]>
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; | |
52 | static unsigned int fib_hash_size; | |
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 | ||
151 | void free_fib_info(struct fib_info *fi) | |
152 | { | |
153 | if (fi->fib_dead == 0) { | |
6a31d2a9 | 154 | pr_warning("Freeing alive fib_info %p\n", fi); |
1da177e4 LT |
155 | return; |
156 | } | |
157 | change_nexthops(fi) { | |
71fceff0 DM |
158 | if (nexthop_nh->nh_dev) |
159 | dev_put(nexthop_nh->nh_dev); | |
160 | nexthop_nh->nh_dev = NULL; | |
1da177e4 LT |
161 | } endfor_nexthops(fi); |
162 | fib_info_cnt--; | |
57d7a600 | 163 | release_net(fi->fib_net); |
1da177e4 LT |
164 | kfree(fi); |
165 | } | |
166 | ||
167 | void fib_release_info(struct fib_info *fi) | |
168 | { | |
832b4c5e | 169 | spin_lock_bh(&fib_info_lock); |
1da177e4 LT |
170 | if (fi && --fi->fib_treeref == 0) { |
171 | hlist_del(&fi->fib_hash); | |
172 | if (fi->fib_prefsrc) | |
173 | hlist_del(&fi->fib_lhash); | |
174 | change_nexthops(fi) { | |
71fceff0 | 175 | if (!nexthop_nh->nh_dev) |
1da177e4 | 176 | continue; |
71fceff0 | 177 | hlist_del(&nexthop_nh->nh_hash); |
1da177e4 LT |
178 | } endfor_nexthops(fi) |
179 | fi->fib_dead = 1; | |
180 | fib_info_put(fi); | |
181 | } | |
832b4c5e | 182 | spin_unlock_bh(&fib_info_lock); |
1da177e4 LT |
183 | } |
184 | ||
6a31d2a9 | 185 | static inline int nh_comp(const struct fib_info *fi, const struct fib_info *ofi) |
1da177e4 LT |
186 | { |
187 | const struct fib_nh *onh = ofi->fib_nh; | |
188 | ||
189 | for_nexthops(fi) { | |
190 | if (nh->nh_oif != onh->nh_oif || | |
191 | nh->nh_gw != onh->nh_gw || | |
192 | nh->nh_scope != onh->nh_scope || | |
193 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
194 | nh->nh_weight != onh->nh_weight || | |
195 | #endif | |
196 | #ifdef CONFIG_NET_CLS_ROUTE | |
197 | nh->nh_tclassid != onh->nh_tclassid || | |
198 | #endif | |
6a31d2a9 | 199 | ((nh->nh_flags ^ onh->nh_flags) & ~RTNH_F_DEAD)) |
1da177e4 LT |
200 | return -1; |
201 | onh++; | |
202 | } endfor_nexthops(fi); | |
203 | return 0; | |
204 | } | |
205 | ||
88ebc72f DM |
206 | static inline unsigned int fib_devindex_hashfn(unsigned int val) |
207 | { | |
208 | unsigned int mask = DEVINDEX_HASHSIZE - 1; | |
209 | ||
210 | return (val ^ | |
211 | (val >> DEVINDEX_HASHBITS) ^ | |
212 | (val >> (DEVINDEX_HASHBITS * 2))) & mask; | |
213 | } | |
214 | ||
1da177e4 LT |
215 | static inline unsigned int fib_info_hashfn(const struct fib_info *fi) |
216 | { | |
217 | unsigned int mask = (fib_hash_size - 1); | |
218 | unsigned int val = fi->fib_nhs; | |
219 | ||
220 | val ^= fi->fib_protocol; | |
81f7bf6c | 221 | val ^= (__force u32)fi->fib_prefsrc; |
1da177e4 | 222 | val ^= fi->fib_priority; |
88ebc72f DM |
223 | for_nexthops(fi) { |
224 | val ^= fib_devindex_hashfn(nh->nh_oif); | |
225 | } endfor_nexthops(fi) | |
1da177e4 LT |
226 | |
227 | return (val ^ (val >> 7) ^ (val >> 12)) & mask; | |
228 | } | |
229 | ||
230 | static struct fib_info *fib_find_info(const struct fib_info *nfi) | |
231 | { | |
232 | struct hlist_head *head; | |
233 | struct hlist_node *node; | |
234 | struct fib_info *fi; | |
235 | unsigned int hash; | |
236 | ||
237 | hash = fib_info_hashfn(nfi); | |
238 | head = &fib_info_hash[hash]; | |
239 | ||
240 | hlist_for_each_entry(fi, node, head, fib_hash) { | |
09ad9bc7 | 241 | if (!net_eq(fi->fib_net, nfi->fib_net)) |
4814bdbd | 242 | continue; |
1da177e4 LT |
243 | if (fi->fib_nhs != nfi->fib_nhs) |
244 | continue; | |
245 | if (nfi->fib_protocol == fi->fib_protocol && | |
246 | nfi->fib_prefsrc == fi->fib_prefsrc && | |
247 | nfi->fib_priority == fi->fib_priority && | |
248 | memcmp(nfi->fib_metrics, fi->fib_metrics, | |
249 | sizeof(fi->fib_metrics)) == 0 && | |
6a31d2a9 | 250 | ((nfi->fib_flags ^ fi->fib_flags) & ~RTNH_F_DEAD) == 0 && |
1da177e4 LT |
251 | (nfi->fib_nhs == 0 || nh_comp(fi, nfi) == 0)) |
252 | return fi; | |
253 | } | |
254 | ||
255 | return NULL; | |
256 | } | |
257 | ||
1da177e4 | 258 | /* Check, that the gateway is already configured. |
6a31d2a9 | 259 | * Used only by redirect accept routine. |
1da177e4 | 260 | */ |
d878e72e | 261 | int ip_fib_check_default(__be32 gw, struct net_device *dev) |
1da177e4 LT |
262 | { |
263 | struct hlist_head *head; | |
264 | struct hlist_node *node; | |
265 | struct fib_nh *nh; | |
266 | unsigned int hash; | |
267 | ||
832b4c5e | 268 | spin_lock(&fib_info_lock); |
1da177e4 LT |
269 | |
270 | hash = fib_devindex_hashfn(dev->ifindex); | |
271 | head = &fib_info_devhash[hash]; | |
272 | hlist_for_each_entry(nh, node, head, nh_hash) { | |
273 | if (nh->nh_dev == dev && | |
274 | nh->nh_gw == gw && | |
6a31d2a9 | 275 | !(nh->nh_flags & RTNH_F_DEAD)) { |
832b4c5e | 276 | spin_unlock(&fib_info_lock); |
1da177e4 LT |
277 | return 0; |
278 | } | |
279 | } | |
280 | ||
832b4c5e | 281 | spin_unlock(&fib_info_lock); |
