?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
[NETNS]: Add netns parameter to fib_get_table/fib_new_table.
[linux-2.6/kvm.git] / net / ipv4 / fib_frontend.c
blob7718823711e3ba178f95af76a1aa235313c72de3
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
18 #include <linux/module.h>
19 #include <asm/uaccess.h>
20 #include <asm/system.h>
21 #include <linux/bitops.h>
22 #include <linux/capability.h>
23 #include <linux/types.h>
24 #include <linux/kernel.h>
25 #include <linux/mm.h>
26 #include <linux/string.h>
27 #include <linux/socket.h>
28 #include <linux/sockios.h>
29 #include <linux/errno.h>
30 #include <linux/in.h>
31 #include <linux/inet.h>
32 #include <linux/inetdevice.h>
33 #include <linux/netdevice.h>
34 #include <linux/if_addr.h>
35 #include <linux/if_arp.h>
36 #include <linux/skbuff.h>
37 #include <linux/init.h>
38 #include <linux/list.h>
39
40 #include <net/ip.h>
41 #include <net/protocol.h>
42 #include <net/route.h>
43 #include <net/tcp.h>
44 #include <net/sock.h>
45 #include <net/icmp.h>
46 #include <net/arp.h>
47 #include <net/ip_fib.h>
48 #include <net/rtnetlink.h>
49
50 #define FFprint(a...) printk(KERN_DEBUG a)
51
52 static struct sock *fibnl;
53 struct hlist_head fib_table_hash[FIB_TABLE_HASHSZ];
54
55 #ifndef CONFIG_IP_MULTIPLE_TABLES
56
57 static int __net_init fib4_rules_init(struct net *net)
58 {
59 struct fib_table *local_table, *main_table;
60
61 local_table = fib_hash_init(RT_TABLE_LOCAL);
62 if (local_table == NULL)
63 return -ENOMEM;
64
65 main_table = fib_hash_init(RT_TABLE_MAIN);
66 if (main_table == NULL)
67 goto fail;
68
69 hlist_add_head_rcu(&local_table->tb_hlist,
70 &fib_table_hash[TABLE_LOCAL_INDEX]);
71 hlist_add_head_rcu(&main_table->tb_hlist,
72 &fib_table_hash[TABLE_MAIN_INDEX]);
73 return 0;
74
75 fail:
76 kfree(local_table);
77 return -ENOMEM;
78 }
79 #else
80
81 struct fib_table *fib_new_table(struct net *net, u32 id)
82 {
83 struct fib_table *tb;
84 unsigned int h;
85
86 if (id == 0)
87 id = RT_TABLE_MAIN;
88 tb = fib_get_table(net, id);
89 if (tb)
90 return tb;
91 tb = fib_hash_init(id);
92 if (!tb)
93 return NULL;
94 h = id & (FIB_TABLE_HASHSZ - 1);
95 hlist_add_head_rcu(&tb->tb_hlist, &fib_table_hash[h]);
96 return tb;
97 }
98
99 struct fib_table *fib_get_table(struct net *net, u32 id)
100 {
101 struct fib_table *tb;
102 struct hlist_node *node;
103 unsigned int h;
104
105 if (id == 0)
106 id = RT_TABLE_MAIN;
107 h = id & (FIB_TABLE_HASHSZ - 1);
108 rcu_read_lock();
109 hlist_for_each_entry_rcu(tb, node, &fib_table_hash[h], tb_hlist) {
110 if (tb->tb_id == id) {
111 rcu_read_unlock();
112 return tb;
113 }
114 }
115 rcu_read_unlock();
116 return NULL;
117 }
118 #endif /* CONFIG_IP_MULTIPLE_TABLES */
119
120 static void fib_flush(void)
121 {
122 int flushed = 0;
123 struct fib_table *tb;
124 struct hlist_node *node;
125 unsigned int h;
126
127 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
128 hlist_for_each_entry(tb, node, &fib_table_hash[h], tb_hlist)
129 flushed += tb->tb_flush(tb);
130 }
131
132 if (flushed)
133 rt_cache_flush(-1);
134 }
135
136 /*
137 * Find the first device with a given source address.
