1 | /* |
2 | * net/tipc/link.c: TIPC link code |
3 | * |
4 | * Copyright (c) 1996-2007, 2012-2016, Ericsson AB |
5 | * Copyright (c) 2004-2007, 2010-2013, Wind River Systems |
6 | * All rights reserved. |
7 | * |
8 | * Redistribution and use in source and binary forms, with or without |
9 | * modification, are permitted provided that the following conditions are met: |
10 | * |
11 | * 1. Redistributions of source code must retain the above copyright |
12 | * notice, this list of conditions and the following disclaimer. |
13 | * 2. Redistributions in binary form must reproduce the above copyright |
14 | * notice, this list of conditions and the following disclaimer in the |
15 | * documentation and/or other materials provided with the distribution. |
16 | * 3. Neither the names of the copyright holders nor the names of its |
17 | * contributors may be used to endorse or promote products derived from |
18 | * this software without specific prior written permission. |
19 | * |
20 | * Alternatively, this software may be distributed under the terms of the |
21 | * GNU General Public License ("GPL") version 2 as published by the Free |
22 | * Software Foundation. |
23 | * |
24 | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" |
25 | * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
26 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
27 | * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE |
28 | * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR |
29 | * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF |
30 | * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS |
31 | * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN |
32 | * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) |
33 | * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
34 | * POSSIBILITY OF SUCH DAMAGE. |
35 | */ |
36 | |
37 | #include "core.h" |
38 | #include "subscr.h" |
39 | #include "link.h" |
40 | #include "bcast.h" |
41 | #include "socket.h" |
42 | #include "name_distr.h" |
43 | #include "discover.h" |
44 | #include "netlink.h" |
45 | #include "monitor.h" |
46 | #include "trace.h" |
47 | #include "crypto.h" |
48 | |
49 | #include <linux/pkt_sched.h> |
50 | |
51 | struct tipc_stats { |
52 | u32 sent_pkts; |
53 | u32 recv_pkts; |
54 | u32 sent_states; |
55 | u32 recv_states; |
56 | u32 sent_probes; |
57 | u32 recv_probes; |
58 | u32 sent_nacks; |
59 | u32 recv_nacks; |
60 | u32 sent_acks; |
61 | u32 sent_bundled; |
62 | u32 sent_bundles; |
63 | u32 recv_bundled; |
64 | u32 recv_bundles; |
65 | u32 retransmitted; |
66 | u32 sent_fragmented; |
67 | u32 sent_fragments; |
68 | u32 recv_fragmented; |
69 | u32 recv_fragments; |
70 | u32 link_congs; /* # port sends blocked by congestion */ |
71 | u32 deferred_recv; |
72 | u32 duplicates; |
73 | u32 max_queue_sz; /* send queue size high water mark */ |
74 | u32 accu_queue_sz; /* used for send queue size profiling */ |
75 | u32 queue_sz_counts; /* used for send queue size profiling */ |
76 | u32 msg_length_counts; /* used for message length profiling */ |
77 | u32 msg_lengths_total; /* used for message length profiling */ |
78 | u32 msg_length_profile[7]; /* used for msg. length profiling */ |
79 | }; |
80 | |
81 | /** |
82 | * struct tipc_link - TIPC link data structure |
83 | * @addr: network address of link's peer node |
84 | * @name: link name character string |
85 | * @net: pointer to namespace struct |
86 | * @peer_session: link session # being used by peer end of link |
87 | * @peer_bearer_id: bearer id used by link's peer endpoint |
88 | * @bearer_id: local bearer id used by link |
89 | * @tolerance: minimum link continuity loss needed to reset link [in ms] |
90 | * @abort_limit: # of unacknowledged continuity probes needed to reset link |
91 | * @state: current state of link FSM |
92 | * @peer_caps: bitmap describing capabilities of peer node |
93 | * @silent_intv_cnt: # of timer intervals without any reception from peer |
94 | * @priority: current link priority |
95 | * @net_plane: current link network plane ('A' through 'H') |
96 | * @mon_state: cookie with information needed by link monitor |
97 | * @mtu: current maximum packet size for this link |
98 | * @advertised_mtu: advertised own mtu when link is being established |
99 | * @backlogq: queue for messages waiting to be sent |
100 | * @ackers: # of peers that needs to ack each packet before it can be released |
101 | * @acked: # last packet acked by a certain peer. Used for broadcast. |
102 | * @rcv_nxt: next sequence number to expect for inbound messages |
103 | * @inputq: buffer queue for messages to be delivered upwards |
104 | * @namedq: buffer queue for name table messages to be delivered upwards |
105 | * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate |
106 | * @reasm_buf: head of partially reassembled inbound message fragments |
107 | * @stats: collects statistics regarding link activity |
108 | * @session: session to be used by link |
109 | * @snd_nxt_state: next send seq number |
110 | * @rcv_nxt_state: next rcv seq number |
111 | * @in_session: have received ACTIVATE_MSG from peer |
112 | * @active: link is active |
113 | * @if_name: associated interface name |
114 | * @rst_cnt: link reset counter |
115 | * @drop_point: seq number for failover handling (FIXME) |
116 | * @failover_reasm_skb: saved failover msg ptr (FIXME) |
117 | * @failover_deferdq: deferred message queue for failover processing (FIXME) |
118 | * @transmq: the link's transmit queue |
119 | * @backlog: link's backlog by priority (importance) |
120 | * @snd_nxt: next sequence number to be used |
121 | * @rcv_unacked: # messages read by user, but not yet acked back to peer |
122 | * @deferdq: deferred receive queue |
123 | * @window: sliding window size for congestion handling |
124 | * @min_win: minimal send window to be used by link |
125 | * @ssthresh: slow start threshold for congestion handling |
126 | * @max_win: maximal send window to be used by link |
127 | * @cong_acks: congestion acks for congestion avoidance (FIXME) |
128 | * @checkpoint: seq number for congestion window size handling |
129 | * @reasm_tnlmsg: fragmentation/reassembly area for tunnel protocol message |
130 | * @last_gap: last gap ack blocks for bcast (FIXME) |
131 | * @last_ga: ptr to gap ack blocks |
132 | * @bc_rcvlink: the peer specific link used for broadcast reception |
133 | * @bc_sndlink: the namespace global link used for broadcast sending |
134 | * @nack_state: bcast nack state |
135 | * @bc_peer_is_up: peer has acked the bcast init msg |
136 | */ |
137 | struct tipc_link { |
138 | u32 addr; |
139 | char name[TIPC_MAX_LINK_NAME]; |
140 | struct net *net; |
141 | |
142 | /* Management and link supervision data */ |
143 | u16 peer_session; |
144 | u16 session; |
145 | u16 snd_nxt_state; |
146 | u16 rcv_nxt_state; |
147 | u32 peer_bearer_id; |
148 | u32 bearer_id; |
149 | u32 tolerance; |
150 | u32 abort_limit; |
151 | u32 state; |
152 | u16 peer_caps; |
153 | bool in_session; |
154 | bool active; |
155 | u32 silent_intv_cnt; |
156 | char if_name[TIPC_MAX_IF_NAME]; |
157 | u32 priority; |
158 | char net_plane; |
159 | struct tipc_mon_state mon_state; |
160 | u16 rst_cnt; |
161 | |
162 | /* Failover/synch */ |
163 | u16 drop_point; |
164 | struct sk_buff *failover_reasm_skb; |
165 | struct sk_buff_head failover_deferdq; |
166 | |
167 | /* Max packet negotiation */ |
168 | u16 mtu; |
169 | u16 advertised_mtu; |
170 | |
171 | /* Sending */ |
172 | struct sk_buff_head transmq; |
173 | struct sk_buff_head backlogq; |
174 | struct { |
175 | u16 len; |
176 | u16 limit; |
177 | struct sk_buff *target_bskb; |
178 | } backlog[5]; |
179 | u16 snd_nxt; |
180 | |
181 | /* Reception */ |
182 | u16 rcv_nxt; |
183 | u32 rcv_unacked; |
184 | struct sk_buff_head deferdq; |
185 | struct sk_buff_head *inputq; |
186 | struct sk_buff_head *namedq; |
187 | |
188 | /* Congestion handling */ |
189 | struct sk_buff_head wakeupq; |
190 | u16 window; |
191 | u16 min_win; |
192 | u16 ssthresh; |
193 | u16 max_win; |
194 | u16 cong_acks; |
195 | u16 checkpoint; |
196 | |
197 | /* Fragmentation/reassembly */ |
198 | struct sk_buff *reasm_buf; |
199 | struct sk_buff *reasm_tnlmsg; |
200 | |
201 | /* Broadcast */ |
202 | u16 ackers; |
203 | u16 acked; |
204 | u16 last_gap; |
205 | struct tipc_gap_ack_blks *last_ga; |
206 | struct tipc_link *bc_rcvlink; |
207 | struct tipc_link *bc_sndlink; |
208 | u8 nack_state; |
209 | bool bc_peer_is_up; |
210 | |
211 | /* Statistics */ |
212 | struct tipc_stats stats; |
213 | }; |
214 | |
215 | /* |
216 | * Error message prefixes |
217 | */ |
218 | static const char *link_co_err = "Link tunneling error, " ; |
219 | static const char *link_rst_msg = "Resetting link " ; |
220 | |
221 | /* Send states for broadcast NACKs |
222 | */ |
223 | enum { |
224 | BC_NACK_SND_CONDITIONAL, |
225 | BC_NACK_SND_UNCONDITIONAL, |
226 | BC_NACK_SND_SUPPRESS, |
227 | }; |
228 | |
229 | #define TIPC_BC_RETR_LIM (jiffies + msecs_to_jiffies(10)) |
230 | #define TIPC_UC_RETR_TIME (jiffies + msecs_to_jiffies(1)) |
231 | |
232 | /* Link FSM states: |
233 | */ |
234 | enum { |
235 | LINK_ESTABLISHED = 0xe, |
236 | LINK_ESTABLISHING = 0xe << 4, |
237 | LINK_RESET = 0x1 << 8, |
238 | LINK_RESETTING = 0x2 << 12, |
239 | LINK_PEER_RESET = 0xd << 16, |
240 | LINK_FAILINGOVER = 0xf << 20, |
241 | LINK_SYNCHING = 0xc << 24 |
242 | }; |
243 | |
244 | /* Link FSM state checking routines |
245 | */ |
246 | static int link_is_up(struct tipc_link *l) |
247 | { |
248 | return l->state & (LINK_ESTABLISHED | LINK_SYNCHING); |
249 | } |
250 | |
251 | static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb, |
252 | struct sk_buff_head *xmitq); |
253 | static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe, |
254 | bool probe_reply, u16 rcvgap, |
255 | int tolerance, int priority, |
256 | struct sk_buff_head *xmitq); |
257 | static void link_print(struct tipc_link *l, const char *str); |
258 | static int tipc_link_build_nack_msg(struct tipc_link *l, |
259 | struct sk_buff_head *xmitq); |
260 | static void tipc_link_build_bc_init_msg(struct tipc_link *l, |
261 | struct sk_buff_head *xmitq); |
262 | static u8 __tipc_build_gap_ack_blks(struct tipc_gap_ack_blks *ga, |
263 | struct tipc_link *l, u8 start_index); |
264 | static u16 tipc_build_gap_ack_blks(struct tipc_link *l, struct tipc_msg *hdr); |
265 | static int tipc_link_advance_transmq(struct tipc_link *l, struct tipc_link *r, |
266 | u16 acked, u16 gap, |
267 | struct tipc_gap_ack_blks *ga, |
268 | struct sk_buff_head *xmitq, |
269 | bool *retransmitted, int *rc); |
270 | static void tipc_link_update_cwin(struct tipc_link *l, int released, |
271 | bool retransmitted); |
272 | /* |
273 | * Simple non-static link routines (i.e. referenced outside this file) |
274 | */ |
275 | bool tipc_link_is_up(struct tipc_link *l) |
276 | { |
277 | return link_is_up(l); |
278 | } |
279 | |
280 | bool tipc_link_peer_is_down(struct tipc_link *l) |
281 | { |
282 | return l->state == LINK_PEER_RESET; |
283 | } |
284 | |
285 | bool tipc_link_is_reset(struct tipc_link *l) |
286 | { |
287 | return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING); |
288 | } |
289 | |
290 | bool tipc_link_is_establishing(struct tipc_link *l) |
291 | { |
292 | return l->state == LINK_ESTABLISHING; |
293 | } |
294 | |
295 | bool tipc_link_is_synching(struct tipc_link *l) |
296 | { |
297 | return l->state == LINK_SYNCHING; |
298 | } |
299 | |
300 | bool tipc_link_is_failingover(struct tipc_link *l) |
301 | { |
302 | return l->state == LINK_FAILINGOVER; |
303 | } |
304 | |
305 | bool tipc_link_is_blocked(struct tipc_link *l) |
306 | { |
307 | return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER); |
308 | } |
309 | |
310 | static bool link_is_bc_sndlink(struct tipc_link *l) |
311 | { |
312 | return !l->bc_sndlink; |
313 | } |
314 | |
315 | static bool link_is_bc_rcvlink(struct tipc_link *l) |
316 | { |
317 | return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l)); |
318 | } |
319 | |
320 | void tipc_link_set_active(struct tipc_link *l, bool active) |
321 | { |
322 | l->active = active; |
323 | } |
324 | |
325 | u32 tipc_link_id(struct tipc_link *l) |
326 | { |
327 | return l->peer_bearer_id << 16 | l->bearer_id; |
328 | } |
329 | |
330 | int tipc_link_min_win(struct tipc_link *l) |
331 | { |
332 | return l->min_win; |
333 | } |
334 | |
335 | int tipc_link_max_win(struct tipc_link *l) |
336 | { |
337 | return l->max_win; |
338 | } |
339 | |
340 | int tipc_link_prio(struct tipc_link *l) |
341 | { |
342 | return l->priority; |
343 | } |
344 | |
345 | unsigned long tipc_link_tolerance(struct tipc_link *l) |
346 | { |
347 | return l->tolerance; |
348 | } |
349 | |
350 | struct sk_buff_head *tipc_link_inputq(struct tipc_link *l) |
351 | { |
352 | return l->inputq; |
353 | } |
354 | |
355 | char tipc_link_plane(struct tipc_link *l) |
356 | { |
357 | return l->net_plane; |
358 | } |
359 | |
360 | struct net *tipc_link_net(struct tipc_link *l) |
361 | { |
362 | return l->net; |
363 | } |
364 | |
365 | void tipc_link_update_caps(struct tipc_link *l, u16 capabilities) |
366 | { |
367 | l->peer_caps = capabilities; |
368 | } |
369 | |
370 | void tipc_link_add_bc_peer(struct tipc_link *snd_l, |
371 | struct tipc_link *uc_l, |
372 | struct sk_buff_head *xmitq) |
373 | { |
374 | struct tipc_link *rcv_l = uc_l->bc_rcvlink; |
375 | |
376 | snd_l->ackers++; |
377 | rcv_l->acked = snd_l->snd_nxt - 1; |
378 | snd_l->state = LINK_ESTABLISHED; |
379 | tipc_link_build_bc_init_msg(l: uc_l, xmitq); |
380 | } |
381 | |
382 | void tipc_link_remove_bc_peer(struct tipc_link *snd_l, |
383 | struct tipc_link *rcv_l, |
384 | struct sk_buff_head *xmitq) |
385 | { |
386 | u16 ack = snd_l->snd_nxt - 1; |
