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>
32 #include <linux/slab.h>
34 #include <net/net_namespace.h>
39 #include <net/protocol.h>
40 #include <net/if_inet6.h>
41 #include <net/ndisc.h>
42 #include <net/addrconf.h>
43 #include <net/ip6_route.h>
45 #include <net/checksum.h>
47 static int ipv6_dev_ac_dec(struct net_device *dev, const struct in6_addr *addr);
50 * socket join an anycast group
53 int ipv6_sock_ac_join(struct sock *sk, int ifindex, const struct in6_addr *addr)
55 struct ipv6_pinfo *np = inet6_sk(sk);
56 struct net_device *dev = NULL;
57 struct inet6_dev *idev;
58 struct ipv6_ac_socklist *pac;
59 struct net *net = sock_net(sk);
60 int ishost = !net->ipv6.devconf_all->forwarding;
65 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
67 if (ipv6_addr_is_multicast(addr))
71 dev = __dev_get_by_index(net, ifindex);
73 if (ipv6_chk_addr_and_flags(net, addr, dev, true, 0, IFA_F_TENTATIVE))
76 pac = sock_kmalloc(sk, sizeof(struct ipv6_ac_socklist), GFP_KERNEL);
80 pac->acl_addr = *addr;
85 rt = rt6_lookup(net, addr, NULL, 0, NULL, 0);
93 /* router, no matching interface: just pick one */
94 dev = __dev_get_by_flags(net, IFF_UP,
95 IFF_UP | IFF_LOOPBACK);
104 idev = __in6_dev_get(dev);
109 err = -EADDRNOTAVAIL;
112 /* reset ishost, now that we have a specific device */
113 ishost = !idev->cnf.forwarding;
115 pac->acl_ifindex = dev->ifindex;
118 * For hosts, allow link-local or matching prefix anycasts.
119 * This obviates the need for propagating anycast routes while
120 * still allowing some non-router anycast participation.
122 if (!ipv6_chk_prefix(addr, dev)) {
124 err = -EADDRNOTAVAIL;
129 err = __ipv6_dev_ac_inc(idev, addr);
131 pac->acl_next = np->ipv6_ac_list;
132 np->ipv6_ac_list = pac;
138 sock_kfree_s(sk, pac, sizeof(*pac));
143 * socket leave an anycast group
145 int ipv6_sock_ac_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
147 struct ipv6_pinfo *np = inet6_sk(sk);
148 struct net_device *dev;
149 struct ipv6_ac_socklist *pac, *prev_pac;
150 struct net *net = sock_net(sk);
155 for (pac = np->ipv6_ac_list; pac; pac = pac->acl_next) {
156 if ((ifindex == 0 || pac->acl_ifindex == ifindex) &&
157 ipv6_addr_equal(&pac->acl_addr, addr))
164 prev_pac->acl_next = pac->acl_next;
166 np->ipv6_ac_list = pac->acl_next;
168 dev = __dev_get_by_index(net, pac->acl_ifindex);
170 ipv6_dev_ac_dec(dev, &pac->acl_addr);
172 sock_kfree_s(sk, pac, sizeof(*pac));
176 void __ipv6_sock_ac_close(struct sock *sk)
178 struct ipv6_pinfo *np = inet6_sk(sk);
179 struct net_device *dev = NULL;
180 struct ipv6_ac_socklist *pac;
181 struct net *net = sock_net(sk);
185 pac = np->ipv6_ac_list;
186 np->ipv6_ac_list = NULL;
190 struct ipv6_ac_socklist *next = pac->acl_next;
192 if (pac->acl_ifindex != prev_index) {
193 dev = __dev_get_by_index(net, pac->acl_ifindex);
194 prev_index = pac->acl_ifindex;
197 ipv6_dev_ac_dec(dev, &pac->acl_addr);
198 sock_kfree_s(sk, pac, sizeof(*pac));
203 void ipv6_sock_ac_close(struct sock *sk)
205 struct ipv6_pinfo *np = inet6_sk(sk);
207 if (!np->ipv6_ac_list)
210 __ipv6_sock_ac_close(sk);
214 static void aca_get(struct ifacaddr6 *aca)
216 refcount_inc(&aca->aca_refcnt);
219 static void aca_put(struct ifacaddr6 *ac)
221 if (refcount_dec_and_test(&ac->aca_refcnt)) {
222 fib6_info_release(ac->aca_rt);
