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 <linux/uaccess.h>
30#include <linux/types.h>
31#include <linux/cache.h>
32#include <linux/capability.h>
33#include <linux/errno.h>
34#include <linux/mm.h>
35#include <linux/kernel.h>
36#include <linux/fcntl.h>
37#include <linux/stat.h>
38#include <linux/socket.h>
39#include <linux/in.h>
40#include <linux/inet.h>
41#include <linux/netdevice.h>
42#include <linux/inetdevice.h>
43#include <linux/igmp.h>
44#include <linux/proc_fs.h>
45#include <linux/seq_file.h>
46#include <linux/mroute.h>
47#include <linux/init.h>
48#include <linux/if_ether.h>
49#include <linux/slab.h>
50#include <net/net_namespace.h>
51#include <net/ip.h>
52#include <net/protocol.h>
53#include <linux/skbuff.h>
54#include <net/route.h>
55#include <net/icmp.h>
56#include <net/udp.h>
57#include <net/raw.h>
58#include <linux/notifier.h>
59#include <linux/if_arp.h>
60#include <linux/netfilter_ipv4.h>
61#include <linux/compat.h>
62#include <linux/export.h>
63#include <net/ip_tunnels.h>
64#include <net/checksum.h>
65#include <net/netlink.h>
66#include <net/fib_rules.h>
67#include <linux/netconf.h>
68#include <net/nexthop.h>
69#include <net/switchdev.h>
70
71struct ipmr_rule {
72 struct fib_rule common;
73};
74
75struct ipmr_result {
76 struct mr_table *mrt;
77};
78
79/* Big lock, protecting vif table, mrt cache and mroute socket state.
80 * Note that the changes are semaphored via rtnl_lock.
81 */
82
83static DEFINE_RWLOCK(mrt_lock);
84
85/* Multicast router control variables */
86
87/* Special spinlock for queue of unresolved entries */
88static DEFINE_SPINLOCK(mfc_unres_lock);
89
90/* We return to original Alan's scheme. Hash table of resolved
91 * entries is changed only in process context and protected
92 * with weak lock mrt_lock. Queue of unresolved entries is protected
93 * with strong spinlock mfc_unres_lock.
94 *
95 * In this case data path is free of exclusive locks at all.
96 */
97
98static struct kmem_cache *mrt_cachep __ro_after_init;
99
100static struct mr_table *ipmr_new_table(struct net *net, u32 id);
101static void ipmr_free_table(struct mr_table *mrt);
102
103static void ip_mr_forward(struct net *net, struct mr_table *mrt,
104 struct net_device *dev, struct sk_buff *skb,
105 struct mfc_cache *cache, int local);
106static int ipmr_cache_report(struct mr_table *mrt,
107 struct sk_buff *pkt, vifi_t vifi, int assert);
108static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
109 int cmd);
110static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt);
111static void mroute_clean_tables(struct mr_table *mrt, bool all);
112static void ipmr_expire_process(struct timer_list *t);
113
114#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
115#define ipmr_for_each_table(mrt, net) \
116 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
117
118static struct mr_table *ipmr_mr_table_iter(struct net *net,
119 struct mr_table *mrt)
120{
121 struct mr_table *ret;
122
123 if (!mrt)
124 ret = list_entry_rcu(net->ipv4.mr_tables.next,
125 struct mr_table, list);
126 else
127 ret = list_entry_rcu(mrt->list.next,
128 struct mr_table, list);
129
130 if (&ret->list == &net->ipv4.mr_tables)
131 return NULL;
132 return ret;
133}
134
135static struct mr_table *ipmr_get_table(struct net *net, u32 id)
136{
137 struct mr_table *mrt;
138
139 ipmr_for_each_table(mrt, net) {
140 if (mrt->id == id)
141 return mrt;
142 }
143 return NULL;
144}
145
146static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
147 struct mr_table **mrt)
148{
149 int err;
150 struct ipmr_result res;
151 struct fib_lookup_arg arg = {
152 .result = &res,
153 .flags = FIB_LOOKUP_NOREF,
154 };
155
156 /* update flow if oif or iif point to device enslaved to l3mdev */
157 l3mdev_update_flow(net, flowi4_to_flowi(flp4));
158
159 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
160 flowi4_to_flowi(flp4), 0, &arg);
161 if (err < 0)
162 return err;
163 *mrt = res.mrt;
164 return 0;
165}
166
167static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
168 int flags, struct fib_lookup_arg *arg)
169{
170 struct ipmr_result *res = arg->result;
171 struct mr_table *mrt;
172
173 switch (rule->action) {
174 case FR_ACT_TO_TBL:
175 break;
176 case FR_ACT_UNREACHABLE:
177 return -ENETUNREACH;
178 case FR_ACT_PROHIBIT:
179 return -EACCES;
180 case FR_ACT_BLACKHOLE:
181 default:
182 return -EINVAL;
183 }
184
185 arg->table = fib_rule_get_table(rule, arg);
186
187 mrt = ipmr_get_table(rule->fr_net, arg->table);
188 if (!mrt)
189 return -EAGAIN;
190 res->mrt = mrt;
191 return 0;
192}
193
194static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
195{
196 return 1;
197}
198
199static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
200 FRA_GENERIC_POLICY,
201};
202
203static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
204 struct fib_rule_hdr *frh, struct nlattr **tb)
205{
206 return 0;
207}
208
209static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
210 struct nlattr **tb)
211{
212 return 1;
213}
214
215static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
216 struct fib_rule_hdr *frh)
217{
218 frh->dst_len = 0;
219 frh->src_len = 0;
220 frh->tos = 0;
221 return 0;
222}
223
224static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
225 .family = RTNL_FAMILY_IPMR,
226 .rule_size = sizeof(struct ipmr_rule),
227 .addr_size = sizeof(u32),
228 .action = ipmr_rule_action,
229 .match = ipmr_rule_match,
230 .configure = ipmr_rule_configure,
231 .compare = ipmr_rule_compare,
232 .fill = ipmr_rule_fill,
233 .nlgroup = RTNLGRP_IPV4_RULE,
234 .policy = ipmr_rule_policy,
235 .owner = THIS_MODULE,
236};
237
238static int __net_init ipmr_rules_init(struct net *net)
239{
240 struct fib_rules_ops *ops;
241 struct mr_table *mrt;
242 int err;
243
244 ops = fib_rules_register(&ipmr_rules_ops_template, net);
245 if (IS_ERR(ops))
246 return PTR_ERR(ops);
247
248 INIT_LIST_HEAD(&net->ipv4.mr_tables);
249
250 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
251 if (IS_ERR(mrt)) {
252 err = PTR_ERR(mrt);
253 goto err1;
254 }
255
256 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
257 if (err < 0)
258 goto err2;
259
260 net->ipv4.mr_rules_ops = ops;
261 return 0;
262
263err2:
264 ipmr_free_table(mrt);
265err1:
266 fib_rules_unregister(ops);
267 return err;
268}
269
270static void __net_exit ipmr_rules_exit(struct net *net)
271{
272 struct mr_table *mrt, *next;
273
274 rtnl_lock();
275 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
276 list_del(&mrt->list);
277 ipmr_free_table(mrt);
278 }
279 fib_rules_unregister(net->ipv4.mr_rules_ops);
280 rtnl_unlock();
281}
282
283static int ipmr_rules_dump(struct net *net, struct notifier_block *nb)
284{
285 return fib_rules_dump(net, nb, RTNL_FAMILY_IPMR);
286}
287
288static unsigned int ipmr_rules_seq_read(struct net *net)
289{
290 return fib_rules_seq_read(net, RTNL_FAMILY_IPMR);
291}
292
293bool ipmr_rule_default(const struct fib_rule *rule)
294{
295 return fib_rule_matchall(rule) && rule->table == RT_TABLE_DEFAULT;
296}
297EXPORT_SYMBOL(ipmr_rule_default);
298#else
299#define ipmr_for_each_table(mrt, net) \
300 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
301
302static struct mr_table *ipmr_mr_table_iter(struct net *net,
303 struct mr_table *mrt)
304{
305 if (!mrt)
306 return net->ipv4.mrt;
307 return NULL;
308}
309
310static struct mr_table *ipmr_get_table(struct net *net, u32 id)
311{
312 return net->ipv4.mrt;
313}
314
315static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
316 struct mr_table **mrt)
317{
318 *mrt = net->ipv4.mrt;
319 return 0;
320}
321
322static int __net_init ipmr_rules_init(struct net *net)
323{
324 struct mr_table *mrt;
325
326 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
327 if (IS_ERR(mrt))
328 return PTR_ERR(mrt);
329 net->ipv4.mrt = mrt;
330 return 0;
331}
332
333static void __net_exit ipmr_rules_exit(struct net *net)
334{
335 rtnl_lock();
336 ipmr_free_table(net->ipv4.mrt);
337 net->ipv4.mrt = NULL;
338 rtnl_unlock();
339}
340
341static int ipmr_rules_dump(struct net *net, struct notifier_block *nb)
342{
343 return 0;
344}
345
346static unsigned int ipmr_rules_seq_read(struct net *net)
347{
348 return 0;
349}
350
351bool ipmr_rule_default(const struct fib_rule *rule)
352{
353 return true;
354}
355EXPORT_SYMBOL(ipmr_rule_default);
356#endif
357
358static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
359 const void *ptr)
360{
361 const struct mfc_cache_cmp_arg *cmparg = arg->key;
362 struct mfc_cache *c = (struct mfc_cache *)ptr;
363
364 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
365 cmparg->mfc_origin != c->mfc_origin;
366}
367
368static const struct rhashtable_params ipmr_rht_params = {
369 .head_offset = offsetof(struct mr_mfc, mnode),
370 .key_offset = offsetof(struct mfc_cache, cmparg),
371 .key_len = sizeof(struct mfc_cache_cmp_arg),
372 .nelem_hint = 3,
373 .locks_mul = 1,
374 .obj_cmpfn = ipmr_hash_cmp,
375 .automatic_shrinking = true,
376};
377
378static void ipmr_new_table_set(struct mr_table *mrt,
379 struct net *net)
380{
381#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
382 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
383#endif
384}
385
386static struct mfc_cache_cmp_arg ipmr_mr_table_ops_cmparg_any = {
387 .mfc_mcastgrp = htonl(INADDR_ANY),
388 .mfc_origin = htonl(INADDR_ANY),
389};
390
391static struct mr_table_ops ipmr_mr_table_ops = {
392 .rht_params = &ipmr_rht_params,
393 .cmparg_any = &ipmr_mr_table_ops_cmparg_any,
394};
395
396static struct mr_table *ipmr_new_table(struct net *net, u32 id)
397{
398 struct mr_table *mrt;
399
400 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
401 if (id != RT_TABLE_DEFAULT && id >= 1000000000)
402 return ERR_PTR(-EINVAL);
403
404 mrt = ipmr_get_table(net, id);
405 if (mrt)
406 return mrt;
407
408 return mr_table_alloc(net, id, &ipmr_mr_table_ops,
409 ipmr_expire_process, ipmr_new_table_set);
410}
411
412static void ipmr_free_table(struct mr_table *mrt)
413{
414 del_timer_sync(&mrt->ipmr_expire_timer);
415 mroute_clean_tables(mrt, true);
416 rhltable_destroy(&mrt->mfc_hash);
417 kfree(mrt);
418}
419
420/* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
421
422static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
423{
424 struct net *net = dev_net(dev);
425
426 dev_close(dev);
427
428 dev = __dev_get_by_name(net, "tunl0");
429 if (dev) {
430 const struct net_device_ops *ops = dev->netdev_ops;
431 struct ifreq ifr;
432 struct ip_tunnel_parm p;
433
434 memset(&p, 0, sizeof(p));
435 p.iph.daddr = v->vifc_rmt_addr.s_addr;
436 p.iph.saddr = v->vifc_lcl_addr.s_addr;
437 p.iph.version = 4;
438 p.iph.ihl = 5;
439 p.iph.protocol = IPPROTO_IPIP;
440 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
441 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
442
443 if (ops->ndo_do_ioctl) {
444 mm_segment_t oldfs = get_fs();
445
446 set_fs(KERNEL_DS);
447 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
448 set_fs(oldfs);
449 }
450 }
451}
452
453/* Initialize ipmr pimreg/tunnel in_device */
454static bool ipmr_init_vif_indev(const struct net_device *dev)
455{
456 struct in_device *in_dev;
457
458 ASSERT_RTNL();
459
460 in_dev = __in_dev_get_rtnl(dev);
461 if (!in_dev)
462 return false;
463 ipv4_devconf_setall(in_dev);
464 neigh_parms_data_state_setall(in_dev->arp_parms);
465 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
466
467 return true;
468}
469
470static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
471{
472 struct net_device *dev;
473
474 dev = __dev_get_by_name(net, "tunl0");
475
476 if (dev) {
477 const struct net_device_ops *ops = dev->netdev_ops;
478 int err;
479 struct ifreq ifr;
480 struct ip_tunnel_parm p;
481
482 memset(&p, 0, sizeof(p));
483 p.iph.daddr = v->vifc_rmt_addr.s_addr;
484 p.iph.saddr = v->vifc_lcl_addr.s_addr;
485 p.iph.version = 4;
486 p.iph.ihl = 5;
487 p.iph.protocol = IPPROTO_IPIP;
488 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
489 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
490
491 if (ops->ndo_do_ioctl) {
492 mm_segment_t oldfs = get_fs();
493
494 set_fs(KERNEL_DS);
495 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
496 set_fs(oldfs);
497 } else {
498 err = -EOPNOTSUPP;
499 }
500 dev = NULL;
501
502 if (err == 0 &&
503 (dev = __dev_get_by_name(net, p.name)) != NULL) {
504 dev->flags |= IFF_MULTICAST;
505 if (!ipmr_init_vif_indev(dev))
506 goto failure;
507 if (dev_open(dev))
508 goto failure;
509 dev_hold(dev);
510 }
511 }
512 return dev;
513
514failure:
515 unregister_netdevice(dev);
516 return NULL;
517}
518
519#if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
520static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
521{
522 struct net *net = dev_net(dev);
523 struct mr_table *mrt;
524 struct flowi4 fl4 = {
525 .flowi4_oif = dev->ifindex,
526 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
527 .flowi4_mark = skb->mark,
528 };
529 int err;
530
531 err = ipmr_fib_lookup(net, &fl4, &mrt);
532 if (err < 0) {
533 kfree_skb(skb);
534 return err;
535 }
536
537 read_lock(&mrt_lock);
538 dev->stats.tx_bytes += skb->len;
539 dev->stats.tx_packets++;
540 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
541 read_unlock(&mrt_lock);
542 kfree_skb(skb);
543 return NETDEV_TX_OK;
544}
545
546static int reg_vif_get_iflink(const struct net_device *dev)
547{
548 return 0;
549}
550
551static const struct net_device_ops reg_vif_netdev_ops = {
552 .ndo_start_xmit = reg_vif_xmit,
553 .ndo_get_iflink = reg_vif_get_iflink,
554};
555
556static void reg_vif_setup(struct net_device *dev)
557{
558 dev->type = ARPHRD_PIMREG;
559 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
560 dev->flags = IFF_NOARP;
561 dev->netdev_ops = ®_vif_netdev_ops;
562 dev->needs_free_netdev = true;
563 dev->features |= NETIF_F_NETNS_LOCAL;
564}
565
566static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
567{
568 struct net_device *dev;
569 char name[IFNAMSIZ];
570
571 if (mrt->id == RT_TABLE_DEFAULT)
572 sprintf(name, "pimreg");
573 else
574 sprintf(name, "pimreg%u", mrt->id);
575
576 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
577
578 if (!dev)
579 return NULL;
580
581 dev_net_set(dev, net);
582
583 if (register_netdevice(dev)) {
584 free_netdev(dev);
585 return NULL;
586 }
587
588 if (!ipmr_init_vif_indev(dev))
589 goto failure;
590 if (dev_open(dev))
591 goto failure;
592
593 dev_hold(dev);
594
595 return dev;
596
597failure:
598 unregister_netdevice(dev);
599 return NULL;
600}
601
602/* called with rcu_read_lock() */
603static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
604 unsigned int pimlen)
605{
606 struct net_device *reg_dev = NULL;
607 struct iphdr *encap;
608
609 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
610 /* Check that:
611 * a. packet is really sent to a multicast group
612 * b. packet is not a NULL-REGISTER
613 * c. packet is not truncated
614 */
615 if (!ipv4_is_multicast(encap->daddr) ||
616 encap->tot_len == 0 ||
617 ntohs(encap->tot_len) + pimlen > skb->len)
618 return 1;
619
620 read_lock(&mrt_lock);
621 if (mrt->mroute_reg_vif_num >= 0)
622 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
623 read_unlock(&mrt_lock);
624
625 if (!reg_dev)
626 return 1;
627
628 skb->mac_header = skb->network_header;
629 skb_pull(skb, (u8 *)encap - skb->data);
630 skb_reset_network_header(skb);
631 skb->protocol = htons(ETH_P_IP);
632 skb->ip_summed = CHECKSUM_NONE;
633
634 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
635
636 netif_rx(skb);
637
638 return NET_RX_SUCCESS;
639}
640#else
641static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
642{
643 return NULL;
644}
645#endif
646
647static int call_ipmr_vif_entry_notifiers(struct net *net,
648 enum fib_event_type event_type,
649 struct vif_device *vif,
650 vifi_t vif_index, u32 tb_id)
651{
652 return mr_call_vif_notifiers(net, RTNL_FAMILY_IPMR, event_type,
653 vif, vif_index, tb_id,
654 &net->ipv4.ipmr_seq);
655}
656
657static int call_ipmr_mfc_entry_notifiers(struct net *net,
658 enum fib_event_type event_type,
659 struct mfc_cache *mfc, u32 tb_id)
660{
661 return mr_call_mfc_notifiers(net, RTNL_FAMILY_IPMR, event_type,
662 &mfc->_c, tb_id, &net->ipv4.ipmr_seq);
663}
664
665/**
666 * vif_delete - Delete a VIF entry
667 * @notify: Set to 1, if the caller is a notifier_call
668 */
669static int vif_delete(struct mr_table *mrt, int vifi, int notify,
670 struct list_head *head)
671{
672 struct net *net = read_pnet(&mrt->net);
673 struct vif_device *v;
674 struct net_device *dev;
675 struct in_device *in_dev;
676
677 if (vifi < 0 || vifi >= mrt->maxvif)
678 return -EADDRNOTAVAIL;
679
680 v = &mrt->vif_table[vifi];
681
682 if (VIF_EXISTS(mrt, vifi))
683 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_DEL, v, vifi,
684 mrt->id);
685
686 write_lock_bh(&mrt_lock);
687 dev = v->dev;
688 v->dev = NULL;
689
690 if (!dev) {
691 write_unlock_bh(&mrt_lock);
692 return -EADDRNOTAVAIL;
693 }
694
695 if (vifi == mrt->mroute_reg_vif_num)
696 mrt->mroute_reg_vif_num = -1;
697
698 if (vifi + 1 == mrt->maxvif) {
699 int tmp;
700
701 for (tmp = vifi - 1; tmp >= 0; tmp--) {
702 if (VIF_EXISTS(mrt, tmp))
703 break;
704 }
705 mrt->maxvif = tmp+1;
706 }
707
708 write_unlock_bh(&mrt_lock);
709
710 dev_set_allmulti(dev, -1);
711
712 in_dev = __in_dev_get_rtnl(dev);
713 if (in_dev) {
714 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
715 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
716 NETCONFA_MC_FORWARDING,
717 dev->ifindex, &in_dev->cnf);
718 ip_rt_multicast_event(in_dev);
719 }
720
721 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
722 unregister_netdevice_queue(dev, head);
723
724 dev_put(dev);
725 return 0;
726}
727
728static void ipmr_cache_free_rcu(struct rcu_head *head)
729{
730 struct mr_mfc *c = container_of(head, struct mr_mfc, rcu);
731
732 kmem_cache_free(mrt_cachep, (struct mfc_cache *)c);
733}
734
735static void ipmr_cache_free(struct mfc_cache *c)
736{
737 call_rcu(&c->_c.rcu, ipmr_cache_free_rcu);
738}
739
740/* Destroy an unresolved cache entry, killing queued skbs
741 * and reporting error to netlink readers.
