2 * Anycast support for IPv6
3 * Linux INET6 implementation
6 * David L Stevens (dlstevens@us.ibm.com)
8 * based heavily on net/ipv6/mcast.c
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
16 #include <linux/capability.h>
17 #include <linux/module.h>
18 #include <linux/errno.h>
19 #include <linux/types.h>
20 #include <linux/random.h>
21 #include <linux/string.h>
22 #include <linux/socket.h>
23 #include <linux/sockios.h>
24 #include <linux/net.h>
25 #include <linux/in6.h>
26 #include <linux/netdevice.h>
27 #include <linux/if_arp.h>
28 #include <linux/route.h>
29 #include <linux/init.h>
30 #include <linux/proc_fs.h>
31 #include <linux/seq_file.h>
33 #include <net/net_namespace.h>
38 #include <net/protocol.h>
39 #include <net/if_inet6.h>
40 #include <net/ndisc.h>
41 #include <net/addrconf.h>
42 #include <net/ip6_route.h>
44 #include <net/checksum.h>
46 static int ipv6_dev_ac_dec(struct net_device *dev, struct in6_addr *addr);
48 /* Big ac list lock for all the sockets */
49 static DEFINE_RWLOCK(ipv6_sk_ac_lock);
52 ip6_onlink(struct in6_addr *addr, struct net_device *dev)
54 struct inet6_dev *idev;
55 struct inet6_ifaddr *ifa;
60 idev = __in6_dev_get(dev);
62 read_lock_bh(&idev->lock);
63 for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
64 onlink = ipv6_prefix_equal(addr, &ifa->addr,
69 read_unlock_bh(&idev->lock);
76 * socket join an anycast group
79 int ipv6_sock_ac_join(struct sock *sk, int ifindex, struct in6_addr *addr)
81 struct ipv6_pinfo *np = inet6_sk(sk);
82 struct net_device *dev = NULL;
83 struct inet6_dev *idev;
84 struct ipv6_ac_socklist *pac;
85 struct net *net = sock_net(sk);
86 int ishost = !ipv6_devconf.forwarding;
89 if (!capable(CAP_NET_ADMIN))
91 if (ipv6_addr_is_multicast(addr))
93 if (ipv6_chk_addr(net, addr, NULL, 0))
96 pac = sock_kmalloc(sk, sizeof(struct ipv6_ac_socklist), GFP_KERNEL);
100 ipv6_addr_copy(&pac->acl_addr, addr);
105 rt = rt6_lookup(net, addr, NULL, 0, 0);
109 dst_release(&rt->u.dst);
111 err = -EADDRNOTAVAIL;
114 /* router, no matching interface: just pick one */
116 dev = dev_get_by_flags(net, IFF_UP, IFF_UP|IFF_LOOPBACK);
119 dev = dev_get_by_index(net, ifindex);
126 idev = in6_dev_get(dev);
131 err = -EADDRNOTAVAIL;
134 /* reset ishost, now that we have a specific device */
135 ishost = !idev->cnf.forwarding;
138 pac->acl_ifindex = dev->ifindex;
141 * For hosts, allow link-local or matching prefix anycasts.
142 * This obviates the need for propagating anycast routes while
143 * still allowing some non-router anycast participation.
145 if (!ip6_onlink(addr, dev)) {
147 err = -EADDRNOTAVAIL;
152 err = ipv6_dev_ac_inc(dev, addr);
156 write_lock_bh(&ipv6_sk_ac_lock);
157 pac->acl_next = np->ipv6_ac_list;
158 np->ipv6_ac_list = pac;
159 write_unlock_bh(&ipv6_sk_ac_lock);
168 sock_kfree_s(sk, pac, sizeof(*pac));
173 * socket leave an anycast group
175 int ipv6_sock_ac_drop(struct sock *sk, int ifindex, struct in6_addr *addr)
177 struct ipv6_pinfo *np = inet6_sk(sk);
178 struct net_device *dev;
179 struct ipv6_ac_socklist *pac, *prev_pac;
180 struct net *net = sock_net(sk);
182 write_lock_bh(&ipv6_sk_ac_lock);
184 for (pac = np->ipv6_ac_list; pac; pac = pac->acl_next) {
185 if ((ifindex == 0 || pac->acl_ifindex == ifindex) &&
186 ipv6_addr_equal(&pac->acl_addr, addr))
191 write_unlock_bh(&ipv6_sk_ac_lock);
195 prev_pac->acl_next = pac->acl_next;
197 np->ipv6_ac_list = pac->acl_next;
199 write_unlock_bh(&ipv6_sk_ac_lock);
201 dev = dev_get_by_index(net, pac->acl_ifindex);
203 ipv6_dev_ac_dec(dev, &pac->acl_addr);
206 sock_kfree_s(sk, pac, sizeof(*pac));
210 void ipv6_sock_ac_close(struct sock *sk)
212 struct ipv6_pinfo *np = inet6_sk(sk);
213 struct net_device *dev = NULL;
214 struct ipv6_ac_socklist *pac;
215 struct net *net = sock_net(sk);
218 write_lock_bh(&ipv6_sk_ac_lock);
219 pac = np->ipv6_ac_list;
220 np->ipv6_ac_list = NULL;
221 write_unlock_bh(&ipv6_sk_ac_lock);
225 struct ipv6_ac_socklist *next = pac->acl_next;
227 if (pac->acl_ifindex != prev_index) {
230 dev = dev_get_by_index(net, pac->acl_ifindex);
231 prev_index = pac->acl_ifindex;
234 ipv6_dev_ac_dec(dev, &pac->acl_addr);
235 sock_kfree_s(sk, pac, sizeof(*pac));
243 /* The function is not used, which is funny. Apparently, author
244 * supposed to use it to filter out datagrams inside udp/raw but forgot.
