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 #ifdef CONFIG_NET_NS
77 struct net *net;
78 #endif
79 u32 id;
80 struct sock __rcu *mroute_sk;
81 struct timer_list ipmr_expire_timer;
82 struct list_head mfc_unres_queue;
83 struct list_head mfc_cache_array[MFC_LINES];
84 struct vif_device vif_table[MAXVIFS];
85 int maxvif;
86 atomic_t cache_resolve_queue_len;
87 bool mroute_do_assert;
88 bool mroute_do_pim;
89 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
90 int mroute_reg_vif_num;
91 #endif
92 };
93
94 struct ipmr_rule {
95 struct fib_rule common;
96 };
97
98 struct ipmr_result {
99 struct mr_table *mrt;
100 };
101
102 /* Big lock, protecting vif table, mrt cache and mroute socket state.
103 * Note that the changes are semaphored via rtnl_lock.
104 */
105
106 static DEFINE_RWLOCK(mrt_lock);
107
108 /*
109 * Multicast router control variables
110 */
111
112 #define VIF_EXISTS(_mrt, _idx) ((_mrt)->vif_table[_idx].dev != NULL)
113
114 /* Special spinlock for queue of unresolved entries */
115 static DEFINE_SPINLOCK(mfc_unres_lock);
116
117 /* We return to original Alan's scheme. Hash table of resolved
118 * entries is changed only in process context and protected
119 * with weak lock mrt_lock. Queue of unresolved entries is protected
120 * with strong spinlock mfc_unres_lock.
121 *
122 * In this case data path is free of exclusive locks at all.
123 */
124
125 static struct kmem_cache *mrt_cachep __read_mostly;
126
127 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
128 static void ipmr_free_table(struct mr_table *mrt);
129
130 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
131 struct sk_buff *skb, struct mfc_cache *cache,
132 int local);
133 static int ipmr_cache_report(struct mr_table *mrt,
134 struct sk_buff *pkt, vifi_t vifi, int assert);
135 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
136 struct mfc_cache *c, struct rtmsg *rtm);
137 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
138 int cmd);
139 static void mroute_clean_tables(struct mr_table *mrt, bool all);
140 static void ipmr_expire_process(unsigned long arg);
141
142 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
143 #define ipmr_for_each_table(mrt, net) \
144 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
145
ipmr_get_table(struct net * net,u32 id)146 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
147 {
148 struct mr_table *mrt;
149
150 ipmr_for_each_table(mrt, net) {
151 if (mrt->id == id)
152 return mrt;
153 }
154 return NULL;
155 }
156
ipmr_fib_lookup(struct net * net,struct flowi4 * flp4,struct mr_table ** mrt)157 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
158 struct mr_table **mrt)
159 {
160 int err;
161 struct ipmr_result res;
162 struct fib_lookup_arg arg = {
163 .result = &res,
164 .flags = FIB_LOOKUP_NOREF,
165 };
166
167 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
168 flowi4_to_flowi(flp4), 0, &arg);
169 if (err < 0)
170 return err;
171 *mrt = res.mrt;
172 return 0;
173 }
174
ipmr_rule_action(struct fib_rule * rule,struct flowi * flp,int flags,struct fib_lookup_arg * arg)175 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
176 int flags, struct fib_lookup_arg *arg)
177 {
178 struct ipmr_result *res = arg->result;
179 struct mr_table *mrt;
180
181 switch (rule->action) {
182 case FR_ACT_TO_TBL:
183 break;
184 case FR_ACT_UNREACHABLE:
185 return -ENETUNREACH;
186 case FR_ACT_PROHIBIT:
187 return -EACCES;
188 case FR_ACT_BLACKHOLE:
189 default:
190 return -EINVAL;
191 }
192
193 mrt = ipmr_get_table(rule->fr_net, rule->table);
194 if (mrt == NULL)
195 return -EAGAIN;
196 res->mrt = mrt;
197 return 0;
198 }
199
ipmr_rule_match(struct fib_rule * rule,struct flowi * fl,int flags)200 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
201 {
202 return 1;
203 }
204
205 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
206 FRA_GENERIC_POLICY,
207 };
208
ipmr_rule_configure(struct fib_rule * rule,struct sk_buff * skb,struct fib_rule_hdr * frh,struct nlattr ** tb)209 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
210 struct fib_rule_hdr *frh, struct nlattr **tb)
211 {
212 return 0;
213 }
214
ipmr_rule_compare(struct fib_rule * rule,struct fib_rule_hdr * frh,struct nlattr ** tb)215 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
216 struct nlattr **tb)
217 {
218 return 1;
219 }
220
ipmr_rule_fill(struct fib_rule * rule,struct sk_buff * skb,struct fib_rule_hdr * frh)221 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
222 struct fib_rule_hdr *frh)
223 {
224 frh->dst_len = 0;
225 frh->src_len = 0;
226 frh->tos = 0;
227 return 0;
228 }
229
230 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
231 .family = RTNL_FAMILY_IPMR,
232 .rule_size = sizeof(struct ipmr_rule),
233 .addr_size = sizeof(u32),
234 .action = ipmr_rule_action,
235 .match = ipmr_rule_match,
236 .configure = ipmr_rule_configure,
237 .compare = ipmr_rule_compare,
238 .default_pref = fib_default_rule_pref,
239 .fill = ipmr_rule_fill,
240 .nlgroup = RTNLGRP_IPV4_RULE,
241 .policy = ipmr_rule_policy,
242 .owner = THIS_MODULE,
243 };
244
ipmr_rules_init(struct net * net)245 static int __net_init ipmr_rules_init(struct net *net)
246 {
247 struct fib_rules_ops *ops;
248 struct mr_table *mrt;
249 int err;
250
251 ops = fib_rules_register(&ipmr_rules_ops_template, net);
252 if (IS_ERR(ops))
253 return PTR_ERR(ops);
254
255 INIT_LIST_HEAD(&net->ipv4.mr_tables);
256
257 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
258 if (mrt == NULL) {
259 err = -ENOMEM;
260 goto err1;
261 }
262
263 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
264 if (err < 0)
265 goto err2;
266
267 net->ipv4.mr_rules_ops = ops;
268 return 0;
269
270 err2:
271 kfree(mrt);
272 err1:
273 fib_rules_unregister(ops);
274 return err;
275 }
276
ipmr_rules_exit(struct net * net)277 static void __net_exit ipmr_rules_exit(struct net *net)
278 {
279 struct mr_table *mrt, *next;
280
281 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
282 list_del(&mrt->list);
283 ipmr_free_table(mrt);
284 }
285 fib_rules_unregister(net->ipv4.mr_rules_ops);
286 }
287 #else
288 #define ipmr_for_each_table(mrt, net) \
289 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
290
ipmr_get_table(struct net * net,u32 id)291 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
292 {
293 return net->ipv4.mrt;
294 }
295
ipmr_fib_lookup(struct net * net,struct flowi4 * flp4,struct mr_table ** mrt)296 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
297 struct mr_table **mrt)
298 {
299 *mrt = net->ipv4.mrt;
300 return 0;
301 }
302
ipmr_rules_init(struct net * net)303 static int __net_init ipmr_rules_init(struct net *net)
304 {
305 net->ipv4.mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
306 return net->ipv4.mrt ? 0 : -ENOMEM;
307 }
308
ipmr_rules_exit(struct net * net)309 static void __net_exit ipmr_rules_exit(struct net *net)
310 {
311 ipmr_free_table(net->ipv4.mrt);
312 }
313 #endif
314
ipmr_new_table(struct net * net,u32 id)315 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
316 {
317 struct mr_table *mrt;
318 unsigned int i;
319
320 mrt = ipmr_get_table(net, id);
321 if (mrt != NULL)
322 return mrt;
323
324 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
325 if (mrt == NULL)
326 return NULL;
327 write_pnet(&mrt->net, net);
328 mrt->id = id;
329
330 /* Forwarding cache */
331 for (i = 0; i < MFC_LINES; i++)
332 INIT_LIST_HEAD(&mrt->mfc_cache_array[i]);
333
334 INIT_LIST_HEAD(&mrt->mfc_unres_queue);
335
336 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
337 (unsigned long)mrt);
338
339 #ifdef CONFIG_IP_PIMSM
340 mrt->mroute_reg_vif_num = -1;
341 #endif
342 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
343 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
344 #endif
345 return mrt;
346 }
347
ipmr_free_table(struct mr_table * mrt)348 static void ipmr_free_table(struct mr_table *mrt)
349 {
350 del_timer_sync(&mrt->ipmr_expire_timer);
351 mroute_clean_tables(mrt, true);
352 kfree(mrt);
353 }
354
355 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
356
ipmr_del_tunnel(struct net_device * dev,struct vifctl * v)357 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
358 {
359 struct net *net = dev_net(dev);
360
361 dev_close(dev);
362
363 dev = __dev_get_by_name(net, "tunl0");
364 if (dev) {
365 const struct net_device_ops *ops = dev->netdev_ops;
366 struct ifreq ifr;
367 struct ip_tunnel_parm p;
368
369 memset(&p, 0, sizeof(p));
370 p.iph.daddr = v->vifc_rmt_addr.s_addr;
371 p.iph.saddr = v->vifc_lcl_addr.s_addr;
372 p.iph.version = 4;
373 p.iph.ihl = 5;
374 p.iph.protocol = IPPROTO_IPIP;
375 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
376 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
377
378 if (ops->ndo_do_ioctl) {
379 mm_segment_t oldfs = get_fs();
380
381 set_fs(KERNEL_DS);
382 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
383 set_fs(oldfs);
384 }
385 }
386 }
387
388 static
ipmr_new_tunnel(struct net * net,struct vifctl * v)389 struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
390 {
391 struct net_device *dev;
392
393 dev = __dev_get_by_name(net, "tunl0");
394
395 if (dev) {
396 const struct net_device_ops *ops = dev->netdev_ops;
397 int err;
398 struct ifreq ifr;
399 struct ip_tunnel_parm p;
400 struct in_device *in_dev;
401
402 memset(&p, 0, sizeof(p));
403 p.iph.daddr = v->vifc_rmt_addr.s_addr;
404 p.iph.saddr = v->vifc_lcl_addr.s_addr;
405 p.iph.version = 4;
406 p.iph.ihl = 5;
407 p.iph.protocol = IPPROTO_IPIP;
408 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
409 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
410
411 if (ops->ndo_do_ioctl) {
412 mm_segment_t oldfs = get_fs();
413
414 set_fs(KERNEL_DS);
415 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
416 set_fs(oldfs);
417 } else {
418 err = -EOPNOTSUPP;
419 }
420 dev = NULL;
421
422 if (err == 0 &&
423 (dev = __dev_get_by_name(net, p.name)) != NULL) {
424 dev->flags |= IFF_MULTICAST;
425
426 in_dev = __in_dev_get_rtnl(dev);
427 if (in_dev == NULL)
428 goto failure;
429
430 ipv4_devconf_setall(in_dev);
431 neigh_parms_data_state_setall(in_dev->arp_parms);
432 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
