]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - net/ipv4/fib_frontend.c
7d9cce3277cad456ae9a282c6477f406eff30f66
[mirror_ubuntu-bionic-kernel.git] / net / ipv4 / fib_frontend.c
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * IPv4 Forwarding Information Base: FIB frontend.
7 *
8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
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
54 static 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
72 fail:
73 fib_free_table(main_table);
74 return -ENOMEM;
75 }
76
77 static bool fib4_has_custom_rules(struct net *net)
78 {
79 return false;
80 }
81 #else
82
83 struct 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 }
116 EXPORT_SYMBOL_GPL(fib_new_table);
117
118 /* caller must hold either rtnl or rcu read lock */
119 struct 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 if (tb->tb_id == id)
132 return tb;
133 }
134 return NULL;
135 }
136
137 static bool fib4_has_custom_rules(struct net *net)
138 {
139 return net->ipv4.fib_has_custom_rules;
140 }
141 #endif /* CONFIG_IP_MULTIPLE_TABLES */
142
143 static void fib_replace_table(struct net *net, struct fib_table *old,
144 struct fib_table *new)
145 {
146 #ifdef CONFIG_IP_MULTIPLE_TABLES
147 switch (new->tb_id) {
148 case RT_TABLE_MAIN:
149 rcu_assign_pointer(net->ipv4.fib_main, new);
150 break;
151 case RT_TABLE_DEFAULT:
152 rcu_assign_pointer(net->ipv4.fib_default, new);
153 break;
154 default:
155 break;
156 }
157
158 #endif
159 /* replace the old table in the hlist */
160 hlist_replace_rcu(&old->tb_hlist, &new->tb_hlist);
161 }
162
163 int fib_unmerge(struct net *net)
164 {
165 struct fib_table *old, *new, *main_table;
166
167 /* attempt to fetch local table if it has been allocated */
168 old = fib_get_table(net, RT_TABLE_LOCAL);
169 if (!old)
170 return 0;
171
172 new = fib_trie_unmerge(old);
173 if (!new)
174 return -ENOMEM;
175
176 /* table is already unmerged */
177 if (new == old)
178 return 0;
179
180 /* replace merged table with clean table */
181 fib_replace_table(net, old, new);
182 fib_free_table(old);
183
184 /* attempt to fetch main table if it has been allocated */
185 main_table = fib_get_table(net, RT_TABLE_MAIN);
186 if (!main_table)
187 return 0;
188
189 /* flush local entries from main table */
190 fib_table_flush_external(main_table);
191
192 return 0;
193 }
194
195 static void fib_flush(struct net *net)
196 {
197 int flushed = 0;
198 unsigned int h;
199
200 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
201 struct hlist_head *head = &net->ipv4.fib_table_hash[h];
202 struct hlist_node *tmp;
203 struct fib_table *tb;
204
205 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist)
206 flushed += fib_table_flush(net, tb);
207 }
208
209 if (flushed)
210 rt_cache_flush(net);
211 }
212
213 /*
214 * Find address type as if only "dev" was present in the system. If
215 * on_dev is NULL then all interfaces are taken into consideration.
216 */
217 static inline unsigned int __inet_dev_addr_type(struct net *net,
218 const struct net_device *dev,
219 __be32 addr, u32 tb_id)
220 {
221 struct flowi4 fl4 = { .daddr = addr };
222 struct fib_result res;
223 unsigned int ret = RTN_BROADCAST;
224 struct fib_table *table;
225
226 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
227 return RTN_BROADCAST;
228 if (ipv4_is_multicast(addr))
229 return RTN_MULTICAST;
230
231 rcu_read_lock();
232
233 table = fib_get_table(net, tb_id);
234 if (table) {
235 ret = RTN_UNICAST;
236 if (!fib_table_lookup(table, &fl4, &res, FIB_LOOKUP_NOREF)) {
237 if (!dev || dev == res.fi->fib_dev)
238 ret = res.type;
239 }
240 }
241
242 rcu_read_unlock();
243 return ret;
244 }
245
246 unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id)
247 {
248 return __inet_dev_addr_type(net, NULL, addr, tb_id);
249 }
250 EXPORT_SYMBOL(inet_addr_type_table);
251
252 unsigned int inet_addr_type(struct net *net, __be32 addr)
253 {
254 return __inet_dev_addr_type(net, NULL, addr, RT_TABLE_LOCAL);
255 }
256 EXPORT_SYMBOL(inet_addr_type);
257
258 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
259 __be32 addr)
260 {
261 u32 rt_table = l3mdev_fib_table(dev) ? : RT_TABLE_LOCAL;
262
263 return __inet_dev_addr_type(net, dev, addr, rt_table);
264 }
265 EXPORT_SYMBOL(inet_dev_addr_type);
266
267 /* inet_addr_type with dev == NULL but using the table from a dev
268 * if one is associated
269 */
270 unsigned int inet_addr_type_dev_table(struct net *net,
271 const struct net_device *dev,
272 __be32 addr)
273 {
274 u32 rt_table = l3mdev_fib_table(dev) ? : RT_TABLE_LOCAL;
275
276 return __inet_dev_addr_type(net, NULL, addr, rt_table);
277 }
278 EXPORT_SYMBOL(inet_addr_type_dev_table);
279
280 __be32 fib_compute_spec_dst(struct sk_buff *skb)
281 {
282 struct net_device *dev = skb->dev;
283 struct in_device *in_dev;
284 struct fib_result res;
285 struct rtable *rt;
286 struct net *net;
287 int scope;
288
289 rt = skb_rtable(skb);
