?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
[NETNS][IPV6]: Make multiple instance of sysctl tables.
[linux-2.6/btrfs-unstable.git] / net / ipv6 / route.c
blob0c7382f4fb85dcac6aeb7ff4a31e9c23d3c621e6
1 /*
2 * Linux INET6 implementation
3 * FIB front-end.
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
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.
25 * Ville Nuorvala
26 * Fixed routing subtrees.
27 */
28
29 #include <linux/capability.h>
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>
40 #include <linux/if_arp.h>
41 #include <linux/proc_fs.h>
42 #include <linux/seq_file.h>
43 #include <net/net_namespace.h>
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>
54 #include <net/netevent.h>
55 #include <net/netlink.h>
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
74 #define CLONE_OFFLINK_ROUTE 0
75
76 static int ip6_rt_max_size = 4096;
77 static int ip6_rt_gc_min_interval = HZ / 2;
78 static int ip6_rt_gc_timeout = 60*HZ;
79 int ip6_rt_gc_interval = 30*HZ;
80 static int ip6_rt_gc_elasticity = 9;
81 static int ip6_rt_mtu_expires = 10*60*HZ;
82 static int ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
83
84 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort);
85 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
86 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
87 static void ip6_dst_destroy(struct dst_entry *);
88 static void ip6_dst_ifdown(struct dst_entry *,
89 struct net_device *dev, int how);
90 static int ip6_dst_gc(void);
91
92 static int ip6_pkt_discard(struct sk_buff *skb);
93 static int ip6_pkt_discard_out(struct sk_buff *skb);
94 static void ip6_link_failure(struct sk_buff *skb);
95 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
96
97 #ifdef CONFIG_IPV6_ROUTE_INFO
98 static struct rt6_info *rt6_add_route_info(struct in6_addr *prefix, int prefixlen,
99 struct in6_addr *gwaddr, int ifindex,
100 unsigned pref);
101 static struct rt6_info *rt6_get_route_info(struct in6_addr *prefix, int prefixlen,
102 struct in6_addr *gwaddr, int ifindex);
103 #endif
104
105 static 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,
116 .local_out = ip6_local_out,
117 .entry_size = sizeof(struct rt6_info),
118 };
119
120 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
121 {
122 }
123
124 static 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
133 struct rt6_info ip6_null_entry = {
134 .u = {
135 .dst = {
136 .__refcnt = ATOMIC_INIT(1),
137 .__use = 1,
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
152 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
153
154 static int ip6_pkt_prohibit(struct sk_buff *skb);
155 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
156
157 struct rt6_info ip6_prohibit_entry = {
158 .u = {
159 .dst = {
160 .__refcnt = ATOMIC_INIT(1),
161 .__use = 1,
162 .obsolete = -1,
163 .error = -EACCES,
164 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
165 .input = ip6_pkt_prohibit,
166 .output = ip6_pkt_prohibit_out,
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
176 struct rt6_info ip6_blk_hole_entry = {
177 .u = {
178 .dst = {
179 .__refcnt = ATOMIC_INIT(1),
180 .__use = 1,
181 .obsolete = -1,
182 .error = -EINVAL,
183 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
184 .input = dst_discard,
185 .output = dst_discard,
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
197 /* allocate dst with ip6_dst_ops */
198 static __inline__ struct rt6_info *ip6_dst_alloc(void)
199 {
200 return (struct rt6_info *)dst_alloc(&ip6_dst_ops);
201 }
202
203 static 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);
211 }
212 }
213
214 static 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;
219 struct net_device *loopback_dev =
220 dev->nd_net->loopback_dev;
221
222 if (dev != loopback_dev && idev != NULL && idev->dev == dev) {
223 struct inet6_dev *loopback_idev =
224 in6_dev_get(loopback_dev);
225 if (loopback_idev != NULL) {
226 rt->rt6i_idev = loopback_idev;
227 in6_dev_put(idev);
228 }
229 }
230 }
231
232 static __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
238 static 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
244 /*
245 * Route lookup. Any table->tb6_lock is implied.
