aboutsummaryrefslogtreecommitdiffstats
path: root/net/bridge
diff options
context:
space:
mode:
authorHerbert Xu <herbert@gondor.apana.org.au>2010-02-27 14:41:45 -0500
committerDavid S. Miller <davem@davemloft.net>2010-02-28 03:48:45 -0500
commiteb1d16414339a6e113d89e2cca2556005d7ce919 (patch)
treeb6c49f870cc36241a894a8c723a38e224d41464b /net/bridge
parent025d89c27f54c69cd0e51666d88aada33666bb6c (diff)
bridge: Add core IGMP snooping support
This patch adds the core functionality of IGMP snooping support without actually hooking it up. So this patch should be a no-op as far as the bridge's external behaviour is concerned. All the new code and data is controlled by the Kconfig option BRIDGE_IGMP_SNOOPING. A run-time toggle is also available. The multicast switching is done using an hash table that is lockless on the read-side through RCU. On the write-side the new multicast_lock is used for all operations. The hash table supports dynamic growth/rehashing. The hash table will be rehashed if any chain length exceeds a preset limit. If rehashing does not reduce the maximum chain length then snooping will be disabled. These features may be added in future (in no particular order): * IGMPv3 source support * Non-querier router detection * IPv6 Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net/bridge')
-rw-r--r--net/bridge/Kconfig12
-rw-r--r--net/bridge/Makefile2
-rw-r--r--net/bridge/br_multicast.c1135
-rw-r--r--net/bridge/br_private.h139
4 files changed, 1288 insertions, 0 deletions
diff --git a/net/bridge/Kconfig b/net/bridge/Kconfig
index e143ca678881..78dd5497210a 100644
--- a/net/bridge/Kconfig
+++ b/net/bridge/Kconfig
@@ -31,3 +31,15 @@ config BRIDGE
31 will be called bridge. 31 will be called bridge.
32 32
33 If unsure, say N. 33 If unsure, say N.
34
35config BRIDGE_IGMP_SNOOPING
36 bool "IGMP snooping"
37 default y
38 ---help---
39 If you say Y here, then the Ethernet bridge will be able selectively
40 forward multicast traffic based on IGMP traffic received from each
41 port.
42
43 Say N to exclude this support and reduce the binary size.
44
45 If unsure, say Y.
diff --git a/net/bridge/Makefile b/net/bridge/Makefile
index f444c12cde5a..d0359ea8ee79 100644
--- a/net/bridge/Makefile
+++ b/net/bridge/Makefile
@@ -12,4 +12,6 @@ bridge-$(CONFIG_SYSFS) += br_sysfs_if.o br_sysfs_br.o
12 12
13bridge-$(CONFIG_BRIDGE_NETFILTER) += br_netfilter.o 13bridge-$(CONFIG_BRIDGE_NETFILTER) += br_netfilter.o
14 14
15bridge-$(CONFIG_BRIDGE_IGMP_SNOOPING) += br_multicast.o
16
15obj-$(CONFIG_BRIDGE_NF_EBTABLES) += netfilter/ 17obj-$(CONFIG_BRIDGE_NF_EBTABLES) += netfilter/
diff --git a/net/bridge/br_multicast.c b/net/bridge/br_multicast.c
new file mode 100644
index 000000000000..746b5a611aae
--- /dev/null
+++ b/net/bridge/br_multicast.c
@@ -0,0 +1,1135 @@
1/*
2 * Bridge multicast support.
3 *
4 * Copyright (c) 2010 Herbert Xu <herbert@gondor.apana.org.au>
5 *
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License as published by the Free
8 * Software Foundation; either version 2 of the License, or (at your option)
9 * any later version.
10 *
11 */
12
13#include <linux/err.h>
14#include <linux/if_ether.h>
15#include <linux/igmp.h>
16#include <linux/jhash.h>
17#include <linux/kernel.h>
18#include <linux/netdevice.h>
19#include <linux/netfilter_bridge.h>
20#include <linux/random.h>
21#include <linux/rculist.h>
22#include <linux/skbuff.h>
23#include <linux/slab.h>
24#include <linux/timer.h>
25#include <net/ip.h>
26
27#include "br_private.h"
28
29static inline int br_ip_hash(struct net_bridge_mdb_htable *mdb, __be32 ip)
30{
31 return jhash_1word(mdb->secret, (u32)ip) & (mdb->max - 1);
32}
33
34static struct net_bridge_mdb_entry *__br_mdb_ip_get(
35 struct net_bridge_mdb_htable *mdb, __be32 dst, int hash)
36{
37 struct net_bridge_mdb_entry *mp;
38 struct hlist_node *p;
39
40 hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
41 if (dst == mp->addr)
42 return mp;
43 }
44
45 return NULL;
46}
47
48static struct net_bridge_mdb_entry *br_mdb_ip_get(
49 struct net_bridge_mdb_htable *mdb, __be32 dst)
50{
51 return __br_mdb_ip_get(mdb, dst, br_ip_hash(mdb, dst));
52}
53
54struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
55 struct sk_buff *skb)
56{
57 struct net_bridge_mdb_htable *mdb = br->mdb;
58
59 if (!mdb || br->multicast_disabled)
60 return NULL;
61
62 switch (skb->protocol) {
63 case htons(ETH_P_IP):
64 if (BR_INPUT_SKB_CB(skb)->igmp)
65 break;