290 if ((rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | RTCF_LOCAL)) ==
291 RTCF_LOCAL)
292 return ip_hdr(skb)->daddr;
293
294 in_dev = __in_dev_get_rcu(dev);
295
296 net = dev_net(dev);
297
298 scope = RT_SCOPE_UNIVERSE;
299 if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) {
300 bool vmark = in_dev && IN_DEV_SRC_VMARK(in_dev);
301 struct flowi4 fl4 = {
302 .flowi4_iif = LOOPBACK_IFINDEX,
303 .flowi4_oif = l3mdev_master_ifindex_rcu(dev),
304 .daddr = ip_hdr(skb)->saddr,
305 .flowi4_tos = RT_TOS(ip_hdr(skb)->tos),
306 .flowi4_scope = scope,
307 .flowi4_mark = vmark ? skb->mark : 0,
308 };
309 if (!fib_lookup(net, &fl4, &res, 0))
310 return FIB_RES_PREFSRC(net, res);
311 } else {
312 scope = RT_SCOPE_LINK;
313 }
314
315 return inet_select_addr(dev, ip_hdr(skb)->saddr, scope);
316 }
317
318 /* Given (packet source, input interface) and optional (dst, oif, tos):
319 * - (main) check, that source is valid i.e. not broadcast or our local
320 * address.
321 * - figure out what "logical" interface this packet arrived
322 * and calculate "specific destination" address.
323 * - check, that packet arrived from expected physical interface.
324 * called with rcu_read_lock()
325 */
326 static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
327 u8 tos, int oif, struct net_device *dev,
328 int rpf, struct in_device *idev, u32 *itag)
329 {
330 int ret, no_addr;
331 struct fib_result res;
332 struct flowi4 fl4;
333 struct net *net = dev_net(dev);
334 bool dev_match;
335
336 fl4.flowi4_oif = 0;
337 fl4.flowi4_iif = l3mdev_master_ifindex_rcu(dev);
338 if (!fl4.flowi4_iif)
339 fl4.flowi4_iif = oif ? : LOOPBACK_IFINDEX;
340 fl4.daddr = src;
341 fl4.saddr = dst;
342 fl4.flowi4_tos = tos;
343 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
344 fl4.flowi4_tun_key.tun_id = 0;
345 fl4.flowi4_flags = 0;
346 fl4.flowi4_uid = sock_net_uid(net, NULL);
347
348 no_addr = idev->ifa_list == NULL;
349
350 fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0;
351
352 trace_fib_validate_source(dev, &fl4);
353
354 if (fib_lookup(net, &fl4, &res, 0))
355 goto last_resort;
356 if (res.type != RTN_UNICAST &&
357 (res.type != RTN_LOCAL || !IN_DEV_ACCEPT_LOCAL(idev)))
358 goto e_inval;
359 fib_combine_itag(itag, &res);
360 dev_match = false;
361
362 #ifdef CONFIG_IP_ROUTE_MULTIPATH
363 for (ret = 0; ret < res.fi->fib_nhs; ret++) {
364 struct fib_nh *nh = &res.fi->fib_nh[ret];
365
366 if (nh->nh_dev == dev) {
367 dev_match = true;
368 break;
369 } else if (l3mdev_master_ifindex_rcu(nh->nh_dev) == dev->ifindex) {
370 dev_match = true;
371 break;
372 }
373 }
374 #else
375 if (FIB_RES_DEV(res) == dev)
376 dev_match = true;
377 #endif
378 if (dev_match) {
379 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
380 return ret;
381 }
382 if (no_addr)
383 goto last_resort;
384 if (rpf == 1)
385 goto e_rpf;
386 fl4.flowi4_oif = dev->ifindex;
387
388 ret = 0;
389 if (fib_lookup(net, &fl4, &res, FIB_LOOKUP_IGNORE_LINKSTATE) == 0) {
390 if (res.type == RTN_UNICAST)
391 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
392 }
393 return ret;
394
395 last_resort:
396 if (rpf)
397 goto e_rpf;
398 *itag = 0;
399 return 0;
400
401 e_inval:
402 return -EINVAL;
403 e_rpf:
404 return -EXDEV;
405 }
406
407 /* Ignore rp_filter for packets protected by IPsec. */
408 int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
409 u8 tos, int oif, struct net_device *dev,
410 struct in_device *idev, u32 *itag)
411 {
412 int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev);
413 struct net *net = dev_net(dev);
414
415 if (!r && !fib_num_tclassid_users(net) &&
416 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) {
417 if (IN_DEV_ACCEPT_LOCAL(idev))
418 goto ok;
419 /* with custom local routes in place, checking local addresses
420 * only will be too optimistic, with custom rules, checking
421 * local addresses only can be too strict, e.g. due to vrf
422 */
423 if (net->ipv4.fib_has_custom_local_routes ||
424 fib4_has_custom_rules(net))
425 goto full_check;
426 if (inet_lookup_ifaddr_rcu(net, src))
427 return -EINVAL;
428
429 ok:
430 *itag = 0;
431 return 0;
432 }
433
434 full_check:
435 return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag);
436 }
437
438 static inline __be32 sk_extract_addr(struct sockaddr *addr)
439 {
440 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
441 }
442
443 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
444 {
445 struct nlattr *nla;
446
447 nla = (struct nlattr *) ((char *) mx + len);
448 nla->nla_type = type;
449 nla->nla_len = nla_attr_size(4);
450 *(u32 *) nla_data(nla) = value;
451
452 return len + nla_total_size(4);
453 }
454
455 static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
456 struct fib_config *cfg)
457 {
458 __be32 addr;
459 int plen;
460
461 memset(cfg, 0, sizeof(*cfg));
462 cfg->fc_nlinfo.nl_net = net;
463
464 if (rt->rt_dst.sa_family != AF_INET)
465 return -EAFNOSUPPORT;
466
467 /*
468 * Check mask for validity:
469 * a) it must be contiguous.
