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 LT |
16 | #include <linux/module.h> |
17 | #include <asm/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> |
1da177e4 | 48 | |
1da177e4 LT |
49 | #ifndef CONFIG_IP_MULTIPLE_TABLES |
50 | ||
7b1a74fd | 51 | static int __net_init fib4_rules_init(struct net *net) |
c3e9a353 | 52 | { |
93456b6d DL |
53 | struct fib_table *local_table, *main_table; |
54 | ||
5348ba85 | 55 | local_table = fib_trie_table(RT_TABLE_LOCAL); |
93456b6d | 56 | if (local_table == NULL) |
dbb50165 DL |
57 | return -ENOMEM; |
58 | ||
5348ba85 | 59 | main_table = fib_trie_table(RT_TABLE_MAIN); |
93456b6d | 60 | if (main_table == NULL) |
dbb50165 DL |
61 | goto fail; |
62 | ||
93456b6d | 63 | hlist_add_head_rcu(&local_table->tb_hlist, |
e4aef8ae | 64 | &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]); |
93456b6d | 65 | hlist_add_head_rcu(&main_table->tb_hlist, |
e4aef8ae | 66 | &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]); |
dbb50165 DL |
67 | return 0; |
68 | ||
69 | fail: | |
8274a97a | 70 | fib_free_table(local_table); |
dbb50165 | 71 | return -ENOMEM; |
c3e9a353 | 72 | } |
1af5a8c4 | 73 | #else |
1da177e4 | 74 | |
8ad4942c | 75 | struct fib_table *fib_new_table(struct net *net, u32 id) |
1da177e4 LT |
76 | { |
77 | struct fib_table *tb; | |
1af5a8c4 | 78 | unsigned int h; |
1da177e4 | 79 | |
1af5a8c4 PM |
80 | if (id == 0) |
81 | id = RT_TABLE_MAIN; | |
8ad4942c | 82 | tb = fib_get_table(net, id); |
1af5a8c4 PM |
83 | if (tb) |
84 | return tb; | |
7f9b8052 | 85 | |
5348ba85 | 86 | tb = fib_trie_table(id); |
1da177e4 LT |
87 | if (!tb) |
88 | return NULL; | |
f4530fa5 DM |
89 | |
90 | switch (id) { | |
91 | case RT_TABLE_LOCAL: | |
92 | net->ipv4.fib_local = tb; | |
93 | break; | |
94 | ||
95 | case RT_TABLE_MAIN: | |
96 | net->ipv4.fib_main = tb; | |
97 | break; | |
98 | ||
99 | case RT_TABLE_DEFAULT: | |
100 | net->ipv4.fib_default = tb; | |
101 | break; | |
102 | ||
103 | default: | |
104 | break; | |
105 | } | |
106 | ||
1af5a8c4 | 107 | h = id & (FIB_TABLE_HASHSZ - 1); |
e4aef8ae | 108 | hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]); |
1da177e4 LT |
109 | return tb; |
110 | } | |
111 | ||
8ad4942c | 112 | struct fib_table *fib_get_table(struct net *net, u32 id) |
1af5a8c4 PM |
113 | { |
114 | struct fib_table *tb; | |
e4aef8ae | 115 | struct hlist_head *head; |
1af5a8c4 | 116 | unsigned int h; |
1da177e4 | 117 | |
1af5a8c4 PM |
118 | if (id == 0) |
119 | id = RT_TABLE_MAIN; | |
120 | h = id & (FIB_TABLE_HASHSZ - 1); | |
e4aef8ae | 121 | |
1af5a8c4 | 122 | rcu_read_lock(); |
e4aef8ae | 123 | head = &net->ipv4.fib_table_hash[h]; |
b67bfe0d | 124 | hlist_for_each_entry_rcu(tb, head, tb_hlist) { |
1af5a8c4 PM |
125 | if (tb->tb_id == id) { |
126 | rcu_read_unlock(); | |
127 | return tb; | |
128 | } | |
129 | } | |
130 | rcu_read_unlock(); | |
131 | return NULL; | |
132 | } | |
1da177e4 LT |
133 | #endif /* CONFIG_IP_MULTIPLE_TABLES */ |
134 | ||
e4aef8ae | 135 | static void fib_flush(struct net *net) |
1da177e4 LT |
136 | { |
137 | int flushed = 0; | |
1da177e4 | 138 | struct fib_table *tb; |
e4aef8ae | 139 | struct hlist_head *head; |
1af5a8c4 | 140 | unsigned int h; |
1da177e4 | 141 | |
1af5a8c4 | 142 | for (h = 0; h < FIB_TABLE_HASHSZ; h++) { |
e4aef8ae | 143 | head = &net->ipv4.fib_table_hash[h]; |
b67bfe0d | 144 | hlist_for_each_entry(tb, head, tb_hlist) |
16c6cf8b | 145 | flushed += fib_table_flush(tb); |
1da177e4 | 146 | } |
1da177e4 LT |
147 | |
148 | if (flushed) | |
4ccfe6d4 | 149 | rt_cache_flush(net); |
1da177e4 LT |
150 | } |
151 | ||
05538116 LAT |
152 | /* |
153 | * Find address type as if only "dev" was present in the system. If | |
154 | * on_dev is NULL then all interfaces are taken into consideration. | |
155 | */ | |
95c96174 ED |
156 | static inline unsigned int __inet_dev_addr_type(struct net *net, |
157 | const struct net_device *dev, | |
158 | __be32 addr) | |
1da177e4 | 159 | { |
9ade2286 | 160 | struct flowi4 fl4 = { .daddr = addr }; |
1da177e4 | 161 | struct fib_result res; |
95c96174 | 162 | unsigned int ret = RTN_BROADCAST; |
03cf786c | 163 | struct fib_table *local_table; |
1da177e4 | 164 | |
1e637c74 | 165 | if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr)) |
1da177e4 | 166 | return RTN_BROADCAST; |
f97c1e0c | 167 | if (ipv4_is_multicast(addr)) |
1da177e4 LT |
168 | return RTN_MULTICAST; |
169 | ||
6b175b26 | 170 | local_table = fib_get_table(net, RT_TABLE_LOCAL); |
03cf786c | 171 | if (local_table) { |
1da177e4 | 172 | ret = RTN_UNICAST; |
ebc0ffae | 173 | rcu_read_lock(); |
9ade2286 | 174 | if (!fib_table_lookup(local_table, &fl4, &res, FIB_LOOKUP_NOREF)) { |
05538116 LAT |
175 | if (!dev || dev == res.fi->fib_dev) |
176 | ret = res.type; | |
1da177e4 | 177 | } |
ebc0ffae | 178 | rcu_read_unlock(); |
1da177e4 LT |
179 | } |
180 | return ret; | |
181 | } | |
182 | ||
6b175b26 | 183 | unsigned int inet_addr_type(struct net *net, __be32 addr) |
05538116 | 184 | { |
6b175b26 | 185 | return __inet_dev_addr_type(net, NULL, addr); |
05538116 | 186 | } |
4bc2f18b | 187 | EXPORT_SYMBOL(inet_addr_type); |
05538116 | 188 | |
6b175b26 EB |
189 | unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev, |
