Loading...
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 <linux/siphash.h>
45#include <net/net_namespace.h>
46#include <net/snmp.h>
47#include <net/ipv6.h>
48#include <net/ip6_fib.h>
49#include <net/ip6_route.h>
50#include <net/ndisc.h>
51#include <net/addrconf.h>
52#include <net/tcp.h>
53#include <linux/rtnetlink.h>
54#include <net/dst.h>
55#include <net/dst_metadata.h>
56#include <net/xfrm.h>
57#include <net/netevent.h>
58#include <net/netlink.h>
59#include <net/rtnh.h>
60#include <net/lwtunnel.h>
61#include <net/ip_tunnels.h>
62#include <net/l3mdev.h>
63#include <net/ip.h>
64#include <linux/uaccess.h>
65#include <linux/btf_ids.h>
66
67#ifdef CONFIG_SYSCTL
68#include <linux/sysctl.h>
69#endif
70
71static int ip6_rt_type_to_error(u8 fib6_type);
72
73#define CREATE_TRACE_POINTS
74#include <trace/events/fib6.h>
75EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
76#undef CREATE_TRACE_POINTS
77
78enum rt6_nud_state {
79 RT6_NUD_FAIL_HARD = -3,
80 RT6_NUD_FAIL_PROBE = -2,
81 RT6_NUD_FAIL_DO_RR = -1,
82 RT6_NUD_SUCCEED = 1
83};
84
85INDIRECT_CALLABLE_SCOPE
86struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
87static unsigned int ip6_default_advmss(const struct dst_entry *dst);
88INDIRECT_CALLABLE_SCOPE
89unsigned int ip6_mtu(const struct dst_entry *dst);
90static void ip6_negative_advice(struct sock *sk,
91 struct dst_entry *dst);
92static void ip6_dst_destroy(struct dst_entry *);
93static void ip6_dst_ifdown(struct dst_entry *,
94 struct net_device *dev);
95static void ip6_dst_gc(struct dst_ops *ops);
96
97static int ip6_pkt_discard(struct sk_buff *skb);
98static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
99static int ip6_pkt_prohibit(struct sk_buff *skb);
100static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
101static void ip6_link_failure(struct sk_buff *skb);
102static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
103 struct sk_buff *skb, u32 mtu,
104 bool confirm_neigh);
105static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
106 struct sk_buff *skb);
107static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
108 int strict);
109static size_t rt6_nlmsg_size(struct fib6_info *f6i);
110static int rt6_fill_node(struct net *net, struct sk_buff *skb,
111 struct fib6_info *rt, struct dst_entry *dst,
112 struct in6_addr *dest, struct in6_addr *src,
113 int iif, int type, u32 portid, u32 seq,
114 unsigned int flags);
115static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
116 const struct in6_addr *daddr,
117 const struct in6_addr *saddr);
118
119#ifdef CONFIG_IPV6_ROUTE_INFO
120static struct fib6_info *rt6_add_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 unsigned int pref);
125static struct fib6_info *rt6_get_route_info(struct net *net,
126 const struct in6_addr *prefix, int prefixlen,
127 const struct in6_addr *gwaddr,
128 struct net_device *dev);
129#endif
130
131struct uncached_list {
132 spinlock_t lock;
133 struct list_head head;
134};
135
136static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
137
138void rt6_uncached_list_add(struct rt6_info *rt)
139{
140 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
141
142 rt->dst.rt_uncached_list = ul;
143
144 spin_lock_bh(&ul->lock);
145 list_add_tail(&rt->dst.rt_uncached, &ul->head);
146 spin_unlock_bh(&ul->lock);
147}
148
149void rt6_uncached_list_del(struct rt6_info *rt)
150{
151 if (!list_empty(&rt->dst.rt_uncached)) {
152 struct uncached_list *ul = rt->dst.rt_uncached_list;
153
154 spin_lock_bh(&ul->lock);
155 list_del_init(&rt->dst.rt_uncached);
156 spin_unlock_bh(&ul->lock);
157 }
158}
159
160static void rt6_uncached_list_flush_dev(struct net_device *dev)
161{
162 int cpu;
163
164 for_each_possible_cpu(cpu) {
165 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
166 struct rt6_info *rt, *safe;
167
168 if (list_empty(&ul->head))
169 continue;
170
171 spin_lock_bh(&ul->lock);
172 list_for_each_entry_safe(rt, safe, &ul->head, dst.rt_uncached) {
173 struct inet6_dev *rt_idev = rt->rt6i_idev;
174 struct net_device *rt_dev = rt->dst.dev;
175 bool handled = false;
176
177 if (rt_idev && rt_idev->dev == dev) {
178 rt->rt6i_idev = in6_dev_get(blackhole_netdev);
179 in6_dev_put(rt_idev);
180 handled = true;
181 }
182
183 if (rt_dev == dev) {
184 rt->dst.dev = blackhole_netdev;
185 netdev_ref_replace(rt_dev, blackhole_netdev,
186 &rt->dst.dev_tracker,
187 GFP_ATOMIC);
188 handled = true;
189 }
190 if (handled)
191 list_del_init(&rt->dst.rt_uncached);
192 }
193 spin_unlock_bh(&ul->lock);
194 }
195}
196
197static inline const void *choose_neigh_daddr(const struct in6_addr *p,
198 struct sk_buff *skb,
199 const void *daddr)
200{
201 if (!ipv6_addr_any(p))
202 return (const void *) p;
203 else if (skb)
204 return &ipv6_hdr(skb)->daddr;
205 return daddr;
206}
207
208struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
209 struct net_device *dev,
210 struct sk_buff *skb,
211 const void *daddr)
212{
213 struct neighbour *n;
214
215 daddr = choose_neigh_daddr(gw, skb, daddr);
216 n = __ipv6_neigh_lookup(dev, daddr);
217 if (n)
218 return n;
219
220 n = neigh_create(&nd_tbl, daddr, dev);
221 return IS_ERR(n) ? NULL : n;
222}
223
224static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
225 struct sk_buff *skb,
226 const void *daddr)
227{
228 const struct rt6_info *rt = dst_rt6_info(dst);
229
230 return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
231 dst->dev, skb, daddr);
232}
233
234static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
235{
236 const struct rt6_info *rt = dst_rt6_info(dst);
237 struct net_device *dev = dst->dev;
238
239 daddr = choose_neigh_daddr(rt6_nexthop(rt, &in6addr_any), NULL, daddr);
240 if (!daddr)
241 return;
242 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
243 return;
244 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
245 return;
246 __ipv6_confirm_neigh(dev, daddr);
247}
248
249static struct dst_ops ip6_dst_ops_template = {
250 .family = AF_INET6,
251 .gc = ip6_dst_gc,
252 .gc_thresh = 1024,
253 .check = ip6_dst_check,
254 .default_advmss = ip6_default_advmss,
255 .mtu = ip6_mtu,
256 .cow_metrics = dst_cow_metrics_generic,
257 .destroy = ip6_dst_destroy,
258 .ifdown = ip6_dst_ifdown,
259 .negative_advice = ip6_negative_advice,
260 .link_failure = ip6_link_failure,
261 .update_pmtu = ip6_rt_update_pmtu,
262 .redirect = rt6_do_redirect,
263 .local_out = __ip6_local_out,
264 .neigh_lookup = ip6_dst_neigh_lookup,
265 .confirm_neigh = ip6_confirm_neigh,
266};
267
268static struct dst_ops ip6_dst_blackhole_ops = {
269 .family = AF_INET6,
270 .default_advmss = ip6_default_advmss,
271 .neigh_lookup = ip6_dst_neigh_lookup,
272 .check = ip6_dst_check,
273 .destroy = ip6_dst_destroy,
274 .cow_metrics = dst_cow_metrics_generic,
275 .update_pmtu = dst_blackhole_update_pmtu,
276 .redirect = dst_blackhole_redirect,
277 .mtu = dst_blackhole_mtu,
278};
279
280static const u32 ip6_template_metrics[RTAX_MAX] = {
281 [RTAX_HOPLIMIT - 1] = 0,
282};
283
284static const struct fib6_info fib6_null_entry_template = {
285 .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
286 .fib6_protocol = RTPROT_KERNEL,
287 .fib6_metric = ~(u32)0,
288 .fib6_ref = REFCOUNT_INIT(1),
289 .fib6_type = RTN_UNREACHABLE,
290 .fib6_metrics = (struct dst_metrics *)&dst_default_metrics,
291};
292
293static const struct rt6_info ip6_null_entry_template = {
294 .dst = {
295 .__rcuref = RCUREF_INIT(1),
296 .__use = 1,
297 .obsolete = DST_OBSOLETE_FORCE_CHK,
298 .error = -ENETUNREACH,
299 .input = ip6_pkt_discard,
300 .output = ip6_pkt_discard_out,
301 },
302 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
303};
304
305#ifdef CONFIG_IPV6_MULTIPLE_TABLES
306
307static const struct rt6_info ip6_prohibit_entry_template = {
308 .dst = {
309 .__rcuref = RCUREF_INIT(1),
310 .__use = 1,
311 .obsolete = DST_OBSOLETE_FORCE_CHK,
312 .error = -EACCES,
313 .input = ip6_pkt_prohibit,
314 .output = ip6_pkt_prohibit_out,
315 },
316 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
317};
318
319static const struct rt6_info ip6_blk_hole_entry_template = {
320 .dst = {
321 .__rcuref = RCUREF_INIT(1),
322 .__use = 1,
323 .obsolete = DST_OBSOLETE_FORCE_CHK,
324 .error = -EINVAL,
325 .input = dst_discard,
326 .output = dst_discard_out,
327 },
328 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
329};
330
331#endif
332
333static void rt6_info_init(struct rt6_info *rt)
334{
335 memset_after(rt, 0, dst);
336}
337
338/* allocate dst with ip6_dst_ops */
339struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
340 int flags)
341{
342 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
343 DST_OBSOLETE_FORCE_CHK, flags);
344
345 if (rt) {
346 rt6_info_init(rt);
347 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
348 }
349
350 return rt;
351}
352EXPORT_SYMBOL(ip6_dst_alloc);
353
354static void ip6_dst_destroy(struct dst_entry *dst)
355{
356 struct rt6_info *rt = dst_rt6_info(dst);
357 struct fib6_info *from;
358 struct inet6_dev *idev;
359
360 ip_dst_metrics_put(dst);
361 rt6_uncached_list_del(rt);
362
363 idev = rt->rt6i_idev;
364 if (idev) {
365 rt->rt6i_idev = NULL;
366 in6_dev_put(idev);
367 }
368
369 from = unrcu_pointer(xchg(&rt->from, NULL));
370 fib6_info_release(from);
371}
372
373static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
374{
375 struct rt6_info *rt = dst_rt6_info(dst);
376 struct inet6_dev *idev = rt->rt6i_idev;
377 struct fib6_info *from;
378
379 if (idev && idev->dev != blackhole_netdev) {
380 struct inet6_dev *blackhole_idev = in6_dev_get(blackhole_netdev);
381
382 if (blackhole_idev) {
383 rt->rt6i_idev = blackhole_idev;
384 in6_dev_put(idev);
385 }
386 }
387 from = unrcu_pointer(xchg(&rt->from, NULL));
388 fib6_info_release(from);
389}
390
391static bool __rt6_check_expired(const struct rt6_info *rt)
392{
393 if (rt->rt6i_flags & RTF_EXPIRES)
394 return time_after(jiffies, rt->dst.expires);
395 else
396 return false;
397}
398
399static bool rt6_check_expired(const struct rt6_info *rt)
400{
401 struct fib6_info *from;
402
403 from = rcu_dereference(rt->from);
404
405 if (rt->rt6i_flags & RTF_EXPIRES) {
406 if (time_after(jiffies, rt->dst.expires))
407 return true;
408 } else if (from) {
409 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
410 fib6_check_expired(from);
411 }
412 return false;
413}
414
415void fib6_select_path(const struct net *net, struct fib6_result *res,
416 struct flowi6 *fl6, int oif, bool have_oif_match,
417 const struct sk_buff *skb, int strict)
418{
419 struct fib6_info *match = res->f6i;
420 struct fib6_info *sibling;
421
422 if (!match->nh && (!match->fib6_nsiblings || have_oif_match))
423 goto out;
424
425 if (match->nh && have_oif_match && res->nh)
426 return;
427
428 if (skb)
429 IP6CB(skb)->flags |= IP6SKB_MULTIPATH;
430
431 /* We might have already computed the hash for ICMPv6 errors. In such
432 * case it will always be non-zero. Otherwise now is the time to do it.
433 */
434 if (!fl6->mp_hash &&
435 (!match->nh || nexthop_is_multipath(match->nh)))
436 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
437
438 if (unlikely(match->nh)) {
439 nexthop_path_fib6_result(res, fl6->mp_hash);
440 return;
441 }
442
443 if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
444 goto out;
445
446 list_for_each_entry_rcu(sibling, &match->fib6_siblings,
447 fib6_siblings) {
448 const struct fib6_nh *nh = sibling->fib6_nh;
449 int nh_upper_bound;
450
451 nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
452 if (fl6->mp_hash > nh_upper_bound)
453 continue;
454 if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
455 break;
456 match = sibling;
457 break;
458 }
459
460out:
461 res->f6i = match;
462 res->nh = match->fib6_nh;
463}
464
465/*
466 * Route lookup. rcu_read_lock() should be held.
467 */
468
469static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
470 const struct in6_addr *saddr, int oif, int flags)
471{
472 const struct net_device *dev;
473
474 if (nh->fib_nh_flags & RTNH_F_DEAD)
475 return false;
476
477 dev = nh->fib_nh_dev;
478 if (oif) {
479 if (dev->ifindex == oif)
480 return true;
481 } else {
482 if (ipv6_chk_addr(net, saddr, dev,
483 flags & RT6_LOOKUP_F_IFACE))
484 return true;
485 }
486
487 return false;
488}
489
490struct fib6_nh_dm_arg {
491 struct net *net;
492 const struct in6_addr *saddr;
493 int oif;
494 int flags;
495 struct fib6_nh *nh;
496};
497
498static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
499{
500 struct fib6_nh_dm_arg *arg = _arg;
501
502 arg->nh = nh;
503 return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
504 arg->flags);
505}
506
507/* returns fib6_nh from nexthop or NULL */
508static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
509 struct fib6_result *res,
510 const struct in6_addr *saddr,
511 int oif, int flags)
512{
513 struct fib6_nh_dm_arg arg = {
514 .net = net,
515 .saddr = saddr,
516 .oif = oif,
517 .flags = flags,
518 };
519
520 if (nexthop_is_blackhole(nh))
521 return NULL;
522
523 if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
524 return arg.nh;
525
526 return NULL;
527}
528
529static void rt6_device_match(struct net *net, struct fib6_result *res,
530 const struct in6_addr *saddr, int oif, int flags)
531{
532 struct fib6_info *f6i = res->f6i;
533 struct fib6_info *spf6i;
534 struct fib6_nh *nh;
535
536 if (!oif && ipv6_addr_any(saddr)) {
537 if (unlikely(f6i->nh)) {
538 nh = nexthop_fib6_nh(f6i->nh);
539 if (nexthop_is_blackhole(f6i->nh))
540 goto out_blackhole;
541 } else {
542 nh = f6i->fib6_nh;
543 }
544 if (!(nh->fib_nh_flags & RTNH_F_DEAD))
545 goto out;
546 }
547
548 for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
549 bool matched = false;
550
551 if (unlikely(spf6i->nh)) {
552 nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
553 oif, flags);
554 if (nh)
555 matched = true;
556 } else {
557 nh = spf6i->fib6_nh;
558 if (__rt6_device_match(net, nh, saddr, oif, flags))
559 matched = true;
560 }
561 if (matched) {
562 res->f6i = spf6i;
563 goto out;
564 }
565 }
566
567 if (oif && flags & RT6_LOOKUP_F_IFACE) {
568 res->f6i = net->ipv6.fib6_null_entry;
569 nh = res->f6i->fib6_nh;
570 goto out;
571 }
572
573 if (unlikely(f6i->nh)) {
574 nh = nexthop_fib6_nh(f6i->nh);
575 if (nexthop_is_blackhole(f6i->nh))
576 goto out_blackhole;
577 } else {
578 nh = f6i->fib6_nh;
579 }
580
581 if (nh->fib_nh_flags & RTNH_F_DEAD) {
582 res->f6i = net->ipv6.fib6_null_entry;
583 nh = res->f6i->fib6_nh;
584 }
585out:
586 res->nh = nh;
587 res->fib6_type = res->f6i->fib6_type;
588 res->fib6_flags = res->f6i->fib6_flags;
589 return;
590
591out_blackhole:
592 res->fib6_flags |= RTF_REJECT;
593 res->fib6_type = RTN_BLACKHOLE;
594 res->nh = nh;
595}
596
597#ifdef CONFIG_IPV6_ROUTER_PREF
598struct __rt6_probe_work {
599 struct work_struct work;
600 struct in6_addr target;
601 struct net_device *dev;
602 netdevice_tracker dev_tracker;
603};
604
605static void rt6_probe_deferred(struct work_struct *w)
606{
607 struct in6_addr mcaddr;
608 struct __rt6_probe_work *work =
609 container_of(w, struct __rt6_probe_work, work);
610
611 addrconf_addr_solict_mult(&work->target, &mcaddr);
612 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
613 netdev_put(work->dev, &work->dev_tracker);
614 kfree(work);
615}
616
617static void rt6_probe(struct fib6_nh *fib6_nh)
618{
619 struct __rt6_probe_work *work = NULL;
620 const struct in6_addr *nh_gw;
621 unsigned long last_probe;
622 struct neighbour *neigh;
623 struct net_device *dev;
624 struct inet6_dev *idev;
625
626 /*
627 * Okay, this does not seem to be appropriate
628 * for now, however, we need to check if it
629 * is really so; aka Router Reachability Probing.
630 *
631 * Router Reachability Probe MUST be rate-limited
632 * to no more than one per minute.
633 */
634 if (!fib6_nh->fib_nh_gw_family)
635 return;
636
637 nh_gw = &fib6_nh->fib_nh_gw6;
638 dev = fib6_nh->fib_nh_dev;
639 rcu_read_lock();
640 last_probe = READ_ONCE(fib6_nh->last_probe);
641 idev = __in6_dev_get(dev);
642 if (!idev)
643 goto out;
644 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
645 if (neigh) {
646 if (READ_ONCE(neigh->nud_state) & NUD_VALID)
647 goto out;
648
649 write_lock_bh(&neigh->lock);
650 if (!(neigh->nud_state & NUD_VALID) &&
651 time_after(jiffies,
652 neigh->updated +
653 READ_ONCE(idev->cnf.rtr_probe_interval))) {
654 work = kmalloc(sizeof(*work), GFP_ATOMIC);
655 if (work)
656 __neigh_set_probe_once(neigh);
657 }
658 write_unlock_bh(&neigh->lock);
659 } else if (time_after(jiffies, last_probe +
660 READ_ONCE(idev->cnf.rtr_probe_interval))) {
661 work = kmalloc(sizeof(*work), GFP_ATOMIC);
662 }
663
664 if (!work || cmpxchg(&fib6_nh->last_probe,
665 last_probe, jiffies) != last_probe) {
666 kfree(work);
667 } else {
668 INIT_WORK(&work->work, rt6_probe_deferred);
669 work->target = *nh_gw;
670 netdev_hold(dev, &work->dev_tracker, GFP_ATOMIC);
671 work->dev = dev;
672 schedule_work(&work->work);
673 }
674
675out:
676 rcu_read_unlock();
677}
678#else
679static inline void rt6_probe(struct fib6_nh *fib6_nh)
680{
681}
682#endif
683
684/*
685 * Default Router Selection (RFC 2461 6.3.6)
686 */
687static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
688{
689 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
690 struct neighbour *neigh;
691
692 rcu_read_lock();
693 neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
694 &fib6_nh->fib_nh_gw6);
695 if (neigh) {
696 u8 nud_state = READ_ONCE(neigh->nud_state);
697
698 if (nud_state & NUD_VALID)
699 ret = RT6_NUD_SUCCEED;
700#ifdef CONFIG_IPV6_ROUTER_PREF
701 else if (!(nud_state & NUD_FAILED))
702 ret = RT6_NUD_SUCCEED;
703 else
704 ret = RT6_NUD_FAIL_PROBE;
705#endif
706 } else {
707 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
708 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
709 }
710 rcu_read_unlock();
711
712 return ret;
713}
714
715static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
716 int strict)
717{
718 int m = 0;
719
720 if (!oif || nh->fib_nh_dev->ifindex == oif)
721 m = 2;
722
723 if (!m && (strict & RT6_LOOKUP_F_IFACE))
724 return RT6_NUD_FAIL_HARD;
725#ifdef CONFIG_IPV6_ROUTER_PREF
726 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
727#endif
728 if ((strict & RT6_LOOKUP_F_REACHABLE) &&
729 !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
730 int n = rt6_check_neigh(nh);
731 if (n < 0)
732 return n;
733 }
734 return m;
735}
736
737static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
738 int oif, int strict, int *mpri, bool *do_rr)
739{
740 bool match_do_rr = false;
741 bool rc = false;
742 int m;
743
744 if (nh->fib_nh_flags & RTNH_F_DEAD)
745 goto out;
746
747 if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
748 nh->fib_nh_flags & RTNH_F_LINKDOWN &&
749 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
750 goto out;
751
752 m = rt6_score_route(nh, fib6_flags, oif, strict);
753 if (m == RT6_NUD_FAIL_DO_RR) {
754 match_do_rr = true;
755 m = 0; /* lowest valid score */
756 } else if (m == RT6_NUD_FAIL_HARD) {
757 goto out;
758 }
759
760 if (strict & RT6_LOOKUP_F_REACHABLE)
761 rt6_probe(nh);
762
763 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
764 if (m > *mpri) {
765 *do_rr = match_do_rr;
766 *mpri = m;
767 rc = true;
768 }
769out:
770 return rc;
771}
772
773struct fib6_nh_frl_arg {
774 u32 flags;
775 int oif;
776 int strict;
777 int *mpri;
778 bool *do_rr;
779 struct fib6_nh *nh;
780};
781
782static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
783{
784 struct fib6_nh_frl_arg *arg = _arg;
785
786 arg->nh = nh;
787 return find_match(nh, arg->flags, arg->oif, arg->strict,
788 arg->mpri, arg->do_rr);
789}
790
791static void __find_rr_leaf(struct fib6_info *f6i_start,
792 struct fib6_info *nomatch, u32 metric,
793 struct fib6_result *res, struct fib6_info **cont,
794 int oif, int strict, bool *do_rr, int *mpri)
795{
796 struct fib6_info *f6i;
797
798 for (f6i = f6i_start;
799 f6i && f6i != nomatch;
800 f6i = rcu_dereference(f6i->fib6_next)) {
801 bool matched = false;
802 struct fib6_nh *nh;
803
804 if (cont && f6i->fib6_metric != metric) {
805 *cont = f6i;
806 return;
807 }
808
809 if (fib6_check_expired(f6i))
810 continue;
811
812 if (unlikely(f6i->nh)) {
813 struct fib6_nh_frl_arg arg = {
814 .flags = f6i->fib6_flags,
815 .oif = oif,
816 .strict = strict,
817 .mpri = mpri,
818 .do_rr = do_rr
819 };
820
821 if (nexthop_is_blackhole(f6i->nh)) {
822 res->fib6_flags = RTF_REJECT;
823 res->fib6_type = RTN_BLACKHOLE;
824 res->f6i = f6i;
825 res->nh = nexthop_fib6_nh(f6i->nh);
826 return;
827 }
828 if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
829 &arg)) {
830 matched = true;
831 nh = arg.nh;
832 }
833 } else {
834 nh = f6i->fib6_nh;
835 if (find_match(nh, f6i->fib6_flags, oif, strict,
836 mpri, do_rr))
837 matched = true;
838 }
839 if (matched) {
840 res->f6i = f6i;
841 res->nh = nh;
842 res->fib6_flags = f6i->fib6_flags;
843 res->fib6_type = f6i->fib6_type;
844 }
845 }
846}
847
848static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
849 struct fib6_info *rr_head, int oif, int strict,
850 bool *do_rr, struct fib6_result *res)
851{
852 u32 metric = rr_head->fib6_metric;
853 struct fib6_info *cont = NULL;
854 int mpri = -1;
855
856 __find_rr_leaf(rr_head, NULL, metric, res, &cont,
857 oif, strict, do_rr, &mpri);
858
859 __find_rr_leaf(leaf, rr_head, metric, res, &cont,
860 oif, strict, do_rr, &mpri);
861
862 if (res->f6i || !cont)
863 return;
864
865 __find_rr_leaf(cont, NULL, metric, res, NULL,
866 oif, strict, do_rr, &mpri);
867}
868
869static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
870 struct fib6_result *res, int strict)
871{
872 struct fib6_info *leaf = rcu_dereference(fn->leaf);
873 struct fib6_info *rt0;
874 bool do_rr = false;
875 int key_plen;
876
877 /* make sure this function or its helpers sets f6i */
878 res->f6i = NULL;
879
880 if (!leaf || leaf == net->ipv6.fib6_null_entry)
881 goto out;
882
883 rt0 = rcu_dereference(fn->rr_ptr);
884 if (!rt0)
885 rt0 = leaf;
886
887 /* Double check to make sure fn is not an intermediate node
888 * and fn->leaf does not points to its child's leaf
889 * (This might happen if all routes under fn are deleted from
890 * the tree and fib6_repair_tree() is called on the node.)
891 */
892 key_plen = rt0->fib6_dst.plen;
893#ifdef CONFIG_IPV6_SUBTREES
894 if (rt0->fib6_src.plen)
895 key_plen = rt0->fib6_src.plen;
896#endif
897 if (fn->fn_bit != key_plen)
898 goto out;
899
900 find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
901 if (do_rr) {
902 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
903
904 /* no entries matched; do round-robin */
905 if (!next || next->fib6_metric != rt0->fib6_metric)
906 next = leaf;
907
908 if (next != rt0) {
909 spin_lock_bh(&leaf->fib6_table->tb6_lock);
910 /* make sure next is not being deleted from the tree */
911 if (next->fib6_node)
912 rcu_assign_pointer(fn->rr_ptr, next);
913 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
914 }
915 }
916
917out:
918 if (!res->f6i) {
919 res->f6i = net->ipv6.fib6_null_entry;
920 res->nh = res->f6i->fib6_nh;
921 res->fib6_flags = res->f6i->fib6_flags;
922 res->fib6_type = res->f6i->fib6_type;
923 }
924}
925
926static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
927{
928 return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
929 res->nh->fib_nh_gw_family;
930}
931
932#ifdef CONFIG_IPV6_ROUTE_INFO
933int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
934 const struct in6_addr *gwaddr)
935{
936 struct net *net = dev_net(dev);
937 struct route_info *rinfo = (struct route_info *) opt;
938 struct in6_addr prefix_buf, *prefix;
939 struct fib6_table *table;
940 unsigned int pref;
941 unsigned long lifetime;
942 struct fib6_info *rt;
943
944 if (len < sizeof(struct route_info)) {
945 return -EINVAL;
946 }
947
948 /* Sanity check for prefix_len and length */
949 if (rinfo->length > 3) {
950 return -EINVAL;
951 } else if (rinfo->prefix_len > 128) {
952 return -EINVAL;
953 } else if (rinfo->prefix_len > 64) {
954 if (rinfo->length < 2) {
955 return -EINVAL;
956 }
957 } else if (rinfo->prefix_len > 0) {
958 if (rinfo->length < 1) {
959 return -EINVAL;
960 }
961 }
962
963 pref = rinfo->route_pref;
964 if (pref == ICMPV6_ROUTER_PREF_INVALID)
965 return -EINVAL;
966
967 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
968
969 if (rinfo->length == 3)
970 prefix = (struct in6_addr *)rinfo->prefix;
971 else {
972 /* this function is safe */
973 ipv6_addr_prefix(&prefix_buf,
974 (struct in6_addr *)rinfo->prefix,
975 rinfo->prefix_len);
976 prefix = &prefix_buf;
977 }
978
979 if (rinfo->prefix_len == 0)
980 rt = rt6_get_dflt_router(net, gwaddr, dev);
981 else
982 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
983 gwaddr, dev);
984
985 if (rt && !lifetime) {
986 ip6_del_rt(net, rt, false);
987 rt = NULL;
988 }
989
990 if (!rt && lifetime)
991 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
992 dev, pref);
993 else if (rt)
994 rt->fib6_flags = RTF_ROUTEINFO |
995 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
996
997 if (rt) {
998 table = rt->fib6_table;
999 spin_lock_bh(&table->tb6_lock);
1000
1001 if (!addrconf_finite_timeout(lifetime)) {
1002 fib6_clean_expires(rt);
1003 fib6_remove_gc_list(rt);
1004 } else {
1005 fib6_set_expires(rt, jiffies + HZ * lifetime);
1006 fib6_add_gc_list(rt);
1007 }
1008
1009 spin_unlock_bh(&table->tb6_lock);
1010
1011 fib6_info_release(rt);
1012 }
1013 return 0;
1014}
1015#endif
1016
1017/*
1018 * Misc support functions
1019 */
1020
1021/* called with rcu_lock held */
1022static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
1023{
1024 struct net_device *dev = res->nh->fib_nh_dev;
1025
1026 if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1027 /* for copies of local routes, dst->dev needs to be the
1028 * device if it is a master device, the master device if
1029 * device is enslaved, and the loopback as the default
1030 */
1031 if (netif_is_l3_slave(dev) &&
1032 !rt6_need_strict(&res->f6i->fib6_dst.addr))
1033 dev = l3mdev_master_dev_rcu(dev);
1034 else if (!netif_is_l3_master(dev))
1035 dev = dev_net(dev)->loopback_dev;
1036 /* last case is netif_is_l3_master(dev) is true in which
1037 * case we want dev returned to be dev
1038 */
1039 }
1040
1041 return dev;
1042}
1043
1044static const int fib6_prop[RTN_MAX + 1] = {
1045 [RTN_UNSPEC] = 0,
1046 [RTN_UNICAST] = 0,
1047 [RTN_LOCAL] = 0,
1048 [RTN_BROADCAST] = 0,
1049 [RTN_ANYCAST] = 0,
1050 [RTN_MULTICAST] = 0,
1051 [RTN_BLACKHOLE] = -EINVAL,
1052 [RTN_UNREACHABLE] = -EHOSTUNREACH,
1053 [RTN_PROHIBIT] = -EACCES,
1054 [RTN_THROW] = -EAGAIN,
1055 [RTN_NAT] = -EINVAL,
1056 [RTN_XRESOLVE] = -EINVAL,
1057};
1058
1059static int ip6_rt_type_to_error(u8 fib6_type)
1060{
1061 return fib6_prop[fib6_type];
1062}
1063
1064static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
1065{
1066 unsigned short flags = 0;
1067
1068 if (rt->dst_nocount)
1069 flags |= DST_NOCOUNT;
1070 if (rt->dst_nopolicy)
1071 flags |= DST_NOPOLICY;
1072
1073 return flags;
1074}
1075
1076static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
1077{
1078 rt->dst.error = ip6_rt_type_to_error(fib6_type);
1079
1080 switch (fib6_type) {
1081 case RTN_BLACKHOLE:
1082 rt->dst.output = dst_discard_out;
1083 rt->dst.input = dst_discard;
1084 break;
1085 case RTN_PROHIBIT:
1086 rt->dst.output = ip6_pkt_prohibit_out;
1087 rt->dst.input = ip6_pkt_prohibit;
1088 break;
1089 case RTN_THROW:
1090 case RTN_UNREACHABLE:
1091 default:
1092 rt->dst.output = ip6_pkt_discard_out;
1093 rt->dst.input = ip6_pkt_discard;
1094 break;
1095 }
1096}
1097
1098static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
1099{
1100 struct fib6_info *f6i = res->f6i;
1101
1102 if (res->fib6_flags & RTF_REJECT) {
1103 ip6_rt_init_dst_reject(rt, res->fib6_type);
1104 return;
1105 }
1106
1107 rt->dst.error = 0;
1108 rt->dst.output = ip6_output;
1109
1110 if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
1111 rt->dst.input = ip6_input;
1112 } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
1113 rt->dst.input = ip6_mc_input;
1114 } else {
1115 rt->dst.input = ip6_forward;
1116 }
1117
1118 if (res->nh->fib_nh_lws) {
1119 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
1120 lwtunnel_set_redirect(&rt->dst);
1121 }
1122
1123 rt->dst.lastuse = jiffies;
1124}
1125
1126/* Caller must already hold reference to @from */
1127static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
1128{
1129 rt->rt6i_flags &= ~RTF_EXPIRES;
1130 rcu_assign_pointer(rt->from, from);
1131 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
1132}
1133
1134/* Caller must already hold reference to f6i in result */
1135static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
1136{
1137 const struct fib6_nh *nh = res->nh;
1138 const struct net_device *dev = nh->fib_nh_dev;
1139 struct fib6_info *f6i = res->f6i;
1140
1141 ip6_rt_init_dst(rt, res);
1142
1143 rt->rt6i_dst = f6i->fib6_dst;
1144 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
1145 rt->rt6i_flags = res->fib6_flags;
1146 if (nh->fib_nh_gw_family) {
1147 rt->rt6i_gateway = nh->fib_nh_gw6;
1148 rt->rt6i_flags |= RTF_GATEWAY;
1149 }
1150 rt6_set_from(rt, f6i);
1151#ifdef CONFIG_IPV6_SUBTREES
1152 rt->rt6i_src = f6i->fib6_src;
1153#endif
1154}
1155
1156static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1157 struct in6_addr *saddr)
1158{
1159 struct fib6_node *pn, *sn;
1160 while (1) {
1161 if (fn->fn_flags & RTN_TL_ROOT)
1162 return NULL;
1163 pn = rcu_dereference(fn->parent);
1164 sn = FIB6_SUBTREE(pn);
1165 if (sn && sn != fn)
1166 fn = fib6_node_lookup(sn, NULL, saddr);
1167 else
1168 fn = pn;
1169 if (fn->fn_flags & RTN_RTINFO)
1170 return fn;
1171 }
1172}
1173
1174static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
1175{
1176 struct rt6_info *rt = *prt;
1177
1178 if (dst_hold_safe(&rt->dst))
1179 return true;
1180 if (net) {
1181 rt = net->ipv6.ip6_null_entry;
1182 dst_hold(&rt->dst);
1183 } else {
1184 rt = NULL;
1185 }
1186 *prt = rt;
1187 return false;
1188}
1189
1190/* called with rcu_lock held */
1191static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
1192{
1193 struct net_device *dev = res->nh->fib_nh_dev;
1194 struct fib6_info *f6i = res->f6i;
1195 unsigned short flags;
1196 struct rt6_info *nrt;
1197
1198 if (!fib6_info_hold_safe(f6i))
1199 goto fallback;
1200
1201 flags = fib6_info_dst_flags(f6i);
1202 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1203 if (!nrt) {
1204 fib6_info_release(f6i);
1205 goto fallback;
1206 }
1207
1208 ip6_rt_copy_init(nrt, res);
1209 return nrt;
1210
1211fallback:
1212 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1213 dst_hold(&nrt->dst);
1214 return nrt;
1215}
1216
1217INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_lookup(struct net *net,
1218 struct fib6_table *table,
1219 struct flowi6 *fl6,
1220 const struct sk_buff *skb,
1221 int flags)
1222{
1223 struct fib6_result res = {};
1224 struct fib6_node *fn;
1225 struct rt6_info *rt;
1226
1227 rcu_read_lock();
1228 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1229restart:
1230 res.f6i = rcu_dereference(fn->leaf);
1231 if (!res.f6i)
1232 res.f6i = net->ipv6.fib6_null_entry;
1233 else
1234 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1235 flags);
1236
1237 if (res.f6i == net->ipv6.fib6_null_entry) {
1238 fn = fib6_backtrack(fn, &fl6->saddr);
1239 if (fn)
1240 goto restart;
1241
1242 rt = net->ipv6.ip6_null_entry;
1243 dst_hold(&rt->dst);
1244 goto out;
1245 } else if (res.fib6_flags & RTF_REJECT) {
1246 goto do_create;
1247 }
1248
1249 fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1250 fl6->flowi6_oif != 0, skb, flags);
1251
1252 /* Search through exception table */
1253 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
1254 if (rt) {
1255 if (ip6_hold_safe(net, &rt))
1256 dst_use_noref(&rt->dst, jiffies);
1257 } else {
1258do_create:
1259 rt = ip6_create_rt_rcu(&res);
1260 }
1261
1262out:
1263 trace_fib6_table_lookup(net, &res, table, fl6);
1264
1265 rcu_read_unlock();
1266
1267 return rt;
1268}
1269
1270struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1271 const struct sk_buff *skb, int flags)
1272{
1273 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1274}
1275EXPORT_SYMBOL_GPL(ip6_route_lookup);
1276
1277struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1278 const struct in6_addr *saddr, int oif,
1279 const struct sk_buff *skb, int strict)
1280{
1281 struct flowi6 fl6 = {
1282 .flowi6_oif = oif,
1283 .daddr = *daddr,
1284 };
1285 struct dst_entry *dst;
1286 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1287
1288 if (saddr) {
1289 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1290 flags |= RT6_LOOKUP_F_HAS_SADDR;
1291 }
1292
1293 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1294 if (dst->error == 0)
1295 return dst_rt6_info(dst);
1296
1297 dst_release(dst);
1298
1299 return NULL;
1300}
1301EXPORT_SYMBOL(rt6_lookup);
1302
1303/* ip6_ins_rt is called with FREE table->tb6_lock.
1304 * It takes new route entry, the addition fails by any reason the
1305 * route is released.
1306 * Caller must hold dst before calling it.
1307 */
1308
1309static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1310 struct netlink_ext_ack *extack)
1311{
1312 int err;
1313 struct fib6_table *table;
1314
1315 table = rt->fib6_table;
1316 spin_lock_bh(&table->tb6_lock);
1317 err = fib6_add(&table->tb6_root, rt, info, extack);
1318 spin_unlock_bh(&table->tb6_lock);
1319
1320 return err;
1321}
1322
1323int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1324{
1325 struct nl_info info = { .nl_net = net, };
1326
1327 return __ip6_ins_rt(rt, &info, NULL);
1328}
1329
1330static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
1331 const struct in6_addr *daddr,
1332 const struct in6_addr *saddr)
1333{
1334 struct fib6_info *f6i = res->f6i;
1335 struct net_device *dev;
1336 struct rt6_info *rt;
1337
1338 /*
1339 * Clone the route.
1340 */
1341
1342 if (!fib6_info_hold_safe(f6i))
1343 return NULL;
1344
1345 dev = ip6_rt_get_dev_rcu(res);
1346 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1347 if (!rt) {
1348 fib6_info_release(f6i);
1349 return NULL;
1350 }
1351
1352 ip6_rt_copy_init(rt, res);
1353 rt->rt6i_flags |= RTF_CACHE;
1354 rt->rt6i_dst.addr = *daddr;
1355 rt->rt6i_dst.plen = 128;
1356
1357 if (!rt6_is_gw_or_nonexthop(res)) {
1358 if (f6i->fib6_dst.plen != 128 &&
1359 ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
1360 rt->rt6i_flags |= RTF_ANYCAST;
1361#ifdef CONFIG_IPV6_SUBTREES
1362 if (rt->rt6i_src.plen && saddr) {
1363 rt->rt6i_src.addr = *saddr;
1364 rt->rt6i_src.plen = 128;
1365 }
1366#endif
1367 }
1368
1369 return rt;
1370}
1371
1372static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
1373{
1374 struct fib6_info *f6i = res->f6i;
1375 unsigned short flags = fib6_info_dst_flags(f6i);
1376 struct net_device *dev;
1377 struct rt6_info *pcpu_rt;
1378
1379 if (!fib6_info_hold_safe(f6i))
1380 return NULL;
1381
1382 rcu_read_lock();
1383 dev = ip6_rt_get_dev_rcu(res);
1384 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT);
1385 rcu_read_unlock();
1386 if (!pcpu_rt) {
1387 fib6_info_release(f6i);
1388 return NULL;
1389 }
1390 ip6_rt_copy_init(pcpu_rt, res);
1391 pcpu_rt->rt6i_flags |= RTF_PCPU;
1392
1393 if (f6i->nh)
1394 pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1395
1396 return pcpu_rt;
1397}
1398
1399static bool rt6_is_valid(const struct rt6_info *rt6)
1400{
1401 return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1402}
1403
1404/* It should be called with rcu_read_lock() acquired */
1405static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
1406{
1407 struct rt6_info *pcpu_rt;
1408
1409 pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
1410
1411 if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1412 struct rt6_info *prev, **p;
1413
1414 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1415 /* Paired with READ_ONCE() in __fib6_drop_pcpu_from() */
1416 prev = xchg(p, NULL);
1417 if (prev) {
1418 dst_dev_put(&prev->dst);
1419 dst_release(&prev->dst);
1420 }
1421
1422 pcpu_rt = NULL;
1423 }
1424
1425 return pcpu_rt;
1426}
1427
1428static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1429 const struct fib6_result *res)
1430{
1431 struct rt6_info *pcpu_rt, *prev, **p;
1432
1433 pcpu_rt = ip6_rt_pcpu_alloc(res);
1434 if (!pcpu_rt)
1435 return NULL;
1436
1437 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1438 prev = cmpxchg(p, NULL, pcpu_rt);
1439 BUG_ON(prev);
1440
1441 if (res->f6i->fib6_destroying) {
1442 struct fib6_info *from;
1443
1444 from = unrcu_pointer(xchg(&pcpu_rt->from, NULL));
1445 fib6_info_release(from);
1446 }
1447
1448 return pcpu_rt;
1449}
1450
1451/* exception hash table implementation
1452 */
1453static DEFINE_SPINLOCK(rt6_exception_lock);
1454
1455/* Remove rt6_ex from hash table and free the memory
1456 * Caller must hold rt6_exception_lock
1457 */
1458static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1459 struct rt6_exception *rt6_ex)
1460{
1461 struct net *net;
1462
1463 if (!bucket || !rt6_ex)
1464 return;
1465
1466 net = dev_net(rt6_ex->rt6i->dst.dev);
1467 net->ipv6.rt6_stats->fib_rt_cache--;
1468
1469 /* purge completely the exception to allow releasing the held resources:
1470 * some [sk] cache may keep the dst around for unlimited time
1471 */
1472 dst_dev_put(&rt6_ex->rt6i->dst);
1473
1474 hlist_del_rcu(&rt6_ex->hlist);
1475 dst_release(&rt6_ex->rt6i->dst);
1476 kfree_rcu(rt6_ex, rcu);
1477 WARN_ON_ONCE(!bucket->depth);
1478 bucket->depth--;
1479}
1480
1481/* Remove oldest rt6_ex in bucket and free the memory
1482 * Caller must hold rt6_exception_lock
1483 */
1484static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1485{
1486 struct rt6_exception *rt6_ex, *oldest = NULL;
1487
1488 if (!bucket)
1489 return;
1490
1491 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1492 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1493 oldest = rt6_ex;
1494 }
1495 rt6_remove_exception(bucket, oldest);
1496}
1497
1498static u32 rt6_exception_hash(const struct in6_addr *dst,
1499 const struct in6_addr *src)
1500{
1501 static siphash_aligned_key_t rt6_exception_key;
1502 struct {
1503 struct in6_addr dst;
1504 struct in6_addr src;
1505 } __aligned(SIPHASH_ALIGNMENT) combined = {
1506 .dst = *dst,
1507 };
1508 u64 val;
1509
1510 net_get_random_once(&rt6_exception_key, sizeof(rt6_exception_key));
1511
1512#ifdef CONFIG_IPV6_SUBTREES
1513 if (src)
1514 combined.src = *src;
1515#endif
1516 val = siphash(&combined, sizeof(combined), &rt6_exception_key);
1517
1518 return hash_64(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1519}
1520
1521/* Helper function to find the cached rt in the hash table
1522 * and update bucket pointer to point to the bucket for this
1523 * (daddr, saddr) pair
1524 * Caller must hold rt6_exception_lock
1525 */
1526static struct rt6_exception *
1527__rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1528 const struct in6_addr *daddr,
1529 const struct in6_addr *saddr)
1530{
1531 struct rt6_exception *rt6_ex;
1532 u32 hval;
1533
1534 if (!(*bucket) || !daddr)
1535 return NULL;
1536
1537 hval = rt6_exception_hash(daddr, saddr);
1538 *bucket += hval;
1539
1540 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1541 struct rt6_info *rt6 = rt6_ex->rt6i;
1542 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1543
1544#ifdef CONFIG_IPV6_SUBTREES
1545 if (matched && saddr)
1546 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1547#endif
1548 if (matched)
1549 return rt6_ex;
1550 }
1551 return NULL;
1552}
1553
1554/* Helper function to find the cached rt in the hash table
1555 * and update bucket pointer to point to the bucket for this
1556 * (daddr, saddr) pair
1557 * Caller must hold rcu_read_lock()
1558 */
1559static struct rt6_exception *
1560__rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1561 const struct in6_addr *daddr,
1562 const struct in6_addr *saddr)
1563{
1564 struct rt6_exception *rt6_ex;
1565 u32 hval;
1566
1567 WARN_ON_ONCE(!rcu_read_lock_held());
1568
1569 if (!(*bucket) || !daddr)
1570 return NULL;
1571
1572 hval = rt6_exception_hash(daddr, saddr);
1573 *bucket += hval;
1574
1575 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1576 struct rt6_info *rt6 = rt6_ex->rt6i;
1577 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1578
1579#ifdef CONFIG_IPV6_SUBTREES
1580 if (matched && saddr)
1581 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1582#endif
1583 if (matched)
1584 return rt6_ex;
1585 }
1586 return NULL;
1587}
1588
1589static unsigned int fib6_mtu(const struct fib6_result *res)
1590{
1591 const struct fib6_nh *nh = res->nh;
1592 unsigned int mtu;
1593
1594 if (res->f6i->fib6_pmtu) {
1595 mtu = res->f6i->fib6_pmtu;
1596 } else {
1597 struct net_device *dev = nh->fib_nh_dev;
1598 struct inet6_dev *idev;
1599
1600 rcu_read_lock();
1601 idev = __in6_dev_get(dev);
1602 mtu = READ_ONCE(idev->cnf.mtu6);
1603 rcu_read_unlock();
1604 }
1605
1606 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1607
1608 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
1609}
1610
1611#define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1612
1613/* used when the flushed bit is not relevant, only access to the bucket
1614 * (ie., all bucket users except rt6_insert_exception);
1615 *
1616 * called under rcu lock; sometimes called with rt6_exception_lock held
1617 */
1618static
1619struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1620 spinlock_t *lock)
1621{
1622 struct rt6_exception_bucket *bucket;
1623
1624 if (lock)
1625 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1626 lockdep_is_held(lock));
1627 else
1628 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1629
1630 /* remove bucket flushed bit if set */
1631 if (bucket) {
1632 unsigned long p = (unsigned long)bucket;
1633
1634 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1635 bucket = (struct rt6_exception_bucket *)p;
1636 }
1637
1638 return bucket;
1639}
1640
1641static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1642{
1643 unsigned long p = (unsigned long)bucket;
1644
1645 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1646}
1647
1648/* called with rt6_exception_lock held */
1649static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1650 spinlock_t *lock)
1651{
1652 struct rt6_exception_bucket *bucket;
1653 unsigned long p;
1654
1655 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1656 lockdep_is_held(lock));
1657
1658 p = (unsigned long)bucket;
1659 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1660 bucket = (struct rt6_exception_bucket *)p;
1661 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1662}
1663
1664static int rt6_insert_exception(struct rt6_info *nrt,
1665 const struct fib6_result *res)
1666{
1667 struct net *net = dev_net(nrt->dst.dev);
1668 struct rt6_exception_bucket *bucket;
1669 struct fib6_info *f6i = res->f6i;
1670 struct in6_addr *src_key = NULL;
1671 struct rt6_exception *rt6_ex;
1672 struct fib6_nh *nh = res->nh;
1673 int max_depth;
1674 int err = 0;
1675
1676 spin_lock_bh(&rt6_exception_lock);
1677
1678 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1679 lockdep_is_held(&rt6_exception_lock));
1680 if (!bucket) {
1681 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1682 GFP_ATOMIC);
1683 if (!bucket) {
1684 err = -ENOMEM;
1685 goto out;
1686 }
1687 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1688 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1689 err = -EINVAL;
1690 goto out;
1691 }
1692
1693#ifdef CONFIG_IPV6_SUBTREES
1694 /* fib6_src.plen != 0 indicates f6i is in subtree
1695 * and exception table is indexed by a hash of
1696 * both fib6_dst and fib6_src.
