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#include <net/inet_dscp.h>
66
67#include <linux/nospec.h>
68
69struct ipmr_rule {
70 struct fib_rule common;
71};
72
73struct ipmr_result {
74 struct mr_table *mrt;
75};
76
77/* Big lock, protecting vif table, mrt cache and mroute socket state.
78 * Note that the changes are semaphored via rtnl_lock.
79 */
80
81static DEFINE_SPINLOCK(mrt_lock);
82
83static struct net_device *vif_dev_read(const struct vif_device *vif)
84{
85 return rcu_dereference(vif->dev);
86}
87
88/* Multicast router control variables */
89
90/* Special spinlock for queue of unresolved entries */
91static DEFINE_SPINLOCK(mfc_unres_lock);
92
93/* We return to original Alan's scheme. Hash table of resolved
94 * entries is changed only in process context and protected
95 * with weak lock mrt_lock. Queue of unresolved entries is protected
96 * with strong spinlock mfc_unres_lock.
97 *
98 * In this case data path is free of exclusive locks at all.
99 */
100
101static struct kmem_cache *mrt_cachep __ro_after_init;
102
103static struct mr_table *ipmr_new_table(struct net *net, u32 id);
104static void ipmr_free_table(struct mr_table *mrt);
105
106static void ip_mr_forward(struct net *net, struct mr_table *mrt,
107 struct net_device *dev, struct sk_buff *skb,
108 struct mfc_cache *cache, int local);
109static int ipmr_cache_report(const struct mr_table *mrt,
110 struct sk_buff *pkt, vifi_t vifi, int assert);
111static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
112 int cmd);
113static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt);
114static void mroute_clean_tables(struct mr_table *mrt, int flags);
115static void ipmr_expire_process(struct timer_list *t);
116
117#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
118#define ipmr_for_each_table(mrt, net) \
119 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list, \
120 lockdep_rtnl_is_held() || \
121 list_empty(&net->ipv4.mr_tables))
122
123static bool ipmr_can_free_table(struct net *net)
124{
125 return !check_net(net) || !net_initialized(net);
126}
127
128static struct mr_table *ipmr_mr_table_iter(struct net *net,
129 struct mr_table *mrt)
130{
131 struct mr_table *ret;
132
133 if (!mrt)
134 ret = list_entry_rcu(net->ipv4.mr_tables.next,
135 struct mr_table, list);
136 else
137 ret = list_entry_rcu(mrt->list.next,
138 struct mr_table, list);
139
140 if (&ret->list == &net->ipv4.mr_tables)
141 return NULL;
142 return ret;
143}
144
145static struct mr_table *__ipmr_get_table(struct net *net, u32 id)
146{
147 struct mr_table *mrt;
148
149 ipmr_for_each_table(mrt, net) {
150 if (mrt->id == id)
151 return mrt;
152 }
153 return NULL;
154}
155
156static struct mr_table *ipmr_get_table(struct net *net, u32 id)
157{
158 struct mr_table *mrt;
159
160 rcu_read_lock();
161 mrt = __ipmr_get_table(net, id);
162 rcu_read_unlock();
163 return mrt;
164}
165
166static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
167 struct mr_table **mrt)
168{
169 int err;
170 struct ipmr_result res;
171 struct fib_lookup_arg arg = {
172 .result = &res,
173 .flags = FIB_LOOKUP_NOREF,
174 };
175
176 /* update flow if oif or iif point to device enslaved to l3mdev */
177 l3mdev_update_flow(net, flowi4_to_flowi(flp4));
178
179 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
180 flowi4_to_flowi(flp4), 0, &arg);
181 if (err < 0)
182 return err;
183 *mrt = res.mrt;
184 return 0;
185}
186
187static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
188 int flags, struct fib_lookup_arg *arg)
189{
190 struct ipmr_result *res = arg->result;
191 struct mr_table *mrt;
192
193 switch (rule->action) {
194 case FR_ACT_TO_TBL:
195 break;
196 case FR_ACT_UNREACHABLE:
197 return -ENETUNREACH;
198 case FR_ACT_PROHIBIT:
199 return -EACCES;
200 case FR_ACT_BLACKHOLE:
201 default:
202 return -EINVAL;
203 }
204
205 arg->table = fib_rule_get_table(rule, arg);
206
207 mrt = __ipmr_get_table(rule->fr_net, arg->table);
208 if (!mrt)
209 return -EAGAIN;
210 res->mrt = mrt;
211 return 0;
212}
213
214static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
215{
216 return 1;
217}
218
219static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
220 struct fib_rule_hdr *frh, struct nlattr **tb,
221 struct netlink_ext_ack *extack)
222{
223 return 0;
224}
225
226static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
227 struct nlattr **tb)
228{
229 return 1;
230}
231
232static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
233 struct fib_rule_hdr *frh)
234{
235 frh->dst_len = 0;
236 frh->src_len = 0;
237 frh->tos = 0;
238 return 0;
239}
240
241static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
242 .family = RTNL_FAMILY_IPMR,
243 .rule_size = sizeof(struct ipmr_rule),
244 .addr_size = sizeof(u32),
245 .action = ipmr_rule_action,
246 .match = ipmr_rule_match,
247 .configure = ipmr_rule_configure,
248 .compare = ipmr_rule_compare,
249 .fill = ipmr_rule_fill,
250 .nlgroup = RTNLGRP_IPV4_RULE,
251 .owner = THIS_MODULE,
252};
253
254static int __net_init ipmr_rules_init(struct net *net)
255{
256 struct fib_rules_ops *ops;
257 struct mr_table *mrt;
258 int err;
259
260 ops = fib_rules_register(&ipmr_rules_ops_template, net);
261 if (IS_ERR(ops))
262 return PTR_ERR(ops);
263
264 INIT_LIST_HEAD(&net->ipv4.mr_tables);
265
266 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
267 if (IS_ERR(mrt)) {
268 err = PTR_ERR(mrt);
269 goto err1;
270 }
271
272 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT);
273 if (err < 0)
274 goto err2;
275
276 net->ipv4.mr_rules_ops = ops;
277 return 0;
278
279err2:
280 rtnl_lock();
281 ipmr_free_table(mrt);
282 rtnl_unlock();
283err1:
284 fib_rules_unregister(ops);
285 return err;
286}
287
288static void __net_exit ipmr_rules_exit(struct net *net)
289{
290 struct mr_table *mrt, *next;
291
292 ASSERT_RTNL();
293 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
294 list_del(&mrt->list);
295 ipmr_free_table(mrt);
296 }
297 fib_rules_unregister(net->ipv4.mr_rules_ops);
298}
299
300static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
301 struct netlink_ext_ack *extack)
302{
303 return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR, extack);
304}
305
306static unsigned int ipmr_rules_seq_read(const struct net *net)
307{
308 return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
309}
310
311bool ipmr_rule_default(const struct fib_rule *rule)
312{
313 return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
314}
315EXPORT_SYMBOL(ipmr_rule_default);
316#else
317#define ipmr_for_each_table(mrt, net) \
318 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
319
320static bool ipmr_can_free_table(struct net *net)
321{
322 return !check_net(net);
323}
324
325static struct mr_table *ipmr_mr_table_iter(struct net *net,
326 struct mr_table *mrt)
327{
328 if (!mrt)
329 return net->ipv4.mrt;
330 return NULL;
331}
332
333static struct mr_table *ipmr_get_table(struct net *net, u32 id)
334{
335 return net->ipv4.mrt;
336}
337
338#define __ipmr_get_table ipmr_get_table
339
340static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
341 struct mr_table **mrt)
342{
343 *mrt = net->ipv4.mrt;
344 return 0;
345}
346
347static int __net_init ipmr_rules_init(struct net *net)
348{
349 struct mr_table *mrt;
350
351 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
352 if (IS_ERR(mrt))
353 return PTR_ERR(mrt);
354 net->ipv4.mrt = mrt;
355 return 0;
356}
357
358static void __net_exit ipmr_rules_exit(struct net *net)
359{
360 ASSERT_RTNL();
361 ipmr_free_table(net->ipv4.mrt);
362 net->ipv4.mrt = NULL;
363}
364
365static int ipmr_rules_dump(struct net *net, struct notifier_block *nb,
366 struct netlink_ext_ack *extack)
367{
368 return 0;
369}
370
371static unsigned int ipmr_rules_seq_read(const struct net *net)
372{
373 return 0;
374}
375
376bool ipmr_rule_default(const struct fib_rule *rule)
377{
378 return true;
379}
380EXPORT_SYMBOL(ipmr_rule_default);
381#endif
382
383static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
384 const void *ptr)
385{
386 const struct mfc_cache_cmp_arg *cmparg = arg->key;
387 const struct mfc_cache *c = ptr;
388
389 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
390 cmparg->mfc_origin != c->mfc_origin;
391}
392
393static const struct rhashtable_params ipmr_rht_params = {
394 .head_offset = offsetof(struct mr_mfc, mnode),
395 .key_offset = offsetof(struct mfc_cache, cmparg),
396 .key_len = sizeof(struct mfc_cache_cmp_arg),
397 .nelem_hint = 3,
398 .obj_cmpfn = ipmr_hash_cmp,
399 .automatic_shrinking = true,
400};
401
402static void ipmr_new_table_set(struct mr_table *mrt,
403 struct net *net)
404{
405#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
406 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
407#endif
408}
409
410static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
411 .mfc_mcastgrp = htonl(INADDR_ANY),
412 .mfc_origin = htonl(INADDR_ANY),
413};
414
415static struct mr_table_ops ipmr_mr_table_ops = {
416 .rht_params = &ipmr_rht_params,
417 .cmparg_any = &ipmr_mr_table_ops_cmparg_any,
418};
419
420static struct mr_table *ipmr_new_table(struct net *net, u32 id)
421{
422 struct mr_table *mrt;
423
424 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
425 if (id != RT_TABLE_DEFAULT && id >= 1000000000)
426 return ERR_PTR(-EINVAL);
427
428 mrt = __ipmr_get_table(net, id);
429 if (mrt)
430 return mrt;
431
432 return mr_table_alloc(net, id, &ipmr_mr_table_ops,
433 ipmr_expire_process, ipmr_new_table_set);
434}
435
436static void ipmr_free_table(struct mr_table *mrt)
437{
438 struct net *net = read_pnet(&mrt->net);
439
440 WARN_ON_ONCE(!ipmr_can_free_table(net));
441
442 timer_shutdown_sync(&mrt->ipmr_expire_timer);
443 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
444 MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC);
445 rhltable_destroy(&mrt->mfc_hash);
446 kfree(mrt);
447}
448
449/* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
450
451/* Initialize ipmr pimreg/tunnel in_device */
452static bool ipmr_init_vif_indev(const struct net_device *dev)
453{
454 struct in_device *in_dev;
455
456 ASSERT_RTNL();
457
458 in_dev = __in_dev_get_rtnl(dev);
459 if (!in_dev)
460 return false;
461 ipv4_devconf_setall(in_dev);
462 neigh_parms_data_state_setall(in_dev->arp_parms);
463 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
464
465 return true;
466}
467
468static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
469{
470 struct net_device *tunnel_dev, *new_dev;
471 struct ip_tunnel_parm_kern p = { };
472 int err;
473
474 tunnel_dev = __dev_get_by_name(net, "tunl0");
475 if (!tunnel_dev)
476 goto out;
477
478 p.iph.daddr = v->vifc_rmt_addr.s_addr;
479 p.iph.saddr = v->vifc_lcl_addr.s_addr;
480 p.iph.version = 4;
481 p.iph.ihl = 5;
482 p.iph.protocol = IPPROTO_IPIP;
483 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
484