387 | |
388 | snd_l->ackers--; |
389 | rcv_l->bc_peer_is_up = true; |
390 | rcv_l->state = LINK_ESTABLISHED; |
391 | tipc_link_bc_ack_rcv(l: rcv_l, acked: ack, gap: 0, NULL, xmitq, NULL); |
392 | trace_tipc_link_reset(l: rcv_l, dqueues: TIPC_DUMP_ALL, header: "bclink removed!" ); |
393 | tipc_link_reset(l: rcv_l); |
394 | rcv_l->state = LINK_RESET; |
395 | if (!snd_l->ackers) { |
396 | trace_tipc_link_reset(l: snd_l, dqueues: TIPC_DUMP_ALL, header: "zero ackers!" ); |
397 | tipc_link_reset(l: snd_l); |
398 | snd_l->state = LINK_RESET; |
399 | __skb_queue_purge(list: xmitq); |
400 | } |
401 | } |
402 | |
403 | int tipc_link_bc_peers(struct tipc_link *l) |
404 | { |
405 | return l->ackers; |
406 | } |
407 | |
408 | static u16 link_bc_rcv_gap(struct tipc_link *l) |
409 | { |
410 | struct sk_buff *skb = skb_peek(list_: &l->deferdq); |
411 | u16 gap = 0; |
412 | |
413 | if (more(left: l->snd_nxt, right: l->rcv_nxt)) |
414 | gap = l->snd_nxt - l->rcv_nxt; |
415 | if (skb) |
416 | gap = buf_seqno(skb) - l->rcv_nxt; |
417 | return gap; |
418 | } |
419 | |
420 | void tipc_link_set_mtu(struct tipc_link *l, int mtu) |
421 | { |
422 | l->mtu = mtu; |
423 | } |
424 | |
425 | int tipc_link_mtu(struct tipc_link *l) |
426 | { |
427 | return l->mtu; |
428 | } |
429 | |
430 | int tipc_link_mss(struct tipc_link *l) |
431 | { |
432 | #ifdef CONFIG_TIPC_CRYPTO |
433 | return l->mtu - INT_H_SIZE - EMSG_OVERHEAD; |
434 | #else |
435 | return l->mtu - INT_H_SIZE; |
436 | #endif |
437 | } |
438 | |
439 | u16 tipc_link_rcv_nxt(struct tipc_link *l) |
440 | { |
441 | return l->rcv_nxt; |
442 | } |
443 | |
444 | u16 tipc_link_acked(struct tipc_link *l) |
445 | { |
446 | return l->acked; |
447 | } |
448 | |
449 | char *tipc_link_name(struct tipc_link *l) |
450 | { |
451 | return l->name; |
452 | } |
453 | |
454 | u32 tipc_link_state(struct tipc_link *l) |
455 | { |
456 | return l->state; |
457 | } |
458 | |
459 | /** |
460 | * tipc_link_create - create a new link |
461 | * @net: pointer to associated network namespace |
462 | * @if_name: associated interface name |
463 | * @bearer_id: id (index) of associated bearer |
464 | * @tolerance: link tolerance to be used by link |
465 | * @net_plane: network plane (A,B,c..) this link belongs to |
466 | * @mtu: mtu to be advertised by link |
467 | * @priority: priority to be used by link |
468 | * @min_win: minimal send window to be used by link |
469 | * @max_win: maximal send window to be used by link |
470 | * @session: session to be used by link |
471 | * @peer: node id of peer node |
472 | * @peer_caps: bitmap describing peer node capabilities |
473 | * @bc_sndlink: the namespace global link used for broadcast sending |
474 | * @bc_rcvlink: the peer specific link used for broadcast reception |
475 | * @inputq: queue to put messages ready for delivery |
476 | * @namedq: queue to put binding table update messages ready for delivery |
477 | * @link: return value, pointer to put the created link |
478 | * @self: local unicast link id |
479 | * @peer_id: 128-bit ID of peer |
480 | * |
481 | * Return: true if link was created, otherwise false |
482 | */ |
483 | bool tipc_link_create(struct net *net, char *if_name, int bearer_id, |
484 | int tolerance, char net_plane, u32 mtu, int priority, |
485 | u32 min_win, u32 max_win, u32 session, u32 self, |
486 | u32 peer, u8 *peer_id, u16 peer_caps, |
487 | struct tipc_link *bc_sndlink, |
488 | struct tipc_link *bc_rcvlink, |
489 | struct sk_buff_head *inputq, |
490 | struct sk_buff_head *namedq, |
491 | struct tipc_link **link) |
492 | { |
493 | char peer_str[NODE_ID_STR_LEN] = {0,}; |
494 | char self_str[NODE_ID_STR_LEN] = {0,}; |
495 | struct tipc_link *l; |
496 | |
497 | l = kzalloc(size: sizeof(*l), GFP_ATOMIC); |
498 | if (!l) |
499 | return false; |
500 | *link = l; |
501 | l->session = session; |
502 | |
503 | /* Set link name for unicast links only */ |
504 | if (peer_id) { |
505 | tipc_nodeid2string(str: self_str, id: tipc_own_id(net)); |
506 | if (strlen(self_str) > 16) |
507 | sprintf(buf: self_str, fmt: "%x" , self); |
508 | tipc_nodeid2string(str: peer_str, id: peer_id); |
509 | if (strlen(peer_str) > 16) |
510 | sprintf(buf: peer_str, fmt: "%x" , peer); |
511 | } |
512 | /* Peer i/f name will be completed by reset/activate message */ |
513 | snprintf(buf: l->name, size: sizeof(l->name), fmt: "%s:%s-%s:unknown" , |
514 | self_str, if_name, peer_str); |
515 | |
516 | strcpy(p: l->if_name, q: if_name); |
517 | l->addr = peer; |
518 | l->peer_caps = peer_caps; |
519 | l->net = net; |
520 | l->in_session = false; |
521 | l->bearer_id = bearer_id; |
522 | l->tolerance = tolerance; |
523 | if (bc_rcvlink) |
524 | bc_rcvlink->tolerance = tolerance; |
525 | l->net_plane = net_plane; |
526 | l->advertised_mtu = mtu; |
527 | l->mtu = mtu; |
528 | l->priority = priority; |
529 | tipc_link_set_queue_limits(l, min_win, max_win); |
530 | l->ackers = 1; |
531 | l->bc_sndlink = bc_sndlink; |
532 | l->bc_rcvlink = bc_rcvlink; |
533 | l->inputq = inputq; |
534 | l->namedq = namedq; |
535 | l->state = LINK_RESETTING; |
536 | __skb_queue_head_init(list: &l->transmq); |
537 | __skb_queue_head_init(list: &l->backlogq); |
538 | __skb_queue_head_init(list: &l->deferdq); |
539 | __skb_queue_head_init(list: &l->failover_deferdq); |
540 | skb_queue_head_init(list: &l->wakeupq); |
541 | skb_queue_head_init(list: l->inputq); |
542 | return true; |
543 | } |
544 | |
545 | /** |
546 | * tipc_link_bc_create - create new link to be used for broadcast |
547 | * @net: pointer to associated network namespace |
548 | * @mtu: mtu to be used initially if no peers |
549 | * @min_win: minimal send window to be used by link |
550 | * @max_win: maximal send window to be used by link |
551 | * @inputq: queue to put messages ready for delivery |
552 | * @namedq: queue to put binding table update messages ready for delivery |
553 | * @link: return value, pointer to put the created link |
554 | * @ownnode: identity of own node |
555 | * @peer: node id of peer node |
556 | * @peer_id: 128-bit ID of peer |
557 | * @peer_caps: bitmap describing peer node capabilities |
558 | * @bc_sndlink: the namespace global link used for broadcast sending |
559 | * |
560 | * Return: true if link was created, otherwise false |
561 | */ |
562 | bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer, u8 *peer_id, |
563 | int mtu, u32 min_win, u32 max_win, u16 peer_caps, |
564 | struct sk_buff_head *inputq, |
565 | struct sk_buff_head *namedq, |
566 | struct tipc_link *bc_sndlink, |
567 | struct tipc_link **link) |
568 | { |
569 | struct tipc_link *l; |
570 | |
571 | if (!tipc_link_create(net, if_name: "" , MAX_BEARERS, tolerance: 0, net_plane: 'Z', mtu, priority: 0, min_win, |
572 | max_win, session: 0, self: ownnode, peer, NULL, peer_caps, |
573 | bc_sndlink, NULL, inputq, namedq, link)) |
574 | return false; |
575 | |
576 | l = *link; |
577 | if (peer_id) { |
578 | char peer_str[NODE_ID_STR_LEN] = {0,}; |
579 | |
580 | tipc_nodeid2string(str: peer_str, id: peer_id); |
581 | if (strlen(peer_str) > 16) |
582 | sprintf(buf: peer_str, fmt: "%x" , peer); |
583 | /* Broadcast receiver link name: "broadcast-link:<peer>" */ |
584 | snprintf(buf: l->name, size: sizeof(l->name), fmt: "%s:%s" , tipc_bclink_name, |
585 | peer_str); |
586 | } else { |
587 | strcpy(p: l->name, q: tipc_bclink_name); |
588 | } |
589 | trace_tipc_link_reset(l, dqueues: TIPC_DUMP_ALL, header: "bclink created!" ); |
590 | tipc_link_reset(l); |
591 | l->state = LINK_RESET; |
592 | l->ackers = 0; |
593 | l->bc_rcvlink = l; |
594 | |
595 | /* Broadcast send link is always up */ |
596 | if (link_is_bc_sndlink(l)) |
597 | l->state = LINK_ESTABLISHED; |
598 | |
599 | /* Disable replicast if even a single peer doesn't support it */ |
600 | if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST)) |
601 | tipc_bcast_toggle_rcast(net, supp: false); |
602 | |
603 | return true; |
604 | } |
605 | |
606 | /** |
607 | * tipc_link_fsm_evt - link finite state machine |
608 | * @l: pointer to link |
609 | * @evt: state machine event to be processed |
610 | */ |
611 | int tipc_link_fsm_evt(struct tipc_link *l, int evt) |
612 | { |
613 | int rc = 0; |
614 | int old_state = l->state; |
615 | |
616 | switch (l->state) { |
617 | case LINK_RESETTING: |
618 | switch (evt) { |
619 | case LINK_PEER_RESET_EVT: |
620 | l->state = LINK_PEER_RESET; |
621 | break; |
622 | case LINK_RESET_EVT: |
623 | l->state = LINK_RESET; |
624 | break; |
625 | case LINK_FAILURE_EVT: |
626 | case LINK_FAILOVER_BEGIN_EVT: |
627 | case LINK_ESTABLISH_EVT: |
628 | case LINK_FAILOVER_END_EVT: |
629 | case LINK_SYNCH_BEGIN_EVT: |
630 | case LINK_SYNCH_END_EVT: |
631 | default: |
632 | goto illegal_evt; |
633 | } |
634 | break; |
635 | case LINK_RESET: |
636 | switch (evt) { |
637 | case LINK_PEER_RESET_EVT: |
638 | l->state = LINK_ESTABLISHING; |
639 | break; |
640 | case LINK_FAILOVER_BEGIN_EVT: |
641 | l->state = LINK_FAILINGOVER; |
642 | break; |
643 | case LINK_FAILURE_EVT: |
644 | case LINK_RESET_EVT: |
645 | case LINK_ESTABLISH_EVT: |
646 | case LINK_FAILOVER_END_EVT: |
647 | break; |
648 | case LINK_SYNCH_BEGIN_EVT: |
649 | case LINK_SYNCH_END_EVT: |
650 | default: |
651 | goto illegal_evt; |
652 | } |
653 | break; |
654 | case LINK_PEER_RESET: |
655 | switch (evt) { |
656 | case LINK_RESET_EVT: |
657 | l->state = LINK_ESTABLISHING; |
658 | break; |
659 | case LINK_PEER_RESET_EVT: |
660 | case LINK_ESTABLISH_EVT: |
661 | case LINK_FAILURE_EVT: |
662 | break; |
663 | case LINK_SYNCH_BEGIN_EVT: |
664 | case LINK_SYNCH_END_EVT: |
665 | case LINK_FAILOVER_BEGIN_EVT: |
666 | case LINK_FAILOVER_END_EVT: |
667 | default: |
668 | goto illegal_evt; |
669 | } |
670 | break; |
671 | case LINK_FAILINGOVER: |
672 | switch (evt) { |
673 | case LINK_FAILOVER_END_EVT: |
674 | l->state = LINK_RESET; |
675 | break; |
676 | case LINK_PEER_RESET_EVT: |
677 | case LINK_RESET_EVT: |
678 | case LINK_ESTABLISH_EVT: |
679 | case LINK_FAILURE_EVT: |
680 | break; |
681 | case LINK_FAILOVER_BEGIN_EVT: |
682 | case LINK_SYNCH_BEGIN_EVT: |
683 | case LINK_SYNCH_END_EVT: |
684 | default: |
685 | goto illegal_evt; |
686 | } |
687 | break; |
688 | case LINK_ESTABLISHING: |
689 | switch (evt) { |
690 | case LINK_ESTABLISH_EVT: |
691 | l->state = LINK_ESTABLISHED; |
692 | break; |
693 | case LINK_FAILOVER_BEGIN_EVT: |
694 | l->state = LINK_FAILINGOVER; |
695 | break; |
696 | case LINK_RESET_EVT: |
697 | l->state = LINK_RESET; |
698 | break; |
699 | case LINK_FAILURE_EVT: |
700 | case LINK_PEER_RESET_EVT: |
701 | case LINK_SYNCH_BEGIN_EVT: |
702 | case LINK_FAILOVER_END_EVT: |
703 | break; |
704 | case LINK_SYNCH_END_EVT: |
705 | default: |
706 | goto illegal_evt; |
707 | } |
708 | break; |
709 | case LINK_ESTABLISHED: |
710 | switch (evt) { |
711 | case LINK_PEER_RESET_EVT: |
712 | l->state = LINK_PEER_RESET; |
713 | rc |= TIPC_LINK_DOWN_EVT; |
714 | break; |
715 | case LINK_FAILURE_EVT: |
716 | l->state = LINK_RESETTING; |
717 | rc |= TIPC_LINK_DOWN_EVT; |
718 | break; |
719 | case LINK_RESET_EVT: |
720 | l->state = LINK_RESET; |
721 | break; |
722 | case LINK_ESTABLISH_EVT: |
723 | case LINK_SYNCH_END_EVT: |
724 | break; |
725 | case LINK_SYNCH_BEGIN_EVT: |
726 | l->state = LINK_SYNCHING; |
727 | break; |
728 | case LINK_FAILOVER_BEGIN_EVT: |
729 | case LINK_FAILOVER_END_EVT: |
730 | default: |
731 | goto illegal_evt; |
732 | } |
733 | break; |
734 | case LINK_SYNCHING: |
735 | switch (evt) { |
736 | case LINK_PEER_RESET_EVT: |
737 | l->state = LINK_PEER_RESET; |
738 | rc |= TIPC_LINK_DOWN_EVT; |
739 | break; |
740 | case LINK_FAILURE_EVT: |
741 | l->state = LINK_RESETTING; |
742 | rc |= TIPC_LINK_DOWN_EVT; |
743 | break; |
744 | case LINK_RESET_EVT: |
745 | l->state = LINK_RESET; |
746 | break; |
747 | case LINK_ESTABLISH_EVT: |
748 | case LINK_SYNCH_BEGIN_EVT: |
749 | break; |
750 | case LINK_SYNCH_END_EVT: |
751 | l->state = LINK_ESTABLISHED; |
752 | break; |
753 | case LINK_FAILOVER_BEGIN_EVT: |
754 | case LINK_FAILOVER_END_EVT: |
755 | default: |
756 | goto illegal_evt; |
757 | } |
758 | break; |
759 | default: |
760 | pr_err("Unknown FSM state %x in %s\n" , l->state, l->name); |
761 | } |
762 | trace_tipc_link_fsm(name: l->name, os: old_state, ns: l->state, evt); |
763 | return rc; |
764 | illegal_evt: |
765 | pr_err("Illegal FSM event %x in state %x on link %s\n" , |
766 | evt, l->state, l->name); |
767 | trace_tipc_link_fsm(name: l->name, os: old_state, ns: l->state, evt); |
768 | return rc; |
769 | } |
770 | |
771 | /* link_profile_stats - update statistical profiling of traffic |
772 | */ |
773 | static void link_profile_stats(struct tipc_link *l) |
774 | { |
775 | struct sk_buff *skb; |
776 | struct tipc_msg *msg; |
777 | int length; |
778 | |
779 | /* Update counters used in statistical profiling of send traffic */ |
780 | l->stats.accu_queue_sz += skb_queue_len(list_: &l->transmq); |
781 | l->stats.queue_sz_counts++; |
782 | |
783 | skb = skb_peek(list_: &l->transmq); |
784 | if (!skb) |
785 | return; |
786 | msg = buf_msg(skb); |
787 | length = msg_size(m: msg); |
788 | |
789 | if (msg_user(m: msg) == MSG_FRAGMENTER) { |
790 | if (msg_type(m: msg) != FIRST_FRAGMENT) |
791 | return; |
792 | length = msg_size(m: msg_inner_hdr(m: msg)); |
793 | } |
794 | l->stats.msg_lengths_total += length; |
795 | l->stats.msg_length_counts++; |
796 | if (length <= 64) |
797 | l->stats.msg_length_profile[0]++; |
798 | else if (length <= 256) |
799 | l->stats.msg_length_profile[1]++; |
800 | else if (length <= 1024) |
801 | l->stats.msg_length_profile[2]++; |
802 | else if (length <= 4096) |
803 | l->stats.msg_length_profile[3]++; |
804 | else if (length <= 16384) |
805 | l->stats.msg_length_profile[4]++; |
806 | else if (length <= 32768) |
