]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - net/ipv4/ipmr.c
projects / mirror_ubuntu-bionic-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
net: ipmr: move the tbl id check in ipmr_new_table
[mirror_ubuntu-bionic-kernel.git] / net / ipv4 / ipmr.c
1 /*
2 * IP multicast routing support for mrouted 3.6/3.8
3 *
4 * (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
5 * Linux Consultancy and Custom Driver Development
6 *
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
11 *
12 * Fixes:
13 * Michael Chastain : Incorrect size of copying.
14 * Alan Cox : Added the cache manager code
15 * Alan Cox : Fixed the clone/copy bug and device race.
16 * Mike McLagan : Routing by source
17 * Malcolm Beattie : Buffer handling fixes.
18 * Alexey Kuznetsov : Double buffer free and other fixes.
19 * SVR Anand : Fixed several multicast bugs and problems.
20 * Alexey Kuznetsov : Status, optimisations and more.
21 * Brad Parker : Better behaviour on mrouted upcall
22 * overflow.
23 * Carlos Picoto : PIMv1 Support
24 * Pavlin Ivanov Radoslavov: PIMv2 Registers must checksum only PIM header
25 * Relax this requirement to work with older peers.
26 *
27 */
28
29 #include <asm/uaccess.h>
30 #include <linux/types.h>
31 #include <linux/capability.h>
32 #include <linux/errno.h>
33 #include <linux/timer.h>
34 #include <linux/mm.h>
35 #include <linux/kernel.h>
36 #include <linux/fcntl.h>
37 #include <linux/stat.h>
38 #include <linux/socket.h>
39 #include <linux/in.h>
40 #include <linux/inet.h>
41 #include <linux/netdevice.h>
42 #include <linux/inetdevice.h>
43 #include <linux/igmp.h>
44 #include <linux/proc_fs.h>
45 #include <linux/seq_file.h>
46 #include <linux/mroute.h>
47 #include <linux/init.h>
48 #include <linux/if_ether.h>
49 #include <linux/slab.h>
50 #include <net/net_namespace.h>
51 #include <net/ip.h>
52 #include <net/protocol.h>
53 #include <linux/skbuff.h>
54 #include <net/route.h>
55 #include <net/sock.h>
56 #include <net/icmp.h>
57 #include <net/udp.h>
58 #include <net/raw.h>
59 #include <linux/notifier.h>
60 #include <linux/if_arp.h>
61 #include <linux/netfilter_ipv4.h>
62 #include <linux/compat.h>
63 #include <linux/export.h>
64 #include <net/ip_tunnels.h>
65 #include <net/checksum.h>
66 #include <net/netlink.h>
67 #include <net/fib_rules.h>
68 #include <linux/netconf.h>
69
70 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
71 #define CONFIG_IP_PIMSM 1
72 #endif
73
74 struct mr_table {
75 struct list_head list;
76 possible_net_t net;
77 u32 id;
78 struct sock __rcu *mroute_sk;
79 struct timer_list ipmr_expire_timer;
80 struct list_head mfc_unres_queue;
81 struct list_head mfc_cache_array[MFC_LINES];
82 struct vif_device vif_table[MAXVIFS];
83 int maxvif;
84 atomic_t cache_resolve_queue_len;
85 bool mroute_do_assert;
86 bool mroute_do_pim;
87 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
88 int mroute_reg_vif_num;
89 #endif
90 };
91
92 struct ipmr_rule {
93 struct fib_rule common;
94 };
95
96 struct ipmr_result {
97 struct mr_table *mrt;
98 };
99
100 /* Big lock, protecting vif table, mrt cache and mroute socket state.
101 * Note that the changes are semaphored via rtnl_lock.
102 */
103
104 static DEFINE_RWLOCK(mrt_lock);
105
106 /*
107 * Multicast router control variables
108 */
109
110 #define VIF_EXISTS(_mrt, _idx) ((_mrt)->vif_table[_idx].dev != NULL)
111
112 /* Special spinlock for queue of unresolved entries */
113 static DEFINE_SPINLOCK(mfc_unres_lock);
114
115 /* We return to original Alan's scheme. Hash table of resolved
116 * entries is changed only in process context and protected
117 * with weak lock mrt_lock. Queue of unresolved entries is protected
118 * with strong spinlock mfc_unres_lock.
119 *
120 * In this case data path is free of exclusive locks at all.
121 */
122
123 static struct kmem_cache *mrt_cachep __read_mostly;
124
125 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
126 static void ipmr_free_table(struct mr_table *mrt);
127
128 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
129 struct sk_buff *skb, struct mfc_cache *cache,
130 int local);
131 static int ipmr_cache_report(struct mr_table *mrt,
132 struct sk_buff *pkt, vifi_t vifi, int assert);
133 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
134 struct mfc_cache *c, struct rtmsg *rtm);
135 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
136 int cmd);
137 static void mroute_clean_tables(struct mr_table *mrt);
138 static void ipmr_expire_process(unsigned long arg);
139
140 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
141 #define ipmr_for_each_table(mrt, net) \
142 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
143
144 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
145 {
146 struct mr_table *mrt;
147
148 ipmr_for_each_table(mrt, net) {
149 if (mrt->id == id)
150 return mrt;
151 }
152 return NULL;
153 }
154
155 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
156 struct mr_table **mrt)
157 {
158 int err;
159 struct ipmr_result res;
160 struct fib_lookup_arg arg = {
161 .result = &res,
162 .flags = FIB_LOOKUP_NOREF,
163 };
164
165 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
166 flowi4_to_flowi(flp4), 0, &arg);
167 if (err < 0)
168 return err;
169 *mrt = res.mrt;
170 return 0;
171 }
172
173 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
174 int flags, struct fib_lookup_arg *arg)
175 {
176 struct ipmr_result *res = arg->result;
177 struct mr_table *mrt;
178
179 switch (rule->action) {
180 case FR_ACT_TO_TBL:
181 break;
182 case FR_ACT_UNREACHABLE:
183 return -ENETUNREACH;
184 case FR_ACT_PROHIBIT:
185 return -EACCES;
186 case FR_ACT_BLACKHOLE:
187 default:
188 return -EINVAL;
189 }
190
191 mrt = ipmr_get_table(rule->fr_net, rule->table);
192 if (!mrt)
193 return -EAGAIN;
194 res->mrt = mrt;
195 return 0;
196 }
197
198 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
199 {
200 return 1;
201 }
202
203 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
204 FRA_GENERIC_POLICY,
205 };
206
207 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
208 struct fib_rule_hdr *frh, struct nlattr **tb)
209 {
210 return 0;
211 }
212
213 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
214 struct nlattr **tb)
215 {
216 return 1;
217 }
218
219 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
220 struct fib_rule_hdr *frh)
221 {
222 frh->dst_len = 0;
223 frh->src_len = 0;
224 frh->tos = 0;
225 return 0;
226 }
227
228 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
229 .family = RTNL_FAMILY_IPMR,
230 .rule_size = sizeof(struct ipmr_rule),
231 .addr_size = sizeof(u32),
232 .action = ipmr_rule_action,
233 .match = ipmr_rule_match,
234 .configure = ipmr_rule_configure,
235 .compare = ipmr_rule_compare,
236 .fill = ipmr_rule_fill,
237 .nlgroup = RTNLGRP_IPV4_RULE,
238 .policy = ipmr_rule_policy,
239 .owner = THIS_MODULE,
240 };
241
242 static int __net_init ipmr_rules_init(struct net *net)
243 {
244 struct fib_rules_ops *ops;
245 struct mr_table *mrt;
246 int err;
247
248 ops = fib_rules_register(&ipmr_rules_ops_template, net);
249 if (IS_ERR(ops))
250 return PTR_ERR(ops);
251
252 INIT_LIST_HEAD(&net->ipv4.mr_tables);
253
254 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
255 if (IS_ERR(mrt)) {
256 err = PTR_ERR(mrt);
257 goto err1;
258 }
259
260 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
261 if (err < 0)
262 goto err2;
263
264 net->ipv4.mr_rules_ops = ops;
265 return 0;
266
267 err2:
268 ipmr_free_table(mrt);
269 err1:
270 fib_rules_unregister(ops);
271 return err;
272 }
273
274 static void __net_exit ipmr_rules_exit(struct net *net)
275 {
276 struct mr_table *mrt, *next;
277
278 rtnl_lock();
279 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
280 list_del(&mrt->list);
281 ipmr_free_table(mrt);
282 }
283 fib_rules_unregister(net->ipv4.mr_rules_ops);
284 rtnl_unlock();
285 }
286 #else
287 #define ipmr_for_each_table(mrt, net) \
288 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
289
290 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
291 {
292 return net->ipv4.mrt;
293 }
294
295 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
296 struct mr_table **mrt)
297 {
298 *mrt = net->ipv4.mrt;
299 return 0;
300 }
301
302 static int __net_init ipmr_rules_init(struct net *net)
303 {
304 struct mr_table *mrt;
305
306 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
307 if (IS_ERR(mrt))
308 return PTR_ERR(mrt);
309 net->ipv4.mrt = mrt;
310 return 0;
311 }
312
313 static void __net_exit ipmr_rules_exit(struct net *net)
314 {
315 rtnl_lock();
316 ipmr_free_table(net->ipv4.mrt);
317 net->ipv4.mrt = NULL;
318 rtnl_unlock();
319 }
320 #endif
321
322 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
323 {
324 struct mr_table *mrt;
325 unsigned int i;
326
327 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
328 if (id != RT_TABLE_DEFAULT && id >= 1000000000)
329 return ERR_PTR(-EINVAL);
330
331 mrt = ipmr_get_table(net, id);
332 if (mrt)
333 return mrt;
334
