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// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * IP multicast routing support for mrouted 3.6/3.8
4 *
5 * (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
6 * Linux Consultancy and Custom Driver Development
7 *
8 * Fixes:
9 * Michael Chastain : Incorrect size of copying.
10 * Alan Cox : Added the cache manager code
11 * Alan Cox : Fixed the clone/copy bug and device race.
12 * Mike McLagan : Routing by source
13 * Malcolm Beattie : Buffer handling fixes.
14 * Alexey Kuznetsov : Double buffer free and other fixes.
15 * SVR Anand : Fixed several multicast bugs and problems.
16 * Alexey Kuznetsov : Status, optimisations and more.
17 * Brad Parker : Better behaviour on mrouted upcall
18 * overflow.
19 * Carlos Picoto : PIMv1 Support
20 * Pavlin Ivanov Radoslavov: PIMv2 Registers must checksum only PIM header
21 * Relax this requirement to work with older peers.
22 */
23
24#include <linux/uaccess.h>
25#include <linux/types.h>
26#include <linux/cache.h>
27#include <linux/capability.h>
28#include <linux/errno.h>
29#include <linux/mm.h>
30#include <linux/kernel.h>
31#include <linux/fcntl.h>
32#include <linux/stat.h>
33#include <linux/socket.h>
34#include <linux/in.h>
35#include <linux/inet.h>
36#include <linux/netdevice.h>
37#include <linux/inetdevice.h>
38#include <linux/igmp.h>
39#include <linux/proc_fs.h>
40#include <linux/seq_file.h>
41#include <linux/mroute.h>
42#include <linux/init.h>
43#include <linux/if_ether.h>
44#include <linux/slab.h>
45#include <net/net_namespace.h>
46#include <net/ip.h>
47#include <net/protocol.h>
48#include <linux/skbuff.h>
49#include <net/route.h>
50#include <net/icmp.h>
51#include <net/udp.h>
52#include <net/raw.h>
53#include <linux/notifier.h>
54#include <linux/if_arp.h>
55#include <linux/netfilter_ipv4.h>
56#include <linux/compat.h>
57#include <linux/export.h>
58#include <linux/rhashtable.h>
59#include <net/ip_tunnels.h>
60#include <net/checksum.h>
61#include <net/netlink.h>
62#include <net/fib_rules.h>
63#include <linux/netconf.h>
64#include <net/rtnh.h>
65
66#include <linux/nospec.h>
67
68struct ipmr_rule {
69 struct fib_rule common;
70};
71
72struct ipmr_result {
73 struct mr_table *mrt;
74};
75
76/* Big lock, protecting vif table, mrt cache and mroute socket state.
77 * Note that the changes are semaphored via rtnl_lock.
78 */
79
80static DEFINE_SPINLOCK(mrt_lock);
81
82static struct net_device *vif_dev_read(const struct vif_device *vif)
83{
84 return rcu_dereference(vif->dev);
85}
86
87/* Multicast router control variables */
88
89/* Special spinlock for queue of unresolved entries */
90static DEFINE_SPINLOCK(mfc_unres_lock);
91
92/* We return to original Alan's scheme. Hash table of resolved
93 * entries is changed only in process context and protected
94 * with weak lock mrt_lock. Queue of unresolved entries is protected
95 * with strong spinlock mfc_unres_lock.
96 *
97 * In this case data path is free of exclusive locks at all.
98 */
99
100static struct kmem_cache *mrt_cachep __ro_after_init;
101
102static struct mr_table *ipmr_new_table(struct net *net, u32 id);
103static void ipmr_free_table(struct mr_table *mrt);
104
105static void ip_mr_forward(struct net *net, struct mr_table *mrt,
106 struct net_device *dev, struct sk_buff *skb,
107 struct mfc_cache *cache, int local);
108static int ipmr_cache_report(const struct mr_table *mrt,
109 struct sk_buff *pkt, vifi_t vifi, int assert);
110static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
111 int cmd);
112static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt);
113static void mroute_clean_tables(struct mr_table *mrt, int flags);
114static void ipmr_expire_process(struct timer_list *t);
115
116#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
117#define ipmr_for_each_table(mrt, net) \
118 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list, \
119 lockdep_rtnl_is_held() || \
120 list_empty(&net->ipv4.mr_tables))
121
122static struct mr_table *ipmr_mr_table_iter(struct net *net,
123 struct mr_table *mrt)
124{
125 struct mr_table *ret;
126
127 if (!mrt)
128 ret = list_entry_rcu(net->ipv4.mr_tables.next,
129 struct mr_table, list);
130 else
131 ret = list_entry_rcu(mrt->list.next,
132 struct mr_table, list);
133
134 if (&ret->list == &net->ipv4.mr_tables)
135 return NULL;
136 return ret;
137}
138
139static struct mr_table *ipmr_get_table(struct net *net, u32 id)
140{
141 struct mr_table *mrt;
142
143 ipmr_for_each_table(mrt, net) {
144 if (mrt->id == id)
145 return mrt;
146 }
147 return NULL;
148}
149
150static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
151 struct mr_table **mrt)
152{
153 int err;
154 struct ipmr_result res;
155 struct fib_lookup_arg arg = {
156 .result = &res,
157 .flags = FIB_LOOKUP_NOREF,
158 };
159
160 /* update flow if oif or iif point to device enslaved to l3mdev */
161 l3mdev_update_flow(net, flowi4_to_flowi(flp4));
162
163 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
164 flowi4_to_flowi(flp4), 0, &arg);
165 if (err < 0)
166 return err;
167 *mrt = res.mrt;
168 return 0;
169}
170
171static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
172 int flags, struct fib_lookup_arg *arg)
173{
174 struct ipmr_result *res = arg->result;
175 struct mr_table *mrt;
176
177 switch (rule->action) {
178 case FR_ACT_TO_TBL:
179 break;
180 case FR_ACT_UNREACHABLE:
181 return -ENETUNREACH;
182 case FR_ACT_PROHIBIT:
183 return -EACCES;
184 case FR_ACT_BLACKHOLE:
185 default:
186 return -EINVAL;
187 }
188
189 arg->table = fib_rule_get_table(rule, arg);
190
191 mrt = ipmr_get_table(rule->fr_net, arg->table);
192 if (!mrt)
193 return -EAGAIN;
194 res->mrt = mrt;
195 return 0;
196}
197
198static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
199{
200 return 1;
201}
202
203static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
204 struct fib_rule_hdr *frh, struct nlattr **tb,
205 struct netlink_ext_ack *extack)
206{
207 return 0;
208}
209
210static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
211 struct nlattr **tb)
212{
213 return 1;
214}
215
216static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
217 struct fib_rule_hdr *frh)
218{
219 frh->dst_len = 0;
220 frh->src_len = 0;
221 frh->tos = 0;
222 return 0;
223}
224
225static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
226 .family = RTNL_FAMILY_IPMR,
227 .rule_size = sizeof(struct ipmr_rule),
228 .addr_size = sizeof(u32),
229 .action = ipmr_rule_action,
230 .match = ipmr_rule_match,
231 .configure = ipmr_rule_configure,
232 .compare = ipmr_rule_compare,
233 .fill = ipmr_rule_fill,
234 .nlgroup = RTNLGRP_IPV4_RULE,
235 .owner = THIS_MODULE,
236};
237
238static int __net_init ipmr_rules_init(struct net *net)
239{
240 struct fib_rules_ops *ops;
241 struct mr_table *mrt;
242 int err;
243
244 ops = fib_rules_register(&ipmr_rules_ops_template, net);
245 if (IS_ERR(ops))
246 return PTR_ERR(ops);
247
248 INIT_LIST_HEAD(&net->ipv4.mr_tables);
249
250 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
251 if (IS_ERR(mrt)) {
252 err = PTR_ERR(mrt);
253 goto err1;
254 }
255
256 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT);
257 if (err < 0)
258 goto err2;
259
260 net->ipv4.mr_rules_ops = ops;
261 return 0;
262
263err2:
264 rtnl_lock();
265 ipmr_free_table(mrt);
266 rtnl_unlock();
267err1:
268 fib_rules_unregister(ops);
269 return err;
270}
271
272static void __net_exit ipmr_rules_exit(struct net *net)
273{
274 struct mr_table *mrt, *next;
275
276 ASSERT_RTNL();
277 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
278 list_del(&mrt->list);
279 ipmr_free_table(mrt);
280 }
281 fib_rules_unregister(net->ipv4.mr_rules_ops);
282}
283
284static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
285 struct netlink_ext_ack *extack)
286{
287 return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack);
288}
289
290static unsigned int ipmr_rules_seq_read(struct net *net)
291{
292 return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
293}
294
295bool ipmr_rule_default(const struct fib_rule *rule)
296{
297 return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
298}
299EXPORT_SYMBOL(ipmr_rule_default);
300#else
301#define ipmr_for_each_table(mrt, net) \
302 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
303
304static struct mr_table *ipmr_mr_table_iter(struct net *net,
305 struct mr_table *mrt)
306{
307 if (!mrt)
308 return net->ipv4.mrt;
309 return NULL;
310}
311
312static struct mr_table *ipmr_get_table(struct net *net, u32 id)
313{
314 return net->ipv4.mrt;
315}
316
317static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
318 struct mr_table **mrt)
319{
320 *mrt = net->ipv4.mrt;
321 return 0;
322}
323
324static int __net_init ipmr_rules_init(struct net *net)
325{
326 struct mr_table *mrt;
327
328 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
329 if (IS_ERR(mrt))
330 return PTR_ERR(mrt);
331 net->ipv4.mrt = mrt;
332 return 0;
333}
334
335static void __net_exit ipmr_rules_exit(struct net *net)
336{
337 ASSERT_RTNL();
338 ipmr_free_table(net->ipv4.mrt);
339 net->ipv4.mrt = NULL;
340}
341
342static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
343 struct netlink_ext_ack *extack)
344{
345 return 0;
346}
347
348static unsigned int ipmr_rules_seq_read(struct net *net)
349{
350 return 0;
351}
352
353bool ipmr_rule_default(const struct fib_rule *rule)
354{
355 return true;
356}
357EXPORT_SYMBOL(ipmr_rule_default);
358#endif
359
360static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
