xref: /openbmc/linux/net/ipv6/anycast.c (revision 2359ccdd)
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 		dst_release(&ac->aca_rt->dst);
217 		kfree(ac);
218 	}
219 }
220 
221 static struct ifacaddr6 *aca_alloc(struct rt6_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 	aca->aca_rt = rt;
235 	aca->aca_users = 1;
236 	/* aca_tstamp should be updated upon changes */
237 	aca->aca_cstamp = aca->aca_tstamp = jiffies;
238 	refcount_set(&aca->aca_refcnt, 1);
239 
240 	return aca;
241 }
242 
243 /*
244  *	device anycast group inc (add if not found)
245  */
246 int __ipv6_dev_ac_inc(struct inet6_dev *idev, const struct in6_addr *addr)
247 {
248 	struct ifacaddr6 *aca;
249 	struct rt6_info *rt;
250 	int err;
251 
252 	ASSERT_RTNL();
253 
254 	write_lock_bh(&idev->lock);
255 	if (idev->dead) {
256 		err = -ENODEV;
257 		goto out;
258 	}
259 
260 	for (aca = idev->ac_list; aca; aca = aca->aca_next) {
261 		if (ipv6_addr_equal(&aca->aca_addr, addr)) {
262 			aca->aca_users++;
263 			err = 0;
264 			goto out;
265 		}
266 	}
267 
268 	rt = addrconf_dst_alloc(idev, addr, true);
269 	if (IS_ERR(rt)) {
270 		err = PTR_ERR(rt);
271 		goto out;
272 	}
273 	aca = aca_alloc(rt, addr);
274 	if (!aca) {
275 		ip6_rt_put(rt);
276 		err = -ENOMEM;
277 		goto out;
278 	}
279 
280 	aca->aca_next = idev->ac_list;
281 	idev->ac_list = aca;
282 
283 	/* Hold this for addrconf_join_solict() below before we unlock,
284 	 * it is already exposed via idev->ac_list.
285 	 */
286 	aca_get(aca);
287 	write_unlock_bh(&idev->lock);
288 
289 	ip6_ins_rt(rt);
290 
291 	addrconf_join_solict(idev->dev, &aca->aca_addr);
292 
293 	aca_put(aca);
294 	return 0;
295 out:
296 	write_unlock_bh(&idev->lock);
297 	return err;
298 }
299 
300 /*
301  *	device anycast group decrement
302  */
303 int __ipv6_dev_ac_dec(struct inet6_dev *idev, const struct in6_addr *addr)
304 {
305 	struct ifacaddr6 *aca, *prev_aca;
306 
307 	ASSERT_RTNL();
308 
309 	write_lock_bh(&idev->lock);
310 	prev_aca = NULL;
311 	for (aca = idev->ac_list; aca; aca = aca->aca_next) {
312 		if (ipv6_addr_equal(&aca->aca_addr, addr))
313 			break;
314 		prev_aca = aca;
315 	}
316 	if (!aca) {
317 		write_unlock_bh(&idev->lock);
318 		return -ENOENT;
319 	}
320 	if (--aca->aca_users > 0) {
321 		write_unlock_bh(&idev->lock);
322 		return 0;
323 	}
324 	if (prev_aca)
325 		prev_aca->aca_next = aca->aca_next;
326 	else
327 		idev->ac_list = aca->aca_next;
328 	write_unlock_bh(&idev->lock);
329 	addrconf_leave_solict(idev, &aca->aca_addr);
330 
331 	dst_hold(&aca->aca_rt->dst);
332 	ip6_del_rt(aca->aca_rt);
333 
334 	aca_put(aca);
335 	return 0;
336 }
337 
338 /* called with rtnl_lock() */
339 static int ipv6_dev_ac_dec(struct net_device *dev, const struct in6_addr *addr)
340 {
341 	struct inet6_dev *idev = __in6_dev_get(dev);
342 
343 	if (!idev)
344 		return -ENODEV;
345 	return __ipv6_dev_ac_dec(idev, addr);
346 }
347 
348 void ipv6_ac_destroy_dev(struct inet6_dev *idev)
349 {
350 	struct ifacaddr6 *aca;
351 
352 	write_lock_bh(&idev->lock);
353 	while ((aca = idev->ac_list) != NULL) {
354 		idev->ac_list = aca->aca_next;
355 		write_unlock_bh(&idev->lock);
356 
357 		addrconf_leave_solict(idev, &aca->aca_addr);
358 
359 		dst_hold(&aca->aca_rt->dst);
360 		ip6_del_rt(aca->aca_rt);
361 
362 		aca_put(aca);
363 
364 		write_lock_bh(&idev->lock);
365 	}
366 	write_unlock_bh(&idev->lock);
367 }
368 
369 /*
370  *	check if the interface has this anycast address
371  *	called with rcu_read_lock()
372  */
373 static bool ipv6_chk_acast_dev(struct net_device *dev, const struct in6_addr *addr)
374 {
375 	struct inet6_dev *idev;
376 	struct ifacaddr6 *aca;
377 
378 	idev = __in6_dev_get(dev);
379 	if (idev) {
380 		read_lock_bh(&idev->lock);
381 		for (aca = idev->ac_list; aca; aca = aca->aca_next)
382 			if (ipv6_addr_equal(&aca->aca_addr, addr))
383 				break;
384 		read_unlock_bh(&idev->lock);
385 		return aca != NULL;
386 	}
387 	return false;
388 }
389 
390 /*
391  *	check if given interface (or any, if dev==0) has this anycast address
392  */
393 bool ipv6_chk_acast_addr(struct net *net, struct net_device *dev,
394 			 const struct in6_addr *addr)
395 {
396 	bool found = false;
397 
398 	rcu_read_lock();
399 	if (dev)
400 		found = ipv6_chk_acast_dev(dev, addr);
401 	else
402 		for_each_netdev_rcu(net, dev)
403 			if (ipv6_chk_acast_dev(dev, addr)) {
404 				found = true;
405 				break;
406 			}
407 	rcu_read_unlock();
408 	return found;
409 }
410 
411 /*	check if this anycast address is link-local on given interface or
412  *	is global
413  */
414 bool ipv6_chk_acast_addr_src(struct net *net, struct net_device *dev,
415 			     const struct in6_addr *addr)
416 {
417 	return ipv6_chk_acast_addr(net,
418 				   (ipv6_addr_type(addr) & IPV6_ADDR_LINKLOCAL ?
