1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * net/core/fib_rules.c Generic Routing Rules
4 *
5 * Authors: Thomas Graf <tgraf@suug.ch>
6 */
7
8 #include <linux/types.h>
9 #include <linux/kernel.h>
10 #include <linux/slab.h>
11 #include <linux/list.h>
12 #include <linux/module.h>
13 #include <net/net_namespace.h>
14 #include <net/sock.h>
15 #include <net/fib_rules.h>
16 #include <net/ip_tunnels.h>
17 #include <linux/indirect_call_wrapper.h>
18
19 #if defined(CONFIG_IPV6) && defined(CONFIG_IPV6_MULTIPLE_TABLES)
20 #ifdef CONFIG_IP_MULTIPLE_TABLES
21 #define INDIRECT_CALL_MT(f, f2, f1, ...) \
22 INDIRECT_CALL_INET(f, f2, f1, __VA_ARGS__)
23 #else
24 #define INDIRECT_CALL_MT(f, f2, f1, ...) INDIRECT_CALL_1(f, f2, __VA_ARGS__)
25 #endif
26 #elif defined(CONFIG_IP_MULTIPLE_TABLES)
27 #define INDIRECT_CALL_MT(f, f2, f1, ...) INDIRECT_CALL_1(f, f1, __VA_ARGS__)
28 #else
29 #define INDIRECT_CALL_MT(f, f2, f1, ...) f(__VA_ARGS__)
30 #endif
31
32 static const struct fib_kuid_range fib_kuid_range_unset = {
33 KUIDT_INIT(0),
34 KUIDT_INIT(~0),
35 };
36
37 bool fib_rule_matchall(const struct fib_rule *rule)
38 {
39 if (rule->iifindex || rule->oifindex || rule->mark || rule->tun_id ||
40 rule->flags)
41 return false;
42 if (rule->suppress_ifgroup != -1 || rule->suppress_prefixlen != -1)
43 return false;
44 if (!uid_eq(rule->uid_range.start, fib_kuid_range_unset.start) ||
45 !uid_eq(rule->uid_range.end, fib_kuid_range_unset.end))
46 return false;
47 if (fib_rule_port_range_set(&rule->sport_range))
48 return false;
49 if (fib_rule_port_range_set(&rule->dport_range))
50 return false;
51 return true;
52 }
53 EXPORT_SYMBOL_GPL(fib_rule_matchall);
54
55 int fib_default_rule_add(struct fib_rules_ops *ops,
56 u32 pref, u32 table)
57 {
58 struct fib_rule *r;
59
60 r = kzalloc(ops->rule_size, GFP_KERNEL_ACCOUNT);
61 if (r == NULL)
62 return -ENOMEM;
63
64 refcount_set(&r->refcnt, 1);
65 r->action = FR_ACT_TO_TBL;
66 r->pref = pref;
67 r->table = table;
68 r->proto = RTPROT_KERNEL;
69 r->fr_net = ops->fro_net;
70 r->uid_range = fib_kuid_range_unset;
71
72 r->suppress_prefixlen = -1;
73 r->suppress_ifgroup = -1;
74
75 /* The lock is not required here, the list in unreacheable
76 * at the moment this function is called */
77 list_add_tail(&r->list, &ops->rules_list);
78 return 0;
79 }
80 EXPORT_SYMBOL(fib_default_rule_add);
81
82 static u32 fib_default_rule_pref(struct fib_rules_ops *ops)
83 {
84 struct list_head *pos;
85 struct fib_rule *rule;
86
87 if (!list_empty(&ops->rules_list)) {
88 pos = ops->rules_list.next;
89 if (pos->next != &ops->rules_list) {
90 rule = list_entry(pos->next, struct fib_rule, list);
91 if (rule->pref)
92 return rule->pref - 1;
93 }
94 }
95
96 return 0;
97 }
98
99 static void notify_rule_change(int event, struct fib_rule *rule,
100 struct fib_rules_ops *ops, struct nlmsghdr *nlh,
101 u32 pid);
102
103 static struct fib_rules_ops *lookup_rules_ops(struct net *net, int family)
104 {
105 struct fib_rules_ops *ops;
106
107 rcu_read_lock();
108 list_for_each_entry_rcu(ops, &net->rules_ops, list) {
109 if (ops->family == family) {
110 if (!try_module_get(ops->owner))
111 ops = NULL;
112 rcu_read_unlock();
113 return ops;
114 }
115 }
116 rcu_read_unlock();
117
118 return NULL;
119 }
120
121 static void rules_ops_put(struct fib_rules_ops *ops)
122 {
123 if (ops)
124 module_put(ops->owner);
125 }
126
127 static void flush_route_cache(struct fib_rules_ops *ops)
128 {
129 if (ops->flush_cache)
130 ops->flush_cache(ops);
131 }
132
133 static int __fib_rules_register(struct fib_rules_ops *ops)
134 {
135 int err = -EEXIST;
136 struct fib_rules_ops *o;
137 struct net *net;
138
139 net = ops->fro_net;
140
141 if (ops->rule_size < sizeof(struct fib_rule))
142 return -EINVAL;
143
144 if (ops->match == NULL || ops->configure == NULL ||
145 ops->compare == NULL || ops->fill == NULL ||
146 ops->action == NULL)
147 return -EINVAL;
148
149 spin_lock(&net->rules_mod_lock);
150 list_for_each_entry(o, &net->rules_ops, list)
151 if (ops->family == o->family)
152 goto errout;
153
154 list_add_tail_rcu(&ops->list, &net->rules_ops);
155 err = 0;
156 errout:
157 spin_unlock(&net->rules_mod_lock);
158
159 return err;
160 }
161
162 struct fib_rules_ops *
163 fib_rules_register(const struct fib_rules_ops *tmpl, struct net *net)
164 {
165 struct fib_rules_ops *ops;
166 int err;
167
168 ops = kmemdup(tmpl, sizeof(*ops), GFP_KERNEL);
169 if (ops == NULL)
170 return ERR_PTR(-ENOMEM);
171
172 INIT_LIST_HEAD(&ops->rules_list);
173 ops->fro_net = net;
174
175 err = __fib_rules_register(ops);
176 if (err) {
177 kfree(ops);
178 ops = ERR_PTR(err);
179 }
180
181 return ops;
182 }
183 EXPORT_SYMBOL_GPL(fib_rules_register);
184