1da177e4 LT |
282 | |
283 | return -1; | |
284 | } | |
285 | ||
339bf98f TG |
286 | static inline size_t fib_nlmsg_size(struct fib_info *fi) |
287 | { | |
288 | size_t payload = NLMSG_ALIGN(sizeof(struct rtmsg)) | |
289 | + nla_total_size(4) /* RTA_TABLE */ | |
290 | + nla_total_size(4) /* RTA_DST */ | |
291 | + nla_total_size(4) /* RTA_PRIORITY */ | |
292 | + nla_total_size(4); /* RTA_PREFSRC */ | |
293 | ||
294 | /* space for nested metrics */ | |
295 | payload += nla_total_size((RTAX_MAX * nla_total_size(4))); | |
296 | ||
297 | if (fi->fib_nhs) { | |
298 | /* Also handles the special case fib_nhs == 1 */ | |
299 | ||
300 | /* each nexthop is packed in an attribute */ | |
301 | size_t nhsize = nla_total_size(sizeof(struct rtnexthop)); | |
302 | ||
303 | /* may contain flow and gateway attribute */ | |
304 | nhsize += 2 * nla_total_size(4); | |
305 | ||
306 | /* all nexthops are packed in a nested attribute */ | |
307 | payload += nla_total_size(fi->fib_nhs * nhsize); | |
308 | } | |
309 | ||
310 | return payload; | |
311 | } | |
312 | ||
81f7bf6c | 313 | void rtmsg_fib(int event, __be32 key, struct fib_alias *fa, |
b8f55831 MK |
314 | int dst_len, u32 tb_id, struct nl_info *info, |
315 | unsigned int nlm_flags) | |
1da177e4 LT |
316 | { |
317 | struct sk_buff *skb; | |
4e902c57 | 318 | u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0; |
f21c7bc5 | 319 | int err = -ENOBUFS; |
1da177e4 | 320 | |
339bf98f | 321 | skb = nlmsg_new(fib_nlmsg_size(fa->fa_info), GFP_KERNEL); |
f21c7bc5 TG |
322 | if (skb == NULL) |
323 | goto errout; | |
1da177e4 | 324 | |
4e902c57 | 325 | err = fib_dump_info(skb, info->pid, seq, event, tb_id, |
be403ea1 | 326 | fa->fa_type, fa->fa_scope, key, dst_len, |
b8f55831 | 327 | fa->fa_tos, fa->fa_info, nlm_flags); |
26932566 PM |
328 | if (err < 0) { |
329 | /* -EMSGSIZE implies BUG in fib_nlmsg_size() */ | |
330 | WARN_ON(err == -EMSGSIZE); | |
331 | kfree_skb(skb); | |
332 | goto errout; | |
333 | } | |
1ce85fe4 PNA |
334 | rtnl_notify(skb, info->nl_net, info->pid, RTNLGRP_IPV4_ROUTE, |
335 | info->nlh, GFP_KERNEL); | |
336 | return; | |
f21c7bc5 TG |
337 | errout: |
338 | if (err < 0) | |
4d1169c1 | 339 | rtnl_set_sk_err(info->nl_net, RTNLGRP_IPV4_ROUTE, err); |
1da177e4 LT |
340 | } |
341 | ||
342 | /* Return the first fib alias matching TOS with | |
343 | * priority less than or equal to PRIO. | |
344 | */ | |
345 | struct fib_alias *fib_find_alias(struct list_head *fah, u8 tos, u32 prio) | |
346 | { | |
347 | if (fah) { | |
348 | struct fib_alias *fa; | |
349 | list_for_each_entry(fa, fah, fa_list) { | |
350 | if (fa->fa_tos > tos) | |
351 | continue; | |
352 | if (fa->fa_info->fib_priority >= prio || | |
353 | fa->fa_tos < tos) | |
354 | return fa; | |
355 | } | |
356 | } | |
357 | return NULL; | |
358 | } | |
359 | ||
360 | int fib_detect_death(struct fib_info *fi, int order, | |
c17860a0 | 361 | struct fib_info **last_resort, int *last_idx, int dflt) |
1da177e4 LT |
362 | { |
363 | struct neighbour *n; | |
364 | int state = NUD_NONE; | |
365 | ||
366 | n = neigh_lookup(&arp_tbl, &fi->fib_nh[0].nh_gw, fi->fib_dev); | |
367 | if (n) { | |
368 | state = n->nud_state; | |
369 | neigh_release(n); | |
370 | } | |
d9319100 | 371 | if (state == NUD_REACHABLE) |
1da177e4 | 372 | return 0; |
6a31d2a9 | 373 | if ((state & NUD_VALID) && order != dflt) |
1da177e4 | 374 | return 0; |
6a31d2a9 ED |
375 | if ((state & NUD_VALID) || |
376 | (*last_idx < 0 && order > dflt)) { | |
1da177e4 LT |
377 | *last_resort = fi; |
378 | *last_idx = order; | |
379 | } | |
380 | return 1; | |
381 | } | |
382 | ||
383 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
384 | ||
4e902c57 | 385 | static int fib_count_nexthops(struct rtnexthop *rtnh, int remaining) |
1da177e4 LT |
386 | { |
387 | int nhs = 0; | |
1da177e4 | 388 | |
4e902c57 | 389 | while (rtnh_ok(rtnh, remaining)) { |
1da177e4 | 390 | nhs++; |
4e902c57 TG |
391 | rtnh = rtnh_next(rtnh, &remaining); |
392 | } | |
393 | ||
394 | /* leftover implies invalid nexthop configuration, discard it */ | |
395 | return remaining > 0 ? 0 : nhs; | |
1da177e4 LT |
396 | } |
397 | ||
4e902c57 TG |
398 | static int fib_get_nhs(struct fib_info *fi, struct rtnexthop *rtnh, |
399 | int remaining, struct fib_config *cfg) | |
1da177e4 | 400 | { |
1da177e4 | 401 | change_nexthops(fi) { |
4e902c57 TG |
402 | int attrlen; |
403 | ||
404 | if (!rtnh_ok(rtnh, remaining)) | |
1da177e4 | 405 | return -EINVAL; |
4e902c57 | 406 | |
71fceff0 DM |
407 | nexthop_nh->nh_flags = |
408 | (cfg->fc_flags & ~0xFF) | rtnh->rtnh_flags; | |
409 | nexthop_nh->nh_oif = rtnh->rtnh_ifindex; | |
410 | nexthop_nh->nh_weight = rtnh->rtnh_hops + 1; | |
4e902c57 TG |
411 | |
412 | attrlen = rtnh_attrlen(rtnh); | |
413 | if (attrlen > 0) { | |
414 | struct nlattr *nla, *attrs = rtnh_attrs(rtnh); | |
415 | ||
416 | nla = nla_find(attrs, attrlen, RTA_GATEWAY); | |
