]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
3 | * operating system. INET is implemented using the BSD Socket | |
4 | * interface as the means of communication with the user level. | |
5 | * | |
6 | * IPv4 Forwarding Information Base: FIB frontend. | |
7 | * | |
8 | * Version: $Id: fib_frontend.c,v 1.26 2001/10/31 21:55:54 davem Exp $ | |
9 | * | |
10 | * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> | |
11 | * | |
12 | * This program is free software; you can redistribute it and/or | |
13 | * modify it under the terms of the GNU General Public License | |
14 | * as published by the Free Software Foundation; either version | |
15 | * 2 of the License, or (at your option) any later version. | |
16 | */ | |
17 | ||
1da177e4 LT |
18 | #include <linux/module.h> |
19 | #include <asm/uaccess.h> | |
20 | #include <asm/system.h> | |
21 | #include <linux/bitops.h> | |
4fc268d2 | 22 | #include <linux/capability.h> |
1da177e4 LT |
23 | #include <linux/types.h> |
24 | #include <linux/kernel.h> | |
25 | #include <linux/sched.h> | |
26 | #include <linux/mm.h> | |
27 | #include <linux/string.h> | |
28 | #include <linux/socket.h> | |
29 | #include <linux/sockios.h> | |
30 | #include <linux/errno.h> | |
31 | #include <linux/in.h> | |
32 | #include <linux/inet.h> | |
14c85021 | 33 | #include <linux/inetdevice.h> |
1da177e4 | 34 | #include <linux/netdevice.h> |
1823730f | 35 | #include <linux/if_addr.h> |
1da177e4 LT |
36 | #include <linux/if_arp.h> |
37 | #include <linux/skbuff.h> | |
38 | #include <linux/netlink.h> | |
39 | #include <linux/init.h> | |
1af5a8c4 | 40 | #include <linux/list.h> |
1da177e4 LT |
41 | |
42 | #include <net/ip.h> | |
43 | #include <net/protocol.h> | |
44 | #include <net/route.h> | |
45 | #include <net/tcp.h> | |
46 | #include <net/sock.h> | |
47 | #include <net/icmp.h> | |
48 | #include <net/arp.h> | |
49 | #include <net/ip_fib.h> | |
50 | ||
51 | #define FFprint(a...) printk(KERN_DEBUG a) | |
52 | ||
53 | #ifndef CONFIG_IP_MULTIPLE_TABLES | |
54 | ||
1da177e4 LT |
55 | struct fib_table *ip_fib_local_table; |
56 | struct fib_table *ip_fib_main_table; | |
57 | ||
1af5a8c4 PM |
58 | #define FIB_TABLE_HASHSZ 1 |
59 | static struct hlist_head fib_table_hash[FIB_TABLE_HASHSZ]; | |
1da177e4 | 60 | |
1af5a8c4 | 61 | #else |
1da177e4 | 62 | |
1af5a8c4 PM |
63 | #define FIB_TABLE_HASHSZ 256 |
64 | static struct hlist_head fib_table_hash[FIB_TABLE_HASHSZ]; | |
1da177e4 | 65 | |
1af5a8c4 | 66 | struct fib_table *fib_new_table(u32 id) |
1da177e4 LT |
67 | { |
68 | struct fib_table *tb; | |
1af5a8c4 | 69 | unsigned int h; |
1da177e4 | 70 | |
1af5a8c4 PM |
71 | if (id == 0) |
72 | id = RT_TABLE_MAIN; | |
73 | tb = fib_get_table(id); | |
74 | if (tb) | |
75 | return tb; | |
1da177e4 LT |
76 | tb = fib_hash_init(id); |
77 | if (!tb) | |
78 | return NULL; | |
1af5a8c4 PM |
79 | h = id & (FIB_TABLE_HASHSZ - 1); |
80 | hlist_add_head_rcu(&tb->tb_hlist, &fib_table_hash[h]); | |
1da177e4 LT |
81 | return tb; |
82 | } | |
83 | ||
1af5a8c4 PM |
84 | struct fib_table *fib_get_table(u32 id) |
85 | { | |
86 | struct fib_table *tb; | |
87 | struct hlist_node *node; | |
88 | unsigned int h; | |
1da177e4 | 89 | |
1af5a8c4 PM |
90 | if (id == 0) |
91 | id = RT_TABLE_MAIN; | |
92 | h = id & (FIB_TABLE_HASHSZ - 1); | |
93 | rcu_read_lock(); | |
94 | hlist_for_each_entry_rcu(tb, node, &fib_table_hash[h], tb_hlist) { | |
95 | if (tb->tb_id == id) { | |
96 | rcu_read_unlock(); | |
97 | return tb; | |
98 | } | |
99 | } | |
100 | rcu_read_unlock(); | |
101 | return NULL; | |
102 | } | |
1da177e4 LT |
103 | #endif /* CONFIG_IP_MULTIPLE_TABLES */ |
104 | ||
1da177e4 LT |
105 | static void fib_flush(void) |
106 | { | |
107 | int flushed = 0; | |
1da177e4 | 108 | struct fib_table *tb; |
1af5a8c4 PM |
109 | struct hlist_node *node; |
110 | unsigned int h; | |