361 const void *ptr)
362{
363 const struct mfc_cache_cmp_arg *cmparg = arg->key;
364 const struct mfc_cache *c = ptr;
365
366 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
367 cmparg->mfc_origin != c->mfc_origin;
368}
369
370static const struct rhashtable_params ipmr_rht_params = {
371 .head_offset = offsetof(struct mr_mfc, mnode),
372 .key_offset = offsetof(struct mfc_cache, cmparg),
373 .key_len = sizeof(struct mfc_cache_cmp_arg),
374 .nelem_hint = 3,
375 .obj_cmpfn = ipmr_hash_cmp,
376 .automatic_shrinking = true,
377};
378
379static void ipmr_new_table_set(struct mr_table *mrt,
380 struct net *net)
381{
382#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
383 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
384#endif
385}
386
387static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
388 .mfc_mcastgrp = htonl(INADDR_ANY),
389 .mfc_origin = htonl(INADDR_ANY),
390};
391
392static struct mr_table_ops ipmr_mr_table_ops = {
393 .rht_params = &ipmr_rht_params,
394 .cmparg_any = &ipmr_mr_table_ops_cmparg_any,
395};
396
397static struct mr_table *ipmr_new_table(struct net *net, u32 id)
398{
399 struct mr_table *mrt;
400
401 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
402 if (id != RT_TABLE_DEFAULT && id >= 1000000000)
403 return ERR_PTR(-EINVAL);
404
405 mrt = ipmr_get_table(net, id);
406 if (mrt)
407 return mrt;
408
409 return mr_table_alloc(net, id, &ipmr_mr_table_ops,
410 ipmr_expire_process, ipmr_new_table_set);
411}
412
413static void ipmr_free_table(struct mr_table *mrt)
414{
415 timer_shutdown_sync(&mrt->ipmr_expire_timer);
416 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
417 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC);
418 rhltable_destroy(&mrt->mfc_hash);
419 kfree(mrt);
420}
421
422/* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
423
424/* Initialize ipmr pimreg/tunnel in_device */
425static bool ipmr_init_vif_indev(const struct net_device *dev)
426{
427 struct in_device *in_dev;
428
429 ASSERT_RTNL();
430
431 in_dev = __in_dev_get_rtnl(dev);
432 if (!in_dev)
433 return false;
434 ipv4_devconf_setall(in_dev);
435 neigh_parms_data_state_setall(in_dev->arp_parms);
436 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
437
438 return true;
439}
440
441static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
442{
443 struct net_device *tunnel_dev, *new_dev;
444 struct ip_tunnel_parm p = { };
445 int err;
446
447 tunnel_dev = __dev_get_by_name(net, "tunl0");
448 if (!tunnel_dev)
449 goto out;
450
451 p.iph.daddr = v->vifc_rmt_addr.s_addr;
452 p.iph.saddr = v->vifc_lcl_addr.s_addr;
453 p.iph.version = 4;
454 p.iph.ihl = 5;
455 p.iph.protocol = IPPROTO_IPIP;
456 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
457
458 if (!tunnel_dev->netdev_ops->ndo_tunnel_ctl)
459 goto out;
460 err = tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
461 SIOCADDTUNNEL);
462 if (err)
463 goto out;
464
465 new_dev = __dev_get_by_name(net, p.name);
466 if (!new_dev)
467 goto out;
468
469 new_dev->flags |= IFF_MULTICAST;
470 if (!ipmr_init_vif_indev(new_dev))
471 goto out_unregister;
472 if (dev_open(new_dev, NULL))
473 goto out_unregister;
474 dev_hold(new_dev);
475 err = dev_set_allmulti(new_dev, 1);
476 if (err) {
477 dev_close(new_dev);
478 tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
479 SIOCDELTUNNEL);
480 dev_put(new_dev);
481 new_dev = ERR_PTR(err);
482 }
483 return new_dev;
484
485out_unregister:
486 unregister_netdevice(new_dev);
487out:
488 return ERR_PTR(-ENOBUFS);
489}
490
491#if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
492static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
493{
494 struct net *net = dev_net(dev);
495 struct mr_table *mrt;
496 struct flowi4 fl4 = {
497 .flowi4_oif = dev->ifindex,
498 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
499 .flowi4_mark = skb->mark,
500 };
501 int err;
502
503 err = ipmr_fib_lookup(net, &fl4, &mrt);
504 if (err < 0) {
505 kfree_skb(skb);
506 return err;
507 }
508
509 DEV_STATS_ADD(dev, tx_bytes, skb->len);
510 DEV_STATS_INC(dev, tx_packets);
511 rcu_read_lock();
512
513 /* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
514 ipmr_cache_report(mrt, skb, READ_ONCE(mrt->mroute_reg_vif_num),
515 IGMPMSG_WHOLEPKT);
516
517 rcu_read_unlock();
518 kfree_skb(skb);
519 return NETDEV_TX_OK;
520}
521
522static int reg_vif_get_iflink(const struct net_device *dev)
523{
524 return 0;
525}
526
527static const struct net_device_ops reg_vif_netdev_ops = {
528 .ndo_start_xmit = reg_vif_xmit,
529 .ndo_get_iflink = reg_vif_get_iflink,
530};
531
532static void reg_vif_setup(struct net_device *dev)
533{
534 dev->type = ARPHRD_PIMREG;
535 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
536 dev->flags = IFF_NOARP;
537 dev->netdev_ops = ®_vif_netdev_ops;
538 dev->needs_free_netdev = true;
539 dev->features |= NETIF_F_NETNS_LOCAL;
540}
541
542static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
543{
544 struct net_device *dev;
545 char name[IFNAMSIZ];
546
547 if (mrt->id == RT_TABLE_DEFAULT)
548 sprintf(name, "pimreg");
549 else
550 sprintf(name, "pimreg%u", mrt->id);
551
552 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
553
554 if (!dev)
555 return NULL;
556
557 dev_net_set(dev, net);
558
559 if (register_netdevice(dev)) {
560 free_netdev(dev);
561 return NULL;
562 }
563
564 if (!ipmr_init_vif_indev(dev))
565 goto failure;
566 if (dev_open(dev, NULL))
567 goto failure;
568
569 dev_hold(dev);
570
571 return dev;
572
573failure:
574 unregister_netdevice(dev);
575 return NULL;
576}
577
578/* called with rcu_read_lock() */
579static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
580 unsigned int pimlen)
581{
582 struct net_device *reg_dev = NULL;
583 struct iphdr *encap;
584 int vif_num;
585
586 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
587 /* Check that:
588 * a. packet is really sent to a multicast group
589 * b. packet is not a NULL-REGISTER
590 * c. packet is not truncated
591 */
592 if (!ipv4_is_multicast(encap->daddr) ||
593 encap->tot_len == 0 ||
594 ntohs(encap->tot_len) + pimlen > skb->len)
595 return 1;
596
597 /* Pairs with WRITE_ONCE() in vif_add()/vid_delete() */
598 vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
599 if (vif_num >= 0)
600 reg_dev = vif_dev_read(&mrt->vif_table[vif_num]);
601 if (!reg_dev)
602 return 1;
603
604 skb->mac_header = skb->network_header;
605 skb_pull(skb, (u8 *)encap - skb->data);
606 skb_reset_network_header(skb);
607 skb->protocol = htons(ETH_P_IP);
608 skb->ip_summed = CHECKSUM_NONE;
609
610 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
611
612 netif_rx(skb);
613
614 return NET_RX_SUCCESS;
615}
616#else
617static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
618{
619 return NULL;
620}
621#endif
622
623static int call_ipmr_vif_entry_notifiers(struct net *net,
624 enum fib_event_type event_type,
625 struct vif_device *vif,
626 struct net_device *vif_dev,
627 vifi_t vif_index, u32 tb_id)
628{
629 return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
630 vif, vif_dev, vif_index, tb_id,
631 &net->ipv4.ipmr_seq);
632}
633
634static int call_ipmr_mfc_entry_notifiers(struct net *net,
635 enum fib_event_type event_type,
636 struct mfc_cache *mfc, u32 tb_id)
637{
638 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
639 &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
640}
641
642/**
643 * vif_delete - Delete a VIF entry
644 * @mrt: Table to delete from
645 * @vifi: VIF identifier to delete
646 * @notify: Set to 1, if the caller is a notifier_call
647 * @head: if unregistering the VIF, place it on this queue
648 */
649static int vif_delete(struct mr_table *mrt, int vifi, int notify,
650 struct list_head *head)
651{
652 struct net *net = read_pnet(&mrt->net);
653 struct vif_device *v;
654 struct net_device *dev;
655 struct in_device *in_dev;
656
657 if (vifi < 0 || vifi >= mrt->maxvif)
658 return -EADDRNOTAVAIL;
659
660 v = &mrt->vif_table[vifi];
661
662 dev = rtnl_dereference(v->dev);
663 if (!dev)
664 return -EADDRNOTAVAIL;
665
666 spin_lock(&mrt_lock);
667 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, dev,
668 vifi, mrt->id);
669 RCU_INIT_POINTER(v->dev, NULL);
670
671 if (vifi == mrt->mroute_reg_vif_num) {
672 /* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
673 WRITE_ONCE(mrt->mroute_reg_vif_num, -1);
674 }
675 if (vifi + 1 == mrt->maxvif) {
676 int tmp;
677
678 for (tmp = vifi - 1; tmp >= 0; tmp--) {
679 if (VIF_EXISTS(mrt, tmp))
680 break;
681 }
682 WRITE_ONCE(mrt->maxvif, tmp + 1);
683 }
684
685 spin_unlock(&mrt_lock);
686
687 dev_set_allmulti(dev, -1);
688
689 in_dev = __in_dev_get_rtnl(dev);
690 if (in_dev) {
691 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
692 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
693 NETCONFA_MC_FORWARDING,
694 dev->ifindex, &in_dev->cnf);
695 ip_rt_multicast_event(in_dev);
696 }
697
698 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
699 unregister_netdevice_queue(dev, head);
700
701 netdev_put(dev, &v->dev_tracker);
702 return 0;
703}
704
705static void ipmr_cache_free_rcu(struct rcu_head *head)
706{
707 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
708
709 kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
710}
711
712static void ipmr_cache_free(struct mfc_cache *c)
713{
714 call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
715}
716
717/* Destroy an unresolved cache entry, killing queued skbs
718 * and reporting error to netlink readers.
