rtc: Namespace fixup
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / core / fib_rules.c
blob42e84e08a1becd4b64a65a74758bf4e13783a8ae
1 /*
2 * net/core/fib_rules.c Generic Routing Rules
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.
8 * Authors: Thomas Graf <tgraf@suug.ch>
9 */
11 #include <linux/types.h>
12 #include <linux/kernel.h>
13 #include <linux/slab.h>
14 #include <linux/list.h>
15 #include <net/net_namespace.h>
16 #include <net/sock.h>
17 #include <net/fib_rules.h>
19 int fib_default_rule_add(struct fib_rules_ops *ops,
20 u32 pref, u32 table, u32 flags)
22 struct fib_rule *r;
24 r = kzalloc(ops->rule_size, GFP_KERNEL);
25 if (r == NULL)
26 return -ENOMEM;
28 atomic_set(&r->refcnt, 1);
29 r->action = FR_ACT_TO_TBL;
30 r->pref = pref;
31 r->table = table;
32 r->flags = flags;
33 r->fr_net = hold_net(ops->fro_net);
35 /* The lock is not required here, the list in unreacheable
36 * at the moment this function is called */
37 list_add_tail(&r->list, &ops->rules_list);
38 return 0;
40 EXPORT_SYMBOL(fib_default_rule_add);
42 u32 fib_default_rule_pref(struct fib_rules_ops *ops)
44 struct list_head *pos;
45 struct fib_rule *rule;
47 if (!list_empty(&ops->rules_list)) {
48 pos = ops->rules_list.next;
49 if (pos->next != &ops->rules_list) {
50 rule = list_entry(pos->next, struct fib_rule, list);
51 if (rule->pref)
52 return rule->pref - 1;
56 return 0;
58 EXPORT_SYMBOL(fib_default_rule_pref);
60 static void notify_rule_change(int event, struct fib_rule *rule,
61 struct fib_rules_ops *ops, struct nlmsghdr *nlh,
62 u32 pid);
64 static struct fib_rules_ops *lookup_rules_ops(struct net *net, int family)
66 struct fib_rules_ops *ops;
68 rcu_read_lock();
69 list_for_each_entry_rcu(ops, &net->rules_ops, list) {
70 if (ops->family == family) {
71 if (!try_module_get(ops->owner))
72 ops = NULL;
73 rcu_read_unlock();
74 return ops;
77 rcu_read_unlock();
79 return NULL;
82 static void rules_ops_put(struct fib_rules_ops *ops)
84 if (ops)
85 module_put(ops->owner);
88 static void flush_route_cache(struct fib_rules_ops *ops)
90 if (ops->flush_cache)
91 ops->flush_cache(ops);
94 static int __fib_rules_register(struct fib_rules_ops *ops)
96 int err = -EEXIST;
97 struct fib_rules_ops *o;
98 struct net *net;
100 net = ops->fro_net;
102 if (ops->rule_size < sizeof(struct fib_rule))
103 return -EINVAL;
105 if (ops->match == NULL || ops->configure == NULL ||
106 ops->compare == NULL || ops->fill == NULL ||
107 ops->action == NULL)
108 return -EINVAL;
110 spin_lock(&net->rules_mod_lock);
111 list_for_each_entry(o, &net->rules_ops, list)
112 if (ops->family == o->family)
113 goto errout;
115 hold_net(net);
116 list_add_tail_rcu(&ops->list, &net->rules_ops);
117 err = 0;
118 errout:
119 spin_unlock(&net->rules_mod_lock);
121 return err;
124 struct fib_rules_ops *
125 fib_rules_register(const struct fib_rules_ops *tmpl, struct net *net)