1da177e4 | 111 | |
1af5a8c4 PM |
112 | for (h = 0; h < FIB_TABLE_HASHSZ; h++) { |
113 | hlist_for_each_entry(tb, node, &fib_table_hash[h], tb_hlist) | |
114 | flushed += tb->tb_flush(tb); | |
1da177e4 | 115 | } |
1da177e4 LT |
116 | |
117 | if (flushed) | |
118 | rt_cache_flush(-1); | |
119 | } | |
120 | ||
121 | /* | |
122 | * Find the first device with a given source address. | |
123 | */ | |
124 | ||
60cad5da | 125 | struct net_device * ip_dev_find(__be32 addr) |
1da177e4 LT |
126 | { |
127 | struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } }; | |
128 | struct fib_result res; | |
129 | struct net_device *dev = NULL; | |
130 | ||
131 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
132 | res.r = NULL; | |
133 | #endif | |
134 | ||
135 | if (!ip_fib_local_table || | |
136 | ip_fib_local_table->tb_lookup(ip_fib_local_table, &fl, &res)) | |
137 | return NULL; | |
138 | if (res.type != RTN_LOCAL) | |
139 | goto out; | |
140 | dev = FIB_RES_DEV(res); | |
141 | ||
142 | if (dev) | |
143 | dev_hold(dev); | |
144 | out: | |
145 | fib_res_put(&res); | |
146 | return dev; | |
147 | } | |
148 | ||
fd683222 | 149 | unsigned inet_addr_type(__be32 addr) |
1da177e4 LT |
150 | { |
151 | struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } }; | |
152 | struct fib_result res; | |
153 | unsigned ret = RTN_BROADCAST; | |
154 | ||
155 | if (ZERONET(addr) || BADCLASS(addr)) | |
156 | return RTN_BROADCAST; | |
157 | if (MULTICAST(addr)) | |
158 | return RTN_MULTICAST; | |
159 | ||
160 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
161 | res.r = NULL; | |
162 | #endif | |
163 | ||
164 | if (ip_fib_local_table) { | |
165 | ret = RTN_UNICAST; | |
166 | if (!ip_fib_local_table->tb_lookup(ip_fib_local_table, | |
167 | &fl, &res)) { | |
168 | ret = res.type; | |
169 | fib_res_put(&res); | |
170 | } | |
171 | } | |
172 | return ret; | |
173 | } | |
174 | ||
175 | /* Given (packet source, input interface) and optional (dst, oif, tos): | |
176 | - (main) check, that source is valid i.e. not broadcast or our local | |
177 | address. | |
178 | - figure out what "logical" interface this packet arrived | |
179 | and calculate "specific destination" address. | |
180 | - check, that packet arrived from expected physical interface. | |
181 | */ | |
182 | ||
d9c9df8c AV |
183 | int fib_validate_source(__be32 src, __be32 dst, u8 tos, int oif, |
184 | struct net_device *dev, __be32 *spec_dst, u32 *itag) | |
1da177e4 LT |
185 | { |
186 | struct in_device *in_dev; | |
187 | struct flowi fl = { .nl_u = { .ip4_u = | |
188 | { .daddr = src, | |
189 | .saddr = dst, | |
190 | .tos = tos } }, | |
191 | .iif = oif }; | |
192 | struct fib_result res; | |
193 | int no_addr, rpf; | |
194 | int ret; | |
195 | ||
196 | no_addr = rpf = 0; | |
197 | rcu_read_lock(); | |
e5ed6399 | 198 | in_dev = __in_dev_get_rcu(dev); |
1da177e4 LT |
199 | if (in_dev) { |
200 | no_addr = in_dev->ifa_list == NULL; | |
201 | rpf = IN_DEV_RPFILTER(in_dev); | |
202 | } | |
203 | rcu_read_unlock(); | |
204 | ||
205 | if (in_dev == NULL) | |
206 | goto e_inval; | |
207 | ||
208 | if (fib_lookup(&fl, &res)) | |
209 | goto last_resort; | |
210 | if (res.type != RTN_UNICAST) | |
211 | goto e_inval_res; | |
212 | *spec_dst = FIB_RES_PREFSRC(res); | |
213 | fib_combine_itag(itag, &res); | |
214 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
215 | if (FIB_RES_DEV(res) == dev || res.fi->fib_nhs > 1) | |
216 | #else | |
217 | if (FIB_RES_DEV(res) == dev) | |
218 | #endif | |
219 | { | |
220 | ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST; | |
221 | fib_res_put(&res); | |
222 | return ret; | |
223 | } | |
224 | fib_res_put(&res); | |
225 | if (no_addr) | |
226 | goto last_resort; | |
227 | if (rpf) | |
228 | goto e_inval; | |
229 | fl.oif = dev->ifindex; | |
230 | ||
231 | ret = 0; | |
