]> bbs.cooldavid.org Git - net-next-2.6.git/blame_incremental - net/ipv4/fib_frontend.c
git://bbs.cooldavid.org / net-next-2.6.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
MIPS: Wire up recvmmsg syscall
[net-next-2.6.git] / net / ipv4 / fib_frontend.c
... / ...
CommitLineData
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
38#include <net/ip.h>
39#include <net/protocol.h>
40#include <net/route.h>
41#include <net/tcp.h>
42#include <net/sock.h>
43#include <net/arp.h>
44#include <net/ip_fib.h>
45#include <net/rtnetlink.h>
46
47#ifndef CONFIG_IP_MULTIPLE_TABLES
48
49static int __net_init fib4_rules_init(struct net *net)
50{
51 struct fib_table *local_table, *main_table;
52
53 local_table = fib_hash_table(RT_TABLE_LOCAL);
54 if (local_table == NULL)
55 return -ENOMEM;
56
57 main_table = fib_hash_table(RT_TABLE_MAIN);
58 if (main_table == NULL)
59 goto fail;
60
61 hlist_add_head_rcu(&local_table->tb_hlist,
62 &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]);
63 hlist_add_head_rcu(&main_table->tb_hlist,
64 &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]);
65 return 0;
66
67fail:
68 kfree(local_table);
69 return -ENOMEM;
70}
71#else
72
73struct fib_table *fib_new_table(struct net *net, u32 id)
74{
75 struct fib_table *tb;
76 unsigned int h;
77
78 if (id == 0)
79 id = RT_TABLE_MAIN;
80 tb = fib_get_table(net, id);
81 if (tb)
82 return tb;
83
84 tb = fib_hash_table(id);
85 if (!tb)
86 return NULL;
87 h = id & (FIB_TABLE_HASHSZ - 1);
88 hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]);
89 return tb;
90}
91
92struct fib_table *fib_get_table(struct net *net, u32 id)
93{
94 struct fib_table *tb;
95 struct hlist_node *node;
96 struct hlist_head *head;
97 unsigned int h;
98
99 if (id == 0)
100 id = RT_TABLE_MAIN;
101 h = id & (FIB_TABLE_HASHSZ - 1);
102
103 rcu_read_lock();
104 head = &net->ipv4.fib_table_hash[h];
105 hlist_for_each_entry_rcu(tb, node, head, tb_hlist) {
106 if (tb->tb_id == id) {
107 rcu_read_unlock();
108 return tb;
109 }
110 }
111 rcu_read_unlock();
112 return NULL;
113}
114#endif /* CONFIG_IP_MULTIPLE_TABLES */
115
116void fib_select_default(struct net *net,
117 const struct flowi *flp, struct fib_result *res)
118{
119 struct fib_table *tb;
120 int table = RT_TABLE_MAIN;
121#ifdef CONFIG_IP_MULTIPLE_TABLES
122 if (res->r == NULL || res->r->action != FR_ACT_TO_TBL)
123 return;
124 table = res->r->table;
125#endif
126 tb = fib_get_table(net, table);
127 if (FIB_RES_GW(*res) && FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK)
128 fib_table_select_default(tb, flp, res);
129}
130
131static void fib_flush(struct net *net)
132{
133 int flushed = 0;
134 struct fib_table *tb;
135 struct hlist_node *node;
136 struct hlist_head *head;
137 unsigned int h;
138
139 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
140 head = &net->ipv4.fib_table_hash[h];
141 hlist_for_each_entry(tb, node, head, tb_hlist)
142 flushed += fib_table_flush(tb);
143 }
144
145 if (flushed)
146 rt_cache_flush(net, -1);
147}
148
149/*
150 * Find the first device with a given source address.
151 */
152
153struct net_device * ip_dev_find(struct net *net, __be32 addr)
154{
155 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } };
156 struct fib_result res;
157 struct net_device *dev = NULL;
158 struct fib_table *local_table;
159
160#ifdef CONFIG_IP_MULTIPLE_TABLES
161 res.r = NULL;
162#endif
163
164 local_table = fib_get_table(net, RT_TABLE_LOCAL);
165 if (!local_table || fib_table_lookup(local_table, &fl, &res))
166 return NULL;
167 if (res.type != RTN_LOCAL)
168 goto out;
169 dev = FIB_RES_DEV(res);
170
171 if (dev)
172 dev_hold(dev);
173out:
174 fib_res_put(&res);
175 return dev;
176}
177
178/*
179 * Find address type as if only "dev" was present in the system. If
180 * on_dev is NULL then all interfaces are taken into consideration.
