blob: 08b7260447b6e90287f65fdbf049a3598e74fa58 [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 Aherne9db0422017-01-17 14:57:36 -080049#include <net/lwtunnel.h>
David Ahernf6d3c192015-08-28 08:42:09 -070050#include <trace/events/fib.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070051
Linus Torvalds1da177e2005-04-16 15:20:36 -070052#ifndef CONFIG_IP_MULTIPLE_TABLES
53
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -080054static int __net_init fib4_rules_init(struct net *net)
Pavel Emelyanovc3e9a352007-11-06 23:34:04 -080055{
Denis V. Lunev93456b62008-01-10 03:23:38 -080056 struct fib_table *local_table, *main_table;
57
Alexander Duyck0ddcf432015-03-06 13:47:00 -080058 main_table = fib_trie_table(RT_TABLE_MAIN, NULL);
Ian Morris51456b22015-04-03 09:17:26 +010059 if (!main_table)
Alexander Duyck61f0d862015-03-11 14:02:16 -070060 return -ENOMEM;
Denis V. Lunevdbb50162008-01-10 03:21:49 -080061
Alexander Duyck0ddcf432015-03-06 13:47:00 -080062 local_table = fib_trie_table(RT_TABLE_LOCAL, main_table);
Ian Morris51456b22015-04-03 09:17:26 +010063 if (!local_table)
Alexander Duyck61f0d862015-03-11 14:02:16 -070064 goto fail;
Alexander Duyck0ddcf432015-03-06 13:47:00 -080065
Denis V. Lunev93456b62008-01-10 03:23:38 -080066 hlist_add_head_rcu(&local_table->tb_hlist,
Denis V. Luneve4aef8a2008-01-10 03:28:24 -080067 &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]);
Denis V. Lunev93456b62008-01-10 03:23:38 -080068 hlist_add_head_rcu(&main_table->tb_hlist,
Denis V. Luneve4aef8a2008-01-10 03:28:24 -080069 &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]);
Denis V. Lunevdbb50162008-01-10 03:21:49 -080070 return 0;
71
72fail:
Alexander Duyck61f0d862015-03-11 14:02:16 -070073 fib_free_table(main_table);
Denis V. Lunevdbb50162008-01-10 03:21:49 -080074 return -ENOMEM;
Pavel Emelyanovc3e9a352007-11-06 23:34:04 -080075}
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#else
77
Denis V. Lunev8ad49422008-01-10 03:24:11 -080078struct fib_table *fib_new_table(struct net *net, u32 id)
Linus Torvalds1da177e2005-04-16 15:20:36 -070079{
Alexander Duyck0ddcf432015-03-06 13:47:00 -080080 struct fib_table *tb, *alias = NULL;
Patrick McHardy1af5a8c2006-08-10 23:10:46 -070081 unsigned int h;
Linus Torvalds1da177e2005-04-16 15:20:36 -070082
Patrick McHardy1af5a8c2006-08-10 23:10:46 -070083 if (id == 0)
84 id = RT_TABLE_MAIN;
Denis V. Lunev8ad49422008-01-10 03:24:11 -080085 tb = fib_get_table(net, id);
Patrick McHardy1af5a8c2006-08-10 23:10:46 -070086 if (tb)
87 return tb;
Stephen Hemminger7f9b8052008-01-14 23:14:20 -080088
Alexander Duyckefc455f2017-01-02 13:32:54 -080089 if (id == RT_TABLE_LOCAL && !net->ipv4.fib_has_custom_rules)
Alexander Duyck0ddcf432015-03-06 13:47:00 -080090 alias = fib_new_table(net, RT_TABLE_MAIN);
91
92 tb = fib_trie_table(id, alias);
Linus Torvalds1da177e2005-04-16 15:20:36 -070093 if (!tb)
94 return NULL;
David S. Millerf4530fa2012-07-05 22:13:13 -070095
96 switch (id) {
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}
David Ahernb3b46632016-05-04 21:46:12 -0700111EXPORT_SYMBOL_GPL(fib_new_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112
Alexander Duyck345e9b52014-12-31 10:56:24 -0800113/* caller must hold either rtnl or rcu read lock */
Denis V. Lunev8ad49422008-01-10 03:24:11 -0800114struct fib_table *fib_get_table(struct net *net, u32 id)
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700115{
116 struct fib_table *tb;
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800117 struct hlist_head *head;
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700118 unsigned int h;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700120 if (id == 0)
121 id = RT_TABLE_MAIN;
122 h = id & (FIB_TABLE_HASHSZ - 1);
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800123
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800124 head = &net->ipv4.fib_table_hash[h];
Sasha Levinb67bfe02013-02-27 17:06:00 -0800125 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
Alexander Duyck345e9b52014-12-31 10:56:24 -0800126 if (tb->tb_id == id)
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700127 return tb;
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700128 }
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700129 return NULL;
130}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131#endif /* CONFIG_IP_MULTIPLE_TABLES */
132
Alexander Duyck0ddcf432015-03-06 13:47:00 -0800133static void fib_replace_table(struct net *net, struct fib_table *old,
134 struct fib_table *new)
135{
136#ifdef CONFIG_IP_MULTIPLE_TABLES
137 switch (new->tb_id) {
Alexander Duyck0ddcf432015-03-06 13:47:00 -0800138 case RT_TABLE_MAIN:
139 rcu_assign_pointer(net->ipv4.fib_main, new);
140 break;
141 case RT_TABLE_DEFAULT:
142 rcu_assign_pointer(net->ipv4.fib_default, new);
143 break;
144 default:
145 break;
146 }
147
148#endif
149 /* replace the old table in the hlist */
150 hlist_replace_rcu(&old->tb_hlist, &new->tb_hlist);
151}
152
153int fib_unmerge(struct net *net)
154{
Alexander Duyck3b709332016-11-15 05:46:06 -0500155 struct fib_table *old, *new, *main_table;
Alexander Duyck0ddcf432015-03-06 13:47:00 -0800156
Alexander Duyck3c9e9f72015-03-12 14:46:23 -0700157 /* attempt to fetch local table if it has been allocated */
Alexander Duyck0ddcf432015-03-06 13:47:00 -0800158 old = fib_get_table(net, RT_TABLE_LOCAL);
Alexander Duyck3c9e9f72015-03-12 14:46:23 -0700159 if (!old)
160 return 0;
Alexander Duyck0ddcf432015-03-06 13:47:00 -0800161
Alexander Duyck3c9e9f72015-03-12 14:46:23 -0700162 new = fib_trie_unmerge(old);
Alexander Duyck0ddcf432015-03-06 13:47:00 -0800163 if (!new)
164 return -ENOMEM;
165
Alexander Duyck3b709332016-11-15 05:46:06 -0500166 /* table is already unmerged */
167 if (new == old)
168 return 0;
169
Alexander Duyck0ddcf432015-03-06 13:47:00 -0800170 /* replace merged table with clean table */
Alexander Duyck3b709332016-11-15 05:46:06 -0500171 fib_replace_table(net, old, new);
172 fib_free_table(old);
173
174 /* attempt to fetch main table if it has been allocated */
175 main_table = fib_get_table(net, RT_TABLE_MAIN);
176 if (!main_table)
177 return 0;
178
179 /* flush local entries from main table */
180 fib_table_flush_external(main_table);
Alexander Duyck0ddcf432015-03-06 13:47:00 -0800181
182 return 0;
183}
184
Denis V. Luneve4aef8a2008-01-10 03:28:24 -0800185static void fib_flush(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186{
187 int flushed = 0;
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700188 unsigned int h;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700190 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
Alexander Duycka7e53532015-03-04 15:02:44 -0800191 struct hlist_head *head = &net->ipv4.fib_table_hash[h];
192 struct hlist_node *tmp;
193 struct fib_table *tb;
194
195 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist)
Jiri Pirkob90eb752016-09-26 12:52:29 +0200196 flushed += fib_table_flush(net, tb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198
199 if (flushed)
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +0000200 rt_cache_flush(net);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201}
202
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800203/*
204 * Find address type as if only "dev" was present in the system. If
205 * on_dev is NULL then all interfaces are taken into consideration.
