?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
route: Mark unused routing attributes as such
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv4 / fib_frontend.c
blob0b2ac6a3d903cbdeaa6d83e37b0d10d309b7b16f
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 #ifndef CONFIG_IP_MULTIPLE_TABLES
51
52 static int __net_init fib4_rules_init(struct net *net)
53 {
54 struct fib_table *local_table, *main_table;
55
56 local_table = fib_hash_table(RT_TABLE_LOCAL);
57 if (local_table == NULL)
58 return -ENOMEM;
59
60 main_table = fib_hash_table(RT_TABLE_MAIN);
61 if (main_table == NULL)
62 goto fail;
63
64 hlist_add_head_rcu(&local_table->tb_hlist,
65 &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]);
66 hlist_add_head_rcu(&main_table->tb_hlist,
67 &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]);
68 return 0;
69
70 fail:
71 kfree(local_table);
72 return -ENOMEM;
73 }
74 #else
75
76 struct fib_table *fib_new_table(struct net *net, u32 id)
77 {
78 struct fib_table *tb;
79 unsigned int h;
80
81 if (id == 0)
82 id = RT_TABLE_MAIN;
83 tb = fib_get_table(net, id);
84 if (tb)
85 return tb;
86
87 tb = fib_hash_table(id);
88 if (!tb)
89 return NULL;
90 h = id & (FIB_TABLE_HASHSZ - 1);
91 hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]);
92 return tb;
93 }
94
95 struct fib_table *fib_get_table(struct net *net, u32 id)
96 {
97 struct fib_table *tb;
98 struct hlist_node *node;
99 struct hlist_head *head;
100 unsigned int h;
101
102 if (id == 0)
103 id = RT_TABLE_MAIN;
104 h = id & (FIB_TABLE_HASHSZ - 1);
105
106 rcu_read_lock();
107 head = &net->ipv4.fib_table_hash[h];
108 hlist_for_each_entry_rcu(tb, node, head, tb_hlist) {
109 if (tb->tb_id == id) {
110 rcu_read_unlock();
111 return tb;
112 }
113 }
114 rcu_read_unlock();
115 return NULL;
116 }
117 #endif /* CONFIG_IP_MULTIPLE_TABLES */
118
119 void fib_select_default(struct net *net,
120 const struct flowi *flp, struct fib_result *res)
121 {
122 struct fib_table *tb;
123 int table = RT_TABLE_MAIN;
124 #ifdef CONFIG_IP_MULTIPLE_TABLES
125 if (res->r == NULL || res->r->action != FR_ACT_TO_TBL)
126 return;
127 table = res->r->table;
128 #endif
129 tb = fib_get_table(net, table);
130 if (FIB_RES_GW(*res) && FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK)
131 tb->tb_select_default(tb, flp, res);
132 }
133
134 static void fib_flush(struct net *net)
135 {
136 int flushed = 0;
137 struct fib_table *tb;
138 struct hlist_node *node;
139 struct hlist_head *head;
140 unsigned int h;
141
142 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
143 head = &net->ipv4.fib_table_hash[h];
144 hlist_for_each_entry(tb, node, head, tb_hlist)
145 flushed += tb->tb_flush(tb);
146 }
147
148 if (flushed)
149 rt_cache_flush(-1);
150 }
151
152 /*
153 * Find the first device with a given source address.
154 */
155
156 struct net_device * ip_dev_find(struct net *net, __be32 addr)
157 {
158 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } };
159 struct fib_result res;
160 struct net_device *dev = NULL;
161 struct fib_table *local_table;
162
163 #ifdef CONFIG_IP_MULTIPLE_TABLES
164 res.r = NULL;
165 #endif
166
167 local_table = fib_get_table(net, RT_TABLE_LOCAL);
168 if (!local_table || local_table->tb_lookup(local_table, &fl, &res))
169 return NULL;
170 if (res.type != RTN_LOCAL)
171 goto out;
172 dev = FIB_RES_DEV(res);
173
174 if (dev)
175 dev_hold(dev);
176 out:
177 fib_res_put(&res);
178 return dev;
179 }
180
181 /*
182 * Find address type as if only "dev" was present in the system. If
183 * on_dev is NULL then all interfaces are taken into consideration.