138 */
139
140 struct net_device * ip_dev_find(__be32 addr)
141 {
142 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } };
143 struct fib_result res;
144 struct net_device *dev = NULL;
145 struct fib_table *local_table;
146
147 #ifdef CONFIG_IP_MULTIPLE_TABLES
148 res.r = NULL;
149 #endif
150
151 local_table = fib_get_table(&init_net, RT_TABLE_LOCAL);
152 if (!local_table || local_table->tb_lookup(local_table, &fl, &res))
153 return NULL;
154 if (res.type != RTN_LOCAL)
155 goto out;
156 dev = FIB_RES_DEV(res);
157
158 if (dev)
159 dev_hold(dev);
160 out:
161 fib_res_put(&res);
162 return dev;
163 }
164
165 /*
166 * Find address type as if only "dev" was present in the system. If
167 * on_dev is NULL then all interfaces are taken into consideration.
168 */
169 static inline unsigned __inet_dev_addr_type(const struct net_device *dev,
170 __be32 addr)
171 {
172 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } };
173 struct fib_result res;
174 unsigned ret = RTN_BROADCAST;
175 struct fib_table *local_table;
176
177 if (ipv4_is_zeronet(addr) || ipv4_is_badclass(addr))
178 return RTN_BROADCAST;
179 if (ipv4_is_multicast(addr))
180 return RTN_MULTICAST;
181
182 #ifdef CONFIG_IP_MULTIPLE_TABLES
183 res.r = NULL;
184 #endif
185
186 local_table = fib_get_table(&init_net, RT_TABLE_LOCAL);
187 if (local_table) {
188 ret = RTN_UNICAST;
189 if (!local_table->tb_lookup(local_table, &fl, &res)) {
190 if (!dev || dev == res.fi->fib_dev)
191 ret = res.type;
192 fib_res_put(&res);
193 }
194 }
195 return ret;
196 }
197
198 unsigned int inet_addr_type(__be32 addr)
199 {
200 return __inet_dev_addr_type(NULL, addr);
201 }
202
203 unsigned int inet_dev_addr_type(const struct net_device *dev, __be32 addr)
204 {
205 return __inet_dev_addr_type(dev, addr);
206 }
207
208 /* Given (packet source, input interface) and optional (dst, oif, tos):
209 - (main) check, that source is valid i.e. not broadcast or our local
210 address.
211 - figure out what "logical" interface this packet arrived
212 and calculate "specific destination" address.
213 - check, that packet arrived from expected physical interface.
214 */
215
216 int fib_validate_source(__be32 src, __be32 dst, u8 tos, int oif,
217 struct net_device *dev, __be32 *spec_dst, u32 *itag)
218 {
219 struct in_device *in_dev;
220 struct flowi fl = { .nl_u = { .ip4_u =
221 { .daddr = src,
222 .saddr = dst,
223 .tos = tos } },
224 .iif = oif };
225 struct fib_result res;
226 int no_addr, rpf;
227 int ret;
228
229 no_addr = rpf = 0;
230 rcu_read_lock();
231 in_dev = __in_dev_get_rcu(dev);
232 if (in_dev) {
233 no_addr = in_dev->ifa_list == NULL;
234 rpf = IN_DEV_RPFILTER(in_dev);
235 }
236 rcu_read_unlock();
237
238 if (in_dev == NULL)
239 goto e_inval;
240
241 if (fib_lookup(&fl, &res))
242 goto last_resort;
243 if (res.type != RTN_UNICAST)
244 goto e_inval_res;
245 *spec_dst = FIB_RES_PREFSRC(res);
246 fib_combine_itag(itag, &res);
247 #ifdef CONFIG_IP_ROUTE_MULTIPATH
248 if (FIB_RES_DEV(res) == dev || res.fi->fib_nhs > 1)
249 #else
250 if (FIB_RES_DEV(res) == dev)
251 #endif
252 {
253 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
254 fib_res_put(&res);
255 return ret;
256 }
257 fib_res_put(&res);
