NFSv4: Fix error handling in nfs4_sp4_select_mode()
[linux-2.6/btrfs-unstable.git] / net / ipv6 / anycast.c
blob4e0ff7031edd55ce6dbb3f2c62e22b9040cc7fec
1 /*
2 * Anycast support for IPv6
3 * Linux INET6 implementation
5 * Authors:
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>
35 #include <net/sock.h>
36 #include <net/snmp.h>
38 #include <net/ipv6.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;
61 int err = 0;
63 ASSERT_RTNL();
65 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
66 return -EPERM;
67 if (ipv6_addr_is_multicast(addr))
68 return -EINVAL;
70 if (ifindex)
71 dev = __dev_get_by_index(net, ifindex);
73 if (ipv6_chk_addr_and_flags(net, addr, dev, true, 0, IFA_F_TENTATIVE))
74 return -EINVAL;
76 pac = sock_kmalloc(sk, sizeof(struct ipv6_ac_socklist), GFP_KERNEL);
77 if (!pac)
78 return -ENOMEM;
79 pac->acl_next = NULL;
80 pac->acl_addr = *addr;
82 if (ifindex == 0) {
83 struct rt6_info *rt;
85 rt = rt6_lookup(net, addr, NULL, 0, NULL, 0);
86 if (rt) {
87 dev = rt->dst.dev;
88 ip6_rt_put(rt);
89 } else if (ishost) {
90 err = -EADDRNOTAVAIL;
91 goto error;
92 } else {
93 /* router, no matching interface: just pick one */
94 dev = __dev_get_by_flags(net, IFF_UP,
95 IFF_UP | IFF_LOOPBACK);
99 if (!dev) {
100 err = -ENODEV;
101 goto error;
104 idev = __in6_dev_get(dev);
105 if (!idev) {
106 if (ifindex)
107 err = -ENODEV;
108 else
109 err = -EADDRNOTAVAIL;
110 goto error;
112 /* reset ishost, now that we have a specific device */
113 ishost = !idev->cnf.forwarding;
115 pac->acl_ifindex = dev->ifindex;
117 /* XXX
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)) {
123 if (ishost)
124 err = -EADDRNOTAVAIL;
125 if (err)
126 goto error;
129 err = __ipv6_dev_ac_inc(idev, addr);
130 if (!err) {
131 pac->acl_next = np->ipv6_ac_list;
132 np->ipv6_ac_list = pac;
133 pac = NULL;
136 error:
137 if (pac)
138 sock_kfree_s(sk, pac, sizeof(*pac));
139 return err;
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);
152 ASSERT_RTNL();
154 prev_pac = NULL;
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))
158 break;
159 prev_pac = pac;
161 if (!pac)
162 return -ENOENT;
163 if (prev_pac)
164 prev_pac->acl_next = pac->acl_next;
165 else
166 np->ipv6_ac_list = pac->acl_next;
168 dev = __dev_get_by_index(net, pac->acl_ifindex);
169 if (dev)
170 ipv6_dev_ac_dec(dev, &pac->acl_addr);
172 sock_kfree_s(sk, pac, sizeof(*pac));
173 return 0;
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);
182 int prev_index;
184 if (!np->ipv6_ac_list)
185 return;
187 rtnl_lock();
188 pac = np->ipv6_ac_list;
189 np->ipv6_ac_list = NULL;
191 prev_index = 0;
192 while (pac) {
193 struct ipv6_ac_socklist *next = pac->acl_next;
195 if (pac->acl_ifindex != prev_index) {
196 dev = __dev_get_by_index(net, pac->acl_ifindex);
197 prev_index = pac->acl_ifindex;
199 if (dev)
200 ipv6_dev_ac_dec(dev, &pac->acl_addr);
201 sock_kfree_s(sk, pac, sizeof(*pac));
202 pac = next;
204 rtnl_unlock();
207 static void aca_get(struct ifacaddr6 *aca)
209 refcount_inc(&aca->aca_refcnt);
212 static void aca_put(struct ifacaddr6 *ac)
214 if (refcount_dec_and_test(&ac->aca_refcnt)) {
215 fib6_info_release(ac->aca_rt);
216 kfree(ac);
220 static struct ifacaddr6 *aca_alloc(struct fib6_info *f6i,
221 const struct in6_addr *addr)
223 struct ifacaddr6 *aca;
225 aca = kzalloc(sizeof(*aca), GFP_ATOMIC);
226 if (!aca)
227 return NULL;
229 aca->aca_addr = *addr;
230 fib6_info_hold(f6i);
231 aca->aca_rt = f6i;
232 aca->aca_users = 1;
233 /* aca_tstamp should be updated upon changes */
234 aca->aca_cstamp = aca->aca_tstamp = jiffies;
235 refcount_set(&aca->aca_refcnt, 1);
237 return aca;
241 * device anycast group inc (add if not found)
