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);
49 /* Big ac list lock for all the sockets */
50 static DEFINE_SPINLOCK(ipv6_sk_ac_lock);
54 * socket join an anycast group
57 int ipv6_sock_ac_join(struct sock *sk, int ifindex, const struct in6_addr *addr)
59 struct ipv6_pinfo *np = inet6_sk(sk);
60 struct net_device *dev = NULL;
61 struct inet6_dev *idev;
62 struct ipv6_ac_socklist *pac;
63 struct net *net = sock_net(sk);
64 int ishost = !net->ipv6.devconf_all->forwarding;
67 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
69 if (ipv6_addr_is_multicast(addr))
71 if (ipv6_chk_addr(net, addr, NULL, 0))
74 pac = sock_kmalloc(sk, sizeof(struct ipv6_ac_socklist), GFP_KERNEL);
78 pac->acl_addr = *addr;
84 rt = rt6_lookup(net, addr, NULL, 0, 0);
92 /* router, no matching interface: just pick one */
93 dev = dev_get_by_flags_rcu(net, IFF_UP,
94 IFF_UP | IFF_LOOPBACK);
97 dev = dev_get_by_index_rcu(net, ifindex);
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(dev, addr);
131 spin_lock_bh(&ipv6_sk_ac_lock);
132 pac->acl_next = np->ipv6_ac_list;
133 np->ipv6_ac_list = pac;
134 spin_unlock_bh(&ipv6_sk_ac_lock);
141 sock_kfree_s(sk, pac, sizeof(*pac));
146 * socket leave an anycast group
148 int ipv6_sock_ac_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
150 struct ipv6_pinfo *np = inet6_sk(sk);
151 struct net_device *dev;
152 struct ipv6_ac_socklist *pac, *prev_pac;
153 struct net *net = sock_net(sk);
155 spin_lock_bh(&ipv6_sk_ac_lock);
157 for (pac = np->ipv6_ac_list; pac; pac = pac->acl_next) {
158 if ((ifindex == 0 || pac->acl_ifindex == ifindex) &&
159 ipv6_addr_equal(&pac->acl_addr, addr))
164 spin_unlock_bh(&ipv6_sk_ac_lock);
168 prev_pac->acl_next = pac->acl_next;
170 np->ipv6_ac_list = pac->acl_next;
172 spin_unlock_bh(&ipv6_sk_ac_lock);
175 dev = dev_get_by_index_rcu(net, pac->acl_ifindex);
177 ipv6_dev_ac_dec(dev, &pac->acl_addr);
180 sock_kfree_s(sk, pac, sizeof(*pac));
184 void ipv6_sock_ac_close(struct sock *sk)
186 struct ipv6_pinfo *np = inet6_sk(sk);
187 struct net_device *dev = NULL;
188 struct ipv6_ac_socklist *pac;
189 struct net *net = sock_net(sk);
192 if (!np->ipv6_ac_list)
195 spin_lock_bh(&ipv6_sk_ac_lock);
196 pac = np->ipv6_ac_list;
197 np->ipv6_ac_list = NULL;
198 spin_unlock_bh(&ipv6_sk_ac_lock);
203 struct ipv6_ac_socklist *next = pac->acl_next;
205 if (pac->acl_ifindex != prev_index) {
206 dev = dev_get_by_index_rcu(net, pac->acl_ifindex);
207 prev_index = pac->acl_ifindex;
210 ipv6_dev_ac_dec(dev, &pac->acl_addr);
211 sock_kfree_s(sk, pac, sizeof(*pac));
217 static void aca_put(struct ifacaddr6 *ac)
219 if (atomic_dec_and_test(&ac->aca_refcnt)) {
220 in6_dev_put(ac->aca_idev);
221 dst_release(&ac->aca_rt->dst);
227 * device anycast group inc (add if not found)
229 int ipv6_dev_ac_inc(struct net_device *dev, const struct in6_addr *addr)
231 struct ifacaddr6 *aca;
232 struct inet6_dev *idev;
236 idev = in6_dev_get(dev);
241 write_lock_bh(&idev->lock);
247 for (aca = idev->ac_list; aca; aca = aca->aca_next) {
248 if (ipv6_addr_equal(&aca->aca_addr, addr)) {
256 * not found: create a new one.