190 | __be32 addr) | |
05538116 | 191 | { |
6a31d2a9 | 192 | return __inet_dev_addr_type(net, dev, addr); |
05538116 | 193 | } |
4bc2f18b | 194 | EXPORT_SYMBOL(inet_dev_addr_type); |
05538116 | 195 | |
35ebf65e DM |
196 | __be32 fib_compute_spec_dst(struct sk_buff *skb) |
197 | { | |
198 | struct net_device *dev = skb->dev; | |
199 | struct in_device *in_dev; | |
200 | struct fib_result res; | |
a207a4b2 | 201 | struct rtable *rt; |
35ebf65e DM |
202 | struct flowi4 fl4; |
203 | struct net *net; | |
a207a4b2 | 204 | int scope; |
35ebf65e | 205 | |
a207a4b2 | 206 | rt = skb_rtable(skb); |
0cc535a2 JA |
207 | if ((rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | RTCF_LOCAL)) == |
208 | RTCF_LOCAL) | |
35ebf65e DM |
209 | return ip_hdr(skb)->daddr; |
210 | ||
211 | in_dev = __in_dev_get_rcu(dev); | |
212 | BUG_ON(!in_dev); | |
35ebf65e DM |
213 | |
214 | net = dev_net(dev); | |
a207a4b2 DM |
215 | |
216 | scope = RT_SCOPE_UNIVERSE; | |
217 | if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) { | |
218 | fl4.flowi4_oif = 0; | |
1fb9489b | 219 | fl4.flowi4_iif = LOOPBACK_IFINDEX; |
a207a4b2 DM |
220 | fl4.daddr = ip_hdr(skb)->saddr; |
221 | fl4.saddr = 0; | |
222 | fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos); | |
223 | fl4.flowi4_scope = scope; | |
224 | fl4.flowi4_mark = IN_DEV_SRC_VMARK(in_dev) ? skb->mark : 0; | |
225 | if (!fib_lookup(net, &fl4, &res)) | |
226 | return FIB_RES_PREFSRC(net, res); | |
227 | } else { | |
228 | scope = RT_SCOPE_LINK; | |
229 | } | |
230 | ||
231 | return inet_select_addr(dev, ip_hdr(skb)->saddr, scope); | |
35ebf65e DM |
232 | } |
233 | ||
1da177e4 | 234 | /* Given (packet source, input interface) and optional (dst, oif, tos): |
6a31d2a9 ED |
235 | * - (main) check, that source is valid i.e. not broadcast or our local |
236 | * address. | |
237 | * - figure out what "logical" interface this packet arrived | |
238 | * and calculate "specific destination" address. | |
239 | * - check, that packet arrived from expected physical interface. | |
ebc0ffae | 240 | * called with rcu_read_lock() |
1da177e4 | 241 | */ |
7a9bc9b8 DM |
242 | static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst, |
243 | u8 tos, int oif, struct net_device *dev, | |
244 | int rpf, struct in_device *idev, u32 *itag) | |
1da177e4 | 245 | { |
1dced6a8 | 246 | int ret, no_addr; |
1da177e4 | 247 | struct fib_result res; |
9e56e380 | 248 | struct flowi4 fl4; |
5b707aaa | 249 | struct net *net; |
9e56e380 | 250 | bool dev_match; |
1da177e4 | 251 | |
9ade2286 | 252 | fl4.flowi4_oif = 0; |
6a662719 | 253 | fl4.flowi4_iif = oif ? : LOOPBACK_IFINDEX; |
9ade2286 DM |
254 | fl4.daddr = src; |
255 | fl4.saddr = dst; | |
256 | fl4.flowi4_tos = tos; | |
257 | fl4.flowi4_scope = RT_SCOPE_UNIVERSE; | |
cc7e17ea | 258 | |
9e56e380 | 259 | no_addr = idev->ifa_list == NULL; |
990078af | 260 | |
9e56e380 | 261 | fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0; |
1da177e4 | 262 | |
c346dca1 | 263 | net = dev_net(dev); |
9ade2286 | 264 | if (fib_lookup(net, &fl4, &res)) |
1da177e4 | 265 | goto last_resort; |
1dced6a8 SB |
266 | if (res.type != RTN_UNICAST && |
267 | (res.type != RTN_LOCAL || !IN_DEV_ACCEPT_LOCAL(idev))) | |
268 | goto e_inval; | |
269 | if (!rpf && !fib_num_tclassid_users(dev_net(dev)) && | |
270 | (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) | |
271 | goto last_resort; | |
1da177e4 | 272 | fib_combine_itag(itag, &res); |
6f86b325 DM |
273 | dev_match = false; |
274 | ||
1da177e4 | 275 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
6f86b325 DM |
276 | for (ret = 0; ret < res.fi->fib_nhs; ret++) { |
277 | struct fib_nh *nh = &res.fi->fib_nh[ret]; | |
278 | ||
279 | if (nh->nh_dev == dev) { | |
280 | dev_match = true; | |
281 | break; | |
282 | } | |
283 | } | |
1da177e4 LT |
284 | #else |
285 | if (FIB_RES_DEV(res) == dev) | |
6f86b325 | 286 | dev_match = true; |
1da177e4 | 287 | #endif |
6f86b325 | 288 | if (dev_match) { |
1da177e4 | 289 | ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST; |
1da177e4 LT |
290 | return ret; |
291 | } | |
1da177e4 LT |
292 | if (no_addr) |
293 | goto last_resort; | |
c1cf8422 | 294 | if (rpf == 1) |
b5f7e755 | 295 | goto e_rpf; |
9ade2286 | 296 | fl4.flowi4_oif = dev->ifindex; |
1da177e4 LT |
297 | |
298 | ret = 0; | |
9ade2286 | 299 | if (fib_lookup(net, &fl4, &res) == 0) { |
41347dcd | 300 | if (res.type == RTN_UNICAST) |
1da177e4 | 301 | ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST; |
1da177e4 LT |
302 | } |
303 | return ret; | |
304 | ||
305 | last_resort: | |
306 | if (rpf) | |
b5f7e755 | 307 | goto e_rpf; |
1da177e4 LT |
308 | *itag = 0; |
309 | return 0; | |
310 | ||
1da177e4 LT |
311 | e_inval: |
312 | return -EINVAL; | |
b5f7e755 ED |
313 | e_rpf: |
314 | return -EXDEV; | |
1da177e4 LT |
315 | } |
316 | ||
7a9bc9b8 DM |
317 | /* Ignore rp_filter for packets protected by IPsec. */ |
318 | int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst, | |
319 | u8 tos, int oif, struct net_device *dev, | |
320 | struct in_device *idev, u32 *itag) | |
321 | { | |
322 | int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev); | |
