xref: /openbmc/linux/net/ipv6/route.c (revision f3db48517f59133610f558f29de8834d7b007691)
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 struct rt6_info * ip6_rt_copy(struct rt6_info *ort);
77 static struct dst_entry	*ip6_dst_check(struct dst_entry *dst, u32 cookie);
78 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
79 static void		ip6_dst_destroy(struct dst_entry *);
80 static void		ip6_dst_ifdown(struct dst_entry *,
81 				       struct net_device *dev, int how);
82 static int		 ip6_dst_gc(struct dst_ops *ops);
83 
84 static int		ip6_pkt_discard(struct sk_buff *skb);
85 static int		ip6_pkt_discard_out(struct sk_buff *skb);
86 static void		ip6_link_failure(struct sk_buff *skb);
87 static void		ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
88 
89 #ifdef CONFIG_IPV6_ROUTE_INFO
90 static struct rt6_info *rt6_add_route_info(struct in6_addr *prefix, int prefixlen,
91 					   struct in6_addr *gwaddr, int ifindex,
92 					   unsigned pref);
93 static struct rt6_info *rt6_get_route_info(struct in6_addr *prefix, int prefixlen,
94 					   struct in6_addr *gwaddr, int ifindex);
95 #endif
96 
97 static struct dst_ops ip6_dst_ops = {
98 	.family			=	AF_INET6,
99 	.protocol		=	__constant_htons(ETH_P_IPV6),
100 	.gc			=	ip6_dst_gc,
101 	.gc_thresh		=	1024,
102 	.check			=	ip6_dst_check,
103 	.destroy		=	ip6_dst_destroy,
104 	.ifdown			=	ip6_dst_ifdown,
105 	.negative_advice	=	ip6_negative_advice,
106 	.link_failure		=	ip6_link_failure,
107 	.update_pmtu		=	ip6_rt_update_pmtu,
108 	.local_out		=	ip6_local_out,
109 	.entry_size		=	sizeof(struct rt6_info),
110 	.entries		=	ATOMIC_INIT(0),
111 };
112 
113 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
114 {
115 }
116 
117 static struct dst_ops ip6_dst_blackhole_ops = {
118 	.family			=	AF_INET6,
119 	.protocol		=	__constant_htons(ETH_P_IPV6),
120 	.destroy		=	ip6_dst_destroy,
121 	.check			=	ip6_dst_check,
122 	.update_pmtu		=	ip6_rt_blackhole_update_pmtu,
123 	.entry_size		=	sizeof(struct rt6_info),
124 	.entries		=	ATOMIC_INIT(0),
125 };
126 
127 struct rt6_info ip6_null_entry = {
128 	.u = {
129 		.dst = {
130 			.__refcnt	= ATOMIC_INIT(1),
131 			.__use		= 1,
132 			.obsolete	= -1,
133 			.error		= -ENETUNREACH,
134 			.metrics	= { [RTAX_HOPLIMIT - 1] = 255, },
135 			.input		= ip6_pkt_discard,
136 			.output		= ip6_pkt_discard_out,
137 			.ops		= &ip6_dst_ops,
138 			.path		= (struct dst_entry*)&ip6_null_entry,
139 		}
140 	},
141 	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
142 	.rt6i_metric	= ~(u32) 0,
143 	.rt6i_ref	= ATOMIC_INIT(1),
144 };
145 
146 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
147 
148 static int ip6_pkt_prohibit(struct sk_buff *skb);
149 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
150 
151 struct rt6_info ip6_prohibit_entry = {
152 	.u = {
153 		.dst = {
154 			.__refcnt	= ATOMIC_INIT(1),
155 			.__use		= 1,
156 			.obsolete	= -1,
157 			.error		= -EACCES,
158 			.metrics	= { [RTAX_HOPLIMIT - 1] = 255, },
159 			.input		= ip6_pkt_prohibit,
160 			.output		= ip6_pkt_prohibit_out,
161 			.ops		= &ip6_dst_ops,
162 			.path		= (struct dst_entry*)&ip6_prohibit_entry,
163 		}
164 	},
165 	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
166 	.rt6i_metric	= ~(u32) 0,
167 	.rt6i_ref	= ATOMIC_INIT(1),
168 };
169 
170 struct rt6_info ip6_blk_hole_entry = {
171 	.u = {
172 		.dst = {
173 			.__refcnt	= ATOMIC_INIT(1),
174 			.__use		= 1,
175 			.obsolete	= -1,
176 			.error		= -EINVAL,
177 			.metrics	= { [RTAX_HOPLIMIT - 1] = 255, },
178 			.input		= dst_discard,
179 			.output		= dst_discard,
180 			.ops		= &ip6_dst_ops,
181 			.path		= (struct dst_entry*)&ip6_blk_hole_entry,
182 		}
183 	},
184 	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
185 	.rt6i_metric	= ~(u32) 0,
186 	.rt6i_ref	= ATOMIC_INIT(1),
187 };
188 
189 #endif
190 
191 /* allocate dst with ip6_dst_ops */
192 static __inline__ struct rt6_info *ip6_dst_alloc(void)
193 {
194 	return (struct rt6_info *)dst_alloc(&ip6_dst_ops);
195 }
196 
197 static void ip6_dst_destroy(struct dst_entry *dst)
198 {
199 	struct rt6_info *rt = (struct rt6_info *)dst;
200 	struct inet6_dev *idev = rt->rt6i_idev;
201 
202 	if (idev != NULL) {
203 		rt->rt6i_idev = NULL;
204 		in6_dev_put(idev);
205 	}
206 }
207 
208 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
209 			   int how)
210 {
211 	struct rt6_info *rt = (struct rt6_info *)dst;
212 	struct inet6_dev *idev = rt->rt6i_idev;
213 	struct net_device *loopback_dev =
214 		dev->nd_net->loopback_dev;
215 
216 	if (dev != loopback_dev && idev != NULL && idev->dev == dev) {
217 		struct inet6_dev *loopback_idev =
218 			in6_dev_get(loopback_dev);
219 		if (loopback_idev != NULL) {
220 			rt->rt6i_idev = loopback_idev;
221 			in6_dev_put(idev);
222 		}
223 	}
224 }
225 
226 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
227 {
228 	return (rt->rt6i_flags & RTF_EXPIRES &&
229 		time_after(jiffies, rt->rt6i_expires));
230 }
231 
232 static inline int rt6_need_strict(struct in6_addr *daddr)
233 {
234 	return (ipv6_addr_type(daddr) &
235 		(IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
236 }
237 
238 /*
239  *	Route lookup. Any table->tb6_lock is implied.
240  */
241 
242 static __inline__ struct rt6_info *rt6_device_match(struct rt6_info *rt,
243 						    int oif,
244 						    int strict)
245 {
246 	struct rt6_info *local = NULL;
247 	struct rt6_info *sprt;
248 
249 	if (oif) {
250 		for (sprt = rt; sprt; sprt = sprt->u.dst.rt6_next) {
251 			struct net_device *dev = sprt->rt6i_dev;
252 			if (dev->ifindex == oif)
253 				return sprt;
254 			if (dev->flags & IFF_LOOPBACK) {
255 				if (sprt->rt6i_idev == NULL ||
256 				    sprt->rt6i_idev->dev->ifindex != oif) {
257 					if (strict && oif)
258 						continue;
259 					if (local && (!oif ||
260 						      local->rt6i_idev->dev->ifindex == oif))
261 						continue;
262 				}
263 				local = sprt;
264 			}
265 		}
266 
267 		if (local)
268 			return local;
269 
270 		if (strict)
271 			return &ip6_null_entry;
272 	}
273 	return rt;
274 }
275 
276 #ifdef CONFIG_IPV6_ROUTER_PREF
277 static void rt6_probe(struct rt6_info *rt)
278 {
279 	struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
280 	/*
281 	 * Okay, this does not seem to be appropriate
282 	 * for now, however, we need to check if it
283 	 * is really so; aka Router Reachability Probing.
284 	 *
285 	 * Router Reachability Probe MUST be rate-limited
286 	 * to no more than one per minute.
