;
1834 		l->stats.recv_pkts++;
1835 
1836 		if (unlikely(msg_user(hdr) == TUNNEL_PROTOCOL))
1837 			rc |= tipc_link_tnl_rcv(l, skb, l->inputq);
1838 		else if (!tipc_data_input(l, skb, l->inputq))
1839 			rc |= tipc_link_input(l, skb, l->inputq, &l->reasm_buf);
1840 		if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1841 			rc |= tipc_link_build_state_msg(l, xmitq);
1842 		if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1843 			break;
1844 	} while ((skb = __tipc_skb_dequeue(defq, l->rcv_nxt)));
1845 
1846 	/* Forward queues and wake up waiting users */
1847 	if (released) {
1848 		tipc_link_update_cwin(l, released, 0);
1849 		tipc_link_advance_backlog(l, xmitq);
1850 		if (unlikely(!skb_queue_empty(&l->wakeupq)))
1851 			link_prepare_wakeup(l);
1852 	}
1853 	return rc;
1854 }
1855 
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)1856 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1857 				      bool probe_reply, u16 rcvgap,
1858 				      int tolerance, int priority,
1859 				      struct sk_buff_head *xmitq)
1860 {
1861 	struct tipc_mon_state *mstate = &l->mon_state;
1862 	struct sk_buff_head *dfq = &l->deferdq;
1863 	struct tipc_link *bcl = l->bc_rcvlink;
1864 	struct tipc_msg *hdr;
1865 	struct sk_buff *skb;
1866 	bool node_up = link_is_up(bcl);
1867 	u16 glen = 0, bc_rcvgap = 0;
1868 	int dlen = 0;
1869 	void *data;
1870 
1871 	/* Don't send protocol message during reset or link failover */
1872 	if (tipc_link_is_blocked(l))
1873 		return;
1874 
1875 	if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1876 		return;
1877 
1878 	if ((probe || probe_reply) && !skb_queue_empty(dfq))
1879 		rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1880 
1881 	skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1882 			      tipc_max_domain_size + MAX_GAP_ACK_BLKS_SZ,
1883 			      l->addr, tipc_own_addr(l->net), 0, 0, 0);
1884 	if (!skb)
1885 		return;
1886 
1887 	hdr = buf_msg(skb);
1888 	data = msg_data(hdr);
1889 	msg_set_session(hdr, l->session);
1890 	msg_set_bearer_id(hdr, l->bearer_id);
1891 	msg_set_net_plane(hdr, l->net_plane);
1892 	msg_set_next_sent(hdr, l->snd_nxt);
1893 	msg_set_ack(hdr, l->rcv_nxt - 1);
1894 	msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1895 	msg_set_bc_ack_invalid(hdr, !node_up);
1896 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1897 	msg_set_link_tolerance(hdr, tolerance);
1898 	msg_set_linkprio(hdr, priority);
1899 	msg_set_redundant_link(hdr, node_up);
1900 	msg_set_seq_gap(hdr, 0);
1901 	msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1902 
1903 	if (mtyp == STATE_MSG) {
1904 		if (l->peer_caps & TIPC_LINK_PROTO_SEQNO)
1905 			msg_set_seqno(hdr, l->snd_nxt_state++);
1906 		msg_set_seq_gap(hdr, rcvgap);
1907 		bc_rcvgap = link_bc_rcv_gap(bcl);
1908 		msg_set_bc_gap(hdr, bc_rcvgap);
1909 		msg_set_probe(hdr, probe);
1910 		msg_set_is_keepalive(hdr, probe || probe_reply);
1911 		if (l->peer_caps & TIPC_GAP_ACK_BLOCK)
1912 			glen = tipc_build_gap_ack_blks(l, hdr);
1913 		tipc_mon_prep(l->net, data + glen, &dlen, mstate, l->bearer_id);
1914 		msg_set_size(hdr, INT_H_SIZE + glen + dlen);
1915 		skb_trim(skb, INT_H_SIZE + glen + dlen);
1916 		l->stats.sent_states++;
1917 		l->rcv_unacked = 0;
1918 	} else {
1919 		/* RESET_MSG or ACTIVATE_MSG */
1920 		if (mtyp == ACTIVATE_MSG) {
1921 			msg_set_dest_session_valid(hdr, 1);
1922 			msg_set_dest_session(hdr, l->peer_session);
1923 		}
1924 		msg_set_max_pkt(hdr, l->advertised_mtu);
1925 		strcpy(data, l->if_name);
1926 		msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1927 		skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1928 	}
1929 	if (probe)
1930 		l->stats.sent_probes++;
1931 	if (rcvgap)
1932 		l->stats.sent_nacks++;
1933 	if (bc_rcvgap)
1934 		bcl->stats.sent_nacks++;
1935 	skb->priority = TC_PRIO_CONTROL;
1936 	__skb_queue_tail(xmitq, skb);
1937 	trace_tipc_proto_build(skb, false, l->name);
1938 }
1939 
tipc_link_create_dummy_tnl_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1940 void tipc_link_create_dummy_tnl_msg(struct tipc_link *l,
1941 				    struct sk_buff_head *xmitq)
1942 {
1943 	u32 onode = tipc_own_addr(l->net);
1944 	struct tipc_msg *hdr, *ihdr;
1945 	struct sk_buff_head tnlq;
1946 	struct sk_buff *skb;
1947 	u32 dnode = l->addr;
1948 
1949 	__skb_queue_head_init(&tnlq);
1950 	skb = tipc_msg_create(TUNNEL_PROTOCOL, FAILOVER_MSG,
1951 			      INT_H_SIZE, BASIC_H_SIZE,
1952 			      dnode, onode, 0, 0, 0);
1953 	if (!skb) {
1954 		pr_warn("%sunable to create tunnel packet\n", link_co_err);
1955 		return;
1956 	}
1957 
1958 	hdr = buf_msg(skb);
1959 	msg_set_msgcnt(hdr, 1);
1960 	msg_set_bearer_id(hdr, l->peer_bearer_id);
1961 
1962 	ihdr = (struct tipc_msg *)msg_data(hdr);
1963 	tipc_msg_init(onode, ihdr, TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1964 		      BASIC_H_SIZE, dnode);
1965 	msg_set_errcode(ihdr, TIPC_ERR_NO_PORT);
1966 	__skb_queue_tail(&tnlq, skb);
1967 	tipc_link_xmit(l, &tnlq, xmitq);
1968 }
1969 
1970 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1971  * with contents of the link's transmit and backlog queues.
