blob: e5b6b0534c5f5e90f9369cafec3c6089d6ebada4 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07008 * 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
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <linux/module.h>
17#include <asm/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018#include <linux/bitops.h>
Randy Dunlap4fc268d2006-01-11 12:17:47 -080019#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#include <linux/types.h>
21#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#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>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020029#include <linux/inetdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070030#include <linux/netdevice.h>
Thomas Graf18237302006-08-04 23:04:54 -070031#include <linux/if_addr.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070032#include <linux/if_arp.h>
33#include <linux/skbuff.h>
David S. Miller7a9bc9b2012-06-29 01:32:45 -070034#include <linux/cache.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070035#include <linux/init.h>
Patrick McHardy1af5a8c2006-08-10 23:10:46 -070036#include <linux/list.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090037#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070038
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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070044#include <net/arp.h>
45#include <net/ip_fib.h>
Thomas Graf63f34442007-03-22 11:55:17 -070046#include <net/rtnetlink.h>
Michael Smith990078a2011-04-07 04:51:51 +000047#include <net/xfrm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070048
Linus Torvalds1da177e2005-04-16 15:20:36 -070049#ifndef CONFIG_IP_MULTIPLE_TABLES
50
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -080051static int __net_init fib4_rules_init(struct net *net)
Pavel Emelyanovc3e9a352007-11-06 23:34:04 -080052{
Denis V. Lunev93456b62008-01-10 03:23:38 -080053 struct fib_table *local_table, *main_table;
54
Alexander Duyck0ddcf432015-03-06 13:47:00 -080055 main_table = fib_trie_table(RT_TABLE_MAIN, NULL);
Denis V. Lunev93456b62008-01-10 03:23:38 -080056 if (main_table == NULL)
Alexander Duyck61f0d862015-03-11 14:02:16 -070057 return -ENOMEM;
Denis V. Lunevdbb50162008-01-10 03:21:49 -080058
Alexander Duyck0ddcf432015-03-06 13:47:00 -080059 local_table = fib_trie_table(RT_TABLE_LOCAL, main_table);
60 if (local_table == NULL)
Alexander Duyck61f0d862015-03-11 14:02:16 -070061 goto fail;
Alexander Duyck0ddcf432015-03-06 13:47:00 -080062
Denis V. Lunev93456b62008-01-10 03:23:38 -080063 hlist_add_head_rcu(&local_table->tb_hlist,
Denis V. Luneve4aef8a2008-01-10 03:28:24 -080064 &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]);
Denis V. Lunev93456b62008-01-10 03:23:38 -080065 hlist_add_head_rcu(&main_table->tb_hlist,
Denis V. Luneve4aef8a2008-01-10 03:28:24 -080066 &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]);
Denis V. Lunevdbb50162008-01-10 03:21:49 -080067 return 0;
68
69fail:
Alexander Duyck61f0d862015-03-11 14:02:16 -070070 fib_free_table(main_table);
Denis V. Lunevdbb50162008-01-10 03:21:49 -080071 return -ENOMEM;
Pavel Emelyanovc3e9a352007-11-06 23:34:04 -080072}
Linus Torvalds1da177e2005-04-16 15:20:36 -070073#else
74
Denis V. Lunev8ad49422008-01-10 03:24:11 -080075struct fib_table *fib_new_table(struct net *net, u32 id)
Linus Torvalds1da177e2005-04-16 15:20:36 -070076{
Alexander Duyck0ddcf432015-03-06 13:47:00 -080077 struct fib_table *tb, *alias = NULL;
Patrick McHardy1af5a8c2006-08-10 23:10:46 -070078 unsigned int h;
Linus Torvalds1da177e2005-04-16 15:20:36 -070079
Patrick McHardy1af5a8c2006-08-10 23:10:46 -070080 if (id == 0)
81 id = RT_TABLE_MAIN;
Denis V. Lunev8ad49422008-01-10 03:24:11 -080082 tb = fib_get_table(net, id);
Patrick McHardy1af5a8c2006-08-10 23:10:46 -070083 if (tb)
84 return tb;
Stephen Hemminger7f9b8052008-01-14 23:14:20 -080085
Alexander Duyck0ddcf432015-03-06 13:47:00 -080086 if (id == RT_TABLE_LOCAL)
87 alias = fib_new_table(net, RT_TABLE_MAIN);
88
89 tb = fib_trie_table(id, alias);
Linus Torvalds1da177e2005-04-16 15:20:36 -070090 if (!tb)
91 return NULL;
David S. Millerf4530fa2012-07-05 22:13:13 -070092
93 switch (id) {
94 case RT_TABLE_LOCAL:
Alexander Duycka7e53532015-03-04 15:02:44 -080095 rcu_assign_pointer(net->ipv4.fib_local, tb);
David S. Millerf4530fa2012-07-05 22:13:13 -070096 break;
David S. Millerf4530fa2012-07-05 22:13:13 -070097 case RT_TABLE_MAIN:
Alexander Duycka7e53532015-03-04 15:02:44 -080098 rcu_assign_pointer(net->ipv4.fib_main, tb);
David S. Millerf4530fa2012-07-05 22:13:13 -070099 break;
David S. Millerf4530fa2012-07-05 22:13:13 -0700100 case RT_TABLE_DEFAULT:
Alexander Duycka7e53532015-03-04 15:02:44 -0800101 rcu_assign_pointer(net->ipv4.fib_default, tb);
David S. Millerf4530fa2012-07-05 22:13:13 -0700102 break;
David S. Millerf4530fa2012-07-05 22:13:13 -0700103 default:
104 break;
105 }
106
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700107 h = id & (FIB_TABLE_HASHSZ - 1);
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800108 hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109 return tb;
110}
111
Alexander Duyck345e9b52014-12-31 10:56:24 -0800112/* caller must hold either rtnl or rcu read lock */
Denis V. Lunev8ad49422008-01-10 03:24:11 -0800113struct fib_table *fib_get_table(struct net *net, u32 id)
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700114{
115 struct fib_table *tb;
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800116 struct hlist_head *head;
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700117 unsigned int h;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700119 if (id == 0)
120 id = RT_TABLE_MAIN;
121 h = id & (FIB_TABLE_HASHSZ - 1);
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800122
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800123 head = &net->ipv4.fib_table_hash[h];
Sasha Levinb67bfe02013-02-27 17:06:00 -0800124 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
Alexander Duyck345e9b52014-12-31 10:56:24 -0800125 if (tb->tb_id == id)
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700126 return tb;
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700127 }
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700128 return NULL;
129}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130#endif /* CONFIG_IP_MULTIPLE_TABLES */
131
Alexander Duyck0ddcf432015-03-06 13:47:00 -0800132static void fib_replace_table(struct net *net, struct fib_table *old,
133 struct fib_table *new)
134{
135#ifdef CONFIG_IP_MULTIPLE_TABLES
136 switch (new->tb_id) {
137 case RT_TABLE_LOCAL:
138 rcu_assign_pointer(net->ipv4.fib_local, new);
139 break;
140 case RT_TABLE_MAIN:
141 rcu_assign_pointer(net->ipv4.fib_main, new);
142 break;
143 case RT_TABLE_DEFAULT:
144 rcu_assign_pointer(net->ipv4.fib_default, new);
145 break;
146 default:
147 break;
148 }
149
150#endif
151 /* replace the old table in the hlist */
152 hlist_replace_rcu(&old->tb_hlist, &new->tb_hlist);
