]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - net/ipv4/ipmr.c
projects / mirror_ubuntu-artful-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
iio: imu: inv_mpu6050: test whoami first and against all known values
[mirror_ubuntu-artful-kernel.git] / net / ipv4 / ipmr.c
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/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 #include <net/nexthop.h>
70
71 struct ipmr_rule {
72 struct fib_rule common;
73 };
74
75 struct 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
83 static DEFINE_RWLOCK(mrt_lock);
84
85 /* Multicast router control variables */
86
87 /* Special spinlock for queue of unresolved entries */
88 static 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
98 static struct kmem_cache *mrt_cachep __read_mostly;
99
100 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
101 static void ipmr_free_table(struct mr_table *mrt);
102
103 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
104 struct sk_buff *skb, struct mfc_cache *cache,
105 int local);
106 static int ipmr_cache_report(struct mr_table *mrt,
107 struct sk_buff *pkt, vifi_t vifi, int assert);
108 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
109 struct mfc_cache *c, struct rtmsg *rtm);
110 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
111 int cmd);
112 static void mroute_clean_tables(struct mr_table *mrt, bool all);
113 static void ipmr_expire_process(unsigned long arg);
114
115 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
116 #define ipmr_for_each_table(mrt, net) \
117 list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
118
119 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
120 {
121 struct mr_table *mrt;
122
123 ipmr_for_each_table(mrt, net) {
124 if (mrt->id == id)
125 return mrt;
126 }
127 return NULL;
128 }
129
130 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
131 struct mr_table **mrt)
132 {
133 int err;
134 struct ipmr_result res;
135 struct fib_lookup_arg arg = {
136 .result = &res,
137 .flags = FIB_LOOKUP_NOREF,
138 };
139
140 /* update flow if oif or iif point to device enslaved to l3mdev */
141 l3mdev_update_flow(net, flowi4_to_flowi(flp4));
142
143 err = fib_rules_lookup(net->ipv4.mr_rules_ops,
144 flowi4_to_flowi(flp4), 0, &arg);
145 if (err < 0)
146 return err;
147 *mrt = res.mrt;
148 return 0;
149 }
150
151 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
152 int flags, struct fib_lookup_arg *arg)
153 {
154 struct ipmr_result *res = arg->result;
155 struct mr_table *mrt;
156
157 switch (rule->action) {
158 case FR_ACT_TO_TBL:
159 break;
160 case FR_ACT_UNREACHABLE:
161 return -ENETUNREACH;
162 case FR_ACT_PROHIBIT:
163 return -EACCES;
164 case FR_ACT_BLACKHOLE:
165 default:
166 return -EINVAL;
167 }
168
169 arg->table = fib_rule_get_table(rule, arg);
170
171 mrt = ipmr_get_table(rule->fr_net, arg->table);
172 if (!mrt)
173 return -EAGAIN;
174 res->mrt = mrt;
175 return 0;
176 }
177
178 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
179 {
180 return 1;
181 }
182
183 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
184 FRA_GENERIC_POLICY,
185 };
186
187 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
188 struct fib_rule_hdr *frh, struct nlattr **tb)
189 {
190 return 0;
191 }
192
193 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
194 struct nlattr **tb)
195 {
196 return 1;
197 }
198
199 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
200 struct fib_rule_hdr *frh)
201 {
202 frh->dst_len = 0;
203 frh->src_len = 0;
204 frh->tos = 0;
205 return 0;
206 }
207
208 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
209 .family = RTNL_FAMILY_IPMR,
210 .rule_size = sizeof(struct ipmr_rule),
211 .addr_size = sizeof(u32),
212 .action = ipmr_rule_action,
213 .match = ipmr_rule_match,
214 .configure = ipmr_rule_configure,
215 .compare = ipmr_rule_compare,
216 .fill = ipmr_rule_fill,
217 .nlgroup = RTNLGRP_IPV4_RULE,
218 .policy = ipmr_rule_policy,
219 .owner = THIS_MODULE,
220 };
221
222 static int __net_init ipmr_rules_init(struct net *net)
223 {
224 struct fib_rules_ops *ops;
225 struct mr_table *mrt;
226 int err;
227
228 ops = fib_rules_register(&ipmr_rules_ops_template, net);
229 if (IS_ERR(ops))
230 return PTR_ERR(ops);
231
232 INIT_LIST_HEAD(&net->ipv4.mr_tables);
233
234 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
235 if (IS_ERR(mrt)) {
236 err = PTR_ERR(mrt);
237 goto err1;
238 }
239
240 err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
241 if (err < 0)
242 goto err2;
243
244 net->ipv4.mr_rules_ops = ops;
245 return 0;
246
247 err2:
248 ipmr_free_table(mrt);
249 err1:
250 fib_rules_unregister(ops);
251 return err;
252 }
253
254 static void __net_exit ipmr_rules_exit(struct net *net)
255 {
256 struct mr_table *mrt, *next;
257
258 rtnl_lock();
259 list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
260 list_del(&mrt->list);
261 ipmr_free_table(mrt);
262 }
263 fib_rules_unregister(net->ipv4.mr_rules_ops);
264 rtnl_unlock();
265 }
266 #else
267 #define ipmr_for_each_table(mrt, net) \
268 for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
269
270 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
271 {
272 return net->ipv4.mrt;
273 }
274
275 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
276 struct mr_table **mrt)
277 {
278 *mrt = net->ipv4.mrt;
279 return 0;
280 }
281
282 static int __net_init ipmr_rules_init(struct net *net)
283 {
284 struct mr_table *mrt;
285
286 mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
287 if (IS_ERR(mrt))
288 return PTR_ERR(mrt);
289 net->ipv4.mrt = mrt;
290 return 0;
291 }
292
293 static void __net_exit ipmr_rules_exit(struct net *net)
294 {
295 rtnl_lock();
296 ipmr_free_table(net->ipv4.mrt);
297 net->ipv4.mrt = NULL;
298 rtnl_unlock();
299 }
300 #endif
301
302 static inline int ipmr_hash_cmp(struct rhashtable_compare_arg *arg,
303 const void *ptr)
304 {
305 const struct mfc_cache_cmp_arg *cmparg = arg->key;
306 struct mfc_cache *c = (struct mfc_cache *)ptr;
307
308 return cmparg->mfc_mcastgrp != c->mfc_mcastgrp ||
309 cmparg->mfc_origin != c->mfc_origin;
310 }
311
312 static const struct rhashtable_params ipmr_rht_params = {
313 .head_offset = offsetof(struct mfc_cache, mnode),
314 .key_offset = offsetof(struct mfc_cache, cmparg),
315 .key_len = sizeof(struct mfc_cache_cmp_arg),
316 .nelem_hint = 3,
317 .locks_mul = 1,
318 .obj_cmpfn = ipmr_hash_cmp,
319 .automatic_shrinking = true,
320 };
321
322 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
323 {
324 struct mr_table *mrt;
325
326 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
327 if (id != RT_TABLE_DEFAULT && id >= 1000000000)
328 return ERR_PTR(-EINVAL);
329
330 mrt = ipmr_get_table(net, id);
331 if (mrt)
332 return mrt;
333
334 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
335 if (!mrt)
336 return ERR_PTR(-ENOMEM);
337 write_pnet(&mrt->net, net);
338 mrt->id = id;
339
340 rhltable_init(&mrt->mfc_hash, &ipmr_rht_params);
341 INIT_LIST_HEAD(&mrt->mfc_cache_list);
342 INIT_LIST_HEAD(&mrt->mfc_unres_queue);
343
344 setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
345 (unsigned long)mrt);
346
347 mrt->mroute_reg_vif_num = -1;
348 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
349 list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
350 #endif
351 return mrt;
352 }
353
354 static void ipmr_free_table(struct mr_table *mrt)
355 {
356 del_timer_sync(&mrt->ipmr_expire_timer);
357 mroute_clean_tables(mrt, true);
358 rhltable_destroy(&mrt->mfc_hash);
359 kfree(mrt);
360 }
361
362 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
363
364 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
365 {
366 struct net *net = dev_net(dev);
367
368 dev_close(dev);
369
370 dev = __dev_get_by_name(net, "tunl0");
371 if (dev) {
372 const struct net_device_ops *ops = dev->netdev_ops;
373 struct ifreq ifr;
374 struct ip_tunnel_parm p;
375
376 memset(&p, 0, sizeof(p));
377 p.iph.daddr = v->vifc_rmt_addr.s_addr;
378 p.iph.saddr = v->vifc_lcl_addr.s_addr;
379 p.iph.version = 4;
380 p.iph.ihl = 5;
381 p.iph.protocol = IPPROTO_IPIP;
382 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
383 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
384
385 if (ops->ndo_do_ioctl) {
386 mm_segment_t oldfs = get_fs();
387
388 set_fs(KERNEL_DS);
389 ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
390 set_fs(oldfs);
391 }
392 }
393 }
394
395 /* Initialize ipmr pimreg/tunnel in_device */
396 static bool ipmr_init_vif_indev(const struct net_device *dev)
397 {
398 struct in_device *in_dev;
399
400 ASSERT_RTNL();
401
402 in_dev = __in_dev_get_rtnl(dev);
403 if (!in_dev)
404 return false;
405 ipv4_devconf_setall(in_dev);
406 neigh_parms_data_state_setall(in_dev->arp_parms);
407 IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
408
409 return true;
410 }
411
412 static struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
413 {
414 struct net_device *dev;
415
416 dev = __dev_get_by_name(net, "tunl0");
417
418 if (dev) {
419 const struct net_device_ops *ops = dev->netdev_ops;
420 int err;
421 struct ifreq ifr;
422 struct ip_tunnel_parm p;
423
424 memset(&p, 0, sizeof(p));
425 p.iph.daddr = v->vifc_rmt_addr.s_addr;
426 p.iph.saddr = v->vifc_lcl_addr.s_addr;
427 p.iph.version = 4;
428 p.iph.ihl = 5;
429 p.iph.protocol = IPPROTO_IPIP;
430 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
431 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
432
433 if (ops->ndo_do_ioctl) {
434 mm_segment_t oldfs = get_fs();
435
436 set_fs(KERNEL_DS);
437 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
438 set_fs(oldfs);
439 } else {
440 err = -EOPNOTSUPP;
441 }
442 dev = NULL;
443
444 if (err == 0 &&
445 (dev = __dev_get_by_name(net, p.name)) != NULL) {
446 dev->flags |= IFF_MULTICAST;
447 if (!ipmr_init_vif_indev(dev))
448 goto failure;
449 if (dev_open(dev))
450 goto failure;
451 dev_hold(dev);
452 }
453 }
454 return dev;
455
456 failure:
457 unregister_netdevice(dev);
458 return NULL;
459 }
460
461 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
462 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
463 {