807 | l->stats.msg_length_profile[5]++; |
808 | else |
809 | l->stats.msg_length_profile[6]++; |
810 | } |
811 | |
812 | /** |
813 | * tipc_link_too_silent - check if link is "too silent" |
814 | * @l: tipc link to be checked |
815 | * |
816 | * Return: true if the link 'silent_intv_cnt' is about to reach the |
817 | * 'abort_limit' value, otherwise false |
818 | */ |
819 | bool tipc_link_too_silent(struct tipc_link *l) |
820 | { |
821 | return (l->silent_intv_cnt + 2 > l->abort_limit); |
822 | } |
823 | |
824 | /* tipc_link_timeout - perform periodic task as instructed from node timeout |
825 | */ |
826 | int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq) |
827 | { |
828 | int mtyp = 0; |
829 | int rc = 0; |
830 | bool state = false; |
831 | bool probe = false; |
832 | bool setup = false; |
833 | u16 bc_snt = l->bc_sndlink->snd_nxt - 1; |
834 | u16 bc_acked = l->bc_rcvlink->acked; |
835 | struct tipc_mon_state *mstate = &l->mon_state; |
836 | |
837 | trace_tipc_link_timeout(l, dqueues: TIPC_DUMP_NONE, header: " " ); |
838 | trace_tipc_link_too_silent(l, dqueues: TIPC_DUMP_ALL, header: " " ); |
839 | switch (l->state) { |
840 | case LINK_ESTABLISHED: |
841 | case LINK_SYNCHING: |
842 | mtyp = STATE_MSG; |
843 | link_profile_stats(l); |
844 | tipc_mon_get_state(net: l->net, addr: l->addr, state: mstate, bearer_id: l->bearer_id); |
845 | if (mstate->reset || (l->silent_intv_cnt > l->abort_limit)) |
846 | return tipc_link_fsm_evt(l, evt: LINK_FAILURE_EVT); |
847 | state = bc_acked != bc_snt; |
848 | state |= l->bc_rcvlink->rcv_unacked; |
849 | state |= l->rcv_unacked; |
850 | state |= !skb_queue_empty(list: &l->transmq); |
851 | probe = mstate->probing; |
852 | probe |= l->silent_intv_cnt; |
853 | if (probe || mstate->monitoring) |
854 | l->silent_intv_cnt++; |
855 | probe |= !skb_queue_empty(list: &l->deferdq); |
856 | if (l->snd_nxt == l->checkpoint) { |
857 | tipc_link_update_cwin(l, released: 0, retransmitted: 0); |
858 | probe = true; |
859 | } |
860 | l->checkpoint = l->snd_nxt; |
861 | break; |
862 | case LINK_RESET: |
863 | setup = l->rst_cnt++ <= 4; |
864 | setup |= !(l->rst_cnt % 16); |
865 | mtyp = RESET_MSG; |
866 | break; |
867 | case LINK_ESTABLISHING: |
868 | setup = true; |
869 | mtyp = ACTIVATE_MSG; |
870 | break; |
871 | case LINK_PEER_RESET: |
872 | case LINK_RESETTING: |
873 | case LINK_FAILINGOVER: |
874 | break; |
875 | default: |
876 | break; |
877 | } |
878 | |
879 | if (state || probe || setup) |
880 | tipc_link_build_proto_msg(l, mtyp, probe, probe_reply: 0, rcvgap: 0, tolerance: 0, priority: 0, xmitq); |
881 | |
882 | return rc; |
883 | } |
884 | |
885 | /** |
886 | * link_schedule_user - schedule a message sender for wakeup after congestion |
887 | * @l: congested link |
888 | * @hdr: header of message that is being sent |
889 | * Create pseudo msg to send back to user when congestion abates |
890 | */ |
891 | static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr) |
892 | { |
893 | u32 dnode = tipc_own_addr(net: l->net); |
894 | u32 dport = msg_origport(m: hdr); |
895 | struct sk_buff *skb; |
896 | |
897 | /* Create and schedule wakeup pseudo message */ |
898 | skb = tipc_msg_create(SOCK_WAKEUP, type: 0, INT_H_SIZE, data_sz: 0, |
899 | dnode, onode: l->addr, dport, oport: 0, errcode: 0); |
900 | if (!skb) |
901 | return -ENOBUFS; |
902 | msg_set_dest_droppable(m: buf_msg(skb), d: true); |
903 | TIPC_SKB_CB(skb)->chain_imp = msg_importance(m: hdr); |
904 | skb_queue_tail(list: &l->wakeupq, newsk: skb); |
905 | l->stats.link_congs++; |
906 | trace_tipc_link_conges(l, dqueues: TIPC_DUMP_ALL, header: "wakeup scheduled!" ); |
907 | return -ELINKCONG; |
908 | } |
909 | |
910 | /** |
911 | * link_prepare_wakeup - prepare users for wakeup after congestion |
912 | * @l: congested link |
913 | * Wake up a number of waiting users, as permitted by available space |
914 | * in the send queue |
915 | */ |
916 | static void link_prepare_wakeup(struct tipc_link *l) |
917 | { |
918 | struct sk_buff_head *wakeupq = &l->wakeupq; |
919 | struct sk_buff_head *inputq = l->inputq; |
920 | struct sk_buff *skb, *tmp; |
921 | struct sk_buff_head tmpq; |
922 | int avail[5] = {0,}; |
923 | int imp = 0; |
924 | |
925 | __skb_queue_head_init(list: &tmpq); |
926 | |
927 | for (; imp <= TIPC_SYSTEM_IMPORTANCE; imp++) |
928 | avail[imp] = l->backlog[imp].limit - l->backlog[imp].len; |
929 | |
930 | skb_queue_walk_safe(wakeupq, skb, tmp) { |
931 | imp = TIPC_SKB_CB(skb)->chain_imp; |
932 | if (avail[imp] <= 0) |
933 | continue; |
934 | avail[imp]--; |
935 | __skb_unlink(skb, list: wakeupq); |
936 | __skb_queue_tail(list: &tmpq, newsk: skb); |
937 | } |
938 | |
939 | spin_lock_bh(lock: &inputq->lock); |
940 | skb_queue_splice_tail(list: &tmpq, head: inputq); |
941 | spin_unlock_bh(lock: &inputq->lock); |
942 | |
943 | } |
944 | |
945 | /** |
946 | * tipc_link_set_skb_retransmit_time - set the time at which retransmission of |
947 | * the given skb should be next attempted |
948 | * @skb: skb to set a future retransmission time for |
949 | * @l: link the skb will be transmitted on |
950 | */ |
951 | static void tipc_link_set_skb_retransmit_time(struct sk_buff *skb, |
952 | struct tipc_link *l) |
953 | { |
954 | if (link_is_bc_sndlink(l)) |
955 | TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM; |
956 | else |
957 | TIPC_SKB_CB(skb)->nxt_retr = TIPC_UC_RETR_TIME; |
958 | } |
959 | |
960 | void tipc_link_reset(struct tipc_link *l) |
961 | { |
962 | struct sk_buff_head list; |
963 | u32 imp; |
964 | |
965 | __skb_queue_head_init(list: &list); |
966 | |
967 | l->in_session = false; |
968 | /* Force re-synch of peer session number before establishing */ |
969 | l->peer_session--; |
970 | l->session++; |
971 | l->mtu = l->advertised_mtu; |
972 | |
973 | spin_lock_bh(lock: &l->wakeupq.lock); |
974 | skb_queue_splice_init(list: &l->wakeupq, head: &list); |
975 | spin_unlock_bh(lock: &l->wakeupq.lock); |
976 | |
977 | spin_lock_bh(lock: &l->inputq->lock); |
978 | skb_queue_splice_init(list: &list, head: l->inputq); |
979 | spin_unlock_bh(lock: &l->inputq->lock); |
980 | |
981 | __skb_queue_purge(list: &l->transmq); |
982 | __skb_queue_purge(list: &l->deferdq); |
983 | __skb_queue_purge(list: &l->backlogq); |
984 | __skb_queue_purge(list: &l->failover_deferdq); |
985 | for (imp = 0; imp <= TIPC_SYSTEM_IMPORTANCE; imp++) { |
986 | l->backlog[imp].len = 0; |
987 | l->backlog[imp].target_bskb = NULL; |
988 | } |
989 | kfree_skb(skb: l->reasm_buf); |
990 | kfree_skb(skb: l->reasm_tnlmsg); |
991 | kfree_skb(skb: l->failover_reasm_skb); |
992 | l->reasm_buf = NULL; |
993 | l->reasm_tnlmsg = NULL; |
994 | l->failover_reasm_skb = NULL; |
995 | l->rcv_unacked = 0; |
996 | l->snd_nxt = 1; |
997 | l->rcv_nxt = 1; |
998 | l->snd_nxt_state = 1; |
999 | l->rcv_nxt_state = 1; |
1000 | l->acked = 0; |
1001 | l->last_gap = 0; |
1002 | kfree(objp: l->last_ga); |
1003 | l->last_ga = NULL; |
1004 | l->silent_intv_cnt = 0; |
1005 | l->rst_cnt = 0; |
1006 | l->bc_peer_is_up = false; |
1007 | memset(&l->mon_state, 0, sizeof(l->mon_state)); |
1008 | tipc_link_reset_stats(l); |
1009 | } |
1010 | |
1011 | /** |
1012 | * tipc_link_xmit(): enqueue buffer list according to queue situation |
1013 | * @l: link to use |
1014 | * @list: chain of buffers containing message |
1015 | * @xmitq: returned list of packets to be sent by caller |
1016 | * |
1017 | * Consumes the buffer chain. |
1018 | * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted |
1019 | * Return: 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS |
1020 | */ |
1021 | int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list, |
1022 | struct sk_buff_head *xmitq) |
1023 | { |
1024 | struct sk_buff_head *backlogq = &l->backlogq; |
1025 | struct sk_buff_head *transmq = &l->transmq; |
1026 | struct sk_buff *skb, *_skb; |
1027 | u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1; |
1028 | u16 ack = l->rcv_nxt - 1; |
1029 | u16 seqno = l->snd_nxt; |
1030 | int pkt_cnt = skb_queue_len(list_: list); |
1031 | unsigned int mss = tipc_link_mss(l); |
1032 | unsigned int cwin = l->window; |
1033 | unsigned int mtu = l->mtu; |
1034 | struct tipc_msg *hdr; |
1035 | bool new_bundle; |
1036 | int rc = 0; |
1037 | int imp; |
1038 | |
1039 | if (pkt_cnt <= 0) |
1040 | return 0; |
1041 | |
1042 | hdr = buf_msg(skb: skb_peek(list_: list)); |
1043 | if (unlikely(msg_size(hdr) > mtu)) { |
1044 | pr_warn("Too large msg, purging xmit list %d %d %d %d %d!\n" , |
1045 | skb_queue_len(list), msg_user(hdr), |
1046 | msg_type(hdr), msg_size(hdr), mtu); |
1047 | __skb_queue_purge(list); |
1048 | return -EMSGSIZE; |
1049 | } |
1050 | |
1051 | imp = msg_importance(m: hdr); |
1052 | /* Allow oversubscription of one data msg per source at congestion */ |
1053 | if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) { |
1054 | if (imp == TIPC_SYSTEM_IMPORTANCE) { |
1055 | pr_warn("%s<%s>, link overflow" , link_rst_msg, l->name); |
1056 | return -ENOBUFS; |
1057 | } |
1058 | rc = link_schedule_user(l, hdr); |
1059 | } |
1060 | |
1061 | if (pkt_cnt > 1) { |
1062 | l->stats.sent_fragmented++; |
1063 | l->stats.sent_fragments += pkt_cnt; |
1064 | } |
1065 | |
1066 | /* Prepare each packet for sending, and add to relevant queue: */ |
1067 | while ((skb = __skb_dequeue(list))) { |
1068 | if (likely(skb_queue_len(transmq) < cwin)) { |
1069 | hdr = buf_msg(skb); |
1070 | msg_set_seqno(m: hdr, n: seqno); |
1071 | msg_set_ack(m: hdr, n: ack); |
1072 | msg_set_bcast_ack(m: hdr, n: bc_ack); |
1073 | _skb = skb_clone(skb, GFP_ATOMIC); |
1074 | if (!_skb) { |
1075 | kfree_skb(skb); |
1076 | __skb_queue_purge(list); |
1077 | return -ENOBUFS; |
1078 | } |
1079 | __skb_queue_tail(list: transmq, newsk: skb); |
1080 | tipc_link_set_skb_retransmit_time(skb, l); |
1081 | __skb_queue_tail(list: xmitq, newsk: _skb); |
1082 | TIPC_SKB_CB(skb)->ackers = l->ackers; |
1083 | l->rcv_unacked = 0; |
1084 | l->stats.sent_pkts++; |
1085 | seqno++; |
1086 | continue; |
1087 | } |
1088 | if (tipc_msg_try_bundle(tskb: l->backlog[imp].target_bskb, skb: &skb, |
1089 | mss, dnode: l->addr, new_bundle: &new_bundle)) { |
1090 | if (skb) { |
1091 | /* Keep a ref. to the skb for next try */ |
1092 | l->backlog[imp].target_bskb = skb; |
1093 | l->backlog[imp].len++; |
1094 | __skb_queue_tail(list: backlogq, newsk: skb); |
1095 | } else { |
1096 | if (new_bundle) { |
1097 | l->stats.sent_bundles++; |
1098 | l->stats.sent_bundled++; |
1099 | } |
1100 | l->stats.sent_bundled++; |
1101 | } |
1102 | continue; |
1103 | } |
1104 | l->backlog[imp].target_bskb = NULL; |
1105 | l->backlog[imp].len += (1 + skb_queue_len(list_: list)); |
1106 | __skb_queue_tail(list: backlogq, newsk: skb); |
1107 | skb_queue_splice_tail_init(list, head: backlogq); |
1108 | } |
1109 | l->snd_nxt = seqno; |
1110 | return rc; |
1111 | } |
1112 | |
1113 | static void tipc_link_update_cwin(struct tipc_link *l, int released, |
1114 | bool retransmitted) |
1115 | { |
1116 | int bklog_len = skb_queue_len(list_: &l->backlogq); |
1117 | struct sk_buff_head *txq = &l->transmq; |
1118 | int txq_len = skb_queue_len(list_: txq); |
1119 | u16 cwin = l->window; |
1120 | |
1121 | /* Enter fast recovery */ |
1122 | if (unlikely(retransmitted)) { |
1123 | l->ssthresh = max_t(u16, l->window / 2, 300); |
1124 | l->window = min_t(u16, l->ssthresh, l->window); |
1125 | return; |
1126 | } |
1127 | /* Enter slow start */ |
1128 | if (unlikely(!released)) { |
1129 | l->ssthresh = max_t(u16, l->window / 2, 300); |
1130 | l->window = l->min_win; |
1131 | return; |
1132 | } |
1133 | /* Don't increase window if no pressure on the transmit queue */ |
1134 | if (txq_len + bklog_len < cwin) |
1135 | return; |
1136 | |
1137 | /* Don't increase window if there are holes the transmit queue */ |
1138 | if (txq_len && l->snd_nxt - buf_seqno(skb: skb_peek(list_: txq)) != txq_len) |
1139 | return; |
1140 | |
1141 | l->cong_acks += released; |
1142 | |
1143 | /* Slow start */ |
1144 | if (cwin <= l->ssthresh) { |
1145 | l->window = min_t(u16, cwin + released, l->max_win); |
1146 | return; |
1147 | } |
1148 | /* Congestion avoidance */ |
1149 | if (l->cong_acks < cwin) |
1150 | return; |
1151 | l->window = min_t(u16, ++cwin, l->max_win); |
1152 | l->cong_acks = 0; |
1153 | } |
1154 | |
1155 | static void tipc_link_advance_backlog(struct tipc_link *l, |
1156 | struct sk_buff_head *xmitq) |
1157 | { |
1158 | u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1; |
1159 | struct sk_buff_head *txq = &l->transmq; |
1160 | struct sk_buff *skb, *_skb; |
1161 | u16 ack = l->rcv_nxt - 1; |
1162 | u16 seqno = l->snd_nxt; |
1163 | struct tipc_msg *hdr; |
1164 | u16 cwin = l->window; |
1165 | u32 imp; |
1166 | |
1167 | while (skb_queue_len(list_: txq) < cwin) { |
1168 | skb = skb_peek(list_: &l->backlogq); |
1169 | if (!skb) |
1170 | break; |
1171 | _skb = skb_clone(skb, GFP_ATOMIC); |
1172 | if (!_skb) |
1173 | break; |
1174 | __skb_dequeue(list: &l->backlogq); |
1175 | hdr = buf_msg(skb); |
1176 | imp = msg_importance(m: hdr); |
1177 | l->backlog[imp].len--; |
1178 | if (unlikely(skb == l->backlog[imp].target_bskb)) |
1179 | l->backlog[imp].target_bskb = NULL; |
1180 | __skb_queue_tail(list: &l->transmq, newsk: skb); |
1181 | tipc_link_set_skb_retransmit_time(skb, l); |
1182 | |
1183 | __skb_queue_tail(list: xmitq, newsk: _skb); |
1184 | TIPC_SKB_CB(skb)->ackers = l->ackers; |
1185 | msg_set_seqno(m: hdr, n: seqno); |
1186 | msg_set_ack(m: hdr, n: ack); |
1187 | msg_set_bcast_ack(m: hdr, n: bc_ack); |
1188 | l->rcv_unacked = 0; |
1189 | l->stats.sent_pkts++; |
1190 | seqno++; |
1191 | } |
1192 | l->snd_nxt = seqno; |
1193 | } |
1194 | |
1195 | /** |
1196 | * link_retransmit_failure() - Detect repeated retransmit failures |