485 if (!tunnel_dev->netdev_ops->ndo_tunnel_ctl)
486 goto out;
487 err = tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
488 SIOCADDTUNNEL);
489 if (err)
490 goto out;
491
492 new_dev = __dev_get_by_name(net, p.name);
493 if (!new_dev)
494 goto out;
495
496 new_dev->flags |= IFF_MULTICAST;
497 if (!ipmr_init_vif_indev(new_dev))
498 goto out_unregister;
499 if (dev_open(new_dev, NULL))
500 goto out_unregister;
501 dev_hold(new_dev);
502 err = dev_set_allmulti(new_dev, 1);
503 if (err) {
504 dev_close(new_dev);
505 tunnel_dev->netdev_ops->ndo_tunnel_ctl(tunnel_dev, &p,
506 SIOCDELTUNNEL);
507 dev_put(new_dev);
508 new_dev = ERR_PTR(err);
509 }
510 return new_dev;
511
512out_unregister:
513 unregister_netdevice(new_dev);
514out:
515 return ERR_PTR(-ENOBUFS);
516}
517
518#if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
519static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
520{
521 struct net *net = dev_net(dev);
522 struct mr_table *mrt;
523 struct flowi4 fl4 = {
524 .flowi4_oif = dev->ifindex,
525 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
526 .flowi4_mark = skb->mark,
527 };
528 int err;
529
530 err = ipmr_fib_lookup(net, &fl4, &mrt);
531 if (err < 0) {
532 kfree_skb(skb);
533 return err;
534 }
535
536 DEV_STATS_ADD(dev, tx_bytes, skb->len);
537 DEV_STATS_INC(dev, tx_packets);
538 rcu_read_lock();
539
540 /* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
541 ipmr_cache_report(mrt, skb, READ_ONCE(mrt->mroute_reg_vif_num),
542 IGMPMSG_WHOLEPKT);
543
544 rcu_read_unlock();
545 kfree_skb(skb);
546 return NETDEV_TX_OK;
547}
548
549static int reg_vif_get_iflink(const struct net_device *dev)
550{
551 return 0;
552}
553
554static const struct net_device_ops reg_vif_netdev_ops = {
555 .ndo_start_xmit = reg_vif_xmit,
556 .ndo_get_iflink = reg_vif_get_iflink,
557};
558
559static void reg_vif_setup(struct net_device *dev)
560{
561 dev->type = ARPHRD_PIMREG;
562 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
563 dev->flags = IFF_NOARP;
564 dev->netdev_ops = ®_vif_netdev_ops;
565 dev->needs_free_netdev = true;
566 dev->netns_local = true;
567}
568
569static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
570{
571 struct net_device *dev;
572 char name[IFNAMSIZ];
573
574 if (mrt->id == RT_TABLE_DEFAULT)
575 sprintf(name, "pimreg");
576 else
577 sprintf(name, "pimreg%u", mrt->id);
578
579 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
580
581 if (!dev)
582 return NULL;
583
584 dev_net_set(dev, net);
585
586 if (register_netdevice(dev)) {
587 free_netdev(dev);
588 return NULL;
589 }
590
591 if (!ipmr_init_vif_indev(dev))
592 goto failure;
593 if (dev_open(dev, NULL))
594 goto failure;
595
596 dev_hold(dev);
597
598 return dev;
599
600failure:
601 unregister_netdevice(dev);
602 return NULL;
603}
604
605/* called with rcu_read_lock() */
606static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
607 unsigned int pimlen)
608{
609 struct net_device *reg_dev = NULL;
610 struct iphdr *encap;
611 int vif_num;
612
613 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
614 /* Check that:
615 * a. packet is really sent to a multicast group
616 * b. packet is not a NULL-REGISTER
617 * c. packet is not truncated
618 */
619 if (!ipv4_is_multicast(encap->daddr) ||
620 encap->tot_len == 0 ||
621 ntohs(encap->tot_len) + pimlen > skb->len)
622 return 1;
623
624 /* Pairs with WRITE_ONCE() in vif_add()/vid_delete() */
625 vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
626 if (vif_num >= 0)
627 reg_dev = vif_dev_read(&mrt->vif_table[vif_num]);
628 if (!reg_dev)
629 return 1;
630
631 skb->mac_header = skb->network_header;
632 skb_pull(skb, (u8 *)encap - skb->data);
633 skb_reset_network_header(skb);
634 skb->protocol = htons(ETH_P_IP);
635 skb->ip_summed = CHECKSUM_NONE;
636
637 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
638
639 netif_rx(skb);
640
641 return NET_RX_SUCCESS;
642}
643#else
644static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
645{
646 return NULL;
647}
648#endif
649
650static int call_ipmr_vif_entry_notifiers(struct net *net,
651 enum fib_event_type event_type,
652 struct vif_device *vif,
653 struct net_device *vif_dev,
654 vifi_t vif_index, u32 tb_id)
655{
656 return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
657 vif, vif_dev, vif_index, tb_id,
658 &net->ipv4.ipmr_seq);
659}
660
661static int call_ipmr_mfc_entry_notifiers(struct net *net,
662 enum fib_event_type event_type,
663 struct mfc_cache *mfc, u32 tb_id)
664{
665 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
666 &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
667}
668
669/**
670 * vif_delete - Delete a VIF entry
671 * @mrt: Table to delete from
672 * @vifi: VIF identifier to delete
673 * @notify: Set to 1, if the caller is a notifier_call
674 * @head: if unregistering the VIF, place it on this queue
675 */
676static int vif_delete(struct mr_table *mrt, int vifi, int notify,
677 struct list_head *head)
678{
679 struct net *net = read_pnet(&mrt->net);
680 struct vif_device *v;
681 struct net_device *dev;
682 struct in_device *in_dev;
683
684 if (vifi < 0 || vifi >= mrt->maxvif)
685 return -EADDRNOTAVAIL;
686
687 v = &mrt->vif_table[vifi];
688
689 dev = rtnl_dereference(v->dev);
690 if (!dev)
691 return -EADDRNOTAVAIL;
692
693 spin_lock(&mrt_lock);
694 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, dev,
695 vifi, mrt->id);
696 RCU_INIT_POINTER(v->dev, NULL);
697
698 if (vifi == mrt->mroute_reg_vif_num) {
699 /* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
700 WRITE_ONCE(mrt->mroute_reg_vif_num, -1);
701 }
702 if (vifi + 1 == mrt->maxvif) {
703 int tmp;
704
705 for (tmp = vifi - 1; tmp >= 0; tmp--) {
706 if (VIF_EXISTS(mrt, tmp))
707 break;
708 }
709 WRITE_ONCE(mrt->maxvif, tmp + 1);
710 }
711
712 spin_unlock(&mrt_lock);
713
714 dev_set_allmulti(dev, -1);
715
716 in_dev = __in_dev_get_rtnl(dev);
717 if (in_dev) {
718 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
719 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
720 NETCONFA_MC_FORWARDING,
721 dev->ifindex, &in_dev->cnf);
722 ip_rt_multicast_event(in_dev);
723 }
724
725 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
726 unregister_netdevice_queue(dev, head);
727
728 netdev_put(dev, &v->dev_tracker);
729 return 0;
730}
731
732static void ipmr_cache_free_rcu(struct rcu_head *head)
733{
734 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
735
736 kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
737}
738
739static void ipmr_cache_free(struct mfc_cache *c)
740{
741 call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
742}
743
744/* Destroy an unresolved cache entry, killing queued skbs
745 * and reporting error to netlink readers.
746 */
747static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
748{
749 struct net *net = read_pnet(&mrt->net);
750 struct sk_buff *skb;
751 struct nlmsgerr *e;
752
753 atomic_dec(&mrt->cache_resolve_queue_len);
754
755 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
756 if (ip_hdr(skb)->version == 0) {
757 struct nlmsghdr *nlh = skb_pull(skb,
758 sizeof(struct iphdr));
759 nlh->nlmsg_type = NLMSG_ERROR;
760 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
761 skb_trim(skb, nlh->nlmsg_len);
762 e = nlmsg_data(nlh);
763 e->error = -ETIMEDOUT;
764 memset(&e->msg, 0, sizeof(e->msg));
765
766 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
767 } else {
768 kfree_skb(skb);
769 }
770 }
771
772 ipmr_cache_free(c);
773}
774
775/* Timer process for the unresolved queue. */
776static void ipmr_expire_process(struct timer_list *t)
777{
778 struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
779 struct mr_mfc *c, *next;
780 unsigned long expires;
781 unsigned long now;
782
783 if (!spin_trylock(&mfc_unres_lock)) {
784 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
785 return;
786 }
787
788 if (list_empty(&mrt->mfc_unres_queue))
789 goto out;
790
791 now = jiffies;
792 expires = 10*HZ;
793
794 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
795 if (time_after(c->mfc_un.unres.expires, now)) {
796 unsigned long interval = c->mfc_un.unres.expires - now;
797 if (interval < expires)
798 expires = interval;
799 continue;
800 }
801
802 list_del(&c->list);
803 mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
804 ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
805 }
806
807 if (!list_empty(&mrt->mfc_unres_queue))
808 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
809
810out:
811 spin_unlock(&mfc_unres_lock);
812}
813
814/* Fill oifs list. It is called under locked mrt_lock. */
815static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
816 unsigned char *ttls)
817{
818 int vifi;
819
820 cache->mfc_un.res.minvif = MAXVIFS;
821 cache->mfc_un.res.maxvif = 0;
822 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
823
824 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
825 if (VIF_EXISTS(mrt, vifi) &&
826 ttls[vifi] && ttls[vifi] < 255) {
827 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
828 if (cache->mfc_un.res.minvif > vifi)
829 cache->mfc_un.res.minvif = vifi;
830 if (cache->mfc_un.res.maxvif <= vifi)
831 cache->mfc_un.res.maxvif = vifi + 1;
832 }
833 }
834 WRITE_ONCE(cache->mfc_un.res.lastuse, jiffies);
835}
836
837static int vif_add(struct net *net, struct mr_table *mrt,
838 struct vifctl *vifc, int mrtsock)
839{
840 struct netdev_phys_item_id ppid = { };
841 int vifi = vifc->vifc_vifi;
842 struct vif_device *v = &mrt->vif_table[vifi];
843 struct net_device *dev;
844 struct in_device *in_dev;
845 int err;
846
847 /* Is vif busy ? */
848 if (VIF_EXISTS(mrt, vifi))
849 return -EADDRINUSE;
850
851 switch (vifc->vifc_flags) {
852 case VIFF_REGISTER:
853 if (!ipmr_pimsm_enabled())
854 return -EINVAL;
855 /* Special Purpose VIF in PIM
856 * All the packets will be sent to the daemon
857 */
858 if (mrt->mroute_reg_vif_num >= 0)
859 return -EADDRINUSE;
860 dev = ipmr_reg_vif(net, mrt);
861 if (!dev)
862 return -ENOBUFS;
863 err = dev_set_allmulti(dev, 1);
864 if (err) {
865 unregister_netdevice(dev);
866 dev_put(dev);
867 return err;
868 }
869 break;
870 case VIFF_TUNNEL:
871 dev = ipmr_new_tunnel(net, vifc);
872 if (IS_ERR(dev))
873 return PTR_ERR(dev);
874 break;
875 case VIFF_USE_IFINDEX:
876 case 0:
877 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
878 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
879 if (dev && !__in_dev_get_rtnl(dev)) {
880 dev_put(dev);
881 return -EADDRNOTAVAIL;
882 }
883 } else {
884 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
885 }
886 if (!dev)
887 return -EADDRNOTAVAIL;
888 err = dev_set_allmulti(dev, 1);
889 if (err) {
890 dev_put(dev);
891 return err;
892 }
893 break;
894 default:
895 return -EINVAL;
896 }
897
898 in_dev = __in_dev_get_rtnl(dev);
899 if (!in_dev) {
900 dev_put(dev);
901 return -EADDRNOTAVAIL;
902 }
903 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