206 */
Eric Dumazet95c96172012-04-15 05:58:06 +0000207static inline unsigned int __inet_dev_addr_type(struct net *net,
208 const struct net_device *dev,
David Ahern9b8ff512015-09-01 14:26:35 -0600209 __be32 addr, u32 tb_id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210{
David S. Miller9ade2282011-03-12 02:02:42 -0500211 struct flowi4 fl4 = { .daddr = addr };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212 struct fib_result res;
Eric Dumazet95c96172012-04-15 05:58:06 +0000213 unsigned int ret = RTN_BROADCAST;
David Ahern15be405e2015-08-13 14:59:04 -0600214 struct fib_table *table;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215
Jan Engelhardt1e637c72008-01-21 03:18:08 -0800216 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 return RTN_BROADCAST;
Joe Perchesf97c1e02007-12-16 13:45:43 -0800218 if (ipv4_is_multicast(addr))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 return RTN_MULTICAST;
220
Alexander Duyck345e9b52014-12-31 10:56:24 -0800221 rcu_read_lock();
222
David Ahern15be405e2015-08-13 14:59:04 -0600223 table = fib_get_table(net, tb_id);
224 if (table) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 ret = RTN_UNICAST;
David Ahern15be405e2015-08-13 14:59:04 -0600226 if (!fib_table_lookup(table, &fl4, &res, FIB_LOOKUP_NOREF)) {
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800227 if (!dev || dev == res.fi->fib_dev)
228 ret = res.type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 }
230 }
Alexander Duyck345e9b52014-12-31 10:56:24 -0800231
232 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233 return ret;
234}
235
David Ahern9b8ff512015-09-01 14:26:35 -0600236unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id)
David Ahern15be405e2015-08-13 14:59:04 -0600237{
238 return __inet_dev_addr_type(net, NULL, addr, tb_id);
239}
240EXPORT_SYMBOL(inet_addr_type_table);
241
Eric W. Biederman6b175b22008-01-10 03:25:28 -0800242unsigned int inet_addr_type(struct net *net, __be32 addr)
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800243{
David Ahern15be405e2015-08-13 14:59:04 -0600244 return __inet_dev_addr_type(net, NULL, addr, RT_TABLE_LOCAL);
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800245}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000246EXPORT_SYMBOL(inet_addr_type);
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800247
Eric W. Biederman6b175b22008-01-10 03:25:28 -0800248unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
249 __be32 addr)
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800250{
David Ahern3236b002015-09-29 20:07:14 -0700251 u32 rt_table = l3mdev_fib_table(dev) ? : RT_TABLE_LOCAL;
David Ahern15be405e2015-08-13 14:59:04 -0600252
253 return __inet_dev_addr_type(net, dev, addr, rt_table);
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800254}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000255EXPORT_SYMBOL(inet_dev_addr_type);
Laszlo Attila Toth05538112007-12-04 23:28:46 -0800256
David Ahern30bbaa12015-08-13 14:59:05 -0600257/* inet_addr_type with dev == NULL but using the table from a dev
258 * if one is associated
259 */
260unsigned int inet_addr_type_dev_table(struct net *net,
261 const struct net_device *dev,
262 __be32 addr)
263{
David Ahern3236b002015-09-29 20:07:14 -0700264 u32 rt_table = l3mdev_fib_table(dev) ? : RT_TABLE_LOCAL;
David Ahern30bbaa12015-08-13 14:59:05 -0600265
266 return __inet_dev_addr_type(net, NULL, addr, rt_table);
267}
268EXPORT_SYMBOL(inet_addr_type_dev_table);
269
David S. Miller35ebf652012-06-28 03:59:11 -0700270__be32 fib_compute_spec_dst(struct sk_buff *skb)
271{
272 struct net_device *dev = skb->dev;
273 struct in_device *in_dev;
274 struct fib_result res;
David S. Millera207a4b2012-06-28 18:33:24 -0700275 struct rtable *rt;
David S. Miller35ebf652012-06-28 03:59:11 -0700276 struct net *net;
David S. Millera207a4b2012-06-28 18:33:24 -0700277 int scope;
David S. Miller35ebf652012-06-28 03:59:11 -0700278
David S. Millera207a4b2012-06-28 18:33:24 -0700279 rt = skb_rtable(skb);
Julian Anastasov0cc535a2012-07-18 21:35:03 +0000280 if ((rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | RTCF_LOCAL)) ==
281 RTCF_LOCAL)
David S. Miller35ebf652012-06-28 03:59:11 -0700282 return ip_hdr(skb)->daddr;
283
284 in_dev = __in_dev_get_rcu(dev);
285 BUG_ON(!in_dev);
David S. Miller35ebf652012-06-28 03:59:11 -0700286
287 net = dev_net(dev);
David S. Millera207a4b2012-06-28 18:33:24 -0700288
289 scope = RT_SCOPE_UNIVERSE;
290 if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) {
Lance Richardson4cfc86f2016-03-22 14:56:57 -0400291 struct flowi4 fl4 = {
292 .flowi4_iif = LOOPBACK_IFINDEX,
293 .daddr = ip_hdr(skb)->saddr,
294 .flowi4_tos = RT_TOS(ip_hdr(skb)->tos),
295 .flowi4_scope = scope,
296 .flowi4_mark = IN_DEV_SRC_VMARK(in_dev) ? skb->mark : 0,
297 };
Andy Gospodarek0eeb0752015-06-23 13:45:37 -0400298 if (!fib_lookup(net, &fl4, &res, 0))
David S. Millera207a4b2012-06-28 18:33:24 -0700299 return FIB_RES_PREFSRC(net, res);
300 } else {
301 scope = RT_SCOPE_LINK;
302 }
303
304 return inet_select_addr(dev, ip_hdr(skb)->saddr, scope);
David S. Miller35ebf652012-06-28 03:59:11 -0700305}
306
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307/* Given (packet source, input interface) and optional (dst, oif, tos):
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000308 * - (main) check, that source is valid i.e. not broadcast or our local
309 * address.
310 * - figure out what "logical" interface this packet arrived
311 * and calculate "specific destination" address.
312 * - check, that packet arrived from expected physical interface.