258 if (no_addr)
259 goto last_resort;
260 if (rpf)
261 goto e_inval;
262 fl.oif = dev->ifindex;
263
264 ret = 0;
265 if (fib_lookup(&fl, &res) == 0) {
266 if (res.type == RTN_UNICAST) {
267 *spec_dst = FIB_RES_PREFSRC(res);
268 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
269 }
270 fib_res_put(&res);
271 }
272 return ret;
273
274 last_resort:
275 if (rpf)
276 goto e_inval;
277 *spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE);
278 *itag = 0;
279 return 0;
280
281 e_inval_res:
282 fib_res_put(&res);
283 e_inval:
284 return -EINVAL;
285 }
286
287 static inline __be32 sk_extract_addr(struct sockaddr *addr)
288 {
289 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
290 }
291
292 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
293 {
294 struct nlattr *nla;
295
296 nla = (struct nlattr *) ((char *) mx + len);
297 nla->nla_type = type;
298 nla->nla_len = nla_attr_size(4);
299 *(u32 *) nla_data(nla) = value;
300
301 return len + nla_total_size(4);
302 }
303
304 static int rtentry_to_fib_config(int cmd, struct rtentry *rt,
305 struct fib_config *cfg)
306 {
307 __be32 addr;
308 int plen;
309
310 memset(cfg, 0, sizeof(*cfg));
311
312 if (rt->rt_dst.sa_family != AF_INET)
313 return -EAFNOSUPPORT;
314
315 /*
316 * Check mask for validity:
317 * a) it must be contiguous.
318 * b) destination must have all host bits clear.
319 * c) if application forgot to set correct family (AF_INET),
320 * reject request unless it is absolutely clear i.e.
321 * both family and mask are zero.
322 */
323 plen = 32;
324 addr = sk_extract_addr(&rt->rt_dst);
325 if (!(rt->rt_flags & RTF_HOST)) {
326 __be32 mask = sk_extract_addr(&rt->rt_genmask);
327
328 if (rt->rt_genmask.sa_family != AF_INET) {
329 if (mask || rt->rt_genmask.sa_family)
330 return -EAFNOSUPPORT;
331 }
332
333 if (bad_mask(mask, addr))
334 return -EINVAL;
335
336 plen = inet_mask_len(mask);
337 }
338
339 cfg->fc_dst_len = plen;
340 cfg->fc_dst = addr;
341
342 if (cmd != SIOCDELRT) {
343 cfg->fc_nlflags = NLM_F_CREATE;
344 cfg->fc_protocol = RTPROT_BOOT;
345 }
346
347 if (rt->rt_metric)
348 cfg->fc_priority = rt->rt_metric - 1;
349
350 if (rt->rt_flags & RTF_REJECT) {
351 cfg->fc_scope = RT_SCOPE_HOST;
352 cfg->fc_type = RTN_UNREACHABLE;
353 return 0;
354 }
355
356 cfg->fc_scope = RT_SCOPE_NOWHERE;
357 cfg->fc_type = RTN_UNICAST;
358
359 if (rt->rt_dev) {
360 char *colon;
361 struct net_device *dev;
362 char devname[IFNAMSIZ];
363
364 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
365 return -EFAULT;
366
367 devname[IFNAMSIZ-1] = 0;
368 colon = strchr(devname, ':');
369 if (colon)
370 *colon = 0;
371 dev = __dev_get_by_name(&init_net, devname);
372 if (!dev)
373 return -ENODEV;
374 cfg->fc_oif = dev->ifindex;
375 if (colon) {
376 struct in_ifaddr *ifa;
377 struct in_device *in_dev = __in_dev_get_rtnl(dev);
378 if (!in_dev)
379 return -ENODEV;
380 *colon = ':';
381 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
382 if (strcmp(ifa->ifa_label, devname) == 0)
383 break;
384 if (ifa == NULL)
385 return -ENODEV;
386 cfg->fc_prefsrc = ifa->ifa_local;
387 }
388 }
389
390 addr = sk_extract_addr(&rt->rt_gateway);
391 if (rt->rt_gateway.sa_family == AF_INET && addr) {