742 */
743static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
744{
745 struct net *net = read_pnet(&mrt->net);
746 struct sk_buff *skb;
747 struct nlmsgerr *e;
748
749 atomic_dec(&mrt->cache_resolve_queue_len);
750
751 while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
752 if (ip_hdr(skb)->version == 0) {
753 struct nlmsghdr *nlh = skb_pull(skb,
754 sizeof(struct iphdr));
755 nlh->nlmsg_type = NLMSG_ERROR;
756 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
757 skb_trim(skb, nlh->nlmsg_len);
758 e = nlmsg_data(nlh);
759 e->error = -ETIMEDOUT;
760 memset(&e->msg, 0, sizeof(e->msg));
761
762 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
763 } else {
764 kfree_skb(skb);
765 }
766 }
767
768 ipmr_cache_free(c);
769}
770
771/* Timer process for the unresolved queue. */
772static void ipmr_expire_process(struct timer_list *t)
773{
774 struct mr_table *mrt = from_timer(mrt, t, ipmr_expire_timer);
775 struct mr_mfc *c, *next;
776 unsigned long expires;
777 unsigned long now;
778
779 if (!spin_trylock(&mfc_unres_lock)) {
780 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
781 return;
782 }
783
784 if (list_empty(&mrt->mfc_unres_queue))
785 goto out;
786
787 now = jiffies;
788 expires = 10*HZ;
789
790 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
791 if (time_after(c->mfc_un.unres.expires, now)) {
792 unsigned long interval = c->mfc_un.unres.expires - now;
793 if (interval < expires)
794 expires = interval;
795 continue;
796 }
797
798 list_del(&c->list);
799 mroute_netlink_event(mrt, (struct mfc_cache *)c, RTM_DELROUTE);
800 ipmr_destroy_unres(mrt, (struct mfc_cache *)c);
801 }
802
803 if (!list_empty(&mrt->mfc_unres_queue))
804 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
805
806out:
807 spin_unlock(&mfc_unres_lock);
808}
809
810/* Fill oifs list. It is called under write locked mrt_lock. */
811static void ipmr_update_thresholds(struct mr_table *mrt, struct mr_mfc *cache,
812 unsigned char *ttls)
813{
814 int vifi;
815
816 cache->mfc_un.res.minvif = MAXVIFS;
817 cache->mfc_un.res.maxvif = 0;
818 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
819
820 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
821 if (VIF_EXISTS(mrt, vifi) &&
822 ttls[vifi] && ttls[vifi] < 255) {
823 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
824 if (cache->mfc_un.res.minvif > vifi)
825 cache->mfc_un.res.minvif = vifi;
826 if (cache->mfc_un.res.maxvif <= vifi)
827 cache->mfc_un.res.maxvif = vifi + 1;
828 }
829 }
830 cache->mfc_un.res.lastuse = jiffies;
831}
832
833static int vif_add(struct net *net, struct mr_table *mrt,
834 struct vifctl *vifc, int mrtsock)
835{
836 int vifi = vifc->vifc_vifi;
837 struct switchdev_attr attr = {
838 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
839 };
840 struct vif_device *v = &mrt->vif_table[vifi];
841 struct net_device *dev;
842 struct in_device *in_dev;
843 int err;
844
845 /* Is vif busy ? */
846 if (VIF_EXISTS(mrt, vifi))
847 return -EADDRINUSE;
848
849 switch (vifc->vifc_flags) {
850 case VIFF_REGISTER:
851 if (!ipmr_pimsm_enabled())
852 return -EINVAL;
853 /* Special Purpose VIF in PIM
854 * All the packets will be sent to the daemon
855 */
856 if (mrt->mroute_reg_vif_num >= 0)
857 return -EADDRINUSE;
858 dev = ipmr_reg_vif(net, mrt);
859 if (!dev)
860 return -ENOBUFS;
861 err = dev_set_allmulti(dev, 1);
862 if (err) {
863 unregister_netdevice(dev);
864 dev_put(dev);
865 return err;
866 }
867 break;
868 case VIFF_TUNNEL:
869 dev = ipmr_new_tunnel(net, vifc);
870 if (!dev)
871 return -ENOBUFS;
872 err = dev_set_allmulti(dev, 1);
873 if (err) {
874 ipmr_del_tunnel(dev, vifc);
875 dev_put(dev);
876 return err;
877 }
878 break;
879 case VIFF_USE_IFINDEX:
880 case 0:
881 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
882 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
883 if (dev && !__in_dev_get_rtnl(dev)) {
884 dev_put(dev);
885 return -EADDRNOTAVAIL;
886 }
887 } else {
888 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
889 }
890 if (!dev)
891 return -EADDRNOTAVAIL;
892 err = dev_set_allmulti(dev, 1);
893 if (err) {
894 dev_put(dev);
895 return err;
896 }
897 break;
898 default:
899 return -EINVAL;
900 }
901
902 in_dev = __in_dev_get_rtnl(dev);
903 if (!in_dev) {
904 dev_put(dev);
905 return -EADDRNOTAVAIL;
906 }
907 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
908 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
909 dev->ifindex, &in_dev->cnf);
910 ip_rt_multicast_event(in_dev);
911
912 /* Fill in the VIF structures */
913 vif_device_init(v, dev, vifc->vifc_rate_limit,
914 vifc->vifc_threshold,
915 vifc->vifc_flags | (!mrtsock ? VIFF_STATIC : 0),
916 (VIFF_TUNNEL | VIFF_REGISTER));
917
918 attr.orig_dev = dev;
919 if (!switchdev_port_attr_get(dev, &attr)) {
920 memcpy(v->dev_parent_id.id, attr.u.ppid.id, attr.u.ppid.id_len);
921 v->dev_parent_id.id_len = attr.u.ppid.id_len;
922 } else {
923 v->dev_parent_id.id_len = 0;
924 }
925
926 v->local = vifc->vifc_lcl_addr.s_addr;
927 v->remote = vifc->vifc_rmt_addr.s_addr;
928
929 /* And finish update writing critical data */
930 write_lock_bh(&mrt_lock);
931 v->dev = dev;
932 if (v->flags & VIFF_REGISTER)
933 mrt->mroute_reg_vif_num = vifi;
934 if (vifi+1 > mrt->maxvif)
935 mrt->maxvif = vifi+1;
936 write_unlock_bh(&mrt_lock);
937 call_ipmr_vif_entry_notifiers(net, FIB_EVENT_VIF_ADD, v, 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 ip_mr_forward(net, mrt, skb->dev, skb, c, 0);
1035 }
1036 }
1037}
1038
1039/* Bounce a cache query up to mrouted and netlink.
1040 *
1041 * Called under mrt_lock.
1042 */
1043static int ipmr_cache_report(struct mr_table *mrt,
1044 struct sk_buff *pkt, vifi_t vifi, int assert)
1045{
1046 const int ihl = ip_hdrlen(pkt);
1047 struct sock *mroute_sk;
1048 struct igmphdr *igmp;
1049 struct igmpmsg *msg;
1050 struct sk_buff *skb;
1051 int ret;
1052
1053 if (assert == IGMPMSG_WHOLEPKT)
1054 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1055 else
1056 skb = alloc_skb(128, GFP_ATOMIC);
1057
1058 if (!skb)
1059 return -ENOBUFS;
1060
1061 if (assert == IGMPMSG_WHOLEPKT) {
1062 /* Ugly, but we have no choice with this interface.
1063 * Duplicate old header, fix ihl, length etc.
1064 * And all this only to mangle msg->im_msgtype and
1065 * to set msg->im_mbz to "mbz" :-)
1066 */
1067 skb_push(skb, sizeof(struct iphdr));
1068 skb_reset_network_header(skb);
1069 skb_reset_transport_header(skb);
1070 msg = (struct igmpmsg *)skb_network_header(skb);
1071 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1072 msg->im_msgtype = IGMPMSG_WHOLEPKT;
1073 msg->im_mbz = 0;
1074 msg->im_vif = mrt->mroute_reg_vif_num;
1075 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1076 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1077 sizeof(struct iphdr));
1078 } else {
1079 /* Copy the IP header */
1080 skb_set_network_header(skb, skb->len);
1081 skb_put(skb, ihl);
1082 skb_copy_to_linear_data(skb, pkt->data, ihl);
1083 /* Flag to the kernel this is a route add */
1084 ip_hdr(skb)->protocol = 0;
1085 msg = (struct igmpmsg *)skb_network_header(skb);
1086 msg->im_vif = vifi;
1087 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1088 /* Add our header */
1089 igmp = skb_put(skb, sizeof(struct igmphdr));
1090 igmp->type = assert;
1091 msg->im_msgtype = assert;
1092 igmp->code = 0;
1093 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1094 skb->transport_header = skb->network_header;
1095 }
1096
1097 rcu_read_lock();
1098 mroute_sk = rcu_dereference(mrt->mroute_sk);
1099 if (!mroute_sk) {
1100 rcu_read_unlock();
1101 kfree_skb(skb);
1102 return -EINVAL;
1103 }
1104
1105 igmpmsg_netlink_event(mrt, skb);
1106
1107 /* Deliver to mrouted */
1108 ret = sock_queue_rcv_skb(mroute_sk, skb);
1109 rcu_read_unlock();
1110 if (ret < 0) {
1111 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1112 kfree_skb(skb);
1113 }
1114
1115 return ret;
1116}
1117
1118/* Queue a packet for resolution. It gets locked cache entry! */
1119static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1120 struct sk_buff *skb, struct net_device *dev)
1121{
1122 const struct iphdr *iph = ip_hdr(skb);
1123 struct mfc_cache *c;
1124 bool found = false;
1125 int err;
1126
1127 spin_lock_bh(&mfc_unres_lock);
1128 list_for_each_entry(c, &mrt->mfc_unres_queue, _c.list) {
1129 if (c->mfc_mcastgrp == iph->daddr &&
1130 c->mfc_origin == iph->saddr) {
1131 found = true;
1132 break;
1133 }
1134 }
1135
1136 if (!found) {
1137 /* Create a new entry if allowable */
1138 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1139 (c = ipmr_cache_alloc_unres()) == NULL) {
1140 spin_unlock_bh(&mfc_unres_lock);
1141
1142 kfree_skb(skb);
1143 return -ENOBUFS;
1144 }
1145
1146 /* Fill in the new cache entry */
1147 c->_c.mfc_parent = -1;
1148 c->mfc_origin = iph->saddr;
1149 c->mfc_mcastgrp = iph->daddr;
1150
1151 /* Reflect first query at mrouted. */
1152 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1153
1154 if (err < 0) {
1155 /* If the report failed throw the cache entry
1156 out - Brad Parker
1157 */
1158 spin_unlock_bh(&mfc_unres_lock);
1159
1160 ipmr_cache_free(c);
1161 kfree_skb(skb);
1162 return err;
1163 }
1164
1165 atomic_inc(&mrt->cache_resolve_queue_len);
1166 list_add(&c->_c.list, &mrt->mfc_unres_queue);
1167 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1168
1169 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1170 mod_timer(&mrt->ipmr_expire_timer,
1171 c->_c.mfc_un.unres.expires);
1172 }
1173
1174 /* See if we can append the packet */
1175 if (c->_c.mfc_un.unres.unresolved.qlen > 3) {
1176 kfree_skb(skb);
1177 err = -ENOBUFS;
1178 } else {
1179 if (dev) {
1180 skb->dev = dev;
1181 skb->skb_iif = dev->ifindex;
1182 }
1183 skb_queue_tail(&c->_c.mfc_un.unres.unresolved, skb);
1184 err = 0;
1185 }
1186
1187 spin_unlock_bh(&mfc_unres_lock);
1188 return err;
1189}
1190
1191/* MFC cache manipulation by user space mroute daemon */
1192
1193static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1194{
1195 struct net *net = read_pnet(&mrt->net);
1196 struct mfc_cache *c;
1197
1198 /* The entries are added/deleted only under RTNL */
1199 rcu_read_lock();
1200 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1201 mfc->mfcc_mcastgrp.s_addr, parent);
1202 rcu_read_unlock();
1203 if (!c)
1204 return -ENOENT;
1205 rhltable_remove(&mrt->mfc_hash, &c->_c.mnode, ipmr_rht_params);
1206 list_del_rcu(&c->_c.list);
1207 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, c, mrt->id);
1208 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1209 mr_cache_put(&c->_c);
1210
1211 return 0;
1212}
1213
1214static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1215 struct mfcctl *mfc, int mrtsock, int parent)
1216{
1217 struct mfc_cache *uc, *c;
1218 struct mr_mfc *_uc;
1219 bool found;
1220 int ret;
1221
1222 if (mfc->mfcc_parent >= MAXVIFS)
1223 return -ENFILE;
1224
1225 /* The entries are added/deleted only under RTNL */
1226 rcu_read_lock();
1227 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1228 mfc->mfcc_mcastgrp.s_addr, parent);
1229 rcu_read_unlock();
1230 if (c) {
1231 write_lock_bh(&mrt_lock);
1232 c->_c.mfc_parent = mfc->mfcc_parent;
1233 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1234 if (!mrtsock)
1235 c->_c.mfc_flags |= MFC_STATIC;
1236 write_unlock_bh(&mrt_lock);
1237 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, c,
1238 mrt->id);
1239 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1240 return 0;
1241 }
1242
1243 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1244 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1245 return -EINVAL;
1246
1247 c = ipmr_cache_alloc();
1248 if (!c)
1249 return -ENOMEM;
1250
1251 c->mfc_origin = mfc->mfcc_origin.s_addr;
1252 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1253 c->_c.mfc_parent = mfc->mfcc_parent;
1254 ipmr_update_thresholds(mrt, &c->_c, mfc->mfcc_ttls);
1255 if (!mrtsock)
1256 c->_c.mfc_flags |= MFC_STATIC;
1257
1258 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->_c.mnode,
1259 ipmr_rht_params);
1260 if (ret) {
1261 pr_err("ipmr: rhtable insert error %d\n", ret);
1262 ipmr_cache_free(c);
1263 return ret;
1264 }
1265 list_add_tail_rcu(&c->_c.list, &mrt->mfc_cache_list);
1266 /* Check to see if we resolved a queued list. If so we
1267 * need to send on the frames and tidy up.
1268 */
1269 found = false;
1270 spin_lock_bh(&mfc_unres_lock);
1271 list_for_each_entry(_uc, &mrt->mfc_unres_queue, list) {
1272 uc = (struct mfc_cache *)_uc;
1273 if (uc->mfc_origin == c->mfc_origin &&
1274 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1275 list_del(&_uc->list);
1276 atomic_dec(&mrt->cache_resolve_queue_len);
1277 found = true;
1278 break;
1279 }
1280 }
1281 if (list_empty(&mrt->mfc_unres_queue))
1282 del_timer(&mrt->ipmr_expire_timer);
1283 spin_unlock_bh(&mfc_unres_lock);
1284
1285 if (found) {
1286 ipmr_cache_resolve(net, mrt, uc, c);
1287 ipmr_cache_free(uc);
1288 }
1289 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_ADD, c, mrt->id);
1290 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1291 return 0;
1292}
1293
1294/* Close the multicast socket, and clear the vif tables etc */
1295static void mroute_clean_tables(struct mr_table *mrt, bool all)
1296{
1297 struct net *net = read_pnet(&mrt->net);
1298 struct mr_mfc *c, *tmp;
1299 struct mfc_cache *cache;
1300 LIST_HEAD(list);
1301 int i;
1302
1303 /* Shut down all active vif entries */
1304 for (i = 0; i < mrt->maxvif; i++) {
1305 if (!all && (mrt->vif_table[i].flags & VIFF_STATIC))
1306 continue;
1307 vif_delete(mrt, i, 0, &list);
1308 }
1309 unregister_netdevice_many(&list);
1310
1311 /* Wipe the cache */
1312 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1313 if (!all && (c->mfc_flags & MFC_STATIC))
1314 continue;
1315 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1316 list_del_rcu(&c->list);
1317 cache = (struct mfc_cache *)c;
1318 call_ipmr_mfc_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, cache,
1319 mrt->id);
1320 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1321 mr_cache_put(c);
1322 }
1323
1324 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1325 spin_lock_bh(&mfc_unres_lock);
1326 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1327 list_del(&c->list);
1328 cache = (struct mfc_cache *)c;
1329 mroute_netlink_event(mrt, cache, RTM_DELROUTE);
1330 ipmr_destroy_unres(mrt, cache);
1331 }
1332 spin_unlock_bh(&mfc_unres_lock);
1333 }
1334}
1335
1336/* called from ip_ra_control(), before an RCU grace period,
1337 * we dont need to call synchronize_rcu() here
1338 */
1339static void mrtsock_destruct(struct sock *sk)
1340{
1341 struct net *net = sock_net(sk);
1342 struct mr_table *mrt;
1343
1344 rtnl_lock();
1345 ipmr_for_each_table(mrt, net) {
1346 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1347 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1348 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1349 NETCONFA_MC_FORWARDING,
1350 NETCONFA_IFINDEX_ALL,
1351 net->ipv4.devconf_all);
1352 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1353 mroute_clean_tables(mrt, false);
1354 }
1355 }
1356 rtnl_unlock();
1357}
1358
1359/* Socket options and virtual interface manipulation. The whole
1360 * virtual interface system is a complete heap, but unfortunately
1361 * that's how BSD mrouted happens to think. Maybe one day with a proper
1362 * MOSPF/PIM router set up we can clean this up.
1363 */
1364
1365int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval,
1366 unsigned int optlen)
1367{
1368 struct net *net = sock_net(sk);
1369 int val, ret = 0, parent = 0;
1370 struct mr_table *mrt;
1371 struct vifctl vif;
1372 struct mfcctl mfc;
1373 u32 uval;
1374
1375 /* There's one exception to the lock - MRT_DONE which needs to unlock */
1376 rtnl_lock();
1377 if (sk->sk_type != SOCK_RAW ||
1378 inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1379 ret = -EOPNOTSUPP;
1380 goto out_unlock;
1381 }
1382
1383 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1384 if (!mrt) {
1385 ret = -ENOENT;
1386 goto out_unlock;
1387 }
1388 if (optname != MRT_INIT) {
1389 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1390 !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1391 ret = -EACCES;
1392 goto out_unlock;
1393 }
1394 }
1395
1396 switch (optname) {
1397 case MRT_INIT:
1398 if (optlen != sizeof(int)) {
1399 ret = -EINVAL;
1400 break;
1401 }
1402 if (rtnl_dereference(mrt->mroute_sk)) {
1403 ret = -EADDRINUSE;
1404 break;
1405 }
1406
1407 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1408 if (ret == 0) {
1409 rcu_assign_pointer(mrt->mroute_sk, sk);
1410 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1411 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1412 NETCONFA_MC_FORWARDING,
1413 NETCONFA_IFINDEX_ALL,
1414 net->ipv4.devconf_all);
1415 }
1416 break;
1417 case MRT_DONE:
1418 if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1419 ret = -EACCES;
1420 } else {
1421 /* We need to unlock here because mrtsock_destruct takes
1422 * care of rtnl itself and we can't change that due to
1423 * the IP_ROUTER_ALERT setsockopt which runs without it.
1424 */
1425 rtnl_unlock();
1426 ret = ip_ra_control(sk, 0, NULL);
1427 goto out;
1428 }
1429 break;
1430 case MRT_ADD_VIF:
1431 case MRT_DEL_VIF:
1432 if (optlen != sizeof(vif)) {
1433 ret = -EINVAL;
1434 break;
1435 }
1436 if (copy_from_user(&vif, optval, sizeof(vif))) {
1437 ret = -EFAULT;
1438 break;
1439 }
1440 if (vif.vifc_vifi >= MAXVIFS) {
1441 ret = -ENFILE;
1442 break;
1443 }
1444 if (optname == MRT_ADD_VIF) {
1445 ret = vif_add(net, mrt, &vif,
1446 sk == rtnl_dereference(mrt->mroute_sk));
1447 } else {
1448 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1449 }
1450 break;
1451 /* Manipulate the forwarding caches. These live
1452 * in a sort of kernel/user symbiosis.
