xref: /linux/net/ipv6/fib6_rules.c (revision c532de5a67a70f8533d495f8f2aaa9a0491c3ad0)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * net/ipv6/fib6_rules.c	IPv6 Routing Policy Rules
4  *
5  * Copyright (C)2003-2006 Helsinki University of Technology
6  * Copyright (C)2003-2006 USAGI/WIDE Project
7  *
8  * Authors
9  *	Thomas Graf		<tgraf@suug.ch>
10  *	Ville Nuorvala		<vnuorval@tcs.hut.fi>
11  */
12 
13 #include <linux/netdevice.h>
14 #include <linux/notifier.h>
15 #include <linux/export.h>
16 #include <linux/indirect_call_wrapper.h>
17 
18 #include <net/fib_rules.h>
19 #include <net/inet_dscp.h>
20 #include <net/ipv6.h>
21 #include <net/addrconf.h>
22 #include <net/ip6_route.h>
23 #include <net/netlink.h>
24 
25 struct fib6_rule {
26 	struct fib_rule		common;
27 	struct rt6key		src;
28 	struct rt6key		dst;
29 	dscp_t			dscp;
30 	u8			dscp_full:1;	/* DSCP or TOS selector */
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->dscp)
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 		    struct netlink_ext_ack *extack)
55 {
56 	return fib_rules_dump(net, nb, AF_INET6, extack);
57 }
58 
59 unsigned int fib6_rules_seq_read(struct net *net)
60 {
61 	return fib_rules_seq_read(net, AF_INET6);
62 }
63 
64 /* called with rcu lock held; no reference taken on fib6_info */
65 int fib6_lookup(struct net *net, int oif, struct flowi6 *fl6,
66 		struct fib6_result *res, int flags)
67 {
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 			.result = res,
75 			.flags = FIB_LOOKUP_NOREF,
76 		};
77 
78 		l3mdev_update_flow(net, flowi6_to_flowi(fl6));
79 
80 		err = fib_rules_lookup(net->ipv6.fib6_rules_ops,
81 				       flowi6_to_flowi(fl6), flags, &arg);
82 	} else {
83 		err = fib6_table_lookup(net, net->ipv6.fib6_local_tbl, oif,
84 					fl6, res, flags);
85 		if (err || res->f6i == net->ipv6.fib6_null_entry)
86 			err = fib6_table_lookup(net, net->ipv6.fib6_main_tbl,
87 						oif, fl6, res, flags);
88 	}
89 
90 	return err;
91 }
92 
93 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6,
94 				   const struct sk_buff *skb,
95 				   int flags, pol_lookup_t lookup)
96 {
97 	if (net->ipv6.fib6_has_custom_rules) {
98 		struct fib6_result res = {};
99 		struct fib_lookup_arg arg = {
100 			.lookup_ptr = lookup,
101 			.lookup_data = skb,
102 			.result = &res,
103 			.flags = FIB_LOOKUP_NOREF,
104 		};
105 
106 		/* update flow if oif or iif point to device enslaved to l3mdev */
107 		l3mdev_update_flow(net, flowi6_to_flowi(fl6));
108 
109 		fib_rules_lookup(net->ipv6.fib6_rules_ops,
110 				 flowi6_to_flowi(fl6), flags, &arg);
111 
112 		if (res.rt6)
113 			return &res.rt6->dst;
114 	} else {
115 		struct rt6_info *rt;
116 
117 		rt = pol_lookup_func(lookup,
118 			     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_flags(rt, flags);
122 		rt = pol_lookup_func(lookup,
123 			     net, net->ipv6.fib6_main_tbl, fl6, skb, flags);
124 		if (rt->dst.error != -EAGAIN)
125 			return &rt->dst;
126 		ip6_rt_put_flags(rt, flags);
127 	}
128 
129 	if (!(flags & RT6_LOOKUP_F_DST_NOREF))
130 		dst_hold(&net->ipv6.ip6_null_entry->dst);
131 	return &net->ipv6.ip6_null_entry->dst;
132 }
133 
134 static int fib6_rule_saddr(struct net *net, struct fib_rule *rule, int flags,
135 			   struct flowi6 *flp6, const struct net_device *dev)
136 {
137 	struct fib6_rule *r = (struct fib6_rule *)rule;
138 
139 	/* If we need to find a source address for this traffic,
140 	 * we check the result if it meets requirement of the rule.
