]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * net/tipc/bcast.c: TIPC broadcast code | |
3 | * | |
4 | * Copyright (c) 2004-2006, Ericsson AB | |
5 | * Copyright (c) 2004, Intel Corporation. | |
6 | * Copyright (c) 2005, Wind River Systems | |
7 | * All rights reserved. | |
8 | * | |
9 | * Redistribution and use in source and binary forms, with or without | |
10 | * modification, are permitted provided that the following conditions are met: | |
11 | * | |
12 | * 1. Redistributions of source code must retain the above copyright | |
13 | * notice, this list of conditions and the following disclaimer. | |
14 | * 2. Redistributions in binary form must reproduce the above copyright | |
15 | * notice, this list of conditions and the following disclaimer in the | |
16 | * documentation and/or other materials provided with the distribution. | |
17 | * 3. Neither the names of the copyright holders nor the names of its | |
18 | * contributors may be used to endorse or promote products derived from | |
19 | * this software without specific prior written permission. | |
20 | * | |
21 | * Alternatively, this software may be distributed under the terms of the | |
22 | * GNU General Public License ("GPL") version 2 as published by the Free | |
23 | * Software Foundation. | |
24 | * | |
25 | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" | |
26 | * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE | |
27 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE | |
28 | * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE | |
29 | * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR | |
30 | * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF | |
31 | * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS | |
32 | * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN | |
33 | * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) | |
34 | * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE | |
35 | * POSSIBILITY OF SUCH DAMAGE. | |
36 | */ | |
37 | ||
38 | #include "core.h" | |
39 | #include "msg.h" | |
40 | #include "dbg.h" | |
41 | #include "link.h" | |
42 | #include "net.h" | |
43 | #include "node.h" | |
44 | #include "port.h" | |
45 | #include "addr.h" | |
46 | #include "node_subscr.h" | |
47 | #include "name_distr.h" | |
48 | #include "bearer.h" | |
49 | #include "name_table.h" | |
50 | #include "bcast.h" | |
51 | ||
52 | #define MAX_PKT_DEFAULT_MCAST 1500 /* bcast link max packet size (fixed) */ | |
53 | ||
54 | #define BCLINK_WIN_DEFAULT 20 /* bcast link window size (default) */ | |
55 | ||
56 | #define BCLINK_LOG_BUF_SIZE 0 | |
57 | ||
58 | /* | |
59 | * Loss rate for incoming broadcast frames; used to test retransmission code. | |
60 | * Set to N to cause every N'th frame to be discarded; 0 => don't discard any. | |
61 | */ | |
62 | ||
63 | #define TIPC_BCAST_LOSS_RATE 0 | |
64 | ||
65 | /** | |
66 | * struct bcbearer_pair - a pair of bearers used by broadcast link | |
67 | * @primary: pointer to primary bearer | |
68 | * @secondary: pointer to secondary bearer | |
69 | * | |
70 | * Bearers must have same priority and same set of reachable destinations | |
71 | * to be paired. | |
72 | */ | |
73 | ||
74 | struct bcbearer_pair { | |
75 | struct bearer *primary; | |
76 | struct bearer *secondary; | |
77 | }; | |
78 | ||
79 | /** | |
80 | * struct bcbearer - bearer used by broadcast link | |
81 | * @bearer: (non-standard) broadcast bearer structure | |
82 | * @media: (non-standard) broadcast media structure | |