904 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
905 dev->ifindex, &in_dev->cnf);
906 ip_rt_multicast_event(in_dev);
907
908 /* Fill in the VIF structures */
909 vif_device_init(v, dev, vifc->vifc_rate_limit,
910 vifc->vifc_threshold,
911 vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
912 (VIFF_TUNNEL | VIFF_REGISTER));
913
914 err = dev_get_port_parent_id(dev, &ppid, true);
915 if (err == 0) {
916 memcpy(v->dev_parent_id.id, ppid.id, ppid.id_len);
917 v->dev_parent_id.id_len = ppid.id_len;
918 } else {
919 v->dev_parent_id.id_len = 0;
920 }
921
922 v->local = vifc->vifc_lcl_addr.s_addr;
923 v->remote = vifc->vifc_rmt_addr.s_addr;
924
925 /* And finish update writing critical data */
926 spin_lock(&mrt_lock);
927 rcu_assign_pointer(v->dev, dev);
928 netdev_tracker_alloc(dev, &v->dev_tracker, GFP_ATOMIC);
929 if (v->flags & VIFF_REGISTER) {
930 /* Pairs with READ_ONCE() in ipmr_cache_report() and reg_vif_xmit() */
931 WRITE_ONCE(mrt->mroute_reg_vif_num, vifi);
932 }
933 if (vifi+1 > mrt->maxvif)
934 WRITE_ONCE(mrt->maxvif, vifi + 1);
935 spin_unlock(&mrt_lock);
936 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, dev,
937 vifi, mrt->id);
938 return 0;
939}
940
941/* called with rcu_read_lock() */
942static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
943 __be32 origin,
944 __be32 mcastgrp)
945{
946 struct mfc_cache_cmp_arg arg = {
947 .mfc_mcastgrp = mcastgrp,
948 .mfc_origin = origin
949 };
950
951 return mr_mfc_find(mrt, &arg);
952}
953
954/* Look for a (*,G) entry */
955static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
956 __be32 mcastgrp, int vifi)
957{
958 struct mfc_cache_cmp_arg arg = {
959 .mfc_mcastgrp = mcastgrp,
960 .mfc_origin = htonl(INADDR_ANY)
961 };
962
963 if (mcastgrp == htonl(INADDR_ANY))
964 return mr_mfc_find_any_parent(mrt, vifi);
965 return mr_mfc_find_any(mrt, vifi, &arg);
966}
967
968/* Look for a (S,G,iif) entry if parent != -1 */
969static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
970 __be32 origin, __be32 mcastgrp,
971 int parent)
972{
973 struct mfc_cache_cmp_arg arg = {
974 .mfc_mcastgrp = mcastgrp,
975 .mfc_origin = origin,
976 };
977
978 return mr_mfc_find_parent(mrt, &arg, parent);
979}
980
981/* Allocate a multicast cache entry */
982static struct mfc_cache *ipmr_cache_alloc(void)
983{
984 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
985
986 if (c) {
987 c->_c.mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
988 c->_c.mfc_un.res.minvif = MAXVIFS;
989 c->_c.free = ipmr_cache_free_rcu;
990 refcount_set(&c->_c.mfc_un.res.refcount, 1);
991 }
992 return c;
993}
994
995static struct mfc_cache *ipmr_cache_alloc_unres(void)
996{
997 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
998
999 if (c) {
1000 skb_queue_head_init(&c->_c.mfc_un.unres.unresolved);
1001 c->_c.mfc_un.unres.expires = jiffies + 10 * HZ;
1002 }
1003 return c;
1004}
1005
1006/* A cache entry has gone into a resolved state from queued */
1007static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
1008 struct mfc_cache *uc, struct mfc_cache *c)
1009{
1010 struct sk_buff *skb;
1011 struct nlmsgerr *e;
1012
1013 /* Play the pending entries through our router */
1014 while ((skb = __skb_dequeue(&uc->_c.mfc_un.unres.unresolved))) {
1015 if (ip_hdr(skb)->version == 0) {
1016 struct nlmsghdr *nlh = skb_pull(skb,
1017 sizeof(struct iphdr));
1018
1019 if (mr_fill_mroute(mrt, skb, &c->_c,
1020 nlmsg_data(nlh)) > 0) {
1021 nlh->nlmsg_len = skb_tail_pointer(skb) -
1022 (u8 *)nlh;
1023 } else {
1024 nlh->nlmsg_type = NLMSG_ERROR;
1025 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
1026 skb_trim(skb, nlh->nlmsg_len);
1027 e = nlmsg_data(nlh);
1028 e->error = -EMSGSIZE;
1029 memset(&e->msg, 0, sizeof(e->msg));
1030 }
1031
1032 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
1033 } else {
1034 rcu_read_lock();
1035 ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1036 rcu_read_unlock();
1037 }
1038 }
1039}
1040
1041/* Bounce a cache query up to mrouted and netlink.
1042 *
1043 * Called under rcu_read_lock().
1044 */
1045static int ipmr_cache_report(const struct mr_table *mrt,
1046 struct sk_buff *pkt, vifi_t vifi, int assert)
1047{
1048 const int ihl = ip_hdrlen(pkt);
1049 struct sock *mroute_sk;
1050 struct igmphdr *igmp;
1051 struct igmpmsg *msg;
1052 struct sk_buff *skb;
1053 int ret;
1054
1055 mroute_sk = rcu_dereference(mrt->mroute_sk);
1056 if (!mroute_sk)
1057 return -EINVAL;
1058
1059 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE)
1060 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1061 else
1062 skb = alloc_skb(128, GFP_ATOMIC);
1063
1064 if (!skb)
1065 return -ENOBUFS;
1066
1067 if (assert == IGMPMSG_WHOLEPKT || assert == IGMPMSG_WRVIFWHOLE) {
1068 /* Ugly, but we have no choice with this interface.
1069 * Duplicate old header, fix ihl, length etc.
1070 * And all this only to mangle msg->im_msgtype and
1071 * to set msg->im_mbz to "mbz" :-)
1072 */
1073 skb_push(skb, sizeof(struct iphdr));
1074 skb_reset_network_header(skb);
1075 skb_reset_transport_header(skb);
1076 msg = (struct igmpmsg *)skb_network_header(skb);
1077 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1078 msg->im_msgtype = assert;
1079 msg->im_mbz = 0;
1080 if (assert == IGMPMSG_WRVIFWHOLE) {
1081 msg->im_vif = vifi;
1082 msg->im_vif_hi = vifi >> 8;
1083 } else {
1084 /* Pairs with WRITE_ONCE() in vif_add() and vif_delete() */
1085 int vif_num = READ_ONCE(mrt->mroute_reg_vif_num);
1086
1087 msg->im_vif = vif_num;
1088 msg->im_vif_hi = vif_num >> 8;
1089 }
1090 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1091 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1092 sizeof(struct iphdr));
1093 } else {
1094 /* Copy the IP header */
1095 skb_set_network_header(skb, skb->len);
1096 skb_put(skb, ihl);
1097 skb_copy_to_linear_data(skb, pkt->data, ihl);
1098 /* Flag to the kernel this is a route add */
1099 ip_hdr(skb)->protocol = 0;
1100 msg = (struct igmpmsg *)skb_network_header(skb);
1101 msg->im_vif = vifi;
1102 msg->im_vif_hi = vifi >> 8;
1103 ipv4_pktinfo_prepare(mroute_sk, pkt, false);
1104 memcpy(skb->cb, pkt->cb, sizeof(skb->cb));
1105 /* Add our header */
1106 igmp = skb_put(skb, sizeof(struct igmphdr));
1107 igmp->type = assert;
1108 msg->im_msgtype = assert;
1109 igmp->code = 0;
1110 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1111 skb->transport_header = skb->network_header;
1112 }
1113
1114 igmpmsg_netlink_event(mrt, skb);
1115
1116 /* Deliver to mrouted */
1117 ret = sock_queue_rcv_skb(mroute_sk, skb);
1118
1119 if (ret < 0) {
1120 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1121 kfree_skb(skb);
1122 }
1123
1124 return ret;
1125}
1126
1127/* Queue a packet for resolution. It gets locked cache entry! */
1128/* Called under rcu_read_lock() */
1129static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1130 struct sk_buff *skb, struct net_device *dev)
1131{
1132 const struct iphdr *iph = ip_hdr(skb);
1133 struct mfc_cache *c;
1134 bool found = false;
1135 int err;
1136
1137 spin_lock_bh(&mfc_unres_lock);
1138 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1139 if (c->mfc_mcastgrp == iph->daddr &&
1140 c->mfc_origin == iph->saddr) {
1141 found = true;
1142 break;
1143 }
1144 }
1145
1146 if (!found) {
1147 /* Create a new entry if allowable */
1148 c = ipmr_cache_alloc_unres();
1149 if (!c) {
1150 spin_unlock_bh(&mfc_unres_lock);
1151
1152 kfree_skb(skb);
1153 return -ENOBUFS;
1154 }
1155
1156 /* Fill in the new cache entry */
1157 c->_c.mfc_parent = -1;
1158 c->mfc_origin = iph->saddr;
1159 c->mfc_mcastgrp = iph->daddr;
1160
1161 /* Reflect first query at mrouted. */
1162 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1163
1164 if (err < 0) {
1165 /* If the report failed throw the cache entry
1166 out - Brad Parker
1167 */
1168 spin_unlock_bh(&mfc_unres_lock);
1169
1170 ipmr_cache_free(c);
1171 kfree_skb(skb);
1172 return err;
1173 }
1174
1175 atomic_inc(&mrt->cache_resolve_queue_len);
1176 list_add(&c->_c.list, &mrt->mfc_unres_queue);
1177 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1178
1179 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1180 mod_timer(&mrt->ipmr_expire_timer,
1181 c->_c.mfc_un.unres.expires);
1182 }
1183
1184 /* See if we can append the packet */
1185 if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1186 kfree_skb(skb);
1187 err = -ENOBUFS;
1188 } else {
1189 if (dev) {
1190 skb->dev = dev;
1191 skb->skb_iif = dev->ifindex;
1192 }
1193 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1194 err = 0;
1195 }
1196
1197 spin_unlock_bh(&mfc_unres_lock);
1198 return err;
1199}
1200
1201/* MFC cache manipulation by user space mroute daemon */
1202
1203static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1204{
1205 struct net *net = read_pnet(&mrt->net);
1206 struct mfc_cache *c;
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 return -ENOENT;
1215 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1216 list_del_rcu(&c->_c.list);
1217 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1218 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1219 mr_cache_put(&c->_c);
1220
1221 return 0;
1222}
1223
1224static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1225 struct mfcctl *mfc, int mrtsock, int parent)
1226{
1227 struct mfc_cache *uc, *c;
1228 struct mr_mfc *_uc;
1229 bool found;
1230 int ret;
1231
1232 if (mfc->mfcc_parent >= MAXVIFS)
1233 return -ENFILE;
1234
1235 /* The entries are added/deleted only under RTNL */
1236 rcu_read_lock();
1237 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1238 mfc->mfcc_mcastgrp.s_addr, parent);
1239 rcu_read_unlock();
1240 if (c) {
1241 spin_lock(&mrt_lock);
1242 c->_c.mfc_parent = mfc->mfcc_parent;
1243 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1244 if (!mrtsock)
1245 c->_c.mfc_flags |= MFC_STATIC;
1246 spin_unlock(&mrt_lock);
1247 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1248 mrt->id);
1249 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1250 return 0;
1251 }
1252
1253 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1254 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1255 return -EINVAL;
1256
1257 c = ipmr_cache_alloc();
1258 if (!c)
1259 return -ENOMEM;
1260
1261 c->mfc_origin = mfc->mfcc_origin.s_addr;
1262 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1263 c->_c.mfc_parent = mfc->mfcc_parent;
1264 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1265 if (!mrtsock)
1266 c->_c.mfc_flags |= MFC_STATIC;
1267
1268 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1269 ipmr_rht_params);
1270 if (ret) {
1271 pr_err("ipmr: rhtable insert error %d\n", ret);
1272 ipmr_cache_free(c);
1273 return ret;
1274 }
1275 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1276 /* Check to see if we resolved a queued list. If so we
1277 * need to send on the frames and tidy up.