66 return br_mdb_ip_get(mdb, ip_hdr(skb)->daddr);
67 }
68
69 return NULL;
70}
71
72static void br_mdb_free(struct rcu_head *head)
73{
74 struct net_bridge_mdb_htable *mdb =
75 container_of(head, struct net_bridge_mdb_htable, rcu);
76 struct net_bridge_mdb_htable *old = mdb->old;
77
78 mdb->old = NULL;
79 kfree(old->mhash);
80 kfree(old);
81}
82
83static int br_mdb_copy(struct net_bridge_mdb_htable *new,
84 struct net_bridge_mdb_htable *old,
85 int elasticity)
86{
87 struct net_bridge_mdb_entry *mp;
88 struct hlist_node *p;
89 int maxlen;
90 int len;
91 int i;
92
93 for (i = 0; i < old->max; i++)
94 hlist_for_each_entry(mp, p, &old->mhash[i], hlist[old->ver])
95 hlist_add_head(&mp->hlist[new->ver],
96 &new->mhash[br_ip_hash(new, mp->addr)]);
97
98 if (!elasticity)
99 return 0;
100
101 maxlen = 0;
102 for (i = 0; i < new->max; i++) {
103 len = 0;
104 hlist_for_each_entry(mp, p, &new->mhash[i], hlist[new->ver])
105 len++;
106 if (len > maxlen)
107 maxlen = len;
108 }
109
110 return maxlen > elasticity ? -EINVAL : 0;
111}
112
113static void br_multicast_free_pg(struct rcu_head *head)
114{
115 struct net_bridge_port_group *p =
116 container_of(head, struct net_bridge_port_group, rcu);
117
118 kfree(p);
119}
120
121static void br_multicast_free_group(struct rcu_head *head)
122{
123 struct net_bridge_mdb_entry *mp =
124 container_of(head, struct net_bridge_mdb_entry, rcu);
125
126 kfree(mp);
127}
128
129static void br_multicast_group_expired(unsigned long data)
130{
131 struct net_bridge_mdb_entry *mp = (void *)data;
132 struct net_bridge *br = mp->br;
133 struct net_bridge_mdb_htable *mdb;
134
135 spin_lock(&br->multicast_lock);
136 if (!netif_running(br->dev) || timer_pending(&mp->timer))
137 goto out;
138
139 if (!hlist_unhashed(&mp->mglist))
140 hlist_del_init(&mp->mglist);
141
142 if (mp->ports)
143 goto out;
144
145 mdb = br->mdb;
146 hlist_del_rcu(&mp->hlist[mdb->ver]);
147 mdb->size--;
148
149 del_timer(&mp->query_timer);
150 call_rcu_bh(&mp->rcu, br_multicast_free_group);
151
152out:
153 spin_unlock(&br->multicast_lock);
154}
155
156static void br_multicast_del_pg(struct net_bridge *br,
157 struct net_bridge_port_group *pg)
158{
159 struct net_bridge_mdb_htable *mdb = br->mdb;
160 struct net_bridge_mdb_entry *mp;
161 struct net_bridge_port_group *p;
162 struct net_bridge_port_group **pp;
163
164 mp = br_mdb_ip_get(mdb, pg->addr);
165 if (WARN_ON(!mp))
166 return;
167
168 for (pp = &mp->ports; (p = *pp); pp = &p->next) {
169 if (p != pg)
170 continue;
171
172 *pp = p->next;
173 hlist_del_init(&p->mglist);
174 del_timer(&p->timer);
175 del_timer(&p->query_timer);
176 call_rcu_bh(&p->rcu, br_multicast_free_pg);
177
178 if (!mp->ports && hlist_unhashed(&mp->mglist) &&
179 netif_running(br->dev))
180 mod_timer(&mp->timer, jiffies);
181
182 return;
183 }
184
185 WARN_ON(1);
186}
187
188static void br_multicast_port_group_expired(unsigned long data)
189{
190 struct net_bridge_port_group *pg = (void *)data;
191 struct net_bridge *br = pg->port->br;
192
193 spin_lock(&br->multicast_lock);
194 if (!netif_running(br->dev) || timer_pending(&pg->timer) ||
195 hlist_unhashed(&pg->mglist))
196 goto out;
197
198 br_multicast_del_pg(br, pg);
199
200out:
201 spin_unlock(&br->multicast_lock);
202}
203
204static int br_mdb_rehash(struct net_bridge_mdb_htable **mdbp, int max,
205 int elasticity)
206{
207 struct net_bridge_mdb_htable *old = *mdbp;
208 struct net_bridge_mdb_htable *mdb;
209 int err;
210
211 mdb = kmalloc(sizeof(*mdb), GFP_ATOMIC);
212 if (!mdb)
213 return -ENOMEM;
214
215 mdb->max = max;
216 mdb->old = old;
217
218 mdb->mhash = kzalloc(max * sizeof(*mdb->mhash), GFP_ATOMIC);
219 if (!mdb->mhash) {
220 kfree(mdb);
221 return -ENOMEM;
222 }
223
224 mdb->size = old ? old->size : 0;
225 mdb->ver = old ? old->ver ^ 1 : 0;
226
227 if (!old || elasticity)
228 get_random_bytes(&mdb->secret, sizeof(mdb->secret));
229 else
230 mdb->secret = old->secret;
231
232 if (!old)
233 goto out;
234
235 err = br_mdb_copy(mdb, old, elasticity);
236 if (err) {
237 kfree(mdb->mhash);
238 kfree(mdb);
239 return err;
240 }
241
242 call_rcu_bh(&mdb->rcu, br_mdb_free);
243
244out:
245 rcu_assign_pointer(*mdbp, mdb);
246
247 return 0;
248}
249
250static struct sk_buff *br_multicast_alloc_query(struct net_bridge *br,
251 __be32 group)
252{
253 struct sk_buff *skb;
254 struct igmphdr *ih;
255 struct ethhdr *eth;
256 struct iphdr *iph;
257
258 skb = netdev_alloc_skb_ip_align(br->dev, sizeof(*eth) + sizeof(*iph) +