1972  */
tipc_link_tnl_prepare(struct tipc_link * l,struct tipc_link * tnl,int mtyp,struct sk_buff_head * xmitq)1973 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1974 			   int mtyp, struct sk_buff_head *xmitq)
1975 {
1976 	struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1977 	struct sk_buff *skb, *tnlskb;
1978 	struct tipc_msg *hdr, tnlhdr;
1979 	struct sk_buff_head *queue = &l->transmq;
1980 	struct sk_buff_head tmpxq, tnlq, frags;
1981 	u16 pktlen, pktcnt, seqno = l->snd_nxt;
1982 	bool pktcnt_need_update = false;
1983 	u16 syncpt;
1984 	int rc;
1985 
1986 	if (!tnl)
1987 		return;
1988 
1989 	__skb_queue_head_init(&tnlq);
1990 	/* Link Synching:
1991 	 * From now on, send only one single ("dummy") SYNCH message
1992 	 * to peer. The SYNCH message does not contain any data, just
1993 	 * a header conveying the synch point to the peer.
1994 	 */
1995 	if (mtyp == SYNCH_MSG && (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1996 		tnlskb = tipc_msg_create(TUNNEL_PROTOCOL, SYNCH_MSG,
1997 					 INT_H_SIZE, 0, l->addr,
1998 					 tipc_own_addr(l->net),
1999 					 0, 0, 0);
2000 		if (!tnlskb) {
2001 			pr_warn("%sunable to create dummy SYNCH_MSG\n",
2002 				link_co_err);
2003 			return;
2004 		}
2005 
2006 		hdr = buf_msg(tnlskb);
2007 		syncpt = l->snd_nxt + skb_queue_len(&l->backlogq) - 1;
2008 		msg_set_syncpt(hdr, syncpt);
2009 		msg_set_bearer_id(hdr, l->peer_bearer_id);
2010 		__skb_queue_tail(&tnlq, tnlskb);
2011 		tipc_link_xmit(tnl, &tnlq, xmitq);
2012 		return;
2013 	}
2014 
2015 	__skb_queue_head_init(&tmpxq);
2016 	__skb_queue_head_init(&frags);
2017 	/* At least one packet required for safe algorithm => add dummy */
2018 	skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
2019 			      BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
2020 			      0, 0, TIPC_ERR_NO_PORT);
2021 	if (!skb) {
2022 		pr_warn("%sunable to create tunnel packet\n", link_co_err);
2023 		return;
2024 	}
2025 	__skb_queue_tail(&tnlq, skb);
2026 	tipc_link_xmit(l, &tnlq, &tmpxq);
2027 	__skb_queue_purge(&tmpxq);
2028 
2029 	/* Initialize reusable tunnel packet header */
2030 	tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
2031 		      mtyp, INT_H_SIZE, l->addr);
2032 	if (mtyp == SYNCH_MSG)
2033 		pktcnt = l->snd_nxt - buf_seqno(skb_peek(&l->transmq));
2034 	else
2035 		pktcnt = skb_queue_len(&l->transmq);
2036 	pktcnt += skb_queue_len(&l->backlogq);
2037 	msg_set_msgcnt(&tnlhdr, pktcnt);
2038 	msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
2039 tnl:
2040 	/* Wrap each packet into a tunnel packet */
2041 	skb_queue_walk(queue, skb) {
2042 		hdr = buf_msg(skb);
2043 		if (queue == &l->backlogq)
2044 			msg_set_seqno(hdr, seqno++);
2045 		pktlen = msg_size(hdr);
2046 
2047 		/* Tunnel link MTU is not large enough? This could be
2048 		 * due to:
2049 		 * 1) Link MTU has just changed or set differently;
2050 		 * 2) Or FAILOVER on the top of a SYNCH message
2051 		 *
2052 		 * The 2nd case should not happen if peer supports
2053 		 * TIPC_TUNNEL_ENHANCED
2054 		 */
2055 		if (pktlen > tnl->mtu - INT_H_SIZE) {
2056 			if (mtyp == FAILOVER_MSG &&
2057 			    (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
2058 				rc = tipc_msg_fragment(skb, &tnlhdr, tnl->mtu,
2059 						       &frags);
2060 				if (rc) {
2061 					pr_warn("%sunable to frag msg: rc %d\n",
2062 						link_co_err, rc);
2063 					return;
2064 				}
2065 				pktcnt += skb_queue_len(&frags) - 1;
2066 				pktcnt_need_update = true;
2067 				skb_queue_splice_tail_init(&frags, &tnlq);
2068 				continue;
2069 			}
2070 			/* Unluckily, peer doesn't have TIPC_TUNNEL_ENHANCED
2071 			 * => Just warn it and return!