470 * b) destination must have all host bits clear.
471 * c) if application forgot to set correct family (AF_INET),
472 * reject request unless it is absolutely clear i.e.
473 * both family and mask are zero.
474 */
475 plen = 32;
476 addr = sk_extract_addr(&rt->rt_dst);
477 if (!(rt->rt_flags & RTF_HOST)) {
478 __be32 mask = sk_extract_addr(&rt->rt_genmask);
479
480 if (rt->rt_genmask.sa_family != AF_INET) {
481 if (mask || rt->rt_genmask.sa_family)
482 return -EAFNOSUPPORT;
483 }
484
485 if (bad_mask(mask, addr))
486 return -EINVAL;
487
488 plen = inet_mask_len(mask);
489 }
490
491 cfg->fc_dst_len = plen;
492 cfg->fc_dst = addr;
493
494 if (cmd != SIOCDELRT) {
495 cfg->fc_nlflags = NLM_F_CREATE;
496 cfg->fc_protocol = RTPROT_BOOT;
497 }
498
499 if (rt->rt_metric)
500 cfg->fc_priority = rt->rt_metric - 1;
501
502 if (rt->rt_flags & RTF_REJECT) {
503 cfg->fc_scope = RT_SCOPE_HOST;
504 cfg->fc_type = RTN_UNREACHABLE;
505 return 0;
506 }
507
508 cfg->fc_scope = RT_SCOPE_NOWHERE;
509 cfg->fc_type = RTN_UNICAST;
510
511 if (rt->rt_dev) {
512 char *colon;
513 struct net_device *dev;
514 char devname[IFNAMSIZ];
515
516 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
517 return -EFAULT;
518
519 devname[IFNAMSIZ-1] = 0;
520 colon = strchr(devname, ':');
521 if (colon)
522 *colon = 0;
523 dev = __dev_get_by_name(net, devname);
524 if (!dev)
525 return -ENODEV;
526 cfg->fc_oif = dev->ifindex;
527 cfg->fc_table = l3mdev_fib_table(dev);
528 if (colon) {
529 struct in_ifaddr *ifa;
530 struct in_device *in_dev = __in_dev_get_rtnl(dev);
531 if (!in_dev)
532 return -ENODEV;
533 *colon = ':';
534 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
535 if (strcmp(ifa->ifa_label, devname) == 0)
536 break;
537 if (!ifa)
538 return -ENODEV;
539 cfg->fc_prefsrc = ifa->ifa_local;
540 }
541 }
542
543 addr = sk_extract_addr(&rt->rt_gateway);
544 if (rt->rt_gateway.sa_family == AF_INET && addr) {
545 unsigned int addr_type;
546
547 cfg->fc_gw = addr;
548 addr_type = inet_addr_type_table(net, addr, cfg->fc_table);
549 if (rt->rt_flags & RTF_GATEWAY &&
550 addr_type == RTN_UNICAST)
551 cfg->fc_scope = RT_SCOPE_UNIVERSE;
552 }
553
554 if (cmd == SIOCDELRT)
555 return 0;
556
557 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
558 return -EINVAL;
559
560 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
561 cfg->fc_scope = RT_SCOPE_LINK;
562
563 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
564 struct nlattr *mx;
565 int len = 0;
566
567 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
568 if (!mx)
569 return -ENOMEM;
570
571 if (rt->rt_flags & RTF_MTU)
572 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
573
574 if (rt->rt_flags & RTF_WINDOW)
575 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
576
577 if (rt->rt_flags & RTF_IRTT)
578 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
579
580 cfg->fc_mx = mx;
581 cfg->fc_mx_len = len;
582 }
583
584 return 0;
585 }
586
587 /*
588 * Handle IP routing ioctl calls.
