xref: /openbmc/linux/net/ipv6/anycast.c (revision 2eb0f624b709e78ec8e2f4c3412947703db99301)
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, const struct in6_addr *addr);
48 
49 /*
50  *	socket join an anycast group
51  */
52 
53 int ipv6_sock_ac_join(struct sock *sk, int ifindex, const struct in6_addr *addr)
54 {
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;
62 
63 	ASSERT_RTNL();
64 
65 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
66 		return -EPERM;
67 	if (ipv6_addr_is_multicast(addr))
68 		return -EINVAL;
69 
70 	if (ifindex)
71 		dev = __dev_get_by_index(net, ifindex);
72 
73 	if (ipv6_chk_addr_and_flags(net, addr, dev, true, 0, IFA_F_TENTATIVE))
74 		return -EINVAL;
75 
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;
81 
82 	if (ifindex == 0) {
83 		struct rt6_info *rt;
84 
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);
96 		}
97 	}
98 
99 	if (!dev) {
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(idev, addr);
130 	if (!err) {
131 		pac->acl_next = np->ipv6_ac_list;
132 		np->ipv6_ac_list = pac;
133 		pac = NULL;
134 	}
135 
136 error:
137 	if (pac)
138 		sock_kfree_s(sk, pac, sizeof(*pac));
139 	return err;
140 }
141 
142 /*
143  *	socket leave an anycast group
144  */
145 int ipv6_sock_ac_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
146 {
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);
151 
152 	ASSERT_RTNL();
153 
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;
160 	}
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;
167 
168 	dev = __dev_get_by_index(net, pac->acl_ifindex);
169 	if (dev)
170 		ipv6_dev_ac_dec(dev, &pac->acl_addr);
171 
172 	sock_kfree_s(sk, pac, sizeof(*pac));
173 	return 0;
174 }
175 
176 void ipv6_sock_ac_close(struct sock *sk)
177 {
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;
183 
184 	if (!np->ipv6_ac_list)
185 		return;
186 
187 	rtnl_lock();
188 	pac = np->ipv6_ac_list;
189 	np->ipv6_ac_list = NULL;
190 
191 	prev_index = 0;
192 	while (pac) {
193 		struct ipv6_ac_socklist *next = pac->acl_next;
194 
195 		if (pac->acl_ifindex != prev_index) {
196 			dev = __dev_get_by_index(net, pac->acl_ifindex);
197 			prev_index = pac->acl_ifindex;
198 		}
199 		if (dev)
200 			ipv6_dev_ac_dec(dev, &pac->acl_addr);
201 		sock_kfree_s(sk, pac, sizeof(*pac));
202 		pac = next;
203 	}
204 	rtnl_unlock();
205 }
206 
207 static void aca_get(struct ifacaddr6 *aca)
208 {
209 	refcount_inc(&aca->aca_refcnt);
210 }
211 
212 static void aca_put(struct ifacaddr6 *ac)
213 {
214 	if (refcount_dec_and_test(&ac->aca_refcnt)) {
215 		in6_dev_put(ac->aca_idev);
216 		fib6_info_release(ac->aca_rt);
217 		kfree(ac);
218 	}
219 }
220 
221 static struct ifacaddr6 *aca_alloc(struct fib6_info *rt,
222 				   const struct in6_addr *addr)
223 {
224 	struct inet6_dev *idev = rt->rt6i_idev;
225 	struct ifacaddr6 *aca;
226 
227 	aca = kzalloc(sizeof(*aca), GFP_ATOMIC);
228 	if (!aca)
229 		return NULL;
230 
231 	aca->aca_addr = *addr;
232 	in6_dev_hold(idev);
233 	aca->aca_idev = idev;
234 	fib6_info_hold(rt);
235 	aca->aca_rt = rt;
236 	aca->aca_users = 1;
237 	/* aca_tstamp should be updated upon changes */
238 	aca->aca_cstamp = aca->aca_tstamp = jiffies;
239 	refcount_set(&aca->aca_refcnt, 1);
240 
241 	return aca;
242 }
243 
244 /*
245  *	device anycast group inc (add if not found)
246  */
247 int __ipv6_dev_ac_inc(struct inet6_dev *idev, const struct in6_addr *addr)
248 {
249 	struct ifacaddr6 *aca;
250 	struct fib6_info *rt;
251 	struct net *net;
252 	int err;
253 
254 	ASSERT_RTNL();
255 
256 	write_lock_bh(&idev->lock);
257 	if (idev->dead) {
258 		err = -ENODEV;
259 		goto out;
260 	}
261 
262 	for (aca = idev->ac_list; aca; aca = aca->aca_next) {
263 		if (ipv6_addr_equal(&aca->aca_addr, addr)) {
264 			aca->aca_users++;
265 			err = 0;
266 			goto out;
267 		}
268 	}
269 
270 	net = dev_net(idev->dev);
271 	rt = addrconf_dst_alloc(net, idev, addr, true, GFP_ATOMIC);
272 	if (IS_ERR(rt)) {
273 		err = PTR_ERR(rt);
274 		goto out;
275 	}
276 	aca = aca_alloc(rt, addr);
277 	if (!aca) {
278 		fib6_info_release(rt);
279 		err = -ENOMEM;
280 		goto out;
281 	}
282 
283 	aca->aca_next = idev->ac_list;
284 	idev->ac_list = aca;
285 
286 	/* Hold this for addrconf_join_solict() below before we unlock,
287 	 * it is already exposed via idev->ac_list.