232 | if (fib_lookup(&fl, &res) == 0) { | |
233 | if (res.type == RTN_UNICAST) { | |
234 | *spec_dst = FIB_RES_PREFSRC(res); | |
235 | ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST; | |
236 | } | |
237 | fib_res_put(&res); | |
238 | } | |
239 | return ret; | |
240 | ||
241 | last_resort: | |
242 | if (rpf) | |
243 | goto e_inval; | |
244 | *spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE); | |
245 | *itag = 0; | |
246 | return 0; | |
247 | ||
248 | e_inval_res: | |
249 | fib_res_put(&res); | |
250 | e_inval: | |
251 | return -EINVAL; | |
252 | } | |
253 | ||
254 | #ifndef CONFIG_IP_NOSIOCRT | |
255 | ||
81f7bf6c | 256 | static inline __be32 sk_extract_addr(struct sockaddr *addr) |
4e902c57 TG |
257 | { |
258 | return ((struct sockaddr_in *) addr)->sin_addr.s_addr; | |
259 | } | |
260 | ||
261 | static int put_rtax(struct nlattr *mx, int len, int type, u32 value) | |
262 | { | |
263 | struct nlattr *nla; | |
264 | ||
265 | nla = (struct nlattr *) ((char *) mx + len); | |
266 | nla->nla_type = type; | |
267 | nla->nla_len = nla_attr_size(4); | |
268 | *(u32 *) nla_data(nla) = value; | |
269 | ||
270 | return len + nla_total_size(4); | |
271 | } | |
272 | ||
273 | static int rtentry_to_fib_config(int cmd, struct rtentry *rt, | |
274 | struct fib_config *cfg) | |
275 | { | |
6d85c10a | 276 | __be32 addr; |
4e902c57 TG |
277 | int plen; |
278 | ||
279 | memset(cfg, 0, sizeof(*cfg)); | |
280 | ||
281 | if (rt->rt_dst.sa_family != AF_INET) | |
282 | return -EAFNOSUPPORT; | |
283 | ||
284 | /* | |
285 | * Check mask for validity: | |
286 | * a) it must be contiguous. | |
287 | * b) destination must have all host bits clear. | |
288 | * c) if application forgot to set correct family (AF_INET), | |
289 | * reject request unless it is absolutely clear i.e. | |
290 | * both family and mask are zero. | |
291 | */ | |
292 | plen = 32; | |
293 | addr = sk_extract_addr(&rt->rt_dst); | |
294 | if (!(rt->rt_flags & RTF_HOST)) { | |
81f7bf6c | 295 | __be32 mask = sk_extract_addr(&rt->rt_genmask); |
4e902c57 TG |
296 | |
297 | if (rt->rt_genmask.sa_family != AF_INET) { | |
298 | if (mask || rt->rt_genmask.sa_family) | |
299 | return -EAFNOSUPPORT; | |
300 | } | |
301 | ||
302 | if (bad_mask(mask, addr)) | |
303 | return -EINVAL; | |
304 | ||
305 | plen = inet_mask_len(mask); | |
306 | } | |
307 | ||
308 | cfg->fc_dst_len = plen; | |
309 | cfg->fc_dst = addr; | |
310 | ||
311 | if (cmd != SIOCDELRT) { | |
312 | cfg->fc_nlflags = NLM_F_CREATE; | |
313 | cfg->fc_protocol = RTPROT_BOOT; | |
314 | } | |
315 | ||
316 | if (rt->rt_metric) | |
317 | cfg->fc_priority = rt->rt_metric - 1; | |
318 | ||
319 | if (rt->rt_flags & RTF_REJECT) { | |
320 | cfg->fc_scope = RT_SCOPE_HOST; | |
321 | cfg->fc_type = RTN_UNREACHABLE; | |
322 | return 0; | |
323 | } | |
324 | ||
325 | cfg->fc_scope = RT_SCOPE_NOWHERE; | |
326 | cfg->fc_type = RTN_UNICAST; | |
327 | ||
328 | if (rt->rt_dev) { | |
329 | char *colon; | |
330 | struct net_device *dev; | |
331 | char devname[IFNAMSIZ]; | |
332 | ||
333 | if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1)) | |
334 | return -EFAULT; | |
335 | ||
336 | devname[IFNAMSIZ-1] = 0; | |
337 | colon = strchr(devname, ':'); | |
338 | if (colon) | |
339 | *colon = 0; | |
340 | dev = __dev_get_by_name(devname); | |
341 | if (!dev) | |
342 | return -ENODEV; | |
343 | cfg->fc_oif = dev->ifindex; | |
344 | if (colon) { | |
345 | struct in_ifaddr *ifa; | |
346 | struct in_device *in_dev = __in_dev_get_rtnl(dev); | |
347 | if (!in_dev) | |
348 | return -ENODEV; | |
349 | *colon = ':'; | |
350 | for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) | |
351 | if (strcmp(ifa->ifa_label, devname) == 0) | |
352 | break; | |
353 | if (ifa == NULL) | |
354 | return -ENODEV; | |
355 | cfg->fc_prefsrc = ifa->ifa_local; | |