259 sizeof(*ih) + 4);
260 if (!skb)
261 goto out;
262
263 skb->protocol = htons(ETH_P_IP);
264
265 skb_reset_mac_header(skb);
266 eth = eth_hdr(skb);
267
268 memcpy(eth->h_source, br->dev->dev_addr, 6);
269 eth->h_dest[0] = 1;
270 eth->h_dest[1] = 0;
271 eth->h_dest[2] = 0x5e;
272 eth->h_dest[3] = 0;
273 eth->h_dest[4] = 0;
274 eth->h_dest[5] = 1;
275 eth->h_proto = htons(ETH_P_IP);
276 skb_put(skb, sizeof(*eth));
277
278 skb_set_network_header(skb, skb->len);
279 iph = ip_hdr(skb);
280
281 iph->version = 4;
282 iph->ihl = 6;
283 iph->tos = 0xc0;
284 iph->tot_len = htons(sizeof(*iph) + sizeof(*ih) + 4);
285 iph->id = 0;
286 iph->frag_off = htons(IP_DF);
287 iph->ttl = 1;
288 iph->protocol = IPPROTO_IGMP;
289 iph->saddr = 0;
290 iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
291 ((u8 *)&iph[1])[0] = IPOPT_RA;
292 ((u8 *)&iph[1])[1] = 4;
293 ((u8 *)&iph[1])[2] = 0;
294 ((u8 *)&iph[1])[3] = 0;
295 ip_send_check(iph);
296 skb_put(skb, 24);
297
298 skb_set_transport_header(skb, skb->len);
299 ih = igmp_hdr(skb);
300 ih->type = IGMP_HOST_MEMBERSHIP_QUERY;
301 ih->code = (group ? br->multicast_last_member_interval :
302 br->multicast_query_response_interval) /
303 (HZ / IGMP_TIMER_SCALE);
304 ih->group = group;
305 ih->csum = 0;
306 ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
307 skb_put(skb, sizeof(*ih));
308
309 __skb_pull(skb, sizeof(*eth));
310
311out:
312 return skb;
313}
314
315static void br_multicast_send_group_query(struct net_bridge_mdb_entry *mp)
316{
317 struct net_bridge *br = mp->br;
318 struct sk_buff *skb;
319
320 skb = br_multicast_alloc_query(br, mp->addr);
321 if (!skb)
322 goto timer;
323
324 netif_rx(skb);
325
326timer:
327 if (++mp->queries_sent < br->multicast_last_member_count)
328 mod_timer(&mp->query_timer,
329 jiffies + br->multicast_last_member_interval);
330}
331
332static void br_multicast_group_query_expired(unsigned long data)
333{
334 struct net_bridge_mdb_entry *mp = (void *)data;
335 struct net_bridge *br = mp->br;
336
337 spin_lock(&br->multicast_lock);
338 if (!netif_running(br->dev) || hlist_unhashed(&mp->mglist) ||
339 mp->queries_sent >= br->multicast_last_member_count)
340 goto out;
341
342 br_multicast_send_group_query(mp);
343
344out:
345 spin_unlock(&br->multicast_lock);
346}
347
348static void br_multicast_send_port_group_query(struct net_bridge_port_group *pg)
349{
350 struct net_bridge_port *port = pg->port;
351 struct net_bridge *br = port->br;
352 struct sk_buff *skb;
353
354 skb = br_multicast_alloc_query(br, pg->addr);
355 if (!skb)
356 goto timer;
357
358 br_deliver(port, skb);
359
360timer:
361 if (++pg->queries_sent < br->multicast_last_member_count)
362 mod_timer(&pg->query_timer,
363 jiffies + br->multicast_last_member_interval);
364}
365
366static void br_multicast_port_group_query_expired(unsigned long data)
367{
368 struct net_bridge_port_group *pg = (void *)data;
369 struct net_bridge_port *port = pg->port;
370 struct net_bridge *br = port->br;
371
372 spin_lock(&br->multicast_lock);
373 if (!netif_running(br->dev) || hlist_unhashed(&pg->mglist) ||
374 pg->queries_sent >= br->multicast_last_member_count)
375 goto out;
376
377 br_multicast_send_port_group_query(pg);
378
379out:
380 spin_unlock(&br->multicast_lock);
381}
382
383static struct net_bridge_mdb_entry *br_multicast_get_group(
384 struct net_bridge *br, struct net_bridge_port *port, __be32 group,
385 int hash)
386{
387 struct net_bridge_mdb_htable *mdb = br->mdb;
388 struct net_bridge_mdb_entry *mp;
389 struct hlist_node *p;
390 unsigned count = 0;
391 unsigned max;
392 int elasticity;
393 int err;
394
395 hlist_for_each_entry(mp, p, &mdb->mhash[hash], hlist[mdb->ver]) {
396 count++;
397 if (unlikely(group == mp->addr)) {
398 return mp;
399 }
400 }
401
402 elasticity = 0;
403 max = mdb->max;
404
405 if (unlikely(count > br->hash_elasticity && count)) {
406 if (net_ratelimit())
407 printk(KERN_INFO "%s: Multicast hash table "
408 "chain limit reached: %s\n",
409 br->dev->name, port ? port->dev->name :
410 br->dev->name);
411
412 elasticity = br->hash_elasticity;
413 }
414
415 if (mdb->size >= max) {
416 max *= 2;
417 if (unlikely(max >= br->hash_max)) {
418 printk(KERN_WARNING "%s: Multicast hash table maximum "
419 "reached, disabling snooping: %s, %d\n",
420 br->dev->name, port ? port->dev->name :
421 br->dev->name,
422 max);
423 err = -E2BIG;
424disable:
425 br->multicast_disabled = 1;
426 goto err;
427 }