71fceff0 | 417 | nexthop_nh->nh_gw = nla ? nla_get_be32(nla) : 0; |
1da177e4 | 418 | #ifdef CONFIG_NET_CLS_ROUTE |
4e902c57 | 419 | nla = nla_find(attrs, attrlen, RTA_FLOW); |
71fceff0 | 420 | nexthop_nh->nh_tclassid = nla ? nla_get_u32(nla) : 0; |
1da177e4 LT |
421 | #endif |
422 | } | |
4e902c57 TG |
423 | |
424 | rtnh = rtnh_next(rtnh, &remaining); | |
1da177e4 | 425 | } endfor_nexthops(fi); |
4e902c57 | 426 | |
1da177e4 LT |
427 | return 0; |
428 | } | |
429 | ||
430 | #endif | |
431 | ||
4e902c57 | 432 | int fib_nh_match(struct fib_config *cfg, struct fib_info *fi) |
1da177e4 LT |
433 | { |
434 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
4e902c57 TG |
435 | struct rtnexthop *rtnh; |
436 | int remaining; | |
1da177e4 LT |
437 | #endif |
438 | ||
4e902c57 | 439 | if (cfg->fc_priority && cfg->fc_priority != fi->fib_priority) |
1da177e4 LT |
440 | return 1; |
441 | ||
4e902c57 TG |
442 | if (cfg->fc_oif || cfg->fc_gw) { |
443 | if ((!cfg->fc_oif || cfg->fc_oif == fi->fib_nh->nh_oif) && | |
444 | (!cfg->fc_gw || cfg->fc_gw == fi->fib_nh->nh_gw)) | |
1da177e4 LT |
445 | return 0; |
446 | return 1; | |
447 | } | |
448 | ||
449 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
4e902c57 | 450 | if (cfg->fc_mp == NULL) |
1da177e4 | 451 | return 0; |
4e902c57 TG |
452 | |
453 | rtnh = cfg->fc_mp; | |
454 | remaining = cfg->fc_mp_len; | |
e905a9ed | 455 | |
1da177e4 | 456 | for_nexthops(fi) { |
4e902c57 | 457 | int attrlen; |
1da177e4 | 458 | |
4e902c57 | 459 | if (!rtnh_ok(rtnh, remaining)) |
1da177e4 | 460 | return -EINVAL; |
4e902c57 TG |
461 | |
462 | if (rtnh->rtnh_ifindex && rtnh->rtnh_ifindex != nh->nh_oif) | |
1da177e4 | 463 | return 1; |
4e902c57 TG |
464 | |
465 | attrlen = rtnh_attrlen(rtnh); | |
466 | if (attrlen < 0) { | |
467 | struct nlattr *nla, *attrs = rtnh_attrs(rtnh); | |
468 | ||
469 | nla = nla_find(attrs, attrlen, RTA_GATEWAY); | |
17fb2c64 | 470 | if (nla && nla_get_be32(nla) != nh->nh_gw) |
1da177e4 LT |
471 | return 1; |
472 | #ifdef CONFIG_NET_CLS_ROUTE | |
4e902c57 TG |
473 | nla = nla_find(attrs, attrlen, RTA_FLOW); |
474 | if (nla && nla_get_u32(nla) != nh->nh_tclassid) | |
1da177e4 LT |
475 | return 1; |
476 | #endif | |
477 | } | |
4e902c57 TG |
478 | |
479 | rtnh = rtnh_next(rtnh, &remaining); | |
1da177e4 LT |
480 | } endfor_nexthops(fi); |
481 | #endif | |
482 | return 0; | |
483 | } | |
484 | ||
485 | ||
486 | /* | |
6a31d2a9 ED |
487 | * Picture |
488 | * ------- | |
489 | * | |
490 | * Semantics of nexthop is very messy by historical reasons. | |
491 | * We have to take into account, that: | |
492 | * a) gateway can be actually local interface address, | |
493 | * so that gatewayed route is direct. | |
494 | * b) gateway must be on-link address, possibly | |
495 | * described not by an ifaddr, but also by a direct route. | |
496 | * c) If both gateway and interface are specified, they should not | |
497 | * contradict. | |
498 | * d) If we use tunnel routes, gateway could be not on-link. | |
499 | * | |
500 | * Attempt to reconcile all of these (alas, self-contradictory) conditions | |
501 | * results in pretty ugly and hairy code with obscure logic. | |
502 | * | |
503 | * I chose to generalized it instead, so that the size | |
504 | * of code does not increase practically, but it becomes | |
505 | * much more general. | |
506 | * Every prefix is assigned a "scope" value: "host" is local address, | |
507 | * "link" is direct route, | |
508 | * [ ... "site" ... "interior" ... ] | |
509 | * and "universe" is true gateway route with global meaning. | |
510 | * | |
511 | * Every prefix refers to a set of "nexthop"s (gw, oif), | |
512 | * where gw must have narrower scope. This recursion stops | |
513 | * when gw has LOCAL scope or if "nexthop" is declared ONLINK, | |
514 | * which means that gw is forced to be on link. | |
515 | * | |
516 | * Code is still hairy, but now it is apparently logically | |
517 | * consistent and very flexible. F.e. as by-product it allows | |
518 | * to co-exists in peace independent exterior and interior | |
519 | * routing processes. | |
520 | * | |
521 | * Normally it looks as following. | |
522 | * | |
523 | * {universe prefix} -> (gw, oif) [scope link] | |
524 | * | | |
525 | * |-> {link prefix} -> (gw, oif) [scope local] | |
526 | * | | |
527 | * |-> {local prefix} (terminal node) | |
1da177e4 | 528 | */ |
4e902c57 TG |
529 | static int fib_check_nh(struct fib_config *cfg, struct fib_info *fi, |
530 | struct fib_nh *nh) | |
1da177e4 LT |
531 | { |
532 | int err; | |
86167a37 | 533 | struct net *net; |
6a31d2a9 | 534 | struct net_device *dev; |
1da177e4 | 535 | |
86167a37 | 536 | net = cfg->fc_nlinfo.nl_net; |
1da177e4 LT |
537 | if (nh->nh_gw) { |
538 | struct fib_result res; | |
539 | ||
6a31d2a9 | 540 | if (nh->nh_flags & RTNH_F_ONLINK) { |
1da177e4 | 541 | |
