]>
Commit | Line | Data |
---|---|---|
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 | * 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 | ||
16 | #include <linux/module.h> | |
17 | #include <asm/uaccess.h> | |
18 | #include <asm/system.h> | |
19 | #include <linux/bitops.h> | |
20 | #include <linux/capability.h> | |
21 | #include <linux/types.h> | |
22 | #include <linux/kernel.h> | |
23 | #include <linux/mm.h> | |
24 | #include <linux/string.h> | |
25 | #include <linux/socket.h> | |
26 | #include <linux/sockios.h> | |
27 | #include <linux/errno.h> | |
28 | #include <linux/in.h> | |
29 | #include <linux/inet.h> | |
30 | #include <linux/inetdevice.h> | |
31 | #include <linux/netdevice.h> | |
32 | #include <linux/if_addr.h> | |
33 | #include <linux/if_arp.h> | |
34 | #include <linux/skbuff.h> | |
35 | #include <linux/init.h> | |
36 | #include <linux/list.h> | |
37 | #include <linux/slab.h> | |
38 | ||
39 | #include <net/ip.h> | |
40 | #include <net/protocol.h> | |
41 | #include <net/route.h> | |
42 | #include <net/tcp.h> | |
43 | #include <net/sock.h> | |
44 | #include <net/arp.h> | |
45 | #include <net/ip_fib.h> | |
46 | #include <net/rtnetlink.h> | |
47 | ||
48 | #ifndef CONFIG_IP_MULTIPLE_TABLES | |
49 | ||
50 | static int __net_init fib4_rules_init(struct net *net) | |
51 | { | |
52 | struct fib_table *local_table, *main_table; | |
53 | ||
54 | local_table = fib_hash_table(RT_TABLE_LOCAL); | |
55 | if (local_table == NULL) | |
56 | return -ENOMEM; | |
57 | ||
58 | main_table = fib_hash_table(RT_TABLE_MAIN); | |
59 | if (main_table == NULL) | |
60 | goto fail; | |
61 | ||
62 | hlist_add_head_rcu(&local_table->tb_hlist, | |
63 | &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]); | |
64 | hlist_add_head_rcu(&main_table->tb_hlist, | |
65 | &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]); | |
66 | return 0; | |
67 | ||
68 | fail: | |
69 | kfree(local_table); | |
70 | return -ENOMEM; | |
71 | } | |
72 | #else | |
73 | ||
74 | struct fib_table *fib_new_table(struct net *net, u32 id) | |
75 | { | |
76 | struct fib_table *tb; | |
77 | unsigned int h; | |
78 | ||
79 | if (id == 0) | |
80 | id = RT_TABLE_MAIN; | |
81 | tb = fib_get_table(net, id); | |
82 | if (tb) | |
83 | return tb; | |
84 | ||
85 | tb = fib_hash_table(id); | |
86 | if (!tb) | |
87 | return NULL; | |
88 | h = id & (FIB_TABLE_HASHSZ - 1); | |
89 | hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]); | |
90 | return tb; | |
91 | } | |
92 | ||
93 | struct fib_table *fib_get_table(struct net *net, u32 id) | |
94 | { | |
95 | struct fib_table *tb; | |
96 | struct hlist_node *node; | |
97 | struct hlist_head *head; | |
98 | unsigned int h; | |
99 | ||
100 | if (id == 0) | |
101 | id = RT_TABLE_MAIN; | |
102 | h = id & (FIB_TABLE_HASHSZ - 1); | |
103 | ||
104 | rcu_read_lock(); | |
105 | head = &net->ipv4.fib_table_hash[h]; | |
106 | hlist_for_each_entry_rcu(tb, node, head, tb_hlist) { | |
107 | if (tb->tb_id == id) { | |
108 | rcu_read_unlock(); | |
109 | return tb; | |
110 | } | |
111 | } | |
112 | rcu_read_unlock(); | |
113 | return NULL; | |
114 | } | |
115 | #endif /* CONFIG_IP_MULTIPLE_TABLES */ | |
116 | ||
117 | void fib_select_default(struct net *net, | |
118 | const struct flowi *flp, struct fib_result *res) | |
119 | { | |
120 | struct fib_table *tb; | |
121 | int table = RT_TABLE_MAIN; | |
122 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
123 | if (res->r == NULL || res->r->action != FR_ACT_TO_TBL) | |
124 | return; | |
125 | table = res->r->table; | |
126 | #endif | |
127 | tb = fib_get_table(net, table); | |
128 | if (FIB_RES_GW(*res) && FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK) | |
129 | fib_table_select_default(tb, flp, res); | |
130 | } | |
131 | ||
132 | static void fib_flush(struct net *net) | |
133 | { | |
134 | int flushed = 0; | |
135 | struct fib_table *tb; | |
136 | struct hlist_node *node; | |
137 | struct hlist_head *head; | |
138 | unsigned int h; | |
139 | ||
140 | for (h = 0; h < FIB_TABLE_HASHSZ; h++) { | |
141 | head = &net->ipv4.fib_table_hash[h]; | |
142 | hlist_for_each_entry(tb, node, head, tb_hlist) | |
143 | flushed += fib_table_flush(tb); | |
144 | } | |
145 | ||
146 | if (flushed) | |
147 | rt_cache_flush(net, -1); | |
148 | } | |
