1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16#include <linux/module.h>
17#include <linux/uaccess.h>
18#include <linux/bitops.h>
19#include <linux/capability.h>
20#include <linux/types.h>
21#include <linux/kernel.h>
22#include <linux/mm.h>
23#include <linux/string.h>
24#include <linux/socket.h>
25#include <linux/sockios.h>
26#include <linux/errno.h>
27#include <linux/in.h>
28#include <linux/inet.h>
29#include <linux/inetdevice.h>
30#include <linux/netdevice.h>
31#include <linux/if_addr.h>
32#include <linux/if_arp.h>
33#include <linux/skbuff.h>
34#include <linux/cache.h>
35#include <linux/init.h>
36#include <linux/list.h>
37#include <linux/slab.h>
38
39#include <net/ip.h>
40#include <net/protocol.h>
41#include <net/route.h>
42#include <net/tcp.h>
43#include <net/sock.h>
44#include <net/arp.h>
45#include <net/ip_fib.h>
46#include <net/rtnetlink.h>
47#include <net/xfrm.h>
48#include <net/l3mdev.h>
49#include <net/lwtunnel.h>
50#include <trace/events/fib.h>
51
52#ifndef CONFIG_IP_MULTIPLE_TABLES
53
54static int __net_init fib4_rules_init(struct net *net)
55{
56 struct fib_table *local_table, *main_table;
57
58 main_table = fib_trie_table(RT_TABLE_MAIN, NULL);
59 if (!main_table)
60 return -ENOMEM;
61
62 local_table = fib_trie_table(RT_TABLE_LOCAL, main_table);
63 if (!local_table)
64 goto fail;
65
66 hlist_add_head_rcu(&local_table->tb_hlist,
67 &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]);
68 hlist_add_head_rcu(&main_table->tb_hlist,
69 &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]);
70 return 0;
71
72fail:
73 fib_free_table(main_table);
74 return -ENOMEM;
75}
76
77static bool fib4_has_custom_rules(struct net *net)
78{
79 return false;
80}
81#else
82
83struct fib_table *fib_new_table(struct net *net, u32 id)
84{
85 struct fib_table *tb, *alias = NULL;
86 unsigned int h;
87
88 if (id == 0)
89 id = RT_TABLE_MAIN;
90 tb = fib_get_table(net, id);
91 if (tb)
92 return tb;
93
94 if (id == RT_TABLE_LOCAL && !net->ipv4.fib_has_custom_rules)
95 alias = fib_new_table(net, RT_TABLE_MAIN);
96
97 tb = fib_trie_table(id, alias);
98 if (!tb)
99 return NULL;
100
101 switch (id) {
102 case RT_TABLE_MAIN:
103 rcu_assign_pointer(net->ipv4.fib_main, tb);
104 break;
105 case RT_TABLE_DEFAULT:
106 rcu_assign_pointer(net->ipv4.fib_default, tb);
107 break;
108 default:
109 break;
110 }
111
112 h = id & (FIB_TABLE_HASHSZ - 1);
113 hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]);
114 return tb;
115}
116EXPORT_SYMBOL_GPL(fib_new_table);
117
118
119struct fib_table *fib_get_table(struct net *net, u32 id)
120{
121 struct fib_table *tb;
122 struct hlist_head *head;
123 unsigned int h;
124
125 if (id == 0)
126 id = RT_TABLE_MAIN;
127 h = id & (FIB_TABLE_HASHSZ - 1);
128
129 head = &net->ipv4.fib_table_hash[h];
130 hlist_for_each_entry_rcu(tb, head, tb_hlist,
131 lockdep_rtnl_is_held()) {
132 if (tb->tb_id == id)
133 return tb;
134 }
135 return NULL;
136}
137
138static bool fib4_has_custom_rules(struct net *net)
139{
140 return net->ipv4.fib_has_custom_rules;
141}
142#endif
143
144static void fib_replace_table(struct net *net, struct fib_table *old,
145 struct fib_table *new)
146{
147#ifdef CONFIG_IP_MULTIPLE_TABLES
148 switch (new->tb_id) {
149 case RT_TABLE_MAIN:
150 rcu_assign_pointer(net->ipv4.fib_main, new);
151 break;
152 case RT_TABLE_DEFAULT:
153 rcu_assign_pointer(net->ipv4.fib_default, new);
154 break;
155 default:
156 break;
157 }
158
159#endif
160
161 hlist_replace_rcu(&old->tb_hlist, &new->tb_hlist);
162}
163
164int fib_unmerge(struct net *net)
165{
166 struct fib_table *old, *new, *main_table;
167
168
169 old = fib_get_table(net, RT_TABLE_LOCAL);
170 if (!old)
171 return 0;
172
173 new = fib_trie_unmerge(old);
174 if (!new)
175 return -ENOMEM;
176
177
178 if (new == old)
179 return 0;
180
181
182 fib_replace_table(net, old, new);
183 fib_free_table(old);
184
185
186 main_table = fib_get_table(net, RT_TABLE_MAIN);
187 if (!main_table)
188 return 0;
189
190
191 fib_table_flush_external(main_table);
192
193 return 0;
194}
195
196static void fib_flush(struct net *net)
197{
198 int flushed = 0;
199 unsigned int h;
200
201 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
202 struct hlist_head *head = &net->ipv4.fib_table_hash[h];
203 struct hlist_node *tmp;
204 struct fib_table *tb;
205
206 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist)
207 flushed += fib_table_flush(net, tb, false);
208 }
209
210 if (flushed)
211 rt_cache_flush(net);
212}
213
214
215
216
217
218static inline unsigned int __inet_dev_addr_type(struct net *net,
219 const struct net_device *dev,
220 __be32 addr, u32 tb_id)
221{
222 struct flowi4 fl4 = { .daddr = addr };
