blob: 47344759306011b5bb940e9cab9237b9e9f624b4 [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>
David Ahern385add92015-09-29 20:07:13 -070048#include <net/l3mdev.h>
David Ahernf6d3c192015-08-28 08:42:09 -070049#include <trace/events/fib.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070050
Linus Torvalds1da177e2005-04-16 15:20:36 -070051#ifndef CONFIG_IP_MULTIPLE_TABLES
52
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -080053static int __net_init fib4_rules_init(struct net *net)
Pavel Emelyanovc3e9a352007-11-06 23:34:04 -080054{
Denis V. Lunev93456b62008-01-10 03:23:38 -080055 struct fib_table *local_table, *main_table;
56
Alexander Duyck0ddcf432015-03-06 13:47:00 -080057 main_table = fib_trie_table(RT_TABLE_MAIN, NULL);
Ian Morris51456b22015-04-03 09:17:26 +010058 if (!main_table)
Alexander Duyck61f0d862015-03-11 14:02:16 -070059 return -ENOMEM;
Denis V. Lunevdbb50162008-01-10 03:21:49 -080060
Alexander Duyck0ddcf432015-03-06 13:47:00 -080061 local_table = fib_trie_table(RT_TABLE_LOCAL, main_table);
Ian Morris51456b22015-04-03 09:17:26 +010062 if (!local_table)
Alexander Duyck61f0d862015-03-11 14:02:16 -070063 goto fail;
Alexander Duyck0ddcf432015-03-06 13:47:00 -080064
Denis V. Lunev93456b62008-01-10 03:23:38 -080065 hlist_add_head_rcu(&local_table->tb_hlist,
Denis V. Luneve4aef8a2008-01-10 03:28:24 -080066 &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]);
Denis V. Lunev93456b62008-01-10 03:23:38 -080067 hlist_add_head_rcu(&main_table->tb_hlist,
Denis V. Luneve4aef8a2008-01-10 03:28:24 -080068 &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]);
Denis V. Lunevdbb50162008-01-10 03:21:49 -080069 return 0;
70
71fail:
Alexander Duyck61f0d862015-03-11 14:02:16 -070072 fib_free_table(main_table);
Denis V. Lunevdbb50162008-01-10 03:21:49 -080073 return -ENOMEM;
Pavel Emelyanovc3e9a352007-11-06 23:34:04 -080074}
Linus Torvalds1da177e2005-04-16 15:20:36 -070075#else
76
Denis V. Lunev8ad49422008-01-10 03:24:11 -080077struct fib_table *fib_new_table(struct net *net, u32 id)
Linus Torvalds1da177e2005-04-16 15:20:36 -070078{
Alexander Duyck0ddcf432015-03-06 13:47:00 -080079 struct fib_table *tb, *alias = NULL;
Patrick McHardy1af5a8c2006-08-10 23:10:46 -070080 unsigned int h;
Linus Torvalds1da177e2005-04-16 15:20:36 -070081
Patrick McHardy1af5a8c2006-08-10 23:10:46 -070082 if (id == 0)
83 id = RT_TABLE_MAIN;
Denis V. Lunev8ad49422008-01-10 03:24:11 -080084 tb = fib_get_table(net, id);
Patrick McHardy1af5a8c2006-08-10 23:10:46 -070085 if (tb)
86 return tb;
Stephen Hemminger7f9b8052008-01-14 23:14:20 -080087
Alexander Duyck0ddcf432015-03-06 13:47:00 -080088 if (id == RT_TABLE_LOCAL)
89 alias = fib_new_table(net, RT_TABLE_MAIN);
90
91 tb = fib_trie_table(id, alias);
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 if (!tb)
93 return NULL;
David S. Millerf4530fa2012-07-05 22:13:13 -070094
95 switch (id) {
96 case RT_TABLE_LOCAL:
Alexander Duycka7e53532015-03-04 15:02:44 -080097 rcu_assign_pointer(net->ipv4.fib_local, tb);
David S. Millerf4530fa2012-07-05 22:13:13 -070098 break;
David S. Millerf4530fa2012-07-05 22:13:13 -070099 case RT_TABLE_MAIN:
Alexander Duycka7e53532015-03-04 15:02:44 -0800100 rcu_assign_pointer(net->ipv4.fib_main, tb);
David S. Millerf4530fa2012-07-05 22:13:13 -0700101 break;
David S. Millerf4530fa2012-07-05 22:13:13 -0700102 case RT_TABLE_DEFAULT:
Alexander Duycka7e53532015-03-04 15:02:44 -0800103 rcu_assign_pointer(net->ipv4.fib_default, tb);
David S. Millerf4530fa2012-07-05 22:13:13 -0700104 break;
David S. Millerf4530fa2012-07-05 22:13:13 -0700105 default:
106 break;
107 }
108
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700109 h = id & (FIB_TABLE_HASHSZ - 1);
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800110 hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111 return tb;
112}
113
Alexander Duyck345e9b52014-12-31 10:56:24 -0800114/* caller must hold either rtnl or rcu read lock */
Denis V. Lunev8ad49422008-01-10 03:24:11 -0800115struct fib_table *fib_get_table(struct net *net, u32 id)
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700116{
117 struct fib_table *tb;
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800118 struct hlist_head *head;
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700119 unsigned int h;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700121 if (id == 0)
122 id = RT_TABLE_MAIN;
123 h = id & (FIB_TABLE_HASHSZ - 1);
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800124
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800125 head = &net->ipv4.fib_table_hash[h];
Sasha Levinb67bfe02013-02-27 17:06:00 -0800126 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
Alexander Duyck345e9b52014-12-31 10:56:24 -0800127 if (tb->tb_id == id)
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700128 return tb;
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700129 }
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700130 return NULL;
131}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132#endif /* CONFIG_IP_MULTIPLE_TABLES */
133
Alexander Duyck0ddcf432015-03-06 13:47:00 -0800134static void fib_replace_table(struct net *net, struct fib_table *old,
135 struct fib_table *new)
136{
137#ifdef CONFIG_IP_MULTIPLE_TABLES
138 switch (new->tb_id) {
139 case RT_TABLE_LOCAL:
140 rcu_assign_pointer(net->ipv4.fib_local, new);
141 break;
142 case RT_TABLE_MAIN:
143 rcu_assign_pointer(net->ipv4.fib_main, new);
144 break;
145 case RT_TABLE_DEFAULT:
146 rcu_assign_pointer(net->ipv4.fib_default, new);
147 break;
148 default:
149 break;
150 }
151
152#endif
153 /* replace the old table in the hlist */
154 hlist_replace_rcu(&old->tb_hlist, &new->tb_hlist);
155}
156
157int fib_unmerge(struct net *net)
158{
159 struct fib_table *old, *new;
160
Alexander Duyck3c9e9f72015-03-12 14:46:23 -0700161 /* attempt to fetch local table if it has been allocated */
Alexander Duyck0ddcf432015-03-06 13:47:00 -0800162 old = fib_get_table(net, RT_TABLE_LOCAL);
Alexander Duyck3c9e9f72015-03-12 14:46:23 -0700163 if (!old)
164 return 0;
Alexander Duyck0ddcf432015-03-06 13:47:00 -0800165
Alexander Duyck3c9e9f72015-03-12 14:46:23 -0700166 new = fib_trie_unmerge(old);
Alexander Duyck0ddcf432015-03-06 13:47:00 -0800167 if (!new)
168 return -ENOMEM;
169
170 /* replace merged table with clean table */
171 if (new != old) {
172 fib_replace_table(net, old, new);
173 fib_free_table(old);
174 }
175
176 return 0;
177}
178
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800179static void fib_flush(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180{
181 int flushed = 0;
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700182 unsigned int h;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700184 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
Alexander Duycka7e53532015-03-04 15:02:44 -0800185 struct hlist_head *head = &net->ipv4.fib_table_hash[h];
186 struct hlist_node *tmp;
187 struct fib_table *tb;
188
189 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist)
Stephen Hemminger16c6cf82009-09-20 10:35:36 +0000190 flushed += fib_table_flush(tb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192
193 if (flushed)
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +0000194 rt_cache_flush(net);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195}
196
Scott Feldman104616e2015-03-05 21:21:16 -0800197void fib_flush_external(struct net *net)
198{
199 struct fib_table *tb;
200 struct hlist_head *head;
201 unsigned int h;
202
203 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
204 head = &net->ipv4.fib_table_hash[h];
205 hlist_for_each_entry(tb, head, tb_hlist)
206 fib_table_flush_external(tb);
207 }
208}
209
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800210/*
211 * Find address type as if only "dev" was present in the system. If
212 * on_dev is NULL then all interfaces are taken into consideration.
