]> git.ipfire.org Git - people/ms/linux.git/blame - net/ipv6/route.c
projects / people / ms / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net: sctp: fix race for one-to-many sockets in sendmsg's auto associate
[people/ms/linux.git] / net / ipv6 / route.c
CommitLineData
1da177e4
LT		 1/*
2 * Linux INET6 implementation
3 * FIB front-end.
4 *
5 * Authors:
1ab1457c 6 * Pedro Roque <roque@di.fc.ul.pt>
1da177e4 7 *
1da177e4
LT		 8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14/* Changes:
15 *
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
c0bece9f
YH		 23 * Ville Nuorvala
24 * Fixed routing subtrees.
1da177e4
LT		 25 */
26
f3213831
JP		 27#define pr_fmt(fmt) "IPv6: " fmt
28
4fc268d2 29#include <linux/capability.h>
1da177e4 30#include <linux/errno.h>
bc3b2d7f 31#include <linux/export.h>
1da177e4
LT		 32#include <linux/types.h>
33#include <linux/times.h>
34#include <linux/socket.h>
35#include <linux/sockios.h>
36#include <linux/net.h>
37#include <linux/route.h>
38#include <linux/netdevice.h>
39#include <linux/in6.h>
7bc570c8 40#include <linux/mroute6.h>
1da177e4 41#include <linux/init.h>
1da177e4 42#include <linux/if_arp.h>
1da177e4
LT		 43#include <linux/proc_fs.h>
44#include <linux/seq_file.h>
5b7c931d 45#include <linux/nsproxy.h>
5a0e3ad6 46#include <linux/slab.h>
457c4cbc 47#include <net/net_namespace.h>
1da177e4
LT		 48#include <net/snmp.h>
49#include <net/ipv6.h>
50#include <net/ip6_fib.h>
51#include <net/ip6_route.h>
52#include <net/ndisc.h>
53#include <net/addrconf.h>
54#include <net/tcp.h>
55#include <linux/rtnetlink.h>
56#include <net/dst.h>
57#include <net/xfrm.h>
8d71740c 58#include <net/netevent.h>
21713ebc 59#include <net/netlink.h>
51ebd318 60#include <net/nexthop.h>
1da177e4
LT		 61
62#include <asm/uaccess.h>
63
64#ifdef CONFIG_SYSCTL
65#include <linux/sysctl.h>
66#endif
67
afc154e9 68enum rt6_nud_state {
7e980569
JB		 69 RT6_NUD_FAIL_HARD = -3,
70 RT6_NUD_FAIL_PROBE = -2,
71 RT6_NUD_FAIL_DO_RR = -1,
afc154e9
HFS		 72 RT6_NUD_SUCCEED = 1
73};
74
1716a961 75static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
21efcfa0 76 const struct in6_addr *dest);
1da177e4 77static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
0dbaee3b 78static unsigned int ip6_default_advmss(const struct dst_entry *dst);
ebb762f2 79static unsigned int ip6_mtu(const struct dst_entry *dst);
1da177e4
LT		 80static struct dst_entry *ip6_negative_advice(struct dst_entry *);
81static void ip6_dst_destroy(struct dst_entry *);
82static void ip6_dst_ifdown(struct dst_entry *,
83 struct net_device *dev, int how);
569d3645 84static int ip6_dst_gc(struct dst_ops *ops);
1da177e4
LT		 85
86static int ip6_pkt_discard(struct sk_buff *skb);
aad88724 87static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb);
7150aede 88static int ip6_pkt_prohibit(struct sk_buff *skb);
aad88724 89static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb);
1da177e4 90static void ip6_link_failure(struct sk_buff *skb);
6700c270
DM		 91static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
92 struct sk_buff *skb, u32 mtu);
93static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
94 struct sk_buff *skb);
52bd4c0c 95static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
1da177e4 96
70ceb4f5 97#ifdef CONFIG_IPV6_ROUTE_INFO
efa2cea0 98static struct rt6_info *rt6_add_route_info(struct net *net,
b71d1d42
ED		 99 const struct in6_addr *prefix, int prefixlen,
100 const struct in6_addr *gwaddr, int ifindex,
95c96174 101 unsigned int pref);
efa2cea0 102static struct rt6_info *rt6_get_route_info(struct net *net,
b71d1d42
ED		 103 const struct in6_addr *prefix, int prefixlen,
104 const struct in6_addr *gwaddr, int ifindex);
70ceb4f5
YH		 105#endif
106
e8243534 107static void rt6_bind_peer(struct rt6_info *rt, int create)
108{
109 struct inet_peer_base *base;
110 struct inet_peer *peer;
111
112 base = inetpeer_base_ptr(rt->_rt6i_peer);
113 if (!base)
114 return;
115
116 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
117 if (peer) {
118 if (!rt6_set_peer(rt, peer))
119 inet_putpeer(peer);
120 }
121}
122
123static struct inet_peer *__rt6_get_peer(struct rt6_info *rt, int create)
124{
125 if (rt6_has_peer(rt))
126 return rt6_peer_ptr(rt);
127
128 rt6_bind_peer(rt, create);
129 return (rt6_has_peer(rt) ? rt6_peer_ptr(rt) : NULL);
130}
131
132static struct inet_peer *rt6_get_peer_create(struct rt6_info *rt)
133{
134 return __rt6_get_peer(rt, 1);
135}
136
06582540
DM		 137static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
138{
139 struct rt6_info *rt = (struct rt6_info *) dst;
140 struct inet_peer *peer;
141 u32 *p = NULL;
142
8e2ec639
YZ		 143 if (!(rt->dst.flags & DST_HOST))
144 return NULL;
145
fbfe95a4 146 peer = rt6_get_peer_create(rt);
06582540
DM		 147 if (peer) {
148 u32 *old_p = __DST_METRICS_PTR(old);
149 unsigned long prev, new;
150
151 p = peer->metrics;
e5fd387a
MK		 152 if (inet_metrics_new(peer) ||
153 (old & DST_METRICS_FORCE_OVERWRITE))
06582540
DM		 154 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
155
156 new = (unsigned long) p;
157 prev = cmpxchg(&dst->_metrics, old, new);
158
159 if (prev != old) {
160 p = __DST_METRICS_PTR(prev);
161 if (prev & DST_METRICS_READ_ONLY)
162 p = NULL;
163 }
164 }
165 return p;
166}
167
f894cbf8
DM		 168static inline const void *choose_neigh_daddr(struct rt6_info *rt,
169 struct sk_buff *skb,
170 const void *daddr)
39232973
DM		 171{
172 struct in6_addr *p = &rt->rt6i_gateway;
173
a7563f34 174 if (!ipv6_addr_any(p))
39232973 175 return (const void *) p;
f894cbf8
DM		 176 else if (skb)
177 return &ipv6_hdr(skb)->daddr;
39232973
DM		 178 return daddr;
179}
180
f894cbf8
DM		 181static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
182 struct sk_buff *skb,
183 const void *daddr)
d3aaeb38 184{
39232973
DM		 185 struct rt6_info *rt = (struct rt6_info *) dst;
186 struct neighbour *n;
187
f894cbf8 188 daddr = choose_neigh_daddr(rt, skb, daddr);
8e022ee6 189 n = __ipv6_neigh_lookup(dst->dev, daddr);
f83c7790
DM		 190 if (n)
191 return n;
192 return neigh_create(&nd_tbl, daddr, dst->dev);
193}
194
9a7ec3a9 195static struct dst_ops ip6_dst_ops_template = {
1da177e4 196 .family = AF_INET6,
09640e63 197 .protocol = cpu_to_be16(ETH_P_IPV6),
1da177e4
LT		 198 .gc = ip6_dst_gc,
199 .gc_thresh = 1024,
200 .check = ip6_dst_check,
0dbaee3b 201 .default_advmss = ip6_default_advmss,
ebb762f2 202 .mtu = ip6_mtu,
06582540 203 .cow_metrics = ipv6_cow_metrics,
1da177e4
LT		 204 .destroy = ip6_dst_destroy,
205 .ifdown = ip6_dst_ifdown,
206 .negative_advice = ip6_negative_advice,
207 .link_failure = ip6_link_failure,
208 .update_pmtu = ip6_rt_update_pmtu,
6e157b6a 209 .redirect = rt6_do_redirect,
1ac06e03 210 .local_out = __ip6_local_out,
d3aaeb38 211 .neigh_lookup = ip6_neigh_lookup,
1da177e4
LT		 212};
213
ebb762f2 214static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
ec831ea7 215{
618f9bc7
SK		 216 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
217
218 return mtu ? : dst->dev->mtu;
ec831ea7
RD		 219}
220
6700c270
DM		 221static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
222 struct sk_buff *skb, u32 mtu)
14e50e57
DM		 223{
224}
225
6700c270
DM		 226static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
227 struct sk_buff *skb)
b587ee3b
DM		 228{
229}
230
0972ddb2
HB		 231static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
232 unsigned long old)
233{
234 return NULL;
235}
236
14e50e57
DM		 237static struct dst_ops ip6_dst_blackhole_ops = {
238 .family = AF_INET6,
09640e63 239 .protocol = cpu_to_be16(ETH_P_IPV6),
14e50e57
DM		 240 .destroy = ip6_dst_destroy,
241 .check = ip6_dst_check,
ebb762f2 242 .mtu = ip6_blackhole_mtu,
214f45c9 243 .default_advmss = ip6_default_advmss,
14e50e57 244 .update_pmtu = ip6_rt_blackhole_update_pmtu,
b587ee3b 245 .redirect = ip6_rt_blackhole_redirect,
0972ddb2 246 .cow_metrics = ip6_rt_blackhole_cow_metrics,
d3aaeb38 247 .neigh_lookup = ip6_neigh_lookup,
14e50e57
DM		 248};
249
62fa8a84 250static const u32 ip6_template_metrics[RTAX_MAX] = {
14edd87d 251 [RTAX_HOPLIMIT - 1] = 0,
62fa8a84
DM		 252};
253
fb0af4c7 254static const struct rt6_info ip6_null_entry_template = {
d8d1f30b
CG		 255 .dst = {
256 .__refcnt = ATOMIC_INIT(1),
257 .__use = 1,
2c20cbd7 258 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 259 .error = -ENETUNREACH,
d8d1f30b
CG		 260 .input = ip6_pkt_discard,
261 .output = ip6_pkt_discard_out,
1da177e4
LT		 262 },
263 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
4f724279 264 .rt6i_protocol = RTPROT_KERNEL,
1da177e4
LT		 265 .rt6i_metric = ~(u32) 0,
266 .rt6i_ref = ATOMIC_INIT(1),
267};
268
101367c2
TG		 269#ifdef CONFIG_IPV6_MULTIPLE_TABLES
270
fb0af4c7 271static const struct rt6_info ip6_prohibit_entry_template = {
d8d1f30b
CG		 272 .dst = {
273 .__refcnt = ATOMIC_INIT(1),
274 .__use = 1,
2c20cbd7 275 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 276 .error = -EACCES,
d8d1f30b
CG		 277 .input = ip6_pkt_prohibit,
278 .output = ip6_pkt_prohibit_out,
101367c2
TG		 279 },
280 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
4f724279 281 .rt6i_protocol = RTPROT_KERNEL,
101367c2
TG		 282 .rt6i_metric = ~(u32) 0,
283 .rt6i_ref = ATOMIC_INIT(1),
284};
285
fb0af4c7 286static const struct rt6_info ip6_blk_hole_entry_template = {
d8d1f30b
CG		 287 .dst = {
288 .__refcnt = ATOMIC_INIT(1),
289 .__use = 1,
2c20cbd7 290 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 291 .error = -EINVAL,
d8d1f30b 292 .input = dst_discard,
aad88724 293 .output = dst_discard_sk,
101367c2
TG		 294 },
295 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
4f724279 296 .rt6i_protocol = RTPROT_KERNEL,
101367c2
TG		 297 .rt6i_metric = ~(u32) 0,
298 .rt6i_ref = ATOMIC_INIT(1),
299};
300
301#endif
302
1da177e4 303/* allocate dst with ip6_dst_ops */
97bab73f 304static inline struct rt6_info *ip6_dst_alloc(struct net *net,
957c665f 305 struct net_device *dev,
8b96d22d
DM		 306 int flags,
307 struct fib6_table *table)
1da177e4 308{
97bab73f 309 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
6f3118b5 310 0, DST_OBSOLETE_FORCE_CHK, flags);
cf911662 311
97bab73f 312 if (rt) {
8104891b
SK		 313 struct dst_entry *dst = &rt->dst;
314
315 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
8b96d22d 316 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
51ebd318 317 INIT_LIST_HEAD(&rt->rt6i_siblings);
97bab73f 318 }
cf911662 319 return rt;
1da177e4
LT		 320}
321
322static void ip6_dst_destroy(struct dst_entry *dst)
323{
324 struct rt6_info *rt = (struct rt6_info *)dst;
325 struct inet6_dev *idev = rt->rt6i_idev;
ecd98837 326 struct dst_entry *from = dst->from;
1da177e4 327
8e2ec639
YZ		 328 if (!(rt->dst.flags & DST_HOST))
329 dst_destroy_metrics_generic(dst);
330
38308473 331 if (idev) {
1da177e4
LT		 332 rt->rt6i_idev = NULL;
333 in6_dev_put(idev);
1ab1457c 334 }
1716a961 335
ecd98837
YH		 336 dst->from = NULL;
337 dst_release(from);
1716a961 338
97bab73f
DM		 339 if (rt6_has_peer(rt)) {
340 struct inet_peer *peer = rt6_peer_ptr(rt);
b3419363
DM		 341 inet_putpeer(peer);
342 }
343}
344
1da177e4
LT		 345static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
346 int how)
347{
348 struct rt6_info *rt = (struct rt6_info *)dst;
349 struct inet6_dev *idev = rt->rt6i_idev;
5a3e55d6 350 struct net_device *loopback_dev =
c346dca1 351 dev_net(dev)->loopback_dev;
1da177e4 352
97cac082
DM		 353 if (dev != loopback_dev) {
354 if (idev && idev->dev == dev) {
355 struct inet6_dev *loopback_idev =
356 in6_dev_get(loopback_dev);
357 if (loopback_idev) {
358 rt->rt6i_idev = loopback_idev;
359 in6_dev_put(idev);
360 }
361 }
1da177e4
LT		 362 }
363}
364
a50feda5 365static bool rt6_check_expired(const struct rt6_info *rt)
1da177e4 366{
1716a961
G		 367 if (rt->rt6i_flags & RTF_EXPIRES) {
368 if (time_after(jiffies, rt->dst.expires))
a50feda5 369 return true;
1716a961 370 } else if (rt->dst.from) {
3fd91fb3 371 return rt6_check_expired((struct rt6_info *) rt->dst.from);
1716a961 372 }
a50feda5 373 return false;
1da177e4
LT		 374}
375
51ebd318
ND		 376/* Multipath route selection:
377 * Hash based function using packet header and flowlabel.
