]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
3 | * operating system. INET is implemented using the BSD Socket | |
4 | * interface as the means of communication with the user level. | |
5 | * | |
6 | * IPv4 Forwarding Information Base: FIB frontend. | |
7 | * | |
1da177e4 LT |
8 | * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> |
9 | * | |
10 | * This program is free software; you can redistribute it and/or | |
11 | * modify it under the terms of the GNU General Public License | |
12 | * as published by the Free Software Foundation; either version | |
13 | * 2 of the License, or (at your option) any later version. | |
14 | */ | |
15 | ||
1da177e4 | 16 | #include <linux/module.h> |
7c0f6ba6 | 17 | #include <linux/uaccess.h> |
1da177e4 | 18 | #include <linux/bitops.h> |
4fc268d2 | 19 | #include <linux/capability.h> |
1da177e4 LT |
20 | #include <linux/types.h> |
21 | #include <linux/kernel.h> | |
1da177e4 LT |
22 | #include <linux/mm.h> |
23 | #include <linux/string.h> | |
24 | #include <linux/socket.h> | |
25 | #include <linux/sockios.h> | |
26 | #include <linux/errno.h> | |
27 | #include <linux/in.h> | |
28 | #include <linux/inet.h> | |
14c85021 | 29 | #include <linux/inetdevice.h> |
1da177e4 | 30 | #include <linux/netdevice.h> |
1823730f | 31 | #include <linux/if_addr.h> |
1da177e4 LT |
32 | #include <linux/if_arp.h> |
33 | #include <linux/skbuff.h> | |
7a9bc9b8 | 34 | #include <linux/cache.h> |
1da177e4 | 35 | #include <linux/init.h> |
1af5a8c4 | 36 | #include <linux/list.h> |
5a0e3ad6 | 37 | #include <linux/slab.h> |
1da177e4 LT |
38 | |
39 | #include <net/ip.h> | |
40 | #include <net/protocol.h> | |
41 | #include <net/route.h> | |
42 | #include <net/tcp.h> | |
43 | #include <net/sock.h> | |
1da177e4 LT |
44 | #include <net/arp.h> |
45 | #include <net/ip_fib.h> | |
63f3444f | 46 | #include <net/rtnetlink.h> |
990078af | 47 | #include <net/xfrm.h> |
385add90 | 48 | #include <net/l3mdev.h> |
9ed59592 | 49 | #include <net/lwtunnel.h> |
f6d3c192 | 50 | #include <trace/events/fib.h> |
1da177e4 | 51 | |
1da177e4 LT |
52 | #ifndef CONFIG_IP_MULTIPLE_TABLES |
53 | ||
7b1a74fd | 54 | static int __net_init fib4_rules_init(struct net *net) |
c3e9a353 | 55 | { |
93456b6d DL |
56 | struct fib_table *local_table, *main_table; |
57 | ||
0ddcf43d | 58 | main_table = fib_trie_table(RT_TABLE_MAIN, NULL); |
51456b29 | 59 | if (!main_table) |
61f0d861 | 60 | return -ENOMEM; |
dbb50165 | 61 | |
0ddcf43d | 62 | local_table = fib_trie_table(RT_TABLE_LOCAL, main_table); |
51456b29 | 63 | if (!local_table) |
61f0d861 | 64 | goto fail; |
0ddcf43d | 65 | |
93456b6d | 66 | hlist_add_head_rcu(&local_table->tb_hlist, |
e4aef8ae | 67 | &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]); |
93456b6d | 68 | hlist_add_head_rcu(&main_table->tb_hlist, |
e4aef8ae | 69 | &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]); |
dbb50165 DL |
70 | return 0; |
71 | ||
72 | fail: | |
61f0d861 | 73 | fib_free_table(main_table); |
dbb50165 | 74 | return -ENOMEM; |
c3e9a353 | 75 | } |
032a4802 PA |
76 | |
77 | static bool fib4_has_custom_rules(struct net *net) | |
78 | { | |
79 | return false; | |
80 | } | |
1af5a8c4 | 81 | #else |
1da177e4 | 82 | |
8ad4942c | 83 | struct fib_table *fib_new_table(struct net *net, u32 id) |
1da177e4 | 84 | { |
0ddcf43d | 85 | struct fib_table *tb, *alias = NULL; |
1af5a8c4 | 86 | unsigned int h; |
1da177e4 | 87 | |
1af5a8c4 PM |
88 | if (id == 0) |
89 | id = RT_TABLE_MAIN; | |
8ad4942c | 90 | tb = fib_get_table(net, id); |
1af5a8c4 PM |
91 | if (tb) |
92 | return tb; | |
7f9b8052 | 93 | |
5350d54f | 94 | if (id == RT_TABLE_LOCAL && !net->ipv4.fib_has_custom_rules) |
0ddcf43d AD |
95 | alias = fib_new_table(net, RT_TABLE_MAIN); |
96 | ||
97 | tb = fib_trie_table(id, alias); | |
1da177e4 LT |
98 | if (!tb) |
99 | return NULL; | |
f4530fa5 DM |
100 | |
101 | switch (id) { | |
f4530fa5 | 102 | case RT_TABLE_MAIN: |
a7e53531 | 103 | rcu_assign_pointer(net->ipv4.fib_main, tb); |
f4530fa5 | 104 | break; |
f4530fa5 | 105 | case RT_TABLE_DEFAULT: |
a7e53531 | 106 | rcu_assign_pointer(net->ipv4.fib_default, tb); |
f4530fa5 | 107 | break; |
f4530fa5 DM |
108 | default: |
109 | break; | |
110 | } | |
111 | ||
1af5a8c4 | 112 | h = id & (FIB_TABLE_HASHSZ - 1); |
e4aef8ae | 113 | hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]); |
1da177e4 LT |
114 | return tb; |
115 | } | |
b3b4663c | 116 | EXPORT_SYMBOL_GPL(fib_new_table); |
1da177e4 | 117 | |
345e9b54 | 118 | /* caller must hold either rtnl or rcu read lock */ |
8ad4942c | 119 | struct fib_table *fib_get_table(struct net *net, u32 id) |
1af5a8c4 PM |
120 | { |
121 | struct fib_table *tb; | |
e4aef8ae | 122 | struct hlist_head *head; |
1af5a8c4 | 123 | unsigned int h; |
1da177e4 | 124 | |
1af5a8c4 PM |
125 | if (id == 0) |
126 | id = RT_TABLE_MAIN; | |
127 | h = id & (FIB_TABLE_HASHSZ - 1); | |
e4aef8ae | 128 | |
e4aef8ae | 129 | head = &net->ipv4.fib_table_hash[h]; |
b67bfe0d | 130 | hlist_for_each_entry_rcu(tb, head, tb_hlist) { |
345e9b54 | 131 | if (tb->tb_id == id) |
1af5a8c4 | 132 | return tb; |
1af5a8c4 | 133 | } |
1af5a8c4 PM |
134 | return NULL; |
135 | } | |
032a4802 PA |
136 | |
137 | static bool fib4_has_custom_rules(struct net *net) | |
138 | { | |
139 | return net->ipv4.fib_has_custom_rules; | |
140 | } | |
1da177e4 LT |
141 | #endif /* CONFIG_IP_MULTIPLE_TABLES */ |
142 | ||
0ddcf43d AD |
143 | static void fib_replace_table(struct net *net, struct fib_table *old, |
144 | struct fib_table *new) | |
145 | { | |
146 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
147 | switch (new->tb_id) { | |
0ddcf43d AD |
148 | case RT_TABLE_MAIN: |
149 | rcu_assign_pointer(net->ipv4.fib_main, new); | |
150 | break; | |
151 | case RT_TABLE_DEFAULT: | |
152 | rcu_assign_pointer(net->ipv4.fib_default, new); | |
153 | break; | |
154 | default: | |
155 | break; | |
156 | } | |
157 | ||
158 | #endif | |
159 | /* replace the old table in the hlist */ | |
160 | hlist_replace_rcu(&old->tb_hlist, &new->tb_hlist); | |
161 | } | |
162 | ||
163 | int fib_unmerge(struct net *net) | |
164 | { | |
3b709334 | 165 | struct fib_table *old, *new, *main_table; |
0ddcf43d | 166 | |
3c9e9f73 | 167 | /* attempt to fetch local table if it has been allocated */ |
0ddcf43d | 168 | old = fib_get_table(net, RT_TABLE_LOCAL); |
3c9e9f73 AD |
169 | if (!old) |
170 | return 0; | |
0ddcf43d | 171 | |
3c9e9f73 | 172 | new = fib_trie_unmerge(old); |
0ddcf43d AD |
173 | if (!new) |
174 | return -ENOMEM; | |