213 */
Eric Dumazet95c96172012-04-15 05:58:06 +0000214static inline unsigned int __inet_dev_addr_type(struct net *net,
215 const struct net_device *dev,
David Ahern9b8ff512015-09-01 14:26:35 -0600216 __be32 addr, u32 tb_id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217{
David S. Miller9ade2282011-03-12 02:02:42 -0500218 struct flowi4 fl4 = { .daddr = addr };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 struct fib_result res;
Eric Dumazet95c96172012-04-15 05:58:06 +0000220 unsigned int ret = RTN_BROADCAST;
David Ahern15be405e2015-08-13 14:59:04 -0600221 struct fib_table *table;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222
Jan Engelhardt1e637c72008-01-21 03:18:08 -0800223 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 return RTN_BROADCAST;
Joe Perchesf97c1e02007-12-16 13:45:43 -0800225 if (ipv4_is_multicast(addr))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226 return RTN_MULTICAST;
227
Alexander Duyck345e9b52014-12-31 10:56:24 -0800228 rcu_read_lock();
229
David Ahern15be405e2015-08-13 14:59:04 -0600230 table = fib_get_table(net, tb_id);
231 if (table) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 ret = RTN_UNICAST;
David Ahern15be405e2015-08-13 14:59:04 -0600233 if (!fib_table_lookup(table, &fl4, &res, FIB_LOOKUP_NOREF)) {
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800234 if (!dev || dev == res.fi->fib_dev)
235 ret = res.type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 }
237 }
Alexander Duyck345e9b52014-12-31 10:56:24 -0800238
239 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240 return ret;
241}
242
David Ahern9b8ff512015-09-01 14:26:35 -0600243unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id)
David Ahern15be405e2015-08-13 14:59:04 -0600244{
245 return __inet_dev_addr_type(net, NULL, addr, tb_id);
246}
247EXPORT_SYMBOL(inet_addr_type_table);
248
Eric W. Biederman6b175b22008-01-10 03:25:28 -0800249unsigned int inet_addr_type(struct net *net, __be32 addr)
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800250{
David Ahern15be405e2015-08-13 14:59:04 -0600251 return __inet_dev_addr_type(net, NULL, addr, RT_TABLE_LOCAL);
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800252}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000253EXPORT_SYMBOL(inet_addr_type);
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800254
Eric W. Biederman6b175b22008-01-10 03:25:28 -0800255unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
256 __be32 addr)
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800257{
David Ahern3236b002015-09-29 20:07:14 -0700258 u32 rt_table = l3mdev_fib_table(dev) ? : RT_TABLE_LOCAL;
David Ahern15be405e2015-08-13 14:59:04 -0600259
260 return __inet_dev_addr_type(net, dev, addr, rt_table);
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800261}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000262EXPORT_SYMBOL(inet_dev_addr_type);
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800263
David Ahern30bbaa12015-08-13 14:59:05 -0600264/* inet_addr_type with dev == NULL but using the table from a dev
265 * if one is associated
266 */
267unsigned int inet_addr_type_dev_table(struct net *net,
268 const struct net_device *dev,
269 __be32 addr)
270{
David Ahern3236b002015-09-29 20:07:14 -0700271 u32 rt_table = l3mdev_fib_table(dev) ? : RT_TABLE_LOCAL;
David Ahern30bbaa12015-08-13 14:59:05 -0600272
273 return __inet_dev_addr_type(net, NULL, addr, rt_table);
274}
275EXPORT_SYMBOL(inet_addr_type_dev_table);
276
David S. Miller35ebf652012-06-28 03:59:11 -0700277__be32 fib_compute_spec_dst(struct sk_buff *skb)
278{
279 struct net_device *dev = skb->dev;
280 struct in_device *in_dev;
281 struct fib_result res;
David S. Millera207a4b2012-06-28 18:33:24 -0700282 struct rtable *rt;
David S. Miller35ebf652012-06-28 03:59:11 -0700283 struct flowi4 fl4;
284 struct net *net;
David S. Millera207a4b2012-06-28 18:33:24 -0700285 int scope;
David S. Miller35ebf652012-06-28 03:59:11 -0700286
David S. Millera207a4b2012-06-28 18:33:24 -0700287 rt = skb_rtable(skb);
Julian Anastasov0cc535a2012-07-18 21:35:03 +0000288 if ((rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | RTCF_LOCAL)) ==
289 RTCF_LOCAL)
David S. Miller35ebf652012-06-28 03:59:11 -0700290 return ip_hdr(skb)->daddr;
291
292 in_dev = __in_dev_get_rcu(dev);
293 BUG_ON(!in_dev);
David S. Miller35ebf652012-06-28 03:59:11 -0700294
295 net = dev_net(dev);
David S. Millera207a4b2012-06-28 18:33:24 -0700296
297 scope = RT_SCOPE_UNIVERSE;
298 if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) {
299 fl4.flowi4_oif = 0;
Pavel Emelyanov1fb94892012-08-08 21:53:36 +0000300 fl4.flowi4_iif = LOOPBACK_IFINDEX;
David S. Millera207a4b2012-06-28 18:33:24 -0700301 fl4.daddr = ip_hdr(skb)->saddr;
302 fl4.saddr = 0;
303 fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
304 fl4.flowi4_scope = scope;
305 fl4.flowi4_mark = IN_DEV_SRC_VMARK(in_dev) ? skb->mark : 0;
Thomas Graf1b7179d2015-07-21 10:43:59 +0200306 fl4.flowi4_tun_key.tun_id = 0;
Andy Gospodarek0eeb0752015-06-23 13:45:37 -0400307 if (!fib_lookup(net, &fl4, &res, 0))
David S. Millera207a4b2012-06-28 18:33:24 -0700308 return FIB_RES_PREFSRC(net, res);
309 } else {
310 scope = RT_SCOPE_LINK;
311 }
312
313 return inet_select_addr(dev, ip_hdr(skb)->saddr, scope);
David S. Miller35ebf652012-06-28 03:59:11 -0700314}
315
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316/* Given (packet source, input interface) and optional (dst, oif, tos):
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000317 * - (main) check, that source is valid i.e. not broadcast or our local
318 * address.
319 * - figure out what "logical" interface this packet arrived
320 * and calculate "specific destination" address.
321 * - check, that packet arrived from expected physical interface.