589 * These are used to manipulate the routing tables
590 */
591 int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
592 {
593 struct fib_config cfg;
594 struct rtentry rt;
595 int err;
596
597 switch (cmd) {
598 case SIOCADDRT: /* Add a route */
599 case SIOCDELRT: /* Delete a route */
600 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
601 return -EPERM;
602
603 if (copy_from_user(&rt, arg, sizeof(rt)))
604 return -EFAULT;
605
606 rtnl_lock();
607 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
608 if (err == 0) {
609 struct fib_table *tb;
610
611 if (cmd == SIOCDELRT) {
612 tb = fib_get_table(net, cfg.fc_table);
613 if (tb)
614 err = fib_table_delete(net, tb, &cfg,
615 NULL);
616 else
617 err = -ESRCH;
618 } else {
619 tb = fib_new_table(net, cfg.fc_table);
620 if (tb)
621 err = fib_table_insert(net, tb,
622 &cfg, NULL);
623 else
624 err = -ENOBUFS;
625 }
626
627 /* allocated by rtentry_to_fib_config() */
628 kfree(cfg.fc_mx);
629 }
630 rtnl_unlock();
631 return err;
632 }
633 return -EINVAL;
634 }
635
636 const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
637 [RTA_DST] = { .type = NLA_U32 },
638 [RTA_SRC] = { .type = NLA_U32 },
639 [RTA_IIF] = { .type = NLA_U32 },
640 [RTA_OIF] = { .type = NLA_U32 },
641 [RTA_GATEWAY] = { .type = NLA_U32 },
642 [RTA_PRIORITY] = { .type = NLA_U32 },
643 [RTA_PREFSRC] = { .type = NLA_U32 },
644 [RTA_METRICS] = { .type = NLA_NESTED },
645 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
646 [RTA_FLOW] = { .type = NLA_U32 },
647 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
648 [RTA_ENCAP] = { .type = NLA_NESTED },
649 [RTA_UID] = { .type = NLA_U32 },
650 [RTA_MARK] = { .type = NLA_U32 },
651 [RTA_TABLE] = { .type = NLA_U32 },
652 };
653
654 static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
655 struct nlmsghdr *nlh, struct fib_config *cfg,
656 struct netlink_ext_ack *extack)
657 {
658 struct nlattr *attr;
659 int err, remaining;
660 struct rtmsg *rtm;
661
662 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy,
663 extack);
664 if (err < 0)
665 goto errout;
666
667 memset(cfg, 0, sizeof(*cfg));
668
669 rtm = nlmsg_data(nlh);
670 cfg->fc_dst_len = rtm->rtm_dst_len;
671 cfg->fc_tos = rtm->rtm_tos;
672 cfg->fc_table = rtm->rtm_table;
673 cfg->fc_protocol = rtm->rtm_protocol;
674 cfg->fc_scope = rtm->rtm_scope;
675 cfg->fc_type = rtm->rtm_type;
676 cfg->fc_flags = rtm->rtm_flags;
677 cfg->fc_nlflags = nlh->nlmsg_flags;
678
679 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
680 cfg->fc_nlinfo.nlh = nlh;
681 cfg->fc_nlinfo.nl_net = net;
682
683 if (cfg->fc_type > RTN_MAX) {
684 NL_SET_ERR_MSG(extack, "Invalid route type");
685 err = -EINVAL;
686 goto errout;
687 }
688
689 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
690 switch (nla_type(attr)) {
691 case RTA_DST:
692 cfg->fc_dst = nla_get_be32(attr);
693 break;
694 case RTA_OIF:
695 cfg->fc_oif = nla_get_u32(attr);
696 break;
697 case RTA_GATEWAY:
698 cfg->fc_gw = nla_get_be32(attr);
699 break;
700 case RTA_PRIORITY:
701 cfg->fc_priority = nla_get_u32(attr);
702 break;
703 case RTA_PREFSRC:
704 cfg->fc_prefsrc = nla_get_be32(attr);
705 break;
706 case RTA_METRICS:
707 cfg->fc_mx = nla_data(attr);
708 cfg->fc_mx_len = nla_len(attr);
709 break;
710 case RTA_MULTIPATH:
711 err = lwtunnel_valid_encap_type_attr(nla_data(attr),
712 nla_len(attr),
713 extack);
714 if (err < 0)
715 goto errout;
716 cfg->fc_mp = nla_data(attr);
717 cfg->fc_mp_len = nla_len(attr);
718 break;
719 case RTA_FLOW:
720 cfg->fc_flow = nla_get_u32(attr);
721 break;
722 case RTA_TABLE:
723 cfg->fc_table = nla_get_u32(attr);
724 break;
725 case RTA_ENCAP:
726 cfg->fc_encap = attr;
727 break;
728 case RTA_ENCAP_TYPE:
729 cfg->fc_encap_type = nla_get_u16(attr);
730 err = lwtunnel_valid_encap_type(cfg->fc_encap_type,
731 extack);
732 if (err < 0)
733 goto errout;
734 break;
735 }
736 }
737
738 return 0;
739 errout:
740 return err;
741 }
742
743 static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
744 struct netlink_ext_ack *extack)
745 {
746 struct net *net = sock_net(skb->sk);
747 struct fib_config cfg;
748 struct fib_table *tb;
749 int err;
750
751 err = rtm_to_fib_config(net, skb, nlh, &cfg, extack);
752 if (err < 0)
753 goto errout;
754
755 tb = fib_get_table(net, cfg.fc_table);
756 if (!tb) {
757 NL_SET_ERR_MSG(extack, "FIB table does not exist");
758 err = -ESRCH;