149 | ||
150 | /** | |
151 | * __ip_dev_find - find the first device with a given source address. | |
152 | * @net: the net namespace | |
153 | * @addr: the source address | |
154 | * @devref: if true, take a reference on the found device | |
155 | * | |
156 | * If a caller uses devref=false, it should be protected by RCU, or RTNL | |
157 | */ | |
158 | struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref) | |
159 | { | |
160 | struct flowi fl = { | |
161 | .fl4_dst = addr, | |
162 | .flags = FLOWI_FLAG_MATCH_ANY_IIF | |
163 | }; | |
164 | struct fib_result res = { 0 }; | |
165 | struct net_device *dev = NULL; | |
166 | ||
167 | rcu_read_lock(); | |
168 | if (fib_lookup(net, &fl, &res)) { | |
169 | rcu_read_unlock(); | |
170 | return NULL; | |
171 | } | |
172 | if (res.type != RTN_LOCAL) | |
173 | goto out; | |
174 | dev = FIB_RES_DEV(res); | |
175 | ||
176 | if (dev && devref) | |
177 | dev_hold(dev); | |
178 | out: | |
179 | rcu_read_unlock(); | |
180 | return dev; | |
181 | } | |
182 | EXPORT_SYMBOL(__ip_dev_find); | |
183 | ||
184 | /* | |
185 | * Find address type as if only "dev" was present in the system. If | |
186 | * on_dev is NULL then all interfaces are taken into consideration. | |
187 | */ | |
188 | static inline unsigned __inet_dev_addr_type(struct net *net, | |
189 | const struct net_device *dev, | |
190 | __be32 addr) | |
191 | { | |
192 | struct flowi fl = { .fl4_dst = addr }; | |
193 | struct fib_result res; | |
194 | unsigned ret = RTN_BROADCAST; | |
195 | struct fib_table *local_table; | |
196 | ||
197 | if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr)) | |
198 | return RTN_BROADCAST; | |
199 | if (ipv4_is_multicast(addr)) | |
200 | return RTN_MULTICAST; | |
201 | ||
202 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
203 | res.r = NULL; | |
204 | #endif | |
205 | ||
206 | local_table = fib_get_table(net, RT_TABLE_LOCAL); | |
207 | if (local_table) { | |
208 | ret = RTN_UNICAST; | |
209 | rcu_read_lock(); | |
210 | if (!fib_table_lookup(local_table, &fl, &res, FIB_LOOKUP_NOREF)) { | |
211 | if (!dev || dev == res.fi->fib_dev) | |
212 | ret = res.type; | |
213 | } | |
214 | rcu_read_unlock(); | |
215 | } | |
216 | return ret; | |
217 | } | |
218 | ||
219 | unsigned int inet_addr_type(struct net *net, __be32 addr) | |
220 | { | |
221 | return __inet_dev_addr_type(net, NULL, addr); | |
222 | } | |
223 | EXPORT_SYMBOL(inet_addr_type); | |
224 | ||
225 | unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev, | |
226 | __be32 addr) | |
227 | { | |
228 | return __inet_dev_addr_type(net, dev, addr); | |
229 | } | |
230 | EXPORT_SYMBOL(inet_dev_addr_type); | |
231 | ||
232 | /* Given (packet source, input interface) and optional (dst, oif, tos): | |
233 | * - (main) check, that source is valid i.e. not broadcast or our local | |
234 | * address. | |
235 | * - figure out what "logical" interface this packet arrived | |
236 | * and calculate "specific destination" address. | |
237 | * - check, that packet arrived from expected physical interface. | |
238 | * called with rcu_read_lock() | |
239 | */ | |
240 | int fib_validate_source(__be32 src, __be32 dst, u8 tos, int oif, | |
241 | struct net_device *dev, __be32 *spec_dst, | |
242 | u32 *itag, u32 mark) | |
243 | { | |
244 | struct in_device *in_dev; | |
245 | struct flowi fl = { | |
246 | .fl4_dst = src, | |
247 | .fl4_src = dst, | |
248 | .fl4_tos = tos, | |
249 | .mark = mark, | |
250 | .iif = oif | |
251 | }; | |
252 | struct fib_result res; | |
253 | int no_addr, rpf, accept_local; | |
254 | bool dev_match; | |
255 | int ret; | |
256 | struct net *net; | |
257 | ||
258 | no_addr = rpf = accept_local = 0; | |
259 | in_dev = __in_dev_get_rcu(dev); | |
260 | if (in_dev) { | |
261 | no_addr = in_dev->ifa_list == NULL; | |
262 | rpf = IN_DEV_RPFILTER(in_dev); | |
263 | accept_local = IN_DEV_ACCEPT_LOCAL(in_dev); | |
264 | if (mark && !IN_DEV_SRC_VMARK(in_dev)) | |
265 | fl.mark = 0; | |
266 | } | |
267 | ||
268 | if (in_dev == NULL) | |
269 | goto e_inval; | |
270 | ||
271 | net = dev_net(dev); | |
272 | if (fib_lookup(net, &fl, &res)) | |
273 | goto last_resort; | |
274 | if (res.type != RTN_UNICAST) { | |
275 | if (res.type != RTN_LOCAL || !accept_local) | |
276 | goto e_inval; | |
277 | } | |