185 static void fib_rules_cleanup_ops(struct fib_rules_ops *ops)
186 {
187 struct fib_rule *rule, *tmp;
188
189 list_for_each_entry_safe(rule, tmp, &ops->rules_list, list) {
190 list_del_rcu(&rule->list);
191 if (ops->delete)
192 ops->delete(rule);
193 fib_rule_put(rule);
194 }
195 }
196
197 void fib_rules_unregister(struct fib_rules_ops *ops)
198 {
199 struct net *net = ops->fro_net;
200
201 spin_lock(&net->rules_mod_lock);
202 list_del_rcu(&ops->list);
203 spin_unlock(&net->rules_mod_lock);
204
205 fib_rules_cleanup_ops(ops);
206 kfree_rcu(ops, rcu);
207 }
208 EXPORT_SYMBOL_GPL(fib_rules_unregister);
209
210 static int uid_range_set(struct fib_kuid_range *range)
211 {
212 return uid_valid(range->start) && uid_valid(range->end);
213 }
214
215 static struct fib_kuid_range nla_get_kuid_range(struct nlattr **tb)
216 {
217 struct fib_rule_uid_range *in;
218 struct fib_kuid_range out;
219
220 in = (struct fib_rule_uid_range *)nla_data(tb[FRA_UID_RANGE]);
221
222 out.start = make_kuid(current_user_ns(), in->start);
223 out.end = make_kuid(current_user_ns(), in->end);
224
225 return out;
226 }
227
228 static int nla_put_uid_range(struct sk_buff *skb, struct fib_kuid_range *range)
229 {
230 struct fib_rule_uid_range out = {
231 from_kuid_munged(current_user_ns(), range->start),
232 from_kuid_munged(current_user_ns(), range->end)
233 };
234
235 return nla_put(skb, FRA_UID_RANGE, sizeof(out), &out);
236 }
237
238 static int nla_get_port_range(struct nlattr *pattr,
239 struct fib_rule_port_range *port_range)
240 {
241 const struct fib_rule_port_range *pr = nla_data(pattr);
242
243 if (!fib_rule_port_range_valid(pr))
244 return -EINVAL;
245
246 port_range->start = pr->start;
247 port_range->end = pr->end;
248
249 return 0;
250 }
251
252 static int nla_put_port_range(struct sk_buff *skb, int attrtype,
253 struct fib_rule_port_range *range)
254 {
255 return nla_put(skb, attrtype, sizeof(*range), range);
256 }
257
258 static int fib_rule_match(struct fib_rule *rule, struct fib_rules_ops *ops,
259 struct flowi *fl, int flags,
260 struct fib_lookup_arg *arg)
261 {
262 int ret = 0;
263
264 if (rule->iifindex && (rule->iifindex != fl->flowi_iif))
265 goto out;
266
267 if (rule->oifindex && (rule->oifindex != fl->flowi_oif))
268 goto out;
269
270 if ((rule->mark ^ fl->flowi_mark) & rule->mark_mask)
271 goto out;
272
273 if (rule->tun_id && (rule->tun_id != fl->flowi_tun_key.tun_id))
274 goto out;
275
276 if (rule->l3mdev && !l3mdev_fib_rule_match(rule->fr_net, fl, arg))
277 goto out;
278
279 if (uid_lt(fl->flowi_uid, rule->uid_range.start) ||
280 uid_gt(fl->flowi_uid, rule->uid_range.end))
281 goto out;
282
283 ret = INDIRECT_CALL_MT(ops->match,
284 fib6_rule_match,
285 fib4_rule_match,
286 rule, fl, flags);
287 out:
288 return (rule->flags & FIB_RULE_INVERT) ? !ret : ret;
289 }
290
291 int fib_rules_lookup(struct fib_rules_ops *ops, struct flowi *fl,
292 int flags, struct fib_lookup_arg *arg)
293 {
294 struct fib_rule *rule;
295 int err;
296
297 rcu_read_lock();
298
299 list_for_each_entry_rcu(rule, &ops->rules_list, list) {
300 jumped:
301 if (!fib_rule_match(rule, ops, fl, flags, arg))
302 continue;
303
304 if (rule->action == FR_ACT_GOTO) {
305 struct fib_rule *target;
306
307 target = rcu_dereference(rule->ctarget);
308 if (target == NULL) {
309 continue;
310 } else {
311 rule = target;
312 goto jumped;
313 }
314 } else if (rule->action == FR_ACT_NOP)
315 continue;
316 else
317 err = INDIRECT_CALL_MT(ops->action,
318 fib6_rule_action,
319 fib4_rule_action,
320 rule, fl, flags, arg);
321
322 if (!err && ops->suppress && INDIRECT_CALL_MT(ops->suppress,
323 fib6_rule_suppress,
324 fib4_rule_suppress,
325 rule, flags, arg))
326 continue;
327
328 if (err != -EAGAIN) {
329 if ((arg->flags & FIB_LOOKUP_NOREF) ||
330 likely(refcount_inc_not_zero(&rule->refcnt))) {
331 arg->rule = rule;
332 goto out;
333 }
334 break;
335 }
336 }
337
338 err = -ESRCH;
339 out:
340 rcu_read_unlock();
341
342 return err;
343 }
344 EXPORT_SYMBOL_GPL(fib_rules_lookup);
345
346 static int call_fib_rule_notifier(struct notifier_block *nb,
347 enum fib_event_type event_type,
348 struct fib_rule *rule, int family,
349 struct netlink_ext_ack *extack)
350 {
351 struct fib_rule_notifier_info info = {
352 .info.family = family,
353 .info.extack = extack,
354 .rule = rule,
355 };
356
357 return call_fib_notifier(nb, event_type, &info.info);
358 }
359
360 static int call_fib_rule_notifiers(struct net *net,
361 enum fib_event_type event_type,
362 struct fib_rule *rule,
363 struct fib_rules_ops *ops,
364 struct netlink_ext_ack *extack)
365 {
366 struct fib_rule_notifier_info info = {
367 .info.family = ops->family,
368 .info.extack = extack,
369 .rule = rule,
370 };
371
372 ops->fib_rules_seq++;