2072 			 */
2073 			pr_warn_ratelimited("%stoo large msg <%d, %d>: %d!\n",
2074 					    link_co_err, msg_user(hdr),
2075 					    msg_type(hdr), msg_size(hdr));
2076 			return;
2077 		}
2078 
2079 		msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
2080 		tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
2081 		if (!tnlskb) {
2082 			pr_warn("%sunable to send packet\n", link_co_err);
2083 			return;
2084 		}
2085 		skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
2086 		skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
2087 		__skb_queue_tail(&tnlq, tnlskb);
2088 	}
2089 	if (queue != &l->backlogq) {
2090 		queue = &l->backlogq;
2091 		goto tnl;
2092 	}
2093 
2094 	if (pktcnt_need_update)
2095 		skb_queue_walk(&tnlq, skb) {
2096 			hdr = buf_msg(skb);
2097 			msg_set_msgcnt(hdr, pktcnt);
2098 		}
2099 
2100 	tipc_link_xmit(tnl, &tnlq, xmitq);
2101 
2102 	if (mtyp == FAILOVER_MSG) {
2103 		tnl->drop_point = l->rcv_nxt;
2104 		tnl->failover_reasm_skb = l->reasm_buf;
2105 		l->reasm_buf = NULL;
2106 
2107 		/* Failover the link's deferdq */
2108 		if (unlikely(!skb_queue_empty(fdefq))) {
2109 			pr_warn("Link failover deferdq not empty: %d!\n",
2110 				skb_queue_len(fdefq));
2111 			__skb_queue_purge(fdefq);
2112 		}
2113 		skb_queue_splice_init(&l->deferdq, fdefq);
2114 	}
2115 }
2116 
2117 /**
2118  * tipc_link_failover_prepare() - prepare tnl for link failover
2119  *
2120  * This is a special version of the precursor - tipc_link_tnl_prepare(),
2121  * see the tipc_node_link_failover() for details
2122  *
2123  * @l: failover link
2124  * @tnl: tunnel link
2125  * @xmitq: queue for messages to be xmited
2126  */
tipc_link_failover_prepare(struct tipc_link * l,struct tipc_link * tnl,struct sk_buff_head * xmitq)2127 void tipc_link_failover_prepare(struct tipc_link *l, struct tipc_link *tnl,
2128 				struct sk_buff_head *xmitq)
2129 {
2130 	struct sk_buff_head *fdefq = &tnl->failover_deferdq;
2131 
2132 	tipc_link_create_dummy_tnl_msg(tnl, xmitq);
2133 
2134 	/* This failover link endpoint was never established before,
2135 	 * so it has not received anything from peer.
2136 	 * Otherwise, it must be a normal failover situation or the
2137 	 * node has entered SELF_DOWN_PEER_LEAVING and both peer nodes
2138 	 * would have to start over from scratch instead.
2139 	 */
2140 	tnl->drop_point = 1;
2141 	tnl->failover_reasm_skb = NULL;
2142 
2143 	/* Initiate the link's failover deferdq */
2144 	if (unlikely(!skb_queue_empty(fdefq))) {
2145 		pr_warn("Link failover deferdq not empty: %d!\n",
2146 			skb_queue_len(fdefq));
2147 		__skb_queue_purge(fdefq);
2148 	}
2149 }
2150 
2151 /* tipc_link_validate_msg(): validate message against current link state
2152  * Returns true if message should be accepted, otherwise false
2153  */
tipc_link_validate_msg(struct tipc_link * l,struct tipc_msg * hdr)2154 bool tipc_link_validate_msg(struct tipc_link *l, struct tipc_msg *hdr)
2155 {
2156 	u16 curr_session = l->peer_session;
2157 	u16 session = msg_session(hdr);
2158 	int mtyp = msg_type(hdr);
2159 
2160 	if (msg_user(hdr) != LINK_PROTOCOL)
2161 		return true;
2162 
2163 	switch (mtyp) {
2164 	case RESET_MSG:
2165 		if (!l->in_session)
2166 			return true;
2167 		/* Accept only RESET with new session number */
2168 		return more(session, curr_session);
2169 	case ACTIVATE_MSG:
2170 		if (!l->in_session)
2171 			return true;
2172 		/* Accept only ACTIVATE with new or current session number */
2173 		return !less(session, curr_session);
2174 	case STATE_MSG:
2175 		/* Accept only STATE with current session number */
2176 		if (!l->in_session)
2177 			return false;
2178 		if (session != curr_session)
2179 			return false;
2180 		/* Extra sanity check */
2181 		if (!link_is_up(l) && msg_ack(hdr))
2182 			return false;
2183 		if (!(l->peer_caps & TIPC_LINK_PROTO_SEQNO))
2184 			return true;
2185 		/* Accept only STATE with new sequence number */
2186 		return !less(msg_seqno(hdr), l->rcv_nxt_state);
2187 	default:
2188 		return false;
2189 	}
2190 }
2191 
2192 /* tipc_link_proto_rcv(): receive link level protocol message :
2193  * Note that network plane id propagates through the network, and may
2194  * change at any time. The node with lowest numerical id determines
2195  * network plane
2196  */
tipc_link_proto_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * xmitq)2197 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
2198 			       struct sk_buff_head *xmitq)
2199 {
2200 	struct tipc_msg *hdr = buf_msg(skb);
2201 	struct tipc_gap_ack_blks *ga = NULL;
2202 	bool reply = msg_probe(hdr), retransmitted = false;
2203 	u32 dlen = msg_data_sz(hdr), glen = 0;
2204 	u16 peers_snd_nxt =  msg_next_sent(hdr);
2205 	u16 peers_tol = msg_link_tolerance(hdr);
2206 	u16 peers_prio = msg_linkprio(hdr);
2207 	u16 gap = msg_seq_gap(hdr);
2208 	u16 ack = msg_ack(hdr);
2209 	u16 rcv_nxt = l->rcv_nxt;
2210 	u16 rcvgap = 0;
2211 	int mtyp = msg_type(hdr);
2212 	int rc = 0, released;
2213 	char *if_name;
2214 	void *data;
2215 
2216 	trace_tipc_proto_rcv(skb, false, l->name);
2217 
2218 	if (dlen > U16_MAX)
2219 		goto exit;
2220 
2221 	if (tipc_link_is_blocked(l) || !xmitq)
2222 		goto exit;
2223 
2224 	if (tipc_own_addr(l->net) > msg_prevnode(hdr))
2225 		l->net_plane = msg_net_plane(hdr);
2226 
2227 	skb_linearize(skb);
2228 	hdr = buf_msg(skb);
2229 	data = msg_data(hdr);
