xref: /linux/net/ipv6/fib6_rules.c (revision e9f0878c4b2004ac19581274c1ae4c61ae3ca70e)
1 /*
2  * net/ipv6/fib6_rules.c	IPv6 Routing Policy Rules
3  *
4  * Copyright (C)2003-2006 Helsinki University of Technology
5  * Copyright (C)2003-2006 USAGI/WIDE Project
6  *
7  *	This program is free software; you can redistribute it and/or
8  *	modify it under the terms of the GNU General Public License as
9  *	published by the Free Software Foundation, version 2.
10  *
11  * Authors
12  *	Thomas Graf		<tgraf@suug.ch>
13  *	Ville Nuorvala		<vnuorval@tcs.hut.fi>
14  */
15 
16 #include <linux/netdevice.h>
17 #include <linux/notifier.h>
18 #include <linux/export.h>
19 
20 #include <net/fib_rules.h>
21 #include <net/ipv6.h>
22 #include <net/addrconf.h>
23 #include <net/ip6_route.h>
24 #include <net/netlink.h>
25 
26 struct fib6_rule {
27 	struct fib_rule		common;
28 	struct rt6key		src;
29 	struct rt6key		dst;
30 	u8			tclass;
31 };
32 
33 static bool fib6_rule_matchall(const struct fib_rule *rule)
34 {
35 	struct fib6_rule *r = container_of(rule, struct fib6_rule, common);
36 
37 	if (r->dst.plen || r->src.plen || r->tclass)
38 		return false;
39 	return fib_rule_matchall(rule);
40 }
41 
42 bool fib6_rule_default(const struct fib_rule *rule)
43 {
44 	if (!fib6_rule_matchall(rule) || rule->action != FR_ACT_TO_TBL ||
45 	    rule->l3mdev)
46 		return false;
47 	if (rule->table != RT6_TABLE_LOCAL && rule->table != RT6_TABLE_MAIN)
48 		return false;
49 	return true;
50 }
51 EXPORT_SYMBOL_GPL(fib6_rule_default);
52 
53 int fib6_rules_dump(struct net *net, struct notifier_block *nb)
54 {
55 	return fib_rules_dump(net, nb, AF_INET6);
56 }
57 
58 unsigned int fib6_rules_seq_read(struct net *net)
59 {
60 	return fib_rules_seq_read(net, AF_INET6);
61 }
62 
63 /* called with rcu lock held; no reference taken on fib6_info */
64 struct fib6_info *fib6_lookup(struct net *net, int oif, struct flowi6 *fl6,
65 			      int flags)
66 {
67 	struct fib6_info *f6i;
68 	int err;
69 
70 	if (net->ipv6.fib6_has_custom_rules) {
71 		struct fib_lookup_arg arg = {
72 			.lookup_ptr = fib6_table_lookup,
73 			.lookup_data = &oif,
74 			.flags = FIB_LOOKUP_NOREF,
75 		};
76 
77 		l3mdev_update_flow(net, flowi6_to_flowi(fl6));
78 
79 		err = fib_rules_lookup(net->ipv6.fib6_rules_ops,
80 				       flowi6_to_flowi(fl6), flags, &arg);
81 		if (err)
82 			return ERR_PTR(err);
83 
84 		f6i = arg.result ? : net->ipv6.fib6_null_entry;
85 	} else {
86 		f6i = fib6_table_lookup(net, net->ipv6.fib6_local_tbl,
87 					oif, fl6, flags);
88 		if (!f6i || f6i == net->ipv6.fib6_null_entry)
89 			f6i = fib6_table_lookup(net, net->ipv6.fib6_main_tbl,
90 						oif, fl6, flags);
91 	}
92 
93 	return f6i;
94 }
95 
96 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6,
97 				   const struct sk_buff *skb,
98 				   int flags, pol_lookup_t lookup)
99 {
100 	if (net->ipv6.fib6_has_custom_rules) {
101 		struct fib_lookup_arg arg = {
102 			.lookup_ptr = lookup,
103 			.lookup_data = skb,
104 			.flags = FIB_LOOKUP_NOREF,
105 		};
106 
107 		/* update flow if oif or iif point to device enslaved to l3mdev */
108 		l3mdev_update_flow(net, flowi6_to_flowi(fl6));
109 
110 		fib_rules_lookup(net->ipv6.fib6_rules_ops,
111 				 flowi6_to_flowi(fl6), flags, &arg);
112 
113 		if (arg.result)
114 			return arg.result;
115 	} else {
116 		struct rt6_info *rt;
117 
118 		rt = lookup(net, net->ipv6.fib6_local_tbl, fl6, skb, flags);
119 		if (rt != net->ipv6.ip6_null_entry && rt->dst.error != -EAGAIN)
120 			return &rt->dst;
121 		ip6_rt_put(rt);
122 		rt = lookup(net, net->ipv6.fib6_main_tbl, fl6, skb, flags);
123 		if (rt->dst.error != -EAGAIN)
124 			return &rt->dst;
125 		ip6_rt_put(rt);
126 	}
127 
128 	dst_hold(&net->ipv6.ip6_null_entry->dst);
129 	return &net->ipv6.ip6_null_entry->dst;
130 }
131 
132 static int fib6_rule_saddr(struct net *net, struct fib_rule *rule, int flags,
133 			   struct flowi6 *flp6, const struct net_device *dev)
134 {
135 	struct fib6_rule *r = (struct fib6_rule *)rule;
136 
137 	/* If we need to find a source address for this traffic,
138 	 * we check the result if it meets requirement of the rule.
