Loading...
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/system.h>
30#include <asm/uaccess.h>
31#include <linux/types.h>
32#include <linux/capability.h>
33#include <linux/errno.h>
34#include <linux/timer.h>
35#include <linux/mm.h>
36#include <linux/kernel.h>
37#include <linux/fcntl.h>
38#include <linux/stat.h>
39#include <linux/socket.h>
40#include <linux/in.h>
41#include <linux/inet.h>
42#include <linux/netdevice.h>
43#include <linux/inetdevice.h>
44#include <linux/igmp.h>
45#include <linux/proc_fs.h>
46#include <linux/seq_file.h>
47#include <linux/mroute.h>
48#include <linux/init.h>
49#include <linux/if_ether.h>
50#include <linux/slab.h>
51#include <net/net_namespace.h>
52#include <net/ip.h>
53#include <net/protocol.h>
54#include <linux/skbuff.h>
55#include <net/route.h>
56#include <net/sock.h>
57#include <net/icmp.h>
58#include <net/udp.h>
59#include <net/raw.h>
60#include <linux/notifier.h>
61#include <linux/if_arp.h>
62#include <linux/netfilter_ipv4.h>
63#include <linux/compat.h>
64#include <net/ipip.h>
65#include <net/checksum.h>
66#include <net/netlink.h>
67#include <net/fib_rules.h>
68
69#if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
70#define CONFIG_IP_PIMSM 1
71#endif
72
73struct mr_table {
74 struct list_head list;
75#ifdef CONFIG_NET_NS
76 struct net *net;
77#endif
78 u32 id;
79 struct sock __rcu *mroute_sk;
80 struct timer_list ipmr_expire_timer;
81 struct list_head mfc_unres_queue;
82 struct list_head mfc_cache_array[MFC_LINES];
83 struct vif_device vif_table[MAXVIFS];
84 int maxvif;
85 atomic_t cache_resolve_queue_len;
86 int mroute_do_assert;
87 int mroute_do_pim;
88#if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
89 int mroute_reg_vif_num;
90#endif
91};
92
93struct ipmr_rule {
94 struct fib_rule common;
95};
96
97struct ipmr_result {
98 struct mr_table *mrt;
99};
100
101/* Big lock, protecting vif table, mrt cache and mroute socket state.
102 * Note that the changes are semaphored via rtnl_lock.
103 */
104
105static DEFINE_RWLOCK(mrt_lock);
106
107/*
108 * Multicast router control variables
109 */
110
111#define VIF_EXISTS(_mrt, _idx) ((_mrt)->vif_table[_idx].dev != NULL)
112
113/* Special spinlock for queue of unresolved entries */
114static DEFINE_SPINLOCK(mfc_unres_lock);
115
116/* We return to original Alan's scheme. Hash table of resolved
117 * entries is changed only in process context and protected
118 * with weak lock mrt_lock. Queue of unresolved entries is protected
119 * with strong spinlock mfc_unres_lock.
120 *
121 * In this case data path is free of exclusive locks at all.
122 */
123
124static struct kmem_cache *mrt_cachep __read_mostly;
125
126static struct mr_table *ipmr_new_table(struct net *net, u32 id);
127static int ip_mr_forward(struct net *net, struct mr_table *mrt,
128 struct sk_buff *skb, struct mfc_cache *cache,
129 int local);
130static int ipmr_cache_report(struct mr_table *mrt,
131 struct sk_buff *pkt, vifi_t vifi, int assert);
132static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
133 struct mfc_cache *c, struct rtmsg *rtm);
134static void ipmr_expire_process(unsigned long arg);
135
136#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
137#define ipmr_for_each_table(mrt, net) \
138 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
139
140static struct mr_table *ipmr_get_table(struct net *net, u32 id)
141{
142 struct mr_table *mrt;
143
144 ipmr_for_each_table(mrt, net) {
145 if (mrt->id == id)
146 return mrt;
147 }
148 return NULL;
149}
150
151static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
152 struct mr_table **mrt)
153{
154 struct ipmr_result res;
155 struct fib_lookup_arg arg = { .result = &res, };
156 int err;
157
158 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
159 flowi4_to_flowi(flp4), 0, &arg);
160 if (err < 0)
161 return err;
162 *mrt = res.mrt;
163 return 0;
164}
165
166static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
167 int flags, struct fib_lookup_arg *arg)
168{
169 struct ipmr_result *res = arg->result;
170 struct mr_table *mrt;
171
172 switch (rule->action) {
173 case FR_ACT_TO_TBL:
174 break;
175 case FR_ACT_UNREACHABLE:
176 return -ENETUNREACH;
177 case FR_ACT_PROHIBIT:
178 return -EACCES;
179 case FR_ACT_BLACKHOLE:
180 default:
181 return -EINVAL;
182 }
183
184 mrt = ipmr_get_table(rule->fr_net, rule->table);
185 if (mrt == NULL)
186 return -EAGAIN;
187 res->mrt = mrt;
188 return 0;
189}
190
191static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
192{
193 return 1;
194}
195
196static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
197 FRA_GENERIC_POLICY,
198};
199
200static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
201 struct fib_rule_hdr *frh, struct nlattr **tb)
202{
203 return 0;
204}
205
206static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
207 struct nlattr **tb)
208{
209 return 1;
210}
211
212static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
213 struct fib_rule_hdr *frh)
214{
215 frh->dst_len = 0;
216 frh->src_len = 0;
217 frh->tos = 0;
218 return 0;
219}
220
221static const struct fib_rules_ops __net_initdata ipmr_rules_ops_template = {
222 .family = RTNL_FAMILY_IPMR,
223 .rule_size = sizeof(struct ipmr_rule),
224 .addr_size = sizeof(u32),
225 .action = ipmr_rule_action,
226 .match = ipmr_rule_match,
227 .configure = ipmr_rule_configure,
228 .compare = ipmr_rule_compare,
229 .default_pref = fib_default_rule_pref,
230 .fill = ipmr_rule_fill,
231 .nlgroup = RTNLGRP_IPV4_RULE,
232 .policy = ipmr_rule_policy,
233 .owner = THIS_MODULE,
234};
235
236static int __net_init ipmr_rules_init(struct net *net)
237{
238 struct fib_rules_ops *ops;
239 struct mr_table *mrt;
240 int err;
241
242 ops = fib_rules_register(&ipmr_rules_ops_template, net);
243 if (IS_ERR(ops))
244 return PTR_ERR(ops);
245
246 INIT_LIST_HEAD(&net->ipv4.mr_tables);
247
248 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
249 if (mrt == NULL) {
250 err = -ENOMEM;
251 goto err1;
252 }
253
254 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
255 if (err < 0)
256 goto err2;
257
258 net->ipv4.mr_rules_ops = ops;
259 return 0;
260
261err2:
262 kfree(mrt);
263err1:
264 fib_rules_unregister(ops);
265 return err;
266}
267
268static void __net_exit ipmr_rules_exit(struct net *net)
269{
270 struct mr_table *mrt, *next;
271
272 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
273 list_del(&mrt->list);
274 kfree(mrt);
275 }
276 fib_rules_unregister(net->ipv4.mr_rules_ops);
277}
278#else
279#define ipmr_for_each_table(mrt, net) \
280 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
281
282static struct mr_table *ipmr_get_table(struct net *net, u32 id)
283{
284 return net->ipv4.mrt;
285}
286
287static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
288 struct mr_table **mrt)
289{
290 *mrt = net->ipv4.mrt;
291 return 0;
292}
293
294static int __net_init ipmr_rules_init(struct net *net)
295{
296 net->ipv4.mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
297 return net->ipv4.mrt ? 0 : -ENOMEM;
298}
299
300static void __net_exit ipmr_rules_exit(struct net *net)
301{
302 kfree(net->ipv4.mrt);
303}
304#endif
305
306static struct mr_table *ipmr_new_table(struct net *net, u32 id)
307{
308 struct mr_table *mrt;
309 unsigned int i;
310
311 mrt = ipmr_get_table(net, id);
312 if (mrt != NULL)
313 return mrt;
314
315 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
316 if (mrt == NULL)
317 return NULL;
318 write_pnet(&mrt->net, net);
319 mrt->id = id;
320
321 /* Forwarding cache */
322 for (i = 0; i < MFC_LINES; i++)
323 INIT_LIST_HEAD(&mrt->mfc_cache_array[i]);
324
325 INIT_LIST_HEAD(&mrt->mfc_unres_queue);
326
327 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
328 (unsigned long)mrt);
329
330#ifdef CONFIG_IP_PIMSM
331 mrt->mroute_reg_vif_num = -1;
332#endif
333#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
334 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
335#endif
336 return mrt;
337}
338
339/* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
340
341static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
342{
343 struct net *net = dev_net(dev);
344
345 dev_close(dev);
346
347 dev = __dev_get_by_name(net, "tunl0");
348 if (dev) {
349 const struct net_device_ops *ops = dev->netdev_ops;
350 struct ifreq ifr;
351 struct ip_tunnel_parm p;
352
353 memset(&p, 0, sizeof(p));
354 p.iph.daddr = v->vifc_rmt_addr.s_addr;
355 p.iph.saddr = v->vifc_lcl_addr.s_addr;
356 p.iph.version = 4;
357 p.iph.ihl = 5;
358 p.iph.protocol = IPPROTO_IPIP;
359 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
360 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
361
362 if (ops->ndo_do_ioctl) {
363 mm_segment_t oldfs = get_fs();
364
365 set_fs(KERNEL_DS);
366 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
367 set_fs(oldfs);
368 }
369 }
370}
371
372static
373struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
374{
375 struct net_device *dev;
376
377 dev = __dev_get_by_name(net, "tunl0");
378
379 if (dev) {
380 const struct net_device_ops *ops = dev->netdev_ops;
381 int err;
382 struct ifreq ifr;
383 struct ip_tunnel_parm p;
384 struct in_device *in_dev;
385
386 memset(&p, 0, sizeof(p));
387 p.iph.daddr = v->vifc_rmt_addr.s_addr;
388 p.iph.saddr = v->vifc_lcl_addr.s_addr;
389 p.iph.version = 4;
390 p.iph.ihl = 5;
391 p.iph.protocol = IPPROTO_IPIP;
392 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
393 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
394
395 if (ops->ndo_do_ioctl) {
396 mm_segment_t oldfs = get_fs();
397
398 set_fs(KERNEL_DS);
399 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
400 set_fs(oldfs);
401 } else {
402 err = -EOPNOTSUPP;
403 }
404 dev = NULL;
405
406 if (err == 0 &&
407 (dev = __dev_get_by_name(net, p.name)) != NULL) {
408 dev->flags |= IFF_MULTICAST;
409
410 in_dev = __in_dev_get_rtnl(dev);
411 if (in_dev == NULL)
412 goto failure;
413
414 ipv4_devconf_setall(in_dev);
415 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
416
417 if (dev_open(dev))
418 goto failure;
419 dev_hold(dev);
420 }
421 }
422 return dev;
423
424failure:
425 /* allow the register to be completed before unregistering. */
426 rtnl_unlock();
427 rtnl_lock();
428
429 unregister_netdevice(dev);
430 return NULL;
431}
432
433#ifdef CONFIG_IP_PIMSM
434
435static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
436{
437 struct net *net = dev_net(dev);
438 struct mr_table *mrt;
439 struct flowi4 fl4 = {
440 .flowi4_oif = dev->ifindex,
441 .flowi4_iif = skb->skb_iif,
442 .flowi4_mark = skb->mark,
443 };
444 int err;
445
446 err = ipmr_fib_lookup(net, &fl4, &mrt);
447 if (err < 0) {
448 kfree_skb(skb);
449 return err;
450 }
451
452 read_lock(&mrt_lock);
453 dev->stats.tx_bytes += skb->len;
454 dev->stats.tx_packets++;
455 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
456 read_unlock(&mrt_lock);
457 kfree_skb(skb);
458 return NETDEV_TX_OK;
459}
460
461static const struct net_device_ops reg_vif_netdev_ops = {
462 .ndo_start_xmit = reg_vif_xmit,
463};
464
465static void reg_vif_setup(struct net_device *dev)
466{
467 dev->type = ARPHRD_PIMREG;
468 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
469 dev->flags = IFF_NOARP;
470 dev->netdev_ops = ®_vif_netdev_ops,
471 dev->destructor = free_netdev;
472 dev->features |= NETIF_F_NETNS_LOCAL;
473}
474
475static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
476{
477 struct net_device *dev;
478 struct in_device *in_dev;
479 char name[IFNAMSIZ];
480
481 if (mrt->id == RT_TABLE_DEFAULT)
482 sprintf(name, "pimreg");
483 else
484 sprintf(name, "pimreg%u", mrt->id);
485
486 dev = alloc_netdev(0, name, reg_vif_setup);
487
488 if (dev == NULL)
489 return NULL;
490
491 dev_net_set(dev, net);
492
493 if (register_netdevice(dev)) {
494 free_netdev(dev);
495 return NULL;
496 }
497 dev->iflink = 0;
498
499 rcu_read_lock();
500 in_dev = __in_dev_get_rcu(dev);
501 if (!in_dev) {
502 rcu_read_unlock();
503 goto failure;
504 }
505
506 ipv4_devconf_setall(in_dev);
507 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
508 rcu_read_unlock();
509
510 if (dev_open(dev))
511 goto failure;
512
513 dev_hold(dev);
514
515 return dev;
516
517failure:
518 /* allow the register to be completed before unregistering. */
519 rtnl_unlock();
520 rtnl_lock();
521
522 unregister_netdevice(dev);
523 return NULL;
524}
525#endif
526
527/*
528 * Delete a VIF entry
529 * @notify: Set to 1, if the caller is a notifier_call
530 */
531
532static int vif_delete(struct mr_table *mrt, int vifi, int notify,
533 struct list_head *head)
534{
535 struct vif_device *v;
536 struct net_device *dev;
537 struct in_device *in_dev;
538
539 if (vifi < 0 || vifi >= mrt->maxvif)
540 return -EADDRNOTAVAIL;
541
542 v = &mrt->vif_table[vifi];
543
544 write_lock_bh(&mrt_lock);
545 dev = v->dev;
546 v->dev = NULL;
547
548 if (!dev) {
549 write_unlock_bh(&mrt_lock);
550 return -EADDRNOTAVAIL;
551 }
552
553#ifdef CONFIG_IP_PIMSM
554 if (vifi == mrt->mroute_reg_vif_num)
555 mrt->mroute_reg_vif_num = -1;
556#endif
557
558 if (vifi + 1 == mrt->maxvif) {
559 int tmp;
560
561 for (tmp = vifi - 1; tmp >= 0; tmp--) {
562 if (VIF_EXISTS(mrt, tmp))
563 break;
564 }
565 mrt->maxvif = tmp+1;
566 }
567
568 write_unlock_bh(&mrt_lock);
569
570 dev_set_allmulti(dev, -1);
571
572 in_dev = __in_dev_get_rtnl(dev);
573 if (in_dev) {
574 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
575 ip_rt_multicast_event(in_dev);
576 }
577
578 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
579 unregister_netdevice_queue(dev, head);
580
581 dev_put(dev);
582 return 0;
583}
584
585static void ipmr_cache_free_rcu(struct rcu_head *head)
586{
587 struct mfc_cache *c = container_of(head, struct mfc_cache, rcu);
588
589 kmem_cache_free(mrt_cachep, c);
590}
591
592static inline void ipmr_cache_free(struct mfc_cache *c)
593{
594 call_rcu(&c->rcu, ipmr_cache_free_rcu);
595}
596
597/* Destroy an unresolved cache entry, killing queued skbs
598 * and reporting error to netlink readers.
