projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
[deliverable/linux.git] / net / ipv4 / fib_frontend.c
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_trie_table(RT_TABLE_LOCAL);
55 if (local_table == NULL)
56 return -ENOMEM;
57
58 main_table = fib_trie_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_trie_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 static void fib_flush(struct net *net)
118 {
119 int flushed = 0;
120 struct fib_table *tb;
121 struct hlist_node *node;
122 struct hlist_head *head;
123 unsigned int h;
124
125 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
126 head = &net->ipv4.fib_table_hash[h];
127 hlist_for_each_entry(tb, node, head, tb_hlist)
128 flushed += fib_table_flush(tb);
129 }
130
131 if (flushed)
132 rt_cache_flush(net, -1);
133 }
134
135 /*
136 * Find address type as if only "dev" was present in the system. If
137 * on_dev is NULL then all interfaces are taken into consideration.
138 */
139 static inline unsigned __inet_dev_addr_type(struct net *net,
140 const struct net_device *dev,
141 __be32 addr)
142 {
143 struct flowi fl = { .fl4_dst = addr };
144 struct fib_result res;
145 unsigned ret = RTN_BROADCAST;
146 struct fib_table *local_table;
147
148 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
149 return RTN_BROADCAST;
150 if (ipv4_is_multicast(addr))
151 return RTN_MULTICAST;
152
153 #ifdef CONFIG_IP_MULTIPLE_TABLES
154 res.r = NULL;
155 #endif
156
157 local_table = fib_get_table(net, RT_TABLE_LOCAL);
158 if (local_table) {
159 ret = RTN_UNICAST;
160 rcu_read_lock();
161 if (!fib_table_lookup(local_table, &fl, &res, FIB_LOOKUP_NOREF)) {
162 if (!dev || dev == res.fi->fib_dev)
163 ret = res.type;
164 }
165 rcu_read_unlock();
166 }
167 return ret;
168 }
169
170 unsigned int inet_addr_type(struct net *net, __be32 addr)
171 {
172 return __inet_dev_addr_type(net, NULL, addr);
173 }
174 EXPORT_SYMBOL(inet_addr_type);
175
176 unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
177 __be32 addr)
178 {
179 return __inet_dev_addr_type(net, dev, addr);
180 }
181 EXPORT_SYMBOL(inet_dev_addr_type);
182
183 /* Given (packet source, input interface) and optional (dst, oif, tos):
184 * - (main) check, that source is valid i.e. not broadcast or our local
185 * address.
186 * - figure out what "logical" interface this packet arrived
187 * and calculate "specific destination" address.
188 * - check, that packet arrived from expected physical interface.
189 * called with rcu_read_lock()
190 */
191 int fib_validate_source(__be32 src, __be32 dst, u8 tos, int oif,
192 struct net_device *dev, __be32 *spec_dst,
193 u32 *itag, u32 mark)
194 {
195 struct in_device *in_dev;
196 struct flowi fl = {
197 .fl4_dst = src,
198 .fl4_src = dst,
199 .fl4_tos = tos,
200 .mark = mark,
201 .iif = oif
202 };
203 struct fib_result res;
204 int no_addr, rpf, accept_local;
205 bool dev_match;
206 int ret;
207 struct net *net;
208
209 no_addr = rpf = accept_local = 0;
210 in_dev = __in_dev_get_rcu(dev);
211 if (in_dev) {
212 no_addr = in_dev->ifa_list == NULL;
213 rpf = IN_DEV_RPFILTER(in_dev);
214 accept_local = IN_DEV_ACCEPT_LOCAL(in_dev);
215 if (mark && !IN_DEV_SRC_VMARK(in_dev))
216 fl.mark = 0;
217 }
218
219 if (in_dev == NULL)
220 goto e_inval;
221
222 net = dev_net(dev);
223 if (fib_lookup(net, &fl, &res))
224 goto last_resort;
225 if (res.type != RTN_UNICAST) {