Eric Dumazetebc0ffa2010-10-05 10:41:36 +0000313 * called with rcu_read_lock()
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 */
David S. Miller7a9bc9b2012-06-29 01:32:45 -0700315static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
316 u8 tos, int oif, struct net_device *dev,
317 int rpf, struct in_device *idev, u32 *itag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318{
Sébastien Barré1dced6a2014-08-17 09:19:54 +0200319 int ret, no_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320 struct fib_result res;
David S. Miller9e56e382012-06-28 18:54:02 -0700321 struct flowi4 fl4;
Julian Anastasov841c0b72017-02-26 15:50:52 +0200322 struct net *net = dev_net(dev);
David S. Miller9e56e382012-06-28 18:54:02 -0700323 bool dev_match;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324
David S. Miller9ade2282011-03-12 02:02:42 -0500325 fl4.flowi4_oif = 0;
David Ahern385add92015-09-29 20:07:13 -0700326 fl4.flowi4_iif = l3mdev_master_ifindex_rcu(dev);
David Aherncd2fbe12015-08-13 14:59:01 -0600327 if (!fl4.flowi4_iif)
328 fl4.flowi4_iif = oif ? : LOOPBACK_IFINDEX;
David S. Miller9ade2282011-03-12 02:02:42 -0500329 fl4.daddr = src;
330 fl4.saddr = dst;
331 fl4.flowi4_tos = tos;
332 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
Thomas Graf1b7179d2015-07-21 10:43:59 +0200333 fl4.flowi4_tun_key.tun_id = 0;
David Ahernb84f7872015-09-29 19:07:07 -0700334 fl4.flowi4_flags = 0;
Julian Anastasov841c0b72017-02-26 15:50:52 +0200335 fl4.flowi4_uid = sock_net_uid(net, NULL);
David S. Millercc7e17e2011-03-09 20:57:50 -0800336
David S. Miller9e56e382012-06-28 18:54:02 -0700337 no_addr = idev->ifa_list == NULL;
Michael Smith990078a2011-04-07 04:51:51 +0000338
David S. Miller9e56e382012-06-28 18:54:02 -0700339 fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340
David Ahernf6d3c192015-08-28 08:42:09 -0700341 trace_fib_validate_source(dev, &fl4);
342
Andy Gospodarek0eeb0752015-06-23 13:45:37 -0400343 if (fib_lookup(net, &fl4, &res, 0))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 goto last_resort;
Sébastien Barré1dced6a2014-08-17 09:19:54 +0200345 if (res.type != RTN_UNICAST &&
346 (res.type != RTN_LOCAL || !IN_DEV_ACCEPT_LOCAL(idev)))
347 goto e_inval;
Julian Anastasov841c0b72017-02-26 15:50:52 +0200348 if (!rpf && !fib_num_tclassid_users(net) &&
Sébastien Barré1dced6a2014-08-17 09:19:54 +0200349 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev)))
350 goto last_resort;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 fib_combine_itag(itag, &res);
David S. Miller6f86b322010-09-06 22:36:19 -0700352 dev_match = false;
353
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354#ifdef CONFIG_IP_ROUTE_MULTIPATH
David S. Miller6f86b322010-09-06 22:36:19 -0700355 for (ret = 0; ret < res.fi->fib_nhs; ret++) {
356 struct fib_nh *nh = &res.fi->fib_nh[ret];
357
358 if (nh->nh_dev == dev) {
359 dev_match = true;
360 break;
David Ahern385add92015-09-29 20:07:13 -0700361 } else if (l3mdev_master_ifindex_rcu(nh->nh_dev) == dev->ifindex) {
David Aherncd2fbe12015-08-13 14:59:01 -0600362 dev_match = true;
363 break;
David S. Miller6f86b322010-09-06 22:36:19 -0700364 }
365 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366#else
367 if (FIB_RES_DEV(res) == dev)
David S. Miller6f86b322010-09-06 22:36:19 -0700368 dev_match = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369#endif
David S. Miller6f86b322010-09-06 22:36:19 -0700370 if (dev_match) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 return ret;
373 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 if (no_addr)
375 goto last_resort;
Stephen Hemmingerc1cf8422009-02-20 08:25:36 +0000376 if (rpf == 1)
Eric Dumazetb5f7e752010-06-02 12:05:27 +0000377 goto e_rpf;
David S. Miller9ade2282011-03-12 02:02:42 -0500378 fl4.flowi4_oif = dev->ifindex;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379
380 ret = 0;
Andy Gospodarek0eeb0752015-06-23 13:45:37 -0400381 if (fib_lookup(net, &fl4, &res, FIB_LOOKUP_IGNORE_LINKSTATE) == 0) {
David S. Miller41347dc2012-06-28 04:05:27 -0700382 if (res.type == RTN_UNICAST)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 }
385 return ret;
386
387last_resort:
388 if (rpf)
Eric Dumazetb5f7e752010-06-02 12:05:27 +0000389 goto e_rpf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 *itag = 0;
391 return 0;
392
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393e_inval:
394 return -EINVAL;
Eric Dumazetb5f7e752010-06-02 12:05:27 +0000395e_rpf:
396 return -EXDEV;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397}
398
David S. Miller7a9bc9b2012-06-29 01:32:45 -0700399/* Ignore rp_filter for packets protected by IPsec. */
400int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
401 u8 tos, int oif, struct net_device *dev,
402 struct in_device *idev, u32 *itag)
403{
404 int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev);
405
Julian Anastasove81da0e2012-10-08 11:41:15 +0000406 if (!r && !fib_num_tclassid_users(dev_net(dev)) &&
Sébastien Barré1dced6a2014-08-17 09:19:54 +0200407 IN_DEV_ACCEPT_LOCAL(idev) &&
Julian Anastasove81da0e2012-10-08 11:41:15 +0000408 (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) {
David S. Miller7a9bc9b2012-06-29 01:32:45 -0700409 *itag = 0;
410 return 0;
411 }
412 return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag);
413}
414
Al Viro81f7bf62006-09-27 18:40:00 -0700415static inline __be32 sk_extract_addr(struct sockaddr *addr)
Thomas Graf4e902c52006-08-17 18:14:52 -0700416{
417 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
418}
419
420static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
421{
422 struct nlattr *nla;
423
424 nla = (struct nlattr *) ((char *) mx + len);
425 nla->nla_type = type;
426 nla->nla_len = nla_attr_size(4);
427 *(u32 *) nla_data(nla) = value;
428
429 return len + nla_total_size(4);
430}
431
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800432static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
Thomas Graf4e902c52006-08-17 18:14:52 -0700433 struct fib_config *cfg)
434{
Al Viro6d85c102006-09-26 22:15:46 -0700435 __be32 addr;
Thomas Graf4e902c52006-08-17 18:14:52 -0700436 int plen;
437
438 memset(cfg, 0, sizeof(*cfg));
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800439 cfg->fc_nlinfo.nl_net = net;
Thomas Graf4e902c52006-08-17 18:14:52 -0700440
441 if (rt->rt_dst.sa_family != AF_INET)
442 return -EAFNOSUPPORT;
443
444 /*
445 * Check mask for validity:
446 * a) it must be contiguous.
447 * b) destination must have all host bits clear.
448 * c) if application forgot to set correct family (AF_INET),
449 * reject request unless it is absolutely clear i.e.
450 * both family and mask are zero.