139 	 */
140 	if ((rule->flags & FIB_RULE_FIND_SADDR) &&
141 	    r->src.plen && !(flags & RT6_LOOKUP_F_HAS_SADDR)) {
142 		struct in6_addr saddr;
143 
144 		if (ipv6_dev_get_saddr(net, dev, &flp6->daddr,
145 				       rt6_flags2srcprefs(flags), &saddr))
146 			return -EAGAIN;
147 
148 		if (!ipv6_prefix_equal(&saddr, &r->src.addr, r->src.plen))
149 			return -EAGAIN;
150 
151 		flp6->saddr = saddr;
152 	}
153 
154 	return 0;
155 }
156 
157 static int fib6_rule_action_alt(struct fib_rule *rule, struct flowi *flp,
158 				int flags, struct fib_lookup_arg *arg)
159 {
160 	struct flowi6 *flp6 = &flp->u.ip6;
161 	struct net *net = rule->fr_net;
162 	struct fib6_table *table;
163 	struct fib6_info *f6i;
164 	int err = -EAGAIN, *oif;
165 	u32 tb_id;
166 
167 	switch (rule->action) {
168 	case FR_ACT_TO_TBL:
169 		break;
170 	case FR_ACT_UNREACHABLE:
171 		return -ENETUNREACH;
172 	case FR_ACT_PROHIBIT:
173 		return -EACCES;
174 	case FR_ACT_BLACKHOLE:
175 	default:
176 		return -EINVAL;
177 	}
178 
179 	tb_id = fib_rule_get_table(rule, arg);
180 	table = fib6_get_table(net, tb_id);
181 	if (!table)
182 		return -EAGAIN;
183 
184 	oif = (int *)arg->lookup_data;
185 	f6i = fib6_table_lookup(net, table, *oif, flp6, flags);
186 	if (f6i != net->ipv6.fib6_null_entry) {
187 		err = fib6_rule_saddr(net, rule, flags, flp6,
188 				      fib6_info_nh_dev(f6i));
189 
190 		if (likely(!err))
191 			arg->result = f6i;
192 	}
193 
194 	return err;
195 }
196 
197 static int __fib6_rule_action(struct fib_rule *rule, struct flowi *flp,
198 			      int flags, struct fib_lookup_arg *arg)
199 {
200 	struct flowi6 *flp6 = &flp->u.ip6;
201 	struct rt6_info *rt = NULL;
202 	struct fib6_table *table;
203 	struct net *net = rule->fr_net;
204 	pol_lookup_t lookup = arg->lookup_ptr;
205 	int err = 0;
206 	u32 tb_id;
207 
208 	switch (rule->action) {
209 	case FR_ACT_TO_TBL:
210 		break;
211 	case FR_ACT_UNREACHABLE:
212 		err = -ENETUNREACH;
213 		rt = net->ipv6.ip6_null_entry;
214 		goto discard_pkt;
215 	default:
216 	case FR_ACT_BLACKHOLE:
217 		err = -EINVAL;
218 		rt = net->ipv6.ip6_blk_hole_entry;
219 		goto discard_pkt;
220 	case FR_ACT_PROHIBIT:
221 		err = -EACCES;
222 		rt = net->ipv6.ip6_prohibit_entry;
223 		goto discard_pkt;
224 	}
225 
226 	tb_id = fib_rule_get_table(rule, arg);
227 	table = fib6_get_table(net, tb_id);
228 	if (!table) {
229 		err = -EAGAIN;
230 		goto out;
231 	}
232 
233 	rt = lookup(net, table, flp6, arg->lookup_data, flags);
234 	if (rt != net->ipv6.ip6_null_entry) {
235 		err = fib6_rule_saddr(net, rule, flags, flp6,
236 				      ip6_dst_idev(&rt->dst)->dev);
237 
238 		if (err == -EAGAIN)
239 			goto again;
240 
241 		err = rt->dst.error;
242 		if (err != -EAGAIN)
243 			goto out;
244 	}
245 again:
246 	ip6_rt_put(rt);
247 	err = -EAGAIN;
248 	rt = NULL;
249 	goto out;
250 
251 discard_pkt:
252 	dst_hold(&rt->dst);
253 out:
254 	arg->result = rt;
255 	return err;
256 }
257 
258 static int fib6_rule_action(struct fib_rule *rule, struct flowi *flp,
259 			    int flags, struct fib_lookup_arg *arg)
260 {
261 	if (arg->lookup_ptr == fib6_table_lookup)
262 		return fib6_rule_action_alt(rule, flp, flags, arg);
263 
264 	return __fib6_rule_action(rule, flp, flags, arg);
265 }
266 
267 static bool fib6_rule_suppress(struct fib_rule *rule, struct fib_lookup_arg *arg)