83 | * @bpairs: array of bearer pairs | |
84 | * @bpairs_temp: temporary array of bearer pairs used by tipc_bcbearer_sort() | |
85 | * @remains: temporary node map used by tipc_bcbearer_send() | |
86 | * @remains_new: temporary node map used tipc_bcbearer_send() | |
87 | * | |
88 | * Note: The fields labelled "temporary" are incorporated into the bearer | |
89 | * to avoid consuming potentially limited stack space through the use of | |
90 | * large local variables within multicast routines. Concurrent access is | |
91 | * prevented through use of the spinlock "bc_lock". | |
92 | */ | |
93 | ||
94 | struct bcbearer { | |
95 | struct bearer bearer; | |
96 | struct media media; | |
97 | struct bcbearer_pair bpairs[MAX_BEARERS]; | |
98 | struct bcbearer_pair bpairs_temp[TIPC_MAX_LINK_PRI + 1]; | |
99 | struct tipc_node_map remains; | |
100 | struct tipc_node_map remains_new; | |
101 | }; | |
102 | ||
103 | /** | |
104 | * struct bclink - link used for broadcast messages | |
105 | * @link: (non-standard) broadcast link structure | |
106 | * @node: (non-standard) node structure representing b'cast link's peer node | |
107 | * | |
108 | * Handles sequence numbering, fragmentation, bundling, etc. | |
109 | */ | |
110 | ||
111 | struct bclink { | |
112 | struct link link; | |
113 | struct tipc_node node; | |
114 | }; | |
115 | ||
116 | ||
117 | static struct bcbearer *bcbearer = NULL; | |
118 | static struct bclink *bclink = NULL; | |
119 | static struct link *bcl = NULL; | |
120 | static DEFINE_SPINLOCK(bc_lock); | |
121 | ||
122 | const char tipc_bclink_name[] = "multicast-link"; | |
123 | ||
124 | ||
125 | static u32 buf_seqno(struct sk_buff *buf) | |
126 | { | |
127 | return msg_seqno(buf_msg(buf)); | |
128 | } | |
129 | ||
130 | static u32 bcbuf_acks(struct sk_buff *buf) | |
131 | { | |
132 | return (u32)(unsigned long)TIPC_SKB_CB(buf)->handle; | |
133 | } | |
134 | ||
135 | static void bcbuf_set_acks(struct sk_buff *buf, u32 acks) | |
136 | { | |
137 | TIPC_SKB_CB(buf)->handle = (void *)(unsigned long)acks; | |
138 | } | |
139 | ||
140 | static void bcbuf_decr_acks(struct sk_buff *buf) | |
141 | { | |
142 | bcbuf_set_acks(buf, bcbuf_acks(buf) - 1); | |
143 | } | |
144 | ||
145 | ||
146 | /** | |
147 | * bclink_set_gap - set gap according to contents of current deferred pkt queue | |
148 | * | |
149 | * Called with 'node' locked, bc_lock unlocked | |
150 | */ | |
151 | ||
152 | static void bclink_set_gap(struct tipc_node *n_ptr) | |
153 | { | |
154 | struct sk_buff *buf = n_ptr->bclink.deferred_head; | |
155 | ||
156 | n_ptr->bclink.gap_after = n_ptr->bclink.gap_to = | |
157 | mod(n_ptr->bclink.last_in); | |
158 | if (unlikely(buf != NULL)) | |
159 | n_ptr->bclink.gap_to = mod(buf_seqno(buf) - 1); | |
160 | } | |
161 | ||
162 | /** | |
163 | * bclink_ack_allowed - test if ACK or NACK message can be sent at this moment | |
164 | * | |
165 | * This mechanism endeavours to prevent all nodes in network from trying | |
166 | * to ACK or NACK at the same time. | |
167 | * | |
168 | * Note: TIPC uses a different trigger to distribute ACKs than it does to | |
169 | * distribute NACKs, but tries to use the same spacing (divide by 16). | |
170 | */ | |
171 | ||
172 | static int bclink_ack_allowed(u32 n) | |
173 | { | |
174 | return((n % TIPC_MIN_LINK_WIN) == tipc_own_tag); | |
175 | } | |
176 | ||
177 | ||
178 | /** | |
179 | * bclink_retransmit_pkt - retransmit broadcast packets | |
180 | * @after: sequence number of last packet to *not* retransmit | |