428 }
429
430 if (max > mdb->max || elasticity) {
431 if (mdb->old) {
432 if (net_ratelimit())
433 printk(KERN_INFO "%s: Multicast hash table "
434 "on fire: %s\n",
435 br->dev->name, port ? port->dev->name :
436 br->dev->name);
437 err = -EEXIST;
438 goto err;
439 }
440
441 err = br_mdb_rehash(&br->mdb, max, elasticity);
442 if (err) {
443 printk(KERN_WARNING "%s: Cannot rehash multicast "
444 "hash table, disabling snooping: "
445 "%s, %d, %d\n",
446 br->dev->name, port ? port->dev->name :
447 br->dev->name,
448 mdb->size, err);
449 goto disable;
450 }
451
452 err = -EAGAIN;
453 goto err;
454 }
455
456 return NULL;
457
458err:
459 mp = ERR_PTR(err);
460 return mp;
461}
462
463static struct net_bridge_mdb_entry *br_multicast_new_group(
464 struct net_bridge *br, struct net_bridge_port *port, __be32 group)
465{
466 struct net_bridge_mdb_htable *mdb = br->mdb;
467 struct net_bridge_mdb_entry *mp;
468 int hash;
469
470 if (!mdb) {
471 if (br_mdb_rehash(&br->mdb, BR_HASH_SIZE, 0))
472 return NULL;
473 goto rehash;
474 }
475
476 hash = br_ip_hash(mdb, group);
477 mp = br_multicast_get_group(br, port, group, hash);
478 switch (PTR_ERR(mp)) {
479 case 0:
480 break;
481
482 case -EAGAIN:
483rehash:
484 mdb = br->mdb;
485 hash = br_ip_hash(mdb, group);
486 break;
487
488 default:
489 goto out;
490 }
491
492 mp = kzalloc(sizeof(*mp), GFP_ATOMIC);
493 if (unlikely(!mp))
494 goto out;
495
496 mp->br = br;
497 mp->addr = group;
498 setup_timer(&mp->timer, br_multicast_group_expired,
499 (unsigned long)mp);
500 setup_timer(&mp->query_timer, br_multicast_group_query_expired,
501 (unsigned long)mp);
502
503 hlist_add_head_rcu(&mp->hlist[mdb->ver], &mdb->mhash[hash]);
504 mdb->size++;
505
506out:
507 return mp;
508}
509
510static int br_multicast_add_group(struct net_bridge *br,
511 struct net_bridge_port *port, __be32 group)
512{
513 struct net_bridge_mdb_entry *mp;
514 struct net_bridge_port_group *p;
515 struct net_bridge_port_group **pp;
516 unsigned long now = jiffies;
517 int err;
518
519 if (ipv4_is_local_multicast(group))
520 return 0;
521
522 spin_lock(&br->multicast_lock);
523 if (!netif_running(br->dev) ||
524 (port && port->state == BR_STATE_DISABLED))
525 goto out;
526
527 mp = br_multicast_new_group(br, port, group);
528 err = PTR_ERR(mp);
529 if (unlikely(IS_ERR(mp) || !mp))
530 goto err;
531
532 if (!port) {
533 hlist_add_head(&mp->mglist, &br->mglist);
534 mod_timer(&mp->timer, now + br->multicast_membership_interval);
535 goto out;
536 }
537
538 for (pp = &mp->ports; (p = *pp); pp = &p->next) {
539 if (p->port == port)
540 goto found;
541 if ((unsigned long)p->port < (unsigned long)port)
542 break;
543 }
544
545 p = kzalloc(sizeof(*p), GFP_ATOMIC);
546 err = -ENOMEM;
547 if (unlikely(!p))
548 goto err;
549
550 p->addr = group;
551 p->port = port;
552 p->next = *pp;
553 hlist_add_head(&p->mglist, &port->mglist);
554 setup_timer(&p->timer, br_multicast_port_group_expired,
555 (unsigned long)p);
556 setup_timer(&p->query_timer, br_multicast_port_group_query_expired,
557 (unsigned long)p);
558
559 rcu_assign_pointer(*pp, p);
560
561found:
562 mod_timer(&p->timer, now + br->multicast_membership_interval);
563out:
564 err = 0;
565
566err:
567 spin_unlock(&br->multicast_lock);
568 return err;
569}
570
571static void br_multicast_router_expired(unsigned long data)
572{
573 struct net_bridge_port *port = (void *)data;
574 struct net_bridge *br = port->br;
575
576 spin_lock(&br->multicast_lock);
577 if (port->multicast_router != 1 ||
578 timer_pending(&port->multicast_router_timer) ||
579 hlist_unhashed(&port->rlist))
580 goto out;
581
582 hlist_del_init_rcu(&port->rlist);
583
584out:
585 spin_unlock(&br->multicast_lock);
586}
587
588static void br_multicast_local_router_expired(unsigned long data)
589{
590}
591
592static void br_multicast_send_query(struct net_bridge *br,
593 struct net_bridge_port *port, u32 sent)
594{
595 unsigned long time;
596 struct sk_buff *skb;
597
598 if (!netif_running(br->dev) || br->multicast_disabled ||
599 timer_pending(&br->multicast_querier_timer))
600 return;
601
602 skb = br_multicast_alloc_query(br, 0);
603 if (!skb)
604 goto timer;
605
606 if (port) {
607 __skb_push(skb, sizeof(struct ethhdr));
608 skb->dev = port->dev;
609 NF_HOOK(PF_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
610 dev_queue_xmit);
611 } else
612 netif_rx(skb);
613
614timer:
615 time = jiffies;
616 time += sent < br->multicast_startup_query_count ?