278 | *spec_dst = FIB_RES_PREFSRC(res); | |
279 | fib_combine_itag(itag, &res); | |
280 | dev_match = false; | |
281 | ||
282 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
283 | for (ret = 0; ret < res.fi->fib_nhs; ret++) { | |
284 | struct fib_nh *nh = &res.fi->fib_nh[ret]; | |
285 | ||
286 | if (nh->nh_dev == dev) { | |
287 | dev_match = true; | |
288 | break; | |
289 | } | |
290 | } | |
291 | #else | |
292 | if (FIB_RES_DEV(res) == dev) | |
293 | dev_match = true; | |
294 | #endif | |
295 | if (dev_match) { | |
296 | ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST; | |
297 | return ret; | |
298 | } | |
299 | if (no_addr) | |
300 | goto last_resort; | |
301 | if (rpf == 1) | |
302 | goto e_rpf; | |
303 | fl.oif = dev->ifindex; | |
304 | ||
305 | ret = 0; | |
306 | if (fib_lookup(net, &fl, &res) == 0) { | |
307 | if (res.type == RTN_UNICAST) { | |
308 | *spec_dst = FIB_RES_PREFSRC(res); | |
309 | ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST; | |
310 | } | |
311 | } | |
312 | return ret; | |
313 | ||
314 | last_resort: | |
315 | if (rpf) | |
316 | goto e_rpf; | |
317 | *spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE); | |
318 | *itag = 0; | |
319 | return 0; | |
320 | ||
321 | e_inval: | |
322 | return -EINVAL; | |
323 | e_rpf: | |
324 | return -EXDEV; | |
325 | } | |
326 | ||
327 | static inline __be32 sk_extract_addr(struct sockaddr *addr) | |
328 | { | |
329 | return ((struct sockaddr_in *) addr)->sin_addr.s_addr; | |
330 | } | |
331 | ||
332 | static int put_rtax(struct nlattr *mx, int len, int type, u32 value) | |
333 | { | |
334 | struct nlattr *nla; | |
335 | ||
336 | nla = (struct nlattr *) ((char *) mx + len); | |
337 | nla->nla_type = type; | |
338 | nla->nla_len = nla_attr_size(4); | |
339 | *(u32 *) nla_data(nla) = value; | |
340 | ||
341 | return len + nla_total_size(4); | |
342 | } | |
343 | ||
344 | static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt, | |
345 | struct fib_config *cfg) | |
346 | { | |
347 | __be32 addr; | |
348 | int plen; | |
349 | ||
350 | memset(cfg, 0, sizeof(*cfg)); | |
351 | cfg->fc_nlinfo.nl_net = net; | |
352 | ||
353 | if (rt->rt_dst.sa_family != AF_INET) | |
354 | return -EAFNOSUPPORT; | |
355 | ||
356 | /* | |
357 | * Check mask for validity: | |
358 | * a) it must be contiguous. | |
359 | * b) destination must have all host bits clear. | |
360 | * c) if application forgot to set correct family (AF_INET), | |
361 | * reject request unless it is absolutely clear i.e. | |
362 | * both family and mask are zero. | |
363 | */ | |
364 | plen = 32; | |
365 | addr = sk_extract_addr(&rt->rt_dst); | |
366 | if (!(rt->rt_flags & RTF_HOST)) { | |
367 | __be32 mask = sk_extract_addr(&rt->rt_genmask); | |
368 | ||
369 | if (rt->rt_genmask.sa_family != AF_INET) { | |
370 | if (mask || rt->rt_genmask.sa_family) | |
371 | return -EAFNOSUPPORT; | |
372 | } | |
373 | ||
374 | if (bad_mask(mask, addr)) | |
375 | return -EINVAL; | |
376 | ||
377 | plen = inet_mask_len(mask); | |
378 | } | |
379 | ||
380 | cfg->fc_dst_len = plen; | |
381 | cfg->fc_dst = addr; | |
382 | ||
383 | if (cmd != SIOCDELRT) { | |
384 | cfg->fc_nlflags = NLM_F_CREATE; | |
385 | cfg->fc_protocol = RTPROT_BOOT; | |
386 | } | |
387 | ||
388 | if (rt->rt_metric) | |
389 | cfg->fc_priority = rt->rt_metric - 1; | |
390 | ||
391 | if (rt->rt_flags & RTF_REJECT) { | |
392 | cfg->fc_scope = RT_SCOPE_HOST; | |
393 | cfg->fc_type = RTN_UNREACHABLE; | |
394 | return 0; | |
395 | } | |
396 | ||
397 | cfg->fc_scope = RT_SCOPE_NOWHERE; | |
398 | cfg->fc_type = RTN_UNICAST; | |
399 | ||
400 | if (rt->rt_dev) { | |
401 | char *colon; | |
402 | struct net_device *dev; | |
403 | char devname[IFNAMSIZ]; | |
404 | ||
405 | if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1)) | |
406 | return -EFAULT; | |
407 | ||
408 | devname[IFNAMSIZ-1] = 0; | |
409 | colon = strchr(devname, ':'); | |
410 | if (colon) | |
411 | *colon = 0; | |
412 | dev = __dev_get_by_name(net, devname); | |
413 | if (!dev) | |
414 | return -ENODEV; | |
415 | cfg->fc_oif = dev->ifindex; | |
416 | if (colon) { | |
417 | struct in_ifaddr *ifa; | |
418 | struct in_device *in_dev = __in_dev_get_rtnl(dev); | |
419 | if (!in_dev) | |
420 | return -ENODEV; | |
421 | *colon = ':'; | |