226 if (res.type != RTN_LOCAL || !accept_local)
227 goto e_inval;
228 }
229 *spec_dst = FIB_RES_PREFSRC(res);
230 fib_combine_itag(itag, &res);
231 dev_match = false;
232
233 #ifdef CONFIG_IP_ROUTE_MULTIPATH
234 for (ret = 0; ret < res.fi->fib_nhs; ret++) {
235 struct fib_nh *nh = &res.fi->fib_nh[ret];
236
237 if (nh->nh_dev == dev) {
238 dev_match = true;
239 break;
240 }
241 }
242 #else
243 if (FIB_RES_DEV(res) == dev)
244 dev_match = true;
245 #endif
246 if (dev_match) {
247 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
248 return ret;
249 }
250 if (no_addr)
251 goto last_resort;
252 if (rpf == 1)
253 goto e_rpf;
254 fl.oif = dev->ifindex;
255
256 ret = 0;
257 if (fib_lookup(net, &fl, &res) == 0) {
258 if (res.type == RTN_UNICAST) {
259 *spec_dst = FIB_RES_PREFSRC(res);
260 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
261 }
262 }
263 return ret;
264
265 last_resort:
266 if (rpf)
267 goto e_rpf;
268 *spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE);
269 *itag = 0;
270 return 0;
271
272 e_inval:
273 return -EINVAL;
274 e_rpf:
275 return -EXDEV;
276 }
277
278 static inline __be32 sk_extract_addr(struct sockaddr *addr)
279 {
280 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
281 }
282
283 static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
284 {
285 struct nlattr *nla;
286
287 nla = (struct nlattr *) ((char *) mx + len);
288 nla->nla_type = type;
289 nla->nla_len = nla_attr_size(4);
290 *(u32 *) nla_data(nla) = value;
291
292 return len + nla_total_size(4);
293 }
294
295 static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
296 struct fib_config *cfg)
297 {
298 __be32 addr;
299 int plen;
300
301 memset(cfg, 0, sizeof(*cfg));
302 cfg->fc_nlinfo.nl_net = net;
303
304 if (rt->rt_dst.sa_family != AF_INET)
305 return -EAFNOSUPPORT;
306
307 /*
308 * Check mask for validity:
309 * a) it must be contiguous.
310 * b) destination must have all host bits clear.
311 * c) if application forgot to set correct family (AF_INET),
312 * reject request unless it is absolutely clear i.e.
313 * both family and mask are zero.
314 */
315 plen = 32;
316 addr = sk_extract_addr(&rt->rt_dst);
317 if (!(rt->rt_flags & RTF_HOST)) {
318 __be32 mask = sk_extract_addr(&rt->rt_genmask);
319
320 if (rt->rt_genmask.sa_family != AF_INET) {
321 if (mask || rt->rt_genmask.sa_family)
322 return -EAFNOSUPPORT;
323 }
324
325 if (bad_mask(mask, addr))
326 return -EINVAL;
327
328 plen = inet_mask_len(mask);
329 }
330
331 cfg->fc_dst_len = plen;
332 cfg->fc_dst = addr;
333
334 if (cmd != SIOCDELRT) {
335 cfg->fc_nlflags = NLM_F_CREATE;
336 cfg->fc_protocol = RTPROT_BOOT;
337 }
338
339 if (rt->rt_metric)
340 cfg->fc_priority = rt->rt_metric - 1;
341
342 if (rt->rt_flags & RTF_REJECT) {
343 cfg->fc_scope = RT_SCOPE_HOST;
344 cfg->fc_type = RTN_UNREACHABLE;
345 return 0;
346 }
347
348 cfg->fc_scope = RT_SCOPE_NOWHERE;
349 cfg->fc_type = RTN_UNICAST;
350
351 if (rt->rt_dev) {
352 char *colon;
353 struct net_device *dev;
354 char devname[IFNAMSIZ];
355
356 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
357 return -EFAULT;
358
359 devname[IFNAMSIZ-1] = 0;
360 colon = strchr(devname, ':');
361 if (colon)
362 *colon = 0;
363 dev = __dev_get_by_name(net, devname);
364 if (!dev)
365 return -ENODEV;
366 cfg->fc_oif = dev->ifindex;
367 if (colon) {
368 struct in_ifaddr *ifa;