599 */
600
601static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
602{
603 struct net *net = read_pnet(&mrt->net);
604 struct sk_buff *skb;
605 struct nlmsgerr *e;
606
607 atomic_dec(&mrt->cache_resolve_queue_len);
608
609 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) {
610 if (ip_hdr(skb)->version == 0) {
611 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
612 nlh->nlmsg_type = NLMSG_ERROR;
613 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
614 skb_trim(skb, nlh->nlmsg_len);
615 e = NLMSG_DATA(nlh);
616 e->error = -ETIMEDOUT;
617 memset(&e->msg, 0, sizeof(e->msg));
618
619 rtnl_unicast(skb, net, NETLINK_CB(skb).pid);
620 } else {
621 kfree_skb(skb);
622 }
623 }
624
625 ipmr_cache_free(c);
626}
627
628
629/* Timer process for the unresolved queue. */
630
631static void ipmr_expire_process(unsigned long arg)
632{
633 struct mr_table *mrt = (struct mr_table *)arg;
634 unsigned long now;
635 unsigned long expires;
636 struct mfc_cache *c, *next;
637
638 if (!spin_trylock(&mfc_unres_lock)) {
639 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
640 return;
641 }
642
643 if (list_empty(&mrt->mfc_unres_queue))
644 goto out;
645
646 now = jiffies;
647 expires = 10*HZ;
648
649 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
650 if (time_after(c->mfc_un.unres.expires, now)) {
651 unsigned long interval = c->mfc_un.unres.expires - now;
652 if (interval < expires)
653 expires = interval;
654 continue;
655 }
656
657 list_del(&c->list);
658 ipmr_destroy_unres(mrt, c);
659 }
660
661 if (!list_empty(&mrt->mfc_unres_queue))
662 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
663
664out:
665 spin_unlock(&mfc_unres_lock);
666}
667
668/* Fill oifs list. It is called under write locked mrt_lock. */
669
670static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache,
671 unsigned char *ttls)
672{
673 int vifi;
674
675 cache->mfc_un.res.minvif = MAXVIFS;
676 cache->mfc_un.res.maxvif = 0;
677 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
678
679 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
680 if (VIF_EXISTS(mrt, vifi) &&
681 ttls[vifi] && ttls[vifi] < 255) {
682 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
683 if (cache->mfc_un.res.minvif > vifi)
684 cache->mfc_un.res.minvif = vifi;
685 if (cache->mfc_un.res.maxvif <= vifi)
686 cache->mfc_un.res.maxvif = vifi + 1;
687 }
688 }
689}
690
691static int vif_add(struct net *net, struct mr_table *mrt,
692 struct vifctl *vifc, int mrtsock)
693{
694 int vifi = vifc->vifc_vifi;
695 struct vif_device *v = &mrt->vif_table[vifi];
696 struct net_device *dev;
697 struct in_device *in_dev;
698 int err;
699
700 /* Is vif busy ? */
701 if (VIF_EXISTS(mrt, vifi))
702 return -EADDRINUSE;
703
704 switch (vifc->vifc_flags) {
705#ifdef CONFIG_IP_PIMSM
706 case VIFF_REGISTER:
707 /*
708 * Special Purpose VIF in PIM
709 * All the packets will be sent to the daemon
710 */
711 if (mrt->mroute_reg_vif_num >= 0)
712 return -EADDRINUSE;
713 dev = ipmr_reg_vif(net, mrt);
714 if (!dev)
715 return -ENOBUFS;
716 err = dev_set_allmulti(dev, 1);
717 if (err) {
718 unregister_netdevice(dev);
719 dev_put(dev);
720 return err;
721 }
722 break;
723#endif
724 case VIFF_TUNNEL:
725 dev = ipmr_new_tunnel(net, vifc);
726 if (!dev)
727 return -ENOBUFS;
728 err = dev_set_allmulti(dev, 1);
729 if (err) {
730 ipmr_del_tunnel(dev, vifc);
731 dev_put(dev);
732 return err;
733 }
734 break;
735
736 case VIFF_USE_IFINDEX:
737 case 0:
738 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
739 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
740 if (dev && __in_dev_get_rtnl(dev) == NULL) {
741 dev_put(dev);
742 return -EADDRNOTAVAIL;
743 }
744 } else {
745 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
746 }
747 if (!dev)
748 return -EADDRNOTAVAIL;
749 err = dev_set_allmulti(dev, 1);
750 if (err) {
751 dev_put(dev);
752 return err;
753 }
754 break;
755 default:
756 return -EINVAL;
757 }
758
759 in_dev = __in_dev_get_rtnl(dev);
760 if (!in_dev) {
761 dev_put(dev);
762 return -EADDRNOTAVAIL;
763 }
764 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
765 ip_rt_multicast_event(in_dev);
766
767 /* Fill in the VIF structures */
768
769 v->rate_limit = vifc->vifc_rate_limit;
770 v->local = vifc->vifc_lcl_addr.s_addr;
771 v->remote = vifc->vifc_rmt_addr.s_addr;
772 v->flags = vifc->vifc_flags;
773 if (!mrtsock)
774 v->flags |= VIFF_STATIC;
775 v->threshold = vifc->vifc_threshold;
776 v->bytes_in = 0;
777 v->bytes_out = 0;
778 v->pkt_in = 0;
779 v->pkt_out = 0;
780 v->link = dev->ifindex;
781 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER))
782 v->link = dev->iflink;
783
784 /* And finish update writing critical data */
785 write_lock_bh(&mrt_lock);
786 v->dev = dev;
787#ifdef CONFIG_IP_PIMSM
788 if (v->flags & VIFF_REGISTER)
789 mrt->mroute_reg_vif_num = vifi;
790#endif
791 if (vifi+1 > mrt->maxvif)
792 mrt->maxvif = vifi+1;
793 write_unlock_bh(&mrt_lock);
794 return 0;
795}
796
797/* called with rcu_read_lock() */
798static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
799 __be32 origin,
800 __be32 mcastgrp)
801{
802 int line = MFC_HASH(mcastgrp, origin);
803 struct mfc_cache *c;
804
805 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) {
806 if (c->mfc_origin == origin && c->mfc_mcastgrp == mcastgrp)
807 return c;
808 }
809 return NULL;
810}
811
812/*
813 * Allocate a multicast cache entry
814 */
815static struct mfc_cache *ipmr_cache_alloc(void)
816{
817 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
818
819 if (c)
820 c->mfc_un.res.minvif = MAXVIFS;
821 return c;
822}
823
824static struct mfc_cache *ipmr_cache_alloc_unres(void)
825{
826 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
827
828 if (c) {
829 skb_queue_head_init(&c->mfc_un.unres.unresolved);
830 c->mfc_un.unres.expires = jiffies + 10*HZ;
831 }
832 return c;
833}
834
835/*
836 * A cache entry has gone into a resolved state from queued
837 */
838
839static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
840 struct mfc_cache *uc, struct mfc_cache *c)
841{
842 struct sk_buff *skb;
843 struct nlmsgerr *e;
844
845 /* Play the pending entries through our router */
846
847 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
848 if (ip_hdr(skb)->version == 0) {
849 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
850
851 if (__ipmr_fill_mroute(mrt, skb, c, NLMSG_DATA(nlh)) > 0) {
852 nlh->nlmsg_len = skb_tail_pointer(skb) -
853 (u8 *)nlh;
854 } else {
855 nlh->nlmsg_type = NLMSG_ERROR;
856 nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
857 skb_trim(skb, nlh->nlmsg_len);
858 e = NLMSG_DATA(nlh);
859 e->error = -EMSGSIZE;
860 memset(&e->msg, 0, sizeof(e->msg));
861 }
862
863 rtnl_unicast(skb, net, NETLINK_CB(skb).pid);
864 } else {
865 ip_mr_forward(net, mrt, skb, c, 0);
866 }
867 }
868}
869
870/*
871 * Bounce a cache query up to mrouted. We could use netlink for this but mrouted
872 * expects the following bizarre scheme.
873 *
874 * Called under mrt_lock.
875 */
876
877static int ipmr_cache_report(struct mr_table *mrt,
878 struct sk_buff *pkt, vifi_t vifi, int assert)
879{
880 struct sk_buff *skb;
881 const int ihl = ip_hdrlen(pkt);
882 struct igmphdr *igmp;
883 struct igmpmsg *msg;
884 struct sock *mroute_sk;
885 int ret;
886
887#ifdef CONFIG_IP_PIMSM
888 if (assert == IGMPMSG_WHOLEPKT)
889 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
890 else
891#endif
892 skb = alloc_skb(128, GFP_ATOMIC);
893
894 if (!skb)
895 return -ENOBUFS;
896
897#ifdef CONFIG_IP_PIMSM
898 if (assert == IGMPMSG_WHOLEPKT) {
899 /* Ugly, but we have no choice with this interface.
900 * Duplicate old header, fix ihl, length etc.
901 * And all this only to mangle msg->im_msgtype and
902 * to set msg->im_mbz to "mbz" :-)
903 */
904 skb_push(skb, sizeof(struct iphdr));
905 skb_reset_network_header(skb);
906 skb_reset_transport_header(skb);
907 msg = (struct igmpmsg *)skb_network_header(skb);
908 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
909 msg->im_msgtype = IGMPMSG_WHOLEPKT;
910 msg->im_mbz = 0;
911 msg->im_vif = mrt->mroute_reg_vif_num;
912 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
913 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
914 sizeof(struct iphdr));
915 } else
916#endif
917 {
918
919 /* Copy the IP header */
920
921 skb->network_header = skb->tail;
922 skb_put(skb, ihl);
923 skb_copy_to_linear_data(skb, pkt->data, ihl);
924 ip_hdr(skb)->protocol = 0; /* Flag to the kernel this is a route add */
925 msg = (struct igmpmsg *)skb_network_header(skb);
926 msg->im_vif = vifi;
927 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
928
929 /* Add our header */
930
931 igmp = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
932 igmp->type =
933 msg->im_msgtype = assert;
934 igmp->code = 0;
935 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
936 skb->transport_header = skb->network_header;
937 }
938
939 rcu_read_lock();
940 mroute_sk = rcu_dereference(mrt->mroute_sk);
941 if (mroute_sk == NULL) {
942 rcu_read_unlock();
943 kfree_skb(skb);
944 return -EINVAL;
945 }
946
947 /* Deliver to mrouted */
948
949 ret = sock_queue_rcv_skb(mroute_sk, skb);
950 rcu_read_unlock();
951 if (ret < 0) {
952 if (net_ratelimit())
953 printk(KERN_WARNING "mroute: pending queue full, dropping entries.\n");
954 kfree_skb(skb);
955 }
956
957 return ret;
958}
959
960/*
961 * Queue a packet for resolution. It gets locked cache entry!
962 */
963
964static int
965ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, struct sk_buff *skb)
966{
967 bool found = false;
968 int err;
969 struct mfc_cache *c;
970 const struct iphdr *iph = ip_hdr(skb);
971
972 spin_lock_bh(&mfc_unres_lock);
973 list_for_each_entry(c, &mrt->mfc_unres_queue, list) {
974 if (c->mfc_mcastgrp == iph->daddr &&
975 c->mfc_origin == iph->saddr) {
976 found = true;
977 break;
978 }
979 }
980
981 if (!found) {
982 /* Create a new entry if allowable */
983
984 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
985 (c = ipmr_cache_alloc_unres()) == NULL) {
986 spin_unlock_bh(&mfc_unres_lock);
987
988 kfree_skb(skb);
989 return -ENOBUFS;
990 }
991
992 /* Fill in the new cache entry */
993
994 c->mfc_parent = -1;
995 c->mfc_origin = iph->saddr;
996 c->mfc_mcastgrp = iph->daddr;
997
998 /* Reflect first query at mrouted. */
999
1000 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1001 if (err < 0) {
1002 /* If the report failed throw the cache entry
1003 out - Brad Parker
1004 */
1005 spin_unlock_bh(&mfc_unres_lock);
1006
1007 ipmr_cache_free(c);
1008 kfree_skb(skb);
1009 return err;
1010 }
1011
1012 atomic_inc(&mrt->cache_resolve_queue_len);
1013 list_add(&c->list, &mrt->mfc_unres_queue);
1014
1015 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1016 mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires);
1017 }
1018
1019 /* See if we can append the packet */
1020
1021 if (c->mfc_un.unres.unresolved.qlen > 3) {
1022 kfree_skb(skb);
1023 err = -ENOBUFS;
1024 } else {
1025 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1026 err = 0;
1027 }
1028
1029 spin_unlock_bh(&mfc_unres_lock);
1030 return err;
1031}
1032
1033/*
1034 * MFC cache manipulation by user space mroute daemon
1035 */
1036
1037static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc)
1038{
1039 int line;
1040 struct mfc_cache *c, *next;
1041
1042 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1043
1044 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[line], list) {
1045 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1046 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr) {
1047 list_del_rcu(&c->list);
1048
1049 ipmr_cache_free(c);
1050 return 0;
1051 }
1052 }
1053 return -ENOENT;
1054}
1055
1056static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1057 struct mfcctl *mfc, int mrtsock)
1058{
1059 bool found = false;
1060 int line;
1061 struct mfc_cache *uc, *c;
1062
1063 if (mfc->mfcc_parent >= MAXVIFS)
1064 return -ENFILE;
1065
1066 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1067
1068 list_for_each_entry(c, &mrt->mfc_cache_array[line], list) {
1069 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1070 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr) {
1071 found = true;
1072 break;
1073 }
1074 }
1075
1076 if (found) {
1077 write_lock_bh(&mrt_lock);
1078 c->mfc_parent = mfc->mfcc_parent;
1079 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1080 if (!mrtsock)
1081 c->mfc_flags |= MFC_STATIC;
1082 write_unlock_bh(&mrt_lock);
1083 return 0;
1084 }
1085
1086 if (!ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1087 return -EINVAL;
1088
1089 c = ipmr_cache_alloc();
1090 if (c == NULL)
1091 return -ENOMEM;
1092
1093 c->mfc_origin = mfc->mfcc_origin.s_addr;
1094 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1095 c->mfc_parent = mfc->mfcc_parent;
1096 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1097 if (!mrtsock)
1098 c->mfc_flags |= MFC_STATIC;
1099
1100 list_add_rcu(&c->list, &mrt->mfc_cache_array[line]);
1101
1102 /*
1103 * Check to see if we resolved a queued list. If so we
1104 * need to send on the frames and tidy up.
1105 */
1106 found = false;
1107 spin_lock_bh(&mfc_unres_lock);
1108 list_for_each_entry(uc, &mrt->mfc_unres_queue, list) {
1109 if (uc->mfc_origin == c->mfc_origin &&
1110 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1111 list_del(&uc->list);
1112 atomic_dec(&mrt->cache_resolve_queue_len);
1113 found = true;
1114 break;
1115 }
1116 }
1117 if (list_empty(&mrt->mfc_unres_queue))
1118 del_timer(&mrt->ipmr_expire_timer);
1119 spin_unlock_bh(&mfc_unres_lock);
1120
1121 if (found) {
1122 ipmr_cache_resolve(net, mrt, uc, c);
1123 ipmr_cache_free(uc);
1124 }
1125 return 0;
1126}
1127
1128/*
1129 * Close the multicast socket, and clear the vif tables etc
1130 */
1131
1132static void mroute_clean_tables(struct mr_table *mrt)
1133{
1134 int i;
1135 LIST_HEAD(list);
1136 struct mfc_cache *c, *next;
1137
1138 /* Shut down all active vif entries */
1139
1140 for (i = 0; i < mrt->maxvif; i++) {
1141 if (!(mrt->vif_table[i].flags & VIFF_STATIC))
1142 vif_delete(mrt, i, 0, &list);
1143 }
1144 unregister_netdevice_many(&list);
1145
1146 /* Wipe the cache */
1147
1148 for (i = 0; i < MFC_LINES; i++) {
1149 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[i], list) {
1150 if (c->mfc_flags & MFC_STATIC)
1151 continue;
1152 list_del_rcu(&c->list);
1153 ipmr_cache_free(c);
1154 }
1155 }
1156
1157 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1158 spin_lock_bh(&mfc_unres_lock);
1159 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
1160 list_del(&c->list);
1161 ipmr_destroy_unres(mrt, c);
1162 }
1163 spin_unlock_bh(&mfc_unres_lock);
1164 }
1165}
1166
1167/* called from ip_ra_control(), before an RCU grace period,
1168 * we dont need to call synchronize_rcu() here
1169 */
1170static void mrtsock_destruct(struct sock *sk)
1171{
1172 struct net *net = sock_net(sk);
1173 struct mr_table *mrt;
1174
1175 rtnl_lock();
1176 ipmr_for_each_table(mrt, net) {
1177 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1178 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1179 rcu_assign_pointer(mrt->mroute_sk, NULL);
1180 mroute_clean_tables(mrt);
1181 }
1182 }
1183 rtnl_unlock();
1184}
1185
1186/*
1187 * Socket options and virtual interface manipulation. The whole
1188 * virtual interface system is a complete heap, but unfortunately
1189 * that's how BSD mrouted happens to think. Maybe one day with a proper
1190 * MOSPF/PIM router set up we can clean this up.
