Loading...
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/inet_dscp.h>
15#include <net/sock.h>
16#include <net/fib_rules.h>
17#include <net/ip_tunnels.h>
18#include <linux/indirect_call_wrapper.h>
19
20#if defined(CONFIG_IPV6) && defined(CONFIG_IPV6_MULTIPLE_TABLES)
21#ifdef CONFIG_IP_MULTIPLE_TABLES
22#define INDIRECT_CALL_MT(f, f2, f1, ...) \
23 INDIRECT_CALL_INET(f, f2, f1, __VA_ARGS__)
24#else
25#define INDIRECT_CALL_MT(f, f2, f1, ...) INDIRECT_CALL_1(f, f2, __VA_ARGS__)
26#endif
27#elif defined(CONFIG_IP_MULTIPLE_TABLES)
28#define INDIRECT_CALL_MT(f, f2, f1, ...) INDIRECT_CALL_1(f, f1, __VA_ARGS__)
29#else
30#define INDIRECT_CALL_MT(f, f2, f1, ...) f(__VA_ARGS__)
31#endif
32
33static const struct fib_kuid_range fib_kuid_range_unset = {
34 KUIDT_INIT(0),
35 KUIDT_INIT(~0),
36};
37
38bool fib_rule_matchall(const struct fib_rule *rule)
39{
40 if (READ_ONCE(rule->iifindex) || READ_ONCE(rule->oifindex) ||
41 rule->mark || rule->tun_id || rule->flags)
42 return false;
43 if (rule->suppress_ifgroup != -1 || rule->suppress_prefixlen != -1)
44 return false;
45 if (!uid_eq(rule->uid_range.start, fib_kuid_range_unset.start) ||
46 !uid_eq(rule->uid_range.end, fib_kuid_range_unset.end))
47 return false;
48 if (fib_rule_port_range_set(&rule->sport_range))
49 return false;
50 if (fib_rule_port_range_set(&rule->dport_range))
51 return false;
52 return true;
53}
54EXPORT_SYMBOL_GPL(fib_rule_matchall);
55
56int fib_default_rule_add(struct fib_rules_ops *ops,
57 u32 pref, u32 table)
58{
59 struct fib_rule *r;
60
61 r = kzalloc(ops->rule_size, GFP_KERNEL_ACCOUNT);
62 if (r == NULL)
63 return -ENOMEM;
64
65 refcount_set(&r->refcnt, 1);
66 r->action = FR_ACT_TO_TBL;
67 r->pref = pref;
68 r->table = table;
69 r->proto = RTPROT_KERNEL;
70 r->fr_net = ops->fro_net;
71 r->uid_range = fib_kuid_range_unset;
72
73 r->suppress_prefixlen = -1;
74 r->suppress_ifgroup = -1;
75
76 /* The lock is not required here, the list in unreachable
77 * at the moment this function is called */
78 list_add_tail(&r->list, &ops->rules_list);
79 return 0;
80}
81EXPORT_SYMBOL(fib_default_rule_add);
82
83static u32 fib_default_rule_pref(struct fib_rules_ops *ops)
84{
85 struct list_head *pos;
86 struct fib_rule *rule;
87
88 if (!list_empty(&ops->rules_list)) {
89 pos = ops->rules_list.next;
90 if (pos->next != &ops->rules_list) {
91 rule = list_entry(pos->next, struct fib_rule, list);
92 if (rule->pref)
93 return rule->pref - 1;
94 }
95 }
96
97 return 0;
98}
99
100static void notify_rule_change(int event, struct fib_rule *rule,
101 struct fib_rules_ops *ops, struct nlmsghdr *nlh,
102 u32 pid);
103
104static struct fib_rules_ops *lookup_rules_ops(const struct net *net,
105 int family)
106{
107 struct fib_rules_ops *ops;
108
109 rcu_read_lock();
110 list_for_each_entry_rcu(ops, &net->rules_ops, list) {
111 if (ops->family == family) {
112 if (!try_module_get(ops->owner))
113 ops = NULL;
114 rcu_read_unlock();
115 return ops;
116 }
117 }
118 rcu_read_unlock();
119
120 return NULL;
121}
122
123static void rules_ops_put(struct fib_rules_ops *ops)
124{
125 if (ops)
126 module_put(ops->owner);
127}
128
129static void flush_route_cache(struct fib_rules_ops *ops)
130{
131 if (ops->flush_cache)
132 ops->flush_cache(ops);
133}
134
135static int __fib_rules_register(struct fib_rules_ops *ops)
136{
137 int err = -EEXIST;
138 struct fib_rules_ops *o;
139 struct net *net;
140
141 net = ops->fro_net;
142
143 if (ops->rule_size < sizeof(struct fib_rule))
144 return -EINVAL;
145
146 if (ops->match == NULL || ops->configure == NULL ||
147 ops->compare == NULL || ops->fill == NULL ||
148 ops->action == NULL)
149 return -EINVAL;
150
151 spin_lock(&net->rules_mod_lock);
152 list_for_each_entry(o, &net->rules_ops, list)
153 if (ops->family == o->family)
154 goto errout;
155
156 list_add_tail_rcu(&ops->list, &net->rules_ops);
157 err = 0;
158errout:
159 spin_unlock(&net->rules_mod_lock);
160
161 return err;
162}
163
164struct fib_rules_ops *
165fib_rules_register(const struct fib_rules_ops *tmpl, struct net *net)
166{
167 struct fib_rules_ops *ops;
168 int err;
169
170 ops = kmemdup(tmpl, sizeof(*ops), GFP_KERNEL);
171 if (ops == NULL)
172 return ERR_PTR(-ENOMEM);
173
174 INIT_LIST_HEAD(&ops->rules_list);
175 ops->fro_net = net;
176
177 err = __fib_rules_register(ops);
178 if (err) {
179 kfree(ops);
180 ops = ERR_PTR(err);
181 }
182
183 return ops;
184}
185EXPORT_SYMBOL_GPL(fib_rules_register);
186
187static void fib_rules_cleanup_ops(struct fib_rules_ops *ops)
188{
189 struct fib_rule *rule, *tmp;
190
191 list_for_each_entry_safe(rule, tmp, &ops->rules_list, list) {
192 list_del_rcu(&rule->list);
193 if (ops->delete)
194 ops->delete(rule);
195 fib_rule_put(rule);
196 }
197}
198
199void fib_rules_unregister(struct fib_rules_ops *ops)
200{
201 struct net *net = ops->fro_net;
202
203 spin_lock(&net->rules_mod_lock);
204 list_del_rcu(&ops->list);
205 spin_unlock(&net->rules_mod_lock);
206
207 fib_rules_cleanup_ops(ops);
208 kfree_rcu(ops, rcu);
209}
210EXPORT_SYMBOL_GPL(fib_rules_unregister);
211
212static int uid_range_set(struct fib_kuid_range *range)
213{
214 return uid_valid(range->start) && uid_valid(range->end);
215}
216
217static struct fib_kuid_range nla_get_kuid_range(struct nlattr **tb)
218{
219 struct fib_rule_uid_range *in;
220 struct fib_kuid_range out;
221
222 in = (struct fib_rule_uid_range *)nla_data(tb[FRA_UID_RANGE]);
223
224 out.start = make_kuid(current_user_ns(), in->start);
225 out.end = make_kuid(current_user_ns(), in->end);
226
227 return out;
228}
229
230static int nla_put_uid_range(struct sk_buff *skb, struct fib_kuid_range *range)
231{
232 struct fib_rule_uid_range out = {
233 from_kuid_munged(current_user_ns(), range->start),
234 from_kuid_munged(current_user_ns(), range->end)
235 };
236
237 return nla_put(skb, FRA_UID_RANGE, sizeof(out), &out);
238}
239
240static int nla_get_port_range(struct nlattr *pattr,
241 struct fib_rule_port_range *port_range)
242{
243 const struct fib_rule_port_range *pr = nla_data(pattr);
244
245 if (!fib_rule_port_range_valid(pr))
246 return -EINVAL;
247
248 port_range->start = pr->start;
249 port_range->end = pr->end;
250
251 return 0;
252}
253
254static int nla_put_port_range(struct sk_buff *skb, int attrtype,
255 struct fib_rule_port_range *range)
256{
257 return nla_put(skb, attrtype, sizeof(*range), range);
258}
259
260static int fib_rule_match(struct fib_rule *rule, struct fib_rules_ops *ops,
261 struct flowi *fl, int flags,
262 struct fib_lookup_arg *arg)
263{
264 int iifindex, oifindex, ret = 0;
265
266 iifindex = READ_ONCE(rule->iifindex);
267 if (iifindex && (iifindex != fl->flowi_iif))
268 goto out;
269
270 oifindex = READ_ONCE(rule->oifindex);
271 if (oifindex && (oifindex != fl->flowi_oif))
272 goto out;
273
274 if ((rule->mark ^ fl->flowi_mark) & rule->mark_mask)
275 goto out;
276
277 if (rule->tun_id && (rule->tun_id != fl->flowi_tun_key.tun_id))
278 goto out;
279
280 if (rule->l3mdev && !l3mdev_fib_rule_match(rule->fr_net, fl, arg))
281 goto out;
282
283 if (uid_lt(fl->flowi_uid, rule->uid_range.start) ||
284 uid_gt(fl->flowi_uid, rule->uid_range.end))
285 goto out;
286
287 ret = INDIRECT_CALL_MT(ops->match,
288 fib6_rule_match,
289 fib4_rule_match,
290 rule, fl, flags);
291out:
292 return (rule->flags & FIB_RULE_INVERT) ? !ret : ret;
293}
294
295int fib_rules_lookup(struct fib_rules_ops *ops, struct flowi *fl,
296 int flags, struct fib_lookup_arg *arg)
297{
