xref: /linux/net/ipv6/fib6_rules.c (revision e814f3fd16acfb7f9966773953de8f740a1e3202)
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 	__be32			flowlabel;
30 	__be32			flowlabel_mask;
31 	dscp_t			dscp;
32 	u8			dscp_full:1;	/* DSCP or TOS selector */
33 };
34 
35 static bool fib6_rule_matchall(const struct fib_rule *rule)
36 {
37 	struct fib6_rule *r = container_of(rule, struct fib6_rule, common);
38 
39 	if (r->dst.plen || r->src.plen || r->dscp || r->flowlabel_mask)
40 		return false;
41 	return fib_rule_matchall(rule);
42 }
43 
44 bool fib6_rule_default(const struct fib_rule *rule)
45 {
46 	if (!fib6_rule_matchall(rule) || rule->action != FR_ACT_TO_TBL ||
47 	    rule->l3mdev)
48 		return false;
49 	if (rule->table != RT6_TABLE_LOCAL && rule->table != RT6_TABLE_MAIN)
50 		return false;
51 	return true;
52 }
53 EXPORT_SYMBOL_GPL(fib6_rule_default);
54 
55 int fib6_rules_dump(struct net *net, struct notifier_block *nb,
56 		    struct netlink_ext_ack *extack)
57 {
58 	return fib_rules_dump(net, nb, AF_INET6, extack);
59 }
60 
61 unsigned int fib6_rules_seq_read(const struct net *net)
62 {
63 	return fib_rules_seq_read(net, AF_INET6);
64 }
65 
66 /* called with rcu lock held; no reference taken on fib6_info */
67 int fib6_lookup(struct net *net, int oif, struct flowi6 *fl6,
68 		struct fib6_result *res, int flags)
69 {
70 	int err;
71 
72 	if (net->ipv6.fib6_has_custom_rules) {
73 		struct fib_lookup_arg arg = {
74 			.lookup_ptr = fib6_table_lookup,
75 			.lookup_data = &oif,
76 			.result = res,
77 			.flags = FIB_LOOKUP_NOREF,
78 		};
79 
80 		l3mdev_update_flow(net, flowi6_to_flowi(fl6));
81 
82 		err = fib_rules_lookup(net->ipv6.fib6_rules_ops,
83 				       flowi6_to_flowi(fl6), flags, &arg);
84 	} else {
85 		err = fib6_table_lookup(net, net->ipv6.fib6_local_tbl, oif,
86 					fl6, res, flags);
87 		if (err || res->f6i == net->ipv6.fib6_null_entry)
88 			err = fib6_table_lookup(net, net->ipv6.fib6_main_tbl,
89 						oif, fl6, res, flags);
90 	}
91 
92 	return err;
93 }
94 
95 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6,
96 				   const struct sk_buff *skb,
97 				   int flags, pol_lookup_t lookup)
98 {
99 	if (net->ipv6.fib6_has_custom_rules) {
100 		struct fib6_result res = {};
101 		struct fib_lookup_arg arg = {
102 			.lookup_ptr = lookup,
103 			.lookup_data = skb,
104 			.result = &res,
105 			.flags = FIB_LOOKUP_NOREF,
106 		};
107 
108 		/* update flow if oif or iif point to device enslaved to l3mdev */
109 		l3mdev_update_flow(net, flowi6_to_flowi(fl6));
110 
111 		fib_rules_lookup(net->ipv6.fib6_rules_ops,
112 				 flowi6_to_flowi(fl6), flags, &arg);
113 
114 		if (res.rt6)
115 			return &res.rt6->dst;
116 	} else {
117 		struct rt6_info *rt;
118 
119 		rt = pol_lookup_func(lookup,
120 			     net, net->ipv6.fib6_local_tbl, fl6, skb, flags);
121 		if (rt != net->ipv6.ip6_null_entry && rt->dst.error != -EAGAIN)
122 			return &rt->dst;
123 		ip6_rt_put_flags(rt, flags);
124 		rt = pol_lookup_func(lookup,
125 			     net, net->ipv6.fib6_main_tbl, fl6, skb, flags);
126 		if (rt->dst.error != -EAGAIN)
127 			return &rt->dst;
128 		ip6_rt_put_flags(rt, flags);
129 	}
130 
131 	if (!(flags & RT6_LOOKUP_F_DST_NOREF))
132 		dst_hold(&net->ipv6.ip6_null_entry->dst);
133 	return &net->ipv6.ip6_null_entry->dst;
134 }
135 
136 static int fib6_rule_saddr(struct net *net, struct fib_rule *rule, int flags,
137 			   struct flowi6 *flp6, const struct net_device *dev)
138 {
139 	struct fib6_rule *r = (struct fib6_rule *)rule;
140 
141 	/* If we need to find a source address for this traffic,
142 	 * we check the result if it meets requirement of the rule.