378 * Adapted from fib_info_hashfn()
379 */
380static int rt6_info_hash_nhsfn(unsigned int candidate_count,
381 const struct flowi6 *fl6)
382{
383 unsigned int val = fl6->flowi6_proto;
384
c08977bb
YH		 385 val ^= ipv6_addr_hash(&fl6->daddr);
386 val ^= ipv6_addr_hash(&fl6->saddr);
51ebd318
ND		 387
388 /* Work only if this not encapsulated */
389 switch (fl6->flowi6_proto) {
390 case IPPROTO_UDP:
391 case IPPROTO_TCP:
392 case IPPROTO_SCTP:
b3ce5ae1
ND		 393 val ^= (__force u16)fl6->fl6_sport;
394 val ^= (__force u16)fl6->fl6_dport;
51ebd318
ND		 395 break;
396
397 case IPPROTO_ICMPV6:
b3ce5ae1
ND		 398 val ^= (__force u16)fl6->fl6_icmp_type;
399 val ^= (__force u16)fl6->fl6_icmp_code;
51ebd318
ND		 400 break;
401 }
402 /* RFC6438 recommands to use flowlabel */
b3ce5ae1 403 val ^= (__force u32)fl6->flowlabel;
51ebd318
ND		 404
405 /* Perhaps, we need to tune, this function? */
406 val = val ^ (val >> 7) ^ (val >> 12);
407 return val % candidate_count;
408}
409
410static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
52bd4c0c
ND		 411 struct flowi6 *fl6, int oif,
412 int strict)
51ebd318
ND		 413{
414 struct rt6_info *sibling, *next_sibling;
415 int route_choosen;
416
417 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
418 /* Don't change the route, if route_choosen == 0
419 * (siblings does not include ourself)
420 */
421 if (route_choosen)
422 list_for_each_entry_safe(sibling, next_sibling,
423 &match->rt6i_siblings, rt6i_siblings) {
424 route_choosen--;
425 if (route_choosen == 0) {
52bd4c0c
ND		 426 if (rt6_score_route(sibling, oif, strict) < 0)
427 break;
51ebd318
ND		 428 match = sibling;
429 break;
430 }
431 }
432 return match;
433}
434
1da177e4 435/*
c71099ac 436 * Route lookup. Any table->tb6_lock is implied.
1da177e4
LT		 437 */
438
8ed67789
DL		 439static inline struct rt6_info *rt6_device_match(struct net *net,
440 struct rt6_info *rt,
b71d1d42 441 const struct in6_addr *saddr,
1da177e4 442 int oif,
d420895e 443 int flags)
1da177e4
LT		 444{
445 struct rt6_info *local = NULL;
446 struct rt6_info *sprt;
447
dd3abc4e
YH		 448 if (!oif && ipv6_addr_any(saddr))
449 goto out;
450
d8d1f30b 451 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
d1918542 452 struct net_device *dev = sprt->dst.dev;
dd3abc4e
YH		 453
454 if (oif) {
1da177e4
LT		 455 if (dev->ifindex == oif)
456 return sprt;
457 if (dev->flags & IFF_LOOPBACK) {
38308473 458 if (!sprt->rt6i_idev ||
1da177e4 459 sprt->rt6i_idev->dev->ifindex != oif) {
d420895e 460 if (flags & RT6_LOOKUP_F_IFACE && oif)
1da177e4 461 continue;
1ab1457c 462 if (local && (!oif ||
1da177e4
LT		 463 local->rt6i_idev->dev->ifindex == oif))
464 continue;
465 }
466 local = sprt;
467 }
dd3abc4e
YH		 468 } else {
469 if (ipv6_chk_addr(net, saddr, dev,
470 flags & RT6_LOOKUP_F_IFACE))
471 return sprt;
1da177e4 472 }
dd3abc4e 473 }
1da177e4 474
dd3abc4e 475 if (oif) {
1da177e4
LT		 476 if (local)
477 return local;
478
d420895e 479 if (flags & RT6_LOOKUP_F_IFACE)
8ed67789 480 return net->ipv6.ip6_null_entry;
1da177e4 481 }
dd3abc4e 482out:
1da177e4
LT		 483 return rt;
484}
485
27097255 486#ifdef CONFIG_IPV6_ROUTER_PREF
c2f17e82
HFS		 487struct __rt6_probe_work {
488 struct work_struct work;
489 struct in6_addr target;
490 struct net_device *dev;
491};
492
493static void rt6_probe_deferred(struct work_struct *w)
494{
495 struct in6_addr mcaddr;
496 struct __rt6_probe_work *work =
497 container_of(w, struct __rt6_probe_work, work);
498
499 addrconf_addr_solict_mult(&work->target, &mcaddr);
500 ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
501 dev_put(work->dev);
502 kfree(w);
503}
504
27097255
YH		 505static void rt6_probe(struct rt6_info *rt)
506{
f2c31e32 507 struct neighbour *neigh;
27097255
YH		 508 /*
509 * Okay, this does not seem to be appropriate
510 * for now, however, we need to check if it
511 * is really so; aka Router Reachability Probing.
512 *
513 * Router Reachability Probe MUST be rate-limited
514 * to no more than one per minute.
515 */
2152caea 516 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
7ff74a59 517 return;
2152caea
YH		 518 rcu_read_lock_bh();
519 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
520 if (neigh) {
521 write_lock(&neigh->lock);
522 if (neigh->nud_state & NUD_VALID)
523 goto out;
7ff74a59 524 }
2152caea
YH		 525
526 if (!neigh ||
52e16356 527 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
c2f17e82 528 struct __rt6_probe_work *work;
27097255 529
c2f17e82
HFS		 530 work = kmalloc(sizeof(*work), GFP_ATOMIC);
531
532 if (neigh && work)
7e980569 533 __neigh_set_probe_once(neigh);
c2f17e82
HFS		 534
535 if (neigh)
2152caea
YH		 536 write_unlock(&neigh->lock);
537
c2f17e82
HFS		 538 if (work) {
539 INIT_WORK(&work->work, rt6_probe_deferred);
540 work->target = rt->rt6i_gateway;
541 dev_hold(rt->dst.dev);
542 work->dev = rt->dst.dev;
543 schedule_work(&work->work);
544 }
f2c31e32 545 } else {
2152caea
YH		 546out:
547 write_unlock(&neigh->lock);
f2c31e32 548 }
2152caea 549 rcu_read_unlock_bh();
27097255
YH		 550}
551#else
552static inline void rt6_probe(struct rt6_info *rt)
553{
27097255
YH		 554}
555#endif
556
1da177e4 557/*
554cfb7e 558 * Default Router Selection (RFC 2461 6.3.6)
1da177e4 559 */
b6f99a21 560static inline int rt6_check_dev(struct rt6_info *rt, int oif)
554cfb7e 561{
d1918542 562 struct net_device *dev = rt->dst.dev;
161980f4 563 if (!oif || dev->ifindex == oif)
554cfb7e 564 return 2;
161980f4
DM		 565 if ((dev->flags & IFF_LOOPBACK) &&
566 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
567 return 1;
568 return 0;
554cfb7e 569}
1da177e4 570
afc154e9 571static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
1da177e4 572{
f2c31e32 573 struct neighbour *neigh;
afc154e9 574 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
f2c31e32 575
4d0c5911
YH		 576 if (rt->rt6i_flags & RTF_NONEXTHOP ||
577 !(rt->rt6i_flags & RTF_GATEWAY))
afc154e9 578 return RT6_NUD_SUCCEED;
145a3621
YH		 579
580 rcu_read_lock_bh();
581 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
582 if (neigh) {
583 read_lock(&neigh->lock);
554cfb7e 584 if (neigh->nud_state & NUD_VALID)
afc154e9 585 ret = RT6_NUD_SUCCEED;
398bcbeb 586#ifdef CONFIG_IPV6_ROUTER_PREF
a5a81f0b 587 else if (!(neigh->nud_state & NUD_FAILED))
afc154e9 588 ret = RT6_NUD_SUCCEED;
7e980569
JB		 589 else
590 ret = RT6_NUD_FAIL_PROBE;
398bcbeb 591#endif
145a3621 592 read_unlock(&neigh->lock);
afc154e9
HFS		 593 } else {
594 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
7e980569 595 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
a5a81f0b 596 }
145a3621
YH		 597 rcu_read_unlock_bh();
598
a5a81f0b 599 return ret;
1da177e4
LT		 600}
601
554cfb7e
YH		 602static int rt6_score_route(struct rt6_info *rt, int oif,
603 int strict)
1da177e4 604{
a5a81f0b 605 int m;
1ab1457c 606
4d0c5911 607 m = rt6_check_dev(rt, oif);
77d16f45 608 if (!m && (strict & RT6_LOOKUP_F_IFACE))
afc154e9 609 return RT6_NUD_FAIL_HARD;
ebacaaa0
YH		 610#ifdef CONFIG_IPV6_ROUTER_PREF
611 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
612#endif
afc154e9
HFS		 613 if (strict & RT6_LOOKUP_F_REACHABLE) {
614 int n = rt6_check_neigh(rt);
615 if (n < 0)
616 return n;
617 }
554cfb7e
YH		 618 return m;
619}
620
f11e6659 621static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
afc154e9
HFS		 622 int *mpri, struct rt6_info *match,
623 bool *do_rr)
554cfb7e 624{
f11e6659 625 int m;
afc154e9 626 bool match_do_rr = false;
f11e6659
DM		 627
628 if (rt6_check_expired(rt))
629 goto out;
630
631 m = rt6_score_route(rt, oif, strict);
7e980569 632 if (m == RT6_NUD_FAIL_DO_RR) {
afc154e9
HFS		 633 match_do_rr = true;
634 m = 0; /* lowest valid score */
7e980569 635 } else if (m == RT6_NUD_FAIL_HARD) {
f11e6659 636 goto out;
afc154e9
HFS		 637 }
638
639 if (strict & RT6_LOOKUP_F_REACHABLE)
640 rt6_probe(rt);
f11e6659 641
7e980569 642 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
f11e6659 643 if (m > *mpri) {
afc154e9 644 *do_rr = match_do_rr;
f11e6659
DM		 645 *mpri = m;
646 match = rt;
f11e6659 647 }
f11e6659
DM		 648out:
649 return match;
650}
651
652static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
653 struct rt6_info *rr_head,
afc154e9
HFS		 654 u32 metric, int oif, int strict,
655 bool *do_rr)
f11e6659
DM		 656{
657 struct rt6_info *rt, *match;
554cfb7e 658 int mpri = -1;
1da177e4 659
f11e6659
DM		 660 match = NULL;
661 for (rt = rr_head; rt && rt->rt6i_metric == metric;
d8d1f30b 662 rt = rt->dst.rt6_next)
afc154e9 663 match = find_match(rt, oif, strict, &mpri, match, do_rr);
f11e6659 664 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
d8d1f30b 665 rt = rt->dst.rt6_next)
afc154e9 666 match = find_match(rt, oif, strict, &mpri, match, do_rr);
1da177e4 667
f11e6659
DM		 668 return match;
669}
1da177e4 670
f11e6659
DM		 671static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
672{
673 struct rt6_info *match, *rt0;
8ed67789 674 struct net *net;
afc154e9 675 bool do_rr = false;
1da177e4 676
f11e6659
DM		 677 rt0 = fn->rr_ptr;
678 if (!rt0)
679 fn->rr_ptr = rt0 = fn->leaf;
1da177e4 680
afc154e9
HFS		 681 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
682 &do_rr);
1da177e4 683
afc154e9 684 if (do_rr) {
d8d1f30b 685 struct rt6_info *next = rt0->dst.rt6_next;
f11e6659 686
554cfb7e 687 /* no entries matched; do round-robin */
f11e6659
DM		 688 if (!next || next->rt6i_metric != rt0->rt6i_metric)
689 next = fn->leaf;
690
691 if (next != rt0)
692 fn->rr_ptr = next;
1da177e4 693 }
1da177e4 694
d1918542 695 net = dev_net(rt0->dst.dev);
a02cec21 696 return match ? match : net->ipv6.ip6_null_entry;
1da177e4
LT		 697}
698
70ceb4f5
YH		 699#ifdef CONFIG_IPV6_ROUTE_INFO
700int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
b71d1d42 701 const struct in6_addr *gwaddr)
70ceb4f5 702{
c346dca1 703 struct net *net = dev_net(dev);
70ceb4f5
YH		 704 struct route_info *rinfo = (struct route_info *) opt;
705 struct in6_addr prefix_buf, *prefix;
706 unsigned int pref;
4bed72e4 707 unsigned long lifetime;
70ceb4f5
YH		 708 struct rt6_info *rt;
709
710 if (len < sizeof(struct route_info)) {
711 return -EINVAL;
712 }
713
714 /* Sanity check for prefix_len and length */
715 if (rinfo->length > 3) {
716 return -EINVAL;
717 } else if (rinfo->prefix_len > 128) {
718 return -EINVAL;
719 } else if (rinfo->prefix_len > 64) {
720 if (rinfo->length < 2) {
721 return -EINVAL;
722 }
723 } else if (rinfo->prefix_len > 0) {
724 if (rinfo->length < 1) {
725 return -EINVAL;
726 }
727 }
728
729 pref = rinfo->route_pref;
730 if (pref == ICMPV6_ROUTER_PREF_INVALID)
3933fc95 731 return -EINVAL;
70ceb4f5 732
4bed72e4 733 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
70ceb4f5
YH		 734
735 if (rinfo->length == 3)
736 prefix = (struct in6_addr *)rinfo->prefix;
737 else {
738 /* this function is safe */
739 ipv6_addr_prefix(&prefix_buf,
740 (struct in6_addr *)rinfo->prefix,
741 rinfo->prefix_len);
742 prefix = &prefix_buf;
743 }
744
f104a567
DJ		 745 if (rinfo->prefix_len == 0)
746 rt = rt6_get_dflt_router(gwaddr, dev);
747 else
748 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
749 gwaddr, dev->ifindex);
70ceb4f5
YH		 750
751 if (rt && !lifetime) {
e0a1ad73 752 ip6_del_rt(rt);
70ceb4f5
YH		 753 rt = NULL;
754 }
755
756 if (!rt && lifetime)
efa2cea0 757 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
70ceb4f5
YH		 758 pref);
759 else if (rt)
760 rt->rt6i_flags = RTF_ROUTEINFO |
761 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
762
763 if (rt) {
1716a961
G		 764 if (!addrconf_finite_timeout(lifetime))
765 rt6_clean_expires(rt);
766 else
767 rt6_set_expires(rt, jiffies + HZ * lifetime);
768
94e187c0 769 ip6_rt_put(rt);
70ceb4f5
YH		 770 }
771 return 0;
772}
773#endif
774
a3c00e46
MKL		 775static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
776 struct in6_addr *saddr)
777{
778 struct fib6_node *pn;
779 while (1) {
780 if (fn->fn_flags & RTN_TL_ROOT)
781 return NULL;
782 pn = fn->parent;
783 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
784 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
785 else
786 fn = pn;
787 if (fn->fn_flags & RTN_RTINFO)
788 return fn;
789 }
790}
c71099ac 791
8ed67789
DL		 792static struct rt6_info *ip6_pol_route_lookup(struct net *net,
793 struct fib6_table *table,
4c9483b2 794 struct flowi6 *fl6, int flags)
1da177e4
LT		 795{
796 struct fib6_node *fn;
797 struct rt6_info *rt;
798
c71099ac 799 read_lock_bh(&table->tb6_lock);
4c9483b2 800 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
c71099ac
TG		 801restart:
802 rt = fn->leaf;
4c9483b2 803 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
51ebd318 804 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
52bd4c0c 805 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
a3c00e46
MKL		 806 if (rt == net->ipv6.ip6_null_entry) {
807 fn = fib6_backtrack(fn, &fl6->saddr);
808 if (fn)
809 goto restart;
810 }
d8d1f30b 811 dst_use(&rt->dst, jiffies);
c71099ac 812 read_unlock_bh(&table->tb6_lock);
c71099ac
TG		 813 return rt;
814
815}
816
67ba4152 817struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
ea6e574e
FW		 818 int flags)
819{
820 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
821}
822EXPORT_SYMBOL_GPL(ip6_route_lookup);
823
9acd9f3a
YH		 824struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
825 const struct in6_addr *saddr, int oif, int strict)
c71099ac 826{
4c9483b2
DM		 827 struct flowi6 fl6 = {
828 .flowi6_oif = oif,
829 .daddr = *daddr,
c71099ac
TG		 830 };
831 struct dst_entry *dst;
77d16f45 832 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
c71099ac 833
adaa70bb 834 if (saddr) {
4c9483b2 835 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
adaa70bb
TG		 836 flags |= RT6_LOOKUP_F_HAS_SADDR;
837 }
838
4c9483b2 839 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
c71099ac
TG		 840 if (dst->error == 0)
841 return (struct rt6_info *) dst;
842
843 dst_release(dst);
844
1da177e4
LT		 845 return NULL;
846}
7159039a
YH		 847EXPORT_SYMBOL(rt6_lookup);
848
c71099ac 849/* ip6_ins_rt is called with FREE table->tb6_lock.