Eric Dumazetebc0ffa2010-10-05 10:41:36 +0000322 * called with rcu_read_lock()
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323 */
David S. Miller7a9bc9b2012-06-29 01:32:45 -0700324static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
325 u8 tos, int oif, struct net_device *dev,
326 int rpf, struct in_device *idev, u32 *itag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327{
Sébastien Barré1dced6a2014-08-17 09:19:54 +0200328 int ret, no_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329 struct fib_result res;
David S. Miller9e56e382012-06-28 18:54:02 -0700330 struct flowi4 fl4;
Denis V. Lunev5b707aa2008-01-21 17:33:15 -0800331 struct net *net;
David S. Miller9e56e382012-06-28 18:54:02 -0700332 bool dev_match;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333
David S. Miller9ade2282011-03-12 02:02:42 -0500334 fl4.flowi4_oif = 0;
David Ahern385add92015-09-29 20:07:13 -0700335 fl4.flowi4_iif = l3mdev_master_ifindex_rcu(dev);
David Aherncd2fbe12015-08-13 14:59:01 -0600336 if (!fl4.flowi4_iif)
337 fl4.flowi4_iif = oif ? : LOOPBACK_IFINDEX;
David S. Miller9ade2282011-03-12 02:02:42 -0500338 fl4.daddr = src;
339 fl4.saddr = dst;
340 fl4.flowi4_tos = tos;
341 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
Thomas Graf1b7179d2015-07-21 10:43:59 +0200342 fl4.flowi4_tun_key.tun_id = 0;
David Ahernb84f7872015-09-29 19:07:07 -0700343 fl4.flowi4_flags = 0;
David S. Millercc7e17e2011-03-09 20:57:50 -0800344
David S. Miller9e56e382012-06-28 18:54:02 -0700345 no_addr = idev->ifa_list == NULL;
Michael Smith990078a2011-04-07 04:51:51 +0000346
David S. Miller9e56e382012-06-28 18:54:02 -0700347 fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348
David Ahernf6d3c192015-08-28 08:42:09 -0700349 trace_fib_validate_source(dev, &fl4);
350
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900351 net = dev_net(dev);
Andy Gospodarek0eeb0752015-06-23 13:45:37 -0400352 if (fib_lookup(net, &fl4, &res, 0))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 goto last_resort;
Sébastien Barré1dced6a2014-08-17 09:19:54 +0200354 if (res.type != RTN_UNICAST &&
355 (res.type != RTN_LOCAL || !IN_DEV_ACCEPT_LOCAL(idev)))
356 goto e_inval;
357 if (!rpf && !fib_num_tclassid_users(dev_net(dev)) &&
358 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev)))
359 goto last_resort;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 fib_combine_itag(itag, &res);
David S. Miller6f86b322010-09-06 22:36:19 -0700361 dev_match = false;
362
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363#ifdef CONFIG_IP_ROUTE_MULTIPATH
David S. Miller6f86b322010-09-06 22:36:19 -0700364 for (ret = 0; ret < res.fi->fib_nhs; ret++) {
365 struct fib_nh *nh = &res.fi->fib_nh[ret];
366
367 if (nh->nh_dev == dev) {
368 dev_match = true;
369 break;
David Ahern385add92015-09-29 20:07:13 -0700370 } else if (l3mdev_master_ifindex_rcu(nh->nh_dev) == dev->ifindex) {
David Aherncd2fbe12015-08-13 14:59:01 -0600371 dev_match = true;
372 break;
David S. Miller6f86b322010-09-06 22:36:19 -0700373 }
374 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375#else
376 if (FIB_RES_DEV(res) == dev)
David S. Miller6f86b322010-09-06 22:36:19 -0700377 dev_match = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378#endif
David S. Miller6f86b322010-09-06 22:36:19 -0700379 if (dev_match) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 return ret;
382 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 if (no_addr)
384 goto last_resort;
Stephen Hemmingerc1cf8422009-02-20 08:25:36 +0000385 if (rpf == 1)
Eric Dumazetb5f7e752010-06-02 12:05:27 +0000386 goto e_rpf;
David S. Miller9ade2282011-03-12 02:02:42 -0500387 fl4.flowi4_oif = dev->ifindex;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388
389 ret = 0;
Andy Gospodarek0eeb0752015-06-23 13:45:37 -0400390 if (fib_lookup(net, &fl4, &res, FIB_LOOKUP_IGNORE_LINKSTATE) == 0) {
David S. Miller41347dc2012-06-28 04:05:27 -0700391 if (res.type == RTN_UNICAST)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 }
394 return ret;
395
396last_resort:
397 if (rpf)
Eric Dumazetb5f7e752010-06-02 12:05:27 +0000398 goto e_rpf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 *itag = 0;
400 return 0;
401
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402e_inval:
403 return -EINVAL;
Eric Dumazetb5f7e752010-06-02 12:05:27 +0000404e_rpf:
405 return -EXDEV;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406}
407
David S. Miller7a9bc9b2012-06-29 01:32:45 -0700408/* Ignore rp_filter for packets protected by IPsec. */
409int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
410 u8 tos, int oif, struct net_device *dev,
411 struct in_device *idev, u32 *itag)
412{
413 int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev);
414
Julian Anastasove81da0e2012-10-08 11:41:15 +0000415 if (!r && !fib_num_tclassid_users(dev_net(dev)) &&
Sébastien Barré1dced6a2014-08-17 09:19:54 +0200416 IN_DEV_ACCEPT_LOCAL(idev) &&
Julian Anastasove81da0e2012-10-08 11:41:15 +0000417 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) {
David S. Miller7a9bc9b2012-06-29 01:32:45 -0700418 *itag = 0;
419 return 0;
420 }
421 return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag);
422}
423
Al Viro81f7bf62006-09-27 18:40:00 -0700424static inline __be32 sk_extract_addr(struct sockaddr *addr)
Thomas Graf4e902c52006-08-17 18:14:52 -0700425{
426 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
427}
428
429static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
430{
431 struct nlattr *nla;
432
433 nla = (struct nlattr *) ((char *) mx + len);
434 nla->nla_type = type;
435 nla->nla_len = nla_attr_size(4);
436 *(u32 *) nla_data(nla) = value;
437
438 return len + nla_total_size(4);
439}
440
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800441static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
Thomas Graf4e902c52006-08-17 18:14:52 -0700442 struct fib_config *cfg)
443{
Al Viro6d85c102006-09-26 22:15:46 -0700444 __be32 addr;
Thomas Graf4e902c52006-08-17 18:14:52 -0700445 int plen;
446
447 memset(cfg, 0, sizeof(*cfg));
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800448 cfg->fc_nlinfo.nl_net = net;
Thomas Graf4e902c52006-08-17 18:14:52 -0700449
450 if (rt->rt_dst.sa_family != AF_INET)
451 return -EAFNOSUPPORT;
452
453 /*
454 * Check mask for validity:
455 * a) it must be contiguous.
456 * b) destination must have all host bits clear.
457 * c) if application forgot to set correct family (AF_INET),
458 * reject request unless it is absolutely clear i.e.
459 * both family and mask are zero.