419 				    dev : NULL),
420 				   addr);
421 }
422 
423 #ifdef CONFIG_PROC_FS
424 struct ac6_iter_state {
425 	struct seq_net_private p;
426 	struct net_device *dev;
427 	struct inet6_dev *idev;
428 };
429 
430 #define ac6_seq_private(seq)	((struct ac6_iter_state *)(seq)->private)
431 
432 static inline struct ifacaddr6 *ac6_get_first(struct seq_file *seq)
433 {
434 	struct ifacaddr6 *im = NULL;
435 	struct ac6_iter_state *state = ac6_seq_private(seq);
436 	struct net *net = seq_file_net(seq);
437 
438 	state->idev = NULL;
439 	for_each_netdev_rcu(net, state->dev) {
440 		struct inet6_dev *idev;
441 		idev = __in6_dev_get(state->dev);
442 		if (!idev)
443 			continue;
444 		read_lock_bh(&idev->lock);
445 		im = idev->ac_list;
446 		if (im) {
447 			state->idev = idev;
448 			break;
449 		}
450 		read_unlock_bh(&idev->lock);
451 	}
452 	return im;
453 }
454 
455 static struct ifacaddr6 *ac6_get_next(struct seq_file *seq, struct ifacaddr6 *im)
456 {
457 	struct ac6_iter_state *state = ac6_seq_private(seq);
458 
459 	im = im->aca_next;
460 	while (!im) {
461 		if (likely(state->idev != NULL))
462 			read_unlock_bh(&state->idev->lock);
463 
464 		state->dev = next_net_device_rcu(state->dev);
465 		if (!state->dev) {
466 			state->idev = NULL;
467 			break;
468 		}
469 		state->idev = __in6_dev_get(state->dev);
470 		if (!state->idev)
471 			continue;
472 		read_lock_bh(&state->idev->lock);
473 		im = state->idev->ac_list;
474 	}
475 	return im;
476 }
477 
478 static struct ifacaddr6 *ac6_get_idx(struct seq_file *seq, loff_t pos)
479 {
480 	struct ifacaddr6 *im = ac6_get_first(seq);
481 	if (im)
482 		while (pos && (im = ac6_get_next(seq, im)) != NULL)
483 			--pos;
484 	return pos ? NULL : im;
485 }
486 
487 static void *ac6_seq_start(struct seq_file *seq, loff_t *pos)
488 	__acquires(RCU)
489 {
490 	rcu_read_lock();
491 	return ac6_get_idx(seq, *pos);
492 }
493 
494 static void *ac6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
495 {
496 	struct ifacaddr6 *im = ac6_get_next(seq, v);
497 
498 	++*pos;
499 	return im;
500 }
501 
502 static void ac6_seq_stop(struct seq_file *seq, void *v)
503 	__releases(RCU)
504 {
505 	struct ac6_iter_state *state = ac6_seq_private(seq);
506 
507 	if (likely(state->idev != NULL)) {
508 		read_unlock_bh(&state->idev->lock);
509 		state->idev = NULL;
510 	}
511 	rcu_read_unlock();
512 }
513 
514 static int ac6_seq_show(struct seq_file *seq, void *v)
515 {
516 	struct ifacaddr6 *im = (struct ifacaddr6 *)v;
517 	struct ac6_iter_state *state = ac6_seq_private(seq);
518 
519 	seq_printf(seq, "%-4d %-15s %pi6 %5d\n",
520 		   state->dev->ifindex, state->dev->name,
521 		   &im->aca_addr, im->aca_users);
522 	return 0;
523 }
524 
525 static const struct seq_operations ac6_seq_ops = {
526 	.start	=	ac6_seq_start,
527 	.next	=	ac6_seq_next,
528 	.stop	=	ac6_seq_stop,
529 	.show	=	ac6_seq_show,
530 };
531 
532 static int ac6_seq_open(struct inode *inode, struct file *file)
533 {
534 	return seq_open_net(inode, file, &ac6_seq_ops,
535 			    sizeof(struct ac6_iter_state));
536 }
537 
538 static const struct file_operations ac6_seq_fops = {
539 	.open		=	ac6_seq_open,
540 	.read		=	seq_read,
541 	.llseek		=	seq_lseek,
542 	.release	=	seq_release_net,
543 };
544 
545 int __net_init ac6_proc_init(struct net *net)
546 {
547 	if (!proc_create("anycast6", 0444, net->proc_net, &ac6_seq_fops))
548 		return -ENOMEM;
549 
550 	return 0;
551 }
552 
553 void ac6_proc_exit(struct net *net)
554 {
555 	remove_proc_entry("anycast6", net->proc_net);
556 }
557 #endif
558