268 {
269 	struct rt6_info *rt = (struct rt6_info *) arg->result;
270 	struct net_device *dev = NULL;
271 
272 	if (rt->rt6i_idev)
273 		dev = rt->rt6i_idev->dev;
274 
275 	/* do not accept result if the route does
276 	 * not meet the required prefix length
277 	 */
278 	if (rt->rt6i_dst.plen <= rule->suppress_prefixlen)
279 		goto suppress_route;
280 
281 	/* do not accept result if the route uses a device
282 	 * belonging to a forbidden interface group
283 	 */
284 	if (rule->suppress_ifgroup != -1 && dev && dev->group == rule->suppress_ifgroup)
285 		goto suppress_route;
286 
287 	return false;
288 
289 suppress_route:
290 	ip6_rt_put(rt);
291 	return true;
292 }
293 
294 static int fib6_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
295 {
296 	struct fib6_rule *r = (struct fib6_rule *) rule;
297 	struct flowi6 *fl6 = &fl->u.ip6;
298 
299 	if (r->dst.plen &&
300 	    !ipv6_prefix_equal(&fl6->daddr, &r->dst.addr, r->dst.plen))
301 		return 0;
302 
303 	/*
304 	 * If FIB_RULE_FIND_SADDR is set and we do not have a
305 	 * source address for the traffic, we defer check for
306 	 * source address.
307 	 */
308 	if (r->src.plen) {
309 		if (flags & RT6_LOOKUP_F_HAS_SADDR) {
310 			if (!ipv6_prefix_equal(&fl6->saddr, &r->src.addr,
311 					       r->src.plen))
312 				return 0;
313 		} else if (!(r->common.flags & FIB_RULE_FIND_SADDR))
314 			return 0;
315 	}
316 
317 	if (r->tclass && r->tclass != ip6_tclass(fl6->flowlabel))
318 		return 0;
319 
320 	if (rule->ip_proto && (rule->ip_proto != fl6->flowi6_proto))
321 		return 0;
322 
323 	if (fib_rule_port_range_set(&rule->sport_range) &&
324 	    !fib_rule_port_inrange(&rule->sport_range, fl6->fl6_sport))
325 		return 0;
326 
327 	if (fib_rule_port_range_set(&rule->dport_range) &&
328 	    !fib_rule_port_inrange(&rule->dport_range, fl6->fl6_dport))
329 		return 0;
330 
331 	return 1;
332 }
333 
334 static const struct nla_policy fib6_rule_policy[FRA_MAX+1] = {
335 	FRA_GENERIC_POLICY,
336 };
337 
338 static int fib6_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
339 			       struct fib_rule_hdr *frh,
340 			       struct nlattr **tb,
341 			       struct netlink_ext_ack *extack)
342 {
343 	int err = -EINVAL;
344 	struct net *net = sock_net(skb->sk);
345 	struct fib6_rule *rule6 = (struct fib6_rule *) rule;
346 
347 	if (rule->action == FR_ACT_TO_TBL && !rule->l3mdev) {
348 		if (rule->table == RT6_TABLE_UNSPEC) {
349 			NL_SET_ERR_MSG(extack, "Invalid table");
350 			goto errout;
351 		}
352 
353 		if (fib6_new_table(net, rule->table) == NULL) {
354 			err = -ENOBUFS;
355 			goto errout;
356 		}
357 	}
358 
359 	if (frh->src_len)
360 		rule6->src.addr = nla_get_in6_addr(tb[FRA_SRC]);
361 
362 	if (frh->dst_len)
363 		rule6->dst.addr = nla_get_in6_addr(tb[FRA_DST]);
364 
365 	rule6->src.plen = frh->src_len;
366 	rule6->dst.plen = frh->dst_len;
367 	rule6->tclass = frh->tos;
368 
369 	if (fib_rule_requires_fldissect(rule))
370 		net->ipv6.fib6_rules_require_fldissect++;
371 
372 	net->ipv6.fib6_has_custom_rules = true;
373 	err = 0;
374 errout:
375 	return err;
376 }
377 
378 static int fib6_rule_delete(struct fib_rule *rule)
379 {
380 	struct net *net = rule->fr_net;
381 
382 	if (net->ipv6.fib6_rules_require_fldissect &&
383 	    fib_rule_requires_fldissect(rule))