184 */
185 static inline unsigned __inet_dev_addr_type(struct net *net,
186 const struct net_device *dev,
187 __be32 addr)
188 {
189 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } };
190 struct fib_result res;
191 unsigned ret = RTN_BROADCAST;
192 struct fib_table *local_table;
193
194 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
195 return RTN_BROADCAST;
196 if (ipv4_is_multicast(addr))
197 return RTN_MULTICAST;
198
199 #ifdef CONFIG_IP_MULTIPLE_TABLES
200 res.r = NULL;
201 #endif
202
203 local_table = fib_get_table(net, RT_TABLE_LOCAL);
204 if (local_table) {
205 ret = RTN_UNICAST;
206 if (!local_table->tb_lookup(local_table, &fl, &res)) {
207 if (!dev || dev == res.fi->fib_dev)
208 ret = res.type;
209 fib_res_put(&res);
210 }
211 }
212 return ret;
213 }
214
215 unsigned int inet_addr_type(struct net *net, __be32 addr)
216 {
217 return __inet_dev_addr_type(net, NULL, addr);
218 }
219
220 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
221 __be32 addr)
222 {
223 return __inet_dev_addr_type(net, dev, addr);
224 }
225
226 /* Given (packet source, input interface) and optional (dst, oif, tos):
227 - (main) check, that source is valid i.e. not broadcast or our local
228 address.
229 - figure out what "logical" interface this packet arrived
230 and calculate "specific destination" address.
231 - check, that packet arrived from expected physical interface.
232 */
233
234 int fib_validate_source(__be32 src, __be32 dst, u8 tos, int oif,
235 struct net_device *dev, __be32 *spec_dst, u32 *itag)
236 {
237 struct in_device *in_dev;
238 struct flowi fl = { .nl_u = { .ip4_u =
239 { .daddr = src,
240 .saddr = dst,
241 .tos = tos } },
242 .iif = oif };
243 struct fib_result res;
244 int no_addr, rpf;
245 int ret;
246 struct net *net;
247
248 no_addr = rpf = 0;
249 rcu_read_lock();
250 in_dev = __in_dev_get_rcu(dev);
251 if (in_dev) {
252 no_addr = in_dev->ifa_list == NULL;
253 rpf = IN_DEV_RPFILTER(in_dev);
254 }
255 rcu_read_unlock();
256
257 if (in_dev == NULL)
258 goto e_inval;
259
260 net = dev_net(dev);
261 if (fib_lookup(net, &fl, &res))
262 goto last_resort;
263 if (res.type != RTN_UNICAST)
264 goto e_inval_res;
265 *spec_dst = FIB_RES_PREFSRC(res);
266 fib_combine_itag(itag, &res);
267 #ifdef CONFIG_IP_ROUTE_MULTIPATH
268 if (FIB_RES_DEV(res) == dev || res.fi->fib_nhs > 1)
269 #else
270 if (FIB_RES_DEV(res) == dev)
271 #endif
272 {
273 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
274 fib_res_put(&res);
275 return ret;
276 }
277 fib_res_put(&res);
278 if (no_addr)
279 goto last_resort;
280 if (rpf)
281 goto e_inval;
282 fl.oif = dev->ifindex;
283
284 ret = 0;
285 if (fib_lookup(net, &fl, &res) == 0) {
286 if (res.type == RTN_UNICAST) {
287 *spec_dst = FIB_RES_PREFSRC(res);
288 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
289 }
290 fib_res_put(&res);
291 }
292 return ret;
293
294 last_resort:
295 if (rpf)
296 goto e_inval;
297 *spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE);
298 *itag = 0;
299 return 0;
300
301 e_inval_res:
302 fib_res_put(&res);
303 e_inval:
304 return -EINVAL;
305 }
306
307 static inline __be32 sk_extract_addr(struct sockaddr *addr)
308 {
309 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
310 }
311
312 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
313 {
314 struct nlattr *nla;
315
316 nla = (struct nlattr *) ((char *) mx + len);
317 nla->nla_type = type;
318 nla->nla_len = nla_attr_size(4);
319 *(u32 *) nla_data(nla) = value;
320
321 return len + nla_total_size(4);
322 }
323
324 static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
325 struct fib_config *cfg)
326 {
327 __be32 addr;
328 int plen;
329
330 memset(cfg, 0, sizeof(*cfg));
331 cfg->fc_nlinfo.nl_net = net;
332
333 if (rt->rt_dst.sa_family != AF_INET)
334 return -EAFNOSUPPORT;
335
336 /*
337 * Check mask for validity:
338 * a) it must be contiguous.