719 */
720static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
721{
722 struct net *net = read_pnet(&mrt->net);
723 struct sk_buff *skb;
724 struct nlmsgerr *e;
725
726 atomic_dec(&mrt->cache_resolve_queue_len);
727
728 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
729 if (ip_hdr(skb)->version == 0) {
730 struct nlmsghdr *nlh = skb_pull(skb,
731 sizeof(struct iphdr));
732 nlh->nlmsg_type = NLMSG_ERROR;
733 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
734 skb_trim(skb, nlh->nlmsg_len);
735 e = nlmsg_data(nlh);
736 e->error = -ETIMEDOUT;
737 memset(&e->msg, 0, sizeof(e->msg));
738
739 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
740 } else {
741 kfree_skb(skb);
742 }
743 }
744
745 ipmr_cache_free(c);
746}
747
748/* Timer process for the unresolved queue. */
749static void ipmr_expire_process(struct timer_list *t)
750{
751 struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
752 struct mr_mfc *c, *next;
753 unsigned long expires;
754 unsigned long now;
755
756 if (!spin_trylock(&mfc_unres_lock)) {
757 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
758 return;
759 }
760
761 if (list_empty(&mrt->mfc_unres_queue))
762 goto out;
763
764 now = jiffies;
765 expires = 10*HZ;
766
767 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
768 if (time_after(c->mfc_un.unres.expires, now)) {
769 unsigned long interval = c->mfc_un.unres.expires - now;
770 if (interval < expires)
771 expires = interval;
772 continue;
773 }
774
775 list_del(&c->list);
776 mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
777 ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
778 }
779
780 if (!list_empty(&mrt->mfc_unres_queue))
781 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
782
783out:
784 spin_unlock(&mfc_unres_lock);
785}
786
787/* Fill oifs list. It is called under locked mrt_lock. */
788static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
789 unsigned char *ttls)
790{
791 int vifi;
792
793 cache->mfc_un.res.minvif = MAXVIFS;
794 cache->mfc_un.res.maxvif = 0;
795 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
796
797 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
798 if (VIF_EXISTS(mrt, vifi) &&
799 ttls[vifi] && ttls[vifi] < 255) {
800 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
801 if (cache->mfc_un.res.minvif > vifi)
802 cache->mfc_un.res.minvif = vifi;
803 if (cache->mfc_un.res.maxvif <= vifi)
804 cache->mfc_un.res.maxvif = vifi + 1;
805 }
806 }
807 cache->mfc_un.res.lastuse = jiffies;
808}
809
810static int vif_add(struct net *net, struct mr_table *mrt,
811 struct vifctl *vifc, int mrtsock)
812{
813 struct netdev_phys_item_id ppid = { };
814 int vifi = vifc->vifc_vifi;
815 struct vif_device *v = &mrt->vif_table[vifi];
816 struct net_device *dev;
817 struct in_device *in_dev;
818 int err;
819
820 /* Is vif busy ? */
821 if (VIF_EXISTS(mrt, vifi))
822 return -EADDRINUSE;
823
824 switch (vifc->vifc_flags) {
825 case VIFF_REGISTER:
826 if (!ipmr_pimsm_enabled())
827 return -EINVAL;
828 /* Special Purpose VIF in PIM
829 * All the packets will be sent to the daemon
830 */
831 if (mrt->mroute_reg_vif_num >= 0)
832 return -EADDRINUSE;
833 dev = ipmr_reg_vif(net, mrt);
834 if (!dev)
835 return -ENOBUFS;
836 err = dev_set_allmulti(dev, 1);
837 if (err) {
838 unregister_netdevice(dev);
839 dev_put(dev);
840 return err;
841 }
842 break;
843 case VIFF_TUNNEL:
844 dev = ipmr_new_tunnel(net, vifc);
845 if (IS_ERR(dev))
846 return PTR_ERR(dev);
847 break;
848 case VIFF_USE_IFINDEX:
849 case 0:
850 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
851 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
852 if (dev && !__in_dev_get_rtnl(dev)) {
853 dev_put(dev);
854 return -EADDRNOTAVAIL;
855 }
856 } else {
857 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
858 }
859 if (!dev)
860 return -EADDRNOTAVAIL;
861 err = dev_set_allmulti(dev, 1);
862 if (err) {
863 dev_put(dev);
864 return err;
865 }
866 break;
867 default:
868 return -EINVAL;
869 }
870
871 in_dev = __in_dev_get_rtnl(dev);
872 if (!in_dev) {
873 dev_put(dev);
874 return -EADDRNOTAVAIL;
875 }
876 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
877 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
878 dev->ifindex, &in_dev->cnf);
879 ip_rt_multicast_event(in_dev);
880
881 /* Fill in the VIF structures */
882 vif_device_init(v, dev, vifc->vifc_rate_limit,
883 vifc->vifc_threshold,
884 vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
885 (VIFF_TUNNEL | VIFF_REGISTER));
886
887 err = dev_get_port_parent_id(dev, &ppid, true);
888 if (err == 0) {
889 memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len);
890 v->dev_parent_id.id_len = ppid.id_len;
891 } else {
892 v->dev_parent_id.id_len = 0;
893 }
894
895 v->local = vifc->vifc_lcl_addr.s_addr;
896 v->remote = vifc->vifc_rmt_addr.s_addr;
897
898 /* And finish update writing critical data */
899 spin_lock(&mrt_lock);
900 rcu_assign_pointer(v->dev, dev);
901 netdev_tracker_alloc(dev, &v->dev_tracker, GFP_ATOMIC);
902 if (v->flags & VIFF_REGISTER) {
903 /* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
904 WRITE_ONCE(mrt->mroute_reg_vif_num, vifi);
905 }
906 if (vifi+1 > mrt->maxvif)
907 WRITE_ONCE(mrt->maxvif, vifi + 1);
908 spin_unlock(&mrt_lock);
909 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, dev,
910 vifi, mrt->id);
911 return 0;
912}
913
914/* called with rcu_read_lock() */
915static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
916 __be32 origin,
917 __be32 mcastgrp)
918{
919 struct mfc_cache_cmp_arg arg = {
920 .mfc_mcastgrp = mcastgrp,
921 .mfc_origin = origin
922 };
923
924 return mr_mfc_find(mrt, &arg);
925}
926
927/* Look for a (*,G) entry */
928static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
929 __be32 mcastgrp, int vifi)
930{
931 struct mfc_cache_cmp_arg arg = {
932 .mfc_mcastgrp = mcastgrp,
933 .mfc_origin = htonl(INADDR_ANY)
934 };
935
936 if (mcastgrp == htonl(INADDR_ANY))
937 return mr_mfc_find_any_parent(mrt, vifi);
938 return mr_mfc_find_any(mrt, vifi, &arg);
939}
940
941/* Look for a (S,G,iif) entry if parent != -1 */
942static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
943 __be32 origin, __be32 mcastgrp,
944 int parent)
945{
946 struct mfc_cache_cmp_arg arg = {
947 .mfc_mcastgrp = mcastgrp,
948 .mfc_origin = origin,
949 };
950
951 return mr_mfc_find_parent(mrt, &arg, parent);
952}
953
954/* Allocate a multicast cache entry */
955static struct mfc_cache *ipmr_cache_alloc(void)
956{
957 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
958
959 if (c) {
960 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
961 c->_c.mfc_un.res.minvif = MAXVIFS;
962 c->_c.free = ipmr_cache_free_rcu;
963 refcount_set(&c->_c.mfc_un.res.refcount, 1);
964 }
965 return c;
966}
967
968static struct mfc_cache *ipmr_cache_alloc_unres(void)
969{
970 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
971
972 if (c) {
973 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
974 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
975 }
976 return c;
977}
978
979/* A cache entry has gone into a resolved state from queued */
980static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
981 struct mfc_cache *uc, struct mfc_cache *c)
982{
983 struct sk_buff *skb;
984 struct nlmsgerr *e;
985
986 /* Play the pending entries through our router */
987 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
988 if (ip_hdr(skb)->version == 0) {
989 struct nlmsghdr *nlh = skb_pull(skb,
990 sizeof(struct iphdr));
991
992 if (mr_fill_mroute(mrt, skb, &c->_c,
993 nlmsg_data(nlh)) > 0) {
994 nlh->nlmsg_len = skb_tail_pointer(skb) -
995 (u8 *)nlh;
996 } else {
997 nlh->nlmsg_type = NLMSG_ERROR;
998 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
999 skb_trim(skb, nlh->nlmsg_len);
1000 e = nlmsg_data(nlh);
1001 e->error = -EMSGSIZE;
1002 memset(&e->msg, 0, sizeof(e->msg));
1003 }
1004
1005 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1006 } else {
1007 rcu_read_lock();
1008 ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1009 rcu_read_unlock();
1010 }
1011 }
1012}
1013
1014/* Bounce a cache query up to mrouted and netlink.
1015 *
1016 * Called under rcu_read_lock().
1017 */
1018static int ipmr_cache_report(const struct mr_table *mrt,
1019 struct sk_buff *pkt, vifi_t vifi, int assert)
1020{
1021 const int ihl = ip_hdrlen(pkt);
1022 struct sock *mroute_sk;
1023 struct igmphdr *igmp;
1024 struct igmpmsg *msg;
1025 struct sk_buff *skb;
1026 int ret;
1027
1028 mroute_sk = rcu_dereference(mrt->mroute_sk);
1029 if (!mroute_sk)
1030 return -EINVAL;
1031
1032 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1033 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1034 else
1035 skb = alloc_skb(128, GFP_ATOMIC);
1036
1037 if (!skb)
1038 return -ENOBUFS;
1039
1040 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1041 /* Ugly, but we have no choice with this interface.
1042 * Duplicate old header, fix ihl, length etc.
1043 * And all this only to mangle msg->im_msgtype and
1044 * to set msg->im_mbz to "mbz" :-)
1045 */
1046 skb_push(skb, sizeof(struct iphdr));
1047 skb_reset_network_header(skb);
1048 skb_reset_transport_header(skb);
1049 msg = (struct igmpmsg *)skb_network_header(skb);
1050 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1051 msg->im_msgtype = assert;
1052 msg->im_mbz = 0;
1053 if (assert == IGMPMSG_WRVIFWHOLE) {
1054 msg->im_vif = vifi;
1055 msg->im_vif_hi = vifi >> 8;
1056 } else {
1057 /* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
1058 int vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
1059
1060 msg->im_vif = vif_num;
1061 msg->im_vif_hi = vif_num >> 8;
1062 }
1063 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1064 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1065 sizeof(struct iphdr));
1066 } else {
1067 /* Copy the IP header */
1068 skb_set_network_header(skb, skb->len);
1069 skb_put(skb, ihl);
1070 skb_copy_to_linear_data(skb, pkt->data, ihl);
1071 /* Flag to the kernel this is a route add */
1072 ip_hdr(skb)->protocol = 0;
1073 msg = (struct igmpmsg *)skb_network_header(skb);
1074 msg->im_vif = vifi;
1075 msg->im_vif_hi = vifi >> 8;
1076 ipv4_pktinfo_prepare(mroute_sk, pkt, false);
1077 memcpy(skb->cb, pkt->cb, sizeof(skb->cb));
1078 /* Add our header */
1079 igmp = skb_put(skb, sizeof(struct igmphdr));
1080 igmp->type = assert;
1081 msg->im_msgtype = assert;
1082 igmp->code = 0;
1083 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1084 skb->transport_header = skb->network_header;
1085 }
1086
1087 igmpmsg_netlink_event(mrt, skb);
1088
1089 /* Deliver to mrouted */
1090 ret = sock_queue_rcv_skb(mroute_sk, skb);
1091
1092 if (ret < 0) {
1093 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1094 kfree_skb(skb);
1095 }
1096
1097 return ret;
1098}
1099
1100/* Queue a packet for resolution. It gets locked cache entry! */
1101/* Called under rcu_read_lock() */
1102static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1103 struct sk_buff *skb, struct net_device *dev)
1104{
1105 const struct iphdr *iph = ip_hdr(skb);
1106 struct mfc_cache *c;
1107 bool found = false;
1108 int err;
1109
1110 spin_lock_bh(&mfc_unres_lock);
1111 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1112 if (c->mfc_mcastgrp == iph->daddr &&
1113 c->mfc_origin == iph->saddr) {
1114 found = true;
1115 break;
1116 }
1117 }
1118
1119 if (!found) {
1120 /* Create a new entry if allowable */
1121 c = ipmr_cache_alloc_unres();
1122 if (!c) {
1123 spin_unlock_bh(&mfc_unres_lock);
1124
1125 kfree_skb(skb);
1126 return -ENOBUFS;
1127 }
1128
1129 /* Fill in the new cache entry */
1130 c->_c.mfc_parent = -1;
1131 c->mfc_origin = iph->saddr;
1132 c->mfc_mcastgrp = iph->daddr;
1133
1134 /* Reflect first query at mrouted. */
1135 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1136
1137 if (err < 0) {
1138 /* If the report failed throw the cache entry
1139 out - Brad Parker
1140 */
1141 spin_unlock_bh(&mfc_unres_lock);
1142
1143 ipmr_cache_free(c);
1144 kfree_skb(skb);
1145 return err;
1146 }
1147
1148 atomic_inc(&mrt->cache_resolve_queue_len);
1149 list_add(&c->_c.list, &mrt->mfc_unres_queue);
1150 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1151
1152 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1153 mod_timer(&mrt->ipmr_expire_timer,
1154 c->_c.mfc_un.unres.expires);
1155 }
1156
1157 /* See if we can append the packet */
1158 if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1159 kfree_skb(skb);
1160 err = -ENOBUFS;
1161 } else {
1162 if (dev) {
1163 skb->dev = dev;
1164 skb->skb_iif = dev->ifindex;
1165 }
1166 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1167 err = 0;
1168 }
1169
1170 spin_unlock_bh(&mfc_unres_lock);
1171 return err;
1172}
1173
1174/* MFC cache manipulation by user space mroute daemon */
1175
1176static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1177{
1178 struct net *net = read_pnet(&mrt->net);
1179 struct mfc_cache *c;
1180
1181 /* The entries are added/deleted only under RTNL */
1182 rcu_read_lock();
1183 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1184 mfc->mfcc_mcastgrp.s_addr, parent);
1185 rcu_read_unlock();
1186 if (!c)
1187 return -ENOENT;
1188 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1189 list_del_rcu(&c->_c.list);
1190 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1191 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1192 mr_cache_put(&c->_c);
1193
1194 return 0;
1195}
1196
1197static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1198 struct mfcctl *mfc, int mrtsock, int parent)
1199{
1200 struct mfc_cache *uc, *c;
1201 struct mr_mfc *_uc;
1202 bool found;
1203 int ret;
1204
1205 if (mfc->mfcc_parent >= MAXVIFS)
1206 return -ENFILE;
1207
1208 /* The entries are added/deleted only under RTNL */
1209 rcu_read_lock();
1210 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1211 mfc->mfcc_mcastgrp.s_addr, parent);
1212 rcu_read_unlock();
1213 if (c) {
1214 spin_lock(&mrt_lock);
1215 c->_c.mfc_parent = mfc->mfcc_parent;
1216 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1217 if (!mrtsock)
1218 c->_c.mfc_flags |= MFC_STATIC;
1219 spin_unlock(&mrt_lock);
1220 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1221 mrt->id);
1222 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1223 return 0;
1224 }
1225
1226 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1227 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1228 return -EINVAL;
1229
1230 c = ipmr_cache_alloc();
1231 if (!c)
1232 return -ENOMEM;
1233
1234 c->mfc_origin = mfc->mfcc_origin.s_addr;
1235 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1236 c->_c.mfc_parent = mfc->mfcc_parent;
1237 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1238 if (!mrtsock)
1239 c->_c.mfc_flags |= MFC_STATIC;
1240
1241 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1242 ipmr_rht_params);
1243 if (ret) {
1244 pr_err("ipmr: rhtable insert error %d\n", ret);
1245 ipmr_cache_free(c);
1246 return ret;
1247 }
1248 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1249 /* Check to see if we resolved a queued list. If so we
1250 * need to send on the frames and tidy up.