1453 */
1454 case MRT_ADD_MFC:
1455 case MRT_DEL_MFC:
1456 parent = -1;
1457 /* fall through */
1458 case MRT_ADD_MFC_PROXY:
1459 case MRT_DEL_MFC_PROXY:
1460 if (optlen != sizeof(mfc)) {
1461 ret = -EINVAL;
1462 break;
1463 }
1464 if (copy_from_user(&mfc, optval, sizeof(mfc))) {
1465 ret = -EFAULT;
1466 break;
1467 }
1468 if (parent == 0)
1469 parent = mfc.mfcc_parent;
1470 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1471 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1472 else
1473 ret = ipmr_mfc_add(net, mrt, &mfc,
1474 sk == rtnl_dereference(mrt->mroute_sk),
1475 parent);
1476 break;
1477 /* Control PIM assert. */
1478 case MRT_ASSERT:
1479 if (optlen != sizeof(val)) {
1480 ret = -EINVAL;
1481 break;
1482 }
1483 if (get_user(val, (int __user *)optval)) {
1484 ret = -EFAULT;
1485 break;
1486 }
1487 mrt->mroute_do_assert = val;
1488 break;
1489 case MRT_PIM:
1490 if (!ipmr_pimsm_enabled()) {
1491 ret = -ENOPROTOOPT;
1492 break;
1493 }
1494 if (optlen != sizeof(val)) {
1495 ret = -EINVAL;
1496 break;
1497 }
1498 if (get_user(val, (int __user *)optval)) {
1499 ret = -EFAULT;
1500 break;
1501 }
1502
1503 val = !!val;
1504 if (val != mrt->mroute_do_pim) {
1505 mrt->mroute_do_pim = val;
1506 mrt->mroute_do_assert = val;
1507 }
1508 break;
1509 case MRT_TABLE:
1510 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1511 ret = -ENOPROTOOPT;
1512 break;
1513 }
1514 if (optlen != sizeof(uval)) {
1515 ret = -EINVAL;
1516 break;
1517 }
1518 if (get_user(uval, (u32 __user *)optval)) {
1519 ret = -EFAULT;
1520 break;
1521 }
1522
1523 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1524 ret = -EBUSY;
1525 } else {
1526 mrt = ipmr_new_table(net, uval);
1527 if (IS_ERR(mrt))
1528 ret = PTR_ERR(mrt);
1529 else
1530 raw_sk(sk)->ipmr_table = uval;
1531 }
1532 break;
1533 /* Spurious command, or MRT_VERSION which you cannot set. */
1534 default:
1535 ret = -ENOPROTOOPT;
1536 }
1537out_unlock:
1538 rtnl_unlock();
1539out:
1540 return ret;
1541}
1542
1543/* Getsock opt support for the multicast routing system. */
1544int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1545{
1546 int olr;
1547 int val;
1548 struct net *net = sock_net(sk);
1549 struct mr_table *mrt;
1550
1551 if (sk->sk_type != SOCK_RAW ||
1552 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1553 return -EOPNOTSUPP;
1554
1555 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1556 if (!mrt)
1557 return -ENOENT;
1558
1559 switch (optname) {
1560 case MRT_VERSION:
1561 val = 0x0305;
1562 break;
1563 case MRT_PIM:
1564 if (!ipmr_pimsm_enabled())
1565 return -ENOPROTOOPT;
1566 val = mrt->mroute_do_pim;
1567 break;
1568 case MRT_ASSERT:
1569 val = mrt->mroute_do_assert;
1570 break;
1571 default:
1572 return -ENOPROTOOPT;
1573 }
1574
1575 if (get_user(olr, optlen))
1576 return -EFAULT;
1577 olr = min_t(unsigned int, olr, sizeof(int));
1578 if (olr < 0)
1579 return -EINVAL;
1580 if (put_user(olr, optlen))
1581 return -EFAULT;
1582 if (copy_to_user(optval, &val, olr))
1583 return -EFAULT;
1584 return 0;
1585}
1586
1587/* The IP multicast ioctl support routines. */
1588int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1589{
1590 struct sioc_sg_req sr;
1591 struct sioc_vif_req vr;
1592 struct vif_device *vif;
1593 struct mfc_cache *c;
1594 struct net *net = sock_net(sk);
1595 struct mr_table *mrt;
1596
1597 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1598 if (!mrt)
1599 return -ENOENT;
1600
1601 switch (cmd) {
1602 case SIOCGETVIFCNT:
1603 if (copy_from_user(&vr, arg, sizeof(vr)))
1604 return -EFAULT;
1605 if (vr.vifi >= mrt->maxvif)
1606 return -EINVAL;
1607 read_lock(&mrt_lock);
1608 vif = &mrt->vif_table[vr.vifi];
1609 if (VIF_EXISTS(mrt, vr.vifi)) {
1610 vr.icount = vif->pkt_in;
1611 vr.ocount = vif->pkt_out;
1612 vr.ibytes = vif->bytes_in;
1613 vr.obytes = vif->bytes_out;
1614 read_unlock(&mrt_lock);
1615
1616 if (copy_to_user(arg, &vr, sizeof(vr)))
1617 return -EFAULT;
1618 return 0;
1619 }
1620 read_unlock(&mrt_lock);
1621 return -EADDRNOTAVAIL;
1622 case SIOCGETSGCNT:
1623 if (copy_from_user(&sr, arg, sizeof(sr)))
1624 return -EFAULT;
1625
1626 rcu_read_lock();
1627 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1628 if (c) {
1629 sr.pktcnt = c->_c.mfc_un.res.pkt;
1630 sr.bytecnt = c->_c.mfc_un.res.bytes;
1631 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1632 rcu_read_unlock();
1633
1634 if (copy_to_user(arg, &sr, sizeof(sr)))
1635 return -EFAULT;
1636 return 0;
1637 }
1638 rcu_read_unlock();
1639 return -EADDRNOTAVAIL;
1640 default:
1641 return -ENOIOCTLCMD;
1642 }
1643}
1644
1645#ifdef CONFIG_COMPAT
1646struct compat_sioc_sg_req {
1647 struct in_addr src;
1648 struct in_addr grp;
1649 compat_ulong_t pktcnt;
1650 compat_ulong_t bytecnt;
1651 compat_ulong_t wrong_if;
1652};
1653
1654struct compat_sioc_vif_req {
1655 vifi_t vifi; /* Which iface */
1656 compat_ulong_t icount;
1657 compat_ulong_t ocount;
1658 compat_ulong_t ibytes;
1659 compat_ulong_t obytes;
1660};
1661
1662int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1663{
1664 struct compat_sioc_sg_req sr;
1665 struct compat_sioc_vif_req vr;
1666 struct vif_device *vif;
1667 struct mfc_cache *c;
1668 struct net *net = sock_net(sk);
1669 struct mr_table *mrt;
1670
1671 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1672 if (!mrt)
1673 return -ENOENT;
1674
1675 switch (cmd) {
1676 case SIOCGETVIFCNT:
1677 if (copy_from_user(&vr, arg, sizeof(vr)))
1678 return -EFAULT;
1679 if (vr.vifi >= mrt->maxvif)
1680 return -EINVAL;
1681 read_lock(&mrt_lock);
1682 vif = &mrt->vif_table[vr.vifi];
1683 if (VIF_EXISTS(mrt, vr.vifi)) {
1684 vr.icount = vif->pkt_in;
1685 vr.ocount = vif->pkt_out;
1686 vr.ibytes = vif->bytes_in;
1687 vr.obytes = vif->bytes_out;
1688 read_unlock(&mrt_lock);
1689
1690 if (copy_to_user(arg, &vr, sizeof(vr)))
1691 return -EFAULT;
1692 return 0;
1693 }
1694 read_unlock(&mrt_lock);
1695 return -EADDRNOTAVAIL;
1696 case SIOCGETSGCNT:
1697 if (copy_from_user(&sr, arg, sizeof(sr)))
1698 return -EFAULT;
1699
1700 rcu_read_lock();
1701 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1702 if (c) {
1703 sr.pktcnt = c->_c.mfc_un.res.pkt;
1704 sr.bytecnt = c->_c.mfc_un.res.bytes;
1705 sr.wrong_if = c->_c.mfc_un.res.wrong_if;
1706 rcu_read_unlock();
1707
1708 if (copy_to_user(arg, &sr, sizeof(sr)))
1709 return -EFAULT;
1710 return 0;
1711 }
1712 rcu_read_unlock();
1713 return -EADDRNOTAVAIL;
1714 default:
1715 return -ENOIOCTLCMD;
1716 }
1717}
1718#endif
1719
1720static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1721{
1722 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1723 struct net *net = dev_net(dev);
1724 struct mr_table *mrt;
1725 struct vif_device *v;
1726 int ct;
1727
1728 if (event != NETDEV_UNREGISTER)
1729 return NOTIFY_DONE;
1730
1731 ipmr_for_each_table(mrt, net) {
1732 v = &mrt->vif_table[0];
1733 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1734 if (v->dev == dev)
1735 vif_delete(mrt, ct, 1, NULL);
1736 }
1737 }
1738 return NOTIFY_DONE;
1739}
1740
1741static struct notifier_block ip_mr_notifier = {
1742 .notifier_call = ipmr_device_event,
1743};
1744
1745/* Encapsulate a packet by attaching a valid IPIP header to it.
1746 * This avoids tunnel drivers and other mess and gives us the speed so
1747 * important for multicast video.
1748 */
1749static void ip_encap(struct net *net, struct sk_buff *skb,
1750 __be32 saddr, __be32 daddr)
1751{
1752 struct iphdr *iph;
1753 const struct iphdr *old_iph = ip_hdr(skb);
1754
1755 skb_push(skb, sizeof(struct iphdr));
1756 skb->transport_header = skb->network_header;
1757 skb_reset_network_header(skb);
1758 iph = ip_hdr(skb);
1759
1760 iph->version = 4;
1761 iph->tos = old_iph->tos;
1762 iph->ttl = old_iph->ttl;
1763 iph->frag_off = 0;
1764 iph->daddr = daddr;
1765 iph->saddr = saddr;
1766 iph->protocol = IPPROTO_IPIP;
1767 iph->ihl = 5;
1768 iph->tot_len = htons(skb->len);
1769 ip_select_ident(net, skb, NULL);
1770 ip_send_check(iph);
1771
1772 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1773 nf_reset(skb);
1774}
1775
1776static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1777 struct sk_buff *skb)
1778{
1779 struct ip_options *opt = &(IPCB(skb)->opt);
1780
1781 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1782 IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1783
1784 if (unlikely(opt->optlen))
1785 ip_forward_options(skb);
1786
1787 return dst_output(net, sk, skb);
1788}
1789
1790#ifdef CONFIG_NET_SWITCHDEV
1791static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1792 int in_vifi, int out_vifi)
1793{
1794 struct vif_device *out_vif = &mrt->vif_table[out_vifi];
1795 struct vif_device *in_vif = &mrt->vif_table[in_vifi];
1796
1797 if (!skb->offload_mr_fwd_mark)
1798 return false;
1799 if (!out_vif->dev_parent_id.id_len || !in_vif->dev_parent_id.id_len)
1800 return false;
1801 return netdev_phys_item_id_same(&out_vif->dev_parent_id,
1802 &in_vif->dev_parent_id);
1803}
1804#else
1805static bool ipmr_forward_offloaded(struct sk_buff *skb, struct mr_table *mrt,
1806 int in_vifi, int out_vifi)
1807{
1808 return false;
1809}
1810#endif
1811
1812/* Processing handlers for ipmr_forward */
1813
1814static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1815 int in_vifi, struct sk_buff *skb,
1816 struct mfc_cache *c, int vifi)
1817{
1818 const struct iphdr *iph = ip_hdr(skb);
1819 struct vif_device *vif = &mrt->vif_table[vifi];
1820 struct net_device *dev;
1821 struct rtable *rt;
1822 struct flowi4 fl4;
1823 int encap = 0;
1824
1825 if (!vif->dev)
1826 goto out_free;
1827
1828 if (vif->flags & VIFF_REGISTER) {
1829 vif->pkt_out++;
1830 vif->bytes_out += skb->len;
1831 vif->dev->stats.tx_bytes += skb->len;
1832 vif->dev->stats.tx_packets++;
1833 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1834 goto out_free;
1835 }
1836
1837 if (ipmr_forward_offloaded(skb, mrt, in_vifi, vifi))
1838 goto out_free;
1839
1840 if (vif->flags & VIFF_TUNNEL) {
1841 rt = ip_route_output_ports(net, &fl4, NULL,
1842 vif->remote, vif->local,
1843 0, 0,
1844 IPPROTO_IPIP,
1845 RT_TOS(iph->tos), vif->link);
1846 if (IS_ERR(rt))
1847 goto out_free;
1848 encap = sizeof(struct iphdr);
1849 } else {
1850 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1851 0, 0,
1852 IPPROTO_IPIP,
1853 RT_TOS(iph->tos), vif->link);
1854 if (IS_ERR(rt))
1855 goto out_free;
1856 }
1857
1858 dev = rt->dst.dev;
1859
1860 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1861 /* Do not fragment multicasts. Alas, IPv4 does not
1862 * allow to send ICMP, so that packets will disappear
1863 * to blackhole.
1864 */
1865 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1866 ip_rt_put(rt);
1867 goto out_free;
1868 }
1869
1870 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1871
1872 if (skb_cow(skb, encap)) {
1873 ip_rt_put(rt);
1874 goto out_free;
1875 }
1876
1877 vif->pkt_out++;
1878 vif->bytes_out += skb->len;
1879
1880 skb_dst_drop(skb);
1881 skb_dst_set(skb, &rt->dst);
1882 ip_decrease_ttl(ip_hdr(skb));
1883
1884 /* FIXME: forward and output firewalls used to be called here.
1885 * What do we do with netfilter? -- RR
1886 */
1887 if (vif->flags & VIFF_TUNNEL) {
1888 ip_encap(net, skb, vif->local, vif->remote);
1889 /* FIXME: extra output firewall step used to be here. --RR */
1890 vif->dev->stats.tx_packets++;
1891 vif->dev->stats.tx_bytes += skb->len;
1892 }
1893
1894 IPCB(skb)->flags |= IPSKB_FORWARDED;
1895
1896 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1897 * not only before forwarding, but after forwarding on all output
1898 * interfaces. It is clear, if mrouter runs a multicasting
1899 * program, it should receive packets not depending to what interface
1900 * program is joined.
1901 * If we will not make it, the program will have to join on all
1902 * interfaces. On the other hand, multihoming host (or router, but
1903 * not mrouter) cannot join to more than one interface - it will
1904 * result in receiving multiple packets.
1905 */
1906 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1907 net, NULL, skb, skb->dev, dev,
1908 ipmr_forward_finish);
1909 return;
1910
1911out_free:
1912 kfree_skb(skb);
1913}
1914
1915static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1916{
1917 int ct;
1918
1919 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1920 if (mrt->vif_table[ct].dev == dev)
1921 break;
1922 }
1923 return ct;
1924}
1925
1926/* "local" means that we should preserve one skb (for local delivery) */
1927static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1928 struct net_device *dev, struct sk_buff *skb,
1929 struct mfc_cache *c, int local)
1930{
1931 int true_vifi = ipmr_find_vif(mrt, dev);
1932 int psend = -1;
1933 int vif, ct;
1934
1935 vif = c->_c.mfc_parent;
1936 c->_c.mfc_un.res.pkt++;
1937 c->_c.mfc_un.res.bytes += skb->len;
1938 c->_c.mfc_un.res.lastuse = jiffies;
1939
1940 if (c->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1941 struct mfc_cache *cache_proxy;
1942
1943 /* For an (*,G) entry, we only check that the incomming
1944 * interface is part of the static tree.
1945 */
1946 cache_proxy = mr_mfc_find_any_parent(mrt, vif);
1947 if (cache_proxy &&
1948 cache_proxy->_c.mfc_un.res.ttls[true_vifi] < 255)
1949 goto forward;
1950 }
1951
1952 /* Wrong interface: drop packet and (maybe) send PIM assert. */
1953 if (mrt->vif_table[vif].dev != dev) {
1954 if (rt_is_output_route(skb_rtable(skb))) {
1955 /* It is our own packet, looped back.
1956 * Very complicated situation...
1957 *
1958 * The best workaround until routing daemons will be
1959 * fixed is not to redistribute packet, if it was
1960 * send through wrong interface. It means, that
1961 * multicast applications WILL NOT work for
1962 * (S,G), which have default multicast route pointing
1963 * to wrong oif. In any case, it is not a good
1964 * idea to use multicasting applications on router.
1965 */
1966 goto dont_forward;
1967 }
1968
1969 c->_c.mfc_un.res.wrong_if++;
1970
1971 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1972 /* pimsm uses asserts, when switching from RPT to SPT,
1973 * so that we cannot check that packet arrived on an oif.
1974 * It is bad, but otherwise we would need to move pretty
1975 * large chunk of pimd to kernel. Ough... --ANK
1976 */
1977 (mrt->mroute_do_pim ||
1978 c->_c.mfc_un.res.ttls[true_vifi] < 255) &&
1979 time_after(jiffies,
1980 c->_c.mfc_un.res.last_assert +
1981 MFC_ASSERT_THRESH)) {
1982 c->_c.mfc_un.res.last_assert = jiffies;
1983 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1984 }
1985 goto dont_forward;
1986 }
1987
1988forward:
1989 mrt->vif_table[vif].pkt_in++;
1990 mrt->vif_table[vif].bytes_in += skb->len;
1991
1992 /* Forward the frame */
1993 if (c->mfc_origin == htonl(INADDR_ANY) &&
1994 c->mfc_mcastgrp == htonl(INADDR_ANY)) {
1995 if (true_vifi >= 0 &&
1996 true_vifi != c->_c.mfc_parent &&
1997 ip_hdr(skb)->ttl >
1998 c->_c.mfc_un.res.ttls[c->_c.mfc_parent]) {
1999 /* It's an (*,*) entry and the packet is not coming from
2000 * the upstream: forward the packet to the upstream
2001 * only.
2002 */
2003 psend = c->_c.mfc_parent;
2004 goto last_forward;
2005 }
2006 goto dont_forward;
2007 }
2008 for (ct = c->_c.mfc_un.res.maxvif - 1;
2009 ct >= c->_c.mfc_un.res.minvif; ct--) {
2010 /* For (*,G) entry, don't forward to the incoming interface */
2011 if ((c->mfc_origin != htonl(INADDR_ANY) ||
2012 ct != true_vifi) &&
2013 ip_hdr(skb)->ttl > c->_c.mfc_un.res.ttls[ct]) {
2014 if (psend != -1) {
2015 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2016
2017 if (skb2)
2018 ipmr_queue_xmit(net, mrt, true_vifi,
2019 skb2, c, psend);
2020 }
2021 psend = ct;
2022 }
2023 }
2024last_forward:
2025 if (psend != -1) {
2026 if (local) {
2027 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2028
2029 if (skb2)
2030 ipmr_queue_xmit(net, mrt, true_vifi, skb2,
2031 c, psend);
2032 } else {
2033 ipmr_queue_xmit(net, mrt, true_vifi, skb, c, psend);
2034 return;
2035 }
2036 }
2037
2038dont_forward:
2039 if (!local)
2040 kfree_skb(skb);
2041}
2042
2043static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
2044{
2045 struct rtable *rt = skb_rtable(skb);
2046 struct iphdr *iph = ip_hdr(skb);
2047 struct flowi4 fl4 = {
2048 .daddr = iph->daddr,
2049 .saddr = iph->saddr,
2050 .flowi4_tos = RT_TOS(iph->tos),
2051 .flowi4_oif = (rt_is_output_route(rt) ?
2052 skb->dev->ifindex : 0),
2053 .flowi4_iif = (rt_is_output_route(rt) ?
2054 LOOPBACK_IFINDEX :
2055 skb->dev->ifindex),
2056 .flowi4_mark = skb->mark,
2057 };
2058 struct mr_table *mrt;
2059 int err;
2060
2061 err = ipmr_fib_lookup(net, &fl4, &mrt);
2062 if (err)
2063 return ERR_PTR(err);
2064 return mrt;
2065}
2066
2067/* Multicast packets for forwarding arrive here
2068 * Called with rcu_read_lock();
2069 */
2070int ip_mr_input(struct sk_buff *skb)
2071{
2072 struct mfc_cache *cache;
2073 struct net *net = dev_net(skb->dev);
2074 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
2075 struct mr_table *mrt;
2076 struct net_device *dev;
2077
2078 /* skb->dev passed in is the loX master dev for vrfs.
2079 * As there are no vifs associated with loopback devices,
2080 * get the proper interface that does have a vif associated with it.
2081 */
2082 dev = skb->dev;
2083 if (netif_is_l3_master(skb->dev)) {
2084 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
2085 if (!dev) {
2086 kfree_skb(skb);
2087 return -ENODEV;
2088 }
2089 }
2090
2091 /* Packet is looped back after forward, it should not be
2092 * forwarded second time, but still can be delivered locally.
2093 */
2094 if (IPCB(skb)->flags & IPSKB_FORWARDED)
2095 goto dont_forward;
2096
2097 mrt = ipmr_rt_fib_lookup(net, skb);
2098 if (IS_ERR(mrt)) {
2099 kfree_skb(skb);
2100 return PTR_ERR(mrt);
2101 }
2102 if (!local) {
2103 if (IPCB(skb)->opt.router_alert) {
2104 if (ip_call_ra_chain(skb))
2105 return 0;
2106 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2107 /* IGMPv1 (and broken IGMPv2 implementations sort of
2108 * Cisco IOS <= 11.2(8)) do not put router alert
2109 * option to IGMP packets destined to routable
2110 * groups. It is very bad, because it means
2111 * that we can forward NO IGMP messages.