339 * b) destination must have all host bits clear.
340 * c) if application forgot to set correct family (AF_INET),
341 * reject request unless it is absolutely clear i.e.
342 * both family and mask are zero.
343 */
344 plen = 32;
345 addr = sk_extract_addr(&rt->rt_dst);
346 if (!(rt->rt_flags & RTF_HOST)) {
347 __be32 mask = sk_extract_addr(&rt->rt_genmask);
348
349 if (rt->rt_genmask.sa_family != AF_INET) {
350 if (mask || rt->rt_genmask.sa_family)
351 return -EAFNOSUPPORT;
352 }
353
354 if (bad_mask(mask, addr))
355 return -EINVAL;
356
357 plen = inet_mask_len(mask);
358 }
359
360 cfg->fc_dst_len = plen;
361 cfg->fc_dst = addr;
362
363 if (cmd != SIOCDELRT) {
364 cfg->fc_nlflags = NLM_F_CREATE;
365 cfg->fc_protocol = RTPROT_BOOT;
366 }
367
368 if (rt->rt_metric)
369 cfg->fc_priority = rt->rt_metric - 1;
370
371 if (rt->rt_flags & RTF_REJECT) {
372 cfg->fc_scope = RT_SCOPE_HOST;
373 cfg->fc_type = RTN_UNREACHABLE;
374 return 0;
375 }
376
377 cfg->fc_scope = RT_SCOPE_NOWHERE;
378 cfg->fc_type = RTN_UNICAST;
379
380 if (rt->rt_dev) {
381 char *colon;
382 struct net_device *dev;
383 char devname[IFNAMSIZ];
384
385 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
386 return -EFAULT;
387
388 devname[IFNAMSIZ-1] = 0;
389 colon = strchr(devname, ':');
390 if (colon)
391 *colon = 0;
392 dev = __dev_get_by_name(net, devname);
393 if (!dev)
394 return -ENODEV;
395 cfg->fc_oif = dev->ifindex;
396 if (colon) {
397 struct in_ifaddr *ifa;
398 struct in_device *in_dev = __in_dev_get_rtnl(dev);
399 if (!in_dev)
400 return -ENODEV;
401 *colon = ':';
402 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
403 if (strcmp(ifa->ifa_label, devname) == 0)
404 break;
405 if (ifa == NULL)
406 return -ENODEV;
407 cfg->fc_prefsrc = ifa->ifa_local;
408 }
409 }
410
411 addr = sk_extract_addr(&rt->rt_gateway);
412 if (rt->rt_gateway.sa_family == AF_INET && addr) {
413 cfg->fc_gw = addr;
414 if (rt->rt_flags & RTF_GATEWAY &&
415 inet_addr_type(net, addr) == RTN_UNICAST)
416 cfg->fc_scope = RT_SCOPE_UNIVERSE;
417 }
418
419 if (cmd == SIOCDELRT)
420 return 0;
421
422 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
423 return -EINVAL;
424
425 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
426 cfg->fc_scope = RT_SCOPE_LINK;
427
428 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
429 struct nlattr *mx;
430 int len = 0;
431
432 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
433 if (mx == NULL)
434 return -ENOMEM;
435
436 if (rt->rt_flags & RTF_MTU)
437 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
438
439 if (rt->rt_flags & RTF_WINDOW)
440 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
441
442 if (rt->rt_flags & RTF_IRTT)
443 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