1697 * Otherwise, the exception table is indexed by
1698 * a hash of only fib6_dst.
1699 */
1700 if (f6i->fib6_src.plen)
1701 src_key = &nrt->rt6i_src.addr;
1702#endif
1703 /* rt6_mtu_change() might lower mtu on f6i.
1704 * Only insert this exception route if its mtu
1705 * is less than f6i's mtu value.
1706 */
1707 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
1708 err = -EINVAL;
1709 goto out;
1710 }
1711
1712 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1713 src_key);
1714 if (rt6_ex)
1715 rt6_remove_exception(bucket, rt6_ex);
1716
1717 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1718 if (!rt6_ex) {
1719 err = -ENOMEM;
1720 goto out;
1721 }
1722 rt6_ex->rt6i = nrt;
1723 rt6_ex->stamp = jiffies;
1724 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1725 bucket->depth++;
1726 net->ipv6.rt6_stats->fib_rt_cache++;
1727
1728 /* Randomize max depth to avoid some side channels attacks. */
1729 max_depth = FIB6_MAX_DEPTH + get_random_u32_below(FIB6_MAX_DEPTH);
1730 while (bucket->depth > max_depth)
1731 rt6_exception_remove_oldest(bucket);
1732
1733out:
1734 spin_unlock_bh(&rt6_exception_lock);
1735
1736 /* Update fn->fn_sernum to invalidate all cached dst */
1737 if (!err) {
1738 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1739 fib6_update_sernum(net, f6i);
1740 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
1741 fib6_force_start_gc(net);
1742 }
1743
1744 return err;
1745}
1746
1747static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
1748{
1749 struct rt6_exception_bucket *bucket;
1750 struct rt6_exception *rt6_ex;
1751 struct hlist_node *tmp;
1752 int i;
1753
1754 spin_lock_bh(&rt6_exception_lock);
1755
1756 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1757 if (!bucket)
1758 goto out;
1759
1760 /* Prevent rt6_insert_exception() to recreate the bucket list */
1761 if (!from)
1762 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1763
1764 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1765 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1766 if (!from ||
1767 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1768 rt6_remove_exception(bucket, rt6_ex);
1769 }
1770 WARN_ON_ONCE(!from && bucket->depth);
1771 bucket++;
1772 }
1773out:
1774 spin_unlock_bh(&rt6_exception_lock);
1775}
1776
1777static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1778{
1779 struct fib6_info *f6i = arg;
1780
1781 fib6_nh_flush_exceptions(nh, f6i);
1782
1783 return 0;
1784}
1785
1786void rt6_flush_exceptions(struct fib6_info *f6i)
1787{
1788 if (f6i->nh)
1789 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1790 f6i);
1791 else
1792 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
1793}
1794
1795/* Find cached rt in the hash table inside passed in rt
1796 * Caller has to hold rcu_read_lock()
1797 */
1798static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
1799 const struct in6_addr *daddr,
1800 const struct in6_addr *saddr)
1801{
1802 const struct in6_addr *src_key = NULL;
1803 struct rt6_exception_bucket *bucket;
1804 struct rt6_exception *rt6_ex;
1805 struct rt6_info *ret = NULL;
1806
1807#ifdef CONFIG_IPV6_SUBTREES
1808 /* fib6i_src.plen != 0 indicates f6i is in subtree
1809 * and exception table is indexed by a hash of
1810 * both fib6_dst and fib6_src.
1811 * However, the src addr used to create the hash
1812 * might not be exactly the passed in saddr which
1813 * is a /128 addr from the flow.
1814 * So we need to use f6i->fib6_src to redo lookup
1815 * if the passed in saddr does not find anything.
1816 * (See the logic in ip6_rt_cache_alloc() on how
1817 * rt->rt6i_src is updated.)
1818 */
1819 if (res->f6i->fib6_src.plen)
1820 src_key = saddr;
1821find_ex:
1822#endif
1823 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
1824 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1825
1826 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1827 ret = rt6_ex->rt6i;
1828
1829#ifdef CONFIG_IPV6_SUBTREES
1830 /* Use fib6_src as src_key and redo lookup */
1831 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1832 src_key = &res->f6i->fib6_src.addr;
1833 goto find_ex;
1834 }
1835#endif
1836
1837 return ret;
1838}
1839
1840/* Remove the passed in cached rt from the hash table that contains it */
1841static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
1842 const struct rt6_info *rt)
1843{
1844 const struct in6_addr *src_key = NULL;
1845 struct rt6_exception_bucket *bucket;
1846 struct rt6_exception *rt6_ex;
1847 int err;
1848
1849 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
1850 return -ENOENT;
1851
1852 spin_lock_bh(&rt6_exception_lock);
1853 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1854
1855#ifdef CONFIG_IPV6_SUBTREES
1856 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1857 * and exception table is indexed by a hash of
1858 * both rt6i_dst and rt6i_src.
1859 * Otherwise, the exception table is indexed by
1860 * a hash of only rt6i_dst.
1861 */
1862 if (plen)
1863 src_key = &rt->rt6i_src.addr;
1864#endif
1865 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1866 &rt->rt6i_dst.addr,
1867 src_key);
1868 if (rt6_ex) {
1869 rt6_remove_exception(bucket, rt6_ex);
1870 err = 0;
1871 } else {
1872 err = -ENOENT;
1873 }
1874
1875 spin_unlock_bh(&rt6_exception_lock);
1876 return err;
1877}
1878
1879struct fib6_nh_excptn_arg {
1880 struct rt6_info *rt;
1881 int plen;
1882};
1883
1884static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1885{
1886 struct fib6_nh_excptn_arg *arg = _arg;
1887 int err;
1888
1889 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1890 if (err == 0)
1891 return 1;
1892
1893 return 0;
1894}
1895
1896static int rt6_remove_exception_rt(struct rt6_info *rt)
1897{
1898 struct fib6_info *from;
1899
1900 from = rcu_dereference(rt->from);
1901 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1902 return -EINVAL;
1903
1904 if (from->nh) {
1905 struct fib6_nh_excptn_arg arg = {
1906 .rt = rt,
1907 .plen = from->fib6_src.plen
1908 };
1909 int rc;
1910
1911 /* rc = 1 means an entry was found */
1912 rc = nexthop_for_each_fib6_nh(from->nh,
1913 rt6_nh_remove_exception_rt,
1914 &arg);
1915 return rc ? 0 : -ENOENT;
1916 }
1917
1918 return fib6_nh_remove_exception(from->fib6_nh,
1919 from->fib6_src.plen, rt);
1920}
1921
1922/* Find rt6_ex which contains the passed in rt cache and
1923 * refresh its stamp
1924 */
1925static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
1926 const struct rt6_info *rt)
1927{
1928 const struct in6_addr *src_key = NULL;
1929 struct rt6_exception_bucket *bucket;
1930 struct rt6_exception *rt6_ex;
1931
1932 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
1933#ifdef CONFIG_IPV6_SUBTREES
1934 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1935 * and exception table is indexed by a hash of
1936 * both rt6i_dst and rt6i_src.
1937 * Otherwise, the exception table is indexed by
1938 * a hash of only rt6i_dst.
1939 */
1940 if (plen)
1941 src_key = &rt->rt6i_src.addr;
1942#endif
1943 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
1944 if (rt6_ex)
1945 rt6_ex->stamp = jiffies;
1946}
1947
1948struct fib6_nh_match_arg {
1949 const struct net_device *dev;
1950 const struct in6_addr *gw;
1951 struct fib6_nh *match;
1952};
1953
1954/* determine if fib6_nh has given device and gateway */
1955static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1956{
1957 struct fib6_nh_match_arg *arg = _arg;
1958
1959 if (arg->dev != nh->fib_nh_dev ||
1960 (arg->gw && !nh->fib_nh_gw_family) ||
1961 (!arg->gw && nh->fib_nh_gw_family) ||
1962 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1963 return 0;
1964
1965 arg->match = nh;
1966
1967 /* found a match, break the loop */
1968 return 1;
1969}
1970
1971static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1972{
1973 struct fib6_info *from;
1974 struct fib6_nh *fib6_nh;
1975
1976 rcu_read_lock();
1977
1978 from = rcu_dereference(rt->from);
1979 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1980 goto unlock;
1981
1982 if (from->nh) {
1983 struct fib6_nh_match_arg arg = {
1984 .dev = rt->dst.dev,
1985 .gw = &rt->rt6i_gateway,
1986 };
1987
1988 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1989
1990 if (!arg.match)
1991 goto unlock;
1992 fib6_nh = arg.match;
1993 } else {
1994 fib6_nh = from->fib6_nh;
1995 }
1996 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
1997unlock:
1998 rcu_read_unlock();
1999}
2000
2001static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
2002 struct rt6_info *rt, int mtu)
2003{
2004 /* If the new MTU is lower than the route PMTU, this new MTU will be the
2005 * lowest MTU in the path: always allow updating the route PMTU to
2006 * reflect PMTU decreases.
2007 *
2008 * If the new MTU is higher, and the route PMTU is equal to the local
2009 * MTU, this means the old MTU is the lowest in the path, so allow
2010 * updating it: if other nodes now have lower MTUs, PMTU discovery will
2011 * handle this.
2012 */
2013
2014 if (dst_mtu(&rt->dst) >= mtu)
2015 return true;
2016
2017 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
2018 return true;
2019
2020 return false;
2021}
2022
2023static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
2024 const struct fib6_nh *nh, int mtu)
2025{
2026 struct rt6_exception_bucket *bucket;
2027 struct rt6_exception *rt6_ex;
2028 int i;
2029
2030 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2031 if (!bucket)
2032 return;
2033
2034 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2035 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2036 struct rt6_info *entry = rt6_ex->rt6i;
2037
2038 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
2039 * route), the metrics of its rt->from have already
2040 * been updated.
2041 */
2042 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
2043 rt6_mtu_change_route_allowed(idev, entry, mtu))
2044 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
2045 }
2046 bucket++;
2047 }
2048}
2049
2050#define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2051
2052static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2053 const struct in6_addr *gateway)
2054{
2055 struct rt6_exception_bucket *bucket;
2056 struct rt6_exception *rt6_ex;
2057 struct hlist_node *tmp;
2058 int i;
2059
2060 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2061 return;
2062
2063 spin_lock_bh(&rt6_exception_lock);
2064 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2065 if (bucket) {
2066 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2067 hlist_for_each_entry_safe(rt6_ex, tmp,
2068 &bucket->chain, hlist) {
2069 struct rt6_info *entry = rt6_ex->rt6i;
2070
2071 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2072 RTF_CACHE_GATEWAY &&
2073 ipv6_addr_equal(gateway,
2074 &entry->rt6i_gateway)) {
2075 rt6_remove_exception(bucket, rt6_ex);
2076 }
2077 }
2078 bucket++;
2079 }
2080 }
2081
2082 spin_unlock_bh(&rt6_exception_lock);
2083}
2084
2085static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2086 struct rt6_exception *rt6_ex,
2087 struct fib6_gc_args *gc_args,
2088 unsigned long now)
2089{
2090 struct rt6_info *rt = rt6_ex->rt6i;
2091
2092 /* we are pruning and obsoleting aged-out and non gateway exceptions
2093 * even if others have still references to them, so that on next
2094 * dst_check() such references can be dropped.
2095 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2096 * expired, independently from their aging, as per RFC 8201 section 4
2097 */
2098 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2099 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2100 pr_debug("aging clone %p\n", rt);
2101 rt6_remove_exception(bucket, rt6_ex);
2102 return;
2103 }
2104 } else if (time_after(jiffies, rt->dst.expires)) {
2105 pr_debug("purging expired route %p\n", rt);
2106 rt6_remove_exception(bucket, rt6_ex);
2107 return;
2108 }
2109
2110 if (rt->rt6i_flags & RTF_GATEWAY) {
2111 struct neighbour *neigh;
2112
2113 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2114
2115 if (!(neigh && (neigh->flags & NTF_ROUTER))) {
2116 pr_debug("purging route %p via non-router but gateway\n",
2117 rt);
2118 rt6_remove_exception(bucket, rt6_ex);
2119 return;
2120 }
2121 }
2122
2123 gc_args->more++;
2124}
2125
2126static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
2127 struct fib6_gc_args *gc_args,
2128 unsigned long now)
2129{
2130 struct rt6_exception_bucket *bucket;
2131 struct rt6_exception *rt6_ex;
2132 struct hlist_node *tmp;
2133 int i;
2134
2135 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2136 return;
2137
2138 rcu_read_lock_bh();
2139 spin_lock(&rt6_exception_lock);
2140 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2141 if (bucket) {
2142 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2143 hlist_for_each_entry_safe(rt6_ex, tmp,
2144 &bucket->chain, hlist) {
2145 rt6_age_examine_exception(bucket, rt6_ex,
2146 gc_args, now);
2147 }
2148 bucket++;
2149 }
2150 }
2151 spin_unlock(&rt6_exception_lock);
2152 rcu_read_unlock_bh();
2153}
2154
2155struct fib6_nh_age_excptn_arg {
2156 struct fib6_gc_args *gc_args;
2157 unsigned long now;
2158};
2159
2160static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2161{
2162 struct fib6_nh_age_excptn_arg *arg = _arg;
2163
2164 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2165 return 0;
2166}
2167
2168void rt6_age_exceptions(struct fib6_info *f6i,
2169 struct fib6_gc_args *gc_args,
2170 unsigned long now)
2171{
2172 if (f6i->nh) {
2173 struct fib6_nh_age_excptn_arg arg = {
2174 .gc_args = gc_args,
2175 .now = now
2176 };
2177
2178 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2179 &arg);
2180 } else {
2181 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2182 }
2183}
2184
2185/* must be called with rcu lock held */
2186int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2187 struct flowi6 *fl6, struct fib6_result *res, int strict)
2188{
2189 struct fib6_node *fn, *saved_fn;
2190
2191 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2192 saved_fn = fn;
2193
2194redo_rt6_select:
2195 rt6_select(net, fn, oif, res, strict);
2196 if (res->f6i == net->ipv6.fib6_null_entry) {
2197 fn = fib6_backtrack(fn, &fl6->saddr);
2198 if (fn)
2199 goto redo_rt6_select;
2200 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2201 /* also consider unreachable route */
2202 strict &= ~RT6_LOOKUP_F_REACHABLE;
2203 fn = saved_fn;
2204 goto redo_rt6_select;
2205 }
2206 }
2207
2208 trace_fib6_table_lookup(net, res, table, fl6);
2209
2210 return 0;
2211}
2212
2213struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2214 int oif, struct flowi6 *fl6,
2215 const struct sk_buff *skb, int flags)
2216{
2217 struct fib6_result res = {};
2218 struct rt6_info *rt = NULL;
2219 int strict = 0;
2220
2221 WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2222 !rcu_read_lock_held());
2223
2224 strict |= flags & RT6_LOOKUP_F_IFACE;
2225 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2226 if (READ_ONCE(net->ipv6.devconf_all->forwarding) == 0)
2227 strict |= RT6_LOOKUP_F_REACHABLE;
2228
2229 rcu_read_lock();
2230
2231 fib6_table_lookup(net, table, oif, fl6, &res, strict);
2232 if (res.f6i == net->ipv6.fib6_null_entry)
2233 goto out;
2234
2235 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
2236
2237 /*Search through exception table */
2238 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
2239 if (rt) {
2240 goto out;
2241 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
2242 !res.nh->fib_nh_gw_family)) {
2243 /* Create a RTF_CACHE clone which will not be
2244 * owned by the fib6 tree. It is for the special case where
2245 * the daddr in the skb during the neighbor look-up is different
2246 * from the fl6->daddr used to look-up route here.
2247 */
2248 rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
2249
2250 if (rt) {
2251 /* 1 refcnt is taken during ip6_rt_cache_alloc().
2252 * As rt6_uncached_list_add() does not consume refcnt,
2253 * this refcnt is always returned to the caller even
2254 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
2255 */
2256 rt6_uncached_list_add(rt);
2257 rcu_read_unlock();
2258
2259 return rt;
2260 }
2261 } else {
2262 /* Get a percpu copy */
2263 local_bh_disable();
2264 rt = rt6_get_pcpu_route(&res);
2265
2266 if (!rt)
2267 rt = rt6_make_pcpu_route(net, &res);
2268
2269 local_bh_enable();
2270 }
2271out:
2272 if (!rt)
2273 rt = net->ipv6.ip6_null_entry;
2274 if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2275 ip6_hold_safe(net, &rt);
2276 rcu_read_unlock();
2277
2278 return rt;
2279}
2280EXPORT_SYMBOL_GPL(ip6_pol_route);
2281
2282INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_input(struct net *net,
2283 struct fib6_table *table,
2284 struct flowi6 *fl6,
2285 const struct sk_buff *skb,
2286 int flags)
2287{
2288 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
2289}
2290
2291struct dst_entry *ip6_route_input_lookup(struct net *net,
2292 struct net_device *dev,
2293 struct flowi6 *fl6,
2294 const struct sk_buff *skb,
2295 int flags)
2296{
2297 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2298 flags |= RT6_LOOKUP_F_IFACE;
2299
2300 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
2301}
2302EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
2303
2304static void ip6_multipath_l3_keys(const struct sk_buff *skb,
2305 struct flow_keys *keys,
2306 struct flow_keys *flkeys)
2307{
2308 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2309 const struct ipv6hdr *key_iph = outer_iph;
2310 struct flow_keys *_flkeys = flkeys;
2311 const struct ipv6hdr *inner_iph;
2312 const struct icmp6hdr *icmph;
2313 struct ipv6hdr _inner_iph;
2314 struct icmp6hdr _icmph;
2315
2316 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2317 goto out;
2318
2319 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2320 sizeof(_icmph), &_icmph);
2321 if (!icmph)
2322 goto out;
2323
2324 if (!icmpv6_is_err(icmph->icmp6_type))
2325 goto out;
2326
2327 inner_iph = skb_header_pointer(skb,
2328 skb_transport_offset(skb) + sizeof(*icmph),
2329 sizeof(_inner_iph), &_inner_iph);
2330 if (!inner_iph)
2331 goto out;
2332
2333 key_iph = inner_iph;
2334 _flkeys = NULL;
2335out:
2336 if (_flkeys) {
2337 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2338 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2339 keys->tags.flow_label = _flkeys->tags.flow_label;
2340 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2341 } else {
2342 keys->addrs.v6addrs.src = key_iph->saddr;
2343 keys->addrs.v6addrs.dst = key_iph->daddr;
2344 keys->tags.flow_label = ip6_flowlabel(key_iph);
2345 keys->basic.ip_proto = key_iph->nexthdr;
2346 }
2347}
2348
2349static u32 rt6_multipath_custom_hash_outer(const struct net *net,
2350 const struct sk_buff *skb,
2351 bool *p_has_inner)
2352{
2353 u32 hash_fields = ip6_multipath_hash_fields(net);
2354 struct flow_keys keys, hash_keys;
2355
2356 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2357 return 0;
2358
2359 memset(&hash_keys, 0, sizeof(hash_keys));
2360 skb_flow_dissect_flow_keys(skb, &keys, FLOW_DISSECTOR_F_STOP_AT_ENCAP);
2361
2362 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2363 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2364 hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2365 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2366 hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2367 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2368 hash_keys.basic.ip_proto = keys.basic.ip_proto;
2369 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2370 hash_keys.tags.flow_label = keys.tags.flow_label;
2371 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2372 hash_keys.ports.src = keys.ports.src;
2373 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2374 hash_keys.ports.dst = keys.ports.dst;
2375
2376 *p_has_inner = !!(keys.control.flags & FLOW_DIS_ENCAPSULATION);
2377 return fib_multipath_hash_from_keys(net, &hash_keys);
2378}
2379
2380static u32 rt6_multipath_custom_hash_inner(const struct net *net,
2381 const struct sk_buff *skb,
2382 bool has_inner)
2383{
2384 u32 hash_fields = ip6_multipath_hash_fields(net);
2385 struct flow_keys keys, hash_keys;
2386
2387 /* We assume the packet carries an encapsulation, but if none was
2388 * encountered during dissection of the outer flow, then there is no
2389 * point in calling the flow dissector again.
2390 */
2391 if (!has_inner)
2392 return 0;
2393
2394 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_MASK))
2395 return 0;
2396
2397 memset(&hash_keys, 0, sizeof(hash_keys));
2398 skb_flow_dissect_flow_keys(skb, &keys, 0);
2399
2400 if (!(keys.control.flags & FLOW_DIS_ENCAPSULATION))
2401 return 0;
2402
2403 if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2404 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2405 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2406 hash_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src;
2407 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2408 hash_keys.addrs.v4addrs.dst = keys.addrs.v4addrs.dst;
2409 } else if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2410 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2411 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2412 hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2413 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2414 hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2415 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL)
2416 hash_keys.tags.flow_label = keys.tags.flow_label;
2417 }
2418
2419 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO)
2420 hash_keys.basic.ip_proto = keys.basic.ip_proto;
2421 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT)
2422 hash_keys.ports.src = keys.ports.src;
2423 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT)
2424 hash_keys.ports.dst = keys.ports.dst;
2425
2426 return fib_multipath_hash_from_keys(net, &hash_keys);
2427}
2428
2429static u32 rt6_multipath_custom_hash_skb(const struct net *net,
2430 const struct sk_buff *skb)
2431{
2432 u32 mhash, mhash_inner;
2433 bool has_inner = true;
2434
2435 mhash = rt6_multipath_custom_hash_outer(net, skb, &has_inner);
2436 mhash_inner = rt6_multipath_custom_hash_inner(net, skb, has_inner);
2437
2438 return jhash_2words(mhash, mhash_inner, 0);
2439}
2440
2441static u32 rt6_multipath_custom_hash_fl6(const struct net *net,
2442 const struct flowi6 *fl6)
2443{
2444 u32 hash_fields = ip6_multipath_hash_fields(net);
2445 struct flow_keys hash_keys;
2446
2447 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2448 return 0;
2449
2450 memset(&hash_keys, 0, sizeof(hash_keys));
2451 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2452 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2453 hash_keys.addrs.v6addrs.src = fl6->saddr;
2454 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2455 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2456 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2457 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2458 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2459 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2460 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2461 hash_keys.ports.src = fl6->fl6_sport;
2462 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2463 hash_keys.ports.dst = fl6->fl6_dport;
2464
2465 return fib_multipath_hash_from_keys(net, &hash_keys);
2466}
2467
2468/* if skb is set it will be used and fl6 can be NULL */
2469u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2470 const struct sk_buff *skb, struct flow_keys *flkeys)
2471{
2472 struct flow_keys hash_keys;
2473 u32 mhash = 0;
2474
2475 switch (ip6_multipath_hash_policy(net)) {
2476 case 0:
2477 memset(&hash_keys, 0, sizeof(hash_keys));
2478 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2479 if (skb) {
2480 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2481 } else {
2482 hash_keys.addrs.v6addrs.src = fl6->saddr;
2483 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2484 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2485 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2486 }
2487 mhash = fib_multipath_hash_from_keys(net, &hash_keys);
2488 break;
2489 case 1:
2490 if (skb) {
2491 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2492 struct flow_keys keys;
2493
2494 /* short-circuit if we already have L4 hash present */
2495 if (skb->l4_hash)
2496 return skb_get_hash_raw(skb) >> 1;
2497
2498 memset(&hash_keys, 0, sizeof(hash_keys));
2499
2500 if (!flkeys) {
2501 skb_flow_dissect_flow_keys(skb, &keys, flag);
2502 flkeys = &keys;
2503 }
2504 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2505 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2506 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2507 hash_keys.ports.src = flkeys->ports.src;
2508 hash_keys.ports.dst = flkeys->ports.dst;
2509 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2510 } else {
2511 memset(&hash_keys, 0, sizeof(hash_keys));
2512 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2513 hash_keys.addrs.v6addrs.src = fl6->saddr;
2514 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2515 hash_keys.ports.src = fl6->fl6_sport;
2516 hash_keys.ports.dst = fl6->fl6_dport;
2517 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2518 }
2519 mhash = fib_multipath_hash_from_keys(net, &hash_keys);
2520 break;
2521 case 2:
2522 memset(&hash_keys, 0, sizeof(hash_keys));
2523 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2524 if (skb) {
2525 struct flow_keys keys;
2526
2527 if (!flkeys) {
2528 skb_flow_dissect_flow_keys(skb, &keys, 0);
2529 flkeys = &keys;
2530 }
2531
2532 /* Inner can be v4 or v6 */
2533 if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2534 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2535 hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2536 hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2537 } else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2538 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2539 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2540 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2541 hash_keys.tags.flow_label = flkeys->tags.flow_label;
2542 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2543 } else {
2544 /* Same as case 0 */
2545 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2546 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2547 }
2548 } else {
2549 /* Same as case 0 */
2550 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2551 hash_keys.addrs.v6addrs.src = fl6->saddr;
2552 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2553 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2554 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2555 }
2556 mhash = fib_multipath_hash_from_keys(net, &hash_keys);
2557 break;
2558 case 3:
2559 if (skb)
2560 mhash = rt6_multipath_custom_hash_skb(net, skb);
2561 else
2562 mhash = rt6_multipath_custom_hash_fl6(net, fl6);
2563 break;
2564 }
2565
2566 return mhash >> 1;
2567}
2568
2569/* Called with rcu held */
2570void ip6_route_input(struct sk_buff *skb)
2571{
2572 const struct ipv6hdr *iph = ipv6_hdr(skb);
2573 struct net *net = dev_net(skb->dev);
2574 int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
2575 struct ip_tunnel_info *tun_info;
2576 struct flowi6 fl6 = {
2577 .flowi6_iif = skb->dev->ifindex,
2578 .daddr = iph->daddr,
2579 .saddr = iph->saddr,
2580 .flowlabel = ip6_flowinfo(iph),
2581 .flowi6_mark = skb->mark,
2582 .flowi6_proto = iph->nexthdr,
2583 };
2584 struct flow_keys *flkeys = NULL, _flkeys;
2585
2586 tun_info = skb_tunnel_info(skb);
2587 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2588 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2589
2590 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2591 flkeys = &_flkeys;
2592
2593 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2594 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2595 skb_dst_drop(skb);
2596 skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2597 &fl6, skb, flags));
2598}
2599
2600INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_output(struct net *net,
2601 struct fib6_table *table,
2602 struct flowi6 *fl6,
2603 const struct sk_buff *skb,
2604 int flags)
2605{
2606 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2607}
2608
2609static struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2610 const struct sock *sk,
2611 struct flowi6 *fl6,
2612 int flags)
2613{
2614 bool any_src;
2615
2616 if (ipv6_addr_type(&fl6->daddr) &
2617 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
2618 struct dst_entry *dst;
2619
2620 /* This function does not take refcnt on the dst */
2621 dst = l3mdev_link_scope_lookup(net, fl6);
2622 if (dst)
2623 return dst;
2624 }
2625
2626 fl6->flowi6_iif = LOOPBACK_IFINDEX;
2627
2628 flags |= RT6_LOOKUP_F_DST_NOREF;
2629 any_src = ipv6_addr_any(&fl6->saddr);
2630 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
2631 (fl6->flowi6_oif && any_src))
2632 flags |= RT6_LOOKUP_F_IFACE;
2633
2634 if (!any_src)
2635 flags |= RT6_LOOKUP_F_HAS_SADDR;
2636 else if (sk)
2637 flags |= rt6_srcprefs2flags(READ_ONCE(inet6_sk(sk)->srcprefs));
2638
2639 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
2640}
2641
2642struct dst_entry *ip6_route_output_flags(struct net *net,
2643 const struct sock *sk,
2644 struct flowi6 *fl6,
2645 int flags)
2646{
2647 struct dst_entry *dst;
2648 struct rt6_info *rt6;
2649
2650 rcu_read_lock();
2651 dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2652 rt6 = dst_rt6_info(dst);
2653 /* For dst cached in uncached_list, refcnt is already taken. */
2654 if (list_empty(&rt6->dst.rt_uncached) && !dst_hold_safe(dst)) {
2655 dst = &net->ipv6.ip6_null_entry->dst;
2656 dst_hold(dst);
2657 }
2658 rcu_read_unlock();
2659
2660 return dst;
2661}
2662EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2663
2664struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2665{
2666 struct rt6_info *rt, *ort = dst_rt6_info(dst_orig);
2667 struct net_device *loopback_dev = net->loopback_dev;
2668 struct dst_entry *new = NULL;
2669
2670 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev,
2671 DST_OBSOLETE_DEAD, 0);
2672 if (rt) {
2673 rt6_info_init(rt);
2674 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2675
2676 new = &rt->dst;
2677 new->__use = 1;
2678 new->input = dst_discard;
2679 new->output = dst_discard_out;
2680
2681 dst_copy_metrics(new, &ort->dst);
2682
2683 rt->rt6i_idev = in6_dev_get(loopback_dev);
2684 rt->rt6i_gateway = ort->rt6i_gateway;
2685 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2686
2687 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2688#ifdef CONFIG_IPV6_SUBTREES
2689 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2690#endif
2691 }
2692
2693 dst_release(dst_orig);
2694 return new ? new : ERR_PTR(-ENOMEM);
2695}
2696
2697/*
2698 * Destination cache support functions
2699 */
2700
2701static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2702{
2703 u32 rt_cookie = 0;
2704
2705 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2706 return false;
2707
2708 if (fib6_check_expired(f6i))
2709 return false;
2710
2711 return true;
2712}
2713
2714static struct dst_entry *rt6_check(struct rt6_info *rt,
2715 struct fib6_info *from,
2716 u32 cookie)
2717{
2718 u32 rt_cookie = 0;
2719
2720 if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
2721 rt_cookie != cookie)
2722 return NULL;
2723
2724 if (rt6_check_expired(rt))
2725 return NULL;
2726
2727 return &rt->dst;
2728}
2729
2730static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2731 struct fib6_info *from,
2732 u32 cookie)
2733{
2734 if (!__rt6_check_expired(rt) &&
2735 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2736 fib6_check(from, cookie))
2737 return &rt->dst;
2738 else
2739 return NULL;
2740}
2741
2742INDIRECT_CALLABLE_SCOPE struct dst_entry *ip6_dst_check(struct dst_entry *dst,
2743 u32 cookie)
2744{
2745 struct dst_entry *dst_ret;
2746 struct fib6_info *from;
2747 struct rt6_info *rt;
2748
2749 rt = dst_rt6_info(dst);
2750
2751 if (rt->sernum)
2752 return rt6_is_valid(rt) ? dst : NULL;
2753
2754 rcu_read_lock();
2755
2756 /* All IPV6 dsts are created with ->obsolete set to the value
2757 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2758 * into this function always.
2759 */
2760
2761 from = rcu_dereference(rt->from);
2762
2763 if (from && (rt->rt6i_flags & RTF_PCPU ||
2764 unlikely(!list_empty(&rt->dst.rt_uncached))))
2765 dst_ret = rt6_dst_from_check(rt, from, cookie);
2766 else
2767 dst_ret = rt6_check(rt, from, cookie);
2768
2769 rcu_read_unlock();
2770
2771 return dst_ret;
2772}
2773EXPORT_INDIRECT_CALLABLE(ip6_dst_check);
2774
2775static void ip6_negative_advice(struct sock *sk,
2776 struct dst_entry *dst)
2777{
2778 struct rt6_info *rt = dst_rt6_info(dst);
2779
2780 if (rt->rt6i_flags & RTF_CACHE) {
2781 rcu_read_lock();
2782 if (rt6_check_expired(rt)) {
2783 /* rt/dst can not be destroyed yet,
2784 * because of rcu_read_lock()
2785 */
2786 sk_dst_reset(sk);
2787 rt6_remove_exception_rt(rt);
2788 }
2789 rcu_read_unlock();
2790 return;
2791 }
2792 sk_dst_reset(sk);
2793}
2794
2795static void ip6_link_failure(struct sk_buff *skb)
2796{
2797 struct rt6_info *rt;
2798
2799 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2800
2801 rt = dst_rt6_info(skb_dst(skb));
2802 if (rt) {
2803 rcu_read_lock();
2804 if (rt->rt6i_flags & RTF_CACHE) {
2805 rt6_remove_exception_rt(rt);
2806 } else {
2807 struct fib6_info *from;
2808 struct fib6_node *fn;
2809
2810 from = rcu_dereference(rt->from);
2811 if (from) {
2812 fn = rcu_dereference(from->fib6_node);
2813 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2814 WRITE_ONCE(fn->fn_sernum, -1);
2815 }
2816 }
2817 rcu_read_unlock();
2818 }
2819}
2820
2821static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2822{
2823 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2824 struct fib6_info *from;
2825
2826 rcu_read_lock();
2827 from = rcu_dereference(rt0->from);
2828 if (from)
2829 rt0->dst.expires = from->expires;
2830 rcu_read_unlock();
2831 }
2832
2833 dst_set_expires(&rt0->dst, timeout);
2834 rt0->rt6i_flags |= RTF_EXPIRES;
2835}
2836
2837static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2838{
2839 struct net *net = dev_net(rt->dst.dev);
2840
2841 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2842 rt->rt6i_flags |= RTF_MODIFIED;
2843 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2844}
2845
2846static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2847{
2848 return !(rt->rt6i_flags & RTF_CACHE) &&
2849 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
2850}
2851
2852static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2853 const struct ipv6hdr *iph, u32 mtu,
2854 bool confirm_neigh)
2855{
2856 const struct in6_addr *daddr, *saddr;
2857 struct rt6_info *rt6 = dst_rt6_info(dst);
2858
2859 /* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU)
2860 * IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it.
2861 * [see also comment in rt6_mtu_change_route()]
2862 */
2863
2864 if (iph) {
2865 daddr = &iph->daddr;
2866 saddr = &iph->saddr;
2867 } else if (sk) {
2868 daddr = &sk->sk_v6_daddr;
2869 saddr = &inet6_sk(sk)->saddr;
2870 } else {
2871 daddr = NULL;
2872 saddr = NULL;
2873 }
2874
2875 if (confirm_neigh)
2876 dst_confirm_neigh(dst, daddr);
2877
2878 if (mtu < IPV6_MIN_MTU)
2879 return;
2880 if (mtu >= dst_mtu(dst))
2881 return;
2882
2883 if (!rt6_cache_allowed_for_pmtu(rt6)) {
2884 rt6_do_update_pmtu(rt6, mtu);
2885 /* update rt6_ex->stamp for cache */
2886 if (rt6->rt6i_flags & RTF_CACHE)
2887 rt6_update_exception_stamp_rt(rt6);
2888 } else if (daddr) {
2889 struct fib6_result res = {};
2890 struct rt6_info *nrt6;
2891
2892 rcu_read_lock();
2893 res.f6i = rcu_dereference(rt6->from);
2894 if (!res.f6i)
2895 goto out_unlock;
2896
2897 res.fib6_flags = res.f6i->fib6_flags;
2898 res.fib6_type = res.f6i->fib6_type;
2899
2900 if (res.f6i->nh) {
2901 struct fib6_nh_match_arg arg = {
2902 .dev = dst->dev,
2903 .gw = &rt6->rt6i_gateway,
2904 };
2905
2906 nexthop_for_each_fib6_nh(res.f6i->nh,
2907 fib6_nh_find_match, &arg);
2908
2909 /* fib6_info uses a nexthop that does not have fib6_nh
2910 * using the dst->dev + gw. Should be impossible.
2911 */
2912 if (!arg.match)
2913 goto out_unlock;
2914
2915 res.nh = arg.match;
2916 } else {
2917 res.nh = res.f6i->fib6_nh;
2918 }
2919
2920 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
2921 if (nrt6) {
2922 rt6_do_update_pmtu(nrt6, mtu);
2923 if (rt6_insert_exception(nrt6, &res))
2924 dst_release_immediate(&nrt6->dst);
2925 }
2926out_unlock:
2927 rcu_read_unlock();
2928 }
2929}
2930
2931static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2932 struct sk_buff *skb, u32 mtu,
2933 bool confirm_neigh)
2934{
2935 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2936 confirm_neigh);
2937}
2938
2939void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2940 int oif, u32 mark, kuid_t uid)
2941{
2942 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2943 struct dst_entry *dst;
2944 struct flowi6 fl6 = {
2945 .flowi6_oif = oif,
2946 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2947 .daddr = iph->daddr,
2948 .saddr = iph->saddr,
2949 .flowlabel = ip6_flowinfo(iph),
2950 .flowi6_uid = uid,
2951 };
2952
2953 dst = ip6_route_output(net, NULL, &fl6);
2954 if (!dst->error)
2955 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
2956 dst_release(dst);
2957}
2958EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2959
2960void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2961{
2962 int oif = sk->sk_bound_dev_if;
2963 struct dst_entry *dst;
2964
2965 if (!oif && skb->dev)
2966 oif = l3mdev_master_ifindex(skb->dev);
2967
2968 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, READ_ONCE(sk->sk_mark),
2969 sk->sk_uid);
2970
2971 dst = __sk_dst_get(sk);
2972 if (!dst || !dst->obsolete ||
2973 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2974 return;
2975
2976 bh_lock_sock(sk);
2977 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2978 ip6_datagram_dst_update(sk, false);
2979 bh_unlock_sock(sk);
2980}
2981EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2982
2983void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2984 const struct flowi6 *fl6)
2985{
2986#ifdef CONFIG_IPV6_SUBTREES
2987 struct ipv6_pinfo *np = inet6_sk(sk);
2988#endif
2989
2990 ip6_dst_store(sk, dst,
2991 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2992 &sk->sk_v6_daddr : NULL,
2993#ifdef CONFIG_IPV6_SUBTREES
2994 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2995 &np->saddr :
2996#endif
2997 NULL);
2998}
2999
3000static bool ip6_redirect_nh_match(const struct fib6_result *res,
3001 struct flowi6 *fl6,
3002 const struct in6_addr *gw,
3003 struct rt6_info **ret)
3004{
3005 const struct fib6_nh *nh = res->nh;
3006
3007 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
3008 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
3009 return false;
3010
3011 /* rt_cache's gateway might be different from its 'parent'
3012 * in the case of an ip redirect.
3013 * So we keep searching in the exception table if the gateway
3014 * is different.
3015 */
3016 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
3017 struct rt6_info *rt_cache;
3018
3019 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
3020 if (rt_cache &&
3021 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
3022 *ret = rt_cache;
3023 return true;
3024 }
3025 return false;
3026 }
3027 return true;
3028}
3029
3030struct fib6_nh_rd_arg {
3031 struct fib6_result *res;
3032 struct flowi6 *fl6;
3033 const struct in6_addr *gw;
3034 struct rt6_info **ret;
3035};
3036
3037static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
3038{
3039 struct fib6_nh_rd_arg *arg = _arg;
3040
3041 arg->res->nh = nh;
3042 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
3043}
3044
3045/* Handle redirects */
3046struct ip6rd_flowi {
3047 struct flowi6 fl6;
3048 struct in6_addr gateway;
3049};
3050
3051INDIRECT_CALLABLE_SCOPE struct rt6_info *__ip6_route_redirect(struct net *net,
3052 struct fib6_table *table,
3053 struct flowi6 *fl6,
3054 const struct sk_buff *skb,
3055 int flags)
3056{
3057 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
3058 struct rt6_info *ret = NULL;
3059 struct fib6_result res = {};
3060 struct fib6_nh_rd_arg arg = {
3061 .res = &res,
3062 .fl6 = fl6,
3063 .gw = &rdfl->gateway,
3064 .ret = &ret
3065 };
3066 struct fib6_info *rt;
3067 struct fib6_node *fn;
3068
3069 /* Get the "current" route for this destination and
3070 * check if the redirect has come from appropriate router.
3071 *
3072 * RFC 4861 specifies that redirects should only be
3073 * accepted if they come from the nexthop to the target.
3074 * Due to the way the routes are chosen, this notion
3075 * is a bit fuzzy and one might need to check all possible
3076 * routes.