422 | for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) | |
423 | if (strcmp(ifa->ifa_label, devname) == 0) | |
424 | break; | |
425 | if (ifa == NULL) | |
426 | return -ENODEV; | |
427 | cfg->fc_prefsrc = ifa->ifa_local; | |
428 | } | |
429 | } | |
430 | ||
431 | addr = sk_extract_addr(&rt->rt_gateway); | |
432 | if (rt->rt_gateway.sa_family == AF_INET && addr) { | |
433 | cfg->fc_gw = addr; | |
434 | if (rt->rt_flags & RTF_GATEWAY && | |
435 | inet_addr_type(net, addr) == RTN_UNICAST) | |
436 | cfg->fc_scope = RT_SCOPE_UNIVERSE; | |
437 | } | |
438 | ||
439 | if (cmd == SIOCDELRT) | |
440 | return 0; | |
441 | ||
442 | if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw) | |
443 | return -EINVAL; | |
444 | ||
445 | if (cfg->fc_scope == RT_SCOPE_NOWHERE) | |
446 | cfg->fc_scope = RT_SCOPE_LINK; | |
447 | ||
448 | if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) { | |
449 | struct nlattr *mx; | |
450 | int len = 0; | |
451 | ||
452 | mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL); | |
453 | if (mx == NULL) | |
454 | return -ENOMEM; | |
455 | ||
456 | if (rt->rt_flags & RTF_MTU) | |
457 | len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40); | |
458 | ||
459 | if (rt->rt_flags & RTF_WINDOW) | |
460 | len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window); | |
461 | ||
462 | if (rt->rt_flags & RTF_IRTT) | |
463 | len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3); | |
464 | ||
465 | cfg->fc_mx = mx; | |
466 | cfg->fc_mx_len = len; | |
467 | } | |
468 | ||
469 | return 0; | |
470 | } | |
471 | ||
472 | /* | |
473 | * Handle IP routing ioctl calls. | |
474 | * These are used to manipulate the routing tables | |
475 | */ | |
476 | int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg) | |
477 | { | |
478 | struct fib_config cfg; | |
479 | struct rtentry rt; | |
480 | int err; | |
481 | ||
482 | switch (cmd) { | |
483 | case SIOCADDRT: /* Add a route */ | |
484 | case SIOCDELRT: /* Delete a route */ | |
485 | if (!capable(CAP_NET_ADMIN)) | |
486 | return -EPERM; | |
487 | ||
488 | if (copy_from_user(&rt, arg, sizeof(rt))) | |
489 | return -EFAULT; | |
490 | ||
491 | rtnl_lock(); | |
492 | err = rtentry_to_fib_config(net, cmd, &rt, &cfg); | |
493 | if (err == 0) { | |
494 | struct fib_table *tb; | |
495 | ||
496 | if (cmd == SIOCDELRT) { | |
497 | tb = fib_get_table(net, cfg.fc_table); | |
498 | if (tb) | |
499 | err = fib_table_delete(tb, &cfg); | |
500 | else | |
501 | err = -ESRCH; | |
502 | } else { | |
503 | tb = fib_new_table(net, cfg.fc_table); | |
504 | if (tb) | |
505 | err = fib_table_insert(tb, &cfg); | |
506 | else | |
507 | err = -ENOBUFS; | |
508 | } | |
509 | ||
510 | /* allocated by rtentry_to_fib_config() */ | |
511 | kfree(cfg.fc_mx); | |
512 | } | |
513 | rtnl_unlock(); | |
514 | return err; | |
515 | } | |
516 | return -EINVAL; | |
517 | } | |
518 | ||
519 | const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = { | |
520 | [RTA_DST] = { .type = NLA_U32 }, | |
521 | [RTA_SRC] = { .type = NLA_U32 }, | |
522 | [RTA_IIF] = { .type = NLA_U32 }, | |
523 | [RTA_OIF] = { .type = NLA_U32 }, | |
524 | [RTA_GATEWAY] = { .type = NLA_U32 }, | |
525 | [RTA_PRIORITY] = { .type = NLA_U32 }, | |
526 | [RTA_PREFSRC] = { .type = NLA_U32 }, | |
527 | [RTA_METRICS] = { .type = NLA_NESTED }, | |
528 | [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) }, | |
529 | [RTA_FLOW] = { .type = NLA_U32 }, | |
530 | }; | |
531 | ||
532 | static int rtm_to_fib_config(struct net *net, struct sk_buff *skb, | |
533 | struct nlmsghdr *nlh, struct fib_config *cfg) | |
534 | { | |
535 | struct nlattr *attr; | |
536 | int err, remaining; | |
537 | struct rtmsg *rtm; | |
538 | ||
539 | err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy); | |
540 | if (err < 0) | |
541 | goto errout; | |
542 | ||
543 | memset(cfg, 0, sizeof(*cfg)); | |
544 | ||
545 | rtm = nlmsg_data(nlh); | |
546 | cfg->fc_dst_len = rtm->rtm_dst_len; | |
547 | cfg->fc_tos = rtm->rtm_tos; | |
548 | cfg->fc_table = rtm->rtm_table; | |
549 | cfg->fc_protocol = rtm->rtm_protocol; | |
550 | cfg->fc_scope = rtm->rtm_scope; | |
551 | cfg->fc_type = rtm->rtm_type; | |
552 | cfg->fc_flags = rtm->rtm_flags; | |
553 | cfg->fc_nlflags = nlh->nlmsg_flags; | |
554 | ||