433
434 if (dev_open(dev))
435 goto failure;
436 dev_hold(dev);
437 }
438 }
439 return dev;
440
441 failure:
442 /* allow the register to be completed before unregistering. */
443 rtnl_unlock();
444 rtnl_lock();
445
446 unregister_netdevice(dev);
447 return NULL;
448 }
449
450 #ifdef CONFIG_IP_PIMSM
451
reg_vif_xmit(struct sk_buff * skb,struct net_device * dev)452 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
453 {
454 struct net *net = dev_net(dev);
455 struct mr_table *mrt;
456 struct flowi4 fl4 = {
457 .flowi4_oif = dev->ifindex,
458 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
459 .flowi4_mark = skb->mark,
460 };
461 int err;
462
463 err = ipmr_fib_lookup(net, &fl4, &mrt);
464 if (err < 0) {
465 kfree_skb(skb);
466 return err;
467 }
468
469 read_lock(&mrt_lock);
470 dev->stats.tx_bytes += skb->len;
471 dev->stats.tx_packets++;
472 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
473 read_unlock(&mrt_lock);
474 kfree_skb(skb);
475 return NETDEV_TX_OK;
476 }
477
478 static const struct net_device_ops reg_vif_netdev_ops = {
479 .ndo_start_xmit = reg_vif_xmit,
480 };
481
reg_vif_setup(struct net_device * dev)482 static void reg_vif_setup(struct net_device *dev)
483 {
484 dev->type = ARPHRD_PIMREG;
485 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
486 dev->flags = IFF_NOARP;
487 dev->netdev_ops = ®_vif_netdev_ops;
488 dev->destructor = free_netdev;
489 dev->features |= NETIF_F_NETNS_LOCAL;
490 }
491
ipmr_reg_vif(struct net * net,struct mr_table * mrt)492 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
493 {
494 struct net_device *dev;
495 struct in_device *in_dev;
496 char name[IFNAMSIZ];
497
498 if (mrt->id == RT_TABLE_DEFAULT)
499 sprintf(name, "pimreg");
500 else
501 sprintf(name, "pimreg%u", mrt->id);
502
503 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
504
505 if (dev == NULL)
506 return NULL;
507
508 dev_net_set(dev, net);
509
510 if (register_netdevice(dev)) {
511 free_netdev(dev);
512 return NULL;
513 }
514 dev->iflink = 0;
515
516 rcu_read_lock();
517 in_dev = __in_dev_get_rcu(dev);
518 if (!in_dev) {
519 rcu_read_unlock();
520 goto failure;
521 }
522
523 ipv4_devconf_setall(in_dev);
524 neigh_parms_data_state_setall(in_dev->arp_parms);
525 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
526 rcu_read_unlock();
527
528 if (dev_open(dev))
529 goto failure;
530
531 dev_hold(dev);
532
533 return dev;
534
535 failure:
536 /* allow the register to be completed before unregistering. */
537 rtnl_unlock();
538 rtnl_lock();
539
540 unregister_netdevice(dev);
541 return NULL;
542 }
543 #endif
544
545 /**
546 * vif_delete - Delete a VIF entry
547 * @notify: Set to 1, if the caller is a notifier_call
548 */
549
vif_delete(struct mr_table * mrt,int vifi,int notify,struct list_head * head)550 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
551 struct list_head *head)
552 {
553 struct vif_device *v;
554 struct net_device *dev;
555 struct in_device *in_dev;
556
557 if (vifi < 0 || vifi >= mrt->maxvif)
558 return -EADDRNOTAVAIL;
559
560 v = &mrt->vif_table[vifi];
561
562 write_lock_bh(&mrt_lock);
563 dev = v->dev;
564 v->dev = NULL;
565
566 if (!dev) {
567 write_unlock_bh(&mrt_lock);
568 return -EADDRNOTAVAIL;
569 }
570
571 #ifdef CONFIG_IP_PIMSM
572 if (vifi == mrt->mroute_reg_vif_num)
573 mrt->mroute_reg_vif_num = -1;
574 #endif
575
576 if (vifi + 1 == mrt->maxvif) {
577 int tmp;
578
579 for (tmp = vifi - 1; tmp >= 0; tmp--) {
580 if (VIF_EXISTS(mrt, tmp))
581 break;
582 }
583 mrt->maxvif = tmp+1;
584 }
585
586 write_unlock_bh(&mrt_lock);
587
588 dev_set_allmulti(dev, -1);
589
590 in_dev = __in_dev_get_rtnl(dev);
591 if (in_dev) {
592 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
593 inet_netconf_notify_devconf(dev_net(dev),
594 NETCONFA_MC_FORWARDING,
595 dev->ifindex, &in_dev->cnf);
596 ip_rt_multicast_event(in_dev);
597 }
598
599 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
600 unregister_netdevice_queue(dev, head);
601
602 dev_put(dev);
603 return 0;
604 }
605
ipmr_cache_free_rcu(struct rcu_head * head)606 static void ipmr_cache_free_rcu(struct rcu_head *head)
607 {
608 struct mfc_cache *c = container_of(head, struct mfc_cache, rcu);
609
610 kmem_cache_free(mrt_cachep, c);
611 }
612
ipmr_cache_free(struct mfc_cache * c)613 static inline void ipmr_cache_free(struct mfc_cache *c)
614 {
615 call_rcu(&c->rcu, ipmr_cache_free_rcu);
616 }
617
618 /* Destroy an unresolved cache entry, killing queued skbs
619 * and reporting error to netlink readers.
620 */
621
ipmr_destroy_unres(struct mr_table * mrt,struct mfc_cache * c)622 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
623 {
624 struct net *net = read_pnet(&mrt->net);
625 struct sk_buff *skb;
626 struct nlmsgerr *e;
627
628 atomic_dec(&mrt->cache_resolve_queue_len);
629
630 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) {
631 if (ip_hdr(skb)->version == 0) {
632 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
633 nlh->nlmsg_type = NLMSG_ERROR;
634 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
635 skb_trim(skb, nlh->nlmsg_len);
636 e = nlmsg_data(nlh);
637 e->error = -ETIMEDOUT;
638 memset(&e->msg, 0, sizeof(e->msg));
639
640 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
641 } else {
642 kfree_skb(skb);
643 }
644 }
645
646 ipmr_cache_free(c);
647 }
648
649
650 /* Timer process for the unresolved queue. */
651
ipmr_expire_process(unsigned long arg)652 static void ipmr_expire_process(unsigned long arg)
653 {
654 struct mr_table *mrt = (struct mr_table *)arg;
655 unsigned long now;
656 unsigned long expires;
657 struct mfc_cache *c, *next;
658
659 if (!spin_trylock(&mfc_unres_lock)) {
660 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
661 return;
662 }
663
664 if (list_empty(&mrt->mfc_unres_queue))
665 goto out;
666
667 now = jiffies;
668 expires = 10*HZ;
669
670 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
671 if (time_after(c->mfc_un.unres.expires, now)) {
672 unsigned long interval = c->mfc_un.unres.expires - now;
673 if (interval < expires)
674 expires = interval;
675 continue;
676 }
677
678 list_del(&c->list);
679 mroute_netlink_event(mrt, c, RTM_DELROUTE);
680 ipmr_destroy_unres(mrt, c);
681 }
682
683 if (!list_empty(&mrt->mfc_unres_queue))
684 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
685
686 out:
687 spin_unlock(&mfc_unres_lock);
688 }
689
690 /* Fill oifs list. It is called under write locked mrt_lock. */
691
ipmr_update_thresholds(struct mr_table * mrt,struct mfc_cache * cache,unsigned char * ttls)692 static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache,
693 unsigned char *ttls)
694 {
695 int vifi;
696
697 cache->mfc_un.res.minvif = MAXVIFS;
698 cache->mfc_un.res.maxvif = 0;
699 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
700
701 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
702 if (VIF_EXISTS(mrt, vifi) &&
703 ttls[vifi] && ttls[vifi] < 255) {
704 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
705 if (cache->mfc_un.res.minvif > vifi)
706 cache->mfc_un.res.minvif = vifi;
707 if (cache->mfc_un.res.maxvif <= vifi)
708 cache->mfc_un.res.maxvif = vifi + 1;
709 }
710 }
711 }
712
vif_add(struct net * net,struct mr_table * mrt,struct vifctl * vifc,int mrtsock)713 static int vif_add(struct net *net, struct mr_table *mrt,
714 struct vifctl *vifc, int mrtsock)
715 {
716 int vifi = vifc->vifc_vifi;
717 struct vif_device *v = &mrt->vif_table[vifi];
718 struct net_device *dev;
719 struct in_device *in_dev;
720 int err;
721
722 /* Is vif busy ? */
723 if (VIF_EXISTS(mrt, vifi))
724 return -EADDRINUSE;
725
726 switch (vifc->vifc_flags) {
727 #ifdef CONFIG_IP_PIMSM
728 case VIFF_REGISTER:
729 /*
730 * Special Purpose VIF in PIM
731 * All the packets will be sent to the daemon
732 */
733 if (mrt->mroute_reg_vif_num >= 0)
734 return -EADDRINUSE;
735 dev = ipmr_reg_vif(net, mrt);
736 if (!dev)
737 return -ENOBUFS;
738 err = dev_set_allmulti(dev, 1);
739 if (err) {
740 unregister_netdevice(dev);
741 dev_put(dev);
742 return err;
743 }
744 break;
745 #endif
746 case VIFF_TUNNEL:
747 dev = ipmr_new_tunnel(net, vifc);
748 if (!dev)
749 return -ENOBUFS;
750 err = dev_set_allmulti(dev, 1);
751 if (err) {
752 ipmr_del_tunnel(dev, vifc);
753 dev_put(dev);
754 return err;
755 }
756 break;
757
758 case VIFF_USE_IFINDEX:
759 case 0:
760 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
761 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
762 if (dev && __in_dev_get_rtnl(dev) == NULL) {
763 dev_put(dev);
764 return -EADDRNOTAVAIL;
765 }
766 } else {
767 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
768 }
769 if (!dev)
770 return -EADDRNOTAVAIL;
771 err = dev_set_allmulti(dev, 1);
772 if (err) {
773 dev_put(dev);
774 return err;
775 }
776 break;
777 default:
778 return -EINVAL;
779 }
780
781 in_dev = __in_dev_get_rtnl(dev);
782 if (!in_dev) {
783 dev_put(dev);
784 return -EADDRNOTAVAIL;
785 }
786 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
787 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, dev->ifindex,
788 &in_dev->cnf);
789 ip_rt_multicast_event(in_dev);
790
791 /* Fill in the VIF structures */
792
793 v->rate_limit = vifc->vifc_rate_limit;
794 v->local = vifc->vifc_lcl_addr.s_addr;
795 v->remote = vifc->vifc_rmt_addr.s_addr;
796 v->flags = vifc->vifc_flags;
797 if (!mrtsock)
798 v->flags |= VIFF_STATIC;
799 v->threshold = vifc->vifc_threshold;
800 v->bytes_in = 0;
801 v->bytes_out = 0;
802 v->pkt_in = 0;
803 v->pkt_out = 0;
804 v->link = dev->ifindex;
805 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER))
806 v->link = dev->iflink;
807
808 /* And finish update writing critical data */
809 write_lock_bh(&mrt_lock);
810 v->dev = dev;
811 #ifdef CONFIG_IP_PIMSM
812 if (v->flags & VIFF_REGISTER)
813 mrt->mroute_reg_vif_num = vifi;
814 #endif
815 if (vifi+1 > mrt->maxvif)
816 mrt->maxvif = vifi+1;
817 write_unlock_bh(&mrt_lock);
818 return 0;
819 }
820
821 /* called with rcu_read_lock() */
ipmr_cache_find(struct mr_table * mrt,__be32 origin,__be32 mcastgrp)822 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
823 __be32 origin,
824 __be32 mcastgrp)
825 {