246 * It is OK, anycasts are not special comparing to delivery to unicasts.
249 int inet6_ac_check(struct sock *sk, struct in6_addr *addr, int ifindex)
251 struct ipv6_ac_socklist *pac;
252 struct ipv6_pinfo *np = inet6_sk(sk);
256 read_lock(&ipv6_sk_ac_lock);
257 for (pac=np->ipv6_ac_list; pac; pac=pac->acl_next) {
258 if (ifindex && pac->acl_ifindex != ifindex)
260 found = ipv6_addr_equal(&pac->acl_addr, addr);
264 read_unlock(&ipv6_sk_ac_lock);
271 static void aca_put(struct ifacaddr6 *ac)
273 if (atomic_dec_and_test(&ac->aca_refcnt)) {
274 in6_dev_put(ac->aca_idev);
275 dst_release(&ac->aca_rt->u.dst);
281 * device anycast group inc (add if not found)
283 int ipv6_dev_ac_inc(struct net_device *dev, struct in6_addr *addr)
285 struct ifacaddr6 *aca;
286 struct inet6_dev *idev;
290 idev = in6_dev_get(dev);
295 write_lock_bh(&idev->lock);
301 for (aca = idev->ac_list; aca; aca = aca->aca_next) {
302 if (ipv6_addr_equal(&aca->aca_addr, addr)) {
310 * not found: create a new one.
313 aca = kzalloc(sizeof(struct ifacaddr6), GFP_ATOMIC);
320 rt = addrconf_dst_alloc(idev, addr, 1);
327 ipv6_addr_copy(&aca->aca_addr, addr);
328 aca->aca_idev = idev;
331 /* aca_tstamp should be updated upon changes */
332 aca->aca_cstamp = aca->aca_tstamp = jiffies;
333 atomic_set(&aca->aca_refcnt, 2);
334 spin_lock_init(&aca->aca_lock);
336 aca->aca_next = idev->ac_list;
338 write_unlock_bh(&idev->lock);
340 dst_hold(&rt->u.dst);
342 dst_release(&rt->u.dst);
344 addrconf_join_solict(dev, &aca->aca_addr);
349 write_unlock_bh(&idev->lock);
355 * device anycast group decrement
357 int __ipv6_dev_ac_dec(struct inet6_dev *idev, struct in6_addr *addr)
359 struct ifacaddr6 *aca, *prev_aca;
361 write_lock_bh(&idev->lock);
363 for (aca = idev->ac_list; aca; aca = aca->aca_next) {
364 if (ipv6_addr_equal(&aca->aca_addr, addr))
369 write_unlock_bh(&idev->lock);
372 if (--aca->aca_users > 0) {
373 write_unlock_bh(&idev->lock);
377 prev_aca->aca_next = aca->aca_next;
379 idev->ac_list = aca->aca_next;
380 write_unlock_bh(&idev->lock);
381 addrconf_leave_solict(idev, &aca->aca_addr);
383 dst_hold(&aca->aca_rt->u.dst);
384 if (ip6_del_rt(aca->aca_rt))
385 dst_free(&aca->aca_rt->u.dst);
387 dst_release(&aca->aca_rt->u.dst);
393 static int ipv6_dev_ac_dec(struct net_device *dev, struct in6_addr *addr)
396 struct inet6_dev *idev = in6_dev_get(dev);
399 ret = __ipv6_dev_ac_dec(idev, addr);
405 * check if the interface has this anycast address
407 static int ipv6_chk_acast_dev(struct net_device *dev, struct in6_addr *addr)
409 struct inet6_dev *idev;
410 struct ifacaddr6 *aca;
412 idev = in6_dev_get(dev);
414 read_lock_bh(&idev->lock);
415 for (aca = idev->ac_list; aca; aca = aca->aca_next)
416 if (ipv6_addr_equal(&aca->aca_addr, addr))
418 read_unlock_bh(&idev->lock);
426 * check if given interface (or any, if dev==0) has this anycast address
428 int ipv6_chk_acast_addr(struct net *net, struct net_device *dev,
