]> bbs.cooldavid.org Git - net-next-2.6.git/blame - net/ipv6/route.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
[NETNS][IPV6]: Make mld_max_msf readonly in other namespaces.
[net-next-2.6.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
LT		 7 *
8 * $Id: route.c,v 1.56 2001/10/31 21:55:55 davem Exp $
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
14 */
15
16/* Changes:
17 *
18 * YOSHIFUJI Hideaki @USAGI
19 * reworked default router selection.
20 * - respect outgoing interface
21 * - select from (probably) reachable routers (i.e.
22 * routers in REACHABLE, STALE, DELAY or PROBE states).
23 * - always select the same router if it is (probably)
24 * reachable. otherwise, round-robin the list.
c0bece9f
YH		 25 * Ville Nuorvala
26 * Fixed routing subtrees.
1da177e4
LT		 27 */
28
4fc268d2 29#include <linux/capability.h>
1da177e4
LT		 30#include <linux/errno.h>
31#include <linux/types.h>
32#include <linux/times.h>
33#include <linux/socket.h>
34#include <linux/sockios.h>
35#include <linux/net.h>
36#include <linux/route.h>
37#include <linux/netdevice.h>
38#include <linux/in6.h>
39#include <linux/init.h>
1da177e4 40#include <linux/if_arp.h>
1da177e4
LT		 41#include <linux/proc_fs.h>
42#include <linux/seq_file.h>
457c4cbc 43#include <net/net_namespace.h>
1da177e4
LT		 44#include <net/snmp.h>
45#include <net/ipv6.h>
46#include <net/ip6_fib.h>
47#include <net/ip6_route.h>
48#include <net/ndisc.h>
49#include <net/addrconf.h>
50#include <net/tcp.h>
51#include <linux/rtnetlink.h>
52#include <net/dst.h>
53#include <net/xfrm.h>
8d71740c 54#include <net/netevent.h>
21713ebc 55#include <net/netlink.h>
1da177e4
LT		 56
57#include <asm/uaccess.h>
58
59#ifdef CONFIG_SYSCTL
60#include <linux/sysctl.h>
61#endif
62
63/* Set to 3 to get tracing. */
64#define RT6_DEBUG 2
65
66#if RT6_DEBUG >= 3
67#define RDBG(x) printk x
68#define RT6_TRACE(x...) printk(KERN_DEBUG x)
69#else
70#define RDBG(x)
71#define RT6_TRACE(x...) do { ; } while (0)
72#endif
73
519fbd87 74#define CLONE_OFFLINK_ROUTE 0
1da177e4
LT		 75
76static int ip6_rt_max_size = 4096;
77static int ip6_rt_gc_min_interval = HZ / 2;
78static int ip6_rt_gc_timeout = 60*HZ;
79int ip6_rt_gc_interval = 30*HZ;
80static int ip6_rt_gc_elasticity = 9;
81static int ip6_rt_mtu_expires = 10*60*HZ;
82static int ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
83
84static struct rt6_info * ip6_rt_copy(struct rt6_info *ort);
85static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
86static struct dst_entry *ip6_negative_advice(struct dst_entry *);
87static void ip6_dst_destroy(struct dst_entry *);
88static void ip6_dst_ifdown(struct dst_entry *,
89 struct net_device *dev, int how);
90static int ip6_dst_gc(void);
91
92static int ip6_pkt_discard(struct sk_buff *skb);
93static int ip6_pkt_discard_out(struct sk_buff *skb);
94static void ip6_link_failure(struct sk_buff *skb);
95static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
96
70ceb4f5
YH		 97#ifdef CONFIG_IPV6_ROUTE_INFO
98static struct rt6_info *rt6_add_route_info(struct in6_addr *prefix, int prefixlen,
99 struct in6_addr *gwaddr, int ifindex,
100 unsigned pref);
101static struct rt6_info *rt6_get_route_info(struct in6_addr *prefix, int prefixlen,
102 struct in6_addr *gwaddr, int ifindex);
103#endif
104
1da177e4
LT		 105static struct dst_ops ip6_dst_ops = {
106 .family = AF_INET6,
107 .protocol = __constant_htons(ETH_P_IPV6),
108 .gc = ip6_dst_gc,
109 .gc_thresh = 1024,
110 .check = ip6_dst_check,
111 .destroy = ip6_dst_destroy,
112 .ifdown = ip6_dst_ifdown,
113 .negative_advice = ip6_negative_advice,
114 .link_failure = ip6_link_failure,
115 .update_pmtu = ip6_rt_update_pmtu,
862b82c6 116 .local_out = ip6_local_out,
1da177e4
LT		 117 .entry_size = sizeof(struct rt6_info),
118};
119
14e50e57
DM		 120static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
121{
122}
123
124static struct dst_ops ip6_dst_blackhole_ops = {
125 .family = AF_INET6,
126 .protocol = __constant_htons(ETH_P_IPV6),
127 .destroy = ip6_dst_destroy,
128 .check = ip6_dst_check,
129 .update_pmtu = ip6_rt_blackhole_update_pmtu,
130 .entry_size = sizeof(struct rt6_info),
131};
132
1da177e4
LT		 133struct rt6_info ip6_null_entry = {
134 .u = {
135 .dst = {
136 .__refcnt = ATOMIC_INIT(1),
137 .__use = 1,
1da177e4
LT		 138 .obsolete = -1,
139 .error = -ENETUNREACH,
140 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
141 .input = ip6_pkt_discard,
142 .output = ip6_pkt_discard_out,
143 .ops = &ip6_dst_ops,
144 .path = (struct dst_entry*)&ip6_null_entry,
145 }
146 },
147 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
148 .rt6i_metric = ~(u32) 0,
149 .rt6i_ref = ATOMIC_INIT(1),
150};
151
101367c2
TG		 152#ifdef CONFIG_IPV6_MULTIPLE_TABLES
153
6723ab54
DM		 154static int ip6_pkt_prohibit(struct sk_buff *skb);
155static int ip6_pkt_prohibit_out(struct sk_buff *skb);
6723ab54 156
101367c2
TG		 157struct rt6_info ip6_prohibit_entry = {
158 .u = {
159 .dst = {
160 .__refcnt = ATOMIC_INIT(1),
161 .__use = 1,
101367c2
TG		 162 .obsolete = -1,
163 .error = -EACCES,
164 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
9ce8ade0
TG		 165 .input = ip6_pkt_prohibit,
166 .output = ip6_pkt_prohibit_out,
101367c2
TG		 167 .ops = &ip6_dst_ops,
168 .path = (struct dst_entry*)&ip6_prohibit_entry,
169 }
170 },
171 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
172 .rt6i_metric = ~(u32) 0,
173 .rt6i_ref = ATOMIC_INIT(1),
174};
175
176struct rt6_info ip6_blk_hole_entry = {
177 .u = {
178 .dst = {
179 .__refcnt = ATOMIC_INIT(1),
180 .__use = 1,
101367c2
TG		 181 .obsolete = -1,
182 .error = -EINVAL,
183 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
352e512c
HX		 184 .input = dst_discard,
185 .output = dst_discard,
101367c2
TG		 186 .ops = &ip6_dst_ops,
187 .path = (struct dst_entry*)&ip6_blk_hole_entry,
188 }
189 },
190 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
191 .rt6i_metric = ~(u32) 0,
192 .rt6i_ref = ATOMIC_INIT(1),
193};
194
195#endif
196
1da177e4
LT		 197/* allocate dst with ip6_dst_ops */
198static __inline__ struct rt6_info *ip6_dst_alloc(void)
199{
200 return (struct rt6_info *)dst_alloc(&ip6_dst_ops);
201}
202
203static void ip6_dst_destroy(struct dst_entry *dst)
204{
205 struct rt6_info *rt = (struct rt6_info *)dst;
206 struct inet6_dev *idev = rt->rt6i_idev;
207
208 if (idev != NULL) {
209 rt->rt6i_idev = NULL;
210 in6_dev_put(idev);
1ab1457c 211 }
1da177e4
LT		 212}
213
214static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
215 int how)
216{
217 struct rt6_info *rt = (struct rt6_info *)dst;
218 struct inet6_dev *idev = rt->rt6i_idev;
5a3e55d6
DL		 219 struct net_device *loopback_dev =
220 dev->nd_net->loopback_dev;
1da177e4 221
5a3e55d6
DL		 222 if (dev != loopback_dev && idev != NULL && idev->dev == dev) {
223 struct inet6_dev *loopback_idev =
224 in6_dev_get(loopback_dev);
1da177e4
LT		 225 if (loopback_idev != NULL) {
226 rt->rt6i_idev = loopback_idev;
227 in6_dev_put(idev);
228 }
229 }
230}
231
232static __inline__ int rt6_check_expired(const struct rt6_info *rt)
233{
234 return (rt->rt6i_flags & RTF_EXPIRES &&
235 time_after(jiffies, rt->rt6i_expires));
236}
237
c71099ac
TG		 238static inline int rt6_need_strict(struct in6_addr *daddr)
239{
240 return (ipv6_addr_type(daddr) &
241 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
242}
243
1da177e4 244/*
c71099ac 245 * Route lookup. Any table->tb6_lock is implied.
1da177e4
LT		 246 */
247
248static __inline__ struct rt6_info *rt6_device_match(struct rt6_info *rt,
249 int oif,
250 int strict)
251{
252 struct rt6_info *local = NULL;
253 struct rt6_info *sprt;
254
255 if (oif) {
7cc48263 256 for (sprt = rt; sprt; sprt = sprt->u.dst.rt6_next) {
1da177e4
LT		 257 struct net_device *dev = sprt->rt6i_dev;
258 if (dev->ifindex == oif)
259 return sprt;
260 if (dev->flags & IFF_LOOPBACK) {
261 if (sprt->rt6i_idev == NULL ||
262 sprt->rt6i_idev->dev->ifindex != oif) {
263 if (strict && oif)
264 continue;
1ab1457c 265 if (local && (!oif ||
1da177e4
LT		 266 local->rt6i_idev->dev->ifindex == oif))
267 continue;
268 }
269 local = sprt;
270 }
271 }
272
273 if (local)
274 return local;
275
276 if (strict)
277 return &ip6_null_entry;
278 }
279 return rt;
280}
281
27097255
YH		 282#ifdef CONFIG_IPV6_ROUTER_PREF
283static void rt6_probe(struct rt6_info *rt)
284{
285 struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
286 /*
287 * Okay, this does not seem to be appropriate
288 * for now, however, we need to check if it
289 * is really so; aka Router Reachability Probing.
290 *
291 * Router Reachability Probe MUST be rate-limited
292 * to no more than one per minute.