1191 */
1192
1193int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1194{
1195 int ret;
1196 struct vifctl vif;
1197 struct mfcctl mfc;
1198 struct net *net = sock_net(sk);
1199 struct mr_table *mrt;
1200
1201 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1202 if (mrt == NULL)
1203 return -ENOENT;
1204
1205 if (optname != MRT_INIT) {
1206 if (sk != rcu_dereference_raw(mrt->mroute_sk) &&
1207 !capable(CAP_NET_ADMIN))
1208 return -EACCES;
1209 }
1210
1211 switch (optname) {
1212 case MRT_INIT:
1213 if (sk->sk_type != SOCK_RAW ||
1214 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1215 return -EOPNOTSUPP;
1216 if (optlen != sizeof(int))
1217 return -ENOPROTOOPT;
1218
1219 rtnl_lock();
1220 if (rtnl_dereference(mrt->mroute_sk)) {
1221 rtnl_unlock();
1222 return -EADDRINUSE;
1223 }
1224
1225 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1226 if (ret == 0) {
1227 rcu_assign_pointer(mrt->mroute_sk, sk);
1228 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1229 }
1230 rtnl_unlock();
1231 return ret;
1232 case MRT_DONE:
1233 if (sk != rcu_dereference_raw(mrt->mroute_sk))
1234 return -EACCES;
1235 return ip_ra_control(sk, 0, NULL);
1236 case MRT_ADD_VIF:
1237 case MRT_DEL_VIF:
1238 if (optlen != sizeof(vif))
1239 return -EINVAL;
1240 if (copy_from_user(&vif, optval, sizeof(vif)))
1241 return -EFAULT;
1242 if (vif.vifc_vifi >= MAXVIFS)
1243 return -ENFILE;
1244 rtnl_lock();
1245 if (optname == MRT_ADD_VIF) {
1246 ret = vif_add(net, mrt, &vif,
1247 sk == rtnl_dereference(mrt->mroute_sk));
1248 } else {
1249 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1250 }
1251 rtnl_unlock();
1252 return ret;
1253
1254 /*
1255 * Manipulate the forwarding caches. These live
1256 * in a sort of kernel/user symbiosis.
1257 */
1258 case MRT_ADD_MFC:
1259 case MRT_DEL_MFC:
1260 if (optlen != sizeof(mfc))
1261 return -EINVAL;
1262 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1263 return -EFAULT;
1264 rtnl_lock();
1265 if (optname == MRT_DEL_MFC)
1266 ret = ipmr_mfc_delete(mrt, &mfc);
1267 else
1268 ret = ipmr_mfc_add(net, mrt, &mfc,
1269 sk == rtnl_dereference(mrt->mroute_sk));
1270 rtnl_unlock();
1271 return ret;
1272 /*
1273 * Control PIM assert.
1274 */
1275 case MRT_ASSERT:
1276 {
1277 int v;
1278 if (get_user(v, (int __user *)optval))
1279 return -EFAULT;
1280 mrt->mroute_do_assert = (v) ? 1 : 0;
1281 return 0;
1282 }
1283#ifdef CONFIG_IP_PIMSM
1284 case MRT_PIM:
1285 {
1286 int v;
1287
1288 if (get_user(v, (int __user *)optval))
1289 return -EFAULT;
1290 v = (v) ? 1 : 0;
1291
1292 rtnl_lock();
1293 ret = 0;
1294 if (v != mrt->mroute_do_pim) {
1295 mrt->mroute_do_pim = v;
1296 mrt->mroute_do_assert = v;
1297 }
1298 rtnl_unlock();
1299 return ret;
1300 }
1301#endif
1302#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
1303 case MRT_TABLE:
1304 {
1305 u32 v;
1306
1307 if (optlen != sizeof(u32))
1308 return -EINVAL;
1309 if (get_user(v, (u32 __user *)optval))
1310 return -EFAULT;
1311
1312 rtnl_lock();
1313 ret = 0;
1314 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1315 ret = -EBUSY;
1316 } else {
1317 if (!ipmr_new_table(net, v))
1318 ret = -ENOMEM;
1319 raw_sk(sk)->ipmr_table = v;
1320 }
1321 rtnl_unlock();
1322 return ret;
1323 }
1324#endif
1325 /*
1326 * Spurious command, or MRT_VERSION which you cannot
1327 * set.
1328 */
1329 default:
1330 return -ENOPROTOOPT;
1331 }
1332}
1333
1334/*
1335 * Getsock opt support for the multicast routing system.
1336 */
1337
1338int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1339{
1340 int olr;
1341 int val;
1342 struct net *net = sock_net(sk);
1343 struct mr_table *mrt;
1344
1345 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1346 if (mrt == NULL)
1347 return -ENOENT;
1348
1349 if (optname != MRT_VERSION &&
1350#ifdef CONFIG_IP_PIMSM
1351 optname != MRT_PIM &&
1352#endif
1353 optname != MRT_ASSERT)
1354 return -ENOPROTOOPT;
1355
1356 if (get_user(olr, optlen))
1357 return -EFAULT;
1358
1359 olr = min_t(unsigned int, olr, sizeof(int));
1360 if (olr < 0)
1361 return -EINVAL;
1362
1363 if (put_user(olr, optlen))
1364 return -EFAULT;
1365 if (optname == MRT_VERSION)
1366 val = 0x0305;
1367#ifdef CONFIG_IP_PIMSM
1368 else if (optname == MRT_PIM)
1369 val = mrt->mroute_do_pim;
1370#endif
1371 else
1372 val = mrt->mroute_do_assert;
1373 if (copy_to_user(optval, &val, olr))
1374 return -EFAULT;
1375 return 0;
1376}
1377
1378/*
1379 * The IP multicast ioctl support routines.
1380 */
1381
1382int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1383{
1384 struct sioc_sg_req sr;
1385 struct sioc_vif_req vr;
1386 struct vif_device *vif;
1387 struct mfc_cache *c;
1388 struct net *net = sock_net(sk);
1389 struct mr_table *mrt;
1390
1391 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1392 if (mrt == NULL)
1393 return -ENOENT;
1394
1395 switch (cmd) {
1396 case SIOCGETVIFCNT:
1397 if (copy_from_user(&vr, arg, sizeof(vr)))
1398 return -EFAULT;
1399 if (vr.vifi >= mrt->maxvif)
1400 return -EINVAL;
1401 read_lock(&mrt_lock);
1402 vif = &mrt->vif_table[vr.vifi];
1403 if (VIF_EXISTS(mrt, vr.vifi)) {
1404 vr.icount = vif->pkt_in;
1405 vr.ocount = vif->pkt_out;
1406 vr.ibytes = vif->bytes_in;
1407 vr.obytes = vif->bytes_out;
1408 read_unlock(&mrt_lock);
1409
1410 if (copy_to_user(arg, &vr, sizeof(vr)))
1411 return -EFAULT;
1412 return 0;
1413 }
1414 read_unlock(&mrt_lock);
1415 return -EADDRNOTAVAIL;
1416 case SIOCGETSGCNT:
1417 if (copy_from_user(&sr, arg, sizeof(sr)))
1418 return -EFAULT;
1419
1420 rcu_read_lock();
1421 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1422 if (c) {
1423 sr.pktcnt = c->mfc_un.res.pkt;
1424 sr.bytecnt = c->mfc_un.res.bytes;
1425 sr.wrong_if = c->mfc_un.res.wrong_if;
1426 rcu_read_unlock();
1427
1428 if (copy_to_user(arg, &sr, sizeof(sr)))
1429 return -EFAULT;
1430 return 0;
1431 }
1432 rcu_read_unlock();
1433 return -EADDRNOTAVAIL;
1434 default:
1435 return -ENOIOCTLCMD;
1436 }
1437}
1438
1439#ifdef CONFIG_COMPAT
1440struct compat_sioc_sg_req {
1441 struct in_addr src;
1442 struct in_addr grp;
1443 compat_ulong_t pktcnt;
1444 compat_ulong_t bytecnt;
1445 compat_ulong_t wrong_if;
1446};
1447
1448struct compat_sioc_vif_req {
1449 vifi_t vifi; /* Which iface */
1450 compat_ulong_t icount;
1451 compat_ulong_t ocount;
1452 compat_ulong_t ibytes;
1453 compat_ulong_t obytes;
1454};
1455
1456int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1457{
1458 struct compat_sioc_sg_req sr;
1459 struct compat_sioc_vif_req vr;
1460 struct vif_device *vif;
1461 struct mfc_cache *c;
1462 struct net *net = sock_net(sk);
1463 struct mr_table *mrt;
1464
1465 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1466 if (mrt == NULL)
1467 return -ENOENT;
1468
1469 switch (cmd) {
1470 case SIOCGETVIFCNT:
1471 if (copy_from_user(&vr, arg, sizeof(vr)))
1472 return -EFAULT;
1473 if (vr.vifi >= mrt->maxvif)
1474 return -EINVAL;
1475 read_lock(&mrt_lock);
1476 vif = &mrt->vif_table[vr.vifi];
1477 if (VIF_EXISTS(mrt, vr.vifi)) {
1478 vr.icount = vif->pkt_in;
1479 vr.ocount = vif->pkt_out;
1480 vr.ibytes = vif->bytes_in;
1481 vr.obytes = vif->bytes_out;
1482 read_unlock(&mrt_lock);
1483
1484 if (copy_to_user(arg, &vr, sizeof(vr)))
1485 return -EFAULT;
1486 return 0;
1487 }
1488 read_unlock(&mrt_lock);
1489 return -EADDRNOTAVAIL;
1490 case SIOCGETSGCNT:
1491 if (copy_from_user(&sr, arg, sizeof(sr)))
1492 return -EFAULT;
1493
1494 rcu_read_lock();
1495 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1496 if (c) {
1497 sr.pktcnt = c->mfc_un.res.pkt;
1498 sr.bytecnt = c->mfc_un.res.bytes;
1499 sr.wrong_if = c->mfc_un.res.wrong_if;
1500 rcu_read_unlock();
1501
1502 if (copy_to_user(arg, &sr, sizeof(sr)))
1503 return -EFAULT;
1504 return 0;
1505 }
1506 rcu_read_unlock();
1507 return -EADDRNOTAVAIL;
1508 default:
1509 return -ENOIOCTLCMD;
1510 }
1511}
1512#endif
1513
1514
1515static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1516{
1517 struct net_device *dev = ptr;
1518 struct net *net = dev_net(dev);
1519 struct mr_table *mrt;
1520 struct vif_device *v;
1521 int ct;
1522 LIST_HEAD(list);
1523
1524 if (event != NETDEV_UNREGISTER)
1525 return NOTIFY_DONE;
1526
1527 ipmr_for_each_table(mrt, net) {
1528 v = &mrt->vif_table[0];
1529 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1530 if (v->dev == dev)
1531 vif_delete(mrt, ct, 1, &list);
1532 }
1533 }
1534 unregister_netdevice_many(&list);
1535 return NOTIFY_DONE;
1536}
1537
1538
1539static struct notifier_block ip_mr_notifier = {
1540 .notifier_call = ipmr_device_event,
1541};
1542
1543/*
1544 * Encapsulate a packet by attaching a valid IPIP header to it.
1545 * This avoids tunnel drivers and other mess and gives us the speed so
1546 * important for multicast video.
1547 */
1548
1549static void ip_encap(struct sk_buff *skb, __be32 saddr, __be32 daddr)
1550{
1551 struct iphdr *iph;
1552 const struct iphdr *old_iph = ip_hdr(skb);
1553
1554 skb_push(skb, sizeof(struct iphdr));
1555 skb->transport_header = skb->network_header;
1556 skb_reset_network_header(skb);
1557 iph = ip_hdr(skb);
1558
1559 iph->version = 4;
1560 iph->tos = old_iph->tos;
1561 iph->ttl = old_iph->ttl;
1562 iph->frag_off = 0;
1563 iph->daddr = daddr;
1564 iph->saddr = saddr;
1565 iph->protocol = IPPROTO_IPIP;
1566 iph->ihl = 5;
1567 iph->tot_len = htons(skb->len);
1568 ip_select_ident(iph, skb_dst(skb), NULL);
1569 ip_send_check(iph);
1570
1571 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1572 nf_reset(skb);
1573}
1574
1575static inline int ipmr_forward_finish(struct sk_buff *skb)
1576{
1577 struct ip_options *opt = &(IPCB(skb)->opt);
1578
1579 IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTFORWDATAGRAMS);
1580
1581 if (unlikely(opt->optlen))
1582 ip_forward_options(skb);
1583
1584 return dst_output(skb);
1585}
1586
1587/*
1588 * Processing handlers for ipmr_forward
1589 */
1590
1591static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1592 struct sk_buff *skb, struct mfc_cache *c, int vifi)
1593{
1594 const struct iphdr *iph = ip_hdr(skb);
1595 struct vif_device *vif = &mrt->vif_table[vifi];
1596 struct net_device *dev;
1597 struct rtable *rt;
1598 struct flowi4 fl4;
1599 int encap = 0;
1600
1601 if (vif->dev == NULL)
1602 goto out_free;
1603
1604#ifdef CONFIG_IP_PIMSM
1605 if (vif->flags & VIFF_REGISTER) {
1606 vif->pkt_out++;
1607 vif->bytes_out += skb->len;
1608 vif->dev->stats.tx_bytes += skb->len;
1609 vif->dev->stats.tx_packets++;
1610 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1611 goto out_free;
1612 }
1613#endif
1614
1615 if (vif->flags & VIFF_TUNNEL) {
1616 rt = ip_route_output_ports(net, &fl4, NULL,
1617 vif->remote, vif->local,
1618 0, 0,
1619 IPPROTO_IPIP,
1620 RT_TOS(iph->tos), vif->link);
1621 if (IS_ERR(rt))
1622 goto out_free;
1623 encap = sizeof(struct iphdr);
1624 } else {
1625 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1626 0, 0,
1627 IPPROTO_IPIP,
1628 RT_TOS(iph->tos), vif->link);
1629 if (IS_ERR(rt))
1630 goto out_free;
1631 }
1632
1633 dev = rt->dst.dev;
1634
1635 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1636 /* Do not fragment multicasts. Alas, IPv4 does not
1637 * allow to send ICMP, so that packets will disappear
1638 * to blackhole.
1639 */
1640
1641 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
1642 ip_rt_put(rt);
1643 goto out_free;
1644 }
1645
1646 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1647
1648 if (skb_cow(skb, encap)) {
1649 ip_rt_put(rt);
1650 goto out_free;
1651 }
1652
1653 vif->pkt_out++;
1654 vif->bytes_out += skb->len;
1655
1656 skb_dst_drop(skb);
1657 skb_dst_set(skb, &rt->dst);
1658 ip_decrease_ttl(ip_hdr(skb));
1659
1660 /* FIXME: forward and output firewalls used to be called here.
1661 * What do we do with netfilter? -- RR
1662 */
1663 if (vif->flags & VIFF_TUNNEL) {
1664 ip_encap(skb, vif->local, vif->remote);
1665 /* FIXME: extra output firewall step used to be here. --RR */
1666 vif->dev->stats.tx_packets++;
1667 vif->dev->stats.tx_bytes += skb->len;
1668 }
1669
1670 IPCB(skb)->flags |= IPSKB_FORWARDED;
1671
1672 /*
1673 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1674 * not only before forwarding, but after forwarding on all output
1675 * interfaces. It is clear, if mrouter runs a multicasting
1676 * program, it should receive packets not depending to what interface
1677 * program is joined.