181 | * @to: sequence number of last packet to retransmit | |
182 | * | |
183 | * Called with bc_lock locked | |
184 | */ | |
185 | ||
186 | static void bclink_retransmit_pkt(u32 after, u32 to) | |
187 | { | |
188 | struct sk_buff *buf; | |
189 | ||
190 | buf = bcl->first_out; | |
191 | while (buf && less_eq(buf_seqno(buf), after)) { | |
192 | buf = buf->next; | |
193 | } | |
194 | tipc_link_retransmit(bcl, buf, mod(to - after)); | |
195 | } | |
196 | ||
197 | /** | |
198 | * tipc_bclink_acknowledge - handle acknowledgement of broadcast packets | |
199 | * @n_ptr: node that sent acknowledgement info | |
200 | * @acked: broadcast sequence # that has been acknowledged | |
201 | * | |
202 | * Node is locked, bc_lock unlocked. | |
203 | */ | |
204 | ||
205 | void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked) | |
206 | { | |
207 | struct sk_buff *crs; | |
208 | struct sk_buff *next; | |
209 | unsigned int released = 0; | |
210 | ||
211 | if (less_eq(acked, n_ptr->bclink.acked)) | |
212 | return; | |
213 | ||
214 | spin_lock_bh(&bc_lock); | |
215 | ||
216 | /* Skip over packets that node has previously acknowledged */ | |
217 | ||
218 | crs = bcl->first_out; | |
219 | while (crs && less_eq(buf_seqno(crs), n_ptr->bclink.acked)) { | |
220 | crs = crs->next; | |
221 | } | |
222 | ||
223 | /* Update packets that node is now acknowledging */ | |
224 | ||
225 | while (crs && less_eq(buf_seqno(crs), acked)) { | |
226 | next = crs->next; | |
227 | bcbuf_decr_acks(crs); | |
228 | if (bcbuf_acks(crs) == 0) { | |
229 | bcl->first_out = next; | |
230 | bcl->out_queue_size--; | |
231 | buf_discard(crs); | |
232 | released = 1; | |
233 | } | |
234 | crs = next; | |
235 | } | |
236 | n_ptr->bclink.acked = acked; | |
237 | ||
238 | /* Try resolving broadcast link congestion, if necessary */ | |
239 | ||
240 | if (unlikely(bcl->next_out)) | |
241 | tipc_link_push_queue(bcl); | |
242 | if (unlikely(released && !list_empty(&bcl->waiting_ports))) | |
243 | tipc_link_wakeup_ports(bcl, 0); | |
244 | spin_unlock_bh(&bc_lock); | |
245 | } | |
246 | ||
247 | /** | |
248 | * bclink_send_ack - unicast an ACK msg | |
249 | * | |
250 | * tipc_net_lock and node lock set | |
251 | */ | |
252 | ||
253 | static void bclink_send_ack(struct tipc_node *n_ptr) | |
254 | { | |
255 | struct link *l_ptr = n_ptr->active_links[n_ptr->addr & 1]; | |
256 | ||
257 | if (l_ptr != NULL) | |
258 | tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0); | |
259 | } | |
260 | ||
261 | /** | |
262 | * bclink_send_nack- broadcast a NACK msg | |
263 | * | |
264 | * tipc_net_lock and node lock set | |
265 | */ | |
266 | ||
267 | static void bclink_send_nack(struct tipc_node *n_ptr) | |
268 | { | |
269 | struct sk_buff *buf; | |
270 | struct tipc_msg *msg; | |
271 | ||
272 | if (!less(n_ptr->bclink.gap_after, n_ptr->bclink.gap_to)) | |
273 | return; | |
274 | ||
275 | buf = buf_acquire(INT_H_SIZE); | |
276 | if (buf) { | |
277 | msg = buf_msg(buf); | |
278 | msg_init(msg, BCAST_PROTOCOL, STATE_MSG, | |
279 | INT_H_SIZE, n_ptr->addr); | |
280 | msg_set_mc_netid(msg, tipc_net_id); | |
281 | msg_set_bcast_ack(msg, mod(n_ptr->bclink.last_in)); | |
282 | msg_set_bcgap_after(msg, n_ptr->bclink.gap_after); | |
283 | msg_set_bcgap_to(msg, n_ptr->bclink.gap_to); | |
284 | msg_set_bcast_tag(msg, tipc_own_tag); | |
285 | ||
286 | if (tipc_bearer_send(&bcbearer->bearer, buf, NULL)) { | |
287 | bcl->stats.sent_nacks++; | |
288 | buf_discard(buf); | |
289 | } else { | |