617 br->multicast_startup_query_interval :
618 br->multicast_query_interval;
619 mod_timer(port ? &port->multicast_query_timer :
620 &br->multicast_query_timer, time);
621}
622
623static void br_multicast_port_query_expired(unsigned long data)
624{
625 struct net_bridge_port *port = (void *)data;
626 struct net_bridge *br = port->br;
627
628 spin_lock(&br->multicast_lock);
629 if (port && (port->state == BR_STATE_DISABLED ||
630 port->state == BR_STATE_BLOCKING))
631 goto out;
632
633 if (port->multicast_startup_queries_sent <
634 br->multicast_startup_query_count)
635 port->multicast_startup_queries_sent++;
636
637 br_multicast_send_query(port->br, port,
638 port->multicast_startup_queries_sent);
639
640out:
641 spin_unlock(&br->multicast_lock);
642}
643
644void br_multicast_add_port(struct net_bridge_port *port)
645{
646 port->multicast_router = 1;
647
648 setup_timer(&port->multicast_router_timer, br_multicast_router_expired,
649 (unsigned long)port);
650 setup_timer(&port->multicast_query_timer,
651 br_multicast_port_query_expired, (unsigned long)port);
652}
653
654void br_multicast_del_port(struct net_bridge_port *port)
655{
656 del_timer_sync(&port->multicast_router_timer);
657}
658
659void br_multicast_enable_port(struct net_bridge_port *port)
660{
661 struct net_bridge *br = port->br;
662
663 spin_lock(&br->multicast_lock);
664 if (br->multicast_disabled || !netif_running(br->dev))
665 goto out;
666
667 port->multicast_startup_queries_sent = 0;
668
669 if (try_to_del_timer_sync(&port->multicast_query_timer) >= 0 ||
670 del_timer(&port->multicast_query_timer))
671 mod_timer(&port->multicast_query_timer, jiffies);
672
673out:
674 spin_unlock(&br->multicast_lock);
675}
676
677void br_multicast_disable_port(struct net_bridge_port *port)
678{
679 struct net_bridge *br = port->br;
680 struct net_bridge_port_group *pg;
681 struct hlist_node *p, *n;
682
683 spin_lock(&br->multicast_lock);
684 hlist_for_each_entry_safe(pg, p, n, &port->mglist, mglist)
685 br_multicast_del_pg(br, pg);
686
687 if (!hlist_unhashed(&port->rlist))
688 hlist_del_init_rcu(&port->rlist);
689 del_timer(&port->multicast_router_timer);
690 del_timer(&port->multicast_query_timer);
691 spin_unlock(&br->multicast_lock);
692}
693
694static int br_multicast_igmp3_report(struct net_bridge *br,
695 struct net_bridge_port *port,
696 struct sk_buff *skb)
697{
698 struct igmpv3_report *ih;
699 struct igmpv3_grec *grec;
700 int i;
701 int len;
702 int num;
703 int type;
704 int err = 0;
705 __be32 group;
706
707 if (!pskb_may_pull(skb, sizeof(*ih)))
708 return -EINVAL;
709
710 ih = igmpv3_report_hdr(skb);
711 num = ntohs(ih->ngrec);
712 len = sizeof(*ih);
713
714 for (i = 0; i < num; i++) {
715 len += sizeof(*grec);
716 if (!pskb_may_pull(skb, len))
717 return -EINVAL;
718
719 grec = (void *)(skb->data + len);
720 group = grec->grec_mca;
721 type = grec->grec_type;
722
723 len += grec->grec_nsrcs * 4;
724 if (!pskb_may_pull(skb, len))
725 return -EINVAL;
726
727 /* We treat this as an IGMPv2 report for now. */
728 switch (type) {
729 case IGMPV3_MODE_IS_INCLUDE:
730 case IGMPV3_MODE_IS_EXCLUDE:
731 case IGMPV3_CHANGE_TO_INCLUDE:
732 case IGMPV3_CHANGE_TO_EXCLUDE:
733 case IGMPV3_ALLOW_NEW_SOURCES:
734 case IGMPV3_BLOCK_OLD_SOURCES:
735 break;
736
737 default:
738 continue;
739 }
740
741 err = br_multicast_add_group(br, port, group);
742 if (err)
743 break;
744 }
745
746 return err;
747}
748
749static void br_multicast_mark_router(struct net_bridge *br,
750 struct net_bridge_port *port)
751{
752 unsigned long now = jiffies;
753 struct hlist_node *p;
754 struct hlist_node **h;
755
756 if (!port) {
757 if (br->multicast_router == 1)
758 mod_timer(&br->multicast_router_timer,
759 now + br->multicast_querier_interval);
760 return;
761 }
762
763 if (port->multicast_router != 1)
764 return;
765
766 if (!hlist_unhashed(&port->rlist))
767 goto timer;
768
769 for (h = &br->router_list.first;
770 (p = *h) &&
771 (unsigned long)container_of(p, struct net_bridge_port, rlist) >
772 (unsigned long)port;
773 h = &p->next)
774 ;
775
776 port->rlist.pprev = h;
777 port->rlist.next = p;
778 rcu_assign_pointer(*h, &port->rlist);
779 if (p)
780 p->pprev = &port->rlist.next;
781
782timer:
783 mod_timer(&port->multicast_router_timer,
784 now + br->multicast_querier_interval);
785}
786
787static void br_multicast_query_received(struct net_bridge *br,
788 struct net_bridge_port *port,
789 __be32 saddr)
790{
791 if (saddr)
792 mod_timer(&br->multicast_querier_timer,
793 jiffies + br->multicast_querier_interval);
794 else if (timer_pending(&br->multicast_querier_timer))
795 return;
796
797 br_multicast_mark_router(br, port);
798}
799
800static int br_multicast_query(struct net_bridge *br,
801 struct net_bridge_port *port,
802 struct sk_buff *skb)
803{
804 struct iphdr *iph = ip_hdr(skb);
805 struct igmphdr *ih = igmp_hdr(skb);
806 struct net_bridge_mdb_entry *mp;
807 struct igmpv3_query *ih3;
808 struct net_bridge_port_group *p;
809 struct net_bridge_port_group **pp;
810 unsigned long max_delay;
811 unsigned long now = jiffies;
812 __be32 group;
813
814 spin_lock(&br->multicast_lock);
815 if (!netif_running(br->dev) ||
816 (port && port->state == BR_STATE_DISABLED))
817 goto out;
818
819 br_multicast_query_received(br, port, iph->saddr);
820
821 group = ih->group;
822
823 if (skb->len == sizeof(*ih)) {
824 max_delay = ih->code * (HZ / IGMP_TIMER_SCALE);
825
826 if (!max_delay) {
827 max_delay = 10 * HZ;
828 group = 0;
829 }
830 } else {
831 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
832 return -EINVAL;
833
834 ih3 = igmpv3_query_hdr(skb);
835 if (ih3->nsrcs)
836 return 0;
837
838 max_delay = ih3->code ? 1 :
839 IGMPV3_MRC(ih3->code) * (HZ / IGMP_TIMER_SCALE);
840 }
841
842 if (!group)
843 goto out;
844
845 mp = br_mdb_ip_get(br->mdb, group);
846 if (!mp)
847 goto out;
848
849 max_delay *= br->multicast_last_member_count;
850
851 if (!hlist_unhashed(&mp->mglist) &&
852 (timer_pending(&mp->timer) ?