1678 * If we will not make it, the program will have to join on all
1679 * interfaces. On the other hand, multihoming host (or router, but
1680 * not mrouter) cannot join to more than one interface - it will
1681 * result in receiving multiple packets.
1682 */
1683 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, skb, skb->dev, dev,
1684 ipmr_forward_finish);
1685 return;
1686
1687out_free:
1688 kfree_skb(skb);
1689}
1690
1691static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1692{
1693 int ct;
1694
1695 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1696 if (mrt->vif_table[ct].dev == dev)
1697 break;
1698 }
1699 return ct;
1700}
1701
1702/* "local" means that we should preserve one skb (for local delivery) */
1703
1704static int ip_mr_forward(struct net *net, struct mr_table *mrt,
1705 struct sk_buff *skb, struct mfc_cache *cache,
1706 int local)
1707{
1708 int psend = -1;
1709 int vif, ct;
1710
1711 vif = cache->mfc_parent;
1712 cache->mfc_un.res.pkt++;
1713 cache->mfc_un.res.bytes += skb->len;
1714
1715 /*
1716 * Wrong interface: drop packet and (maybe) send PIM assert.
1717 */
1718 if (mrt->vif_table[vif].dev != skb->dev) {
1719 int true_vifi;
1720
1721 if (rt_is_output_route(skb_rtable(skb))) {
1722 /* It is our own packet, looped back.
1723 * Very complicated situation...
1724 *
1725 * The best workaround until routing daemons will be
1726 * fixed is not to redistribute packet, if it was
1727 * send through wrong interface. It means, that
1728 * multicast applications WILL NOT work for
1729 * (S,G), which have default multicast route pointing
1730 * to wrong oif. In any case, it is not a good
1731 * idea to use multicasting applications on router.
1732 */
1733 goto dont_forward;
1734 }
1735
1736 cache->mfc_un.res.wrong_if++;
1737 true_vifi = ipmr_find_vif(mrt, skb->dev);
1738
1739 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1740 /* pimsm uses asserts, when switching from RPT to SPT,
1741 * so that we cannot check that packet arrived on an oif.
1742 * It is bad, but otherwise we would need to move pretty
1743 * large chunk of pimd to kernel. Ough... --ANK
1744 */
1745 (mrt->mroute_do_pim ||
1746 cache->mfc_un.res.ttls[true_vifi] < 255) &&
1747 time_after(jiffies,
1748 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1749 cache->mfc_un.res.last_assert = jiffies;
1750 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1751 }
1752 goto dont_forward;
1753 }
1754
1755 mrt->vif_table[vif].pkt_in++;
1756 mrt->vif_table[vif].bytes_in += skb->len;
1757
1758 /*
1759 * Forward the frame
1760 */
1761 for (ct = cache->mfc_un.res.maxvif - 1;
1762 ct >= cache->mfc_un.res.minvif; ct--) {
1763 if (ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
1764 if (psend != -1) {
1765 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1766
1767 if (skb2)
1768 ipmr_queue_xmit(net, mrt, skb2, cache,
1769 psend);
1770 }
1771 psend = ct;
1772 }
1773 }
1774 if (psend != -1) {
1775 if (local) {
1776 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1777
1778 if (skb2)
1779 ipmr_queue_xmit(net, mrt, skb2, cache, psend);
1780 } else {
1781 ipmr_queue_xmit(net, mrt, skb, cache, psend);
1782 return 0;
1783 }
1784 }
1785
1786dont_forward:
1787 if (!local)
1788 kfree_skb(skb);
1789 return 0;
1790}
1791
1792static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
1793{
1794 struct rtable *rt = skb_rtable(skb);
1795 struct iphdr *iph = ip_hdr(skb);
1796 struct flowi4 fl4 = {
1797 .daddr = iph->daddr,
1798 .saddr = iph->saddr,
1799 .flowi4_tos = RT_TOS(iph->tos),
1800 .flowi4_oif = rt->rt_oif,
1801 .flowi4_iif = rt->rt_iif,
1802 .flowi4_mark = rt->rt_mark,
1803 };
1804 struct mr_table *mrt;
1805 int err;
1806
1807 err = ipmr_fib_lookup(net, &fl4, &mrt);
1808 if (err)
1809 return ERR_PTR(err);
1810 return mrt;
1811}
1812
1813/*
1814 * Multicast packets for forwarding arrive here
1815 * Called with rcu_read_lock();
1816 */
1817
1818int ip_mr_input(struct sk_buff *skb)
1819{
1820 struct mfc_cache *cache;
1821 struct net *net = dev_net(skb->dev);
1822 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
1823 struct mr_table *mrt;
1824
1825 /* Packet is looped back after forward, it should not be
1826 * forwarded second time, but still can be delivered locally.
1827 */
1828 if (IPCB(skb)->flags & IPSKB_FORWARDED)
1829 goto dont_forward;
1830
1831 mrt = ipmr_rt_fib_lookup(net, skb);
1832 if (IS_ERR(mrt)) {
1833 kfree_skb(skb);
1834 return PTR_ERR(mrt);
1835 }
1836 if (!local) {
1837 if (IPCB(skb)->opt.router_alert) {
1838 if (ip_call_ra_chain(skb))
1839 return 0;
1840 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
1841 /* IGMPv1 (and broken IGMPv2 implementations sort of
1842 * Cisco IOS <= 11.2(8)) do not put router alert
1843 * option to IGMP packets destined to routable
1844 * groups. It is very bad, because it means
1845 * that we can forward NO IGMP messages.
1846 */
1847 struct sock *mroute_sk;
1848
1849 mroute_sk = rcu_dereference(mrt->mroute_sk);
1850 if (mroute_sk) {
1851 nf_reset(skb);
1852 raw_rcv(mroute_sk, skb);
1853 return 0;
1854 }
1855 }
1856 }
1857
1858 /* already under rcu_read_lock() */
1859 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
1860
1861 /*
1862 * No usable cache entry
1863 */
1864 if (cache == NULL) {
1865 int vif;
1866
1867 if (local) {
1868 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1869 ip_local_deliver(skb);
1870 if (skb2 == NULL)
1871 return -ENOBUFS;
1872 skb = skb2;
1873 }
1874
1875 read_lock(&mrt_lock);
1876 vif = ipmr_find_vif(mrt, skb->dev);
1877 if (vif >= 0) {
1878 int err2 = ipmr_cache_unresolved(mrt, vif, skb);
1879 read_unlock(&mrt_lock);
1880
1881 return err2;
1882 }
1883 read_unlock(&mrt_lock);
1884 kfree_skb(skb);
1885 return -ENODEV;
1886 }
1887
1888 read_lock(&mrt_lock);
1889 ip_mr_forward(net, mrt, skb, cache, local);
1890 read_unlock(&mrt_lock);
1891
1892 if (local)
1893 return ip_local_deliver(skb);
1894
1895 return 0;
1896
1897dont_forward:
1898 if (local)
1899 return ip_local_deliver(skb);
1900 kfree_skb(skb);
1901 return 0;
1902}
1903
1904#ifdef CONFIG_IP_PIMSM
1905/* called with rcu_read_lock() */
1906static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
1907 unsigned int pimlen)
1908{
1909 struct net_device *reg_dev = NULL;
1910 struct iphdr *encap;
1911
1912 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
1913 /*
1914 * Check that:
1915 * a. packet is really sent to a multicast group
1916 * b. packet is not a NULL-REGISTER
1917 * c. packet is not truncated
1918 */
1919 if (!ipv4_is_multicast(encap->daddr) ||
1920 encap->tot_len == 0 ||
1921 ntohs(encap->tot_len) + pimlen > skb->len)
1922 return 1;
1923
1924 read_lock(&mrt_lock);
1925 if (mrt->mroute_reg_vif_num >= 0)
1926 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
1927 read_unlock(&mrt_lock);
1928
1929 if (reg_dev == NULL)
1930 return 1;
1931
1932 skb->mac_header = skb->network_header;
1933 skb_pull(skb, (u8 *)encap - skb->data);
1934 skb_reset_network_header(skb);
1935 skb->protocol = htons(ETH_P_IP);
1936 skb->ip_summed = CHECKSUM_NONE;
1937 skb->pkt_type = PACKET_HOST;
1938
1939 skb_tunnel_rx(skb, reg_dev);
1940
1941 netif_rx(skb);
1942
1943 return NET_RX_SUCCESS;
1944}
1945#endif
1946
1947#ifdef CONFIG_IP_PIMSM_V1
1948/*
1949 * Handle IGMP messages of PIMv1
1950 */
1951
1952int pim_rcv_v1(struct sk_buff *skb)
1953{
1954 struct igmphdr *pim;
1955 struct net *net = dev_net(skb->dev);
1956 struct mr_table *mrt;
1957
1958 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
1959 goto drop;
1960
1961 pim = igmp_hdr(skb);
1962
1963 mrt = ipmr_rt_fib_lookup(net, skb);
1964 if (IS_ERR(mrt))
1965 goto drop;
1966 if (!mrt->mroute_do_pim ||
1967 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
1968 goto drop;
1969
1970 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
1971drop:
1972 kfree_skb(skb);
1973 }
1974 return 0;
1975}
1976#endif
1977
1978#ifdef CONFIG_IP_PIMSM_V2
1979static int pim_rcv(struct sk_buff *skb)
1980{
1981 struct pimreghdr *pim;
1982 struct net *net = dev_net(skb->dev);
1983 struct mr_table *mrt;
1984
1985 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
1986 goto drop;
1987
1988 pim = (struct pimreghdr *)skb_transport_header(skb);
1989 if (pim->type != ((PIM_VERSION << 4) | (PIM_REGISTER)) ||
1990 (pim->flags & PIM_NULL_REGISTER) ||
1991 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
1992 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
1993 goto drop;
1994
1995 mrt = ipmr_rt_fib_lookup(net, skb);
1996 if (IS_ERR(mrt))
1997 goto drop;
1998 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
1999drop:
2000 kfree_skb(skb);
2001 }
2002 return 0;
2003}
2004#endif
2005
2006static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2007 struct mfc_cache *c, struct rtmsg *rtm)
2008{
2009 int ct;
2010 struct rtnexthop *nhp;
2011 u8 *b = skb_tail_pointer(skb);
2012 struct rtattr *mp_head;
2013
2014 /* If cache is unresolved, don't try to parse IIF and OIF */
2015 if (c->mfc_parent >= MAXVIFS)
2016 return -ENOENT;
2017
2018 if (VIF_EXISTS(mrt, c->mfc_parent))
2019 RTA_PUT(skb, RTA_IIF, 4, &mrt->vif_table[c->mfc_parent].dev->ifindex);
2020
2021 mp_head = (struct rtattr *)skb_put(skb, RTA_LENGTH(0));
2022
2023 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2024 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2025 if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
2026 goto rtattr_failure;
2027 nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp)));
2028 nhp->rtnh_flags = 0;
2029 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2030 nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex;
2031 nhp->rtnh_len = sizeof(*nhp);
2032 }
2033 }
2034 mp_head->rta_type = RTA_MULTIPATH;
2035 mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head;
2036 rtm->rtm_type = RTN_MULTICAST;
2037 return 1;
2038
2039rtattr_failure:
2040 nlmsg_trim(skb, b);
2041 return -EMSGSIZE;
2042}
2043
2044int ipmr_get_route(struct net *net, struct sk_buff *skb,
2045 __be32 saddr, __be32 daddr,
2046 struct rtmsg *rtm, int nowait)
2047{
2048 struct mfc_cache *cache;
2049 struct mr_table *mrt;
2050 int err;
2051
2052 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2053 if (mrt == NULL)
2054 return -ENOENT;
2055
2056 rcu_read_lock();
2057 cache = ipmr_cache_find(mrt, saddr, daddr);
2058
2059 if (cache == NULL) {
2060 struct sk_buff *skb2;
2061 struct iphdr *iph;
2062 struct net_device *dev;
2063 int vif = -1;
2064
2065 if (nowait) {
2066 rcu_read_unlock();
2067 return -EAGAIN;
2068 }
2069
2070 dev = skb->dev;
2071 read_lock(&mrt_lock);
2072 if (dev)
2073 vif = ipmr_find_vif(mrt, dev);
2074 if (vif < 0) {
2075 read_unlock(&mrt_lock);
2076 rcu_read_unlock();
2077 return -ENODEV;
2078 }
2079 skb2 = skb_clone(skb, GFP_ATOMIC);
2080 if (!skb2) {
2081 read_unlock(&mrt_lock);
2082 rcu_read_unlock();
2083 return -ENOMEM;
2084 }
2085
2086 skb_push(skb2, sizeof(struct iphdr));
2087 skb_reset_network_header(skb2);
2088 iph = ip_hdr(skb2);
2089 iph->ihl = sizeof(struct iphdr) >> 2;
2090 iph->saddr = saddr;
2091 iph->daddr = daddr;
2092 iph->version = 0;
2093 err = ipmr_cache_unresolved(mrt, vif, skb2);
2094 read_unlock(&mrt_lock);
2095 rcu_read_unlock();
2096 return err;
2097 }
2098
2099 read_lock(&mrt_lock);
2100 if (!nowait && (rtm->rtm_flags & RTM_F_NOTIFY))
2101 cache->mfc_flags |= MFC_NOTIFY;
2102 err = __ipmr_fill_mroute(mrt, skb, cache, rtm);
2103 read_unlock(&mrt_lock);
2104 rcu_read_unlock();
2105 return err;
2106}
2107
2108static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2109 u32 pid, u32 seq, struct mfc_cache *c)
2110{
2111 struct nlmsghdr *nlh;
2112 struct rtmsg *rtm;
2113
2114 nlh = nlmsg_put(skb, pid, seq, RTM_NEWROUTE, sizeof(*rtm), NLM_F_MULTI);
2115 if (nlh == NULL)
2116 return -EMSGSIZE;
2117
2118 rtm = nlmsg_data(nlh);
2119 rtm->rtm_family = RTNL_FAMILY_IPMR;
2120 rtm->rtm_dst_len = 32;
2121 rtm->rtm_src_len = 32;
2122 rtm->rtm_tos = 0;
2123 rtm->rtm_table = mrt->id;
2124 NLA_PUT_U32(skb, RTA_TABLE, mrt->id);
2125 rtm->rtm_type = RTN_MULTICAST;
2126 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2127 rtm->rtm_protocol = RTPROT_UNSPEC;
2128 rtm->rtm_flags = 0;
2129
2130 NLA_PUT_BE32(skb, RTA_SRC, c->mfc_origin);
2131 NLA_PUT_BE32(skb, RTA_DST, c->mfc_mcastgrp);
2132
2133 if (__ipmr_fill_mroute(mrt, skb, c, rtm) < 0)
2134 goto nla_put_failure;
2135
2136 return nlmsg_end(skb, nlh);
2137
2138nla_put_failure:
2139 nlmsg_cancel(skb, nlh);
2140 return -EMSGSIZE;
2141}
2142
2143static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2144{
2145 struct net *net = sock_net(skb->sk);
2146 struct mr_table *mrt;
2147 struct mfc_cache *mfc;
2148 unsigned int t = 0, s_t;
2149 unsigned int h = 0, s_h;
2150 unsigned int e = 0, s_e;
2151
2152 s_t = cb->args[0];
2153 s_h = cb->args[1];
2154 s_e = cb->args[2];
2155
2156 rcu_read_lock();
2157 ipmr_for_each_table(mrt, net) {
2158 if (t < s_t)
2159 goto next_table;
2160 if (t > s_t)
2161 s_h = 0;
2162 for (h = s_h; h < MFC_LINES; h++) {
2163 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_array[h], list) {
2164 if (e < s_e)
2165 goto next_entry;
2166 if (ipmr_fill_mroute(mrt, skb,
2167 NETLINK_CB(cb->skb).pid,
2168 cb->nlh->nlmsg_seq,
2169 mfc) < 0)
2170 goto done;
2171next_entry:
2172 e++;
2173 }
2174 e = s_e = 0;
2175 }
2176 s_h = 0;
2177next_table:
2178 t++;
2179 }
2180done:
2181 rcu_read_unlock();
2182
2183 cb->args[2] = e;
2184 cb->args[1] = h;
2185 cb->args[0] = t;
2186
2187 return skb->len;
2188}
2189
2190#ifdef CONFIG_PROC_FS
2191/*
2192 * The /proc interfaces to multicast routing :
2193 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2194 */
2195struct ipmr_vif_iter {
2196 struct seq_net_private p;
2197 struct mr_table *mrt;
2198 int ct;
2199};
2200
2201static struct vif_device *ipmr_vif_seq_idx(struct net *net,
2202 struct ipmr_vif_iter *iter,
2203 loff_t pos)
2204{
2205 struct mr_table *mrt = iter->mrt;
2206
2207 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
2208 if (!VIF_EXISTS(mrt, iter->ct))
2209 continue;
2210 if (pos-- == 0)
2211 return &mrt->vif_table[iter->ct];
2212 }
2213 return NULL;
2214}
2215
2216static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2217 __acquires(mrt_lock)
2218{
2219 struct ipmr_vif_iter *iter = seq->private;
2220 struct net *net = seq_file_net(seq);
2221 struct mr_table *mrt;
2222
2223 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2224 if (mrt == NULL)
2225 return ERR_PTR(-ENOENT);
2226
2227 iter->mrt = mrt;
2228
2229 read_lock(&mrt_lock);
2230 return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
2231 : SEQ_START_TOKEN;
2232}
2233
2234static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2235{
2236 struct ipmr_vif_iter *iter = seq->private;
2237 struct net *net = seq_file_net(seq);
2238 struct mr_table *mrt = iter->mrt;
2239
2240 ++*pos;
2241 if (v == SEQ_START_TOKEN)