290 | tipc_bearer_schedule(bcl->b_ptr, bcl); | |
291 | bcl->proto_msg_queue = buf; | |
292 | bcl->stats.bearer_congs++; | |
293 | } | |
294 | ||
295 | /* | |
296 | * Ensure we doesn't send another NACK msg to the node | |
297 | * until 16 more deferred messages arrive from it | |
298 | * (i.e. helps prevent all nodes from NACK'ing at same time) | |
299 | */ | |
300 | ||
301 | n_ptr->bclink.nack_sync = tipc_own_tag; | |
302 | } | |
303 | } | |
304 | ||
305 | /** | |
306 | * tipc_bclink_check_gap - send a NACK if a sequence gap exists | |
307 | * | |
308 | * tipc_net_lock and node lock set | |
309 | */ | |
310 | ||
311 | void tipc_bclink_check_gap(struct tipc_node *n_ptr, u32 last_sent) | |
312 | { | |
313 | if (!n_ptr->bclink.supported || | |
314 | less_eq(last_sent, mod(n_ptr->bclink.last_in))) | |
315 | return; | |
316 | ||
317 | bclink_set_gap(n_ptr); | |
318 | if (n_ptr->bclink.gap_after == n_ptr->bclink.gap_to) | |
319 | n_ptr->bclink.gap_to = last_sent; | |
320 | bclink_send_nack(n_ptr); | |
321 | } | |
322 | ||
323 | /** | |
324 | * tipc_bclink_peek_nack - process a NACK msg meant for another node | |
325 | * | |
326 | * Only tipc_net_lock set. | |
327 | */ | |
328 | ||
329 | static void tipc_bclink_peek_nack(u32 dest, u32 sender_tag, u32 gap_after, u32 gap_to) | |
330 | { | |
331 | struct tipc_node *n_ptr = tipc_node_find(dest); | |
332 | u32 my_after, my_to; | |
333 | ||
334 | if (unlikely(!n_ptr || !tipc_node_is_up(n_ptr))) | |
335 | return; | |
336 | tipc_node_lock(n_ptr); | |
337 | /* | |
338 | * Modify gap to suppress unnecessary NACKs from this node | |
339 | */ | |
340 | my_after = n_ptr->bclink.gap_after; | |
341 | my_to = n_ptr->bclink.gap_to; | |
342 | ||
343 | if (less_eq(gap_after, my_after)) { | |
344 | if (less(my_after, gap_to) && less(gap_to, my_to)) | |
345 | n_ptr->bclink.gap_after = gap_to; | |
346 | else if (less_eq(my_to, gap_to)) | |
347 | n_ptr->bclink.gap_to = n_ptr->bclink.gap_after; | |
348 | } else if (less_eq(gap_after, my_to)) { | |
349 | if (less_eq(my_to, gap_to)) | |
350 | n_ptr->bclink.gap_to = gap_after; | |
351 | } else { | |
352 | /* | |
353 | * Expand gap if missing bufs not in deferred queue: | |
354 | */ | |
355 | struct sk_buff *buf = n_ptr->bclink.deferred_head; | |
356 | u32 prev = n_ptr->bclink.gap_to; | |
357 | ||
358 | for (; buf; buf = buf->next) { | |
359 | u32 seqno = buf_seqno(buf); | |
360 | ||
361 | if (mod(seqno - prev) != 1) { | |
362 | buf = NULL; | |
363 | break; | |
364 | } | |
365 | if (seqno == gap_after) | |
366 | break; | |
367 | prev = seqno; | |
368 | } | |
369 | if (buf == NULL) | |
370 | n_ptr->bclink.gap_to = gap_after; | |
371 | } | |
372 | /* | |
373 | * Some nodes may send a complementary NACK now: | |
374 | */ | |
375 | if (bclink_ack_allowed(sender_tag + 1)) { | |
376 | if (n_ptr->bclink.gap_to != n_ptr->bclink.gap_after) { | |
377 | bclink_send_nack(n_ptr); | |
378 | bclink_set_gap(n_ptr); | |
379 | } | |
380 | } | |
381 | tipc_node_unlock(n_ptr); | |
382 | } | |
383 | ||
384 | /** | |
385 | * tipc_bclink_send_msg - broadcast a packet to all nodes in cluster | |
386 | */ | |
387 | ||
388 | int tipc_bclink_send_msg(struct sk_buff *buf) | |
389 | { | |
390 | int res; | |
391 | ||
392 | spin_lock_bh(&bc_lock); | |
393 | ||
394 | res = tipc_link_send_buf(bcl, buf); | |
395 | if (unlikely(res == -ELINKCONG)) | |
396 | buf_discard(buf); | |
397 | else | |
398 | bcl->stats.sent_info++; | |
399 | ||
400 | if (bcl->out_queue_size > bcl->stats.max_queue_sz) | |