826 int line = MFC_HASH(mcastgrp, origin);
827 struct mfc_cache *c;
828
829 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) {
830 if (c->mfc_origin == origin && c->mfc_mcastgrp == mcastgrp)
831 return c;
832 }
833 return NULL;
834 }
835
836 /* Look for a (*,*,oif) entry */
ipmr_cache_find_any_parent(struct mr_table * mrt,int vifi)837 static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt,
838 int vifi)
839 {
840 int line = MFC_HASH(htonl(INADDR_ANY), htonl(INADDR_ANY));
841 struct mfc_cache *c;
842
843 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
844 if (c->mfc_origin == htonl(INADDR_ANY) &&
845 c->mfc_mcastgrp == htonl(INADDR_ANY) &&
846 c->mfc_un.res.ttls[vifi] < 255)
847 return c;
848
849 return NULL;
850 }
851
852 /* Look for a (*,G) entry */
ipmr_cache_find_any(struct mr_table * mrt,__be32 mcastgrp,int vifi)853 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
854 __be32 mcastgrp, int vifi)
855 {
856 int line = MFC_HASH(mcastgrp, htonl(INADDR_ANY));
857 struct mfc_cache *c, *proxy;
858
859 if (mcastgrp == htonl(INADDR_ANY))
860 goto skip;
861
862 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
863 if (c->mfc_origin == htonl(INADDR_ANY) &&
864 c->mfc_mcastgrp == mcastgrp) {
865 if (c->mfc_un.res.ttls[vifi] < 255)
866 return c;
867
868 /* It's ok if the vifi is part of the static tree */
869 proxy = ipmr_cache_find_any_parent(mrt,
870 c->mfc_parent);
871 if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
872 return c;
873 }
874
875 skip:
876 return ipmr_cache_find_any_parent(mrt, vifi);
877 }
878
879 /*
880 * Allocate a multicast cache entry
881 */
ipmr_cache_alloc(void)882 static struct mfc_cache *ipmr_cache_alloc(void)
883 {
884 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
885
886 if (c)
887 c->mfc_un.res.minvif = MAXVIFS;
888 return c;
889 }
890
ipmr_cache_alloc_unres(void)891 static struct mfc_cache *ipmr_cache_alloc_unres(void)
892 {
893 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
894
895 if (c) {
896 skb_queue_head_init(&c->mfc_un.unres.unresolved);
897 c->mfc_un.unres.expires = jiffies + 10*HZ;
898 }
899 return c;
900 }
901
902 /*
903 * A cache entry has gone into a resolved state from queued
904 */
905
ipmr_cache_resolve(struct net * net,struct mr_table * mrt,struct mfc_cache * uc,struct mfc_cache * c)906 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
907 struct mfc_cache *uc, struct mfc_cache *c)
908 {
909 struct sk_buff *skb;
910 struct nlmsgerr *e;
911
912 /* Play the pending entries through our router */
913
914 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
915 if (ip_hdr(skb)->version == 0) {
916 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
917
918 if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
919 nlh->nlmsg_len = skb_tail_pointer(skb) -
920 (u8 *)nlh;
921 } else {
922 nlh->nlmsg_type = NLMSG_ERROR;
923 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
924 skb_trim(skb, nlh->nlmsg_len);
925 e = nlmsg_data(nlh);
926 e->error = -EMSGSIZE;
927 memset(&e->msg, 0, sizeof(e->msg));
928 }
929
930 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
931 } else {
932 ip_mr_forward(net, mrt, skb, c, 0);
933 }
934 }
935 }
936
937 /*
938 * Bounce a cache query up to mrouted. We could use netlink for this but mrouted
939 * expects the following bizarre scheme.
940 *
941 * Called under mrt_lock.
942 */
943
ipmr_cache_report(struct mr_table * mrt,struct sk_buff * pkt,vifi_t vifi,int assert)944 static int ipmr_cache_report(struct mr_table *mrt,
945 struct sk_buff *pkt, vifi_t vifi, int assert)
946 {
947 struct sk_buff *skb;
948 const int ihl = ip_hdrlen(pkt);
949 struct igmphdr *igmp;
950 struct igmpmsg *msg;
951 struct sock *mroute_sk;
952 int ret;
953
954 #ifdef CONFIG_IP_PIMSM
955 if (assert == IGMPMSG_WHOLEPKT)
956 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
957 else
958 #endif
959 skb = alloc_skb(128, GFP_ATOMIC);
960
961 if (!skb)
962 return -ENOBUFS;
963
964 #ifdef CONFIG_IP_PIMSM
965 if (assert == IGMPMSG_WHOLEPKT) {
966 /* Ugly, but we have no choice with this interface.
967 * Duplicate old header, fix ihl, length etc.
968 * And all this only to mangle msg->im_msgtype and
969 * to set msg->im_mbz to "mbz" :-)
970 */
971 skb_push(skb, sizeof(struct iphdr));
972 skb_reset_network_header(skb);
973 skb_reset_transport_header(skb);
974 msg = (struct igmpmsg *)skb_network_header(skb);
975 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
976 msg->im_msgtype = IGMPMSG_WHOLEPKT;
977 msg->im_mbz = 0;
978 msg->im_vif = mrt->mroute_reg_vif_num;
979 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
980 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
981 sizeof(struct iphdr));
982 } else
983 #endif
984 {
985
986 /* Copy the IP header */
987
988 skb_set_network_header(skb, skb->len);
989 skb_put(skb, ihl);
990 skb_copy_to_linear_data(skb, pkt->data, ihl);
991 ip_hdr(skb)->protocol = 0; /* Flag to the kernel this is a route add */
992 msg = (struct igmpmsg *)skb_network_header(skb);
993 msg->im_vif = vifi;
994 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
995
996 /* Add our header */
997
998 igmp = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
999 igmp->type =
1000 msg->im_msgtype = assert;
1001 igmp->code = 0;
1002 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1003 skb->transport_header = skb->network_header;
1004 }
1005
1006 rcu_read_lock();
1007 mroute_sk = rcu_dereference(mrt->mroute_sk);
1008 if (mroute_sk == NULL) {
1009 rcu_read_unlock();
1010 kfree_skb(skb);
1011 return -EINVAL;
1012 }
1013
1014 /* Deliver to mrouted */
1015
1016 ret = sock_queue_rcv_skb(mroute_sk, skb);
1017 rcu_read_unlock();
1018 if (ret < 0) {
1019 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1020 kfree_skb(skb);
1021 }
1022
1023 return ret;
1024 }
1025
1026 /*
1027 * Queue a packet for resolution. It gets locked cache entry!
1028 */
1029
1030 static int
ipmr_cache_unresolved(struct mr_table * mrt,vifi_t vifi,struct sk_buff * skb)1031 ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, struct sk_buff *skb)
1032 {
1033 bool found = false;
1034 int err;
1035 struct mfc_cache *c;
1036 const struct iphdr *iph = ip_hdr(skb);
1037
1038 spin_lock_bh(&mfc_unres_lock);
1039 list_for_each_entry(c, &mrt->mfc_unres_queue, list) {
1040 if (c->mfc_mcastgrp == iph->daddr &&
1041 c->mfc_origin == iph->saddr) {
1042 found = true;
1043 break;
1044 }
1045 }
1046
1047 if (!found) {
1048 /* Create a new entry if allowable */
1049
1050 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1051 (c = ipmr_cache_alloc_unres()) == NULL) {
1052 spin_unlock_bh(&mfc_unres_lock);
1053
1054 kfree_skb(skb);
1055 return -ENOBUFS;
1056 }
1057
1058 /* Fill in the new cache entry */
1059
1060 c->mfc_parent = -1;
1061 c->mfc_origin = iph->saddr;
1062 c->mfc_mcastgrp = iph->daddr;
1063
1064 /* Reflect first query at mrouted. */
1065
1066 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1067 if (err < 0) {
1068 /* If the report failed throw the cache entry
1069 out - Brad Parker
1070 */
1071 spin_unlock_bh(&mfc_unres_lock);
1072
1073 ipmr_cache_free(c);
1074 kfree_skb(skb);
1075 return err;
1076 }
1077
1078 atomic_inc(&mrt->cache_resolve_queue_len);
1079 list_add(&c->list, &mrt->mfc_unres_queue);
1080 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1081
1082 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1083 mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires);
1084 }
1085
1086 /* See if we can append the packet */
1087
1088 if (c->mfc_un.unres.unresolved.qlen > 3) {
1089 kfree_skb(skb);
1090 err = -ENOBUFS;
1091 } else {
1092 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1093 err = 0;
1094 }
1095
1096 spin_unlock_bh(&mfc_unres_lock);
1097 return err;
1098 }
1099
1100 /*
1101 * MFC cache manipulation by user space mroute daemon
1102 */
1103
ipmr_mfc_delete(struct mr_table * mrt,struct mfcctl * mfc,int parent)1104 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1105 {
1106 int line;
1107 struct mfc_cache *c, *next;
1108
1109 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1110
1111 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[line], list) {
1112 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1113 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1114 (parent == -1 || parent == c->mfc_parent)) {
1115 list_del_rcu(&c->list);
1116 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1117 ipmr_cache_free(c);
1118 return 0;
1119 }
1120 }
1121 return -ENOENT;
1122 }
1123
ipmr_mfc_add(struct net * net,struct mr_table * mrt,struct mfcctl * mfc,int mrtsock,int parent)1124 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1125 struct mfcctl *mfc, int mrtsock, int parent)
1126 {
1127 bool found = false;
1128 int line;
1129 struct mfc_cache *uc, *c;
1130
1131 if (mfc->mfcc_parent >= MAXVIFS)
1132 return -ENFILE;
1133
1134 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1135
1136 list_for_each_entry(c, &mrt->mfc_cache_array[line], list) {
1137 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1138 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1139 (parent == -1 || parent == c->mfc_parent)) {
1140 found = true;
1141 break;
1142 }
1143 }
1144
1145 if (found) {
1146 write_lock_bh(&mrt_lock);
1147 c->mfc_parent = mfc->mfcc_parent;
1148 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1149 if (!mrtsock)
1150 c->mfc_flags |= MFC_STATIC;
1151 write_unlock_bh(&mrt_lock);
1152 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1153 return 0;
1154 }
1155
1156 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1157 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1158 return -EINVAL;
1159
1160 c = ipmr_cache_alloc();
1161 if (c == NULL)
1162 return -ENOMEM;
1163
1164 c->mfc_origin = mfc->mfcc_origin.s_addr;
1165 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1166 c->mfc_parent = mfc->mfcc_parent;
1167 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1168 if (!mrtsock)
1169 c->mfc_flags |= MFC_STATIC;
1170
1171 list_add_rcu(&c->list, &mrt->mfc_cache_array[line]);
1172
1173 /*
1174 * Check to see if we resolved a queued list. If so we
1175 * need to send on the frames and tidy up.