141 	 */
142 	if ((rule->flags & FIB_RULE_FIND_SADDR) &&
143 	    r->src.plen && !(flags & RT6_LOOKUP_F_HAS_SADDR)) {
144 		struct in6_addr saddr;
145 
146 		if (ipv6_dev_get_saddr(net, dev, &flp6->daddr,
147 				       rt6_flags2srcprefs(flags), &saddr))
148 			return -EAGAIN;
149 
150 		if (!ipv6_prefix_equal(&saddr, &r->src.addr, r->src.plen))
151 			return -EAGAIN;
152 
153 		flp6->saddr = saddr;
154 	}
155 
156 	return 0;
157 }
158 
159 static int fib6_rule_action_alt(struct fib_rule *rule, struct flowi *flp,
160 				int flags, struct fib_lookup_arg *arg)
161 {
162 	struct fib6_result *res = arg->result;
163 	struct flowi6 *flp6 = &flp->u.ip6;
164 	struct net *net = rule->fr_net;
165 	struct fib6_table *table;
166 	int err, *oif;
167 	u32 tb_id;
168 
169 	switch (rule->action) {
170 	case FR_ACT_TO_TBL:
171 		break;
172 	case FR_ACT_UNREACHABLE:
173 		return -ENETUNREACH;
174 	case FR_ACT_PROHIBIT:
175 		return -EACCES;
176 	case FR_ACT_BLACKHOLE:
177 	default:
178 		return -EINVAL;
179 	}
180 
181 	tb_id = fib_rule_get_table(rule, arg);
182 	table = fib6_get_table(net, tb_id);
183 	if (!table)
184 		return -EAGAIN;
185 
186 	oif = (int *)arg->lookup_data;
187 	err = fib6_table_lookup(net, table, *oif, flp6, res, flags);
188 	if (!err && res->f6i != net->ipv6.fib6_null_entry)
189 		err = fib6_rule_saddr(net, rule, flags, flp6,
190 				      res->nh->fib_nh_dev);
191 	else
192 		err = -EAGAIN;
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 fib6_result *res = arg->result;
201 	struct flowi6 *flp6 = &flp->u.ip6;
202 	struct rt6_info *rt = NULL;
203 	struct fib6_table *table;
204 	struct net *net = rule->fr_net;
205 	pol_lookup_t lookup = arg->lookup_ptr;
206 	int err = 0;
207 	u32 tb_id;
208 
209 	switch (rule->action) {
210 	case FR_ACT_TO_TBL:
211 		break;
212 	case FR_ACT_UNREACHABLE:
213 		err = -ENETUNREACH;
214 		rt = net->ipv6.ip6_null_entry;
215 		goto discard_pkt;
216 	default:
217 	case FR_ACT_BLACKHOLE:
218 		err = -EINVAL;
219 		rt = net->ipv6.ip6_blk_hole_entry;
220 		goto discard_pkt;
221 	case FR_ACT_PROHIBIT:
222 		err = -EACCES;
223 		rt = net->ipv6.ip6_prohibit_entry;
224 		goto discard_pkt;
225 	}
226 
227 	tb_id = fib_rule_get_table(rule, arg);
228 	table = fib6_get_table(net, tb_id);
229 	if (!table) {
230 		err = -EAGAIN;
231 		goto out;
232 	}
233 
234 	rt = pol_lookup_func(lookup,
235 			     net, table, flp6, arg->lookup_data, flags);
236 	if (rt != net->ipv6.ip6_null_entry) {
237 		struct inet6_dev *idev = ip6_dst_idev(&rt->dst);
238 
239 		if (!idev)
240 			goto again;
241 		err = fib6_rule_saddr(net, rule, flags, flp6,
242 				      idev->dev);
243 
244 		if (err == -EAGAIN)
245 			goto again;
246 
247 		err = rt->dst.error;
248 		if (err != -EAGAIN)
249 			goto out;
250 	}
251 again:
252 	ip6_rt_put_flags(rt, flags);
253 	err = -EAGAIN;
254 	rt = NULL;
255 	goto out;
256 
257 discard_pkt:
258 	if (!(flags & RT6_LOOKUP_F_DST_NOREF))
259 		dst_hold(&rt->dst);
260 out:
261 	res->rt6 = rt;
262 	return err;
263 }
264 
265 INDIRECT_CALLABLE_SCOPE int fib6_rule_action(struct fib_rule *rule,
266 					     struct flowi *flp, int flags,
267 					     struct fib_lookup_arg *arg)
268 {
269 	if (arg->lookup_ptr == fib6_table_lookup)
270 		return fib6_rule_action_alt(rule, flp, flags, arg);
271 
272 	return __fib6_rule_action(rule, flp, flags, arg);
273 }
274 
275 INDIRECT_CALLABLE_SCOPE bool fib6_rule_suppress(struct fib_rule *rule,
276 						int flags,
277 						struct fib_lookup_arg *arg)
278 {
279 	struct fib6_result *res = arg->result;
280 	struct rt6_info *rt = res->rt6;
281 	struct net_device *dev = NULL;
282 
283 	if (!rt)
284 		return false;
285 
286 	if (rt->rt6i_idev)
287 		dev = rt->rt6i_idev->dev;
288 
289 	/* do not accept result if the route does
290 	 * not meet the required prefix length
291 	 */
292 	if (rt->rt6i_dst.plen <= rule->suppress_prefixlen)
293 		goto suppress_route;
294 
295 	/* do not accept result if the route uses a device
296 	 * belonging to a forbidden interface group
297 	 */
298 	if (rule->suppress_ifgroup != -1 && dev && dev->group == rule->suppress_ifgroup)
299 		goto suppress_route;
300 
301 	return false;
302 
303 suppress_route:
304 	ip6_rt_put_flags(rt, flags);
305 	return true;
306 }
307 
308 INDIRECT_CALLABLE_SCOPE int fib6_rule_match(struct fib_rule *rule,
309 					    struct flowi *fl, int flags)
310 {
311 	struct fib6_rule *r = (struct fib6_rule *) rule;
312 	struct flowi6 *fl6 = &fl->u.ip6;
313 
314 	if (r->dst.plen &&
315 	    !ipv6_prefix_equal(&fl6->daddr, &r->dst.addr, r->dst.plen))
316 		return 0;
317 
318 	/*
319 	 * If FIB_RULE_FIND_SADDR is set and we do not have a
320 	 * source address for the traffic, we defer check for
321 	 * source address.
322 	 */
323 	if (r->src.plen) {
324 		if (flags & RT6_LOOKUP_F_HAS_SADDR) {
325 			if (!ipv6_prefix_equal(&fl6->saddr, &r->src.addr,
326 					       r->src.plen))
327 				return 0;
328 		} else if (!(r->common.flags & FIB_RULE_FIND_SADDR))
329 			return 0;
330 	}
331 
332 	if (r->dscp && r->dscp != ip6_dscp(fl6->flowlabel))
333 		return 0;
334 
335 	if (rule->ip_proto && (rule->ip_proto != fl6->flowi6_proto))
336 		return 0;
337 
338 	if (fib_rule_port_range_set(&rule->sport_range) &&
339 	    !fib_rule_port_inrange(&rule->sport_range, fl6->fl6_sport))
340 		return 0;
341 
342 	if (fib_rule_port_range_set(&rule->dport_range) &&
343 	    !fib_rule_port_inrange(&rule->dport_range, fl6->fl6_dport))
344 		return 0;
345 
346 	return 1;
347 }
348 
349 static int fib6_nl2rule_dscp(const struct nlattr *nla, struct fib6_rule *rule6,
350 			     struct netlink_ext_ack *extack)
351 {
352 	if (rule6->dscp) {
353 		NL_SET_ERR_MSG(extack, "Cannot specify both TOS and DSCP");
354 		return -EINVAL;
355 	}
356 
357 	rule6->dscp = inet_dsfield_to_dscp(nla_get_u8(nla) << 2);