401 | bcl->stats.max_queue_sz = bcl->out_queue_size; | |
402 | bcl->stats.queue_sz_counts++; | |
403 | bcl->stats.accu_queue_sz += bcl->out_queue_size; | |
404 | ||
405 | spin_unlock_bh(&bc_lock); | |
406 | return res; | |
407 | } | |
408 | ||
409 | /** | |
410 | * tipc_bclink_recv_pkt - receive a broadcast packet, and deliver upwards | |
411 | * | |
412 | * tipc_net_lock is read_locked, no other locks set | |
413 | */ | |
414 | ||
415 | void tipc_bclink_recv_pkt(struct sk_buff *buf) | |
416 | { | |
417 | #if (TIPC_BCAST_LOSS_RATE) | |
418 | static int rx_count = 0; | |
419 | #endif | |
420 | struct tipc_msg *msg = buf_msg(buf); | |
421 | struct tipc_node* node = tipc_node_find(msg_prevnode(msg)); | |
422 | u32 next_in; | |
423 | u32 seqno; | |
424 | struct sk_buff *deferred; | |
425 | ||
426 | msg_dbg(msg, "<BC<<<"); | |
427 | ||
428 | if (unlikely(!node || !tipc_node_is_up(node) || !node->bclink.supported || | |
429 | (msg_mc_netid(msg) != tipc_net_id))) { | |
430 | buf_discard(buf); | |
431 | return; | |
432 | } | |
433 | ||
434 | if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) { | |
435 | msg_dbg(msg, "<BCNACK<<<"); | |
436 | if (msg_destnode(msg) == tipc_own_addr) { | |
437 | tipc_node_lock(node); | |
438 | tipc_bclink_acknowledge(node, msg_bcast_ack(msg)); | |
439 | tipc_node_unlock(node); | |
440 | spin_lock_bh(&bc_lock); | |
441 | bcl->stats.recv_nacks++; | |
442 | bcl->owner->next = node; /* remember requestor */ | |
443 | bclink_retransmit_pkt(msg_bcgap_after(msg), | |
444 | msg_bcgap_to(msg)); | |
445 | bcl->owner->next = NULL; | |
446 | spin_unlock_bh(&bc_lock); | |
447 | } else { | |
448 | tipc_bclink_peek_nack(msg_destnode(msg), | |
449 | msg_bcast_tag(msg), | |
450 | msg_bcgap_after(msg), | |
451 | msg_bcgap_to(msg)); | |
452 | } | |
453 | buf_discard(buf); | |
454 | return; | |
455 | } | |
456 | ||
457 | #if (TIPC_BCAST_LOSS_RATE) | |
458 | if (++rx_count == TIPC_BCAST_LOSS_RATE) { | |
459 | rx_count = 0; | |
460 | buf_discard(buf); | |
461 | return; | |
462 | } | |
463 | #endif | |
464 | ||
465 | tipc_node_lock(node); | |
466 | receive: | |
467 | deferred = node->bclink.deferred_head; | |
468 | next_in = mod(node->bclink.last_in + 1); | |
469 | seqno = msg_seqno(msg); | |
470 | ||
471 | if (likely(seqno == next_in)) { | |
472 | bcl->stats.recv_info++; | |
473 | node->bclink.last_in++; | |
474 | bclink_set_gap(node); | |
475 | if (unlikely(bclink_ack_allowed(seqno))) { | |
476 | bclink_send_ack(node); | |
477 | bcl->stats.sent_acks++; | |
478 | } | |
479 | if (likely(msg_isdata(msg))) { | |
480 | tipc_node_unlock(node); | |
481 | tipc_port_recv_mcast(buf, NULL); | |
482 | } else if (msg_user(msg) == MSG_BUNDLER) { | |
483 | bcl->stats.recv_bundles++; | |
484 | bcl->stats.recv_bundled += msg_msgcnt(msg); | |
485 | tipc_node_unlock(node); | |
486 | tipc_link_recv_bundle(buf); | |
487 | } else if (msg_user(msg) == MSG_FRAGMENTER) { | |
488 | bcl->stats.recv_fragments++; | |
489 | if (tipc_link_recv_fragment(&node->bclink.defragm, | |
490 | &buf, &msg)) | |
491 | bcl->stats.recv_fragmented++; | |
492 | tipc_node_unlock(node); | |
493 | tipc_net_route_msg(buf); | |
494 | } else { | |
495 | tipc_node_unlock(node); | |
496 | tipc_net_route_msg(buf); | |
497 | } | |
498 | if (deferred && (buf_seqno(deferred) == mod(next_in + 1))) { | |
499 | tipc_node_lock(node); | |
500 | buf = deferred; | |
501 | msg = buf_msg(buf); | |
502 | node->bclink.deferred_head = deferred->next; | |
503 | goto receive; | |
504 | } | |
505 | return; | |