143 	 */
144 	if ((rule->flags & FIB_RULE_FIND_SADDR) &&
145 	    r->src.plen && !(flags & RT6_LOOKUP_F_HAS_SADDR)) {
146 		struct in6_addr saddr;
147 
148 		if (ipv6_dev_get_saddr(net, dev, &flp6->daddr,
149 				       rt6_flags2srcprefs(flags), &saddr))
150 			return -EAGAIN;
151 
152 		if (!ipv6_prefix_equal(&saddr, &r->src.addr, r->src.plen))
153 			return -EAGAIN;
154 
155 		flp6->saddr = saddr;
156 	}
157 
158 	return 0;
159 }
160 
161 static int fib6_rule_action_alt(struct fib_rule *rule, struct flowi *flp,
162 				int flags, struct fib_lookup_arg *arg)
163 {
164 	struct fib6_result *res = arg->result;
165 	struct flowi6 *flp6 = &flp->u.ip6;
166 	struct net *net = rule->fr_net;
167 	struct fib6_table *table;
168 	int err, *oif;
169 	u32 tb_id;
170 
171 	switch (rule->action) {
172 	case FR_ACT_TO_TBL:
173 		break;
174 	case FR_ACT_UNREACHABLE:
175 		return -ENETUNREACH;
176 	case FR_ACT_PROHIBIT:
177 		return -EACCES;
178 	case FR_ACT_BLACKHOLE:
179 	default:
180 		return -EINVAL;
181 	}
182 
183 	tb_id = fib_rule_get_table(rule, arg);
184 	table = fib6_get_table(net, tb_id);
185 	if (!table)
186 		return -EAGAIN;
187 
188 	oif = (int *)arg->lookup_data;
189 	err = fib6_table_lookup(net, table, *oif, flp6, res, flags);
190 	if (!err && res->f6i != net->ipv6.fib6_null_entry)
191 		err = fib6_rule_saddr(net, rule, flags, flp6,
192 				      res->nh->fib_nh_dev);
193 	else
194 		err = -EAGAIN;
195 
196 	return err;
197 }
198 
199 static int __fib6_rule_action(struct fib_rule *rule, struct flowi *flp,
200 			      int flags, struct fib_lookup_arg *arg)
201 {
202 	struct fib6_result *res = arg->result;
203 	struct flowi6 *flp6 = &flp->u.ip6;
204 	struct rt6_info *rt = NULL;
205 	struct fib6_table *table;
206 	struct net *net = rule->fr_net;
207 	pol_lookup_t lookup = arg->lookup_ptr;
208 	int err = 0;
209 	u32 tb_id;
210 
211 	switch (rule->action) {
212 	case FR_ACT_TO_TBL:
213 		break;
214 	case FR_ACT_UNREACHABLE:
215 		err = -ENETUNREACH;
216 		rt = net->ipv6.ip6_null_entry;
217 		goto discard_pkt;
218 	default:
219 	case FR_ACT_BLACKHOLE:
220 		err = -EINVAL;
221 		rt = net->ipv6.ip6_blk_hole_entry;
222 		goto discard_pkt;
223 	case FR_ACT_PROHIBIT:
224 		err = -EACCES;
225 		rt = net->ipv6.ip6_prohibit_entry;
226 		goto discard_pkt;
227 	}
228 
229 	tb_id = fib_rule_get_table(rule, arg);
230 	table = fib6_get_table(net, tb_id);
231 	if (!table) {
232 		err = -EAGAIN;
