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: semantics. | |
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 <asm/uaccess.h> |
17 | #include <asm/system.h> | |
18 | #include <linux/bitops.h> | |
19 | #include <linux/types.h> | |
20 | #include <linux/kernel.h> | |
21 | #include <linux/jiffies.h> | |
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 LT |
30 | #include <linux/netdevice.h> |
31 | #include <linux/if_arp.h> | |
32 | #include <linux/proc_fs.h> | |
33 | #include <linux/skbuff.h> | |
1da177e4 | 34 | #include <linux/init.h> |
5a0e3ad6 | 35 | #include <linux/slab.h> |
1da177e4 | 36 | |
14c85021 | 37 | #include <net/arp.h> |
1da177e4 LT |
38 | #include <net/ip.h> |
39 | #include <net/protocol.h> | |
40 | #include <net/route.h> | |
41 | #include <net/tcp.h> | |
42 | #include <net/sock.h> | |
43 | #include <net/ip_fib.h> | |
f21c7bc5 | 44 | #include <net/netlink.h> |
4e902c57 | 45 | #include <net/nexthop.h> |
1da177e4 LT |
46 | |
47 | #include "fib_lookup.h" | |
48 | ||
832b4c5e | 49 | static DEFINE_SPINLOCK(fib_info_lock); |
1da177e4 LT |
50 | static struct hlist_head *fib_info_hash; |
51 | static struct hlist_head *fib_info_laddrhash; | |
123b9731 | 52 | static unsigned int fib_info_hash_size; |
1da177e4 LT |
53 | static unsigned int fib_info_cnt; |
54 | ||
55 | #define DEVINDEX_HASHBITS 8 | |
56 | #define DEVINDEX_HASHSIZE (1U << DEVINDEX_HASHBITS) | |
57 | static struct hlist_head fib_info_devhash[DEVINDEX_HASHSIZE]; | |
58 | ||
59 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
60 | ||
61 | static DEFINE_SPINLOCK(fib_multipath_lock); | |
62 | ||
6a31d2a9 ED |
63 | #define for_nexthops(fi) { \ |
64 | int nhsel; const struct fib_nh *nh; \ | |
65 | for (nhsel = 0, nh = (fi)->fib_nh; \ | |
66 | nhsel < (fi)->fib_nhs; \ | |
67 | nh++, nhsel++) | |
68 | ||
69 | #define change_nexthops(fi) { \ | |
70 | int nhsel; struct fib_nh *nexthop_nh; \ | |
71 | for (nhsel = 0, nexthop_nh = (struct fib_nh *)((fi)->fib_nh); \ | |
72 | nhsel < (fi)->fib_nhs; \ | |
73 | nexthop_nh++, nhsel++) | |
1da177e4 LT |
74 | |
75 | #else /* CONFIG_IP_ROUTE_MULTIPATH */ | |
76 | ||
77 | /* Hope, that gcc will optimize it to get rid of dummy loop */ | |
78 | ||
6a31d2a9 ED |
79 | #define for_nexthops(fi) { \ |
80 | int nhsel; const struct fib_nh *nh = (fi)->fib_nh; \ | |
81 | for (nhsel = 0; nhsel < 1; nhsel++) | |
1da177e4 | 82 | |
6a31d2a9 ED |
83 | #define change_nexthops(fi) { \ |
84 | int nhsel; \ | |
85 | struct fib_nh *nexthop_nh = (struct fib_nh *)((fi)->fib_nh); \ | |
86 | for (nhsel = 0; nhsel < 1; nhsel++) | |
1da177e4 LT |
87 | |
88 | #endif /* CONFIG_IP_ROUTE_MULTIPATH */ | |
89 | ||
90 | #define endfor_nexthops(fi) } | |
91 | ||
92 | ||
e905a9ed | 93 | static const struct |
1da177e4 LT |
94 | { |
95 | int error; | |
96 | u8 scope; | |
a0ee18b9 | 97 | } fib_props[RTN_MAX + 1] = { |
6a31d2a9 | 98 | [RTN_UNSPEC] = { |
1da177e4 LT |
99 | .error = 0, |
100 | .scope = RT_SCOPE_NOWHERE, | |
6a31d2a9 ED |
101 | }, |
102 | [RTN_UNICAST] = { | |
1da177e4 LT |
103 | .error = 0, |
104 | .scope = RT_SCOPE_UNIVERSE, | |
6a31d2a9 ED |
105 | }, |
106 | [RTN_LOCAL] = { | |
1da177e4 LT |
107 | .error = 0, |
108 | .scope = RT_SCOPE_HOST, | |
6a31d2a9 ED |
109 | }, |
110 | [RTN_BROADCAST] = { | |
1da177e4 LT |
111 | .error = 0, |
112 | .scope = RT_SCOPE_LINK, | |
6a31d2a9 ED |
113 | }, |
114 | [RTN_ANYCAST] = { | |
1da177e4 LT |
115 | .error = 0, |
116 | .scope = RT_SCOPE_LINK, | |
6a31d2a9 ED |
117 | }, |
118 | [RTN_MULTICAST] = { | |
1da177e4 LT |
119 | .error = 0, |
120 | .scope = RT_SCOPE_UNIVERSE, | |
6a31d2a9 ED |
121 | }, |
122 | [RTN_BLACKHOLE] = { | |
1da177e4 LT |
123 | .error = -EINVAL, |
124 | .scope = RT_SCOPE_UNIVERSE, | |
6a31d2a9 ED |
125 | }, |
126 | [RTN_UNREACHABLE] = { | |
1da177e4 LT |
127 | .error = -EHOSTUNREACH, |
128 | .scope = RT_SCOPE_UNIVERSE, | |
6a31d2a9 ED |
129 | }, |
130 | [RTN_PROHIBIT] = { | |
1da177e4 LT |
131 | .error = -EACCES, |
132 | .scope = RT_SCOPE_UNIVERSE, | |
6a31d2a9 ED |
133 | }, |
134 | [RTN_THROW] = { | |
1da177e4 LT |
135 | .error = -EAGAIN, |
136 | .scope = RT_SCOPE_UNIVERSE, | |
6a31d2a9 ED |
137 | }, |
138 | [RTN_NAT] = { | |
1da177e4 LT |
139 | .error = -EINVAL, |
140 | .scope = RT_SCOPE_NOWHERE, | |
6a31d2a9 ED |
141 | }, |
142 | [RTN_XRESOLVE] = { | |
1da177e4 LT |
143 | .error = -EINVAL, |
144 | .scope = RT_SCOPE_NOWHERE, | |
6a31d2a9 | 145 | }, |
1da177e4 LT |
146 | }; |
147 | ||
148 | ||
149 | /* Release a nexthop info record */ | |
150 | ||
ebc0ffae ED |
151 | static void free_fib_info_rcu(struct rcu_head *head) |
152 | { | |
153 | struct fib_info *fi = container_of(head, struct fib_info, rcu); | |
154 | ||
725d1e1b DM |
155 | if (fi->fib_metrics != (u32 *) dst_default_metrics) |
156 | kfree(fi->fib_metrics); | |
ebc0ffae ED |
157 | kfree(fi); |
158 | } | |
159 | ||
1da177e4 LT |
160 | void free_fib_info(struct fib_info *fi) |
161 | { | |
162 | if (fi->fib_dead == 0) { | |
6a31d2a9 | 163 | pr_warning("Freeing alive fib_info %p\n", fi); |
1da177e4 LT |
164 | return; |
165 | } | |
166 | change_nexthops(fi) { | |
71fceff0 DM |
167 | if (nexthop_nh->nh_dev) |
168 | dev_put(nexthop_nh->nh_dev); | |
169 | nexthop_nh->nh_dev = NULL; | |
1da177e4 LT |
170 | } endfor_nexthops(fi); |
171 | fib_info_cnt--; | |
57d7a600 | 172 | release_net(fi->fib_net); |
ebc0ffae | 173 | call_rcu(&fi->rcu, free_fib_info_rcu); |
1da177e4 LT |
174 | } |
175 | ||
176 | void fib_release_info(struct fib_info *fi) | |
177 | { | |
832b4c5e | 178 | spin_lock_bh(&fib_info_lock); |
1da177e4 LT |
179 | if (fi && --fi->fib_treeref == 0) { |
180 | hlist_del(&fi->fib_hash); | |
181 | if (fi->fib_prefsrc) | |
182 | hlist_del(&fi->fib_lhash); | |
183 | change_nexthops(fi) { | |
71fceff0 | 184 | if (!nexthop_nh->nh_dev) |
1da177e4 | 185 | continue; |