1278 */
1279 found = false;
1280 spin_lock_bh(&mfc_unres_lock);
1281 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1282 uc = (struct mfc_cache *)_uc;
1283 if (uc->mfc_origin == c->mfc_origin &&
1284 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1285 list_del(&_uc->list);
1286 atomic_dec(&mrt->cache_resolve_queue_len);
1287 found = true;
1288 break;
1289 }
1290 }
1291 if (list_empty(&mrt->mfc_unres_queue))
1292 del_timer(&mrt->ipmr_expire_timer);
1293 spin_unlock_bh(&mfc_unres_lock);
1294
1295 if (found) {
1296 ipmr_cache_resolve(net, mrt, uc, c);
1297 ipmr_cache_free(uc);
1298 }
1299 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1300 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1301 return 0;
1302}
1303
1304/* Close the multicast socket, and clear the vif tables etc */
1305static void mroute_clean_tables(struct mr_table *mrt, int flags)
1306{
1307 struct net *net = read_pnet(&mrt->net);
1308 struct mr_mfc *c, *tmp;
1309 struct mfc_cache *cache;
1310 LIST_HEAD(list);
1311 int i;
1312
1313 /* Shut down all active vif entries */
1314 if (flags & (MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC)) {
1315 for (i = 0; i < mrt->maxvif; i++) {
1316 if (((mrt->vif_table[i].flags & VIFF_STATIC) &&
1317 !(flags & MRT_FLUSH_VIFS_STATIC)) ||
1318 (!(mrt->vif_table[i].flags & VIFF_STATIC) && !(flags & MRT_FLUSH_VIFS)))
1319 continue;
1320 vif_delete(mrt, i, 0, &list);
1321 }
1322 unregister_netdevice_many(&list);
1323 }
1324
1325 /* Wipe the cache */
1326 if (flags & (MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC)) {
1327 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1328 if (((c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC_STATIC)) ||
1329 (!(c->mfc_flags & MFC_STATIC) && !(flags & MRT_FLUSH_MFC)))
1330 continue;
1331 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1332 list_del_rcu(&c->list);
1333 cache = (struct mfc_cache *)c;
1334 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1335 mrt->id);
1336 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1337 mr_cache_put(c);
1338 }
1339 }
1340
1341 if (flags & MRT_FLUSH_MFC) {
1342 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1343 spin_lock_bh(&mfc_unres_lock);
1344 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1345 list_del(&c->list);
1346 cache = (struct mfc_cache *)c;
1347 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1348 ipmr_destroy_unres(mrt, cache);
1349 }
1350 spin_unlock_bh(&mfc_unres_lock);
1351 }
1352 }
1353}
1354
1355/* called from ip_ra_control(), before an RCU grace period,
1356 * we don't need to call synchronize_rcu() here
1357 */
1358static void mrtsock_destruct(struct sock *sk)
1359{
1360 struct net *net = sock_net(sk);
1361 struct mr_table *mrt;
1362
1363 rtnl_lock();
1364 ipmr_for_each_table(mrt, net) {
1365 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1366 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1367 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1368 NETCONFA_MC_FORWARDING,
1369 NETCONFA_IFINDEX_ALL,
1370 net->ipv4.devconf_all);
1371 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1372 mroute_clean_tables(mrt, MRT_FLUSH_VIFS | MRT_FLUSH_MFC);
1373 }
1374 }
1375 rtnl_unlock();
1376}
1377
1378/* Socket options and virtual interface manipulation. The whole
1379 * virtual interface system is a complete heap, but unfortunately
1380 * that's how BSD mrouted happens to think. Maybe one day with a proper
1381 * MOSPF/PIM router set up we can clean this up.
1382 */
1383
1384int ip_mroute_setsockopt(struct sock *sk, int optname, sockptr_t optval,
1385 unsigned int optlen)
1386{
1387 struct net *net = sock_net(sk);
1388 int val, ret = 0, parent = 0;
1389 struct mr_table *mrt;
1390 struct vifctl vif;
1391 struct mfcctl mfc;
1392 bool do_wrvifwhole;
1393 u32 uval;
1394
1395 /* There's one exception to the lock - MRT_DONE which needs to unlock */
1396 rtnl_lock();
1397 if (sk->sk_type != SOCK_RAW ||
1398 inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1399 ret = -EOPNOTSUPP;
1400 goto out_unlock;
1401 }
1402
1403 mrt = __ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1404 if (!mrt) {
1405 ret = -ENOENT;
1406 goto out_unlock;
1407 }
1408 if (optname != MRT_INIT) {
1409 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1410 !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1411 ret = -EACCES;
1412 goto out_unlock;
1413 }
1414 }
1415
1416 switch (optname) {
1417 case MRT_INIT:
1418 if (optlen != sizeof(int)) {
1419 ret = -EINVAL;
1420 break;
1421 }
1422 if (rtnl_dereference(mrt->mroute_sk)) {
1423 ret = -EADDRINUSE;
1424 break;
1425 }
1426
1427 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1428 if (ret == 0) {
1429 rcu_assign_pointer(mrt->mroute_sk, sk);
1430 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1431 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1432 NETCONFA_MC_FORWARDING,
1433 NETCONFA_IFINDEX_ALL,
1434 net->ipv4.devconf_all);
1435 }
1436 break;
1437 case MRT_DONE:
1438 if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1439 ret = -EACCES;
1440 } else {
1441 /* We need to unlock here because mrtsock_destruct takes
1442 * care of rtnl itself and we can't change that due to
1443 * the IP_ROUTER_ALERT setsockopt which runs without it.
1444 */
1445 rtnl_unlock();
1446 ret = ip_ra_control(sk, 0, NULL);
1447 goto out;
1448 }
1449 break;
1450 case MRT_ADD_VIF:
1451 case MRT_DEL_VIF:
1452 if (optlen != sizeof(vif)) {
1453 ret = -EINVAL;
1454 break;
1455 }
1456 if (copy_from_sockptr(&vif, optval, sizeof(vif))) {
1457 ret = -EFAULT;
1458 break;
1459 }
1460 if (vif.vifc_vifi >= MAXVIFS) {
1461 ret = -ENFILE;
1462 break;
1463 }
1464 if (optname == MRT_ADD_VIF) {
1465 ret = vif_add(net, mrt, &vif,
1466 sk == rtnl_dereference(mrt->mroute_sk));
1467 } else {
1468 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1469 }
1470 break;
1471 /* Manipulate the forwarding caches. These live
1472 * in a sort of kernel/user symbiosis.
1473 */
1474 case MRT_ADD_MFC:
1475 case MRT_DEL_MFC:
1476 parent = -1;
1477 fallthrough;
1478 case MRT_ADD_MFC_PROXY:
1479 case MRT_DEL_MFC_PROXY:
1480 if (optlen != sizeof(mfc)) {
1481 ret = -EINVAL;
1482 break;
1483 }
1484 if (copy_from_sockptr(&mfc, optval, sizeof(mfc))) {
1485 ret = -EFAULT;
1486 break;
1487 }
1488 if (parent == 0)
1489 parent = mfc.mfcc_parent;
1490 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1491 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1492 else
1493 ret = ipmr_mfc_add(net, mrt, &mfc,
1494 sk == rtnl_dereference(mrt->mroute_sk),
1495 parent);
1496 break;
1497 case MRT_FLUSH:
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 mroute_clean_tables(mrt, val);
1507 break;
1508 /* Control PIM assert. */
1509 case MRT_ASSERT:
1510 if (optlen != sizeof(val)) {
1511 ret = -EINVAL;
1512 break;
1513 }
1514 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1515 ret = -EFAULT;
1516 break;
1517 }
1518 mrt->mroute_do_assert = val;
1519 break;
1520 case MRT_PIM:
1521 if (!ipmr_pimsm_enabled()) {
1522 ret = -ENOPROTOOPT;
1523 break;
1524 }
1525 if (optlen != sizeof(val)) {
1526 ret = -EINVAL;
1527 break;
1528 }
1529 if (copy_from_sockptr(&val, optval, sizeof(val))) {
1530 ret = -EFAULT;
1531 break;
1532 }
1533
1534 do_wrvifwhole = (val == IGMPMSG_WRVIFWHOLE);
1535 val = !!val;
1536 if (val != mrt->mroute_do_pim) {
1537 mrt->mroute_do_pim = val;
1538 mrt->mroute_do_assert = val;
1539 mrt->mroute_do_wrvifwhole = do_wrvifwhole;
1540 }
1541 break;
1542 case MRT_TABLE:
1543 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1544 ret = -ENOPROTOOPT;
1545 break;
1546 }
1547 if (optlen != sizeof(uval)) {
1548 ret = -EINVAL;
1549 break;
1550 }
1551 if (copy_from_sockptr(&uval, optval, sizeof(uval))) {
1552 ret = -EFAULT;
1553 break;
1554 }
1555
1556 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1557 ret = -EBUSY;
1558 } else {
1559 mrt = ipmr_new_table(net, uval);
1560 if (IS_ERR(mrt))
1561 ret = PTR_ERR(mrt);
1562 else
1563 raw_sk(sk)->ipmr_table = uval;
1564 }
1565 break;
1566 /* Spurious command, or MRT_VERSION which you cannot set. */
1567 default:
1568 ret = -ENOPROTOOPT;
1569 }
1570out_unlock:
1571 rtnl_unlock();
1572out:
1573 return ret;
1574}
1575
1576/* Execute if this ioctl is a special mroute ioctl */
1577int ipmr_sk_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1578{
1579 switch (cmd) {
1580 /* These userspace buffers will be consumed by ipmr_ioctl() */
1581 case SIOCGETVIFCNT: {
1582 struct sioc_vif_req buffer;
1583
1584 return sock_ioctl_inout(sk, cmd, arg, &buffer,
1585 sizeof(buffer));
1586 }
1587 case SIOCGETSGCNT: {
1588 struct sioc_sg_req buffer;
1589
1590 return sock_ioctl_inout(sk, cmd, arg, &buffer,
1591 sizeof(buffer));
1592 }
1593 }
1594 /* return code > 0 means that the ioctl was not executed */
1595 return 1;
1596}
1597
1598/* Getsock opt support for the multicast routing system. */
1599int ip_mroute_getsockopt(struct sock *sk, int optname, sockptr_t optval,
1600 sockptr_t optlen)
1601{
1602 int olr;
1603 int val;
1604 struct net *net = sock_net(sk);
1605 struct mr_table *mrt;
1606
1607 if (sk->sk_type != SOCK_RAW ||
1608 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1609 return -EOPNOTSUPP;
1610
1611 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1612 if (!mrt)
1613 return -ENOENT;
1614
1615 switch (optname) {
1616 case MRT_VERSION:
1617 val = 0x0305;
1618 break;
1619 case MRT_PIM:
1620 if (!ipmr_pimsm_enabled())
1621 return -ENOPROTOOPT;
1622 val = mrt->mroute_do_pim;
1623 break;
1624 case MRT_ASSERT:
1625 val = mrt->mroute_do_assert;
1626 break;
1627 default:
1628 return -ENOPROTOOPT;
1629 }
1630
1631 if (copy_from_sockptr(&olr, optlen, sizeof(int)))
1632 return -EFAULT;
1633 if (olr < 0)
1634 return -EINVAL;
1635
1636 olr = min_t(unsigned int, olr, sizeof(int));
1637
1638 if (copy_to_sockptr(optlen, &olr, sizeof(int)))
1639 return -EFAULT;
1640 if (copy_to_sockptr(optval, &val, olr))
1641 return -EFAULT;
1642 return 0;
1643}
1644
1645/* The IP multicast ioctl support routines. */
1646int ipmr_ioctl(struct sock *sk, int cmd, void *arg)
1647{
1648 struct vif_device *vif;
1649 struct mfc_cache *c;
1650 struct net *net = sock_net(sk);
1651 struct sioc_vif_req *vr;
1652 struct sioc_sg_req *sr;
1653 struct mr_table *mrt;
1654
1655 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1656 if (!mrt)
1657 return -ENOENT;
1658
1659 switch (cmd) {
1660 case SIOCGETVIFCNT:
1661 vr = (struct sioc_vif_req *)arg;
1662 if (vr->vifi >= mrt->maxvif)
1663 return -EINVAL;
1664 vr->vifi = array_index_nospec(vr->vifi, mrt->maxvif);
1665 rcu_read_lock();
1666 vif = &mrt->vif_table[vr->vifi];
1667 if (VIF_EXISTS(mrt, vr->vifi)) {
1668 vr->icount = READ_ONCE(vif->pkt_in);
1669 vr->ocount = READ_ONCE(vif->pkt_out);
1670 vr->ibytes = READ_ONCE(vif->bytes_in);