759 goto errout;
760 }
761
762 err = fib_table_delete(net, tb, &cfg, extack);
763 errout:
764 return err;
765 }
766
767 static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
768 struct netlink_ext_ack *extack)
769 {
770 struct net *net = sock_net(skb->sk);
771 struct fib_config cfg;
772 struct fib_table *tb;
773 int err;
774
775 err = rtm_to_fib_config(net, skb, nlh, &cfg, extack);
776 if (err < 0)
777 goto errout;
778
779 tb = fib_new_table(net, cfg.fc_table);
780 if (!tb) {
781 err = -ENOBUFS;
782 goto errout;
783 }
784
785 err = fib_table_insert(net, tb, &cfg, extack);
786 if (!err && cfg.fc_type == RTN_LOCAL)
787 net->ipv4.fib_has_custom_local_routes = true;
788 errout:
789 return err;
790 }
791
792 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
793 {
794 struct net *net = sock_net(skb->sk);
795 unsigned int h, s_h;
796 unsigned int e = 0, s_e;
797 struct fib_table *tb;
798 struct hlist_head *head;
799 int dumped = 0, err;
800
801 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
802 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
803 return skb->len;
804
805 s_h = cb->args[0];
806 s_e = cb->args[1];
807
808 rcu_read_lock();
809
810 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
811 e = 0;
812 head = &net->ipv4.fib_table_hash[h];
813 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
814 if (e < s_e)
815 goto next;
816 if (dumped)
817 memset(&cb->args[2], 0, sizeof(cb->args) -
818 2 * sizeof(cb->args[0]));
819 err = fib_table_dump(tb, skb, cb);
820 if (err < 0) {
821 if (likely(skb->len))
822 goto out;
823
824 goto out_err;
825 }
826 dumped = 1;
827 next:
828 e++;
829 }
830 }
831 out:
832 err = skb->len;
833 out_err:
834 rcu_read_unlock();
835
836 cb->args[1] = e;
837 cb->args[0] = h;
838
839 return err;
840 }
841
842 /* Prepare and feed intra-kernel routing request.
843 * Really, it should be netlink message, but :-( netlink
844 * can be not configured, so that we feed it directly
845 * to fib engine. It is legal, because all events occur
846 * only when netlink is already locked.
847 */
848 static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
849 {
850 struct net *net = dev_net(ifa->ifa_dev->dev);
851 u32 tb_id = l3mdev_fib_table(ifa->ifa_dev->dev);
852 struct fib_table *tb;
853 struct fib_config cfg = {
854 .fc_protocol = RTPROT_KERNEL,
855 .fc_type = type,
856 .fc_dst = dst,
857 .fc_dst_len = dst_len,
858 .fc_prefsrc = ifa->ifa_local,
859 .fc_oif = ifa->ifa_dev->dev->ifindex,
860 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
861 .fc_nlinfo = {
862 .nl_net = net,
863 },
864 };
865
866 if (!tb_id)
867 tb_id = (type == RTN_UNICAST) ? RT_TABLE_MAIN : RT_TABLE_LOCAL;
868
869 tb = fib_new_table(net, tb_id);
870 if (!tb)
871 return;
872
873 cfg.fc_table = tb->tb_id;
874
875 if (type != RTN_LOCAL)
876 cfg.fc_scope = RT_SCOPE_LINK;
877 else
878 cfg.fc_scope = RT_SCOPE_HOST;
879
880 if (cmd == RTM_NEWROUTE)
881 fib_table_insert(net, tb, &cfg, NULL);
882 else
883 fib_table_delete(net, tb, &cfg, NULL);
884 }
885
886 void fib_add_ifaddr(struct in_ifaddr *ifa)
887 {
888 struct in_device *in_dev = ifa->ifa_dev;
889 struct net_device *dev = in_dev->dev;
890 struct in_ifaddr *prim = ifa;
891 __be32 mask = ifa->ifa_mask;
892 __be32 addr = ifa->ifa_local;
893 __be32 prefix = ifa->ifa_address & mask;
894
895 if (ifa->ifa_flags & IFA_F_SECONDARY) {
896 prim = inet_ifa_byprefix(in_dev, prefix, mask);
897 if (!prim) {
898 pr_warn("%s: bug: prim == NULL\n", __func__);
899 return;
900 }
901 }
902
903 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
904
905 if (!(dev->flags & IFF_UP))
906 return;
907
908 /* Add broadcast address, if it is explicitly assigned. */
909 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
910 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
911
912 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
913 (prefix != addr || ifa->ifa_prefixlen < 32)) {
914 if (!(ifa->ifa_flags & IFA_F_NOPREFIXROUTE))
915 fib_magic(RTM_NEWROUTE,
916 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
917 prefix, ifa->ifa_prefixlen, prim);
918
919 /* Add network specific broadcasts, when it takes a sense */
920 if (ifa->ifa_prefixlen < 31) {
921 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
922 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
923 32, prim);
924 }
925 }
926 }
927
928 /* Delete primary or secondary address.
