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