2112 */
2113 struct sock *mroute_sk;
2114
2115 mroute_sk = rcu_dereference(mrt->mroute_sk);
2116 if (mroute_sk) {
2117 nf_reset(skb);
2118 raw_rcv(mroute_sk, skb);
2119 return 0;
2120 }
2121 }
2122 }
2123
2124 /* already under rcu_read_lock() */
2125 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2126 if (!cache) {
2127 int vif = ipmr_find_vif(mrt, dev);
2128
2129 if (vif >= 0)
2130 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2131 vif);
2132 }
2133
2134 /* No usable cache entry */
2135 if (!cache) {
2136 int vif;
2137
2138 if (local) {
2139 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2140 ip_local_deliver(skb);
2141 if (!skb2)
2142 return -ENOBUFS;
2143 skb = skb2;
2144 }
2145
2146 read_lock(&mrt_lock);
2147 vif = ipmr_find_vif(mrt, dev);
2148 if (vif >= 0) {
2149 int err2 = ipmr_cache_unresolved(mrt, vif, skb, dev);
2150 read_unlock(&mrt_lock);
2151
2152 return err2;
2153 }
2154 read_unlock(&mrt_lock);
2155 kfree_skb(skb);
2156 return -ENODEV;
2157 }
2158
2159 read_lock(&mrt_lock);
2160 ip_mr_forward(net, mrt, dev, skb, cache, local);
2161 read_unlock(&mrt_lock);
2162
2163 if (local)
2164 return ip_local_deliver(skb);
2165
2166 return 0;
2167
2168dont_forward:
2169 if (local)
2170 return ip_local_deliver(skb);
2171 kfree_skb(skb);
2172 return 0;
2173}
2174
2175#ifdef CONFIG_IP_PIMSM_V1
2176/* Handle IGMP messages of PIMv1 */
2177int pim_rcv_v1(struct sk_buff *skb)
2178{
2179 struct igmphdr *pim;
2180 struct net *net = dev_net(skb->dev);
2181 struct mr_table *mrt;
2182
2183 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2184 goto drop;
2185
2186 pim = igmp_hdr(skb);
2187
2188 mrt = ipmr_rt_fib_lookup(net, skb);
2189 if (IS_ERR(mrt))
2190 goto drop;
2191 if (!mrt->mroute_do_pim ||
2192 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2193 goto drop;
2194
2195 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2196drop:
2197 kfree_skb(skb);
2198 }
2199 return 0;
2200}
2201#endif
2202
2203#ifdef CONFIG_IP_PIMSM_V2
2204static int pim_rcv(struct sk_buff *skb)
2205{
2206 struct pimreghdr *pim;
2207 struct net *net = dev_net(skb->dev);
2208 struct mr_table *mrt;
2209
2210 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2211 goto drop;
2212
2213 pim = (struct pimreghdr *)skb_transport_header(skb);
2214 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2215 (pim->flags & PIM_NULL_REGISTER) ||
2216 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2217 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2218 goto drop;
2219
2220 mrt = ipmr_rt_fib_lookup(net, skb);
2221 if (IS_ERR(mrt))
2222 goto drop;
2223 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2224drop:
2225 kfree_skb(skb);
2226 }
2227 return 0;
2228}
2229#endif
2230
2231int ipmr_get_route(struct net *net, struct sk_buff *skb,
2232 __be32 saddr, __be32 daddr,
2233 struct rtmsg *rtm, u32 portid)
2234{
2235 struct mfc_cache *cache;
2236 struct mr_table *mrt;
2237 int err;
2238
2239 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2240 if (!mrt)
2241 return -ENOENT;
2242
2243 rcu_read_lock();
2244 cache = ipmr_cache_find(mrt, saddr, daddr);
2245 if (!cache && skb->dev) {
2246 int vif = ipmr_find_vif(mrt, skb->dev);
2247
2248 if (vif >= 0)
2249 cache = ipmr_cache_find_any(mrt, daddr, vif);
2250 }
2251 if (!cache) {
2252 struct sk_buff *skb2;
2253 struct iphdr *iph;
2254 struct net_device *dev;
2255 int vif = -1;
2256
2257 dev = skb->dev;
2258 read_lock(&mrt_lock);
2259 if (dev)
2260 vif = ipmr_find_vif(mrt, dev);
2261 if (vif < 0) {
2262 read_unlock(&mrt_lock);
2263 rcu_read_unlock();
2264 return -ENODEV;
2265 }
2266 skb2 = skb_clone(skb, GFP_ATOMIC);
2267 if (!skb2) {
2268 read_unlock(&mrt_lock);
2269 rcu_read_unlock();
2270 return -ENOMEM;
2271 }
2272
2273 NETLINK_CB(skb2).portid = portid;
2274 skb_push(skb2, sizeof(struct iphdr));
2275 skb_reset_network_header(skb2);
2276 iph = ip_hdr(skb2);
2277 iph->ihl = sizeof(struct iphdr) >> 2;
2278 iph->saddr = saddr;
2279 iph->daddr = daddr;
2280 iph->version = 0;
2281 err = ipmr_cache_unresolved(mrt, vif, skb2, dev);
2282 read_unlock(&mrt_lock);
2283 rcu_read_unlock();
2284 return err;
2285 }
2286
2287 read_lock(&mrt_lock);
2288 err = mr_fill_mroute(mrt, skb, &cache->_c, rtm);
2289 read_unlock(&mrt_lock);
2290 rcu_read_unlock();
2291 return err;
2292}
2293
2294static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2295 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2296 int flags)
2297{
2298 struct nlmsghdr *nlh;
2299 struct rtmsg *rtm;
2300 int err;
2301
2302 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2303 if (!nlh)
2304 return -EMSGSIZE;
2305
2306 rtm = nlmsg_data(nlh);
2307 rtm->rtm_family = RTNL_FAMILY_IPMR;
2308 rtm->rtm_dst_len = 32;
2309 rtm->rtm_src_len = 32;
2310 rtm->rtm_tos = 0;
2311 rtm->rtm_table = mrt->id;
2312 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2313 goto nla_put_failure;
2314 rtm->rtm_type = RTN_MULTICAST;
2315 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2316 if (c->_c.mfc_flags & MFC_STATIC)
2317 rtm->rtm_protocol = RTPROT_STATIC;
2318 else
2319 rtm->rtm_protocol = RTPROT_MROUTED;
2320 rtm->rtm_flags = 0;
2321
2322 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2323 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2324 goto nla_put_failure;
2325 err = mr_fill_mroute(mrt, skb, &c->_c, rtm);
2326 /* do not break the dump if cache is unresolved */
2327 if (err < 0 && err != -ENOENT)
2328 goto nla_put_failure;
2329
2330 nlmsg_end(skb, nlh);
2331 return 0;
2332
2333nla_put_failure:
2334 nlmsg_cancel(skb, nlh);
2335 return -EMSGSIZE;
2336}
2337
2338static int _ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2339 u32 portid, u32 seq, struct mr_mfc *c, int cmd,
2340 int flags)
2341{
2342 return ipmr_fill_mroute(mrt, skb, portid, seq, (struct mfc_cache *)c,
2343 cmd, flags);
2344}
2345
2346static size_t mroute_msgsize(bool unresolved, int maxvif)
2347{
2348 size_t len =
2349 NLMSG_ALIGN(sizeof(struct rtmsg))
2350 + nla_total_size(4) /* RTA_TABLE */
2351 + nla_total_size(4) /* RTA_SRC */
2352 + nla_total_size(4) /* RTA_DST */
2353 ;
2354
2355 if (!unresolved)
2356 len = len
2357 + nla_total_size(4) /* RTA_IIF */
2358 + nla_total_size(0) /* RTA_MULTIPATH */
2359 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2360 /* RTA_MFC_STATS */
2361 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2362 ;
2363
2364 return len;
2365}
2366
2367static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2368 int cmd)
2369{
2370 struct net *net = read_pnet(&mrt->net);
2371 struct sk_buff *skb;
2372 int err = -ENOBUFS;
2373
2374 skb = nlmsg_new(mroute_msgsize(mfc->_c.mfc_parent >= MAXVIFS,
2375 mrt->maxvif),
2376 GFP_ATOMIC);
2377 if (!skb)
2378 goto errout;
2379
2380 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2381 if (err < 0)
2382 goto errout;
2383
2384 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2385 return;
2386
2387errout:
2388 kfree_skb(skb);
2389 if (err < 0)
2390 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2391}
2392
2393static size_t igmpmsg_netlink_msgsize(size_t payloadlen)
2394{
2395 size_t len =
2396 NLMSG_ALIGN(sizeof(struct rtgenmsg))
2397 + nla_total_size(1) /* IPMRA_CREPORT_MSGTYPE */
2398 + nla_total_size(4) /* IPMRA_CREPORT_VIF_ID */
2399 + nla_total_size(4) /* IPMRA_CREPORT_SRC_ADDR */
2400 + nla_total_size(4) /* IPMRA_CREPORT_DST_ADDR */
2401 /* IPMRA_CREPORT_PKT */
2402 + nla_total_size(payloadlen)
2403 ;
2404
2405 return len;
2406}
2407
2408static void igmpmsg_netlink_event(struct mr_table *mrt, struct sk_buff *pkt)
2409{
2410 struct net *net = read_pnet(&mrt->net);
2411 struct nlmsghdr *nlh;
2412 struct rtgenmsg *rtgenm;
2413 struct igmpmsg *msg;
2414 struct sk_buff *skb;
2415 struct nlattr *nla;
2416 int payloadlen;
2417
2418 payloadlen = pkt->len - sizeof(struct igmpmsg);
2419 msg = (struct igmpmsg *)skb_network_header(pkt);
2420
2421 skb = nlmsg_new(igmpmsg_netlink_msgsize(payloadlen), GFP_ATOMIC);
2422 if (!skb)
2423 goto errout;
2424
2425 nlh = nlmsg_put(skb, 0, 0, RTM_NEWCACHEREPORT,
2426 sizeof(struct rtgenmsg), 0);
2427 if (!nlh)
2428 goto errout;
2429 rtgenm = nlmsg_data(nlh);
2430 rtgenm->rtgen_family = RTNL_FAMILY_IPMR;
2431 if (nla_put_u8(skb, IPMRA_CREPORT_MSGTYPE, msg->im_msgtype) ||
2432 nla_put_u32(skb, IPMRA_CREPORT_VIF_ID, msg->im_vif) ||
2433 nla_put_in_addr(skb, IPMRA_CREPORT_SRC_ADDR,
2434 msg->im_src.s_addr) ||
2435 nla_put_in_addr(skb, IPMRA_CREPORT_DST_ADDR,
2436 msg->im_dst.s_addr))
2437 goto nla_put_failure;
2438
2439 nla = nla_reserve(skb, IPMRA_CREPORT_PKT, payloadlen);
2440 if (!nla || skb_copy_bits(pkt, sizeof(struct igmpmsg),
2441 nla_data(nla), payloadlen))
2442 goto nla_put_failure;
2443
2444 nlmsg_end(skb, nlh);
2445
2446 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE_R, NULL, GFP_ATOMIC);
2447 return;
2448
2449nla_put_failure:
2450 nlmsg_cancel(skb, nlh);
2451errout:
2452 kfree_skb(skb);
2453 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE_R, -ENOBUFS);
2454}
2455
2456static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2457 struct netlink_ext_ack *extack)
2458{
2459 struct net *net = sock_net(in_skb->sk);
2460 struct nlattr *tb[RTA_MAX + 1];
2461 struct sk_buff *skb = NULL;
2462 struct mfc_cache *cache;
2463 struct mr_table *mrt;
2464 struct rtmsg *rtm;
2465 __be32 src, grp;
2466 u32 tableid;
2467 int err;
2468
2469 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX,
2470 rtm_ipv4_policy, extack);
2471 if (err < 0)
2472 goto errout;
2473
2474 rtm = nlmsg_data(nlh);
2475
2476 src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
2477 grp = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
2478 tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
2479
2480 mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
2481 if (!mrt) {
2482 err = -ENOENT;
2483 goto errout_free;
2484 }
2485
2486 /* entries are added/deleted only under RTNL */
2487 rcu_read_lock();
2488 cache = ipmr_cache_find(mrt, src, grp);
2489 rcu_read_unlock();
2490 if (!cache) {
2491 err = -ENOENT;
2492 goto errout_free;
2493 }
2494
2495 skb = nlmsg_new(mroute_msgsize(false, mrt->maxvif), GFP_KERNEL);
2496 if (!skb) {
2497 err = -ENOBUFS;
2498 goto errout_free;
2499 }
2500
2501 err = ipmr_fill_mroute(mrt, skb, NETLINK_CB(in_skb).portid,
2502 nlh->nlmsg_seq, cache,
2503 RTM_NEWROUTE, 0);
2504 if (err < 0)
2505 goto errout_free;
2506
2507 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2508
2509errout:
2510 return err;
2511
2512errout_free:
2513 kfree_skb(skb);
2514 goto errout;
2515}
2516
2517static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2518{
2519 return mr_rtm_dumproute(skb, cb, ipmr_mr_table_iter,
2520 _ipmr_fill_mroute, &mfc_unres_lock);
2521}
2522
2523static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2524 [RTA_SRC] = { .type = NLA_U32 },
2525 [RTA_DST] = { .type = NLA_U32 },
2526 [RTA_IIF] = { .type = NLA_U32 },
2527 [RTA_TABLE] = { .type = NLA_U32 },
2528 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2529};
2530
2531static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2532{
2533 switch (rtm_protocol) {
2534 case RTPROT_STATIC:
2535 case RTPROT_MROUTED:
2536 return true;
2537 }
2538 return false;
2539}
2540
2541static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2542{
2543 struct rtnexthop *rtnh = nla_data(nla);
2544 int remaining = nla_len(nla), vifi = 0;
2545
2546 while (rtnh_ok(rtnh, remaining)) {
2547 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2548 if (++vifi == MAXVIFS)
2549 break;
2550 rtnh = rtnh_next(rtnh, &remaining);
2551 }
2552
2553 return remaining > 0 ? -EINVAL : vifi;
2554}
2555
2556/* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2557static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2558 struct mfcctl *mfcc, int *mrtsock,
2559 struct mr_table **mrtret,
2560 struct netlink_ext_ack *extack)
2561{
2562 struct net_device *dev = NULL;
2563 u32 tblid = RT_TABLE_DEFAULT;
2564 struct mr_table *mrt;
2565 struct nlattr *attr;
2566 struct rtmsg *rtm;
2567 int ret, rem;
2568
2569 ret = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipmr_policy,
2570 extack);
2571 if (ret < 0)
2572 goto out;
2573 rtm = nlmsg_data(nlh);
2574
2575 ret = -EINVAL;
2576 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2577 rtm->rtm_type != RTN_MULTICAST ||
2578 rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2579 !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2580 goto out;
2581
2582 memset(mfcc, 0, sizeof(*mfcc));
2583 mfcc->mfcc_parent = -1;
2584 ret = 0;
2585 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2586 switch (nla_type(attr)) {
2587 case RTA_SRC:
2588 mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2589 break;
2590 case RTA_DST:
2591 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2592 break;
2593 case RTA_IIF:
2594 dev = __dev_get_by_index(net, nla_get_u32(attr));
2595 if (!dev) {
2596 ret = -ENODEV;
2597 goto out;
2598 }
2599 break;
2600 case RTA_MULTIPATH:
2601 if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2602 ret = -EINVAL;
2603 goto out;
2604 }
2605 break;
2606 case RTA_PREFSRC:
2607 ret = 1;
2608 break;
2609 case RTA_TABLE:
2610 tblid = nla_get_u32(attr);
2611 break;
2612 }
2613 }
2614 mrt = ipmr_get_table(net, tblid);
2615 if (!mrt) {
2616 ret = -ENOENT;
2617 goto out;
2618 }
2619 *mrtret = mrt;
2620 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2621 if (dev)
2622 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2623
2624out:
2625 return ret;
2626}
2627
2628/* takes care of both newroute and delroute */
2629static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2630 struct netlink_ext_ack *extack)
2631{
2632 struct net *net = sock_net(skb->sk);
2633 int ret, mrtsock, parent;
2634 struct mr_table *tbl;
2635 struct mfcctl mfcc;
2636
2637 mrtsock = 0;
2638 tbl = NULL;
2639 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2640 if (ret < 0)
2641 return ret;
2642
2643 parent = ret ? mfcc.mfcc_parent : -1;
2644 if (nlh->nlmsg_type == RTM_NEWROUTE)
2645 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2646 else
2647 return ipmr_mfc_delete(tbl, &mfcc, parent);
2648}
2649
2650static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
2651{
2652 u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
2653
2654 if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
2655 nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
2656 nla_put_s32(skb, IPMRA_TABLE_MROUTE_REG_VIF_NUM,
2657 mrt->mroute_reg_vif_num) ||
2658 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_ASSERT,
2659 mrt->mroute_do_assert) ||
2660 nla_put_u8(skb, IPMRA_TABLE_MROUTE_DO_PIM, mrt->mroute_do_pim))
2661 return false;
2662
2663 return true;
2664}
2665
2666static bool ipmr_fill_vif(struct mr_table *mrt, u32 vifid, struct sk_buff *skb)
2667{
2668 struct nlattr *vif_nest;
2669 struct vif_device *vif;
2670
2671 /* if the VIF doesn't exist just continue */
2672 if (!VIF_EXISTS(mrt, vifid))
2673 return true;
2674
2675 vif = &mrt->vif_table[vifid];
2676 vif_nest = nla_nest_start(skb, IPMRA_VIF);
2677 if (!vif_nest)
2678 return false;
2679 if (nla_put_u32(skb, IPMRA_VIFA_IFINDEX, vif->dev->ifindex) ||
2680 nla_put_u32(skb, IPMRA_VIFA_VIF_ID, vifid) ||
2681 nla_put_u16(skb, IPMRA_VIFA_FLAGS, vif->flags) ||
2682 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_IN, vif->bytes_in,
2683 IPMRA_VIFA_PAD) ||
2684 nla_put_u64_64bit(skb, IPMRA_VIFA_BYTES_OUT, vif->bytes_out,
2685 IPMRA_VIFA_PAD) ||
2686 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_IN, vif->pkt_in,
2687 IPMRA_VIFA_PAD) ||
2688 nla_put_u64_64bit(skb, IPMRA_VIFA_PACKETS_OUT, vif->pkt_out,
2689 IPMRA_VIFA_PAD) ||
2690 nla_put_be32(skb, IPMRA_VIFA_LOCAL_ADDR, vif->local) ||
2691 nla_put_be32(skb, IPMRA_VIFA_REMOTE_ADDR, vif->remote)) {
2692 nla_nest_cancel(skb, vif_nest);
2693 return false;
2694 }
2695 nla_nest_end(skb, vif_nest);
2696
2697 return true;
2698}
2699
2700static int ipmr_rtm_dumplink(struct sk_buff *skb, struct netlink_callback *cb)
2701{
2702 struct net *net = sock_net(skb->sk);
2703 struct nlmsghdr *nlh = NULL;
2704 unsigned int t = 0, s_t;
2705 unsigned int e = 0, s_e;
2706 struct mr_table *mrt;
2707
2708 s_t = cb->args[0];
2709 s_e = cb->args[1];
2710
2711 ipmr_for_each_table(mrt, net) {
2712 struct nlattr *vifs, *af;
2713 struct ifinfomsg *hdr;
2714 u32 i;
2715
2716 if (t < s_t)
2717 goto skip_table;
2718 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
2719 cb->nlh->nlmsg_seq, RTM_NEWLINK,
2720 sizeof(*hdr), NLM_F_MULTI);
2721 if (!nlh)
2722 break;
2723
2724 hdr = nlmsg_data(nlh);
2725 memset(hdr, 0, sizeof(*hdr));
2726 hdr->ifi_family = RTNL_FAMILY_IPMR;
2727
2728 af = nla_nest_start(skb, IFLA_AF_SPEC);
2729 if (!af) {
2730 nlmsg_cancel(skb, nlh);
2731 goto out;
2732 }
2733
2734 if (!ipmr_fill_table(mrt, skb)) {
2735 nlmsg_cancel(skb, nlh);
2736 goto out;
2737 }
2738
2739 vifs = nla_nest_start(skb, IPMRA_TABLE_VIFS);
2740 if (!vifs) {
2741 nla_nest_end(skb, af);
2742 nlmsg_end(skb, nlh);
2743 goto out;
2744 }
2745 for (i = 0; i < mrt->maxvif; i++) {
2746 if (e < s_e)
2747 goto skip_entry;
2748 if (!ipmr_fill_vif(mrt, i, skb)) {
2749 nla_nest_end(skb, vifs);
2750 nla_nest_end(skb, af);
2751 nlmsg_end(skb, nlh);
2752 goto out;
2753 }
2754skip_entry:
2755 e++;
2756 }
2757 s_e = 0;
2758 e = 0;
2759 nla_nest_end(skb, vifs);
2760 nla_nest_end(skb, af);
2761 nlmsg_end(skb, nlh);
2762skip_table:
2763 t++;
2764 }
2765
2766out:
2767 cb->args[1] = e;
2768 cb->args[0] = t;
2769
2770 return skb->len;
2771}
2772
2773#ifdef CONFIG_PROC_FS
2774/* The /proc interfaces to multicast routing :
2775 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2776 */
2777
2778static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2779 __acquires(mrt_lock)
2780{
2781 struct mr_vif_iter *iter = seq->private;
2782 struct net *net = seq_file_net(seq);
2783 struct mr_table *mrt;
2784
2785 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2786 if (!mrt)
2787 return ERR_PTR(-ENOENT);
2788
2789 iter->mrt = mrt;
2790
2791 read_lock(&mrt_lock);
2792 return mr_vif_seq_start(seq, pos);
2793}
2794
2795static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2796 __releases(mrt_lock)
2797{
2798 read_unlock(&mrt_lock);
2799}
2800
2801static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2802{
2803 struct mr_vif_iter *iter = seq->private;
2804 struct mr_table *mrt = iter->mrt;
2805
2806 if (v == SEQ_START_TOKEN) {
2807 seq_puts(seq,
2808 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2809 } else {
2810 const struct vif_device *vif = v;
2811 const char *name = vif->dev ?