335 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
336 if (!mrt)
337 return ERR_PTR(-ENOMEM);
338 write_pnet(&mrt->net, net);
339 mrt->id = id;
340
341 /* Forwarding cache */
342 for (i = 0; i < MFC_LINES; i++)
343 INIT_LIST_HEAD(&mrt->mfc_cache_array[i]);
344
345 INIT_LIST_HEAD(&mrt->mfc_unres_queue);
346
347 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
348 (unsigned long)mrt);
349
350 #ifdef CONFIG_IP_PIMSM
351 mrt->mroute_reg_vif_num = -1;
352 #endif
353 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
354 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
355 #endif
356 return mrt;
357 }
358
359 static void ipmr_free_table(struct mr_table *mrt)
360 {
361 del_timer_sync(&mrt->ipmr_expire_timer);
362 mroute_clean_tables(mrt);
363 kfree(mrt);
364 }
365
366 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
367
368 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
369 {
370 struct net *net = dev_net(dev);
371
372 dev_close(dev);
373
374 dev = __dev_get_by_name(net, "tunl0");
375 if (dev) {
376 const struct net_device_ops *ops = dev->netdev_ops;
377 struct ifreq ifr;
378 struct ip_tunnel_parm p;
379
380 memset(&p, 0, sizeof(p));
381 p.iph.daddr = v->vifc_rmt_addr.s_addr;
382 p.iph.saddr = v->vifc_lcl_addr.s_addr;
383 p.iph.version = 4;
384 p.iph.ihl = 5;
385 p.iph.protocol = IPPROTO_IPIP;
386 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
387 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
388
389 if (ops->ndo_do_ioctl) {
390 mm_segment_t oldfs = get_fs();
391
392 set_fs(KERNEL_DS);
393 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
394 set_fs(oldfs);
395 }
396 }
397 }
398
399 static
400 struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
401 {
402 struct net_device *dev;
403
404 dev = __dev_get_by_name(net, "tunl0");
405
406 if (dev) {
407 const struct net_device_ops *ops = dev->netdev_ops;
408 int err;
409 struct ifreq ifr;
410 struct ip_tunnel_parm p;
411 struct in_device *in_dev;
412
413 memset(&p, 0, sizeof(p));
414 p.iph.daddr = v->vifc_rmt_addr.s_addr;
415 p.iph.saddr = v->vifc_lcl_addr.s_addr;
416 p.iph.version = 4;
417 p.iph.ihl = 5;
418 p.iph.protocol = IPPROTO_IPIP;
419 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
420 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
421
422 if (ops->ndo_do_ioctl) {
423 mm_segment_t oldfs = get_fs();
424
425 set_fs(KERNEL_DS);
426 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
427 set_fs(oldfs);
428 } else {
429 err = -EOPNOTSUPP;
430 }
431 dev = NULL;
432
433 if (err == 0 &&
434 (dev = __dev_get_by_name(net, p.name)) != NULL) {
435 dev->flags |= IFF_MULTICAST;
436
437 in_dev = __in_dev_get_rtnl(dev);
438 if (!in_dev)
439 goto failure;
440
441 ipv4_devconf_setall(in_dev);
442 neigh_parms_data_state_setall(in_dev->arp_parms);
443 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
444
445 if (dev_open(dev))
446 goto failure;
447 dev_hold(dev);
448 }
449 }
450 return dev;
451
452 failure:
453 /* allow the register to be completed before unregistering. */
454 rtnl_unlock();
455 rtnl_lock();
456
457 unregister_netdevice(dev);
458 return NULL;
459 }
460
461 #ifdef CONFIG_IP_PIMSM
462
463 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
464 {
465 struct net *net = dev_net(dev);
466 struct mr_table *mrt;
467 struct flowi4 fl4 = {
468 .flowi4_oif = dev->ifindex,
469 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
470 .flowi4_mark = skb->mark,
471 };
472 int err;
473
474 err = ipmr_fib_lookup(net, &fl4, &mrt);
475 if (err < 0) {
476 kfree_skb(skb);
477 return err;
478 }
479
480 read_lock(&mrt_lock);
481 dev->stats.tx_bytes += skb->len;
482 dev->stats.tx_packets++;
483 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
484 read_unlock(&mrt_lock);
485 kfree_skb(skb);
486 return NETDEV_TX_OK;
487 }
488
489 static int reg_vif_get_iflink(const struct net_device *dev)
490 {
491 return 0;
492 }
493
494 static const struct net_device_ops reg_vif_netdev_ops = {
495 .ndo_start_xmit = reg_vif_xmit,
496 .ndo_get_iflink = reg_vif_get_iflink,
497 };
498
499 static void reg_vif_setup(struct net_device *dev)
500 {
501 dev->type = ARPHRD_PIMREG;
502 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
503 dev->flags = IFF_NOARP;
504 dev->netdev_ops = &reg_vif_netdev_ops;
505 dev->destructor = free_netdev;
506 dev->features |= NETIF_F_NETNS_LOCAL;
507 }
508
509 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
510 {
511 struct net_device *dev;
512 struct in_device *in_dev;
513 char name[IFNAMSIZ];
514
515 if (mrt->id == RT_TABLE_DEFAULT)
516 sprintf(name, "pimreg");
517 else
518 sprintf(name, "pimreg%u", mrt->id);
519
520 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
521
522 if (!dev)
523 return NULL;
524
525 dev_net_set(dev, net);
526
527 if (register_netdevice(dev)) {
528 free_netdev(dev);
529 return NULL;
530 }
531
532 rcu_read_lock();
533 in_dev = __in_dev_get_rcu(dev);
534 if (!in_dev) {
535 rcu_read_unlock();
536 goto failure;
537 }
538
539 ipv4_devconf_setall(in_dev);
540 neigh_parms_data_state_setall(in_dev->arp_parms);
541 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
542 rcu_read_unlock();
543
544 if (dev_open(dev))
545 goto failure;
546
547 dev_hold(dev);
548
549 return dev;
550
551 failure:
552 /* allow the register to be completed before unregistering. */
553 rtnl_unlock();
554 rtnl_lock();
555
556 unregister_netdevice(dev);
557 return NULL;
558 }
559 #endif
560
561 /**
562 * vif_delete - Delete a VIF entry
563 * @notify: Set to 1, if the caller is a notifier_call
564 */
565
566 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
567 struct list_head *head)
568 {
569 struct vif_device *v;
570 struct net_device *dev;
571 struct in_device *in_dev;
572
573 if (vifi < 0 || vifi >= mrt->maxvif)
574 return -EADDRNOTAVAIL;
575
576 v = &mrt->vif_table[vifi];
577
578 write_lock_bh(&mrt_lock);
579 dev = v->dev;
580 v->dev = NULL;
581
582 if (!dev) {
583 write_unlock_bh(&mrt_lock);
584 return -EADDRNOTAVAIL;
585 }
586
587 #ifdef CONFIG_IP_PIMSM
588 if (vifi == mrt->mroute_reg_vif_num)
589 mrt->mroute_reg_vif_num = -1;
590 #endif
591
592 if (vifi + 1 == mrt->maxvif) {
593 int tmp;
594
595 for (tmp = vifi - 1; tmp >= 0; tmp--) {
596 if (VIF_EXISTS(mrt, tmp))
597 break;
598 }
599 mrt->maxvif = tmp+1;
600 }
601
602 write_unlock_bh(&mrt_lock);
603
604 dev_set_allmulti(dev, -1);
605
606 in_dev = __in_dev_get_rtnl(dev);
607 if (in_dev) {
608 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
609 inet_netconf_notify_devconf(dev_net(dev),
610 NETCONFA_MC_FORWARDING,
611 dev->ifindex, &in_dev->cnf);
612 ip_rt_multicast_event(in_dev);
613 }
614
615 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
616 unregister_netdevice_queue(dev, head);
617
618 dev_put(dev);
619 return 0;
620 }
621
622 static void ipmr_cache_free_rcu(struct rcu_head *head)
623 {
624 struct mfc_cache *c = container_of(head, struct mfc_cache, rcu);
625
626 kmem_cache_free(mrt_cachep, c);
627 }
628
629 static inline void ipmr_cache_free(struct mfc_cache *c)
630 {
631 call_rcu(&c->rcu, ipmr_cache_free_rcu);
632 }
633
634 /* Destroy an unresolved cache entry, killing queued skbs
635 * and reporting error to netlink readers.
636 */
637
638 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
639 {
640 struct net *net = read_pnet(&mrt->net);
641 struct sk_buff *skb;
642 struct nlmsgerr *e;
643
644 atomic_dec(&mrt->cache_resolve_queue_len);
645
646 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) {
647 if (ip_hdr(skb)->version == 0) {
648 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
649 nlh->nlmsg_type = NLMSG_ERROR;
650 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
651 skb_trim(skb, nlh->nlmsg_len);
652 e = nlmsg_data(nlh);
653 e->error = -ETIMEDOUT;
654 memset(&e->msg, 0, sizeof(e->msg));
655
656 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
657 } else {
658 kfree_skb(skb);
659 }
660 }
661
662 ipmr_cache_free(c);
663 }
664
665
666 /* Timer process for the unresolved queue. */
667
668 static void ipmr_expire_process(unsigned long arg)
669 {
670 struct mr_table *mrt = (struct mr_table *)arg;
671 unsigned long now;
672 unsigned long expires;
673 struct mfc_cache *c, *next;
674
675 if (!spin_trylock(&mfc_unres_lock)) {
676 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
677 return;
678 }
679
680 if (list_empty(&mrt->mfc_unres_queue))
681 goto out;
682
683 now = jiffies;
684 expires = 10*HZ;
685
686 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
687 if (time_after(c->mfc_un.unres.expires, now)) {
688 unsigned long interval = c->mfc_un.unres.expires - now;
689 if (interval < expires)
690 expires = interval;
691 continue;
692 }
693
694 list_del(&c->list);
695 mroute_netlink_event(mrt, c, RTM_DELROUTE);
696 ipmr_destroy_unres(mrt, c);
697 }
698
699 if (!list_empty(&mrt->mfc_unres_queue))
700 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
701
702 out:
703 spin_unlock(&mfc_unres_lock);
704 }
705
706 /* Fill oifs list. It is called under write locked mrt_lock. */