3077 */
3078
3079 rcu_read_lock();
3080 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
3081restart:
3082 for_each_fib6_node_rt_rcu(fn) {
3083 res.f6i = rt;
3084 if (fib6_check_expired(rt))
3085 continue;
3086 if (rt->fib6_flags & RTF_REJECT)
3087 break;
3088 if (unlikely(rt->nh)) {
3089 if (nexthop_is_blackhole(rt->nh))
3090 continue;
3091 /* on match, res->nh is filled in and potentially ret */
3092 if (nexthop_for_each_fib6_nh(rt->nh,
3093 fib6_nh_redirect_match,
3094 &arg))
3095 goto out;
3096 } else {
3097 res.nh = rt->fib6_nh;
3098 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
3099 &ret))
3100 goto out;
3101 }
3102 }
3103
3104 if (!rt)
3105 rt = net->ipv6.fib6_null_entry;
3106 else if (rt->fib6_flags & RTF_REJECT) {
3107 ret = net->ipv6.ip6_null_entry;
3108 goto out;
3109 }
3110
3111 if (rt == net->ipv6.fib6_null_entry) {
3112 fn = fib6_backtrack(fn, &fl6->saddr);
3113 if (fn)
3114 goto restart;
3115 }
3116
3117 res.f6i = rt;
3118 res.nh = rt->fib6_nh;
3119out:
3120 if (ret) {
3121 ip6_hold_safe(net, &ret);
3122 } else {
3123 res.fib6_flags = res.f6i->fib6_flags;
3124 res.fib6_type = res.f6i->fib6_type;
3125 ret = ip6_create_rt_rcu(&res);
3126 }
3127
3128 rcu_read_unlock();
3129
3130 trace_fib6_table_lookup(net, &res, table, fl6);
3131 return ret;
3132};
3133
3134static struct dst_entry *ip6_route_redirect(struct net *net,
3135 const struct flowi6 *fl6,
3136 const struct sk_buff *skb,
3137 const struct in6_addr *gateway)
3138{
3139 int flags = RT6_LOOKUP_F_HAS_SADDR;
3140 struct ip6rd_flowi rdfl;
3141
3142 rdfl.fl6 = *fl6;
3143 rdfl.gateway = *gateway;
3144
3145 return fib6_rule_lookup(net, &rdfl.fl6, skb,
3146 flags, __ip6_route_redirect);
3147}
3148
3149void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
3150 kuid_t uid)
3151{
3152 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
3153 struct dst_entry *dst;
3154 struct flowi6 fl6 = {
3155 .flowi6_iif = LOOPBACK_IFINDEX,
3156 .flowi6_oif = oif,
3157 .flowi6_mark = mark,
3158 .daddr = iph->daddr,
3159 .saddr = iph->saddr,
3160 .flowlabel = ip6_flowinfo(iph),
3161 .flowi6_uid = uid,
3162 };
3163
3164 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
3165 rt6_do_redirect(dst, NULL, skb);
3166 dst_release(dst);
3167}
3168EXPORT_SYMBOL_GPL(ip6_redirect);
3169
3170void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
3171{
3172 const struct ipv6hdr *iph = ipv6_hdr(skb);
3173 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3174 struct dst_entry *dst;
3175 struct flowi6 fl6 = {
3176 .flowi6_iif = LOOPBACK_IFINDEX,
3177 .flowi6_oif = oif,
3178 .daddr = msg->dest,
3179 .saddr = iph->daddr,
3180 .flowi6_uid = sock_net_uid(net, NULL),
3181 };
3182
3183 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
3184 rt6_do_redirect(dst, NULL, skb);
3185 dst_release(dst);
3186}
3187
3188void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3189{
3190 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if,
3191 READ_ONCE(sk->sk_mark), sk->sk_uid);
3192}
3193EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3194
3195static unsigned int ip6_default_advmss(const struct dst_entry *dst)
3196{
3197 struct net_device *dev = dst->dev;
3198 unsigned int mtu = dst_mtu(dst);
3199 struct net *net;
3200
3201 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3202
3203 rcu_read_lock();
3204
3205 net = dev_net_rcu(dev);
3206 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3207 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3208
3209 rcu_read_unlock();
3210
3211 /*
3212 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3213 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3214 * IPV6_MAXPLEN is also valid and means: "any MSS,
3215 * rely only on pmtu discovery"
3216 */
3217 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3218 mtu = IPV6_MAXPLEN;
3219 return mtu;
3220}
3221
3222INDIRECT_CALLABLE_SCOPE unsigned int ip6_mtu(const struct dst_entry *dst)
3223{
3224 return ip6_dst_mtu_maybe_forward(dst, false);
3225}
3226EXPORT_INDIRECT_CALLABLE(ip6_mtu);
3227
3228/* MTU selection:
3229 * 1. mtu on route is locked - use it
3230 * 2. mtu from nexthop exception
3231 * 3. mtu from egress device
3232 *
3233 * based on ip6_dst_mtu_forward and exception logic of
3234 * rt6_find_cached_rt; called with rcu_read_lock
3235 */
3236u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3237 const struct in6_addr *daddr,
3238 const struct in6_addr *saddr)
3239{
3240 const struct fib6_nh *nh = res->nh;
3241 struct fib6_info *f6i = res->f6i;
3242 struct inet6_dev *idev;
3243 struct rt6_info *rt;
3244 u32 mtu = 0;
3245
3246 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3247 mtu = f6i->fib6_pmtu;
3248 if (mtu)
3249 goto out;
3250 }
3251
3252 rt = rt6_find_cached_rt(res, daddr, saddr);
3253 if (unlikely(rt)) {
3254 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3255 } else {
3256 struct net_device *dev = nh->fib_nh_dev;
3257
3258 mtu = IPV6_MIN_MTU;
3259 idev = __in6_dev_get(dev);
3260 if (idev)
3261 mtu = max_t(u32, mtu, READ_ONCE(idev->cnf.mtu6));
3262 }
3263
3264 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3265out:
3266 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3267}
3268
3269struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3270 struct flowi6 *fl6)
3271{
3272 struct dst_entry *dst;
3273 struct rt6_info *rt;
3274 struct inet6_dev *idev = in6_dev_get(dev);
3275 struct net *net = dev_net(dev);
3276
3277 if (unlikely(!idev))
3278 return ERR_PTR(-ENODEV);
3279
3280 rt = ip6_dst_alloc(net, dev, 0);
3281 if (unlikely(!rt)) {
3282 in6_dev_put(idev);
3283 dst = ERR_PTR(-ENOMEM);
3284 goto out;
3285 }
3286
3287 rt->dst.input = ip6_input;
3288 rt->dst.output = ip6_output;
3289 rt->rt6i_gateway = fl6->daddr;
3290 rt->rt6i_dst.addr = fl6->daddr;
3291 rt->rt6i_dst.plen = 128;
3292 rt->rt6i_idev = idev;
3293 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3294
3295 /* Add this dst into uncached_list so that rt6_disable_ip() can
3296 * do proper release of the net_device
3297 */
3298 rt6_uncached_list_add(rt);
3299
3300 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3301
3302out:
3303 return dst;
3304}
3305
3306static void ip6_dst_gc(struct dst_ops *ops)
3307{
3308 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3309 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3310 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3311 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3312 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3313 unsigned int val;
3314 int entries;
3315
3316 if (time_after(rt_last_gc + rt_min_interval, jiffies))
3317 goto out;
3318
3319 fib6_run_gc(atomic_inc_return(&net->ipv6.ip6_rt_gc_expire), net, true);
3320 entries = dst_entries_get_slow(ops);
3321 if (entries < ops->gc_thresh)
3322 atomic_set(&net->ipv6.ip6_rt_gc_expire, rt_gc_timeout >> 1);
3323out:
3324 val = atomic_read(&net->ipv6.ip6_rt_gc_expire);
3325 atomic_set(&net->ipv6.ip6_rt_gc_expire, val - (val >> rt_elasticity));
3326}
3327
3328static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3329 const struct in6_addr *gw_addr, u32 tbid,
3330 int flags, struct fib6_result *res)
3331{
3332 struct flowi6 fl6 = {
3333 .flowi6_oif = cfg->fc_ifindex,
3334 .daddr = *gw_addr,
3335 .saddr = cfg->fc_prefsrc,
3336 };
3337 struct fib6_table *table;
3338 int err;
3339
3340 table = fib6_get_table(net, tbid);
3341 if (!table)
3342 return -EINVAL;
3343
3344 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3345 flags |= RT6_LOOKUP_F_HAS_SADDR;
3346
3347 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3348
3349 err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3350 if (!err && res->f6i != net->ipv6.fib6_null_entry)
3351 fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3352 cfg->fc_ifindex != 0, NULL, flags);
3353
3354 return err;
3355}
3356
3357static int ip6_route_check_nh_onlink(struct net *net,
3358 struct fib6_config *cfg,
3359 const struct net_device *dev,
3360 struct netlink_ext_ack *extack)
3361{
3362 u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3363 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3364 struct fib6_result res = {};
3365 int err;
3366
3367 err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3368 if (!err && !(res.fib6_flags & RTF_REJECT) &&
3369 /* ignore match if it is the default route */
3370 !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3371 (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3372 NL_SET_ERR_MSG(extack,
3373 "Nexthop has invalid gateway or device mismatch");
3374 err = -EINVAL;
3375 }
3376
3377 return err;
3378}
3379
3380static int ip6_route_check_nh(struct net *net,
3381 struct fib6_config *cfg,
3382 struct net_device **_dev,
3383 netdevice_tracker *dev_tracker,
3384 struct inet6_dev **idev)
3385{
3386 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3387 struct net_device *dev = _dev ? *_dev : NULL;
3388 int flags = RT6_LOOKUP_F_IFACE;
3389 struct fib6_result res = {};
3390 int err = -EHOSTUNREACH;
3391
3392 if (cfg->fc_table) {
3393 err = ip6_nh_lookup_table(net, cfg, gw_addr,
3394 cfg->fc_table, flags, &res);
3395 /* gw_addr can not require a gateway or resolve to a reject
3396 * route. If a device is given, it must match the result.
3397 */
3398 if (err || res.fib6_flags & RTF_REJECT ||
3399 res.nh->fib_nh_gw_family ||
3400 (dev && dev != res.nh->fib_nh_dev))
3401 err = -EHOSTUNREACH;
3402 }
3403
3404 if (err < 0) {
3405 struct flowi6 fl6 = {
3406 .flowi6_oif = cfg->fc_ifindex,
3407 .daddr = *gw_addr,
3408 };
3409
3410 err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3411 if (err || res.fib6_flags & RTF_REJECT ||
3412 res.nh->fib_nh_gw_family)
3413 err = -EHOSTUNREACH;
3414
3415 if (err)
3416 return err;
3417
3418 fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3419 cfg->fc_ifindex != 0, NULL, flags);
3420 }
3421
3422 err = 0;
3423 if (dev) {
3424 if (dev != res.nh->fib_nh_dev)
3425 err = -EHOSTUNREACH;
3426 } else {
3427 *_dev = dev = res.nh->fib_nh_dev;
3428 netdev_hold(dev, dev_tracker, GFP_ATOMIC);
3429 *idev = in6_dev_get(dev);
3430 }
3431
3432 return err;
3433}
3434
3435static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3436 struct net_device **_dev,
3437 netdevice_tracker *dev_tracker,
3438 struct inet6_dev **idev,
3439 struct netlink_ext_ack *extack)
3440{
3441 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3442 int gwa_type = ipv6_addr_type(gw_addr);
3443 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3444 const struct net_device *dev = *_dev;
3445 bool need_addr_check = !dev;
3446 int err = -EINVAL;
3447
3448 /* if gw_addr is local we will fail to detect this in case
3449 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3450 * will return already-added prefix route via interface that
3451 * prefix route was assigned to, which might be non-loopback.
3452 */
3453 if (dev &&
3454 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3455 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3456 goto out;
3457 }
3458
3459 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3460 /* IPv6 strictly inhibits using not link-local
3461 * addresses as nexthop address.
3462 * Otherwise, router will not able to send redirects.
3463 * It is very good, but in some (rare!) circumstances
3464 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3465 * some exceptions. --ANK
3466 * We allow IPv4-mapped nexthops to support RFC4798-type
3467 * addressing
3468 */
3469 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3470 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3471 goto out;
3472 }
3473
3474 rcu_read_lock();
3475
3476 if (cfg->fc_flags & RTNH_F_ONLINK)
3477 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3478 else
3479 err = ip6_route_check_nh(net, cfg, _dev, dev_tracker,
3480 idev);
3481
3482 rcu_read_unlock();
3483
3484 if (err)
3485 goto out;
3486 }
3487
3488 /* reload in case device was changed */
3489 dev = *_dev;
3490
3491 err = -EINVAL;
3492 if (!dev) {
3493 NL_SET_ERR_MSG(extack, "Egress device not specified");
3494 goto out;
3495 } else if (dev->flags & IFF_LOOPBACK) {
3496 NL_SET_ERR_MSG(extack,
3497 "Egress device can not be loopback device for this route");
3498 goto out;
3499 }
3500
3501 /* if we did not check gw_addr above, do so now that the
3502 * egress device has been resolved.
3503 */
3504 if (need_addr_check &&
3505 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3506 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3507 goto out;
3508 }
3509
3510 err = 0;
3511out:
3512 return err;
3513}
3514
3515static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3516{
3517 if ((flags & RTF_REJECT) ||
3518 (dev && (dev->flags & IFF_LOOPBACK) &&
3519 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3520 !(flags & (RTF_ANYCAST | RTF_LOCAL))))
3521 return true;
3522
3523 return false;
3524}
3525
3526int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3527 struct fib6_config *cfg, gfp_t gfp_flags,
3528 struct netlink_ext_ack *extack)
3529{
3530 netdevice_tracker *dev_tracker = &fib6_nh->fib_nh_dev_tracker;
3531 struct net_device *dev = NULL;
3532 struct inet6_dev *idev = NULL;
3533 int addr_type;
3534 int err;
3535
3536 fib6_nh->fib_nh_family = AF_INET6;
3537#ifdef CONFIG_IPV6_ROUTER_PREF
3538 fib6_nh->last_probe = jiffies;
3539#endif
3540 if (cfg->fc_is_fdb) {
3541 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3542 fib6_nh->fib_nh_gw_family = AF_INET6;
3543 return 0;
3544 }
3545
3546 err = -ENODEV;
3547 if (cfg->fc_ifindex) {
3548 dev = netdev_get_by_index(net, cfg->fc_ifindex,
3549 dev_tracker, gfp_flags);
3550 if (!dev)
3551 goto out;
3552 idev = in6_dev_get(dev);
3553 if (!idev)
3554 goto out;
3555 }
3556
3557 if (cfg->fc_flags & RTNH_F_ONLINK) {
3558 if (!dev) {
3559 NL_SET_ERR_MSG(extack,
3560 "Nexthop device required for onlink");
3561 goto out;
3562 }
3563
3564 if (!(dev->flags & IFF_UP)) {
3565 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3566 err = -ENETDOWN;
3567 goto out;
3568 }
3569
3570 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3571 }
3572
3573 fib6_nh->fib_nh_weight = 1;
3574
3575 /* We cannot add true routes via loopback here,
3576 * they would result in kernel looping; promote them to reject routes
3577 */
3578 addr_type = ipv6_addr_type(&cfg->fc_dst);
3579 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3580 /* hold loopback dev/idev if we haven't done so. */
3581 if (dev != net->loopback_dev) {
3582 if (dev) {
3583 netdev_put(dev, dev_tracker);
3584 in6_dev_put(idev);
3585 }
3586 dev = net->loopback_dev;
3587 netdev_hold(dev, dev_tracker, gfp_flags);
3588 idev = in6_dev_get(dev);
3589 if (!idev) {
3590 err = -ENODEV;
3591 goto out;
3592 }
3593 }
3594 goto pcpu_alloc;
3595 }
3596
3597 if (cfg->fc_flags & RTF_GATEWAY) {
3598 err = ip6_validate_gw(net, cfg, &dev, dev_tracker,
3599 &idev, extack);
3600 if (err)
3601 goto out;
3602
3603 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3604 fib6_nh->fib_nh_gw_family = AF_INET6;
3605 }
3606
3607 err = -ENODEV;
3608 if (!dev)
3609 goto out;
3610
3611 if (!idev || idev->cnf.disable_ipv6) {
3612 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3613 err = -EACCES;
3614 goto out;
3615 }
3616
3617 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3618 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3619 err = -ENETDOWN;
3620 goto out;
3621 }
3622
3623 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3624 !netif_carrier_ok(dev))
3625 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3626
3627 err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
3628 cfg->fc_encap_type, cfg, gfp_flags, extack);
3629 if (err)
3630 goto out;
3631
3632pcpu_alloc:
3633 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3634 if (!fib6_nh->rt6i_pcpu) {
3635 err = -ENOMEM;
3636 goto out;
3637 }
3638
3639 fib6_nh->fib_nh_dev = dev;
3640 fib6_nh->fib_nh_oif = dev->ifindex;
3641 err = 0;
3642out:
3643 if (idev)
3644 in6_dev_put(idev);
3645
3646 if (err) {
3647 lwtstate_put(fib6_nh->fib_nh_lws);
3648 fib6_nh->fib_nh_lws = NULL;
3649 netdev_put(dev, dev_tracker);
3650 }
3651
3652 return err;
3653}
3654
3655void fib6_nh_release(struct fib6_nh *fib6_nh)
3656{
3657 struct rt6_exception_bucket *bucket;
3658
3659 rcu_read_lock();
3660
3661 fib6_nh_flush_exceptions(fib6_nh, NULL);
3662 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3663 if (bucket) {
3664 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3665 kfree(bucket);
3666 }
3667
3668 rcu_read_unlock();
3669
3670 fib6_nh_release_dsts(fib6_nh);
3671 free_percpu(fib6_nh->rt6i_pcpu);
3672
3673 fib_nh_common_release(&fib6_nh->nh_common);
3674}
3675
3676void fib6_nh_release_dsts(struct fib6_nh *fib6_nh)
3677{
3678 int cpu;
3679
3680 if (!fib6_nh->rt6i_pcpu)
3681 return;
3682
3683 for_each_possible_cpu(cpu) {
3684 struct rt6_info *pcpu_rt, **ppcpu_rt;
3685
3686 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3687 pcpu_rt = xchg(ppcpu_rt, NULL);
3688 if (pcpu_rt) {
3689 dst_dev_put(&pcpu_rt->dst);
3690 dst_release(&pcpu_rt->dst);
3691 }
3692 }
3693}
3694
3695static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3696 gfp_t gfp_flags,
3697 struct netlink_ext_ack *extack)
3698{
3699 struct net *net = cfg->fc_nlinfo.nl_net;
3700 struct fib6_info *rt = NULL;
3701 struct nexthop *nh = NULL;
3702 struct fib6_table *table;
3703 struct fib6_nh *fib6_nh;
3704 int err = -EINVAL;
3705 int addr_type;
3706
3707 /* RTF_PCPU is an internal flag; can not be set by userspace */
3708 if (cfg->fc_flags & RTF_PCPU) {
3709 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3710 goto out;
3711 }
3712
3713 /* RTF_CACHE is an internal flag; can not be set by userspace */
3714 if (cfg->fc_flags & RTF_CACHE) {
3715 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3716 goto out;
3717 }
3718
3719 if (cfg->fc_type > RTN_MAX) {
3720 NL_SET_ERR_MSG(extack, "Invalid route type");
3721 goto out;
3722 }
3723
3724 if (cfg->fc_dst_len > 128) {
3725 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3726 goto out;
3727 }
3728 if (cfg->fc_src_len > 128) {
3729 NL_SET_ERR_MSG(extack, "Invalid source address length");
3730 goto out;
3731 }
3732#ifndef CONFIG_IPV6_SUBTREES
3733 if (cfg->fc_src_len) {
3734 NL_SET_ERR_MSG(extack,
3735 "Specifying source address requires IPV6_SUBTREES to be enabled");
3736 goto out;
3737 }
3738#endif
3739 if (cfg->fc_nh_id) {
3740 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3741 if (!nh) {
3742 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3743 goto out;
3744 }
3745 err = fib6_check_nexthop(nh, cfg, extack);
3746 if (err)
3747 goto out;
3748 }
3749
3750 err = -ENOBUFS;
3751 if (cfg->fc_nlinfo.nlh &&
3752 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3753 table = fib6_get_table(net, cfg->fc_table);
3754 if (!table) {
3755 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3756 table = fib6_new_table(net, cfg->fc_table);
3757 }
3758 } else {
3759 table = fib6_new_table(net, cfg->fc_table);
3760 }
3761
3762 if (!table)
3763 goto out;
3764
3765 err = -ENOMEM;
3766 rt = fib6_info_alloc(gfp_flags, !nh);
3767 if (!rt)
3768 goto out;
3769
3770 rt->fib6_metrics = ip_fib_metrics_init(cfg->fc_mx, cfg->fc_mx_len,
3771 extack);
3772 if (IS_ERR(rt->fib6_metrics)) {
3773 err = PTR_ERR(rt->fib6_metrics);
3774 /* Do not leave garbage there. */
3775 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3776 goto out_free;
3777 }
3778
3779 if (cfg->fc_flags & RTF_ADDRCONF)
3780 rt->dst_nocount = true;
3781
3782 if (cfg->fc_flags & RTF_EXPIRES)
3783 fib6_set_expires(rt, jiffies +
3784 clock_t_to_jiffies(cfg->fc_expires));
3785
3786 if (cfg->fc_protocol == RTPROT_UNSPEC)
3787 cfg->fc_protocol = RTPROT_BOOT;
3788 rt->fib6_protocol = cfg->fc_protocol;
3789
3790 rt->fib6_table = table;
3791 rt->fib6_metric = cfg->fc_metric;
3792 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3793 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3794
3795 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3796 rt->fib6_dst.plen = cfg->fc_dst_len;
3797
3798#ifdef CONFIG_IPV6_SUBTREES
3799 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3800 rt->fib6_src.plen = cfg->fc_src_len;
3801#endif
3802 if (nh) {
3803 if (rt->fib6_src.plen) {
3804 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3805 goto out_free;
3806 }
3807 if (!nexthop_get(nh)) {
3808 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3809 goto out_free;
3810 }
3811 rt->nh = nh;
3812 fib6_nh = nexthop_fib6_nh(rt->nh);
3813 } else {
3814 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3815 if (err)
3816 goto out;
3817
3818 fib6_nh = rt->fib6_nh;
3819
3820 /* We cannot add true routes via loopback here, they would
3821 * result in kernel looping; promote them to reject routes
3822 */
3823 addr_type = ipv6_addr_type(&cfg->fc_dst);
3824 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3825 addr_type))
3826 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3827 }
3828
3829 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3830 struct net_device *dev = fib6_nh->fib_nh_dev;
3831
3832 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3833 NL_SET_ERR_MSG(extack, "Invalid source address");
3834 err = -EINVAL;
3835 goto out;
3836 }
3837 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3838 rt->fib6_prefsrc.plen = 128;
3839 } else
3840 rt->fib6_prefsrc.plen = 0;
3841
3842 return rt;
3843out:
3844 fib6_info_release(rt);
3845 return ERR_PTR(err);
3846out_free:
3847 ip_fib_metrics_put(rt->fib6_metrics);
3848 kfree(rt);
3849 return ERR_PTR(err);
3850}
3851
3852int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3853 struct netlink_ext_ack *extack)
3854{
3855 struct fib6_info *rt;
3856 int err;
3857
3858 rt = ip6_route_info_create(cfg, gfp_flags, extack);
3859 if (IS_ERR(rt))
3860 return PTR_ERR(rt);
3861
3862 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3863 fib6_info_release(rt);
3864
3865 return err;
3866}
3867
3868static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3869{
3870 struct net *net = info->nl_net;
3871 struct fib6_table *table;
3872 int err;
3873
3874 if (rt == net->ipv6.fib6_null_entry) {
3875 err = -ENOENT;
3876 goto out;
3877 }
3878
3879 table = rt->fib6_table;
3880 spin_lock_bh(&table->tb6_lock);
3881 err = fib6_del(rt, info);
3882 spin_unlock_bh(&table->tb6_lock);
3883
3884out:
3885 fib6_info_release(rt);
3886 return err;
3887}
3888
3889int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
3890{
3891 struct nl_info info = {
3892 .nl_net = net,
3893 .skip_notify = skip_notify
3894 };
3895
3896 return __ip6_del_rt(rt, &info);
3897}
3898
3899static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3900{
3901 struct nl_info *info = &cfg->fc_nlinfo;
3902 struct net *net = info->nl_net;
3903 struct sk_buff *skb = NULL;
3904 struct fib6_table *table;
3905 int err = -ENOENT;
3906
3907 if (rt == net->ipv6.fib6_null_entry)
3908 goto out_put;
3909 table = rt->fib6_table;
3910 spin_lock_bh(&table->tb6_lock);
3911
3912 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3913 struct fib6_info *sibling, *next_sibling;
3914 struct fib6_node *fn;
3915
3916 /* prefer to send a single notification with all hops */
3917 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3918 if (skb) {
3919 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3920
3921 if (rt6_fill_node(net, skb, rt, NULL,
3922 NULL, NULL, 0, RTM_DELROUTE,
3923 info->portid, seq, 0) < 0) {
3924 kfree_skb(skb);
3925 skb = NULL;
3926 } else
3927 info->skip_notify = 1;
3928 }
3929
3930 /* 'rt' points to the first sibling route. If it is not the
3931 * leaf, then we do not need to send a notification. Otherwise,
3932 * we need to check if the last sibling has a next route or not
3933 * and emit a replace or delete notification, respectively.
3934 */
3935 info->skip_notify_kernel = 1;
3936 fn = rcu_dereference_protected(rt->fib6_node,
3937 lockdep_is_held(&table->tb6_lock));
3938 if (rcu_access_pointer(fn->leaf) == rt) {
3939 struct fib6_info *last_sibling, *replace_rt;
3940
3941 last_sibling = list_last_entry(&rt->fib6_siblings,
3942 struct fib6_info,
3943 fib6_siblings);
3944 replace_rt = rcu_dereference_protected(
3945 last_sibling->fib6_next,
3946 lockdep_is_held(&table->tb6_lock));
3947 if (replace_rt)
3948 call_fib6_entry_notifiers_replace(net,
3949 replace_rt);
3950 else
3951 call_fib6_multipath_entry_notifiers(net,
3952 FIB_EVENT_ENTRY_DEL,
3953 rt, rt->fib6_nsiblings,
3954 NULL);
3955 }
3956 list_for_each_entry_safe(sibling, next_sibling,
3957 &rt->fib6_siblings,
3958 fib6_siblings) {
3959 err = fib6_del(sibling, info);
3960 if (err)
3961 goto out_unlock;
3962 }
3963 }
3964
3965 err = fib6_del(rt, info);
3966out_unlock:
3967 spin_unlock_bh(&table->tb6_lock);
3968out_put:
3969 fib6_info_release(rt);
3970
3971 if (skb) {
3972 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3973 info->nlh, gfp_any());
3974 }
3975 return err;
3976}
3977
3978static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3979{
3980 int rc = -ESRCH;
3981
3982 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3983 goto out;
3984
3985 if (cfg->fc_flags & RTF_GATEWAY &&
3986 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3987 goto out;
3988
3989 rc = rt6_remove_exception_rt(rt);
3990out:
3991 return rc;
3992}
3993
3994static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3995 struct fib6_nh *nh)
3996{
3997 struct fib6_result res = {
3998 .f6i = rt,
3999 .nh = nh,
4000 };
4001 struct rt6_info *rt_cache;
4002
4003 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
4004 if (rt_cache)
4005 return __ip6_del_cached_rt(rt_cache, cfg);
4006
4007 return 0;
4008}
4009
4010struct fib6_nh_del_cached_rt_arg {
4011 struct fib6_config *cfg;
4012 struct fib6_info *f6i;
4013};
4014
4015static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
4016{
4017 struct fib6_nh_del_cached_rt_arg *arg = _arg;
4018 int rc;
4019
4020 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
4021 return rc != -ESRCH ? rc : 0;
4022}
4023
4024static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
4025{
4026 struct fib6_nh_del_cached_rt_arg arg = {
4027 .cfg = cfg,
4028 .f6i = f6i
4029 };
4030
4031 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
4032}
4033
4034static int ip6_route_del(struct fib6_config *cfg,
4035 struct netlink_ext_ack *extack)
4036{
4037 struct fib6_table *table;
4038 struct fib6_info *rt;
4039 struct fib6_node *fn;
4040 int err = -ESRCH;
4041
4042 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
4043 if (!table) {
4044 NL_SET_ERR_MSG(extack, "FIB table does not exist");
4045 return err;
4046 }
4047
4048 rcu_read_lock();
4049
4050 fn = fib6_locate(&table->tb6_root,
4051 &cfg->fc_dst, cfg->fc_dst_len,
4052 &cfg->fc_src, cfg->fc_src_len,
4053 !(cfg->fc_flags & RTF_CACHE));
4054
4055 if (fn) {
4056 for_each_fib6_node_rt_rcu(fn) {
4057 struct fib6_nh *nh;
4058
4059 if (rt->nh && cfg->fc_nh_id &&
4060 rt->nh->id != cfg->fc_nh_id)
4061 continue;
4062
4063 if (cfg->fc_flags & RTF_CACHE) {
4064 int rc = 0;
4065
4066 if (rt->nh) {
4067 rc = ip6_del_cached_rt_nh(cfg, rt);
4068 } else if (cfg->fc_nh_id) {
4069 continue;
4070 } else {
4071 nh = rt->fib6_nh;
4072 rc = ip6_del_cached_rt(cfg, rt, nh);
4073 }
4074 if (rc != -ESRCH) {
4075 rcu_read_unlock();
4076 return rc;
4077 }
4078 continue;
4079 }
4080
4081 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
4082 continue;
4083 if (cfg->fc_protocol &&
4084 cfg->fc_protocol != rt->fib6_protocol)
4085 continue;
4086
4087 if (rt->nh) {
4088 if (!fib6_info_hold_safe(rt))
4089 continue;
4090 rcu_read_unlock();
4091
4092 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4093 }
4094 if (cfg->fc_nh_id)
4095 continue;
4096
4097 nh = rt->fib6_nh;
4098 if (cfg->fc_ifindex &&
4099 (!nh->fib_nh_dev ||
4100 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
4101 continue;
4102 if (cfg->fc_flags & RTF_GATEWAY &&
4103 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
4104 continue;
4105 if (!fib6_info_hold_safe(rt))
4106 continue;
4107 rcu_read_unlock();
4108
4109 /* if gateway was specified only delete the one hop */
4110 if (cfg->fc_flags & RTF_GATEWAY)
4111 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4112
4113 return __ip6_del_rt_siblings(rt, cfg);
4114 }
4115 }
4116 rcu_read_unlock();
4117
4118 return err;
4119}
4120
4121static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
4122{
4123 struct netevent_redirect netevent;
4124 struct rt6_info *rt, *nrt = NULL;
4125 struct fib6_result res = {};
4126 struct ndisc_options ndopts;
4127 struct inet6_dev *in6_dev;
4128 struct neighbour *neigh;
4129 struct rd_msg *msg;
4130 int optlen, on_link;
4131 u8 *lladdr;
4132
4133 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
4134 optlen -= sizeof(*msg);
4135
4136 if (optlen < 0) {
4137 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
4138 return;
4139 }
4140
4141 msg = (struct rd_msg *)icmp6_hdr(skb);
4142
4143 if (ipv6_addr_is_multicast(&msg->dest)) {
4144 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
4145 return;
4146 }
4147
4148 on_link = 0;
4149 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
4150 on_link = 1;
4151 } else if (ipv6_addr_type(&msg->target) !=
4152 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
4153 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
4154 return;
4155 }
4156
4157 in6_dev = __in6_dev_get(skb->dev);
4158 if (!in6_dev)
4159 return;
4160 if (READ_ONCE(in6_dev->cnf.forwarding) ||
4161 !READ_ONCE(in6_dev->cnf.accept_redirects))
4162 return;
4163
4164 /* RFC2461 8.1:
4165 * The IP source address of the Redirect MUST be the same as the current
4166 * first-hop router for the specified ICMP Destination Address.
4167 */
4168
4169 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
4170 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4171 return;
4172 }
4173
4174 lladdr = NULL;
4175 if (ndopts.nd_opts_tgt_lladdr) {
4176 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4177 skb->dev);
4178 if (!lladdr) {
4179 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4180 return;
4181 }
4182 }
4183
4184 rt = dst_rt6_info(dst);
4185 if (rt->rt6i_flags & RTF_REJECT) {
4186 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
4187 return;
4188 }
4189
4190 /* Redirect received -> path was valid.
4191 * Look, redirects are sent only in response to data packets,
4192 * so that this nexthop apparently is reachable. --ANK
4193 */
4194 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
4195
4196 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
4197 if (!neigh)
4198 return;
4199
4200 /*
4201 * We have finally decided to accept it.
4202 */
4203
4204 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
4205 NEIGH_UPDATE_F_WEAK_OVERRIDE|
4206 NEIGH_UPDATE_F_OVERRIDE|
4207 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
4208 NEIGH_UPDATE_F_ISROUTER)),
4209 NDISC_REDIRECT, &ndopts);
4210
4211 rcu_read_lock();
4212 res.f6i = rcu_dereference(rt->from);
4213 if (!res.f6i)
4214 goto out;
4215
4216 if (res.f6i->nh) {
4217 struct fib6_nh_match_arg arg = {
4218 .dev = dst->dev,
4219 .gw = &rt->rt6i_gateway,
4220 };
4221
4222 nexthop_for_each_fib6_nh(res.f6i->nh,
4223 fib6_nh_find_match, &arg);
4224
4225 /* fib6_info uses a nexthop that does not have fib6_nh
4226 * using the dst->dev. Should be impossible
4227 */
4228 if (!arg.match)
4229 goto out;
4230 res.nh = arg.match;
4231 } else {
4232 res.nh = res.f6i->fib6_nh;
4233 }
4234
4235 res.fib6_flags = res.f6i->fib6_flags;
4236 res.fib6_type = res.f6i->fib6_type;
4237 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4238 if (!nrt)
4239 goto out;
4240
4241 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4242 if (on_link)
4243 nrt->rt6i_flags &= ~RTF_GATEWAY;
4244
4245 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4246
4247 /* rt6_insert_exception() will take care of duplicated exceptions */
4248 if (rt6_insert_exception(nrt, &res)) {
4249 dst_release_immediate(&nrt->dst);
4250 goto out;
4251 }
4252
4253 netevent.old = &rt->dst;
4254 netevent.new = &nrt->dst;
4255 netevent.daddr = &msg->dest;
4256 netevent.neigh = neigh;
4257 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4258
4259out:
4260 rcu_read_unlock();
4261 neigh_release(neigh);
4262}
4263
4264#ifdef CONFIG_IPV6_ROUTE_INFO
4265static struct fib6_info *rt6_get_route_info(struct net *net,
4266 const struct in6_addr *prefix, int prefixlen,
4267 const struct in6_addr *gwaddr,
4268 struct net_device *dev)
4269{
4270 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4271 int ifindex = dev->ifindex;
4272 struct fib6_node *fn;
4273 struct fib6_info *rt = NULL;
4274 struct fib6_table *table;
4275
4276 table = fib6_get_table(net, tb_id);
4277 if (!table)
4278 return NULL;
4279
4280 rcu_read_lock();
4281 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4282 if (!fn)
4283 goto out;
4284
4285 for_each_fib6_node_rt_rcu(fn) {
4286 /* these routes do not use nexthops */
4287 if (rt->nh)
4288 continue;
4289 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4290 continue;
4291 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4292 !rt->fib6_nh->fib_nh_gw_family)
4293 continue;
4294 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4295 continue;
4296 if (!fib6_info_hold_safe(rt))
4297 continue;
4298 break;
4299 }
4300out:
4301 rcu_read_unlock();
4302 return rt;
4303}
4304
4305static struct fib6_info *rt6_add_route_info(struct net *net,
4306 const struct in6_addr *prefix, int prefixlen,
4307 const struct in6_addr *gwaddr,
4308 struct net_device *dev,
4309 unsigned int pref)
4310{
4311 struct fib6_config cfg = {
4312 .fc_metric = IP6_RT_PRIO_USER,
4313 .fc_ifindex = dev->ifindex,
4314 .fc_dst_len = prefixlen,
4315 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4316 RTF_UP | RTF_PREF(pref),
4317 .fc_protocol = RTPROT_RA,
4318 .fc_type = RTN_UNICAST,
4319 .fc_nlinfo.portid = 0,
4320 .fc_nlinfo.nlh = NULL,
4321 .fc_nlinfo.nl_net = net,
4322 };
4323
4324 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4325 cfg.fc_dst = *prefix;
4326 cfg.fc_gateway = *gwaddr;
4327
4328 /* We should treat it as a default route if prefix length is 0. */
4329 if (!prefixlen)
4330 cfg.fc_flags |= RTF_DEFAULT;
4331
4332 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4333
4334 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4335}
4336#endif
4337
4338struct fib6_info *rt6_get_dflt_router(struct net *net,
4339 const struct in6_addr *addr,
4340 struct net_device *dev)
4341{
4342 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4343 struct fib6_info *rt;
4344 struct fib6_table *table;
4345
4346 table = fib6_get_table(net, tb_id);
4347 if (!table)
4348 return NULL;
4349
4350 rcu_read_lock();
4351 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4352 struct fib6_nh *nh;
4353
4354 /* RA routes do not use nexthops */
4355 if (rt->nh)
4356 continue;
4357
4358 nh = rt->fib6_nh;
4359 if (dev == nh->fib_nh_dev &&
4360 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4361 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4362 break;
4363 }
4364 if (rt && !fib6_info_hold_safe(rt))
4365 rt = NULL;
4366 rcu_read_unlock();
4367 return rt;
4368}
4369
4370struct fib6_info *rt6_add_dflt_router(struct net *net,
4371 const struct in6_addr *gwaddr,
4372 struct net_device *dev,
4373 unsigned int pref,
4374 u32 defrtr_usr_metric,
4375 int lifetime)
4376{
4377 struct fib6_config cfg = {
4378 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4379 .fc_metric = defrtr_usr_metric,
4380 .fc_ifindex = dev->ifindex,
4381 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4382 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4383 .fc_protocol = RTPROT_RA,
4384 .fc_type = RTN_UNICAST,
4385 .fc_nlinfo.portid = 0,
4386 .fc_nlinfo.nlh = NULL,
4387 .fc_nlinfo.nl_net = net,
4388 .fc_expires = jiffies_to_clock_t(lifetime * HZ),
4389 };
4390
4391 cfg.fc_gateway = *gwaddr;
4392
4393 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4394 struct fib6_table *table;
4395
4396 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4397 if (table)
4398 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4399 }
4400
4401 return rt6_get_dflt_router(net, gwaddr, dev);
4402}
4403
4404static void __rt6_purge_dflt_routers(struct net *net,
4405 struct fib6_table *table)
4406{
4407 struct fib6_info *rt;
4408
4409restart:
4410 rcu_read_lock();
4411 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4412 struct net_device *dev = fib6_info_nh_dev(rt);
4413 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4414
4415 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4416 (!idev || idev->cnf.accept_ra != 2) &&
4417 fib6_info_hold_safe(rt)) {
4418 rcu_read_unlock();
4419 ip6_del_rt(net, rt, false);
4420 goto restart;
4421 }
4422 }
4423 rcu_read_unlock();
4424
4425 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4426}
4427
4428void rt6_purge_dflt_routers(struct net *net)
4429{
4430 struct fib6_table *table;
4431 struct hlist_head *head;
4432 unsigned int h;
4433
4434 rcu_read_lock();
4435
4436 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4437 head = &net->ipv6.fib_table_hash[h];
4438 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4439 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4440 __rt6_purge_dflt_routers(net, table);
4441 }
4442 }
4443
4444 rcu_read_unlock();
4445}
4446
4447static void rtmsg_to_fib6_config(struct net *net,
4448 struct in6_rtmsg *rtmsg,
4449 struct fib6_config *cfg)
4450{
4451 *cfg = (struct fib6_config){
4452 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4453 : RT6_TABLE_MAIN,
4454 .fc_ifindex = rtmsg->rtmsg_ifindex,
4455 .fc_metric = rtmsg->rtmsg_metric,
4456 .fc_expires = rtmsg->rtmsg_info,
4457 .fc_dst_len = rtmsg->rtmsg_dst_len,
4458 .fc_src_len = rtmsg->rtmsg_src_len,
4459 .fc_flags = rtmsg->rtmsg_flags,
4460 .fc_type = rtmsg->rtmsg_type,
4461
4462 .fc_nlinfo.nl_net = net,
4463
4464 .fc_dst = rtmsg->rtmsg_dst,
4465 .fc_src = rtmsg->rtmsg_src,
4466 .fc_gateway = rtmsg->rtmsg_gateway,
4467 };
4468}
4469
4470int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
4471{
4472 struct fib6_config cfg;
4473 int err;
4474
4475 if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4476 return -EINVAL;
4477 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4478 return -EPERM;
4479
4480 rtmsg_to_fib6_config(net, rtmsg, &cfg);
4481
4482 rtnl_lock();
4483 switch (cmd) {
4484 case SIOCADDRT:
4485 /* Only do the default setting of fc_metric in route adding */
4486 if (cfg.fc_metric == 0)
4487 cfg.fc_metric = IP6_RT_PRIO_USER;
4488 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4489 break;
4490 case SIOCDELRT:
4491 err = ip6_route_del(&cfg, NULL);
4492 break;
4493 }
4494 rtnl_unlock();
4495 return err;
4496}
4497
4498/*
4499 * Drop the packet on the floor
4500 */
4501
4502static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4503{
4504 struct dst_entry *dst = skb_dst(skb);
4505 struct net *net = dev_net(dst->dev);
4506 struct inet6_dev *idev;
4507 SKB_DR(reason);
4508 int type;
4509
4510 if (netif_is_l3_master(skb->dev) ||
4511 dst->dev == net->loopback_dev)
4512 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4513 else
4514 idev = ip6_dst_idev(dst);
4515
4516 switch (ipstats_mib_noroutes) {
4517 case IPSTATS_MIB_INNOROUTES:
4518 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4519 if (type == IPV6_ADDR_ANY) {
4520 SKB_DR_SET(reason, IP_INADDRERRORS);
4521 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4522 break;
4523 }
4524 SKB_DR_SET(reason, IP_INNOROUTES);
4525 fallthrough;
4526 case IPSTATS_MIB_OUTNOROUTES:
4527 SKB_DR_OR(reason, IP_OUTNOROUTES);
4528 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4529 break;
4530 }
4531
4532 /* Start over by dropping the dst for l3mdev case */
4533 if (netif_is_l3_master(skb->dev))
4534 skb_dst_drop(skb);
4535
4536 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4537 kfree_skb_reason(skb, reason);
4538 return 0;
4539}
4540
4541static int ip6_pkt_discard(struct sk_buff *skb)
4542{
4543 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4544}
4545
4546static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4547{
4548 skb->dev = skb_dst(skb)->dev;
4549 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4550}
4551
4552static int ip6_pkt_prohibit(struct sk_buff *skb)
4553{
4554 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4555}
4556
4557static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4558{
4559 skb->dev = skb_dst(skb)->dev;
4560 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4561}
4562
4563/*
4564 * Allocate a dst for local (unicast / anycast) address.
4565 */
4566
4567struct fib6_info *addrconf_f6i_alloc(struct net *net,
4568 struct inet6_dev *idev,
4569 const struct in6_addr *addr,
4570 bool anycast, gfp_t gfp_flags,
4571 struct netlink_ext_ack *extack)
4572{
4573 struct fib6_config cfg = {
4574 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4575 .fc_ifindex = idev->dev->ifindex,
4576 .fc_flags = RTF_UP | RTF_NONEXTHOP,
4577 .fc_dst = *addr,
4578 .fc_dst_len = 128,
4579 .fc_protocol = RTPROT_KERNEL,
4580 .fc_nlinfo.nl_net = net,
4581 .fc_ignore_dev_down = true,
4582 };
4583 struct fib6_info *f6i;
4584
4585 if (anycast) {
4586 cfg.fc_type = RTN_ANYCAST;
4587 cfg.fc_flags |= RTF_ANYCAST;
4588 } else {
4589 cfg.fc_type = RTN_LOCAL;
4590 cfg.fc_flags |= RTF_LOCAL;
4591 }
4592
4593 f6i = ip6_route_info_create(&cfg, gfp_flags, extack);
4594 if (!IS_ERR(f6i)) {
4595 f6i->dst_nocount = true;
4596
4597 if (!anycast &&
4598 (READ_ONCE(net->ipv6.devconf_all->disable_policy) ||
4599 READ_ONCE(idev->cnf.disable_policy)))
4600 f6i->dst_nopolicy = true;
4601 }
4602
4603 return f6i;
4604}
4605
4606/* remove deleted ip from prefsrc entries */
4607struct arg_dev_net_ip {
4608 struct net *net;
4609 struct in6_addr *addr;
4610};
4611
4612static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4613{
4614 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4615 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4616
4617 if (!rt->nh &&
4618 rt != net->ipv6.fib6_null_entry &&
4619 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr) &&
4620 !ipv6_chk_addr(net, addr, rt->fib6_nh->fib_nh_dev, 0)) {
4621 spin_lock_bh(&rt6_exception_lock);
4622 /* remove prefsrc entry */
4623 rt->fib6_prefsrc.plen = 0;
4624 spin_unlock_bh(&rt6_exception_lock);
4625 }
4626 return 0;
4627}
4628
4629void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4630{
4631 struct net *net = dev_net(ifp->idev->dev);
4632 struct arg_dev_net_ip adni = {
4633 .net = net,
4634 .addr = &ifp->addr,
4635 };
4636 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4637}
4638
4639#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
4640
4641/* Remove routers and update dst entries when gateway turn into host. */
4642static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4643{
4644 struct in6_addr *gateway = (struct in6_addr *)arg;
4645 struct fib6_nh *nh;
4646
4647 /* RA routes do not use nexthops */
4648 if (rt->nh)
4649 return 0;
4650
4651 nh = rt->fib6_nh;
4652 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4653 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4654 return -1;
4655
4656 /* Further clean up cached routes in exception table.