1251 */
1252 found = false;
1253 spin_lock_bh(&mfc_unres_lock);
1254 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1255 uc = (struct mfc_cache *)_uc;
1256 if (uc->mfc_origin == c->mfc_origin &&
1257 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1258 list_del(&_uc->list);
1259 atomic_dec(&mrt->cache_resolve_queue_len);
1260 found = true;
1261 break;
1262 }
1263 }
1264 if (list_empty(&mrt->mfc_unres_queue))
1265 del_timer(&mrt->ipmr_expire_timer);
1266 spin_unlock_bh(&mfc_unres_lock);
1267
1268 if (found) {
1269 ipmr_cache_resolve(net, mrt, uc, c);
1270 ipmr_cache_free(uc);
1271 }
1272 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1273 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1274 return 0;
1275}
1276
1277/* Close the multicast socket, and clear the vif tables etc */
1278static void mroute_clean_tables(struct mr_table *mrt, int flags)
1279{
1280 struct net *net = read_pnet(&mrt->net);
1281 struct mr_mfc *c, *tmp;
1282 struct mfc_cache *cache;
1283 LIST_HEAD(list);
1284 int i;
1285
1286 /* Shut down all active vif entries */
1287 if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1288 for (i = 0; i < mrt->maxvif; i++) {
1289 if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1290 !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1291 (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1292 continue;
1293 vif_delete(mrt, i, 0, &list);
1294 }
1295 unregister_netdevice_many(&list);
1296 }
1297
1298 /* Wipe the cache */
1299 if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1300 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1301 if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1302 (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1303 continue;
1304 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1305 list_del_rcu(&c->list);
1306 cache = (struct mfc_cache *)c;
1307 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1308 mrt->id);
1309 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1310 mr_cache_put(c);
1311 }
1312 }
1313
1314 if (flags & MRT_FLUSH_MFC) {
1315 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1316 spin_lock_bh(&mfc_unres_lock);
1317 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1318 list_del(&c->list);
1319 cache = (struct mfc_cache *)c;
1320 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1321 ipmr_destroy_unres(mrt, cache);
1322 }
1323 spin_unlock_bh(&mfc_unres_lock);
1324 }
1325 }
1326}
1327
1328/* called from ip_ra_control(), before an RCU grace period,
1329 * we don't need to call synchronize_rcu() here
1330 */
1331static void mrtsock_destruct(struct sock *sk)
1332{
1333 struct net *net = sock_net(sk);
1334 struct mr_table *mrt;
1335
1336 rtnl_lock();
1337 ipmr_for_each_table(mrt, net) {
1338 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1339 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1340 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1341 NETCONFA_MC_FORWARDING,
1342 NETCONFA_IFINDEX_ALL,
1343 net->ipv4.devconf_all);
1344 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1345 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC);
1346 }
1347 }
1348 rtnl_unlock();
1349}
1350
1351/* Socket options and virtual interface manipulation. The whole
1352 * virtual interface system is a complete heap, but unfortunately
1353 * that's how BSD mrouted happens to think. Maybe one day with a proper
1354 * MOSPF/PIM router set up we can clean this up.
1355 */
1356
1357int ip_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval,
1358 unsigned int optlen)
1359{
1360 struct net *net = sock_net(sk);
1361 int val, ret = 0, parent = 0;
1362 struct mr_table *mrt;
1363 struct vifctl vif;
1364 struct mfcctl mfc;
1365 bool do_wrvifwhole;
1366 u32 uval;
1367
1368 /* There's one exception to the lock - MRT_DONE which needs to unlock */
1369 rtnl_lock();
1370 if (sk->sk_type != SOCK_RAW ||
1371 inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1372 ret = -EOPNOTSUPP;
1373 goto out_unlock;
1374 }
1375
1376 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1377 if (!mrt) {
1378 ret = -ENOENT;
1379 goto out_unlock;
1380 }
1381 if (optname != MRT_INIT) {
1382 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1383 !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1384 ret = -EACCES;
1385 goto out_unlock;
1386 }
1387 }
1388
1389 switch (optname) {
1390 case MRT_INIT:
1391 if (optlen != sizeof(int)) {
1392 ret = -EINVAL;
1393 break;
1394 }
1395 if (rtnl_dereference(mrt->mroute_sk)) {
1396 ret = -EADDRINUSE;
1397 break;
1398 }
1399
1400 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1401 if (ret == 0) {
1402 rcu_assign_pointer(mrt->mroute_sk, sk);
1403 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1404 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1405 NETCONFA_MC_FORWARDING,
1406 NETCONFA_IFINDEX_ALL,
1407 net->ipv4.devconf_all);
1408 }
1409 break;
1410 case MRT_DONE:
1411 if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1412 ret = -EACCES;
1413 } else {
1414 /* We need to unlock here because mrtsock_destruct takes
1415 * care of rtnl itself and we can't change that due to
1416 * the IP_ROUTER_ALERT setsockopt which runs without it.
1417 */
1418 rtnl_unlock();
1419 ret = ip_ra_control(sk, 0, NULL);
1420 goto out;
1421 }
1422 break;
1423 case MRT_ADD_VIF:
1424 case MRT_DEL_VIF:
1425 if (optlen != sizeof(vif)) {
1426 ret = -EINVAL;
1427 break;
1428 }
1429 if (copy_from_sockptr(&vif, optval, sizeof(vif))) {
1430 ret = -EFAULT;
1431 break;
1432 }
1433 if (vif.vifc_vifi >= MAXVIFS) {
1434 ret = -ENFILE;
1435 break;
1436 }
1437 if (optname == MRT_ADD_VIF) {
1438 ret = vif_add(net, mrt, &vif,
1439 sk == rtnl_dereference(mrt->mroute_sk));
1440 } else {
1441 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1442 }
1443 break;
1444 /* Manipulate the forwarding caches. These live
1445 * in a sort of kernel/user symbiosis.
1446 */
1447 case MRT_ADD_MFC:
1448 case MRT_DEL_MFC:
1449 parent = -1;
1450 fallthrough;
1451 case MRT_ADD_MFC_PROXY:
1452 case MRT_DEL_MFC_PROXY:
1453 if (optlen != sizeof(mfc)) {
1454 ret = -EINVAL;
1455 break;
1456 }
1457 if (copy_from_sockptr(&mfc, optval, sizeof(mfc))) {
1458 ret = -EFAULT;
1459 break;
1460 }
1461 if (parent == 0)
1462 parent = mfc.mfcc_parent;
1463 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1464 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1465 else
1466 ret = ipmr_mfc_add(net, mrt, &mfc,
1467 sk == rtnl_dereference(mrt->mroute_sk),
1468 parent);
1469 break;
1470 case MRT_FLUSH:
1471 if (optlen != sizeof(val)) {
1472 ret = -EINVAL;
1473 break;
1474 }
1475 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1476 ret = -EFAULT;
1477 break;
1478 }
1479 mroute_clean_tables(mrt, val);
1480 break;
1481 /* Control PIM assert. */
1482 case MRT_ASSERT:
1483 if (optlen != sizeof(val)) {
1484 ret = -EINVAL;
1485 break;
1486 }
1487 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1488 ret = -EFAULT;
1489 break;
1490 }
1491 mrt->mroute_do_assert = val;
1492 break;
1493 case MRT_PIM:
1494 if (!ipmr_pimsm_enabled()) {
1495 ret = -ENOPROTOOPT;
1496 break;
1497 }
1498 if (optlen != sizeof(val)) {
1499 ret = -EINVAL;
1500 break;
1501 }
1502 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1503 ret = -EFAULT;
1504 break;
1505 }
1506
1507 do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1508 val = !!val;
1509 if (val != mrt->mroute_do_pim) {
1510 mrt->mroute_do_pim = val;
1511 mrt->mroute_do_assert = val;
1512 mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1513 }
1514 break;
1515 case MRT_TABLE:
1516 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1517 ret = -ENOPROTOOPT;
1518 break;
1519 }
1520 if (optlen != sizeof(uval)) {
1521 ret = -EINVAL;
1522 break;
1523 }
1524 if (copy_from_sockptr(&uval, optval, sizeof(uval))) {
1525 ret = -EFAULT;
1526 break;
1527 }
1528
1529 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1530 ret = -EBUSY;
1531 } else {
1532 mrt = ipmr_new_table(net, uval);
1533 if (IS_ERR(mrt))
1534 ret = PTR_ERR(mrt);
1535 else
1536 raw_sk(sk)->ipmr_table = uval;
1537 }
1538 break;
1539 /* Spurious command, or MRT_VERSION which you cannot set. */
1540 default:
1541 ret = -ENOPROTOOPT;
1542 }
1543out_unlock:
1544 rtnl_unlock();
1545out:
1546 return ret;
1547}
1548
1549/* Execute if this ioctl is a special mroute ioctl */
1550int ipmr_sk_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1551{
1552 switch (cmd) {
1553 /* These userspace buffers will be consumed by ipmr_ioctl() */
1554 case SIOCGETVIFCNT: {
1555 struct sioc_vif_req buffer;
1556
1557 return sock_ioctl_inout(sk, cmd, arg, &buffer,
1558 sizeof(buffer));
1559 }
1560 case SIOCGETSGCNT: {
1561 struct sioc_sg_req buffer;
1562
1563 return sock_ioctl_inout(sk, cmd, arg, &buffer,
1564 sizeof(buffer));
1565 }
1566 }
1567 /* return code > 0 means that the ioctl was not executed */
1568 return 1;
1569}
1570
1571/* Getsock opt support for the multicast routing system. */
1572int ip_mroute_getsockopt(struct sock *sk, int optname, sockptr_t optval,
1573 sockptr_t optlen)
1574{
1575 int olr;
1576 int val;
1577 struct net *net = sock_net(sk);
1578 struct mr_table *mrt;
1579
1580 if (sk->sk_type != SOCK_RAW ||
1581 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1582 return -EOPNOTSUPP;
1583
1584 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1585 if (!mrt)
1586 return -ENOENT;
1587
1588 switch (optname) {
1589 case MRT_VERSION:
1590 val = 0x0305;
1591 break;
1592 case MRT_PIM:
1593 if (!ipmr_pimsm_enabled())
1594 return -ENOPROTOOPT;
1595 val = mrt->mroute_do_pim;
1596 break;
1597 case MRT_ASSERT:
1598 val = mrt->mroute_do_assert;
1599 break;
1600 default:
1601 return -ENOPROTOOPT;
1602 }
1603
1604 if (copy_from_sockptr(&olr, optlen, sizeof(int)))
1605 return -EFAULT;
1606 olr = min_t(unsigned int, olr, sizeof(int));
1607 if (olr < 0)
1608 return -EINVAL;
1609 if (copy_to_sockptr(optlen, &olr, sizeof(int)))
1610 return -EFAULT;
1611 if (copy_to_sockptr(optval, &val, olr))
1612 return -EFAULT;
1613 return 0;
1614}
1615
1616/* The IP multicast ioctl support routines. */
1617int ipmr_ioctl(struct sock *sk, int cmd, void *arg)
1618{
1619 struct vif_device *vif;
1620 struct mfc_cache *c;