2230 
2231 	if (!tipc_link_validate_msg(l, hdr)) {
2232 		trace_tipc_skb_dump(skb, false, "PROTO invalid (1)!");
2233 		trace_tipc_link_dump(l, TIPC_DUMP_NONE, "PROTO invalid (1)!");
2234 		goto exit;
2235 	}
2236 
2237 	switch (mtyp) {
2238 	case RESET_MSG:
2239 	case ACTIVATE_MSG:
2240 		/* Complete own link name with peer's interface name */
2241 		if_name =  strrchr(l->name, ':') + 1;
2242 		if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
2243 			break;
2244 		if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
2245 			break;
2246 		strncpy(if_name, data, TIPC_MAX_IF_NAME);
2247 
2248 		/* Update own tolerance if peer indicates a non-zero value */
2249 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2250 			l->tolerance = peers_tol;
2251 			l->bc_rcvlink->tolerance = peers_tol;
2252 		}
2253 		/* Update own priority if peer's priority is higher */
2254 		if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
2255 			l->priority = peers_prio;
2256 
2257 		/* If peer is going down we want full re-establish cycle */
2258 		if (msg_peer_stopping(hdr)) {
2259 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2260 			break;
2261 		}
2262 
2263 		/* If this endpoint was re-created while peer was ESTABLISHING
2264 		 * it doesn't know current session number. Force re-synch.
2265 		 */
2266 		if (mtyp == ACTIVATE_MSG && msg_dest_session_valid(hdr) &&
2267 		    l->session != msg_dest_session(hdr)) {
2268 			if (less(l->session, msg_dest_session(hdr)))
2269 				l->session = msg_dest_session(hdr) + 1;
2270 			break;
2271 		}
2272 
2273 		/* ACTIVATE_MSG serves as PEER_RESET if link is already down */
2274 		if (mtyp == RESET_MSG || !link_is_up(l))
2275 			rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
2276 
2277 		/* ACTIVATE_MSG takes up link if it was already locally reset */
2278 		if (mtyp == ACTIVATE_MSG && l->state == LINK_ESTABLISHING)
2279 			rc = TIPC_LINK_UP_EVT;
2280 
2281 		l->peer_session = msg_session(hdr);
2282 		l->in_session = true;
2283 		l->peer_bearer_id = msg_bearer_id(hdr);
2284 		if (l->mtu > msg_max_pkt(hdr))
2285 			l->mtu = msg_max_pkt(hdr);
2286 		break;
2287 
2288 	case STATE_MSG:
2289 		/* Validate Gap ACK blocks, drop if invalid */
2290 		glen = tipc_get_gap_ack_blks(&ga, l, hdr, true);
2291 		if (glen > dlen)
2292 			break;
2293 
2294 		l->rcv_nxt_state = msg_seqno(hdr) + 1;
2295 
2296 		/* Update own tolerance if peer indicates a non-zero value */
2297 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2298 			l->tolerance = peers_tol;
2299 			l->bc_rcvlink->tolerance = peers_tol;
2300 		}
2301 		/* Update own prio if peer indicates a different value */
2302 		if ((peers_prio != l->priority) &&
2303 		    in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
2304 			l->priority = peers_prio;
2305 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2306 		}
2307 
2308 		l->silent_intv_cnt = 0;
2309 		l->stats.recv_states++;
2310 		if (msg_probe(hdr))
2311 			l->stats.recv_probes++;
2312 
2313 		if (!link_is_up(l)) {
2314 			if (l->state == LINK_ESTABLISHING)
2315 				rc = TIPC_LINK_UP_EVT;
2316 			break;
2317 		}
2318 
2319 		tipc_mon_rcv(l->net, data + glen, dlen - glen, l->addr,
2320 			     &l->mon_state, l->bearer_id);
2321 
2322 		/* Send NACK if peer has sent pkts we haven't received yet */
2323 		if ((reply || msg_is_keepalive(hdr)) &&
2324 		    more(peers_snd_nxt, rcv_nxt) &&
2325 		    !tipc_link_is_synching(l) &&
2326 		    skb_queue_empty(&l->deferdq))
2327 			rcvgap = peers_snd_nxt - l->rcv_nxt;
2328 		if (rcvgap || reply)
2329 			tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
2330 						  rcvgap, 0, 0, xmitq);
2331 
2332 		released = tipc_link_advance_transmq(l, l, ack, gap, ga, xmitq,
2333 						     &retransmitted, &rc);
2334 		if (gap)
2335 			l->stats.recv_nacks++;
2336 		if (released || retransmitted)
2337 			tipc_link_update_cwin(l, released, retransmitted);
2338 		if (released)
2339 			tipc_link_advance_backlog(l, xmitq);
2340 		if (unlikely(!skb_queue_empty(&l->wakeupq)))
2341 			link_prepare_wakeup(l);
2342 	}
2343 exit:
2344 	kfree_skb(skb);
2345 	return rc;
2346 }
2347 
2348 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
2349  */
tipc_link_build_bc_proto_msg(struct tipc_link * l,bool bcast,u16 peers_snd_nxt,struct sk_buff_head * xmitq)2350 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
2351 					 u16 peers_snd_nxt,
2352 					 struct sk_buff_head *xmitq)
2353 {
2354 	struct sk_buff *skb;
2355 	struct tipc_msg *hdr;
2356 	struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
2357 	u16 ack = l->rcv_nxt - 1;
2358 	u16 gap_to = peers_snd_nxt - 1;
2359 
2360 	skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
2361 			      0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
2362 	if (!skb)
2363 		return false;
2364 	hdr = buf_msg(skb);
2365 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
2366 	msg_set_bcast_ack(hdr, ack);
2367 	msg_set_bcgap_after(hdr, ack);
2368 	if (dfrd_skb)
2369 		gap_to = buf_seqno(dfrd_skb) - 1;
2370 	msg_set_bcgap_to(hdr, gap_to);
2371 	msg_set_non_seq(hdr, bcast);
2372 	__skb_queue_tail(xmitq, skb);
2373 	return true;
2374 }
2375 
2376 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
2377  *
2378  * Give a newly added peer node the sequence number where it should
2379  * start receiving and acking broadcast packets.
