Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/geert/linux...
[pandora-kernel.git] / net / ipv6 / anycast.c
1 /*
2  *      Anycast support for IPv6
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      David L Stevens (dlstevens@us.ibm.com)
7  *
8  *      based heavily on net/ipv6/mcast.c
9  *
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.
14  */
15
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>
33
34 #include <net/net_namespace.h>
35 #include <net/sock.h>
36 #include <net/snmp.h>
37
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>
44
45 #include <net/checksum.h>
46
47 static int ipv6_dev_ac_dec(struct net_device *dev, struct in6_addr *addr);
48
49 /* Big ac list lock for all the sockets */
50 static DEFINE_RWLOCK(ipv6_sk_ac_lock);
51
52
53 /*
54  *      socket join an anycast group
55  */
56
57 int ipv6_sock_ac_join(struct sock *sk, int ifindex, struct in6_addr *addr)
58 {
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;
65         int     err = 0;
66
67         if (!capable(CAP_NET_ADMIN))
68                 return -EPERM;
69         if (ipv6_addr_is_multicast(addr))
70                 return -EINVAL;
71         if (ipv6_chk_addr(net, addr, NULL, 0))
72                 return -EINVAL;
73
74         pac = sock_kmalloc(sk, sizeof(struct ipv6_ac_socklist), GFP_KERNEL);
75         if (pac == NULL)
76                 return -ENOMEM;
77         pac->acl_next = NULL;
78         ipv6_addr_copy(&pac->acl_addr, addr);
79
80         rcu_read_lock();
81         if (ifindex == 0) {
82                 struct rt6_info *rt;
83
84                 rt = rt6_lookup(net, addr, NULL, 0, 0);
85                 if (rt) {
86                         dev = rt->rt6i_dev;
87                         dst_release(&rt->dst);
88                 } else if (ishost) {
89                         err = -EADDRNOTAVAIL;
90                         goto error;
91                 } else {
92                         /* router, no matching interface: just pick one */
93                         dev = dev_get_by_flags_rcu(net, IFF_UP,
94                                                    IFF_UP | IFF_LOOPBACK);
95                 }
96         } else
97                 dev = dev_get_by_index_rcu(net, ifindex);
98
99         if (dev == NULL) {
100                 err = -ENODEV;
101                 goto error;
102         }
103
104         idev = __in6_dev_get(dev);
105         if (!idev) {
106                 if (ifindex)
107                         err = -ENODEV;
108                 else
109                         err = -EADDRNOTAVAIL;
110                 goto error;
111         }
112         /* reset ishost, now that we have a specific device */
113         ishost = !idev->cnf.forwarding;
114
115         pac->acl_ifindex = dev->ifindex;
116
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.
121          */
122         if (!ipv6_chk_prefix(addr, dev)) {
123                 if (ishost)
124                         err = -EADDRNOTAVAIL;
125                 if (err)
126                         goto error;
127         }
128
129         err = ipv6_dev_ac_inc(dev, addr);
130         if (!err) {
131                 write_lock_bh(&ipv6_sk_ac_lock);
132                 pac->acl_next = np->ipv6_ac_list;
133                 np->ipv6_ac_list = pac;
134                 write_unlock_bh(&ipv6_sk_ac_lock);
135                 pac = NULL;
136         }
137
138 error:
139         rcu_read_unlock();
140         if (pac)
141                 sock_kfree_s(sk, pac, sizeof(*pac));
142         return err;
143 }
144
145 /*
146  *      socket leave an anycast group
147  */
148 int ipv6_sock_ac_drop(struct sock *sk, int ifindex, struct in6_addr *addr)
149 {
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);
154
155         write_lock_bh(&ipv6_sk_ac_lock);