71fceff0 | 186 | hlist_del(&nexthop_nh->nh_hash); |
1da177e4 LT |
187 | } endfor_nexthops(fi) |
188 | fi->fib_dead = 1; | |
189 | fib_info_put(fi); | |
190 | } | |
832b4c5e | 191 | spin_unlock_bh(&fib_info_lock); |
1da177e4 LT |
192 | } |
193 | ||
6a31d2a9 | 194 | static inline int nh_comp(const struct fib_info *fi, const struct fib_info *ofi) |
1da177e4 LT |
195 | { |
196 | const struct fib_nh *onh = ofi->fib_nh; | |
197 | ||
198 | for_nexthops(fi) { | |
199 | if (nh->nh_oif != onh->nh_oif || | |
200 | nh->nh_gw != onh->nh_gw || | |
201 | nh->nh_scope != onh->nh_scope || | |
202 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
203 | nh->nh_weight != onh->nh_weight || | |
204 | #endif | |
c7066f70 | 205 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1da177e4 LT |
206 | nh->nh_tclassid != onh->nh_tclassid || |
207 | #endif | |
6a31d2a9 | 208 | ((nh->nh_flags ^ onh->nh_flags) & ~RTNH_F_DEAD)) |
1da177e4 LT |
209 | return -1; |
210 | onh++; | |
211 | } endfor_nexthops(fi); | |
212 | return 0; | |
213 | } | |
214 | ||
88ebc72f DM |
215 | static inline unsigned int fib_devindex_hashfn(unsigned int val) |
216 | { | |
217 | unsigned int mask = DEVINDEX_HASHSIZE - 1; | |
218 | ||
219 | return (val ^ | |
220 | (val >> DEVINDEX_HASHBITS) ^ | |
221 | (val >> (DEVINDEX_HASHBITS * 2))) & mask; | |
222 | } | |
223 | ||
1da177e4 LT |
224 | static inline unsigned int fib_info_hashfn(const struct fib_info *fi) |
225 | { | |
123b9731 | 226 | unsigned int mask = (fib_info_hash_size - 1); |
1da177e4 LT |
227 | unsigned int val = fi->fib_nhs; |
228 | ||
229 | val ^= fi->fib_protocol; | |
81f7bf6c | 230 | val ^= (__force u32)fi->fib_prefsrc; |
1da177e4 | 231 | val ^= fi->fib_priority; |
88ebc72f DM |
232 | for_nexthops(fi) { |
233 | val ^= fib_devindex_hashfn(nh->nh_oif); | |
234 | } endfor_nexthops(fi) | |
1da177e4 LT |
235 | |
236 | return (val ^ (val >> 7) ^ (val >> 12)) & mask; | |
237 | } | |
238 | ||
239 | static struct fib_info *fib_find_info(const struct fib_info *nfi) | |
240 | { | |
241 | struct hlist_head *head; | |
242 | struct hlist_node *node; | |
243 | struct fib_info *fi; | |
244 | unsigned int hash; | |
245 | ||
246 | hash = fib_info_hashfn(nfi); | |
247 | head = &fib_info_hash[hash]; | |
248 | ||
249 | hlist_for_each_entry(fi, node, head, fib_hash) { | |
09ad9bc7 | 250 | if (!net_eq(fi->fib_net, nfi->fib_net)) |
4814bdbd | 251 | continue; |
1da177e4 LT |
252 | if (fi->fib_nhs != nfi->fib_nhs) |
253 | continue; | |
254 | if (nfi->fib_protocol == fi->fib_protocol && | |
255 | nfi->fib_prefsrc == fi->fib_prefsrc && | |
256 | nfi->fib_priority == fi->fib_priority && | |
257 | memcmp(nfi->fib_metrics, fi->fib_metrics, | |
258 | sizeof(fi->fib_metrics)) == 0 && | |
6a31d2a9 | 259 | ((nfi->fib_flags ^ fi->fib_flags) & ~RTNH_F_DEAD) == 0 && |
1da177e4 LT |
260 | (nfi->fib_nhs == 0 || nh_comp(fi, nfi) == 0)) |
261 | return fi; | |
262 | } | |
263 | ||
264 | return NULL; | |
265 | } | |
266 | ||
1da177e4 | 267 | /* Check, that the gateway is already configured. |
6a31d2a9 | 268 | * Used only by redirect accept routine. |
1da177e4 | 269 | */ |
d878e72e | 270 | int ip_fib_check_default(__be32 gw, struct net_device *dev) |
1da177e4 LT |
271 | { |
272 | struct hlist_head *head; | |
273 | struct hlist_node *node; | |
274 | struct fib_nh *nh; | |
275 | unsigned int hash; | |
276 | ||
832b4c5e | 277 | spin_lock(&fib_info_lock); |
1da177e4 LT |
278 | |
279 | hash = fib_devindex_hashfn(dev->ifindex); | |
280 | head = &fib_info_devhash[hash]; | |
281 | hlist_for_each_entry(nh, node, head, nh_hash) { | |
282 | if (nh->nh_dev == dev && | |
283 | nh->nh_gw == gw && | |
6a31d2a9 | 284 | !(nh->nh_flags & RTNH_F_DEAD)) { |
832b4c5e | 285 | spin_unlock(&fib_info_lock); |
1da177e4 LT |
286 | return 0; |
287 | } | |
288 | } | |
289 | ||
832b4c5e | 290 | spin_unlock(&fib_info_lock); |
1da177e4 LT |
291 | |
292 | return -1; | |
293 | } | |
294 | ||
339bf98f TG |
295 | static inline size_t fib_nlmsg_size(struct fib_info *fi) |
296 | { | |
297 | size_t payload = NLMSG_ALIGN(sizeof(struct rtmsg)) | |
298 | + nla_total_size(4) /* RTA_TABLE */ | |
299 | + nla_total_size(4) /* RTA_DST */ | |
300 | + nla_total_size(4) /* RTA_PRIORITY */ | |
301 | + nla_total_size(4); /* RTA_PREFSRC */ | |
302 | ||
303 | /* space for nested metrics */ | |
304 | payload += nla_total_size((RTAX_MAX * nla_total_size(4))); | |
305 | ||
306 | if (fi->fib_nhs) { | |
307 | /* Also handles the special case fib_nhs == 1 */ | |
308 | ||
309 | /* each nexthop is packed in an attribute */ | |
310 | size_t nhsize = nla_total_size(sizeof(struct rtnexthop)); | |
311 | ||
312 | /* may contain flow and gateway attribute */ | |
313 | nhsize += 2 * nla_total_size(4); | |
314 | ||
315 | /* all nexthops are packed in a nested attribute */ | |
316 | payload += nla_total_size(fi->fib_nhs * nhsize); | |
317 | } | |
318 | ||
319 | return payload; | |
320 | } | |
321 | ||
81f7bf6c | 322 | void rtmsg_fib(int event, __be32 key, struct fib_alias *fa, |
b8f55831 MK |
323 | int dst_len, u32 tb_id, struct nl_info *info, |
324 | unsigned int nlm_flags) | |
1da177e4 LT |
325 | { |
326 | struct sk_buff *skb; | |
4e902c57 | 327 | u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0; |
f21c7bc5 | 328 | int err = -ENOBUFS; |
1da177e4 | 329 | |
339bf98f | 330 | skb = nlmsg_new(fib_nlmsg_size(fa->fa_info), GFP_KERNEL); |
f21c7bc5 TG |
331 | if (skb == NULL) |
332 | goto errout; | |
1da177e4 | 333 | |
4e902c57 | 334 | err = fib_dump_info(skb, info->pid, seq, event, tb_id, |
be403ea1 | 335 | fa->fa_type, fa->fa_scope, key, dst_len, |
b8f55831 | 336 | fa->fa_tos, fa->fa_info, nlm_flags); |
26932566 PM |
337 | if (err < 0) { |
338 | /* -EMSGSIZE implies BUG in fib_nlmsg_size() */ | |
339 | WARN_ON(err == -EMSGSIZE); | |
340 | kfree_skb(skb); | |
