]>
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 | 16 | #include <asm/uaccess.h> |
1da177e4 LT |
17 | #include <linux/bitops.h> |
18 | #include <linux/types.h> | |
19 | #include <linux/kernel.h> | |
20 | #include <linux/jiffies.h> | |
21 | #include <linux/mm.h> | |
22 | #include <linux/string.h> | |
23 | #include <linux/socket.h> | |
24 | #include <linux/sockios.h> | |
25 | #include <linux/errno.h> | |
26 | #include <linux/in.h> | |
27 | #include <linux/inet.h> | |
14c85021 | 28 | #include <linux/inetdevice.h> |
1da177e4 LT |
29 | #include <linux/netdevice.h> |
30 | #include <linux/if_arp.h> | |
31 | #include <linux/proc_fs.h> | |
32 | #include <linux/skbuff.h> | |
1da177e4 | 33 | #include <linux/init.h> |
5a0e3ad6 | 34 | #include <linux/slab.h> |
1da177e4 | 35 | |
14c85021 | 36 | #include <net/arp.h> |
1da177e4 LT |
37 | #include <net/ip.h> |
38 | #include <net/protocol.h> | |
39 | #include <net/route.h> | |
40 | #include <net/tcp.h> | |
41 | #include <net/sock.h> | |
42 | #include <net/ip_fib.h> | |
f21c7bc5 | 43 | #include <net/netlink.h> |
4e902c57 | 44 | #include <net/nexthop.h> |
1da177e4 LT |
45 | |
46 | #include "fib_lookup.h" | |
47 | ||
832b4c5e | 48 | static DEFINE_SPINLOCK(fib_info_lock); |
1da177e4 LT |
49 | static struct hlist_head *fib_info_hash; |
50 | static struct hlist_head *fib_info_laddrhash; | |
123b9731 | 51 | static unsigned int fib_info_hash_size; |
1da177e4 LT |
52 | static unsigned int fib_info_cnt; |
53 | ||
54 | #define DEVINDEX_HASHBITS 8 | |
55 | #define DEVINDEX_HASHSIZE (1U << DEVINDEX_HASHBITS) | |
56 | static struct hlist_head fib_info_devhash[DEVINDEX_HASHSIZE]; | |
57 | ||
58 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
59 | ||
60 | static DEFINE_SPINLOCK(fib_multipath_lock); | |
61 | ||
6a31d2a9 ED |
62 | #define for_nexthops(fi) { \ |
63 | int nhsel; const struct fib_nh *nh; \ | |
64 | for (nhsel = 0, nh = (fi)->fib_nh; \ | |
65 | nhsel < (fi)->fib_nhs; \ | |
66 | nh++, nhsel++) | |
67 | ||
68 | #define change_nexthops(fi) { \ | |
69 | int nhsel; struct fib_nh *nexthop_nh; \ | |
70 | for (nhsel = 0, nexthop_nh = (struct fib_nh *)((fi)->fib_nh); \ | |
71 | nhsel < (fi)->fib_nhs; \ | |
72 | nexthop_nh++, nhsel++) | |
1da177e4 LT |
73 | |
74 | #else /* CONFIG_IP_ROUTE_MULTIPATH */ | |
75 | ||
76 | /* Hope, that gcc will optimize it to get rid of dummy loop */ | |
77 | ||
6a31d2a9 ED |
78 | #define for_nexthops(fi) { \ |
79 | int nhsel; const struct fib_nh *nh = (fi)->fib_nh; \ | |
80 | for (nhsel = 0; nhsel < 1; nhsel++) | |
1da177e4 | 81 | |
6a31d2a9 ED |
82 | #define change_nexthops(fi) { \ |
83 | int nhsel; \ | |
84 | struct fib_nh *nexthop_nh = (struct fib_nh *)((fi)->fib_nh); \ | |
85 | for (nhsel = 0; nhsel < 1; nhsel++) | |
1da177e4 LT |
86 | |
87 | #endif /* CONFIG_IP_ROUTE_MULTIPATH */ | |
88 | ||
89 | #define endfor_nexthops(fi) } | |
90 | ||
91 | ||
3be0686b | 92 | const struct fib_prop fib_props[RTN_MAX + 1] = { |
6a31d2a9 | 93 | [RTN_UNSPEC] = { |
1da177e4 LT |
94 | .error = 0, |
95 | .scope = RT_SCOPE_NOWHERE, | |
6a31d2a9 ED |
96 | }, |
97 | [RTN_UNICAST] = { | |
1da177e4 LT |
98 | .error = 0, |
99 | .scope = RT_SCOPE_UNIVERSE, | |
6a31d2a9 ED |
100 | }, |
101 | [RTN_LOCAL] = { | |
1da177e4 LT |
102 | .error = 0, |
103 | .scope = RT_SCOPE_HOST, | |
6a31d2a9 ED |
104 | }, |
105 | [RTN_BROADCAST] = { | |
1da177e4 LT |
106 | .error = 0, |
107 | .scope = RT_SCOPE_LINK, | |
6a31d2a9 ED |
108 | }, |
109 | [RTN_ANYCAST] = { | |
1da177e4 LT |
110 | .error = 0, |
111 | .scope = RT_SCOPE_LINK, | |
6a31d2a9 ED |
112 | }, |
113 | [RTN_MULTICAST] = { | |
1da177e4 LT |
114 | .error = 0, |
115 | .scope = RT_SCOPE_UNIVERSE, | |
6a31d2a9 ED |
116 | }, |
117 | [RTN_BLACKHOLE] = { | |
1da177e4 LT |
118 | .error = -EINVAL, |
119 | .scope = RT_SCOPE_UNIVERSE, | |
6a31d2a9 ED |
120 | }, |
121 | [RTN_UNREACHABLE] = { | |
1da177e4 LT |
122 | .error = -EHOSTUNREACH, |
123 | .scope = RT_SCOPE_UNIVERSE, | |
6a31d2a9 ED |
124 | }, |
125 | [RTN_PROHIBIT] = { | |
1da177e4 LT |
126 | .error = -EACCES, |
127 | .scope = RT_SCOPE_UNIVERSE, | |
6a31d2a9 ED |
128 | }, |
129 | [RTN_THROW] = { | |
1da177e4 LT |
130 | .error = -EAGAIN, |
131 | .scope = RT_SCOPE_UNIVERSE, | |
6a31d2a9 ED |
132 | }, |
133 | [RTN_NAT] = { | |
1da177e4 LT |
134 | .error = -EINVAL, |
135 | .scope = RT_SCOPE_NOWHERE, | |
6a31d2a9 ED |
136 | }, |
137 | [RTN_XRESOLVE] = { | |
1da177e4 LT |
138 | .error = -EINVAL, |
139 | .scope = RT_SCOPE_NOWHERE, | |
6a31d2a9 | 140 | }, |
1da177e4 LT |
141 | }; |
142 | ||
1da177e4 | 143 | /* Release a nexthop info record */ |
19c1ea14 YZ |
144 | static void free_fib_info_rcu(struct rcu_head *head) |
145 | { | |
146 | struct fib_info *fi = container_of(head, struct fib_info, rcu); | |
147 | ||
e49cc0da YZ |
148 | change_nexthops(fi) { |
149 | if (nexthop_nh->nh_dev) | |
150 | dev_put(nexthop_nh->nh_dev); | |
151 | } endfor_nexthops(fi); | |
152 | ||
153 | release_net(fi->fib_net); | |
19c1ea14 YZ |
154 | if (fi->fib_metrics != (u32 *) dst_default_metrics) |
155 | kfree(fi->fib_metrics); | |
156 | kfree(fi); | |
157 | } | |
1da177e4 LT |
158 | |
159 | void free_fib_info(struct fib_info *fi) | |
160 | { | |
161 | if (fi->fib_dead == 0) { | |
058bd4d2 | 162 | pr_warn("Freeing alive fib_info %p\n", fi); |
1da177e4 LT |
163 | return; |
164 | } | |
1da177e4 | 165 | fib_info_cnt--; |
19c1ea14 | 166 | call_rcu(&fi->rcu, free_fib_info_rcu); |
1da177e4 LT |
167 | } |
168 | ||
169 | void fib_release_info(struct fib_info *fi) | |
170 | { | |
832b4c5e | 171 | spin_lock_bh(&fib_info_lock); |
1da177e4 LT |
172 | if (fi && --fi->fib_treeref == 0) { |
173 | hlist_del(&fi->fib_hash); | |