1671 vr->obytes = READ_ONCE(vif->bytes_out);
1672 rcu_read_unlock();
1673
1674 return 0;
1675 }
1676 rcu_read_unlock();
1677 return -EADDRNOTAVAIL;
1678 case SIOCGETSGCNT:
1679 sr = (struct sioc_sg_req *)arg;
1680
1681 rcu_read_lock();
1682 c = ipmr_cache_find(mrt, sr->src.s_addr, sr->grp.s_addr);
1683 if (c) {
1684 sr->pktcnt = atomic_long_read(&c->_c.mfc_un.res.pkt);
1685 sr->bytecnt = atomic_long_read(&c->_c.mfc_un.res.bytes);
1686 sr->wrong_if = atomic_long_read(&c->_c.mfc_un.res.wrong_if);
1687 rcu_read_unlock();
1688 return 0;
1689 }
1690 rcu_read_unlock();
1691 return -EADDRNOTAVAIL;
1692 default:
1693 return -ENOIOCTLCMD;
1694 }
1695}
1696
1697#ifdef CONFIG_COMPAT
1698struct compat_sioc_sg_req {
1699 struct in_addr src;
1700 struct in_addr grp;
1701 compat_ulong_t pktcnt;
1702 compat_ulong_t bytecnt;
1703 compat_ulong_t wrong_if;
1704};
1705
1706struct compat_sioc_vif_req {
1707 vifi_t vifi; /* Which iface */
1708 compat_ulong_t icount;
1709 compat_ulong_t ocount;
1710 compat_ulong_t ibytes;
1711 compat_ulong_t obytes;
1712};
1713
1714int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1715{
1716 struct compat_sioc_sg_req sr;
1717 struct compat_sioc_vif_req vr;
1718 struct vif_device *vif;
1719 struct mfc_cache *c;
1720 struct net *net = sock_net(sk);
1721 struct mr_table *mrt;
1722
1723 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1724 if (!mrt)
1725 return -ENOENT;
1726
1727 switch (cmd) {
1728 case SIOCGETVIFCNT:
1729 if (copy_from_user(&vr, arg, sizeof(vr)))
1730 return -EFAULT;
1731 if (vr.vifi >= mrt->maxvif)
1732 return -EINVAL;
1733 vr.vifi = array_index_nospec(vr.vifi, mrt->maxvif);
1734 rcu_read_lock();
1735 vif = &mrt->vif_table[vr.vifi];
1736 if (VIF_EXISTS(mrt, vr.vifi)) {
1737 vr.icount = READ_ONCE(vif->pkt_in);
1738 vr.ocount = READ_ONCE(vif->pkt_out);
1739 vr.ibytes = READ_ONCE(vif->bytes_in);
1740 vr.obytes = READ_ONCE(vif->bytes_out);
1741 rcu_read_unlock();
1742
1743 if (copy_to_user(arg, &vr, sizeof(vr)))
1744 return -EFAULT;
1745 return 0;
1746 }
1747 rcu_read_unlock();
1748 return -EADDRNOTAVAIL;
1749 case SIOCGETSGCNT:
1750 if (copy_from_user(&sr, arg, sizeof(sr)))
1751 return -EFAULT;
1752
1753 rcu_read_lock();
1754 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1755 if (c) {
1756 sr.pktcnt = atomic_long_read(&c->_c.mfc_un.res.pkt);
1757 sr.bytecnt = atomic_long_read(&c->_c.mfc_un.res.bytes);
1758 sr.wrong_if = atomic_long_read(&c->_c.mfc_un.res.wrong_if);
1759 rcu_read_unlock();
1760
1761 if (copy_to_user(arg, &sr, sizeof(sr)))
1762 return -EFAULT;
1763 return 0;
1764 }
1765 rcu_read_unlock();
1766 return -EADDRNOTAVAIL;
1767 default:
1768 return -ENOIOCTLCMD;
1769 }
1770}
1771#endif
1772
1773static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1774{
1775 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1776 struct net *net = dev_net(dev);
1777 struct mr_table *mrt;
1778 struct vif_device *v;
1779 int ct;
1780
1781 if (event != NETDEV_UNREGISTER)
1782 return NOTIFY_DONE;
1783
1784 ipmr_for_each_table(mrt, net) {
1785 v = &mrt->vif_table[0];
1786 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1787 if (rcu_access_pointer(v->dev) == dev)
1788 vif_delete(mrt, ct, 1, NULL);
1789 }
1790 }
1791 return NOTIFY_DONE;
1792}
1793
1794static struct notifier_block ip_mr_notifier = {
1795 .notifier_call = ipmr_device_event,
1796};
1797
1798/* Encapsulate a packet by attaching a valid IPIP header to it.
1799 * This avoids tunnel drivers and other mess and gives us the speed so
1800 * important for multicast video.
1801 */
1802static void ip_encap(struct net *net, struct sk_buff *skb,
1803 __be32 saddr, __be32 daddr)
1804{
1805 struct iphdr *iph;
1806 const struct iphdr *old_iph = ip_hdr(skb);
1807
1808 skb_push(skb, sizeof(struct iphdr));
1809 skb->transport_header = skb->network_header;
1810 skb_reset_network_header(skb);
1811 iph = ip_hdr(skb);
1812
1813 iph->version = 4;
1814 iph->tos = old_iph->tos;
1815 iph->ttl = old_iph->ttl;
1816 iph->frag_off = 0;
1817 iph->daddr = daddr;
1818 iph->saddr = saddr;
1819 iph->protocol = IPPROTO_IPIP;
1820 iph->ihl = 5;
1821 iph->tot_len = htons(skb->len);
1822 ip_select_ident(net, skb, NULL);
1823 ip_send_check(iph);
1824
1825 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1826 nf_reset_ct(skb);
1827}
1828
1829static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1830 struct sk_buff *skb)
1831{
1832 struct ip_options *opt = &(IPCB(skb)->opt);
1833
1834 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1835
1836 if (unlikely(opt->optlen))
1837 ip_forward_options(skb);
1838
1839 return dst_output(net, sk, skb);
1840}
1841
1842#ifdef CONFIG_NET_SWITCHDEV
1843static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1844 int in_vifi, int out_vifi)
1845{
1846 struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1847 struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1848
1849 if (!skb->offload_l3_fwd_mark)
1850 return false;
1851 if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1852 return false;
1853 return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1854 &in_vif->dev_parent_id);
1855}
1856#else
1857static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1858 int in_vifi, int out_vifi)
1859{
1860 return false;
1861}
1862#endif
1863
1864/* Processing handlers for ipmr_forward, under rcu_read_lock() */
1865
1866static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1867 int in_vifi, struct sk_buff *skb, int vifi)
1868{
1869 const struct iphdr *iph = ip_hdr(skb);
1870 struct vif_device *vif = &mrt->vif_table[vifi];
1871 struct net_device *vif_dev;
1872 struct net_device *dev;
1873 struct rtable *rt;
1874 struct flowi4 fl4;
1875 int encap = 0;
1876
1877 vif_dev = vif_dev_read(vif);
1878 if (!vif_dev)
1879 goto out_free;
1880
1881 if (vif->flags & VIFF_REGISTER) {
1882 WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1883 WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1884 DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
1885 DEV_STATS_INC(vif_dev, tx_packets);
1886 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1887 goto out_free;
1888 }
1889
1890 if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1891 goto out_free;
1892
1893 if (vif->flags & VIFF_TUNNEL) {
1894 rt = ip_route_output_ports(net, &fl4, NULL,
1895 vif->remote, vif->local,
1896 0, 0,
1897 IPPROTO_IPIP,
1898 iph->tos & INET_DSCP_MASK, vif->link);
1899 if (IS_ERR(rt))
1900 goto out_free;
1901 encap = sizeof(struct iphdr);
1902 } else {
1903 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1904 0, 0,
1905 IPPROTO_IPIP,
1906 iph->tos & INET_DSCP_MASK, vif->link);
1907 if (IS_ERR(rt))
1908 goto out_free;
1909 }
1910
1911 dev = rt->dst.dev;
1912
1913 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1914 /* Do not fragment multicasts. Alas, IPv4 does not
1915 * allow to send ICMP, so that packets will disappear
1916 * to blackhole.
1917 */
1918 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1919 ip_rt_put(rt);
1920 goto out_free;
1921 }
1922
1923 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1924
1925 if (skb_cow(skb, encap)) {
1926 ip_rt_put(rt);
1927 goto out_free;
1928 }
1929
1930 WRITE_ONCE(vif->pkt_out, vif->pkt_out + 1);
1931 WRITE_ONCE(vif->bytes_out, vif->bytes_out + skb->len);
1932
1933 skb_dst_drop(skb);
1934 skb_dst_set(skb, &rt->dst);
1935 ip_decrease_ttl(ip_hdr(skb));
1936
1937 /* FIXME: forward and output firewalls used to be called here.
1938 * What do we do with netfilter? -- RR
1939 */
1940 if (vif->flags & VIFF_TUNNEL) {
1941 ip_encap(net, skb, vif->local, vif->remote);
1942 /* FIXME: extra output firewall step used to be here. --RR */
1943 DEV_STATS_INC(vif_dev, tx_packets);
1944 DEV_STATS_ADD(vif_dev, tx_bytes, skb->len);
1945 }
1946
1947 IPCB(skb)->flags |= IPSKB_FORWARDED;
1948
1949 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1950 * not only before forwarding, but after forwarding on all output
1951 * interfaces. It is clear, if mrouter runs a multicasting
1952 * program, it should receive packets not depending to what interface
1953 * program is joined.
1954 * If we will not make it, the program will have to join on all
1955 * interfaces. On the other hand, multihoming host (or router, but
1956 * not mrouter) cannot join to more than one interface - it will
1957 * result in receiving multiple packets.
1958 */
1959 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1960 net, NULL, skb, skb->dev, dev,
1961 ipmr_forward_finish);
1962 return;
1963
1964out_free:
1965 kfree_skb(skb);
1966}
1967
1968/* Called with mrt_lock or rcu_read_lock() */
1969static int ipmr_find_vif(const struct mr_table *mrt, struct net_device *dev)
1970{
1971 int ct;
1972 /* Pairs with WRITE_ONCE() in vif_delete()/vif_add() */
1973 for (ct = READ_ONCE(mrt->maxvif) - 1; ct >= 0; ct--) {
1974 if (rcu_access_pointer(mrt->vif_table[ct].dev) == dev)
1975 break;
1976 }
1977 return ct;
1978}
1979
1980/* "local" means that we should preserve one skb (for local delivery) */
1981/* Called uner rcu_read_lock() */
1982static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1983 struct net_device *dev, struct sk_buff *skb,
1984 struct mfc_cache *c, int local)
1985{
1986 int true_vifi = ipmr_find_vif(mrt, dev);
1987 int psend = -1;
1988 int vif, ct;
1989
1990 vif = c->_c.mfc_parent;
1991 atomic_long_inc(&c->_c.mfc_un.res.pkt);
1992 atomic_long_add(skb->len, &c->_c.mfc_un.res.bytes);
1993 WRITE_ONCE(c->_c.mfc_un.res.lastuse, jiffies);
1994
1995 if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1996 struct mfc_cache *cache_proxy;
1997
1998 /* For an (*,G) entry, we only check that the incoming
1999 * interface is part of the static tree.
2000 */
2001 cache_proxy = mr_mfc_find_any_parent(mrt, vif);
2002 if (cache_proxy &&
2003 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
2004 goto forward;
2005 }
2006
2007 /* Wrong interface: drop packet and (maybe) send PIM assert. */
2008 if (rcu_access_pointer(mrt->vif_table[vif].dev) != dev) {
2009 if (rt_is_output_route(skb_rtable(skb))) {
2010 /* It is our own packet, looped back.
2011 * Very complicated situation...
2012 *
2013 * The best workaround until routing daemons will be
2014 * fixed is not to redistribute packet, if it was
2015 * send through wrong interface. It means, that
2016 * multicast applications WILL NOT work for
2017 * (S,G), which have default multicast route pointing
2018 * to wrong oif. In any case, it is not a good
2019 * idea to use multicasting applications on router.
2020 */
2021 goto dont_forward;
2022 }
2023
2024 atomic_long_inc(&c->_c.mfc_un.res.wrong_if);
2025
2026 if (true_vifi >= 0 && mrt->mroute_do_assert &&
2027 /* pimsm uses asserts, when switching from RPT to SPT,
2028 * so that we cannot check that packet arrived on an oif.