298 struct fib_rule *rule;
299 int err;
300
301 rcu_read_lock();
302
303 list_for_each_entry_rcu(rule, &ops->rules_list, list) {
304jumped:
305 if (!fib_rule_match(rule, ops, fl, flags, arg))
306 continue;
307
308 if (rule->action == FR_ACT_GOTO) {
309 struct fib_rule *target;
310
311 target = rcu_dereference(rule->ctarget);
312 if (target == NULL) {
313 continue;
314 } else {
315 rule = target;
316 goto jumped;
317 }
318 } else if (rule->action == FR_ACT_NOP)
319 continue;
320 else
321 err = INDIRECT_CALL_MT(ops->action,
322 fib6_rule_action,
323 fib4_rule_action,
324 rule, fl, flags, arg);
325
326 if (!err && ops->suppress && INDIRECT_CALL_MT(ops->suppress,
327 fib6_rule_suppress,
328 fib4_rule_suppress,
329 rule, flags, arg))
330 continue;
331
332 if (err != -EAGAIN) {
333 if ((arg->flags & FIB_LOOKUP_NOREF) ||
334 likely(refcount_inc_not_zero(&rule->refcnt))) {
335 arg->rule = rule;
336 goto out;
337 }
338 break;
339 }
340 }
341
342 err = -ESRCH;
343out:
344 rcu_read_unlock();
345
346 return err;
347}
348EXPORT_SYMBOL_GPL(fib_rules_lookup);
349
350static int call_fib_rule_notifier(struct notifier_block *nb,
351 enum fib_event_type event_type,
352 struct fib_rule *rule, int family,
353 struct netlink_ext_ack *extack)
354{
355 struct fib_rule_notifier_info info = {
356 .info.family = family,
357 .info.extack = extack,
358 .rule = rule,
359 };
360
361 return call_fib_notifier(nb, event_type, &info.info);
362}
363
364static int call_fib_rule_notifiers(struct net *net,
365 enum fib_event_type event_type,
366 struct fib_rule *rule,
367 struct fib_rules_ops *ops,
368 struct netlink_ext_ack *extack)
369{
370 struct fib_rule_notifier_info info = {
371 .info.family = ops->family,
372 .info.extack = extack,
373 .rule = rule,
374 };
375
376 ASSERT_RTNL();
377 /* Paired with READ_ONCE() in fib_rules_seq() */
378 WRITE_ONCE(ops->fib_rules_seq, ops->fib_rules_seq + 1);
379 return call_fib_notifiers(net, event_type, &info.info);
380}
381
382/* Called with rcu_read_lock() */
383int fib_rules_dump(struct net *net, struct notifier_block *nb, int family,
384 struct netlink_ext_ack *extack)
385{
386 struct fib_rules_ops *ops;
387 struct fib_rule *rule;
388 int err = 0;
389
390 ops = lookup_rules_ops(net, family);
391 if (!ops)
392 return -EAFNOSUPPORT;
393 list_for_each_entry_rcu(rule, &ops->rules_list, list) {
394 err = call_fib_rule_notifier(nb, FIB_EVENT_RULE_ADD,
395 rule, family, extack);
396 if (err)
397 break;
398 }
399 rules_ops_put(ops);
400
401 return err;
402}
403EXPORT_SYMBOL_GPL(fib_rules_dump);
404
405unsigned int fib_rules_seq_read(const struct net *net, int family)
406{
407 unsigned int fib_rules_seq;
408 struct fib_rules_ops *ops;
409
410 ops = lookup_rules_ops(net, family);
411 if (!ops)
412 return 0;
413 /* Paired with WRITE_ONCE() in call_fib_rule_notifiers() */
414 fib_rules_seq = READ_ONCE(ops->fib_rules_seq);
415 rules_ops_put(ops);
416
417 return fib_rules_seq;
418}
419EXPORT_SYMBOL_GPL(fib_rules_seq_read);
420
421static struct fib_rule *rule_find(struct fib_rules_ops *ops,
422 struct fib_rule_hdr *frh,
423 struct nlattr **tb,
424 struct fib_rule *rule,
425 bool user_priority)
426{
427 struct fib_rule *r;
428
429 list_for_each_entry(r, &ops->rules_list, list) {
430 if (rule->action && r->action != rule->action)
431 continue;
432
433 if (rule->table && r->table != rule->table)
434 continue;
435
436 if (user_priority && r->pref != rule->pref)
437 continue;
438
439 if (rule->iifname[0] &&
440 memcmp(r->iifname, rule->iifname, IFNAMSIZ))
441 continue;
442
443 if (rule->oifname[0] &&
444 memcmp(r->oifname, rule->oifname, IFNAMSIZ))
445 continue;
446
447 if (rule->mark && r->mark != rule->mark)
448 continue;
449
450 if (rule->suppress_ifgroup != -1 &&
451 r->suppress_ifgroup != rule->suppress_ifgroup)
452 continue;
453
454 if (rule->suppress_prefixlen != -1 &&
455 r->suppress_prefixlen != rule->suppress_prefixlen)
456 continue;
457
458 if (rule->mark_mask && r->mark_mask != rule->mark_mask)
459 continue;
460
461 if (rule->tun_id && r->tun_id != rule->tun_id)
462 continue;
463
464 if (r->fr_net != rule->fr_net)
465 continue;
466
467 if (rule->l3mdev && r->l3mdev != rule->l3mdev)
468 continue;
469
470 if (uid_range_set(&rule->uid_range) &&
471 (!uid_eq(r->uid_range.start, rule->uid_range.start) ||
472 !uid_eq(r->uid_range.end, rule->uid_range.end)))
473 continue;
474
475 if (rule->ip_proto && r->ip_proto != rule->ip_proto)
476 continue;
477
478 if (rule->proto && r->proto != rule->proto)
479 continue;
480
481 if (fib_rule_port_range_set(&rule->sport_range) &&
482 !fib_rule_port_range_compare(&r->sport_range,
483 &rule->sport_range))
484 continue;
485
486 if (fib_rule_port_range_set(&rule->dport_range) &&
487 !fib_rule_port_range_compare(&r->dport_range,
488 &rule->dport_range))
489 continue;
490
491 if (!ops->compare(r, frh, tb))
492 continue;
493 return r;
494 }
495
496 return NULL;
497}
498
499#ifdef CONFIG_NET_L3_MASTER_DEV
500static int fib_nl2rule_l3mdev(struct nlattr *nla, struct fib_rule *nlrule,
501 struct netlink_ext_ack *extack)
502{
503 nlrule->l3mdev = nla_get_u8(nla);
504 if (nlrule->l3mdev != 1) {
505 NL_SET_ERR_MSG(extack, "Invalid l3mdev attribute");
506 return -1;
507 }
508
509 return 0;
510}
511#else
512static int fib_nl2rule_l3mdev(struct nlattr *nla, struct fib_rule *nlrule,
513 struct netlink_ext_ack *extack)
514{
515 NL_SET_ERR_MSG(extack, "l3mdev support is not enabled in kernel");
516 return -1;
517}
518#endif
519
520static int fib_nl2rule(struct sk_buff *skb, struct nlmsghdr *nlh,
521 struct netlink_ext_ack *extack,
522 struct fib_rules_ops *ops,
523 struct nlattr *tb[],
524 struct fib_rule **rule,
525 bool *user_priority)
526{
527 struct net *net = sock_net(skb->sk);
528 struct fib_rule_hdr *frh = nlmsg_data(nlh);
529 struct fib_rule *nlrule = NULL;
530 int err = -EINVAL;
531
532 if (frh->src_len)
533 if (!tb[FRA_SRC] ||
534 frh->src_len > (ops->addr_size * 8) ||
535 nla_len(tb[FRA_SRC]) != ops->addr_size) {
536 NL_SET_ERR_MSG(extack, "Invalid source address");
537 goto errout;
538 }
539
540 if (frh->dst_len)
541 if (!tb[FRA_DST] ||
542 frh->dst_len > (ops->addr_size * 8) ||
543 nla_len(tb[FRA_DST]) != ops->addr_size) {
544 NL_SET_ERR_MSG(extack, "Invalid dst address");
545 goto errout;
546 }
547
548 nlrule = kzalloc(ops->rule_size, GFP_KERNEL_ACCOUNT);
549 if (!nlrule) {
550 err = -ENOMEM;
551 goto errout;
552 }
553 refcount_set(&nlrule->refcnt, 1);
554 nlrule->fr_net = net;
555
556 if (tb[FRA_PRIORITY]) {
557 nlrule->pref = nla_get_u32(tb[FRA_PRIORITY]);
558 *user_priority = true;
559 } else {
560 nlrule->pref = fib_default_rule_pref(ops);
561 }
562
563 nlrule->proto = nla_get_u8_default(tb[FRA_PROTOCOL], RTPROT_UNSPEC);
564
565 if (tb[FRA_IIFNAME]) {
566 struct net_device *dev;
567
568 nlrule->iifindex = -1;
569 nla_strscpy(nlrule->iifname, tb[FRA_IIFNAME], IFNAMSIZ);
570 dev = __dev_get_by_name(net, nlrule->iifname);
571 if (dev)
572 nlrule->iifindex = dev->ifindex;
573 }
574
575 if (tb[FRA_OIFNAME]) {
576 struct net_device *dev;
577
578 nlrule->oifindex = -1;
579 nla_strscpy(nlrule->oifname, tb[FRA_OIFNAME], IFNAMSIZ);
580 dev = __dev_get_by_name(net, nlrule->oifname);
581 if (dev)
582 nlrule->oifindex = dev->ifindex;
583 }
584
585 if (tb[FRA_FWMARK]) {
586 nlrule->mark = nla_get_u32(tb[FRA_FWMARK]);
587 if (nlrule->mark)
588 /* compatibility: if the mark value is non-zero all bits
589 * are compared unless a mask is explicitly specified.