451 */
452 plen = 32;
453 addr = sk_extract_addr(&rt->rt_dst);
454 if (!(rt->rt_flags & RTF_HOST)) {
Al Viro81f7bf62006-09-27 18:40:00 -0700455 __be32 mask = sk_extract_addr(&rt->rt_genmask);
Thomas Graf4e902c52006-08-17 18:14:52 -0700456
457 if (rt->rt_genmask.sa_family != AF_INET) {
458 if (mask || rt->rt_genmask.sa_family)
459 return -EAFNOSUPPORT;
460 }
461
462 if (bad_mask(mask, addr))
463 return -EINVAL;
464
465 plen = inet_mask_len(mask);
466 }
467
468 cfg->fc_dst_len = plen;
469 cfg->fc_dst = addr;
470
471 if (cmd != SIOCDELRT) {
472 cfg->fc_nlflags = NLM_F_CREATE;
473 cfg->fc_protocol = RTPROT_BOOT;
474 }
475
476 if (rt->rt_metric)
477 cfg->fc_priority = rt->rt_metric - 1;
478
479 if (rt->rt_flags & RTF_REJECT) {
480 cfg->fc_scope = RT_SCOPE_HOST;
481 cfg->fc_type = RTN_UNREACHABLE;
482 return 0;
483 }
484
485 cfg->fc_scope = RT_SCOPE_NOWHERE;
486 cfg->fc_type = RTN_UNICAST;
487
488 if (rt->rt_dev) {
489 char *colon;
490 struct net_device *dev;
491 char devname[IFNAMSIZ];
492
493 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
494 return -EFAULT;
495
496 devname[IFNAMSIZ-1] = 0;
497 colon = strchr(devname, ':');
498 if (colon)
499 *colon = 0;
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800500 dev = __dev_get_by_name(net, devname);
Thomas Graf4e902c52006-08-17 18:14:52 -0700501 if (!dev)
502 return -ENODEV;
503 cfg->fc_oif = dev->ifindex;
Mark Tomlinson5a56a0b2016-09-05 10:20:20 +1200504 cfg->fc_table = l3mdev_fib_table(dev);
Thomas Graf4e902c52006-08-17 18:14:52 -0700505 if (colon) {
506 struct in_ifaddr *ifa;
507 struct in_device *in_dev = __in_dev_get_rtnl(dev);
508 if (!in_dev)
509 return -ENODEV;
510 *colon = ':';
511 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
512 if (strcmp(ifa->ifa_label, devname) == 0)
513 break;
Ian Morris51456b22015-04-03 09:17:26 +0100514 if (!ifa)
Thomas Graf4e902c52006-08-17 18:14:52 -0700515 return -ENODEV;
516 cfg->fc_prefsrc = ifa->ifa_local;
517 }
518 }
519
520 addr = sk_extract_addr(&rt->rt_gateway);
521 if (rt->rt_gateway.sa_family == AF_INET && addr) {
David Ahern30bbaa12015-08-13 14:59:05 -0600522 unsigned int addr_type;
523
Thomas Graf4e902c52006-08-17 18:14:52 -0700524 cfg->fc_gw = addr;
David Ahern30bbaa12015-08-13 14:59:05 -0600525 addr_type = inet_addr_type_table(net, addr, cfg->fc_table);
Thomas Graf4e902c52006-08-17 18:14:52 -0700526 if (rt->rt_flags & RTF_GATEWAY &&
David Ahern30bbaa12015-08-13 14:59:05 -0600527 addr_type == RTN_UNICAST)
Thomas Graf4e902c52006-08-17 18:14:52 -0700528 cfg->fc_scope = RT_SCOPE_UNIVERSE;
529 }
530
531 if (cmd == SIOCDELRT)
532 return 0;
533
534 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
535 return -EINVAL;
536
537 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
538 cfg->fc_scope = RT_SCOPE_LINK;
539
540 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
541 struct nlattr *mx;
542 int len = 0;
543
544 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
Ian Morris51456b22015-04-03 09:17:26 +0100545 if (!mx)
Thomas Graf4e902c52006-08-17 18:14:52 -0700546 return -ENOMEM;
547
548 if (rt->rt_flags & RTF_MTU)
549 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
550
551 if (rt->rt_flags & RTF_WINDOW)
552 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
553
554 if (rt->rt_flags & RTF_IRTT)
555 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
556
557 cfg->fc_mx = mx;
558 cfg->fc_mx_len = len;
559 }
560
561 return 0;
562}
563
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564/*
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000565 * Handle IP routing ioctl calls.
566 * These are used to manipulate the routing tables
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 */
Denis V. Lunev1bad1182008-01-10 03:29:53 -0800568int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569{
Thomas Graf4e902c52006-08-17 18:14:52 -0700570 struct fib_config cfg;
571 struct rtentry rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573
574 switch (cmd) {
575 case SIOCADDRT: /* Add a route */
576 case SIOCDELRT: /* Delete a route */
Eric W. Biederman52e804c2012-11-16 03:03:05 +0000577 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 return -EPERM;
Thomas Graf4e902c52006-08-17 18:14:52 -0700579
580 if (copy_from_user(&rt, arg, sizeof(rt)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 return -EFAULT;
Thomas Graf4e902c52006-08-17 18:14:52 -0700582
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583 rtnl_lock();
Denis V. Lunev1bad1182008-01-10 03:29:53 -0800584 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585 if (err == 0) {
Thomas Graf4e902c52006-08-17 18:14:52 -0700586 struct fib_table *tb;
587
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588 if (cmd == SIOCDELRT) {
Denis V. Lunev1bad1182008-01-10 03:29:53 -0800589 tb = fib_get_table(net, cfg.fc_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590 if (tb)
Jiri Pirkob90eb752016-09-26 12:52:29 +0200591 err = fib_table_delete(net, tb, &cfg);
Thomas Graf4e902c52006-08-17 18:14:52 -0700592 else
593 err = -ESRCH;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594 } else {
Denis V. Lunev1bad1182008-01-10 03:29:53 -0800595 tb = fib_new_table(net, cfg.fc_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 if (tb)
Jiri Pirkob90eb752016-09-26 12:52:29 +0200597 err = fib_table_insert(net, tb, &cfg);
Thomas Graf4e902c52006-08-17 18:14:52 -0700598 else
599 err = -ENOBUFS;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 }
Thomas Graf4e902c52006-08-17 18:14:52 -0700601
602 /* allocated by rtentry_to_fib_config() */
603 kfree(cfg.fc_mx);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 }
605 rtnl_unlock();
606 return err;
607 }
608 return -EINVAL;
609}
610
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000611const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
Thomas Graf4e902c52006-08-17 18:14:52 -0700612 [RTA_DST] = { .type = NLA_U32 },
613 [RTA_SRC] = { .type = NLA_U32 },
614 [RTA_IIF] = { .type = NLA_U32 },
615 [RTA_OIF] = { .type = NLA_U32 },
616 [RTA_GATEWAY] = { .type = NLA_U32 },
617 [RTA_PRIORITY] = { .type = NLA_U32 },
618 [RTA_PREFSRC] = { .type = NLA_U32 },
619 [RTA_METRICS] = { .type = NLA_NESTED },
Thomas Graf5176f912006-08-26 20:13:18 -0700620 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
Thomas Graf4e902c52006-08-17 18:14:52 -0700621 [RTA_FLOW] = { .type = NLA_U32 },
Roopa Prabhu571e7222015-07-21 10:43:47 +0200622 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
623 [RTA_ENCAP] = { .type = NLA_NESTED },
Lorenzo Colitti2ec93fe2016-11-04 02:23:42 +0900624 [RTA_UID] = { .type = NLA_U32 },
Thomas Graf4e902c52006-08-17 18:14:52 -0700625};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800627static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000628 struct nlmsghdr *nlh, struct fib_config *cfg)
Thomas Graf4e902c52006-08-17 18:14:52 -0700629{
630 struct nlattr *attr;
631 int err, remaining;
632 struct rtmsg *rtm;
633
634 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
635 if (err < 0)
636 goto errout;
637
638 memset(cfg, 0, sizeof(*cfg));
639
640 rtm = nlmsg_data(nlh);
Thomas Graf4e902c52006-08-17 18:14:52 -0700641 cfg->fc_dst_len = rtm->rtm_dst_len;
Thomas Graf4e902c52006-08-17 18:14:52 -0700642 cfg->fc_tos = rtm->rtm_tos;
643 cfg->fc_table = rtm->rtm_table;
644 cfg->fc_protocol = rtm->rtm_protocol;
645 cfg->fc_scope = rtm->rtm_scope;
646 cfg->fc_type = rtm->rtm_type;
647 cfg->fc_flags = rtm->rtm_flags;
648 cfg->fc_nlflags = nlh->nlmsg_flags;
649
Eric W. Biederman15e47302012-09-07 20:12:54 +0000650 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
Thomas Graf4e902c52006-08-17 18:14:52 -0700651 cfg->fc_nlinfo.nlh = nlh;