707
708 static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache,
709 unsigned char *ttls)
710 {
711 int vifi;
712
713 cache->mfc_un.res.minvif = MAXVIFS;
714 cache->mfc_un.res.maxvif = 0;
715 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
716
717 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
718 if (VIF_EXISTS(mrt, vifi) &&
719 ttls[vifi] && ttls[vifi] < 255) {
720 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
721 if (cache->mfc_un.res.minvif > vifi)
722 cache->mfc_un.res.minvif = vifi;
723 if (cache->mfc_un.res.maxvif <= vifi)
724 cache->mfc_un.res.maxvif = vifi + 1;
725 }
726 }
727 }
728
729 static int vif_add(struct net *net, struct mr_table *mrt,
730 struct vifctl *vifc, int mrtsock)
731 {
732 int vifi = vifc->vifc_vifi;
733 struct vif_device *v = &mrt->vif_table[vifi];
734 struct net_device *dev;
735 struct in_device *in_dev;
736 int err;
737
738 /* Is vif busy ? */
739 if (VIF_EXISTS(mrt, vifi))
740 return -EADDRINUSE;
741
742 switch (vifc->vifc_flags) {
743 #ifdef CONFIG_IP_PIMSM
744 case VIFF_REGISTER:
745 /*
746 * Special Purpose VIF in PIM
747 * All the packets will be sent to the daemon
748 */
749 if (mrt->mroute_reg_vif_num >= 0)
750 return -EADDRINUSE;
751 dev = ipmr_reg_vif(net, mrt);
752 if (!dev)
753 return -ENOBUFS;
754 err = dev_set_allmulti(dev, 1);
755 if (err) {
756 unregister_netdevice(dev);
757 dev_put(dev);
758 return err;
759 }
760 break;
761 #endif
762 case VIFF_TUNNEL:
763 dev = ipmr_new_tunnel(net, vifc);
764 if (!dev)
765 return -ENOBUFS;
766 err = dev_set_allmulti(dev, 1);
767 if (err) {
768 ipmr_del_tunnel(dev, vifc);
769 dev_put(dev);
770 return err;
771 }
772 break;
773
774 case VIFF_USE_IFINDEX:
775 case 0:
776 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
777 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
778 if (dev && !__in_dev_get_rtnl(dev)) {
779 dev_put(dev);
780 return -EADDRNOTAVAIL;
781 }
782 } else {
783 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
784 }
785 if (!dev)
786 return -EADDRNOTAVAIL;
787 err = dev_set_allmulti(dev, 1);
788 if (err) {
789 dev_put(dev);
790 return err;
791 }
792 break;
793 default:
794 return -EINVAL;
795 }
796
797 in_dev = __in_dev_get_rtnl(dev);
798 if (!in_dev) {
799 dev_put(dev);
800 return -EADDRNOTAVAIL;
801 }
802 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
803 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, dev->ifindex,
804 &in_dev->cnf);
805 ip_rt_multicast_event(in_dev);
806
807 /* Fill in the VIF structures */
808
809 v->rate_limit = vifc->vifc_rate_limit;
810 v->local = vifc->vifc_lcl_addr.s_addr;
811 v->remote = vifc->vifc_rmt_addr.s_addr;
812 v->flags = vifc->vifc_flags;
813 if (!mrtsock)
814 v->flags |= VIFF_STATIC;
815 v->threshold = vifc->vifc_threshold;
816 v->bytes_in = 0;
817 v->bytes_out = 0;
818 v->pkt_in = 0;
819 v->pkt_out = 0;
820 v->link = dev->ifindex;
821 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER))
822 v->link = dev_get_iflink(dev);
823
824 /* And finish update writing critical data */
825 write_lock_bh(&mrt_lock);
826 v->dev = dev;
827 #ifdef CONFIG_IP_PIMSM
828 if (v->flags & VIFF_REGISTER)
829 mrt->mroute_reg_vif_num = vifi;
830 #endif
831 if (vifi+1 > mrt->maxvif)
832 mrt->maxvif = vifi+1;
833 write_unlock_bh(&mrt_lock);
834 return 0;
835 }
836
837 /* called with rcu_read_lock() */
838 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
839 __be32 origin,
840 __be32 mcastgrp)
841 {
842 int line = MFC_HASH(mcastgrp, origin);
843 struct mfc_cache *c;
844
845 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) {
846 if (c->mfc_origin == origin && c->mfc_mcastgrp == mcastgrp)
847 return c;
848 }
849 return NULL;
850 }
851
852 /* Look for a (*,*,oif) entry */
853 static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt,
854 int vifi)
855 {
856 int line = MFC_HASH(htonl(INADDR_ANY), htonl(INADDR_ANY));
857 struct mfc_cache *c;
858
859 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
860 if (c->mfc_origin == htonl(INADDR_ANY) &&
861 c->mfc_mcastgrp == htonl(INADDR_ANY) &&
862 c->mfc_un.res.ttls[vifi] < 255)
863 return c;
864
865 return NULL;
866 }
867
868 /* Look for a (*,G) entry */
869 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
870 __be32 mcastgrp, int vifi)
871 {
872 int line = MFC_HASH(mcastgrp, htonl(INADDR_ANY));
873 struct mfc_cache *c, *proxy;
874
875 if (mcastgrp == htonl(INADDR_ANY))
876 goto skip;
877
878 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
879 if (c->mfc_origin == htonl(INADDR_ANY) &&
880 c->mfc_mcastgrp == mcastgrp) {
881 if (c->mfc_un.res.ttls[vifi] < 255)
882 return c;
883
884 /* It's ok if the vifi is part of the static tree */
885 proxy = ipmr_cache_find_any_parent(mrt,
886 c->mfc_parent);
887 if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
888 return c;
889 }
890
891 skip:
892 return ipmr_cache_find_any_parent(mrt, vifi);
893 }
894
895 /*
896 * Allocate a multicast cache entry
897 */
898 static struct mfc_cache *ipmr_cache_alloc(void)
899 {
900 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
901
902 if (c)
903 c->mfc_un.res.minvif = MAXVIFS;
904 return c;
905 }
906
907 static struct mfc_cache *ipmr_cache_alloc_unres(void)
908 {
909 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
910
911 if (c) {
912 skb_queue_head_init(&c->mfc_un.unres.unresolved);
913 c->mfc_un.unres.expires = jiffies + 10*HZ;
914 }
915 return c;
916 }
917
918 /*
919 * A cache entry has gone into a resolved state from queued
920 */
921
922 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
923 struct mfc_cache *uc, struct mfc_cache *c)
924 {
925 struct sk_buff *skb;
926 struct nlmsgerr *e;
927
928 /* Play the pending entries through our router */
929
930 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
931 if (ip_hdr(skb)->version == 0) {
932 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
933
934 if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
935 nlh->nlmsg_len = skb_tail_pointer(skb) -
936 (u8 *)nlh;
937 } else {
938 nlh->nlmsg_type = NLMSG_ERROR;
939 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
940 skb_trim(skb, nlh->nlmsg_len);
941 e = nlmsg_data(nlh);
942 e->error = -EMSGSIZE;
943 memset(&e->msg, 0, sizeof(e->msg));
944 }
945
946 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
947 } else {
948 ip_mr_forward(net, mrt, skb, c, 0);
949 }
950 }
951 }
952
953 /*
954 * Bounce a cache query up to mrouted. We could use netlink for this but mrouted
955 * expects the following bizarre scheme.
956 *
957 * Called under mrt_lock.
958 */
959
960 static int ipmr_cache_report(struct mr_table *mrt,
961 struct sk_buff *pkt, vifi_t vifi, int assert)
962 {
963 struct sk_buff *skb;
964 const int ihl = ip_hdrlen(pkt);
965 struct igmphdr *igmp;
966 struct igmpmsg *msg;
967 struct sock *mroute_sk;
968 int ret;
969
970 #ifdef CONFIG_IP_PIMSM
971 if (assert == IGMPMSG_WHOLEPKT)
972 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
973 else
974 #endif
975 skb = alloc_skb(128, GFP_ATOMIC);
976
977 if (!skb)
978 return -ENOBUFS;
979
980 #ifdef CONFIG_IP_PIMSM
981 if (assert == IGMPMSG_WHOLEPKT) {
982 /* Ugly, but we have no choice with this interface.
983 * Duplicate old header, fix ihl, length etc.
984 * And all this only to mangle msg->im_msgtype and
985 * to set msg->im_mbz to "mbz" :-)
986 */
987 skb_push(skb, sizeof(struct iphdr));
988 skb_reset_network_header(skb);
989 skb_reset_transport_header(skb);
990 msg = (struct igmpmsg *)skb_network_header(skb);
991 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
992 msg->im_msgtype = IGMPMSG_WHOLEPKT;
993 msg->im_mbz = 0;
994 msg->im_vif = mrt->mroute_reg_vif_num;
995 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
996 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
997 sizeof(struct iphdr));
998 } else
999 #endif
1000 {
1001
1002 /* Copy the IP header */
1003
1004 skb_set_network_header(skb, skb->len);
1005 skb_put(skb, ihl);
1006 skb_copy_to_linear_data(skb, pkt->data, ihl);
1007 ip_hdr(skb)->protocol = 0; /* Flag to the kernel this is a route add */
1008 msg = (struct igmpmsg *)skb_network_header(skb);
1009 msg->im_vif = vifi;
1010 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1011
1012 /* Add our header */
1013
1014 igmp = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
1015 igmp->type =
1016 msg->im_msgtype = assert;
1017 igmp->code = 0;
1018 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1019 skb->transport_header = skb->network_header;
1020 }
1021
1022 rcu_read_lock();
1023 mroute_sk = rcu_dereference(mrt->mroute_sk);
1024 if (!mroute_sk) {
1025 rcu_read_unlock();
1026 kfree_skb(skb);
1027 return -EINVAL;
1028 }
1029
1030 /* Deliver to mrouted */
1031
1032 ret = sock_queue_rcv_skb(mroute_sk, skb);
1033 rcu_read_unlock();
1034 if (ret < 0) {
1035 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1036 kfree_skb(skb);
1037 }
1038
1039 return ret;
1040 }
1041
1042 /*
1043 * Queue a packet for resolution. It gets locked cache entry!