243 int __ipv6_dev_ac_inc(struct inet6_dev *idev, const struct in6_addr *addr)
245 struct ifacaddr6 *aca;
246 struct fib6_info *f6i;
247 struct net *net;
248 int err;
250 ASSERT_RTNL();
252 write_lock_bh(&idev->lock);
253 if (idev->dead) {
254 err = -ENODEV;
255 goto out;
258 for (aca = idev->ac_list; aca; aca = aca->aca_next) {
259 if (ipv6_addr_equal(&aca->aca_addr, addr)) {
260 aca->aca_users++;
261 err = 0;
262 goto out;
266 net = dev_net(idev->dev);
267 f6i = addrconf_f6i_alloc(net, idev, addr, true, GFP_ATOMIC);
268 if (IS_ERR(f6i)) {
269 err = PTR_ERR(f6i);
270 goto out;
272 aca = aca_alloc(f6i, addr);
273 if (!aca) {
274 fib6_info_release(f6i);
275 err = -ENOMEM;
276 goto out;
279 aca->aca_next = idev->ac_list;
280 idev->ac_list = aca;
282 /* Hold this for addrconf_join_solict() below before we unlock,
283 * it is already exposed via idev->ac_list.
285 aca_get(aca);
286 write_unlock_bh(&idev->lock);
288 ip6_ins_rt(net, f6i);
290 addrconf_join_solict(idev->dev, &aca->aca_addr);
292 aca_put(aca);
293 return 0;
294 out:
295 write_unlock_bh(&idev->lock);
296 return err;
300 * device anycast group decrement
302 int __ipv6_dev_ac_dec(struct inet6_dev *idev, const struct in6_addr *addr)
304 struct ifacaddr6 *aca, *prev_aca;
306 ASSERT_RTNL();
308 write_lock_bh(&idev->lock);
309 prev_aca = NULL;
310 for (aca = idev->ac_list; aca; aca = aca->aca_next) {
311 if (ipv6_addr_equal(&aca->aca_addr, addr))
312 break;
313 prev_aca = aca;
315 if (!aca) {
316 write_unlock_bh(&idev->lock);
317 return -ENOENT;
319 if (--aca->aca_users > 0) {
320 write_unlock_bh(&idev->lock);
321 return 0;
323 if (prev_aca)
324 prev_aca->aca_next = aca->aca_next;
325 else
326 idev->ac_list = aca->aca_next;
327 write_unlock_bh(&idev->lock);
328 addrconf_leave_solict(idev, &aca->aca_addr);
330 ip6_del_rt(dev_net(idev->dev), aca->aca_rt);
332 aca_put(aca);
333 return 0;
336 /* called with rtnl_lock() */
337 static int ipv6_dev_ac_dec(struct net_device *dev, const struct in6_addr *addr)
339 struct inet6_dev *idev = __in6_dev_get(dev);
341 if (!idev)
342 return -ENODEV;
343 return __ipv6_dev_ac_dec(idev, addr);
346 void ipv6_ac_destroy_dev(struct inet6_dev *idev)
348 struct ifacaddr6 *aca;
350 write_lock_bh(&idev->lock);
351 while ((aca = idev->ac_list) != NULL) {
352 idev->ac_list = aca->aca_next;
353 write_unlock_bh(&idev->lock);
355 addrconf_leave_solict(idev, &aca->aca_addr);
357 ip6_del_rt(dev_net(idev->dev), aca->aca_rt);
359 aca_put(aca);
361 write_lock_bh(&idev->lock);
363 write_unlock_bh(&idev->lock);
367 * check if the interface has this anycast address
368 * called with rcu_read_lock()
370 static bool ipv6_chk_acast_dev(struct net_device *dev, const struct in6_addr *addr)
372 struct inet6_dev *idev;
373 struct ifacaddr6 *aca;
375 idev = __in6_dev_get(dev);
376 if (idev) {
377 read_lock_bh(&idev->lock);
378 for (aca = idev->ac_list; aca; aca = aca->aca_next)
379 if (ipv6_addr_equal(&aca->aca_addr, addr))
380 break;
381 read_unlock_bh(&idev->lock);
382 return aca != NULL;
384 return false;
388 * check if given interface (or any, if dev==0) has this anycast address
390 bool ipv6_chk_acast_addr(struct net *net, struct net_device *dev,
391 const struct in6_addr *addr)
393 bool found = false;
395 rcu_read_lock();
396 if (dev)
397 found = ipv6_chk_acast_dev(dev, addr);
398 else
399 for_each_netdev_rcu(net, dev)
400 if (ipv6_chk_acast_dev(dev, addr)) {
401 found = true;
402 break;
404 rcu_read_unlock();
405 return found;
408 /* check if this anycast address is link-local on given interface or
409 * is global
411 bool ipv6_chk_acast_addr_src(struct net *net, struct net_device *dev,
412 const struct in6_addr *addr)
414 return ipv6_chk_acast_addr(net,
415 (ipv6_addr_type(addr) & IPV6_ADDR_LINKLOCAL ?