1176 */
1177 found = false;
1178 spin_lock_bh(&mfc_unres_lock);
1179 list_for_each_entry(uc, &mrt->mfc_unres_queue, list) {
1180 if (uc->mfc_origin == c->mfc_origin &&
1181 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1182 list_del(&uc->list);
1183 atomic_dec(&mrt->cache_resolve_queue_len);
1184 found = true;
1185 break;
1186 }
1187 }
1188 if (list_empty(&mrt->mfc_unres_queue))
1189 del_timer(&mrt->ipmr_expire_timer);
1190 spin_unlock_bh(&mfc_unres_lock);
1191
1192 if (found) {
1193 ipmr_cache_resolve(net, mrt, uc, c);
1194 ipmr_cache_free(uc);
1195 }
1196 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1197 return 0;
1198 }
1199
1200 /*
1201 * Close the multicast socket, and clear the vif tables etc
1202 */
1203
mroute_clean_tables(struct mr_table * mrt,bool all)1204 static void mroute_clean_tables(struct mr_table *mrt, bool all)
1205 {
1206 int i;
1207 LIST_HEAD(list);
1208 struct mfc_cache *c, *next;
1209
1210 /* Shut down all active vif entries */
1211
1212 for (i = 0; i < mrt->maxvif; i++) {
1213 if (!all && (mrt->vif_table[i].flags & VIFF_STATIC))
1214 continue;
1215 vif_delete(mrt, i, 0, &list);
1216 }
1217 unregister_netdevice_many(&list);
1218
1219 /* Wipe the cache */
1220
1221 for (i = 0; i < MFC_LINES; i++) {
1222 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[i], list) {
1223 if (!all && (c->mfc_flags & MFC_STATIC))
1224 continue;
1225 list_del_rcu(&c->list);
1226 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1227 ipmr_cache_free(c);
1228 }
1229 }
1230
1231 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1232 spin_lock_bh(&mfc_unres_lock);
1233 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
1234 list_del(&c->list);
1235 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1236 ipmr_destroy_unres(mrt, c);
1237 }
1238 spin_unlock_bh(&mfc_unres_lock);
1239 }
1240 }
1241
1242 /* called from ip_ra_control(), before an RCU grace period,
1243 * we dont need to call synchronize_rcu() here
1244 */
mrtsock_destruct(struct sock * sk)1245 static void mrtsock_destruct(struct sock *sk)
1246 {
1247 struct net *net = sock_net(sk);
1248 struct mr_table *mrt;
1249
1250 rtnl_lock();
1251 ipmr_for_each_table(mrt, net) {
1252 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1253 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1254 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1255 NETCONFA_IFINDEX_ALL,
1256 net->ipv4.devconf_all);
1257 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1258 mroute_clean_tables(mrt, false);
1259 }
1260 }
1261 rtnl_unlock();
1262 }
1263
1264 /*
1265 * Socket options and virtual interface manipulation. The whole
1266 * virtual interface system is a complete heap, but unfortunately
1267 * that's how BSD mrouted happens to think. Maybe one day with a proper
1268 * MOSPF/PIM router set up we can clean this up.
1269 */
1270
ip_mroute_setsockopt(struct sock * sk,int optname,char __user * optval,unsigned int optlen)1271 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1272 {
1273 int ret, parent = 0;
1274 struct vifctl vif;
1275 struct mfcctl mfc;
1276 struct net *net = sock_net(sk);
1277 struct mr_table *mrt;
1278
1279 if (sk->sk_type != SOCK_RAW ||
1280 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1281 return -EOPNOTSUPP;
1282
1283 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1284 if (mrt == NULL)
1285 return -ENOENT;
1286
1287 if (optname != MRT_INIT) {
1288 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1289 !ns_capable(net->user_ns, CAP_NET_ADMIN))
1290 return -EACCES;
1291 }
1292
1293 switch (optname) {
1294 case MRT_INIT:
1295 if (optlen != sizeof(int))
1296 return -EINVAL;
1297
1298 rtnl_lock();
1299 if (rtnl_dereference(mrt->mroute_sk)) {
1300 rtnl_unlock();
1301 return -EADDRINUSE;
1302 }
1303
1304 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1305 if (ret == 0) {
1306 rcu_assign_pointer(mrt->mroute_sk, sk);
1307 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1308 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1309 NETCONFA_IFINDEX_ALL,
1310 net->ipv4.devconf_all);
1311 }
1312 rtnl_unlock();
1313 return ret;
1314 case MRT_DONE:
1315 if (sk != rcu_access_pointer(mrt->mroute_sk))
1316 return -EACCES;
1317 return ip_ra_control(sk, 0, NULL);
1318 case MRT_ADD_VIF:
1319 case MRT_DEL_VIF:
1320 if (optlen != sizeof(vif))
1321 return -EINVAL;
1322 if (copy_from_user(&vif, optval, sizeof(vif)))
1323 return -EFAULT;
1324 if (vif.vifc_vifi >= MAXVIFS)
1325 return -ENFILE;
1326 rtnl_lock();
1327 if (optname == MRT_ADD_VIF) {
1328 ret = vif_add(net, mrt, &vif,
1329 sk == rtnl_dereference(mrt->mroute_sk));
1330 } else {
1331 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1332 }
1333 rtnl_unlock();
1334 return ret;
1335
1336 /*
1337 * Manipulate the forwarding caches. These live
1338 * in a sort of kernel/user symbiosis.
1339 */
1340 case MRT_ADD_MFC:
1341 case MRT_DEL_MFC:
1342 parent = -1;
1343 case MRT_ADD_MFC_PROXY:
1344 case MRT_DEL_MFC_PROXY:
1345 if (optlen != sizeof(mfc))
1346 return -EINVAL;
1347 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1348 return -EFAULT;
1349 if (parent == 0)
1350 parent = mfc.mfcc_parent;
1351 rtnl_lock();
1352 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1353 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1354 else
1355 ret = ipmr_mfc_add(net, mrt, &mfc,
1356 sk == rtnl_dereference(mrt->mroute_sk),
1357 parent);
1358 rtnl_unlock();
1359 return ret;
1360 /*
1361 * Control PIM assert.
1362 */
1363 case MRT_ASSERT:
1364 {
1365 int v;
1366 if (optlen != sizeof(v))
1367 return -EINVAL;
1368 if (get_user(v, (int __user *)optval))
1369 return -EFAULT;
1370 mrt->mroute_do_assert = v;
1371 return 0;
1372 }
1373 #ifdef CONFIG_IP_PIMSM
1374 case MRT_PIM:
1375 {
1376 int v;
1377
1378 if (optlen != sizeof(v))
1379 return -EINVAL;
1380 if (get_user(v, (int __user *)optval))
1381 return -EFAULT;
1382 v = !!v;
1383
1384 rtnl_lock();
1385 ret = 0;
1386 if (v != mrt->mroute_do_pim) {
1387 mrt->mroute_do_pim = v;
1388 mrt->mroute_do_assert = v;
1389 }
1390 rtnl_unlock();
1391 return ret;
1392 }
1393 #endif
1394 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
1395 case MRT_TABLE:
1396 {
1397 u32 v;
1398
1399 if (optlen != sizeof(u32))
1400 return -EINVAL;
1401 if (get_user(v, (u32 __user *)optval))
1402 return -EFAULT;
1403
1404 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
1405 if (v != RT_TABLE_DEFAULT && v >= 1000000000)
1406 return -EINVAL;
1407
1408 rtnl_lock();
1409 ret = 0;
1410 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1411 ret = -EBUSY;
1412 } else {
1413 if (!ipmr_new_table(net, v))
1414 ret = -ENOMEM;
1415 else
1416 raw_sk(sk)->ipmr_table = v;
1417 }
1418 rtnl_unlock();
1419 return ret;
1420 }
1421 #endif
1422 /*
1423 * Spurious command, or MRT_VERSION which you cannot
1424 * set.
1425 */
1426 default:
1427 return -ENOPROTOOPT;
1428 }
1429 }
1430
1431 /*
1432 * Getsock opt support for the multicast routing system.
1433 */
1434
ip_mroute_getsockopt(struct sock * sk,int optname,char __user * optval,int __user * optlen)1435 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1436 {
1437 int olr;
1438 int val;
1439 struct net *net = sock_net(sk);
1440 struct mr_table *mrt;
1441
1442 if (sk->sk_type != SOCK_RAW ||
1443 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1444 return -EOPNOTSUPP;
1445
1446 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1447 if (mrt == NULL)
1448 return -ENOENT;
1449
1450 if (optname != MRT_VERSION &&
1451 #ifdef CONFIG_IP_PIMSM
1452 optname != MRT_PIM &&
1453 #endif
1454 optname != MRT_ASSERT)
1455 return -ENOPROTOOPT;
1456
1457 if (get_user(olr, optlen))
1458 return -EFAULT;
1459
1460 olr = min_t(unsigned int, olr, sizeof(int));
1461 if (olr < 0)
1462 return -EINVAL;
1463
1464 if (put_user(olr, optlen))
1465 return -EFAULT;
1466 if (optname == MRT_VERSION)
1467 val = 0x0305;
1468 #ifdef CONFIG_IP_PIMSM
1469 else if (optname == MRT_PIM)
1470 val = mrt->mroute_do_pim;
1471 #endif
1472 else
1473 val = mrt->mroute_do_assert;
1474 if (copy_to_user(optval, &val, olr))
1475 return -EFAULT;
1476 return 0;
1477 }
1478
1479 /*
1480 * The IP multicast ioctl support routines.