4657 * This is needed because cached route may have a different
4658 * gateway than its 'parent' in the case of an ip redirect.
4659 */
4660 fib6_nh_exceptions_clean_tohost(nh, gateway);
4661
4662 return 0;
4663}
4664
4665void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4666{
4667 fib6_clean_all(net, fib6_clean_tohost, gateway);
4668}
4669
4670struct arg_netdev_event {
4671 const struct net_device *dev;
4672 union {
4673 unsigned char nh_flags;
4674 unsigned long event;
4675 };
4676};
4677
4678static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4679{
4680 struct fib6_info *iter;
4681 struct fib6_node *fn;
4682
4683 fn = rcu_dereference_protected(rt->fib6_node,
4684 lockdep_is_held(&rt->fib6_table->tb6_lock));
4685 iter = rcu_dereference_protected(fn->leaf,
4686 lockdep_is_held(&rt->fib6_table->tb6_lock));
4687 while (iter) {
4688 if (iter->fib6_metric == rt->fib6_metric &&
4689 rt6_qualify_for_ecmp(iter))
4690 return iter;
4691 iter = rcu_dereference_protected(iter->fib6_next,
4692 lockdep_is_held(&rt->fib6_table->tb6_lock));
4693 }
4694
4695 return NULL;
4696}
4697
4698/* only called for fib entries with builtin fib6_nh */
4699static bool rt6_is_dead(const struct fib6_info *rt)
4700{
4701 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4702 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4703 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4704 return true;
4705
4706 return false;
4707}
4708
4709static int rt6_multipath_total_weight(const struct fib6_info *rt)
4710{
4711 struct fib6_info *iter;
4712 int total = 0;
4713
4714 if (!rt6_is_dead(rt))
4715 total += rt->fib6_nh->fib_nh_weight;
4716
4717 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4718 if (!rt6_is_dead(iter))
4719 total += iter->fib6_nh->fib_nh_weight;
4720 }
4721
4722 return total;
4723}
4724
4725static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4726{
4727 int upper_bound = -1;
4728
4729 if (!rt6_is_dead(rt)) {
4730 *weight += rt->fib6_nh->fib_nh_weight;
4731 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4732 total) - 1;
4733 }
4734 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4735}
4736
4737static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4738{
4739 struct fib6_info *iter;
4740 int weight = 0;
4741
4742 rt6_upper_bound_set(rt, &weight, total);
4743
4744 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4745 rt6_upper_bound_set(iter, &weight, total);
4746}
4747
4748void rt6_multipath_rebalance(struct fib6_info *rt)
4749{
4750 struct fib6_info *first;
4751 int total;
4752
4753 /* In case the entire multipath route was marked for flushing,
4754 * then there is no need to rebalance upon the removal of every
4755 * sibling route.
4756 */
4757 if (!rt->fib6_nsiblings || rt->should_flush)
4758 return;
4759
4760 /* During lookup routes are evaluated in order, so we need to
4761 * make sure upper bounds are assigned from the first sibling
4762 * onwards.
4763 */
4764 first = rt6_multipath_first_sibling(rt);
4765 if (WARN_ON_ONCE(!first))
4766 return;
4767
4768 total = rt6_multipath_total_weight(first);
4769 rt6_multipath_upper_bound_set(first, total);
4770}
4771
4772static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4773{
4774 const struct arg_netdev_event *arg = p_arg;
4775 struct net *net = dev_net(arg->dev);
4776
4777 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4778 rt->fib6_nh->fib_nh_dev == arg->dev) {
4779 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4780 fib6_update_sernum_upto_root(net, rt);
4781 rt6_multipath_rebalance(rt);
4782 }
4783
4784 return 0;
4785}
4786
4787void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4788{
4789 struct arg_netdev_event arg = {
4790 .dev = dev,
4791 {
4792 .nh_flags = nh_flags,
4793 },
4794 };
4795
4796 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4797 arg.nh_flags |= RTNH_F_LINKDOWN;
4798
4799 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4800}
4801
4802/* only called for fib entries with inline fib6_nh */
4803static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4804 const struct net_device *dev)
4805{
4806 struct fib6_info *iter;
4807
4808 if (rt->fib6_nh->fib_nh_dev == dev)
4809 return true;
4810 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4811 if (iter->fib6_nh->fib_nh_dev == dev)
4812 return true;
4813
4814 return false;
4815}
4816
4817static void rt6_multipath_flush(struct fib6_info *rt)
4818{
4819 struct fib6_info *iter;
4820
4821 rt->should_flush = 1;
4822 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4823 iter->should_flush = 1;
4824}
4825
4826static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4827 const struct net_device *down_dev)
4828{
4829 struct fib6_info *iter;
4830 unsigned int dead = 0;
4831
4832 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4833 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4834 dead++;
4835 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4836 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4837 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4838 dead++;
4839
4840 return dead;
4841}
4842
4843static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4844 const struct net_device *dev,
4845 unsigned char nh_flags)
4846{
4847 struct fib6_info *iter;
4848
4849 if (rt->fib6_nh->fib_nh_dev == dev)
4850 rt->fib6_nh->fib_nh_flags |= nh_flags;
4851 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4852 if (iter->fib6_nh->fib_nh_dev == dev)
4853 iter->fib6_nh->fib_nh_flags |= nh_flags;
4854}
4855
4856/* called with write lock held for table with rt */
4857static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4858{
4859 const struct arg_netdev_event *arg = p_arg;
4860 const struct net_device *dev = arg->dev;
4861 struct net *net = dev_net(dev);
4862
4863 if (rt == net->ipv6.fib6_null_entry || rt->nh)
4864 return 0;
4865
4866 switch (arg->event) {
4867 case NETDEV_UNREGISTER:
4868 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4869 case NETDEV_DOWN:
4870 if (rt->should_flush)
4871 return -1;
4872 if (!rt->fib6_nsiblings)
4873 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4874 if (rt6_multipath_uses_dev(rt, dev)) {
4875 unsigned int count;
4876
4877 count = rt6_multipath_dead_count(rt, dev);
4878 if (rt->fib6_nsiblings + 1 == count) {
4879 rt6_multipath_flush(rt);
4880 return -1;
4881 }
4882 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4883 RTNH_F_LINKDOWN);
4884 fib6_update_sernum(net, rt);
4885 rt6_multipath_rebalance(rt);
4886 }
4887 return -2;
4888 case NETDEV_CHANGE:
4889 if (rt->fib6_nh->fib_nh_dev != dev ||
4890 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4891 break;
4892 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4893 rt6_multipath_rebalance(rt);
4894 break;
4895 }
4896
4897 return 0;
4898}
4899
4900void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4901{
4902 struct arg_netdev_event arg = {
4903 .dev = dev,
4904 {
4905 .event = event,
4906 },
4907 };
4908 struct net *net = dev_net(dev);
4909
4910 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4911 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4912 else
4913 fib6_clean_all(net, fib6_ifdown, &arg);
4914}
4915
4916void rt6_disable_ip(struct net_device *dev, unsigned long event)
4917{
4918 rt6_sync_down_dev(dev, event);
4919 rt6_uncached_list_flush_dev(dev);
4920 neigh_ifdown(&nd_tbl, dev);
4921}
4922
4923struct rt6_mtu_change_arg {
4924 struct net_device *dev;
4925 unsigned int mtu;
4926 struct fib6_info *f6i;
4927};
4928
4929static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4930{
4931 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4932 struct fib6_info *f6i = arg->f6i;
4933
4934 /* For administrative MTU increase, there is no way to discover
4935 * IPv6 PMTU increase, so PMTU increase should be updated here.
4936 * Since RFC 1981 doesn't include administrative MTU increase
4937 * update PMTU increase is a MUST. (i.e. jumbo frame)
4938 */
4939 if (nh->fib_nh_dev == arg->dev) {
4940 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4941 u32 mtu = f6i->fib6_pmtu;
4942
4943 if (mtu >= arg->mtu ||
4944 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4945 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4946
4947 spin_lock_bh(&rt6_exception_lock);
4948 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4949 spin_unlock_bh(&rt6_exception_lock);
4950 }
4951
4952 return 0;
4953}
4954
4955static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4956{
4957 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4958 struct inet6_dev *idev;
4959
4960 /* In IPv6 pmtu discovery is not optional,
4961 so that RTAX_MTU lock cannot disable it.
4962 We still use this lock to block changes
4963 caused by addrconf/ndisc.
4964 */
4965
4966 idev = __in6_dev_get(arg->dev);
4967 if (!idev)
4968 return 0;
4969
4970 if (fib6_metric_locked(f6i, RTAX_MTU))
4971 return 0;
4972
4973 arg->f6i = f6i;
4974 if (f6i->nh) {
4975 /* fib6_nh_mtu_change only returns 0, so this is safe */
4976 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4977 arg);
4978 }
4979
4980 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4981}
4982
4983void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4984{
4985 struct rt6_mtu_change_arg arg = {
4986 .dev = dev,
4987 .mtu = mtu,
4988 };
4989
4990 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4991}
4992
4993static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4994 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
4995 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
4996 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
4997 [RTA_OIF] = { .type = NLA_U32 },
4998 [RTA_IIF] = { .type = NLA_U32 },
4999 [RTA_PRIORITY] = { .type = NLA_U32 },
5000 [RTA_METRICS] = { .type = NLA_NESTED },
5001 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
5002 [RTA_PREF] = { .type = NLA_U8 },
5003 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
5004 [RTA_ENCAP] = { .type = NLA_NESTED },
5005 [RTA_EXPIRES] = { .type = NLA_U32 },
5006 [RTA_UID] = { .type = NLA_U32 },
5007 [RTA_MARK] = { .type = NLA_U32 },
5008 [RTA_TABLE] = { .type = NLA_U32 },
5009 [RTA_IP_PROTO] = { .type = NLA_U8 },
5010 [RTA_SPORT] = { .type = NLA_U16 },
5011 [RTA_DPORT] = { .type = NLA_U16 },
5012 [RTA_NH_ID] = { .type = NLA_U32 },
5013};
5014
5015static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
5016 struct fib6_config *cfg,
5017 struct netlink_ext_ack *extack)
5018{
5019 struct rtmsg *rtm;
5020 struct nlattr *tb[RTA_MAX+1];
5021 unsigned int pref;
5022 int err;
5023
5024 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5025 rtm_ipv6_policy, extack);
5026 if (err < 0)
5027 goto errout;
5028
5029 err = -EINVAL;
5030 rtm = nlmsg_data(nlh);
5031
5032 if (rtm->rtm_tos) {
5033 NL_SET_ERR_MSG(extack,
5034 "Invalid dsfield (tos): option not available for IPv6");
5035 goto errout;
5036 }
5037
5038 *cfg = (struct fib6_config){
5039 .fc_table = rtm->rtm_table,
5040 .fc_dst_len = rtm->rtm_dst_len,
5041 .fc_src_len = rtm->rtm_src_len,
5042 .fc_flags = RTF_UP,
5043 .fc_protocol = rtm->rtm_protocol,
5044 .fc_type = rtm->rtm_type,
5045
5046 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
5047 .fc_nlinfo.nlh = nlh,
5048 .fc_nlinfo.nl_net = sock_net(skb->sk),
5049 };
5050
5051 if (rtm->rtm_type == RTN_UNREACHABLE ||
5052 rtm->rtm_type == RTN_BLACKHOLE ||
5053 rtm->rtm_type == RTN_PROHIBIT ||
5054 rtm->rtm_type == RTN_THROW)
5055 cfg->fc_flags |= RTF_REJECT;
5056
5057 if (rtm->rtm_type == RTN_LOCAL)
5058 cfg->fc_flags |= RTF_LOCAL;
5059
5060 if (rtm->rtm_flags & RTM_F_CLONED)
5061 cfg->fc_flags |= RTF_CACHE;
5062
5063 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
5064
5065 if (tb[RTA_NH_ID]) {
5066 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
5067 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
5068 NL_SET_ERR_MSG(extack,
5069 "Nexthop specification and nexthop id are mutually exclusive");
5070 goto errout;
5071 }
5072 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
5073 }
5074
5075 if (tb[RTA_GATEWAY]) {
5076 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
5077 cfg->fc_flags |= RTF_GATEWAY;
5078 }
5079 if (tb[RTA_VIA]) {
5080 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
5081 goto errout;
5082 }
5083
5084 if (tb[RTA_DST]) {
5085 int plen = (rtm->rtm_dst_len + 7) >> 3;
5086
5087 if (nla_len(tb[RTA_DST]) < plen)
5088 goto errout;
5089
5090 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
5091 }
5092
5093 if (tb[RTA_SRC]) {
5094 int plen = (rtm->rtm_src_len + 7) >> 3;
5095
5096 if (nla_len(tb[RTA_SRC]) < plen)
5097 goto errout;
5098
5099 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
5100 }
5101
5102 if (tb[RTA_PREFSRC])
5103 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
5104
5105 if (tb[RTA_OIF])
5106 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
5107
5108 if (tb[RTA_PRIORITY])
5109 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
5110
5111 if (tb[RTA_METRICS]) {
5112 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
5113 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
5114 }
5115
5116 if (tb[RTA_TABLE])
5117 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
5118
5119 if (tb[RTA_MULTIPATH]) {
5120 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
5121 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
5122
5123 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
5124 cfg->fc_mp_len, extack);
5125 if (err < 0)
5126 goto errout;
5127 }
5128
5129 if (tb[RTA_PREF]) {
5130 pref = nla_get_u8(tb[RTA_PREF]);
5131 if (pref != ICMPV6_ROUTER_PREF_LOW &&
5132 pref != ICMPV6_ROUTER_PREF_HIGH)
5133 pref = ICMPV6_ROUTER_PREF_MEDIUM;
5134 cfg->fc_flags |= RTF_PREF(pref);
5135 }
5136
5137 if (tb[RTA_ENCAP])
5138 cfg->fc_encap = tb[RTA_ENCAP];
5139
5140 if (tb[RTA_ENCAP_TYPE]) {
5141 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5142
5143 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
5144 if (err < 0)
5145 goto errout;
5146 }
5147
5148 if (tb[RTA_EXPIRES]) {
5149 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5150
5151 if (addrconf_finite_timeout(timeout)) {
5152 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5153 cfg->fc_flags |= RTF_EXPIRES;
5154 }
5155 }
5156
5157 err = 0;
5158errout:
5159 return err;
5160}
5161
5162struct rt6_nh {
5163 struct fib6_info *fib6_info;
5164 struct fib6_config r_cfg;
5165 struct list_head next;
5166};
5167
5168static int ip6_route_info_append(struct net *net,
5169 struct list_head *rt6_nh_list,
5170 struct fib6_info *rt,
5171 struct fib6_config *r_cfg)
5172{
5173 struct rt6_nh *nh;
5174 int err = -EEXIST;
5175
5176 list_for_each_entry(nh, rt6_nh_list, next) {
5177 /* check if fib6_info already exists */
5178 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
5179 return err;
5180 }
5181
5182 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5183 if (!nh)
5184 return -ENOMEM;
5185 nh->fib6_info = rt;
5186 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5187 list_add_tail(&nh->next, rt6_nh_list);
5188
5189 return 0;
5190}
5191
5192static void ip6_route_mpath_notify(struct fib6_info *rt,
5193 struct fib6_info *rt_last,
5194 struct nl_info *info,
5195 __u16 nlflags)
5196{
5197 /* if this is an APPEND route, then rt points to the first route
5198 * inserted and rt_last points to last route inserted. Userspace
5199 * wants a consistent dump of the route which starts at the first
5200 * nexthop. Since sibling routes are always added at the end of
5201 * the list, find the first sibling of the last route appended
5202 */
5203 rcu_read_lock();
5204
5205 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5206 rt = list_first_or_null_rcu(&rt_last->fib6_siblings,
5207 struct fib6_info,
5208 fib6_siblings);
5209 }
5210
5211 if (rt)
5212 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5213
5214 rcu_read_unlock();
5215}
5216
5217static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5218{
5219 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5220 bool should_notify = false;
5221 struct fib6_info *leaf;
5222 struct fib6_node *fn;
5223
5224 rcu_read_lock();
5225 fn = rcu_dereference(rt->fib6_node);
5226 if (!fn)
5227 goto out;
5228
5229 leaf = rcu_dereference(fn->leaf);
5230 if (!leaf)
5231 goto out;
5232
5233 if (rt == leaf ||
5234 (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5235 rt6_qualify_for_ecmp(leaf)))
5236 should_notify = true;
5237out:
5238 rcu_read_unlock();
5239
5240 return should_notify;
5241}
5242
5243static int fib6_gw_from_attr(struct in6_addr *gw, struct nlattr *nla,
5244 struct netlink_ext_ack *extack)
5245{
5246 if (nla_len(nla) < sizeof(*gw)) {
5247 NL_SET_ERR_MSG(extack, "Invalid IPv6 address in RTA_GATEWAY");
5248 return -EINVAL;
5249 }
5250
5251 *gw = nla_get_in6_addr(nla);
5252
5253 return 0;
5254}
5255
5256static int ip6_route_multipath_add(struct fib6_config *cfg,
5257 struct netlink_ext_ack *extack)
5258{
5259 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5260 struct nl_info *info = &cfg->fc_nlinfo;
5261 struct fib6_config r_cfg;
5262 struct rtnexthop *rtnh;
5263 struct fib6_info *rt;
5264 struct rt6_nh *err_nh;
5265 struct rt6_nh *nh, *nh_safe;
5266 __u16 nlflags;
5267 int remaining;
5268 int attrlen;
5269 int err = 1;
5270 int nhn = 0;
5271 int replace = (cfg->fc_nlinfo.nlh &&
5272 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5273 LIST_HEAD(rt6_nh_list);
5274
5275 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5276 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5277 nlflags |= NLM_F_APPEND;
5278
5279 remaining = cfg->fc_mp_len;
5280 rtnh = (struct rtnexthop *)cfg->fc_mp;
5281
5282 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
5283 * fib6_info structs per nexthop
5284 */
5285 while (rtnh_ok(rtnh, remaining)) {
5286 memcpy(&r_cfg, cfg, sizeof(*cfg));
5287 if (rtnh->rtnh_ifindex)
5288 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5289
5290 attrlen = rtnh_attrlen(rtnh);
5291 if (attrlen > 0) {
5292 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5293
5294 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5295 if (nla) {
5296 err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5297 extack);
5298 if (err)
5299 goto cleanup;
5300
5301 r_cfg.fc_flags |= RTF_GATEWAY;
5302 }
5303 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5304
5305 /* RTA_ENCAP_TYPE length checked in
5306 * lwtunnel_valid_encap_type_attr
5307 */
5308 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5309 if (nla)
5310 r_cfg.fc_encap_type = nla_get_u16(nla);
5311 }
5312
5313 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5314 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5315 if (IS_ERR(rt)) {
5316 err = PTR_ERR(rt);
5317 rt = NULL;
5318 goto cleanup;
5319 }
5320 if (!rt6_qualify_for_ecmp(rt)) {
5321 err = -EINVAL;
5322 NL_SET_ERR_MSG(extack,
5323 "Device only routes can not be added for IPv6 using the multipath API.");
5324 fib6_info_release(rt);
5325 goto cleanup;
5326 }
5327
5328 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5329
5330 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5331 rt, &r_cfg);
5332 if (err) {
5333 fib6_info_release(rt);
5334 goto cleanup;
5335 }
5336
5337 rtnh = rtnh_next(rtnh, &remaining);
5338 }
5339
5340 if (list_empty(&rt6_nh_list)) {
5341 NL_SET_ERR_MSG(extack,
5342 "Invalid nexthop configuration - no valid nexthops");
5343 return -EINVAL;
5344 }
5345
5346 /* for add and replace send one notification with all nexthops.
5347 * Skip the notification in fib6_add_rt2node and send one with
5348 * the full route when done
5349 */
5350 info->skip_notify = 1;
5351
5352 /* For add and replace, send one notification with all nexthops. For
5353 * append, send one notification with all appended nexthops.
5354 */
5355 info->skip_notify_kernel = 1;
5356
5357 err_nh = NULL;
5358 list_for_each_entry(nh, &rt6_nh_list, next) {
5359 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5360
5361 if (err) {
5362 if (replace && nhn)
5363 NL_SET_ERR_MSG_MOD(extack,
5364 "multipath route replace failed (check consistency of installed routes)");
5365 err_nh = nh;
5366 goto add_errout;
5367 }
5368 /* save reference to last route successfully inserted */
5369 rt_last = nh->fib6_info;
5370
5371 /* save reference to first route for notification */
5372 if (!rt_notif)
5373 rt_notif = nh->fib6_info;
5374
5375 /* Because each route is added like a single route we remove
5376 * these flags after the first nexthop: if there is a collision,
5377 * we have already failed to add the first nexthop:
5378 * fib6_add_rt2node() has rejected it; when replacing, old
5379 * nexthops have been replaced by first new, the rest should
5380 * be added to it.
5381 */
5382 if (cfg->fc_nlinfo.nlh) {
5383 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5384 NLM_F_REPLACE);
5385 cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
5386 }
5387 nhn++;
5388 }
5389
5390 /* An in-kernel notification should only be sent in case the new
5391 * multipath route is added as the first route in the node, or if
5392 * it was appended to it. We pass 'rt_notif' since it is the first
5393 * sibling and might allow us to skip some checks in the replace case.
5394 */
5395 if (ip6_route_mpath_should_notify(rt_notif)) {
5396 enum fib_event_type fib_event;
5397
5398 if (rt_notif->fib6_nsiblings != nhn - 1)
5399 fib_event = FIB_EVENT_ENTRY_APPEND;
5400 else
5401 fib_event = FIB_EVENT_ENTRY_REPLACE;
5402
5403 err = call_fib6_multipath_entry_notifiers(info->nl_net,
5404 fib_event, rt_notif,
5405 nhn - 1, extack);
5406 if (err) {
5407 /* Delete all the siblings that were just added */
5408 err_nh = NULL;
5409 goto add_errout;
5410 }
5411 }
5412
5413 /* success ... tell user about new route */
5414 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5415 goto cleanup;
5416
5417add_errout:
5418 /* send notification for routes that were added so that
5419 * the delete notifications sent by ip6_route_del are
5420 * coherent
5421 */
5422 if (rt_notif)
5423 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5424
5425 /* Delete routes that were already added */
5426 list_for_each_entry(nh, &rt6_nh_list, next) {
5427 if (err_nh == nh)
5428 break;
5429 ip6_route_del(&nh->r_cfg, extack);
5430 }
5431
5432cleanup:
5433 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5434 fib6_info_release(nh->fib6_info);
5435 list_del(&nh->next);
5436 kfree(nh);
5437 }
5438
5439 return err;
5440}
5441
5442static int ip6_route_multipath_del(struct fib6_config *cfg,
5443 struct netlink_ext_ack *extack)
5444{
5445 struct fib6_config r_cfg;
5446 struct rtnexthop *rtnh;
5447 int last_err = 0;
5448 int remaining;
5449 int attrlen;
5450 int err;
5451
5452 remaining = cfg->fc_mp_len;
5453 rtnh = (struct rtnexthop *)cfg->fc_mp;
5454
5455 /* Parse a Multipath Entry */
5456 while (rtnh_ok(rtnh, remaining)) {
5457 memcpy(&r_cfg, cfg, sizeof(*cfg));
5458 if (rtnh->rtnh_ifindex)
5459 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5460
5461 attrlen = rtnh_attrlen(rtnh);
5462 if (attrlen > 0) {
5463 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5464
5465 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5466 if (nla) {
5467 err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5468 extack);
5469 if (err) {
5470 last_err = err;
5471 goto next_rtnh;
5472 }
5473
5474 r_cfg.fc_flags |= RTF_GATEWAY;
5475 }
5476 }
5477 err = ip6_route_del(&r_cfg, extack);
5478 if (err)
5479 last_err = err;
5480
5481next_rtnh:
5482 rtnh = rtnh_next(rtnh, &remaining);
5483 }
5484
5485 return last_err;
5486}
5487
5488static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5489 struct netlink_ext_ack *extack)
5490{
5491 struct fib6_config cfg;
5492 int err;
5493
5494 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5495 if (err < 0)
5496 return err;
5497
5498 if (cfg.fc_nh_id &&
5499 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5500 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5501 return -EINVAL;
5502 }
5503
5504 if (cfg.fc_mp)
5505 return ip6_route_multipath_del(&cfg, extack);
5506 else {
5507 cfg.fc_delete_all_nh = 1;
5508 return ip6_route_del(&cfg, extack);
5509 }
5510}
5511
5512static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5513 struct netlink_ext_ack *extack)
5514{
5515 struct fib6_config cfg;
5516 int err;
5517
5518 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5519 if (err < 0)
5520 return err;
5521
5522 if (cfg.fc_metric == 0)
5523 cfg.fc_metric = IP6_RT_PRIO_USER;
5524
5525 if (cfg.fc_mp)
5526 return ip6_route_multipath_add(&cfg, extack);
5527 else
5528 return ip6_route_add(&cfg, GFP_KERNEL, extack);
5529}
5530
5531/* add the overhead of this fib6_nh to nexthop_len */
5532static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5533{
5534 int *nexthop_len = arg;
5535
5536 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5537 + NLA_ALIGN(sizeof(struct rtnexthop))
5538 + nla_total_size(16); /* RTA_GATEWAY */
5539
5540 if (nh->fib_nh_lws) {
5541 /* RTA_ENCAP_TYPE */
5542 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5543 /* RTA_ENCAP */
5544 *nexthop_len += nla_total_size(2);
5545 }
5546
5547 return 0;
5548}
5549
5550static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5551{
5552 int nexthop_len;
5553
5554 if (f6i->nh) {
5555 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5556 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5557 &nexthop_len);
5558 } else {
5559 struct fib6_nh *nh = f6i->fib6_nh;
5560 struct fib6_info *sibling;
5561
5562 nexthop_len = 0;
5563 if (f6i->fib6_nsiblings) {
5564 rt6_nh_nlmsg_size(nh, &nexthop_len);
5565
5566 rcu_read_lock();
5567
5568 list_for_each_entry_rcu(sibling, &f6i->fib6_siblings,
5569 fib6_siblings) {
5570 rt6_nh_nlmsg_size(sibling->fib6_nh, &nexthop_len);
5571 }
5572
5573 rcu_read_unlock();
5574 }
5575 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5576 }
5577
5578 return NLMSG_ALIGN(sizeof(struct rtmsg))
5579 + nla_total_size(16) /* RTA_SRC */
5580 + nla_total_size(16) /* RTA_DST */
5581 + nla_total_size(16) /* RTA_GATEWAY */
5582 + nla_total_size(16) /* RTA_PREFSRC */
5583 + nla_total_size(4) /* RTA_TABLE */
5584 + nla_total_size(4) /* RTA_IIF */
5585 + nla_total_size(4) /* RTA_OIF */
5586 + nla_total_size(4) /* RTA_PRIORITY */
5587 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5588 + nla_total_size(sizeof(struct rta_cacheinfo))
5589 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5590 + nla_total_size(1) /* RTA_PREF */
5591 + nexthop_len;
5592}
5593
5594static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5595 unsigned char *flags)
5596{
5597 if (nexthop_is_multipath(nh)) {
5598 struct nlattr *mp;
5599
5600 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5601 if (!mp)
5602 goto nla_put_failure;
5603
5604 if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
5605 goto nla_put_failure;
5606
5607 nla_nest_end(skb, mp);
5608 } else {
5609 struct fib6_nh *fib6_nh;
5610
5611 fib6_nh = nexthop_fib6_nh(nh);
5612 if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
5613 flags, false) < 0)
5614 goto nla_put_failure;
5615 }
5616
5617 return 0;
5618
5619nla_put_failure:
5620 return -EMSGSIZE;
5621}
5622
5623static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5624 struct fib6_info *rt, struct dst_entry *dst,
5625 struct in6_addr *dest, struct in6_addr *src,
5626 int iif, int type, u32 portid, u32 seq,
5627 unsigned int flags)
5628{
5629 struct rt6_info *rt6 = dst_rt6_info(dst);
5630 struct rt6key *rt6_dst, *rt6_src;
5631 u32 *pmetrics, table, rt6_flags;
5632 unsigned char nh_flags = 0;
5633 struct nlmsghdr *nlh;
5634 struct rtmsg *rtm;
5635 long expires = 0;
5636
5637 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5638 if (!nlh)
5639 return -EMSGSIZE;
5640
5641 if (rt6) {
5642 rt6_dst = &rt6->rt6i_dst;
5643 rt6_src = &rt6->rt6i_src;
5644 rt6_flags = rt6->rt6i_flags;
5645 } else {
5646 rt6_dst = &rt->fib6_dst;
5647 rt6_src = &rt->fib6_src;
5648 rt6_flags = rt->fib6_flags;
5649 }
5650
5651 rtm = nlmsg_data(nlh);
5652 rtm->rtm_family = AF_INET6;
5653 rtm->rtm_dst_len = rt6_dst->plen;
5654 rtm->rtm_src_len = rt6_src->plen;
5655 rtm->rtm_tos = 0;
5656 if (rt->fib6_table)
5657 table = rt->fib6_table->tb6_id;
5658 else
5659 table = RT6_TABLE_UNSPEC;
5660 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5661 if (nla_put_u32(skb, RTA_TABLE, table))
5662 goto nla_put_failure;
5663
5664 rtm->rtm_type = rt->fib6_type;
5665 rtm->rtm_flags = 0;
5666 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5667 rtm->rtm_protocol = rt->fib6_protocol;
5668
5669 if (rt6_flags & RTF_CACHE)
5670 rtm->rtm_flags |= RTM_F_CLONED;
5671
5672 if (dest) {
5673 if (nla_put_in6_addr(skb, RTA_DST, dest))
5674 goto nla_put_failure;
5675 rtm->rtm_dst_len = 128;
5676 } else if (rtm->rtm_dst_len)
5677 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5678 goto nla_put_failure;
5679#ifdef CONFIG_IPV6_SUBTREES
5680 if (src) {
5681 if (nla_put_in6_addr(skb, RTA_SRC, src))
5682 goto nla_put_failure;
5683 rtm->rtm_src_len = 128;
5684 } else if (rtm->rtm_src_len &&
5685 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5686 goto nla_put_failure;
5687#endif
5688 if (iif) {
5689#ifdef CONFIG_IPV6_MROUTE
5690 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5691 int err = ip6mr_get_route(net, skb, rtm, portid);
5692
5693 if (err == 0)
5694 return 0;
5695 if (err < 0)
5696 goto nla_put_failure;
5697 } else
5698#endif
5699 if (nla_put_u32(skb, RTA_IIF, iif))
5700 goto nla_put_failure;
5701 } else if (dest) {
5702 struct in6_addr saddr_buf;
5703 if (ip6_route_get_saddr(net, rt, dest, 0, 0, &saddr_buf) == 0 &&
5704 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5705 goto nla_put_failure;
5706 }
5707
5708 if (rt->fib6_prefsrc.plen) {
5709 struct in6_addr saddr_buf;
5710 saddr_buf = rt->fib6_prefsrc.addr;
5711 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5712 goto nla_put_failure;
5713 }
5714
5715 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5716 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5717 goto nla_put_failure;
5718
5719 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5720 goto nla_put_failure;
5721
5722 /* For multipath routes, walk the siblings list and add
5723 * each as a nexthop within RTA_MULTIPATH.
5724 */
5725 if (rt6) {
5726 if (rt6_flags & RTF_GATEWAY &&
5727 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5728 goto nla_put_failure;
5729
5730 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5731 goto nla_put_failure;
5732
5733 if (dst->lwtstate &&
5734 lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
5735 goto nla_put_failure;
5736 } else if (rt->fib6_nsiblings) {
5737 struct fib6_info *sibling;
5738 struct nlattr *mp;
5739
5740 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5741 if (!mp)
5742 goto nla_put_failure;
5743
5744 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5745 rt->fib6_nh->fib_nh_weight, AF_INET6,
5746 0) < 0)
5747 goto nla_put_failure;
5748
5749 rcu_read_lock();
5750
5751 list_for_each_entry_rcu(sibling, &rt->fib6_siblings,
5752 fib6_siblings) {
5753 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5754 sibling->fib6_nh->fib_nh_weight,
5755 AF_INET6, 0) < 0) {
5756 rcu_read_unlock();
5757
5758 goto nla_put_failure;
5759 }
5760 }
5761
5762 rcu_read_unlock();
5763
5764 nla_nest_end(skb, mp);
5765 } else if (rt->nh) {
5766 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5767 goto nla_put_failure;
5768
5769 if (nexthop_is_blackhole(rt->nh))
5770 rtm->rtm_type = RTN_BLACKHOLE;
5771
5772 if (READ_ONCE(net->ipv4.sysctl_nexthop_compat_mode) &&
5773 rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5774 goto nla_put_failure;
5775
5776 rtm->rtm_flags |= nh_flags;
5777 } else {
5778 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
5779 &nh_flags, false) < 0)
5780 goto nla_put_failure;
5781
5782 rtm->rtm_flags |= nh_flags;
5783 }
5784
5785 if (rt6_flags & RTF_EXPIRES) {
5786 expires = dst ? dst->expires : rt->expires;
5787 expires -= jiffies;
5788 }
5789
5790 if (!dst) {
5791 if (READ_ONCE(rt->offload))
5792 rtm->rtm_flags |= RTM_F_OFFLOAD;
5793 if (READ_ONCE(rt->trap))
5794 rtm->rtm_flags |= RTM_F_TRAP;
5795 if (READ_ONCE(rt->offload_failed))
5796 rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
5797 }
5798
5799 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5800 goto nla_put_failure;
5801
5802 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5803 goto nla_put_failure;
5804
5805
5806 nlmsg_end(skb, nlh);
5807 return 0;
5808
5809nla_put_failure:
5810 nlmsg_cancel(skb, nlh);
5811 return -EMSGSIZE;
5812}
5813
5814static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5815{
5816 const struct net_device *dev = arg;
5817
5818 if (nh->fib_nh_dev == dev)
5819 return 1;
5820
5821 return 0;
5822}
5823
5824static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5825 const struct net_device *dev)
5826{
5827 if (f6i->nh) {
5828 struct net_device *_dev = (struct net_device *)dev;
5829
5830 return !!nexthop_for_each_fib6_nh(f6i->nh,
5831 fib6_info_nh_uses_dev,
5832 _dev);
5833 }
5834
5835 if (f6i->fib6_nh->fib_nh_dev == dev)
5836 return true;
5837
5838 if (f6i->fib6_nsiblings) {
5839 struct fib6_info *sibling, *next_sibling;
5840
5841 list_for_each_entry_safe(sibling, next_sibling,
5842 &f6i->fib6_siblings, fib6_siblings) {
5843 if (sibling->fib6_nh->fib_nh_dev == dev)
5844 return true;
5845 }
5846 }
5847
5848 return false;
5849}
5850
5851struct fib6_nh_exception_dump_walker {
5852 struct rt6_rtnl_dump_arg *dump;
5853 struct fib6_info *rt;
5854 unsigned int flags;
5855 unsigned int skip;
5856 unsigned int count;
5857};
5858
5859static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5860{
5861 struct fib6_nh_exception_dump_walker *w = arg;
5862 struct rt6_rtnl_dump_arg *dump = w->dump;
5863 struct rt6_exception_bucket *bucket;
5864 struct rt6_exception *rt6_ex;
5865 int i, err;
5866
5867 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5868 if (!bucket)
5869 return 0;
5870
5871 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5872 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5873 if (w->skip) {
5874 w->skip--;
5875 continue;
5876 }
5877
5878 /* Expiration of entries doesn't bump sernum, insertion
5879 * does. Removal is triggered by insertion, so we can
5880 * rely on the fact that if entries change between two
5881 * partial dumps, this node is scanned again completely,
5882 * see rt6_insert_exception() and fib6_dump_table().
5883 *
5884 * Count expired entries we go through as handled
5885 * entries that we'll skip next time, in case of partial
5886 * node dump. Otherwise, if entries expire meanwhile,
5887 * we'll skip the wrong amount.
5888 */
5889 if (rt6_check_expired(rt6_ex->rt6i)) {
5890 w->count++;
5891 continue;
5892 }
5893
5894 err = rt6_fill_node(dump->net, dump->skb, w->rt,
5895 &rt6_ex->rt6i->dst, NULL, NULL, 0,
5896 RTM_NEWROUTE,
5897 NETLINK_CB(dump->cb->skb).portid,
5898 dump->cb->nlh->nlmsg_seq, w->flags);
5899 if (err)
5900 return err;
5901
5902 w->count++;
5903 }
5904 bucket++;
5905 }
5906
5907 return 0;
5908}
5909
5910/* Return -1 if done with node, number of handled routes on partial dump */
5911int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5912{
5913 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5914 struct fib_dump_filter *filter = &arg->filter;
5915 unsigned int flags = NLM_F_MULTI;
5916 struct net *net = arg->net;
5917 int count = 0;
5918
5919 if (rt == net->ipv6.fib6_null_entry)
5920 return -1;
5921
5922 if ((filter->flags & RTM_F_PREFIX) &&
5923 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5924 /* success since this is not a prefix route */
5925 return -1;
5926 }
5927 if (filter->filter_set &&
5928 ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5929 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5930 (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5931 return -1;
5932 }
5933
5934 if (filter->filter_set ||
5935 !filter->dump_routes || !filter->dump_exceptions) {
5936 flags |= NLM_F_DUMP_FILTERED;
5937 }
5938
5939 if (filter->dump_routes) {
5940 if (skip) {
5941 skip--;
5942 } else {
5943 if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5944 0, RTM_NEWROUTE,
5945 NETLINK_CB(arg->cb->skb).portid,
5946 arg->cb->nlh->nlmsg_seq, flags)) {
5947 return 0;
5948 }
5949 count++;
5950 }
5951 }
5952
5953 if (filter->dump_exceptions) {
5954 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5955 .rt = rt,
5956 .flags = flags,
5957 .skip = skip,
5958 .count = 0 };
5959 int err;
5960
5961 rcu_read_lock();
5962 if (rt->nh) {
5963 err = nexthop_for_each_fib6_nh(rt->nh,
5964 rt6_nh_dump_exceptions,
5965 &w);
5966 } else {
5967 err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5968 }
5969 rcu_read_unlock();
5970
5971 if (err)
5972 return count + w.count;
5973 }
5974
5975 return -1;
5976}
5977
5978static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5979 const struct nlmsghdr *nlh,
5980 struct nlattr **tb,
5981 struct netlink_ext_ack *extack)
5982{
5983 struct rtmsg *rtm;
5984 int i, err;
5985
5986 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5987 NL_SET_ERR_MSG_MOD(extack,
5988 "Invalid header for get route request");
5989 return -EINVAL;
5990 }
5991
5992 if (!netlink_strict_get_check(skb))
5993 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5994 rtm_ipv6_policy, extack);
5995
5996 rtm = nlmsg_data(nlh);
5997 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5998 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5999 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
6000 rtm->rtm_type) {
6001 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
6002 return -EINVAL;
6003 }
6004 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
6005 NL_SET_ERR_MSG_MOD(extack,
6006 "Invalid flags for get route request");
6007 return -EINVAL;
6008 }
6009
6010 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
6011 rtm_ipv6_policy, extack);
6012 if (err)
6013 return err;
6014
6015 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
6016 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
6017 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
6018 return -EINVAL;
6019 }
6020
6021 for (i = 0; i <= RTA_MAX; i++) {
6022 if (!tb[i])
6023 continue;
6024
6025 switch (i) {
6026 case RTA_SRC:
6027 case RTA_DST:
6028 case RTA_IIF:
6029 case RTA_OIF:
6030 case RTA_MARK:
6031 case RTA_UID:
6032 case RTA_SPORT:
6033 case RTA_DPORT:
6034 case RTA_IP_PROTO:
6035 break;
6036 default:
6037 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
6038 return -EINVAL;
6039 }
6040 }
6041
6042 return 0;
6043}
6044
6045static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
6046 struct netlink_ext_ack *extack)
6047{
6048 struct net *net = sock_net(in_skb->sk);
6049 struct nlattr *tb[RTA_MAX+1];
6050 int err, iif = 0, oif = 0;
6051 struct fib6_info *from;
6052 struct dst_entry *dst;
6053 struct rt6_info *rt;
6054 struct sk_buff *skb;
6055 struct rtmsg *rtm;
6056 struct flowi6 fl6 = {};
6057 bool fibmatch;
6058
6059 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
6060 if (err < 0)
6061 goto errout;
6062
6063 err = -EINVAL;
6064 rtm = nlmsg_data(nlh);
6065 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
6066 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
6067
6068 if (tb[RTA_SRC]) {
6069 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
6070 goto errout;
6071
6072 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
6073 }
6074
6075 if (tb[RTA_DST]) {
6076 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
6077 goto errout;
6078
6079 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
6080 }
6081
6082 if (tb[RTA_IIF])
6083 iif = nla_get_u32(tb[RTA_IIF]);
6084
6085 if (tb[RTA_OIF])
6086 oif = nla_get_u32(tb[RTA_OIF]);
6087
6088 if (tb[RTA_MARK])
6089 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
6090
6091 if (tb[RTA_UID])
6092 fl6.flowi6_uid = make_kuid(current_user_ns(),
6093 nla_get_u32(tb[RTA_UID]));
6094 else
6095 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
6096
6097 if (tb[RTA_SPORT])
6098 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
6099
6100 if (tb[RTA_DPORT])
6101 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
6102
6103 if (tb[RTA_IP_PROTO]) {
6104 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
6105 &fl6.flowi6_proto, AF_INET6,
6106 extack);
6107 if (err)
6108 goto errout;
6109 }
6110
6111 if (iif) {
6112 struct net_device *dev;
6113 int flags = 0;
6114
6115 rcu_read_lock();
6116
6117 dev = dev_get_by_index_rcu(net, iif);
6118 if (!dev) {
6119 rcu_read_unlock();
6120 err = -ENODEV;
6121 goto errout;
6122 }
6123
6124 fl6.flowi6_iif = iif;
6125
6126 if (!ipv6_addr_any(&fl6.saddr))
6127 flags |= RT6_LOOKUP_F_HAS_SADDR;
6128
6129 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
6130
6131 rcu_read_unlock();
6132 } else {
6133 fl6.flowi6_oif = oif;
6134
6135 dst = ip6_route_output(net, NULL, &fl6);
6136 }
6137
6138
6139 rt = dst_rt6_info(dst);
6140 if (rt->dst.error) {
6141 err = rt->dst.error;
6142 ip6_rt_put(rt);
6143 goto errout;
6144 }
6145
6146 if (rt == net->ipv6.ip6_null_entry) {
6147 err = rt->dst.error;
6148 ip6_rt_put(rt);
6149 goto errout;
6150 }
6151
6152 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
6153 if (!skb) {
6154 ip6_rt_put(rt);
6155 err = -ENOBUFS;
6156 goto errout;
6157 }
6158
6159 skb_dst_set(skb, &rt->dst);
6160
6161 rcu_read_lock();
6162 from = rcu_dereference(rt->from);
6163 if (from) {
6164 if (fibmatch)
6165 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
6166 iif, RTM_NEWROUTE,
6167 NETLINK_CB(in_skb).portid,
6168 nlh->nlmsg_seq, 0);
6169 else
6170 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
6171 &fl6.saddr, iif, RTM_NEWROUTE,
6172 NETLINK_CB(in_skb).portid,
6173 nlh->nlmsg_seq, 0);
6174 } else {
6175 err = -ENETUNREACH;
6176 }
6177 rcu_read_unlock();
6178
6179 if (err < 0) {
6180 kfree_skb(skb);
6181 goto errout;
6182 }
6183
6184 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
6185errout:
6186 return err;
6187}
6188
6189void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
6190 unsigned int nlm_flags)
6191{
6192 struct sk_buff *skb;
6193 struct net *net = info->nl_net;
6194 u32 seq;
6195 int err;
6196
6197 err = -ENOBUFS;
6198 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6199
6200 skb = nlmsg_new(rt6_nlmsg_size(rt), GFP_ATOMIC);
6201 if (!skb)
6202 goto errout;
6203
6204 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6205 event, info->portid, seq, nlm_flags);
6206 if (err < 0) {
6207 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6208 WARN_ON(err == -EMSGSIZE);
6209 kfree_skb(skb);
6210 goto errout;
6211 }
6212 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6213 info->nlh, GFP_ATOMIC);
6214 return;
6215errout:
6216 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6217}
6218
6219void fib6_rt_update(struct net *net, struct fib6_info *rt,
6220 struct nl_info *info)
6221{
6222 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6223 struct sk_buff *skb;
6224 int err = -ENOBUFS;
6225
6226 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6227 if (!skb)
6228 goto errout;
6229
6230 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6231 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6232 if (err < 0) {
6233 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6234 WARN_ON(err == -EMSGSIZE);
6235 kfree_skb(skb);
6236 goto errout;
6237 }
6238 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6239 info->nlh, gfp_any());
6240 return;
6241errout:
6242 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6243}
6244
6245void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
6246 bool offload, bool trap, bool offload_failed)
6247{
6248 struct sk_buff *skb;
6249 int err;
6250
6251 if (READ_ONCE(f6i->offload) == offload &&
6252 READ_ONCE(f6i->trap) == trap &&
6253 READ_ONCE(f6i->offload_failed) == offload_failed)
6254 return;
6255
6256 WRITE_ONCE(f6i->offload, offload);
6257 WRITE_ONCE(f6i->trap, trap);
6258
6259 /* 2 means send notifications only if offload_failed was changed. */
6260 if (net->ipv6.sysctl.fib_notify_on_flag_change == 2 &&
6261 READ_ONCE(f6i->offload_failed) == offload_failed)
6262 return;
6263
6264 WRITE_ONCE(f6i->offload_failed, offload_failed);
6265
6266 if (!rcu_access_pointer(f6i->fib6_node))
6267 /* The route was removed from the tree, do not send
6268 * notification.