464 struct net *net = dev_net(dev);
465 struct mr_table *mrt;
466 struct flowi4 fl4 = {
467 .flowi4_oif = dev->ifindex,
468 .flowi4_iif = skb->skb_iif ? : LOOPBACK_IFINDEX,
469 .flowi4_mark = skb->mark,
470 };
471 int err;
472
473 err = ipmr_fib_lookup(net, &fl4, &mrt);
474 if (err < 0) {
475 kfree_skb(skb);
476 return err;
477 }
478
479 read_lock(&mrt_lock);
480 dev->stats.tx_bytes += skb->len;
481 dev->stats.tx_packets++;
482 ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
483 read_unlock(&mrt_lock);
484 kfree_skb(skb);
485 return NETDEV_TX_OK;
486 }
487
488 static int reg_vif_get_iflink(const struct net_device *dev)
489 {
490 return 0;
491 }
492
493 static const struct net_device_ops reg_vif_netdev_ops = {
494 .ndo_start_xmit = reg_vif_xmit,
495 .ndo_get_iflink = reg_vif_get_iflink,
496 };
497
498 static void reg_vif_setup(struct net_device *dev)
499 {
500 dev->type = ARPHRD_PIMREG;
501 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
502 dev->flags = IFF_NOARP;
503 dev->netdev_ops = &reg_vif_netdev_ops;
504 dev->destructor = free_netdev;
505 dev->features |= NETIF_F_NETNS_LOCAL;
506 }
507
508 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
509 {
510 struct net_device *dev;
511 char name[IFNAMSIZ];
512
513 if (mrt->id == RT_TABLE_DEFAULT)
514 sprintf(name, "pimreg");
515 else
516 sprintf(name, "pimreg%u", mrt->id);
517
518 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, reg_vif_setup);
519
520 if (!dev)
521 return NULL;
522
523 dev_net_set(dev, net);
524
525 if (register_netdevice(dev)) {
526 free_netdev(dev);
527 return NULL;
528 }
529
530 if (!ipmr_init_vif_indev(dev))
531 goto failure;
532 if (dev_open(dev))
533 goto failure;
534
535 dev_hold(dev);
536
537 return dev;
538
539 failure:
540 unregister_netdevice(dev);
541 return NULL;
542 }
543
544 /* called with rcu_read_lock() */
545 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
546 unsigned int pimlen)
547 {
548 struct net_device *reg_dev = NULL;
549 struct iphdr *encap;
550
551 encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
552 /* Check that:
553 * a. packet is really sent to a multicast group
554 * b. packet is not a NULL-REGISTER
555 * c. packet is not truncated
556 */
557 if (!ipv4_is_multicast(encap->daddr) ||
558 encap->tot_len == 0 ||
559 ntohs(encap->tot_len) + pimlen > skb->len)
560 return 1;
561
562 read_lock(&mrt_lock);
563 if (mrt->mroute_reg_vif_num >= 0)
564 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
565 read_unlock(&mrt_lock);
566
567 if (!reg_dev)
568 return 1;
569
570 skb->mac_header = skb->network_header;
571 skb_pull(skb, (u8 *)encap - skb->data);
572 skb_reset_network_header(skb);
573 skb->protocol = htons(ETH_P_IP);
574 skb->ip_summed = CHECKSUM_NONE;
575
576 skb_tunnel_rx(skb, reg_dev, dev_net(reg_dev));
577
578 netif_rx(skb);
579
580 return NET_RX_SUCCESS;
581 }
582 #else
583 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
584 {
585 return NULL;
586 }
587 #endif
588
589 /**
590 * vif_delete - Delete a VIF entry
591 * @notify: Set to 1, if the caller is a notifier_call
592 */
593 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
594 struct list_head *head)
595 {
596 struct vif_device *v;
597 struct net_device *dev;
598 struct in_device *in_dev;
599
600 if (vifi < 0 || vifi >= mrt->maxvif)
601 return -EADDRNOTAVAIL;
602
603 v = &mrt->vif_table[vifi];
604
605 write_lock_bh(&mrt_lock);
606 dev = v->dev;
607 v->dev = NULL;
608
609 if (!dev) {
610 write_unlock_bh(&mrt_lock);
611 return -EADDRNOTAVAIL;
612 }
613
614 if (vifi == mrt->mroute_reg_vif_num)
615 mrt->mroute_reg_vif_num = -1;
616
617 if (vifi + 1 == mrt->maxvif) {
618 int tmp;
619
620 for (tmp = vifi - 1; tmp >= 0; tmp--) {
621 if (VIF_EXISTS(mrt, tmp))
622 break;
623 }
624 mrt->maxvif = tmp+1;
625 }
626
627 write_unlock_bh(&mrt_lock);
628
629 dev_set_allmulti(dev, -1);
630
631 in_dev = __in_dev_get_rtnl(dev);
632 if (in_dev) {
633 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
634 inet_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
635 NETCONFA_MC_FORWARDING,
636 dev->ifindex, &in_dev->cnf);
637 ip_rt_multicast_event(in_dev);
638 }
639
640 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
641 unregister_netdevice_queue(dev, head);
642
643 dev_put(dev);
644 return 0;
645 }
646
647 static void ipmr_cache_free_rcu(struct rcu_head *head)
648 {
649 struct mfc_cache *c = container_of(head, struct mfc_cache, rcu);
650
651 kmem_cache_free(mrt_cachep, c);
652 }
653
654 static inline void ipmr_cache_free(struct mfc_cache *c)
655 {
656 call_rcu(&c->rcu, ipmr_cache_free_rcu);
657 }
658
659 /* Destroy an unresolved cache entry, killing queued skbs
660 * and reporting error to netlink readers.
661 */
662 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
663 {
664 struct net *net = read_pnet(&mrt->net);
665 struct sk_buff *skb;
666 struct nlmsgerr *e;
667
668 atomic_dec(&mrt->cache_resolve_queue_len);
669
670 while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) {
671 if (ip_hdr(skb)->version == 0) {
672 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
673 nlh->nlmsg_type = NLMSG_ERROR;
674 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
675 skb_trim(skb, nlh->nlmsg_len);
676 e = nlmsg_data(nlh);
677 e->error = -ETIMEDOUT;
678 memset(&e->msg, 0, sizeof(e->msg));
679
680 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
681 } else {
682 kfree_skb(skb);
683 }
684 }
685
686 ipmr_cache_free(c);
687 }
688
689 /* Timer process for the unresolved queue. */
690 static void ipmr_expire_process(unsigned long arg)
691 {
692 struct mr_table *mrt = (struct mr_table *)arg;
693 unsigned long now;
694 unsigned long expires;
695 struct mfc_cache *c, *next;
696
697 if (!spin_trylock(&mfc_unres_lock)) {
698 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
699 return;
700 }
701
702 if (list_empty(&mrt->mfc_unres_queue))
703 goto out;
704
705 now = jiffies;
706 expires = 10*HZ;
707
708 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
709 if (time_after(c->mfc_un.unres.expires, now)) {
710 unsigned long interval = c->mfc_un.unres.expires - now;
711 if (interval < expires)
712 expires = interval;
713 continue;
714 }
715
716 list_del(&c->list);
717 mroute_netlink_event(mrt, c, RTM_DELROUTE);
718 ipmr_destroy_unres(mrt, c);
719 }
720
721 if (!list_empty(&mrt->mfc_unres_queue))
722 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
723
724 out:
725 spin_unlock(&mfc_unres_lock);
726 }
727
728 /* Fill oifs list. It is called under write locked mrt_lock. */
729 static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache,
730 unsigned char *ttls)
731 {
732 int vifi;
733
734 cache->mfc_un.res.minvif = MAXVIFS;
735 cache->mfc_un.res.maxvif = 0;
736 memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
737
738 for (vifi = 0; vifi < mrt->maxvif; vifi++) {
739 if (VIF_EXISTS(mrt, vifi) &&
740 ttls[vifi] && ttls[vifi] < 255) {
741 cache->mfc_un.res.ttls[vifi] = ttls[vifi];
742 if (cache->mfc_un.res.minvif > vifi)
743 cache->mfc_un.res.minvif = vifi;
744 if (cache->mfc_un.res.maxvif <= vifi)
745 cache->mfc_un.res.maxvif = vifi + 1;
746 }
747 }
748 cache->mfc_un.res.lastuse = jiffies;
749 }
750
751 static int vif_add(struct net *net, struct mr_table *mrt,
752 struct vifctl *vifc, int mrtsock)
753 {
754 int vifi = vifc->vifc_vifi;
755 struct vif_device *v = &mrt->vif_table[vifi];
756 struct net_device *dev;
757 struct in_device *in_dev;
758 int err;
759
760 /* Is vif busy ? */
761 if (VIF_EXISTS(mrt, vifi))
762 return -EADDRINUSE;
763
764 switch (vifc->vifc_flags) {
765 case VIFF_REGISTER:
766 if (!ipmr_pimsm_enabled())
767 return -EINVAL;
768 /* Special Purpose VIF in PIM
769 * All the packets will be sent to the daemon
770 */
771 if (mrt->mroute_reg_vif_num >= 0)
772 return -EADDRINUSE;
773 dev = ipmr_reg_vif(net, mrt);
774 if (!dev)
775 return -ENOBUFS;
776 err = dev_set_allmulti(dev, 1);
777 if (err) {
778 unregister_netdevice(dev);
779 dev_put(dev);
780 return err;
781 }
782 break;
783 case VIFF_TUNNEL:
784 dev = ipmr_new_tunnel(net, vifc);
785 if (!dev)
786 return -ENOBUFS;
787 err = dev_set_allmulti(dev, 1);
788 if (err) {
789 ipmr_del_tunnel(dev, vifc);
790 dev_put(dev);
791 return err;
792 }
793 break;
794 case VIFF_USE_IFINDEX:
795 case 0:
796 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
797 dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
798 if (dev && !__in_dev_get_rtnl(dev)) {
799 dev_put(dev);
800 return -EADDRNOTAVAIL;
801 }
802 } else {
803 dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
804 }
805 if (!dev)
806 return -EADDRNOTAVAIL;
807 err = dev_set_allmulti(dev, 1);
808 if (err) {
809 dev_put(dev);
810 return err;
811 }
812 break;
813 default:
814 return -EINVAL;
815 }
816
817 in_dev = __in_dev_get_rtnl(dev);
818 if (!in_dev) {
819 dev_put(dev);
820 return -EADDRNOTAVAIL;
821 }
822 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
823 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_MC_FORWARDING,
824 dev->ifindex, &in_dev->cnf);
825 ip_rt_multicast_event(in_dev);
826
827 /* Fill in the VIF structures */
828
829 v->rate_limit = vifc->vifc_rate_limit;
830 v->local = vifc->vifc_lcl_addr.s_addr;
831 v->remote = vifc->vifc_rmt_addr.s_addr;
832 v->flags = vifc->vifc_flags;
833 if (!mrtsock)
834 v->flags |= VIFF_STATIC;
835 v->threshold = vifc->vifc_threshold;
836 v->bytes_in = 0;
837 v->bytes_out = 0;
838 v->pkt_in = 0;
839 v->pkt_out = 0;
840 v->link = dev->ifindex;
841 if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER))
842 v->link = dev_get_iflink(dev);
843
844 /* And finish update writing critical data */
845 write_lock_bh(&mrt_lock);
846 v->dev = dev;
847 if (v->flags & VIFF_REGISTER)
848 mrt->mroute_reg_vif_num = vifi;
849 if (vifi+1 > mrt->maxvif)
850 mrt->maxvif = vifi+1;
851 write_unlock_bh(&mrt_lock);
852 return 0;
853 }
854
855 /* called with rcu_read_lock() */
856 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
857 __be32 origin,
858 __be32 mcastgrp)
859 {
860 struct mfc_cache_cmp_arg arg = {
861 .mfc_mcastgrp = mcastgrp,
862 .mfc_origin = origin
863 };
864 struct rhlist_head *tmp, *list;
865 struct mfc_cache *c;