356 | } | |
357 | } | |
358 | ||
359 | addr = sk_extract_addr(&rt->rt_gateway); | |
360 | if (rt->rt_gateway.sa_family == AF_INET && addr) { | |
361 | cfg->fc_gw = addr; | |
362 | if (rt->rt_flags & RTF_GATEWAY && | |
363 | inet_addr_type(addr) == RTN_UNICAST) | |
364 | cfg->fc_scope = RT_SCOPE_UNIVERSE; | |
365 | } | |
366 | ||
367 | if (cmd == SIOCDELRT) | |
368 | return 0; | |
369 | ||
370 | if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw) | |
371 | return -EINVAL; | |
372 | ||
373 | if (cfg->fc_scope == RT_SCOPE_NOWHERE) | |
374 | cfg->fc_scope = RT_SCOPE_LINK; | |
375 | ||
376 | if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) { | |
377 | struct nlattr *mx; | |
378 | int len = 0; | |
379 | ||
380 | mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL); | |
381 | if (mx == NULL) | |
382 | return -ENOMEM; | |
383 | ||
384 | if (rt->rt_flags & RTF_MTU) | |
385 | len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40); | |
386 | ||
387 | if (rt->rt_flags & RTF_WINDOW) | |
388 | len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window); | |
389 | ||
390 | if (rt->rt_flags & RTF_IRTT) | |
391 | len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3); | |
392 | ||
393 | cfg->fc_mx = mx; | |
394 | cfg->fc_mx_len = len; | |
395 | } | |
396 | ||
397 | return 0; | |
398 | } | |
399 | ||
1da177e4 LT |
400 | /* |
401 | * Handle IP routing ioctl calls. These are used to manipulate the routing tables | |
402 | */ | |
403 | ||
404 | int ip_rt_ioctl(unsigned int cmd, void __user *arg) | |
405 | { | |
4e902c57 TG |
406 | struct fib_config cfg; |
407 | struct rtentry rt; | |
1da177e4 | 408 | int err; |
1da177e4 LT |
409 | |
410 | switch (cmd) { | |
411 | case SIOCADDRT: /* Add a route */ | |
412 | case SIOCDELRT: /* Delete a route */ | |
413 | if (!capable(CAP_NET_ADMIN)) | |
414 | return -EPERM; | |
4e902c57 TG |
415 | |
416 | if (copy_from_user(&rt, arg, sizeof(rt))) | |
1da177e4 | 417 | return -EFAULT; |
4e902c57 | 418 | |
1da177e4 | 419 | rtnl_lock(); |
4e902c57 | 420 | err = rtentry_to_fib_config(cmd, &rt, &cfg); |
1da177e4 | 421 | if (err == 0) { |
4e902c57 TG |
422 | struct fib_table *tb; |
423 | ||
1da177e4 | 424 | if (cmd == SIOCDELRT) { |
4e902c57 | 425 | tb = fib_get_table(cfg.fc_table); |
1da177e4 | 426 | if (tb) |
4e902c57 TG |
427 | err = tb->tb_delete(tb, &cfg); |
428 | else | |
429 | err = -ESRCH; | |
1da177e4 | 430 | } else { |
4e902c57 | 431 | tb = fib_new_table(cfg.fc_table); |
1da177e4 | 432 | if (tb) |
4e902c57 TG |
433 | err = tb->tb_insert(tb, &cfg); |
434 | else | |
435 | err = -ENOBUFS; | |
1da177e4 | 436 | } |
4e902c57 TG |
437 | |
438 | /* allocated by rtentry_to_fib_config() */ | |
439 | kfree(cfg.fc_mx); | |
1da177e4 LT |
440 | } |
441 | rtnl_unlock(); | |
442 | return err; | |
443 | } | |
444 | return -EINVAL; | |
445 | } | |
446 | ||
447 | #else | |
448 | ||
449 | int ip_rt_ioctl(unsigned int cmd, void *arg) | |
450 | { | |
451 | return -EINVAL; | |
452 | } | |
453 | ||
454 | #endif | |
455 | ||
d889ce3b | 456 | struct nla_policy rtm_ipv4_policy[RTA_MAX+1] __read_mostly = { |
4e902c57 TG |
457 | [RTA_DST] = { .type = NLA_U32 }, |
458 | [RTA_SRC] = { .type = NLA_U32 }, | |
459 | [RTA_IIF] = { .type = NLA_U32 }, | |
460 | [RTA_OIF] = { .type = NLA_U32 }, | |
461 | [RTA_GATEWAY] = { .type = NLA_U32 }, | |
462 | [RTA_PRIORITY] = { .type = NLA_U32 }, | |
463 | [RTA_PREFSRC] = { .type = NLA_U32 }, | |
464 | [RTA_METRICS] = { .type = NLA_NESTED }, | |
5176f91e | 465 | [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) }, |
4e902c57 TG |
466 | [RTA_PROTOINFO] = { .type = NLA_U32 }, |
467 | [RTA_FLOW] = { .type = NLA_U32 }, | |
468 | [RTA_MP_ALGO] = { .type = NLA_U32 }, | |
469 | }; | |
470 | ||
471 | static int rtm_to_fib_config(struct sk_buff *skb, struct nlmsghdr *nlh, | |
472 | struct fib_config *cfg) | |