1da177e4
LT		 850 It takes new route entry, the addition fails by any reason the
851 route is freed. In any case, if caller does not hold it, it may
852 be destroyed.
853 */
854
e5fd387a
MK		 855static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
856 struct nlattr *mx, int mx_len)
1da177e4
LT		 857{
858 int err;
c71099ac 859 struct fib6_table *table;
1da177e4 860
c71099ac
TG		 861 table = rt->rt6i_table;
862 write_lock_bh(&table->tb6_lock);
e5fd387a 863 err = fib6_add(&table->tb6_root, rt, info, mx, mx_len);
c71099ac 864 write_unlock_bh(&table->tb6_lock);
1da177e4
LT		 865
866 return err;
867}
868
40e22e8f
TG		 869int ip6_ins_rt(struct rt6_info *rt)
870{
4d1169c1 871 struct nl_info info = {
d1918542 872 .nl_net = dev_net(rt->dst.dev),
4d1169c1 873 };
e5fd387a 874 return __ip6_ins_rt(rt, &info, NULL, 0);
40e22e8f
TG		 875}
876
1716a961 877static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
21efcfa0 878 const struct in6_addr *daddr,
b71d1d42 879 const struct in6_addr *saddr)
1da177e4 880{
1da177e4
LT		 881 struct rt6_info *rt;
882
883 /*
884 * Clone the route.
885 */
886
21efcfa0 887 rt = ip6_rt_copy(ort, daddr);
1da177e4
LT		 888
889 if (rt) {
249a3630
DJ		 890 if (ort->rt6i_dst.plen != 128 &&
891 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
892 rt->rt6i_flags |= RTF_ANYCAST;
1da177e4 893
1da177e4 894 rt->rt6i_flags |= RTF_CACHE;
1da177e4
LT		 895
896#ifdef CONFIG_IPV6_SUBTREES
897 if (rt->rt6i_src.plen && saddr) {
4e3fd7a0 898 rt->rt6i_src.addr = *saddr;
1da177e4
LT		 899 rt->rt6i_src.plen = 128;
900 }
901#endif
95a9a5ba 902 }
1da177e4 903
95a9a5ba
YH		 904 return rt;
905}
1da177e4 906
21efcfa0
ED		 907static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
908 const struct in6_addr *daddr)
299d9939 909{
21efcfa0
ED		 910 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
911
887c95cc 912 if (rt)
299d9939 913 rt->rt6i_flags |= RTF_CACHE;
299d9939
YH		 914 return rt;
915}
916
8ed67789 917static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
4c9483b2 918 struct flowi6 *fl6, int flags)
1da177e4 919{
367efcb9 920 struct fib6_node *fn, *saved_fn;
519fbd87 921 struct rt6_info *rt, *nrt;
c71099ac 922 int strict = 0;
1da177e4 923 int attempts = 3;
519fbd87 924 int err;
1da177e4 925
77d16f45 926 strict |= flags & RT6_LOOKUP_F_IFACE;
367efcb9
MKL		 927 if (net->ipv6.devconf_all->forwarding == 0)
928 strict |= RT6_LOOKUP_F_REACHABLE;
1da177e4 929
a3c00e46 930redo_fib6_lookup_lock:
c71099ac 931 read_lock_bh(&table->tb6_lock);
1da177e4 932
4c9483b2 933 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
367efcb9 934 saved_fn = fn;
1da177e4 935
a3c00e46 936redo_rt6_select:
367efcb9 937 rt = rt6_select(fn, oif, strict);
52bd4c0c 938 if (rt->rt6i_nsiblings)
367efcb9 939 rt = rt6_multipath_select(rt, fl6, oif, strict);
a3c00e46
MKL		 940 if (rt == net->ipv6.ip6_null_entry) {
941 fn = fib6_backtrack(fn, &fl6->saddr);
942 if (fn)
943 goto redo_rt6_select;
367efcb9
MKL		 944 else if (strict & RT6_LOOKUP_F_REACHABLE) {
945 /* also consider unreachable route */
946 strict &= ~RT6_LOOKUP_F_REACHABLE;
947 fn = saved_fn;
948 goto redo_rt6_select;
949 } else {
950 dst_hold(&rt->dst);
951 read_unlock_bh(&table->tb6_lock);
952 goto out2;
953 }
a3c00e46
MKL		 954 }
955
d8d1f30b 956 dst_hold(&rt->dst);
c71099ac 957 read_unlock_bh(&table->tb6_lock);
fb9de91e 958
94c77bb4
MKL		 959 if (rt->rt6i_flags & RTF_CACHE)
960 goto out2;
961
c440f160 962 if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
4c9483b2 963 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
7343ff31 964 else if (!(rt->dst.flags & DST_HOST))
4c9483b2 965 nrt = rt6_alloc_clone(rt, &fl6->daddr);
7343ff31
DM		 966 else
967 goto out2;
e40cf353 968
94e187c0 969 ip6_rt_put(rt);
8ed67789 970 rt = nrt ? : net->ipv6.ip6_null_entry;
1da177e4 971
d8d1f30b 972 dst_hold(&rt->dst);
519fbd87 973 if (nrt) {
40e22e8f 974 err = ip6_ins_rt(nrt);
519fbd87 975 if (!err)
1da177e4 976 goto out2;
1da177e4 977 }
1da177e4 978
519fbd87
YH		 979 if (--attempts <= 0)
980 goto out2;
981
982 /*
c71099ac 983 * Race condition! In the gap, when table->tb6_lock was
519fbd87
YH		 984 * released someone could insert this route. Relookup.
985 */
94e187c0 986 ip6_rt_put(rt);
a3c00e46 987 goto redo_fib6_lookup_lock;
519fbd87 988
1da177e4 989out2:
d8d1f30b
CG		 990 rt->dst.lastuse = jiffies;
991 rt->dst.__use++;
c71099ac
TG		 992
993 return rt;
1da177e4
LT		 994}
995
8ed67789 996static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
4c9483b2 997 struct flowi6 *fl6, int flags)
4acad72d 998{
4c9483b2 999 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
4acad72d
PE		 1000}
1001
72331bc0
SL		 1002static struct dst_entry *ip6_route_input_lookup(struct net *net,
1003 struct net_device *dev,
1004 struct flowi6 *fl6, int flags)
1005{
1006 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1007 flags |= RT6_LOOKUP_F_IFACE;
1008
1009 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1010}
1011
c71099ac
TG		 1012void ip6_route_input(struct sk_buff *skb)
1013{
b71d1d42 1014 const struct ipv6hdr *iph = ipv6_hdr(skb);
c346dca1 1015 struct net *net = dev_net(skb->dev);
adaa70bb 1016 int flags = RT6_LOOKUP_F_HAS_SADDR;
4c9483b2
DM		 1017 struct flowi6 fl6 = {
1018 .flowi6_iif = skb->dev->ifindex,
1019 .daddr = iph->daddr,
1020 .saddr = iph->saddr,
6502ca52 1021 .flowlabel = ip6_flowinfo(iph),
4c9483b2
DM		 1022 .flowi6_mark = skb->mark,
1023 .flowi6_proto = iph->nexthdr,
c71099ac 1024 };
adaa70bb 1025
72331bc0 1026 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
c71099ac
TG		 1027}
1028
8ed67789 1029static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
4c9483b2 1030 struct flowi6 *fl6, int flags)
1da177e4 1031{
4c9483b2 1032 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
c71099ac
TG		 1033}
1034
67ba4152 1035struct dst_entry *ip6_route_output(struct net *net, const struct sock *sk,
4c9483b2 1036 struct flowi6 *fl6)
c71099ac
TG		 1037{
1038 int flags = 0;
1039
1fb9489b 1040 fl6->flowi6_iif = LOOPBACK_IFINDEX;
4dc27d1c 1041
4c9483b2 1042 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
77d16f45 1043 flags |= RT6_LOOKUP_F_IFACE;
c71099ac 1044
4c9483b2 1045 if (!ipv6_addr_any(&fl6->saddr))
adaa70bb 1046 flags |= RT6_LOOKUP_F_HAS_SADDR;
0c9a2ac1
YH		 1047 else if (sk)
1048 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
adaa70bb 1049
4c9483b2 1050 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1da177e4 1051}
7159039a 1052EXPORT_SYMBOL(ip6_route_output);
1da177e4 1053
2774c131 1054struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
14e50e57 1055{
5c1e6aa3 1056 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
14e50e57
DM		 1057 struct dst_entry *new = NULL;
1058
f5b0a874 1059 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
14e50e57 1060 if (rt) {
d8d1f30b 1061 new = &rt->dst;
14e50e57 1062
8104891b
SK		 1063 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1064 rt6_init_peer(rt, net->ipv6.peers);
1065
14e50e57 1066 new->__use = 1;
352e512c 1067 new->input = dst_discard;
aad88724 1068 new->output = dst_discard_sk;
14e50e57 1069
21efcfa0
ED		 1070 if (dst_metrics_read_only(&ort->dst))
1071 new->_metrics = ort->dst._metrics;
1072 else
1073 dst_copy_metrics(new, &ort->dst);
14e50e57
DM		 1074 rt->rt6i_idev = ort->rt6i_idev;
1075 if (rt->rt6i_idev)
1076 in6_dev_hold(rt->rt6i_idev);
14e50e57 1077
4e3fd7a0 1078 rt->rt6i_gateway = ort->rt6i_gateway;
1716a961 1079 rt->rt6i_flags = ort->rt6i_flags;
14e50e57
DM		 1080 rt->rt6i_metric = 0;
1081
1082 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1083#ifdef CONFIG_IPV6_SUBTREES
1084 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1085#endif
1086
1087 dst_free(new);
1088 }
1089
69ead7af
DM		 1090 dst_release(dst_orig);
1091 return new ? new : ERR_PTR(-ENOMEM);
14e50e57 1092}
14e50e57 1093
1da177e4
LT		 1094/*
1095 * Destination cache support functions
1096 */
1097
1098static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1099{
1100 struct rt6_info *rt;
1101
1102 rt = (struct rt6_info *) dst;
1103
6f3118b5
ND		 1104 /* All IPV6 dsts are created with ->obsolete set to the value
1105 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1106 * into this function always.
1107 */
e3bc10bd
HFS		 1108 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1109 return NULL;
a4477c4d 1110
e3bc10bd
HFS		 1111 if (rt6_check_expired(rt))
1112 return NULL;
1113
1114 return dst;
1da177e4
LT		 1115}
1116
1117static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1118{
1119 struct rt6_info *rt = (struct rt6_info *) dst;
1120
1121 if (rt) {
54c1a859
YH		 1122 if (rt->rt6i_flags & RTF_CACHE) {
1123 if (rt6_check_expired(rt)) {
1124 ip6_del_rt(rt);
1125 dst = NULL;
1126 }
1127 } else {
1da177e4 1128 dst_release(dst);
54c1a859
YH		 1129 dst = NULL;
1130 }
1da177e4 1131 }
54c1a859 1132 return dst;
1da177e4
LT		 1133}
1134
1135static void ip6_link_failure(struct sk_buff *skb)
1136{
1137 struct rt6_info *rt;
1138
3ffe533c 1139 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1da177e4 1140
adf30907 1141 rt = (struct rt6_info *) skb_dst(skb);
1da177e4 1142 if (rt) {
1eb4f758
HFS		 1143 if (rt->rt6i_flags & RTF_CACHE) {
1144 dst_hold(&rt->dst);
1145 if (ip6_del_rt(rt))
1146 dst_free(&rt->dst);
1147 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1da177e4 1148 rt->rt6i_node->fn_sernum = -1;
1eb4f758 1149 }
1da177e4
LT		 1150 }
1151}
1152
6700c270
DM		 1153static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1154 struct sk_buff *skb, u32 mtu)
1da177e4 1155{
67ba4152 1156 struct rt6_info *rt6 = (struct rt6_info *)dst;
1da177e4 1157
81aded24 1158 dst_confirm(dst);
1da177e4 1159 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
81aded24
DM		 1160 struct net *net = dev_net(dst->dev);
1161
1da177e4
LT		 1162 rt6->rt6i_flags |= RTF_MODIFIED;
1163 if (mtu < IPV6_MIN_MTU) {
defb3519 1164 u32 features = dst_metric(dst, RTAX_FEATURES);
1da177e4 1165 mtu = IPV6_MIN_MTU;
defb3519
DM		 1166 features |= RTAX_FEATURE_ALLFRAG;
1167 dst_metric_set(dst, RTAX_FEATURES, features);
1da177e4 1168 }
defb3519 1169 dst_metric_set(dst, RTAX_MTU, mtu);
81aded24 1170 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1da177e4
LT		 1171 }
1172}
1173
42ae66c8
DM		 1174void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1175 int oif, u32 mark)
81aded24
DM		 1176{
1177 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1178 struct dst_entry *dst;
1179 struct flowi6 fl6;
1180
1181 memset(&fl6, 0, sizeof(fl6));
1182 fl6.flowi6_oif = oif;
1b3c61dc 1183 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
81aded24
DM		 1184 fl6.daddr = iph->daddr;
1185 fl6.saddr = iph->saddr;
6502ca52 1186 fl6.flowlabel = ip6_flowinfo(iph);
81aded24
DM		 1187
1188 dst = ip6_route_output(net, NULL, &fl6);
1189 if (!dst->error)
6700c270 1190 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
81aded24
DM		 1191 dst_release(dst);
1192}
1193EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1194
1195void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1196{
1197 ip6_update_pmtu(skb, sock_net(sk), mtu,
1198 sk->sk_bound_dev_if, sk->sk_mark);
1199}
1200EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1201
b55b76b2
DJ		 1202/* Handle redirects */
1203struct ip6rd_flowi {
1204 struct flowi6 fl6;
1205 struct in6_addr gateway;
1206};
1207
1208static struct rt6_info *__ip6_route_redirect(struct net *net,
1209 struct fib6_table *table,
1210 struct flowi6 *fl6,
1211 int flags)
1212{
1213 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1214 struct rt6_info *rt;
1215 struct fib6_node *fn;
1216
1217 /* Get the "current" route for this destination and
1218 * check if the redirect has come from approriate router.
1219 *
1220 * RFC 4861 specifies that redirects should only be
1221 * accepted if they come from the nexthop to the target.
1222 * Due to the way the routes are chosen, this notion
1223 * is a bit fuzzy and one might need to check all possible
1224 * routes.