866
867 list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params);
868 rhl_for_each_entry_rcu(c, tmp, list, mnode)
869 return c;
870
871 return NULL;
872 }
873
874 /* Look for a (*,*,oif) entry */
875 static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt,
876 int vifi)
877 {
878 struct mfc_cache_cmp_arg arg = {
879 .mfc_mcastgrp = htonl(INADDR_ANY),
880 .mfc_origin = htonl(INADDR_ANY)
881 };
882 struct rhlist_head *tmp, *list;
883 struct mfc_cache *c;
884
885 list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params);
886 rhl_for_each_entry_rcu(c, tmp, list, mnode)
887 if (c->mfc_un.res.ttls[vifi] < 255)
888 return c;
889
890 return NULL;
891 }
892
893 /* Look for a (*,G) entry */
894 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
895 __be32 mcastgrp, int vifi)
896 {
897 struct mfc_cache_cmp_arg arg = {
898 .mfc_mcastgrp = mcastgrp,
899 .mfc_origin = htonl(INADDR_ANY)
900 };
901 struct rhlist_head *tmp, *list;
902 struct mfc_cache *c, *proxy;
903
904 if (mcastgrp == htonl(INADDR_ANY))
905 goto skip;
906
907 list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params);
908 rhl_for_each_entry_rcu(c, tmp, list, mnode) {
909 if (c->mfc_un.res.ttls[vifi] < 255)
910 return c;
911
912 /* It's ok if the vifi is part of the static tree */
913 proxy = ipmr_cache_find_any_parent(mrt, c->mfc_parent);
914 if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
915 return c;
916 }
917
918 skip:
919 return ipmr_cache_find_any_parent(mrt, vifi);
920 }
921
922 /* Look for a (S,G,iif) entry if parent != -1 */
923 static struct mfc_cache *ipmr_cache_find_parent(struct mr_table *mrt,
924 __be32 origin, __be32 mcastgrp,
925 int parent)
926 {
927 struct mfc_cache_cmp_arg arg = {
928 .mfc_mcastgrp = mcastgrp,
929 .mfc_origin = origin,
930 };
931 struct rhlist_head *tmp, *list;
932 struct mfc_cache *c;
933
934 list = rhltable_lookup(&mrt->mfc_hash, &arg, ipmr_rht_params);
935 rhl_for_each_entry_rcu(c, tmp, list, mnode)
936 if (parent == -1 || parent == c->mfc_parent)
937 return c;
938
939 return NULL;
940 }
941
942 /* Allocate a multicast cache entry */
943 static struct mfc_cache *ipmr_cache_alloc(void)
944 {
945 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
946
947 if (c) {
948 c->mfc_un.res.last_assert = jiffies - MFC_ASSERT_THRESH - 1;
949 c->mfc_un.res.minvif = MAXVIFS;
950 }
951 return c;
952 }
953
954 static struct mfc_cache *ipmr_cache_alloc_unres(void)
955 {
956 struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
957
958 if (c) {
959 skb_queue_head_init(&c->mfc_un.unres.unresolved);
960 c->mfc_un.unres.expires = jiffies + 10*HZ;
961 }
962 return c;
963 }
964
965 /* A cache entry has gone into a resolved state from queued */
966 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
967 struct mfc_cache *uc, struct mfc_cache *c)
968 {
969 struct sk_buff *skb;
970 struct nlmsgerr *e;
971
972 /* Play the pending entries through our router */
973 while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
974 if (ip_hdr(skb)->version == 0) {
975 struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
976
977 if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
978 nlh->nlmsg_len = skb_tail_pointer(skb) -
979 (u8 *)nlh;
980 } else {
981 nlh->nlmsg_type = NLMSG_ERROR;
982 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
983 skb_trim(skb, nlh->nlmsg_len);
984 e = nlmsg_data(nlh);
985 e->error = -EMSGSIZE;
986 memset(&e->msg, 0, sizeof(e->msg));
987 }
988
989 rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
990 } else {
991 ip_mr_forward(net, mrt, skb, c, 0);
992 }
993 }
994 }
995
996 /* Bounce a cache query up to mrouted. We could use netlink for this but mrouted
997 * expects the following bizarre scheme.
998 *
999 * Called under mrt_lock.
1000 */
1001 static int ipmr_cache_report(struct mr_table *mrt,
1002 struct sk_buff *pkt, vifi_t vifi, int assert)
1003 {
1004 const int ihl = ip_hdrlen(pkt);
1005 struct sock *mroute_sk;
1006 struct igmphdr *igmp;
1007 struct igmpmsg *msg;
1008 struct sk_buff *skb;
1009 int ret;
1010
1011 if (assert == IGMPMSG_WHOLEPKT)
1012 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
1013 else
1014 skb = alloc_skb(128, GFP_ATOMIC);
1015
1016 if (!skb)
1017 return -ENOBUFS;
1018
1019 if (assert == IGMPMSG_WHOLEPKT) {
1020 /* Ugly, but we have no choice with this interface.
1021 * Duplicate old header, fix ihl, length etc.
1022 * And all this only to mangle msg->im_msgtype and
1023 * to set msg->im_mbz to "mbz" :-)
1024 */
1025 skb_push(skb, sizeof(struct iphdr));
1026 skb_reset_network_header(skb);
1027 skb_reset_transport_header(skb);
1028 msg = (struct igmpmsg *)skb_network_header(skb);
1029 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
1030 msg->im_msgtype = IGMPMSG_WHOLEPKT;
1031 msg->im_mbz = 0;
1032 msg->im_vif = mrt->mroute_reg_vif_num;
1033 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
1034 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
1035 sizeof(struct iphdr));
1036 } else {
1037 /* Copy the IP header */
1038 skb_set_network_header(skb, skb->len);
1039 skb_put(skb, ihl);
1040 skb_copy_to_linear_data(skb, pkt->data, ihl);
1041 /* Flag to the kernel this is a route add */
1042 ip_hdr(skb)->protocol = 0;
1043 msg = (struct igmpmsg *)skb_network_header(skb);
1044 msg->im_vif = vifi;
1045 skb_dst_set(skb, dst_clone(skb_dst(pkt)));
1046 /* Add our header */
1047 igmp = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
1048 igmp->type = assert;
1049 msg->im_msgtype = assert;
1050 igmp->code = 0;
1051 ip_hdr(skb)->tot_len = htons(skb->len); /* Fix the length */
1052 skb->transport_header = skb->network_header;
1053 }
1054
1055 rcu_read_lock();
1056 mroute_sk = rcu_dereference(mrt->mroute_sk);
1057 if (!mroute_sk) {
1058 rcu_read_unlock();
1059 kfree_skb(skb);
1060 return -EINVAL;
1061 }
1062
1063 /* Deliver to mrouted */
1064 ret = sock_queue_rcv_skb(mroute_sk, skb);
1065 rcu_read_unlock();
1066 if (ret < 0) {
1067 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1068 kfree_skb(skb);
1069 }
1070
1071 return ret;
1072 }
1073
1074 /* Queue a packet for resolution. It gets locked cache entry! */
1075 static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
1076 struct sk_buff *skb)
1077 {
1078 const struct iphdr *iph = ip_hdr(skb);
1079 struct mfc_cache *c;
1080 bool found = false;
1081 int err;
1082
1083 spin_lock_bh(&mfc_unres_lock);
1084 list_for_each_entry(c, &mrt->mfc_unres_queue, list) {
1085 if (c->mfc_mcastgrp == iph->daddr &&
1086 c->mfc_origin == iph->saddr) {
1087 found = true;
1088 break;
1089 }
1090 }
1091
1092 if (!found) {
1093 /* Create a new entry if allowable */
1094 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1095 (c = ipmr_cache_alloc_unres()) == NULL) {
1096 spin_unlock_bh(&mfc_unres_lock);
1097
1098 kfree_skb(skb);
1099 return -ENOBUFS;
1100 }
1101
1102 /* Fill in the new cache entry */
1103 c->mfc_parent = -1;
1104 c->mfc_origin = iph->saddr;
1105 c->mfc_mcastgrp = iph->daddr;
1106
1107 /* Reflect first query at mrouted. */
1108 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1109 if (err < 0) {
1110 /* If the report failed throw the cache entry
1111 out - Brad Parker
1112 */
1113 spin_unlock_bh(&mfc_unres_lock);
1114
1115 ipmr_cache_free(c);
1116 kfree_skb(skb);
1117 return err;
1118 }
1119
1120 atomic_inc(&mrt->cache_resolve_queue_len);
1121 list_add(&c->list, &mrt->mfc_unres_queue);
1122 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1123
1124 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1125 mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires);
1126 }
1127
1128 /* See if we can append the packet */
1129 if (c->mfc_un.unres.unresolved.qlen > 3) {
1130 kfree_skb(skb);
1131 err = -ENOBUFS;
1132 } else {
1133 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1134 err = 0;
1135 }
1136
1137 spin_unlock_bh(&mfc_unres_lock);
1138 return err;
1139 }
1140
1141 /* MFC cache manipulation by user space mroute daemon */
1142
1143 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1144 {
1145 struct mfc_cache *c;
1146
1147 /* The entries are added/deleted only under RTNL */
1148 rcu_read_lock();
1149 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1150 mfc->mfcc_mcastgrp.s_addr, parent);
1151 rcu_read_unlock();
1152 if (!c)
1153 return -ENOENT;
1154 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1155 list_del_rcu(&c->list);
1156 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1157 ipmr_cache_free(c);
1158
1159 return 0;
1160 }
1161
1162 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1163 struct mfcctl *mfc, int mrtsock, int parent)
1164 {
1165 struct mfc_cache *uc, *c;
1166 bool found;
1167 int ret;
1168
1169 if (mfc->mfcc_parent >= MAXVIFS)
1170 return -ENFILE;
1171
1172 /* The entries are added/deleted only under RTNL */
1173 rcu_read_lock();
1174 c = ipmr_cache_find_parent(mrt, mfc->mfcc_origin.s_addr,
1175 mfc->mfcc_mcastgrp.s_addr, parent);
1176 rcu_read_unlock();
1177 if (c) {
1178 write_lock_bh(&mrt_lock);
1179 c->mfc_parent = mfc->mfcc_parent;
1180 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1181 if (!mrtsock)
1182 c->mfc_flags |= MFC_STATIC;
1183 write_unlock_bh(&mrt_lock);
1184 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1185 return 0;
1186 }
1187
1188 if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1189 !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1190 return -EINVAL;
1191
1192 c = ipmr_cache_alloc();
1193 if (!c)
1194 return -ENOMEM;
1195
1196 c->mfc_origin = mfc->mfcc_origin.s_addr;
1197 c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1198 c->mfc_parent = mfc->mfcc_parent;
1199 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1200 if (!mrtsock)
1201 c->mfc_flags |= MFC_STATIC;
1202
1203 ret = rhltable_insert_key(&mrt->mfc_hash, &c->cmparg, &c->mnode,
1204 ipmr_rht_params);
1205 if (ret) {
1206 pr_err("ipmr: rhtable insert error %d\n", ret);
1207 ipmr_cache_free(c);
1208 return ret;
1209 }
1210 list_add_tail_rcu(&c->list, &mrt->mfc_cache_list);
1211 /* Check to see if we resolved a queued list. If so we
1212 * need to send on the frames and tidy up.