369 struct in_device *in_dev = __in_dev_get_rtnl(dev);
370 if (!in_dev)
371 return -ENODEV;
372 *colon = ':';
373 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
374 if (strcmp(ifa->ifa_label, devname) == 0)
375 break;
376 if (ifa == NULL)
377 return -ENODEV;
378 cfg->fc_prefsrc = ifa->ifa_local;
379 }
380 }
381
382 addr = sk_extract_addr(&rt->rt_gateway);
383 if (rt->rt_gateway.sa_family == AF_INET && addr) {
384 cfg->fc_gw = addr;
385 if (rt->rt_flags & RTF_GATEWAY &&
386 inet_addr_type(net, addr) == RTN_UNICAST)
387 cfg->fc_scope = RT_SCOPE_UNIVERSE;
388 }
389
390 if (cmd == SIOCDELRT)
391 return 0;
392
393 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
394 return -EINVAL;
395
396 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
397 cfg->fc_scope = RT_SCOPE_LINK;
398
399 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
400 struct nlattr *mx;
401 int len = 0;
402
403 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
404 if (mx == NULL)
405 return -ENOMEM;
406
407 if (rt->rt_flags & RTF_MTU)
408 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
409
410 if (rt->rt_flags & RTF_WINDOW)
411 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
412
413 if (rt->rt_flags & RTF_IRTT)
414 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
415
416 cfg->fc_mx = mx;
417 cfg->fc_mx_len = len;
418 }
419
420 return 0;
421 }
422
423 /*
424 * Handle IP routing ioctl calls.
425 * These are used to manipulate the routing tables
426 */
427 int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
428 {
429 struct fib_config cfg;
430 struct rtentry rt;
431 int err;
432
433 switch (cmd) {
434 case SIOCADDRT: /* Add a route */
435 case SIOCDELRT: /* Delete a route */
436 if (!capable(CAP_NET_ADMIN))
437 return -EPERM;
438
439 if (copy_from_user(&rt, arg, sizeof(rt)))
440 return -EFAULT;
441
442 rtnl_lock();
443 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
444 if (err == 0) {
445 struct fib_table *tb;
446
447 if (cmd == SIOCDELRT) {
448 tb = fib_get_table(net, cfg.fc_table);
449 if (tb)
450 err = fib_table_delete(tb, &cfg);
451 else
452 err = -ESRCH;
453 } else {
454 tb = fib_new_table(net, cfg.fc_table);
455 if (tb)
456 err = fib_table_insert(tb, &cfg);
457 else
458 err = -ENOBUFS;
459 }
460
461 /* allocated by rtentry_to_fib_config() */
462 kfree(cfg.fc_mx);
463 }
464 rtnl_unlock();
465 return err;
466 }
467 return -EINVAL;
468 }
469
470 const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
471 [RTA_DST] = { .type = NLA_U32 },
472 [RTA_SRC] = { .type = NLA_U32 },
473 [RTA_IIF] = { .type = NLA_U32 },
474 [RTA_OIF] = { .type = NLA_U32 },
475 [RTA_GATEWAY] = { .type = NLA_U32 },
476 [RTA_PRIORITY] = { .type = NLA_U32 },
477 [RTA_PREFSRC] = { .type = NLA_U32 },
478 [RTA_METRICS] = { .type = NLA_NESTED },
479 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
480 [RTA_FLOW] = { .type = NLA_U32 },
481 };
482
483 static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
484 struct nlmsghdr *nlh, struct fib_config *cfg)
485 {
486 struct nlattr *attr;
487 int err, remaining;
488 struct rtmsg *rtm;
489
490 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
491 if (err < 0)
492 goto errout;
493
494 memset(cfg, 0, sizeof(*cfg));
495
496 rtm = nlmsg_data(nlh);
497 cfg->fc_dst_len = rtm->rtm_dst_len;
498 cfg->fc_tos = rtm->rtm_tos;
499 cfg->fc_table = rtm->rtm_table;
500 cfg->fc_protocol = rtm->rtm_protocol;