590 */
591 nlrule->mark_mask = 0xFFFFFFFF;
592 }
593
594 if (tb[FRA_FWMASK])
595 nlrule->mark_mask = nla_get_u32(tb[FRA_FWMASK]);
596
597 if (tb[FRA_TUN_ID])
598 nlrule->tun_id = nla_get_be64(tb[FRA_TUN_ID]);
599
600 if (tb[FRA_L3MDEV] &&
601 fib_nl2rule_l3mdev(tb[FRA_L3MDEV], nlrule, extack) < 0)
602 goto errout_free;
603
604 nlrule->action = frh->action;
605 nlrule->flags = frh->flags;
606 nlrule->table = frh_get_table(frh, tb);
607 if (tb[FRA_SUPPRESS_PREFIXLEN])
608 nlrule->suppress_prefixlen = nla_get_u32(tb[FRA_SUPPRESS_PREFIXLEN]);
609 else
610 nlrule->suppress_prefixlen = -1;
611
612 if (tb[FRA_SUPPRESS_IFGROUP])
613 nlrule->suppress_ifgroup = nla_get_u32(tb[FRA_SUPPRESS_IFGROUP]);
614 else
615 nlrule->suppress_ifgroup = -1;
616
617 if (tb[FRA_GOTO]) {
618 if (nlrule->action != FR_ACT_GOTO) {
619 NL_SET_ERR_MSG(extack, "Unexpected goto");
620 goto errout_free;
621 }
622
623 nlrule->target = nla_get_u32(tb[FRA_GOTO]);
624 /* Backward jumps are prohibited to avoid endless loops */
625 if (nlrule->target <= nlrule->pref) {
626 NL_SET_ERR_MSG(extack, "Backward goto not supported");
627 goto errout_free;
628 }
629 } else if (nlrule->action == FR_ACT_GOTO) {
630 NL_SET_ERR_MSG(extack, "Missing goto target for action goto");
631 goto errout_free;
632 }
633
634 if (nlrule->l3mdev && nlrule->table) {
635 NL_SET_ERR_MSG(extack, "l3mdev and table are mutually exclusive");
636 goto errout_free;
637 }
638
639 if (tb[FRA_UID_RANGE]) {
640 if (current_user_ns() != net->user_ns) {
641 err = -EPERM;
642 NL_SET_ERR_MSG(extack, "No permission to set uid");
643 goto errout_free;
644 }
645
646 nlrule->uid_range = nla_get_kuid_range(tb);
647
648 if (!uid_range_set(&nlrule->uid_range) ||
649 !uid_lte(nlrule->uid_range.start, nlrule->uid_range.end)) {
650 NL_SET_ERR_MSG(extack, "Invalid uid range");
651 goto errout_free;
652 }
653 } else {
654 nlrule->uid_range = fib_kuid_range_unset;
655 }
656
657 if (tb[FRA_IP_PROTO])
658 nlrule->ip_proto = nla_get_u8(tb[FRA_IP_PROTO]);
659
660 if (tb[FRA_SPORT_RANGE]) {
661 err = nla_get_port_range(tb[FRA_SPORT_RANGE],
662 &nlrule->sport_range);
663 if (err) {
664 NL_SET_ERR_MSG(extack, "Invalid sport range");
665 goto errout_free;
666 }
667 }
668
669 if (tb[FRA_DPORT_RANGE]) {
670 err = nla_get_port_range(tb[FRA_DPORT_RANGE],
671 &nlrule->dport_range);
672 if (err) {
673 NL_SET_ERR_MSG(extack, "Invalid dport range");
674 goto errout_free;
675 }
676 }
677
678 *rule = nlrule;
679
680 return 0;
681
682errout_free:
683 kfree(nlrule);
684errout:
685 return err;
686}
687
688static int rule_exists(struct fib_rules_ops *ops, struct fib_rule_hdr *frh,
689 struct nlattr **tb, struct fib_rule *rule)
690{
691 struct fib_rule *r;
692
693 list_for_each_entry(r, &ops->rules_list, list) {
694 if (r->action != rule->action)
695 continue;
696
697 if (r->table != rule->table)
698 continue;
699
700 if (r->pref != rule->pref)
701 continue;
702
703 if (memcmp(r->iifname, rule->iifname, IFNAMSIZ))
704 continue;
705
706 if (memcmp(r->oifname, rule->oifname, IFNAMSIZ))
707 continue;
708
709 if (r->mark != rule->mark)
710 continue;
711
712 if (r->suppress_ifgroup != rule->suppress_ifgroup)
713 continue;
714
715 if (r->suppress_prefixlen != rule->suppress_prefixlen)
716 continue;
717
718 if (r->mark_mask != rule->mark_mask)
719 continue;
720
721 if (r->tun_id != rule->tun_id)
722 continue;
723
724 if (r->fr_net != rule->fr_net)
725 continue;
726
727 if (r->l3mdev != rule->l3mdev)
728 continue;
729
730 if (!uid_eq(r->uid_range.start, rule->uid_range.start) ||
731 !uid_eq(r->uid_range.end, rule->uid_range.end))
732 continue;
733
734 if (r->ip_proto != rule->ip_proto)
735 continue;
736
737 if (r->proto != rule->proto)
738 continue;
739
740 if (!fib_rule_port_range_compare(&r->sport_range,
741 &rule->sport_range))
742 continue;
743
744 if (!fib_rule_port_range_compare(&r->dport_range,
745 &rule->dport_range))
746 continue;
747
748 if (!ops->compare(r, frh, tb))
749 continue;
750 return 1;
751 }
752 return 0;
753}
754
755static const struct nla_policy fib_rule_policy[FRA_MAX + 1] = {
756 [FRA_UNSPEC] = { .strict_start_type = FRA_DPORT_RANGE + 1 },
757 [FRA_IIFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
758 [FRA_OIFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
759 [FRA_PRIORITY] = { .type = NLA_U32 },
760 [FRA_FWMARK] = { .type = NLA_U32 },
761 [FRA_FLOW] = { .type = NLA_U32 },
762 [FRA_TUN_ID] = { .type = NLA_U64 },
763 [FRA_FWMASK] = { .type = NLA_U32 },
764 [FRA_TABLE] = { .type = NLA_U32 },
765 [FRA_SUPPRESS_PREFIXLEN] = { .type = NLA_U32 },
766 [FRA_SUPPRESS_IFGROUP] = { .type = NLA_U32 },
767 [FRA_GOTO] = { .type = NLA_U32 },
768 [FRA_L3MDEV] = { .type = NLA_U8 },
769 [FRA_UID_RANGE] = { .len = sizeof(struct fib_rule_uid_range) },
770 [FRA_PROTOCOL] = { .type = NLA_U8 },
771 [FRA_IP_PROTO] = { .type = NLA_U8 },
772 [FRA_SPORT_RANGE] = { .len = sizeof(struct fib_rule_port_range) },
773 [FRA_DPORT_RANGE] = { .len = sizeof(struct fib_rule_port_range) },
774 [FRA_DSCP] = NLA_POLICY_MAX(NLA_U8, INET_DSCP_MASK >> 2),
775};
776
777int fib_nl_newrule(struct sk_buff *skb, struct nlmsghdr *nlh,
778 struct netlink_ext_ack *extack)
779{
780 struct net *net = sock_net(skb->sk);
781 struct fib_rule_hdr *frh = nlmsg_data(nlh);
782 struct fib_rules_ops *ops = NULL;
783 struct fib_rule *rule = NULL, *r, *last = NULL;
784 struct nlattr *tb[FRA_MAX + 1];
785 int err = -EINVAL, unresolved = 0;
786 bool user_priority = false;
787
788 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh))) {
789 NL_SET_ERR_MSG(extack, "Invalid msg length");
790 goto errout;
791 }
792
793 ops = lookup_rules_ops(net, frh->family);
794 if (!ops) {
795 err = -EAFNOSUPPORT;
796 NL_SET_ERR_MSG(extack, "Rule family not supported");
797 goto errout;
798 }
799
800 err = nlmsg_parse_deprecated(nlh, sizeof(*frh), tb, FRA_MAX,
801 fib_rule_policy, extack);
802 if (err < 0) {
803 NL_SET_ERR_MSG(extack, "Error parsing msg");
804 goto errout;
805 }
806
807 err = fib_nl2rule(skb, nlh, extack, ops, tb, &rule, &user_priority);
808 if (err)
809 goto errout;
810
811 if ((nlh->nlmsg_flags & NLM_F_EXCL) &&
812 rule_exists(ops, frh, tb, rule)) {
813 err = -EEXIST;
814 goto errout_free;
815 }
816
817 err = ops->configure(rule, skb, frh, tb, extack);
818 if (err < 0)
819 goto errout_free;
820
821 err = call_fib_rule_notifiers(net, FIB_EVENT_RULE_ADD, rule, ops,
822 extack);
823 if (err < 0)
824 goto errout_free;
825
826 list_for_each_entry(r, &ops->rules_list, list) {
827 if (r->pref == rule->target) {
828 RCU_INIT_POINTER(rule->ctarget, r);
829 break;
830 }
831 }
832
833 if (rcu_dereference_protected(rule->ctarget, 1) == NULL)
834 unresolved = 1;
835
836 list_for_each_entry(r, &ops->rules_list, list) {
837 if (r->pref > rule->pref)
838 break;
839 last = r;
840 }
841
842 if (last)
843 list_add_rcu(&rule->list, &last->list);
844 else
845 list_add_rcu(&rule->list, &ops->rules_list);
846
847 if (ops->unresolved_rules) {
848 /*
849 * There are unresolved goto rules in the list, check if
850 * any of them are pointing to this new rule.
851 */
852 list_for_each_entry(r, &ops->rules_list, list) {
853 if (r->action == FR_ACT_GOTO &&
854 r->target == rule->pref &&
855 rtnl_dereference(r->ctarget) == NULL) {
856 rcu_assign_pointer(r->ctarget, rule);
857 if (--ops->unresolved_rules == 0)
858 break;
859 }
860 }
861 }
862
863 if (rule->action == FR_ACT_GOTO)
864 ops->nr_goto_rules++;
865
866 if (unresolved)
867 ops->unresolved_rules++;
868
869 if (rule->tun_id)
870 ip_tunnel_need_metadata();
871
872 notify_rule_change(RTM_NEWRULE, rule, ops, nlh, NETLINK_CB(skb).portid);
873 flush_route_cache(ops);
874 rules_ops_put(ops);
875 return 0;
876
877errout_free:
878 kfree(rule);
879errout:
880 rules_ops_put(ops);
881 return err;
882}
883EXPORT_SYMBOL_GPL(fib_nl_newrule);
884
885int fib_nl_delrule(struct sk_buff *skb, struct nlmsghdr *nlh,
886 struct netlink_ext_ack *extack)
887{
888 struct net *net = sock_net(skb->sk);
889 struct fib_rule_hdr *frh = nlmsg_data(nlh);
890 struct fib_rules_ops *ops = NULL;
891 struct fib_rule *rule = NULL, *r, *nlrule = NULL;
892 struct nlattr *tb[FRA_MAX+1];
893 int err = -EINVAL;
894 bool user_priority = false;
895
896 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh))) {
897 NL_SET_ERR_MSG(extack, "Invalid msg length");
898 goto errout;
899 }
900
901 ops = lookup_rules_ops(net, frh->family);
902 if (ops == NULL) {
903 err = -EAFNOSUPPORT;
904 NL_SET_ERR_MSG(extack, "Rule family not supported");
905 goto errout;
906 }
907
908 err = nlmsg_parse_deprecated(nlh, sizeof(*frh), tb, FRA_MAX,
909 fib_rule_policy, extack);
910 if (err < 0) {
911 NL_SET_ERR_MSG(extack, "Error parsing msg");
912 goto errout;
913 }
914
915 err = fib_nl2rule(skb, nlh, extack, ops, tb, &nlrule, &user_priority);
916 if (err)
917 goto errout;
918
919 rule = rule_find(ops, frh, tb, nlrule, user_priority);
920 if (!rule) {
921 err = -ENOENT;
922 goto errout;
923 }
924
925 if (rule->flags & FIB_RULE_PERMANENT) {
926 err = -EPERM;
927 goto errout;
928 }
929
930 if (ops->delete) {
931 err = ops->delete(rule);
932 if (err)
933 goto errout;
934 }
935
936 if (rule->tun_id)
937 ip_tunnel_unneed_metadata();
938
939 list_del_rcu(&rule->list);
940
941 if (rule->action == FR_ACT_GOTO) {
942 ops->nr_goto_rules--;
943 if (rtnl_dereference(rule->ctarget) == NULL)
944 ops->unresolved_rules--;
945 }
946
947 /*
948 * Check if this rule is a target to any of them. If so,
949 * adjust to the next one with the same preference or
950 * disable them. As this operation is eventually very
951 * expensive, it is only performed if goto rules, except
952 * current if it is goto rule, have actually been added.