1197 | * @l: tipc link sender |
1198 | * @r: tipc link receiver (= l in case of unicast) |
1199 | * @rc: returned code |
1200 | * |
1201 | * Return: true if the repeated retransmit failures happens, otherwise |
1202 | * false |
1203 | */ |
1204 | static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r, |
1205 | int *rc) |
1206 | { |
1207 | struct sk_buff *skb = skb_peek(list_: &l->transmq); |
1208 | struct tipc_msg *hdr; |
1209 | |
1210 | if (!skb) |
1211 | return false; |
1212 | |
1213 | if (!TIPC_SKB_CB(skb)->retr_cnt) |
1214 | return false; |
1215 | |
1216 | if (!time_after(jiffies, TIPC_SKB_CB(skb)->retr_stamp + |
1217 | msecs_to_jiffies(r->tolerance * 10))) |
1218 | return false; |
1219 | |
1220 | hdr = buf_msg(skb); |
1221 | if (link_is_bc_sndlink(l) && !less(left: r->acked, right: msg_seqno(m: hdr))) |
1222 | return false; |
1223 | |
1224 | pr_warn("Retransmission failure on link <%s>\n" , l->name); |
1225 | link_print(l, str: "State of link " ); |
1226 | pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n" , |
1227 | msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr)); |
1228 | pr_info("sqno %u, prev: %x, dest: %x\n" , |
1229 | msg_seqno(hdr), msg_prevnode(hdr), msg_destnode(hdr)); |
1230 | pr_info("retr_stamp %d, retr_cnt %d\n" , |
1231 | jiffies_to_msecs(TIPC_SKB_CB(skb)->retr_stamp), |
1232 | TIPC_SKB_CB(skb)->retr_cnt); |
1233 | |
1234 | trace_tipc_list_dump(list: &l->transmq, more: true, header: "retrans failure!" ); |
1235 | trace_tipc_link_dump(l, dqueues: TIPC_DUMP_NONE, header: "retrans failure!" ); |
1236 | trace_tipc_link_dump(l: r, dqueues: TIPC_DUMP_NONE, header: "retrans failure!" ); |
1237 | |
1238 | if (link_is_bc_sndlink(l)) { |
1239 | r->state = LINK_RESET; |
1240 | *rc |= TIPC_LINK_DOWN_EVT; |
1241 | } else { |
1242 | *rc |= tipc_link_fsm_evt(l, evt: LINK_FAILURE_EVT); |
1243 | } |
1244 | |
1245 | return true; |
1246 | } |
1247 | |
1248 | /* tipc_data_input - deliver data and name distr msgs to upper layer |
1249 | * |
1250 | * Consumes buffer if message is of right type |
1251 | * Node lock must be held |
1252 | */ |
1253 | static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb, |
1254 | struct sk_buff_head *inputq) |
1255 | { |
1256 | struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq; |
1257 | struct tipc_msg *hdr = buf_msg(skb); |
1258 | |
1259 | switch (msg_user(m: hdr)) { |
1260 | case TIPC_LOW_IMPORTANCE: |
1261 | case TIPC_MEDIUM_IMPORTANCE: |
1262 | case TIPC_HIGH_IMPORTANCE: |
1263 | case TIPC_CRITICAL_IMPORTANCE: |
1264 | if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) { |
1265 | skb_queue_tail(list: mc_inputq, newsk: skb); |
1266 | return true; |
1267 | } |
1268 | fallthrough; |
1269 | case CONN_MANAGER: |
1270 | skb_queue_tail(list: inputq, newsk: skb); |
1271 | return true; |
1272 | case GROUP_PROTOCOL: |
1273 | skb_queue_tail(list: mc_inputq, newsk: skb); |
1274 | return true; |
1275 | case NAME_DISTRIBUTOR: |
1276 | l->bc_rcvlink->state = LINK_ESTABLISHED; |
1277 | skb_queue_tail(list: l->namedq, newsk: skb); |
1278 | return true; |
1279 | case MSG_BUNDLER: |
1280 | case TUNNEL_PROTOCOL: |
1281 | case MSG_FRAGMENTER: |
1282 | case BCAST_PROTOCOL: |
1283 | return false; |
1284 | #ifdef CONFIG_TIPC_CRYPTO |
1285 | case MSG_CRYPTO: |
1286 | if (sysctl_tipc_key_exchange_enabled && |
1287 | TIPC_SKB_CB(skb)->decrypted) { |
1288 | tipc_crypto_msg_rcv(net: l->net, skb); |
1289 | return true; |
1290 | } |
1291 | fallthrough; |
1292 | #endif |
1293 | default: |
1294 | pr_warn("Dropping received illegal msg type\n" ); |
1295 | kfree_skb(skb); |
1296 | return true; |
1297 | } |
1298 | } |
1299 | |
1300 | /* tipc_link_input - process packet that has passed link protocol check |
1301 | * |
1302 | * Consumes buffer |
1303 | */ |
1304 | static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb, |
1305 | struct sk_buff_head *inputq, |
1306 | struct sk_buff **reasm_skb) |
1307 | { |
1308 | struct tipc_msg *hdr = buf_msg(skb); |
1309 | struct sk_buff *iskb; |
1310 | struct sk_buff_head tmpq; |
1311 | int usr = msg_user(m: hdr); |
1312 | int pos = 0; |
1313 | |
1314 | if (usr == MSG_BUNDLER) { |
1315 | skb_queue_head_init(list: &tmpq); |
1316 | l->stats.recv_bundles++; |
1317 | l->stats.recv_bundled += msg_msgcnt(m: hdr); |
1318 | while (tipc_msg_extract(skb, iskb: &iskb, pos: &pos)) |
1319 | tipc_data_input(l, skb: iskb, inputq: &tmpq); |
1320 | tipc_skb_queue_splice_tail(list: &tmpq, head: inputq); |
1321 | return 0; |
1322 | } else if (usr == MSG_FRAGMENTER) { |
1323 | l->stats.recv_fragments++; |
1324 | if (tipc_buf_append(headbuf: reasm_skb, buf: &skb)) { |
1325 | l->stats.recv_fragmented++; |
1326 | tipc_data_input(l, skb, inputq); |
1327 | } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) { |
1328 | pr_warn_ratelimited("Unable to build fragment list\n" ); |
1329 | return tipc_link_fsm_evt(l, evt: LINK_FAILURE_EVT); |
1330 | } |
1331 | return 0; |
1332 | } else if (usr == BCAST_PROTOCOL) { |
1333 | tipc_bcast_lock(net: l->net); |
1334 | tipc_link_bc_init_rcv(l: l->bc_rcvlink, hdr); |
1335 | tipc_bcast_unlock(net: l->net); |
1336 | } |
1337 | |
1338 | kfree_skb(skb); |
1339 | return 0; |
1340 | } |
1341 | |
1342 | /* tipc_link_tnl_rcv() - receive TUNNEL_PROTOCOL message, drop or process the |
1343 | * inner message along with the ones in the old link's |
1344 | * deferdq |
1345 | * @l: tunnel link |
1346 | * @skb: TUNNEL_PROTOCOL message |
1347 | * @inputq: queue to put messages ready for delivery |
1348 | */ |
1349 | static int tipc_link_tnl_rcv(struct tipc_link *l, struct sk_buff *skb, |
1350 | struct sk_buff_head *inputq) |
1351 | { |
1352 | struct sk_buff **reasm_skb = &l->failover_reasm_skb; |
1353 | struct sk_buff **reasm_tnlmsg = &l->reasm_tnlmsg; |
1354 | struct sk_buff_head *fdefq = &l->failover_deferdq; |
1355 | struct tipc_msg *hdr = buf_msg(skb); |
1356 | struct sk_buff *iskb; |
1357 | int ipos = 0; |
1358 | int rc = 0; |
1359 | u16 seqno; |
1360 | |
1361 | if (msg_type(m: hdr) == SYNCH_MSG) { |
1362 | kfree_skb(skb); |
1363 | return 0; |
1364 | } |
1365 | |
1366 | /* Not a fragment? */ |
1367 | if (likely(!msg_nof_fragms(hdr))) { |
1368 | if (unlikely(!tipc_msg_extract(skb, &iskb, &ipos))) { |
1369 | pr_warn_ratelimited("Unable to extract msg, defq: %d\n" , |
1370 | skb_queue_len(fdefq)); |
1371 | return 0; |
1372 | } |
1373 | kfree_skb(skb); |
1374 | } else { |
1375 | /* Set fragment type for buf_append */ |
1376 | if (msg_fragm_no(m: hdr) == 1) |
1377 | msg_set_type(m: hdr, FIRST_FRAGMENT); |
1378 | else if (msg_fragm_no(m: hdr) < msg_nof_fragms(m: hdr)) |
1379 | msg_set_type(m: hdr, FRAGMENT); |
1380 | else |
1381 | msg_set_type(m: hdr, LAST_FRAGMENT); |
1382 | |
1383 | if (!tipc_buf_append(headbuf: reasm_tnlmsg, buf: &skb)) { |
1384 | /* Successful but non-complete reassembly? */ |
1385 | if (*reasm_tnlmsg || link_is_bc_rcvlink(l)) |
1386 | return 0; |
1387 | pr_warn_ratelimited("Unable to reassemble tunnel msg\n" ); |
1388 | return tipc_link_fsm_evt(l, evt: LINK_FAILURE_EVT); |
1389 | } |
1390 | iskb = skb; |
1391 | } |
1392 | |
1393 | do { |
1394 | seqno = buf_seqno(skb: iskb); |
1395 | if (unlikely(less(seqno, l->drop_point))) { |
1396 | kfree_skb(skb: iskb); |
1397 | continue; |
1398 | } |
1399 | if (unlikely(seqno != l->drop_point)) { |
1400 | __tipc_skb_queue_sorted(list: fdefq, seqno, skb: iskb); |
1401 | continue; |
1402 | } |
1403 | |
1404 | l->drop_point++; |
1405 | if (!tipc_data_input(l, skb: iskb, inputq)) |
1406 | rc |= tipc_link_input(l, skb: iskb, inputq, reasm_skb); |
1407 | if (unlikely(rc)) |
1408 | break; |
1409 | } while ((iskb = __tipc_skb_dequeue(list: fdefq, seqno: l->drop_point))); |
1410 | |
1411 | return rc; |
1412 | } |
1413 | |
1414 | /** |
1415 | * tipc_get_gap_ack_blks - get Gap ACK blocks from PROTOCOL/STATE_MSG |
1416 | * @ga: returned pointer to the Gap ACK blocks if any |
1417 | * @l: the tipc link |
1418 | * @hdr: the PROTOCOL/STATE_MSG header |
1419 | * @uc: desired Gap ACK blocks type, i.e. unicast (= 1) or broadcast (= 0) |
1420 | * |
1421 | * Return: the total Gap ACK blocks size |
1422 | */ |
1423 | u16 tipc_get_gap_ack_blks(struct tipc_gap_ack_blks **ga, struct tipc_link *l, |
1424 | struct tipc_msg *hdr, bool uc) |
1425 | { |
1426 | struct tipc_gap_ack_blks *p; |
1427 | u16 sz = 0; |
1428 | |
1429 | /* Does peer support the Gap ACK blocks feature? */ |
1430 | if (l->peer_caps & TIPC_GAP_ACK_BLOCK) { |
1431 | p = (struct tipc_gap_ack_blks *)msg_data(m: hdr); |
1432 | sz = ntohs(p->len); |
1433 | /* Sanity check */ |
1434 | if (sz == struct_size(p, gacks, size_add(p->ugack_cnt, p->bgack_cnt))) { |
1435 | /* Good, check if the desired type exists */ |
1436 | if ((uc && p->ugack_cnt) || (!uc && p->bgack_cnt)) |
1437 | goto ok; |
1438 | /* Backward compatible: peer might not support bc, but uc? */ |
1439 | } else if (uc && sz == struct_size(p, gacks, p->ugack_cnt)) { |
1440 | if (p->ugack_cnt) { |
1441 | p->bgack_cnt = 0; |
1442 | goto ok; |
1443 | } |
1444 | } |
1445 | } |
1446 | /* Other cases: ignore! */ |
1447 | p = NULL; |
1448 | |
1449 | ok: |
1450 | *ga = p; |
1451 | return sz; |
1452 | } |
1453 | |
1454 | static u8 __tipc_build_gap_ack_blks(struct tipc_gap_ack_blks *ga, |
1455 | struct tipc_link *l, u8 start_index) |
1456 | { |
1457 | struct tipc_gap_ack *gacks = &ga->gacks[start_index]; |
1458 | struct sk_buff *skb = skb_peek(list_: &l->deferdq); |
1459 | u16 expect, seqno = 0; |
1460 | u8 n = 0; |
1461 | |
1462 | if (!skb) |
1463 | return 0; |
1464 | |
1465 | expect = buf_seqno(skb); |
1466 | skb_queue_walk(&l->deferdq, skb) { |
1467 | seqno = buf_seqno(skb); |
1468 | if (unlikely(more(seqno, expect))) { |
1469 | gacks[n].ack = htons(expect - 1); |
1470 | gacks[n].gap = htons(seqno - expect); |
1471 | if (++n >= MAX_GAP_ACK_BLKS / 2) { |
1472 | pr_info_ratelimited("Gacks on %s: %d, ql: %d!\n" , |
1473 | l->name, n, |
1474 | skb_queue_len(&l->deferdq)); |
1475 | return n; |
1476 | } |
1477 | } else if (unlikely(less(seqno, expect))) { |
1478 | pr_warn("Unexpected skb in deferdq!\n" ); |
1479 | continue; |
1480 | } |
1481 | expect = seqno + 1; |
1482 | } |
1483 | |
1484 | /* last block */ |
1485 | gacks[n].ack = htons(seqno); |
1486 | gacks[n].gap = 0; |
1487 | n++; |
1488 | return n; |
1489 | } |
1490 | |
1491 | /* tipc_build_gap_ack_blks - build Gap ACK blocks |
1492 | * @l: tipc unicast link |
1493 | * @hdr: the tipc message buffer to store the Gap ACK blocks after built |
1494 | * |
1495 | * The function builds Gap ACK blocks for both the unicast & broadcast receiver |
1496 | * links of a certain peer, the buffer after built has the network data format |
1497 | * as found at the struct tipc_gap_ack_blks definition. |
1498 | * |
1499 | * returns the actual allocated memory size |
1500 | */ |
1501 | static u16 tipc_build_gap_ack_blks(struct tipc_link *l, struct tipc_msg *hdr) |
1502 | { |
1503 | struct tipc_link *bcl = l->bc_rcvlink; |
1504 | struct tipc_gap_ack_blks *ga; |
1505 | u16 len; |
1506 | |
1507 | ga = (struct tipc_gap_ack_blks *)msg_data(m: hdr); |
1508 | |
1509 | /* Start with broadcast link first */ |
1510 | tipc_bcast_lock(net: bcl->net); |
1511 | msg_set_bcast_ack(m: hdr, n: bcl->rcv_nxt - 1); |
1512 | msg_set_bc_gap(m: hdr, n: link_bc_rcv_gap(l: bcl)); |
1513 | ga->bgack_cnt = __tipc_build_gap_ack_blks(ga, l: bcl, start_index: 0); |
1514 | tipc_bcast_unlock(net: bcl->net); |
1515 | |
1516 | /* Now for unicast link, but an explicit NACK only (???) */ |
1517 | ga->ugack_cnt = (msg_seq_gap(m: hdr)) ? |
1518 | __tipc_build_gap_ack_blks(ga, l, start_index: ga->bgack_cnt) : 0; |
1519 | |
1520 | /* Total len */ |
1521 | len = struct_size(ga, gacks, size_add(ga->bgack_cnt, ga->ugack_cnt)); |
1522 | ga->len = htons(len); |
1523 | return len; |
1524 | } |
1525 | |
1526 | /* tipc_link_advance_transmq - advance TIPC link transmq queue by releasing |
1527 | * acked packets, also doing retransmissions if |
1528 | * gaps found |
1529 | * @l: tipc link with transmq queue to be advanced |
1530 | * @r: tipc link "receiver" i.e. in case of broadcast (= "l" if unicast) |
1531 | * @acked: seqno of last packet acked by peer without any gaps before |
1532 | * @gap: # of gap packets |
1533 | * @ga: buffer pointer to Gap ACK blocks from peer |
1534 | * @xmitq: queue for accumulating the retransmitted packets if any |
1535 | * @retransmitted: returned boolean value if a retransmission is really issued |
1536 | * @rc: returned code e.g. TIPC_LINK_DOWN_EVT if a repeated retransmit failures |
1537 | * happens (- unlikely case) |
1538 | * |
1539 | * Return: the number of packets released from the link transmq |
1540 | */ |
1541 | static int tipc_link_advance_transmq(struct tipc_link *l, struct tipc_link *r, |
1542 | u16 acked, u16 gap, |
1543 | struct tipc_gap_ack_blks *ga, |
1544 | struct sk_buff_head *xmitq, |
1545 | bool *retransmitted, int *rc) |
1546 | { |
1547 | struct tipc_gap_ack_blks *last_ga = r->last_ga, *this_ga = NULL; |
1548 | struct tipc_gap_ack *gacks = NULL; |
1549 | struct sk_buff *skb, *_skb, *tmp; |
1550 | struct tipc_msg *hdr; |
1551 | u32 qlen = skb_queue_len(list_: &l->transmq); |
1552 | u16 nacked = acked, ngap = gap, gack_cnt = 0; |
1553 | u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1; |
1554 | u16 ack = l->rcv_nxt - 1; |
1555 | u16 seqno, n = 0; |
1556 | u16 end = r->acked, start = end, offset = r->last_gap; |
1557 | u16 si = (last_ga) ? last_ga->start_index : 0; |
1558 | bool is_uc = !link_is_bc_sndlink(l); |
1559 | bool bc_has_acked = false; |
1560 | |
1561 | trace_tipc_link_retrans(r, f: acked + 1, t: acked + gap, tq: &l->transmq); |
1562 | |
1563 | /* Determine Gap ACK blocks if any for the particular link */ |
1564 | if (ga && is_uc) { |
1565 | /* Get the Gap ACKs, uc part */ |