1213 */
1214 found = false;
1215 spin_lock_bh(&mfc_unres_lock);
1216 list_for_each_entry(uc, &mrt->mfc_unres_queue, list) {
1217 if (uc->mfc_origin == c->mfc_origin &&
1218 uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1219 list_del(&uc->list);
1220 atomic_dec(&mrt->cache_resolve_queue_len);
1221 found = true;
1222 break;
1223 }
1224 }
1225 if (list_empty(&mrt->mfc_unres_queue))
1226 del_timer(&mrt->ipmr_expire_timer);
1227 spin_unlock_bh(&mfc_unres_lock);
1228
1229 if (found) {
1230 ipmr_cache_resolve(net, mrt, uc, c);
1231 ipmr_cache_free(uc);
1232 }
1233 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1234 return 0;
1235 }
1236
1237 /* Close the multicast socket, and clear the vif tables etc */
1238 static void mroute_clean_tables(struct mr_table *mrt, bool all)
1239 {
1240 struct mfc_cache *c, *tmp;
1241 LIST_HEAD(list);
1242 int i;
1243
1244 /* Shut down all active vif entries */
1245 for (i = 0; i < mrt->maxvif; i++) {
1246 if (!all && (mrt->vif_table[i].flags & VIFF_STATIC))
1247 continue;
1248 vif_delete(mrt, i, 0, &list);
1249 }
1250 unregister_netdevice_many(&list);
1251
1252 /* Wipe the cache */
1253 list_for_each_entry_safe(c, tmp, &mrt->mfc_cache_list, list) {
1254 if (!all && (c->mfc_flags & MFC_STATIC))
1255 continue;
1256 rhltable_remove(&mrt->mfc_hash, &c->mnode, ipmr_rht_params);
1257 list_del_rcu(&c->list);
1258 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1259 ipmr_cache_free(c);
1260 }
1261
1262 if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1263 spin_lock_bh(&mfc_unres_lock);
1264 list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
1265 list_del(&c->list);
1266 mroute_netlink_event(mrt, c, RTM_DELROUTE);
1267 ipmr_destroy_unres(mrt, c);
1268 }
1269 spin_unlock_bh(&mfc_unres_lock);
1270 }
1271 }
1272
1273 /* called from ip_ra_control(), before an RCU grace period,
1274 * we dont need to call synchronize_rcu() here
1275 */
1276 static void mrtsock_destruct(struct sock *sk)
1277 {
1278 struct net *net = sock_net(sk);
1279 struct mr_table *mrt;
1280
1281 ASSERT_RTNL();
1282 ipmr_for_each_table(mrt, net) {
1283 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1284 IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1285 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1286 NETCONFA_MC_FORWARDING,
1287 NETCONFA_IFINDEX_ALL,
1288 net->ipv4.devconf_all);
1289 RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1290 mroute_clean_tables(mrt, false);
1291 }
1292 }
1293 }
1294
1295 /* Socket options and virtual interface manipulation. The whole
1296 * virtual interface system is a complete heap, but unfortunately
1297 * that's how BSD mrouted happens to think. Maybe one day with a proper
1298 * MOSPF/PIM router set up we can clean this up.
1299 */
1300
1301 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval,
1302 unsigned int optlen)
1303 {
1304 struct net *net = sock_net(sk);
1305 int val, ret = 0, parent = 0;
1306 struct mr_table *mrt;
1307 struct vifctl vif;
1308 struct mfcctl mfc;
1309 u32 uval;
1310
1311 /* There's one exception to the lock - MRT_DONE which needs to unlock */
1312 rtnl_lock();
1313 if (sk->sk_type != SOCK_RAW ||
1314 inet_sk(sk)->inet_num != IPPROTO_IGMP) {
1315 ret = -EOPNOTSUPP;
1316 goto out_unlock;
1317 }
1318
1319 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1320 if (!mrt) {
1321 ret = -ENOENT;
1322 goto out_unlock;
1323 }
1324 if (optname != MRT_INIT) {
1325 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1326 !ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1327 ret = -EACCES;
1328 goto out_unlock;
1329 }
1330 }
1331
1332 switch (optname) {
1333 case MRT_INIT:
1334 if (optlen != sizeof(int)) {
1335 ret = -EINVAL;
1336 break;
1337 }
1338 if (rtnl_dereference(mrt->mroute_sk)) {
1339 ret = -EADDRINUSE;
1340 break;
1341 }
1342
1343 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1344 if (ret == 0) {
1345 rcu_assign_pointer(mrt->mroute_sk, sk);
1346 IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1347 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
1348 NETCONFA_MC_FORWARDING,
1349 NETCONFA_IFINDEX_ALL,
1350 net->ipv4.devconf_all);
1351 }
1352 break;
1353 case MRT_DONE:
1354 if (sk != rcu_access_pointer(mrt->mroute_sk)) {
1355 ret = -EACCES;
1356 } else {
1357 ret = ip_ra_control(sk, 0, NULL);
1358 goto out_unlock;
1359 }
1360 break;
1361 case MRT_ADD_VIF:
1362 case MRT_DEL_VIF:
1363 if (optlen != sizeof(vif)) {
1364 ret = -EINVAL;
1365 break;
1366 }
1367 if (copy_from_user(&vif, optval, sizeof(vif))) {
1368 ret = -EFAULT;
1369 break;
1370 }
1371 if (vif.vifc_vifi >= MAXVIFS) {
1372 ret = -ENFILE;
1373 break;
1374 }
1375 if (optname == MRT_ADD_VIF) {
1376 ret = vif_add(net, mrt, &vif,
1377 sk == rtnl_dereference(mrt->mroute_sk));
1378 } else {
1379 ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1380 }
1381 break;
1382 /* Manipulate the forwarding caches. These live
1383 * in a sort of kernel/user symbiosis.
1384 */
1385 case MRT_ADD_MFC:
1386 case MRT_DEL_MFC:
1387 parent = -1;
1388 case MRT_ADD_MFC_PROXY:
1389 case MRT_DEL_MFC_PROXY:
1390 if (optlen != sizeof(mfc)) {
1391 ret = -EINVAL;
1392 break;
1393 }
1394 if (copy_from_user(&mfc, optval, sizeof(mfc))) {
1395 ret = -EFAULT;
1396 break;
1397 }
1398 if (parent == 0)
1399 parent = mfc.mfcc_parent;
1400 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1401 ret = ipmr_mfc_delete(mrt, &mfc, parent);
1402 else
1403 ret = ipmr_mfc_add(net, mrt, &mfc,
1404 sk == rtnl_dereference(mrt->mroute_sk),
1405 parent);
1406 break;
1407 /* Control PIM assert. */
1408 case MRT_ASSERT:
1409 if (optlen != sizeof(val)) {
1410 ret = -EINVAL;
1411 break;
1412 }
1413 if (get_user(val, (int __user *)optval)) {
1414 ret = -EFAULT;
1415 break;
1416 }
1417 mrt->mroute_do_assert = val;
1418 break;
1419 case MRT_PIM:
1420 if (!ipmr_pimsm_enabled()) {
1421 ret = -ENOPROTOOPT;
1422 break;
1423 }
1424 if (optlen != sizeof(val)) {
1425 ret = -EINVAL;
1426 break;
1427 }
1428 if (get_user(val, (int __user *)optval)) {
1429 ret = -EFAULT;
1430 break;
1431 }
1432
1433 val = !!val;
1434 if (val != mrt->mroute_do_pim) {
1435 mrt->mroute_do_pim = val;
1436 mrt->mroute_do_assert = val;
1437 }
1438 break;
1439 case MRT_TABLE:
1440 if (!IS_BUILTIN(CONFIG_IP_MROUTE_MULTIPLE_TABLES)) {
1441 ret = -ENOPROTOOPT;
1442 break;
1443 }
1444 if (optlen != sizeof(uval)) {
1445 ret = -EINVAL;
1446 break;
1447 }
1448 if (get_user(uval, (u32 __user *)optval)) {
1449 ret = -EFAULT;
1450 break;
1451 }
1452
1453 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1454 ret = -EBUSY;
1455 } else {
1456 mrt = ipmr_new_table(net, uval);
1457 if (IS_ERR(mrt))
1458 ret = PTR_ERR(mrt);
1459 else
1460 raw_sk(sk)->ipmr_table = uval;
1461 }
1462 break;
1463 /* Spurious command, or MRT_VERSION which you cannot set. */
1464 default:
1465 ret = -ENOPROTOOPT;
1466 }
1467 out_unlock:
1468 rtnl_unlock();
1469 return ret;
1470 }
1471
1472 /* Getsock opt support for the multicast routing system. */
1473 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1474 {
1475 int olr;
1476 int val;
1477 struct net *net = sock_net(sk);
1478 struct mr_table *mrt;
1479
1480 if (sk->sk_type != SOCK_RAW ||
1481 inet_sk(sk)->inet_num != IPPROTO_IGMP)
1482 return -EOPNOTSUPP;
1483
1484 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1485 if (!mrt)
1486 return -ENOENT;
1487
1488 switch (optname) {
1489 case MRT_VERSION:
1490 val = 0x0305;
1491 break;
1492 case MRT_PIM:
1493 if (!ipmr_pimsm_enabled())
1494 return -ENOPROTOOPT;
1495 val = mrt->mroute_do_pim;
1496 break;
1497 case MRT_ASSERT:
1498 val = mrt->mroute_do_assert;
1499 break;
1500 default:
1501 return -ENOPROTOOPT;
1502 }
1503
1504 if (get_user(olr, optlen))
1505 return -EFAULT;
1506 olr = min_t(unsigned int, olr, sizeof(int));
1507 if (olr < 0)
1508 return -EINVAL;
1509 if (put_user(olr, optlen))
1510 return -EFAULT;
1511 if (copy_to_user(optval, &val, olr))
1512 return -EFAULT;
1513 return 0;
1514 }
1515
1516 /* The IP multicast ioctl support routines. */
1517 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1518 {
1519 struct sioc_sg_req sr;
1520 struct sioc_vif_req vr;
1521 struct vif_device *vif;
1522 struct mfc_cache *c;
1523 struct net *net = sock_net(sk);
1524 struct mr_table *mrt;
1525
1526 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1527 if (!mrt)
1528 return -ENOENT;
1529
1530 switch (cmd) {
1531 case SIOCGETVIFCNT:
1532 if (copy_from_user(&vr, arg, sizeof(vr)))
1533 return -EFAULT;
1534 if (vr.vifi >= mrt->maxvif)
1535 return -EINVAL;
1536 read_lock(&mrt_lock);
1537 vif = &mrt->vif_table[vr.vifi];
1538 if (VIF_EXISTS(mrt, vr.vifi)) {
1539 vr.icount = vif->pkt_in;
1540 vr.ocount = vif->pkt_out;
1541 vr.ibytes = vif->bytes_in;
1542 vr.obytes = vif->bytes_out;
1543 read_unlock(&mrt_lock);
1544
1545 if (copy_to_user(arg, &vr, sizeof(vr)))
1546 return -EFAULT;
1547 return 0;
1548 }
1549 read_unlock(&mrt_lock);
1550 return -EADDRNOTAVAIL;
1551 case SIOCGETSGCNT:
1552 if (copy_from_user(&sr, arg, sizeof(sr)))
1553 return -EFAULT;
1554
1555 rcu_read_lock();
1556 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1557 if (c) {