953 */
954 if (ops->nr_goto_rules > 0) {
955 struct fib_rule *n;
956
957 n = list_next_entry(rule, list);
958 if (&n->list == &ops->rules_list || n->pref != rule->pref)
959 n = NULL;
960 list_for_each_entry(r, &ops->rules_list, list) {
961 if (rtnl_dereference(r->ctarget) != rule)
962 continue;
963 rcu_assign_pointer(r->ctarget, n);
964 if (!n)
965 ops->unresolved_rules++;
966 }
967 }
968
969 call_fib_rule_notifiers(net, FIB_EVENT_RULE_DEL, rule, ops,
970 NULL);
971 notify_rule_change(RTM_DELRULE, rule, ops, nlh,
972 NETLINK_CB(skb).portid);
973 fib_rule_put(rule);
974 flush_route_cache(ops);
975 rules_ops_put(ops);
976 kfree(nlrule);
977 return 0;
978
979errout:
980 kfree(nlrule);
981 rules_ops_put(ops);
982 return err;
983}
984EXPORT_SYMBOL_GPL(fib_nl_delrule);
985
986static inline size_t fib_rule_nlmsg_size(struct fib_rules_ops *ops,
987 struct fib_rule *rule)
988{
989 size_t payload = NLMSG_ALIGN(sizeof(struct fib_rule_hdr))
990 + nla_total_size(IFNAMSIZ) /* FRA_IIFNAME */
991 + nla_total_size(IFNAMSIZ) /* FRA_OIFNAME */
992 + nla_total_size(4) /* FRA_PRIORITY */
993 + nla_total_size(4) /* FRA_TABLE */
994 + nla_total_size(4) /* FRA_SUPPRESS_PREFIXLEN */
995 + nla_total_size(4) /* FRA_SUPPRESS_IFGROUP */
996 + nla_total_size(4) /* FRA_FWMARK */
997 + nla_total_size(4) /* FRA_FWMASK */
998 + nla_total_size_64bit(8) /* FRA_TUN_ID */
999 + nla_total_size(sizeof(struct fib_kuid_range))
1000 + nla_total_size(1) /* FRA_PROTOCOL */
1001 + nla_total_size(1) /* FRA_IP_PROTO */
1002 + nla_total_size(sizeof(struct fib_rule_port_range)) /* FRA_SPORT_RANGE */
1003 + nla_total_size(sizeof(struct fib_rule_port_range)); /* FRA_DPORT_RANGE */
1004
1005 if (ops->nlmsg_payload)
1006 payload += ops->nlmsg_payload(rule);
1007
1008 return payload;
1009}
1010
1011static int fib_nl_fill_rule(struct sk_buff *skb, struct fib_rule *rule,
1012 u32 pid, u32 seq, int type, int flags,
1013 struct fib_rules_ops *ops)
1014{
1015 struct nlmsghdr *nlh;
1016 struct fib_rule_hdr *frh;
1017
1018 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*frh), flags);
1019 if (nlh == NULL)
1020 return -EMSGSIZE;
1021
1022 frh = nlmsg_data(nlh);
1023 frh->family = ops->family;
1024 frh->table = rule->table < 256 ? rule->table : RT_TABLE_COMPAT;
1025 if (nla_put_u32(skb, FRA_TABLE, rule->table))
1026 goto nla_put_failure;
1027 if (nla_put_u32(skb, FRA_SUPPRESS_PREFIXLEN, rule->suppress_prefixlen))
1028 goto nla_put_failure;
1029 frh->res1 = 0;
1030 frh->res2 = 0;
1031 frh->action = rule->action;
1032 frh->flags = rule->flags;
1033
1034 if (nla_put_u8(skb, FRA_PROTOCOL, rule->proto))
1035 goto nla_put_failure;
1036
1037 if (rule->action == FR_ACT_GOTO &&
1038 rcu_access_pointer(rule->ctarget) == NULL)
1039 frh->flags |= FIB_RULE_UNRESOLVED;
1040
1041 if (rule->iifname[0]) {
1042 if (nla_put_string(skb, FRA_IIFNAME, rule->iifname))
1043 goto nla_put_failure;
1044 if (READ_ONCE(rule->iifindex) == -1)
1045 frh->flags |= FIB_RULE_IIF_DETACHED;
1046 }
1047
1048 if (rule->oifname[0]) {
1049 if (nla_put_string(skb, FRA_OIFNAME, rule->oifname))
1050 goto nla_put_failure;
1051 if (READ_ONCE(rule->oifindex) == -1)
1052 frh->flags |= FIB_RULE_OIF_DETACHED;
1053 }
1054
1055 if ((rule->pref &&
1056 nla_put_u32(skb, FRA_PRIORITY, rule->pref)) ||
1057 (rule->mark &&
1058 nla_put_u32(skb, FRA_FWMARK, rule->mark)) ||
1059 ((rule->mark_mask || rule->mark) &&
1060 nla_put_u32(skb, FRA_FWMASK, rule->mark_mask)) ||
1061 (rule->target &&
1062 nla_put_u32(skb, FRA_GOTO, rule->target)) ||
1063 (rule->tun_id &&
1064 nla_put_be64(skb, FRA_TUN_ID, rule->tun_id, FRA_PAD)) ||
1065 (rule->l3mdev &&
1066 nla_put_u8(skb, FRA_L3MDEV, rule->l3mdev)) ||
1067 (uid_range_set(&rule->uid_range) &&
1068 nla_put_uid_range(skb, &rule->uid_range)) ||
1069 (fib_rule_port_range_set(&rule->sport_range) &&
1070 nla_put_port_range(skb, FRA_SPORT_RANGE, &rule->sport_range)) ||
1071 (fib_rule_port_range_set(&rule->dport_range) &&
1072 nla_put_port_range(skb, FRA_DPORT_RANGE, &rule->dport_range)) ||
1073 (rule->ip_proto && nla_put_u8(skb, FRA_IP_PROTO, rule->ip_proto)))
1074 goto nla_put_failure;
1075
1076 if (rule->suppress_ifgroup != -1) {
1077 if (nla_put_u32(skb, FRA_SUPPRESS_IFGROUP, rule->suppress_ifgroup))
1078 goto nla_put_failure;
1079 }
1080
1081 if (ops->fill(rule, skb, frh) < 0)
1082 goto nla_put_failure;
1083
1084 nlmsg_end(skb, nlh);
1085 return 0;
1086
1087nla_put_failure:
1088 nlmsg_cancel(skb, nlh);
1089 return -EMSGSIZE;
1090}
1091
1092static int dump_rules(struct sk_buff *skb, struct netlink_callback *cb,
1093 struct fib_rules_ops *ops)
1094{
1095 int idx = 0;
1096 struct fib_rule *rule;
1097 int err = 0;
1098
1099 rcu_read_lock();
1100 list_for_each_entry_rcu(rule, &ops->rules_list, list) {
1101 if (idx < cb->args[1])
1102 goto skip;
1103
1104 err = fib_nl_fill_rule(skb, rule, NETLINK_CB(cb->skb).portid,
1105 cb->nlh->nlmsg_seq, RTM_NEWRULE,
1106 NLM_F_MULTI, ops);
1107 if (err)
1108 break;
1109skip:
1110 idx++;
1111 }
1112 rcu_read_unlock();
1113 cb->args[1] = idx;
1114 rules_ops_put(ops);
1115
1116 return err;
1117}
1118
1119static int fib_valid_dumprule_req(const struct nlmsghdr *nlh,
1120 struct netlink_ext_ack *extack)
1121{
1122 struct fib_rule_hdr *frh;
1123
1124 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh))) {
1125 NL_SET_ERR_MSG(extack, "Invalid header for fib rule dump request");
1126 return -EINVAL;
1127 }
1128
1129 frh = nlmsg_data(nlh);
1130 if (frh->dst_len || frh->src_len || frh->tos || frh->table ||
1131 frh->res1 || frh->res2 || frh->action || frh->flags) {
1132 NL_SET_ERR_MSG(extack,
1133 "Invalid values in header for fib rule dump request");
1134 return -EINVAL;
1135 }
1136
1137 if (nlmsg_attrlen(nlh, sizeof(*frh))) {
1138 NL_SET_ERR_MSG(extack, "Invalid data after header in fib rule dump request");
1139 return -EINVAL;
1140 }
1141
1142 return 0;
1143}
1144
1145static int fib_nl_dumprule(struct sk_buff *skb, struct netlink_callback *cb)
1146{
1147 const struct nlmsghdr *nlh = cb->nlh;
1148 struct net *net = sock_net(skb->sk);
1149 struct fib_rules_ops *ops;
1150 int err, idx = 0, family;
1151
1152 if (cb->strict_check) {
1153 err = fib_valid_dumprule_req(nlh, cb->extack);
1154
1155 if (err < 0)
1156 return err;
1157 }
1158
1159 family = rtnl_msg_family(nlh);
1160 if (family != AF_UNSPEC) {
1161 /* Protocol specific dump request */
1162 ops = lookup_rules_ops(net, family);
1163 if (ops == NULL)
1164 return -EAFNOSUPPORT;
1165
1166 return dump_rules(skb, cb, ops);
1167 }
1168
1169 err = 0;
1170 rcu_read_lock();
1171 list_for_each_entry_rcu(ops, &net->rules_ops, list) {
1172 if (idx < cb->args[0] || !try_module_get(ops->owner))
1173 goto skip;
1174
1175 err = dump_rules(skb, cb, ops);
1176 if (err < 0)
1177 break;
1178
1179 cb->args[1] = 0;
1180skip:
1181 idx++;
1182 }
1183 rcu_read_unlock();
1184 cb->args[0] = idx;
1185
1186 return err;
1187}
1188
1189static void notify_rule_change(int event, struct fib_rule *rule,
1190 struct fib_rules_ops *ops, struct nlmsghdr *nlh,
1191 u32 pid)
1192{
1193 struct net *net;
1194 struct sk_buff *skb;
1195 int err = -ENOMEM;
1196
1197 net = ops->fro_net;
1198 skb = nlmsg_new(fib_rule_nlmsg_size(ops, rule), GFP_KERNEL);
1199 if (skb == NULL)
1200 goto errout;
1201
1202 err = fib_nl_fill_rule(skb, rule, pid, nlh->nlmsg_seq, event, 0, ops);
1203 if (err < 0) {
1204 /* -EMSGSIZE implies BUG in fib_rule_nlmsg_size() */
1205 WARN_ON(err == -EMSGSIZE);
1206 kfree_skb(skb);
1207 goto errout;
1208 }
1209
1210 rtnl_notify(skb, net, pid, ops->nlgroup, nlh, GFP_KERNEL);
1211 return;
1212errout:
1213 rtnl_set_sk_err(net, ops->nlgroup, err);
1214}
1215
1216static void attach_rules(struct list_head *rules, struct net_device *dev)
1217{
1218 struct fib_rule *rule;
1219
1220 list_for_each_entry(rule, rules, list) {
1221 if (rule->iifindex == -1 &&
1222 strcmp(dev->name, rule->iifname) == 0)
1223 WRITE_ONCE(rule->iifindex, dev->ifindex);
1224 if (rule->oifindex == -1 &&
1225 strcmp(dev->name, rule->oifname) == 0)
1226 WRITE_ONCE(rule->oifindex, dev->ifindex);