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800652 cfg->fc_nlinfo.nl_net = net;
Thomas Graf4e902c52006-08-17 18:14:52 -0700653
Thomas Grafa0ee18b2007-03-24 20:32:54 -0700654 if (cfg->fc_type > RTN_MAX) {
655 err = -EINVAL;
656 goto errout;
657 }
658
Thomas Graf4e902c52006-08-17 18:14:52 -0700659 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
Thomas Graf8f4c1f92007-09-12 14:44:36 +0200660 switch (nla_type(attr)) {
Thomas Graf4e902c52006-08-17 18:14:52 -0700661 case RTA_DST:
Al Viro17fb2c62006-09-26 22:15:25 -0700662 cfg->fc_dst = nla_get_be32(attr);
Thomas Graf4e902c52006-08-17 18:14:52 -0700663 break;
Thomas Graf4e902c52006-08-17 18:14:52 -0700664 case RTA_OIF:
665 cfg->fc_oif = nla_get_u32(attr);
666 break;
667 case RTA_GATEWAY:
Al Viro17fb2c62006-09-26 22:15:25 -0700668 cfg->fc_gw = nla_get_be32(attr);
Thomas Graf4e902c52006-08-17 18:14:52 -0700669 break;
670 case RTA_PRIORITY:
671 cfg->fc_priority = nla_get_u32(attr);
672 break;
673 case RTA_PREFSRC:
Al Viro17fb2c62006-09-26 22:15:25 -0700674 cfg->fc_prefsrc = nla_get_be32(attr);
Thomas Graf4e902c52006-08-17 18:14:52 -0700675 break;
676 case RTA_METRICS:
677 cfg->fc_mx = nla_data(attr);
678 cfg->fc_mx_len = nla_len(attr);
679 break;
680 case RTA_MULTIPATH:
David Aherne9db0422017-01-17 14:57:36 -0800681 err = lwtunnel_valid_encap_type_attr(nla_data(attr),
682 nla_len(attr));
683 if (err < 0)
684 goto errout;
Thomas Graf4e902c52006-08-17 18:14:52 -0700685 cfg->fc_mp = nla_data(attr);
686 cfg->fc_mp_len = nla_len(attr);
687 break;
688 case RTA_FLOW:
689 cfg->fc_flow = nla_get_u32(attr);
690 break;
Thomas Graf4e902c52006-08-17 18:14:52 -0700691 case RTA_TABLE:
692 cfg->fc_table = nla_get_u32(attr);
693 break;
Roopa Prabhu571e7222015-07-21 10:43:47 +0200694 case RTA_ENCAP:
695 cfg->fc_encap = attr;
696 break;
697 case RTA_ENCAP_TYPE:
698 cfg->fc_encap_type = nla_get_u16(attr);
David Aherne9db0422017-01-17 14:57:36 -0800699 err = lwtunnel_valid_encap_type(cfg->fc_encap_type);
700 if (err < 0)
701 goto errout;
Roopa Prabhu571e7222015-07-21 10:43:47 +0200702 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
Jiri Pirkob90eb752016-09-26 12:52:29 +0200728 err = fib_table_delete(net, 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
Jiri Pirkob90eb752016-09-26 12:52:29 +0200750 err = fib_table_insert(net, 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;
David Ahern68647612017-05-15 23:19:17 -0700762 int dumped = 0, err;
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]));
David Ahern68647612017-05-15 23:19:17 -0700782 err = fib_table_dump(tb, skb, cb);
783 if (err < 0) {
784 if (likely(skb->len))
785 goto out;
786
787 goto out_err;
788 }
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700789 dumped = 1;
790next:
791 e++;
792 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 }
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700794out:
David Ahern68647612017-05-15 23:19:17 -0700795 err = skb->len;
796out_err:
Alexander Duycka7e53532015-03-04 15:02:44 -0800797 rcu_read_unlock();
798
Patrick McHardy1af5a8c2006-08-10 23:10:46 -0700799 cb->args[1] = e;
800 cb->args[0] = h;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801
David Ahern68647612017-05-15 23:19:17 -0700802 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803}
804
805/* Prepare and feed intra-kernel routing request.
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000806 * Really, it should be netlink message, but :-( netlink
807 * can be not configured, so that we feed it directly
808 * to fib engine. It is legal, because all events occur
809 * only when netlink is already locked.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 */
Al Viro81f7bf62006-09-27 18:40:00 -0700811static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812{
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900813 struct net *net = dev_net(ifa->ifa_dev->dev);
David Ahern3236b002015-09-29 20:07:14 -0700814 u32 tb_id = l3mdev_fib_table(ifa->ifa_dev->dev);
Thomas Graf4e902c52006-08-17 18:14:52 -0700815 struct fib_table *tb;
816 struct fib_config cfg = {
817 .fc_protocol = RTPROT_KERNEL,
818 .fc_type = type,
819 .fc_dst = dst,
820 .fc_dst_len = dst_len,
821 .fc_prefsrc = ifa->ifa_local,
822 .fc_oif = ifa->ifa_dev->dev->ifindex,
823 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
Denis V. Lunev4d1169c2008-01-10 03:26:13 -0800824 .fc_nlinfo = {
Denis V. Lunev4b5d47d2008-01-10 03:29:23 -0800825 .nl_net = net,
Denis V. Lunev4d1169c2008-01-10 03:26:13 -0800826 },
Thomas Graf4e902c52006-08-17 18:14:52 -0700827 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828
David Ahern021dd3b2015-08-13 14:59:06 -0600829 if (!tb_id)
830 tb_id = (type == RTN_UNICAST) ? RT_TABLE_MAIN : RT_TABLE_LOCAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831
David Ahern021dd3b2015-08-13 14:59:06 -0600832 tb = fib_new_table(net, tb_id);
Ian Morris51456b22015-04-03 09:17:26 +0100833 if (!tb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834 return;
835
Thomas Graf4e902c52006-08-17 18:14:52 -0700836 cfg.fc_table = tb->tb_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837
Thomas Graf4e902c52006-08-17 18:14:52 -0700838 if (type != RTN_LOCAL)
839 cfg.fc_scope = RT_SCOPE_LINK;
840 else
841 cfg.fc_scope = RT_SCOPE_HOST;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842
843 if (cmd == RTM_NEWROUTE)
Jiri Pirkob90eb752016-09-26 12:52:29 +0200844 fib_table_insert(net, tb, &cfg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 else
Jiri Pirkob90eb752016-09-26 12:52:29 +0200846 fib_table_delete(net, tb, &cfg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847}
848
Jamal Hadi Salim0ff60a42005-11-22 14:47:37 -0800849void fib_add_ifaddr(struct in_ifaddr *ifa)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850{
851 struct in_device *in_dev = ifa->ifa_dev;
852 struct net_device *dev = in_dev->dev;
853 struct in_ifaddr *prim = ifa;
Al Viroa144ea42006-09-28 18:00:55 -0700854 __be32 mask = ifa->ifa_mask;
855 __be32 addr = ifa->ifa_local;
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000856 __be32 prefix = ifa->ifa_address & mask;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000858 if (ifa->ifa_flags & IFA_F_SECONDARY) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 prim = inet_ifa_byprefix(in_dev, prefix, mask);
Ian Morris51456b22015-04-03 09:17:26 +0100860 if (!prim) {
Joe Perches058bd4d2012-03-11 18:36:11 +0000861 pr_warn("%s: bug: prim == NULL\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 return;
863 }
864 }
865
866 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
867
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000868 if (!(dev->flags & IFF_UP))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 return;
870
871 /* Add broadcast address, if it is explicitly assigned. */
Al Viroa144ea42006-09-28 18:00:55 -0700872 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
874
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000875 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876 (prefix != addr || ifa->ifa_prefixlen < 32)) {
Paolo Abeni7b131182015-10-20 10:28:45 +0200877 if (!(ifa->ifa_flags & IFA_F_NOPREFIXROUTE))
878 fib_magic(RTM_NEWROUTE,
879 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
880 prefix, ifa->ifa_prefixlen, prim);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881
882 /* Add network specific broadcasts, when it takes a sense */
883 if (ifa->ifa_prefixlen < 31) {
884 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000885 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
886 32, prim);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 }
888 }
889}
890
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000891/* Delete primary or secondary address.