1621 struct net *net = sock_net(sk);
1622 struct sioc_vif_req *vr;
1623 struct sioc_sg_req *sr;
1624 struct mr_table *mrt;
1625
1626 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1627 if (!mrt)
1628 return -ENOENT;
1629
1630 switch (cmd) {
1631 case SIOCGETVIFCNT:
1632 vr = (struct sioc_vif_req *)arg;
1633 if (vr->vifi >= mrt->maxvif)
1634 return -EINVAL;
1635 vr->vifi = array_index_nospec(vr->vifi, mrt->maxvif);
1636 rcu_read_lock();
1637 vif = &mrt->vif_table[vr->vifi];
1638 if (VIF_EXISTS(mrt, vr->vifi)) {
1639 vr->icount = READ_ONCE(vif->pkt_in);
1640 vr->ocount = READ_ONCE(vif->pkt_out);
1641 vr->ibytes = READ_ONCE(vif->bytes_in);
1642 vr->obytes = READ_ONCE(vif->bytes_out);
1643 rcu_read_unlock();
1644
1645 return 0;
1646 }
1647 rcu_read_unlock();
1648 return -EADDRNOTAVAIL;
1649 case SIOCGETSGCNT:
1650 sr = (struct sioc_sg_req *)arg;
1651
1652 rcu_read_lock();
1653 c = ipmr_cache_find(mrt, sr->src.s_addr, sr->grp.s_addr);
1654 if (c) {
1655 sr->pktcnt = c->_c.mfc_un.res.pkt;
1656 sr->bytecnt = c->_c.mfc_un.res.bytes;
1657 sr->wrong_if = c->_c.mfc_un.res.wrong_if;
1658 rcu_read_unlock();
1659 return 0;
1660 }
1661 rcu_read_unlock();
1662 return -EADDRNOTAVAIL;
1663 default:
1664 return -ENOIOCTLCMD;
1665 }
1666}
1667
1668#ifdef CONFIG_COMPAT
1669struct compat_sioc_sg_req {
1670 struct in_addr src;
1671 struct in_addr grp;
1672 compat_ulong_t pktcnt;
1673 compat_ulong_t bytecnt;
1674 compat_ulong_t wrong_if;
1675};
1676
1677struct compat_sioc_vif_req {
1678 vifi_t vifi; /* Which iface */
1679 compat_ulong_t icount;
1680 compat_ulong_t ocount;
1681 compat_ulong_t ibytes;
1682 compat_ulong_t obytes;
1683};
1684
1685int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1686{
1687 struct compat_sioc_sg_req sr;
1688 struct compat_sioc_vif_req vr;
1689 struct vif_device *vif;
1690 struct mfc_cache *c;
1691 struct net *net = sock_net(sk);
1692 struct mr_table *mrt;
1693
1694 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1695 if (!mrt)
1696 return -ENOENT;
1697
1698 switch (cmd) {
1699 case SIOCGETVIFCNT:
1700 if (copy_from_user(&vr, arg, sizeof(vr)))
1701 return -EFAULT;
1702 if (vr.vifi >= mrt->maxvif)
1703 return -EINVAL;
1704 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1705 rcu_read_lock();
1706 vif = &mrt->vif_table[vr.vifi];
1707 if (VIF_EXISTS(mrt, vr.vifi)) {
1708 vr.icount = READ_ONCE(vif->pkt_in);
1709 vr.ocount = READ_ONCE(vif->pkt_out);
1710 vr.ibytes = READ_ONCE(vif->bytes_in);
1711 vr.obytes = READ_ONCE(vif->bytes_out);
1712 rcu_read_unlock();
1713
1714 if (copy_to_user(arg, &vr, sizeof(vr)))
1715 return -EFAULT;
1716 return 0;
1717 }
1718 rcu_read_unlock();
1719 return -EADDRNOTAVAIL;
1720 case SIOCGETSGCNT:
1721 if (copy_from_user(&sr, arg, sizeof(sr)))
1722 return -EFAULT;
1723
1724 rcu_read_lock();
1725 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1726 if (c) {
1727 sr.pktcnt = c->_c.mfc_un.res.pkt;
1728 sr.bytecnt = c->_c.mfc_un.res.bytes;
1729 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1730 rcu_read_unlock();
1731
1732 if (copy_to_user(arg, &sr, sizeof(sr)))
1733 return -EFAULT;
1734 return 0;
1735 }
1736 rcu_read_unlock();
1737 return -EADDRNOTAVAIL;
1738 default:
1739 return -ENOIOCTLCMD;
1740 }
1741}
1742#endif
1743
1744static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1745{
1746 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1747 struct net *net = dev_net(dev);
1748 struct mr_table *mrt;
1749 struct vif_device *v;
1750 int ct;
1751
1752 if (event != NETDEV_UNREGISTER)
1753 return NOTIFY_DONE;
1754
1755 ipmr_for_each_table(mrt, net) {
1756 v = &mrt->vif_table[0];
1757 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1758 if (rcu_access_pointer(v->dev) == dev)
1759 vif_delete(mrt, ct, 1, NULL);
1760 }
1761 }
1762 return NOTIFY_DONE;
1763}
1764
1765static struct notifier_block ip_mr_notifier = {
1766 .notifier_call = ipmr_device_event,
1767};
1768
1769/* Encapsulate a packet by attaching a valid IPIP header to it.
1770 * This avoids tunnel drivers and other mess and gives us the speed so
1771 * important for multicast video.
1772 */
1773static void ip_encap(struct net *net, struct sk_buff *skb,
1774 __be32 saddr, __be32 daddr)
1775{
1776 struct iphdr *iph;
1777 const struct iphdr *old_iph = ip_hdr(skb);
1778
1779 skb_push(skb, sizeof(struct iphdr));
1780 skb->transport_header = skb->network_header;
1781 skb_reset_network_header(skb);
1782 iph = ip_hdr(skb);
1783
1784 iph->version = 4;
1785 iph->tos = old_iph->tos;
1786 iph->ttl = old_iph->ttl;
1787 iph->frag_off = 0;
1788 iph->daddr = daddr;
1789 iph->saddr = saddr;
1790 iph->protocol = IPPROTO_IPIP;
1791 iph->ihl = 5;
1792 iph->tot_len = htons(skb->len);
1793 ip_select_ident(net, skb, NULL);
1794 ip_send_check(iph);
1795
1796 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1797 nf_reset_ct(skb);
1798}
1799
1800static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1801 struct sk_buff *skb)
1802{
1803 struct ip_options *opt = &(IPCB(skb)->opt);
1804
1805 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1806
1807 if (unlikely(opt->optlen))
1808 ip_forward_options(skb);
1809
1810 return dst_output(net, sk, skb);
1811}
1812
1813#ifdef CONFIG_NET_SWITCHDEV
1814static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1815 int in_vifi, int out_vifi)
1816{
1817 struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1818 struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1819
1820 if (!skb->offload_l3_fwd_mark)
1821 return false;
1822 if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1823 return false;
1824 return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1825 &in_vif->dev_parent_id);
1826}
1827#else
1828static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1829 int in_vifi, int out_vifi)
1830{
1831 return false;
1832}
1833#endif
1834
1835/* Processing handlers for ipmr_forward, under rcu_read_lock() */
1836
1837static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1838 int in_vifi, struct sk_buff *skb, int vifi)
1839{
1840 const struct iphdr *iph = ip_hdr(skb);
1841 struct vif_device *vif = &mrt->vif_table[vifi];
1842 struct net_device *vif_dev;
1843 struct net_device *dev;
1844 struct rtable *rt;
1845 struct flowi4 fl4;
1846 int encap = 0;
1847
1848 vif_dev = vif_dev_read(vif);
1849 if (!vif_dev)
1850 goto out_free;
1851
1852 if (vif->flags & VIFF_REGISTER) {
1853 WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1854 WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1855 DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
1856 DEV_STATS_INC(vif_dev, tx_packets);
1857 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1858 goto out_free;
1859 }
1860
1861 if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1862 goto out_free;
1863
1864 if (vif->flags & VIFF_TUNNEL) {
1865 rt = ip_route_output_ports(net, &fl4, NULL,
1866 vif->remote, vif->local,
1867 0, 0,
1868 IPPROTO_IPIP,
1869 RT_TOS(iph->tos), vif->link);
1870 if (IS_ERR(rt))
1871 goto out_free;
1872 encap = sizeof(struct iphdr);
1873 } else {
1874 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1875 0, 0,
1876 IPPROTO_IPIP,
1877 RT_TOS(iph->tos), vif->link);
1878 if (IS_ERR(rt))
1879 goto out_free;
1880 }
1881
1882 dev = rt->dst.dev;
1883
1884 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1885 /* Do not fragment multicasts. Alas, IPv4 does not
1886 * allow to send ICMP, so that packets will disappear
1887 * to blackhole.
1888 */
1889 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1890 ip_rt_put(rt);
1891 goto out_free;
1892 }
1893
1894 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1895
1896 if (skb_cow(skb, encap)) {
1897 ip_rt_put(rt);
1898 goto out_free;
1899 }
1900
1901 WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1902 WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1903
1904 skb_dst_drop(skb);
1905 skb_dst_set(skb, &rt->dst);
1906 ip_decrease_ttl(ip_hdr(skb));
1907
1908 /* FIXME: forward and output firewalls used to be called here.
1909 * What do we do with netfilter? -- RR
1910 */
1911 if (vif->flags & VIFF_TUNNEL) {
1912 ip_encap(net, skb, vif->local, vif->remote);
1913 /* FIXME: extra output firewall step used to be here. --RR */
1914 DEV_STATS_INC(vif_dev, tx_packets);
1915 DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
1916 }
1917
1918 IPCB(skb)->flags |= IPSKB_FORWARDED;
1919
1920 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1921 * not only before forwarding, but after forwarding on all output
1922 * interfaces. It is clear, if mrouter runs a multicasting
1923 * program, it should receive packets not depending to what interface
1924 * program is joined.
1925 * If we will not make it, the program will have to join on all
1926 * interfaces. On the other hand, multihoming host (or router, but
1927 * not mrouter) cannot join to more than one interface - it will
1928 * result in receiving multiple packets.
1929 */
1930 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1931 net, NULL, skb, skb->dev, dev,
1932 ipmr_forward_finish);
1933 return;
1934
1935out_free:
1936 kfree_skb(skb);
1937}
1938
1939/* Called with mrt_lock or rcu_read_lock() */
1940static int ipmr_find_vif(const struct mr_table *mrt, struct net_device *dev)
1941{
1942 int ct;
1943 /* Pairs with WRITE_ONCE() in vif_delete()/vif_add() */
1944 for (ct = READ_ONCE(mrt->maxvif) - 1; ct >= 0; ct--) {
1945 if (rcu_access_pointer(mrt->vif_table[ct].dev) == dev)
1946 break;
1947 }
1948 return ct;
1949}
1950
1951/* "local" means that we should preserve one skb (for local delivery) */
1952/* Called uner rcu_read_lock() */
1953static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1954 struct net_device *dev, struct sk_buff *skb,
1955 struct mfc_cache *c, int local)
1956{
1957 int true_vifi = ipmr_find_vif(mrt, dev);
1958 int psend = -1;
1959 int vif, ct;
1960
1961 vif = c->_c.mfc_parent;
1962 c->_c.mfc_un.res.pkt++;
1963 c->_c.mfc_un.res.bytes += skb->len;
1964 c->_c.mfc_un.res.lastuse = jiffies;
1965
1966 if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1967 struct mfc_cache *cache_proxy;
1968
1969 /* For an (*,G) entry, we only check that the incoming
1970 * interface is part of the static tree.