323 | ||
e81da0e1 | 324 | if (!r && !fib_num_tclassid_users(dev_net(dev)) && |
1dced6a8 | 325 | IN_DEV_ACCEPT_LOCAL(idev) && |
e81da0e1 | 326 | (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) { |
7a9bc9b8 DM |
327 | *itag = 0; |
328 | return 0; | |
329 | } | |
330 | return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag); | |
331 | } | |
332 | ||
81f7bf6c | 333 | static inline __be32 sk_extract_addr(struct sockaddr *addr) |
4e902c57 TG |
334 | { |
335 | return ((struct sockaddr_in *) addr)->sin_addr.s_addr; | |
336 | } | |
337 | ||
338 | static int put_rtax(struct nlattr *mx, int len, int type, u32 value) | |
339 | { | |
340 | struct nlattr *nla; | |
341 | ||
342 | nla = (struct nlattr *) ((char *) mx + len); | |
343 | nla->nla_type = type; | |
344 | nla->nla_len = nla_attr_size(4); | |
345 | *(u32 *) nla_data(nla) = value; | |
346 | ||
347 | return len + nla_total_size(4); | |
348 | } | |
349 | ||
4b5d47d4 | 350 | static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt, |
4e902c57 TG |
351 | struct fib_config *cfg) |
352 | { | |
6d85c10a | 353 | __be32 addr; |
4e902c57 TG |
354 | int plen; |
355 | ||
356 | memset(cfg, 0, sizeof(*cfg)); | |
4b5d47d4 | 357 | cfg->fc_nlinfo.nl_net = net; |
4e902c57 TG |
358 | |
359 | if (rt->rt_dst.sa_family != AF_INET) | |
360 | return -EAFNOSUPPORT; | |
361 | ||
362 | /* | |
363 | * Check mask for validity: | |
364 | * a) it must be contiguous. | |
365 | * b) destination must have all host bits clear. | |
366 | * c) if application forgot to set correct family (AF_INET), | |
367 | * reject request unless it is absolutely clear i.e. | |
368 | * both family and mask are zero. | |
369 | */ | |
370 | plen = 32; | |
371 | addr = sk_extract_addr(&rt->rt_dst); | |
372 | if (!(rt->rt_flags & RTF_HOST)) { | |
81f7bf6c | 373 | __be32 mask = sk_extract_addr(&rt->rt_genmask); |
4e902c57 TG |
374 | |
375 | if (rt->rt_genmask.sa_family != AF_INET) { | |
376 | if (mask || rt->rt_genmask.sa_family) | |
377 | return -EAFNOSUPPORT; | |
378 | } | |
379 | ||
380 | if (bad_mask(mask, addr)) | |
381 | return -EINVAL; | |
382 | ||
383 | plen = inet_mask_len(mask); | |
384 | } | |
385 | ||
386 | cfg->fc_dst_len = plen; | |
387 | cfg->fc_dst = addr; | |
388 | ||
389 | if (cmd != SIOCDELRT) { | |
390 | cfg->fc_nlflags = NLM_F_CREATE; | |
391 | cfg->fc_protocol = RTPROT_BOOT; | |
392 | } | |
393 | ||
394 | if (rt->rt_metric) | |
395 | cfg->fc_priority = rt->rt_metric - 1; | |
396 | ||
397 | if (rt->rt_flags & RTF_REJECT) { | |
398 | cfg->fc_scope = RT_SCOPE_HOST; | |
399 | cfg->fc_type = RTN_UNREACHABLE; | |
400 | return 0; | |
401 | } | |
402 | ||
403 | cfg->fc_scope = RT_SCOPE_NOWHERE; | |
404 | cfg->fc_type = RTN_UNICAST; | |
405 | ||
406 | if (rt->rt_dev) { | |
407 | char *colon; | |
408 | struct net_device *dev; | |
409 | char devname[IFNAMSIZ]; | |
410 | ||
411 | if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1)) | |
412 | return -EFAULT; | |
413 | ||
414 | devname[IFNAMSIZ-1] = 0; | |
415 | colon = strchr(devname, ':'); | |
416 | if (colon) | |
417 | *colon = 0; | |
4b5d47d4 | 418 | dev = __dev_get_by_name(net, devname); |
4e902c57 TG |
419 | if (!dev) |
420 | return -ENODEV; | |
421 | cfg->fc_oif = dev->ifindex; | |
422 | if (colon) { | |
423 | struct in_ifaddr *ifa; | |
424 | struct in_device *in_dev = __in_dev_get_rtnl(dev); | |
425 | if (!in_dev) | |
426 | return -ENODEV; | |
427 | *colon = ':'; | |
428 | for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) | |
429 | if (strcmp(ifa->ifa_label, devname) == 0) | |
430 | break; | |
431 | if (ifa == NULL) | |
432 | return -ENODEV; | |
433 | cfg->fc_prefsrc = ifa->ifa_local; | |
434 | } | |
435 | } | |
436 | ||
437 | addr = sk_extract_addr(&rt->rt_gateway); | |
438 | if (rt->rt_gateway.sa_family == AF_INET && addr) { | |
439 | cfg->fc_gw = addr; | |
440 | if (rt->rt_flags & RTF_GATEWAY && | |
4b5d47d4 | 441 | inet_addr_type(net, addr) == RTN_UNICAST) |
4e902c57 TG |
442 | cfg->fc_scope = RT_SCOPE_UNIVERSE; |
443 | } | |
444 | ||
445 | if (cmd == SIOCDELRT) | |
446 | return 0; | |
447 | ||
448 | if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw) | |
449 | return -EINVAL; | |
450 | ||
451 | if (cfg->fc_scope == RT_SCOPE_NOWHERE) | |
452 | cfg->fc_scope = RT_SCOPE_LINK; | |
453 | ||
454 | if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) { | |
455 | struct nlattr *mx; | |
456 | int len = 0; | |
457 | ||
458 | mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL); | |
e905a9ed | 459 | if (mx == NULL) |
4e902c57 TG |
460 | return -ENOMEM; |
461 | ||
462 | if (rt->rt_flags & RTF_MTU) | |
463 | len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40); | |
464 | ||
465 | if (rt->rt_flags & RTF_WINDOW) | |
466 | len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window); | |
467 | ||
468 | if (rt->rt_flags & RTF_IRTT) | |
469 | len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3); | |
470 | ||
471 | cfg->fc_mx = mx; | |
472 | cfg->fc_mx_len = len; | |
473 | } | |
474 | ||
475 | return 0; | |
476 | } | |
477 | ||
1da177e4 | 478 | /* |
6a31d2a9 ED |
479 | * Handle IP routing ioctl calls. |
480 | * These are used to manipulate the routing tables | |