506 | } else if (less(next_in, seqno)) { | |
507 | u32 gap_after = node->bclink.gap_after; | |
508 | u32 gap_to = node->bclink.gap_to; | |
509 | ||
510 | if (tipc_link_defer_pkt(&node->bclink.deferred_head, | |
511 | &node->bclink.deferred_tail, | |
512 | buf)) { | |
513 | node->bclink.nack_sync++; | |
514 | bcl->stats.deferred_recv++; | |
515 | if (seqno == mod(gap_after + 1)) | |
516 | node->bclink.gap_after = seqno; | |
517 | else if (less(gap_after, seqno) && less(seqno, gap_to)) | |
518 | node->bclink.gap_to = seqno; | |
519 | } | |
520 | if (bclink_ack_allowed(node->bclink.nack_sync)) { | |
521 | if (gap_to != gap_after) | |
522 | bclink_send_nack(node); | |
523 | bclink_set_gap(node); | |
524 | } | |
525 | } else { | |
526 | bcl->stats.duplicates++; | |
527 | buf_discard(buf); | |
528 | } | |
529 | tipc_node_unlock(node); | |
530 | } | |
531 | ||
532 | u32 tipc_bclink_get_last_sent(void) | |
533 | { | |
534 | u32 last_sent = mod(bcl->next_out_no - 1); | |
535 | ||
536 | if (bcl->next_out) | |
537 | last_sent = mod(buf_seqno(bcl->next_out) - 1); | |
538 | return last_sent; | |
539 | } | |
540 | ||
541 | u32 tipc_bclink_acks_missing(struct tipc_node *n_ptr) | |
542 | { | |
543 | return (n_ptr->bclink.supported && | |
544 | (tipc_bclink_get_last_sent() != n_ptr->bclink.acked)); | |
545 | } | |
546 | ||
547 | ||
548 | /** | |
549 | * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer | |
550 | * | |
551 | * Send through as many bearers as necessary to reach all nodes | |
552 | * that support TIPC multicasting. | |
553 | * | |
554 | * Returns 0 if packet sent successfully, non-zero if not | |
555 | */ | |
556 | ||
557 | static int tipc_bcbearer_send(struct sk_buff *buf, | |
558 | struct tipc_bearer *unused1, | |
559 | struct tipc_media_addr *unused2) | |
560 | { | |
561 | static int send_count = 0; | |
562 | ||
563 | int bp_index; | |
564 | int swap_time; | |
565 | ||
566 | /* Prepare buffer for broadcasting (if first time trying to send it) */ | |
567 | ||
568 | if (likely(!msg_non_seq(buf_msg(buf)))) { | |
569 | struct tipc_msg *msg; | |
570 | ||
571 | assert(tipc_cltr_bcast_nodes.count != 0); | |
572 | bcbuf_set_acks(buf, tipc_cltr_bcast_nodes.count); | |
573 | msg = buf_msg(buf); | |
574 | msg_set_non_seq(msg, 1); | |
575 | msg_set_mc_netid(msg, tipc_net_id); | |
576 | } | |
577 | ||
578 | /* Determine if bearer pairs should be swapped following this attempt */ | |
579 | ||
580 | if ((swap_time = (++send_count >= 10))) | |
581 | send_count = 0; | |
582 | ||
583 | /* Send buffer over bearers until all targets reached */ | |
584 | ||
585 | bcbearer->remains = tipc_cltr_bcast_nodes; | |
586 | ||
587 | for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) { | |
588 | struct bearer *p = bcbearer->bpairs[bp_index].primary; | |
589 | struct bearer *s = bcbearer->bpairs[bp_index].secondary; | |
590 | ||
591 | if (!p) | |
592 | break; /* no more bearers to try */ | |
593 | ||
594 | tipc_nmap_diff(&bcbearer->remains, &p->nodes, &bcbearer->remains_new); | |
595 | if (bcbearer->remains_new.count == bcbearer->remains.count) | |
596 | continue; /* bearer pair doesn't add anything */ | |
597 | ||
598 | if (!p->publ.blocked && | |
599 | !p->media->send_msg(buf, &p->publ, &p->media->bcast_addr)) { | |
600 | if (swap_time && s && !s->publ.blocked) | |
601 | goto swap; | |
602 | else | |
603 | goto update; | |
604 | } | |
605 | ||
606 | if (!s || s->publ.blocked || | |
607 | s->media->send_msg(buf, &s->publ, &s->media->bcast_addr)) | |