853 time_after(mp->timer.expires, now + max_delay) :
854 try_to_del_timer_sync(&mp->timer) >= 0))
855 mod_timer(&mp->timer, now + max_delay);
856
857 for (pp = &mp->ports; (p = *pp); pp = &p->next) {
858 if (timer_pending(&p->timer) ?
859 time_after(p->timer.expires, now + max_delay) :
860 try_to_del_timer_sync(&p->timer) >= 0)
861 mod_timer(&mp->timer, now + max_delay);
862 }
863
864out:
865 spin_unlock(&br->multicast_lock);
866 return 0;
867}
868
869static void br_multicast_leave_group(struct net_bridge *br,
870 struct net_bridge_port *port,
871 __be32 group)
872{
873 struct net_bridge_mdb_htable *mdb;
874 struct net_bridge_mdb_entry *mp;
875 struct net_bridge_port_group *p;
876 unsigned long now;
877 unsigned long time;
878
879 if (ipv4_is_local_multicast(group))
880 return;
881
882 spin_lock(&br->multicast_lock);
883 if (!netif_running(br->dev) ||
884 (port && port->state == BR_STATE_DISABLED) ||
885 timer_pending(&br->multicast_querier_timer))
886 goto out;
887
888 mdb = br->mdb;
889 mp = br_mdb_ip_get(mdb, group);
890 if (!mp)
891 goto out;
892
893 now = jiffies;
894 time = now + br->multicast_last_member_count *
895 br->multicast_last_member_interval;
896
897 if (!port) {
898 if (!hlist_unhashed(&mp->mglist) &&
899 (timer_pending(&mp->timer) ?
900 time_after(mp->timer.expires, time) :
901 try_to_del_timer_sync(&mp->timer) >= 0)) {
902 mod_timer(&mp->timer, time);
903
904 mp->queries_sent = 0;
905 mod_timer(&mp->query_timer, now);
906 }
907
908 goto out;
909 }
910
911 for (p = mp->ports; p; p = p->next) {
912 if (p->port != port)
913 continue;
914
915 if (!hlist_unhashed(&p->mglist) &&
916 (timer_pending(&p->timer) ?
917 time_after(p->timer.expires, time) :
918 try_to_del_timer_sync(&p->timer) >= 0)) {
919 mod_timer(&p->timer, time);
920
921 p->queries_sent = 0;
922 mod_timer(&p->query_timer, now);
923 }
924
925 break;
926 }
927
928out:
929 spin_unlock(&br->multicast_lock);
930}
931
932static int br_multicast_ipv4_rcv(struct net_bridge *br,
933 struct net_bridge_port *port,
934 struct sk_buff *skb)
935{
936 struct sk_buff *skb2 = skb;
937 struct iphdr *iph;
938 struct igmphdr *ih;
939 unsigned len;
940 unsigned offset;
941 int err;
942
943 BR_INPUT_SKB_CB(skb)->igmp = 0;
944 BR_INPUT_SKB_CB(skb)->mrouters_only = 0;
945
946 /* We treat OOM as packet loss for now. */
947 if (!pskb_may_pull(skb, sizeof(*iph)))
948 return -EINVAL;
949
950 iph = ip_hdr(skb);
951
952 if (iph->ihl < 5 || iph->version != 4)
953 return -EINVAL;
954
955 if (!pskb_may_pull(skb, ip_hdrlen(skb)))
956 return -EINVAL;
957
958 iph = ip_hdr(skb);
959
960 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
961 return -EINVAL;
962
963 if (iph->protocol != IPPROTO_IGMP)
964 return 0;
965
966 len = ntohs(iph->tot_len);
967 if (skb->len < len || len < ip_hdrlen(skb))
968 return -EINVAL;
969
970 if (skb->len > len) {
971 skb2 = skb_clone(skb, GFP_ATOMIC);
972 if (!skb2)
973 return -ENOMEM;
974
975 err = pskb_trim_rcsum(skb2, len);
976 if (err)
977 return err;
978 }
979
980 len -= ip_hdrlen(skb2);
981 offset = skb_network_offset(skb2) + ip_hdrlen(skb2);
982 __skb_pull(skb2, offset);
983 skb_reset_transport_header(skb2);
984
985 err = -EINVAL;
986 if (!pskb_may_pull(skb2, sizeof(*ih)))
987 goto out;
988
989 iph = ip_hdr(skb2);
990
991 switch (skb2->ip_summed) {
992 case CHECKSUM_COMPLETE:
993 if (!csum_fold(skb2->csum))
994 break;
995 /* fall through */
996 case CHECKSUM_NONE:
997 skb2->csum = 0;
998 if (skb_checksum_complete(skb2))
999 return -EINVAL;
1000 }
1001
1002 err = 0;
1003
1004 BR_INPUT_SKB_CB(skb)->igmp = 1;
1005 ih = igmp_hdr(skb2);
1006
1007 switch (ih->type) {
1008 case IGMP_HOST_MEMBERSHIP_REPORT:
1009 case IGMPV2_HOST_MEMBERSHIP_REPORT:
1010 BR_INPUT_SKB_CB(skb2)->mrouters_only = 1;
1011 err = br_multicast_add_group(br, port, ih->group);
1012 break;
1013 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1014 err = br_multicast_igmp3_report(br, port, skb2);
1015 break;
1016 case IGMP_HOST_MEMBERSHIP_QUERY:
1017 err = br_multicast_query(br, port, skb2);
1018 break;
1019 case IGMP_HOST_LEAVE_MESSAGE:
1020 br_multicast_leave_group(br, port, ih->group);
1021 break;
1022 }
1023
1024out:
1025 __skb_push(skb2, offset);
1026 if (skb2 != skb)
1027 kfree_skb(skb2);
1028 return err;
1029}
1030
1031int br_multicast_rcv(struct net_bridge *br, struct net_bridge_port *port,