1da177e4 | 481 | */ |
1bad118a | 482 | int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg) |
1da177e4 | 483 | { |
4e902c57 TG |
484 | struct fib_config cfg; |
485 | struct rtentry rt; | |
1da177e4 | 486 | int err; |
1da177e4 LT |
487 | |
488 | switch (cmd) { | |
489 | case SIOCADDRT: /* Add a route */ | |
490 | case SIOCDELRT: /* Delete a route */ | |
52e804c6 | 491 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) |
1da177e4 | 492 | return -EPERM; |
4e902c57 TG |
493 | |
494 | if (copy_from_user(&rt, arg, sizeof(rt))) | |
1da177e4 | 495 | return -EFAULT; |
4e902c57 | 496 | |
1da177e4 | 497 | rtnl_lock(); |
1bad118a | 498 | err = rtentry_to_fib_config(net, cmd, &rt, &cfg); |
1da177e4 | 499 | if (err == 0) { |
4e902c57 TG |
500 | struct fib_table *tb; |
501 | ||
1da177e4 | 502 | if (cmd == SIOCDELRT) { |
1bad118a | 503 | tb = fib_get_table(net, cfg.fc_table); |
1da177e4 | 504 | if (tb) |
16c6cf8b | 505 | err = fib_table_delete(tb, &cfg); |
4e902c57 TG |
506 | else |
507 | err = -ESRCH; | |
1da177e4 | 508 | } else { |
1bad118a | 509 | tb = fib_new_table(net, cfg.fc_table); |
1da177e4 | 510 | if (tb) |
16c6cf8b | 511 | err = fib_table_insert(tb, &cfg); |
4e902c57 TG |
512 | else |
513 | err = -ENOBUFS; | |
1da177e4 | 514 | } |
4e902c57 TG |
515 | |
516 | /* allocated by rtentry_to_fib_config() */ | |
517 | kfree(cfg.fc_mx); | |
1da177e4 LT |
518 | } |
519 | rtnl_unlock(); | |
520 | return err; | |
521 | } | |
522 | return -EINVAL; | |
523 | } | |
524 | ||
6a31d2a9 | 525 | const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = { |
4e902c57 TG |
526 | [RTA_DST] = { .type = NLA_U32 }, |
527 | [RTA_SRC] = { .type = NLA_U32 }, | |
528 | [RTA_IIF] = { .type = NLA_U32 }, | |
529 | [RTA_OIF] = { .type = NLA_U32 }, | |
530 | [RTA_GATEWAY] = { .type = NLA_U32 }, | |
531 | [RTA_PRIORITY] = { .type = NLA_U32 }, | |
532 | [RTA_PREFSRC] = { .type = NLA_U32 }, | |
533 | [RTA_METRICS] = { .type = NLA_NESTED }, | |
5176f91e | 534 | [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) }, |
4e902c57 | 535 | [RTA_FLOW] = { .type = NLA_U32 }, |
4e902c57 TG |
536 | }; |
537 | ||
4b5d47d4 | 538 | static int rtm_to_fib_config(struct net *net, struct sk_buff *skb, |
6a31d2a9 | 539 | struct nlmsghdr *nlh, struct fib_config *cfg) |
1da177e4 | 540 | { |
4e902c57 TG |
541 | struct nlattr *attr; |
542 | int err, remaining; | |
543 | struct rtmsg *rtm; | |
544 | ||
545 | err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy); | |
546 | if (err < 0) | |
547 | goto errout; | |
548 | ||
549 | memset(cfg, 0, sizeof(*cfg)); | |
550 | ||
551 | rtm = nlmsg_data(nlh); | |
4e902c57 | 552 | cfg->fc_dst_len = rtm->rtm_dst_len; |
4e902c57 TG |
553 | cfg->fc_tos = rtm->rtm_tos; |
554 | cfg->fc_table = rtm->rtm_table; | |
555 | cfg->fc_protocol = rtm->rtm_protocol; | |
556 | cfg->fc_scope = rtm->rtm_scope; | |
557 | cfg->fc_type = rtm->rtm_type; | |
558 | cfg->fc_flags = rtm->rtm_flags; | |
559 | cfg->fc_nlflags = nlh->nlmsg_flags; | |
560 | ||
15e47304 | 561 | cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid; |
4e902c57 | 562 | cfg->fc_nlinfo.nlh = nlh; |
4b5d47d4 | 563 | cfg->fc_nlinfo.nl_net = net; |
4e902c57 | 564 | |
a0ee18b9 TG |
565 | if (cfg->fc_type > RTN_MAX) { |
566 | err = -EINVAL; | |
567 | goto errout; | |
568 | } | |
569 | ||
4e902c57 | 570 | nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) { |
8f4c1f9b | 571 | switch (nla_type(attr)) { |
4e902c57 | 572 | case RTA_DST: |
17fb2c64 | 573 | cfg->fc_dst = nla_get_be32(attr); |
4e902c57 | 574 | break; |
4e902c57 TG |
575 | case RTA_OIF: |
576 | cfg->fc_oif = nla_get_u32(attr); | |
577 | break; | |
578 | case RTA_GATEWAY: | |
17fb2c64 | 579 | cfg->fc_gw = nla_get_be32(attr); |
4e902c57 TG |
580 | break; |
581 | case RTA_PRIORITY: | |
582 | cfg->fc_priority = nla_get_u32(attr); | |
583 | break; | |
584 | case RTA_PREFSRC: | |
17fb2c64 | 585 | cfg->fc_prefsrc = nla_get_be32(attr); |
4e902c57 TG |
586 | break; |
587 | case RTA_METRICS: | |
588 | cfg->fc_mx = nla_data(attr); | |
589 | cfg->fc_mx_len = nla_len(attr); | |
590 | break; | |
591 | case RTA_MULTIPATH: | |
592 | cfg->fc_mp = nla_data(attr); | |
593 | cfg->fc_mp_len = nla_len(attr); | |
594 | break; | |
595 | case RTA_FLOW: | |
596 | cfg->fc_flow = nla_get_u32(attr); | |
597 | break; | |
4e902c57 TG |
598 | case RTA_TABLE: |
599 | cfg->fc_table = nla_get_u32(attr); | |
600 | break; | |
1da177e4 LT |
601 | } |
602 | } | |
4e902c57 | 603 | |
1da177e4 | 604 | return 0; |
4e902c57 TG |
605 | errout: |
606 | return err; | |
1da177e4 LT |
607 | } |
608 | ||
661d2967 | 609 | static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh) |
1da177e4 | 610 | { |
3b1e0a65 | 611 | struct net *net = sock_net(skb->sk); |
4e902c57 TG |
612 | struct fib_config cfg; |
613 | struct fib_table *tb; | |
614 | int err; | |
1da177e4 | 615 | |
4b5d47d4 | 616 | err = rtm_to_fib_config(net, skb, nlh, &cfg); |
4e902c57 TG |
617 | if (err < 0) |
618 | goto errout; | |
1da177e4 | 619 | |
8ad4942c | 620 | tb = fib_get_table(net, cfg.fc_table); |
4e902c57 TG |
621 | if (tb == NULL) { |
622 | err = -ESRCH; | |
623 | goto errout; | |
624 | } | |
625 | ||
16c6cf8b | 626 | err = fib_table_delete(tb, &cfg); |