555 | cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid; | |
556 | cfg->fc_nlinfo.nlh = nlh; | |
557 | cfg->fc_nlinfo.nl_net = net; | |
558 | ||
559 | if (cfg->fc_type > RTN_MAX) { | |
560 | err = -EINVAL; | |
561 | goto errout; | |
562 | } | |
563 | ||
564 | nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) { | |
565 | switch (nla_type(attr)) { | |
566 | case RTA_DST: | |
567 | cfg->fc_dst = nla_get_be32(attr); | |
568 | break; | |
569 | case RTA_OIF: | |
570 | cfg->fc_oif = nla_get_u32(attr); | |
571 | break; | |
572 | case RTA_GATEWAY: | |
573 | cfg->fc_gw = nla_get_be32(attr); | |
574 | break; | |
575 | case RTA_PRIORITY: | |
576 | cfg->fc_priority = nla_get_u32(attr); | |
577 | break; | |
578 | case RTA_PREFSRC: | |
579 | cfg->fc_prefsrc = nla_get_be32(attr); | |
580 | break; | |
581 | case RTA_METRICS: | |
582 | cfg->fc_mx = nla_data(attr); | |
583 | cfg->fc_mx_len = nla_len(attr); | |
584 | break; | |
585 | case RTA_MULTIPATH: | |
586 | cfg->fc_mp = nla_data(attr); | |
587 | cfg->fc_mp_len = nla_len(attr); | |
588 | break; | |
589 | case RTA_FLOW: | |
590 | cfg->fc_flow = nla_get_u32(attr); | |
591 | break; | |
592 | case RTA_TABLE: | |
593 | cfg->fc_table = nla_get_u32(attr); | |
594 | break; | |
595 | } | |
596 | } | |
597 | ||
598 | return 0; | |
599 | errout: | |
600 | return err; | |
601 | } | |
602 | ||
603 | static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) | |
604 | { | |
605 | struct net *net = sock_net(skb->sk); | |
606 | struct fib_config cfg; | |
607 | struct fib_table *tb; | |
608 | int err; | |
609 | ||
610 | err = rtm_to_fib_config(net, skb, nlh, &cfg); | |
611 | if (err < 0) | |
612 | goto errout; | |
613 | ||
614 | tb = fib_get_table(net, cfg.fc_table); | |
615 | if (tb == NULL) { | |
616 | err = -ESRCH; | |
617 | goto errout; | |
618 | } | |
619 | ||
620 | err = fib_table_delete(tb, &cfg); | |
621 | errout: | |
622 | return err; | |
623 | } | |
624 | ||
625 | static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) | |
626 | { | |
627 | struct net *net = sock_net(skb->sk); | |
628 | struct fib_config cfg; | |
629 | struct fib_table *tb; | |
630 | int err; | |
631 | ||
632 | err = rtm_to_fib_config(net, skb, nlh, &cfg); | |
633 | if (err < 0) | |
634 | goto errout; | |
635 | ||
636 | tb = fib_new_table(net, cfg.fc_table); | |
637 | if (tb == NULL) { | |
638 | err = -ENOBUFS; | |
639 | goto errout; | |
640 | } | |
641 | ||
642 | err = fib_table_insert(tb, &cfg); | |
643 | errout: | |
644 | return err; | |
645 | } | |
646 | ||
647 | static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb) | |
648 | { | |
649 | struct net *net = sock_net(skb->sk); | |
650 | unsigned int h, s_h; | |
651 | unsigned int e = 0, s_e; | |
652 | struct fib_table *tb; | |
653 | struct hlist_node *node; | |
654 | struct hlist_head *head; | |
655 | int dumped = 0; | |
656 | ||
657 | if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) && | |
658 | ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED) | |
659 | return ip_rt_dump(skb, cb); | |
660 | ||
661 | s_h = cb->args[0]; | |
662 | s_e = cb->args[1]; | |
663 | ||
664 | for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) { | |
665 | e = 0; | |
666 | head = &net->ipv4.fib_table_hash[h]; | |
667 | hlist_for_each_entry(tb, node, head, tb_hlist) { | |
668 | if (e < s_e) | |
669 | goto next; | |
670 | if (dumped) | |
671 | memset(&cb->args[2], 0, sizeof(cb->args) - | |
672 | 2 * sizeof(cb->args[0])); | |
673 | if (fib_table_dump(tb, skb, cb) < 0) | |
674 | goto out; | |
675 | dumped = 1; | |
676 | next: | |
677 | e++; | |
678 | } | |
679 | } | |
680 | out: | |
681 | cb->args[1] = e; | |
682 | cb->args[0] = h; | |
683 | ||
684 | return skb->len; | |
685 | } | |
686 | ||
687 | /* Prepare and feed intra-kernel routing request. | |
688 | * Really, it should be netlink message, but :-( netlink | |
689 | * can be not configured, so that we feed it directly | |
690 | * to fib engine. It is legal, because all events occur | |
691 | * only when netlink is already locked. | |
692 | */ | |
693 | static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa) | |
694 | { | |
695 | struct net *net = dev_net(ifa->ifa_dev->dev); | |
696 | struct fib_table *tb; | |
697 | struct fib_config cfg = { | |
698 | .fc_protocol = RTPROT_KERNEL, | |