1227 }
1228}
1229
1230static void detach_rules(struct list_head *rules, struct net_device *dev)
1231{
1232 struct fib_rule *rule;
1233
1234 list_for_each_entry(rule, rules, list) {
1235 if (rule->iifindex == dev->ifindex)
1236 WRITE_ONCE(rule->iifindex, -1);
1237 if (rule->oifindex == dev->ifindex)
1238 WRITE_ONCE(rule->oifindex, -1);
1239 }
1240}
1241
1242
1243static int fib_rules_event(struct notifier_block *this, unsigned long event,
1244 void *ptr)
1245{
1246 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1247 struct net *net = dev_net(dev);
1248 struct fib_rules_ops *ops;
1249
1250 ASSERT_RTNL();
1251
1252 switch (event) {
1253 case NETDEV_REGISTER:
1254 list_for_each_entry(ops, &net->rules_ops, list)
1255 attach_rules(&ops->rules_list, dev);
1256 break;
1257
1258 case NETDEV_CHANGENAME:
1259 list_for_each_entry(ops, &net->rules_ops, list) {
1260 detach_rules(&ops->rules_list, dev);
1261 attach_rules(&ops->rules_list, dev);
1262 }
1263 break;
1264
1265 case NETDEV_UNREGISTER:
1266 list_for_each_entry(ops, &net->rules_ops, list)
1267 detach_rules(&ops->rules_list, dev);
1268 break;
1269 }
1270
1271 return NOTIFY_DONE;
1272}
1273
1274static struct notifier_block fib_rules_notifier = {
1275 .notifier_call = fib_rules_event,
1276};
1277
1278static int __net_init fib_rules_net_init(struct net *net)
1279{
1280 INIT_LIST_HEAD(&net->rules_ops);
1281 spin_lock_init(&net->rules_mod_lock);
1282 return 0;
1283}
1284
1285static void __net_exit fib_rules_net_exit(struct net *net)
1286{
1287 WARN_ON_ONCE(!list_empty(&net->rules_ops));
1288}
1289
1290static struct pernet_operations fib_rules_net_ops = {
1291 .init = fib_rules_net_init,
1292 .exit = fib_rules_net_exit,
1293};
1294
1295static const struct rtnl_msg_handler fib_rules_rtnl_msg_handlers[] __initconst = {
1296 {.msgtype = RTM_NEWRULE, .doit = fib_nl_newrule},
1297 {.msgtype = RTM_DELRULE, .doit = fib_nl_delrule},
1298 {.msgtype = RTM_GETRULE, .dumpit = fib_nl_dumprule,
1299 .flags = RTNL_FLAG_DUMP_UNLOCKED},
1300};
1301
1302static int __init fib_rules_init(void)
1303{
1304 int err;
1305
1306 rtnl_register_many(fib_rules_rtnl_msg_handlers);
1307
1308 err = register_pernet_subsys(&fib_rules_net_ops);
1309 if (err < 0)
1310 goto fail;
1311
1312 err = register_netdevice_notifier(&fib_rules_notifier);
1313 if (err < 0)
1314 goto fail_unregister;
1315
1316 return 0;
1317
1318fail_unregister:
1319 unregister_pernet_subsys(&fib_rules_net_ops);
1320fail:
1321 rtnl_unregister_many(fib_rules_rtnl_msg_handlers);
1322 return err;
1323}
1324
1325subsys_initcall(fib_rules_init);
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
32static const struct fib_kuid_range fib_kuid_range_unset = {
33 KUIDT_INIT(0),
34 KUIDT_INIT(~0),
35};
36
37bool 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}
53EXPORT_SYMBOL_GPL(fib_rule_matchall);
54
55int fib_default_rule_add(struct fib_rules_ops *ops,
56 u32 pref, u32 table, u32 flags)
57{
58 struct fib_rule *r;
59
60 r = kzalloc(ops->rule_size, GFP_KERNEL);
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->flags = flags;
69 r->proto = RTPROT_KERNEL;
70 r->fr_net = ops->fro_net;
71 r->uid_range = fib_kuid_range_unset;
72
73 r->suppress_prefixlen = -1;
74 r->suppress_ifgroup = -1;
75
76 /* The lock is not required here, the list in unreacheable
77 * at the moment this function is called */
78 list_add_tail(&r->list, &ops->rules_list);
79 return 0;
80}
81EXPORT_SYMBOL(fib_default_rule_add);
82
83static u32 fib_default_rule_pref(struct fib_rules_ops *ops)
84{
85 struct list_head *pos;
86 struct fib_rule *rule;
87
88 if (!list_empty(&ops->rules_list)) {
89 pos = ops->rules_list.next;
90 if (pos->next != &ops->rules_list) {
91 rule = list_entry(pos->next, struct fib_rule, list);
92 if (rule->pref)
93 return rule->pref - 1;
94 }
95 }
96
97 return 0;
98}
99
100static void notify_rule_change(int event, struct fib_rule *rule,
101 struct fib_rules_ops *ops, struct nlmsghdr *nlh,
102 u32 pid);
103
104static struct fib_rules_ops *lookup_rules_ops(struct net *net, int family)
105{
106 struct fib_rules_ops *ops;
107
108 rcu_read_lock();
109 list_for_each_entry_rcu(ops, &net->rules_ops, list) {
110 if (ops->family == family) {
111 if (!try_module_get(ops->owner))
112 ops = NULL;
113 rcu_read_unlock();
114 return ops;
115 }
116 }
117 rcu_read_unlock();
118
119 return NULL;
120}
121
122static void rules_ops_put(struct fib_rules_ops *ops)
123{
124 if (ops)
125 module_put(ops->owner);
126}
127
128static void flush_route_cache(struct fib_rules_ops *ops)
129{
130 if (ops->flush_cache)
131 ops->flush_cache(ops);
132}
133
134static int __fib_rules_register(struct fib_rules_ops *ops)
135{
136 int err = -EEXIST;
137 struct fib_rules_ops *o;
138 struct net *net;
139
140 net = ops->fro_net;
141
142 if (ops->rule_size < sizeof(struct fib_rule))
143 return -EINVAL;
144
145 if (ops->match == NULL || ops->configure == NULL ||
146 ops->compare == NULL || ops->fill == NULL ||
147 ops->action == NULL)
148 return -EINVAL;
149
150 spin_lock(&net->rules_mod_lock);
151 list_for_each_entry(o, &net->rules_ops, list)
152 if (ops->family == o->family)
153 goto errout;
154
155 list_add_tail_rcu(&ops->list, &net->rules_ops);
156 err = 0;
157errout:
158 spin_unlock(&net->rules_mod_lock);
159
160 return err;
161}
162
163struct fib_rules_ops *
164fib_rules_register(const struct fib_rules_ops *tmpl, struct net *net)
165{
166 struct fib_rules_ops *ops;
167 int err;
168
169 ops = kmemdup(tmpl, sizeof(*ops), GFP_KERNEL);
170 if (ops == NULL)
171 return ERR_PTR(-ENOMEM);
172
173 INIT_LIST_HEAD(&ops->rules_list);
174 ops->fro_net = net;
175
176 err = __fib_rules_register(ops);
177 if (err) {
178 kfree(ops);
179 ops = ERR_PTR(err);
180 }
181
182 return ops;
183}
184EXPORT_SYMBOL_GPL(fib_rules_register);
185
186static void fib_rules_cleanup_ops(struct fib_rules_ops *ops)
187{
188 struct fib_rule *rule, *tmp;
189
190 list_for_each_entry_safe(rule, tmp, &ops->rules_list, list) {
191 list_del_rcu(&rule->list);
192 if (ops->delete)
193 ops->delete(rule);
194 fib_rule_put(rule);
195 }
196}
197
198void fib_rules_unregister(struct fib_rules_ops *ops)
199{
200 struct net *net = ops->fro_net;
201
202 spin_lock(&net->rules_mod_lock);
203 list_del_rcu(&ops->list);
204 spin_unlock(&net->rules_mod_lock);
205
206 fib_rules_cleanup_ops(ops);
207 kfree_rcu(ops, rcu);
208}
209EXPORT_SYMBOL_GPL(fib_rules_unregister);
210
211static int uid_range_set(struct fib_kuid_range *range)
212{
213 return uid_valid(range->start) && uid_valid(range->end);
214}
215
216static struct fib_kuid_range nla_get_kuid_range(struct nlattr **tb)
217{
218 struct fib_rule_uid_range *in;
219 struct fib_kuid_range out;
220
221 in = (struct fib_rule_uid_range *)nla_data(tb[FRA_UID_RANGE]);
222
223 out.start = make_kuid(current_user_ns(), in->start);
224 out.end = make_kuid(current_user_ns(), in->end);
225
226 return out;
227}
228
229static int nla_put_uid_range(struct sk_buff *skb, struct fib_kuid_range *range)
230{
231 struct fib_rule_uid_range out = {
232 from_kuid_munged(current_user_ns(), range->start),
233 from_kuid_munged(current_user_ns(), range->end)
234 };
235
236 return nla_put(skb, FRA_UID_RANGE, sizeof(out), &out);
237}
238
239static int nla_get_port_range(struct nlattr *pattr,
240 struct fib_rule_port_range *port_range)
241{
242 const struct fib_rule_port_range *pr = nla_data(pattr);
243
244 if (!fib_rule_port_range_valid(pr))
245 return -EINVAL;
246
247 port_range->start = pr->start;
248 port_range->end = pr->end;
249
250 return 0;
251}
252
253static int nla_put_port_range(struct sk_buff *skb, int attrtype,
254 struct fib_rule_port_range *range)
255{
256 return nla_put(skb, attrtype, sizeof(*range), range);
257}
258
259static int fib_rule_match(struct fib_rule *rule, struct fib_rules_ops *ops,
260 struct flowi *fl, int flags,
261 struct fib_lookup_arg *arg)
262{
263 int ret = 0;
264
265 if (rule->iifindex && (rule->iifindex != fl->flowi_iif))
266 goto out;
267