373 return call_fib_notifiers(net, event_type, &info.info);
374 }
375
376 /* Called with rcu_read_lock() */
377 int fib_rules_dump(struct net *net, struct notifier_block *nb, int family,
378 struct netlink_ext_ack *extack)
379 {
380 struct fib_rules_ops *ops;
381 struct fib_rule *rule;
382 int err = 0;
383
384 ops = lookup_rules_ops(net, family);
385 if (!ops)
386 return -EAFNOSUPPORT;
387 list_for_each_entry_rcu(rule, &ops->rules_list, list) {
388 err = call_fib_rule_notifier(nb, FIB_EVENT_RULE_ADD,
389 rule, family, extack);
390 if (err)
391 break;
392 }
393 rules_ops_put(ops);
394
395 return err;
396 }
397 EXPORT_SYMBOL_GPL(fib_rules_dump);
398
399 unsigned int fib_rules_seq_read(struct net *net, int family)
400 {
401 unsigned int fib_rules_seq;
402 struct fib_rules_ops *ops;
403
404 ASSERT_RTNL();
405
406 ops = lookup_rules_ops(net, family);
407 if (!ops)
408 return 0;
409 fib_rules_seq = ops->fib_rules_seq;
410 rules_ops_put(ops);
411
412 return fib_rules_seq;
413 }
414 EXPORT_SYMBOL_GPL(fib_rules_seq_read);
415
416 static struct fib_rule *rule_find(struct fib_rules_ops *ops,
417 struct fib_rule_hdr *frh,
418 struct nlattr **tb,
419 struct fib_rule *rule,
420 bool user_priority)
421 {
422 struct fib_rule *r;
423
424 list_for_each_entry(r, &ops->rules_list, list) {
425 if (rule->action && r->action != rule->action)
426 continue;
427
428 if (rule->table && r->table != rule->table)
429 continue;
430
431 if (user_priority && r->pref != rule->pref)
432 continue;
433
434 if (rule->iifname[0] &&
435 memcmp(r->iifname, rule->iifname, IFNAMSIZ))
436 continue;
437
438 if (rule->oifname[0] &&
439 memcmp(r->oifname, rule->oifname, IFNAMSIZ))
440 continue;
441
442 if (rule->mark && r->mark != rule->mark)
443 continue;
444
445 if (rule->suppress_ifgroup != -1 &&
446 r->suppress_ifgroup != rule->suppress_ifgroup)
447 continue;
448
449 if (rule->suppress_prefixlen != -1 &&
450 r->suppress_prefixlen != rule->suppress_prefixlen)
451 continue;
452
453 if (rule->mark_mask && r->mark_mask != rule->mark_mask)
454 continue;
455
456 if (rule->tun_id && r->tun_id != rule->tun_id)
457 continue;
458
459 if (r->fr_net != rule->fr_net)
460 continue;
461
462 if (rule->l3mdev && r->l3mdev != rule->l3mdev)
463 continue;
464
465 if (uid_range_set(&rule->uid_range) &&
466 (!uid_eq(r->uid_range.start, rule->uid_range.start) ||
467 !uid_eq(r->uid_range.end, rule->uid_range.end)))
468 continue;
469
470 if (rule->ip_proto && r->ip_proto != rule->ip_proto)
471 continue;
472
473 if (rule->proto && r->proto != rule->proto)
474 continue;
475
476 if (fib_rule_port_range_set(&rule->sport_range) &&
477 !fib_rule_port_range_compare(&r->sport_range,
478 &rule->sport_range))
479 continue;
480
481 if (fib_rule_port_range_set(&rule->dport_range) &&
482 !fib_rule_port_range_compare(&r->dport_range,
483 &rule->dport_range))
484 continue;
485
486 if (!ops->compare(r, frh, tb))
487 continue;
488 return r;
489 }
490
491 return NULL;
492 }
493
494 #ifdef CONFIG_NET_L3_MASTER_DEV
495 static int fib_nl2rule_l3mdev(struct nlattr *nla, struct fib_rule *nlrule,
496 struct netlink_ext_ack *extack)
497 {
498 nlrule->l3mdev = nla_get_u8(nla);
499 if (nlrule->l3mdev != 1) {
500 NL_SET_ERR_MSG(extack, "Invalid l3mdev attribute");
501 return -1;
502 }
503
504 return 0;
505 }
506 #else
507 static int fib_nl2rule_l3mdev(struct nlattr *nla, struct fib_rule *nlrule,
508 struct netlink_ext_ack *extack)
509 {
510 NL_SET_ERR_MSG(extack, "l3mdev support is not enabled in kernel");
511 return -1;
512 }
513 #endif
514
515 static int fib_nl2rule(struct sk_buff *skb, struct nlmsghdr *nlh,
516 struct netlink_ext_ack *extack,
517 struct fib_rules_ops *ops,
518 struct nlattr *tb[],
519 struct fib_rule **rule,
520 bool *user_priority)
521 {
522 struct net *net = sock_net(skb->sk);
523 struct fib_rule_hdr *frh = nlmsg_data(nlh);
524 struct fib_rule *nlrule = NULL;
525 int err = -EINVAL;
526
527 if (frh->src_len)
528 if (!tb[FRA_SRC] ||
529 frh->src_len > (ops->addr_size * 8) ||
530 nla_len(tb[FRA_SRC]) != ops->addr_size) {
531 NL_SET_ERR_MSG(extack, "Invalid source address");
532 goto errout;
533 }
534
535 if (frh->dst_len)
536 if (!tb[FRA_DST] ||
537 frh->dst_len > (ops->addr_size * 8) ||
538 nla_len(tb[FRA_DST]) != ops->addr_size) {
539 NL_SET_ERR_MSG(extack, "Invalid dst address");
540 goto errout;
541 }
542
543 nlrule = kzalloc(ops->rule_size, GFP_KERNEL_ACCOUNT);
544 if (!nlrule) {
545 err = -ENOMEM;
546 goto errout;
547 }
548 refcount_set(&nlrule->refcnt, 1);
549 nlrule->fr_net = net;
550
551 if (tb[FRA_PRIORITY]) {
552 nlrule->pref = nla_get_u32(tb[FRA_PRIORITY]);
553 *user_priority = true;
554 } else {
555 nlrule->pref = fib_default_rule_pref(ops);
556 }
557
558 nlrule->proto = tb[FRA_PROTOCOL] ?