460 */
461 plen = 32;
462 addr = sk_extract_addr(&rt->rt_dst);
463 if (!(rt->rt_flags & RTF_HOST)) {
Al Viro81f7bf62006-09-27 18:40:00 -0700464 __be32 mask = sk_extract_addr(&rt->rt_genmask);
Thomas Graf4e902c52006-08-17 18:14:52 -0700465
466 if (rt->rt_genmask.sa_family != AF_INET) {
467 if (mask || rt->rt_genmask.sa_family)
468 return -EAFNOSUPPORT;
469 }
470
471 if (bad_mask(mask, addr))
472 return -EINVAL;
473
474 plen = inet_mask_len(mask);
475 }
476
477 cfg->fc_dst_len = plen;
478 cfg->fc_dst = addr;
479
480 if (cmd != SIOCDELRT) {
481 cfg->fc_nlflags = NLM_F_CREATE;
482 cfg->fc_protocol = RTPROT_BOOT;
483 }
484
485 if (rt->rt_metric)
486 cfg->fc_priority = rt->rt_metric - 1;
487
488 if (rt->rt_flags & RTF_REJECT) {
489 cfg->fc_scope = RT_SCOPE_HOST;
490 cfg->fc_type = RTN_UNREACHABLE;
491 return 0;
492 }
493
494 cfg->fc_scope = RT_SCOPE_NOWHERE;
495 cfg->fc_type = RTN_UNICAST;
496
497 if (rt->rt_dev) {
498 char *colon;
499 struct net_device *dev;
500 char devname[IFNAMSIZ];
501
502 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
503 return -EFAULT;
504
505 devname[IFNAMSIZ-1] = 0;
506 colon = strchr(devname, ':');
507 if (colon)
508 *colon = 0;
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800509 dev = __dev_get_by_name(net, devname);
Thomas Graf4e902c52006-08-17 18:14:52 -0700510 if (!dev)
511 return -ENODEV;
512 cfg->fc_oif = dev->ifindex;
513 if (colon) {
514 struct in_ifaddr *ifa;
515 struct in_device *in_dev = __in_dev_get_rtnl(dev);
516 if (!in_dev)
517 return -ENODEV;
518 *colon = ':';
519 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
520 if (strcmp(ifa->ifa_label, devname) == 0)
521 break;
Ian Morris51456b22015-04-03 09:17:26 +0100522 if (!ifa)
Thomas Graf4e902c52006-08-17 18:14:52 -0700523 return -ENODEV;
524 cfg->fc_prefsrc = ifa->ifa_local;
525 }
526 }
527
528 addr = sk_extract_addr(&rt->rt_gateway);
529 if (rt->rt_gateway.sa_family == AF_INET && addr) {
David Ahern30bbaa12015-08-13 14:59:05 -0600530 unsigned int addr_type;
531
Thomas Graf4e902c52006-08-17 18:14:52 -0700532 cfg->fc_gw = addr;
David Ahern30bbaa12015-08-13 14:59:05 -0600533 addr_type = inet_addr_type_table(net, addr, cfg->fc_table);
Thomas Graf4e902c52006-08-17 18:14:52 -0700534 if (rt->rt_flags & RTF_GATEWAY &&
David Ahern30bbaa12015-08-13 14:59:05 -0600535 addr_type == RTN_UNICAST)
Thomas Graf4e902c52006-08-17 18:14:52 -0700536 cfg->fc_scope = RT_SCOPE_UNIVERSE;
537 }
538
539 if (cmd == SIOCDELRT)
540 return 0;
541
542 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
543 return -EINVAL;
544
545 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
546 cfg->fc_scope = RT_SCOPE_LINK;
547
548 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
549 struct nlattr *mx;
550 int len = 0;
551
552 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
Ian Morris51456b22015-04-03 09:17:26 +0100553 if (!mx)
Thomas Graf4e902c52006-08-17 18:14:52 -0700554 return -ENOMEM;
555
556 if (rt->rt_flags & RTF_MTU)
557 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
558
559 if (rt->rt_flags & RTF_WINDOW)
560 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
561
562 if (rt->rt_flags & RTF_IRTT)
563 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
564
565 cfg->fc_mx = mx;
566 cfg->fc_mx_len = len;
567 }
568
569 return 0;
570}
571
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572/*
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000573 * Handle IP routing ioctl calls.
574 * These are used to manipulate the routing tables
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 */
Denis V. Lunev1bad1182008-01-10 03:29:53 -0800576int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577{
Thomas Graf4e902c52006-08-17 18:14:52 -0700578 struct fib_config cfg;
579 struct rtentry rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581
582 switch (cmd) {
583 case SIOCADDRT: /* Add a route */
584 case SIOCDELRT: /* Delete a route */
Eric W. Biederman52e804c2012-11-16 03:03:05 +0000585 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 return -EPERM;
Thomas Graf4e902c52006-08-17 18:14:52 -0700587
588 if (copy_from_user(&rt, arg, sizeof(rt)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 return -EFAULT;
Thomas Graf4e902c52006-08-17 18:14:52 -0700590
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 rtnl_lock();
Denis V. Lunev1bad1182008-01-10 03:29:53 -0800592 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593 if (err == 0) {
Thomas Graf4e902c52006-08-17 18:14:52 -0700594 struct fib_table *tb;
595
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 if (cmd == SIOCDELRT) {
Denis V. Lunev1bad1182008-01-10 03:29:53 -0800597 tb = fib_get_table(net, cfg.fc_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598 if (tb)
Stephen Hemminger16c6cf82009-09-20 10:35:36 +0000599 err = fib_table_delete(tb, &cfg);
Thomas Graf4e902c52006-08-17 18:14:52 -0700600 else
601 err = -ESRCH;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 } else {
Denis V. Lunev1bad1182008-01-10 03:29:53 -0800603 tb = fib_new_table(net, cfg.fc_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 if (tb)
Stephen Hemminger16c6cf82009-09-20 10:35:36 +0000605 err = fib_table_insert(tb, &cfg);
Thomas Graf4e902c52006-08-17 18:14:52 -0700606 else
607 err = -ENOBUFS;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 }
Thomas Graf4e902c52006-08-17 18:14:52 -0700609
610 /* allocated by rtentry_to_fib_config() */
611 kfree(cfg.fc_mx);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612 }
613 rtnl_unlock();
614 return err;
615 }
616 return -EINVAL;
617}
618
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000619const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
Thomas Graf4e902c52006-08-17 18:14:52 -0700620 [RTA_DST] = { .type = NLA_U32 },
621 [RTA_SRC] = { .type = NLA_U32 },
622 [RTA_IIF] = { .type = NLA_U32 },
623 [RTA_OIF] = { .type = NLA_U32 },
624 [RTA_GATEWAY] = { .type = NLA_U32 },
625 [RTA_PRIORITY] = { .type = NLA_U32 },
626 [RTA_PREFSRC] = { .type = NLA_U32 },
627 [RTA_METRICS] = { .type = NLA_NESTED },
Thomas Graf5176f912006-08-26 20:13:18 -0700628 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
Thomas Graf4e902c52006-08-17 18:14:52 -0700629 [RTA_FLOW] = { .type = NLA_U32 },
Roopa Prabhu571e7222015-07-21 10:43:47 +0200630 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
631 [RTA_ENCAP] = { .type = NLA_NESTED },
Thomas Graf4e902c52006-08-17 18:14:52 -0700632};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800634static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000635 struct nlmsghdr *nlh, struct fib_config *cfg)
Thomas Graf4e902c52006-08-17 18:14:52 -0700636{
637 struct nlattr *attr;
638 int err, remaining;
639 struct rtmsg *rtm;
640
641 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
642 if (err < 0)
643 goto errout;
644
645 memset(cfg, 0, sizeof(*cfg));
646
647 rtm = nlmsg_data(nlh);
Thomas Graf4e902c52006-08-17 18:14:52 -0700648 cfg->fc_dst_len = rtm->rtm_dst_len;
Thomas Graf4e902c52006-08-17 18:14:52 -0700649 cfg->fc_tos = rtm->rtm_tos;
650 cfg->fc_table = rtm->rtm_table;
651 cfg->fc_protocol = rtm->rtm_protocol;
652 cfg->fc_scope = rtm->rtm_scope;
653 cfg->fc_type = rtm->rtm_type;
654 cfg->fc_flags = rtm->rtm_flags;
655 cfg->fc_nlflags = nlh->nlmsg_flags;
656
Eric W. Biederman15e47302012-09-07 20:12:54 +0000657 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
Thomas Graf4e902c52006-08-17 18:14:52 -0700658 cfg->fc_nlinfo.nlh = nlh;
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800659 cfg->fc_nlinfo.nl_net = net;
Thomas Graf4e902c52006-08-17 18:14:52 -0700660
Thomas Grafa0ee18b2007-03-24 20:32:54 -0700661 if (cfg->fc_type > RTN_MAX) {
662 err = -EINVAL;
663 goto errout;
664 }
665
Thomas Graf4e902c52006-08-17 18:14:52 -0700666 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
Thomas Graf8f4c1f92007-09-12 14:44:36 +0200667 switch (nla_type(attr)) {
Thomas Graf4e902c52006-08-17 18:14:52 -0700668 case RTA_DST:
Al Viro17fb2c62006-09-26 22:15:25 -0700669 cfg->fc_dst = nla_get_be32(attr);
Thomas Graf4e902c52006-08-17 18:14:52 -0700670 break;
Thomas Graf4e902c52006-08-17 18:14:52 -0700671 case RTA_OIF:
672 cfg->fc_oif = nla_get_u32(attr);
673 break;
674 case RTA_GATEWAY:
Al Viro17fb2c62006-09-26 22:15:25 -0700675 cfg->fc_gw = nla_get_be32(attr);
Thomas Graf4e902c52006-08-17 18:14:52 -0700676 break;
677 case RTA_PRIORITY:
678 cfg->fc_priority = nla_get_u32(attr);
679 break;
680 case RTA_PREFSRC:
Al Viro17fb2c62006-09-26 22:15:25 -0700681 cfg->fc_prefsrc = nla_get_be32(attr);
Thomas Graf4e902c52006-08-17 18:14:52 -0700682 break;
683 case RTA_METRICS:
684 cfg->fc_mx = nla_data(attr);
685 cfg->fc_mx_len = nla_len(attr);
686 break;
687 case RTA_MULTIPATH:
688 cfg->fc_mp = nla_data(attr);
689 cfg->fc_mp_len = nla_len(attr);
690 break;
691 case RTA_FLOW:
692 cfg->fc_flow = nla_get_u32(attr);
693 break;
Thomas Graf4e902c52006-08-17 18:14:52 -0700694 case RTA_TABLE:
695 cfg->fc_table = nla_get_u32(attr);
696 break;
Roopa Prabhu571e7222015-07-21 10:43:47 +0200697 case RTA_ENCAP:
698 cfg->fc_encap = attr;
699 break;
700 case RTA_ENCAP_TYPE:
701 cfg->fc_encap_type = nla_get_u16(attr);
702 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 }
704 }
Thomas Graf4e902c52006-08-17 18:14:52 -0700705
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706 return 0;
Thomas Graf4e902c52006-08-17 18:14:52 -0700707errout:
708 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709}
710
Thomas Graf661d2962013-03-21 07:45:29 +0000711static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900713 struct net *net = sock_net(skb->sk);
Thomas Graf4e902c52006-08-17 18:14:52 -0700714 struct fib_config cfg;
715 struct fib_table *tb;
716 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800718 err = rtm_to_fib_config(net, skb, nlh, &cfg);
Thomas Graf4e902c52006-08-17 18:14:52 -0700719 if (err < 0)
720 goto errout;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721
Denis V. Lunev8ad49422008-01-10 03:24:11 -0800722 tb = fib_get_table(net, cfg.fc_table);
Ian Morris51456b22015-04-03 09:17:26 +0100723 if (!tb) {
Thomas Graf4e902c52006-08-17 18:14:52 -0700724 err = -ESRCH;
725 goto errout;
726 }
727
Stephen Hemminger16c6cf82009-09-20 10:35:36 +0000728 err = fib_table_delete(tb, &cfg);
Thomas Graf4e902c52006-08-17 18:14:52 -0700729errout:
730 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731}
732
Thomas Graf661d2962013-03-21 07:45:29 +0000733static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900735 struct net *net = sock_net(skb->sk);
Thomas Graf4e902c52006-08-17 18:14:52 -0700736 struct fib_config cfg;
737 struct fib_table *tb;
738 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800740 err = rtm_to_fib_config(net, skb, nlh, &cfg);
Thomas Graf4e902c52006-08-17 18:14:52 -0700741 if (err < 0)
742 goto errout;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743
Denis V. Lunev226b0b4a52008-01-10 03:30:24 -0800744 tb = fib_new_table(net, cfg.fc_table);
Ian Morris51456b22015-04-03 09:17:26 +0100745 if (!tb) {
Thomas Graf4e902c52006-08-17 18:14:52 -0700746 err = -ENOBUFS;
747 goto errout;
748 }
749
Stephen Hemminger16c6cf82009-09-20 10:35:36 +0000750 err = fib_table_insert(tb, &cfg);
Thomas Graf4e902c52006-08-17 18:14:52 -0700751errout:
752 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753}
754
Thomas Graf63f34442007-03-22 11:55:17 -0700755static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900757 struct net *net = sock_net(skb->sk);
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700758 unsigned int h, s_h;
759 unsigned int e = 0, s_e;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760 struct fib_table *tb;
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800761 struct hlist_head *head;
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700762 int dumped = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763
Thomas Grafbe403ea2006-08-17 18:15:17 -0700764 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
765 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
Li RongQing0b8c7f62014-03-21 11:33:10 +0800766 return skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700768 s_h = cb->args[0];
769 s_e = cb->args[1];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770
Alexander Duycka7e53532015-03-04 15:02:44 -0800771 rcu_read_lock();
772
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700773 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
774 e = 0;
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800775 head = &net->ipv4.fib_table_hash[h];
Alexander Duycka7e53532015-03-04 15:02:44 -0800776 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700777 if (e < s_e)
778 goto next;
779 if (dumped)
780 memset(&cb->args[2], 0, sizeof(cb->args) -
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900781 2 * sizeof(cb->args[0]));
Stephen Hemminger16c6cf82009-09-20 10:35:36 +0000782 if (fib_table_dump(tb, skb, cb) < 0)
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700783 goto out;
784 dumped = 1;
785next:
786 e++;
787 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 }
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700789out:
Alexander Duycka7e53532015-03-04 15:02:44 -0800790 rcu_read_unlock();
791
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700792 cb->args[1] = e;
793 cb->args[0] = h;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794
795 return skb->len;
796}
797
798/* Prepare and feed intra-kernel routing request.
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000799 * Really, it should be netlink message, but :-( netlink
800 * can be not configured, so that we feed it directly
801 * to fib engine. It is legal, because all events occur
802 * only when netlink is already locked.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 */
Al Viro81f7bf62006-09-27 18:40:00 -0700804static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805{
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900806 struct net *net = dev_net(ifa->ifa_dev->dev);
David Ahern3236b002015-09-29 20:07:14 -0700807 u32 tb_id = l3mdev_fib_table(ifa->ifa_dev->dev);
Thomas Graf4e902c52006-08-17 18:14:52 -0700808 struct fib_table *tb;
809 struct fib_config cfg = {
810 .fc_protocol = RTPROT_KERNEL,
811 .fc_type = type,
812 .fc_dst = dst,
813 .fc_dst_len = dst_len,
814 .fc_prefsrc = ifa->ifa_local,
815 .fc_oif = ifa->ifa_dev->dev->ifindex,
816 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
Denis V. Lunev4d1169c2008-01-10 03:26:13 -0800817 .fc_nlinfo = {
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800818 .nl_net = net,
Denis V. Lunev4d1169c2008-01-10 03:26:13 -0800819 },
Thomas Graf4e902c52006-08-17 18:14:52 -0700820 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821
David Ahern021dd3b2015-08-13 14:59:06 -0600822 if (!tb_id)
823 tb_id = (type == RTN_UNICAST) ? RT_TABLE_MAIN : RT_TABLE_LOCAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824
David Ahern021dd3b2015-08-13 14:59:06 -0600825 tb = fib_new_table(net, tb_id);
Ian Morris51456b22015-04-03 09:17:26 +0100826 if (!tb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 return;
828
Thomas Graf4e902c52006-08-17 18:14:52 -0700829 cfg.fc_table = tb->tb_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830
Thomas Graf4e902c52006-08-17 18:14:52 -0700831 if (type != RTN_LOCAL)
832 cfg.fc_scope = RT_SCOPE_LINK;
833 else
834 cfg.fc_scope = RT_SCOPE_HOST;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835
836 if (cmd == RTM_NEWROUTE)
Stephen Hemminger16c6cf82009-09-20 10:35:36 +0000837 fib_table_insert(tb, &cfg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 else
Stephen Hemminger16c6cf82009-09-20 10:35:36 +0000839 fib_table_delete(tb, &cfg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840}
841
Jamal Hadi Salim0ff60a42005-11-22 14:47:37 -0800842void fib_add_ifaddr(struct in_ifaddr *ifa)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843{
844 struct in_device *in_dev = ifa->ifa_dev;
845 struct net_device *dev = in_dev->dev;
846 struct in_ifaddr *prim = ifa;
Al Viroa144ea42006-09-28 18:00:55 -0700847 __be32 mask = ifa->ifa_mask;
848 __be32 addr = ifa->ifa_local;
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000849 __be32 prefix = ifa->ifa_address & mask;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000851 if (ifa->ifa_flags & IFA_F_SECONDARY) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852 prim = inet_ifa_byprefix(in_dev, prefix, mask);
Ian Morris51456b22015-04-03 09:17:26 +0100853 if (!prim) {
Joe Perches058bd4d2012-03-11 18:36:11 +0000854 pr_warn("%s: bug: prim == NULL\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 return;
856 }
857 }
858
859 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
860
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000861 if (!(dev->flags & IFF_UP))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 return;
863
864 /* Add broadcast address, if it is explicitly assigned. */
Al Viroa144ea42006-09-28 18:00:55 -0700865 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
867
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000868 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 (prefix != addr || ifa->ifa_prefixlen < 32)) {
Paolo Abeni7b131182015-10-20 10:28:45 +0200870 if (!(ifa->ifa_flags & IFA_F_NOPREFIXROUTE))
871 fib_magic(RTM_NEWROUTE,
872 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
873 prefix, ifa->ifa_prefixlen, prim);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874
875 /* Add network specific broadcasts, when it takes a sense */
876 if (ifa->ifa_prefixlen < 31) {
877 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000878 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
879 32, prim);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 }
881 }
882}
883
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000884/* Delete primary or secondary address.