1225 */
1226
1227 read_lock_bh(&table->tb6_lock);
1228 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1229restart:
1230 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1231 if (rt6_check_expired(rt))
1232 continue;
1233 if (rt->dst.error)
1234 break;
1235 if (!(rt->rt6i_flags & RTF_GATEWAY))
1236 continue;
1237 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1238 continue;
1239 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1240 continue;
1241 break;
1242 }
1243
1244 if (!rt)
1245 rt = net->ipv6.ip6_null_entry;
1246 else if (rt->dst.error) {
1247 rt = net->ipv6.ip6_null_entry;
a3c00e46
MKL		 1248 } else if (rt == net->ipv6.ip6_null_entry) {
1249 fn = fib6_backtrack(fn, &fl6->saddr);
1250 if (fn)
1251 goto restart;
b55b76b2 1252 }
a3c00e46 1253
b55b76b2
DJ		 1254 dst_hold(&rt->dst);
1255
1256 read_unlock_bh(&table->tb6_lock);
1257
1258 return rt;
1259};
1260
1261static struct dst_entry *ip6_route_redirect(struct net *net,
1262 const struct flowi6 *fl6,
1263 const struct in6_addr *gateway)
1264{
1265 int flags = RT6_LOOKUP_F_HAS_SADDR;
1266 struct ip6rd_flowi rdfl;
1267
1268 rdfl.fl6 = *fl6;
1269 rdfl.gateway = *gateway;
1270
1271 return fib6_rule_lookup(net, &rdfl.fl6,
1272 flags, __ip6_route_redirect);
1273}
1274
3a5ad2ee
DM		 1275void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1276{
1277 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1278 struct dst_entry *dst;
1279 struct flowi6 fl6;
1280
1281 memset(&fl6, 0, sizeof(fl6));
e374c618 1282 fl6.flowi6_iif = LOOPBACK_IFINDEX;
3a5ad2ee
DM		 1283 fl6.flowi6_oif = oif;
1284 fl6.flowi6_mark = mark;
3a5ad2ee
DM		 1285 fl6.daddr = iph->daddr;
1286 fl6.saddr = iph->saddr;
6502ca52 1287 fl6.flowlabel = ip6_flowinfo(iph);
3a5ad2ee 1288
b55b76b2
DJ		 1289 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1290 rt6_do_redirect(dst, NULL, skb);
3a5ad2ee
DM		 1291 dst_release(dst);
1292}
1293EXPORT_SYMBOL_GPL(ip6_redirect);
1294
c92a59ec
DJ		 1295void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1296 u32 mark)
1297{
1298 const struct ipv6hdr *iph = ipv6_hdr(skb);
1299 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1300 struct dst_entry *dst;
1301 struct flowi6 fl6;
1302
1303 memset(&fl6, 0, sizeof(fl6));
e374c618 1304 fl6.flowi6_iif = LOOPBACK_IFINDEX;
c92a59ec
DJ		 1305 fl6.flowi6_oif = oif;
1306 fl6.flowi6_mark = mark;
c92a59ec
DJ		 1307 fl6.daddr = msg->dest;
1308 fl6.saddr = iph->daddr;
1309
b55b76b2
DJ		 1310 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1311 rt6_do_redirect(dst, NULL, skb);
c92a59ec
DJ		 1312 dst_release(dst);
1313}
1314
3a5ad2ee
DM		 1315void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1316{
1317 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1318}
1319EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1320
0dbaee3b 1321static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1da177e4 1322{
0dbaee3b
DM		 1323 struct net_device *dev = dst->dev;
1324 unsigned int mtu = dst_mtu(dst);
1325 struct net *net = dev_net(dev);
1326
1da177e4
LT		 1327 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1328
5578689a
DL		 1329 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1330 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1da177e4
LT		 1331
1332 /*
1ab1457c
YH		 1333 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1334 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1335 * IPV6_MAXPLEN is also valid and means: "any MSS,
1da177e4
LT		 1336 * rely only on pmtu discovery"
1337 */
1338 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1339 mtu = IPV6_MAXPLEN;
1340 return mtu;
1341}
1342
ebb762f2 1343static unsigned int ip6_mtu(const struct dst_entry *dst)
d33e4553 1344{
d33e4553 1345 struct inet6_dev *idev;
618f9bc7
SK		 1346 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1347
1348 if (mtu)
30f78d8e 1349 goto out;
618f9bc7
SK		 1350
1351 mtu = IPV6_MIN_MTU;
d33e4553
DM		 1352
1353 rcu_read_lock();
1354 idev = __in6_dev_get(dst->dev);
1355 if (idev)
1356 mtu = idev->cnf.mtu6;
1357 rcu_read_unlock();
1358
30f78d8e
ED		 1359out:
1360 return min_t(unsigned int, mtu, IP6_MAX_MTU);
d33e4553
DM		 1361}
1362
3b00944c
YH		 1363static struct dst_entry *icmp6_dst_gc_list;
1364static DEFINE_SPINLOCK(icmp6_dst_lock);
5d0bbeeb 1365
3b00944c 1366struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
87a11578 1367 struct flowi6 *fl6)
1da177e4 1368{
87a11578 1369 struct dst_entry *dst;
1da177e4
LT		 1370 struct rt6_info *rt;
1371 struct inet6_dev *idev = in6_dev_get(dev);
c346dca1 1372 struct net *net = dev_net(dev);
1da177e4 1373
38308473 1374 if (unlikely(!idev))
122bdf67 1375 return ERR_PTR(-ENODEV);
1da177e4 1376
8b96d22d 1377 rt = ip6_dst_alloc(net, dev, 0, NULL);
38308473 1378 if (unlikely(!rt)) {
1da177e4 1379 in6_dev_put(idev);
87a11578 1380 dst = ERR_PTR(-ENOMEM);
1da177e4
LT		 1381 goto out;
1382 }
1383
8e2ec639
YZ		 1384 rt->dst.flags |= DST_HOST;
1385 rt->dst.output = ip6_output;
d8d1f30b 1386 atomic_set(&rt->dst.__refcnt, 1);
550bab42 1387 rt->rt6i_gateway = fl6->daddr;
87a11578 1388 rt->rt6i_dst.addr = fl6->daddr;
8e2ec639
YZ		 1389 rt->rt6i_dst.plen = 128;
1390 rt->rt6i_idev = idev;
14edd87d 1391 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1da177e4 1392
3b00944c 1393 spin_lock_bh(&icmp6_dst_lock);
d8d1f30b
CG		 1394 rt->dst.next = icmp6_dst_gc_list;
1395 icmp6_dst_gc_list = &rt->dst;
3b00944c 1396 spin_unlock_bh(&icmp6_dst_lock);
1da177e4 1397
5578689a 1398 fib6_force_start_gc(net);
1da177e4 1399
87a11578
DM		 1400 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1401
1da177e4 1402out:
87a11578 1403 return dst;
1da177e4
LT		 1404}
1405
3d0f24a7 1406int icmp6_dst_gc(void)
1da177e4 1407{
e9476e95 1408 struct dst_entry *dst, **pprev;
3d0f24a7 1409 int more = 0;
1da177e4 1410
3b00944c
YH		 1411 spin_lock_bh(&icmp6_dst_lock);
1412 pprev = &icmp6_dst_gc_list;
5d0bbeeb 1413
1da177e4
LT		 1414 while ((dst = *pprev) != NULL) {
1415 if (!atomic_read(&dst->__refcnt)) {
1416 *pprev = dst->next;
1417 dst_free(dst);
1da177e4
LT		 1418 } else {
1419 pprev = &dst->next;
3d0f24a7 1420 ++more;
1da177e4
LT		 1421 }
1422 }
1423
3b00944c 1424 spin_unlock_bh(&icmp6_dst_lock);
5d0bbeeb 1425
3d0f24a7 1426 return more;
1da177e4
LT		 1427}
1428
1e493d19
DM		 1429static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1430 void *arg)
1431{
1432 struct dst_entry *dst, **pprev;
1433
1434 spin_lock_bh(&icmp6_dst_lock);
1435 pprev = &icmp6_dst_gc_list;
1436 while ((dst = *pprev) != NULL) {
1437 struct rt6_info *rt = (struct rt6_info *) dst;
1438 if (func(rt, arg)) {
1439 *pprev = dst->next;
1440 dst_free(dst);
1441 } else {
1442 pprev = &dst->next;
1443 }
1444 }
1445 spin_unlock_bh(&icmp6_dst_lock);
1446}
1447
569d3645 1448static int ip6_dst_gc(struct dst_ops *ops)
1da177e4 1449{
86393e52 1450 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
7019b78e
DL		 1451 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1452 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1453 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1454 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1455 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
fc66f95c 1456 int entries;
7019b78e 1457
fc66f95c 1458 entries = dst_entries_get_fast(ops);
49a18d86 1459 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
fc66f95c 1460 entries <= rt_max_size)
1da177e4
LT		 1461 goto out;
1462
6891a346 1463 net->ipv6.ip6_rt_gc_expire++;
14956643 1464 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
fc66f95c
ED		 1465 entries = dst_entries_get_slow(ops);
1466 if (entries < ops->gc_thresh)
7019b78e 1467 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1da177e4 1468out:
7019b78e 1469 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
fc66f95c 1470 return entries > rt_max_size;
1da177e4
LT		 1471}
1472
1da177e4
LT		 1473/*
1474 *
1475 */
1476
86872cb5 1477int ip6_route_add(struct fib6_config *cfg)
1da177e4
LT		 1478{
1479 int err;
5578689a 1480 struct net *net = cfg->fc_nlinfo.nl_net;
1da177e4
LT		 1481 struct rt6_info *rt = NULL;
1482 struct net_device *dev = NULL;
1483 struct inet6_dev *idev = NULL;
c71099ac 1484 struct fib6_table *table;
1da177e4
LT		 1485 int addr_type;
1486
86872cb5 1487 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1da177e4
LT		 1488 return -EINVAL;
1489#ifndef CONFIG_IPV6_SUBTREES
86872cb5 1490 if (cfg->fc_src_len)
1da177e4
LT		 1491 return -EINVAL;
1492#endif
86872cb5 1493 if (cfg->fc_ifindex) {
1da177e4 1494 err = -ENODEV;
5578689a 1495 dev = dev_get_by_index(net, cfg->fc_ifindex);
1da177e4
LT		 1496 if (!dev)
1497 goto out;
1498 idev = in6_dev_get(dev);
1499 if (!idev)
1500 goto out;
1501 }
1502
86872cb5
TG		 1503 if (cfg->fc_metric == 0)
1504 cfg->fc_metric = IP6_RT_PRIO_USER;
1da177e4 1505
d71314b4 1506 err = -ENOBUFS;
38308473
DM		 1507 if (cfg->fc_nlinfo.nlh &&
1508 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
d71314b4 1509 table = fib6_get_table(net, cfg->fc_table);
38308473 1510 if (!table) {
f3213831 1511 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
d71314b4
MV		 1512 table = fib6_new_table(net, cfg->fc_table);
1513 }
1514 } else {
1515 table = fib6_new_table(net, cfg->fc_table);
1516 }
38308473
DM		 1517
1518 if (!table)
c71099ac 1519 goto out;
c71099ac 1520
c88507fb 1521 rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1da177e4 1522
38308473 1523 if (!rt) {
1da177e4
LT		 1524 err = -ENOMEM;
1525 goto out;
1526 }
1527
1716a961
G		 1528 if (cfg->fc_flags & RTF_EXPIRES)
1529 rt6_set_expires(rt, jiffies +
1530 clock_t_to_jiffies(cfg->fc_expires));
1531 else
1532 rt6_clean_expires(rt);
1da177e4 1533
86872cb5
TG		 1534 if (cfg->fc_protocol == RTPROT_UNSPEC)
1535 cfg->fc_protocol = RTPROT_BOOT;
1536 rt->rt6i_protocol = cfg->fc_protocol;
1537
1538 addr_type = ipv6_addr_type(&cfg->fc_dst);
1da177e4
LT		 1539
1540 if (addr_type & IPV6_ADDR_MULTICAST)
d8d1f30b 1541 rt->dst.input = ip6_mc_input;
ab79ad14
MŻ		 1542 else if (cfg->fc_flags & RTF_LOCAL)
1543 rt->dst.input = ip6_input;
1da177e4 1544 else
d8d1f30b 1545 rt->dst.input = ip6_forward;
1da177e4 1546
d8d1f30b 1547 rt->dst.output = ip6_output;
1da177e4 1548
86872cb5
TG		 1549 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1550 rt->rt6i_dst.plen = cfg->fc_dst_len;
e5fd387a
MK		 1551 if (rt->rt6i_dst.plen == 128) {
1552 rt->dst.flags |= DST_HOST;
1553 dst_metrics_set_force_overwrite(&rt->dst);
8e2ec639 1554 }
e5fd387a 1555
1da177e4 1556#ifdef CONFIG_IPV6_SUBTREES
86872cb5
TG		 1557 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1558 rt->rt6i_src.plen = cfg->fc_src_len;
1da177e4
LT		 1559#endif
1560
86872cb5 1561 rt->rt6i_metric = cfg->fc_metric;
1da177e4
LT		 1562
1563 /* We cannot add true routes via loopback here,
1564 they would result in kernel looping; promote them to reject routes
1565 */
86872cb5 1566 if ((cfg->fc_flags & RTF_REJECT) ||
38308473
DM		 1567 (dev && (dev->flags & IFF_LOOPBACK) &&
1568 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1569 !(cfg->fc_flags & RTF_LOCAL))) {
1da177e4 1570 /* hold loopback dev/idev if we haven't done so. */
5578689a 1571 if (dev != net->loopback_dev) {
1da177e4
LT		 1572 if (dev) {
1573 dev_put(dev);
1574 in6_dev_put(idev);
1575 }
5578689a 1576 dev = net->loopback_dev;
1da177e4
LT		 1577 dev_hold(dev);
1578 idev = in6_dev_get(dev);
1579 if (!idev) {
1580 err = -ENODEV;
1581 goto out;
1582 }
1583 }
1da177e4 1584 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
ef2c7d7b
ND		 1585 switch (cfg->fc_type) {
1586 case RTN_BLACKHOLE:
1587 rt->dst.error = -EINVAL;
aad88724 1588 rt->dst.output = dst_discard_sk;
7150aede 1589 rt->dst.input = dst_discard;
ef2c7d7b
ND		 1590 break;
1591 case RTN_PROHIBIT:
1592 rt->dst.error = -EACCES;
7150aede
K		 1593 rt->dst.output = ip6_pkt_prohibit_out;
1594 rt->dst.input = ip6_pkt_prohibit;
ef2c7d7b 1595 break;
b4949ab2 1596 case RTN_THROW:
ef2c7d7b 1597 default:
7150aede
K		 1598 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1599 : -ENETUNREACH;
1600 rt->dst.output = ip6_pkt_discard_out;
1601 rt->dst.input = ip6_pkt_discard;
ef2c7d7b
ND		 1602 break;
1603 }
1da177e4
LT		 1604 goto install_route;
1605 }
1606
86872cb5 1607 if (cfg->fc_flags & RTF_GATEWAY) {
b71d1d42 1608 const struct in6_addr *gw_addr;
1da177e4
LT		 1609 int gwa_type;
1610
86872cb5 1611 gw_addr = &cfg->fc_gateway;
4e3fd7a0 1612 rt->rt6i_gateway = *gw_addr;
1da177e4
LT		 1613 gwa_type = ipv6_addr_type(gw_addr);
1614
1615 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1616 struct rt6_info *grt;
1617
1618 /* IPv6 strictly inhibits using not link-local
1619 addresses as nexthop address.