288 	 */
289 	aca_get(aca);
290 	write_unlock_bh(&idev->lock);
291 
292 	ip6_ins_rt(net, rt);
293 
294 	addrconf_join_solict(idev->dev, &aca->aca_addr);
295 
296 	aca_put(aca);
297 	return 0;
298 out:
299 	write_unlock_bh(&idev->lock);
300 	return err;
301 }
302 
303 /*
304  *	device anycast group decrement
305  */
306 int __ipv6_dev_ac_dec(struct inet6_dev *idev, const struct in6_addr *addr)
307 {
308 	struct ifacaddr6 *aca, *prev_aca;
309 
310 	ASSERT_RTNL();
311 
312 	write_lock_bh(&idev->lock);
313 	prev_aca = NULL;
314 	for (aca = idev->ac_list; aca; aca = aca->aca_next) {
315 		if (ipv6_addr_equal(&aca->aca_addr, addr))
316 			break;
317 		prev_aca = aca;
318 	}
319 	if (!aca) {
320 		write_unlock_bh(&idev->lock);
321 		return -ENOENT;
322 	}
323 	if (--aca->aca_users > 0) {
324 		write_unlock_bh(&idev->lock);
325 		return 0;
326 	}
327 	if (prev_aca)
328 		prev_aca->aca_next = aca->aca_next;
329 	else
330 		idev->ac_list = aca->aca_next;
331 	write_unlock_bh(&idev->lock);
332 	addrconf_leave_solict(idev, &aca->aca_addr);
333 
334 	ip6_del_rt(dev_net(idev->dev), aca->aca_rt);
335 
336 	aca_put(aca);
337 	return 0;
338 }
339 
340 /* called with rtnl_lock() */
341 static int ipv6_dev_ac_dec(struct net_device *dev, const struct in6_addr *addr)
342 {
343 	struct inet6_dev *idev = __in6_dev_get(dev);
344 
345 	if (!idev)
346 		return -ENODEV;
347 	return __ipv6_dev_ac_dec(idev, addr);
348 }
349 
350 void ipv6_ac_destroy_dev(struct inet6_dev *idev)
351 {
352 	struct ifacaddr6 *aca;
353 
354 	write_lock_bh(&idev->lock);
355 	while ((aca = idev->ac_list) != NULL) {
356 		idev->ac_list = aca->aca_next;
357 		write_unlock_bh(&idev->lock);
358 
359 		addrconf_leave_solict(idev, &aca->aca_addr);
360 
361 		ip6_del_rt(dev_net(idev->dev), aca->aca_rt);
362 
363 		aca_put(aca);
364 
365 		write_lock_bh(&idev->lock);
366 	}
367 	write_unlock_bh(&idev->lock);
368 }
369 
370 /*
371  *	check if the interface has this anycast address
372  *	called with rcu_read_lock()
373  */
374 static bool ipv6_chk_acast_dev(struct net_device *dev, const 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 false;
389 }
390 
391 /*
392  *	check if given interface (or any, if dev==0) has this anycast address
393  */
394 bool ipv6_chk_acast_addr(struct net *net, struct net_device *dev,
395 			 const struct in6_addr *addr)
396 {
397 	bool found = false;
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 = true;
406 				break;
407 			}
408 	rcu_read_unlock();
409 	return found;
410 }
411 
412 /*	check if this anycast address is link-local on given interface or
413  *	is global
414  */
415 bool ipv6_chk_acast_addr_src(struct net *net, struct net_device *dev,
416 			     const struct in6_addr *addr)
417 {
418 	return ipv6_chk_acast_addr(net,
419 				   (ipv6_addr_type(addr) & IPV6_ADDR_LINKLOCAL ?