153}
154
155int fib_unmerge(struct net *net)
156{
157 struct fib_table *old, *new;
158
Alexander Duyck3c9e9f72015-03-12 14:46:23 -0700159 /* attempt to fetch local table if it has been allocated */
Alexander Duyck0ddcf432015-03-06 13:47:00 -0800160 old = fib_get_table(net, RT_TABLE_LOCAL);
Alexander Duyck3c9e9f72015-03-12 14:46:23 -0700161 if (!old)
162 return 0;
Alexander Duyck0ddcf432015-03-06 13:47:00 -0800163
Alexander Duyck3c9e9f72015-03-12 14:46:23 -0700164 new = fib_trie_unmerge(old);
Alexander Duyck0ddcf432015-03-06 13:47:00 -0800165 if (!new)
166 return -ENOMEM;
167
168 /* replace merged table with clean table */
169 if (new != old) {
170 fib_replace_table(net, old, new);
171 fib_free_table(old);
172 }
173
174 return 0;
175}
176
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800177static void fib_flush(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178{
179 int flushed = 0;
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700180 unsigned int h;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700182 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
Alexander Duycka7e53532015-03-04 15:02:44 -0800183 struct hlist_head *head = &net->ipv4.fib_table_hash[h];
184 struct hlist_node *tmp;
185 struct fib_table *tb;
186
187 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist)
Stephen Hemminger16c6cf82009-09-20 10:35:36 +0000188 flushed += fib_table_flush(tb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190
191 if (flushed)
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +0000192 rt_cache_flush(net);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193}
194
Scott Feldman104616e2015-03-05 21:21:16 -0800195void fib_flush_external(struct net *net)
196{
197 struct fib_table *tb;
198 struct hlist_head *head;
199 unsigned int h;
200
201 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
202 head = &net->ipv4.fib_table_hash[h];
203 hlist_for_each_entry(tb, head, tb_hlist)
204 fib_table_flush_external(tb);
205 }
206}
207
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800208/*
209 * Find address type as if only "dev" was present in the system. If
210 * on_dev is NULL then all interfaces are taken into consideration.
211 */
Eric Dumazet95c96172012-04-15 05:58:06 +0000212static inline unsigned int __inet_dev_addr_type(struct net *net,
213 const struct net_device *dev,
214 __be32 addr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215{
David S. Miller9ade2282011-03-12 02:02:42 -0500216 struct flowi4 fl4 = { .daddr = addr };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 struct fib_result res;
Eric Dumazet95c96172012-04-15 05:58:06 +0000218 unsigned int ret = RTN_BROADCAST;
Pavel Emelyanov03cf7862007-10-23 21:17:27 -0700219 struct fib_table *local_table;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220
Jan Engelhardt1e637c72008-01-21 03:18:08 -0800221 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222 return RTN_BROADCAST;
Joe Perchesf97c1e02007-12-16 13:45:43 -0800223 if (ipv4_is_multicast(addr))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 return RTN_MULTICAST;
225
Alexander Duyck345e9b52014-12-31 10:56:24 -0800226 rcu_read_lock();
227
Eric W. Biederman6b175b22008-01-10 03:25:28 -0800228 local_table = fib_get_table(net, RT_TABLE_LOCAL);
Pavel Emelyanov03cf7862007-10-23 21:17:27 -0700229 if (local_table) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 ret = RTN_UNICAST;
David S. Miller9ade2282011-03-12 02:02:42 -0500231 if (!fib_table_lookup(local_table, &fl4, &res, FIB_LOOKUP_NOREF)) {
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800232 if (!dev || dev == res.fi->fib_dev)
233 ret = res.type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 }
235 }
Alexander Duyck345e9b52014-12-31 10:56:24 -0800236
237 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 return ret;
239}
240
Eric W. Biederman6b175b22008-01-10 03:25:28 -0800241unsigned int inet_addr_type(struct net *net, __be32 addr)
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800242{
Eric W. Biederman6b175b22008-01-10 03:25:28 -0800243 return __inet_dev_addr_type(net, NULL, addr);
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800244}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000245EXPORT_SYMBOL(inet_addr_type);
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800246
Eric W. Biederman6b175b22008-01-10 03:25:28 -0800247unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
248 __be32 addr)
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800249{
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000250 return __inet_dev_addr_type(net, dev, addr);
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800251}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000252EXPORT_SYMBOL(inet_dev_addr_type);
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800253
David S. Miller35ebf652012-06-28 03:59:11 -0700254__be32 fib_compute_spec_dst(struct sk_buff *skb)
255{
256 struct net_device *dev = skb->dev;
257 struct in_device *in_dev;
258 struct fib_result res;
David S. Millera207a4b2012-06-28 18:33:24 -0700259 struct rtable *rt;
David S. Miller35ebf652012-06-28 03:59:11 -0700260 struct flowi4 fl4;
261 struct net *net;
David S. Millera207a4b2012-06-28 18:33:24 -0700262 int scope;
David S. Miller35ebf652012-06-28 03:59:11 -0700263
David S. Millera207a4b2012-06-28 18:33:24 -0700264 rt = skb_rtable(skb);
Julian Anastasov0cc535a2012-07-18 21:35:03 +0000265 if ((rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | RTCF_LOCAL)) ==
266 RTCF_LOCAL)
David S. Miller35ebf652012-06-28 03:59:11 -0700267 return ip_hdr(skb)->daddr;
268
269 in_dev = __in_dev_get_rcu(dev);
270 BUG_ON(!in_dev);
David S. Miller35ebf652012-06-28 03:59:11 -0700271
272 net = dev_net(dev);
David S. Millera207a4b2012-06-28 18:33:24 -0700273
274 scope = RT_SCOPE_UNIVERSE;
275 if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) {
276 fl4.flowi4_oif = 0;
Pavel Emelyanov1fb94892012-08-08 21:53:36 +0000277 fl4.flowi4_iif = LOOPBACK_IFINDEX;
David S. Millera207a4b2012-06-28 18:33:24 -0700278 fl4.daddr = ip_hdr(skb)->saddr;
279 fl4.saddr = 0;
280 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
281 fl4.flowi4_scope = scope;
282 fl4.flowi4_mark = IN_DEV_SRC_VMARK(in_dev) ? skb->mark : 0;
283 if (!fib_lookup(net, &fl4, &res))
284 return FIB_RES_PREFSRC(net, res);
285 } else {
286 scope = RT_SCOPE_LINK;
287 }
288
289 return inet_select_addr(dev, ip_hdr(skb)->saddr, scope);
David S. Miller35ebf652012-06-28 03:59:11 -0700290}
291
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292/* Given (packet source, input interface) and optional (dst, oif, tos):
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000293 * - (main) check, that source is valid i.e. not broadcast or our local
294 * address.
295 * - figure out what "logical" interface this packet arrived
296 * and calculate "specific destination" address.
297 * - check, that packet arrived from expected physical interface.