127 struct fib_rules_ops *ops;
128 int err;
130 ops = kmemdup(tmpl, sizeof(*ops), GFP_KERNEL);
131 if (ops == NULL)
132 return ERR_PTR(-ENOMEM);
134 INIT_LIST_HEAD(&ops->rules_list);
135 ops->fro_net = net;
137 err = __fib_rules_register(ops);
138 if (err) {
139 kfree(ops);
140 ops = ERR_PTR(err);
143 return ops;
145 EXPORT_SYMBOL_GPL(fib_rules_register);
147 void fib_rules_cleanup_ops(struct fib_rules_ops *ops)
149 struct fib_rule *rule, *tmp;
151 list_for_each_entry_safe(rule, tmp, &ops->rules_list, list) {
152 list_del_rcu(&rule->list);
153 fib_rule_put(rule);
156 EXPORT_SYMBOL_GPL(fib_rules_cleanup_ops);
158 static void fib_rules_put_rcu(struct rcu_head *head)
160 struct fib_rules_ops *ops = container_of(head, struct fib_rules_ops, rcu);
161 struct net *net = ops->fro_net;
163 release_net(net);
164 kfree(ops);
167 void fib_rules_unregister(struct fib_rules_ops *ops)
169 struct net *net = ops->fro_net;
171 spin_lock(&net->rules_mod_lock);
172 list_del_rcu(&ops->list);
173 fib_rules_cleanup_ops(ops);
174 spin_unlock(&net->rules_mod_lock);
176 call_rcu(&ops->rcu, fib_rules_put_rcu);
178 EXPORT_SYMBOL_GPL(fib_rules_unregister);
180 static int fib_rule_match(struct fib_rule *rule, struct fib_rules_ops *ops,
181 struct flowi *fl, int flags)
183 int ret = 0;
185 if (rule->iifindex && (rule->iifindex != fl->iif))
186 goto out;
188 if (rule->oifindex && (rule->oifindex != fl->oif))
189 goto out;
191 if ((rule->mark ^ fl->mark) & rule->mark_mask)
192 goto out;
194 ret = ops->match(rule, fl, flags);
195 out:
196 return (rule->flags & FIB_RULE_INVERT) ? !ret : ret;
199 int fib_rules_lookup(struct fib_rules_ops *ops, struct flowi *fl,
200 int flags, struct fib_lookup_arg *arg)
202 struct fib_rule *rule;
203 int err;
205 rcu_read_lock();
207 list_for_each_entry_rcu(rule, &ops->rules_list, list) {
208 jumped:
209 if (!fib_rule_match(rule, ops, fl, flags))
210 continue;
212 if (rule->action == FR_ACT_GOTO) {
213 struct fib_rule *target;
215 target = rcu_dereference(rule->ctarget);
216 if (target == NULL) {
217 continue;
218 } else {
219 rule = target;
220 goto jumped;
222 } else if (rule->action == FR_ACT_NOP)
223 continue;
224 else
225 err = ops->action(rule, fl, flags, arg);
227 if (err != -EAGAIN) {
228 fib_rule_get(rule);
229 arg->rule = rule;
230 goto out;
234 err = -ESRCH;
235 out:
236 rcu_read_unlock();
238 return err;
240 EXPORT_SYMBOL_GPL(fib_rules_lookup);
242 static int validate_rulemsg(struct fib_rule_hdr *frh, struct nlattr **tb,
243 struct fib_rules_ops *ops)
245 int err = -EINVAL;
247 if (frh->src_len)
248 if (tb[FRA_SRC] == NULL ||
249 frh->src_len > (ops->addr_size * 8) ||
250 nla_len(tb[FRA_SRC]) != ops->addr_size)
251 goto errout;
253 if (frh->dst_len)
254 if (tb[FRA_DST] == NULL ||
255 frh->dst_len > (ops->addr_size * 8) ||
256 nla_len(tb[FRA_DST]) != ops->addr_size)