2812 vif->dev->name : "none";
2813
2814 seq_printf(seq,
2815 "%2td %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2816 vif - mrt->vif_table,
2817 name, vif->bytes_in, vif->pkt_in,
2818 vif->bytes_out, vif->pkt_out,
2819 vif->flags, vif->local, vif->remote);
2820 }
2821 return 0;
2822}
2823
2824static const struct seq_operations ipmr_vif_seq_ops = {
2825 .start = ipmr_vif_seq_start,
2826 .next = mr_vif_seq_next,
2827 .stop = ipmr_vif_seq_stop,
2828 .show = ipmr_vif_seq_show,
2829};
2830
2831static int ipmr_vif_open(struct inode *inode, struct file *file)
2832{
2833 return seq_open_net(inode, file, &ipmr_vif_seq_ops,
2834 sizeof(struct mr_vif_iter));
2835}
2836
2837static const struct file_operations ipmr_vif_fops = {
2838 .open = ipmr_vif_open,
2839 .read = seq_read,
2840 .llseek = seq_lseek,
2841 .release = seq_release_net,
2842};
2843
2844static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2845{
2846 struct net *net = seq_file_net(seq);
2847 struct mr_table *mrt;
2848
2849 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2850 if (!mrt)
2851 return ERR_PTR(-ENOENT);
2852
2853 return mr_mfc_seq_start(seq, pos, mrt, &mfc_unres_lock);
2854}
2855
2856static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2857{
2858 int n;
2859
2860 if (v == SEQ_START_TOKEN) {
2861 seq_puts(seq,
2862 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2863 } else {
2864 const struct mfc_cache *mfc = v;
2865 const struct mr_mfc_iter *it = seq->private;
2866 const struct mr_table *mrt = it->mrt;
2867
2868 seq_printf(seq, "%08X %08X %-3hd",
2869 (__force u32) mfc->mfc_mcastgrp,
2870 (__force u32) mfc->mfc_origin,
2871 mfc->_c.mfc_parent);
2872
2873 if (it->cache != &mrt->mfc_unres_queue) {
2874 seq_printf(seq, " %8lu %8lu %8lu",
2875 mfc->_c.mfc_un.res.pkt,
2876 mfc->_c.mfc_un.res.bytes,
2877 mfc->_c.mfc_un.res.wrong_if);
2878 for (n = mfc->_c.mfc_un.res.minvif;
2879 n < mfc->_c.mfc_un.res.maxvif; n++) {
2880 if (VIF_EXISTS(mrt, n) &&
2881 mfc->_c.mfc_un.res.ttls[n] < 255)
2882 seq_printf(seq,
2883 " %2d:%-3d",
2884 n, mfc->_c.mfc_un.res.ttls[n]);
2885 }
2886 } else {
2887 /* unresolved mfc_caches don't contain
2888 * pkt, bytes and wrong_if values
2889 */
2890 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2891 }
2892 seq_putc(seq, '\n');
2893 }
2894 return 0;
2895}
2896
2897static const struct seq_operations ipmr_mfc_seq_ops = {
2898 .start = ipmr_mfc_seq_start,
2899 .next = mr_mfc_seq_next,
2900 .stop = mr_mfc_seq_stop,
2901 .show = ipmr_mfc_seq_show,
2902};
2903
2904static int ipmr_mfc_open(struct inode *inode, struct file *file)
2905{
2906 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
2907 sizeof(struct mr_mfc_iter));
2908}
2909
2910static const struct file_operations ipmr_mfc_fops = {
2911 .open = ipmr_mfc_open,
2912 .read = seq_read,
2913 .llseek = seq_lseek,
2914 .release = seq_release_net,
2915};
2916#endif
2917
2918#ifdef CONFIG_IP_PIMSM_V2
2919static const struct net_protocol pim_protocol = {
2920 .handler = pim_rcv,
2921 .netns_ok = 1,
2922};
2923#endif
2924
2925static unsigned int ipmr_seq_read(struct net *net)
2926{
2927 ASSERT_RTNL();
2928
2929 return net->ipv4.ipmr_seq + ipmr_rules_seq_read(net);
2930}
2931
2932static int ipmr_dump(struct net *net, struct notifier_block *nb)
2933{
2934 return mr_dump(net, nb, RTNL_FAMILY_IPMR, ipmr_rules_dump,
2935 ipmr_mr_table_iter, &mrt_lock);
2936}
2937
2938static const struct fib_notifier_ops ipmr_notifier_ops_template = {
2939 .family = RTNL_FAMILY_IPMR,
2940 .fib_seq_read = ipmr_seq_read,
2941 .fib_dump = ipmr_dump,
2942 .owner = THIS_MODULE,
2943};
2944
2945static int __net_init ipmr_notifier_init(struct net *net)
2946{
2947 struct fib_notifier_ops *ops;
2948
2949 net->ipv4.ipmr_seq = 0;
2950
2951 ops = fib_notifier_ops_register(&ipmr_notifier_ops_template, net);
2952 if (IS_ERR(ops))
2953 return PTR_ERR(ops);
2954 net->ipv4.ipmr_notifier_ops = ops;
2955
2956 return 0;
2957}
2958
2959static void __net_exit ipmr_notifier_exit(struct net *net)
2960{
2961 fib_notifier_ops_unregister(net->ipv4.ipmr_notifier_ops);
2962 net->ipv4.ipmr_notifier_ops = NULL;
2963}
2964
2965/* Setup for IP multicast routing */
2966static int __net_init ipmr_net_init(struct net *net)
2967{
2968 int err;
2969
2970 err = ipmr_notifier_init(net);
2971 if (err)
2972 goto ipmr_notifier_fail;
2973
2974 err = ipmr_rules_init(net);
2975 if (err < 0)
2976 goto ipmr_rules_fail;
2977
2978#ifdef CONFIG_PROC_FS
2979 err = -ENOMEM;
2980 if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops))
2981 goto proc_vif_fail;
2982 if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops))
2983 goto proc_cache_fail;
2984#endif
2985 return 0;
2986
2987#ifdef CONFIG_PROC_FS
2988proc_cache_fail:
2989 remove_proc_entry("ip_mr_vif", net->proc_net);
2990proc_vif_fail:
2991 ipmr_rules_exit(net);
2992#endif
2993ipmr_rules_fail:
2994 ipmr_notifier_exit(net);
2995ipmr_notifier_fail:
2996 return err;
2997}
2998
2999static void __net_exit ipmr_net_exit(struct net *net)
3000{
3001#ifdef CONFIG_PROC_FS
3002 remove_proc_entry("ip_mr_cache", net->proc_net);
3003 remove_proc_entry("ip_mr_vif", net->proc_net);
3004#endif
3005 ipmr_notifier_exit(net);
3006 ipmr_rules_exit(net);
3007}
3008
3009static struct pernet_operations ipmr_net_ops = {
3010 .init = ipmr_net_init,
3011 .exit = ipmr_net_exit,
3012};
3013
3014int __init ip_mr_init(void)
3015{
3016 int err;
3017
3018 mrt_cachep = kmem_cache_create("ip_mrt_cache",
3019 sizeof(struct mfc_cache),
3020 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
3021 NULL);
3022
3023 err = register_pernet_subsys(&ipmr_net_ops);
3024 if (err)
3025 goto reg_pernet_fail;
3026
3027 err = register_netdevice_notifier(&ip_mr_notifier);
3028 if (err)
3029 goto reg_notif_fail;
3030#ifdef CONFIG_IP_PIMSM_V2
3031 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
3032 pr_err("%s: can't add PIM protocol\n", __func__);
3033 err = -EAGAIN;
3034 goto add_proto_fail;
3035 }
3036#endif
3037 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
3038 ipmr_rtm_getroute, ipmr_rtm_dumproute, 0);
3039 rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
3040 ipmr_rtm_route, NULL, 0);
3041 rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
3042 ipmr_rtm_route, NULL, 0);
3043
3044 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETLINK,
3045 NULL, ipmr_rtm_dumplink, 0);
3046 return 0;
3047
3048#ifdef CONFIG_IP_PIMSM_V2
3049add_proto_fail:
3050 unregister_netdevice_notifier(&ip_mr_notifier);
3051#endif
3052reg_notif_fail:
3053 unregister_pernet_subsys(&ipmr_net_ops);
3054reg_pernet_fail:
3055 kmem_cache_destroy(mrt_cachep);
3056 return err;
3057}
1/*
2 * IP multicast routing support for mrouted 3.6/3.8
3 *
4 * (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
5 * Linux Consultancy and Custom Driver Development
6 *
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
11 *
12 * Fixes:
13 * Michael Chastain : Incorrect size of copying.
14 * Alan Cox : Added the cache manager code
15 * Alan Cox : Fixed the clone/copy bug and device race.
16 * Mike McLagan : Routing by source
17 * Malcolm Beattie : Buffer handling fixes.
18 * Alexey Kuznetsov : Double buffer free and other fixes.
19 * SVR Anand : Fixed several multicast bugs and problems.
20 * Alexey Kuznetsov : Status, optimisations and more.
21 * Brad Parker : Better behaviour on mrouted upcall
22 * overflow.
23 * Carlos Picoto : PIMv1 Support
24 * Pavlin Ivanov Radoslavov: PIMv2 Registers must checksum only PIM header
25 * Relax this requirement to work with older peers.
26 *
27 */
28
29#include <asm/uaccess.h>
30#include <linux/types.h>
31#include <linux/capability.h>
32#include <linux/errno.h>
33#include <linux/timer.h>
34#include <linux/mm.h>
35#include <linux/kernel.h>
36#include <linux/fcntl.h>
37#include <linux/stat.h>
38#include <linux/socket.h>
39#include <linux/in.h>
40#include <linux/inet.h>
41#include <linux/netdevice.h>
42#include <linux/inetdevice.h>
43#include <linux/igmp.h>
44#include <linux/proc_fs.h>
45#include <linux/seq_file.h>
46#include <linux/mroute.h>
47#include <linux/init.h>
48#include <linux/if_ether.h>
49#include <linux/slab.h>
50#include <net/net_namespace.h>
51#include <net/ip.h>
52#include <net/protocol.h>
53#include <linux/skbuff.h>
54#include <net/route.h>
55#include <net/sock.h>
56#include <net/icmp.h>
57#include <net/udp.h>
58#include <net/raw.h>
59#include <linux/notifier.h>
60#include <linux/if_arp.h>
61#include <linux/netfilter_ipv4.h>
62#include <linux/compat.h>
63#include <linux/export.h>
64#include <net/ip_tunnels.h>
65#include <net/checksum.h>
66#include <net/netlink.h>
67#include <net/fib_rules.h>
68#include <linux/netconf.h>
69
70#if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
71#define CONFIG_IP_PIMSM 1
72#endif
73
74struct mr_table {
75 struct list_head list;
76#ifdef CONFIG_NET_NS
77 struct net *net;
78#endif
79 u32 id;
80 struct sock __rcu *mroute_sk;
81 struct timer_list ipmr_expire_timer;
82 struct list_head mfc_unres_queue;
83 struct list_head mfc_cache_array[MFC_LINES];
84 struct vif_device vif_table[MAXVIFS];
85 int maxvif;
86 atomic_t cache_resolve_queue_len;
87 bool mroute_do_assert;
88 bool mroute_do_pim;
89#if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
90 int mroute_reg_vif_num;
91#endif
92};
93
94struct ipmr_rule {
95 struct fib_rule common;
96};
97
98struct ipmr_result {
99 struct mr_table *mrt;
100};
101
102/* Big lock, protecting vif table, mrt cache and mroute socket state.
103 * Note that the changes are semaphored via rtnl_lock.
104 */
105
106static DEFINE_RWLOCK(mrt_lock);
107
108/*
109 * Multicast router control variables
110 */
111
112#define VIF_EXISTS(_mrt, _idx) ((_mrt)->vif_table[_idx].dev != NULL)
113
114/* Special spinlock for queue of unresolved entries */
115static DEFINE_SPINLOCK(mfc_unres_lock);
116
117/* We return to original Alan's scheme. Hash table of resolved
118 * entries is changed only in process context and protected
119 * with weak lock mrt_lock. Queue of unresolved entries is protected
120 * with strong spinlock mfc_unres_lock.
121 *
122 * In this case data path is free of exclusive locks at all.
123 */
124
125static struct kmem_cache *mrt_cachep __read_mostly;
126
127static struct mr_table *ipmr_new_table(struct net *net, u32 id);
128static void ipmr_free_table(struct mr_table *mrt);
129
130static void ip_mr_forward(struct net *net, struct mr_table *mrt,
131 struct sk_buff *skb, struct mfc_cache *cache,
132 int local);
133static int ipmr_cache_report(struct mr_table *mrt,
134 struct sk_buff *pkt, vifi_t vifi, int assert);
135static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
136 struct mfc_cache *c, struct rtmsg *rtm);
137static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
138 int cmd);
139static void mroute_clean_tables(struct mr_table *mrt);
140static void ipmr_expire_process(unsigned long arg);
141
142#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
143#define ipmr_for_each_table(mrt, net) \
144 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
145
146static struct mr_table *ipmr_get_table(struct net *net, u32 id)
147{
148 struct mr_table *mrt;
149
150 ipmr_for_each_table(mrt, net) {
151 if (mrt->id == id)
152 return mrt;
153 }
154 return NULL;
155}
156
157static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
158 struct mr_table **mrt)
159{
160 int err;
161 struct ipmr_result res;
162 struct fib_lookup_arg arg = {
163 .result = &res,
164 .flags = FIB_LOOKUP_NOREF,
165 };
166
167 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
168 flowi4_to_flowi(flp4), 0, &arg);
169 if (err < 0)
170 return err;
171 *mrt = res.mrt;
172 return 0;
173}
174
175static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
176 int flags, struct fib_lookup_arg *arg)
177{
178 struct ipmr_result *res = arg->result;
179 struct mr_table *mrt;
180
181 switch (rule->action) {
182 case FR_ACT_TO_TBL:
183 break;
184 case FR_ACT_UNREACHABLE:
185 return -ENETUNREACH;
186 case FR_ACT_PROHIBIT:
187 return -EACCES;
188 case FR_ACT_BLACKHOLE:
189 default:
190 return -EINVAL;
191 }
192
193 mrt = ipmr_get_table(rule->fr_net, rule->table);
194 if (mrt == NULL)
195 return -EAGAIN;
196 res->mrt = mrt;
197 return 0;
198}
199
200static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
201{
202 return 1;
203}
204
205static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
206 FRA_GENERIC_POLICY,
207};
208
209static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
210 struct fib_rule_hdr *frh, struct nlattr **tb)
211{
212 return 0;
213}
214
215static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
216 struct nlattr **tb)
217{
218 return 1;
219}
220
221static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
222 struct fib_rule_hdr *frh)
223{
224 frh->dst_len = 0;
225 frh->src_len = 0;
226 frh->tos = 0;
227 return 0;
228}
229
230static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
231 .family = RTNL_FAMILY_IPMR,
232 .rule_size = sizeof(struct ipmr_rule),
233 .addr_size = sizeof(u32),
234 .action = ipmr_rule_action,
235 .match = ipmr_rule_match,
236 .configure = ipmr_rule_configure,
237 .compare = ipmr_rule_compare,
238 .default_pref = fib_default_rule_pref,
239 .fill = ipmr_rule_fill,
240 .nlgroup = RTNLGRP_IPV4_RULE,
241 .policy = ipmr_rule_policy,
242 .owner = THIS_MODULE,
243};
244
245static int __net_init ipmr_rules_init(struct net *net)
246{
247 struct fib_rules_ops *ops;
248 struct mr_table *mrt;
249 int err;
250
251 ops = fib_rules_register(&ipmr_rules_ops_template, net);
252 if (IS_ERR(ops))
253 return PTR_ERR(ops);
254
255 INIT_LIST_HEAD(&net->ipv4.mr_tables);
256
257 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
258 if (mrt == NULL) {
259 err = -ENOMEM;
260 goto err1;
261 }
262
263 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
264 if (err < 0)
265 goto err2;
266
267 net->ipv4.mr_rules_ops = ops;
268 return 0;
269
270err2:
271 kfree(mrt);
272err1:
273 fib_rules_unregister(ops);
274 return err;
275}
276
277static void __net_exit ipmr_rules_exit(struct net *net)
278{
279 struct mr_table *mrt, *next;
280
281 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
282 list_del(&mrt->list);
283 ipmr_free_table(mrt);
284 }
285 fib_rules_unregister(net->ipv4.mr_rules_ops);
286}
287#else
288#define ipmr_for_each_table(mrt, net) \
289 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
290
291static struct mr_table *ipmr_get_table(struct net *net, u32 id)
292{
293 return net->ipv4.mrt;
294}
295
296static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
297 struct mr_table **mrt)
298{
299 *mrt = net->ipv4.mrt;
300 return 0;
301}
302
303static int __net_init ipmr_rules_init(struct net *net)
304{
305 net->ipv4.mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
306 return net->ipv4.mrt ? 0 : -ENOMEM;
307}
308
309static void __net_exit ipmr_rules_exit(struct net *net)
310{
311 ipmr_free_table(net->ipv4.mrt);
312}
313#endif
314
315static struct mr_table *ipmr_new_table(struct net *net, u32 id)
316{
317 struct mr_table *mrt;
318 unsigned int i;
319
320 mrt = ipmr_get_table(net, id);
321 if (mrt != NULL)
322 return mrt;
323
324 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
325 if (mrt == NULL)
326 return NULL;
327 write_pnet(&mrt->net, net);
328 mrt->id = id;
329
330 /* Forwarding cache */
331 for (i = 0; i < MFC_LINES; i++)
332 INIT_LIST_HEAD(&mrt->mfc_cache_array[i]);
333
334 INIT_LIST_HEAD(&mrt->mfc_unres_queue);
335
336 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
337 (unsigned long)mrt);
338
339#ifdef CONFIG_IP_PIMSM
340 mrt->mroute_reg_vif_num = -1;
341#endif
342#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
343 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
344#endif
345 return mrt;
346}
347
348static void ipmr_free_table(struct mr_table *mrt)
349{
350 del_timer_sync(&mrt->ipmr_expire_timer);
351 mroute_clean_tables(mrt);
352 kfree(mrt);
353}
354
355/* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
356
357static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
358{
359 struct net *net = dev_net(dev);
360
361 dev_close(dev);
362
363 dev = __dev_get_by_name(net, "tunl0");
364 if (dev) {
365 const struct net_device_ops *ops = dev->netdev_ops;
366 struct ifreq ifr;
367 struct ip_tunnel_parm p;
368
369 memset(&p, 0, sizeof(p));
370 p.iph.daddr = v->vifc_rmt_addr.s_addr;
371 p.iph.saddr = v->vifc_lcl_addr.s_addr;
372 p.iph.version = 4;
373 p.iph.ihl = 5;
374 p.iph.protocol = IPPROTO_IPIP;
375 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
376 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
377
378 if (ops->ndo_do_ioctl) {
379 mm_segment_t oldfs = get_fs();
380
381 set_fs(KERNEL_DS);
382 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
383 set_fs(oldfs);
384 }
385 }
386}
387
388static
389struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
390{
391 struct net_device *dev;
392
393 dev = __dev_get_by_name(net, "tunl0");
394
395 if (dev) {
396 const struct net_device_ops *ops = dev->netdev_ops;
397 int err;
398 struct ifreq ifr;
399 struct ip_tunnel_parm p;
400 struct in_device *in_dev;
401
402 memset(&p, 0, sizeof(p));
403 p.iph.daddr = v->vifc_rmt_addr.s_addr;
404 p.iph.saddr = v->vifc_lcl_addr.s_addr;
405 p.iph.version = 4;
406 p.iph.ihl = 5;
407 p.iph.protocol = IPPROTO_IPIP;
408 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
409 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
410
411 if (ops->ndo_do_ioctl) {
412 mm_segment_t oldfs = get_fs();
413
414 set_fs(KERNEL_DS);
415 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
416 set_fs(oldfs);
417 } else {
418 err = -EOPNOTSUPP;
419 }
420 dev = NULL;
421
422 if (err == 0 &&
423 (dev = __dev_get_by_name(net, p.name)) != NULL) {
424 dev->flags |= IFF_MULTICAST;
425
426 in_dev = __in_dev_get_rtnl(dev);
427 if (in_dev == NULL)
428 goto failure;
429
430 ipv4_devconf_setall(in_dev);
431 neigh_parms_data_state_setall(in_dev->arp_parms);
432 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
433
434 if (dev_open(dev))
435 goto failure;
436 dev_hold(dev);
437 }
438 }
439 return dev;
440
441failure:
442 /* allow the register to be completed before unregistering. */
443 rtnl_unlock();
444 rtnl_lock();
445
446 unregister_netdevice(dev);
447 return NULL;
448}
449
450#ifdef CONFIG_IP_PIMSM
451
452static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
453{
454 struct net *net = dev_net(dev);
455 struct mr_table *mrt;
456 struct flowi4 fl4 = {
457 .flowi4_oif = dev->ifindex,
458 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
459 .flowi4_mark = skb->mark,
460 };
461 int err;
462
463 err = ipmr_fib_lookup(net, &fl4, &mrt);
464 if (err < 0) {
465 kfree_skb(skb);
466 return err;
467 }
468
469 read_lock(&mrt_lock);
470 dev->stats.tx_bytes += skb->len;
471 dev->stats.tx_packets++;
472 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
473 read_unlock(&mrt_lock);
474 kfree_skb(skb);
475 return NETDEV_TX_OK;
476}
477
478static const struct net_device_ops reg_vif_netdev_ops = {
479 .ndo_start_xmit = reg_vif_xmit,
480};
481
482static void reg_vif_setup(struct net_device *dev)
483{
484 dev->type = ARPHRD_PIMREG;
485 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
486 dev->flags = IFF_NOARP;
487 dev->netdev_ops = ®_vif_netdev_ops,
488 dev->destructor = free_netdev;
489 dev->features |= NETIF_F_NETNS_LOCAL;
490}
491
492static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
493{
494 struct net_device *dev;
495 struct in_device *in_dev;
496 char name[IFNAMSIZ];
497
498 if (mrt->id == RT_TABLE_DEFAULT)
499 sprintf(name, "pimreg");
500 else
501 sprintf(name, "pimreg%u", mrt->id);
502
503 dev = alloc_netdev(0, name, reg_vif_setup);
504
505 if (dev == NULL)
506 return NULL;
507
508 dev_net_set(dev, net);
509
510 if (register_netdevice(dev)) {
511 free_netdev(dev);
512 return NULL;
513 }
514 dev->iflink = 0;
515
516 rcu_read_lock();
517 in_dev = __in_dev_get_rcu(dev);
518 if (!in_dev) {
519 rcu_read_unlock();
520 goto failure;
521 }
522
523 ipv4_devconf_setall(in_dev);
524 neigh_parms_data_state_setall(in_dev->arp_parms);
525 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
526 rcu_read_unlock();
527
528 if (dev_open(dev))
529 goto failure;
530
531 dev_hold(dev);
532
533 return dev;
534
535failure:
536 /* allow the register to be completed before unregistering. */
537 rtnl_unlock();
538 rtnl_lock();
539
540 unregister_netdevice(dev);
541 return NULL;
542}
543#endif
544
545/**
546 * vif_delete - Delete a VIF entry
547 * @notify: Set to 1, if the caller is a notifier_call
548 */
549
550static int vif_delete(struct mr_table *mrt, int vifi, int notify,
551 struct list_head *head)
552{
553 struct vif_device *v;
554 struct net_device *dev;
555 struct in_device *in_dev;
556
557 if (vifi < 0 || vifi >= mrt->maxvif)
558 return -EADDRNOTAVAIL;
559
560 v = &mrt->vif_table[vifi];
561
562 write_lock_bh(&mrt_lock);
563 dev = v->dev;
564 v->dev = NULL;
565
566 if (!dev) {
567 write_unlock_bh(&mrt_lock);
568 return -EADDRNOTAVAIL;
569 }
570
571#ifdef CONFIG_IP_PIMSM
572 if (vifi == mrt->mroute_reg_vif_num)
573 mrt->mroute_reg_vif_num = -1;
574#endif
575
576 if (vifi + 1 == mrt->maxvif) {
577 int tmp;
578
579 for (tmp = vifi - 1; tmp >= 0; tmp--) {
580 if (VIF_EXISTS(mrt, tmp))
581 break;
582 }
583 mrt->maxvif = tmp+1;
584 }
585
586 write_unlock_bh(&mrt_lock);
587
588 dev_set_allmulti(dev, -1);
589
590 in_dev = __in_dev_get_rtnl(dev);
591 if (in_dev) {
592 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
593 inet_netconf_notify_devconf(dev_net(dev),
594 NETCONFA_MC_FORWARDING,
595 dev->ifindex, &in_dev->cnf);
596 ip_rt_multicast_event(in_dev);
597 }
598
599 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
600 unregister_netdevice_queue(dev, head);
601
602 dev_put(dev);
603 return 0;
604}
605
606static void ipmr_cache_free_rcu(struct rcu_head *head)
607{
608 struct mfc_cache *c = container_of(head, struct mfc_cache, rcu);
609
610 kmem_cache_free(mrt_cachep, c);
611}
612
613static inline void ipmr_cache_free(struct mfc_cache *c)
614{
615 call_rcu(&c->rcu, ipmr_cache_free_rcu);
616}
617
618/* Destroy an unresolved cache entry, killing queued skbs
619 * and reporting error to netlink readers.