1558 sr.pktcnt = c->mfc_un.res.pkt;
1559 sr.bytecnt = c->mfc_un.res.bytes;
1560 sr.wrong_if = c->mfc_un.res.wrong_if;
1561 rcu_read_unlock();
1562
1563 if (copy_to_user(arg, &sr, sizeof(sr)))
1564 return -EFAULT;
1565 return 0;
1566 }
1567 rcu_read_unlock();
1568 return -EADDRNOTAVAIL;
1569 default:
1570 return -ENOIOCTLCMD;
1571 }
1572 }
1573
1574 #ifdef CONFIG_COMPAT
1575 struct compat_sioc_sg_req {
1576 struct in_addr src;
1577 struct in_addr grp;
1578 compat_ulong_t pktcnt;
1579 compat_ulong_t bytecnt;
1580 compat_ulong_t wrong_if;
1581 };
1582
1583 struct compat_sioc_vif_req {
1584 vifi_t vifi; /* Which iface */
1585 compat_ulong_t icount;
1586 compat_ulong_t ocount;
1587 compat_ulong_t ibytes;
1588 compat_ulong_t obytes;
1589 };
1590
1591 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1592 {
1593 struct compat_sioc_sg_req sr;
1594 struct compat_sioc_vif_req vr;
1595 struct vif_device *vif;
1596 struct mfc_cache *c;
1597 struct net *net = sock_net(sk);
1598 struct mr_table *mrt;
1599
1600 mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1601 if (!mrt)
1602 return -ENOENT;
1603
1604 switch (cmd) {
1605 case SIOCGETVIFCNT:
1606 if (copy_from_user(&vr, arg, sizeof(vr)))
1607 return -EFAULT;
1608 if (vr.vifi >= mrt->maxvif)
1609 return -EINVAL;
1610 read_lock(&mrt_lock);
1611 vif = &mrt->vif_table[vr.vifi];
1612 if (VIF_EXISTS(mrt, vr.vifi)) {
1613 vr.icount = vif->pkt_in;
1614 vr.ocount = vif->pkt_out;
1615 vr.ibytes = vif->bytes_in;
1616 vr.obytes = vif->bytes_out;
1617 read_unlock(&mrt_lock);
1618
1619 if (copy_to_user(arg, &vr, sizeof(vr)))
1620 return -EFAULT;
1621 return 0;
1622 }
1623 read_unlock(&mrt_lock);
1624 return -EADDRNOTAVAIL;
1625 case SIOCGETSGCNT:
1626 if (copy_from_user(&sr, arg, sizeof(sr)))
1627 return -EFAULT;
1628
1629 rcu_read_lock();
1630 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1631 if (c) {
1632 sr.pktcnt = c->mfc_un.res.pkt;
1633 sr.bytecnt = c->mfc_un.res.bytes;
1634 sr.wrong_if = c->mfc_un.res.wrong_if;
1635 rcu_read_unlock();
1636
1637 if (copy_to_user(arg, &sr, sizeof(sr)))
1638 return -EFAULT;
1639 return 0;
1640 }
1641 rcu_read_unlock();
1642 return -EADDRNOTAVAIL;
1643 default:
1644 return -ENOIOCTLCMD;
1645 }
1646 }
1647 #endif
1648
1649 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1650 {
1651 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1652 struct net *net = dev_net(dev);
1653 struct mr_table *mrt;
1654 struct vif_device *v;
1655 int ct;
1656
1657 if (event != NETDEV_UNREGISTER)
1658 return NOTIFY_DONE;
1659
1660 ipmr_for_each_table(mrt, net) {
1661 v = &mrt->vif_table[0];
1662 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1663 if (v->dev == dev)
1664 vif_delete(mrt, ct, 1, NULL);
1665 }
1666 }
1667 return NOTIFY_DONE;
1668 }
1669
1670 static struct notifier_block ip_mr_notifier = {
1671 .notifier_call = ipmr_device_event,
1672 };
1673
1674 /* Encapsulate a packet by attaching a valid IPIP header to it.
1675 * This avoids tunnel drivers and other mess and gives us the speed so
1676 * important for multicast video.
1677 */
1678 static void ip_encap(struct net *net, struct sk_buff *skb,
1679 __be32 saddr, __be32 daddr)
1680 {
1681 struct iphdr *iph;
1682 const struct iphdr *old_iph = ip_hdr(skb);
1683
1684 skb_push(skb, sizeof(struct iphdr));
1685 skb->transport_header = skb->network_header;
1686 skb_reset_network_header(skb);
1687 iph = ip_hdr(skb);
1688
1689 iph->version = 4;
1690 iph->tos = old_iph->tos;
1691 iph->ttl = old_iph->ttl;
1692 iph->frag_off = 0;
1693 iph->daddr = daddr;
1694 iph->saddr = saddr;
1695 iph->protocol = IPPROTO_IPIP;
1696 iph->ihl = 5;
1697 iph->tot_len = htons(skb->len);
1698 ip_select_ident(net, skb, NULL);
1699 ip_send_check(iph);
1700
1701 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1702 nf_reset(skb);
1703 }
1704
1705 static inline int ipmr_forward_finish(struct net *net, struct sock *sk,
1706 struct sk_buff *skb)
1707 {
1708 struct ip_options *opt = &(IPCB(skb)->opt);
1709
1710 IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
1711 IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1712
1713 if (unlikely(opt->optlen))
1714 ip_forward_options(skb);
1715
1716 return dst_output(net, sk, skb);
1717 }
1718
1719 /* Processing handlers for ipmr_forward */
1720
1721 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1722 struct sk_buff *skb, struct mfc_cache *c, int vifi)
1723 {
1724 const struct iphdr *iph = ip_hdr(skb);
1725 struct vif_device *vif = &mrt->vif_table[vifi];
1726 struct net_device *dev;
1727 struct rtable *rt;
1728 struct flowi4 fl4;
1729 int encap = 0;
1730
1731 if (!vif->dev)
1732 goto out_free;
1733
1734 if (vif->flags & VIFF_REGISTER) {
1735 vif->pkt_out++;
1736 vif->bytes_out += skb->len;
1737 vif->dev->stats.tx_bytes += skb->len;
1738 vif->dev->stats.tx_packets++;
1739 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1740 goto out_free;
1741 }
1742
1743 if (vif->flags & VIFF_TUNNEL) {
1744 rt = ip_route_output_ports(net, &fl4, NULL,
1745 vif->remote, vif->local,
1746 0, 0,
1747 IPPROTO_IPIP,
1748 RT_TOS(iph->tos), vif->link);
1749 if (IS_ERR(rt))
1750 goto out_free;
1751 encap = sizeof(struct iphdr);
1752 } else {
1753 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1754 0, 0,
1755 IPPROTO_IPIP,
1756 RT_TOS(iph->tos), vif->link);
1757 if (IS_ERR(rt))
1758 goto out_free;
1759 }
1760
1761 dev = rt->dst.dev;
1762
1763 if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1764 /* Do not fragment multicasts. Alas, IPv4 does not
1765 * allow to send ICMP, so that packets will disappear
1766 * to blackhole.
1767 */
1768 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
1769 ip_rt_put(rt);
1770 goto out_free;
1771 }
1772
1773 encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1774
1775 if (skb_cow(skb, encap)) {
1776 ip_rt_put(rt);
1777 goto out_free;
1778 }
1779
1780 vif->pkt_out++;
1781 vif->bytes_out += skb->len;
1782
1783 skb_dst_drop(skb);
1784 skb_dst_set(skb, &rt->dst);
1785 ip_decrease_ttl(ip_hdr(skb));
1786
1787 /* FIXME: forward and output firewalls used to be called here.
1788 * What do we do with netfilter? -- RR
1789 */
1790 if (vif->flags & VIFF_TUNNEL) {
1791 ip_encap(net, skb, vif->local, vif->remote);
1792 /* FIXME: extra output firewall step used to be here. --RR */
1793 vif->dev->stats.tx_packets++;
1794 vif->dev->stats.tx_bytes += skb->len;
1795 }
1796
1797 IPCB(skb)->flags |= IPSKB_FORWARDED;
1798
1799 /* RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1800 * not only before forwarding, but after forwarding on all output
1801 * interfaces. It is clear, if mrouter runs a multicasting
1802 * program, it should receive packets not depending to what interface
1803 * program is joined.
1804 * If we will not make it, the program will have to join on all
1805 * interfaces. On the other hand, multihoming host (or router, but
1806 * not mrouter) cannot join to more than one interface - it will
1807 * result in receiving multiple packets.
1808 */
1809 NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
1810 net, NULL, skb, skb->dev, dev,
1811 ipmr_forward_finish);
1812 return;
1813
1814 out_free:
1815 kfree_skb(skb);
1816 }
1817
1818 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1819 {
1820 int ct;
1821
1822 for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1823 if (mrt->vif_table[ct].dev == dev)
1824 break;
1825 }
1826 return ct;
1827 }
1828
1829 /* "local" means that we should preserve one skb (for local delivery) */
1830 static void ip_mr_forward(struct net *net, struct mr_table *mrt,
1831 struct sk_buff *skb, struct mfc_cache *cache,
1832 int local)
1833 {
1834 int true_vifi = ipmr_find_vif(mrt, skb->dev);
1835 int psend = -1;
1836 int vif, ct;
1837
1838 vif = cache->mfc_parent;
1839 cache->mfc_un.res.pkt++;
1840 cache->mfc_un.res.bytes += skb->len;
1841 cache->mfc_un.res.lastuse = jiffies;
1842
1843 if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1844 struct mfc_cache *cache_proxy;
1845
1846 /* For an (*,G) entry, we only check that the incomming
1847 * interface is part of the static tree.
1848 */
1849 cache_proxy = ipmr_cache_find_any_parent(mrt, vif);
1850 if (cache_proxy &&
1851 cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
1852 goto forward;
1853 }
1854
1855 /* Wrong interface: drop packet and (maybe) send PIM assert. */
1856 if (mrt->vif_table[vif].dev != skb->dev) {
1857 struct net_device *mdev;
1858
1859 mdev = l3mdev_master_dev_rcu(mrt->vif_table[vif].dev);
1860 if (mdev == skb->dev)
1861 goto forward;
1862
1863 if (rt_is_output_route(skb_rtable(skb))) {
1864 /* It is our own packet, looped back.
1865 * Very complicated situation...
1866 *
1867 * The best workaround until routing daemons will be
1868 * fixed is not to redistribute packet, if it was
1869 * send through wrong interface. It means, that
1870 * multicast applications WILL NOT work for
1871 * (S,G), which have default multicast route pointing
1872 * to wrong oif. In any case, it is not a good
1873 * idea to use multicasting applications on router.