293 */
294 if (!neigh || (neigh->nud_state & NUD_VALID))
295 return;
296 read_lock_bh(&neigh->lock);
297 if (!(neigh->nud_state & NUD_VALID) &&
52e16356 298 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
27097255
YH		 299 struct in6_addr mcaddr;
300 struct in6_addr *target;
301
302 neigh->updated = jiffies;
303 read_unlock_bh(&neigh->lock);
304
305 target = (struct in6_addr *)&neigh->primary_key;
306 addrconf_addr_solict_mult(target, &mcaddr);
307 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
308 } else
309 read_unlock_bh(&neigh->lock);
310}
311#else
312static inline void rt6_probe(struct rt6_info *rt)
313{
314 return;
315}
316#endif
317
1da177e4 318/*
554cfb7e 319 * Default Router Selection (RFC 2461 6.3.6)
1da177e4 320 */
b6f99a21 321static inline int rt6_check_dev(struct rt6_info *rt, int oif)
554cfb7e
YH		 322{
323 struct net_device *dev = rt->rt6i_dev;
161980f4 324 if (!oif || dev->ifindex == oif)
554cfb7e 325 return 2;
161980f4
DM		 326 if ((dev->flags & IFF_LOOPBACK) &&
327 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
328 return 1;
329 return 0;
554cfb7e 330}
1da177e4 331
b6f99a21 332static inline int rt6_check_neigh(struct rt6_info *rt)
1da177e4 333{
554cfb7e 334 struct neighbour *neigh = rt->rt6i_nexthop;
398bcbeb 335 int m;
4d0c5911
YH		 336 if (rt->rt6i_flags & RTF_NONEXTHOP ||
337 !(rt->rt6i_flags & RTF_GATEWAY))
338 m = 1;
339 else if (neigh) {
554cfb7e
YH		 340 read_lock_bh(&neigh->lock);
341 if (neigh->nud_state & NUD_VALID)
4d0c5911 342 m = 2;
398bcbeb
YH		 343#ifdef CONFIG_IPV6_ROUTER_PREF
344 else if (neigh->nud_state & NUD_FAILED)
345 m = 0;
346#endif
347 else
ea73ee23 348 m = 1;
554cfb7e 349 read_unlock_bh(&neigh->lock);
398bcbeb
YH		 350 } else
351 m = 0;
554cfb7e 352 return m;
1da177e4
LT		 353}
354
554cfb7e
YH		 355static int rt6_score_route(struct rt6_info *rt, int oif,
356 int strict)
1da177e4 357{
4d0c5911 358 int m, n;
1ab1457c 359
4d0c5911 360 m = rt6_check_dev(rt, oif);
77d16f45 361 if (!m && (strict & RT6_LOOKUP_F_IFACE))
554cfb7e 362 return -1;
ebacaaa0
YH		 363#ifdef CONFIG_IPV6_ROUTER_PREF
364 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
365#endif
4d0c5911 366 n = rt6_check_neigh(rt);
557e92ef 367 if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
554cfb7e
YH		 368 return -1;
369 return m;
370}
371
f11e6659
DM		 372static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
373 int *mpri, struct rt6_info *match)
554cfb7e 374{
f11e6659
DM		 375 int m;
376
377 if (rt6_check_expired(rt))
378 goto out;
379
380 m = rt6_score_route(rt, oif, strict);
381 if (m < 0)
382 goto out;
383
384 if (m > *mpri) {
385 if (strict & RT6_LOOKUP_F_REACHABLE)
386 rt6_probe(match);
387 *mpri = m;
388 match = rt;
389 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
390 rt6_probe(rt);
391 }
392
393out:
394 return match;
395}
396
397static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
398 struct rt6_info *rr_head,
399 u32 metric, int oif, int strict)
400{
401 struct rt6_info *rt, *match;
554cfb7e 402 int mpri = -1;
1da177e4 403
f11e6659
DM		 404 match = NULL;
405 for (rt = rr_head; rt && rt->rt6i_metric == metric;
406 rt = rt->u.dst.rt6_next)
407 match = find_match(rt, oif, strict, &mpri, match);
408 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
409 rt = rt->u.dst.rt6_next)
410 match = find_match(rt, oif, strict, &mpri, match);
1da177e4 411
f11e6659
DM		 412 return match;
413}
1da177e4 414
f11e6659
DM		 415static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
416{
417 struct rt6_info *match, *rt0;
1da177e4 418
f11e6659
DM		 419 RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
420 __FUNCTION__, fn->leaf, oif);
554cfb7e 421
f11e6659
DM		 422 rt0 = fn->rr_ptr;
423 if (!rt0)
424 fn->rr_ptr = rt0 = fn->leaf;
1da177e4 425
f11e6659 426 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
1da177e4 427
554cfb7e 428 if (!match &&
f11e6659
DM		 429 (strict & RT6_LOOKUP_F_REACHABLE)) {
430 struct rt6_info *next = rt0->u.dst.rt6_next;
431
554cfb7e 432 /* no entries matched; do round-robin */
f11e6659
DM		 433 if (!next || next->rt6i_metric != rt0->rt6i_metric)
434 next = fn->leaf;
435
436 if (next != rt0)
437 fn->rr_ptr = next;
1da177e4 438 }
1da177e4 439
f11e6659
DM		 440 RT6_TRACE("%s() => %p\n",
441 __FUNCTION__, match);
1da177e4 442
554cfb7e 443 return (match ? match : &ip6_null_entry);
1da177e4
LT		 444}
445
70ceb4f5
YH		 446#ifdef CONFIG_IPV6_ROUTE_INFO
447int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
448 struct in6_addr *gwaddr)
449{
450 struct route_info *rinfo = (struct route_info *) opt;
451 struct in6_addr prefix_buf, *prefix;
452 unsigned int pref;
453 u32 lifetime;
454 struct rt6_info *rt;
455
456 if (len < sizeof(struct route_info)) {
457 return -EINVAL;
458 }
459
460 /* Sanity check for prefix_len and length */
461 if (rinfo->length > 3) {
462 return -EINVAL;
463 } else if (rinfo->prefix_len > 128) {
464 return -EINVAL;
465 } else if (rinfo->prefix_len > 64) {
466 if (rinfo->length < 2) {
467 return -EINVAL;
468 }
469 } else if (rinfo->prefix_len > 0) {
470 if (rinfo->length < 1) {
471 return -EINVAL;
472 }
473 }
474
475 pref = rinfo->route_pref;
476 if (pref == ICMPV6_ROUTER_PREF_INVALID)
477 pref = ICMPV6_ROUTER_PREF_MEDIUM;
478
e69a4adc 479 lifetime = ntohl(rinfo->lifetime);
70ceb4f5
YH		 480 if (lifetime == 0xffffffff) {
481 /* infinity */
482 } else if (lifetime > 0x7fffffff/HZ) {
483 /* Avoid arithmetic overflow */
484 lifetime = 0x7fffffff/HZ - 1;
485 }
486
487 if (rinfo->length == 3)
488 prefix = (struct in6_addr *)rinfo->prefix;
489 else {
490 /* this function is safe */
491 ipv6_addr_prefix(&prefix_buf,
492 (struct in6_addr *)rinfo->prefix,
493 rinfo->prefix_len);
494 prefix = &prefix_buf;
495 }
496
497 rt = rt6_get_route_info(prefix, rinfo->prefix_len, gwaddr, dev->ifindex);
498
499 if (rt && !lifetime) {
e0a1ad73 500 ip6_del_rt(rt);
70ceb4f5
YH		 501 rt = NULL;
502 }
503
504 if (!rt && lifetime)
505 rt = rt6_add_route_info(prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
506 pref);
507 else if (rt)
508 rt->rt6i_flags = RTF_ROUTEINFO |
509 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
510
511 if (rt) {
512 if (lifetime == 0xffffffff) {
513 rt->rt6i_flags &= ~RTF_EXPIRES;
514 } else {
515 rt->rt6i_expires = jiffies + HZ * lifetime;
516 rt->rt6i_flags |= RTF_EXPIRES;
517 }
518 dst_release(&rt->u.dst);
519 }
520 return 0;
521}
522#endif
523
982f56f3
YH		 524#define BACKTRACK(saddr) \
525do { \
526 if (rt == &ip6_null_entry) { \
527 struct fib6_node *pn; \
e0eda7bb 528 while (1) { \
982f56f3
YH		 529 if (fn->fn_flags & RTN_TL_ROOT) \
530 goto out; \
531 pn = fn->parent; \
532 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
8bce65b9 533 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
982f56f3
YH		 534 else \
535 fn = pn; \
536 if (fn->fn_flags & RTN_RTINFO) \
537 goto restart; \
c71099ac 538 } \
c71099ac 539 } \
982f56f3 540} while(0)
c71099ac
TG		 541
542static struct rt6_info *ip6_pol_route_lookup(struct fib6_table *table,
543 struct flowi *fl, int flags)
1da177e4
LT		 544{
545 struct fib6_node *fn;
546 struct rt6_info *rt;
547
c71099ac
TG		 548 read_lock_bh(&table->tb6_lock);
549 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
550restart:
551 rt = fn->leaf;
77d16f45 552 rt = rt6_device_match(rt, fl->oif, flags);
982f56f3 553 BACKTRACK(&fl->fl6_src);
c71099ac 554out:
03f49f34 555 dst_use(&rt->u.dst, jiffies);
c71099ac 556 read_unlock_bh(&table->tb6_lock);
c71099ac
TG		 557 return rt;
558
559}
560
561struct rt6_info *rt6_lookup(struct in6_addr *daddr, struct in6_addr *saddr,
562 int oif, int strict)
563{
564 struct flowi fl = {
565 .oif = oif,
566 .nl_u = {
567 .ip6_u = {
568 .daddr = *daddr,
c71099ac
TG		 569 },
570 },
571 };
572 struct dst_entry *dst;
77d16f45 573 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
c71099ac 574
adaa70bb
TG		 575 if (saddr) {
576 memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
577 flags |= RT6_LOOKUP_F_HAS_SADDR;
578 }
579
c71099ac
TG		 580 dst = fib6_rule_lookup(&fl, flags, ip6_pol_route_lookup);
581 if (dst->error == 0)
582 return (struct rt6_info *) dst;
583
584 dst_release(dst);
585
1da177e4
LT		 586 return NULL;
587}
588
7159039a
YH		 589EXPORT_SYMBOL(rt6_lookup);
590
c71099ac 591/* ip6_ins_rt is called with FREE table->tb6_lock.
1da177e4
LT		 592 It takes new route entry, the addition fails by any reason the
593 route is freed. In any case, if caller does not hold it, it may
594 be destroyed.
595 */
596
86872cb5 597static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
1da177e4
LT		 598{
599 int err;
c71099ac 600 struct fib6_table *table;
1da177e4 601
c71099ac
TG		 602 table = rt->rt6i_table;
603 write_lock_bh(&table->tb6_lock);
86872cb5 604 err = fib6_add(&table->tb6_root, rt, info);
c71099ac 605 write_unlock_bh(&table->tb6_lock);
1da177e4
LT		 606
607 return err;
608}
609
40e22e8f
TG		 610int ip6_ins_rt(struct rt6_info *rt)
611{
528c4ceb
DL		 612 struct nl_info info = {};
613 return __ip6_ins_rt(rt, &info);
40e22e8f
TG		 614}
615
95a9a5ba
YH		 616static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
617 struct in6_addr *saddr)
1da177e4 618{
1da177e4
LT		 619 struct rt6_info *rt;
620
621 /*
622 * Clone the route.