392 cfg->fc_gw = addr;
393 if (rt->rt_flags & RTF_GATEWAY &&
394 inet_addr_type(addr) == RTN_UNICAST)
395 cfg->fc_scope = RT_SCOPE_UNIVERSE;
396 }
397
398 if (cmd == SIOCDELRT)
399 return 0;
400
401 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
402 return -EINVAL;
403
404 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
405 cfg->fc_scope = RT_SCOPE_LINK;
406
407 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
408 struct nlattr *mx;
409 int len = 0;
410
411 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
412 if (mx == NULL)
413 return -ENOMEM;
414
415 if (rt->rt_flags & RTF_MTU)
416 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
417
418 if (rt->rt_flags & RTF_WINDOW)
419 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
420
421 if (rt->rt_flags & RTF_IRTT)
422 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
423
424 cfg->fc_mx = mx;
425 cfg->fc_mx_len = len;
426 }
427
428 return 0;
429 }
430
431 /*
432 * Handle IP routing ioctl calls. These are used to manipulate the routing tables
433 */
434
435 int ip_rt_ioctl(unsigned int cmd, void __user *arg)
436 {
437 struct fib_config cfg;
438 struct rtentry rt;
439 int err;
440
441 switch (cmd) {
442 case SIOCADDRT: /* Add a route */
443 case SIOCDELRT: /* Delete a route */
444 if (!capable(CAP_NET_ADMIN))
445 return -EPERM;
446
447 if (copy_from_user(&rt, arg, sizeof(rt)))
448 return -EFAULT;
449
450 rtnl_lock();
451 err = rtentry_to_fib_config(cmd, &rt, &cfg);
452 if (err == 0) {
453 struct fib_table *tb;
454
455 if (cmd == SIOCDELRT) {
456 tb = fib_get_table(&init_net, cfg.fc_table);
457 if (tb)
458 err = tb->tb_delete(tb, &cfg);
459 else
460 err = -ESRCH;
461 } else {
462 tb = fib_new_table(&init_net, cfg.fc_table);
463 if (tb)
464 err = tb->tb_insert(tb, &cfg);
465 else
466 err = -ENOBUFS;
467 }
468
469 /* allocated by rtentry_to_fib_config() */
470 kfree(cfg.fc_mx);
471 }
472 rtnl_unlock();
473 return err;
474 }
475 return -EINVAL;
476 }
477
478 const struct nla_policy rtm_ipv4_policy[RTA_MAX+1] = {
479 [RTA_DST] = { .type = NLA_U32 },
480 [RTA_SRC] = { .type = NLA_U32 },
481 [RTA_IIF] = { .type = NLA_U32 },
482 [RTA_OIF] = { .type = NLA_U32 },
483 [RTA_GATEWAY] = { .type = NLA_U32 },
484 [RTA_PRIORITY] = { .type = NLA_U32 },
485 [RTA_PREFSRC] = { .type = NLA_U32 },
486 [RTA_METRICS] = { .type = NLA_NESTED },
487 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
488 [RTA_PROTOINFO] = { .type = NLA_U32 },
489 [RTA_FLOW] = { .type = NLA_U32 },
490 };
491
492 static int rtm_to_fib_config(struct sk_buff *skb, struct nlmsghdr *nlh,
493 struct fib_config *cfg)
494 {
495 struct nlattr *attr;
496 int err, remaining;
497 struct rtmsg *rtm;
498
499 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
500 if (err < 0)
501 goto errout;
502
503 memset(cfg, 0, sizeof(*cfg));
504
505 rtm = nlmsg_data(nlh);
506 cfg->fc_dst_len = rtm->rtm_dst_len;
507 cfg->fc_tos = rtm->rtm_tos;
508 cfg->fc_table = rtm->rtm_table;
509 cfg->fc_protocol = rtm->rtm_protocol;
510 cfg->fc_scope = rtm->rtm_scope;
511 cfg->fc_type = rtm->rtm_type;
512 cfg->fc_flags = rtm->rtm_flags;
513 cfg->fc_nlflags = nlh->nlmsg_flags;
514
515 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