Eric Dumazetebc0ffa2010-10-05 10:41:36 +0000298 * called with rcu_read_lock()
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 */
David S. Miller7a9bc9b2012-06-29 01:32:45 -0700300static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
301 u8 tos, int oif, struct net_device *dev,
302 int rpf, struct in_device *idev, u32 *itag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303{
Sébastien Barré1dced6a2014-08-17 09:19:54 +0200304 int ret, no_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 struct fib_result res;
David S. Miller9e56e382012-06-28 18:54:02 -0700306 struct flowi4 fl4;
Denis V. Lunev5b707aa2008-01-21 17:33:15 -0800307 struct net *net;
David S. Miller9e56e382012-06-28 18:54:02 -0700308 bool dev_match;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309
David S. Miller9ade2282011-03-12 02:02:42 -0500310 fl4.flowi4_oif = 0;
Cong Wang6a662712014-04-15 16:25:34 -0700311 fl4.flowi4_iif = oif ? : LOOPBACK_IFINDEX;
David S. Miller9ade2282011-03-12 02:02:42 -0500312 fl4.daddr = src;
313 fl4.saddr = dst;
314 fl4.flowi4_tos = tos;
315 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
David S. Millercc7e17e2011-03-09 20:57:50 -0800316
David S. Miller9e56e382012-06-28 18:54:02 -0700317 no_addr = idev->ifa_list == NULL;
Michael Smith990078a2011-04-07 04:51:51 +0000318
David S. Miller9e56e382012-06-28 18:54:02 -0700319 fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900321 net = dev_net(dev);
David S. Miller9ade2282011-03-12 02:02:42 -0500322 if (fib_lookup(net, &fl4, &res))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323 goto last_resort;
Sébastien Barré1dced6a2014-08-17 09:19:54 +0200324 if (res.type != RTN_UNICAST &&
325 (res.type != RTN_LOCAL || !IN_DEV_ACCEPT_LOCAL(idev)))
326 goto e_inval;
327 if (!rpf && !fib_num_tclassid_users(dev_net(dev)) &&
328 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev)))
329 goto last_resort;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330 fib_combine_itag(itag, &res);
David S. Miller6f86b322010-09-06 22:36:19 -0700331 dev_match = false;
332
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333#ifdef CONFIG_IP_ROUTE_MULTIPATH
David S. Miller6f86b322010-09-06 22:36:19 -0700334 for (ret = 0; ret < res.fi->fib_nhs; ret++) {
335 struct fib_nh *nh = &res.fi->fib_nh[ret];
336
337 if (nh->nh_dev == dev) {
338 dev_match = true;
339 break;
340 }
341 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342#else
343 if (FIB_RES_DEV(res) == dev)
David S. Miller6f86b322010-09-06 22:36:19 -0700344 dev_match = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345#endif
David S. Miller6f86b322010-09-06 22:36:19 -0700346 if (dev_match) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 return ret;
349 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350 if (no_addr)
351 goto last_resort;
Stephen Hemmingerc1cf8422009-02-20 08:25:36 +0000352 if (rpf == 1)
Eric Dumazetb5f7e752010-06-02 12:05:27 +0000353 goto e_rpf;
David S. Miller9ade2282011-03-12 02:02:42 -0500354 fl4.flowi4_oif = dev->ifindex;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355
356 ret = 0;
David S. Miller9ade2282011-03-12 02:02:42 -0500357 if (fib_lookup(net, &fl4, &res) == 0) {
David S. Miller41347dc2012-06-28 04:05:27 -0700358 if (res.type == RTN_UNICAST)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 }
361 return ret;
362
363last_resort:
364 if (rpf)
Eric Dumazetb5f7e752010-06-02 12:05:27 +0000365 goto e_rpf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 *itag = 0;
367 return 0;
368
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369e_inval:
370 return -EINVAL;
Eric Dumazetb5f7e752010-06-02 12:05:27 +0000371e_rpf:
372 return -EXDEV;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373}
374
David S. Miller7a9bc9b2012-06-29 01:32:45 -0700375/* Ignore rp_filter for packets protected by IPsec. */
376int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
377 u8 tos, int oif, struct net_device *dev,
378 struct in_device *idev, u32 *itag)
379{
380 int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev);
381
Julian Anastasove81da0e2012-10-08 11:41:15 +0000382 if (!r && !fib_num_tclassid_users(dev_net(dev)) &&
Sébastien Barré1dced6a2014-08-17 09:19:54 +0200383 IN_DEV_ACCEPT_LOCAL(idev) &&
Julian Anastasove81da0e2012-10-08 11:41:15 +0000384 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) {
David S. Miller7a9bc9b2012-06-29 01:32:45 -0700385 *itag = 0;
386 return 0;
387 }
388 return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag);
389}
390
Al Viro81f7bf62006-09-27 18:40:00 -0700391static inline __be32 sk_extract_addr(struct sockaddr *addr)
Thomas Graf4e902c52006-08-17 18:14:52 -0700392{
393 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
394}
395
396static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
397{
398 struct nlattr *nla;
399
400 nla = (struct nlattr *) ((char *) mx + len);
401 nla->nla_type = type;
402 nla->nla_len = nla_attr_size(4);
403 *(u32 *) nla_data(nla) = value;
404
405 return len + nla_total_size(4);
406}
407
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800408static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
Thomas Graf4e902c52006-08-17 18:14:52 -0700409 struct fib_config *cfg)
410{
Al Viro6d85c102006-09-26 22:15:46 -0700411 __be32 addr;
Thomas Graf4e902c52006-08-17 18:14:52 -0700412 int plen;
413
414 memset(cfg, 0, sizeof(*cfg));
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800415 cfg->fc_nlinfo.nl_net = net;
Thomas Graf4e902c52006-08-17 18:14:52 -0700416
417 if (rt->rt_dst.sa_family != AF_INET)
418 return -EAFNOSUPPORT;
419
420 /*
421 * Check mask for validity:
422 * a) it must be contiguous.
423 * b) destination must have all host bits clear.
424 * c) if application forgot to set correct family (AF_INET),
425 * reject request unless it is absolutely clear i.e.
426 * both family and mask are zero.