1566 | gack_cnt = ga->ugack_cnt; |
1567 | gacks = &ga->gacks[ga->bgack_cnt]; |
1568 | } else if (ga) { |
1569 | /* Copy the Gap ACKs, bc part, for later renewal if needed */ |
1570 | this_ga = kmemdup(p: ga, struct_size(ga, gacks, ga->bgack_cnt), |
1571 | GFP_ATOMIC); |
1572 | if (likely(this_ga)) { |
1573 | this_ga->start_index = 0; |
1574 | /* Start with the bc Gap ACKs */ |
1575 | gack_cnt = this_ga->bgack_cnt; |
1576 | gacks = &this_ga->gacks[0]; |
1577 | } else { |
1578 | /* Hmm, we can get in trouble..., simply ignore it */ |
1579 | pr_warn_ratelimited("Ignoring bc Gap ACKs, no memory\n" ); |
1580 | } |
1581 | } |
1582 | |
1583 | /* Advance the link transmq */ |
1584 | skb_queue_walk_safe(&l->transmq, skb, tmp) { |
1585 | seqno = buf_seqno(skb); |
1586 | |
1587 | next_gap_ack: |
1588 | if (less_eq(left: seqno, right: nacked)) { |
1589 | if (is_uc) |
1590 | goto release; |
1591 | /* Skip packets peer has already acked */ |
1592 | if (!more(left: seqno, right: r->acked)) |
1593 | continue; |
1594 | /* Get the next of last Gap ACK blocks */ |
1595 | while (more(left: seqno, right: end)) { |
1596 | if (!last_ga || si >= last_ga->bgack_cnt) |
1597 | break; |
1598 | start = end + offset + 1; |
1599 | end = ntohs(last_ga->gacks[si].ack); |
1600 | offset = ntohs(last_ga->gacks[si].gap); |
1601 | si++; |
1602 | WARN_ONCE(more(start, end) || |
1603 | (!offset && |
1604 | si < last_ga->bgack_cnt) || |
1605 | si > MAX_GAP_ACK_BLKS, |
1606 | "Corrupted Gap ACK: %d %d %d %d %d\n" , |
1607 | start, end, offset, si, |
1608 | last_ga->bgack_cnt); |
1609 | } |
1610 | /* Check against the last Gap ACK block */ |
1611 | if (tipc_in_range(val: seqno, min: start, max: end)) |
1612 | continue; |
1613 | /* Update/release the packet peer is acking */ |
1614 | bc_has_acked = true; |
1615 | if (--TIPC_SKB_CB(skb)->ackers) |
1616 | continue; |
1617 | release: |
1618 | /* release skb */ |
1619 | __skb_unlink(skb, list: &l->transmq); |
1620 | kfree_skb(skb); |
1621 | } else if (less_eq(left: seqno, right: nacked + ngap)) { |
1622 | /* First gap: check if repeated retrans failures? */ |
1623 | if (unlikely(seqno == acked + 1 && |
1624 | link_retransmit_failure(l, r, rc))) { |
1625 | /* Ignore this bc Gap ACKs if any */ |
1626 | kfree(objp: this_ga); |
1627 | this_ga = NULL; |
1628 | break; |
1629 | } |
1630 | /* retransmit skb if unrestricted*/ |
1631 | if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr)) |
1632 | continue; |
1633 | tipc_link_set_skb_retransmit_time(skb, l); |
1634 | _skb = pskb_copy(skb, GFP_ATOMIC); |
1635 | if (!_skb) |
1636 | continue; |
1637 | hdr = buf_msg(skb: _skb); |
1638 | msg_set_ack(m: hdr, n: ack); |
1639 | msg_set_bcast_ack(m: hdr, n: bc_ack); |
1640 | _skb->priority = TC_PRIO_CONTROL; |
1641 | __skb_queue_tail(list: xmitq, newsk: _skb); |
1642 | l->stats.retransmitted++; |
1643 | if (!is_uc) |
1644 | r->stats.retransmitted++; |
1645 | *retransmitted = true; |
1646 | /* Increase actual retrans counter & mark first time */ |
1647 | if (!TIPC_SKB_CB(skb)->retr_cnt++) |
1648 | TIPC_SKB_CB(skb)->retr_stamp = jiffies; |
1649 | } else { |
1650 | /* retry with Gap ACK blocks if any */ |
1651 | if (n >= gack_cnt) |
1652 | break; |
1653 | nacked = ntohs(gacks[n].ack); |
1654 | ngap = ntohs(gacks[n].gap); |
1655 | n++; |
1656 | goto next_gap_ack; |
1657 | } |
1658 | } |
1659 | |
1660 | /* Renew last Gap ACK blocks for bc if needed */ |
1661 | if (bc_has_acked) { |
1662 | if (this_ga) { |
1663 | kfree(objp: last_ga); |
1664 | r->last_ga = this_ga; |
1665 | r->last_gap = gap; |
1666 | } else if (last_ga) { |
1667 | if (less(left: acked, right: start)) { |
1668 | si--; |
1669 | offset = start - acked - 1; |
1670 | } else if (less(left: acked, right: end)) { |
1671 | acked = end; |
1672 | } |
1673 | if (si < last_ga->bgack_cnt) { |
1674 | last_ga->start_index = si; |
1675 | r->last_gap = offset; |
1676 | } else { |
1677 | kfree(objp: last_ga); |
1678 | r->last_ga = NULL; |
1679 | r->last_gap = 0; |
1680 | } |
1681 | } else { |
1682 | r->last_gap = 0; |
1683 | } |
1684 | r->acked = acked; |
1685 | } else { |
1686 | kfree(objp: this_ga); |
1687 | } |
1688 | |
1689 | return qlen - skb_queue_len(list_: &l->transmq); |
1690 | } |
1691 | |
1692 | /* tipc_link_build_state_msg: prepare link state message for transmission |
1693 | * |
1694 | * Note that sending of broadcast ack is coordinated among nodes, to reduce |
1695 | * risk of ack storms towards the sender |
1696 | */ |
1697 | int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq) |
1698 | { |
1699 | if (!l) |
1700 | return 0; |
1701 | |
1702 | /* Broadcast ACK must be sent via a unicast link => defer to caller */ |
1703 | if (link_is_bc_rcvlink(l)) { |
1704 | if (((l->rcv_nxt ^ tipc_own_addr(net: l->net)) & 0xf) != 0xf) |
1705 | return 0; |
1706 | l->rcv_unacked = 0; |
1707 | |
1708 | /* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */ |
1709 | l->snd_nxt = l->rcv_nxt; |
1710 | return TIPC_LINK_SND_STATE; |
1711 | } |
1712 | /* Unicast ACK */ |
1713 | l->rcv_unacked = 0; |
1714 | l->stats.sent_acks++; |
1715 | tipc_link_build_proto_msg(l, STATE_MSG, probe: 0, probe_reply: 0, rcvgap: 0, tolerance: 0, priority: 0, xmitq); |
1716 | return 0; |
1717 | } |
1718 | |
1719 | /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message |
1720 | */ |
1721 | void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq) |
1722 | { |
1723 | int mtyp = RESET_MSG; |
1724 | struct sk_buff *skb; |
1725 | |
1726 | if (l->state == LINK_ESTABLISHING) |
1727 | mtyp = ACTIVATE_MSG; |
1728 | |
1729 | tipc_link_build_proto_msg(l, mtyp, probe: 0, probe_reply: 0, rcvgap: 0, tolerance: 0, priority: 0, xmitq); |
1730 | |
1731 | /* Inform peer that this endpoint is going down if applicable */ |
1732 | skb = skb_peek_tail(list_: xmitq); |
1733 | if (skb && (l->state == LINK_RESET)) |
1734 | msg_set_peer_stopping(m: buf_msg(skb), s: 1); |
1735 | } |
1736 | |
1737 | /* tipc_link_build_nack_msg: prepare link nack message for transmission |
1738 | * Note that sending of broadcast NACK is coordinated among nodes, to |
1739 | * reduce the risk of NACK storms towards the sender |
1740 | */ |
1741 | static int tipc_link_build_nack_msg(struct tipc_link *l, |
1742 | struct sk_buff_head *xmitq) |
1743 | { |
1744 | u32 def_cnt = ++l->stats.deferred_recv; |
1745 | struct sk_buff_head *dfq = &l->deferdq; |
1746 | u32 defq_len = skb_queue_len(list_: dfq); |
1747 | int match1, match2; |
1748 | |
1749 | if (link_is_bc_rcvlink(l)) { |
1750 | match1 = def_cnt & 0xf; |
1751 | match2 = tipc_own_addr(net: l->net) & 0xf; |
1752 | if (match1 == match2) |
1753 | return TIPC_LINK_SND_STATE; |
1754 | return 0; |
1755 | } |
1756 | |
1757 | if (defq_len >= 3 && !((defq_len - 3) % 16)) { |
1758 | u16 rcvgap = buf_seqno(skb: skb_peek(list_: dfq)) - l->rcv_nxt; |
1759 | |
1760 | tipc_link_build_proto_msg(l, STATE_MSG, probe: 0, probe_reply: 0, |
1761 | rcvgap, tolerance: 0, priority: 0, xmitq); |
1762 | } |
1763 | return 0; |
1764 | } |
1765 | |
1766 | /* tipc_link_rcv - process TIPC packets/messages arriving from off-node |
1767 | * @l: the link that should handle the message |
1768 | * @skb: TIPC packet |
1769 | * @xmitq: queue to place packets to be sent after this call |
1770 | */ |
1771 | int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb, |
1772 | struct sk_buff_head *xmitq) |
1773 | { |
1774 | struct sk_buff_head *defq = &l->deferdq; |
1775 | struct tipc_msg *hdr = buf_msg(skb); |
1776 | u16 seqno, rcv_nxt, win_lim; |
1777 | int released = 0; |
1778 | int rc = 0; |
1779 | |
1780 | /* Verify and update link state */ |
1781 | if (unlikely(msg_user(hdr) == LINK_PROTOCOL)) |
1782 | return tipc_link_proto_rcv(l, skb, xmitq); |
1783 | |
1784 | /* Don't send probe at next timeout expiration */ |
1785 | l->silent_intv_cnt = 0; |
1786 | |
1787 | do { |
1788 | hdr = buf_msg(skb); |
1789 | seqno = msg_seqno(m: hdr); |
1790 | rcv_nxt = l->rcv_nxt; |
1791 | win_lim = rcv_nxt + TIPC_MAX_LINK_WIN; |
1792 | |
1793 | if (unlikely(!link_is_up(l))) { |
1794 | if (l->state == LINK_ESTABLISHING) |
1795 | rc = TIPC_LINK_UP_EVT; |
1796 | kfree_skb(skb); |
1797 | break; |
1798 | } |
1799 | |
1800 | /* Drop if outside receive window */ |
1801 | if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) { |
1802 | l->stats.duplicates++; |
1803 | kfree_skb(skb); |
1804 | break; |
1805 | } |
1806 | released += tipc_link_advance_transmq(l, r: l, acked: msg_ack(m: hdr), gap: 0, |
1807 | NULL, NULL, NULL, NULL); |
1808 | |
1809 | /* Defer delivery if sequence gap */ |
1810 | if (unlikely(seqno != rcv_nxt)) { |
1811 | if (!__tipc_skb_queue_sorted(list: defq, seqno, skb)) |
1812 | l->stats.duplicates++; |
1813 | rc |= tipc_link_build_nack_msg(l, xmitq); |
1814 | break; |
1815 | } |
1816 | |
1817 | /* Deliver packet */ |
1818 | l->rcv_nxt++; |
1819 | l->stats.recv_pkts++; |
1820 | |
1821 | if (unlikely(msg_user(hdr) == TUNNEL_PROTOCOL)) |
1822 | rc |= tipc_link_tnl_rcv(l, skb, inputq: l->inputq); |
1823 | else if (!tipc_data_input(l, skb, inputq: l->inputq)) |
1824 | rc |= tipc_link_input(l, skb, inputq: l->inputq, reasm_skb: &l->reasm_buf); |
1825 | if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN)) |
1826 | rc |= tipc_link_build_state_msg(l, xmitq); |
1827 | if (unlikely(rc & ~TIPC_LINK_SND_STATE)) |
1828 | break; |
1829 | } while ((skb = __tipc_skb_dequeue(list: defq, seqno: l->rcv_nxt))); |
1830 | |
1831 | /* Forward queues and wake up waiting users */ |
1832 | if (released) { |
1833 | tipc_link_update_cwin(l, released, retransmitted: 0); |
1834 | tipc_link_advance_backlog(l, xmitq); |
1835 | if (unlikely(!skb_queue_empty(&l->wakeupq))) |
1836 | link_prepare_wakeup(l); |
1837 | } |
1838 | return rc; |
1839 | } |
1840 | |
1841 | static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe, |
1842 | bool probe_reply, u16 rcvgap, |
1843 | int tolerance, int priority, |
1844 | struct sk_buff_head *xmitq) |
1845 | { |
1846 | struct tipc_mon_state *mstate = &l->mon_state; |
1847 | struct sk_buff_head *dfq = &l->deferdq; |
1848 | struct tipc_link *bcl = l->bc_rcvlink; |
1849 | struct tipc_msg *hdr; |
1850 | struct sk_buff *skb; |
1851 | bool node_up = link_is_up(l: bcl); |
1852 | u16 glen = 0, bc_rcvgap = 0; |
1853 | int dlen = 0; |
1854 | void *data; |
1855 | |
1856 | /* Don't send protocol message during reset or link failover */ |
1857 | if (tipc_link_is_blocked(l)) |
1858 | return; |
1859 | |
1860 | if (!tipc_link_is_up(l) && (mtyp == STATE_MSG)) |
1861 | return; |
1862 | |
1863 | if ((probe || probe_reply) && !skb_queue_empty(list: dfq)) |
1864 | rcvgap = buf_seqno(skb: skb_peek(list_: dfq)) - l->rcv_nxt; |
1865 | |
1866 | skb = tipc_msg_create(LINK_PROTOCOL, type: mtyp, INT_H_SIZE, |
1867 | data_sz: tipc_max_domain_size + MAX_GAP_ACK_BLKS_SZ, |
1868 | dnode: l->addr, onode: tipc_own_addr(net: l->net), dport: 0, oport: 0, errcode: 0); |
1869 | if (!skb) |
1870 | return; |
1871 | |
1872 | hdr = buf_msg(skb); |
1873 | data = msg_data(m: hdr); |
1874 | msg_set_session(m: hdr, n: l->session); |
1875 | msg_set_bearer_id(m: hdr, n: l->bearer_id); |
1876 | msg_set_net_plane(m: hdr, n: l->net_plane); |
1877 | msg_set_next_sent(m: hdr, n: l->snd_nxt); |
1878 | msg_set_ack(m: hdr, n: l->rcv_nxt - 1); |
1879 | msg_set_bcast_ack(m: hdr, n: bcl->rcv_nxt - 1); |
1880 | msg_set_bc_ack_invalid(m: hdr, invalid: !node_up); |
1881 | msg_set_last_bcast(m: hdr, n: l->bc_sndlink->snd_nxt - 1); |
1882 | msg_set_link_tolerance(m: hdr, n: tolerance); |
1883 | msg_set_linkprio(m: hdr, n: priority); |
1884 | msg_set_redundant_link(m: hdr, r: node_up); |
1885 | msg_set_seq_gap(m: hdr, n: 0); |
1886 | msg_set_seqno(m: hdr, n: l->snd_nxt + U16_MAX / 2); |
1887 | |
1888 | if (mtyp == STATE_MSG) { |
1889 | if (l->peer_caps & TIPC_LINK_PROTO_SEQNO) |
1890 | msg_set_seqno(m: hdr, n: l->snd_nxt_state++); |
1891 | msg_set_seq_gap(m: hdr, n: rcvgap); |
1892 | bc_rcvgap = link_bc_rcv_gap(l: bcl); |
1893 | msg_set_bc_gap(m: hdr, n: bc_rcvgap); |
1894 | msg_set_probe(m: hdr, val: probe); |
1895 | msg_set_is_keepalive(m: hdr, d: probe || probe_reply); |
1896 | if (l->peer_caps & TIPC_GAP_ACK_BLOCK) |
1897 | glen = tipc_build_gap_ack_blks(l, hdr); |
1898 | tipc_mon_prep(net: l->net, data: data + glen, dlen: &dlen, state: mstate, bearer_id: l->bearer_id); |
1899 | msg_set_size(m: hdr, INT_H_SIZE + glen + dlen); |
1900 | skb_trim(skb, INT_H_SIZE + glen + dlen); |
1901 | l->stats.sent_states++; |
1902 | l->rcv_unacked = 0; |
1903 | } else { |
1904 | /* RESET_MSG or ACTIVATE_MSG */ |
1905 | if (mtyp == ACTIVATE_MSG) { |
1906 | msg_set_dest_session_valid(m: hdr, valid: 1); |
1907 | msg_set_dest_session(m: hdr, n: l->peer_session); |
1908 | } |
1909 | msg_set_max_pkt(m: hdr, n: l->advertised_mtu); |
1910 | strcpy(p: data, q: l->if_name); |
1911 | msg_set_size(m: hdr, INT_H_SIZE + TIPC_MAX_IF_NAME); |
1912 | skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME); |
1913 | } |
1914 | if (probe) |
1915 | l->stats.sent_probes++; |
1916 | if (rcvgap) |
1917 | l->stats.sent_nacks++; |
1918 | if (bc_rcvgap) |
1919 | bcl->stats.sent_nacks++; |
1920 | skb->priority = TC_PRIO_CONTROL; |
1921 | __skb_queue_tail(list: xmitq, newsk: skb); |
1922 | trace_tipc_proto_build(skb, more: false, header: l->name); |
1923 | } |
1924 | |
1925 | void tipc_link_create_dummy_tnl_msg(struct tipc_link *l, |
1926 | struct sk_buff_head *xmitq) |
1927 | { |
1928 | u32 onode = tipc_own_addr(net: l->net); |
1929 | struct tipc_msg *hdr, *ihdr; |
1930 | struct sk_buff_head tnlq; |
1931 | struct sk_buff *skb; |
1932 | u32 dnode = l->addr; |
1933 | |
1934 | __skb_queue_head_init(list: &tnlq); |
1935 | skb = tipc_msg_create(TUNNEL_PROTOCOL, FAILOVER_MSG, |
1936 | INT_H_SIZE, BASIC_H_SIZE, |
1937 | dnode, onode, dport: 0, oport: 0, errcode: 0); |