174 | if (fi->fib_prefsrc) | |
175 | hlist_del(&fi->fib_lhash); | |
176 | change_nexthops(fi) { | |
71fceff0 | 177 | if (!nexthop_nh->nh_dev) |
1da177e4 | 178 | continue; |
71fceff0 | 179 | hlist_del(&nexthop_nh->nh_hash); |
1da177e4 LT |
180 | } endfor_nexthops(fi) |
181 | fi->fib_dead = 1; | |
182 | fib_info_put(fi); | |
183 | } | |
832b4c5e | 184 | spin_unlock_bh(&fib_info_lock); |
1da177e4 LT |
185 | } |
186 | ||
6a31d2a9 | 187 | static inline int nh_comp(const struct fib_info *fi, const struct fib_info *ofi) |
1da177e4 LT |
188 | { |
189 | const struct fib_nh *onh = ofi->fib_nh; | |
190 | ||
191 | for_nexthops(fi) { | |
192 | if (nh->nh_oif != onh->nh_oif || | |
193 | nh->nh_gw != onh->nh_gw || | |
194 | nh->nh_scope != onh->nh_scope || | |
195 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
196 | nh->nh_weight != onh->nh_weight || | |
197 | #endif | |
c7066f70 | 198 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1da177e4 LT |
199 | nh->nh_tclassid != onh->nh_tclassid || |
200 | #endif | |
6a31d2a9 | 201 | ((nh->nh_flags ^ onh->nh_flags) & ~RTNH_F_DEAD)) |
1da177e4 LT |
202 | return -1; |
203 | onh++; | |
204 | } endfor_nexthops(fi); | |
205 | return 0; | |
206 | } | |
207 | ||
88ebc72f DM |
208 | static inline unsigned int fib_devindex_hashfn(unsigned int val) |
209 | { | |
210 | unsigned int mask = DEVINDEX_HASHSIZE - 1; | |
211 | ||
212 | return (val ^ | |
213 | (val >> DEVINDEX_HASHBITS) ^ | |
214 | (val >> (DEVINDEX_HASHBITS * 2))) & mask; | |
215 | } | |
216 | ||
1da177e4 LT |
217 | static inline unsigned int fib_info_hashfn(const struct fib_info *fi) |
218 | { | |
123b9731 | 219 | unsigned int mask = (fib_info_hash_size - 1); |
1da177e4 LT |
220 | unsigned int val = fi->fib_nhs; |
221 | ||
37e826c5 | 222 | val ^= (fi->fib_protocol << 8) | fi->fib_scope; |
81f7bf6c | 223 | val ^= (__force u32)fi->fib_prefsrc; |
1da177e4 | 224 | val ^= fi->fib_priority; |
88ebc72f DM |
225 | for_nexthops(fi) { |
226 | val ^= fib_devindex_hashfn(nh->nh_oif); | |
227 | } endfor_nexthops(fi) | |
1da177e4 LT |
228 | |
229 | return (val ^ (val >> 7) ^ (val >> 12)) & mask; | |
230 | } | |
231 | ||
232 | static struct fib_info *fib_find_info(const struct fib_info *nfi) | |
233 | { | |
234 | struct hlist_head *head; | |
235 | struct hlist_node *node; | |
236 | struct fib_info *fi; | |
237 | unsigned int hash; | |
238 | ||
239 | hash = fib_info_hashfn(nfi); | |
240 | head = &fib_info_hash[hash]; | |
241 | ||
242 | hlist_for_each_entry(fi, node, head, fib_hash) { | |
09ad9bc7 | 243 | if (!net_eq(fi->fib_net, nfi->fib_net)) |
4814bdbd | 244 | continue; |
1da177e4 LT |
245 | if (fi->fib_nhs != nfi->fib_nhs) |
246 | continue; | |
247 | if (nfi->fib_protocol == fi->fib_protocol && | |
37e826c5 | 248 | nfi->fib_scope == fi->fib_scope && |
1da177e4 LT |
249 | nfi->fib_prefsrc == fi->fib_prefsrc && |
250 | nfi->fib_priority == fi->fib_priority && | |
251 | memcmp(nfi->fib_metrics, fi->fib_metrics, | |
fcd13f42 | 252 | sizeof(u32) * RTAX_MAX) == 0 && |
6a31d2a9 | 253 | ((nfi->fib_flags ^ fi->fib_flags) & ~RTNH_F_DEAD) == 0 && |
1da177e4 LT |
254 | (nfi->fib_nhs == 0 || nh_comp(fi, nfi) == 0)) |
255 | return fi; | |
256 | } | |
257 | ||
258 | return NULL; | |
259 | } | |
260 | ||
1da177e4 | 261 | /* Check, that the gateway is already configured. |
6a31d2a9 | 262 | * Used only by redirect accept routine. |
1da177e4 | 263 | */ |
d878e72e | 264 | int ip_fib_check_default(__be32 gw, struct net_device *dev) |
1da177e4 LT |
265 | { |
266 | struct hlist_head *head; | |
267 | struct hlist_node *node; | |
268 | struct fib_nh *nh; | |
269 | unsigned int hash; | |
270 | ||
832b4c5e | 271 | spin_lock(&fib_info_lock); |
1da177e4 LT |
272 | |
273 | hash = fib_devindex_hashfn(dev->ifindex); | |
274 | head = &fib_info_devhash[hash]; | |
275 | hlist_for_each_entry(nh, node, head, nh_hash) { | |
276 | if (nh->nh_dev == dev && | |
277 | nh->nh_gw == gw && | |
6a31d2a9 | 278 | !(nh->nh_flags & RTNH_F_DEAD)) { |
832b4c5e | 279 | spin_unlock(&fib_info_lock); |
1da177e4 LT |
280 | return 0; |
281 | } | |
282 | } | |
283 | ||
832b4c5e | 284 | spin_unlock(&fib_info_lock); |
1da177e4 LT |
285 | |
286 | return -1; | |
287 | } | |
288 | ||
339bf98f TG |
289 | static inline size_t fib_nlmsg_size(struct fib_info *fi) |
290 | { | |
291 | size_t payload = NLMSG_ALIGN(sizeof(struct rtmsg)) | |
292 | + nla_total_size(4) /* RTA_TABLE */ | |
293 | + nla_total_size(4) /* RTA_DST */ | |
294 | + nla_total_size(4) /* RTA_PRIORITY */ | |
295 | + nla_total_size(4); /* RTA_PREFSRC */ | |
296 | ||
297 | /* space for nested metrics */ | |
298 | payload += nla_total_size((RTAX_MAX * nla_total_size(4))); | |
299 | ||
300 | if (fi->fib_nhs) { | |
301 | /* Also handles the special case fib_nhs == 1 */ | |
302 | ||
303 | /* each nexthop is packed in an attribute */ | |
304 | size_t nhsize = nla_total_size(sizeof(struct rtnexthop)); | |
305 | ||
306 | /* may contain flow and gateway attribute */ | |
307 | nhsize += 2 * nla_total_size(4); | |
308 | ||
309 | /* all nexthops are packed in a nested attribute */ | |
310 | payload += nla_total_size(fi->fib_nhs * nhsize); | |
311 | } | |
312 | ||
313 | return payload; | |
314 | } | |
315 | ||
81f7bf6c | 316 | void rtmsg_fib(int event, __be32 key, struct fib_alias *fa, |
b8f55831 MK |
317 | int dst_len, u32 tb_id, struct nl_info *info, |
318 | unsigned int nlm_flags) | |
1da177e4 LT |
319 | { |
320 | struct sk_buff *skb; | |
4e902c57 | 321 | u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0; |
f21c7bc5 | 322 | int err = -ENOBUFS; |
1da177e4 | 323 | |
339bf98f | 324 | skb = nlmsg_new(fib_nlmsg_size(fa->fa_info), GFP_KERNEL); |
f21c7bc5 TG |