341 | goto errout; | |
342 | } | |
1ce85fe4 PNA |
343 | rtnl_notify(skb, info->nl_net, info->pid, RTNLGRP_IPV4_ROUTE, |
344 | info->nlh, GFP_KERNEL); | |
345 | return; | |
f21c7bc5 TG |
346 | errout: |
347 | if (err < 0) | |
4d1169c1 | 348 | rtnl_set_sk_err(info->nl_net, RTNLGRP_IPV4_ROUTE, err); |
1da177e4 LT |
349 | } |
350 | ||
351 | /* Return the first fib alias matching TOS with | |
352 | * priority less than or equal to PRIO. | |
353 | */ | |
354 | struct fib_alias *fib_find_alias(struct list_head *fah, u8 tos, u32 prio) | |
355 | { | |
356 | if (fah) { | |
357 | struct fib_alias *fa; | |
358 | list_for_each_entry(fa, fah, fa_list) { | |
359 | if (fa->fa_tos > tos) | |
360 | continue; | |
361 | if (fa->fa_info->fib_priority >= prio || | |
362 | fa->fa_tos < tos) | |
363 | return fa; | |
364 | } | |
365 | } | |
366 | return NULL; | |
367 | } | |
368 | ||
369 | int fib_detect_death(struct fib_info *fi, int order, | |
c17860a0 | 370 | struct fib_info **last_resort, int *last_idx, int dflt) |
1da177e4 LT |
371 | { |
372 | struct neighbour *n; | |
373 | int state = NUD_NONE; | |
374 | ||
375 | n = neigh_lookup(&arp_tbl, &fi->fib_nh[0].nh_gw, fi->fib_dev); | |
376 | if (n) { | |
377 | state = n->nud_state; | |
378 | neigh_release(n); | |
379 | } | |
d9319100 | 380 | if (state == NUD_REACHABLE) |
1da177e4 | 381 | return 0; |
6a31d2a9 | 382 | if ((state & NUD_VALID) && order != dflt) |
1da177e4 | 383 | return 0; |
6a31d2a9 ED |
384 | if ((state & NUD_VALID) || |
385 | (*last_idx < 0 && order > dflt)) { | |
1da177e4 LT |
386 | *last_resort = fi; |
387 | *last_idx = order; | |
388 | } | |
389 | return 1; | |
390 | } | |
391 | ||
392 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
393 | ||
4e902c57 | 394 | static int fib_count_nexthops(struct rtnexthop *rtnh, int remaining) |
1da177e4 LT |
395 | { |
396 | int nhs = 0; | |
1da177e4 | 397 | |
4e902c57 | 398 | while (rtnh_ok(rtnh, remaining)) { |
1da177e4 | 399 | nhs++; |
4e902c57 TG |
400 | rtnh = rtnh_next(rtnh, &remaining); |
401 | } | |
402 | ||
403 | /* leftover implies invalid nexthop configuration, discard it */ | |
404 | return remaining > 0 ? 0 : nhs; | |
1da177e4 LT |
405 | } |
406 | ||
4e902c57 TG |
407 | static int fib_get_nhs(struct fib_info *fi, struct rtnexthop *rtnh, |
408 | int remaining, struct fib_config *cfg) | |
1da177e4 | 409 | { |
1da177e4 | 410 | change_nexthops(fi) { |
4e902c57 TG |
411 | int attrlen; |
412 | ||
413 | if (!rtnh_ok(rtnh, remaining)) | |
1da177e4 | 414 | return -EINVAL; |
4e902c57 | 415 | |
71fceff0 DM |
416 | nexthop_nh->nh_flags = |
417 | (cfg->fc_flags & ~0xFF) | rtnh->rtnh_flags; | |
418 | nexthop_nh->nh_oif = rtnh->rtnh_ifindex; | |
419 | nexthop_nh->nh_weight = rtnh->rtnh_hops + 1; | |
4e902c57 TG |
420 | |
421 | attrlen = rtnh_attrlen(rtnh); | |
422 | if (attrlen > 0) { | |
423 | struct nlattr *nla, *attrs = rtnh_attrs(rtnh); | |
424 | ||
425 | nla = nla_find(attrs, attrlen, RTA_GATEWAY); | |
71fceff0 | 426 | nexthop_nh->nh_gw = nla ? nla_get_be32(nla) : 0; |
c7066f70 | 427 | #ifdef CONFIG_IP_ROUTE_CLASSID |
4e902c57 | 428 | nla = nla_find(attrs, attrlen, RTA_FLOW); |
71fceff0 | 429 | nexthop_nh->nh_tclassid = nla ? nla_get_u32(nla) : 0; |
1da177e4 LT |
430 | #endif |
431 | } | |
4e902c57 TG |
432 | |
433 | rtnh = rtnh_next(rtnh, &remaining); | |
1da177e4 | 434 | } endfor_nexthops(fi); |
4e902c57 | 435 | |
1da177e4 LT |
436 | return 0; |
437 | } | |
438 | ||
439 | #endif | |
440 | ||
4e902c57 | 441 | int fib_nh_match(struct fib_config *cfg, struct fib_info *fi) |
1da177e4 LT |
442 | { |
443 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
4e902c57 TG |
444 | struct rtnexthop *rtnh; |
445 | int remaining; | |
1da177e4 LT |
446 | #endif |
447 | ||
4e902c57 | 448 | if (cfg->fc_priority && cfg->fc_priority != fi->fib_priority) |
1da177e4 LT |
449 | return 1; |
450 | ||
4e902c57 TG |
451 | if (cfg->fc_oif || cfg->fc_gw) { |
452 | if ((!cfg->fc_oif || cfg->fc_oif == fi->fib_nh->nh_oif) && | |
453 | (!cfg->fc_gw || cfg->fc_gw == fi->fib_nh->nh_gw)) | |
1da177e4 LT |
454 | return 0; |
455 | return 1; | |
456 | } | |
457 | ||
458 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
4e902c57 | 459 | if (cfg->fc_mp == NULL) |
1da177e4 | 460 | return 0; |
4e902c57 TG |
461 | |
462 | rtnh = cfg->fc_mp; | |
463 | remaining = cfg->fc_mp_len; | |
e905a9ed | 464 | |
1da177e4 | 465 | for_nexthops(fi) { |
4e902c57 | 466 | int attrlen; |
1da177e4 | 467 | |
4e902c57 | 468 | if (!rtnh_ok(rtnh, remaining)) |
1da177e4 | 469 | return -EINVAL; |
4e902c57 TG |
470 | |
471 | if (rtnh->rtnh_ifindex && rtnh->rtnh_ifindex != nh->nh_oif) | |
1da177e4 | 472 | return 1; |
4e902c57 TG |
473 | |
474 | attrlen = rtnh_attrlen(rtnh); | |
475 | if (attrlen < 0) { | |
476 | struct nlattr *nla, *attrs = rtnh_attrs(rtnh); | |
477 | ||
478 | nla = nla_find(attrs, attrlen, RTA_GATEWAY); | |
17fb2c64 | 479 | if (nla && nla_get_be32(nla) != nh->nh_gw) |
1da177e4 | 480 | return 1; |
c7066f70 | 481 | #ifdef CONFIG_IP_ROUTE_CLASSID |
4e902c57 TG |
482 | nla = nla_find(attrs, attrlen, RTA_FLOW); |
483 | if (nla && nla_get_u32(nla) != nh->nh_tclassid) | |
1da177e4 LT |
484 | return 1; |
485 | #endif | |
486 | } | |
4e902c57 TG |
487 | |
488 | rtnh = rtnh_next(rtnh, &remaining); | |
1da177e4 LT |
489 | } endfor_nexthops(fi); |
490 | #endif | |
491 | return 0; | |
492 | } | |
493 | ||
494 | ||
495 | /* | |
6a31d2a9 ED |
496 | * Picture |
497 | * ------- | |
498 | * | |
499 | * Semantics of nexthop is very messy by historical reasons. | |
500 | * We have to take into account, that: | |
501 | * a) gateway can be actually local interface address, | |
502 | * so that gatewayed route is direct. | |
503 | * b) gateway must be on-link address, possibly | |
504 | * described not by an ifaddr, but also by a direct route. | |