1938 | if (!skb) { |
1939 | pr_warn("%sunable to create tunnel packet\n" , link_co_err); |
1940 | return; |
1941 | } |
1942 | |
1943 | hdr = buf_msg(skb); |
1944 | msg_set_msgcnt(m: hdr, n: 1); |
1945 | msg_set_bearer_id(m: hdr, n: l->peer_bearer_id); |
1946 | |
1947 | ihdr = (struct tipc_msg *)msg_data(m: hdr); |
1948 | tipc_msg_init(own_addr: onode, m: ihdr, TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG, |
1949 | BASIC_H_SIZE, destnode: dnode); |
1950 | msg_set_errcode(m: ihdr, TIPC_ERR_NO_PORT); |
1951 | __skb_queue_tail(list: &tnlq, newsk: skb); |
1952 | tipc_link_xmit(l, list: &tnlq, xmitq); |
1953 | } |
1954 | |
1955 | /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets |
1956 | * with contents of the link's transmit and backlog queues. |
1957 | */ |
1958 | void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl, |
1959 | int mtyp, struct sk_buff_head *xmitq) |
1960 | { |
1961 | struct sk_buff_head *fdefq = &tnl->failover_deferdq; |
1962 | struct sk_buff *skb, *tnlskb; |
1963 | struct tipc_msg *hdr, tnlhdr; |
1964 | struct sk_buff_head *queue = &l->transmq; |
1965 | struct sk_buff_head tmpxq, tnlq, frags; |
1966 | u16 pktlen, pktcnt, seqno = l->snd_nxt; |
1967 | bool pktcnt_need_update = false; |
1968 | u16 syncpt; |
1969 | int rc; |
1970 | |
1971 | if (!tnl) |
1972 | return; |
1973 | |
1974 | __skb_queue_head_init(list: &tnlq); |
1975 | /* Link Synching: |
1976 | * From now on, send only one single ("dummy") SYNCH message |
1977 | * to peer. The SYNCH message does not contain any data, just |
1978 | * a header conveying the synch point to the peer. |
1979 | */ |
1980 | if (mtyp == SYNCH_MSG && (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) { |
1981 | tnlskb = tipc_msg_create(TUNNEL_PROTOCOL, SYNCH_MSG, |
1982 | INT_H_SIZE, data_sz: 0, dnode: l->addr, |
1983 | onode: tipc_own_addr(net: l->net), |
1984 | dport: 0, oport: 0, errcode: 0); |
1985 | if (!tnlskb) { |
1986 | pr_warn("%sunable to create dummy SYNCH_MSG\n" , |
1987 | link_co_err); |
1988 | return; |
1989 | } |
1990 | |
1991 | hdr = buf_msg(skb: tnlskb); |
1992 | syncpt = l->snd_nxt + skb_queue_len(list_: &l->backlogq) - 1; |
1993 | msg_set_syncpt(m: hdr, n: syncpt); |
1994 | msg_set_bearer_id(m: hdr, n: l->peer_bearer_id); |
1995 | __skb_queue_tail(list: &tnlq, newsk: tnlskb); |
1996 | tipc_link_xmit(l: tnl, list: &tnlq, xmitq); |
1997 | return; |
1998 | } |
1999 | |
2000 | __skb_queue_head_init(list: &tmpxq); |
2001 | __skb_queue_head_init(list: &frags); |
2002 | /* At least one packet required for safe algorithm => add dummy */ |
2003 | skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG, |
2004 | BASIC_H_SIZE, data_sz: 0, dnode: l->addr, onode: tipc_own_addr(net: l->net), |
2005 | dport: 0, oport: 0, TIPC_ERR_NO_PORT); |
2006 | if (!skb) { |
2007 | pr_warn("%sunable to create tunnel packet\n" , link_co_err); |
2008 | return; |
2009 | } |
2010 | __skb_queue_tail(list: &tnlq, newsk: skb); |
2011 | tipc_link_xmit(l, list: &tnlq, xmitq: &tmpxq); |
2012 | __skb_queue_purge(list: &tmpxq); |
2013 | |
2014 | /* Initialize reusable tunnel packet header */ |
2015 | tipc_msg_init(own_addr: tipc_own_addr(net: l->net), m: &tnlhdr, TUNNEL_PROTOCOL, |
2016 | type: mtyp, INT_H_SIZE, destnode: l->addr); |
2017 | if (mtyp == SYNCH_MSG) |
2018 | pktcnt = l->snd_nxt - buf_seqno(skb: skb_peek(list_: &l->transmq)); |
2019 | else |
2020 | pktcnt = skb_queue_len(list_: &l->transmq); |
2021 | pktcnt += skb_queue_len(list_: &l->backlogq); |
2022 | msg_set_msgcnt(m: &tnlhdr, n: pktcnt); |
2023 | msg_set_bearer_id(m: &tnlhdr, n: l->peer_bearer_id); |
2024 | tnl: |
2025 | /* Wrap each packet into a tunnel packet */ |
2026 | skb_queue_walk(queue, skb) { |
2027 | hdr = buf_msg(skb); |
2028 | if (queue == &l->backlogq) |
2029 | msg_set_seqno(m: hdr, n: seqno++); |
2030 | pktlen = msg_size(m: hdr); |
2031 | |
2032 | /* Tunnel link MTU is not large enough? This could be |
2033 | * due to: |
2034 | * 1) Link MTU has just changed or set differently; |
2035 | * 2) Or FAILOVER on the top of a SYNCH message |
2036 | * |
2037 | * The 2nd case should not happen if peer supports |
2038 | * TIPC_TUNNEL_ENHANCED |
2039 | */ |
2040 | if (pktlen > tnl->mtu - INT_H_SIZE) { |
2041 | if (mtyp == FAILOVER_MSG && |
2042 | (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) { |
2043 | rc = tipc_msg_fragment(skb, hdr: &tnlhdr, pktmax: tnl->mtu, |
2044 | frags: &frags); |
2045 | if (rc) { |
2046 | pr_warn("%sunable to frag msg: rc %d\n" , |
2047 | link_co_err, rc); |
2048 | return; |
2049 | } |
2050 | pktcnt += skb_queue_len(list_: &frags) - 1; |
2051 | pktcnt_need_update = true; |
2052 | skb_queue_splice_tail_init(list: &frags, head: &tnlq); |
2053 | continue; |
2054 | } |
2055 | /* Unluckily, peer doesn't have TIPC_TUNNEL_ENHANCED |
2056 | * => Just warn it and return! |
2057 | */ |
2058 | pr_warn_ratelimited("%stoo large msg <%d, %d>: %d!\n" , |
2059 | link_co_err, msg_user(hdr), |
2060 | msg_type(hdr), msg_size(hdr)); |
2061 | return; |
2062 | } |
2063 | |
2064 | msg_set_size(m: &tnlhdr, sz: pktlen + INT_H_SIZE); |
2065 | tnlskb = tipc_buf_acquire(size: pktlen + INT_H_SIZE, GFP_ATOMIC); |
2066 | if (!tnlskb) { |
2067 | pr_warn("%sunable to send packet\n" , link_co_err); |
2068 | return; |
2069 | } |
2070 | skb_copy_to_linear_data(skb: tnlskb, from: &tnlhdr, INT_H_SIZE); |
2071 | skb_copy_to_linear_data_offset(skb: tnlskb, INT_H_SIZE, from: hdr, len: pktlen); |
2072 | __skb_queue_tail(list: &tnlq, newsk: tnlskb); |
2073 | } |
2074 | if (queue != &l->backlogq) { |
2075 | queue = &l->backlogq; |
2076 | goto tnl; |
2077 | } |
2078 | |
2079 | if (pktcnt_need_update) |
2080 | skb_queue_walk(&tnlq, skb) { |
2081 | hdr = buf_msg(skb); |
2082 | msg_set_msgcnt(m: hdr, n: pktcnt); |
2083 | } |
2084 | |
2085 | tipc_link_xmit(l: tnl, list: &tnlq, xmitq); |
2086 | |
2087 | if (mtyp == FAILOVER_MSG) { |
2088 | tnl->drop_point = l->rcv_nxt; |
2089 | tnl->failover_reasm_skb = l->reasm_buf; |
2090 | l->reasm_buf = NULL; |
2091 | |
2092 | /* Failover the link's deferdq */ |
2093 | if (unlikely(!skb_queue_empty(fdefq))) { |
2094 | pr_warn("Link failover deferdq not empty: %d!\n" , |
2095 | skb_queue_len(fdefq)); |
2096 | __skb_queue_purge(list: fdefq); |
2097 | } |
2098 | skb_queue_splice_init(list: &l->deferdq, head: fdefq); |
2099 | } |
2100 | } |
2101 | |
2102 | /** |
2103 | * tipc_link_failover_prepare() - prepare tnl for link failover |
2104 | * |
2105 | * This is a special version of the precursor - tipc_link_tnl_prepare(), |
2106 | * see the tipc_node_link_failover() for details |
2107 | * |
2108 | * @l: failover link |
2109 | * @tnl: tunnel link |
2110 | * @xmitq: queue for messages to be xmited |
2111 | */ |
2112 | void tipc_link_failover_prepare(struct tipc_link *l, struct tipc_link *tnl, |
2113 | struct sk_buff_head *xmitq) |
2114 | { |
2115 | struct sk_buff_head *fdefq = &tnl->failover_deferdq; |
2116 | |
2117 | tipc_link_create_dummy_tnl_msg(l: tnl, xmitq); |
2118 | |
2119 | /* This failover link endpoint was never established before, |
2120 | * so it has not received anything from peer. |
2121 | * Otherwise, it must be a normal failover situation or the |
2122 | * node has entered SELF_DOWN_PEER_LEAVING and both peer nodes |
2123 | * would have to start over from scratch instead. |
2124 | */ |
2125 | tnl->drop_point = 1; |
2126 | tnl->failover_reasm_skb = NULL; |
2127 | |
2128 | /* Initiate the link's failover deferdq */ |
2129 | if (unlikely(!skb_queue_empty(fdefq))) { |
2130 | pr_warn("Link failover deferdq not empty: %d!\n" , |
2131 | skb_queue_len(fdefq)); |
2132 | __skb_queue_purge(list: fdefq); |
2133 | } |
2134 | } |
2135 | |
2136 | /* tipc_link_validate_msg(): validate message against current link state |
2137 | * Returns true if message should be accepted, otherwise false |
2138 | */ |
2139 | bool tipc_link_validate_msg(struct tipc_link *l, struct tipc_msg *hdr) |
2140 | { |
2141 | u16 curr_session = l->peer_session; |
2142 | u16 session = msg_session(m: hdr); |
2143 | int mtyp = msg_type(m: hdr); |
2144 | |
2145 | if (msg_user(m: hdr) != LINK_PROTOCOL) |
2146 | return true; |
2147 | |
2148 | switch (mtyp) { |
2149 | case RESET_MSG: |
2150 | if (!l->in_session) |
2151 | return true; |
2152 | /* Accept only RESET with new session number */ |
2153 | return more(left: session, right: curr_session); |
2154 | case ACTIVATE_MSG: |
2155 | if (!l->in_session) |
2156 | return true; |
2157 | /* Accept only ACTIVATE with new or current session number */ |
2158 | return !less(left: session, right: curr_session); |
2159 | case STATE_MSG: |
2160 | /* Accept only STATE with current session number */ |
2161 | if (!l->in_session) |
2162 | return false; |
2163 | if (session != curr_session) |
2164 | return false; |
2165 | /* Extra sanity check */ |
2166 | if (!link_is_up(l) && msg_ack(m: hdr)) |
2167 | return false; |
2168 | if (!(l->peer_caps & TIPC_LINK_PROTO_SEQNO)) |
2169 | return true; |
2170 | /* Accept only STATE with new sequence number */ |
2171 | return !less(left: msg_seqno(m: hdr), right: l->rcv_nxt_state); |
2172 | default: |
2173 | return false; |
2174 | } |
2175 | } |
2176 | |
2177 | /* tipc_link_proto_rcv(): receive link level protocol message : |
2178 | * Note that network plane id propagates through the network, and may |
2179 | * change at any time. The node with lowest numerical id determines |
2180 | * network plane |
2181 | */ |
2182 | static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb, |
2183 | struct sk_buff_head *xmitq) |
2184 | { |
2185 | struct tipc_msg *hdr = buf_msg(skb); |
2186 | struct tipc_gap_ack_blks *ga = NULL; |
2187 | bool reply = msg_probe(m: hdr), retransmitted = false; |
2188 | u32 dlen = msg_data_sz(m: hdr), glen = 0, msg_max; |
2189 | u16 peers_snd_nxt = msg_next_sent(m: hdr); |
2190 | u16 peers_tol = msg_link_tolerance(m: hdr); |
2191 | u16 peers_prio = msg_linkprio(m: hdr); |
2192 | u16 gap = msg_seq_gap(m: hdr); |
2193 | u16 ack = msg_ack(m: hdr); |
2194 | u16 rcv_nxt = l->rcv_nxt; |
2195 | u16 rcvgap = 0; |
2196 | int mtyp = msg_type(m: hdr); |
2197 | int rc = 0, released; |
2198 | char *if_name; |
2199 | void *data; |
2200 | |
2201 | trace_tipc_proto_rcv(skb, more: false, header: l->name); |
2202 | |
2203 | if (dlen > U16_MAX) |
2204 | goto exit; |
2205 | |
2206 | if (tipc_link_is_blocked(l) || !xmitq) |
2207 | goto exit; |
2208 | |
2209 | if (tipc_own_addr(net: l->net) > msg_prevnode(m: hdr)) |
2210 | l->net_plane = msg_net_plane(m: hdr); |
2211 | |
2212 | if (skb_linearize(skb)) |
2213 | goto exit; |
2214 | |
2215 | hdr = buf_msg(skb); |
2216 | data = msg_data(m: hdr); |
2217 | |
2218 | if (!tipc_link_validate_msg(l, hdr)) { |
2219 | trace_tipc_skb_dump(skb, more: false, header: "PROTO invalid (1)!" ); |
2220 | trace_tipc_link_dump(l, dqueues: TIPC_DUMP_NONE, header: "PROTO invalid (1)!" ); |
2221 | goto exit; |
2222 | } |
2223 | |
2224 | switch (mtyp) { |
2225 | case RESET_MSG: |
2226 | case ACTIVATE_MSG: |
2227 | msg_max = msg_max_pkt(m: hdr); |
2228 | if (msg_max < tipc_bearer_min_mtu(net: l->net, bearer_id: l->bearer_id)) |
2229 | break; |
2230 | /* Complete own link name with peer's interface name */ |
2231 | if_name = strrchr(l->name, ':') + 1; |
2232 | if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME) |
2233 | break; |
2234 | if (msg_data_sz(m: hdr) < TIPC_MAX_IF_NAME) |
2235 | break; |
2236 | strncpy(p: if_name, q: data, TIPC_MAX_IF_NAME); |
2237 | |
2238 | /* Update own tolerance if peer indicates a non-zero value */ |
2239 | if (tipc_in_range(val: peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) { |
2240 | l->tolerance = peers_tol; |
2241 | l->bc_rcvlink->tolerance = peers_tol; |
2242 | } |
2243 | /* Update own priority if peer's priority is higher */ |
2244 | if (tipc_in_range(val: peers_prio, min: l->priority + 1, TIPC_MAX_LINK_PRI)) |
2245 | l->priority = peers_prio; |
2246 | |
2247 | /* If peer is going down we want full re-establish cycle */ |
2248 | if (msg_peer_stopping(m: hdr)) { |
2249 | rc = tipc_link_fsm_evt(l, evt: LINK_FAILURE_EVT); |
2250 | break; |
2251 | } |
2252 | |
2253 | /* If this endpoint was re-created while peer was ESTABLISHING |
2254 | * it doesn't know current session number. Force re-synch. |
2255 | */ |
2256 | if (mtyp == ACTIVATE_MSG && msg_dest_session_valid(m: hdr) && |
2257 | l->session != msg_dest_session(m: hdr)) { |
2258 | if (less(left: l->session, right: msg_dest_session(m: hdr))) |
2259 | l->session = msg_dest_session(m: hdr) + 1; |
2260 | break; |
2261 | } |
2262 | |
2263 | /* ACTIVATE_MSG serves as PEER_RESET if link is already down */ |
2264 | if (mtyp == RESET_MSG || !link_is_up(l)) |
2265 | rc = tipc_link_fsm_evt(l, evt: LINK_PEER_RESET_EVT); |
2266 | |
2267 | /* ACTIVATE_MSG takes up link if it was already locally reset */ |
2268 | if (mtyp == ACTIVATE_MSG && l->state == LINK_ESTABLISHING) |
2269 | rc = TIPC_LINK_UP_EVT; |
2270 | |
2271 | l->peer_session = msg_session(m: hdr); |
2272 | l->in_session = true; |
2273 | l->peer_bearer_id = msg_bearer_id(m: hdr); |
2274 | if (l->mtu > msg_max) |
2275 | l->mtu = msg_max; |
2276 | break; |
2277 | |
2278 | case STATE_MSG: |
2279 | /* Validate Gap ACK blocks, drop if invalid */ |
2280 | glen = tipc_get_gap_ack_blks(ga: &ga, l, hdr, uc: true); |
2281 | if (glen > dlen) |
2282 | break; |
2283 | |
2284 | l->rcv_nxt_state = msg_seqno(m: hdr) + 1; |
2285 | |
2286 | /* Update own tolerance if peer indicates a non-zero value */ |
2287 | if (tipc_in_range(val: peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) { |
2288 | l->tolerance = peers_tol; |
2289 | l->bc_rcvlink->tolerance = peers_tol; |
2290 | } |
2291 | /* Update own prio if peer indicates a different value */ |
2292 | if ((peers_prio != l->priority) && |