2380  */
tipc_link_build_bc_init_msg(struct tipc_link * l,struct sk_buff_head * xmitq)2381 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
2382 					struct sk_buff_head *xmitq)
2383 {
2384 	struct sk_buff_head list;
2385 
2386 	__skb_queue_head_init(&list);
2387 	if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
2388 		return;
2389 	msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
2390 	tipc_link_xmit(l, &list, xmitq);
2391 }
2392 
2393 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
2394  */
tipc_link_bc_init_rcv(struct tipc_link * l,struct tipc_msg * hdr)2395 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
2396 {
2397 	int mtyp = msg_type(hdr);
2398 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2399 
2400 	if (link_is_up(l))
2401 		return;
2402 
2403 	if (msg_user(hdr) == BCAST_PROTOCOL) {
2404 		l->rcv_nxt = peers_snd_nxt;
2405 		l->state = LINK_ESTABLISHED;
2406 		return;
2407 	}
2408 
2409 	if (l->peer_caps & TIPC_BCAST_SYNCH)
2410 		return;
2411 
2412 	if (msg_peer_node_is_up(hdr))
2413 		return;
2414 
2415 	/* Compatibility: accept older, less safe initial synch data */
2416 	if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
2417 		l->rcv_nxt = peers_snd_nxt;
2418 }
2419 
2420 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
2421  */
tipc_link_bc_sync_rcv(struct tipc_link * l,struct tipc_msg * hdr,struct sk_buff_head * xmitq)2422 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
2423 			  struct sk_buff_head *xmitq)
2424 {
2425 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2426 	int rc = 0;
2427 
2428 	if (!link_is_up(l))
2429 		return rc;
2430 
2431 	if (!msg_peer_node_is_up(hdr))
2432 		return rc;
2433 
2434 	/* Open when peer acknowledges our bcast init msg (pkt #1) */
2435 	if (msg_ack(hdr))
2436 		l->bc_peer_is_up = true;
2437 
2438 	if (!l->bc_peer_is_up)
2439 		return rc;
2440 
2441 	/* Ignore if peers_snd_nxt goes beyond receive window */
2442 	if (more(peers_snd_nxt, l->rcv_nxt + l->window))
2443 		return rc;
2444 
2445 	l->snd_nxt = peers_snd_nxt;
2446 	if (link_bc_rcv_gap(l))
2447 		rc |= TIPC_LINK_SND_STATE;
2448 
2449 	/* Return now if sender supports nack via STATE messages */
2450 	if (l->peer_caps & TIPC_BCAST_STATE_NACK)
2451 		return rc;
2452 
2453 	/* Otherwise, be backwards compatible */
2454 
2455 	if (!more(peers_snd_nxt, l->rcv_nxt)) {
2456 		l->nack_state = BC_NACK_SND_CONDITIONAL;
2457 		return 0;
2458 	}
2459 
2460 	/* Don't NACK if one was recently sent or peeked */
2461 	if (l->nack_state == BC_NACK_SND_SUPPRESS) {
2462 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2463 		return 0;
2464 	}
2465 
2466 	/* Conditionally delay NACK sending until next synch rcv */
2467 	if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
2468 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2469 		if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
2470 			return 0;
2471 	}
2472 
2473 	/* Send NACK now but suppress next one */
2474 	tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
2475 	l->nack_state = BC_NACK_SND_SUPPRESS;
2476 	return 0;
2477 }
2478 
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)2479 int tipc_link_bc_ack_rcv(struct tipc_link *r, u16 acked, u16 gap,
2480 			 struct tipc_gap_ack_blks *ga,
2481 			 struct sk_buff_head *xmitq,
2482 			 struct sk_buff_head *retrq)
2483 {
2484 	struct tipc_link *l = r->bc_sndlink;
2485 	bool unused = false;
2486 	int rc = 0;
2487 
2488 	if (!link_is_up(r) || !r->bc_peer_is_up)
2489 		return 0;
2490 
2491 	if (gap) {
2492 		l->stats.recv_nacks++;
2493 		r->stats.recv_nacks++;
2494 	}
2495 
2496 	if (less(acked, r->acked) || (acked == r->acked && !gap && !ga))
2497 		return 0;
2498 
2499 	trace_tipc_link_bc_ack(r, acked, gap, &l->transmq);
2500 	tipc_link_advance_transmq(l, r, acked, gap, ga, retrq, &unused, &rc);
2501 
2502 	tipc_link_advance_backlog(l, xmitq);
2503 	if (unlikely(!skb_queue_empty(&l->wakeupq)))
2504 		link_prepare_wakeup(l);
2505 
2506 	return rc;
2507 }
2508 
2509 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
2510  * This function is here for backwards compatibility, since
2511  * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
2512  */
tipc_link_bc_nack_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * xmitq)2513 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
2514 			  struct sk_buff_head *xmitq)
2515 {
2516 	struct tipc_msg *hdr = buf_msg(skb);
2517 	u32 dnode = msg_destnode(hdr);
2518 	int mtyp = msg_type(hdr);
2519 	u16 acked = msg_bcast_ack(hdr);
2520 	u16 from = acked + 1;
2521 	u16 to = msg_bcgap_to(hdr);
2522 	u16 peers_snd_nxt = to + 1;
2523 	int rc = 0;
2524 
2525 	kfree_skb(skb);
2526 
2527 	if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
2528 		return 0;
2529 
2530 	if (mtyp != STATE_MSG)
2531 		return 0;
2532 
2533 	if (dnode == tipc_own_addr(l->net)) {
2534 		rc = tipc_link_bc_ack_rcv(l, acked, to - acked, NULL, xmitq,
2535 					  xmitq);
2536 		l->stats.recv_nacks++;
2537 		return rc;
2538 	}
2539 
2540 	/* Msg for other node => suppress own NACK at next sync if applicable */
2541 	if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
2542 		l->nack_state = BC_NACK_SND_SUPPRESS;
2543 
2544 	return 0;
2545 }
2546 
tipc_link_set_queue_limits(struct tipc_link * l,u32 min_win,u32 max_win)2547 void tipc_link_set_queue_limits(struct tipc_link *l, u32 min_win, u32 max_win)
2548 {
2549 	int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
2550 
2551 	l->min_win = min_win;
2552 	l->ssthresh = max_win;
2553 	l->max_win = max_win;
2554 	l->window = min_win;
2555 	l->backlog[TIPC_LOW_IMPORTANCE].limit      = min_win * 2;
2556 	l->backlog[TIPC_MEDIUM_IMPORTANCE].limit   = min_win * 4;
2557 	l->backlog[TIPC_HIGH_IMPORTANCE].limit     = min_win * 6;
2558 	l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = min_win * 8;
2559 	l->backlog[TIPC_SYSTEM_IMPORTANCE].limit   = max_bulk;
2560 }
2561 
2562 /**
2563  * tipc_link_reset_stats - reset link statistics
