projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / ipv6 / route.c
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 .entry_size = sizeof(struct rt6_info),
117 };
118
119 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
120 {
121 }
122
123 static struct dst_ops ip6_dst_blackhole_ops = {
124 .family = AF_INET6,
125 .protocol = __constant_htons(ETH_P_IPV6),
126 .destroy = ip6_dst_destroy,
127 .check = ip6_dst_check,
128 .update_pmtu = ip6_rt_blackhole_update_pmtu,
129 .entry_size = sizeof(struct rt6_info),
130 };
131
132 struct rt6_info ip6_null_entry = {
133 .u = {
134 .dst = {
135 .__refcnt = ATOMIC_INIT(1),
136 .__use = 1,
137 .obsolete = -1,
138 .error = -ENETUNREACH,
139 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
140 .input = ip6_pkt_discard,
141 .output = ip6_pkt_discard_out,
142 .ops = &ip6_dst_ops,
143 .path = (struct dst_entry*)&ip6_null_entry,
144 }
145 },
146 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
147 .rt6i_metric = ~(u32) 0,
148 .rt6i_ref = ATOMIC_INIT(1),
149 };
150
151 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
152
153 static int ip6_pkt_prohibit(struct sk_buff *skb);
154 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
155 static int ip6_pkt_blk_hole(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 = ip6_pkt_blk_hole,
185 .output = ip6_pkt_blk_hole,
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
220 if (dev != init_net.loopback_dev && idev != NULL && idev->dev == dev) {
221 struct inet6_dev *loopback_idev = in6_dev_get(init_net.loopback_dev);
222 if (loopback_idev != NULL) {
223 rt->rt6i_idev = loopback_idev;
224 in6_dev_put(idev);
225 }
226 }
227 }
228
229 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
230 {
231 return (rt->rt6i_flags & RTF_EXPIRES &&
232 time_after(jiffies, rt->rt6i_expires));
233 }
234
235 static inline int rt6_need_strict(struct in6_addr *daddr)
236 {
237 return (ipv6_addr_type(daddr) &
238 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
239 }
240
241 /*
242 * Route lookup. Any table->tb6_lock is implied.
243 */
244
245 static __inline__ struct rt6_info *rt6_device_match(struct rt6_info *rt,
246 int oif,
247 int strict)
248 {
249 struct rt6_info *local = NULL;
250 struct rt6_info *sprt;
251
252 if (oif) {
253 for (sprt = rt; sprt; sprt = sprt->u.dst.rt6_next) {
254 struct net_device *dev = sprt->rt6i_dev;
255 if (dev->ifindex == oif)
256 return sprt;
257 if (dev->flags & IFF_LOOPBACK) {
258 if (sprt->rt6i_idev == NULL ||
259 sprt->rt6i_idev->dev->ifindex != oif) {
260 if (strict && oif)
261 continue;
262 if (local && (!oif ||
263 local->rt6i_idev->dev->ifindex == oif))
264 continue;
265 }
266 local = sprt;
267 }
268 }
269
270 if (local)
271 return local;
272
273 if (strict)
274 return &ip6_null_entry;
275 }
276 return rt;
277 }
278
279 #ifdef CONFIG_IPV6_ROUTER_PREF
280 static void rt6_probe(struct rt6_info *rt)
281 {
282 struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
283 /*
284 * Okay, this does not seem to be appropriate
285 * for now, however, we need to check if it
286 * is really so; aka Router Reachability Probing.
287 *
288 * Router Reachability Probe MUST be rate-limited
289 * to no more than one per minute.
290 */
291 if (!neigh || (neigh->nud_state & NUD_VALID))
292 return;
293 read_lock_bh(&neigh->lock);
294 if (!(neigh->nud_state & NUD_VALID) &&
295 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
296 struct in6_addr mcaddr;
297 struct in6_addr *target;
298
299 neigh->updated = jiffies;
300 read_unlock_bh(&neigh->lock);
301
302 target = (struct in6_addr *)&neigh->primary_key;
303 addrconf_addr_solict_mult(target, &mcaddr);
304 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
305 } else
306 read_unlock_bh(&neigh->lock);
307 }
308 #else
309 static inline void rt6_probe(struct rt6_info *rt)
310 {
311 return;
312 }
313 #endif
314
315 /*
316 * Default Router Selection (RFC 2461 6.3.6)
317 */
318 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
319 {
320 struct net_device *dev = rt->rt6i_dev;
321 if (!oif || dev->ifindex == oif)
322 return 2;
323 if ((dev->flags & IFF_LOOPBACK) &&
324 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
325 return 1;
326 return 0;
327 }
328
329 static inline int rt6_check_neigh(struct rt6_info *rt)
330 {
331 struct neighbour *neigh = rt->rt6i_nexthop;
332 int m = 0;
333 if (rt->rt6i_flags & RTF_NONEXTHOP ||
334 !(rt->rt6i_flags & RTF_GATEWAY))
335 m = 1;
336 else if (neigh) {
337 read_lock_bh(&neigh->lock);
338 if (neigh->nud_state & NUD_VALID)
339 m = 2;
340 else if (!(neigh->nud_state & NUD_FAILED))
341 m = 1;
342 read_unlock_bh(&neigh->lock);
343 }
344 return m;
345 }
346
347 static int rt6_score_route(struct rt6_info *rt, int oif,
348 int strict)
349 {
350 int m, n;
351
352 m = rt6_check_dev(rt, oif);
353 if (!m && (strict & RT6_LOOKUP_F_IFACE))
354 return -1;
355 #ifdef CONFIG_IPV6_ROUTER_PREF
356 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
357 #endif
358 n = rt6_check_neigh(rt);
359 if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
360 return -1;
361 return m;
362 }
363
364 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
365 int *mpri, struct rt6_info *match)
366 {
367 int m;
368
369 if (rt6_check_expired(rt))
370 goto out;
371
372 m = rt6_score_route(rt, oif, strict);
373 if (m < 0)
374 goto out;
375
376 if (m > *mpri) {
377 if (strict & RT6_LOOKUP_F_REACHABLE)
378 rt6_probe(match);
379 *mpri = m;
380 match = rt;
381 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
382 rt6_probe(rt);
383 }
384
385 out:
386 return match;
387 }
388
389 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
390 struct rt6_info *rr_head,
391 u32 metric, int oif, int strict)
392 {
393 struct rt6_info *rt, *match;
394 int mpri = -1;
395
396 match = NULL;
397 for (rt = rr_head; rt && rt->rt6i_metric == metric;
398 rt = rt->u.dst.rt6_next)
399 match = find_match(rt, oif, strict, &mpri, match);
400 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
401 rt = rt->u.dst.rt6_next)
402 match = find_match(rt, oif, strict, &mpri, match);
403
404 return match;
405 }
406
407 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
408 {
409 struct rt6_info *match, *rt0;
410
411 RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
412 __FUNCTION__, fn->leaf, oif);
413
414 rt0 = fn->rr_ptr;
415 if (!rt0)
416 fn->rr_ptr = rt0 = fn->leaf;
417
418 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
419
420 if (!match &&
421 (strict & RT6_LOOKUP_F_REACHABLE)) {
422 struct rt6_info *next = rt0->u.dst.rt6_next;
423
424 /* no entries matched; do round-robin */
425 if (!next || next->rt6i_metric != rt0->rt6i_metric)
426 next = fn->leaf;
427
428 if (next != rt0)
429 fn->rr_ptr = next;
430 }
431
432 RT6_TRACE("%s() => %p\n",
433 __FUNCTION__, match);
434
435 return (match ? match : &ip6_null_entry);