444
445 cfg->fc_mx = mx;
446 cfg->fc_mx_len = len;
447 }
448
449 return 0;
450 }
451
452 /*
453 * Handle IP routing ioctl calls. These are used to manipulate the routing tables
454 */
455
456 int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
457 {
458 struct fib_config cfg;
459 struct rtentry rt;
460 int err;
461
462 switch (cmd) {
463 case SIOCADDRT: /* Add a route */
464 case SIOCDELRT: /* Delete a route */
465 if (!capable(CAP_NET_ADMIN))
466 return -EPERM;
467
468 if (copy_from_user(&rt, arg, sizeof(rt)))
469 return -EFAULT;
470
471 rtnl_lock();
472 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
473 if (err == 0) {
474 struct fib_table *tb;
475
476 if (cmd == SIOCDELRT) {
477 tb = fib_get_table(net, cfg.fc_table);
478 if (tb)
479 err = tb->tb_delete(tb, &cfg);
480 else
481 err = -ESRCH;
482 } else {
483 tb = fib_new_table(net, cfg.fc_table);
484 if (tb)
485 err = tb->tb_insert(tb, &cfg);
486 else
487 err = -ENOBUFS;
488 }
489
490 /* allocated by rtentry_to_fib_config() */
491 kfree(cfg.fc_mx);
492 }
493 rtnl_unlock();
494 return err;
495 }
496 return -EINVAL;
497 }
498
499 const struct nla_policy rtm_ipv4_policy[RTA_MAX+1] = {
500 [RTA_DST] = { .type = NLA_U32 },
501 [RTA_SRC] = { .type = NLA_U32 },
502 [RTA_IIF] = { .type = NLA_U32 },
503 [RTA_OIF] = { .type = NLA_U32 },
504 [RTA_GATEWAY] = { .type = NLA_U32 },
505 [RTA_PRIORITY] = { .type = NLA_U32 },
506 [RTA_PREFSRC] = { .type = NLA_U32 },
507 [RTA_METRICS] = { .type = NLA_NESTED },
508 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
509 [RTA_FLOW] = { .type = NLA_U32 },
510 };
511
512 static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
513 struct nlmsghdr *nlh, struct fib_config *cfg)
514 {
515 struct nlattr *attr;
516 int err, remaining;
517 struct rtmsg *rtm;
518
519 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
520 if (err < 0)
521 goto errout;
522
523 memset(cfg, 0, sizeof(*cfg));
524
525 rtm = nlmsg_data(nlh);
526 cfg->fc_dst_len = rtm->rtm_dst_len;
527 cfg->fc_tos = rtm->rtm_tos;
528 cfg->fc_table = rtm->rtm_table;
529 cfg->fc_protocol = rtm->rtm_protocol;
530 cfg->fc_scope = rtm->rtm_scope;
531 cfg->fc_type = rtm->rtm_type;
532 cfg->fc_flags = rtm->rtm_flags;
533 cfg->fc_nlflags = nlh->nlmsg_flags;
534
535 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
536 cfg->fc_nlinfo.nlh = nlh;
537 cfg->fc_nlinfo.nl_net = net;
538
539 if (cfg->fc_type > RTN_MAX) {
540 err = -EINVAL;
541 goto errout;
542 }
543
544 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
545 switch (nla_type(attr)) {
546 case RTA_DST:
547 cfg->fc_dst = nla_get_be32(attr);
548 break;
549 case RTA_OIF:
550 cfg->fc_oif = nla_get_u32(attr);
551 break;
552 case RTA_GATEWAY:
553 cfg->fc_gw = nla_get_be32(attr);
554 break;
555 case RTA_PRIORITY:
556 cfg->fc_priority = nla_get_u32(attr);
557 break;
558 case RTA_PREFSRC:
559 cfg->fc_prefsrc = nla_get_be32(attr);
560 break;
561 case RTA_METRICS:
562 cfg->fc_mx = nla_data(attr);
563 cfg->fc_mx_len = nla_len(attr);
564 break;
565 case RTA_MULTIPATH:
566 cfg->fc_mp = nla_data(attr);
567 cfg->fc_mp_len = nla_len(attr);
568 break;
569 case RTA_FLOW:
570 cfg->fc_flow = nla_get_u32(attr);
571 break;
572 case RTA_TABLE:
573 cfg->fc_table = nla_get_u32(attr);
574 break;
575 }
576 }
577
578 return 0;
579 errout:
580 return err;
581 }
582
583 static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
584 {
585 struct net *net = sock_net(skb->sk);
586 struct fib_config cfg;
587 struct fib_table *tb;
588 int err;
589
590 err = rtm_to_fib_config(net, skb, nlh, &cfg);
591 if (err < 0)
592 goto errout;
593
594 tb = fib_get_table(net, cfg.fc_table);
595 if (tb == NULL) {
596 err = -ESRCH;
597 goto errout;
598 }
599
600 err = tb->tb_delete(tb, &cfg);
601 errout:
602 return err;
603 }
604
605 static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
606 {
607 struct net *net = sock_net(skb->sk);
608 struct fib_config cfg;
609 struct fib_table *tb;
610 int err;
611
612 err = rtm_to_fib_config(net, skb, nlh, &cfg);
613 if (err < 0)
614 goto errout;
615
616 tb = fib_new_table(net, cfg.fc_table);
617 if (tb == NULL) {
618 err = -ENOBUFS;
619 goto errout;
620 }
621
622 err = tb->tb_insert(tb, &cfg);
623 errout:
624 return err;
625 }
626
627 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
628 {
629 struct net *net = sock_net(skb->sk);
630 unsigned int h, s_h;
631 unsigned int e = 0, s_e;
632 struct fib_table *tb;
633 struct hlist_node *node;
634 struct hlist_head *head;
635 int dumped = 0;
636
637 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
638 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
639 return ip_rt_dump(skb, cb);
640
641 s_h = cb->args[0];
642 s_e = cb->args[1];
643
644 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
645 e = 0;
646 head = &net->ipv4.fib_table_hash[h];
647 hlist_for_each_entry(tb, node, head, tb_hlist) {
648 if (e < s_e)
649 goto next;
650 if (dumped)
651 memset(&cb->args[2], 0, sizeof(cb->args) -
652 2 * sizeof(cb->args[0]));
653 if (tb->tb_dump(tb, skb, cb) < 0)
654 goto out;
655 dumped = 1;
656 next:
657 e++;
658 }
659 }
660 out:
661 cb->args[1] = e;
662 cb->args[0] = h;
663
664 return skb->len;
665 }
666
667 /* Prepare and feed intra-kernel routing request.
668 Really, it should be netlink message, but :-( netlink
669 can be not configured, so that we feed it directly
670 to fib engine. It is legal, because all events occur
671 only when netlink is already locked.