268 if (rule->oifindex && (rule->oifindex != fl->flowi_oif))
269 goto out;
270
271 if ((rule->mark ^ fl->flowi_mark) & rule->mark_mask)
272 goto out;
273
274 if (rule->tun_id && (rule->tun_id != fl->flowi_tun_key.tun_id))
275 goto out;
276
277 if (rule->l3mdev && !l3mdev_fib_rule_match(rule->fr_net, fl, arg))
278 goto out;
279
280 if (uid_lt(fl->flowi_uid, rule->uid_range.start) ||
281 uid_gt(fl->flowi_uid, rule->uid_range.end))
282 goto out;
283
284 ret = INDIRECT_CALL_MT(ops->match,
285 fib6_rule_match,
286 fib4_rule_match,
287 rule, fl, flags);
288out:
289 return (rule->flags & FIB_RULE_INVERT) ? !ret : ret;
290}
291
292int fib_rules_lookup(struct fib_rules_ops *ops, struct flowi *fl,
293 int flags, struct fib_lookup_arg *arg)
294{
295 struct fib_rule *rule;
296 int err;
297
298 rcu_read_lock();
299
300 list_for_each_entry_rcu(rule, &ops->rules_list, list) {
301jumped:
302 if (!fib_rule_match(rule, ops, fl, flags, arg))
303 continue;
304
305 if (rule->action == FR_ACT_GOTO) {
306 struct fib_rule *target;
307
308 target = rcu_dereference(rule->ctarget);
309 if (target == NULL) {
310 continue;
311 } else {
312 rule = target;
313 goto jumped;
314 }
315 } else if (rule->action == FR_ACT_NOP)
316 continue;
317 else
318 err = INDIRECT_CALL_MT(ops->action,
319 fib6_rule_action,
320 fib4_rule_action,
321 rule, fl, flags, arg);
322
323 if (!err && ops->suppress && INDIRECT_CALL_MT(ops->suppress,
324 fib6_rule_suppress,
325 fib4_rule_suppress,
326 rule, arg))
327 continue;
328
329 if (err != -EAGAIN) {
330 if ((arg->flags & FIB_LOOKUP_NOREF) ||
331 likely(refcount_inc_not_zero(&rule->refcnt))) {
332 arg->rule = rule;
333 goto out;
334 }
335 break;
336 }
337 }
338
339 err = -ESRCH;
340out:
341 rcu_read_unlock();
342
343 return err;
344}
345EXPORT_SYMBOL_GPL(fib_rules_lookup);
346
347static int call_fib_rule_notifier(struct notifier_block *nb,
348 enum fib_event_type event_type,
349 struct fib_rule *rule, int family,
350 struct netlink_ext_ack *extack)
351{
352 struct fib_rule_notifier_info info = {
353 .info.family = family,
354 .info.extack = extack,
355 .rule = rule,
356 };
357
358 return call_fib_notifier(nb, event_type, &info.info);
359}
360
361static int call_fib_rule_notifiers(struct net *net,
362 enum fib_event_type event_type,
363 struct fib_rule *rule,
364 struct fib_rules_ops *ops,
365 struct netlink_ext_ack *extack)
366{
367 struct fib_rule_notifier_info info = {
368 .info.family = ops->family,
369 .info.extack = extack,
370 .rule = rule,
371 };
372
373 ops->fib_rules_seq++;
374 return call_fib_notifiers(net, event_type, &info.info);
375}
376
377/* Called with rcu_read_lock() */
378int fib_rules_dump(struct net *net, struct notifier_block *nb, int family,
379 struct netlink_ext_ack *extack)
380{
381 struct fib_rules_ops *ops;
382 struct fib_rule *rule;
383 int err = 0;
384
385 ops = lookup_rules_ops(net, family);
386 if (!ops)
387 return -EAFNOSUPPORT;
388 list_for_each_entry_rcu(rule, &ops->rules_list, list) {
389 err = call_fib_rule_notifier(nb, FIB_EVENT_RULE_ADD,
390 rule, family, extack);
391 if (err)
392 break;
393 }
394 rules_ops_put(ops);
395
396 return err;
397}
398EXPORT_SYMBOL_GPL(fib_rules_dump);
399
400unsigned int fib_rules_seq_read(struct net *net, int family)
401{
402 unsigned int fib_rules_seq;
403 struct fib_rules_ops *ops;
404
405 ASSERT_RTNL();
406
407 ops = lookup_rules_ops(net, family);
408 if (!ops)
409 return 0;
410 fib_rules_seq = ops->fib_rules_seq;
411 rules_ops_put(ops);
412
413 return fib_rules_seq;
414}
415EXPORT_SYMBOL_GPL(fib_rules_seq_read);
416
417static struct fib_rule *rule_find(struct fib_rules_ops *ops,
418 struct fib_rule_hdr *frh,
419 struct nlattr **tb,
420 struct fib_rule *rule,
421 bool user_priority)
422{
423 struct fib_rule *r;
424
425 list_for_each_entry(r, &ops->rules_list, list) {
426 if (rule->action && r->action != rule->action)
427 continue;
428
429 if (rule->table && r->table != rule->table)
430 continue;
431
432 if (user_priority && r->pref != rule->pref)
433 continue;
434
435 if (rule->iifname[0] &&
436 memcmp(r->iifname, rule->iifname, IFNAMSIZ))
437 continue;
438
439 if (rule->oifname[0] &&
440 memcmp(r->oifname, rule->oifname, IFNAMSIZ))
441 continue;
442
443 if (rule->mark && r->mark != rule->mark)
444 continue;
445
446 if (rule->suppress_ifgroup != -1 &&
447 r->suppress_ifgroup != rule->suppress_ifgroup)
448 continue;
449
450 if (rule->suppress_prefixlen != -1 &&
451 r->suppress_prefixlen != rule->suppress_prefixlen)
452 continue;
453
454 if (rule->mark_mask && r->mark_mask != rule->mark_mask)
455 continue;
456
457 if (rule->tun_id && r->tun_id != rule->tun_id)
458 continue;
459
460 if (r->fr_net != rule->fr_net)
461 continue;
462
463 if (rule->l3mdev && r->l3mdev != rule->l3mdev)
464 continue;
465
466 if (uid_range_set(&rule->uid_range) &&
467 (!uid_eq(r->uid_range.start, rule->uid_range.start) ||
468 !uid_eq(r->uid_range.end, rule->uid_range.end)))
469 continue;
470
471 if (rule->ip_proto && r->ip_proto != rule->ip_proto)
472 continue;
473
474 if (rule->proto && r->proto != rule->proto)
475 continue;
476
477 if (fib_rule_port_range_set(&rule->sport_range) &&
478 !fib_rule_port_range_compare(&r->sport_range,
479 &rule->sport_range))
480 continue;
481
482 if (fib_rule_port_range_set(&rule->dport_range) &&
483 !fib_rule_port_range_compare(&r->dport_range,
484 &rule->dport_range))
485 continue;
486
487 if (!ops->compare(r, frh, tb))
488 continue;
489 return r;
490 }
491
492 return NULL;
493}
494
495#ifdef CONFIG_NET_L3_MASTER_DEV
496static int fib_nl2rule_l3mdev(struct nlattr *nla, struct fib_rule *nlrule,
497 struct netlink_ext_ack *extack)
498{
499 nlrule->l3mdev = nla_get_u8(nla);
500 if (nlrule->l3mdev != 1) {
501 NL_SET_ERR_MSG(extack, "Invalid l3mdev attribute");
502 return -1;
503 }
504
505 return 0;
506}
507#else
508static int fib_nl2rule_l3mdev(struct nlattr *nla, struct fib_rule *nlrule,
509 struct netlink_ext_ack *extack)
510{
511 NL_SET_ERR_MSG(extack, "l3mdev support is not enabled in kernel");
512 return -1;
513}
514#endif
515
516static int fib_nl2rule(struct sk_buff *skb, struct nlmsghdr *nlh,
517 struct netlink_ext_ack *extack,
518 struct fib_rules_ops *ops,
519 struct nlattr *tb[],
520 struct fib_rule **rule,
521 bool *user_priority)
522{
523 struct net *net = sock_net(skb->sk);
524 struct fib_rule_hdr *frh = nlmsg_data(nlh);
525 struct fib_rule *nlrule = NULL;
526 int err = -EINVAL;
527
528 if (frh->src_len)
529 if (!tb[FRA_SRC] ||
530 frh->src_len > (ops->addr_size * 8) ||
531 nla_len(tb[FRA_SRC]) != ops->addr_size) {
532 NL_SET_ERR_MSG(extack, "Invalid source address");
533 goto errout;
534 }
535
536 if (frh->dst_len)
537 if (!tb[FRA_DST] ||
538 frh->dst_len > (ops->addr_size * 8) ||
539 nla_len(tb[FRA_DST]) != ops->addr_size) {
540 NL_SET_ERR_MSG(extack, "Invalid dst address");
541 goto errout;
542 }
543
544 nlrule = kzalloc(ops->rule_size, GFP_KERNEL);
545 if (!nlrule) {
546 err = -ENOMEM;
547 goto errout;
548 }
549 refcount_set(&nlrule->refcnt, 1);
550 nlrule->fr_net = net;
551
552 if (tb[FRA_PRIORITY]) {
553 nlrule->pref = nla_get_u32(tb[FRA_PRIORITY]);
554 *user_priority = true;
555 } else {
556 nlrule->pref = fib_default_rule_pref(ops);
557 }
558
559 nlrule->proto = tb[FRA_PROTOCOL] ?
560 nla_get_u8(tb[FRA_PROTOCOL]) : RTPROT_UNSPEC;
561
562 if (tb[FRA_IIFNAME]) {
563 struct net_device *dev;
564
565 nlrule->iifindex = -1;
566 nla_strlcpy(nlrule->iifname, tb[FRA_IIFNAME], IFNAMSIZ);
567 dev = __dev_get_by_name(net, nlrule->iifname);
568 if (dev)
569 nlrule->iifindex = dev->ifindex;
570 }
571
572 if (tb[FRA_OIFNAME]) {
573 struct net_device *dev;
574
575 nlrule->oifindex = -1;
576 nla_strlcpy(nlrule->oifname, tb[FRA_OIFNAME], IFNAMSIZ);
577 dev = __dev_get_by_name(net, nlrule->oifname);
578 if (dev)
579 nlrule->oifindex = dev->ifindex;
580 }
581
582 if (tb[FRA_FWMARK]) {
583 nlrule->mark = nla_get_u32(tb[FRA_FWMARK]);
584 if (nlrule->mark)
585 /* compatibility: if the mark value is non-zero all bits
586 * are compared unless a mask is explicitly specified.