181 */
182static inline unsigned __inet_dev_addr_type(struct net *net,
183 const struct net_device *dev,
184 __be32 addr)
185{
186 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } };
187 struct fib_result res;
188 unsigned ret = RTN_BROADCAST;
189 struct fib_table *local_table;
190
191 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
192 return RTN_BROADCAST;
193 if (ipv4_is_multicast(addr))
194 return RTN_MULTICAST;
195
196#ifdef CONFIG_IP_MULTIPLE_TABLES
197 res.r = NULL;
198#endif
199
200 local_table = fib_get_table(net, RT_TABLE_LOCAL);
201 if (local_table) {
202 ret = RTN_UNICAST;
203 if (!fib_table_lookup(local_table, &fl, &res)) {
204 if (!dev || dev == res.fi->fib_dev)
205 ret = res.type;
206 fib_res_put(&res);
207 }
208 }
209 return ret;
210}
211
212unsigned int inet_addr_type(struct net *net, __be32 addr)
213{
214 return __inet_dev_addr_type(net, NULL, addr);
215}
216
217unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
218 __be32 addr)
219{
220 return __inet_dev_addr_type(net, dev, addr);
221}
222
223/* Given (packet source, input interface) and optional (dst, oif, tos):
224 - (main) check, that source is valid i.e. not broadcast or our local
225 address.
226 - figure out what "logical" interface this packet arrived
227 and calculate "specific destination" address.
228 - check, that packet arrived from expected physical interface.
229 */
230
231int fib_validate_source(__be32 src, __be32 dst, u8 tos, int oif,
232 struct net_device *dev, __be32 *spec_dst,
233 u32 *itag, u32 mark)
234{
235 struct in_device *in_dev;
236 struct flowi fl = { .nl_u = { .ip4_u =
237 { .daddr = src,
238 .saddr = dst,
239 .tos = tos } },
240 .mark = mark,
241 .iif = oif };
242
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 == 1)
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
294last_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
301e_inval_res:
302 fib_res_put(&res);
303e_inval:
304 return -EINVAL;
305}
306
307static inline __be32 sk_extract_addr(struct sockaddr *addr)
308{
309 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
310}
311
312static 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
324static 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
456int 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 = fib_table_delete(tb, &cfg);
480 else
481 err = -ESRCH;
482 } else {
483 tb = fib_new_table(net, cfg.fc_table);
484 if (tb)
485 err = fib_table_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
499const 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
512static 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;
579errout:
580 return err;
581}
582
583static 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 = fib_table_delete(tb, &cfg);
601errout:
602 return err;
603}
604
605static 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 = fib_table_insert(tb, &cfg);
623errout:
624 return err;
625}
626
627static 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 (fib_table_dump(tb, skb, cb) < 0)
654 goto out;
655 dumped = 1;
656next:
657 e++;
658 }
659 }
660out:
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
674static 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 fib_table_insert(tb, &cfg);
708 else
709 fib_table_delete(tb, &cfg);
710}
711
712void 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
751static 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
820static 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 = fib_table_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
851static 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
881static 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
892static void nl_fib_lookup_exit(struct net *net)
893{
894 netlink_kernel_release(net->ipv4.fibnl);
895 net->ipv4.fibnl = NULL;
896}
897
898static void fib_disable_ip(struct net_device *dev, int force, int delay)
899{
900 if (fib_sync_down_dev(dev, force))
901 fib_flush(dev_net(dev));
902 rt_cache_flush(dev_net(dev), delay);
903 arp_ifdown(dev);
904}
905
906static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
907{
908 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
909 struct net_device *dev = ifa->ifa_dev->dev;
910
911 switch (event) {
912 case NETDEV_UP:
913 fib_add_ifaddr(ifa);
914#ifdef CONFIG_IP_ROUTE_MULTIPATH
915 fib_sync_up(dev);
916#endif
917 rt_cache_flush(dev_net(dev), -1);
918 break;
919 case NETDEV_DOWN:
920 fib_del_ifaddr(ifa);
921 if (ifa->ifa_dev->ifa_list == NULL) {
922 /* Last address was deleted from this interface.