672 */
673
674 static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
675 {
676 struct net *net = dev_net(ifa->ifa_dev->dev);
677 struct fib_table *tb;
678 struct fib_config cfg = {
679 .fc_protocol = RTPROT_KERNEL,
680 .fc_type = type,
681 .fc_dst = dst,
682 .fc_dst_len = dst_len,
683 .fc_prefsrc = ifa->ifa_local,
684 .fc_oif = ifa->ifa_dev->dev->ifindex,
685 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
686 .fc_nlinfo = {
687 .nl_net = net,
688 },
689 };
690
691 if (type == RTN_UNICAST)
692 tb = fib_new_table(net, RT_TABLE_MAIN);
693 else
694 tb = fib_new_table(net, RT_TABLE_LOCAL);
695
696 if (tb == NULL)
697 return;
698
699 cfg.fc_table = tb->tb_id;
700
701 if (type != RTN_LOCAL)
702 cfg.fc_scope = RT_SCOPE_LINK;
703 else
704 cfg.fc_scope = RT_SCOPE_HOST;
705
706 if (cmd == RTM_NEWROUTE)
707 tb->tb_insert(tb, &cfg);
708 else
709 tb->tb_delete(tb, &cfg);
710 }
711
712 void fib_add_ifaddr(struct in_ifaddr *ifa)
713 {
714 struct in_device *in_dev = ifa->ifa_dev;
715 struct net_device *dev = in_dev->dev;
716 struct in_ifaddr *prim = ifa;
717 __be32 mask = ifa->ifa_mask;
718 __be32 addr = ifa->ifa_local;
719 __be32 prefix = ifa->ifa_address&mask;
720
721 if (ifa->ifa_flags&IFA_F_SECONDARY) {
722 prim = inet_ifa_byprefix(in_dev, prefix, mask);
723 if (prim == NULL) {
724 printk(KERN_WARNING "fib_add_ifaddr: bug: prim == NULL\n");
725 return;
726 }
727 }
728
729 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
730
731 if (!(dev->flags&IFF_UP))
732 return;
733
734 /* Add broadcast address, if it is explicitly assigned. */
735 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
736 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
737
738 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags&IFA_F_SECONDARY) &&
739 (prefix != addr || ifa->ifa_prefixlen < 32)) {
740 fib_magic(RTM_NEWROUTE, dev->flags&IFF_LOOPBACK ? RTN_LOCAL :
741 RTN_UNICAST, prefix, ifa->ifa_prefixlen, prim);
742
743 /* Add network specific broadcasts, when it takes a sense */
744 if (ifa->ifa_prefixlen < 31) {
745 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
746 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix|~mask, 32, prim);
747 }
748 }
749 }
750
751 static void fib_del_ifaddr(struct in_ifaddr *ifa)
752 {
753 struct in_device *in_dev = ifa->ifa_dev;
754 struct net_device *dev = in_dev->dev;
755 struct in_ifaddr *ifa1;
756 struct in_ifaddr *prim = ifa;
757 __be32 brd = ifa->ifa_address|~ifa->ifa_mask;
758 __be32 any = ifa->ifa_address&ifa->ifa_mask;
759 #define LOCAL_OK 1
760 #define BRD_OK 2
761 #define BRD0_OK 4
762 #define BRD1_OK 8
763 unsigned ok = 0;
764
765 if (!(ifa->ifa_flags&IFA_F_SECONDARY))
766 fib_magic(RTM_DELROUTE, dev->flags&IFF_LOOPBACK ? RTN_LOCAL :
767 RTN_UNICAST, any, ifa->ifa_prefixlen, prim);
768 else {
769 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
770 if (prim == NULL) {
771 printk(KERN_WARNING "fib_del_ifaddr: bug: prim == NULL\n");
772 return;
773 }
774 }
775
776 /* Deletion is more complicated than add.
777 We should take care of not to delete too much :-)
778
779 Scan address list to be sure that addresses are really gone.
780 */
781
782 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
783 if (ifa->ifa_local == ifa1->ifa_local)
784 ok |= LOCAL_OK;
785 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
786 ok |= BRD_OK;
787 if (brd == ifa1->ifa_broadcast)
788 ok |= BRD1_OK;
789 if (any == ifa1->ifa_broadcast)
790 ok |= BRD0_OK;
791 }
792
793 if (!(ok&BRD_OK))
794 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
795 if (!(ok&BRD1_OK))
796 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
797 if (!(ok&BRD0_OK))
798 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
799 if (!(ok&LOCAL_OK)) {
800 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
801
802 /* Check, that this local address finally disappeared. */
803 if (inet_addr_type(dev_net(dev), ifa->ifa_local) != RTN_LOCAL) {
804 /* And the last, but not the least thing.
805 We must flush stray FIB entries.
806
807 First of all, we scan fib_info list searching
808 for stray nexthop entries, then ignite fib_flush.