257 goto errout;
259 err = 0;
260 errout:
261 return err;
264 static int fib_nl_newrule(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
266 struct net *net = sock_net(skb->sk);
267 struct fib_rule_hdr *frh = nlmsg_data(nlh);
268 struct fib_rules_ops *ops = NULL;
269 struct fib_rule *rule, *r, *last = NULL;
270 struct nlattr *tb[FRA_MAX+1];
271 int err = -EINVAL, unresolved = 0;
273 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh)))
274 goto errout;
276 ops = lookup_rules_ops(net, frh->family);
277 if (ops == NULL) {
278 err = -EAFNOSUPPORT;
279 goto errout;
282 err = nlmsg_parse(nlh, sizeof(*frh), tb, FRA_MAX, ops->policy);
283 if (err < 0)
284 goto errout;
286 err = validate_rulemsg(frh, tb, ops);
287 if (err < 0)
288 goto errout;
290 rule = kzalloc(ops->rule_size, GFP_KERNEL);
291 if (rule == NULL) {
292 err = -ENOMEM;
293 goto errout;
295 rule->fr_net = hold_net(net);
297 if (tb[FRA_PRIORITY])
298 rule->pref = nla_get_u32(tb[FRA_PRIORITY]);
300 if (tb[FRA_IIFNAME]) {
301 struct net_device *dev;
303 rule->iifindex = -1;
304 nla_strlcpy(rule->iifname, tb[FRA_IIFNAME], IFNAMSIZ);
305 dev = __dev_get_by_name(net, rule->iifname);
306 if (dev)
307 rule->iifindex = dev->ifindex;
310 if (tb[FRA_OIFNAME]) {
311 struct net_device *dev;
313 rule->oifindex = -1;
314 nla_strlcpy(rule->oifname, tb[FRA_OIFNAME], IFNAMSIZ);
315 dev = __dev_get_by_name(net, rule->oifname);
316 if (dev)
317 rule->oifindex = dev->ifindex;
320 if (tb[FRA_FWMARK]) {
321 rule->mark = nla_get_u32(tb[FRA_FWMARK]);
322 if (rule->mark)
323 /* compatibility: if the mark value is non-zero all bits
324 * are compared unless a mask is explicitly specified.
326 rule->mark_mask = 0xFFFFFFFF;
329 if (tb[FRA_FWMASK])
330 rule->mark_mask = nla_get_u32(tb[FRA_FWMASK]);
332 rule->action = frh->action;
333 rule->flags = frh->flags;
334 rule->table = frh_get_table(frh, tb);
336 if (!tb[FRA_PRIORITY] && ops->default_pref)
337 rule->pref = ops->default_pref(ops);
339 err = -EINVAL;
340 if (tb[FRA_GOTO]) {
341 if (rule->action != FR_ACT_GOTO)
342 goto errout_free;
344 rule->target = nla_get_u32(tb[FRA_GOTO]);
345 /* Backward jumps are prohibited to avoid endless loops */
346 if (rule->target <= rule->pref)
347 goto errout_free;
349 list_for_each_entry(r, &ops->rules_list, list) {
350 if (r->pref == rule->target) {
351 rule->ctarget = r;
352 break;
356 if (rule->ctarget == NULL)
357 unresolved = 1;
358 } else if (rule->action == FR_ACT_GOTO)
359 goto errout_free;
361 err = ops->configure(rule, skb, frh, tb);
362 if (err < 0)
363 goto errout_free;
365 list_for_each_entry(r, &ops->rules_list, list) {
366 if (r->pref > rule->pref)
367 break;
368 last = r;
371 fib_rule_get(rule);
373 if (ops->unresolved_rules) {
375 * There are unresolved goto rules in the list, check if
376 * any of them are pointing to this new rule.