233 		goto out;
234 	}
235 
236 	rt = pol_lookup_func(lookup,
237 			     net, table, flp6, arg->lookup_data, flags);
238 	if (rt != net->ipv6.ip6_null_entry) {
239 		struct inet6_dev *idev = ip6_dst_idev(&rt->dst);
240 
241 		if (!idev)
242 			goto again;
243 		err = fib6_rule_saddr(net, rule, flags, flp6,
244 				      idev->dev);
245 
246 		if (err == -EAGAIN)
247 			goto again;
248 
249 		err = rt->dst.error;
250 		if (err != -EAGAIN)
251 			goto out;
252 	}
253 again:
254 	ip6_rt_put_flags(rt, flags);
255 	err = -EAGAIN;
256 	rt = NULL;
257 	goto out;
258 
259 discard_pkt:
260 	if (!(flags & RT6_LOOKUP_F_DST_NOREF))
261 		dst_hold(&rt->dst);
262 out:
263 	res->rt6 = rt;
264 	return err;
265 }
266 
267 INDIRECT_CALLABLE_SCOPE int fib6_rule_action(struct fib_rule *rule,
268 					     struct flowi *flp, int flags,
269 					     struct fib_lookup_arg *arg)
270 {
271 	if (arg->lookup_ptr == fib6_table_lookup)
272 		return fib6_rule_action_alt(rule, flp, flags, arg);
273 
274 	return __fib6_rule_action(rule, flp, flags, arg);
275 }
276 
277 INDIRECT_CALLABLE_SCOPE bool fib6_rule_suppress(struct fib_rule *rule,
278 						int flags,
279 						struct fib_lookup_arg *arg)
280 {
281 	struct fib6_result *res = arg->result;
282 	struct rt6_info *rt = res->rt6;
283 	struct net_device *dev = NULL;
284 
285 	if (!rt)
286 		return false;
287 
288 	if (rt->rt6i_idev)
289 		dev = rt->rt6i_idev->dev;
290 
291 	/* do not accept result if the route does
292 	 * not meet the required prefix length
293 	 */
294 	if (rt->rt6i_dst.plen <= rule->suppress_prefixlen)
295 		goto suppress_route;
296 
297 	/* do not accept result if the route uses a device
298 	 * belonging to a forbidden interface group
299 	 */
300 	if (rule->suppress_ifgroup != -1 && dev && dev->group == rule->suppress_ifgroup)
301 		goto suppress_route;
302 
303 	return false;
304 
305 suppress_route:
306 	ip6_rt_put_flags(rt, flags);
307 	return true;
308 }
309 
310 INDIRECT_CALLABLE_SCOPE int fib6_rule_match(struct fib_rule *rule,
311 					    struct flowi *fl, int flags)
312 {
313 	struct fib6_rule *r = (struct fib6_rule *) rule;
314 	struct flowi6 *fl6 = &fl->u.ip6;
315 
316 	if (r->dst.plen &&
317 	    !ipv6_prefix_equal(&fl6->daddr, &r->dst.addr, r->dst.plen))
318 		return 0;
319 
320 	/*
321 	 * If FIB_RULE_FIND_SADDR is set and we do not have a
322 	 * source address for the traffic, we defer check for
323 	 * source address.