429 struct in6_addr *addr)
434 return ipv6_chk_acast_dev(dev, addr);
435 read_lock(&dev_base_lock);
436 for_each_netdev(net, dev)
437 if (ipv6_chk_acast_dev(dev, addr)) {
441 read_unlock(&dev_base_lock);
446 #ifdef CONFIG_PROC_FS
447 struct ac6_iter_state {
448 struct seq_net_private p;
449 struct net_device *dev;
450 struct inet6_dev *idev;
453 #define ac6_seq_private(seq) ((struct ac6_iter_state *)(seq)->private)
455 static inline struct ifacaddr6 *ac6_get_first(struct seq_file *seq)
457 struct ifacaddr6 *im = NULL;
458 struct ac6_iter_state *state = ac6_seq_private(seq);
459 struct net *net = seq_file_net(seq);
462 for_each_netdev(net, state->dev) {
463 struct inet6_dev *idev;
464 idev = in6_dev_get(state->dev);
467 read_lock_bh(&idev->lock);
473 read_unlock_bh(&idev->lock);
479 static struct ifacaddr6 *ac6_get_next(struct seq_file *seq, struct ifacaddr6 *im)
481 struct ac6_iter_state *state = ac6_seq_private(seq);
485 if (likely(state->idev != NULL)) {
486 read_unlock_bh(&state->idev->lock);
487 in6_dev_put(state->idev);
489 state->dev = next_net_device(state->dev);
494 state->idev = in6_dev_get(state->dev);
497 read_lock_bh(&state->idev->lock);
498 im = state->idev->ac_list;
503 static struct ifacaddr6 *ac6_get_idx(struct seq_file *seq, loff_t pos)
505 struct ifacaddr6 *im = ac6_get_first(seq);
507 while (pos && (im = ac6_get_next(seq, im)) != NULL)
509 return pos ? NULL : im;
512 static void *ac6_seq_start(struct seq_file *seq, loff_t *pos)
513 __acquires(dev_base_lock)
515 read_lock(&dev_base_lock);
516 return ac6_get_idx(seq, *pos);
519 static void *ac6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
521 struct ifacaddr6 *im;
522 im = ac6_get_next(seq, v);
527 static void ac6_seq_stop(struct seq_file *seq, void *v)
528 __releases(dev_base_lock)
530 struct ac6_iter_state *state = ac6_seq_private(seq);
531 if (likely(state->idev != NULL)) {
532 read_unlock_bh(&state->idev->lock);
533 in6_dev_put(state->idev);
535 read_unlock(&dev_base_lock);
538 static int ac6_seq_show(struct seq_file *seq, void *v)
540 struct ifacaddr6 *im = (struct ifacaddr6 *)v;
541 struct ac6_iter_state *state = ac6_seq_private(seq);
544 "%-4d %-15s " NIP6_SEQFMT " %5d\n",
545 state->dev->ifindex, state->dev->name,
551 static const struct seq_operations ac6_seq_ops = {
552 .start = ac6_seq_start,
553 .next = ac6_seq_next,
554 .stop = ac6_seq_stop,
555 .show = ac6_seq_show,
558 static int ac6_seq_open(struct inode *inode, struct file *file)
560 return seq_open_net(inode, file, &ac6_seq_ops,
561 sizeof(struct ac6_iter_state));
564 static const struct file_operations ac6_seq_fops = {
565 .owner = THIS_MODULE,
566 .open = ac6_seq_open,
569 .release = seq_release_net,
572 int ac6_proc_init(struct net *net)
574 if (!proc_net_fops_create(net, "anycast6", S_IRUGO, &ac6_seq_fops))
580 void ac6_proc_exit(struct net *net)
582 proc_net_remove(net, "anycast6");