516 cfg->fc_nlinfo.nlh = nlh;
517
518 if (cfg->fc_type > RTN_MAX) {
519 err = -EINVAL;
520 goto errout;
521 }
522
523 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
524 switch (nla_type(attr)) {
525 case RTA_DST:
526 cfg->fc_dst = nla_get_be32(attr);
527 break;
528 case RTA_OIF:
529 cfg->fc_oif = nla_get_u32(attr);
530 break;
531 case RTA_GATEWAY:
532 cfg->fc_gw = nla_get_be32(attr);
533 break;
534 case RTA_PRIORITY:
535 cfg->fc_priority = nla_get_u32(attr);
536 break;
537 case RTA_PREFSRC:
538 cfg->fc_prefsrc = nla_get_be32(attr);
539 break;
540 case RTA_METRICS:
541 cfg->fc_mx = nla_data(attr);
542 cfg->fc_mx_len = nla_len(attr);
543 break;
544 case RTA_MULTIPATH:
545 cfg->fc_mp = nla_data(attr);
546 cfg->fc_mp_len = nla_len(attr);
547 break;
548 case RTA_FLOW:
549 cfg->fc_flow = nla_get_u32(attr);
550 break;
551 case RTA_TABLE:
552 cfg->fc_table = nla_get_u32(attr);
553 break;
554 }
555 }
556
557 return 0;
558 errout:
559 return err;
560 }
561
562 static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
563 {
564 struct net *net = skb->sk->sk_net;
565 struct fib_config cfg;
566 struct fib_table *tb;
567 int err;
568
569 if (net != &init_net)
570 return -EINVAL;
571
572 err = rtm_to_fib_config(skb, nlh, &cfg);
573 if (err < 0)
574 goto errout;
575
576 tb = fib_get_table(net, cfg.fc_table);
577 if (tb == NULL) {
578 err = -ESRCH;
579 goto errout;
580 }
581
582 err = tb->tb_delete(tb, &cfg);
583 errout:
584 return err;
585 }
586
587 static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
588 {
589 struct net *net = skb->sk->sk_net;
590 struct fib_config cfg;
591 struct fib_table *tb;
592 int err;
593
594 if (net != &init_net)
595 return -EINVAL;
596
597 err = rtm_to_fib_config(skb, nlh, &cfg);
598 if (err < 0)
599 goto errout;
600
601 tb = fib_new_table(&init_net, cfg.fc_table);
602 if (tb == NULL) {
603 err = -ENOBUFS;
604 goto errout;
605 }
606
607 err = tb->tb_insert(tb, &cfg);
608 errout:
609 return err;
610 }
611
612 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
613 {
614 struct net *net = skb->sk->sk_net;
615 unsigned int h, s_h;
616 unsigned int e = 0, s_e;
617 struct fib_table *tb;
618 struct hlist_node *node;
619 int dumped = 0;
620
621 if (net != &init_net)
622 return 0;
623
624 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
625 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
626 return ip_rt_dump(skb, cb);
627
628 s_h = cb->args[0];
629 s_e = cb->args[1];
630
631 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
632 e = 0;
633 hlist_for_each_entry(tb, node, &fib_table_hash[h], tb_hlist) {
634 if (e < s_e)
635 goto next;
636 if (dumped)
637 memset(&cb->args[2], 0, sizeof(cb->args) -
638 2 * sizeof(cb->args[0]));
639 if (tb->tb_dump(tb, skb, cb) < 0)
640 goto out;
641 dumped = 1;
642 next:
643 e++;
644 }
645 }
646 out:
647 cb->args[1] = e;
648 cb->args[0] = h;
649
650 return skb->len;
651 }
652
653 /* Prepare and feed intra-kernel routing request.
654 Really, it should be netlink message, but :-( netlink
655 can be not configured, so that we feed it directly
656 to fib engine. It is legal, because all events occur
657 only when netlink is already locked.