4e902c57 TG |
627 | errout: |
628 | return err; | |
1da177e4 LT |
629 | } |
630 | ||
661d2967 | 631 | static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh) |
1da177e4 | 632 | { |
3b1e0a65 | 633 | struct net *net = sock_net(skb->sk); |
4e902c57 TG |
634 | struct fib_config cfg; |
635 | struct fib_table *tb; | |
636 | int err; | |
1da177e4 | 637 | |
4b5d47d4 | 638 | err = rtm_to_fib_config(net, skb, nlh, &cfg); |
4e902c57 TG |
639 | if (err < 0) |
640 | goto errout; | |
1da177e4 | 641 | |
226b0b4a | 642 | tb = fib_new_table(net, cfg.fc_table); |
4e902c57 TG |
643 | if (tb == NULL) { |
644 | err = -ENOBUFS; | |
645 | goto errout; | |
646 | } | |
647 | ||
16c6cf8b | 648 | err = fib_table_insert(tb, &cfg); |
4e902c57 TG |
649 | errout: |
650 | return err; | |
1da177e4 LT |
651 | } |
652 | ||
63f3444f | 653 | static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb) |
1da177e4 | 654 | { |
3b1e0a65 | 655 | struct net *net = sock_net(skb->sk); |
1af5a8c4 PM |
656 | unsigned int h, s_h; |
657 | unsigned int e = 0, s_e; | |
1da177e4 | 658 | struct fib_table *tb; |
e4aef8ae | 659 | struct hlist_head *head; |
1af5a8c4 | 660 | int dumped = 0; |
1da177e4 | 661 | |
be403ea1 TG |
662 | if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) && |
663 | ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED) | |
0b8c7f6f | 664 | return skb->len; |
1da177e4 | 665 | |
1af5a8c4 PM |
666 | s_h = cb->args[0]; |
667 | s_e = cb->args[1]; | |
668 | ||
669 | for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) { | |
670 | e = 0; | |
e4aef8ae | 671 | head = &net->ipv4.fib_table_hash[h]; |
b67bfe0d | 672 | hlist_for_each_entry(tb, head, tb_hlist) { |
1af5a8c4 PM |
673 | if (e < s_e) |
674 | goto next; | |
675 | if (dumped) | |
676 | memset(&cb->args[2], 0, sizeof(cb->args) - | |
e905a9ed | 677 | 2 * sizeof(cb->args[0])); |
16c6cf8b | 678 | if (fib_table_dump(tb, skb, cb) < 0) |
1af5a8c4 PM |
679 | goto out; |
680 | dumped = 1; | |
681 | next: | |
682 | e++; | |
683 | } | |
1da177e4 | 684 | } |
1af5a8c4 PM |
685 | out: |
686 | cb->args[1] = e; | |
687 | cb->args[0] = h; | |
1da177e4 LT |
688 | |
689 | return skb->len; | |
690 | } | |
691 | ||
692 | /* Prepare and feed intra-kernel routing request. | |
6a31d2a9 ED |
693 | * Really, it should be netlink message, but :-( netlink |
694 | * can be not configured, so that we feed it directly | |
695 | * to fib engine. It is legal, because all events occur | |
696 | * only when netlink is already locked. | |
1da177e4 | 697 | */ |
81f7bf6c | 698 | static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa) |
1da177e4 | 699 | { |
c346dca1 | 700 | struct net *net = dev_net(ifa->ifa_dev->dev); |
4e902c57 TG |
701 | struct fib_table *tb; |
702 | struct fib_config cfg = { | |
703 | .fc_protocol = RTPROT_KERNEL, | |
704 | .fc_type = type, | |
705 | .fc_dst = dst, | |
706 | .fc_dst_len = dst_len, | |
707 | .fc_prefsrc = ifa->ifa_local, | |
708 | .fc_oif = ifa->ifa_dev->dev->ifindex, | |
709 | .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND, | |
4d1169c1 | 710 | .fc_nlinfo = { |
4b5d47d4 | 711 | .nl_net = net, |
4d1169c1 | 712 | }, |
4e902c57 | 713 | }; |
1da177e4 LT |
714 | |
715 | if (type == RTN_UNICAST) | |
4b5d47d4 | 716 | tb = fib_new_table(net, RT_TABLE_MAIN); |
1da177e4 | 717 | else |
4b5d47d4 | 718 | tb = fib_new_table(net, RT_TABLE_LOCAL); |
1da177e4 LT |
719 | |
720 | if (tb == NULL) | |
721 | return; | |
722 | ||
4e902c57 | 723 | cfg.fc_table = tb->tb_id; |
1da177e4 | 724 | |
4e902c57 TG |
725 | if (type != RTN_LOCAL) |
726 | cfg.fc_scope = RT_SCOPE_LINK; | |
727 | else | |
728 | cfg.fc_scope = RT_SCOPE_HOST; | |
1da177e4 LT |
729 | |
730 | if (cmd == RTM_NEWROUTE) | |
16c6cf8b | 731 | fib_table_insert(tb, &cfg); |
1da177e4 | 732 | else |
16c6cf8b | 733 | fib_table_delete(tb, &cfg); |
1da177e4 LT |
734 | } |
735 | ||
0ff60a45 | 736 | void fib_add_ifaddr(struct in_ifaddr *ifa) |
1da177e4 LT |
737 | { |
738 | struct in_device *in_dev = ifa->ifa_dev; | |
739 | struct net_device *dev = in_dev->dev; | |
740 | struct in_ifaddr *prim = ifa; | |
a144ea4b AV |
741 | __be32 mask = ifa->ifa_mask; |
742 | __be32 addr = ifa->ifa_local; | |
6a31d2a9 | 743 | __be32 prefix = ifa->ifa_address & mask; |
1da177e4 | 744 | |
6a31d2a9 | 745 | if (ifa->ifa_flags & IFA_F_SECONDARY) { |
1da177e4 LT |
746 | prim = inet_ifa_byprefix(in_dev, prefix, mask); |
747 | if (prim == NULL) { | |
058bd4d2 | 748 | pr_warn("%s: bug: prim == NULL\n", __func__); |
1da177e4 LT |
749 | return; |
750 | } | |
751 | } | |
752 | ||
753 | fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim); | |
754 | ||
6a31d2a9 | 755 | if (!(dev->flags & IFF_UP)) |
1da177e4 LT |
756 | return; |
757 | ||
758 | /* Add broadcast address, if it is explicitly assigned. */ | |
a144ea4b | 759 | if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF)) |
1da177e4 LT |
760 | fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim); |
761 | ||
6a31d2a9 | 762 | if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) && |
1da177e4 | 763 | (prefix != addr || ifa->ifa_prefixlen < 32)) { |
6a31d2a9 ED |
764 | fib_magic(RTM_NEWROUTE, |