227 static struct ifacaddr6 *aca_alloc(struct fib6_info *f6i,
228 const struct in6_addr *addr)
230 struct ifacaddr6 *aca;
232 aca = kzalloc(sizeof(*aca), GFP_ATOMIC);
236 aca->aca_addr = *addr;
240 /* aca_tstamp should be updated upon changes */
241 aca->aca_cstamp = aca->aca_tstamp = jiffies;
242 refcount_set(&aca->aca_refcnt, 1);
248 * device anycast group inc (add if not found)
250 int __ipv6_dev_ac_inc(struct inet6_dev *idev, const struct in6_addr *addr)
252 struct ifacaddr6 *aca;
253 struct fib6_info *f6i;
259 write_lock_bh(&idev->lock);
265 for (aca = idev->ac_list; aca; aca = aca->aca_next) {
266 if (ipv6_addr_equal(&aca->aca_addr, addr)) {
273 net = dev_net(idev->dev);
274 f6i = addrconf_f6i_alloc(net, idev, addr, true, GFP_ATOMIC);
279 aca = aca_alloc(f6i, addr);
281 fib6_info_release(f6i);
286 aca->aca_next = idev->ac_list;
289 /* Hold this for addrconf_join_solict() below before we unlock,
290 * it is already exposed via idev->ac_list.
293 write_unlock_bh(&idev->lock);
295 ip6_ins_rt(net, f6i);
297 addrconf_join_solict(idev->dev, &aca->aca_addr);
302 write_unlock_bh(&idev->lock);
307 * device anycast group decrement
309 int __ipv6_dev_ac_dec(struct inet6_dev *idev, const struct in6_addr *addr)
311 struct ifacaddr6 *aca, *prev_aca;
315 write_lock_bh(&idev->lock);
317 for (aca = idev->ac_list; aca; aca = aca->aca_next) {
318 if (ipv6_addr_equal(&aca->aca_addr, addr))
323 write_unlock_bh(&idev->lock);
326 if (--aca->aca_users > 0) {
327 write_unlock_bh(&idev->lock);
331 prev_aca->aca_next = aca->aca_next;
333 idev->ac_list = aca->aca_next;
334 write_unlock_bh(&idev->lock);
335 addrconf_leave_solict(idev, &aca->aca_addr);
337 ip6_del_rt(dev_net(idev->dev), aca->aca_rt);
343 /* called with rtnl_lock() */
344 static int ipv6_dev_ac_dec(struct net_device *dev, const struct in6_addr *addr)
346 struct inet6_dev *idev = __in6_dev_get(dev);
350 return __ipv6_dev_ac_dec(idev, addr);
353 void ipv6_ac_destroy_dev(struct inet6_dev *idev)
355 struct ifacaddr6 *aca;
357 write_lock_bh(&idev->lock);
358 while ((aca = idev->ac_list) != NULL) {
359 idev->ac_list = aca->aca_next;
360 write_unlock_bh(&idev->lock);
362 addrconf_leave_solict(idev, &aca->aca_addr);
364 ip6_del_rt(dev_net(idev->dev), aca->aca_rt);
368 write_lock_bh(&idev->lock);
370 write_unlock_bh(&idev->lock);
374 * check if the interface has this anycast address
375 * called with rcu_read_lock()
377 static bool ipv6_chk_acast_dev(struct net_device *dev, const struct in6_addr *addr)
379 struct inet6_dev *idev;
380 struct ifacaddr6 *aca;
382 idev = __in6_dev_get(dev);
384 read_lock_bh(&idev->lock);
385 for (aca = idev->ac_list; aca; aca = aca->aca_next)
386 if (ipv6_addr_equal(&aca->aca_addr, addr))
388 read_unlock_bh(&idev->lock);
395 * check if given interface (or any, if dev==0) has this anycast address
397 bool ipv6_chk_acast_addr(struct net *net, struct net_device *dev,
398 const struct in6_addr *addr)
404 found = ipv6_chk_acast_dev(dev, addr);
406 for_each_netdev_rcu(net, dev)
407 if (ipv6_chk_acast_dev(dev, addr)) {
415 /* check if this anycast address is link-local on given interface or
418 bool ipv6_chk_acast_addr_src(struct net *net, struct net_device *dev,
419 const struct in6_addr *addr)
421 return ipv6_chk_acast_addr(net,
422 (ipv6_addr_type(addr) & IPV6_ADDR_LINKLOCAL ?