246 */
247
248 static __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) {
256 for (sprt = rt; sprt; sprt = sprt->u.dst.rt6_next) {
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;
265 if (local && (!oif ||
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
282 #ifdef CONFIG_IPV6_ROUTER_PREF
283 static 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) &&
298 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
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
312 static inline void rt6_probe(struct rt6_info *rt)
313 {
314 return;
315 }
316 #endif
317
318 /*
319 * Default Router Selection (RFC 2461 6.3.6)
320 */
321 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
322 {
323 struct net_device *dev = rt->rt6i_dev;
324 if (!oif || dev->ifindex == oif)
325 return 2;
326 if ((dev->flags & IFF_LOOPBACK) &&
327 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
328 return 1;
329 return 0;
330 }
331
332 static inline int rt6_check_neigh(struct rt6_info *rt)
333 {
334 struct neighbour *neigh = rt->rt6i_nexthop;
335 int m;
336 if (rt->rt6i_flags & RTF_NONEXTHOP ||
337 !(rt->rt6i_flags & RTF_GATEWAY))
338 m = 1;
339 else if (neigh) {
340 read_lock_bh(&neigh->lock);
341 if (neigh->nud_state & NUD_VALID)
342 m = 2;
343 #ifdef CONFIG_IPV6_ROUTER_PREF
344 else if (neigh->nud_state & NUD_FAILED)
345 m = 0;
346 #endif
347 else
348 m = 1;
349 read_unlock_bh(&neigh->lock);
350 } else
351 m = 0;
352 return m;
353 }
354
355 static int rt6_score_route(struct rt6_info *rt, int oif,
356 int strict)
357 {
358 int m, n;
359
360 m = rt6_check_dev(rt, oif);
361 if (!m && (strict & RT6_LOOKUP_F_IFACE))
362 return -1;
363 #ifdef CONFIG_IPV6_ROUTER_PREF
364 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
365 #endif
366 n = rt6_check_neigh(rt);
367 if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
368 return -1;
369 return m;
370 }
371
372 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
373 int *mpri, struct rt6_info *match)
374 {
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
393 out:
394 return match;
395 }
396
397 static 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;
402 int mpri = -1;
403
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);
411
412 return match;
413 }
414
415 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
416 {
417 struct rt6_info *match, *rt0;
418
419 RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
420 __FUNCTION__, fn->leaf, oif);
421
422 rt0 = fn->rr_ptr;
423 if (!rt0)
424 fn->rr_ptr = rt0 = fn->leaf;
425
426 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
427
428 if (!match &&
429 (strict & RT6_LOOKUP_F_REACHABLE)) {
430 struct rt6_info *next = rt0->u.dst.rt6_next;
431
432 /* no entries matched; do round-robin */
433 if (!next || next->rt6i_metric != rt0->rt6i_metric)
434 next = fn->leaf;
435
436 if (next != rt0)
437 fn->rr_ptr = next;
438 }
439
440 RT6_TRACE("%s() => %p\n",
441 __FUNCTION__, match);
442
443 return (match ? match : &ip6_null_entry);
444 }
445
446 #ifdef CONFIG_IPV6_ROUTE_INFO
447 int 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
479 lifetime = ntohl(rinfo->lifetime);
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) {
500 ip6_del_rt(rt);
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
524 #define BACKTRACK(saddr) \
525 do { \
526 if (rt == &ip6_null_entry) { \
527 struct fib6_node *pn; \
528 while (1) { \
529 if (fn->fn_flags & RTN_TL_ROOT) \
530 goto out; \
531 pn = fn->parent; \
532 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
533 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
534 else \
535 fn = pn; \
536 if (fn->fn_flags & RTN_RTINFO) \
537 goto restart; \
538 } \
539 } \
540 } while(0)
541
542 static struct rt6_info *ip6_pol_route_lookup(struct fib6_table *table,
543 struct flowi *fl, int flags)
544 {
545 struct fib6_node *fn;
546 struct rt6_info *rt;
547
548 read_lock_bh(&table->tb6_lock);
549 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
550 restart:
551 rt = fn->leaf;
552 rt = rt6_device_match(rt, fl->oif, flags);
553 BACKTRACK(&fl->fl6_src);
554 out:
555 dst_use(&rt->u.dst, jiffies);
556 read_unlock_bh(&table->tb6_lock);
557 return rt;
558
559 }
560
561 struct 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,
569 },
570 },
571 };
572 struct dst_entry *dst;
573 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
574
575 if (saddr) {
576 memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
577 flags |= RT6_LOOKUP_F_HAS_SADDR;
578 }
579
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
586 return NULL;
587 }
588
589 EXPORT_SYMBOL(rt6_lookup);
590
591 /* ip6_ins_rt is called with FREE table->tb6_lock.
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
597 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
598 {
599 int err;
600 struct fib6_table *table;
601
602 table = rt->rt6i_table;
603 write_lock_bh(&table->tb6_lock);
604 err = fib6_add(&table->tb6_root, rt, info);
605 write_unlock_bh(&table->tb6_lock);
606
607 return err;
608 }
609
610 int ip6_ins_rt(struct rt6_info *rt)
611 {
612 struct nl_info info = {};
613 return __ip6_ins_rt(rt, &info);
614 }
615
616 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
617 struct in6_addr *saddr)
618 {
619 struct rt6_info *rt;
620
621 /*
622 * Clone the route.
623 */
624
625 rt = ip6_rt_copy(ort);
626
627 if (rt) {
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;
632 ipv6_addr_copy(&rt->rt6i_gateway, daddr);
633 }
634
635 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
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
649 }
650
651 return rt;
652 }
653
654 static 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;
661 rt->u.dst.flags |= DST_HOST;
662 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
663 }
664 return rt;
665 }
666
667 static struct rt6_info *ip6_pol_route(struct fib6_table *table, int oif,
668 struct flowi *fl, int flags)
669 {
670 struct fib6_node *fn;
671 struct rt6_info *rt, *nrt;
672 int strict = 0;
673 int attempts = 3;
674 int err;
675 int reachable = ipv6_devconf.forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
676
677 strict |= flags & RT6_LOOKUP_F_IFACE;
678
679 relookup:
680 read_lock_bh(&table->tb6_lock);
681
682 restart_2:
683 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
684
685 restart:
686 rt = rt6_select(fn, oif, strict | reachable);
687 BACKTRACK(&fl->fl6_src);
688 if (rt == &ip6_null_entry ||
689 rt->rt6i_flags & RTF_CACHE)
690 goto out;
691
692 dst_hold(&rt->u.dst);
693 read_unlock_bh(&table->tb6_lock);
694
695 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
696 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
697 else {
698 #if CLONE_OFFLINK_ROUTE
699 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
700 #else
701 goto out2;
702 #endif
703 }
704
705 dst_release(&rt->u.dst);
706 rt = nrt ? : &ip6_null_entry;
707
708 dst_hold(&rt->u.dst);
709 if (nrt) {
710 err = ip6_ins_rt(nrt);
711 if (!err)
712 goto out2;
713 }
714
715 if (--attempts <= 0)
716 goto out2;
717
718 /*
719 * Race condition! In the gap, when table->tb6_lock was
720 * released someone could insert this route. Relookup.