765 | dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST, | |
766 | prefix, ifa->ifa_prefixlen, prim); | |
1da177e4 LT |
767 | |
768 | /* Add network specific broadcasts, when it takes a sense */ | |
769 | if (ifa->ifa_prefixlen < 31) { | |
770 | fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim); | |
6a31d2a9 ED |
771 | fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask, |
772 | 32, prim); | |
1da177e4 LT |
773 | } |
774 | } | |
775 | } | |
776 | ||
e6abbaa2 JA |
777 | /* Delete primary or secondary address. |
778 | * Optionally, on secondary address promotion consider the addresses | |
779 | * from subnet iprim as deleted, even if they are in device list. | |
780 | * In this case the secondary ifa can be in device list. | |
781 | */ | |
782 | void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim) | |
1da177e4 LT |
783 | { |
784 | struct in_device *in_dev = ifa->ifa_dev; | |
785 | struct net_device *dev = in_dev->dev; | |
786 | struct in_ifaddr *ifa1; | |
e6abbaa2 | 787 | struct in_ifaddr *prim = ifa, *prim1 = NULL; |
6a31d2a9 ED |
788 | __be32 brd = ifa->ifa_address | ~ifa->ifa_mask; |
789 | __be32 any = ifa->ifa_address & ifa->ifa_mask; | |
1da177e4 LT |
790 | #define LOCAL_OK 1 |
791 | #define BRD_OK 2 | |
792 | #define BRD0_OK 4 | |
793 | #define BRD1_OK 8 | |
95c96174 | 794 | unsigned int ok = 0; |
e6abbaa2 JA |
795 | int subnet = 0; /* Primary network */ |
796 | int gone = 1; /* Address is missing */ | |
797 | int same_prefsrc = 0; /* Another primary with same IP */ | |
1da177e4 | 798 | |
e6abbaa2 | 799 | if (ifa->ifa_flags & IFA_F_SECONDARY) { |
1da177e4 LT |
800 | prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask); |
801 | if (prim == NULL) { | |
058bd4d2 | 802 | pr_warn("%s: bug: prim == NULL\n", __func__); |
1da177e4 LT |
803 | return; |
804 | } | |
e6abbaa2 | 805 | if (iprim && iprim != prim) { |
058bd4d2 | 806 | pr_warn("%s: bug: iprim != prim\n", __func__); |
e6abbaa2 JA |
807 | return; |
808 | } | |
809 | } else if (!ipv4_is_zeronet(any) && | |
810 | (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) { | |
811 | fib_magic(RTM_DELROUTE, | |
812 | dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST, | |
813 | any, ifa->ifa_prefixlen, prim); | |
814 | subnet = 1; | |
1da177e4 LT |
815 | } |
816 | ||
817 | /* Deletion is more complicated than add. | |
6a31d2a9 ED |
818 | * We should take care of not to delete too much :-) |
819 | * | |
820 | * Scan address list to be sure that addresses are really gone. | |
1da177e4 LT |
821 | */ |
822 | ||
823 | for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) { | |
e6abbaa2 JA |
824 | if (ifa1 == ifa) { |
825 | /* promotion, keep the IP */ | |
826 | gone = 0; | |
827 | continue; | |
828 | } | |
829 | /* Ignore IFAs from our subnet */ | |
830 | if (iprim && ifa1->ifa_mask == iprim->ifa_mask && | |
831 | inet_ifa_match(ifa1->ifa_address, iprim)) | |
832 | continue; | |
833 | ||
834 | /* Ignore ifa1 if it uses different primary IP (prefsrc) */ | |
835 | if (ifa1->ifa_flags & IFA_F_SECONDARY) { | |
836 | /* Another address from our subnet? */ | |
837 | if (ifa1->ifa_mask == prim->ifa_mask && | |
838 | inet_ifa_match(ifa1->ifa_address, prim)) | |
839 | prim1 = prim; | |
840 | else { | |
841 | /* We reached the secondaries, so | |
842 | * same_prefsrc should be determined. | |
843 | */ | |
844 | if (!same_prefsrc) | |
845 | continue; | |
846 | /* Search new prim1 if ifa1 is not | |
847 | * using the current prim1 | |
848 | */ | |
849 | if (!prim1 || | |
850 | ifa1->ifa_mask != prim1->ifa_mask || | |
851 | !inet_ifa_match(ifa1->ifa_address, prim1)) | |
852 | prim1 = inet_ifa_byprefix(in_dev, | |
853 | ifa1->ifa_address, | |
854 | ifa1->ifa_mask); | |
855 | if (!prim1) | |
856 | continue; | |
857 | if (prim1->ifa_local != prim->ifa_local) | |
858 | continue; | |
859 | } | |
860 | } else { | |
861 | if (prim->ifa_local != ifa1->ifa_local) | |
862 | continue; | |
863 | prim1 = ifa1; | |
864 | if (prim != prim1) | |
865 | same_prefsrc = 1; | |
866 | } | |
1da177e4 LT |
867 | if (ifa->ifa_local == ifa1->ifa_local) |
868 | ok |= LOCAL_OK; | |
869 | if (ifa->ifa_broadcast == ifa1->ifa_broadcast) | |
870 | ok |= BRD_OK; | |
871 | if (brd == ifa1->ifa_broadcast) | |
872 | ok |= BRD1_OK; | |
873 | if (any == ifa1->ifa_broadcast) | |
874 | ok |= BRD0_OK; | |
e6abbaa2 JA |
875 | /* primary has network specific broadcasts */ |
876 | if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) { | |
877 | __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask; | |
878 | __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask; | |
879 | ||
880 | if (!ipv4_is_zeronet(any1)) { | |
881 | if (ifa->ifa_broadcast == brd1 || | |
882 | ifa->ifa_broadcast == any1) | |
883 | ok |= BRD_OK; | |
884 | if (brd == brd1 || brd == any1) | |
885 | ok |= BRD1_OK; | |
886 | if (any == brd1 || any == any1) | |
887 | ok |= BRD0_OK; | |
888 | } | |
889 | } | |
1da177e4 LT |
890 | } |
891 | ||
6a31d2a9 | 892 | if (!(ok & BRD_OK)) |
1da177e4 | 893 | fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim); |
e6abbaa2 JA |
894 | if (subnet && ifa->ifa_prefixlen < 31) { |
895 | if (!(ok & BRD1_OK)) | |
896 | fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim); | |