324 	 */
325 	if (r->src.plen) {
326 		if (flags & RT6_LOOKUP_F_HAS_SADDR) {
327 			if (!ipv6_prefix_equal(&fl6->saddr, &r->src.addr,
328 					       r->src.plen))
329 				return 0;
330 		} else if (!(r->common.flags & FIB_RULE_FIND_SADDR))
331 			return 0;
332 	}
333 
334 	if (r->dscp && r->dscp != ip6_dscp(fl6->flowlabel))
335 		return 0;
336 
337 	if ((r->flowlabel ^ flowi6_get_flowlabel(fl6)) & r->flowlabel_mask)
338 		return 0;
339 
340 	if (rule->ip_proto && (rule->ip_proto != fl6->flowi6_proto))
341 		return 0;
342 
343 	if (fib_rule_port_range_set(&rule->sport_range) &&
344 	    !fib_rule_port_inrange(&rule->sport_range, fl6->fl6_sport))
345 		return 0;
346 
347 	if (fib_rule_port_range_set(&rule->dport_range) &&
348 	    !fib_rule_port_inrange(&rule->dport_range, fl6->fl6_dport))
349 		return 0;
350 
351 	return 1;
352 }
353 
354 static int fib6_nl2rule_dscp(const struct nlattr *nla, struct fib6_rule *rule6,
355 			     struct netlink_ext_ack *extack)
356 {
357 	if (rule6->dscp) {
358 		NL_SET_ERR_MSG(extack, "Cannot specify both TOS and DSCP");
359 		return -EINVAL;
360 	}
361 
362 	rule6->dscp = inet_dsfield_to_dscp(nla_get_u8(nla) << 2);
363 	rule6->dscp_full = true;
364 
365 	return 0;
366 }
367 
368 static int fib6_nl2rule_flowlabel(struct nlattr **tb, struct fib6_rule *rule6,
369 				  struct netlink_ext_ack *extack)
370 {
371 	__be32 flowlabel, flowlabel_mask;
372 
373 	if (NL_REQ_ATTR_CHECK(extack, NULL, tb, FRA_FLOWLABEL) ||
374 	    NL_REQ_ATTR_CHECK(extack, NULL, tb, FRA_FLOWLABEL_MASK))
375 		return -EINVAL;
376 
377 	flowlabel = nla_get_be32(tb[FRA_FLOWLABEL]);
378 	flowlabel_mask = nla_get_be32(tb[FRA_FLOWLABEL_MASK]);
379 
380 	if (flowlabel_mask & ~IPV6_FLOWLABEL_MASK) {
381 		NL_SET_ERR_MSG_ATTR(extack, tb[FRA_FLOWLABEL_MASK],
382 				    "Invalid flow label mask");
383 		return -EINVAL;
384 	}
385 
386 	if (flowlabel & ~flowlabel_mask) {
387 		NL_SET_ERR_MSG(extack, "Flow label and mask do not match");
388 		return -EINVAL;
389 	}
390 
391 	rule6->flowlabel = flowlabel;
392 	rule6->flowlabel_mask = flowlabel_mask;
393 
394 	return 0;
395 }
396 
397 static int fib6_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
398 			       struct fib_rule_hdr *frh,
399 			       struct nlattr **tb,
400 			       struct netlink_ext_ack *extack)
401 {
402 	int err = -EINVAL;
403 	struct net *net = sock_net(skb->sk);
404 	struct fib6_rule *rule6 = (struct fib6_rule *) rule;
405 
406 	if (!inet_validate_dscp(frh->tos)) {
407 		NL_SET_ERR_MSG(extack,
408 			       "Invalid dsfield (tos): ECN bits must be 0");
409 		goto errout;
410 	}
411 	rule6->dscp = inet_dsfield_to_dscp(frh->tos);
412 
413 	if (tb[FRA_DSCP] && fib6_nl2rule_dscp(tb[FRA_DSCP], rule6, extack) < 0)
414 		goto errout;
415 
416 	if ((tb[FRA_FLOWLABEL] || tb[FRA_FLOWLABEL_MASK]) &&