384 		net->ipv6.fib6_rules_require_fldissect--;
385 
386 	return 0;
387 }
388 
389 static int fib6_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
390 			     struct nlattr **tb)
391 {
392 	struct fib6_rule *rule6 = (struct fib6_rule *) rule;
393 
394 	if (frh->src_len && (rule6->src.plen != frh->src_len))
395 		return 0;
396 
397 	if (frh->dst_len && (rule6->dst.plen != frh->dst_len))
398 		return 0;
399 
400 	if (frh->tos && (rule6->tclass != frh->tos))
401 		return 0;
402 
403 	if (frh->src_len &&
404 	    nla_memcmp(tb[FRA_SRC], &rule6->src.addr, sizeof(struct in6_addr)))
405 		return 0;
406 
407 	if (frh->dst_len &&
408 	    nla_memcmp(tb[FRA_DST], &rule6->dst.addr, sizeof(struct in6_addr)))
409 		return 0;
410 
411 	return 1;
412 }
413 
414 static int fib6_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
415 			  struct fib_rule_hdr *frh)
416 {
417 	struct fib6_rule *rule6 = (struct fib6_rule *) rule;
418 
419 	frh->dst_len = rule6->dst.plen;
420 	frh->src_len = rule6->src.plen;
421 	frh->tos = rule6->tclass;
422 
423 	if ((rule6->dst.plen &&
424 	     nla_put_in6_addr(skb, FRA_DST, &rule6->dst.addr)) ||
425 	    (rule6->src.plen &&
426 	     nla_put_in6_addr(skb, FRA_SRC, &rule6->src.addr)))
427 		goto nla_put_failure;
428 	return 0;
429 
430 nla_put_failure:
431 	return -ENOBUFS;
432 }
433 
434 static size_t fib6_rule_nlmsg_payload(struct fib_rule *rule)
435 {
436 	return nla_total_size(16) /* dst */
437 	       + nla_total_size(16); /* src */
438 }
439 
440 static const struct fib_rules_ops __net_initconst fib6_rules_ops_template = {
441 	.family			= AF_INET6,
442 	.rule_size		= sizeof(struct fib6_rule),
443 	.addr_size		= sizeof(struct in6_addr),
444 	.action			= fib6_rule_action,
445 	.match			= fib6_rule_match,
446 	.suppress		= fib6_rule_suppress,
447 	.configure		= fib6_rule_configure,
448 	.delete			= fib6_rule_delete,
449 	.compare		= fib6_rule_compare,
450 	.fill			= fib6_rule_fill,
451 	.nlmsg_payload		= fib6_rule_nlmsg_payload,
452 	.nlgroup		= RTNLGRP_IPV6_RULE,
453 	.policy			= fib6_rule_policy,
454 	.owner			= THIS_MODULE,
455 	.fro_net		= &init_net,
456 };
457 
458 static int __net_init fib6_rules_net_init(struct net *net)
459 {
460 	struct fib_rules_ops *ops;
461 	int err = -ENOMEM;
462 
463 	ops = fib_rules_register(&fib6_rules_ops_template, net);
464 	if (IS_ERR(ops))
465 		return PTR_ERR(ops);
466 
467 	err = fib_default_rule_add(ops, 0, RT6_TABLE_LOCAL, 0);
468 	if (err)
469 		goto out_fib6_rules_ops;
470 
471 	err = fib_default_rule_add(ops, 0x7FFE, RT6_TABLE_MAIN, 0);
472 	if (err)
473 		goto out_fib6_rules_ops;
474 
475 	net->ipv6.fib6_rules_ops = ops;
476 	net->ipv6.fib6_rules_require_fldissect = 0;
477 out:
478 	return err;
479 
480 out_fib6_rules_ops:
481 	fib_rules_unregister(ops);
482 	goto out;
483 }
484 
485 static void __net_exit fib6_rules_net_exit(struct net *net)
486 {
487 	rtnl_lock();
488 	fib_rules_unregister(net->ipv6.fib6_rules_ops);
489 	rtnl_unlock();
490 }
491 
492 static struct pernet_operations fib6_rules_net_ops = {
493 	.init = fib6_rules_net_init,
494 	.exit = fib6_rules_net_exit,
495 };
496 
497 int __init fib6_rules_init(void)
498 {
499 	return register_pernet_subsys(&fib6_rules_net_ops);
500 }
501 
502 
503 void fib6_rules_cleanup(void)
504 {
505 	unregister_pernet_subsys(&fib6_rules_net_ops);
506 }
507