378 list_for_each_entry(r, &ops->rules_list, list) {
379 if (r->action == FR_ACT_GOTO &&
380 r->target == rule->pref) {
381 BUG_ON(r->ctarget != NULL);
382 rcu_assign_pointer(r->ctarget, rule);
383 if (--ops->unresolved_rules == 0)
384 break;
389 if (rule->action == FR_ACT_GOTO)
390 ops->nr_goto_rules++;
392 if (unresolved)
393 ops->unresolved_rules++;
395 if (last)
396 list_add_rcu(&rule->list, &last->list);
397 else
398 list_add_rcu(&rule->list, &ops->rules_list);
400 notify_rule_change(RTM_NEWRULE, rule, ops, nlh, NETLINK_CB(skb).pid);
401 flush_route_cache(ops);
402 rules_ops_put(ops);
403 return 0;
405 errout_free:
406 release_net(rule->fr_net);
407 kfree(rule);
408 errout:
409 rules_ops_put(ops);
410 return err;
413 static int fib_nl_delrule(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
415 struct net *net = sock_net(skb->sk);
416 struct fib_rule_hdr *frh = nlmsg_data(nlh);
417 struct fib_rules_ops *ops = NULL;
418 struct fib_rule *rule, *tmp;
419 struct nlattr *tb[FRA_MAX+1];
420 int err = -EINVAL;
422 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh)))
423 goto errout;
425 ops = lookup_rules_ops(net, frh->family);
426 if (ops == NULL) {
427 err = -EAFNOSUPPORT;
428 goto errout;
431 err = nlmsg_parse(nlh, sizeof(*frh), tb, FRA_MAX, ops->policy);
432 if (err < 0)
433 goto errout;
435 err = validate_rulemsg(frh, tb, ops);
436 if (err < 0)
437 goto errout;
439 list_for_each_entry(rule, &ops->rules_list, list) {
440 if (frh->action && (frh->action != rule->action))
441 continue;
443 if (frh->table && (frh_get_table(frh, tb) != rule->table))
444 continue;
446 if (tb[FRA_PRIORITY] &&
447 (rule->pref != nla_get_u32(tb[FRA_PRIORITY])))
448 continue;
450 if (tb[FRA_IIFNAME] &&
451 nla_strcmp(tb[FRA_IIFNAME], rule->iifname))
452 continue;
454 if (tb[FRA_OIFNAME] &&
455 nla_strcmp(tb[FRA_OIFNAME], rule->oifname))
456 continue;
458 if (tb[FRA_FWMARK] &&
459 (rule->mark != nla_get_u32(tb[FRA_FWMARK])))
460 continue;
462 if (tb[FRA_FWMASK] &&
463 (rule->mark_mask != nla_get_u32(tb[FRA_FWMASK])))
464 continue;
466 if (!ops->compare(rule, frh, tb))
467 continue;
469 if (rule->flags & FIB_RULE_PERMANENT) {
470 err = -EPERM;
471 goto errout;
474 list_del_rcu(&rule->list);
476 if (rule->action == FR_ACT_GOTO)
477 ops->nr_goto_rules--;
480 * Check if this rule is a target to any of them. If so,
481 * disable them. As this operation is eventually very
482 * expensive, it is only performed if goto rules have
483 * actually been added.
485 if (ops->nr_goto_rules > 0) {
486 list_for_each_entry(tmp, &ops->rules_list, list) {
487 if (tmp->ctarget == rule) {
488 rcu_assign_pointer(tmp->ctarget, NULL);
489 ops->unresolved_rules++;
494 synchronize_rcu();
495 notify_rule_change(RTM_DELRULE, rule, ops, nlh,
496 NETLINK_CB(skb).pid);
497 fib_rule_put(rule);
498 flush_route_cache(ops);
499 rules_ops_put(ops);
500 return 0;
503 err = -ENOENT;
504 errout:
505 rules_ops_put(ops);
506 return err;
509 static inline size_t fib_rule_nlmsg_size(struct fib_rules_ops *ops,