721 */
722 dst_release(&rt->u.dst);
723 goto relookup;
724
725 out:
726 if (reachable) {
727 reachable = 0;
728 goto restart_2;
729 }
730 dst_hold(&rt->u.dst);
731 read_unlock_bh(&table->tb6_lock);
732 out2:
733 rt->u.dst.lastuse = jiffies;
734 rt->u.dst.__use++;
735
736 return rt;
737 }
738
739 static 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
745 void ip6_route_input(struct sk_buff *skb)
746 {
747 struct ipv6hdr *iph = ipv6_hdr(skb);
748 int flags = RT6_LOOKUP_F_HAS_SADDR;
749 struct flowi fl = {
750 .iif = skb->dev->ifindex,
751 .nl_u = {
752 .ip6_u = {
753 .daddr = iph->daddr,
754 .saddr = iph->saddr,
755 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
756 },
757 },
758 .mark = skb->mark,
759 .proto = iph->nexthdr,
760 };
761
762 if (rt6_need_strict(&iph->daddr))
763 flags |= RT6_LOOKUP_F_IFACE;
764
765 skb->dst = fib6_rule_lookup(&fl, flags, ip6_pol_route_input);
766 }
767
768 static struct rt6_info *ip6_pol_route_output(struct fib6_table *table,
769 struct flowi *fl, int flags)
770 {
771 return ip6_pol_route(table, fl->oif, fl, flags);
772 }
773
774 struct 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))
779 flags |= RT6_LOOKUP_F_IFACE;
780
781 if (!ipv6_addr_any(&fl->fl6_src))
782 flags |= RT6_LOOKUP_F_HAS_SADDR;
783
784 return fib6_rule_lookup(fl, flags, ip6_pol_route_output);
785 }
786
787 EXPORT_SYMBOL(ip6_route_output);
788
789 int 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;
801 new->input = dst_discard;
802 new->output = dst_discard;
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 }
829 EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
830
831 /*
832 * Destination cache support functions
833 */
834
835 static 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
847 static 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)
853 ip6_del_rt(rt);
854 else
855 dst_release(dst);
856 }
857 return NULL;
858 }
859
860 static 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
876 static 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;
887 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
888 }
889 }
890
891 static int ipv6_get_mtu(struct net_device *dev);
892
893 static 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 /*
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,
904 * rely only on pmtu discovery"
905 */
906 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
907 mtu = IPV6_MAXPLEN;
908 return mtu;
909 }
910
911 static struct dst_entry *ndisc_dst_gc_list;
912 static DEFINE_SPINLOCK(ndisc_lock);
913
914 struct dst_entry *ndisc_dst_alloc(struct net_device *dev,
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 */
947 rt->u.dst.flags = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
948 ? DST_HOST
949 : 0;
950 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
951 rt->rt6i_dst.plen = 128;
952 #endif
953
954 spin_lock_bh(&ndisc_lock);
955 rt->u.dst.next = ndisc_dst_gc_list;
956 ndisc_dst_gc_list = &rt->u.dst;
957 spin_unlock_bh(&ndisc_lock);
958
959 fib6_force_start_gc();
960
961 out:
962 return &rt->u.dst;
963 }
964
965 int ndisc_dst_gc(int *more)
966 {
967 struct dst_entry *dst, *next, **pprev;
968 int freed;
969
970 next = NULL;
971 freed = 0;
972
973 spin_lock_bh(&ndisc_lock);
974 pprev = &ndisc_dst_gc_list;
975
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
987 spin_unlock_bh(&ndisc_lock);
988
989 return freed;
990 }
991
992 static 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
1008 out:
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
1019 static 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
1032 int 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
1049 int ip6_route_add(struct fib6_config *cfg)
1050 {
1051 int err;
1052 struct rt6_info *rt = NULL;
1053 struct net_device *dev = NULL;
1054 struct inet6_dev *idev = NULL;
1055 struct fib6_table *table;
1056 int addr_type;
1057
1058 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1059 return -EINVAL;
1060 #ifndef CONFIG_IPV6_SUBTREES
1061 if (cfg->fc_src_len)
1062 return -EINVAL;
1063 #endif
1064 if (cfg->fc_ifindex) {
1065 err = -ENODEV;
1066 dev = dev_get_by_index(&init_net, cfg->fc_ifindex);
1067 if (!dev)
1068 goto out;
1069 idev = in6_dev_get(dev);
1070 if (!idev)
1071 goto out;
1072 }
1073
1074 if (cfg->fc_metric == 0)
1075 cfg->fc_metric = IP6_RT_PRIO_USER;
1076
1077 table = fib6_new_table(cfg->fc_table);
1078 if (table == NULL) {
1079 err = -ENOBUFS;
1080 goto out;
1081 }
1082
1083 rt = ip6_dst_alloc();
1084
1085 if (rt == NULL) {
1086 err = -ENOMEM;
1087 goto out;
1088 }
1089
1090 rt->u.dst.obsolete = -1;
1091 rt->rt6i_expires = jiffies + clock_t_to_jiffies(cfg->fc_expires);
1092
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);
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
1106 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1107 rt->rt6i_dst.plen = cfg->fc_dst_len;
1108 if (rt->rt6i_dst.plen == 128)
1109 rt->u.dst.flags = DST_HOST;