427 */
428 plen = 32;
429 addr = sk_extract_addr(&rt->rt_dst);
430 if (!(rt->rt_flags & RTF_HOST)) {
Al Viro81f7bf62006-09-27 18:40:00 -0700431 __be32 mask = sk_extract_addr(&rt->rt_genmask);
Thomas Graf4e902c52006-08-17 18:14:52 -0700432
433 if (rt->rt_genmask.sa_family != AF_INET) {
434 if (mask || rt->rt_genmask.sa_family)
435 return -EAFNOSUPPORT;
436 }
437
438 if (bad_mask(mask, addr))
439 return -EINVAL;
440
441 plen = inet_mask_len(mask);
442 }
443
444 cfg->fc_dst_len = plen;
445 cfg->fc_dst = addr;
446
447 if (cmd != SIOCDELRT) {
448 cfg->fc_nlflags = NLM_F_CREATE;
449 cfg->fc_protocol = RTPROT_BOOT;
450 }
451
452 if (rt->rt_metric)
453 cfg->fc_priority = rt->rt_metric - 1;
454
455 if (rt->rt_flags & RTF_REJECT) {
456 cfg->fc_scope = RT_SCOPE_HOST;
457 cfg->fc_type = RTN_UNREACHABLE;
458 return 0;
459 }
460
461 cfg->fc_scope = RT_SCOPE_NOWHERE;
462 cfg->fc_type = RTN_UNICAST;
463
464 if (rt->rt_dev) {
465 char *colon;
466 struct net_device *dev;
467 char devname[IFNAMSIZ];
468
469 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
470 return -EFAULT;
471
472 devname[IFNAMSIZ-1] = 0;
473 colon = strchr(devname, ':');
474 if (colon)
475 *colon = 0;
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800476 dev = __dev_get_by_name(net, devname);
Thomas Graf4e902c52006-08-17 18:14:52 -0700477 if (!dev)
478 return -ENODEV;
479 cfg->fc_oif = dev->ifindex;
480 if (colon) {
481 struct in_ifaddr *ifa;
482 struct in_device *in_dev = __in_dev_get_rtnl(dev);
483 if (!in_dev)
484 return -ENODEV;
485 *colon = ':';
486 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
487 if (strcmp(ifa->ifa_label, devname) == 0)
488 break;
489 if (ifa == NULL)
490 return -ENODEV;
491 cfg->fc_prefsrc = ifa->ifa_local;
492 }
493 }
494
495 addr = sk_extract_addr(&rt->rt_gateway);
496 if (rt->rt_gateway.sa_family == AF_INET && addr) {
497 cfg->fc_gw = addr;
498 if (rt->rt_flags & RTF_GATEWAY &&
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800499 inet_addr_type(net, addr) == RTN_UNICAST)
Thomas Graf4e902c52006-08-17 18:14:52 -0700500 cfg->fc_scope = RT_SCOPE_UNIVERSE;
501 }
502
503 if (cmd == SIOCDELRT)
504 return 0;
505
506 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
507 return -EINVAL;
508
509 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
510 cfg->fc_scope = RT_SCOPE_LINK;
511
512 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
513 struct nlattr *mx;
514 int len = 0;
515
516 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900517 if (mx == NULL)
Thomas Graf4e902c52006-08-17 18:14:52 -0700518 return -ENOMEM;
519
520 if (rt->rt_flags & RTF_MTU)
521 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
522
523 if (rt->rt_flags & RTF_WINDOW)
524 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
525
526 if (rt->rt_flags & RTF_IRTT)
527 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
528
529 cfg->fc_mx = mx;
530 cfg->fc_mx_len = len;
531 }
532
533 return 0;
534}
535
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536/*
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000537 * Handle IP routing ioctl calls.
538 * These are used to manipulate the routing tables
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 */
Denis V. Lunev1bad1182008-01-10 03:29:53 -0800540int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541{
Thomas Graf4e902c52006-08-17 18:14:52 -0700542 struct fib_config cfg;
543 struct rtentry rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545
546 switch (cmd) {
547 case SIOCADDRT: /* Add a route */
548 case SIOCDELRT: /* Delete a route */
Eric W. Biederman52e804c2012-11-16 03:03:05 +0000549 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 return -EPERM;
Thomas Graf4e902c52006-08-17 18:14:52 -0700551
552 if (copy_from_user(&rt, arg, sizeof(rt)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 return -EFAULT;
Thomas Graf4e902c52006-08-17 18:14:52 -0700554
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 rtnl_lock();
Denis V. Lunev1bad1182008-01-10 03:29:53 -0800556 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 if (err == 0) {
Thomas Graf4e902c52006-08-17 18:14:52 -0700558 struct fib_table *tb;
559
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 if (cmd == SIOCDELRT) {
Denis V. Lunev1bad1182008-01-10 03:29:53 -0800561 tb = fib_get_table(net, cfg.fc_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562 if (tb)
Stephen Hemminger16c6cf82009-09-20 10:35:36 +0000563 err = fib_table_delete(tb, &cfg);
Thomas Graf4e902c52006-08-17 18:14:52 -0700564 else
565 err = -ESRCH;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 } else {
Denis V. Lunev1bad1182008-01-10 03:29:53 -0800567 tb = fib_new_table(net, cfg.fc_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568 if (tb)
Stephen Hemminger16c6cf82009-09-20 10:35:36 +0000569 err = fib_table_insert(tb, &cfg);
Thomas Graf4e902c52006-08-17 18:14:52 -0700570 else
571 err = -ENOBUFS;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 }
Thomas Graf4e902c52006-08-17 18:14:52 -0700573
574 /* allocated by rtentry_to_fib_config() */
575 kfree(cfg.fc_mx);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 }
577 rtnl_unlock();
578 return err;
579 }
580 return -EINVAL;
581}
582
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000583const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
Thomas Graf4e902c52006-08-17 18:14:52 -0700584 [RTA_DST] = { .type = NLA_U32 },
585 [RTA_SRC] = { .type = NLA_U32 },
586 [RTA_IIF] = { .type = NLA_U32 },
587 [RTA_OIF] = { .type = NLA_U32 },
588 [RTA_GATEWAY] = { .type = NLA_U32 },
589 [RTA_PRIORITY] = { .type = NLA_U32 },
590 [RTA_PREFSRC] = { .type = NLA_U32 },
591 [RTA_METRICS] = { .type = NLA_NESTED },
Thomas Graf5176f912006-08-26 20:13:18 -0700592 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
Thomas Graf4e902c52006-08-17 18:14:52 -0700593 [RTA_FLOW] = { .type = NLA_U32 },
Thomas Graf4e902c52006-08-17 18:14:52 -0700594};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800596static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000597 struct nlmsghdr *nlh, struct fib_config *cfg)
Thomas Graf4e902c52006-08-17 18:14:52 -0700598{
599 struct nlattr *attr;
600 int err, remaining;
601 struct rtmsg *rtm;
602
603 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
604 if (err < 0)
605 goto errout;
606
607 memset(cfg, 0, sizeof(*cfg));