6269 */
6270 return;
6271
6272 if (!net->ipv6.sysctl.fib_notify_on_flag_change)
6273 return;
6274
6275 skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
6276 if (!skb) {
6277 err = -ENOBUFS;
6278 goto errout;
6279 }
6280
6281 err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
6282 0, 0);
6283 if (err < 0) {
6284 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6285 WARN_ON(err == -EMSGSIZE);
6286 kfree_skb(skb);
6287 goto errout;
6288 }
6289
6290 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
6291 return;
6292
6293errout:
6294 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6295}
6296EXPORT_SYMBOL(fib6_info_hw_flags_set);
6297
6298static int ip6_route_dev_notify(struct notifier_block *this,
6299 unsigned long event, void *ptr)
6300{
6301 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6302 struct net *net = dev_net(dev);
6303
6304 if (!(dev->flags & IFF_LOOPBACK))
6305 return NOTIFY_OK;
6306
6307 if (event == NETDEV_REGISTER) {
6308 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
6309 net->ipv6.ip6_null_entry->dst.dev = dev;
6310 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6311#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6312 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
6313 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
6314 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
6315 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
6316#endif
6317 } else if (event == NETDEV_UNREGISTER &&
6318 dev->reg_state != NETREG_UNREGISTERED) {
6319 /* NETDEV_UNREGISTER could be fired for multiple times by
6320 * netdev_wait_allrefs(). Make sure we only call this once.
6321 */
6322 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
6323#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6324 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6325 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
6326#endif
6327 }
6328
6329 return NOTIFY_OK;
6330}
6331
6332/*
6333 * /proc
6334 */
6335
6336#ifdef CONFIG_PROC_FS
6337static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6338{
6339 struct net *net = (struct net *)seq->private;
6340 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6341 net->ipv6.rt6_stats->fib_nodes,
6342 net->ipv6.rt6_stats->fib_route_nodes,
6343 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6344 net->ipv6.rt6_stats->fib_rt_entries,
6345 net->ipv6.rt6_stats->fib_rt_cache,
6346 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6347 net->ipv6.rt6_stats->fib_discarded_routes);
6348
6349 return 0;
6350}
6351#endif /* CONFIG_PROC_FS */
6352
6353#ifdef CONFIG_SYSCTL
6354
6355static int ipv6_sysctl_rtcache_flush(const struct ctl_table *ctl, int write,
6356 void *buffer, size_t *lenp, loff_t *ppos)
6357{
6358 struct net *net;
6359 int delay;
6360 int ret;
6361 if (!write)
6362 return -EINVAL;
6363
6364 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6365 if (ret)
6366 return ret;
6367
6368 net = (struct net *)ctl->extra1;
6369 delay = net->ipv6.sysctl.flush_delay;
6370 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6371 return 0;
6372}
6373
6374static struct ctl_table ipv6_route_table_template[] = {
6375 {
6376 .procname = "max_size",
6377 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
6378 .maxlen = sizeof(int),
6379 .mode = 0644,
6380 .proc_handler = proc_dointvec,
6381 },
6382 {
6383 .procname = "gc_thresh",
6384 .data = &ip6_dst_ops_template.gc_thresh,
6385 .maxlen = sizeof(int),
6386 .mode = 0644,
6387 .proc_handler = proc_dointvec,
6388 },
6389 {
6390 .procname = "flush",
6391 .data = &init_net.ipv6.sysctl.flush_delay,
6392 .maxlen = sizeof(int),
6393 .mode = 0200,
6394 .proc_handler = ipv6_sysctl_rtcache_flush
6395 },
6396 {
6397 .procname = "gc_min_interval",
6398 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6399 .maxlen = sizeof(int),
6400 .mode = 0644,
6401 .proc_handler = proc_dointvec_jiffies,
6402 },
6403 {
6404 .procname = "gc_timeout",
6405 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6406 .maxlen = sizeof(int),
6407 .mode = 0644,
6408 .proc_handler = proc_dointvec_jiffies,
6409 },
6410 {
6411 .procname = "gc_interval",
6412 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
6413 .maxlen = sizeof(int),
6414 .mode = 0644,
6415 .proc_handler = proc_dointvec_jiffies,
6416 },
6417 {
6418 .procname = "gc_elasticity",
6419 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6420 .maxlen = sizeof(int),
6421 .mode = 0644,
6422 .proc_handler = proc_dointvec,
6423 },
6424 {
6425 .procname = "mtu_expires",
6426 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6427 .maxlen = sizeof(int),
6428 .mode = 0644,
6429 .proc_handler = proc_dointvec_jiffies,
6430 },
6431 {
6432 .procname = "min_adv_mss",
6433 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
6434 .maxlen = sizeof(int),
6435 .mode = 0644,
6436 .proc_handler = proc_dointvec,
6437 },
6438 {
6439 .procname = "gc_min_interval_ms",
6440 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6441 .maxlen = sizeof(int),
6442 .mode = 0644,
6443 .proc_handler = proc_dointvec_ms_jiffies,
6444 },
6445 {
6446 .procname = "skip_notify_on_dev_down",
6447 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6448 .maxlen = sizeof(u8),
6449 .mode = 0644,
6450 .proc_handler = proc_dou8vec_minmax,
6451 .extra1 = SYSCTL_ZERO,
6452 .extra2 = SYSCTL_ONE,
6453 },
6454};
6455
6456struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6457{
6458 struct ctl_table *table;
6459
6460 table = kmemdup(ipv6_route_table_template,
6461 sizeof(ipv6_route_table_template),
6462 GFP_KERNEL);
6463
6464 if (table) {
6465 table[0].data = &net->ipv6.sysctl.ip6_rt_max_size;
6466 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6467 table[2].data = &net->ipv6.sysctl.flush_delay;
6468 table[2].extra1 = net;
6469 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6470 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6471 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6472 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6473 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6474 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6475 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6476 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6477 }
6478
6479 return table;
6480}
6481
6482size_t ipv6_route_sysctl_table_size(struct net *net)
6483{
6484 /* Don't export sysctls to unprivileged users */
6485 if (net->user_ns != &init_user_ns)
6486 return 1;
6487
6488 return ARRAY_SIZE(ipv6_route_table_template);
6489}
6490#endif
6491
6492static int __net_init ip6_route_net_init(struct net *net)
6493{
6494 int ret = -ENOMEM;
6495
6496 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6497 sizeof(net->ipv6.ip6_dst_ops));
6498
6499 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6500 goto out_ip6_dst_ops;
6501
6502 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6503 if (!net->ipv6.fib6_null_entry)
6504 goto out_ip6_dst_entries;
6505 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6506 sizeof(*net->ipv6.fib6_null_entry));
6507
6508 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6509 sizeof(*net->ipv6.ip6_null_entry),
6510 GFP_KERNEL);
6511 if (!net->ipv6.ip6_null_entry)
6512 goto out_fib6_null_entry;
6513 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6514 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6515 ip6_template_metrics, true);
6516 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->dst.rt_uncached);
6517
6518#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6519 net->ipv6.fib6_has_custom_rules = false;
6520 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6521 sizeof(*net->ipv6.ip6_prohibit_entry),
6522 GFP_KERNEL);
6523 if (!net->ipv6.ip6_prohibit_entry)
6524 goto out_ip6_null_entry;
6525 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6526 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6527 ip6_template_metrics, true);
6528 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->dst.rt_uncached);
6529
6530 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6531 sizeof(*net->ipv6.ip6_blk_hole_entry),
6532 GFP_KERNEL);
6533 if (!net->ipv6.ip6_blk_hole_entry)
6534 goto out_ip6_prohibit_entry;
6535 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6536 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6537 ip6_template_metrics, true);
6538 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->dst.rt_uncached);
6539#ifdef CONFIG_IPV6_SUBTREES
6540 net->ipv6.fib6_routes_require_src = 0;
6541#endif
6542#endif
6543
6544 net->ipv6.sysctl.flush_delay = 0;
6545 net->ipv6.sysctl.ip6_rt_max_size = INT_MAX;
6546 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6547 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6548 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6549 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6550 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6551 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6552 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6553
6554 atomic_set(&net->ipv6.ip6_rt_gc_expire, 30*HZ);
6555
6556 ret = 0;
6557out:
6558 return ret;
6559
6560#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6561out_ip6_prohibit_entry:
6562 kfree(net->ipv6.ip6_prohibit_entry);
6563out_ip6_null_entry:
6564 kfree(net->ipv6.ip6_null_entry);
6565#endif
6566out_fib6_null_entry:
6567 kfree(net->ipv6.fib6_null_entry);
6568out_ip6_dst_entries:
6569 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6570out_ip6_dst_ops:
6571 goto out;
6572}
6573
6574static void __net_exit ip6_route_net_exit(struct net *net)
6575{
6576 kfree(net->ipv6.fib6_null_entry);
6577 kfree(net->ipv6.ip6_null_entry);
6578#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6579 kfree(net->ipv6.ip6_prohibit_entry);
6580 kfree(net->ipv6.ip6_blk_hole_entry);
6581#endif
6582 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6583}
6584
6585static int __net_init ip6_route_net_init_late(struct net *net)
6586{
6587#ifdef CONFIG_PROC_FS
6588 if (!proc_create_net("ipv6_route", 0, net->proc_net,
6589 &ipv6_route_seq_ops,
6590 sizeof(struct ipv6_route_iter)))
6591 return -ENOMEM;
6592
6593 if (!proc_create_net_single("rt6_stats", 0444, net->proc_net,
6594 rt6_stats_seq_show, NULL)) {
6595 remove_proc_entry("ipv6_route", net->proc_net);
6596 return -ENOMEM;
6597 }
6598#endif
6599 return 0;
6600}
6601
6602static void __net_exit ip6_route_net_exit_late(struct net *net)
6603{
6604#ifdef CONFIG_PROC_FS
6605 remove_proc_entry("ipv6_route", net->proc_net);
6606 remove_proc_entry("rt6_stats", net->proc_net);
6607#endif
6608}
6609
6610static struct pernet_operations ip6_route_net_ops = {
6611 .init = ip6_route_net_init,
6612 .exit = ip6_route_net_exit,
6613};
6614
6615static int __net_init ipv6_inetpeer_init(struct net *net)
6616{
6617 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6618
6619 if (!bp)
6620 return -ENOMEM;
6621 inet_peer_base_init(bp);
6622 net->ipv6.peers = bp;
6623 return 0;
6624}
6625
6626static void __net_exit ipv6_inetpeer_exit(struct net *net)
6627{
6628 struct inet_peer_base *bp = net->ipv6.peers;
6629
6630 net->ipv6.peers = NULL;
6631 inetpeer_invalidate_tree(bp);
6632 kfree(bp);
6633}
6634
6635static struct pernet_operations ipv6_inetpeer_ops = {
6636 .init = ipv6_inetpeer_init,
6637 .exit = ipv6_inetpeer_exit,
6638};
6639
6640static struct pernet_operations ip6_route_net_late_ops = {
6641 .init = ip6_route_net_init_late,
6642 .exit = ip6_route_net_exit_late,
6643};
6644
6645static struct notifier_block ip6_route_dev_notifier = {
6646 .notifier_call = ip6_route_dev_notify,
6647 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6648};
6649
6650void __init ip6_route_init_special_entries(void)
6651{
6652 /* Registering of the loopback is done before this portion of code,
6653 * the loopback reference in rt6_info will not be taken, do it
6654 * manually for init_net */
6655 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6656 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6657 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6658 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6659 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6660 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6661 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6662 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6663 #endif
6664}
6665
6666#if IS_BUILTIN(CONFIG_IPV6)
6667#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6668DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6669
6670BTF_ID_LIST(btf_fib6_info_id)
6671BTF_ID(struct, fib6_info)
6672
6673static const struct bpf_iter_seq_info ipv6_route_seq_info = {
6674 .seq_ops = &ipv6_route_seq_ops,
6675 .init_seq_private = bpf_iter_init_seq_net,
6676 .fini_seq_private = bpf_iter_fini_seq_net,
6677 .seq_priv_size = sizeof(struct ipv6_route_iter),
6678};
6679
6680static struct bpf_iter_reg ipv6_route_reg_info = {
6681 .target = "ipv6_route",
6682 .ctx_arg_info_size = 1,
6683 .ctx_arg_info = {
6684 { offsetof(struct bpf_iter__ipv6_route, rt),
6685 PTR_TO_BTF_ID_OR_NULL },
6686 },
6687 .seq_info = &ipv6_route_seq_info,
6688};
6689
6690static int __init bpf_iter_register(void)
6691{
6692 ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id;
6693 return bpf_iter_reg_target(&ipv6_route_reg_info);
6694}
6695
6696static void bpf_iter_unregister(void)
6697{
6698 bpf_iter_unreg_target(&ipv6_route_reg_info);
6699}
6700#endif
6701#endif
6702
6703static const struct rtnl_msg_handler ip6_route_rtnl_msg_handlers[] __initconst_or_module = {
6704 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_NEWROUTE,
6705 .doit = inet6_rtm_newroute},
6706 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_DELROUTE,
6707 .doit = inet6_rtm_delroute},
6708 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETROUTE,
6709 .doit = inet6_rtm_getroute, .flags = RTNL_FLAG_DOIT_UNLOCKED},
6710};
6711
6712int __init ip6_route_init(void)
6713{
6714 int ret;
6715 int cpu;
6716
6717 ret = -ENOMEM;
6718 ip6_dst_ops_template.kmem_cachep =
6719 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6720 SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT, NULL);
6721 if (!ip6_dst_ops_template.kmem_cachep)
6722 goto out;
6723
6724 ret = dst_entries_init(&ip6_dst_blackhole_ops);
6725 if (ret)
6726 goto out_kmem_cache;
6727
6728 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6729 if (ret)
6730 goto out_dst_entries;
6731
6732 ret = register_pernet_subsys(&ip6_route_net_ops);
6733 if (ret)
6734 goto out_register_inetpeer;
6735
6736 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6737
6738 ret = fib6_init();
6739 if (ret)
6740 goto out_register_subsys;
6741
6742 ret = xfrm6_init();
6743 if (ret)
6744 goto out_fib6_init;
6745
6746 ret = fib6_rules_init();
6747 if (ret)
6748 goto xfrm6_init;
6749
6750 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6751 if (ret)
6752 goto fib6_rules_init;
6753
6754 ret = rtnl_register_many(ip6_route_rtnl_msg_handlers);
6755 if (ret < 0)
6756 goto out_register_late_subsys;
6757
6758 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6759 if (ret)
6760 goto out_register_late_subsys;
6761
6762#if IS_BUILTIN(CONFIG_IPV6)
6763#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6764 ret = bpf_iter_register();
6765 if (ret)
6766 goto out_register_late_subsys;
6767#endif
6768#endif
6769
6770 for_each_possible_cpu(cpu) {
6771 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6772
6773 INIT_LIST_HEAD(&ul->head);
6774 spin_lock_init(&ul->lock);
6775 }
6776
6777out:
6778 return ret;
6779
6780out_register_late_subsys:
6781 rtnl_unregister_all(PF_INET6);
6782 unregister_pernet_subsys(&ip6_route_net_late_ops);
6783fib6_rules_init:
6784 fib6_rules_cleanup();
6785xfrm6_init:
6786 xfrm6_fini();
6787out_fib6_init:
6788 fib6_gc_cleanup();
6789out_register_subsys:
6790 unregister_pernet_subsys(&ip6_route_net_ops);
6791out_register_inetpeer:
6792 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6793out_dst_entries:
6794 dst_entries_destroy(&ip6_dst_blackhole_ops);
6795out_kmem_cache:
6796 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6797 goto out;
6798}
6799
6800void ip6_route_cleanup(void)
6801{
6802#if IS_BUILTIN(CONFIG_IPV6)
6803#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6804 bpf_iter_unregister();
6805#endif
6806#endif
6807 unregister_netdevice_notifier(&ip6_route_dev_notifier);
6808 unregister_pernet_subsys(&ip6_route_net_late_ops);
6809 fib6_rules_cleanup();
6810 xfrm6_fini();
6811 fib6_gc_cleanup();
6812 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6813 unregister_pernet_subsys(&ip6_route_net_ops);
6814 dst_entries_destroy(&ip6_dst_blackhole_ops);
6815 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6816}
1/*
2 * Linux INET6 implementation
3 * FIB front-end.
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14/* Changes:
15 *
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
23 * Ville Nuorvala
24 * Fixed routing subtrees.
25 */
26
27#define pr_fmt(fmt) "IPv6: " fmt
28
29#include <linux/capability.h>
30#include <linux/errno.h>
31#include <linux/export.h>
32#include <linux/types.h>
33#include <linux/times.h>
34#include <linux/socket.h>
35#include <linux/sockios.h>
36#include <linux/net.h>
37#include <linux/route.h>
38#include <linux/netdevice.h>
39#include <linux/in6.h>
40#include <linux/mroute6.h>
41#include <linux/init.h>
42#include <linux/if_arp.h>
43#include <linux/proc_fs.h>
44#include <linux/seq_file.h>
45#include <linux/nsproxy.h>
46#include <linux/slab.h>
47#include <linux/jhash.h>
48#include <net/net_namespace.h>
49#include <net/snmp.h>
50#include <net/ipv6.h>
51#include <net/ip6_fib.h>
52#include <net/ip6_route.h>
53#include <net/ndisc.h>
54#include <net/addrconf.h>
55#include <net/tcp.h>
56#include <linux/rtnetlink.h>
57#include <net/dst.h>
58#include <net/dst_metadata.h>
59#include <net/xfrm.h>
60#include <net/netevent.h>
61#include <net/netlink.h>
62#include <net/nexthop.h>
63#include <net/lwtunnel.h>
64#include <net/ip_tunnels.h>
65#include <net/l3mdev.h>
66#include <trace/events/fib6.h>
67
68#include <linux/uaccess.h>
69
70#ifdef CONFIG_SYSCTL
71#include <linux/sysctl.h>
72#endif
73
74enum rt6_nud_state {
75 RT6_NUD_FAIL_HARD = -3,
76 RT6_NUD_FAIL_PROBE = -2,
77 RT6_NUD_FAIL_DO_RR = -1,
78 RT6_NUD_SUCCEED = 1
79};
80
81static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort);
82static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
83static unsigned int ip6_default_advmss(const struct dst_entry *dst);
84static unsigned int ip6_mtu(const struct dst_entry *dst);
85static struct dst_entry *ip6_negative_advice(struct dst_entry *);
86static void ip6_dst_destroy(struct dst_entry *);
87static void ip6_dst_ifdown(struct dst_entry *,
88 struct net_device *dev, int how);
89static int ip6_dst_gc(struct dst_ops *ops);
90
91static int ip6_pkt_discard(struct sk_buff *skb);
92static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
93static int ip6_pkt_prohibit(struct sk_buff *skb);
94static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
95static void ip6_link_failure(struct sk_buff *skb);
96static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
97 struct sk_buff *skb, u32 mtu);
98static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
99 struct sk_buff *skb);
100static void rt6_dst_from_metrics_check(struct rt6_info *rt);
101static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
102static size_t rt6_nlmsg_size(struct rt6_info *rt);
103static int rt6_fill_node(struct net *net,
104 struct sk_buff *skb, struct rt6_info *rt,
105 struct in6_addr *dst, struct in6_addr *src,
106 int iif, int type, u32 portid, u32 seq,
107 unsigned int flags);
108static struct rt6_info *rt6_find_cached_rt(struct rt6_info *rt,
109 struct in6_addr *daddr,
110 struct in6_addr *saddr);
111
112#ifdef CONFIG_IPV6_ROUTE_INFO
113static struct rt6_info *rt6_add_route_info(struct net *net,
114 const struct in6_addr *prefix, int prefixlen,
115 const struct in6_addr *gwaddr,
116 struct net_device *dev,
117 unsigned int pref);
118static struct rt6_info *rt6_get_route_info(struct net *net,
119 const struct in6_addr *prefix, int prefixlen,
120 const struct in6_addr *gwaddr,
121 struct net_device *dev);
122#endif
123
124struct uncached_list {
125 spinlock_t lock;
126 struct list_head head;
127};
128
129static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
130
131void rt6_uncached_list_add(struct rt6_info *rt)
132{
133 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
134
135 rt->rt6i_uncached_list = ul;
136
137 spin_lock_bh(&ul->lock);
138 list_add_tail(&rt->rt6i_uncached, &ul->head);
139 spin_unlock_bh(&ul->lock);
140}
141
142void rt6_uncached_list_del(struct rt6_info *rt)
143{
144 if (!list_empty(&rt->rt6i_uncached)) {
145 struct uncached_list *ul = rt->rt6i_uncached_list;
146 struct net *net = dev_net(rt->dst.dev);
147
148 spin_lock_bh(&ul->lock);
149 list_del(&rt->rt6i_uncached);
150 atomic_dec(&net->ipv6.rt6_stats->fib_rt_uncache);
151 spin_unlock_bh(&ul->lock);
152 }
153}
154
155static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
156{
157 struct net_device *loopback_dev = net->loopback_dev;
158 int cpu;
159
160 if (dev == loopback_dev)
161 return;
162
163 for_each_possible_cpu(cpu) {
164 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
165 struct rt6_info *rt;
166
167 spin_lock_bh(&ul->lock);
168 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
169 struct inet6_dev *rt_idev = rt->rt6i_idev;
170 struct net_device *rt_dev = rt->dst.dev;
171
172 if (rt_idev->dev == dev) {
173 rt->rt6i_idev = in6_dev_get(loopback_dev);
174 in6_dev_put(rt_idev);
175 }
176
177 if (rt_dev == dev) {
178 rt->dst.dev = loopback_dev;
179 dev_hold(rt->dst.dev);
180 dev_put(rt_dev);
181 }
182 }
183 spin_unlock_bh(&ul->lock);
184 }
185}
186
187static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
188{
189 return dst_metrics_write_ptr(&rt->from->dst);
190}
191
192static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
193{
194 struct rt6_info *rt = (struct rt6_info *)dst;
195
196 if (rt->rt6i_flags & RTF_PCPU)
197 return rt6_pcpu_cow_metrics(rt);
198 else if (rt->rt6i_flags & RTF_CACHE)
199 return NULL;
200 else
201 return dst_cow_metrics_generic(dst, old);
202}
203
204static inline const void *choose_neigh_daddr(struct rt6_info *rt,
205 struct sk_buff *skb,
206 const void *daddr)
207{
208 struct in6_addr *p = &rt->rt6i_gateway;
209
210 if (!ipv6_addr_any(p))
211 return (const void *) p;
212 else if (skb)
213 return &ipv6_hdr(skb)->daddr;
214 return daddr;
215}
216
217static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
218 struct sk_buff *skb,
219 const void *daddr)
220{
221 struct rt6_info *rt = (struct rt6_info *) dst;
222 struct neighbour *n;
223
224 daddr = choose_neigh_daddr(rt, skb, daddr);
225 n = __ipv6_neigh_lookup(dst->dev, daddr);
226 if (n)
227 return n;
228 return neigh_create(&nd_tbl, daddr, dst->dev);
229}
230
231static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
232{
233 struct net_device *dev = dst->dev;
234 struct rt6_info *rt = (struct rt6_info *)dst;
235
236 daddr = choose_neigh_daddr(rt, NULL, daddr);
237 if (!daddr)
238 return;
239 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
240 return;
241 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
242 return;
243 __ipv6_confirm_neigh(dev, daddr);
244}
245
246static struct dst_ops ip6_dst_ops_template = {
247 .family = AF_INET6,
248 .gc = ip6_dst_gc,
249 .gc_thresh = 1024,
250 .check = ip6_dst_check,
251 .default_advmss = ip6_default_advmss,
252 .mtu = ip6_mtu,
253 .cow_metrics = ipv6_cow_metrics,
254 .destroy = ip6_dst_destroy,
255 .ifdown = ip6_dst_ifdown,
256 .negative_advice = ip6_negative_advice,
257 .link_failure = ip6_link_failure,
258 .update_pmtu = ip6_rt_update_pmtu,
259 .redirect = rt6_do_redirect,
260 .local_out = __ip6_local_out,
261 .neigh_lookup = ip6_neigh_lookup,
262 .confirm_neigh = ip6_confirm_neigh,
263};
264
265static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
266{
267 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
268
269 return mtu ? : dst->dev->mtu;
270}
271
272static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
273 struct sk_buff *skb, u32 mtu)
274{
275}
276
277static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
278 struct sk_buff *skb)
279{
280}
281
282static struct dst_ops ip6_dst_blackhole_ops = {
283 .family = AF_INET6,
284 .destroy = ip6_dst_destroy,
285 .check = ip6_dst_check,
286 .mtu = ip6_blackhole_mtu,
287 .default_advmss = ip6_default_advmss,
288 .update_pmtu = ip6_rt_blackhole_update_pmtu,
289 .redirect = ip6_rt_blackhole_redirect,
290 .cow_metrics = dst_cow_metrics_generic,
291 .neigh_lookup = ip6_neigh_lookup,
292};
293
294static const u32 ip6_template_metrics[RTAX_MAX] = {
295 [RTAX_HOPLIMIT - 1] = 0,
296};
297
298static const struct rt6_info ip6_null_entry_template = {
299 .dst = {
300 .__refcnt = ATOMIC_INIT(1),
301 .__use = 1,
302 .obsolete = DST_OBSOLETE_FORCE_CHK,
303 .error = -ENETUNREACH,
304 .input = ip6_pkt_discard,
305 .output = ip6_pkt_discard_out,
306 },
307 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
308 .rt6i_protocol = RTPROT_KERNEL,
309 .rt6i_metric = ~(u32) 0,
310 .rt6i_ref = ATOMIC_INIT(1),
311};
312
313#ifdef CONFIG_IPV6_MULTIPLE_TABLES
314
315static const struct rt6_info ip6_prohibit_entry_template = {
316 .dst = {
317 .__refcnt = ATOMIC_INIT(1),
318 .__use = 1,
319 .obsolete = DST_OBSOLETE_FORCE_CHK,
320 .error = -EACCES,
321 .input = ip6_pkt_prohibit,
322 .output = ip6_pkt_prohibit_out,
323 },
324 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
325 .rt6i_protocol = RTPROT_KERNEL,
326 .rt6i_metric = ~(u32) 0,
327 .rt6i_ref = ATOMIC_INIT(1),
328};
329
330static const struct rt6_info ip6_blk_hole_entry_template = {
331 .dst = {
332 .__refcnt = ATOMIC_INIT(1),
333 .__use = 1,
334 .obsolete = DST_OBSOLETE_FORCE_CHK,
335 .error = -EINVAL,
336 .input = dst_discard,
337 .output = dst_discard_out,
338 },
339 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
340 .rt6i_protocol = RTPROT_KERNEL,
341 .rt6i_metric = ~(u32) 0,
342 .rt6i_ref = ATOMIC_INIT(1),
343};
344
345#endif
346
347static void rt6_info_init(struct rt6_info *rt)
348{
349 struct dst_entry *dst = &rt->dst;
350
351 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
352 INIT_LIST_HEAD(&rt->rt6i_siblings);
353 INIT_LIST_HEAD(&rt->rt6i_uncached);
354}
355
356/* allocate dst with ip6_dst_ops */
357static struct rt6_info *__ip6_dst_alloc(struct net *net,
358 struct net_device *dev,
359 int flags)
360{
361 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
362 1, DST_OBSOLETE_FORCE_CHK, flags);
363
364 if (rt) {
365 rt6_info_init(rt);
366 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
367 }
368
369 return rt;
370}
371
372struct rt6_info *ip6_dst_alloc(struct net *net,
373 struct net_device *dev,
374 int flags)
375{
376 struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
377
378 if (rt) {
379 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
380 if (!rt->rt6i_pcpu) {
381 dst_release_immediate(&rt->dst);
382 return NULL;
383 }
384 }
385
386 return rt;
387}
388EXPORT_SYMBOL(ip6_dst_alloc);
389
390static void ip6_dst_destroy(struct dst_entry *dst)
391{
392 struct rt6_info *rt = (struct rt6_info *)dst;
393 struct rt6_exception_bucket *bucket;
394 struct rt6_info *from = rt->from;
395 struct inet6_dev *idev;
396
397 dst_destroy_metrics_generic(dst);
398 free_percpu(rt->rt6i_pcpu);
399 rt6_uncached_list_del(rt);
400
401 idev = rt->rt6i_idev;
402 if (idev) {
403 rt->rt6i_idev = NULL;
404 in6_dev_put(idev);
405 }
406 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket, 1);
407 if (bucket) {
408 rt->rt6i_exception_bucket = NULL;
409 kfree(bucket);
410 }
411
412 rt->from = NULL;
413 dst_release(&from->dst);
414}
415
416static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
417 int how)
418{
419 struct rt6_info *rt = (struct rt6_info *)dst;
420 struct inet6_dev *idev = rt->rt6i_idev;
421 struct net_device *loopback_dev =
422 dev_net(dev)->loopback_dev;
423
424 if (idev && idev->dev != loopback_dev) {
425 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
426 if (loopback_idev) {
427 rt->rt6i_idev = loopback_idev;
428 in6_dev_put(idev);
429 }
430 }
431}
432
433static bool __rt6_check_expired(const struct rt6_info *rt)
434{
435 if (rt->rt6i_flags & RTF_EXPIRES)
436 return time_after(jiffies, rt->dst.expires);
437 else
438 return false;
439}
440
441static bool rt6_check_expired(const struct rt6_info *rt)
442{
443 if (rt->rt6i_flags & RTF_EXPIRES) {
444 if (time_after(jiffies, rt->dst.expires))
445 return true;
446 } else if (rt->from) {
447 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
448 rt6_check_expired(rt->from);
449 }
450 return false;
451}
452
453static struct rt6_info *rt6_multipath_select(const struct net *net,
454 struct rt6_info *match,
455 struct flowi6 *fl6, int oif,
456 const struct sk_buff *skb,
457 int strict)
458{
459 struct rt6_info *sibling, *next_sibling;
460
461 /* We might have already computed the hash for ICMPv6 errors. In such
462 * case it will always be non-zero. Otherwise now is the time to do it.
463 */
464 if (!fl6->mp_hash)
465 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
466
467 if (fl6->mp_hash <= atomic_read(&match->rt6i_nh_upper_bound))
468 return match;
469
470 list_for_each_entry_safe(sibling, next_sibling, &match->rt6i_siblings,
471 rt6i_siblings) {
472 if (fl6->mp_hash > atomic_read(&sibling->rt6i_nh_upper_bound))
473 continue;
474 if (rt6_score_route(sibling, oif, strict) < 0)
475 break;
476 match = sibling;
477 break;
478 }
479
480 return match;
481}
482
483/*
484 * Route lookup. rcu_read_lock() should be held.
485 */
486
487static inline struct rt6_info *rt6_device_match(struct net *net,
488 struct rt6_info *rt,
489 const struct in6_addr *saddr,
490 int oif,
491 int flags)
492{
493 struct rt6_info *local = NULL;
494 struct rt6_info *sprt;
495
496 if (!oif && ipv6_addr_any(saddr) && !(rt->rt6i_nh_flags & RTNH_F_DEAD))
497 return rt;
498
499 for (sprt = rt; sprt; sprt = rcu_dereference(sprt->rt6_next)) {
500 struct net_device *dev = sprt->dst.dev;
501
502 if (sprt->rt6i_nh_flags & RTNH_F_DEAD)
503 continue;
504
505 if (oif) {
506 if (dev->ifindex == oif)
507 return sprt;
508 if (dev->flags & IFF_LOOPBACK) {
509 if (!sprt->rt6i_idev ||
510 sprt->rt6i_idev->dev->ifindex != oif) {
511 if (flags & RT6_LOOKUP_F_IFACE)
512 continue;
513 if (local &&
514 local->rt6i_idev->dev->ifindex == oif)
515 continue;
516 }
517 local = sprt;
518 }
519 } else {
520 if (ipv6_chk_addr(net, saddr, dev,
521 flags & RT6_LOOKUP_F_IFACE))
522 return sprt;
523 }
524 }
525
526 if (oif) {
527 if (local)
528 return local;
529
530 if (flags & RT6_LOOKUP_F_IFACE)
531 return net->ipv6.ip6_null_entry;
532 }
533
534 return rt->rt6i_nh_flags & RTNH_F_DEAD ? net->ipv6.ip6_null_entry : rt;
535}
536
537#ifdef CONFIG_IPV6_ROUTER_PREF
538struct __rt6_probe_work {
539 struct work_struct work;
540 struct in6_addr target;
541 struct net_device *dev;
542};
543
544static void rt6_probe_deferred(struct work_struct *w)
545{
546 struct in6_addr mcaddr;
547 struct __rt6_probe_work *work =
548 container_of(w, struct __rt6_probe_work, work);
549
550 addrconf_addr_solict_mult(&work->target, &mcaddr);
551 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
552 dev_put(work->dev);
553 kfree(work);
554}
555
556static void rt6_probe(struct rt6_info *rt)
557{
558 struct __rt6_probe_work *work;
559 struct neighbour *neigh;
560 /*
561 * Okay, this does not seem to be appropriate
562 * for now, however, we need to check if it
563 * is really so; aka Router Reachability Probing.
564 *
565 * Router Reachability Probe MUST be rate-limited
566 * to no more than one per minute.
567 */
568 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
569 return;
570 rcu_read_lock_bh();
571 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
572 if (neigh) {
573 if (neigh->nud_state & NUD_VALID)
574 goto out;
575
576 work = NULL;
577 write_lock(&neigh->lock);
578 if (!(neigh->nud_state & NUD_VALID) &&
579 time_after(jiffies,
580 neigh->updated +
581 rt->rt6i_idev->cnf.rtr_probe_interval)) {
582 work = kmalloc(sizeof(*work), GFP_ATOMIC);
583 if (work)
584 __neigh_set_probe_once(neigh);
585 }
586 write_unlock(&neigh->lock);
587 } else {
588 work = kmalloc(sizeof(*work), GFP_ATOMIC);
589 }
590
591 if (work) {
592 INIT_WORK(&work->work, rt6_probe_deferred);
593 work->target = rt->rt6i_gateway;
594 dev_hold(rt->dst.dev);
595 work->dev = rt->dst.dev;
596 schedule_work(&work->work);
597 }
598
599out:
600 rcu_read_unlock_bh();
601}
602#else
603static inline void rt6_probe(struct rt6_info *rt)
604{
605}
606#endif
607
608/*
609 * Default Router Selection (RFC 2461 6.3.6)
610 */
611static inline int rt6_check_dev(struct rt6_info *rt, int oif)
612{
613 struct net_device *dev = rt->dst.dev;
614 if (!oif || dev->ifindex == oif)
615 return 2;
616 if ((dev->flags & IFF_LOOPBACK) &&
617 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
618 return 1;
619 return 0;
620}
621
622static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
623{
624 struct neighbour *neigh;
625 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
626
627 if (rt->rt6i_flags & RTF_NONEXTHOP ||
628 !(rt->rt6i_flags & RTF_GATEWAY))
629 return RT6_NUD_SUCCEED;
630
631 rcu_read_lock_bh();
632 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
633 if (neigh) {
634 read_lock(&neigh->lock);
635 if (neigh->nud_state & NUD_VALID)
636 ret = RT6_NUD_SUCCEED;
637#ifdef CONFIG_IPV6_ROUTER_PREF
638 else if (!(neigh->nud_state & NUD_FAILED))
639 ret = RT6_NUD_SUCCEED;
640 else
641 ret = RT6_NUD_FAIL_PROBE;
642#endif
643 read_unlock(&neigh->lock);
644 } else {
645 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
646 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
647 }
648 rcu_read_unlock_bh();
649
650 return ret;
651}
652
653static int rt6_score_route(struct rt6_info *rt, int oif,
654 int strict)
655{
656 int m;
657
658 m = rt6_check_dev(rt, oif);
659 if (!m && (strict & RT6_LOOKUP_F_IFACE))
660 return RT6_NUD_FAIL_HARD;
661#ifdef CONFIG_IPV6_ROUTER_PREF
662 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
663#endif
664 if (strict & RT6_LOOKUP_F_REACHABLE) {
665 int n = rt6_check_neigh(rt);
666 if (n < 0)
667 return n;
668 }
669 return m;
670}
671
672static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
673 int *mpri, struct rt6_info *match,
674 bool *do_rr)
675{
676 int m;
677 bool match_do_rr = false;
678 struct inet6_dev *idev = rt->rt6i_idev;
679
680 if (rt->rt6i_nh_flags & RTNH_F_DEAD)
681 goto out;
682
683 if (idev->cnf.ignore_routes_with_linkdown &&
684 rt->rt6i_nh_flags & RTNH_F_LINKDOWN &&
685 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
686 goto out;
687
688 if (rt6_check_expired(rt))
689 goto out;
690
691 m = rt6_score_route(rt, oif, strict);
692 if (m == RT6_NUD_FAIL_DO_RR) {
693 match_do_rr = true;
694 m = 0; /* lowest valid score */
695 } else if (m == RT6_NUD_FAIL_HARD) {
696 goto out;
697 }
698
699 if (strict & RT6_LOOKUP_F_REACHABLE)
700 rt6_probe(rt);
701
702 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
703 if (m > *mpri) {
704 *do_rr = match_do_rr;
705 *mpri = m;
706 match = rt;
707 }
708out:
709 return match;
710}
711
712static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
713 struct rt6_info *leaf,
714 struct rt6_info *rr_head,
715 u32 metric, int oif, int strict,
716 bool *do_rr)
717{
718 struct rt6_info *rt, *match, *cont;
719 int mpri = -1;
720
721 match = NULL;
722 cont = NULL;
723 for (rt = rr_head; rt; rt = rcu_dereference(rt->rt6_next)) {
724 if (rt->rt6i_metric != metric) {
725 cont = rt;
726 break;
727 }
728
729 match = find_match(rt, oif, strict, &mpri, match, do_rr);
730 }
731
732 for (rt = leaf; rt && rt != rr_head;
733 rt = rcu_dereference(rt->rt6_next)) {
734 if (rt->rt6i_metric != metric) {
735 cont = rt;
736 break;
737 }
738
739 match = find_match(rt, oif, strict, &mpri, match, do_rr);
740 }
741
742 if (match || !cont)
743 return match;
744
745 for (rt = cont; rt; rt = rcu_dereference(rt->rt6_next))
746 match = find_match(rt, oif, strict, &mpri, match, do_rr);
747
748 return match;
749}
750
751static struct rt6_info *rt6_select(struct net *net, struct fib6_node *fn,
752 int oif, int strict)
753{
754 struct rt6_info *leaf = rcu_dereference(fn->leaf);
755 struct rt6_info *match, *rt0;
756 bool do_rr = false;
757 int key_plen;
758
759 if (!leaf || leaf == net->ipv6.ip6_null_entry)
760 return net->ipv6.ip6_null_entry;
761
762 rt0 = rcu_dereference(fn->rr_ptr);
763 if (!rt0)
764 rt0 = leaf;
765
766 /* Double check to make sure fn is not an intermediate node
767 * and fn->leaf does not points to its child's leaf
768 * (This might happen if all routes under fn are deleted from
769 * the tree and fib6_repair_tree() is called on the node.)
770 */
771 key_plen = rt0->rt6i_dst.plen;
772#ifdef CONFIG_IPV6_SUBTREES
773 if (rt0->rt6i_src.plen)
774 key_plen = rt0->rt6i_src.plen;
775#endif
776 if (fn->fn_bit != key_plen)
777 return net->ipv6.ip6_null_entry;
778
779 match = find_rr_leaf(fn, leaf, rt0, rt0->rt6i_metric, oif, strict,
780 &do_rr);
781
782 if (do_rr) {
783 struct rt6_info *next = rcu_dereference(rt0->rt6_next);
784
785 /* no entries matched; do round-robin */
786 if (!next || next->rt6i_metric != rt0->rt6i_metric)
787 next = leaf;
788
789 if (next != rt0) {
790 spin_lock_bh(&leaf->rt6i_table->tb6_lock);
791 /* make sure next is not being deleted from the tree */
792 if (next->rt6i_node)
793 rcu_assign_pointer(fn->rr_ptr, next);
794 spin_unlock_bh(&leaf->rt6i_table->tb6_lock);
795 }
796 }
797
798 return match ? match : net->ipv6.ip6_null_entry;
799}
800
801static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
802{
803 return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
804}
805
806#ifdef CONFIG_IPV6_ROUTE_INFO
807int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
808 const struct in6_addr *gwaddr)
809{
810 struct net *net = dev_net(dev);
811 struct route_info *rinfo = (struct route_info *) opt;
812 struct in6_addr prefix_buf, *prefix;
813 unsigned int pref;
814 unsigned long lifetime;
815 struct rt6_info *rt;
816
817 if (len < sizeof(struct route_info)) {
818 return -EINVAL;
819 }
820
821 /* Sanity check for prefix_len and length */
822 if (rinfo->length > 3) {
823 return -EINVAL;
824 } else if (rinfo->prefix_len > 128) {
825 return -EINVAL;
826 } else if (rinfo->prefix_len > 64) {
827 if (rinfo->length < 2) {
828 return -EINVAL;
829 }
830 } else if (rinfo->prefix_len > 0) {
831 if (rinfo->length < 1) {
832 return -EINVAL;
833 }
834 }
835
836 pref = rinfo->route_pref;
837 if (pref == ICMPV6_ROUTER_PREF_INVALID)
838 return -EINVAL;
839
840 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
841
842 if (rinfo->length == 3)
843 prefix = (struct in6_addr *)rinfo->prefix;
844 else {
845 /* this function is safe */
846 ipv6_addr_prefix(&prefix_buf,
847 (struct in6_addr *)rinfo->prefix,
848 rinfo->prefix_len);
849 prefix = &prefix_buf;
850 }
851
852 if (rinfo->prefix_len == 0)
853 rt = rt6_get_dflt_router(gwaddr, dev);
854 else
855 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
856 gwaddr, dev);
857
858 if (rt && !lifetime) {
859 ip6_del_rt(rt);
860 rt = NULL;
861 }
862
863 if (!rt && lifetime)
864 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
865 dev, pref);
866 else if (rt)
867 rt->rt6i_flags = RTF_ROUTEINFO |
868 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
869
870 if (rt) {
871 if (!addrconf_finite_timeout(lifetime))
872 rt6_clean_expires(rt);
873 else
874 rt6_set_expires(rt, jiffies + HZ * lifetime);
875
876 ip6_rt_put(rt);
877 }
878 return 0;
879}
880#endif
881
882static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
883 struct in6_addr *saddr)
884{
885 struct fib6_node *pn, *sn;
886 while (1) {
887 if (fn->fn_flags & RTN_TL_ROOT)
888 return NULL;
889 pn = rcu_dereference(fn->parent);
890 sn = FIB6_SUBTREE(pn);
891 if (sn && sn != fn)
892 fn = fib6_lookup(sn, NULL, saddr);
893 else
894 fn = pn;
895 if (fn->fn_flags & RTN_RTINFO)
896 return fn;
897 }
898}
899
900static bool ip6_hold_safe(struct net *net, struct rt6_info **prt,
901 bool null_fallback)
902{
903 struct rt6_info *rt = *prt;
904
905 if (dst_hold_safe(&rt->dst))
906 return true;
907 if (null_fallback) {
908 rt = net->ipv6.ip6_null_entry;
909 dst_hold(&rt->dst);
910 } else {
911 rt = NULL;
912 }
913 *prt = rt;
914 return false;
915}
916
917static struct rt6_info *ip6_pol_route_lookup(struct net *net,
918 struct fib6_table *table,
919 struct flowi6 *fl6,
920 const struct sk_buff *skb,
921 int flags)
922{
923 struct rt6_info *rt, *rt_cache;
924 struct fib6_node *fn;
925
926 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
927 flags &= ~RT6_LOOKUP_F_IFACE;
928
929 rcu_read_lock();
930 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
931restart:
932 rt = rcu_dereference(fn->leaf);
933 if (!rt) {
934 rt = net->ipv6.ip6_null_entry;
935 } else {
936 rt = rt6_device_match(net, rt, &fl6->saddr,
937 fl6->flowi6_oif, flags);
938 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
939 rt = rt6_multipath_select(net, rt, fl6, fl6->flowi6_oif,
940 skb, flags);
941 }
942 if (rt == net->ipv6.ip6_null_entry) {
943 fn = fib6_backtrack(fn, &fl6->saddr);
944 if (fn)
945 goto restart;
946 }
947 /* Search through exception table */
948 rt_cache = rt6_find_cached_rt(rt, &fl6->daddr, &fl6->saddr);
949 if (rt_cache)
950 rt = rt_cache;
951
952 if (ip6_hold_safe(net, &rt, true))
953 dst_use_noref(&rt->dst, jiffies);
954
955 rcu_read_unlock();
956
957 trace_fib6_table_lookup(net, rt, table, fl6);
958
959 return rt;
960
961}
962
963struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
964 const struct sk_buff *skb, int flags)
965{
966 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
967}
968EXPORT_SYMBOL_GPL(ip6_route_lookup);
969
970struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
971 const struct in6_addr *saddr, int oif,
972 const struct sk_buff *skb, int strict)
973{
974 struct flowi6 fl6 = {
975 .flowi6_oif = oif,
976 .daddr = *daddr,
977 };
978 struct dst_entry *dst;
979 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
980
981 if (saddr) {
982 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
983 flags |= RT6_LOOKUP_F_HAS_SADDR;
984 }
985
986 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
987 if (dst->error == 0)
988 return (struct rt6_info *) dst;
989
990 dst_release(dst);
991
992 return NULL;
993}
994EXPORT_SYMBOL(rt6_lookup);
995
996/* ip6_ins_rt is called with FREE table->tb6_lock.
