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;
85 rt = rt6_lookup(net, addr, NULL, 0, 0);
93 /* router, no matching interface: just pick one */
94 dev = dev_get_by_flags_rcu(net, IFF_UP,
95 IFF_UP | IFF_LOOPBACK);
98 dev = dev_get_by_index_rcu(net, ifindex);
105 idev = __in6_dev_get(dev);
110 err = -EADDRNOTAVAIL;
113 /* reset ishost, now that we have a specific device */
114 ishost = !idev->cnf.forwarding;
116 pac->acl_ifindex = dev->ifindex;
119 * For hosts, allow link-local or matching prefix anycasts.
120 * This obviates the need for propagating anycast routes while
121 * still allowing some non-router anycast participation.
123 if (!ipv6_chk_prefix(addr, dev)) {
125 err = -EADDRNOTAVAIL;
130 err = ipv6_dev_ac_inc(dev, addr);
132 spin_lock_bh(&ipv6_sk_ac_lock);
133 pac->acl_next = np->ipv6_ac_list;
134 np->ipv6_ac_list = pac;
135 spin_unlock_bh(&ipv6_sk_ac_lock);
143 sock_kfree_s(sk, pac, sizeof(*pac));
148 * socket leave an anycast group
150 int ipv6_sock_ac_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
152 struct ipv6_pinfo *np = inet6_sk(sk);
153 struct net_device *dev;
154 struct ipv6_ac_socklist *pac, *prev_pac;
155 struct net *net = sock_net(sk);
157 spin_lock_bh(&ipv6_sk_ac_lock);
159 for (pac = np->ipv6_ac_list; pac; pac = pac->acl_next) {
160 if ((ifindex == 0 || pac->acl_ifindex == ifindex) &&
161 ipv6_addr_equal(&pac->acl_addr, addr))
166 spin_unlock_bh(&ipv6_sk_ac_lock);
170 prev_pac->acl_next = pac->acl_next;
172 np->ipv6_ac_list = pac->acl_next;
174 spin_unlock_bh(&ipv6_sk_ac_lock);
178 dev = dev_get_by_index_rcu(net, pac->acl_ifindex);
180 ipv6_dev_ac_dec(dev, &pac->acl_addr);
184 sock_kfree_s(sk, pac, sizeof(*pac));
188 void ipv6_sock_ac_close(struct sock *sk)
190 struct ipv6_pinfo *np = inet6_sk(sk);
191 struct net_device *dev = NULL;
192 struct ipv6_ac_socklist *pac;
193 struct net *net = sock_net(sk);
196 if (!np->ipv6_ac_list)
199 spin_lock_bh(&ipv6_sk_ac_lock);
200 pac = np->ipv6_ac_list;
201 np->ipv6_ac_list = NULL;
202 spin_unlock_bh(&ipv6_sk_ac_lock);
208 struct ipv6_ac_socklist *next = pac->acl_next;
210 if (pac->acl_ifindex != prev_index) {
211 dev = dev_get_by_index_rcu(net, pac->acl_ifindex);
212 prev_index = pac->acl_ifindex;
215 ipv6_dev_ac_dec(dev, &pac->acl_addr);
216 sock_kfree_s(sk, pac, sizeof(*pac));
223 static void aca_put(struct ifacaddr6 *ac)
225 if (atomic_dec_and_test(&ac->aca_refcnt)) {
226 in6_dev_put(ac->aca_idev);
227 dst_release(&ac->aca_rt->dst);
233 * device anycast group inc (add if not found)
235 int ipv6_dev_ac_inc(struct net_device *dev, const struct in6_addr *addr)
237 struct ifacaddr6 *aca;
238 struct inet6_dev *idev;
244 idev = in6_dev_get(dev);
249 write_lock_bh(&idev->lock);
255 for (aca = idev->ac_list; aca; aca = aca->aca_next) {
256 if (ipv6_addr_equal(&aca->aca_addr, addr)) {
264 * not found: create a new one.