4e902c57 | 542 | if (cfg->fc_scope >= RT_SCOPE_LINK) |
1da177e4 | 543 | return -EINVAL; |
86167a37 | 544 | if (inet_addr_type(net, nh->nh_gw) != RTN_UNICAST) |
1da177e4 | 545 | return -EINVAL; |
6a31d2a9 ED |
546 | dev = __dev_get_by_index(net, nh->nh_oif); |
547 | if (!dev) | |
1da177e4 | 548 | return -ENODEV; |
6a31d2a9 | 549 | if (!(dev->flags & IFF_UP)) |
1da177e4 LT |
550 | return -ENETDOWN; |
551 | nh->nh_dev = dev; | |
552 | dev_hold(dev); | |
553 | nh->nh_scope = RT_SCOPE_LINK; | |
554 | return 0; | |
555 | } | |
556 | { | |
4e902c57 TG |
557 | struct flowi fl = { |
558 | .nl_u = { | |
559 | .ip4_u = { | |
560 | .daddr = nh->nh_gw, | |
561 | .scope = cfg->fc_scope + 1, | |
562 | }, | |
563 | }, | |
564 | .oif = nh->nh_oif, | |
565 | }; | |
1da177e4 LT |
566 | |
567 | /* It is not necessary, but requires a bit of thinking */ | |
568 | if (fl.fl4_scope < RT_SCOPE_LINK) | |
569 | fl.fl4_scope = RT_SCOPE_LINK; | |
6a31d2a9 ED |
570 | err = fib_lookup(net, &fl, &res); |
571 | if (err) | |
1da177e4 LT |
572 | return err; |
573 | } | |
574 | err = -EINVAL; | |
575 | if (res.type != RTN_UNICAST && res.type != RTN_LOCAL) | |
576 | goto out; | |
577 | nh->nh_scope = res.scope; | |
578 | nh->nh_oif = FIB_RES_OIF(res); | |
6a31d2a9 ED |
579 | nh->nh_dev = dev = FIB_RES_DEV(res); |
580 | if (!dev) | |
1da177e4 | 581 | goto out; |
6a31d2a9 | 582 | dev_hold(dev); |
1da177e4 | 583 | err = -ENETDOWN; |
6a31d2a9 | 584 | if (!(dev->flags & IFF_UP)) |
1da177e4 LT |
585 | goto out; |
586 | err = 0; | |
587 | out: | |
588 | fib_res_put(&res); | |
589 | return err; | |
590 | } else { | |
591 | struct in_device *in_dev; | |
592 | ||
6a31d2a9 | 593 | if (nh->nh_flags & (RTNH_F_PERVASIVE | RTNH_F_ONLINK)) |
1da177e4 LT |
594 | return -EINVAL; |
595 | ||
86167a37 | 596 | in_dev = inetdev_by_index(net, nh->nh_oif); |
1da177e4 LT |
597 | if (in_dev == NULL) |
598 | return -ENODEV; | |
6a31d2a9 | 599 | if (!(in_dev->dev->flags & IFF_UP)) { |
1da177e4 LT |
600 | in_dev_put(in_dev); |
601 | return -ENETDOWN; | |
602 | } | |
603 | nh->nh_dev = in_dev->dev; | |
604 | dev_hold(nh->nh_dev); | |
605 | nh->nh_scope = RT_SCOPE_HOST; | |
606 | in_dev_put(in_dev); | |
607 | } | |
608 | return 0; | |
609 | } | |
610 | ||
81f7bf6c | 611 | static inline unsigned int fib_laddr_hashfn(__be32 val) |
1da177e4 LT |
612 | { |
613 | unsigned int mask = (fib_hash_size - 1); | |
614 | ||
6a31d2a9 ED |
615 | return ((__force u32)val ^ |
616 | ((__force u32)val >> 7) ^ | |
617 | ((__force u32)val >> 14)) & mask; | |
1da177e4 LT |
618 | } |
619 | ||
620 | static struct hlist_head *fib_hash_alloc(int bytes) | |
621 | { | |
622 | if (bytes <= PAGE_SIZE) | |
88f83491 | 623 | return kzalloc(bytes, GFP_KERNEL); |
1da177e4 LT |
624 | else |
625 | return (struct hlist_head *) | |
6a31d2a9 ED |
626 | __get_free_pages(GFP_KERNEL | __GFP_ZERO, |
627 | get_order(bytes)); | |
1da177e4 LT |
628 | } |
629 | ||
630 | static void fib_hash_free(struct hlist_head *hash, int bytes) | |
631 | { | |
632 | if (!hash) | |
633 | return; | |
634 | ||
635 | if (bytes <= PAGE_SIZE) | |
636 | kfree(hash); | |
637 | else | |
638 | free_pages((unsigned long) hash, get_order(bytes)); | |
639 | } | |
640 | ||
641 | static void fib_hash_move(struct hlist_head *new_info_hash, | |
642 | struct hlist_head *new_laddrhash, | |
643 | unsigned int new_size) | |
644 | { | |
b7656e7f | 645 | struct hlist_head *old_info_hash, *old_laddrhash; |
1da177e4 | 646 | unsigned int old_size = fib_hash_size; |
b7656e7f | 647 | unsigned int i, bytes; |
1da177e4 | 648 | |
832b4c5e | 649 | spin_lock_bh(&fib_info_lock); |
b7656e7f DM |
650 | old_info_hash = fib_info_hash; |
651 | old_laddrhash = fib_info_laddrhash; | |
1da177e4 LT |
652 | fib_hash_size = new_size; |
653 | ||
654 | for (i = 0; i < old_size; i++) { | |
655 | struct hlist_head *head = &fib_info_hash[i]; | |
656 | struct hlist_node *node, *n; | |
657 | struct fib_info *fi; | |
658 | ||
659 | hlist_for_each_entry_safe(fi, node, n, head, fib_hash) { | |
660 | struct hlist_head *dest; | |
661 | unsigned int new_hash; | |
662 | ||
663 | hlist_del(&fi->fib_hash); | |
664 | ||
665 | new_hash = fib_info_hashfn(fi); | |
666 | dest = &new_info_hash[new_hash]; | |
667 | hlist_add_head(&fi->fib_hash, dest); | |
668 | } | |
669 | } | |
670 | fib_info_hash = new_info_hash; | |
671 | ||
672 | for (i = 0; i < old_size; i++) { | |
673 | struct hlist_head *lhead = &fib_info_laddrhash[i]; | |
674 | struct hlist_node *node, *n; | |
675 | struct fib_info *fi; | |
676 | ||
677 | hlist_for_each_entry_safe(fi, node, n, lhead, fib_lhash) { | |
678 | struct hlist_head *ldest; | |
679 | unsigned int new_hash; | |
680 | ||
681 | hlist_del(&fi->fib_lhash); | |
682 | ||
683 | new_hash = fib_laddr_hashfn(fi->fib_prefsrc); | |
684 | ldest = &new_laddrhash[new_hash]; | |
685 | hlist_add_head(&fi->fib_lhash, ldest); | |