505 | * c) If both gateway and interface are specified, they should not | |
506 | * contradict. | |
507 | * d) If we use tunnel routes, gateway could be not on-link. | |
508 | * | |
509 | * Attempt to reconcile all of these (alas, self-contradictory) conditions | |
510 | * results in pretty ugly and hairy code with obscure logic. | |
511 | * | |
512 | * I chose to generalized it instead, so that the size | |
513 | * of code does not increase practically, but it becomes | |
514 | * much more general. | |
515 | * Every prefix is assigned a "scope" value: "host" is local address, | |
516 | * "link" is direct route, | |
517 | * [ ... "site" ... "interior" ... ] | |
518 | * and "universe" is true gateway route with global meaning. | |
519 | * | |
520 | * Every prefix refers to a set of "nexthop"s (gw, oif), | |
521 | * where gw must have narrower scope. This recursion stops | |
522 | * when gw has LOCAL scope or if "nexthop" is declared ONLINK, | |
523 | * which means that gw is forced to be on link. | |
524 | * | |
525 | * Code is still hairy, but now it is apparently logically | |
526 | * consistent and very flexible. F.e. as by-product it allows | |
527 | * to co-exists in peace independent exterior and interior | |
528 | * routing processes. | |
529 | * | |
530 | * Normally it looks as following. | |
531 | * | |
532 | * {universe prefix} -> (gw, oif) [scope link] | |
533 | * | | |
534 | * |-> {link prefix} -> (gw, oif) [scope local] | |
535 | * | | |
536 | * |-> {local prefix} (terminal node) | |
1da177e4 | 537 | */ |
4e902c57 TG |
538 | static int fib_check_nh(struct fib_config *cfg, struct fib_info *fi, |
539 | struct fib_nh *nh) | |
1da177e4 LT |
540 | { |
541 | int err; | |
86167a37 | 542 | struct net *net; |
6a31d2a9 | 543 | struct net_device *dev; |
1da177e4 | 544 | |
86167a37 | 545 | net = cfg->fc_nlinfo.nl_net; |
1da177e4 LT |
546 | if (nh->nh_gw) { |
547 | struct fib_result res; | |
548 | ||
6a31d2a9 | 549 | if (nh->nh_flags & RTNH_F_ONLINK) { |
1da177e4 | 550 | |
4e902c57 | 551 | if (cfg->fc_scope >= RT_SCOPE_LINK) |
1da177e4 | 552 | return -EINVAL; |
86167a37 | 553 | if (inet_addr_type(net, nh->nh_gw) != RTN_UNICAST) |
1da177e4 | 554 | return -EINVAL; |
6a31d2a9 ED |
555 | dev = __dev_get_by_index(net, nh->nh_oif); |
556 | if (!dev) | |
1da177e4 | 557 | return -ENODEV; |
6a31d2a9 | 558 | if (!(dev->flags & IFF_UP)) |
1da177e4 LT |
559 | return -ENETDOWN; |
560 | nh->nh_dev = dev; | |
561 | dev_hold(dev); | |
562 | nh->nh_scope = RT_SCOPE_LINK; | |
563 | return 0; | |
564 | } | |
ebc0ffae | 565 | rcu_read_lock(); |
1da177e4 | 566 | { |
4e902c57 | 567 | struct flowi fl = { |
5811662b CG |
568 | .fl4_dst = nh->nh_gw, |
569 | .fl4_scope = cfg->fc_scope + 1, | |
4e902c57 TG |
570 | .oif = nh->nh_oif, |
571 | }; | |
1da177e4 LT |
572 | |
573 | /* It is not necessary, but requires a bit of thinking */ | |
574 | if (fl.fl4_scope < RT_SCOPE_LINK) | |
575 | fl.fl4_scope = RT_SCOPE_LINK; | |
6a31d2a9 | 576 | err = fib_lookup(net, &fl, &res); |
ebc0ffae ED |
577 | if (err) { |
578 | rcu_read_unlock(); | |
1da177e4 | 579 | return err; |
ebc0ffae | 580 | } |
1da177e4 LT |
581 | } |
582 | err = -EINVAL; | |
583 | if (res.type != RTN_UNICAST && res.type != RTN_LOCAL) | |
584 | goto out; | |
585 | nh->nh_scope = res.scope; | |
586 | nh->nh_oif = FIB_RES_OIF(res); | |
6a31d2a9 ED |
587 | nh->nh_dev = dev = FIB_RES_DEV(res); |
588 | if (!dev) | |
1da177e4 | 589 | goto out; |
6a31d2a9 | 590 | dev_hold(dev); |
8723e1b4 | 591 | err = (dev->flags & IFF_UP) ? 0 : -ENETDOWN; |
1da177e4 LT |
592 | } else { |
593 | struct in_device *in_dev; | |
594 | ||
6a31d2a9 | 595 | if (nh->nh_flags & (RTNH_F_PERVASIVE | RTNH_F_ONLINK)) |
1da177e4 LT |
596 | return -EINVAL; |
597 | ||
8723e1b4 ED |
598 | rcu_read_lock(); |
599 | err = -ENODEV; | |
86167a37 | 600 | in_dev = inetdev_by_index(net, nh->nh_oif); |
1da177e4 | 601 | if (in_dev == NULL) |
8723e1b4 ED |
602 | goto out; |
603 | err = -ENETDOWN; | |
604 | if (!(in_dev->dev->flags & IFF_UP)) | |
605 | goto out; | |
1da177e4 LT |
606 | nh->nh_dev = in_dev->dev; |
607 | dev_hold(nh->nh_dev); | |
608 | nh->nh_scope = RT_SCOPE_HOST; | |
8723e1b4 | 609 | err = 0; |
1da177e4 | 610 | } |
8723e1b4 ED |
611 | out: |
612 | rcu_read_unlock(); | |
613 | return err; | |
1da177e4 LT |
614 | } |
615 | ||
81f7bf6c | 616 | static inline unsigned int fib_laddr_hashfn(__be32 val) |
1da177e4 | 617 | { |
123b9731 | 618 | unsigned int mask = (fib_info_hash_size - 1); |
1da177e4 | 619 | |
6a31d2a9 ED |
620 | return ((__force u32)val ^ |
621 | ((__force u32)val >> 7) ^ | |
622 | ((__force u32)val >> 14)) & mask; | |
1da177e4 LT |
623 | } |
624 | ||
123b9731 | 625 | static struct hlist_head *fib_info_hash_alloc(int bytes) |
1da177e4 LT |
626 | { |
627 | if (bytes <= PAGE_SIZE) | |
88f83491 | 628 | return kzalloc(bytes, GFP_KERNEL); |
1da177e4 LT |
629 | else |
630 | return (struct hlist_head *) | |
6a31d2a9 ED |
631 | __get_free_pages(GFP_KERNEL | __GFP_ZERO, |
632 | get_order(bytes)); | |
1da177e4 LT |
633 | } |
634 | ||
123b9731 | 635 | static void fib_info_hash_free(struct hlist_head *hash, int bytes) |
1da177e4 LT |
636 | { |
637 | if (!hash) | |
638 | return; | |
639 | ||
640 | if (bytes <= PAGE_SIZE) | |
641 | kfree(hash); | |
642 | else | |
643 | free_pages((unsigned long) hash, get_order(bytes)); | |
644 | } | |
645 | ||
123b9731 DM |
646 | static void fib_info_hash_move(struct hlist_head *new_info_hash, |
647 | struct hlist_head *new_laddrhash, | |
648 | unsigned int new_size) | |
1da177e4 | 649 | { |
b7656e7f | 650 | struct hlist_head *old_info_hash, *old_laddrhash; |
123b9731 | 651 | unsigned int old_size = fib_info_hash_size; |
b7656e7f | 652 | unsigned int i, bytes; |
1da177e4 | 653 | |
832b4c5e | 654 | spin_lock_bh(&fib_info_lock); |