1044 */
1045
1046 static int
1047 ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, struct sk_buff *skb)
1048 {
1049 bool found = false;
1050 int err;
1051 struct mfc_cache *c;
1052 const struct iphdr *iph = ip_hdr(skb);
1053
1054 spin_lock_bh(&mfc_unres_lock);
1055 list_for_each_entry(c, &mrt->mfc_unres_queue, list) {
1056 if (c->mfc_mcastgrp == iph->daddr &&
1057 c->mfc_origin == iph->saddr) {
1058 found = true;
1059 break;
1060 }
1061 }
1062
1063 if (!found) {
1064 /* Create a new entry if allowable */
1065
1066 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1067 (c = ipmr_cache_alloc_unres()) == NULL) {
1068 spin_unlock_bh(&mfc_unres_lock);
1069
1070 kfree_skb(skb);
1071 return -ENOBUFS;
1072 }
1073
1074 /* Fill in the new cache entry */
1075
1076 c->mfc_parent = -1;
1077 c->mfc_origin = iph->saddr;
1078 c->mfc_mcastgrp = iph->daddr;
1079
1080 /* Reflect first query at mrouted. */
1081
1082 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1083 if (err < 0) {
1084 /* If the report failed throw the cache entry
1085 out - Brad Parker
1086 */
1087 spin_unlock_bh(&mfc_unres_lock);
1088
1089 ipmr_cache_free(c);
1090 kfree_skb(skb);
1091 return err;
1092 }
1093
1094 atomic_inc(&mrt->cache_resolve_queue_len);
1095 list_add(&c->list, &mrt->mfc_unres_queue);
1096 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1097
1098 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1099 mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires);
1100 }
1101
1102 /* See if we can append the packet */
1103
1104 if (c->mfc_un.unres.unresolved.qlen > 3) {
1105 kfree_skb(skb);
1106 err = -ENOBUFS;
1107 } else {
1108 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1109 err = 0;
1110 }
1111
1112 spin_unlock_bh(&mfc_unres_lock);
1113 return err;
1114 }
1115
1116 /*
1117 * MFC cache manipulation by user space mroute daemon
1118 */
1119
1120 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1121 {
1122 int line;
1123 struct mfc_cache *c, *next;
1124
1125 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1126
1127 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[line], list) {
1128 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1129 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1130 (parent == -1 || parent == c->mfc_parent)) {
1131 list_del_rcu(&c->list);
1132 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1133 ipmr_cache_free(c);
1134 return 0;
1135 }
1136 }
1137 return -ENOENT;
1138 }
1139
1140 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1141 struct mfcctl *mfc, int mrtsock, int parent)
1142 {
1143 bool found = false;
1144 int line;
1145 struct mfc_cache *uc, *c;
1146
1147 if (mfc->mfcc_parent >= MAXVIFS)
1148 return -ENFILE;
1149
1150 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1151
1152 list_for_each_entry(c, &mrt->mfc_cache_array[line], list) {
1153 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1154 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1155 (parent == -1 || parent == c->mfc_parent)) {
1156 found = true;
1157 break;
1158 }
1159 }
1160
1161 if (found) {
1162 write_lock_bh(&mrt_lock);
1163 c->mfc_parent = mfc->mfcc_parent;
1164 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1165 if (!mrtsock)
1166 c->mfc_flags |= MFC_STATIC;
1167 write_unlock_bh(&mrt_lock);
1168 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1169 return 0;
1170 }
1171
1172 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1173 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1174 return -EINVAL;
1175
1176 c = ipmr_cache_alloc();
1177 if (!c)
1178 return -ENOMEM;
1179
1180 c->mfc_origin = mfc->mfcc_origin.s_addr;
1181 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1182 c->mfc_parent = mfc->mfcc_parent;
1183 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1184 if (!mrtsock)
1185 c->mfc_flags |= MFC_STATIC;
1186
1187 list_add_rcu(&c->list, &mrt->mfc_cache_array[line]);
1188
1189 /*
1190 * Check to see if we resolved a queued list. If so we
1191 * need to send on the frames and tidy up.
1192 */
1193 found = false;
1194 spin_lock_bh(&mfc_unres_lock);
1195 list_for_each_entry(uc, &mrt->mfc_unres_queue, list) {
1196 if (uc->mfc_origin == c->mfc_origin &&
1197 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1198 list_del(&uc->list);
1199 atomic_dec(&mrt->cache_resolve_queue_len);
1200 found = true;
1201 break;
1202 }
1203 }
1204 if (list_empty(&mrt->mfc_unres_queue))
1205 del_timer(&mrt->ipmr_expire_timer);
1206 spin_unlock_bh(&mfc_unres_lock);
1207
1208 if (found) {
1209 ipmr_cache_resolve(net, mrt, uc, c);
1210 ipmr_cache_free(uc);
1211 }
1212 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1213 return 0;
1214 }
1215
1216 /*
1217 * Close the multicast socket, and clear the vif tables etc
1218 */
1219
1220 static void mroute_clean_tables(struct mr_table *mrt)
1221 {
1222 int i;
1223 LIST_HEAD(list);
1224 struct mfc_cache *c, *next;
1225
1226 /* Shut down all active vif entries */
1227
1228 for (i = 0; i < mrt->maxvif; i++) {
1229 if (!(mrt->vif_table[i].flags & VIFF_STATIC))
1230 vif_delete(mrt, i, 0, &list);
1231 }
1232 unregister_netdevice_many(&list);
1233
1234 /* Wipe the cache */
1235
1236 for (i = 0; i < MFC_LINES; i++) {
1237 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[i], list) {
1238 if (c->mfc_flags & MFC_STATIC)
1239 continue;
1240 list_del_rcu(&c->list);
1241 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1242 ipmr_cache_free(c);
1243 }
1244 }
1245
1246 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1247 spin_lock_bh(&mfc_unres_lock);
1248 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
1249 list_del(&c->list);
1250 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1251 ipmr_destroy_unres(mrt, c);
1252 }
1253 spin_unlock_bh(&mfc_unres_lock);
1254 }
1255 }
1256
1257 /* called from ip_ra_control(), before an RCU grace period,
1258 * we dont need to call synchronize_rcu() here
1259 */
1260 static void mrtsock_destruct(struct sock *sk)
1261 {
1262 struct net *net = sock_net(sk);
1263 struct mr_table *mrt;
1264
1265 rtnl_lock();
1266 ipmr_for_each_table(mrt, net) {
1267 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1268 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1269 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1270 NETCONFA_IFINDEX_ALL,
1271 net->ipv4.devconf_all);
1272 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1273 mroute_clean_tables(mrt);
1274 }
1275 }
1276 rtnl_unlock();
1277 }
1278
1279 /*
1280 * Socket options and virtual interface manipulation. The whole
1281 * virtual interface system is a complete heap, but unfortunately
1282 * that's how BSD mrouted happens to think. Maybe one day with a proper
1283 * MOSPF/PIM router set up we can clean this up.
1284 */
1285
1286 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1287 {
1288 int ret, parent = 0;
1289 struct vifctl vif;
1290 struct mfcctl mfc;
1291 struct net *net = sock_net(sk);
1292 struct mr_table *mrt;
1293
1294 if (sk->sk_type != SOCK_RAW ||
1295 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1296 return -EOPNOTSUPP;
1297
1298 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1299 if (!mrt)
1300 return -ENOENT;
1301
1302 if (optname != MRT_INIT) {
1303 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1304 !ns_capable(net->user_ns, CAP_NET_ADMIN))
1305 return -EACCES;
1306 }
1307
1308 switch (optname) {
1309 case MRT_INIT:
1310 if (optlen != sizeof(int))
1311 return -EINVAL;
1312
1313 rtnl_lock();
1314 if (rtnl_dereference(mrt->mroute_sk)) {
1315 rtnl_unlock();
1316 return -EADDRINUSE;
1317 }
1318
1319 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1320 if (ret == 0) {
1321 rcu_assign_pointer(mrt->mroute_sk, sk);
1322 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1323 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1324 NETCONFA_IFINDEX_ALL,
1325 net->ipv4.devconf_all);
1326 }
1327 rtnl_unlock();
1328 return ret;
1329 case MRT_DONE:
1330 if (sk != rcu_access_pointer(mrt->mroute_sk))
1331 return -EACCES;
1332 return ip_ra_control(sk, 0, NULL);
1333 case MRT_ADD_VIF:
1334 case MRT_DEL_VIF:
1335 if (optlen != sizeof(vif))
1336 return -EINVAL;
1337 if (copy_from_user(&vif, optval, sizeof(vif)))
1338 return -EFAULT;
1339 if (vif.vifc_vifi >= MAXVIFS)
1340 return -ENFILE;
1341 rtnl_lock();
1342 if (optname == MRT_ADD_VIF) {
1343 ret = vif_add(net, mrt, &vif,
1344 sk == rtnl_dereference(mrt->mroute_sk));
1345 } else {
1346 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1347 }
1348 rtnl_unlock();
1349 return ret;
1350
1351 /*
1352 * Manipulate the forwarding caches. These live
1353 * in a sort of kernel/user symbiosis.
1354 */
1355 case MRT_ADD_MFC:
1356 case MRT_DEL_MFC:
1357 parent = -1;
1358 case MRT_ADD_MFC_PROXY:
1359 case MRT_DEL_MFC_PROXY:
1360 if (optlen != sizeof(mfc))
1361 return -EINVAL;
1362 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1363 return -EFAULT;
1364 if (parent == 0)
1365 parent = mfc.mfcc_parent;
1366 rtnl_lock();
1367 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1368 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1369 else
1370 ret = ipmr_mfc_add(net, mrt, &mfc,
1371 sk == rtnl_dereference(mrt->mroute_sk),
1372 parent);
1373 rtnl_unlock();
1374 return ret;
1375 /*
1376 * Control PIM assert.
1377 */
1378 case MRT_ASSERT:
1379 {
1380 int v;
1381 if (optlen != sizeof(v))
1382 return -EINVAL;
1383 if (get_user(v, (int __user *)optval))
1384 return -EFAULT;
1385 mrt->mroute_do_assert = v;
1386 return 0;
1387 }
1388 #ifdef CONFIG_IP_PIMSM
1389 case MRT_PIM:
1390 {
1391 int v;
1392
1393 if (optlen != sizeof(v))
1394 return -EINVAL;
1395 if (get_user(v, (int __user *)optval))
1396 return -EFAULT;
1397 v = !!v;
1398
1399 rtnl_lock();
1400 ret = 0;
1401 if (v != mrt->mroute_do_pim) {
1402 mrt->mroute_do_pim = v;
1403 mrt->mroute_do_assert = v;
1404 }
1405 rtnl_unlock();
1406 return ret;
1407 }
1408 #endif
1409 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
1410 case MRT_TABLE:
1411 {
1412 u32 v;
1413
1414 if (optlen != sizeof(u32))
1415 return -EINVAL;
1416 if (get_user(v, (u32 __user *)optval))
1417 return -EFAULT;
1418
1419 rtnl_lock();
1420 ret = 0;
1421 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1422 ret = -EBUSY;
1423 } else {
1424 mrt = ipmr_new_table(net, v);
1425 if (IS_ERR(mrt))
1426 ret = PTR_ERR(mrt);
1427 else
1428 raw_sk(sk)->ipmr_table = v;
1429 }
1430 rtnl_unlock();
1431 return ret;
1432 }
1433 #endif
1434 /*
1435 * Spurious command, or MRT_VERSION which you cannot
1436 * set.
1437 */
1438 default:
1439 return -ENOPROTOOPT;
1440 }
1441 }
1442
1443 /*
1444 * Getsock opt support for the multicast routing system.