1da177e4 | 473 | { |
4e902c57 TG |
474 | struct nlattr *attr; |
475 | int err, remaining; | |
476 | struct rtmsg *rtm; | |
477 | ||
478 | err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy); | |
479 | if (err < 0) | |
480 | goto errout; | |
481 | ||
482 | memset(cfg, 0, sizeof(*cfg)); | |
483 | ||
484 | rtm = nlmsg_data(nlh); | |
4e902c57 | 485 | cfg->fc_dst_len = rtm->rtm_dst_len; |
4e902c57 TG |
486 | cfg->fc_tos = rtm->rtm_tos; |
487 | cfg->fc_table = rtm->rtm_table; | |
488 | cfg->fc_protocol = rtm->rtm_protocol; | |
489 | cfg->fc_scope = rtm->rtm_scope; | |
490 | cfg->fc_type = rtm->rtm_type; | |
491 | cfg->fc_flags = rtm->rtm_flags; | |
492 | cfg->fc_nlflags = nlh->nlmsg_flags; | |
493 | ||
494 | cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid; | |
495 | cfg->fc_nlinfo.nlh = nlh; | |
496 | ||
497 | nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) { | |
498 | switch (attr->nla_type) { | |
499 | case RTA_DST: | |
17fb2c64 | 500 | cfg->fc_dst = nla_get_be32(attr); |
4e902c57 | 501 | break; |
4e902c57 TG |
502 | case RTA_OIF: |
503 | cfg->fc_oif = nla_get_u32(attr); | |
504 | break; | |
505 | case RTA_GATEWAY: | |
17fb2c64 | 506 | cfg->fc_gw = nla_get_be32(attr); |
4e902c57 TG |
507 | break; |
508 | case RTA_PRIORITY: | |
509 | cfg->fc_priority = nla_get_u32(attr); | |
510 | break; | |
511 | case RTA_PREFSRC: | |
17fb2c64 | 512 | cfg->fc_prefsrc = nla_get_be32(attr); |
4e902c57 TG |
513 | break; |
514 | case RTA_METRICS: | |
515 | cfg->fc_mx = nla_data(attr); | |
516 | cfg->fc_mx_len = nla_len(attr); | |
517 | break; | |
518 | case RTA_MULTIPATH: | |
519 | cfg->fc_mp = nla_data(attr); | |
520 | cfg->fc_mp_len = nla_len(attr); | |
521 | break; | |
522 | case RTA_FLOW: | |
523 | cfg->fc_flow = nla_get_u32(attr); | |
524 | break; | |
525 | case RTA_MP_ALGO: | |
526 | cfg->fc_mp_alg = nla_get_u32(attr); | |
527 | break; | |
528 | case RTA_TABLE: | |
529 | cfg->fc_table = nla_get_u32(attr); | |
530 | break; | |
1da177e4 LT |
531 | } |
532 | } | |
4e902c57 | 533 | |
1da177e4 | 534 | return 0; |
4e902c57 TG |
535 | errout: |
536 | return err; | |
1da177e4 LT |
537 | } |
538 | ||
539 | int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg) | |
540 | { | |
4e902c57 TG |
541 | struct fib_config cfg; |
542 | struct fib_table *tb; | |
543 | int err; | |
1da177e4 | 544 | |
4e902c57 TG |
545 | err = rtm_to_fib_config(skb, nlh, &cfg); |
546 | if (err < 0) | |
547 | goto errout; | |
1da177e4 | 548 | |
4e902c57 TG |
549 | tb = fib_get_table(cfg.fc_table); |
550 | if (tb == NULL) { | |
551 | err = -ESRCH; | |
552 | goto errout; | |
553 | } | |
554 | ||
555 | err = tb->tb_delete(tb, &cfg); | |
556 | errout: | |
557 | return err; | |
1da177e4 LT |
558 | } |
559 | ||
560 | int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg) | |
561 | { | |
4e902c57 TG |
562 | struct fib_config cfg; |
563 | struct fib_table *tb; | |
564 | int err; | |
1da177e4 | 565 | |
4e902c57 TG |
566 | err = rtm_to_fib_config(skb, nlh, &cfg); |
567 | if (err < 0) | |
568 | goto errout; | |
1da177e4 | 569 | |
4e902c57 TG |
570 | tb = fib_new_table(cfg.fc_table); |
571 | if (tb == NULL) { | |
572 | err = -ENOBUFS; | |
573 | goto errout; | |
574 | } | |
575 | ||
576 | err = tb->tb_insert(tb, &cfg); | |
577 | errout: | |
578 | return err; | |
1da177e4 LT |
579 | } |
580 | ||
581 | int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb) | |
582 | { | |
1af5a8c4 PM |
583 | unsigned int h, s_h; |
584 | unsigned int e = 0, s_e; | |
1da177e4 | 585 | struct fib_table *tb; |
1af5a8c4 PM |
586 | struct hlist_node *node; |
587 | int dumped = 0; | |
1da177e4 | 588 | |
be403ea1 TG |
589 | if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) && |
590 | ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED) | |
1da177e4 LT |
591 | return ip_rt_dump(skb, cb); |