223 struct fib_result res;
224 unsigned int ret = RTN_BROADCAST;
225 struct fib_table *table;
226
227 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
228 return RTN_BROADCAST;
229 if (ipv4_is_multicast(addr))
230 return RTN_MULTICAST;
231
232 rcu_read_lock();
233
234 table = fib_get_table(net, tb_id);
235 if (table) {
236 ret = RTN_UNICAST;
237 if (!fib_table_lookup(table, &fl4, &res, FIB_LOOKUP_NOREF)) {
238 if (!dev || dev == res.fi->fib_dev)
239 ret = res.type;
240 }
241 }
242
243 rcu_read_unlock();
244 return ret;
245}
246
247unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id)
248{
249 return __inet_dev_addr_type(net, NULL, addr, tb_id);
250}
251EXPORT_SYMBOL(inet_addr_type_table);
252
253unsigned int inet_addr_type(struct net *net, __be32 addr)
254{
255 return __inet_dev_addr_type(net, NULL, addr, RT_TABLE_LOCAL);
256}
257EXPORT_SYMBOL(inet_addr_type);
258
259unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
260 __be32 addr)
261{
262 u32 rt_table = l3mdev_fib_table(dev) ? : RT_TABLE_LOCAL;
263
264 return __inet_dev_addr_type(net, dev, addr, rt_table);
265}
266EXPORT_SYMBOL(inet_dev_addr_type);
267
268
269
270
271unsigned int inet_addr_type_dev_table(struct net *net,
272 const struct net_device *dev,
273 __be32 addr)
274{
275 u32 rt_table = l3mdev_fib_table(dev) ? : RT_TABLE_LOCAL;
276
277 return __inet_dev_addr_type(net, NULL, addr, rt_table);
278}
279EXPORT_SYMBOL(inet_addr_type_dev_table);
280
281__be32 fib_compute_spec_dst(struct sk_buff *skb)
282{
283 struct net_device *dev = skb->dev;
284 struct in_device *in_dev;
285 struct fib_result res;
286 struct rtable *rt;
287 struct net *net;
288 int scope;
289
290 rt = skb_rtable(skb);
291 if ((rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | RTCF_LOCAL)) ==
292 RTCF_LOCAL)
293 return ip_hdr(skb)->daddr;
294
295 in_dev = __in_dev_get_rcu(dev);
296
297 net = dev_net(dev);
298
299 scope = RT_SCOPE_UNIVERSE;
300 if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) {
301 bool vmark = in_dev && IN_DEV_SRC_VMARK(in_dev);
302 struct flowi4 fl4 = {
303 .flowi4_iif = LOOPBACK_IFINDEX,
304 .flowi4_oif = l3mdev_master_ifindex_rcu(dev),
305 .daddr = ip_hdr(skb)->saddr,
306 .flowi4_tos = ip_hdr(skb)->tos & IPTOS_RT_MASK,
307 .flowi4_scope = scope,
308 .flowi4_mark = vmark ? skb->mark : 0,
309 };
310 if (!fib_lookup(net, &fl4, &res, 0))
311 return FIB_RES_PREFSRC(net, res);
312 } else {
313 scope = RT_SCOPE_LINK;
314 }
315
316 return inet_select_addr(dev, ip_hdr(skb)->saddr, scope);
317}
318
319bool fib_info_nh_uses_dev(struct fib_info *fi, const struct net_device *dev)
320{
321 bool dev_match = false;
322#ifdef CONFIG_IP_ROUTE_MULTIPATH
323 int ret;
324
325 for (ret = 0; ret < fi->fib_nhs; ret++) {
326 struct fib_nh *nh = &fi->fib_nh[ret];
327
328 if (nh->nh_dev == dev) {
329 dev_match = true;
330 break;
331 } else if (l3mdev_master_ifindex_rcu(nh->nh_dev) == dev->ifindex) {
332 dev_match = true;
333 break;
334 }
335 }
336#else
337 if (fi->fib_nh[0].nh_dev == dev)
338 dev_match = true;
339#endif
340
341 return dev_match;
342}
343EXPORT_SYMBOL_GPL(fib_info_nh_uses_dev);
344
345
346
347
348
349
350
351
352
353static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
354 u8 tos, int oif, struct net_device *dev,
355 int rpf, struct in_device *idev, u32 *itag)
356{
357 struct net *net = dev_net(dev);
358 struct flow_keys flkeys;
359 int ret, no_addr;
360 struct fib_result res;
361 struct flowi4 fl4;
362 bool dev_match;
363
364 fl4.flowi4_oif = 0;
365 fl4.flowi4_iif = l3mdev_master_ifindex_rcu(dev);
366 if (!fl4.flowi4_iif)
367 fl4.flowi4_iif = oif ? : LOOPBACK_IFINDEX;
368 fl4.daddr = src;
369 fl4.saddr = dst;
370 fl4.flowi4_tos = tos;
371 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
372 fl4.flowi4_tun_key.tun_id = 0;
373 fl4.flowi4_flags = 0;
374 fl4.flowi4_uid = sock_net_uid(net, NULL);
375
376 no_addr = idev->ifa_list == NULL;
377
378 fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0;
379 if (!fib4_rules_early_flow_dissect(net, skb, &fl4, &flkeys)) {
380 fl4.flowi4_proto = 0;
381 fl4.fl4_sport = 0;
382 fl4.fl4_dport = 0;
383 } else {
384 swap(fl4.fl4_sport, fl4.fl4_dport);
385 }
386
387 if (fib_lookup(net, &fl4, &res, 0))
388 goto last_resort;
389 if (res.type != RTN_UNICAST &&
390 (res.type != RTN_LOCAL || !IN_DEV_ACCEPT_LOCAL(idev)))
391 goto e_inval;
392 fib_combine_itag(itag, &res);
393
394 dev_match = fib_info_nh_uses_dev(res.fi, dev);
395 if (dev_match) {
396 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
397 return ret;
398 }
399 if (no_addr)
400 goto last_resort;
401 if (rpf == 1)
402 goto e_rpf;
403 fl4.flowi4_oif = dev->ifindex;
404
405 ret = 0;