436 }
437
438 #ifdef CONFIG_IPV6_ROUTE_INFO
439 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
440 struct in6_addr *gwaddr)
441 {
442 struct route_info *rinfo = (struct route_info *) opt;
443 struct in6_addr prefix_buf, *prefix;
444 unsigned int pref;
445 u32 lifetime;
446 struct rt6_info *rt;
447
448 if (len < sizeof(struct route_info)) {
449 return -EINVAL;
450 }
451
452 /* Sanity check for prefix_len and length */
453 if (rinfo->length > 3) {
454 return -EINVAL;
455 } else if (rinfo->prefix_len > 128) {
456 return -EINVAL;
457 } else if (rinfo->prefix_len > 64) {
458 if (rinfo->length < 2) {
459 return -EINVAL;
460 }
461 } else if (rinfo->prefix_len > 0) {
462 if (rinfo->length < 1) {
463 return -EINVAL;
464 }
465 }
466
467 pref = rinfo->route_pref;
468 if (pref == ICMPV6_ROUTER_PREF_INVALID)
469 pref = ICMPV6_ROUTER_PREF_MEDIUM;
470
471 lifetime = ntohl(rinfo->lifetime);
472 if (lifetime == 0xffffffff) {
473 /* infinity */
474 } else if (lifetime > 0x7fffffff/HZ) {
475 /* Avoid arithmetic overflow */
476 lifetime = 0x7fffffff/HZ - 1;
477 }
478
479 if (rinfo->length == 3)
480 prefix = (struct in6_addr *)rinfo->prefix;
481 else {
482 /* this function is safe */
483 ipv6_addr_prefix(&prefix_buf,
484 (struct in6_addr *)rinfo->prefix,
485 rinfo->prefix_len);
486 prefix = &prefix_buf;
487 }
488
489 rt = rt6_get_route_info(prefix, rinfo->prefix_len, gwaddr, dev->ifindex);
490
491 if (rt && !lifetime) {
492 ip6_del_rt(rt);
493 rt = NULL;
494 }
495
496 if (!rt && lifetime)
497 rt = rt6_add_route_info(prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
498 pref);
499 else if (rt)
500 rt->rt6i_flags = RTF_ROUTEINFO |
501 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
502
503 if (rt) {
504 if (lifetime == 0xffffffff) {
505 rt->rt6i_flags &= ~RTF_EXPIRES;
506 } else {
507 rt->rt6i_expires = jiffies + HZ * lifetime;
508 rt->rt6i_flags |= RTF_EXPIRES;
509 }
510 dst_release(&rt->u.dst);
511 }
512 return 0;
513 }
514 #endif
515
516 #define BACKTRACK(saddr) \
517 do { \
518 if (rt == &ip6_null_entry) { \
519 struct fib6_node *pn; \
520 while (1) { \
521 if (fn->fn_flags & RTN_TL_ROOT) \
522 goto out; \
523 pn = fn->parent; \
524 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
525 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
526 else \
527 fn = pn; \
528 if (fn->fn_flags & RTN_RTINFO) \
529 goto restart; \
530 } \
531 } \
532 } while(0)
533
534 static struct rt6_info *ip6_pol_route_lookup(struct fib6_table *table,
535 struct flowi *fl, int flags)
536 {
537 struct fib6_node *fn;
538 struct rt6_info *rt;
539
540 read_lock_bh(&table->tb6_lock);
541 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
542 restart:
543 rt = fn->leaf;
544 rt = rt6_device_match(rt, fl->oif, flags);
545 BACKTRACK(&fl->fl6_src);
546 out:
547 dst_hold(&rt->u.dst);
548 read_unlock_bh(&table->tb6_lock);
549
550 rt->u.dst.lastuse = jiffies;
551 rt->u.dst.__use++;
552
553 return rt;
554
555 }
556
557 struct rt6_info *rt6_lookup(struct in6_addr *daddr, struct in6_addr *saddr,
558 int oif, int strict)
559 {
560 struct flowi fl = {
561 .oif = oif,
562 .nl_u = {
563 .ip6_u = {
564 .daddr = *daddr,
565 },
566 },
567 };
568 struct dst_entry *dst;
569 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
570
571 if (saddr) {
572 memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
573 flags |= RT6_LOOKUP_F_HAS_SADDR;
574 }
575
576 dst = fib6_rule_lookup(&fl, flags, ip6_pol_route_lookup);
577 if (dst->error == 0)
578 return (struct rt6_info *) dst;
579
580 dst_release(dst);
581
582 return NULL;
583 }
584
585 EXPORT_SYMBOL(rt6_lookup);
586
587 /* ip6_ins_rt is called with FREE table->tb6_lock.
588 It takes new route entry, the addition fails by any reason the
589 route is freed. In any case, if caller does not hold it, it may
590 be destroyed.
591 */
592
593 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
594 {
595 int err;
596 struct fib6_table *table;
597
598 table = rt->rt6i_table;
599 write_lock_bh(&table->tb6_lock);
600 err = fib6_add(&table->tb6_root, rt, info);
601 write_unlock_bh(&table->tb6_lock);
602
603 return err;
604 }
605
606 int ip6_ins_rt(struct rt6_info *rt)
607 {
608 return __ip6_ins_rt(rt, NULL);
609 }
610
611 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
612 struct in6_addr *saddr)
613 {
614 struct rt6_info *rt;
615
616 /*
617 * Clone the route.
618 */
619
620 rt = ip6_rt_copy(ort);
621
622 if (rt) {
623 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
624 if (rt->rt6i_dst.plen != 128 &&
625 ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
626 rt->rt6i_flags |= RTF_ANYCAST;
627 ipv6_addr_copy(&rt->rt6i_gateway, daddr);
628 }
629
630 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
631 rt->rt6i_dst.plen = 128;
632 rt->rt6i_flags |= RTF_CACHE;
633 rt->u.dst.flags |= DST_HOST;
634
635 #ifdef CONFIG_IPV6_SUBTREES
636 if (rt->rt6i_src.plen && saddr) {
637 ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
638 rt->rt6i_src.plen = 128;
639 }
640 #endif
641
642 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
643
644 }
645
646 return rt;
647 }
648
649 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
650 {
651 struct rt6_info *rt = ip6_rt_copy(ort);
652 if (rt) {
653 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
654 rt->rt6i_dst.plen = 128;
655 rt->rt6i_flags |= RTF_CACHE;
656 rt->u.dst.flags |= DST_HOST;
657 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
658 }
659 return rt;
660 }
661
662 static struct rt6_info *ip6_pol_route(struct fib6_table *table, int oif,
663 struct flowi *fl, int flags)
664 {
665 struct fib6_node *fn;
666 struct rt6_info *rt, *nrt;
667 int strict = 0;
668 int attempts = 3;
669 int err;
670 int reachable = ipv6_devconf.forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
671
672 strict |= flags & RT6_LOOKUP_F_IFACE;
673
674 relookup:
675 read_lock_bh(&table->tb6_lock);
676
677 restart_2:
678 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
679
680 restart:
681 rt = rt6_select(fn, oif, strict | reachable);
682 BACKTRACK(&fl->fl6_src);
683 if (rt == &ip6_null_entry ||
684 rt->rt6i_flags & RTF_CACHE)
685 goto out;
686
687 dst_hold(&rt->u.dst);
688 read_unlock_bh(&table->tb6_lock);
689
690 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
691 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
692 else {
693 #if CLONE_OFFLINK_ROUTE
694 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
695 #else
696 goto out2;
697 #endif
698 }
699
700 dst_release(&rt->u.dst);
701 rt = nrt ? : &ip6_null_entry;
702
703 dst_hold(&rt->u.dst);
704 if (nrt) {
705 err = ip6_ins_rt(nrt);
706 if (!err)
707 goto out2;
708 }
709
710 if (--attempts <= 0)
711 goto out2;
712
713 /*
714 * Race condition! In the gap, when table->tb6_lock was
715 * released someone could insert this route. Relookup.