501 cfg->fc_scope = rtm->rtm_scope;
502 cfg->fc_type = rtm->rtm_type;
503 cfg->fc_flags = rtm->rtm_flags;
504 cfg->fc_nlflags = nlh->nlmsg_flags;
505
506 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
507 cfg->fc_nlinfo.nlh = nlh;
508 cfg->fc_nlinfo.nl_net = net;
509
510 if (cfg->fc_type > RTN_MAX) {
511 err = -EINVAL;
512 goto errout;
513 }
514
515 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
516 switch (nla_type(attr)) {
517 case RTA_DST:
518 cfg->fc_dst = nla_get_be32(attr);
519 break;
520 case RTA_OIF:
521 cfg->fc_oif = nla_get_u32(attr);
522 break;
523 case RTA_GATEWAY:
524 cfg->fc_gw = nla_get_be32(attr);
525 break;
526 case RTA_PRIORITY:
527 cfg->fc_priority = nla_get_u32(attr);
528 break;
529 case RTA_PREFSRC:
530 cfg->fc_prefsrc = nla_get_be32(attr);
531 break;
532 case RTA_METRICS:
533 cfg->fc_mx = nla_data(attr);
534 cfg->fc_mx_len = nla_len(attr);
535 break;
536 case RTA_MULTIPATH:
537 cfg->fc_mp = nla_data(attr);
538 cfg->fc_mp_len = nla_len(attr);
539 break;
540 case RTA_FLOW:
541 cfg->fc_flow = nla_get_u32(attr);
542 break;
543 case RTA_TABLE:
544 cfg->fc_table = nla_get_u32(attr);
545 break;
546 }
547 }
548
549 return 0;
550 errout:
551 return err;
552 }
553
554 static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
555 {
556 struct net *net = sock_net(skb->sk);
557 struct fib_config cfg;
558 struct fib_table *tb;
559 int err;
560
561 err = rtm_to_fib_config(net, skb, nlh, &cfg);
562 if (err < 0)
563 goto errout;
564
565 tb = fib_get_table(net, cfg.fc_table);
566 if (tb == NULL) {
567 err = -ESRCH;
568 goto errout;
569 }
570
571 err = fib_table_delete(tb, &cfg);
572 errout:
573 return err;
574 }
575
576 static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
577 {
578 struct net *net = sock_net(skb->sk);
579 struct fib_config cfg;
580 struct fib_table *tb;
581 int err;
582
583 err = rtm_to_fib_config(net, skb, nlh, &cfg);
584 if (err < 0)
585 goto errout;
586
587 tb = fib_new_table(net, cfg.fc_table);
588 if (tb == NULL) {
589 err = -ENOBUFS;
590 goto errout;
591 }
592
593 err = fib_table_insert(tb, &cfg);
594 errout:
595 return err;
596 }
597
598 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
599 {
600 struct net *net = sock_net(skb->sk);
601 unsigned int h, s_h;
602 unsigned int e = 0, s_e;
603 struct fib_table *tb;
604 struct hlist_node *node;
605 struct hlist_head *head;
606 int dumped = 0;
607
608 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
609 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
610 return ip_rt_dump(skb, cb);
611
612 s_h = cb->args[0];
613 s_e = cb->args[1];
614
615 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
616 e = 0;
617 head = &net->ipv4.fib_table_hash[h];
618 hlist_for_each_entry(tb, node, head, tb_hlist) {
619 if (e < s_e)
620 goto next;
621 if (dumped)
622 memset(&cb->args[2], 0, sizeof(cb->args) -
623 2 * sizeof(cb->args[0]));
624 if (fib_table_dump(tb, skb, cb) < 0)
625 goto out;
626 dumped = 1;
627 next:
628 e++;
629 }
630 }
631 out:
632 cb->args[1] = e;
633 cb->args[0] = h;
634
635 return skb->len;
636 }
637
638 /* Prepare and feed intra-kernel routing request.
639 * Really, it should be netlink message, but :-( netlink
640 * can be not configured, so that we feed it directly
641 * to fib engine. It is legal, because all events occur
642 * only when netlink is already locked.