406 if (fib_lookup(net, &fl4, &res, FIB_LOOKUP_IGNORE_LINKSTATE) == 0) {
407 if (res.type == RTN_UNICAST)
408 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
409 }
410 return ret;
411
412last_resort:
413 if (rpf)
414 goto e_rpf;
415 *itag = 0;
416 return 0;
417
418e_inval:
419 return -EINVAL;
420e_rpf:
421 return -EXDEV;
422}
423
424
425int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
426 u8 tos, int oif, struct net_device *dev,
427 struct in_device *idev, u32 *itag)
428{
429 int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev);
430 struct net *net = dev_net(dev);
431
432 if (!r && !fib_num_tclassid_users(net) &&
433 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) {
434 if (IN_DEV_ACCEPT_LOCAL(idev))
435 goto ok;
436
437
438
439
440 if (net->ipv4.fib_has_custom_local_routes ||
441 fib4_has_custom_rules(net))
442 goto full_check;
443 if (inet_lookup_ifaddr_rcu(net, src))
444 return -EINVAL;
445
446ok:
447 *itag = 0;
448 return 0;
449 }
450
451full_check:
452 return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag);
453}
454
455static inline __be32 sk_extract_addr(struct sockaddr *addr)
456{
457 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
458}
459
460static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
461{
462 struct nlattr *nla;
463
464 nla = (struct nlattr *) ((char *) mx + len);
465 nla->nla_type = type;
466 nla->nla_len = nla_attr_size(4);
467 *(u32 *) nla_data(nla) = value;
468
469 return len + nla_total_size(4);
470}
471
472static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
473 struct fib_config *cfg)
474{
475 __be32 addr;
476 int plen;
477
478 memset(cfg, 0, sizeof(*cfg));
479 cfg->fc_nlinfo.nl_net = net;
480
481 if (rt->rt_dst.sa_family != AF_INET)
482 return -EAFNOSUPPORT;
483
484
485
486
487
488
489
490
491
492 plen = 32;
493 addr = sk_extract_addr(&rt->rt_dst);
494 if (!(rt->rt_flags & RTF_HOST)) {
495 __be32 mask = sk_extract_addr(&rt->rt_genmask);
496
497 if (rt->rt_genmask.sa_family != AF_INET) {
498 if (mask || rt->rt_genmask.sa_family)
499 return -EAFNOSUPPORT;
500 }
501
502 if (bad_mask(mask, addr))
503 return -EINVAL;
504
505 plen = inet_mask_len(mask);
506 }
507
508 cfg->fc_dst_len = plen;
509 cfg->fc_dst = addr;
510
511 if (cmd != SIOCDELRT) {
512 cfg->fc_nlflags = NLM_F_CREATE;
513 cfg->fc_protocol = RTPROT_BOOT;
514 }
515
516 if (rt->rt_metric)
517 cfg->fc_priority = rt->rt_metric - 1;
518
519 if (rt->rt_flags & RTF_REJECT) {
520 cfg->fc_scope = RT_SCOPE_HOST;
521 cfg->fc_type = RTN_UNREACHABLE;
522 return 0;
523 }
524
525 cfg->fc_scope = RT_SCOPE_NOWHERE;
526 cfg->fc_type = RTN_UNICAST;
527
528 if (rt->rt_dev) {
529 char *colon;
530 struct net_device *dev;
531 char devname[IFNAMSIZ];
532
533 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
534 return -EFAULT;
535
536 devname[IFNAMSIZ-1] = 0;
537 colon = strchr(devname, ':');
538 if (colon)
539 *colon = 0;
540 dev = __dev_get_by_name(net, devname);
541 if (!dev)
542 return -ENODEV;
543 cfg->fc_oif = dev->ifindex;
544 cfg->fc_table = l3mdev_fib_table(dev);
545 if (colon) {
546 const struct in_ifaddr *ifa;
547 struct in_device *in_dev;
548
549 in_dev = __in_dev_get_rtnl(dev);
550 if (!in_dev)
551 return -ENODEV;
552
553 *colon = ':';
554
555 rcu_read_lock();
556 in_dev_for_each_ifa_rcu(ifa, in_dev) {
557 if (strcmp(ifa->ifa_label, devname) == 0)
558 break;
559 }
560 rcu_read_unlock();
561
562 if (!ifa)
563 return -ENODEV;
564 cfg->fc_prefsrc = ifa->ifa_local;
565 }
566 }
567
568 addr = sk_extract_addr(&rt->rt_gateway);
569 if (rt->rt_gateway.sa_family == AF_INET && addr) {
570 unsigned int addr_type;
571
572 cfg->fc_gw = addr;
573 addr_type = inet_addr_type_table(net, addr, cfg->fc_table);
574 if (rt->rt_flags & RTF_GATEWAY &&
575 addr_type == RTN_UNICAST)
576 cfg->fc_scope = RT_SCOPE_UNIVERSE;
577 }
578
579 if (cmd == SIOCDELRT)
580 return 0;
581
582 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
583 return -EINVAL;
584
585 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
586 cfg->fc_scope = RT_SCOPE_LINK;
587
588 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
589 struct nlattr *mx;
590 int len = 0;
591
592 mx = kcalloc(3, nla_total_size(4), GFP_KERNEL);
593 if (!mx)
594 return -ENOMEM;
595
596 if (rt->rt_flags & RTF_MTU)
597 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
598
599 if (rt->rt_flags & RTF_WINDOW)
600 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
601
602 if (rt->rt_flags & RTF_IRTT)
603 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
604
605 cfg->fc_mx = mx;
606 cfg->fc_mx_len = len;
607 }
608
609 return 0;
610}
611
612
613
614
615
616int ip_rt_ioctl(struct net *net, unsigned int cmd, struct rtentry *rt)