885 * Optionally, on secondary address promotion consider the addresses
886 * from subnet iprim as deleted, even if they are in device list.
887 * In this case the secondary ifa can be in device list.
888 */
889void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890{
891 struct in_device *in_dev = ifa->ifa_dev;
892 struct net_device *dev = in_dev->dev;
893 struct in_ifaddr *ifa1;
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000894 struct in_ifaddr *prim = ifa, *prim1 = NULL;
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000895 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
896 __be32 any = ifa->ifa_address & ifa->ifa_mask;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897#define LOCAL_OK 1
898#define BRD_OK 2
899#define BRD0_OK 4
900#define BRD1_OK 8
Eric Dumazet95c96172012-04-15 05:58:06 +0000901 unsigned int ok = 0;
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000902 int subnet = 0; /* Primary network */
903 int gone = 1; /* Address is missing */
904 int same_prefsrc = 0; /* Another primary with same IP */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000906 if (ifa->ifa_flags & IFA_F_SECONDARY) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
Ian Morris51456b22015-04-03 09:17:26 +0100908 if (!prim) {
Joe Perches058bd4d2012-03-11 18:36:11 +0000909 pr_warn("%s: bug: prim == NULL\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 return;
911 }
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000912 if (iprim && iprim != prim) {
Joe Perches058bd4d2012-03-11 18:36:11 +0000913 pr_warn("%s: bug: iprim != prim\n", __func__);
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000914 return;
915 }
916 } else if (!ipv4_is_zeronet(any) &&
917 (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) {
Paolo Abeni7b131182015-10-20 10:28:45 +0200918 if (!(ifa->ifa_flags & IFA_F_NOPREFIXROUTE))
919 fib_magic(RTM_DELROUTE,
920 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
921 any, ifa->ifa_prefixlen, prim);
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000922 subnet = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923 }
924
925 /* Deletion is more complicated than add.
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000926 * We should take care of not to delete too much :-)
927 *
928 * Scan address list to be sure that addresses are really gone.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929 */
930
931 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000932 if (ifa1 == ifa) {
933 /* promotion, keep the IP */
934 gone = 0;
935 continue;
936 }
937 /* Ignore IFAs from our subnet */
938 if (iprim && ifa1->ifa_mask == iprim->ifa_mask &&
939 inet_ifa_match(ifa1->ifa_address, iprim))
940 continue;
941
942 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
943 if (ifa1->ifa_flags & IFA_F_SECONDARY) {
944 /* Another address from our subnet? */
945 if (ifa1->ifa_mask == prim->ifa_mask &&
946 inet_ifa_match(ifa1->ifa_address, prim))
947 prim1 = prim;
948 else {
949 /* We reached the secondaries, so
950 * same_prefsrc should be determined.
951 */
952 if (!same_prefsrc)
953 continue;
954 /* Search new prim1 if ifa1 is not
955 * using the current prim1
956 */
957 if (!prim1 ||
958 ifa1->ifa_mask != prim1->ifa_mask ||
959 !inet_ifa_match(ifa1->ifa_address, prim1))
960 prim1 = inet_ifa_byprefix(in_dev,
961 ifa1->ifa_address,
962 ifa1->ifa_mask);
963 if (!prim1)
964 continue;
965 if (prim1->ifa_local != prim->ifa_local)
966 continue;
967 }
968 } else {
969 if (prim->ifa_local != ifa1->ifa_local)
970 continue;
971 prim1 = ifa1;
972 if (prim != prim1)
973 same_prefsrc = 1;
974 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 if (ifa->ifa_local == ifa1->ifa_local)
976 ok |= LOCAL_OK;
977 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
978 ok |= BRD_OK;
979 if (brd == ifa1->ifa_broadcast)
980 ok |= BRD1_OK;
981 if (any == ifa1->ifa_broadcast)
982 ok |= BRD0_OK;
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000983 /* primary has network specific broadcasts */
984 if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) {
985 __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask;
986 __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask;
987
988 if (!ipv4_is_zeronet(any1)) {
989 if (ifa->ifa_broadcast == brd1 ||
990 ifa->ifa_broadcast == any1)
991 ok |= BRD_OK;
992 if (brd == brd1 || brd == any1)
993 ok |= BRD1_OK;
994 if (any == brd1 || any == any1)
995 ok |= BRD0_OK;
996 }
997 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998 }
999
Eric Dumazet6a31d2a2010-10-04 20:00:18 +00001000 if (!(ok & BRD_OK))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
Julian Anastasove6abbaa2011-03-19 12:13:49 +00001002 if (subnet && ifa->ifa_prefixlen < 31) {
1003 if (!(ok & BRD1_OK))
1004 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
1005 if (!(ok & BRD0_OK))
1006 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
1007 }
Eric Dumazet6a31d2a2010-10-04 20:00:18 +00001008 if (!(ok & LOCAL_OK)) {
David Ahern30bbaa12015-08-13 14:59:05 -06001009 unsigned int addr_type;
1010
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
1012
1013 /* Check, that this local address finally disappeared. */
David Ahern30bbaa12015-08-13 14:59:05 -06001014 addr_type = inet_addr_type_dev_table(dev_net(dev), dev,
1015 ifa->ifa_local);
1016 if (gone && addr_type != RTN_LOCAL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 /* And the last, but not the least thing.
Eric Dumazet6a31d2a2010-10-04 20:00:18 +00001018 * We must flush stray FIB entries.
1019 *
1020 * First of all, we scan fib_info list searching
1021 * for stray nexthop entries, then ignite fib_flush.