892 * Optionally, on secondary address promotion consider the addresses
893 * from subnet iprim as deleted, even if they are in device list.
894 * In this case the secondary ifa can be in device list.
895 */
896void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897{
898 struct in_device *in_dev = ifa->ifa_dev;
899 struct net_device *dev = in_dev->dev;
900 struct in_ifaddr *ifa1;
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000901 struct in_ifaddr *prim = ifa, *prim1 = NULL;
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000902 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
903 __be32 any = ifa->ifa_address & ifa->ifa_mask;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904#define LOCAL_OK 1
905#define BRD_OK 2
906#define BRD0_OK 4
907#define BRD1_OK 8
Eric Dumazet95c96172012-04-15 05:58:06 +0000908 unsigned int ok = 0;
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000909 int subnet = 0; /* Primary network */
910 int gone = 1; /* Address is missing */
911 int same_prefsrc = 0; /* Another primary with same IP */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000913 if (ifa->ifa_flags & IFA_F_SECONDARY) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
Ian Morris51456b22015-04-03 09:17:26 +0100915 if (!prim) {
Paolo Abeni391a2032016-04-21 22:23:31 +0200916 /* if the device has been deleted, we don't perform
917 * address promotion
918 */
919 if (!in_dev->dead)
920 pr_warn("%s: bug: prim == NULL\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921 return;
922 }
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000923 if (iprim && iprim != prim) {
Joe Perches058bd4d2012-03-11 18:36:11 +0000924 pr_warn("%s: bug: iprim != prim\n", __func__);
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000925 return;
926 }
927 } else if (!ipv4_is_zeronet(any) &&
928 (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) {
Paolo Abeni7b131182015-10-20 10:28:45 +0200929 if (!(ifa->ifa_flags & IFA_F_NOPREFIXROUTE))
930 fib_magic(RTM_DELROUTE,
931 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
932 any, ifa->ifa_prefixlen, prim);
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000933 subnet = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934 }
935
David S. Millerfbd40ea2016-03-13 23:28:00 -0400936 if (in_dev->dead)
937 goto no_promotions;
938
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 /* Deletion is more complicated than add.
Eric Dumazet6a31d2a2010-10-04 20:00:18 +0000940 * We should take care of not to delete too much :-)
941 *
942 * Scan address list to be sure that addresses are really gone.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943 */
944
945 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000946 if (ifa1 == ifa) {
947 /* promotion, keep the IP */
948 gone = 0;
949 continue;
950 }
951 /* Ignore IFAs from our subnet */
952 if (iprim && ifa1->ifa_mask == iprim->ifa_mask &&
953 inet_ifa_match(ifa1->ifa_address, iprim))
954 continue;
955
956 /* Ignore ifa1 if it uses different primary IP (prefsrc) */
957 if (ifa1->ifa_flags & IFA_F_SECONDARY) {
958 /* Another address from our subnet? */
959 if (ifa1->ifa_mask == prim->ifa_mask &&
960 inet_ifa_match(ifa1->ifa_address, prim))
961 prim1 = prim;
962 else {
963 /* We reached the secondaries, so
964 * same_prefsrc should be determined.
965 */
966 if (!same_prefsrc)
967 continue;
968 /* Search new prim1 if ifa1 is not
969 * using the current prim1
970 */
971 if (!prim1 ||
972 ifa1->ifa_mask != prim1->ifa_mask ||
973 !inet_ifa_match(ifa1->ifa_address, prim1))
974 prim1 = inet_ifa_byprefix(in_dev,
975 ifa1->ifa_address,
976 ifa1->ifa_mask);
977 if (!prim1)
978 continue;
979 if (prim1->ifa_local != prim->ifa_local)
980 continue;
981 }
982 } else {
983 if (prim->ifa_local != ifa1->ifa_local)
984 continue;
985 prim1 = ifa1;
986 if (prim != prim1)
987 same_prefsrc = 1;
988 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 if (ifa->ifa_local == ifa1->ifa_local)
990 ok |= LOCAL_OK;
991 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
992 ok |= BRD_OK;
993 if (brd == ifa1->ifa_broadcast)
994 ok |= BRD1_OK;
995 if (any == ifa1->ifa_broadcast)
996 ok |= BRD0_OK;
Julian Anastasove6abbaa2011-03-19 12:13:49 +0000997 /* primary has network specific broadcasts */
998 if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) {
999 __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask;
1000 __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask;
1001
1002 if (!ipv4_is_zeronet(any1)) {
1003 if (ifa->ifa_broadcast == brd1 ||
1004 ifa->ifa_broadcast == any1)
1005 ok |= BRD_OK;
1006 if (brd == brd1 || brd == any1)
1007 ok |= BRD1_OK;
1008 if (any == brd1 || any == any1)
1009 ok |= BRD0_OK;
1010 }
1011 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 }
1013
David S. Millerfbd40ea2016-03-13 23:28:00 -04001014no_promotions:
Eric Dumazet6a31d2a2010-10-04 20:00:18 +00001015 if (!(ok & BRD_OK))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
Julian Anastasove6abbaa2011-03-19 12:13:49 +00001017 if (subnet && ifa->ifa_prefixlen < 31) {
1018 if (!(ok & BRD1_OK))
1019 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
1020 if (!(ok & BRD0_OK))
1021 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
1022 }
Eric Dumazet6a31d2a2010-10-04 20:00:18 +00001023 if (!(ok & LOCAL_OK)) {
David Ahern30bbaa12015-08-13 14:59:05 -06001024 unsigned int addr_type;
1025
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
1027
1028 /* Check, that this local address finally disappeared. */
David Ahern30bbaa12015-08-13 14:59:05 -06001029 addr_type = inet_addr_type_dev_table(dev_net(dev), dev,
1030 ifa->ifa_local);
1031 if (gone && addr_type != RTN_LOCAL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 /* And the last, but not the least thing.
Eric Dumazet6a31d2a2010-10-04 20:00:18 +00001033 * We must flush stray FIB entries.
1034 *
1035 * First of all, we scan fib_info list searching
1036 * for stray nexthop entries, then ignite fib_flush.