325 | if (skb == NULL) |
326 | goto errout; | |
1da177e4 | 327 | |
4e902c57 | 328 | err = fib_dump_info(skb, info->pid, seq, event, tb_id, |
37e826c5 | 329 | fa->fa_type, key, dst_len, |
b8f55831 | 330 | fa->fa_tos, fa->fa_info, nlm_flags); |
26932566 PM |
331 | if (err < 0) { |
332 | /* -EMSGSIZE implies BUG in fib_nlmsg_size() */ | |
333 | WARN_ON(err == -EMSGSIZE); | |
334 | kfree_skb(skb); | |
335 | goto errout; | |
336 | } | |
1ce85fe4 PNA |
337 | rtnl_notify(skb, info->nl_net, info->pid, RTNLGRP_IPV4_ROUTE, |
338 | info->nlh, GFP_KERNEL); | |
339 | return; | |
f21c7bc5 TG |
340 | errout: |
341 | if (err < 0) | |
4d1169c1 | 342 | rtnl_set_sk_err(info->nl_net, RTNLGRP_IPV4_ROUTE, err); |
1da177e4 LT |
343 | } |
344 | ||
345 | /* Return the first fib alias matching TOS with | |
346 | * priority less than or equal to PRIO. | |
347 | */ | |
348 | struct fib_alias *fib_find_alias(struct list_head *fah, u8 tos, u32 prio) | |
349 | { | |
350 | if (fah) { | |
351 | struct fib_alias *fa; | |
352 | list_for_each_entry(fa, fah, fa_list) { | |
353 | if (fa->fa_tos > tos) | |
354 | continue; | |
355 | if (fa->fa_info->fib_priority >= prio || | |
356 | fa->fa_tos < tos) | |
357 | return fa; | |
358 | } | |
359 | } | |
360 | return NULL; | |
361 | } | |
362 | ||
363 | int fib_detect_death(struct fib_info *fi, int order, | |
c17860a0 | 364 | struct fib_info **last_resort, int *last_idx, int dflt) |
1da177e4 LT |
365 | { |
366 | struct neighbour *n; | |
367 | int state = NUD_NONE; | |
368 | ||
369 | n = neigh_lookup(&arp_tbl, &fi->fib_nh[0].nh_gw, fi->fib_dev); | |
370 | if (n) { | |
371 | state = n->nud_state; | |
372 | neigh_release(n); | |
373 | } | |
d9319100 | 374 | if (state == NUD_REACHABLE) |
1da177e4 | 375 | return 0; |
6a31d2a9 | 376 | if ((state & NUD_VALID) && order != dflt) |
1da177e4 | 377 | return 0; |
6a31d2a9 ED |
378 | if ((state & NUD_VALID) || |
379 | (*last_idx < 0 && order > dflt)) { | |
1da177e4 LT |
380 | *last_resort = fi; |
381 | *last_idx = order; | |
382 | } | |
383 | return 1; | |
384 | } | |
385 | ||
386 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
387 | ||
4e902c57 | 388 | static int fib_count_nexthops(struct rtnexthop *rtnh, int remaining) |
1da177e4 LT |
389 | { |
390 | int nhs = 0; | |
1da177e4 | 391 | |
4e902c57 | 392 | while (rtnh_ok(rtnh, remaining)) { |
1da177e4 | 393 | nhs++; |
4e902c57 TG |
394 | rtnh = rtnh_next(rtnh, &remaining); |
395 | } | |
396 | ||
397 | /* leftover implies invalid nexthop configuration, discard it */ | |
398 | return remaining > 0 ? 0 : nhs; | |
1da177e4 LT |
399 | } |
400 | ||
4e902c57 TG |
401 | static int fib_get_nhs(struct fib_info *fi, struct rtnexthop *rtnh, |
402 | int remaining, struct fib_config *cfg) | |
1da177e4 | 403 | { |
1da177e4 | 404 | change_nexthops(fi) { |
4e902c57 TG |
405 | int attrlen; |
406 | ||
407 | if (!rtnh_ok(rtnh, remaining)) | |
1da177e4 | 408 | return -EINVAL; |
4e902c57 | 409 | |
71fceff0 DM |
410 | nexthop_nh->nh_flags = |
411 | (cfg->fc_flags & ~0xFF) | rtnh->rtnh_flags; | |
412 | nexthop_nh->nh_oif = rtnh->rtnh_ifindex; | |
413 | nexthop_nh->nh_weight = rtnh->rtnh_hops + 1; | |
4e902c57 TG |
414 | |
415 | attrlen = rtnh_attrlen(rtnh); | |
416 | if (attrlen > 0) { | |
417 | struct nlattr *nla, *attrs = rtnh_attrs(rtnh); | |
418 | ||
419 | nla = nla_find(attrs, attrlen, RTA_GATEWAY); | |
71fceff0 | 420 | nexthop_nh->nh_gw = nla ? nla_get_be32(nla) : 0; |
c7066f70 | 421 | #ifdef CONFIG_IP_ROUTE_CLASSID |
4e902c57 | 422 | nla = nla_find(attrs, attrlen, RTA_FLOW); |
71fceff0 | 423 | nexthop_nh->nh_tclassid = nla ? nla_get_u32(nla) : 0; |
1da177e4 LT |
424 | #endif |
425 | } | |
4e902c57 TG |
426 | |
427 | rtnh = rtnh_next(rtnh, &remaining); | |
1da177e4 | 428 | } endfor_nexthops(fi); |
4e902c57 | 429 | |
1da177e4 LT |
430 | return 0; |
431 | } | |
432 | ||
433 | #endif | |
434 | ||
4e902c57 | 435 | int fib_nh_match(struct fib_config *cfg, struct fib_info *fi) |
1da177e4 LT |
436 | { |
437 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
4e902c57 TG |
438 | struct rtnexthop *rtnh; |
439 | int remaining; | |
1da177e4 LT |
440 | #endif |
441 | ||
4e902c57 | 442 | if (cfg->fc_priority && cfg->fc_priority != fi->fib_priority) |
1da177e4 LT |
443 | return 1; |
444 | ||
4e902c57 TG |
445 | if (cfg->fc_oif || cfg->fc_gw) { |
446 | if ((!cfg->fc_oif || cfg->fc_oif == fi->fib_nh->nh_oif) && | |
447 | (!cfg->fc_gw || cfg->fc_gw == fi->fib_nh->nh_gw)) | |
1da177e4 LT |
448 | return 0; |
449 | return 1; | |
450 | } | |
451 | ||
452 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
4e902c57 | 453 | if (cfg->fc_mp == NULL) |
1da177e4 | 454 | return 0; |
4e902c57 TG |
455 | |
456 | rtnh = cfg->fc_mp; | |
457 | remaining = cfg->fc_mp_len; | |
e905a9ed | 458 | |
1da177e4 | 459 | for_nexthops(fi) { |
4e902c57 | 460 | int attrlen; |
1da177e4 | 461 | |
4e902c57 | 462 | if (!rtnh_ok(rtnh, remaining)) |
1da177e4 | 463 | return -EINVAL; |
4e902c57 TG |
464 | |
465 | if (rtnh->rtnh_ifindex && rtnh->rtnh_ifindex != nh->nh_oif) | |
1da177e4 | 466 | return 1; |
4e902c57 TG |
467 | |
468 | attrlen = rtnh_attrlen(rtnh); | |
469 | if (attrlen < 0) { | |
470 | struct nlattr *nla, *attrs = rtnh_attrs(rtnh); | |
471 | ||
472 | nla = nla_find(attrs, attrlen, RTA_GATEWAY); | |
17fb2c64 | 473 | if (nla && nla_get_be32(nla) != nh->nh_gw) |
1da177e4 | 474 | return 1; |
c7066f70 | 475 | #ifdef CONFIG_IP_ROUTE_CLASSID |
4e902c57 TG |
476 | nla = nla_find(attrs, attrlen, RTA_FLOW); |
477 | if (nla && nla_get_u32(nla) != nh->nh_tclassid) | |
1da177e4 LT |
478 | return 1; |
479 | #endif | |
480 | } | |
4e902c57 TG |
481 | |
482 | rtnh = rtnh_next(rtnh, &remaining); | |