417 	    fib6_nl2rule_flowlabel(tb, rule6, extack) < 0)
418 		goto errout;
419 
420 	if (rule->action == FR_ACT_TO_TBL && !rule->l3mdev) {
421 		if (rule->table == RT6_TABLE_UNSPEC) {
422 			NL_SET_ERR_MSG(extack, "Invalid table");
423 			goto errout;
424 		}
425 
426 		if (fib6_new_table(net, rule->table) == NULL) {
427 			err = -ENOBUFS;
428 			goto errout;
429 		}
430 	}
431 
432 	if (frh->src_len)
433 		rule6->src.addr = nla_get_in6_addr(tb[FRA_SRC]);
434 
435 	if (frh->dst_len)
436 		rule6->dst.addr = nla_get_in6_addr(tb[FRA_DST]);
437 
438 	rule6->src.plen = frh->src_len;
439 	rule6->dst.plen = frh->dst_len;
440 
441 	if (fib_rule_requires_fldissect(rule))
442 		net->ipv6.fib6_rules_require_fldissect++;
443 
444 	net->ipv6.fib6_has_custom_rules = true;
445 	err = 0;
446 errout:
447 	return err;
448 }
449 
450 static int fib6_rule_delete(struct fib_rule *rule)
451 {
452 	struct net *net = rule->fr_net;
453 
454 	if (net->ipv6.fib6_rules_require_fldissect &&
455 	    fib_rule_requires_fldissect(rule))
456 		net->ipv6.fib6_rules_require_fldissect--;
457 
458 	return 0;
459 }
460 
461 static int fib6_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
462 			     struct nlattr **tb)
463 {
464 	struct fib6_rule *rule6 = (struct fib6_rule *) rule;
465 
466 	if (frh->src_len && (rule6->src.plen != frh->src_len))
467 		return 0;
468 
469 	if (frh->dst_len && (rule6->dst.plen != frh->dst_len))
470 		return 0;
471 
472 	if (frh->tos &&
473 	    (rule6->dscp_full ||
474 	     inet_dscp_to_dsfield(rule6->dscp) != frh->tos))
475 		return 0;
476 
477 	if (tb[FRA_DSCP]) {
478 		dscp_t dscp;
479 
480 		dscp = inet_dsfield_to_dscp(nla_get_u8(tb[FRA_DSCP]) << 2);
481 		if (!rule6->dscp_full || rule6->dscp != dscp)
482 			return 0;
483 	}
484 
485 	if (tb[FRA_FLOWLABEL] &&
486 	    nla_get_be32(tb[FRA_FLOWLABEL]) != rule6->flowlabel)
487 		return 0;
488 
489 	if (tb[FRA_FLOWLABEL_MASK] &&
490 	    nla_get_be32(tb[FRA_FLOWLABEL_MASK]) != rule6->flowlabel_mask)
491 		return 0;
492 
493 	if (frh->src_len &&
494 	    nla_memcmp(tb[FRA_SRC], &rule6->src.addr, sizeof(struct in6_addr)))
495 		return 0;
496 
497 	if (frh->dst_len &&
498 	    nla_memcmp(tb[FRA_DST], &rule6->dst.addr, sizeof(struct in6_addr)))
499 		return 0;
500 
501 	return 1;
502 }
503 
504 static int fib6_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
505 			  struct fib_rule_hdr *frh)
506 {
507 	struct fib6_rule *rule6 = (struct fib6_rule *) rule;
508 
509 	frh->dst_len = rule6->dst.plen;
510 	frh->src_len = rule6->src.plen;
511 
512 	if (rule6->dscp_full) {
513 		frh->tos = 0;
514 		if (nla_put_u8(skb, FRA_DSCP,
515 			       inet_dscp_to_dsfield(rule6->dscp) >> 2))
516 			goto nla_put_failure;
517 	} else {
518 		frh->tos = inet_dscp_to_dsfield(rule6->dscp);
519 	}
520 
521 	if (rule6->flowlabel_mask &&
