]> git.proxmox.com Git - mirror_ubuntu-focal-kernel.git/blame_incremental - net/ipv4/fib_frontend.c
projects / mirror_ubuntu-focal-kernel.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
bonding: add retransmit membership reports tunable
[mirror_ubuntu-focal-kernel.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#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
50static int __net_init fib4_rules_init(struct net *net)
51{
52 struct fib_table *local_table, *main_table;
53
54 local_table = fib_hash_table(RT_TABLE_LOCAL);
55 if (local_table == NULL)
56 return -ENOMEM;
57
58 main_table = fib_hash_table(RT_TABLE_MAIN);
59 if (main_table == NULL)
60 goto fail;
61
62 hlist_add_head_rcu(&local_table->tb_hlist,
63 &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]);
64 hlist_add_head_rcu(&main_table->tb_hlist,
65 &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]);
66 return 0;
67
68fail:
69 kfree(local_table);
70 return -ENOMEM;
71}
72#else
73
74struct fib_table *fib_new_table(struct net *net, u32 id)
75{
76 struct fib_table *tb;
77 unsigned int h;
78
79 if (id == 0)
80 id = RT_TABLE_MAIN;
81 tb = fib_get_table(net, id);
82 if (tb)
83 return tb;
84
85 tb = fib_hash_table(id);
86 if (!tb)
87 return NULL;
88 h = id & (FIB_TABLE_HASHSZ - 1);
89 hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]);
90 return tb;
91}
92
93struct 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
117void fib_select_default(struct net *net,
118 const struct flowi *flp, struct fib_result *res)
119{
120 struct fib_table *tb;
121 int table = RT_TABLE_MAIN;
122#ifdef CONFIG_IP_MULTIPLE_TABLES
123 if (res->r == NULL || res->r->action != FR_ACT_TO_TBL)
124 return;
125 table = res->r->table;
126#endif
127 tb = fib_get_table(net, table);
128 if (FIB_RES_GW(*res) && FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK)
129 fib_table_select_default(tb, flp, res);
130}
131
132static void fib_flush(struct net *net)
133{
134 int flushed = 0;
135 struct fib_table *tb;
136 struct hlist_node *node;
137 struct hlist_head *head;
138 unsigned int h;
139
140 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
141 head = &net->ipv4.fib_table_hash[h];
142 hlist_for_each_entry(tb, node, head, tb_hlist)
143 flushed += fib_table_flush(tb);
144 }
145
146 if (flushed)
147 rt_cache_flush(net, -1);
148}
149
150/**
151 * __ip_dev_find - find the first device with a given source address.
152 * @net: the net namespace
153 * @addr: the source address
154 * @devref: if true, take a reference on the found device
155 *
156 * If a caller uses devref=false, it should be protected by RCU
157 */
158struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref)
159{
160 struct flowi fl = {
161 .nl_u = {
162 .ip4_u = {
163 .daddr = addr
164 }
165 },
166 .flags = FLOWI_FLAG_MATCH_ANY_IIF
167 };
168 struct fib_result res = { 0 };
169 struct net_device *dev = NULL;
170
171 if (fib_lookup(net, &fl, &res))
172 return NULL;
173 if (res.type != RTN_LOCAL)
174 goto out;
175 dev = FIB_RES_DEV(res);
176
177 if (dev && devref)
178 dev_hold(dev);
179out:
180 fib_res_put(&res);
181 return dev;
182}
183EXPORT_SYMBOL(__ip_dev_find);
184
185/*
186 * Find address type as if only "dev" was present in the system. If
187 * on_dev is NULL then all interfaces are taken into consideration.