686 | } | |
687 | } | |
688 | fib_info_laddrhash = new_laddrhash; | |
689 | ||
832b4c5e | 690 | spin_unlock_bh(&fib_info_lock); |
b7656e7f DM |
691 | |
692 | bytes = old_size * sizeof(struct hlist_head *); | |
693 | fib_hash_free(old_info_hash, bytes); | |
694 | fib_hash_free(old_laddrhash, bytes); | |
1da177e4 LT |
695 | } |
696 | ||
4e902c57 | 697 | struct fib_info *fib_create_info(struct fib_config *cfg) |
1da177e4 LT |
698 | { |
699 | int err; | |
700 | struct fib_info *fi = NULL; | |
701 | struct fib_info *ofi; | |
1da177e4 | 702 | int nhs = 1; |
7462bd74 | 703 | struct net *net = cfg->fc_nlinfo.nl_net; |
1da177e4 LT |
704 | |
705 | /* Fast check to catch the most weird cases */ | |
4e902c57 | 706 | if (fib_props[cfg->fc_type].scope > cfg->fc_scope) |
1da177e4 LT |
707 | goto err_inval; |
708 | ||
709 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
4e902c57 TG |
710 | if (cfg->fc_mp) { |
711 | nhs = fib_count_nexthops(cfg->fc_mp, cfg->fc_mp_len); | |
1da177e4 LT |
712 | if (nhs == 0) |
713 | goto err_inval; | |
714 | } | |
715 | #endif | |
1da177e4 LT |
716 | |
717 | err = -ENOBUFS; | |
718 | if (fib_info_cnt >= fib_hash_size) { | |
719 | unsigned int new_size = fib_hash_size << 1; | |
720 | struct hlist_head *new_info_hash; | |
721 | struct hlist_head *new_laddrhash; | |
722 | unsigned int bytes; | |
723 | ||
724 | if (!new_size) | |
725 | new_size = 1; | |
726 | bytes = new_size * sizeof(struct hlist_head *); | |
727 | new_info_hash = fib_hash_alloc(bytes); | |
728 | new_laddrhash = fib_hash_alloc(bytes); | |
729 | if (!new_info_hash || !new_laddrhash) { | |
730 | fib_hash_free(new_info_hash, bytes); | |
731 | fib_hash_free(new_laddrhash, bytes); | |
88f83491 | 732 | } else |
1da177e4 | 733 | fib_hash_move(new_info_hash, new_laddrhash, new_size); |
1da177e4 LT |
734 | |
735 | if (!fib_hash_size) | |
736 | goto failure; | |
737 | } | |
738 | ||
0da974f4 | 739 | fi = kzalloc(sizeof(*fi)+nhs*sizeof(struct fib_nh), GFP_KERNEL); |
1da177e4 LT |
740 | if (fi == NULL) |
741 | goto failure; | |
742 | fib_info_cnt++; | |
1da177e4 | 743 | |
57d7a600 | 744 | fi->fib_net = hold_net(net); |
4e902c57 TG |
745 | fi->fib_protocol = cfg->fc_protocol; |
746 | fi->fib_flags = cfg->fc_flags; | |
747 | fi->fib_priority = cfg->fc_priority; | |
748 | fi->fib_prefsrc = cfg->fc_prefsrc; | |
1da177e4 LT |
749 | |
750 | fi->fib_nhs = nhs; | |
751 | change_nexthops(fi) { | |
71fceff0 | 752 | nexthop_nh->nh_parent = fi; |
1da177e4 LT |
753 | } endfor_nexthops(fi) |
754 | ||
4e902c57 TG |
755 | if (cfg->fc_mx) { |
756 | struct nlattr *nla; | |
757 | int remaining; | |
758 | ||
759 | nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) { | |
8f4c1f9b | 760 | int type = nla_type(nla); |
4e902c57 TG |
761 | |
762 | if (type) { | |
763 | if (type > RTAX_MAX) | |
1da177e4 | 764 | goto err_inval; |
4e902c57 | 765 | fi->fib_metrics[type - 1] = nla_get_u32(nla); |
1da177e4 | 766 | } |
1da177e4 LT |
767 | } |
768 | } | |
1da177e4 | 769 | |
4e902c57 | 770 | if (cfg->fc_mp) { |
1da177e4 | 771 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
4e902c57 TG |
772 | err = fib_get_nhs(fi, cfg->fc_mp, cfg->fc_mp_len, cfg); |
773 | if (err != 0) | |
1da177e4 | 774 | goto failure; |
4e902c57 | 775 | if (cfg->fc_oif && fi->fib_nh->nh_oif != cfg->fc_oif) |
1da177e4 | 776 | goto err_inval; |
4e902c57 | 777 | if (cfg->fc_gw && fi->fib_nh->nh_gw != cfg->fc_gw) |
1da177e4 LT |
778 | goto err_inval; |
779 | #ifdef CONFIG_NET_CLS_ROUTE | |
4e902c57 | 780 | if (cfg->fc_flow && fi->fib_nh->nh_tclassid != cfg->fc_flow) |
1da177e4 LT |
781 | goto err_inval; |
782 | #endif | |
783 | #else | |
784 | goto err_inval; | |
785 | #endif | |
786 | } else { | |
787 | struct fib_nh *nh = fi->fib_nh; | |
4e902c57 TG |
788 | |
789 | nh->nh_oif = cfg->fc_oif; | |
790 | nh->nh_gw = cfg->fc_gw; | |
791 | nh->nh_flags = cfg->fc_flags; | |
1da177e4 | 792 | #ifdef CONFIG_NET_CLS_ROUTE |
4e902c57 | 793 | nh->nh_tclassid = cfg->fc_flow; |
1da177e4 | 794 | #endif |
1da177e4 LT |
795 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
796 | nh->nh_weight = 1; | |
797 | #endif | |
798 | } | |
799 | ||
4e902c57 TG |
800 | if (fib_props[cfg->fc_type].error) { |
801 | if (cfg->fc_gw || cfg->fc_oif || cfg->fc_mp) | |
1da177e4 LT |
802 | goto err_inval; |
803 | goto link_it; | |
804 | } | |
805 | ||
4e902c57 | 806 | if (cfg->fc_scope > RT_SCOPE_HOST) |
1da177e4 LT |
807 | goto err_inval; |
808 | ||
4e902c57 | 809 | if (cfg->fc_scope == RT_SCOPE_HOST) { |
1da177e4 LT |
810 | struct fib_nh *nh = fi->fib_nh; |
811 | ||
812 | /* Local address is added. */ | |
813 | if (nhs != 1 || nh->nh_gw) | |
814 | goto err_inval; | |
815 | nh->nh_scope = RT_SCOPE_NOWHERE; | |
7462bd74 | 816 | nh->nh_dev = dev_get_by_index(net, fi->fib_nh->nh_oif); |
1da177e4 LT |
817 | err = -ENODEV; |
818 | if (nh->nh_dev == NULL) | |
819 | goto failure; | |