623 */
624
625 rt = ip6_rt_copy(ort);
626
627 if (rt) {
58c4fb86
YH		 628 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
629 if (rt->rt6i_dst.plen != 128 &&
630 ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
631 rt->rt6i_flags |= RTF_ANYCAST;
1da177e4 632 ipv6_addr_copy(&rt->rt6i_gateway, daddr);
58c4fb86 633 }
1da177e4 634
58c4fb86 635 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
1da177e4
LT		 636 rt->rt6i_dst.plen = 128;
637 rt->rt6i_flags |= RTF_CACHE;
638 rt->u.dst.flags |= DST_HOST;
639
640#ifdef CONFIG_IPV6_SUBTREES
641 if (rt->rt6i_src.plen && saddr) {
642 ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
643 rt->rt6i_src.plen = 128;
644 }
645#endif
646
647 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
648
95a9a5ba 649 }
1da177e4 650
95a9a5ba
YH		 651 return rt;
652}
1da177e4 653
299d9939
YH		 654static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
655{
656 struct rt6_info *rt = ip6_rt_copy(ort);
657 if (rt) {
658 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
659 rt->rt6i_dst.plen = 128;
660 rt->rt6i_flags |= RTF_CACHE;
299d9939
YH		 661 rt->u.dst.flags |= DST_HOST;
662 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
663 }
664 return rt;
665}
666
4acad72d 667static struct rt6_info *ip6_pol_route(struct fib6_table *table, int oif,
8ce11e6a 668 struct flowi *fl, int flags)
1da177e4
LT		 669{
670 struct fib6_node *fn;
519fbd87 671 struct rt6_info *rt, *nrt;
c71099ac 672 int strict = 0;
1da177e4 673 int attempts = 3;
519fbd87 674 int err;
ea659e07 675 int reachable = ipv6_devconf.forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
1da177e4 676
77d16f45 677 strict |= flags & RT6_LOOKUP_F_IFACE;
1da177e4
LT		 678
679relookup:
c71099ac 680 read_lock_bh(&table->tb6_lock);
1da177e4 681
8238dd06 682restart_2:
c71099ac 683 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1da177e4
LT		 684
685restart:
4acad72d 686 rt = rt6_select(fn, oif, strict | reachable);
982f56f3 687 BACKTRACK(&fl->fl6_src);
8238dd06
YH		 688 if (rt == &ip6_null_entry ||
689 rt->rt6i_flags & RTF_CACHE)
1ddef044 690 goto out;
1da177e4 691
fb9de91e 692 dst_hold(&rt->u.dst);
c71099ac 693 read_unlock_bh(&table->tb6_lock);
fb9de91e 694
519fbd87 695 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
c71099ac 696 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
519fbd87
YH		 697 else {
698#if CLONE_OFFLINK_ROUTE
c71099ac 699 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
519fbd87
YH		 700#else
701 goto out2;
702#endif
703 }
e40cf353 704
519fbd87
YH		 705 dst_release(&rt->u.dst);
706 rt = nrt ? : &ip6_null_entry;
1da177e4 707
519fbd87
YH		 708 dst_hold(&rt->u.dst);
709 if (nrt) {
40e22e8f 710 err = ip6_ins_rt(nrt);
519fbd87 711 if (!err)
1da177e4 712 goto out2;
1da177e4 713 }
1da177e4 714
519fbd87
YH		 715 if (--attempts <= 0)
716 goto out2;
717
718 /*
c71099ac 719 * Race condition! In the gap, when table->tb6_lock was
519fbd87
YH		 720 * released someone could insert this route. Relookup.
721 */
722 dst_release(&rt->u.dst);
723 goto relookup;
724
725out:
8238dd06
YH		 726 if (reachable) {
727 reachable = 0;
728 goto restart_2;
729 }
519fbd87 730 dst_hold(&rt->u.dst);
c71099ac 731 read_unlock_bh(&table->tb6_lock);
1da177e4
LT		 732out2:
733 rt->u.dst.lastuse = jiffies;
734 rt->u.dst.__use++;
c71099ac
TG		 735
736 return rt;
1da177e4
LT		 737}
738
4acad72d
PE		 739static struct rt6_info *ip6_pol_route_input(struct fib6_table *table,
740 struct flowi *fl, int flags)
741{
742 return ip6_pol_route(table, fl->iif, fl, flags);
743}
744
c71099ac
TG		 745void ip6_route_input(struct sk_buff *skb)
746{
0660e03f 747 struct ipv6hdr *iph = ipv6_hdr(skb);
adaa70bb 748 int flags = RT6_LOOKUP_F_HAS_SADDR;
c71099ac
TG		 749 struct flowi fl = {
750 .iif = skb->dev->ifindex,
751 .nl_u = {
752 .ip6_u = {
753 .daddr = iph->daddr,
754 .saddr = iph->saddr,
90bcaf7b 755 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
c71099ac
TG		 756 },
757 },
1ab1457c 758 .mark = skb->mark,
c71099ac
TG		 759 .proto = iph->nexthdr,
760 };
adaa70bb
TG		 761
762 if (rt6_need_strict(&iph->daddr))
763 flags |= RT6_LOOKUP_F_IFACE;
c71099ac
TG		 764
765 skb->dst = fib6_rule_lookup(&fl, flags, ip6_pol_route_input);
766}
767
768static struct rt6_info *ip6_pol_route_output(struct fib6_table *table,
769 struct flowi *fl, int flags)
1da177e4 770{
4acad72d 771 return ip6_pol_route(table, fl->oif, fl, flags);
c71099ac
TG		 772}
773
774struct dst_entry * ip6_route_output(struct sock *sk, struct flowi *fl)
775{
776 int flags = 0;
777
778 if (rt6_need_strict(&fl->fl6_dst))
77d16f45 779 flags |= RT6_LOOKUP_F_IFACE;
c71099ac 780
adaa70bb
TG		 781 if (!ipv6_addr_any(&fl->fl6_src))
782 flags |= RT6_LOOKUP_F_HAS_SADDR;
783
c71099ac 784 return fib6_rule_lookup(fl, flags, ip6_pol_route_output);
1da177e4
LT		 785}
786
7159039a 787EXPORT_SYMBOL(ip6_route_output);
1da177e4 788
14e50e57
DM		 789int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
790{
791 struct rt6_info *ort = (struct rt6_info *) *dstp;
792 struct rt6_info *rt = (struct rt6_info *)
793 dst_alloc(&ip6_dst_blackhole_ops);
794 struct dst_entry *new = NULL;
795
796 if (rt) {
797 new = &rt->u.dst;
798
799 atomic_set(&new->__refcnt, 1);
800 new->__use = 1;
352e512c
HX		 801 new->input = dst_discard;
802 new->output = dst_discard;
14e50e57
DM		 803
804 memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
805 new->dev = ort->u.dst.dev;
806 if (new->dev)
807 dev_hold(new->dev);
808 rt->rt6i_idev = ort->rt6i_idev;
809 if (rt->rt6i_idev)
810 in6_dev_hold(rt->rt6i_idev);
811 rt->rt6i_expires = 0;
812
813 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
814 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
815 rt->rt6i_metric = 0;
816
817 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
818#ifdef CONFIG_IPV6_SUBTREES
819 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
820#endif
821
822 dst_free(new);
823 }
824
825 dst_release(*dstp);
826 *dstp = new;
827 return (new ? 0 : -ENOMEM);
828}
829EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
830
1da177e4
LT		 831/*
832 * Destination cache support functions
833 */
834
835static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
836{
837 struct rt6_info *rt;
838
839 rt = (struct rt6_info *) dst;
840
841 if (rt && rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
842 return dst;
843
844 return NULL;
845}
846
847static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
848{
849 struct rt6_info *rt = (struct rt6_info *) dst;
850
851 if (rt) {
852 if (rt->rt6i_flags & RTF_CACHE)
e0a1ad73 853 ip6_del_rt(rt);
1da177e4
LT		 854 else
855 dst_release(dst);
856 }
857 return NULL;
858}
859
860static void ip6_link_failure(struct sk_buff *skb)
861{
862 struct rt6_info *rt;
863
864 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
865
866 rt = (struct rt6_info *) skb->dst;
867 if (rt) {
868 if (rt->rt6i_flags&RTF_CACHE) {
869 dst_set_expires(&rt->u.dst, 0);
870 rt->rt6i_flags |= RTF_EXPIRES;
871 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
872 rt->rt6i_node->fn_sernum = -1;
873 }
874}
875
876static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
877{
878 struct rt6_info *rt6 = (struct rt6_info*)dst;
879
880 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
881 rt6->rt6i_flags |= RTF_MODIFIED;
882 if (mtu < IPV6_MIN_MTU) {
883 mtu = IPV6_MIN_MTU;
884 dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
885 }
886 dst->metrics[RTAX_MTU-1] = mtu;
8d71740c 887 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
1da177e4
LT		 888 }
889}
890
1da177e4
LT		 891static int ipv6_get_mtu(struct net_device *dev);
892
893static inline unsigned int ipv6_advmss(unsigned int mtu)
894{
895 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
896
897 if (mtu < ip6_rt_min_advmss)
898 mtu = ip6_rt_min_advmss;
899
900 /*
1ab1457c
YH		 901 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
902 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
903 * IPV6_MAXPLEN is also valid and means: "any MSS,
1da177e4
LT		 904 * rely only on pmtu discovery"
905 */
906 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
907 mtu = IPV6_MAXPLEN;
908 return mtu;
909}
910
5d0bbeeb 911static struct dst_entry *ndisc_dst_gc_list;
8ce11e6a 912static DEFINE_SPINLOCK(ndisc_lock);
5d0bbeeb 913
1ab1457c 914struct dst_entry *ndisc_dst_alloc(struct net_device *dev,
1da177e4
LT		 915 struct neighbour *neigh,
916 struct in6_addr *addr,
917 int (*output)(struct sk_buff *))
918{
919 struct rt6_info *rt;
920 struct inet6_dev *idev = in6_dev_get(dev);
921
922 if (unlikely(idev == NULL))
923 return NULL;
924
925 rt = ip6_dst_alloc();
926 if (unlikely(rt == NULL)) {
927 in6_dev_put(idev);
928 goto out;
929 }
930
931 dev_hold(dev);
932 if (neigh)
933 neigh_hold(neigh);
934 else
935 neigh = ndisc_get_neigh(dev, addr);
936
937 rt->rt6i_dev = dev;
938 rt->rt6i_idev = idev;
939 rt->rt6i_nexthop = neigh;
940 atomic_set(&rt->u.dst.__refcnt, 1);
941 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
942 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
943 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
944 rt->u.dst.output = output;
945
946#if 0 /* there's no chance to use these for ndisc */
1ab1457c
YH		 947 rt->u.dst.flags = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
948 ? DST_HOST
1da177e4
LT		 949 : 0;
950 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
951 rt->rt6i_dst.plen = 128;
952#endif
953
5d0bbeeb 954 spin_lock_bh(&ndisc_lock);
1da177e4
LT		 955 rt->u.dst.next = ndisc_dst_gc_list;
956 ndisc_dst_gc_list = &rt->u.dst;
5d0bbeeb 957 spin_unlock_bh(&ndisc_lock);
1da177e4
LT		 958
959 fib6_force_start_gc();
960
961out:
40aa7b90 962 return &rt->u.dst;
1da177e4
LT		 963}
964
965int ndisc_dst_gc(int *more)
966{
967 struct dst_entry *dst, *next, **pprev;
968 int freed;
969
970 next = NULL;
1ab1457c 971 freed = 0;
5d0bbeeb
TG		 972
973 spin_lock_bh(&ndisc_lock);
1da177e4 974 pprev = &ndisc_dst_gc_list;
5d0bbeeb 975
1da177e4
LT		 976 while ((dst = *pprev) != NULL) {
977 if (!atomic_read(&dst->__refcnt)) {
978 *pprev = dst->next;
979 dst_free(dst);
980 freed++;
981 } else {
982 pprev = &dst->next;
983 (*more)++;
984 }
985 }
986
5d0bbeeb
TG		 987 spin_unlock_bh(&ndisc_lock);
988
1da177e4
LT		 989 return freed;
990}
991
992static int ip6_dst_gc(void)
993{
994 static unsigned expire = 30*HZ;
995 static unsigned long last_gc;
996 unsigned long now = jiffies;
997
998 if (time_after(last_gc + ip6_rt_gc_min_interval, now) &&
999 atomic_read(&ip6_dst_ops.entries) <= ip6_rt_max_size)
1000 goto out;
1001
1002 expire++;
1003 fib6_run_gc(expire);
1004 last_gc = now;
1005 if (atomic_read(&ip6_dst_ops.entries) < ip6_dst_ops.gc_thresh)
1006 expire = ip6_rt_gc_timeout>>1;
1007
1008out:
1009 expire -= expire>>ip6_rt_gc_elasticity;
1010 return (atomic_read(&ip6_dst_ops.entries) > ip6_rt_max_size);
1011}
1012
1013/* Clean host part of a prefix. Not necessary in radix tree,
1014 but results in cleaner routing tables.