287 	 */
288 	if (!neigh || (neigh->nud_state & NUD_VALID))
289 		return;
290 	read_lock_bh(&neigh->lock);
291 	if (!(neigh->nud_state & NUD_VALID) &&
292 	    time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
293 		struct in6_addr mcaddr;
294 		struct in6_addr *target;
295 
296 		neigh->updated = jiffies;
297 		read_unlock_bh(&neigh->lock);
298 
299 		target = (struct in6_addr *)&neigh->primary_key;
300 		addrconf_addr_solict_mult(target, &mcaddr);
301 		ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
302 	} else
303 		read_unlock_bh(&neigh->lock);
304 }
305 #else
306 static inline void rt6_probe(struct rt6_info *rt)
307 {
308 	return;
309 }
310 #endif
311 
312 /*
313  * Default Router Selection (RFC 2461 6.3.6)
314  */
315 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
316 {
317 	struct net_device *dev = rt->rt6i_dev;
318 	if (!oif || dev->ifindex == oif)
319 		return 2;
320 	if ((dev->flags & IFF_LOOPBACK) &&
321 	    rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
322 		return 1;
323 	return 0;
324 }
325 
326 static inline int rt6_check_neigh(struct rt6_info *rt)
327 {
328 	struct neighbour *neigh = rt->rt6i_nexthop;
329 	int m;
330 	if (rt->rt6i_flags & RTF_NONEXTHOP ||
331 	    !(rt->rt6i_flags & RTF_GATEWAY))
332 		m = 1;
333 	else if (neigh) {
334 		read_lock_bh(&neigh->lock);
335 		if (neigh->nud_state & NUD_VALID)
336 			m = 2;
337 #ifdef CONFIG_IPV6_ROUTER_PREF
338 		else if (neigh->nud_state & NUD_FAILED)
339 			m = 0;
340 #endif
341 		else
342 			m = 1;
343 		read_unlock_bh(&neigh->lock);
344 	} else
345 		m = 0;
346 	return m;
347 }
348 
349 static int rt6_score_route(struct rt6_info *rt, int oif,
350 			   int strict)
351 {
352 	int m, n;
353 
354 	m = rt6_check_dev(rt, oif);
355 	if (!m && (strict & RT6_LOOKUP_F_IFACE))
356 		return -1;
357 #ifdef CONFIG_IPV6_ROUTER_PREF
358 	m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
359 #endif
360 	n = rt6_check_neigh(rt);
361 	if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
362 		return -1;
363 	return m;
364 }
365 
366 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
367 				   int *mpri, struct rt6_info *match)
368 {
369 	int m;
370 
371 	if (rt6_check_expired(rt))
372 		goto out;
373 
374 	m = rt6_score_route(rt, oif, strict);
375 	if (m < 0)
376 		goto out;
377 
378 	if (m > *mpri) {
379 		if (strict & RT6_LOOKUP_F_REACHABLE)
380 			rt6_probe(match);
381 		*mpri = m;
382 		match = rt;
383 	} else if (strict & RT6_LOOKUP_F_REACHABLE) {
384 		rt6_probe(rt);
385 	}
386 
387 out:
388 	return match;
389 }
390 
391 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
392 				     struct rt6_info *rr_head,
393 				     u32 metric, int oif, int strict)
394 {
395 	struct rt6_info *rt, *match;
396 	int mpri = -1;
397 
398 	match = NULL;
399 	for (rt = rr_head; rt && rt->rt6i_metric == metric;
400 	     rt = rt->u.dst.rt6_next)
401 		match = find_match(rt, oif, strict, &mpri, match);
402 	for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
403 	     rt = rt->u.dst.rt6_next)
404 		match = find_match(rt, oif, strict, &mpri, match);
405 
406 	return match;
407 }
408 
409 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
410 {
411 	struct rt6_info *match, *rt0;
412 
413 	RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
414 		  __FUNCTION__, fn->leaf, oif);
415 
416 	rt0 = fn->rr_ptr;
417 	if (!rt0)
418 		fn->rr_ptr = rt0 = fn->leaf;
419 
420 	match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
421 
422 	if (!match &&
423 	    (strict & RT6_LOOKUP_F_REACHABLE)) {
424 		struct rt6_info *next = rt0->u.dst.rt6_next;
425 
426 		/* no entries matched; do round-robin */
427 		if (!next || next->rt6i_metric != rt0->rt6i_metric)
428 			next = fn->leaf;
429 
430 		if (next != rt0)
431 			fn->rr_ptr = next;
432 	}
433 
434 	RT6_TRACE("%s() => %p\n",
435 		  __FUNCTION__, match);
436 
437 	return (match ? match : &ip6_null_entry);
438 }
439 
440 #ifdef CONFIG_IPV6_ROUTE_INFO
441 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
442 		  struct in6_addr *gwaddr)
443 {
444 	struct route_info *rinfo = (struct route_info *) opt;
445 	struct in6_addr prefix_buf, *prefix;
446 	unsigned int pref;
447 	u32 lifetime;
448 	struct rt6_info *rt;
449 
450 	if (len < sizeof(struct route_info)) {
451 		return -EINVAL;
452 	}
453 
454 	/* Sanity check for prefix_len and length */
455 	if (rinfo->length > 3) {
456 		return -EINVAL;
457 	} else if (rinfo->prefix_len > 128) {
458 		return -EINVAL;
459 	} else if (rinfo->prefix_len > 64) {
460 		if (rinfo->length < 2) {
461 			return -EINVAL;
462 		}
463 	} else if (rinfo->prefix_len > 0) {
464 		if (rinfo->length < 1) {
465 			return -EINVAL;
466 		}
467 	}
468 
469 	pref = rinfo->route_pref;
470 	if (pref == ICMPV6_ROUTER_PREF_INVALID)
471 		pref = ICMPV6_ROUTER_PREF_MEDIUM;
472 
473 	lifetime = ntohl(rinfo->lifetime);
474 	if (lifetime == 0xffffffff) {
475 		/* infinity */
476 	} else if (lifetime > 0x7fffffff/HZ) {
477 		/* Avoid arithmetic overflow */
478 		lifetime = 0x7fffffff/HZ - 1;
479 	}
480 
481 	if (rinfo->length == 3)
482 		prefix = (struct in6_addr *)rinfo->prefix;
483 	else {
484 		/* this function is safe */
485 		ipv6_addr_prefix(&prefix_buf,
486 				 (struct in6_addr *)rinfo->prefix,
487 				 rinfo->prefix_len);
488 		prefix = &prefix_buf;
489 	}
490 
491 	rt = rt6_get_route_info(prefix, rinfo->prefix_len, gwaddr, dev->ifindex);
492 
493 	if (rt && !lifetime) {
494 		ip6_del_rt(rt);
495 		rt = NULL;
496 	}
497 
498 	if (!rt && lifetime)
499 		rt = rt6_add_route_info(prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
500 					pref);
501 	else if (rt)
502 		rt->rt6i_flags = RTF_ROUTEINFO |
503 				 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
504 
505 	if (rt) {
506 		if (lifetime == 0xffffffff) {
507 			rt->rt6i_flags &= ~RTF_EXPIRES;
508 		} else {
509 			rt->rt6i_expires = jiffies + HZ * lifetime;
510 			rt->rt6i_flags |= RTF_EXPIRES;
511 		}
512 		dst_release(&rt->u.dst);
513 	}
514 	return 0;
515 }
516 #endif
517 
518 #define BACKTRACK(saddr) \
519 do { \
520 	if (rt == &ip6_null_entry) { \
521 		struct fib6_node *pn; \
522 		while (1) { \
523 			if (fn->fn_flags & RTN_TL_ROOT) \
524 				goto out; \
525 			pn = fn->parent; \
526 			if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
527 				fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
528 			else \
529 				fn = pn; \
530 			if (fn->fn_flags & RTN_RTINFO) \
531 				goto restart; \
532 		} \
533 	} \
534 } while(0)
535 
536 static struct rt6_info *ip6_pol_route_lookup(struct fib6_table *table,
537 					     struct flowi *fl, int flags)
538 {
539 	struct fib6_node *fn;
540 	struct rt6_info *rt;
541 
542 	read_lock_bh(&table->tb6_lock);
543 	fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
544 restart:
545 	rt = fn->leaf;
546 	rt = rt6_device_match(rt, fl->oif, flags);
547 	BACKTRACK(&fl->fl6_src);
548 out:
549 	dst_use(&rt->u.dst, jiffies);
550 	read_unlock_bh(&table->tb6_lock);
551 	return rt;
552 
553 }
554 
555 struct rt6_info *rt6_lookup(struct in6_addr *daddr, struct in6_addr *saddr,
556 			    int oif, int strict)
557 {
558 	struct flowi fl = {
559 		.oif = oif,
560 		.nl_u = {
561 			.ip6_u = {
562 				.daddr = *daddr,
563 			},
564 		},
565 	};
566 	struct dst_entry *dst;
567 	int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
568 
569 	if (saddr) {
570 		memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
571 		flags |= RT6_LOOKUP_F_HAS_SADDR;
572 	}
573 
574 	dst = fib6_rule_lookup(&init_net, &fl, flags, ip6_pol_route_lookup);
575 	if (dst->error == 0)
576 		return (struct rt6_info *) dst;
577 
578 	dst_release(dst);
579 
580 	return NULL;
581 }
582 
583 EXPORT_SYMBOL(rt6_lookup);
584 
585 /* ip6_ins_rt is called with FREE table->tb6_lock.
586    It takes new route entry, the addition fails by any reason the
587    route is freed. In any case, if caller does not hold it, it may
588    be destroyed.
589  */
590 
591 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
592 {
593 	int err;
594 	struct fib6_table *table;
595 
596 	table = rt->rt6i_table;
597 	write_lock_bh(&table->tb6_lock);
598 	err = fib6_add(&table->tb6_root, rt, info);
599 	write_unlock_bh(&table->tb6_lock);
600 
601 	return err;
602 }
603 
604 int ip6_ins_rt(struct rt6_info *rt)
605 {
606 	struct nl_info info = {
607 		.nl_net = &init_net,
608 	};
609 	return __ip6_ins_rt(rt, &info);
610 }
611 
612 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
613 				      struct in6_addr *saddr)
614 {
615 	struct rt6_info *rt;
616 
617 	/*
618 	 *	Clone the route.