929 * Optionally, on secondary address promotion consider the addresses
930 * from subnet iprim as deleted, even if they are in device list.
931 * In this case the secondary ifa can be in device list.
932 */
933 void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim)
934 {
935 struct in_device *in_dev = ifa->ifa_dev;
936 struct net_device *dev = in_dev->dev;
937 struct in_ifaddr *ifa1;
938 struct in_ifaddr *prim = ifa, *prim1 = NULL;
939 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
940 __be32 any = ifa->ifa_address & ifa->ifa_mask;
941 #define LOCAL_OK 1
942 #define BRD_OK 2
943 #define BRD0_OK 4
944 #define BRD1_OK 8
945 unsigned int ok = 0;
946 int subnet = 0; /* Primary network */
947 int gone = 1; /* Address is missing */
948 int same_prefsrc = 0; /* Another primary with same IP */
949
950 if (ifa->ifa_flags & IFA_F_SECONDARY) {
951 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
952 if (!prim) {
953 /* if the device has been deleted, we don't perform
954 * address promotion
955 */
956 if (!in_dev->dead)
957 pr_warn("%s: bug: prim == NULL\n", __func__);
958 return;
959 }
960 if (iprim && iprim != prim) {
961 pr_warn("%s: bug: iprim != prim\n", __func__);
962 return;
963 }
964 } else if (!ipv4_is_zeronet(any) &&
965 (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) {
966 if (!(ifa->ifa_flags & IFA_F_NOPREFIXROUTE))
967 fib_magic(RTM_DELROUTE,
968 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
969 any, ifa->ifa_prefixlen, prim);
970 subnet = 1;
971 }
972
973 if (in_dev->dead)
974 goto no_promotions;
975
976 /* Deletion is more complicated than add.
977 * We should take care of not to delete too much :-)
978 *
979 * Scan address list to be sure that addresses are really gone.
980 */
981
982 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
983 if (ifa1 == ifa) {
984 /* promotion, keep the IP */
985 gone = 0;
986 continue;
987 }
988 /* Ignore IFAs from our subnet */
989 if (iprim && ifa1->ifa_mask == iprim->ifa_mask &&
990 inet_ifa_match(ifa1->ifa_address, iprim))
991 continue;
992
993 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
994 if (ifa1->ifa_flags & IFA_F_SECONDARY) {
995 /* Another address from our subnet? */
996 if (ifa1->ifa_mask == prim->ifa_mask &&
997 inet_ifa_match(ifa1->ifa_address, prim))
998 prim1 = prim;
999 else {
1000 /* We reached the secondaries, so
1001 * same_prefsrc should be determined.
1002 */
1003 if (!same_prefsrc)
1004 continue;
1005 /* Search new prim1 if ifa1 is not
1006 * using the current prim1
1007 */
1008 if (!prim1 ||
1009 ifa1->ifa_mask != prim1->ifa_mask ||
1010 !inet_ifa_match(ifa1->ifa_address, prim1))
1011 prim1 = inet_ifa_byprefix(in_dev,
1012 ifa1->ifa_address,
1013 ifa1->ifa_mask);
1014 if (!prim1)
1015 continue;
1016 if (prim1->ifa_local != prim->ifa_local)
1017 continue;
1018 }
1019 } else {
1020 if (prim->ifa_local != ifa1->ifa_local)
1021 continue;
1022 prim1 = ifa1;
1023 if (prim != prim1)
1024 same_prefsrc = 1;
1025 }
1026 if (ifa->ifa_local == ifa1->ifa_local)
1027 ok |= LOCAL_OK;
1028 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
1029 ok |= BRD_OK;
1030 if (brd == ifa1->ifa_broadcast)
1031 ok |= BRD1_OK;
1032 if (any == ifa1->ifa_broadcast)
1033 ok |= BRD0_OK;
1034 /* primary has network specific broadcasts */
1035 if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) {
1036 __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask;
1037 __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask;
1038
1039 if (!ipv4_is_zeronet(any1)) {
1040 if (ifa->ifa_broadcast == brd1 ||
1041 ifa->ifa_broadcast == any1)
1042 ok |= BRD_OK;
1043 if (brd == brd1 || brd == any1)
1044 ok |= BRD1_OK;
1045 if (any == brd1 || any == any1)
1046 ok |= BRD0_OK;
1047 }
1048 }
1049 }
1050
1051 no_promotions:
1052 if (!(ok & BRD_OK))
1053 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
1054 if (subnet && ifa->ifa_prefixlen < 31) {
1055 if (!(ok & BRD1_OK))
1056 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
1057 if (!(ok & BRD0_OK))
1058 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
1059 }
1060 if (!(ok & LOCAL_OK)) {
1061 unsigned int addr_type;
1062
1063 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
1064
1065 /* Check, that this local address finally disappeared. */
1066 addr_type = inet_addr_type_dev_table(dev_net(dev), dev,
1067 ifa->ifa_local);
1068 if (gone && addr_type != RTN_LOCAL) {
1069 /* And the last, but not the least thing.
1070 * We must flush stray FIB entries.
1071 *
1072 * First of all, we scan fib_info list searching
1073 * for stray nexthop entries, then ignite fib_flush.