1481 */
1482
ipmr_ioctl(struct sock * sk,int cmd,void __user * arg)1483 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1484 {
1485 struct sioc_sg_req sr;
1486 struct sioc_vif_req vr;
1487 struct vif_device *vif;
1488 struct mfc_cache *c;
1489 struct net *net = sock_net(sk);
1490 struct mr_table *mrt;
1491
1492 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1493 if (mrt == NULL)
1494 return -ENOENT;
1495
1496 switch (cmd) {
1497 case SIOCGETVIFCNT:
1498 if (copy_from_user(&vr, arg, sizeof(vr)))
1499 return -EFAULT;
1500 if (vr.vifi >= mrt->maxvif)
1501 return -EINVAL;
1502 read_lock(&mrt_lock);
1503 vif = &mrt->vif_table[vr.vifi];
1504 if (VIF_EXISTS(mrt, vr.vifi)) {
1505 vr.icount = vif->pkt_in;
1506 vr.ocount = vif->pkt_out;
1507 vr.ibytes = vif->bytes_in;
1508 vr.obytes = vif->bytes_out;
1509 read_unlock(&mrt_lock);
1510
1511 if (copy_to_user(arg, &vr, sizeof(vr)))
1512 return -EFAULT;
1513 return 0;
1514 }
1515 read_unlock(&mrt_lock);
1516 return -EADDRNOTAVAIL;
1517 case SIOCGETSGCNT:
1518 if (copy_from_user(&sr, arg, sizeof(sr)))
1519 return -EFAULT;
1520
1521 rcu_read_lock();
1522 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1523 if (c) {
1524 sr.pktcnt = c->mfc_un.res.pkt;
1525 sr.bytecnt = c->mfc_un.res.bytes;
1526 sr.wrong_if = c->mfc_un.res.wrong_if;
1527 rcu_read_unlock();
1528
1529 if (copy_to_user(arg, &sr, sizeof(sr)))
1530 return -EFAULT;
1531 return 0;
1532 }
1533 rcu_read_unlock();
1534 return -EADDRNOTAVAIL;
1535 default:
1536 return -ENOIOCTLCMD;
1537 }
1538 }
1539
1540 #ifdef CONFIG_COMPAT
1541 struct compat_sioc_sg_req {
1542 struct in_addr src;
1543 struct in_addr grp;
1544 compat_ulong_t pktcnt;
1545 compat_ulong_t bytecnt;
1546 compat_ulong_t wrong_if;
1547 };
1548
1549 struct compat_sioc_vif_req {
1550 vifi_t vifi; /* Which iface */
1551 compat_ulong_t icount;
1552 compat_ulong_t ocount;
1553 compat_ulong_t ibytes;
1554 compat_ulong_t obytes;
1555 };
1556
ipmr_compat_ioctl(struct sock * sk,unsigned int cmd,void __user * arg)1557 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1558 {
1559 struct compat_sioc_sg_req sr;
1560 struct compat_sioc_vif_req vr;
1561 struct vif_device *vif;
1562 struct mfc_cache *c;
1563 struct net *net = sock_net(sk);
1564 struct mr_table *mrt;
1565
1566 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1567 if (mrt == NULL)
1568 return -ENOENT;
1569
1570 switch (cmd) {
1571 case SIOCGETVIFCNT:
1572 if (copy_from_user(&vr, arg, sizeof(vr)))
1573 return -EFAULT;
1574 if (vr.vifi >= mrt->maxvif)
1575 return -EINVAL;
1576 read_lock(&mrt_lock);
1577 vif = &mrt->vif_table[vr.vifi];
1578 if (VIF_EXISTS(mrt, vr.vifi)) {
1579 vr.icount = vif->pkt_in;
1580 vr.ocount = vif->pkt_out;
1581 vr.ibytes = vif->bytes_in;
1582 vr.obytes = vif->bytes_out;
1583 read_unlock(&mrt_lock);
1584
1585 if (copy_to_user(arg, &vr, sizeof(vr)))
1586 return -EFAULT;
1587 return 0;
1588 }
1589 read_unlock(&mrt_lock);
1590 return -EADDRNOTAVAIL;
1591 case SIOCGETSGCNT:
1592 if (copy_from_user(&sr, arg, sizeof(sr)))
1593 return -EFAULT;
1594
1595 rcu_read_lock();
1596 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1597 if (c) {
1598 sr.pktcnt = c->mfc_un.res.pkt;
1599 sr.bytecnt = c->mfc_un.res.bytes;
1600 sr.wrong_if = c->mfc_un.res.wrong_if;
1601 rcu_read_unlock();
1602
1603 if (copy_to_user(arg, &sr, sizeof(sr)))
1604 return -EFAULT;
1605 return 0;
1606 }
1607 rcu_read_unlock();
1608 return -EADDRNOTAVAIL;
1609 default:
1610 return -ENOIOCTLCMD;
1611 }
1612 }
1613 #endif
1614
1615
ipmr_device_event(struct notifier_block * this,unsigned long event,void * ptr)1616 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1617 {
1618 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1619 struct net *net = dev_net(dev);
1620 struct mr_table *mrt;
1621 struct vif_device *v;
1622 int ct;
1623
1624 if (event != NETDEV_UNREGISTER)
1625 return NOTIFY_DONE;
1626
1627 ipmr_for_each_table(mrt, net) {
1628 v = &mrt->vif_table[0];
1629 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1630 if (v->dev == dev)
1631 vif_delete(mrt, ct, 1, NULL);
1632 }
1633 }
1634 return NOTIFY_DONE;
1635 }
1636
1637
1638 static struct notifier_block ip_mr_notifier = {
1639 .notifier_call = ipmr_device_event,
1640 };
1641
1642 /*
1643 * Encapsulate a packet by attaching a valid IPIP header to it.
1644 * This avoids tunnel drivers and other mess and gives us the speed so
1645 * important for multicast video.
1646 */
1647
ip_encap(struct sk_buff * skb,__be32 saddr,__be32 daddr)1648 static void ip_encap(struct sk_buff *skb, __be32 saddr, __be32 daddr)
1649 {
1650 struct iphdr *iph;
1651 const struct iphdr *old_iph = ip_hdr(skb);
1652
1653 skb_push(skb, sizeof(struct iphdr));
1654 skb->transport_header = skb->network_header;
1655 skb_reset_network_header(skb);
1656 iph = ip_hdr(skb);
1657
1658 iph->version = 4;
1659 iph->tos = old_iph->tos;
1660 iph->ttl = old_iph->ttl;
1661 iph->frag_off = 0;
1662 iph->daddr = daddr;
1663 iph->saddr = saddr;
1664 iph->protocol = IPPROTO_IPIP;
1665 iph->ihl = 5;
1666 iph->tot_len = htons(skb->len);
1667 ip_select_ident(skb, NULL);
1668 ip_send_check(iph);
1669
1670 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1671 nf_reset(skb);
1672 }
1673
ipmr_forward_finish(struct sk_buff * skb)1674 static inline int ipmr_forward_finish(struct sk_buff *skb)
1675 {
1676 struct ip_options *opt = &(IPCB(skb)->opt);
1677
1678 IP_INC_STATS(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTFORWDATAGRAMS);
1679 IP_ADD_STATS(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTOCTETS, skb->len);
1680
1681 if (unlikely(opt->optlen))
1682 ip_forward_options(skb);
1683
1684 return dst_output(skb);
1685 }
1686
1687 /*
1688 * Processing handlers for ipmr_forward
1689 */
1690
ipmr_queue_xmit(struct net * net,struct mr_table * mrt,struct sk_buff * skb,struct mfc_cache * c,int vifi)1691 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1692 struct sk_buff *skb, struct mfc_cache *c, int vifi)
1693 {
1694 const struct iphdr *iph = ip_hdr(skb);
1695 struct vif_device *vif = &mrt->vif_table[vifi];
1696 struct net_device *dev;
1697 struct rtable *rt;
1698 struct flowi4 fl4;
1699 int encap = 0;
1700
1701 if (vif->dev == NULL)
1702 goto out_free;
1703
1704 #ifdef CONFIG_IP_PIMSM
1705 if (vif->flags & VIFF_REGISTER) {
1706 vif->pkt_out++;
1707 vif->bytes_out += skb->len;
1708 vif->dev->stats.tx_bytes += skb->len;
1709 vif->dev->stats.tx_packets++;
1710 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1711 goto out_free;
1712 }
1713 #endif
1714
1715 if (vif->flags & VIFF_TUNNEL) {
1716 rt = ip_route_output_ports(net, &fl4, NULL,
1717 vif->remote, vif->local,
1718 0, 0,
1719 IPPROTO_IPIP,
1720 RT_TOS(iph->tos), vif->link);
1721 if (IS_ERR(rt))
1722 goto out_free;
1723 encap = sizeof(struct iphdr);
1724 } else {
1725 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1726 0, 0,
1727 IPPROTO_IPIP,
1728 RT_TOS(iph->tos), vif->link);
1729 if (IS_ERR(rt))
1730 goto out_free;
1731 }
1732
1733 dev = rt->dst.dev;
1734
1735 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1736 /* Do not fragment multicasts. Alas, IPv4 does not
1737 * allow to send ICMP, so that packets will disappear
1738 * to blackhole.
1739 */
1740
1741 IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
1742 ip_rt_put(rt);
1743 goto out_free;
1744 }
1745
1746 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1747
1748 if (skb_cow(skb, encap)) {
1749 ip_rt_put(rt);
1750 goto out_free;
1751 }
1752
1753 vif->pkt_out++;
1754 vif->bytes_out += skb->len;
1755
1756 skb_dst_drop(skb);
1757 skb_dst_set(skb, &rt->dst);
1758 ip_decrease_ttl(ip_hdr(skb));
1759
1760 /* FIXME: forward and output firewalls used to be called here.
1761 * What do we do with netfilter? -- RR
1762 */
1763 if (vif->flags & VIFF_TUNNEL) {
1764 ip_encap(skb, vif->local, vif->remote);
1765 /* FIXME: extra output firewall step used to be here. --RR */
1766 vif->dev->stats.tx_packets++;
1767 vif->dev->stats.tx_bytes += skb->len;
1768 }
1769
1770 IPCB(skb)->flags |= IPSKB_FORWARDED;
1771
1772 /*
1773 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1774 * not only before forwarding, but after forwarding on all output
1775 * interfaces. It is clear, if mrouter runs a multicasting
1776 * program, it should receive packets not depending to what interface
1777 * program is joined.
1778 * If we will not make it, the program will have to join on all
1779 * interfaces. On the other hand, multihoming host (or router, but
1780 * not mrouter) cannot join to more than one interface - it will
1781 * result in receiving multiple packets.
1782 */
1783 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, skb, skb->dev, dev,
1784 ipmr_forward_finish);
1785 return;
1786
1787 out_free:
1788 kfree_skb(skb);
1789 }
1790
ipmr_find_vif(struct mr_table * mrt,struct net_device * dev)1791 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1792 {
1793 int ct;
1794
1795 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1796 if (mrt->vif_table[ct].dev == dev)
1797 break;
1798 }
1799 return ct;
1800 }
1801
1802 /* "local" means that we should preserve one skb (for local delivery) */
1803
ip_mr_forward(struct net * net,struct mr_table * mrt,struct sk_buff * skb,struct mfc_cache * cache,int local)1804 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1805 struct sk_buff *skb, struct mfc_cache *cache,
1806 int local)
1807 {
1808 int psend = -1;
1809 int vif, ct;
1810 int true_vifi = ipmr_find_vif(mrt, skb->dev);
1811
1812 vif = cache->mfc_parent;
1813 cache->mfc_un.res.pkt++;
1814 cache->mfc_un.res.bytes += skb->len;
1815
1816 if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1817 struct mfc_cache *cache_proxy;
1818
1819 /* For an (*,G) entry, we only check that the incomming
1820 * interface is part of the static tree.