1445 */
1446
1447 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1448 {
1449 int olr;
1450 int val;
1451 struct net *net = sock_net(sk);
1452 struct mr_table *mrt;
1453
1454 if (sk->sk_type != SOCK_RAW ||
1455 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1456 return -EOPNOTSUPP;
1457
1458 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1459 if (!mrt)
1460 return -ENOENT;
1461
1462 if (optname != MRT_VERSION &&
1463 #ifdef CONFIG_IP_PIMSM
1464 optname != MRT_PIM &&
1465 #endif
1466 optname != MRT_ASSERT)
1467 return -ENOPROTOOPT;
1468
1469 if (get_user(olr, optlen))
1470 return -EFAULT;
1471
1472 olr = min_t(unsigned int, olr, sizeof(int));
1473 if (olr < 0)
1474 return -EINVAL;
1475
1476 if (put_user(olr, optlen))
1477 return -EFAULT;
1478 if (optname == MRT_VERSION)
1479 val = 0x0305;
1480 #ifdef CONFIG_IP_PIMSM
1481 else if (optname == MRT_PIM)
1482 val = mrt->mroute_do_pim;
1483 #endif
1484 else
1485 val = mrt->mroute_do_assert;
1486 if (copy_to_user(optval, &val, olr))
1487 return -EFAULT;
1488 return 0;
1489 }
1490
1491 /*
1492 * The IP multicast ioctl support routines.
1493 */
1494
1495 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1496 {
1497 struct sioc_sg_req sr;
1498 struct sioc_vif_req vr;
1499 struct vif_device *vif;
1500 struct mfc_cache *c;
1501 struct net *net = sock_net(sk);
1502 struct mr_table *mrt;
1503
1504 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1505 if (!mrt)
1506 return -ENOENT;
1507
1508 switch (cmd) {
1509 case SIOCGETVIFCNT:
1510 if (copy_from_user(&vr, arg, sizeof(vr)))
1511 return -EFAULT;
1512 if (vr.vifi >= mrt->maxvif)
1513 return -EINVAL;
1514 read_lock(&mrt_lock);
1515 vif = &mrt->vif_table[vr.vifi];
1516 if (VIF_EXISTS(mrt, vr.vifi)) {
1517 vr.icount = vif->pkt_in;
1518 vr.ocount = vif->pkt_out;
1519 vr.ibytes = vif->bytes_in;
1520 vr.obytes = vif->bytes_out;
1521 read_unlock(&mrt_lock);
1522
1523 if (copy_to_user(arg, &vr, sizeof(vr)))
1524 return -EFAULT;
1525 return 0;
1526 }
1527 read_unlock(&mrt_lock);
1528 return -EADDRNOTAVAIL;
1529 case SIOCGETSGCNT:
1530 if (copy_from_user(&sr, arg, sizeof(sr)))
1531 return -EFAULT;
1532
1533 rcu_read_lock();
1534 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1535 if (c) {
1536 sr.pktcnt = c->mfc_un.res.pkt;
1537 sr.bytecnt = c->mfc_un.res.bytes;
1538 sr.wrong_if = c->mfc_un.res.wrong_if;
1539 rcu_read_unlock();
1540
1541 if (copy_to_user(arg, &sr, sizeof(sr)))
1542 return -EFAULT;
1543 return 0;
1544 }
1545 rcu_read_unlock();
1546 return -EADDRNOTAVAIL;
1547 default:
1548 return -ENOIOCTLCMD;
1549 }
1550 }
1551
1552 #ifdef CONFIG_COMPAT
1553 struct compat_sioc_sg_req {
1554 struct in_addr src;
1555 struct in_addr grp;
1556 compat_ulong_t pktcnt;
1557 compat_ulong_t bytecnt;
1558 compat_ulong_t wrong_if;
1559 };
1560
1561 struct compat_sioc_vif_req {
1562 vifi_t vifi; /* Which iface */
1563 compat_ulong_t icount;
1564 compat_ulong_t ocount;
1565 compat_ulong_t ibytes;
1566 compat_ulong_t obytes;
1567 };
1568
1569 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1570 {
1571 struct compat_sioc_sg_req sr;
1572 struct compat_sioc_vif_req vr;
1573 struct vif_device *vif;
1574 struct mfc_cache *c;
1575 struct net *net = sock_net(sk);
1576 struct mr_table *mrt;
1577
1578 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1579 if (!mrt)
1580 return -ENOENT;
1581
1582 switch (cmd) {
1583 case SIOCGETVIFCNT:
1584 if (copy_from_user(&vr, arg, sizeof(vr)))
1585 return -EFAULT;
1586 if (vr.vifi >= mrt->maxvif)
1587 return -EINVAL;
1588 read_lock(&mrt_lock);
1589 vif = &mrt->vif_table[vr.vifi];
1590 if (VIF_EXISTS(mrt, vr.vifi)) {
1591 vr.icount = vif->pkt_in;
1592 vr.ocount = vif->pkt_out;
1593 vr.ibytes = vif->bytes_in;
1594 vr.obytes = vif->bytes_out;
1595 read_unlock(&mrt_lock);
1596
1597 if (copy_to_user(arg, &vr, sizeof(vr)))
1598 return -EFAULT;
1599 return 0;
1600 }
1601 read_unlock(&mrt_lock);
1602 return -EADDRNOTAVAIL;
1603 case SIOCGETSGCNT:
1604 if (copy_from_user(&sr, arg, sizeof(sr)))
1605 return -EFAULT;
1606
1607 rcu_read_lock();
1608 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1609 if (c) {
1610 sr.pktcnt = c->mfc_un.res.pkt;
1611 sr.bytecnt = c->mfc_un.res.bytes;
1612 sr.wrong_if = c->mfc_un.res.wrong_if;
1613 rcu_read_unlock();
1614
1615 if (copy_to_user(arg, &sr, sizeof(sr)))
1616 return -EFAULT;
1617 return 0;
1618 }
1619 rcu_read_unlock();
1620 return -EADDRNOTAVAIL;
1621 default:
1622 return -ENOIOCTLCMD;
1623 }
1624 }
1625 #endif
1626
1627
1628 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1629 {
1630 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1631 struct net *net = dev_net(dev);
1632 struct mr_table *mrt;
1633 struct vif_device *v;
1634 int ct;
1635
1636 if (event != NETDEV_UNREGISTER)
1637 return NOTIFY_DONE;
1638
1639 ipmr_for_each_table(mrt, net) {
1640 v = &mrt->vif_table[0];
1641 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1642 if (v->dev == dev)
1643 vif_delete(mrt, ct, 1, NULL);
1644 }
1645 }
1646 return NOTIFY_DONE;
1647 }
1648
1649
1650 static struct notifier_block ip_mr_notifier = {
1651 .notifier_call = ipmr_device_event,
1652 };
1653
1654 /*
1655 * Encapsulate a packet by attaching a valid IPIP header to it.
1656 * This avoids tunnel drivers and other mess and gives us the speed so
1657 * important for multicast video.
1658 */
1659
1660 static void ip_encap(struct net *net, struct sk_buff *skb,
1661 __be32 saddr, __be32 daddr)
1662 {
1663 struct iphdr *iph;
1664 const struct iphdr *old_iph = ip_hdr(skb);
1665
1666 skb_push(skb, sizeof(struct iphdr));
1667 skb->transport_header = skb->network_header;
1668 skb_reset_network_header(skb);
1669 iph = ip_hdr(skb);
1670
1671 iph->version = 4;
1672 iph->tos = old_iph->tos;
1673 iph->ttl = old_iph->ttl;
1674 iph->frag_off = 0;
1675 iph->daddr = daddr;
1676 iph->saddr = saddr;
1677 iph->protocol = IPPROTO_IPIP;
1678 iph->ihl = 5;
1679 iph->tot_len = htons(skb->len);
1680 ip_select_ident(net, skb, NULL);
1681 ip_send_check(iph);
1682
1683 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1684 nf_reset(skb);
1685 }
1686
1687 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1688 struct sk_buff *skb)
1689 {
1690 struct ip_options *opt = &(IPCB(skb)->opt);
1691
1692 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1693 IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1694
1695 if (unlikely(opt->optlen))
1696 ip_forward_options(skb);
1697
1698 return dst_output(net, sk, skb);
1699 }
1700
1701 /*
1702 * Processing handlers for ipmr_forward
1703 */
1704
1705 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1706 struct sk_buff *skb, struct mfc_cache *c, int vifi)
1707 {
1708 const struct iphdr *iph = ip_hdr(skb);
1709 struct vif_device *vif = &mrt->vif_table[vifi];
1710 struct net_device *dev;
1711 struct rtable *rt;
1712 struct flowi4 fl4;
1713 int encap = 0;
1714
1715 if (!vif->dev)
1716 goto out_free;
1717
1718 #ifdef CONFIG_IP_PIMSM
1719 if (vif->flags & VIFF_REGISTER) {
1720 vif->pkt_out++;
1721 vif->bytes_out += skb->len;
1722 vif->dev->stats.tx_bytes += skb->len;
1723 vif->dev->stats.tx_packets++;
1724 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1725 goto out_free;
1726 }
1727 #endif
1728
1729 if (vif->flags & VIFF_TUNNEL) {
1730 rt = ip_route_output_ports(net, &fl4, NULL,
1731 vif->remote, vif->local,
1732 0, 0,
1733 IPPROTO_IPIP,
1734 RT_TOS(iph->tos), vif->link);
1735 if (IS_ERR(rt))
1736 goto out_free;
1737 encap = sizeof(struct iphdr);
1738 } else {
1739 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1740 0, 0,
1741 IPPROTO_IPIP,
1742 RT_TOS(iph->tos), vif->link);
1743 if (IS_ERR(rt))
1744 goto out_free;
1745 }
1746
1747 dev = rt->dst.dev;
1748
1749 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1750 /* Do not fragment multicasts. Alas, IPv4 does not
1751 * allow to send ICMP, so that packets will disappear
1752 * to blackhole.
1753 */
1754
1755 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1756 ip_rt_put(rt);
1757 goto out_free;
1758 }
1759
1760 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1761
1762 if (skb_cow(skb, encap)) {
1763 ip_rt_put(rt);
1764 goto out_free;
1765 }
1766
1767 vif->pkt_out++;
1768 vif->bytes_out += skb->len;
1769
1770 skb_dst_drop(skb);
1771 skb_dst_set(skb, &rt->dst);
1772 ip_decrease_ttl(ip_hdr(skb));
1773
1774 /* FIXME: forward and output firewalls used to be called here.
1775 * What do we do with netfilter? -- RR
1776 */
1777 if (vif->flags & VIFF_TUNNEL) {
1778 ip_encap(net, skb, vif->local, vif->remote);
1779 /* FIXME: extra output firewall step used to be here. --RR */
1780 vif->dev->stats.tx_packets++;
1781 vif->dev->stats.tx_bytes += skb->len;
1782 }
1783
1784 IPCB(skb)->flags |= IPSKB_FORWARDED;
1785
1786 /*
1787 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1788 * not only before forwarding, but after forwarding on all output
1789 * interfaces. It is clear, if mrouter runs a multicasting
1790 * program, it should receive packets not depending to what interface
1791 * program is joined.