1da177e4 LT |
483 | } endfor_nexthops(fi); |
484 | #endif | |
485 | return 0; | |
486 | } | |
487 | ||
488 | ||
489 | /* | |
6a31d2a9 ED |
490 | * Picture |
491 | * ------- | |
492 | * | |
493 | * Semantics of nexthop is very messy by historical reasons. | |
494 | * We have to take into account, that: | |
495 | * a) gateway can be actually local interface address, | |
496 | * so that gatewayed route is direct. | |
497 | * b) gateway must be on-link address, possibly | |
498 | * described not by an ifaddr, but also by a direct route. | |
499 | * c) If both gateway and interface are specified, they should not | |
500 | * contradict. | |
501 | * d) If we use tunnel routes, gateway could be not on-link. | |
502 | * | |
503 | * Attempt to reconcile all of these (alas, self-contradictory) conditions | |
504 | * results in pretty ugly and hairy code with obscure logic. | |
505 | * | |
506 | * I chose to generalized it instead, so that the size | |
507 | * of code does not increase practically, but it becomes | |
508 | * much more general. | |
509 | * Every prefix is assigned a "scope" value: "host" is local address, | |
510 | * "link" is direct route, | |
511 | * [ ... "site" ... "interior" ... ] | |
512 | * and "universe" is true gateway route with global meaning. | |
513 | * | |
514 | * Every prefix refers to a set of "nexthop"s (gw, oif), | |
515 | * where gw must have narrower scope. This recursion stops | |
516 | * when gw has LOCAL scope or if "nexthop" is declared ONLINK, | |
517 | * which means that gw is forced to be on link. | |
518 | * | |
519 | * Code is still hairy, but now it is apparently logically | |
520 | * consistent and very flexible. F.e. as by-product it allows | |
521 | * to co-exists in peace independent exterior and interior | |
522 | * routing processes. | |
523 | * | |
524 | * Normally it looks as following. | |
525 | * | |
526 | * {universe prefix} -> (gw, oif) [scope link] | |
527 | * | | |
528 | * |-> {link prefix} -> (gw, oif) [scope local] | |
529 | * | | |
530 | * |-> {local prefix} (terminal node) | |
1da177e4 | 531 | */ |
4e902c57 TG |
532 | static int fib_check_nh(struct fib_config *cfg, struct fib_info *fi, |
533 | struct fib_nh *nh) | |
1da177e4 LT |
534 | { |
535 | int err; | |
86167a37 | 536 | struct net *net; |
6a31d2a9 | 537 | struct net_device *dev; |
1da177e4 | 538 | |
86167a37 | 539 | net = cfg->fc_nlinfo.nl_net; |
1da177e4 LT |
540 | if (nh->nh_gw) { |
541 | struct fib_result res; | |
542 | ||
6a31d2a9 | 543 | if (nh->nh_flags & RTNH_F_ONLINK) { |
1da177e4 | 544 | |
4e902c57 | 545 | if (cfg->fc_scope >= RT_SCOPE_LINK) |
1da177e4 | 546 | return -EINVAL; |
86167a37 | 547 | if (inet_addr_type(net, nh->nh_gw) != RTN_UNICAST) |
1da177e4 | 548 | return -EINVAL; |
6a31d2a9 ED |
549 | dev = __dev_get_by_index(net, nh->nh_oif); |
550 | if (!dev) | |
1da177e4 | 551 | return -ENODEV; |
6a31d2a9 | 552 | if (!(dev->flags & IFF_UP)) |
1da177e4 LT |
553 | return -ENETDOWN; |
554 | nh->nh_dev = dev; | |
555 | dev_hold(dev); | |
556 | nh->nh_scope = RT_SCOPE_LINK; | |
557 | return 0; | |
558 | } | |
ebc0ffae | 559 | rcu_read_lock(); |
1da177e4 | 560 | { |
9ade2286 DM |
561 | struct flowi4 fl4 = { |
562 | .daddr = nh->nh_gw, | |
563 | .flowi4_scope = cfg->fc_scope + 1, | |
564 | .flowi4_oif = nh->nh_oif, | |
4e902c57 | 565 | }; |
1da177e4 LT |
566 | |
567 | /* It is not necessary, but requires a bit of thinking */ | |
9ade2286 DM |
568 | if (fl4.flowi4_scope < RT_SCOPE_LINK) |
569 | fl4.flowi4_scope = RT_SCOPE_LINK; | |
570 | err = fib_lookup(net, &fl4, &res); | |
ebc0ffae ED |
571 | if (err) { |
572 | rcu_read_unlock(); | |
1da177e4 | 573 | return err; |
ebc0ffae | 574 | } |
1da177e4 LT |
575 | } |
576 | err = -EINVAL; | |
577 | if (res.type != RTN_UNICAST && res.type != RTN_LOCAL) | |
578 | goto out; | |
579 | nh->nh_scope = res.scope; | |
580 | nh->nh_oif = FIB_RES_OIF(res); | |
6a31d2a9 ED |
581 | nh->nh_dev = dev = FIB_RES_DEV(res); |
582 | if (!dev) | |
1da177e4 | 583 | goto out; |
6a31d2a9 | 584 | dev_hold(dev); |
8723e1b4 | 585 | err = (dev->flags & IFF_UP) ? 0 : -ENETDOWN; |
1da177e4 LT |
586 | } else { |
587 | struct in_device *in_dev; | |
588 | ||
6a31d2a9 | 589 | if (nh->nh_flags & (RTNH_F_PERVASIVE | RTNH_F_ONLINK)) |
1da177e4 LT |
590 | return -EINVAL; |
591 | ||
8723e1b4 ED |
592 | rcu_read_lock(); |
593 | err = -ENODEV; | |
86167a37 | 594 | in_dev = inetdev_by_index(net, nh->nh_oif); |
1da177e4 | 595 | if (in_dev == NULL) |
8723e1b4 ED |
596 | goto out; |
597 | err = -ENETDOWN; | |
598 | if (!(in_dev->dev->flags & IFF_UP)) | |
599 | goto out; | |
1da177e4 LT |
600 | nh->nh_dev = in_dev->dev; |
601 | dev_hold(nh->nh_dev); | |
602 | nh->nh_scope = RT_SCOPE_HOST; | |
8723e1b4 | 603 | err = 0; |
1da177e4 | 604 | } |
8723e1b4 ED |
605 | out: |
606 | rcu_read_unlock(); | |
607 | return err; | |
1da177e4 LT |
608 | } |
609 | ||
81f7bf6c | 610 | static inline unsigned int fib_laddr_hashfn(__be32 val) |
1da177e4 | 611 | { |
123b9731 | 612 | unsigned int mask = (fib_info_hash_size - 1); |
1da177e4 | 613 | |
6a31d2a9 ED |
614 | return ((__force u32)val ^ |
615 | ((__force u32)val >> 7) ^ | |
616 | ((__force u32)val >> 14)) & mask; | |
1da177e4 LT |
617 | } |
618 | ||
123b9731 | 619 | static struct hlist_head *fib_info_hash_alloc(int bytes) |
1da177e4 LT |
620 | { |
621 | if (bytes <= PAGE_SIZE) | |
88f83491 | 622 | return kzalloc(bytes, GFP_KERNEL); |
1da177e4 LT |
623 | else |
624 | return (struct hlist_head *) | |
6a31d2a9 ED |
625 | __get_free_pages(GFP_KERNEL | __GFP_ZERO, |
626 | get_order(bytes)); | |
1da177e4 LT |
627 | } |
628 | ||