592 | ||
1af5a8c4 PM |
593 | s_h = cb->args[0]; |
594 | s_e = cb->args[1]; | |
595 | ||
596 | for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) { | |
597 | e = 0; | |
598 | hlist_for_each_entry(tb, node, &fib_table_hash[h], tb_hlist) { | |
599 | if (e < s_e) | |
600 | goto next; | |
601 | if (dumped) | |
602 | memset(&cb->args[2], 0, sizeof(cb->args) - | |
603 | 2 * sizeof(cb->args[0])); | |
604 | if (tb->tb_dump(tb, skb, cb) < 0) | |
605 | goto out; | |
606 | dumped = 1; | |
607 | next: | |
608 | e++; | |
609 | } | |
1da177e4 | 610 | } |
1af5a8c4 PM |
611 | out: |
612 | cb->args[1] = e; | |
613 | cb->args[0] = h; | |
1da177e4 LT |
614 | |
615 | return skb->len; | |
616 | } | |
617 | ||
618 | /* Prepare and feed intra-kernel routing request. | |
619 | Really, it should be netlink message, but :-( netlink | |
620 | can be not configured, so that we feed it directly | |
621 | to fib engine. It is legal, because all events occur | |
622 | only when netlink is already locked. | |
623 | */ | |
624 | ||
81f7bf6c | 625 | static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa) |
1da177e4 | 626 | { |
4e902c57 TG |
627 | struct fib_table *tb; |
628 | struct fib_config cfg = { | |
629 | .fc_protocol = RTPROT_KERNEL, | |
630 | .fc_type = type, | |
631 | .fc_dst = dst, | |
632 | .fc_dst_len = dst_len, | |
633 | .fc_prefsrc = ifa->ifa_local, | |
634 | .fc_oif = ifa->ifa_dev->dev->ifindex, | |
635 | .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND, | |
636 | }; | |
1da177e4 LT |
637 | |
638 | if (type == RTN_UNICAST) | |
639 | tb = fib_new_table(RT_TABLE_MAIN); | |
640 | else | |
641 | tb = fib_new_table(RT_TABLE_LOCAL); | |
642 | ||
643 | if (tb == NULL) | |
644 | return; | |
645 | ||
4e902c57 | 646 | cfg.fc_table = tb->tb_id; |
1da177e4 | 647 | |
4e902c57 TG |
648 | if (type != RTN_LOCAL) |
649 | cfg.fc_scope = RT_SCOPE_LINK; | |
650 | else | |
651 | cfg.fc_scope = RT_SCOPE_HOST; | |
1da177e4 LT |
652 | |
653 | if (cmd == RTM_NEWROUTE) | |
4e902c57 | 654 | tb->tb_insert(tb, &cfg); |
1da177e4 | 655 | else |
4e902c57 | 656 | tb->tb_delete(tb, &cfg); |
1da177e4 LT |
657 | } |
658 | ||
0ff60a45 | 659 | void fib_add_ifaddr(struct in_ifaddr *ifa) |
1da177e4 LT |
660 | { |
661 | struct in_device *in_dev = ifa->ifa_dev; | |
662 | struct net_device *dev = in_dev->dev; | |
663 | struct in_ifaddr *prim = ifa; | |
a144ea4b AV |
664 | __be32 mask = ifa->ifa_mask; |
665 | __be32 addr = ifa->ifa_local; | |
666 | __be32 prefix = ifa->ifa_address&mask; | |
1da177e4 LT |
667 | |
668 | if (ifa->ifa_flags&IFA_F_SECONDARY) { | |
669 | prim = inet_ifa_byprefix(in_dev, prefix, mask); | |
670 | if (prim == NULL) { | |
671 | printk(KERN_DEBUG "fib_add_ifaddr: bug: prim == NULL\n"); | |
672 | return; | |
673 | } | |
674 | } | |
675 | ||
676 | fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim); | |
677 | ||
678 | if (!(dev->flags&IFF_UP)) | |
679 | return; | |
680 | ||
681 | /* Add broadcast address, if it is explicitly assigned. */ | |
a144ea4b | 682 | if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF)) |
1da177e4 LT |
683 | fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim); |
684 | ||
685 | if (!ZERONET(prefix) && !(ifa->ifa_flags&IFA_F_SECONDARY) && | |
686 | (prefix != addr || ifa->ifa_prefixlen < 32)) { | |
687 | fib_magic(RTM_NEWROUTE, dev->flags&IFF_LOOPBACK ? RTN_LOCAL : | |
688 | RTN_UNICAST, prefix, ifa->ifa_prefixlen, prim); | |
689 | ||
690 | /* Add network specific broadcasts, when it takes a sense */ | |
691 | if (ifa->ifa_prefixlen < 31) { | |
692 | fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim); | |
693 | fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix|~mask, 32, prim); | |
694 | } | |
695 | } | |
696 | } | |
697 | ||
698 | static void fib_del_ifaddr(struct in_ifaddr *ifa) | |
699 | { | |