1015
1016 Remove it only when all the things will work!
1017 */
1018
1019static int ipv6_get_mtu(struct net_device *dev)
1020{
1021 int mtu = IPV6_MIN_MTU;
1022 struct inet6_dev *idev;
1023
1024 idev = in6_dev_get(dev);
1025 if (idev) {
1026 mtu = idev->cnf.mtu6;
1027 in6_dev_put(idev);
1028 }
1029 return mtu;
1030}
1031
1032int ipv6_get_hoplimit(struct net_device *dev)
1033{
1034 int hoplimit = ipv6_devconf.hop_limit;
1035 struct inet6_dev *idev;
1036
1037 idev = in6_dev_get(dev);
1038 if (idev) {
1039 hoplimit = idev->cnf.hop_limit;
1040 in6_dev_put(idev);
1041 }
1042 return hoplimit;
1043}
1044
1045/*
1046 *
1047 */
1048
86872cb5 1049int ip6_route_add(struct fib6_config *cfg)
1da177e4
LT		 1050{
1051 int err;
1da177e4
LT		 1052 struct rt6_info *rt = NULL;
1053 struct net_device *dev = NULL;
1054 struct inet6_dev *idev = NULL;
c71099ac 1055 struct fib6_table *table;
1da177e4
LT		 1056 int addr_type;
1057
86872cb5 1058 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1da177e4
LT		 1059 return -EINVAL;
1060#ifndef CONFIG_IPV6_SUBTREES
86872cb5 1061 if (cfg->fc_src_len)
1da177e4
LT		 1062 return -EINVAL;
1063#endif
86872cb5 1064 if (cfg->fc_ifindex) {
1da177e4 1065 err = -ENODEV;
881d966b 1066 dev = dev_get_by_index(&init_net, cfg->fc_ifindex);
1da177e4
LT		 1067 if (!dev)
1068 goto out;
1069 idev = in6_dev_get(dev);
1070 if (!idev)
1071 goto out;
1072 }
1073
86872cb5
TG		 1074 if (cfg->fc_metric == 0)
1075 cfg->fc_metric = IP6_RT_PRIO_USER;
1da177e4 1076
86872cb5 1077 table = fib6_new_table(cfg->fc_table);
c71099ac
TG		 1078 if (table == NULL) {
1079 err = -ENOBUFS;
1080 goto out;
1081 }
1082
1da177e4
LT		 1083 rt = ip6_dst_alloc();
1084
1085 if (rt == NULL) {
1086 err = -ENOMEM;
1087 goto out;
1088 }
1089
1090 rt->u.dst.obsolete = -1;
86872cb5 1091 rt->rt6i_expires = jiffies + clock_t_to_jiffies(cfg->fc_expires);
1da177e4 1092
86872cb5
TG		 1093 if (cfg->fc_protocol == RTPROT_UNSPEC)
1094 cfg->fc_protocol = RTPROT_BOOT;
1095 rt->rt6i_protocol = cfg->fc_protocol;
1096
1097 addr_type = ipv6_addr_type(&cfg->fc_dst);
1da177e4
LT		 1098
1099 if (addr_type & IPV6_ADDR_MULTICAST)
1100 rt->u.dst.input = ip6_mc_input;
1101 else
1102 rt->u.dst.input = ip6_forward;
1103
1104 rt->u.dst.output = ip6_output;
1105
86872cb5
TG		 1106 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1107 rt->rt6i_dst.plen = cfg->fc_dst_len;
1da177e4
LT		 1108 if (rt->rt6i_dst.plen == 128)
1109 rt->u.dst.flags = DST_HOST;
1110
1111#ifdef CONFIG_IPV6_SUBTREES
86872cb5
TG		 1112 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1113 rt->rt6i_src.plen = cfg->fc_src_len;
1da177e4
LT		 1114#endif
1115
86872cb5 1116 rt->rt6i_metric = cfg->fc_metric;
1da177e4
LT		 1117
1118 /* We cannot add true routes via loopback here,
1119 they would result in kernel looping; promote them to reject routes
1120 */
86872cb5 1121 if ((cfg->fc_flags & RTF_REJECT) ||
1da177e4
LT		 1122 (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1123 /* hold loopback dev/idev if we haven't done so. */
2774c7ab 1124 if (dev != init_net.loopback_dev) {
1da177e4
LT		 1125 if (dev) {
1126 dev_put(dev);
1127 in6_dev_put(idev);
1128 }
2774c7ab 1129 dev = init_net.loopback_dev;
1da177e4
LT		 1130 dev_hold(dev);
1131 idev = in6_dev_get(dev);
1132 if (!idev) {
1133 err = -ENODEV;
1134 goto out;
1135 }
1136 }
1137 rt->u.dst.output = ip6_pkt_discard_out;
1138 rt->u.dst.input = ip6_pkt_discard;
1139 rt->u.dst.error = -ENETUNREACH;
1140 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1141 goto install_route;
1142 }
1143
86872cb5 1144 if (cfg->fc_flags & RTF_GATEWAY) {
1da177e4
LT		 1145 struct in6_addr *gw_addr;
1146 int gwa_type;
1147
86872cb5
TG		 1148 gw_addr = &cfg->fc_gateway;
1149 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1da177e4
LT		 1150 gwa_type = ipv6_addr_type(gw_addr);
1151
1152 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1153 struct rt6_info *grt;
1154
1155 /* IPv6 strictly inhibits using not link-local
1156 addresses as nexthop address.
1157 Otherwise, router will not able to send redirects.
1158 It is very good, but in some (rare!) circumstances
1159 (SIT, PtP, NBMA NOARP links) it is handy to allow
1160 some exceptions. --ANK
1161 */
1162 err = -EINVAL;
1163 if (!(gwa_type&IPV6_ADDR_UNICAST))
1164 goto out;
1165
86872cb5 1166 grt = rt6_lookup(gw_addr, NULL, cfg->fc_ifindex, 1);
1da177e4
LT		 1167
1168 err = -EHOSTUNREACH;
1169 if (grt == NULL)
1170 goto out;
1171 if (dev) {
1172 if (dev != grt->rt6i_dev) {
1173 dst_release(&grt->u.dst);
1174 goto out;
1175 }
1176 } else {
1177 dev = grt->rt6i_dev;
1178 idev = grt->rt6i_idev;
1179 dev_hold(dev);
1180 in6_dev_hold(grt->rt6i_idev);
1181 }
1182 if (!(grt->rt6i_flags&RTF_GATEWAY))
1183 err = 0;
1184 dst_release(&grt->u.dst);
1185
1186 if (err)
1187 goto out;
1188 }
1189 err = -EINVAL;
1190 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1191 goto out;
1192 }
1193
1194 err = -ENODEV;
1195 if (dev == NULL)
1196 goto out;
1197
86872cb5 1198 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1da177e4
LT		 1199 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1200 if (IS_ERR(rt->rt6i_nexthop)) {
1201 err = PTR_ERR(rt->rt6i_nexthop);
1202 rt->rt6i_nexthop = NULL;
1203 goto out;
1204 }
1205 }
1206
86872cb5 1207 rt->rt6i_flags = cfg->fc_flags;
1da177e4
LT		 1208
1209install_route:
86872cb5
TG		 1210 if (cfg->fc_mx) {
1211 struct nlattr *nla;
1212 int remaining;
1213
1214 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
8f4c1f9b 1215 int type = nla_type(nla);
86872cb5
TG		 1216
1217 if (type) {
1218 if (type > RTAX_MAX) {
1da177e4
LT		 1219 err = -EINVAL;
1220 goto out;
1221 }
86872cb5
TG		 1222
1223 rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1da177e4 1224 }
1da177e4
LT		 1225 }
1226 }
1227
1228 if (rt->u.dst.metrics[RTAX_HOPLIMIT-1] == 0)
1229 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1230 if (!rt->u.dst.metrics[RTAX_MTU-1])
1231 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1232 if (!rt->u.dst.metrics[RTAX_ADVMSS-1])
1233 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
1234 rt->u.dst.dev = dev;
1235 rt->rt6i_idev = idev;
c71099ac 1236 rt->rt6i_table = table;
86872cb5 1237 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1da177e4
LT		 1238
1239out:
1240 if (dev)
1241 dev_put(dev);
1242 if (idev)
1243 in6_dev_put(idev);
1244 if (rt)
40aa7b90 1245 dst_free(&rt->u.dst);
1da177e4
LT		 1246 return err;
1247}
1248
86872cb5 1249static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1da177e4
LT		 1250{
1251 int err;
c71099ac 1252 struct fib6_table *table;
1da177e4 1253
6c813a72
PM		 1254 if (rt == &ip6_null_entry)
1255 return -ENOENT;
1256
c71099ac
TG		 1257 table = rt->rt6i_table;
1258 write_lock_bh(&table->tb6_lock);
1da177e4 1259
86872cb5 1260 err = fib6_del(rt, info);
1da177e4
LT		 1261 dst_release(&rt->u.dst);
1262
c71099ac 1263 write_unlock_bh(&table->tb6_lock);
1da177e4
LT		 1264
1265 return err;
1266}
1267
e0a1ad73
TG		 1268int ip6_del_rt(struct rt6_info *rt)
1269{
528c4ceb
DL		 1270 struct nl_info info = {};
1271 return __ip6_del_rt(rt, &info);
e0a1ad73
TG		 1272}
1273
86872cb5 1274static int ip6_route_del(struct fib6_config *cfg)
1da177e4 1275{
c71099ac 1276 struct fib6_table *table;
1da177e4
LT		 1277 struct fib6_node *fn;
1278 struct rt6_info *rt;
1279 int err = -ESRCH;
1280
86872cb5 1281 table = fib6_get_table(cfg->fc_table);
c71099ac
TG		 1282 if (table == NULL)
1283 return err;
1284
1285 read_lock_bh(&table->tb6_lock);
1da177e4 1286
c71099ac 1287 fn = fib6_locate(&table->tb6_root,
86872cb5
TG		 1288 &cfg->fc_dst, cfg->fc_dst_len,
1289 &cfg->fc_src, cfg->fc_src_len);
1ab1457c 1290
1da177e4 1291 if (fn) {
7cc48263 1292 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
86872cb5 1293 if (cfg->fc_ifindex &&
1da177e4 1294 (rt->rt6i_dev == NULL ||
86872cb5 1295 rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1da177e4 1296 continue;
86872cb5
TG		 1297 if (cfg->fc_flags & RTF_GATEWAY &&
1298 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1da177e4 1299 continue;
86872cb5 1300 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1da177e4
LT		 1301 continue;
1302 dst_hold(&rt->u.dst);
c71099ac 1303 read_unlock_bh(&table->tb6_lock);
1da177e4 1304
86872cb5 1305 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1da177e4
LT		 1306 }
1307 }
c71099ac 1308 read_unlock_bh(&table->tb6_lock);
1da177e4
LT		 1309
1310 return err;
1311}
1312
1313/*
1314 * Handle redirects
1315 */
a6279458
YH		 1316struct ip6rd_flowi {
1317 struct flowi fl;
1318 struct in6_addr gateway;
1319};
1320
1321static struct rt6_info *__ip6_route_redirect(struct fib6_table *table,
1322 struct flowi *fl,
1323 int flags)
1da177e4 1324{
a6279458
YH		 1325 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1326 struct rt6_info *rt;
e843b9e1 1327 struct fib6_node *fn;
c71099ac 1328
1da177e4 1329 /*
e843b9e1
YH		 1330 * Get the "current" route for this destination and
1331 * check if the redirect has come from approriate router.