1032 struct sk_buff *skb)
1033{
1034 if (br->multicast_disabled)
1035 return 0;
1036
1037 switch (skb->protocol) {
1038 case htons(ETH_P_IP):
1039 return br_multicast_ipv4_rcv(br, port, skb);
1040 }
1041
1042 return 0;
1043}
1044
1045static void br_multicast_query_expired(unsigned long data)
1046{
1047 struct net_bridge *br = (void *)data;
1048
1049 spin_lock(&br->multicast_lock);
1050 if (br->multicast_startup_queries_sent <
1051 br->multicast_startup_query_count)
1052 br->multicast_startup_queries_sent++;
1053
1054 br_multicast_send_query(br, NULL, br->multicast_startup_queries_sent);
1055
1056 spin_unlock(&br->multicast_lock);
1057}
1058
1059void br_multicast_init(struct net_bridge *br)
1060{
1061 br->hash_elasticity = 4;
1062 br->hash_max = 512;
1063
1064 br->multicast_router = 1;
1065 br->multicast_last_member_count = 2;
1066 br->multicast_startup_query_count = 2;
1067
1068 br->multicast_last_member_interval = HZ;
1069 br->multicast_query_response_interval = 10 * HZ;
1070 br->multicast_startup_query_interval = 125 * HZ / 4;
1071 br->multicast_query_interval = 125 * HZ;
1072 br->multicast_querier_interval = 255 * HZ;
1073 br->multicast_membership_interval = 260 * HZ;
1074
1075 spin_lock_init(&br->multicast_lock);
1076 setup_timer(&br->multicast_router_timer,
1077 br_multicast_local_router_expired, 0);
1078 setup_timer(&br->multicast_querier_timer,
1079 br_multicast_local_router_expired, 0);
1080 setup_timer(&br->multicast_query_timer, br_multicast_query_expired,
1081 (unsigned long)br);
1082}
1083
1084void br_multicast_open(struct net_bridge *br)
1085{
1086 br->multicast_startup_queries_sent = 0;
1087
1088 if (br->multicast_disabled)
1089 return;
1090
1091 mod_timer(&br->multicast_query_timer, jiffies);
1092}
1093
1094void br_multicast_stop(struct net_bridge *br)
1095{
1096 struct net_bridge_mdb_htable *mdb;
1097 struct net_bridge_mdb_entry *mp;
1098 struct hlist_node *p, *n;
1099 u32 ver;
1100 int i;
1101
1102 del_timer_sync(&br->multicast_router_timer);
1103 del_timer_sync(&br->multicast_querier_timer);
1104 del_timer_sync(&br->multicast_query_timer);
1105
1106 spin_lock_bh(&br->multicast_lock);
1107 mdb = br->mdb;
1108 if (!mdb)
1109 goto out;
1110
1111 br->mdb = NULL;
1112
1113 ver = mdb->ver;
1114 for (i = 0; i < mdb->max; i++) {
1115 hlist_for_each_entry_safe(mp, p, n, &mdb->mhash[i],
1116 hlist[ver]) {
1117 del_timer(&mp->timer);
1118 del_timer(&mp->query_timer);
1119 call_rcu_bh(&mp->rcu, br_multicast_free_group);
1120 }
1121 }
1122
1123 if (mdb->old) {
1124 spin_unlock_bh(&br->multicast_lock);
1125 synchronize_rcu_bh();
1126 spin_lock_bh(&br->multicast_lock);
1127 WARN_ON(mdb->old);
1128 }
1129
1130 mdb->old = mdb;
1131 call_rcu_bh(&mdb->rcu, br_mdb_free);
1132
1133out:
1134 spin_unlock_bh(&br->multicast_lock);
1135}
diff --git a/net/bridge/br_private.h b/net/bridge/br_private.h
index fad5a2669d34..44345c9afdd3 100644
--- a/net/bridge/br_private.h
+++ b/net/bridge/br_private.h
@@ -57,6 +57,41 @@ struct net_bridge_fdb_entry
57 unsigned char is_static; 57 unsigned char is_static;
58}; 58};
59 59
60struct net_bridge_port_group {
61 struct net_bridge_port *port;
62 struct net_bridge_port_group *next;
63 struct hlist_node mglist;
64 struct rcu_head rcu;
65 struct timer_list timer;
66 struct timer_list query_timer;
67 __be32 addr;
68 u32 queries_sent;
69};
70
71struct net_bridge_mdb_entry
72{
73 struct hlist_node hlist[2];
74 struct hlist_node mglist;
75 struct net_bridge *br;
76 struct net_bridge_port_group *ports;
77 struct rcu_head rcu;
78 struct timer_list timer;
79 struct timer_list query_timer;
80 __be32 addr;
81 u32 queries_sent;
82};
83
84struct net_bridge_mdb_htable
85{
86 struct hlist_head *mhash;
87 struct rcu_head rcu;
88 struct net_bridge_mdb_htable *old;
89 u32 size;
90 u32 max;
91 u32 secret;
92 u32 ver;
93};
94
60struct net_bridge_port 95struct net_bridge_port
61{ 96{
62 struct net_bridge *br; 97 struct net_bridge *br;
@@ -84,6 +119,15 @@ struct net_bridge_port
84 119
85 unsigned long flags; 120 unsigned long flags;
86#define BR_HAIRPIN_MODE 0x00000001 121#define BR_HAIRPIN_MODE 0x00000001
122
123#ifdef CONFIG_BRIDGE_IGMP_SNOOPING
124 u32 multicast_startup_queries_sent;
125 unsigned char multicast_router;
126 struct timer_list multicast_router_timer;
127 struct timer_list multicast_query_timer;