510 struct fib_rule *rule)
512 size_t payload = NLMSG_ALIGN(sizeof(struct fib_rule_hdr))
513 + nla_total_size(IFNAMSIZ) /* FRA_IIFNAME */
514 + nla_total_size(IFNAMSIZ) /* FRA_OIFNAME */
515 + nla_total_size(4) /* FRA_PRIORITY */
516 + nla_total_size(4) /* FRA_TABLE */
517 + nla_total_size(4) /* FRA_FWMARK */
518 + nla_total_size(4); /* FRA_FWMASK */
520 if (ops->nlmsg_payload)
521 payload += ops->nlmsg_payload(rule);
523 return payload;
526 static int fib_nl_fill_rule(struct sk_buff *skb, struct fib_rule *rule,
527 u32 pid, u32 seq, int type, int flags,
528 struct fib_rules_ops *ops)
530 struct nlmsghdr *nlh;
531 struct fib_rule_hdr *frh;
533 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*frh), flags);
534 if (nlh == NULL)
535 return -EMSGSIZE;
537 frh = nlmsg_data(nlh);
538 frh->family = ops->family;
539 frh->table = rule->table;
540 NLA_PUT_U32(skb, FRA_TABLE, rule->table);
541 frh->res1 = 0;
542 frh->res2 = 0;
543 frh->action = rule->action;
544 frh->flags = rule->flags;
546 if (rule->action == FR_ACT_GOTO && rule->ctarget == NULL)
547 frh->flags |= FIB_RULE_UNRESOLVED;
549 if (rule->iifname[0]) {
550 NLA_PUT_STRING(skb, FRA_IIFNAME, rule->iifname);
552 if (rule->iifindex == -1)
553 frh->flags |= FIB_RULE_IIF_DETACHED;
556 if (rule->oifname[0]) {
557 NLA_PUT_STRING(skb, FRA_OIFNAME, rule->oifname);
559 if (rule->oifindex == -1)
560 frh->flags |= FIB_RULE_OIF_DETACHED;
563 if (rule->pref)
564 NLA_PUT_U32(skb, FRA_PRIORITY, rule->pref);
566 if (rule->mark)
567 NLA_PUT_U32(skb, FRA_FWMARK, rule->mark);
569 if (rule->mark_mask || rule->mark)
570 NLA_PUT_U32(skb, FRA_FWMASK, rule->mark_mask);
572 if (rule->target)
573 NLA_PUT_U32(skb, FRA_GOTO, rule->target);
575 if (ops->fill(rule, skb, frh) < 0)
576 goto nla_put_failure;
578 return nlmsg_end(skb, nlh);
580 nla_put_failure:
581 nlmsg_cancel(skb, nlh);
582 return -EMSGSIZE;
585 static int dump_rules(struct sk_buff *skb, struct netlink_callback *cb,
586 struct fib_rules_ops *ops)
588 int idx = 0;
589 struct fib_rule *rule;
591 list_for_each_entry(rule, &ops->rules_list, list) {
592 if (idx < cb->args[1])
593 goto skip;
595 if (fib_nl_fill_rule(skb, rule, NETLINK_CB(cb->skb).pid,
596 cb->nlh->nlmsg_seq, RTM_NEWRULE,
597 NLM_F_MULTI, ops) < 0)
598 break;
599 skip:
600 idx++;
602 cb->args[1] = idx;
603 rules_ops_put(ops);
605 return skb->len;
608 static int fib_nl_dumprule(struct sk_buff *skb, struct netlink_callback *cb)
610 struct net *net = sock_net(skb->sk);
611 struct fib_rules_ops *ops;
612 int idx = 0, family;
614 family = rtnl_msg_family(cb->nlh);
615 if (family != AF_UNSPEC) {
616 /* Protocol specific dump request */
617 ops = lookup_rules_ops(net, family);
618 if (ops == NULL)
619 return -EAFNOSUPPORT;
621 return dump_rules(skb, cb, ops);
624 rcu_read_lock();
625 list_for_each_entry_rcu(ops, &net->rules_ops, list) {
626 if (idx < cb->args[0] || !try_module_get(ops->owner))
627 goto skip;
629 if (dump_rules(skb, cb, ops) < 0)
630 break;
632 cb->args[1] = 0;
633 skip:
634 idx++;