1874 */
1875 goto dont_forward;
1876 }
1877
1878 cache->mfc_un.res.wrong_if++;
1879
1880 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1881 /* pimsm uses asserts, when switching from RPT to SPT,
1882 * so that we cannot check that packet arrived on an oif.
1883 * It is bad, but otherwise we would need to move pretty
1884 * large chunk of pimd to kernel. Ough... --ANK
1885 */
1886 (mrt->mroute_do_pim ||
1887 cache->mfc_un.res.ttls[true_vifi] < 255) &&
1888 time_after(jiffies,
1889 cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1890 cache->mfc_un.res.last_assert = jiffies;
1891 ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1892 }
1893 goto dont_forward;
1894 }
1895
1896 forward:
1897 mrt->vif_table[vif].pkt_in++;
1898 mrt->vif_table[vif].bytes_in += skb->len;
1899
1900 /* Forward the frame */
1901 if (cache->mfc_origin == htonl(INADDR_ANY) &&
1902 cache->mfc_mcastgrp == htonl(INADDR_ANY)) {
1903 if (true_vifi >= 0 &&
1904 true_vifi != cache->mfc_parent &&
1905 ip_hdr(skb)->ttl >
1906 cache->mfc_un.res.ttls[cache->mfc_parent]) {
1907 /* It's an (*,*) entry and the packet is not coming from
1908 * the upstream: forward the packet to the upstream
1909 * only.
1910 */
1911 psend = cache->mfc_parent;
1912 goto last_forward;
1913 }
1914 goto dont_forward;
1915 }
1916 for (ct = cache->mfc_un.res.maxvif - 1;
1917 ct >= cache->mfc_un.res.minvif; ct--) {
1918 /* For (*,G) entry, don't forward to the incoming interface */
1919 if ((cache->mfc_origin != htonl(INADDR_ANY) ||
1920 ct != true_vifi) &&
1921 ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
1922 if (psend != -1) {
1923 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1924
1925 if (skb2)
1926 ipmr_queue_xmit(net, mrt, skb2, cache,
1927 psend);
1928 }
1929 psend = ct;
1930 }
1931 }
1932 last_forward:
1933 if (psend != -1) {
1934 if (local) {
1935 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1936
1937 if (skb2)
1938 ipmr_queue_xmit(net, mrt, skb2, cache, psend);
1939 } else {
1940 ipmr_queue_xmit(net, mrt, skb, cache, psend);
1941 return;
1942 }
1943 }
1944
1945 dont_forward:
1946 if (!local)
1947 kfree_skb(skb);
1948 }
1949
1950 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
1951 {
1952 struct rtable *rt = skb_rtable(skb);
1953 struct iphdr *iph = ip_hdr(skb);
1954 struct flowi4 fl4 = {
1955 .daddr = iph->daddr,
1956 .saddr = iph->saddr,
1957 .flowi4_tos = RT_TOS(iph->tos),
1958 .flowi4_oif = (rt_is_output_route(rt) ?
1959 skb->dev->ifindex : 0),
1960 .flowi4_iif = (rt_is_output_route(rt) ?
1961 LOOPBACK_IFINDEX :
1962 skb->dev->ifindex),
1963 .flowi4_mark = skb->mark,
1964 };
1965 struct mr_table *mrt;
1966 int err;
1967
1968 err = ipmr_fib_lookup(net, &fl4, &mrt);
1969 if (err)
1970 return ERR_PTR(err);
1971 return mrt;
1972 }
1973
1974 /* Multicast packets for forwarding arrive here
1975 * Called with rcu_read_lock();
1976 */
1977 int ip_mr_input(struct sk_buff *skb)
1978 {
1979 struct mfc_cache *cache;
1980 struct net *net = dev_net(skb->dev);
1981 int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
1982 struct mr_table *mrt;
1983 struct net_device *dev;
1984
1985 /* skb->dev passed in is the loX master dev for vrfs.
1986 * As there are no vifs associated with loopback devices,
1987 * get the proper interface that does have a vif associated with it.
1988 */
1989 dev = skb->dev;
1990 if (netif_is_l3_master(skb->dev)) {
1991 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif);
1992 if (!dev) {
1993 kfree_skb(skb);
1994 return -ENODEV;
1995 }
1996 }
1997
1998 /* Packet is looped back after forward, it should not be
1999 * forwarded second time, but still can be delivered locally.
2000 */
2001 if (IPCB(skb)->flags & IPSKB_FORWARDED)
2002 goto dont_forward;
2003
2004 mrt = ipmr_rt_fib_lookup(net, skb);
2005 if (IS_ERR(mrt)) {
2006 kfree_skb(skb);
2007 return PTR_ERR(mrt);
2008 }
2009 if (!local) {
2010 if (IPCB(skb)->opt.router_alert) {
2011 if (ip_call_ra_chain(skb))
2012 return 0;
2013 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
2014 /* IGMPv1 (and broken IGMPv2 implementations sort of
2015 * Cisco IOS <= 11.2(8)) do not put router alert
2016 * option to IGMP packets destined to routable
2017 * groups. It is very bad, because it means
2018 * that we can forward NO IGMP messages.
2019 */
2020 struct sock *mroute_sk;
2021
2022 mroute_sk = rcu_dereference(mrt->mroute_sk);
2023 if (mroute_sk) {
2024 nf_reset(skb);
2025 raw_rcv(mroute_sk, skb);
2026 return 0;
2027 }
2028 }
2029 }
2030
2031 /* already under rcu_read_lock() */
2032 cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
2033 if (!cache) {
2034 int vif = ipmr_find_vif(mrt, dev);
2035
2036 if (vif >= 0)
2037 cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
2038 vif);
2039 }
2040
2041 /* No usable cache entry */
2042 if (!cache) {
2043 int vif;
2044
2045 if (local) {
2046 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2047 ip_local_deliver(skb);
2048 if (!skb2)
2049 return -ENOBUFS;
2050 skb = skb2;
2051 }
2052
2053 read_lock(&mrt_lock);
2054 vif = ipmr_find_vif(mrt, dev);
2055 if (vif >= 0) {
2056 int err2 = ipmr_cache_unresolved(mrt, vif, skb);
2057 read_unlock(&mrt_lock);
2058
2059 return err2;
2060 }
2061 read_unlock(&mrt_lock);
2062 kfree_skb(skb);
2063 return -ENODEV;
2064 }
2065
2066 read_lock(&mrt_lock);
2067 ip_mr_forward(net, mrt, skb, cache, local);
2068 read_unlock(&mrt_lock);
2069
2070 if (local)
2071 return ip_local_deliver(skb);
2072
2073 return 0;
2074
2075 dont_forward:
2076 if (local)
2077 return ip_local_deliver(skb);
2078 kfree_skb(skb);
2079 return 0;
2080 }
2081
2082 #ifdef CONFIG_IP_PIMSM_V1
2083 /* Handle IGMP messages of PIMv1 */
2084 int pim_rcv_v1(struct sk_buff *skb)
2085 {
2086 struct igmphdr *pim;
2087 struct net *net = dev_net(skb->dev);
2088 struct mr_table *mrt;
2089
2090 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2091 goto drop;
2092
2093 pim = igmp_hdr(skb);
2094
2095 mrt = ipmr_rt_fib_lookup(net, skb);
2096 if (IS_ERR(mrt))
2097 goto drop;
2098 if (!mrt->mroute_do_pim ||
2099 pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2100 goto drop;
2101
2102 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2103 drop:
2104 kfree_skb(skb);
2105 }
2106 return 0;
2107 }
2108 #endif
2109
2110 #ifdef CONFIG_IP_PIMSM_V2
2111 static int pim_rcv(struct sk_buff *skb)
2112 {
2113 struct pimreghdr *pim;
2114 struct net *net = dev_net(skb->dev);
2115 struct mr_table *mrt;
2116
2117 if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2118 goto drop;
2119
2120 pim = (struct pimreghdr *)skb_transport_header(skb);
2121 if (pim->type != ((PIM_VERSION << 4) | (PIM_TYPE_REGISTER)) ||
2122 (pim->flags & PIM_NULL_REGISTER) ||
2123 (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2124 csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2125 goto drop;
2126
2127 mrt = ipmr_rt_fib_lookup(net, skb);
2128 if (IS_ERR(mrt))
2129 goto drop;
2130 if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2131 drop:
2132 kfree_skb(skb);
2133 }
2134 return 0;
2135 }
2136 #endif
2137
2138 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2139 struct mfc_cache *c, struct rtmsg *rtm)
2140 {
2141 struct rta_mfc_stats mfcs;
2142 struct nlattr *mp_attr;
2143 struct rtnexthop *nhp;
2144 unsigned long lastuse;
2145 int ct;
2146
2147 /* If cache is unresolved, don't try to parse IIF and OIF */
2148 if (c->mfc_parent >= MAXVIFS) {
2149 rtm->rtm_flags |= RTNH_F_UNRESOLVED;
2150 return -ENOENT;
2151 }
2152
2153 if (VIF_EXISTS(mrt, c->mfc_parent) &&
2154 nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
2155 return -EMSGSIZE;
2156
2157 if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH)))
2158 return -EMSGSIZE;
2159
2160 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2161 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2162 if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) {
2163 nla_nest_cancel(skb, mp_attr);
2164 return -EMSGSIZE;
2165 }
2166
2167 nhp->rtnh_flags = 0;
2168 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2169 nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex;
2170 nhp->rtnh_len = sizeof(*nhp);
2171 }
2172 }
2173
2174 nla_nest_end(skb, mp_attr);
2175
2176 lastuse = READ_ONCE(c->mfc_un.res.lastuse);
2177 lastuse = time_after_eq(jiffies, lastuse) ? jiffies - lastuse : 0;
2178
2179 mfcs.mfcs_packets = c->mfc_un.res.pkt;
2180 mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2181 mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2182 if (nla_put_64bit(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs, RTA_PAD) ||
2183 nla_put_u64_64bit(skb, RTA_EXPIRES, jiffies_to_clock_t(lastuse),
2184 RTA_PAD))
2185 return -EMSGSIZE;
2186
2187 rtm->rtm_type = RTN_MULTICAST;
2188 return 1;
2189 }
2190
2191 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2192 __be32 saddr, __be32 daddr,
2193 struct rtmsg *rtm, u32 portid)
2194 {
2195 struct mfc_cache *cache;
2196 struct mr_table *mrt;
2197 int err;
2198