2242 return ipmr_vif_seq_idx(net, iter, 0);
2243
2244 while (++iter->ct < mrt->maxvif) {
2245 if (!VIF_EXISTS(mrt, iter->ct))
2246 continue;
2247 return &mrt->vif_table[iter->ct];
2248 }
2249 return NULL;
2250}
2251
2252static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2253 __releases(mrt_lock)
2254{
2255 read_unlock(&mrt_lock);
2256}
2257
2258static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2259{
2260 struct ipmr_vif_iter *iter = seq->private;
2261 struct mr_table *mrt = iter->mrt;
2262
2263 if (v == SEQ_START_TOKEN) {
2264 seq_puts(seq,
2265 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2266 } else {
2267 const struct vif_device *vif = v;
2268 const char *name = vif->dev ? vif->dev->name : "none";
2269
2270 seq_printf(seq,
2271 "%2Zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2272 vif - mrt->vif_table,
2273 name, vif->bytes_in, vif->pkt_in,
2274 vif->bytes_out, vif->pkt_out,
2275 vif->flags, vif->local, vif->remote);
2276 }
2277 return 0;
2278}
2279
2280static const struct seq_operations ipmr_vif_seq_ops = {
2281 .start = ipmr_vif_seq_start,
2282 .next = ipmr_vif_seq_next,
2283 .stop = ipmr_vif_seq_stop,
2284 .show = ipmr_vif_seq_show,
2285};
2286
2287static int ipmr_vif_open(struct inode *inode, struct file *file)
2288{
2289 return seq_open_net(inode, file, &ipmr_vif_seq_ops,
2290 sizeof(struct ipmr_vif_iter));
2291}
2292
2293static const struct file_operations ipmr_vif_fops = {
2294 .owner = THIS_MODULE,
2295 .open = ipmr_vif_open,
2296 .read = seq_read,
2297 .llseek = seq_lseek,
2298 .release = seq_release_net,
2299};
2300
2301struct ipmr_mfc_iter {
2302 struct seq_net_private p;
2303 struct mr_table *mrt;
2304 struct list_head *cache;
2305 int ct;
2306};
2307
2308
2309static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
2310 struct ipmr_mfc_iter *it, loff_t pos)
2311{
2312 struct mr_table *mrt = it->mrt;
2313 struct mfc_cache *mfc;
2314
2315 rcu_read_lock();
2316 for (it->ct = 0; it->ct < MFC_LINES; it->ct++) {
2317 it->cache = &mrt->mfc_cache_array[it->ct];
2318 list_for_each_entry_rcu(mfc, it->cache, list)
2319 if (pos-- == 0)
2320 return mfc;
2321 }
2322 rcu_read_unlock();
2323
2324 spin_lock_bh(&mfc_unres_lock);
2325 it->cache = &mrt->mfc_unres_queue;
2326 list_for_each_entry(mfc, it->cache, list)
2327 if (pos-- == 0)
2328 return mfc;
2329 spin_unlock_bh(&mfc_unres_lock);
2330
2331 it->cache = NULL;
2332 return NULL;
2333}
2334
2335
2336static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2337{
2338 struct ipmr_mfc_iter *it = seq->private;
2339 struct net *net = seq_file_net(seq);
2340 struct mr_table *mrt;
2341
2342 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2343 if (mrt == NULL)
2344 return ERR_PTR(-ENOENT);
2345
2346 it->mrt = mrt;
2347 it->cache = NULL;
2348 it->ct = 0;
2349 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
2350 : SEQ_START_TOKEN;
2351}
2352
2353static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2354{
2355 struct mfc_cache *mfc = v;
2356 struct ipmr_mfc_iter *it = seq->private;
2357 struct net *net = seq_file_net(seq);
2358 struct mr_table *mrt = it->mrt;
2359
2360 ++*pos;
2361
2362 if (v == SEQ_START_TOKEN)
2363 return ipmr_mfc_seq_idx(net, seq->private, 0);
2364
2365 if (mfc->list.next != it->cache)
2366 return list_entry(mfc->list.next, struct mfc_cache, list);
2367
2368 if (it->cache == &mrt->mfc_unres_queue)
2369 goto end_of_list;
2370
2371 BUG_ON(it->cache != &mrt->mfc_cache_array[it->ct]);
2372
2373 while (++it->ct < MFC_LINES) {
2374 it->cache = &mrt->mfc_cache_array[it->ct];
2375 if (list_empty(it->cache))
2376 continue;
2377 return list_first_entry(it->cache, struct mfc_cache, list);
2378 }
2379
2380 /* exhausted cache_array, show unresolved */
2381 rcu_read_unlock();
2382 it->cache = &mrt->mfc_unres_queue;
2383 it->ct = 0;
2384
2385 spin_lock_bh(&mfc_unres_lock);
2386 if (!list_empty(it->cache))
2387 return list_first_entry(it->cache, struct mfc_cache, list);
2388
2389end_of_list:
2390 spin_unlock_bh(&mfc_unres_lock);
2391 it->cache = NULL;
2392
2393 return NULL;
2394}
2395
2396static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
2397{
2398 struct ipmr_mfc_iter *it = seq->private;
2399 struct mr_table *mrt = it->mrt;
2400
2401 if (it->cache == &mrt->mfc_unres_queue)
2402 spin_unlock_bh(&mfc_unres_lock);
2403 else if (it->cache == &mrt->mfc_cache_array[it->ct])
2404 rcu_read_unlock();
2405}
2406
2407static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2408{
2409 int n;
2410
2411 if (v == SEQ_START_TOKEN) {
2412 seq_puts(seq,
2413 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2414 } else {
2415 const struct mfc_cache *mfc = v;
2416 const struct ipmr_mfc_iter *it = seq->private;
2417 const struct mr_table *mrt = it->mrt;
2418
2419 seq_printf(seq, "%08X %08X %-3hd",
2420 (__force u32) mfc->mfc_mcastgrp,
2421 (__force u32) mfc->mfc_origin,
2422 mfc->mfc_parent);
2423
2424 if (it->cache != &mrt->mfc_unres_queue) {
2425 seq_printf(seq, " %8lu %8lu %8lu",
2426 mfc->mfc_un.res.pkt,
2427 mfc->mfc_un.res.bytes,
2428 mfc->mfc_un.res.wrong_if);
2429 for (n = mfc->mfc_un.res.minvif;
2430 n < mfc->mfc_un.res.maxvif; n++) {
2431 if (VIF_EXISTS(mrt, n) &&
2432 mfc->mfc_un.res.ttls[n] < 255)
2433 seq_printf(seq,
2434 " %2d:%-3d",
2435 n, mfc->mfc_un.res.ttls[n]);
2436 }
2437 } else {
2438 /* unresolved mfc_caches don't contain
2439 * pkt, bytes and wrong_if values
2440 */
2441 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2442 }
2443 seq_putc(seq, '\n');
2444 }
2445 return 0;
2446}
2447
2448static const struct seq_operations ipmr_mfc_seq_ops = {
2449 .start = ipmr_mfc_seq_start,
2450 .next = ipmr_mfc_seq_next,
2451 .stop = ipmr_mfc_seq_stop,
2452 .show = ipmr_mfc_seq_show,
2453};
2454
2455static int ipmr_mfc_open(struct inode *inode, struct file *file)
2456{
2457 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
2458 sizeof(struct ipmr_mfc_iter));
2459}
2460
2461static const struct file_operations ipmr_mfc_fops = {
2462 .owner = THIS_MODULE,
2463 .open = ipmr_mfc_open,
2464 .read = seq_read,
2465 .llseek = seq_lseek,
2466 .release = seq_release_net,
2467};
2468#endif
2469
2470#ifdef CONFIG_IP_PIMSM_V2
2471static const struct net_protocol pim_protocol = {
2472 .handler = pim_rcv,
2473 .netns_ok = 1,
2474};
2475#endif
2476
2477
2478/*
2479 * Setup for IP multicast routing
2480 */
2481static int __net_init ipmr_net_init(struct net *net)
2482{
2483 int err;
2484
2485 err = ipmr_rules_init(net);
2486 if (err < 0)
2487 goto fail;
2488
2489#ifdef CONFIG_PROC_FS
2490 err = -ENOMEM;
2491 if (!proc_net_fops_create(net, "ip_mr_vif", 0, &ipmr_vif_fops))
2492 goto proc_vif_fail;
2493 if (!proc_net_fops_create(net, "ip_mr_cache", 0, &ipmr_mfc_fops))
2494 goto proc_cache_fail;
2495#endif
2496 return 0;
2497
2498#ifdef CONFIG_PROC_FS
2499proc_cache_fail:
2500 proc_net_remove(net, "ip_mr_vif");
2501proc_vif_fail:
2502 ipmr_rules_exit(net);
2503#endif
2504fail:
2505 return err;
2506}
2507
2508static void __net_exit ipmr_net_exit(struct net *net)
2509{
2510#ifdef CONFIG_PROC_FS
2511 proc_net_remove(net, "ip_mr_cache");
2512 proc_net_remove(net, "ip_mr_vif");
2513#endif
2514 ipmr_rules_exit(net);
2515}
2516
2517static struct pernet_operations ipmr_net_ops = {
2518 .init = ipmr_net_init,
2519 .exit = ipmr_net_exit,
2520};
2521
2522int __init ip_mr_init(void)
2523{
2524 int err;
2525
2526 mrt_cachep = kmem_cache_create("ip_mrt_cache",
2527 sizeof(struct mfc_cache),
2528 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
2529 NULL);
2530 if (!mrt_cachep)
2531 return -ENOMEM;
2532
2533 err = register_pernet_subsys(&ipmr_net_ops);
2534 if (err)
2535 goto reg_pernet_fail;
2536
2537 err = register_netdevice_notifier(&ip_mr_notifier);
2538 if (err)
2539 goto reg_notif_fail;
2540#ifdef CONFIG_IP_PIMSM_V2
2541 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
2542 printk(KERN_ERR "ip_mr_init: can't add PIM protocol\n");
2543 err = -EAGAIN;
2544 goto add_proto_fail;
2545 }
2546#endif
2547 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
2548 NULL, ipmr_rtm_dumproute, NULL);
2549 return 0;
2550
2551#ifdef CONFIG_IP_PIMSM_V2
2552add_proto_fail:
2553 unregister_netdevice_notifier(&ip_mr_notifier);
2554#endif
2555reg_notif_fail:
2556 unregister_pernet_subsys(&ipmr_net_ops);
2557reg_pernet_fail:
2558 kmem_cache_destroy(mrt_cachep);
2559 return err;
2560}
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
74struct 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
94struct ipmr_rule {
95 struct fib_rule common;
96};
97
98struct 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
106static 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 */
115static 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
125static struct kmem_cache *mrt_cachep __read_mostly;
126
127static struct mr_table *ipmr_new_table(struct net *net, u32 id);
128static void ipmr_free_table(struct mr_table *mrt);
129
130static void ip_mr_forward(struct net *net, struct mr_table *mrt,
131 struct sk_buff *skb, struct mfc_cache *cache,
132 int local);
133static int ipmr_cache_report(struct mr_table *mrt,
134 struct sk_buff *pkt, vifi_t vifi, int assert);
135static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
136 struct mfc_cache *c, struct rtmsg *rtm);
137static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
138 int cmd);
139static void mroute_clean_tables(struct mr_table *mrt);
140static 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
146static 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
157static 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
175static 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
200static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
201{
202 return 1;
203}
204
205static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
206 FRA_GENERIC_POLICY,
207};
208
209static 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
215static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
216 struct nlattr **tb)
217{
218 return 1;
219}
220
221static 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
230static 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
245static 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
270err2:
271 kfree(mrt);
272err1:
273 fib_rules_unregister(ops);
274 return err;
275}
276
277static 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
291static struct mr_table *ipmr_get_table(struct net *net, u32 id)
292{
293 return net->ipv4.mrt;
294}
295
296static 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
303static 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
309static void __net_exit ipmr_rules_exit(struct net *net)
310{
311 ipmr_free_table(net->ipv4.mrt);
312}
313#endif
314
315static 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
348static void ipmr_free_table(struct mr_table *mrt)
349{
350 del_timer_sync(&mrt->ipmr_expire_timer);
351 mroute_clean_tables(mrt);
352 kfree(mrt);
353}
354
355/* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
356
357static 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
388static
389struct 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
441failure:
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
452static 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
478static const struct net_device_ops reg_vif_netdev_ops = {
479 .ndo_start_xmit = reg_vif_xmit,
480};
481
482static 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
492static 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, 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
535failure:
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
550static 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
606static 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
613static 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
622static 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
652static 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
686out:
687 spin_unlock(&mfc_unres_lock);
688}
689
690/* Fill oifs list. It is called under write locked mrt_lock. */
691
692static 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
713static 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() */
822static 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 */
837static 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 */
853static 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
875skip:
876 return ipmr_cache_find_any_parent(mrt, vifi);
877}
878
879/*
880 * Allocate a multicast cache entry
881 */
882static 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
891static 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
906static 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
944static 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
1030static int
1031ipmr_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
1104static 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
1124static 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
1204static void mroute_clean_tables(struct mr_table *mrt)
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 (!(mrt->vif_table[i].flags & VIFF_STATIC))
1214 vif_delete(mrt, i, 0, &list);
1215 }
1216 unregister_netdevice_many(&list);
1217
1218 /* Wipe the cache */
1219
1220 for (i = 0; i < MFC_LINES; i++) {
1221 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[i], list) {
1222 if (c->mfc_flags & MFC_STATIC)
1223 continue;
1224 list_del_rcu(&c->list);
1225 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1226 ipmr_cache_free(c);
1227 }
1228 }
1229
1230 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1231 spin_lock_bh(&mfc_unres_lock);
1232 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
1233 list_del(&c->list);
1234 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1235 ipmr_destroy_unres(mrt, c);
1236 }
1237 spin_unlock_bh(&mfc_unres_lock);
1238 }
1239}
1240
1241/* called from ip_ra_control(), before an RCU grace period,
1242 * we dont need to call synchronize_rcu() here
1243 */
1244static void mrtsock_destruct(struct sock *sk)
1245{
1246 struct net *net = sock_net(sk);
1247 struct mr_table *mrt;
1248
1249 rtnl_lock();
1250 ipmr_for_each_table(mrt, net) {
1251 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1252 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1253 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1254 NETCONFA_IFINDEX_ALL,
1255 net->ipv4.devconf_all);
1256 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1257 mroute_clean_tables(mrt);
1258 }
1259 }
1260 rtnl_unlock();
1261}
1262
1263/*
1264 * Socket options and virtual interface manipulation. The whole
1265 * virtual interface system is a complete heap, but unfortunately
1266 * that's how BSD mrouted happens to think. Maybe one day with a proper
1267 * MOSPF/PIM router set up we can clean this up.