1971 */
1972 cache_proxy = mr_mfc_find_any_parent(mrt, vif);
1973 if (cache_proxy &&
1974 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
1975 goto forward;
1976 }
1977
1978 /* Wrong interface: drop packet and (maybe) send PIM assert. */
1979 if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev) {
1980 if (rt_is_output_route(skb_rtable(skb))) {
1981 /* It is our own packet, looped back.
1982 * Very complicated situation...
1983 *
1984 * The best workaround until routing daemons will be
1985 * fixed is not to redistribute packet, if it was
1986 * send through wrong interface. It means, that
1987 * multicast applications WILL NOT work for
1988 * (S,G), which have default multicast route pointing
1989 * to wrong oif. In any case, it is not a good
1990 * idea to use multicasting applications on router.
1991 */
1992 goto dont_forward;
1993 }
1994
1995 c->_c.mfc_un.res.wrong_if++;
1996
1997 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1998 /* pimsm uses asserts, when switching from RPT to SPT,
1999 * so that we cannot check that packet arrived on an oif.
2000 * It is bad, but otherwise we would need to move pretty
2001 * large chunk of pimd to kernel. Ough... --ANK
2002 */
2003 (mrt->mroute_do_pim ||
2004 c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
2005 time_after(jiffies,
2006 c->_c.mfc_un.res.last_assert +
2007 MFC_ASSERT_THRESH)) {
2008 c->_c.mfc_un.res.last_assert = jiffies;
2009 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
2010 if (mrt->mroute_do_wrvifwhole)
2011 ipmr_cache_report(mrt, skb, true_vifi,
2012 IGMPMSG_WRVIFWHOLE);
2013 }
2014 goto dont_forward;
2015 }
2016
2017forward:
2018 WRITE_ONCE(mrt->vif_table[vif].pkt_in,
2019 mrt->vif_table[vif].pkt_in + 1);
2020 WRITE_ONCE(mrt->vif_table[vif].bytes_in,
2021 mrt->vif_table[vif].bytes_in + skb->len);
2022
2023 /* Forward the frame */
2024 if (c->mfc_origin == htonl(INADDR_ANY) &&
2025 c->mfc_mcastgrp == htonl(INADDR_ANY)) {
2026 if (true_vifi >= 0 &&
2027 true_vifi != c->_c.mfc_parent &&
2028 ip_hdr(skb)->ttl >
2029 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2030 /* It's an (*,*) entry and the packet is not coming from
2031 * the upstream: forward the packet to the upstream
2032 * only.
2033 */
2034 psend = c->_c.mfc_parent;
2035 goto last_forward;
2036 }
2037 goto dont_forward;
2038 }
2039 for (ct = c->_c.mfc_un.res.maxvif - 1;
2040 ct >= c->_c.mfc_un.res.minvif; ct--) {
2041 /* For (*,G) entry, don't forward to the incoming interface */
2042 if ((c->mfc_origin != htonl(INADDR_ANY) ||
2043 ct != true_vifi) &&
2044 ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2045 if (psend != -1) {
2046 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2047
2048 if (skb2)
2049 ipmr_queue_xmit(net, mrt, true_vifi,
2050 skb2, psend);
2051 }
2052 psend = ct;
2053 }
2054 }
2055last_forward:
2056 if (psend != -1) {
2057 if (local) {
2058 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2059
2060 if (skb2)
2061 ipmr_queue_xmit(net, mrt, true_vifi, skb2,
2062 psend);
2063 } else {
2064 ipmr_queue_xmit(net, mrt, true_vifi, skb, psend);
2065 return;
2066 }
2067 }
2068
2069dont_forward:
2070 if (!local)
2071 kfree_skb(skb);
2072}
2073
2074static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2075{
2076 struct rtable *rt = skb_rtable(skb);
2077 struct iphdr *iph = ip_hdr(skb);
2078 struct flowi4 fl4 = {
2079 .daddr = iph->daddr,
2080 .saddr = iph->saddr,
2081 .flowi4_tos = RT_TOS(iph->tos),
2082 .flowi4_oif = (rt_is_output_route(rt) ?
2083 skb->dev->ifindex : 0),
2084 .flowi4_iif = (rt_is_output_route(rt) ?
2085 LOOPBACK_IFINDEX :
2086 skb->dev->ifindex),
2087 .flowi4_mark = skb->mark,
2088 };
2089 struct mr_table *mrt;
2090 int err;
2091
2092 err = ipmr_fib_lookup(net, &fl4, &mrt);
2093 if (err)
2094 return ERR_PTR(err);
2095 return mrt;
2096}
2097
2098/* Multicast packets for forwarding arrive here
2099 * Called with rcu_read_lock();
2100 */
2101int ip_mr_input(struct sk_buff *skb)
2102{
2103 struct mfc_cache *cache;
2104 struct net *net = dev_net(skb->dev);
2105 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2106 struct mr_table *mrt;
2107 struct net_device *dev;
2108
2109 /* skb->dev passed in is the loX master dev for vrfs.
2110 * As there are no vifs associated with loopback devices,
2111 * get the proper interface that does have a vif associated with it.
2112 */
2113 dev = skb->dev;
2114 if (netif_is_l3_master(skb->dev)) {
2115 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2116 if (!dev) {
2117 kfree_skb(skb);
2118 return -ENODEV;
2119 }
2120 }
2121
2122 /* Packet is looped back after forward, it should not be
2123 * forwarded second time, but still can be delivered locally.
2124 */
2125 if (IPCB(skb)->flags & IPSKB_FORWARDED)
2126 goto dont_forward;
2127
2128 mrt = ipmr_rt_fib_lookup(net, skb);
2129 if (IS_ERR(mrt)) {
2130 kfree_skb(skb);
2131 return PTR_ERR(mrt);
2132 }
2133 if (!local) {
2134 if (IPCB(skb)->opt.router_alert) {
2135 if (ip_call_ra_chain(skb))
2136 return 0;
2137 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2138 /* IGMPv1 (and broken IGMPv2 implementations sort of
2139 * Cisco IOS <= 11.2(8)) do not put router alert
2140 * option to IGMP packets destined to routable
2141 * groups. It is very bad, because it means
2142 * that we can forward NO IGMP messages.
2143 */
2144 struct sock *mroute_sk;
2145
2146 mroute_sk = rcu_dereference(mrt->mroute_sk);
2147 if (mroute_sk) {
2148 nf_reset_ct(skb);
2149 raw_rcv(mroute_sk, skb);
2150 return 0;
2151 }
2152 }
2153 }
2154
2155 /* already under rcu_read_lock() */
2156 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2157 if (!cache) {
2158 int vif = ipmr_find_vif(mrt, dev);
2159
2160 if (vif >= 0)
2161 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2162 vif);
2163 }
2164
2165 /* No usable cache entry */
2166 if (!cache) {
2167 int vif;
2168
2169 if (local) {
2170 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2171 ip_local_deliver(skb);
2172 if (!skb2)
2173 return -ENOBUFS;
2174 skb = skb2;
2175 }
2176
2177 vif = ipmr_find_vif(mrt, dev);
2178 if (vif >= 0)
2179 return ipmr_cache_unresolved(mrt, vif, skb, dev);
2180 kfree_skb(skb);
2181 return -ENODEV;
2182 }
2183
2184 ip_mr_forward(net, mrt, dev, skb, cache, local);
2185
2186 if (local)
2187 return ip_local_deliver(skb);
2188
2189 return 0;
2190
2191dont_forward:
2192 if (local)
2193 return ip_local_deliver(skb);
2194 kfree_skb(skb);
2195 return 0;
2196}
2197
2198#ifdef CONFIG_IP_PIMSM_V1
2199/* Handle IGMP messages of PIMv1 */
2200int pim_rcv_v1(struct sk_buff *skb)
2201{
2202 struct igmphdr *pim;
2203 struct net *net = dev_net(skb->dev);
2204 struct mr_table *mrt;
2205
2206 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2207 goto drop;
2208
2209 pim = igmp_hdr(skb);
2210
2211 mrt = ipmr_rt_fib_lookup(net, skb);
2212 if (IS_ERR(mrt))
2213 goto drop;
2214 if (!mrt->mroute_do_pim ||
2215 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2216 goto drop;
2217
2218 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2219drop:
2220 kfree_skb(skb);
2221 }
2222 return 0;
2223}
2224#endif
2225
2226#ifdef CONFIG_IP_PIMSM_V2
2227static int pim_rcv(struct sk_buff *skb)
2228{
2229 struct pimreghdr *pim;
2230 struct net *net = dev_net(skb->dev);
2231 struct mr_table *mrt;
2232
2233 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2234 goto drop;
2235
2236 pim = (struct pimreghdr *)skb_transport_header(skb);
2237 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2238 (pim->flags & PIM_NULL_REGISTER) ||
2239 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2240 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2241 goto drop;
2242
2243 mrt = ipmr_rt_fib_lookup(net, skb);
2244 if (IS_ERR(mrt))
2245 goto drop;
2246 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2247drop:
2248 kfree_skb(skb);
2249 }
2250 return 0;
2251}
2252#endif
2253
2254int ipmr_get_route(struct net *net, struct sk_buff *skb,
2255 __be32 saddr, __be32 daddr,
2256 struct rtmsg *rtm, u32 portid)
2257{
2258 struct mfc_cache *cache;
2259 struct mr_table *mrt;
2260 int err;
2261
2262 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2263 if (!mrt)
2264 return -ENOENT;
2265
2266 rcu_read_lock();
2267 cache = ipmr_cache_find(mrt, saddr, daddr);
2268 if (!cache && skb->dev) {
2269 int vif = ipmr_find_vif(mrt, skb->dev);
2270
2271 if (vif >= 0)
2272 cache = ipmr_cache_find_any(mrt, daddr, vif);
2273 }
2274 if (!cache) {
2275 struct sk_buff *skb2;
2276 struct iphdr *iph;
2277 struct net_device *dev;
2278 int vif = -1;
2279
2280 dev = skb->dev;
2281 if (dev)
2282 vif = ipmr_find_vif(mrt, dev);
2283 if (vif < 0) {
2284 rcu_read_unlock();
2285 return -ENODEV;
2286 }
2287
2288 skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr));
2289 if (!skb2) {
2290 rcu_read_unlock();
2291 return -ENOMEM;
2292 }
2293
2294 NETLINK_CB(skb2).portid = portid;
2295 skb_push(skb2, sizeof(struct iphdr));
2296 skb_reset_network_header(skb2);
2297 iph = ip_hdr(skb2);
2298 iph->ihl = sizeof(struct iphdr) >> 2;
2299 iph->saddr = saddr;
2300 iph->daddr = daddr;
2301 iph->version = 0;
2302 err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2303 rcu_read_unlock();
2304 return err;
2305 }
2306
2307 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2308 rcu_read_unlock();
2309 return err;
2310}
2311
2312static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2313 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2314 int flags)
2315{
2316 struct nlmsghdr *nlh;
2317 struct rtmsg *rtm;
2318 int err;
2319
2320 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2321 if (!nlh)
2322 return -EMSGSIZE;
2323
2324 rtm = nlmsg_data(nlh);
2325 rtm->rtm_family = RTNL_FAMILY_IPMR;
2326 rtm->rtm_dst_len = 32;
2327 rtm->rtm_src_len = 32;
2328 rtm->rtm_tos = 0;
2329 rtm->rtm_table = mrt->id;
2330 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2331 goto nla_put_failure;
2332 rtm->rtm_type = RTN_MULTICAST;
2333 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2334 if (c->_c.mfc_flags & MFC_STATIC)
2335 rtm->rtm_protocol = RTPROT_STATIC;
2336 else
2337 rtm->rtm_protocol = RTPROT_MROUTED;
2338 rtm->rtm_flags = 0;
2339
2340 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2341 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2342 goto nla_put_failure;
2343 err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2344 /* do not break the dump if cache is unresolved */
2345 if (err < 0 && err != -ENOENT)
2346 goto nla_put_failure;
2347
2348 nlmsg_end(skb, nlh);
2349 return 0;
2350
2351nla_put_failure:
2352 nlmsg_cancel(skb, nlh);
2353 return -EMSGSIZE;
2354}
2355
2356static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2357 u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2358 int flags)
2359{
2360 return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2361 cmd, flags);
2362}
2363
2364static size_t mroute_msgsize(bool unresolved, int maxvif)
2365{
2366 size_t len =
2367 NLMSG_ALIGN(sizeof(struct rtmsg))
2368 + nla_total_size(4) /* RTA_TABLE */