1037 */
Mark Tomlinson5a56a0b2016-09-05 10:20:20 +12001038 if (fib_sync_down_addr(dev, ifa->ifa_local))
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001039 fib_flush(dev_net(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 }
1041 }
1042#undef LOCAL_OK
1043#undef BRD_OK
1044#undef BRD0_OK
1045#undef BRD1_OK
1046}
1047
Alexander Duyck345e9b52014-12-31 10:56:24 -08001048static void nl_fib_lookup(struct net *net, struct fib_result_nl *frn)
Robert Olsson246955f2005-06-20 13:36:39 -07001049{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001050
Robert Olsson246955f2005-06-20 13:36:39 -07001051 struct fib_result res;
David S. Miller9ade2282011-03-12 02:02:42 -05001052 struct flowi4 fl4 = {
1053 .flowi4_mark = frn->fl_mark,
1054 .daddr = frn->fl_addr,
1055 .flowi4_tos = frn->fl_tos,
1056 .flowi4_scope = frn->fl_scope,
Eric Dumazet6a31d2a2010-10-04 20:00:18 +00001057 };
Alexander Duyck345e9b52014-12-31 10:56:24 -08001058 struct fib_table *tb;
1059
1060 rcu_read_lock();
1061
1062 tb = fib_get_table(net, frn->tb_id_in);
Alexey Kuznetsov1194ed02007-04-25 13:07:28 -07001063
1064 frn->err = -ENOENT;
Robert Olsson246955f2005-06-20 13:36:39 -07001065 if (tb) {
1066 local_bh_disable();
1067
1068 frn->tb_id = tb->tb_id;
David S. Miller9ade2282011-03-12 02:02:42 -05001069 frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF);
Robert Olsson246955f2005-06-20 13:36:39 -07001070
1071 if (!frn->err) {
1072 frn->prefixlen = res.prefixlen;
1073 frn->nh_sel = res.nh_sel;
1074 frn->type = res.type;
1075 frn->scope = res.scope;
1076 }
1077 local_bh_enable();
1078 }
Alexander Duyck345e9b52014-12-31 10:56:24 -08001079
1080 rcu_read_unlock();
Robert Olsson246955f2005-06-20 13:36:39 -07001081}
1082
David S. Miller28f7b0362007-10-10 21:32:39 -07001083static void nl_fib_input(struct sk_buff *skb)
Robert Olsson246955f2005-06-20 13:36:39 -07001084{
Denis V. Lunev6bd48fc2008-01-10 03:28:55 -08001085 struct net *net;
Robert Olsson246955f2005-06-20 13:36:39 -07001086 struct fib_result_nl *frn;
David S. Miller28f7b0362007-10-10 21:32:39 -07001087 struct nlmsghdr *nlh;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001088 u32 portid;
Alexey Kuznetsov1194ed02007-04-25 13:07:28 -07001089
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001090 net = sock_net(skb->sk);
Arnaldo Carvalho de Melob529ccf2007-04-25 19:08:35 -07001091 nlh = nlmsg_hdr(skb);
Eric Dumazet62e85fe2017-03-21 19:22:28 -07001092 if (skb->len < nlmsg_total_size(sizeof(*frn)) ||
1093 skb->len < nlh->nlmsg_len ||
Hong zhi guo573ce262013-03-27 06:47:04 +00001094 nlmsg_len(nlh) < sizeof(*frn))
Thomas Grafea865752005-12-01 14:30:00 -08001095 return;
Denis V. Lunevd883a032007-12-21 02:01:53 -08001096
Pablo Neira3a365152013-06-28 03:04:23 +02001097 skb = netlink_skb_clone(skb, GFP_KERNEL);
Ian Morris51456b22015-04-03 09:17:26 +01001098 if (!skb)
Denis V. Lunevd883a032007-12-21 02:01:53 -08001099 return;
1100 nlh = nlmsg_hdr(skb);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001101
Hong zhi guo573ce262013-03-27 06:47:04 +00001102 frn = (struct fib_result_nl *) nlmsg_data(nlh);
Alexander Duyck345e9b52014-12-31 10:56:24 -08001103 nl_fib_lookup(net, frn);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001104
Rami Rosen28a28282013-01-11 09:59:20 +00001105 portid = NETLINK_CB(skb).portid; /* netlink portid */
Eric W. Biederman15e47302012-09-07 20:12:54 +00001106 NETLINK_CB(skb).portid = 0; /* from kernel */
Patrick McHardyac6d4392005-08-14 19:29:52 -07001107 NETLINK_CB(skb).dst_group = 0; /* unicast */
Eric W. Biederman15e47302012-09-07 20:12:54 +00001108 netlink_unicast(net->ipv4.fibnl, skb, portid, MSG_DONTWAIT);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001109}
Robert Olsson246955f2005-06-20 13:36:39 -07001110
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00001111static int __net_init nl_fib_lookup_init(struct net *net)
Robert Olsson246955f2005-06-20 13:36:39 -07001112{
Denis V. Lunev6bd48fc2008-01-10 03:28:55 -08001113 struct sock *sk;
Pablo Neira Ayusoa31f2d12012-06-29 06:15:21 +00001114 struct netlink_kernel_cfg cfg = {
1115 .input = nl_fib_input,
1116 };
1117
Pablo Neira Ayuso9f00d972012-09-08 02:53:54 +00001118 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, &cfg);
Ian Morris51456b22015-04-03 09:17:26 +01001119 if (!sk)
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001120 return -EAFNOSUPPORT;
Denis V. Lunev6bd48fc2008-01-10 03:28:55 -08001121 net->ipv4.fibnl = sk;
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001122 return 0;
1123}
1124
1125static void nl_fib_lookup_exit(struct net *net)
1126{
Denis V. Lunevb7c6ba62008-01-28 14:41:19 -08001127 netlink_kernel_release(net->ipv4.fibnl);
Denis V. Lunev775516b2008-01-18 23:55:19 -08001128 net->ipv4.fibnl = NULL;
Robert Olsson246955f2005-06-20 13:36:39 -07001129}
1130
Julian Anastasov4f823de2015-10-30 10:23:33 +02001131static void fib_disable_ip(struct net_device *dev, unsigned long event,
1132 bool force)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133{
Julian Anastasov4f823de2015-10-30 10:23:33 +02001134 if (fib_sync_down_dev(dev, event, force))
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001135 fib_flush(dev_net(dev));
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001136 rt_cache_flush(dev_net(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 arp_ifdown(dev);
1138}
1139
1140static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
1141{
Jianjun Kong6ed25332008-11-03 00:25:16 -08001142 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
Denis V. Lunev76e6ebf2008-07-05 19:00:44 -07001143 struct net_device *dev = ifa->ifa_dev->dev;
David S. Miller436c3b62011-03-24 17:42:21 -07001144 struct net *net = dev_net(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145
1146 switch (event) {
1147 case NETDEV_UP:
1148 fib_add_ifaddr(ifa);
1149#ifdef CONFIG_IP_ROUTE_MULTIPATH
Andy Gospodarek8a3d0312015-06-23 13:45:36 -04001150 fib_sync_up(dev, RTNH_F_DEAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151#endif
David S. Miller436c3b62011-03-24 17:42:21 -07001152 atomic_inc(&net->ipv4.dev_addr_genid);
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001153 rt_cache_flush(dev_net(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154 break;
1155 case NETDEV_DOWN:
Julian Anastasove6abbaa2011-03-19 12:13:49 +00001156 fib_del_ifaddr(ifa, NULL);
David S. Miller436c3b62011-03-24 17:42:21 -07001157 atomic_inc(&net->ipv4.dev_addr_genid);
Ian Morris51456b22015-04-03 09:17:26 +01001158 if (!ifa->ifa_dev->ifa_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 /* Last address was deleted from this interface.