617{
618 struct fib_config cfg;
619 int err;
620
621 switch (cmd) {
622 case SIOCADDRT:
623 case SIOCDELRT:
624 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
625 return -EPERM;
626
627 rtnl_lock();
628 err = rtentry_to_fib_config(net, cmd, rt, &cfg);
629 if (err == 0) {
630 struct fib_table *tb;
631
632 if (cmd == SIOCDELRT) {
633 tb = fib_get_table(net, cfg.fc_table);
634 if (tb)
635 err = fib_table_delete(net, tb, &cfg,
636 NULL);
637 else
638 err = -ESRCH;
639 } else {
640 tb = fib_new_table(net, cfg.fc_table);
641 if (tb)
642 err = fib_table_insert(net, tb,
643 &cfg, NULL);
644 else
645 err = -ENOBUFS;
646 }
647
648
649 kfree(cfg.fc_mx);
650 }
651 rtnl_unlock();
652 return err;
653 }
654 return -EINVAL;
655}
656
657const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
658 [RTA_DST] = { .type = NLA_U32 },
659 [RTA_SRC] = { .type = NLA_U32 },
660 [RTA_IIF] = { .type = NLA_U32 },
661 [RTA_OIF] = { .type = NLA_U32 },
662 [RTA_GATEWAY] = { .type = NLA_U32 },
663 [RTA_PRIORITY] = { .type = NLA_U32 },
664 [RTA_PREFSRC] = { .type = NLA_U32 },
665 [RTA_METRICS] = { .type = NLA_NESTED },
666 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
667 [RTA_FLOW] = { .type = NLA_U32 },
668 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
669 [RTA_ENCAP] = { .type = NLA_NESTED },
670 [RTA_UID] = { .type = NLA_U32 },
671 [RTA_MARK] = { .type = NLA_U32 },
672 [RTA_TABLE] = { .type = NLA_U32 },
673 [RTA_IP_PROTO] = { .type = NLA_U8 },
674 [RTA_SPORT] = { .type = NLA_U16 },
675 [RTA_DPORT] = { .type = NLA_U16 },
676};
677
678static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
679 struct nlmsghdr *nlh, struct fib_config *cfg,
680 struct netlink_ext_ack *extack)
681{
682 struct nlattr *attr;
683 int err, remaining;
684 struct rtmsg *rtm;
685
686 err = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
687 rtm_ipv4_policy, extack);
688 if (err < 0)
689 goto errout;
690
691 memset(cfg, 0, sizeof(*cfg));
692
693 rtm = nlmsg_data(nlh);
694 cfg->fc_dst_len = rtm->rtm_dst_len;
695 cfg->fc_tos = rtm->rtm_tos;
696 cfg->fc_table = rtm->rtm_table;
697 cfg->fc_protocol = rtm->rtm_protocol;
698 cfg->fc_scope = rtm->rtm_scope;
699 cfg->fc_type = rtm->rtm_type;
700 cfg->fc_flags = rtm->rtm_flags;
701 cfg->fc_nlflags = nlh->nlmsg_flags;
702
703 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
704 cfg->fc_nlinfo.nlh = nlh;
705 cfg->fc_nlinfo.nl_net = net;
706
707 if (cfg->fc_type > RTN_MAX) {
708 NL_SET_ERR_MSG(extack, "Invalid route type");
709 err = -EINVAL;
710 goto errout;
711 }
712
713 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
714 switch (nla_type(attr)) {
715 case RTA_DST:
716 cfg->fc_dst = nla_get_be32(attr);
717 break;
718 case RTA_OIF:
719 cfg->fc_oif = nla_get_u32(attr);
720 break;
721 case RTA_GATEWAY:
722 cfg->fc_gw = nla_get_be32(attr);
723 break;
724 case RTA_VIA:
725 NL_SET_ERR_MSG(extack, "IPv4 does not support RTA_VIA attribute");
726 err = -EINVAL;
727 goto errout;
728 case RTA_PRIORITY:
729 cfg->fc_priority = nla_get_u32(attr);
730 break;
731 case RTA_PREFSRC:
732 cfg->fc_prefsrc = nla_get_be32(attr);
733 break;
734 case RTA_METRICS:
735 cfg->fc_mx = nla_data(attr);
736 cfg->fc_mx_len = nla_len(attr);
737 break;
738 case RTA_MULTIPATH:
739 err = lwtunnel_valid_encap_type_attr(nla_data(attr),
740 nla_len(attr),
741 extack);
742 if (err < 0)
743 goto errout;
744 cfg->fc_mp = nla_data(attr);
745 cfg->fc_mp_len = nla_len(attr);
746 break;
747 case RTA_FLOW:
748 cfg->fc_flow = nla_get_u32(attr);
749 break;
750 case RTA_TABLE:
751 cfg->fc_table = nla_get_u32(attr);
752 break;
753 case RTA_ENCAP:
754 cfg->fc_encap = attr;
755 break;
756 case RTA_ENCAP_TYPE:
757 cfg->fc_encap_type = nla_get_u16(attr);
758 err = lwtunnel_valid_encap_type(cfg->fc_encap_type,
759 extack);
760 if (err < 0)
761 goto errout;
762 break;
763 }
764 }
765
766 return 0;
767errout:
768 return err;
769}
770
771static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
772 struct netlink_ext_ack *extack)
773{
774 struct net *net = sock_net(skb->sk);
775 struct fib_config cfg;
776 struct fib_table *tb;
777 int err;
778
779 err = rtm_to_fib_config(net, skb, nlh, &cfg, extack);
780 if (err < 0)
781 goto errout;
782
783 tb = fib_get_table(net, cfg.fc_table);
784 if (!tb) {
785 NL_SET_ERR_MSG(extack, "FIB table does not exist");
786 err = -ESRCH;
787 goto errout;
788 }
789
790 err = fib_table_delete(net, tb, &cfg, extack);
791errout:
792 return err;
793}
794
795static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
796 struct netlink_ext_ack *extack)
797{
798 struct net *net = sock_net(skb->sk);
799 struct fib_config cfg;
800 struct fib_table *tb;
801 int err;