1022 */
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001023 if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local))
1024 fib_flush(dev_net(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 }
1026 }
1027#undef LOCAL_OK
1028#undef BRD_OK
1029#undef BRD0_OK
1030#undef BRD1_OK
1031}
1032
Alexander Duyck345e9b52014-12-31 10:56:24 -08001033static void nl_fib_lookup(struct net *net, struct fib_result_nl *frn)
Robert Olsson246955f2005-06-20 13:36:39 -07001034{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001035
Robert Olsson246955f2005-06-20 13:36:39 -07001036 struct fib_result res;
David S. Miller9ade2282011-03-12 02:02:42 -05001037 struct flowi4 fl4 = {
1038 .flowi4_mark = frn->fl_mark,
1039 .daddr = frn->fl_addr,
1040 .flowi4_tos = frn->fl_tos,
1041 .flowi4_scope = frn->fl_scope,
Eric Dumazet6a31d2a2010-10-04 20:00:18 +00001042 };
Alexander Duyck345e9b52014-12-31 10:56:24 -08001043 struct fib_table *tb;
1044
1045 rcu_read_lock();
1046
1047 tb = fib_get_table(net, frn->tb_id_in);
Alexey Kuznetsov1194ed02007-04-25 13:07:28 -07001048
1049 frn->err = -ENOENT;
Robert Olsson246955f2005-06-20 13:36:39 -07001050 if (tb) {
1051 local_bh_disable();
1052
1053 frn->tb_id = tb->tb_id;
David S. Miller9ade2282011-03-12 02:02:42 -05001054 frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF);
Robert Olsson246955f2005-06-20 13:36:39 -07001055
1056 if (!frn->err) {
1057 frn->prefixlen = res.prefixlen;
1058 frn->nh_sel = res.nh_sel;
1059 frn->type = res.type;
1060 frn->scope = res.scope;
1061 }
1062 local_bh_enable();
1063 }
Alexander Duyck345e9b52014-12-31 10:56:24 -08001064
1065 rcu_read_unlock();
Robert Olsson246955f2005-06-20 13:36:39 -07001066}
1067
David S. Miller28f7b0362007-10-10 21:32:39 -07001068static void nl_fib_input(struct sk_buff *skb)
Robert Olsson246955f2005-06-20 13:36:39 -07001069{
Denis V. Lunev6bd48fc2008-01-10 03:28:55 -08001070 struct net *net;
Robert Olsson246955f2005-06-20 13:36:39 -07001071 struct fib_result_nl *frn;
David S. Miller28f7b0362007-10-10 21:32:39 -07001072 struct nlmsghdr *nlh;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001073 u32 portid;
Alexey Kuznetsov1194ed02007-04-25 13:07:28 -07001074
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001075 net = sock_net(skb->sk);
Arnaldo Carvalho de Melob529ccf2007-04-25 19:08:35 -07001076 nlh = nlmsg_hdr(skb);
Hong zhi guo573ce262013-03-27 06:47:04 +00001077 if (skb->len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len ||
1078 nlmsg_len(nlh) < sizeof(*frn))
Thomas Grafea865752005-12-01 14:30:00 -08001079 return;
Denis V. Lunevd883a032007-12-21 02:01:53 -08001080
Pablo Neira3a365152013-06-28 03:04:23 +02001081 skb = netlink_skb_clone(skb, GFP_KERNEL);
Ian Morris51456b22015-04-03 09:17:26 +01001082 if (!skb)
Denis V. Lunevd883a032007-12-21 02:01:53 -08001083 return;
1084 nlh = nlmsg_hdr(skb);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001085
Hong zhi guo573ce262013-03-27 06:47:04 +00001086 frn = (struct fib_result_nl *) nlmsg_data(nlh);
Alexander Duyck345e9b52014-12-31 10:56:24 -08001087 nl_fib_lookup(net, frn);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001088
Rami Rosen28a28282013-01-11 09:59:20 +00001089 portid = NETLINK_CB(skb).portid; /* netlink portid */
Eric W. Biederman15e47302012-09-07 20:12:54 +00001090 NETLINK_CB(skb).portid = 0; /* from kernel */
Patrick McHardyac6d4392005-08-14 19:29:52 -07001091 NETLINK_CB(skb).dst_group = 0; /* unicast */
Eric W. Biederman15e47302012-09-07 20:12:54 +00001092 netlink_unicast(net->ipv4.fibnl, skb, portid, MSG_DONTWAIT);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001093}
Robert Olsson246955f2005-06-20 13:36:39 -07001094
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00001095static int __net_init nl_fib_lookup_init(struct net *net)
Robert Olsson246955f2005-06-20 13:36:39 -07001096{
Denis V. Lunev6bd48fc2008-01-10 03:28:55 -08001097 struct sock *sk;
Pablo Neira Ayusoa31f2d12012-06-29 06:15:21 +00001098 struct netlink_kernel_cfg cfg = {
1099 .input = nl_fib_input,
1100 };
1101
Pablo Neira Ayuso9f00d972012-09-08 02:53:54 +00001102 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, &cfg);
Ian Morris51456b22015-04-03 09:17:26 +01001103 if (!sk)
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001104 return -EAFNOSUPPORT;
Denis V. Lunev6bd48fc2008-01-10 03:28:55 -08001105 net->ipv4.fibnl = sk;
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001106 return 0;
1107}
1108
1109static void nl_fib_lookup_exit(struct net *net)
1110{
Denis V. Lunevb7c6ba62008-01-28 14:41:19 -08001111 netlink_kernel_release(net->ipv4.fibnl);
Denis V. Lunev775516b2008-01-18 23:55:19 -08001112 net->ipv4.fibnl = NULL;
Robert Olsson246955f2005-06-20 13:36:39 -07001113}
1114
Julian Anastasov4f823de2015-10-30 10:23:33 +02001115static void fib_disable_ip(struct net_device *dev, unsigned long event,
1116 bool force)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117{
Julian Anastasov4f823de2015-10-30 10:23:33 +02001118 if (fib_sync_down_dev(dev, event, force))
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001119 fib_flush(dev_net(dev));
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001120 rt_cache_flush(dev_net(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 arp_ifdown(dev);
1122}
1123
1124static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
1125{
Jianjun Kong6ed25332008-11-03 00:25:16 -08001126 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
Denis V. Lunev76e6ebf2008-07-05 19:00:44 -07001127 struct net_device *dev = ifa->ifa_dev->dev;
David S. Miller436c3b62011-03-24 17:42:21 -07001128 struct net *net = dev_net(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129
1130 switch (event) {
1131 case NETDEV_UP:
1132 fib_add_ifaddr(ifa);
1133#ifdef CONFIG_IP_ROUTE_MULTIPATH
Andy Gospodarek8a3d0312015-06-23 13:45:36 -04001134 fib_sync_up(dev, RTNH_F_DEAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135#endif
David S. Miller436c3b62011-03-24 17:42:21 -07001136 atomic_inc(&net->ipv4.dev_addr_genid);
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001137 rt_cache_flush(dev_net(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138 break;
1139 case NETDEV_DOWN:
Julian Anastasove6abbaa2011-03-19 12:13:49 +00001140 fib_del_ifaddr(ifa, NULL);
David S. Miller436c3b62011-03-24 17:42:21 -07001141 atomic_inc(&net->ipv4.dev_addr_genid);
Ian Morris51456b22015-04-03 09:17:26 +01001142 if (!ifa->ifa_dev->ifa_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 /* Last address was deleted from this interface.