716 */
717 dst_release(&rt->u.dst);
718 goto relookup;
719
720 out:
721 if (reachable) {
722 reachable = 0;
723 goto restart_2;
724 }
725 dst_hold(&rt->u.dst);
726 read_unlock_bh(&table->tb6_lock);
727 out2:
728 rt->u.dst.lastuse = jiffies;
729 rt->u.dst.__use++;
730
731 return rt;
732 }
733
734 static struct rt6_info *ip6_pol_route_input(struct fib6_table *table,
735 struct flowi *fl, int flags)
736 {
737 return ip6_pol_route(table, fl->iif, fl, flags);
738 }
739
740 void ip6_route_input(struct sk_buff *skb)
741 {
742 struct ipv6hdr *iph = ipv6_hdr(skb);
743 int flags = RT6_LOOKUP_F_HAS_SADDR;
744 struct flowi fl = {
745 .iif = skb->dev->ifindex,
746 .nl_u = {
747 .ip6_u = {
748 .daddr = iph->daddr,
749 .saddr = iph->saddr,
750 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
751 },
752 },
753 .mark = skb->mark,
754 .proto = iph->nexthdr,
755 };
756
757 if (rt6_need_strict(&iph->daddr))
758 flags |= RT6_LOOKUP_F_IFACE;
759
760 skb->dst = fib6_rule_lookup(&fl, flags, ip6_pol_route_input);
761 }
762
763 static struct rt6_info *ip6_pol_route_output(struct fib6_table *table,
764 struct flowi *fl, int flags)
765 {
766 return ip6_pol_route(table, fl->oif, fl, flags);
767 }
768
769 struct dst_entry * ip6_route_output(struct sock *sk, struct flowi *fl)
770 {
771 int flags = 0;
772
773 if (rt6_need_strict(&fl->fl6_dst))
774 flags |= RT6_LOOKUP_F_IFACE;
775
776 if (!ipv6_addr_any(&fl->fl6_src))
777 flags |= RT6_LOOKUP_F_HAS_SADDR;
778
779 return fib6_rule_lookup(fl, flags, ip6_pol_route_output);
780 }
781
782 EXPORT_SYMBOL(ip6_route_output);
783
784 static int ip6_blackhole_output(struct sk_buff *skb)
785 {
786 kfree_skb(skb);
787 return 0;
788 }
789
790 int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
791 {
792 struct rt6_info *ort = (struct rt6_info *) *dstp;
793 struct rt6_info *rt = (struct rt6_info *)
794 dst_alloc(&ip6_dst_blackhole_ops);
795 struct dst_entry *new = NULL;
796
797 if (rt) {
798 new = &rt->u.dst;
799
800 atomic_set(&new->__refcnt, 1);
801 new->__use = 1;
802 new->input = ip6_blackhole_output;
803 new->output = ip6_blackhole_output;
804
805 memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
806 new->dev = ort->u.dst.dev;
807 if (new->dev)
808 dev_hold(new->dev);
809 rt->rt6i_idev = ort->rt6i_idev;
810 if (rt->rt6i_idev)
811 in6_dev_hold(rt->rt6i_idev);
812 rt->rt6i_expires = 0;
813
814 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
815 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
816 rt->rt6i_metric = 0;
817
818 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
819 #ifdef CONFIG_IPV6_SUBTREES
820 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
821 #endif
822
823 dst_free(new);
824 }
825
826 dst_release(*dstp);
827 *dstp = new;
828 return (new ? 0 : -ENOMEM);
829 }
830 EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
831
832 /*
833 * Destination cache support functions
834 */
835
836 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
837 {
838 struct rt6_info *rt;
839
840 rt = (struct rt6_info *) dst;
841
842 if (rt && rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
843 return dst;
844
845 return NULL;
846 }
847
848 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
849 {
850 struct rt6_info *rt = (struct rt6_info *) dst;
851
852 if (rt) {
853 if (rt->rt6i_flags & RTF_CACHE)
854 ip6_del_rt(rt);
855 else
856 dst_release(dst);
857 }
858 return NULL;
859 }
860
861 static void ip6_link_failure(struct sk_buff *skb)
862 {
863 struct rt6_info *rt;
864
865 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
866
867 rt = (struct rt6_info *) skb->dst;
868 if (rt) {
869 if (rt->rt6i_flags&RTF_CACHE) {
870 dst_set_expires(&rt->u.dst, 0);
871 rt->rt6i_flags |= RTF_EXPIRES;
872 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
873 rt->rt6i_node->fn_sernum = -1;
874 }
875 }
876
877 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
878 {
879 struct rt6_info *rt6 = (struct rt6_info*)dst;
880
881 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
882 rt6->rt6i_flags |= RTF_MODIFIED;
883 if (mtu < IPV6_MIN_MTU) {
884 mtu = IPV6_MIN_MTU;
885 dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
886 }
887 dst->metrics[RTAX_MTU-1] = mtu;
888 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
889 }
890 }
891
892 static int ipv6_get_mtu(struct net_device *dev);
893
894 static inline unsigned int ipv6_advmss(unsigned int mtu)
895 {
896 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
897
898 if (mtu < ip6_rt_min_advmss)
899 mtu = ip6_rt_min_advmss;
900
901 /*
902 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
903 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
904 * IPV6_MAXPLEN is also valid and means: "any MSS,
905 * rely only on pmtu discovery"
906 */
907 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
908 mtu = IPV6_MAXPLEN;
909 return mtu;
910 }
911
912 static struct dst_entry *ndisc_dst_gc_list;
913 static DEFINE_SPINLOCK(ndisc_lock);
914
915 struct dst_entry *ndisc_dst_alloc(struct net_device *dev,
916 struct neighbour *neigh,
917 struct in6_addr *addr,
918 int (*output)(struct sk_buff *))
919 {
920 struct rt6_info *rt;
921 struct inet6_dev *idev = in6_dev_get(dev);
922
923 if (unlikely(idev == NULL))
924 return NULL;
925
926 rt = ip6_dst_alloc();
927 if (unlikely(rt == NULL)) {
928 in6_dev_put(idev);
929 goto out;
930 }
931
932 dev_hold(dev);
933 if (neigh)
934 neigh_hold(neigh);
935 else
936 neigh = ndisc_get_neigh(dev, addr);
937
938 rt->rt6i_dev = dev;
939 rt->rt6i_idev = idev;