802
803 err = rtm_to_fib_config(net, skb, nlh, &cfg, extack);
804 if (err < 0)
805 goto errout;
806
807 tb = fib_new_table(net, cfg.fc_table);
808 if (!tb) {
809 err = -ENOBUFS;
810 goto errout;
811 }
812
813 err = fib_table_insert(net, tb, &cfg, extack);
814 if (!err && cfg.fc_type == RTN_LOCAL)
815 net->ipv4.fib_has_custom_local_routes = true;
816errout:
817 return err;
818}
819
820int ip_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
821 struct fib_dump_filter *filter,
822 struct netlink_callback *cb)
823{
824 struct netlink_ext_ack *extack = cb->extack;
825 struct nlattr *tb[RTA_MAX + 1];
826 struct rtmsg *rtm;
827 int err, i;
828
829 ASSERT_RTNL();
830
831 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
832 NL_SET_ERR_MSG(extack, "Invalid header for FIB dump request");
833 return -EINVAL;
834 }
835
836 rtm = nlmsg_data(nlh);
837 if (rtm->rtm_dst_len || rtm->rtm_src_len || rtm->rtm_tos ||
838 rtm->rtm_scope) {
839 NL_SET_ERR_MSG(extack, "Invalid values in header for FIB dump request");
840 return -EINVAL;
841 }
842
843 if (rtm->rtm_flags & ~(RTM_F_CLONED | RTM_F_PREFIX)) {
844 NL_SET_ERR_MSG(extack, "Invalid flags for FIB dump request");
845 return -EINVAL;
846 }
847 if (rtm->rtm_flags & RTM_F_CLONED)
848 filter->dump_routes = false;
849 else
850 filter->dump_exceptions = false;
851
852 filter->flags = rtm->rtm_flags;
853 filter->protocol = rtm->rtm_protocol;
854 filter->rt_type = rtm->rtm_type;
855 filter->table_id = rtm->rtm_table;
856
857 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
858 rtm_ipv4_policy, extack);
859 if (err < 0)
860 return err;
861
862 for (i = 0; i <= RTA_MAX; ++i) {
863 int ifindex;
864
865 if (!tb[i])
866 continue;
867
868 switch (i) {
869 case RTA_TABLE:
870 filter->table_id = nla_get_u32(tb[i]);
871 break;
872 case RTA_OIF:
873 ifindex = nla_get_u32(tb[i]);
874 filter->dev = __dev_get_by_index(net, ifindex);
875 if (!filter->dev)
876 return -ENODEV;
877 break;
878 default:
879 NL_SET_ERR_MSG(extack, "Unsupported attribute in dump request");
880 return -EINVAL;
881 }
882 }
883
884 if (filter->flags || filter->protocol || filter->rt_type ||
885 filter->table_id || filter->dev) {
886 filter->filter_set = 1;
887 cb->answer_flags = NLM_F_DUMP_FILTERED;
888 }
889
890 return 0;
891}
892EXPORT_SYMBOL_GPL(ip_valid_fib_dump_req);
893
894static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
895{
896 struct fib_dump_filter filter = { .dump_routes = true,
897 .dump_exceptions = true };
898 const struct nlmsghdr *nlh = cb->nlh;
899 struct net *net = sock_net(skb->sk);
900 unsigned int h, s_h;
901 unsigned int e = 0, s_e;
902 struct fib_table *tb;
903 struct hlist_head *head;
904 int dumped = 0, err;
905
906 if (cb->strict_check) {
907 err = ip_valid_fib_dump_req(net, nlh, &filter, cb);
908 if (err < 0)
909 return err;
910 } else if (nlmsg_len(nlh) >= sizeof(struct rtmsg)) {
911 struct rtmsg *rtm = nlmsg_data(nlh);
912
913 filter.flags = rtm->rtm_flags & (RTM_F_PREFIX | RTM_F_CLONED);
914 }
915
916
917 if (filter.flags & RTM_F_PREFIX)
918 return skb->len;
919
920 if (filter.table_id) {
921 tb = fib_get_table(net, filter.table_id);
922 if (!tb) {
923 if (rtnl_msg_family(cb->nlh) != PF_INET)
924 return skb->len;
925
926 NL_SET_ERR_MSG(cb->extack, "ipv4: FIB table does not exist");
927 return -ENOENT;
928 }
929
930 rcu_read_lock();
931 err = fib_table_dump(tb, skb, cb, &filter);
932 rcu_read_unlock();
933 return skb->len ? : err;
934 }
935
936 s_h = cb->args[0];
937 s_e = cb->args[1];
938
939 rcu_read_lock();
940
941 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
942 e = 0;
943 head = &net->ipv4.fib_table_hash[h];
944 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
945 if (e < s_e)
946 goto next;
947 if (dumped)
948 memset(&cb->args[2], 0, sizeof(cb->args) -
949 2 * sizeof(cb->args[0]));
950 err = fib_table_dump(tb, skb, cb, &filter);
951 if (err < 0) {
952 if (likely(skb->len))
953 goto out;
954
955 goto out_err;
956 }
957 dumped = 1;
958next:
959 e++;
960 }
961 }
962out:
963 err = skb->len;
964out_err:
965 rcu_read_unlock();
966
967 cb->args[1] = e;
968 cb->args[0] = h;
969
970 return err;
971}
972
973
974
975
976
977
978
979static void fib_magic(int cmd, int type, __be32 dst, int dst_len,
980 struct in_ifaddr *ifa, u32 rt_priority)
981{
982 struct net *net = dev_net(ifa->ifa_dev->dev);
983 u32 tb_id = l3mdev_fib_table(ifa->ifa_dev->dev);
984 struct fib_table *tb;
985 struct fib_config cfg = {
986 .fc_protocol = RTPROT_KERNEL,
987 .fc_type = type,
988 .fc_dst = dst,
989 .fc_dst_len = dst_len,
990 .fc_priority = rt_priority,
991 .fc_prefsrc = ifa->ifa_local,
992 .fc_oif = ifa->ifa_dev->dev->ifindex,