267 aca = kzalloc(sizeof(struct ifacaddr6), GFP_ATOMIC);
274 rt = addrconf_dst_alloc(idev, addr, true);
281 aca->aca_addr = *addr;
282 aca->aca_idev = idev;
285 /* aca_tstamp should be updated upon changes */
286 aca->aca_cstamp = aca->aca_tstamp = jiffies;
287 atomic_set(&aca->aca_refcnt, 2);
288 spin_lock_init(&aca->aca_lock);
290 aca->aca_next = idev->ac_list;
292 write_unlock_bh(&idev->lock);
296 addrconf_join_solict(dev, &aca->aca_addr);
301 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 dst_hold(&aca->aca_rt->dst);
338 ip6_del_rt(aca->aca_rt);
344 /* called with rcu_read_lock() */
345 static int ipv6_dev_ac_dec(struct net_device *dev, const struct in6_addr *addr)
347 struct inet6_dev *idev = __in6_dev_get(dev);
351 return __ipv6_dev_ac_dec(idev, addr);
354 void ipv6_ac_destroy_dev(struct inet6_dev *idev)
356 struct ifacaddr6 *aca;
358 write_lock_bh(&idev->lock);
359 while ((aca = idev->ac_list) != NULL) {
360 idev->ac_list = aca->aca_next;
361 write_unlock_bh(&idev->lock);
363 addrconf_leave_solict(idev, &aca->aca_addr);
365 dst_hold(&aca->aca_rt->dst);
366 ip6_del_rt(aca->aca_rt);
370 write_lock_bh(&idev->lock);
372 write_unlock_bh(&idev->lock);
376 * check if the interface has this anycast address
377 * called with rcu_read_lock()
379 static bool ipv6_chk_acast_dev(struct net_device *dev, const struct in6_addr *addr)
381 struct inet6_dev *idev;
382 struct ifacaddr6 *aca;
384 idev = __in6_dev_get(dev);
386 read_lock_bh(&idev->lock);
387 for (aca = idev->ac_list; aca; aca = aca->aca_next)
388 if (ipv6_addr_equal(&aca->aca_addr, addr))
390 read_unlock_bh(&idev->lock);
397 * check if given interface (or any, if dev==0) has this anycast address
399 bool ipv6_chk_acast_addr(struct net *net, struct net_device *dev,
400 const struct in6_addr *addr)
406 found = ipv6_chk_acast_dev(dev, addr);
408 for_each_netdev_rcu(net, dev)
409 if (ipv6_chk_acast_dev(dev, addr)) {
417 /* check if this anycast address is link-local on given interface or
420 bool ipv6_chk_acast_addr_src(struct net *net, struct net_device *dev,
421 const struct in6_addr *addr)
423 return ipv6_chk_acast_addr(net,
424 (ipv6_addr_type(addr) & IPV6_ADDR_LINKLOCAL ?
429 #ifdef CONFIG_PROC_FS
430 struct ac6_iter_state {
431 struct seq_net_private p;
432 struct net_device *dev;
433 struct inet6_dev *idev;
436 #define ac6_seq_private(seq) ((struct ac6_iter_state *)(seq)->private)
438 static inline struct ifacaddr6 *ac6_get_first(struct seq_file *seq)
440 struct ifacaddr6 *im = NULL;
441 struct ac6_iter_state *state = ac6_seq_private(seq);
442 struct net *net = seq_file_net(seq);
445 for_each_netdev_rcu(net, state->dev) {
446 struct inet6_dev *idev;
447 idev = __in6_dev_get(state->dev);
450 read_lock_bh(&idev->lock);
456 read_unlock_bh(&idev->lock);
461 static struct ifacaddr6 *ac6_get_next(struct seq_file *seq, struct ifacaddr6 *im)
463 struct ac6_iter_state *state = ac6_seq_private(seq);
467 if (likely(state->idev != NULL))
468 read_unlock_bh(&state->idev->lock);
470 state->dev = next_net_device_rcu(state->dev);
475 state->idev = __in6_dev_get(state->dev);
478 read_lock_bh(&state->idev->lock);
479 im = state->idev->ac_list;
484 static struct ifacaddr6 *ac6_get_idx(struct seq_file *seq, loff_t pos)
486 struct ifacaddr6 *im = ac6_get_first(seq);
488 while (pos && (im = ac6_get_next(seq, im)) != NULL)
490 return pos ? NULL : im;
493 static void *ac6_seq_start(struct seq_file *seq, loff_t *pos)
497 return ac6_get_idx(seq, *pos);
500 static void *ac6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
502 struct ifacaddr6 *im = ac6_get_next(seq, v);
508 static void ac6_seq_stop(struct seq_file *seq, void *v)
511 struct ac6_iter_state *state = ac6_seq_private(seq);
513 if (likely(state->idev != NULL)) {
514 read_unlock_bh(&state->idev->lock);
520 static int ac6_seq_show(struct seq_file *seq, void *v)
522 struct ifacaddr6 *im = (struct ifacaddr6 *)v;
523 struct ac6_iter_state *state = ac6_seq_private(seq);
525 seq_printf(seq, "%-4d %-15s %pi6 %5d\n",
526 state->dev->ifindex, state->dev->name,
527 &im->aca_addr, im->aca_users);
531 static const struct seq_operations ac6_seq_ops = {
532 .start = ac6_seq_start,
533 .next = ac6_seq_next,
534 .stop = ac6_seq_stop,
535 .show = ac6_seq_show,
538 static int ac6_seq_open(struct inode *inode, struct file *file)
540 return seq_open_net(inode, file, &ac6_seq_ops,
541 sizeof(struct ac6_iter_state));
544 static const struct file_operations ac6_seq_fops = {
545 .owner = THIS_MODULE,
546 .open = ac6_seq_open,
549 .release = seq_release_net,
552 int __net_init ac6_proc_init(struct net *net)
554 if (!proc_create("anycast6", S_IRUGO, net->proc_net, &ac6_seq_fops))
560 void ac6_proc_exit(struct net *net)
562 remove_proc_entry("anycast6", net->proc_net);