188 */
189static inline unsigned __inet_dev_addr_type(struct net *net,
190 const struct net_device *dev,
191 __be32 addr)
192{
193 struct flowi fl = { .nl_u = { .ip4_u = { .daddr = addr } } };
194 struct fib_result res;
195 unsigned ret = RTN_BROADCAST;
196 struct fib_table *local_table;
197
198 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
199 return RTN_BROADCAST;
200 if (ipv4_is_multicast(addr))
201 return RTN_MULTICAST;
202
203#ifdef CONFIG_IP_MULTIPLE_TABLES
204 res.r = NULL;
205#endif
206
207 local_table = fib_get_table(net, RT_TABLE_LOCAL);
208 if (local_table) {
209 ret = RTN_UNICAST;
210 if (!fib_table_lookup(local_table, &fl, &res)) {
211 if (!dev || dev == res.fi->fib_dev)
212 ret = res.type;
213 fib_res_put(&res);
214 }
215 }
216 return ret;
217}
218
219unsigned int inet_addr_type(struct net *net, __be32 addr)
220{
221 return __inet_dev_addr_type(net, NULL, addr);
222}
223EXPORT_SYMBOL(inet_addr_type);
224
225unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
226 __be32 addr)
227{
228 return __inet_dev_addr_type(net, dev, addr);
229}
230EXPORT_SYMBOL(inet_dev_addr_type);
231
232/* Given (packet source, input interface) and optional (dst, oif, tos):
233 * - (main) check, that source is valid i.e. not broadcast or our local
234 * address.
235 * - figure out what "logical" interface this packet arrived
236 * and calculate "specific destination" address.
237 * - check, that packet arrived from expected physical interface.
238 */
239int fib_validate_source(__be32 src, __be32 dst, u8 tos, int oif,
240 struct net_device *dev, __be32 *spec_dst,
241 u32 *itag, u32 mark)
242{
243 struct in_device *in_dev;
244 struct flowi fl = {
245 .nl_u = {
246 .ip4_u = {
247 .daddr = src,
248 .saddr = dst,
249 .tos = tos
250 }
251 },
252 .mark = mark,
253 .iif = oif
254 };
255 struct fib_result res;
256 int no_addr, rpf, accept_local;
257 bool dev_match;
258 int ret;
259 struct net *net;
260
261 no_addr = rpf = accept_local = 0;
262 rcu_read_lock();
263 in_dev = __in_dev_get_rcu(dev);
264 if (in_dev) {
265 no_addr = in_dev->ifa_list == NULL;
266 rpf = IN_DEV_RPFILTER(in_dev);
267 accept_local = IN_DEV_ACCEPT_LOCAL(in_dev);
268 if (mark && !IN_DEV_SRC_VMARK(in_dev))
269 fl.mark = 0;
270 }
271 rcu_read_unlock();
272
273 if (in_dev == NULL)
274 goto e_inval;
275
276 net = dev_net(dev);
277 if (fib_lookup(net, &fl, &res))
278 goto last_resort;
279 if (res.type != RTN_UNICAST) {
280 if (res.type != RTN_LOCAL || !accept_local)
281 goto e_inval_res;
282 }
283 *spec_dst = FIB_RES_PREFSRC(res);
284 fib_combine_itag(itag, &res);
285 dev_match = false;
286
287#ifdef CONFIG_IP_ROUTE_MULTIPATH
288 for (ret = 0; ret < res.fi->fib_nhs; ret++) {
289 struct fib_nh *nh = &res.fi->fib_nh[ret];
290
291 if (nh->nh_dev == dev) {
292 dev_match = true;
293 break;
294 }
295 }
296#else
297 if (FIB_RES_DEV(res) == dev)
298 dev_match = true;
299#endif
300 if (dev_match) {
301 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
302 fib_res_put(&res);
303 return ret;
304 }
305 fib_res_put(&res);
306 if (no_addr)
307 goto last_resort;
308 if (rpf == 1)
309 goto e_rpf;
310 fl.oif = dev->ifindex;
311
312 ret = 0;
313 if (fib_lookup(net, &fl, &res) == 0) {
314 if (res.type == RTN_UNICAST) {
315 *spec_dst = FIB_RES_PREFSRC(res);
316 ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
317 }
318 fib_res_put(&res);
319 }
320 return ret;
321
322last_resort:
323 if (rpf)
324 goto e_rpf;
325 *spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE);
326 *itag = 0;
327 return 0;
328
329e_inval_res:
330 fib_res_put(&res);
331e_inval:
332 return -EINVAL;
333e_rpf:
334 return -EXDEV;
335}
336
337static inline __be32 sk_extract_addr(struct sockaddr *addr)
338{
339 return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
340}
341
342static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
343{
344 struct nlattr *nla;
345
346 nla = (struct nlattr *) ((char *) mx + len);
347 nla->nla_type = type;
348 nla->nla_len = nla_attr_size(4);
349 *(u32 *) nla_data(nla) = value;
350
351 return len + nla_total_size(4);
352}
353
354static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
355 struct fib_config *cfg)
356{
357 __be32 addr;
358 int plen;
359
360 memset(cfg, 0, sizeof(*cfg));
361 cfg->fc_nlinfo.nl_net = net;
362
363 if (rt->rt_dst.sa_family != AF_INET)
364 return -EAFNOSUPPORT;
365
366 /*
367 * Check mask for validity:
368 * a) it must be contiguous.