1110
1111 #ifdef CONFIG_IPV6_SUBTREES
1112 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1113 rt->rt6i_src.plen = cfg->fc_src_len;
1114 #endif
1115
1116 rt->rt6i_metric = cfg->fc_metric;
1117
1118 /* We cannot add true routes via loopback here,
1119 they would result in kernel looping; promote them to reject routes
1120 */
1121 if ((cfg->fc_flags & RTF_REJECT) ||
1122 (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1123 /* hold loopback dev/idev if we haven't done so. */
1124 if (dev != init_net.loopback_dev) {
1125 if (dev) {
1126 dev_put(dev);
1127 in6_dev_put(idev);
1128 }
1129 dev = init_net.loopback_dev;
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
1144 if (cfg->fc_flags & RTF_GATEWAY) {
1145 struct in6_addr *gw_addr;
1146 int gwa_type;
1147
1148 gw_addr = &cfg->fc_gateway;
1149 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
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
1166 grt = rt6_lookup(gw_addr, NULL, cfg->fc_ifindex, 1);
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
1198 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
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
1207 rt->rt6i_flags = cfg->fc_flags;
1208
1209 install_route:
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) {
1215 int type = nla_type(nla);
1216
1217 if (type) {
1218 if (type > RTAX_MAX) {
1219 err = -EINVAL;
1220 goto out;
1221 }
1222
1223 rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1224 }
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;
1236 rt->rt6i_table = table;
1237 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1238
1239 out:
1240 if (dev)
1241 dev_put(dev);
1242 if (idev)
1243 in6_dev_put(idev);
1244 if (rt)
1245 dst_free(&rt->u.dst);
1246 return err;
1247 }
1248
1249 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1250 {
1251 int err;
1252 struct fib6_table *table;
1253
1254 if (rt == &ip6_null_entry)
1255 return -ENOENT;
1256
1257 table = rt->rt6i_table;
1258 write_lock_bh(&table->tb6_lock);
1259
1260 err = fib6_del(rt, info);
1261 dst_release(&rt->u.dst);
1262
1263 write_unlock_bh(&table->tb6_lock);
1264
1265 return err;
1266 }
1267
1268 int ip6_del_rt(struct rt6_info *rt)
1269 {
1270 struct nl_info info = {};
1271 return __ip6_del_rt(rt, &info);
1272 }
1273
1274 static int ip6_route_del(struct fib6_config *cfg)
1275 {
1276 struct fib6_table *table;
1277 struct fib6_node *fn;
1278 struct rt6_info *rt;
1279 int err = -ESRCH;
1280
1281 table = fib6_get_table(cfg->fc_table);
1282 if (table == NULL)
1283 return err;
1284
1285 read_lock_bh(&table->tb6_lock);
1286
1287 fn = fib6_locate(&table->tb6_root,
1288 &cfg->fc_dst, cfg->fc_dst_len,
1289 &cfg->fc_src, cfg->fc_src_len);
1290
1291 if (fn) {
1292 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1293 if (cfg->fc_ifindex &&
1294 (rt->rt6i_dev == NULL ||
1295 rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1296 continue;
1297 if (cfg->fc_flags & RTF_GATEWAY &&
1298 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1299 continue;
1300 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1301 continue;
1302 dst_hold(&rt->u.dst);
1303 read_unlock_bh(&table->tb6_lock);
1304
1305 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1306 }
1307 }
1308 read_unlock_bh(&table->tb6_lock);
1309
1310 return err;
1311 }
1312
1313 /*
1314 * Handle redirects
1315 */
1316 struct ip6rd_flowi {
1317 struct flowi fl;
1318 struct in6_addr gateway;
1319 };
1320
1321 static struct rt6_info *__ip6_route_redirect(struct fib6_table *table,
1322 struct flowi *fl,
1323 int flags)
1324 {
1325 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1326 struct rt6_info *rt;
1327 struct fib6_node *fn;
1328
1329 /*
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.
1338 */
1339
1340 read_lock_bh(&table->tb6_lock);
1341 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1342 restart:
1343 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
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;
1356 if (fl->oif != rt->rt6i_dev->ifindex)
1357 continue;
1358 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1359 continue;
1360 break;
1361 }
1362
1363 if (!rt)
1364 rt = &ip6_null_entry;
1365 BACKTRACK(&fl->fl6_src);
1366 out:
1367 dst_hold(&rt->u.dst);
1368
1369 read_unlock_bh(&table->tb6_lock);
1370
1371 return rt;
1372 };
1373
1374 static 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 {
1379 int flags = RT6_LOOKUP_F_HAS_SADDR;
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 };
1392
1393 if (rt6_need_strict(dest))
1394 flags |= RT6_LOOKUP_F_IFACE;
1395
1396 return (struct rt6_info *)fib6_rule_lookup((struct flowi *)&rdfl, flags, __ip6_route_redirect);
1397 }
1398
1399 void 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) {
1409 if (net_ratelimit())
1410 printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1411 "for redirect target\n");
1412 goto out;
1413 }
1414
1415 /*
1416 * We have finally decided to accept it.