897 | if (!(ok & BRD0_OK)) | |
898 | fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim); | |
899 | } | |
6a31d2a9 | 900 | if (!(ok & LOCAL_OK)) { |
1da177e4 LT |
901 | fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim); |
902 | ||
903 | /* Check, that this local address finally disappeared. */ | |
e6abbaa2 JA |
904 | if (gone && |
905 | inet_addr_type(dev_net(dev), ifa->ifa_local) != RTN_LOCAL) { | |
1da177e4 | 906 | /* And the last, but not the least thing. |
6a31d2a9 ED |
907 | * We must flush stray FIB entries. |
908 | * | |
909 | * First of all, we scan fib_info list searching | |
910 | * for stray nexthop entries, then ignite fib_flush. | |
911 | */ | |
c346dca1 YH |
912 | if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local)) |
913 | fib_flush(dev_net(dev)); | |
1da177e4 LT |
914 | } |
915 | } | |
916 | #undef LOCAL_OK | |
917 | #undef BRD_OK | |
918 | #undef BRD0_OK | |
919 | #undef BRD1_OK | |
920 | } | |
921 | ||
6a31d2a9 | 922 | static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb) |
246955fe | 923 | { |
e905a9ed | 924 | |
246955fe | 925 | struct fib_result res; |
9ade2286 DM |
926 | struct flowi4 fl4 = { |
927 | .flowi4_mark = frn->fl_mark, | |
928 | .daddr = frn->fl_addr, | |
929 | .flowi4_tos = frn->fl_tos, | |
930 | .flowi4_scope = frn->fl_scope, | |
6a31d2a9 | 931 | }; |
1194ed0a AK |
932 | |
933 | frn->err = -ENOENT; | |
246955fe RO |
934 | if (tb) { |
935 | local_bh_disable(); | |
936 | ||
937 | frn->tb_id = tb->tb_id; | |
9ade2286 | 938 | frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF); |
246955fe RO |
939 | |
940 | if (!frn->err) { | |
941 | frn->prefixlen = res.prefixlen; | |
942 | frn->nh_sel = res.nh_sel; | |
943 | frn->type = res.type; | |
944 | frn->scope = res.scope; | |
945 | } | |
946 | local_bh_enable(); | |
947 | } | |
948 | } | |
949 | ||
28f7b036 | 950 | static void nl_fib_input(struct sk_buff *skb) |
246955fe | 951 | { |
6bd48fcf | 952 | struct net *net; |
246955fe | 953 | struct fib_result_nl *frn; |
28f7b036 | 954 | struct nlmsghdr *nlh; |
246955fe | 955 | struct fib_table *tb; |
15e47304 | 956 | u32 portid; |
1194ed0a | 957 | |
3b1e0a65 | 958 | net = sock_net(skb->sk); |
b529ccf2 | 959 | nlh = nlmsg_hdr(skb); |
573ce260 H |
960 | if (skb->len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len || |
961 | nlmsg_len(nlh) < sizeof(*frn)) | |
ea86575e | 962 | return; |
d883a036 | 963 | |
3a36515f | 964 | skb = netlink_skb_clone(skb, GFP_KERNEL); |
d883a036 DL |
965 | if (skb == NULL) |
966 | return; | |
967 | nlh = nlmsg_hdr(skb); | |
e905a9ed | 968 | |
573ce260 | 969 | frn = (struct fib_result_nl *) nlmsg_data(nlh); |
6bd48fcf | 970 | tb = fib_get_table(net, frn->tb_id_in); |
246955fe RO |
971 | |
972 | nl_fib_lookup(frn, tb); | |
e905a9ed | 973 | |
28a28283 | 974 | portid = NETLINK_CB(skb).portid; /* netlink portid */ |
15e47304 | 975 | NETLINK_CB(skb).portid = 0; /* from kernel */ |
ac6d439d | 976 | NETLINK_CB(skb).dst_group = 0; /* unicast */ |
15e47304 | 977 | netlink_unicast(net->ipv4.fibnl, skb, portid, MSG_DONTWAIT); |
e905a9ed | 978 | } |
246955fe | 979 | |
2c8c1e72 | 980 | static int __net_init nl_fib_lookup_init(struct net *net) |
246955fe | 981 | { |
6bd48fcf | 982 | struct sock *sk; |
a31f2d17 PNA |
983 | struct netlink_kernel_cfg cfg = { |
984 | .input = nl_fib_input, | |
985 | }; | |
986 | ||
9f00d977 | 987 | sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, &cfg); |
6bd48fcf | 988 | if (sk == NULL) |
7b1a74fd | 989 | return -EAFNOSUPPORT; |
6bd48fcf | 990 | net->ipv4.fibnl = sk; |
7b1a74fd DL |
991 | return 0; |
992 | } | |
993 | ||
994 | static void nl_fib_lookup_exit(struct net *net) | |
995 | { | |
b7c6ba6e | 996 | netlink_kernel_release(net->ipv4.fibnl); |
775516bf | 997 | net->ipv4.fibnl = NULL; |
246955fe RO |
998 | } |
999 | ||
4ccfe6d4 | 1000 | static void fib_disable_ip(struct net_device *dev, int force) |
1da177e4 | 1001 | { |
85326fa5 | 1002 | if (fib_sync_down_dev(dev, force)) |
c346dca1 | 1003 | fib_flush(dev_net(dev)); |
4ccfe6d4 | 1004 | rt_cache_flush(dev_net(dev)); |
1da177e4 LT |
1005 | arp_ifdown(dev); |
1006 | } | |
1007 | ||
1008 | static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr) | |
1009 | { | |
6ed2533e | 1010 | struct in_ifaddr *ifa = (struct in_ifaddr *)ptr; |
76e6ebfb | 1011 | struct net_device *dev = ifa->ifa_dev->dev; |
436c3b66 | 1012 | struct net *net = dev_net(dev); |
1da177e4 LT |
1013 | |
1014 | switch (event) { | |
1015 | case NETDEV_UP: | |
1016 | fib_add_ifaddr(ifa); | |
1017 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
76e6ebfb | 1018 | fib_sync_up(dev); |
1da177e4 | 1019 | #endif |
436c3b66 | 1020 | atomic_inc(&net->ipv4.dev_addr_genid); |
4ccfe6d4 | 1021 | rt_cache_flush(dev_net(dev)); |
1da177e4 LT |
1022 | break; |
1023 | case NETDEV_DOWN: | |
e6abbaa2 | 1024 | fib_del_ifaddr(ifa, NULL); |
436c3b66 | 1025 | atomic_inc(&net->ipv4.dev_addr_genid); |
9fcc2e8a | 1026 | if (ifa->ifa_dev->ifa_list == NULL) { |
1da177e4 | 1027 | /* Last address was deleted from this interface. |
6a31d2a9 | 1028 | * Disable IP. |
1da177e4 | 1029 | */ |
4ccfe6d4 | 1030 | fib_disable_ip(dev, 1); |