123b9731 | 629 | static void fib_info_hash_free(struct hlist_head *hash, int bytes) |
1da177e4 LT |
630 | { |
631 | if (!hash) | |
632 | return; | |
633 | ||
634 | if (bytes <= PAGE_SIZE) | |
635 | kfree(hash); | |
636 | else | |
637 | free_pages((unsigned long) hash, get_order(bytes)); | |
638 | } | |
639 | ||
123b9731 DM |
640 | static void fib_info_hash_move(struct hlist_head *new_info_hash, |
641 | struct hlist_head *new_laddrhash, | |
642 | unsigned int new_size) | |
1da177e4 | 643 | { |
b7656e7f | 644 | struct hlist_head *old_info_hash, *old_laddrhash; |
123b9731 | 645 | unsigned int old_size = fib_info_hash_size; |
b7656e7f | 646 | unsigned int i, bytes; |
1da177e4 | 647 | |
832b4c5e | 648 | spin_lock_bh(&fib_info_lock); |
b7656e7f DM |
649 | old_info_hash = fib_info_hash; |
650 | old_laddrhash = fib_info_laddrhash; | |
123b9731 | 651 | fib_info_hash_size = new_size; |
1da177e4 LT |
652 | |
653 | for (i = 0; i < old_size; i++) { | |
654 | struct hlist_head *head = &fib_info_hash[i]; | |
655 | struct hlist_node *node, *n; | |
656 | struct fib_info *fi; | |
657 | ||
658 | hlist_for_each_entry_safe(fi, node, n, head, fib_hash) { | |
659 | struct hlist_head *dest; | |
660 | unsigned int new_hash; | |
661 | ||
662 | hlist_del(&fi->fib_hash); | |
663 | ||
664 | new_hash = fib_info_hashfn(fi); | |
665 | dest = &new_info_hash[new_hash]; | |
666 | hlist_add_head(&fi->fib_hash, dest); | |
667 | } | |
668 | } | |
669 | fib_info_hash = new_info_hash; | |
670 | ||
671 | for (i = 0; i < old_size; i++) { | |
672 | struct hlist_head *lhead = &fib_info_laddrhash[i]; | |
673 | struct hlist_node *node, *n; | |
674 | struct fib_info *fi; | |
675 | ||
676 | hlist_for_each_entry_safe(fi, node, n, lhead, fib_lhash) { | |
677 | struct hlist_head *ldest; | |
678 | unsigned int new_hash; | |
679 | ||
680 | hlist_del(&fi->fib_lhash); | |
681 | ||
682 | new_hash = fib_laddr_hashfn(fi->fib_prefsrc); | |
683 | ldest = &new_laddrhash[new_hash]; | |
684 | hlist_add_head(&fi->fib_lhash, ldest); | |
685 | } | |
686 | } | |
687 | fib_info_laddrhash = new_laddrhash; | |
688 | ||
832b4c5e | 689 | spin_unlock_bh(&fib_info_lock); |
b7656e7f DM |
690 | |
691 | bytes = old_size * sizeof(struct hlist_head *); | |
123b9731 DM |
692 | fib_info_hash_free(old_info_hash, bytes); |
693 | fib_info_hash_free(old_laddrhash, bytes); | |
1da177e4 LT |
694 | } |
695 | ||
436c3b66 DM |
696 | __be32 fib_info_update_nh_saddr(struct net *net, struct fib_nh *nh) |
697 | { | |
698 | nh->nh_saddr = inet_select_addr(nh->nh_dev, | |
699 | nh->nh_gw, | |
37e826c5 | 700 | nh->nh_parent->fib_scope); |
436c3b66 DM |
701 | nh->nh_saddr_genid = atomic_read(&net->ipv4.dev_addr_genid); |
702 | ||
703 | return nh->nh_saddr; | |
704 | } | |
705 | ||
4e902c57 | 706 | struct fib_info *fib_create_info(struct fib_config *cfg) |
1da177e4 LT |
707 | { |
708 | int err; | |
709 | struct fib_info *fi = NULL; | |
710 | struct fib_info *ofi; | |
1da177e4 | 711 | int nhs = 1; |
7462bd74 | 712 | struct net *net = cfg->fc_nlinfo.nl_net; |
1da177e4 | 713 | |
4c8237cd DM |
714 | if (cfg->fc_type > RTN_MAX) |
715 | goto err_inval; | |
716 | ||
1da177e4 | 717 | /* Fast check to catch the most weird cases */ |
4e902c57 | 718 | if (fib_props[cfg->fc_type].scope > cfg->fc_scope) |
1da177e4 LT |
719 | goto err_inval; |
720 | ||
721 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
4e902c57 TG |
722 | if (cfg->fc_mp) { |
723 | nhs = fib_count_nexthops(cfg->fc_mp, cfg->fc_mp_len); | |
1da177e4 LT |
724 | if (nhs == 0) |
725 | goto err_inval; | |
726 | } | |
727 | #endif | |
1da177e4 LT |
728 | |
729 | err = -ENOBUFS; | |
123b9731 DM |
730 | if (fib_info_cnt >= fib_info_hash_size) { |
731 | unsigned int new_size = fib_info_hash_size << 1; | |
1da177e4 LT |
732 | struct hlist_head *new_info_hash; |
733 | struct hlist_head *new_laddrhash; | |
734 | unsigned int bytes; | |
735 | ||
736 | if (!new_size) | |
737 | new_size = 1; | |
738 | bytes = new_size * sizeof(struct hlist_head *); | |
123b9731 DM |
739 | new_info_hash = fib_info_hash_alloc(bytes); |
740 | new_laddrhash = fib_info_hash_alloc(bytes); | |
1da177e4 | 741 | if (!new_info_hash || !new_laddrhash) { |
123b9731 DM |
742 | fib_info_hash_free(new_info_hash, bytes); |
743 | fib_info_hash_free(new_laddrhash, bytes); | |
88f83491 | 744 | } else |
123b9731 | 745 | fib_info_hash_move(new_info_hash, new_laddrhash, new_size); |
1da177e4 | 746 | |
123b9731 | 747 | if (!fib_info_hash_size) |
1da177e4 LT |
748 | goto failure; |
749 | } | |
750 | ||
0da974f4 | 751 | fi = kzalloc(sizeof(*fi)+nhs*sizeof(struct fib_nh), GFP_KERNEL); |
1da177e4 LT |
752 | if (fi == NULL) |
753 | goto failure; | |
725d1e1b DM |
754 | if (cfg->fc_mx) { |
755 | fi->fib_metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL); | |
756 | if (!fi->fib_metrics) | |
757 | goto failure; | |
758 | } else | |
759 | fi->fib_metrics = (u32 *) dst_default_metrics; | |
1da177e4 | 760 | fib_info_cnt++; |
1da177e4 | 761 | |
57d7a600 | 762 | fi->fib_net = hold_net(net); |
4e902c57 | 763 | fi->fib_protocol = cfg->fc_protocol; |
37e826c5 | 764 | fi->fib_scope = cfg->fc_scope; |
4e902c57 TG |
765 | fi->fib_flags = cfg->fc_flags; |
766 | fi->fib_priority = cfg->fc_priority; | |
767 | fi->fib_prefsrc = cfg->fc_prefsrc; | |
1da177e4 LT |
768 | |
769 | fi->fib_nhs = nhs; | |
770 | change_nexthops(fi) { | |
71fceff0 | 771 | nexthop_nh->nh_parent = fi; |
1da177e4 LT |
772 | } endfor_nexthops(fi) |
773 | ||
4e902c57 TG |
774 | if (cfg->fc_mx) { |
775 | struct nlattr *nla; | |
776 | int remaining; | |
777 | ||
778 | nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) { | |
8f4c1f9b | 779 | int type = nla_type(nla); |
4e902c57 TG |
780 | |