358 	rule6->dscp_full = true;
359 
360 	return 0;
361 }
362 
363 static int fib6_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
364 			       struct fib_rule_hdr *frh,
365 			       struct nlattr **tb,
366 			       struct netlink_ext_ack *extack)
367 {
368 	int err = -EINVAL;
369 	struct net *net = sock_net(skb->sk);
370 	struct fib6_rule *rule6 = (struct fib6_rule *) rule;
371 
372 	if (!inet_validate_dscp(frh->tos)) {
373 		NL_SET_ERR_MSG(extack,
374 			       "Invalid dsfield (tos): ECN bits must be 0");
375 		goto errout;
376 	}
377 	rule6->dscp = inet_dsfield_to_dscp(frh->tos);
378 
379 	if (tb[FRA_DSCP] && fib6_nl2rule_dscp(tb[FRA_DSCP], rule6, extack) < 0)
380 		goto errout;
381 
382 	if (rule->action == FR_ACT_TO_TBL && !rule->l3mdev) {
383 		if (rule->table == RT6_TABLE_UNSPEC) {
384 			NL_SET_ERR_MSG(extack, "Invalid table");
385 			goto errout;
386 		}
387 
388 		if (fib6_new_table(net, rule->table) == NULL) {
389 			err = -ENOBUFS;
390 			goto errout;
391 		}
392 	}
393 
394 	if (frh->src_len)
395 		rule6->src.addr = nla_get_in6_addr(tb[FRA_SRC]);
396 
397 	if (frh->dst_len)
398 		rule6->dst.addr = nla_get_in6_addr(tb[FRA_DST]);
399 
400 	rule6->src.plen = frh->src_len;
401 	rule6->dst.plen = frh->dst_len;
402 
403 	if (fib_rule_requires_fldissect(rule))
404 		net->ipv6.fib6_rules_require_fldissect++;
405 
406 	net->ipv6.fib6_has_custom_rules = true;
407 	err = 0;
408 errout:
409 	return err;
410 }
411 
412 static int fib6_rule_delete(struct fib_rule *rule)
413 {
414 	struct net *net = rule->fr_net;
415 
416 	if (net->ipv6.fib6_rules_require_fldissect &&
417 	    fib_rule_requires_fldissect(rule))
418 		net->ipv6.fib6_rules_require_fldissect--;
419 
420 	return 0;
421 }
422 
423 static int fib6_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
424 			     struct nlattr **tb)
425 {
426 	struct fib6_rule *rule6 = (struct fib6_rule *) rule;
427 
428 	if (frh->src_len && (rule6->src.plen != frh->src_len))
429 		return 0;
430 
431 	if (frh->dst_len && (rule6->dst.plen != frh->dst_len))
432 		return 0;
433 
434 	if (frh->tos &&
435 	    (rule6->dscp_full ||
436 	     inet_dscp_to_dsfield(rule6->dscp) != frh->tos))
437 		return 0;
438 
439 	if (tb[FRA_DSCP]) {
440 		dscp_t dscp;
441 
442 		dscp = inet_dsfield_to_dscp(nla_get_u8(tb[FRA_DSCP]) << 2);
443 		if (!rule6->dscp_full || rule6->dscp != dscp)
444 			return 0;
445 	}
446 
447 	if (frh->src_len &&
448 	    nla_memcmp(tb[FRA_SRC], &rule6->src.addr, sizeof(struct in6_addr)))
449 		return 0;
450 
451 	if (frh->dst_len &&
452 	    nla_memcmp(tb[FRA_DST], &rule6->dst.addr, sizeof(struct in6_addr)))
453 		return 0;
454 
455 	return 1;
456 }
457 
458 static int fib6_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
459 			  struct fib_rule_hdr *frh)
460 {
461 	struct fib6_rule *rule6 = (struct fib6_rule *) rule;
462 
463 	frh->dst_len = rule6->dst.plen;