1268 */
1269
1270int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1271{
1272 int ret, parent = 0;
1273 struct vifctl vif;
1274 struct mfcctl mfc;
1275 struct net *net = sock_net(sk);
1276 struct mr_table *mrt;
1277
1278 if (sk->sk_type != SOCK_RAW ||
1279 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1280 return -EOPNOTSUPP;
1281
1282 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1283 if (mrt == NULL)
1284 return -ENOENT;
1285
1286 if (optname != MRT_INIT) {
1287 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1288 !ns_capable(net->user_ns, CAP_NET_ADMIN))
1289 return -EACCES;
1290 }
1291
1292 switch (optname) {
1293 case MRT_INIT:
1294 if (optlen != sizeof(int))
1295 return -EINVAL;
1296
1297 rtnl_lock();
1298 if (rtnl_dereference(mrt->mroute_sk)) {
1299 rtnl_unlock();
1300 return -EADDRINUSE;
1301 }
1302
1303 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1304 if (ret == 0) {
1305 rcu_assign_pointer(mrt->mroute_sk, sk);
1306 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1307 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1308 NETCONFA_IFINDEX_ALL,
1309 net->ipv4.devconf_all);
1310 }
1311 rtnl_unlock();
1312 return ret;
1313 case MRT_DONE:
1314 if (sk != rcu_access_pointer(mrt->mroute_sk))
1315 return -EACCES;
1316 return ip_ra_control(sk, 0, NULL);
1317 case MRT_ADD_VIF:
1318 case MRT_DEL_VIF:
1319 if (optlen != sizeof(vif))
1320 return -EINVAL;
1321 if (copy_from_user(&vif, optval, sizeof(vif)))
1322 return -EFAULT;
1323 if (vif.vifc_vifi >= MAXVIFS)
1324 return -ENFILE;
1325 rtnl_lock();
1326 if (optname == MRT_ADD_VIF) {
1327 ret = vif_add(net, mrt, &vif,
1328 sk == rtnl_dereference(mrt->mroute_sk));
1329 } else {
1330 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1331 }
1332 rtnl_unlock();
1333 return ret;
1334
1335 /*
1336 * Manipulate the forwarding caches. These live
1337 * in a sort of kernel/user symbiosis.
1338 */
1339 case MRT_ADD_MFC:
1340 case MRT_DEL_MFC:
1341 parent = -1;
1342 case MRT_ADD_MFC_PROXY:
1343 case MRT_DEL_MFC_PROXY:
1344 if (optlen != sizeof(mfc))
1345 return -EINVAL;
1346 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1347 return -EFAULT;
1348 if (parent == 0)
1349 parent = mfc.mfcc_parent;
1350 rtnl_lock();
1351 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1352 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1353 else
1354 ret = ipmr_mfc_add(net, mrt, &mfc,
1355 sk == rtnl_dereference(mrt->mroute_sk),
1356 parent);
1357 rtnl_unlock();
1358 return ret;
1359 /*
1360 * Control PIM assert.
1361 */
1362 case MRT_ASSERT:
1363 {
1364 int v;
1365 if (optlen != sizeof(v))
1366 return -EINVAL;
1367 if (get_user(v, (int __user *)optval))
1368 return -EFAULT;
1369 mrt->mroute_do_assert = v;
1370 return 0;
1371 }
1372#ifdef CONFIG_IP_PIMSM
1373 case MRT_PIM:
1374 {
1375 int v;
1376
1377 if (optlen != sizeof(v))
1378 return -EINVAL;
1379 if (get_user(v, (int __user *)optval))
1380 return -EFAULT;
1381 v = !!v;
1382
1383 rtnl_lock();
1384 ret = 0;
1385 if (v != mrt->mroute_do_pim) {
1386 mrt->mroute_do_pim = v;
1387 mrt->mroute_do_assert = v;
1388 }
1389 rtnl_unlock();
1390 return ret;
1391 }
1392#endif
1393#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
1394 case MRT_TABLE:
1395 {
1396 u32 v;
1397
1398 if (optlen != sizeof(u32))
1399 return -EINVAL;
1400 if (get_user(v, (u32 __user *)optval))
1401 return -EFAULT;
1402
1403 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
1404 if (v != RT_TABLE_DEFAULT && v >= 1000000000)
1405 return -EINVAL;
1406
1407 rtnl_lock();
1408 ret = 0;
1409 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1410 ret = -EBUSY;
1411 } else {
1412 if (!ipmr_new_table(net, v))
1413 ret = -ENOMEM;
1414 else
1415 raw_sk(sk)->ipmr_table = v;
1416 }
1417 rtnl_unlock();
1418 return ret;
1419 }
1420#endif
1421 /*
1422 * Spurious command, or MRT_VERSION which you cannot
1423 * set.
1424 */
1425 default:
1426 return -ENOPROTOOPT;
1427 }
1428}
1429
1430/*
1431 * Getsock opt support for the multicast routing system.
1432 */
1433
1434int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1435{
1436 int olr;
1437 int val;
1438 struct net *net = sock_net(sk);
1439 struct mr_table *mrt;
1440
1441 if (sk->sk_type != SOCK_RAW ||
1442 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1443 return -EOPNOTSUPP;
1444
1445 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1446 if (mrt == NULL)
1447 return -ENOENT;
1448
1449 if (optname != MRT_VERSION &&
1450#ifdef CONFIG_IP_PIMSM
1451 optname != MRT_PIM &&
1452#endif
1453 optname != MRT_ASSERT)
1454 return -ENOPROTOOPT;
1455
1456 if (get_user(olr, optlen))
1457 return -EFAULT;
1458
1459 olr = min_t(unsigned int, olr, sizeof(int));
1460 if (olr < 0)
1461 return -EINVAL;
1462
1463 if (put_user(olr, optlen))
1464 return -EFAULT;
1465 if (optname == MRT_VERSION)
1466 val = 0x0305;
1467#ifdef CONFIG_IP_PIMSM
1468 else if (optname == MRT_PIM)
1469 val = mrt->mroute_do_pim;
1470#endif
1471 else
1472 val = mrt->mroute_do_assert;
1473 if (copy_to_user(optval, &val, olr))
1474 return -EFAULT;
1475 return 0;
1476}
1477
1478/*
1479 * The IP multicast ioctl support routines.
1480 */
1481
1482int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1483{
1484 struct sioc_sg_req sr;
1485 struct sioc_vif_req vr;
1486 struct vif_device *vif;
1487 struct mfc_cache *c;
1488 struct net *net = sock_net(sk);
1489 struct mr_table *mrt;
1490
1491 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1492 if (mrt == NULL)
1493 return -ENOENT;
1494
1495 switch (cmd) {
1496 case SIOCGETVIFCNT:
1497 if (copy_from_user(&vr, arg, sizeof(vr)))
1498 return -EFAULT;
1499 if (vr.vifi >= mrt->maxvif)
1500 return -EINVAL;
1501 read_lock(&mrt_lock);
1502 vif = &mrt->vif_table[vr.vifi];
1503 if (VIF_EXISTS(mrt, vr.vifi)) {
1504 vr.icount = vif->pkt_in;
1505 vr.ocount = vif->pkt_out;
1506 vr.ibytes = vif->bytes_in;
1507 vr.obytes = vif->bytes_out;
1508 read_unlock(&mrt_lock);
1509
1510 if (copy_to_user(arg, &vr, sizeof(vr)))
1511 return -EFAULT;
1512 return 0;
1513 }
1514 read_unlock(&mrt_lock);
1515 return -EADDRNOTAVAIL;
1516 case SIOCGETSGCNT:
1517 if (copy_from_user(&sr, arg, sizeof(sr)))
1518 return -EFAULT;
1519
1520 rcu_read_lock();
1521 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1522 if (c) {
1523 sr.pktcnt = c->mfc_un.res.pkt;
1524 sr.bytecnt = c->mfc_un.res.bytes;
1525 sr.wrong_if = c->mfc_un.res.wrong_if;
1526 rcu_read_unlock();
1527
1528 if (copy_to_user(arg, &sr, sizeof(sr)))
1529 return -EFAULT;
1530 return 0;
1531 }
1532 rcu_read_unlock();
1533 return -EADDRNOTAVAIL;
1534 default:
1535 return -ENOIOCTLCMD;
1536 }
1537}
1538
1539#ifdef CONFIG_COMPAT
1540struct compat_sioc_sg_req {
1541 struct in_addr src;
1542 struct in_addr grp;
1543 compat_ulong_t pktcnt;
1544 compat_ulong_t bytecnt;
1545 compat_ulong_t wrong_if;
1546};
1547
1548struct compat_sioc_vif_req {
1549 vifi_t vifi; /* Which iface */
1550 compat_ulong_t icount;
1551 compat_ulong_t ocount;
1552 compat_ulong_t ibytes;
1553 compat_ulong_t obytes;
1554};
1555
1556int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1557{
1558 struct compat_sioc_sg_req sr;
1559 struct compat_sioc_vif_req vr;
1560 struct vif_device *vif;
1561 struct mfc_cache *c;
1562 struct net *net = sock_net(sk);
1563 struct mr_table *mrt;
1564
1565 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1566 if (mrt == NULL)
1567 return -ENOENT;
1568
1569 switch (cmd) {
1570 case SIOCGETVIFCNT:
1571 if (copy_from_user(&vr, arg, sizeof(vr)))
1572 return -EFAULT;
1573 if (vr.vifi >= mrt->maxvif)
1574 return -EINVAL;
1575 read_lock(&mrt_lock);
1576 vif = &mrt->vif_table[vr.vifi];
1577 if (VIF_EXISTS(mrt, vr.vifi)) {
1578 vr.icount = vif->pkt_in;
1579 vr.ocount = vif->pkt_out;
1580 vr.ibytes = vif->bytes_in;
1581 vr.obytes = vif->bytes_out;
1582 read_unlock(&mrt_lock);
1583
1584 if (copy_to_user(arg, &vr, sizeof(vr)))
1585 return -EFAULT;
1586 return 0;
1587 }
1588 read_unlock(&mrt_lock);
1589 return -EADDRNOTAVAIL;
1590 case SIOCGETSGCNT:
1591 if (copy_from_user(&sr, arg, sizeof(sr)))
1592 return -EFAULT;
1593
1594 rcu_read_lock();
1595 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1596 if (c) {
1597 sr.pktcnt = c->mfc_un.res.pkt;
1598 sr.bytecnt = c->mfc_un.res.bytes;
1599 sr.wrong_if = c->mfc_un.res.wrong_if;
1600 rcu_read_unlock();
1601
1602 if (copy_to_user(arg, &sr, sizeof(sr)))
1603 return -EFAULT;
1604 return 0;
1605 }
1606 rcu_read_unlock();
1607 return -EADDRNOTAVAIL;
1608 default:
1609 return -ENOIOCTLCMD;
1610 }
1611}
1612#endif
1613
1614
1615static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1616{
1617 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1618 struct net *net = dev_net(dev);
1619 struct mr_table *mrt;
1620 struct vif_device *v;
1621 int ct;
1622
1623 if (event != NETDEV_UNREGISTER)
1624 return NOTIFY_DONE;
1625
1626 ipmr_for_each_table(mrt, net) {
1627 v = &mrt->vif_table[0];
1628 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1629 if (v->dev == dev)
1630 vif_delete(mrt, ct, 1, NULL);
1631 }
1632 }
1633 return NOTIFY_DONE;
1634}
1635
1636
1637static struct notifier_block ip_mr_notifier = {
1638 .notifier_call = ipmr_device_event,
1639};
1640
1641/*
1642 * Encapsulate a packet by attaching a valid IPIP header to it.
1643 * This avoids tunnel drivers and other mess and gives us the speed so
1644 * important for multicast video.
1645 */
1646
1647static void ip_encap(struct sk_buff *skb, __be32 saddr, __be32 daddr)
1648{
1649 struct iphdr *iph;
1650 const struct iphdr *old_iph = ip_hdr(skb);
1651
1652 skb_push(skb, sizeof(struct iphdr));
1653 skb->transport_header = skb->network_header;
1654 skb_reset_network_header(skb);
1655 iph = ip_hdr(skb);
1656
1657 iph->version = 4;
1658 iph->tos = old_iph->tos;
1659 iph->ttl = old_iph->ttl;
1660 iph->frag_off = 0;
1661 iph->daddr = daddr;
1662 iph->saddr = saddr;
1663 iph->protocol = IPPROTO_IPIP;
1664 iph->ihl = 5;
1665 iph->tot_len = htons(skb->len);
1666 ip_select_ident(skb, skb_dst(skb), NULL);
1667 ip_send_check(iph);
1668
1669 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1670 nf_reset(skb);
1671}
1672
1673static inline int ipmr_forward_finish(struct sk_buff *skb)
1674{
1675 struct ip_options *opt = &(IPCB(skb)->opt);
1676
1677 IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTFORWDATAGRAMS);
1678 IP_ADD_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTOCTETS, skb->len);
1679
1680 if (unlikely(opt->optlen))
1681 ip_forward_options(skb);
1682
1683 return dst_output(skb);
1684}
1685
1686/*
1687 * Processing handlers for ipmr_forward
1688 */
1689
1690static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1691 struct sk_buff *skb, struct mfc_cache *c, int vifi)
1692{
1693 const struct iphdr *iph = ip_hdr(skb);
1694 struct vif_device *vif = &mrt->vif_table[vifi];
1695 struct net_device *dev;
1696 struct rtable *rt;
1697 struct flowi4 fl4;
1698 int encap = 0;
1699
1700 if (vif->dev == NULL)
1701 goto out_free;
1702
1703#ifdef CONFIG_IP_PIMSM
1704 if (vif->flags & VIFF_REGISTER) {
1705 vif->pkt_out++;
1706 vif->bytes_out += skb->len;
1707 vif->dev->stats.tx_bytes += skb->len;
1708 vif->dev->stats.tx_packets++;
1709 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1710 goto out_free;
1711 }
1712#endif
1713
1714 if (vif->flags & VIFF_TUNNEL) {
1715 rt = ip_route_output_ports(net, &fl4, NULL,
1716 vif->remote, vif->local,
1717 0, 0,
1718 IPPROTO_IPIP,
1719 RT_TOS(iph->tos), vif->link);
1720 if (IS_ERR(rt))
1721 goto out_free;
1722 encap = sizeof(struct iphdr);
1723 } else {
1724 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1725 0, 0,
1726 IPPROTO_IPIP,
1727 RT_TOS(iph->tos), vif->link);
1728 if (IS_ERR(rt))
1729 goto out_free;
1730 }
1731
1732 dev = rt->dst.dev;
1733
1734 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1735 /* Do not fragment multicasts. Alas, IPv4 does not
1736 * allow to send ICMP, so that packets will disappear
1737 * to blackhole.
1738 */
1739
1740 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
1741 ip_rt_put(rt);
1742 goto out_free;
1743 }
1744
1745 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1746
1747 if (skb_cow(skb, encap)) {
1748 ip_rt_put(rt);
1749 goto out_free;
1750 }
1751
1752 vif->pkt_out++;
1753 vif->bytes_out += skb->len;
1754
1755 skb_dst_drop(skb);
1756 skb_dst_set(skb, &rt->dst);
1757 ip_decrease_ttl(ip_hdr(skb));
1758
1759 /* FIXME: forward and output firewalls used to be called here.