620 */
621
622static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
623{
624 struct net *net = read_pnet(&mrt->net);
625 struct sk_buff *skb;
626 struct nlmsgerr *e;
627
628 atomic_dec(&mrt->cache_resolve_queue_len);
629
630 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) {
631 if (ip_hdr(skb)->version == 0) {
632 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
633 nlh->nlmsg_type = NLMSG_ERROR;
634 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
635 skb_trim(skb, nlh->nlmsg_len);
636 e = nlmsg_data(nlh);
637 e->error = -ETIMEDOUT;
638 memset(&e->msg, 0, sizeof(e->msg));
639
640 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
641 } else {
642 kfree_skb(skb);
643 }
644 }
645
646 ipmr_cache_free(c);
647}
648
649
650/* Timer process for the unresolved queue. */
651
652static void ipmr_expire_process(unsigned long arg)
653{
654 struct mr_table *mrt = (struct mr_table *)arg;
655 unsigned long now;
656 unsigned long expires;
657 struct mfc_cache *c, *next;
658
659 if (!spin_trylock(&mfc_unres_lock)) {
660 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
661 return;
662 }
663
664 if (list_empty(&mrt->mfc_unres_queue))
665 goto out;
666
667 now = jiffies;
668 expires = 10*HZ;
669
670 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
671 if (time_after(c->mfc_un.unres.expires, now)) {
672 unsigned long interval = c->mfc_un.unres.expires - now;
673 if (interval < expires)
674 expires = interval;
675 continue;
676 }
677
678 list_del(&c->list);
679 mroute_netlink_event(mrt, c, RTM_DELROUTE);
680 ipmr_destroy_unres(mrt, c);
681 }
682
683 if (!list_empty(&mrt->mfc_unres_queue))
684 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
685
686out:
687 spin_unlock(&mfc_unres_lock);
688}
689
690/* Fill oifs list. It is called under write locked mrt_lock. */
691
692static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache,
693 unsigned char *ttls)
694{
695 int vifi;
696
697 cache->mfc_un.res.minvif = MAXVIFS;
698 cache->mfc_un.res.maxvif = 0;
699 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
700
701 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
702 if (VIF_EXISTS(mrt, vifi) &&
703 ttls[vifi] && ttls[vifi] < 255) {
704 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
705 if (cache->mfc_un.res.minvif > vifi)
706 cache->mfc_un.res.minvif = vifi;
707 if (cache->mfc_un.res.maxvif <= vifi)
708 cache->mfc_un.res.maxvif = vifi + 1;
709 }
710 }
711}
712
713static int vif_add(struct net *net, struct mr_table *mrt,
714 struct vifctl *vifc, int mrtsock)
715{
716 int vifi = vifc->vifc_vifi;
717 struct vif_device *v = &mrt->vif_table[vifi];
718 struct net_device *dev;
719 struct in_device *in_dev;
720 int err;
721
722 /* Is vif busy ? */
723 if (VIF_EXISTS(mrt, vifi))
724 return -EADDRINUSE;
725
726 switch (vifc->vifc_flags) {
727#ifdef CONFIG_IP_PIMSM
728 case VIFF_REGISTER:
729 /*
730 * Special Purpose VIF in PIM
731 * All the packets will be sent to the daemon
732 */
733 if (mrt->mroute_reg_vif_num >= 0)
734 return -EADDRINUSE;
735 dev = ipmr_reg_vif(net, mrt);
736 if (!dev)
737 return -ENOBUFS;
738 err = dev_set_allmulti(dev, 1);
739 if (err) {
740 unregister_netdevice(dev);
741 dev_put(dev);
742 return err;
743 }
744 break;
745#endif
746 case VIFF_TUNNEL:
747 dev = ipmr_new_tunnel(net, vifc);
748 if (!dev)
749 return -ENOBUFS;
750 err = dev_set_allmulti(dev, 1);
751 if (err) {
752 ipmr_del_tunnel(dev, vifc);
753 dev_put(dev);
754 return err;
755 }
756 break;
757
758 case VIFF_USE_IFINDEX:
759 case 0:
760 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
761 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
762 if (dev && __in_dev_get_rtnl(dev) == NULL) {
763 dev_put(dev);
764 return -EADDRNOTAVAIL;
765 }
766 } else {
767 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
768 }
769 if (!dev)
770 return -EADDRNOTAVAIL;
771 err = dev_set_allmulti(dev, 1);
772 if (err) {
773 dev_put(dev);
774 return err;
775 }
776 break;
777 default:
778 return -EINVAL;
779 }
780
781 in_dev = __in_dev_get_rtnl(dev);
782 if (!in_dev) {
783 dev_put(dev);
784 return -EADDRNOTAVAIL;
785 }
786 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
787 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, dev->ifindex,
788 &in_dev->cnf);
789 ip_rt_multicast_event(in_dev);
790
791 /* Fill in the VIF structures */
792
793 v->rate_limit = vifc->vifc_rate_limit;
794 v->local = vifc->vifc_lcl_addr.s_addr;
795 v->remote = vifc->vifc_rmt_addr.s_addr;
796 v->flags = vifc->vifc_flags;
797 if (!mrtsock)
798 v->flags |= VIFF_STATIC;
799 v->threshold = vifc->vifc_threshold;
800 v->bytes_in = 0;
801 v->bytes_out = 0;
802 v->pkt_in = 0;
803 v->pkt_out = 0;
804 v->link = dev->ifindex;
805 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER))
806 v->link = dev->iflink;
807
808 /* And finish update writing critical data */
809 write_lock_bh(&mrt_lock);
810 v->dev = dev;
811#ifdef CONFIG_IP_PIMSM
812 if (v->flags & VIFF_REGISTER)
813 mrt->mroute_reg_vif_num = vifi;
814#endif
815 if (vifi+1 > mrt->maxvif)
816 mrt->maxvif = vifi+1;
817 write_unlock_bh(&mrt_lock);
818 return 0;
819}
820
821/* called with rcu_read_lock() */
822static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
823 __be32 origin,
824 __be32 mcastgrp)
825{
826 int line = MFC_HASH(mcastgrp, origin);
827 struct mfc_cache *c;
828
829 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) {
830 if (c->mfc_origin == origin && c->mfc_mcastgrp == mcastgrp)
831 return c;
832 }
833 return NULL;
834}
835
836/* Look for a (*,*,oif) entry */
837static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt,
838 int vifi)
839{
840 int line = MFC_HASH(htonl(INADDR_ANY), htonl(INADDR_ANY));
841 struct mfc_cache *c;
842
843 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
844 if (c->mfc_origin == htonl(INADDR_ANY) &&
845 c->mfc_mcastgrp == htonl(INADDR_ANY) &&
846 c->mfc_un.res.ttls[vifi] < 255)
847 return c;
848
849 return NULL;
850}
851
852/* Look for a (*,G) entry */
853static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
854 __be32 mcastgrp, int vifi)
855{
856 int line = MFC_HASH(mcastgrp, htonl(INADDR_ANY));
857 struct mfc_cache *c, *proxy;
858
859 if (mcastgrp == htonl(INADDR_ANY))
860 goto skip;
861
862 list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
863 if (c->mfc_origin == htonl(INADDR_ANY) &&
864 c->mfc_mcastgrp == mcastgrp) {
865 if (c->mfc_un.res.ttls[vifi] < 255)
866 return c;
867
868 /* It's ok if the vifi is part of the static tree */
869 proxy = ipmr_cache_find_any_parent(mrt,
870 c->mfc_parent);
871 if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
872 return c;
873 }
874
875skip:
876 return ipmr_cache_find_any_parent(mrt, vifi);
877}
878
879/*
880 * Allocate a multicast cache entry
881 */
882static struct mfc_cache *ipmr_cache_alloc(void)
883{
884 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
885
886 if (c)
887 c->mfc_un.res.minvif = MAXVIFS;
888 return c;
889}
890
891static struct mfc_cache *ipmr_cache_alloc_unres(void)
892{
893 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
894
895 if (c) {
896 skb_queue_head_init(&c->mfc_un.unres.unresolved);
897 c->mfc_un.unres.expires = jiffies + 10*HZ;
898 }
899 return c;
900}
901
902/*
903 * A cache entry has gone into a resolved state from queued
904 */
905
906static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
907 struct mfc_cache *uc, struct mfc_cache *c)
908{
909 struct sk_buff *skb;
910 struct nlmsgerr *e;
911
912 /* Play the pending entries through our router */
913
914 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
915 if (ip_hdr(skb)->version == 0) {
916 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
917
918 if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
919 nlh->nlmsg_len = skb_tail_pointer(skb) -
920 (u8 *)nlh;
921 } else {
922 nlh->nlmsg_type = NLMSG_ERROR;
923 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
924 skb_trim(skb, nlh->nlmsg_len);
925 e = nlmsg_data(nlh);
926 e->error = -EMSGSIZE;
927 memset(&e->msg, 0, sizeof(e->msg));
928 }
929
930 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
931 } else {
932 ip_mr_forward(net, mrt, skb, c, 0);
933 }
934 }
935}
936
937/*
938 * Bounce a cache query up to mrouted. We could use netlink for this but mrouted
939 * expects the following bizarre scheme.
940 *
941 * Called under mrt_lock.
942 */
943
944static int ipmr_cache_report(struct mr_table *mrt,
945 struct sk_buff *pkt, vifi_t vifi, int assert)
946{
947 struct sk_buff *skb;
948 const int ihl = ip_hdrlen(pkt);
949 struct igmphdr *igmp;
950 struct igmpmsg *msg;
951 struct sock *mroute_sk;
952 int ret;
953
954#ifdef CONFIG_IP_PIMSM
955 if (assert == IGMPMSG_WHOLEPKT)
956 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
957 else
958#endif
959 skb = alloc_skb(128, GFP_ATOMIC);
960
961 if (!skb)
962 return -ENOBUFS;
963
964#ifdef CONFIG_IP_PIMSM
965 if (assert == IGMPMSG_WHOLEPKT) {
966 /* Ugly, but we have no choice with this interface.
967 * Duplicate old header, fix ihl, length etc.
968 * And all this only to mangle msg->im_msgtype and
969 * to set msg->im_mbz to "mbz" :-)
970 */
971 skb_push(skb, sizeof(struct iphdr));
972 skb_reset_network_header(skb);
973 skb_reset_transport_header(skb);
974 msg = (struct igmpmsg *)skb_network_header(skb);
975 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
976 msg->im_msgtype = IGMPMSG_WHOLEPKT;
977 msg->im_mbz = 0;
978 msg->im_vif = mrt->mroute_reg_vif_num;
979 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
980 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
981 sizeof(struct iphdr));
982 } else
983#endif
984 {
985
986 /* Copy the IP header */
987
988 skb_set_network_header(skb, skb->len);
989 skb_put(skb, ihl);
990 skb_copy_to_linear_data(skb, pkt->data, ihl);
991 ip_hdr(skb)->protocol = 0; /* Flag to the kernel this is a route add */
992 msg = (struct igmpmsg *)skb_network_header(skb);
993 msg->im_vif = vifi;
994 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
995
996 /* Add our header */
997
998 igmp = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
999 igmp->type =
1000 msg->im_msgtype = assert;
1001 igmp->code = 0;
1002 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1003 skb->transport_header = skb->network_header;
1004 }
1005
1006 rcu_read_lock();
1007 mroute_sk = rcu_dereference(mrt->mroute_sk);
1008 if (mroute_sk == NULL) {
1009 rcu_read_unlock();
1010 kfree_skb(skb);
1011 return -EINVAL;
1012 }
1013
1014 /* Deliver to mrouted */
1015
1016 ret = sock_queue_rcv_skb(mroute_sk, skb);
1017 rcu_read_unlock();
1018 if (ret < 0) {
1019 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1020 kfree_skb(skb);
1021 }
1022
1023 return ret;
1024}
1025
1026/*
1027 * Queue a packet for resolution. It gets locked cache entry!
1028 */
1029
1030static int
1031ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, struct sk_buff *skb)
1032{
1033 bool found = false;
1034 int err;
1035 struct mfc_cache *c;
1036 const struct iphdr *iph = ip_hdr(skb);
1037
1038 spin_lock_bh(&mfc_unres_lock);
1039 list_for_each_entry(c, &mrt->mfc_unres_queue, list) {
1040 if (c->mfc_mcastgrp == iph->daddr &&
1041 c->mfc_origin == iph->saddr) {
1042 found = true;
1043 break;
1044 }
1045 }
1046
1047 if (!found) {
1048 /* Create a new entry if allowable */
1049
1050 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1051 (c = ipmr_cache_alloc_unres()) == NULL) {
1052 spin_unlock_bh(&mfc_unres_lock);
1053
1054 kfree_skb(skb);
1055 return -ENOBUFS;
1056 }
1057
1058 /* Fill in the new cache entry */
1059
1060 c->mfc_parent = -1;
1061 c->mfc_origin = iph->saddr;
1062 c->mfc_mcastgrp = iph->daddr;
1063
1064 /* Reflect first query at mrouted. */
1065
1066 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1067 if (err < 0) {
1068 /* If the report failed throw the cache entry
1069 out - Brad Parker
1070 */
1071 spin_unlock_bh(&mfc_unres_lock);
1072
1073 ipmr_cache_free(c);
1074 kfree_skb(skb);
1075 return err;
1076 }
1077
1078 atomic_inc(&mrt->cache_resolve_queue_len);
1079 list_add(&c->list, &mrt->mfc_unres_queue);
1080 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1081
1082 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1083 mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires);
1084 }
1085
1086 /* See if we can append the packet */
1087
1088 if (c->mfc_un.unres.unresolved.qlen > 3) {
1089 kfree_skb(skb);
1090 err = -ENOBUFS;
1091 } else {
1092 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1093 err = 0;
1094 }
1095
1096 spin_unlock_bh(&mfc_unres_lock);
1097 return err;
1098}
1099
1100/*
1101 * MFC cache manipulation by user space mroute daemon
1102 */
1103
1104static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1105{
1106 int line;
1107 struct mfc_cache *c, *next;
1108
1109 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1110
1111 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[line], list) {
1112 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1113 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1114 (parent == -1 || parent == c->mfc_parent)) {
1115 list_del_rcu(&c->list);
1116 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1117 ipmr_cache_free(c);
1118 return 0;
1119 }
1120 }
1121 return -ENOENT;
1122}
1123
1124static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1125 struct mfcctl *mfc, int mrtsock, int parent)
1126{
1127 bool found = false;
1128 int line;
1129 struct mfc_cache *uc, *c;
1130
1131 if (mfc->mfcc_parent >= MAXVIFS)
1132 return -ENFILE;
1133
1134 line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1135
1136 list_for_each_entry(c, &mrt->mfc_cache_array[line], list) {
1137 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1138 c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1139 (parent == -1 || parent == c->mfc_parent)) {
1140 found = true;
1141 break;
1142 }
1143 }
1144
1145 if (found) {
1146 write_lock_bh(&mrt_lock);
1147 c->mfc_parent = mfc->mfcc_parent;
1148 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1149 if (!mrtsock)
1150 c->mfc_flags |= MFC_STATIC;
1151 write_unlock_bh(&mrt_lock);
1152 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1153 return 0;
1154 }
1155
1156 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1157 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1158 return -EINVAL;
1159
1160 c = ipmr_cache_alloc();
1161 if (c == NULL)
1162 return -ENOMEM;
1163
1164 c->mfc_origin = mfc->mfcc_origin.s_addr;
1165 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1166 c->mfc_parent = mfc->mfcc_parent;
1167 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1168 if (!mrtsock)
1169 c->mfc_flags |= MFC_STATIC;
1170
1171 list_add_rcu(&c->list, &mrt->mfc_cache_array[line]);
1172
1173 /*
1174 * Check to see if we resolved a queued list. If so we
1175 * need to send on the frames and tidy up.
1176 */
1177 found = false;
1178 spin_lock_bh(&mfc_unres_lock);
1179 list_for_each_entry(uc, &mrt->mfc_unres_queue, list) {
1180 if (uc->mfc_origin == c->mfc_origin &&
1181 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1182 list_del(&uc->list);
1183 atomic_dec(&mrt->cache_resolve_queue_len);
1184 found = true;
1185 break;
1186 }
1187 }
1188 if (list_empty(&mrt->mfc_unres_queue))
1189 del_timer(&mrt->ipmr_expire_timer);
1190 spin_unlock_bh(&mfc_unres_lock);
1191
1192 if (found) {
1193 ipmr_cache_resolve(net, mrt, uc, c);
1194 ipmr_cache_free(uc);
1195 }
1196 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1197 return 0;
1198}
1199
1200/*
1201 * Close the multicast socket, and clear the vif tables etc
1202 */
1203
1204static void mroute_clean_tables(struct mr_table *mrt)
1205{
1206 int i;
1207 LIST_HEAD(list);
1208 struct mfc_cache *c, *next;
1209
1210 /* Shut down all active vif entries */
1211
1212 for (i = 0; i < mrt->maxvif; i++) {
1213 if (!(mrt->vif_table[i].flags & VIFF_STATIC))
1214 vif_delete(mrt, i, 0, &list);
1215 }
1216 unregister_netdevice_many(&list);
1217
1218 /* Wipe the cache */
1219
1220 for (i = 0; i < MFC_LINES; i++) {
1221 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[i], list) {
1222 if (c->mfc_flags & MFC_STATIC)
1223 continue;
1224 list_del_rcu(&c->list);
1225 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1226 ipmr_cache_free(c);
1227 }
1228 }
1229
1230 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1231 spin_lock_bh(&mfc_unres_lock);
1232 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
1233 list_del(&c->list);
1234 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1235 ipmr_destroy_unres(mrt, c);
1236 }
1237 spin_unlock_bh(&mfc_unres_lock);
1238 }
1239}
1240
1241/* called from ip_ra_control(), before an RCU grace period,
1242 * we dont need to call synchronize_rcu() here
1243 */
1244static void mrtsock_destruct(struct sock *sk)
1245{
1246 struct net *net = sock_net(sk);
1247 struct mr_table *mrt;
1248
1249 rtnl_lock();
1250 ipmr_for_each_table(mrt, net) {
1251 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1252 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1253 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1254 NETCONFA_IFINDEX_ALL,
1255 net->ipv4.devconf_all);
1256 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1257 mroute_clean_tables(mrt);
1258 }
1259 }
1260 rtnl_unlock();
1261}
1262
1263/*
1264 * Socket options and virtual interface manipulation. The whole
1265 * virtual interface system is a complete heap, but unfortunately
1266 * that's how BSD mrouted happens to think. Maybe one day with a proper
1267 * MOSPF/PIM router set up we can clean this up.
1268 */
1269
1270int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1271{
1272 int ret, parent = 0;
1273 struct vifctl vif;
1274 struct mfcctl mfc;
1275 struct net *net = sock_net(sk);
1276 struct mr_table *mrt;
1277
1278 if (sk->sk_type != SOCK_RAW ||
1279 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1280 return -EOPNOTSUPP;
1281
1282 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1283 if (mrt == NULL)
1284 return -ENOENT;
1285
1286 if (optname != MRT_INIT) {
1287 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1288 !ns_capable(net->user_ns, CAP_NET_ADMIN))
1289 return -EACCES;
1290 }
1291
1292 switch (optname) {
1293 case MRT_INIT:
1294 if (optlen != sizeof(int))
1295 return -EINVAL;
1296
1297 rtnl_lock();
1298 if (rtnl_dereference(mrt->mroute_sk)) {
1299 rtnl_unlock();
1300 return -EADDRINUSE;
1301 }
1302
1303 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1304 if (ret == 0) {
1305 rcu_assign_pointer(mrt->mroute_sk, sk);
1306 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1307 inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1308 NETCONFA_IFINDEX_ALL,
1309 net->ipv4.devconf_all);
1310 }
1311 rtnl_unlock();
1312 return ret;
1313 case MRT_DONE:
1314 if (sk != rcu_access_pointer(mrt->mroute_sk))
1315 return -EACCES;
1316 return ip_ra_control(sk, 0, NULL);
1317 case MRT_ADD_VIF:
1318 case MRT_DEL_VIF:
1319 if (optlen != sizeof(vif))
1320 return -EINVAL;
1321 if (copy_from_user(&vif, optval, sizeof(vif)))
1322 return -EFAULT;
1323 if (vif.vifc_vifi >= MAXVIFS)
1324 return -ENFILE;
1325 rtnl_lock();
1326 if (optname == MRT_ADD_VIF) {
1327 ret = vif_add(net, mrt, &vif,
1328 sk == rtnl_dereference(mrt->mroute_sk));
1329 } else {
1330 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1331 }
1332 rtnl_unlock();
1333 return ret;
1334
1335 /*
1336 * Manipulate the forwarding caches. These live
1337 * in a sort of kernel/user symbiosis.
1338 */
1339 case MRT_ADD_MFC:
1340 case MRT_DEL_MFC:
1341 parent = -1;
1342 case MRT_ADD_MFC_PROXY:
1343 case MRT_DEL_MFC_PROXY:
1344 if (optlen != sizeof(mfc))
1345 return -EINVAL;
1346 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1347 return -EFAULT;
1348 if (parent == 0)
1349 parent = mfc.mfcc_parent;
1350 rtnl_lock();
1351 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1352 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1353 else
1354 ret = ipmr_mfc_add(net, mrt, &mfc,
1355 sk == rtnl_dereference(mrt->mroute_sk),
1356 parent);
1357 rtnl_unlock();
1358 return ret;
1359 /*
1360 * Control PIM assert.