559 nla_get_u8(tb[FRA_PROTOCOL]) : RTPROT_UNSPEC;
560
561 if (tb[FRA_IIFNAME]) {
562 struct net_device *dev;
563
564 nlrule->iifindex = -1;
565 nla_strscpy(nlrule->iifname, tb[FRA_IIFNAME], IFNAMSIZ);
566 dev = __dev_get_by_name(net, nlrule->iifname);
567 if (dev)
568 nlrule->iifindex = dev->ifindex;
569 }
570
571 if (tb[FRA_OIFNAME]) {
572 struct net_device *dev;
573
574 nlrule->oifindex = -1;
575 nla_strscpy(nlrule->oifname, tb[FRA_OIFNAME], IFNAMSIZ);
576 dev = __dev_get_by_name(net, nlrule->oifname);
577 if (dev)
578 nlrule->oifindex = dev->ifindex;
579 }
580
581 if (tb[FRA_FWMARK]) {
582 nlrule->mark = nla_get_u32(tb[FRA_FWMARK]);
583 if (nlrule->mark)
584 /* compatibility: if the mark value is non-zero all bits
585 * are compared unless a mask is explicitly specified.
586 */
587 nlrule->mark_mask = 0xFFFFFFFF;
588 }
589
590 if (tb[FRA_FWMASK])
591 nlrule->mark_mask = nla_get_u32(tb[FRA_FWMASK]);
592
593 if (tb[FRA_TUN_ID])
594 nlrule->tun_id = nla_get_be64(tb[FRA_TUN_ID]);
595
596 if (tb[FRA_L3MDEV] &&
597 fib_nl2rule_l3mdev(tb[FRA_L3MDEV], nlrule, extack) < 0)
598 goto errout_free;
599
600 nlrule->action = frh->action;
601 nlrule->flags = frh->flags;
602 nlrule->table = frh_get_table(frh, tb);
603 if (tb[FRA_SUPPRESS_PREFIXLEN])
604 nlrule->suppress_prefixlen = nla_get_u32(tb[FRA_SUPPRESS_PREFIXLEN]);
605 else
606 nlrule->suppress_prefixlen = -1;
607
608 if (tb[FRA_SUPPRESS_IFGROUP])
609 nlrule->suppress_ifgroup = nla_get_u32(tb[FRA_SUPPRESS_IFGROUP]);
610 else
611 nlrule->suppress_ifgroup = -1;
612
613 if (tb[FRA_GOTO]) {
614 if (nlrule->action != FR_ACT_GOTO) {
615 NL_SET_ERR_MSG(extack, "Unexpected goto");
616 goto errout_free;
617 }
618
619 nlrule->target = nla_get_u32(tb[FRA_GOTO]);
620 /* Backward jumps are prohibited to avoid endless loops */
621 if (nlrule->target <= nlrule->pref) {
622 NL_SET_ERR_MSG(extack, "Backward goto not supported");
623 goto errout_free;
624 }
625 } else if (nlrule->action == FR_ACT_GOTO) {
626 NL_SET_ERR_MSG(extack, "Missing goto target for action goto");
627 goto errout_free;
628 }
629
630 if (nlrule->l3mdev && nlrule->table) {
631 NL_SET_ERR_MSG(extack, "l3mdev and table are mutually exclusive");
632 goto errout_free;
633 }
634
635 if (tb[FRA_UID_RANGE]) {
636 if (current_user_ns() != net->user_ns) {
637 err = -EPERM;
638 NL_SET_ERR_MSG(extack, "No permission to set uid");
639 goto errout_free;
640 }
641
642 nlrule->uid_range = nla_get_kuid_range(tb);
643
644 if (!uid_range_set(&nlrule->uid_range) ||
645 !uid_lte(nlrule->uid_range.start, nlrule->uid_range.end)) {
646 NL_SET_ERR_MSG(extack, "Invalid uid range");
647 goto errout_free;
648 }
649 } else {
650 nlrule->uid_range = fib_kuid_range_unset;
651 }
652
653 if (tb[FRA_IP_PROTO])
654 nlrule->ip_proto = nla_get_u8(tb[FRA_IP_PROTO]);
655
656 if (tb[FRA_SPORT_RANGE]) {
657 err = nla_get_port_range(tb[FRA_SPORT_RANGE],
658 &nlrule->sport_range);
659 if (err) {
660 NL_SET_ERR_MSG(extack, "Invalid sport range");
661 goto errout_free;
662 }
663 }
664
665 if (tb[FRA_DPORT_RANGE]) {
666 err = nla_get_port_range(tb[FRA_DPORT_RANGE],
667 &nlrule->dport_range);
668 if (err) {
669 NL_SET_ERR_MSG(extack, "Invalid dport range");
670 goto errout_free;
671 }
672 }
673
674 *rule = nlrule;
675
676 return 0;
677
678 errout_free:
679 kfree(nlrule);
680 errout:
681 return err;
682 }
683
684 static int rule_exists(struct fib_rules_ops *ops, struct fib_rule_hdr *frh,
685 struct nlattr **tb, struct fib_rule *rule)
686 {
687 struct fib_rule *r;
688
689 list_for_each_entry(r, &ops->rules_list, list) {
690 if (r->action != rule->action)
691 continue;
692
693 if (r->table != rule->table)
694 continue;
695
696 if (r->pref != rule->pref)
697 continue;
698
699 if (memcmp(r->iifname, rule->iifname, IFNAMSIZ))
700 continue;
701
702 if (memcmp(r->oifname, rule->oifname, IFNAMSIZ))
703 continue;
704
705 if (r->mark != rule->mark)
706 continue;