1620 Otherwise, router will not able to send redirects.
1621 It is very good, but in some (rare!) circumstances
1622 (SIT, PtP, NBMA NOARP links) it is handy to allow
1623 some exceptions. --ANK
1624 */
1625 err = -EINVAL;
38308473 1626 if (!(gwa_type & IPV6_ADDR_UNICAST))
1da177e4
LT		 1627 goto out;
1628
5578689a 1629 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1da177e4
LT		 1630
1631 err = -EHOSTUNREACH;
38308473 1632 if (!grt)
1da177e4
LT		 1633 goto out;
1634 if (dev) {
d1918542 1635 if (dev != grt->dst.dev) {
94e187c0 1636 ip6_rt_put(grt);
1da177e4
LT		 1637 goto out;
1638 }
1639 } else {
d1918542 1640 dev = grt->dst.dev;
1da177e4
LT		 1641 idev = grt->rt6i_idev;
1642 dev_hold(dev);
1643 in6_dev_hold(grt->rt6i_idev);
1644 }
38308473 1645 if (!(grt->rt6i_flags & RTF_GATEWAY))
1da177e4 1646 err = 0;
94e187c0 1647 ip6_rt_put(grt);
1da177e4
LT		 1648
1649 if (err)
1650 goto out;
1651 }
1652 err = -EINVAL;
38308473 1653 if (!dev || (dev->flags & IFF_LOOPBACK))
1da177e4
LT		 1654 goto out;
1655 }
1656
1657 err = -ENODEV;
38308473 1658 if (!dev)
1da177e4
LT		 1659 goto out;
1660
c3968a85
DW		 1661 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1662 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1663 err = -EINVAL;
1664 goto out;
1665 }
4e3fd7a0 1666 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
c3968a85
DW		 1667 rt->rt6i_prefsrc.plen = 128;
1668 } else
1669 rt->rt6i_prefsrc.plen = 0;
1670
86872cb5 1671 rt->rt6i_flags = cfg->fc_flags;
1da177e4
LT		 1672
1673install_route:
d8d1f30b 1674 rt->dst.dev = dev;
1da177e4 1675 rt->rt6i_idev = idev;
c71099ac 1676 rt->rt6i_table = table;
63152fc0 1677
c346dca1 1678 cfg->fc_nlinfo.nl_net = dev_net(dev);
63152fc0 1679
e5fd387a 1680 return __ip6_ins_rt(rt, &cfg->fc_nlinfo, cfg->fc_mx, cfg->fc_mx_len);
1da177e4
LT		 1681
1682out:
1683 if (dev)
1684 dev_put(dev);
1685 if (idev)
1686 in6_dev_put(idev);
1687 if (rt)
d8d1f30b 1688 dst_free(&rt->dst);
1da177e4
LT		 1689 return err;
1690}
1691
86872cb5 1692static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1da177e4
LT		 1693{
1694 int err;
c71099ac 1695 struct fib6_table *table;
d1918542 1696 struct net *net = dev_net(rt->dst.dev);
1da177e4 1697
6825a26c
G		 1698 if (rt == net->ipv6.ip6_null_entry) {
1699 err = -ENOENT;
1700 goto out;
1701 }
6c813a72 1702
c71099ac
TG		 1703 table = rt->rt6i_table;
1704 write_lock_bh(&table->tb6_lock);
86872cb5 1705 err = fib6_del(rt, info);
c71099ac 1706 write_unlock_bh(&table->tb6_lock);
1da177e4 1707
6825a26c 1708out:
94e187c0 1709 ip6_rt_put(rt);
1da177e4
LT		 1710 return err;
1711}
1712
e0a1ad73
TG		 1713int ip6_del_rt(struct rt6_info *rt)
1714{
4d1169c1 1715 struct nl_info info = {
d1918542 1716 .nl_net = dev_net(rt->dst.dev),
4d1169c1 1717 };
528c4ceb 1718 return __ip6_del_rt(rt, &info);
e0a1ad73
TG		 1719}
1720
86872cb5 1721static int ip6_route_del(struct fib6_config *cfg)
1da177e4 1722{
c71099ac 1723 struct fib6_table *table;
1da177e4
LT		 1724 struct fib6_node *fn;
1725 struct rt6_info *rt;
1726 int err = -ESRCH;
1727
5578689a 1728 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
38308473 1729 if (!table)
c71099ac
TG		 1730 return err;
1731
1732 read_lock_bh(&table->tb6_lock);
1da177e4 1733
c71099ac 1734 fn = fib6_locate(&table->tb6_root,
86872cb5
TG		 1735 &cfg->fc_dst, cfg->fc_dst_len,
1736 &cfg->fc_src, cfg->fc_src_len);
1ab1457c 1737
1da177e4 1738 if (fn) {
d8d1f30b 1739 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
86872cb5 1740 if (cfg->fc_ifindex &&
d1918542
DM		 1741 (!rt->dst.dev ||
1742 rt->dst.dev->ifindex != cfg->fc_ifindex))
1da177e4 1743 continue;
86872cb5
TG		 1744 if (cfg->fc_flags & RTF_GATEWAY &&
1745 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1da177e4 1746 continue;
86872cb5 1747 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1da177e4 1748 continue;
d8d1f30b 1749 dst_hold(&rt->dst);
c71099ac 1750 read_unlock_bh(&table->tb6_lock);
1da177e4 1751
86872cb5 1752 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1da177e4
LT		 1753 }
1754 }
c71099ac 1755 read_unlock_bh(&table->tb6_lock);
1da177e4
LT		 1756
1757 return err;
1758}
1759
6700c270 1760static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
a6279458 1761{
e8599ff4 1762 struct net *net = dev_net(skb->dev);
a6279458 1763 struct netevent_redirect netevent;
e8599ff4 1764 struct rt6_info *rt, *nrt = NULL;
e8599ff4
DM		 1765 struct ndisc_options ndopts;
1766 struct inet6_dev *in6_dev;
1767 struct neighbour *neigh;
71bcdba0 1768 struct rd_msg *msg;
6e157b6a
DM		 1769 int optlen, on_link;
1770 u8 *lladdr;
e8599ff4 1771
29a3cad5 1772 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
71bcdba0 1773 optlen -= sizeof(*msg);
e8599ff4
DM		 1774
1775 if (optlen < 0) {
6e157b6a 1776 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
e8599ff4
DM		 1777 return;
1778 }
1779
71bcdba0 1780 msg = (struct rd_msg *)icmp6_hdr(skb);
e8599ff4 1781
71bcdba0 1782 if (ipv6_addr_is_multicast(&msg->dest)) {
6e157b6a 1783 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
e8599ff4
DM		 1784 return;
1785 }
1786
6e157b6a 1787 on_link = 0;
71bcdba0 1788 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
e8599ff4 1789 on_link = 1;
71bcdba0 1790 } else if (ipv6_addr_type(&msg->target) !=
e8599ff4 1791 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
6e157b6a 1792 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
e8599ff4
DM		 1793 return;
1794 }
1795
1796 in6_dev = __in6_dev_get(skb->dev);
1797 if (!in6_dev)
1798 return;
1799 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1800 return;
1801
1802 /* RFC2461 8.1:
1803 * The IP source address of the Redirect MUST be the same as the current
1804 * first-hop router for the specified ICMP Destination Address.
1805 */
1806
71bcdba0 1807 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
e8599ff4
DM		 1808 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1809 return;
1810 }
6e157b6a
DM		 1811
1812 lladdr = NULL;
e8599ff4
DM		 1813 if (ndopts.nd_opts_tgt_lladdr) {
1814 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1815 skb->dev);
1816 if (!lladdr) {
1817 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1818 return;
1819 }
1820 }
1821
6e157b6a
DM		 1822 rt = (struct rt6_info *) dst;
1823 if (rt == net->ipv6.ip6_null_entry) {
1824 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
e8599ff4 1825 return;
6e157b6a 1826 }
e8599ff4 1827
6e157b6a
DM		 1828 /* Redirect received -> path was valid.
1829 * Look, redirects are sent only in response to data packets,
1830 * so that this nexthop apparently is reachable. --ANK
1831 */
1832 dst_confirm(&rt->dst);
a6279458 1833
71bcdba0 1834 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
6e157b6a
DM		 1835 if (!neigh)
1836 return;
a6279458 1837
1da177e4
LT		 1838 /*
1839 * We have finally decided to accept it.
1840 */
1841
1ab1457c 1842 neigh_update(neigh, lladdr, NUD_STALE,
1da177e4
LT		 1843 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1844 NEIGH_UPDATE_F_OVERRIDE|
1845 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1846 NEIGH_UPDATE_F_ISROUTER))
1847 );
1848
71bcdba0 1849 nrt = ip6_rt_copy(rt, &msg->dest);
38308473 1850 if (!nrt)
1da177e4
LT		 1851 goto out;
1852
1853 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1854 if (on_link)
1855 nrt->rt6i_flags &= ~RTF_GATEWAY;
1856
4e3fd7a0 1857 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1da177e4 1858
40e22e8f 1859 if (ip6_ins_rt(nrt))
1da177e4
LT		 1860 goto out;
1861
d8d1f30b
CG		 1862 netevent.old = &rt->dst;
1863 netevent.new = &nrt->dst;
71bcdba0 1864 netevent.daddr = &msg->dest;
60592833 1865 netevent.neigh = neigh;
8d71740c
TT		 1866 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1867
38308473 1868 if (rt->rt6i_flags & RTF_CACHE) {
6e157b6a 1869 rt = (struct rt6_info *) dst_clone(&rt->dst);
e0a1ad73 1870 ip6_del_rt(rt);
1da177e4
LT		 1871 }
1872
1873out:
e8599ff4 1874 neigh_release(neigh);
6e157b6a
DM		 1875}
1876
1da177e4
LT		 1877/*
1878 * Misc support functions
1879 */
1880
1716a961 1881static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
21efcfa0 1882 const struct in6_addr *dest)
1da177e4 1883{
d1918542 1884 struct net *net = dev_net(ort->dst.dev);
8b96d22d
DM		 1885 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1886 ort->rt6i_table);
1da177e4
LT		 1887
1888 if (rt) {
d8d1f30b
CG		 1889 rt->dst.input = ort->dst.input;
1890 rt->dst.output = ort->dst.output;
8e2ec639 1891 rt->dst.flags |= DST_HOST;
d8d1f30b 1892
4e3fd7a0 1893 rt->rt6i_dst.addr = *dest;
8e2ec639 1894 rt->rt6i_dst.plen = 128;
defb3519 1895 dst_copy_metrics(&rt->dst, &ort->dst);
d8d1f30b 1896 rt->dst.error = ort->dst.error;
1da177e4
LT		 1897 rt->rt6i_idev = ort->rt6i_idev;
1898 if (rt->rt6i_idev)
1899 in6_dev_hold(rt->rt6i_idev);
d8d1f30b 1900 rt->dst.lastuse = jiffies;
1da177e4 1901
550bab42
JA		 1902 if (ort->rt6i_flags & RTF_GATEWAY)
1903 rt->rt6i_gateway = ort->rt6i_gateway;
1904 else
1905 rt->rt6i_gateway = *dest;
1716a961 1906 rt->rt6i_flags = ort->rt6i_flags;
24f5b855 1907 rt6_set_from(rt, ort);
1da177e4
LT		 1908 rt->rt6i_metric = 0;
1909
1da177e4
LT		 1910#ifdef CONFIG_IPV6_SUBTREES
1911 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1912#endif
0f6c6392 1913 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
c71099ac 1914 rt->rt6i_table = ort->rt6i_table;
1da177e4
LT		 1915 }
1916 return rt;
1917}
1918
70ceb4f5 1919#ifdef CONFIG_IPV6_ROUTE_INFO
efa2cea0 1920static struct rt6_info *rt6_get_route_info(struct net *net,
b71d1d42
ED		 1921 const struct in6_addr *prefix, int prefixlen,
1922 const struct in6_addr *gwaddr, int ifindex)
70ceb4f5
YH		 1923{
1924 struct fib6_node *fn;
1925 struct rt6_info *rt = NULL;
c71099ac
TG		 1926 struct fib6_table *table;
1927
efa2cea0 1928 table = fib6_get_table(net, RT6_TABLE_INFO);
38308473 1929 if (!table)
c71099ac 1930 return NULL;
70ceb4f5 1931
5744dd9b 1932 read_lock_bh(&table->tb6_lock);
67ba4152 1933 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
70ceb4f5
YH		 1934 if (!fn)
1935 goto out;
1936
d8d1f30b 1937 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
d1918542 1938 if (rt->dst.dev->ifindex != ifindex)
70ceb4f5
YH		 1939 continue;
1940 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1941 continue;
1942 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1943 continue;
d8d1f30b 1944 dst_hold(&rt->dst);
70ceb4f5
YH		 1945 break;
1946 }
1947out:
5744dd9b 1948 read_unlock_bh(&table->tb6_lock);
70ceb4f5
YH		 1949 return rt;
1950}
1951
efa2cea0 1952static struct rt6_info *rt6_add_route_info(struct net *net,
b71d1d42
ED		 1953 const struct in6_addr *prefix, int prefixlen,
1954 const struct in6_addr *gwaddr, int ifindex,
95c96174 1955 unsigned int pref)
70ceb4f5 1956{
86872cb5
TG		 1957 struct fib6_config cfg = {
1958 .fc_table = RT6_TABLE_INFO,
238fc7ea 1959 .fc_metric = IP6_RT_PRIO_USER,
86872cb5
TG		 1960 .fc_ifindex = ifindex,
1961 .fc_dst_len = prefixlen,
1962 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1963 RTF_UP | RTF_PREF(pref),
15e47304 1964 .fc_nlinfo.portid = 0,
efa2cea0
DL		 1965 .fc_nlinfo.nlh = NULL,
1966 .fc_nlinfo.nl_net = net,
86872cb5
TG		 1967 };
1968
4e3fd7a0
AD		 1969 cfg.fc_dst = *prefix;
1970 cfg.fc_gateway = *gwaddr;
70ceb4f5 1971
e317da96
YH		 1972 /* We should treat it as a default route if prefix length is 0. */
1973 if (!prefixlen)
86872cb5 1974 cfg.fc_flags |= RTF_DEFAULT;
70ceb4f5 1975
86872cb5 1976 ip6_route_add(&cfg);
70ceb4f5 1977
efa2cea0 1978 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
70ceb4f5
YH		 1979}
1980#endif
1981
b71d1d42 1982struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1ab1457c 1983{
1da177e4 1984 struct rt6_info *rt;
c71099ac 1985 struct fib6_table *table;
1da177e4 1986
c346dca1 1987 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
38308473 1988 if (!table)
c71099ac 1989 return NULL;
1da177e4 1990
5744dd9b 1991 read_lock_bh(&table->tb6_lock);
67ba4152 1992 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
d1918542 1993 if (dev == rt->dst.dev &&
045927ff 1994 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1da177e4
LT		 1995 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1996 break;
1997 }
1998 if (rt)
d8d1f30b 1999 dst_hold(&rt->dst);
5744dd9b 2000 read_unlock_bh(&table->tb6_lock);
1da177e4
LT		 2001 return rt;
2002}
2003
b71d1d42 2004struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
ebacaaa0
YH		 2005 struct net_device *dev,
2006 unsigned int pref)
1da177e4 2007{
86872cb5
TG		 2008 struct fib6_config cfg = {
2009 .fc_table = RT6_TABLE_DFLT,
238fc7ea 2010 .fc_metric = IP6_RT_PRIO_USER,
86872cb5
TG		 2011 .fc_ifindex = dev->ifindex,
2012 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2013 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
15e47304 2014 .fc_nlinfo.portid = 0,
5578689a 2015 .fc_nlinfo.nlh = NULL,
c346dca1 2016 .fc_nlinfo.nl_net = dev_net(dev),
86872cb5 2017 };
1da177e4 2018
4e3fd7a0 2019 cfg.fc_gateway = *gwaddr;
1da177e4 2020
86872cb5 2021 ip6_route_add(&cfg);
1da177e4 2022
1da177e4
LT		 2023 return rt6_get_dflt_router(gwaddr, dev);
2024}
2025
7b4da532 2026void rt6_purge_dflt_routers(struct net *net)
1da177e4
LT		 2027{
2028 struct rt6_info *rt;
c71099ac
TG		 2029 struct fib6_table *table;
2030
2031 /* NOTE: Keep consistent with rt6_get_dflt_router */
7b4da532 2032 table = fib6_get_table(net, RT6_TABLE_DFLT);
38308473 2033 if (!table)
c71099ac 2034 return;
1da177e4
LT		 2035
2036restart:
c71099ac 2037 read_lock_bh(&table->tb6_lock);
d8d1f30b 2038 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
3e8b0ac3
LC		 2039 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2040 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
d8d1f30b 2041 dst_hold(&rt->dst);
c71099ac 2042 read_unlock_bh(&table->tb6_lock);
e0a1ad73 2043 ip6_del_rt(rt);
1da177e4