369 * b) destination must have all host bits clear.
370 * c) if application forgot to set correct family (AF_INET),
371 * reject request unless it is absolutely clear i.e.
372 * both family and mask are zero.
373 */
374 plen = 32;
375 addr = sk_extract_addr(&rt->rt_dst);
376 if (!(rt->rt_flags & RTF_HOST)) {
377 __be32 mask = sk_extract_addr(&rt->rt_genmask);
378
379 if (rt->rt_genmask.sa_family != AF_INET) {
380 if (mask || rt->rt_genmask.sa_family)
381 return -EAFNOSUPPORT;
382 }
383
384 if (bad_mask(mask, addr))
385 return -EINVAL;
386
387 plen = inet_mask_len(mask);
388 }
389
390 cfg->fc_dst_len = plen;
391 cfg->fc_dst = addr;
392
393 if (cmd != SIOCDELRT) {
394 cfg->fc_nlflags = NLM_F_CREATE;
395 cfg->fc_protocol = RTPROT_BOOT;
396 }
397
398 if (rt->rt_metric)
399 cfg->fc_priority = rt->rt_metric - 1;
400
401 if (rt->rt_flags & RTF_REJECT) {
402 cfg->fc_scope = RT_SCOPE_HOST;
403 cfg->fc_type = RTN_UNREACHABLE;
404 return 0;
405 }
406
407 cfg->fc_scope = RT_SCOPE_NOWHERE;
408 cfg->fc_type = RTN_UNICAST;
409
410 if (rt->rt_dev) {
411 char *colon;
412 struct net_device *dev;
413 char devname[IFNAMSIZ];
414
415 if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
416 return -EFAULT;
417
418 devname[IFNAMSIZ-1] = 0;
419 colon = strchr(devname, ':');
420 if (colon)
421 *colon = 0;
422 dev = __dev_get_by_name(net, devname);
423 if (!dev)
424 return -ENODEV;
425 cfg->fc_oif = dev->ifindex;
426 if (colon) {
427 struct in_ifaddr *ifa;
428 struct in_device *in_dev = __in_dev_get_rtnl(dev);
429 if (!in_dev)
430 return -ENODEV;
431 *colon = ':';
432 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
433 if (strcmp(ifa->ifa_label, devname) == 0)
434 break;
435 if (ifa == NULL)
436 return -ENODEV;
437 cfg->fc_prefsrc = ifa->ifa_local;
438 }
439 }
440
441 addr = sk_extract_addr(&rt->rt_gateway);
442 if (rt->rt_gateway.sa_family == AF_INET && addr) {
443 cfg->fc_gw = addr;
444 if (rt->rt_flags & RTF_GATEWAY &&
445 inet_addr_type(net, addr) == RTN_UNICAST)
446 cfg->fc_scope = RT_SCOPE_UNIVERSE;
447 }
448
449 if (cmd == SIOCDELRT)
450 return 0;
451
452 if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
453 return -EINVAL;
454
455 if (cfg->fc_scope == RT_SCOPE_NOWHERE)
456 cfg->fc_scope = RT_SCOPE_LINK;
457
458 if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
459 struct nlattr *mx;
460 int len = 0;
461
462 mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
463 if (mx == NULL)
464 return -ENOMEM;
465
466 if (rt->rt_flags & RTF_MTU)
467 len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
468
469 if (rt->rt_flags & RTF_WINDOW)
470 len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
471
472 if (rt->rt_flags & RTF_IRTT)
473 len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
474
475 cfg->fc_mx = mx;
476 cfg->fc_mx_len = len;
477 }
478
479 return 0;
480}
481
482/*
483 * Handle IP routing ioctl calls.