1792 * If we will not make it, the program will have to join on all
1793 * interfaces. On the other hand, multihoming host (or router, but
1794 * not mrouter) cannot join to more than one interface - it will
1795 * result in receiving multiple packets.
1796 */
1797 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1798 net, NULL, skb, skb->dev, dev,
1799 ipmr_forward_finish);
1800 return;
1801
1802 out_free:
1803 kfree_skb(skb);
1804 }
1805
1806 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1807 {
1808 int ct;
1809
1810 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1811 if (mrt->vif_table[ct].dev == dev)
1812 break;
1813 }
1814 return ct;
1815 }
1816
1817 /* "local" means that we should preserve one skb (for local delivery) */
1818
1819 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1820 struct sk_buff *skb, struct mfc_cache *cache,
1821 int local)
1822 {
1823 int psend = -1;
1824 int vif, ct;
1825 int true_vifi = ipmr_find_vif(mrt, skb->dev);
1826
1827 vif = cache->mfc_parent;
1828 cache->mfc_un.res.pkt++;
1829 cache->mfc_un.res.bytes += skb->len;
1830
1831 if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1832 struct mfc_cache *cache_proxy;
1833
1834 /* For an (*,G) entry, we only check that the incomming
1835 * interface is part of the static tree.
1836 */
1837 cache_proxy = ipmr_cache_find_any_parent(mrt, vif);
1838 if (cache_proxy &&
1839 cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
1840 goto forward;
1841 }
1842
1843 /*
1844 * Wrong interface: drop packet and (maybe) send PIM assert.
1845 */
1846 if (mrt->vif_table[vif].dev != skb->dev) {
1847 if (rt_is_output_route(skb_rtable(skb))) {
1848 /* It is our own packet, looped back.
1849 * Very complicated situation...
1850 *
1851 * The best workaround until routing daemons will be
1852 * fixed is not to redistribute packet, if it was
1853 * send through wrong interface. It means, that
1854 * multicast applications WILL NOT work for
1855 * (S,G), which have default multicast route pointing
1856 * to wrong oif. In any case, it is not a good
1857 * idea to use multicasting applications on router.
1858 */
1859 goto dont_forward;
1860 }
1861
1862 cache->mfc_un.res.wrong_if++;
1863
1864 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1865 /* pimsm uses asserts, when switching from RPT to SPT,
1866 * so that we cannot check that packet arrived on an oif.
1867 * It is bad, but otherwise we would need to move pretty
1868 * large chunk of pimd to kernel. Ough... --ANK
1869 */
1870 (mrt->mroute_do_pim ||
1871 cache->mfc_un.res.ttls[true_vifi] < 255) &&
1872 time_after(jiffies,
1873 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1874 cache->mfc_un.res.last_assert = jiffies;
1875 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1876 }
1877 goto dont_forward;
1878 }
1879
1880 forward:
1881 mrt->vif_table[vif].pkt_in++;
1882 mrt->vif_table[vif].bytes_in += skb->len;
1883
1884 /*
1885 * Forward the frame
1886 */
1887 if (cache->mfc_origin == htonl(INADDR_ANY) &&
1888 cache->mfc_mcastgrp == htonl(INADDR_ANY)) {
1889 if (true_vifi >= 0 &&
1890 true_vifi != cache->mfc_parent &&
1891 ip_hdr(skb)->ttl >
1892 cache->mfc_un.res.ttls[cache->mfc_parent]) {
1893 /* It's an (*,*) entry and the packet is not coming from
1894 * the upstream: forward the packet to the upstream
1895 * only.
1896 */
1897 psend = cache->mfc_parent;
1898 goto last_forward;
1899 }
1900 goto dont_forward;
1901 }
1902 for (ct = cache->mfc_un.res.maxvif - 1;
1903 ct >= cache->mfc_un.res.minvif; ct--) {
1904 /* For (*,G) entry, don't forward to the incoming interface */
1905 if ((cache->mfc_origin != htonl(INADDR_ANY) ||
1906 ct != true_vifi) &&
1907 ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
1908 if (psend != -1) {
1909 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1910
1911 if (skb2)
1912 ipmr_queue_xmit(net, mrt, skb2, cache,
1913 psend);
1914 }
1915 psend = ct;
1916 }
1917 }
1918 last_forward:
1919 if (psend != -1) {
1920 if (local) {
1921 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1922
1923 if (skb2)
1924 ipmr_queue_xmit(net, mrt, skb2, cache, psend);
1925 } else {
1926 ipmr_queue_xmit(net, mrt, skb, cache, psend);
1927 return;
1928 }
1929 }
1930
1931 dont_forward:
1932 if (!local)
1933 kfree_skb(skb);
1934 }
1935
1936 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
1937 {
1938 struct rtable *rt = skb_rtable(skb);
1939 struct iphdr *iph = ip_hdr(skb);
1940 struct flowi4 fl4 = {
1941 .daddr = iph->daddr,
1942 .saddr = iph->saddr,
1943 .flowi4_tos = RT_TOS(iph->tos),
1944 .flowi4_oif = (rt_is_output_route(rt) ?
1945 skb->dev->ifindex : 0),
1946 .flowi4_iif = (rt_is_output_route(rt) ?
1947 LOOPBACK_IFINDEX :
1948 skb->dev->ifindex),
1949 .flowi4_mark = skb->mark,
1950 };
1951 struct mr_table *mrt;
1952 int err;
1953
1954 err = ipmr_fib_lookup(net, &fl4, &mrt);
1955 if (err)
1956 return ERR_PTR(err);
1957 return mrt;
1958 }
1959
1960 /*
1961 * Multicast packets for forwarding arrive here
1962 * Called with rcu_read_lock();
1963 */
1964
1965 int ip_mr_input(struct sk_buff *skb)
1966 {
1967 struct mfc_cache *cache;
1968 struct net *net = dev_net(skb->dev);
1969 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
1970 struct mr_table *mrt;
1971
1972 /* Packet is looped back after forward, it should not be
1973 * forwarded second time, but still can be delivered locally.
1974 */
1975 if (IPCB(skb)->flags & IPSKB_FORWARDED)
1976 goto dont_forward;
1977
1978 mrt = ipmr_rt_fib_lookup(net, skb);
1979 if (IS_ERR(mrt)) {
1980 kfree_skb(skb);
1981 return PTR_ERR(mrt);
1982 }
1983 if (!local) {
1984 if (IPCB(skb)->opt.router_alert) {
1985 if (ip_call_ra_chain(skb))
1986 return 0;
1987 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
1988 /* IGMPv1 (and broken IGMPv2 implementations sort of
1989 * Cisco IOS <= 11.2(8)) do not put router alert
1990 * option to IGMP packets destined to routable
1991 * groups. It is very bad, because it means
1992 * that we can forward NO IGMP messages.
1993 */
1994 struct sock *mroute_sk;
1995
1996 mroute_sk = rcu_dereference(mrt->mroute_sk);
1997 if (mroute_sk) {
1998 nf_reset(skb);
1999 raw_rcv(mroute_sk, skb);
2000 return 0;
2001 }
2002 }
2003 }
2004
2005 /* already under rcu_read_lock() */
2006 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2007 if (!cache) {
2008 int vif = ipmr_find_vif(mrt, skb->dev);
2009
2010 if (vif >= 0)
2011 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2012 vif);
2013 }
2014
2015 /*
2016 * No usable cache entry
2017 */
2018 if (!cache) {
2019 int vif;
2020
2021 if (local) {
2022 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2023 ip_local_deliver(skb);
2024 if (!skb2)
2025 return -ENOBUFS;
2026 skb = skb2;
2027 }
2028
2029 read_lock(&mrt_lock);
2030 vif = ipmr_find_vif(mrt, skb->dev);
2031 if (vif >= 0) {
2032 int err2 = ipmr_cache_unresolved(mrt, vif, skb);
2033 read_unlock(&mrt_lock);
2034
2035 return err2;
2036 }
2037 read_unlock(&mrt_lock);
2038 kfree_skb(skb);
2039 return -ENODEV;
2040 }
2041
2042 read_lock(&mrt_lock);
2043 ip_mr_forward(net, mrt, skb, cache, local);
2044 read_unlock(&mrt_lock);
2045
2046 if (local)
2047 return ip_local_deliver(skb);
2048
2049 return 0;
2050
2051 dont_forward:
2052 if (local)
2053 return ip_local_deliver(skb);
2054 kfree_skb(skb);
2055 return 0;
2056 }
2057
2058 #ifdef CONFIG_IP_PIMSM
2059 /* called with rcu_read_lock() */
2060 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
2061 unsigned int pimlen)
2062 {
2063 struct net_device *reg_dev = NULL;
2064 struct iphdr *encap;
2065
2066 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
2067 /*
2068 * Check that:
2069 * a. packet is really sent to a multicast group
2070 * b. packet is not a NULL-REGISTER
2071 * c. packet is not truncated
2072 */
2073 if (!ipv4_is_multicast(encap->daddr) ||
2074 encap->tot_len == 0 ||
2075 ntohs(encap->tot_len) + pimlen > skb->len)
2076 return 1;
2077
2078 read_lock(&mrt_lock);
2079 if (mrt->mroute_reg_vif_num >= 0)
2080 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
2081 read_unlock(&mrt_lock);
2082
2083 if (!reg_dev)
2084 return 1;
2085
2086 skb->mac_header = skb->network_header;
2087 skb_pull(skb, (u8 *)encap - skb->data);
2088 skb_reset_network_header(skb);
2089 skb->protocol = htons(ETH_P_IP);
2090 skb->ip_summed = CHECKSUM_NONE;
2091
2092 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
2093
2094 netif_rx(skb);
2095
2096 return NET_RX_SUCCESS;
2097 }
2098 #endif
2099
2100 #ifdef CONFIG_IP_PIMSM_V1
2101 /*
2102 * Handle IGMP messages of PIMv1
2103 */
2104
2105 int pim_rcv_v1(struct sk_buff *skb)
2106 {
2107 struct igmphdr *pim;
2108 struct net *net = dev_net(skb->dev);
2109 struct mr_table *mrt;
2110
2111 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2112 goto drop;
2113
2114 pim = igmp_hdr(skb);
2115
2116 mrt = ipmr_rt_fib_lookup(net, skb);
2117 if (IS_ERR(mrt))
2118 goto drop;