1760 * What do we do with netfilter? -- RR
1761 */
1762 if (vif->flags & VIFF_TUNNEL) {
1763 ip_encap(skb, vif->local, vif->remote);
1764 /* FIXME: extra output firewall step used to be here. --RR */
1765 vif->dev->stats.tx_packets++;
1766 vif->dev->stats.tx_bytes += skb->len;
1767 }
1768
1769 IPCB(skb)->flags |= IPSKB_FORWARDED;
1770
1771 /*
1772 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1773 * not only before forwarding, but after forwarding on all output
1774 * interfaces. It is clear, if mrouter runs a multicasting
1775 * program, it should receive packets not depending to what interface
1776 * program is joined.
1777 * If we will not make it, the program will have to join on all
1778 * interfaces. On the other hand, multihoming host (or router, but
1779 * not mrouter) cannot join to more than one interface - it will
1780 * result in receiving multiple packets.
1781 */
1782 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, skb, skb->dev, dev,
1783 ipmr_forward_finish);
1784 return;
1785
1786out_free:
1787 kfree_skb(skb);
1788}
1789
1790static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1791{
1792 int ct;
1793
1794 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1795 if (mrt->vif_table[ct].dev == dev)
1796 break;
1797 }
1798 return ct;
1799}
1800
1801/* "local" means that we should preserve one skb (for local delivery) */
1802
1803static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1804 struct sk_buff *skb, struct mfc_cache *cache,
1805 int local)
1806{
1807 int psend = -1;
1808 int vif, ct;
1809 int true_vifi = ipmr_find_vif(mrt, skb->dev);
1810
1811 vif = cache->mfc_parent;
1812 cache->mfc_un.res.pkt++;
1813 cache->mfc_un.res.bytes += skb->len;
1814
1815 if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1816 struct mfc_cache *cache_proxy;
1817
1818 /* For an (*,G) entry, we only check that the incomming
1819 * interface is part of the static tree.
1820 */
1821 cache_proxy = ipmr_cache_find_any_parent(mrt, vif);
1822 if (cache_proxy &&
1823 cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
1824 goto forward;
1825 }
1826
1827 /*
1828 * Wrong interface: drop packet and (maybe) send PIM assert.
1829 */
1830 if (mrt->vif_table[vif].dev != skb->dev) {
1831 if (rt_is_output_route(skb_rtable(skb))) {
1832 /* It is our own packet, looped back.
1833 * Very complicated situation...
1834 *
1835 * The best workaround until routing daemons will be
1836 * fixed is not to redistribute packet, if it was
1837 * send through wrong interface. It means, that
1838 * multicast applications WILL NOT work for
1839 * (S,G), which have default multicast route pointing
1840 * to wrong oif. In any case, it is not a good
1841 * idea to use multicasting applications on router.
1842 */
1843 goto dont_forward;
1844 }
1845
1846 cache->mfc_un.res.wrong_if++;
1847
1848 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1849 /* pimsm uses asserts, when switching from RPT to SPT,
1850 * so that we cannot check that packet arrived on an oif.
1851 * It is bad, but otherwise we would need to move pretty
1852 * large chunk of pimd to kernel. Ough... --ANK
1853 */
1854 (mrt->mroute_do_pim ||
1855 cache->mfc_un.res.ttls[true_vifi] < 255) &&
1856 time_after(jiffies,
1857 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1858 cache->mfc_un.res.last_assert = jiffies;
1859 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1860 }
1861 goto dont_forward;
1862 }
1863
1864forward:
1865 mrt->vif_table[vif].pkt_in++;
1866 mrt->vif_table[vif].bytes_in += skb->len;
1867
1868 /*
1869 * Forward the frame
1870 */
1871 if (cache->mfc_origin == htonl(INADDR_ANY) &&
1872 cache->mfc_mcastgrp == htonl(INADDR_ANY)) {
1873 if (true_vifi >= 0 &&
1874 true_vifi != cache->mfc_parent &&
1875 ip_hdr(skb)->ttl >
1876 cache->mfc_un.res.ttls[cache->mfc_parent]) {
1877 /* It's an (*,*) entry and the packet is not coming from
1878 * the upstream: forward the packet to the upstream
1879 * only.
1880 */
1881 psend = cache->mfc_parent;
1882 goto last_forward;
1883 }
1884 goto dont_forward;
1885 }
1886 for (ct = cache->mfc_un.res.maxvif - 1;
1887 ct >= cache->mfc_un.res.minvif; ct--) {
1888 /* For (*,G) entry, don't forward to the incoming interface */
1889 if ((cache->mfc_origin != htonl(INADDR_ANY) ||
1890 ct != true_vifi) &&
1891 ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
1892 if (psend != -1) {
1893 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1894
1895 if (skb2)
1896 ipmr_queue_xmit(net, mrt, skb2, cache,
1897 psend);
1898 }
1899 psend = ct;
1900 }
1901 }
1902last_forward:
1903 if (psend != -1) {
1904 if (local) {
1905 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1906
1907 if (skb2)
1908 ipmr_queue_xmit(net, mrt, skb2, cache, psend);
1909 } else {
1910 ipmr_queue_xmit(net, mrt, skb, cache, psend);
1911 return;
1912 }
1913 }
1914
1915dont_forward:
1916 if (!local)
1917 kfree_skb(skb);
1918}
1919
1920static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
1921{
1922 struct rtable *rt = skb_rtable(skb);
1923 struct iphdr *iph = ip_hdr(skb);
1924 struct flowi4 fl4 = {
1925 .daddr = iph->daddr,
1926 .saddr = iph->saddr,
1927 .flowi4_tos = RT_TOS(iph->tos),
1928 .flowi4_oif = (rt_is_output_route(rt) ?
1929 skb->dev->ifindex : 0),
1930 .flowi4_iif = (rt_is_output_route(rt) ?
1931 LOOPBACK_IFINDEX :
1932 skb->dev->ifindex),
1933 .flowi4_mark = skb->mark,
1934 };
1935 struct mr_table *mrt;
1936 int err;
1937
1938 err = ipmr_fib_lookup(net, &fl4, &mrt);
1939 if (err)
1940 return ERR_PTR(err);
1941 return mrt;
1942}
1943
1944/*
1945 * Multicast packets for forwarding arrive here
1946 * Called with rcu_read_lock();
1947 */
1948
1949int ip_mr_input(struct sk_buff *skb)
1950{
1951 struct mfc_cache *cache;
1952 struct net *net = dev_net(skb->dev);
1953 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
1954 struct mr_table *mrt;
1955
1956 /* Packet is looped back after forward, it should not be
1957 * forwarded second time, but still can be delivered locally.
1958 */
1959 if (IPCB(skb)->flags & IPSKB_FORWARDED)
1960 goto dont_forward;
1961
1962 mrt = ipmr_rt_fib_lookup(net, skb);
1963 if (IS_ERR(mrt)) {
1964 kfree_skb(skb);
1965 return PTR_ERR(mrt);
1966 }
1967 if (!local) {
1968 if (IPCB(skb)->opt.router_alert) {
1969 if (ip_call_ra_chain(skb))
1970 return 0;
1971 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
1972 /* IGMPv1 (and broken IGMPv2 implementations sort of
1973 * Cisco IOS <= 11.2(8)) do not put router alert
1974 * option to IGMP packets destined to routable
1975 * groups. It is very bad, because it means
1976 * that we can forward NO IGMP messages.
1977 */
1978 struct sock *mroute_sk;
1979
1980 mroute_sk = rcu_dereference(mrt->mroute_sk);
1981 if (mroute_sk) {
1982 nf_reset(skb);
1983 raw_rcv(mroute_sk, skb);
1984 return 0;
1985 }
1986 }
1987 }
1988
1989 /* already under rcu_read_lock() */
1990 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
1991 if (cache == NULL) {
1992 int vif = ipmr_find_vif(mrt, skb->dev);
1993
1994 if (vif >= 0)
1995 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
1996 vif);
1997 }
1998
1999 /*
2000 * No usable cache entry
2001 */
2002 if (cache == NULL) {
2003 int vif;
2004
2005 if (local) {
2006 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2007 ip_local_deliver(skb);
2008 if (skb2 == NULL)
2009 return -ENOBUFS;
2010 skb = skb2;
2011 }
2012
2013 read_lock(&mrt_lock);
2014 vif = ipmr_find_vif(mrt, skb->dev);
2015 if (vif >= 0) {
2016 int err2 = ipmr_cache_unresolved(mrt, vif, skb);
2017 read_unlock(&mrt_lock);
2018
2019 return err2;
2020 }
2021 read_unlock(&mrt_lock);
2022 kfree_skb(skb);
2023 return -ENODEV;
2024 }
2025
2026 read_lock(&mrt_lock);
2027 ip_mr_forward(net, mrt, skb, cache, local);
2028 read_unlock(&mrt_lock);
2029
2030 if (local)
2031 return ip_local_deliver(skb);
2032
2033 return 0;
2034
2035dont_forward:
2036 if (local)
2037 return ip_local_deliver(skb);
2038 kfree_skb(skb);
2039 return 0;
2040}
2041
2042#ifdef CONFIG_IP_PIMSM
2043/* called with rcu_read_lock() */
2044static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
2045 unsigned int pimlen)
2046{
2047 struct net_device *reg_dev = NULL;
2048 struct iphdr *encap;
2049
2050 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
2051 /*
2052 * Check that:
2053 * a. packet is really sent to a multicast group
2054 * b. packet is not a NULL-REGISTER
2055 * c. packet is not truncated
2056 */
2057 if (!ipv4_is_multicast(encap->daddr) ||
2058 encap->tot_len == 0 ||
2059 ntohs(encap->tot_len) + pimlen > skb->len)
2060 return 1;
2061
2062 read_lock(&mrt_lock);
2063 if (mrt->mroute_reg_vif_num >= 0)
2064 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
2065 read_unlock(&mrt_lock);
2066
2067 if (reg_dev == NULL)
2068 return 1;
2069
2070 skb->mac_header = skb->network_header;
2071 skb_pull(skb, (u8 *)encap - skb->data);
2072 skb_reset_network_header(skb);
2073 skb->protocol = htons(ETH_P_IP);
2074 skb->ip_summed = CHECKSUM_NONE;
2075
2076 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
2077
2078 netif_rx(skb);
2079
2080 return NET_RX_SUCCESS;
2081}
2082#endif
2083
2084#ifdef CONFIG_IP_PIMSM_V1
2085/*
2086 * Handle IGMP messages of PIMv1
2087 */
2088
2089int pim_rcv_v1(struct sk_buff *skb)
2090{
2091 struct igmphdr *pim;
2092 struct net *net = dev_net(skb->dev);
2093 struct mr_table *mrt;
2094
2095 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2096 goto drop;
2097
2098 pim = igmp_hdr(skb);
2099
2100 mrt = ipmr_rt_fib_lookup(net, skb);
2101 if (IS_ERR(mrt))
2102 goto drop;
2103 if (!mrt->mroute_do_pim ||
2104 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2105 goto drop;
2106
2107 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2108drop:
2109 kfree_skb(skb);
2110 }
2111 return 0;
2112}
2113#endif
2114
2115#ifdef CONFIG_IP_PIMSM_V2
2116static int pim_rcv(struct sk_buff *skb)
2117{
2118 struct pimreghdr *pim;
2119 struct net *net = dev_net(skb->dev);
2120 struct mr_table *mrt;
2121
2122 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2123 goto drop;
2124
2125 pim = (struct pimreghdr *)skb_transport_header(skb);
2126 if (pim->type != ((PIM_VERSION << 4) | (PIM_REGISTER)) ||
2127 (pim->flags & PIM_NULL_REGISTER) ||
2128 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2129 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2130 goto drop;
2131
2132 mrt = ipmr_rt_fib_lookup(net, skb);
2133 if (IS_ERR(mrt))
2134 goto drop;
2135 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2136drop:
2137 kfree_skb(skb);
2138 }
2139 return 0;
2140}
2141#endif
2142
2143static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2144 struct mfc_cache *c, struct rtmsg *rtm)
2145{
2146 int ct;
2147 struct rtnexthop *nhp;
2148 struct nlattr *mp_attr;
2149 struct rta_mfc_stats mfcs;
2150
2151 /* If cache is unresolved, don't try to parse IIF and OIF */
2152 if (c->mfc_parent >= MAXVIFS)
2153 return -ENOENT;
2154
2155 if (VIF_EXISTS(mrt, c->mfc_parent) &&
2156 nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
2157 return -EMSGSIZE;
2158
2159 if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH)))
2160 return -EMSGSIZE;
2161
2162 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2163 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2164 if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) {
2165 nla_nest_cancel(skb, mp_attr);
2166 return -EMSGSIZE;
2167 }
2168
2169 nhp->rtnh_flags = 0;
2170 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2171 nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex;
2172 nhp->rtnh_len = sizeof(*nhp);
2173 }
2174 }
2175
2176 nla_nest_end(skb, mp_attr);
2177
2178 mfcs.mfcs_packets = c->mfc_un.res.pkt;
2179 mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2180 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2181 if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0)
2182 return -EMSGSIZE;
2183
2184 rtm->rtm_type = RTN_MULTICAST;
2185 return 1;
2186}
2187
2188int ipmr_get_route(struct net *net, struct sk_buff *skb,
2189 __be32 saddr, __be32 daddr,
2190 struct rtmsg *rtm, int nowait)
2191{
2192 struct mfc_cache *cache;
2193 struct mr_table *mrt;
2194 int err;
2195
2196 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2197 if (mrt == NULL)
2198 return -ENOENT;
2199
2200 rcu_read_lock();
2201 cache = ipmr_cache_find(mrt, saddr, daddr);
2202 if (cache == NULL && skb->dev) {
2203 int vif = ipmr_find_vif(mrt, skb->dev);
2204
2205 if (vif >= 0)
2206 cache = ipmr_cache_find_any(mrt, daddr, vif);
2207 }
2208 if (cache == NULL) {
2209 struct sk_buff *skb2;
2210 struct iphdr *iph;
2211 struct net_device *dev;
2212 int vif = -1;
2213
2214 if (nowait) {
2215 rcu_read_unlock();
2216 return -EAGAIN;
2217 }
2218
2219 dev = skb->dev;
2220 read_lock(&mrt_lock);
2221 if (dev)
2222 vif = ipmr_find_vif(mrt, dev);
2223 if (vif < 0) {
2224 read_unlock(&mrt_lock);
2225 rcu_read_unlock();
2226 return -ENODEV;
2227 }
2228 skb2 = skb_clone(skb, GFP_ATOMIC);
2229 if (!skb2) {
2230 read_unlock(&mrt_lock);
2231 rcu_read_unlock();
2232 return -ENOMEM;
2233 }
2234
2235 skb_push(skb2, sizeof(struct iphdr));
2236 skb_reset_network_header(skb2);
2237 iph = ip_hdr(skb2);
2238 iph->ihl = sizeof(struct iphdr) >> 2;
2239 iph->saddr = saddr;
2240 iph->daddr = daddr;
2241 iph->version = 0;
2242 err = ipmr_cache_unresolved(mrt, vif, skb2);
2243 read_unlock(&mrt_lock);
2244 rcu_read_unlock();
2245 return err;
2246 }
2247
2248 read_lock(&mrt_lock);
2249 if (!nowait && (rtm->rtm_flags & RTM_F_NOTIFY))
2250 cache->mfc_flags |= MFC_NOTIFY;
2251 err = __ipmr_fill_mroute(mrt, skb, cache, rtm);
2252 read_unlock(&mrt_lock);
2253 rcu_read_unlock();
2254 return err;