699 | .fc_type = type, | |
700 | .fc_dst = dst, | |
701 | .fc_dst_len = dst_len, | |
702 | .fc_prefsrc = ifa->ifa_local, | |
703 | .fc_oif = ifa->ifa_dev->dev->ifindex, | |
704 | .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND, | |
705 | .fc_nlinfo = { | |
706 | .nl_net = net, | |
707 | }, | |
708 | }; | |
709 | ||
710 | if (type == RTN_UNICAST) | |
711 | tb = fib_new_table(net, RT_TABLE_MAIN); | |
712 | else | |
713 | tb = fib_new_table(net, RT_TABLE_LOCAL); | |
714 | ||
715 | if (tb == NULL) | |
716 | return; | |
717 | ||
718 | cfg.fc_table = tb->tb_id; | |
719 | ||
720 | if (type != RTN_LOCAL) | |
721 | cfg.fc_scope = RT_SCOPE_LINK; | |
722 | else | |
723 | cfg.fc_scope = RT_SCOPE_HOST; | |
724 | ||
725 | if (cmd == RTM_NEWROUTE) | |
726 | fib_table_insert(tb, &cfg); | |
727 | else | |
728 | fib_table_delete(tb, &cfg); | |
729 | } | |
730 | ||
731 | void fib_add_ifaddr(struct in_ifaddr *ifa) | |
732 | { | |
733 | struct in_device *in_dev = ifa->ifa_dev; | |
734 | struct net_device *dev = in_dev->dev; | |
735 | struct in_ifaddr *prim = ifa; | |
736 | __be32 mask = ifa->ifa_mask; | |
737 | __be32 addr = ifa->ifa_local; | |
738 | __be32 prefix = ifa->ifa_address & mask; | |
739 | ||
740 | if (ifa->ifa_flags & IFA_F_SECONDARY) { | |
741 | prim = inet_ifa_byprefix(in_dev, prefix, mask); | |
742 | if (prim == NULL) { | |
743 | printk(KERN_WARNING "fib_add_ifaddr: bug: prim == NULL\n"); | |
744 | return; | |
745 | } | |
746 | } | |
747 | ||
748 | fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim); | |
749 | ||
750 | if (!(dev->flags & IFF_UP)) | |
751 | return; | |
752 | ||
753 | /* Add broadcast address, if it is explicitly assigned. */ | |
754 | if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF)) | |
755 | fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim); | |
756 | ||
757 | if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) && | |
758 | (prefix != addr || ifa->ifa_prefixlen < 32)) { | |
759 | fib_magic(RTM_NEWROUTE, | |
760 | dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST, | |
761 | prefix, ifa->ifa_prefixlen, prim); | |
762 | ||
763 | /* Add network specific broadcasts, when it takes a sense */ | |
764 | if (ifa->ifa_prefixlen < 31) { | |
765 | fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim); | |
766 | fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask, | |
767 | 32, prim); | |
768 | } | |
769 | } | |
770 | } | |
771 | ||
772 | static void fib_del_ifaddr(struct in_ifaddr *ifa) | |
773 | { | |
774 | struct in_device *in_dev = ifa->ifa_dev; | |
775 | struct net_device *dev = in_dev->dev; | |
776 | struct in_ifaddr *ifa1; | |
777 | struct in_ifaddr *prim = ifa; | |
778 | __be32 brd = ifa->ifa_address | ~ifa->ifa_mask; | |
779 | __be32 any = ifa->ifa_address & ifa->ifa_mask; | |
780 | #define LOCAL_OK 1 | |
781 | #define BRD_OK 2 | |
782 | #define BRD0_OK 4 | |
783 | #define BRD1_OK 8 | |
784 | unsigned ok = 0; | |
785 | ||
786 | if (!(ifa->ifa_flags & IFA_F_SECONDARY)) | |
787 | fib_magic(RTM_DELROUTE, | |
788 | dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST, | |
789 | any, ifa->ifa_prefixlen, prim); | |
790 | else { | |
791 | prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask); | |
792 | if (prim == NULL) { | |
793 | printk(KERN_WARNING "fib_del_ifaddr: bug: prim == NULL\n"); | |
794 | return; | |
795 | } | |
796 | } | |
797 | ||
798 | /* Deletion is more complicated than add. | |
799 | * We should take care of not to delete too much :-) | |
800 | * | |
801 | * Scan address list to be sure that addresses are really gone. | |
802 | */ | |
803 | ||
804 | for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) { | |
805 | if (ifa->ifa_local == ifa1->ifa_local) | |
806 | ok |= LOCAL_OK; | |
807 | if (ifa->ifa_broadcast == ifa1->ifa_broadcast) | |
808 | ok |= BRD_OK; | |
809 | if (brd == ifa1->ifa_broadcast) | |
810 | ok |= BRD1_OK; | |
811 | if (any == ifa1->ifa_broadcast) | |
812 | ok |= BRD0_OK; | |
813 | } | |
814 | ||
815 | if (!(ok & BRD_OK)) | |
816 | fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim); | |
817 | if (!(ok & BRD1_OK)) | |
818 | fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim); | |
819 | if (!(ok & BRD0_OK)) | |
820 | fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim); | |