608
609 rtm = nlmsg_data(nlh);
Thomas Graf4e902c52006-08-17 18:14:52 -0700610 cfg->fc_dst_len = rtm->rtm_dst_len;
Thomas Graf4e902c52006-08-17 18:14:52 -0700611 cfg->fc_tos = rtm->rtm_tos;
612 cfg->fc_table = rtm->rtm_table;
613 cfg->fc_protocol = rtm->rtm_protocol;
614 cfg->fc_scope = rtm->rtm_scope;
615 cfg->fc_type = rtm->rtm_type;
616 cfg->fc_flags = rtm->rtm_flags;
617 cfg->fc_nlflags = nlh->nlmsg_flags;
618
Eric W. Biederman15e47302012-09-07 20:12:54 +0000619 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
Thomas Graf4e902c52006-08-17 18:14:52 -0700620 cfg->fc_nlinfo.nlh = nlh;
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800621 cfg->fc_nlinfo.nl_net = net;
Thomas Graf4e902c52006-08-17 18:14:52 -0700622
Thomas Grafa0ee18b2007-03-24 20:32:54 -0700623 if (cfg->fc_type > RTN_MAX) {
624 err = -EINVAL;
625 goto errout;
626 }
627
Thomas Graf4e902c52006-08-17 18:14:52 -0700628 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
Thomas Graf8f4c1f92007-09-12 14:44:36 +0200629 switch (nla_type(attr)) {
Thomas Graf4e902c52006-08-17 18:14:52 -0700630 case RTA_DST:
Al Viro17fb2c62006-09-26 22:15:25 -0700631 cfg->fc_dst = nla_get_be32(attr);
Thomas Graf4e902c52006-08-17 18:14:52 -0700632 break;
Thomas Graf4e902c52006-08-17 18:14:52 -0700633 case RTA_OIF:
634 cfg->fc_oif = nla_get_u32(attr);
635 break;
636 case RTA_GATEWAY:
Al Viro17fb2c62006-09-26 22:15:25 -0700637 cfg->fc_gw = nla_get_be32(attr);
Thomas Graf4e902c52006-08-17 18:14:52 -0700638 break;
639 case RTA_PRIORITY:
640 cfg->fc_priority = nla_get_u32(attr);
641 break;
642 case RTA_PREFSRC:
Al Viro17fb2c62006-09-26 22:15:25 -0700643 cfg->fc_prefsrc = nla_get_be32(attr);
Thomas Graf4e902c52006-08-17 18:14:52 -0700644 break;
645 case RTA_METRICS:
646 cfg->fc_mx = nla_data(attr);
647 cfg->fc_mx_len = nla_len(attr);
648 break;
649 case RTA_MULTIPATH:
650 cfg->fc_mp = nla_data(attr);
651 cfg->fc_mp_len = nla_len(attr);
652 break;
653 case RTA_FLOW:
654 cfg->fc_flow = nla_get_u32(attr);
655 break;
Thomas Graf4e902c52006-08-17 18:14:52 -0700656 case RTA_TABLE:
657 cfg->fc_table = nla_get_u32(attr);
658 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 }
660 }
Thomas Graf4e902c52006-08-17 18:14:52 -0700661
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 return 0;
Thomas Graf4e902c52006-08-17 18:14:52 -0700663errout:
664 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665}
666
Thomas Graf661d2962013-03-21 07:45:29 +0000667static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900669 struct net *net = sock_net(skb->sk);
Thomas Graf4e902c52006-08-17 18:14:52 -0700670 struct fib_config cfg;
671 struct fib_table *tb;
672 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800674 err = rtm_to_fib_config(net, skb, nlh, &cfg);
Thomas Graf4e902c52006-08-17 18:14:52 -0700675 if (err < 0)
676 goto errout;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677
Denis V. Lunev8ad49422008-01-10 03:24:11 -0800678 tb = fib_get_table(net, cfg.fc_table);
Thomas Graf4e902c52006-08-17 18:14:52 -0700679 if (tb == NULL) {
680 err = -ESRCH;
681 goto errout;
682 }
683
Stephen Hemminger16c6cf82009-09-20 10:35:36 +0000684 err = fib_table_delete(tb, &cfg);
Thomas Graf4e902c52006-08-17 18:14:52 -0700685errout:
686 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687}
688
Thomas Graf661d2962013-03-21 07:45:29 +0000689static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900691 struct net *net = sock_net(skb->sk);
Thomas Graf4e902c52006-08-17 18:14:52 -0700692 struct fib_config cfg;
693 struct fib_table *tb;
694 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800696 err = rtm_to_fib_config(net, skb, nlh, &cfg);
Thomas Graf4e902c52006-08-17 18:14:52 -0700697 if (err < 0)
698 goto errout;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699
Denis V. Lunev226b0b4a52008-01-10 03:30:24 -0800700 tb = fib_new_table(net, cfg.fc_table);
Thomas Graf4e902c52006-08-17 18:14:52 -0700701 if (tb == NULL) {
702 err = -ENOBUFS;
703 goto errout;
704 }
705
Stephen Hemminger16c6cf82009-09-20 10:35:36 +0000706 err = fib_table_insert(tb, &cfg);
Thomas Graf4e902c52006-08-17 18:14:52 -0700707errout:
708 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709}
710
Thomas Graf63f34442007-03-22 11:55:17 -0700711static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900713 struct net *net = sock_net(skb->sk);
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700714 unsigned int h, s_h;
715 unsigned int e = 0, s_e;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 struct fib_table *tb;
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800717 struct hlist_head *head;
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700718 int dumped = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719
Thomas Grafbe403ea2006-08-17 18:15:17 -0700720 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
721 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
Li RongQing0b8c7f62014-03-21 11:33:10 +0800722 return skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700724 s_h = cb->args[0];
725 s_e = cb->args[1];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726
Alexander Duycka7e53532015-03-04 15:02:44 -0800727 rcu_read_lock();
728
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700729 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
730 e = 0;
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800731 head = &net->ipv4.fib_table_hash[h];
Alexander Duycka7e53532015-03-04 15:02:44 -0800732 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700733 if (e < s_e)
734 goto next;
735 if (dumped)
736 memset(&cb->args[2], 0, sizeof(cb->args) -
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900737 2 * sizeof(cb->args[0]));
Stephen Hemminger16c6cf82009-09-20 10:35:36 +0000738 if (fib_table_dump(tb, skb, cb) < 0)
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700739 goto out;
740 dumped = 1;
741next:
742 e++;
743 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 }
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700745out:
Alexander Duycka7e53532015-03-04 15:02:44 -0800746 rcu_read_unlock();
747
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700748 cb->args[1] = e;
749 cb->args[0] = h;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750
751 return skb->len;
752}
753
754/* Prepare and feed intra-kernel routing request.