1361 */
1362 case MRT_ASSERT:
1363 {
1364 int v;
1365 if (optlen != sizeof(v))
1366 return -EINVAL;
1367 if (get_user(v, (int __user *)optval))
1368 return -EFAULT;
1369 mrt->mroute_do_assert = v;
1370 return 0;
1371 }
1372#ifdef CONFIG_IP_PIMSM
1373 case MRT_PIM:
1374 {
1375 int v;
1376
1377 if (optlen != sizeof(v))
1378 return -EINVAL;
1379 if (get_user(v, (int __user *)optval))
1380 return -EFAULT;
1381 v = !!v;
1382
1383 rtnl_lock();
1384 ret = 0;
1385 if (v != mrt->mroute_do_pim) {
1386 mrt->mroute_do_pim = v;
1387 mrt->mroute_do_assert = v;
1388 }
1389 rtnl_unlock();
1390 return ret;
1391 }
1392#endif
1393#ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
1394 case MRT_TABLE:
1395 {
1396 u32 v;
1397
1398 if (optlen != sizeof(u32))
1399 return -EINVAL;
1400 if (get_user(v, (u32 __user *)optval))
1401 return -EFAULT;
1402
1403 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
1404 if (v != RT_TABLE_DEFAULT && v >= 1000000000)
1405 return -EINVAL;
1406
1407 rtnl_lock();
1408 ret = 0;
1409 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1410 ret = -EBUSY;
1411 } else {
1412 if (!ipmr_new_table(net, v))
1413 ret = -ENOMEM;
1414 else
1415 raw_sk(sk)->ipmr_table = v;
1416 }
1417 rtnl_unlock();
1418 return ret;
1419 }
1420#endif
1421 /*
1422 * Spurious command, or MRT_VERSION which you cannot
1423 * set.
1424 */
1425 default:
1426 return -ENOPROTOOPT;
1427 }
1428}
1429
1430/*
1431 * Getsock opt support for the multicast routing system.
1432 */
1433
1434int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1435{
1436 int olr;
1437 int val;
1438 struct net *net = sock_net(sk);
1439 struct mr_table *mrt;
1440
1441 if (sk->sk_type != SOCK_RAW ||
1442 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1443 return -EOPNOTSUPP;
1444
1445 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1446 if (mrt == NULL)
1447 return -ENOENT;
1448
1449 if (optname != MRT_VERSION &&
1450#ifdef CONFIG_IP_PIMSM
1451 optname != MRT_PIM &&
1452#endif
1453 optname != MRT_ASSERT)
1454 return -ENOPROTOOPT;
1455
1456 if (get_user(olr, optlen))
1457 return -EFAULT;
1458
1459 olr = min_t(unsigned int, olr, sizeof(int));
1460 if (olr < 0)
1461 return -EINVAL;
1462
1463 if (put_user(olr, optlen))
1464 return -EFAULT;
1465 if (optname == MRT_VERSION)
1466 val = 0x0305;
1467#ifdef CONFIG_IP_PIMSM
1468 else if (optname == MRT_PIM)
1469 val = mrt->mroute_do_pim;
1470#endif
1471 else
1472 val = mrt->mroute_do_assert;
1473 if (copy_to_user(optval, &val, olr))
1474 return -EFAULT;
1475 return 0;
1476}
1477
1478/*
1479 * The IP multicast ioctl support routines.
1480 */
1481
1482int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1483{
1484 struct sioc_sg_req sr;
1485 struct sioc_vif_req vr;
1486 struct vif_device *vif;
1487 struct mfc_cache *c;
1488 struct net *net = sock_net(sk);
1489 struct mr_table *mrt;
1490
1491 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1492 if (mrt == NULL)
1493 return -ENOENT;
1494
1495 switch (cmd) {
1496 case SIOCGETVIFCNT:
1497 if (copy_from_user(&vr, arg, sizeof(vr)))
1498 return -EFAULT;
1499 if (vr.vifi >= mrt->maxvif)
1500 return -EINVAL;
1501 read_lock(&mrt_lock);
1502 vif = &mrt->vif_table[vr.vifi];
1503 if (VIF_EXISTS(mrt, vr.vifi)) {
1504 vr.icount = vif->pkt_in;
1505 vr.ocount = vif->pkt_out;
1506 vr.ibytes = vif->bytes_in;
1507 vr.obytes = vif->bytes_out;
1508 read_unlock(&mrt_lock);
1509
1510 if (copy_to_user(arg, &vr, sizeof(vr)))
1511 return -EFAULT;
1512 return 0;
1513 }
1514 read_unlock(&mrt_lock);
1515 return -EADDRNOTAVAIL;
1516 case SIOCGETSGCNT:
1517 if (copy_from_user(&sr, arg, sizeof(sr)))
1518 return -EFAULT;
1519
1520 rcu_read_lock();
1521 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1522 if (c) {
1523 sr.pktcnt = c->mfc_un.res.pkt;
1524 sr.bytecnt = c->mfc_un.res.bytes;
1525 sr.wrong_if = c->mfc_un.res.wrong_if;
1526 rcu_read_unlock();
1527
1528 if (copy_to_user(arg, &sr, sizeof(sr)))
1529 return -EFAULT;
1530 return 0;
1531 }
1532 rcu_read_unlock();
1533 return -EADDRNOTAVAIL;
1534 default:
1535 return -ENOIOCTLCMD;
1536 }
1537}
1538
1539#ifdef CONFIG_COMPAT
1540struct compat_sioc_sg_req {
1541 struct in_addr src;
1542 struct in_addr grp;
1543 compat_ulong_t pktcnt;
1544 compat_ulong_t bytecnt;
1545 compat_ulong_t wrong_if;
1546};
1547
1548struct compat_sioc_vif_req {
1549 vifi_t vifi; /* Which iface */
1550 compat_ulong_t icount;
1551 compat_ulong_t ocount;
1552 compat_ulong_t ibytes;
1553 compat_ulong_t obytes;
1554};
1555
1556int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1557{
1558 struct compat_sioc_sg_req sr;
1559 struct compat_sioc_vif_req vr;
1560 struct vif_device *vif;
1561 struct mfc_cache *c;
1562 struct net *net = sock_net(sk);
1563 struct mr_table *mrt;
1564
1565 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1566 if (mrt == NULL)
1567 return -ENOENT;
1568
1569 switch (cmd) {
1570 case SIOCGETVIFCNT:
1571 if (copy_from_user(&vr, arg, sizeof(vr)))
1572 return -EFAULT;
1573 if (vr.vifi >= mrt->maxvif)
1574 return -EINVAL;
1575 read_lock(&mrt_lock);
1576 vif = &mrt->vif_table[vr.vifi];
1577 if (VIF_EXISTS(mrt, vr.vifi)) {
1578 vr.icount = vif->pkt_in;
1579 vr.ocount = vif->pkt_out;
1580 vr.ibytes = vif->bytes_in;
1581 vr.obytes = vif->bytes_out;
1582 read_unlock(&mrt_lock);
1583
1584 if (copy_to_user(arg, &vr, sizeof(vr)))
1585 return -EFAULT;
1586 return 0;
1587 }
1588 read_unlock(&mrt_lock);
1589 return -EADDRNOTAVAIL;
1590 case SIOCGETSGCNT:
1591 if (copy_from_user(&sr, arg, sizeof(sr)))
1592 return -EFAULT;
1593
1594 rcu_read_lock();
1595 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1596 if (c) {
1597 sr.pktcnt = c->mfc_un.res.pkt;
1598 sr.bytecnt = c->mfc_un.res.bytes;
1599 sr.wrong_if = c->mfc_un.res.wrong_if;
1600 rcu_read_unlock();
1601
1602 if (copy_to_user(arg, &sr, sizeof(sr)))
1603 return -EFAULT;
1604 return 0;
1605 }
1606 rcu_read_unlock();
1607 return -EADDRNOTAVAIL;
1608 default:
1609 return -ENOIOCTLCMD;
1610 }
1611}
1612#endif
1613
1614
1615static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1616{
1617 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1618 struct net *net = dev_net(dev);
1619 struct mr_table *mrt;
1620 struct vif_device *v;
1621 int ct;
1622
1623 if (event != NETDEV_UNREGISTER)
1624 return NOTIFY_DONE;
1625
1626 ipmr_for_each_table(mrt, net) {
1627 v = &mrt->vif_table[0];
1628 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1629 if (v->dev == dev)
1630 vif_delete(mrt, ct, 1, NULL);
1631 }
1632 }
1633 return NOTIFY_DONE;
1634}
1635
1636
1637static struct notifier_block ip_mr_notifier = {
1638 .notifier_call = ipmr_device_event,
1639};
1640
1641/*
1642 * Encapsulate a packet by attaching a valid IPIP header to it.
1643 * This avoids tunnel drivers and other mess and gives us the speed so
1644 * important for multicast video.
1645 */
1646
1647static void ip_encap(struct sk_buff *skb, __be32 saddr, __be32 daddr)
1648{
1649 struct iphdr *iph;
1650 const struct iphdr *old_iph = ip_hdr(skb);
1651
1652 skb_push(skb, sizeof(struct iphdr));
1653 skb->transport_header = skb->network_header;
1654 skb_reset_network_header(skb);
1655 iph = ip_hdr(skb);
1656
1657 iph->version = 4;
1658 iph->tos = old_iph->tos;
1659 iph->ttl = old_iph->ttl;
1660 iph->frag_off = 0;
1661 iph->daddr = daddr;
1662 iph->saddr = saddr;
1663 iph->protocol = IPPROTO_IPIP;
1664 iph->ihl = 5;
1665 iph->tot_len = htons(skb->len);
1666 ip_select_ident(skb, skb_dst(skb), NULL);
1667 ip_send_check(iph);
1668
1669 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1670 nf_reset(skb);
1671}
1672
1673static inline int ipmr_forward_finish(struct sk_buff *skb)
1674{
1675 struct ip_options *opt = &(IPCB(skb)->opt);
1676
1677 IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTFORWDATAGRAMS);
1678 IP_ADD_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTOCTETS, skb->len);
1679
1680 if (unlikely(opt->optlen))
1681 ip_forward_options(skb);
1682
1683 return dst_output(skb);
1684}
1685
1686/*
1687 * Processing handlers for ipmr_forward
1688 */
1689
1690static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1691 struct sk_buff *skb, struct mfc_cache *c, int vifi)
1692{
1693 const struct iphdr *iph = ip_hdr(skb);
1694 struct vif_device *vif = &mrt->vif_table[vifi];
1695 struct net_device *dev;
1696 struct rtable *rt;
1697 struct flowi4 fl4;
1698 int encap = 0;
1699
1700 if (vif->dev == NULL)
1701 goto out_free;
1702
1703#ifdef CONFIG_IP_PIMSM
1704 if (vif->flags & VIFF_REGISTER) {
1705 vif->pkt_out++;
1706 vif->bytes_out += skb->len;
1707 vif->dev->stats.tx_bytes += skb->len;
1708 vif->dev->stats.tx_packets++;
1709 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1710 goto out_free;
1711 }
1712#endif
1713
1714 if (vif->flags & VIFF_TUNNEL) {
1715 rt = ip_route_output_ports(net, &fl4, NULL,
1716 vif->remote, vif->local,
1717 0, 0,
1718 IPPROTO_IPIP,
1719 RT_TOS(iph->tos), vif->link);
1720 if (IS_ERR(rt))
1721 goto out_free;
1722 encap = sizeof(struct iphdr);
1723 } else {
1724 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1725 0, 0,
1726 IPPROTO_IPIP,
1727 RT_TOS(iph->tos), vif->link);
1728 if (IS_ERR(rt))
1729 goto out_free;
1730 }
1731
1732 dev = rt->dst.dev;
1733
1734 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1735 /* Do not fragment multicasts. Alas, IPv4 does not
1736 * allow to send ICMP, so that packets will disappear
1737 * to blackhole.
1738 */
1739
1740 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
1741 ip_rt_put(rt);
1742 goto out_free;
1743 }
1744
1745 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1746
1747 if (skb_cow(skb, encap)) {
1748 ip_rt_put(rt);
1749 goto out_free;
1750 }
1751
1752 vif->pkt_out++;
1753 vif->bytes_out += skb->len;
1754
1755 skb_dst_drop(skb);
1756 skb_dst_set(skb, &rt->dst);
1757 ip_decrease_ttl(ip_hdr(skb));
1758
1759 /* FIXME: forward and output firewalls used to be called here.
1760 * What do we do with netfilter? -- RR
1761 */
1762 if (vif->flags & VIFF_TUNNEL) {
1763 ip_encap(skb, vif->local, vif->remote);
1764 /* FIXME: extra output firewall step used to be here. --RR */
1765 vif->dev->stats.tx_packets++;
1766 vif->dev->stats.tx_bytes += skb->len;
1767 }
1768
1769 IPCB(skb)->flags |= IPSKB_FORWARDED;
1770
1771 /*
1772 * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1773 * not only before forwarding, but after forwarding on all output
1774 * interfaces. It is clear, if mrouter runs a multicasting
1775 * program, it should receive packets not depending to what interface
1776 * program is joined.
1777 * If we will not make it, the program will have to join on all
1778 * interfaces. On the other hand, multihoming host (or router, but
1779 * not mrouter) cannot join to more than one interface - it will
1780 * result in receiving multiple packets.
1781 */
1782 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, skb, skb->dev, dev,
1783 ipmr_forward_finish);
1784 return;
1785
1786out_free:
1787 kfree_skb(skb);
1788}
1789
1790static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1791{
1792 int ct;
1793
1794 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1795 if (mrt->vif_table[ct].dev == dev)
1796 break;
1797 }
1798 return ct;
1799}
1800
1801/* "local" means that we should preserve one skb (for local delivery) */
1802
1803static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1804 struct sk_buff *skb, struct mfc_cache *cache,
1805 int local)
1806{
1807 int psend = -1;
1808 int vif, ct;
1809 int true_vifi = ipmr_find_vif(mrt, skb->dev);
1810
1811 vif = cache->mfc_parent;
1812 cache->mfc_un.res.pkt++;
1813 cache->mfc_un.res.bytes += skb->len;
1814
1815 if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1816 struct mfc_cache *cache_proxy;
1817
1818 /* For an (*,G) entry, we only check that the incomming
1819 * interface is part of the static tree.
1820 */
1821 cache_proxy = ipmr_cache_find_any_parent(mrt, vif);
1822 if (cache_proxy &&
1823 cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
1824 goto forward;
1825 }
1826
1827 /*
1828 * Wrong interface: drop packet and (maybe) send PIM assert.
1829 */
1830 if (mrt->vif_table[vif].dev != skb->dev) {
1831 if (rt_is_output_route(skb_rtable(skb))) {
1832 /* It is our own packet, looped back.
1833 * Very complicated situation...
1834 *
1835 * The best workaround until routing daemons will be
1836 * fixed is not to redistribute packet, if it was
1837 * send through wrong interface. It means, that
1838 * multicast applications WILL NOT work for
1839 * (S,G), which have default multicast route pointing
1840 * to wrong oif. In any case, it is not a good
1841 * idea to use multicasting applications on router.
1842 */
1843 goto dont_forward;
1844 }
1845
1846 cache->mfc_un.res.wrong_if++;
1847
1848 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1849 /* pimsm uses asserts, when switching from RPT to SPT,
1850 * so that we cannot check that packet arrived on an oif.
1851 * It is bad, but otherwise we would need to move pretty
1852 * large chunk of pimd to kernel. Ough... --ANK
1853 */
1854 (mrt->mroute_do_pim ||
1855 cache->mfc_un.res.ttls[true_vifi] < 255) &&
1856 time_after(jiffies,
1857 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1858 cache->mfc_un.res.last_assert = jiffies;
1859 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1860 }
1861 goto dont_forward;
1862 }
1863
1864forward:
1865 mrt->vif_table[vif].pkt_in++;
1866 mrt->vif_table[vif].bytes_in += skb->len;
1867
1868 /*
1869 * Forward the frame
1870 */
1871 if (cache->mfc_origin == htonl(INADDR_ANY) &&
1872 cache->mfc_mcastgrp == htonl(INADDR_ANY)) {
1873 if (true_vifi >= 0 &&
1874 true_vifi != cache->mfc_parent &&
1875 ip_hdr(skb)->ttl >
1876 cache->mfc_un.res.ttls[cache->mfc_parent]) {
1877 /* It's an (*,*) entry and the packet is not coming from
1878 * the upstream: forward the packet to the upstream
1879 * only.
1880 */
1881 psend = cache->mfc_parent;
1882 goto last_forward;
1883 }
1884 goto dont_forward;
1885 }
1886 for (ct = cache->mfc_un.res.maxvif - 1;
1887 ct >= cache->mfc_un.res.minvif; ct--) {
1888 /* For (*,G) entry, don't forward to the incoming interface */
1889 if ((cache->mfc_origin != htonl(INADDR_ANY) ||
1890 ct != true_vifi) &&
1891 ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
1892 if (psend != -1) {
1893 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1894
1895 if (skb2)
1896 ipmr_queue_xmit(net, mrt, skb2, cache,
1897 psend);
1898 }
1899 psend = ct;
1900 }
1901 }
1902last_forward:
1903 if (psend != -1) {
1904 if (local) {
1905 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1906
1907 if (skb2)
1908 ipmr_queue_xmit(net, mrt, skb2, cache, psend);
1909 } else {
1910 ipmr_queue_xmit(net, mrt, skb, cache, psend);
1911 return;
1912 }
1913 }
1914
1915dont_forward:
1916 if (!local)
1917 kfree_skb(skb);
1918}
1919
1920static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
1921{
1922 struct rtable *rt = skb_rtable(skb);
1923 struct iphdr *iph = ip_hdr(skb);
1924 struct flowi4 fl4 = {
1925 .daddr = iph->daddr,
1926 .saddr = iph->saddr,
1927 .flowi4_tos = RT_TOS(iph->tos),
1928 .flowi4_oif = (rt_is_output_route(rt) ?
1929 skb->dev->ifindex : 0),
1930 .flowi4_iif = (rt_is_output_route(rt) ?
1931 LOOPBACK_IFINDEX :
1932 skb->dev->ifindex),
1933 .flowi4_mark = skb->mark,
1934 };
1935 struct mr_table *mrt;
1936 int err;
1937
1938 err = ipmr_fib_lookup(net, &fl4, &mrt);
1939 if (err)
1940 return ERR_PTR(err);
1941 return mrt;
1942}
1943
1944/*
1945 * Multicast packets for forwarding arrive here
1946 * Called with rcu_read_lock();
1947 */
1948
1949int ip_mr_input(struct sk_buff *skb)
1950{
1951 struct mfc_cache *cache;
1952 struct net *net = dev_net(skb->dev);
1953 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
1954 struct mr_table *mrt;
1955
1956 /* Packet is looped back after forward, it should not be
1957 * forwarded second time, but still can be delivered locally.
1958 */
1959 if (IPCB(skb)->flags & IPSKB_FORWARDED)
1960 goto dont_forward;
1961
1962 mrt = ipmr_rt_fib_lookup(net, skb);
1963 if (IS_ERR(mrt)) {
1964 kfree_skb(skb);
1965 return PTR_ERR(mrt);
1966 }
1967 if (!local) {
1968 if (IPCB(skb)->opt.router_alert) {
1969 if (ip_call_ra_chain(skb))
1970 return 0;
1971 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
1972 /* IGMPv1 (and broken IGMPv2 implementations sort of
1973 * Cisco IOS <= 11.2(8)) do not put router alert
1974 * option to IGMP packets destined to routable
1975 * groups. It is very bad, because it means
1976 * that we can forward NO IGMP messages.