2119 if (!mrt->mroute_do_pim ||
2120 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2121 goto drop;
2122
2123 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2124 drop:
2125 kfree_skb(skb);
2126 }
2127 return 0;
2128 }
2129 #endif
2130
2131 #ifdef CONFIG_IP_PIMSM_V2
2132 static int pim_rcv(struct sk_buff *skb)
2133 {
2134 struct pimreghdr *pim;
2135 struct net *net = dev_net(skb->dev);
2136 struct mr_table *mrt;
2137
2138 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2139 goto drop;
2140
2141 pim = (struct pimreghdr *)skb_transport_header(skb);
2142 if (pim->type != ((PIM_VERSION << 4) | (PIM_REGISTER)) ||
2143 (pim->flags & PIM_NULL_REGISTER) ||
2144 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2145 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2146 goto drop;
2147
2148 mrt = ipmr_rt_fib_lookup(net, skb);
2149 if (IS_ERR(mrt))
2150 goto drop;
2151 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2152 drop:
2153 kfree_skb(skb);
2154 }
2155 return 0;
2156 }
2157 #endif
2158
2159 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2160 struct mfc_cache *c, struct rtmsg *rtm)
2161 {
2162 int ct;
2163 struct rtnexthop *nhp;
2164 struct nlattr *mp_attr;
2165 struct rta_mfc_stats mfcs;
2166
2167 /* If cache is unresolved, don't try to parse IIF and OIF */
2168 if (c->mfc_parent >= MAXVIFS)
2169 return -ENOENT;
2170
2171 if (VIF_EXISTS(mrt, c->mfc_parent) &&
2172 nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
2173 return -EMSGSIZE;
2174
2175 if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH)))
2176 return -EMSGSIZE;
2177
2178 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2179 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2180 if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) {
2181 nla_nest_cancel(skb, mp_attr);
2182 return -EMSGSIZE;
2183 }
2184
2185 nhp->rtnh_flags = 0;
2186 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2187 nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex;
2188 nhp->rtnh_len = sizeof(*nhp);
2189 }
2190 }
2191
2192 nla_nest_end(skb, mp_attr);
2193
2194 mfcs.mfcs_packets = c->mfc_un.res.pkt;
2195 mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2196 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2197 if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0)
2198 return -EMSGSIZE;
2199
2200 rtm->rtm_type = RTN_MULTICAST;
2201 return 1;
2202 }
2203
2204 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2205 __be32 saddr, __be32 daddr,
2206 struct rtmsg *rtm, int nowait)
2207 {
2208 struct mfc_cache *cache;
2209 struct mr_table *mrt;
2210 int err;
2211
2212 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2213 if (!mrt)
2214 return -ENOENT;
2215
2216 rcu_read_lock();
2217 cache = ipmr_cache_find(mrt, saddr, daddr);
2218 if (!cache && skb->dev) {
2219 int vif = ipmr_find_vif(mrt, skb->dev);
2220
2221 if (vif >= 0)
2222 cache = ipmr_cache_find_any(mrt, daddr, vif);
2223 }
2224 if (!cache) {
2225 struct sk_buff *skb2;
2226 struct iphdr *iph;
2227 struct net_device *dev;
2228 int vif = -1;
2229
2230 if (nowait) {
2231 rcu_read_unlock();
2232 return -EAGAIN;
2233 }
2234
2235 dev = skb->dev;
2236 read_lock(&mrt_lock);
2237 if (dev)
2238 vif = ipmr_find_vif(mrt, dev);
2239 if (vif < 0) {
2240 read_unlock(&mrt_lock);
2241 rcu_read_unlock();
2242 return -ENODEV;
2243 }
2244 skb2 = skb_clone(skb, GFP_ATOMIC);
2245 if (!skb2) {
2246 read_unlock(&mrt_lock);
2247 rcu_read_unlock();
2248 return -ENOMEM;
2249 }
2250
2251 skb_push(skb2, sizeof(struct iphdr));
2252 skb_reset_network_header(skb2);
2253 iph = ip_hdr(skb2);
2254 iph->ihl = sizeof(struct iphdr) >> 2;
2255 iph->saddr = saddr;
2256 iph->daddr = daddr;
2257 iph->version = 0;
2258 err = ipmr_cache_unresolved(mrt, vif, skb2);
2259 read_unlock(&mrt_lock);
2260 rcu_read_unlock();
2261 return err;
2262 }
2263
2264 read_lock(&mrt_lock);
2265 if (!nowait && (rtm->rtm_flags & RTM_F_NOTIFY))
2266 cache->mfc_flags |= MFC_NOTIFY;
2267 err = __ipmr_fill_mroute(mrt, skb, cache, rtm);
2268 read_unlock(&mrt_lock);
2269 rcu_read_unlock();
2270 return err;
2271 }
2272
2273 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2274 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2275 int flags)
2276 {
2277 struct nlmsghdr *nlh;
2278 struct rtmsg *rtm;
2279 int err;
2280
2281 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2282 if (!nlh)
2283 return -EMSGSIZE;
2284
2285 rtm = nlmsg_data(nlh);
2286 rtm->rtm_family = RTNL_FAMILY_IPMR;
2287 rtm->rtm_dst_len = 32;
2288 rtm->rtm_src_len = 32;
2289 rtm->rtm_tos = 0;
2290 rtm->rtm_table = mrt->id;
2291 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2292 goto nla_put_failure;
2293 rtm->rtm_type = RTN_MULTICAST;
2294 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2295 if (c->mfc_flags & MFC_STATIC)
2296 rtm->rtm_protocol = RTPROT_STATIC;
2297 else
2298 rtm->rtm_protocol = RTPROT_MROUTED;
2299 rtm->rtm_flags = 0;
2300
2301 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2302 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2303 goto nla_put_failure;
2304 err = __ipmr_fill_mroute(mrt, skb, c, rtm);
2305 /* do not break the dump if cache is unresolved */
2306 if (err < 0 && err != -ENOENT)
2307 goto nla_put_failure;
2308
2309 nlmsg_end(skb, nlh);
2310 return 0;
2311
2312 nla_put_failure:
2313 nlmsg_cancel(skb, nlh);
2314 return -EMSGSIZE;
2315 }
2316
2317 static size_t mroute_msgsize(bool unresolved, int maxvif)
2318 {
2319 size_t len =
2320 NLMSG_ALIGN(sizeof(struct rtmsg))
2321 + nla_total_size(4) /* RTA_TABLE */
2322 + nla_total_size(4) /* RTA_SRC */
2323 + nla_total_size(4) /* RTA_DST */
2324 ;
2325
2326 if (!unresolved)
2327 len = len
2328 + nla_total_size(4) /* RTA_IIF */
2329 + nla_total_size(0) /* RTA_MULTIPATH */
2330 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2331 /* RTA_MFC_STATS */
2332 + nla_total_size(sizeof(struct rta_mfc_stats))
2333 ;
2334
2335 return len;
2336 }
2337
2338 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2339 int cmd)
2340 {
2341 struct net *net = read_pnet(&mrt->net);
2342 struct sk_buff *skb;
2343 int err = -ENOBUFS;
2344
2345 skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif),
2346 GFP_ATOMIC);
2347 if (!skb)
2348 goto errout;
2349
2350 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2351 if (err < 0)
2352 goto errout;
2353
2354 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2355 return;
2356
2357 errout:
2358 kfree_skb(skb);
2359 if (err < 0)
2360 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2361 }
2362
2363 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2364 {
2365 struct net *net = sock_net(skb->sk);
2366 struct mr_table *mrt;
2367 struct mfc_cache *mfc;
2368 unsigned int t = 0, s_t;
2369 unsigned int h = 0, s_h;
2370 unsigned int e = 0, s_e;
2371
2372 s_t = cb->args[0];
2373 s_h = cb->args[1];
2374 s_e = cb->args[2];
2375
2376 rcu_read_lock();
2377 ipmr_for_each_table(mrt, net) {
2378 if (t < s_t)
2379 goto next_table;
2380 if (t > s_t)
2381 s_h = 0;
2382 for (h = s_h; h < MFC_LINES; h++) {
2383 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_array[h], list) {
2384 if (e < s_e)
2385 goto next_entry;
2386 if (ipmr_fill_mroute(mrt, skb,
2387 NETLINK_CB(cb->skb).portid,
2388 cb->nlh->nlmsg_seq,
2389 mfc, RTM_NEWROUTE,
2390 NLM_F_MULTI) < 0)
2391 goto done;
2392 next_entry:
2393 e++;
2394 }
2395 e = s_e = 0;
2396 }
2397 spin_lock_bh(&mfc_unres_lock);
2398 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
2399 if (e < s_e)
2400 goto next_entry2;
2401 if (ipmr_fill_mroute(mrt, skb,
2402 NETLINK_CB(cb->skb).portid,
2403 cb->nlh->nlmsg_seq,
2404 mfc, RTM_NEWROUTE,
2405 NLM_F_MULTI) < 0) {
2406 spin_unlock_bh(&mfc_unres_lock);
2407 goto done;
2408 }
2409 next_entry2:
2410 e++;
2411 }
2412 spin_unlock_bh(&mfc_unres_lock);
2413 e = s_e = 0;
2414 s_h = 0;
2415 next_table:
2416 t++;
2417 }
2418 done:
2419 rcu_read_unlock();
2420
2421 cb->args[2] = e;
2422 cb->args[1] = h;
2423 cb->args[0] = t;
2424
2425 return skb->len;
2426 }
2427
2428 #ifdef CONFIG_PROC_FS
2429 /*
2430 * The /proc interfaces to multicast routing :
2431 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2432 */
2433 struct ipmr_vif_iter {
2434 struct seq_net_private p;
2435 struct mr_table *mrt;
2436 int ct;
2437 };
2438
2439 static struct vif_device *ipmr_vif_seq_idx(struct net *net,
2440 struct ipmr_vif_iter *iter,
2441 loff_t pos)
2442 {
2443 struct mr_table *mrt = iter->mrt;
2444
2445 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
2446 if (!VIF_EXISTS(mrt, iter->ct))
2447 continue;
2448 if (pos-- == 0)
2449 return &mrt->vif_table[iter->ct];
2450 }
2451 return NULL;
2452 }
2453
2454 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2455 __acquires(mrt_lock)
2456 {
2457 struct ipmr_vif_iter *iter = seq->private;
2458 struct net *net = seq_file_net(seq);
2459 struct mr_table *mrt;
2460