781 | if (type) { | |
6fac2625 DM |
782 | u32 val; |
783 | ||
4e902c57 | 784 | if (type > RTAX_MAX) |
1da177e4 | 785 | goto err_inval; |
6fac2625 DM |
786 | val = nla_get_u32(nla); |
787 | if (type == RTAX_ADVMSS && val > 65535 - 40) | |
788 | val = 65535 - 40; | |
789 | fi->fib_metrics[type - 1] = val; | |
1da177e4 | 790 | } |
1da177e4 LT |
791 | } |
792 | } | |
1da177e4 | 793 | |
4e902c57 | 794 | if (cfg->fc_mp) { |
1da177e4 | 795 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
4e902c57 TG |
796 | err = fib_get_nhs(fi, cfg->fc_mp, cfg->fc_mp_len, cfg); |
797 | if (err != 0) | |
1da177e4 | 798 | goto failure; |
4e902c57 | 799 | if (cfg->fc_oif && fi->fib_nh->nh_oif != cfg->fc_oif) |
1da177e4 | 800 | goto err_inval; |
4e902c57 | 801 | if (cfg->fc_gw && fi->fib_nh->nh_gw != cfg->fc_gw) |
1da177e4 | 802 | goto err_inval; |
c7066f70 | 803 | #ifdef CONFIG_IP_ROUTE_CLASSID |
4e902c57 | 804 | if (cfg->fc_flow && fi->fib_nh->nh_tclassid != cfg->fc_flow) |
1da177e4 LT |
805 | goto err_inval; |
806 | #endif | |
807 | #else | |
808 | goto err_inval; | |
809 | #endif | |
810 | } else { | |
811 | struct fib_nh *nh = fi->fib_nh; | |
4e902c57 TG |
812 | |
813 | nh->nh_oif = cfg->fc_oif; | |
814 | nh->nh_gw = cfg->fc_gw; | |
815 | nh->nh_flags = cfg->fc_flags; | |
c7066f70 | 816 | #ifdef CONFIG_IP_ROUTE_CLASSID |
4e902c57 | 817 | nh->nh_tclassid = cfg->fc_flow; |
1da177e4 | 818 | #endif |
1da177e4 LT |
819 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
820 | nh->nh_weight = 1; | |
821 | #endif | |
822 | } | |
823 | ||
4e902c57 TG |
824 | if (fib_props[cfg->fc_type].error) { |
825 | if (cfg->fc_gw || cfg->fc_oif || cfg->fc_mp) | |
1da177e4 LT |
826 | goto err_inval; |
827 | goto link_it; | |
4c8237cd DM |
828 | } else { |
829 | switch (cfg->fc_type) { | |
830 | case RTN_UNICAST: | |
831 | case RTN_LOCAL: | |
832 | case RTN_BROADCAST: | |
833 | case RTN_ANYCAST: | |
834 | case RTN_MULTICAST: | |
835 | break; | |
836 | default: | |
837 | goto err_inval; | |
838 | } | |
1da177e4 LT |
839 | } |
840 | ||
4e902c57 | 841 | if (cfg->fc_scope > RT_SCOPE_HOST) |
1da177e4 LT |
842 | goto err_inval; |
843 | ||
4e902c57 | 844 | if (cfg->fc_scope == RT_SCOPE_HOST) { |
1da177e4 LT |
845 | struct fib_nh *nh = fi->fib_nh; |
846 | ||
847 | /* Local address is added. */ | |
848 | if (nhs != 1 || nh->nh_gw) | |
849 | goto err_inval; | |
850 | nh->nh_scope = RT_SCOPE_NOWHERE; | |
7462bd74 | 851 | nh->nh_dev = dev_get_by_index(net, fi->fib_nh->nh_oif); |
1da177e4 LT |
852 | err = -ENODEV; |
853 | if (nh->nh_dev == NULL) | |
854 | goto failure; | |
855 | } else { | |
856 | change_nexthops(fi) { | |
6a31d2a9 ED |
857 | err = fib_check_nh(cfg, fi, nexthop_nh); |
858 | if (err != 0) | |
1da177e4 LT |
859 | goto failure; |
860 | } endfor_nexthops(fi) | |
861 | } | |
862 | ||
863 | if (fi->fib_prefsrc) { | |
4e902c57 TG |
864 | if (cfg->fc_type != RTN_LOCAL || !cfg->fc_dst || |
865 | fi->fib_prefsrc != cfg->fc_dst) | |
7462bd74 | 866 | if (inet_addr_type(net, fi->fib_prefsrc) != RTN_LOCAL) |
1da177e4 LT |
867 | goto err_inval; |
868 | } | |
869 | ||
1fc050a1 | 870 | change_nexthops(fi) { |
436c3b66 | 871 | fib_info_update_nh_saddr(net, nexthop_nh); |
1fc050a1 DM |
872 | } endfor_nexthops(fi) |
873 | ||
1da177e4 | 874 | link_it: |
6a31d2a9 ED |
875 | ofi = fib_find_info(fi); |
876 | if (ofi) { | |
1da177e4 LT |
877 | fi->fib_dead = 1; |
878 | free_fib_info(fi); | |
879 | ofi->fib_treeref++; | |
880 | return ofi; | |
881 | } | |
882 | ||
883 | fi->fib_treeref++; | |
884 | atomic_inc(&fi->fib_clntref); | |
832b4c5e | 885 | spin_lock_bh(&fib_info_lock); |
1da177e4 LT |
886 | hlist_add_head(&fi->fib_hash, |
887 | &fib_info_hash[fib_info_hashfn(fi)]); | |
888 | if (fi->fib_prefsrc) { | |
889 | struct hlist_head *head; | |
890 | ||
891 | head = &fib_info_laddrhash[fib_laddr_hashfn(fi->fib_prefsrc)]; | |
892 | hlist_add_head(&fi->fib_lhash, head); | |
893 | } | |
894 | change_nexthops(fi) { | |
895 | struct hlist_head *head; | |
896 | unsigned int hash; | |
897 | ||
71fceff0 | 898 | if (!nexthop_nh->nh_dev) |
1da177e4 | 899 | continue; |
71fceff0 | 900 | hash = fib_devindex_hashfn(nexthop_nh->nh_dev->ifindex); |
1da177e4 | 901 | head = &fib_info_devhash[hash]; |
71fceff0 | 902 | hlist_add_head(&nexthop_nh->nh_hash, head); |
1da177e4 | 903 | } endfor_nexthops(fi) |
832b4c5e | 904 | spin_unlock_bh(&fib_info_lock); |
1da177e4 LT |
905 | return fi; |
906 | ||
907 | err_inval: | |
908 | err = -EINVAL; | |
909 | ||
910 | failure: | |
e905a9ed | 911 | if (fi) { |
1da177e4 LT |
912 | fi->fib_dead = 1; |
913 | free_fib_info(fi); | |
914 | } | |
4e902c57 TG |
915 | |
916 | return ERR_PTR(err); | |
1da177e4 LT |
917 | } |
918 | ||
be403ea1 | 919 | int fib_dump_info(struct sk_buff *skb, u32 pid, u32 seq, int event, |
37e826c5 | 920 | u32 tb_id, u8 type, __be32 dst, int dst_len, u8 tos, |
be403ea1 | 921 | struct fib_info *fi, unsigned int flags) |
1da177e4 | 922 | { |
be403ea1 | 923 | struct nlmsghdr *nlh; |
1da177e4 | 924 | struct rtmsg *rtm; |
1da177e4 | 925 | |
be403ea1 TG |
926 | nlh = nlmsg_put(skb, pid, seq, event, sizeof(*rtm), flags); |
927 | if (nlh == NULL) | |
26932566 | 928 | return -EMSGSIZE; |
be403ea1 TG |
929 | |
930 | rtm = nlmsg_data(nlh); | |
1da177e4 LT |
931 | rtm->rtm_family = AF_INET; |
932 | rtm->rtm_dst_len = dst_len; | |
933 | rtm->rtm_src_len = 0; | |
934 | rtm->rtm_tos = tos; | |
709772e6 KPO |
935 | if (tb_id < 256) |
936 | rtm->rtm_table = tb_id; | |
937 | else | |
938 | rtm->rtm_table = RT_TABLE_COMPAT; | |
f3756b79 DM |
939 | if (nla_put_u32(skb, RTA_TABLE, tb_id)) |
940 | goto nla_put_failure; | |
1da177e4 LT |
941 | rtm->rtm_type = type; |