2199 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2200 if (!mrt)
2201 return -ENOENT;
2202
2203 rcu_read_lock();
2204 cache = ipmr_cache_find(mrt, saddr, daddr);
2205 if (!cache && skb->dev) {
2206 int vif = ipmr_find_vif(mrt, skb->dev);
2207
2208 if (vif >= 0)
2209 cache = ipmr_cache_find_any(mrt, daddr, vif);
2210 }
2211 if (!cache) {
2212 struct sk_buff *skb2;
2213 struct iphdr *iph;
2214 struct net_device *dev;
2215 int vif = -1;
2216
2217 dev = skb->dev;
2218 read_lock(&mrt_lock);
2219 if (dev)
2220 vif = ipmr_find_vif(mrt, dev);
2221 if (vif < 0) {
2222 read_unlock(&mrt_lock);
2223 rcu_read_unlock();
2224 return -ENODEV;
2225 }
2226 skb2 = skb_clone(skb, GFP_ATOMIC);
2227 if (!skb2) {
2228 read_unlock(&mrt_lock);
2229 rcu_read_unlock();
2230 return -ENOMEM;
2231 }
2232
2233 NETLINK_CB(skb2).portid = portid;
2234 skb_push(skb2, sizeof(struct iphdr));
2235 skb_reset_network_header(skb2);
2236 iph = ip_hdr(skb2);
2237 iph->ihl = sizeof(struct iphdr) >> 2;
2238 iph->saddr = saddr;
2239 iph->daddr = daddr;
2240 iph->version = 0;
2241 err = ipmr_cache_unresolved(mrt, vif, skb2);
2242 read_unlock(&mrt_lock);
2243 rcu_read_unlock();
2244 return err;
2245 }
2246
2247 read_lock(&mrt_lock);
2248 err = __ipmr_fill_mroute(mrt, skb, cache, rtm);
2249 read_unlock(&mrt_lock);
2250 rcu_read_unlock();
2251 return err;
2252 }
2253
2254 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2255 u32 portid, u32 seq, struct mfc_cache *c, int cmd,
2256 int flags)
2257 {
2258 struct nlmsghdr *nlh;
2259 struct rtmsg *rtm;
2260 int err;
2261
2262 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), flags);
2263 if (!nlh)
2264 return -EMSGSIZE;
2265
2266 rtm = nlmsg_data(nlh);
2267 rtm->rtm_family = RTNL_FAMILY_IPMR;
2268 rtm->rtm_dst_len = 32;
2269 rtm->rtm_src_len = 32;
2270 rtm->rtm_tos = 0;
2271 rtm->rtm_table = mrt->id;
2272 if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2273 goto nla_put_failure;
2274 rtm->rtm_type = RTN_MULTICAST;
2275 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2276 if (c->mfc_flags & MFC_STATIC)
2277 rtm->rtm_protocol = RTPROT_STATIC;
2278 else
2279 rtm->rtm_protocol = RTPROT_MROUTED;
2280 rtm->rtm_flags = 0;
2281
2282 if (nla_put_in_addr(skb, RTA_SRC, c->mfc_origin) ||
2283 nla_put_in_addr(skb, RTA_DST, c->mfc_mcastgrp))
2284 goto nla_put_failure;
2285 err = __ipmr_fill_mroute(mrt, skb, c, rtm);
2286 /* do not break the dump if cache is unresolved */
2287 if (err < 0 && err != -ENOENT)
2288 goto nla_put_failure;
2289
2290 nlmsg_end(skb, nlh);
2291 return 0;
2292
2293 nla_put_failure:
2294 nlmsg_cancel(skb, nlh);
2295 return -EMSGSIZE;
2296 }
2297
2298 static size_t mroute_msgsize(bool unresolved, int maxvif)
2299 {
2300 size_t len =
2301 NLMSG_ALIGN(sizeof(struct rtmsg))
2302 + nla_total_size(4) /* RTA_TABLE */
2303 + nla_total_size(4) /* RTA_SRC */
2304 + nla_total_size(4) /* RTA_DST */
2305 ;
2306
2307 if (!unresolved)
2308 len = len
2309 + nla_total_size(4) /* RTA_IIF */
2310 + nla_total_size(0) /* RTA_MULTIPATH */
2311 + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2312 /* RTA_MFC_STATS */
2313 + nla_total_size_64bit(sizeof(struct rta_mfc_stats))
2314 ;
2315
2316 return len;
2317 }
2318
2319 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2320 int cmd)
2321 {
2322 struct net *net = read_pnet(&mrt->net);
2323 struct sk_buff *skb;
2324 int err = -ENOBUFS;
2325
2326 skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif),
2327 GFP_ATOMIC);
2328 if (!skb)
2329 goto errout;
2330
2331 err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd, 0);
2332 if (err < 0)
2333 goto errout;
2334
2335 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2336 return;
2337
2338 errout:
2339 kfree_skb(skb);
2340 if (err < 0)
2341 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2342 }
2343
2344 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2345 {
2346 struct net *net = sock_net(skb->sk);
2347 struct mr_table *mrt;
2348 struct mfc_cache *mfc;
2349 unsigned int t = 0, s_t;
2350 unsigned int e = 0, s_e;
2351
2352 s_t = cb->args[0];
2353 s_e = cb->args[1];
2354
2355 rcu_read_lock();
2356 ipmr_for_each_table(mrt, net) {
2357 if (t < s_t)
2358 goto next_table;
2359 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list) {
2360 if (e < s_e)
2361 goto next_entry;
2362 if (ipmr_fill_mroute(mrt, skb,
2363 NETLINK_CB(cb->skb).portid,
2364 cb->nlh->nlmsg_seq,
2365 mfc, RTM_NEWROUTE,
2366 NLM_F_MULTI) < 0)
2367 goto done;
2368 next_entry:
2369 e++;
2370 }
2371 e = 0;
2372 s_e = 0;
2373
2374 spin_lock_bh(&mfc_unres_lock);
2375 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
2376 if (e < s_e)
2377 goto next_entry2;
2378 if (ipmr_fill_mroute(mrt, skb,
2379 NETLINK_CB(cb->skb).portid,
2380 cb->nlh->nlmsg_seq,
2381 mfc, RTM_NEWROUTE,
2382 NLM_F_MULTI) < 0) {
2383 spin_unlock_bh(&mfc_unres_lock);
2384 goto done;
2385 }
2386 next_entry2:
2387 e++;
2388 }
2389 spin_unlock_bh(&mfc_unres_lock);
2390 e = 0;
2391 s_e = 0;
2392 next_table:
2393 t++;
2394 }
2395 done:
2396 rcu_read_unlock();
2397
2398 cb->args[1] = e;
2399 cb->args[0] = t;
2400
2401 return skb->len;
2402 }
2403
2404 static const struct nla_policy rtm_ipmr_policy[RTA_MAX + 1] = {
2405 [RTA_SRC] = { .type = NLA_U32 },
2406 [RTA_DST] = { .type = NLA_U32 },
2407 [RTA_IIF] = { .type = NLA_U32 },
2408 [RTA_TABLE] = { .type = NLA_U32 },
2409 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2410 };
2411
2412 static bool ipmr_rtm_validate_proto(unsigned char rtm_protocol)
2413 {
2414 switch (rtm_protocol) {
2415 case RTPROT_STATIC:
2416 case RTPROT_MROUTED:
2417 return true;
2418 }
2419 return false;
2420 }
2421
2422 static int ipmr_nla_get_ttls(const struct nlattr *nla, struct mfcctl *mfcc)
2423 {
2424 struct rtnexthop *rtnh = nla_data(nla);
2425 int remaining = nla_len(nla), vifi = 0;
2426
2427 while (rtnh_ok(rtnh, remaining)) {
2428 mfcc->mfcc_ttls[vifi] = rtnh->rtnh_hops;
2429 if (++vifi == MAXVIFS)
2430 break;
2431 rtnh = rtnh_next(rtnh, &remaining);
2432 }
2433
2434 return remaining > 0 ? -EINVAL : vifi;
2435 }
2436
2437 /* returns < 0 on error, 0 for ADD_MFC and 1 for ADD_MFC_PROXY */
2438 static int rtm_to_ipmr_mfcc(struct net *net, struct nlmsghdr *nlh,
2439 struct mfcctl *mfcc, int *mrtsock,
2440 struct mr_table **mrtret,
2441 struct netlink_ext_ack *extack)
2442 {
2443 struct net_device *dev = NULL;
2444 u32 tblid = RT_TABLE_DEFAULT;
2445 struct mr_table *mrt;
2446 struct nlattr *attr;
2447 struct rtmsg *rtm;
2448 int ret, rem;
2449
2450 ret = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipmr_policy,
2451 extack);
2452 if (ret < 0)
2453 goto out;
2454 rtm = nlmsg_data(nlh);
2455
2456 ret = -EINVAL;
2457 if (rtm->rtm_family != RTNL_FAMILY_IPMR || rtm->rtm_dst_len != 32 ||
2458 rtm->rtm_type != RTN_MULTICAST ||
2459 rtm->rtm_scope != RT_SCOPE_UNIVERSE ||
2460 !ipmr_rtm_validate_proto(rtm->rtm_protocol))
2461 goto out;
2462
2463 memset(mfcc, 0, sizeof(*mfcc));
2464 mfcc->mfcc_parent = -1;
2465 ret = 0;
2466 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), rem) {
2467 switch (nla_type(attr)) {
2468 case RTA_SRC:
2469 mfcc->mfcc_origin.s_addr = nla_get_be32(attr);
2470 break;
2471 case RTA_DST:
2472 mfcc->mfcc_mcastgrp.s_addr = nla_get_be32(attr);
2473 break;
2474 case RTA_IIF:
2475 dev = __dev_get_by_index(net, nla_get_u32(attr));
2476 if (!dev) {
2477 ret = -ENODEV;
2478 goto out;
2479 }
2480 break;
2481 case RTA_MULTIPATH:
2482 if (ipmr_nla_get_ttls(attr, mfcc) < 0) {
2483 ret = -EINVAL;
2484 goto out;
2485 }
2486 break;
2487 case RTA_PREFSRC:
2488 ret = 1;
2489 break;
2490 case RTA_TABLE:
2491 tblid = nla_get_u32(attr);
2492 break;
2493 }
2494 }
2495 mrt = ipmr_get_table(net, tblid);
2496 if (!mrt) {
2497 ret = -ENOENT;
2498 goto out;
2499 }
2500 *mrtret = mrt;
2501 *mrtsock = rtm->rtm_protocol == RTPROT_MROUTED ? 1 : 0;
2502 if (dev)
2503 mfcc->mfcc_parent = ipmr_find_vif(mrt, dev);
2504
2505 out:
2506 return ret;
2507 }
2508
2509 /* takes care of both newroute and delroute */
2510 static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
2511 struct netlink_ext_ack *extack)
2512 {
2513 struct net *net = sock_net(skb->sk);
2514 int ret, mrtsock, parent;
2515 struct mr_table *tbl;
2516 struct mfcctl mfcc;
2517
2518 mrtsock = 0;
2519 tbl = NULL;
2520 ret = rtm_to_ipmr_mfcc(net, nlh, &mfcc, &mrtsock, &tbl, extack);
2521 if (ret < 0)
2522 return ret;
2523
2524 parent = ret ? mfcc.mfcc_parent : -1;
2525 if (nlh->nlmsg_type == RTM_NEWROUTE)
2526 return ipmr_mfc_add(net, tbl, &mfcc, mrtsock, parent);
2527 else
2528 return ipmr_mfc_delete(tbl, &mfcc, parent);
2529 }
2530
2531 #ifdef CONFIG_PROC_FS
2532 /* The /proc interfaces to multicast routing :
2533 * /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2534 */
2535 struct ipmr_vif_iter {
2536 struct seq_net_private p;
2537 struct mr_table *mrt;
2538 int ct;
2539 };
2540
2541 static struct vif_device *ipmr_vif_seq_idx(struct net *net,
2542 struct ipmr_vif_iter *iter,
2543 loff_t pos)
2544 {
2545 struct mr_table *mrt = iter->mrt;
2546