619 	 */
620 
621 	rt = ip6_rt_copy(ort);
622 
623 	if (rt) {
624 		if (!(rt->rt6i_flags&RTF_GATEWAY)) {
625 			if (rt->rt6i_dst.plen != 128 &&
626 			    ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
627 				rt->rt6i_flags |= RTF_ANYCAST;
628 			ipv6_addr_copy(&rt->rt6i_gateway, daddr);
629 		}
630 
631 		ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
632 		rt->rt6i_dst.plen = 128;
633 		rt->rt6i_flags |= RTF_CACHE;
634 		rt->u.dst.flags |= DST_HOST;
635 
636 #ifdef CONFIG_IPV6_SUBTREES
637 		if (rt->rt6i_src.plen && saddr) {
638 			ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
639 			rt->rt6i_src.plen = 128;
640 		}
641 #endif
642 
643 		rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
644 
645 	}
646 
647 	return rt;
648 }
649 
650 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
651 {
652 	struct rt6_info *rt = ip6_rt_copy(ort);
653 	if (rt) {
654 		ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
655 		rt->rt6i_dst.plen = 128;
656 		rt->rt6i_flags |= RTF_CACHE;
657 		rt->u.dst.flags |= DST_HOST;
658 		rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
659 	}
660 	return rt;
661 }
662 
663 static struct rt6_info *ip6_pol_route(struct fib6_table *table, int oif,
664 					    struct flowi *fl, int flags)
665 {
666 	struct fib6_node *fn;
667 	struct rt6_info *rt, *nrt;
668 	int strict = 0;
669 	int attempts = 3;
670 	int err;
671 	int reachable = ipv6_devconf.forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
672 
673 	strict |= flags & RT6_LOOKUP_F_IFACE;
674 
675 relookup:
676 	read_lock_bh(&table->tb6_lock);
677 
678 restart_2:
679 	fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
680 
681 restart:
682 	rt = rt6_select(fn, oif, strict | reachable);
683 	BACKTRACK(&fl->fl6_src);
684 	if (rt == &ip6_null_entry ||
685 	    rt->rt6i_flags & RTF_CACHE)
686 		goto out;
687 
688 	dst_hold(&rt->u.dst);
689 	read_unlock_bh(&table->tb6_lock);
690 
691 	if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
692 		nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
693 	else {
694 #if CLONE_OFFLINK_ROUTE
695 		nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
696 #else
697 		goto out2;
698 #endif
699 	}
700 
701 	dst_release(&rt->u.dst);
702 	rt = nrt ? : &ip6_null_entry;
703 
704 	dst_hold(&rt->u.dst);
705 	if (nrt) {
706 		err = ip6_ins_rt(nrt);
707 		if (!err)
708 			goto out2;
709 	}
710 
711 	if (--attempts <= 0)
712 		goto out2;
713 
714 	/*
715 	 * Race condition! In the gap, when table->tb6_lock was
716 	 * released someone could insert this route.  Relookup.
717 	 */
718 	dst_release(&rt->u.dst);
719 	goto relookup;
720 
721 out:
722 	if (reachable) {
723 		reachable = 0;
724 		goto restart_2;
725 	}
726 	dst_hold(&rt->u.dst);
727 	read_unlock_bh(&table->tb6_lock);
728 out2:
729 	rt->u.dst.lastuse = jiffies;
730 	rt->u.dst.__use++;
731 
732 	return rt;
733 }
734 
735 static struct rt6_info *ip6_pol_route_input(struct fib6_table *table,
736 					    struct flowi *fl, int flags)
737 {
738 	return ip6_pol_route(table, fl->iif, fl, flags);
739 }
740 
741 void ip6_route_input(struct sk_buff *skb)
742 {
743 	struct ipv6hdr *iph = ipv6_hdr(skb);
744 	int flags = RT6_LOOKUP_F_HAS_SADDR;
745 	struct flowi fl = {
746 		.iif = skb->dev->ifindex,
747 		.nl_u = {
748 			.ip6_u = {
749 				.daddr = iph->daddr,
750 				.saddr = iph->saddr,
751 				.flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
752 			},
753 		},
754 		.mark = skb->mark,
755 		.proto = iph->nexthdr,
756 	};
757 
758 	if (rt6_need_strict(&iph->daddr))
759 		flags |= RT6_LOOKUP_F_IFACE;
760 
761 	skb->dst = fib6_rule_lookup(&init_net, &fl, flags, ip6_pol_route_input);
762 }
763 
764 static struct rt6_info *ip6_pol_route_output(struct fib6_table *table,
765 					     struct flowi *fl, int flags)
766 {
767 	return ip6_pol_route(table, fl->oif, fl, flags);
768 }
769 
770 struct dst_entry * ip6_route_output(struct sock *sk, struct flowi *fl)
771 {
772 	int flags = 0;
773 
774 	if (rt6_need_strict(&fl->fl6_dst))
775 		flags |= RT6_LOOKUP_F_IFACE;
776 
777 	if (!ipv6_addr_any(&fl->fl6_src))
778 		flags |= RT6_LOOKUP_F_HAS_SADDR;
779 
780 	return fib6_rule_lookup(&init_net, fl, flags, ip6_pol_route_output);
781 }
782 
783 EXPORT_SYMBOL(ip6_route_output);
784 
785 int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
786 {
787 	struct rt6_info *ort = (struct rt6_info *) *dstp;
788 	struct rt6_info *rt = (struct rt6_info *)
789 		dst_alloc(&ip6_dst_blackhole_ops);
790 	struct dst_entry *new = NULL;
791 
792 	if (rt) {
793 		new = &rt->u.dst;
794 
795 		atomic_set(&new->__refcnt, 1);
796 		new->__use = 1;
797 		new->input = dst_discard;
798 		new->output = dst_discard;
799 
800 		memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
801 		new->dev = ort->u.dst.dev;
802 		if (new->dev)
803 			dev_hold(new->dev);
804 		rt->rt6i_idev = ort->rt6i_idev;
805 		if (rt->rt6i_idev)
806 			in6_dev_hold(rt->rt6i_idev);
807 		rt->rt6i_expires = 0;
808 
809 		ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
810 		rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
811 		rt->rt6i_metric = 0;
812 
813 		memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
814 #ifdef CONFIG_IPV6_SUBTREES
815 		memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
816 #endif
817 
818 		dst_free(new);
819 	}
820 
821 	dst_release(*dstp);
822 	*dstp = new;
823 	return (new ? 0 : -ENOMEM);
824 }
825 EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
826 
827 /*
828  *	Destination cache support functions
829  */
830 
831 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
832 {
833 	struct rt6_info *rt;
834 
835 	rt = (struct rt6_info *) dst;
836 
837 	if (rt && rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
838 		return dst;
839 
840 	return NULL;
841 }
842 
843 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
844 {
845 	struct rt6_info *rt = (struct rt6_info *) dst;
846 
847 	if (rt) {
848 		if (rt->rt6i_flags & RTF_CACHE)
849 			ip6_del_rt(rt);
850 		else
851 			dst_release(dst);
852 	}
853 	return NULL;
854 }
855 
856 static void ip6_link_failure(struct sk_buff *skb)
857 {
858 	struct rt6_info *rt;
859 
860 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
861 
862 	rt = (struct rt6_info *) skb->dst;
863 	if (rt) {
864 		if (rt->rt6i_flags&RTF_CACHE) {
865 			dst_set_expires(&rt->u.dst, 0);
866 			rt->rt6i_flags |= RTF_EXPIRES;
867 		} else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
868 			rt->rt6i_node->fn_sernum = -1;
869 	}
870 }
871 
872 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
873 {
874 	struct rt6_info *rt6 = (struct rt6_info*)dst;
875 
876 	if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
877 		rt6->rt6i_flags |= RTF_MODIFIED;
878 		if (mtu < IPV6_MIN_MTU) {
879 			mtu = IPV6_MIN_MTU;
880 			dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
881 		}
882 		dst->metrics[RTAX_MTU-1] = mtu;
883 		call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
884 	}
885 }
886 
887 static int ipv6_get_mtu(struct net_device *dev);
888 
889 static inline unsigned int ipv6_advmss(unsigned int mtu)
890 {
891 	mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
892 
893 	if (mtu < init_net.ipv6.sysctl.ip6_rt_min_advmss)
894 		mtu = init_net.ipv6.sysctl.ip6_rt_min_advmss;
895 
896 	/*
897 	 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
898 	 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
899 	 * IPV6_MAXPLEN is also valid and means: "any MSS,
900 	 * rely only on pmtu discovery"
901 	 */
902 	if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
903 		mtu = IPV6_MAXPLEN;
904 	return mtu;
905 }
906 
907 static struct dst_entry *icmp6_dst_gc_list;
908 static DEFINE_SPINLOCK(icmp6_dst_lock);
909 
910 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
911 				  struct neighbour *neigh,
912 				  struct in6_addr *addr)
913 {
914 	struct rt6_info *rt;
915 	struct inet6_dev *idev = in6_dev_get(dev);
916 
917 	if (unlikely(idev == NULL))
918 		return NULL;
919 
920 	rt = ip6_dst_alloc();
921 	if (unlikely(rt == NULL)) {
922 		in6_dev_put(idev);
923 		goto out;
924 	}
925 
926 	dev_hold(dev);
927 	if (neigh)
928 		neigh_hold(neigh);
929 	else
930 		neigh = ndisc_get_neigh(dev, addr);
931 
932 	rt->rt6i_dev	  = dev;
933 	rt->rt6i_idev     = idev;
934 	rt->rt6i_nexthop  = neigh;
935 	atomic_set(&rt->u.dst.__refcnt, 1);
936 	rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
937 	rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
938 	rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
939 	rt->u.dst.output  = ip6_output;
940 
941 #if 0	/* there's no chance to use these for ndisc */
942 	rt->u.dst.flags   = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
943 				? DST_HOST
944 				: 0;
945 	ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
946 	rt->rt6i_dst.plen = 128;
947 #endif
948 
949 	spin_lock_bh(&icmp6_dst_lock);
950 	rt->u.dst.next = icmp6_dst_gc_list;
951 	icmp6_dst_gc_list = &rt->u.dst;
952 	spin_unlock_bh(&icmp6_dst_lock);
953 
954 	fib6_force_start_gc();
955 
956 out:
957 	return &rt->u.dst;
958 }
959 
960 int icmp6_dst_gc(int *more)
961 {
962 	struct dst_entry *dst, *next, **pprev;
963 	int freed;
964 
965 	next = NULL;
966 	freed = 0;
967 
968 	spin_lock_bh(&icmp6_dst_lock);
969 	pprev = &icmp6_dst_gc_list;
970 
971 	while ((dst = *pprev) != NULL) {
972 		if (!atomic_read(&dst->__refcnt)) {
973 			*pprev = dst->next;
974 			dst_free(dst);
975 			freed++;
976 		} else {
977 			pprev = &dst->next;
978 			(*more)++;
979 		}
980 	}
981 
982 	spin_unlock_bh(&icmp6_dst_lock);
983 
984 	return freed;
985 }
986 
987 static int ip6_dst_gc(struct dst_ops *ops)
988 {
989 	static unsigned expire = 30*HZ;
990 	static unsigned long last_gc;
991 	unsigned long now = jiffies;
992 
993 	if (time_after(last_gc + init_net.ipv6.sysctl.ip6_rt_gc_min_interval, now) &&
994 	    atomic_read(&ip6_dst_ops.entries) <= init_net.ipv6.sysctl.ip6_rt_max_size)
995 		goto out;
996 
997 	expire++;
998 	fib6_run_gc(expire);
999 	last_gc = now;
1000 	if (atomic_read(&ip6_dst_ops.entries) < ip6_dst_ops.gc_thresh)
1001 		expire = init_net.ipv6.sysctl.ip6_rt_gc_timeout>>1;
1002 
1003 out:
1004 	expire -= expire>>init_net.ipv6.sysctl.ip6_rt_gc_elasticity;
1005 	return (atomic_read(&ip6_dst_ops.entries) > init_net.ipv6.sysctl.ip6_rt_max_size);
1006 }
1007 
1008 /* Clean host part of a prefix. Not necessary in radix tree,
1009    but results in cleaner routing tables.