940 rt->rt6i_nexthop = neigh;
941 atomic_set(&rt->u.dst.__refcnt, 1);
942 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
943 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
944 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
945 rt->u.dst.output = output;
946
947 #if 0 /* there's no chance to use these for ndisc */
948 rt->u.dst.flags = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
949 ? DST_HOST
950 : 0;
951 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
952 rt->rt6i_dst.plen = 128;
953 #endif
954
955 spin_lock_bh(&ndisc_lock);
956 rt->u.dst.next = ndisc_dst_gc_list;
957 ndisc_dst_gc_list = &rt->u.dst;
958 spin_unlock_bh(&ndisc_lock);
959
960 fib6_force_start_gc();
961
962 out:
963 return &rt->u.dst;
964 }
965
966 int ndisc_dst_gc(int *more)
967 {
968 struct dst_entry *dst, *next, **pprev;
969 int freed;
970
971 next = NULL;
972 freed = 0;
973
974 spin_lock_bh(&ndisc_lock);
975 pprev = &ndisc_dst_gc_list;
976
977 while ((dst = *pprev) != NULL) {
978 if (!atomic_read(&dst->__refcnt)) {
979 *pprev = dst->next;
980 dst_free(dst);
981 freed++;
982 } else {
983 pprev = &dst->next;
984 (*more)++;
985 }
986 }
987
988 spin_unlock_bh(&ndisc_lock);
989
990 return freed;
991 }
992
993 static int ip6_dst_gc(void)
994 {
995 static unsigned expire = 30*HZ;
996 static unsigned long last_gc;
997 unsigned long now = jiffies;
998
999 if (time_after(last_gc + ip6_rt_gc_min_interval, now) &&
1000 atomic_read(&ip6_dst_ops.entries) <= ip6_rt_max_size)
1001 goto out;
1002
1003 expire++;
1004 fib6_run_gc(expire);
1005 last_gc = now;
1006 if (atomic_read(&ip6_dst_ops.entries) < ip6_dst_ops.gc_thresh)
1007 expire = ip6_rt_gc_timeout>>1;
1008
1009 out:
1010 expire -= expire>>ip6_rt_gc_elasticity;
1011 return (atomic_read(&ip6_dst_ops.entries) > ip6_rt_max_size);
1012 }
1013
1014 /* Clean host part of a prefix. Not necessary in radix tree,
1015 but results in cleaner routing tables.
1016
1017 Remove it only when all the things will work!
1018 */
1019
1020 static int ipv6_get_mtu(struct net_device *dev)
1021 {
1022 int mtu = IPV6_MIN_MTU;
1023 struct inet6_dev *idev;
1024
1025 idev = in6_dev_get(dev);
1026 if (idev) {
1027 mtu = idev->cnf.mtu6;
1028 in6_dev_put(idev);
1029 }
1030 return mtu;
1031 }
1032
1033 int ipv6_get_hoplimit(struct net_device *dev)
1034 {
1035 int hoplimit = ipv6_devconf.hop_limit;
1036 struct inet6_dev *idev;
1037
1038 idev = in6_dev_get(dev);
1039 if (idev) {
1040 hoplimit = idev->cnf.hop_limit;
1041 in6_dev_put(idev);
1042 }
1043 return hoplimit;
1044 }
1045
1046 /*
1047 *
1048 */
1049
1050 int ip6_route_add(struct fib6_config *cfg)
1051 {
1052 int err;
1053 struct rt6_info *rt = NULL;
1054 struct net_device *dev = NULL;
1055 struct inet6_dev *idev = NULL;
1056 struct fib6_table *table;
1057 int addr_type;
1058
1059 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1060 return -EINVAL;
1061 #ifndef CONFIG_IPV6_SUBTREES
1062 if (cfg->fc_src_len)
1063 return -EINVAL;
1064 #endif
1065 if (cfg->fc_ifindex) {
1066 err = -ENODEV;
1067 dev = dev_get_by_index(&init_net, cfg->fc_ifindex);
1068 if (!dev)
1069 goto out;
1070 idev = in6_dev_get(dev);
1071 if (!idev)
1072 goto out;
1073 }
1074
1075 if (cfg->fc_metric == 0)
1076 cfg->fc_metric = IP6_RT_PRIO_USER;
1077
1078 table = fib6_new_table(cfg->fc_table);
1079 if (table == NULL) {
1080 err = -ENOBUFS;
1081 goto out;
1082 }
1083
1084 rt = ip6_dst_alloc();
1085
1086 if (rt == NULL) {
1087 err = -ENOMEM;
1088 goto out;
1089 }
1090
1091 rt->u.dst.obsolete = -1;
1092 rt->rt6i_expires = jiffies + clock_t_to_jiffies(cfg->fc_expires);
1093
1094 if (cfg->fc_protocol == RTPROT_UNSPEC)
1095 cfg->fc_protocol = RTPROT_BOOT;
1096 rt->rt6i_protocol = cfg->fc_protocol;
1097
1098 addr_type = ipv6_addr_type(&cfg->fc_dst);
1099
1100 if (addr_type & IPV6_ADDR_MULTICAST)
1101 rt->u.dst.input = ip6_mc_input;
1102 else
1103 rt->u.dst.input = ip6_forward;
1104
1105 rt->u.dst.output = ip6_output;
1106
1107 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1108 rt->rt6i_dst.plen = cfg->fc_dst_len;
1109 if (rt->rt6i_dst.plen == 128)
1110 rt->u.dst.flags = DST_HOST;
1111
1112 #ifdef CONFIG_IPV6_SUBTREES
1113 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1114 rt->rt6i_src.plen = cfg->fc_src_len;
1115 #endif
1116
1117 rt->rt6i_metric = cfg->fc_metric;
1118
1119 /* We cannot add true routes via loopback here,
1120 they would result in kernel looping; promote them to reject routes
1121 */
1122 if ((cfg->fc_flags & RTF_REJECT) ||
1123 (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1124 /* hold loopback dev/idev if we haven't done so. */
1125 if (dev != init_net.loopback_dev) {
1126 if (dev) {
1127 dev_put(dev);
1128 in6_dev_put(idev);
1129 }
1130 dev = init_net.loopback_dev;
1131 dev_hold(dev);
1132 idev = in6_dev_get(dev);
1133 if (!idev) {
1134 err = -ENODEV;
1135 goto out;
1136 }
1137 }
1138 rt->u.dst.output = ip6_pkt_discard_out;
1139 rt->u.dst.input = ip6_pkt_discard;
1140 rt->u.dst.error = -ENETUNREACH;
1141 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1142 goto install_route;
1143 }
1144
1145 if (cfg->fc_flags & RTF_GATEWAY) {
1146 struct in6_addr *gw_addr;
1147 int gwa_type;
1148
1149 gw_addr = &cfg->fc_gateway;
1150 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1151 gwa_type = ipv6_addr_type(gw_addr);
1152
1153 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1154 struct rt6_info *grt;
1155
1156 /* IPv6 strictly inhibits using not link-local
1157 addresses as nexthop address.