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000755 * Really, it should be netlink message, but :-( netlink
756 * can be not configured, so that we feed it directly
757 * to fib engine. It is legal, because all events occur
758 * only when netlink is already locked.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 */
Al Viro81f7bf62006-09-27 18:40:00 -0700760static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761{
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900762 struct net *net = dev_net(ifa->ifa_dev->dev);
Thomas Graf4e902c52006-08-17 18:14:52 -0700763 struct fib_table *tb;
764 struct fib_config cfg = {
765 .fc_protocol = RTPROT_KERNEL,
766 .fc_type = type,
767 .fc_dst = dst,
768 .fc_dst_len = dst_len,
769 .fc_prefsrc = ifa->ifa_local,
770 .fc_oif = ifa->ifa_dev->dev->ifindex,
771 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
Denis V. Lunev4d1169c2008-01-10 03:26:13 -0800772 .fc_nlinfo = {
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800773 .nl_net = net,
Denis V. Lunev4d1169c2008-01-10 03:26:13 -0800774 },
Thomas Graf4e902c52006-08-17 18:14:52 -0700775 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776
777 if (type == RTN_UNICAST)
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800778 tb = fib_new_table(net, RT_TABLE_MAIN);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 else
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800780 tb = fib_new_table(net, RT_TABLE_LOCAL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781
782 if (tb == NULL)
783 return;
784
Thomas Graf4e902c52006-08-17 18:14:52 -0700785 cfg.fc_table = tb->tb_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786
Thomas Graf4e902c52006-08-17 18:14:52 -0700787 if (type != RTN_LOCAL)
788 cfg.fc_scope = RT_SCOPE_LINK;
789 else
790 cfg.fc_scope = RT_SCOPE_HOST;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791
792 if (cmd == RTM_NEWROUTE)
Stephen Hemminger16c6cf82009-09-20 10:35:36 +0000793 fib_table_insert(tb, &cfg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 else
Stephen Hemminger16c6cf82009-09-20 10:35:36 +0000795 fib_table_delete(tb, &cfg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796}
797
Jamal Hadi Salim0ff60a42005-11-22 14:47:37 -0800798void fib_add_ifaddr(struct in_ifaddr *ifa)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799{
800 struct in_device *in_dev = ifa->ifa_dev;
801 struct net_device *dev = in_dev->dev;
802 struct in_ifaddr *prim = ifa;
Al Viroa144ea42006-09-28 18:00:55 -0700803 __be32 mask = ifa->ifa_mask;
804 __be32 addr = ifa->ifa_local;
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000805 __be32 prefix = ifa->ifa_address & mask;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000807 if (ifa->ifa_flags & IFA_F_SECONDARY) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808 prim = inet_ifa_byprefix(in_dev, prefix, mask);
809 if (prim == NULL) {
Joe Perches058bd4d2012-03-11 18:36:11 +0000810 pr_warn("%s: bug: prim == NULL\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811 return;
812 }
813 }
814
815 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
816
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000817 if (!(dev->flags & IFF_UP))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 return;
819
820 /* Add broadcast address, if it is explicitly assigned. */
Al Viroa144ea42006-09-28 18:00:55 -0700821 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
823
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000824 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 (prefix != addr || ifa->ifa_prefixlen < 32)) {
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000826 fib_magic(RTM_NEWROUTE,
827 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
828 prefix, ifa->ifa_prefixlen, prim);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829
830 /* Add network specific broadcasts, when it takes a sense */
831 if (ifa->ifa_prefixlen < 31) {
832 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000833 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
834 32, prim);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 }
836 }
837}
838
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000839/* Delete primary or secondary address.
840 * Optionally, on secondary address promotion consider the addresses
841 * from subnet iprim as deleted, even if they are in device list.
842 * In this case the secondary ifa can be in device list.
843 */
844void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845{
846 struct in_device *in_dev = ifa->ifa_dev;
847 struct net_device *dev = in_dev->dev;
848 struct in_ifaddr *ifa1;
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000849 struct in_ifaddr *prim = ifa, *prim1 = NULL;
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000850 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
851 __be32 any = ifa->ifa_address & ifa->ifa_mask;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852#define LOCAL_OK 1
853#define BRD_OK 2
854#define BRD0_OK 4
855#define BRD1_OK 8
Eric Dumazet95c96172012-04-15 05:58:06 +0000856 unsigned int ok = 0;
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000857 int subnet = 0; /* Primary network */
858 int gone = 1; /* Address is missing */
859 int same_prefsrc = 0; /* Another primary with same IP */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000861 if (ifa->ifa_flags & IFA_F_SECONDARY) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
863 if (prim == NULL) {
Joe Perches058bd4d2012-03-11 18:36:11 +0000864 pr_warn("%s: bug: prim == NULL\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 return;
866 }
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000867 if (iprim && iprim != prim) {
Joe Perches058bd4d2012-03-11 18:36:11 +0000868 pr_warn("%s: bug: iprim != prim\n", __func__);
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000869 return;
870 }
871 } else if (!ipv4_is_zeronet(any) &&
872 (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) {
873 fib_magic(RTM_DELROUTE,
874 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
875 any, ifa->ifa_prefixlen, prim);
876 subnet = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 }
878
879 /* Deletion is more complicated than add.
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000880 * We should take care of not to delete too much :-)
881 *
882 * Scan address list to be sure that addresses are really gone.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 */
884
885 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000886 if (ifa1 == ifa) {
887 /* promotion, keep the IP */
888 gone = 0;
889 continue;
890 }
891 /* Ignore IFAs from our subnet */
892 if (iprim && ifa1->ifa_mask == iprim->ifa_mask &&
893 inet_ifa_match(ifa1->ifa_address, iprim))
894 continue;
895
896 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
897 if (ifa1->ifa_flags & IFA_F_SECONDARY) {
898 /* Another address from our subnet? */
899 if (ifa1->ifa_mask == prim->ifa_mask &&
900 inet_ifa_match(ifa1->ifa_address, prim))
901 prim1 = prim;
902 else {
903 /* We reached the secondaries, so
904 * same_prefsrc should be determined.
905 */
906 if (!same_prefsrc)
907 continue;
908 /* Search new prim1 if ifa1 is not
909 * using the current prim1
910 */
911 if (!prim1 ||
912 ifa1->ifa_mask != prim1->ifa_mask ||
913 !inet_ifa_match(ifa1->ifa_address, prim1))
914 prim1 = inet_ifa_byprefix(in_dev,
915 ifa1->ifa_address,
916 ifa1->ifa_mask);
917 if (!prim1)
918 continue;
919 if (prim1->ifa_local != prim->ifa_local)
920 continue;
921 }
922 } else {
923 if (prim->ifa_local != ifa1->ifa_local)
924 continue;
925 prim1 = ifa1;
926 if (prim != prim1)
927 same_prefsrc = 1;
928 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929 if (ifa->ifa_local == ifa1->ifa_local)
930 ok |= LOCAL_OK;
931 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
932 ok |= BRD_OK;
933 if (brd == ifa1->ifa_broadcast)
934 ok |= BRD1_OK;
935 if (any == ifa1->ifa_broadcast)
936 ok |= BRD0_OK;
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000937 /* primary has network specific broadcasts */
938 if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) {
939 __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask;
940 __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask;
941
942 if (!ipv4_is_zeronet(any1)) {
943 if (ifa->ifa_broadcast == brd1 ||
944 ifa->ifa_broadcast == any1)
945 ok |= BRD_OK;
946 if (brd == brd1 || brd == any1)
947 ok |= BRD1_OK;
948 if (any == brd1 || any == any1)
949 ok |= BRD0_OK;
950 }
951 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952 }
953
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000954 if (!(ok & BRD_OK))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000956 if (subnet && ifa->ifa_prefixlen < 31) {
957 if (!(ok & BRD1_OK))
958 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
959 if (!(ok & BRD0_OK))
960 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
961 }
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000962 if (!(ok & LOCAL_OK)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
964
965 /* Check, that this local address finally disappeared. */
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000966 if (gone &&
967 inet_addr_type(dev_net(dev), ifa->ifa_local) != RTN_LOCAL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 /* And the last, but not the least thing.
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000969 * We must flush stray FIB entries.
970 *
971 * First of all, we scan fib_info list searching
972 * for stray nexthop entries, then ignite fib_flush.