587 */
588 nlrule->mark_mask = 0xFFFFFFFF;
589 }
590
591 if (tb[FRA_FWMASK])
592 nlrule->mark_mask = nla_get_u32(tb[FRA_FWMASK]);
593
594 if (tb[FRA_TUN_ID])
595 nlrule->tun_id = nla_get_be64(tb[FRA_TUN_ID]);
596
597 err = -EINVAL;
598 if (tb[FRA_L3MDEV] &&
599 fib_nl2rule_l3mdev(tb[FRA_L3MDEV], nlrule, extack) < 0)
600 goto errout_free;
601
602 nlrule->action = frh->action;
603 nlrule->flags = frh->flags;
604 nlrule->table = frh_get_table(frh, tb);
605 if (tb[FRA_SUPPRESS_PREFIXLEN])
606 nlrule->suppress_prefixlen = nla_get_u32(tb[FRA_SUPPRESS_PREFIXLEN]);
607 else
608 nlrule->suppress_prefixlen = -1;
609
610 if (tb[FRA_SUPPRESS_IFGROUP])
611 nlrule->suppress_ifgroup = nla_get_u32(tb[FRA_SUPPRESS_IFGROUP]);
612 else
613 nlrule->suppress_ifgroup = -1;
614
615 if (tb[FRA_GOTO]) {
616 if (nlrule->action != FR_ACT_GOTO) {
617 NL_SET_ERR_MSG(extack, "Unexpected goto");
618 goto errout_free;
619 }
620
621 nlrule->target = nla_get_u32(tb[FRA_GOTO]);
622 /* Backward jumps are prohibited to avoid endless loops */
623 if (nlrule->target <= nlrule->pref) {
624 NL_SET_ERR_MSG(extack, "Backward goto not supported");
625 goto errout_free;
626 }
627 } else if (nlrule->action == FR_ACT_GOTO) {
628 NL_SET_ERR_MSG(extack, "Missing goto target for action goto");
629 goto errout_free;
630 }
631
632 if (nlrule->l3mdev && nlrule->table) {
633 NL_SET_ERR_MSG(extack, "l3mdev and table are mutually exclusive");
634 goto errout_free;
635 }
636
637 if (tb[FRA_UID_RANGE]) {
638 if (current_user_ns() != net->user_ns) {
639 err = -EPERM;
640 NL_SET_ERR_MSG(extack, "No permission to set uid");
641 goto errout_free;
642 }
643
644 nlrule->uid_range = nla_get_kuid_range(tb);
645
646 if (!uid_range_set(&nlrule->uid_range) ||
647 !uid_lte(nlrule->uid_range.start, nlrule->uid_range.end)) {
648 NL_SET_ERR_MSG(extack, "Invalid uid range");
649 goto errout_free;
650 }
651 } else {
652 nlrule->uid_range = fib_kuid_range_unset;
653 }
654
655 if (tb[FRA_IP_PROTO])
656 nlrule->ip_proto = nla_get_u8(tb[FRA_IP_PROTO]);
657
658 if (tb[FRA_SPORT_RANGE]) {
659 err = nla_get_port_range(tb[FRA_SPORT_RANGE],
660 &nlrule->sport_range);
661 if (err) {
662 NL_SET_ERR_MSG(extack, "Invalid sport range");
663 goto errout_free;
664 }
665 }
666
667 if (tb[FRA_DPORT_RANGE]) {
668 err = nla_get_port_range(tb[FRA_DPORT_RANGE],
669 &nlrule->dport_range);
670 if (err) {
671 NL_SET_ERR_MSG(extack, "Invalid dport range");
672 goto errout_free;
673 }
674 }
675
676 *rule = nlrule;
677
678 return 0;
679
680errout_free:
681 kfree(nlrule);
682errout:
683 return err;
684}
685
686static int rule_exists(struct fib_rules_ops *ops, struct fib_rule_hdr *frh,
687 struct nlattr **tb, struct fib_rule *rule)
688{
689 struct fib_rule *r;
690
691 list_for_each_entry(r, &ops->rules_list, list) {
692 if (r->action != rule->action)
693 continue;
694
695 if (r->table != rule->table)
696 continue;
697
698 if (r->pref != rule->pref)
699 continue;
700
701 if (memcmp(r->iifname, rule->iifname, IFNAMSIZ))
702 continue;
703
704 if (memcmp(r->oifname, rule->oifname, IFNAMSIZ))
705 continue;
706
707 if (r->mark != rule->mark)
708 continue;
709
710 if (r->suppress_ifgroup != rule->suppress_ifgroup)
711 continue;
712
713 if (r->suppress_prefixlen != rule->suppress_prefixlen)
714 continue;
715
716 if (r->mark_mask != rule->mark_mask)
717 continue;
718
719 if (r->tun_id != rule->tun_id)
720 continue;
721
722 if (r->fr_net != rule->fr_net)
723 continue;
724
725 if (r->l3mdev != rule->l3mdev)
726 continue;
727
728 if (!uid_eq(r->uid_range.start, rule->uid_range.start) ||
729 !uid_eq(r->uid_range.end, rule->uid_range.end))
730 continue;
731
732 if (r->ip_proto != rule->ip_proto)
733 continue;
734
735 if (r->proto != rule->proto)
736 continue;
737
738 if (!fib_rule_port_range_compare(&r->sport_range,
739 &rule->sport_range))
740 continue;
741
742 if (!fib_rule_port_range_compare(&r->dport_range,
743 &rule->dport_range))
744 continue;
745
746 if (!ops->compare(r, frh, tb))
747 continue;
748 return 1;
749 }
750 return 0;
751}
752
753int fib_nl_newrule(struct sk_buff *skb, struct nlmsghdr *nlh,
754 struct netlink_ext_ack *extack)
755{
756 struct net *net = sock_net(skb->sk);
757 struct fib_rule_hdr *frh = nlmsg_data(nlh);
758 struct fib_rules_ops *ops = NULL;
759 struct fib_rule *rule = NULL, *r, *last = NULL;
760 struct nlattr *tb[FRA_MAX + 1];
761 int err = -EINVAL, unresolved = 0;
762 bool user_priority = false;
763
764 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh))) {
765 NL_SET_ERR_MSG(extack, "Invalid msg length");
766 goto errout;
767 }
768
769 ops = lookup_rules_ops(net, frh->family);
770 if (!ops) {
771 err = -EAFNOSUPPORT;
772 NL_SET_ERR_MSG(extack, "Rule family not supported");
773 goto errout;
774 }
775
776 err = nlmsg_parse_deprecated(nlh, sizeof(*frh), tb, FRA_MAX,
777 ops->policy, extack);
778 if (err < 0) {
779 NL_SET_ERR_MSG(extack, "Error parsing msg");
780 goto errout;
781 }
782
783 err = fib_nl2rule(skb, nlh, extack, ops, tb, &rule, &user_priority);
784 if (err)
785 goto errout;
786
787 if ((nlh->nlmsg_flags & NLM_F_EXCL) &&
788 rule_exists(ops, frh, tb, rule)) {
789 err = -EEXIST;
790 goto errout_free;
791 }
792
793 err = ops->configure(rule, skb, frh, tb, extack);
794 if (err < 0)
795 goto errout_free;
796
797 err = call_fib_rule_notifiers(net, FIB_EVENT_RULE_ADD, rule, ops,
798 extack);
799 if (err < 0)
800 goto errout_free;
801
802 list_for_each_entry(r, &ops->rules_list, list) {
803 if (r->pref == rule->target) {
804 RCU_INIT_POINTER(rule->ctarget, r);
805 break;
806 }
807 }
808
809 if (rcu_dereference_protected(rule->ctarget, 1) == NULL)
810 unresolved = 1;
811
812 list_for_each_entry(r, &ops->rules_list, list) {
813 if (r->pref > rule->pref)
814 break;
815 last = r;
816 }
817
818 if (last)
819 list_add_rcu(&rule->list, &last->list);
820 else
821 list_add_rcu(&rule->list, &ops->rules_list);
822
823 if (ops->unresolved_rules) {
824 /*
825 * There are unresolved goto rules in the list, check if
826 * any of them are pointing to this new rule.
827 */
828 list_for_each_entry(r, &ops->rules_list, list) {
829 if (r->action == FR_ACT_GOTO &&
830 r->target == rule->pref &&
831 rtnl_dereference(r->ctarget) == NULL) {
832 rcu_assign_pointer(r->ctarget, rule);
833 if (--ops->unresolved_rules == 0)
834 break;
835 }
836 }
837 }
838
839 if (rule->action == FR_ACT_GOTO)
840 ops->nr_goto_rules++;
841
842 if (unresolved)
843 ops->unresolved_rules++;
844
845 if (rule->tun_id)
846 ip_tunnel_need_metadata();
847
848 notify_rule_change(RTM_NEWRULE, rule, ops, nlh, NETLINK_CB(skb).portid);
849 flush_route_cache(ops);
850 rules_ops_put(ops);
851 return 0;
852
853errout_free:
854 kfree(rule);
855errout:
856 rules_ops_put(ops);
857 return err;
858}
859EXPORT_SYMBOL_GPL(fib_nl_newrule);
860
861int fib_nl_delrule(struct sk_buff *skb, struct nlmsghdr *nlh,
862 struct netlink_ext_ack *extack)
863{
864 struct net *net = sock_net(skb->sk);
865 struct fib_rule_hdr *frh = nlmsg_data(nlh);
866 struct fib_rules_ops *ops = NULL;
867 struct fib_rule *rule = NULL, *r, *nlrule = NULL;
868 struct nlattr *tb[FRA_MAX+1];
869 int err = -EINVAL;
870 bool user_priority = false;
871
872 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh))) {
873 NL_SET_ERR_MSG(extack, "Invalid msg length");
874 goto errout;
875 }
876
877 ops = lookup_rules_ops(net, frh->family);
878 if (ops == NULL) {
879 err = -EAFNOSUPPORT;
880 NL_SET_ERR_MSG(extack, "Rule family not supported");
881 goto errout;
882 }
883
884 err = nlmsg_parse_deprecated(nlh, sizeof(*frh), tb, FRA_MAX,
885 ops->policy, extack);
886 if (err < 0) {
887 NL_SET_ERR_MSG(extack, "Error parsing msg");
888 goto errout;
889 }
890
891 err = fib_nl2rule(skb, nlh, extack, ops, tb, &nlrule, &user_priority);
892 if (err)
893 goto errout;
894
895 rule = rule_find(ops, frh, tb, nlrule, user_priority);
896 if (!rule) {
897 err = -ENOENT;
898 goto errout;
899 }
900
901 if (rule->flags & FIB_RULE_PERMANENT) {
902 err = -EPERM;
903 goto errout;
904 }
905
906 if (ops->delete) {
907 err = ops->delete(rule);
908 if (err)
909 goto errout;
910 }
911
912 if (rule->tun_id)
913 ip_tunnel_unneed_metadata();
914
915 list_del_rcu(&rule->list);
916
917 if (rule->action == FR_ACT_GOTO) {
918 ops->nr_goto_rules--;
919 if (rtnl_dereference(rule->ctarget) == NULL)
920 ops->unresolved_rules--;
921 }
922
923 /*
924 * Check if this rule is a target to any of them. If so,
925 * adjust to the next one with the same preference or
926 * disable them. As this operation is eventually very
927 * expensive, it is only performed if goto rules, except
928 * current if it is goto rule, have actually been added.