1332 *
1333 * RFC 2461 specifies that redirects should only be
1334 * accepted if they come from the nexthop to the target.
1335 * Due to the way the routes are chosen, this notion
1336 * is a bit fuzzy and one might need to check all possible
1337 * routes.
1da177e4 1338 */
1da177e4 1339
c71099ac 1340 read_lock_bh(&table->tb6_lock);
a6279458 1341 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
e843b9e1 1342restart:
7cc48263 1343 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
e843b9e1
YH		 1344 /*
1345 * Current route is on-link; redirect is always invalid.
1346 *
1347 * Seems, previous statement is not true. It could
1348 * be node, which looks for us as on-link (f.e. proxy ndisc)
1349 * But then router serving it might decide, that we should
1350 * know truth 8)8) --ANK (980726).
1351 */
1352 if (rt6_check_expired(rt))
1353 continue;
1354 if (!(rt->rt6i_flags & RTF_GATEWAY))
1355 continue;
a6279458 1356 if (fl->oif != rt->rt6i_dev->ifindex)
e843b9e1 1357 continue;
a6279458 1358 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
e843b9e1
YH		 1359 continue;
1360 break;
1361 }
a6279458 1362
cb15d9c2 1363 if (!rt)
a6279458 1364 rt = &ip6_null_entry;
cb15d9c2
YH		 1365 BACKTRACK(&fl->fl6_src);
1366out:
a6279458
YH		 1367 dst_hold(&rt->u.dst);
1368
c71099ac 1369 read_unlock_bh(&table->tb6_lock);
e843b9e1 1370
a6279458
YH		 1371 return rt;
1372};
1373
1374static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1375 struct in6_addr *src,
1376 struct in6_addr *gateway,
1377 struct net_device *dev)
1378{
adaa70bb 1379 int flags = RT6_LOOKUP_F_HAS_SADDR;
a6279458
YH		 1380 struct ip6rd_flowi rdfl = {
1381 .fl = {
1382 .oif = dev->ifindex,
1383 .nl_u = {
1384 .ip6_u = {
1385 .daddr = *dest,
1386 .saddr = *src,
1387 },
1388 },
1389 },
1390 .gateway = *gateway,
1391 };
adaa70bb
TG		 1392
1393 if (rt6_need_strict(dest))
1394 flags |= RT6_LOOKUP_F_IFACE;
a6279458
YH		 1395
1396 return (struct rt6_info *)fib6_rule_lookup((struct flowi *)&rdfl, flags, __ip6_route_redirect);
1397}
1398
1399void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1400 struct in6_addr *saddr,
1401 struct neighbour *neigh, u8 *lladdr, int on_link)
1402{
1403 struct rt6_info *rt, *nrt = NULL;
1404 struct netevent_redirect netevent;
1405
1406 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1407
1408 if (rt == &ip6_null_entry) {
1da177e4
LT		 1409 if (net_ratelimit())
1410 printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1411 "for redirect target\n");
a6279458 1412 goto out;
1da177e4
LT		 1413 }
1414
1da177e4
LT		 1415 /*
1416 * We have finally decided to accept it.
1417 */
1418
1ab1457c 1419 neigh_update(neigh, lladdr, NUD_STALE,
1da177e4
LT		 1420 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1421 NEIGH_UPDATE_F_OVERRIDE|
1422 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1423 NEIGH_UPDATE_F_ISROUTER))
1424 );
1425
1426 /*
1427 * Redirect received -> path was valid.
1428 * Look, redirects are sent only in response to data packets,
1429 * so that this nexthop apparently is reachable. --ANK
1430 */
1431 dst_confirm(&rt->u.dst);
1432
1433 /* Duplicate redirect: silently ignore. */
1434 if (neigh == rt->u.dst.neighbour)
1435 goto out;
1436
1437 nrt = ip6_rt_copy(rt);
1438 if (nrt == NULL)
1439 goto out;
1440
1441 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1442 if (on_link)
1443 nrt->rt6i_flags &= ~RTF_GATEWAY;
1444
1445 ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1446 nrt->rt6i_dst.plen = 128;
1447 nrt->u.dst.flags |= DST_HOST;
1448
1449 ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1450 nrt->rt6i_nexthop = neigh_clone(neigh);
1451 /* Reset pmtu, it may be better */
1452 nrt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1453 nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&nrt->u.dst));
1454
40e22e8f 1455 if (ip6_ins_rt(nrt))
1da177e4
LT		 1456 goto out;
1457
8d71740c
TT		 1458 netevent.old = &rt->u.dst;
1459 netevent.new = &nrt->u.dst;
1460 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1461
1da177e4 1462 if (rt->rt6i_flags&RTF_CACHE) {
e0a1ad73 1463 ip6_del_rt(rt);
1da177e4
LT		 1464 return;
1465 }
1466
1467out:
1ab1457c 1468 dst_release(&rt->u.dst);
1da177e4
LT		 1469 return;
1470}
1471
1472/*
1473 * Handle ICMP "packet too big" messages
1474 * i.e. Path MTU discovery
1475 */
1476
1477void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1478 struct net_device *dev, u32 pmtu)
1479{
1480 struct rt6_info *rt, *nrt;
1481 int allfrag = 0;
1482
1483 rt = rt6_lookup(daddr, saddr, dev->ifindex, 0);
1484 if (rt == NULL)
1485 return;
1486
1487 if (pmtu >= dst_mtu(&rt->u.dst))
1488 goto out;
1489
1490 if (pmtu < IPV6_MIN_MTU) {
1491 /*
1ab1457c 1492 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1da177e4
LT		 1493 * MTU (1280) and a fragment header should always be included
1494 * after a node receiving Too Big message reporting PMTU is
1495 * less than the IPv6 Minimum Link MTU.
1496 */
1497 pmtu = IPV6_MIN_MTU;
1498 allfrag = 1;
1499 }
1500
1501 /* New mtu received -> path was valid.
1502 They are sent only in response to data packets,
1503 so that this nexthop apparently is reachable. --ANK
1504 */
1505 dst_confirm(&rt->u.dst);
1506
1507 /* Host route. If it is static, it would be better
1508 not to override it, but add new one, so that
1509 when cache entry will expire old pmtu
1510 would return automatically.
1511 */
1512 if (rt->rt6i_flags & RTF_CACHE) {
1513 rt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1514 if (allfrag)
1515 rt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1516 dst_set_expires(&rt->u.dst, ip6_rt_mtu_expires);
1517 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1518 goto out;
1519 }
1520
1521 /* Network route.
1522 Two cases are possible:
1523 1. It is connected route. Action: COW
1524 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1525 */
d5315b50 1526 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
a1e78363 1527 nrt = rt6_alloc_cow(rt, daddr, saddr);
d5315b50
YH		 1528 else
1529 nrt = rt6_alloc_clone(rt, daddr);
a1e78363 1530
d5315b50 1531 if (nrt) {
a1e78363
YH		 1532 nrt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1533 if (allfrag)
1534 nrt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1535
1536 /* According to RFC 1981, detecting PMTU increase shouldn't be
1537 * happened within 5 mins, the recommended timer is 10 mins.
1538 * Here this route expiration time is set to ip6_rt_mtu_expires
1539 * which is 10 mins. After 10 mins the decreased pmtu is expired
1540 * and detecting PMTU increase will be automatically happened.
1541 */
1542 dst_set_expires(&nrt->u.dst, ip6_rt_mtu_expires);
1543 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1544
40e22e8f 1545 ip6_ins_rt(nrt);
1da177e4 1546 }
1da177e4
LT		 1547out:
1548 dst_release(&rt->u.dst);
1549}
1550
1551/*
1552 * Misc support functions
1553 */
1554
1555static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1556{
1557 struct rt6_info *rt = ip6_dst_alloc();
1558
1559 if (rt) {
1560 rt->u.dst.input = ort->u.dst.input;
1561 rt->u.dst.output = ort->u.dst.output;
1562
1563 memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
22e1e4d8 1564 rt->u.dst.error = ort->u.dst.error;
1da177e4
LT		 1565 rt->u.dst.dev = ort->u.dst.dev;
1566 if (rt->u.dst.dev)
1567 dev_hold(rt->u.dst.dev);
1568 rt->rt6i_idev = ort->rt6i_idev;
1569 if (rt->rt6i_idev)
1570 in6_dev_hold(rt->rt6i_idev);
1571 rt->u.dst.lastuse = jiffies;
1572 rt->rt6i_expires = 0;
1573
1574 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1575 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1576 rt->rt6i_metric = 0;
1577
1578 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1579#ifdef CONFIG_IPV6_SUBTREES
1580 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1581#endif
c71099ac 1582 rt->rt6i_table = ort->rt6i_table;
1da177e4
LT		 1583 }
1584 return rt;
1585}
1586
70ceb4f5
YH		 1587#ifdef CONFIG_IPV6_ROUTE_INFO
1588static struct rt6_info *rt6_get_route_info(struct in6_addr *prefix, int prefixlen,
1589 struct in6_addr *gwaddr, int ifindex)
1590{
1591 struct fib6_node *fn;
1592 struct rt6_info *rt = NULL;
c71099ac
TG		 1593 struct fib6_table *table;
1594
1595 table = fib6_get_table(RT6_TABLE_INFO);
1596 if (table == NULL)
1597 return NULL;
70ceb4f5 1598
c71099ac
TG		 1599 write_lock_bh(&table->tb6_lock);
1600 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
70ceb4f5
YH		 1601 if (!fn)
1602 goto out;
1603
7cc48263 1604 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
70ceb4f5
YH		 1605 if (rt->rt6i_dev->ifindex != ifindex)
1606 continue;
1607 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1608 continue;
1609 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1610 continue;
1611 dst_hold(&rt->u.dst);
1612 break;
1613 }
1614out:
c71099ac 1615 write_unlock_bh(&table->tb6_lock);
70ceb4f5
YH		 1616 return rt;
1617}
1618
1619static struct rt6_info *rt6_add_route_info(struct in6_addr *prefix, int prefixlen,
1620 struct in6_addr *gwaddr, int ifindex,
1621 unsigned pref)
1622{
86872cb5
TG		 1623 struct fib6_config cfg = {
1624 .fc_table = RT6_TABLE_INFO,
1625 .fc_metric = 1024,
1626 .fc_ifindex = ifindex,
1627 .fc_dst_len = prefixlen,
1628 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1629 RTF_UP | RTF_PREF(pref),
1630 };
1631
1632 ipv6_addr_copy(&cfg.fc_dst, prefix);
1633 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
70ceb4f5 1634
e317da96
YH		 1635 /* We should treat it as a default route if prefix length is 0. */
1636 if (!prefixlen)
86872cb5 1637 cfg.fc_flags |= RTF_DEFAULT;
70ceb4f5 1638
86872cb5 1639 ip6_route_add(&cfg);
70ceb4f5
YH		 1640
1641 return rt6_get_route_info(prefix, prefixlen, gwaddr, ifindex);
1642}
1643#endif
1644
1da177e4 1645struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1ab1457c 1646{
1da177e4 1647 struct rt6_info *rt;
c71099ac 1648 struct fib6_table *table;
1da177e4 1649
c71099ac
TG		 1650 table = fib6_get_table(RT6_TABLE_DFLT);
1651 if (table == NULL)
1652 return NULL;
1da177e4 1653
c71099ac 1654 write_lock_bh(&table->tb6_lock);
7cc48263 1655 for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
1da177e4 1656 if (dev == rt->rt6i_dev &&
045927ff 1657 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1da177e4
LT		 1658 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1659 break;
1660 }
1661 if (rt)
1662 dst_hold(&rt->u.dst);
c71099ac 1663 write_unlock_bh(&table->tb6_lock);
1da177e4
LT		 1664 return rt;
1665}
1666
c7dc89c0
FT		 1667EXPORT_SYMBOL(rt6_get_dflt_router);
1668
1da177e4 1669struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
ebacaaa0
YH		 1670 struct net_device *dev,
1671 unsigned int pref)
1da177e4 1672{
86872cb5
TG		 1673 struct fib6_config cfg = {
1674 .fc_table = RT6_TABLE_DFLT,
1675 .fc_metric = 1024,
1676 .fc_ifindex = dev->ifindex,
1677 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1678 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1679 };
1da177e4 1680
86872cb5 1681 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1da177e4 1682
86872cb5 1683 ip6_route_add(&cfg);
1da177e4 1684
1da177e4
LT		 1685 return rt6_get_dflt_router(gwaddr, dev);
1686}
1687
1688void rt6_purge_dflt_routers(void)
1689{
1690 struct rt6_info *rt;
c71099ac
TG		 1691 struct fib6_table *table;
1692
1693 /* NOTE: Keep consistent with rt6_get_dflt_router */
1694 table = fib6_get_table(RT6_TABLE_DFLT);
1695 if (table == NULL)
1696 return;
1da177e4
LT		 1697
1698restart:
c71099ac 1699 read_lock_bh(&table->tb6_lock);
7cc48263 1700 for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1da177e4
LT		 1701 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1702 dst_hold(&rt->u.dst);
c71099ac 1703 read_unlock_bh(&table->tb6_lock);
e0a1ad73 1704 ip6_del_rt(rt);
1da177e4
LT		 1705 goto restart;
1706 }
1707 }
c71099ac 1708 read_unlock_bh(&table->tb6_lock);
1da177e4
LT		 1709}
1710
86872cb5
TG		 1711static void rtmsg_to_fib6_config(struct in6_rtmsg *rtmsg,
1712 struct fib6_config *cfg)
1713{
1714 memset(cfg, 0, sizeof(*cfg));
1715
1716 cfg->fc_table = RT6_TABLE_MAIN;
1717 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1718 cfg->fc_metric = rtmsg->rtmsg_metric;
1719 cfg->fc_expires = rtmsg->rtmsg_info;
1720 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1721 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1722 cfg->fc_flags = rtmsg->rtmsg_flags;
1723
1724 ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1725 ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1726 ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1727}
1728
1da177e4
LT		 1729int ipv6_route_ioctl(unsigned int cmd, void __user *arg)
1730{
86872cb5 1731 struct fib6_config cfg;
1da177e4
LT		 1732 struct in6_rtmsg rtmsg;
1733 int err;
1734
1735 switch(cmd) {
1736 case SIOCADDRT: /* Add a route */
1737 case SIOCDELRT: /* Delete a route */
1738 if (!capable(CAP_NET_ADMIN))
1739 return -EPERM;
1740 err = copy_from_user(&rtmsg, arg,
1741 sizeof(struct in6_rtmsg));
1742 if (err)
1743 return -EFAULT;
86872cb5
TG		 1744
1745 rtmsg_to_fib6_config(&rtmsg, &cfg);
1746
1da177e4
LT		 1747 rtnl_lock();
1748 switch (cmd) {
1749 case SIOCADDRT:
86872cb5 1750 err = ip6_route_add(&cfg);
1da177e4
LT		 1751 break;
1752 case SIOCDELRT:
86872cb5 1753 err = ip6_route_del(&cfg);
1da177e4
LT		 1754 break;
1755 default:
1756 err = -EINVAL;
1757 }
1758 rtnl_unlock();
1759
1760 return err;
3ff50b79 1761 }
1da177e4
LT		 1762
1763 return -EINVAL;
1764}
1765
1766/*
1767 * Drop the packet on the floor
1768 */
1769
612f09e8
YH		 1770static inline int ip6_pkt_drop(struct sk_buff *skb, int code,
1771 int ipstats_mib_noroutes)
1da177e4 1772{
612f09e8
YH		 1773 int type;
1774 switch (ipstats_mib_noroutes) {
1775 case IPSTATS_MIB_INNOROUTES:
0660e03f 1776 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
612f09e8
YH		 1777 if (type == IPV6_ADDR_ANY || type == IPV6_ADDR_RESERVED) {
1778 IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_INADDRERRORS);
1779 break;
1780 }
1781 /* FALLTHROUGH */
1782 case IPSTATS_MIB_OUTNOROUTES:
1783 IP6_INC_STATS(ip6_dst_idev(skb->dst), ipstats_mib_noroutes);
1784 break;
1785 }
9ce8ade0 1786 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0, skb->dev);
1da177e4
LT		 1787 kfree_skb(skb);
1788 return 0;
1789}
1790
9ce8ade0
TG		 1791static int ip6_pkt_discard(struct sk_buff *skb)
1792{
612f09e8 1793 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 1794}
1795
20380731 1796static int ip6_pkt_discard_out(struct sk_buff *skb)
1da177e4
LT		 1797{
1798 skb->dev = skb->dst->dev;
612f09e8 1799 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1da177e4
LT		 1800}
1801
6723ab54
DM		 1802#ifdef CONFIG_IPV6_MULTIPLE_TABLES
1803
9ce8ade0
TG		 1804static int ip6_pkt_prohibit(struct sk_buff *skb)
1805{
612f09e8 1806 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 1807}
1808
1809static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1810{
1811 skb->dev = skb->dst->dev;
612f09e8 1812 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
9ce8ade0
TG		 1813}
1814
6723ab54
DM		 1815#endif
1816
1da177e4
LT		 1817/*
1818 * Allocate a dst for local (unicast / anycast) address.
1819 */
1820
1821struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1822 const struct in6_addr *addr,
1823 int anycast)
1824{
1825 struct rt6_info *rt = ip6_dst_alloc();
1826
1827 if (rt == NULL)
1828 return ERR_PTR(-ENOMEM);
1829
2774c7ab 1830 dev_hold(init_net.loopback_dev);
1da177e4
LT		 1831 in6_dev_hold(idev);
1832
1833 rt->u.dst.flags = DST_HOST;
1834 rt->u.dst.input = ip6_input;
1835 rt->u.dst.output = ip6_output;
2774c7ab 1836 rt->rt6i_dev = init_net.loopback_dev;
1da177e4
LT		 1837 rt->rt6i_idev = idev;
1838 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1839 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
1840 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1841 rt->u.dst.obsolete = -1;
1842
1843 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
58c4fb86
YH		 1844 if (anycast)
1845 rt->rt6i_flags |= RTF_ANYCAST;
1846 else
1da177e4
LT		 1847 rt->rt6i_flags |= RTF_LOCAL;
1848 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1849 if (rt->rt6i_nexthop == NULL) {
40aa7b90 1850 dst_free(&rt->u.dst);
1da177e4
LT		 1851 return ERR_PTR(-ENOMEM);
1852 }
1853
1854 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1855 rt->rt6i_dst.plen = 128;
c71099ac 1856 rt->rt6i_table = fib6_get_table(RT6_TABLE_LOCAL);
1da177e4
LT		 1857
1858 atomic_set(&rt->u.dst.__refcnt, 1);
1859
1860 return rt;
1861}
1862
1863static int fib6_ifdown(struct rt6_info *rt, void *arg)
1864{
1865 if (((void*)rt->rt6i_dev == arg || arg == NULL) &&
1866 rt != &ip6_null_entry) {
1867 RT6_TRACE("deleted by ifdown %p\n", rt);
1868 return -1;
1869 }
1870 return 0;
1871}
1872
1873void rt6_ifdown(struct net_device *dev)
1874{
c71099ac 1875 fib6_clean_all(fib6_ifdown, 0, dev);
1da177e4
LT		 1876}
1877
1878struct rt6_mtu_change_arg
1879{
1880 struct net_device *dev;
1881 unsigned mtu;
1882};
1883
1884static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
1885{
1886 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
1887 struct inet6_dev *idev;
1888
1889 /* In IPv6 pmtu discovery is not optional,
1890 so that RTAX_MTU lock cannot disable it.
1891 We still use this lock to block changes
1892 caused by addrconf/ndisc.
1893 */
1894
1895 idev = __in6_dev_get(arg->dev);
1896 if (idev == NULL)
1897 return 0;
1898
1899 /* For administrative MTU increase, there is no way to discover
1900 IPv6 PMTU increase, so PMTU increase should be updated here.
1901 Since RFC 1981 doesn't include administrative MTU increase
1902 update PMTU increase is a MUST. (i.e. jumbo frame)