156         prev_pac = NULL;
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))
160                         break;
161                 prev_pac = pac;
162         }
163         if (!pac) {
164                 write_unlock_bh(&ipv6_sk_ac_lock);
165                 return -ENOENT;
166         }
167         if (prev_pac)
168                 prev_pac->acl_next = pac->acl_next;
169         else
170                 np->ipv6_ac_list = pac->acl_next;
171
172         write_unlock_bh(&ipv6_sk_ac_lock);
173
174         rcu_read_lock();
175         dev = dev_get_by_index_rcu(net, pac->acl_ifindex);
176         if (dev)
177                 ipv6_dev_ac_dec(dev, &pac->acl_addr);
178         rcu_read_unlock();
179
180         sock_kfree_s(sk, pac, sizeof(*pac));
181         return 0;
182 }
183
184 void ipv6_sock_ac_close(struct sock *sk)
185 {
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);
190         int     prev_index;
191
192         write_lock_bh(&ipv6_sk_ac_lock);
193         pac = np->ipv6_ac_list;
194         np->ipv6_ac_list = NULL;
195         write_unlock_bh(&ipv6_sk_ac_lock);
196
197         prev_index = 0;
198         rcu_read_lock();
199         while (pac) {
200                 struct ipv6_ac_socklist *next = pac->acl_next;
201
202                 if (pac->acl_ifindex != prev_index) {
203                         dev = dev_get_by_index_rcu(net, pac->acl_ifindex);
204                         prev_index = pac->acl_ifindex;
205                 }
206                 if (dev)
207                         ipv6_dev_ac_dec(dev, &pac->acl_addr);
208                 sock_kfree_s(sk, pac, sizeof(*pac));
209                 pac = next;
210         }
211         rcu_read_unlock();
212 }
213
214 #if 0
215 /* The function is not used, which is funny. Apparently, author
216  * supposed to use it to filter out datagrams inside udp/raw but forgot.
217  *
218  * It is OK, anycasts are not special comparing to delivery to unicasts.
219  */
220
221 int inet6_ac_check(struct sock *sk, struct in6_addr *addr, int ifindex)
222 {
223         struct ipv6_ac_socklist *pac;
224         struct ipv6_pinfo *np = inet6_sk(sk);
225         int     found;
226
227         found = 0;
228         read_lock(&ipv6_sk_ac_lock);
229         for (pac=np->ipv6_ac_list; pac; pac=pac->acl_next) {
230                 if (ifindex && pac->acl_ifindex != ifindex)
231                         continue;
232                 found = ipv6_addr_equal(&pac->acl_addr, addr);
233                 if (found)
234                         break;
235         }
236         read_unlock(&ipv6_sk_ac_lock);
237
238         return found;
239 }
240
241 #endif
242
243 static void aca_put(struct ifacaddr6 *ac)
244 {
245         if (atomic_dec_and_test(&ac->aca_refcnt)) {
246                 in6_dev_put(ac->aca_idev);
247                 dst_release(&ac->aca_rt->dst);
248                 kfree(ac);
249         }
250 }
251
252 /*
253  *      device anycast group inc (add if not found)
254  */
255 int ipv6_dev_ac_inc(struct net_device *dev, struct in6_addr *addr)
256 {
257         struct ifacaddr6 *aca;
258         struct inet6_dev *idev;
259         struct rt6_info *rt;
260         int err;
261
262         idev = in6_dev_get(dev);
263
264         if (idev == NULL)
265                 return -EINVAL;
266
267         write_lock_bh(&idev->lock);
268         if (idev->dead) {
269                 err = -ENODEV;
270                 goto out;
271         }
272
273         for (aca = idev->ac_list; aca; aca = aca->aca_next) {
274                 if (ipv6_addr_equal(&aca->aca_addr, addr)) {
275                         aca->aca_users++;
276                         err = 0;
277                         goto out;
278                 }
279         }
280
281         /*
282          *      not found: create a new one.