Eric Dumazet6a31d2a2010-10-04 20:00:18 +00001160 * Disable IP.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161 */
Julian Anastasov4f823de2015-10-30 10:23:33 +02001162 fib_disable_ip(dev, event, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163 } else {
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001164 rt_cache_flush(dev_net(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165 }
1166 break;
1167 }
1168 return NOTIFY_DONE;
1169}
1170
1171static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1172{
Jiri Pirko351638e2013-05-28 01:30:21 +00001173 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
David Ahern7f49e7a2015-12-10 10:25:24 -08001174 struct netdev_notifier_changeupper_info *info;
Eric Dumazet0115e8e2012-08-22 17:19:46 +00001175 struct in_device *in_dev;
David S. Miller436c3b62011-03-24 17:42:21 -07001176 struct net *net = dev_net(dev);
Andy Gospodarek8a3d0312015-06-23 13:45:36 -04001177 unsigned int flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178
1179 if (event == NETDEV_UNREGISTER) {
Julian Anastasov4f823de2015-10-30 10:23:33 +02001180 fib_disable_ip(dev, event, true);
David S. Millercaacf052012-07-31 15:06:50 -07001181 rt_flush_dev(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 return NOTIFY_DONE;
1183 }
1184
Eric Dumazet0115e8e2012-08-22 17:19:46 +00001185 in_dev = __in_dev_get_rtnl(dev);
Oliver Hartkoppa0065f22014-01-23 10:19:34 +01001186 if (!in_dev)
1187 return NOTIFY_DONE;
Eric Dumazet0115e8e2012-08-22 17:19:46 +00001188
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189 switch (event) {
1190 case NETDEV_UP:
1191 for_ifa(in_dev) {
1192 fib_add_ifaddr(ifa);
1193 } endfor_ifa(in_dev);
1194#ifdef CONFIG_IP_ROUTE_MULTIPATH
Andy Gospodarek8a3d0312015-06-23 13:45:36 -04001195 fib_sync_up(dev, RTNH_F_DEAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196#endif
David S. Miller436c3b62011-03-24 17:42:21 -07001197 atomic_inc(&net->ipv4.dev_addr_genid);
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001198 rt_cache_flush(net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199 break;
1200 case NETDEV_DOWN:
Julian Anastasov4f823de2015-10-30 10:23:33 +02001201 fib_disable_ip(dev, event, false);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203 case NETDEV_CHANGE:
Andy Gospodarek8a3d0312015-06-23 13:45:36 -04001204 flags = dev_get_flags(dev);
1205 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1206 fib_sync_up(dev, RTNH_F_LINKDOWN);
1207 else
Julian Anastasov4f823de2015-10-30 10:23:33 +02001208 fib_sync_down_dev(dev, event, false);
Andy Gospodarek8a3d0312015-06-23 13:45:36 -04001209 /* fall through */
1210 case NETDEV_CHANGEMTU:
Nicolas Dichtel4ccfe6d2012-09-07 00:45:29 +00001211 rt_cache_flush(net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212 break;
David Ahern7f49e7a2015-12-10 10:25:24 -08001213 case NETDEV_CHANGEUPPER:
1214 info = ptr;
1215 /* flush all routes if dev is linked to or unlinked from
1216 * an L3 master device (e.g., VRF)
1217 */
1218 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
1219 fib_disable_ip(dev, NETDEV_DOWN, true);
1220 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221 }
1222 return NOTIFY_DONE;
1223}
1224
1225static struct notifier_block fib_inetaddr_notifier = {
Jianjun Kong6ed25332008-11-03 00:25:16 -08001226 .notifier_call = fib_inetaddr_event,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227};
1228
1229static struct notifier_block fib_netdev_notifier = {
Jianjun Kong6ed25332008-11-03 00:25:16 -08001230 .notifier_call = fib_netdev_event,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231};
1232
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001233static int __net_init ip_fib_net_init(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234{
Denis V. Lunevdce5cbe2008-01-31 18:44:53 -08001235 int err;
Eric Dumazet10da66f2010-10-13 08:22:03 +00001236 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
Patrick McHardy1af5a8c2006-08-10 23:10:46 -07001237
Eric Dumazet10da66f2010-10-13 08:22:03 +00001238 /* Avoid false sharing : Use at least a full cache line */
1239 size = max_t(size_t, size, L1_CACHE_BYTES);
1240
1241 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
Ian Morris51456b22015-04-03 09:17:26 +01001242 if (!net->ipv4.fib_table_hash)
Denis V. Luneve4aef8a2008-01-10 03:28:24 -08001243 return -ENOMEM;
1244
Denis V. Lunevdce5cbe2008-01-31 18:44:53 -08001245 err = fib4_rules_init(net);
1246 if (err < 0)
1247 goto fail;
1248 return 0;
1249
1250fail:
1251 kfree(net->ipv4.fib_table_hash);
1252 return err;
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001253}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00001255static void ip_fib_net_exit(struct net *net)
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001256{
Ido Schimmele4f66982017-12-20 19:34:19 +02001257 int i;
Thomas Graf63f34442007-03-22 11:55:17 -07001258
Sabrina Dubroca6dede752015-03-10 21:03:54 +01001259 rtnl_lock();
Alexander Duyck6e47d6c2015-03-27 14:14:22 -07001260#ifdef CONFIG_IP_MULTIPLE_TABLES
Alexander Duyck6e47d6c2015-03-27 14:14:22 -07001261 RCU_INIT_POINTER(net->ipv4.fib_main, NULL);
1262 RCU_INIT_POINTER(net->ipv4.fib_default, NULL);
1263#endif
Ido Schimmele4f66982017-12-20 19:34:19 +02001264 /* Destroy the tables in reverse order to guarantee that the
1265 * local table, ID 255, is destroyed before the main table, ID
1266 * 254. This is necessary as the local table may contain
1267 * references to data contained in the main table.
1268 */
1269 for (i = FIB_TABLE_HASHSZ - 1; i >= 0; i--) {
Alexander Duycka7e53532015-03-04 15:02:44 -08001270 struct hlist_head *head = &net->ipv4.fib_table_hash[i];
Sasha Levinb67bfe02013-02-27 17:06:00 -08001271 struct hlist_node *tmp;
Alexander Duycka7e53532015-03-04 15:02:44 -08001272 struct fib_table *tb;
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001273
Alexander Duycka7e53532015-03-04 15:02:44 -08001274 hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) {
Alexander Duycka7e53532015-03-04 15:02:44 -08001275 hlist_del(&tb->tb_hlist);
Jiri Pirkob90eb752016-09-26 12:52:29 +02001276 fib_table_flush(net, tb);
Pavel Emelyanov4aa2c462010-10-28 02:00:43 +00001277 fib_free_table(tb);
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001278 }
1279 }
Alexander Duyckad88d052015-03-27 14:14:16 -07001280
1281#ifdef CONFIG_IP_MULTIPLE_TABLES
1282 fib4_rules_exit(net);
1283#endif
Eric Dumazete2666f82011-03-30 16:57:46 -07001284 rtnl_unlock();
Denis V. Luneve4aef8a2008-01-10 03:28:24 -08001285 kfree(net->ipv4.fib_table_hash);
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001286}
1287
1288static int __net_init fib_net_init(struct net *net)
1289{
1290 int error;
1291
David S. Millerf4530fa2012-07-05 22:13:13 -07001292#ifdef CONFIG_IP_ROUTE_CLASSID
1293 net->ipv4.fib_num_tclassid_users = 0;
1294#endif
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001295 error = ip_fib_net_init(net);
1296 if (error < 0)
1297 goto out;
1298 error = nl_fib_lookup_init(net);
1299 if (error < 0)
1300 goto out_nlfl;
1301 error = fib_proc_init(net);
1302 if (error < 0)
1303 goto out_proc;
1304out:
1305 return error;
1306
1307out_proc:
1308 nl_fib_lookup_exit(net);
1309out_nlfl:
1310 ip_fib_net_exit(net);
1311 goto out;
1312}
1313
1314static void __net_exit fib_net_exit(struct net *net)
1315{
1316 fib_proc_exit(net);
1317 nl_fib_lookup_exit(net);
1318 ip_fib_net_exit(net);
1319}
1320
1321static struct pernet_operations fib_net_ops = {
1322 .init = fib_net_init,
1323 .exit = fib_net_exit,
1324};
1325
1326void __init ip_fib_init(void)
1327{
Mahesh Bandewar0c47f112017-07-19 15:41:33 -07001328 fib_trie_init();
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001329
1330 register_pernet_subsys(&fib_net_ops);
Mahesh Bandewar0c47f112017-07-19 15:41:33 -07001331
Denis V. Lunev7b1a74f2008-01-10 03:22:17 -08001332 register_netdevice_notifier(&fib_netdev_notifier);
1333 register_inetaddr_notifier(&fib_inetaddr_notifier);
Stephen Hemminger7f9b8052008-01-14 23:14:20 -08001334
Mahesh Bandewar0c47f112017-07-19 15:41:33 -07001335 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, NULL);
1336 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, NULL);
1337 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338}