997 * It takes new route entry, the addition fails by any reason the
998 * route is released.
999 * Caller must hold dst before calling it.
1000 */
1001
1002static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
1003 struct mx6_config *mxc,
1004 struct netlink_ext_ack *extack)
1005{
1006 int err;
1007 struct fib6_table *table;
1008
1009 table = rt->rt6i_table;
1010 spin_lock_bh(&table->tb6_lock);
1011 err = fib6_add(&table->tb6_root, rt, info, mxc, extack);
1012 spin_unlock_bh(&table->tb6_lock);
1013
1014 return err;
1015}
1016
1017int ip6_ins_rt(struct rt6_info *rt)
1018{
1019 struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
1020 struct mx6_config mxc = { .mx = NULL, };
1021
1022 /* Hold dst to account for the reference from the fib6 tree */
1023 dst_hold(&rt->dst);
1024 return __ip6_ins_rt(rt, &info, &mxc, NULL);
1025}
1026
1027/* called with rcu_lock held */
1028static struct net_device *ip6_rt_get_dev_rcu(struct rt6_info *rt)
1029{
1030 struct net_device *dev = rt->dst.dev;
1031
1032 if (rt->rt6i_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1033 /* for copies of local routes, dst->dev needs to be the
1034 * device if it is a master device, the master device if
1035 * device is enslaved, and the loopback as the default
1036 */
1037 if (netif_is_l3_slave(dev) &&
1038 !rt6_need_strict(&rt->rt6i_dst.addr))
1039 dev = l3mdev_master_dev_rcu(dev);
1040 else if (!netif_is_l3_master(dev))
1041 dev = dev_net(dev)->loopback_dev;
1042 /* last case is netif_is_l3_master(dev) is true in which
1043 * case we want dev returned to be dev
1044 */
1045 }
1046
1047 return dev;
1048}
1049
1050static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
1051 const struct in6_addr *daddr,
1052 const struct in6_addr *saddr)
1053{
1054 struct net_device *dev;
1055 struct rt6_info *rt;
1056
1057 /*
1058 * Clone the route.
1059 */
1060
1061 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
1062 ort = ort->from;
1063
1064 rcu_read_lock();
1065 dev = ip6_rt_get_dev_rcu(ort);
1066 rt = __ip6_dst_alloc(dev_net(dev), dev, 0);
1067 rcu_read_unlock();
1068 if (!rt)
1069 return NULL;
1070
1071 ip6_rt_copy_init(rt, ort);
1072 rt->rt6i_flags |= RTF_CACHE;
1073 rt->rt6i_metric = 0;
1074 rt->dst.flags |= DST_HOST;
1075 rt->rt6i_dst.addr = *daddr;
1076 rt->rt6i_dst.plen = 128;
1077
1078 if (!rt6_is_gw_or_nonexthop(ort)) {
1079 if (ort->rt6i_dst.plen != 128 &&
1080 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
1081 rt->rt6i_flags |= RTF_ANYCAST;
1082#ifdef CONFIG_IPV6_SUBTREES
1083 if (rt->rt6i_src.plen && saddr) {
1084 rt->rt6i_src.addr = *saddr;
1085 rt->rt6i_src.plen = 128;
1086 }
1087#endif
1088 }
1089
1090 return rt;
1091}
1092
1093static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
1094{
1095 struct net_device *dev;
1096 struct rt6_info *pcpu_rt;
1097
1098 rcu_read_lock();
1099 dev = ip6_rt_get_dev_rcu(rt);
1100 pcpu_rt = __ip6_dst_alloc(dev_net(dev), dev, rt->dst.flags);
1101 rcu_read_unlock();
1102 if (!pcpu_rt)
1103 return NULL;
1104 ip6_rt_copy_init(pcpu_rt, rt);
1105 pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
1106 pcpu_rt->rt6i_flags |= RTF_PCPU;
1107 return pcpu_rt;
1108}
1109
1110/* It should be called with rcu_read_lock() acquired */
1111static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
1112{
1113 struct rt6_info *pcpu_rt, **p;
1114
1115 p = this_cpu_ptr(rt->rt6i_pcpu);
1116 pcpu_rt = *p;
1117
1118 if (pcpu_rt && ip6_hold_safe(NULL, &pcpu_rt, false))
1119 rt6_dst_from_metrics_check(pcpu_rt);
1120
1121 return pcpu_rt;
1122}
1123
1124static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1125{
1126 struct rt6_info *pcpu_rt, *prev, **p;
1127
1128 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1129 if (!pcpu_rt) {
1130 struct net *net = dev_net(rt->dst.dev);
1131
1132 dst_hold(&net->ipv6.ip6_null_entry->dst);
1133 return net->ipv6.ip6_null_entry;
1134 }
1135
1136 dst_hold(&pcpu_rt->dst);
1137 p = this_cpu_ptr(rt->rt6i_pcpu);
1138 prev = cmpxchg(p, NULL, pcpu_rt);
1139 BUG_ON(prev);
1140
1141 rt6_dst_from_metrics_check(pcpu_rt);
1142 return pcpu_rt;
1143}
1144
1145/* exception hash table implementation
1146 */
1147static DEFINE_SPINLOCK(rt6_exception_lock);
1148
1149/* Remove rt6_ex from hash table and free the memory
1150 * Caller must hold rt6_exception_lock
1151 */
1152static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1153 struct rt6_exception *rt6_ex)
1154{
1155 struct net *net;
1156
1157 if (!bucket || !rt6_ex)
1158 return;
1159
1160 net = dev_net(rt6_ex->rt6i->dst.dev);
1161 rt6_ex->rt6i->rt6i_node = NULL;
1162 hlist_del_rcu(&rt6_ex->hlist);
1163 rt6_release(rt6_ex->rt6i);
1164 kfree_rcu(rt6_ex, rcu);
1165 WARN_ON_ONCE(!bucket->depth);
1166 bucket->depth--;
1167 net->ipv6.rt6_stats->fib_rt_cache--;
1168}
1169
1170/* Remove oldest rt6_ex in bucket and free the memory
1171 * Caller must hold rt6_exception_lock
1172 */
1173static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1174{
1175 struct rt6_exception *rt6_ex, *oldest = NULL;
1176
1177 if (!bucket)
1178 return;
1179
1180 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1181 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1182 oldest = rt6_ex;
1183 }
1184 rt6_remove_exception(bucket, oldest);
1185}
1186
1187static u32 rt6_exception_hash(const struct in6_addr *dst,
1188 const struct in6_addr *src)
1189{
1190 static u32 seed __read_mostly;
1191 u32 val;
1192
1193 net_get_random_once(&seed, sizeof(seed));
1194 val = jhash(dst, sizeof(*dst), seed);
1195
1196#ifdef CONFIG_IPV6_SUBTREES
1197 if (src)
1198 val = jhash(src, sizeof(*src), val);
1199#endif
1200 return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1201}
1202
1203/* Helper function to find the cached rt in the hash table
1204 * and update bucket pointer to point to the bucket for this
1205 * (daddr, saddr) pair
1206 * Caller must hold rt6_exception_lock
1207 */
1208static struct rt6_exception *
1209__rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1210 const struct in6_addr *daddr,
1211 const struct in6_addr *saddr)
1212{
1213 struct rt6_exception *rt6_ex;
1214 u32 hval;
1215
1216 if (!(*bucket) || !daddr)
1217 return NULL;
1218
1219 hval = rt6_exception_hash(daddr, saddr);
1220 *bucket += hval;
1221
1222 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1223 struct rt6_info *rt6 = rt6_ex->rt6i;
1224 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1225
1226#ifdef CONFIG_IPV6_SUBTREES
1227 if (matched && saddr)
1228 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1229#endif
1230 if (matched)
1231 return rt6_ex;
1232 }
1233 return NULL;
1234}
1235
1236/* Helper function to find the cached rt in the hash table
1237 * and update bucket pointer to point to the bucket for this
1238 * (daddr, saddr) pair
1239 * Caller must hold rcu_read_lock()
1240 */
1241static struct rt6_exception *
1242__rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1243 const struct in6_addr *daddr,
1244 const struct in6_addr *saddr)
1245{
1246 struct rt6_exception *rt6_ex;
1247 u32 hval;
1248
1249 WARN_ON_ONCE(!rcu_read_lock_held());
1250
1251 if (!(*bucket) || !daddr)
1252 return NULL;
1253
1254 hval = rt6_exception_hash(daddr, saddr);
1255 *bucket += hval;
1256
1257 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1258 struct rt6_info *rt6 = rt6_ex->rt6i;
1259 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1260
1261#ifdef CONFIG_IPV6_SUBTREES
1262 if (matched && saddr)
1263 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1264#endif
1265 if (matched)
1266 return rt6_ex;
1267 }
1268 return NULL;
1269}
1270
1271static int rt6_insert_exception(struct rt6_info *nrt,
1272 struct rt6_info *ort)
1273{
1274 struct net *net = dev_net(ort->dst.dev);
1275 struct rt6_exception_bucket *bucket;
1276 struct in6_addr *src_key = NULL;
1277 struct rt6_exception *rt6_ex;
1278 int err = 0;
1279
1280 /* ort can't be a cache or pcpu route */
1281 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
1282 ort = ort->from;
1283 WARN_ON_ONCE(ort->rt6i_flags & (RTF_CACHE | RTF_PCPU));
1284
1285 spin_lock_bh(&rt6_exception_lock);
1286
1287 if (ort->exception_bucket_flushed) {
1288 err = -EINVAL;
1289 goto out;
1290 }
1291
1292 bucket = rcu_dereference_protected(ort->rt6i_exception_bucket,
1293 lockdep_is_held(&rt6_exception_lock));
1294 if (!bucket) {
1295 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1296 GFP_ATOMIC);
1297 if (!bucket) {
1298 err = -ENOMEM;
1299 goto out;
1300 }
1301 rcu_assign_pointer(ort->rt6i_exception_bucket, bucket);
1302 }
1303
1304#ifdef CONFIG_IPV6_SUBTREES
1305 /* rt6i_src.plen != 0 indicates ort is in subtree
1306 * and exception table is indexed by a hash of
1307 * both rt6i_dst and rt6i_src.
1308 * Otherwise, the exception table is indexed by
1309 * a hash of only rt6i_dst.
1310 */
1311 if (ort->rt6i_src.plen)
1312 src_key = &nrt->rt6i_src.addr;
1313#endif
1314
1315 /* Update rt6i_prefsrc as it could be changed
1316 * in rt6_remove_prefsrc()
1317 */
1318 nrt->rt6i_prefsrc = ort->rt6i_prefsrc;
1319 /* rt6_mtu_change() might lower mtu on ort.
1320 * Only insert this exception route if its mtu
1321 * is less than ort's mtu value.
1322 */
1323 if (nrt->rt6i_pmtu >= dst_mtu(&ort->dst)) {
1324 err = -EINVAL;
1325 goto out;
1326 }
1327
1328 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1329 src_key);
1330 if (rt6_ex)
1331 rt6_remove_exception(bucket, rt6_ex);
1332
1333 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1334 if (!rt6_ex) {
1335 err = -ENOMEM;
1336 goto out;
1337 }
1338 rt6_ex->rt6i = nrt;
1339 rt6_ex->stamp = jiffies;
1340 atomic_inc(&nrt->rt6i_ref);
1341 nrt->rt6i_node = ort->rt6i_node;
1342 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1343 bucket->depth++;
1344 net->ipv6.rt6_stats->fib_rt_cache++;
1345
1346 if (bucket->depth > FIB6_MAX_DEPTH)
1347 rt6_exception_remove_oldest(bucket);
1348
1349out:
1350 spin_unlock_bh(&rt6_exception_lock);
1351
1352 /* Update fn->fn_sernum to invalidate all cached dst */
1353 if (!err) {
1354 spin_lock_bh(&ort->rt6i_table->tb6_lock);
1355 fib6_update_sernum(ort);
1356 spin_unlock_bh(&ort->rt6i_table->tb6_lock);
1357 fib6_force_start_gc(net);
1358 }
1359
1360 return err;
1361}
1362
1363void rt6_flush_exceptions(struct rt6_info *rt)
1364{
1365 struct rt6_exception_bucket *bucket;
1366 struct rt6_exception *rt6_ex;
1367 struct hlist_node *tmp;
1368 int i;
1369
1370 spin_lock_bh(&rt6_exception_lock);
1371 /* Prevent rt6_insert_exception() to recreate the bucket list */
1372 rt->exception_bucket_flushed = 1;
1373
1374 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1375 lockdep_is_held(&rt6_exception_lock));
1376 if (!bucket)
1377 goto out;
1378
1379 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1380 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist)
1381 rt6_remove_exception(bucket, rt6_ex);
1382 WARN_ON_ONCE(bucket->depth);
1383 bucket++;
1384 }
1385
1386out:
1387 spin_unlock_bh(&rt6_exception_lock);
1388}
1389
1390/* Find cached rt in the hash table inside passed in rt
1391 * Caller has to hold rcu_read_lock()
1392 */
1393static struct rt6_info *rt6_find_cached_rt(struct rt6_info *rt,
1394 struct in6_addr *daddr,
1395 struct in6_addr *saddr)
1396{
1397 struct rt6_exception_bucket *bucket;
1398 struct in6_addr *src_key = NULL;
1399 struct rt6_exception *rt6_ex;
1400 struct rt6_info *res = NULL;
1401
1402 bucket = rcu_dereference(rt->rt6i_exception_bucket);
1403
1404#ifdef CONFIG_IPV6_SUBTREES
1405 /* rt6i_src.plen != 0 indicates rt is in subtree
1406 * and exception table is indexed by a hash of
1407 * both rt6i_dst and rt6i_src.
1408 * Otherwise, the exception table is indexed by
1409 * a hash of only rt6i_dst.
1410 */
1411 if (rt->rt6i_src.plen)
1412 src_key = saddr;
1413#endif
1414 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1415
1416 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1417 res = rt6_ex->rt6i;
1418
1419 return res;
1420}
1421
1422/* Remove the passed in cached rt from the hash table that contains it */
1423int rt6_remove_exception_rt(struct rt6_info *rt)
1424{
1425 struct rt6_exception_bucket *bucket;
1426 struct rt6_info *from = rt->from;
1427 struct in6_addr *src_key = NULL;
1428 struct rt6_exception *rt6_ex;
1429 int err;
1430
1431 if (!from ||
1432 !(rt->rt6i_flags & RTF_CACHE))
1433 return -EINVAL;
1434
1435 if (!rcu_access_pointer(from->rt6i_exception_bucket))
1436 return -ENOENT;
1437
1438 spin_lock_bh(&rt6_exception_lock);
1439 bucket = rcu_dereference_protected(from->rt6i_exception_bucket,
1440 lockdep_is_held(&rt6_exception_lock));
1441#ifdef CONFIG_IPV6_SUBTREES
1442 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1443 * and exception table is indexed by a hash of
1444 * both rt6i_dst and rt6i_src.
1445 * Otherwise, the exception table is indexed by
1446 * a hash of only rt6i_dst.
1447 */
1448 if (from->rt6i_src.plen)
1449 src_key = &rt->rt6i_src.addr;
1450#endif
1451 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1452 &rt->rt6i_dst.addr,
1453 src_key);
1454 if (rt6_ex) {
1455 rt6_remove_exception(bucket, rt6_ex);
1456 err = 0;
1457 } else {
1458 err = -ENOENT;
1459 }
1460
1461 spin_unlock_bh(&rt6_exception_lock);
1462 return err;
1463}
1464
1465/* Find rt6_ex which contains the passed in rt cache and
1466 * refresh its stamp
1467 */
1468static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1469{
1470 struct rt6_exception_bucket *bucket;
1471 struct rt6_info *from = rt->from;
1472 struct in6_addr *src_key = NULL;
1473 struct rt6_exception *rt6_ex;
1474
1475 if (!from ||
1476 !(rt->rt6i_flags & RTF_CACHE))
1477 return;
1478
1479 rcu_read_lock();
1480 bucket = rcu_dereference(from->rt6i_exception_bucket);
1481
1482#ifdef CONFIG_IPV6_SUBTREES
1483 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1484 * and exception table is indexed by a hash of
1485 * both rt6i_dst and rt6i_src.
1486 * Otherwise, the exception table is indexed by
1487 * a hash of only rt6i_dst.
1488 */
1489 if (from->rt6i_src.plen)
1490 src_key = &rt->rt6i_src.addr;
1491#endif
1492 rt6_ex = __rt6_find_exception_rcu(&bucket,
1493 &rt->rt6i_dst.addr,
1494 src_key);
1495 if (rt6_ex)
1496 rt6_ex->stamp = jiffies;
1497
1498 rcu_read_unlock();
1499}
1500
1501static void rt6_exceptions_remove_prefsrc(struct rt6_info *rt)
1502{
1503 struct rt6_exception_bucket *bucket;
1504 struct rt6_exception *rt6_ex;
1505 int i;
1506
1507 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1508 lockdep_is_held(&rt6_exception_lock));
1509
1510 if (bucket) {
1511 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1512 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1513 rt6_ex->rt6i->rt6i_prefsrc.plen = 0;
1514 }
1515 bucket++;
1516 }
1517 }
1518}
1519
1520static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1521 struct rt6_info *rt, int mtu)
1522{
1523 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1524 * lowest MTU in the path: always allow updating the route PMTU to
1525 * reflect PMTU decreases.
1526 *
1527 * If the new MTU is higher, and the route PMTU is equal to the local
1528 * MTU, this means the old MTU is the lowest in the path, so allow
1529 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1530 * handle this.
1531 */
1532
1533 if (dst_mtu(&rt->dst) >= mtu)
1534 return true;
1535
1536 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
1537 return true;
1538
1539 return false;
1540}
1541
1542static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
1543 struct rt6_info *rt, int mtu)
1544{
1545 struct rt6_exception_bucket *bucket;
1546 struct rt6_exception *rt6_ex;
1547 int i;
1548
1549 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1550 lockdep_is_held(&rt6_exception_lock));
1551
1552 if (!bucket)
1553 return;
1554
1555 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1556 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1557 struct rt6_info *entry = rt6_ex->rt6i;
1558
1559 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
1560 * route), the metrics of its rt->dst.from have already
1561 * been updated.
1562 */
1563 if (entry->rt6i_pmtu &&
1564 rt6_mtu_change_route_allowed(idev, entry, mtu))
1565 entry->rt6i_pmtu = mtu;
1566 }
1567 bucket++;
1568 }
1569}
1570
1571#define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
1572
1573static void rt6_exceptions_clean_tohost(struct rt6_info *rt,
1574 struct in6_addr *gateway)
1575{
1576 struct rt6_exception_bucket *bucket;
1577 struct rt6_exception *rt6_ex;
1578 struct hlist_node *tmp;
1579 int i;
1580
1581 if (!rcu_access_pointer(rt->rt6i_exception_bucket))
1582 return;
1583
1584 spin_lock_bh(&rt6_exception_lock);
1585 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1586 lockdep_is_held(&rt6_exception_lock));
1587
1588 if (bucket) {
1589 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1590 hlist_for_each_entry_safe(rt6_ex, tmp,
1591 &bucket->chain, hlist) {
1592 struct rt6_info *entry = rt6_ex->rt6i;
1593
1594 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
1595 RTF_CACHE_GATEWAY &&
1596 ipv6_addr_equal(gateway,
1597 &entry->rt6i_gateway)) {
1598 rt6_remove_exception(bucket, rt6_ex);
1599 }
1600 }
1601 bucket++;
1602 }
1603 }
1604
1605 spin_unlock_bh(&rt6_exception_lock);
1606}
1607
1608static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
1609 struct rt6_exception *rt6_ex,
1610 struct fib6_gc_args *gc_args,
1611 unsigned long now)
1612{
1613 struct rt6_info *rt = rt6_ex->rt6i;
1614
1615 /* we are pruning and obsoleting aged-out and non gateway exceptions
1616 * even if others have still references to them, so that on next
1617 * dst_check() such references can be dropped.
1618 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
1619 * expired, independently from their aging, as per RFC 8201 section 4
1620 */
1621 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1622 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
1623 RT6_TRACE("aging clone %p\n", rt);
1624 rt6_remove_exception(bucket, rt6_ex);
1625 return;
1626 }
1627 } else if (time_after(jiffies, rt->dst.expires)) {
1628 RT6_TRACE("purging expired route %p\n", rt);
1629 rt6_remove_exception(bucket, rt6_ex);
1630 return;
1631 }
1632
1633 if (rt->rt6i_flags & RTF_GATEWAY) {
1634 struct neighbour *neigh;
1635 __u8 neigh_flags = 0;
1636
1637 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
1638 if (neigh)
1639 neigh_flags = neigh->flags;
1640
1641 if (!(neigh_flags & NTF_ROUTER)) {
1642 RT6_TRACE("purging route %p via non-router but gateway\n",
1643 rt);
1644 rt6_remove_exception(bucket, rt6_ex);
1645 return;
1646 }
1647 }
1648
1649 gc_args->more++;
1650}
1651
1652void rt6_age_exceptions(struct rt6_info *rt,
1653 struct fib6_gc_args *gc_args,
1654 unsigned long now)
1655{
1656 struct rt6_exception_bucket *bucket;
1657 struct rt6_exception *rt6_ex;
1658 struct hlist_node *tmp;
1659 int i;
1660
1661 if (!rcu_access_pointer(rt->rt6i_exception_bucket))
1662 return;
1663
1664 rcu_read_lock_bh();
1665 spin_lock(&rt6_exception_lock);
1666 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1667 lockdep_is_held(&rt6_exception_lock));
1668
1669 if (bucket) {
1670 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1671 hlist_for_each_entry_safe(rt6_ex, tmp,
1672 &bucket->chain, hlist) {
1673 rt6_age_examine_exception(bucket, rt6_ex,
1674 gc_args, now);
1675 }
1676 bucket++;
1677 }
1678 }
1679 spin_unlock(&rt6_exception_lock);
1680 rcu_read_unlock_bh();
1681}
1682
1683struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
1684 int oif, struct flowi6 *fl6,
1685 const struct sk_buff *skb, int flags)
1686{
1687 struct fib6_node *fn, *saved_fn;
1688 struct rt6_info *rt, *rt_cache;
1689 int strict = 0;
1690
1691 strict |= flags & RT6_LOOKUP_F_IFACE;
1692 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
1693 if (net->ipv6.devconf_all->forwarding == 0)
1694 strict |= RT6_LOOKUP_F_REACHABLE;
1695
1696 rcu_read_lock();
1697
1698 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1699 saved_fn = fn;
1700
1701 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1702 oif = 0;
1703
1704redo_rt6_select:
1705 rt = rt6_select(net, fn, oif, strict);
1706 if (rt->rt6i_nsiblings)
1707 rt = rt6_multipath_select(net, rt, fl6, oif, skb, strict);
1708 if (rt == net->ipv6.ip6_null_entry) {
1709 fn = fib6_backtrack(fn, &fl6->saddr);
1710 if (fn)
1711 goto redo_rt6_select;
1712 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1713 /* also consider unreachable route */
1714 strict &= ~RT6_LOOKUP_F_REACHABLE;
1715 fn = saved_fn;
1716 goto redo_rt6_select;
1717 }
1718 }
1719
1720 /*Search through exception table */
1721 rt_cache = rt6_find_cached_rt(rt, &fl6->daddr, &fl6->saddr);
1722 if (rt_cache)
1723 rt = rt_cache;
1724
1725 if (rt == net->ipv6.ip6_null_entry) {
1726 rcu_read_unlock();
1727 dst_hold(&rt->dst);
1728 trace_fib6_table_lookup(net, rt, table, fl6);
1729 return rt;
1730 } else if (rt->rt6i_flags & RTF_CACHE) {
1731 if (ip6_hold_safe(net, &rt, true)) {
1732 dst_use_noref(&rt->dst, jiffies);
1733 rt6_dst_from_metrics_check(rt);
1734 }
1735 rcu_read_unlock();
1736 trace_fib6_table_lookup(net, rt, table, fl6);
1737 return rt;
1738 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1739 !(rt->rt6i_flags & RTF_GATEWAY))) {
1740 /* Create a RTF_CACHE clone which will not be
1741 * owned by the fib6 tree. It is for the special case where
1742 * the daddr in the skb during the neighbor look-up is different
1743 * from the fl6->daddr used to look-up route here.
1744 */
1745
1746 struct rt6_info *uncached_rt;
1747
1748 if (ip6_hold_safe(net, &rt, true)) {
1749 dst_use_noref(&rt->dst, jiffies);
1750 } else {
1751 rcu_read_unlock();
1752 uncached_rt = rt;
1753 goto uncached_rt_out;
1754 }
1755 rcu_read_unlock();
1756
1757 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1758 dst_release(&rt->dst);
1759
1760 if (uncached_rt) {
1761 /* Uncached_rt's refcnt is taken during ip6_rt_cache_alloc()
1762 * No need for another dst_hold()
1763 */
1764 rt6_uncached_list_add(uncached_rt);
1765 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
1766 } else {
1767 uncached_rt = net->ipv6.ip6_null_entry;
1768 dst_hold(&uncached_rt->dst);
1769 }
1770
1771uncached_rt_out:
1772 trace_fib6_table_lookup(net, uncached_rt, table, fl6);
1773 return uncached_rt;
1774
1775 } else {
1776 /* Get a percpu copy */
1777
1778 struct rt6_info *pcpu_rt;
1779
1780 dst_use_noref(&rt->dst, jiffies);
1781 local_bh_disable();
1782 pcpu_rt = rt6_get_pcpu_route(rt);
1783
1784 if (!pcpu_rt) {
1785 /* atomic_inc_not_zero() is needed when using rcu */
1786 if (atomic_inc_not_zero(&rt->rt6i_ref)) {
1787 /* No dst_hold() on rt is needed because grabbing
1788 * rt->rt6i_ref makes sure rt can't be released.
1789 */
1790 pcpu_rt = rt6_make_pcpu_route(rt);
1791 rt6_release(rt);
1792 } else {
1793 /* rt is already removed from tree */
1794 pcpu_rt = net->ipv6.ip6_null_entry;
1795 dst_hold(&pcpu_rt->dst);
1796 }
1797 }
1798 local_bh_enable();
1799 rcu_read_unlock();
1800 trace_fib6_table_lookup(net, pcpu_rt, table, fl6);
1801 return pcpu_rt;
1802 }
1803}
1804EXPORT_SYMBOL_GPL(ip6_pol_route);
1805
1806static struct rt6_info *ip6_pol_route_input(struct net *net,
1807 struct fib6_table *table,
1808 struct flowi6 *fl6,
1809 const struct sk_buff *skb,
1810 int flags)
1811{
1812 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
1813}
1814
1815struct dst_entry *ip6_route_input_lookup(struct net *net,
1816 struct net_device *dev,
1817 struct flowi6 *fl6,
1818 const struct sk_buff *skb,
1819 int flags)
1820{
1821 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1822 flags |= RT6_LOOKUP_F_IFACE;
1823
1824 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
1825}
1826EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
1827
1828static void ip6_multipath_l3_keys(const struct sk_buff *skb,
1829 struct flow_keys *keys,
1830 struct flow_keys *flkeys)
1831{
1832 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
1833 const struct ipv6hdr *key_iph = outer_iph;
1834 struct flow_keys *_flkeys = flkeys;
1835 const struct ipv6hdr *inner_iph;
1836 const struct icmp6hdr *icmph;
1837 struct ipv6hdr _inner_iph;
1838 struct icmp6hdr _icmph;
1839
1840 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
1841 goto out;
1842
1843 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
1844 sizeof(_icmph), &_icmph);
1845 if (!icmph)
1846 goto out;
1847
1848 if (icmph->icmp6_type != ICMPV6_DEST_UNREACH &&
1849 icmph->icmp6_type != ICMPV6_PKT_TOOBIG &&
1850 icmph->icmp6_type != ICMPV6_TIME_EXCEED &&
1851 icmph->icmp6_type != ICMPV6_PARAMPROB)
1852 goto out;
1853
1854 inner_iph = skb_header_pointer(skb,
1855 skb_transport_offset(skb) + sizeof(*icmph),
1856 sizeof(_inner_iph), &_inner_iph);
1857 if (!inner_iph)
1858 goto out;
1859
1860 key_iph = inner_iph;
1861 _flkeys = NULL;
1862out:
1863 if (_flkeys) {
1864 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
1865 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
1866 keys->tags.flow_label = _flkeys->tags.flow_label;
1867 keys->basic.ip_proto = _flkeys->basic.ip_proto;
1868 } else {
1869 keys->addrs.v6addrs.src = key_iph->saddr;
1870 keys->addrs.v6addrs.dst = key_iph->daddr;
1871 keys->tags.flow_label = ip6_flowinfo(key_iph);
1872 keys->basic.ip_proto = key_iph->nexthdr;
1873 }
1874}
1875
1876/* if skb is set it will be used and fl6 can be NULL */
1877u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
1878 const struct sk_buff *skb, struct flow_keys *flkeys)
1879{
1880 struct flow_keys hash_keys;
1881 u32 mhash;
1882
1883 switch (ip6_multipath_hash_policy(net)) {
1884 case 0:
1885 memset(&hash_keys, 0, sizeof(hash_keys));
1886 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1887 if (skb) {
1888 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
1889 } else {
1890 hash_keys.addrs.v6addrs.src = fl6->saddr;
1891 hash_keys.addrs.v6addrs.dst = fl6->daddr;
1892 hash_keys.tags.flow_label = (__force u32)fl6->flowlabel;
1893 hash_keys.basic.ip_proto = fl6->flowi6_proto;
1894 }
1895 break;
1896 case 1:
1897 if (skb) {
1898 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
1899 struct flow_keys keys;
1900
1901 /* short-circuit if we already have L4 hash present */
1902 if (skb->l4_hash)
1903 return skb_get_hash_raw(skb) >> 1;
1904
1905 memset(&hash_keys, 0, sizeof(hash_keys));
1906
1907 if (!flkeys) {
1908 skb_flow_dissect_flow_keys(skb, &keys, flag);
1909 flkeys = &keys;
1910 }
1911 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1912 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
1913 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
1914 hash_keys.ports.src = flkeys->ports.src;
1915 hash_keys.ports.dst = flkeys->ports.dst;
1916 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
1917 } else {
1918 memset(&hash_keys, 0, sizeof(hash_keys));
1919 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1920 hash_keys.addrs.v6addrs.src = fl6->saddr;
1921 hash_keys.addrs.v6addrs.dst = fl6->daddr;
1922 hash_keys.ports.src = fl6->fl6_sport;
1923 hash_keys.ports.dst = fl6->fl6_dport;
1924 hash_keys.basic.ip_proto = fl6->flowi6_proto;
1925 }
1926 break;
1927 }
1928 mhash = flow_hash_from_keys(&hash_keys);
1929
1930 return mhash >> 1;
1931}
1932
1933void ip6_route_input(struct sk_buff *skb)
1934{
1935 const struct ipv6hdr *iph = ipv6_hdr(skb);
1936 struct net *net = dev_net(skb->dev);
1937 int flags = RT6_LOOKUP_F_HAS_SADDR;
1938 struct ip_tunnel_info *tun_info;
1939 struct flowi6 fl6 = {
1940 .flowi6_iif = skb->dev->ifindex,
1941 .daddr = iph->daddr,
1942 .saddr = iph->saddr,
1943 .flowlabel = ip6_flowinfo(iph),
1944 .flowi6_mark = skb->mark,
1945 .flowi6_proto = iph->nexthdr,
1946 };
1947 struct flow_keys *flkeys = NULL, _flkeys;
1948
1949 tun_info = skb_tunnel_info(skb);
1950 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1951 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1952
1953 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
1954 flkeys = &_flkeys;
1955
1956 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
1957 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
1958 skb_dst_drop(skb);
1959 skb_dst_set(skb,
1960 ip6_route_input_lookup(net, skb->dev, &fl6, skb, flags));
1961}
1962
1963static struct rt6_info *ip6_pol_route_output(struct net *net,
1964 struct fib6_table *table,
1965 struct flowi6 *fl6,
1966 const struct sk_buff *skb,
1967 int flags)
1968{
1969 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
1970}
1971
1972struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1973 struct flowi6 *fl6, int flags)
1974{
1975 bool any_src;
1976
1977 if (rt6_need_strict(&fl6->daddr)) {
1978 struct dst_entry *dst;
1979
1980 dst = l3mdev_link_scope_lookup(net, fl6);
1981 if (dst)
1982 return dst;
1983 }
1984
1985 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1986
1987 any_src = ipv6_addr_any(&fl6->saddr);
1988 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1989 (fl6->flowi6_oif && any_src))
1990 flags |= RT6_LOOKUP_F_IFACE;
1991
1992 if (!any_src)
1993 flags |= RT6_LOOKUP_F_HAS_SADDR;
1994 else if (sk)
1995 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1996
1997 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
1998}
1999EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2000
2001struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2002{
2003 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
2004 struct net_device *loopback_dev = net->loopback_dev;
2005 struct dst_entry *new = NULL;
2006
2007 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
2008 DST_OBSOLETE_DEAD, 0);
2009 if (rt) {
2010 rt6_info_init(rt);
2011 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2012
2013 new = &rt->dst;
2014 new->__use = 1;
2015 new->input = dst_discard;
2016 new->output = dst_discard_out;
2017
2018 dst_copy_metrics(new, &ort->dst);
2019
2020 rt->rt6i_idev = in6_dev_get(loopback_dev);
2021 rt->rt6i_gateway = ort->rt6i_gateway;
2022 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2023 rt->rt6i_metric = 0;
2024
2025 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2026#ifdef CONFIG_IPV6_SUBTREES
2027 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2028#endif
2029 }
2030
2031 dst_release(dst_orig);
2032 return new ? new : ERR_PTR(-ENOMEM);
2033}
2034
2035/*
2036 * Destination cache support functions
2037 */
2038
2039static void rt6_dst_from_metrics_check(struct rt6_info *rt)
2040{
2041 if (rt->from &&
2042 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(&rt->from->dst))
2043 dst_init_metrics(&rt->dst, dst_metrics_ptr(&rt->from->dst), true);
2044}
2045
2046static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
2047{
2048 u32 rt_cookie = 0;
2049
2050 if (!rt6_get_cookie_safe(rt, &rt_cookie) || rt_cookie != cookie)
2051 return NULL;
2052
2053 if (rt6_check_expired(rt))
2054 return NULL;
2055
2056 return &rt->dst;
2057}
2058
2059static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
2060{
2061 if (!__rt6_check_expired(rt) &&
2062 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2063 rt6_check(rt->from, cookie))
2064 return &rt->dst;
2065 else
2066 return NULL;
2067}
2068
2069static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
2070{
2071 struct rt6_info *rt;
2072
2073 rt = (struct rt6_info *) dst;
2074
2075 /* All IPV6 dsts are created with ->obsolete set to the value
2076 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2077 * into this function always.
2078 */
2079
2080 rt6_dst_from_metrics_check(rt);
2081
2082 if (rt->rt6i_flags & RTF_PCPU ||
2083 (unlikely(!list_empty(&rt->rt6i_uncached)) && rt->from))
2084 return rt6_dst_from_check(rt, cookie);
2085 else
2086 return rt6_check(rt, cookie);
2087}
2088
2089static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2090{
2091 struct rt6_info *rt = (struct rt6_info *) dst;
2092
2093 if (rt) {
2094 if (rt->rt6i_flags & RTF_CACHE) {
2095 if (rt6_check_expired(rt)) {
2096 ip6_del_rt(rt);
2097 dst = NULL;
2098 }
2099 } else {
2100 dst_release(dst);
2101 dst = NULL;
2102 }
2103 }
2104 return dst;
2105}
2106
2107static void ip6_link_failure(struct sk_buff *skb)
2108{
2109 struct rt6_info *rt;
2110
2111 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2112
2113 rt = (struct rt6_info *) skb_dst(skb);
2114 if (rt) {
2115 if (rt->rt6i_flags & RTF_CACHE) {
2116 if (dst_hold_safe(&rt->dst))
2117 ip6_del_rt(rt);
2118 } else {
2119 struct fib6_node *fn;
2120
2121 rcu_read_lock();
2122 fn = rcu_dereference(rt->rt6i_node);
2123 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2124 fn->fn_sernum = -1;
2125 rcu_read_unlock();
2126 }
2127 }
2128}
2129
2130static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2131{
2132 struct net *net = dev_net(rt->dst.dev);
2133
2134 rt->rt6i_flags |= RTF_MODIFIED;
2135 rt->rt6i_pmtu = mtu;
2136 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2137}
2138
2139static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2140{
2141 return !(rt->rt6i_flags & RTF_CACHE) &&
2142 (rt->rt6i_flags & RTF_PCPU ||
2143 rcu_access_pointer(rt->rt6i_node));
2144}
2145
2146static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2147 const struct ipv6hdr *iph, u32 mtu)
2148{
2149 const struct in6_addr *daddr, *saddr;
2150 struct rt6_info *rt6 = (struct rt6_info *)dst;
2151
2152 if (rt6->rt6i_flags & RTF_LOCAL)
2153 return;
2154
2155 if (dst_metric_locked(dst, RTAX_MTU))
2156 return;
2157
2158 if (iph) {
2159 daddr = &iph->daddr;
2160 saddr = &iph->saddr;
2161 } else if (sk) {
2162 daddr = &sk->sk_v6_daddr;
2163 saddr = &inet6_sk(sk)->saddr;
2164 } else {
2165 daddr = NULL;
2166 saddr = NULL;
2167 }
2168 dst_confirm_neigh(dst, daddr);
2169 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
2170 if (mtu >= dst_mtu(dst))
2171 return;
2172
2173 if (!rt6_cache_allowed_for_pmtu(rt6)) {
2174 rt6_do_update_pmtu(rt6, mtu);
2175 /* update rt6_ex->stamp for cache */
2176 if (rt6->rt6i_flags & RTF_CACHE)
2177 rt6_update_exception_stamp_rt(rt6);
2178 } else if (daddr) {
2179 struct rt6_info *nrt6;
2180
2181 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
2182 if (nrt6) {
2183 rt6_do_update_pmtu(nrt6, mtu);
2184 if (rt6_insert_exception(nrt6, rt6))
2185 dst_release_immediate(&nrt6->dst);
2186 }
2187 }
2188}
2189
2190static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2191 struct sk_buff *skb, u32 mtu)
2192{
2193 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
2194}
2195
2196void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2197 int oif, u32 mark, kuid_t uid)
2198{
2199 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2200 struct dst_entry *dst;
2201 struct flowi6 fl6;
2202
2203 memset(&fl6, 0, sizeof(fl6));
2204 fl6.flowi6_oif = oif;
2205 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
2206 fl6.daddr = iph->daddr;
2207 fl6.saddr = iph->saddr;
2208 fl6.flowlabel = ip6_flowinfo(iph);
2209 fl6.flowi6_uid = uid;
2210
2211 dst = ip6_route_output(net, NULL, &fl6);
2212 if (!dst->error)
2213 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
2214 dst_release(dst);
2215}
2216EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2217
2218void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2219{
2220 struct dst_entry *dst;
2221
2222 ip6_update_pmtu(skb, sock_net(sk), mtu,
2223 sk->sk_bound_dev_if, sk->sk_mark, sk->sk_uid);
2224
2225 dst = __sk_dst_get(sk);
2226 if (!dst || !dst->obsolete ||
2227 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2228 return;
2229
2230 bh_lock_sock(sk);
2231 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2232 ip6_datagram_dst_update(sk, false);
2233 bh_unlock_sock(sk);
2234}
2235EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2236
2237void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2238 const struct flowi6 *fl6)
2239{
2240#ifdef CONFIG_IPV6_SUBTREES
2241 struct ipv6_pinfo *np = inet6_sk(sk);
2242#endif
2243
2244 ip6_dst_store(sk, dst,
2245 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2246 &sk->sk_v6_daddr : NULL,
2247#ifdef CONFIG_IPV6_SUBTREES
2248 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2249 &np->saddr :
2250#endif
2251 NULL);
2252}
2253
2254/* Handle redirects */
2255struct ip6rd_flowi {
2256 struct flowi6 fl6;
2257 struct in6_addr gateway;
2258};
2259
2260static struct rt6_info *__ip6_route_redirect(struct net *net,
2261 struct fib6_table *table,
2262 struct flowi6 *fl6,
2263 const struct sk_buff *skb,
2264 int flags)
2265{
2266 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
2267 struct rt6_info *rt, *rt_cache;
2268 struct fib6_node *fn;
2269
2270 /* Get the "current" route for this destination and
2271 * check if the redirect has come from appropriate router.
2272 *
2273 * RFC 4861 specifies that redirects should only be
2274 * accepted if they come from the nexthop to the target.
2275 * Due to the way the routes are chosen, this notion
2276 * is a bit fuzzy and one might need to check all possible
2277 * routes.
2278 */
2279
2280 rcu_read_lock();
2281 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2282restart:
2283 for_each_fib6_node_rt_rcu(fn) {
2284 if (rt->rt6i_nh_flags & RTNH_F_DEAD)
2285 continue;
2286 if (rt6_check_expired(rt))
2287 continue;
2288 if (rt->dst.error)
2289 break;
2290 if (!(rt->rt6i_flags & RTF_GATEWAY))
2291 continue;
2292 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
2293 continue;
2294 /* rt_cache's gateway might be different from its 'parent'
2295 * in the case of an ip redirect.
2296 * So we keep searching in the exception table if the gateway
2297 * is different.