b7656e7f DM |
655 | old_info_hash = fib_info_hash; |
656 | old_laddrhash = fib_info_laddrhash; | |
123b9731 | 657 | fib_info_hash_size = new_size; |
1da177e4 LT |
658 | |
659 | for (i = 0; i < old_size; i++) { | |
660 | struct hlist_head *head = &fib_info_hash[i]; | |
661 | struct hlist_node *node, *n; | |
662 | struct fib_info *fi; | |
663 | ||
664 | hlist_for_each_entry_safe(fi, node, n, head, fib_hash) { | |
665 | struct hlist_head *dest; | |
666 | unsigned int new_hash; | |
667 | ||
668 | hlist_del(&fi->fib_hash); | |
669 | ||
670 | new_hash = fib_info_hashfn(fi); | |
671 | dest = &new_info_hash[new_hash]; | |
672 | hlist_add_head(&fi->fib_hash, dest); | |
673 | } | |
674 | } | |
675 | fib_info_hash = new_info_hash; | |
676 | ||
677 | for (i = 0; i < old_size; i++) { | |
678 | struct hlist_head *lhead = &fib_info_laddrhash[i]; | |
679 | struct hlist_node *node, *n; | |
680 | struct fib_info *fi; | |
681 | ||
682 | hlist_for_each_entry_safe(fi, node, n, lhead, fib_lhash) { | |
683 | struct hlist_head *ldest; | |
684 | unsigned int new_hash; | |
685 | ||
686 | hlist_del(&fi->fib_lhash); | |
687 | ||
688 | new_hash = fib_laddr_hashfn(fi->fib_prefsrc); | |
689 | ldest = &new_laddrhash[new_hash]; | |
690 | hlist_add_head(&fi->fib_lhash, ldest); | |
691 | } | |
692 | } | |
693 | fib_info_laddrhash = new_laddrhash; | |
694 | ||
832b4c5e | 695 | spin_unlock_bh(&fib_info_lock); |
b7656e7f DM |
696 | |
697 | bytes = old_size * sizeof(struct hlist_head *); | |
123b9731 DM |
698 | fib_info_hash_free(old_info_hash, bytes); |
699 | fib_info_hash_free(old_laddrhash, bytes); | |
1da177e4 LT |
700 | } |
701 | ||
4e902c57 | 702 | struct fib_info *fib_create_info(struct fib_config *cfg) |
1da177e4 LT |
703 | { |
704 | int err; | |
705 | struct fib_info *fi = NULL; | |
706 | struct fib_info *ofi; | |
1da177e4 | 707 | int nhs = 1; |
7462bd74 | 708 | struct net *net = cfg->fc_nlinfo.nl_net; |
1da177e4 LT |
709 | |
710 | /* Fast check to catch the most weird cases */ | |
4e902c57 | 711 | if (fib_props[cfg->fc_type].scope > cfg->fc_scope) |
1da177e4 LT |
712 | goto err_inval; |
713 | ||
714 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
4e902c57 TG |
715 | if (cfg->fc_mp) { |
716 | nhs = fib_count_nexthops(cfg->fc_mp, cfg->fc_mp_len); | |
1da177e4 LT |
717 | if (nhs == 0) |
718 | goto err_inval; | |
719 | } | |
720 | #endif | |
1da177e4 LT |
721 | |
722 | err = -ENOBUFS; | |
123b9731 DM |
723 | if (fib_info_cnt >= fib_info_hash_size) { |
724 | unsigned int new_size = fib_info_hash_size << 1; | |
1da177e4 LT |
725 | struct hlist_head *new_info_hash; |
726 | struct hlist_head *new_laddrhash; | |
727 | unsigned int bytes; | |
728 | ||
729 | if (!new_size) | |
730 | new_size = 1; | |
731 | bytes = new_size * sizeof(struct hlist_head *); | |
123b9731 DM |
732 | new_info_hash = fib_info_hash_alloc(bytes); |
733 | new_laddrhash = fib_info_hash_alloc(bytes); | |
1da177e4 | 734 | if (!new_info_hash || !new_laddrhash) { |
123b9731 DM |
735 | fib_info_hash_free(new_info_hash, bytes); |
736 | fib_info_hash_free(new_laddrhash, bytes); | |
88f83491 | 737 | } else |
123b9731 | 738 | fib_info_hash_move(new_info_hash, new_laddrhash, new_size); |
1da177e4 | 739 | |
123b9731 | 740 | if (!fib_info_hash_size) |
1da177e4 LT |
741 | goto failure; |
742 | } | |
743 | ||
0da974f4 | 744 | fi = kzalloc(sizeof(*fi)+nhs*sizeof(struct fib_nh), GFP_KERNEL); |
1da177e4 LT |
745 | if (fi == NULL) |
746 | goto failure; | |
725d1e1b DM |
747 | if (cfg->fc_mx) { |
748 | fi->fib_metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL); | |
749 | if (!fi->fib_metrics) | |
750 | goto failure; | |
751 | } else | |
752 | fi->fib_metrics = (u32 *) dst_default_metrics; | |
1da177e4 | 753 | fib_info_cnt++; |
1da177e4 | 754 | |
57d7a600 | 755 | fi->fib_net = hold_net(net); |
4e902c57 TG |
756 | fi->fib_protocol = cfg->fc_protocol; |
757 | fi->fib_flags = cfg->fc_flags; | |
758 | fi->fib_priority = cfg->fc_priority; | |
759 | fi->fib_prefsrc = cfg->fc_prefsrc; | |
1da177e4 LT |
760 | |
761 | fi->fib_nhs = nhs; | |
762 | change_nexthops(fi) { | |
71fceff0 | 763 | nexthop_nh->nh_parent = fi; |
1da177e4 LT |
764 | } endfor_nexthops(fi) |
765 | ||
4e902c57 TG |
766 | if (cfg->fc_mx) { |
767 | struct nlattr *nla; | |
768 | int remaining; | |
769 | ||
770 | nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) { | |
8f4c1f9b | 771 | int type = nla_type(nla); |
4e902c57 TG |
772 | |
773 | if (type) { | |
774 | if (type > RTAX_MAX) | |
1da177e4 | 775 | goto err_inval; |
4e902c57 | 776 | fi->fib_metrics[type - 1] = nla_get_u32(nla); |
1da177e4 | 777 | } |
1da177e4 LT |
778 | } |
779 | } | |
1da177e4 | 780 | |
4e902c57 | 781 | if (cfg->fc_mp) { |
1da177e4 | 782 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
4e902c57 TG |
783 | err = fib_get_nhs(fi, cfg->fc_mp, cfg->fc_mp_len, cfg); |
784 | if (err != 0) | |
1da177e4 | 785 | goto failure; |
4e902c57 | 786 | if (cfg->fc_oif && fi->fib_nh->nh_oif != cfg->fc_oif) |
1da177e4 | 787 | goto err_inval; |
4e902c57 | 788 | if (cfg->fc_gw && fi->fib_nh->nh_gw != cfg->fc_gw) |
1da177e4 | 789 | goto err_inval; |
c7066f70 | 790 | #ifdef CONFIG_IP_ROUTE_CLASSID |
4e902c57 | 791 | if (cfg->fc_flow && fi->fib_nh->nh_tclassid != cfg->fc_flow) |
1da177e4 LT |
792 | goto err_inval; |
793 | #endif | |
794 | #else | |
795 | goto err_inval; | |
796 | #endif | |
797 | } else { | |
798 | struct fib_nh *nh = fi->fib_nh; | |
4e902c57 TG |
799 | |
800 | nh->nh_oif = cfg->fc_oif; | |
801 | nh->nh_gw = cfg->fc_gw; | |
802 | nh->nh_flags = cfg->fc_flags; | |
c7066f70 | 803 | #ifdef CONFIG_IP_ROUTE_CLASSID |
4e902c57 | 804 | nh->nh_tclassid = cfg->fc_flow; |
1da177e4 | 805 | #endif |
1da177e4 LT |
806 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
807 | nh->nh_weight = 1; | |
808 | #endif | |