283          */
284
285         aca = kzalloc(sizeof(struct ifacaddr6), GFP_ATOMIC);
286
287         if (aca == NULL) {
288                 err = -ENOMEM;
289                 goto out;
290         }
291
292         rt = addrconf_dst_alloc(idev, addr, 1);
293         if (IS_ERR(rt)) {
294                 kfree(aca);
295                 err = PTR_ERR(rt);
296                 goto out;
297         }
298
299         ipv6_addr_copy(&aca->aca_addr, addr);
300         aca->aca_idev = idev;
301         aca->aca_rt = rt;
302         aca->aca_users = 1;
303         /* aca_tstamp should be updated upon changes */
304         aca->aca_cstamp = aca->aca_tstamp = jiffies;
305         atomic_set(&aca->aca_refcnt, 2);
306         spin_lock_init(&aca->aca_lock);
307
308         aca->aca_next = idev->ac_list;
309         idev->ac_list = aca;
310         write_unlock_bh(&idev->lock);
311
312         ip6_ins_rt(rt);
313
314         addrconf_join_solict(dev, &aca->aca_addr);
315
316         aca_put(aca);
317         return 0;
318 out:
319         write_unlock_bh(&idev->lock);
320         in6_dev_put(idev);
321         return err;
322 }
323
324 /*
325  *      device anycast group decrement
326  */
327 int __ipv6_dev_ac_dec(struct inet6_dev *idev, struct in6_addr *addr)
328 {
329         struct ifacaddr6 *aca, *prev_aca;
330
331         write_lock_bh(&idev->lock);
332         prev_aca = NULL;
333         for (aca = idev->ac_list; aca; aca = aca->aca_next) {
334                 if (ipv6_addr_equal(&aca->aca_addr, addr))
335                         break;
336                 prev_aca = aca;
337         }
338         if (!aca) {
339                 write_unlock_bh(&idev->lock);
340                 return -ENOENT;
341         }
342         if (--aca->aca_users > 0) {
343                 write_unlock_bh(&idev->lock);
344                 return 0;
345         }
346         if (prev_aca)
347                 prev_aca->aca_next = aca->aca_next;
348         else
349                 idev->ac_list = aca->aca_next;
350         write_unlock_bh(&idev->lock);
351         addrconf_leave_solict(idev, &aca->aca_addr);
352
353         dst_hold(&aca->aca_rt->dst);
354         ip6_del_rt(aca->aca_rt);
355
356         aca_put(aca);
357         return 0;
358 }
359
360 /* called with rcu_read_lock() */
361 static int ipv6_dev_ac_dec(struct net_device *dev, struct in6_addr *addr)
362 {
363         struct inet6_dev *idev = __in6_dev_get(dev);
364
365         if (idev == NULL)
366                 return -ENODEV;
367         return __ipv6_dev_ac_dec(idev, addr);
368 }
369
370 /*
371  *      check if the interface has this anycast address
372  *      called with rcu_read_lock()
373  */
374 static int ipv6_chk_acast_dev(struct net_device *dev, struct in6_addr *addr)
375 {
376         struct inet6_dev *idev;
377         struct ifacaddr6 *aca;
378
379         idev = __in6_dev_get(dev);
380         if (idev) {
381                 read_lock_bh(&idev->lock);
382                 for (aca = idev->ac_list; aca; aca = aca->aca_next)
383                         if (ipv6_addr_equal(&aca->aca_addr, addr))
384                                 break;
385                 read_unlock_bh(&idev->lock);
386                 return aca != NULL;
387         }
388         return 0;
389 }
390
391 /*
392  *      check if given interface (or any, if dev==0) has this anycast address
393  */
394 int ipv6_chk_acast_addr(struct net *net, struct net_device *dev,
395                         struct in6_addr *addr)
396 {
397         int found = 0;
398
399         rcu_read_lock();
400         if (dev)
401                 found = ipv6_chk_acast_dev(dev, addr);
402         else
403                 for_each_netdev_rcu(net, dev)
404                         if (ipv6_chk_acast_dev(dev, addr)) {
405                                 found = 1;
406                                 break;
407                         }
408         rcu_read_unlock();
409         return found;
410 }
411
412
413 #ifdef CONFIG_PROC_FS
414 struct ac6_iter_state {
415         struct seq_net_private p;
416         struct net_device *dev;
417         struct inet6_dev *idev;
418 };
419
420 #define ac6_seq_private(seq)    ((struct ac6_iter_state *)(seq)->private)