1821 */
1822 cache_proxy = ipmr_cache_find_any_parent(mrt, vif);
1823 if (cache_proxy &&
1824 cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
1825 goto forward;
1826 }
1827
1828 /*
1829 * Wrong interface: drop packet and (maybe) send PIM assert.
1830 */
1831 if (mrt->vif_table[vif].dev != skb->dev) {
1832 if (rt_is_output_route(skb_rtable(skb))) {
1833 /* It is our own packet, looped back.
1834 * Very complicated situation...
1835 *
1836 * The best workaround until routing daemons will be
1837 * fixed is not to redistribute packet, if it was
1838 * send through wrong interface. It means, that
1839 * multicast applications WILL NOT work for
1840 * (S,G), which have default multicast route pointing
1841 * to wrong oif. In any case, it is not a good
1842 * idea to use multicasting applications on router.
1843 */
1844 goto dont_forward;
1845 }
1846
1847 cache->mfc_un.res.wrong_if++;
1848
1849 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1850 /* pimsm uses asserts, when switching from RPT to SPT,
1851 * so that we cannot check that packet arrived on an oif.
1852 * It is bad, but otherwise we would need to move pretty
1853 * large chunk of pimd to kernel. Ough... --ANK
1854 */
1855 (mrt->mroute_do_pim ||
1856 cache->mfc_un.res.ttls[true_vifi] < 255) &&
1857 time_after(jiffies,
1858 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1859 cache->mfc_un.res.last_assert = jiffies;
1860 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1861 }
1862 goto dont_forward;
1863 }
1864
1865 forward:
1866 mrt->vif_table[vif].pkt_in++;
1867 mrt->vif_table[vif].bytes_in += skb->len;
1868
1869 /*
1870 * Forward the frame
1871 */
1872 if (cache->mfc_origin == htonl(INADDR_ANY) &&
1873 cache->mfc_mcastgrp == htonl(INADDR_ANY)) {
1874 if (true_vifi >= 0 &&
1875 true_vifi != cache->mfc_parent &&
1876 ip_hdr(skb)->ttl >
1877 cache->mfc_un.res.ttls[cache->mfc_parent]) {
1878 /* It's an (*,*) entry and the packet is not coming from
1879 * the upstream: forward the packet to the upstream
1880 * only.
1881 */
1882 psend = cache->mfc_parent;
1883 goto last_forward;
1884 }
1885 goto dont_forward;
1886 }
1887 for (ct = cache->mfc_un.res.maxvif - 1;
1888 ct >= cache->mfc_un.res.minvif; ct--) {
1889 /* For (*,G) entry, don't forward to the incoming interface */
1890 if ((cache->mfc_origin != htonl(INADDR_ANY) ||
1891 ct != true_vifi) &&
1892 ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
1893 if (psend != -1) {
1894 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1895
1896 if (skb2)
1897 ipmr_queue_xmit(net, mrt, skb2, cache,
1898 psend);
1899 }
1900 psend = ct;
1901 }
1902 }
1903 last_forward:
1904 if (psend != -1) {
1905 if (local) {
1906 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1907
1908 if (skb2)
1909 ipmr_queue_xmit(net, mrt, skb2, cache, psend);
1910 } else {
1911 ipmr_queue_xmit(net, mrt, skb, cache, psend);
1912 return;
1913 }
1914 }
1915
1916 dont_forward:
1917 if (!local)
1918 kfree_skb(skb);
1919 }
1920
ipmr_rt_fib_lookup(struct net * net,struct sk_buff * skb)1921 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
1922 {
1923 struct rtable *rt = skb_rtable(skb);
1924 struct iphdr *iph = ip_hdr(skb);
1925 struct flowi4 fl4 = {
1926 .daddr = iph->daddr,
1927 .saddr = iph->saddr,
1928 .flowi4_tos = RT_TOS(iph->tos),
1929 .flowi4_oif = (rt_is_output_route(rt) ?
1930 skb->dev->ifindex : 0),
1931 .flowi4_iif = (rt_is_output_route(rt) ?
1932 LOOPBACK_IFINDEX :
1933 skb->dev->ifindex),
1934 .flowi4_mark = skb->mark,
1935 };
1936 struct mr_table *mrt;
1937 int err;
1938
1939 err = ipmr_fib_lookup(net, &fl4, &mrt);
1940 if (err)
1941 return ERR_PTR(err);
1942 return mrt;
1943 }
1944
1945 /*
1946 * Multicast packets for forwarding arrive here
1947 * Called with rcu_read_lock();
1948 */
1949
ip_mr_input(struct sk_buff * skb)1950 int ip_mr_input(struct sk_buff *skb)
1951 {
1952 struct mfc_cache *cache;
1953 struct net *net = dev_net(skb->dev);
1954 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
1955 struct mr_table *mrt;
1956
1957 /* Packet is looped back after forward, it should not be
1958 * forwarded second time, but still can be delivered locally.
1959 */
1960 if (IPCB(skb)->flags & IPSKB_FORWARDED)
1961 goto dont_forward;
1962
1963 mrt = ipmr_rt_fib_lookup(net, skb);
1964 if (IS_ERR(mrt)) {
1965 kfree_skb(skb);
1966 return PTR_ERR(mrt);
1967 }
1968 if (!local) {
1969 if (IPCB(skb)->opt.router_alert) {
1970 if (ip_call_ra_chain(skb))
1971 return 0;
1972 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
1973 /* IGMPv1 (and broken IGMPv2 implementations sort of
1974 * Cisco IOS <= 11.2(8)) do not put router alert
1975 * option to IGMP packets destined to routable
1976 * groups. It is very bad, because it means
1977 * that we can forward NO IGMP messages.
1978 */
1979 struct sock *mroute_sk;
1980
1981 mroute_sk = rcu_dereference(mrt->mroute_sk);
1982 if (mroute_sk) {
1983 nf_reset(skb);
1984 raw_rcv(mroute_sk, skb);
1985 return 0;
1986 }
1987 }
1988 }
1989
1990 /* already under rcu_read_lock() */
1991 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
1992 if (cache == NULL) {
1993 int vif = ipmr_find_vif(mrt, skb->dev);
1994
1995 if (vif >= 0)
1996 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
1997 vif);
1998 }
1999
2000 /*
2001 * No usable cache entry
2002 */
2003 if (cache == NULL) {
2004 int vif;
2005
2006 if (local) {
2007 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2008 ip_local_deliver(skb);
2009 if (skb2 == NULL)
2010 return -ENOBUFS;
2011 skb = skb2;
2012 }
2013
2014 read_lock(&mrt_lock);
2015 vif = ipmr_find_vif(mrt, skb->dev);
2016 if (vif >= 0) {
2017 int err2 = ipmr_cache_unresolved(mrt, vif, skb);
2018 read_unlock(&mrt_lock);
2019
2020 return err2;
2021 }
2022 read_unlock(&mrt_lock);
2023 kfree_skb(skb);
2024 return -ENODEV;
2025 }
2026
2027 read_lock(&mrt_lock);
2028 ip_mr_forward(net, mrt, skb, cache, local);
2029 read_unlock(&mrt_lock);
2030
2031 if (local)
2032 return ip_local_deliver(skb);
2033
2034 return 0;
2035
2036 dont_forward:
2037 if (local)
2038 return ip_local_deliver(skb);
2039 kfree_skb(skb);
2040 return 0;
2041 }
2042
2043 #ifdef CONFIG_IP_PIMSM
2044 /* called with rcu_read_lock() */
__pim_rcv(struct mr_table * mrt,struct sk_buff * skb,unsigned int pimlen)2045 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
2046 unsigned int pimlen)
2047 {
2048 struct net_device *reg_dev = NULL;
2049 struct iphdr *encap;
2050
2051 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
2052 /*
2053 * Check that:
2054 * a. packet is really sent to a multicast group
2055 * b. packet is not a NULL-REGISTER
2056 * c. packet is not truncated
2057 */
2058 if (!ipv4_is_multicast(encap->daddr) ||
2059 encap->tot_len == 0 ||
2060 ntohs(encap->tot_len) + pimlen > skb->len)
2061 return 1;
2062
2063 read_lock(&mrt_lock);
2064 if (mrt->mroute_reg_vif_num >= 0)
2065 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
2066 read_unlock(&mrt_lock);
2067
2068 if (reg_dev == NULL)
2069 return 1;
2070
2071 skb->mac_header = skb->network_header;
2072 skb_pull(skb, (u8 *)encap - skb->data);
2073 skb_reset_network_header(skb);
2074 skb->protocol = htons(ETH_P_IP);
2075 skb->ip_summed = CHECKSUM_NONE;
2076
2077 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
2078
2079 netif_rx(skb);
2080
2081 return NET_RX_SUCCESS;
2082 }
2083 #endif
2084
2085 #ifdef CONFIG_IP_PIMSM_V1
2086 /*
2087 * Handle IGMP messages of PIMv1
2088 */
2089
pim_rcv_v1(struct sk_buff * skb)2090 int pim_rcv_v1(struct sk_buff *skb)
2091 {
2092 struct igmphdr *pim;
2093 struct net *net = dev_net(skb->dev);
2094 struct mr_table *mrt;
2095
2096 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2097 goto drop;
2098
2099 pim = igmp_hdr(skb);
2100
2101 mrt = ipmr_rt_fib_lookup(net, skb);
2102 if (IS_ERR(mrt))
2103 goto drop;
2104 if (!mrt->mroute_do_pim ||
2105 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2106 goto drop;
2107
2108 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2109 drop:
2110 kfree_skb(skb);
2111 }
2112 return 0;
2113 }
2114 #endif
2115
2116 #ifdef CONFIG_IP_PIMSM_V2
pim_rcv(struct sk_buff * skb)2117 static int pim_rcv(struct sk_buff *skb)
2118 {
2119 struct pimreghdr *pim;
2120 struct net *net = dev_net(skb->dev);
2121 struct mr_table *mrt;
2122
2123 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2124 goto drop;
2125
2126 pim = (struct pimreghdr *)skb_transport_header(skb);
2127 if (pim->type != ((PIM_VERSION << 4) | (PIM_REGISTER)) ||
2128 (pim->flags & PIM_NULL_REGISTER) ||
2129 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2130 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2131 goto drop;
2132
2133 mrt = ipmr_rt_fib_lookup(net, skb);
2134 if (IS_ERR(mrt))
2135 goto drop;
2136 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2137 drop:
2138 kfree_skb(skb);
2139 }
2140 return 0;
2141 }
2142 #endif
2143