820 | } else { | |
821 | change_nexthops(fi) { | |
6a31d2a9 ED |
822 | err = fib_check_nh(cfg, fi, nexthop_nh); |
823 | if (err != 0) | |
1da177e4 LT |
824 | goto failure; |
825 | } endfor_nexthops(fi) | |
826 | } | |
827 | ||
828 | if (fi->fib_prefsrc) { | |
4e902c57 TG |
829 | if (cfg->fc_type != RTN_LOCAL || !cfg->fc_dst || |
830 | fi->fib_prefsrc != cfg->fc_dst) | |
7462bd74 | 831 | if (inet_addr_type(net, fi->fib_prefsrc) != RTN_LOCAL) |
1da177e4 LT |
832 | goto err_inval; |
833 | } | |
834 | ||
835 | link_it: | |
6a31d2a9 ED |
836 | ofi = fib_find_info(fi); |
837 | if (ofi) { | |
1da177e4 LT |
838 | fi->fib_dead = 1; |
839 | free_fib_info(fi); | |
840 | ofi->fib_treeref++; | |
841 | return ofi; | |
842 | } | |
843 | ||
844 | fi->fib_treeref++; | |
845 | atomic_inc(&fi->fib_clntref); | |
832b4c5e | 846 | spin_lock_bh(&fib_info_lock); |
1da177e4 LT |
847 | hlist_add_head(&fi->fib_hash, |
848 | &fib_info_hash[fib_info_hashfn(fi)]); | |
849 | if (fi->fib_prefsrc) { | |
850 | struct hlist_head *head; | |
851 | ||
852 | head = &fib_info_laddrhash[fib_laddr_hashfn(fi->fib_prefsrc)]; | |
853 | hlist_add_head(&fi->fib_lhash, head); | |
854 | } | |
855 | change_nexthops(fi) { | |
856 | struct hlist_head *head; | |
857 | unsigned int hash; | |
858 | ||
71fceff0 | 859 | if (!nexthop_nh->nh_dev) |
1da177e4 | 860 | continue; |
71fceff0 | 861 | hash = fib_devindex_hashfn(nexthop_nh->nh_dev->ifindex); |
1da177e4 | 862 | head = &fib_info_devhash[hash]; |
71fceff0 | 863 | hlist_add_head(&nexthop_nh->nh_hash, head); |
1da177e4 | 864 | } endfor_nexthops(fi) |
832b4c5e | 865 | spin_unlock_bh(&fib_info_lock); |
1da177e4 LT |
866 | return fi; |
867 | ||
868 | err_inval: | |
869 | err = -EINVAL; | |
870 | ||
871 | failure: | |
e905a9ed | 872 | if (fi) { |
1da177e4 LT |
873 | fi->fib_dead = 1; |
874 | free_fib_info(fi); | |
875 | } | |
4e902c57 TG |
876 | |
877 | return ERR_PTR(err); | |
1da177e4 LT |
878 | } |
879 | ||
e5b43760 | 880 | /* Note! fib_semantic_match intentionally uses RCU list functions. */ |
1da177e4 | 881 | int fib_semantic_match(struct list_head *head, const struct flowi *flp, |
e204a345 | 882 | struct fib_result *res, int prefixlen) |
1da177e4 LT |
883 | { |
884 | struct fib_alias *fa; | |
885 | int nh_sel = 0; | |
886 | ||
e5b43760 | 887 | list_for_each_entry_rcu(fa, head, fa_list) { |
1da177e4 LT |
888 | int err; |
889 | ||
890 | if (fa->fa_tos && | |
891 | fa->fa_tos != flp->fl4_tos) | |
892 | continue; | |
893 | ||
894 | if (fa->fa_scope < flp->fl4_scope) | |
895 | continue; | |
896 | ||
897 | fa->fa_state |= FA_S_ACCESSED; | |
898 | ||
899 | err = fib_props[fa->fa_type].error; | |
900 | if (err == 0) { | |
901 | struct fib_info *fi = fa->fa_info; | |
902 | ||
903 | if (fi->fib_flags & RTNH_F_DEAD) | |
904 | continue; | |
905 | ||
906 | switch (fa->fa_type) { | |
907 | case RTN_UNICAST: | |
908 | case RTN_LOCAL: | |
909 | case RTN_BROADCAST: | |
910 | case RTN_ANYCAST: | |
911 | case RTN_MULTICAST: | |
912 | for_nexthops(fi) { | |
6a31d2a9 | 913 | if (nh->nh_flags & RTNH_F_DEAD) |
1da177e4 LT |
914 | continue; |
915 | if (!flp->oif || flp->oif == nh->nh_oif) | |
916 | break; | |
917 | } | |
918 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
919 | if (nhsel < fi->fib_nhs) { | |
920 | nh_sel = nhsel; | |
921 | goto out_fill_res; | |
922 | } | |
923 | #else | |
6a31d2a9 | 924 | if (nhsel < 1) |
1da177e4 | 925 | goto out_fill_res; |
1da177e4 LT |
926 | #endif |
927 | endfor_nexthops(fi); | |
928 | continue; | |
929 | ||
930 | default: | |
6a31d2a9 ED |
931 | pr_warning("fib_semantic_match bad type %#x\n", |
932 | fa->fa_type); | |
1da177e4 | 933 | return -EINVAL; |
3ff50b79 | 934 | } |
1da177e4 LT |
935 | } |
936 | return err; | |
937 | } | |
938 | return 1; | |
939 | ||
940 | out_fill_res: | |
941 | res->prefixlen = prefixlen; | |
942 | res->nh_sel = nh_sel; | |
943 | res->type = fa->fa_type; | |
944 | res->scope = fa->fa_scope; | |
945 | res->fi = fa->fa_info; | |
1da177e4 LT |
946 | atomic_inc(&res->fi->fib_clntref); |
947 | return 0; | |
948 | } | |
949 | ||
950 | /* Find appropriate source address to this destination */ | |
951 | ||
b83738ae | 952 | __be32 __fib_res_prefsrc(struct fib_result *res) |
1da177e4 LT |
953 | { |
954 | return inet_select_addr(FIB_RES_DEV(*res), FIB_RES_GW(*res), res->scope); | |
955 | } | |
956 | ||
be403ea1 | 957 | int fib_dump_info(struct sk_buff *skb, u32 pid, u32 seq, int event, |
81f7bf6c | 958 | u32 tb_id, u8 type, u8 scope, __be32 dst, int dst_len, u8 tos, |
be403ea1 | 959 | struct fib_info *fi, unsigned int flags) |
1da177e4 | 960 | { |
be403ea1 | 961 | struct nlmsghdr *nlh; |
1da177e4 | 962 | struct rtmsg *rtm; |
1da177e4 | 963 | |
be403ea1 TG |
964 | nlh = nlmsg_put(skb, pid, seq, event, sizeof(*rtm), flags); |
965 | if (nlh == NULL) | |
26932566 | 966 | return -EMSGSIZE; |
be403ea1 TG |
967 | |