929 */
930 if (ops->nr_goto_rules > 0) {
931 struct fib_rule *n;
932
933 n = list_next_entry(rule, list);
934 if (&n->list == &ops->rules_list || n->pref != rule->pref)
935 n = NULL;
936 list_for_each_entry(r, &ops->rules_list, list) {
937 if (rtnl_dereference(r->ctarget) != rule)
938 continue;
939 rcu_assign_pointer(r->ctarget, n);
940 if (!n)
941 ops->unresolved_rules++;
942 }
943 }
944
945 call_fib_rule_notifiers(net, FIB_EVENT_RULE_DEL, rule, ops,
946 NULL);
947 notify_rule_change(RTM_DELRULE, rule, ops, nlh,
948 NETLINK_CB(skb).portid);
949 fib_rule_put(rule);
950 flush_route_cache(ops);
951 rules_ops_put(ops);
952 kfree(nlrule);
953 return 0;
954
955errout:
956 kfree(nlrule);
957 rules_ops_put(ops);
958 return err;
959}
960EXPORT_SYMBOL_GPL(fib_nl_delrule);
961
962static inline size_t fib_rule_nlmsg_size(struct fib_rules_ops *ops,
963 struct fib_rule *rule)
964{
965 size_t payload = NLMSG_ALIGN(sizeof(struct fib_rule_hdr))
966 + nla_total_size(IFNAMSIZ) /* FRA_IIFNAME */
967 + nla_total_size(IFNAMSIZ) /* FRA_OIFNAME */
968 + nla_total_size(4) /* FRA_PRIORITY */
969 + nla_total_size(4) /* FRA_TABLE */
970 + nla_total_size(4) /* FRA_SUPPRESS_PREFIXLEN */
971 + nla_total_size(4) /* FRA_SUPPRESS_IFGROUP */
972 + nla_total_size(4) /* FRA_FWMARK */
973 + nla_total_size(4) /* FRA_FWMASK */
974 + nla_total_size_64bit(8) /* FRA_TUN_ID */
975 + nla_total_size(sizeof(struct fib_kuid_range))
976 + nla_total_size(1) /* FRA_PROTOCOL */
977 + nla_total_size(1) /* FRA_IP_PROTO */
978 + nla_total_size(sizeof(struct fib_rule_port_range)) /* FRA_SPORT_RANGE */
979 + nla_total_size(sizeof(struct fib_rule_port_range)); /* FRA_DPORT_RANGE */
980
981 if (ops->nlmsg_payload)
982 payload += ops->nlmsg_payload(rule);
983
984 return payload;
985}
986
987static int fib_nl_fill_rule(struct sk_buff *skb, struct fib_rule *rule,
988 u32 pid, u32 seq, int type, int flags,
989 struct fib_rules_ops *ops)
990{
991 struct nlmsghdr *nlh;
992 struct fib_rule_hdr *frh;
993
994 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*frh), flags);
995 if (nlh == NULL)
996 return -EMSGSIZE;
997
998 frh = nlmsg_data(nlh);
999 frh->family = ops->family;
1000 frh->table = rule->table < 256 ? rule->table : RT_TABLE_COMPAT;
1001 if (nla_put_u32(skb, FRA_TABLE, rule->table))
1002 goto nla_put_failure;
1003 if (nla_put_u32(skb, FRA_SUPPRESS_PREFIXLEN, rule->suppress_prefixlen))
1004 goto nla_put_failure;
1005 frh->res1 = 0;
1006 frh->res2 = 0;
1007 frh->action = rule->action;
1008 frh->flags = rule->flags;
1009
1010 if (nla_put_u8(skb, FRA_PROTOCOL, rule->proto))
1011 goto nla_put_failure;
1012
1013 if (rule->action == FR_ACT_GOTO &&
1014 rcu_access_pointer(rule->ctarget) == NULL)
1015 frh->flags |= FIB_RULE_UNRESOLVED;
1016
1017 if (rule->iifname[0]) {
1018 if (nla_put_string(skb, FRA_IIFNAME, rule->iifname))
1019 goto nla_put_failure;
1020 if (rule->iifindex == -1)
1021 frh->flags |= FIB_RULE_IIF_DETACHED;
1022 }
1023
1024 if (rule->oifname[0]) {
1025 if (nla_put_string(skb, FRA_OIFNAME, rule->oifname))
1026 goto nla_put_failure;
1027 if (rule->oifindex == -1)
1028 frh->flags |= FIB_RULE_OIF_DETACHED;
1029 }
1030
1031 if ((rule->pref &&
1032 nla_put_u32(skb, FRA_PRIORITY, rule->pref)) ||
1033 (rule->mark &&
1034 nla_put_u32(skb, FRA_FWMARK, rule->mark)) ||
1035 ((rule->mark_mask || rule->mark) &&
1036 nla_put_u32(skb, FRA_FWMASK, rule->mark_mask)) ||
1037 (rule->target &&
1038 nla_put_u32(skb, FRA_GOTO, rule->target)) ||
1039 (rule->tun_id &&
1040 nla_put_be64(skb, FRA_TUN_ID, rule->tun_id, FRA_PAD)) ||
1041 (rule->l3mdev &&
1042 nla_put_u8(skb, FRA_L3MDEV, rule->l3mdev)) ||
1043 (uid_range_set(&rule->uid_range) &&
1044 nla_put_uid_range(skb, &rule->uid_range)) ||
1045 (fib_rule_port_range_set(&rule->sport_range) &&
1046 nla_put_port_range(skb, FRA_SPORT_RANGE, &rule->sport_range)) ||
1047 (fib_rule_port_range_set(&rule->dport_range) &&
1048 nla_put_port_range(skb, FRA_DPORT_RANGE, &rule->dport_range)) ||
1049 (rule->ip_proto && nla_put_u8(skb, FRA_IP_PROTO, rule->ip_proto)))
1050 goto nla_put_failure;
1051
1052 if (rule->suppress_ifgroup != -1) {
1053 if (nla_put_u32(skb, FRA_SUPPRESS_IFGROUP, rule->suppress_ifgroup))
1054 goto nla_put_failure;
1055 }
1056
1057 if (ops->fill(rule, skb, frh) < 0)
1058 goto nla_put_failure;
1059
1060 nlmsg_end(skb, nlh);
1061 return 0;
1062
1063nla_put_failure:
1064 nlmsg_cancel(skb, nlh);
1065 return -EMSGSIZE;
1066}
1067
1068static int dump_rules(struct sk_buff *skb, struct netlink_callback *cb,
1069 struct fib_rules_ops *ops)
1070{
1071 int idx = 0;
1072 struct fib_rule *rule;
1073 int err = 0;
1074
1075 rcu_read_lock();
1076 list_for_each_entry_rcu(rule, &ops->rules_list, list) {
1077 if (idx < cb->args[1])
1078 goto skip;
1079
1080 err = fib_nl_fill_rule(skb, rule, NETLINK_CB(cb->skb).portid,
1081 cb->nlh->nlmsg_seq, RTM_NEWRULE,
1082 NLM_F_MULTI, ops);
1083 if (err)
1084 break;
1085skip:
1086 idx++;
1087 }
1088 rcu_read_unlock();
1089 cb->args[1] = idx;
1090 rules_ops_put(ops);
1091
1092 return err;
1093}
1094
1095static int fib_valid_dumprule_req(const struct nlmsghdr *nlh,
1096 struct netlink_ext_ack *extack)
1097{
1098 struct fib_rule_hdr *frh;
1099
1100 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh))) {
1101 NL_SET_ERR_MSG(extack, "Invalid header for fib rule dump request");
1102 return -EINVAL;
1103 }
1104
1105 frh = nlmsg_data(nlh);
1106 if (frh->dst_len || frh->src_len || frh->tos || frh->table ||
1107 frh->res1 || frh->res2 || frh->action || frh->flags) {
1108 NL_SET_ERR_MSG(extack,
1109 "Invalid values in header for fib rule dump request");
1110 return -EINVAL;
1111 }
1112
1113 if (nlmsg_attrlen(nlh, sizeof(*frh))) {
1114 NL_SET_ERR_MSG(extack, "Invalid data after header in fib rule dump request");
1115 return -EINVAL;
1116 }
1117
1118 return 0;
1119}
1120
1121static int fib_nl_dumprule(struct sk_buff *skb, struct netlink_callback *cb)
1122{
1123 const struct nlmsghdr *nlh = cb->nlh;
1124 struct net *net = sock_net(skb->sk);
1125 struct fib_rules_ops *ops;
1126 int idx = 0, family;
1127
1128 if (cb->strict_check) {
1129 int err = fib_valid_dumprule_req(nlh, cb->extack);
1130
1131 if (err < 0)
1132 return err;
1133 }
1134
1135 family = rtnl_msg_family(nlh);
1136 if (family != AF_UNSPEC) {
1137 /* Protocol specific dump request */
1138 ops = lookup_rules_ops(net, family);
1139 if (ops == NULL)
1140 return -EAFNOSUPPORT;
1141
1142 dump_rules(skb, cb, ops);
1143
1144 return skb->len;
1145 }
1146
1147 rcu_read_lock();
1148 list_for_each_entry_rcu(ops, &net->rules_ops, list) {
1149 if (idx < cb->args[0] || !try_module_get(ops->owner))
1150 goto skip;
1151
1152 if (dump_rules(skb, cb, ops) < 0)
1153 break;
1154
1155 cb->args[1] = 0;
1156skip:
1157 idx++;
1158 }
1159 rcu_read_unlock();
1160 cb->args[0] = idx;
1161
1162 return skb->len;
1163}
1164
1165static void notify_rule_change(int event, struct fib_rule *rule,
1166 struct fib_rules_ops *ops, struct nlmsghdr *nlh,
1167 u32 pid)
1168{
1169 struct net *net;
1170 struct sk_buff *skb;
1171 int err = -ENOBUFS;
1172
1173 net = ops->fro_net;
1174 skb = nlmsg_new(fib_rule_nlmsg_size(ops, rule), GFP_KERNEL);
1175 if (skb == NULL)
1176 goto errout;
1177
1178 err = fib_nl_fill_rule(skb, rule, pid, nlh->nlmsg_seq, event, 0, ops);
1179 if (err < 0) {
1180 /* -EMSGSIZE implies BUG in fib_rule_nlmsg_size() */
1181 WARN_ON(err == -EMSGSIZE);
1182 kfree_skb(skb);
1183 goto errout;
1184 }
1185
1186 rtnl_notify(skb, net, pid, ops->nlgroup, nlh, GFP_KERNEL);
1187 return;
1188errout:
1189 if (err < 0)
1190 rtnl_set_sk_err(net, ops->nlgroup, err);
1191}
1192
1193static void attach_rules(struct list_head *rules, struct net_device *dev)
1194{
1195 struct fib_rule *rule;
1196
1197 list_for_each_entry(rule, rules, list) {
1198 if (rule->iifindex == -1 &&
1199 strcmp(dev->name, rule->iifname) == 0)
1200 rule->iifindex = dev->ifindex;
1201 if (rule->oifindex == -1 &&
1202 strcmp(dev->name, rule->oifname) == 0)
1203 rule->oifindex = dev->ifindex;
1204 }
1205}
1206
1207static void detach_rules(struct list_head *rules, struct net_device *dev)
1208{
1209 struct fib_rule *rule;
1210
1211 list_for_each_entry(rule, rules, list) {
1212 if (rule->iifindex == dev->ifindex)
1213 rule->iifindex = -1;
1214 if (rule->oifindex == dev->ifindex)
1215 rule->oifindex = -1;
1216 }
1217}
1218
1219
1220static int fib_rules_event(struct notifier_block *this, unsigned long event,
1221 void *ptr)
1222{
1223 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1224 struct net *net = dev_net(dev);
1225 struct fib_rules_ops *ops;
1226
1227 ASSERT_RTNL();
1228
1229 switch (event) {
1230 case NETDEV_REGISTER:
1231 list_for_each_entry(ops, &net->rules_ops, list)
1232 attach_rules(&ops->rules_list, dev);
1233 break;
1234
1235 case NETDEV_CHANGENAME:
1236 list_for_each_entry(ops, &net->rules_ops, list) {
1237 detach_rules(&ops->rules_list, dev);
1238 attach_rules(&ops->rules_list, dev);
1239 }
1240 break;
1241
1242 case NETDEV_UNREGISTER:
1243 list_for_each_entry(ops, &net->rules_ops, list)
1244 detach_rules(&ops->rules_list, dev);
1245 break;
1246 }
1247
1248 return NOTIFY_DONE;
1249}
1250
1251static struct notifier_block fib_rules_notifier = {
1252 .notifier_call = fib_rules_event,
1253};
1254
1255static int __net_init fib_rules_net_init(struct net *net)
1256{
1257 INIT_LIST_HEAD(&net->rules_ops);
1258 spin_lock_init(&net->rules_mod_lock);
1259 return 0;
1260}
1261
1262static void __net_exit fib_rules_net_exit(struct net *net)
1263{
1264 WARN_ON_ONCE(!list_empty(&net->rules_ops));
1265}
1266
1267static struct pernet_operations fib_rules_net_ops = {
1268 .init = fib_rules_net_init,
1269 .exit = fib_rules_net_exit,
1270};
1271
1272static int __init fib_rules_init(void)
1273{
1274 int err;
1275 rtnl_register(PF_UNSPEC, RTM_NEWRULE, fib_nl_newrule, NULL, 0);
1276 rtnl_register(PF_UNSPEC, RTM_DELRULE, fib_nl_delrule, NULL, 0);
1277 rtnl_register(PF_UNSPEC, RTM_GETRULE, NULL, fib_nl_dumprule, 0);
1278
1279 err = register_pernet_subsys(&fib_rules_net_ops);
1280 if (err < 0)
1281 goto fail;
1282
1283 err = register_netdevice_notifier(&fib_rules_notifier);
1284 if (err < 0)
1285 goto fail_unregister;
1286
1287 return 0;
1288
1289fail_unregister:
1290 unregister_pernet_subsys(&fib_rules_net_ops);
1291fail:
1292 rtnl_unregister(PF_UNSPEC, RTM_NEWRULE);
1293 rtnl_unregister(PF_UNSPEC, RTM_DELRULE);
1294 rtnl_unregister(PF_UNSPEC, RTM_GETRULE);
1295 return err;
1296}
1297
1298subsys_initcall(fib_rules_init);