1417 */
1418
1419 neigh_update(neigh, lladdr, NUD_STALE,
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
1455 if (ip6_ins_rt(nrt))
1456 goto out;
1457
1458 netevent.old = &rt->u.dst;
1459 netevent.new = &nrt->u.dst;
1460 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1461
1462 if (rt->rt6i_flags&RTF_CACHE) {
1463 ip6_del_rt(rt);
1464 return;
1465 }
1466
1467 out:
1468 dst_release(&rt->u.dst);
1469 return;
1470 }
1471
1472 /*
1473 * Handle ICMP "packet too big" messages
1474 * i.e. Path MTU discovery
1475 */
1476
1477 void 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 /*
1492 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
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 */
1526 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1527 nrt = rt6_alloc_cow(rt, daddr, saddr);
1528 else
1529 nrt = rt6_alloc_clone(rt, daddr);
1530
1531 if (nrt) {
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
1545 ip6_ins_rt(nrt);
1546 }
1547 out:
1548 dst_release(&rt->u.dst);
1549 }
1550
1551 /*
1552 * Misc support functions
1553 */
1554
1555 static 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));
1564 rt->u.dst.error = ort->u.dst.error;
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
1582 rt->rt6i_table = ort->rt6i_table;
1583 }
1584 return rt;
1585 }
1586
1587 #ifdef CONFIG_IPV6_ROUTE_INFO
1588 static 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;
1593 struct fib6_table *table;
1594
1595 table = fib6_get_table(RT6_TABLE_INFO);
1596 if (table == NULL)
1597 return NULL;
1598
1599 write_lock_bh(&table->tb6_lock);
1600 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1601 if (!fn)
1602 goto out;
1603
1604 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
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 }
1614 out:
1615 write_unlock_bh(&table->tb6_lock);
1616 return rt;
1617 }
1618
1619 static struct rt6_info *rt6_add_route_info(struct in6_addr *prefix, int prefixlen,
1620 struct in6_addr *gwaddr, int ifindex,
1621 unsigned pref)
1622 {
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);
1634
1635 /* We should treat it as a default route if prefix length is 0. */
1636 if (!prefixlen)
1637 cfg.fc_flags |= RTF_DEFAULT;
1638
1639 ip6_route_add(&cfg);
1640
1641 return rt6_get_route_info(prefix, prefixlen, gwaddr, ifindex);
1642 }
1643 #endif
1644
1645 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1646 {
1647 struct rt6_info *rt;
1648 struct fib6_table *table;
1649
1650 table = fib6_get_table(RT6_TABLE_DFLT);
1651 if (table == NULL)
1652 return NULL;
1653
1654 write_lock_bh(&table->tb6_lock);
1655 for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
1656 if (dev == rt->rt6i_dev &&
1657 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1658 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1659 break;
1660 }
1661 if (rt)
1662 dst_hold(&rt->u.dst);
1663 write_unlock_bh(&table->tb6_lock);
1664 return rt;
1665 }
1666
1667 EXPORT_SYMBOL(rt6_get_dflt_router);
1668
1669 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1670 struct net_device *dev,
1671 unsigned int pref)
1672 {
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 };
1680
1681 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1682
1683 ip6_route_add(&cfg);
1684
1685 return rt6_get_dflt_router(gwaddr, dev);
1686 }
1687
1688 void rt6_purge_dflt_routers(void)
1689 {
1690 struct rt6_info *rt;
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;
1697
1698 restart:
1699 read_lock_bh(&table->tb6_lock);
1700 for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1701 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1702 dst_hold(&rt->u.dst);
1703 read_unlock_bh(&table->tb6_lock);
1704 ip6_del_rt(rt);
1705 goto restart;
1706 }
1707 }
1708 read_unlock_bh(&table->tb6_lock);
1709 }
1710
1711 static 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
1729 int ipv6_route_ioctl(unsigned int cmd, void __user *arg)
1730 {
1731 struct fib6_config cfg;
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;
1744
1745 rtmsg_to_fib6_config(&rtmsg, &cfg);
1746
1747 rtnl_lock();
1748 switch (cmd) {
1749 case SIOCADDRT:
1750 err = ip6_route_add(&cfg);
1751 break;
1752 case SIOCDELRT:
1753 err = ip6_route_del(&cfg);
1754 break;
1755 default:
1756 err = -EINVAL;
1757 }
1758 rtnl_unlock();
1759
1760 return err;
1761 }
1762
1763 return -EINVAL;
1764 }
1765
1766 /*
1767 * Drop the packet on the floor
1768 */
1769
1770 static inline int ip6_pkt_drop(struct sk_buff *skb, int code,
1771 int ipstats_mib_noroutes)
1772 {
1773 int type;
1774 switch (ipstats_mib_noroutes) {
1775 case IPSTATS_MIB_INNOROUTES:
1776 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
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 }
1786 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0, skb->dev);
1787 kfree_skb(skb);
1788 return 0;
1789 }
1790
1791 static int ip6_pkt_discard(struct sk_buff *skb)
1792 {
1793 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1794 }
1795
1796 static int ip6_pkt_discard_out(struct sk_buff *skb)
1797 {
1798 skb->dev = skb->dst->dev;
1799 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1800 }
1801
1802 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1803
1804 static int ip6_pkt_prohibit(struct sk_buff *skb)
1805 {
1806 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1807 }
1808
1809 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1810 {
1811 skb->dev = skb->dst->dev;
1812 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1813 }
1814
1815 #endif
1816
1817 /*
1818 * Allocate a dst for local (unicast / anycast) address.