707
708 if (r->suppress_ifgroup != rule->suppress_ifgroup)
709 continue;
710
711 if (r->suppress_prefixlen != rule->suppress_prefixlen)
712 continue;
713
714 if (r->mark_mask != rule->mark_mask)
715 continue;
716
717 if (r->tun_id != rule->tun_id)
718 continue;
719
720 if (r->fr_net != rule->fr_net)
721 continue;
722
723 if (r->l3mdev != rule->l3mdev)
724 continue;
725
726 if (!uid_eq(r->uid_range.start, rule->uid_range.start) ||
727 !uid_eq(r->uid_range.end, rule->uid_range.end))
728 continue;
729
730 if (r->ip_proto != rule->ip_proto)
731 continue;
732
733 if (r->proto != rule->proto)
734 continue;
735
736 if (!fib_rule_port_range_compare(&r->sport_range,
737 &rule->sport_range))
738 continue;
739
740 if (!fib_rule_port_range_compare(&r->dport_range,
741 &rule->dport_range))
742 continue;
743
744 if (!ops->compare(r, frh, tb))
745 continue;
746 return 1;
747 }
748 return 0;
749 }
750
751 static const struct nla_policy fib_rule_policy[FRA_MAX + 1] = {
752 [FRA_UNSPEC] = { .strict_start_type = FRA_DPORT_RANGE + 1 },
753 [FRA_IIFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
754 [FRA_OIFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
755 [FRA_PRIORITY] = { .type = NLA_U32 },
756 [FRA_FWMARK] = { .type = NLA_U32 },
757 [FRA_FLOW] = { .type = NLA_U32 },
758 [FRA_TUN_ID] = { .type = NLA_U64 },
759 [FRA_FWMASK] = { .type = NLA_U32 },
760 [FRA_TABLE] = { .type = NLA_U32 },
761 [FRA_SUPPRESS_PREFIXLEN] = { .type = NLA_U32 },
762 [FRA_SUPPRESS_IFGROUP] = { .type = NLA_U32 },
763 [FRA_GOTO] = { .type = NLA_U32 },
764 [FRA_L3MDEV] = { .type = NLA_U8 },
765 [FRA_UID_RANGE] = { .len = sizeof(struct fib_rule_uid_range) },
766 [FRA_PROTOCOL] = { .type = NLA_U8 },
767 [FRA_IP_PROTO] = { .type = NLA_U8 },
768 [FRA_SPORT_RANGE] = { .len = sizeof(struct fib_rule_port_range) },
769 [FRA_DPORT_RANGE] = { .len = sizeof(struct fib_rule_port_range) }
770 };
771
772 int fib_nl_newrule(struct sk_buff *skb, struct nlmsghdr *nlh,
773 struct netlink_ext_ack *extack)
774 {
775 struct net *net = sock_net(skb->sk);
776 struct fib_rule_hdr *frh = nlmsg_data(nlh);
777 struct fib_rules_ops *ops = NULL;
778 struct fib_rule *rule = NULL, *r, *last = NULL;
779 struct nlattr *tb[FRA_MAX + 1];
780 int err = -EINVAL, unresolved = 0;
781 bool user_priority = false;
782
783 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh))) {
784 NL_SET_ERR_MSG(extack, "Invalid msg length");
785 goto errout;
786 }
787
788 ops = lookup_rules_ops(net, frh->family);
789 if (!ops) {
790 err = -EAFNOSUPPORT;
791 NL_SET_ERR_MSG(extack, "Rule family not supported");
792 goto errout;
793 }
794
795 err = nlmsg_parse_deprecated(nlh, sizeof(*frh), tb, FRA_MAX,
796 fib_rule_policy, extack);
797 if (err < 0) {
798 NL_SET_ERR_MSG(extack, "Error parsing msg");
799 goto errout;
800 }
801
802 err = fib_nl2rule(skb, nlh, extack, ops, tb, &rule, &user_priority);
803 if (err)
804 goto errout;
805
806 if ((nlh->nlmsg_flags & NLM_F_EXCL) &&
807 rule_exists(ops, frh, tb, rule)) {
808 err = -EEXIST;
809 goto errout_free;
810 }
811
812 err = ops->configure(rule, skb, frh, tb, extack);
813 if (err < 0)
814 goto errout_free;
815
816 err = call_fib_rule_notifiers(net, FIB_EVENT_RULE_ADD, rule, ops,
817 extack);
818 if (err < 0)
819 goto errout_free;
820
821 list_for_each_entry(r, &ops->rules_list, list) {
822 if (r->pref == rule->target) {
823 RCU_INIT_POINTER(rule->ctarget, r);
824 break;
825 }
826 }
827
828 if (rcu_dereference_protected(rule->ctarget, 1) == NULL)
829 unresolved = 1;
830
831 list_for_each_entry(r, &ops->rules_list, list) {
832 if (r->pref > rule->pref)
833 break;
834 last = r;
835 }
836
837 if (last)
838 list_add_rcu(&rule->list, &last->list);
839 else
840 list_add_rcu(&rule->list, &ops->rules_list);
841
842 if (ops->unresolved_rules) {
843 /*
844 * There are unresolved goto rules in the list, check if
845 * any of them are pointing to this new rule.