1074 */
1075 if (fib_sync_down_addr(dev, ifa->ifa_local))
1076 fib_flush(dev_net(dev));
1077 }
1078 }
1079 #undef LOCAL_OK
1080 #undef BRD_OK
1081 #undef BRD0_OK
1082 #undef BRD1_OK
1083 }
1084
1085 static void nl_fib_lookup(struct net *net, struct fib_result_nl *frn)
1086 {
1087
1088 struct fib_result res;
1089 struct flowi4 fl4 = {
1090 .flowi4_mark = frn->fl_mark,
1091 .daddr = frn->fl_addr,
1092 .flowi4_tos = frn->fl_tos,
1093 .flowi4_scope = frn->fl_scope,
1094 };
1095 struct fib_table *tb;
1096
1097 rcu_read_lock();
1098
1099 tb = fib_get_table(net, frn->tb_id_in);
1100
1101 frn->err = -ENOENT;
1102 if (tb) {
1103 local_bh_disable();
1104
1105 frn->tb_id = tb->tb_id;
1106 frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF);
1107
1108 if (!frn->err) {
1109 frn->prefixlen = res.prefixlen;
1110 frn->nh_sel = res.nh_sel;
1111 frn->type = res.type;
1112 frn->scope = res.scope;
1113 }
1114 local_bh_enable();
1115 }
1116
1117 rcu_read_unlock();
1118 }
1119
1120 static void nl_fib_input(struct sk_buff *skb)
1121 {
1122 struct net *net;
1123 struct fib_result_nl *frn;
1124 struct nlmsghdr *nlh;
1125 u32 portid;
1126
1127 net = sock_net(skb->sk);
1128 nlh = nlmsg_hdr(skb);
1129 if (skb->len < nlmsg_total_size(sizeof(*frn)) ||
1130 skb->len < nlh->nlmsg_len ||
1131 nlmsg_len(nlh) < sizeof(*frn))
1132 return;
1133
1134 skb = netlink_skb_clone(skb, GFP_KERNEL);
1135 if (!skb)
1136 return;
1137 nlh = nlmsg_hdr(skb);
1138
1139 frn = (struct fib_result_nl *) nlmsg_data(nlh);
1140 nl_fib_lookup(net, frn);
1141
1142 portid = NETLINK_CB(skb).portid; /* netlink portid */
1143 NETLINK_CB(skb).portid = 0; /* from kernel */
1144 NETLINK_CB(skb).dst_group = 0; /* unicast */
1145 netlink_unicast(net->ipv4.fibnl, skb, portid, MSG_DONTWAIT);
1146 }
1147
1148 static int __net_init nl_fib_lookup_init(struct net *net)
1149 {
1150 struct sock *sk;
1151 struct netlink_kernel_cfg cfg = {
1152 .input = nl_fib_input,
1153 };
1154
1155 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, &cfg);
1156 if (!sk)
1157 return -EAFNOSUPPORT;
1158 net->ipv4.fibnl = sk;
1159 return 0;
1160 }
1161
1162 static void nl_fib_lookup_exit(struct net *net)
1163 {
1164 netlink_kernel_release(net->ipv4.fibnl);
1165 net->ipv4.fibnl = NULL;
1166 }
1167
1168 static void fib_disable_ip(struct net_device *dev, unsigned long event,
1169 bool force)
1170 {
1171 if (fib_sync_down_dev(dev, event, force))
1172 fib_flush(dev_net(dev));
1173 else
1174 rt_cache_flush(dev_net(dev));
1175 arp_ifdown(dev);
1176 }
1177
1178 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
1179 {
1180 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
1181 struct net_device *dev = ifa->ifa_dev->dev;
1182 struct net *net = dev_net(dev);
1183
1184 switch (event) {
1185 case NETDEV_UP:
1186 fib_add_ifaddr(ifa);
1187 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1188 fib_sync_up(dev, RTNH_F_DEAD);
1189 #endif
1190 atomic_inc(&net->ipv4.dev_addr_genid);
1191 rt_cache_flush(dev_net(dev));
1192 break;
1193 case NETDEV_DOWN:
1194 fib_del_ifaddr(ifa, NULL);
1195 atomic_inc(&net->ipv4.dev_addr_genid);
1196 if (!ifa->ifa_dev->ifa_list) {
1197 /* Last address was deleted from this interface.
1198 * Disable IP.