522 	    (nla_put_be32(skb, FRA_FLOWLABEL, rule6->flowlabel) ||
523 	     nla_put_be32(skb, FRA_FLOWLABEL_MASK, rule6->flowlabel_mask)))
524 		goto nla_put_failure;
525 
526 	if ((rule6->dst.plen &&
527 	     nla_put_in6_addr(skb, FRA_DST, &rule6->dst.addr)) ||
528 	    (rule6->src.plen &&
529 	     nla_put_in6_addr(skb, FRA_SRC, &rule6->src.addr)))
530 		goto nla_put_failure;
531 	return 0;
532 
533 nla_put_failure:
534 	return -ENOBUFS;
535 }
536 
537 static size_t fib6_rule_nlmsg_payload(struct fib_rule *rule)
538 {
539 	return nla_total_size(16) /* dst */
540 	       + nla_total_size(16) /* src */
541 	       + nla_total_size(1) /* dscp */
542 	       + nla_total_size(4) /* flowlabel */
543 	       + nla_total_size(4); /* flowlabel mask */
544 }
545 
546 static void fib6_rule_flush_cache(struct fib_rules_ops *ops)
547 {
548 	rt_genid_bump_ipv6(ops->fro_net);
549 }
550 
551 static const struct fib_rules_ops __net_initconst fib6_rules_ops_template = {
552 	.family			= AF_INET6,
553 	.rule_size		= sizeof(struct fib6_rule),
554 	.addr_size		= sizeof(struct in6_addr),
555 	.action			= fib6_rule_action,
556 	.match			= fib6_rule_match,
557 	.suppress		= fib6_rule_suppress,
558 	.configure		= fib6_rule_configure,
559 	.delete			= fib6_rule_delete,
560 	.compare		= fib6_rule_compare,
561 	.fill			= fib6_rule_fill,
562 	.nlmsg_payload		= fib6_rule_nlmsg_payload,
563 	.flush_cache		= fib6_rule_flush_cache,
564 	.nlgroup		= RTNLGRP_IPV6_RULE,
565 	.owner			= THIS_MODULE,
566 	.fro_net		= &init_net,
567 };
568 
569 static int __net_init fib6_rules_net_init(struct net *net)
570 {
571 	struct fib_rules_ops *ops;
572 	int err;
573 
574 	ops = fib_rules_register(&fib6_rules_ops_template, net);
575 	if (IS_ERR(ops))
576 		return PTR_ERR(ops);
577 
578 	err = fib_default_rule_add(ops, 0, RT6_TABLE_LOCAL);
579 	if (err)
580 		goto out_fib6_rules_ops;
581 
582 	err = fib_default_rule_add(ops, 0x7FFE, RT6_TABLE_MAIN);
583 	if (err)
584 		goto out_fib6_rules_ops;
585 
586 	net->ipv6.fib6_rules_ops = ops;
587 	net->ipv6.fib6_rules_require_fldissect = 0;
588 out:
589 	return err;
590 
591 out_fib6_rules_ops:
592 	fib_rules_unregister(ops);
593 	goto out;
594 }
595 
596 static void __net_exit fib6_rules_net_exit_batch(struct list_head *net_list)
597 {
598 	struct net *net;
599 
600 	rtnl_lock();
601 	list_for_each_entry(net, net_list, exit_list) {
602 		fib_rules_unregister(net->ipv6.fib6_rules_ops);
603 		cond_resched();
604 	}
605 	rtnl_unlock();
606 }
607 
608 static struct pernet_operations fib6_rules_net_ops = {
609 	.init = fib6_rules_net_init,
610 	.exit_batch = fib6_rules_net_exit_batch,
611 };
612 
613 int __init fib6_rules_init(void)
614 {
615 	return register_pernet_subsys(&fib6_rules_net_ops);
616 }
617 
618 
619 void fib6_rules_cleanup(void)
620 {
621 	unregister_pernet_subsys(&fib6_rules_net_ops);
622 }
623