1158 Otherwise, router will not able to send redirects.
1159 It is very good, but in some (rare!) circumstances
1160 (SIT, PtP, NBMA NOARP links) it is handy to allow
1161 some exceptions. --ANK
1162 */
1163 err = -EINVAL;
1164 if (!(gwa_type&IPV6_ADDR_UNICAST))
1165 goto out;
1166
1167 grt = rt6_lookup(gw_addr, NULL, cfg->fc_ifindex, 1);
1168
1169 err = -EHOSTUNREACH;
1170 if (grt == NULL)
1171 goto out;
1172 if (dev) {
1173 if (dev != grt->rt6i_dev) {
1174 dst_release(&grt->u.dst);
1175 goto out;
1176 }
1177 } else {
1178 dev = grt->rt6i_dev;
1179 idev = grt->rt6i_idev;
1180 dev_hold(dev);
1181 in6_dev_hold(grt->rt6i_idev);
1182 }
1183 if (!(grt->rt6i_flags&RTF_GATEWAY))
1184 err = 0;
1185 dst_release(&grt->u.dst);
1186
1187 if (err)
1188 goto out;
1189 }
1190 err = -EINVAL;
1191 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1192 goto out;
1193 }
1194
1195 err = -ENODEV;
1196 if (dev == NULL)
1197 goto out;
1198
1199 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1200 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1201 if (IS_ERR(rt->rt6i_nexthop)) {
1202 err = PTR_ERR(rt->rt6i_nexthop);
1203 rt->rt6i_nexthop = NULL;
1204 goto out;
1205 }
1206 }
1207
1208 rt->rt6i_flags = cfg->fc_flags;
1209
1210 install_route:
1211 if (cfg->fc_mx) {
1212 struct nlattr *nla;
1213 int remaining;
1214
1215 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1216 int type = nla_type(nla);
1217
1218 if (type) {
1219 if (type > RTAX_MAX) {
1220 err = -EINVAL;
1221 goto out;
1222 }
1223
1224 rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1225 }
1226 }
1227 }
1228
1229 if (rt->u.dst.metrics[RTAX_HOPLIMIT-1] == 0)
1230 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1231 if (!rt->u.dst.metrics[RTAX_MTU-1])
1232 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1233 if (!rt->u.dst.metrics[RTAX_ADVMSS-1])
1234 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
1235 rt->u.dst.dev = dev;
1236 rt->rt6i_idev = idev;
1237 rt->rt6i_table = table;
1238 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1239
1240 out:
1241 if (dev)
1242 dev_put(dev);
1243 if (idev)
1244 in6_dev_put(idev);
1245 if (rt)
1246 dst_free(&rt->u.dst);
1247 return err;
1248 }
1249
1250 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1251 {
1252 int err;
1253 struct fib6_table *table;
1254
1255 if (rt == &ip6_null_entry)
1256 return -ENOENT;
1257
1258 table = rt->rt6i_table;
1259 write_lock_bh(&table->tb6_lock);
1260
1261 err = fib6_del(rt, info);
1262 dst_release(&rt->u.dst);
1263
1264 write_unlock_bh(&table->tb6_lock);
1265
1266 return err;
1267 }
1268
1269 int ip6_del_rt(struct rt6_info *rt)
1270 {
1271 return __ip6_del_rt(rt, NULL);
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 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1668 struct net_device *dev,
1669 unsigned int pref)
1670 {
1671 struct fib6_config cfg = {
1672 .fc_table = RT6_TABLE_DFLT,
1673 .fc_metric = 1024,
1674 .fc_ifindex = dev->ifindex,
1675 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1676 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1677 };
1678
1679 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1680
1681 ip6_route_add(&cfg);
1682
1683 return rt6_get_dflt_router(gwaddr, dev);
1684 }
1685
1686 void rt6_purge_dflt_routers(void)
1687 {
1688 struct rt6_info *rt;
1689 struct fib6_table *table;
1690
1691 /* NOTE: Keep consistent with rt6_get_dflt_router */
1692 table = fib6_get_table(RT6_TABLE_DFLT);
1693 if (table == NULL)
1694 return;
1695
1696 restart:
1697 read_lock_bh(&table->tb6_lock);
1698 for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1699 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1700 dst_hold(&rt->u.dst);
1701 read_unlock_bh(&table->tb6_lock);
1702 ip6_del_rt(rt);
1703 goto restart;
1704 }
1705 }
1706 read_unlock_bh(&table->tb6_lock);
1707 }
1708
1709 static void rtmsg_to_fib6_config(struct in6_rtmsg *rtmsg,
1710 struct fib6_config *cfg)
1711 {
1712 memset(cfg, 0, sizeof(*cfg));
1713
1714 cfg->fc_table = RT6_TABLE_MAIN;
1715 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1716 cfg->fc_metric = rtmsg->rtmsg_metric;
1717 cfg->fc_expires = rtmsg->rtmsg_info;
1718 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1719 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1720 cfg->fc_flags = rtmsg->rtmsg_flags;
1721
1722 ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1723 ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1724 ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1725 }
1726
1727 int ipv6_route_ioctl(unsigned int cmd, void __user *arg)
1728 {
1729 struct fib6_config cfg;
1730 struct in6_rtmsg rtmsg;
1731 int err;
1732
1733 switch(cmd) {
1734 case SIOCADDRT: /* Add a route */
1735 case SIOCDELRT: /* Delete a route */
1736 if (!capable(CAP_NET_ADMIN))
1737 return -EPERM;
1738 err = copy_from_user(&rtmsg, arg,
1739 sizeof(struct in6_rtmsg));
1740 if (err)
1741 return -EFAULT;
1742
1743 rtmsg_to_fib6_config(&rtmsg, &cfg);
1744
1745 rtnl_lock();
1746 switch (cmd) {
1747 case SIOCADDRT:
1748 err = ip6_route_add(&cfg);
1749 break;
1750 case SIOCDELRT:
1751 err = ip6_route_del(&cfg);
1752 break;
1753 default:
1754 err = -EINVAL;
1755 }
1756 rtnl_unlock();
1757
1758 return err;
1759 }
1760
1761 return -EINVAL;
1762 }
1763
1764 /*
1765 * Drop the packet on the floor
1766 */
1767
1768 static inline int ip6_pkt_drop(struct sk_buff *skb, int code,
1769 int ipstats_mib_noroutes)
1770 {
1771 int type;
1772 switch (ipstats_mib_noroutes) {
1773 case IPSTATS_MIB_INNOROUTES:
1774 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1775 if (type == IPV6_ADDR_ANY || type == IPV6_ADDR_RESERVED) {
1776 IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_INADDRERRORS);
1777 break;
1778 }
1779 /* FALLTHROUGH */
1780 case IPSTATS_MIB_OUTNOROUTES:
1781 IP6_INC_STATS(ip6_dst_idev(skb->dst), ipstats_mib_noroutes);
1782 break;
1783 }
1784 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0, skb->dev);
1785 kfree_skb(skb);
1786 return 0;
1787 }
1788
1789 static int ip6_pkt_discard(struct sk_buff *skb)
1790 {
1791 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1792 }
1793
1794 static int ip6_pkt_discard_out(struct sk_buff *skb)
1795 {
1796 skb->dev = skb->dst->dev;
1797 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1798 }
1799
1800 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1801
1802 static int ip6_pkt_prohibit(struct sk_buff *skb)
1803 {
1804 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1805 }
1806
1807 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1808 {
1809 skb->dev = skb->dst->dev;
1810 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1811 }
1812
1813 static int ip6_pkt_blk_hole(struct sk_buff *skb)
1814 {
1815 kfree_skb(skb);
1816 return 0;
1817 }
1818
1819 #endif
1820
1821 /*
1822 * Allocate a dst for local (unicast / anycast) address.