427 #ifdef CONFIG_PROC_FS
428 struct ac6_iter_state {
429 struct seq_net_private p;
430 struct net_device *dev;
431 struct inet6_dev *idev;
434 #define ac6_seq_private(seq) ((struct ac6_iter_state *)(seq)->private)
436 static inline struct ifacaddr6 *ac6_get_first(struct seq_file *seq)
438 struct ifacaddr6 *im = NULL;
439 struct ac6_iter_state *state = ac6_seq_private(seq);
440 struct net *net = seq_file_net(seq);
443 for_each_netdev_rcu(net, state->dev) {
444 struct inet6_dev *idev;
445 idev = __in6_dev_get(state->dev);
448 read_lock_bh(&idev->lock);
454 read_unlock_bh(&idev->lock);
459 static struct ifacaddr6 *ac6_get_next(struct seq_file *seq, struct ifacaddr6 *im)
461 struct ac6_iter_state *state = ac6_seq_private(seq);
465 if (likely(state->idev != NULL))
466 read_unlock_bh(&state->idev->lock);
468 state->dev = next_net_device_rcu(state->dev);
473 state->idev = __in6_dev_get(state->dev);
476 read_lock_bh(&state->idev->lock);
477 im = state->idev->ac_list;
482 static struct ifacaddr6 *ac6_get_idx(struct seq_file *seq, loff_t pos)
484 struct ifacaddr6 *im = ac6_get_first(seq);
486 while (pos && (im = ac6_get_next(seq, im)) != NULL)
488 return pos ? NULL : im;
491 static void *ac6_seq_start(struct seq_file *seq, loff_t *pos)
495 return ac6_get_idx(seq, *pos);
498 static void *ac6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
500 struct ifacaddr6 *im = ac6_get_next(seq, v);
506 static void ac6_seq_stop(struct seq_file *seq, void *v)
509 struct ac6_iter_state *state = ac6_seq_private(seq);
511 if (likely(state->idev != NULL)) {
512 read_unlock_bh(&state->idev->lock);
518 static int ac6_seq_show(struct seq_file *seq, void *v)
520 struct ifacaddr6 *im = (struct ifacaddr6 *)v;
521 struct ac6_iter_state *state = ac6_seq_private(seq);
523 seq_printf(seq, "%-4d %-15s %pi6 %5d\n",
524 state->dev->ifindex, state->dev->name,
525 &im->aca_addr, im->aca_users);
529 static const struct seq_operations ac6_seq_ops = {
530 .start = ac6_seq_start,
531 .next = ac6_seq_next,
532 .stop = ac6_seq_stop,
533 .show = ac6_seq_show,
536 int __net_init ac6_proc_init(struct net *net)
538 if (!proc_create_net("anycast6", 0444, net->proc_net, &ac6_seq_ops,
539 sizeof(struct ac6_iter_state)))
545 void ac6_proc_exit(struct net *net)
547 remove_proc_entry("anycast6", net->proc_net);