2293 | tipc_in_range(val: peers_prio, min: 1, TIPC_MAX_LINK_PRI)) { |
2294 | l->priority = peers_prio; |
2295 | rc = tipc_link_fsm_evt(l, evt: LINK_FAILURE_EVT); |
2296 | } |
2297 | |
2298 | l->silent_intv_cnt = 0; |
2299 | l->stats.recv_states++; |
2300 | if (msg_probe(m: hdr)) |
2301 | l->stats.recv_probes++; |
2302 | |
2303 | if (!link_is_up(l)) { |
2304 | if (l->state == LINK_ESTABLISHING) |
2305 | rc = TIPC_LINK_UP_EVT; |
2306 | break; |
2307 | } |
2308 | |
2309 | tipc_mon_rcv(net: l->net, data: data + glen, dlen: dlen - glen, addr: l->addr, |
2310 | state: &l->mon_state, bearer_id: l->bearer_id); |
2311 | |
2312 | /* Send NACK if peer has sent pkts we haven't received yet */ |
2313 | if ((reply || msg_is_keepalive(m: hdr)) && |
2314 | more(left: peers_snd_nxt, right: rcv_nxt) && |
2315 | !tipc_link_is_synching(l) && |
2316 | skb_queue_empty(list: &l->deferdq)) |
2317 | rcvgap = peers_snd_nxt - l->rcv_nxt; |
2318 | if (rcvgap || reply) |
2319 | tipc_link_build_proto_msg(l, STATE_MSG, probe: 0, probe_reply: reply, |
2320 | rcvgap, tolerance: 0, priority: 0, xmitq); |
2321 | |
2322 | released = tipc_link_advance_transmq(l, r: l, acked: ack, gap, ga, xmitq, |
2323 | retransmitted: &retransmitted, rc: &rc); |
2324 | if (gap) |
2325 | l->stats.recv_nacks++; |
2326 | if (released || retransmitted) |
2327 | tipc_link_update_cwin(l, released, retransmitted); |
2328 | if (released) |
2329 | tipc_link_advance_backlog(l, xmitq); |
2330 | if (unlikely(!skb_queue_empty(&l->wakeupq))) |
2331 | link_prepare_wakeup(l); |
2332 | } |
2333 | exit: |
2334 | kfree_skb(skb); |
2335 | return rc; |
2336 | } |
2337 | |
2338 | /* tipc_link_build_bc_proto_msg() - create broadcast protocol message |
2339 | */ |
2340 | static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast, |
2341 | u16 peers_snd_nxt, |
2342 | struct sk_buff_head *xmitq) |
2343 | { |
2344 | struct sk_buff *skb; |
2345 | struct tipc_msg *hdr; |
2346 | struct sk_buff *dfrd_skb = skb_peek(list_: &l->deferdq); |
2347 | u16 ack = l->rcv_nxt - 1; |
2348 | u16 gap_to = peers_snd_nxt - 1; |
2349 | |
2350 | skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE, |
2351 | data_sz: 0, dnode: l->addr, onode: tipc_own_addr(net: l->net), dport: 0, oport: 0, errcode: 0); |
2352 | if (!skb) |
2353 | return false; |
2354 | hdr = buf_msg(skb); |
2355 | msg_set_last_bcast(m: hdr, n: l->bc_sndlink->snd_nxt - 1); |
2356 | msg_set_bcast_ack(m: hdr, n: ack); |
2357 | msg_set_bcgap_after(m: hdr, n: ack); |
2358 | if (dfrd_skb) |
2359 | gap_to = buf_seqno(skb: dfrd_skb) - 1; |
2360 | msg_set_bcgap_to(m: hdr, n: gap_to); |
2361 | msg_set_non_seq(m: hdr, n: bcast); |
2362 | __skb_queue_tail(list: xmitq, newsk: skb); |
2363 | return true; |
2364 | } |
2365 | |
2366 | /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints. |
2367 | * |
2368 | * Give a newly added peer node the sequence number where it should |
2369 | * start receiving and acking broadcast packets. |
2370 | */ |
2371 | static void tipc_link_build_bc_init_msg(struct tipc_link *l, |
2372 | struct sk_buff_head *xmitq) |
2373 | { |
2374 | struct sk_buff_head list; |
2375 | |
2376 | __skb_queue_head_init(list: &list); |
2377 | if (!tipc_link_build_bc_proto_msg(l: l->bc_rcvlink, bcast: false, peers_snd_nxt: 0, xmitq: &list)) |
2378 | return; |
2379 | msg_set_bc_ack_invalid(m: buf_msg(skb: skb_peek(list_: &list)), invalid: true); |
2380 | tipc_link_xmit(l, list: &list, xmitq); |
2381 | } |
2382 | |
2383 | /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer |
2384 | */ |
2385 | void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr) |
2386 | { |
2387 | int mtyp = msg_type(m: hdr); |
2388 | u16 peers_snd_nxt = msg_bc_snd_nxt(m: hdr); |
2389 | |
2390 | if (link_is_up(l)) |
2391 | return; |
2392 | |
2393 | if (msg_user(m: hdr) == BCAST_PROTOCOL) { |
2394 | l->rcv_nxt = peers_snd_nxt; |
2395 | l->state = LINK_ESTABLISHED; |
2396 | return; |
2397 | } |
2398 | |
2399 | if (l->peer_caps & TIPC_BCAST_SYNCH) |
2400 | return; |
2401 | |
2402 | if (msg_peer_node_is_up(m: hdr)) |
2403 | return; |
2404 | |
2405 | /* Compatibility: accept older, less safe initial synch data */ |
2406 | if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG)) |
2407 | l->rcv_nxt = peers_snd_nxt; |
2408 | } |
2409 | |
2410 | /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state |
2411 | */ |
2412 | int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr, |
2413 | struct sk_buff_head *xmitq) |
2414 | { |
2415 | u16 peers_snd_nxt = msg_bc_snd_nxt(m: hdr); |
2416 | int rc = 0; |
2417 | |
2418 | if (!link_is_up(l)) |
2419 | return rc; |
2420 | |
2421 | if (!msg_peer_node_is_up(m: hdr)) |
2422 | return rc; |
2423 | |
2424 | /* Open when peer acknowledges our bcast init msg (pkt #1) */ |
2425 | if (msg_ack(m: hdr)) |
2426 | l->bc_peer_is_up = true; |
2427 | |
2428 | if (!l->bc_peer_is_up) |
2429 | return rc; |
2430 | |
2431 | /* Ignore if peers_snd_nxt goes beyond receive window */ |
2432 | if (more(left: peers_snd_nxt, right: l->rcv_nxt + l->window)) |
2433 | return rc; |
2434 | |
2435 | l->snd_nxt = peers_snd_nxt; |
2436 | if (link_bc_rcv_gap(l)) |
2437 | rc |= TIPC_LINK_SND_STATE; |
2438 | |
2439 | /* Return now if sender supports nack via STATE messages */ |
2440 | if (l->peer_caps & TIPC_BCAST_STATE_NACK) |
2441 | return rc; |
2442 | |
2443 | /* Otherwise, be backwards compatible */ |
2444 | |
2445 | if (!more(left: peers_snd_nxt, right: l->rcv_nxt)) { |
2446 | l->nack_state = BC_NACK_SND_CONDITIONAL; |
2447 | return 0; |
2448 | } |
2449 | |
2450 | /* Don't NACK if one was recently sent or peeked */ |
2451 | if (l->nack_state == BC_NACK_SND_SUPPRESS) { |
2452 | l->nack_state = BC_NACK_SND_UNCONDITIONAL; |
2453 | return 0; |
2454 | } |
2455 | |
2456 | /* Conditionally delay NACK sending until next synch rcv */ |
2457 | if (l->nack_state == BC_NACK_SND_CONDITIONAL) { |
2458 | l->nack_state = BC_NACK_SND_UNCONDITIONAL; |
2459 | if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN) |
2460 | return 0; |
2461 | } |
2462 | |
2463 | /* Send NACK now but suppress next one */ |
2464 | tipc_link_build_bc_proto_msg(l, bcast: true, peers_snd_nxt, xmitq); |
2465 | l->nack_state = BC_NACK_SND_SUPPRESS; |
2466 | return 0; |
2467 | } |
2468 | |
2469 | int tipc_link_bc_ack_rcv(struct tipc_link *r, u16 acked, u16 gap, |
2470 | struct tipc_gap_ack_blks *ga, |
2471 | struct sk_buff_head *xmitq, |
2472 | struct sk_buff_head *retrq) |
2473 | { |
2474 | struct tipc_link *l = r->bc_sndlink; |
2475 | bool unused = false; |
2476 | int rc = 0; |
2477 | |
2478 | if (!link_is_up(l: r) || !r->bc_peer_is_up) |
2479 | return 0; |
2480 | |
2481 | if (gap) { |
2482 | l->stats.recv_nacks++; |
2483 | r->stats.recv_nacks++; |
2484 | } |
2485 | |
2486 | if (less(left: acked, right: r->acked) || (acked == r->acked && !gap && !ga)) |
2487 | return 0; |
2488 | |
2489 | trace_tipc_link_bc_ack(r, f: acked, t: gap, tq: &l->transmq); |
2490 | tipc_link_advance_transmq(l, r, acked, gap, ga, xmitq: retrq, retransmitted: &unused, rc: &rc); |
2491 | |
2492 | tipc_link_advance_backlog(l, xmitq); |
2493 | if (unlikely(!skb_queue_empty(&l->wakeupq))) |
2494 | link_prepare_wakeup(l); |
2495 | |
2496 | return rc; |
2497 | } |
2498 | |
2499 | /* tipc_link_bc_nack_rcv(): receive broadcast nack message |
2500 | * This function is here for backwards compatibility, since |
2501 | * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5. |
2502 | */ |
2503 | int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb, |
2504 | struct sk_buff_head *xmitq) |
2505 | { |
2506 | struct tipc_msg *hdr = buf_msg(skb); |
2507 | u32 dnode = msg_destnode(m: hdr); |
2508 | int mtyp = msg_type(m: hdr); |
2509 | u16 acked = msg_bcast_ack(m: hdr); |
2510 | u16 from = acked + 1; |
2511 | u16 to = msg_bcgap_to(m: hdr); |
2512 | u16 peers_snd_nxt = to + 1; |
2513 | int rc = 0; |
2514 | |
2515 | kfree_skb(skb); |
2516 | |
2517 | if (!tipc_link_is_up(l) || !l->bc_peer_is_up) |
2518 | return 0; |
2519 | |
2520 | if (mtyp != STATE_MSG) |
2521 | return 0; |
2522 | |
2523 | if (dnode == tipc_own_addr(net: l->net)) { |
2524 | rc = tipc_link_bc_ack_rcv(r: l, acked, gap: to - acked, NULL, xmitq, |
2525 | retrq: xmitq); |
2526 | l->stats.recv_nacks++; |
2527 | return rc; |
2528 | } |
2529 | |
2530 | /* Msg for other node => suppress own NACK at next sync if applicable */ |
2531 | if (more(left: peers_snd_nxt, right: l->rcv_nxt) && !less(left: l->rcv_nxt, right: from)) |
2532 | l->nack_state = BC_NACK_SND_SUPPRESS; |
2533 | |
2534 | return 0; |
2535 | } |
2536 | |
2537 | void tipc_link_set_queue_limits(struct tipc_link *l, u32 min_win, u32 max_win) |
2538 | { |
2539 | int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE); |
2540 | |
2541 | l->min_win = min_win; |
2542 | l->ssthresh = max_win; |
2543 | l->max_win = max_win; |
2544 | l->window = min_win; |
2545 | l->backlog[TIPC_LOW_IMPORTANCE].limit = min_win * 2; |
2546 | l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = min_win * 4; |
2547 | l->backlog[TIPC_HIGH_IMPORTANCE].limit = min_win * 6; |
2548 | l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = min_win * 8; |
2549 | l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk; |
2550 | } |
2551 | |
2552 | /** |
2553 | * tipc_link_reset_stats - reset link statistics |
2554 | * @l: pointer to link |
2555 | */ |
2556 | void tipc_link_reset_stats(struct tipc_link *l) |
2557 | { |
2558 | memset(&l->stats, 0, sizeof(l->stats)); |
2559 | } |
2560 | |
2561 | static void link_print(struct tipc_link *l, const char *str) |
2562 | { |
2563 | struct sk_buff *hskb = skb_peek(list_: &l->transmq); |
2564 | u16 head = hskb ? msg_seqno(m: buf_msg(skb: hskb)) : l->snd_nxt - 1; |
2565 | u16 tail = l->snd_nxt - 1; |
2566 | |
2567 | pr_info("%s Link <%s> state %x\n" , str, l->name, l->state); |
2568 | pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n" , |
2569 | skb_queue_len(&l->transmq), head, tail, |
2570 | skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt); |
2571 | } |
2572 | |
2573 | /* Parse and validate nested (link) properties valid for media, bearer and link |
2574 | */ |
2575 | int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[]) |
2576 | { |
2577 | int err; |
2578 | |
2579 | err = nla_parse_nested_deprecated(tb: props, maxtype: TIPC_NLA_PROP_MAX, nla: prop, |
2580 | policy: tipc_nl_prop_policy, NULL); |
2581 | if (err) |
2582 | return err; |
2583 | |
2584 | if (props[TIPC_NLA_PROP_PRIO]) { |
2585 | u32 prio; |
2586 | |
2587 | prio = nla_get_u32(nla: props[TIPC_NLA_PROP_PRIO]); |
2588 | if (prio > TIPC_MAX_LINK_PRI) |
2589 | return -EINVAL; |
2590 | } |
2591 | |
2592 | if (props[TIPC_NLA_PROP_TOL]) { |
2593 | u32 tol; |
2594 | |
2595 | tol = nla_get_u32(nla: props[TIPC_NLA_PROP_TOL]); |
2596 | if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL)) |
2597 | return -EINVAL; |
2598 | } |
2599 | |
2600 | if (props[TIPC_NLA_PROP_WIN]) { |
2601 | u32 max_win; |
2602 | |
2603 | max_win = nla_get_u32(nla: props[TIPC_NLA_PROP_WIN]); |
2604 | if (max_win < TIPC_DEF_LINK_WIN || max_win > TIPC_MAX_LINK_WIN) |
2605 | return -EINVAL; |
2606 | } |
2607 | |
2608 | return 0; |
2609 | } |
2610 | |
2611 | static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s) |
2612 | { |
2613 | int i; |
2614 | struct nlattr *stats; |
2615 | |
2616 | struct nla_map { |
2617 | u32 key; |
2618 | u32 val; |
2619 | }; |
2620 | |
2621 | struct nla_map map[] = { |
2622 | {TIPC_NLA_STATS_RX_INFO, 0}, |
2623 | {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments}, |
2624 | {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented}, |
2625 | {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles}, |
2626 | {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled}, |
2627 | {TIPC_NLA_STATS_TX_INFO, 0}, |
2628 | {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments}, |
2629 | {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented}, |
2630 | {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles}, |
2631 | {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled}, |
2632 | {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ? |
2633 | s->msg_length_counts : 1}, |
2634 | {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts}, |
2635 | {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total}, |
2636 | {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]}, |
2637 | {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]}, |
2638 | {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]}, |
2639 | {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]}, |
2640 | {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]}, |
2641 | {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]}, |
2642 | {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]}, |
2643 | {TIPC_NLA_STATS_RX_STATES, s->recv_states}, |
2644 | {TIPC_NLA_STATS_RX_PROBES, s->recv_probes}, |
2645 | {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks}, |
2646 | {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv}, |
2647 | {TIPC_NLA_STATS_TX_STATES, s->sent_states}, |
2648 | {TIPC_NLA_STATS_TX_PROBES, s->sent_probes}, |
2649 | {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks}, |
2650 | {TIPC_NLA_STATS_TX_ACKS, s->sent_acks}, |
2651 | {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted}, |
2652 | {TIPC_NLA_STATS_DUPLICATES, s->duplicates}, |
2653 | {TIPC_NLA_STATS_LINK_CONGS, s->link_congs}, |
2654 | {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz}, |
2655 | {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ? |
2656 | (s->accu_queue_sz / s->queue_sz_counts) : 0} |
2657 | }; |
2658 | |
2659 | stats = nla_nest_start_noflag(skb, attrtype: TIPC_NLA_LINK_STATS); |
2660 | if (!stats) |
2661 | return -EMSGSIZE; |
2662 | |