2029 * It is bad, but otherwise we would need to move pretty
2030 * large chunk of pimd to kernel. Ough... --ANK
2031 */
2032 (mrt->mroute_do_pim ||
2033 c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
2034 time_after(jiffies,
2035 c->_c.mfc_un.res.last_assert +
2036 MFC_ASSERT_THRESH)) {
2037 c->_c.mfc_un.res.last_assert = jiffies;
2038 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
2039 if (mrt->mroute_do_wrvifwhole)
2040 ipmr_cache_report(mrt, skb, true_vifi,
2041 IGMPMSG_WRVIFWHOLE);
2042 }
2043 goto dont_forward;
2044 }
2045
2046forward:
2047 WRITE_ONCE(mrt->vif_table[vif].pkt_in,
2048 mrt->vif_table[vif].pkt_in + 1);
2049 WRITE_ONCE(mrt->vif_table[vif].bytes_in,
2050 mrt->vif_table[vif].bytes_in + skb->len);
2051
2052 /* Forward the frame */
2053 if (c->mfc_origin == htonl(INADDR_ANY) &&
2054 c->mfc_mcastgrp == htonl(INADDR_ANY)) {
2055 if (true_vifi >= 0 &&
2056 true_vifi != c->_c.mfc_parent &&
2057 ip_hdr(skb)->ttl >
2058 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
2059 /* It's an (*,*) entry and the packet is not coming from
2060 * the upstream: forward the packet to the upstream
2061 * only.
2062 */
2063 psend = c->_c.mfc_parent;
2064 goto last_forward;
2065 }
2066 goto dont_forward;
2067 }
2068 for (ct = c->_c.mfc_un.res.maxvif - 1;
2069 ct >= c->_c.mfc_un.res.minvif; ct--) {
2070 /* For (*,G) entry, don't forward to the incoming interface */
2071 if ((c->mfc_origin != htonl(INADDR_ANY) ||
2072 ct != true_vifi) &&
2073 ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2074 if (psend != -1) {
2075 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2076
2077 if (skb2)
2078 ipmr_queue_xmit(net, mrt, true_vifi,
2079 skb2, psend);
2080 }
2081 psend = ct;
2082 }
2083 }
2084last_forward:
2085 if (psend != -1) {
2086 if (local) {
2087 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2088
2089 if (skb2)
2090 ipmr_queue_xmit(net, mrt, true_vifi, skb2,
2091 psend);
2092 } else {
2093 ipmr_queue_xmit(net, mrt, true_vifi, skb, psend);
2094 return;
2095 }
2096 }
2097
2098dont_forward:
2099 if (!local)
2100 kfree_skb(skb);
2101}
2102
2103static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2104{
2105 struct rtable *rt = skb_rtable(skb);
2106 struct iphdr *iph = ip_hdr(skb);
2107 struct flowi4 fl4 = {
2108 .daddr = iph->daddr,
2109 .saddr = iph->saddr,
2110 .flowi4_tos = inet_dscp_to_dsfield(ip4h_dscp(iph)),
2111 .flowi4_oif = (rt_is_output_route(rt) ?
2112 skb->dev->ifindex : 0),
2113 .flowi4_iif = (rt_is_output_route(rt) ?
2114 LOOPBACK_IFINDEX :
2115 skb->dev->ifindex),
2116 .flowi4_mark = skb->mark,
2117 };
2118 struct mr_table *mrt;
2119 int err;
2120
2121 err = ipmr_fib_lookup(net, &fl4, &mrt);
2122 if (err)
2123 return ERR_PTR(err);
2124 return mrt;
2125}
2126
2127/* Multicast packets for forwarding arrive here
2128 * Called with rcu_read_lock();
2129 */
2130int ip_mr_input(struct sk_buff *skb)
2131{
2132 struct mfc_cache *cache;
2133 struct net *net = dev_net(skb->dev);
2134 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2135 struct mr_table *mrt;
2136 struct net_device *dev;
2137
2138 /* skb->dev passed in is the loX master dev for vrfs.
2139 * As there are no vifs associated with loopback devices,
2140 * get the proper interface that does have a vif associated with it.
2141 */
2142 dev = skb->dev;
2143 if (netif_is_l3_master(skb->dev)) {
2144 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2145 if (!dev) {
2146 kfree_skb(skb);
2147 return -ENODEV;
2148 }
2149 }
2150
2151 /* Packet is looped back after forward, it should not be
2152 * forwarded second time, but still can be delivered locally.
2153 */
2154 if (IPCB(skb)->flags & IPSKB_FORWARDED)
2155 goto dont_forward;
2156
2157 mrt = ipmr_rt_fib_lookup(net, skb);
2158 if (IS_ERR(mrt)) {
2159 kfree_skb(skb);
2160 return PTR_ERR(mrt);
2161 }
2162 if (!local) {
2163 if (IPCB(skb)->opt.router_alert) {
2164 if (ip_call_ra_chain(skb))
2165 return 0;
2166 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2167 /* IGMPv1 (and broken IGMPv2 implementations sort of
2168 * Cisco IOS <= 11.2(8)) do not put router alert
2169 * option to IGMP packets destined to routable
2170 * groups. It is very bad, because it means
2171 * that we can forward NO IGMP messages.
2172 */
2173 struct sock *mroute_sk;
2174
2175 mroute_sk = rcu_dereference(mrt->mroute_sk);
2176 if (mroute_sk) {
2177 nf_reset_ct(skb);
2178 raw_rcv(mroute_sk, skb);
2179 return 0;
2180 }
2181 }
2182 }
2183
2184 /* already under rcu_read_lock() */
2185 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2186 if (!cache) {
2187 int vif = ipmr_find_vif(mrt, dev);
2188
2189 if (vif >= 0)
2190 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2191 vif);
2192 }
2193
2194 /* No usable cache entry */
2195 if (!cache) {
2196 int vif;
2197
2198 if (local) {
2199 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2200 ip_local_deliver(skb);
2201 if (!skb2)
2202 return -ENOBUFS;
2203 skb = skb2;
2204 }
2205
2206 vif = ipmr_find_vif(mrt, dev);
2207 if (vif >= 0)
2208 return ipmr_cache_unresolved(mrt, vif, skb, dev);
2209 kfree_skb(skb);
2210 return -ENODEV;
2211 }
2212
2213 ip_mr_forward(net, mrt, dev, skb, cache, local);
2214
2215 if (local)
2216 return ip_local_deliver(skb);
2217
2218 return 0;
2219
2220dont_forward:
2221 if (local)
2222 return ip_local_deliver(skb);
2223 kfree_skb(skb);
2224 return 0;
2225}
2226
2227#ifdef CONFIG_IP_PIMSM_V1
2228/* Handle IGMP messages of PIMv1 */
2229int pim_rcv_v1(struct sk_buff *skb)
2230{
2231 struct igmphdr *pim;
2232 struct net *net = dev_net(skb->dev);
2233 struct mr_table *mrt;
2234
2235 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2236 goto drop;
2237
2238 pim = igmp_hdr(skb);
2239
2240 mrt = ipmr_rt_fib_lookup(net, skb);
2241 if (IS_ERR(mrt))
2242 goto drop;
2243 if (!mrt->mroute_do_pim ||
2244 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2245 goto drop;
2246
2247 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2248drop:
2249 kfree_skb(skb);
2250 }
2251 return 0;
2252}
2253#endif
2254
2255#ifdef CONFIG_IP_PIMSM_V2
2256static int pim_rcv(struct sk_buff *skb)
2257{
2258 struct pimreghdr *pim;
2259 struct net *net = dev_net(skb->dev);
2260 struct mr_table *mrt;
2261
2262 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2263 goto drop;
2264
2265 pim = (struct pimreghdr *)skb_transport_header(skb);
2266 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2267 (pim->flags & PIM_NULL_REGISTER) ||
2268 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2269 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2270 goto drop;
2271
2272 mrt = ipmr_rt_fib_lookup(net, skb);
2273 if (IS_ERR(mrt))
2274 goto drop;
2275 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2276drop:
2277 kfree_skb(skb);
2278 }
2279 return 0;
2280}
2281#endif
2282
2283int ipmr_get_route(struct net *net, struct sk_buff *skb,
2284 __be32 saddr, __be32 daddr,
2285 struct rtmsg *rtm, u32 portid)
2286{
2287 struct mfc_cache *cache;
2288 struct mr_table *mrt;
2289 int err;
2290
2291 rcu_read_lock();
2292 mrt = __ipmr_get_table(net, RT_TABLE_DEFAULT);
2293 if (!mrt) {
2294 rcu_read_unlock();
2295 return -ENOENT;
2296 }
2297
2298 cache = ipmr_cache_find(mrt, saddr, daddr);
2299 if (!cache && skb->dev) {
2300 int vif = ipmr_find_vif(mrt, skb->dev);
2301
2302 if (vif >= 0)
2303 cache = ipmr_cache_find_any(mrt, daddr, vif);
2304 }
2305 if (!cache) {
2306 struct sk_buff *skb2;
2307 struct iphdr *iph;
2308 struct net_device *dev;
2309 int vif = -1;
2310
2311 dev = skb->dev;
2312 if (dev)
2313 vif = ipmr_find_vif(mrt, dev);
2314 if (vif < 0) {
2315 rcu_read_unlock();
2316 return -ENODEV;
2317 }
2318
2319 skb2 = skb_realloc_headroom(skb, sizeof(struct iphdr));
2320 if (!skb2) {
2321 rcu_read_unlock();
2322 return -ENOMEM;
2323 }
2324
2325 NETLINK_CB(skb2).portid = portid;
2326 skb_push(skb2, sizeof(struct iphdr));
2327 skb_reset_network_header(skb2);
2328 iph = ip_hdr(skb2);
2329 iph->ihl = sizeof(struct iphdr) >> 2;
2330 iph->saddr = saddr;
2331 iph->daddr = daddr;
2332 iph->version = 0;
2333 err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2334 rcu_read_unlock();
2335 return err;
2336 }
2337
2338 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2339 rcu_read_unlock();
2340 return err;
2341}
2342
2343static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2344 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2345 int flags)
2346{
2347 struct nlmsghdr *nlh;
2348 struct rtmsg *rtm;
2349 int err;
2350
2351 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2352 if (!nlh)
2353 return -EMSGSIZE;
2354
2355 rtm = nlmsg_data(nlh);
2356 rtm->rtm_family = RTNL_FAMILY_IPMR;
2357 rtm->rtm_dst_len = 32;
2358 rtm->rtm_src_len = 32;
2359 rtm->rtm_tos = 0;
2360 rtm->rtm_table = mrt->id;
2361 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2362 goto nla_put_failure;
2363 rtm->rtm_type = RTN_MULTICAST;
2364 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2365 if (c->_c.mfc_flags & MFC_STATIC)
2366 rtm->rtm_protocol = RTPROT_STATIC;
2367 else
2368 rtm->rtm_protocol = RTPROT_MROUTED;
2369 rtm->rtm_flags = 0;
2370
2371 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2372 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2373 goto nla_put_failure;
2374 err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2375 /* do not break the dump if cache is unresolved */
2376 if (err < 0 && err != -ENOENT)
2377 goto nla_put_failure;
2378
2379 nlmsg_end(skb, nlh);
2380 return 0;
2381
2382nla_put_failure:
2383 nlmsg_cancel(skb, nlh);
2384 return -EMSGSIZE;
2385}
2386
2387static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2388 u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2389 int flags)
2390{
2391 return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2392 cmd, flags);
2393}
2394
2395static size_t mroute_msgsize(bool unresolved, int maxvif)
2396{
2397 size_t len =
2398 NLMSG_ALIGN(sizeof(struct rtmsg))
2399 + nla_total_size(4) /* RTA_TABLE */
2400 + nla_total_size(4) /* RTA_SRC */
2401 + nla_total_size(4) /* RTA_DST */
2402 ;
2403
2404 if (!unresolved)
2405 len = len
2406 + nla_total_size(4) /* RTA_IIF */
2407 + nla_total_size(0) /* RTA_MULTIPATH */
2408 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2409 /* RTA_MFC_STATS */
2410 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2411 ;
2412
2413 return len;
2414}
2415
2416static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2417 int cmd)
2418{
2419 struct net *net = read_pnet(&mrt->net);
2420 struct sk_buff *skb;
2421 int err = -ENOBUFS;
2422
2423 skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2424 mrt->maxvif),
2425 GFP_ATOMIC);
2426 if (!skb)
2427 goto errout;
2428
2429 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2430 if (err < 0)
2431 goto errout;
2432