658 */
659
660 static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
661 {
662 struct fib_table *tb;
663 struct fib_config cfg = {
664 .fc_protocol = RTPROT_KERNEL,
665 .fc_type = type,
666 .fc_dst = dst,
667 .fc_dst_len = dst_len,
668 .fc_prefsrc = ifa->ifa_local,
669 .fc_oif = ifa->ifa_dev->dev->ifindex,
670 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
671 };
672
673 if (type == RTN_UNICAST)
674 tb = fib_new_table(&init_net, RT_TABLE_MAIN);
675 else
676 tb = fib_new_table(&init_net, RT_TABLE_LOCAL);
677
678 if (tb == NULL)
679 return;
680
681 cfg.fc_table = tb->tb_id;
682
683 if (type != RTN_LOCAL)
684 cfg.fc_scope = RT_SCOPE_LINK;
685 else
686 cfg.fc_scope = RT_SCOPE_HOST;
687
688 if (cmd == RTM_NEWROUTE)
689 tb->tb_insert(tb, &cfg);
690 else
691 tb->tb_delete(tb, &cfg);
692 }
693
694 void fib_add_ifaddr(struct in_ifaddr *ifa)
695 {
696 struct in_device *in_dev = ifa->ifa_dev;
697 struct net_device *dev = in_dev->dev;
698 struct in_ifaddr *prim = ifa;
699 __be32 mask = ifa->ifa_mask;
700 __be32 addr = ifa->ifa_local;
701 __be32 prefix = ifa->ifa_address&mask;
702
703 if (ifa->ifa_flags&IFA_F_SECONDARY) {
704 prim = inet_ifa_byprefix(in_dev, prefix, mask);
705 if (prim == NULL) {
706 printk(KERN_DEBUG "fib_add_ifaddr: bug: prim == NULL\n");
707 return;
708 }
709 }
710
711 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
712
713 if (!(dev->flags&IFF_UP))
714 return;
715
716 /* Add broadcast address, if it is explicitly assigned. */
717 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
718 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
719
720 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags&IFA_F_SECONDARY) &&
721 (prefix != addr || ifa->ifa_prefixlen < 32)) {
722 fib_magic(RTM_NEWROUTE, dev->flags&IFF_LOOPBACK ? RTN_LOCAL :
723 RTN_UNICAST, prefix, ifa->ifa_prefixlen, prim);
724
725 /* Add network specific broadcasts, when it takes a sense */
726 if (ifa->ifa_prefixlen < 31) {
727 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
728 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix|~mask, 32, prim);
729 }
730 }
731 }
732
733 static void fib_del_ifaddr(struct in_ifaddr *ifa)
734 {
735 struct in_device *in_dev = ifa->ifa_dev;
736 struct net_device *dev = in_dev->dev;
737 struct in_ifaddr *ifa1;
738 struct in_ifaddr *prim = ifa;
739 __be32 brd = ifa->ifa_address|~ifa->ifa_mask;
740 __be32 any = ifa->ifa_address&ifa->ifa_mask;
741 #define LOCAL_OK 1
742 #define BRD_OK 2
743 #define BRD0_OK 4
744 #define BRD1_OK 8
745 unsigned ok = 0;
746
747 if (!(ifa->ifa_flags&IFA_F_SECONDARY))
748 fib_magic(RTM_DELROUTE, dev->flags&IFF_LOOPBACK ? RTN_LOCAL :
749 RTN_UNICAST, any, ifa->ifa_prefixlen, prim);
750 else {
751 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
752 if (prim == NULL) {
753 printk(KERN_DEBUG "fib_del_ifaddr: bug: prim == NULL\n");
754 return;
755 }
756 }
757
758 /* Deletion is more complicated than add.
759 We should take care of not to delete too much :-)
760
761 Scan address list to be sure that addresses are really gone.
762 */
763
764 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
765 if (ifa->ifa_local == ifa1->ifa_local)
766 ok |= LOCAL_OK;
767 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
768 ok |= BRD_OK;
769 if (brd == ifa1->ifa_broadcast)
770 ok |= BRD1_OK;
771 if (any == ifa1->ifa_broadcast)
772 ok |= BRD0_OK;
773 }
774
775 if (!(ok&BRD_OK))
776 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
777 if (!(ok&BRD1_OK))
778 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
779 if (!(ok&BRD0_OK))
780 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
781 if (!(ok&LOCAL_OK)) {
782 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
783
784 /* Check, that this local address finally disappeared. */
785 if (inet_addr_type(ifa->ifa_local) != RTN_LOCAL) {
786 /* And the last, but not the least thing.