700 | struct in_device *in_dev = ifa->ifa_dev; | |
701 | struct net_device *dev = in_dev->dev; | |
702 | struct in_ifaddr *ifa1; | |
703 | struct in_ifaddr *prim = ifa; | |
a144ea4b AV |
704 | __be32 brd = ifa->ifa_address|~ifa->ifa_mask; |
705 | __be32 any = ifa->ifa_address&ifa->ifa_mask; | |
1da177e4 LT |
706 | #define LOCAL_OK 1 |
707 | #define BRD_OK 2 | |
708 | #define BRD0_OK 4 | |
709 | #define BRD1_OK 8 | |
710 | unsigned ok = 0; | |
711 | ||
712 | if (!(ifa->ifa_flags&IFA_F_SECONDARY)) | |
713 | fib_magic(RTM_DELROUTE, dev->flags&IFF_LOOPBACK ? RTN_LOCAL : | |
714 | RTN_UNICAST, any, ifa->ifa_prefixlen, prim); | |
715 | else { | |
716 | prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask); | |
717 | if (prim == NULL) { | |
718 | printk(KERN_DEBUG "fib_del_ifaddr: bug: prim == NULL\n"); | |
719 | return; | |
720 | } | |
721 | } | |
722 | ||
723 | /* Deletion is more complicated than add. | |
724 | We should take care of not to delete too much :-) | |
725 | ||
726 | Scan address list to be sure that addresses are really gone. | |
727 | */ | |
728 | ||
729 | for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) { | |
730 | if (ifa->ifa_local == ifa1->ifa_local) | |
731 | ok |= LOCAL_OK; | |
732 | if (ifa->ifa_broadcast == ifa1->ifa_broadcast) | |
733 | ok |= BRD_OK; | |
734 | if (brd == ifa1->ifa_broadcast) | |
735 | ok |= BRD1_OK; | |
736 | if (any == ifa1->ifa_broadcast) | |
737 | ok |= BRD0_OK; | |
738 | } | |
739 | ||
740 | if (!(ok&BRD_OK)) | |
741 | fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim); | |
742 | if (!(ok&BRD1_OK)) | |
743 | fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim); | |
744 | if (!(ok&BRD0_OK)) | |
745 | fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim); | |
746 | if (!(ok&LOCAL_OK)) { | |
747 | fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim); | |
748 | ||
749 | /* Check, that this local address finally disappeared. */ | |
750 | if (inet_addr_type(ifa->ifa_local) != RTN_LOCAL) { | |
751 | /* And the last, but not the least thing. | |
752 | We must flush stray FIB entries. | |
753 | ||
754 | First of all, we scan fib_info list searching | |
755 | for stray nexthop entries, then ignite fib_flush. | |
756 | */ | |
757 | if (fib_sync_down(ifa->ifa_local, NULL, 0)) | |
758 | fib_flush(); | |
759 | } | |
760 | } | |
761 | #undef LOCAL_OK | |
762 | #undef BRD_OK | |
763 | #undef BRD0_OK | |
764 | #undef BRD1_OK | |
765 | } | |
766 | ||
246955fe RO |
767 | static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb ) |
768 | { | |
769 | ||
770 | struct fib_result res; | |
5f300893 | 771 | struct flowi fl = { .mark = frn->fl_mark, |
47dcf0cb | 772 | .nl_u = { .ip4_u = { .daddr = frn->fl_addr, |
246955fe RO |
773 | .tos = frn->fl_tos, |
774 | .scope = frn->fl_scope } } }; | |
775 | if (tb) { | |
776 | local_bh_disable(); | |
777 | ||
778 | frn->tb_id = tb->tb_id; | |
779 | frn->err = tb->tb_lookup(tb, &fl, &res); | |
780 | ||
781 | if (!frn->err) { | |
782 | frn->prefixlen = res.prefixlen; | |
783 | frn->nh_sel = res.nh_sel; | |
784 | frn->type = res.type; | |
785 | frn->scope = res.scope; | |
786 | } | |
787 | local_bh_enable(); | |
788 | } | |
789 | } | |
790 | ||
791 | static void nl_fib_input(struct sock *sk, int len) | |
792 | { | |
793 | struct sk_buff *skb = NULL; | |
794 | struct nlmsghdr *nlh = NULL; | |
795 | struct fib_result_nl *frn; | |
246955fe RO |
796 | u32 pid; |
797 | struct fib_table *tb; | |
798 | ||
ea86575e | 799 | skb = skb_dequeue(&sk->sk_receive_queue); |
246955fe | 800 | nlh = (struct nlmsghdr *)skb->data; |
ea86575e TG |
801 | if (skb->len < NLMSG_SPACE(0) || skb->len < nlh->nlmsg_len || |
802 | nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*frn))) { | |
803 | kfree_skb(skb); | |
804 | return; | |
805 | } | |
246955fe RO |
806 | |