2298 */
2299 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway)) {
2300 rt_cache = rt6_find_cached_rt(rt,
2301 &fl6->daddr,
2302 &fl6->saddr);
2303 if (rt_cache &&
2304 ipv6_addr_equal(&rdfl->gateway,
2305 &rt_cache->rt6i_gateway)) {
2306 rt = rt_cache;
2307 break;
2308 }
2309 continue;
2310 }
2311 break;
2312 }
2313
2314 if (!rt)
2315 rt = net->ipv6.ip6_null_entry;
2316 else if (rt->dst.error) {
2317 rt = net->ipv6.ip6_null_entry;
2318 goto out;
2319 }
2320
2321 if (rt == net->ipv6.ip6_null_entry) {
2322 fn = fib6_backtrack(fn, &fl6->saddr);
2323 if (fn)
2324 goto restart;
2325 }
2326
2327out:
2328 ip6_hold_safe(net, &rt, true);
2329
2330 rcu_read_unlock();
2331
2332 trace_fib6_table_lookup(net, rt, table, fl6);
2333 return rt;
2334};
2335
2336static struct dst_entry *ip6_route_redirect(struct net *net,
2337 const struct flowi6 *fl6,
2338 const struct sk_buff *skb,
2339 const struct in6_addr *gateway)
2340{
2341 int flags = RT6_LOOKUP_F_HAS_SADDR;
2342 struct ip6rd_flowi rdfl;
2343
2344 rdfl.fl6 = *fl6;
2345 rdfl.gateway = *gateway;
2346
2347 return fib6_rule_lookup(net, &rdfl.fl6, skb,
2348 flags, __ip6_route_redirect);
2349}
2350
2351void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
2352 kuid_t uid)
2353{
2354 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2355 struct dst_entry *dst;
2356 struct flowi6 fl6;
2357
2358 memset(&fl6, 0, sizeof(fl6));
2359 fl6.flowi6_iif = LOOPBACK_IFINDEX;
2360 fl6.flowi6_oif = oif;
2361 fl6.flowi6_mark = mark;
2362 fl6.daddr = iph->daddr;
2363 fl6.saddr = iph->saddr;
2364 fl6.flowlabel = ip6_flowinfo(iph);
2365 fl6.flowi6_uid = uid;
2366
2367 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
2368 rt6_do_redirect(dst, NULL, skb);
2369 dst_release(dst);
2370}
2371EXPORT_SYMBOL_GPL(ip6_redirect);
2372
2373void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
2374 u32 mark)
2375{
2376 const struct ipv6hdr *iph = ipv6_hdr(skb);
2377 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
2378 struct dst_entry *dst;
2379 struct flowi6 fl6;
2380
2381 memset(&fl6, 0, sizeof(fl6));
2382 fl6.flowi6_iif = LOOPBACK_IFINDEX;
2383 fl6.flowi6_oif = oif;
2384 fl6.flowi6_mark = mark;
2385 fl6.daddr = msg->dest;
2386 fl6.saddr = iph->daddr;
2387 fl6.flowi6_uid = sock_net_uid(net, NULL);
2388
2389 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
2390 rt6_do_redirect(dst, NULL, skb);
2391 dst_release(dst);
2392}
2393
2394void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
2395{
2396 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
2397 sk->sk_uid);
2398}
2399EXPORT_SYMBOL_GPL(ip6_sk_redirect);
2400
2401static unsigned int ip6_default_advmss(const struct dst_entry *dst)
2402{
2403 struct net_device *dev = dst->dev;
2404 unsigned int mtu = dst_mtu(dst);
2405 struct net *net = dev_net(dev);
2406
2407 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
2408
2409 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
2410 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
2411
2412 /*
2413 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
2414 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
2415 * IPV6_MAXPLEN is also valid and means: "any MSS,
2416 * rely only on pmtu discovery"
2417 */
2418 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
2419 mtu = IPV6_MAXPLEN;
2420 return mtu;
2421}
2422
2423static unsigned int ip6_mtu(const struct dst_entry *dst)
2424{
2425 const struct rt6_info *rt = (const struct rt6_info *)dst;
2426 unsigned int mtu = rt->rt6i_pmtu;
2427 struct inet6_dev *idev;
2428
2429 if (mtu)
2430 goto out;
2431
2432 mtu = dst_metric_raw(dst, RTAX_MTU);
2433 if (mtu)
2434 goto out;
2435
2436 mtu = IPV6_MIN_MTU;
2437
2438 rcu_read_lock();
2439 idev = __in6_dev_get(dst->dev);
2440 if (idev)
2441 mtu = idev->cnf.mtu6;
2442 rcu_read_unlock();
2443
2444out:
2445 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
2446
2447 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
2448}
2449
2450struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
2451 struct flowi6 *fl6)
2452{
2453 struct dst_entry *dst;
2454 struct rt6_info *rt;
2455 struct inet6_dev *idev = in6_dev_get(dev);
2456 struct net *net = dev_net(dev);
2457
2458 if (unlikely(!idev))
2459 return ERR_PTR(-ENODEV);
2460
2461 rt = ip6_dst_alloc(net, dev, 0);
2462 if (unlikely(!rt)) {
2463 in6_dev_put(idev);
2464 dst = ERR_PTR(-ENOMEM);
2465 goto out;
2466 }
2467
2468 rt->dst.flags |= DST_HOST;
2469 rt->dst.input = ip6_input;
2470 rt->dst.output = ip6_output;
2471 rt->rt6i_gateway = fl6->daddr;
2472 rt->rt6i_dst.addr = fl6->daddr;
2473 rt->rt6i_dst.plen = 128;
2474 rt->rt6i_idev = idev;
2475 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
2476
2477 /* Add this dst into uncached_list so that rt6_disable_ip() can
2478 * do proper release of the net_device
2479 */
2480 rt6_uncached_list_add(rt);
2481 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
2482
2483 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
2484
2485out:
2486 return dst;
2487}
2488
2489static int ip6_dst_gc(struct dst_ops *ops)
2490{
2491 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
2492 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
2493 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
2494 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
2495 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
2496 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
2497 int entries;
2498
2499 entries = dst_entries_get_fast(ops);
2500 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
2501 entries <= rt_max_size)
2502 goto out;
2503
2504 net->ipv6.ip6_rt_gc_expire++;
2505 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
2506 entries = dst_entries_get_slow(ops);
2507 if (entries < ops->gc_thresh)
2508 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
2509out:
2510 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
2511 return entries > rt_max_size;
2512}
2513
2514static int ip6_convert_metrics(struct mx6_config *mxc,
2515 const struct fib6_config *cfg)
2516{
2517 struct net *net = cfg->fc_nlinfo.nl_net;
2518 bool ecn_ca = false;
2519 struct nlattr *nla;
2520 int remaining;
2521 u32 *mp;
2522
2523 if (!cfg->fc_mx)
2524 return 0;
2525
2526 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
2527 if (unlikely(!mp))
2528 return -ENOMEM;
2529
2530 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
2531 int type = nla_type(nla);
2532 u32 val;
2533
2534 if (!type)
2535 continue;
2536 if (unlikely(type > RTAX_MAX))
2537 goto err;
2538
2539 if (type == RTAX_CC_ALGO) {
2540 char tmp[TCP_CA_NAME_MAX];
2541
2542 nla_strlcpy(tmp, nla, sizeof(tmp));
2543 val = tcp_ca_get_key_by_name(net, tmp, &ecn_ca);
2544 if (val == TCP_CA_UNSPEC)
2545 goto err;
2546 } else {
2547 val = nla_get_u32(nla);
2548 }
2549 if (type == RTAX_HOPLIMIT && val > 255)
2550 val = 255;
2551 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
2552 goto err;
2553
2554 mp[type - 1] = val;
2555 __set_bit(type - 1, mxc->mx_valid);
2556 }
2557
2558 if (ecn_ca) {
2559 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
2560 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
2561 }
2562
2563 mxc->mx = mp;
2564 return 0;
2565 err:
2566 kfree(mp);
2567 return -EINVAL;
2568}
2569
2570static struct rt6_info *ip6_nh_lookup_table(struct net *net,
2571 struct fib6_config *cfg,
2572 const struct in6_addr *gw_addr,
2573 u32 tbid, int flags)
2574{
2575 struct flowi6 fl6 = {
2576 .flowi6_oif = cfg->fc_ifindex,
2577 .daddr = *gw_addr,
2578 .saddr = cfg->fc_prefsrc,
2579 };
2580 struct fib6_table *table;
2581 struct rt6_info *rt;
2582
2583 table = fib6_get_table(net, tbid);
2584 if (!table)
2585 return NULL;
2586
2587 if (!ipv6_addr_any(&cfg->fc_prefsrc))
2588 flags |= RT6_LOOKUP_F_HAS_SADDR;
2589
2590 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
2591 rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, NULL, flags);
2592
2593 /* if table lookup failed, fall back to full lookup */
2594 if (rt == net->ipv6.ip6_null_entry) {
2595 ip6_rt_put(rt);
2596 rt = NULL;
2597 }
2598
2599 return rt;
2600}
2601
2602static int ip6_route_check_nh_onlink(struct net *net,
2603 struct fib6_config *cfg,
2604 const struct net_device *dev,
2605 struct netlink_ext_ack *extack)
2606{
2607 u32 tbid = l3mdev_fib_table(dev) ? : RT_TABLE_MAIN;
2608 const struct in6_addr *gw_addr = &cfg->fc_gateway;
2609 u32 flags = RTF_LOCAL | RTF_ANYCAST | RTF_REJECT;
2610 struct rt6_info *grt;
2611 int err;
2612
2613 err = 0;
2614 grt = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0);
2615 if (grt) {
2616 if (!grt->dst.error &&
2617 (grt->rt6i_flags & flags || dev != grt->dst.dev)) {
2618 NL_SET_ERR_MSG(extack,
2619 "Nexthop has invalid gateway or device mismatch");
2620 err = -EINVAL;
2621 }
2622
2623 ip6_rt_put(grt);
2624 }
2625
2626 return err;
2627}
2628
2629static int ip6_route_check_nh(struct net *net,
2630 struct fib6_config *cfg,
2631 struct net_device **_dev,
2632 struct inet6_dev **idev)
2633{
2634 const struct in6_addr *gw_addr = &cfg->fc_gateway;
2635 struct net_device *dev = _dev ? *_dev : NULL;
2636 struct rt6_info *grt = NULL;
2637 int err = -EHOSTUNREACH;
2638
2639 if (cfg->fc_table) {
2640 int flags = RT6_LOOKUP_F_IFACE;
2641
2642 grt = ip6_nh_lookup_table(net, cfg, gw_addr,
2643 cfg->fc_table, flags);
2644 if (grt) {
2645 if (grt->rt6i_flags & RTF_GATEWAY ||
2646 (dev && dev != grt->dst.dev)) {
2647 ip6_rt_put(grt);
2648 grt = NULL;
2649 }
2650 }
2651 }
2652
2653 if (!grt)
2654 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, NULL, 1);
2655
2656 if (!grt)
2657 goto out;
2658
2659 if (dev) {
2660 if (dev != grt->dst.dev) {
2661 ip6_rt_put(grt);
2662 goto out;
2663 }
2664 } else {
2665 *_dev = dev = grt->dst.dev;
2666 *idev = grt->rt6i_idev;
2667 dev_hold(dev);
2668 in6_dev_hold(grt->rt6i_idev);
2669 }
2670
2671 if (!(grt->rt6i_flags & RTF_GATEWAY))
2672 err = 0;
2673
2674 ip6_rt_put(grt);
2675
2676out:
2677 return err;
2678}
2679
2680static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
2681 struct net_device **_dev, struct inet6_dev **idev,
2682 struct netlink_ext_ack *extack)
2683{
2684 const struct in6_addr *gw_addr = &cfg->fc_gateway;
2685 int gwa_type = ipv6_addr_type(gw_addr);
2686 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
2687 const struct net_device *dev = *_dev;
2688 bool need_addr_check = !dev;
2689 int err = -EINVAL;
2690
2691 /* if gw_addr is local we will fail to detect this in case
2692 * address is still TENTATIVE (DAD in progress). rt6_lookup()
2693 * will return already-added prefix route via interface that
2694 * prefix route was assigned to, which might be non-loopback.
2695 */
2696 if (dev &&
2697 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
2698 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
2699 goto out;
2700 }
2701
2702 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
2703 /* IPv6 strictly inhibits using not link-local
2704 * addresses as nexthop address.
2705 * Otherwise, router will not able to send redirects.
2706 * It is very good, but in some (rare!) circumstances
2707 * (SIT, PtP, NBMA NOARP links) it is handy to allow
2708 * some exceptions. --ANK
2709 * We allow IPv4-mapped nexthops to support RFC4798-type
2710 * addressing
2711 */
2712 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
2713 NL_SET_ERR_MSG(extack, "Invalid gateway address");
2714 goto out;
2715 }
2716
2717 if (cfg->fc_flags & RTNH_F_ONLINK)
2718 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
2719 else
2720 err = ip6_route_check_nh(net, cfg, _dev, idev);
2721
2722 if (err)
2723 goto out;
2724 }
2725
2726 /* reload in case device was changed */
2727 dev = *_dev;
2728
2729 err = -EINVAL;
2730 if (!dev) {
2731 NL_SET_ERR_MSG(extack, "Egress device not specified");
2732 goto out;
2733 } else if (dev->flags & IFF_LOOPBACK) {
2734 NL_SET_ERR_MSG(extack,
2735 "Egress device can not be loopback device for this route");
2736 goto out;
2737 }
2738
2739 /* if we did not check gw_addr above, do so now that the
2740 * egress device has been resolved.
2741 */
2742 if (need_addr_check &&
2743 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
2744 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
2745 goto out;
2746 }
2747
2748 err = 0;
2749out:
2750 return err;
2751}
2752
2753static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg,
2754 struct netlink_ext_ack *extack)
2755{
2756 struct net *net = cfg->fc_nlinfo.nl_net;
2757 struct rt6_info *rt = NULL;
2758 struct net_device *dev = NULL;
2759 struct inet6_dev *idev = NULL;
2760 struct fib6_table *table;
2761 int addr_type;
2762 int err = -EINVAL;
2763
2764 /* RTF_PCPU is an internal flag; can not be set by userspace */
2765 if (cfg->fc_flags & RTF_PCPU) {
2766 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
2767 goto out;
2768 }
2769
2770 /* RTF_CACHE is an internal flag; can not be set by userspace */
2771 if (cfg->fc_flags & RTF_CACHE) {
2772 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
2773 goto out;
2774 }
2775
2776 if (cfg->fc_dst_len > 128) {
2777 NL_SET_ERR_MSG(extack, "Invalid prefix length");
2778 goto out;
2779 }
2780 if (cfg->fc_src_len > 128) {
2781 NL_SET_ERR_MSG(extack, "Invalid source address length");
2782 goto out;
2783 }
2784#ifndef CONFIG_IPV6_SUBTREES
2785 if (cfg->fc_src_len) {
2786 NL_SET_ERR_MSG(extack,
2787 "Specifying source address requires IPV6_SUBTREES to be enabled");
2788 goto out;
2789 }
2790#endif
2791 if (cfg->fc_ifindex) {
2792 err = -ENODEV;
2793 dev = dev_get_by_index(net, cfg->fc_ifindex);
2794 if (!dev)
2795 goto out;
2796 idev = in6_dev_get(dev);
2797 if (!idev)
2798 goto out;
2799 }
2800
2801 if (cfg->fc_metric == 0)
2802 cfg->fc_metric = IP6_RT_PRIO_USER;
2803
2804 if (cfg->fc_flags & RTNH_F_ONLINK) {
2805 if (!dev) {
2806 NL_SET_ERR_MSG(extack,
2807 "Nexthop device required for onlink");
2808 err = -ENODEV;
2809 goto out;
2810 }
2811
2812 if (!(dev->flags & IFF_UP)) {
2813 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
2814 err = -ENETDOWN;
2815 goto out;
2816 }
2817 }
2818
2819 err = -ENOBUFS;
2820 if (cfg->fc_nlinfo.nlh &&
2821 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
2822 table = fib6_get_table(net, cfg->fc_table);
2823 if (!table) {
2824 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
2825 table = fib6_new_table(net, cfg->fc_table);
2826 }
2827 } else {
2828 table = fib6_new_table(net, cfg->fc_table);
2829 }
2830
2831 if (!table)
2832 goto out;
2833
2834 rt = ip6_dst_alloc(net, NULL,
2835 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
2836
2837 if (!rt) {
2838 err = -ENOMEM;
2839 goto out;
2840 }
2841
2842 if (cfg->fc_flags & RTF_EXPIRES)
2843 rt6_set_expires(rt, jiffies +
2844 clock_t_to_jiffies(cfg->fc_expires));
2845 else
2846 rt6_clean_expires(rt);
2847
2848 if (cfg->fc_protocol == RTPROT_UNSPEC)
2849 cfg->fc_protocol = RTPROT_BOOT;
2850 rt->rt6i_protocol = cfg->fc_protocol;
2851
2852 addr_type = ipv6_addr_type(&cfg->fc_dst);
2853
2854 if (addr_type & IPV6_ADDR_MULTICAST)
2855 rt->dst.input = ip6_mc_input;
2856 else if (cfg->fc_flags & RTF_LOCAL)
2857 rt->dst.input = ip6_input;
2858 else
2859 rt->dst.input = ip6_forward;
2860
2861 rt->dst.output = ip6_output;
2862
2863 if (cfg->fc_encap) {
2864 struct lwtunnel_state *lwtstate;
2865
2866 err = lwtunnel_build_state(cfg->fc_encap_type,
2867 cfg->fc_encap, AF_INET6, cfg,
2868 &lwtstate, extack);
2869 if (err)
2870 goto out;
2871 rt->dst.lwtstate = lwtstate_get(lwtstate);
2872 lwtunnel_set_redirect(&rt->dst);
2873 }
2874
2875 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
2876 rt->rt6i_dst.plen = cfg->fc_dst_len;
2877 if (rt->rt6i_dst.plen == 128)
2878 rt->dst.flags |= DST_HOST;
2879
2880#ifdef CONFIG_IPV6_SUBTREES
2881 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
2882 rt->rt6i_src.plen = cfg->fc_src_len;
2883#endif
2884
2885 rt->rt6i_metric = cfg->fc_metric;
2886 rt->rt6i_nh_weight = 1;
2887
2888 /* We cannot add true routes via loopback here,
2889 they would result in kernel looping; promote them to reject routes
2890 */
2891 if ((cfg->fc_flags & RTF_REJECT) ||
2892 (dev && (dev->flags & IFF_LOOPBACK) &&
2893 !(addr_type & IPV6_ADDR_LOOPBACK) &&
2894 !(cfg->fc_flags & RTF_LOCAL))) {
2895 /* hold loopback dev/idev if we haven't done so. */
2896 if (dev != net->loopback_dev) {
2897 if (dev) {
2898 dev_put(dev);
2899 in6_dev_put(idev);
2900 }
2901 dev = net->loopback_dev;
2902 dev_hold(dev);
2903 idev = in6_dev_get(dev);
2904 if (!idev) {
2905 err = -ENODEV;
2906 goto out;
2907 }
2908 }
2909 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
2910 switch (cfg->fc_type) {
2911 case RTN_BLACKHOLE:
2912 rt->dst.error = -EINVAL;
2913 rt->dst.output = dst_discard_out;
2914 rt->dst.input = dst_discard;
2915 break;
2916 case RTN_PROHIBIT:
2917 rt->dst.error = -EACCES;
2918 rt->dst.output = ip6_pkt_prohibit_out;
2919 rt->dst.input = ip6_pkt_prohibit;
2920 break;
2921 case RTN_THROW:
2922 case RTN_UNREACHABLE:
2923 default:
2924 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
2925 : (cfg->fc_type == RTN_UNREACHABLE)
2926 ? -EHOSTUNREACH : -ENETUNREACH;
2927 rt->dst.output = ip6_pkt_discard_out;
2928 rt->dst.input = ip6_pkt_discard;
2929 break;
2930 }
2931 goto install_route;
2932 }
2933
2934 if (cfg->fc_flags & RTF_GATEWAY) {
2935 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
2936 if (err)
2937 goto out;
2938
2939 rt->rt6i_gateway = cfg->fc_gateway;
2940 }
2941
2942 err = -ENODEV;
2943 if (!dev)
2944 goto out;
2945
2946 if (idev->cnf.disable_ipv6) {
2947 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
2948 err = -EACCES;
2949 goto out;
2950 }
2951
2952 if (!(dev->flags & IFF_UP)) {
2953 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
2954 err = -ENETDOWN;
2955 goto out;
2956 }
2957
2958 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
2959 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
2960 NL_SET_ERR_MSG(extack, "Invalid source address");
2961 err = -EINVAL;
2962 goto out;
2963 }
2964 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
2965 rt->rt6i_prefsrc.plen = 128;
2966 } else
2967 rt->rt6i_prefsrc.plen = 0;
2968
2969 rt->rt6i_flags = cfg->fc_flags;
2970
2971install_route:
2972 if (!(rt->rt6i_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
2973 !netif_carrier_ok(dev))
2974 rt->rt6i_nh_flags |= RTNH_F_LINKDOWN;
2975 rt->rt6i_nh_flags |= (cfg->fc_flags & RTNH_F_ONLINK);
2976 rt->dst.dev = dev;
2977 rt->rt6i_idev = idev;
2978 rt->rt6i_table = table;
2979
2980 cfg->fc_nlinfo.nl_net = dev_net(dev);
2981
2982 return rt;
2983out:
2984 if (dev)
2985 dev_put(dev);
2986 if (idev)
2987 in6_dev_put(idev);
2988 if (rt)
2989 dst_release_immediate(&rt->dst);
2990
2991 return ERR_PTR(err);
2992}
2993
2994int ip6_route_add(struct fib6_config *cfg,
2995 struct netlink_ext_ack *extack)
2996{
2997 struct mx6_config mxc = { .mx = NULL, };
2998 struct rt6_info *rt;
2999 int err;
3000
3001 rt = ip6_route_info_create(cfg, extack);
3002 if (IS_ERR(rt)) {
3003 err = PTR_ERR(rt);
3004 rt = NULL;
3005 goto out;
3006 }
3007
3008 err = ip6_convert_metrics(&mxc, cfg);
3009 if (err)
3010 goto out;
3011
3012 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc, extack);
3013
3014 kfree(mxc.mx);
3015
3016 return err;
3017out:
3018 if (rt)
3019 dst_release_immediate(&rt->dst);
3020
3021 return err;
3022}
3023
3024static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
3025{
3026 int err;
3027 struct fib6_table *table;
3028 struct net *net = dev_net(rt->dst.dev);
3029
3030 if (rt == net->ipv6.ip6_null_entry) {
3031 err = -ENOENT;
3032 goto out;
3033 }
3034
3035 table = rt->rt6i_table;
3036 spin_lock_bh(&table->tb6_lock);
3037 err = fib6_del(rt, info);
3038 spin_unlock_bh(&table->tb6_lock);
3039
3040out:
3041 ip6_rt_put(rt);
3042 return err;
3043}
3044
3045int ip6_del_rt(struct rt6_info *rt)
3046{
3047 struct nl_info info = {
3048 .nl_net = dev_net(rt->dst.dev),
3049 };
3050 return __ip6_del_rt(rt, &info);
3051}
3052
3053static int __ip6_del_rt_siblings(struct rt6_info *rt, struct fib6_config *cfg)
3054{
3055 struct nl_info *info = &cfg->fc_nlinfo;
3056 struct net *net = info->nl_net;
3057 struct sk_buff *skb = NULL;
3058 struct fib6_table *table;
3059 int err = -ENOENT;
3060
3061 if (rt == net->ipv6.ip6_null_entry)
3062 goto out_put;
3063 table = rt->rt6i_table;
3064 spin_lock_bh(&table->tb6_lock);
3065
3066 if (rt->rt6i_nsiblings && cfg->fc_delete_all_nh) {
3067 struct rt6_info *sibling, *next_sibling;
3068
3069 /* prefer to send a single notification with all hops */
3070 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3071 if (skb) {
3072 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3073
3074 if (rt6_fill_node(net, skb, rt,
3075 NULL, NULL, 0, RTM_DELROUTE,
3076 info->portid, seq, 0) < 0) {
3077 kfree_skb(skb);
3078 skb = NULL;
3079 } else
3080 info->skip_notify = 1;
3081 }
3082
3083 list_for_each_entry_safe(sibling, next_sibling,
3084 &rt->rt6i_siblings,
3085 rt6i_siblings) {
3086 err = fib6_del(sibling, info);
3087 if (err)
3088 goto out_unlock;
3089 }
3090 }
3091
3092 err = fib6_del(rt, info);
3093out_unlock:
3094 spin_unlock_bh(&table->tb6_lock);
3095out_put:
3096 ip6_rt_put(rt);
3097
3098 if (skb) {
3099 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3100 info->nlh, gfp_any());
3101 }
3102 return err;
3103}
3104
3105static int ip6_route_del(struct fib6_config *cfg,
3106 struct netlink_ext_ack *extack)
3107{
3108 struct rt6_info *rt, *rt_cache;
3109 struct fib6_table *table;
3110 struct fib6_node *fn;
3111 int err = -ESRCH;
3112
3113 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
3114 if (!table) {
3115 NL_SET_ERR_MSG(extack, "FIB table does not exist");
3116 return err;
3117 }
3118
3119 rcu_read_lock();
3120
3121 fn = fib6_locate(&table->tb6_root,
3122 &cfg->fc_dst, cfg->fc_dst_len,
3123 &cfg->fc_src, cfg->fc_src_len,
3124 !(cfg->fc_flags & RTF_CACHE));
3125
3126 if (fn) {
3127 for_each_fib6_node_rt_rcu(fn) {
3128 if (cfg->fc_flags & RTF_CACHE) {
3129 rt_cache = rt6_find_cached_rt(rt, &cfg->fc_dst,
3130 &cfg->fc_src);
3131 if (!rt_cache)
3132 continue;
3133 rt = rt_cache;
3134 }
3135 if (cfg->fc_ifindex &&
3136 (!rt->dst.dev ||
3137 rt->dst.dev->ifindex != cfg->fc_ifindex))
3138 continue;
3139 if (cfg->fc_flags & RTF_GATEWAY &&
3140 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3141 continue;
3142 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
3143 continue;
3144 if (cfg->fc_protocol && cfg->fc_protocol != rt->rt6i_protocol)
3145 continue;
3146 if (!dst_hold_safe(&rt->dst))
3147 break;
3148 rcu_read_unlock();
3149
3150 /* if gateway was specified only delete the one hop */
3151 if (cfg->fc_flags & RTF_GATEWAY)
3152 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3153
3154 return __ip6_del_rt_siblings(rt, cfg);
3155 }
3156 }
3157 rcu_read_unlock();
3158
3159 return err;
3160}
3161
3162static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
3163{
3164 struct netevent_redirect netevent;
3165 struct rt6_info *rt, *nrt = NULL;
3166 struct ndisc_options ndopts;
3167 struct inet6_dev *in6_dev;
3168 struct neighbour *neigh;
3169 struct rd_msg *msg;
3170 int optlen, on_link;
3171 u8 *lladdr;
3172
3173 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
3174 optlen -= sizeof(*msg);
3175
3176 if (optlen < 0) {
3177 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
3178 return;
3179 }
3180
3181 msg = (struct rd_msg *)icmp6_hdr(skb);
3182
3183 if (ipv6_addr_is_multicast(&msg->dest)) {
3184 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
3185 return;
3186 }
3187
3188 on_link = 0;
3189 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
3190 on_link = 1;
3191 } else if (ipv6_addr_type(&msg->target) !=
3192 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
3193 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
3194 return;
3195 }
3196
3197 in6_dev = __in6_dev_get(skb->dev);
3198 if (!in6_dev)
3199 return;
3200 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
3201 return;
3202
3203 /* RFC2461 8.1:
3204 * The IP source address of the Redirect MUST be the same as the current
3205 * first-hop router for the specified ICMP Destination Address.
3206 */
3207
3208 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
3209 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
3210 return;
3211 }
3212
3213 lladdr = NULL;
3214 if (ndopts.nd_opts_tgt_lladdr) {
3215 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
3216 skb->dev);
3217 if (!lladdr) {
3218 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
3219 return;
3220 }
3221 }
3222
3223 rt = (struct rt6_info *) dst;
3224 if (rt->rt6i_flags & RTF_REJECT) {
3225 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
3226 return;
3227 }
3228
3229 /* Redirect received -> path was valid.
3230 * Look, redirects are sent only in response to data packets,
3231 * so that this nexthop apparently is reachable. --ANK
3232 */
3233 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
3234
3235 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
3236 if (!neigh)
3237 return;
3238
3239 /*
3240 * We have finally decided to accept it.
3241 */
3242
3243 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
3244 NEIGH_UPDATE_F_WEAK_OVERRIDE|
3245 NEIGH_UPDATE_F_OVERRIDE|
3246 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
3247 NEIGH_UPDATE_F_ISROUTER)),
3248 NDISC_REDIRECT, &ndopts);
3249
3250 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
3251 if (!nrt)
3252 goto out;
3253
3254 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
3255 if (on_link)
3256 nrt->rt6i_flags &= ~RTF_GATEWAY;
3257
3258 nrt->rt6i_protocol = RTPROT_REDIRECT;
3259 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
3260
3261 /* No need to remove rt from the exception table if rt is
3262 * a cached route because rt6_insert_exception() will
3263 * takes care of it
3264 */
3265 if (rt6_insert_exception(nrt, rt)) {
3266 dst_release_immediate(&nrt->dst);
3267 goto out;
3268 }
3269
3270 netevent.old = &rt->dst;
3271 netevent.new = &nrt->dst;
3272 netevent.daddr = &msg->dest;
3273 netevent.neigh = neigh;
3274 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
3275
3276out:
3277 neigh_release(neigh);
3278}
3279
3280/*
3281 * Misc support functions
3282 */
3283
3284static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
3285{
3286 BUG_ON(from->from);
3287
3288 rt->rt6i_flags &= ~RTF_EXPIRES;
3289 dst_hold(&from->dst);
3290 rt->from = from;
3291 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
3292}
3293
3294static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
3295{
3296 rt->dst.input = ort->dst.input;
3297 rt->dst.output = ort->dst.output;
3298 rt->rt6i_dst = ort->rt6i_dst;
3299 rt->dst.error = ort->dst.error;
3300 rt->rt6i_idev = ort->rt6i_idev;
3301 if (rt->rt6i_idev)
3302 in6_dev_hold(rt->rt6i_idev);
3303 rt->dst.lastuse = jiffies;
3304 rt->rt6i_gateway = ort->rt6i_gateway;
3305 rt->rt6i_flags = ort->rt6i_flags;
3306 rt6_set_from(rt, ort);
3307 rt->rt6i_metric = ort->rt6i_metric;
3308#ifdef CONFIG_IPV6_SUBTREES
3309 rt->rt6i_src = ort->rt6i_src;
3310#endif
3311 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
3312 rt->rt6i_table = ort->rt6i_table;
3313 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
3314}
3315
3316#ifdef CONFIG_IPV6_ROUTE_INFO
3317static struct rt6_info *rt6_get_route_info(struct net *net,
3318 const struct in6_addr *prefix, int prefixlen,
3319 const struct in6_addr *gwaddr,
3320 struct net_device *dev)
3321{
3322 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
3323 int ifindex = dev->ifindex;
3324 struct fib6_node *fn;
3325 struct rt6_info *rt = NULL;
3326 struct fib6_table *table;
3327
3328 table = fib6_get_table(net, tb_id);
3329 if (!table)
3330 return NULL;
3331
3332 rcu_read_lock();
3333 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
3334 if (!fn)
3335 goto out;
3336
3337 for_each_fib6_node_rt_rcu(fn) {
3338 if (rt->dst.dev->ifindex != ifindex)
3339 continue;
3340 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
3341 continue;
3342 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
3343 continue;
3344 ip6_hold_safe(NULL, &rt, false);
3345 break;
3346 }
3347out:
3348 rcu_read_unlock();
3349 return rt;
3350}
3351
3352static struct rt6_info *rt6_add_route_info(struct net *net,
3353 const struct in6_addr *prefix, int prefixlen,
3354 const struct in6_addr *gwaddr,
3355 struct net_device *dev,
3356 unsigned int pref)
3357{
3358 struct fib6_config cfg = {
3359 .fc_metric = IP6_RT_PRIO_USER,
3360 .fc_ifindex = dev->ifindex,
3361 .fc_dst_len = prefixlen,
3362 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
3363 RTF_UP | RTF_PREF(pref),
3364 .fc_protocol = RTPROT_RA,
3365 .fc_nlinfo.portid = 0,
3366 .fc_nlinfo.nlh = NULL,
3367 .fc_nlinfo.nl_net = net,
3368 };
3369
3370 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
3371 cfg.fc_dst = *prefix;
3372 cfg.fc_gateway = *gwaddr;
3373
3374 /* We should treat it as a default route if prefix length is 0. */
3375 if (!prefixlen)
3376 cfg.fc_flags |= RTF_DEFAULT;
3377
3378 ip6_route_add(&cfg, NULL);
3379
3380 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
3381}
3382#endif
3383
3384struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
3385{
3386 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
3387 struct rt6_info *rt;
3388 struct fib6_table *table;
3389
3390 table = fib6_get_table(dev_net(dev), tb_id);
3391 if (!table)
3392 return NULL;
3393
3394 rcu_read_lock();
3395 for_each_fib6_node_rt_rcu(&table->tb6_root) {
3396 if (dev == rt->dst.dev &&
3397 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
3398 ipv6_addr_equal(&rt->rt6i_gateway, addr))
3399 break;
3400 }
3401 if (rt)
3402 ip6_hold_safe(NULL, &rt, false);
3403 rcu_read_unlock();
3404 return rt;
3405}
3406
3407struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
3408 struct net_device *dev,
3409 unsigned int pref)
3410{
3411 struct fib6_config cfg = {
3412 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
3413 .fc_metric = IP6_RT_PRIO_USER,
3414 .fc_ifindex = dev->ifindex,
3415 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
3416 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
3417 .fc_protocol = RTPROT_RA,
3418 .fc_nlinfo.portid = 0,
3419 .fc_nlinfo.nlh = NULL,
3420 .fc_nlinfo.nl_net = dev_net(dev),
3421 };
3422
3423 cfg.fc_gateway = *gwaddr;
3424
3425 if (!ip6_route_add(&cfg, NULL)) {
3426 struct fib6_table *table;
3427
3428 table = fib6_get_table(dev_net(dev), cfg.fc_table);
3429 if (table)
3430 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
3431 }
3432
3433 return rt6_get_dflt_router(gwaddr, dev);
3434}
3435
3436static void __rt6_purge_dflt_routers(struct fib6_table *table)
3437{
3438 struct rt6_info *rt;
3439
3440restart:
3441 rcu_read_lock();
3442 for_each_fib6_node_rt_rcu(&table->tb6_root) {
3443 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
3444 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
3445 if (dst_hold_safe(&rt->dst)) {
3446 rcu_read_unlock();
3447 ip6_del_rt(rt);
3448 } else {
3449 rcu_read_unlock();
3450 }
3451 goto restart;
3452 }
3453 }
3454 rcu_read_unlock();
3455
3456 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
3457}
3458
3459void rt6_purge_dflt_routers(struct net *net)
3460{
3461 struct fib6_table *table;
3462 struct hlist_head *head;
3463 unsigned int h;
3464
3465 rcu_read_lock();
3466
3467 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
3468 head = &net->ipv6.fib_table_hash[h];
3469 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
3470 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
3471 __rt6_purge_dflt_routers(table);
3472 }
3473 }
3474
3475 rcu_read_unlock();
3476}
3477
3478static void rtmsg_to_fib6_config(struct net *net,
3479 struct in6_rtmsg *rtmsg,
3480 struct fib6_config *cfg)
3481{
3482 memset(cfg, 0, sizeof(*cfg));
3483
3484 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
3485 : RT6_TABLE_MAIN;
3486 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
3487 cfg->fc_metric = rtmsg->rtmsg_metric;
3488 cfg->fc_expires = rtmsg->rtmsg_info;
3489 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
3490 cfg->fc_src_len = rtmsg->rtmsg_src_len;
3491 cfg->fc_flags = rtmsg->rtmsg_flags;
3492
3493 cfg->fc_nlinfo.nl_net = net;
3494
3495 cfg->fc_dst = rtmsg->rtmsg_dst;
3496 cfg->fc_src = rtmsg->rtmsg_src;
3497 cfg->fc_gateway = rtmsg->rtmsg_gateway;
3498}
3499
3500int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
3501{
3502 struct fib6_config cfg;
3503 struct in6_rtmsg rtmsg;
3504 int err;
3505
3506 switch (cmd) {
3507 case SIOCADDRT: /* Add a route */
3508 case SIOCDELRT: /* Delete a route */
3509 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3510 return -EPERM;
3511 err = copy_from_user(&rtmsg, arg,
3512 sizeof(struct in6_rtmsg));
3513 if (err)
3514 return -EFAULT;
3515
3516 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
3517
3518 rtnl_lock();
3519 switch (cmd) {
3520 case SIOCADDRT:
3521 err = ip6_route_add(&cfg, NULL);
3522 break;
3523 case SIOCDELRT:
3524 err = ip6_route_del(&cfg, NULL);
3525 break;
3526 default:
3527 err = -EINVAL;
3528 }
3529 rtnl_unlock();
3530
3531 return err;
3532 }
3533
3534 return -EINVAL;
3535}
3536
3537/*
3538 * Drop the packet on the floor
3539 */
3540
3541static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
3542{
3543 int type;
3544 struct dst_entry *dst = skb_dst(skb);
3545 switch (ipstats_mib_noroutes) {
3546 case IPSTATS_MIB_INNOROUTES:
3547 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
3548 if (type == IPV6_ADDR_ANY) {
3549 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
3550 IPSTATS_MIB_INADDRERRORS);
3551 break;
3552 }
3553 /* FALLTHROUGH */
3554 case IPSTATS_MIB_OUTNOROUTES:
3555 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
3556 ipstats_mib_noroutes);
3557 break;
3558 }
3559 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
3560 kfree_skb(skb);
3561 return 0;
3562}
3563
3564static int ip6_pkt_discard(struct sk_buff *skb)
3565{
3566 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
3567}
3568
3569static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
3570{
3571 skb->dev = skb_dst(skb)->dev;
3572 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
3573}
3574
3575static int ip6_pkt_prohibit(struct sk_buff *skb)
3576{
3577 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
3578}
3579
3580static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
3581{
3582 skb->dev = skb_dst(skb)->dev;
3583 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
3584}
3585
3586/*
3587 * Allocate a dst for local (unicast / anycast) address.
3588 */
3589
3590struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
3591 const struct in6_addr *addr,
3592 bool anycast)
3593{
3594 u32 tb_id;
3595 struct net *net = dev_net(idev->dev);
3596 struct net_device *dev = idev->dev;
3597 struct rt6_info *rt;
3598
3599 rt = ip6_dst_alloc(net, dev, DST_NOCOUNT);
3600 if (!rt)
3601 return ERR_PTR(-ENOMEM);
3602
3603 in6_dev_hold(idev);
3604
3605 rt->dst.flags |= DST_HOST;
3606 rt->dst.input = ip6_input;
3607 rt->dst.output = ip6_output;
3608 rt->rt6i_idev = idev;
3609
3610 rt->rt6i_protocol = RTPROT_KERNEL;
3611 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
3612 if (anycast)
3613 rt->rt6i_flags |= RTF_ANYCAST;
3614 else
3615 rt->rt6i_flags |= RTF_LOCAL;
3616
3617 rt->rt6i_gateway = *addr;
3618 rt->rt6i_dst.addr = *addr;
3619 rt->rt6i_dst.plen = 128;
3620 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
3621 rt->rt6i_table = fib6_get_table(net, tb_id);
3622
3623 return rt;
3624}
3625
3626/* remove deleted ip from prefsrc entries */
3627struct arg_dev_net_ip {
3628 struct net_device *dev;
3629 struct net *net;
3630 struct in6_addr *addr;
3631};
3632
3633static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
3634{
3635 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
3636 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
3637 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
3638
3639 if (((void *)rt->dst.dev == dev || !dev) &&
3640 rt != net->ipv6.ip6_null_entry &&
3641 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
3642 spin_lock_bh(&rt6_exception_lock);
3643 /* remove prefsrc entry */
3644 rt->rt6i_prefsrc.plen = 0;
3645 /* need to update cache as well */
3646 rt6_exceptions_remove_prefsrc(rt);
3647 spin_unlock_bh(&rt6_exception_lock);
3648 }
3649 return 0;
3650}
3651
3652void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
3653{
3654 struct net *net = dev_net(ifp->idev->dev);
3655 struct arg_dev_net_ip adni = {
3656 .dev = ifp->idev->dev,
3657 .net = net,
3658 .addr = &ifp->addr,
3659 };
3660 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
3661}
3662
3663#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
3664
3665/* Remove routers and update dst entries when gateway turn into host. */
3666static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
3667{
3668 struct in6_addr *gateway = (struct in6_addr *)arg;
3669
3670 if (((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
3671 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
3672 return -1;
3673 }
3674
3675 /* Further clean up cached routes in exception table.
3676 * This is needed because cached route may have a different
3677 * gateway than its 'parent' in the case of an ip redirect.
3678 */
3679 rt6_exceptions_clean_tohost(rt, gateway);
3680
3681 return 0;
3682}
3683
3684void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
3685{
3686 fib6_clean_all(net, fib6_clean_tohost, gateway);
3687}
3688
3689struct arg_netdev_event {
3690 const struct net_device *dev;
3691 union {
3692 unsigned int nh_flags;
3693 unsigned long event;
3694 };
3695};
3696
3697static struct rt6_info *rt6_multipath_first_sibling(const struct rt6_info *rt)
3698{
3699 struct rt6_info *iter;
3700 struct fib6_node *fn;
3701
3702 fn = rcu_dereference_protected(rt->rt6i_node,
3703 lockdep_is_held(&rt->rt6i_table->tb6_lock));
3704 iter = rcu_dereference_protected(fn->leaf,
3705 lockdep_is_held(&rt->rt6i_table->tb6_lock));
3706 while (iter) {
3707 if (iter->rt6i_metric == rt->rt6i_metric &&
3708 rt6_qualify_for_ecmp(iter))
3709 return iter;
3710 iter = rcu_dereference_protected(iter->rt6_next,
3711 lockdep_is_held(&rt->rt6i_table->tb6_lock));
3712 }
3713
3714 return NULL;
3715}
3716
3717static bool rt6_is_dead(const struct rt6_info *rt)
3718{
3719 if (rt->rt6i_nh_flags & RTNH_F_DEAD ||
3720 (rt->rt6i_nh_flags & RTNH_F_LINKDOWN &&
3721 rt->rt6i_idev->cnf.ignore_routes_with_linkdown))
3722 return true;
3723
3724 return false;
3725}
3726
3727static int rt6_multipath_total_weight(const struct rt6_info *rt)
3728{
3729 struct rt6_info *iter;
3730 int total = 0;
3731
3732 if (!rt6_is_dead(rt))
3733 total += rt->rt6i_nh_weight;
3734
3735 list_for_each_entry(iter, &rt->rt6i_siblings, rt6i_siblings) {
3736 if (!rt6_is_dead(iter))
3737 total += iter->rt6i_nh_weight;
3738 }
3739
3740 return total;
3741}
3742
3743static void rt6_upper_bound_set(struct rt6_info *rt, int *weight, int total)
3744{
3745 int upper_bound = -1;
3746
3747 if (!rt6_is_dead(rt)) {
3748 *weight += rt->rt6i_nh_weight;
3749 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
3750 total) - 1;
3751 }
3752 atomic_set(&rt->rt6i_nh_upper_bound, upper_bound);
3753}
3754
3755static void rt6_multipath_upper_bound_set(struct rt6_info *rt, int total)
3756{
3757 struct rt6_info *iter;
3758 int weight = 0;
3759
3760 rt6_upper_bound_set(rt, &weight, total);
3761
3762 list_for_each_entry(iter, &rt->rt6i_siblings, rt6i_siblings)
3763 rt6_upper_bound_set(iter, &weight, total);
3764}
3765
3766void rt6_multipath_rebalance(struct rt6_info *rt)
3767{
3768 struct rt6_info *first;
3769 int total;
3770
3771 /* In case the entire multipath route was marked for flushing,
3772 * then there is no need to rebalance upon the removal of every
3773 * sibling route.
3774 */
3775 if (!rt->rt6i_nsiblings || rt->should_flush)
3776 return;
3777
3778 /* During lookup routes are evaluated in order, so we need to
3779 * make sure upper bounds are assigned from the first sibling
3780 * onwards.
3781 */
3782 first = rt6_multipath_first_sibling(rt);
3783 if (WARN_ON_ONCE(!first))
3784 return;
3785
3786 total = rt6_multipath_total_weight(first);
3787 rt6_multipath_upper_bound_set(first, total);
3788}
3789
3790static int fib6_ifup(struct rt6_info *rt, void *p_arg)
3791{
3792 const struct arg_netdev_event *arg = p_arg;
3793 const struct net *net = dev_net(arg->dev);
3794
3795 if (rt != net->ipv6.ip6_null_entry && rt->dst.dev == arg->dev) {
3796 rt->rt6i_nh_flags &= ~arg->nh_flags;
3797 fib6_update_sernum_upto_root(dev_net(rt->dst.dev), rt);
3798 rt6_multipath_rebalance(rt);
3799 }
3800
3801 return 0;
3802}
3803
3804void rt6_sync_up(struct net_device *dev, unsigned int nh_flags)
3805{
3806 struct arg_netdev_event arg = {
3807 .dev = dev,
3808 {
3809 .nh_flags = nh_flags,
3810 },
3811 };
3812
3813 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
3814 arg.nh_flags |= RTNH_F_LINKDOWN;
3815
3816 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
3817}
3818
3819static bool rt6_multipath_uses_dev(const struct rt6_info *rt,
3820 const struct net_device *dev)
3821{
3822 struct rt6_info *iter;
3823
3824 if (rt->dst.dev == dev)
3825 return true;
3826 list_for_each_entry(iter, &rt->rt6i_siblings, rt6i_siblings)
3827 if (iter->dst.dev == dev)
3828 return true;
3829
3830 return false;
3831}
3832
3833static void rt6_multipath_flush(struct rt6_info *rt)
3834{
3835 struct rt6_info *iter;
3836
3837 rt->should_flush = 1;
3838 list_for_each_entry(iter, &rt->rt6i_siblings, rt6i_siblings)
3839 iter->should_flush = 1;
3840}
3841
3842static unsigned int rt6_multipath_dead_count(const struct rt6_info *rt,
3843 const struct net_device *down_dev)
3844{
3845 struct rt6_info *iter;
3846 unsigned int dead = 0;
3847
3848 if (rt->dst.dev == down_dev || rt->rt6i_nh_flags & RTNH_F_DEAD)
3849 dead++;
3850 list_for_each_entry(iter, &rt->rt6i_siblings, rt6i_siblings)
3851 if (iter->dst.dev == down_dev ||
3852 iter->rt6i_nh_flags & RTNH_F_DEAD)
3853 dead++;
3854
3855 return dead;
3856}
3857
3858static void rt6_multipath_nh_flags_set(struct rt6_info *rt,
3859 const struct net_device *dev,
3860 unsigned int nh_flags)
3861{
3862 struct rt6_info *iter;
3863
3864 if (rt->dst.dev == dev)
3865 rt->rt6i_nh_flags |= nh_flags;
3866 list_for_each_entry(iter, &rt->rt6i_siblings, rt6i_siblings)
3867 if (iter->dst.dev == dev)
3868 iter->rt6i_nh_flags |= nh_flags;
3869}
3870
3871/* called with write lock held for table with rt */
3872static int fib6_ifdown(struct rt6_info *rt, void *p_arg)
3873{
3874 const struct arg_netdev_event *arg = p_arg;
3875 const struct net_device *dev = arg->dev;
3876 const struct net *net = dev_net(dev);
3877
3878 if (rt == net->ipv6.ip6_null_entry)
3879 return 0;
3880
3881 switch (arg->event) {
3882 case NETDEV_UNREGISTER:
3883 return rt->dst.dev == dev ? -1 : 0;
3884 case NETDEV_DOWN:
3885 if (rt->should_flush)
3886 return -1;
3887 if (!rt->rt6i_nsiblings)
3888 return rt->dst.dev == dev ? -1 : 0;
3889 if (rt6_multipath_uses_dev(rt, dev)) {
3890 unsigned int count;
3891
3892 count = rt6_multipath_dead_count(rt, dev);
3893 if (rt->rt6i_nsiblings + 1 == count) {
3894 rt6_multipath_flush(rt);
3895 return -1;
3896 }
3897 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
3898 RTNH_F_LINKDOWN);
3899 fib6_update_sernum(rt);
3900 rt6_multipath_rebalance(rt);
3901 }
3902 return -2;
3903 case NETDEV_CHANGE:
3904 if (rt->dst.dev != dev ||
3905 rt->rt6i_flags & (RTF_LOCAL | RTF_ANYCAST))
3906 break;
3907 rt->rt6i_nh_flags |= RTNH_F_LINKDOWN;
3908 rt6_multipath_rebalance(rt);
3909 break;
3910 }
3911
3912 return 0;
3913}
3914
3915void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
3916{
3917 struct arg_netdev_event arg = {
3918 .dev = dev,
3919 {
3920 .event = event,
3921 },
3922 };
3923
3924 fib6_clean_all(dev_net(dev), fib6_ifdown, &arg);
3925}
3926
3927void rt6_disable_ip(struct net_device *dev, unsigned long event)
3928{
3929 rt6_sync_down_dev(dev, event);
3930 rt6_uncached_list_flush_dev(dev_net(dev), dev);
3931 neigh_ifdown(&nd_tbl, dev);
3932}
3933
3934struct rt6_mtu_change_arg {
3935 struct net_device *dev;
3936 unsigned int mtu;
3937};
3938
3939static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
3940{
3941 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
3942 struct inet6_dev *idev;
3943
3944 /* In IPv6 pmtu discovery is not optional,
3945 so that RTAX_MTU lock cannot disable it.
3946 We still use this lock to block changes
3947 caused by addrconf/ndisc.
3948 */
3949
3950 idev = __in6_dev_get(arg->dev);
3951 if (!idev)
3952 return 0;
3953
3954 /* For administrative MTU increase, there is no way to discover
3955 IPv6 PMTU increase, so PMTU increase should be updated here.
3956 Since RFC 1981 doesn't include administrative MTU increase
3957 update PMTU increase is a MUST. (i.e. jumbo frame)
3958 */
3959 if (rt->dst.dev == arg->dev &&
3960 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
3961 spin_lock_bh(&rt6_exception_lock);
3962 if (dst_metric_raw(&rt->dst, RTAX_MTU) &&
3963 rt6_mtu_change_route_allowed(idev, rt, arg->mtu))
3964 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
3965 rt6_exceptions_update_pmtu(idev, rt, arg->mtu);
3966 spin_unlock_bh(&rt6_exception_lock);
3967 }
3968 return 0;
3969}
3970
3971void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
3972{
3973 struct rt6_mtu_change_arg arg = {
3974 .dev = dev,
3975 .mtu = mtu,
3976 };
3977
3978 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
3979}
3980
3981static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
3982 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
3983 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
3984 [RTA_OIF] = { .type = NLA_U32 },
3985 [RTA_IIF] = { .type = NLA_U32 },
3986 [RTA_PRIORITY] = { .type = NLA_U32 },
3987 [RTA_METRICS] = { .type = NLA_NESTED },
3988 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
3989 [RTA_PREF] = { .type = NLA_U8 },
3990 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
3991 [RTA_ENCAP] = { .type = NLA_NESTED },
3992 [RTA_EXPIRES] = { .type = NLA_U32 },
3993 [RTA_UID] = { .type = NLA_U32 },
3994 [RTA_MARK] = { .type = NLA_U32 },
3995 [RTA_TABLE] = { .type = NLA_U32 },
3996};
3997
3998static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
3999 struct fib6_config *cfg,
4000 struct netlink_ext_ack *extack)
4001{
4002 struct rtmsg *rtm;
4003 struct nlattr *tb[RTA_MAX+1];
4004 unsigned int pref;
4005 int err;
4006
4007 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
4008 NULL);
4009 if (err < 0)
4010 goto errout;
4011
4012 err = -EINVAL;
4013 rtm = nlmsg_data(nlh);
4014 memset(cfg, 0, sizeof(*cfg));
4015
4016 cfg->fc_table = rtm->rtm_table;
4017 cfg->fc_dst_len = rtm->rtm_dst_len;
4018 cfg->fc_src_len = rtm->rtm_src_len;
4019 cfg->fc_flags = RTF_UP;
4020 cfg->fc_protocol = rtm->rtm_protocol;
4021 cfg->fc_type = rtm->rtm_type;
4022
4023 if (rtm->rtm_type == RTN_UNREACHABLE ||
4024 rtm->rtm_type == RTN_BLACKHOLE ||
4025 rtm->rtm_type == RTN_PROHIBIT ||
4026 rtm->rtm_type == RTN_THROW)
4027 cfg->fc_flags |= RTF_REJECT;
4028
4029 if (rtm->rtm_type == RTN_LOCAL)
4030 cfg->fc_flags |= RTF_LOCAL;
4031
4032 if (rtm->rtm_flags & RTM_F_CLONED)
4033 cfg->fc_flags |= RTF_CACHE;
4034
4035 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4036
4037 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
4038 cfg->fc_nlinfo.nlh = nlh;
4039 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
4040
4041 if (tb[RTA_GATEWAY]) {
4042 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
4043 cfg->fc_flags |= RTF_GATEWAY;
4044 }
4045
4046 if (tb[RTA_DST]) {
4047 int plen = (rtm->rtm_dst_len + 7) >> 3;
4048
4049 if (nla_len(tb[RTA_DST]) < plen)
4050 goto errout;
4051
4052 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
4053 }
4054
4055 if (tb[RTA_SRC]) {
4056 int plen = (rtm->rtm_src_len + 7) >> 3;
4057
4058 if (nla_len(tb[RTA_SRC]) < plen)
4059 goto errout;
4060
4061 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
4062 }
4063
4064 if (tb[RTA_PREFSRC])
4065 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
4066
4067 if (tb[RTA_OIF])
4068 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4069
4070 if (tb[RTA_PRIORITY])
4071 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4072
4073 if (tb[RTA_METRICS]) {
4074 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4075 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
4076 }
4077
4078 if (tb[RTA_TABLE])
4079 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4080
4081 if (tb[RTA_MULTIPATH]) {
4082 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4083 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
4084
4085 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
4086 cfg->fc_mp_len, extack);
4087 if (err < 0)
4088 goto errout;
4089 }
4090
4091 if (tb[RTA_PREF]) {
4092 pref = nla_get_u8(tb[RTA_PREF]);
4093 if (pref != ICMPV6_ROUTER_PREF_LOW &&
4094 pref != ICMPV6_ROUTER_PREF_HIGH)
4095 pref = ICMPV6_ROUTER_PREF_MEDIUM;
4096 cfg->fc_flags |= RTF_PREF(pref);
4097 }
4098
4099 if (tb[RTA_ENCAP])
4100 cfg->fc_encap = tb[RTA_ENCAP];
4101
4102 if (tb[RTA_ENCAP_TYPE]) {
4103 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
4104
4105 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
4106 if (err < 0)
4107 goto errout;
4108 }
4109
4110 if (tb[RTA_EXPIRES]) {
4111 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
4112
4113 if (addrconf_finite_timeout(timeout)) {
4114 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
4115 cfg->fc_flags |= RTF_EXPIRES;
4116 }
4117 }
4118
4119 err = 0;
4120errout:
4121 return err;
4122}
4123
4124struct rt6_nh {
4125 struct rt6_info *rt6_info;
4126 struct fib6_config r_cfg;
4127 struct mx6_config mxc;
4128 struct list_head next;
4129};
4130
4131static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
4132{
4133 struct rt6_nh *nh;
4134
4135 list_for_each_entry(nh, rt6_nh_list, next) {
4136 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6c nexthop %pI6c ifi %d\n",
4137 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
4138 nh->r_cfg.fc_ifindex);
4139 }
4140}
4141
4142static int ip6_route_info_append(struct list_head *rt6_nh_list,
4143 struct rt6_info *rt, struct fib6_config *r_cfg)
4144{
4145 struct rt6_nh *nh;
4146 int err = -EEXIST;
4147
4148 list_for_each_entry(nh, rt6_nh_list, next) {
4149 /* check if rt6_info already exists */
4150 if (rt6_duplicate_nexthop(nh->rt6_info, rt))
4151 return err;
4152 }
4153
4154 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
4155 if (!nh)
4156 return -ENOMEM;
4157 nh->rt6_info = rt;
4158 err = ip6_convert_metrics(&nh->mxc, r_cfg);
4159 if (err) {
4160 kfree(nh);
4161 return err;
4162 }
4163 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
4164 list_add_tail(&nh->next, rt6_nh_list);
4165
4166 return 0;
4167}
4168
4169static void ip6_route_mpath_notify(struct rt6_info *rt,
4170 struct rt6_info *rt_last,
4171 struct nl_info *info,
4172 __u16 nlflags)
4173{
4174 /* if this is an APPEND route, then rt points to the first route
4175 * inserted and rt_last points to last route inserted. Userspace
4176 * wants a consistent dump of the route which starts at the first
4177 * nexthop. Since sibling routes are always added at the end of
4178 * the list, find the first sibling of the last route appended
4179 */
4180 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->rt6i_nsiblings) {
4181 rt = list_first_entry(&rt_last->rt6i_siblings,
4182 struct rt6_info,
4183 rt6i_siblings);
4184 }
4185
4186 if (rt)
4187 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
4188}
4189
4190static int ip6_route_multipath_add(struct fib6_config *cfg,
4191 struct netlink_ext_ack *extack)
4192{
4193 struct rt6_info *rt_notif = NULL, *rt_last = NULL;
4194 struct nl_info *info = &cfg->fc_nlinfo;
4195 struct fib6_config r_cfg;
4196 struct rtnexthop *rtnh;
4197 struct rt6_info *rt;
4198 struct rt6_nh *err_nh;
4199 struct rt6_nh *nh, *nh_safe;
4200 __u16 nlflags;
4201 int remaining;
4202 int attrlen;
4203 int err = 1;
4204 int nhn = 0;
4205 int replace = (cfg->fc_nlinfo.nlh &&
4206 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
4207 LIST_HEAD(rt6_nh_list);
4208
4209 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
4210 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
4211 nlflags |= NLM_F_APPEND;
4212
4213 remaining = cfg->fc_mp_len;
4214 rtnh = (struct rtnexthop *)cfg->fc_mp;
4215
4216 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
4217 * rt6_info structs per nexthop
4218 */
4219 while (rtnh_ok(rtnh, remaining)) {
4220 memcpy(&r_cfg, cfg, sizeof(*cfg));
4221 if (rtnh->rtnh_ifindex)
4222 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
4223
4224 attrlen = rtnh_attrlen(rtnh);
4225 if (attrlen > 0) {
4226 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
4227
4228 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
4229 if (nla) {
4230 r_cfg.fc_gateway = nla_get_in6_addr(nla);
4231 r_cfg.fc_flags |= RTF_GATEWAY;
4232 }
4233 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
4234 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
4235 if (nla)
4236 r_cfg.fc_encap_type = nla_get_u16(nla);
4237 }
4238
4239 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
4240 rt = ip6_route_info_create(&r_cfg, extack);
4241 if (IS_ERR(rt)) {
4242 err = PTR_ERR(rt);
4243 rt = NULL;
4244 goto cleanup;
4245 }
4246
4247 rt->rt6i_nh_weight = rtnh->rtnh_hops + 1;
4248
4249 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
4250 if (err) {
4251 dst_release_immediate(&rt->dst);
4252 goto cleanup;
4253 }
4254
4255 rtnh = rtnh_next(rtnh, &remaining);
4256 }
4257
4258 /* for add and replace send one notification with all nexthops.
4259 * Skip the notification in fib6_add_rt2node and send one with
4260 * the full route when done
4261 */
4262 info->skip_notify = 1;
4263
4264 err_nh = NULL;
4265 list_for_each_entry(nh, &rt6_nh_list, next) {
4266 rt_last = nh->rt6_info;
4267 err = __ip6_ins_rt(nh->rt6_info, info, &nh->mxc, extack);
4268 /* save reference to first route for notification */
4269 if (!rt_notif && !err)
4270 rt_notif = nh->rt6_info;
4271
4272 /* nh->rt6_info is used or freed at this point, reset to NULL*/
4273 nh->rt6_info = NULL;
4274 if (err) {
4275 if (replace && nhn)
4276 ip6_print_replace_route_err(&rt6_nh_list);
4277 err_nh = nh;
4278 goto add_errout;
4279 }
4280
4281 /* Because each route is added like a single route we remove
4282 * these flags after the first nexthop: if there is a collision,
4283 * we have already failed to add the first nexthop:
4284 * fib6_add_rt2node() has rejected it; when replacing, old
4285 * nexthops have been replaced by first new, the rest should
4286 * be added to it.
4287 */
4288 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
4289 NLM_F_REPLACE);
4290 nhn++;
4291 }
4292
4293 /* success ... tell user about new route */
4294 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
4295 goto cleanup;
4296
4297add_errout:
4298 /* send notification for routes that were added so that
4299 * the delete notifications sent by ip6_route_del are
4300 * coherent
4301 */
4302 if (rt_notif)
4303 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
4304
4305 /* Delete routes that were already added */
4306 list_for_each_entry(nh, &rt6_nh_list, next) {
4307 if (err_nh == nh)
4308 break;
4309 ip6_route_del(&nh->r_cfg, extack);
4310 }
4311
4312cleanup:
4313 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
4314 if (nh->rt6_info)
4315 dst_release_immediate(&nh->rt6_info->dst);
4316 kfree(nh->mxc.mx);
4317 list_del(&nh->next);
4318 kfree(nh);
4319 }
4320
4321 return err;
4322}
4323
4324static int ip6_route_multipath_del(struct fib6_config *cfg,
4325 struct netlink_ext_ack *extack)
4326{
4327 struct fib6_config r_cfg;
4328 struct rtnexthop *rtnh;
4329 int remaining;
4330 int attrlen;
4331 int err = 1, last_err = 0;
4332
4333 remaining = cfg->fc_mp_len;
4334 rtnh = (struct rtnexthop *)cfg->fc_mp;
4335
4336 /* Parse a Multipath Entry */
4337 while (rtnh_ok(rtnh, remaining)) {
4338 memcpy(&r_cfg, cfg, sizeof(*cfg));
4339 if (rtnh->rtnh_ifindex)
4340 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
4341
4342 attrlen = rtnh_attrlen(rtnh);
4343 if (attrlen > 0) {
4344 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
4345
4346 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
4347 if (nla) {
4348 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
4349 r_cfg.fc_flags |= RTF_GATEWAY;
4350 }
4351 }
4352 err = ip6_route_del(&r_cfg, extack);
4353 if (err)
4354 last_err = err;
4355
4356 rtnh = rtnh_next(rtnh, &remaining);
4357 }
4358
4359 return last_err;
4360}
4361
4362static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
4363 struct netlink_ext_ack *extack)
4364{
4365 struct fib6_config cfg;
4366 int err;
4367
4368 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
4369 if (err < 0)
4370 return err;
4371
4372 if (cfg.fc_mp)
4373 return ip6_route_multipath_del(&cfg, extack);
4374 else {
4375 cfg.fc_delete_all_nh = 1;
4376 return ip6_route_del(&cfg, extack);
4377 }
4378}
4379
4380static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
4381 struct netlink_ext_ack *extack)
4382{
4383 struct fib6_config cfg;
4384 int err;
4385
4386 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
4387 if (err < 0)
4388 return err;
4389
4390 if (cfg.fc_mp)
4391 return ip6_route_multipath_add(&cfg, extack);
4392 else
4393 return ip6_route_add(&cfg, extack);
4394}
4395
4396static size_t rt6_nlmsg_size(struct rt6_info *rt)
4397{
4398 int nexthop_len = 0;
4399
4400 if (rt->rt6i_nsiblings) {
4401 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
4402 + NLA_ALIGN(sizeof(struct rtnexthop))
4403 + nla_total_size(16) /* RTA_GATEWAY */
4404 + lwtunnel_get_encap_size(rt->dst.lwtstate);
4405
4406 nexthop_len *= rt->rt6i_nsiblings;
4407 }
4408
4409 return NLMSG_ALIGN(sizeof(struct rtmsg))
4410 + nla_total_size(16) /* RTA_SRC */
4411 + nla_total_size(16) /* RTA_DST */
4412 + nla_total_size(16) /* RTA_GATEWAY */
4413 + nla_total_size(16) /* RTA_PREFSRC */
4414 + nla_total_size(4) /* RTA_TABLE */
4415 + nla_total_size(4) /* RTA_IIF */
4416 + nla_total_size(4) /* RTA_OIF */
4417 + nla_total_size(4) /* RTA_PRIORITY */
4418 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
4419 + nla_total_size(sizeof(struct rta_cacheinfo))
4420 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
4421 + nla_total_size(1) /* RTA_PREF */
4422 + lwtunnel_get_encap_size(rt->dst.lwtstate)
4423 + nexthop_len;
4424}
4425
4426static int rt6_nexthop_info(struct sk_buff *skb, struct rt6_info *rt,
4427 unsigned int *flags, bool skip_oif)
4428{
4429 if (rt->rt6i_nh_flags & RTNH_F_DEAD)
4430 *flags |= RTNH_F_DEAD;
4431
4432 if (rt->rt6i_nh_flags & RTNH_F_LINKDOWN) {
4433 *flags |= RTNH_F_LINKDOWN;
4434 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
4435 *flags |= RTNH_F_DEAD;
4436 }
4437
4438 if (rt->rt6i_flags & RTF_GATEWAY) {
4439 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
4440 goto nla_put_failure;
4441 }
4442
4443 *flags |= (rt->rt6i_nh_flags & RTNH_F_ONLINK);
4444 if (rt->rt6i_nh_flags & RTNH_F_OFFLOAD)
4445 *flags |= RTNH_F_OFFLOAD;
4446
4447 /* not needed for multipath encoding b/c it has a rtnexthop struct */
4448 if (!skip_oif && rt->dst.dev &&
4449 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
4450 goto nla_put_failure;
4451
4452 if (rt->dst.lwtstate &&
4453 lwtunnel_fill_encap(skb, rt->dst.lwtstate) < 0)
4454 goto nla_put_failure;
4455
4456 return 0;
4457
4458nla_put_failure:
4459 return -EMSGSIZE;
4460}
4461
4462/* add multipath next hop */
4463static int rt6_add_nexthop(struct sk_buff *skb, struct rt6_info *rt)
4464{
4465 struct rtnexthop *rtnh;
4466 unsigned int flags = 0;
4467
4468 rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
4469 if (!rtnh)
4470 goto nla_put_failure;
4471
4472 rtnh->rtnh_hops = rt->rt6i_nh_weight - 1;
4473 rtnh->rtnh_ifindex = rt->dst.dev ? rt->dst.dev->ifindex : 0;
4474
4475 if (rt6_nexthop_info(skb, rt, &flags, true) < 0)
4476 goto nla_put_failure;
4477
4478 rtnh->rtnh_flags = flags;
4479
4480 /* length of rtnetlink header + attributes */
4481 rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
4482
4483 return 0;
4484
4485nla_put_failure:
4486 return -EMSGSIZE;
4487}
4488
4489static int rt6_fill_node(struct net *net,
4490 struct sk_buff *skb, struct rt6_info *rt,
4491 struct in6_addr *dst, struct in6_addr *src,
4492 int iif, int type, u32 portid, u32 seq,
4493 unsigned int flags)
4494{
4495 u32 metrics[RTAX_MAX];
4496 struct rtmsg *rtm;
4497 struct nlmsghdr *nlh;
4498 long expires;
4499 u32 table;
4500
4501 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
4502 if (!nlh)
4503 return -EMSGSIZE;
4504
4505 rtm = nlmsg_data(nlh);
4506 rtm->rtm_family = AF_INET6;
4507 rtm->rtm_dst_len = rt->rt6i_dst.plen;
4508 rtm->rtm_src_len = rt->rt6i_src.plen;
4509 rtm->rtm_tos = 0;
4510 if (rt->rt6i_table)
4511 table = rt->rt6i_table->tb6_id;
4512 else
4513 table = RT6_TABLE_UNSPEC;
4514 rtm->rtm_table = table;
4515 if (nla_put_u32(skb, RTA_TABLE, table))
4516 goto nla_put_failure;
4517 if (rt->rt6i_flags & RTF_REJECT) {
4518 switch (rt->dst.error) {
4519 case -EINVAL:
4520 rtm->rtm_type = RTN_BLACKHOLE;
4521 break;
4522 case -EACCES:
4523 rtm->rtm_type = RTN_PROHIBIT;
4524 break;
4525 case -EAGAIN:
4526 rtm->rtm_type = RTN_THROW;
4527 break;
4528 default:
4529 rtm->rtm_type = RTN_UNREACHABLE;
4530 break;
4531 }
4532 }
4533 else if (rt->rt6i_flags & RTF_LOCAL)
4534 rtm->rtm_type = RTN_LOCAL;
4535 else if (rt->rt6i_flags & RTF_ANYCAST)
4536 rtm->rtm_type = RTN_ANYCAST;
4537 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
4538 rtm->rtm_type = RTN_LOCAL;
4539 else
4540 rtm->rtm_type = RTN_UNICAST;
4541 rtm->rtm_flags = 0;
4542 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
4543 rtm->rtm_protocol = rt->rt6i_protocol;
4544
4545 if (rt->rt6i_flags & RTF_CACHE)
4546 rtm->rtm_flags |= RTM_F_CLONED;
4547
4548 if (dst) {
4549 if (nla_put_in6_addr(skb, RTA_DST, dst))
4550 goto nla_put_failure;
4551 rtm->rtm_dst_len = 128;
4552 } else if (rtm->rtm_dst_len)
4553 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
4554 goto nla_put_failure;
4555#ifdef CONFIG_IPV6_SUBTREES
4556 if (src) {
4557 if (nla_put_in6_addr(skb, RTA_SRC, src))
4558 goto nla_put_failure;
4559 rtm->rtm_src_len = 128;
4560 } else if (rtm->rtm_src_len &&
4561 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
4562 goto nla_put_failure;
4563#endif
4564 if (iif) {
4565#ifdef CONFIG_IPV6_MROUTE
4566 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
4567 int err = ip6mr_get_route(net, skb, rtm, portid);
4568
4569 if (err == 0)
4570 return 0;
4571 if (err < 0)
4572 goto nla_put_failure;
4573 } else
4574#endif
4575 if (nla_put_u32(skb, RTA_IIF, iif))
4576 goto nla_put_failure;
4577 } else if (dst) {
4578 struct in6_addr saddr_buf;
4579 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
4580 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
4581 goto nla_put_failure;
4582 }
4583
4584 if (rt->rt6i_prefsrc.plen) {
4585 struct in6_addr saddr_buf;
4586 saddr_buf = rt->rt6i_prefsrc.addr;
4587 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
4588 goto nla_put_failure;
4589 }
4590
4591 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
4592 if (rt->rt6i_pmtu)
4593 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
4594 if (rtnetlink_put_metrics(skb, metrics) < 0)
4595 goto nla_put_failure;
4596
4597 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
4598 goto nla_put_failure;
4599
4600 /* For multipath routes, walk the siblings list and add
4601 * each as a nexthop within RTA_MULTIPATH.
4602 */
4603 if (rt->rt6i_nsiblings) {
4604 struct rt6_info *sibling, *next_sibling;
4605 struct nlattr *mp;
4606
4607 mp = nla_nest_start(skb, RTA_MULTIPATH);
4608 if (!mp)
4609 goto nla_put_failure;
4610
4611 if (rt6_add_nexthop(skb, rt) < 0)
4612 goto nla_put_failure;
4613
4614 list_for_each_entry_safe(sibling, next_sibling,
4615 &rt->rt6i_siblings, rt6i_siblings) {
4616 if (rt6_add_nexthop(skb, sibling) < 0)
4617 goto nla_put_failure;
4618 }
4619
4620 nla_nest_end(skb, mp);
4621 } else {
4622 if (rt6_nexthop_info(skb, rt, &rtm->rtm_flags, false) < 0)
4623 goto nla_put_failure;
4624 }
4625
4626 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
4627
4628 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
4629 goto nla_put_failure;
4630
4631 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
4632 goto nla_put_failure;
4633
4634
4635 nlmsg_end(skb, nlh);
4636 return 0;
4637
4638nla_put_failure:
4639 nlmsg_cancel(skb, nlh);
4640 return -EMSGSIZE;
4641}
4642
4643int rt6_dump_route(struct rt6_info *rt, void *p_arg)
4644{
4645 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
4646 struct net *net = arg->net;
4647
4648 if (rt == net->ipv6.ip6_null_entry)
4649 return 0;
4650
4651 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
4652 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
4653
4654 /* user wants prefix routes only */
4655 if (rtm->rtm_flags & RTM_F_PREFIX &&
4656 !(rt->rt6i_flags & RTF_PREFIX_RT)) {
4657 /* success since this is not a prefix route */
4658 return 1;
4659 }
4660 }
4661
4662 return rt6_fill_node(net,
4663 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
4664 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
4665 NLM_F_MULTI);
4666}
4667
4668static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4669 struct netlink_ext_ack *extack)
4670{
4671 struct net *net = sock_net(in_skb->sk);
4672 struct nlattr *tb[RTA_MAX+1];
4673 int err, iif = 0, oif = 0;
4674 struct dst_entry *dst;
4675 struct rt6_info *rt;
4676 struct sk_buff *skb;
4677 struct rtmsg *rtm;
4678 struct flowi6 fl6;
4679 bool fibmatch;
4680
4681 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
4682 extack);
4683 if (err < 0)
4684 goto errout;
4685
4686 err = -EINVAL;
4687 memset(&fl6, 0, sizeof(fl6));
4688 rtm = nlmsg_data(nlh);
4689 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
4690 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
4691
4692 if (tb[RTA_SRC]) {
4693 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
4694 goto errout;
4695
4696 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
4697 }
4698
4699 if (tb[RTA_DST]) {
4700 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
4701 goto errout;
4702
4703 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
4704 }
4705
4706 if (tb[RTA_IIF])
4707 iif = nla_get_u32(tb[RTA_IIF]);
4708
4709 if (tb[RTA_OIF])
4710 oif = nla_get_u32(tb[RTA_OIF]);
4711
4712 if (tb[RTA_MARK])
4713 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
4714
4715 if (tb[RTA_UID])
4716 fl6.flowi6_uid = make_kuid(current_user_ns(),
4717 nla_get_u32(tb[RTA_UID]));
4718 else
4719 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
4720
4721 if (iif) {
4722 struct net_device *dev;
4723 int flags = 0;
4724
4725 rcu_read_lock();
4726
4727 dev = dev_get_by_index_rcu(net, iif);
4728 if (!dev) {
4729 rcu_read_unlock();
4730 err = -ENODEV;
4731 goto errout;
4732 }
4733
4734 fl6.flowi6_iif = iif;
4735
4736 if (!ipv6_addr_any(&fl6.saddr))
4737 flags |= RT6_LOOKUP_F_HAS_SADDR;
4738
4739 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
4740
4741 rcu_read_unlock();
4742 } else {
4743 fl6.flowi6_oif = oif;
4744
4745 dst = ip6_route_output(net, NULL, &fl6);
4746 }
4747
4748
4749 rt = container_of(dst, struct rt6_info, dst);
4750 if (rt->dst.error) {
4751 err = rt->dst.error;
4752 ip6_rt_put(rt);
4753 goto errout;
4754 }
4755
4756 if (rt == net->ipv6.ip6_null_entry) {
4757 err = rt->dst.error;
4758 ip6_rt_put(rt);
4759 goto errout;
4760 }
4761
4762 if (fibmatch && rt->from) {
4763 struct rt6_info *ort = rt->from;
4764
4765 dst_hold(&ort->dst);
4766 ip6_rt_put(rt);
4767 rt = ort;
4768 }
4769
4770 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
4771 if (!skb) {
4772 ip6_rt_put(rt);
4773 err = -ENOBUFS;
4774 goto errout;
4775 }
4776
4777 skb_dst_set(skb, &rt->dst);
4778 if (fibmatch)
4779 err = rt6_fill_node(net, skb, rt, NULL, NULL, iif,
4780 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
4781 nlh->nlmsg_seq, 0);
4782 else
4783 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
4784 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
4785 nlh->nlmsg_seq, 0);
4786 if (err < 0) {
4787 kfree_skb(skb);
4788 goto errout;
4789 }
4790
4791 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4792errout:
4793 return err;
4794}
4795
4796void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
4797 unsigned int nlm_flags)
4798{
4799 struct sk_buff *skb;
4800 struct net *net = info->nl_net;
4801 u32 seq;
4802 int err;
4803
4804 err = -ENOBUFS;
4805 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
4806
4807 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
4808 if (!skb)
4809 goto errout;
4810
4811 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
4812 event, info->portid, seq, nlm_flags);
4813 if (err < 0) {
4814 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
4815 WARN_ON(err == -EMSGSIZE);
4816 kfree_skb(skb);
4817 goto errout;
4818 }
4819 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
4820 info->nlh, gfp_any());
4821 return;
4822errout:
4823 if (err < 0)
4824 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
4825}
4826
4827static int ip6_route_dev_notify(struct notifier_block *this,
4828 unsigned long event, void *ptr)
4829{
4830 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4831 struct net *net = dev_net(dev);
4832
4833 if (!(dev->flags & IFF_LOOPBACK))
4834 return NOTIFY_OK;
4835
4836 if (event == NETDEV_REGISTER) {
4837 net->ipv6.ip6_null_entry->dst.dev = dev;
4838 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
4839#ifdef CONFIG_IPV6_MULTIPLE_TABLES
4840 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
4841 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
4842 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
4843 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
4844#endif
4845 } else if (event == NETDEV_UNREGISTER &&
4846 dev->reg_state != NETREG_UNREGISTERED) {
4847 /* NETDEV_UNREGISTER could be fired for multiple times by
4848 * netdev_wait_allrefs(). Make sure we only call this once.
4849 */
4850 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
4851#ifdef CONFIG_IPV6_MULTIPLE_TABLES
4852 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
4853 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
4854#endif
4855 }
4856
4857 return NOTIFY_OK;
4858}
4859
4860/*
4861 * /proc
4862 */
4863
4864#ifdef CONFIG_PROC_FS
4865
4866static const struct file_operations ipv6_route_proc_fops = {
4867 .open = ipv6_route_open,
4868 .read = seq_read,
4869 .llseek = seq_lseek,
4870 .release = seq_release_net,
4871};
4872
4873static int rt6_stats_seq_show(struct seq_file *seq, void *v)
4874{
4875 struct net *net = (struct net *)seq->private;
4876 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
4877 net->ipv6.rt6_stats->fib_nodes,
4878 net->ipv6.rt6_stats->fib_route_nodes,
4879 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
4880 net->ipv6.rt6_stats->fib_rt_entries,
4881 net->ipv6.rt6_stats->fib_rt_cache,
4882 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
4883 net->ipv6.rt6_stats->fib_discarded_routes);
4884
4885 return 0;
4886}
4887
4888static int rt6_stats_seq_open(struct inode *inode, struct file *file)
4889{
4890 return single_open_net(inode, file, rt6_stats_seq_show);
4891}
4892
4893static const struct file_operations rt6_stats_seq_fops = {
4894 .open = rt6_stats_seq_open,
4895 .read = seq_read,
4896 .llseek = seq_lseek,
4897 .release = single_release_net,
4898};
4899#endif /* CONFIG_PROC_FS */
4900
4901#ifdef CONFIG_SYSCTL
4902
4903static
4904int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
4905 void __user *buffer, size_t *lenp, loff_t *ppos)
4906{
4907 struct net *net;
4908 int delay;
4909 if (!write)
4910 return -EINVAL;
4911
4912 net = (struct net *)ctl->extra1;
4913 delay = net->ipv6.sysctl.flush_delay;
4914 proc_dointvec(ctl, write, buffer, lenp, ppos);
4915 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
4916 return 0;
4917}
4918
4919struct ctl_table ipv6_route_table_template[] = {
4920 {
4921 .procname = "flush",
4922 .data = &init_net.ipv6.sysctl.flush_delay,
4923 .maxlen = sizeof(int),
4924 .mode = 0200,
4925 .proc_handler = ipv6_sysctl_rtcache_flush
4926 },
4927 {
4928 .procname = "gc_thresh",
4929 .data = &ip6_dst_ops_template.gc_thresh,
4930 .maxlen = sizeof(int),
4931 .mode = 0644,
4932 .proc_handler = proc_dointvec,
4933 },
4934 {
4935 .procname = "max_size",
4936 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
4937 .maxlen = sizeof(int),
4938 .mode = 0644,
4939 .proc_handler = proc_dointvec,
4940 },
4941 {
4942 .procname = "gc_min_interval",
4943 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
4944 .maxlen = sizeof(int),
4945 .mode = 0644,
4946 .proc_handler = proc_dointvec_jiffies,
4947 },
4948 {
4949 .procname = "gc_timeout",
4950 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
4951 .maxlen = sizeof(int),
4952 .mode = 0644,
4953 .proc_handler = proc_dointvec_jiffies,
4954 },
4955 {
4956 .procname = "gc_interval",
4957 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
4958 .maxlen = sizeof(int),
4959 .mode = 0644,
4960 .proc_handler = proc_dointvec_jiffies,
4961 },
4962 {
4963 .procname = "gc_elasticity",
4964 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
4965 .maxlen = sizeof(int),
4966 .mode = 0644,
4967 .proc_handler = proc_dointvec,
4968 },
4969 {
4970 .procname = "mtu_expires",
4971 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
4972 .maxlen = sizeof(int),
4973 .mode = 0644,
4974 .proc_handler = proc_dointvec_jiffies,
4975 },
4976 {
4977 .procname = "min_adv_mss",
4978 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
4979 .maxlen = sizeof(int),
4980 .mode = 0644,
4981 .proc_handler = proc_dointvec,
4982 },
4983 {
4984 .procname = "gc_min_interval_ms",
4985 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
4986 .maxlen = sizeof(int),
4987 .mode = 0644,
4988 .proc_handler = proc_dointvec_ms_jiffies,
4989 },
4990 { }
4991};
4992
4993struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
4994{
4995 struct ctl_table *table;
4996
4997 table = kmemdup(ipv6_route_table_template,
4998 sizeof(ipv6_route_table_template),
4999 GFP_KERNEL);
5000
5001 if (table) {
5002 table[0].data = &net->ipv6.sysctl.flush_delay;
5003 table[0].extra1 = net;
5004 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
5005 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
5006 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
5007 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
5008 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
5009 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
5010 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
5011 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
5012 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
5013
5014 /* Don't export sysctls to unprivileged users */
5015 if (net->user_ns != &init_user_ns)
5016 table[0].procname = NULL;
5017 }
5018
5019 return table;
5020}
5021#endif
5022
5023static int __net_init ip6_route_net_init(struct net *net)
5024{
5025 int ret = -ENOMEM;
5026
5027 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
5028 sizeof(net->ipv6.ip6_dst_ops));
5029
5030 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
5031 goto out_ip6_dst_ops;
5032
5033 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
5034 sizeof(*net->ipv6.ip6_null_entry),
5035 GFP_KERNEL);
5036 if (!net->ipv6.ip6_null_entry)
5037 goto out_ip6_dst_entries;
5038 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
5039 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
5040 ip6_template_metrics, true);
5041
5042#ifdef CONFIG_IPV6_MULTIPLE_TABLES
5043 net->ipv6.fib6_has_custom_rules = false;
5044 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
5045 sizeof(*net->ipv6.ip6_prohibit_entry),
5046 GFP_KERNEL);
5047 if (!net->ipv6.ip6_prohibit_entry)
5048 goto out_ip6_null_entry;
5049 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
5050 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
5051 ip6_template_metrics, true);
5052
5053 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
5054 sizeof(*net->ipv6.ip6_blk_hole_entry),
5055 GFP_KERNEL);
5056 if (!net->ipv6.ip6_blk_hole_entry)
5057 goto out_ip6_prohibit_entry;
5058 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
5059 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
5060 ip6_template_metrics, true);
5061#endif
5062
5063 net->ipv6.sysctl.flush_delay = 0;
5064 net->ipv6.sysctl.ip6_rt_max_size = 4096;
5065 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
5066 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
5067 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
5068 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
5069 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
5070 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
5071
5072 net->ipv6.ip6_rt_gc_expire = 30*HZ;
5073
5074 ret = 0;
5075out:
5076 return ret;
5077
5078#ifdef CONFIG_IPV6_MULTIPLE_TABLES
5079out_ip6_prohibit_entry:
5080 kfree(net->ipv6.ip6_prohibit_entry);
5081out_ip6_null_entry:
5082 kfree(net->ipv6.ip6_null_entry);
5083#endif
5084out_ip6_dst_entries:
5085 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
5086out_ip6_dst_ops:
5087 goto out;
5088}
5089
5090static void __net_exit ip6_route_net_exit(struct net *net)
5091{
5092 kfree(net->ipv6.ip6_null_entry);
5093#ifdef CONFIG_IPV6_MULTIPLE_TABLES
5094 kfree(net->ipv6.ip6_prohibit_entry);
5095 kfree(net->ipv6.ip6_blk_hole_entry);
5096#endif
5097 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
5098}
5099
5100static int __net_init ip6_route_net_init_late(struct net *net)
5101{
5102#ifdef CONFIG_PROC_FS
5103 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
5104 proc_create("rt6_stats", 0444, net->proc_net, &rt6_stats_seq_fops);
5105#endif
5106 return 0;
5107}
5108
5109static void __net_exit ip6_route_net_exit_late(struct net *net)
5110{
5111#ifdef CONFIG_PROC_FS
5112 remove_proc_entry("ipv6_route", net->proc_net);
5113 remove_proc_entry("rt6_stats", net->proc_net);
5114#endif
5115}
5116
5117static struct pernet_operations ip6_route_net_ops = {
5118 .init = ip6_route_net_init,
5119 .exit = ip6_route_net_exit,
5120};
5121
5122static int __net_init ipv6_inetpeer_init(struct net *net)
5123{
5124 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
5125
5126 if (!bp)
5127 return -ENOMEM;
5128 inet_peer_base_init(bp);
5129 net->ipv6.peers = bp;
5130 return 0;
5131}
5132
5133static void __net_exit ipv6_inetpeer_exit(struct net *net)
5134{
5135 struct inet_peer_base *bp = net->ipv6.peers;
5136
5137 net->ipv6.peers = NULL;
5138 inetpeer_invalidate_tree(bp);
5139 kfree(bp);
5140}
5141
5142static struct pernet_operations ipv6_inetpeer_ops = {
5143 .init = ipv6_inetpeer_init,
5144 .exit = ipv6_inetpeer_exit,
5145};
5146
5147static struct pernet_operations ip6_route_net_late_ops = {
5148 .init = ip6_route_net_init_late,
5149 .exit = ip6_route_net_exit_late,
5150};
5151
5152static struct notifier_block ip6_route_dev_notifier = {
5153 .notifier_call = ip6_route_dev_notify,
5154 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
5155};
5156
5157void __init ip6_route_init_special_entries(void)
5158{
5159 /* Registering of the loopback is done before this portion of code,
5160 * the loopback reference in rt6_info will not be taken, do it
5161 * manually for init_net */
5162 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
5163 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
5164 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
5165 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
5166 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
5167 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
5168 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
5169 #endif
5170}
5171
5172int __init ip6_route_init(void)
5173{
5174 int ret;
5175 int cpu;
5176
5177 ret = -ENOMEM;
5178 ip6_dst_ops_template.kmem_cachep =
5179 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
5180 SLAB_HWCACHE_ALIGN, NULL);
5181 if (!ip6_dst_ops_template.kmem_cachep)
5182 goto out;
5183
5184 ret = dst_entries_init(&ip6_dst_blackhole_ops);
5185 if (ret)
5186 goto out_kmem_cache;
5187
5188 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
5189 if (ret)
5190 goto out_dst_entries;
5191
5192 ret = register_pernet_subsys(&ip6_route_net_ops);
5193 if (ret)
5194 goto out_register_inetpeer;
5195
5196 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
5197
5198 ret = fib6_init();
5199 if (ret)
5200 goto out_register_subsys;
5201
5202 ret = xfrm6_init();
5203 if (ret)
5204 goto out_fib6_init;
5205
5206 ret = fib6_rules_init();
5207 if (ret)
5208 goto xfrm6_init;
5209
5210 ret = register_pernet_subsys(&ip6_route_net_late_ops);
5211 if (ret)
5212 goto fib6_rules_init;
5213
5214 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
5215 inet6_rtm_newroute, NULL, 0);
5216 if (ret < 0)
5217 goto out_register_late_subsys;
5218
5219 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
5220 inet6_rtm_delroute, NULL, 0);
5221 if (ret < 0)
5222 goto out_register_late_subsys;
5223
5224 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
5225 inet6_rtm_getroute, NULL,
5226 RTNL_FLAG_DOIT_UNLOCKED);
5227 if (ret < 0)
5228 goto out_register_late_subsys;
5229
5230 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
5231 if (ret)
5232 goto out_register_late_subsys;
5233
5234 for_each_possible_cpu(cpu) {
5235 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
5236
5237 INIT_LIST_HEAD(&ul->head);
5238 spin_lock_init(&ul->lock);
5239 }
5240
5241out:
5242 return ret;
5243
5244out_register_late_subsys:
5245 rtnl_unregister_all(PF_INET6);
5246 unregister_pernet_subsys(&ip6_route_net_late_ops);
5247fib6_rules_init:
5248 fib6_rules_cleanup();
5249xfrm6_init:
5250 xfrm6_fini();
5251out_fib6_init:
5252 fib6_gc_cleanup();
5253out_register_subsys:
5254 unregister_pernet_subsys(&ip6_route_net_ops);
5255out_register_inetpeer:
5256 unregister_pernet_subsys(&ipv6_inetpeer_ops);
5257out_dst_entries:
5258 dst_entries_destroy(&ip6_dst_blackhole_ops);
5259out_kmem_cache:
5260 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
5261 goto out;
5262}
5263
5264void ip6_route_cleanup(void)
5265{
5266 unregister_netdevice_notifier(&ip6_route_dev_notifier);
5267 unregister_pernet_subsys(&ip6_route_net_late_ops);
5268 fib6_rules_cleanup();
5269 xfrm6_fini();
5270 fib6_gc_cleanup();
5271 unregister_pernet_subsys(&ipv6_inetpeer_ops);
5272 unregister_pernet_subsys(&ip6_route_net_ops);
5273 dst_entries_destroy(&ip6_dst_blackhole_ops);
5274 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
5275}