809 */
810 if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local))
811 fib_flush(dev_net(dev));
812 }
813 }
814 #undef LOCAL_OK
815 #undef BRD_OK
816 #undef BRD0_OK
817 #undef BRD1_OK
818 }
819
820 static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb )
821 {
822
823 struct fib_result res;
824 struct flowi fl = { .mark = frn->fl_mark,
825 .nl_u = { .ip4_u = { .daddr = frn->fl_addr,
826 .tos = frn->fl_tos,
827 .scope = frn->fl_scope } } };
828
829 #ifdef CONFIG_IP_MULTIPLE_TABLES
830 res.r = NULL;
831 #endif
832
833 frn->err = -ENOENT;
834 if (tb) {
835 local_bh_disable();
836
837 frn->tb_id = tb->tb_id;
838 frn->err = tb->tb_lookup(tb, &fl, &res);
839
840 if (!frn->err) {
841 frn->prefixlen = res.prefixlen;
842 frn->nh_sel = res.nh_sel;
843 frn->type = res.type;
844 frn->scope = res.scope;
845 fib_res_put(&res);
846 }
847 local_bh_enable();
848 }
849 }
850
851 static void nl_fib_input(struct sk_buff *skb)
852 {
853 struct net *net;
854 struct fib_result_nl *frn;
855 struct nlmsghdr *nlh;
856 struct fib_table *tb;
857 u32 pid;
858
859 net = sock_net(skb->sk);
860 nlh = nlmsg_hdr(skb);
861 if (skb->len < NLMSG_SPACE(0) || skb->len < nlh->nlmsg_len ||
862 nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*frn)))
863 return;
864
865 skb = skb_clone(skb, GFP_KERNEL);
866 if (skb == NULL)
867 return;
868 nlh = nlmsg_hdr(skb);
869
870 frn = (struct fib_result_nl *) NLMSG_DATA(nlh);
871 tb = fib_get_table(net, frn->tb_id_in);
872
873 nl_fib_lookup(frn, tb);
874
875 pid = NETLINK_CB(skb).pid; /* pid of sending process */
876 NETLINK_CB(skb).pid = 0; /* from kernel */
877 NETLINK_CB(skb).dst_group = 0; /* unicast */
878 netlink_unicast(net->ipv4.fibnl, skb, pid, MSG_DONTWAIT);
879 }
880
881 static int nl_fib_lookup_init(struct net *net)
882 {
883 struct sock *sk;
884 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, 0,
885 nl_fib_input, NULL, THIS_MODULE);
886 if (sk == NULL)
887 return -EAFNOSUPPORT;
888 net->ipv4.fibnl = sk;
889 return 0;
890 }
891
892 static void nl_fib_lookup_exit(struct net *net)
893 {
894 netlink_kernel_release(net->ipv4.fibnl);
895 net->ipv4.fibnl = NULL;
896 }
897
898 static void fib_disable_ip(struct net_device *dev, int force)
899 {
900 if (fib_sync_down_dev(dev, force))
901 fib_flush(dev_net(dev));
902 rt_cache_flush(0);
903 arp_ifdown(dev);
904 }
905
906 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
907 {
908 struct in_ifaddr *ifa = (struct in_ifaddr*)ptr;
909
910 switch (event) {
911 case NETDEV_UP:
912 fib_add_ifaddr(ifa);
913 #ifdef CONFIG_IP_ROUTE_MULTIPATH
914 fib_sync_up(ifa->ifa_dev->dev);
915 #endif
916 rt_cache_flush(-1);
917 break;
918 case NETDEV_DOWN:
919 fib_del_ifaddr(ifa);
920 if (ifa->ifa_dev->ifa_list == NULL) {
921 /* Last address was deleted from this interface.