Eric Dumazet6a31d2a2010-10-04 20:00:18 +00001144 * Disable IP.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145 */
Julian Anastasov4f823de2015-10-30 10:23:33 +02001146 fib_disable_ip(dev, event, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147 } else {
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001148 rt_cache_flush(dev_net(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149 }
1150 break;
1151 }
1152 return NOTIFY_DONE;
1153}
1154
1155static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1156{
Jiri Pirko351638e2013-05-28 01:30:21 +00001157 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
David Ahern7f49e7a2015-12-10 10:25:24 -08001158 struct netdev_notifier_changeupper_info *info;
Eric Dumazet0115e8e2012-08-22 17:19:46 +00001159 struct in_device *in_dev;
David S. Miller436c3b62011-03-24 17:42:21 -07001160 struct net *net = dev_net(dev);
Andy Gospodarek8a3d0312015-06-23 13:45:36 -04001161 unsigned int flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162
1163 if (event == NETDEV_UNREGISTER) {
Julian Anastasov4f823de2015-10-30 10:23:33 +02001164 fib_disable_ip(dev, event, true);
David S. Millercaacf052012-07-31 15:06:50 -07001165 rt_flush_dev(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166 return NOTIFY_DONE;
1167 }
1168
Eric Dumazet0115e8e2012-08-22 17:19:46 +00001169 in_dev = __in_dev_get_rtnl(dev);
Oliver Hartkoppa0065f22014-01-23 10:19:34 +01001170 if (!in_dev)
1171 return NOTIFY_DONE;
Eric Dumazet0115e8e2012-08-22 17:19:46 +00001172
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 switch (event) {
1174 case NETDEV_UP:
1175 for_ifa(in_dev) {
1176 fib_add_ifaddr(ifa);
1177 } endfor_ifa(in_dev);
1178#ifdef CONFIG_IP_ROUTE_MULTIPATH
Andy Gospodarek8a3d0312015-06-23 13:45:36 -04001179 fib_sync_up(dev, RTNH_F_DEAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180#endif
David S. Miller436c3b62011-03-24 17:42:21 -07001181 atomic_inc(&net->ipv4.dev_addr_genid);
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001182 rt_cache_flush(net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183 break;
1184 case NETDEV_DOWN:
Julian Anastasov4f823de2015-10-30 10:23:33 +02001185 fib_disable_ip(dev, event, false);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187 case NETDEV_CHANGE:
Andy Gospodarek8a3d0312015-06-23 13:45:36 -04001188 flags = dev_get_flags(dev);
1189 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1190 fib_sync_up(dev, RTNH_F_LINKDOWN);
1191 else
Julian Anastasov4f823de2015-10-30 10:23:33 +02001192 fib_sync_down_dev(dev, event, false);
Andy Gospodarek8a3d0312015-06-23 13:45:36 -04001193 /* fall through */
1194 case NETDEV_CHANGEMTU:
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001195 rt_cache_flush(net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 break;
David Ahern7f49e7a2015-12-10 10:25:24 -08001197 case NETDEV_CHANGEUPPER:
1198 info = ptr;
1199 /* flush all routes if dev is linked to or unlinked from
1200 * an L3 master device (e.g., VRF)
1201 */
1202 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
1203 fib_disable_ip(dev, NETDEV_DOWN, true);
1204 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205 }
1206 return NOTIFY_DONE;
1207}
1208
1209static struct notifier_block fib_inetaddr_notifier = {
Jianjun Kong6ed25332008-11-03 00:25:16 -08001210 .notifier_call = fib_inetaddr_event,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211};
1212
1213static struct notifier_block fib_netdev_notifier = {
Jianjun Kong6ed25332008-11-03 00:25:16 -08001214 .notifier_call = fib_netdev_event,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215};
1216
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001217static int __net_init ip_fib_net_init(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218{
Denis V. Lunevdce5cbe2008-01-31 18:44:53 -08001219 int err;
Eric Dumazet10da66f2010-10-13 08:22:03 +00001220 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
Patrick McHardy1af5a8c2006-08-10 23:10:46 -07001221
Eric Dumazet10da66f2010-10-13 08:22:03 +00001222 /* Avoid false sharing : Use at least a full cache line */
1223 size = max_t(size_t, size, L1_CACHE_BYTES);
1224
1225 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
Ian Morris51456b22015-04-03 09:17:26 +01001226 if (!net->ipv4.fib_table_hash)
Denis V. Luneve4aef8a2008-01-10 03:28:24 -08001227 return -ENOMEM;
1228
Denis V. Lunevdce5cbe2008-01-31 18:44:53 -08001229 err = fib4_rules_init(net);
1230 if (err < 0)
1231 goto fail;
1232 return 0;
1233
1234fail:
1235 kfree(net->ipv4.fib_table_hash);
1236 return err;
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001237}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00001239static void ip_fib_net_exit(struct net *net)
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001240{
1241 unsigned int i;
Thomas Graf63f34442007-03-22 11:55:17 -07001242
Sabrina Dubroca6dede752015-03-10 21:03:54 +01001243 rtnl_lock();
Alexander Duyck6e47d6c2015-03-27 14:14:22 -07001244#ifdef CONFIG_IP_MULTIPLE_TABLES
1245 RCU_INIT_POINTER(net->ipv4.fib_local, NULL);
1246 RCU_INIT_POINTER(net->ipv4.fib_main, NULL);
1247 RCU_INIT_POINTER(net->ipv4.fib_default, NULL);
1248#endif
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001249 for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
Alexander Duycka7e53532015-03-04 15:02:44 -08001250 struct hlist_head *head = &net->ipv4.fib_table_hash[i];
Sasha Levinb67bfe02013-02-27 17:06:00 -08001251 struct hlist_node *tmp;
Alexander Duycka7e53532015-03-04 15:02:44 -08001252 struct fib_table *tb;
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001253
Alexander Duycka7e53532015-03-04 15:02:44 -08001254 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) {
Alexander Duycka7e53532015-03-04 15:02:44 -08001255 hlist_del(&tb->tb_hlist);
Alexander Duyck6e47d6c2015-03-27 14:14:22 -07001256 fib_table_flush(tb);
Pavel Emelyanov4aa2c462010-10-28 02:00:43 +00001257 fib_free_table(tb);
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001258 }
1259 }
Alexander Duyckad88d052015-03-27 14:14:16 -07001260
1261#ifdef CONFIG_IP_MULTIPLE_TABLES
1262 fib4_rules_exit(net);
1263#endif
Eric Dumazete2666f82011-03-30 16:57:46 -07001264 rtnl_unlock();
Denis V. Luneve4aef8a2008-01-10 03:28:24 -08001265 kfree(net->ipv4.fib_table_hash);
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001266}
1267
1268static int __net_init fib_net_init(struct net *net)
1269{
1270 int error;
1271
David S. Millerf4530fa2012-07-05 22:13:13 -07001272#ifdef CONFIG_IP_ROUTE_CLASSID
1273 net->ipv4.fib_num_tclassid_users = 0;
1274#endif
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001275 error = ip_fib_net_init(net);
1276 if (error < 0)
1277 goto out;
1278 error = nl_fib_lookup_init(net);
1279 if (error < 0)
1280 goto out_nlfl;
1281 error = fib_proc_init(net);
1282 if (error < 0)
1283 goto out_proc;
1284out:
1285 return error;
1286
1287out_proc:
1288 nl_fib_lookup_exit(net);
1289out_nlfl:
1290 ip_fib_net_exit(net);
1291 goto out;
1292}
1293
1294static void __net_exit fib_net_exit(struct net *net)
1295{
1296 fib_proc_exit(net);
1297 nl_fib_lookup_exit(net);
1298 ip_fib_net_exit(net);
1299}
1300
1301static struct pernet_operations fib_net_ops = {
1302 .init = fib_net_init,
1303 .exit = fib_net_exit,
1304};
1305
1306void __init ip_fib_init(void)
1307{
Greg Rosec7ac8672011-06-10 01:27:09 +00001308 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, NULL);
1309 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, NULL);
1310 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, NULL);
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001311
1312 register_pernet_subsys(&fib_net_ops);
1313 register_netdevice_notifier(&fib_netdev_notifier);
1314 register_inetaddr_notifier(&fib_inetaddr_notifier);
Stephen Hemminger7f9b8052008-01-14 23:14:20 -08001315
David S. Miller5348ba82011-02-01 15:30:56 -08001316 fib_trie_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317}