923 Disable IP.
924 */
925 fib_disable_ip(dev, 1, 0);
926 } else {
927 rt_cache_flush(dev_net(dev), -1);
928 }
929 break;
930 }
931 return NOTIFY_DONE;
932}
933
934static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
935{
936 struct net_device *dev = ptr;
937 struct in_device *in_dev = __in_dev_get_rtnl(dev);
938
939 if (event == NETDEV_UNREGISTER) {
940 fib_disable_ip(dev, 2, -1);
941 return NOTIFY_DONE;
942 }
943
944 if (!in_dev)
945 return NOTIFY_DONE;
946
947 switch (event) {
948 case NETDEV_UP:
949 for_ifa(in_dev) {
950 fib_add_ifaddr(ifa);
951 } endfor_ifa(in_dev);
952#ifdef CONFIG_IP_ROUTE_MULTIPATH
953 fib_sync_up(dev);
954#endif
955 rt_cache_flush(dev_net(dev), -1);
956 break;
957 case NETDEV_DOWN:
958 fib_disable_ip(dev, 0, 0);
959 break;
960 case NETDEV_CHANGEMTU:
961 case NETDEV_CHANGE:
962 case NETDEV_UNREGISTER_PERNET:
963 rt_cache_flush(dev_net(dev), 0);
964 break;
965 }
966 return NOTIFY_DONE;
967}
968
969static struct notifier_block fib_inetaddr_notifier = {
970 .notifier_call = fib_inetaddr_event,
971};
972
973static struct notifier_block fib_netdev_notifier = {
974 .notifier_call = fib_netdev_event,
975};
976
977static int __net_init ip_fib_net_init(struct net *net)
978{
979 int err;
980 unsigned int i;
981
982 net->ipv4.fib_table_hash = kzalloc(
983 sizeof(struct hlist_head)*FIB_TABLE_HASHSZ, GFP_KERNEL);
984 if (net->ipv4.fib_table_hash == NULL)
985 return -ENOMEM;
986
987 for (i = 0; i < FIB_TABLE_HASHSZ; i++)
988 INIT_HLIST_HEAD(&net->ipv4.fib_table_hash[i]);
989
990 err = fib4_rules_init(net);
991 if (err < 0)
992 goto fail;
993 return 0;
994
995fail:
996 kfree(net->ipv4.fib_table_hash);
997 return err;
998}
999
1000static void __net_exit ip_fib_net_exit(struct net *net)
1001{
1002 unsigned int i;
1003
1004#ifdef CONFIG_IP_MULTIPLE_TABLES
1005 fib4_rules_exit(net);
1006#endif
1007
1008 for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
1009 struct fib_table *tb;
1010 struct hlist_head *head;
1011 struct hlist_node *node, *tmp;
1012
1013 head = &net->ipv4.fib_table_hash[i];
1014 hlist_for_each_entry_safe(tb, node, tmp, head, tb_hlist) {
1015 hlist_del(node);
1016 fib_table_flush(tb);
1017 kfree(tb);
1018 }
1019 }
1020 kfree(net->ipv4.fib_table_hash);
1021}
1022
1023static int __net_init fib_net_init(struct net *net)
1024{
1025 int error;
1026
1027 error = ip_fib_net_init(net);
1028 if (error < 0)
1029 goto out;
1030 error = nl_fib_lookup_init(net);
1031 if (error < 0)
1032 goto out_nlfl;
1033 error = fib_proc_init(net);
1034 if (error < 0)
1035 goto out_proc;
1036out:
1037 return error;
1038
1039out_proc:
1040 nl_fib_lookup_exit(net);
1041out_nlfl:
1042 ip_fib_net_exit(net);
1043 goto out;
1044}
1045
1046static void __net_exit fib_net_exit(struct net *net)
1047{
1048 fib_proc_exit(net);
1049 nl_fib_lookup_exit(net);
1050 ip_fib_net_exit(net);
1051}
1052
1053static struct pernet_operations fib_net_ops = {
1054 .init = fib_net_init,
1055 .exit = fib_net_exit,
1056};
1057
1058void __init ip_fib_init(void)
1059{
1060 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL);
1061 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL);
1062 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib);
1063
1064 register_pernet_subsys(&fib_net_ops);
1065 register_netdevice_notifier(&fib_netdev_notifier);
1066 register_inetaddr_notifier(&fib_inetaddr_notifier);
1067
1068 fib_hash_init();
1069}
1070
1071EXPORT_SYMBOL(inet_addr_type);
1072EXPORT_SYMBOL(inet_dev_addr_type);
1073EXPORT_SYMBOL(ip_dev_find);
Clone of davem's net-next-2.6 tree