973 */
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900974 if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local))
975 fib_flush(dev_net(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976 }
977 }
978#undef LOCAL_OK
979#undef BRD_OK
980#undef BRD0_OK
981#undef BRD1_OK
982}
983
Alexander Duyck345e9b52014-12-31 10:56:24 -0800984static void nl_fib_lookup(struct net *net, struct fib_result_nl *frn)
Robert Olsson246955f2005-06-20 13:36:39 -0700985{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900986
Robert Olsson246955f2005-06-20 13:36:39 -0700987 struct fib_result res;
David S. Miller9ade2282011-03-12 02:02:42 -0500988 struct flowi4 fl4 = {
989 .flowi4_mark = frn->fl_mark,
990 .daddr = frn->fl_addr,
991 .flowi4_tos = frn->fl_tos,
992 .flowi4_scope = frn->fl_scope,
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000993 };
Alexander Duyck345e9b52014-12-31 10:56:24 -0800994 struct fib_table *tb;
995
996 rcu_read_lock();
997
998 tb = fib_get_table(net, frn->tb_id_in);
Alexey Kuznetsov1194ed02007-04-25 13:07:28 -0700999
1000 frn->err = -ENOENT;
Robert Olsson246955f2005-06-20 13:36:39 -07001001 if (tb) {
1002 local_bh_disable();
1003
1004 frn->tb_id = tb->tb_id;
David S. Miller9ade2282011-03-12 02:02:42 -05001005 frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF);
Robert Olsson246955f2005-06-20 13:36:39 -07001006
1007 if (!frn->err) {
1008 frn->prefixlen = res.prefixlen;
1009 frn->nh_sel = res.nh_sel;
1010 frn->type = res.type;
1011 frn->scope = res.scope;
1012 }
1013 local_bh_enable();
1014 }
Alexander Duyck345e9b52014-12-31 10:56:24 -08001015
1016 rcu_read_unlock();
Robert Olsson246955f2005-06-20 13:36:39 -07001017}
1018
David S. Miller28f7b0362007-10-10 21:32:39 -07001019static void nl_fib_input(struct sk_buff *skb)
Robert Olsson246955f2005-06-20 13:36:39 -07001020{
Denis V. Lunev6bd48fc2008-01-10 03:28:55 -08001021 struct net *net;
Robert Olsson246955f2005-06-20 13:36:39 -07001022 struct fib_result_nl *frn;
David S. Miller28f7b0362007-10-10 21:32:39 -07001023 struct nlmsghdr *nlh;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001024 u32 portid;
Alexey Kuznetsov1194ed02007-04-25 13:07:28 -07001025
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001026 net = sock_net(skb->sk);
Arnaldo Carvalho de Melob529ccf2007-04-25 19:08:35 -07001027 nlh = nlmsg_hdr(skb);
Hong zhi guo573ce262013-03-27 06:47:04 +00001028 if (skb->len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len ||
1029 nlmsg_len(nlh) < sizeof(*frn))
Thomas Grafea865752005-12-01 14:30:00 -08001030 return;
Denis V. Lunevd883a032007-12-21 02:01:53 -08001031
Pablo Neira3a365152013-06-28 03:04:23 +02001032 skb = netlink_skb_clone(skb, GFP_KERNEL);
Denis V. Lunevd883a032007-12-21 02:01:53 -08001033 if (skb == NULL)
1034 return;
1035 nlh = nlmsg_hdr(skb);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001036
Hong zhi guo573ce262013-03-27 06:47:04 +00001037 frn = (struct fib_result_nl *) nlmsg_data(nlh);
Alexander Duyck345e9b52014-12-31 10:56:24 -08001038 nl_fib_lookup(net, frn);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001039
Rami Rosen28a28282013-01-11 09:59:20 +00001040 portid = NETLINK_CB(skb).portid; /* netlink portid */
Eric W. Biederman15e47302012-09-07 20:12:54 +00001041 NETLINK_CB(skb).portid = 0; /* from kernel */
Patrick McHardyac6d4392005-08-14 19:29:52 -07001042 NETLINK_CB(skb).dst_group = 0; /* unicast */
Eric W. Biederman15e47302012-09-07 20:12:54 +00001043 netlink_unicast(net->ipv4.fibnl, skb, portid, MSG_DONTWAIT);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001044}
Robert Olsson246955f2005-06-20 13:36:39 -07001045
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00001046static int __net_init nl_fib_lookup_init(struct net *net)
Robert Olsson246955f2005-06-20 13:36:39 -07001047{
Denis V. Lunev6bd48fc2008-01-10 03:28:55 -08001048 struct sock *sk;
Pablo Neira Ayusoa31f2d12012-06-29 06:15:21 +00001049 struct netlink_kernel_cfg cfg = {
1050 .input = nl_fib_input,
1051 };
1052
Pablo Neira Ayuso9f00d972012-09-08 02:53:54 +00001053 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, &cfg);
Denis V. Lunev6bd48fc2008-01-10 03:28:55 -08001054 if (sk == NULL)
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001055 return -EAFNOSUPPORT;
Denis V. Lunev6bd48fc2008-01-10 03:28:55 -08001056 net->ipv4.fibnl = sk;
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001057 return 0;
1058}
1059
1060static void nl_fib_lookup_exit(struct net *net)
1061{
Denis V. Lunevb7c6ba62008-01-28 14:41:19 -08001062 netlink_kernel_release(net->ipv4.fibnl);
Denis V. Lunev775516b2008-01-18 23:55:19 -08001063 net->ipv4.fibnl = NULL;
Robert Olsson246955f2005-06-20 13:36:39 -07001064}
1065
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001066static void fib_disable_ip(struct net_device *dev, int force)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067{
Denis V. Lunev85326fa2008-01-31 18:48:47 -08001068 if (fib_sync_down_dev(dev, force))
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001069 fib_flush(dev_net(dev));
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001070 rt_cache_flush(dev_net(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 arp_ifdown(dev);
1072}
1073
1074static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
1075{
Jianjun Kong6ed25332008-11-03 00:25:16 -08001076 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
Denis V. Lunev76e6ebf2008-07-05 19:00:44 -07001077 struct net_device *dev = ifa->ifa_dev->dev;
David S. Miller436c3b62011-03-24 17:42:21 -07001078 struct net *net = dev_net(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079
1080 switch (event) {
1081 case NETDEV_UP:
1082 fib_add_ifaddr(ifa);
1083#ifdef CONFIG_IP_ROUTE_MULTIPATH
Denis V. Lunev76e6ebf2008-07-05 19:00:44 -07001084 fib_sync_up(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085#endif
David S. Miller436c3b62011-03-24 17:42:21 -07001086 atomic_inc(&net->ipv4.dev_addr_genid);
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001087 rt_cache_flush(dev_net(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088 break;
1089 case NETDEV_DOWN:
Julian Anastasove6abbaa2011-03-19 12:13:49 +00001090 fib_del_ifaddr(ifa, NULL);
David S. Miller436c3b62011-03-24 17:42:21 -07001091 atomic_inc(&net->ipv4.dev_addr_genid);
Jayachandran C9fcc2e82005-10-27 15:10:01 -07001092 if (ifa->ifa_dev->ifa_list == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093 /* Last address was deleted from this interface.