636 rcu_read_unlock();
637 cb->args[0] = idx;
639 return skb->len;
642 static void notify_rule_change(int event, struct fib_rule *rule,
643 struct fib_rules_ops *ops, struct nlmsghdr *nlh,
644 u32 pid)
646 struct net *net;
647 struct sk_buff *skb;
648 int err = -ENOBUFS;
650 net = ops->fro_net;
651 skb = nlmsg_new(fib_rule_nlmsg_size(ops, rule), GFP_KERNEL);
652 if (skb == NULL)
653 goto errout;
655 err = fib_nl_fill_rule(skb, rule, pid, nlh->nlmsg_seq, event, 0, ops);
656 if (err < 0) {
657 /* -EMSGSIZE implies BUG in fib_rule_nlmsg_size() */
658 WARN_ON(err == -EMSGSIZE);
659 kfree_skb(skb);
660 goto errout;
663 rtnl_notify(skb, net, pid, ops->nlgroup, nlh, GFP_KERNEL);
664 return;
665 errout:
666 if (err < 0)
667 rtnl_set_sk_err(net, ops->nlgroup, err);
670 static void attach_rules(struct list_head *rules, struct net_device *dev)
672 struct fib_rule *rule;
674 list_for_each_entry(rule, rules, list) {
675 if (rule->iifindex == -1 &&
676 strcmp(dev->name, rule->iifname) == 0)
677 rule->iifindex = dev->ifindex;
678 if (rule->oifindex == -1 &&
679 strcmp(dev->name, rule->oifname) == 0)
680 rule->oifindex = dev->ifindex;
684 static void detach_rules(struct list_head *rules, struct net_device *dev)
686 struct fib_rule *rule;
688 list_for_each_entry(rule, rules, list) {
689 if (rule->iifindex == dev->ifindex)
690 rule->iifindex = -1;
691 if (rule->oifindex == dev->ifindex)
692 rule->oifindex = -1;
697 static int fib_rules_event(struct notifier_block *this, unsigned long event,
698 void *ptr)
700 struct net_device *dev = ptr;
701 struct net *net = dev_net(dev);
702 struct fib_rules_ops *ops;
704 ASSERT_RTNL();
706 switch (event) {
707 case NETDEV_REGISTER:
708 list_for_each_entry(ops, &net->rules_ops, list)
709 attach_rules(&ops->rules_list, dev);
710 break;
712 case NETDEV_UNREGISTER:
713 list_for_each_entry(ops, &net->rules_ops, list)
714 detach_rules(&ops->rules_list, dev);
715 break;
718 return NOTIFY_DONE;
721 static struct notifier_block fib_rules_notifier = {
722 .notifier_call = fib_rules_event,
725 static int __net_init fib_rules_net_init(struct net *net)
727 INIT_LIST_HEAD(&net->rules_ops);
728 spin_lock_init(&net->rules_mod_lock);
729 return 0;
732 static struct pernet_operations fib_rules_net_ops = {
733 .init = fib_rules_net_init,
736 static int __init fib_rules_init(void)
738 int err;
739 rtnl_register(PF_UNSPEC, RTM_NEWRULE, fib_nl_newrule, NULL);
740 rtnl_register(PF_UNSPEC, RTM_DELRULE, fib_nl_delrule, NULL);
741 rtnl_register(PF_UNSPEC, RTM_GETRULE, NULL, fib_nl_dumprule);
743 err = register_pernet_subsys(&fib_rules_net_ops);
744 if (err < 0)
745 goto fail;
747 err = register_netdevice_notifier(&fib_rules_notifier);
748 if (err < 0)
749 goto fail_unregister;
751 return 0;
753 fail_unregister:
754 unregister_pernet_subsys(&fib_rules_net_ops);
755 fail:
756 rtnl_unregister(PF_UNSPEC, RTM_NEWRULE);
757 rtnl_unregister(PF_UNSPEC, RTM_DELRULE);
758 rtnl_unregister(PF_UNSPEC, RTM_GETRULE);
759 return err;
762 subsys_initcall(fib_rules_init);