1819 */
1820
1821 struct 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
1830 dev_hold(init_net.loopback_dev);
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;
1836 rt->rt6i_dev = init_net.loopback_dev;
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;
1844 if (anycast)
1845 rt->rt6i_flags |= RTF_ANYCAST;
1846 else
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) {
1850 dst_free(&rt->u.dst);
1851 return ERR_PTR(-ENOMEM);
1852 }
1853
1854 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1855 rt->rt6i_dst.plen = 128;
1856 rt->rt6i_table = fib6_get_table(RT6_TABLE_LOCAL);
1857
1858 atomic_set(&rt->u.dst.__refcnt, 1);
1859
1860 return rt;
1861 }
1862
1863 static 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
1873 void rt6_ifdown(struct net_device *dev)
1874 {
1875 fib6_clean_all(fib6_ifdown, 0, dev);
1876 }
1877
1878 struct rt6_mtu_change_arg
1879 {
1880 struct net_device *dev;
1881 unsigned mtu;
1882 };
1883
1884 static 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) &&
1915 (dst_mtu(&rt->u.dst) > arg->mtu ||
1916 (dst_mtu(&rt->u.dst) < arg->mtu &&
1917 dst_mtu(&rt->u.dst) == idev->cnf.mtu6))) {
1918 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
1919 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(arg->mtu);
1920 }
1921 return 0;
1922 }
1923
1924 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
1925 {
1926 struct rt6_mtu_change_arg arg = {
1927 .dev = dev,
1928 .mtu = mtu,
1929 };
1930
1931 fib6_clean_all(rt6_mtu_change_route, 0, &arg);
1932 }
1933
1934 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
1935 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
1936 [RTA_OIF] = { .type = NLA_U32 },
1937 [RTA_IIF] = { .type = NLA_U32 },
1938 [RTA_PRIORITY] = { .type = NLA_U32 },
1939 [RTA_METRICS] = { .type = NLA_NESTED },
1940 };
1941
1942 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
1943 struct fib6_config *cfg)
1944 {
1945 struct rtmsg *rtm;
1946 struct nlattr *tb[RTA_MAX+1];
1947 int err;
1948
1949 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
1950 if (err < 0)
1951 goto errout;
1952
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;
1972 }
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);
1981 }
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);
1990 }
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]);
2001 }
2002
2003 if (tb[RTA_TABLE])
2004 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2005
2006 err = 0;
2007 errout:
2008 return err;
2009 }
2010
2011 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2012 {
2013 struct net *net = skb->sk->sk_net;
2014 struct fib6_config cfg;
2015 int err;
2016
2017 if (net != &init_net)
2018 return -EINVAL;
2019
2020 err = rtm_to_fib6_config(skb, nlh, &cfg);
2021 if (err < 0)
2022 return err;
2023
2024 return ip6_route_del(&cfg);
2025 }
2026
2027 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2028 {
2029 struct net *net = skb->sk->sk_net;
2030 struct fib6_config cfg;
2031 int err;
2032
2033 if (net != &init_net)
2034 return -EINVAL;
2035
2036 err = rtm_to_fib6_config(skb, nlh, &cfg);
2037 if (err < 0)
2038 return err;
2039
2040 return ip6_route_add(&cfg);
2041 }
2042
2043 static 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 */
2054 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2055 + nla_total_size(sizeof(struct rta_cacheinfo));
2056 }
2057
2058 static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
2059 struct in6_addr *dst, struct in6_addr *src,
2060 int iif, int type, u32 pid, u32 seq,
2061 int prefix, unsigned int flags)
2062 {
2063 struct rtmsg *rtm;
2064 struct nlmsghdr *nlh;
2065 long expires;
2066 u32 table;
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
2075 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2076 if (nlh == NULL)
2077 return -EMSGSIZE;
2078
2079 rtm = nlmsg_data(nlh);
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;
2084 if (rt->rt6i_table)
2085 table = rt->rt6i_table->tb6_id;
2086 else
2087 table = RT6_TABLE_UNSPEC;
2088 rtm->rtm_table = table;
2089 NLA_PUT_U32(skb, RTA_TABLE, table);
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) {
2110 NLA_PUT(skb, RTA_DST, 16, dst);
2111 rtm->rtm_dst_len = 128;
2112 } else if (rtm->rtm_dst_len)
2113 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2114 #ifdef CONFIG_IPV6_SUBTREES
2115 if (src) {
2116 NLA_PUT(skb, RTA_SRC, 16, src);
2117 rtm->rtm_src_len = 128;
2118 } else if (rtm->rtm_src_len)
2119 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2120 #endif
2121 if (iif)
2122 NLA_PUT_U32(skb, RTA_IIF, iif);
2123 else if (dst) {
2124 struct in6_addr saddr_buf;
2125 if (ipv6_get_saddr(&rt->u.dst, dst, &saddr_buf) == 0)
2126 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2127 }
2128
2129 if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2130 goto nla_put_failure;
2131
2132 if (rt->u.dst.neighbour)
2133 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2134
2135 if (rt->u.dst.dev)
2136 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2137
2138 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
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;
2144
2145 return nlmsg_end(skb, nlh);
2146
2147 nla_put_failure:
2148 nlmsg_cancel(skb, nlh);
2149 return -EMSGSIZE;
2150 }
2151
2152 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2153 {
2154 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2155 int prefix;
2156
2157 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2158 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
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,
2165 prefix, NLM_F_MULTI);
2166 }
2167
2168 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2169 {
2170 struct net *net = in_skb->sk->sk_net;
2171 struct nlattr *tb[RTA_MAX+1];
2172 struct rt6_info *rt;
2173 struct sk_buff *skb;
2174 struct rtmsg *rtm;
2175 struct flowi fl;
2176 int err, iif = 0;
2177
2178 if (net != &init_net)
2179 return -EINVAL;
2180
2181 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2182 if (err < 0)
2183 goto errout;
2184
2185 err = -EINVAL;
2186 memset(&fl, 0, sizeof(fl));
2187
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]);
2207
2208 if (iif) {
2209 struct net_device *dev;
2210 dev = __dev_get_by_index(&init_net, iif);
2211 if (!dev) {
2212 err = -ENODEV;
2213 goto errout;
2214 }
2215 }
2216
2217 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2218 if (skb == NULL) {
2219 err = -ENOBUFS;
2220 goto errout;
2221 }
2222
2223 /* Reserve room for dummy headers, this skb can pass
2224 through good chunk of routing engine.