787 We must flush stray FIB entries.
788
789 First of all, we scan fib_info list searching
790 for stray nexthop entries, then ignite fib_flush.
791 */
792 if (fib_sync_down(ifa->ifa_local, NULL, 0))
793 fib_flush();
794 }
795 }
796 #undef LOCAL_OK
797 #undef BRD_OK
798 #undef BRD0_OK
799 #undef BRD1_OK
800 }
801
802 static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb )
803 {
804
805 struct fib_result res;
806 struct flowi fl = { .mark = frn->fl_mark,
807 .nl_u = { .ip4_u = { .daddr = frn->fl_addr,
808 .tos = frn->fl_tos,
809 .scope = frn->fl_scope } } };
810
811 #ifdef CONFIG_IP_MULTIPLE_TABLES
812 res.r = NULL;
813 #endif
814
815 frn->err = -ENOENT;
816 if (tb) {
817 local_bh_disable();
818
819 frn->tb_id = tb->tb_id;
820 frn->err = tb->tb_lookup(tb, &fl, &res);
821
822 if (!frn->err) {
823 frn->prefixlen = res.prefixlen;
824 frn->nh_sel = res.nh_sel;
825 frn->type = res.type;
826 frn->scope = res.scope;
827 fib_res_put(&res);
828 }
829 local_bh_enable();
830 }
831 }
832
833 static void nl_fib_input(struct sk_buff *skb)
834 {
835 struct fib_result_nl *frn;
836 struct nlmsghdr *nlh;
837 struct fib_table *tb;
838 u32 pid;
839
840 nlh = nlmsg_hdr(skb);
841 if (skb->len < NLMSG_SPACE(0) || skb->len < nlh->nlmsg_len ||
842 nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*frn)))
843 return;
844
845 skb = skb_clone(skb, GFP_KERNEL);
846 if (skb == NULL)
847 return;
848 nlh = nlmsg_hdr(skb);
849
850 frn = (struct fib_result_nl *) NLMSG_DATA(nlh);
851 tb = fib_get_table(&init_net, frn->tb_id_in);
852
853 nl_fib_lookup(frn, tb);
854
855 pid = NETLINK_CB(skb).pid; /* pid of sending process */
856 NETLINK_CB(skb).pid = 0; /* from kernel */
857 NETLINK_CB(skb).dst_group = 0; /* unicast */
858 netlink_unicast(fibnl, skb, pid, MSG_DONTWAIT);
859 }
860
861 static int nl_fib_lookup_init(struct net *net)
862 {
863 fibnl = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, 0,
864 nl_fib_input, NULL, THIS_MODULE);
865 if (fibnl == NULL)
866 return -EAFNOSUPPORT;
867 return 0;
868 }
869
870 static void nl_fib_lookup_exit(struct net *net)
871 {
872 sock_put(fibnl);
873 }
874
875 static void fib_disable_ip(struct net_device *dev, int force)
876 {
877 if (fib_sync_down(0, dev, force))
878 fib_flush();
879 rt_cache_flush(0);
880 arp_ifdown(dev);
881 }
882
883 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
884 {
885 struct in_ifaddr *ifa = (struct in_ifaddr*)ptr;
886
887 switch (event) {
888 case NETDEV_UP:
889 fib_add_ifaddr(ifa);
890 #ifdef CONFIG_IP_ROUTE_MULTIPATH
891 fib_sync_up(ifa->ifa_dev->dev);
892 #endif
893 rt_cache_flush(-1);
894 break;
895 case NETDEV_DOWN:
896 fib_del_ifaddr(ifa);
897 if (ifa->ifa_dev->ifa_list == NULL) {
898 /* Last address was deleted from this interface.