821 | if (!(ok & LOCAL_OK)) { | |
822 | fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim); | |
823 | ||
824 | /* Check, that this local address finally disappeared. */ | |
825 | if (inet_addr_type(dev_net(dev), ifa->ifa_local) != RTN_LOCAL) { | |
826 | /* And the last, but not the least thing. | |
827 | * We must flush stray FIB entries. | |
828 | * | |
829 | * First of all, we scan fib_info list searching | |
830 | * for stray nexthop entries, then ignite fib_flush. | |
831 | */ | |
832 | if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local)) | |
833 | fib_flush(dev_net(dev)); | |
834 | } | |
835 | } | |
836 | #undef LOCAL_OK | |
837 | #undef BRD_OK | |
838 | #undef BRD0_OK | |
839 | #undef BRD1_OK | |
840 | } | |
841 | ||
842 | static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb) | |
843 | { | |
844 | ||
845 | struct fib_result res; | |
846 | struct flowi fl = { | |
847 | .mark = frn->fl_mark, | |
848 | .fl4_dst = frn->fl_addr, | |
849 | .fl4_tos = frn->fl_tos, | |
850 | .fl4_scope = frn->fl_scope, | |
851 | }; | |
852 | ||
853 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
854 | res.r = NULL; | |
855 | #endif | |
856 | ||
857 | frn->err = -ENOENT; | |
858 | if (tb) { | |
859 | local_bh_disable(); | |
860 | ||
861 | frn->tb_id = tb->tb_id; | |
862 | rcu_read_lock(); | |
863 | frn->err = fib_table_lookup(tb, &fl, &res, FIB_LOOKUP_NOREF); | |
864 | ||
865 | if (!frn->err) { | |
866 | frn->prefixlen = res.prefixlen; | |
867 | frn->nh_sel = res.nh_sel; | |
868 | frn->type = res.type; | |
869 | frn->scope = res.scope; | |
870 | } | |
871 | rcu_read_unlock(); | |
872 | local_bh_enable(); | |
873 | } | |
874 | } | |
875 | ||
876 | static void nl_fib_input(struct sk_buff *skb) | |
877 | { | |
878 | struct net *net; | |
879 | struct fib_result_nl *frn; | |
880 | struct nlmsghdr *nlh; | |
881 | struct fib_table *tb; | |
882 | u32 pid; | |
883 | ||
884 | net = sock_net(skb->sk); | |
885 | nlh = nlmsg_hdr(skb); | |
886 | if (skb->len < NLMSG_SPACE(0) || skb->len < nlh->nlmsg_len || | |
887 | nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*frn))) | |
888 | return; | |
889 | ||
890 | skb = skb_clone(skb, GFP_KERNEL); | |
891 | if (skb == NULL) | |
892 | return; | |
893 | nlh = nlmsg_hdr(skb); | |
894 | ||
895 | frn = (struct fib_result_nl *) NLMSG_DATA(nlh); | |
896 | tb = fib_get_table(net, frn->tb_id_in); | |
897 | ||
898 | nl_fib_lookup(frn, tb); | |
899 | ||
900 | pid = NETLINK_CB(skb).pid; /* pid of sending process */ | |
901 | NETLINK_CB(skb).pid = 0; /* from kernel */ | |
902 | NETLINK_CB(skb).dst_group = 0; /* unicast */ | |
903 | netlink_unicast(net->ipv4.fibnl, skb, pid, MSG_DONTWAIT); | |
904 | } | |
905 | ||
906 | static int __net_init nl_fib_lookup_init(struct net *net) | |
907 | { | |
908 | struct sock *sk; | |
909 | sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, 0, | |
910 | nl_fib_input, NULL, THIS_MODULE); | |
911 | if (sk == NULL) | |
912 | return -EAFNOSUPPORT; | |
913 | net->ipv4.fibnl = sk; | |
914 | return 0; | |
915 | } | |
916 | ||
917 | static void nl_fib_lookup_exit(struct net *net) | |
918 | { | |
919 | netlink_kernel_release(net->ipv4.fibnl); | |
920 | net->ipv4.fibnl = NULL; | |
921 | } | |
922 | ||
923 | static void fib_disable_ip(struct net_device *dev, int force, int delay) | |
924 | { | |
925 | if (fib_sync_down_dev(dev, force)) | |
926 | fib_flush(dev_net(dev)); | |
927 | rt_cache_flush(dev_net(dev), delay); | |
928 | arp_ifdown(dev); | |
929 | } | |
930 | ||
931 | static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr) | |
932 | { | |
933 | struct in_ifaddr *ifa = (struct in_ifaddr *)ptr; | |
934 | struct net_device *dev = ifa->ifa_dev->dev; | |
935 | ||
936 | switch (event) { | |
937 | case NETDEV_UP: | |
938 | fib_add_ifaddr(ifa); | |
939 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
940 | fib_sync_up(dev); | |
941 | #endif | |
942 | rt_cache_flush(dev_net(dev), -1); | |
943 | break; | |
944 | case NETDEV_DOWN: | |
945 | fib_del_ifaddr(ifa); | |
946 | if (ifa->ifa_dev->ifa_list == NULL) { | |
947 | /* Last address was deleted from this interface. | |
948 | * Disable IP. | |
949 | */ | |
950 | fib_disable_ip(dev, 1, 0); | |
951 | } else { | |
952 | rt_cache_flush(dev_net(dev), -1); | |