968 | rtm = nlmsg_data(nlh); | |
1da177e4 LT |
969 | rtm->rtm_family = AF_INET; |
970 | rtm->rtm_dst_len = dst_len; | |
971 | rtm->rtm_src_len = 0; | |
972 | rtm->rtm_tos = tos; | |
709772e6 KPO |
973 | if (tb_id < 256) |
974 | rtm->rtm_table = tb_id; | |
975 | else | |
976 | rtm->rtm_table = RT_TABLE_COMPAT; | |
be403ea1 | 977 | NLA_PUT_U32(skb, RTA_TABLE, tb_id); |
1da177e4 LT |
978 | rtm->rtm_type = type; |
979 | rtm->rtm_flags = fi->fib_flags; | |
980 | rtm->rtm_scope = scope; | |
1da177e4 | 981 | rtm->rtm_protocol = fi->fib_protocol; |
be403ea1 TG |
982 | |
983 | if (rtm->rtm_dst_len) | |
17fb2c64 | 984 | NLA_PUT_BE32(skb, RTA_DST, dst); |
be403ea1 | 985 | |
1da177e4 | 986 | if (fi->fib_priority) |
be403ea1 TG |
987 | NLA_PUT_U32(skb, RTA_PRIORITY, fi->fib_priority); |
988 | ||
1da177e4 | 989 | if (rtnetlink_put_metrics(skb, fi->fib_metrics) < 0) |
be403ea1 TG |
990 | goto nla_put_failure; |
991 | ||
1da177e4 | 992 | if (fi->fib_prefsrc) |
17fb2c64 | 993 | NLA_PUT_BE32(skb, RTA_PREFSRC, fi->fib_prefsrc); |
be403ea1 | 994 | |
1da177e4 LT |
995 | if (fi->fib_nhs == 1) { |
996 | if (fi->fib_nh->nh_gw) | |
17fb2c64 | 997 | NLA_PUT_BE32(skb, RTA_GATEWAY, fi->fib_nh->nh_gw); |
be403ea1 | 998 | |
1da177e4 | 999 | if (fi->fib_nh->nh_oif) |
be403ea1 | 1000 | NLA_PUT_U32(skb, RTA_OIF, fi->fib_nh->nh_oif); |
8265abc0 PM |
1001 | #ifdef CONFIG_NET_CLS_ROUTE |
1002 | if (fi->fib_nh[0].nh_tclassid) | |
be403ea1 | 1003 | NLA_PUT_U32(skb, RTA_FLOW, fi->fib_nh[0].nh_tclassid); |
8265abc0 | 1004 | #endif |
1da177e4 LT |
1005 | } |
1006 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
1007 | if (fi->fib_nhs > 1) { | |
be403ea1 TG |
1008 | struct rtnexthop *rtnh; |
1009 | struct nlattr *mp; | |
1010 | ||
1011 | mp = nla_nest_start(skb, RTA_MULTIPATH); | |
1012 | if (mp == NULL) | |
1013 | goto nla_put_failure; | |
1da177e4 LT |
1014 | |
1015 | for_nexthops(fi) { | |
be403ea1 TG |
1016 | rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh)); |
1017 | if (rtnh == NULL) | |
1018 | goto nla_put_failure; | |
1019 | ||
1020 | rtnh->rtnh_flags = nh->nh_flags & 0xFF; | |
1021 | rtnh->rtnh_hops = nh->nh_weight - 1; | |
1022 | rtnh->rtnh_ifindex = nh->nh_oif; | |
1023 | ||
1da177e4 | 1024 | if (nh->nh_gw) |
17fb2c64 | 1025 | NLA_PUT_BE32(skb, RTA_GATEWAY, nh->nh_gw); |
8265abc0 PM |
1026 | #ifdef CONFIG_NET_CLS_ROUTE |
1027 | if (nh->nh_tclassid) | |
be403ea1 | 1028 | NLA_PUT_U32(skb, RTA_FLOW, nh->nh_tclassid); |
8265abc0 | 1029 | #endif |
be403ea1 TG |
1030 | /* length of rtnetlink header + attributes */ |
1031 | rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *) rtnh; | |
1da177e4 | 1032 | } endfor_nexthops(fi); |
be403ea1 TG |
1033 | |
1034 | nla_nest_end(skb, mp); | |
1da177e4 LT |
1035 | } |
1036 | #endif | |
be403ea1 | 1037 | return nlmsg_end(skb, nlh); |
1da177e4 | 1038 | |
be403ea1 | 1039 | nla_put_failure: |
26932566 PM |
1040 | nlmsg_cancel(skb, nlh); |
1041 | return -EMSGSIZE; | |
1da177e4 LT |
1042 | } |
1043 | ||
1da177e4 | 1044 | /* |
6a31d2a9 ED |
1045 | * Update FIB if: |
1046 | * - local address disappeared -> we must delete all the entries | |
1047 | * referring to it. | |
1048 | * - device went down -> we must shutdown all nexthops going via it. | |
1da177e4 | 1049 | */ |
4814bdbd | 1050 | int fib_sync_down_addr(struct net *net, __be32 local) |
1da177e4 LT |
1051 | { |
1052 | int ret = 0; | |
85326fa5 DL |
1053 | unsigned int hash = fib_laddr_hashfn(local); |
1054 | struct hlist_head *head = &fib_info_laddrhash[hash]; | |
1055 | struct hlist_node *node; | |
1056 | struct fib_info *fi; | |
1da177e4 | 1057 | |
85326fa5 DL |
1058 | if (fib_info_laddrhash == NULL || local == 0) |
1059 | return 0; | |
1da177e4 | 1060 | |
85326fa5 | 1061 | hlist_for_each_entry(fi, node, head, fib_lhash) { |
09ad9bc7 | 1062 | if (!net_eq(fi->fib_net, net)) |
4814bdbd | 1063 | continue; |
85326fa5 DL |
1064 | if (fi->fib_prefsrc == local) { |
1065 | fi->fib_flags |= RTNH_F_DEAD; | |
1066 | ret++; | |
1da177e4 LT |
1067 | } |
1068 | } | |
85326fa5 DL |
1069 | return ret; |
1070 | } | |
1071 | ||
1072 | int fib_sync_down_dev(struct net_device *dev, int force) | |
1073 | { | |
1074 | int ret = 0; | |
1075 | int scope = RT_SCOPE_NOWHERE; | |
1076 | struct fib_info *prev_fi = NULL; | |
1077 | unsigned int hash = fib_devindex_hashfn(dev->ifindex); | |
1078 | struct hlist_head *head = &fib_info_devhash[hash]; | |
1079 | struct hlist_node *node; | |
1080 | struct fib_nh *nh; | |
1da177e4 | 1081 | |
85326fa5 DL |
1082 | if (force) |
1083 | scope = -1; | |
1da177e4 | 1084 | |
85326fa5 DL |
1085 | hlist_for_each_entry(nh, node, head, nh_hash) { |
1086 | struct fib_info *fi = nh->nh_parent; | |
1087 | int dead; | |
1da177e4 | 1088 | |
85326fa5 DL |
1089 | BUG_ON(!fi->fib_nhs); |
1090 | if (nh->nh_dev != dev || fi == prev_fi) | |
1091 | continue; | |
1092 | prev_fi = fi; | |
1093 | dead = 0; | |