1977 */
1978 struct sock *mroute_sk;
1979
1980 mroute_sk = rcu_dereference(mrt->mroute_sk);
1981 if (mroute_sk) {
1982 nf_reset(skb);
1983 raw_rcv(mroute_sk, skb);
1984 return 0;
1985 }
1986 }
1987 }
1988
1989 /* already under rcu_read_lock() */
1990 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
1991 if (cache == NULL) {
1992 int vif = ipmr_find_vif(mrt, skb->dev);
1993
1994 if (vif >= 0)
1995 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
1996 vif);
1997 }
1998
1999 /*
2000 * No usable cache entry
2001 */
2002 if (cache == NULL) {
2003 int vif;
2004
2005 if (local) {
2006 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2007 ip_local_deliver(skb);
2008 if (skb2 == NULL)
2009 return -ENOBUFS;
2010 skb = skb2;
2011 }
2012
2013 read_lock(&mrt_lock);
2014 vif = ipmr_find_vif(mrt, skb->dev);
2015 if (vif >= 0) {
2016 int err2 = ipmr_cache_unresolved(mrt, vif, skb);
2017 read_unlock(&mrt_lock);
2018
2019 return err2;
2020 }
2021 read_unlock(&mrt_lock);
2022 kfree_skb(skb);
2023 return -ENODEV;
2024 }
2025
2026 read_lock(&mrt_lock);
2027 ip_mr_forward(net, mrt, skb, cache, local);
2028 read_unlock(&mrt_lock);
2029
2030 if (local)
2031 return ip_local_deliver(skb);
2032
2033 return 0;
2034
2035dont_forward:
2036 if (local)
2037 return ip_local_deliver(skb);
2038 kfree_skb(skb);
2039 return 0;
2040}
2041
2042#ifdef CONFIG_IP_PIMSM
2043/* called with rcu_read_lock() */
2044static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
2045 unsigned int pimlen)
2046{
2047 struct net_device *reg_dev = NULL;
2048 struct iphdr *encap;
2049
2050 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
2051 /*
2052 * Check that:
2053 * a. packet is really sent to a multicast group
2054 * b. packet is not a NULL-REGISTER
2055 * c. packet is not truncated
2056 */
2057 if (!ipv4_is_multicast(encap->daddr) ||
2058 encap->tot_len == 0 ||
2059 ntohs(encap->tot_len) + pimlen > skb->len)
2060 return 1;
2061
2062 read_lock(&mrt_lock);
2063 if (mrt->mroute_reg_vif_num >= 0)
2064 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
2065 read_unlock(&mrt_lock);
2066
2067 if (reg_dev == NULL)
2068 return 1;
2069
2070 skb->mac_header = skb->network_header;
2071 skb_pull(skb, (u8 *)encap - skb->data);
2072 skb_reset_network_header(skb);
2073 skb->protocol = htons(ETH_P_IP);
2074 skb->ip_summed = CHECKSUM_NONE;
2075
2076 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
2077
2078 netif_rx(skb);
2079
2080 return NET_RX_SUCCESS;
2081}
2082#endif
2083
2084#ifdef CONFIG_IP_PIMSM_V1
2085/*
2086 * Handle IGMP messages of PIMv1
2087 */
2088
2089int pim_rcv_v1(struct sk_buff *skb)
2090{
2091 struct igmphdr *pim;
2092 struct net *net = dev_net(skb->dev);
2093 struct mr_table *mrt;
2094
2095 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2096 goto drop;
2097
2098 pim = igmp_hdr(skb);
2099
2100 mrt = ipmr_rt_fib_lookup(net, skb);
2101 if (IS_ERR(mrt))
2102 goto drop;
2103 if (!mrt->mroute_do_pim ||
2104 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2105 goto drop;
2106
2107 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2108drop:
2109 kfree_skb(skb);
2110 }
2111 return 0;
2112}
2113#endif
2114
2115#ifdef CONFIG_IP_PIMSM_V2
2116static int pim_rcv(struct sk_buff *skb)
2117{
2118 struct pimreghdr *pim;
2119 struct net *net = dev_net(skb->dev);
2120 struct mr_table *mrt;
2121
2122 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2123 goto drop;
2124
2125 pim = (struct pimreghdr *)skb_transport_header(skb);
2126 if (pim->type != ((PIM_VERSION << 4) | (PIM_REGISTER)) ||
2127 (pim->flags & PIM_NULL_REGISTER) ||
2128 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2129 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2130 goto drop;
2131
2132 mrt = ipmr_rt_fib_lookup(net, skb);
2133 if (IS_ERR(mrt))
2134 goto drop;
2135 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2136drop:
2137 kfree_skb(skb);
2138 }
2139 return 0;
2140}
2141#endif
2142
2143static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2144 struct mfc_cache *c, struct rtmsg *rtm)
2145{
2146 int ct;
2147 struct rtnexthop *nhp;
2148 struct nlattr *mp_attr;
2149 struct rta_mfc_stats mfcs;
2150
2151 /* If cache is unresolved, don't try to parse IIF and OIF */
2152 if (c->mfc_parent >= MAXVIFS)
2153 return -ENOENT;
2154
2155 if (VIF_EXISTS(mrt, c->mfc_parent) &&
2156 nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
2157 return -EMSGSIZE;
2158
2159 if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH)))
2160 return -EMSGSIZE;
2161
2162 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2163 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2164 if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) {
2165 nla_nest_cancel(skb, mp_attr);
2166 return -EMSGSIZE;
2167 }
2168
2169 nhp->rtnh_flags = 0;
2170 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2171 nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex;
2172 nhp->rtnh_len = sizeof(*nhp);
2173 }
2174 }
2175
2176 nla_nest_end(skb, mp_attr);
2177
2178 mfcs.mfcs_packets = c->mfc_un.res.pkt;
2179 mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2180 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2181 if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0)
2182 return -EMSGSIZE;
2183
2184 rtm->rtm_type = RTN_MULTICAST;
2185 return 1;
2186}
2187
2188int ipmr_get_route(struct net *net, struct sk_buff *skb,
2189 __be32 saddr, __be32 daddr,
2190 struct rtmsg *rtm, int nowait)
2191{
2192 struct mfc_cache *cache;
2193 struct mr_table *mrt;
2194 int err;
2195
2196 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2197 if (mrt == NULL)
2198 return -ENOENT;
2199
2200 rcu_read_lock();
2201 cache = ipmr_cache_find(mrt, saddr, daddr);
2202 if (cache == NULL && skb->dev) {
2203 int vif = ipmr_find_vif(mrt, skb->dev);
2204
2205 if (vif >= 0)
2206 cache = ipmr_cache_find_any(mrt, daddr, vif);
2207 }
2208 if (cache == NULL) {
2209 struct sk_buff *skb2;
2210 struct iphdr *iph;
2211 struct net_device *dev;
2212 int vif = -1;
2213
2214 if (nowait) {
2215 rcu_read_unlock();
2216 return -EAGAIN;
2217 }
2218
2219 dev = skb->dev;
2220 read_lock(&mrt_lock);
2221 if (dev)
2222 vif = ipmr_find_vif(mrt, dev);
2223 if (vif < 0) {
2224 read_unlock(&mrt_lock);
2225 rcu_read_unlock();
2226 return -ENODEV;
2227 }
2228 skb2 = skb_clone(skb, GFP_ATOMIC);
2229 if (!skb2) {
2230 read_unlock(&mrt_lock);
2231 rcu_read_unlock();
2232 return -ENOMEM;
2233 }
2234
2235 skb_push(skb2, sizeof(struct iphdr));
2236 skb_reset_network_header(skb2);
2237 iph = ip_hdr(skb2);
2238 iph->ihl = sizeof(struct iphdr) >> 2;
2239 iph->saddr = saddr;
2240 iph->daddr = daddr;
2241 iph->version = 0;
2242 err = ipmr_cache_unresolved(mrt, vif, skb2);
2243 read_unlock(&mrt_lock);
2244 rcu_read_unlock();
2245 return err;
2246 }
2247
2248 read_lock(&mrt_lock);
2249 if (!nowait && (rtm->rtm_flags & RTM_F_NOTIFY))
2250 cache->mfc_flags |= MFC_NOTIFY;
2251 err = __ipmr_fill_mroute(mrt, skb, cache, rtm);
2252 read_unlock(&mrt_lock);
2253 rcu_read_unlock();
2254 return err;
2255}
2256
2257static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2258 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2259 int flags)
2260{
2261 struct nlmsghdr *nlh;
2262 struct rtmsg *rtm;
2263 int err;
2264
2265 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2266 if (nlh == NULL)
2267 return -EMSGSIZE;
2268
2269 rtm = nlmsg_data(nlh);
2270 rtm->rtm_family = RTNL_FAMILY_IPMR;
2271 rtm->rtm_dst_len = 32;
2272 rtm->rtm_src_len = 32;
2273 rtm->rtm_tos = 0;
2274 rtm->rtm_table = mrt->id;
2275 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2276 goto nla_put_failure;
2277 rtm->rtm_type = RTN_MULTICAST;
2278 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2279 if (c->mfc_flags & MFC_STATIC)
2280 rtm->rtm_protocol = RTPROT_STATIC;
2281 else
2282 rtm->rtm_protocol = RTPROT_MROUTED;
2283 rtm->rtm_flags = 0;
2284
2285 if (nla_put_be32(skb, RTA_SRC, c->mfc_origin) ||
2286 nla_put_be32(skb, RTA_DST, c->mfc_mcastgrp))
2287 goto nla_put_failure;
2288 err = __ipmr_fill_mroute(mrt, skb, c, rtm);
2289 /* do not break the dump if cache is unresolved */
2290 if (err < 0 && err != -ENOENT)
2291 goto nla_put_failure;
2292
2293 return nlmsg_end(skb, nlh);
2294
2295nla_put_failure:
2296 nlmsg_cancel(skb, nlh);
2297 return -EMSGSIZE;
2298}
2299
2300static size_t mroute_msgsize(bool unresolved, int maxvif)
2301{
2302 size_t len =
2303 NLMSG_ALIGN(sizeof(struct rtmsg))
2304 + nla_total_size(4) /* RTA_TABLE */
2305 + nla_total_size(4) /* RTA_SRC */
2306 + nla_total_size(4) /* RTA_DST */
2307 ;
2308
2309 if (!unresolved)
2310 len = len
2311 + nla_total_size(4) /* RTA_IIF */
2312 + nla_total_size(0) /* RTA_MULTIPATH */
2313 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2314 /* RTA_MFC_STATS */
2315 + nla_total_size(sizeof(struct rta_mfc_stats))
2316 ;
2317
2318 return len;
2319}
2320
2321static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2322 int cmd)
2323{
2324 struct net *net = read_pnet(&mrt->net);
2325 struct sk_buff *skb;
2326 int err = -ENOBUFS;
2327
2328 skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif),
2329 GFP_ATOMIC);
2330 if (skb == NULL)
2331 goto errout;
2332
2333 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2334 if (err < 0)
2335 goto errout;
2336
2337 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2338 return;
2339
2340errout:
2341 kfree_skb(skb);
2342 if (err < 0)
2343 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2344}
2345
2346static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2347{
2348 struct net *net = sock_net(skb->sk);
2349 struct mr_table *mrt;
2350 struct mfc_cache *mfc;
2351 unsigned int t = 0, s_t;
2352 unsigned int h = 0, s_h;
2353 unsigned int e = 0, s_e;
2354
2355 s_t = cb->args[0];
2356 s_h = cb->args[1];
2357 s_e = cb->args[2];
2358
2359 rcu_read_lock();
2360 ipmr_for_each_table(mrt, net) {
2361 if (t < s_t)
2362 goto next_table;
2363 if (t > s_t)
2364 s_h = 0;
2365 for (h = s_h; h < MFC_LINES; h++) {
2366 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_array[h], list) {
2367 if (e < s_e)
2368 goto next_entry;
2369 if (ipmr_fill_mroute(mrt, skb,
2370 NETLINK_CB(cb->skb).portid,
2371 cb->nlh->nlmsg_seq,
2372 mfc, RTM_NEWROUTE,
2373 NLM_F_MULTI) < 0)
2374 goto done;
2375next_entry:
2376 e++;
2377 }
2378 e = s_e = 0;
2379 }
2380 spin_lock_bh(&mfc_unres_lock);
2381 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
2382 if (e < s_e)
2383 goto next_entry2;
2384 if (ipmr_fill_mroute(mrt, skb,
2385 NETLINK_CB(cb->skb).portid,
2386 cb->nlh->nlmsg_seq,
2387 mfc, RTM_NEWROUTE,
2388 NLM_F_MULTI) < 0) {
2389 spin_unlock_bh(&mfc_unres_lock);
2390 goto done;
2391 }
2392next_entry2:
2393 e++;
2394 }
2395 spin_unlock_bh(&mfc_unres_lock);
2396 e = s_e = 0;
2397 s_h = 0;
2398next_table:
2399 t++;
2400 }
2401done:
2402 rcu_read_unlock();
2403
2404 cb->args[2] = e;
2405 cb->args[1] = h;
2406 cb->args[0] = t;
2407
2408 return skb->len;
2409}
2410
2411#ifdef CONFIG_PROC_FS
2412/*
2413 * The /proc interfaces to multicast routing :
2414 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2415 */
2416struct ipmr_vif_iter {
2417 struct seq_net_private p;
2418 struct mr_table *mrt;
2419 int ct;
2420};
2421
2422static struct vif_device *ipmr_vif_seq_idx(struct net *net,
2423 struct ipmr_vif_iter *iter,
2424 loff_t pos)
2425{
2426 struct mr_table *mrt = iter->mrt;
2427
2428 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
2429 if (!VIF_EXISTS(mrt, iter->ct))
2430 continue;
2431 if (pos-- == 0)
2432 return &mrt->vif_table[iter->ct];
2433 }
2434 return NULL;
2435}
2436
2437static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2438 __acquires(mrt_lock)
2439{
2440 struct ipmr_vif_iter *iter = seq->private;
2441 struct net *net = seq_file_net(seq);
2442 struct mr_table *mrt;
2443
2444 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2445 if (mrt == NULL)
2446 return ERR_PTR(-ENOENT);
2447
2448 iter->mrt = mrt;
2449
2450 read_lock(&mrt_lock);
2451 return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
2452 : SEQ_START_TOKEN;
2453}
2454
2455static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2456{
2457 struct ipmr_vif_iter *iter = seq->private;
2458 struct net *net = seq_file_net(seq);
2459 struct mr_table *mrt = iter->mrt;
2460
2461 ++*pos;
2462 if (v == SEQ_START_TOKEN)
2463 return ipmr_vif_seq_idx(net, iter, 0);
2464
2465 while (++iter->ct < mrt->maxvif) {
2466 if (!VIF_EXISTS(mrt, iter->ct))
2467 continue;
2468 return &mrt->vif_table[iter->ct];
2469 }
2470 return NULL;
2471}
2472
2473static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2474 __releases(mrt_lock)
2475{
2476 read_unlock(&mrt_lock);
2477}
2478
2479static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2480{
2481 struct ipmr_vif_iter *iter = seq->private;
2482 struct mr_table *mrt = iter->mrt;
2483
2484 if (v == SEQ_START_TOKEN) {
2485 seq_puts(seq,
2486 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2487 } else {
2488 const struct vif_device *vif = v;
2489 const char *name = vif->dev ? vif->dev->name : "none";
2490
2491 seq_printf(seq,
2492 "%2Zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2493 vif - mrt->vif_table,
2494 name, vif->bytes_in, vif->pkt_in,
2495 vif->bytes_out, vif->pkt_out,
2496 vif->flags, vif->local, vif->remote);
2497 }
2498 return 0;
2499}
2500
2501static const struct seq_operations ipmr_vif_seq_ops = {
2502 .start = ipmr_vif_seq_start,
2503 .next = ipmr_vif_seq_next,
2504 .stop = ipmr_vif_seq_stop,
2505 .show = ipmr_vif_seq_show,
2506};
2507
2508static int ipmr_vif_open(struct inode *inode, struct file *file)
2509{
2510 return seq_open_net(inode, file, &ipmr_vif_seq_ops,
2511 sizeof(struct ipmr_vif_iter));
2512}
2513
2514static const struct file_operations ipmr_vif_fops = {
2515 .owner = THIS_MODULE,
2516 .open = ipmr_vif_open,
2517 .read = seq_read,
2518 .llseek = seq_lseek,
2519 .release = seq_release_net,
2520};
2521
2522struct ipmr_mfc_iter {
2523 struct seq_net_private p;
2524 struct mr_table *mrt;
2525 struct list_head *cache;
2526 int ct;
2527};
2528
2529
2530static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
2531 struct ipmr_mfc_iter *it, loff_t pos)
2532{
2533 struct mr_table *mrt = it->mrt;
2534 struct mfc_cache *mfc;
2535
2536 rcu_read_lock();
2537 for (it->ct = 0; it->ct < MFC_LINES; it->ct++) {
2538 it->cache = &mrt->mfc_cache_array[it->ct];
2539 list_for_each_entry_rcu(mfc, it->cache, list)
2540 if (pos-- == 0)
2541 return mfc;
2542 }
2543 rcu_read_unlock();
2544
2545 spin_lock_bh(&mfc_unres_lock);
2546 it->cache = &mrt->mfc_unres_queue;
2547 list_for_each_entry(mfc, it->cache, list)
2548 if (pos-- == 0)
2549 return mfc;
2550 spin_unlock_bh(&mfc_unres_lock);
2551
2552 it->cache = NULL;
2553 return NULL;
2554}
2555
2556
2557static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2558{
2559 struct ipmr_mfc_iter *it = seq->private;
2560 struct net *net = seq_file_net(seq);
2561 struct mr_table *mrt;
2562
2563 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2564 if (mrt == NULL)
2565 return ERR_PTR(-ENOENT);
2566
2567 it->mrt = mrt;
2568 it->cache = NULL;
2569 it->ct = 0;
2570 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
2571 : SEQ_START_TOKEN;
2572}
2573
2574static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2575{
2576 struct mfc_cache *mfc = v;
2577 struct ipmr_mfc_iter *it = seq->private;
2578 struct net *net = seq_file_net(seq);
2579 struct mr_table *mrt = it->mrt;
2580
2581 ++*pos;
2582
2583 if (v == SEQ_START_TOKEN)
2584 return ipmr_mfc_seq_idx(net, seq->private, 0);
2585
2586 if (mfc->list.next != it->cache)
2587 return list_entry(mfc->list.next, struct mfc_cache, list);
2588
2589 if (it->cache == &mrt->mfc_unres_queue)
2590 goto end_of_list;
2591
2592 BUG_ON(it->cache != &mrt->mfc_cache_array[it->ct]);
2593
2594 while (++it->ct < MFC_LINES) {
2595 it->cache = &mrt->mfc_cache_array[it->ct];
2596 if (list_empty(it->cache))
2597 continue;
2598 return list_first_entry(it->cache, struct mfc_cache, list);
2599 }
2600
2601 /* exhausted cache_array, show unresolved */
2602 rcu_read_unlock();
2603 it->cache = &mrt->mfc_unres_queue;
2604 it->ct = 0;
2605
2606 spin_lock_bh(&mfc_unres_lock);
2607 if (!list_empty(it->cache))
2608 return list_first_entry(it->cache, struct mfc_cache, list);
2609
2610end_of_list:
2611 spin_unlock_bh(&mfc_unres_lock);
2612 it->cache = NULL;
2613
2614 return NULL;
2615}
2616
2617static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
2618{
2619 struct ipmr_mfc_iter *it = seq->private;
2620 struct mr_table *mrt = it->mrt;
2621
2622 if (it->cache == &mrt->mfc_unres_queue)
2623 spin_unlock_bh(&mfc_unres_lock);
2624 else if (it->cache == &mrt->mfc_cache_array[it->ct])
2625 rcu_read_unlock();
2626}
2627
2628static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2629{
2630 int n;
2631
2632 if (v == SEQ_START_TOKEN) {
2633 seq_puts(seq,
2634 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2635 } else {
2636 const struct mfc_cache *mfc = v;
2637 const struct ipmr_mfc_iter *it = seq->private;
2638 const struct mr_table *mrt = it->mrt;
2639
2640 seq_printf(seq, "%08X %08X %-3hd",
2641 (__force u32) mfc->mfc_mcastgrp,
2642 (__force u32) mfc->mfc_origin,
2643 mfc->mfc_parent);
2644
2645 if (it->cache != &mrt->mfc_unres_queue) {
2646 seq_printf(seq, " %8lu %8lu %8lu",
2647 mfc->mfc_un.res.pkt,
2648 mfc->mfc_un.res.bytes,
2649 mfc->mfc_un.res.wrong_if);
2650 for (n = mfc->mfc_un.res.minvif;
2651 n < mfc->mfc_un.res.maxvif; n++) {
2652 if (VIF_EXISTS(mrt, n) &&
2653 mfc->mfc_un.res.ttls[n] < 255)
2654 seq_printf(seq,
2655 " %2d:%-3d",
2656 n, mfc->mfc_un.res.ttls[n]);
2657 }
2658 } else {
2659 /* unresolved mfc_caches don't contain
2660 * pkt, bytes and wrong_if values
2661 */
2662 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2663 }
2664 seq_putc(seq, '\n');
2665 }
2666 return 0;
2667}
2668
2669static const struct seq_operations ipmr_mfc_seq_ops = {
2670 .start = ipmr_mfc_seq_start,
2671 .next = ipmr_mfc_seq_next,
2672 .stop = ipmr_mfc_seq_stop,
2673 .show = ipmr_mfc_seq_show,
2674};
2675
2676static int ipmr_mfc_open(struct inode *inode, struct file *file)
2677{
2678 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
2679 sizeof(struct ipmr_mfc_iter));
2680}
2681
2682static const struct file_operations ipmr_mfc_fops = {
2683 .owner = THIS_MODULE,
2684 .open = ipmr_mfc_open,
2685 .read = seq_read,
2686 .llseek = seq_lseek,
2687 .release = seq_release_net,
2688};
2689#endif
2690
2691#ifdef CONFIG_IP_PIMSM_V2
2692static const struct net_protocol pim_protocol = {
2693 .handler = pim_rcv,
2694 .netns_ok = 1,
2695};
2696#endif
2697
2698
2699/*
2700 * Setup for IP multicast routing
2701 */
2702static int __net_init ipmr_net_init(struct net *net)
2703{
2704 int err;
2705
2706 err = ipmr_rules_init(net);
2707 if (err < 0)
2708 goto fail;
2709
2710#ifdef CONFIG_PROC_FS
2711 err = -ENOMEM;
2712 if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops))
2713 goto proc_vif_fail;
2714 if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops))
2715 goto proc_cache_fail;
2716#endif
2717 return 0;
2718
2719#ifdef CONFIG_PROC_FS
2720proc_cache_fail:
2721 remove_proc_entry("ip_mr_vif", net->proc_net);
2722proc_vif_fail:
2723 ipmr_rules_exit(net);
2724#endif
2725fail:
2726 return err;
2727}
2728
2729static void __net_exit ipmr_net_exit(struct net *net)
2730{
2731#ifdef CONFIG_PROC_FS
2732 remove_proc_entry("ip_mr_cache", net->proc_net);
2733 remove_proc_entry("ip_mr_vif", net->proc_net);
2734#endif
2735 ipmr_rules_exit(net);
2736}
2737
2738static struct pernet_operations ipmr_net_ops = {
2739 .init = ipmr_net_init,
2740 .exit = ipmr_net_exit,
2741};
2742
2743int __init ip_mr_init(void)
2744{
2745 int err;
2746
2747 mrt_cachep = kmem_cache_create("ip_mrt_cache",
2748 sizeof(struct mfc_cache),
2749 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
2750 NULL);
2751 if (!mrt_cachep)
2752 return -ENOMEM;
2753
2754 err = register_pernet_subsys(&ipmr_net_ops);
2755 if (err)
2756 goto reg_pernet_fail;
2757
2758 err = register_netdevice_notifier(&ip_mr_notifier);
2759 if (err)
2760 goto reg_notif_fail;
2761#ifdef CONFIG_IP_PIMSM_V2
2762 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
2763 pr_err("%s: can't add PIM protocol\n", __func__);
2764 err = -EAGAIN;
2765 goto add_proto_fail;
2766 }
2767#endif
2768 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
2769 NULL, ipmr_rtm_dumproute, NULL);
2770 return 0;
2771
2772#ifdef CONFIG_IP_PIMSM_V2
2773add_proto_fail:
2774 unregister_netdevice_notifier(&ip_mr_notifier);
2775#endif
2776reg_notif_fail:
2777 unregister_pernet_subsys(&ipmr_net_ops);
2778reg_pernet_fail:
2779 kmem_cache_destroy(mrt_cachep);
2780 return err;
2781}