643 */
644 static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
645 {
646 struct net *net = dev_net(ifa->ifa_dev->dev);
647 struct fib_table *tb;
648 struct fib_config cfg = {
649 .fc_protocol = RTPROT_KERNEL,
650 .fc_type = type,
651 .fc_dst = dst,
652 .fc_dst_len = dst_len,
653 .fc_prefsrc = ifa->ifa_local,
654 .fc_oif = ifa->ifa_dev->dev->ifindex,
655 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
656 .fc_nlinfo = {
657 .nl_net = net,
658 },
659 };
660
661 if (type == RTN_UNICAST)
662 tb = fib_new_table(net, RT_TABLE_MAIN);
663 else
664 tb = fib_new_table(net, RT_TABLE_LOCAL);
665
666 if (tb == NULL)
667 return;
668
669 cfg.fc_table = tb->tb_id;
670
671 if (type != RTN_LOCAL)
672 cfg.fc_scope = RT_SCOPE_LINK;
673 else
674 cfg.fc_scope = RT_SCOPE_HOST;
675
676 if (cmd == RTM_NEWROUTE)
677 fib_table_insert(tb, &cfg);
678 else
679 fib_table_delete(tb, &cfg);
680 }
681
682 void fib_add_ifaddr(struct in_ifaddr *ifa)
683 {
684 struct in_device *in_dev = ifa->ifa_dev;
685 struct net_device *dev = in_dev->dev;
686 struct in_ifaddr *prim = ifa;
687 __be32 mask = ifa->ifa_mask;
688 __be32 addr = ifa->ifa_local;
689 __be32 prefix = ifa->ifa_address & mask;
690
691 if (ifa->ifa_flags & IFA_F_SECONDARY) {
692 prim = inet_ifa_byprefix(in_dev, prefix, mask);
693 if (prim == NULL) {
694 printk(KERN_WARNING "fib_add_ifaddr: bug: prim == NULL\n");
695 return;
696 }
697 }
698
699 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
700
701 if (!(dev->flags & IFF_UP))
702 return;
703
704 /* Add broadcast address, if it is explicitly assigned. */
705 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
706 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
707
708 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
709 (prefix != addr || ifa->ifa_prefixlen < 32)) {
710 fib_magic(RTM_NEWROUTE,
711 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
712 prefix, ifa->ifa_prefixlen, prim);
713
714 /* Add network specific broadcasts, when it takes a sense */
715 if (ifa->ifa_prefixlen < 31) {
716 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
717 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
718 32, prim);
719 }
720 }
721 }
722
723 static void fib_del_ifaddr(struct in_ifaddr *ifa)
724 {
725 struct in_device *in_dev = ifa->ifa_dev;
726 struct net_device *dev = in_dev->dev;
727 struct in_ifaddr *ifa1;
728 struct in_ifaddr *prim = ifa;
729 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
730 __be32 any = ifa->ifa_address & ifa->ifa_mask;
731 #define LOCAL_OK 1
732 #define BRD_OK 2
733 #define BRD0_OK 4
734 #define BRD1_OK 8
735 unsigned ok = 0;
736
737 if (!(ifa->ifa_flags & IFA_F_SECONDARY))
738 fib_magic(RTM_DELROUTE,
739 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
740 any, ifa->ifa_prefixlen, prim);
741 else {
742 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
743 if (prim == NULL) {
744 printk(KERN_WARNING "fib_del_ifaddr: bug: prim == NULL\n");
745 return;
746 }
747 }
748
749 /* Deletion is more complicated than add.
750 * We should take care of not to delete too much :-)
751 *
752 * Scan address list to be sure that addresses are really gone.
753 */
754
755 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
756 if (ifa->ifa_local == ifa1->ifa_local)
757 ok |= LOCAL_OK;
758 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
759 ok |= BRD_OK;
760 if (brd == ifa1->ifa_broadcast)
761 ok |= BRD1_OK;
762 if (any == ifa1->ifa_broadcast)
763 ok |= BRD0_OK;
764 }
765
766 if (!(ok & BRD_OK))
767 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
768 if (!(ok & BRD1_OK))
769 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
770 if (!(ok & BRD0_OK))
771 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
772 if (!(ok & LOCAL_OK)) {
773 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
774
775 /* Check, that this local address finally disappeared. */
776 if (inet_addr_type(dev_net(dev), ifa->ifa_local) != RTN_LOCAL) {
777 /* And the last, but not the least thing.