259 aca = kzalloc(sizeof(struct ifacaddr6), GFP_ATOMIC);
266 rt = addrconf_dst_alloc(idev, addr, true);
273 aca->aca_addr = *addr;
274 aca->aca_idev = idev;
277 /* aca_tstamp should be updated upon changes */
278 aca->aca_cstamp = aca->aca_tstamp = jiffies;
279 atomic_set(&aca->aca_refcnt, 2);
280 spin_lock_init(&aca->aca_lock);
282 aca->aca_next = idev->ac_list;
284 write_unlock_bh(&idev->lock);
288 addrconf_join_solict(dev, &aca->aca_addr);
293 write_unlock_bh(&idev->lock);
299 * device anycast group decrement
301 int __ipv6_dev_ac_dec(struct inet6_dev *idev, const struct in6_addr *addr)
303 struct ifacaddr6 *aca, *prev_aca;
305 write_lock_bh(&idev->lock);
307 for (aca = idev->ac_list; aca; aca = aca->aca_next) {
308 if (ipv6_addr_equal(&aca->aca_addr, addr))
313 write_unlock_bh(&idev->lock);
316 if (--aca->aca_users > 0) {
317 write_unlock_bh(&idev->lock);
321 prev_aca->aca_next = aca->aca_next;
323 idev->ac_list = aca->aca_next;
324 write_unlock_bh(&idev->lock);
325 addrconf_leave_solict(idev, &aca->aca_addr);
327 dst_hold(&aca->aca_rt->dst);
328 ip6_del_rt(aca->aca_rt);
334 /* called with rcu_read_lock() */
335 static int ipv6_dev_ac_dec(struct net_device *dev, const struct in6_addr *addr)
337 struct inet6_dev *idev = __in6_dev_get(dev);
341 return __ipv6_dev_ac_dec(idev, addr);
345 * check if the interface has this anycast address
346 * called with rcu_read_lock()
348 static bool ipv6_chk_acast_dev(struct net_device *dev, const struct in6_addr *addr)
350 struct inet6_dev *idev;
351 struct ifacaddr6 *aca;
353 idev = __in6_dev_get(dev);
355 read_lock_bh(&idev->lock);
356 for (aca = idev->ac_list; aca; aca = aca->aca_next)
357 if (ipv6_addr_equal(&aca->aca_addr, addr))
359 read_unlock_bh(&idev->lock);
366 * check if given interface (or any, if dev==0) has this anycast address
368 bool ipv6_chk_acast_addr(struct net *net, struct net_device *dev,
369 const struct in6_addr *addr)
375 found = ipv6_chk_acast_dev(dev, addr);
377 for_each_netdev_rcu(net, dev)
378 if (ipv6_chk_acast_dev(dev, addr)) {
386 /* check if this anycast address is link-local on given interface or
389 bool ipv6_chk_acast_addr_src(struct net *net, struct net_device *dev,
390 const struct in6_addr *addr)
392 return ipv6_chk_acast_addr(net,
393 (ipv6_addr_type(addr) & IPV6_ADDR_LINKLOCAL ?
398 #ifdef CONFIG_PROC_FS
399 struct ac6_iter_state {
400 struct seq_net_private p;
401 struct net_device *dev;
402 struct inet6_dev *idev;
405 #define ac6_seq_private(seq) ((struct ac6_iter_state *)(seq)->private)
407 static inline struct ifacaddr6 *ac6_get_first(struct seq_file *seq)
409 struct ifacaddr6 *im = NULL;
410 struct ac6_iter_state *state = ac6_seq_private(seq);
411 struct net *net = seq_file_net(seq);
414 for_each_netdev_rcu(net, state->dev) {
415 struct inet6_dev *idev;
416 idev = __in6_dev_get(state->dev);
419 read_lock_bh(&idev->lock);
425 read_unlock_bh(&idev->lock);
430 static struct ifacaddr6 *ac6_get_next(struct seq_file *seq, struct ifacaddr6 *im)
432 struct ac6_iter_state *state = ac6_seq_private(seq);
436 if (likely(state->idev != NULL))
437 read_unlock_bh(&state->idev->lock);
439 state->dev = next_net_device_rcu(state->dev);
444 state->idev = __in6_dev_get(state->dev);
447 read_lock_bh(&state->idev->lock);
448 im = state->idev->ac_list;
453 static struct ifacaddr6 *ac6_get_idx(struct seq_file *seq, loff_t pos)
455 struct ifacaddr6 *im = ac6_get_first(seq);
457 while (pos && (im = ac6_get_next(seq, im)) != NULL)
459 return pos ? NULL : im;
462 static void *ac6_seq_start(struct seq_file *seq, loff_t *pos)
466 return ac6_get_idx(seq, *pos);
469 static void *ac6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
471 struct ifacaddr6 *im = ac6_get_next(seq, v);
477 static void ac6_seq_stop(struct seq_file *seq, void *v)
480 struct ac6_iter_state *state = ac6_seq_private(seq);
482 if (likely(state->idev != NULL)) {
483 read_unlock_bh(&state->idev->lock);
489 static int ac6_seq_show(struct seq_file *seq, void *v)
491 struct ifacaddr6 *im = (struct ifacaddr6 *)v;
492 struct ac6_iter_state *state = ac6_seq_private(seq);
494 seq_printf(seq, "%-4d %-15s %pi6 %5d\n",
495 state->dev->ifindex, state->dev->name,
496 &im->aca_addr, im->aca_users);
500 static const struct seq_operations ac6_seq_ops = {
501 .start = ac6_seq_start,
502 .next = ac6_seq_next,
503 .stop = ac6_seq_stop,
504 .show = ac6_seq_show,
507 static int ac6_seq_open(struct inode *inode, struct file *file)
509 return seq_open_net(inode, file, &ac6_seq_ops,
510 sizeof(struct ac6_iter_state));
513 static const struct file_operations ac6_seq_fops = {
514 .owner = THIS_MODULE,
515 .open = ac6_seq_open,
518 .release = seq_release_net,
521 int __net_init ac6_proc_init(struct net *net)
523 if (!proc_create("anycast6", S_IRUGO, net->proc_net, &ac6_seq_fops))
529 void ac6_proc_exit(struct net *net)
531 remove_proc_entry("anycast6", net->proc_net);