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