807 | frn = (struct fib_result_nl *) NLMSG_DATA(nlh); | |
808 | tb = fib_get_table(frn->tb_id_in); | |
809 | ||
810 | nl_fib_lookup(frn, tb); | |
811 | ||
812 | pid = nlh->nlmsg_pid; /*pid of sending process */ | |
246955fe | 813 | NETLINK_CB(skb).pid = 0; /* from kernel */ |
ac6d439d | 814 | NETLINK_CB(skb).dst_group = 0; /* unicast */ |
246955fe RO |
815 | netlink_unicast(sk, skb, pid, MSG_DONTWAIT); |
816 | } | |
817 | ||
818 | static void nl_fib_lookup_init(void) | |
819 | { | |
06628607 | 820 | netlink_kernel_create(NETLINK_FIB_LOOKUP, 0, nl_fib_input, THIS_MODULE); |
246955fe RO |
821 | } |
822 | ||
1da177e4 LT |
823 | static void fib_disable_ip(struct net_device *dev, int force) |
824 | { | |
825 | if (fib_sync_down(0, dev, force)) | |
826 | fib_flush(); | |
827 | rt_cache_flush(0); | |
828 | arp_ifdown(dev); | |
829 | } | |
830 | ||
831 | static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr) | |
832 | { | |
833 | struct in_ifaddr *ifa = (struct in_ifaddr*)ptr; | |
834 | ||
835 | switch (event) { | |
836 | case NETDEV_UP: | |
837 | fib_add_ifaddr(ifa); | |
838 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
839 | fib_sync_up(ifa->ifa_dev->dev); | |
840 | #endif | |
841 | rt_cache_flush(-1); | |
842 | break; | |
843 | case NETDEV_DOWN: | |
844 | fib_del_ifaddr(ifa); | |
9fcc2e8a | 845 | if (ifa->ifa_dev->ifa_list == NULL) { |
1da177e4 LT |
846 | /* Last address was deleted from this interface. |
847 | Disable IP. | |
848 | */ | |
849 | fib_disable_ip(ifa->ifa_dev->dev, 1); | |
850 | } else { | |
851 | rt_cache_flush(-1); | |
852 | } | |
853 | break; | |
854 | } | |
855 | return NOTIFY_DONE; | |
856 | } | |
857 | ||
858 | static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr) | |
859 | { | |
860 | struct net_device *dev = ptr; | |
e5ed6399 | 861 | struct in_device *in_dev = __in_dev_get_rtnl(dev); |
1da177e4 LT |
862 | |
863 | if (event == NETDEV_UNREGISTER) { | |
864 | fib_disable_ip(dev, 2); | |
865 | return NOTIFY_DONE; | |
866 | } | |
867 | ||
868 | if (!in_dev) | |
869 | return NOTIFY_DONE; | |
870 | ||
871 | switch (event) { | |
872 | case NETDEV_UP: | |
873 | for_ifa(in_dev) { | |
874 | fib_add_ifaddr(ifa); | |
875 | } endfor_ifa(in_dev); | |
876 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
877 | fib_sync_up(dev); | |
878 | #endif | |
879 | rt_cache_flush(-1); | |
880 | break; | |
881 | case NETDEV_DOWN: | |
882 | fib_disable_ip(dev, 0); | |
883 | break; | |
884 | case NETDEV_CHANGEMTU: | |
885 | case NETDEV_CHANGE: | |
886 | rt_cache_flush(0); | |
887 | break; | |
888 | } | |
889 | return NOTIFY_DONE; | |
890 | } | |
891 | ||
892 | static struct notifier_block fib_inetaddr_notifier = { | |
893 | .notifier_call =fib_inetaddr_event, | |
894 | }; | |
895 | ||
896 | static struct notifier_block fib_netdev_notifier = { | |
897 | .notifier_call =fib_netdev_event, | |
898 | }; | |
899 | ||
900 | void __init ip_fib_init(void) | |
901 | { | |
1af5a8c4 PM |
902 | unsigned int i; |
903 | ||
904 | for (i = 0; i < FIB_TABLE_HASHSZ; i++) | |
905 | INIT_HLIST_HEAD(&fib_table_hash[i]); | |
1da177e4 LT |
906 | #ifndef CONFIG_IP_MULTIPLE_TABLES |
907 | ip_fib_local_table = fib_hash_init(RT_TABLE_LOCAL); | |
1af5a8c4 | 908 | hlist_add_head_rcu(&ip_fib_local_table->tb_hlist, &fib_table_hash[0]); |
1da177e4 | 909 | ip_fib_main_table = fib_hash_init(RT_TABLE_MAIN); |
1af5a8c4 | 910 | hlist_add_head_rcu(&ip_fib_main_table->tb_hlist, &fib_table_hash[0]); |
1da177e4 | 911 | #else |
e1ef4bf2 | 912 | fib4_rules_init(); |
1da177e4 LT |
913 | #endif |
914 | ||
915 | register_netdevice_notifier(&fib_netdev_notifier); | |
916 | register_inetaddr_notifier(&fib_inetaddr_notifier); | |
246955fe | 917 | nl_fib_lookup_init(); |
1da177e4 LT |
918 | } |
919 | ||
920 | EXPORT_SYMBOL(inet_addr_type); | |
a1e8733e | 921 | EXPORT_SYMBOL(ip_dev_find); |