2225 */
2226 skb_reset_mac_header(skb);
2227 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2228
2229 rt = (struct rt6_info*) ip6_route_output(NULL, &fl);
2230 skb->dst = &rt->u.dst;
2231
2232 err = rt6_fill_node(skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2233 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2234 nlh->nlmsg_seq, 0, 0);
2235 if (err < 0) {
2236 kfree_skb(skb);
2237 goto errout;
2238 }
2239
2240 err = rtnl_unicast(skb, &init_net, NETLINK_CB(in_skb).pid);
2241 errout:
2242 return err;
2243 }
2244
2245 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2246 {
2247 struct sk_buff *skb;
2248 u32 seq;
2249 int err;
2250
2251 err = -ENOBUFS;
2252 seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2253
2254 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2255 if (skb == NULL)
2256 goto errout;
2257
2258 err = rt6_fill_node(skb, rt, NULL, NULL, 0,
2259 event, info->pid, seq, 0, 0);
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 }
2266 err = rtnl_notify(skb, &init_net, info->pid,
2267 RTNLGRP_IPV6_ROUTE, info->nlh, gfp_any());
2268 errout:
2269 if (err < 0)
2270 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_ROUTE, err);
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
2281 struct rt6_proc_arg
2282 {
2283 char *buffer;
2284 int offset;
2285 int length;
2286 int skip;
2287 int len;
2288 };
2289
2290 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2291 {
2292 struct seq_file *m = p_arg;
2293
2294 seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_dst.addr),
2295 rt->rt6i_dst.plen);
2296
2297 #ifdef CONFIG_IPV6_SUBTREES
2298 seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_src.addr),
2299 rt->rt6i_src.plen);
2300 #else
2301 seq_puts(m, "00000000000000000000000000000000 00 ");
2302 #endif
2303
2304 if (rt->rt6i_nexthop) {
2305 seq_printf(m, NIP6_SEQFMT,
2306 NIP6(*((struct in6_addr *)rt->rt6i_nexthop->primary_key)));
2307 } else {
2308 seq_puts(m, "00000000000000000000000000000000");
2309 }
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 : "");
2314 return 0;
2315 }
2316
2317 static int ipv6_route_show(struct seq_file *m, void *v)
2318 {
2319 fib6_clean_all(rt6_info_route, 0, m);
2320 return 0;
2321 }
2322
2323 static int ipv6_route_open(struct inode *inode, struct file *file)
2324 {
2325 return single_open(file, ipv6_route_show, NULL);
2326 }
2327
2328 static 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
2336 static 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
2348 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2349 {
2350 return single_open(file, rt6_stats_seq_show, NULL);
2351 }
2352
2353 static const struct file_operations rt6_stats_seq_fops = {
2354 .owner = THIS_MODULE,
2355 .open = rt6_stats_seq_open,
2356 .read = seq_read,
2357 .llseek = seq_lseek,
2358 .release = single_release,
2359 };
2360
2361 static 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;
2373 out:
2374 return ret;
2375 out_ipv6_route:
2376 proc_net_remove(net, "ipv6_route");
2377 goto out;
2378 }
2379
2380 static 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
2386 static inline int ipv6_route_proc_init(struct net *net)
2387 {
2388 return 0;
2389 }
2390 static inline void ipv6_route_proc_fini(struct net *net)
2391 {
2392 return ;
2393 }
2394 #endif /* CONFIG_PROC_FS */
2395
2396 #ifdef CONFIG_SYSCTL
2397
2398 static int flush_delay;
2399
2400 static
2401 int 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
2412 ctl_table ipv6_route_table_template[] = {
2413 {
2414 .procname = "flush",
2415 .data = &flush_delay,
2416 .maxlen = sizeof(int),
2417 .mode = 0200,
2418 .proc_handler = &ipv6_sysctl_rtcache_flush