484 * These are used to manipulate the routing tables
485 */
486int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
487{
488 struct fib_config cfg;
489 struct rtentry rt;
490 int err;
491
492 switch (cmd) {
493 case SIOCADDRT: /* Add a route */
494 case SIOCDELRT: /* Delete a route */
495 if (!capable(CAP_NET_ADMIN))
496 return -EPERM;
497
498 if (copy_from_user(&rt, arg, sizeof(rt)))
499 return -EFAULT;
500
501 rtnl_lock();
502 err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
503 if (err == 0) {
504 struct fib_table *tb;
505
506 if (cmd == SIOCDELRT) {
507 tb = fib_get_table(net, cfg.fc_table);
508 if (tb)
509 err = fib_table_delete(tb, &cfg);
510 else
511 err = -ESRCH;
512 } else {
513 tb = fib_new_table(net, cfg.fc_table);
514 if (tb)
515 err = fib_table_insert(tb, &cfg);
516 else
517 err = -ENOBUFS;
518 }
519
520 /* allocated by rtentry_to_fib_config() */
521 kfree(cfg.fc_mx);
522 }
523 rtnl_unlock();
524 return err;
525 }
526 return -EINVAL;
527}
528
529const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
530 [RTA_DST] = { .type = NLA_U32 },
531 [RTA_SRC] = { .type = NLA_U32 },
532 [RTA_IIF] = { .type = NLA_U32 },
533 [RTA_OIF] = { .type = NLA_U32 },
534 [RTA_GATEWAY] = { .type = NLA_U32 },
535 [RTA_PRIORITY] = { .type = NLA_U32 },
536 [RTA_PREFSRC] = { .type = NLA_U32 },
537 [RTA_METRICS] = { .type = NLA_NESTED },
538 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
539 [RTA_FLOW] = { .type = NLA_U32 },
540};
541
542static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
543 struct nlmsghdr *nlh, struct fib_config *cfg)
544{
545 struct nlattr *attr;
546 int err, remaining;
547 struct rtmsg *rtm;
548
549 err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
550 if (err < 0)
551 goto errout;
552
553 memset(cfg, 0, sizeof(*cfg));
554
555 rtm = nlmsg_data(nlh);
556 cfg->fc_dst_len = rtm->rtm_dst_len;
557 cfg->fc_tos = rtm->rtm_tos;
558 cfg->fc_table = rtm->rtm_table;
559 cfg->fc_protocol = rtm->rtm_protocol;
560 cfg->fc_scope = rtm->rtm_scope;
561 cfg->fc_type = rtm->rtm_type;
562 cfg->fc_flags = rtm->rtm_flags;
563 cfg->fc_nlflags = nlh->nlmsg_flags;
564
565 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
566 cfg->fc_nlinfo.nlh = nlh;
567 cfg->fc_nlinfo.nl_net = net;
568
569 if (cfg->fc_type > RTN_MAX) {
570 err = -EINVAL;
571 goto errout;
572 }
573
574 nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
575 switch (nla_type(attr)) {
576 case RTA_DST:
577 cfg->fc_dst = nla_get_be32(attr);
578 break;
579 case RTA_OIF:
580 cfg->fc_oif = nla_get_u32(attr);
581 break;
582 case RTA_GATEWAY:
583 cfg->fc_gw = nla_get_be32(attr);
584 break;
585 case RTA_PRIORITY:
586 cfg->fc_priority = nla_get_u32(attr);
587 break;
588 case RTA_PREFSRC:
589 cfg->fc_prefsrc = nla_get_be32(attr);
590 break;
591 case RTA_METRICS:
592 cfg->fc_mx = nla_data(attr);
593 cfg->fc_mx_len = nla_len(attr);
594 break;
595 case RTA_MULTIPATH:
596 cfg->fc_mp = nla_data(attr);
597 cfg->fc_mp_len = nla_len(attr);
598 break;
599 case RTA_FLOW:
600 cfg->fc_flow = nla_get_u32(attr);
601 break;
602 case RTA_TABLE:
603 cfg->fc_table = nla_get_u32(attr);
604 break;
605 }
606 }
607
608 return 0;
609errout:
610 return err;
611}
612
613static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
614{
615 struct net *net = sock_net(skb->sk);
616 struct fib_config cfg;
617 struct fib_table *tb;
618 int err;
619
620 err = rtm_to_fib_config(net, skb, nlh, &cfg);
621 if (err < 0)
622 goto errout;
623
624 tb = fib_get_table(net, cfg.fc_table);
625 if (tb == NULL) {
626 err = -ESRCH;
627 goto errout;
628 }
629
630 err = fib_table_delete(tb, &cfg);
631errout:
632 return err;
633}
634
635static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
636{
637 struct net *net = sock_net(skb->sk);
638 struct fib_config cfg;
639 struct fib_table *tb;
640 int err;
641
642 err = rtm_to_fib_config(net, skb, nlh, &cfg);
643 if (err < 0)
644 goto errout;
645
646 tb = fib_new_table(net, cfg.fc_table);
647 if (tb == NULL) {
648 err = -ENOBUFS;
649 goto errout;
650 }
651
652 err = fib_table_insert(tb, &cfg);
653errout:
654 return err;
655}
656
657static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
658{
659 struct net *net = sock_net(skb->sk);
660 unsigned int h, s_h;
661 unsigned int e = 0, s_e;
662 struct fib_table *tb;
663 struct hlist_node *node;
664 struct hlist_head *head;
665 int dumped = 0;
666
667 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
668 ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
669 return ip_rt_dump(skb, cb);
670
671 s_h = cb->args[0];
672 s_e = cb->args[1];
673
674 for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
675 e = 0;
676 head = &net->ipv4.fib_table_hash[h];
677 hlist_for_each_entry(tb, node, head, tb_hlist) {
678 if (e < s_e)
679 goto next;
680 if (dumped)
681 memset(&cb->args[2], 0, sizeof(cb->args) -
682 2 * sizeof(cb->args[0]));
683 if (fib_table_dump(tb, skb, cb) < 0)
684 goto out;
685 dumped = 1;
686next:
687 e++;
688 }
689 }
690out:
691 cb->args[1] = e;
692 cb->args[0] = h;
693
694 return skb->len;
695}
696
697/* Prepare and feed intra-kernel routing request.
698 * Really, it should be netlink message, but :-( netlink
699 * can be not configured, so that we feed it directly
700 * to fib engine. It is legal, because all events occur
701 * only when netlink is already locked.