1010 
1011    Remove it only when all the things will work!
1012  */
1013 
1014 static int ipv6_get_mtu(struct net_device *dev)
1015 {
1016 	int mtu = IPV6_MIN_MTU;
1017 	struct inet6_dev *idev;
1018 
1019 	idev = in6_dev_get(dev);
1020 	if (idev) {
1021 		mtu = idev->cnf.mtu6;
1022 		in6_dev_put(idev);
1023 	}
1024 	return mtu;
1025 }
1026 
1027 int ipv6_get_hoplimit(struct net_device *dev)
1028 {
1029 	int hoplimit = ipv6_devconf.hop_limit;
1030 	struct inet6_dev *idev;
1031 
1032 	idev = in6_dev_get(dev);
1033 	if (idev) {
1034 		hoplimit = idev->cnf.hop_limit;
1035 		in6_dev_put(idev);
1036 	}
1037 	return hoplimit;
1038 }
1039 
1040 /*
1041  *
1042  */
1043 
1044 int ip6_route_add(struct fib6_config *cfg)
1045 {
1046 	int err;
1047 	struct rt6_info *rt = NULL;
1048 	struct net_device *dev = NULL;
1049 	struct inet6_dev *idev = NULL;
1050 	struct fib6_table *table;
1051 	int addr_type;
1052 
1053 	if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1054 		return -EINVAL;
1055 #ifndef CONFIG_IPV6_SUBTREES
1056 	if (cfg->fc_src_len)
1057 		return -EINVAL;
1058 #endif
1059 	if (cfg->fc_ifindex) {
1060 		err = -ENODEV;
1061 		dev = dev_get_by_index(&init_net, cfg->fc_ifindex);
1062 		if (!dev)
1063 			goto out;
1064 		idev = in6_dev_get(dev);
1065 		if (!idev)
1066 			goto out;
1067 	}
1068 
1069 	if (cfg->fc_metric == 0)
1070 		cfg->fc_metric = IP6_RT_PRIO_USER;
1071 
1072 	table = fib6_new_table(&init_net, cfg->fc_table);
1073 	if (table == NULL) {
1074 		err = -ENOBUFS;
1075 		goto out;
1076 	}
1077 
1078 	rt = ip6_dst_alloc();
1079 
1080 	if (rt == NULL) {
1081 		err = -ENOMEM;
1082 		goto out;
1083 	}
1084 
1085 	rt->u.dst.obsolete = -1;
1086 	rt->rt6i_expires = jiffies + clock_t_to_jiffies(cfg->fc_expires);
1087 
1088 	if (cfg->fc_protocol == RTPROT_UNSPEC)
1089 		cfg->fc_protocol = RTPROT_BOOT;
1090 	rt->rt6i_protocol = cfg->fc_protocol;
1091 
1092 	addr_type = ipv6_addr_type(&cfg->fc_dst);
1093 
1094 	if (addr_type & IPV6_ADDR_MULTICAST)
1095 		rt->u.dst.input = ip6_mc_input;
1096 	else
1097 		rt->u.dst.input = ip6_forward;
1098 
1099 	rt->u.dst.output = ip6_output;
1100 
1101 	ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1102 	rt->rt6i_dst.plen = cfg->fc_dst_len;
1103 	if (rt->rt6i_dst.plen == 128)
1104 	       rt->u.dst.flags = DST_HOST;
1105 
1106 #ifdef CONFIG_IPV6_SUBTREES
1107 	ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1108 	rt->rt6i_src.plen = cfg->fc_src_len;
1109 #endif
1110 
1111 	rt->rt6i_metric = cfg->fc_metric;
1112 
1113 	/* We cannot add true routes via loopback here,
1114 	   they would result in kernel looping; promote them to reject routes
1115 	 */
1116 	if ((cfg->fc_flags & RTF_REJECT) ||
1117 	    (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1118 		/* hold loopback dev/idev if we haven't done so. */
1119 		if (dev != init_net.loopback_dev) {
1120 			if (dev) {
1121 				dev_put(dev);
1122 				in6_dev_put(idev);
1123 			}
1124 			dev = init_net.loopback_dev;
1125 			dev_hold(dev);
1126 			idev = in6_dev_get(dev);
1127 			if (!idev) {
1128 				err = -ENODEV;
1129 				goto out;
1130 			}
1131 		}
1132 		rt->u.dst.output = ip6_pkt_discard_out;
1133 		rt->u.dst.input = ip6_pkt_discard;
1134 		rt->u.dst.error = -ENETUNREACH;
1135 		rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1136 		goto install_route;
1137 	}
1138 
1139 	if (cfg->fc_flags & RTF_GATEWAY) {
1140 		struct in6_addr *gw_addr;
1141 		int gwa_type;
1142 
1143 		gw_addr = &cfg->fc_gateway;
1144 		ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1145 		gwa_type = ipv6_addr_type(gw_addr);
1146 
1147 		if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1148 			struct rt6_info *grt;
1149 
1150 			/* IPv6 strictly inhibits using not link-local
1151 			   addresses as nexthop address.
1152 			   Otherwise, router will not able to send redirects.
1153 			   It is very good, but in some (rare!) circumstances
1154 			   (SIT, PtP, NBMA NOARP links) it is handy to allow
1155 			   some exceptions. --ANK
1156 			 */
1157 			err = -EINVAL;
1158 			if (!(gwa_type&IPV6_ADDR_UNICAST))
1159 				goto out;
1160 
1161 			grt = rt6_lookup(gw_addr, NULL, cfg->fc_ifindex, 1);
1162 
1163 			err = -EHOSTUNREACH;
1164 			if (grt == NULL)
1165 				goto out;
1166 			if (dev) {
1167 				if (dev != grt->rt6i_dev) {
1168 					dst_release(&grt->u.dst);
1169 					goto out;
1170 				}
1171 			} else {
1172 				dev = grt->rt6i_dev;
1173 				idev = grt->rt6i_idev;
1174 				dev_hold(dev);
1175 				in6_dev_hold(grt->rt6i_idev);
1176 			}
1177 			if (!(grt->rt6i_flags&RTF_GATEWAY))
1178 				err = 0;
1179 			dst_release(&grt->u.dst);
1180 
1181 			if (err)
1182 				goto out;
1183 		}
1184 		err = -EINVAL;
1185 		if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1186 			goto out;
1187 	}
1188 
1189 	err = -ENODEV;
1190 	if (dev == NULL)
1191 		goto out;
1192 
1193 	if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1194 		rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1195 		if (IS_ERR(rt->rt6i_nexthop)) {
1196 			err = PTR_ERR(rt->rt6i_nexthop);
1197 			rt->rt6i_nexthop = NULL;
1198 			goto out;
1199 		}
1200 	}
1201 
1202 	rt->rt6i_flags = cfg->fc_flags;
1203 
1204 install_route:
1205 	if (cfg->fc_mx) {
1206 		struct nlattr *nla;
1207 		int remaining;
1208 
1209 		nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1210 			int type = nla_type(nla);
1211 
1212 			if (type) {
1213 				if (type > RTAX_MAX) {
1214 					err = -EINVAL;
1215 					goto out;
1216 				}
1217 
1218 				rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1219 			}
1220 		}
1221 	}
1222 
1223 	if (rt->u.dst.metrics[RTAX_HOPLIMIT-1] == 0)
1224 		rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1225 	if (!rt->u.dst.metrics[RTAX_MTU-1])
1226 		rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1227 	if (!rt->u.dst.metrics[RTAX_ADVMSS-1])
1228 		rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
1229 	rt->u.dst.dev = dev;
1230 	rt->rt6i_idev = idev;
1231 	rt->rt6i_table = table;
1232 	return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1233 
1234 out:
1235 	if (dev)
1236 		dev_put(dev);
1237 	if (idev)
1238 		in6_dev_put(idev);
1239 	if (rt)
1240 		dst_free(&rt->u.dst);
1241 	return err;
1242 }
1243 
1244 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1245 {
1246 	int err;
1247 	struct fib6_table *table;
1248 
1249 	if (rt == &ip6_null_entry)
1250 		return -ENOENT;
1251 
1252 	table = rt->rt6i_table;
1253 	write_lock_bh(&table->tb6_lock);
1254 
1255 	err = fib6_del(rt, info);
1256 	dst_release(&rt->u.dst);
1257 
1258 	write_unlock_bh(&table->tb6_lock);
1259 
1260 	return err;
1261 }
1262 
1263 int ip6_del_rt(struct rt6_info *rt)
1264 {
1265 	struct nl_info info = {
1266 		.nl_net = &init_net,
1267 	};
1268 	return __ip6_del_rt(rt, &info);
1269 }
1270 
1271 static int ip6_route_del(struct fib6_config *cfg)
1272 {
1273 	struct fib6_table *table;
1274 	struct fib6_node *fn;
1275 	struct rt6_info *rt;
1276 	int err = -ESRCH;
1277 
1278 	table = fib6_get_table(&init_net, cfg->fc_table);
1279 	if (table == NULL)
1280 		return err;
1281 
1282 	read_lock_bh(&table->tb6_lock);
1283 
1284 	fn = fib6_locate(&table->tb6_root,
1285 			 &cfg->fc_dst, cfg->fc_dst_len,
1286 			 &cfg->fc_src, cfg->fc_src_len);
1287 
1288 	if (fn) {
1289 		for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1290 			if (cfg->fc_ifindex &&
1291 			    (rt->rt6i_dev == NULL ||
1292 			     rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1293 				continue;
1294 			if (cfg->fc_flags & RTF_GATEWAY &&
1295 			    !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1296 				continue;
1297 			if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1298 				continue;
1299 			dst_hold(&rt->u.dst);
1300 			read_unlock_bh(&table->tb6_lock);
1301 
1302 			return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1303 		}
1304 	}
1305 	read_unlock_bh(&table->tb6_lock);
1306 
1307 	return err;
1308 }
1309 
1310 /*
1311  *	Handle redirects
1312  */
1313 struct ip6rd_flowi {
1314 	struct flowi fl;
1315 	struct in6_addr gateway;
1316 };
1317 
1318 static struct rt6_info *__ip6_route_redirect(struct fib6_table *table,
1319 					     struct flowi *fl,
1320 					     int flags)
1321 {
1322 	struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1323 	struct rt6_info *rt;
1324 	struct fib6_node *fn;
1325 
1326 	/*
1327 	 * Get the "current" route for this destination and
1328 	 * check if the redirect has come from approriate router.