LT		 2044 goto restart;
2045 }
2046 }
c71099ac 2047 read_unlock_bh(&table->tb6_lock);
1da177e4
LT		 2048}
2049
5578689a
DL		 2050static void rtmsg_to_fib6_config(struct net *net,
2051 struct in6_rtmsg *rtmsg,
86872cb5
TG		 2052 struct fib6_config *cfg)
2053{
2054 memset(cfg, 0, sizeof(*cfg));
2055
2056 cfg->fc_table = RT6_TABLE_MAIN;
2057 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2058 cfg->fc_metric = rtmsg->rtmsg_metric;
2059 cfg->fc_expires = rtmsg->rtmsg_info;
2060 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2061 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2062 cfg->fc_flags = rtmsg->rtmsg_flags;
2063
5578689a 2064 cfg->fc_nlinfo.nl_net = net;
f1243c2d 2065
4e3fd7a0
AD		 2066 cfg->fc_dst = rtmsg->rtmsg_dst;
2067 cfg->fc_src = rtmsg->rtmsg_src;
2068 cfg->fc_gateway = rtmsg->rtmsg_gateway;
86872cb5
TG		 2069}
2070
5578689a 2071int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1da177e4 2072{
86872cb5 2073 struct fib6_config cfg;
1da177e4
LT		 2074 struct in6_rtmsg rtmsg;
2075 int err;
2076
67ba4152 2077 switch (cmd) {
1da177e4
LT		 2078 case SIOCADDRT: /* Add a route */
2079 case SIOCDELRT: /* Delete a route */
af31f412 2080 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4
LT		 2081 return -EPERM;
2082 err = copy_from_user(&rtmsg, arg,
2083 sizeof(struct in6_rtmsg));
2084 if (err)
2085 return -EFAULT;
86872cb5 2086
5578689a 2087 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
86872cb5 2088
1da177e4
LT		 2089 rtnl_lock();
2090 switch (cmd) {
2091 case SIOCADDRT:
86872cb5 2092 err = ip6_route_add(&cfg);
1da177e4
LT		 2093 break;
2094 case SIOCDELRT:
86872cb5 2095 err = ip6_route_del(&cfg);
1da177e4
LT		 2096 break;
2097 default:
2098 err = -EINVAL;
2099 }
2100 rtnl_unlock();
2101
2102 return err;
3ff50b79 2103 }
1da177e4
LT		 2104
2105 return -EINVAL;
2106}
2107
2108/*
2109 * Drop the packet on the floor
2110 */
2111
d5fdd6ba 2112static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1da177e4 2113{
612f09e8 2114 int type;
adf30907 2115 struct dst_entry *dst = skb_dst(skb);
612f09e8
YH		 2116 switch (ipstats_mib_noroutes) {
2117 case IPSTATS_MIB_INNOROUTES:
0660e03f 2118 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
45bb0060 2119 if (type == IPV6_ADDR_ANY) {
3bd653c8
DL		 2120 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2121 IPSTATS_MIB_INADDRERRORS);
612f09e8
YH		 2122 break;
2123 }
2124 /* FALLTHROUGH */
2125 case IPSTATS_MIB_OUTNOROUTES:
3bd653c8
DL		 2126 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2127 ipstats_mib_noroutes);
612f09e8
YH		 2128 break;
2129 }
3ffe533c 2130 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1da177e4
LT		 2131 kfree_skb(skb);
2132 return 0;
2133}
2134
9ce8ade0
TG		 2135static int ip6_pkt_discard(struct sk_buff *skb)
2136{
612f09e8 2137 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 2138}
2139
aad88724 2140static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
1da177e4 2141{
adf30907 2142 skb->dev = skb_dst(skb)->dev;
612f09e8 2143 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1da177e4
LT		 2144}
2145
9ce8ade0
TG		 2146static int ip6_pkt_prohibit(struct sk_buff *skb)
2147{
612f09e8 2148 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 2149}
2150
aad88724 2151static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
9ce8ade0 2152{
adf30907 2153 skb->dev = skb_dst(skb)->dev;
612f09e8 2154 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
9ce8ade0
TG		 2155}
2156
1da177e4
LT		 2157/*
2158 * Allocate a dst for local (unicast / anycast) address.
2159 */
2160
2161struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2162 const struct in6_addr *addr,
8f031519 2163 bool anycast)
1da177e4 2164{
c346dca1 2165 struct net *net = dev_net(idev->dev);
a3300ef4
HFS		 2166 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2167 DST_NOCOUNT, NULL);
2168 if (!rt)
1da177e4
LT		 2169 return ERR_PTR(-ENOMEM);
2170
1da177e4
LT		 2171 in6_dev_hold(idev);
2172
11d53b49 2173 rt->dst.flags |= DST_HOST;
d8d1f30b
CG		 2174 rt->dst.input = ip6_input;
2175 rt->dst.output = ip6_output;
1da177e4 2176 rt->rt6i_idev = idev;
1da177e4
LT		 2177
2178 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
58c4fb86
YH		 2179 if (anycast)
2180 rt->rt6i_flags |= RTF_ANYCAST;
2181 else
1da177e4 2182 rt->rt6i_flags |= RTF_LOCAL;
1da177e4 2183
550bab42 2184 rt->rt6i_gateway = *addr;
4e3fd7a0 2185 rt->rt6i_dst.addr = *addr;
1da177e4 2186 rt->rt6i_dst.plen = 128;
5578689a 2187 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
1da177e4 2188
d8d1f30b 2189 atomic_set(&rt->dst.__refcnt, 1);
1da177e4
LT		 2190
2191 return rt;
2192}
2193
c3968a85
DW		 2194int ip6_route_get_saddr(struct net *net,
2195 struct rt6_info *rt,
b71d1d42 2196 const struct in6_addr *daddr,
c3968a85
DW		 2197 unsigned int prefs,
2198 struct in6_addr *saddr)
2199{
67ba4152 2200 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry *)rt);
c3968a85
DW		 2201 int err = 0;
2202 if (rt->rt6i_prefsrc.plen)
4e3fd7a0 2203 *saddr = rt->rt6i_prefsrc.addr;
c3968a85
DW		 2204 else
2205 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2206 daddr, prefs, saddr);
2207 return err;
2208}
2209
2210/* remove deleted ip from prefsrc entries */
2211struct arg_dev_net_ip {
2212 struct net_device *dev;
2213 struct net *net;
2214 struct in6_addr *addr;
2215};
2216
2217static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2218{
2219 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2220 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2221 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2222
d1918542 2223 if (((void *)rt->dst.dev == dev || !dev) &&
c3968a85
DW		 2224 rt != net->ipv6.ip6_null_entry &&
2225 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2226 /* remove prefsrc entry */
2227 rt->rt6i_prefsrc.plen = 0;
2228 }
2229 return 0;
2230}
2231
2232void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2233{
2234 struct net *net = dev_net(ifp->idev->dev);
2235 struct arg_dev_net_ip adni = {
2236 .dev = ifp->idev->dev,
2237 .net = net,
2238 .addr = &ifp->addr,
2239 };
0c3584d5 2240 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
c3968a85
DW		 2241}
2242
be7a010d
DJ		 2243#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2244#define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2245
2246/* Remove routers and update dst entries when gateway turn into host. */
2247static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2248{
2249 struct in6_addr *gateway = (struct in6_addr *)arg;
2250
2251 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2252 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2253 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2254 return -1;
2255 }
2256 return 0;
2257}
2258
2259void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2260{
2261 fib6_clean_all(net, fib6_clean_tohost, gateway);
2262}
2263
8ed67789
DL		 2264struct arg_dev_net {
2265 struct net_device *dev;
2266 struct net *net;
2267};
2268
1da177e4
LT		 2269static int fib6_ifdown(struct rt6_info *rt, void *arg)
2270{
bc3ef660 2271 const struct arg_dev_net *adn = arg;
2272 const struct net_device *dev = adn->dev;
8ed67789 2273
d1918542 2274 if ((rt->dst.dev == dev || !dev) &&
c159d30c 2275 rt != adn->net->ipv6.ip6_null_entry)
1da177e4 2276 return -1;
c159d30c 2277
1da177e4
LT		 2278 return 0;
2279}
2280
f3db4851 2281void rt6_ifdown(struct net *net, struct net_device *dev)
1da177e4 2282{
8ed67789
DL		 2283 struct arg_dev_net adn = {
2284 .dev = dev,
2285 .net = net,
2286 };
2287
0c3584d5 2288 fib6_clean_all(net, fib6_ifdown, &adn);
1e493d19 2289 icmp6_clean_all(fib6_ifdown, &adn);
1da177e4
LT		 2290}
2291
95c96174 2292struct rt6_mtu_change_arg {
1da177e4 2293 struct net_device *dev;
95c96174 2294 unsigned int mtu;
1da177e4
LT		 2295};
2296
2297static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2298{
2299 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2300 struct inet6_dev *idev;
2301
2302 /* In IPv6 pmtu discovery is not optional,
2303 so that RTAX_MTU lock cannot disable it.
2304 We still use this lock to block changes
2305 caused by addrconf/ndisc.
2306 */
2307
2308 idev = __in6_dev_get(arg->dev);
38308473 2309 if (!idev)
1da177e4
LT		 2310 return 0;
2311
2312 /* For administrative MTU increase, there is no way to discover
2313 IPv6 PMTU increase, so PMTU increase should be updated here.
2314 Since RFC 1981 doesn't include administrative MTU increase
2315 update PMTU increase is a MUST. (i.e. jumbo frame)
2316 */
2317 /*
2318 If new MTU is less than route PMTU, this new MTU will be the
2319 lowest MTU in the path, update the route PMTU to reflect PMTU
2320 decreases; if new MTU is greater than route PMTU, and the
2321 old MTU is the lowest MTU in the path, update the route PMTU
2322 to reflect the increase. In this case if the other nodes' MTU
2323 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2324 PMTU discouvery.
2325 */
d1918542 2326 if (rt->dst.dev == arg->dev &&
d8d1f30b
CG		 2327 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2328 (dst_mtu(&rt->dst) >= arg->mtu ||
2329 (dst_mtu(&rt->dst) < arg->mtu &&
2330 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
defb3519 2331 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
566cfd8f 2332 }
1da177e4
LT		 2333 return 0;
2334}
2335
95c96174 2336void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
1da177e4 2337{
c71099ac
TG		 2338 struct rt6_mtu_change_arg arg = {
2339 .dev = dev,
2340 .mtu = mtu,
2341 };
1da177e4 2342
0c3584d5 2343 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
1da177e4
LT		 2344}
2345
ef7c79ed 2346static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
5176f91e 2347 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
86872cb5 2348 [RTA_OIF] = { .type = NLA_U32 },
ab364a6f 2349 [RTA_IIF] = { .type = NLA_U32 },
86872cb5
TG		 2350 [RTA_PRIORITY] = { .type = NLA_U32 },
2351 [RTA_METRICS] = { .type = NLA_NESTED },
51ebd318 2352 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
86872cb5
TG		 2353};
2354
2355static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2356 struct fib6_config *cfg)
1da177e4 2357{
86872cb5
TG		 2358 struct rtmsg *rtm;
2359 struct nlattr *tb[RTA_MAX+1];
2360 int err;
1da177e4 2361
86872cb5
TG		 2362 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2363 if (err < 0)
2364 goto errout;
1da177e4 2365
86872cb5
TG		 2366 err = -EINVAL;
2367 rtm = nlmsg_data(nlh);
2368 memset(cfg, 0, sizeof(*cfg));
2369
2370 cfg->fc_table = rtm->rtm_table;
2371 cfg->fc_dst_len = rtm->rtm_dst_len;
2372 cfg->fc_src_len = rtm->rtm_src_len;
2373 cfg->fc_flags = RTF_UP;
2374 cfg->fc_protocol = rtm->rtm_protocol;
ef2c7d7b 2375 cfg->fc_type = rtm->rtm_type;
86872cb5 2376
ef2c7d7b
ND		 2377 if (rtm->rtm_type == RTN_UNREACHABLE ||
2378 rtm->rtm_type == RTN_BLACKHOLE ||
b4949ab2
ND		 2379 rtm->rtm_type == RTN_PROHIBIT ||
2380 rtm->rtm_type == RTN_THROW)
86872cb5
TG		 2381 cfg->fc_flags |= RTF_REJECT;
2382
ab79ad14
MŻ		 2383 if (rtm->rtm_type == RTN_LOCAL)
2384 cfg->fc_flags |= RTF_LOCAL;
2385
15e47304 2386 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
86872cb5 2387 cfg->fc_nlinfo.nlh = nlh;
3b1e0a65 2388 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
86872cb5
TG		 2389
2390 if (tb[RTA_GATEWAY]) {
2391 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2392 cfg->fc_flags |= RTF_GATEWAY;
1da177e4 2393 }
86872cb5
TG		 2394
2395 if (tb[RTA_DST]) {
2396 int plen = (rtm->rtm_dst_len + 7) >> 3;
2397
2398 if (nla_len(tb[RTA_DST]) < plen)
2399 goto errout;
2400
2401 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1da177e4 2402 }
86872cb5
TG		 2403
2404 if (tb[RTA_SRC]) {
2405 int plen = (rtm->rtm_src_len + 7) >> 3;
2406
2407 if (nla_len(tb[RTA_SRC]) < plen)
2408 goto errout;
2409
2410 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1da177e4 2411 }
86872cb5 2412
c3968a85
DW		 2413 if (tb[RTA_PREFSRC])
2414 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2415
86872cb5
TG		 2416 if (tb[RTA_OIF])
2417 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2418
2419 if (tb[RTA_PRIORITY])
2420 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2421
2422 if (tb[RTA_METRICS]) {
2423 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2424 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
1da177e4 2425 }
86872cb5
TG		 2426
2427 if (tb[RTA_TABLE])
2428 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2429
51ebd318
ND		 2430 if (tb[RTA_MULTIPATH]) {
2431 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2432 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2433 }
2434
86872cb5
TG		 2435 err = 0;
2436errout:
2437 return err;
1da177e4
LT		 2438}
2439
51ebd318
ND		 2440static int ip6_route_multipath(struct fib6_config *cfg, int add)
2441{
2442 struct fib6_config r_cfg;
2443 struct rtnexthop *rtnh;
2444 int remaining;
2445 int attrlen;
2446 int err = 0, last_err = 0;
2447
2448beginning:
2449 rtnh = (struct rtnexthop *)cfg->fc_mp;
2450 remaining = cfg->fc_mp_len;
2451
2452 /* Parse a Multipath Entry */
2453 while (rtnh_ok(rtnh, remaining)) {
2454 memcpy(&r_cfg, cfg, sizeof(*cfg));
2455 if (rtnh->rtnh_ifindex)
2456 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2457
2458 attrlen = rtnh_attrlen(rtnh);
2459 if (attrlen > 0) {
2460 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2461
2462 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2463 if (nla) {
2464 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2465 r_cfg.fc_flags |= RTF_GATEWAY;
2466 }
2467 }
2468 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2469 if (err) {
2470 last_err = err;
2471 /* If we are trying to remove a route, do not stop the
2472 * loop when ip6_route_del() fails (because next hop is
2473 * already gone), we should try to remove all next hops.