702 */
703static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
704{
705 struct net *net = dev_net(ifa->ifa_dev->dev);
706 struct fib_table *tb;
707 struct fib_config cfg = {
708 .fc_protocol = RTPROT_KERNEL,
709 .fc_type = type,
710 .fc_dst = dst,
711 .fc_dst_len = dst_len,
712 .fc_prefsrc = ifa->ifa_local,
713 .fc_oif = ifa->ifa_dev->dev->ifindex,
714 .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
715 .fc_nlinfo = {
716 .nl_net = net,
717 },
718 };
719
720 if (type == RTN_UNICAST)
721 tb = fib_new_table(net, RT_TABLE_MAIN);
722 else
723 tb = fib_new_table(net, RT_TABLE_LOCAL);
724
725 if (tb == NULL)
726 return;
727
728 cfg.fc_table = tb->tb_id;
729
730 if (type != RTN_LOCAL)
731 cfg.fc_scope = RT_SCOPE_LINK;
732 else
733 cfg.fc_scope = RT_SCOPE_HOST;
734
735 if (cmd == RTM_NEWROUTE)
736 fib_table_insert(tb, &cfg);
737 else
738 fib_table_delete(tb, &cfg);
739}
740
741void fib_add_ifaddr(struct in_ifaddr *ifa)
742{
743 struct in_device *in_dev = ifa->ifa_dev;
744 struct net_device *dev = in_dev->dev;
745 struct in_ifaddr *prim = ifa;
746 __be32 mask = ifa->ifa_mask;
747 __be32 addr = ifa->ifa_local;
748 __be32 prefix = ifa->ifa_address & mask;
749
750 if (ifa->ifa_flags & IFA_F_SECONDARY) {
751 prim = inet_ifa_byprefix(in_dev, prefix, mask);
752 if (prim == NULL) {
753 printk(KERN_WARNING "fib_add_ifaddr: bug: prim == NULL\n");
754 return;
755 }
756 }
757
758 fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
759
760 if (!(dev->flags & IFF_UP))
761 return;
762
763 /* Add broadcast address, if it is explicitly assigned. */
764 if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
765 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
766
767 if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
768 (prefix != addr || ifa->ifa_prefixlen < 32)) {
769 fib_magic(RTM_NEWROUTE,
770 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
771 prefix, ifa->ifa_prefixlen, prim);
772
773 /* Add network specific broadcasts, when it takes a sense */
774 if (ifa->ifa_prefixlen < 31) {
775 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
776 fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
777 32, prim);
778 }
779 }
780}
781
782static void fib_del_ifaddr(struct in_ifaddr *ifa)
783{
784 struct in_device *in_dev = ifa->ifa_dev;
785 struct net_device *dev = in_dev->dev;
786 struct in_ifaddr *ifa1;
787 struct in_ifaddr *prim = ifa;
788 __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
789 __be32 any = ifa->ifa_address & ifa->ifa_mask;
790#define LOCAL_OK 1
791#define BRD_OK 2
792#define BRD0_OK 4
793#define BRD1_OK 8
794 unsigned ok = 0;
795
796 if (!(ifa->ifa_flags & IFA_F_SECONDARY))
797 fib_magic(RTM_DELROUTE,
798 dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
799 any, ifa->ifa_prefixlen, prim);
800 else {
801 prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
802 if (prim == NULL) {
803 printk(KERN_WARNING "fib_del_ifaddr: bug: prim == NULL\n");
804 return;
805 }
806 }
807
808 /* Deletion is more complicated than add.
809 * We should take care of not to delete too much :-)
810 *
811 * Scan address list to be sure that addresses are really gone.
812 */
813
814 for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
815 if (ifa->ifa_local == ifa1->ifa_local)
816 ok |= LOCAL_OK;
817 if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
818 ok |= BRD_OK;
819 if (brd == ifa1->ifa_broadcast)
820 ok |= BRD1_OK;
821 if (any == ifa1->ifa_broadcast)
822 ok |= BRD0_OK;
823 }
824
825 if (!(ok & BRD_OK))
826 fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
827 if (!(ok & BRD1_OK))
828 fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
829 if (!(ok & BRD0_OK))
830 fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
831 if (!(ok & LOCAL_OK)) {
832 fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
833
834 /* Check, that this local address finally disappeared. */
835 if (inet_addr_type(dev_net(dev), ifa->ifa_local) != RTN_LOCAL) {
836 /* And the last, but not the least thing.
837 * We must flush stray FIB entries.
838 *
839 * First of all, we scan fib_info list searching
840 * for stray nexthop entries, then ignite fib_flush.