1329 	 *
1330 	 * RFC 2461 specifies that redirects should only be
1331 	 * accepted if they come from the nexthop to the target.
1332 	 * Due to the way the routes are chosen, this notion
1333 	 * is a bit fuzzy and one might need to check all possible
1334 	 * routes.
1335 	 */
1336 
1337 	read_lock_bh(&table->tb6_lock);
1338 	fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1339 restart:
1340 	for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1341 		/*
1342 		 * Current route is on-link; redirect is always invalid.
1343 		 *
1344 		 * Seems, previous statement is not true. It could
1345 		 * be node, which looks for us as on-link (f.e. proxy ndisc)
1346 		 * But then router serving it might decide, that we should
1347 		 * know truth 8)8) --ANK (980726).
1348 		 */
1349 		if (rt6_check_expired(rt))
1350 			continue;
1351 		if (!(rt->rt6i_flags & RTF_GATEWAY))
1352 			continue;
1353 		if (fl->oif != rt->rt6i_dev->ifindex)
1354 			continue;
1355 		if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1356 			continue;
1357 		break;
1358 	}
1359 
1360 	if (!rt)
1361 		rt = &ip6_null_entry;
1362 	BACKTRACK(&fl->fl6_src);
1363 out:
1364 	dst_hold(&rt->u.dst);
1365 
1366 	read_unlock_bh(&table->tb6_lock);
1367 
1368 	return rt;
1369 };
1370 
1371 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1372 					   struct in6_addr *src,
1373 					   struct in6_addr *gateway,
1374 					   struct net_device *dev)
1375 {
1376 	int flags = RT6_LOOKUP_F_HAS_SADDR;
1377 	struct ip6rd_flowi rdfl = {
1378 		.fl = {
1379 			.oif = dev->ifindex,
1380 			.nl_u = {
1381 				.ip6_u = {
1382 					.daddr = *dest,
1383 					.saddr = *src,
1384 				},
1385 			},
1386 		},
1387 		.gateway = *gateway,
1388 	};
1389 
1390 	if (rt6_need_strict(dest))
1391 		flags |= RT6_LOOKUP_F_IFACE;
1392 
1393 	return (struct rt6_info *)fib6_rule_lookup(&init_net,
1394 						   (struct flowi *)&rdfl,
1395 						   flags, __ip6_route_redirect);
1396 }
1397 
1398 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1399 		  struct in6_addr *saddr,
1400 		  struct neighbour *neigh, u8 *lladdr, int on_link)
1401 {
1402 	struct rt6_info *rt, *nrt = NULL;
1403 	struct netevent_redirect netevent;
1404 
1405 	rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1406 
1407 	if (rt == &ip6_null_entry) {
1408 		if (net_ratelimit())
1409 			printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1410 			       "for redirect target\n");
1411 		goto out;
1412 	}
1413 
1414 	/*
1415 	 *	We have finally decided to accept it.
1416 	 */
1417 
1418 	neigh_update(neigh, lladdr, NUD_STALE,
1419 		     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1420 		     NEIGH_UPDATE_F_OVERRIDE|
1421 		     (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1422 				     NEIGH_UPDATE_F_ISROUTER))
1423 		     );
1424 
1425 	/*
1426 	 * Redirect received -> path was valid.
1427 	 * Look, redirects are sent only in response to data packets,
1428 	 * so that this nexthop apparently is reachable. --ANK
1429 	 */
1430 	dst_confirm(&rt->u.dst);
1431 
1432 	/* Duplicate redirect: silently ignore. */
1433 	if (neigh == rt->u.dst.neighbour)
1434 		goto out;
1435 
1436 	nrt = ip6_rt_copy(rt);
1437 	if (nrt == NULL)
1438 		goto out;
1439 
1440 	nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1441 	if (on_link)
1442 		nrt->rt6i_flags &= ~RTF_GATEWAY;
1443 
1444 	ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1445 	nrt->rt6i_dst.plen = 128;
1446 	nrt->u.dst.flags |= DST_HOST;
1447 
1448 	ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1449 	nrt->rt6i_nexthop = neigh_clone(neigh);
1450 	/* Reset pmtu, it may be better */
1451 	nrt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1452 	nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&nrt->u.dst));
1453 
1454 	if (ip6_ins_rt(nrt))
1455 		goto out;
1456 
1457 	netevent.old = &rt->u.dst;
1458 	netevent.new = &nrt->u.dst;
1459 	call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1460 
1461 	if (rt->rt6i_flags&RTF_CACHE) {
1462 		ip6_del_rt(rt);
1463 		return;
1464 	}
1465 
1466 out:
1467 	dst_release(&rt->u.dst);
1468 	return;
1469 }
1470 
1471 /*
1472  *	Handle ICMP "packet too big" messages
1473  *	i.e. Path MTU discovery
1474  */
1475 
1476 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1477 			struct net_device *dev, u32 pmtu)
1478 {
1479 	struct rt6_info *rt, *nrt;
1480 	int allfrag = 0;
1481 
1482 	rt = rt6_lookup(daddr, saddr, dev->ifindex, 0);
1483 	if (rt == NULL)
1484 		return;
1485 
1486 	if (pmtu >= dst_mtu(&rt->u.dst))
1487 		goto out;
1488 
1489 	if (pmtu < IPV6_MIN_MTU) {
1490 		/*
1491 		 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1492 		 * MTU (1280) and a fragment header should always be included
1493 		 * after a node receiving Too Big message reporting PMTU is
1494 		 * less than the IPv6 Minimum Link MTU.
1495 		 */
1496 		pmtu = IPV6_MIN_MTU;
1497 		allfrag = 1;
1498 	}
1499 
1500 	/* New mtu received -> path was valid.
1501 	   They are sent only in response to data packets,
1502 	   so that this nexthop apparently is reachable. --ANK
1503 	 */
1504 	dst_confirm(&rt->u.dst);
1505 
1506 	/* Host route. If it is static, it would be better
1507 	   not to override it, but add new one, so that
1508 	   when cache entry will expire old pmtu
1509 	   would return automatically.
1510 	 */
1511 	if (rt->rt6i_flags & RTF_CACHE) {
1512 		rt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1513 		if (allfrag)
1514 			rt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1515 		dst_set_expires(&rt->u.dst, init_net.ipv6.sysctl.ip6_rt_mtu_expires);
1516 		rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1517 		goto out;
1518 	}
1519 
1520 	/* Network route.
1521 	   Two cases are possible:
1522 	   1. It is connected route. Action: COW
1523 	   2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1524 	 */
1525 	if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1526 		nrt = rt6_alloc_cow(rt, daddr, saddr);
1527 	else
1528 		nrt = rt6_alloc_clone(rt, daddr);
1529 
1530 	if (nrt) {
1531 		nrt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1532 		if (allfrag)
1533 			nrt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1534 
1535 		/* According to RFC 1981, detecting PMTU increase shouldn't be
1536 		 * happened within 5 mins, the recommended timer is 10 mins.
1537 		 * Here this route expiration time is set to ip6_rt_mtu_expires
1538 		 * which is 10 mins. After 10 mins the decreased pmtu is expired
1539 		 * and detecting PMTU increase will be automatically happened.