1823 */
1824
1825 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1826 const struct in6_addr *addr,
1827 int anycast)
1828 {
1829 struct rt6_info *rt = ip6_dst_alloc();
1830
1831 if (rt == NULL)
1832 return ERR_PTR(-ENOMEM);
1833
1834 dev_hold(init_net.loopback_dev);
1835 in6_dev_hold(idev);
1836
1837 rt->u.dst.flags = DST_HOST;
1838 rt->u.dst.input = ip6_input;
1839 rt->u.dst.output = ip6_output;
1840 rt->rt6i_dev = init_net.loopback_dev;
1841 rt->rt6i_idev = idev;
1842 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1843 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
1844 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1845 rt->u.dst.obsolete = -1;
1846
1847 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1848 if (anycast)
1849 rt->rt6i_flags |= RTF_ANYCAST;
1850 else
1851 rt->rt6i_flags |= RTF_LOCAL;
1852 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1853 if (rt->rt6i_nexthop == NULL) {
1854 dst_free(&rt->u.dst);
1855 return ERR_PTR(-ENOMEM);
1856 }
1857
1858 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1859 rt->rt6i_dst.plen = 128;
1860 rt->rt6i_table = fib6_get_table(RT6_TABLE_LOCAL);
1861
1862 atomic_set(&rt->u.dst.__refcnt, 1);
1863
1864 return rt;
1865 }
1866
1867 static int fib6_ifdown(struct rt6_info *rt, void *arg)
1868 {
1869 if (((void*)rt->rt6i_dev == arg || arg == NULL) &&
1870 rt != &ip6_null_entry) {
1871 RT6_TRACE("deleted by ifdown %p\n", rt);
1872 return -1;
1873 }
1874 return 0;
1875 }
1876
1877 void rt6_ifdown(struct net_device *dev)
1878 {
1879 fib6_clean_all(fib6_ifdown, 0, dev);
1880 }
1881
1882 struct rt6_mtu_change_arg
1883 {
1884 struct net_device *dev;
1885 unsigned mtu;
1886 };
1887
1888 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
1889 {
1890 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
1891 struct inet6_dev *idev;
1892
1893 /* In IPv6 pmtu discovery is not optional,
1894 so that RTAX_MTU lock cannot disable it.
1895 We still use this lock to block changes
1896 caused by addrconf/ndisc.
1897 */
1898
1899 idev = __in6_dev_get(arg->dev);
1900 if (idev == NULL)
1901 return 0;
1902
1903 /* For administrative MTU increase, there is no way to discover
1904 IPv6 PMTU increase, so PMTU increase should be updated here.
1905 Since RFC 1981 doesn't include administrative MTU increase
1906 update PMTU increase is a MUST. (i.e. jumbo frame)
1907 */
1908 /*
1909 If new MTU is less than route PMTU, this new MTU will be the
1910 lowest MTU in the path, update the route PMTU to reflect PMTU
1911 decreases; if new MTU is greater than route PMTU, and the
1912 old MTU is the lowest MTU in the path, update the route PMTU
1913 to reflect the increase. In this case if the other nodes' MTU
1914 also have the lowest MTU, TOO BIG MESSAGE will be lead to
1915 PMTU discouvery.
1916 */
1917 if (rt->rt6i_dev == arg->dev &&
1918 !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
1919 (dst_mtu(&rt->u.dst) > arg->mtu ||
1920 (dst_mtu(&rt->u.dst) < arg->mtu &&
1921 dst_mtu(&rt->u.dst) == idev->cnf.mtu6))) {
1922 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
1923 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(arg->mtu);
1924 }
1925 return 0;
1926 }
1927
1928 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
1929 {
1930 struct rt6_mtu_change_arg arg = {
1931 .dev = dev,
1932 .mtu = mtu,
1933 };
1934
1935 fib6_clean_all(rt6_mtu_change_route, 0, &arg);
1936 }
1937
1938 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
1939 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
1940 [RTA_OIF] = { .type = NLA_U32 },
1941 [RTA_IIF] = { .type = NLA_U32 },
1942 [RTA_PRIORITY] = { .type = NLA_U32 },
1943 [RTA_METRICS] = { .type = NLA_NESTED },
1944 };
1945
1946 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
1947 struct fib6_config *cfg)
1948 {
1949 struct rtmsg *rtm;
1950 struct nlattr *tb[RTA_MAX+1];
1951 int err;
1952
1953 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
1954 if (err < 0)
1955 goto errout;
1956
1957 err = -EINVAL;
1958 rtm = nlmsg_data(nlh);
1959 memset(cfg, 0, sizeof(*cfg));
1960
1961 cfg->fc_table = rtm->rtm_table;
1962 cfg->fc_dst_len = rtm->rtm_dst_len;
1963 cfg->fc_src_len = rtm->rtm_src_len;
1964 cfg->fc_flags = RTF_UP;
1965 cfg->fc_protocol = rtm->rtm_protocol;
1966
1967 if (rtm->rtm_type == RTN_UNREACHABLE)
1968 cfg->fc_flags |= RTF_REJECT;
1969
1970 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
1971 cfg->fc_nlinfo.nlh = nlh;
1972
1973 if (tb[RTA_GATEWAY]) {
1974 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
1975 cfg->fc_flags |= RTF_GATEWAY;
1976 }
1977
1978 if (tb[RTA_DST]) {
1979 int plen = (rtm->rtm_dst_len + 7) >> 3;
1980
1981 if (nla_len(tb[RTA_DST]) < plen)
1982 goto errout;
1983
1984 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1985 }
1986
1987 if (tb[RTA_SRC]) {
1988 int plen = (rtm->rtm_src_len + 7) >> 3;
1989
1990 if (nla_len(tb[RTA_SRC]) < plen)
1991 goto errout;
1992
1993 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1994 }
1995
1996 if (tb[RTA_OIF])
1997 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
1998
1999 if (tb[RTA_PRIORITY])
2000 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2001
2002 if (tb[RTA_METRICS]) {
2003 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2004 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2005 }
2006
2007 if (tb[RTA_TABLE])
2008 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2009
2010 err = 0;
2011 errout:
2012 return err;
2013 }
2014
2015 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2016 {
2017 struct fib6_config cfg;
2018 int err;
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 fib6_config cfg;
2030 int err;
2031
2032 err = rtm_to_fib6_config(skb, nlh, &cfg);
2033 if (err < 0)
2034 return err;
2035
2036 return ip6_route_add(&cfg);
2037 }
2038
2039 static inline size_t rt6_nlmsg_size(void)
2040 {
2041 return NLMSG_ALIGN(sizeof(struct rtmsg))
2042 + nla_total_size(16) /* RTA_SRC */
2043 + nla_total_size(16) /* RTA_DST */
2044 + nla_total_size(16) /* RTA_GATEWAY */
2045 + nla_total_size(16) /* RTA_PREFSRC */
2046 + nla_total_size(4) /* RTA_TABLE */
2047 + nla_total_size(4) /* RTA_IIF */
2048 + nla_total_size(4) /* RTA_OIF */
2049 + nla_total_size(4) /* RTA_PRIORITY */
2050 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2051 + nla_total_size(sizeof(struct rta_cacheinfo));