942 | rtm->rtm_flags = fi->fib_flags; | |
37e826c5 | 943 | rtm->rtm_scope = fi->fib_scope; |
1da177e4 | 944 | rtm->rtm_protocol = fi->fib_protocol; |
be403ea1 | 945 | |
f3756b79 DM |
946 | if (rtm->rtm_dst_len && |
947 | nla_put_be32(skb, RTA_DST, dst)) | |
948 | goto nla_put_failure; | |
949 | if (fi->fib_priority && | |
950 | nla_put_u32(skb, RTA_PRIORITY, fi->fib_priority)) | |
951 | goto nla_put_failure; | |
1da177e4 | 952 | if (rtnetlink_put_metrics(skb, fi->fib_metrics) < 0) |
be403ea1 TG |
953 | goto nla_put_failure; |
954 | ||
f3756b79 DM |
955 | if (fi->fib_prefsrc && |
956 | nla_put_be32(skb, RTA_PREFSRC, fi->fib_prefsrc)) | |
957 | goto nla_put_failure; | |
1da177e4 | 958 | if (fi->fib_nhs == 1) { |
f3756b79 DM |
959 | if (fi->fib_nh->nh_gw && |
960 | nla_put_be32(skb, RTA_GATEWAY, fi->fib_nh->nh_gw)) | |
961 | goto nla_put_failure; | |
962 | if (fi->fib_nh->nh_oif && | |
963 | nla_put_u32(skb, RTA_OIF, fi->fib_nh->nh_oif)) | |
964 | goto nla_put_failure; | |
c7066f70 | 965 | #ifdef CONFIG_IP_ROUTE_CLASSID |
f3756b79 DM |
966 | if (fi->fib_nh[0].nh_tclassid && |
967 | nla_put_u32(skb, RTA_FLOW, fi->fib_nh[0].nh_tclassid)) | |
968 | goto nla_put_failure; | |
8265abc0 | 969 | #endif |
1da177e4 LT |
970 | } |
971 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
972 | if (fi->fib_nhs > 1) { | |
be403ea1 TG |
973 | struct rtnexthop *rtnh; |
974 | struct nlattr *mp; | |
975 | ||
976 | mp = nla_nest_start(skb, RTA_MULTIPATH); | |
977 | if (mp == NULL) | |
978 | goto nla_put_failure; | |
1da177e4 LT |
979 | |
980 | for_nexthops(fi) { | |
be403ea1 TG |
981 | rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh)); |
982 | if (rtnh == NULL) | |
983 | goto nla_put_failure; | |
984 | ||
985 | rtnh->rtnh_flags = nh->nh_flags & 0xFF; | |
986 | rtnh->rtnh_hops = nh->nh_weight - 1; | |
987 | rtnh->rtnh_ifindex = nh->nh_oif; | |
988 | ||
f3756b79 DM |
989 | if (nh->nh_gw && |
990 | nla_put_be32(skb, RTA_GATEWAY, nh->nh_gw)) | |
991 | goto nla_put_failure; | |
c7066f70 | 992 | #ifdef CONFIG_IP_ROUTE_CLASSID |
f3756b79 DM |
993 | if (nh->nh_tclassid && |
994 | nla_put_u32(skb, RTA_FLOW, nh->nh_tclassid)) | |
995 | goto nla_put_failure; | |
8265abc0 | 996 | #endif |
be403ea1 TG |
997 | /* length of rtnetlink header + attributes */ |
998 | rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *) rtnh; | |
1da177e4 | 999 | } endfor_nexthops(fi); |
be403ea1 TG |
1000 | |
1001 | nla_nest_end(skb, mp); | |
1da177e4 LT |
1002 | } |
1003 | #endif | |
be403ea1 | 1004 | return nlmsg_end(skb, nlh); |
1da177e4 | 1005 | |
be403ea1 | 1006 | nla_put_failure: |
26932566 PM |
1007 | nlmsg_cancel(skb, nlh); |
1008 | return -EMSGSIZE; | |
1da177e4 LT |
1009 | } |
1010 | ||
1da177e4 | 1011 | /* |
6a31d2a9 ED |
1012 | * Update FIB if: |
1013 | * - local address disappeared -> we must delete all the entries | |
1014 | * referring to it. | |
1015 | * - device went down -> we must shutdown all nexthops going via it. | |
1da177e4 | 1016 | */ |
4814bdbd | 1017 | int fib_sync_down_addr(struct net *net, __be32 local) |
1da177e4 LT |
1018 | { |
1019 | int ret = 0; | |
85326fa5 DL |
1020 | unsigned int hash = fib_laddr_hashfn(local); |
1021 | struct hlist_head *head = &fib_info_laddrhash[hash]; | |
1022 | struct hlist_node *node; | |
1023 | struct fib_info *fi; | |
1da177e4 | 1024 | |
85326fa5 DL |
1025 | if (fib_info_laddrhash == NULL || local == 0) |
1026 | return 0; | |
1da177e4 | 1027 | |
85326fa5 | 1028 | hlist_for_each_entry(fi, node, head, fib_lhash) { |
09ad9bc7 | 1029 | if (!net_eq(fi->fib_net, net)) |
4814bdbd | 1030 | continue; |
85326fa5 DL |
1031 | if (fi->fib_prefsrc == local) { |
1032 | fi->fib_flags |= RTNH_F_DEAD; | |
1033 | ret++; | |
1da177e4 LT |
1034 | } |
1035 | } | |
85326fa5 DL |
1036 | return ret; |
1037 | } | |
1038 | ||
1039 | int fib_sync_down_dev(struct net_device *dev, int force) | |
1040 | { | |
1041 | int ret = 0; | |
1042 | int scope = RT_SCOPE_NOWHERE; | |
1043 | struct fib_info *prev_fi = NULL; | |
1044 | unsigned int hash = fib_devindex_hashfn(dev->ifindex); | |
1045 | struct hlist_head *head = &fib_info_devhash[hash]; | |
1046 | struct hlist_node *node; | |
1047 | struct fib_nh *nh; | |
1da177e4 | 1048 | |
85326fa5 DL |
1049 | if (force) |
1050 | scope = -1; | |
1da177e4 | 1051 | |
85326fa5 DL |
1052 | hlist_for_each_entry(nh, node, head, nh_hash) { |
1053 | struct fib_info *fi = nh->nh_parent; | |
1054 | int dead; | |
1da177e4 | 1055 | |
85326fa5 DL |
1056 | BUG_ON(!fi->fib_nhs); |
1057 | if (nh->nh_dev != dev || fi == prev_fi) | |
1058 | continue; | |
1059 | prev_fi = fi; | |
1060 | dead = 0; | |
1061 | change_nexthops(fi) { | |
6a31d2a9 | 1062 | if (nexthop_nh->nh_flags & RTNH_F_DEAD) |
85326fa5 | 1063 | dead++; |
71fceff0 DM |
1064 | else if (nexthop_nh->nh_dev == dev && |
1065 | nexthop_nh->nh_scope != scope) { | |
1066 | nexthop_nh->nh_flags |= RTNH_F_DEAD; | |
1da177e4 | 1067 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
85326fa5 | 1068 | spin_lock_bh(&fib_multipath_lock); |
71fceff0 DM |
1069 | fi->fib_power -= nexthop_nh->nh_power; |
1070 | nexthop_nh->nh_power = 0; | |
85326fa5 | 1071 | spin_unlock_bh(&fib_multipath_lock); |
1da177e4 | 1072 | #endif |
85326fa5 DL |
1073 | dead++; |
1074 | } | |
1da177e4 | 1075 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
71fceff0 | 1076 | if (force > 1 && nexthop_nh->nh_dev == dev) { |
85326fa5 DL |
1077 | dead = fi->fib_nhs; |
1078 | break; | |
1da177e4 | 1079 | } |
85326fa5 DL |
1080 | #endif |
1081 | } endfor_nexthops(fi) | |
1082 | if (dead == fi->fib_nhs) { | |
1083 | fi->fib_flags |= RTNH_F_DEAD; | |
1084 | ret++; | |
1da177e4 LT |
1085 | } |