2564  * @l: pointer to link
2565  */
tipc_link_reset_stats(struct tipc_link * l)2566 void tipc_link_reset_stats(struct tipc_link *l)
2567 {
2568 	memset(&l->stats, 0, sizeof(l->stats));
2569 }
2570 
link_print(struct tipc_link * l,const char * str)2571 static void link_print(struct tipc_link *l, const char *str)
2572 {
2573 	struct sk_buff *hskb = skb_peek(&l->transmq);
2574 	u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
2575 	u16 tail = l->snd_nxt - 1;
2576 
2577 	pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
2578 	pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
2579 		skb_queue_len(&l->transmq), head, tail,
2580 		skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
2581 }
2582 
2583 /* Parse and validate nested (link) properties valid for media, bearer and link
2584  */
tipc_nl_parse_link_prop(struct nlattr * prop,struct nlattr * props[])2585 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
2586 {
2587 	int err;
2588 
2589 	err = nla_parse_nested_deprecated(props, TIPC_NLA_PROP_MAX, prop,
2590 					  tipc_nl_prop_policy, NULL);
2591 	if (err)
2592 		return err;
2593 
2594 	if (props[TIPC_NLA_PROP_PRIO]) {
2595 		u32 prio;
2596 
2597 		prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
2598 		if (prio > TIPC_MAX_LINK_PRI)
2599 			return -EINVAL;
2600 	}
2601 
2602 	if (props[TIPC_NLA_PROP_TOL]) {
2603 		u32 tol;
2604 
2605 		tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
2606 		if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
2607 			return -EINVAL;
2608 	}
2609 
2610 	if (props[TIPC_NLA_PROP_WIN]) {
2611 		u32 max_win;
2612 
2613 		max_win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
2614 		if (max_win < TIPC_DEF_LINK_WIN || max_win > TIPC_MAX_LINK_WIN)
2615 			return -EINVAL;
2616 	}
2617 
2618 	return 0;
2619 }
2620 
__tipc_nl_add_stats(struct sk_buff * skb,struct tipc_stats * s)2621 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
2622 {
2623 	int i;
2624 	struct nlattr *stats;
2625 
2626 	struct nla_map {
2627 		u32 key;
2628 		u32 val;
2629 	};
2630 
2631 	struct nla_map map[] = {
2632 		{TIPC_NLA_STATS_RX_INFO, 0},
2633 		{TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
2634 		{TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
2635 		{TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
2636 		{TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
2637 		{TIPC_NLA_STATS_TX_INFO, 0},
2638 		{TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
2639 		{TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
2640 		{TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
2641 		{TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
2642 		{TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
2643 			s->msg_length_counts : 1},
2644 		{TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
2645 		{TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
2646 		{TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
2647 		{TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
2648 		{TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
2649 		{TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
2650 		{TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
2651 		{TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
2652 		{TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
2653 		{TIPC_NLA_STATS_RX_STATES, s->recv_states},
2654 		{TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
2655 		{TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
2656 		{TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
2657 		{TIPC_NLA_STATS_TX_STATES, s->sent_states},
2658 		{TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
2659 		{TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
2660 		{TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
2661 		{TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
2662 		{TIPC_NLA_STATS_DUPLICATES, s->duplicates},
2663 		{TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
2664 		{TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
2665 		{TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
2666 			(s->accu_queue_sz / s->queue_sz_counts) : 0}
2667 	};
2668 
2669 	stats = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2670 	if (!stats)
2671 		return -EMSGSIZE;
2672 
2673 	for (i = 0; i <  ARRAY_SIZE(map); i++)
2674 		if (nla_put_u32(skb, map[i].key, map[i].val))
2675 			goto msg_full;
2676 
2677 	nla_nest_end(skb, stats);
2678 
2679 	return 0;
2680 msg_full:
2681 	nla_nest_cancel(skb, stats);
2682 
2683 	return -EMSGSIZE;
2684 }
2685 
2686 /* 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)2687 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
2688 		       struct tipc_link *link, int nlflags)
2689 {
2690 	u32 self = tipc_own_addr(net);
2691 	struct nlattr *attrs;
2692 	struct nlattr *prop;
2693 	void *hdr;
2694 	int err;
2695 
2696 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2697 			  nlflags, TIPC_NL_LINK_GET);
2698 	if (!hdr)
2699 		return -EMSGSIZE;
2700 
2701 	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2702 	if (!attrs)
2703 		goto msg_full;
2704 
2705 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
2706 		goto attr_msg_full;
2707 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
2708 		goto attr_msg_full;
2709 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2710 		goto attr_msg_full;
2711 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
2712 		goto attr_msg_full;
2713 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
2714 		goto attr_msg_full;
2715 
2716 	if (tipc_link_is_up(link))
2717 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2718 			goto attr_msg_full;
2719 	if (link->active)
2720 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2721 			goto attr_msg_full;
2722 
2723 	prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2724 	if (!prop)
2725 		goto attr_msg_full;