922 Disable IP.
923 */
924 fib_disable_ip(ifa->ifa_dev->dev, 1);
925 } else {
926 rt_cache_flush(-1);
927 }
928 break;
929 }
930 return NOTIFY_DONE;
931 }
932
933 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
934 {
935 struct net_device *dev = ptr;
936 struct in_device *in_dev = __in_dev_get_rtnl(dev);
937
938 if (event == NETDEV_UNREGISTER) {
939 fib_disable_ip(dev, 2);
940 return NOTIFY_DONE;
941 }
942
943 if (!in_dev)
944 return NOTIFY_DONE;
945
946 switch (event) {
947 case NETDEV_UP:
948 for_ifa(in_dev) {
949 fib_add_ifaddr(ifa);
950 } endfor_ifa(in_dev);
951 #ifdef CONFIG_IP_ROUTE_MULTIPATH
952 fib_sync_up(dev);
953 #endif
954 rt_cache_flush(-1);
955 break;
956 case NETDEV_DOWN:
957 fib_disable_ip(dev, 0);
958 break;
959 case NETDEV_CHANGEMTU:
960 case NETDEV_CHANGE:
961 rt_cache_flush(0);
962 break;
963 }
964 return NOTIFY_DONE;
965 }
966
967 static struct notifier_block fib_inetaddr_notifier = {
968 .notifier_call =fib_inetaddr_event,
969 };
970
971 static struct notifier_block fib_netdev_notifier = {
972 .notifier_call =fib_netdev_event,
973 };
974
975 static int __net_init ip_fib_net_init(struct net *net)
976 {
977 int err;
978 unsigned int i;
979
980 net->ipv4.fib_table_hash = kzalloc(
981 sizeof(struct hlist_head)*FIB_TABLE_HASHSZ, GFP_KERNEL);
982 if (net->ipv4.fib_table_hash == NULL)
983 return -ENOMEM;
984
985 for (i = 0; i < FIB_TABLE_HASHSZ; i++)
986 INIT_HLIST_HEAD(&net->ipv4.fib_table_hash[i]);
987
988 err = fib4_rules_init(net);
989 if (err < 0)
990 goto fail;
991 return 0;
992
993 fail:
994 kfree(net->ipv4.fib_table_hash);
995 return err;
996 }
997
998 static void __net_exit ip_fib_net_exit(struct net *net)
999 {
1000 unsigned int i;
1001
1002 #ifdef CONFIG_IP_MULTIPLE_TABLES
1003 fib4_rules_exit(net);
1004 #endif
1005
1006 for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
1007 struct fib_table *tb;
1008 struct hlist_head *head;
1009 struct hlist_node *node, *tmp;
1010
1011 head = &net->ipv4.fib_table_hash[i];
1012 hlist_for_each_entry_safe(tb, node, tmp, head, tb_hlist) {
1013 hlist_del(node);
1014 tb->tb_flush(tb);
1015 kfree(tb);
1016 }
1017 }
1018 kfree(net->ipv4.fib_table_hash);
1019 }
1020
1021 static int __net_init fib_net_init(struct net *net)
1022 {
1023 int error;
1024
1025 error = ip_fib_net_init(net);
1026 if (error < 0)
1027 goto out;
1028 error = nl_fib_lookup_init(net);
1029 if (error < 0)
1030 goto out_nlfl;
1031 error = fib_proc_init(net);
1032 if (error < 0)
1033 goto out_proc;
1034 out:
1035 return error;
1036
1037 out_proc:
1038 nl_fib_lookup_exit(net);
1039 out_nlfl:
1040 ip_fib_net_exit(net);
1041 goto out;
1042 }
1043
1044 static void __net_exit fib_net_exit(struct net *net)
1045 {
1046 fib_proc_exit(net);
1047 nl_fib_lookup_exit(net);
1048 ip_fib_net_exit(net);
1049 }
1050
1051 static struct pernet_operations fib_net_ops = {
1052 .init = fib_net_init,
1053 .exit = fib_net_exit,
1054 };
1055
1056 void __init ip_fib_init(void)
1057 {
1058 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL);
1059 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL);
1060 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib);
1061
1062 register_pernet_subsys(&fib_net_ops);
1063 register_netdevice_notifier(&fib_netdev_notifier);
1064 register_inetaddr_notifier(&fib_inetaddr_notifier);
1065
1066 fib_hash_init();
1067 }
1068
1069 EXPORT_SYMBOL(inet_addr_type);
1070 EXPORT_SYMBOL(inet_dev_addr_type);
1071 EXPORT_SYMBOL(ip_dev_find);
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree