?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
bus: mhi: core: Fix some error return code
[linux/fpc-iii.git] / net / ipv6 / route.c
blob8d418038fe32914c2cd828ec4abd81ea0c26f26a
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * Linux INET6 implementation
4 * FIB front-end.
5 *
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
8 */
9
10 /* Changes:
11 *
12 * YOSHIFUJI Hideaki @USAGI
13 * reworked default router selection.
14 * - respect outgoing interface
15 * - select from (probably) reachable routers (i.e.
16 * routers in REACHABLE, STALE, DELAY or PROBE states).
17 * - always select the same router if it is (probably)
18 * reachable. otherwise, round-robin the list.
19 * Ville Nuorvala
20 * Fixed routing subtrees.
21 */
22
23 #define pr_fmt(fmt) "IPv6: " fmt
24
25 #include <linux/capability.h>
26 #include <linux/errno.h>
27 #include <linux/export.h>
28 #include <linux/types.h>
29 #include <linux/times.h>
30 #include <linux/socket.h>
31 #include <linux/sockios.h>
32 #include <linux/net.h>
33 #include <linux/route.h>
34 #include <linux/netdevice.h>
35 #include <linux/in6.h>
36 #include <linux/mroute6.h>
37 #include <linux/init.h>
38 #include <linux/if_arp.h>
39 #include <linux/proc_fs.h>
40 #include <linux/seq_file.h>
41 #include <linux/nsproxy.h>
42 #include <linux/slab.h>
43 #include <linux/jhash.h>
44 #include <net/net_namespace.h>
45 #include <net/snmp.h>
46 #include <net/ipv6.h>
47 #include <net/ip6_fib.h>
48 #include <net/ip6_route.h>
49 #include <net/ndisc.h>
50 #include <net/addrconf.h>
51 #include <net/tcp.h>
52 #include <linux/rtnetlink.h>
53 #include <net/dst.h>
54 #include <net/dst_metadata.h>
55 #include <net/xfrm.h>
56 #include <net/netevent.h>
57 #include <net/netlink.h>
58 #include <net/rtnh.h>
59 #include <net/lwtunnel.h>
60 #include <net/ip_tunnels.h>
61 #include <net/l3mdev.h>
62 #include <net/ip.h>
63 #include <linux/uaccess.h>
64
65 #ifdef CONFIG_SYSCTL
66 #include <linux/sysctl.h>
67 #endif
68
69 static int ip6_rt_type_to_error(u8 fib6_type);
70
71 #define CREATE_TRACE_POINTS
72 #include <trace/events/fib6.h>
73 EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
74 #undef CREATE_TRACE_POINTS
75
76 enum rt6_nud_state {
77 RT6_NUD_FAIL_HARD = -3,
78 RT6_NUD_FAIL_PROBE = -2,
79 RT6_NUD_FAIL_DO_RR = -1,
80 RT6_NUD_SUCCEED = 1
81 };
82
83 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
84 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
85 static unsigned int ip6_mtu(const struct dst_entry *dst);
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(struct dst_ops *ops);
91
92 static int ip6_pkt_discard(struct sk_buff *skb);
93 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
94 static int ip6_pkt_prohibit(struct sk_buff *skb);
95 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
96 static void ip6_link_failure(struct sk_buff *skb);
97 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
98 struct sk_buff *skb, u32 mtu,
99 bool confirm_neigh);
100 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
101 struct sk_buff *skb);
102 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
103 int strict);
104 static size_t rt6_nlmsg_size(struct fib6_info *f6i);
105 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
106 struct fib6_info *rt, struct dst_entry *dst,
107 struct in6_addr *dest, struct in6_addr *src,
108 int iif, int type, u32 portid, u32 seq,
109 unsigned int flags);
110 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
111 const struct in6_addr *daddr,
112 const struct in6_addr *saddr);
113
114 #ifdef CONFIG_IPV6_ROUTE_INFO
115 static struct fib6_info *rt6_add_route_info(struct net *net,
116 const struct in6_addr *prefix, int prefixlen,
117 const struct in6_addr *gwaddr,
118 struct net_device *dev,
119 unsigned int pref);
120 static struct fib6_info *rt6_get_route_info(struct net *net,
121 const struct in6_addr *prefix, int prefixlen,
122 const struct in6_addr *gwaddr,
123 struct net_device *dev);
124 #endif
125
126 struct uncached_list {
127 spinlock_t lock;
128 struct list_head head;
129 };
130
131 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
132
133 void rt6_uncached_list_add(struct rt6_info *rt)
134 {
135 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
136
137 rt->rt6i_uncached_list = ul;
138
139 spin_lock_bh(&ul->lock);
140 list_add_tail(&rt->rt6i_uncached, &ul->head);
141 spin_unlock_bh(&ul->lock);
142 }
143
144 void rt6_uncached_list_del(struct rt6_info *rt)
145 {
146 if (!list_empty(&rt->rt6i_uncached)) {
147 struct uncached_list *ul = rt->rt6i_uncached_list;
148 struct net *net = dev_net(rt->dst.dev);
149
150 spin_lock_bh(&ul->lock);
151 list_del(&rt->rt6i_uncached);
152 atomic_dec(&net->ipv6.rt6_stats->fib_rt_uncache);
153 spin_unlock_bh(&ul->lock);
154 }
155 }
156
157 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
158 {
159 struct net_device *loopback_dev = net->loopback_dev;
160 int cpu;
161
162 if (dev == loopback_dev)
163 return;
164
165 for_each_possible_cpu(cpu) {
166 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
167 struct rt6_info *rt;
168
169 spin_lock_bh(&ul->lock);
170 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
171 struct inet6_dev *rt_idev = rt->rt6i_idev;
172 struct net_device *rt_dev = rt->dst.dev;
173
174 if (rt_idev->dev == dev) {
175 rt->rt6i_idev = in6_dev_get(loopback_dev);
176 in6_dev_put(rt_idev);
177 }
178
179 if (rt_dev == dev) {
180 rt->dst.dev = blackhole_netdev;
181 dev_hold(rt->dst.dev);
182 dev_put(rt_dev);
183 }
184 }
185 spin_unlock_bh(&ul->lock);
186 }
187 }
188
189 static inline const void *choose_neigh_daddr(const struct in6_addr *p,
190 struct sk_buff *skb,
191 const void *daddr)
192 {
193 if (!ipv6_addr_any(p))
194 return (const void *) p;
195 else if (skb)
196 return &ipv6_hdr(skb)->daddr;
197 return daddr;
198 }
199
200 struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
201 struct net_device *dev,
202 struct sk_buff *skb,
203 const void *daddr)
204 {
205 struct neighbour *n;
206
207 daddr = choose_neigh_daddr(gw, skb, daddr);
208 n = __ipv6_neigh_lookup(dev, daddr);
209 if (n)
210 return n;
211
212 n = neigh_create(&nd_tbl, daddr, dev);
213 return IS_ERR(n) ? NULL : n;
214 }
215
216 static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
217 struct sk_buff *skb,
218 const void *daddr)
219 {
220 const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
221
222 return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
223 dst->dev, skb, daddr);
224 }
225
226 static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
227 {
228 struct net_device *dev = dst->dev;
229 struct rt6_info *rt = (struct rt6_info *)dst;
230
231 daddr = choose_neigh_daddr(rt6_nexthop(rt, &in6addr_any), NULL, daddr);
232 if (!daddr)
233 return;
234 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
235 return;
236 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
237 return;
238 __ipv6_confirm_neigh(dev, daddr);
239 }
240
241 static struct dst_ops ip6_dst_ops_template = {
242 .family = AF_INET6,
243 .gc = ip6_dst_gc,
244 .gc_thresh = 1024,
245 .check = ip6_dst_check,
246 .default_advmss = ip6_default_advmss,
247 .mtu = ip6_mtu,
248 .cow_metrics = dst_cow_metrics_generic,
249 .destroy = ip6_dst_destroy,
250 .ifdown = ip6_dst_ifdown,
251 .negative_advice = ip6_negative_advice,
252 .link_failure = ip6_link_failure,
253 .update_pmtu = ip6_rt_update_pmtu,
254 .redirect = rt6_do_redirect,
255 .local_out = __ip6_local_out,
256 .neigh_lookup = ip6_dst_neigh_lookup,
257 .confirm_neigh = ip6_confirm_neigh,
258 };
259
260 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
261 {
262 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
263
264 return mtu ? : dst->dev->mtu;
265 }
266
267 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
268 struct sk_buff *skb, u32 mtu,
269 bool confirm_neigh)
270 {
271 }
272
273 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
274 struct sk_buff *skb)
275 {
276 }
277
278 static struct dst_ops ip6_dst_blackhole_ops = {
279 .family = AF_INET6,
280 .destroy = ip6_dst_destroy,
281 .check = ip6_dst_check,
282 .mtu = ip6_blackhole_mtu,
283 .default_advmss = ip6_default_advmss,
284 .update_pmtu = ip6_rt_blackhole_update_pmtu,
285 .redirect = ip6_rt_blackhole_redirect,
286 .cow_metrics = dst_cow_metrics_generic,
287 .neigh_lookup = ip6_dst_neigh_lookup,
288 };
289
290 static const u32 ip6_template_metrics[RTAX_MAX] = {
291 [RTAX_HOPLIMIT - 1] = 0,
292 };
293
294 static const struct fib6_info fib6_null_entry_template = {
295 .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
296 .fib6_protocol = RTPROT_KERNEL,
297 .fib6_metric = ~(u32)0,
298 .fib6_ref = REFCOUNT_INIT(1),
299 .fib6_type = RTN_UNREACHABLE,
300 .fib6_metrics = (struct dst_metrics *)&dst_default_metrics,
301 };
302
303 static const struct rt6_info ip6_null_entry_template = {
304 .dst = {
305 .__refcnt = ATOMIC_INIT(1),
306 .__use = 1,
307 .obsolete = DST_OBSOLETE_FORCE_CHK,
308 .error = -ENETUNREACH,
309 .input = ip6_pkt_discard,
310 .output = ip6_pkt_discard_out,
311 },
312 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
313 };
314
315 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
316
317 static const struct rt6_info ip6_prohibit_entry_template = {
318 .dst = {
319 .__refcnt = ATOMIC_INIT(1),
320 .__use = 1,
321 .obsolete = DST_OBSOLETE_FORCE_CHK,
322 .error = -EACCES,
323 .input = ip6_pkt_prohibit,
324 .output = ip6_pkt_prohibit_out,
325 },
326 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
327 };
328
329 static const struct rt6_info ip6_blk_hole_entry_template = {
330 .dst = {
331 .__refcnt = ATOMIC_INIT(1),
332 .__use = 1,
333 .obsolete = DST_OBSOLETE_FORCE_CHK,
334 .error = -EINVAL,
335 .input = dst_discard,
336 .output = dst_discard_out,
337 },
338 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
339 };
340
341 #endif
342
343 static void rt6_info_init(struct rt6_info *rt)
344 {
345 struct dst_entry *dst = &rt->dst;
346
347 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
348 INIT_LIST_HEAD(&rt->rt6i_uncached);
349 }
350
351 /* allocate dst with ip6_dst_ops */
352 struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
353 int flags)
354 {
355 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
356 1, DST_OBSOLETE_FORCE_CHK, flags);
357
358 if (rt) {
359 rt6_info_init(rt);
360 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
361 }
362
363 return rt;
364 }
365 EXPORT_SYMBOL(ip6_dst_alloc);
366
367 static void ip6_dst_destroy(struct dst_entry *dst)
368 {
369 struct rt6_info *rt = (struct rt6_info *)dst;
370 struct fib6_info *from;
371 struct inet6_dev *idev;
372
373 ip_dst_metrics_put(dst);
374 rt6_uncached_list_del(rt);
375
376 idev = rt->rt6i_idev;
377 if (idev) {
378 rt->rt6i_idev = NULL;
379 in6_dev_put(idev);
380 }
381
382 from = xchg((__force struct fib6_info **)&rt->from, NULL);
383 fib6_info_release(from);
384 }
385
386 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
387 int how)
388 {
389 struct rt6_info *rt = (struct rt6_info *)dst;
390 struct inet6_dev *idev = rt->rt6i_idev;
391 struct net_device *loopback_dev =
392 dev_net(dev)->loopback_dev;
393
394 if (idev && idev->dev != loopback_dev) {
395 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
396 if (loopback_idev) {
397 rt->rt6i_idev = loopback_idev;
398 in6_dev_put(idev);
399 }
400 }
401 }
402
403 static bool __rt6_check_expired(const struct rt6_info *rt)
404 {
405 if (rt->rt6i_flags & RTF_EXPIRES)
406 return time_after(jiffies, rt->dst.expires);
407 else
408 return false;
409 }
410
411 static bool rt6_check_expired(const struct rt6_info *rt)
412 {
413 struct fib6_info *from;
414
415 from = rcu_dereference(rt->from);
416
417 if (rt->rt6i_flags & RTF_EXPIRES) {
418 if (time_after(jiffies, rt->dst.expires))
419 return true;
420 } else if (from) {
421 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
422 fib6_check_expired(from);
423 }
424 return false;
425 }
426
427 void fib6_select_path(const struct net *net, struct fib6_result *res,
428 struct flowi6 *fl6, int oif, bool have_oif_match,
429 const struct sk_buff *skb, int strict)
430 {
431 struct fib6_info *sibling, *next_sibling;
432 struct fib6_info *match = res->f6i;
433
434 if ((!match->fib6_nsiblings && !match->nh) || have_oif_match)
435 goto out;
436
437 /* We might have already computed the hash for ICMPv6 errors. In such
438 * case it will always be non-zero. Otherwise now is the time to do it.
439 */
440 if (!fl6->mp_hash &&
441 (!match->nh || nexthop_is_multipath(match->nh)))
442 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
443
444 if (unlikely(match->nh)) {
445 nexthop_path_fib6_result(res, fl6->mp_hash);
446 return;
447 }
448
449 if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
450 goto out;
451
452 list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
453 fib6_siblings) {
454 const struct fib6_nh *nh = sibling->fib6_nh;
455 int nh_upper_bound;
456
457 nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
458 if (fl6->mp_hash > nh_upper_bound)
459 continue;
460 if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
461 break;
462 match = sibling;
463 break;
464 }
465
466 out:
467 res->f6i = match;
468 res->nh = match->fib6_nh;
469 }
470
471 /*
472 * Route lookup. rcu_read_lock() should be held.
473 */
474
475 static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
476 const struct in6_addr *saddr, int oif, int flags)
477 {
478 const struct net_device *dev;
479
480 if (nh->fib_nh_flags & RTNH_F_DEAD)
481 return false;
482
483 dev = nh->fib_nh_dev;
484 if (oif) {
485 if (dev->ifindex == oif)
486 return true;
487 } else {
488 if (ipv6_chk_addr(net, saddr, dev,
489 flags & RT6_LOOKUP_F_IFACE))
490 return true;
491 }
492
493 return false;
494 }
495
496 struct fib6_nh_dm_arg {
497 struct net *net;
498 const struct in6_addr *saddr;
499 int oif;
500 int flags;
501 struct fib6_nh *nh;
502 };
503
504 static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
505 {
506 struct fib6_nh_dm_arg *arg = _arg;
507
508 arg->nh = nh;
509 return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
510 arg->flags);
511 }
512
513 /* returns fib6_nh from nexthop or NULL */
514 static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
515 struct fib6_result *res,
516 const struct in6_addr *saddr,
517 int oif, int flags)
518 {
519 struct fib6_nh_dm_arg arg = {
520 .net = net,
521 .saddr = saddr,
522 .oif = oif,
523 .flags = flags,
524 };
525
526 if (nexthop_is_blackhole(nh))
527 return NULL;
528
529 if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
530 return arg.nh;
531
532 return NULL;
533 }
534
535 static void rt6_device_match(struct net *net, struct fib6_result *res,
536 const struct in6_addr *saddr, int oif, int flags)
537 {
538 struct fib6_info *f6i = res->f6i;
539 struct fib6_info *spf6i;
540 struct fib6_nh *nh;
541
542 if (!oif && ipv6_addr_any(saddr)) {
543 if (unlikely(f6i->nh)) {
544 nh = nexthop_fib6_nh(f6i->nh);
545 if (nexthop_is_blackhole(f6i->nh))
546 goto out_blackhole;
547 } else {
548 nh = f6i->fib6_nh;
549 }
550 if (!(nh->fib_nh_flags & RTNH_F_DEAD))
551 goto out;
552 }
553
554 for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
555 bool matched = false;
556
557 if (unlikely(spf6i->nh)) {
558 nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
559 oif, flags);
560 if (nh)
561 matched = true;
562 } else {
563 nh = spf6i->fib6_nh;
564 if (__rt6_device_match(net, nh, saddr, oif, flags))
565 matched = true;
566 }
567 if (matched) {
568 res->f6i = spf6i;
569 goto out;
570 }
571 }
572
573 if (oif && flags & RT6_LOOKUP_F_IFACE) {
574 res->f6i = net->ipv6.fib6_null_entry;
575 nh = res->f6i->fib6_nh;
576 goto out;
577 }
578
579 if (unlikely(f6i->nh)) {
580 nh = nexthop_fib6_nh(f6i->nh);
581 if (nexthop_is_blackhole(f6i->nh))
582 goto out_blackhole;
583 } else {
584 nh = f6i->fib6_nh;
585 }
586
587 if (nh->fib_nh_flags & RTNH_F_DEAD) {
588 res->f6i = net->ipv6.fib6_null_entry;
589 nh = res->f6i->fib6_nh;
590 }
591 out:
592 res->nh = nh;
593 res->fib6_type = res->f6i->fib6_type;
594 res->fib6_flags = res->f6i->fib6_flags;
595 return;
596
597 out_blackhole:
598 res->fib6_flags |= RTF_REJECT;
599 res->fib6_type = RTN_BLACKHOLE;
600 res->nh = nh;
601 }
602
603 #ifdef CONFIG_IPV6_ROUTER_PREF
604 struct __rt6_probe_work {
605 struct work_struct work;
606 struct in6_addr target;
607 struct net_device *dev;
608 };
609
610 static void rt6_probe_deferred(struct work_struct *w)
611 {
612 struct in6_addr mcaddr;
613 struct __rt6_probe_work *work =
614 container_of(w, struct __rt6_probe_work, work);
615
616 addrconf_addr_solict_mult(&work->target, &mcaddr);
617 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
618 dev_put(work->dev);
619 kfree(work);
620 }
621
622 static void rt6_probe(struct fib6_nh *fib6_nh)
623 {
624 struct __rt6_probe_work *work = NULL;
625 const struct in6_addr *nh_gw;
626 unsigned long last_probe;
627 struct neighbour *neigh;
628 struct net_device *dev;
629 struct inet6_dev *idev;
630
631 /*
632 * Okay, this does not seem to be appropriate
633 * for now, however, we need to check if it
634 * is really so; aka Router Reachability Probing.
635 *
636 * Router Reachability Probe MUST be rate-limited
637 * to no more than one per minute.
638 */
639 if (!fib6_nh->fib_nh_gw_family)
640 return;
641
642 nh_gw = &fib6_nh->fib_nh_gw6;
643 dev = fib6_nh->fib_nh_dev;
644 rcu_read_lock_bh();
645 last_probe = READ_ONCE(fib6_nh->last_probe);
646 idev = __in6_dev_get(dev);
647 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
648 if (neigh) {
649 if (neigh->nud_state & NUD_VALID)
650 goto out;
651
652 write_lock(&neigh->lock);
653 if (!(neigh->nud_state & NUD_VALID) &&
654 time_after(jiffies,
655 neigh->updated + idev->cnf.rtr_probe_interval)) {
656 work = kmalloc(sizeof(*work), GFP_ATOMIC);
657 if (work)
658 __neigh_set_probe_once(neigh);
659 }
660 write_unlock(&neigh->lock);
661 } else if (time_after(jiffies, last_probe +
662 idev->cnf.rtr_probe_interval)) {
663 work = kmalloc(sizeof(*work), GFP_ATOMIC);
664 }
665
666 if (!work || cmpxchg(&fib6_nh->last_probe,
667 last_probe, jiffies) != last_probe) {
668 kfree(work);
669 } else {
670 INIT_WORK(&work->work, rt6_probe_deferred);
671 work->target = *nh_gw;
672 dev_hold(dev);
673 work->dev = dev;
674 schedule_work(&work->work);
675 }
676
677 out:
678 rcu_read_unlock_bh();
679 }
680 #else
681 static inline void rt6_probe(struct fib6_nh *fib6_nh)
682 {
683 }
684 #endif
685
686 /*
687 * Default Router Selection (RFC 2461 6.3.6)
688 */
689 static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
690 {
691 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
692 struct neighbour *neigh;
693
694 rcu_read_lock_bh();
695 neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
696 &fib6_nh->fib_nh_gw6);
697 if (neigh) {
698 read_lock(&neigh->lock);
699 if (neigh->nud_state & NUD_VALID)
700 ret = RT6_NUD_SUCCEED;
701 #ifdef CONFIG_IPV6_ROUTER_PREF
702 else if (!(neigh->nud_state & NUD_FAILED))
703 ret = RT6_NUD_SUCCEED;
704 else
705 ret = RT6_NUD_FAIL_PROBE;
706 #endif
707 read_unlock(&neigh->lock);
708 } else {
709 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
710 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
711 }
712 rcu_read_unlock_bh();
713
714 return ret;
715 }
716
717 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
718 int strict)
719 {
720 int m = 0;
721
722 if (!oif || nh->fib_nh_dev->ifindex == oif)
723 m = 2;
724
725 if (!m && (strict & RT6_LOOKUP_F_IFACE))
726 return RT6_NUD_FAIL_HARD;
727 #ifdef CONFIG_IPV6_ROUTER_PREF
728 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
729 #endif
730 if ((strict & RT6_LOOKUP_F_REACHABLE) &&
731 !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
732 int n = rt6_check_neigh(nh);
733 if (n < 0)
734 return n;
735 }
736 return m;
737 }
738
739 static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
740 int oif, int strict, int *mpri, bool *do_rr)
741 {
742 bool match_do_rr = false;
743 bool rc = false;
744 int m;
745
746 if (nh->fib_nh_flags & RTNH_F_DEAD)
747 goto out;
748
749 if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
750 nh->fib_nh_flags & RTNH_F_LINKDOWN &&
751 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
752 goto out;
753
754 m = rt6_score_route(nh, fib6_flags, oif, strict);
755 if (m == RT6_NUD_FAIL_DO_RR) {
756 match_do_rr = true;
757 m = 0; /* lowest valid score */
758 } else if (m == RT6_NUD_FAIL_HARD) {
759 goto out;
760 }
761
762 if (strict & RT6_LOOKUP_F_REACHABLE)
763 rt6_probe(nh);
764
765 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
766 if (m > *mpri) {
767 *do_rr = match_do_rr;
768 *mpri = m;
769 rc = true;
770 }
771 out:
772 return rc;
773 }
774
775 struct fib6_nh_frl_arg {
776 u32 flags;
777 int oif;
778 int strict;
779 int *mpri;
780 bool *do_rr;
781 struct fib6_nh *nh;
782 };
783
784 static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
785 {
786 struct fib6_nh_frl_arg *arg = _arg;
787
788 arg->nh = nh;
789 return find_match(nh, arg->flags, arg->oif, arg->strict,
790 arg->mpri, arg->do_rr);
791 }
792
793 static void __find_rr_leaf(struct fib6_info *f6i_start,
794 struct fib6_info *nomatch, u32 metric,
795 struct fib6_result *res, struct fib6_info **cont,
796 int oif, int strict, bool *do_rr, int *mpri)
797 {
798 struct fib6_info *f6i;
799
800 for (f6i = f6i_start;
801 f6i && f6i != nomatch;
802 f6i = rcu_dereference(f6i->fib6_next)) {
803 bool matched = false;
804 struct fib6_nh *nh;
805
806 if (cont && f6i->fib6_metric != metric) {
807 *cont = f6i;
808 return;
809 }
810
811 if (fib6_check_expired(f6i))
812 continue;
813
814 if (unlikely(f6i->nh)) {
815 struct fib6_nh_frl_arg arg = {
816 .flags = f6i->fib6_flags,
817 .oif = oif,
818 .strict = strict,
819 .mpri = mpri,
820 .do_rr = do_rr
821 };
822
823 if (nexthop_is_blackhole(f6i->nh)) {
824 res->fib6_flags = RTF_REJECT;
825 res->fib6_type = RTN_BLACKHOLE;
826 res->f6i = f6i;
827 res->nh = nexthop_fib6_nh(f6i->nh);
828 return;
829 }
830 if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
831 &arg)) {
832 matched = true;
833 nh = arg.nh;
834 }
835 } else {
836 nh = f6i->fib6_nh;
837 if (find_match(nh, f6i->fib6_flags, oif, strict,
838 mpri, do_rr))
839 matched = true;
840 }
841 if (matched) {
842 res->f6i = f6i;
843 res->nh = nh;
844 res->fib6_flags = f6i->fib6_flags;
845 res->fib6_type = f6i->fib6_type;
846 }
847 }
848 }
849
850 static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
851 struct fib6_info *rr_head, int oif, int strict,
852 bool *do_rr, struct fib6_result *res)
853 {
854 u32 metric = rr_head->fib6_metric;
855 struct fib6_info *cont = NULL;
856 int mpri = -1;
857
858 __find_rr_leaf(rr_head, NULL, metric, res, &cont,
859 oif, strict, do_rr, &mpri);
860
861 __find_rr_leaf(leaf, rr_head, metric, res, &cont,
862 oif, strict, do_rr, &mpri);
863
864 if (res->f6i || !cont)
865 return;
866
867 __find_rr_leaf(cont, NULL, metric, res, NULL,
868 oif, strict, do_rr, &mpri);
869 }
870
871 static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
872 struct fib6_result *res, int strict)
873 {
874 struct fib6_info *leaf = rcu_dereference(fn->leaf);
875 struct fib6_info *rt0;
876 bool do_rr = false;
877 int key_plen;
878
879 /* make sure this function or its helpers sets f6i */
880 res->f6i = NULL;
881
882 if (!leaf || leaf == net->ipv6.fib6_null_entry)
883 goto out;
884
885 rt0 = rcu_dereference(fn->rr_ptr);
886 if (!rt0)
887 rt0 = leaf;
888
889 /* Double check to make sure fn is not an intermediate node
890 * and fn->leaf does not points to its child's leaf
891 * (This might happen if all routes under fn are deleted from
892 * the tree and fib6_repair_tree() is called on the node.)
893 */
894 key_plen = rt0->fib6_dst.plen;
895 #ifdef CONFIG_IPV6_SUBTREES
896 if (rt0->fib6_src.plen)
897 key_plen = rt0->fib6_src.plen;
898 #endif
899 if (fn->fn_bit != key_plen)
900 goto out;
901
902 find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
903 if (do_rr) {
904 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
905
906 /* no entries matched; do round-robin */
907 if (!next || next->fib6_metric != rt0->fib6_metric)
908 next = leaf;
909
910 if (next != rt0) {
911 spin_lock_bh(&leaf->fib6_table->tb6_lock);
912 /* make sure next is not being deleted from the tree */
913 if (next->fib6_node)
914 rcu_assign_pointer(fn->rr_ptr, next);
915 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
916 }
917 }
918
919 out:
920 if (!res->f6i) {
921 res->f6i = net->ipv6.fib6_null_entry;
922 res->nh = res->f6i->fib6_nh;
923 res->fib6_flags = res->f6i->fib6_flags;
924 res->fib6_type = res->f6i->fib6_type;
925 }
926 }
927
928 static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
929 {
930 return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
931 res->nh->fib_nh_gw_family;
932 }
933
934 #ifdef CONFIG_IPV6_ROUTE_INFO
935 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
936 const struct in6_addr *gwaddr)
937 {
938 struct net *net = dev_net(dev);
939 struct route_info *rinfo = (struct route_info *) opt;
940 struct in6_addr prefix_buf, *prefix;
941 unsigned int pref;
942 unsigned long lifetime;
943 struct fib6_info *rt;
944
945 if (len < sizeof(struct route_info)) {
946 return -EINVAL;
947 }
948
949 /* Sanity check for prefix_len and length */
950 if (rinfo->length > 3) {
951 return -EINVAL;
952 } else if (rinfo->prefix_len > 128) {
953 return -EINVAL;
954 } else if (rinfo->prefix_len > 64) {
955 if (rinfo->length < 2) {
956 return -EINVAL;
957 }
958 } else if (rinfo->prefix_len > 0) {
959 if (rinfo->length < 1) {
960 return -EINVAL;
961 }
962 }
963
964 pref = rinfo->route_pref;
965 if (pref == ICMPV6_ROUTER_PREF_INVALID)
966 return -EINVAL;
967
968 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
969
970 if (rinfo->length == 3)
971 prefix = (struct in6_addr *)rinfo->prefix;
972 else {
973 /* this function is safe */
974 ipv6_addr_prefix(&prefix_buf,
975 (struct in6_addr *)rinfo->prefix,
976 rinfo->prefix_len);
977 prefix = &prefix_buf;
978 }
979
980 if (rinfo->prefix_len == 0)
981 rt = rt6_get_dflt_router(net, gwaddr, dev);
982 else
983 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
984 gwaddr, dev);
985
986 if (rt && !lifetime) {
987 ip6_del_rt(net, rt);
988 rt = NULL;
989 }
990
991 if (!rt && lifetime)
992 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
993 dev, pref);
994 else if (rt)
995 rt->fib6_flags = RTF_ROUTEINFO |
996 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
997
998 if (rt) {
999 if (!addrconf_finite_timeout(lifetime))
1000 fib6_clean_expires(rt);
1001 else
1002 fib6_set_expires(rt, jiffies + HZ * lifetime);
1003
1004 fib6_info_release(rt);
1005 }
1006 return 0;
1007 }
1008 #endif
1009
1010 /*
1011 * Misc support functions
1012 */
1013
1014 /* called with rcu_lock held */
1015 static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
1016 {
1017 struct net_device *dev = res->nh->fib_nh_dev;
1018
1019 if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1020 /* for copies of local routes, dst->dev needs to be the
1021 * device if it is a master device, the master device if
1022 * device is enslaved, and the loopback as the default
1023 */
1024 if (netif_is_l3_slave(dev) &&
1025 !rt6_need_strict(&res->f6i->fib6_dst.addr))
1026 dev = l3mdev_master_dev_rcu(dev);
1027 else if (!netif_is_l3_master(dev))
1028 dev = dev_net(dev)->loopback_dev;
1029 /* last case is netif_is_l3_master(dev) is true in which
1030 * case we want dev returned to be dev
1031 */
1032 }
1033
1034 return dev;
1035 }
1036
1037 static const int fib6_prop[RTN_MAX + 1] = {
1038 [RTN_UNSPEC] = 0,
1039 [RTN_UNICAST] = 0,
1040 [RTN_LOCAL] = 0,
1041 [RTN_BROADCAST] = 0,
1042 [RTN_ANYCAST] = 0,
1043 [RTN_MULTICAST] = 0,
1044 [RTN_BLACKHOLE] = -EINVAL,
1045 [RTN_UNREACHABLE] = -EHOSTUNREACH,
1046 [RTN_PROHIBIT] = -EACCES,
1047 [RTN_THROW] = -EAGAIN,
1048 [RTN_NAT] = -EINVAL,
1049 [RTN_XRESOLVE] = -EINVAL,
1050 };
1051
1052 static int ip6_rt_type_to_error(u8 fib6_type)
1053 {
1054 return fib6_prop[fib6_type];
1055 }
1056
1057 static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
1058 {
1059 unsigned short flags = 0;
1060
1061 if (rt->dst_nocount)
1062 flags |= DST_NOCOUNT;
1063 if (rt->dst_nopolicy)
1064 flags |= DST_NOPOLICY;
1065
1066 return flags;
1067 }
1068
1069 static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
1070 {
1071 rt->dst.error = ip6_rt_type_to_error(fib6_type);
1072
1073 switch (fib6_type) {
1074 case RTN_BLACKHOLE:
1075 rt->dst.output = dst_discard_out;
1076 rt->dst.input = dst_discard;
1077 break;
1078 case RTN_PROHIBIT:
1079 rt->dst.output = ip6_pkt_prohibit_out;
1080 rt->dst.input = ip6_pkt_prohibit;
1081 break;
1082 case RTN_THROW:
1083 case RTN_UNREACHABLE:
1084 default:
1085 rt->dst.output = ip6_pkt_discard_out;
1086 rt->dst.input = ip6_pkt_discard;
1087 break;
1088 }
1089 }
1090
1091 static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
1092 {
1093 struct fib6_info *f6i = res->f6i;
1094
1095 if (res->fib6_flags & RTF_REJECT) {
1096 ip6_rt_init_dst_reject(rt, res->fib6_type);
1097 return;
1098 }
1099
1100 rt->dst.error = 0;
1101 rt->dst.output = ip6_output;
1102
1103 if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
1104 rt->dst.input = ip6_input;
1105 } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
1106 rt->dst.input = ip6_mc_input;
1107 } else {
1108 rt->dst.input = ip6_forward;
1109 }
1110
1111 if (res->nh->fib_nh_lws) {
1112 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
1113 lwtunnel_set_redirect(&rt->dst);
1114 }
1115
1116 rt->dst.lastuse = jiffies;
1117 }
1118
1119 /* Caller must already hold reference to @from */
1120 static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
1121 {
1122 rt->rt6i_flags &= ~RTF_EXPIRES;
1123 rcu_assign_pointer(rt->from, from);
1124 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
1125 }
1126
1127 /* Caller must already hold reference to f6i in result */
1128 static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
1129 {
1130 const struct fib6_nh *nh = res->nh;
1131 const struct net_device *dev = nh->fib_nh_dev;
1132 struct fib6_info *f6i = res->f6i;
1133
1134 ip6_rt_init_dst(rt, res);
1135
1136 rt->rt6i_dst = f6i->fib6_dst;
1137 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
1138 rt->rt6i_flags = res->fib6_flags;
1139 if (nh->fib_nh_gw_family) {
1140 rt->rt6i_gateway = nh->fib_nh_gw6;
1141 rt->rt6i_flags |= RTF_GATEWAY;
1142 }
1143 rt6_set_from(rt, f6i);
1144 #ifdef CONFIG_IPV6_SUBTREES
1145 rt->rt6i_src = f6i->fib6_src;
1146 #endif
1147 }
1148
1149 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1150 struct in6_addr *saddr)
1151 {
1152 struct fib6_node *pn, *sn;
1153 while (1) {
1154 if (fn->fn_flags & RTN_TL_ROOT)
1155 return NULL;
1156 pn = rcu_dereference(fn->parent);
1157 sn = FIB6_SUBTREE(pn);
1158 if (sn && sn != fn)
1159 fn = fib6_node_lookup(sn, NULL, saddr);
1160 else
1161 fn = pn;
1162 if (fn->fn_flags & RTN_RTINFO)
1163 return fn;
1164 }
1165 }
1166
1167 static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
1168 {
1169 struct rt6_info *rt = *prt;
1170
1171 if (dst_hold_safe(&rt->dst))
1172 return true;
1173 if (net) {
1174 rt = net->ipv6.ip6_null_entry;
1175 dst_hold(&rt->dst);
1176 } else {
1177 rt = NULL;
1178 }
1179 *prt = rt;
1180 return false;
1181 }
1182
1183 /* called with rcu_lock held */
1184 static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
1185 {
1186 struct net_device *dev = res->nh->fib_nh_dev;
1187 struct fib6_info *f6i = res->f6i;
1188 unsigned short flags;
1189 struct rt6_info *nrt;
1190
1191 if (!fib6_info_hold_safe(f6i))
1192 goto fallback;
1193
1194 flags = fib6_info_dst_flags(f6i);
1195 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1196 if (!nrt) {
1197 fib6_info_release(f6i);
1198 goto fallback;
1199 }
1200
1201 ip6_rt_copy_init(nrt, res);
1202 return nrt;
1203
1204 fallback:
1205 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1206 dst_hold(&nrt->dst);
1207 return nrt;
1208 }
1209
1210 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
1211 struct fib6_table *table,
1212 struct flowi6 *fl6,
1213 const struct sk_buff *skb,
1214 int flags)
1215 {
1216 struct fib6_result res = {};
1217 struct fib6_node *fn;
1218 struct rt6_info *rt;
1219
1220 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1221 flags &= ~RT6_LOOKUP_F_IFACE;
1222
1223 rcu_read_lock();
1224 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1225 restart:
1226 res.f6i = rcu_dereference(fn->leaf);
1227 if (!res.f6i)
1228 res.f6i = net->ipv6.fib6_null_entry;
1229 else
1230 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1231 flags);
1232
1233 if (res.f6i == net->ipv6.fib6_null_entry) {
1234 fn = fib6_backtrack(fn, &fl6->saddr);
1235 if (fn)
1236 goto restart;
1237
1238 rt = net->ipv6.ip6_null_entry;
1239 dst_hold(&rt->dst);
1240 goto out;
1241 } else if (res.fib6_flags & RTF_REJECT) {
1242 goto do_create;
1243 }
1244
1245 fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1246 fl6->flowi6_oif != 0, skb, flags);
1247
1248 /* Search through exception table */
1249 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
1250 if (rt) {
1251 if (ip6_hold_safe(net, &rt))
1252 dst_use_noref(&rt->dst, jiffies);
1253 } else {
1254 do_create:
1255 rt = ip6_create_rt_rcu(&res);
1256 }
1257
1258 out:
1259 trace_fib6_table_lookup(net, &res, table, fl6);
1260
1261 rcu_read_unlock();
1262
1263 return rt;
1264 }
1265
1266 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1267 const struct sk_buff *skb, int flags)
1268 {
1269 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1270 }
1271 EXPORT_SYMBOL_GPL(ip6_route_lookup);
1272
1273 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1274 const struct in6_addr *saddr, int oif,
1275 const struct sk_buff *skb, int strict)
1276 {
1277 struct flowi6 fl6 = {
1278 .flowi6_oif = oif,
1279 .daddr = *daddr,
1280 };
1281 struct dst_entry *dst;
1282 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1283
1284 if (saddr) {
1285 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1286 flags |= RT6_LOOKUP_F_HAS_SADDR;
1287 }
1288
1289 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1290 if (dst->error == 0)
1291 return (struct rt6_info *) dst;
1292
1293 dst_release(dst);
1294
1295 return NULL;
1296 }
1297 EXPORT_SYMBOL(rt6_lookup);
1298
1299 /* ip6_ins_rt is called with FREE table->tb6_lock.
1300 * It takes new route entry, the addition fails by any reason the
1301 * route is released.
1302 * Caller must hold dst before calling it.
1303 */
1304
1305 static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1306 struct netlink_ext_ack *extack)
1307 {
1308 int err;
1309 struct fib6_table *table;
1310
1311 table = rt->fib6_table;
1312 spin_lock_bh(&table->tb6_lock);
1313 err = fib6_add(&table->tb6_root, rt, info, extack);
1314 spin_unlock_bh(&table->tb6_lock);
1315
1316 return err;
1317 }
1318
1319 int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1320 {
1321 struct nl_info info = { .nl_net = net, };
1322
1323 return __ip6_ins_rt(rt, &info, NULL);
1324 }
1325
1326 static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
1327 const struct in6_addr *daddr,
1328 const struct in6_addr *saddr)
1329 {
1330 struct fib6_info *f6i = res->f6i;
1331 struct net_device *dev;
1332 struct rt6_info *rt;
1333
1334 /*
1335 * Clone the route.
1336 */
1337
1338 if (!fib6_info_hold_safe(f6i))
1339 return NULL;
1340
1341 dev = ip6_rt_get_dev_rcu(res);
1342 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1343 if (!rt) {
1344 fib6_info_release(f6i);
1345 return NULL;
1346 }
1347
1348 ip6_rt_copy_init(rt, res);
1349 rt->rt6i_flags |= RTF_CACHE;
1350 rt->rt6i_dst.addr = *daddr;
1351 rt->rt6i_dst.plen = 128;
1352
1353 if (!rt6_is_gw_or_nonexthop(res)) {
1354 if (f6i->fib6_dst.plen != 128 &&
1355 ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
1356 rt->rt6i_flags |= RTF_ANYCAST;
1357 #ifdef CONFIG_IPV6_SUBTREES
1358 if (rt->rt6i_src.plen && saddr) {
1359 rt->rt6i_src.addr = *saddr;
1360 rt->rt6i_src.plen = 128;
1361 }
1362 #endif
1363 }
1364
1365 return rt;
1366 }
1367
1368 static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
1369 {
1370 struct fib6_info *f6i = res->f6i;
1371 unsigned short flags = fib6_info_dst_flags(f6i);
1372 struct net_device *dev;
1373 struct rt6_info *pcpu_rt;
1374
1375 if (!fib6_info_hold_safe(f6i))
1376 return NULL;
1377
1378 rcu_read_lock();
1379 dev = ip6_rt_get_dev_rcu(res);
1380 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags);
1381 rcu_read_unlock();
1382 if (!pcpu_rt) {
1383 fib6_info_release(f6i);
1384 return NULL;
1385 }
1386 ip6_rt_copy_init(pcpu_rt, res);
1387 pcpu_rt->rt6i_flags |= RTF_PCPU;
1388
1389 if (f6i->nh)
1390 pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1391
1392 return pcpu_rt;
1393 }
1394
1395 static bool rt6_is_valid(const struct rt6_info *rt6)
1396 {
1397 return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1398 }
1399
1400 /* It should be called with rcu_read_lock() acquired */
1401 static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
1402 {
1403 struct rt6_info *pcpu_rt;
1404
1405 pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
1406
1407 if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1408 struct rt6_info *prev, **p;
1409
1410 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1411 prev = xchg(p, NULL);
1412 if (prev) {
1413 dst_dev_put(&prev->dst);
1414 dst_release(&prev->dst);
1415 }
1416
1417 pcpu_rt = NULL;
1418 }
1419
1420 return pcpu_rt;
1421 }
1422
1423 static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1424 const struct fib6_result *res)
1425 {
1426 struct rt6_info *pcpu_rt, *prev, **p;
1427
1428 pcpu_rt = ip6_rt_pcpu_alloc(res);
1429 if (!pcpu_rt)
1430 return NULL;
1431
1432 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1433 prev = cmpxchg(p, NULL, pcpu_rt);
1434 BUG_ON(prev);
1435
1436 if (res->f6i->fib6_destroying) {
1437 struct fib6_info *from;
1438
1439 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1440 fib6_info_release(from);
1441 }
1442
1443 return pcpu_rt;
1444 }
1445
1446 /* exception hash table implementation
1447 */
1448 static DEFINE_SPINLOCK(rt6_exception_lock);
1449
1450 /* Remove rt6_ex from hash table and free the memory
1451 * Caller must hold rt6_exception_lock
1452 */
1453 static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1454 struct rt6_exception *rt6_ex)
1455 {
1456 struct fib6_info *from;
1457 struct net *net;
1458
1459 if (!bucket || !rt6_ex)
1460 return;
1461
1462 net = dev_net(rt6_ex->rt6i->dst.dev);
1463 net->ipv6.rt6_stats->fib_rt_cache--;
1464
1465 /* purge completely the exception to allow releasing the held resources:
1466 * some [sk] cache may keep the dst around for unlimited time
1467 */
1468 from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
1469 fib6_info_release(from);
1470 dst_dev_put(&rt6_ex->rt6i->dst);
1471
1472 hlist_del_rcu(&rt6_ex->hlist);
1473 dst_release(&rt6_ex->rt6i->dst);
1474 kfree_rcu(rt6_ex, rcu);
1475 WARN_ON_ONCE(!bucket->depth);
1476 bucket->depth--;
1477 }
1478
1479 /* Remove oldest rt6_ex in bucket and free the memory
1480 * Caller must hold rt6_exception_lock
1481 */
1482 static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1483 {
1484 struct rt6_exception *rt6_ex, *oldest = NULL;
1485
1486 if (!bucket)
1487 return;
1488
1489 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1490 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1491 oldest = rt6_ex;
1492 }
1493 rt6_remove_exception(bucket, oldest);
1494 }
1495
1496 static u32 rt6_exception_hash(const struct in6_addr *dst,
1497 const struct in6_addr *src)
1498 {
1499 static u32 seed __read_mostly;
1500 u32 val;
1501
1502 net_get_random_once(&seed, sizeof(seed));
1503 val = jhash2((const u32 *)dst, sizeof(*dst)/sizeof(u32), seed);
1504
1505 #ifdef CONFIG_IPV6_SUBTREES
1506 if (src)
1507 val = jhash2((const u32 *)src, sizeof(*src)/sizeof(u32), val);
1508 #endif
1509 return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1510 }
1511
1512 /* Helper function to find the cached rt in the hash table
1513 * and update bucket pointer to point to the bucket for this
1514 * (daddr, saddr) pair
1515 * Caller must hold rt6_exception_lock
1516 */
1517 static struct rt6_exception *
1518 __rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1519 const struct in6_addr *daddr,
1520 const struct in6_addr *saddr)
1521 {
1522 struct rt6_exception *rt6_ex;
1523 u32 hval;
1524
1525 if (!(*bucket) || !daddr)
1526 return NULL;
1527
1528 hval = rt6_exception_hash(daddr, saddr);
1529 *bucket += hval;
1530
1531 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1532 struct rt6_info *rt6 = rt6_ex->rt6i;
1533 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1534
1535 #ifdef CONFIG_IPV6_SUBTREES
1536 if (matched && saddr)
1537 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1538 #endif
1539 if (matched)
1540 return rt6_ex;
1541 }
1542 return NULL;
1543 }
1544
1545 /* Helper function to find the cached rt in the hash table
1546 * and update bucket pointer to point to the bucket for this
1547 * (daddr, saddr) pair
1548 * Caller must hold rcu_read_lock()
1549 */
1550 static struct rt6_exception *
1551 __rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1552 const struct in6_addr *daddr,
1553 const struct in6_addr *saddr)
1554 {
1555 struct rt6_exception *rt6_ex;
1556 u32 hval;
1557
1558 WARN_ON_ONCE(!rcu_read_lock_held());
1559
1560 if (!(*bucket) || !daddr)
1561 return NULL;
1562
1563 hval = rt6_exception_hash(daddr, saddr);
1564 *bucket += hval;
1565
1566 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1567 struct rt6_info *rt6 = rt6_ex->rt6i;
1568 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1569
1570 #ifdef CONFIG_IPV6_SUBTREES
1571 if (matched && saddr)
1572 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1573 #endif
1574 if (matched)
1575 return rt6_ex;
1576 }
1577 return NULL;
1578 }
1579
1580 static unsigned int fib6_mtu(const struct fib6_result *res)
1581 {
1582 const struct fib6_nh *nh = res->nh;
1583 unsigned int mtu;
1584
1585 if (res->f6i->fib6_pmtu) {
1586 mtu = res->f6i->fib6_pmtu;
1587 } else {
1588 struct net_device *dev = nh->fib_nh_dev;
1589 struct inet6_dev *idev;
1590
1591 rcu_read_lock();
1592 idev = __in6_dev_get(dev);
1593 mtu = idev->cnf.mtu6;
1594 rcu_read_unlock();
1595 }
1596
1597 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1598
1599 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
1600 }
1601
1602 #define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1603
1604 /* used when the flushed bit is not relevant, only access to the bucket
1605 * (ie., all bucket users except rt6_insert_exception);
1606 *
1607 * called under rcu lock; sometimes called with rt6_exception_lock held
1608 */
1609 static
1610 struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1611 spinlock_t *lock)
1612 {
1613 struct rt6_exception_bucket *bucket;
1614
1615 if (lock)
1616 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1617 lockdep_is_held(lock));
1618 else
1619 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1620
1621 /* remove bucket flushed bit if set */
1622 if (bucket) {
1623 unsigned long p = (unsigned long)bucket;
1624
1625 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1626 bucket = (struct rt6_exception_bucket *)p;
1627 }
1628
1629 return bucket;
1630 }
1631
1632 static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1633 {
1634 unsigned long p = (unsigned long)bucket;
1635
1636 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1637 }
1638
1639 /* called with rt6_exception_lock held */
1640 static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1641 spinlock_t *lock)
1642 {
1643 struct rt6_exception_bucket *bucket;
1644 unsigned long p;
1645
1646 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1647 lockdep_is_held(lock));
1648
1649 p = (unsigned long)bucket;
1650 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1651 bucket = (struct rt6_exception_bucket *)p;
1652 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1653 }
1654
1655 static int rt6_insert_exception(struct rt6_info *nrt,
1656 const struct fib6_result *res)
1657 {
1658 struct net *net = dev_net(nrt->dst.dev);
1659 struct rt6_exception_bucket *bucket;
1660 struct fib6_info *f6i = res->f6i;
1661 struct in6_addr *src_key = NULL;
1662 struct rt6_exception *rt6_ex;
1663 struct fib6_nh *nh = res->nh;
1664 int err = 0;
1665
1666 spin_lock_bh(&rt6_exception_lock);
1667
1668 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1669 lockdep_is_held(&rt6_exception_lock));
1670 if (!bucket) {
1671 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1672 GFP_ATOMIC);
1673 if (!bucket) {
1674 err = -ENOMEM;
1675 goto out;
1676 }
1677 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1678 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1679 err = -EINVAL;
1680 goto out;
1681 }
1682
1683 #ifdef CONFIG_IPV6_SUBTREES
1684 /* fib6_src.plen != 0 indicates f6i is in subtree
1685 * and exception table is indexed by a hash of
1686 * both fib6_dst and fib6_src.
1687 * Otherwise, the exception table is indexed by
1688 * a hash of only fib6_dst.
1689 */
1690 if (f6i->fib6_src.plen)
1691 src_key = &nrt->rt6i_src.addr;
1692 #endif
1693 /* rt6_mtu_change() might lower mtu on f6i.
1694 * Only insert this exception route if its mtu
1695 * is less than f6i's mtu value.
1696 */
1697 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
1698 err = -EINVAL;
1699 goto out;
1700 }
1701
1702 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1703 src_key);
1704 if (rt6_ex)
1705 rt6_remove_exception(bucket, rt6_ex);
1706
1707 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1708 if (!rt6_ex) {
1709 err = -ENOMEM;
1710 goto out;
1711 }
1712 rt6_ex->rt6i = nrt;
1713 rt6_ex->stamp = jiffies;
1714 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1715 bucket->depth++;
1716 net->ipv6.rt6_stats->fib_rt_cache++;
1717
1718 if (bucket->depth > FIB6_MAX_DEPTH)
1719 rt6_exception_remove_oldest(bucket);
1720
1721 out:
1722 spin_unlock_bh(&rt6_exception_lock);
1723
1724 /* Update fn->fn_sernum to invalidate all cached dst */
1725 if (!err) {
1726 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1727 fib6_update_sernum(net, f6i);
1728 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
1729 fib6_force_start_gc(net);
1730 }
1731
1732 return err;
1733 }
1734
1735 static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
1736 {
1737 struct rt6_exception_bucket *bucket;
1738 struct rt6_exception *rt6_ex;
1739 struct hlist_node *tmp;
1740 int i;
1741
1742 spin_lock_bh(&rt6_exception_lock);
1743
1744 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1745 if (!bucket)
1746 goto out;
1747
1748 /* Prevent rt6_insert_exception() to recreate the bucket list */
1749 if (!from)
1750 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1751
1752 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1753 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1754 if (!from ||
1755 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1756 rt6_remove_exception(bucket, rt6_ex);
1757 }
1758 WARN_ON_ONCE(!from && bucket->depth);
1759 bucket++;
1760 }
1761 out:
1762 spin_unlock_bh(&rt6_exception_lock);
1763 }
1764
1765 static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1766 {
1767 struct fib6_info *f6i = arg;
1768
1769 fib6_nh_flush_exceptions(nh, f6i);
1770
1771 return 0;
1772 }
1773
1774 void rt6_flush_exceptions(struct fib6_info *f6i)
1775 {
1776 if (f6i->nh)
1777 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1778 f6i);
1779 else
1780 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
1781 }
1782
1783 /* Find cached rt in the hash table inside passed in rt
1784 * Caller has to hold rcu_read_lock()
1785 */
1786 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
1787 const struct in6_addr *daddr,
1788 const struct in6_addr *saddr)
1789 {
1790 const struct in6_addr *src_key = NULL;
1791 struct rt6_exception_bucket *bucket;
1792 struct rt6_exception *rt6_ex;
1793 struct rt6_info *ret = NULL;
1794
1795 #ifdef CONFIG_IPV6_SUBTREES
1796 /* fib6i_src.plen != 0 indicates f6i is in subtree
1797 * and exception table is indexed by a hash of
1798 * both fib6_dst and fib6_src.
1799 * However, the src addr used to create the hash
1800 * might not be exactly the passed in saddr which
1801 * is a /128 addr from the flow.
1802 * So we need to use f6i->fib6_src to redo lookup
1803 * if the passed in saddr does not find anything.
1804 * (See the logic in ip6_rt_cache_alloc() on how
1805 * rt->rt6i_src is updated.)
1806 */
1807 if (res->f6i->fib6_src.plen)
1808 src_key = saddr;
1809 find_ex:
1810 #endif
1811 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
1812 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1813
1814 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1815 ret = rt6_ex->rt6i;
1816
1817 #ifdef CONFIG_IPV6_SUBTREES
1818 /* Use fib6_src as src_key and redo lookup */
1819 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1820 src_key = &res->f6i->fib6_src.addr;
1821 goto find_ex;
1822 }
1823 #endif
1824
1825 return ret;
1826 }
1827
1828 /* Remove the passed in cached rt from the hash table that contains it */
1829 static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
1830 const struct rt6_info *rt)
1831 {
1832 const struct in6_addr *src_key = NULL;
1833 struct rt6_exception_bucket *bucket;
1834 struct rt6_exception *rt6_ex;
1835 int err;
1836
1837 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
1838 return -ENOENT;
1839
1840 spin_lock_bh(&rt6_exception_lock);
1841 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1842
1843 #ifdef CONFIG_IPV6_SUBTREES
1844 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1845 * and exception table is indexed by a hash of
1846 * both rt6i_dst and rt6i_src.
1847 * Otherwise, the exception table is indexed by
1848 * a hash of only rt6i_dst.
1849 */
1850 if (plen)
1851 src_key = &rt->rt6i_src.addr;
1852 #endif
1853 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1854 &rt->rt6i_dst.addr,
1855 src_key);
1856 if (rt6_ex) {
1857 rt6_remove_exception(bucket, rt6_ex);
1858 err = 0;
1859 } else {
1860 err = -ENOENT;
1861 }
1862
1863 spin_unlock_bh(&rt6_exception_lock);
1864 return err;
1865 }
1866
1867 struct fib6_nh_excptn_arg {
1868 struct rt6_info *rt;
1869 int plen;
1870 };
1871
1872 static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1873 {
1874 struct fib6_nh_excptn_arg *arg = _arg;
1875 int err;
1876
1877 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1878 if (err == 0)
1879 return 1;
1880
1881 return 0;
1882 }
1883
1884 static int rt6_remove_exception_rt(struct rt6_info *rt)
1885 {
1886 struct fib6_info *from;
1887
1888 from = rcu_dereference(rt->from);
1889 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1890 return -EINVAL;
1891
1892 if (from->nh) {
1893 struct fib6_nh_excptn_arg arg = {
1894 .rt = rt,
1895 .plen = from->fib6_src.plen
1896 };
1897 int rc;
1898
1899 /* rc = 1 means an entry was found */
1900 rc = nexthop_for_each_fib6_nh(from->nh,
1901 rt6_nh_remove_exception_rt,
1902 &arg);
1903 return rc ? 0 : -ENOENT;
1904 }
1905
1906 return fib6_nh_remove_exception(from->fib6_nh,
1907 from->fib6_src.plen, rt);
1908 }
1909
1910 /* Find rt6_ex which contains the passed in rt cache and
1911 * refresh its stamp
1912 */
1913 static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
1914 const struct rt6_info *rt)
1915 {
1916 const struct in6_addr *src_key = NULL;
1917 struct rt6_exception_bucket *bucket;
1918 struct rt6_exception *rt6_ex;
1919
1920 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
1921 #ifdef CONFIG_IPV6_SUBTREES
1922 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1923 * and exception table is indexed by a hash of
1924 * both rt6i_dst and rt6i_src.
1925 * Otherwise, the exception table is indexed by
1926 * a hash of only rt6i_dst.
1927 */
1928 if (plen)
1929 src_key = &rt->rt6i_src.addr;
1930 #endif
1931 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
1932 if (rt6_ex)
1933 rt6_ex->stamp = jiffies;
1934 }
1935
1936 struct fib6_nh_match_arg {
1937 const struct net_device *dev;
1938 const struct in6_addr *gw;
1939 struct fib6_nh *match;
1940 };
1941
1942 /* determine if fib6_nh has given device and gateway */
1943 static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1944 {
1945 struct fib6_nh_match_arg *arg = _arg;
1946
1947 if (arg->dev != nh->fib_nh_dev ||
1948 (arg->gw && !nh->fib_nh_gw_family) ||
1949 (!arg->gw && nh->fib_nh_gw_family) ||
1950 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1951 return 0;
1952
1953 arg->match = nh;
1954
1955 /* found a match, break the loop */
1956 return 1;
1957 }
1958
1959 static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1960 {
1961 struct fib6_info *from;
1962 struct fib6_nh *fib6_nh;
1963
1964 rcu_read_lock();
1965
1966 from = rcu_dereference(rt->from);
1967 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1968 goto unlock;
1969
1970 if (from->nh) {
1971 struct fib6_nh_match_arg arg = {
1972 .dev = rt->dst.dev,
1973 .gw = &rt->rt6i_gateway,
1974 };
1975
1976 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1977
1978 if (!arg.match)
1979 goto unlock;
1980 fib6_nh = arg.match;
1981 } else {
1982 fib6_nh = from->fib6_nh;
1983 }
1984 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
1985 unlock:
1986 rcu_read_unlock();
1987 }
1988
1989 static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1990 struct rt6_info *rt, int mtu)
1991 {
1992 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1993 * lowest MTU in the path: always allow updating the route PMTU to
1994 * reflect PMTU decreases.
1995 *
1996 * If the new MTU is higher, and the route PMTU is equal to the local
1997 * MTU, this means the old MTU is the lowest in the path, so allow
1998 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1999 * handle this.
2000 */
2001
2002 if (dst_mtu(&rt->dst) >= mtu)
2003 return true;
2004
2005 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
2006 return true;
2007
2008 return false;
2009 }
2010
2011 static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
2012 const struct fib6_nh *nh, int mtu)
2013 {
2014 struct rt6_exception_bucket *bucket;
2015 struct rt6_exception *rt6_ex;
2016 int i;
2017
2018 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2019 if (!bucket)
2020 return;
2021
2022 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2023 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2024 struct rt6_info *entry = rt6_ex->rt6i;
2025
2026 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
2027 * route), the metrics of its rt->from have already
2028 * been updated.
2029 */
2030 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
2031 rt6_mtu_change_route_allowed(idev, entry, mtu))
2032 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
2033 }
2034 bucket++;
2035 }
2036 }
2037
2038 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2039
2040 static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2041 const struct in6_addr *gateway)
2042 {
2043 struct rt6_exception_bucket *bucket;
2044 struct rt6_exception *rt6_ex;
2045 struct hlist_node *tmp;
2046 int i;
2047
2048 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2049 return;
2050
2051 spin_lock_bh(&rt6_exception_lock);
2052 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2053 if (bucket) {
2054 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2055 hlist_for_each_entry_safe(rt6_ex, tmp,
2056 &bucket->chain, hlist) {
2057 struct rt6_info *entry = rt6_ex->rt6i;
2058
2059 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2060 RTF_CACHE_GATEWAY &&
2061 ipv6_addr_equal(gateway,
2062 &entry->rt6i_gateway)) {
2063 rt6_remove_exception(bucket, rt6_ex);
2064 }
2065 }
2066 bucket++;
2067 }
2068 }
2069
2070 spin_unlock_bh(&rt6_exception_lock);
2071 }
2072
2073 static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2074 struct rt6_exception *rt6_ex,
2075 struct fib6_gc_args *gc_args,
2076 unsigned long now)
2077 {
2078 struct rt6_info *rt = rt6_ex->rt6i;
2079
2080 /* we are pruning and obsoleting aged-out and non gateway exceptions
2081 * even if others have still references to them, so that on next
2082 * dst_check() such references can be dropped.
2083 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2084 * expired, independently from their aging, as per RFC 8201 section 4
2085 */
2086 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2087 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2088 RT6_TRACE("aging clone %p\n", rt);
2089 rt6_remove_exception(bucket, rt6_ex);
2090 return;
2091 }
2092 } else if (time_after(jiffies, rt->dst.expires)) {
2093 RT6_TRACE("purging expired route %p\n", rt);
2094 rt6_remove_exception(bucket, rt6_ex);
2095 return;
2096 }
2097
2098 if (rt->rt6i_flags & RTF_GATEWAY) {
2099 struct neighbour *neigh;
2100 __u8 neigh_flags = 0;
2101
2102 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2103 if (neigh)
2104 neigh_flags = neigh->flags;
2105
2106 if (!(neigh_flags & NTF_ROUTER)) {
2107 RT6_TRACE("purging route %p via non-router but gateway\n",
2108 rt);
2109 rt6_remove_exception(bucket, rt6_ex);
2110 return;
2111 }
2112 }
2113
2114 gc_args->more++;
2115 }
2116
2117 static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
2118 struct fib6_gc_args *gc_args,
2119 unsigned long now)
2120 {
2121 struct rt6_exception_bucket *bucket;
2122 struct rt6_exception *rt6_ex;
2123 struct hlist_node *tmp;
2124 int i;
2125
2126 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2127 return;
2128
2129 rcu_read_lock_bh();
2130 spin_lock(&rt6_exception_lock);
2131 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2132 if (bucket) {
2133 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2134 hlist_for_each_entry_safe(rt6_ex, tmp,
2135 &bucket->chain, hlist) {
2136 rt6_age_examine_exception(bucket, rt6_ex,
2137 gc_args, now);
2138 }
2139 bucket++;
2140 }
2141 }
2142 spin_unlock(&rt6_exception_lock);
2143 rcu_read_unlock_bh();
2144 }
2145
2146 struct fib6_nh_age_excptn_arg {
2147 struct fib6_gc_args *gc_args;
2148 unsigned long now;
2149 };
2150
2151 static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2152 {
2153 struct fib6_nh_age_excptn_arg *arg = _arg;
2154
2155 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2156 return 0;
2157 }
2158
2159 void rt6_age_exceptions(struct fib6_info *f6i,
2160 struct fib6_gc_args *gc_args,
2161 unsigned long now)
2162 {
2163 if (f6i->nh) {
2164 struct fib6_nh_age_excptn_arg arg = {
2165 .gc_args = gc_args,
2166 .now = now
2167 };
2168
2169 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2170 &arg);
2171 } else {
2172 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2173 }
2174 }
2175
2176 /* must be called with rcu lock held */
2177 int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2178 struct flowi6 *fl6, struct fib6_result *res, int strict)
2179 {
2180 struct fib6_node *fn, *saved_fn;
2181
2182 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2183 saved_fn = fn;
2184
2185 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2186 oif = 0;
2187
2188 redo_rt6_select:
2189 rt6_select(net, fn, oif, res, strict);
2190 if (res->f6i == net->ipv6.fib6_null_entry) {
2191 fn = fib6_backtrack(fn, &fl6->saddr);
2192 if (fn)
2193 goto redo_rt6_select;
2194 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2195 /* also consider unreachable route */
2196 strict &= ~RT6_LOOKUP_F_REACHABLE;
2197 fn = saved_fn;
2198 goto redo_rt6_select;
2199 }
2200 }
2201
2202 trace_fib6_table_lookup(net, res, table, fl6);
2203
2204 return 0;
2205 }
2206
2207 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2208 int oif, struct flowi6 *fl6,
2209 const struct sk_buff *skb, int flags)
2210 {
2211 struct fib6_result res = {};
2212 struct rt6_info *rt = NULL;
2213 int strict = 0;
2214
2215 WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2216 !rcu_read_lock_held());
2217
2218 strict |= flags & RT6_LOOKUP_F_IFACE;
2219 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2220 if (net->ipv6.devconf_all->forwarding == 0)
2221 strict |= RT6_LOOKUP_F_REACHABLE;
2222
2223 rcu_read_lock();
2224
2225 fib6_table_lookup(net, table, oif, fl6, &res, strict);
2226 if (res.f6i == net->ipv6.fib6_null_entry)
2227 goto out;
2228
2229 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
2230
2231 /*Search through exception table */
2232 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
2233 if (rt) {
2234 goto out;
2235 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
2236 !res.nh->fib_nh_gw_family)) {
2237 /* Create a RTF_CACHE clone which will not be
2238 * owned by the fib6 tree. It is for the special case where
2239 * the daddr in the skb during the neighbor look-up is different
2240 * from the fl6->daddr used to look-up route here.
2241 */
2242 rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
2243
2244 if (rt) {
2245 /* 1 refcnt is taken during ip6_rt_cache_alloc().
2246 * As rt6_uncached_list_add() does not consume refcnt,
2247 * this refcnt is always returned to the caller even
2248 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
2249 */
2250 rt6_uncached_list_add(rt);
2251 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
2252 rcu_read_unlock();
2253
2254 return rt;
2255 }
2256 } else {
2257 /* Get a percpu copy */
2258 local_bh_disable();
2259 rt = rt6_get_pcpu_route(&res);
2260
2261 if (!rt)
2262 rt = rt6_make_pcpu_route(net, &res);
2263
2264 local_bh_enable();
2265 }
2266 out:
2267 if (!rt)
2268 rt = net->ipv6.ip6_null_entry;
2269 if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2270 ip6_hold_safe(net, &rt);
2271 rcu_read_unlock();
2272
2273 return rt;
2274 }
2275 EXPORT_SYMBOL_GPL(ip6_pol_route);
2276
2277 static struct rt6_info *ip6_pol_route_input(struct net *net,
2278 struct fib6_table *table,
2279 struct flowi6 *fl6,
2280 const struct sk_buff *skb,
2281 int flags)
2282 {
2283 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
2284 }
2285
2286 struct dst_entry *ip6_route_input_lookup(struct net *net,
2287 struct net_device *dev,
2288 struct flowi6 *fl6,
2289 const struct sk_buff *skb,
2290 int flags)
2291 {
2292 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2293 flags |= RT6_LOOKUP_F_IFACE;
2294
2295 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
2296 }
2297 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
2298
2299 static void ip6_multipath_l3_keys(const struct sk_buff *skb,
2300 struct flow_keys *keys,
2301 struct flow_keys *flkeys)
2302 {
2303 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2304 const struct ipv6hdr *key_iph = outer_iph;
2305 struct flow_keys *_flkeys = flkeys;
2306 const struct ipv6hdr *inner_iph;
2307 const struct icmp6hdr *icmph;
2308 struct ipv6hdr _inner_iph;
2309 struct icmp6hdr _icmph;
2310
2311 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2312 goto out;
2313
2314 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2315 sizeof(_icmph), &_icmph);
2316 if (!icmph)
2317 goto out;
2318
2319 if (!icmpv6_is_err(icmph->icmp6_type))
2320 goto out;
2321
2322 inner_iph = skb_header_pointer(skb,
2323 skb_transport_offset(skb) + sizeof(*icmph),
2324 sizeof(_inner_iph), &_inner_iph);
2325 if (!inner_iph)
2326 goto out;
2327
2328 key_iph = inner_iph;
2329 _flkeys = NULL;
2330 out:
2331 if (_flkeys) {
2332 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2333 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2334 keys->tags.flow_label = _flkeys->tags.flow_label;
2335 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2336 } else {
2337 keys->addrs.v6addrs.src = key_iph->saddr;
2338 keys->addrs.v6addrs.dst = key_iph->daddr;
2339 keys->tags.flow_label = ip6_flowlabel(key_iph);
2340 keys->basic.ip_proto = key_iph->nexthdr;
2341 }
2342 }
2343
2344 /* if skb is set it will be used and fl6 can be NULL */
2345 u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2346 const struct sk_buff *skb, struct flow_keys *flkeys)
2347 {
2348 struct flow_keys hash_keys;
2349 u32 mhash;
2350
2351 switch (ip6_multipath_hash_policy(net)) {
2352 case 0:
2353 memset(&hash_keys, 0, sizeof(hash_keys));
2354 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2355 if (skb) {
2356 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2357 } else {
2358 hash_keys.addrs.v6addrs.src = fl6->saddr;
2359 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2360 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2361 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2362 }
2363 break;
2364 case 1:
2365 if (skb) {
2366 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2367 struct flow_keys keys;
2368
2369 /* short-circuit if we already have L4 hash present */
2370 if (skb->l4_hash)
2371 return skb_get_hash_raw(skb) >> 1;
2372
2373 memset(&hash_keys, 0, sizeof(hash_keys));
2374
2375 if (!flkeys) {
2376 skb_flow_dissect_flow_keys(skb, &keys, flag);
2377 flkeys = &keys;
2378 }
2379 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2380 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2381 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2382 hash_keys.ports.src = flkeys->ports.src;
2383 hash_keys.ports.dst = flkeys->ports.dst;
2384 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2385 } else {
2386 memset(&hash_keys, 0, sizeof(hash_keys));
2387 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2388 hash_keys.addrs.v6addrs.src = fl6->saddr;
2389 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2390 hash_keys.ports.src = fl6->fl6_sport;
2391 hash_keys.ports.dst = fl6->fl6_dport;
2392 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2393 }
2394 break;
2395 case 2:
2396 memset(&hash_keys, 0, sizeof(hash_keys));
2397 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2398 if (skb) {
2399 struct flow_keys keys;
2400
2401 if (!flkeys) {
2402 skb_flow_dissect_flow_keys(skb, &keys, 0);
2403 flkeys = &keys;
2404 }
2405
2406 /* Inner can be v4 or v6 */
2407 if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2408 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2409 hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2410 hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2411 } else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2412 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2413 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2414 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2415 hash_keys.tags.flow_label = flkeys->tags.flow_label;
2416 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2417 } else {
2418 /* Same as case 0 */
2419 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2420 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2421 }
2422 } else {
2423 /* Same as case 0 */
2424 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2425 hash_keys.addrs.v6addrs.src = fl6->saddr;
2426 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2427 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2428 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2429 }
2430 break;
2431 }
2432 mhash = flow_hash_from_keys(&hash_keys);
2433
2434 return mhash >> 1;
2435 }
2436
2437 /* Called with rcu held */
2438 void ip6_route_input(struct sk_buff *skb)
2439 {
2440 const struct ipv6hdr *iph = ipv6_hdr(skb);
2441 struct net *net = dev_net(skb->dev);
2442 int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
2443 struct ip_tunnel_info *tun_info;
2444 struct flowi6 fl6 = {
2445 .flowi6_iif = skb->dev->ifindex,
2446 .daddr = iph->daddr,
2447 .saddr = iph->saddr,
2448 .flowlabel = ip6_flowinfo(iph),
2449 .flowi6_mark = skb->mark,
2450 .flowi6_proto = iph->nexthdr,
2451 };
2452 struct flow_keys *flkeys = NULL, _flkeys;
2453
2454 tun_info = skb_tunnel_info(skb);
2455 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2456 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2457
2458 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2459 flkeys = &_flkeys;
2460
2461 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2462 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2463 skb_dst_drop(skb);
2464 skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2465 &fl6, skb, flags));
2466 }
2467
2468 static struct rt6_info *ip6_pol_route_output(struct net *net,
2469 struct fib6_table *table,
2470 struct flowi6 *fl6,
2471 const struct sk_buff *skb,
2472 int flags)
2473 {
2474 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2475 }
2476
2477 struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2478 const struct sock *sk,
2479 struct flowi6 *fl6, int flags)
2480 {
2481 bool any_src;
2482
2483 if (ipv6_addr_type(&fl6->daddr) &
2484 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
2485 struct dst_entry *dst;
2486
2487 /* This function does not take refcnt on the dst */
2488 dst = l3mdev_link_scope_lookup(net, fl6);
2489 if (dst)
2490 return dst;
2491 }
2492
2493 fl6->flowi6_iif = LOOPBACK_IFINDEX;
2494
2495 flags |= RT6_LOOKUP_F_DST_NOREF;
2496 any_src = ipv6_addr_any(&fl6->saddr);
2497 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
2498 (fl6->flowi6_oif && any_src))
2499 flags |= RT6_LOOKUP_F_IFACE;
2500
2501 if (!any_src)
2502 flags |= RT6_LOOKUP_F_HAS_SADDR;
2503 else if (sk)
2504 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
2505
2506 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
2507 }
2508 EXPORT_SYMBOL_GPL(ip6_route_output_flags_noref);
2509
2510 struct dst_entry *ip6_route_output_flags(struct net *net,
2511 const struct sock *sk,
2512 struct flowi6 *fl6,
2513 int flags)
2514 {
2515 struct dst_entry *dst;
2516 struct rt6_info *rt6;
2517
2518 rcu_read_lock();
2519 dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2520 rt6 = (struct rt6_info *)dst;
2521 /* For dst cached in uncached_list, refcnt is already taken. */
2522 if (list_empty(&rt6->rt6i_uncached) && !dst_hold_safe(dst)) {
2523 dst = &net->ipv6.ip6_null_entry->dst;
2524 dst_hold(dst);
2525 }
2526 rcu_read_unlock();
2527
2528 return dst;
2529 }
2530 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2531
2532 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2533 {
2534 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
2535 struct net_device *loopback_dev = net->loopback_dev;
2536 struct dst_entry *new = NULL;
2537
2538 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
2539 DST_OBSOLETE_DEAD, 0);
2540 if (rt) {
2541 rt6_info_init(rt);
2542 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2543
2544 new = &rt->dst;
2545 new->__use = 1;
2546 new->input = dst_discard;
2547 new->output = dst_discard_out;
2548
2549 dst_copy_metrics(new, &ort->dst);
2550
2551 rt->rt6i_idev = in6_dev_get(loopback_dev);
2552 rt->rt6i_gateway = ort->rt6i_gateway;
2553 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2554
2555 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2556 #ifdef CONFIG_IPV6_SUBTREES
2557 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2558 #endif
2559 }
2560
2561 dst_release(dst_orig);
2562 return new ? new : ERR_PTR(-ENOMEM);
2563 }
2564
2565 /*
2566 * Destination cache support functions
2567 */
2568
2569 static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2570 {
2571 u32 rt_cookie = 0;
2572
2573 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2574 return false;
2575
2576 if (fib6_check_expired(f6i))
2577 return false;
2578
2579 return true;
2580 }
2581
2582 static struct dst_entry *rt6_check(struct rt6_info *rt,
2583 struct fib6_info *from,
2584 u32 cookie)
2585 {
2586 u32 rt_cookie = 0;
2587
2588 if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
2589 rt_cookie != cookie)
2590 return NULL;
2591
2592 if (rt6_check_expired(rt))
2593 return NULL;
2594
2595 return &rt->dst;
2596 }
2597
2598 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2599 struct fib6_info *from,
2600 u32 cookie)
2601 {
2602 if (!__rt6_check_expired(rt) &&
2603 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2604 fib6_check(from, cookie))
2605 return &rt->dst;
2606 else
2607 return NULL;
2608 }
2609
2610 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
2611 {
2612 struct dst_entry *dst_ret;
2613 struct fib6_info *from;
2614 struct rt6_info *rt;
2615
2616 rt = container_of(dst, struct rt6_info, dst);
2617
2618 if (rt->sernum)
2619 return rt6_is_valid(rt) ? dst : NULL;
2620
2621 rcu_read_lock();
2622
2623 /* All IPV6 dsts are created with ->obsolete set to the value
2624 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2625 * into this function always.
2626 */
2627
2628 from = rcu_dereference(rt->from);
2629
2630 if (from && (rt->rt6i_flags & RTF_PCPU ||
2631 unlikely(!list_empty(&rt->rt6i_uncached))))
2632 dst_ret = rt6_dst_from_check(rt, from, cookie);
2633 else
2634 dst_ret = rt6_check(rt, from, cookie);
2635
2636 rcu_read_unlock();
2637
2638 return dst_ret;
2639 }
2640
2641 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2642 {
2643 struct rt6_info *rt = (struct rt6_info *) dst;
2644
2645 if (rt) {
2646 if (rt->rt6i_flags & RTF_CACHE) {
2647 rcu_read_lock();
2648 if (rt6_check_expired(rt)) {
2649 rt6_remove_exception_rt(rt);
2650 dst = NULL;
2651 }
2652 rcu_read_unlock();
2653 } else {
2654 dst_release(dst);
2655 dst = NULL;
2656 }
2657 }
2658 return dst;
2659 }
2660
2661 static void ip6_link_failure(struct sk_buff *skb)
2662 {
2663 struct rt6_info *rt;
2664
2665 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2666
2667 rt = (struct rt6_info *) skb_dst(skb);
2668 if (rt) {
2669 rcu_read_lock();
2670 if (rt->rt6i_flags & RTF_CACHE) {
2671 rt6_remove_exception_rt(rt);
2672 } else {
2673 struct fib6_info *from;
2674 struct fib6_node *fn;
2675
2676 from = rcu_dereference(rt->from);
2677 if (from) {
2678 fn = rcu_dereference(from->fib6_node);
2679 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2680 fn->fn_sernum = -1;
2681 }
2682 }
2683 rcu_read_unlock();
2684 }
2685 }
2686
2687 static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2688 {
2689 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2690 struct fib6_info *from;
2691
2692 rcu_read_lock();
2693 from = rcu_dereference(rt0->from);
2694 if (from)
2695 rt0->dst.expires = from->expires;
2696 rcu_read_unlock();
2697 }
2698
2699 dst_set_expires(&rt0->dst, timeout);
2700 rt0->rt6i_flags |= RTF_EXPIRES;
2701 }
2702
2703 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2704 {
2705 struct net *net = dev_net(rt->dst.dev);
2706
2707 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2708 rt->rt6i_flags |= RTF_MODIFIED;
2709 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2710 }
2711
2712 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2713 {
2714 return !(rt->rt6i_flags & RTF_CACHE) &&
2715 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
2716 }
2717
2718 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2719 const struct ipv6hdr *iph, u32 mtu,
2720 bool confirm_neigh)
2721 {
2722 const struct in6_addr *daddr, *saddr;
2723 struct rt6_info *rt6 = (struct rt6_info *)dst;
2724
2725 if (dst_metric_locked(dst, RTAX_MTU))
2726 return;
2727
2728 if (iph) {
2729 daddr = &iph->daddr;
2730 saddr = &iph->saddr;
2731 } else if (sk) {
2732 daddr = &sk->sk_v6_daddr;
2733 saddr = &inet6_sk(sk)->saddr;
2734 } else {
2735 daddr = NULL;
2736 saddr = NULL;
2737 }
2738
2739 if (confirm_neigh)
2740 dst_confirm_neigh(dst, daddr);
2741
2742 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
2743 if (mtu >= dst_mtu(dst))
2744 return;
2745
2746 if (!rt6_cache_allowed_for_pmtu(rt6)) {
2747 rt6_do_update_pmtu(rt6, mtu);
2748 /* update rt6_ex->stamp for cache */
2749 if (rt6->rt6i_flags & RTF_CACHE)
2750 rt6_update_exception_stamp_rt(rt6);
2751 } else if (daddr) {
2752 struct fib6_result res = {};
2753 struct rt6_info *nrt6;
2754
2755 rcu_read_lock();
2756 res.f6i = rcu_dereference(rt6->from);
2757 if (!res.f6i)
2758 goto out_unlock;
2759
2760 res.fib6_flags = res.f6i->fib6_flags;
2761 res.fib6_type = res.f6i->fib6_type;
2762
2763 if (res.f6i->nh) {
2764 struct fib6_nh_match_arg arg = {
2765 .dev = dst->dev,
2766 .gw = &rt6->rt6i_gateway,
2767 };
2768
2769 nexthop_for_each_fib6_nh(res.f6i->nh,
2770 fib6_nh_find_match, &arg);
2771
2772 /* fib6_info uses a nexthop that does not have fib6_nh
2773 * using the dst->dev + gw. Should be impossible.
2774 */
2775 if (!arg.match)
2776 goto out_unlock;
2777
2778 res.nh = arg.match;
2779 } else {
2780 res.nh = res.f6i->fib6_nh;
2781 }
2782
2783 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
2784 if (nrt6) {
2785 rt6_do_update_pmtu(nrt6, mtu);
2786 if (rt6_insert_exception(nrt6, &res))
2787 dst_release_immediate(&nrt6->dst);
2788 }
2789 out_unlock:
2790 rcu_read_unlock();
2791 }
2792 }
2793
2794 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2795 struct sk_buff *skb, u32 mtu,
2796 bool confirm_neigh)
2797 {
2798 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2799 confirm_neigh);
2800 }
2801
2802 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2803 int oif, u32 mark, kuid_t uid)
2804 {
2805 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2806 struct dst_entry *dst;
2807 struct flowi6 fl6 = {
2808 .flowi6_oif = oif,
2809 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2810 .daddr = iph->daddr,
2811 .saddr = iph->saddr,
2812 .flowlabel = ip6_flowinfo(iph),
2813 .flowi6_uid = uid,
2814 };
2815
2816 dst = ip6_route_output(net, NULL, &fl6);
2817 if (!dst->error)
2818 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
2819 dst_release(dst);
2820 }
2821 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2822
2823 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2824 {
2825 int oif = sk->sk_bound_dev_if;
2826 struct dst_entry *dst;
2827
2828 if (!oif && skb->dev)
2829 oif = l3mdev_master_ifindex(skb->dev);
2830
2831 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
2832
2833 dst = __sk_dst_get(sk);
2834 if (!dst || !dst->obsolete ||
2835 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2836 return;
2837
2838 bh_lock_sock(sk);
2839 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2840 ip6_datagram_dst_update(sk, false);
2841 bh_unlock_sock(sk);
2842 }
2843 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2844
2845 void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2846 const struct flowi6 *fl6)
2847 {
2848 #ifdef CONFIG_IPV6_SUBTREES
2849 struct ipv6_pinfo *np = inet6_sk(sk);
2850 #endif
2851
2852 ip6_dst_store(sk, dst,
2853 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2854 &sk->sk_v6_daddr : NULL,
2855 #ifdef CONFIG_IPV6_SUBTREES
2856 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2857 &np->saddr :
2858 #endif
2859 NULL);
2860 }
2861
2862 static bool ip6_redirect_nh_match(const struct fib6_result *res,
2863 struct flowi6 *fl6,
2864 const struct in6_addr *gw,
2865 struct rt6_info **ret)
2866 {
2867 const struct fib6_nh *nh = res->nh;
2868
2869 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2870 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2871 return false;
2872
2873 /* rt_cache's gateway might be different from its 'parent'
2874 * in the case of an ip redirect.
2875 * So we keep searching in the exception table if the gateway
2876 * is different.
2877 */
2878 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
2879 struct rt6_info *rt_cache;
2880
2881 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
2882 if (rt_cache &&
2883 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
2884 *ret = rt_cache;
2885 return true;
2886 }
2887 return false;
2888 }
2889 return true;
2890 }
2891
2892 struct fib6_nh_rd_arg {
2893 struct fib6_result *res;
2894 struct flowi6 *fl6;
2895 const struct in6_addr *gw;
2896 struct rt6_info **ret;
2897 };
2898
2899 static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
2900 {
2901 struct fib6_nh_rd_arg *arg = _arg;
2902
2903 arg->res->nh = nh;
2904 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
2905 }
2906
2907 /* Handle redirects */
2908 struct ip6rd_flowi {
2909 struct flowi6 fl6;
2910 struct in6_addr gateway;
2911 };
2912
2913 static struct rt6_info *__ip6_route_redirect(struct net *net,
2914 struct fib6_table *table,
2915 struct flowi6 *fl6,
2916 const struct sk_buff *skb,
2917 int flags)
2918 {
2919 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
2920 struct rt6_info *ret = NULL;
2921 struct fib6_result res = {};
2922 struct fib6_nh_rd_arg arg = {
2923 .res = &res,
2924 .fl6 = fl6,
2925 .gw = &rdfl->gateway,
2926 .ret = &ret
2927 };
2928 struct fib6_info *rt;
2929 struct fib6_node *fn;
2930
2931 /* l3mdev_update_flow overrides oif if the device is enslaved; in
2932 * this case we must match on the real ingress device, so reset it
2933 */
2934 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2935 fl6->flowi6_oif = skb->dev->ifindex;
2936
2937 /* Get the "current" route for this destination and
2938 * check if the redirect has come from appropriate router.
2939 *
2940 * RFC 4861 specifies that redirects should only be
2941 * accepted if they come from the nexthop to the target.
2942 * Due to the way the routes are chosen, this notion
2943 * is a bit fuzzy and one might need to check all possible
2944 * routes.
2945 */
2946
2947 rcu_read_lock();
2948 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2949 restart:
2950 for_each_fib6_node_rt_rcu(fn) {
2951 res.f6i = rt;
2952 if (fib6_check_expired(rt))
2953 continue;
2954 if (rt->fib6_flags & RTF_REJECT)
2955 break;
2956 if (unlikely(rt->nh)) {
2957 if (nexthop_is_blackhole(rt->nh))
2958 continue;
2959 /* on match, res->nh is filled in and potentially ret */
2960 if (nexthop_for_each_fib6_nh(rt->nh,
2961 fib6_nh_redirect_match,
2962 &arg))
2963 goto out;
2964 } else {
2965 res.nh = rt->fib6_nh;
2966 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
2967 &ret))
2968 goto out;
2969 }
2970 }
2971
2972 if (!rt)
2973 rt = net->ipv6.fib6_null_entry;
2974 else if (rt->fib6_flags & RTF_REJECT) {
2975 ret = net->ipv6.ip6_null_entry;
2976 goto out;
2977 }
2978
2979 if (rt == net->ipv6.fib6_null_entry) {
2980 fn = fib6_backtrack(fn, &fl6->saddr);
2981 if (fn)
2982 goto restart;
2983 }
2984
2985 res.f6i = rt;
2986 res.nh = rt->fib6_nh;
2987 out:
2988 if (ret) {
2989 ip6_hold_safe(net, &ret);
2990 } else {
2991 res.fib6_flags = res.f6i->fib6_flags;
2992 res.fib6_type = res.f6i->fib6_type;
2993 ret = ip6_create_rt_rcu(&res);
2994 }
2995
2996 rcu_read_unlock();
2997
2998 trace_fib6_table_lookup(net, &res, table, fl6);
2999 return ret;
3000 };
3001
3002 static struct dst_entry *ip6_route_redirect(struct net *net,
3003 const struct flowi6 *fl6,
3004 const struct sk_buff *skb,
3005 const struct in6_addr *gateway)
3006 {
3007 int flags = RT6_LOOKUP_F_HAS_SADDR;
3008 struct ip6rd_flowi rdfl;
3009
3010 rdfl.fl6 = *fl6;
3011 rdfl.gateway = *gateway;
3012
3013 return fib6_rule_lookup(net, &rdfl.fl6, skb,
3014 flags, __ip6_route_redirect);
3015 }
3016
3017 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
3018 kuid_t uid)
3019 {
3020 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
3021 struct dst_entry *dst;
3022 struct flowi6 fl6 = {
3023 .flowi6_iif = LOOPBACK_IFINDEX,
3024 .flowi6_oif = oif,
3025 .flowi6_mark = mark,
3026 .daddr = iph->daddr,
3027 .saddr = iph->saddr,
3028 .flowlabel = ip6_flowinfo(iph),
3029 .flowi6_uid = uid,
3030 };
3031
3032 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
3033 rt6_do_redirect(dst, NULL, skb);
3034 dst_release(dst);
3035 }
3036 EXPORT_SYMBOL_GPL(ip6_redirect);
3037
3038 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
3039 {
3040 const struct ipv6hdr *iph = ipv6_hdr(skb);
3041 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3042 struct dst_entry *dst;
3043 struct flowi6 fl6 = {
3044 .flowi6_iif = LOOPBACK_IFINDEX,
3045 .flowi6_oif = oif,
3046 .daddr = msg->dest,
3047 .saddr = iph->daddr,
3048 .flowi6_uid = sock_net_uid(net, NULL),
3049 };
3050
3051 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
3052 rt6_do_redirect(dst, NULL, skb);
3053 dst_release(dst);
3054 }
3055
3056 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3057 {
3058 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
3059 sk->sk_uid);
3060 }
3061 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3062
3063 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
3064 {
3065 struct net_device *dev = dst->dev;
3066 unsigned int mtu = dst_mtu(dst);
3067 struct net *net = dev_net(dev);
3068
3069 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3070
3071 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3072 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3073
3074 /*
3075 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3076 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3077 * IPV6_MAXPLEN is also valid and means: "any MSS,
3078 * rely only on pmtu discovery"
3079 */
3080 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3081 mtu = IPV6_MAXPLEN;
3082 return mtu;
3083 }
3084
3085 static unsigned int ip6_mtu(const struct dst_entry *dst)
3086 {
3087 struct inet6_dev *idev;
3088 unsigned int mtu;
3089
3090 mtu = dst_metric_raw(dst, RTAX_MTU);
3091 if (mtu)
3092 goto out;
3093
3094 mtu = IPV6_MIN_MTU;
3095
3096 rcu_read_lock();
3097 idev = __in6_dev_get(dst->dev);
3098 if (idev)
3099 mtu = idev->cnf.mtu6;
3100 rcu_read_unlock();
3101
3102 out:
3103 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3104
3105 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
3106 }
3107
3108 /* MTU selection:
3109 * 1. mtu on route is locked - use it
3110 * 2. mtu from nexthop exception
3111 * 3. mtu from egress device
3112 *
3113 * based on ip6_dst_mtu_forward and exception logic of
3114 * rt6_find_cached_rt; called with rcu_read_lock
3115 */
3116 u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3117 const struct in6_addr *daddr,
3118 const struct in6_addr *saddr)
3119 {
3120 const struct fib6_nh *nh = res->nh;
3121 struct fib6_info *f6i = res->f6i;
3122 struct inet6_dev *idev;
3123 struct rt6_info *rt;
3124 u32 mtu = 0;
3125
3126 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3127 mtu = f6i->fib6_pmtu;
3128 if (mtu)
3129 goto out;
3130 }
3131
3132 rt = rt6_find_cached_rt(res, daddr, saddr);
3133 if (unlikely(rt)) {
3134 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3135 } else {
3136 struct net_device *dev = nh->fib_nh_dev;
3137
3138 mtu = IPV6_MIN_MTU;
3139 idev = __in6_dev_get(dev);
3140 if (idev && idev->cnf.mtu6 > mtu)
3141 mtu = idev->cnf.mtu6;
3142 }
3143
3144 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3145 out:
3146 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3147 }
3148
3149 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3150 struct flowi6 *fl6)
3151 {
3152 struct dst_entry *dst;
3153 struct rt6_info *rt;
3154 struct inet6_dev *idev = in6_dev_get(dev);
3155 struct net *net = dev_net(dev);
3156
3157 if (unlikely(!idev))
3158 return ERR_PTR(-ENODEV);
3159
3160 rt = ip6_dst_alloc(net, dev, 0);
3161 if (unlikely(!rt)) {
3162 in6_dev_put(idev);
3163 dst = ERR_PTR(-ENOMEM);
3164 goto out;
3165 }
3166
3167 rt->dst.input = ip6_input;
3168 rt->dst.output = ip6_output;
3169 rt->rt6i_gateway = fl6->daddr;
3170 rt->rt6i_dst.addr = fl6->daddr;
3171 rt->rt6i_dst.plen = 128;
3172 rt->rt6i_idev = idev;
3173 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3174
3175 /* Add this dst into uncached_list so that rt6_disable_ip() can
3176 * do proper release of the net_device
3177 */
3178 rt6_uncached_list_add(rt);
3179 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
3180
3181 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3182
3183 out:
3184 return dst;
3185 }
3186
3187 static int ip6_dst_gc(struct dst_ops *ops)
3188 {
3189 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3190 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3191 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3192 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3193 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3194 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3195 int entries;
3196
3197 entries = dst_entries_get_fast(ops);
3198 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
3199 entries <= rt_max_size)
3200 goto out;
3201
3202 net->ipv6.ip6_rt_gc_expire++;
3203 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
3204 entries = dst_entries_get_slow(ops);
3205 if (entries < ops->gc_thresh)
3206 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
3207 out:
3208 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
3209 return entries > rt_max_size;
3210 }
3211
3212 static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3213 const struct in6_addr *gw_addr, u32 tbid,
3214 int flags, struct fib6_result *res)
3215 {
3216 struct flowi6 fl6 = {
3217 .flowi6_oif = cfg->fc_ifindex,
3218 .daddr = *gw_addr,
3219 .saddr = cfg->fc_prefsrc,
3220 };
3221 struct fib6_table *table;
3222 int err;
3223
3224 table = fib6_get_table(net, tbid);
3225 if (!table)
3226 return -EINVAL;
3227
3228 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3229 flags |= RT6_LOOKUP_F_HAS_SADDR;
3230
3231 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3232
3233 err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3234 if (!err && res->f6i != net->ipv6.fib6_null_entry)
3235 fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3236 cfg->fc_ifindex != 0, NULL, flags);
3237
3238 return err;
3239 }
3240
3241 static int ip6_route_check_nh_onlink(struct net *net,
3242 struct fib6_config *cfg,
3243 const struct net_device *dev,
3244 struct netlink_ext_ack *extack)
3245 {
3246 u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3247 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3248 struct fib6_result res = {};
3249 int err;
3250
3251 err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3252 if (!err && !(res.fib6_flags & RTF_REJECT) &&
3253 /* ignore match if it is the default route */
3254 !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3255 (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3256 NL_SET_ERR_MSG(extack,
3257 "Nexthop has invalid gateway or device mismatch");
3258 err = -EINVAL;
3259 }
3260
3261 return err;
3262 }
3263
3264 static int ip6_route_check_nh(struct net *net,
3265 struct fib6_config *cfg,
3266 struct net_device **_dev,
3267 struct inet6_dev **idev)
3268 {
3269 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3270 struct net_device *dev = _dev ? *_dev : NULL;
3271 int flags = RT6_LOOKUP_F_IFACE;
3272 struct fib6_result res = {};
3273 int err = -EHOSTUNREACH;
3274
3275 if (cfg->fc_table) {
3276 err = ip6_nh_lookup_table(net, cfg, gw_addr,
3277 cfg->fc_table, flags, &res);
3278 /* gw_addr can not require a gateway or resolve to a reject
3279 * route. If a device is given, it must match the result.
3280 */
3281 if (err || res.fib6_flags & RTF_REJECT ||
3282 res.nh->fib_nh_gw_family ||
3283 (dev && dev != res.nh->fib_nh_dev))
3284 err = -EHOSTUNREACH;
3285 }
3286
3287 if (err < 0) {
3288 struct flowi6 fl6 = {
3289 .flowi6_oif = cfg->fc_ifindex,
3290 .daddr = *gw_addr,
3291 };
3292
3293 err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3294 if (err || res.fib6_flags & RTF_REJECT ||
3295 res.nh->fib_nh_gw_family)
3296 err = -EHOSTUNREACH;
3297
3298 if (err)
3299 return err;
3300
3301 fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3302 cfg->fc_ifindex != 0, NULL, flags);
3303 }
3304
3305 err = 0;
3306 if (dev) {
3307 if (dev != res.nh->fib_nh_dev)
3308 err = -EHOSTUNREACH;
3309 } else {
3310 *_dev = dev = res.nh->fib_nh_dev;
3311 dev_hold(dev);
3312 *idev = in6_dev_get(dev);
3313 }
3314
3315 return err;
3316 }
3317
3318 static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3319 struct net_device **_dev, struct inet6_dev **idev,
3320 struct netlink_ext_ack *extack)
3321 {
3322 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3323 int gwa_type = ipv6_addr_type(gw_addr);
3324 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3325 const struct net_device *dev = *_dev;
3326 bool need_addr_check = !dev;
3327 int err = -EINVAL;
3328
3329 /* if gw_addr is local we will fail to detect this in case
3330 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3331 * will return already-added prefix route via interface that
3332 * prefix route was assigned to, which might be non-loopback.
3333 */
3334 if (dev &&
3335 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3336 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3337 goto out;
3338 }
3339
3340 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3341 /* IPv6 strictly inhibits using not link-local
3342 * addresses as nexthop address.
3343 * Otherwise, router will not able to send redirects.
3344 * It is very good, but in some (rare!) circumstances
3345 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3346 * some exceptions. --ANK
3347 * We allow IPv4-mapped nexthops to support RFC4798-type
3348 * addressing
3349 */
3350 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3351 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3352 goto out;
3353 }
3354
3355 rcu_read_lock();
3356
3357 if (cfg->fc_flags & RTNH_F_ONLINK)
3358 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3359 else
3360 err = ip6_route_check_nh(net, cfg, _dev, idev);
3361
3362 rcu_read_unlock();
3363
3364 if (err)
3365 goto out;
3366 }
3367
3368 /* reload in case device was changed */
3369 dev = *_dev;
3370
3371 err = -EINVAL;
3372 if (!dev) {
3373 NL_SET_ERR_MSG(extack, "Egress device not specified");
3374 goto out;
3375 } else if (dev->flags & IFF_LOOPBACK) {
3376 NL_SET_ERR_MSG(extack,
3377 "Egress device can not be loopback device for this route");
3378 goto out;
3379 }
3380
3381 /* if we did not check gw_addr above, do so now that the
3382 * egress device has been resolved.
3383 */
3384 if (need_addr_check &&
3385 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3386 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3387 goto out;
3388 }
3389
3390 err = 0;
3391 out:
3392 return err;
3393 }
3394
3395 static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3396 {
3397 if ((flags & RTF_REJECT) ||
3398 (dev && (dev->flags & IFF_LOOPBACK) &&
3399 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3400 !(flags & RTF_LOCAL)))
3401 return true;
3402
3403 return false;
3404 }
3405
3406 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3407 struct fib6_config *cfg, gfp_t gfp_flags,
3408 struct netlink_ext_ack *extack)
3409 {
3410 struct net_device *dev = NULL;
3411 struct inet6_dev *idev = NULL;
3412 int addr_type;
3413 int err;
3414
3415 fib6_nh->fib_nh_family = AF_INET6;
3416 #ifdef CONFIG_IPV6_ROUTER_PREF
3417 fib6_nh->last_probe = jiffies;
3418 #endif
3419
3420 err = -ENODEV;
3421 if (cfg->fc_ifindex) {
3422 dev = dev_get_by_index(net, cfg->fc_ifindex);
3423 if (!dev)
3424 goto out;
3425 idev = in6_dev_get(dev);
3426 if (!idev)
3427 goto out;
3428 }
3429
3430 if (cfg->fc_flags & RTNH_F_ONLINK) {
3431 if (!dev) {
3432 NL_SET_ERR_MSG(extack,
3433 "Nexthop device required for onlink");
3434 goto out;
3435 }
3436
3437 if (!(dev->flags & IFF_UP)) {
3438 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3439 err = -ENETDOWN;
3440 goto out;
3441 }
3442
3443 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3444 }
3445
3446 fib6_nh->fib_nh_weight = 1;
3447
3448 /* We cannot add true routes via loopback here,
3449 * they would result in kernel looping; promote them to reject routes
3450 */
3451 addr_type = ipv6_addr_type(&cfg->fc_dst);
3452 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3453 /* hold loopback dev/idev if we haven't done so. */
3454 if (dev != net->loopback_dev) {
3455 if (dev) {
3456 dev_put(dev);
3457 in6_dev_put(idev);
3458 }
3459 dev = net->loopback_dev;
3460 dev_hold(dev);
3461 idev = in6_dev_get(dev);
3462 if (!idev) {
3463 err = -ENODEV;
3464 goto out;
3465 }
3466 }
3467 goto pcpu_alloc;
3468 }
3469
3470 if (cfg->fc_flags & RTF_GATEWAY) {
3471 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3472 if (err)
3473 goto out;
3474
3475 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3476 fib6_nh->fib_nh_gw_family = AF_INET6;
3477 }
3478
3479 err = -ENODEV;
3480 if (!dev)
3481 goto out;
3482
3483 if (idev->cnf.disable_ipv6) {
3484 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3485 err = -EACCES;
3486 goto out;
3487 }
3488
3489 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3490 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3491 err = -ENETDOWN;
3492 goto out;
3493 }
3494
3495 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3496 !netif_carrier_ok(dev))
3497 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3498
3499 err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
3500 cfg->fc_encap_type, cfg, gfp_flags, extack);
3501 if (err)
3502 goto out;
3503
3504 pcpu_alloc:
3505 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3506 if (!fib6_nh->rt6i_pcpu) {
3507 err = -ENOMEM;
3508 goto out;
3509 }
3510
3511 fib6_nh->fib_nh_dev = dev;
3512 fib6_nh->fib_nh_oif = dev->ifindex;
3513 err = 0;
3514 out:
3515 if (idev)
3516 in6_dev_put(idev);
3517
3518 if (err) {
3519 lwtstate_put(fib6_nh->fib_nh_lws);
3520 fib6_nh->fib_nh_lws = NULL;
3521 if (dev)
3522 dev_put(dev);
3523 }
3524
3525 return err;
3526 }
3527
3528 void fib6_nh_release(struct fib6_nh *fib6_nh)
3529 {
3530 struct rt6_exception_bucket *bucket;
3531
3532 rcu_read_lock();
3533
3534 fib6_nh_flush_exceptions(fib6_nh, NULL);
3535 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3536 if (bucket) {
3537 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3538 kfree(bucket);
3539 }
3540
3541 rcu_read_unlock();
3542
3543 if (fib6_nh->rt6i_pcpu) {
3544 int cpu;
3545
3546 for_each_possible_cpu(cpu) {
3547 struct rt6_info **ppcpu_rt;
3548 struct rt6_info *pcpu_rt;
3549
3550 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3551 pcpu_rt = *ppcpu_rt;
3552 if (pcpu_rt) {
3553 dst_dev_put(&pcpu_rt->dst);
3554 dst_release(&pcpu_rt->dst);
3555 *ppcpu_rt = NULL;
3556 }
3557 }
3558
3559 free_percpu(fib6_nh->rt6i_pcpu);
3560 }
3561
3562 fib_nh_common_release(&fib6_nh->nh_common);
3563 }
3564
3565 static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3566 gfp_t gfp_flags,
3567 struct netlink_ext_ack *extack)
3568 {
3569 struct net *net = cfg->fc_nlinfo.nl_net;
3570 struct fib6_info *rt = NULL;
3571 struct nexthop *nh = NULL;
3572 struct fib6_table *table;
3573 struct fib6_nh *fib6_nh;
3574 int err = -EINVAL;
3575 int addr_type;
3576
3577 /* RTF_PCPU is an internal flag; can not be set by userspace */
3578 if (cfg->fc_flags & RTF_PCPU) {
3579 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3580 goto out;
3581 }
3582
3583 /* RTF_CACHE is an internal flag; can not be set by userspace */
3584 if (cfg->fc_flags & RTF_CACHE) {
3585 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3586 goto out;
3587 }
3588
3589 if (cfg->fc_type > RTN_MAX) {
3590 NL_SET_ERR_MSG(extack, "Invalid route type");
3591 goto out;
3592 }
3593
3594 if (cfg->fc_dst_len > 128) {
3595 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3596 goto out;
3597 }
3598 if (cfg->fc_src_len > 128) {
3599 NL_SET_ERR_MSG(extack, "Invalid source address length");
3600 goto out;
3601 }
3602 #ifndef CONFIG_IPV6_SUBTREES
3603 if (cfg->fc_src_len) {
3604 NL_SET_ERR_MSG(extack,
3605 "Specifying source address requires IPV6_SUBTREES to be enabled");
3606 goto out;
3607 }
3608 #endif
3609 if (cfg->fc_nh_id) {
3610 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3611 if (!nh) {
3612 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3613 goto out;
3614 }
3615 err = fib6_check_nexthop(nh, cfg, extack);
3616 if (err)
3617 goto out;
3618 }
3619
3620 err = -ENOBUFS;
3621 if (cfg->fc_nlinfo.nlh &&
3622 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3623 table = fib6_get_table(net, cfg->fc_table);
3624 if (!table) {
3625 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3626 table = fib6_new_table(net, cfg->fc_table);
3627 }
3628 } else {
3629 table = fib6_new_table(net, cfg->fc_table);
3630 }
3631
3632 if (!table)
3633 goto out;
3634
3635 err = -ENOMEM;
3636 rt = fib6_info_alloc(gfp_flags, !nh);
3637 if (!rt)
3638 goto out;
3639
3640 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3641 extack);
3642 if (IS_ERR(rt->fib6_metrics)) {
3643 err = PTR_ERR(rt->fib6_metrics);
3644 /* Do not leave garbage there. */
3645 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3646 goto out;
3647 }
3648
3649 if (cfg->fc_flags & RTF_ADDRCONF)
3650 rt->dst_nocount = true;
3651
3652 if (cfg->fc_flags & RTF_EXPIRES)
3653 fib6_set_expires(rt, jiffies +
3654 clock_t_to_jiffies(cfg->fc_expires));
3655 else
3656 fib6_clean_expires(rt);
3657
3658 if (cfg->fc_protocol == RTPROT_UNSPEC)
3659 cfg->fc_protocol = RTPROT_BOOT;
3660 rt->fib6_protocol = cfg->fc_protocol;
3661
3662 rt->fib6_table = table;
3663 rt->fib6_metric = cfg->fc_metric;
3664 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3665 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3666
3667 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3668 rt->fib6_dst.plen = cfg->fc_dst_len;
3669
3670 #ifdef CONFIG_IPV6_SUBTREES
3671 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3672 rt->fib6_src.plen = cfg->fc_src_len;
3673 #endif
3674 if (nh) {
3675 if (!nexthop_get(nh)) {
3676 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3677 goto out;
3678 }
3679 if (rt->fib6_src.plen) {
3680 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3681 goto out;
3682 }
3683 rt->nh = nh;
3684 fib6_nh = nexthop_fib6_nh(rt->nh);
3685 } else {
3686 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3687 if (err)
3688 goto out;
3689
3690 fib6_nh = rt->fib6_nh;
3691
3692 /* We cannot add true routes via loopback here, they would
3693 * result in kernel looping; promote them to reject routes
3694 */
3695 addr_type = ipv6_addr_type(&cfg->fc_dst);
3696 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3697 addr_type))
3698 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3699 }
3700
3701 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3702 struct net_device *dev = fib6_nh->fib_nh_dev;
3703
3704 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3705 NL_SET_ERR_MSG(extack, "Invalid source address");
3706 err = -EINVAL;
3707 goto out;
3708 }
3709 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3710 rt->fib6_prefsrc.plen = 128;
3711 } else
3712 rt->fib6_prefsrc.plen = 0;
3713
3714 return rt;
3715 out:
3716 fib6_info_release(rt);
3717 return ERR_PTR(err);
3718 }
3719
3720 int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3721 struct netlink_ext_ack *extack)
3722 {
3723 struct fib6_info *rt;
3724 int err;
3725
3726 rt = ip6_route_info_create(cfg, gfp_flags, extack);
3727 if (IS_ERR(rt))
3728 return PTR_ERR(rt);
3729
3730 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3731 fib6_info_release(rt);
3732
3733 return err;
3734 }
3735
3736 static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3737 {
3738 struct net *net = info->nl_net;
3739 struct fib6_table *table;
3740 int err;
3741
3742 if (rt == net->ipv6.fib6_null_entry) {
3743 err = -ENOENT;
3744 goto out;
3745 }
3746
3747 table = rt->fib6_table;
3748 spin_lock_bh(&table->tb6_lock);
3749 err = fib6_del(rt, info);
3750 spin_unlock_bh(&table->tb6_lock);
3751
3752 out:
3753 fib6_info_release(rt);
3754 return err;
3755 }
3756
3757 int ip6_del_rt(struct net *net, struct fib6_info *rt)
3758 {
3759 struct nl_info info = { .nl_net = net };
3760
3761 return __ip6_del_rt(rt, &info);
3762 }
3763
3764 static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3765 {
3766 struct nl_info *info = &cfg->fc_nlinfo;
3767 struct net *net = info->nl_net;
3768 struct sk_buff *skb = NULL;
3769 struct fib6_table *table;
3770 int err = -ENOENT;
3771
3772 if (rt == net->ipv6.fib6_null_entry)
3773 goto out_put;
3774 table = rt->fib6_table;
3775 spin_lock_bh(&table->tb6_lock);
3776
3777 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3778 struct fib6_info *sibling, *next_sibling;
3779 struct fib6_node *fn;
3780
3781 /* prefer to send a single notification with all hops */
3782 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3783 if (skb) {
3784 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3785
3786 if (rt6_fill_node(net, skb, rt, NULL,
3787 NULL, NULL, 0, RTM_DELROUTE,
3788 info->portid, seq, 0) < 0) {
3789 kfree_skb(skb);
3790 skb = NULL;
3791 } else
3792 info->skip_notify = 1;
3793 }
3794
3795 /* 'rt' points to the first sibling route. If it is not the
3796 * leaf, then we do not need to send a notification. Otherwise,
3797 * we need to check if the last sibling has a next route or not
3798 * and emit a replace or delete notification, respectively.
3799 */
3800 info->skip_notify_kernel = 1;
3801 fn = rcu_dereference_protected(rt->fib6_node,
3802 lockdep_is_held(&table->tb6_lock));
3803 if (rcu_access_pointer(fn->leaf) == rt) {
3804 struct fib6_info *last_sibling, *replace_rt;
3805
3806 last_sibling = list_last_entry(&rt->fib6_siblings,
3807 struct fib6_info,
3808 fib6_siblings);
3809 replace_rt = rcu_dereference_protected(
3810 last_sibling->fib6_next,
3811 lockdep_is_held(&table->tb6_lock));
3812 if (replace_rt)
3813 call_fib6_entry_notifiers_replace(net,
3814 replace_rt);
3815 else
3816 call_fib6_multipath_entry_notifiers(net,
3817 FIB_EVENT_ENTRY_DEL,
3818 rt, rt->fib6_nsiblings,
3819 NULL);
3820 }
3821 list_for_each_entry_safe(sibling, next_sibling,
3822 &rt->fib6_siblings,
3823 fib6_siblings) {
3824 err = fib6_del(sibling, info);
3825 if (err)
3826 goto out_unlock;
3827 }
3828 }
3829
3830 err = fib6_del(rt, info);
3831 out_unlock:
3832 spin_unlock_bh(&table->tb6_lock);
3833 out_put:
3834 fib6_info_release(rt);
3835
3836 if (skb) {
3837 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3838 info->nlh, gfp_any());
3839 }
3840 return err;
3841 }
3842
3843 static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3844 {
3845 int rc = -ESRCH;
3846
3847 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3848 goto out;
3849
3850 if (cfg->fc_flags & RTF_GATEWAY &&
3851 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3852 goto out;
3853
3854 rc = rt6_remove_exception_rt(rt);
3855 out:
3856 return rc;
3857 }
3858
3859 static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3860 struct fib6_nh *nh)
3861 {
3862 struct fib6_result res = {
3863 .f6i = rt,
3864 .nh = nh,
3865 };
3866 struct rt6_info *rt_cache;
3867
3868 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3869 if (rt_cache)
3870 return __ip6_del_cached_rt(rt_cache, cfg);
3871
3872 return 0;
3873 }
3874
3875 struct fib6_nh_del_cached_rt_arg {
3876 struct fib6_config *cfg;
3877 struct fib6_info *f6i;
3878 };
3879
3880 static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
3881 {
3882 struct fib6_nh_del_cached_rt_arg *arg = _arg;
3883 int rc;
3884
3885 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
3886 return rc != -ESRCH ? rc : 0;
3887 }
3888
3889 static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
3890 {
3891 struct fib6_nh_del_cached_rt_arg arg = {
3892 .cfg = cfg,
3893 .f6i = f6i
3894 };
3895
3896 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
3897 }
3898
3899 static int ip6_route_del(struct fib6_config *cfg,
3900 struct netlink_ext_ack *extack)
3901 {
3902 struct fib6_table *table;
3903 struct fib6_info *rt;
3904 struct fib6_node *fn;
3905 int err = -ESRCH;
3906
3907 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
3908 if (!table) {
3909 NL_SET_ERR_MSG(extack, "FIB table does not exist");
3910 return err;
3911 }
3912
3913 rcu_read_lock();
3914
3915 fn = fib6_locate(&table->tb6_root,
3916 &cfg->fc_dst, cfg->fc_dst_len,
3917 &cfg->fc_src, cfg->fc_src_len,
3918 !(cfg->fc_flags & RTF_CACHE));
3919
3920 if (fn) {
3921 for_each_fib6_node_rt_rcu(fn) {
3922 struct fib6_nh *nh;
3923
3924 if (rt->nh && cfg->fc_nh_id &&
3925 rt->nh->id != cfg->fc_nh_id)
3926 continue;
3927
3928 if (cfg->fc_flags & RTF_CACHE) {
3929 int rc = 0;
3930
3931 if (rt->nh) {
3932 rc = ip6_del_cached_rt_nh(cfg, rt);
3933 } else if (cfg->fc_nh_id) {
3934 continue;
3935 } else {
3936 nh = rt->fib6_nh;
3937 rc = ip6_del_cached_rt(cfg, rt, nh);
3938 }
3939 if (rc != -ESRCH) {
3940 rcu_read_unlock();
3941 return rc;
3942 }
3943 continue;
3944 }
3945
3946 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
3947 continue;
3948 if (cfg->fc_protocol &&
3949 cfg->fc_protocol != rt->fib6_protocol)
3950 continue;
3951
3952 if (rt->nh) {
3953 if (!fib6_info_hold_safe(rt))
3954 continue;
3955 rcu_read_unlock();
3956
3957 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3958 }
3959 if (cfg->fc_nh_id)
3960 continue;
3961
3962 nh = rt->fib6_nh;
3963 if (cfg->fc_ifindex &&
3964 (!nh->fib_nh_dev ||
3965 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
3966 continue;
3967 if (cfg->fc_flags & RTF_GATEWAY &&
3968 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
3969 continue;
3970 if (!fib6_info_hold_safe(rt))
3971 continue;
3972 rcu_read_unlock();
3973
3974 /* if gateway was specified only delete the one hop */
3975 if (cfg->fc_flags & RTF_GATEWAY)
3976 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3977
3978 return __ip6_del_rt_siblings(rt, cfg);
3979 }
3980 }
3981 rcu_read_unlock();
3982
3983 return err;
3984 }
3985
3986 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
3987 {
3988 struct netevent_redirect netevent;
3989 struct rt6_info *rt, *nrt = NULL;
3990 struct fib6_result res = {};
3991 struct ndisc_options ndopts;
3992 struct inet6_dev *in6_dev;
3993 struct neighbour *neigh;
3994 struct rd_msg *msg;
3995 int optlen, on_link;
3996 u8 *lladdr;
3997
3998 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
3999 optlen -= sizeof(*msg);
4000
4001 if (optlen < 0) {
4002 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
4003 return;
4004 }
4005
4006 msg = (struct rd_msg *)icmp6_hdr(skb);
4007
4008 if (ipv6_addr_is_multicast(&msg->dest)) {
4009 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
4010 return;
4011 }
4012
4013 on_link = 0;
4014 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
4015 on_link = 1;
4016 } else if (ipv6_addr_type(&msg->target) !=
4017 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
4018 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
4019 return;
4020 }
4021
4022 in6_dev = __in6_dev_get(skb->dev);
4023 if (!in6_dev)
4024 return;
4025 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
4026 return;
4027
4028 /* RFC2461 8.1:
4029 * The IP source address of the Redirect MUST be the same as the current
4030 * first-hop router for the specified ICMP Destination Address.
4031 */
4032
4033 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
4034 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4035 return;
4036 }
4037
4038 lladdr = NULL;
4039 if (ndopts.nd_opts_tgt_lladdr) {
4040 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4041 skb->dev);
4042 if (!lladdr) {
4043 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4044 return;
4045 }
4046 }
4047
4048 rt = (struct rt6_info *) dst;
4049 if (rt->rt6i_flags & RTF_REJECT) {
4050 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
4051 return;
4052 }
4053
4054 /* Redirect received -> path was valid.
4055 * Look, redirects are sent only in response to data packets,
4056 * so that this nexthop apparently is reachable. --ANK
4057 */
4058 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
4059
4060 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
4061 if (!neigh)
4062 return;
4063
4064 /*
4065 * We have finally decided to accept it.
4066 */
4067
4068 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
4069 NEIGH_UPDATE_F_WEAK_OVERRIDE|
4070 NEIGH_UPDATE_F_OVERRIDE|
4071 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
4072 NEIGH_UPDATE_F_ISROUTER)),
4073 NDISC_REDIRECT, &ndopts);
4074
4075 rcu_read_lock();
4076 res.f6i = rcu_dereference(rt->from);
4077 if (!res.f6i)
4078 goto out;
4079
4080 if (res.f6i->nh) {
4081 struct fib6_nh_match_arg arg = {
4082 .dev = dst->dev,
4083 .gw = &rt->rt6i_gateway,
4084 };
4085
4086 nexthop_for_each_fib6_nh(res.f6i->nh,
4087 fib6_nh_find_match, &arg);
4088
4089 /* fib6_info uses a nexthop that does not have fib6_nh
4090 * using the dst->dev. Should be impossible
4091 */
4092 if (!arg.match)
4093 goto out;
4094 res.nh = arg.match;
4095 } else {
4096 res.nh = res.f6i->fib6_nh;
4097 }
4098
4099 res.fib6_flags = res.f6i->fib6_flags;
4100 res.fib6_type = res.f6i->fib6_type;
4101 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4102 if (!nrt)
4103 goto out;
4104
4105 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4106 if (on_link)
4107 nrt->rt6i_flags &= ~RTF_GATEWAY;
4108
4109 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4110
4111 /* rt6_insert_exception() will take care of duplicated exceptions */
4112 if (rt6_insert_exception(nrt, &res)) {
4113 dst_release_immediate(&nrt->dst);
4114 goto out;
4115 }
4116
4117 netevent.old = &rt->dst;
4118 netevent.new = &nrt->dst;
4119 netevent.daddr = &msg->dest;
4120 netevent.neigh = neigh;
4121 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4122
4123 out:
4124 rcu_read_unlock();
4125 neigh_release(neigh);
4126 }
4127
4128 #ifdef CONFIG_IPV6_ROUTE_INFO
4129 static struct fib6_info *rt6_get_route_info(struct net *net,
4130 const struct in6_addr *prefix, int prefixlen,
4131 const struct in6_addr *gwaddr,
4132 struct net_device *dev)
4133 {
4134 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4135 int ifindex = dev->ifindex;
4136 struct fib6_node *fn;
4137 struct fib6_info *rt = NULL;
4138 struct fib6_table *table;
4139
4140 table = fib6_get_table(net, tb_id);
4141 if (!table)
4142 return NULL;
4143
4144 rcu_read_lock();
4145 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4146 if (!fn)
4147 goto out;
4148
4149 for_each_fib6_node_rt_rcu(fn) {
4150 /* these routes do not use nexthops */
4151 if (rt->nh)
4152 continue;
4153 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4154 continue;
4155 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4156 !rt->fib6_nh->fib_nh_gw_family)
4157 continue;
4158 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4159 continue;
4160 if (!fib6_info_hold_safe(rt))
4161 continue;
4162 break;
4163 }
4164 out:
4165 rcu_read_unlock();
4166 return rt;
4167 }
4168
4169 static struct fib6_info *rt6_add_route_info(struct net *net,
4170 const struct in6_addr *prefix, int prefixlen,
4171 const struct in6_addr *gwaddr,
4172 struct net_device *dev,
4173 unsigned int pref)
4174 {
4175 struct fib6_config cfg = {
4176 .fc_metric = IP6_RT_PRIO_USER,
4177 .fc_ifindex = dev->ifindex,
4178 .fc_dst_len = prefixlen,
4179 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4180 RTF_UP | RTF_PREF(pref),
4181 .fc_protocol = RTPROT_RA,
4182 .fc_type = RTN_UNICAST,
4183 .fc_nlinfo.portid = 0,
4184 .fc_nlinfo.nlh = NULL,
4185 .fc_nlinfo.nl_net = net,
4186 };
4187
4188 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
4189 cfg.fc_dst = *prefix;
4190 cfg.fc_gateway = *gwaddr;
4191
4192 /* We should treat it as a default route if prefix length is 0. */
4193 if (!prefixlen)
4194 cfg.fc_flags |= RTF_DEFAULT;
4195
4196 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4197
4198 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4199 }
4200 #endif
4201
4202 struct fib6_info *rt6_get_dflt_router(struct net *net,
4203 const struct in6_addr *addr,
4204 struct net_device *dev)
4205 {
4206 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4207 struct fib6_info *rt;
4208 struct fib6_table *table;
4209
4210 table = fib6_get_table(net, tb_id);
4211 if (!table)
4212 return NULL;
4213
4214 rcu_read_lock();
4215 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4216 struct fib6_nh *nh;
4217
4218 /* RA routes do not use nexthops */
4219 if (rt->nh)
4220 continue;
4221
4222 nh = rt->fib6_nh;
4223 if (dev == nh->fib_nh_dev &&
4224 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4225 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4226 break;
4227 }
4228 if (rt && !fib6_info_hold_safe(rt))
4229 rt = NULL;
4230 rcu_read_unlock();
4231 return rt;
4232 }
4233
4234 struct fib6_info *rt6_add_dflt_router(struct net *net,
4235 const struct in6_addr *gwaddr,
4236 struct net_device *dev,
4237 unsigned int pref)
4238 {
4239 struct fib6_config cfg = {
4240 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4241 .fc_metric = IP6_RT_PRIO_USER,
4242 .fc_ifindex = dev->ifindex,
4243 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4244 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4245 .fc_protocol = RTPROT_RA,
4246 .fc_type = RTN_UNICAST,
4247 .fc_nlinfo.portid = 0,
4248 .fc_nlinfo.nlh = NULL,
4249 .fc_nlinfo.nl_net = net,
4250 };
4251
4252 cfg.fc_gateway = *gwaddr;
4253
4254 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4255 struct fib6_table *table;
4256
4257 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4258 if (table)
4259 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4260 }
4261
4262 return rt6_get_dflt_router(net, gwaddr, dev);
4263 }
4264
4265 static void __rt6_purge_dflt_routers(struct net *net,
4266 struct fib6_table *table)
4267 {
4268 struct fib6_info *rt;
4269
4270 restart:
4271 rcu_read_lock();
4272 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4273 struct net_device *dev = fib6_info_nh_dev(rt);
4274 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4275
4276 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4277 (!idev || idev->cnf.accept_ra != 2) &&
4278 fib6_info_hold_safe(rt)) {
4279 rcu_read_unlock();
4280 ip6_del_rt(net, rt);
4281 goto restart;
4282 }
4283 }
4284 rcu_read_unlock();
4285
4286 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4287 }
4288
4289 void rt6_purge_dflt_routers(struct net *net)
4290 {
4291 struct fib6_table *table;
4292 struct hlist_head *head;
4293 unsigned int h;
4294
4295 rcu_read_lock();
4296
4297 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4298 head = &net->ipv6.fib_table_hash[h];
4299 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4300 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4301 __rt6_purge_dflt_routers(net, table);
4302 }
4303 }
4304
4305 rcu_read_unlock();
4306 }
4307
4308 static void rtmsg_to_fib6_config(struct net *net,
4309 struct in6_rtmsg *rtmsg,
4310 struct fib6_config *cfg)
4311 {
4312 *cfg = (struct fib6_config){
4313 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4314 : RT6_TABLE_MAIN,
4315 .fc_ifindex = rtmsg->rtmsg_ifindex,
4316 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
4317 .fc_expires = rtmsg->rtmsg_info,
4318 .fc_dst_len = rtmsg->rtmsg_dst_len,
4319 .fc_src_len = rtmsg->rtmsg_src_len,
4320 .fc_flags = rtmsg->rtmsg_flags,
4321 .fc_type = rtmsg->rtmsg_type,
4322
4323 .fc_nlinfo.nl_net = net,
4324
4325 .fc_dst = rtmsg->rtmsg_dst,
4326 .fc_src = rtmsg->rtmsg_src,
4327 .fc_gateway = rtmsg->rtmsg_gateway,
4328 };
4329 }
4330
4331 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
4332 {
4333 struct fib6_config cfg;
4334 struct in6_rtmsg rtmsg;
4335 int err;
4336
4337 switch (cmd) {
4338 case SIOCADDRT: /* Add a route */
4339 case SIOCDELRT: /* Delete a route */
4340 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4341 return -EPERM;
4342 err = copy_from_user(&rtmsg, arg,
4343 sizeof(struct in6_rtmsg));
4344 if (err)
4345 return -EFAULT;
4346
4347 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
4348
4349 rtnl_lock();
4350 switch (cmd) {
4351 case SIOCADDRT:
4352 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4353 break;
4354 case SIOCDELRT:
4355 err = ip6_route_del(&cfg, NULL);
4356 break;
4357 default:
4358 err = -EINVAL;
4359 }
4360 rtnl_unlock();
4361
4362 return err;
4363 }
4364
4365 return -EINVAL;
4366 }
4367
4368 /*
4369 * Drop the packet on the floor
4370 */
4371
4372 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4373 {
4374 struct dst_entry *dst = skb_dst(skb);
4375 struct net *net = dev_net(dst->dev);
4376 struct inet6_dev *idev;
4377 int type;
4378
4379 if (netif_is_l3_master(skb->dev) &&
4380 dst->dev == net->loopback_dev)
4381 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4382 else
4383 idev = ip6_dst_idev(dst);
4384
4385 switch (ipstats_mib_noroutes) {
4386 case IPSTATS_MIB_INNOROUTES:
4387 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4388 if (type == IPV6_ADDR_ANY) {
4389 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4390 break;
4391 }
4392 fallthrough;
4393 case IPSTATS_MIB_OUTNOROUTES:
4394 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4395 break;
4396 }
4397
4398 /* Start over by dropping the dst for l3mdev case */
4399 if (netif_is_l3_master(skb->dev))
4400 skb_dst_drop(skb);
4401
4402 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4403 kfree_skb(skb);
4404 return 0;
4405 }
4406
4407 static int ip6_pkt_discard(struct sk_buff *skb)
4408 {
4409 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4410 }
4411
4412 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4413 {
4414 skb->dev = skb_dst(skb)->dev;
4415 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4416 }
4417
4418 static int ip6_pkt_prohibit(struct sk_buff *skb)
4419 {
4420 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4421 }
4422
4423 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4424 {
4425 skb->dev = skb_dst(skb)->dev;
4426 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4427 }
4428
4429 /*
4430 * Allocate a dst for local (unicast / anycast) address.
4431 */
4432
4433 struct fib6_info *addrconf_f6i_alloc(struct net *net,
4434 struct inet6_dev *idev,
4435 const struct in6_addr *addr,
4436 bool anycast, gfp_t gfp_flags)
4437 {
4438 struct fib6_config cfg = {
4439 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4440 .fc_ifindex = idev->dev->ifindex,
4441 .fc_flags = RTF_UP | RTF_NONEXTHOP,
4442 .fc_dst = *addr,
4443 .fc_dst_len = 128,
4444 .fc_protocol = RTPROT_KERNEL,
4445 .fc_nlinfo.nl_net = net,
4446 .fc_ignore_dev_down = true,
4447 };
4448 struct fib6_info *f6i;
4449
4450 if (anycast) {
4451 cfg.fc_type = RTN_ANYCAST;
4452 cfg.fc_flags |= RTF_ANYCAST;
4453 } else {
4454 cfg.fc_type = RTN_LOCAL;
4455 cfg.fc_flags |= RTF_LOCAL;
4456 }
4457
4458 f6i = ip6_route_info_create(&cfg, gfp_flags, NULL);
4459 if (!IS_ERR(f6i))
4460 f6i->dst_nocount = true;
4461 return f6i;
4462 }
4463
4464 /* remove deleted ip from prefsrc entries */
4465 struct arg_dev_net_ip {
4466 struct net_device *dev;
4467 struct net *net;
4468 struct in6_addr *addr;
4469 };
4470
4471 static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4472 {
4473 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4474 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4475 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4476
4477 if (!rt->nh &&
4478 ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
4479 rt != net->ipv6.fib6_null_entry &&
4480 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
4481 spin_lock_bh(&rt6_exception_lock);
4482 /* remove prefsrc entry */
4483 rt->fib6_prefsrc.plen = 0;
4484 spin_unlock_bh(&rt6_exception_lock);
4485 }
4486 return 0;
4487 }
4488
4489 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4490 {
4491 struct net *net = dev_net(ifp->idev->dev);
4492 struct arg_dev_net_ip adni = {
4493 .dev = ifp->idev->dev,
4494 .net = net,
4495 .addr = &ifp->addr,
4496 };
4497 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4498 }
4499
4500 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
4501
4502 /* Remove routers and update dst entries when gateway turn into host. */
4503 static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4504 {
4505 struct in6_addr *gateway = (struct in6_addr *)arg;
4506 struct fib6_nh *nh;
4507
4508 /* RA routes do not use nexthops */
4509 if (rt->nh)
4510 return 0;
4511
4512 nh = rt->fib6_nh;
4513 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4514 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4515 return -1;
4516
4517 /* Further clean up cached routes in exception table.
4518 * This is needed because cached route may have a different
4519 * gateway than its 'parent' in the case of an ip redirect.
4520 */
4521 fib6_nh_exceptions_clean_tohost(nh, gateway);
4522
4523 return 0;
4524 }
4525
4526 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4527 {
4528 fib6_clean_all(net, fib6_clean_tohost, gateway);
4529 }
4530
4531 struct arg_netdev_event {
4532 const struct net_device *dev;
4533 union {
4534 unsigned char nh_flags;
4535 unsigned long event;
4536 };
4537 };
4538
4539 static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4540 {
4541 struct fib6_info *iter;
4542 struct fib6_node *fn;
4543
4544 fn = rcu_dereference_protected(rt->fib6_node,
4545 lockdep_is_held(&rt->fib6_table->tb6_lock));
4546 iter = rcu_dereference_protected(fn->leaf,
4547 lockdep_is_held(&rt->fib6_table->tb6_lock));
4548 while (iter) {
4549 if (iter->fib6_metric == rt->fib6_metric &&
4550 rt6_qualify_for_ecmp(iter))
4551 return iter;
4552 iter = rcu_dereference_protected(iter->fib6_next,
4553 lockdep_is_held(&rt->fib6_table->tb6_lock));
4554 }
4555
4556 return NULL;
4557 }
4558
4559 /* only called for fib entries with builtin fib6_nh */
4560 static bool rt6_is_dead(const struct fib6_info *rt)
4561 {
4562 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4563 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4564 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4565 return true;
4566
4567 return false;
4568 }
4569
4570 static int rt6_multipath_total_weight(const struct fib6_info *rt)
4571 {
4572 struct fib6_info *iter;
4573 int total = 0;
4574
4575 if (!rt6_is_dead(rt))
4576 total += rt->fib6_nh->fib_nh_weight;
4577
4578 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4579 if (!rt6_is_dead(iter))
4580 total += iter->fib6_nh->fib_nh_weight;
4581 }
4582
4583 return total;
4584 }
4585
4586 static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4587 {
4588 int upper_bound = -1;
4589
4590 if (!rt6_is_dead(rt)) {
4591 *weight += rt->fib6_nh->fib_nh_weight;
4592 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4593 total) - 1;
4594 }
4595 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4596 }
4597
4598 static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4599 {
4600 struct fib6_info *iter;
4601 int weight = 0;
4602
4603 rt6_upper_bound_set(rt, &weight, total);
4604
4605 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4606 rt6_upper_bound_set(iter, &weight, total);
4607 }
4608
4609 void rt6_multipath_rebalance(struct fib6_info *rt)
4610 {
4611 struct fib6_info *first;
4612 int total;
4613
4614 /* In case the entire multipath route was marked for flushing,
4615 * then there is no need to rebalance upon the removal of every
4616 * sibling route.
4617 */
4618 if (!rt->fib6_nsiblings || rt->should_flush)
4619 return;
4620
4621 /* During lookup routes are evaluated in order, so we need to
4622 * make sure upper bounds are assigned from the first sibling
4623 * onwards.
4624 */
4625 first = rt6_multipath_first_sibling(rt);
4626 if (WARN_ON_ONCE(!first))
4627 return;
4628
4629 total = rt6_multipath_total_weight(first);
4630 rt6_multipath_upper_bound_set(first, total);
4631 }
4632
4633 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4634 {
4635 const struct arg_netdev_event *arg = p_arg;
4636 struct net *net = dev_net(arg->dev);
4637
4638 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4639 rt->fib6_nh->fib_nh_dev == arg->dev) {
4640 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4641 fib6_update_sernum_upto_root(net, rt);
4642 rt6_multipath_rebalance(rt);
4643 }
4644
4645 return 0;
4646 }
4647
4648 void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4649 {
4650 struct arg_netdev_event arg = {
4651 .dev = dev,
4652 {
4653 .nh_flags = nh_flags,
4654 },
4655 };
4656
4657 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4658 arg.nh_flags |= RTNH_F_LINKDOWN;
4659
4660 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4661 }
4662
4663 /* only called for fib entries with inline fib6_nh */
4664 static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4665 const struct net_device *dev)
4666 {
4667 struct fib6_info *iter;
4668
4669 if (rt->fib6_nh->fib_nh_dev == dev)
4670 return true;
4671 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4672 if (iter->fib6_nh->fib_nh_dev == dev)
4673 return true;
4674
4675 return false;
4676 }
4677
4678 static void rt6_multipath_flush(struct fib6_info *rt)
4679 {
4680 struct fib6_info *iter;
4681
4682 rt->should_flush = 1;
4683 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4684 iter->should_flush = 1;
4685 }
4686
4687 static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4688 const struct net_device *down_dev)
4689 {
4690 struct fib6_info *iter;
4691 unsigned int dead = 0;
4692
4693 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4694 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4695 dead++;
4696 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4697 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4698 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4699 dead++;
4700
4701 return dead;
4702 }
4703
4704 static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4705 const struct net_device *dev,
4706 unsigned char nh_flags)
4707 {
4708 struct fib6_info *iter;
4709
4710 if (rt->fib6_nh->fib_nh_dev == dev)
4711 rt->fib6_nh->fib_nh_flags |= nh_flags;
4712 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4713 if (iter->fib6_nh->fib_nh_dev == dev)
4714 iter->fib6_nh->fib_nh_flags |= nh_flags;
4715 }
4716
4717 /* called with write lock held for table with rt */
4718 static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4719 {
4720 const struct arg_netdev_event *arg = p_arg;
4721 const struct net_device *dev = arg->dev;
4722 struct net *net = dev_net(dev);
4723
4724 if (rt == net->ipv6.fib6_null_entry || rt->nh)
4725 return 0;
4726
4727 switch (arg->event) {
4728 case NETDEV_UNREGISTER:
4729 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4730 case NETDEV_DOWN:
4731 if (rt->should_flush)
4732 return -1;
4733 if (!rt->fib6_nsiblings)
4734 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4735 if (rt6_multipath_uses_dev(rt, dev)) {
4736 unsigned int count;
4737
4738 count = rt6_multipath_dead_count(rt, dev);
4739 if (rt->fib6_nsiblings + 1 == count) {
4740 rt6_multipath_flush(rt);
4741 return -1;
4742 }
4743 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4744 RTNH_F_LINKDOWN);
4745 fib6_update_sernum(net, rt);
4746 rt6_multipath_rebalance(rt);
4747 }
4748 return -2;
4749 case NETDEV_CHANGE:
4750 if (rt->fib6_nh->fib_nh_dev != dev ||
4751 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4752 break;
4753 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4754 rt6_multipath_rebalance(rt);
4755 break;
4756 }
4757
4758 return 0;
4759 }
4760
4761 void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4762 {
4763 struct arg_netdev_event arg = {
4764 .dev = dev,
4765 {
4766 .event = event,
4767 },
4768 };
4769 struct net *net = dev_net(dev);
4770
4771 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4772 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4773 else
4774 fib6_clean_all(net, fib6_ifdown, &arg);
4775 }
4776
4777 void rt6_disable_ip(struct net_device *dev, unsigned long event)
4778 {
4779 rt6_sync_down_dev(dev, event);
4780 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4781 neigh_ifdown(&nd_tbl, dev);
4782 }
4783
4784 struct rt6_mtu_change_arg {
4785 struct net_device *dev;
4786 unsigned int mtu;
4787 struct fib6_info *f6i;
4788 };
4789
4790 static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4791 {
4792 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4793 struct fib6_info *f6i = arg->f6i;
4794
4795 /* For administrative MTU increase, there is no way to discover
4796 * IPv6 PMTU increase, so PMTU increase should be updated here.
4797 * Since RFC 1981 doesn't include administrative MTU increase
4798 * update PMTU increase is a MUST. (i.e. jumbo frame)
4799 */
4800 if (nh->fib_nh_dev == arg->dev) {
4801 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4802 u32 mtu = f6i->fib6_pmtu;
4803
4804 if (mtu >= arg->mtu ||
4805 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4806 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4807
4808 spin_lock_bh(&rt6_exception_lock);
4809 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4810 spin_unlock_bh(&rt6_exception_lock);
4811 }
4812
4813 return 0;
4814 }
4815
4816 static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4817 {
4818 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4819 struct inet6_dev *idev;
4820
4821 /* In IPv6 pmtu discovery is not optional,
4822 so that RTAX_MTU lock cannot disable it.
4823 We still use this lock to block changes
4824 caused by addrconf/ndisc.
4825 */
4826
4827 idev = __in6_dev_get(arg->dev);
4828 if (!idev)
4829 return 0;
4830
4831 if (fib6_metric_locked(f6i, RTAX_MTU))
4832 return 0;
4833
4834 arg->f6i = f6i;
4835 if (f6i->nh) {
4836 /* fib6_nh_mtu_change only returns 0, so this is safe */
4837 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4838 arg);
4839 }
4840
4841 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4842 }
4843
4844 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4845 {
4846 struct rt6_mtu_change_arg arg = {
4847 .dev = dev,
4848 .mtu = mtu,
4849 };
4850
4851 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4852 }
4853
4854 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4855 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
4856 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
4857 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
4858 [RTA_OIF] = { .type = NLA_U32 },
4859 [RTA_IIF] = { .type = NLA_U32 },
4860 [RTA_PRIORITY] = { .type = NLA_U32 },
4861 [RTA_METRICS] = { .type = NLA_NESTED },
4862 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
4863 [RTA_PREF] = { .type = NLA_U8 },
4864 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4865 [RTA_ENCAP] = { .type = NLA_NESTED },
4866 [RTA_EXPIRES] = { .type = NLA_U32 },
4867 [RTA_UID] = { .type = NLA_U32 },
4868 [RTA_MARK] = { .type = NLA_U32 },
4869 [RTA_TABLE] = { .type = NLA_U32 },
4870 [RTA_IP_PROTO] = { .type = NLA_U8 },
4871 [RTA_SPORT] = { .type = NLA_U16 },
4872 [RTA_DPORT] = { .type = NLA_U16 },
4873 [RTA_NH_ID] = { .type = NLA_U32 },
4874 };
4875
4876 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
4877 struct fib6_config *cfg,
4878 struct netlink_ext_ack *extack)
4879 {
4880 struct rtmsg *rtm;
4881 struct nlattr *tb[RTA_MAX+1];
4882 unsigned int pref;
4883 int err;
4884
4885 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
4886 rtm_ipv6_policy, extack);
4887 if (err < 0)
4888 goto errout;
4889
4890 err = -EINVAL;
4891 rtm = nlmsg_data(nlh);
4892
4893 *cfg = (struct fib6_config){
4894 .fc_table = rtm->rtm_table,
4895 .fc_dst_len = rtm->rtm_dst_len,
4896 .fc_src_len = rtm->rtm_src_len,
4897 .fc_flags = RTF_UP,
4898 .fc_protocol = rtm->rtm_protocol,
4899 .fc_type = rtm->rtm_type,
4900
4901 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
4902 .fc_nlinfo.nlh = nlh,
4903 .fc_nlinfo.nl_net = sock_net(skb->sk),
4904 };
4905
4906 if (rtm->rtm_type == RTN_UNREACHABLE ||
4907 rtm->rtm_type == RTN_BLACKHOLE ||
4908 rtm->rtm_type == RTN_PROHIBIT ||
4909 rtm->rtm_type == RTN_THROW)
4910 cfg->fc_flags |= RTF_REJECT;
4911
4912 if (rtm->rtm_type == RTN_LOCAL)
4913 cfg->fc_flags |= RTF_LOCAL;
4914
4915 if (rtm->rtm_flags & RTM_F_CLONED)
4916 cfg->fc_flags |= RTF_CACHE;
4917
4918 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4919
4920 if (tb[RTA_NH_ID]) {
4921 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
4922 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
4923 NL_SET_ERR_MSG(extack,
4924 "Nexthop specification and nexthop id are mutually exclusive");
4925 goto errout;
4926 }
4927 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
4928 }
4929
4930 if (tb[RTA_GATEWAY]) {
4931 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
4932 cfg->fc_flags |= RTF_GATEWAY;
4933 }
4934 if (tb[RTA_VIA]) {
4935 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
4936 goto errout;
4937 }
4938
4939 if (tb[RTA_DST]) {
4940 int plen = (rtm->rtm_dst_len + 7) >> 3;
4941
4942 if (nla_len(tb[RTA_DST]) < plen)
4943 goto errout;
4944
4945 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
4946 }
4947
4948 if (tb[RTA_SRC]) {
4949 int plen = (rtm->rtm_src_len + 7) >> 3;
4950
4951 if (nla_len(tb[RTA_SRC]) < plen)
4952 goto errout;
4953
4954 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
4955 }
4956
4957 if (tb[RTA_PREFSRC])
4958 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
4959
4960 if (tb[RTA_OIF])
4961 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4962
4963 if (tb[RTA_PRIORITY])
4964 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4965
4966 if (tb[RTA_METRICS]) {
4967 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4968 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
4969 }
4970
4971 if (tb[RTA_TABLE])
4972 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4973
4974 if (tb[RTA_MULTIPATH]) {
4975 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4976 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
4977
4978 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
4979 cfg->fc_mp_len, extack);
4980 if (err < 0)
4981 goto errout;
4982 }
4983
4984 if (tb[RTA_PREF]) {
4985 pref = nla_get_u8(tb[RTA_PREF]);
4986 if (pref != ICMPV6_ROUTER_PREF_LOW &&
4987 pref != ICMPV6_ROUTER_PREF_HIGH)
4988 pref = ICMPV6_ROUTER_PREF_MEDIUM;
4989 cfg->fc_flags |= RTF_PREF(pref);
4990 }
4991
4992 if (tb[RTA_ENCAP])
4993 cfg->fc_encap = tb[RTA_ENCAP];
4994
4995 if (tb[RTA_ENCAP_TYPE]) {
4996 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
4997
4998 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
4999 if (err < 0)
5000 goto errout;
5001 }
5002
5003 if (tb[RTA_EXPIRES]) {
5004 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5005
5006 if (addrconf_finite_timeout(timeout)) {
5007 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5008 cfg->fc_flags |= RTF_EXPIRES;
5009 }
5010 }
5011
5012 err = 0;
5013 errout:
5014 return err;
5015 }
5016
5017 struct rt6_nh {
5018 struct fib6_info *fib6_info;
5019 struct fib6_config r_cfg;
5020 struct list_head next;
5021 };
5022
5023 static int ip6_route_info_append(struct net *net,
5024 struct list_head *rt6_nh_list,
5025 struct fib6_info *rt,
5026 struct fib6_config *r_cfg)
5027 {
5028 struct rt6_nh *nh;
5029 int err = -EEXIST;
5030
5031 list_for_each_entry(nh, rt6_nh_list, next) {
5032 /* check if fib6_info already exists */
5033 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
5034 return err;
5035 }
5036
5037 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5038 if (!nh)
5039 return -ENOMEM;
5040 nh->fib6_info = rt;
5041 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5042 list_add_tail(&nh->next, rt6_nh_list);
5043
5044 return 0;
5045 }
5046
5047 static void ip6_route_mpath_notify(struct fib6_info *rt,
5048 struct fib6_info *rt_last,
5049 struct nl_info *info,
5050 __u16 nlflags)
5051 {
5052 /* if this is an APPEND route, then rt points to the first route
5053 * inserted and rt_last points to last route inserted. Userspace
5054 * wants a consistent dump of the route which starts at the first
5055 * nexthop. Since sibling routes are always added at the end of
5056 * the list, find the first sibling of the last route appended
5057 */
5058 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5059 rt = list_first_entry(&rt_last->fib6_siblings,
5060 struct fib6_info,
5061 fib6_siblings);
5062 }
5063
5064 if (rt)
5065 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5066 }
5067
5068 static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5069 {
5070 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5071 bool should_notify = false;
5072 struct fib6_info *leaf;
5073 struct fib6_node *fn;
5074
5075 rcu_read_lock();
5076 fn = rcu_dereference(rt->fib6_node);
5077 if (!fn)
5078 goto out;
5079
5080 leaf = rcu_dereference(fn->leaf);
5081 if (!leaf)
5082 goto out;
5083
5084 if (rt == leaf ||
5085 (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5086 rt6_qualify_for_ecmp(leaf)))
5087 should_notify = true;
5088 out:
5089 rcu_read_unlock();
5090
5091 return should_notify;
5092 }
5093
5094 static int ip6_route_multipath_add(struct fib6_config *cfg,
5095 struct netlink_ext_ack *extack)
5096 {
5097 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5098 struct nl_info *info = &cfg->fc_nlinfo;
5099 struct fib6_config r_cfg;
5100 struct rtnexthop *rtnh;
5101 struct fib6_info *rt;
5102 struct rt6_nh *err_nh;
5103 struct rt6_nh *nh, *nh_safe;
5104 __u16 nlflags;
5105 int remaining;
5106 int attrlen;
5107 int err = 1;
5108 int nhn = 0;
5109 int replace = (cfg->fc_nlinfo.nlh &&
5110 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5111 LIST_HEAD(rt6_nh_list);
5112
5113 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5114 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5115 nlflags |= NLM_F_APPEND;
5116
5117 remaining = cfg->fc_mp_len;
5118 rtnh = (struct rtnexthop *)cfg->fc_mp;
5119
5120 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
5121 * fib6_info structs per nexthop
5122 */
5123 while (rtnh_ok(rtnh, remaining)) {
5124 memcpy(&r_cfg, cfg, sizeof(*cfg));
5125 if (rtnh->rtnh_ifindex)
5126 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5127
5128 attrlen = rtnh_attrlen(rtnh);
5129 if (attrlen > 0) {
5130 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5131
5132 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5133 if (nla) {
5134 r_cfg.fc_gateway = nla_get_in6_addr(nla);
5135 r_cfg.fc_flags |= RTF_GATEWAY;
5136 }
5137 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5138 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5139 if (nla)
5140 r_cfg.fc_encap_type = nla_get_u16(nla);
5141 }
5142
5143 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5144 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5145 if (IS_ERR(rt)) {
5146 err = PTR_ERR(rt);
5147 rt = NULL;
5148 goto cleanup;
5149 }
5150 if (!rt6_qualify_for_ecmp(rt)) {
5151 err = -EINVAL;
5152 NL_SET_ERR_MSG(extack,
5153 "Device only routes can not be added for IPv6 using the multipath API.");
5154 fib6_info_release(rt);
5155 goto cleanup;
5156 }
5157
5158 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5159
5160 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5161 rt, &r_cfg);
5162 if (err) {
5163 fib6_info_release(rt);
5164 goto cleanup;
5165 }
5166
5167 rtnh = rtnh_next(rtnh, &remaining);
5168 }
5169
5170 if (list_empty(&rt6_nh_list)) {
5171 NL_SET_ERR_MSG(extack,
5172 "Invalid nexthop configuration - no valid nexthops");
5173 return -EINVAL;
5174 }
5175
5176 /* for add and replace send one notification with all nexthops.
5177 * Skip the notification in fib6_add_rt2node and send one with
5178 * the full route when done
5179 */
5180 info->skip_notify = 1;
5181
5182 /* For add and replace, send one notification with all nexthops. For
5183 * append, send one notification with all appended nexthops.
5184 */
5185 info->skip_notify_kernel = 1;
5186
5187 err_nh = NULL;
5188 list_for_each_entry(nh, &rt6_nh_list, next) {
5189 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5190 fib6_info_release(nh->fib6_info);
5191
5192 if (!err) {
5193 /* save reference to last route successfully inserted */
5194 rt_last = nh->fib6_info;
5195
5196 /* save reference to first route for notification */
5197 if (!rt_notif)
5198 rt_notif = nh->fib6_info;
5199 }
5200
5201 /* nh->fib6_info is used or freed at this point, reset to NULL*/
5202 nh->fib6_info = NULL;
5203 if (err) {
5204 if (replace && nhn)
5205 NL_SET_ERR_MSG_MOD(extack,
5206 "multipath route replace failed (check consistency of installed routes)");
5207 err_nh = nh;
5208 goto add_errout;
5209 }
5210
5211 /* Because each route is added like a single route we remove
5212 * these flags after the first nexthop: if there is a collision,
5213 * we have already failed to add the first nexthop:
5214 * fib6_add_rt2node() has rejected it; when replacing, old
5215 * nexthops have been replaced by first new, the rest should
5216 * be added to it.
5217 */
5218 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5219 NLM_F_REPLACE);
5220 cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
5221 nhn++;
5222 }
5223
5224 /* An in-kernel notification should only be sent in case the new
5225 * multipath route is added as the first route in the node, or if
5226 * it was appended to it. We pass 'rt_notif' since it is the first
5227 * sibling and might allow us to skip some checks in the replace case.
5228 */
5229 if (ip6_route_mpath_should_notify(rt_notif)) {
5230 enum fib_event_type fib_event;
5231
5232 if (rt_notif->fib6_nsiblings != nhn - 1)
5233 fib_event = FIB_EVENT_ENTRY_APPEND;
5234 else
5235 fib_event = FIB_EVENT_ENTRY_REPLACE;
5236
5237 err = call_fib6_multipath_entry_notifiers(info->nl_net,
5238 fib_event, rt_notif,
5239 nhn - 1, extack);
5240 if (err) {
5241 /* Delete all the siblings that were just added */
5242 err_nh = NULL;
5243 goto add_errout;
5244 }
5245 }
5246
5247 /* success ... tell user about new route */
5248 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5249 goto cleanup;
5250
5251 add_errout:
5252 /* send notification for routes that were added so that
5253 * the delete notifications sent by ip6_route_del are
5254 * coherent
5255 */
5256 if (rt_notif)
5257 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5258
5259 /* Delete routes that were already added */
5260 list_for_each_entry(nh, &rt6_nh_list, next) {
5261 if (err_nh == nh)
5262 break;
5263 ip6_route_del(&nh->r_cfg, extack);
5264 }
5265
5266 cleanup:
5267 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5268 if (nh->fib6_info)
5269 fib6_info_release(nh->fib6_info);
5270 list_del(&nh->next);
5271 kfree(nh);
5272 }
5273
5274 return err;
5275 }
5276
5277 static int ip6_route_multipath_del(struct fib6_config *cfg,
5278 struct netlink_ext_ack *extack)
5279 {
5280 struct fib6_config r_cfg;
5281 struct rtnexthop *rtnh;
5282 int remaining;
5283 int attrlen;
5284 int err = 1, last_err = 0;
5285
5286 remaining = cfg->fc_mp_len;
5287 rtnh = (struct rtnexthop *)cfg->fc_mp;
5288
5289 /* Parse a Multipath Entry */
5290 while (rtnh_ok(rtnh, remaining)) {
5291 memcpy(&r_cfg, cfg, sizeof(*cfg));
5292 if (rtnh->rtnh_ifindex)
5293 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5294
5295 attrlen = rtnh_attrlen(rtnh);
5296 if (attrlen > 0) {
5297 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5298
5299 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5300 if (nla) {
5301 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
5302 r_cfg.fc_flags |= RTF_GATEWAY;
5303 }
5304 }
5305 err = ip6_route_del(&r_cfg, extack);
5306 if (err)
5307 last_err = err;
5308
5309 rtnh = rtnh_next(rtnh, &remaining);
5310 }
5311
5312 return last_err;
5313 }
5314
5315 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5316 struct netlink_ext_ack *extack)
5317 {
5318 struct fib6_config cfg;
5319 int err;
5320
5321 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5322 if (err < 0)
5323 return err;
5324
5325 if (cfg.fc_nh_id &&
5326 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5327 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5328 return -EINVAL;
5329 }
5330
5331 if (cfg.fc_mp)
5332 return ip6_route_multipath_del(&cfg, extack);
5333 else {
5334 cfg.fc_delete_all_nh = 1;
5335 return ip6_route_del(&cfg, extack);
5336 }
5337 }
5338
5339 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5340 struct netlink_ext_ack *extack)
5341 {
5342 struct fib6_config cfg;
5343 int err;
5344
5345 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5346 if (err < 0)
5347 return err;
5348
5349 if (cfg.fc_metric == 0)
5350 cfg.fc_metric = IP6_RT_PRIO_USER;
5351
5352 if (cfg.fc_mp)
5353 return ip6_route_multipath_add(&cfg, extack);
5354 else
5355 return ip6_route_add(&cfg, GFP_KERNEL, extack);
5356 }
5357
5358 /* add the overhead of this fib6_nh to nexthop_len */
5359 static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5360 {
5361 int *nexthop_len = arg;
5362
5363 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5364 + NLA_ALIGN(sizeof(struct rtnexthop))
5365 + nla_total_size(16); /* RTA_GATEWAY */
5366
5367 if (nh->fib_nh_lws) {
5368 /* RTA_ENCAP_TYPE */
5369 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5370 /* RTA_ENCAP */
5371 *nexthop_len += nla_total_size(2);
5372 }
5373
5374 return 0;
5375 }
5376
5377 static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5378 {
5379 int nexthop_len;
5380
5381 if (f6i->nh) {
5382 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5383 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5384 &nexthop_len);
5385 } else {
5386 struct fib6_nh *nh = f6i->fib6_nh;
5387
5388 nexthop_len = 0;
5389 if (f6i->fib6_nsiblings) {
5390 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
5391 + NLA_ALIGN(sizeof(struct rtnexthop))
5392 + nla_total_size(16) /* RTA_GATEWAY */
5393 + lwtunnel_get_encap_size(nh->fib_nh_lws);
5394
5395 nexthop_len *= f6i->fib6_nsiblings;
5396 }
5397 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5398 }
5399
5400 return NLMSG_ALIGN(sizeof(struct rtmsg))
5401 + nla_total_size(16) /* RTA_SRC */
5402 + nla_total_size(16) /* RTA_DST */
5403 + nla_total_size(16) /* RTA_GATEWAY */
5404 + nla_total_size(16) /* RTA_PREFSRC */
5405 + nla_total_size(4) /* RTA_TABLE */
5406 + nla_total_size(4) /* RTA_IIF */
5407 + nla_total_size(4) /* RTA_OIF */
5408 + nla_total_size(4) /* RTA_PRIORITY */
5409 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5410 + nla_total_size(sizeof(struct rta_cacheinfo))
5411 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5412 + nla_total_size(1) /* RTA_PREF */
5413 + nexthop_len;
5414 }
5415
5416 static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5417 unsigned char *flags)
5418 {
5419 if (nexthop_is_multipath(nh)) {
5420 struct nlattr *mp;
5421
5422 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5423 if (!mp)
5424 goto nla_put_failure;
5425
5426 if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
5427 goto nla_put_failure;
5428
5429 nla_nest_end(skb, mp);
5430 } else {
5431 struct fib6_nh *fib6_nh;
5432
5433 fib6_nh = nexthop_fib6_nh(nh);
5434 if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
5435 flags, false) < 0)
5436 goto nla_put_failure;
5437 }
5438
5439 return 0;
5440
5441 nla_put_failure:
5442 return -EMSGSIZE;
5443 }
5444
5445 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5446 struct fib6_info *rt, struct dst_entry *dst,
5447 struct in6_addr *dest, struct in6_addr *src,
5448 int iif, int type, u32 portid, u32 seq,
5449 unsigned int flags)
5450 {
5451 struct rt6_info *rt6 = (struct rt6_info *)dst;
5452 struct rt6key *rt6_dst, *rt6_src;
5453 u32 *pmetrics, table, rt6_flags;
5454 unsigned char nh_flags = 0;
5455 struct nlmsghdr *nlh;
5456 struct rtmsg *rtm;
5457 long expires = 0;
5458
5459 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5460 if (!nlh)
5461 return -EMSGSIZE;
5462
5463 if (rt6) {
5464 rt6_dst = &rt6->rt6i_dst;
5465 rt6_src = &rt6->rt6i_src;
5466 rt6_flags = rt6->rt6i_flags;
5467 } else {
5468 rt6_dst = &rt->fib6_dst;
5469 rt6_src = &rt->fib6_src;
5470 rt6_flags = rt->fib6_flags;
5471 }
5472
5473 rtm = nlmsg_data(nlh);
5474 rtm->rtm_family = AF_INET6;
5475 rtm->rtm_dst_len = rt6_dst->plen;
5476 rtm->rtm_src_len = rt6_src->plen;
5477 rtm->rtm_tos = 0;
5478 if (rt->fib6_table)
5479 table = rt->fib6_table->tb6_id;
5480 else
5481 table = RT6_TABLE_UNSPEC;
5482 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5483 if (nla_put_u32(skb, RTA_TABLE, table))
5484 goto nla_put_failure;
5485
5486 rtm->rtm_type = rt->fib6_type;
5487 rtm->rtm_flags = 0;
5488 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5489 rtm->rtm_protocol = rt->fib6_protocol;
5490
5491 if (rt6_flags & RTF_CACHE)
5492 rtm->rtm_flags |= RTM_F_CLONED;
5493
5494 if (dest) {
5495 if (nla_put_in6_addr(skb, RTA_DST, dest))
5496 goto nla_put_failure;
5497 rtm->rtm_dst_len = 128;
5498 } else if (rtm->rtm_dst_len)
5499 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5500 goto nla_put_failure;
5501 #ifdef CONFIG_IPV6_SUBTREES
5502 if (src) {
5503 if (nla_put_in6_addr(skb, RTA_SRC, src))
5504 goto nla_put_failure;
5505 rtm->rtm_src_len = 128;
5506 } else if (rtm->rtm_src_len &&
5507 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5508 goto nla_put_failure;
5509 #endif
5510 if (iif) {
5511 #ifdef CONFIG_IPV6_MROUTE
5512 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5513 int err = ip6mr_get_route(net, skb, rtm, portid);
5514
5515 if (err == 0)
5516 return 0;
5517 if (err < 0)
5518 goto nla_put_failure;
5519 } else
5520 #endif
5521 if (nla_put_u32(skb, RTA_IIF, iif))
5522 goto nla_put_failure;
5523 } else if (dest) {
5524 struct in6_addr saddr_buf;
5525 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
5526 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5527 goto nla_put_failure;
5528 }
5529
5530 if (rt->fib6_prefsrc.plen) {
5531 struct in6_addr saddr_buf;
5532 saddr_buf = rt->fib6_prefsrc.addr;
5533 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5534 goto nla_put_failure;
5535 }
5536
5537 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5538 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5539 goto nla_put_failure;
5540
5541 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5542 goto nla_put_failure;
5543
5544 /* For multipath routes, walk the siblings list and add
5545 * each as a nexthop within RTA_MULTIPATH.
5546 */
5547 if (rt6) {
5548 if (rt6_flags & RTF_GATEWAY &&
5549 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5550 goto nla_put_failure;
5551
5552 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5553 goto nla_put_failure;
5554 } else if (rt->fib6_nsiblings) {
5555 struct fib6_info *sibling, *next_sibling;
5556 struct nlattr *mp;
5557
5558 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5559 if (!mp)
5560 goto nla_put_failure;
5561
5562 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5563 rt->fib6_nh->fib_nh_weight, AF_INET6) < 0)
5564 goto nla_put_failure;
5565
5566 list_for_each_entry_safe(sibling, next_sibling,
5567 &rt->fib6_siblings, fib6_siblings) {
5568 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5569 sibling->fib6_nh->fib_nh_weight,
5570 AF_INET6) < 0)
5571 goto nla_put_failure;
5572 }
5573
5574 nla_nest_end(skb, mp);
5575 } else if (rt->nh) {
5576 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5577 goto nla_put_failure;
5578
5579 if (nexthop_is_blackhole(rt->nh))
5580 rtm->rtm_type = RTN_BLACKHOLE;
5581
5582 if (rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5583 goto nla_put_failure;
5584
5585 rtm->rtm_flags |= nh_flags;
5586 } else {
5587 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
5588 &nh_flags, false) < 0)
5589 goto nla_put_failure;
5590
5591 rtm->rtm_flags |= nh_flags;
5592 }
5593
5594 if (rt6_flags & RTF_EXPIRES) {
5595 expires = dst ? dst->expires : rt->expires;
5596 expires -= jiffies;
5597 }
5598
5599 if (!dst) {
5600 if (rt->offload)
5601 rtm->rtm_flags |= RTM_F_OFFLOAD;
5602 if (rt->trap)
5603 rtm->rtm_flags |= RTM_F_TRAP;
5604 }
5605
5606 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5607 goto nla_put_failure;
5608
5609 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5610 goto nla_put_failure;
5611
5612
5613 nlmsg_end(skb, nlh);
5614 return 0;
5615
5616 nla_put_failure:
5617 nlmsg_cancel(skb, nlh);
5618 return -EMSGSIZE;
5619 }
5620
5621 static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5622 {
5623 const struct net_device *dev = arg;
5624
5625 if (nh->fib_nh_dev == dev)
5626 return 1;
5627
5628 return 0;
5629 }
5630
5631 static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5632 const struct net_device *dev)
5633 {
5634 if (f6i->nh) {
5635 struct net_device *_dev = (struct net_device *)dev;
5636
5637 return !!nexthop_for_each_fib6_nh(f6i->nh,
5638 fib6_info_nh_uses_dev,
5639 _dev);
5640 }
5641
5642 if (f6i->fib6_nh->fib_nh_dev == dev)
5643 return true;
5644
5645 if (f6i->fib6_nsiblings) {
5646 struct fib6_info *sibling, *next_sibling;
5647
5648 list_for_each_entry_safe(sibling, next_sibling,
5649 &f6i->fib6_siblings, fib6_siblings) {
5650 if (sibling->fib6_nh->fib_nh_dev == dev)
5651 return true;
5652 }
5653 }
5654
5655 return false;
5656 }
5657
5658 struct fib6_nh_exception_dump_walker {
5659 struct rt6_rtnl_dump_arg *dump;
5660 struct fib6_info *rt;
5661 unsigned int flags;
5662 unsigned int skip;
5663 unsigned int count;
5664 };
5665
5666 static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5667 {
5668 struct fib6_nh_exception_dump_walker *w = arg;
5669 struct rt6_rtnl_dump_arg *dump = w->dump;
5670 struct rt6_exception_bucket *bucket;
5671 struct rt6_exception *rt6_ex;
5672 int i, err;
5673
5674 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5675 if (!bucket)
5676 return 0;
5677
5678 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5679 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5680 if (w->skip) {
5681 w->skip--;
5682 continue;
5683 }
5684
5685 /* Expiration of entries doesn't bump sernum, insertion
5686 * does. Removal is triggered by insertion, so we can
5687 * rely on the fact that if entries change between two
5688 * partial dumps, this node is scanned again completely,
5689 * see rt6_insert_exception() and fib6_dump_table().
5690 *
5691 * Count expired entries we go through as handled
5692 * entries that we'll skip next time, in case of partial
5693 * node dump. Otherwise, if entries expire meanwhile,
5694 * we'll skip the wrong amount.
5695 */
5696 if (rt6_check_expired(rt6_ex->rt6i)) {
5697 w->count++;
5698 continue;
5699 }
5700
5701 err = rt6_fill_node(dump->net, dump->skb, w->rt,
5702 &rt6_ex->rt6i->dst, NULL, NULL, 0,
5703 RTM_NEWROUTE,
5704 NETLINK_CB(dump->cb->skb).portid,
5705 dump->cb->nlh->nlmsg_seq, w->flags);
5706 if (err)
5707 return err;
5708
5709 w->count++;
5710 }
5711 bucket++;
5712 }
5713
5714 return 0;
5715 }
5716
5717 /* Return -1 if done with node, number of handled routes on partial dump */
5718 int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5719 {
5720 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5721 struct fib_dump_filter *filter = &arg->filter;
5722 unsigned int flags = NLM_F_MULTI;
5723 struct net *net = arg->net;
5724 int count = 0;
5725
5726 if (rt == net->ipv6.fib6_null_entry)
5727 return -1;
5728
5729 if ((filter->flags & RTM_F_PREFIX) &&
5730 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5731 /* success since this is not a prefix route */
5732 return -1;
5733 }
5734 if (filter->filter_set &&
5735 ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5736 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5737 (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5738 return -1;
5739 }
5740
5741 if (filter->filter_set ||
5742 !filter->dump_routes || !filter->dump_exceptions) {
5743 flags |= NLM_F_DUMP_FILTERED;
5744 }
5745
5746 if (filter->dump_routes) {
5747 if (skip) {
5748 skip--;
5749 } else {
5750 if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5751 0, RTM_NEWROUTE,
5752 NETLINK_CB(arg->cb->skb).portid,
5753 arg->cb->nlh->nlmsg_seq, flags)) {
5754 return 0;
5755 }
5756 count++;
5757 }
5758 }
5759
5760 if (filter->dump_exceptions) {
5761 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5762 .rt = rt,
5763 .flags = flags,
5764 .skip = skip,
5765 .count = 0 };
5766 int err;
5767
5768 rcu_read_lock();
5769 if (rt->nh) {
5770 err = nexthop_for_each_fib6_nh(rt->nh,
5771 rt6_nh_dump_exceptions,
5772 &w);
5773 } else {
5774 err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5775 }
5776 rcu_read_unlock();
5777
5778 if (err)
5779 return count += w.count;
5780 }
5781
5782 return -1;
5783 }
5784
5785 static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5786 const struct nlmsghdr *nlh,
5787 struct nlattr **tb,
5788 struct netlink_ext_ack *extack)
5789 {
5790 struct rtmsg *rtm;
5791 int i, err;
5792
5793 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5794 NL_SET_ERR_MSG_MOD(extack,
5795 "Invalid header for get route request");
5796 return -EINVAL;
5797 }
5798
5799 if (!netlink_strict_get_check(skb))
5800 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5801 rtm_ipv6_policy, extack);
5802
5803 rtm = nlmsg_data(nlh);
5804 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5805 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5806 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5807 rtm->rtm_type) {
5808 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5809 return -EINVAL;
5810 }
5811 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5812 NL_SET_ERR_MSG_MOD(extack,
5813 "Invalid flags for get route request");
5814 return -EINVAL;
5815 }
5816
5817 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5818 rtm_ipv6_policy, extack);
5819 if (err)
5820 return err;
5821
5822 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5823 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5824 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5825 return -EINVAL;
5826 }
5827
5828 for (i = 0; i <= RTA_MAX; i++) {
5829 if (!tb[i])
5830 continue;
5831
5832 switch (i) {
5833 case RTA_SRC:
5834 case RTA_DST:
5835 case RTA_IIF:
5836 case RTA_OIF:
5837 case RTA_MARK:
5838 case RTA_UID:
5839 case RTA_SPORT:
5840 case RTA_DPORT:
5841 case RTA_IP_PROTO:
5842 break;
5843 default:
5844 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
5845 return -EINVAL;
5846 }
5847 }
5848
5849 return 0;
5850 }
5851
5852 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5853 struct netlink_ext_ack *extack)
5854 {
5855 struct net *net = sock_net(in_skb->sk);
5856 struct nlattr *tb[RTA_MAX+1];
5857 int err, iif = 0, oif = 0;
5858 struct fib6_info *from;
5859 struct dst_entry *dst;
5860 struct rt6_info *rt;
5861 struct sk_buff *skb;
5862 struct rtmsg *rtm;
5863 struct flowi6 fl6 = {};
5864 bool fibmatch;
5865
5866 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
5867 if (err < 0)
5868 goto errout;
5869
5870 err = -EINVAL;
5871 rtm = nlmsg_data(nlh);
5872 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
5873 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
5874
5875 if (tb[RTA_SRC]) {
5876 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
5877 goto errout;
5878
5879 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
5880 }
5881
5882 if (tb[RTA_DST]) {
5883 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
5884 goto errout;
5885
5886 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
5887 }
5888
5889 if (tb[RTA_IIF])
5890 iif = nla_get_u32(tb[RTA_IIF]);
5891
5892 if (tb[RTA_OIF])
5893 oif = nla_get_u32(tb[RTA_OIF]);
5894
5895 if (tb[RTA_MARK])
5896 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
5897
5898 if (tb[RTA_UID])
5899 fl6.flowi6_uid = make_kuid(current_user_ns(),
5900 nla_get_u32(tb[RTA_UID]));
5901 else
5902 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
5903
5904 if (tb[RTA_SPORT])
5905 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
5906
5907 if (tb[RTA_DPORT])
5908 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
5909
5910 if (tb[RTA_IP_PROTO]) {
5911 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
5912 &fl6.flowi6_proto, AF_INET6,
5913 extack);
5914 if (err)
5915 goto errout;
5916 }
5917
5918 if (iif) {
5919 struct net_device *dev;
5920 int flags = 0;
5921
5922 rcu_read_lock();
5923
5924 dev = dev_get_by_index_rcu(net, iif);
5925 if (!dev) {
5926 rcu_read_unlock();
5927 err = -ENODEV;
5928 goto errout;
5929 }
5930
5931 fl6.flowi6_iif = iif;
5932
5933 if (!ipv6_addr_any(&fl6.saddr))
5934 flags |= RT6_LOOKUP_F_HAS_SADDR;
5935
5936 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
5937
5938 rcu_read_unlock();
5939 } else {
5940 fl6.flowi6_oif = oif;
5941
5942 dst = ip6_route_output(net, NULL, &fl6);
5943 }
5944
5945
5946 rt = container_of(dst, struct rt6_info, dst);
5947 if (rt->dst.error) {
5948 err = rt->dst.error;
5949 ip6_rt_put(rt);
5950 goto errout;
5951 }
5952
5953 if (rt == net->ipv6.ip6_null_entry) {
5954 err = rt->dst.error;
5955 ip6_rt_put(rt);
5956 goto errout;
5957 }
5958
5959 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
5960 if (!skb) {
5961 ip6_rt_put(rt);
5962 err = -ENOBUFS;
5963 goto errout;
5964 }
5965
5966 skb_dst_set(skb, &rt->dst);
5967
5968 rcu_read_lock();
5969 from = rcu_dereference(rt->from);
5970 if (from) {
5971 if (fibmatch)
5972 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
5973 iif, RTM_NEWROUTE,
5974 NETLINK_CB(in_skb).portid,
5975 nlh->nlmsg_seq, 0);
5976 else
5977 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
5978 &fl6.saddr, iif, RTM_NEWROUTE,
5979 NETLINK_CB(in_skb).portid,
5980 nlh->nlmsg_seq, 0);
5981 } else {
5982 err = -ENETUNREACH;
5983 }
5984 rcu_read_unlock();
5985
5986 if (err < 0) {
5987 kfree_skb(skb);
5988 goto errout;
5989 }
5990
5991 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5992 errout:
5993 return err;
5994 }
5995
5996 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
5997 unsigned int nlm_flags)
5998 {
5999 struct sk_buff *skb;
6000 struct net *net = info->nl_net;
6001 u32 seq;
6002 int err;
6003
6004 err = -ENOBUFS;
6005 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6006
6007 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6008 if (!skb)
6009 goto errout;
6010
6011 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6012 event, info->portid, seq, nlm_flags);
6013 if (err < 0) {
6014 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6015 WARN_ON(err == -EMSGSIZE);
6016 kfree_skb(skb);
6017 goto errout;
6018 }
6019 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6020 info->nlh, gfp_any());
6021 return;
6022 errout:
6023 if (err < 0)
6024 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6025 }
6026
6027 void fib6_rt_update(struct net *net, struct fib6_info *rt,
6028 struct nl_info *info)
6029 {
6030 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6031 struct sk_buff *skb;
6032 int err = -ENOBUFS;
6033
6034 /* call_fib6_entry_notifiers will be removed when in-kernel notifier
6035 * is implemented and supported for nexthop objects
6036 */
6037 call_fib6_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, rt, NULL);
6038
6039 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6040 if (!skb)
6041 goto errout;
6042
6043 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6044 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6045 if (err < 0) {
6046 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6047 WARN_ON(err == -EMSGSIZE);
6048 kfree_skb(skb);
6049 goto errout;
6050 }
6051 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6052 info->nlh, gfp_any());
6053 return;
6054 errout:
6055 if (err < 0)
6056 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6057 }
6058
6059 static int ip6_route_dev_notify(struct notifier_block *this,
6060 unsigned long event, void *ptr)
6061 {
6062 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6063 struct net *net = dev_net(dev);
6064
6065 if (!(dev->flags & IFF_LOOPBACK))
6066 return NOTIFY_OK;
6067
6068 if (event == NETDEV_REGISTER) {
6069 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
6070 net->ipv6.ip6_null_entry->dst.dev = dev;
6071 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6072 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6073 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
6074 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
6075 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
6076 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
6077 #endif
6078 } else if (event == NETDEV_UNREGISTER &&
6079 dev->reg_state != NETREG_UNREGISTERED) {
6080 /* NETDEV_UNREGISTER could be fired for multiple times by
6081 * netdev_wait_allrefs(). Make sure we only call this once.
6082 */
6083 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
6084 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6085 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6086 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
6087 #endif
6088 }
6089
6090 return NOTIFY_OK;
6091 }
6092
6093 /*
6094 * /proc
6095 */
6096
6097 #ifdef CONFIG_PROC_FS
6098 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6099 {
6100 struct net *net = (struct net *)seq->private;
6101 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6102 net->ipv6.rt6_stats->fib_nodes,
6103 net->ipv6.rt6_stats->fib_route_nodes,
6104 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6105 net->ipv6.rt6_stats->fib_rt_entries,
6106 net->ipv6.rt6_stats->fib_rt_cache,
6107 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6108 net->ipv6.rt6_stats->fib_discarded_routes);
6109
6110 return 0;
6111 }
6112 #endif /* CONFIG_PROC_FS */
6113
6114 #ifdef CONFIG_SYSCTL
6115
6116 static
6117 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
6118 void __user *buffer, size_t *lenp, loff_t *ppos)
6119 {
6120 struct net *net;
6121 int delay;
6122 int ret;
6123 if (!write)
6124 return -EINVAL;
6125
6126 net = (struct net *)ctl->extra1;
6127 delay = net->ipv6.sysctl.flush_delay;
6128 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6129 if (ret)
6130 return ret;
6131
6132 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6133 return 0;
6134 }
6135
6136 static struct ctl_table ipv6_route_table_template[] = {
6137 {
6138 .procname = "flush",
6139 .data = &init_net.ipv6.sysctl.flush_delay,
6140 .maxlen = sizeof(int),
6141 .mode = 0200,
6142 .proc_handler = ipv6_sysctl_rtcache_flush
6143 },
6144 {
6145 .procname = "gc_thresh",
6146 .data = &ip6_dst_ops_template.gc_thresh,
6147 .maxlen = sizeof(int),
6148 .mode = 0644,
6149 .proc_handler = proc_dointvec,
6150 },
6151 {
6152 .procname = "max_size",
6153 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
6154 .maxlen = sizeof(int),
6155 .mode = 0644,
6156 .proc_handler = proc_dointvec,
6157 },
6158 {
6159 .procname = "gc_min_interval",
6160 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6161 .maxlen = sizeof(int),
6162 .mode = 0644,
6163 .proc_handler = proc_dointvec_jiffies,
6164 },
6165 {
6166 .procname = "gc_timeout",
6167 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6168 .maxlen = sizeof(int),
6169 .mode = 0644,
6170 .proc_handler = proc_dointvec_jiffies,
6171 },
6172 {
6173 .procname = "gc_interval",
6174 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
6175 .maxlen = sizeof(int),
6176 .mode = 0644,
6177 .proc_handler = proc_dointvec_jiffies,
6178 },
6179 {
6180 .procname = "gc_elasticity",
6181 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6182 .maxlen = sizeof(int),
6183 .mode = 0644,
6184 .proc_handler = proc_dointvec,
6185 },
6186 {
6187 .procname = "mtu_expires",
6188 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6189 .maxlen = sizeof(int),
6190 .mode = 0644,
6191 .proc_handler = proc_dointvec_jiffies,
6192 },
6193 {
6194 .procname = "min_adv_mss",
6195 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
6196 .maxlen = sizeof(int),
6197 .mode = 0644,
6198 .proc_handler = proc_dointvec,
6199 },
6200 {
6201 .procname = "gc_min_interval_ms",
6202 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6203 .maxlen = sizeof(int),
6204 .mode = 0644,
6205 .proc_handler = proc_dointvec_ms_jiffies,
6206 },
6207 {
6208 .procname = "skip_notify_on_dev_down",
6209 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6210 .maxlen = sizeof(int),
6211 .mode = 0644,
6212 .proc_handler = proc_dointvec_minmax,
6213 .extra1 = SYSCTL_ZERO,
6214 .extra2 = SYSCTL_ONE,
6215 },
6216 { }
6217 };
6218
6219 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6220 {
6221 struct ctl_table *table;
6222
6223 table = kmemdup(ipv6_route_table_template,
6224 sizeof(ipv6_route_table_template),
6225 GFP_KERNEL);
6226
6227 if (table) {
6228 table[0].data = &net->ipv6.sysctl.flush_delay;
6229 table[0].extra1 = net;
6230 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6231 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
6232 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6233 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6234 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6235 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6236 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6237 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6238 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6239 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6240
6241 /* Don't export sysctls to unprivileged users */
6242 if (net->user_ns != &init_user_ns)
6243 table[0].procname = NULL;
6244 }
6245
6246 return table;
6247 }
6248 #endif
6249
6250 static int __net_init ip6_route_net_init(struct net *net)
6251 {
6252 int ret = -ENOMEM;
6253
6254 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6255 sizeof(net->ipv6.ip6_dst_ops));
6256
6257 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6258 goto out_ip6_dst_ops;
6259
6260 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6261 if (!net->ipv6.fib6_null_entry)
6262 goto out_ip6_dst_entries;
6263 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6264 sizeof(*net->ipv6.fib6_null_entry));
6265
6266 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6267 sizeof(*net->ipv6.ip6_null_entry),
6268 GFP_KERNEL);
6269 if (!net->ipv6.ip6_null_entry)
6270 goto out_fib6_null_entry;
6271 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6272 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6273 ip6_template_metrics, true);
6274 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->rt6i_uncached);
6275
6276 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6277 net->ipv6.fib6_has_custom_rules = false;
6278 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6279 sizeof(*net->ipv6.ip6_prohibit_entry),
6280 GFP_KERNEL);
6281 if (!net->ipv6.ip6_prohibit_entry)
6282 goto out_ip6_null_entry;
6283 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6284 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6285 ip6_template_metrics, true);
6286 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->rt6i_uncached);
6287
6288 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6289 sizeof(*net->ipv6.ip6_blk_hole_entry),
6290 GFP_KERNEL);
6291 if (!net->ipv6.ip6_blk_hole_entry)
6292 goto out_ip6_prohibit_entry;
6293 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6294 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6295 ip6_template_metrics, true);
6296 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->rt6i_uncached);
6297 #ifdef CONFIG_IPV6_SUBTREES
6298 net->ipv6.fib6_routes_require_src = 0;
6299 #endif
6300 #endif
6301
6302 net->ipv6.sysctl.flush_delay = 0;
6303 net->ipv6.sysctl.ip6_rt_max_size = 4096;
6304 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6305 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6306 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6307 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6308 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6309 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6310 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6311
6312 net->ipv6.ip6_rt_gc_expire = 30*HZ;
6313
6314 ret = 0;
6315 out:
6316 return ret;
6317
6318 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6319 out_ip6_prohibit_entry:
6320 kfree(net->ipv6.ip6_prohibit_entry);
6321 out_ip6_null_entry:
6322 kfree(net->ipv6.ip6_null_entry);
6323 #endif
6324 out_fib6_null_entry:
6325 kfree(net->ipv6.fib6_null_entry);
6326 out_ip6_dst_entries:
6327 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6328 out_ip6_dst_ops:
6329 goto out;
6330 }
6331
6332 static void __net_exit ip6_route_net_exit(struct net *net)
6333 {
6334 kfree(net->ipv6.fib6_null_entry);
6335 kfree(net->ipv6.ip6_null_entry);
6336 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6337 kfree(net->ipv6.ip6_prohibit_entry);
6338 kfree(net->ipv6.ip6_blk_hole_entry);
6339 #endif
6340 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6341 }
6342
6343 static int __net_init ip6_route_net_init_late(struct net *net)
6344 {
6345 #ifdef CONFIG_PROC_FS
6346 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6347 sizeof(struct ipv6_route_iter));
6348 proc_create_net_single("rt6_stats", 0444, net->proc_net,
6349 rt6_stats_seq_show, NULL);
6350 #endif
6351 return 0;
6352 }
6353
6354 static void __net_exit ip6_route_net_exit_late(struct net *net)
6355 {
6356 #ifdef CONFIG_PROC_FS
6357 remove_proc_entry("ipv6_route", net->proc_net);
6358 remove_proc_entry("rt6_stats", net->proc_net);
6359 #endif
6360 }
6361
6362 static struct pernet_operations ip6_route_net_ops = {
6363 .init = ip6_route_net_init,
6364 .exit = ip6_route_net_exit,
6365 };
6366
6367 static int __net_init ipv6_inetpeer_init(struct net *net)
6368 {
6369 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6370
6371 if (!bp)
6372 return -ENOMEM;
6373 inet_peer_base_init(bp);
6374 net->ipv6.peers = bp;
6375 return 0;
6376 }
6377
6378 static void __net_exit ipv6_inetpeer_exit(struct net *net)
6379 {
6380 struct inet_peer_base *bp = net->ipv6.peers;
6381
6382 net->ipv6.peers = NULL;
6383 inetpeer_invalidate_tree(bp);
6384 kfree(bp);
6385 }
6386
6387 static struct pernet_operations ipv6_inetpeer_ops = {
6388 .init = ipv6_inetpeer_init,
6389 .exit = ipv6_inetpeer_exit,
6390 };
6391
6392 static struct pernet_operations ip6_route_net_late_ops = {
6393 .init = ip6_route_net_init_late,
6394 .exit = ip6_route_net_exit_late,
6395 };
6396
6397 static struct notifier_block ip6_route_dev_notifier = {
6398 .notifier_call = ip6_route_dev_notify,
6399 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6400 };
6401
6402 void __init ip6_route_init_special_entries(void)
6403 {
6404 /* Registering of the loopback is done before this portion of code,
6405 * the loopback reference in rt6_info will not be taken, do it
6406 * manually for init_net */
6407 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6408 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6409 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6410 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6411 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6412 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6413 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6414 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6415 #endif
6416 }
6417
6418 int __init ip6_route_init(void)
6419 {
6420 int ret;
6421 int cpu;
6422
6423 ret = -ENOMEM;
6424 ip6_dst_ops_template.kmem_cachep =
6425 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6426 SLAB_HWCACHE_ALIGN, NULL);
6427 if (!ip6_dst_ops_template.kmem_cachep)
6428 goto out;
6429
6430 ret = dst_entries_init(&ip6_dst_blackhole_ops);
6431 if (ret)
6432 goto out_kmem_cache;
6433
6434 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6435 if (ret)
6436 goto out_dst_entries;
6437
6438 ret = register_pernet_subsys(&ip6_route_net_ops);
6439 if (ret)
6440 goto out_register_inetpeer;
6441
6442 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6443
6444 ret = fib6_init();
6445 if (ret)
6446 goto out_register_subsys;
6447
6448 ret = xfrm6_init();
6449 if (ret)
6450 goto out_fib6_init;
6451
6452 ret = fib6_rules_init();
6453 if (ret)
6454 goto xfrm6_init;
6455
6456 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6457 if (ret)
6458 goto fib6_rules_init;
6459
6460 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6461 inet6_rtm_newroute, NULL, 0);
6462 if (ret < 0)
6463 goto out_register_late_subsys;
6464
6465 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6466 inet6_rtm_delroute, NULL, 0);
6467 if (ret < 0)
6468 goto out_register_late_subsys;
6469
6470 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6471 inet6_rtm_getroute, NULL,
6472 RTNL_FLAG_DOIT_UNLOCKED);
6473 if (ret < 0)
6474 goto out_register_late_subsys;
6475
6476 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6477 if (ret)
6478 goto out_register_late_subsys;
6479
6480 for_each_possible_cpu(cpu) {
6481 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6482
6483 INIT_LIST_HEAD(&ul->head);
6484 spin_lock_init(&ul->lock);
6485 }
6486
6487 out:
6488 return ret;
6489
6490 out_register_late_subsys:
6491 rtnl_unregister_all(PF_INET6);
6492 unregister_pernet_subsys(&ip6_route_net_late_ops);
6493 fib6_rules_init:
6494 fib6_rules_cleanup();
6495 xfrm6_init:
6496 xfrm6_fini();
6497 out_fib6_init:
6498 fib6_gc_cleanup();
6499 out_register_subsys:
6500 unregister_pernet_subsys(&ip6_route_net_ops);
6501 out_register_inetpeer:
6502 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6503 out_dst_entries:
6504 dst_entries_destroy(&ip6_dst_blackhole_ops);
6505 out_kmem_cache:
6506 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6507 goto out;
6508 }
6509
6510 void ip6_route_cleanup(void)
6511 {
6512 unregister_netdevice_notifier(&ip6_route_dev_notifier);
6513 unregister_pernet_subsys(&ip6_route_net_late_ops);
6514 fib6_rules_cleanup();
6515 xfrm6_fini();
6516 fib6_gc_cleanup();
6517 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6518 unregister_pernet_subsys(&ip6_route_net_ops);
6519 dst_entries_destroy(&ip6_dst_blackhole_ops);
6520 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6521 }
Linux with added FPC-III support