899 Disable IP.
900 */
901 fib_disable_ip(ifa->ifa_dev->dev, 1);
902 } else {
903 rt_cache_flush(-1);
904 }
905 break;
906 }
907 return NOTIFY_DONE;
908 }
909
910 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
911 {
912 struct net_device *dev = ptr;
913 struct in_device *in_dev = __in_dev_get_rtnl(dev);
914
915 if (dev->nd_net != &init_net)
916 return NOTIFY_DONE;
917
918 if (event == NETDEV_UNREGISTER) {
919 fib_disable_ip(dev, 2);
920 return NOTIFY_DONE;
921 }
922
923 if (!in_dev)
924 return NOTIFY_DONE;
925
926 switch (event) {
927 case NETDEV_UP:
928 for_ifa(in_dev) {
929 fib_add_ifaddr(ifa);
930 } endfor_ifa(in_dev);
931 #ifdef CONFIG_IP_ROUTE_MULTIPATH
932 fib_sync_up(dev);
933 #endif
934 rt_cache_flush(-1);
935 break;
936 case NETDEV_DOWN:
937 fib_disable_ip(dev, 0);
938 break;
939 case NETDEV_CHANGEMTU:
940 case NETDEV_CHANGE:
941 rt_cache_flush(0);
942 break;
943 }
944 return NOTIFY_DONE;
945 }
946
947 static struct notifier_block fib_inetaddr_notifier = {
948 .notifier_call =fib_inetaddr_event,
949 };
950
951 static struct notifier_block fib_netdev_notifier = {
952 .notifier_call =fib_netdev_event,
953 };
954
955 static int __net_init ip_fib_net_init(struct net *net)
956 {
957 unsigned int i;
958
959 for (i = 0; i < FIB_TABLE_HASHSZ; i++)
960 INIT_HLIST_HEAD(&fib_table_hash[i]);
961
962 return fib4_rules_init(net);
963 }
964
965 static void __net_exit ip_fib_net_exit(struct net *net)
966 {
967 unsigned int i;
968
969 #ifdef CONFIG_IP_MULTIPLE_TABLES
970 fib4_rules_exit(net);
971 #endif
972
973 for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
974 struct fib_table *tb;
975 struct hlist_head *head;
976 struct hlist_node *node, *tmp;
977
978 head = &fib_table_hash[i];
979 hlist_for_each_entry_safe(tb, node, tmp, head, tb_hlist) {
980 hlist_del(node);
981 tb->tb_flush(tb);
982 kfree(tb);
983 }
984 }
985 }
986
987 static int __net_init fib_net_init(struct net *net)
988 {
989 int error;
990
991 error = 0;
992 if (net != &init_net)
993 goto out;
994
995 error = ip_fib_net_init(net);
996 if (error < 0)
997 goto out;
998 error = nl_fib_lookup_init(net);
999 if (error < 0)
1000 goto out_nlfl;
1001 error = fib_proc_init(net);
1002 if (error < 0)
1003 goto out_proc;
1004 out:
1005 return error;
1006
1007 out_proc:
1008 nl_fib_lookup_exit(net);
1009 out_nlfl:
1010 ip_fib_net_exit(net);
1011 goto out;
1012 }
1013
1014 static void __net_exit fib_net_exit(struct net *net)
1015 {
1016 fib_proc_exit(net);
1017 nl_fib_lookup_exit(net);
1018 ip_fib_net_exit(net);
1019 }
1020
1021 static struct pernet_operations fib_net_ops = {
1022 .init = fib_net_init,
1023 .exit = fib_net_exit,
1024 };
1025
1026 void __init ip_fib_init(void)
1027 {
1028 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL);
1029 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL);
1030 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib);
1031
1032 register_pernet_subsys(&fib_net_ops);
1033 register_netdevice_notifier(&fib_netdev_notifier);
1034 register_inetaddr_notifier(&fib_inetaddr_notifier);
1035 }
1036
1037 EXPORT_SYMBOL(inet_addr_type);
1038 EXPORT_SYMBOL(inet_dev_addr_type);
1039 EXPORT_SYMBOL(ip_dev_find);
kernel-based virtual machine