1903 */
1904 /*
1905 If new MTU is less than route PMTU, this new MTU will be the
1906 lowest MTU in the path, update the route PMTU to reflect PMTU
1907 decreases; if new MTU is greater than route PMTU, and the
1908 old MTU is the lowest MTU in the path, update the route PMTU
1909 to reflect the increase. In this case if the other nodes' MTU
1910 also have the lowest MTU, TOO BIG MESSAGE will be lead to
1911 PMTU discouvery.
1912 */
1913 if (rt->rt6i_dev == arg->dev &&
1914 !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
1ab1457c
YH		 1915 (dst_mtu(&rt->u.dst) > arg->mtu ||
1916 (dst_mtu(&rt->u.dst) < arg->mtu &&
566cfd8f 1917 dst_mtu(&rt->u.dst) == idev->cnf.mtu6))) {
1da177e4 1918 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
566cfd8f
SA		 1919 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(arg->mtu);
1920 }
1da177e4
LT		 1921 return 0;
1922}
1923
1924void rt6_mtu_change(struct net_device *dev, unsigned mtu)
1925{
c71099ac
TG		 1926 struct rt6_mtu_change_arg arg = {
1927 .dev = dev,
1928 .mtu = mtu,
1929 };
1da177e4 1930
c71099ac 1931 fib6_clean_all(rt6_mtu_change_route, 0, &arg);
1da177e4
LT		 1932}
1933
ef7c79ed 1934static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
5176f91e 1935 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
86872cb5 1936 [RTA_OIF] = { .type = NLA_U32 },
ab364a6f 1937 [RTA_IIF] = { .type = NLA_U32 },
86872cb5
TG		 1938 [RTA_PRIORITY] = { .type = NLA_U32 },
1939 [RTA_METRICS] = { .type = NLA_NESTED },
1940};
1941
1942static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
1943 struct fib6_config *cfg)
1da177e4 1944{
86872cb5
TG		 1945 struct rtmsg *rtm;
1946 struct nlattr *tb[RTA_MAX+1];
1947 int err;
1da177e4 1948
86872cb5
TG		 1949 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
1950 if (err < 0)
1951 goto errout;
1da177e4 1952
86872cb5
TG		 1953 err = -EINVAL;
1954 rtm = nlmsg_data(nlh);
1955 memset(cfg, 0, sizeof(*cfg));
1956
1957 cfg->fc_table = rtm->rtm_table;
1958 cfg->fc_dst_len = rtm->rtm_dst_len;
1959 cfg->fc_src_len = rtm->rtm_src_len;
1960 cfg->fc_flags = RTF_UP;
1961 cfg->fc_protocol = rtm->rtm_protocol;
1962
1963 if (rtm->rtm_type == RTN_UNREACHABLE)
1964 cfg->fc_flags |= RTF_REJECT;
1965
1966 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
1967 cfg->fc_nlinfo.nlh = nlh;
1968
1969 if (tb[RTA_GATEWAY]) {
1970 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
1971 cfg->fc_flags |= RTF_GATEWAY;
1da177e4 1972 }
86872cb5
TG		 1973
1974 if (tb[RTA_DST]) {
1975 int plen = (rtm->rtm_dst_len + 7) >> 3;
1976
1977 if (nla_len(tb[RTA_DST]) < plen)
1978 goto errout;
1979
1980 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1da177e4 1981 }
86872cb5
TG		 1982
1983 if (tb[RTA_SRC]) {
1984 int plen = (rtm->rtm_src_len + 7) >> 3;
1985
1986 if (nla_len(tb[RTA_SRC]) < plen)
1987 goto errout;
1988
1989 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1da177e4 1990 }
86872cb5
TG		 1991
1992 if (tb[RTA_OIF])
1993 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
1994
1995 if (tb[RTA_PRIORITY])
1996 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
1997
1998 if (tb[RTA_METRICS]) {
1999 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2000 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
1da177e4 2001 }
86872cb5
TG		 2002
2003 if (tb[RTA_TABLE])
2004 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2005
2006 err = 0;
2007errout:
2008 return err;
1da177e4
LT		 2009}
2010
c127ea2c 2011static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1da177e4 2012{
b854272b 2013 struct net *net = skb->sk->sk_net;
86872cb5
TG		 2014 struct fib6_config cfg;
2015 int err;
1da177e4 2016
b854272b
DL		 2017 if (net != &init_net)
2018 return -EINVAL;
2019
86872cb5
TG		 2020 err = rtm_to_fib6_config(skb, nlh, &cfg);
2021 if (err < 0)
2022 return err;
2023
2024 return ip6_route_del(&cfg);
1da177e4
LT		 2025}
2026
c127ea2c 2027static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1da177e4 2028{
b854272b 2029 struct net *net = skb->sk->sk_net;
86872cb5
TG		 2030 struct fib6_config cfg;
2031 int err;
1da177e4 2032
b854272b
DL		 2033 if (net != &init_net)
2034 return -EINVAL;
2035
86872cb5
TG		 2036 err = rtm_to_fib6_config(skb, nlh, &cfg);
2037 if (err < 0)
2038 return err;
2039
2040 return ip6_route_add(&cfg);
1da177e4
LT		 2041}
2042
339bf98f
TG		 2043static inline size_t rt6_nlmsg_size(void)
2044{
2045 return NLMSG_ALIGN(sizeof(struct rtmsg))
2046 + nla_total_size(16) /* RTA_SRC */
2047 + nla_total_size(16) /* RTA_DST */
2048 + nla_total_size(16) /* RTA_GATEWAY */
2049 + nla_total_size(16) /* RTA_PREFSRC */
2050 + nla_total_size(4) /* RTA_TABLE */
2051 + nla_total_size(4) /* RTA_IIF */
2052 + nla_total_size(4) /* RTA_OIF */
2053 + nla_total_size(4) /* RTA_PRIORITY */
6a2b9ce0 2054 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
339bf98f
TG		 2055 + nla_total_size(sizeof(struct rta_cacheinfo));
2056}
2057
1da177e4 2058static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
0d51aa80
JHS		 2059 struct in6_addr *dst, struct in6_addr *src,
2060 int iif, int type, u32 pid, u32 seq,
2061 int prefix, unsigned int flags)
1da177e4
LT		 2062{
2063 struct rtmsg *rtm;
2d7202bf 2064 struct nlmsghdr *nlh;
e3703b3d 2065 long expires;
9e762a4a 2066 u32 table;
1da177e4
LT		 2067
2068 if (prefix) { /* user wants prefix routes only */
2069 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2070 /* success since this is not a prefix route */
2071 return 1;
2072 }
2073 }
2074
2d7202bf
TG		 2075 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2076 if (nlh == NULL)
26932566 2077 return -EMSGSIZE;
2d7202bf
TG		 2078
2079 rtm = nlmsg_data(nlh);
1da177e4
LT		 2080 rtm->rtm_family = AF_INET6;
2081 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2082 rtm->rtm_src_len = rt->rt6i_src.plen;
2083 rtm->rtm_tos = 0;
c71099ac 2084 if (rt->rt6i_table)
9e762a4a 2085 table = rt->rt6i_table->tb6_id;
c71099ac 2086 else
9e762a4a
PM		 2087 table = RT6_TABLE_UNSPEC;
2088 rtm->rtm_table = table;
2d7202bf 2089 NLA_PUT_U32(skb, RTA_TABLE, table);
1da177e4
LT		 2090 if (rt->rt6i_flags&RTF_REJECT)
2091 rtm->rtm_type = RTN_UNREACHABLE;
2092 else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2093 rtm->rtm_type = RTN_LOCAL;
2094 else
2095 rtm->rtm_type = RTN_UNICAST;
2096 rtm->rtm_flags = 0;
2097 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2098 rtm->rtm_protocol = rt->rt6i_protocol;
2099 if (rt->rt6i_flags&RTF_DYNAMIC)
2100 rtm->rtm_protocol = RTPROT_REDIRECT;
2101 else if (rt->rt6i_flags & RTF_ADDRCONF)
2102 rtm->rtm_protocol = RTPROT_KERNEL;
2103 else if (rt->rt6i_flags&RTF_DEFAULT)
2104 rtm->rtm_protocol = RTPROT_RA;
2105
2106 if (rt->rt6i_flags&RTF_CACHE)
2107 rtm->rtm_flags |= RTM_F_CLONED;
2108
2109 if (dst) {
2d7202bf 2110 NLA_PUT(skb, RTA_DST, 16, dst);
1ab1457c 2111 rtm->rtm_dst_len = 128;
1da177e4 2112 } else if (rtm->rtm_dst_len)
2d7202bf 2113 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
1da177e4
LT		 2114#ifdef CONFIG_IPV6_SUBTREES
2115 if (src) {
2d7202bf 2116 NLA_PUT(skb, RTA_SRC, 16, src);
1ab1457c 2117 rtm->rtm_src_len = 128;
1da177e4 2118 } else if (rtm->rtm_src_len)
2d7202bf 2119 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
1da177e4
LT		 2120#endif
2121 if (iif)
2d7202bf 2122 NLA_PUT_U32(skb, RTA_IIF, iif);
1da177e4
LT		 2123 else if (dst) {
2124 struct in6_addr saddr_buf;
2125 if (ipv6_get_saddr(&rt->u.dst, dst, &saddr_buf) == 0)
2d7202bf 2126 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
1da177e4 2127 }
2d7202bf 2128
1da177e4 2129 if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2d7202bf
TG		 2130 goto nla_put_failure;
2131
1da177e4 2132 if (rt->u.dst.neighbour)
2d7202bf
TG		 2133 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2134
1da177e4 2135 if (rt->u.dst.dev)
2d7202bf
TG		 2136 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2137
2138 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
e3703b3d
TG		 2139
2140 expires = rt->rt6i_expires ? rt->rt6i_expires - jiffies : 0;
2141 if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
2142 expires, rt->u.dst.error) < 0)
2143 goto nla_put_failure;
2d7202bf
TG		 2144
2145 return nlmsg_end(skb, nlh);
2146
2147nla_put_failure:
26932566
PM		 2148 nlmsg_cancel(skb, nlh);
2149 return -EMSGSIZE;
1da177e4
LT		 2150}
2151
1b43af54 2152int rt6_dump_route(struct rt6_info *rt, void *p_arg)
1da177e4
LT		 2153{
2154 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2155 int prefix;
2156
2d7202bf
TG		 2157 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2158 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
1da177e4
LT		 2159 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2160 } else
2161 prefix = 0;
2162
2163 return rt6_fill_node(arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2164 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
0d51aa80 2165 prefix, NLM_F_MULTI);
1da177e4
LT		 2166}
2167
c127ea2c 2168static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
1da177e4 2169{
b854272b 2170 struct net *net = in_skb->sk->sk_net;
ab364a6f
TG		 2171 struct nlattr *tb[RTA_MAX+1];
2172 struct rt6_info *rt;
1da177e4 2173 struct sk_buff *skb;
ab364a6f 2174 struct rtmsg *rtm;
1da177e4 2175 struct flowi fl;
ab364a6f 2176 int err, iif = 0;
1da177e4 2177
b854272b
DL		 2178 if (net != &init_net)
2179 return -EINVAL;
2180
ab364a6f
TG		 2181 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2182 if (err < 0)
2183 goto errout;
1da177e4 2184
ab364a6f 2185 err = -EINVAL;
1da177e4 2186 memset(&fl, 0, sizeof(fl));
1da177e4 2187
ab364a6f
TG		 2188 if (tb[RTA_SRC]) {
2189 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2190 goto errout;
2191
2192 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2193 }
2194
2195 if (tb[RTA_DST]) {
2196 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2197 goto errout;
2198
2199 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2200 }
2201
2202 if (tb[RTA_IIF])
2203 iif = nla_get_u32(tb[RTA_IIF]);
2204
2205 if (tb[RTA_OIF])
2206 fl.oif = nla_get_u32(tb[RTA_OIF]);
1da177e4
LT		 2207
2208 if (iif) {
2209 struct net_device *dev;
881d966b 2210 dev = __dev_get_by_index(&init_net, iif);
1da177e4
LT		 2211 if (!dev) {
2212 err = -ENODEV;
ab364a6f 2213 goto errout;
1da177e4
LT		 2214 }
2215 }
2216
ab364a6f
TG		 2217 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2218 if (skb == NULL) {
2219 err = -ENOBUFS;
2220 goto errout;
2221 }
1da177e4 2222
ab364a6f
TG		 2223 /* Reserve room for dummy headers, this skb can pass
2224 through good chunk of routing engine.