846 */
847 list_for_each_entry(r, &ops->rules_list, list) {
848 if (r->action == FR_ACT_GOTO &&
849 r->target == rule->pref &&
850 rtnl_dereference(r->ctarget) == NULL) {
851 rcu_assign_pointer(r->ctarget, rule);
852 if (--ops->unresolved_rules == 0)
853 break;
854 }
855 }
856 }
857
858 if (rule->action == FR_ACT_GOTO)
859 ops->nr_goto_rules++;
860
861 if (unresolved)
862 ops->unresolved_rules++;
863
864 if (rule->tun_id)
865 ip_tunnel_need_metadata();
866
867 notify_rule_change(RTM_NEWRULE, rule, ops, nlh, NETLINK_CB(skb).portid);
868 flush_route_cache(ops);
869 rules_ops_put(ops);
870 return 0;
871
872 errout_free:
873 kfree(rule);
874 errout:
875 rules_ops_put(ops);
876 return err;
877 }
878 EXPORT_SYMBOL_GPL(fib_nl_newrule);
879
880 int fib_nl_delrule(struct sk_buff *skb, struct nlmsghdr *nlh,
881 struct netlink_ext_ack *extack)
882 {
883 struct net *net = sock_net(skb->sk);
884 struct fib_rule_hdr *frh = nlmsg_data(nlh);
885 struct fib_rules_ops *ops = NULL;
886 struct fib_rule *rule = NULL, *r, *nlrule = NULL;
887 struct nlattr *tb[FRA_MAX+1];
888 int err = -EINVAL;
889 bool user_priority = false;
890
891 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh))) {
892 NL_SET_ERR_MSG(extack, "Invalid msg length");
893 goto errout;
894 }
895
896 ops = lookup_rules_ops(net, frh->family);
897 if (ops == NULL) {
898 err = -EAFNOSUPPORT;
899 NL_SET_ERR_MSG(extack, "Rule family not supported");
900 goto errout;
901 }
902
903 err = nlmsg_parse_deprecated(nlh, sizeof(*frh), tb, FRA_MAX,
904 fib_rule_policy, extack);
905 if (err < 0) {
906 NL_SET_ERR_MSG(extack, "Error parsing msg");
907 goto errout;
908 }
909
910 err = fib_nl2rule(skb, nlh, extack, ops, tb, &nlrule, &user_priority);
911 if (err)
912 goto errout;
913
914 rule = rule_find(ops, frh, tb, nlrule, user_priority);
915 if (!rule) {
916 err = -ENOENT;
917 goto errout;
918 }
919
920 if (rule->flags & FIB_RULE_PERMANENT) {
921 err = -EPERM;
922 goto errout;
923 }
924
925 if (ops->delete) {
926 err = ops->delete(rule);
927 if (err)
928 goto errout;
929 }
930
931 if (rule->tun_id)
932 ip_tunnel_unneed_metadata();
933
934 list_del_rcu(&rule->list);
935
936 if (rule->action == FR_ACT_GOTO) {
937 ops->nr_goto_rules--;
938 if (rtnl_dereference(rule->ctarget) == NULL)
939 ops->unresolved_rules--;
940 }
941
942 /*
943 * Check if this rule is a target to any of them. If so,
944 * adjust to the next one with the same preference or
945 * disable them. As this operation is eventually very
946 * expensive, it is only performed if goto rules, except
947 * current if it is goto rule, have actually been added.
948 */
949 if (ops->nr_goto_rules > 0) {
950 struct fib_rule *n;
951
952 n = list_next_entry(rule, list);
953 if (&n->list == &ops->rules_list || n->pref != rule->pref)
954 n = NULL;
955 list_for_each_entry(r, &ops->rules_list, list) {
956 if (rtnl_dereference(r->ctarget) != rule)
957 continue;
958 rcu_assign_pointer(r->ctarget, n);
959 if (!n)
960 ops->unresolved_rules++;
961 }
962 }
963
964 call_fib_rule_notifiers(net, FIB_EVENT_RULE_DEL, rule, ops,
965 NULL);
966 notify_rule_change(RTM_DELRULE, rule, ops, nlh,
967 NETLINK_CB(skb).portid);
968 fib_rule_put(rule);
969 flush_route_cache(ops);
970 rules_ops_put(ops);
971 kfree(nlrule);
972 return 0;
973
974 errout:
975 kfree(nlrule);
976 rules_ops_put(ops);
977 return err;
978 }
979 EXPORT_SYMBOL_GPL(fib_nl_delrule);
980
981 static inline size_t fib_rule_nlmsg_size(struct fib_rules_ops *ops,
982 struct fib_rule *rule)
983 {
984 size_t payload = NLMSG_ALIGN(sizeof(struct fib_rule_hdr))
985 + nla_total_size(IFNAMSIZ) /* FRA_IIFNAME */
986 + nla_total_size(IFNAMSIZ) /* FRA_OIFNAME */
987 + nla_total_size(4) /* FRA_PRIORITY */
988 + nla_total_size(4) /* FRA_TABLE */
989 + nla_total_size(4) /* FRA_SUPPRESS_PREFIXLEN */
990 + nla_total_size(4) /* FRA_SUPPRESS_IFGROUP */
991 + nla_total_size(4) /* FRA_FWMARK */
992 + nla_total_size(4) /* FRA_FWMASK */
993 + nla_total_size_64bit(8) /* FRA_TUN_ID */
994 + nla_total_size(sizeof(struct fib_kuid_range))
995 + nla_total_size(1) /* FRA_PROTOCOL */
996 + nla_total_size(1) /* FRA_IP_PROTO */
997 + nla_total_size(sizeof(struct fib_rule_port_range)) /* FRA_SPORT_RANGE */
998 + nla_total_size(sizeof(struct fib_rule_port_range)); /* FRA_DPORT_RANGE */
999
1000 if (ops->nlmsg_payload)
1001 payload += ops->nlmsg_payload(rule);
1002
1003 return payload;
1004 }
1005