1540 		 */
1541 		dst_set_expires(&nrt->u.dst, init_net.ipv6.sysctl.ip6_rt_mtu_expires);
1542 		nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1543 
1544 		ip6_ins_rt(nrt);
1545 	}
1546 out:
1547 	dst_release(&rt->u.dst);
1548 }
1549 
1550 /*
1551  *	Misc support functions
1552  */
1553 
1554 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1555 {
1556 	struct rt6_info *rt = ip6_dst_alloc();
1557 
1558 	if (rt) {
1559 		rt->u.dst.input = ort->u.dst.input;
1560 		rt->u.dst.output = ort->u.dst.output;
1561 
1562 		memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
1563 		rt->u.dst.error = ort->u.dst.error;
1564 		rt->u.dst.dev = ort->u.dst.dev;
1565 		if (rt->u.dst.dev)
1566 			dev_hold(rt->u.dst.dev);
1567 		rt->rt6i_idev = ort->rt6i_idev;
1568 		if (rt->rt6i_idev)
1569 			in6_dev_hold(rt->rt6i_idev);
1570 		rt->u.dst.lastuse = jiffies;
1571 		rt->rt6i_expires = 0;
1572 
1573 		ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1574 		rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1575 		rt->rt6i_metric = 0;
1576 
1577 		memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1578 #ifdef CONFIG_IPV6_SUBTREES
1579 		memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1580 #endif
1581 		rt->rt6i_table = ort->rt6i_table;
1582 	}
1583 	return rt;
1584 }
1585 
1586 #ifdef CONFIG_IPV6_ROUTE_INFO
1587 static struct rt6_info *rt6_get_route_info(struct in6_addr *prefix, int prefixlen,
1588 					   struct in6_addr *gwaddr, int ifindex)
1589 {
1590 	struct fib6_node *fn;
1591 	struct rt6_info *rt = NULL;
1592 	struct fib6_table *table;
1593 
1594 	table = fib6_get_table(&init_net, RT6_TABLE_INFO);
1595 	if (table == NULL)
1596 		return NULL;
1597 
1598 	write_lock_bh(&table->tb6_lock);
1599 	fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1600 	if (!fn)
1601 		goto out;
1602 
1603 	for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1604 		if (rt->rt6i_dev->ifindex != ifindex)
1605 			continue;
1606 		if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1607 			continue;
1608 		if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1609 			continue;
1610 		dst_hold(&rt->u.dst);
1611 		break;
1612 	}
1613 out:
1614 	write_unlock_bh(&table->tb6_lock);
1615 	return rt;
1616 }
1617 
1618 static struct rt6_info *rt6_add_route_info(struct in6_addr *prefix, int prefixlen,
1619 					   struct in6_addr *gwaddr, int ifindex,
1620 					   unsigned pref)
1621 {
1622 	struct fib6_config cfg = {
1623 		.fc_table	= RT6_TABLE_INFO,
1624 		.fc_metric	= IP6_RT_PRIO_USER,
1625 		.fc_ifindex	= ifindex,
1626 		.fc_dst_len	= prefixlen,
1627 		.fc_flags	= RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1628 				  RTF_UP | RTF_PREF(pref),
1629 	};
1630 
1631 	ipv6_addr_copy(&cfg.fc_dst, prefix);
1632 	ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1633 
1634 	/* We should treat it as a default route if prefix length is 0. */
1635 	if (!prefixlen)
1636 		cfg.fc_flags |= RTF_DEFAULT;
1637 
1638 	ip6_route_add(&cfg);
1639 
1640 	return rt6_get_route_info(prefix, prefixlen, gwaddr, ifindex);
1641 }
1642 #endif
1643 
1644 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1645 {
1646 	struct rt6_info *rt;
1647 	struct fib6_table *table;
1648 
1649 	table = fib6_get_table(&init_net, RT6_TABLE_DFLT);
1650 	if (table == NULL)
1651 		return NULL;
1652 
1653 	write_lock_bh(&table->tb6_lock);
1654 	for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
1655 		if (dev == rt->rt6i_dev &&
1656 		    ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1657 		    ipv6_addr_equal(&rt->rt6i_gateway, addr))
1658 			break;
1659 	}
1660 	if (rt)
1661 		dst_hold(&rt->u.dst);
1662 	write_unlock_bh(&table->tb6_lock);
1663 	return rt;
1664 }
1665 
1666 EXPORT_SYMBOL(rt6_get_dflt_router);
1667 
1668 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1669 				     struct net_device *dev,
1670 				     unsigned int pref)
1671 {
1672 	struct fib6_config cfg = {
1673 		.fc_table	= RT6_TABLE_DFLT,
1674 		.fc_metric	= IP6_RT_PRIO_USER,
1675 		.fc_ifindex	= dev->ifindex,
1676 		.fc_flags	= RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1677 				  RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1678 	};
1679 
1680 	ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1681 
1682 	ip6_route_add(&cfg);
1683 
1684 	return rt6_get_dflt_router(gwaddr, dev);
1685 }
1686 
1687 void rt6_purge_dflt_routers(void)
1688 {
1689 	struct rt6_info *rt;
1690 	struct fib6_table *table;
1691 
1692 	/* NOTE: Keep consistent with rt6_get_dflt_router */
1693 	table = fib6_get_table(&init_net, RT6_TABLE_DFLT);
1694 	if (table == NULL)
1695 		return;
1696 
1697 restart:
1698 	read_lock_bh(&table->tb6_lock);
1699 	for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1700 		if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1701 			dst_hold(&rt->u.dst);
1702 			read_unlock_bh(&table->tb6_lock);
1703 			ip6_del_rt(rt);
1704 			goto restart;
1705 		}
1706 	}
1707 	read_unlock_bh(&table->tb6_lock);
1708 }
1709 
1710 static void rtmsg_to_fib6_config(struct in6_rtmsg *rtmsg,
1711 				 struct fib6_config *cfg)
1712 {
1713 	memset(cfg, 0, sizeof(*cfg));
1714 
1715 	cfg->fc_table = RT6_TABLE_MAIN;
1716 	cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1717 	cfg->fc_metric = rtmsg->rtmsg_metric;
1718 	cfg->fc_expires = rtmsg->rtmsg_info;
1719 	cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1720 	cfg->fc_src_len = rtmsg->rtmsg_src_len;
1721 	cfg->fc_flags = rtmsg->rtmsg_flags;
1722 
1723 	cfg->fc_nlinfo.nl_net = &init_net;
1724 
1725 	ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1726 	ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1727 	ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1728 }
1729 
1730 int ipv6_route_ioctl(unsigned int cmd, void __user *arg)
1731 {
1732 	struct fib6_config cfg;
1733 	struct in6_rtmsg rtmsg;
1734 	int err;
1735 
1736 	switch(cmd) {
1737 	case SIOCADDRT:		/* Add a route */
1738 	case SIOCDELRT:		/* Delete a route */
1739 		if (!capable(CAP_NET_ADMIN))
1740 			return -EPERM;
1741 		err = copy_from_user(&rtmsg, arg,
1742 				     sizeof(struct in6_rtmsg));
1743 		if (err)
1744 			return -EFAULT;
1745 
1746 		rtmsg_to_fib6_config(&rtmsg, &cfg);
1747 
1748 		rtnl_lock();
1749 		switch (cmd) {
1750 		case SIOCADDRT:
1751 			err = ip6_route_add(&cfg);
1752 			break;
1753 		case SIOCDELRT:
1754 			err = ip6_route_del(&cfg);
1755 			break;
1756 		default:
1757 			err = -EINVAL;
1758 		}
1759 		rtnl_unlock();
1760 
1761 		return err;
1762 	}
1763 
1764 	return -EINVAL;
1765 }
1766 
1767 /*
1768  *	Drop the packet on the floor
1769  */
1770 
1771 static int ip6_pkt_drop(struct sk_buff *skb, int code, 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(&init_net, 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 *net, struct net_device *dev)
1874 {
1875 	fib6_clean_all(net, 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(dev->nd_net, 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 	cfg->fc_nlinfo.nl_net = skb->sk->sk_net;
1969 
1970 	if (tb[RTA_GATEWAY]) {
1971 		nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
1972 		cfg->fc_flags |= RTF_GATEWAY;
1973 	}
1974 
1975 	if (tb[RTA_DST]) {
1976 		int plen = (rtm->rtm_dst_len + 7) >> 3;
1977 
1978 		if (nla_len(tb[RTA_DST]) < plen)
1979 			goto errout;
1980 
1981 		nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1982 	}
1983 
1984 	if (tb[RTA_SRC]) {
1985 		int plen = (rtm->rtm_src_len + 7) >> 3;
1986 
1987 		if (nla_len(tb[RTA_SRC]) < plen)
1988 			goto errout;
1989 
1990 		nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1991 	}
1992 
1993 	if (tb[RTA_OIF])
1994 		cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
1995 
1996 	if (tb[RTA_PRIORITY])
1997 		cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
1998 
1999 	if (tb[RTA_METRICS]) {
2000 		cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2001 		cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2002 	}
2003 
2004 	if (tb[RTA_TABLE])
2005 		cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2006 
2007 	err = 0;
2008 errout:
2009 	return err;
2010 }
2011 
2012 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2013 {
2014 	struct net *net = skb->sk->sk_net;
2015 	struct fib6_config cfg;
2016 	int err;
2017 
2018 	if (net != &init_net)
2019 		return -EINVAL;
2020 
2021 	err = rtm_to_fib6_config(skb, nlh, &cfg);
2022 	if (err < 0)
2023 		return err;
2024 
2025 	return ip6_route_del(&cfg);
2026 }
2027 