2726 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2727 		goto prop_msg_full;
2728 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2729 		goto prop_msg_full;
2730 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2731 			link->window))
2732 		goto prop_msg_full;
2733 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2734 		goto prop_msg_full;
2735 	nla_nest_end(msg->skb, prop);
2736 
2737 	err = __tipc_nl_add_stats(msg->skb, &link->stats);
2738 	if (err)
2739 		goto attr_msg_full;
2740 
2741 	nla_nest_end(msg->skb, attrs);
2742 	genlmsg_end(msg->skb, hdr);
2743 
2744 	return 0;
2745 
2746 prop_msg_full:
2747 	nla_nest_cancel(msg->skb, prop);
2748 attr_msg_full:
2749 	nla_nest_cancel(msg->skb, attrs);
2750 msg_full:
2751 	genlmsg_cancel(msg->skb, hdr);
2752 
2753 	return -EMSGSIZE;
2754 }
2755 
__tipc_nl_add_bc_link_stat(struct sk_buff * skb,struct tipc_stats * stats)2756 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2757 				      struct tipc_stats *stats)
2758 {
2759 	int i;
2760 	struct nlattr *nest;
2761 
2762 	struct nla_map {
2763 		__u32 key;
2764 		__u32 val;
2765 	};
2766 
2767 	struct nla_map map[] = {
2768 		{TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2769 		{TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2770 		{TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2771 		{TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2772 		{TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2773 		{TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2774 		{TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2775 		{TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2776 		{TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2777 		{TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2778 		{TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2779 		{TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2780 		{TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2781 		{TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2782 		{TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2783 		{TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2784 		{TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2785 		{TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2786 		{TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2787 			(stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2788 	};
2789 
2790 	nest = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2791 	if (!nest)
2792 		return -EMSGSIZE;
2793 
2794 	for (i = 0; i <  ARRAY_SIZE(map); i++)
2795 		if (nla_put_u32(skb, map[i].key, map[i].val))
2796 			goto msg_full;
2797 
2798 	nla_nest_end(skb, nest);
2799 
2800 	return 0;
2801 msg_full:
2802 	nla_nest_cancel(skb, nest);
2803 
2804 	return -EMSGSIZE;
2805 }
2806 
tipc_nl_add_bc_link(struct net * net,struct tipc_nl_msg * msg,struct tipc_link * bcl)2807 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg,
2808 			struct tipc_link *bcl)
2809 {
2810 	int err;
2811 	void *hdr;
2812 	struct nlattr *attrs;
2813 	struct nlattr *prop;
2814 	u32 bc_mode = tipc_bcast_get_mode(net);
2815 	u32 bc_ratio = tipc_bcast_get_broadcast_ratio(net);
2816 
2817 	if (!bcl)
2818 		return 0;
2819 
2820 	tipc_bcast_lock(net);
2821 
2822 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2823 			  NLM_F_MULTI, TIPC_NL_LINK_GET);
2824 	if (!hdr) {
2825 		tipc_bcast_unlock(net);
2826 		return -EMSGSIZE;
2827 	}
2828 
2829 	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2830 	if (!attrs)
2831 		goto msg_full;
2832 
2833 	/* The broadcast link is always up */
2834 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2835 		goto attr_msg_full;
2836 
2837 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2838 		goto attr_msg_full;
2839 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2840 		goto attr_msg_full;
2841 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2842 		goto attr_msg_full;
2843 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2844 		goto attr_msg_full;
2845 
2846 	prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2847 	if (!prop)
2848 		goto attr_msg_full;
2849 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->max_win))
2850 		goto prop_msg_full;
2851 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST, bc_mode))
2852 		goto prop_msg_full;
2853 	if (bc_mode & BCLINK_MODE_SEL)
2854 		if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST_RATIO,
2855 				bc_ratio))
2856 			goto prop_msg_full;
2857 	nla_nest_end(msg->skb, prop);
2858 
2859 	err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2860 	if (err)
2861 		goto attr_msg_full;
2862 
2863 	tipc_bcast_unlock(net);
2864 	nla_nest_end(msg->skb, attrs);
2865 	genlmsg_end(msg->skb, hdr);
2866 
2867 	return 0;
2868 
2869 prop_msg_full:
2870 	nla_nest_cancel(msg->skb, prop);
2871 attr_msg_full:
2872 	nla_nest_cancel(msg->skb, attrs);
2873 msg_full:
2874 	tipc_bcast_unlock(net);
2875 	genlmsg_cancel(msg->skb, hdr);
2876 
2877 	return -EMSGSIZE;
2878 }
2879 
tipc_link_set_tolerance(struct tipc_link * l,u32 tol,struct sk_buff_head * xmitq)2880 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2881 			     struct sk_buff_head *xmitq)
2882 {
2883 	l->tolerance = tol;
2884 	if (l->bc_rcvlink)
2885 		l->bc_rcvlink->tolerance = tol;
2886 	if (link_is_up(l))
2887 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2888 }
2889 
tipc_link_set_prio(struct tipc_link * l,u32 prio,struct sk_buff_head * xmitq)2890 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2891 			struct sk_buff_head *xmitq)
2892 {
2893 	l->priority = prio;
2894 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2895 }
2896 
tipc_link_set_abort_limit(struct tipc_link * l,u32 limit)2897 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2898 {
2899 	l->abort_limit = limit;
2900 }
2901 
2902 /**
2903  * tipc_link_dump - dump TIPC link data
2904  * @l: tipc link to be dumped
2905  * @dqueues: bitmask to decide if any link queue to be dumped?