1006 static int fib_nl_fill_rule(struct sk_buff *skb, struct fib_rule *rule,
1007 u32 pid, u32 seq, int type, int flags,
1008 struct fib_rules_ops *ops)
1009 {
1010 struct nlmsghdr *nlh;
1011 struct fib_rule_hdr *frh;
1012
1013 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*frh), flags);
1014 if (nlh == NULL)
1015 return -EMSGSIZE;
1016
1017 frh = nlmsg_data(nlh);
1018 frh->family = ops->family;
1019 frh->table = rule->table < 256 ? rule->table : RT_TABLE_COMPAT;
1020 if (nla_put_u32(skb, FRA_TABLE, rule->table))
1021 goto nla_put_failure;
1022 if (nla_put_u32(skb, FRA_SUPPRESS_PREFIXLEN, rule->suppress_prefixlen))
1023 goto nla_put_failure;
1024 frh->res1 = 0;
1025 frh->res2 = 0;
1026 frh->action = rule->action;
1027 frh->flags = rule->flags;
1028
1029 if (nla_put_u8(skb, FRA_PROTOCOL, rule->proto))
1030 goto nla_put_failure;
1031
1032 if (rule->action == FR_ACT_GOTO &&
1033 rcu_access_pointer(rule->ctarget) == NULL)
1034 frh->flags |= FIB_RULE_UNRESOLVED;
1035
1036 if (rule->iifname[0]) {
1037 if (nla_put_string(skb, FRA_IIFNAME, rule->iifname))
1038 goto nla_put_failure;
1039 if (rule->iifindex == -1)
1040 frh->flags |= FIB_RULE_IIF_DETACHED;
1041 }
1042
1043 if (rule->oifname[0]) {
1044 if (nla_put_string(skb, FRA_OIFNAME, rule->oifname))
1045 goto nla_put_failure;
1046 if (rule->oifindex == -1)
1047 frh->flags |= FIB_RULE_OIF_DETACHED;
1048 }
1049
1050 if ((rule->pref &&
1051 nla_put_u32(skb, FRA_PRIORITY, rule->pref)) ||
1052 (rule->mark &&
1053 nla_put_u32(skb, FRA_FWMARK, rule->mark)) ||
1054 ((rule->mark_mask || rule->mark) &&
1055 nla_put_u32(skb, FRA_FWMASK, rule->mark_mask)) ||
1056 (rule->target &&
1057 nla_put_u32(skb, FRA_GOTO, rule->target)) ||
1058 (rule->tun_id &&
1059 nla_put_be64(skb, FRA_TUN_ID, rule->tun_id, FRA_PAD)) ||
1060 (rule->l3mdev &&
1061 nla_put_u8(skb, FRA_L3MDEV, rule->l3mdev)) ||
1062 (uid_range_set(&rule->uid_range) &&
1063 nla_put_uid_range(skb, &rule->uid_range)) ||
1064 (fib_rule_port_range_set(&rule->sport_range) &&
1065 nla_put_port_range(skb, FRA_SPORT_RANGE, &rule->sport_range)) ||
1066 (fib_rule_port_range_set(&rule->dport_range) &&
1067 nla_put_port_range(skb, FRA_DPORT_RANGE, &rule->dport_range)) ||
1068 (rule->ip_proto && nla_put_u8(skb, FRA_IP_PROTO, rule->ip_proto)))
1069 goto nla_put_failure;
1070
1071 if (rule->suppress_ifgroup != -1) {
1072 if (nla_put_u32(skb, FRA_SUPPRESS_IFGROUP, rule->suppress_ifgroup))
1073 goto nla_put_failure;
1074 }
1075
1076 if (ops->fill(rule, skb, frh) < 0)
1077 goto nla_put_failure;
1078
1079 nlmsg_end(skb, nlh);
1080 return 0;
1081
1082 nla_put_failure:
1083 nlmsg_cancel(skb, nlh);
1084 return -EMSGSIZE;
1085 }
1086
1087 static int dump_rules(struct sk_buff *skb, struct netlink_callback *cb,
1088 struct fib_rules_ops *ops)
1089 {
1090 int idx = 0;
1091 struct fib_rule *rule;
1092 int err = 0;
1093
1094 rcu_read_lock();
1095 list_for_each_entry_rcu(rule, &ops->rules_list, list) {
1096 if (idx < cb->args[1])
1097 goto skip;
1098
1099 err = fib_nl_fill_rule(skb, rule, NETLINK_CB(cb->skb).portid,
1100 cb->nlh->nlmsg_seq, RTM_NEWRULE,
1101 NLM_F_MULTI, ops);
1102 if (err)
1103 break;
1104 skip:
1105 idx++;
1106 }
1107 rcu_read_unlock();
1108 cb->args[1] = idx;
1109 rules_ops_put(ops);
1110
1111 return err;
1112 }
1113
1114 static int fib_valid_dumprule_req(const struct nlmsghdr *nlh,
1115 struct netlink_ext_ack *extack)
1116 {
1117 struct fib_rule_hdr *frh;
1118
1119 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh))) {
1120 NL_SET_ERR_MSG(extack, "Invalid header for fib rule dump request");
1121 return -EINVAL;
1122 }
1123
1124 frh = nlmsg_data(nlh);
1125 if (frh->dst_len || frh->src_len || frh->tos || frh->table ||
1126 frh->res1 || frh->res2 || frh->action || frh->flags) {
1127 NL_SET_ERR_MSG(extack,
1128 "Invalid values in header for fib rule dump request");
1129 return -EINVAL;
1130 }
1131
1132 if (nlmsg_attrlen(nlh, sizeof(*frh))) {
1133 NL_SET_ERR_MSG(extack, "Invalid data after header in fib rule dump request");
1134 return -EINVAL;
1135 }
1136
1137 return 0;
1138 }
1139
1140 static int fib_nl_dumprule(struct sk_buff *skb, struct netlink_callback *cb)
1141 {
1142 const struct nlmsghdr *nlh = cb->nlh;
1143 struct net *net = sock_net(skb->sk);
1144 struct fib_rules_ops *ops;
1145 int err, idx = 0, family;
1146
1147 if (cb->strict_check) {
1148 err = fib_valid_dumprule_req(nlh, cb->extack);
1149
1150 if (err < 0)
1151 return err;