__ipmr_fill_mroute(struct mr_table * mrt,struct sk_buff * skb,struct mfc_cache * c,struct rtmsg * rtm)2144 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2145 struct mfc_cache *c, struct rtmsg *rtm)
2146 {
2147 int ct;
2148 struct rtnexthop *nhp;
2149 struct nlattr *mp_attr;
2150 struct rta_mfc_stats mfcs;
2151
2152 /* If cache is unresolved, don't try to parse IIF and OIF */
2153 if (c->mfc_parent >= MAXVIFS)
2154 return -ENOENT;
2155
2156 if (VIF_EXISTS(mrt, c->mfc_parent) &&
2157 nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
2158 return -EMSGSIZE;
2159
2160 if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH)))
2161 return -EMSGSIZE;
2162
2163 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2164 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2165 if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) {
2166 nla_nest_cancel(skb, mp_attr);
2167 return -EMSGSIZE;
2168 }
2169
2170 nhp->rtnh_flags = 0;
2171 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2172 nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex;
2173 nhp->rtnh_len = sizeof(*nhp);
2174 }
2175 }
2176
2177 nla_nest_end(skb, mp_attr);
2178
2179 mfcs.mfcs_packets = c->mfc_un.res.pkt;
2180 mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2181 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2182 if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0)
2183 return -EMSGSIZE;
2184
2185 rtm->rtm_type = RTN_MULTICAST;
2186 return 1;
2187 }
2188
ipmr_get_route(struct net * net,struct sk_buff * skb,__be32 saddr,__be32 daddr,struct rtmsg * rtm,int nowait,u32 portid)2189 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2190 __be32 saddr, __be32 daddr,
2191 struct rtmsg *rtm, int nowait, u32 portid)
2192 {
2193 struct mfc_cache *cache;
2194 struct mr_table *mrt;
2195 int err;
2196
2197 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2198 if (mrt == NULL)
2199 return -ENOENT;
2200
2201 rcu_read_lock();
2202 cache = ipmr_cache_find(mrt, saddr, daddr);
2203 if (cache == NULL && skb->dev) {
2204 int vif = ipmr_find_vif(mrt, skb->dev);
2205
2206 if (vif >= 0)
2207 cache = ipmr_cache_find_any(mrt, daddr, vif);
2208 }
2209 if (cache == NULL) {
2210 struct sk_buff *skb2;
2211 struct iphdr *iph;
2212 struct net_device *dev;
2213 int vif = -1;
2214
2215 if (nowait) {
2216 rcu_read_unlock();
2217 return -EAGAIN;
2218 }
2219
2220 dev = skb->dev;
2221 read_lock(&mrt_lock);
2222 if (dev)
2223 vif = ipmr_find_vif(mrt, dev);
2224 if (vif < 0) {
2225 read_unlock(&mrt_lock);
2226 rcu_read_unlock();
2227 return -ENODEV;
2228 }
2229 skb2 = skb_clone(skb, GFP_ATOMIC);
2230 if (!skb2) {
2231 read_unlock(&mrt_lock);
2232 rcu_read_unlock();
2233 return -ENOMEM;
2234 }
2235
2236 NETLINK_CB(skb2).portid = portid;
2237 skb_push(skb2, sizeof(struct iphdr));
2238 skb_reset_network_header(skb2);
2239 iph = ip_hdr(skb2);
2240 iph->ihl = sizeof(struct iphdr) >> 2;
2241 iph->saddr = saddr;
2242 iph->daddr = daddr;
2243 iph->version = 0;
2244 err = ipmr_cache_unresolved(mrt, vif, skb2);
2245 read_unlock(&mrt_lock);
2246 rcu_read_unlock();
2247 return err;
2248 }
2249
2250 read_lock(&mrt_lock);
2251 if (!nowait && (rtm->rtm_flags & RTM_F_NOTIFY))
2252 cache->mfc_flags |= MFC_NOTIFY;
2253 err = __ipmr_fill_mroute(mrt, skb, cache, rtm);
2254 read_unlock(&mrt_lock);
2255 rcu_read_unlock();
2256 return err;
2257 }
2258
ipmr_fill_mroute(struct mr_table * mrt,struct sk_buff * skb,u32 portid,u32 seq,struct mfc_cache * c,int cmd,int flags)2259 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2260 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2261 int flags)
2262 {
2263 struct nlmsghdr *nlh;
2264 struct rtmsg *rtm;
2265 int err;
2266
2267 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2268 if (nlh == NULL)
2269 return -EMSGSIZE;
2270
2271 rtm = nlmsg_data(nlh);
2272 rtm->rtm_family = RTNL_FAMILY_IPMR;
2273 rtm->rtm_dst_len = 32;
2274 rtm->rtm_src_len = 32;
2275 rtm->rtm_tos = 0;
2276 rtm->rtm_table = mrt->id;
2277 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2278 goto nla_put_failure;
2279 rtm->rtm_type = RTN_MULTICAST;
2280 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2281 if (c->mfc_flags & MFC_STATIC)
2282 rtm->rtm_protocol = RTPROT_STATIC;
2283 else
2284 rtm->rtm_protocol = RTPROT_MROUTED;
2285 rtm->rtm_flags = 0;
2286
2287 if (nla_put_be32(skb, RTA_SRC, c->mfc_origin) ||
2288 nla_put_be32(skb, RTA_DST, c->mfc_mcastgrp))
2289 goto nla_put_failure;
2290 err = __ipmr_fill_mroute(mrt, skb, c, rtm);
2291 /* do not break the dump if cache is unresolved */
2292 if (err < 0 && err != -ENOENT)
2293 goto nla_put_failure;
2294
2295 return nlmsg_end(skb, nlh);
2296
2297 nla_put_failure:
2298 nlmsg_cancel(skb, nlh);
2299 return -EMSGSIZE;
2300 }
2301
mroute_msgsize(bool unresolved,int maxvif)2302 static size_t mroute_msgsize(bool unresolved, int maxvif)
2303 {
2304 size_t len =
2305 NLMSG_ALIGN(sizeof(struct rtmsg))
2306 + nla_total_size(4) /* RTA_TABLE */
2307 + nla_total_size(4) /* RTA_SRC */
2308 + nla_total_size(4) /* RTA_DST */
2309 ;
2310
2311 if (!unresolved)
2312 len = len
2313 + nla_total_size(4) /* RTA_IIF */
2314 + nla_total_size(0) /* RTA_MULTIPATH */
2315 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2316 /* RTA_MFC_STATS */
2317 + nla_total_size(sizeof(struct rta_mfc_stats))
2318 ;
2319
2320 return len;
2321 }
2322
mroute_netlink_event(struct mr_table * mrt,struct mfc_cache * mfc,int cmd)2323 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2324 int cmd)
2325 {
2326 struct net *net = read_pnet(&mrt->net);
2327 struct sk_buff *skb;
2328 int err = -ENOBUFS;
2329
2330 skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif),
2331 GFP_ATOMIC);
2332 if (skb == NULL)
2333 goto errout;
2334
2335 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2336 if (err < 0)
2337 goto errout;
2338
2339 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2340 return;
2341
2342 errout:
2343 kfree_skb(skb);
2344 if (err < 0)
2345 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2346 }
2347
ipmr_rtm_dumproute(struct sk_buff * skb,struct netlink_callback * cb)2348 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2349 {
2350 struct net *net = sock_net(skb->sk);
2351 struct mr_table *mrt;
2352 struct mfc_cache *mfc;
2353 unsigned int t = 0, s_t;
2354 unsigned int h = 0, s_h;
2355 unsigned int e = 0, s_e;
2356
2357 s_t = cb->args[0];
2358 s_h = cb->args[1];
2359 s_e = cb->args[2];
2360
2361 rcu_read_lock();
2362 ipmr_for_each_table(mrt, net) {
2363 if (t < s_t)
2364 goto next_table;
2365 if (t > s_t)
2366 s_h = 0;
2367 for (h = s_h; h < MFC_LINES; h++) {
2368 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_array[h], list) {
2369 if (e < s_e)
2370 goto next_entry;
2371 if (ipmr_fill_mroute(mrt, skb,
2372 NETLINK_CB(cb->skb).portid,
2373 cb->nlh->nlmsg_seq,
2374 mfc, RTM_NEWROUTE,
2375 NLM_F_MULTI) < 0)
2376 goto done;
2377 next_entry:
2378 e++;
2379 }
2380 e = s_e = 0;
2381 }
2382 spin_lock_bh(&mfc_unres_lock);
2383 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
2384 if (e < s_e)
2385 goto next_entry2;
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 spin_unlock_bh(&mfc_unres_lock);
2392 goto done;
2393 }
2394 next_entry2:
2395 e++;
2396 }
2397 spin_unlock_bh(&mfc_unres_lock);
2398 e = s_e = 0;
2399 s_h = 0;
2400 next_table:
2401 t++;
2402 }
2403 done:
2404 rcu_read_unlock();
2405
2406 cb->args[2] = e;
2407 cb->args[1] = h;
2408 cb->args[0] = t;
2409
2410 return skb->len;
2411 }
2412
2413 #ifdef CONFIG_PROC_FS
2414 /*
2415 * The /proc interfaces to multicast routing :
2416 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2417 */
2418 struct ipmr_vif_iter {
2419 struct seq_net_private p;
2420 struct mr_table *mrt;
2421 int ct;
2422 };
2423
ipmr_vif_seq_idx(struct net * net,struct ipmr_vif_iter * iter,loff_t pos)2424 static struct vif_device *ipmr_vif_seq_idx(struct net *net,
2425 struct ipmr_vif_iter *iter,
2426 loff_t pos)
2427 {
2428 struct mr_table *mrt = iter->mrt;
2429
2430 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
2431 if (!VIF_EXISTS(mrt, iter->ct))
2432 continue;
2433 if (pos-- == 0)
2434 return &mrt->vif_table[iter->ct];
2435 }
2436 return NULL;
2437 }
2438
ipmr_vif_seq_start(struct seq_file * seq,loff_t * pos)2439 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2440 __acquires(mrt_lock)
2441 {
2442 struct ipmr_vif_iter *iter = seq->private;
2443 struct net *net = seq_file_net(seq);
2444 struct mr_table *mrt;
2445
2446 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2447 if (mrt == NULL)
2448 return ERR_PTR(-ENOENT);
2449
2450 iter->mrt = mrt;
2451
2452 read_lock(&mrt_lock);
2453 return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
2454 : SEQ_START_TOKEN;
2455 }
2456
ipmr_vif_seq_next(struct seq_file * seq,void * v,loff_t * pos)2457 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2458 {
2459 struct ipmr_vif_iter *iter = seq->private;
2460 struct net *net = seq_file_net(seq);
2461 struct mr_table *mrt = iter->mrt;
2462
2463 ++*pos;
2464 if (v == SEQ_START_TOKEN)