809 | } | |
810 | ||
4e902c57 TG |
811 | if (fib_props[cfg->fc_type].error) { |
812 | if (cfg->fc_gw || cfg->fc_oif || cfg->fc_mp) | |
1da177e4 LT |
813 | goto err_inval; |
814 | goto link_it; | |
815 | } | |
816 | ||
4e902c57 | 817 | if (cfg->fc_scope > RT_SCOPE_HOST) |
1da177e4 LT |
818 | goto err_inval; |
819 | ||
4e902c57 | 820 | if (cfg->fc_scope == RT_SCOPE_HOST) { |
1da177e4 LT |
821 | struct fib_nh *nh = fi->fib_nh; |
822 | ||
823 | /* Local address is added. */ | |
824 | if (nhs != 1 || nh->nh_gw) | |
825 | goto err_inval; | |
826 | nh->nh_scope = RT_SCOPE_NOWHERE; | |
7462bd74 | 827 | nh->nh_dev = dev_get_by_index(net, fi->fib_nh->nh_oif); |
1da177e4 LT |
828 | err = -ENODEV; |
829 | if (nh->nh_dev == NULL) | |
830 | goto failure; | |
831 | } else { | |
832 | change_nexthops(fi) { | |
6a31d2a9 ED |
833 | err = fib_check_nh(cfg, fi, nexthop_nh); |
834 | if (err != 0) | |
1da177e4 LT |
835 | goto failure; |
836 | } endfor_nexthops(fi) | |
837 | } | |
838 | ||
839 | if (fi->fib_prefsrc) { | |
4e902c57 TG |
840 | if (cfg->fc_type != RTN_LOCAL || !cfg->fc_dst || |
841 | fi->fib_prefsrc != cfg->fc_dst) | |
7462bd74 | 842 | if (inet_addr_type(net, fi->fib_prefsrc) != RTN_LOCAL) |
1da177e4 LT |
843 | goto err_inval; |
844 | } | |
845 | ||
846 | link_it: | |
6a31d2a9 ED |
847 | ofi = fib_find_info(fi); |
848 | if (ofi) { | |
1da177e4 LT |
849 | fi->fib_dead = 1; |
850 | free_fib_info(fi); | |
851 | ofi->fib_treeref++; | |
852 | return ofi; | |
853 | } | |
854 | ||
855 | fi->fib_treeref++; | |
856 | atomic_inc(&fi->fib_clntref); | |
832b4c5e | 857 | spin_lock_bh(&fib_info_lock); |
1da177e4 LT |
858 | hlist_add_head(&fi->fib_hash, |
859 | &fib_info_hash[fib_info_hashfn(fi)]); | |
860 | if (fi->fib_prefsrc) { | |
861 | struct hlist_head *head; | |
862 | ||
863 | head = &fib_info_laddrhash[fib_laddr_hashfn(fi->fib_prefsrc)]; | |
864 | hlist_add_head(&fi->fib_lhash, head); | |
865 | } | |
866 | change_nexthops(fi) { | |
867 | struct hlist_head *head; | |
868 | unsigned int hash; | |
869 | ||
71fceff0 | 870 | if (!nexthop_nh->nh_dev) |
1da177e4 | 871 | continue; |
71fceff0 | 872 | hash = fib_devindex_hashfn(nexthop_nh->nh_dev->ifindex); |
1da177e4 | 873 | head = &fib_info_devhash[hash]; |
71fceff0 | 874 | hlist_add_head(&nexthop_nh->nh_hash, head); |
1da177e4 | 875 | } endfor_nexthops(fi) |
832b4c5e | 876 | spin_unlock_bh(&fib_info_lock); |
1da177e4 LT |
877 | return fi; |
878 | ||
879 | err_inval: | |
880 | err = -EINVAL; | |
881 | ||
882 | failure: | |
e905a9ed | 883 | if (fi) { |
1da177e4 LT |
884 | fi->fib_dead = 1; |
885 | free_fib_info(fi); | |
886 | } | |
4e902c57 TG |
887 | |
888 | return ERR_PTR(err); | |
1da177e4 LT |
889 | } |
890 | ||
e5b43760 | 891 | /* Note! fib_semantic_match intentionally uses RCU list functions. */ |
5b470441 DM |
892 | int fib_semantic_match(struct fib_table *tb, struct list_head *head, |
893 | const struct flowi *flp, struct fib_result *res, | |
894 | int prefixlen, int fib_flags) | |
1da177e4 LT |
895 | { |
896 | struct fib_alias *fa; | |
897 | int nh_sel = 0; | |
898 | ||
e5b43760 | 899 | list_for_each_entry_rcu(fa, head, fa_list) { |
1da177e4 LT |
900 | int err; |
901 | ||
902 | if (fa->fa_tos && | |
903 | fa->fa_tos != flp->fl4_tos) | |
904 | continue; | |
905 | ||
906 | if (fa->fa_scope < flp->fl4_scope) | |
907 | continue; | |
908 | ||
9b0c290e | 909 | fib_alias_accessed(fa); |
1da177e4 LT |
910 | |
911 | err = fib_props[fa->fa_type].error; | |
912 | if (err == 0) { | |
913 | struct fib_info *fi = fa->fa_info; | |
914 | ||
915 | if (fi->fib_flags & RTNH_F_DEAD) | |
916 | continue; | |
917 | ||
918 | switch (fa->fa_type) { | |
919 | case RTN_UNICAST: | |
920 | case RTN_LOCAL: | |
921 | case RTN_BROADCAST: | |
922 | case RTN_ANYCAST: | |
923 | case RTN_MULTICAST: | |
924 | for_nexthops(fi) { | |
6a31d2a9 | 925 | if (nh->nh_flags & RTNH_F_DEAD) |
1da177e4 LT |
926 | continue; |
927 | if (!flp->oif || flp->oif == nh->nh_oif) | |
928 | break; | |
929 | } | |
930 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
931 | if (nhsel < fi->fib_nhs) { | |
932 | nh_sel = nhsel; | |
933 | goto out_fill_res; | |
934 | } | |
935 | #else | |
6a31d2a9 | 936 | if (nhsel < 1) |
1da177e4 | 937 | goto out_fill_res; |
1da177e4 LT |
938 | #endif |
939 | endfor_nexthops(fi); | |
940 | continue; | |
941 | ||
942 | default: | |
6a31d2a9 ED |
943 | pr_warning("fib_semantic_match bad type %#x\n", |
944 | fa->fa_type); | |
1da177e4 | 945 | return -EINVAL; |
3ff50b79 | 946 | } |
1da177e4 LT |
947 | } |
948 | return err; | |
949 | } | |
950 | return 1; | |
951 | ||
952 | out_fill_res: | |
953 | res->prefixlen = prefixlen; | |
954 | res->nh_sel = nh_sel; | |
955 | res->type = fa->fa_type; | |
956 | res->scope = fa->fa_scope; | |
957 | res->fi = fa->fa_info; | |
5b470441 DM |
958 | res->table = tb; |
959 | res->fa_head = head; | |
ebc0ffae ED |
960 | if (!(fib_flags & FIB_LOOKUP_NOREF)) |
961 | atomic_inc(&res->fi->fib_clntref); | |
1da177e4 LT |
962 | return 0; |
963 | } | |
964 | ||
965 | /* Find appropriate source address to this destination */ | |
966 | ||
b83738ae | 967 | __be32 __fib_res_prefsrc(struct fib_result *res) |
1da177e4 LT |
968 | { |
969 | return inet_select_addr(FIB_RES_DEV(*res), FIB_RES_GW(*res), res->scope); | |
970 | } | |
971 | ||
be403ea1 | 972 | int fib_dump_info(struct sk_buff *skb, u32 pid, u32 seq, int event, |
81f7bf6c | 973 | u32 tb_id, u8 type, u8 scope, __be32 dst, int dst_len, u8 tos, |
be403ea1 | 974 | struct fib_info *fi, unsigned int flags) |
1da177e4 | 975 | { |
be403ea1 | 976 | struct nlmsghdr *nlh; |
1da177e4 | 977 | struct rtmsg *rtm; |
1da177e4 | 978 | |
be403ea1 TG |
979 | nlh = nlmsg_put(skb, pid, seq, event, sizeof(*rtm), flags); |
980 | if (nlh == NULL) | |
26932566 | 981 | return -EMSGSIZE; |
be403ea1 TG |
982 | |