2433 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2434 return;
2435
2436errout:
2437 kfree_skb(skb);
2438 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2439}
2440
2441static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2442{
2443 size_t len =
2444 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2445 + nla_total_size(1) /* IPMRA_CREPORT_MSGTYPE */
2446 + nla_total_size(4) /* IPMRA_CREPORT_VIF_ID */
2447 + nla_total_size(4) /* IPMRA_CREPORT_SRC_ADDR */
2448 + nla_total_size(4) /* IPMRA_CREPORT_DST_ADDR */
2449 + nla_total_size(4) /* IPMRA_CREPORT_TABLE */
2450 /* IPMRA_CREPORT_PKT */
2451 + nla_total_size(payloadlen)
2452 ;
2453
2454 return len;
2455}
2456
2457static void igmpmsg_netlink_event(const struct mr_table *mrt, struct sk_buff *pkt)
2458{
2459 struct net *net = read_pnet(&mrt->net);
2460 struct nlmsghdr *nlh;
2461 struct rtgenmsg *rtgenm;
2462 struct igmpmsg *msg;
2463 struct sk_buff *skb;
2464 struct nlattr *nla;
2465 int payloadlen;
2466
2467 payloadlen = pkt->len - sizeof(struct igmpmsg);
2468 msg = (struct igmpmsg *)skb_network_header(pkt);
2469
2470 skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2471 if (!skb)
2472 goto errout;
2473
2474 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2475 sizeof(struct rtgenmsg), 0);
2476 if (!nlh)
2477 goto errout;
2478 rtgenm = nlmsg_data(nlh);
2479 rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2480 if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2481 nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif | (msg->im_vif_hi << 8)) ||
2482 nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2483 msg->im_src.s_addr) ||
2484 nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2485 msg->im_dst.s_addr) ||
2486 nla_put_u32(skb, IPMRA_CREPORT_TABLE, mrt->id))
2487 goto nla_put_failure;
2488
2489 nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2490 if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2491 nla_data(nla), payloadlen))
2492 goto nla_put_failure;
2493
2494 nlmsg_end(skb, nlh);
2495
2496 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2497 return;
2498
2499nla_put_failure:
2500 nlmsg_cancel(skb, nlh);
2501errout:
2502 kfree_skb(skb);
2503 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2504}
2505
2506static int ipmr_rtm_valid_getroute_req(struct sk_buff *skb,
2507 const struct nlmsghdr *nlh,
2508 struct nlattr **tb,
2509 struct netlink_ext_ack *extack)
2510{
2511 struct rtmsg *rtm;
2512 int i, err;
2513
2514 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2515 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for multicast route get request");
2516 return -EINVAL;
2517 }
2518
2519 if (!netlink_strict_get_check(skb))
2520 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2521 rtm_ipv4_policy, extack);
2522
2523 rtm = nlmsg_data(nlh);
2524 if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
2525 (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
2526 rtm->rtm_tos || rtm->rtm_table || rtm->rtm_protocol ||
2527 rtm->rtm_scope || rtm->rtm_type || rtm->rtm_flags) {
2528 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for multicast route get request");
2529 return -EINVAL;
2530 }
2531
2532 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2533 rtm_ipv4_policy, extack);
2534 if (err)
2535 return err;
2536
2537 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
2538 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
2539 NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
2540 return -EINVAL;
2541 }
2542
2543 for (i = 0; i <= RTA_MAX; i++) {
2544 if (!tb[i])
2545 continue;
2546
2547 switch (i) {
2548 case RTA_SRC:
2549 case RTA_DST:
2550 case RTA_TABLE:
2551 break;
2552 default:
2553 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in multicast route get request");
2554 return -EINVAL;
2555 }
2556 }
2557
2558 return 0;
2559}
2560
2561static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2562 struct netlink_ext_ack *extack)
2563{
2564 struct net *net = sock_net(in_skb->sk);
2565 struct nlattr *tb[RTA_MAX + 1];
2566 struct sk_buff *skb = NULL;
2567 struct mfc_cache *cache;
2568 struct mr_table *mrt;
2569 __be32 src, grp;
2570 u32 tableid;
2571 int err;
2572
2573 err = ipmr_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
2574 if (err < 0)
2575 goto errout;
2576
2577 src = nla_get_in_addr_default(tb[RTA_SRC], 0);
2578 grp = nla_get_in_addr_default(tb[RTA_DST], 0);
2579 tableid = nla_get_u32_default(tb[RTA_TABLE], 0);
2580
2581 mrt = __ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2582 if (!mrt) {
2583 err = -ENOENT;
2584 goto errout_free;
2585 }
2586
2587 /* entries are added/deleted only under RTNL */
2588 rcu_read_lock();
2589 cache = ipmr_cache_find(mrt, src, grp);
2590 rcu_read_unlock();
2591 if (!cache) {
2592 err = -ENOENT;
2593 goto errout_free;
2594 }
2595
2596 skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2597 if (!skb) {
2598 err = -ENOBUFS;
2599 goto errout_free;
2600 }
2601
2602 err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2603 nlh->nlmsg_seq, cache,
2604 RTM_NEWROUTE, 0);
2605 if (err < 0)
2606 goto errout_free;
2607
2608 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2609
2610errout:
2611 return err;
2612
2613errout_free:
2614 kfree_skb(skb);
2615 goto errout;
2616}
2617
2618static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2619{
2620 struct fib_dump_filter filter = {
2621 .rtnl_held = true,
2622 };
2623 int err;
2624
2625 if (cb->strict_check) {
2626 err = ip_valid_fib_dump_req(sock_net(skb->sk), cb->nlh,
2627 &filter, cb);
2628 if (err < 0)
2629 return err;
2630 }
2631
2632 if (filter.table_id) {
2633 struct mr_table *mrt;
2634
2635 mrt = __ipmr_get_table(sock_net(skb->sk), filter.table_id);
2636 if (!mrt) {
2637 if (rtnl_msg_family(cb->nlh) != RTNL_FAMILY_IPMR)
2638 return skb->len;
2639
2640 NL_SET_ERR_MSG(cb->extack, "ipv4: MR table does not exist");
2641 return -ENOENT;
2642 }
2643 err = mr_table_dump(mrt, skb, cb, _ipmr_fill_mroute,
2644 &mfc_unres_lock, &filter);
2645 return skb->len ? : err;
2646 }
2647
2648 return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2649 _ipmr_fill_mroute, &mfc_unres_lock, &filter);
2650}
2651
2652static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2653 [RTA_SRC] = { .type = NLA_U32 },
2654 [RTA_DST] = { .type = NLA_U32 },
2655 [RTA_IIF] = { .type = NLA_U32 },
2656 [RTA_TABLE] = { .type = NLA_U32 },
2657 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2658};
2659
2660static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2661{
2662 switch (rtm_protocol) {
2663 case RTPROT_STATIC:
2664 case RTPROT_MROUTED:
2665 return true;
2666 }
2667 return false;
2668}
2669
2670static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2671{
2672 struct rtnexthop *rtnh = nla_data(nla);
2673 int remaining = nla_len(nla), vifi = 0;
2674
2675 while (rtnh_ok(rtnh, remaining)) {
2676 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2677 if (++vifi == MAXVIFS)
2678 break;
2679 rtnh = rtnh_next(rtnh, &remaining);
2680 }
2681
2682 return remaining > 0 ? -EINVAL : vifi;
2683}
2684
2685/* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2686static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2687 struct mfcctl *mfcc, int *mrtsock,
2688 struct mr_table **mrtret,
2689 struct netlink_ext_ack *extack)
2690{
2691 struct net_device *dev = NULL;
2692 u32 tblid = RT_TABLE_DEFAULT;
2693 struct mr_table *mrt;
2694 struct nlattr *attr;
2695 struct rtmsg *rtm;
2696 int ret, rem;
2697
2698 ret = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
2699 rtm_ipmr_policy, extack);
2700 if (ret < 0)
2701 goto out;
2702 rtm = nlmsg_data(nlh);
2703
2704 ret = -EINVAL;
2705 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2706 rtm->rtm_type != RTN_MULTICAST ||
2707 rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2708 !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2709 goto out;
2710
2711 memset(mfcc, 0, sizeof(*mfcc));
2712 mfcc->mfcc_parent = -1;
2713 ret = 0;
2714 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2715 switch (nla_type(attr)) {
2716 case RTA_SRC:
2717 mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2718 break;
2719 case RTA_DST:
2720 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2721 break;
2722 case RTA_IIF:
2723 dev = __dev_get_by_index(net, nla_get_u32(attr));
2724 if (!dev) {
2725 ret = -ENODEV;
2726 goto out;
2727 }
2728 break;
2729 case RTA_MULTIPATH:
2730 if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2731 ret = -EINVAL;
2732 goto out;
2733 }
2734 break;
2735 case RTA_PREFSRC:
2736 ret = 1;
2737 break;
2738 case RTA_TABLE:
2739 tblid = nla_get_u32(attr);
2740 break;
2741 }
2742 }
2743 mrt = __ipmr_get_table(net, tblid);
2744 if (!mrt) {
2745 ret = -ENOENT;
2746 goto out;
2747 }
2748 *mrtret = mrt;
2749 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2750 if (dev)
2751 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2752
2753out:
2754 return ret;
2755}
2756
2757/* takes care of both newroute and delroute */
2758static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2759 struct netlink_ext_ack *extack)
2760{
2761 struct net *net = sock_net(skb->sk);
2762 int ret, mrtsock, parent;
2763 struct mr_table *tbl;
2764 struct mfcctl mfcc;
2765
2766 mrtsock = 0;
2767 tbl = NULL;
2768 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2769 if (ret < 0)
2770 return ret;
2771
2772 parent = ret ? mfcc.mfcc_parent : -1;
2773 if (nlh->nlmsg_type == RTM_NEWROUTE)
2774 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2775 else
2776 return ipmr_mfc_delete(tbl, &mfcc, parent);
2777}
2778
2779static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2780{
2781 u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2782
2783 if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2784 nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2785 nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2786 mrt->mroute_reg_vif_num) ||
2787 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2788 mrt->mroute_do_assert) ||
2789 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim) ||
2790 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_WRVIFWHOLE,
2791 mrt->mroute_do_wrvifwhole))
2792 return false;
2793
2794 return true;
2795}
2796
2797static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2798{
2799 struct net_device *vif_dev;
2800 struct nlattr *vif_nest;
2801 struct vif_device *vif;
2802
2803 vif = &mrt->vif_table[vifid];
2804 vif_dev = rtnl_dereference(vif->dev);
2805 /* if the VIF doesn't exist just continue */
2806 if (!vif_dev)
2807 return true;
2808
2809 vif_nest = nla_nest_start_noflag(skb, IPMRA_VIF);
2810 if (!vif_nest)
2811 return false;
2812
2813 if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif_dev->ifindex) ||