778 * We must flush stray FIB entries.
779 *
780 * First of all, we scan fib_info list searching
781 * for stray nexthop entries, then ignite fib_flush.
782 */
783 if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local))
784 fib_flush(dev_net(dev));
785 }
786 }
787 #undef LOCAL_OK
788 #undef BRD_OK
789 #undef BRD0_OK
790 #undef BRD1_OK
791 }
792
793 static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb)
794 {
795
796 struct fib_result res;
797 struct flowi fl = {
798 .mark = frn->fl_mark,
799 .fl4_dst = frn->fl_addr,
800 .fl4_tos = frn->fl_tos,
801 .fl4_scope = frn->fl_scope,
802 };
803
804 #ifdef CONFIG_IP_MULTIPLE_TABLES
805 res.r = NULL;
806 #endif
807
808 frn->err = -ENOENT;
809 if (tb) {
810 local_bh_disable();
811
812 frn->tb_id = tb->tb_id;
813 rcu_read_lock();
814 frn->err = fib_table_lookup(tb, &fl, &res, FIB_LOOKUP_NOREF);
815
816 if (!frn->err) {
817 frn->prefixlen = res.prefixlen;
818 frn->nh_sel = res.nh_sel;
819 frn->type = res.type;
820 frn->scope = res.scope;
821 }
822 rcu_read_unlock();
823 local_bh_enable();
824 }
825 }
826
827 static void nl_fib_input(struct sk_buff *skb)
828 {
829 struct net *net;
830 struct fib_result_nl *frn;
831 struct nlmsghdr *nlh;
832 struct fib_table *tb;
833 u32 pid;
834
835 net = sock_net(skb->sk);
836 nlh = nlmsg_hdr(skb);
837 if (skb->len < NLMSG_SPACE(0) || skb->len < nlh->nlmsg_len ||
838 nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*frn)))
839 return;
840
841 skb = skb_clone(skb, GFP_KERNEL);
842 if (skb == NULL)
843 return;
844 nlh = nlmsg_hdr(skb);
845
846 frn = (struct fib_result_nl *) NLMSG_DATA(nlh);
847 tb = fib_get_table(net, frn->tb_id_in);
848
849 nl_fib_lookup(frn, tb);
850
851 pid = NETLINK_CB(skb).pid; /* pid of sending process */
852 NETLINK_CB(skb).pid = 0; /* from kernel */
853 NETLINK_CB(skb).dst_group = 0; /* unicast */
854 netlink_unicast(net->ipv4.fibnl, skb, pid, MSG_DONTWAIT);
855 }
856
857 static int __net_init nl_fib_lookup_init(struct net *net)
858 {
859 struct sock *sk;
860 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, 0,
861 nl_fib_input, NULL, THIS_MODULE);
862 if (sk == NULL)
863 return -EAFNOSUPPORT;
864 net->ipv4.fibnl = sk;
865 return 0;
866 }
867
868 static void nl_fib_lookup_exit(struct net *net)
869 {
870 netlink_kernel_release(net->ipv4.fibnl);
871 net->ipv4.fibnl = NULL;
872 }
873
874 static void fib_disable_ip(struct net_device *dev, int force, int delay)
875 {
876 if (fib_sync_down_dev(dev, force))
877 fib_flush(dev_net(dev));
878 rt_cache_flush(dev_net(dev), delay);
879 arp_ifdown(dev);
880 }
881
882 static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
883 {
884 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
885 struct net_device *dev = ifa->ifa_dev->dev;
886
887 switch (event) {
888 case NETDEV_UP:
889 fib_add_ifaddr(ifa);
890 #ifdef CONFIG_IP_ROUTE_MULTIPATH
891 fib_sync_up(dev);
892 #endif
893 rt_cache_flush(dev_net(dev), -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(dev, 1, 0);
902 } else {
903 rt_cache_flush(dev_net(dev), -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 (event == NETDEV_UNREGISTER) {
916 fib_disable_ip(dev, 2, -1);
917 return NOTIFY_DONE;
918 }
919
920 if (!in_dev)
921 return NOTIFY_DONE;
922
923 switch (event) {
924 case NETDEV_UP:
925 for_ifa(in_dev) {
926 fib_add_ifaddr(ifa);
927 } endfor_ifa(in_dev);
928 #ifdef CONFIG_IP_ROUTE_MULTIPATH
929 fib_sync_up(dev);
930 #endif
931 rt_cache_flush(dev_net(dev), -1);
932 break;
933 case NETDEV_DOWN:
934 fib_disable_ip(dev, 0, 0);
935 break;
936 case NETDEV_CHANGEMTU:
937 case NETDEV_CHANGE:
938 rt_cache_flush(dev_net(dev), 0);
939 break;
940 case NETDEV_UNREGISTER_BATCH:
941 /* The batch unregister is only called on the first
942 * device in the list of devices being unregistered.