175 | ||
3b709334 AD |
176 | /* table is already unmerged */ |
177 | if (new == old) | |
178 | return 0; | |
179 | ||
0ddcf43d | 180 | /* replace merged table with clean table */ |
3b709334 AD |
181 | fib_replace_table(net, old, new); |
182 | fib_free_table(old); | |
183 | ||
184 | /* attempt to fetch main table if it has been allocated */ | |
185 | main_table = fib_get_table(net, RT_TABLE_MAIN); | |
186 | if (!main_table) | |
187 | return 0; | |
188 | ||
189 | /* flush local entries from main table */ | |
190 | fib_table_flush_external(main_table); | |
0ddcf43d AD |
191 | |
192 | return 0; | |
193 | } | |
194 | ||
e4aef8ae | 195 | static void fib_flush(struct net *net) |
1da177e4 LT |
196 | { |
197 | int flushed = 0; | |
1af5a8c4 | 198 | unsigned int h; |
1da177e4 | 199 | |
1af5a8c4 | 200 | for (h = 0; h < FIB_TABLE_HASHSZ; h++) { |
a7e53531 AD |
201 | struct hlist_head *head = &net->ipv4.fib_table_hash[h]; |
202 | struct hlist_node *tmp; | |
203 | struct fib_table *tb; | |
204 | ||
205 | hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) | |
b90eb754 | 206 | flushed += fib_table_flush(net, tb); |
1da177e4 | 207 | } |
1da177e4 LT |
208 | |
209 | if (flushed) | |
4ccfe6d4 | 210 | rt_cache_flush(net); |
1da177e4 LT |
211 | } |
212 | ||
05538116 LAT |
213 | /* |
214 | * Find address type as if only "dev" was present in the system. If | |
215 | * on_dev is NULL then all interfaces are taken into consideration. | |
216 | */ | |
95c96174 ED |
217 | static inline unsigned int __inet_dev_addr_type(struct net *net, |
218 | const struct net_device *dev, | |
9b8ff518 | 219 | __be32 addr, u32 tb_id) |
1da177e4 | 220 | { |
9ade2286 | 221 | struct flowi4 fl4 = { .daddr = addr }; |
1da177e4 | 222 | struct fib_result res; |
95c96174 | 223 | unsigned int ret = RTN_BROADCAST; |
15be405e | 224 | struct fib_table *table; |
1da177e4 | 225 | |
1e637c74 | 226 | if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr)) |
1da177e4 | 227 | return RTN_BROADCAST; |
f97c1e0c | 228 | if (ipv4_is_multicast(addr)) |
1da177e4 LT |
229 | return RTN_MULTICAST; |
230 | ||
345e9b54 AD |
231 | rcu_read_lock(); |
232 | ||
15be405e DA |
233 | table = fib_get_table(net, tb_id); |
234 | if (table) { | |
1da177e4 | 235 | ret = RTN_UNICAST; |
15be405e | 236 | if (!fib_table_lookup(table, &fl4, &res, FIB_LOOKUP_NOREF)) { |
05538116 LAT |
237 | if (!dev || dev == res.fi->fib_dev) |
238 | ret = res.type; | |
1da177e4 LT |
239 | } |
240 | } | |
345e9b54 AD |
241 | |
242 | rcu_read_unlock(); | |
1da177e4 LT |
243 | return ret; |
244 | } | |
245 | ||
9b8ff518 | 246 | unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id) |
15be405e DA |
247 | { |
248 | return __inet_dev_addr_type(net, NULL, addr, tb_id); | |
249 | } | |
250 | EXPORT_SYMBOL(inet_addr_type_table); | |
251 | ||
6b175b26 | 252 | unsigned int inet_addr_type(struct net *net, __be32 addr) |
05538116 | 253 | { |
15be405e | 254 | return __inet_dev_addr_type(net, NULL, addr, RT_TABLE_LOCAL); |
05538116 | 255 | } |
4bc2f18b | 256 | EXPORT_SYMBOL(inet_addr_type); |
05538116 | 257 | |
6b175b26 EB |
258 | unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev, |
259 | __be32 addr) | |
05538116 | 260 | { |
3236b004 | 261 | u32 rt_table = l3mdev_fib_table(dev) ? : RT_TABLE_LOCAL; |
15be405e DA |
262 | |
263 | return __inet_dev_addr_type(net, dev, addr, rt_table); | |
05538116 | 264 | } |
4bc2f18b | 265 | EXPORT_SYMBOL(inet_dev_addr_type); |
05538116 | 266 | |
30bbaa19 DA |
267 | /* inet_addr_type with dev == NULL but using the table from a dev |
268 | * if one is associated | |
269 | */ | |
270 | unsigned int inet_addr_type_dev_table(struct net *net, | |
271 | const struct net_device *dev, | |
272 | __be32 addr) | |
273 | { | |
3236b004 | 274 | u32 rt_table = l3mdev_fib_table(dev) ? : RT_TABLE_LOCAL; |
30bbaa19 DA |
275 | |
276 | return __inet_dev_addr_type(net, NULL, addr, rt_table); | |
277 | } | |
278 | EXPORT_SYMBOL(inet_addr_type_dev_table); | |
279 | ||
35ebf65e DM |
280 | __be32 fib_compute_spec_dst(struct sk_buff *skb) |
281 | { | |
282 | struct net_device *dev = skb->dev; | |
283 | struct in_device *in_dev; | |
284 | struct fib_result res; | |
a207a4b2 | 285 | struct rtable *rt; |
35ebf65e | 286 | struct net *net; |
a207a4b2 | 287 | int scope; |
35ebf65e | 288 | |
a207a4b2 | 289 | rt = skb_rtable(skb); |
0cc535a2 JA |
290 | if ((rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | RTCF_LOCAL)) == |
291 | RTCF_LOCAL) | |
35ebf65e DM |
292 | return ip_hdr(skb)->daddr; |
293 | ||
294 | in_dev = __in_dev_get_rcu(dev); | |
295 | BUG_ON(!in_dev); | |
35ebf65e DM |
296 | |
297 | net = dev_net(dev); | |
a207a4b2 DM |
298 | |
299 | scope = RT_SCOPE_UNIVERSE; | |
300 | if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) { | |
4cfc86f3 LR |
301 | struct flowi4 fl4 = { |
302 | .flowi4_iif = LOOPBACK_IFINDEX, | |
303 | .daddr = ip_hdr(skb)->saddr, | |
304 | .flowi4_tos = RT_TOS(ip_hdr(skb)->tos), | |
305 | .flowi4_scope = scope, | |
306 | .flowi4_mark = IN_DEV_SRC_VMARK(in_dev) ? skb->mark : 0, | |
307 | }; | |
0eeb075f | 308 | if (!fib_lookup(net, &fl4, &res, 0)) |
a207a4b2 DM |
309 | return FIB_RES_PREFSRC(net, res); |
310 | } else { | |
311 | scope = RT_SCOPE_LINK; | |
312 | } | |
313 | ||
314 | return inet_select_addr(dev, ip_hdr(skb)->saddr, scope); | |
35ebf65e DM |
315 | } |
316 | ||
1da177e4 | 317 | /* Given (packet source, input interface) and optional (dst, oif, tos): |
6a31d2a9 ED |
318 | * - (main) check, that source is valid i.e. not broadcast or our local |
319 | * address. | |
320 | * - figure out what "logical" interface this packet arrived | |
321 | * and calculate "specific destination" address. | |
322 | * - check, that packet arrived from expected physical interface. | |
ebc0ffae | 323 | * called with rcu_read_lock() |
1da177e4 | 324 | */ |
7a9bc9b8 DM |
325 | static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst, |
326 | u8 tos, int oif, struct net_device *dev, | |
327 | int rpf, struct in_device *idev, u32 *itag) | |
1da177e4 | 328 | { |
1dced6a8 | 329 | int ret, no_addr; |
1da177e4 | 330 | struct fib_result res; |
9e56e380 | 331 | struct flowi4 fl4; |
8bcfd092 | 332 | struct net *net = dev_net(dev); |
9e56e380 | 333 | bool dev_match; |
1da177e4 | 334 | |
9ade2286 | 335 | fl4.flowi4_oif = 0; |
385add90 | 336 | fl4.flowi4_iif = l3mdev_master_ifindex_rcu(dev); |
cd2fbe1b DA |
337 | if (!fl4.flowi4_iif) |
338 | fl4.flowi4_iif = oif ? : LOOPBACK_IFINDEX; | |
9ade2286 DM |