1086 | } | |
1087 | ||
1088 | return ret; | |
1089 | } | |
1090 | ||
0c838ff1 DM |
1091 | /* Must be invoked inside of an RCU protected region. */ |
1092 | void fib_select_default(struct fib_result *res) | |
1093 | { | |
1094 | struct fib_info *fi = NULL, *last_resort = NULL; | |
1095 | struct list_head *fa_head = res->fa_head; | |
1096 | struct fib_table *tb = res->table; | |
1097 | int order = -1, last_idx = -1; | |
1098 | struct fib_alias *fa; | |
1099 | ||
1100 | list_for_each_entry_rcu(fa, fa_head, fa_list) { | |
1101 | struct fib_info *next_fi = fa->fa_info; | |
1102 | ||
37e826c5 | 1103 | if (next_fi->fib_scope != res->scope || |
0c838ff1 DM |
1104 | fa->fa_type != RTN_UNICAST) |
1105 | continue; | |
1106 | ||
1107 | if (next_fi->fib_priority > res->fi->fib_priority) | |
1108 | break; | |
1109 | if (!next_fi->fib_nh[0].nh_gw || | |
1110 | next_fi->fib_nh[0].nh_scope != RT_SCOPE_LINK) | |
1111 | continue; | |
1112 | ||
1113 | fib_alias_accessed(fa); | |
1114 | ||
1115 | if (fi == NULL) { | |
1116 | if (next_fi != res->fi) | |
1117 | break; | |
1118 | } else if (!fib_detect_death(fi, order, &last_resort, | |
1119 | &last_idx, tb->tb_default)) { | |
1120 | fib_result_assign(res, fi); | |
1121 | tb->tb_default = order; | |
1122 | goto out; | |
1123 | } | |
1124 | fi = next_fi; | |
1125 | order++; | |
1126 | } | |
1127 | ||
1128 | if (order <= 0 || fi == NULL) { | |
1129 | tb->tb_default = -1; | |
1130 | goto out; | |
1131 | } | |
1132 | ||
1133 | if (!fib_detect_death(fi, order, &last_resort, &last_idx, | |
1134 | tb->tb_default)) { | |
1135 | fib_result_assign(res, fi); | |
1136 | tb->tb_default = order; | |
1137 | goto out; | |
1138 | } | |
1139 | ||
1140 | if (last_idx >= 0) | |
1141 | fib_result_assign(res, last_resort); | |
1142 | tb->tb_default = last_idx; | |
1143 | out: | |
31d40937 | 1144 | return; |
0c838ff1 DM |
1145 | } |
1146 | ||
1da177e4 LT |
1147 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
1148 | ||
1149 | /* | |
6a31d2a9 ED |
1150 | * Dead device goes up. We wake up dead nexthops. |
1151 | * It takes sense only on multipath routes. | |
1da177e4 | 1152 | */ |
1da177e4 LT |
1153 | int fib_sync_up(struct net_device *dev) |
1154 | { | |
1155 | struct fib_info *prev_fi; | |
1156 | unsigned int hash; | |
1157 | struct hlist_head *head; | |
1158 | struct hlist_node *node; | |
1159 | struct fib_nh *nh; | |
1160 | int ret; | |
1161 | ||
6a31d2a9 | 1162 | if (!(dev->flags & IFF_UP)) |
1da177e4 LT |
1163 | return 0; |
1164 | ||
1165 | prev_fi = NULL; | |
1166 | hash = fib_devindex_hashfn(dev->ifindex); | |
1167 | head = &fib_info_devhash[hash]; | |
1168 | ret = 0; | |
1169 | ||
1170 | hlist_for_each_entry(nh, node, head, nh_hash) { | |
1171 | struct fib_info *fi = nh->nh_parent; | |
1172 | int alive; | |
1173 | ||
1174 | BUG_ON(!fi->fib_nhs); | |
1175 | if (nh->nh_dev != dev || fi == prev_fi) | |
1176 | continue; | |
1177 | ||
1178 | prev_fi = fi; | |
1179 | alive = 0; | |
1180 | change_nexthops(fi) { | |
6a31d2a9 | 1181 | if (!(nexthop_nh->nh_flags & RTNH_F_DEAD)) { |
1da177e4 LT |
1182 | alive++; |
1183 | continue; | |
1184 | } | |
71fceff0 | 1185 | if (nexthop_nh->nh_dev == NULL || |
6a31d2a9 | 1186 | !(nexthop_nh->nh_dev->flags & IFF_UP)) |
1da177e4 | 1187 | continue; |
71fceff0 DM |
1188 | if (nexthop_nh->nh_dev != dev || |
1189 | !__in_dev_get_rtnl(dev)) | |
1da177e4 LT |
1190 | continue; |
1191 | alive++; | |
1192 | spin_lock_bh(&fib_multipath_lock); | |
71fceff0 DM |
1193 | nexthop_nh->nh_power = 0; |
1194 | nexthop_nh->nh_flags &= ~RTNH_F_DEAD; | |
1da177e4 LT |
1195 | spin_unlock_bh(&fib_multipath_lock); |
1196 | } endfor_nexthops(fi) | |
1197 | ||
1198 | if (alive > 0) { | |
1199 | fi->fib_flags &= ~RTNH_F_DEAD; | |
1200 | ret++; | |
1201 | } | |
1202 | } | |
1203 | ||
1204 | return ret; | |
1205 | } | |
1206 | ||
1207 | /* | |
6a31d2a9 ED |
1208 | * The algorithm is suboptimal, but it provides really |
1209 | * fair weighted route distribution. | |
1da177e4 | 1210 | */ |
1b7fe593 | 1211 | void fib_select_multipath(struct fib_result *res) |
1da177e4 LT |
1212 | { |
1213 | struct fib_info *fi = res->fi; | |
1214 | int w; | |
1215 | ||
1216 | spin_lock_bh(&fib_multipath_lock); | |
1217 | if (fi->fib_power <= 0) { | |
1218 | int power = 0; | |
1219 | change_nexthops(fi) { | |
6a31d2a9 | 1220 | if (!(nexthop_nh->nh_flags & RTNH_F_DEAD)) { |
71fceff0 DM |
1221 | power += nexthop_nh->nh_weight; |
1222 | nexthop_nh->nh_power = nexthop_nh->nh_weight; | |
1da177e4 LT |
1223 | } |
1224 | } endfor_nexthops(fi); | |
1225 | fi->fib_power = power; | |
1226 | if (power <= 0) { | |
1227 | spin_unlock_bh(&fib_multipath_lock); | |
1228 | /* Race condition: route has just become dead. */ | |
1229 | res->nh_sel = 0; | |
1230 | return; | |
1231 | } | |
1232 | } | |
1233 | ||
1234 | ||
1235 | /* w should be random number [0..fi->fib_power-1], | |
6a31d2a9 | 1236 | * it is pretty bad approximation. |
1da177e4 LT |
1237 | */ |
1238 | ||
1239 | w = jiffies % fi->fib_power; | |
1240 | ||
1241 | change_nexthops(fi) { | |
6a31d2a9 | 1242 | if (!(nexthop_nh->nh_flags & RTNH_F_DEAD) && |
71fceff0 | 1243 | nexthop_nh->nh_power) { |
6a31d2a9 ED |
1244 | w -= nexthop_nh->nh_power; |
1245 | if (w <= 0) { | |
71fceff0 | 1246 | nexthop_nh->nh_power--; |
1da177e4 LT |
1247 | fi->fib_power--; |
1248 | res->nh_sel = nhsel; | |
1249 | spin_unlock_bh(&fib_multipath_lock); | |
1250 | return; | |
1251 | } | |
1252 | } | |
1253 | } endfor_nexthops(fi); | |
1254 | ||
1255 | /* Race condition: route has just become dead. */ | |
1256 | res->nh_sel = 0; | |
1257 | spin_unlock_bh(&fib_multipath_lock); | |
1258 | } | |
1259 | #endif |