2028 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2029 {
2030 	struct net *net = skb->sk->sk_net;
2031 	struct fib6_config cfg;
2032 	int err;
2033 
2034 	if (net != &init_net)
2035 		return -EINVAL;
2036 
2037 	err = rtm_to_fib6_config(skb, nlh, &cfg);
2038 	if (err < 0)
2039 		return err;
2040 
2041 	return ip6_route_add(&cfg);
2042 }
2043 
2044 static inline size_t rt6_nlmsg_size(void)
2045 {
2046 	return NLMSG_ALIGN(sizeof(struct rtmsg))
2047 	       + nla_total_size(16) /* RTA_SRC */
2048 	       + nla_total_size(16) /* RTA_DST */
2049 	       + nla_total_size(16) /* RTA_GATEWAY */
2050 	       + nla_total_size(16) /* RTA_PREFSRC */
2051 	       + nla_total_size(4) /* RTA_TABLE */
2052 	       + nla_total_size(4) /* RTA_IIF */
2053 	       + nla_total_size(4) /* RTA_OIF */
2054 	       + nla_total_size(4) /* RTA_PRIORITY */
2055 	       + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2056 	       + nla_total_size(sizeof(struct rta_cacheinfo));
2057 }
2058 
2059 static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
2060 			 struct in6_addr *dst, struct in6_addr *src,
2061 			 int iif, int type, u32 pid, u32 seq,
2062 			 int prefix, unsigned int flags)
2063 {
2064 	struct rtmsg *rtm;
2065 	struct nlmsghdr *nlh;
2066 	long expires;
2067 	u32 table;
2068 
2069 	if (prefix) {	/* user wants prefix routes only */
2070 		if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2071 			/* success since this is not a prefix route */
2072 			return 1;
2073 		}
2074 	}
2075 
2076 	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2077 	if (nlh == NULL)
2078 		return -EMSGSIZE;
2079 
2080 	rtm = nlmsg_data(nlh);
2081 	rtm->rtm_family = AF_INET6;
2082 	rtm->rtm_dst_len = rt->rt6i_dst.plen;
2083 	rtm->rtm_src_len = rt->rt6i_src.plen;
2084 	rtm->rtm_tos = 0;
2085 	if (rt->rt6i_table)
2086 		table = rt->rt6i_table->tb6_id;
2087 	else
2088 		table = RT6_TABLE_UNSPEC;
2089 	rtm->rtm_table = table;
2090 	NLA_PUT_U32(skb, RTA_TABLE, table);
2091 	if (rt->rt6i_flags&RTF_REJECT)
2092 		rtm->rtm_type = RTN_UNREACHABLE;
2093 	else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2094 		rtm->rtm_type = RTN_LOCAL;
2095 	else
2096 		rtm->rtm_type = RTN_UNICAST;
2097 	rtm->rtm_flags = 0;
2098 	rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2099 	rtm->rtm_protocol = rt->rt6i_protocol;
2100 	if (rt->rt6i_flags&RTF_DYNAMIC)
2101 		rtm->rtm_protocol = RTPROT_REDIRECT;
2102 	else if (rt->rt6i_flags & RTF_ADDRCONF)
2103 		rtm->rtm_protocol = RTPROT_KERNEL;
2104 	else if (rt->rt6i_flags&RTF_DEFAULT)
2105 		rtm->rtm_protocol = RTPROT_RA;
2106 
2107 	if (rt->rt6i_flags&RTF_CACHE)
2108 		rtm->rtm_flags |= RTM_F_CLONED;
2109 
2110 	if (dst) {
2111 		NLA_PUT(skb, RTA_DST, 16, dst);
2112 		rtm->rtm_dst_len = 128;
2113 	} else if (rtm->rtm_dst_len)
2114 		NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2115 #ifdef CONFIG_IPV6_SUBTREES
2116 	if (src) {
2117 		NLA_PUT(skb, RTA_SRC, 16, src);
2118 		rtm->rtm_src_len = 128;
2119 	} else if (rtm->rtm_src_len)
2120 		NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2121 #endif
2122 	if (iif)
2123 		NLA_PUT_U32(skb, RTA_IIF, iif);
2124 	else if (dst) {
2125 		struct in6_addr saddr_buf;
2126 		if (ipv6_dev_get_saddr(ip6_dst_idev(&rt->u.dst)->dev,
2127 				       dst, &saddr_buf) == 0)
2128 			NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2129 	}
2130 
2131 	if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2132 		goto nla_put_failure;
2133 
2134 	if (rt->u.dst.neighbour)
2135 		NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2136 
2137 	if (rt->u.dst.dev)
2138 		NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2139 
2140 	NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2141 
2142 	expires = rt->rt6i_expires ? rt->rt6i_expires - jiffies : 0;
2143 	if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
2144 			       expires, rt->u.dst.error) < 0)
2145 		goto nla_put_failure;
2146 
2147 	return nlmsg_end(skb, nlh);
2148 
2149 nla_put_failure:
2150 	nlmsg_cancel(skb, nlh);
2151 	return -EMSGSIZE;
2152 }
2153 
2154 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2155 {
2156 	struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2157 	int prefix;
2158 
2159 	if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2160 		struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2161 		prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2162 	} else
2163 		prefix = 0;
2164 
2165 	return rt6_fill_node(arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2166 		     NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2167 		     prefix, NLM_F_MULTI);
2168 }
2169 
2170 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2171 {
2172 	struct net *net = in_skb->sk->sk_net;
2173 	struct nlattr *tb[RTA_MAX+1];
2174 	struct rt6_info *rt;
2175 	struct sk_buff *skb;
2176 	struct rtmsg *rtm;
2177 	struct flowi fl;
2178 	int err, iif = 0;
2179 
2180 	if (net != &init_net)
2181 		return -EINVAL;
2182 
2183 	err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2184 	if (err < 0)
2185 		goto errout;
2186 
2187 	err = -EINVAL;
2188 	memset(&fl, 0, sizeof(fl));
2189 
2190 	if (tb[RTA_SRC]) {
2191 		if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2192 			goto errout;
2193 
2194 		ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2195 	}
2196 
2197 	if (tb[RTA_DST]) {
2198 		if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2199 			goto errout;
2200 
2201 		ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2202 	}
2203 
2204 	if (tb[RTA_IIF])
2205 		iif = nla_get_u32(tb[RTA_IIF]);
2206 
2207 	if (tb[RTA_OIF])
2208 		fl.oif = nla_get_u32(tb[RTA_OIF]);
2209 
2210 	if (iif) {
2211 		struct net_device *dev;
2212 		dev = __dev_get_by_index(&init_net, iif);
2213 		if (!dev) {
2214 			err = -ENODEV;
2215 			goto errout;
2216 		}
2217 	}
2218 
2219 	skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2220 	if (skb == NULL) {
2221 		err = -ENOBUFS;
2222 		goto errout;
2223 	}
2224 
2225 	/* Reserve room for dummy headers, this skb can pass
2226 	   through good chunk of routing engine.
2227 	 */
2228 	skb_reset_mac_header(skb);
2229 	skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2230 
2231 	rt = (struct rt6_info*) ip6_route_output(NULL, &fl);
2232 	skb->dst = &rt->u.dst;
2233 
2234 	err = rt6_fill_node(skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2235 			    RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2236 			    nlh->nlmsg_seq, 0, 0);
2237 	if (err < 0) {
2238 		kfree_skb(skb);
2239 		goto errout;
2240 	}
2241 
2242 	err = rtnl_unicast(skb, &init_net, NETLINK_CB(in_skb).pid);
2243 errout:
2244 	return err;
2245 }
2246 
2247 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2248 {
2249 	struct sk_buff *skb;
2250 	u32 seq;
2251 	int err;
2252 
2253 	err = -ENOBUFS;
2254 	seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2255 
2256 	skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2257 	if (skb == NULL)
2258 		goto errout;
2259 
2260 	err = rt6_fill_node(skb, rt, NULL, NULL, 0,
2261 				event, info->pid, seq, 0, 0);
2262 	if (err < 0) {
2263 		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2264 		WARN_ON(err == -EMSGSIZE);
2265 		kfree_skb(skb);
2266 		goto errout;
2267 	}
2268 	err = rtnl_notify(skb, &init_net, info->pid,
2269 				RTNLGRP_IPV6_ROUTE, info->nlh, gfp_any());
2270 errout:
2271 	if (err < 0)
2272 		rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_ROUTE, err);
2273 }
2274 
2275 /*
2276  *	/proc
2277  */
2278 
2279 #ifdef CONFIG_PROC_FS
2280 
2281 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2282 
2283 struct rt6_proc_arg
2284 {
2285 	char *buffer;
2286 	int offset;
2287 	int length;
2288 	int skip;
2289 	int len;
2290 };
2291 
2292 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2293 {
2294 	struct seq_file *m = p_arg;
2295 
2296 	seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_dst.addr),
2297 		   rt->rt6i_dst.plen);
2298 
2299 #ifdef CONFIG_IPV6_SUBTREES
2300 	seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_src.addr),
2301 		   rt->rt6i_src.plen);
2302 #else
2303 	seq_puts(m, "00000000000000000000000000000000 00 ");
2304 #endif
2305 
2306 	if (rt->rt6i_nexthop) {
2307 		seq_printf(m, NIP6_SEQFMT,
2308 			   NIP6(*((struct in6_addr *)rt->rt6i_nexthop->primary_key)));
2309 	} else {
2310 		seq_puts(m, "00000000000000000000000000000000");
2311 	}
2312 	seq_printf(m, " %08x %08x %08x %08x %8s\n",