608 | continue; /* unable to send using bearer pair */ | |
609 | swap: | |
610 | bcbearer->bpairs[bp_index].primary = s; | |
611 | bcbearer->bpairs[bp_index].secondary = p; | |
612 | update: | |
613 | if (bcbearer->remains_new.count == 0) | |
614 | return 0; | |
615 | ||
616 | bcbearer->remains = bcbearer->remains_new; | |
617 | } | |
618 | ||
619 | /* Unable to reach all targets */ | |
620 | ||
621 | bcbearer->bearer.publ.blocked = 1; | |
622 | bcl->stats.bearer_congs++; | |
623 | return 1; | |
624 | } | |
625 | ||
626 | /** | |
627 | * tipc_bcbearer_sort - create sets of bearer pairs used by broadcast bearer | |
628 | */ | |
629 | ||
630 | void tipc_bcbearer_sort(void) | |
631 | { | |
632 | struct bcbearer_pair *bp_temp = bcbearer->bpairs_temp; | |
633 | struct bcbearer_pair *bp_curr; | |
634 | int b_index; | |
635 | int pri; | |
636 | ||
637 | spin_lock_bh(&bc_lock); | |
638 | ||
639 | /* Group bearers by priority (can assume max of two per priority) */ | |
640 | ||
641 | memset(bp_temp, 0, sizeof(bcbearer->bpairs_temp)); | |
642 | ||
643 | for (b_index = 0; b_index < MAX_BEARERS; b_index++) { | |
644 | struct bearer *b = &tipc_bearers[b_index]; | |
645 | ||
646 | if (!b->active || !b->nodes.count) | |
647 | continue; | |
648 | ||
649 | if (!bp_temp[b->priority].primary) | |
650 | bp_temp[b->priority].primary = b; | |
651 | else | |
652 | bp_temp[b->priority].secondary = b; | |
653 | } | |
654 | ||
655 | /* Create array of bearer pairs for broadcasting */ | |
656 | ||
657 | bp_curr = bcbearer->bpairs; | |
658 | memset(bcbearer->bpairs, 0, sizeof(bcbearer->bpairs)); | |
659 | ||
660 | for (pri = TIPC_MAX_LINK_PRI; pri >= 0; pri--) { | |
661 | ||
662 | if (!bp_temp[pri].primary) | |
663 | continue; | |
664 | ||
665 | bp_curr->primary = bp_temp[pri].primary; | |
666 | ||
667 | if (bp_temp[pri].secondary) { | |
668 | if (tipc_nmap_equal(&bp_temp[pri].primary->nodes, | |
669 | &bp_temp[pri].secondary->nodes)) { | |
670 | bp_curr->secondary = bp_temp[pri].secondary; | |
671 | } else { | |
672 | bp_curr++; | |
673 | bp_curr->primary = bp_temp[pri].secondary; | |
674 | } | |
675 | } | |
676 | ||
677 | bp_curr++; | |
678 | } | |
679 | ||
680 | spin_unlock_bh(&bc_lock); | |
681 | } | |
682 | ||
683 | /** | |
684 | * tipc_bcbearer_push - resolve bearer congestion | |
685 | * | |
686 | * Forces bclink to push out any unsent packets, until all packets are gone | |
687 | * or congestion reoccurs. | |
688 | * No locks set when function called | |
689 | */ | |
690 | ||
691 | void tipc_bcbearer_push(void) | |
692 | { | |
693 | struct bearer *b_ptr; | |
694 | ||
695 | spin_lock_bh(&bc_lock); | |
696 | b_ptr = &bcbearer->bearer; | |
697 | if (b_ptr->publ.blocked) { | |
698 | b_ptr->publ.blocked = 0; | |
699 | tipc_bearer_lock_push(b_ptr); | |
700 | } | |
701 | spin_unlock_bh(&bc_lock); | |
702 | } | |
703 | ||
704 | ||
705 | int tipc_bclink_stats(char *buf, const u32 buf_size) | |
706 | { | |
707 | struct print_buf pb; | |
708 | ||
709 | if (!bcl) | |
710 | return 0; | |
711 | ||
712 | tipc_printbuf_init(&pb, buf, buf_size); | |
713 | ||
714 | spin_lock_bh(&bc_lock); | |
715 | ||
716 | tipc_printf(&pb, "Link <%s>\n" | |
717 | " Window:%u packets\n", | |
718 | bcl->name, bcl->queue_limit[0]); | |
719 | tipc_printf(&pb, " RX packets:%u fragments:%u/%u bundles:%u/%u\n", | |
720 | bcl->stats.recv_info, | |
721 | bcl->stats.recv_fragments, | |
722 | bcl->stats.recv_fragmented, | |
723 | bcl->stats.recv_bundles, | |
724 | bcl->stats.recv_bundled); | |