421
422 static inline struct ifacaddr6 *ac6_get_first(struct seq_file *seq)
423 {
424         struct ifacaddr6 *im = NULL;
425         struct ac6_iter_state *state = ac6_seq_private(seq);
426         struct net *net = seq_file_net(seq);
427
428         state->idev = NULL;
429         for_each_netdev_rcu(net, state->dev) {
430                 struct inet6_dev *idev;
431                 idev = __in6_dev_get(state->dev);
432                 if (!idev)
433                         continue;
434                 read_lock_bh(&idev->lock);
435                 im = idev->ac_list;
436                 if (im) {
437                         state->idev = idev;
438                         break;
439                 }
440                 read_unlock_bh(&idev->lock);
441         }
442         return im;
443 }
444
445 static struct ifacaddr6 *ac6_get_next(struct seq_file *seq, struct ifacaddr6 *im)
446 {
447         struct ac6_iter_state *state = ac6_seq_private(seq);
448
449         im = im->aca_next;
450         while (!im) {
451                 if (likely(state->idev != NULL))
452                         read_unlock_bh(&state->idev->lock);
453
454                 state->dev = next_net_device_rcu(state->dev);
455                 if (!state->dev) {
456                         state->idev = NULL;
457                         break;
458                 }
459                 state->idev = __in6_dev_get(state->dev);
460                 if (!state->idev)
461                         continue;
462                 read_lock_bh(&state->idev->lock);
463                 im = state->idev->ac_list;
464         }
465         return im;
466 }
467
468 static struct ifacaddr6 *ac6_get_idx(struct seq_file *seq, loff_t pos)
469 {
470         struct ifacaddr6 *im = ac6_get_first(seq);
471         if (im)
472                 while (pos && (im = ac6_get_next(seq, im)) != NULL)
473                         --pos;
474         return pos ? NULL : im;
475 }
476
477 static void *ac6_seq_start(struct seq_file *seq, loff_t *pos)
478         __acquires(RCU)
479 {
480         rcu_read_lock();
481         return ac6_get_idx(seq, *pos);
482 }
483
484 static void *ac6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
485 {
486         struct ifacaddr6 *im = ac6_get_next(seq, v);
487
488         ++*pos;
489         return im;
490 }
491
492 static void ac6_seq_stop(struct seq_file *seq, void *v)
493         __releases(RCU)
494 {
495         struct ac6_iter_state *state = ac6_seq_private(seq);
496
497         if (likely(state->idev != NULL)) {
498                 read_unlock_bh(&state->idev->lock);
499                 state->idev = NULL;
500         }
501         rcu_read_unlock();
502 }
503
504 static int ac6_seq_show(struct seq_file *seq, void *v)
505 {
506         struct ifacaddr6 *im = (struct ifacaddr6 *)v;
507         struct ac6_iter_state *state = ac6_seq_private(seq);
508
509         seq_printf(seq, "%-4d %-15s %pi6 %5d\n",
510                    state->dev->ifindex, state->dev->name,
511                    &im->aca_addr, im->aca_users);
512         return 0;
513 }
514
515 static const struct seq_operations ac6_seq_ops = {
516         .start  =       ac6_seq_start,
517         .next   =       ac6_seq_next,
518         .stop   =       ac6_seq_stop,
519         .show   =       ac6_seq_show,
520 };
521
522 static int ac6_seq_open(struct inode *inode, struct file *file)
523 {
524         return seq_open_net(inode, file, &ac6_seq_ops,
525                             sizeof(struct ac6_iter_state));
526 }
527
528 static const struct file_operations ac6_seq_fops = {
529         .owner          =       THIS_MODULE,
530         .open           =       ac6_seq_open,
531         .read           =       seq_read,
532         .llseek         =       seq_lseek,
533         .release        =       seq_release_net,
534 };
535
536 int __net_init ac6_proc_init(struct net *net)
537 {
538         if (!proc_net_fops_create(net, "anycast6", S_IRUGO, &ac6_seq_fops))
539                 return -ENOMEM;
540
541         return 0;
542 }
543
544 void ac6_proc_exit(struct net *net)
545 {
546         proc_net_remove(net, "anycast6");
547 }
548 #endif
549