841 */
842 if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local))
843 fib_flush(dev_net(dev));
844 }
845 }
846#undef LOCAL_OK
847#undef BRD_OK
848#undef BRD0_OK
849#undef BRD1_OK
850}
851
852static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb)
853{
854
855 struct fib_result res;
856 struct flowi fl = {
857 .mark = frn->fl_mark,
858 .nl_u = {
859 .ip4_u = {
860 .daddr = frn->fl_addr,
861 .tos = frn->fl_tos,
862 .scope = frn->fl_scope
863 }
864 }
865 };
866
867#ifdef CONFIG_IP_MULTIPLE_TABLES
868 res.r = NULL;
869#endif
870
871 frn->err = -ENOENT;
872 if (tb) {
873 local_bh_disable();
874
875 frn->tb_id = tb->tb_id;
876 frn->err = fib_table_lookup(tb, &fl, &res);
877
878 if (!frn->err) {
879 frn->prefixlen = res.prefixlen;
880 frn->nh_sel = res.nh_sel;
881 frn->type = res.type;
882 frn->scope = res.scope;
883 fib_res_put(&res);
884 }
885 local_bh_enable();
886 }
887}
888
889static void nl_fib_input(struct sk_buff *skb)
890{
891 struct net *net;
892 struct fib_result_nl *frn;
893 struct nlmsghdr *nlh;
894 struct fib_table *tb;
895 u32 pid;
896
897 net = sock_net(skb->sk);
898 nlh = nlmsg_hdr(skb);
899 if (skb->len < NLMSG_SPACE(0) || skb->len < nlh->nlmsg_len ||
900 nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*frn)))
901 return;
902
903 skb = skb_clone(skb, GFP_KERNEL);
904 if (skb == NULL)
905 return;
906 nlh = nlmsg_hdr(skb);
907
908 frn = (struct fib_result_nl *) NLMSG_DATA(nlh);
909 tb = fib_get_table(net, frn->tb_id_in);
910
911 nl_fib_lookup(frn, tb);
912
913 pid = NETLINK_CB(skb).pid; /* pid of sending process */
914 NETLINK_CB(skb).pid = 0; /* from kernel */
915 NETLINK_CB(skb).dst_group = 0; /* unicast */
916 netlink_unicast(net->ipv4.fibnl, skb, pid, MSG_DONTWAIT);
917}
918
919static int __net_init nl_fib_lookup_init(struct net *net)
920{
921 struct sock *sk;
922 sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, 0,
923 nl_fib_input, NULL, THIS_MODULE);
924 if (sk == NULL)
925 return -EAFNOSUPPORT;
926 net->ipv4.fibnl = sk;
927 return 0;
928}
929
930static void nl_fib_lookup_exit(struct net *net)
931{
932 netlink_kernel_release(net->ipv4.fibnl);
933 net->ipv4.fibnl = NULL;
934}
935
936static void fib_disable_ip(struct net_device *dev, int force, int delay)
937{
938 if (fib_sync_down_dev(dev, force))
939 fib_flush(dev_net(dev));
940 rt_cache_flush(dev_net(dev), delay);
941 arp_ifdown(dev);
942}
943
944static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
945{
946 struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
947 struct net_device *dev = ifa->ifa_dev->dev;
948
949 switch (event) {
950 case NETDEV_UP:
951 fib_add_ifaddr(ifa);
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_del_ifaddr(ifa);
959 if (ifa->ifa_dev->ifa_list == NULL) {
960 /* Last address was deleted from this interface.
961 * Disable IP.