953 | } | |
954 | break; | |
955 | } | |
956 | return NOTIFY_DONE; | |
957 | } | |
958 | ||
959 | static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr) | |
960 | { | |
961 | struct net_device *dev = ptr; | |
962 | struct in_device *in_dev = __in_dev_get_rtnl(dev); | |
963 | ||
964 | if (event == NETDEV_UNREGISTER) { | |
965 | fib_disable_ip(dev, 2, -1); | |
966 | return NOTIFY_DONE; | |
967 | } | |
968 | ||
969 | if (!in_dev) | |
970 | return NOTIFY_DONE; | |
971 | ||
972 | switch (event) { | |
973 | case NETDEV_UP: | |
974 | for_ifa(in_dev) { | |
975 | fib_add_ifaddr(ifa); | |
976 | } endfor_ifa(in_dev); | |
977 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
978 | fib_sync_up(dev); | |
979 | #endif | |
980 | rt_cache_flush(dev_net(dev), -1); | |
981 | break; | |
982 | case NETDEV_DOWN: | |
983 | fib_disable_ip(dev, 0, 0); | |
984 | break; | |
985 | case NETDEV_CHANGEMTU: | |
986 | case NETDEV_CHANGE: | |
987 | rt_cache_flush(dev_net(dev), 0); | |
988 | break; | |
989 | case NETDEV_UNREGISTER_BATCH: | |
990 | rt_cache_flush_batch(); | |
991 | break; | |
992 | } | |
993 | return NOTIFY_DONE; | |
994 | } | |
995 | ||
996 | static struct notifier_block fib_inetaddr_notifier = { | |
997 | .notifier_call = fib_inetaddr_event, | |
998 | }; | |
999 | ||
1000 | static struct notifier_block fib_netdev_notifier = { | |
1001 | .notifier_call = fib_netdev_event, | |
1002 | }; | |
1003 | ||
1004 | static int __net_init ip_fib_net_init(struct net *net) | |
1005 | { | |
1006 | int err; | |
1007 | size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ; | |
1008 | ||
1009 | /* Avoid false sharing : Use at least a full cache line */ | |
1010 | size = max_t(size_t, size, L1_CACHE_BYTES); | |
1011 | ||
1012 | net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL); | |
1013 | if (net->ipv4.fib_table_hash == NULL) | |
1014 | return -ENOMEM; | |
1015 | ||
1016 | err = fib4_rules_init(net); | |
1017 | if (err < 0) | |
1018 | goto fail; | |
1019 | return 0; | |
1020 | ||
1021 | fail: | |
1022 | kfree(net->ipv4.fib_table_hash); | |
1023 | return err; | |
1024 | } | |
1025 | ||
1026 | static void ip_fib_net_exit(struct net *net) | |
1027 | { | |
1028 | unsigned int i; | |
1029 | ||
1030 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
1031 | fib4_rules_exit(net); | |
1032 | #endif | |
1033 | ||
1034 | for (i = 0; i < FIB_TABLE_HASHSZ; i++) { | |
1035 | struct fib_table *tb; | |
1036 | struct hlist_head *head; | |
1037 | struct hlist_node *node, *tmp; | |
1038 | ||
1039 | head = &net->ipv4.fib_table_hash[i]; | |
1040 | hlist_for_each_entry_safe(tb, node, tmp, head, tb_hlist) { | |
1041 | hlist_del(node); | |
1042 | fib_table_flush(tb); | |
1043 | fib_free_table(tb); | |
1044 | } | |
1045 | } | |
1046 | kfree(net->ipv4.fib_table_hash); | |
1047 | } | |
1048 | ||
1049 | static int __net_init fib_net_init(struct net *net) | |
1050 | { | |
1051 | int error; | |
1052 | ||
1053 | error = ip_fib_net_init(net); | |
1054 | if (error < 0) | |
1055 | goto out; | |
1056 | error = nl_fib_lookup_init(net); | |
1057 | if (error < 0) | |
1058 | goto out_nlfl; | |
1059 | error = fib_proc_init(net); | |
1060 | if (error < 0) | |
1061 | goto out_proc; | |
1062 | out: | |
1063 | return error; | |
1064 | ||
1065 | out_proc: | |
1066 | nl_fib_lookup_exit(net); | |
1067 | out_nlfl: | |
1068 | ip_fib_net_exit(net); | |
1069 | goto out; | |
1070 | } | |
1071 | ||
1072 | static void __net_exit fib_net_exit(struct net *net) | |
1073 | { | |
1074 | fib_proc_exit(net); | |
1075 | nl_fib_lookup_exit(net); | |
1076 | ip_fib_net_exit(net); | |
1077 | } | |
1078 | ||
1079 | static struct pernet_operations fib_net_ops = { | |
1080 | .init = fib_net_init, | |
1081 | .exit = fib_net_exit, | |
1082 | }; | |
1083 | ||
1084 | void __init ip_fib_init(void) | |
1085 | { | |
1086 | rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL); | |
1087 | rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL); | |
1088 | rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib); | |
1089 | ||
1090 | register_pernet_subsys(&fib_net_ops); | |
1091 | register_netdevice_notifier(&fib_netdev_notifier); | |
1092 | register_inetaddr_notifier(&fib_inetaddr_notifier); | |
1093 | ||
1094 | fib_hash_init(); | |
1095 | } |