983 | rtm = nlmsg_data(nlh); | |
1da177e4 LT |
984 | rtm->rtm_family = AF_INET; |
985 | rtm->rtm_dst_len = dst_len; | |
986 | rtm->rtm_src_len = 0; | |
987 | rtm->rtm_tos = tos; | |
709772e6 KPO |
988 | if (tb_id < 256) |
989 | rtm->rtm_table = tb_id; | |
990 | else | |
991 | rtm->rtm_table = RT_TABLE_COMPAT; | |
be403ea1 | 992 | NLA_PUT_U32(skb, RTA_TABLE, tb_id); |
1da177e4 LT |
993 | rtm->rtm_type = type; |
994 | rtm->rtm_flags = fi->fib_flags; | |
995 | rtm->rtm_scope = scope; | |
1da177e4 | 996 | rtm->rtm_protocol = fi->fib_protocol; |
be403ea1 TG |
997 | |
998 | if (rtm->rtm_dst_len) | |
17fb2c64 | 999 | NLA_PUT_BE32(skb, RTA_DST, dst); |
be403ea1 | 1000 | |
1da177e4 | 1001 | if (fi->fib_priority) |
be403ea1 TG |
1002 | NLA_PUT_U32(skb, RTA_PRIORITY, fi->fib_priority); |
1003 | ||
1da177e4 | 1004 | if (rtnetlink_put_metrics(skb, fi->fib_metrics) < 0) |
be403ea1 TG |
1005 | goto nla_put_failure; |
1006 | ||
1da177e4 | 1007 | if (fi->fib_prefsrc) |
17fb2c64 | 1008 | NLA_PUT_BE32(skb, RTA_PREFSRC, fi->fib_prefsrc); |
be403ea1 | 1009 | |
1da177e4 LT |
1010 | if (fi->fib_nhs == 1) { |
1011 | if (fi->fib_nh->nh_gw) | |
17fb2c64 | 1012 | NLA_PUT_BE32(skb, RTA_GATEWAY, fi->fib_nh->nh_gw); |
be403ea1 | 1013 | |
1da177e4 | 1014 | if (fi->fib_nh->nh_oif) |
be403ea1 | 1015 | NLA_PUT_U32(skb, RTA_OIF, fi->fib_nh->nh_oif); |
c7066f70 | 1016 | #ifdef CONFIG_IP_ROUTE_CLASSID |
8265abc0 | 1017 | if (fi->fib_nh[0].nh_tclassid) |
be403ea1 | 1018 | NLA_PUT_U32(skb, RTA_FLOW, fi->fib_nh[0].nh_tclassid); |
8265abc0 | 1019 | #endif |
1da177e4 LT |
1020 | } |
1021 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
1022 | if (fi->fib_nhs > 1) { | |
be403ea1 TG |
1023 | struct rtnexthop *rtnh; |
1024 | struct nlattr *mp; | |
1025 | ||
1026 | mp = nla_nest_start(skb, RTA_MULTIPATH); | |
1027 | if (mp == NULL) | |
1028 | goto nla_put_failure; | |
1da177e4 LT |
1029 | |
1030 | for_nexthops(fi) { | |
be403ea1 TG |
1031 | rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh)); |
1032 | if (rtnh == NULL) | |
1033 | goto nla_put_failure; | |
1034 | ||
1035 | rtnh->rtnh_flags = nh->nh_flags & 0xFF; | |
1036 | rtnh->rtnh_hops = nh->nh_weight - 1; | |
1037 | rtnh->rtnh_ifindex = nh->nh_oif; | |
1038 | ||
1da177e4 | 1039 | if (nh->nh_gw) |
17fb2c64 | 1040 | NLA_PUT_BE32(skb, RTA_GATEWAY, nh->nh_gw); |
c7066f70 | 1041 | #ifdef CONFIG_IP_ROUTE_CLASSID |
8265abc0 | 1042 | if (nh->nh_tclassid) |
be403ea1 | 1043 | NLA_PUT_U32(skb, RTA_FLOW, nh->nh_tclassid); |
8265abc0 | 1044 | #endif |
be403ea1 TG |
1045 | /* length of rtnetlink header + attributes */ |
1046 | rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *) rtnh; | |
1da177e4 | 1047 | } endfor_nexthops(fi); |
be403ea1 TG |
1048 | |
1049 | nla_nest_end(skb, mp); | |
1da177e4 LT |
1050 | } |
1051 | #endif | |
be403ea1 | 1052 | return nlmsg_end(skb, nlh); |
1da177e4 | 1053 | |
be403ea1 | 1054 | nla_put_failure: |
26932566 PM |
1055 | nlmsg_cancel(skb, nlh); |
1056 | return -EMSGSIZE; | |
1da177e4 LT |
1057 | } |
1058 | ||
1da177e4 | 1059 | /* |
6a31d2a9 ED |
1060 | * Update FIB if: |
1061 | * - local address disappeared -> we must delete all the entries | |
1062 | * referring to it. | |
1063 | * - device went down -> we must shutdown all nexthops going via it. | |
1da177e4 | 1064 | */ |
4814bdbd | 1065 | int fib_sync_down_addr(struct net *net, __be32 local) |
1da177e4 LT |
1066 | { |
1067 | int ret = 0; | |
85326fa5 DL |
1068 | unsigned int hash = fib_laddr_hashfn(local); |
1069 | struct hlist_head *head = &fib_info_laddrhash[hash]; | |
1070 | struct hlist_node *node; | |
1071 | struct fib_info *fi; | |
1da177e4 | 1072 | |
85326fa5 DL |
1073 | if (fib_info_laddrhash == NULL || local == 0) |
1074 | return 0; | |
1da177e4 | 1075 | |
85326fa5 | 1076 | hlist_for_each_entry(fi, node, head, fib_lhash) { |
09ad9bc7 | 1077 | if (!net_eq(fi->fib_net, net)) |
4814bdbd | 1078 | continue; |
85326fa5 DL |
1079 | if (fi->fib_prefsrc == local) { |
1080 | fi->fib_flags |= RTNH_F_DEAD; | |
1081 | ret++; | |
1da177e4 LT |
1082 | } |
1083 | } | |
85326fa5 DL |
1084 | return ret; |
1085 | } | |
1086 | ||
1087 | int fib_sync_down_dev(struct net_device *dev, int force) | |
1088 | { | |
1089 | int ret = 0; | |
1090 | int scope = RT_SCOPE_NOWHERE; | |
1091 | struct fib_info *prev_fi = NULL; | |
1092 | unsigned int hash = fib_devindex_hashfn(dev->ifindex); | |
1093 | struct hlist_head *head = &fib_info_devhash[hash]; | |
1094 | struct hlist_node *node; | |
1095 | struct fib_nh *nh; | |
1da177e4 | 1096 | |
85326fa5 DL |
1097 | if (force) |
1098 | scope = -1; | |
1da177e4 | 1099 | |
85326fa5 DL |
1100 | hlist_for_each_entry(nh, node, head, nh_hash) { |
1101 | struct fib_info *fi = nh->nh_parent; | |
1102 | int dead; | |
1da177e4 | 1103 | |
85326fa5 DL |
1104 | BUG_ON(!fi->fib_nhs); |
1105 | if (nh->nh_dev != dev || fi == prev_fi) | |
1106 | continue; | |
1107 | prev_fi = fi; | |
1108 | dead = 0; | |
1109 | change_nexthops(fi) { | |
6a31d2a9 | 1110 | if (nexthop_nh->nh_flags & RTNH_F_DEAD) |
85326fa5 | 1111 | dead++; |
71fceff0 DM |
1112 | else if (nexthop_nh->nh_dev == dev && |
1113 | nexthop_nh->nh_scope != scope) { | |
1114 | nexthop_nh->nh_flags |= RTNH_F_DEAD; | |
1da177e4 | 1115 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
85326fa5 | 1116 | spin_lock_bh(&fib_multipath_lock); |
71fceff0 DM |
1117 | fi->fib_power -= nexthop_nh->nh_power; |
1118 | nexthop_nh->nh_power = 0; | |
85326fa5 | 1119 | spin_unlock_bh(&fib_multipath_lock); |
1da177e4 | 1120 | #endif |
85326fa5 DL |
1121 | dead++; |
1122 | } | |
1da177e4 | 1123 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
71fceff0 | 1124 | if (force > 1 && nexthop_nh->nh_dev == dev) { |
85326fa5 DL |
1125 | dead = fi->fib_nhs; |
1126 | break; | |
1da177e4 | 1127 | } |
85326fa5 DL |
1128 | #endif |