2225 */
459a98ed 2226 skb_reset_mac_header(skb);
ab364a6f 2227 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
1da177e4 2228
ab364a6f 2229 rt = (struct rt6_info*) ip6_route_output(NULL, &fl);
1da177e4
LT		 2230 skb->dst = &rt->u.dst;
2231
ab364a6f 2232 err = rt6_fill_node(skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
1da177e4 2233 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
0d51aa80 2234 nlh->nlmsg_seq, 0, 0);
1da177e4 2235 if (err < 0) {
ab364a6f
TG		 2236 kfree_skb(skb);
2237 goto errout;
1da177e4
LT		 2238 }
2239
97c53cac 2240 err = rtnl_unicast(skb, &init_net, NETLINK_CB(in_skb).pid);
ab364a6f 2241errout:
1da177e4 2242 return err;
1da177e4
LT		 2243}
2244
86872cb5 2245void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
1da177e4
LT		 2246{
2247 struct sk_buff *skb;
528c4ceb
DL		 2248 u32 seq;
2249 int err;
2250
2251 err = -ENOBUFS;
2252 seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
86872cb5 2253
339bf98f 2254 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
21713ebc
TG		 2255 if (skb == NULL)
2256 goto errout;
2257
528c4ceb
DL		 2258 err = rt6_fill_node(skb, rt, NULL, NULL, 0,
2259 event, info->pid, seq, 0, 0);
26932566
PM		 2260 if (err < 0) {
2261 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2262 WARN_ON(err == -EMSGSIZE);
2263 kfree_skb(skb);
2264 goto errout;
2265 }
528c4ceb
DL		 2266 err = rtnl_notify(skb, &init_net, info->pid,
2267 RTNLGRP_IPV6_ROUTE, info->nlh, gfp_any());
21713ebc
TG		 2268errout:
2269 if (err < 0)
97c53cac 2270 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_ROUTE, err);
1da177e4
LT		 2271}
2272
2273/*
2274 * /proc
2275 */
2276
2277#ifdef CONFIG_PROC_FS
2278
2279#define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2280
2281struct rt6_proc_arg
2282{
2283 char *buffer;
2284 int offset;
2285 int length;
2286 int skip;
2287 int len;
2288};
2289
2290static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2291{
33120b30 2292 struct seq_file *m = p_arg;
1da177e4 2293
33120b30
AD		 2294 seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_dst.addr),
2295 rt->rt6i_dst.plen);
1da177e4
LT		 2296
2297#ifdef CONFIG_IPV6_SUBTREES
33120b30
AD		 2298 seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_src.addr),
2299 rt->rt6i_src.plen);
1da177e4 2300#else
33120b30 2301 seq_puts(m, "00000000000000000000000000000000 00 ");
1da177e4
LT		 2302#endif
2303
2304 if (rt->rt6i_nexthop) {
33120b30
AD		 2305 seq_printf(m, NIP6_SEQFMT,
2306 NIP6(*((struct in6_addr *)rt->rt6i_nexthop->primary_key)));
1da177e4 2307 } else {
33120b30 2308 seq_puts(m, "00000000000000000000000000000000");
1da177e4 2309 }
33120b30
AD		 2310 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2311 rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2312 rt->u.dst.__use, rt->rt6i_flags,
2313 rt->rt6i_dev ? rt->rt6i_dev->name : "");
1da177e4
LT		 2314 return 0;
2315}
2316
33120b30 2317static int ipv6_route_show(struct seq_file *m, void *v)
1da177e4 2318{
33120b30
AD		 2319 fib6_clean_all(rt6_info_route, 0, m);
2320 return 0;
2321}
1da177e4 2322
33120b30
AD		 2323static int ipv6_route_open(struct inode *inode, struct file *file)
2324{
2325 return single_open(file, ipv6_route_show, NULL);
1da177e4
LT		 2326}
2327
33120b30
AD		 2328static const struct file_operations ipv6_route_proc_fops = {
2329 .owner = THIS_MODULE,
2330 .open = ipv6_route_open,
2331 .read = seq_read,
2332 .llseek = seq_lseek,
2333 .release = single_release,
2334};
2335
1da177e4
LT		 2336static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2337{
2338 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2339 rt6_stats.fib_nodes, rt6_stats.fib_route_nodes,
2340 rt6_stats.fib_rt_alloc, rt6_stats.fib_rt_entries,
2341 rt6_stats.fib_rt_cache,
2342 atomic_read(&ip6_dst_ops.entries),
2343 rt6_stats.fib_discarded_routes);
2344
2345 return 0;
2346}
2347
2348static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2349{
2350 return single_open(file, rt6_stats_seq_show, NULL);
2351}
2352
9a32144e 2353static const struct file_operations rt6_stats_seq_fops = {
1da177e4
LT		 2354 .owner = THIS_MODULE,
2355 .open = rt6_stats_seq_open,
2356 .read = seq_read,
2357 .llseek = seq_lseek,
2358 .release = single_release,
2359};
75314fb3
DL		 2360
2361static int ipv6_route_proc_init(struct net *net)
2362{
2363 int ret = -ENOMEM;
2364 if (!proc_net_fops_create(net, "ipv6_route",
2365 0, &ipv6_route_proc_fops))
2366 goto out;
2367
2368 if (!proc_net_fops_create(net, "rt6_stats",
2369 S_IRUGO, &rt6_stats_seq_fops))
2370 goto out_ipv6_route;
2371
2372 ret = 0;
2373out:
2374 return ret;
2375out_ipv6_route:
2376 proc_net_remove(net, "ipv6_route");
2377 goto out;
2378}
2379
2380static void ipv6_route_proc_fini(struct net *net)
2381{
2382 proc_net_remove(net, "ipv6_route");
2383 proc_net_remove(net, "rt6_stats");
2384}
2385#else
2386static inline int ipv6_route_proc_init(struct net *net)
2387{
2388 return 0;
2389}
2390static inline void ipv6_route_proc_fini(struct net *net)
2391{
2392 return ;
2393}
1da177e4
LT		 2394#endif /* CONFIG_PROC_FS */
2395
2396#ifdef CONFIG_SYSCTL
2397
2398static int flush_delay;
2399
2400static
2401int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write, struct file * filp,
2402 void __user *buffer, size_t *lenp, loff_t *ppos)
2403{
2404 if (write) {
2405 proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
2406 fib6_run_gc(flush_delay <= 0 ? ~0UL : (unsigned long)flush_delay);
2407 return 0;
2408 } else
2409 return -EINVAL;
2410}
2411
760f2d01 2412ctl_table ipv6_route_table_template[] = {
1ab1457c 2413 {
1da177e4 2414 .procname = "flush",
1ab1457c 2415 .data = &flush_delay,
1da177e4 2416 .maxlen = sizeof(int),
89c8b3a1 2417 .mode = 0200,
1ab1457c 2418 .proc_handler = &ipv6_sysctl_rtcache_flush
1da177e4
LT		 2419 },
2420 {
2421 .ctl_name = NET_IPV6_ROUTE_GC_THRESH,
2422 .procname = "gc_thresh",
1ab1457c 2423 .data = &ip6_dst_ops.gc_thresh,
1da177e4
LT		 2424 .maxlen = sizeof(int),
2425 .mode = 0644,
1ab1457c 2426 .proc_handler = &proc_dointvec,
1da177e4
LT		 2427 },
2428 {
2429 .ctl_name = NET_IPV6_ROUTE_MAX_SIZE,
2430 .procname = "max_size",
1ab1457c 2431 .data = &ip6_rt_max_size,
1da177e4
LT		 2432 .maxlen = sizeof(int),
2433 .mode = 0644,
1ab1457c 2434 .proc_handler = &proc_dointvec,
1da177e4
LT		 2435 },
2436 {
2437 .ctl_name = NET_IPV6_ROUTE_GC_MIN_INTERVAL,
2438 .procname = "gc_min_interval",
1ab1457c 2439 .data = &ip6_rt_gc_min_interval,
1da177e4
LT		 2440 .maxlen = sizeof(int),
2441 .mode = 0644,
1ab1457c 2442 .proc_handler = &proc_dointvec_jiffies,
1da177e4
LT		 2443 .strategy = &sysctl_jiffies,
2444 },
2445 {
2446 .ctl_name = NET_IPV6_ROUTE_GC_TIMEOUT,
2447 .procname = "gc_timeout",
1ab1457c 2448 .data = &ip6_rt_gc_timeout,
1da177e4
LT		 2449 .maxlen = sizeof(int),
2450 .mode = 0644,
1ab1457c 2451 .proc_handler = &proc_dointvec_jiffies,
1da177e4
LT		 2452 .strategy = &sysctl_jiffies,
2453 },
2454 {
2455 .ctl_name = NET_IPV6_ROUTE_GC_INTERVAL,
2456 .procname = "gc_interval",
1ab1457c 2457 .data = &ip6_rt_gc_interval,
1da177e4
LT		 2458 .maxlen = sizeof(int),
2459 .mode = 0644,
1ab1457c 2460 .proc_handler = &proc_dointvec_jiffies,
1da177e4
LT		 2461 .strategy = &sysctl_jiffies,
2462 },
2463 {
2464 .ctl_name = NET_IPV6_ROUTE_GC_ELASTICITY,
2465 .procname = "gc_elasticity",
1ab1457c 2466 .data = &ip6_rt_gc_elasticity,
1da177e4
LT		 2467 .maxlen = sizeof(int),
2468 .mode = 0644,
1ab1457c 2469 .proc_handler = &proc_dointvec_jiffies,
1da177e4
LT		 2470 .strategy = &sysctl_jiffies,
2471 },
2472 {
2473 .ctl_name = NET_IPV6_ROUTE_MTU_EXPIRES,
2474 .procname = "mtu_expires",
1ab1457c 2475 .data = &ip6_rt_mtu_expires,
1da177e4
LT		 2476 .maxlen = sizeof(int),
2477 .mode = 0644,
1ab1457c 2478 .proc_handler = &proc_dointvec_jiffies,
1da177e4
LT		 2479 .strategy = &sysctl_jiffies,
2480 },
2481 {
2482 .ctl_name = NET_IPV6_ROUTE_MIN_ADVMSS,
2483 .procname = "min_adv_mss",
1ab1457c 2484 .data = &ip6_rt_min_advmss,
1da177e4
LT		 2485 .maxlen = sizeof(int),
2486 .mode = 0644,
1ab1457c 2487 .proc_handler = &proc_dointvec_jiffies,
1da177e4
LT		 2488 .strategy = &sysctl_jiffies,
2489 },
2490 {
2491 .ctl_name = NET_IPV6_ROUTE_GC_MIN_INTERVAL_MS,
2492 .procname = "gc_min_interval_ms",
1ab1457c 2493 .data = &ip6_rt_gc_min_interval,
1da177e4
LT		 2494 .maxlen = sizeof(int),
2495 .mode = 0644,
1ab1457c 2496 .proc_handler = &proc_dointvec_ms_jiffies,
1da177e4
LT		 2497 .strategy = &sysctl_ms_jiffies,
2498 },
2499 { .ctl_name = 0 }
2500};
2501
760f2d01
DL		 2502struct ctl_table *ipv6_route_sysctl_init(struct net *net)
2503{
2504 struct ctl_table *table;
2505
2506 table = kmemdup(ipv6_route_table_template,
2507 sizeof(ipv6_route_table_template),
2508 GFP_KERNEL);
2509 return table;
2510}
1da177e4
LT		 2511#endif
2512
433d49c3 2513int __init ip6_route_init(void)
1da177e4 2514{
433d49c3
DL		 2515 int ret;
2516
e5d679f3
AD		 2517 ip6_dst_ops.kmem_cachep =
2518 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
f845ab6b
DL		 2519 SLAB_HWCACHE_ALIGN, NULL);
2520 if (!ip6_dst_ops.kmem_cachep)
2521 return -ENOMEM;
2522
14e50e57
DM		 2523 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops.kmem_cachep;
2524
433d49c3
DL		 2525 ret = fib6_init();
2526 if (ret)
2527 goto out_kmem_cache;
2528
75314fb3
DL		 2529 ret = ipv6_route_proc_init(&init_net);
2530 if (ret)
433d49c3
DL		 2531 goto out_fib6_init;
2532
433d49c3
DL		 2533 ret = xfrm6_init();
2534 if (ret)
75314fb3 2535 goto out_proc_init;
c35b7e72 2536
433d49c3
DL		 2537 ret = fib6_rules_init();
2538 if (ret)
2539 goto xfrm6_init;
7e5449c2 2540
433d49c3
DL		 2541 ret = -ENOBUFS;
2542 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2543 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2544 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2545 goto fib6_rules_init;
c127ea2c 2546
433d49c3
DL		 2547 ret = 0;
2548out:
2549 return ret;
2550
2551fib6_rules_init:
433d49c3
DL		 2552 fib6_rules_cleanup();
2553xfrm6_init:
433d49c3 2554 xfrm6_fini();
75314fb3
DL		 2555out_proc_init:
2556 ipv6_route_proc_fini(&init_net);
433d49c3 2557out_fib6_init:
433d49c3
DL		 2558 rt6_ifdown(NULL);
2559 fib6_gc_cleanup();
2560out_kmem_cache:
2561 kmem_cache_destroy(ip6_dst_ops.kmem_cachep);
2562 goto out;
1da177e4
LT		 2563}
2564
2565void ip6_route_cleanup(void)
2566{
101367c2 2567 fib6_rules_cleanup();
75314fb3 2568 ipv6_route_proc_fini(&init_net);
1da177e4 2569 xfrm6_fini();
1da177e4
LT		 2570 rt6_ifdown(NULL);
2571 fib6_gc_cleanup();
2572 kmem_cache_destroy(ip6_dst_ops.kmem_cachep);
2573}
Clone of davem's net-next-2.6 tree