2369 + nla_total_size(4) /* RTA_SRC */
2370 + nla_total_size(4) /* RTA_DST */
2371 ;
2372
2373 if (!unresolved)
2374 len = len
2375 + nla_total_size(4) /* RTA_IIF */
2376 + nla_total_size(0) /* RTA_MULTIPATH */
2377 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2378 /* RTA_MFC_STATS */
2379 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2380 ;
2381
2382 return len;
2383}
2384
2385static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2386 int cmd)
2387{
2388 struct net *net = read_pnet(&mrt->net);
2389 struct sk_buff *skb;
2390 int err = -ENOBUFS;
2391
2392 skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2393 mrt->maxvif),
2394 GFP_ATOMIC);
2395 if (!skb)
2396 goto errout;
2397
2398 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2399 if (err < 0)
2400 goto errout;
2401
2402 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2403 return;
2404
2405errout:
2406 kfree_skb(skb);
2407 if (err < 0)
2408 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2409}
2410
2411static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2412{
2413 size_t len =
2414 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2415 + nla_total_size(1) /* IPMRA_CREPORT_MSGTYPE */
2416 + nla_total_size(4) /* IPMRA_CREPORT_VIF_ID */
2417 + nla_total_size(4) /* IPMRA_CREPORT_SRC_ADDR */
2418 + nla_total_size(4) /* IPMRA_CREPORT_DST_ADDR */
2419 + nla_total_size(4) /* IPMRA_CREPORT_TABLE */
2420 /* IPMRA_CREPORT_PKT */
2421 + nla_total_size(payloadlen)
2422 ;
2423
2424 return len;
2425}
2426
2427static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt)
2428{
2429 struct net *net = read_pnet(&mrt->net);
2430 struct nlmsghdr *nlh;
2431 struct rtgenmsg *rtgenm;
2432 struct igmpmsg *msg;
2433 struct sk_buff *skb;
2434 struct nlattr *nla;
2435 int payloadlen;
2436
2437 payloadlen = pkt->len - sizeof(struct igmpmsg);
2438 msg = (struct igmpmsg *)skb_network_header(pkt);
2439
2440 skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2441 if (!skb)
2442 goto errout;
2443
2444 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2445 sizeof(struct rtgenmsg), 0);
2446 if (!nlh)
2447 goto errout;
2448 rtgenm = nlmsg_data(nlh);
2449 rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2450 if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2451 nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif | (msg->im_vif_hi << 8)) ||
2452 nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2453 msg->im_src.s_addr) ||
2454 nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2455 msg->im_dst.s_addr) ||
2456 nla_put_u32(skb, IPMRA_CREPORT_TABLE, mrt->id))
2457 goto nla_put_failure;
2458
2459 nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2460 if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2461 nla_data(nla), payloadlen))
2462 goto nla_put_failure;
2463
2464 nlmsg_end(skb, nlh);
2465
2466 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2467 return;
2468
2469nla_put_failure:
2470 nlmsg_cancel(skb, nlh);
2471errout:
2472 kfree_skb(skb);
2473 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2474}
2475
2476static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2477 const struct nlmsghdr *nlh,
2478 struct nlattr **tb,
2479 struct netlink_ext_ack *extack)
2480{
2481 struct rtmsg *rtm;
2482 int i, err;
2483
2484 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2485 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2486 return -EINVAL;
2487 }
2488
2489 if (!netlink_strict_get_check(skb))
2490 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2491 rtm_ipv4_policy, extack);
2492
2493 rtm = nlmsg_data(nlh);
2494 if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2495 (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2496 rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2497 rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2498 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2499 return -EINVAL;
2500 }
2501
2502 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2503 rtm_ipv4_policy, extack);
2504 if (err)
2505 return err;
2506
2507 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2508 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2509 NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2510 return -EINVAL;
2511 }
2512
2513 for (i = 0; i <= RTA_MAX; i++) {
2514 if (!tb[i])
2515 continue;
2516
2517 switch (i) {
2518 case RTA_SRC:
2519 case RTA_DST:
2520 case RTA_TABLE:
2521 break;
2522 default:
2523 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2524 return -EINVAL;
2525 }
2526 }
2527
2528 return 0;
2529}
2530
2531static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2532 struct netlink_ext_ack *extack)
2533{
2534 struct net *net = sock_net(in_skb->sk);
2535 struct nlattr *tb[RTA_MAX + 1];
2536 struct sk_buff *skb = NULL;
2537 struct mfc_cache *cache;
2538 struct mr_table *mrt;
2539 __be32 src, grp;
2540 u32 tableid;
2541 int err;
2542
2543 err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2544 if (err < 0)
2545 goto errout;
2546
2547 src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2548 grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2549 tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2550
2551 mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2552 if (!mrt) {
2553 err = -ENOENT;
2554 goto errout_free;
2555 }
2556
2557 /* entries are added/deleted only under RTNL */
2558 rcu_read_lock();
2559 cache = ipmr_cache_find(mrt, src, grp);
2560 rcu_read_unlock();
2561 if (!cache) {
2562 err = -ENOENT;
2563 goto errout_free;
2564 }
2565
2566 skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2567 if (!skb) {
2568 err = -ENOBUFS;
2569 goto errout_free;
2570 }
2571
2572 err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2573 nlh->nlmsg_seq, cache,
2574 RTM_NEWROUTE, 0);
2575 if (err < 0)
2576 goto errout_free;
2577
2578 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2579
2580errout:
2581 return err;
2582
2583errout_free:
2584 kfree_skb(skb);
2585 goto errout;
2586}
2587
2588static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2589{
2590 struct fib_dump_filter filter = {};
2591 int err;
2592
2593 if (cb->strict_check) {
2594 err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
2595 &filter, cb);
2596 if (err < 0)
2597 return err;
2598 }
2599
2600 if (filter.table_id) {
2601 struct mr_table *mrt;
2602
2603 mrt = ipmr_get_table(sock_net(skb->sk), filter.table_id);
2604 if (!mrt) {
2605 if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IPMR)
2606 return skb->len;
2607
2608 NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2609 return -ENOENT;
2610 }
2611 err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2612 &mfc_unres_lock, &filter);
2613 return skb->len ? : err;
2614 }
2615
2616 return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2617 _ipmr_fill_mroute, &mfc_unres_lock, &filter);
2618}
2619
2620static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2621 [RTA_SRC] = { .type = NLA_U32 },
2622 [RTA_DST] = { .type = NLA_U32 },
2623 [RTA_IIF] = { .type = NLA_U32 },
2624 [RTA_TABLE] = { .type = NLA_U32 },
2625 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2626};
2627
2628static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2629{
2630 switch (rtm_protocol) {
2631 case RTPROT_STATIC:
2632 case RTPROT_MROUTED:
2633 return true;
2634 }
2635 return false;
2636}
2637
2638static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2639{
2640 struct rtnexthop *rtnh = nla_data(nla);
2641 int remaining = nla_len(nla), vifi = 0;
2642
2643 while (rtnh_ok(rtnh, remaining)) {
2644 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2645 if (++vifi == MAXVIFS)
2646 break;
2647 rtnh = rtnh_next(rtnh, &remaining);
2648 }
2649
2650 return remaining > 0 ? -EINVAL : vifi;
2651}
2652
2653/* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2654static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2655 struct mfcctl *mfcc, int *mrtsock,
2656 struct mr_table **mrtret,
2657 struct netlink_ext_ack *extack)
2658{
2659 struct net_device *dev = NULL;
2660 u32 tblid = RT_TABLE_DEFAULT;
2661 struct mr_table *mrt;
2662 struct nlattr *attr;
2663 struct rtmsg *rtm;
2664 int ret, rem;
2665
2666 ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2667 rtm_ipmr_policy, extack);
2668 if (ret < 0)
2669 goto out;
2670 rtm = nlmsg_data(nlh);
2671
2672 ret = -EINVAL;
2673 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2674 rtm->rtm_type != RTN_MULTICAST ||
2675 rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2676 !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2677 goto out;
2678
2679 memset(mfcc, 0, sizeof(*mfcc));
2680 mfcc->mfcc_parent = -1;
2681 ret = 0;
2682 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2683 switch (nla_type(attr)) {
2684 case RTA_SRC:
2685 mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2686 break;
2687 case RTA_DST:
2688 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2689 break;
2690 case RTA_IIF:
2691 dev = __dev_get_by_index(net, nla_get_u32(attr));
2692 if (!dev) {
2693 ret = -ENODEV;
2694 goto out;
2695 }
2696 break;
2697 case RTA_MULTIPATH:
2698 if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2699 ret = -EINVAL;
2700 goto out;
2701 }
2702 break;
2703 case RTA_PREFSRC:
2704 ret = 1;
2705 break;
2706 case RTA_TABLE:
2707 tblid = nla_get_u32(attr);
2708 break;
2709 }
2710 }
2711 mrt = ipmr_get_table(net, tblid);
2712 if (!mrt) {
2713 ret = -ENOENT;
2714 goto out;
2715 }
2716 *mrtret = mrt;
2717 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2718 if (dev)
2719 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2720
2721out:
2722 return ret;
2723}
2724
2725/* takes care of both newroute and delroute */
2726static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2727 struct netlink_ext_ack *extack)
2728{
2729 struct net *net = sock_net(skb->sk);
2730 int ret, mrtsock, parent;
2731 struct mr_table *tbl;
2732 struct mfcctl mfcc;
2733
2734 mrtsock = 0;
2735 tbl = NULL;
2736 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2737 if (ret < 0)
2738 return ret;
2739
2740 parent = ret ? mfcc.mfcc_parent : -1;
2741 if (nlh->nlmsg_type == RTM_NEWROUTE)
2742 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2743 else
2744 return ipmr_mfc_delete(tbl, &mfcc, parent);
2745}
2746
2747static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2748{
2749 u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2750
2751 if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2752 nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2753 nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2754 mrt->mroute_reg_vif_num) ||
2755 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2756 mrt->mroute_do_assert) ||
2757 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) ||
2758 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2759 mrt->mroute_do_wrvifwhole))
2760 return false;
2761
2762 return true;
2763}
2764
2765static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2766{