2461 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2462 if (!mrt)
2463 return ERR_PTR(-ENOENT);
2464
2465 iter->mrt = mrt;
2466
2467 read_lock(&mrt_lock);
2468 return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
2469 : SEQ_START_TOKEN;
2470 }
2471
2472 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2473 {
2474 struct ipmr_vif_iter *iter = seq->private;
2475 struct net *net = seq_file_net(seq);
2476 struct mr_table *mrt = iter->mrt;
2477
2478 ++*pos;
2479 if (v == SEQ_START_TOKEN)
2480 return ipmr_vif_seq_idx(net, iter, 0);
2481
2482 while (++iter->ct < mrt->maxvif) {
2483 if (!VIF_EXISTS(mrt, iter->ct))
2484 continue;
2485 return &mrt->vif_table[iter->ct];
2486 }
2487 return NULL;
2488 }
2489
2490 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2491 __releases(mrt_lock)
2492 {
2493 read_unlock(&mrt_lock);
2494 }
2495
2496 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2497 {
2498 struct ipmr_vif_iter *iter = seq->private;
2499 struct mr_table *mrt = iter->mrt;
2500
2501 if (v == SEQ_START_TOKEN) {
2502 seq_puts(seq,
2503 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2504 } else {
2505 const struct vif_device *vif = v;
2506 const char *name = vif->dev ? vif->dev->name : "none";
2507
2508 seq_printf(seq,
2509 "%2Zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2510 vif - mrt->vif_table,
2511 name, vif->bytes_in, vif->pkt_in,
2512 vif->bytes_out, vif->pkt_out,
2513 vif->flags, vif->local, vif->remote);
2514 }
2515 return 0;
2516 }
2517
2518 static const struct seq_operations ipmr_vif_seq_ops = {
2519 .start = ipmr_vif_seq_start,
2520 .next = ipmr_vif_seq_next,
2521 .stop = ipmr_vif_seq_stop,
2522 .show = ipmr_vif_seq_show,
2523 };
2524
2525 static int ipmr_vif_open(struct inode *inode, struct file *file)
2526 {
2527 return seq_open_net(inode, file, &ipmr_vif_seq_ops,
2528 sizeof(struct ipmr_vif_iter));
2529 }
2530
2531 static const struct file_operations ipmr_vif_fops = {
2532 .owner = THIS_MODULE,
2533 .open = ipmr_vif_open,
2534 .read = seq_read,
2535 .llseek = seq_lseek,
2536 .release = seq_release_net,
2537 };
2538
2539 struct ipmr_mfc_iter {
2540 struct seq_net_private p;
2541 struct mr_table *mrt;
2542 struct list_head *cache;
2543 int ct;
2544 };
2545
2546
2547 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
2548 struct ipmr_mfc_iter *it, loff_t pos)
2549 {
2550 struct mr_table *mrt = it->mrt;
2551 struct mfc_cache *mfc;
2552
2553 rcu_read_lock();
2554 for (it->ct = 0; it->ct < MFC_LINES; it->ct++) {
2555 it->cache = &mrt->mfc_cache_array[it->ct];
2556 list_for_each_entry_rcu(mfc, it->cache, list)
2557 if (pos-- == 0)
2558 return mfc;
2559 }
2560 rcu_read_unlock();
2561
2562 spin_lock_bh(&mfc_unres_lock);
2563 it->cache = &mrt->mfc_unres_queue;
2564 list_for_each_entry(mfc, it->cache, list)
2565 if (pos-- == 0)
2566 return mfc;
2567 spin_unlock_bh(&mfc_unres_lock);
2568
2569 it->cache = NULL;
2570 return NULL;
2571 }
2572
2573
2574 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2575 {
2576 struct ipmr_mfc_iter *it = seq->private;
2577 struct net *net = seq_file_net(seq);
2578 struct mr_table *mrt;
2579
2580 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2581 if (!mrt)
2582 return ERR_PTR(-ENOENT);
2583
2584 it->mrt = mrt;
2585 it->cache = NULL;
2586 it->ct = 0;
2587 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
2588 : SEQ_START_TOKEN;
2589 }
2590
2591 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2592 {
2593 struct mfc_cache *mfc = v;
2594 struct ipmr_mfc_iter *it = seq->private;
2595 struct net *net = seq_file_net(seq);
2596 struct mr_table *mrt = it->mrt;
2597
2598 ++*pos;
2599
2600 if (v == SEQ_START_TOKEN)
2601 return ipmr_mfc_seq_idx(net, seq->private, 0);
2602
2603 if (mfc->list.next != it->cache)
2604 return list_entry(mfc->list.next, struct mfc_cache, list);
2605
2606 if (it->cache == &mrt->mfc_unres_queue)
2607 goto end_of_list;
2608
2609 BUG_ON(it->cache != &mrt->mfc_cache_array[it->ct]);
2610
2611 while (++it->ct < MFC_LINES) {
2612 it->cache = &mrt->mfc_cache_array[it->ct];
2613 if (list_empty(it->cache))
2614 continue;
2615 return list_first_entry(it->cache, struct mfc_cache, list);
2616 }
2617
2618 /* exhausted cache_array, show unresolved */
2619 rcu_read_unlock();
2620 it->cache = &mrt->mfc_unres_queue;
2621 it->ct = 0;
2622
2623 spin_lock_bh(&mfc_unres_lock);
2624 if (!list_empty(it->cache))
2625 return list_first_entry(it->cache, struct mfc_cache, list);
2626
2627 end_of_list:
2628 spin_unlock_bh(&mfc_unres_lock);
2629 it->cache = NULL;
2630
2631 return NULL;
2632 }
2633
2634 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
2635 {
2636 struct ipmr_mfc_iter *it = seq->private;
2637 struct mr_table *mrt = it->mrt;
2638
2639 if (it->cache == &mrt->mfc_unres_queue)
2640 spin_unlock_bh(&mfc_unres_lock);
2641 else if (it->cache == &mrt->mfc_cache_array[it->ct])
2642 rcu_read_unlock();
2643 }
2644
2645 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2646 {
2647 int n;
2648
2649 if (v == SEQ_START_TOKEN) {
2650 seq_puts(seq,
2651 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2652 } else {
2653 const struct mfc_cache *mfc = v;
2654 const struct ipmr_mfc_iter *it = seq->private;
2655 const struct mr_table *mrt = it->mrt;
2656
2657 seq_printf(seq, "%08X %08X %-3hd",
2658 (__force u32) mfc->mfc_mcastgrp,
2659 (__force u32) mfc->mfc_origin,
2660 mfc->mfc_parent);
2661
2662 if (it->cache != &mrt->mfc_unres_queue) {
2663 seq_printf(seq, " %8lu %8lu %8lu",
2664 mfc->mfc_un.res.pkt,
2665 mfc->mfc_un.res.bytes,
2666 mfc->mfc_un.res.wrong_if);
2667 for (n = mfc->mfc_un.res.minvif;
2668 n < mfc->mfc_un.res.maxvif; n++) {
2669 if (VIF_EXISTS(mrt, n) &&
2670 mfc->mfc_un.res.ttls[n] < 255)
2671 seq_printf(seq,
2672 " %2d:%-3d",
2673 n, mfc->mfc_un.res.ttls[n]);
2674 }
2675 } else {
2676 /* unresolved mfc_caches don't contain
2677 * pkt, bytes and wrong_if values
2678 */
2679 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2680 }
2681 seq_putc(seq, '\n');
2682 }
2683 return 0;
2684 }
2685
2686 static const struct seq_operations ipmr_mfc_seq_ops = {
2687 .start = ipmr_mfc_seq_start,
2688 .next = ipmr_mfc_seq_next,
2689 .stop = ipmr_mfc_seq_stop,
2690 .show = ipmr_mfc_seq_show,
2691 };
2692
2693 static int ipmr_mfc_open(struct inode *inode, struct file *file)
2694 {
2695 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
2696 sizeof(struct ipmr_mfc_iter));
2697 }
2698
2699 static const struct file_operations ipmr_mfc_fops = {
2700 .owner = THIS_MODULE,
2701 .open = ipmr_mfc_open,
2702 .read = seq_read,
2703 .llseek = seq_lseek,
2704 .release = seq_release_net,
2705 };
2706 #endif
2707
2708 #ifdef CONFIG_IP_PIMSM_V2
2709 static const struct net_protocol pim_protocol = {
2710 .handler = pim_rcv,
2711 .netns_ok = 1,
2712 };
2713 #endif
2714
2715
2716 /*
2717 * Setup for IP multicast routing
2718 */
2719 static int __net_init ipmr_net_init(struct net *net)
2720 {
2721 int err;
2722
2723 err = ipmr_rules_init(net);
2724 if (err < 0)
2725 goto fail;
2726
2727 #ifdef CONFIG_PROC_FS
2728 err = -ENOMEM;
2729 if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops))
2730 goto proc_vif_fail;
2731 if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops))
2732 goto proc_cache_fail;
2733 #endif
2734 return 0;
2735
2736 #ifdef CONFIG_PROC_FS
2737 proc_cache_fail:
2738 remove_proc_entry("ip_mr_vif", net->proc_net);
2739 proc_vif_fail:
2740 ipmr_rules_exit(net);
2741 #endif
2742 fail:
2743 return err;
2744 }
2745
2746 static void __net_exit ipmr_net_exit(struct net *net)
2747 {
2748 #ifdef CONFIG_PROC_FS
2749 remove_proc_entry("ip_mr_cache", net->proc_net);
2750 remove_proc_entry("ip_mr_vif", net->proc_net);
2751 #endif
2752 ipmr_rules_exit(net);
2753 }
2754
2755 static struct pernet_operations ipmr_net_ops = {
2756 .init = ipmr_net_init,
2757 .exit = ipmr_net_exit,
2758 };
2759
2760 int __init ip_mr_init(void)
2761 {
2762 int err;
2763
2764 mrt_cachep = kmem_cache_create("ip_mrt_cache",
2765 sizeof(struct mfc_cache),
2766 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
2767 NULL);
2768 if (!mrt_cachep)
2769 return -ENOMEM;
2770
2771 err = register_pernet_subsys(&ipmr_net_ops);
2772 if (err)
2773 goto reg_pernet_fail;
2774
2775 err = register_netdevice_notifier(&ip_mr_notifier);
2776 if (err)
2777 goto reg_notif_fail;
2778 #ifdef CONFIG_IP_PIMSM_V2
2779 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
2780 pr_err("%s: can't add PIM protocol\n", __func__);
2781 err = -EAGAIN;
2782 goto add_proto_fail;
2783 }
2784 #endif
2785 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
2786 NULL, ipmr_rtm_dumproute, NULL);
2787 return 0;
2788
2789 #ifdef CONFIG_IP_PIMSM_V2
2790 add_proto_fail:
2791 unregister_netdevice_notifier(&ip_mr_notifier);
2792 #endif
2793 reg_notif_fail:
2794 unregister_pernet_subsys(&ipmr_net_ops);
2795 reg_pernet_fail:
2796 kmem_cache_destroy(mrt_cachep);
2797 return err;
2798 }
ubuntu-bionic-kernel mirror