2052 }
2053
2054 static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
2055 struct in6_addr *dst, struct in6_addr *src,
2056 int iif, int type, u32 pid, u32 seq,
2057 int prefix, unsigned int flags)
2058 {
2059 struct rtmsg *rtm;
2060 struct nlmsghdr *nlh;
2061 long expires;
2062 u32 table;
2063
2064 if (prefix) { /* user wants prefix routes only */
2065 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2066 /* success since this is not a prefix route */
2067 return 1;
2068 }
2069 }
2070
2071 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2072 if (nlh == NULL)
2073 return -EMSGSIZE;
2074
2075 rtm = nlmsg_data(nlh);
2076 rtm->rtm_family = AF_INET6;
2077 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2078 rtm->rtm_src_len = rt->rt6i_src.plen;
2079 rtm->rtm_tos = 0;
2080 if (rt->rt6i_table)
2081 table = rt->rt6i_table->tb6_id;
2082 else
2083 table = RT6_TABLE_UNSPEC;
2084 rtm->rtm_table = table;
2085 NLA_PUT_U32(skb, RTA_TABLE, table);
2086 if (rt->rt6i_flags&RTF_REJECT)
2087 rtm->rtm_type = RTN_UNREACHABLE;
2088 else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2089 rtm->rtm_type = RTN_LOCAL;
2090 else
2091 rtm->rtm_type = RTN_UNICAST;
2092 rtm->rtm_flags = 0;
2093 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2094 rtm->rtm_protocol = rt->rt6i_protocol;
2095 if (rt->rt6i_flags&RTF_DYNAMIC)
2096 rtm->rtm_protocol = RTPROT_REDIRECT;
2097 else if (rt->rt6i_flags & RTF_ADDRCONF)
2098 rtm->rtm_protocol = RTPROT_KERNEL;
2099 else if (rt->rt6i_flags&RTF_DEFAULT)
2100 rtm->rtm_protocol = RTPROT_RA;
2101
2102 if (rt->rt6i_flags&RTF_CACHE)
2103 rtm->rtm_flags |= RTM_F_CLONED;
2104
2105 if (dst) {
2106 NLA_PUT(skb, RTA_DST, 16, dst);
2107 rtm->rtm_dst_len = 128;
2108 } else if (rtm->rtm_dst_len)
2109 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2110 #ifdef CONFIG_IPV6_SUBTREES
2111 if (src) {
2112 NLA_PUT(skb, RTA_SRC, 16, src);
2113 rtm->rtm_src_len = 128;
2114 } else if (rtm->rtm_src_len)
2115 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2116 #endif
2117 if (iif)
2118 NLA_PUT_U32(skb, RTA_IIF, iif);
2119 else if (dst) {
2120 struct in6_addr saddr_buf;
2121 if (ipv6_get_saddr(&rt->u.dst, dst, &saddr_buf) == 0)
2122 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2123 }
2124
2125 if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2126 goto nla_put_failure;
2127
2128 if (rt->u.dst.neighbour)
2129 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2130
2131 if (rt->u.dst.dev)
2132 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2133
2134 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2135
2136 expires = rt->rt6i_expires ? rt->rt6i_expires - jiffies : 0;
2137 if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
2138 expires, rt->u.dst.error) < 0)
2139 goto nla_put_failure;
2140
2141 return nlmsg_end(skb, nlh);
2142
2143 nla_put_failure:
2144 nlmsg_cancel(skb, nlh);
2145 return -EMSGSIZE;
2146 }
2147
2148 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2149 {
2150 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2151 int prefix;
2152
2153 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2154 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2155 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2156 } else
2157 prefix = 0;
2158
2159 return rt6_fill_node(arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2160 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2161 prefix, NLM_F_MULTI);
2162 }
2163
2164 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2165 {
2166 struct nlattr *tb[RTA_MAX+1];
2167 struct rt6_info *rt;
2168 struct sk_buff *skb;
2169 struct rtmsg *rtm;
2170 struct flowi fl;
2171 int err, iif = 0;
2172
2173 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2174 if (err < 0)
2175 goto errout;
2176
2177 err = -EINVAL;
2178 memset(&fl, 0, sizeof(fl));
2179
2180 if (tb[RTA_SRC]) {
2181 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2182 goto errout;
2183
2184 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2185 }
2186
2187 if (tb[RTA_DST]) {
2188 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2189 goto errout;
2190
2191 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2192 }
2193
2194 if (tb[RTA_IIF])
2195 iif = nla_get_u32(tb[RTA_IIF]);
2196
2197 if (tb[RTA_OIF])
2198 fl.oif = nla_get_u32(tb[RTA_OIF]);
2199
2200 if (iif) {
2201 struct net_device *dev;
2202 dev = __dev_get_by_index(&init_net, iif);
2203 if (!dev) {
2204 err = -ENODEV;
2205 goto errout;
2206 }
2207 }
2208
2209 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2210 if (skb == NULL) {
2211 err = -ENOBUFS;
2212 goto errout;
2213 }
2214
2215 /* Reserve room for dummy headers, this skb can pass
2216 through good chunk of routing engine.
2217 */
2218 skb_reset_mac_header(skb);
2219 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2220
2221 rt = (struct rt6_info*) ip6_route_output(NULL, &fl);
2222 skb->dst = &rt->u.dst;
2223
2224 err = rt6_fill_node(skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2225 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2226 nlh->nlmsg_seq, 0, 0);
2227 if (err < 0) {
2228 kfree_skb(skb);
2229 goto errout;
2230 }
2231
2232 err = rtnl_unicast(skb, NETLINK_CB(in_skb).pid);
2233 errout:
2234 return err;
2235 }
2236
2237 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2238 {
2239 struct sk_buff *skb;
2240 u32 pid = 0, seq = 0;
2241 struct nlmsghdr *nlh = NULL;
2242 int err = -ENOBUFS;
2243
2244 if (info) {
2245 pid = info->pid;
2246 nlh = info->nlh;
2247 if (nlh)
2248 seq = nlh->nlmsg_seq;
2249 }
2250
2251 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2252 if (skb == NULL)
2253 goto errout;
2254
2255 err = rt6_fill_node(skb, rt, NULL, NULL, 0, event, pid, seq, 0, 0);
2256 if (err < 0) {
2257 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2258 WARN_ON(err == -EMSGSIZE);
2259 kfree_skb(skb);
2260 goto errout;
2261 }
2262 err = rtnl_notify(skb, pid, RTNLGRP_IPV6_ROUTE, nlh, gfp_any());
2263 errout:
2264 if (err < 0)
2265 rtnl_set_sk_err(RTNLGRP_IPV6_ROUTE, err);
2266 }
2267
2268 /*
2269 * /proc
2270 */
2271
2272 #ifdef CONFIG_PROC_FS
2273
2274 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2275
2276 struct rt6_proc_arg
2277 {
2278 char *buffer;
2279 int offset;
2280 int length;
2281 int skip;
2282 int len;
2283 };