962 */
963 fib_disable_ip(dev, 1, 0);
964 } else {
965 rt_cache_flush(dev_net(dev), -1);
966 }
967 break;
968 }
969 return NOTIFY_DONE;
970}
971
972static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
973{
974 struct net_device *dev = ptr;
975 struct in_device *in_dev = __in_dev_get_rtnl(dev);
976
977 if (event == NETDEV_UNREGISTER) {
978 fib_disable_ip(dev, 2, -1);
979 return NOTIFY_DONE;
980 }
981
982 if (!in_dev)
983 return NOTIFY_DONE;
984
985 switch (event) {
986 case NETDEV_UP:
987 for_ifa(in_dev) {
988 fib_add_ifaddr(ifa);
989 } endfor_ifa(in_dev);
990#ifdef CONFIG_IP_ROUTE_MULTIPATH
991 fib_sync_up(dev);
992#endif
993 rt_cache_flush(dev_net(dev), -1);
994 break;
995 case NETDEV_DOWN:
996 fib_disable_ip(dev, 0, 0);
997 break;
998 case NETDEV_CHANGEMTU:
999 case NETDEV_CHANGE:
1000 rt_cache_flush(dev_net(dev), 0);
1001 break;
1002 case NETDEV_UNREGISTER_BATCH:
1003 rt_cache_flush_batch();
1004 break;
1005 }
1006 return NOTIFY_DONE;
1007}
1008
1009static struct notifier_block fib_inetaddr_notifier = {
1010 .notifier_call = fib_inetaddr_event,
1011};
1012
1013static struct notifier_block fib_netdev_notifier = {
1014 .notifier_call = fib_netdev_event,
1015};
1016
1017static int __net_init ip_fib_net_init(struct net *net)
1018{
1019 int err;
1020 unsigned int i;
1021
1022 net->ipv4.fib_table_hash = kzalloc(
1023 sizeof(struct hlist_head)*FIB_TABLE_HASHSZ, GFP_KERNEL);
1024 if (net->ipv4.fib_table_hash == NULL)
1025 return -ENOMEM;
1026
1027 for (i = 0; i < FIB_TABLE_HASHSZ; i++)
1028 INIT_HLIST_HEAD(&net->ipv4.fib_table_hash[i]);
1029
1030 err = fib4_rules_init(net);
1031 if (err < 0)
1032 goto fail;
1033 return 0;
1034
1035fail:
1036 kfree(net->ipv4.fib_table_hash);
1037 return err;
1038}
1039
1040static void ip_fib_net_exit(struct net *net)
1041{
1042 unsigned int i;
1043
1044#ifdef CONFIG_IP_MULTIPLE_TABLES
1045 fib4_rules_exit(net);
1046#endif
1047
1048 for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
1049 struct fib_table *tb;
1050 struct hlist_head *head;
1051 struct hlist_node *node, *tmp;
1052
1053 head = &net->ipv4.fib_table_hash[i];
1054 hlist_for_each_entry_safe(tb, node, tmp, head, tb_hlist) {
1055 hlist_del(node);
1056 fib_table_flush(tb);
1057 kfree(tb);
1058 }
1059 }
1060 kfree(net->ipv4.fib_table_hash);
1061}
1062
1063static int __net_init fib_net_init(struct net *net)
1064{
1065 int error;
1066
1067 error = ip_fib_net_init(net);
1068 if (error < 0)
1069 goto out;
1070 error = nl_fib_lookup_init(net);
1071 if (error < 0)
1072 goto out_nlfl;
1073 error = fib_proc_init(net);
1074 if (error < 0)
1075 goto out_proc;
1076out:
1077 return error;
1078
1079out_proc:
1080 nl_fib_lookup_exit(net);
1081out_nlfl:
1082 ip_fib_net_exit(net);
1083 goto out;
1084}
1085
1086static void __net_exit fib_net_exit(struct net *net)
1087{
1088 fib_proc_exit(net);
1089 nl_fib_lookup_exit(net);
1090 ip_fib_net_exit(net);
1091}
1092
1093static struct pernet_operations fib_net_ops = {
1094 .init = fib_net_init,
1095 .exit = fib_net_exit,
1096};
1097
1098void __init ip_fib_init(void)
1099{
1100 rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL);
1101 rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL);
1102 rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib);
1103
1104 register_pernet_subsys(&fib_net_ops);
1105 register_netdevice_notifier(&fib_netdev_notifier);
1106 register_inetaddr_notifier(&fib_inetaddr_notifier);
1107
1108 fib_hash_init();
1109}
Ubuntu Focal Linux kernel mirror