2474 */
2475 if (add) {
2476 /* If add fails, we should try to delete all
2477 * next hops that have been already added.
2478 */
2479 add = 0;
2480 goto beginning;
2481 }
2482 }
1a72418b
ND		 2483 /* Because each route is added like a single route we remove
2484 * this flag after the first nexthop (if there is a collision,
2485 * we have already fail to add the first nexthop:
2486 * fib6_add_rt2node() has reject it).
2487 */
2488 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
51ebd318
ND		 2489 rtnh = rtnh_next(rtnh, &remaining);
2490 }
2491
2492 return last_err;
2493}
2494
67ba4152 2495static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
1da177e4 2496{
86872cb5
TG		 2497 struct fib6_config cfg;
2498 int err;
1da177e4 2499
86872cb5
TG		 2500 err = rtm_to_fib6_config(skb, nlh, &cfg);
2501 if (err < 0)
2502 return err;
2503
51ebd318
ND		 2504 if (cfg.fc_mp)
2505 return ip6_route_multipath(&cfg, 0);
2506 else
2507 return ip6_route_del(&cfg);
1da177e4
LT		 2508}
2509
67ba4152 2510static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
1da177e4 2511{
86872cb5
TG		 2512 struct fib6_config cfg;
2513 int err;
1da177e4 2514
86872cb5
TG		 2515 err = rtm_to_fib6_config(skb, nlh, &cfg);
2516 if (err < 0)
2517 return err;
2518
51ebd318
ND		 2519 if (cfg.fc_mp)
2520 return ip6_route_multipath(&cfg, 1);
2521 else
2522 return ip6_route_add(&cfg);
1da177e4
LT		 2523}
2524
339bf98f
TG		 2525static inline size_t rt6_nlmsg_size(void)
2526{
2527 return NLMSG_ALIGN(sizeof(struct rtmsg))
2528 + nla_total_size(16) /* RTA_SRC */
2529 + nla_total_size(16) /* RTA_DST */
2530 + nla_total_size(16) /* RTA_GATEWAY */
2531 + nla_total_size(16) /* RTA_PREFSRC */
2532 + nla_total_size(4) /* RTA_TABLE */
2533 + nla_total_size(4) /* RTA_IIF */
2534 + nla_total_size(4) /* RTA_OIF */
2535 + nla_total_size(4) /* RTA_PRIORITY */
6a2b9ce0 2536 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
339bf98f
TG		 2537 + nla_total_size(sizeof(struct rta_cacheinfo));
2538}
2539
191cd582
BH		 2540static int rt6_fill_node(struct net *net,
2541 struct sk_buff *skb, struct rt6_info *rt,
0d51aa80 2542 struct in6_addr *dst, struct in6_addr *src,
15e47304 2543 int iif, int type, u32 portid, u32 seq,
7bc570c8 2544 int prefix, int nowait, unsigned int flags)
1da177e4
LT		 2545{
2546 struct rtmsg *rtm;
2d7202bf 2547 struct nlmsghdr *nlh;
e3703b3d 2548 long expires;
9e762a4a 2549 u32 table;
1da177e4
LT		 2550
2551 if (prefix) { /* user wants prefix routes only */
2552 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2553 /* success since this is not a prefix route */
2554 return 1;
2555 }
2556 }
2557
15e47304 2558 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
38308473 2559 if (!nlh)
26932566 2560 return -EMSGSIZE;
2d7202bf
TG		 2561
2562 rtm = nlmsg_data(nlh);
1da177e4
LT		 2563 rtm->rtm_family = AF_INET6;
2564 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2565 rtm->rtm_src_len = rt->rt6i_src.plen;
2566 rtm->rtm_tos = 0;
c71099ac 2567 if (rt->rt6i_table)
9e762a4a 2568 table = rt->rt6i_table->tb6_id;
c71099ac 2569 else
9e762a4a
PM		 2570 table = RT6_TABLE_UNSPEC;
2571 rtm->rtm_table = table;
c78679e8
DM		 2572 if (nla_put_u32(skb, RTA_TABLE, table))
2573 goto nla_put_failure;
ef2c7d7b
ND		 2574 if (rt->rt6i_flags & RTF_REJECT) {
2575 switch (rt->dst.error) {
2576 case -EINVAL:
2577 rtm->rtm_type = RTN_BLACKHOLE;
2578 break;
2579 case -EACCES:
2580 rtm->rtm_type = RTN_PROHIBIT;
2581 break;
b4949ab2
ND		 2582 case -EAGAIN:
2583 rtm->rtm_type = RTN_THROW;
2584 break;
ef2c7d7b
ND		 2585 default:
2586 rtm->rtm_type = RTN_UNREACHABLE;
2587 break;
2588 }
2589 }
38308473 2590 else if (rt->rt6i_flags & RTF_LOCAL)
ab79ad14 2591 rtm->rtm_type = RTN_LOCAL;
d1918542 2592 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
1da177e4
LT		 2593 rtm->rtm_type = RTN_LOCAL;
2594 else
2595 rtm->rtm_type = RTN_UNICAST;
2596 rtm->rtm_flags = 0;
2597 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2598 rtm->rtm_protocol = rt->rt6i_protocol;
38308473 2599 if (rt->rt6i_flags & RTF_DYNAMIC)
1da177e4 2600 rtm->rtm_protocol = RTPROT_REDIRECT;
f0396f60
DO		 2601 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2602 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2603 rtm->rtm_protocol = RTPROT_RA;
2604 else
2605 rtm->rtm_protocol = RTPROT_KERNEL;
2606 }
1da177e4 2607
38308473 2608 if (rt->rt6i_flags & RTF_CACHE)
1da177e4
LT		 2609 rtm->rtm_flags |= RTM_F_CLONED;
2610
2611 if (dst) {
c78679e8
DM		 2612 if (nla_put(skb, RTA_DST, 16, dst))
2613 goto nla_put_failure;
1ab1457c 2614 rtm->rtm_dst_len = 128;
1da177e4 2615 } else if (rtm->rtm_dst_len)
c78679e8
DM		 2616 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2617 goto nla_put_failure;
1da177e4
LT		 2618#ifdef CONFIG_IPV6_SUBTREES
2619 if (src) {
c78679e8
DM		 2620 if (nla_put(skb, RTA_SRC, 16, src))
2621 goto nla_put_failure;
1ab1457c 2622 rtm->rtm_src_len = 128;
c78679e8
DM		 2623 } else if (rtm->rtm_src_len &&
2624 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2625 goto nla_put_failure;
1da177e4 2626#endif
7bc570c8
YH		 2627 if (iif) {
2628#ifdef CONFIG_IPV6_MROUTE
2629 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
8229efda 2630 int err = ip6mr_get_route(net, skb, rtm, nowait);
7bc570c8
YH		 2631 if (err <= 0) {
2632 if (!nowait) {
2633 if (err == 0)
2634 return 0;
2635 goto nla_put_failure;
2636 } else {
2637 if (err == -EMSGSIZE)
2638 goto nla_put_failure;
2639 }
2640 }
2641 } else
2642#endif
c78679e8
DM		 2643 if (nla_put_u32(skb, RTA_IIF, iif))
2644 goto nla_put_failure;
7bc570c8 2645 } else if (dst) {
1da177e4 2646 struct in6_addr saddr_buf;
c78679e8
DM		 2647 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2648 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2649 goto nla_put_failure;
1da177e4 2650 }
2d7202bf 2651
c3968a85
DW		 2652 if (rt->rt6i_prefsrc.plen) {
2653 struct in6_addr saddr_buf;
4e3fd7a0 2654 saddr_buf = rt->rt6i_prefsrc.addr;
c78679e8
DM		 2655 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2656 goto nla_put_failure;
c3968a85
DW		 2657 }
2658
defb3519 2659 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2d7202bf
TG		 2660 goto nla_put_failure;
2661
dd0cbf29
YH		 2662 if (rt->rt6i_flags & RTF_GATEWAY) {
2663 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
94f826b8 2664 goto nla_put_failure;
94f826b8 2665 }
2d7202bf 2666
c78679e8
DM		 2667 if (rt->dst.dev &&
2668 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2669 goto nla_put_failure;
2670 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2671 goto nla_put_failure;
8253947e
LW		 2672
2673 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
69cdf8f9 2674
87a50699 2675 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
e3703b3d 2676 goto nla_put_failure;
2d7202bf
TG		 2677
2678 return nlmsg_end(skb, nlh);
2679
2680nla_put_failure:
26932566
PM		 2681 nlmsg_cancel(skb, nlh);
2682 return -EMSGSIZE;
1da177e4
LT		 2683}
2684
1b43af54 2685int rt6_dump_route(struct rt6_info *rt, void *p_arg)
1da177e4
LT		 2686{
2687 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2688 int prefix;
2689
2d7202bf
TG		 2690 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2691 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
1da177e4
LT		 2692 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2693 } else
2694 prefix = 0;
2695
191cd582
BH		 2696 return rt6_fill_node(arg->net,
2697 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
15e47304 2698 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
7bc570c8 2699 prefix, 0, NLM_F_MULTI);
1da177e4
LT		 2700}
2701
67ba4152 2702static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
1da177e4 2703{
3b1e0a65 2704 struct net *net = sock_net(in_skb->sk);
ab364a6f
TG		 2705 struct nlattr *tb[RTA_MAX+1];
2706 struct rt6_info *rt;
1da177e4 2707 struct sk_buff *skb;
ab364a6f 2708 struct rtmsg *rtm;
4c9483b2 2709 struct flowi6 fl6;
72331bc0 2710 int err, iif = 0, oif = 0;
1da177e4 2711
ab364a6f
TG		 2712 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2713 if (err < 0)
2714 goto errout;
1da177e4 2715
ab364a6f 2716 err = -EINVAL;
4c9483b2 2717 memset(&fl6, 0, sizeof(fl6));
1da177e4 2718
ab364a6f
TG		 2719 if (tb[RTA_SRC]) {
2720 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2721 goto errout;
2722
4e3fd7a0 2723 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
ab364a6f
TG		 2724 }
2725
2726 if (tb[RTA_DST]) {
2727 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2728 goto errout;
2729
4e3fd7a0 2730 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
ab364a6f
TG		 2731 }
2732
2733 if (tb[RTA_IIF])
2734 iif = nla_get_u32(tb[RTA_IIF]);
2735
2736 if (tb[RTA_OIF])
72331bc0 2737 oif = nla_get_u32(tb[RTA_OIF]);
1da177e4 2738
2e47b291
LC		 2739 if (tb[RTA_MARK])
2740 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
2741
1da177e4
LT		 2742 if (iif) {
2743 struct net_device *dev;
72331bc0
SL		 2744 int flags = 0;
2745
5578689a 2746 dev = __dev_get_by_index(net, iif);
1da177e4
LT		 2747 if (!dev) {
2748 err = -ENODEV;
ab364a6f 2749 goto errout;
1da177e4 2750 }
72331bc0
SL		 2751
2752 fl6.flowi6_iif = iif;
2753
2754 if (!ipv6_addr_any(&fl6.saddr))
2755 flags |= RT6_LOOKUP_F_HAS_SADDR;
2756
2757 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2758 flags);
2759 } else {
2760 fl6.flowi6_oif = oif;
2761
2762 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
1da177e4
LT		 2763 }
2764
ab364a6f 2765 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
38308473 2766 if (!skb) {
94e187c0 2767 ip6_rt_put(rt);
ab364a6f
TG		 2768 err = -ENOBUFS;
2769 goto errout;
2770 }
1da177e4 2771
ab364a6f
TG		 2772 /* Reserve room for dummy headers, this skb can pass
2773 through good chunk of routing engine.