2465 return ipmr_vif_seq_idx(net, iter, 0);
2466
2467 while (++iter->ct < mrt->maxvif) {
2468 if (!VIF_EXISTS(mrt, iter->ct))
2469 continue;
2470 return &mrt->vif_table[iter->ct];
2471 }
2472 return NULL;
2473 }
2474
ipmr_vif_seq_stop(struct seq_file * seq,void * v)2475 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2476 __releases(mrt_lock)
2477 {
2478 read_unlock(&mrt_lock);
2479 }
2480
ipmr_vif_seq_show(struct seq_file * seq,void * v)2481 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2482 {
2483 struct ipmr_vif_iter *iter = seq->private;
2484 struct mr_table *mrt = iter->mrt;
2485
2486 if (v == SEQ_START_TOKEN) {
2487 seq_puts(seq,
2488 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2489 } else {
2490 const struct vif_device *vif = v;
2491 const char *name = vif->dev ? vif->dev->name : "none";
2492
2493 seq_printf(seq,
2494 "%2Zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2495 vif - mrt->vif_table,
2496 name, vif->bytes_in, vif->pkt_in,
2497 vif->bytes_out, vif->pkt_out,
2498 vif->flags, vif->local, vif->remote);
2499 }
2500 return 0;
2501 }
2502
2503 static const struct seq_operations ipmr_vif_seq_ops = {
2504 .start = ipmr_vif_seq_start,
2505 .next = ipmr_vif_seq_next,
2506 .stop = ipmr_vif_seq_stop,
2507 .show = ipmr_vif_seq_show,
2508 };
2509
ipmr_vif_open(struct inode * inode,struct file * file)2510 static int ipmr_vif_open(struct inode *inode, struct file *file)
2511 {
2512 return seq_open_net(inode, file, &ipmr_vif_seq_ops,
2513 sizeof(struct ipmr_vif_iter));
2514 }
2515
2516 static const struct file_operations ipmr_vif_fops = {
2517 .owner = THIS_MODULE,
2518 .open = ipmr_vif_open,
2519 .read = seq_read,
2520 .llseek = seq_lseek,
2521 .release = seq_release_net,
2522 };
2523
2524 struct ipmr_mfc_iter {
2525 struct seq_net_private p;
2526 struct mr_table *mrt;
2527 struct list_head *cache;
2528 int ct;
2529 };
2530
2531
ipmr_mfc_seq_idx(struct net * net,struct ipmr_mfc_iter * it,loff_t pos)2532 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
2533 struct ipmr_mfc_iter *it, loff_t pos)
2534 {
2535 struct mr_table *mrt = it->mrt;
2536 struct mfc_cache *mfc;
2537
2538 rcu_read_lock();
2539 for (it->ct = 0; it->ct < MFC_LINES; it->ct++) {
2540 it->cache = &mrt->mfc_cache_array[it->ct];
2541 list_for_each_entry_rcu(mfc, it->cache, list)
2542 if (pos-- == 0)
2543 return mfc;
2544 }
2545 rcu_read_unlock();
2546
2547 spin_lock_bh(&mfc_unres_lock);
2548 it->cache = &mrt->mfc_unres_queue;
2549 list_for_each_entry(mfc, it->cache, list)
2550 if (pos-- == 0)
2551 return mfc;
2552 spin_unlock_bh(&mfc_unres_lock);
2553
2554 it->cache = NULL;
2555 return NULL;
2556 }
2557
2558
ipmr_mfc_seq_start(struct seq_file * seq,loff_t * pos)2559 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2560 {
2561 struct ipmr_mfc_iter *it = seq->private;
2562 struct net *net = seq_file_net(seq);
2563 struct mr_table *mrt;
2564
2565 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2566 if (mrt == NULL)
2567 return ERR_PTR(-ENOENT);
2568
2569 it->mrt = mrt;
2570 it->cache = NULL;
2571 it->ct = 0;
2572 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
2573 : SEQ_START_TOKEN;
2574 }
2575
ipmr_mfc_seq_next(struct seq_file * seq,void * v,loff_t * pos)2576 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2577 {
2578 struct mfc_cache *mfc = v;
2579 struct ipmr_mfc_iter *it = seq->private;
2580 struct net *net = seq_file_net(seq);
2581 struct mr_table *mrt = it->mrt;
2582
2583 ++*pos;
2584
2585 if (v == SEQ_START_TOKEN)
2586 return ipmr_mfc_seq_idx(net, seq->private, 0);
2587
2588 if (mfc->list.next != it->cache)
2589 return list_entry(mfc->list.next, struct mfc_cache, list);
2590
2591 if (it->cache == &mrt->mfc_unres_queue)
2592 goto end_of_list;
2593
2594 BUG_ON(it->cache != &mrt->mfc_cache_array[it->ct]);
2595
2596 while (++it->ct < MFC_LINES) {
2597 it->cache = &mrt->mfc_cache_array[it->ct];
2598 if (list_empty(it->cache))
2599 continue;
2600 return list_first_entry(it->cache, struct mfc_cache, list);
2601 }
2602
2603 /* exhausted cache_array, show unresolved */
2604 rcu_read_unlock();
2605 it->cache = &mrt->mfc_unres_queue;
2606 it->ct = 0;
2607
2608 spin_lock_bh(&mfc_unres_lock);
2609 if (!list_empty(it->cache))
2610 return list_first_entry(it->cache, struct mfc_cache, list);
2611
2612 end_of_list:
2613 spin_unlock_bh(&mfc_unres_lock);
2614 it->cache = NULL;
2615
2616 return NULL;
2617 }
2618
ipmr_mfc_seq_stop(struct seq_file * seq,void * v)2619 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
2620 {
2621 struct ipmr_mfc_iter *it = seq->private;
2622 struct mr_table *mrt = it->mrt;
2623
2624 if (it->cache == &mrt->mfc_unres_queue)
2625 spin_unlock_bh(&mfc_unres_lock);
2626 else if (it->cache == &mrt->mfc_cache_array[it->ct])
2627 rcu_read_unlock();
2628 }
2629
ipmr_mfc_seq_show(struct seq_file * seq,void * v)2630 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2631 {
2632 int n;
2633
2634 if (v == SEQ_START_TOKEN) {
2635 seq_puts(seq,
2636 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2637 } else {
2638 const struct mfc_cache *mfc = v;
2639 const struct ipmr_mfc_iter *it = seq->private;
2640 const struct mr_table *mrt = it->mrt;
2641
2642 seq_printf(seq, "%08X %08X %-3hd",
2643 (__force u32) mfc->mfc_mcastgrp,
2644 (__force u32) mfc->mfc_origin,
2645 mfc->mfc_parent);
2646
2647 if (it->cache != &mrt->mfc_unres_queue) {
2648 seq_printf(seq, " %8lu %8lu %8lu",
2649 mfc->mfc_un.res.pkt,
2650 mfc->mfc_un.res.bytes,
2651 mfc->mfc_un.res.wrong_if);
2652 for (n = mfc->mfc_un.res.minvif;
2653 n < mfc->mfc_un.res.maxvif; n++) {
2654 if (VIF_EXISTS(mrt, n) &&
2655 mfc->mfc_un.res.ttls[n] < 255)
2656 seq_printf(seq,
2657 " %2d:%-3d",
2658 n, mfc->mfc_un.res.ttls[n]);
2659 }
2660 } else {
2661 /* unresolved mfc_caches don't contain
2662 * pkt, bytes and wrong_if values
2663 */
2664 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2665 }
2666 seq_putc(seq, '\n');
2667 }
2668 return 0;
2669 }
2670
2671 static const struct seq_operations ipmr_mfc_seq_ops = {
2672 .start = ipmr_mfc_seq_start,
2673 .next = ipmr_mfc_seq_next,
2674 .stop = ipmr_mfc_seq_stop,
2675 .show = ipmr_mfc_seq_show,
2676 };
2677
ipmr_mfc_open(struct inode * inode,struct file * file)2678 static int ipmr_mfc_open(struct inode *inode, struct file *file)
2679 {
2680 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
2681 sizeof(struct ipmr_mfc_iter));
2682 }
2683
2684 static const struct file_operations ipmr_mfc_fops = {
2685 .owner = THIS_MODULE,
2686 .open = ipmr_mfc_open,
2687 .read = seq_read,
2688 .llseek = seq_lseek,
2689 .release = seq_release_net,
2690 };
2691 #endif
2692
2693 #ifdef CONFIG_IP_PIMSM_V2
2694 static const struct net_protocol pim_protocol = {
2695 .handler = pim_rcv,
2696 .netns_ok = 1,
2697 };
2698 #endif
2699
2700
2701 /*
2702 * Setup for IP multicast routing
2703 */
ipmr_net_init(struct net * net)2704 static int __net_init ipmr_net_init(struct net *net)
2705 {
2706 int err;
2707
2708 err = ipmr_rules_init(net);
2709 if (err < 0)
2710 goto fail;
2711
2712 #ifdef CONFIG_PROC_FS
2713 err = -ENOMEM;
2714 if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops))
2715 goto proc_vif_fail;
2716 if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops))
2717 goto proc_cache_fail;
2718 #endif
2719 return 0;
2720
2721 #ifdef CONFIG_PROC_FS
2722 proc_cache_fail:
2723 remove_proc_entry("ip_mr_vif", net->proc_net);
2724 proc_vif_fail:
2725 ipmr_rules_exit(net);
2726 #endif
2727 fail:
2728 return err;
2729 }
2730
ipmr_net_exit(struct net * net)2731 static void __net_exit ipmr_net_exit(struct net *net)
2732 {
2733 #ifdef CONFIG_PROC_FS
2734 remove_proc_entry("ip_mr_cache", net->proc_net);
2735 remove_proc_entry("ip_mr_vif", net->proc_net);
2736 #endif
2737 ipmr_rules_exit(net);
2738 }
2739
2740 static struct pernet_operations ipmr_net_ops = {
2741 .init = ipmr_net_init,
2742 .exit = ipmr_net_exit,
2743 };
2744
ip_mr_init(void)2745 int __init ip_mr_init(void)
2746 {
2747 int err;
2748
2749 mrt_cachep = kmem_cache_create("ip_mrt_cache",
2750 sizeof(struct mfc_cache),
2751 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
2752 NULL);
2753 if (!mrt_cachep)
2754 return -ENOMEM;
2755
2756 err = register_pernet_subsys(&ipmr_net_ops);
2757 if (err)
2758 goto reg_pernet_fail;
2759
2760 err = register_netdevice_notifier(&ip_mr_notifier);
2761 if (err)
2762 goto reg_notif_fail;
2763 #ifdef CONFIG_IP_PIMSM_V2
2764 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
2765 pr_err("%s: can't add PIM protocol\n", __func__);
2766 err = -EAGAIN;
2767 goto add_proto_fail;
2768 }
2769 #endif
2770 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
2771 NULL, ipmr_rtm_dumproute, NULL);
2772 return 0;
2773
2774 #ifdef CONFIG_IP_PIMSM_V2
2775 add_proto_fail:
2776 unregister_netdevice_notifier(&ip_mr_notifier);
2777 #endif
2778 reg_notif_fail:
2779 unregister_pernet_subsys(&ipmr_net_ops);
2780 reg_pernet_fail:
2781 kmem_cache_destroy(mrt_cachep);
2782 return err;
2783 }
2784