2284
2285 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2286 {
2287 struct seq_file *m = p_arg;
2288
2289 seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_dst.addr),
2290 rt->rt6i_dst.plen);
2291
2292 #ifdef CONFIG_IPV6_SUBTREES
2293 seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_src.addr),
2294 rt->rt6i_src.plen);
2295 #else
2296 seq_puts(m, "00000000000000000000000000000000 00 ");
2297 #endif
2298
2299 if (rt->rt6i_nexthop) {
2300 seq_printf(m, NIP6_SEQFMT,
2301 NIP6(*((struct in6_addr *)rt->rt6i_nexthop->primary_key)));
2302 } else {
2303 seq_puts(m, "00000000000000000000000000000000");
2304 }
2305 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2306 rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2307 rt->u.dst.__use, rt->rt6i_flags,
2308 rt->rt6i_dev ? rt->rt6i_dev->name : "");
2309 return 0;
2310 }
2311
2312 static int ipv6_route_show(struct seq_file *m, void *v)
2313 {
2314 fib6_clean_all(rt6_info_route, 0, m);
2315 return 0;
2316 }
2317
2318 static int ipv6_route_open(struct inode *inode, struct file *file)
2319 {
2320 return single_open(file, ipv6_route_show, NULL);
2321 }
2322
2323 static const struct file_operations ipv6_route_proc_fops = {
2324 .owner = THIS_MODULE,
2325 .open = ipv6_route_open,
2326 .read = seq_read,
2327 .llseek = seq_lseek,
2328 .release = single_release,
2329 };
2330
2331 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2332 {
2333 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2334 rt6_stats.fib_nodes, rt6_stats.fib_route_nodes,
2335 rt6_stats.fib_rt_alloc, rt6_stats.fib_rt_entries,
2336 rt6_stats.fib_rt_cache,
2337 atomic_read(&ip6_dst_ops.entries),
2338 rt6_stats.fib_discarded_routes);
2339
2340 return 0;
2341 }
2342
2343 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2344 {
2345 return single_open(file, rt6_stats_seq_show, NULL);
2346 }
2347
2348 static const struct file_operations rt6_stats_seq_fops = {
2349 .owner = THIS_MODULE,
2350 .open = rt6_stats_seq_open,
2351 .read = seq_read,
2352 .llseek = seq_lseek,
2353 .release = single_release,
2354 };
2355 #endif /* CONFIG_PROC_FS */
2356
2357 #ifdef CONFIG_SYSCTL
2358
2359 static int flush_delay;
2360
2361 static
2362 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write, struct file * filp,
2363 void __user *buffer, size_t *lenp, loff_t *ppos)
2364 {
2365 if (write) {
2366 proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
2367 fib6_run_gc(flush_delay <= 0 ? ~0UL : (unsigned long)flush_delay);
2368 return 0;
2369 } else
2370 return -EINVAL;
2371 }
2372
2373 ctl_table ipv6_route_table[] = {
2374 {
2375 .procname = "flush",
2376 .data = &flush_delay,
2377 .maxlen = sizeof(int),
2378 .mode = 0200,
2379 .proc_handler = &ipv6_sysctl_rtcache_flush
2380 },
2381 {
2382 .ctl_name = NET_IPV6_ROUTE_GC_THRESH,
2383 .procname = "gc_thresh",
2384 .data = &ip6_dst_ops.gc_thresh,
2385 .maxlen = sizeof(int),
2386 .mode = 0644,
2387 .proc_handler = &proc_dointvec,
2388 },
2389 {
2390 .ctl_name = NET_IPV6_ROUTE_MAX_SIZE,
2391 .procname = "max_size",
2392 .data = &ip6_rt_max_size,
2393 .maxlen = sizeof(int),
2394 .mode = 0644,
2395 .proc_handler = &proc_dointvec,
2396 },
2397 {
2398 .ctl_name = NET_IPV6_ROUTE_GC_MIN_INTERVAL,
2399 .procname = "gc_min_interval",
2400 .data = &ip6_rt_gc_min_interval,
2401 .maxlen = sizeof(int),
2402 .mode = 0644,
2403 .proc_handler = &proc_dointvec_jiffies,
2404 .strategy = &sysctl_jiffies,
2405 },
2406 {
2407 .ctl_name = NET_IPV6_ROUTE_GC_TIMEOUT,
2408 .procname = "gc_timeout",
2409 .data = &ip6_rt_gc_timeout,
2410 .maxlen = sizeof(int),
2411 .mode = 0644,
2412 .proc_handler = &proc_dointvec_jiffies,
2413 .strategy = &sysctl_jiffies,
2414 },
2415 {
2416 .ctl_name = NET_IPV6_ROUTE_GC_INTERVAL,
2417 .procname = "gc_interval",
2418 .data = &ip6_rt_gc_interval,
2419 .maxlen = sizeof(int),
2420 .mode = 0644,
2421 .proc_handler = &proc_dointvec_jiffies,
2422 .strategy = &sysctl_jiffies,
2423 },
2424 {
2425 .ctl_name = NET_IPV6_ROUTE_GC_ELASTICITY,
2426 .procname = "gc_elasticity",
2427 .data = &ip6_rt_gc_elasticity,
2428 .maxlen = sizeof(int),
2429 .mode = 0644,
2430 .proc_handler = &proc_dointvec_jiffies,
2431 .strategy = &sysctl_jiffies,
2432 },
2433 {
2434 .ctl_name = NET_IPV6_ROUTE_MTU_EXPIRES,
2435 .procname = "mtu_expires",
2436 .data = &ip6_rt_mtu_expires,
2437 .maxlen = sizeof(int),
2438 .mode = 0644,
2439 .proc_handler = &proc_dointvec_jiffies,
2440 .strategy = &sysctl_jiffies,
2441 },
2442 {
2443 .ctl_name = NET_IPV6_ROUTE_MIN_ADVMSS,
2444 .procname = "min_adv_mss",
2445 .data = &ip6_rt_min_advmss,
2446 .maxlen = sizeof(int),
2447 .mode = 0644,
2448 .proc_handler = &proc_dointvec_jiffies,
2449 .strategy = &sysctl_jiffies,
2450 },
2451 {
2452 .ctl_name = NET_IPV6_ROUTE_GC_MIN_INTERVAL_MS,
2453 .procname = "gc_min_interval_ms",
2454 .data = &ip6_rt_gc_min_interval,
2455 .maxlen = sizeof(int),
2456 .mode = 0644,
2457 .proc_handler = &proc_dointvec_ms_jiffies,
2458 .strategy = &sysctl_ms_jiffies,
2459 },
2460 { .ctl_name = 0 }
2461 };
2462
2463 #endif
2464
2465 void __init ip6_route_init(void)
2466 {
2467 ip6_dst_ops.kmem_cachep =
2468 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2469 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
2470 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops.kmem_cachep;
2471
2472 fib6_init();
2473 proc_net_fops_create(&init_net, "ipv6_route", 0, &ipv6_route_proc_fops);
2474 proc_net_fops_create(&init_net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2475 #ifdef CONFIG_XFRM
2476 xfrm6_init();
2477 #endif
2478 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2479 fib6_rules_init();
2480 #endif
2481
2482 __rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL);
2483 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL);
2484 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL);
2485 }
2486
2487 void ip6_route_cleanup(void)
2488 {
2489 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2490 fib6_rules_cleanup();
2491 #endif
2492 #ifdef CONFIG_PROC_FS
2493 proc_net_remove(&init_net, "ipv6_route");
2494 proc_net_remove(&init_net, "rt6_stats");
2495 #endif
2496 #ifdef CONFIG_XFRM
2497 xfrm6_fini();
2498 #endif
2499 rt6_ifdown(NULL);
2500 fib6_gc_cleanup();
2501 kmem_cache_destroy(ip6_dst_ops.kmem_cachep);
2502 }
kernel source for telekom puls (alcatel/mediatek)