2774 */
459a98ed 2775 skb_reset_mac_header(skb);
ab364a6f 2776 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
1da177e4 2777
d8d1f30b 2778 skb_dst_set(skb, &rt->dst);
1da177e4 2779
4c9483b2 2780 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
15e47304 2781 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
7bc570c8 2782 nlh->nlmsg_seq, 0, 0, 0);
1da177e4 2783 if (err < 0) {
ab364a6f
TG		 2784 kfree_skb(skb);
2785 goto errout;
1da177e4
LT		 2786 }
2787
15e47304 2788 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
ab364a6f 2789errout:
1da177e4 2790 return err;
1da177e4
LT		 2791}
2792
86872cb5 2793void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
1da177e4
LT		 2794{
2795 struct sk_buff *skb;
5578689a 2796 struct net *net = info->nl_net;
528c4ceb
DL		 2797 u32 seq;
2798 int err;
2799
2800 err = -ENOBUFS;
38308473 2801 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
86872cb5 2802
339bf98f 2803 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
38308473 2804 if (!skb)
21713ebc
TG		 2805 goto errout;
2806
191cd582 2807 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
15e47304 2808 event, info->portid, seq, 0, 0, 0);
26932566
PM		 2809 if (err < 0) {
2810 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2811 WARN_ON(err == -EMSGSIZE);
2812 kfree_skb(skb);
2813 goto errout;
2814 }
15e47304 2815 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
1ce85fe4
PNA		 2816 info->nlh, gfp_any());
2817 return;
21713ebc
TG		 2818errout:
2819 if (err < 0)
5578689a 2820 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
1da177e4
LT		 2821}
2822
8ed67789 2823static int ip6_route_dev_notify(struct notifier_block *this,
351638e7 2824 unsigned long event, void *ptr)
8ed67789 2825{
351638e7 2826 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
c346dca1 2827 struct net *net = dev_net(dev);
8ed67789
DL		 2828
2829 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
d8d1f30b 2830 net->ipv6.ip6_null_entry->dst.dev = dev;
8ed67789
DL		 2831 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2832#ifdef CONFIG_IPV6_MULTIPLE_TABLES
d8d1f30b 2833 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
8ed67789 2834 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
d8d1f30b 2835 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
8ed67789
DL		 2836 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2837#endif
2838 }
2839
2840 return NOTIFY_OK;
2841}
2842
1da177e4
LT		 2843/*
2844 * /proc
2845 */
2846
2847#ifdef CONFIG_PROC_FS
2848
33120b30
AD		 2849static const struct file_operations ipv6_route_proc_fops = {
2850 .owner = THIS_MODULE,
2851 .open = ipv6_route_open,
2852 .read = seq_read,
2853 .llseek = seq_lseek,
8d2ca1d7 2854 .release = seq_release_net,
33120b30
AD		 2855};
2856
1da177e4
LT		 2857static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2858{
69ddb805 2859 struct net *net = (struct net *)seq->private;
1da177e4 2860 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
69ddb805
DL		 2861 net->ipv6.rt6_stats->fib_nodes,
2862 net->ipv6.rt6_stats->fib_route_nodes,
2863 net->ipv6.rt6_stats->fib_rt_alloc,
2864 net->ipv6.rt6_stats->fib_rt_entries,
2865 net->ipv6.rt6_stats->fib_rt_cache,
fc66f95c 2866 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
69ddb805 2867 net->ipv6.rt6_stats->fib_discarded_routes);
1da177e4
LT		 2868
2869 return 0;
2870}
2871
2872static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2873{
de05c557 2874 return single_open_net(inode, file, rt6_stats_seq_show);
69ddb805
DL		 2875}
2876
9a32144e 2877static const struct file_operations rt6_stats_seq_fops = {
1da177e4
LT		 2878 .owner = THIS_MODULE,
2879 .open = rt6_stats_seq_open,
2880 .read = seq_read,
2881 .llseek = seq_lseek,
b6fcbdb4 2882 .release = single_release_net,
1da177e4
LT		 2883};
2884#endif /* CONFIG_PROC_FS */
2885
2886#ifdef CONFIG_SYSCTL
2887
1da177e4 2888static
fe2c6338 2889int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
1da177e4
LT		 2890 void __user *buffer, size_t *lenp, loff_t *ppos)
2891{
c486da34
LAG		 2892 struct net *net;
2893 int delay;
2894 if (!write)
1da177e4 2895 return -EINVAL;
c486da34
LAG		 2896
2897 net = (struct net *)ctl->extra1;
2898 delay = net->ipv6.sysctl.flush_delay;
2899 proc_dointvec(ctl, write, buffer, lenp, ppos);
2ac3ac8f 2900 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
c486da34 2901 return 0;
1da177e4
LT		 2902}
2903
fe2c6338 2904struct ctl_table ipv6_route_table_template[] = {
1ab1457c 2905 {
1da177e4 2906 .procname = "flush",
4990509f 2907 .data = &init_net.ipv6.sysctl.flush_delay,
1da177e4 2908 .maxlen = sizeof(int),
89c8b3a1 2909 .mode = 0200,
6d9f239a 2910 .proc_handler = ipv6_sysctl_rtcache_flush
1da177e4
LT		 2911 },
2912 {
1da177e4 2913 .procname = "gc_thresh",
9a7ec3a9 2914 .data = &ip6_dst_ops_template.gc_thresh,
1da177e4
LT		 2915 .maxlen = sizeof(int),
2916 .mode = 0644,
6d9f239a 2917 .proc_handler = proc_dointvec,
1da177e4
LT		 2918 },
2919 {
1da177e4 2920 .procname = "max_size",
4990509f 2921 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
1da177e4
LT		 2922 .maxlen = sizeof(int),
2923 .mode = 0644,
6d9f239a 2924 .proc_handler = proc_dointvec,
1da177e4
LT		 2925 },
2926 {
1da177e4 2927 .procname = "gc_min_interval",
4990509f 2928 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 2929 .maxlen = sizeof(int),
2930 .mode = 0644,
6d9f239a 2931 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 2932 },
2933 {
1da177e4 2934 .procname = "gc_timeout",
4990509f 2935 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
1da177e4
LT		 2936 .maxlen = sizeof(int),
2937 .mode = 0644,
6d9f239a 2938 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 2939 },
2940 {
1da177e4 2941 .procname = "gc_interval",
4990509f 2942 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
1da177e4
LT		 2943 .maxlen = sizeof(int),
2944 .mode = 0644,
6d9f239a 2945 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 2946 },
2947 {
1da177e4 2948 .procname = "gc_elasticity",
4990509f 2949 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
1da177e4
LT		 2950 .maxlen = sizeof(int),
2951 .mode = 0644,
f3d3f616 2952 .proc_handler = proc_dointvec,
1da177e4
LT		 2953 },
2954 {
1da177e4 2955 .procname = "mtu_expires",
4990509f 2956 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
1da177e4
LT		 2957 .maxlen = sizeof(int),
2958 .mode = 0644,
6d9f239a 2959 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 2960 },
2961 {
1da177e4 2962 .procname = "min_adv_mss",
4990509f 2963 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
1da177e4
LT		 2964 .maxlen = sizeof(int),
2965 .mode = 0644,
f3d3f616 2966 .proc_handler = proc_dointvec,
1da177e4
LT		 2967 },
2968 {
1da177e4 2969 .procname = "gc_min_interval_ms",
4990509f 2970 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 2971 .maxlen = sizeof(int),
2972 .mode = 0644,
6d9f239a 2973 .proc_handler = proc_dointvec_ms_jiffies,
1da177e4 2974 },
f8572d8f 2975 { }
1da177e4
LT		 2976};
2977
2c8c1e72 2978struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
760f2d01
DL		 2979{
2980 struct ctl_table *table;
2981
2982 table = kmemdup(ipv6_route_table_template,
2983 sizeof(ipv6_route_table_template),
2984 GFP_KERNEL);
5ee09105
YH		 2985
2986 if (table) {
2987 table[0].data = &net->ipv6.sysctl.flush_delay;
c486da34 2988 table[0].extra1 = net;
86393e52 2989 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
5ee09105
YH		 2990 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2991 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2992 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2993 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2994 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2995 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2996 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
9c69fabe 2997 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
464dc801
EB		 2998
2999 /* Don't export sysctls to unprivileged users */
3000 if (net->user_ns != &init_user_ns)
3001 table[0].procname = NULL;
5ee09105
YH		 3002 }
3003
760f2d01
DL		 3004 return table;
3005}
1da177e4
LT		 3006#endif
3007
2c8c1e72 3008static int __net_init ip6_route_net_init(struct net *net)
cdb18761 3009{
633d424b 3010 int ret = -ENOMEM;
8ed67789 3011
86393e52
AD		 3012 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3013 sizeof(net->ipv6.ip6_dst_ops));
f2fc6a54 3014
fc66f95c
ED		 3015 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3016 goto out_ip6_dst_ops;
3017
8ed67789
DL		 3018 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3019 sizeof(*net->ipv6.ip6_null_entry),
3020 GFP_KERNEL);
3021 if (!net->ipv6.ip6_null_entry)
fc66f95c 3022 goto out_ip6_dst_entries;
d8d1f30b 3023 net->ipv6.ip6_null_entry->dst.path =
8ed67789 3024 (struct dst_entry *)net->ipv6.ip6_null_entry;
d8d1f30b 3025 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 3026 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3027 ip6_template_metrics, true);
8ed67789
DL		 3028
3029#ifdef CONFIG_IPV6_MULTIPLE_TABLES
3030 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3031 sizeof(*net->ipv6.ip6_prohibit_entry),
3032 GFP_KERNEL);
68fffc67
PZ		 3033 if (!net->ipv6.ip6_prohibit_entry)
3034 goto out_ip6_null_entry;
d8d1f30b 3035 net->ipv6.ip6_prohibit_entry->dst.path =
8ed67789 3036 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
d8d1f30b 3037 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 3038 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3039 ip6_template_metrics, true);
8ed67789
DL		 3040
3041 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3042 sizeof(*net->ipv6.ip6_blk_hole_entry),
3043 GFP_KERNEL);
68fffc67
PZ		 3044 if (!net->ipv6.ip6_blk_hole_entry)
3045 goto out_ip6_prohibit_entry;
d8d1f30b 3046 net->ipv6.ip6_blk_hole_entry->dst.path =
8ed67789 3047 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
d8d1f30b 3048 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 3049 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3050 ip6_template_metrics, true);
8ed67789
DL		 3051#endif
3052
b339a47c
PZ		 3053 net->ipv6.sysctl.flush_delay = 0;
3054 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3055 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3056 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3057 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3058 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3059 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3060 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3061
6891a346
BT		 3062 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3063
8ed67789
DL		 3064 ret = 0;
3065out:
3066 return ret;
f2fc6a54 3067
68fffc67
PZ		 3068#ifdef CONFIG_IPV6_MULTIPLE_TABLES
3069out_ip6_prohibit_entry:
3070 kfree(net->ipv6.ip6_prohibit_entry);
3071out_ip6_null_entry:
3072 kfree(net->ipv6.ip6_null_entry);
3073#endif
fc66f95c
ED		 3074out_ip6_dst_entries:
3075 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
f2fc6a54 3076out_ip6_dst_ops:
f2fc6a54 3077 goto out;
cdb18761
DL		 3078}
3079
2c8c1e72 3080static void __net_exit ip6_route_net_exit(struct net *net)
cdb18761 3081{
8ed67789
DL		 3082 kfree(net->ipv6.ip6_null_entry);
3083#ifdef CONFIG_IPV6_MULTIPLE_TABLES
3084 kfree(net->ipv6.ip6_prohibit_entry);
3085 kfree(net->ipv6.ip6_blk_hole_entry);
3086#endif
41bb78b4 3087 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
cdb18761
DL		 3088}
3089
d189634e
TG		 3090static int __net_init ip6_route_net_init_late(struct net *net)
3091{
3092#ifdef CONFIG_PROC_FS
d4beaa66
G		 3093 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3094 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
d189634e
TG		 3095#endif
3096 return 0;
3097}
3098
3099static void __net_exit ip6_route_net_exit_late(struct net *net)
3100{
3101#ifdef CONFIG_PROC_FS
ece31ffd
G		 3102 remove_proc_entry("ipv6_route", net->proc_net);
3103 remove_proc_entry("rt6_stats", net->proc_net);
d189634e
TG		 3104#endif
3105}
3106
cdb18761
DL		 3107static struct pernet_operations ip6_route_net_ops = {
3108 .init = ip6_route_net_init,
3109 .exit = ip6_route_net_exit,
3110};
3111
c3426b47
DM		 3112static int __net_init ipv6_inetpeer_init(struct net *net)
3113{
3114 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3115
3116 if (!bp)
3117 return -ENOMEM;
3118 inet_peer_base_init(bp);
3119 net->ipv6.peers = bp;
3120 return 0;
3121}
3122
3123static void __net_exit ipv6_inetpeer_exit(struct net *net)
3124{
3125 struct inet_peer_base *bp = net->ipv6.peers;
3126
3127 net->ipv6.peers = NULL;
56a6b248 3128 inetpeer_invalidate_tree(bp);
c3426b47
DM		 3129 kfree(bp);
3130}
3131
2b823f72 3132static struct pernet_operations ipv6_inetpeer_ops = {
c3426b47
DM		 3133 .init = ipv6_inetpeer_init,
3134 .exit = ipv6_inetpeer_exit,
3135};
3136
d189634e
TG		 3137static struct pernet_operations ip6_route_net_late_ops = {
3138 .init = ip6_route_net_init_late,
3139 .exit = ip6_route_net_exit_late,
3140};
3141
8ed67789
DL		 3142static struct notifier_block ip6_route_dev_notifier = {
3143 .notifier_call = ip6_route_dev_notify,
3144 .priority = 0,
3145};
3146
433d49c3 3147int __init ip6_route_init(void)
1da177e4 3148{
433d49c3
DL		 3149 int ret;
3150
9a7ec3a9
DL		 3151 ret = -ENOMEM;
3152 ip6_dst_ops_template.kmem_cachep =
e5d679f3 3153 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
f845ab6b 3154 SLAB_HWCACHE_ALIGN, NULL);
9a7ec3a9 3155 if (!ip6_dst_ops_template.kmem_cachep)
c19a28e1 3156 goto out;
14e50e57 3157
fc66f95c 3158 ret = dst_entries_init(&ip6_dst_blackhole_ops);
8ed67789 3159 if (ret)
bdb3289f 3160 goto out_kmem_cache;
bdb3289f 3161
c3426b47
DM		 3162 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3163 if (ret)
e8803b6c 3164 goto out_dst_entries;
2a0c451a 3165
7e52b33b
DM		 3166 ret = register_pernet_subsys(&ip6_route_net_ops);
3167 if (ret)
3168 goto out_register_inetpeer;
c3426b47 3169
5dc121e9
AE		 3170 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3171
8ed67789
DL		 3172 /* Registering of the loopback is done before this portion of code,
3173 * the loopback reference in rt6_info will not be taken, do it
3174 * manually for init_net */
d8d1f30b 3175 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
8ed67789
DL		 3176 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3177 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
d8d1f30b 3178 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
8ed67789 3179 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
d8d1f30b 3180 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
8ed67789
DL		 3181 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3182 #endif
e8803b6c 3183 ret = fib6_init();
433d49c3 3184 if (ret)
8ed67789 3185 goto out_register_subsys;
433d49c3 3186
433d49c3
DL		 3187 ret = xfrm6_init();
3188 if (ret)
e8803b6c 3189 goto out_fib6_init;
c35b7e72 3190
433d49c3
DL		 3191 ret = fib6_rules_init();
3192 if (ret)
3193 goto xfrm6_init;
7e5449c2 3194
d189634e
TG		 3195 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3196 if (ret)
3197 goto fib6_rules_init;
3198
433d49c3 3199 ret = -ENOBUFS;
c7ac8679
GR		 3200 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3201 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3202 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
d189634e 3203 goto out_register_late_subsys;
c127ea2c 3204
8ed67789 3205 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
cdb18761 3206 if (ret)
d189634e 3207 goto out_register_late_subsys;
8ed67789 3208
433d49c3
DL		 3209out:
3210 return ret;
3211
d189634e
TG		 3212out_register_late_subsys:
3213 unregister_pernet_subsys(&ip6_route_net_late_ops);
433d49c3 3214fib6_rules_init:
433d49c3
DL		 3215 fib6_rules_cleanup();
3216xfrm6_init:
433d49c3 3217 xfrm6_fini();
2a0c451a
TG		 3218out_fib6_init:
3219 fib6_gc_cleanup();
8ed67789
DL		 3220out_register_subsys:
3221 unregister_pernet_subsys(&ip6_route_net_ops);
7e52b33b
DM		 3222out_register_inetpeer:
3223 unregister_pernet_subsys(&ipv6_inetpeer_ops);
fc66f95c
ED		 3224out_dst_entries:
3225 dst_entries_destroy(&ip6_dst_blackhole_ops);
433d49c3 3226out_kmem_cache:
f2fc6a54 3227 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
433d49c3 3228 goto out;
1da177e4
LT		 3229}
3230
3231void ip6_route_cleanup(void)
3232{
8ed67789 3233 unregister_netdevice_notifier(&ip6_route_dev_notifier);
d189634e 3234 unregister_pernet_subsys(&ip6_route_net_late_ops);
101367c2 3235 fib6_rules_cleanup();
1da177e4 3236 xfrm6_fini();
1da177e4 3237 fib6_gc_cleanup();
c3426b47 3238 unregister_pernet_subsys(&ipv6_inetpeer_ops);
8ed67789 3239 unregister_pernet_subsys(&ip6_route_net_ops);
41bb78b4 3240 dst_entries_destroy(&ip6_dst_blackhole_ops);
f2fc6a54 3241 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
1da177e4 3242}
Michael's Linux tree.