339 | fl4.daddr = src; |
340 | fl4.saddr = dst; | |
341 | fl4.flowi4_tos = tos; | |
342 | fl4.flowi4_scope = RT_SCOPE_UNIVERSE; | |
1b7179d3 | 343 | fl4.flowi4_tun_key.tun_id = 0; |
b84f7878 | 344 | fl4.flowi4_flags = 0; |
8bcfd092 | 345 | fl4.flowi4_uid = sock_net_uid(net, NULL); |
cc7e17ea | 346 | |
9e56e380 | 347 | no_addr = idev->ifa_list == NULL; |
990078af | 348 | |
9e56e380 | 349 | fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0; |
1da177e4 | 350 | |
f6d3c192 DA |
351 | trace_fib_validate_source(dev, &fl4); |
352 | ||
0eeb075f | 353 | if (fib_lookup(net, &fl4, &res, 0)) |
1da177e4 | 354 | goto last_resort; |
1dced6a8 SB |
355 | if (res.type != RTN_UNICAST && |
356 | (res.type != RTN_LOCAL || !IN_DEV_ACCEPT_LOCAL(idev))) | |
357 | goto e_inval; | |
1da177e4 | 358 | fib_combine_itag(itag, &res); |
6f86b325 DM |
359 | dev_match = false; |
360 | ||
1da177e4 | 361 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
6f86b325 DM |
362 | for (ret = 0; ret < res.fi->fib_nhs; ret++) { |
363 | struct fib_nh *nh = &res.fi->fib_nh[ret]; | |
364 | ||
365 | if (nh->nh_dev == dev) { | |
366 | dev_match = true; | |
367 | break; | |
385add90 | 368 | } else if (l3mdev_master_ifindex_rcu(nh->nh_dev) == dev->ifindex) { |
cd2fbe1b DA |
369 | dev_match = true; |
370 | break; | |
6f86b325 DM |
371 | } |
372 | } | |
1da177e4 LT |
373 | #else |
374 | if (FIB_RES_DEV(res) == dev) | |
6f86b325 | 375 | dev_match = true; |
1da177e4 | 376 | #endif |
6f86b325 | 377 | if (dev_match) { |
1da177e4 | 378 | ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST; |
1da177e4 LT |
379 | return ret; |
380 | } | |
1da177e4 LT |
381 | if (no_addr) |
382 | goto last_resort; | |
c1cf8422 | 383 | if (rpf == 1) |
b5f7e755 | 384 | goto e_rpf; |
9ade2286 | 385 | fl4.flowi4_oif = dev->ifindex; |
1da177e4 LT |
386 | |
387 | ret = 0; | |
0eeb075f | 388 | if (fib_lookup(net, &fl4, &res, FIB_LOOKUP_IGNORE_LINKSTATE) == 0) { |
41347dcd | 389 | if (res.type == RTN_UNICAST) |
1da177e4 | 390 | ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST; |
1da177e4 LT |
391 | } |
392 | return ret; | |
393 | ||
394 | last_resort: | |
395 | if (rpf) | |
b5f7e755 | 396 | goto e_rpf; |
1da177e4 LT |
397 | *itag = 0; |
398 | return 0; | |
399 | ||
1da177e4 LT |
400 | e_inval: |
401 | return -EINVAL; | |
b5f7e755 ED |
402 | e_rpf: |
403 | return -EXDEV; | |
1da177e4 LT |
404 | } |
405 | ||
7a9bc9b8 DM |
406 | /* Ignore rp_filter for packets protected by IPsec. */ |
407 | int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst, | |
408 | u8 tos, int oif, struct net_device *dev, | |
409 | struct in_device *idev, u32 *itag) | |
410 | { | |
411 | int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev); | |
6e617de8 | 412 | struct net *net = dev_net(dev); |
7a9bc9b8 | 413 | |
6e617de8 | 414 | if (!r && !fib_num_tclassid_users(net) && |
e81da0e1 | 415 | (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) { |
6e617de8 PA |
416 | if (IN_DEV_ACCEPT_LOCAL(idev)) |
417 | goto ok; | |
032a4802 PA |
418 | /* with custom local routes in place, checking local addresses |
419 | * only will be too optimistic, with custom rules, checking | |
420 | * local addresses only can be too strict, e.g. due to vrf | |
6e617de8 | 421 | */ |
032a4802 PA |
422 | if (net->ipv4.fib_has_custom_local_routes || |
423 | fib4_has_custom_rules(net)) | |
6e617de8 PA |
424 | goto full_check; |
425 | if (inet_lookup_ifaddr_rcu(net, src)) | |
426 | return -EINVAL; | |
427 | ||
428 | ok: | |
7a9bc9b8 DM |
429 | *itag = 0; |
430 | return 0; | |
431 | } | |
6e617de8 PA |
432 | |
433 | full_check: | |
7a9bc9b8 DM |
434 | return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag); |
435 | } | |
436 | ||
81f7bf6c | 437 | static inline __be32 sk_extract_addr(struct sockaddr *addr) |
4e902c57 TG |
438 | { |
439 | return ((struct sockaddr_in *) addr)->sin_addr.s_addr; | |
440 | } | |
441 | ||
442 | static int put_rtax(struct nlattr *mx, int len, int type, u32 value) | |
443 | { | |
444 | struct nlattr *nla; | |
445 | ||
446 | nla = (struct nlattr *) ((char *) mx + len); | |
447 | nla->nla_type = type; | |
448 | nla->nla_len = nla_attr_size(4); | |
449 | *(u32 *) nla_data(nla) = value; | |
450 | ||
451 | return len + nla_total_size(4); | |
452 | } | |
453 | ||
4b5d47d4 | 454 | static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt, |
4e902c57 TG |
455 | struct fib_config *cfg) |
456 | { | |
6d85c10a | 457 | __be32 addr; |
4e902c57 TG |
458 | int plen; |
459 | ||
460 | memset(cfg, 0, sizeof(*cfg)); | |
4b5d47d4 | 461 | cfg->fc_nlinfo.nl_net = net; |
4e902c57 TG |
462 | |
463 | if (rt->rt_dst.sa_family != AF_INET) | |
464 | return -EAFNOSUPPORT; | |
465 | ||
466 | /* | |
467 | * Check mask for validity: | |
468 | * a) it must be contiguous. | |
469 | * b) destination must have all host bits clear. | |
470 | * c) if application forgot to set correct family (AF_INET), | |
471 | * reject request unless it is absolutely clear i.e. | |
472 | * both family and mask are zero. | |
473 | */ | |
474 | plen = 32; | |
475 | addr = sk_extract_addr(&rt->rt_dst); | |
476 | if (!(rt->rt_flags & RTF_HOST)) { | |
81f7bf6c | 477 | __be32 mask = sk_extract_addr(&rt->rt_genmask); |
4e902c57 TG |
478 | |
479 | if (rt->rt_genmask.sa_family != AF_INET) { | |
480 | if (mask || rt->rt_genmask.sa_family) | |
481 | return -EAFNOSUPPORT; | |
482 | } | |
483 | ||
484 | if (bad_mask(mask, addr)) | |
485 | return -EINVAL; | |
486 | ||
487 | plen = inet_mask_len(mask); | |
488 | } | |
489 | ||
490 | cfg->fc_dst_len = plen; | |
491 | cfg->fc_dst = addr; | |
492 | ||
493 | if (cmd != SIOCDELRT) { | |
494 | cfg->fc_nlflags = NLM_F_CREATE; | |
495 | cfg->fc_protocol = RTPROT_BOOT; | |
496 | } | |
497 | ||
498 | if (rt->rt_metric) | |
499 | cfg->fc_priority = rt->rt_metric - 1; | |
500 | ||
501 | if (rt->rt_flags & RTF_REJECT) { | |
502 | cfg->fc_scope = RT_SCOPE_HOST; | |
503 | cfg->fc_type = RTN_UNREACHABLE; | |
504 | return 0; | |
505 | } | |
506 | ||
507 | cfg->fc_scope = RT_SCOPE_NOWHERE; | |
508 | cfg->fc_type = RTN_UNICAST; | |
509 | ||
510 | if (rt->rt_dev) { | |
511 | char *colon; | |
512 | struct net_device *dev; | |
513 | char devname[IFNAMSIZ]; | |
514 | ||
515 | if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1)) | |
516 | return -EFAULT; | |
517 | ||
518 | devname[IFNAMSIZ-1] = 0; | |
519 | colon = strchr(devname, ':'); | |
520 | if (colon) | |
521 | *colon = 0; | |
4b5d47d4 | 522 | dev = __dev_get_by_name(net, devname); |
4e902c57 TG |
523 | if (!dev) |
524 | return -ENODEV; | |
525 | cfg->fc_oif = dev->ifindex; | |
5a56a0b3 | 526 | cfg->fc_table = l3mdev_fib_table(dev); |
4e902c57 TG |
527 | if (colon) { |
528 | struct in_ifaddr *ifa; | |
529 | struct in_device *in_dev = __in_dev_get_rtnl(dev); | |
530 | if (!in_dev) | |
531 | return -ENODEV; | |
532 | *colon = ':'; | |
533 | for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) | |
534 | if (strcmp(ifa->ifa_label, devname) == 0) | |
535 | break; | |
51456b29 | 536 | if (!ifa) |
4e902c57 TG |
537 | return -ENODEV; |
538 | cfg->fc_prefsrc = ifa->ifa_local; | |
539 | } | |
540 | } | |
541 | ||
542 | addr = sk_extract_addr(&rt->rt_gateway); | |
543 | if (rt->rt_gateway.sa_family == AF_INET && addr) { | |
30bbaa19 DA |
544 | unsigned int addr_type; |
545 | ||
4e902c57 | 546 | cfg->fc_gw = addr; |
30bbaa19 | 547 | addr_type = inet_addr_type_table(net, addr, cfg->fc_table); |
4e902c57 | 548 | if (rt->rt_flags & RTF_GATEWAY && |
30bbaa19 | 549 | addr_type == RTN_UNICAST) |
4e902c57 TG |
550 | cfg->fc_scope = RT_SCOPE_UNIVERSE; |
551 | } | |
552 | ||
553 | if (cmd == SIOCDELRT) | |
554 | return 0; | |
555 | ||
556 | if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw) | |
557 | return -EINVAL; | |
558 | ||
559 | if (cfg->fc_scope == RT_SCOPE_NOWHERE) | |
560 | cfg->fc_scope = RT_SCOPE_LINK; | |
561 | ||
562 | if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) { | |
563 | struct nlattr *mx; | |
564 | int len = 0; | |
565 | ||
566 | mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL); | |
51456b29 | 567 | if (!mx) |
4e902c57 TG |
568 | return -ENOMEM; |
569 | ||
570 | if (rt->rt_flags & RTF_MTU) | |
571 | len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40); | |
572 | ||
573 | if (rt->rt_flags & RTF_WINDOW) | |
574 | len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window); | |
575 | ||
576 | if (rt->rt_flags & RTF_IRTT) | |
577 | len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3); | |
578 | ||
579 | cfg->fc_mx = mx; | |
580 | cfg->fc_mx_len = len; | |
581 | } | |
582 | ||
583 | return 0; | |
584 | } | |
585 | ||
1da177e4 | 586 | /* |
6a31d2a9 ED |
587 | * Handle IP routing ioctl calls. |
588 | * These are used to manipulate the routing tables | |
1da177e4 | 589 | */ |
1bad118a | 590 | int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg) |
1da177e4 | 591 | { |
4e902c57 TG |
592 | struct fib_config cfg; |
593 | struct rtentry rt; | |
1da177e4 | 594 | int err; |
1da177e4 LT |
595 | |
596 | switch (cmd) { | |
597 | case SIOCADDRT: /* Add a route */ | |
598 | case SIOCDELRT: /* Delete a route */ | |
52e804c6 | 599 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) |
1da177e4 | 600 | return -EPERM; |
4e902c57 TG |
601 | |
602 | if (copy_from_user(&rt, arg, sizeof(rt))) | |
1da177e4 | 603 | return -EFAULT; |
4e902c57 | 604 | |
1da177e4 | 605 | rtnl_lock(); |
1bad118a | 606 | err = rtentry_to_fib_config(net, cmd, &rt, &cfg); |
1da177e4 | 607 | if (err == 0) { |
4e902c57 TG |
608 | struct fib_table *tb; |
609 | ||
1da177e4 | 610 | if (cmd == SIOCDELRT) { |
1bad118a | 611 | tb = fib_get_table(net, cfg.fc_table); |
1da177e4 | 612 | if (tb) |
78055998 DA |
613 | err = fib_table_delete(net, tb, &cfg, |
614 | NULL); | |
4e902c57 TG |
615 | else |
616 | err = -ESRCH; | |
1da177e4 | 617 | } else { |
1bad118a | 618 | tb = fib_new_table(net, cfg.fc_table); |
1da177e4 | 619 | if (tb) |
6d8422a1 DA |
620 | err = fib_table_insert(net, tb, |
621 | &cfg, NULL); | |
4e902c57 TG |
622 | else |
623 | err = -ENOBUFS; | |
1da177e4 | 624 | } |
4e902c57 TG |
625 | |
626 | /* allocated by rtentry_to_fib_config() */ | |
627 | kfree(cfg.fc_mx); | |
1da177e4 LT |
628 | } |
629 | rtnl_unlock(); | |
630 | return err; | |
631 | } | |
632 | return -EINVAL; | |
633 | } | |
634 | ||
6a31d2a9 | 635 | const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = { |
4e902c57 TG |
636 | [RTA_DST] = { .type = NLA_U32 }, |
637 | [RTA_SRC] = { .type = NLA_U32 }, | |
638 | [RTA_IIF] = { .type = NLA_U32 }, | |
639 | [RTA_OIF] = { .type = NLA_U32 }, | |
640 | [RTA_GATEWAY] = { .type = NLA_U32 }, | |
641 | [RTA_PRIORITY] = { .type = NLA_U32 }, | |
642 | [RTA_PREFSRC] = { .type = NLA_U32 }, | |
643 | [RTA_METRICS] = { .type = NLA_NESTED }, | |
5176f91e | 644 | [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) }, |
4e902c57 | 645 | [RTA_FLOW] = { .type = NLA_U32 }, |
571e7226 RP |
646 | [RTA_ENCAP_TYPE] = { .type = NLA_U16 }, |
647 | [RTA_ENCAP] = { .type = NLA_NESTED }, | |
622ec2c9 | 648 | [RTA_UID] = { .type = NLA_U32 }, |
3b45a410 | 649 | [RTA_MARK] = { .type = NLA_U32 }, |
4e902c57 TG |
650 | }; |
651 | ||
4b5d47d4 | 652 | static int rtm_to_fib_config(struct net *net, struct sk_buff *skb, |
6d8422a1 DA |
653 | struct nlmsghdr *nlh, struct fib_config *cfg, |
654 | struct netlink_ext_ack *extack) | |
1da177e4 | 655 | { |
4e902c57 TG |
656 | struct nlattr *attr; |
657 | int err, remaining; | |
658 | struct rtmsg *rtm; | |
659 | ||
fceb6435 | 660 | err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy, |
6d8422a1 | 661 | extack); |
4e902c57 TG |
662 | if (err < 0) |
663 | goto errout; | |
664 | ||
665 | memset(cfg, 0, sizeof(*cfg)); | |
666 | ||
667 | rtm = nlmsg_data(nlh); | |
4e902c57 | 668 | cfg->fc_dst_len = rtm->rtm_dst_len; |
4e902c57 TG |
669 | cfg->fc_tos = rtm->rtm_tos; |
670 | cfg->fc_table = rtm->rtm_table; | |
671 | cfg->fc_protocol = rtm->rtm_protocol; | |
672 | cfg->fc_scope = rtm->rtm_scope; | |
673 | cfg->fc_type = rtm->rtm_type; | |
674 | cfg->fc_flags = rtm->rtm_flags; | |
675 | cfg->fc_nlflags = nlh->nlmsg_flags; | |
676 | ||
15e47304 | 677 | cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid; |
4e902c57 | 678 | cfg->fc_nlinfo.nlh = nlh; |
4b5d47d4 | 679 | cfg->fc_nlinfo.nl_net = net; |
4e902c57 | 680 | |
a0ee18b9 | 681 | if (cfg->fc_type > RTN_MAX) { |
c3ab2b4e | 682 | NL_SET_ERR_MSG(extack, "Invalid route type"); |
a0ee18b9 TG |
683 | err = -EINVAL; |
684 | goto errout; | |
685 | } | |
686 | ||
4e902c57 | 687 | nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) { |
8f4c1f9b | 688 | switch (nla_type(attr)) { |
4e902c57 | 689 | case RTA_DST: |
17fb2c64 | 690 | cfg->fc_dst = nla_get_be32(attr); |
4e902c57 | 691 | break; |
4e902c57 TG |
692 | case RTA_OIF: |
693 | cfg->fc_oif = nla_get_u32(attr); | |
694 | break; | |
695 | case RTA_GATEWAY: | |
17fb2c64 | 696 | cfg->fc_gw = nla_get_be32(attr); |
4e902c57 TG |
697 | break; |
698 | case RTA_PRIORITY: | |
699 | cfg->fc_priority = nla_get_u32(attr); | |
700 | break; | |
701 | case RTA_PREFSRC: | |
17fb2c64 | 702 | cfg->fc_prefsrc = nla_get_be32(attr); |
4e902c57 TG |
703 | break; |
704 | case RTA_METRICS: | |
705 | cfg->fc_mx = nla_data(attr); | |
706 | cfg->fc_mx_len = nla_len(attr); | |
707 | break; | |
708 | case RTA_MULTIPATH: | |
9ed59592 | 709 | err = lwtunnel_valid_encap_type_attr(nla_data(attr), |
c255bd68 DA |
710 | nla_len(attr), |
711 | extack); | |
9ed59592 DA |
712 | if (err < 0) |
713 | goto errout; | |
4e902c57 TG |
714 | cfg->fc_mp = nla_data(attr); |
715 | cfg->fc_mp_len = nla_len(attr); | |
716 | break; | |
717 | case RTA_FLOW: | |
718 | cfg->fc_flow = nla_get_u32(attr); | |
719 | break; | |
4e902c57 TG |
720 | case RTA_TABLE: |
721 | cfg->fc_table = nla_get_u32(attr); | |
722 | break; | |
571e7226 RP |
723 | case RTA_ENCAP: |
724 | cfg->fc_encap = attr; | |
725 | break; | |
726 | case RTA_ENCAP_TYPE: | |
727 | cfg->fc_encap_type = nla_get_u16(attr); | |
c255bd68 DA |
728 | err = lwtunnel_valid_encap_type(cfg->fc_encap_type, |
729 | extack); | |
9ed59592 DA |
730 | if (err < 0) |
731 | goto errout; | |
571e7226 | 732 | break; |
1da177e4 LT |
733 | } |
734 | } | |
4e902c57 | 735 | |
1da177e4 | 736 | return 0; |
4e902c57 TG |
737 | errout: |
738 | return err; | |
1da177e4 LT |
739 | } |
740 | ||
c21ef3e3 DA |
741 | static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh, |
742 | struct netlink_ext_ack *extack) | |
1da177e4 | 743 | { |
3b1e0a65 | 744 | struct net *net = sock_net(skb->sk); |
4e902c57 TG |
745 | struct fib_config cfg; |
746 | struct fib_table *tb; | |
747 | int err; | |
1da177e4 | 748 | |
6d8422a1 | 749 | err = rtm_to_fib_config(net, skb, nlh, &cfg, extack); |
4e902c57 TG |
750 | if (err < 0) |
751 | goto errout; | |
1da177e4 | 752 | |
8ad4942c | 753 | tb = fib_get_table(net, cfg.fc_table); |
51456b29 | 754 | if (!tb) { |
c3ab2b4e | 755 | NL_SET_ERR_MSG(extack, "FIB table does not exist"); |
4e902c57 TG |
756 | err = -ESRCH; |
757 | goto errout; | |
758 | } | |
759 | ||
78055998 | 760 | err = fib_table_delete(net, tb, &cfg, extack); |
4e902c57 TG |
761 | errout: |
762 | return err; | |
1da177e4 LT |
763 | } |
764 | ||
c21ef3e3 DA |
765 | static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh, |
766 | struct netlink_ext_ack *extack) | |
1da177e4 | 767 | { |
3b1e0a65 | 768 | struct net *net = sock_net(skb->sk); |
4e902c57 TG |
769 | struct fib_config cfg; |
770 | struct fib_table *tb; | |
771 | int err; | |
1da177e4 | 772 | |
6d8422a1 | 773 | err = rtm_to_fib_config(net, skb, nlh, &cfg, extack); |
4e902c57 TG |
774 | if (err < 0) |
775 | goto errout; | |
1da177e4 | 776 | |
226b0b4a | 777 | tb = fib_new_table(net, cfg.fc_table); |
51456b29 | 778 | if (!tb) { |
4e902c57 TG |
779 | err = -ENOBUFS; |
780 | goto errout; | |
781 | } | |
782 | ||
6d8422a1 | 783 | err = fib_table_insert(net, tb, &cfg, extack); |
6e617de8 PA |
784 | if (!err && cfg.fc_type == RTN_LOCAL) |
785 | net->ipv4.fib_has_custom_local_routes = true; | |
4e902c57 TG |
786 | errout: |
787 | return err; | |
1da177e4 LT |
788 | } |
789 | ||
63f3444f | 790 | static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb) |
1da177e4 | 791 | { |
3b1e0a65 | 792 | struct net *net = sock_net(skb->sk); |
1af5a8c4 PM |
793 | unsigned int h, s_h; |
794 | unsigned int e = 0, s_e; | |
1da177e4 | 795 | struct fib_table *tb; |
e4aef8ae | 796 | struct hlist_head *head; |
f6c5775f | 797 | int dumped = 0, err; |
1da177e4 | 798 | |
be403ea1 TG |
799 | if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) && |
800 | ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED) | |
0b8c7f6f | 801 | return skb->len; |
1da177e4 | 802 | |
1af5a8c4 PM |
803 | s_h = cb->args[0]; |
804 | s_e = cb->args[1]; | |
805 | ||
a7e53531 AD |
806 | rcu_read_lock(); |
807 | ||
1af5a8c4 PM |
808 | for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) { |
809 | e = 0; | |
e4aef8ae | 810 | head = &net->ipv4.fib_table_hash[h]; |
a7e53531 | 811 | hlist_for_each_entry_rcu(tb, head, tb_hlist) { |
1af5a8c4 PM |
812 | if (e < s_e) |
813 | goto next; | |
814 | if (dumped) | |
815 | memset(&cb->args[2], 0, sizeof(cb->args) - | |
e905a9ed | 816 | 2 * sizeof(cb->args[0])); |
f6c5775f DA |
817 | err = fib_table_dump(tb, skb, cb); |
818 | if (err < 0) { | |
819 | if (likely(skb->len)) | |
820 | goto out; | |
821 | ||
822 | goto out_err; | |
823 | } | |
1af5a8c4 PM |
824 | dumped = 1; |
825 | next: | |
826 | e++; | |
827 | } | |
1da177e4 | 828 | } |
1af5a8c4 | 829 | out: |
f6c5775f DA |
830 | err = skb->len; |
831 | out_err: | |
a7e53531 AD |
832 | rcu_read_unlock(); |
833 | ||
1af5a8c4 PM |
834 | cb->args[1] = e; |
835 | cb->args[0] = h; | |
1da177e4 | 836 | |
f6c5775f | 837 | return err; |
1da177e4 LT |
838 | } |
839 | ||
840 | /* Prepare and feed intra-kernel routing request. | |
6a31d2a9 ED |
841 | * Really, it should be netlink message, but :-( netlink |
842 | * can be not configured, so that we feed it directly | |
843 | * to fib engine. It is legal, because all events occur | |
844 | * only when netlink is already locked. | |
1da177e4 | 845 | */ |
81f7bf6c | 846 | static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa) |
1da177e4 | 847 | { |
c346dca1 | 848 | struct net *net = dev_net(ifa->ifa_dev->dev); |
3236b004 | 849 | u32 tb_id = l3mdev_fib_table(ifa->ifa_dev->dev); |
4e902c57 TG |
850 | struct fib_table *tb; |
851 | struct fib_config cfg = { | |
852 | .fc_protocol = RTPROT_KERNEL, | |
853 | .fc_type = type, | |
854 | .fc_dst = dst, | |
855 | .fc_dst_len = dst_len, | |
856 | .fc_prefsrc = ifa->ifa_local, | |
857 | .fc_oif = ifa->ifa_dev->dev->ifindex, | |
858 | .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND, | |
4d1169c1 | 859 | .fc_nlinfo = { |
4b5d47d4 | 860 | .nl_net = net, |
4d1169c1 | 861 | }, |
4e902c57 | 862 | }; |
1da177e4 | 863 | |
021dd3b8 DA |
864 | if (!tb_id) |
865 | tb_id = (type == RTN_UNICAST) ? RT_TABLE_MAIN : RT_TABLE_LOCAL; | |
1da177e4 | 866 | |
021dd3b8 | 867 | tb = fib_new_table(net, tb_id); |
51456b29 | 868 | if (!tb) |
1da177e4 LT |
869 | return; |
870 | ||
4e902c57 | 871 | cfg.fc_table = tb->tb_id; |
1da177e4 | 872 | |
4e902c57 TG |
873 | if (type != RTN_LOCAL) |
874 | cfg.fc_scope = RT_SCOPE_LINK; | |
875 | else | |
876 | cfg.fc_scope = RT_SCOPE_HOST; | |
1da177e4 LT |
877 | |
878 | if (cmd == RTM_NEWROUTE) | |
6d8422a1 | 879 | fib_table_insert(net, tb, &cfg, NULL); |
1da177e4 | 880 | else |
78055998 | 881 | fib_table_delete(net, tb, &cfg, NULL); |
1da177e4 LT |
882 | } |
883 | ||
0ff60a45 | 884 | void fib_add_ifaddr(struct in_ifaddr *ifa) |
1da177e4 LT |
885 | { |
886 | struct in_device *in_dev = ifa->ifa_dev; | |
887 | struct net_device *dev = in_dev->dev; | |
888 | struct in_ifaddr *prim = ifa; | |
a144ea4b AV |
889 | __be32 mask = ifa->ifa_mask; |
890 | __be32 addr = ifa->ifa_local; | |
6a31d2a9 | 891 | __be32 prefix = ifa->ifa_address & mask; |
1da177e4 | 892 | |
6a31d2a9 | 893 | if (ifa->ifa_flags & IFA_F_SECONDARY) { |
1da177e4 | 894 | prim = inet_ifa_byprefix(in_dev, prefix, mask); |
51456b29 | 895 | if (!prim) { |
058bd4d2 | 896 | pr_warn("%s: bug: prim == NULL\n", __func__); |
1da177e4 LT |
897 | return; |
898 | } | |
899 | } | |
900 | ||
901 | fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim); | |
902 | ||
6a31d2a9 | 903 | if (!(dev->flags & IFF_UP)) |
1da177e4 LT |
904 | return; |
905 | ||
906 | /* Add broadcast address, if it is explicitly assigned. */ | |
a144ea4b | 907 | if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF)) |
1da177e4 LT |
908 | fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim); |
909 | ||
6a31d2a9 | 910 | if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) && |
1da177e4 | 911 | (prefix != addr || ifa->ifa_prefixlen < 32)) { |
7b131180 PA |
912 | if (!(ifa->ifa_flags & IFA_F_NOPREFIXROUTE)) |
913 | fib_magic(RTM_NEWROUTE, | |
914 | dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST, | |
915 | prefix, ifa->ifa_prefixlen, prim); | |
1da177e4 LT |
916 | |
917 | /* Add network specific broadcasts, when it takes a sense */ | |
918 | if (ifa->ifa_prefixlen < 31) { | |
919 | fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim); | |
6a31d2a9 ED |
920 | fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask, |
921 | 32, prim); | |
1da177e4 LT |
922 | } |
923 | } | |
924 | } | |
925 | ||
e6abbaa2 JA |
926 | /* Delete primary or secondary address. |
927 | * Optionally, on secondary address promotion consider the addresses | |
928 | * from subnet iprim as deleted, even if they are in device list. | |
929 | * In this case the secondary ifa can be in device list. | |
930 | */ | |
931 | void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim) | |
1da177e4 LT |
932 | { |
933 | struct in_device *in_dev = ifa->ifa_dev; | |
934 | struct net_device *dev = in_dev->dev; | |
935 | struct in_ifaddr *ifa1; | |
e6abbaa2 | 936 | struct in_ifaddr *prim = ifa, *prim1 = NULL; |
6a31d2a9 ED |
937 | __be32 brd = ifa->ifa_address | ~ifa->ifa_mask; |
938 | __be32 any = ifa->ifa_address & ifa->ifa_mask; | |
1da177e4 LT |
939 | #define LOCAL_OK 1 |
940 | #define BRD_OK 2 | |
941 | #define BRD0_OK 4 | |
942 | #define BRD1_OK 8 | |
95c96174 | 943 | unsigned int ok = 0; |
e6abbaa2 JA |
944 | int subnet = 0; /* Primary network */ |
945 | int gone = 1; /* Address is missing */ | |
946 | int same_prefsrc = 0; /* Another primary with same IP */ | |
1da177e4 | 947 | |
e6abbaa2 | 948 | if (ifa->ifa_flags & IFA_F_SECONDARY) { |
1da177e4 | 949 | prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask); |
51456b29 | 950 | if (!prim) { |
391a2033 PA |
951 | /* if the device has been deleted, we don't perform |
952 | * address promotion | |
953 | */ | |
954 | if (!in_dev->dead) | |
955 | pr_warn("%s: bug: prim == NULL\n", __func__); | |
1da177e4 LT |
956 | return; |
957 | } | |
e6abbaa2 | 958 | if (iprim && iprim != prim) { |
058bd4d2 | 959 | pr_warn("%s: bug: iprim != prim\n", __func__); |
e6abbaa2 JA |
960 | return; |
961 | } | |
962 | } else if (!ipv4_is_zeronet(any) && | |
963 | (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) { | |
7b131180 PA |
964 | if (!(ifa->ifa_flags & IFA_F_NOPREFIXROUTE)) |
965 | fib_magic(RTM_DELROUTE, | |
966 | dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST, | |
967 | any, ifa->ifa_prefixlen, prim); | |
e6abbaa2 | 968 | subnet = 1; |
1da177e4 LT |
969 | } |
970 | ||
fbd40ea0 DM |
971 | if (in_dev->dead) |
972 | goto no_promotions; | |
973 | ||
1da177e4 | 974 | /* Deletion is more complicated than add. |
6a31d2a9 ED |
975 | * We should take care of not to delete too much :-) |
976 | * | |
977 | * Scan address list to be sure that addresses are really gone. | |
1da177e4 LT |
978 | */ |
979 | ||
980 | for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) { | |
e6abbaa2 JA |
981 | if (ifa1 == ifa) { |
982 | /* promotion, keep the IP */ | |
983 | gone = 0; | |
984 | continue; | |
985 | } | |
986 | /* Ignore IFAs from our subnet */ | |
987 | if (iprim && ifa1->ifa_mask == iprim->ifa_mask && | |
988 | inet_ifa_match(ifa1->ifa_address, iprim)) | |
989 | continue; | |
990 | ||
991 | /* Ignore ifa1 if it uses different primary IP (prefsrc) */ | |
992 | if (ifa1->ifa_flags & IFA_F_SECONDARY) { | |
993 | /* Another address from our subnet? */ | |
994 | if (ifa1->ifa_mask == prim->ifa_mask && | |
995 | inet_ifa_match(ifa1->ifa_address, prim)) | |
996 | prim1 = prim; | |
997 | else { | |
998 | /* We reached the secondaries, so | |
999 | * same_prefsrc should be determined. | |
1000 | */ | |
1001 | if (!same_prefsrc) | |
1002 | continue; | |
1003 | /* Search new prim1 if ifa1 is not | |
1004 | * using the current prim1 | |
1005 | */ | |
1006 | if (!prim1 || | |
1007 | ifa1->ifa_mask != prim1->ifa_mask || | |
1008 | !inet_ifa_match(ifa1->ifa_address, prim1)) | |
1009 | prim1 = inet_ifa_byprefix(in_dev, | |
1010 | ifa1->ifa_address, | |
1011 | ifa1->ifa_mask); | |
1012 | if (!prim1) | |
1013 | continue; | |
1014 | if (prim1->ifa_local != prim->ifa_local) | |
1015 | continue; | |
1016 | } | |
1017 | } else { | |
1018 | if (prim->ifa_local != ifa1->ifa_local) | |
1019 | continue; | |
1020 | prim1 = ifa1; | |
1021 | if (prim != prim1) | |
1022 | same_prefsrc = 1; | |
1023 | } | |
1da177e4 LT |
1024 | if (ifa->ifa_local == ifa1->ifa_local) |
1025 | ok |= LOCAL_OK; | |
1026 | if (ifa->ifa_broadcast == ifa1->ifa_broadcast) | |
1027 | ok |= BRD_OK; | |
1028 | if (brd == ifa1->ifa_broadcast) | |
1029 | ok |= BRD1_OK; | |
1030 | if (any == ifa1->ifa_broadcast) | |
1031 | ok |= BRD0_OK; | |
e6abbaa2 JA |
1032 | /* primary has network specific broadcasts */ |
1033 | if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) { | |
1034 | __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask; | |
1035 | __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask; | |
1036 | ||
1037 | if (!ipv4_is_zeronet(any1)) { | |
1038 | if (ifa->ifa_broadcast == brd1 || | |
1039 | ifa->ifa_broadcast == any1) | |
1040 | ok |= BRD_OK; | |
1041 | if (brd == brd1 || brd == any1) | |
1042 | ok |= BRD1_OK; | |
1043 | if (any == brd1 || any == any1) | |
1044 | ok |= BRD0_OK; | |
1045 | } | |
1046 | } | |
1da177e4 LT |
1047 | } |
1048 | ||
fbd40ea0 | 1049 | no_promotions: |
6a31d2a9 | 1050 | if (!(ok & BRD_OK)) |
1da177e4 | 1051 | fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim); |
e6abbaa2 JA |
1052 | if (subnet && ifa->ifa_prefixlen < 31) { |
1053 | if (!(ok & BRD1_OK)) | |
1054 | fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim); | |
1055 | if (!(ok & BRD0_OK)) | |
1056 | fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim); | |
1057 | } | |
6a31d2a9 | 1058 | if (!(ok & LOCAL_OK)) { |
30bbaa19 DA |
1059 | unsigned int addr_type; |
1060 | ||
1da177e4 LT |
1061 | fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim); |
1062 | ||
1063 | /* Check, that this local address finally disappeared. */ | |
30bbaa19 DA |
1064 | addr_type = inet_addr_type_dev_table(dev_net(dev), dev, |
1065 | ifa->ifa_local); | |
1066 | if (gone && addr_type != RTN_LOCAL) { | |
1da177e4 | 1067 | /* And the last, but not the least thing. |
6a31d2a9 ED |
1068 | * We must flush stray FIB entries. |
1069 | * | |
1070 | * First of all, we scan fib_info list searching | |
1071 | * for stray nexthop entries, then ignite fib_flush. | |
1072 | */ | |
5a56a0b3 | 1073 | if (fib_sync_down_addr(dev, ifa->ifa_local)) |
c346dca1 | 1074 | fib_flush(dev_net(dev)); |
1da177e4 LT |
1075 | } |
1076 | } | |
1077 | #undef LOCAL_OK | |
1078 | #undef BRD_OK | |
1079 | #undef BRD0_OK | |
1080 | #undef BRD1_OK | |
1081 | } | |
1082 | ||
345e9b54 | 1083 | static void nl_fib_lookup(struct net *net, struct fib_result_nl *frn) |
246955fe | 1084 | { |
e905a9ed | 1085 | |
246955fe | 1086 | struct fib_result res; |
9ade2286 DM |
1087 | struct flowi4 fl4 = { |
1088 | .flowi4_mark = frn->fl_mark, | |
1089 | .daddr = frn->fl_addr, | |
1090 | .flowi4_tos = frn->fl_tos, | |
1091 | .flowi4_scope = frn->fl_scope, | |
6a31d2a9 | 1092 | }; |
345e9b54 AD |
1093 | struct fib_table *tb; |
1094 | ||
1095 | rcu_read_lock(); | |
1096 | ||
1097 | tb = fib_get_table(net, frn->tb_id_in); | |
1194ed0a AK |
1098 | |
1099 | frn->err = -ENOENT; | |
246955fe RO |
1100 | if (tb) { |
1101 | local_bh_disable(); | |
1102 | ||
1103 | frn->tb_id = tb->tb_id; | |
9ade2286 | 1104 | frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF); |
246955fe RO |
1105 | |
1106 | if (!frn->err) { | |
1107 | frn->prefixlen = res.prefixlen; | |
1108 | frn->nh_sel = res.nh_sel; | |
1109 | frn->type = res.type; | |
1110 | frn->scope = res.scope; | |
1111 | } | |
1112 | local_bh_enable(); | |
1113 | } | |
345e9b54 AD |
1114 | |
1115 | rcu_read_unlock(); | |
246955fe RO |
1116 | } |
1117 | ||
28f7b036 | 1118 | static void nl_fib_input(struct sk_buff *skb) |
246955fe | 1119 | { |
6bd48fcf | 1120 | struct net *net; |
246955fe | 1121 | struct fib_result_nl *frn; |
28f7b036 | 1122 | struct nlmsghdr *nlh; |
15e47304 | 1123 | u32 portid; |
1194ed0a | 1124 | |
3b1e0a65 | 1125 | net = sock_net(skb->sk); |
b529ccf2 | 1126 | nlh = nlmsg_hdr(skb); |
c64c0b3c ED |
1127 | if (skb->len < nlmsg_total_size(sizeof(*frn)) || |
1128 | skb->len < nlh->nlmsg_len || | |
573ce260 | 1129 | nlmsg_len(nlh) < sizeof(*frn)) |
ea86575e | 1130 | return; |
d883a036 | 1131 | |
3a36515f | 1132 | skb = netlink_skb_clone(skb, GFP_KERNEL); |
51456b29 | 1133 | if (!skb) |
d883a036 DL |
1134 | return; |
1135 | nlh = nlmsg_hdr(skb); | |
e905a9ed | 1136 | |
573ce260 | 1137 | frn = (struct fib_result_nl *) nlmsg_data(nlh); |
345e9b54 | 1138 | nl_fib_lookup(net, frn); |
e905a9ed | 1139 | |
28a28283 | 1140 | portid = NETLINK_CB(skb).portid; /* netlink portid */ |
15e47304 | 1141 | NETLINK_CB(skb).portid = 0; /* from kernel */ |
ac6d439d | 1142 | NETLINK_CB(skb).dst_group = 0; /* unicast */ |
15e47304 | 1143 | netlink_unicast(net->ipv4.fibnl, skb, portid, MSG_DONTWAIT); |
e905a9ed | 1144 | } |
246955fe | 1145 | |
2c8c1e72 | 1146 | static int __net_init nl_fib_lookup_init(struct net *net) |
246955fe | 1147 | { |
6bd48fcf | 1148 | struct sock *sk; |
a31f2d17 PNA |
1149 | struct netlink_kernel_cfg cfg = { |
1150 | .input = nl_fib_input, | |
1151 | }; | |
1152 | ||
9f00d977 | 1153 | sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, &cfg); |
51456b29 | 1154 | if (!sk) |
7b1a74fd | 1155 | return -EAFNOSUPPORT; |
6bd48fcf | 1156 | net->ipv4.fibnl = sk; |
7b1a74fd DL |
1157 | return 0; |
1158 | } | |
1159 | ||
1160 | static void nl_fib_lookup_exit(struct net *net) | |
1161 | { | |
b7c6ba6e | 1162 | netlink_kernel_release(net->ipv4.fibnl); |
775516bf | 1163 | net->ipv4.fibnl = NULL; |
246955fe RO |
1164 | } |
1165 | ||
4f823def JA |
1166 | static void fib_disable_ip(struct net_device *dev, unsigned long event, |
1167 | bool force) | |
1da177e4 | 1168 | { |
4f823def | 1169 | if (fib_sync_down_dev(dev, event, force)) |
c346dca1 | 1170 | fib_flush(dev_net(dev)); |
06b4fc52 GF |
1171 | else |
1172 | rt_cache_flush(dev_net(dev)); | |
1da177e4 LT |
1173 | arp_ifdown(dev); |
1174 | } | |
1175 | ||
1176 | static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr) | |
1177 | { | |
6ed2533e | 1178 | struct in_ifaddr *ifa = (struct in_ifaddr *)ptr; |
76e6ebfb | 1179 | struct net_device *dev = ifa->ifa_dev->dev; |
436c3b66 | 1180 | struct net *net = dev_net(dev); |
1da177e4 LT |
1181 | |
1182 | switch (event) { | |
1183 | case NETDEV_UP: | |
1184 | fib_add_ifaddr(ifa); | |
1185 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
8a3d0316 | 1186 | fib_sync_up(dev, RTNH_F_DEAD); |
1da177e4 | 1187 | #endif |
436c3b66 | 1188 | atomic_inc(&net->ipv4.dev_addr_genid); |
4ccfe6d4 | 1189 | rt_cache_flush(dev_net(dev)); |
1da177e4 LT |
1190 | break; |
1191 | case NETDEV_DOWN: | |
e6abbaa2 | 1192 | fib_del_ifaddr(ifa, NULL); |
436c3b66 | 1193 | atomic_inc(&net->ipv4.dev_addr_genid); |
51456b29 | 1194 | if (!ifa->ifa_dev->ifa_list) { |
1da177e4 | 1195 | /* Last address was deleted from this interface. |
6a31d2a9 | 1196 | * Disable IP. |
1da177e4 | 1197 | */ |
4f823def | 1198 | fib_disable_ip(dev, event, true); |
1da177e4 | 1199 | } else { |
4ccfe6d4 | 1200 | rt_cache_flush(dev_net(dev)); |
1da177e4 LT |
1201 | } |
1202 | break; | |
1203 | } | |
1204 | return NOTIFY_DONE; | |
1205 | } | |
1206 | ||
1207 | static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr) | |
1208 | { | |
351638e7 | 1209 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); |
7f49e7a3 | 1210 | struct netdev_notifier_changeupper_info *info; |
0115e8e3 | 1211 | struct in_device *in_dev; |
436c3b66 | 1212 | struct net *net = dev_net(dev); |
8a3d0316 | 1213 | unsigned int flags; |
1da177e4 LT |
1214 | |
1215 | if (event == NETDEV_UNREGISTER) { | |
4f823def | 1216 | fib_disable_ip(dev, event, true); |
caacf05e | 1217 | rt_flush_dev(dev); |
1da177e4 LT |
1218 | return NOTIFY_DONE; |
1219 | } | |
1220 | ||
0115e8e3 | 1221 | in_dev = __in_dev_get_rtnl(dev); |
a0065f26 OH |
1222 | if (!in_dev) |
1223 | return NOTIFY_DONE; | |
0115e8e3 | 1224 | |
1da177e4 LT |
1225 | switch (event) { |
1226 | case NETDEV_UP: | |
1227 | for_ifa(in_dev) { | |
1228 | fib_add_ifaddr(ifa); | |
1229 | } endfor_ifa(in_dev); | |
1230 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
8a3d0316 | 1231 | fib_sync_up(dev, RTNH_F_DEAD); |
1da177e4 | 1232 | #endif |
436c3b66 | 1233 | atomic_inc(&net->ipv4.dev_addr_genid); |
4ccfe6d4 | 1234 | rt_cache_flush(net); |
1da177e4 LT |
1235 | break; |
1236 | case NETDEV_DOWN: | |
4f823def | 1237 | fib_disable_ip(dev, event, false); |
1da177e4 | 1238 | break; |
1da177e4 | 1239 | case NETDEV_CHANGE: |
8a3d0316 AG |
1240 | flags = dev_get_flags(dev); |
1241 | if (flags & (IFF_RUNNING | IFF_LOWER_UP)) | |
1242 | fib_sync_up(dev, RTNH_F_LINKDOWN); | |
1243 | else | |
4f823def | 1244 | fib_sync_down_dev(dev, event, false); |
8a3d0316 AG |
1245 | /* fall through */ |
1246 | case NETDEV_CHANGEMTU: | |
4ccfe6d4 | 1247 | rt_cache_flush(net); |
1da177e4 | 1248 | break; |
7f49e7a3 DA |
1249 | case NETDEV_CHANGEUPPER: |
1250 | info = ptr; | |
1251 | /* flush all routes if dev is linked to or unlinked from | |
1252 | * an L3 master device (e.g., VRF) | |
1253 | */ | |
1254 | if (info->upper_dev && netif_is_l3_master(info->upper_dev)) | |
1255 | fib_disable_ip(dev, NETDEV_DOWN, true); | |
1256 | break; | |
1da177e4 LT |
1257 | } |
1258 | return NOTIFY_DONE; | |
1259 | } | |
1260 | ||
1261 | static struct notifier_block fib_inetaddr_notifier = { | |
6ed2533e | 1262 | .notifier_call = fib_inetaddr_event, |
1da177e4 LT |
1263 | }; |
1264 | ||
1265 | static struct notifier_block fib_netdev_notifier = { | |
6ed2533e | 1266 | .notifier_call = fib_netdev_event, |
1da177e4 LT |
1267 | }; |
1268 | ||
7b1a74fd | 1269 | static int __net_init ip_fib_net_init(struct net *net) |
1da177e4 | 1270 | { |
dce5cbee | 1271 | int err; |
10da66f7 ED |
1272 | size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ; |
1273 | ||
04b1d4e5 IS |
1274 | err = fib4_notifier_init(net); |
1275 | if (err) | |
1276 | return err; | |
cacaad11 | 1277 | |
10da66f7 ED |
1278 | /* Avoid false sharing : Use at least a full cache line */ |
1279 | size = max_t(size_t, size, L1_CACHE_BYTES); | |
1af5a8c4 | 1280 | |
10da66f7 | 1281 | net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL); |
04b1d4e5 IS |
1282 | if (!net->ipv4.fib_table_hash) { |
1283 | err = -ENOMEM; | |
1284 | goto err_table_hash_alloc; | |
1285 | } | |
e4aef8ae | 1286 | |
dce5cbee DL |
1287 | err = fib4_rules_init(net); |
1288 | if (err < 0) | |
04b1d4e5 | 1289 | goto err_rules_init; |
dce5cbee DL |
1290 | return 0; |
1291 | ||
04b1d4e5 | 1292 | err_rules_init: |
dce5cbee | 1293 | kfree(net->ipv4.fib_table_hash); |
04b1d4e5 IS |
1294 | err_table_hash_alloc: |
1295 | fib4_notifier_exit(net); | |
dce5cbee | 1296 | return err; |
7b1a74fd | 1297 | } |
1da177e4 | 1298 | |
2c8c1e72 | 1299 | static void ip_fib_net_exit(struct net *net) |
7b1a74fd | 1300 | { |
b4681c28 | 1301 | int i; |
7b1a74fd | 1302 | |
6dede75b | 1303 | rtnl_lock(); |
6e47d6ca | 1304 | #ifdef CONFIG_IP_MULTIPLE_TABLES |
6e47d6ca AD |
1305 | RCU_INIT_POINTER(net->ipv4.fib_main, NULL); |
1306 | RCU_INIT_POINTER(net->ipv4.fib_default, NULL); | |
1307 | #endif | |
b4681c28 IS |
1308 | /* Destroy the tables in reverse order to guarantee that the |
1309 | * local table, ID 255, is destroyed before the main table, ID | |
1310 | * 254. This is necessary as the local table may contain | |
1311 | * references to data contained in the main table. | |
1312 | */ | |
1313 | for (i = FIB_TABLE_HASHSZ - 1; i >= 0; i--) { | |
a7e53531 | 1314 | struct hlist_head *head = &net->ipv4.fib_table_hash[i]; |
b67bfe0d | 1315 | struct hlist_node *tmp; |
a7e53531 AD |
1316 | struct fib_table *tb; |
1317 | ||
b67bfe0d SL |
1318 | hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) { |
1319 | hlist_del(&tb->tb_hlist); | |
b90eb754 | 1320 | fib_table_flush(net, tb); |
4aa2c466 | 1321 | fib_free_table(tb); |
7b1a74fd DL |
1322 | } |
1323 | } | |
ad88d051 AD |
1324 | |
1325 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
1326 | fib4_rules_exit(net); | |
1327 | #endif | |
e2666f84 | 1328 | rtnl_unlock(); |
e4aef8ae | 1329 | kfree(net->ipv4.fib_table_hash); |
04b1d4e5 | 1330 | fib4_notifier_exit(net); |
7b1a74fd DL |
1331 | } |
1332 | ||
1333 | static int __net_init fib_net_init(struct net *net) | |
1334 | { | |
1335 | int error; | |
1336 | ||
f4530fa5 DM |
1337 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1338 | net->ipv4.fib_num_tclassid_users = 0; | |
1339 | #endif | |
7b1a74fd DL |
1340 | error = ip_fib_net_init(net); |
1341 | if (error < 0) | |
1342 | goto out; | |
1343 | error = nl_fib_lookup_init(net); | |
1344 | if (error < 0) | |
1345 | goto out_nlfl; | |
1346 | error = fib_proc_init(net); | |
1347 | if (error < 0) | |
1348 | goto out_proc; | |
1349 | out: | |
1350 | return error; | |
1351 | ||
1352 | out_proc: | |
1353 | nl_fib_lookup_exit(net); | |
1354 | out_nlfl: | |
1355 | ip_fib_net_exit(net); | |
1356 | goto out; | |
1357 | } | |
1358 | ||
1359 | static void __net_exit fib_net_exit(struct net *net) | |
1360 | { | |
1361 | fib_proc_exit(net); | |
1362 | nl_fib_lookup_exit(net); | |
1363 | ip_fib_net_exit(net); | |
1364 | } | |
1365 | ||
1366 | static struct pernet_operations fib_net_ops = { | |
1367 | .init = fib_net_init, | |
1368 | .exit = fib_net_exit, | |
1369 | }; | |
1370 | ||
1371 | void __init ip_fib_init(void) | |
1372 | { | |
8799a221 | 1373 | fib_trie_init(); |
7b1a74fd DL |
1374 | |
1375 | register_pernet_subsys(&fib_net_ops); | |
8799a221 | 1376 | |
7b1a74fd DL |
1377 | register_netdevice_notifier(&fib_netdev_notifier); |
1378 | register_inetaddr_notifier(&fib_inetaddr_notifier); | |
7f9b8052 | 1379 | |
b97bac64 FW |
1380 | rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, 0); |
1381 | rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, 0); | |
1382 | rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, 0); | |
1da177e4 | 1383 | } |