128 struct hlist_head mglist;
129 struct hlist_node rlist;
130#endif
87}; 131};
88 132
89struct net_bridge 133struct net_bridge
@@ -124,6 +168,35 @@ struct net_bridge
124 unsigned char topology_change; 168 unsigned char topology_change;
125 unsigned char topology_change_detected; 169 unsigned char topology_change_detected;
126 170
171#ifdef CONFIG_BRIDGE_IGMP_SNOOPING
172 unsigned char multicast_router;
173
174 u8 multicast_disabled:1;
175
176 u32 hash_elasticity;
177 u32 hash_max;
178
179 u32 multicast_last_member_count;
180 u32 multicast_startup_queries_sent;
181 u32 multicast_startup_query_count;
182
183 unsigned long multicast_last_member_interval;
184 unsigned long multicast_membership_interval;
185 unsigned long multicast_querier_interval;
186 unsigned long multicast_query_interval;
187 unsigned long multicast_query_response_interval;
188 unsigned long multicast_startup_query_interval;
189
190 spinlock_t multicast_lock;
191 struct net_bridge_mdb_htable *mdb;
192 struct hlist_head router_list;
193 struct hlist_head mglist;
194
195 struct timer_list multicast_router_timer;
196 struct timer_list multicast_querier_timer;
197 struct timer_list multicast_query_timer;
198#endif
199
127 struct timer_list hello_timer; 200 struct timer_list hello_timer;
128 struct timer_list tcn_timer; 201 struct timer_list tcn_timer;
129 struct timer_list topology_change_timer; 202 struct timer_list topology_change_timer;
@@ -133,6 +206,8 @@ struct net_bridge
133 206
134struct br_input_skb_cb { 207struct br_input_skb_cb {
135 struct net_device *brdev; 208 struct net_device *brdev;
209 int igmp;
210 int mrouters_only;
136}; 211};
137 212
138#define BR_INPUT_SKB_CB(__skb) ((struct br_input_skb_cb *)(__skb)->cb) 213#define BR_INPUT_SKB_CB(__skb) ((struct br_input_skb_cb *)(__skb)->cb)
@@ -204,6 +279,70 @@ extern struct sk_buff *br_handle_frame(struct net_bridge_port *p,
204extern int br_dev_ioctl(struct net_device *dev, struct ifreq *rq, int cmd); 279extern int br_dev_ioctl(struct net_device *dev, struct ifreq *rq, int cmd);
205extern int br_ioctl_deviceless_stub(struct net *net, unsigned int cmd, void __user *arg); 280extern int br_ioctl_deviceless_stub(struct net *net, unsigned int cmd, void __user *arg);
206 281
282/* br_multicast.c */
283#ifdef CONFIG_BRIDGE_IGMP_SNOOPING
284extern int br_multicast_rcv(struct net_bridge *br,
285 struct net_bridge_port *port,
286 struct sk_buff *skb);
287extern struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
288 struct sk_buff *skb);
289extern void br_multicast_add_port(struct net_bridge_port *port);
290extern void br_multicast_del_port(struct net_bridge_port *port);
291extern void br_multicast_enable_port(struct net_bridge_port *port);
292extern void br_multicast_disable_port(struct net_bridge_port *port);
293extern void br_multicast_init(struct net_bridge *br);
294extern void br_multicast_open(struct net_bridge *br);
295extern void br_multicast_stop(struct net_bridge *br);
296#else
297static inline int br_multicast_rcv(struct net_bridge *br,
298 struct net_bridge_port *port,
299 struct sk_buff *skb)
300{
301 return 0;
302}
303
304static inline struct net_bridge_mdb_entry *br_mdb_get(struct net_bridge *br,
305 struct sk_buff *skb)
306{
307 return NULL;
308}
309
310static inline void br_multicast_add_port(struct net_bridge_port *port)
311{
312}
313
314static inline void br_multicast_del_port(struct net_bridge_port *port)
315{
316}
317
318static inline void br_multicast_enable_port(struct net_bridge_port *port)
319{
320}
321
322static inline void br_multicast_disable_port(struct net_bridge_port *port)
323{
324}
325
326static inline void br_multicast_init(struct net_bridge *br)
327{
328}
329
330static inline void br_multicast_open(struct net_bridge *br)
331{
332}
333
334static inline void br_multicast_stop(struct net_bridge *br)
335{
336}
337#endif
338
339static inline bool br_multicast_is_router(struct net_bridge *br)
340{
341 return br->multicast_router == 2 ||
342 (br->multicast_router == 1 &&
343 timer_pending(&br->multicast_router_timer));
344}
345
207/* br_netfilter.c */ 346/* br_netfilter.c */
208#ifdef CONFIG_BRIDGE_NETFILTER 347#ifdef CONFIG_BRIDGE_NETFILTER
209extern int br_netfilter_init(void); 348extern int br_netfilter_init(void);
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-28 02:59:48 -0500