416 dev : NULL),
417 addr);
420 #ifdef CONFIG_PROC_FS
421 struct ac6_iter_state {
422 struct seq_net_private p;
423 struct net_device *dev;
424 struct inet6_dev *idev;
427 #define ac6_seq_private(seq) ((struct ac6_iter_state *)(seq)->private)
429 static inline struct ifacaddr6 *ac6_get_first(struct seq_file *seq)
431 struct ifacaddr6 *im = NULL;
432 struct ac6_iter_state *state = ac6_seq_private(seq);
433 struct net *net = seq_file_net(seq);
435 state->idev = NULL;
436 for_each_netdev_rcu(net, state->dev) {
437 struct inet6_dev *idev;
438 idev = __in6_dev_get(state->dev);
439 if (!idev)
440 continue;
441 read_lock_bh(&idev->lock);
442 im = idev->ac_list;
443 if (im) {
444 state->idev = idev;
445 break;
447 read_unlock_bh(&idev->lock);
449 return im;
452 static struct ifacaddr6 *ac6_get_next(struct seq_file *seq, struct ifacaddr6 *im)
454 struct ac6_iter_state *state = ac6_seq_private(seq);
456 im = im->aca_next;
457 while (!im) {
458 if (likely(state->idev != NULL))
459 read_unlock_bh(&state->idev->lock);
461 state->dev = next_net_device_rcu(state->dev);
462 if (!state->dev) {
463 state->idev = NULL;
464 break;
466 state->idev = __in6_dev_get(state->dev);
467 if (!state->idev)
468 continue;
469 read_lock_bh(&state->idev->lock);
470 im = state->idev->ac_list;
472 return im;
475 static struct ifacaddr6 *ac6_get_idx(struct seq_file *seq, loff_t pos)
477 struct ifacaddr6 *im = ac6_get_first(seq);
478 if (im)
479 while (pos && (im = ac6_get_next(seq, im)) != NULL)
480 --pos;
481 return pos ? NULL : im;
484 static void *ac6_seq_start(struct seq_file *seq, loff_t *pos)
485 __acquires(RCU)
487 rcu_read_lock();
488 return ac6_get_idx(seq, *pos);
491 static void *ac6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
493 struct ifacaddr6 *im = ac6_get_next(seq, v);
495 ++*pos;
496 return im;
499 static void ac6_seq_stop(struct seq_file *seq, void *v)
500 __releases(RCU)
502 struct ac6_iter_state *state = ac6_seq_private(seq);
504 if (likely(state->idev != NULL)) {
505 read_unlock_bh(&state->idev->lock);
506 state->idev = NULL;
508 rcu_read_unlock();
511 static int ac6_seq_show(struct seq_file *seq, void *v)
513 struct ifacaddr6 *im = (struct ifacaddr6 *)v;
514 struct ac6_iter_state *state = ac6_seq_private(seq);
516 seq_printf(seq, "%-4d %-15s %pi6 %5d\n",
517 state->dev->ifindex, state->dev->name,
518 &im->aca_addr, im->aca_users);
519 return 0;
522 static const struct seq_operations ac6_seq_ops = {
523 .start = ac6_seq_start,
524 .next = ac6_seq_next,
525 .stop = ac6_seq_stop,
526 .show = ac6_seq_show,
529 int __net_init ac6_proc_init(struct net *net)
531 if (!proc_create_net("anycast6", 0444, net->proc_net, &ac6_seq_ops,
532 sizeof(struct ac6_iter_state)))
533 return -ENOMEM;
535 return 0;
538 void ac6_proc_exit(struct net *net)
540 remove_proc_entry("anycast6", net->proc_net);
542 #endif