1129 | } endfor_nexthops(fi) | |
1130 | if (dead == fi->fib_nhs) { | |
1131 | fi->fib_flags |= RTNH_F_DEAD; | |
1132 | ret++; | |
1da177e4 LT |
1133 | } |
1134 | } | |
1135 | ||
1136 | return ret; | |
1137 | } | |
1138 | ||
0c838ff1 DM |
1139 | /* Must be invoked inside of an RCU protected region. */ |
1140 | void fib_select_default(struct fib_result *res) | |
1141 | { | |
1142 | struct fib_info *fi = NULL, *last_resort = NULL; | |
1143 | struct list_head *fa_head = res->fa_head; | |
1144 | struct fib_table *tb = res->table; | |
1145 | int order = -1, last_idx = -1; | |
1146 | struct fib_alias *fa; | |
1147 | ||
1148 | list_for_each_entry_rcu(fa, fa_head, fa_list) { | |
1149 | struct fib_info *next_fi = fa->fa_info; | |
1150 | ||
1151 | if (fa->fa_scope != res->scope || | |
1152 | fa->fa_type != RTN_UNICAST) | |
1153 | continue; | |
1154 | ||
1155 | if (next_fi->fib_priority > res->fi->fib_priority) | |
1156 | break; | |
1157 | if (!next_fi->fib_nh[0].nh_gw || | |
1158 | next_fi->fib_nh[0].nh_scope != RT_SCOPE_LINK) | |
1159 | continue; | |
1160 | ||
1161 | fib_alias_accessed(fa); | |
1162 | ||
1163 | if (fi == NULL) { | |
1164 | if (next_fi != res->fi) | |
1165 | break; | |
1166 | } else if (!fib_detect_death(fi, order, &last_resort, | |
1167 | &last_idx, tb->tb_default)) { | |
1168 | fib_result_assign(res, fi); | |
1169 | tb->tb_default = order; | |
1170 | goto out; | |
1171 | } | |
1172 | fi = next_fi; | |
1173 | order++; | |
1174 | } | |
1175 | ||
1176 | if (order <= 0 || fi == NULL) { | |
1177 | tb->tb_default = -1; | |
1178 | goto out; | |
1179 | } | |
1180 | ||
1181 | if (!fib_detect_death(fi, order, &last_resort, &last_idx, | |
1182 | tb->tb_default)) { | |
1183 | fib_result_assign(res, fi); | |
1184 | tb->tb_default = order; | |
1185 | goto out; | |
1186 | } | |
1187 | ||
1188 | if (last_idx >= 0) | |
1189 | fib_result_assign(res, last_resort); | |
1190 | tb->tb_default = last_idx; | |
1191 | out: | |
31d40937 | 1192 | return; |
0c838ff1 DM |
1193 | } |
1194 | ||
1da177e4 LT |
1195 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
1196 | ||
1197 | /* | |
6a31d2a9 ED |
1198 | * Dead device goes up. We wake up dead nexthops. |
1199 | * It takes sense only on multipath routes. | |
1da177e4 | 1200 | */ |
1da177e4 LT |
1201 | int fib_sync_up(struct net_device *dev) |
1202 | { | |
1203 | struct fib_info *prev_fi; | |
1204 | unsigned int hash; | |
1205 | struct hlist_head *head; | |
1206 | struct hlist_node *node; | |
1207 | struct fib_nh *nh; | |
1208 | int ret; | |
1209 | ||
6a31d2a9 | 1210 | if (!(dev->flags & IFF_UP)) |
1da177e4 LT |
1211 | return 0; |
1212 | ||
1213 | prev_fi = NULL; | |
1214 | hash = fib_devindex_hashfn(dev->ifindex); | |
1215 | head = &fib_info_devhash[hash]; | |
1216 | ret = 0; | |
1217 | ||
1218 | hlist_for_each_entry(nh, node, head, nh_hash) { | |
1219 | struct fib_info *fi = nh->nh_parent; | |
1220 | int alive; | |
1221 | ||
1222 | BUG_ON(!fi->fib_nhs); | |
1223 | if (nh->nh_dev != dev || fi == prev_fi) | |
1224 | continue; | |
1225 | ||
1226 | prev_fi = fi; | |
1227 | alive = 0; | |
1228 | change_nexthops(fi) { | |
6a31d2a9 | 1229 | if (!(nexthop_nh->nh_flags & RTNH_F_DEAD)) { |
1da177e4 LT |
1230 | alive++; |
1231 | continue; | |
1232 | } | |
71fceff0 | 1233 | if (nexthop_nh->nh_dev == NULL || |
6a31d2a9 | 1234 | !(nexthop_nh->nh_dev->flags & IFF_UP)) |
1da177e4 | 1235 | continue; |
71fceff0 DM |
1236 | if (nexthop_nh->nh_dev != dev || |
1237 | !__in_dev_get_rtnl(dev)) | |
1da177e4 LT |
1238 | continue; |
1239 | alive++; | |
1240 | spin_lock_bh(&fib_multipath_lock); | |
71fceff0 DM |
1241 | nexthop_nh->nh_power = 0; |
1242 | nexthop_nh->nh_flags &= ~RTNH_F_DEAD; | |
1da177e4 LT |
1243 | spin_unlock_bh(&fib_multipath_lock); |
1244 | } endfor_nexthops(fi) | |
1245 | ||
1246 | if (alive > 0) { | |
1247 | fi->fib_flags &= ~RTNH_F_DEAD; | |
1248 | ret++; | |
1249 | } | |
1250 | } | |
1251 | ||
1252 | return ret; | |
1253 | } | |
1254 | ||
1255 | /* | |
6a31d2a9 ED |
1256 | * The algorithm is suboptimal, but it provides really |
1257 | * fair weighted route distribution. | |
1da177e4 | 1258 | */ |
1da177e4 LT |
1259 | void fib_select_multipath(const struct flowi *flp, struct fib_result *res) |
1260 | { | |
1261 | struct fib_info *fi = res->fi; | |
1262 | int w; | |
1263 | ||
1264 | spin_lock_bh(&fib_multipath_lock); | |
1265 | if (fi->fib_power <= 0) { | |
1266 | int power = 0; | |
1267 | change_nexthops(fi) { | |
6a31d2a9 | 1268 | if (!(nexthop_nh->nh_flags & RTNH_F_DEAD)) { |
71fceff0 DM |
1269 | power += nexthop_nh->nh_weight; |
1270 | nexthop_nh->nh_power = nexthop_nh->nh_weight; | |
1da177e4 LT |
1271 | } |
1272 | } endfor_nexthops(fi); | |
1273 | fi->fib_power = power; | |
1274 | if (power <= 0) { | |
1275 | spin_unlock_bh(&fib_multipath_lock); | |
1276 | /* Race condition: route has just become dead. */ | |
1277 | res->nh_sel = 0; | |
1278 | return; | |
1279 | } | |
1280 | } | |
1281 | ||
1282 | ||
1283 | /* w should be random number [0..fi->fib_power-1], | |
6a31d2a9 | 1284 | * it is pretty bad approximation. |
1da177e4 LT |
1285 | */ |
1286 | ||
1287 | w = jiffies % fi->fib_power; | |
1288 | ||
1289 | change_nexthops(fi) { | |
6a31d2a9 | 1290 | if (!(nexthop_nh->nh_flags & RTNH_F_DEAD) && |
71fceff0 | 1291 | nexthop_nh->nh_power) { |
6a31d2a9 ED |
1292 | w -= nexthop_nh->nh_power; |
1293 | if (w <= 0) { | |
71fceff0 | 1294 | nexthop_nh->nh_power--; |
1da177e4 LT |
1295 | fi->fib_power--; |
1296 | res->nh_sel = nhsel; | |
1297 | spin_unlock_bh(&fib_multipath_lock); | |
1298 | return; | |
1299 | } | |
1300 | } | |
1301 | } endfor_nexthops(fi); | |
1302 | ||
1303 | /* Race condition: route has just become dead. */ | |
1304 | res->nh_sel = 0; | |
1305 | spin_unlock_bh(&fib_multipath_lock); | |
1306 | } | |
1307 | #endif |