2255}
2256
2257static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2258 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2259 int flags)
2260{
2261 struct nlmsghdr *nlh;
2262 struct rtmsg *rtm;
2263 int err;
2264
2265 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2266 if (nlh == NULL)
2267 return -EMSGSIZE;
2268
2269 rtm = nlmsg_data(nlh);
2270 rtm->rtm_family = RTNL_FAMILY_IPMR;
2271 rtm->rtm_dst_len = 32;
2272 rtm->rtm_src_len = 32;
2273 rtm->rtm_tos = 0;
2274 rtm->rtm_table = mrt->id;
2275 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2276 goto nla_put_failure;
2277 rtm->rtm_type = RTN_MULTICAST;
2278 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2279 if (c->mfc_flags & MFC_STATIC)
2280 rtm->rtm_protocol = RTPROT_STATIC;
2281 else
2282 rtm->rtm_protocol = RTPROT_MROUTED;
2283 rtm->rtm_flags = 0;
2284
2285 if (nla_put_be32(skb, RTA_SRC, c->mfc_origin) ||
2286 nla_put_be32(skb, RTA_DST, c->mfc_mcastgrp))
2287 goto nla_put_failure;
2288 err = __ipmr_fill_mroute(mrt, skb, c, rtm);
2289 /* do not break the dump if cache is unresolved */
2290 if (err < 0 && err != -ENOENT)
2291 goto nla_put_failure;
2292
2293 return nlmsg_end(skb, nlh);
2294
2295nla_put_failure:
2296 nlmsg_cancel(skb, nlh);
2297 return -EMSGSIZE;
2298}
2299
2300static size_t mroute_msgsize(bool unresolved, int maxvif)
2301{
2302 size_t len =
2303 NLMSG_ALIGN(sizeof(struct rtmsg))
2304 + nla_total_size(4) /* RTA_TABLE */
2305 + nla_total_size(4) /* RTA_SRC */
2306 + nla_total_size(4) /* RTA_DST */
2307 ;
2308
2309 if (!unresolved)
2310 len = len
2311 + nla_total_size(4) /* RTA_IIF */
2312 + nla_total_size(0) /* RTA_MULTIPATH */
2313 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2314 /* RTA_MFC_STATS */
2315 + nla_total_size(sizeof(struct rta_mfc_stats))
2316 ;
2317
2318 return len;
2319}
2320
2321static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2322 int cmd)
2323{
2324 struct net *net = read_pnet(&mrt->net);
2325 struct sk_buff *skb;
2326 int err = -ENOBUFS;
2327
2328 skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif),
2329 GFP_ATOMIC);
2330 if (skb == NULL)
2331 goto errout;
2332
2333 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2334 if (err < 0)
2335 goto errout;
2336
2337 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2338 return;
2339
2340errout:
2341 kfree_skb(skb);
2342 if (err < 0)
2343 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2344}
2345
2346static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2347{
2348 struct net *net = sock_net(skb->sk);
2349 struct mr_table *mrt;
2350 struct mfc_cache *mfc;
2351 unsigned int t = 0, s_t;
2352 unsigned int h = 0, s_h;
2353 unsigned int e = 0, s_e;
2354
2355 s_t = cb->args[0];
2356 s_h = cb->args[1];
2357 s_e = cb->args[2];
2358
2359 rcu_read_lock();
2360 ipmr_for_each_table(mrt, net) {
2361 if (t < s_t)
2362 goto next_table;
2363 if (t > s_t)
2364 s_h = 0;
2365 for (h = s_h; h < MFC_LINES; h++) {
2366 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_array[h], list) {
2367 if (e < s_e)
2368 goto next_entry;
2369 if (ipmr_fill_mroute(mrt, skb,
2370 NETLINK_CB(cb->skb).portid,
2371 cb->nlh->nlmsg_seq,
2372 mfc, RTM_NEWROUTE,
2373 NLM_F_MULTI) < 0)
2374 goto done;
2375next_entry:
2376 e++;
2377 }
2378 e = s_e = 0;
2379 }
2380 spin_lock_bh(&mfc_unres_lock);
2381 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
2382 if (e < s_e)
2383 goto next_entry2;
2384 if (ipmr_fill_mroute(mrt, skb,
2385 NETLINK_CB(cb->skb).portid,
2386 cb->nlh->nlmsg_seq,
2387 mfc, RTM_NEWROUTE,
2388 NLM_F_MULTI) < 0) {
2389 spin_unlock_bh(&mfc_unres_lock);
2390 goto done;
2391 }
2392next_entry2:
2393 e++;
2394 }
2395 spin_unlock_bh(&mfc_unres_lock);
2396 e = s_e = 0;
2397 s_h = 0;
2398next_table:
2399 t++;
2400 }
2401done:
2402 rcu_read_unlock();
2403
2404 cb->args[2] = e;
2405 cb->args[1] = h;
2406 cb->args[0] = t;
2407
2408 return skb->len;
2409}
2410
2411#ifdef CONFIG_PROC_FS
2412/*
2413 * The /proc interfaces to multicast routing :
2414 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2415 */
2416struct ipmr_vif_iter {
2417 struct seq_net_private p;
2418 struct mr_table *mrt;
2419 int ct;
2420};
2421
2422static struct vif_device *ipmr_vif_seq_idx(struct net *net,
2423 struct ipmr_vif_iter *iter,
2424 loff_t pos)
2425{
2426 struct mr_table *mrt = iter->mrt;
2427
2428 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
2429 if (!VIF_EXISTS(mrt, iter->ct))
2430 continue;
2431 if (pos-- == 0)
2432 return &mrt->vif_table[iter->ct];
2433 }
2434 return NULL;
2435}
2436
2437static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2438 __acquires(mrt_lock)
2439{
2440 struct ipmr_vif_iter *iter = seq->private;
2441 struct net *net = seq_file_net(seq);
2442 struct mr_table *mrt;
2443
2444 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2445 if (mrt == NULL)
2446 return ERR_PTR(-ENOENT);
2447
2448 iter->mrt = mrt;
2449
2450 read_lock(&mrt_lock);
2451 return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
2452 : SEQ_START_TOKEN;
2453}
2454
2455static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2456{
2457 struct ipmr_vif_iter *iter = seq->private;
2458 struct net *net = seq_file_net(seq);
2459 struct mr_table *mrt = iter->mrt;
2460
2461 ++*pos;
2462 if (v == SEQ_START_TOKEN)
2463 return ipmr_vif_seq_idx(net, iter, 0);
2464
2465 while (++iter->ct < mrt->maxvif) {
2466 if (!VIF_EXISTS(mrt, iter->ct))
2467 continue;
2468 return &mrt->vif_table[iter->ct];
2469 }
2470 return NULL;
2471}
2472
2473static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2474 __releases(mrt_lock)
2475{
2476 read_unlock(&mrt_lock);
2477}
2478
2479static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2480{
2481 struct ipmr_vif_iter *iter = seq->private;
2482 struct mr_table *mrt = iter->mrt;
2483
2484 if (v == SEQ_START_TOKEN) {
2485 seq_puts(seq,
2486 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2487 } else {
2488 const struct vif_device *vif = v;
2489 const char *name = vif->dev ? vif->dev->name : "none";
2490
2491 seq_printf(seq,
2492 "%2Zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2493 vif - mrt->vif_table,
2494 name, vif->bytes_in, vif->pkt_in,
2495 vif->bytes_out, vif->pkt_out,
2496 vif->flags, vif->local, vif->remote);
2497 }
2498 return 0;
2499}
2500
2501static const struct seq_operations ipmr_vif_seq_ops = {
2502 .start = ipmr_vif_seq_start,
2503 .next = ipmr_vif_seq_next,
2504 .stop = ipmr_vif_seq_stop,
2505 .show = ipmr_vif_seq_show,
2506};
2507
2508static int ipmr_vif_open(struct inode *inode, struct file *file)
2509{
2510 return seq_open_net(inode, file, &ipmr_vif_seq_ops,
2511 sizeof(struct ipmr_vif_iter));
2512}
2513
2514static const struct file_operations ipmr_vif_fops = {
2515 .owner = THIS_MODULE,
2516 .open = ipmr_vif_open,
2517 .read = seq_read,
2518 .llseek = seq_lseek,
2519 .release = seq_release_net,
2520};
2521
2522struct ipmr_mfc_iter {
2523 struct seq_net_private p;
2524 struct mr_table *mrt;
2525 struct list_head *cache;
2526 int ct;
2527};
2528
2529
2530static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
2531 struct ipmr_mfc_iter *it, loff_t pos)
2532{
2533 struct mr_table *mrt = it->mrt;
2534 struct mfc_cache *mfc;
2535
2536 rcu_read_lock();
2537 for (it->ct = 0; it->ct < MFC_LINES; it->ct++) {
2538 it->cache = &mrt->mfc_cache_array[it->ct];
2539 list_for_each_entry_rcu(mfc, it->cache, list)
2540 if (pos-- == 0)
2541 return mfc;
2542 }
2543 rcu_read_unlock();
2544
2545 spin_lock_bh(&mfc_unres_lock);
2546 it->cache = &mrt->mfc_unres_queue;
2547 list_for_each_entry(mfc, it->cache, list)
2548 if (pos-- == 0)
2549 return mfc;
2550 spin_unlock_bh(&mfc_unres_lock);
2551
2552 it->cache = NULL;
2553 return NULL;
2554}
2555
2556
2557static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2558{
2559 struct ipmr_mfc_iter *it = seq->private;
2560 struct net *net = seq_file_net(seq);
2561 struct mr_table *mrt;
2562
2563 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2564 if (mrt == NULL)
2565 return ERR_PTR(-ENOENT);
2566
2567 it->mrt = mrt;
2568 it->cache = NULL;
2569 it->ct = 0;
2570 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
2571 : SEQ_START_TOKEN;
2572}
2573
2574static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2575{
2576 struct mfc_cache *mfc = v;
2577 struct ipmr_mfc_iter *it = seq->private;
2578 struct net *net = seq_file_net(seq);
2579 struct mr_table *mrt = it->mrt;
2580
2581 ++*pos;
2582
2583 if (v == SEQ_START_TOKEN)
2584 return ipmr_mfc_seq_idx(net, seq->private, 0);
2585
2586 if (mfc->list.next != it->cache)
2587 return list_entry(mfc->list.next, struct mfc_cache, list);
2588
2589 if (it->cache == &mrt->mfc_unres_queue)
2590 goto end_of_list;
2591
2592 BUG_ON(it->cache != &mrt->mfc_cache_array[it->ct]);
2593
2594 while (++it->ct < MFC_LINES) {
2595 it->cache = &mrt->mfc_cache_array[it->ct];
2596 if (list_empty(it->cache))
2597 continue;
2598 return list_first_entry(it->cache, struct mfc_cache, list);
2599 }
2600
2601 /* exhausted cache_array, show unresolved */
2602 rcu_read_unlock();
2603 it->cache = &mrt->mfc_unres_queue;
2604 it->ct = 0;
2605
2606 spin_lock_bh(&mfc_unres_lock);
2607 if (!list_empty(it->cache))
2608 return list_first_entry(it->cache, struct mfc_cache, list);
2609
2610end_of_list:
2611 spin_unlock_bh(&mfc_unres_lock);
2612 it->cache = NULL;
2613
2614 return NULL;
2615}
2616
2617static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
2618{
2619 struct ipmr_mfc_iter *it = seq->private;
2620 struct mr_table *mrt = it->mrt;
2621
2622 if (it->cache == &mrt->mfc_unres_queue)
2623 spin_unlock_bh(&mfc_unres_lock);
2624 else if (it->cache == &mrt->mfc_cache_array[it->ct])
2625 rcu_read_unlock();
2626}
2627
2628static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2629{
2630 int n;
2631
2632 if (v == SEQ_START_TOKEN) {
2633 seq_puts(seq,
2634 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2635 } else {
2636 const struct mfc_cache *mfc = v;
2637 const struct ipmr_mfc_iter *it = seq->private;
2638 const struct mr_table *mrt = it->mrt;
2639
2640 seq_printf(seq, "%08X %08X %-3hd",
2641 (__force u32) mfc->mfc_mcastgrp,
2642 (__force u32) mfc->mfc_origin,
2643 mfc->mfc_parent);
2644
2645 if (it->cache != &mrt->mfc_unres_queue) {
2646 seq_printf(seq, " %8lu %8lu %8lu",
2647 mfc->mfc_un.res.pkt,
2648 mfc->mfc_un.res.bytes,
2649 mfc->mfc_un.res.wrong_if);
2650 for (n = mfc->mfc_un.res.minvif;
2651 n < mfc->mfc_un.res.maxvif; n++) {
2652 if (VIF_EXISTS(mrt, n) &&
2653 mfc->mfc_un.res.ttls[n] < 255)
2654 seq_printf(seq,
2655 " %2d:%-3d",
2656 n, mfc->mfc_un.res.ttls[n]);
2657 }
2658 } else {
2659 /* unresolved mfc_caches don't contain
2660 * pkt, bytes and wrong_if values
2661 */
2662 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2663 }
2664 seq_putc(seq, '\n');
2665 }
2666 return 0;
2667}
2668
2669static const struct seq_operations ipmr_mfc_seq_ops = {
2670 .start = ipmr_mfc_seq_start,
2671 .next = ipmr_mfc_seq_next,
2672 .stop = ipmr_mfc_seq_stop,
2673 .show = ipmr_mfc_seq_show,
2674};
2675
2676static int ipmr_mfc_open(struct inode *inode, struct file *file)
2677{
2678 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
2679 sizeof(struct ipmr_mfc_iter));
2680}
2681
2682static const struct file_operations ipmr_mfc_fops = {
2683 .owner = THIS_MODULE,
2684 .open = ipmr_mfc_open,
2685 .read = seq_read,
2686 .llseek = seq_lseek,
2687 .release = seq_release_net,
2688};
2689#endif
2690
2691#ifdef CONFIG_IP_PIMSM_V2
2692static const struct net_protocol pim_protocol = {
2693 .handler = pim_rcv,
2694 .netns_ok = 1,
2695};
2696#endif
2697
2698
2699/*
2700 * Setup for IP multicast routing
2701 */
2702static int __net_init ipmr_net_init(struct net *net)
2703{
2704 int err;
2705
2706 err = ipmr_rules_init(net);
2707 if (err < 0)
2708 goto fail;
2709
2710#ifdef CONFIG_PROC_FS
2711 err = -ENOMEM;
2712 if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops))
2713 goto proc_vif_fail;
2714 if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops))
2715 goto proc_cache_fail;
2716#endif
2717 return 0;
2718
2719#ifdef CONFIG_PROC_FS
2720proc_cache_fail:
2721 remove_proc_entry("ip_mr_vif", net->proc_net);
2722proc_vif_fail:
2723 ipmr_rules_exit(net);
2724#endif
2725fail:
2726 return err;
2727}
2728
2729static void __net_exit ipmr_net_exit(struct net *net)
2730{
2731#ifdef CONFIG_PROC_FS
2732 remove_proc_entry("ip_mr_cache", net->proc_net);
2733 remove_proc_entry("ip_mr_vif", net->proc_net);
2734#endif
2735 ipmr_rules_exit(net);
2736}
2737
2738static struct pernet_operations ipmr_net_ops = {
2739 .init = ipmr_net_init,
2740 .exit = ipmr_net_exit,
2741};
2742
2743int __init ip_mr_init(void)
2744{
2745 int err;
2746
2747 mrt_cachep = kmem_cache_create("ip_mrt_cache",
2748 sizeof(struct mfc_cache),
2749 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
2750 NULL);
2751 if (!mrt_cachep)
2752 return -ENOMEM;
2753
2754 err = register_pernet_subsys(&ipmr_net_ops);
2755 if (err)
2756 goto reg_pernet_fail;
2757
2758 err = register_netdevice_notifier(&ip_mr_notifier);
2759 if (err)
2760 goto reg_notif_fail;
2761#ifdef CONFIG_IP_PIMSM_V2
2762 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
2763 pr_err("%s: can't add PIM protocol\n", __func__);
2764 err = -EAGAIN;
2765 goto add_proto_fail;
2766 }
2767#endif
2768 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
2769 NULL, ipmr_rtm_dumproute, NULL);
2770 return 0;
2771
2772#ifdef CONFIG_IP_PIMSM_V2
2773add_proto_fail:
2774 unregister_netdevice_notifier(&ip_mr_notifier);
2775#endif
2776reg_notif_fail:
2777 unregister_pernet_subsys(&ipmr_net_ops);
2778reg_pernet_fail:
2779 kmem_cache_destroy(mrt_cachep);
2780 return err;
2781}