1152 }
1153
1154 family = rtnl_msg_family(nlh);
1155 if (family != AF_UNSPEC) {
1156 /* Protocol specific dump request */
1157 ops = lookup_rules_ops(net, family);
1158 if (ops == NULL)
1159 return -EAFNOSUPPORT;
1160
1161 return dump_rules(skb, cb, ops);
1162 }
1163
1164 err = 0;
1165 rcu_read_lock();
1166 list_for_each_entry_rcu(ops, &net->rules_ops, list) {
1167 if (idx < cb->args[0] || !try_module_get(ops->owner))
1168 goto skip;
1169
1170 err = dump_rules(skb, cb, ops);
1171 if (err < 0)
1172 break;
1173
1174 cb->args[1] = 0;
1175 skip:
1176 idx++;
1177 }
1178 rcu_read_unlock();
1179 cb->args[0] = idx;
1180
1181 return err;
1182 }
1183
1184 static void notify_rule_change(int event, struct fib_rule *rule,
1185 struct fib_rules_ops *ops, struct nlmsghdr *nlh,
1186 u32 pid)
1187 {
1188 struct net *net;
1189 struct sk_buff *skb;
1190 int err = -ENOMEM;
1191
1192 net = ops->fro_net;
1193 skb = nlmsg_new(fib_rule_nlmsg_size(ops, rule), GFP_KERNEL);
1194 if (skb == NULL)
1195 goto errout;
1196
1197 err = fib_nl_fill_rule(skb, rule, pid, nlh->nlmsg_seq, event, 0, ops);
1198 if (err < 0) {
1199 /* -EMSGSIZE implies BUG in fib_rule_nlmsg_size() */
1200 WARN_ON(err == -EMSGSIZE);
1201 kfree_skb(skb);
1202 goto errout;
1203 }
1204
1205 rtnl_notify(skb, net, pid, ops->nlgroup, nlh, GFP_KERNEL);
1206 return;
1207 errout:
1208 if (err < 0)
1209 rtnl_set_sk_err(net, ops->nlgroup, err);
1210 }
1211
1212 static void attach_rules(struct list_head *rules, struct net_device *dev)
1213 {
1214 struct fib_rule *rule;
1215
1216 list_for_each_entry(rule, rules, list) {
1217 if (rule->iifindex == -1 &&
1218 strcmp(dev->name, rule->iifname) == 0)
1219 rule->iifindex = dev->ifindex;
1220 if (rule->oifindex == -1 &&
1221 strcmp(dev->name, rule->oifname) == 0)
1222 rule->oifindex = dev->ifindex;
1223 }
1224 }
1225
1226 static void detach_rules(struct list_head *rules, struct net_device *dev)
1227 {
1228 struct fib_rule *rule;
1229
1230 list_for_each_entry(rule, rules, list) {
1231 if (rule->iifindex == dev->ifindex)
1232 rule->iifindex = -1;
1233 if (rule->oifindex == dev->ifindex)
1234 rule->oifindex = -1;
1235 }
1236 }
1237
1238
1239 static int fib_rules_event(struct notifier_block *this, unsigned long event,
1240 void *ptr)
1241 {
1242 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1243 struct net *net = dev_net(dev);
1244 struct fib_rules_ops *ops;
1245
1246 ASSERT_RTNL();
1247
1248 switch (event) {
1249 case NETDEV_REGISTER:
1250 list_for_each_entry(ops, &net->rules_ops, list)
1251 attach_rules(&ops->rules_list, dev);
1252 break;
1253
1254 case NETDEV_CHANGENAME:
1255 list_for_each_entry(ops, &net->rules_ops, list) {
1256 detach_rules(&ops->rules_list, dev);
1257 attach_rules(&ops->rules_list, dev);
1258 }
1259 break;
1260
1261 case NETDEV_UNREGISTER:
1262 list_for_each_entry(ops, &net->rules_ops, list)
1263 detach_rules(&ops->rules_list, dev);
1264 break;
1265 }
1266
1267 return NOTIFY_DONE;
1268 }
1269
1270 static struct notifier_block fib_rules_notifier = {
1271 .notifier_call = fib_rules_event,
1272 };
1273
1274 static int __net_init fib_rules_net_init(struct net *net)
1275 {
1276 INIT_LIST_HEAD(&net->rules_ops);
1277 spin_lock_init(&net->rules_mod_lock);
1278 return 0;
1279 }
1280
1281 static void __net_exit fib_rules_net_exit(struct net *net)
1282 {
1283 WARN_ON_ONCE(!list_empty(&net->rules_ops));
1284 }
1285
1286 static struct pernet_operations fib_rules_net_ops = {
1287 .init = fib_rules_net_init,
1288 .exit = fib_rules_net_exit,
1289 };
1290
1291 static int __init fib_rules_init(void)
1292 {
1293 int err;
1294 rtnl_register(PF_UNSPEC, RTM_NEWRULE, fib_nl_newrule, NULL, 0);
1295 rtnl_register(PF_UNSPEC, RTM_DELRULE, fib_nl_delrule, NULL, 0);
1296 rtnl_register(PF_UNSPEC, RTM_GETRULE, NULL, fib_nl_dumprule,
1297 RTNL_FLAG_DUMP_UNLOCKED);
1298
1299 err = register_pernet_subsys(&fib_rules_net_ops);
1300 if (err < 0)
1301 goto fail;
1302
1303 err = register_netdevice_notifier(&fib_rules_notifier);
1304 if (err < 0)
1305 goto fail_unregister;
1306
1307 return 0;
1308
1309 fail_unregister:
1310 unregister_pernet_subsys(&fib_rules_net_ops);
1311 fail:
1312 rtnl_unregister(PF_UNSPEC, RTM_NEWRULE);
1313 rtnl_unregister(PF_UNSPEC, RTM_DELRULE);
1314 rtnl_unregister(PF_UNSPEC, RTM_GETRULE);
1315 return err;
1316 }
1317
1318 subsys_initcall(fib_rules_init);
1319
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.