2313 		   rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2314 		   rt->u.dst.__use, rt->rt6i_flags,
2315 		   rt->rt6i_dev ? rt->rt6i_dev->name : "");
2316 	return 0;
2317 }
2318 
2319 static int ipv6_route_show(struct seq_file *m, void *v)
2320 {
2321 	struct net *net = (struct net *)m->private;
2322 	fib6_clean_all(net, rt6_info_route, 0, m);
2323 	return 0;
2324 }
2325 
2326 static int ipv6_route_open(struct inode *inode, struct file *file)
2327 {
2328 	struct net *net = get_proc_net(inode);
2329 	if (!net)
2330 		return -ENXIO;
2331 	return single_open(file, ipv6_route_show, net);
2332 }
2333 
2334 static int ipv6_route_release(struct inode *inode, struct file *file)
2335 {
2336 	struct seq_file *seq = file->private_data;
2337 	struct net *net = seq->private;
2338 	put_net(net);
2339 	return single_release(inode, file);
2340 }
2341 
2342 static const struct file_operations ipv6_route_proc_fops = {
2343 	.owner		= THIS_MODULE,
2344 	.open		= ipv6_route_open,
2345 	.read		= seq_read,
2346 	.llseek		= seq_lseek,
2347 	.release	= ipv6_route_release,
2348 };
2349 
2350 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2351 {
2352 	seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2353 		      rt6_stats.fib_nodes, rt6_stats.fib_route_nodes,
2354 		      rt6_stats.fib_rt_alloc, rt6_stats.fib_rt_entries,
2355 		      rt6_stats.fib_rt_cache,
2356 		      atomic_read(&ip6_dst_ops.entries),
2357 		      rt6_stats.fib_discarded_routes);
2358 
2359 	return 0;
2360 }
2361 
2362 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2363 {
2364 	return single_open(file, rt6_stats_seq_show, NULL);
2365 }
2366 
2367 static const struct file_operations rt6_stats_seq_fops = {
2368 	.owner	 = THIS_MODULE,
2369 	.open	 = rt6_stats_seq_open,
2370 	.read	 = seq_read,
2371 	.llseek	 = seq_lseek,
2372 	.release = single_release,
2373 };
2374 
2375 static int ipv6_route_proc_init(struct net *net)
2376 {
2377 	int ret = -ENOMEM;
2378 	if (!proc_net_fops_create(net, "ipv6_route",
2379 				  0, &ipv6_route_proc_fops))
2380 		goto out;
2381 
2382 	if (!proc_net_fops_create(net, "rt6_stats",
2383 				  S_IRUGO, &rt6_stats_seq_fops))
2384 		goto out_ipv6_route;
2385 
2386 	ret = 0;
2387 out:
2388 	return ret;
2389 out_ipv6_route:
2390 	proc_net_remove(net, "ipv6_route");
2391 	goto out;
2392 }
2393 
2394 static void ipv6_route_proc_fini(struct net *net)
2395 {
2396 	proc_net_remove(net, "ipv6_route");
2397 	proc_net_remove(net, "rt6_stats");
2398 }
2399 #else
2400 static inline int ipv6_route_proc_init(struct net *net)
2401 {
2402 	return 0;
2403 }
2404 static inline void ipv6_route_proc_fini(struct net *net)
2405 {
2406 	return ;
2407 }
2408 #endif	/* CONFIG_PROC_FS */
2409 
2410 #ifdef CONFIG_SYSCTL
2411 
2412 static
2413 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write, struct file * filp,
2414 			      void __user *buffer, size_t *lenp, loff_t *ppos)
2415 {
2416 	int delay = init_net.ipv6.sysctl.flush_delay;
2417 	if (write) {
2418 		proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
2419 		fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay);
2420 		return 0;
2421 	} else
2422 		return -EINVAL;
2423 }
2424 
2425 ctl_table ipv6_route_table_template[] = {
2426 	{
2427 		.procname	=	"flush",
2428 		.data		=	&init_net.ipv6.sysctl.flush_delay,
2429 		.maxlen		=	sizeof(int),
2430 		.mode		=	0200,
2431 		.proc_handler	=	&ipv6_sysctl_rtcache_flush
2432 	},
2433 	{
2434 		.ctl_name	=	NET_IPV6_ROUTE_GC_THRESH,
2435 		.procname	=	"gc_thresh",
2436 		.data		=	&ip6_dst_ops.gc_thresh,
2437 		.maxlen		=	sizeof(int),
2438 		.mode		=	0644,
2439 		.proc_handler	=	&proc_dointvec,
2440 	},
2441 	{
2442 		.ctl_name	=	NET_IPV6_ROUTE_MAX_SIZE,
2443 		.procname	=	"max_size",
2444 		.data		=	&init_net.ipv6.sysctl.ip6_rt_max_size,
2445 		.maxlen		=	sizeof(int),
2446 		.mode		=	0644,
2447 		.proc_handler	=	&proc_dointvec,
2448 	},
2449 	{
2450 		.ctl_name	=	NET_IPV6_ROUTE_GC_MIN_INTERVAL,
2451 		.procname	=	"gc_min_interval",
2452 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2453 		.maxlen		=	sizeof(int),
2454 		.mode		=	0644,
2455 		.proc_handler	=	&proc_dointvec_jiffies,
2456 		.strategy	=	&sysctl_jiffies,
2457 	},
2458 	{
2459 		.ctl_name	=	NET_IPV6_ROUTE_GC_TIMEOUT,
2460 		.procname	=	"gc_timeout",
2461 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2462 		.maxlen		=	sizeof(int),
2463 		.mode		=	0644,
2464 		.proc_handler	=	&proc_dointvec_jiffies,
2465 		.strategy	=	&sysctl_jiffies,
2466 	},
2467 	{
2468 		.ctl_name	=	NET_IPV6_ROUTE_GC_INTERVAL,
2469 		.procname	=	"gc_interval",
2470 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_interval,
2471 		.maxlen		=	sizeof(int),
2472 		.mode		=	0644,
2473 		.proc_handler	=	&proc_dointvec_jiffies,
2474 		.strategy	=	&sysctl_jiffies,
2475 	},
2476 	{
2477 		.ctl_name	=	NET_IPV6_ROUTE_GC_ELASTICITY,
2478 		.procname	=	"gc_elasticity",
2479 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2480 		.maxlen		=	sizeof(int),
2481 		.mode		=	0644,
2482 		.proc_handler	=	&proc_dointvec_jiffies,
2483 		.strategy	=	&sysctl_jiffies,
2484 	},
2485 	{
2486 		.ctl_name	=	NET_IPV6_ROUTE_MTU_EXPIRES,
2487 		.procname	=	"mtu_expires",
2488 		.data		=	&init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2489 		.maxlen		=	sizeof(int),
2490 		.mode		=	0644,
2491 		.proc_handler	=	&proc_dointvec_jiffies,
2492 		.strategy	=	&sysctl_jiffies,
2493 	},
2494 	{
2495 		.ctl_name	=	NET_IPV6_ROUTE_MIN_ADVMSS,
2496 		.procname	=	"min_adv_mss",
2497 		.data		=	&init_net.ipv6.sysctl.ip6_rt_min_advmss,
2498 		.maxlen		=	sizeof(int),
2499 		.mode		=	0644,
2500 		.proc_handler	=	&proc_dointvec_jiffies,
2501 		.strategy	=	&sysctl_jiffies,
2502 	},
2503 	{
2504 		.ctl_name	=	NET_IPV6_ROUTE_GC_MIN_INTERVAL_MS,
2505 		.procname	=	"gc_min_interval_ms",
2506 		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2507 		.maxlen		=	sizeof(int),
2508 		.mode		=	0644,
2509 		.proc_handler	=	&proc_dointvec_ms_jiffies,
2510 		.strategy	=	&sysctl_ms_jiffies,
2511 	},
2512 	{ .ctl_name = 0 }
2513 };
2514 
2515 struct ctl_table *ipv6_route_sysctl_init(struct net *net)
2516 {
2517 	struct ctl_table *table;
2518 
2519 	table = kmemdup(ipv6_route_table_template,
2520 			sizeof(ipv6_route_table_template),
2521 			GFP_KERNEL);
2522 
2523 	if (table) {
2524 		table[0].data = &net->ipv6.sysctl.flush_delay;
2525 		/* table[1].data will be handled when we have
2526 		   routes per namespace */
2527 		table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2528 		table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2529 		table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2530 		table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2531 		table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2532 		table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2533 		table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2534 	}
2535 
2536 	return table;
2537 }
2538 #endif
2539 
2540 int __init ip6_route_init(void)
2541 {
2542 	int ret;
2543 
2544 	ip6_dst_ops.kmem_cachep =
2545 		kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2546 				  SLAB_HWCACHE_ALIGN, NULL);
2547 	if (!ip6_dst_ops.kmem_cachep)
2548 		return -ENOMEM;
2549 
2550 	ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops.kmem_cachep;
2551 
2552 	ret = fib6_init();
2553 	if (ret)
2554 		goto out_kmem_cache;
2555 
2556 	ret = ipv6_route_proc_init(&init_net);
2557 	if (ret)
2558 		goto out_fib6_init;
2559 
2560 	ret = xfrm6_init();
2561 	if (ret)
2562 		goto out_proc_init;
2563 
2564 	ret = fib6_rules_init();
2565 	if (ret)
2566 		goto xfrm6_init;
2567 
2568 	ret = -ENOBUFS;
2569 	if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2570 	    __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2571 	    __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2572 		goto fib6_rules_init;
2573 
2574 	ret = 0;
2575 out:
2576 	return ret;
2577 
2578 fib6_rules_init:
2579 	fib6_rules_cleanup();
2580 xfrm6_init:
2581 	xfrm6_fini();
2582 out_proc_init:
2583 	ipv6_route_proc_fini(&init_net);
2584 out_fib6_init:
2585 	rt6_ifdown(&init_net, NULL);
2586 	fib6_gc_cleanup();
2587 out_kmem_cache:
2588 	kmem_cache_destroy(ip6_dst_ops.kmem_cachep);
2589 	goto out;
2590 }
2591 
2592 void ip6_route_cleanup(void)
2593 {
2594 	fib6_rules_cleanup();
2595 	ipv6_route_proc_fini(&init_net);
2596 	xfrm6_fini();
2597 	rt6_ifdown(&init_net, NULL);
2598 	fib6_gc_cleanup();
2599 	kmem_cache_destroy(ip6_dst_ops.kmem_cachep);
2600 }
2601