1199 */
1200 fib_disable_ip(dev, event, true);
1201 } else {
1202 rt_cache_flush(dev_net(dev));
1203 }
1204 break;
1205 }
1206 return NOTIFY_DONE;
1207 }
1208
1209 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1210 {
1211 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1212 struct netdev_notifier_changeupper_info *info;
1213 struct in_device *in_dev;
1214 struct net *net = dev_net(dev);
1215 unsigned int flags;
1216
1217 if (event == NETDEV_UNREGISTER) {
1218 fib_disable_ip(dev, event, true);
1219 rt_flush_dev(dev);
1220 return NOTIFY_DONE;
1221 }
1222
1223 in_dev = __in_dev_get_rtnl(dev);
1224 if (!in_dev)
1225 return NOTIFY_DONE;
1226
1227 switch (event) {
1228 case NETDEV_UP:
1229 for_ifa(in_dev) {
1230 fib_add_ifaddr(ifa);
1231 } endfor_ifa(in_dev);
1232 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1233 fib_sync_up(dev, RTNH_F_DEAD);
1234 #endif
1235 atomic_inc(&net->ipv4.dev_addr_genid);
1236 rt_cache_flush(net);
1237 break;
1238 case NETDEV_DOWN:
1239 fib_disable_ip(dev, event, false);
1240 break;
1241 case NETDEV_CHANGE:
1242 flags = dev_get_flags(dev);
1243 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1244 fib_sync_up(dev, RTNH_F_LINKDOWN);
1245 else
1246 fib_sync_down_dev(dev, event, false);
1247 /* fall through */
1248 case NETDEV_CHANGEMTU:
1249 rt_cache_flush(net);
1250 break;
1251 case NETDEV_CHANGEUPPER:
1252 info = ptr;
1253 /* flush all routes if dev is linked to or unlinked from
1254 * an L3 master device (e.g., VRF)
1255 */
1256 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
1257 fib_disable_ip(dev, NETDEV_DOWN, true);
1258 break;
1259 }
1260 return NOTIFY_DONE;
1261 }
1262
1263 static struct notifier_block fib_inetaddr_notifier = {
1264 .notifier_call = fib_inetaddr_event,
1265 };
1266
1267 static struct notifier_block fib_netdev_notifier = {
1268 .notifier_call = fib_netdev_event,
1269 };
1270
1271 static int __net_init ip_fib_net_init(struct net *net)
1272 {
1273 int err;
1274 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
1275
1276 err = fib4_notifier_init(net);
1277 if (err)
1278 return err;
1279
1280 /* Avoid false sharing : Use at least a full cache line */
1281 size = max_t(size_t, size, L1_CACHE_BYTES);
1282
1283 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
1284 if (!net->ipv4.fib_table_hash) {
1285 err = -ENOMEM;
1286 goto err_table_hash_alloc;
1287 }
1288
1289 err = fib4_rules_init(net);
1290 if (err < 0)
1291 goto err_rules_init;
1292 return 0;
1293
1294 err_rules_init:
1295 kfree(net->ipv4.fib_table_hash);
1296 err_table_hash_alloc:
1297 fib4_notifier_exit(net);
1298 return err;
1299 }
1300
1301 static void ip_fib_net_exit(struct net *net)
1302 {
1303 int i;
1304
1305 rtnl_lock();
1306 #ifdef CONFIG_IP_MULTIPLE_TABLES
1307 RCU_INIT_POINTER(net->ipv4.fib_main, NULL);
1308 RCU_INIT_POINTER(net->ipv4.fib_default, NULL);
1309 #endif
1310 /* Destroy the tables in reverse order to guarantee that the
1311 * local table, ID 255, is destroyed before the main table, ID
1312 * 254. This is necessary as the local table may contain
1313 * references to data contained in the main table.
1314 */
1315 for (i = FIB_TABLE_HASHSZ - 1; i >= 0; i--) {
1316 struct hlist_head *head = &net->ipv4.fib_table_hash[i];
1317 struct hlist_node *tmp;
1318 struct fib_table *tb;
1319
1320 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) {
1321 hlist_del(&tb->tb_hlist);
1322 fib_table_flush(net, tb);
1323 fib_free_table(tb);
1324 }
1325 }
1326
1327 #ifdef CONFIG_IP_MULTIPLE_TABLES
1328 fib4_rules_exit(net);
1329 #endif
1330 rtnl_unlock();
1331 kfree(net->ipv4.fib_table_hash);
1332 fib4_notifier_exit(net);
1333 }
1334
1335 static int __net_init fib_net_init(struct net *net)
1336 {
1337 int error;
1338
1339 #ifdef CONFIG_IP_ROUTE_CLASSID
1340 net->ipv4.fib_num_tclassid_users = 0;
1341 #endif
1342 error = ip_fib_net_init(net);
1343 if (error < 0)
1344 goto out;
1345 error = nl_fib_lookup_init(net);
1346 if (error < 0)
1347 goto out_nlfl;
1348 error = fib_proc_init(net);
1349 if (error < 0)
1350 goto out_proc;
1351 out:
1352 return error;
1353
1354 out_proc:
1355 nl_fib_lookup_exit(net);
1356 out_nlfl:
1357 ip_fib_net_exit(net);
1358 goto out;
1359 }
1360
1361 static void __net_exit fib_net_exit(struct net *net)
1362 {
1363 fib_proc_exit(net);
1364 nl_fib_lookup_exit(net);
1365 ip_fib_net_exit(net);
1366 }
1367
1368 static struct pernet_operations fib_net_ops = {
1369 .init = fib_net_init,
1370 .exit = fib_net_exit,
1371 };
1372
1373 void __init ip_fib_init(void)
1374 {
1375 fib_trie_init();
1376
1377 register_pernet_subsys(&fib_net_ops);
1378
1379 register_netdevice_notifier(&fib_netdev_notifier);
1380 register_inetaddr_notifier(&fib_inetaddr_notifier);
1381
1382 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, 0);
1383 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, 0);
1384 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, 0);
1385 }