2814 nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2815 nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2816 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2817 IPMRA_VIFA_PAD) ||
2818 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2819 IPMRA_VIFA_PAD) ||
2820 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2821 IPMRA_VIFA_PAD) ||
2822 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2823 IPMRA_VIFA_PAD) ||
2824 nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2825 nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2826 nla_nest_cancel(skb, vif_nest);
2827 return false;
2828 }
2829 nla_nest_end(skb, vif_nest);
2830
2831 return true;
2832}
2833
2834static int ipmr_valid_dumplink(const struct nlmsghdr *nlh,
2835 struct netlink_ext_ack *extack)
2836{
2837 struct ifinfomsg *ifm;
2838
2839 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2840 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for ipmr link dump");
2841 return -EINVAL;
2842 }
2843
2844 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
2845 NL_SET_ERR_MSG(extack, "Invalid data after header in ipmr link dump");
2846 return -EINVAL;
2847 }
2848
2849 ifm = nlmsg_data(nlh);
2850 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2851 ifm->ifi_change || ifm->ifi_index) {
2852 NL_SET_ERR_MSG(extack, "Invalid values in header for ipmr link dump request");
2853 return -EINVAL;
2854 }
2855
2856 return 0;
2857}
2858
2859static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2860{
2861 struct net *net = sock_net(skb->sk);
2862 struct nlmsghdr *nlh = NULL;
2863 unsigned int t = 0, s_t;
2864 unsigned int e = 0, s_e;
2865 struct mr_table *mrt;
2866
2867 if (cb->strict_check) {
2868 int err = ipmr_valid_dumplink(cb->nlh, cb->extack);
2869
2870 if (err < 0)
2871 return err;
2872 }
2873
2874 s_t = cb->args[0];
2875 s_e = cb->args[1];
2876
2877 ipmr_for_each_table(mrt, net) {
2878 struct nlattr *vifs, *af;
2879 struct ifinfomsg *hdr;
2880 u32 i;
2881
2882 if (t < s_t)
2883 goto skip_table;
2884 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2885 cb->nlh->nlmsg_seq, RTM_NEWLINK,
2886 sizeof(*hdr), NLM_F_MULTI);
2887 if (!nlh)
2888 break;
2889
2890 hdr = nlmsg_data(nlh);
2891 memset(hdr, 0, sizeof(*hdr));
2892 hdr->ifi_family = RTNL_FAMILY_IPMR;
2893
2894 af = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
2895 if (!af) {
2896 nlmsg_cancel(skb, nlh);
2897 goto out;
2898 }
2899
2900 if (!ipmr_fill_table(mrt, skb)) {
2901 nlmsg_cancel(skb, nlh);
2902 goto out;
2903 }
2904
2905 vifs = nla_nest_start_noflag(skb, IPMRA_TABLE_VIFS);
2906 if (!vifs) {
2907 nla_nest_end(skb, af);
2908 nlmsg_end(skb, nlh);
2909 goto out;
2910 }
2911 for (i = 0; i < mrt->maxvif; i++) {
2912 if (e < s_e)
2913 goto skip_entry;
2914 if (!ipmr_fill_vif(mrt, i, skb)) {
2915 nla_nest_end(skb, vifs);
2916 nla_nest_end(skb, af);
2917 nlmsg_end(skb, nlh);
2918 goto out;
2919 }
2920skip_entry:
2921 e++;
2922 }
2923 s_e = 0;
2924 e = 0;
2925 nla_nest_end(skb, vifs);
2926 nla_nest_end(skb, af);
2927 nlmsg_end(skb, nlh);
2928skip_table:
2929 t++;
2930 }
2931
2932out:
2933 cb->args[1] = e;
2934 cb->args[0] = t;
2935
2936 return skb->len;
2937}
2938
2939#ifdef CONFIG_PROC_FS
2940/* The /proc interfaces to multicast routing :
2941 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2942 */
2943
2944static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2945 __acquires(RCU)
2946{
2947 struct mr_vif_iter *iter = seq->private;
2948 struct net *net = seq_file_net(seq);
2949 struct mr_table *mrt;
2950
2951 rcu_read_lock();
2952 mrt = __ipmr_get_table(net, RT_TABLE_DEFAULT);
2953 if (!mrt) {
2954 rcu_read_unlock();
2955 return ERR_PTR(-ENOENT);
2956 }
2957
2958 iter->mrt = mrt;
2959
2960 return mr_vif_seq_start(seq, pos);
2961}
2962
2963static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2964 __releases(RCU)
2965{
2966 rcu_read_unlock();
2967}
2968
2969static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2970{
2971 struct mr_vif_iter *iter = seq->private;
2972 struct mr_table *mrt = iter->mrt;
2973
2974 if (v == SEQ_START_TOKEN) {
2975 seq_puts(seq,
2976 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2977 } else {
2978 const struct vif_device *vif = v;
2979 const struct net_device *vif_dev;
2980 const char *name;
2981
2982 vif_dev = vif_dev_read(vif);
2983 name = vif_dev ? vif_dev->name : "none";
2984 seq_printf(seq,
2985 "%2td %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2986 vif - mrt->vif_table,
2987 name, vif->bytes_in, vif->pkt_in,
2988 vif->bytes_out, vif->pkt_out,
2989 vif->flags, vif->local, vif->remote);
2990 }
2991 return 0;
2992}
2993
2994static const struct seq_operations ipmr_vif_seq_ops = {
2995 .start = ipmr_vif_seq_start,
2996 .next = mr_vif_seq_next,
2997 .stop = ipmr_vif_seq_stop,
2998 .show = ipmr_vif_seq_show,
2999};
3000
3001static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
3002{
3003 struct net *net = seq_file_net(seq);
3004 struct mr_table *mrt;
3005
3006 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
3007 if (!mrt)
3008 return ERR_PTR(-ENOENT);
3009
3010 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
3011}
3012
3013static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
3014{
3015 int n;
3016
3017 if (v == SEQ_START_TOKEN) {
3018 seq_puts(seq,
3019 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
3020 } else {
3021 const struct mfc_cache *mfc = v;
3022 const struct mr_mfc_iter *it = seq->private;
3023 const struct mr_table *mrt = it->mrt;
3024
3025 seq_printf(seq, "%08X %08X %-3hd",
3026 (__force u32) mfc->mfc_mcastgrp,
3027 (__force u32) mfc->mfc_origin,
3028 mfc->_c.mfc_parent);
3029
3030 if (it->cache != &mrt->mfc_unres_queue) {
3031 seq_printf(seq, " %8lu %8lu %8lu",
3032 atomic_long_read(&mfc->_c.mfc_un.res.pkt),
3033 atomic_long_read(&mfc->_c.mfc_un.res.bytes),
3034 atomic_long_read(&mfc->_c.mfc_un.res.wrong_if));
3035 for (n = mfc->_c.mfc_un.res.minvif;
3036 n < mfc->_c.mfc_un.res.maxvif; n++) {
3037 if (VIF_EXISTS(mrt, n) &&
3038 mfc->_c.mfc_un.res.ttls[n] < 255)
3039 seq_printf(seq,
3040 " %2d:%-3d",
3041 n, mfc->_c.mfc_un.res.ttls[n]);
3042 }
3043 } else {
3044 /* unresolved mfc_caches don't contain
3045 * pkt, bytes and wrong_if values
3046 */
3047 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
3048 }
3049 seq_putc(seq, '\n');
3050 }
3051 return 0;
3052}
3053
3054static const struct seq_operations ipmr_mfc_seq_ops = {
3055 .start = ipmr_mfc_seq_start,
3056 .next = mr_mfc_seq_next,
3057 .stop = mr_mfc_seq_stop,
3058 .show = ipmr_mfc_seq_show,
3059};
3060#endif
3061
3062#ifdef CONFIG_IP_PIMSM_V2
3063static const struct net_protocol pim_protocol = {
3064 .handler = pim_rcv,
3065};
3066#endif
3067
3068static unsigned int ipmr_seq_read(const struct net *net)
3069{
3070 return READ_ONCE(net->ipv4.ipmr_seq) + ipmr_rules_seq_read(net);
3071}
3072
3073static int ipmr_dump(struct net *net, struct notifier_block *nb,
3074 struct netlink_ext_ack *extack)
3075{
3076 return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
3077 ipmr_mr_table_iter, extack);
3078}
3079
3080static const struct fib_notifier_ops ipmr_notifier_ops_template = {
3081 .family = RTNL_FAMILY_IPMR,
3082 .fib_seq_read = ipmr_seq_read,
3083 .fib_dump = ipmr_dump,
3084 .owner = THIS_MODULE,
3085};
3086
3087static int __net_init ipmr_notifier_init(struct net *net)
3088{
3089 struct fib_notifier_ops *ops;
3090
3091 net->ipv4.ipmr_seq = 0;
3092
3093 ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
3094 if (IS_ERR(ops))
3095 return PTR_ERR(ops);
3096 net->ipv4.ipmr_notifier_ops = ops;
3097
3098 return 0;
3099}
3100
3101static void __net_exit ipmr_notifier_exit(struct net *net)
3102{
3103 fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
3104 net->ipv4.ipmr_notifier_ops = NULL;
3105}
3106
3107/* Setup for IP multicast routing */
3108static int __net_init ipmr_net_init(struct net *net)
3109{
3110 int err;
3111
3112 err = ipmr_notifier_init(net);
3113 if (err)
3114 goto ipmr_notifier_fail;
3115
3116 err = ipmr_rules_init(net);
3117 if (err < 0)
3118 goto ipmr_rules_fail;
3119
3120#ifdef CONFIG_PROC_FS
3121 err = -ENOMEM;
3122 if (!proc_create_net("ip_mr_vif", 0, net->proc_net, &ipmr_vif_seq_ops,
3123 sizeof(struct mr_vif_iter)))
3124 goto proc_vif_fail;
3125 if (!proc_create_net("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_seq_ops,
3126 sizeof(struct mr_mfc_iter)))
3127 goto proc_cache_fail;
3128#endif
3129 return 0;
3130
3131#ifdef CONFIG_PROC_FS
3132proc_cache_fail:
3133 remove_proc_entry("ip_mr_vif", net->proc_net);
3134proc_vif_fail:
3135 rtnl_lock();
3136 ipmr_rules_exit(net);
3137 rtnl_unlock();
3138#endif
3139ipmr_rules_fail:
3140 ipmr_notifier_exit(net);
3141ipmr_notifier_fail:
3142 return err;
3143}
3144
3145static void __net_exit ipmr_net_exit(struct net *net)
3146{
3147#ifdef CONFIG_PROC_FS
3148 remove_proc_entry("ip_mr_cache", net->proc_net);
3149 remove_proc_entry("ip_mr_vif", net->proc_net);
3150#endif
3151 ipmr_notifier_exit(net);
3152}
3153
3154static void __net_exit ipmr_net_exit_batch(struct list_head *net_list)
3155{
3156 struct net *net;
3157
3158 rtnl_lock();
3159 list_for_each_entry(net, net_list, exit_list)
3160 ipmr_rules_exit(net);
3161 rtnl_unlock();
3162}
3163
3164static struct pernet_operations ipmr_net_ops = {
3165 .init = ipmr_net_init,
3166 .exit = ipmr_net_exit,
3167 .exit_batch = ipmr_net_exit_batch,
3168};
3169
3170static const struct rtnl_msg_handler ipmr_rtnl_msg_handlers[] __initconst = {
3171 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_GETLINK,
3172 .dumpit = ipmr_rtm_dumplink},
3173 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_NEWROUTE,
3174 .doit = ipmr_rtm_route},
3175 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_DELROUTE,
3176 .doit = ipmr_rtm_route},
3177 {.protocol = RTNL_FAMILY_IPMR, .msgtype = RTM_GETROUTE,
3178 .doit = ipmr_rtm_getroute, .dumpit = ipmr_rtm_dumproute},
3179};
3180
3181int __init ip_mr_init(void)
3182{
3183 int err;
3184
3185 mrt_cachep = KMEM_CACHE(mfc_cache, SLAB_HWCACHE_ALIGN | SLAB_PANIC);
3186
3187 err = register_pernet_subsys(&ipmr_net_ops);
3188 if (err)
3189 goto reg_pernet_fail;
3190
3191 err = register_netdevice_notifier(&ip_mr_notifier);
3192 if (err)
3193 goto reg_notif_fail;
3194#ifdef CONFIG_IP_PIMSM_V2
3195 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3196 pr_err("%s: can't add PIM protocol\n", __func__);
3197 err = -EAGAIN;
3198 goto add_proto_fail;
3199 }
3200#endif
3201 rtnl_register_many(ipmr_rtnl_msg_handlers);
3202
3203 return 0;
3204
3205#ifdef CONFIG_IP_PIMSM_V2
3206add_proto_fail:
3207 unregister_netdevice_notifier(&ip_mr_notifier);
3208#endif
3209reg_notif_fail:
3210 unregister_pernet_subsys(&ipmr_net_ops);
3211reg_pernet_fail:
3212 kmem_cache_destroy(mrt_cachep);
3213 return err;
3214}