2663 | for (i = 0; i < ARRAY_SIZE(map); i++) |
2664 | if (nla_put_u32(skb, attrtype: map[i].key, value: map[i].val)) |
2665 | goto msg_full; |
2666 | |
2667 | nla_nest_end(skb, start: stats); |
2668 | |
2669 | return 0; |
2670 | msg_full: |
2671 | nla_nest_cancel(skb, start: stats); |
2672 | |
2673 | return -EMSGSIZE; |
2674 | } |
2675 | |
2676 | /* Caller should hold appropriate locks to protect the link */ |
2677 | int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg, |
2678 | struct tipc_link *link, int nlflags) |
2679 | { |
2680 | u32 self = tipc_own_addr(net); |
2681 | struct nlattr *attrs; |
2682 | struct nlattr *prop; |
2683 | void *hdr; |
2684 | int err; |
2685 | |
2686 | hdr = genlmsg_put(skb: msg->skb, portid: msg->portid, seq: msg->seq, family: &tipc_genl_family, |
2687 | flags: nlflags, cmd: TIPC_NL_LINK_GET); |
2688 | if (!hdr) |
2689 | return -EMSGSIZE; |
2690 | |
2691 | attrs = nla_nest_start_noflag(skb: msg->skb, attrtype: TIPC_NLA_LINK); |
2692 | if (!attrs) |
2693 | goto msg_full; |
2694 | |
2695 | if (nla_put_string(skb: msg->skb, attrtype: TIPC_NLA_LINK_NAME, str: link->name)) |
2696 | goto attr_msg_full; |
2697 | if (nla_put_u32(skb: msg->skb, attrtype: TIPC_NLA_LINK_DEST, value: tipc_cluster_mask(addr: self))) |
2698 | goto attr_msg_full; |
2699 | if (nla_put_u32(skb: msg->skb, attrtype: TIPC_NLA_LINK_MTU, value: link->mtu)) |
2700 | goto attr_msg_full; |
2701 | if (nla_put_u32(skb: msg->skb, attrtype: TIPC_NLA_LINK_RX, value: link->stats.recv_pkts)) |
2702 | goto attr_msg_full; |
2703 | if (nla_put_u32(skb: msg->skb, attrtype: TIPC_NLA_LINK_TX, value: link->stats.sent_pkts)) |
2704 | goto attr_msg_full; |
2705 | |
2706 | if (tipc_link_is_up(l: link)) |
2707 | if (nla_put_flag(skb: msg->skb, attrtype: TIPC_NLA_LINK_UP)) |
2708 | goto attr_msg_full; |
2709 | if (link->active) |
2710 | if (nla_put_flag(skb: msg->skb, attrtype: TIPC_NLA_LINK_ACTIVE)) |
2711 | goto attr_msg_full; |
2712 | |
2713 | prop = nla_nest_start_noflag(skb: msg->skb, attrtype: TIPC_NLA_LINK_PROP); |
2714 | if (!prop) |
2715 | goto attr_msg_full; |
2716 | if (nla_put_u32(skb: msg->skb, attrtype: TIPC_NLA_PROP_PRIO, value: link->priority)) |
2717 | goto prop_msg_full; |
2718 | if (nla_put_u32(skb: msg->skb, attrtype: TIPC_NLA_PROP_TOL, value: link->tolerance)) |
2719 | goto prop_msg_full; |
2720 | if (nla_put_u32(skb: msg->skb, attrtype: TIPC_NLA_PROP_WIN, |
2721 | value: link->window)) |
2722 | goto prop_msg_full; |
2723 | if (nla_put_u32(skb: msg->skb, attrtype: TIPC_NLA_PROP_PRIO, value: link->priority)) |
2724 | goto prop_msg_full; |
2725 | nla_nest_end(skb: msg->skb, start: prop); |
2726 | |
2727 | err = __tipc_nl_add_stats(skb: msg->skb, s: &link->stats); |
2728 | if (err) |
2729 | goto attr_msg_full; |
2730 | |
2731 | nla_nest_end(skb: msg->skb, start: attrs); |
2732 | genlmsg_end(skb: msg->skb, hdr); |
2733 | |
2734 | return 0; |
2735 | |
2736 | prop_msg_full: |
2737 | nla_nest_cancel(skb: msg->skb, start: prop); |
2738 | attr_msg_full: |
2739 | nla_nest_cancel(skb: msg->skb, start: attrs); |
2740 | msg_full: |
2741 | genlmsg_cancel(skb: msg->skb, hdr); |
2742 | |
2743 | return -EMSGSIZE; |
2744 | } |
2745 | |
2746 | static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb, |
2747 | struct tipc_stats *stats) |
2748 | { |
2749 | int i; |
2750 | struct nlattr *nest; |
2751 | |
2752 | struct nla_map { |
2753 | __u32 key; |
2754 | __u32 val; |
2755 | }; |
2756 | |
2757 | struct nla_map map[] = { |
2758 | {TIPC_NLA_STATS_RX_INFO, stats->recv_pkts}, |
2759 | {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments}, |
2760 | {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented}, |
2761 | {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles}, |
2762 | {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled}, |
2763 | {TIPC_NLA_STATS_TX_INFO, stats->sent_pkts}, |
2764 | {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments}, |
2765 | {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented}, |
2766 | {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles}, |
2767 | {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled}, |
2768 | {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks}, |
2769 | {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv}, |
2770 | {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks}, |
2771 | {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks}, |
2772 | {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted}, |
2773 | {TIPC_NLA_STATS_DUPLICATES, stats->duplicates}, |
2774 | {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs}, |
2775 | {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz}, |
2776 | {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ? |
2777 | (stats->accu_queue_sz / stats->queue_sz_counts) : 0} |
2778 | }; |
2779 | |
2780 | nest = nla_nest_start_noflag(skb, attrtype: TIPC_NLA_LINK_STATS); |
2781 | if (!nest) |
2782 | return -EMSGSIZE; |
2783 | |
2784 | for (i = 0; i < ARRAY_SIZE(map); i++) |
2785 | if (nla_put_u32(skb, attrtype: map[i].key, value: map[i].val)) |
2786 | goto msg_full; |
2787 | |
2788 | nla_nest_end(skb, start: nest); |
2789 | |
2790 | return 0; |
2791 | msg_full: |
2792 | nla_nest_cancel(skb, start: nest); |
2793 | |
2794 | return -EMSGSIZE; |
2795 | } |
2796 | |
2797 | int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg, |
2798 | struct tipc_link *bcl) |
2799 | { |
2800 | int err; |
2801 | void *hdr; |
2802 | struct nlattr *attrs; |
2803 | struct nlattr *prop; |
2804 | u32 bc_mode = tipc_bcast_get_mode(net); |
2805 | u32 bc_ratio = tipc_bcast_get_broadcast_ratio(net); |
2806 | |
2807 | if (!bcl) |
2808 | return 0; |
2809 | |
2810 | tipc_bcast_lock(net); |
2811 | |
2812 | hdr = genlmsg_put(skb: msg->skb, portid: msg->portid, seq: msg->seq, family: &tipc_genl_family, |
2813 | NLM_F_MULTI, cmd: TIPC_NL_LINK_GET); |
2814 | if (!hdr) { |
2815 | tipc_bcast_unlock(net); |
2816 | return -EMSGSIZE; |
2817 | } |
2818 | |
2819 | attrs = nla_nest_start_noflag(skb: msg->skb, attrtype: TIPC_NLA_LINK); |
2820 | if (!attrs) |
2821 | goto msg_full; |
2822 | |
2823 | /* The broadcast link is always up */ |
2824 | if (nla_put_flag(skb: msg->skb, attrtype: TIPC_NLA_LINK_UP)) |
2825 | goto attr_msg_full; |
2826 | |
2827 | if (nla_put_flag(skb: msg->skb, attrtype: TIPC_NLA_LINK_BROADCAST)) |
2828 | goto attr_msg_full; |
2829 | if (nla_put_string(skb: msg->skb, attrtype: TIPC_NLA_LINK_NAME, str: bcl->name)) |
2830 | goto attr_msg_full; |
2831 | if (nla_put_u32(skb: msg->skb, attrtype: TIPC_NLA_LINK_RX, value: 0)) |
2832 | goto attr_msg_full; |
2833 | if (nla_put_u32(skb: msg->skb, attrtype: TIPC_NLA_LINK_TX, value: 0)) |
2834 | goto attr_msg_full; |
2835 | |
2836 | prop = nla_nest_start_noflag(skb: msg->skb, attrtype: TIPC_NLA_LINK_PROP); |
2837 | if (!prop) |
2838 | goto attr_msg_full; |
2839 | if (nla_put_u32(skb: msg->skb, attrtype: TIPC_NLA_PROP_WIN, value: bcl->max_win)) |
2840 | goto prop_msg_full; |
2841 | if (nla_put_u32(skb: msg->skb, attrtype: TIPC_NLA_PROP_BROADCAST, value: bc_mode)) |
2842 | goto prop_msg_full; |
2843 | if (bc_mode & BCLINK_MODE_SEL) |
2844 | if (nla_put_u32(skb: msg->skb, attrtype: TIPC_NLA_PROP_BROADCAST_RATIO, |
2845 | value: bc_ratio)) |
2846 | goto prop_msg_full; |
2847 | nla_nest_end(skb: msg->skb, start: prop); |
2848 | |
2849 | err = __tipc_nl_add_bc_link_stat(skb: msg->skb, stats: &bcl->stats); |
2850 | if (err) |
2851 | goto attr_msg_full; |
2852 | |
2853 | tipc_bcast_unlock(net); |
2854 | nla_nest_end(skb: msg->skb, start: attrs); |
2855 | genlmsg_end(skb: msg->skb, hdr); |
2856 | |
2857 | return 0; |
2858 | |
2859 | prop_msg_full: |
2860 | nla_nest_cancel(skb: msg->skb, start: prop); |
2861 | attr_msg_full: |
2862 | nla_nest_cancel(skb: msg->skb, start: attrs); |
2863 | msg_full: |
2864 | tipc_bcast_unlock(net); |
2865 | genlmsg_cancel(skb: msg->skb, hdr); |
2866 | |
2867 | return -EMSGSIZE; |
2868 | } |
2869 | |
2870 | void tipc_link_set_tolerance(struct tipc_link *l, u32 tol, |
2871 | struct sk_buff_head *xmitq) |
2872 | { |
2873 | l->tolerance = tol; |
2874 | if (l->bc_rcvlink) |
2875 | l->bc_rcvlink->tolerance = tol; |
2876 | if (link_is_up(l)) |
2877 | tipc_link_build_proto_msg(l, STATE_MSG, probe: 0, probe_reply: 0, rcvgap: 0, tolerance: tol, priority: 0, xmitq); |
2878 | } |
2879 | |
2880 | void tipc_link_set_prio(struct tipc_link *l, u32 prio, |
2881 | struct sk_buff_head *xmitq) |
2882 | { |
2883 | l->priority = prio; |
2884 | tipc_link_build_proto_msg(l, STATE_MSG, probe: 0, probe_reply: 0, rcvgap: 0, tolerance: 0, priority: prio, xmitq); |
2885 | } |
2886 | |
2887 | void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit) |
2888 | { |
2889 | l->abort_limit = limit; |
2890 | } |
2891 | |
2892 | /** |
2893 | * tipc_link_dump - dump TIPC link data |
2894 | * @l: tipc link to be dumped |
2895 | * @dqueues: bitmask to decide if any link queue to be dumped? |
2896 | * - TIPC_DUMP_NONE: don't dump link queues |
2897 | * - TIPC_DUMP_TRANSMQ: dump link transmq queue |
2898 | * - TIPC_DUMP_BACKLOGQ: dump link backlog queue |
2899 | * - TIPC_DUMP_DEFERDQ: dump link deferd queue |
2900 | * - TIPC_DUMP_INPUTQ: dump link input queue |
2901 | * - TIPC_DUMP_WAKEUP: dump link wakeup queue |
2902 | * - TIPC_DUMP_ALL: dump all the link queues above |
2903 | * @buf: returned buffer of dump data in format |
2904 | */ |
2905 | int tipc_link_dump(struct tipc_link *l, u16 dqueues, char *buf) |
2906 | { |
2907 | int i = 0; |
2908 | size_t sz = (dqueues) ? LINK_LMAX : LINK_LMIN; |
2909 | struct sk_buff_head *list; |
2910 | struct sk_buff *hskb, *tskb; |
2911 | u32 len; |
2912 | |
2913 | if (!l) { |
2914 | i += scnprintf(buf, size: sz, fmt: "link data: (null)\n" ); |
2915 | return i; |
2916 | } |
2917 | |
2918 | i += scnprintf(buf, size: sz, fmt: "link data: %x" , l->addr); |
2919 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " %x" , l->state); |
2920 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " %u" , l->in_session); |
2921 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " %u" , l->session); |
2922 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " %u" , l->peer_session); |
2923 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " %u" , l->snd_nxt); |
2924 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " %u" , l->rcv_nxt); |
2925 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " %u" , l->snd_nxt_state); |
2926 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " %u" , l->rcv_nxt_state); |
2927 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " %x" , l->peer_caps); |
2928 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " %u" , l->silent_intv_cnt); |
2929 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " %u" , l->rst_cnt); |
2930 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " %u" , 0); |
2931 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " %u" , 0); |
2932 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " %u" , l->acked); |
2933 | |
2934 | list = &l->transmq; |
2935 | len = skb_queue_len(list_: list); |
2936 | hskb = skb_peek(list_: list); |
2937 | tskb = skb_peek_tail(list_: list); |
2938 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " | %u %u %u" , len, |
2939 | (hskb) ? msg_seqno(m: buf_msg(skb: hskb)) : 0, |
2940 | (tskb) ? msg_seqno(m: buf_msg(skb: tskb)) : 0); |
2941 | |
2942 | list = &l->deferdq; |
2943 | len = skb_queue_len(list_: list); |
2944 | hskb = skb_peek(list_: list); |
2945 | tskb = skb_peek_tail(list_: list); |
2946 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " | %u %u %u" , len, |
2947 | (hskb) ? msg_seqno(m: buf_msg(skb: hskb)) : 0, |
2948 | (tskb) ? msg_seqno(m: buf_msg(skb: tskb)) : 0); |
2949 | |
2950 | list = &l->backlogq; |
2951 | len = skb_queue_len(list_: list); |
2952 | hskb = skb_peek(list_: list); |
2953 | tskb = skb_peek_tail(list_: list); |
2954 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " | %u %u %u" , len, |
2955 | (hskb) ? msg_seqno(m: buf_msg(skb: hskb)) : 0, |
2956 | (tskb) ? msg_seqno(m: buf_msg(skb: tskb)) : 0); |
2957 | |
2958 | list = l->inputq; |
2959 | len = skb_queue_len(list_: list); |
2960 | hskb = skb_peek(list_: list); |
2961 | tskb = skb_peek_tail(list_: list); |
2962 | i += scnprintf(buf: buf + i, size: sz - i, fmt: " | %u %u %u\n" , len, |
2963 | (hskb) ? msg_seqno(m: buf_msg(skb: hskb)) : 0, |
2964 | (tskb) ? msg_seqno(m: buf_msg(skb: tskb)) : 0); |
2965 | |
2966 | if (dqueues & TIPC_DUMP_TRANSMQ) { |
2967 | i += scnprintf(buf: buf + i, size: sz - i, fmt: "transmq: " ); |
2968 | i += tipc_list_dump(list: &l->transmq, more: false, buf: buf + i); |
2969 | } |
2970 | if (dqueues & TIPC_DUMP_BACKLOGQ) { |
2971 | i += scnprintf(buf: buf + i, size: sz - i, |
2972 | fmt: "backlogq: <%u %u %u %u %u>, " , |
2973 | l->backlog[TIPC_LOW_IMPORTANCE].len, |
2974 | l->backlog[TIPC_MEDIUM_IMPORTANCE].len, |
2975 | l->backlog[TIPC_HIGH_IMPORTANCE].len, |
2976 | l->backlog[TIPC_CRITICAL_IMPORTANCE].len, |
2977 | l->backlog[TIPC_SYSTEM_IMPORTANCE].len); |
2978 | i += tipc_list_dump(list: &l->backlogq, more: false, buf: buf + i); |
2979 | } |
2980 | if (dqueues & TIPC_DUMP_DEFERDQ) { |
2981 | i += scnprintf(buf: buf + i, size: sz - i, fmt: "deferdq: " ); |
2982 | i += tipc_list_dump(list: &l->deferdq, more: false, buf: buf + i); |
2983 | } |
2984 | if (dqueues & TIPC_DUMP_INPUTQ) { |
2985 | i += scnprintf(buf: buf + i, size: sz - i, fmt: "inputq: " ); |
2986 | i += tipc_list_dump(list: l->inputq, more: false, buf: buf + i); |
2987 | } |
2988 | if (dqueues & TIPC_DUMP_WAKEUP) { |
2989 | i += scnprintf(buf: buf + i, size: sz - i, fmt: "wakeup: " ); |
2990 | i += tipc_list_dump(list: &l->wakeupq, more: false, buf: buf + i); |
2991 | } |
2992 | |
2993 | return i; |
2994 | } |
2995 | |