420 				    dev : NULL),
421 				   addr);
422 }
423 
424 #ifdef CONFIG_PROC_FS
425 struct ac6_iter_state {
426 	struct seq_net_private p;
427 	struct net_device *dev;
428 	struct inet6_dev *idev;
429 };
430 
431 #define ac6_seq_private(seq)	((struct ac6_iter_state *)(seq)->private)
432 
433 static inline struct ifacaddr6 *ac6_get_first(struct seq_file *seq)
434 {
435 	struct ifacaddr6 *im = NULL;
436 	struct ac6_iter_state *state = ac6_seq_private(seq);
437 	struct net *net = seq_file_net(seq);
438 
439 	state->idev = NULL;
440 	for_each_netdev_rcu(net, state->dev) {
441 		struct inet6_dev *idev;
442 		idev = __in6_dev_get(state->dev);
443 		if (!idev)
444 			continue;
445 		read_lock_bh(&idev->lock);
446 		im = idev->ac_list;
447 		if (im) {
448 			state->idev = idev;
449 			break;
450 		}
451 		read_unlock_bh(&idev->lock);
452 	}
453 	return im;
454 }
455 
456 static struct ifacaddr6 *ac6_get_next(struct seq_file *seq, struct ifacaddr6 *im)
457 {
458 	struct ac6_iter_state *state = ac6_seq_private(seq);
459 
460 	im = im->aca_next;
461 	while (!im) {
462 		if (likely(state->idev != NULL))
463 			read_unlock_bh(&state->idev->lock);
464 
465 		state->dev = next_net_device_rcu(state->dev);
466 		if (!state->dev) {
467 			state->idev = NULL;
468 			break;
469 		}
470 		state->idev = __in6_dev_get(state->dev);
471 		if (!state->idev)
472 			continue;
473 		read_lock_bh(&state->idev->lock);
474 		im = state->idev->ac_list;
475 	}
476 	return im;
477 }
478 
479 static struct ifacaddr6 *ac6_get_idx(struct seq_file *seq, loff_t pos)
480 {
481 	struct ifacaddr6 *im = ac6_get_first(seq);
482 	if (im)
483 		while (pos && (im = ac6_get_next(seq, im)) != NULL)
484 			--pos;
485 	return pos ? NULL : im;
486 }
487 
488 static void *ac6_seq_start(struct seq_file *seq, loff_t *pos)
489 	__acquires(RCU)
490 {
491 	rcu_read_lock();
492 	return ac6_get_idx(seq, *pos);
493 }
494 
495 static void *ac6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
496 {
497 	struct ifacaddr6 *im = ac6_get_next(seq, v);
498 
499 	++*pos;
500 	return im;
501 }
502 
503 static void ac6_seq_stop(struct seq_file *seq, void *v)
504 	__releases(RCU)
505 {
506 	struct ac6_iter_state *state = ac6_seq_private(seq);
507 
508 	if (likely(state->idev != NULL)) {
509 		read_unlock_bh(&state->idev->lock);
510 		state->idev = NULL;
511 	}
512 	rcu_read_unlock();
513 }
514 
515 static int ac6_seq_show(struct seq_file *seq, void *v)
516 {
517 	struct ifacaddr6 *im = (struct ifacaddr6 *)v;
518 	struct ac6_iter_state *state = ac6_seq_private(seq);
519 
520 	seq_printf(seq, "%-4d %-15s %pi6 %5d\n",
521 		   state->dev->ifindex, state->dev->name,
522 		   &im->aca_addr, im->aca_users);
523 	return 0;
524 }
525 
526 static const struct seq_operations ac6_seq_ops = {
527 	.start	=	ac6_seq_start,
528 	.next	=	ac6_seq_next,
529 	.stop	=	ac6_seq_stop,
530 	.show	=	ac6_seq_show,
531 };
532 
533 static int ac6_seq_open(struct inode *inode, struct file *file)
534 {
535 	return seq_open_net(inode, file, &ac6_seq_ops,
536 			    sizeof(struct ac6_iter_state));
537 }
538 
539 static const struct file_operations ac6_seq_fops = {
540 	.open		=	ac6_seq_open,
541 	.read		=	seq_read,
542 	.llseek		=	seq_lseek,
543 	.release	=	seq_release_net,
544 };
545 
546 int __net_init ac6_proc_init(struct net *net)
547 {
548 	if (!proc_create("anycast6", 0444, net->proc_net, &ac6_seq_fops))
549 		return -ENOMEM;
550 
551 	return 0;
552 }
553 
554 void ac6_proc_exit(struct net *net)
555 {
556 	remove_proc_entry("anycast6", net->proc_net);
557 }
558 #endif
559