Eric Dumazet6a31d2a2010-10-04 20:00:18 +00001094 * Disable IP.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095 */
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001096 fib_disable_ip(dev, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097 } else {
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001098 rt_cache_flush(dev_net(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099 }
1100 break;
1101 }
1102 return NOTIFY_DONE;
1103}
1104
1105static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1106{
Jiri Pirko351638e2013-05-28 01:30:21 +00001107 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
Eric Dumazet0115e8e2012-08-22 17:19:46 +00001108 struct in_device *in_dev;
David S. Miller436c3b62011-03-24 17:42:21 -07001109 struct net *net = dev_net(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110
1111 if (event == NETDEV_UNREGISTER) {
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001112 fib_disable_ip(dev, 2);
David S. Millercaacf052012-07-31 15:06:50 -07001113 rt_flush_dev(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114 return NOTIFY_DONE;
1115 }
1116
Eric Dumazet0115e8e2012-08-22 17:19:46 +00001117 in_dev = __in_dev_get_rtnl(dev);
Oliver Hartkoppa0065f22014-01-23 10:19:34 +01001118 if (!in_dev)
1119 return NOTIFY_DONE;
Eric Dumazet0115e8e2012-08-22 17:19:46 +00001120
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 switch (event) {
1122 case NETDEV_UP:
1123 for_ifa(in_dev) {
1124 fib_add_ifaddr(ifa);
1125 } endfor_ifa(in_dev);
1126#ifdef CONFIG_IP_ROUTE_MULTIPATH
1127 fib_sync_up(dev);
1128#endif
David S. Miller436c3b62011-03-24 17:42:21 -07001129 atomic_inc(&net->ipv4.dev_addr_genid);
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001130 rt_cache_flush(net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 break;
1132 case NETDEV_DOWN:
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001133 fib_disable_ip(dev, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134 break;
1135 case NETDEV_CHANGEMTU:
1136 case NETDEV_CHANGE:
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001137 rt_cache_flush(net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138 break;
1139 }
1140 return NOTIFY_DONE;
1141}
1142
1143static struct notifier_block fib_inetaddr_notifier = {
Jianjun Kong6ed25332008-11-03 00:25:16 -08001144 .notifier_call = fib_inetaddr_event,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145};
1146
1147static struct notifier_block fib_netdev_notifier = {
Jianjun Kong6ed25332008-11-03 00:25:16 -08001148 .notifier_call = fib_netdev_event,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149};
1150
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001151static int __net_init ip_fib_net_init(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152{
Denis V. Lunevdce5cbe2008-01-31 18:44:53 -08001153 int err;
Eric Dumazet10da66f2010-10-13 08:22:03 +00001154 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
Patrick McHardy1af5a8c2006-08-10 23:10:46 -07001155
Eric Dumazet10da66f2010-10-13 08:22:03 +00001156 /* Avoid false sharing : Use at least a full cache line */
1157 size = max_t(size_t, size, L1_CACHE_BYTES);
1158
1159 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
Denis V. Luneve4aef8a2008-01-10 03:28:24 -08001160 if (net->ipv4.fib_table_hash == NULL)
1161 return -ENOMEM;
1162
Denis V. Lunevdce5cbe2008-01-31 18:44:53 -08001163 err = fib4_rules_init(net);
1164 if (err < 0)
1165 goto fail;
1166 return 0;
1167
1168fail:
1169 kfree(net->ipv4.fib_table_hash);
1170 return err;
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001171}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00001173static void ip_fib_net_exit(struct net *net)
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001174{
1175 unsigned int i;
Thomas Graf63f34442007-03-22 11:55:17 -07001176
Sabrina Dubroca6dede752015-03-10 21:03:54 +01001177 rtnl_lock();
1178
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001179#ifdef CONFIG_IP_MULTIPLE_TABLES
1180 fib4_rules_exit(net);
1181#endif
1182
1183 for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
Alexander Duycka7e53532015-03-04 15:02:44 -08001184 struct hlist_head *head = &net->ipv4.fib_table_hash[i];
Sasha Levinb67bfe02013-02-27 17:06:00 -08001185 struct hlist_node *tmp;
Alexander Duycka7e53532015-03-04 15:02:44 -08001186 struct fib_table *tb;
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001187
Alexander Duycka7e53532015-03-04 15:02:44 -08001188 /* this is done in two passes as flushing the table could
1189 * cause it to be reallocated in order to accommodate new
1190 * tnodes at the root as the table shrinks.
1191 */
1192 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist)
Stephen Hemminger16c6cf82009-09-20 10:35:36 +00001193 fib_table_flush(tb);
Alexander Duycka7e53532015-03-04 15:02:44 -08001194
1195 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) {
1196#ifdef CONFIG_IP_MULTIPLE_TABLES
1197 switch (tb->tb_id) {
1198 case RT_TABLE_LOCAL:
1199 RCU_INIT_POINTER(net->ipv4.fib_local, NULL);
1200 break;
1201 case RT_TABLE_MAIN:
1202 RCU_INIT_POINTER(net->ipv4.fib_main, NULL);
1203 break;
1204 case RT_TABLE_DEFAULT:
1205 RCU_INIT_POINTER(net->ipv4.fib_default, NULL);
1206 break;
1207 default:
1208 break;
1209 }
1210#endif
1211 hlist_del(&tb->tb_hlist);
Pavel Emelyanov4aa2c462010-10-28 02:00:43 +00001212 fib_free_table(tb);
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001213 }
1214 }
Eric Dumazete2666f82011-03-30 16:57:46 -07001215 rtnl_unlock();
Denis V. Luneve4aef8a2008-01-10 03:28:24 -08001216 kfree(net->ipv4.fib_table_hash);
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001217}
1218
1219static int __net_init fib_net_init(struct net *net)
1220{
1221 int error;
1222
David S. Millerf4530fa2012-07-05 22:13:13 -07001223#ifdef CONFIG_IP_ROUTE_CLASSID
1224 net->ipv4.fib_num_tclassid_users = 0;
1225#endif
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001226 error = ip_fib_net_init(net);
1227 if (error < 0)
1228 goto out;
1229 error = nl_fib_lookup_init(net);
1230 if (error < 0)
1231 goto out_nlfl;
1232 error = fib_proc_init(net);
1233 if (error < 0)
1234 goto out_proc;
1235out:
1236 return error;
1237
1238out_proc:
1239 nl_fib_lookup_exit(net);
1240out_nlfl:
1241 ip_fib_net_exit(net);
1242 goto out;
1243}
1244
1245static void __net_exit fib_net_exit(struct net *net)
1246{
1247 fib_proc_exit(net);
1248 nl_fib_lookup_exit(net);
1249 ip_fib_net_exit(net);
1250}
1251
1252static struct pernet_operations fib_net_ops = {
1253 .init = fib_net_init,
1254 .exit = fib_net_exit,
1255};
1256
1257void __init ip_fib_init(void)
1258{
Greg Rosec7ac8672011-06-10 01:27:09 +00001259 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, NULL);
1260 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, NULL);
1261 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, NULL);
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001262
1263 register_pernet_subsys(&fib_net_ops);
1264 register_netdevice_notifier(&fib_netdev_notifier);
1265 register_inetaddr_notifier(&fib_inetaddr_notifier);
Stephen Hemminger7f9b8052008-01-14 23:14:20 -08001266
David S. Miller5348ba82011-02-01 15:30:56 -08001267 fib_trie_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268}