2906  *           - TIPC_DUMP_NONE: don't dump link queues
2907  *           - TIPC_DUMP_TRANSMQ: dump link transmq queue
2908  *           - TIPC_DUMP_BACKLOGQ: dump link backlog queue
2909  *           - TIPC_DUMP_DEFERDQ: dump link deferd queue
2910  *           - TIPC_DUMP_INPUTQ: dump link input queue
2911  *           - TIPC_DUMP_WAKEUP: dump link wakeup queue
2912  *           - TIPC_DUMP_ALL: dump all the link queues above
2913  * @buf: returned buffer of dump data in format
2914  */
tipc_link_dump(struct tipc_link * l,u16 dqueues,char * buf)2915 int tipc_link_dump(struct tipc_link *l, u16 dqueues, char *buf)
2916 {
2917 	int i = 0;
2918 	size_t sz = (dqueues) ? LINK_LMAX : LINK_LMIN;
2919 	struct sk_buff_head *list;
2920 	struct sk_buff *hskb, *tskb;
2921 	u32 len;
2922 
2923 	if (!l) {
2924 		i += scnprintf(buf, sz, "link data: (null)\n");
2925 		return i;
2926 	}
2927 
2928 	i += scnprintf(buf, sz, "link data: %x", l->addr);
2929 	i += scnprintf(buf + i, sz - i, " %x", l->state);
2930 	i += scnprintf(buf + i, sz - i, " %u", l->in_session);
2931 	i += scnprintf(buf + i, sz - i, " %u", l->session);
2932 	i += scnprintf(buf + i, sz - i, " %u", l->peer_session);
2933 	i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt);
2934 	i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt);
2935 	i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt_state);
2936 	i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt_state);
2937 	i += scnprintf(buf + i, sz - i, " %x", l->peer_caps);
2938 	i += scnprintf(buf + i, sz - i, " %u", l->silent_intv_cnt);
2939 	i += scnprintf(buf + i, sz - i, " %u", l->rst_cnt);
2940 	i += scnprintf(buf + i, sz - i, " %u", 0);
2941 	i += scnprintf(buf + i, sz - i, " %u", 0);
2942 	i += scnprintf(buf + i, sz - i, " %u", l->acked);
2943 
2944 	list = &l->transmq;
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->deferdq;
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", len,
2957 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2958 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2959 
2960 	list = &l->backlogq;
2961 	len = skb_queue_len(list);
2962 	hskb = skb_peek(list);
2963 	tskb = skb_peek_tail(list);
2964 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2965 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2966 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2967 
2968 	list = l->inputq;
2969 	len = skb_queue_len(list);
2970 	hskb = skb_peek(list);
2971 	tskb = skb_peek_tail(list);
2972 	i += scnprintf(buf + i, sz - i, " | %u %u %u\n", len,
2973 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2974 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2975 
2976 	if (dqueues & TIPC_DUMP_TRANSMQ) {
2977 		i += scnprintf(buf + i, sz - i, "transmq: ");
2978 		i += tipc_list_dump(&l->transmq, false, buf + i);
2979 	}
2980 	if (dqueues & TIPC_DUMP_BACKLOGQ) {
2981 		i += scnprintf(buf + i, sz - i,
2982 			       "backlogq: <%u %u %u %u %u>, ",
2983 			       l->backlog[TIPC_LOW_IMPORTANCE].len,
2984 			       l->backlog[TIPC_MEDIUM_IMPORTANCE].len,
2985 			       l->backlog[TIPC_HIGH_IMPORTANCE].len,
2986 			       l->backlog[TIPC_CRITICAL_IMPORTANCE].len,
2987 			       l->backlog[TIPC_SYSTEM_IMPORTANCE].len);
2988 		i += tipc_list_dump(&l->backlogq, false, buf + i);
2989 	}
2990 	if (dqueues & TIPC_DUMP_DEFERDQ) {
2991 		i += scnprintf(buf + i, sz - i, "deferdq: ");
2992 		i += tipc_list_dump(&l->deferdq, false, buf + i);
2993 	}
2994 	if (dqueues & TIPC_DUMP_INPUTQ) {
2995 		i += scnprintf(buf + i, sz - i, "inputq: ");
2996 		i += tipc_list_dump(l->inputq, false, buf + i);
2997 	}
2998 	if (dqueues & TIPC_DUMP_WAKEUP) {
2999 		i += scnprintf(buf + i, sz - i, "wakeup: ");
3000 		i += tipc_list_dump(&l->wakeupq, false, buf + i);
3001 	}
3002 
3003 	return i;
3004 }
3005