2767 struct net_device *vif_dev;
2768 struct nlattr *vif_nest;
2769 struct vif_device *vif;
2770
2771 vif = &mrt->vif_table[vifid];
2772 vif_dev = rtnl_dereference(vif->dev);
2773 /* if the VIF doesn't exist just continue */
2774 if (!vif_dev)
2775 return true;
2776
2777 vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
2778 if (!vif_nest)
2779 return false;
2780
2781 if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif_dev->ifindex) ||
2782 nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2783 nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2784 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2785 IPMRA_VIFA_PAD) ||
2786 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2787 IPMRA_VIFA_PAD) ||
2788 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2789 IPMRA_VIFA_PAD) ||
2790 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2791 IPMRA_VIFA_PAD) ||
2792 nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2793 nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2794 nla_nest_cancel(skb, vif_nest);
2795 return false;
2796 }
2797 nla_nest_end(skb, vif_nest);
2798
2799 return true;
2800}
2801
2802static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
2803 struct netlink_ext_ack *extack)
2804{
2805 struct ifinfomsg *ifm;
2806
2807 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2808 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
2809 return -EINVAL;
2810 }
2811
2812 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
2813 NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
2814 return -EINVAL;
2815 }
2816
2817 ifm = nlmsg_data(nlh);
2818 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2819 ifm->ifi_change || ifm->ifi_index) {
2820 NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
2821 return -EINVAL;
2822 }
2823
2824 return 0;
2825}
2826
2827static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2828{
2829 struct net *net = sock_net(skb->sk);
2830 struct nlmsghdr *nlh = NULL;
2831 unsigned int t = 0, s_t;
2832 unsigned int e = 0, s_e;
2833 struct mr_table *mrt;
2834
2835 if (cb->strict_check) {
2836 int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
2837
2838 if (err < 0)
2839 return err;
2840 }
2841
2842 s_t = cb->args[0];
2843 s_e = cb->args[1];
2844
2845 ipmr_for_each_table(mrt, net) {
2846 struct nlattr *vifs, *af;
2847 struct ifinfomsg *hdr;
2848 u32 i;
2849
2850 if (t < s_t)
2851 goto skip_table;
2852 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2853 cb->nlh->nlmsg_seq, RTM_NEWLINK,
2854 sizeof(*hdr), NLM_F_MULTI);
2855 if (!nlh)
2856 break;
2857
2858 hdr = nlmsg_data(nlh);
2859 memset(hdr, 0, sizeof(*hdr));
2860 hdr->ifi_family = RTNL_FAMILY_IPMR;
2861
2862 af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
2863 if (!af) {
2864 nlmsg_cancel(skb, nlh);
2865 goto out;
2866 }
2867
2868 if (!ipmr_fill_table(mrt, skb)) {
2869 nlmsg_cancel(skb, nlh);
2870 goto out;
2871 }
2872
2873 vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
2874 if (!vifs) {
2875 nla_nest_end(skb, af);
2876 nlmsg_end(skb, nlh);
2877 goto out;
2878 }
2879 for (i = 0; i < mrt->maxvif; i++) {
2880 if (e < s_e)
2881 goto skip_entry;
2882 if (!ipmr_fill_vif(mrt, i, skb)) {
2883 nla_nest_end(skb, vifs);
2884 nla_nest_end(skb, af);
2885 nlmsg_end(skb, nlh);
2886 goto out;
2887 }
2888skip_entry:
2889 e++;
2890 }
2891 s_e = 0;
2892 e = 0;
2893 nla_nest_end(skb, vifs);
2894 nla_nest_end(skb, af);
2895 nlmsg_end(skb, nlh);
2896skip_table:
2897 t++;
2898 }
2899
2900out:
2901 cb->args[1] = e;
2902 cb->args[0] = t;
2903
2904 return skb->len;
2905}
2906
2907#ifdef CONFIG_PROC_FS
2908/* The /proc interfaces to multicast routing :
2909 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2910 */
2911
2912static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2913 __acquires(RCU)
2914{
2915 struct mr_vif_iter *iter = seq->private;
2916 struct net *net = seq_file_net(seq);
2917 struct mr_table *mrt;
2918
2919 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2920 if (!mrt)
2921 return ERR_PTR(-ENOENT);
2922
2923 iter->mrt = mrt;
2924
2925 rcu_read_lock();
2926 return mr_vif_seq_start(seq, pos);
2927}
2928
2929static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2930 __releases(RCU)
2931{
2932 rcu_read_unlock();
2933}
2934
2935static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2936{
2937 struct mr_vif_iter *iter = seq->private;
2938 struct mr_table *mrt = iter->mrt;
2939
2940 if (v == SEQ_START_TOKEN) {
2941 seq_puts(seq,
2942 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2943 } else {
2944 const struct vif_device *vif = v;
2945 const struct net_device *vif_dev;
2946 const char *name;
2947
2948 vif_dev = vif_dev_read(vif);
2949 name = vif_dev ? vif_dev->name : "none";
2950 seq_printf(seq,
2951 "%2td %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2952 vif - mrt->vif_table,
2953 name, vif->bytes_in, vif->pkt_in,
2954 vif->bytes_out, vif->pkt_out,
2955 vif->flags, vif->local, vif->remote);
2956 }
2957 return 0;
2958}
2959
2960static const struct seq_operations ipmr_vif_seq_ops = {
2961 .start = ipmr_vif_seq_start,
2962 .next = mr_vif_seq_next,
2963 .stop = ipmr_vif_seq_stop,
2964 .show = ipmr_vif_seq_show,
2965};
2966
2967static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2968{
2969 struct net *net = seq_file_net(seq);
2970 struct mr_table *mrt;
2971
2972 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2973 if (!mrt)
2974 return ERR_PTR(-ENOENT);
2975
2976 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
2977}
2978
2979static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2980{
2981 int n;
2982
2983 if (v == SEQ_START_TOKEN) {
2984 seq_puts(seq,
2985 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2986 } else {
2987 const struct mfc_cache *mfc = v;
2988 const struct mr_mfc_iter *it = seq->private;
2989 const struct mr_table *mrt = it->mrt;
2990
2991 seq_printf(seq, "%08X %08X %-3hd",
2992 (__force u32) mfc->mfc_mcastgrp,
2993 (__force u32) mfc->mfc_origin,
2994 mfc->_c.mfc_parent);
2995
2996 if (it->cache != &mrt->mfc_unres_queue) {
2997 seq_printf(seq, " %8lu %8lu %8lu",
2998 mfc->_c.mfc_un.res.pkt,
2999 mfc->_c.mfc_un.res.bytes,
3000 mfc->_c.mfc_un.res.wrong_if);
3001 for (n = mfc->_c.mfc_un.res.minvif;
3002 n < mfc->_c.mfc_un.res.maxvif; n++) {
3003 if (VIF_EXISTS(mrt, n) &&
3004 mfc->_c.mfc_un.res.ttls[n] < 255)
3005 seq_printf(seq,
3006 " %2d:%-3d",
3007 n, mfc->_c.mfc_un.res.ttls[n]);
3008 }
3009 } else {
3010 /* unresolved mfc_caches don't contain
3011 * pkt, bytes and wrong_if values
3012 */
3013 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
3014 }
3015 seq_putc(seq, '\n');
3016 }
3017 return 0;
3018}
3019
3020static const struct seq_operations ipmr_mfc_seq_ops = {
3021 .start = ipmr_mfc_seq_start,
3022 .next = mr_mfc_seq_next,
3023 .stop = mr_mfc_seq_stop,
3024 .show = ipmr_mfc_seq_show,
3025};
3026#endif
3027
3028#ifdef CONFIG_IP_PIMSM_V2
3029static const struct net_protocol pim_protocol = {
3030 .handler = pim_rcv,
3031};
3032#endif
3033
3034static unsigned int ipmr_seq_read(struct net *net)
3035{
3036 ASSERT_RTNL();
3037
3038 return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
3039}
3040
3041static int ipmr_dump(struct net *net, struct notifier_block *nb,
3042 struct netlink_ext_ack *extack)
3043{
3044 return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
3045 ipmr_mr_table_iter, extack);
3046}
3047
3048static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3049 .family = RTNL_FAMILY_IPMR,
3050 .fib_seq_read = ipmr_seq_read,
3051 .fib_dump = ipmr_dump,
3052 .owner = THIS_MODULE,
3053};
3054
3055static int __net_init ipmr_notifier_init(struct net *net)
3056{
3057 struct fib_notifier_ops *ops;
3058
3059 net->ipv4.ipmr_seq = 0;
3060
3061 ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3062 if (IS_ERR(ops))
3063 return PTR_ERR(ops);
3064 net->ipv4.ipmr_notifier_ops = ops;
3065
3066 return 0;
3067}
3068
3069static void __net_exit ipmr_notifier_exit(struct net *net)
3070{
3071 fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3072 net->ipv4.ipmr_notifier_ops = NULL;
3073}
3074
3075/* Setup for IP multicast routing */
3076static int __net_init ipmr_net_init(struct net *net)
3077{
3078 int err;
3079
3080 err = ipmr_notifier_init(net);
3081 if (err)
3082 goto ipmr_notifier_fail;
3083
3084 err = ipmr_rules_init(net);
3085 if (err < 0)
3086 goto ipmr_rules_fail;
3087
3088#ifdef CONFIG_PROC_FS
3089 err = -ENOMEM;
3090 if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3091 sizeof(struct mr_vif_iter)))
3092 goto proc_vif_fail;
3093 if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3094 sizeof(struct mr_mfc_iter)))
3095 goto proc_cache_fail;
3096#endif
3097 return 0;
3098
3099#ifdef CONFIG_PROC_FS
3100proc_cache_fail:
3101 remove_proc_entry("ip_mr_vif", net->proc_net);
3102proc_vif_fail:
3103 rtnl_lock();
3104 ipmr_rules_exit(net);
3105 rtnl_unlock();
3106#endif
3107ipmr_rules_fail:
3108 ipmr_notifier_exit(net);
3109ipmr_notifier_fail:
3110 return err;
3111}
3112
3113static void __net_exit ipmr_net_exit(struct net *net)
3114{
3115#ifdef CONFIG_PROC_FS
3116 remove_proc_entry("ip_mr_cache", net->proc_net);
3117 remove_proc_entry("ip_mr_vif", net->proc_net);
3118#endif
3119 ipmr_notifier_exit(net);
3120}
3121
3122static void __net_exit ipmr_net_exit_batch(struct list_head *net_list)
3123{
3124 struct net *net;
3125
3126 rtnl_lock();
3127 list_for_each_entry(net, net_list, exit_list)
3128 ipmr_rules_exit(net);
3129 rtnl_unlock();
3130}
3131
3132static struct pernet_operations ipmr_net_ops = {
3133 .init = ipmr_net_init,
3134 .exit = ipmr_net_exit,
3135 .exit_batch = ipmr_net_exit_batch,
3136};
3137
3138int __init ip_mr_init(void)
3139{
3140 int err;
3141
3142 mrt_cachep = kmem_cache_create("ip_mrt_cache",
3143 sizeof(struct mfc_cache),
3144 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
3145 NULL);
3146
3147 err = register_pernet_subsys(&ipmr_net_ops);
3148 if (err)
3149 goto reg_pernet_fail;
3150
3151 err = register_netdevice_notifier(&ip_mr_notifier);
3152 if (err)
3153 goto reg_notif_fail;
3154#ifdef CONFIG_IP_PIMSM_V2
3155 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3156 pr_err("%s: can't add PIM protocol\n", __func__);
3157 err = -EAGAIN;
3158 goto add_proto_fail;
3159 }
3160#endif
3161 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
3162 ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
3163 rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
3164 ipmr_rtm_route, NULL, 0);
3165 rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
3166 ipmr_rtm_route, NULL, 0);
3167
3168 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
3169 NULL, ipmr_rtm_dumplink, 0);
3170 return 0;
3171
3172#ifdef CONFIG_IP_PIMSM_V2
3173add_proto_fail:
3174 unregister_netdevice_notifier(&ip_mr_notifier);
3175#endif
3176reg_notif_fail:
3177 unregister_pernet_subsys(&ipmr_net_ops);
3178reg_pernet_fail:
3179 kmem_cache_destroy(mrt_cachep);
3180 return err;
3181}