1094 | change_nexthops(fi) { | |
6a31d2a9 | 1095 | if (nexthop_nh->nh_flags & RTNH_F_DEAD) |
85326fa5 | 1096 | dead++; |
71fceff0 DM |
1097 | else if (nexthop_nh->nh_dev == dev && |
1098 | nexthop_nh->nh_scope != scope) { | |
1099 | nexthop_nh->nh_flags |= RTNH_F_DEAD; | |
1da177e4 | 1100 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
85326fa5 | 1101 | spin_lock_bh(&fib_multipath_lock); |
71fceff0 DM |
1102 | fi->fib_power -= nexthop_nh->nh_power; |
1103 | nexthop_nh->nh_power = 0; | |
85326fa5 | 1104 | spin_unlock_bh(&fib_multipath_lock); |
1da177e4 | 1105 | #endif |
85326fa5 DL |
1106 | dead++; |
1107 | } | |
1da177e4 | 1108 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
71fceff0 | 1109 | if (force > 1 && nexthop_nh->nh_dev == dev) { |
85326fa5 DL |
1110 | dead = fi->fib_nhs; |
1111 | break; | |
1da177e4 | 1112 | } |
85326fa5 DL |
1113 | #endif |
1114 | } endfor_nexthops(fi) | |
1115 | if (dead == fi->fib_nhs) { | |
1116 | fi->fib_flags |= RTNH_F_DEAD; | |
1117 | ret++; | |
1da177e4 LT |
1118 | } |
1119 | } | |
1120 | ||
1121 | return ret; | |
1122 | } | |
1123 | ||
1124 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
1125 | ||
1126 | /* | |
6a31d2a9 ED |
1127 | * Dead device goes up. We wake up dead nexthops. |
1128 | * It takes sense only on multipath routes. | |
1da177e4 | 1129 | */ |
1da177e4 LT |
1130 | int fib_sync_up(struct net_device *dev) |
1131 | { | |
1132 | struct fib_info *prev_fi; | |
1133 | unsigned int hash; | |
1134 | struct hlist_head *head; | |
1135 | struct hlist_node *node; | |
1136 | struct fib_nh *nh; | |
1137 | int ret; | |
1138 | ||
6a31d2a9 | 1139 | if (!(dev->flags & IFF_UP)) |
1da177e4 LT |
1140 | return 0; |
1141 | ||
1142 | prev_fi = NULL; | |
1143 | hash = fib_devindex_hashfn(dev->ifindex); | |
1144 | head = &fib_info_devhash[hash]; | |
1145 | ret = 0; | |
1146 | ||
1147 | hlist_for_each_entry(nh, node, head, nh_hash) { | |
1148 | struct fib_info *fi = nh->nh_parent; | |
1149 | int alive; | |
1150 | ||
1151 | BUG_ON(!fi->fib_nhs); | |
1152 | if (nh->nh_dev != dev || fi == prev_fi) | |
1153 | continue; | |
1154 | ||
1155 | prev_fi = fi; | |
1156 | alive = 0; | |
1157 | change_nexthops(fi) { | |
6a31d2a9 | 1158 | if (!(nexthop_nh->nh_flags & RTNH_F_DEAD)) { |
1da177e4 LT |
1159 | alive++; |
1160 | continue; | |
1161 | } | |
71fceff0 | 1162 | if (nexthop_nh->nh_dev == NULL || |
6a31d2a9 | 1163 | !(nexthop_nh->nh_dev->flags & IFF_UP)) |
1da177e4 | 1164 | continue; |
71fceff0 DM |
1165 | if (nexthop_nh->nh_dev != dev || |
1166 | !__in_dev_get_rtnl(dev)) | |
1da177e4 LT |
1167 | continue; |
1168 | alive++; | |
1169 | spin_lock_bh(&fib_multipath_lock); | |
71fceff0 DM |
1170 | nexthop_nh->nh_power = 0; |
1171 | nexthop_nh->nh_flags &= ~RTNH_F_DEAD; | |
1da177e4 LT |
1172 | spin_unlock_bh(&fib_multipath_lock); |
1173 | } endfor_nexthops(fi) | |
1174 | ||
1175 | if (alive > 0) { | |
1176 | fi->fib_flags &= ~RTNH_F_DEAD; | |
1177 | ret++; | |
1178 | } | |
1179 | } | |
1180 | ||
1181 | return ret; | |
1182 | } | |
1183 | ||
1184 | /* | |
6a31d2a9 ED |
1185 | * The algorithm is suboptimal, but it provides really |
1186 | * fair weighted route distribution. | |
1da177e4 | 1187 | */ |
1da177e4 LT |
1188 | void fib_select_multipath(const struct flowi *flp, struct fib_result *res) |
1189 | { | |
1190 | struct fib_info *fi = res->fi; | |
1191 | int w; | |
1192 | ||
1193 | spin_lock_bh(&fib_multipath_lock); | |
1194 | if (fi->fib_power <= 0) { | |
1195 | int power = 0; | |
1196 | change_nexthops(fi) { | |
6a31d2a9 | 1197 | if (!(nexthop_nh->nh_flags & RTNH_F_DEAD)) { |
71fceff0 DM |
1198 | power += nexthop_nh->nh_weight; |
1199 | nexthop_nh->nh_power = nexthop_nh->nh_weight; | |
1da177e4 LT |
1200 | } |
1201 | } endfor_nexthops(fi); | |
1202 | fi->fib_power = power; | |
1203 | if (power <= 0) { | |
1204 | spin_unlock_bh(&fib_multipath_lock); | |
1205 | /* Race condition: route has just become dead. */ | |
1206 | res->nh_sel = 0; | |
1207 | return; | |
1208 | } | |
1209 | } | |
1210 | ||
1211 | ||
1212 | /* w should be random number [0..fi->fib_power-1], | |
6a31d2a9 | 1213 | * it is pretty bad approximation. |
1da177e4 LT |
1214 | */ |
1215 | ||
1216 | w = jiffies % fi->fib_power; | |
1217 | ||
1218 | change_nexthops(fi) { | |
6a31d2a9 | 1219 | if (!(nexthop_nh->nh_flags & RTNH_F_DEAD) && |
71fceff0 | 1220 | nexthop_nh->nh_power) { |
6a31d2a9 ED |
1221 | w -= nexthop_nh->nh_power; |
1222 | if (w <= 0) { | |
71fceff0 | 1223 | nexthop_nh->nh_power--; |
1da177e4 LT |
1224 | fi->fib_power--; |
1225 | res->nh_sel = nhsel; | |
1226 | spin_unlock_bh(&fib_multipath_lock); | |
1227 | return; | |
1228 | } | |
1229 | } | |
1230 | } endfor_nexthops(fi); | |
1231 | ||
1232 | /* Race condition: route has just become dead. */ | |
1233 | res->nh_sel = 0; | |
1234 | spin_unlock_bh(&fib_multipath_lock); | |
1235 | } | |
1236 | #endif |