2547 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
2548 if (!VIF_EXISTS(mrt, iter->ct))
2549 continue;
2550 if (pos-- == 0)
2551 return &mrt->vif_table[iter->ct];
2552 }
2553 return NULL;
2554 }
2555
2556 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2557 __acquires(mrt_lock)
2558 {
2559 struct ipmr_vif_iter *iter = 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)
2565 return ERR_PTR(-ENOENT);
2566
2567 iter->mrt = mrt;
2568
2569 read_lock(&mrt_lock);
2570 return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
2571 : SEQ_START_TOKEN;
2572 }
2573
2574 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2575 {
2576 struct ipmr_vif_iter *iter = seq->private;
2577 struct net *net = seq_file_net(seq);
2578 struct mr_table *mrt = iter->mrt;
2579
2580 ++*pos;
2581 if (v == SEQ_START_TOKEN)
2582 return ipmr_vif_seq_idx(net, iter, 0);
2583
2584 while (++iter->ct < mrt->maxvif) {
2585 if (!VIF_EXISTS(mrt, iter->ct))
2586 continue;
2587 return &mrt->vif_table[iter->ct];
2588 }
2589 return NULL;
2590 }
2591
2592 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2593 __releases(mrt_lock)
2594 {
2595 read_unlock(&mrt_lock);
2596 }
2597
2598 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2599 {
2600 struct ipmr_vif_iter *iter = seq->private;
2601 struct mr_table *mrt = iter->mrt;
2602
2603 if (v == SEQ_START_TOKEN) {
2604 seq_puts(seq,
2605 "Interface BytesIn PktsIn BytesOut PktsOut Flags Local Remote\n");
2606 } else {
2607 const struct vif_device *vif = v;
2608 const char *name = vif->dev ? vif->dev->name : "none";
2609
2610 seq_printf(seq,
2611 "%2zd %-10s %8ld %7ld %8ld %7ld %05X %08X %08X\n",
2612 vif - mrt->vif_table,
2613 name, vif->bytes_in, vif->pkt_in,
2614 vif->bytes_out, vif->pkt_out,
2615 vif->flags, vif->local, vif->remote);
2616 }
2617 return 0;
2618 }
2619
2620 static const struct seq_operations ipmr_vif_seq_ops = {
2621 .start = ipmr_vif_seq_start,
2622 .next = ipmr_vif_seq_next,
2623 .stop = ipmr_vif_seq_stop,
2624 .show = ipmr_vif_seq_show,
2625 };
2626
2627 static int ipmr_vif_open(struct inode *inode, struct file *file)
2628 {
2629 return seq_open_net(inode, file, &ipmr_vif_seq_ops,
2630 sizeof(struct ipmr_vif_iter));
2631 }
2632
2633 static const struct file_operations ipmr_vif_fops = {
2634 .owner = THIS_MODULE,
2635 .open = ipmr_vif_open,
2636 .read = seq_read,
2637 .llseek = seq_lseek,
2638 .release = seq_release_net,
2639 };
2640
2641 struct ipmr_mfc_iter {
2642 struct seq_net_private p;
2643 struct mr_table *mrt;
2644 struct list_head *cache;
2645 };
2646
2647 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
2648 struct ipmr_mfc_iter *it, loff_t pos)
2649 {
2650 struct mr_table *mrt = it->mrt;
2651 struct mfc_cache *mfc;
2652
2653 rcu_read_lock();
2654 it->cache = &mrt->mfc_cache_list;
2655 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list)
2656 if (pos-- == 0)
2657 return mfc;
2658 rcu_read_unlock();
2659
2660 spin_lock_bh(&mfc_unres_lock);
2661 it->cache = &mrt->mfc_unres_queue;
2662 list_for_each_entry(mfc, it->cache, list)
2663 if (pos-- == 0)
2664 return mfc;
2665 spin_unlock_bh(&mfc_unres_lock);
2666
2667 it->cache = NULL;
2668 return NULL;
2669 }
2670
2671
2672 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2673 {
2674 struct ipmr_mfc_iter *it = seq->private;
2675 struct net *net = seq_file_net(seq);
2676 struct mr_table *mrt;
2677
2678 mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2679 if (!mrt)
2680 return ERR_PTR(-ENOENT);
2681
2682 it->mrt = mrt;
2683 it->cache = NULL;
2684 return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
2685 : SEQ_START_TOKEN;
2686 }
2687
2688 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2689 {
2690 struct ipmr_mfc_iter *it = seq->private;
2691 struct net *net = seq_file_net(seq);
2692 struct mr_table *mrt = it->mrt;
2693 struct mfc_cache *mfc = v;
2694
2695 ++*pos;
2696
2697 if (v == SEQ_START_TOKEN)
2698 return ipmr_mfc_seq_idx(net, seq->private, 0);
2699
2700 if (mfc->list.next != it->cache)
2701 return list_entry(mfc->list.next, struct mfc_cache, list);
2702
2703 if (it->cache == &mrt->mfc_unres_queue)
2704 goto end_of_list;
2705
2706 /* exhausted cache_array, show unresolved */
2707 rcu_read_unlock();
2708 it->cache = &mrt->mfc_unres_queue;
2709
2710 spin_lock_bh(&mfc_unres_lock);
2711 if (!list_empty(it->cache))
2712 return list_first_entry(it->cache, struct mfc_cache, list);
2713
2714 end_of_list:
2715 spin_unlock_bh(&mfc_unres_lock);
2716 it->cache = NULL;
2717
2718 return NULL;
2719 }
2720
2721 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
2722 {
2723 struct ipmr_mfc_iter *it = seq->private;
2724 struct mr_table *mrt = it->mrt;
2725
2726 if (it->cache == &mrt->mfc_unres_queue)
2727 spin_unlock_bh(&mfc_unres_lock);
2728 else if (it->cache == &mrt->mfc_cache_list)
2729 rcu_read_unlock();
2730 }
2731
2732 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2733 {
2734 int n;
2735
2736 if (v == SEQ_START_TOKEN) {
2737 seq_puts(seq,
2738 "Group Origin Iif Pkts Bytes Wrong Oifs\n");
2739 } else {
2740 const struct mfc_cache *mfc = v;
2741 const struct ipmr_mfc_iter *it = seq->private;
2742 const struct mr_table *mrt = it->mrt;
2743
2744 seq_printf(seq, "%08X %08X %-3hd",
2745 (__force u32) mfc->mfc_mcastgrp,
2746 (__force u32) mfc->mfc_origin,
2747 mfc->mfc_parent);
2748
2749 if (it->cache != &mrt->mfc_unres_queue) {
2750 seq_printf(seq, " %8lu %8lu %8lu",
2751 mfc->mfc_un.res.pkt,
2752 mfc->mfc_un.res.bytes,
2753 mfc->mfc_un.res.wrong_if);
2754 for (n = mfc->mfc_un.res.minvif;
2755 n < mfc->mfc_un.res.maxvif; n++) {
2756 if (VIF_EXISTS(mrt, n) &&
2757 mfc->mfc_un.res.ttls[n] < 255)
2758 seq_printf(seq,
2759 " %2d:%-3d",
2760 n, mfc->mfc_un.res.ttls[n]);
2761 }
2762 } else {
2763 /* unresolved mfc_caches don't contain
2764 * pkt, bytes and wrong_if values
2765 */
2766 seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2767 }
2768 seq_putc(seq, '\n');
2769 }
2770 return 0;
2771 }
2772
2773 static const struct seq_operations ipmr_mfc_seq_ops = {
2774 .start = ipmr_mfc_seq_start,
2775 .next = ipmr_mfc_seq_next,
2776 .stop = ipmr_mfc_seq_stop,
2777 .show = ipmr_mfc_seq_show,
2778 };
2779
2780 static int ipmr_mfc_open(struct inode *inode, struct file *file)
2781 {
2782 return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
2783 sizeof(struct ipmr_mfc_iter));
2784 }
2785
2786 static const struct file_operations ipmr_mfc_fops = {
2787 .owner = THIS_MODULE,
2788 .open = ipmr_mfc_open,
2789 .read = seq_read,
2790 .llseek = seq_lseek,
2791 .release = seq_release_net,
2792 };
2793 #endif
2794
2795 #ifdef CONFIG_IP_PIMSM_V2
2796 static const struct net_protocol pim_protocol = {
2797 .handler = pim_rcv,
2798 .netns_ok = 1,
2799 };
2800 #endif
2801
2802 /* Setup for IP multicast routing */
2803 static int __net_init ipmr_net_init(struct net *net)
2804 {
2805 int err;
2806
2807 err = ipmr_rules_init(net);
2808 if (err < 0)
2809 goto fail;
2810
2811 #ifdef CONFIG_PROC_FS
2812 err = -ENOMEM;
2813 if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops))
2814 goto proc_vif_fail;
2815 if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops))
2816 goto proc_cache_fail;
2817 #endif
2818 return 0;
2819
2820 #ifdef CONFIG_PROC_FS
2821 proc_cache_fail:
2822 remove_proc_entry("ip_mr_vif", net->proc_net);
2823 proc_vif_fail:
2824 ipmr_rules_exit(net);
2825 #endif
2826 fail:
2827 return err;
2828 }
2829
2830 static void __net_exit ipmr_net_exit(struct net *net)
2831 {
2832 #ifdef CONFIG_PROC_FS
2833 remove_proc_entry("ip_mr_cache", net->proc_net);
2834 remove_proc_entry("ip_mr_vif", net->proc_net);
2835 #endif
2836 ipmr_rules_exit(net);
2837 }
2838
2839 static struct pernet_operations ipmr_net_ops = {
2840 .init = ipmr_net_init,
2841 .exit = ipmr_net_exit,
2842 };
2843
2844 int __init ip_mr_init(void)
2845 {
2846 int err;
2847
2848 mrt_cachep = kmem_cache_create("ip_mrt_cache",
2849 sizeof(struct mfc_cache),
2850 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
2851 NULL);
2852
2853 err = register_pernet_subsys(&ipmr_net_ops);
2854 if (err)
2855 goto reg_pernet_fail;
2856
2857 err = register_netdevice_notifier(&ip_mr_notifier);
2858 if (err)
2859 goto reg_notif_fail;
2860 #ifdef CONFIG_IP_PIMSM_V2
2861 if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
2862 pr_err("%s: can't add PIM protocol\n", __func__);
2863 err = -EAGAIN;
2864 goto add_proto_fail;
2865 }
2866 #endif
2867 rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
2868 NULL, ipmr_rtm_dumproute, NULL);
2869 rtnl_register(RTNL_FAMILY_IPMR, RTM_NEWROUTE,
2870 ipmr_rtm_route, NULL, NULL);
2871 rtnl_register(RTNL_FAMILY_IPMR, RTM_DELROUTE,
2872 ipmr_rtm_route, NULL, NULL);
2873 return 0;
2874
2875 #ifdef CONFIG_IP_PIMSM_V2
2876 add_proto_fail:
2877 unregister_netdevice_notifier(&ip_mr_notifier);
2878 #endif
2879 reg_notif_fail:
2880 unregister_pernet_subsys(&ipmr_net_ops);
2881 reg_pernet_fail:
2882 kmem_cache_destroy(mrt_cachep);
2883 return err;
2884 }
Ubuntu Artful kernel mirror