464 	frh->src_len = rule6->src.plen;
465 
466 	if (rule6->dscp_full) {
467 		frh->tos = 0;
468 		if (nla_put_u8(skb, FRA_DSCP,
469 			       inet_dscp_to_dsfield(rule6->dscp) >> 2))
470 			goto nla_put_failure;
471 	} else {
472 		frh->tos = inet_dscp_to_dsfield(rule6->dscp);
473 	}
474 
475 	if ((rule6->dst.plen &&
476 	     nla_put_in6_addr(skb, FRA_DST, &rule6->dst.addr)) ||
477 	    (rule6->src.plen &&
478 	     nla_put_in6_addr(skb, FRA_SRC, &rule6->src.addr)))
479 		goto nla_put_failure;
480 	return 0;
481 
482 nla_put_failure:
483 	return -ENOBUFS;
484 }
485 
486 static size_t fib6_rule_nlmsg_payload(struct fib_rule *rule)
487 {
488 	return nla_total_size(16) /* dst */
489 	       + nla_total_size(16) /* src */
490 	       + nla_total_size(1); /* dscp */
491 }
492 
493 static void fib6_rule_flush_cache(struct fib_rules_ops *ops)
494 {
495 	rt_genid_bump_ipv6(ops->fro_net);
496 }
497 
498 static const struct fib_rules_ops __net_initconst fib6_rules_ops_template = {
499 	.family			= AF_INET6,
500 	.rule_size		= sizeof(struct fib6_rule),
501 	.addr_size		= sizeof(struct in6_addr),
502 	.action			= fib6_rule_action,
503 	.match			= fib6_rule_match,
504 	.suppress		= fib6_rule_suppress,
505 	.configure		= fib6_rule_configure,
506 	.delete			= fib6_rule_delete,
507 	.compare		= fib6_rule_compare,
508 	.fill			= fib6_rule_fill,
509 	.nlmsg_payload		= fib6_rule_nlmsg_payload,
510 	.flush_cache		= fib6_rule_flush_cache,
511 	.nlgroup		= RTNLGRP_IPV6_RULE,
512 	.owner			= THIS_MODULE,
513 	.fro_net		= &init_net,
514 };
515 
516 static int __net_init fib6_rules_net_init(struct net *net)
517 {
518 	struct fib_rules_ops *ops;
519 	int err;
520 
521 	ops = fib_rules_register(&fib6_rules_ops_template, net);
522 	if (IS_ERR(ops))
523 		return PTR_ERR(ops);
524 
525 	err = fib_default_rule_add(ops, 0, RT6_TABLE_LOCAL);
526 	if (err)
527 		goto out_fib6_rules_ops;
528 
529 	err = fib_default_rule_add(ops, 0x7FFE, RT6_TABLE_MAIN);
530 	if (err)
531 		goto out_fib6_rules_ops;
532 
533 	net->ipv6.fib6_rules_ops = ops;
534 	net->ipv6.fib6_rules_require_fldissect = 0;
535 out:
536 	return err;
537 
538 out_fib6_rules_ops:
539 	fib_rules_unregister(ops);
540 	goto out;
541 }
542 
543 static void __net_exit fib6_rules_net_exit_batch(struct list_head *net_list)
544 {
545 	struct net *net;
546 
547 	rtnl_lock();
548 	list_for_each_entry(net, net_list, exit_list) {
549 		fib_rules_unregister(net->ipv6.fib6_rules_ops);
550 		cond_resched();
551 	}
552 	rtnl_unlock();
553 }
554 
555 static struct pernet_operations fib6_rules_net_ops = {
556 	.init = fib6_rules_net_init,
557 	.exit_batch = fib6_rules_net_exit_batch,
558 };
559 
560 int __init fib6_rules_init(void)
561 {
562 	return register_pernet_subsys(&fib6_rules_net_ops);
563 }
564 
565 
566 void fib6_rules_cleanup(void)
567 {
568 	unregister_pernet_subsys(&fib6_rules_net_ops);
569 }
570