2419 },
2420 {
2421 .ctl_name = NET_IPV6_ROUTE_GC_THRESH,
2422 .procname = "gc_thresh",
2423 .data = &ip6_dst_ops.gc_thresh,
2424 .maxlen = sizeof(int),
2425 .mode = 0644,
2426 .proc_handler = &proc_dointvec,
2427 },
2428 {
2429 .ctl_name = NET_IPV6_ROUTE_MAX_SIZE,
2430 .procname = "max_size",
2431 .data = &ip6_rt_max_size,
2432 .maxlen = sizeof(int),
2433 .mode = 0644,
2434 .proc_handler = &proc_dointvec,
2435 },
2436 {
2437 .ctl_name = NET_IPV6_ROUTE_GC_MIN_INTERVAL,
2438 .procname = "gc_min_interval",
2439 .data = &ip6_rt_gc_min_interval,
2440 .maxlen = sizeof(int),
2441 .mode = 0644,
2442 .proc_handler = &proc_dointvec_jiffies,
2443 .strategy = &sysctl_jiffies,
2444 },
2445 {
2446 .ctl_name = NET_IPV6_ROUTE_GC_TIMEOUT,
2447 .procname = "gc_timeout",
2448 .data = &ip6_rt_gc_timeout,
2449 .maxlen = sizeof(int),
2450 .mode = 0644,
2451 .proc_handler = &proc_dointvec_jiffies,
2452 .strategy = &sysctl_jiffies,
2453 },
2454 {
2455 .ctl_name = NET_IPV6_ROUTE_GC_INTERVAL,
2456 .procname = "gc_interval",
2457 .data = &ip6_rt_gc_interval,
2458 .maxlen = sizeof(int),
2459 .mode = 0644,
2460 .proc_handler = &proc_dointvec_jiffies,
2461 .strategy = &sysctl_jiffies,
2462 },
2463 {
2464 .ctl_name = NET_IPV6_ROUTE_GC_ELASTICITY,
2465 .procname = "gc_elasticity",
2466 .data = &ip6_rt_gc_elasticity,
2467 .maxlen = sizeof(int),
2468 .mode = 0644,
2469 .proc_handler = &proc_dointvec_jiffies,
2470 .strategy = &sysctl_jiffies,
2471 },
2472 {
2473 .ctl_name = NET_IPV6_ROUTE_MTU_EXPIRES,
2474 .procname = "mtu_expires",
2475 .data = &ip6_rt_mtu_expires,
2476 .maxlen = sizeof(int),
2477 .mode = 0644,
2478 .proc_handler = &proc_dointvec_jiffies,
2479 .strategy = &sysctl_jiffies,
2480 },
2481 {
2482 .ctl_name = NET_IPV6_ROUTE_MIN_ADVMSS,
2483 .procname = "min_adv_mss",
2484 .data = &ip6_rt_min_advmss,
2485 .maxlen = sizeof(int),
2486 .mode = 0644,
2487 .proc_handler = &proc_dointvec_jiffies,
2488 .strategy = &sysctl_jiffies,
2489 },
2490 {
2491 .ctl_name = NET_IPV6_ROUTE_GC_MIN_INTERVAL_MS,
2492 .procname = "gc_min_interval_ms",
2493 .data = &ip6_rt_gc_min_interval,
2494 .maxlen = sizeof(int),
2495 .mode = 0644,
2496 .proc_handler = &proc_dointvec_ms_jiffies,
2497 .strategy = &sysctl_ms_jiffies,
2498 },
2499 { .ctl_name = 0 }
2500 };
2501
2502 struct 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 }
2511 #endif
2512
2513 int __init ip6_route_init(void)
2514 {
2515 int ret;
2516
2517 ip6_dst_ops.kmem_cachep =
2518 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2519 SLAB_HWCACHE_ALIGN, NULL);
2520 if (!ip6_dst_ops.kmem_cachep)
2521 return -ENOMEM;
2522
2523 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops.kmem_cachep;
2524
2525 ret = fib6_init();
2526 if (ret)
2527 goto out_kmem_cache;
2528
2529 ret = ipv6_route_proc_init(&init_net);
2530 if (ret)
2531 goto out_fib6_init;
2532
2533 ret = xfrm6_init();
2534 if (ret)
2535 goto out_proc_init;
2536
2537 ret = fib6_rules_init();
2538 if (ret)
2539 goto xfrm6_init;
2540
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;
2546
2547 ret = 0;
2548 out:
2549 return ret;
2550
2551 fib6_rules_init:
2552 fib6_rules_cleanup();
2553 xfrm6_init:
2554 xfrm6_fini();
2555 out_proc_init:
2556 ipv6_route_proc_fini(&init_net);
2557 out_fib6_init:
2558 rt6_ifdown(NULL);
2559 fib6_gc_cleanup();
2560 out_kmem_cache:
2561 kmem_cache_destroy(ip6_dst_ops.kmem_cachep);
2562 goto out;
2563 }
2564
2565 void ip6_route_cleanup(void)
2566 {
2567 fib6_rules_cleanup();
2568 ipv6_route_proc_fini(&init_net);
2569 xfrm6_fini();
2570 rt6_ifdown(NULL);
2571 fib6_gc_cleanup();
2572 kmem_cache_destroy(ip6_dst_ops.kmem_cachep);
2573 }
(deprecated) Btrfs devel tree