725 | tipc_printf(&pb, " TX packets:%u fragments:%u/%u bundles:%u/%u\n", | |
726 | bcl->stats.sent_info, | |
727 | bcl->stats.sent_fragments, | |
728 | bcl->stats.sent_fragmented, | |
729 | bcl->stats.sent_bundles, | |
730 | bcl->stats.sent_bundled); | |
731 | tipc_printf(&pb, " RX naks:%u defs:%u dups:%u\n", | |
732 | bcl->stats.recv_nacks, | |
733 | bcl->stats.deferred_recv, | |
734 | bcl->stats.duplicates); | |
735 | tipc_printf(&pb, " TX naks:%u acks:%u dups:%u\n", | |
736 | bcl->stats.sent_nacks, | |
737 | bcl->stats.sent_acks, | |
738 | bcl->stats.retransmitted); | |
739 | tipc_printf(&pb, " Congestion bearer:%u link:%u Send queue max:%u avg:%u\n", | |
740 | bcl->stats.bearer_congs, | |
741 | bcl->stats.link_congs, | |
742 | bcl->stats.max_queue_sz, | |
743 | bcl->stats.queue_sz_counts | |
744 | ? (bcl->stats.accu_queue_sz / bcl->stats.queue_sz_counts) | |
745 | : 0); | |
746 | ||
747 | spin_unlock_bh(&bc_lock); | |
748 | return tipc_printbuf_validate(&pb); | |
749 | } | |
750 | ||
751 | int tipc_bclink_reset_stats(void) | |
752 | { | |
753 | if (!bcl) | |
754 | return -ENOPROTOOPT; | |
755 | ||
756 | spin_lock_bh(&bc_lock); | |
757 | memset(&bcl->stats, 0, sizeof(bcl->stats)); | |
758 | spin_unlock_bh(&bc_lock); | |
759 | return 0; | |
760 | } | |
761 | ||
762 | int tipc_bclink_set_queue_limits(u32 limit) | |
763 | { | |
764 | if (!bcl) | |
765 | return -ENOPROTOOPT; | |
766 | if ((limit < TIPC_MIN_LINK_WIN) || (limit > TIPC_MAX_LINK_WIN)) | |
767 | return -EINVAL; | |
768 | ||
769 | spin_lock_bh(&bc_lock); | |
770 | tipc_link_set_queue_limits(bcl, limit); | |
771 | spin_unlock_bh(&bc_lock); | |
772 | return 0; | |
773 | } | |
774 | ||
775 | int tipc_bclink_init(void) | |
776 | { | |
777 | bcbearer = kzalloc(sizeof(*bcbearer), GFP_ATOMIC); | |
778 | bclink = kzalloc(sizeof(*bclink), GFP_ATOMIC); | |
779 | if (!bcbearer || !bclink) { | |
780 | nomem: | |
781 | warn("Multicast link creation failed, no memory\n"); | |
782 | kfree(bcbearer); | |
783 | bcbearer = NULL; | |
784 | kfree(bclink); | |
785 | bclink = NULL; | |
786 | return -ENOMEM; | |
787 | } | |
788 | ||
789 | INIT_LIST_HEAD(&bcbearer->bearer.cong_links); | |
790 | bcbearer->bearer.media = &bcbearer->media; | |
791 | bcbearer->media.send_msg = tipc_bcbearer_send; | |
792 | sprintf(bcbearer->media.name, "tipc-multicast"); | |
793 | ||
794 | bcl = &bclink->link; | |
795 | INIT_LIST_HEAD(&bcl->waiting_ports); | |
796 | bcl->next_out_no = 1; | |
797 | spin_lock_init(&bclink->node.lock); | |
798 | bcl->owner = &bclink->node; | |
799 | bcl->max_pkt = MAX_PKT_DEFAULT_MCAST; | |
800 | tipc_link_set_queue_limits(bcl, BCLINK_WIN_DEFAULT); | |
801 | bcl->b_ptr = &bcbearer->bearer; | |
802 | bcl->state = WORKING_WORKING; | |
803 | strlcpy(bcl->name, tipc_bclink_name, TIPC_MAX_LINK_NAME); | |
804 | ||
805 | if (BCLINK_LOG_BUF_SIZE) { | |
806 | char *pb = kmalloc(BCLINK_LOG_BUF_SIZE, GFP_ATOMIC); | |
807 | ||
808 | if (!pb) | |
809 | goto nomem; | |
810 | tipc_printbuf_init(&bcl->print_buf, pb, BCLINK_LOG_BUF_SIZE); | |
811 | } | |
812 | ||
813 | return 0; | |
814 | } | |
815 | ||
816 | void tipc_bclink_stop(void) | |
817 | { | |
818 | spin_lock_bh(&bc_lock); | |
819 | if (bcbearer) { | |
820 | tipc_link_stop(bcl); | |
821 | if (BCLINK_LOG_BUF_SIZE) | |
822 | kfree(bcl->print_buf.buf); | |
823 | bcl = NULL; | |
824 | kfree(bclink); | |
825 | bclink = NULL; | |
826 | kfree(bcbearer); | |
827 | bcbearer = NULL; | |
828 | } | |
829 | spin_unlock_bh(&bc_lock); | |
830 | } | |
831 |