943 * Therefore we should not pass dev_net(dev) in here.
944 */
945 rt_cache_flush_batch(NULL);
946 break;
947 }
948 return NOTIFY_DONE;
949 }
950
951 static struct notifier_block fib_inetaddr_notifier = {
952 .notifier_call = fib_inetaddr_event,
953 };
954
955 static struct notifier_block fib_netdev_notifier = {
956 .notifier_call = fib_netdev_event,
957 };
958
959 static int __net_init ip_fib_net_init(struct net *net)
960 {
961 int err;
962 size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
963
964 /* Avoid false sharing : Use at least a full cache line */
965 size = max_t(size_t, size, L1_CACHE_BYTES);
966
967 net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
968 if (net->ipv4.fib_table_hash == NULL)
969 return -ENOMEM;
970
971 err = fib4_rules_init(net);
972 if (err < 0)
973 goto fail;
974 return 0;
975
976 fail:
977 kfree(net->ipv4.fib_table_hash);
978 return err;
979 }
980
981 static void ip_fib_net_exit(struct net *net)
982 {
983 unsigned int i;
984
985 #ifdef CONFIG_IP_MULTIPLE_TABLES
986 fib4_rules_exit(net);
987 #endif
988
989 for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
990 struct fib_table *tb;
991 struct hlist_head *head;
992 struct hlist_node *node, *tmp;
993
994 head = &net->ipv4.fib_table_hash[i];
995 hlist_for_each_entry_safe(tb, node, tmp, head, tb_hlist) {
996 hlist_del(node);
997 fib_table_flush(tb);
998 fib_free_table(tb);
999 }
1000 }
1001 kfree(net->ipv4.fib_table_hash);
1002 }
1003
1004 static int __net_init fib_net_init(struct net *net)
1005 {
1006 int error;
1007
1008 error = ip_fib_net_init(net);
1009 if (error < 0)
1010 goto out;
1011 error = nl_fib_lookup_init(net);
1012 if (error < 0)
1013 goto out_nlfl;
1014 error = fib_proc_init(net);
1015 if (error < 0)
1016 goto out_proc;
1017 out:
1018 return error;
1019
1020 out_proc:
1021 nl_fib_lookup_exit(net);
1022 out_nlfl:
1023 ip_fib_net_exit(net);
1024 goto out;
1025 }
1026
1027 static void __net_exit fib_net_exit(struct net *net)
1028 {
1029 fib_proc_exit(net);
1030 nl_fib_lookup_exit(net);
1031 ip_fib_net_exit(net);
1032 }
1033
1034 static struct pernet_operations fib_net_ops = {
1035 .init = fib_net_init,
1036 .exit = fib_net_exit,
1037 };
1038
1039 void __init ip_fib_init(void)
1040 {
1041 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL);
1042 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL);
1043 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib);
1044
1045 register_pernet_subsys(&fib_net_ops);
1046 register_netdevice_notifier(&fib_netdev_notifier);
1047 register_inetaddr_notifier(&fib_inetaddr_notifier);
1048
1049 fib_trie_init();
1050 }
Linux kernel - Deliverables
This page took 0.053707 seconds and 6 git commands to generate.