993 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
994 .fc_nlinfo = {
995 .nl_net = net,
996 },
997 };
998
999 if (!tb_id)
1000 tb_id = (type == RTN_UNICAST) ? RT_TABLE_MAIN : RT_TABLE_LOCAL;
1001
1002 tb = fib_new_table(net, tb_id);
1003 if (!tb)
1004 return;
1005
1006 cfg.fc_table = tb->tb_id;
1007
1008 if (type != RTN_LOCAL)
1009 cfg.fc_scope = RT_SCOPE_LINK;
1010 else
1011 cfg.fc_scope = RT_SCOPE_HOST;
1012
1013 if (cmd == RTM_NEWROUTE)
1014 fib_table_insert(net, tb, &cfg, NULL);
1015 else
1016 fib_table_delete(net, tb, &cfg, NULL);
1017}
1018
1019void fib_add_ifaddr(struct in_ifaddr *ifa)
1020{
1021 struct in_device *in_dev = ifa->ifa_dev;
1022 struct net_device *dev = in_dev->dev;
1023 struct in_ifaddr *prim = ifa;
1024 __be32 mask = ifa->ifa_mask;
1025 __be32 addr = ifa->ifa_local;
1026 __be32 prefix = ifa->ifa_address & mask;
1027
1028 if (ifa->ifa_flags & IFA_F_SECONDARY) {
1029 prim = inet_ifa_byprefix(in_dev, prefix, mask);
1030 if (!prim) {
1031 pr_warn("%s: bug: prim == NULL\n", __func__);
1032 return;
1033 }
1034 }
1035
1036 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim, 0);
1037
1038 if (!(dev->flags & IFF_UP))
1039 return;
1040
1041
1042 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
1043 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32,
1044 prim, 0);
1045
1046 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
1047 (prefix != addr || ifa->ifa_prefixlen < 32)) {
1048 if (!(ifa->ifa_flags & IFA_F_NOPREFIXROUTE))
1049 fib_magic(RTM_NEWROUTE,
1050 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
1051 prefix, ifa->ifa_prefixlen, prim,
1052 ifa->ifa_rt_priority);
1053
1054
1055 if (ifa->ifa_prefixlen < 31) {
1056 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32,
1057 prim, 0);
1058 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
1059 32, prim, 0);
1060 }
1061 }
1062}
1063
1064void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric)
1065{
1066 __be32 prefix = ifa->ifa_address & ifa->ifa_mask;
1067 struct in_device *in_dev = ifa->ifa_dev;
1068 struct net_device *dev = in_dev->dev;
1069
1070 if (!(dev->flags & IFF_UP) ||
1071 ifa->ifa_flags & (IFA_F_SECONDARY | IFA_F_NOPREFIXROUTE) ||
1072 ipv4_is_zeronet(prefix) ||
1073 (prefix == ifa->ifa_local && ifa->ifa_prefixlen == 32))
1074 return;
1075
1076
1077 fib_magic(RTM_NEWROUTE,
1078 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
1079 prefix, ifa->ifa_prefixlen, ifa, new_metric);
1080
1081
1082 fib_magic(RTM_DELROUTE,
1083 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
1084 prefix, ifa->ifa_prefixlen, ifa, ifa->ifa_rt_priority);
1085}
1086
1087
1088
1089
1090
1091
1092void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim)
1093{
1094 struct in_device *in_dev = ifa->ifa_dev;
1095 struct net_device *dev = in_dev->dev;
1096 struct in_ifaddr *ifa1;
1097 struct in_ifaddr *prim = ifa, *prim1 = NULL;
1098 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
1099 __be32 any = ifa->ifa_address & ifa->ifa_mask;
1100#define LOCAL_OK 1
1101#define BRD_OK 2
1102#define BRD0_OK 4
1103#define BRD1_OK 8
1104 unsigned int ok = 0;
1105 int subnet = 0;
1106 int gone = 1;
1107 int same_prefsrc = 0;
1108
1109 if (ifa->ifa_flags & IFA_F_SECONDARY) {
1110 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
1111 if (!prim) {
1112
1113
1114
1115 if (!in_dev->dead)
1116 pr_warn("%s: bug: prim == NULL\n", __func__);
1117 return;
1118 }
1119 if (iprim && iprim != prim) {
1120 pr_warn("%s: bug: iprim != prim\n", __func__);
1121 return;
1122 }
1123 } else if (!ipv4_is_zeronet(any) &&
1124 (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) {
1125 if (!(ifa->ifa_flags & IFA_F_NOPREFIXROUTE))
1126 fib_magic(RTM_DELROUTE,
1127 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
1128 any, ifa->ifa_prefixlen, prim, 0);
1129 subnet = 1;
1130 }
1131
1132 if (in_dev->dead)
1133 goto no_promotions;
1134
1135
1136
1137
1138
1139
1140 rcu_read_lock();
1141 in_dev_for_each_ifa_rcu(ifa1, in_dev) {
1142 if (ifa1 == ifa) {
1143
1144 gone = 0;
1145 continue;
1146 }
1147
1148 if (iprim && ifa1->ifa_mask == iprim->ifa_mask &&
1149 inet_ifa_match(ifa1->ifa_address, iprim))
1150 continue;
1151
1152
1153 if (ifa1->ifa_flags & IFA_F_SECONDARY) {
1154
1155 if (ifa1->ifa_mask == prim->ifa_mask &&
1156 inet_ifa_match(ifa1->ifa_address, prim))
1157 prim1 = prim;
1158 else {
1159
1160
1161
1162 if (!same_prefsrc)
1163 continue;
1164
1165
1166
1167 if (!prim1 ||
1168 ifa1->ifa_mask != prim1->ifa_mask ||
1169 !inet_ifa_match(ifa1->ifa_address, prim1))
1170 prim1 = inet_ifa_byprefix(in_dev,
1171 ifa1->ifa_address,
1172 ifa1->ifa_mask);
1173 if (!prim1)
1174 continue;
1175 if (prim1->ifa_local != prim->ifa_local)
1176 continue;
1177 }
1178 } else {
1179 if (prim->ifa_local != ifa1->ifa_local)
1180 continue;
1181 prim1 = ifa1;
1182 if (prim != prim1)
1183 same_prefsrc = 1;
1184 }
1185 if (ifa->ifa_local == ifa1->ifa_local)
1186 ok |= LOCAL_OK;
1187 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
1188 ok |= BRD_OK;
1189 if (brd == ifa1->ifa_broadcast)
1190 ok |= BRD1_OK;
1191 if (any == ifa1->ifa_broadcast)
1192 ok |= BRD0_OK;
1193
1194 if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) {
1195 __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask;
1196 __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask;
1197
1198 if (!ipv4_is_zeronet(any1)) {
1199 if (ifa->ifa_broadcast == brd1 ||
1200 ifa->ifa_broadcast == any1)
1201 ok |= BRD_OK;
1202 if (brd == brd1 || brd == any1)
1203 ok |= BRD1_OK;
1204 if (any == brd1 || any == any1)
1205 ok |= BRD0_OK;
1206 }
1207 }
1208 }
1209 rcu_read_unlock();
1210
1211no_promotions:
1212 if (!(ok & BRD_OK))
1213 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32,
1214 prim, 0);
1215 if (subnet && ifa->ifa_prefixlen < 31) {
1216 if (!(ok & BRD1_OK))
1217 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32,
1218 prim, 0);
1219 if (!(ok & BRD0_OK))
1220 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32,
1221 prim, 0);
1222 }
1223 if (!(ok & LOCAL_OK)) {
1224 unsigned int addr_type;
1225
1226 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim, 0);
1227
1228
1229 addr_type = inet_addr_type_dev_table(dev_net(dev), dev,
1230 ifa->ifa_local);
1231 if (gone && addr_type != RTN_LOCAL) {
1232
1233
1234
1235
1236
1237
1238 if (fib_sync_down_addr(dev, ifa->ifa_local))
1239 fib_flush(dev_net(dev));
1240 }
1241 }
1242#undef LOCAL_OK
1243#undef BRD_OK
1244#undef BRD0_OK
1245#undef BRD1_OK
1246}
1247
1248static void nl_fib_lookup(struct net *net, struct fib_result_nl *frn)
1249{
1250
1251 struct fib_result res;
1252 struct flowi4 fl4 = {
1253 .flowi4_mark = frn->fl_mark,
1254 .daddr = frn->fl_addr,
1255 .flowi4_tos = frn->fl_tos,
1256 .flowi4_scope = frn->fl_scope,
1257 };
1258 struct fib_table *tb;
1259
1260 rcu_read_lock();
1261
1262 tb = fib_get_table(net, frn->tb_id_in);
1263
1264 frn->err = -ENOENT;
1265 if (tb) {
1266 local_bh_disable();
1267
1268 frn->tb_id = tb->tb_id;
1269 frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF);
1270
1271 if (!frn->err) {
1272 frn->prefixlen = res.prefixlen;
1273 frn->nh_sel = res.nh_sel;
1274 frn->type = res.type;
1275 frn->scope = res.scope;
1276 }
1277 local_bh_enable();
1278 }
1279
1280 rcu_read_unlock();
1281}
1282
1283static void nl_fib_input(struct sk_buff *skb)
1284{
1285 struct net *net;
1286 struct fib_result_nl *frn;
1287 struct nlmsghdr *nlh;
1288 u32 portid;
1289
1290 net = sock_net(skb->sk);
1291 nlh = nlmsg_hdr(skb);
1292 if (skb->len < nlmsg_total_size(sizeof(*frn)) ||
1293 skb->len < nlh->nlmsg_len ||
1294 nlmsg_len(nlh) < sizeof(*frn))
1295 return;
1296
1297 skb = netlink_skb_clone(skb, GFP_KERNEL);
1298 if (!skb)
1299 return;
1300 nlh = nlmsg_hdr(skb);
1301
1302 frn = (struct fib_result_nl *) nlmsg_data(nlh);
1303 nl_fib_lookup(net, frn);
1304
1305 portid = NETLINK_CB(skb).portid;
1306 NETLINK_CB(skb).portid = 0;
1307 NETLINK_CB(skb).dst_group = 0;
1308 nlmsg_unicast(net->ipv4.fibnl, skb, portid);
1309}
1310
1311static int __net_init nl_fib_lookup_init(struct net *net)
1312{
1313 struct sock *sk;
1314 struct netlink_kernel_cfg cfg = {
1315 .input = nl_fib_input,
1316 };
1317
1318 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, &cfg);
1319 if (!sk)
1320 return -EAFNOSUPPORT;
1321 net->ipv4.fibnl = sk;
1322 return 0;
1323}
1324
1325static void nl_fib_lookup_exit(struct net *net)
1326{
1327 netlink_kernel_release(net->ipv4.fibnl);
1328 net->ipv4.fibnl = NULL;
1329}
1330
1331static void fib_disable_ip(struct net_device *dev, unsigned long event,
1332 bool force)
1333{
1334 if (fib_sync_down_dev(dev, event, force))
1335 fib_flush(dev_net(dev));
1336 else
1337 rt_cache_flush(dev_net(dev));
1338 arp_ifdown(dev);
1339}
1340
1341static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
1342{
1343 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
1344 struct net_device *dev = ifa->ifa_dev->dev;
1345 struct net *net = dev_net(dev);
1346
1347 switch (event) {
1348 case NETDEV_UP:
1349 fib_add_ifaddr(ifa);
1350#ifdef CONFIG_IP_ROUTE_MULTIPATH
1351 fib_sync_up(dev, RTNH_F_DEAD);
1352#endif
1353 atomic_inc(&net->ipv4.dev_addr_genid);
1354 rt_cache_flush(dev_net(dev));
1355 break;
1356 case NETDEV_DOWN:
1357 fib_del_ifaddr(ifa, NULL);
1358 atomic_inc(&net->ipv4.dev_addr_genid);
1359 if (!ifa->ifa_dev->ifa_list) {
1360
1361
1362
1363 fib_disable_ip(dev, event, true);
1364 } else {
1365 rt_cache_flush(dev_net(dev));
1366 }
1367 break;
1368 }
1369 return NOTIFY_DONE;
1370}
1371
1372static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1373{
1374 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1375 struct netdev_notifier_changeupper_info *upper_info = ptr;
1376 struct netdev_notifier_info_ext *info_ext = ptr;
1377 struct in_device *in_dev;
1378 struct net *net = dev_net(dev);
1379 struct in_ifaddr *ifa;
1380 unsigned int flags;
1381
1382 if (event == NETDEV_UNREGISTER) {
1383 fib_disable_ip(dev, event, true);
1384 rt_flush_dev(dev);
1385 return NOTIFY_DONE;
1386 }
1387
1388 in_dev = __in_dev_get_rtnl(dev);
1389 if (!in_dev)
1390 return NOTIFY_DONE;
1391
1392 switch (event) {
1393 case NETDEV_UP:
1394 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1395 fib_add_ifaddr(ifa);
1396 }
1397#ifdef CONFIG_IP_ROUTE_MULTIPATH
1398 fib_sync_up(dev, RTNH_F_DEAD);
1399#endif
1400 atomic_inc(&net->ipv4.dev_addr_genid);
1401 rt_cache_flush(net);
1402 break;
1403 case NETDEV_DOWN:
1404 fib_disable_ip(dev, event, false);
1405 break;
1406 case NETDEV_CHANGE:
1407 flags = dev_get_flags(dev);
1408 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1409 fib_sync_up(dev, RTNH_F_LINKDOWN);
1410 else
1411 fib_sync_down_dev(dev, event, false);
1412 rt_cache_flush(net);
1413 break;
1414 case NETDEV_CHANGEMTU:
1415 fib_sync_mtu(dev, info_ext->ext.mtu);
1416 rt_cache_flush(net);
1417 break;
1418 case NETDEV_CHANGEUPPER:
1419 upper_info = ptr;
1420
1421
1422
1423 if (upper_info->upper_dev &&
1424 netif_is_l3_master(upper_info->upper_dev))
1425 fib_disable_ip(dev, NETDEV_DOWN, true);
1426 break;
1427 }
1428 return NOTIFY_DONE;
1429}
1430
1431static struct notifier_block fib_inetaddr_notifier = {
1432 .notifier_call = fib_inetaddr_event,
1433};
1434
1435static struct notifier_block fib_netdev_notifier = {
1436 .notifier_call = fib_netdev_event,
1437};
1438
1439static int __net_init ip_fib_net_init(struct net *net)
1440{
1441 int err;
1442 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
1443
1444 err = fib4_notifier_init(net);
1445 if (err)
1446 return err;
1447
1448
1449 size = max_t(size_t, size, L1_CACHE_BYTES);
1450
1451 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
1452 if (!net->ipv4.fib_table_hash) {
1453 err = -ENOMEM;
1454 goto err_table_hash_alloc;
1455 }
1456
1457 err = fib4_rules_init(net);
1458 if (err < 0)
1459 goto err_rules_init;
1460 return 0;
1461
1462err_rules_init:
1463 kfree(net->ipv4.fib_table_hash);
1464err_table_hash_alloc:
1465 fib4_notifier_exit(net);
1466 return err;
1467}
1468
1469static void ip_fib_net_exit(struct net *net)
1470{
1471 int i;
1472
1473 rtnl_lock();
1474#ifdef CONFIG_IP_MULTIPLE_TABLES
1475 RCU_INIT_POINTER(net->ipv4.fib_main, NULL);
1476 RCU_INIT_POINTER(net->ipv4.fib_default, NULL);
1477#endif
1478
1479
1480
1481
1482
1483 for (i = FIB_TABLE_HASHSZ - 1; i >= 0; i--) {
1484 struct hlist_head *head = &net->ipv4.fib_table_hash[i];
1485 struct hlist_node *tmp;
1486 struct fib_table *tb;
1487
1488 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) {
1489 hlist_del(&tb->tb_hlist);
1490 fib_table_flush(net, tb, true);
1491 fib_free_table(tb);
1492 }
1493 }
1494
1495#ifdef CONFIG_IP_MULTIPLE_TABLES
1496 fib4_rules_exit(net);
1497#endif
1498 rtnl_unlock();
1499 kfree(net->ipv4.fib_table_hash);
1500 fib4_notifier_exit(net);
1501}
1502
1503static int __net_init fib_net_init(struct net *net)
1504{
1505 int error;
1506
1507#ifdef CONFIG_IP_ROUTE_CLASSID
1508 net->ipv4.fib_num_tclassid_users = 0;
1509#endif
1510 error = ip_fib_net_init(net);
1511 if (error < 0)
1512 goto out;
1513 error = nl_fib_lookup_init(net);
1514 if (error < 0)
1515 goto out_nlfl;
1516 error = fib_proc_init(net);
1517 if (error < 0)
1518 goto out_proc;
1519out:
1520 return error;
1521
1522out_proc:
1523 nl_fib_lookup_exit(net);
1524out_nlfl:
1525 ip_fib_net_exit(net);
1526 goto out;
1527}
1528
1529static void __net_exit fib_net_exit(struct net *net)
1530{
1531 fib_proc_exit(net);
1532 nl_fib_lookup_exit(net);
1533 ip_fib_net_exit(net);
1534}
1535
1536static struct pernet_operations fib_net_ops = {
1537 .init = fib_net_init,
1538 .exit = fib_net_exit,
1539};
1540
1541void __init ip_fib_init(void)
1542{
1543 fib_trie_init();
1544
1545 register_pernet_subsys(&fib_net_ops);
1546
1547 register_netdevice_notifier(&fib_netdev_notifier);
1548 register_inetaddr_notifier(&fib_inetaddr_notifier);
1549
1550 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, 0);
1551 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, 0);
1552 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, 0);
1553}
1554
