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