1da177e4 | 1031 | } else { |
4ccfe6d4 | 1032 | rt_cache_flush(dev_net(dev)); |
1da177e4 LT |
1033 | } |
1034 | break; | |
1035 | } | |
1036 | return NOTIFY_DONE; | |
1037 | } | |
1038 | ||
1039 | static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr) | |
1040 | { | |
351638e7 | 1041 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); |
0115e8e3 | 1042 | struct in_device *in_dev; |
436c3b66 | 1043 | struct net *net = dev_net(dev); |
1da177e4 LT |
1044 | |
1045 | if (event == NETDEV_UNREGISTER) { | |
4ccfe6d4 | 1046 | fib_disable_ip(dev, 2); |
caacf05e | 1047 | rt_flush_dev(dev); |
1da177e4 LT |
1048 | return NOTIFY_DONE; |
1049 | } | |
1050 | ||
0115e8e3 | 1051 | in_dev = __in_dev_get_rtnl(dev); |
a0065f26 OH |
1052 | if (!in_dev) |
1053 | return NOTIFY_DONE; | |
0115e8e3 | 1054 | |
1da177e4 LT |
1055 | switch (event) { |
1056 | case NETDEV_UP: | |
1057 | for_ifa(in_dev) { | |
1058 | fib_add_ifaddr(ifa); | |
1059 | } endfor_ifa(in_dev); | |
1060 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
1061 | fib_sync_up(dev); | |
1062 | #endif | |
436c3b66 | 1063 | atomic_inc(&net->ipv4.dev_addr_genid); |
4ccfe6d4 | 1064 | rt_cache_flush(net); |
1da177e4 LT |
1065 | break; |
1066 | case NETDEV_DOWN: | |
4ccfe6d4 | 1067 | fib_disable_ip(dev, 0); |
1da177e4 LT |
1068 | break; |
1069 | case NETDEV_CHANGEMTU: | |
1070 | case NETDEV_CHANGE: | |
4ccfe6d4 | 1071 | rt_cache_flush(net); |
1da177e4 LT |
1072 | break; |
1073 | } | |
1074 | return NOTIFY_DONE; | |
1075 | } | |
1076 | ||
1077 | static struct notifier_block fib_inetaddr_notifier = { | |
6ed2533e | 1078 | .notifier_call = fib_inetaddr_event, |
1da177e4 LT |
1079 | }; |
1080 | ||
1081 | static struct notifier_block fib_netdev_notifier = { | |
6ed2533e | 1082 | .notifier_call = fib_netdev_event, |
1da177e4 LT |
1083 | }; |
1084 | ||
7b1a74fd | 1085 | static int __net_init ip_fib_net_init(struct net *net) |
1da177e4 | 1086 | { |
dce5cbee | 1087 | int err; |
10da66f7 ED |
1088 | size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ; |
1089 | ||
1090 | /* Avoid false sharing : Use at least a full cache line */ | |
1091 | size = max_t(size_t, size, L1_CACHE_BYTES); | |
1af5a8c4 | 1092 | |
10da66f7 | 1093 | net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL); |
e4aef8ae DL |
1094 | if (net->ipv4.fib_table_hash == NULL) |
1095 | return -ENOMEM; | |
1096 | ||
dce5cbee DL |
1097 | err = fib4_rules_init(net); |
1098 | if (err < 0) | |
1099 | goto fail; | |
1100 | return 0; | |
1101 | ||
1102 | fail: | |
1103 | kfree(net->ipv4.fib_table_hash); | |
1104 | return err; | |
7b1a74fd | 1105 | } |
1da177e4 | 1106 | |
2c8c1e72 | 1107 | static void ip_fib_net_exit(struct net *net) |
7b1a74fd DL |
1108 | { |
1109 | unsigned int i; | |
1110 | ||
1111 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
1112 | fib4_rules_exit(net); | |
1113 | #endif | |
1114 | ||
e2666f84 | 1115 | rtnl_lock(); |
7b1a74fd DL |
1116 | for (i = 0; i < FIB_TABLE_HASHSZ; i++) { |
1117 | struct fib_table *tb; | |
1118 | struct hlist_head *head; | |
b67bfe0d | 1119 | struct hlist_node *tmp; |
63f3444f | 1120 | |
e4aef8ae | 1121 | head = &net->ipv4.fib_table_hash[i]; |
b67bfe0d SL |
1122 | hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) { |
1123 | hlist_del(&tb->tb_hlist); | |
16c6cf8b | 1124 | fib_table_flush(tb); |
4aa2c466 | 1125 | fib_free_table(tb); |
7b1a74fd DL |
1126 | } |
1127 | } | |
e2666f84 | 1128 | rtnl_unlock(); |
e4aef8ae | 1129 | kfree(net->ipv4.fib_table_hash); |
7b1a74fd DL |
1130 | } |
1131 | ||
1132 | static int __net_init fib_net_init(struct net *net) | |
1133 | { | |
1134 | int error; | |
1135 | ||
f4530fa5 DM |
1136 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1137 | net->ipv4.fib_num_tclassid_users = 0; | |
1138 | #endif | |
7b1a74fd DL |
1139 | error = ip_fib_net_init(net); |
1140 | if (error < 0) | |
1141 | goto out; | |
1142 | error = nl_fib_lookup_init(net); | |
1143 | if (error < 0) | |
1144 | goto out_nlfl; | |
1145 | error = fib_proc_init(net); | |
1146 | if (error < 0) | |
1147 | goto out_proc; | |
1148 | out: | |
1149 | return error; | |
1150 | ||
1151 | out_proc: | |
1152 | nl_fib_lookup_exit(net); | |
1153 | out_nlfl: | |
1154 | ip_fib_net_exit(net); | |
1155 | goto out; | |
1156 | } | |
1157 | ||
1158 | static void __net_exit fib_net_exit(struct net *net) | |
1159 | { | |
1160 | fib_proc_exit(net); | |
1161 | nl_fib_lookup_exit(net); | |
1162 | ip_fib_net_exit(net); | |
1163 | } | |
1164 | ||
1165 | static struct pernet_operations fib_net_ops = { | |
1166 | .init = fib_net_init, | |
1167 | .exit = fib_net_exit, | |
1168 | }; | |
1169 | ||
1170 | void __init ip_fib_init(void) | |
1171 | { | |
c7ac8679 GR |
1172 | rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, NULL); |
1173 | rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, NULL); | |
1174 | rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, NULL); | |
7b1a74fd DL |
1175 | |
1176 | register_pernet_subsys(&fib_net_ops); | |
1177 | register_netdevice_notifier(&fib_netdev_notifier); | |
1178 | register_inetaddr_notifier(&fib_inetaddr_notifier); | |
7f9b8052 | 1179 | |
5348ba85 | 1180 | fib_trie_init(); |
1da177e4 | 1181 | } |