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
71 static int ip6_rt_type_to_error(u8 fib6_type);
72
73 #define CREATE_TRACE_POINTS
74 #include <trace/events/fib6.h>
75 EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
76 #undef CREATE_TRACE_POINTS
77
78 enum 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
85 INDIRECT_CALLABLE_SCOPE
86 struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
87 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
88 INDIRECT_CALLABLE_SCOPE
89 unsigned int ip6_mtu(const struct dst_entry *dst);
90 static void ip6_negative_advice(struct sock *sk,
91 struct dst_entry *dst);
92 static void ip6_dst_destroy(struct dst_entry *);
93 static void ip6_dst_ifdown(struct dst_entry *,
94 struct net_device *dev);
95 static void ip6_dst_gc(struct dst_ops *ops);
96
97 static int ip6_pkt_discard(struct sk_buff *skb);
98 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
99 static int ip6_pkt_prohibit(struct sk_buff *skb);
100 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
101 static void ip6_link_failure(struct sk_buff *skb);
102 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
103 struct sk_buff *skb, u32 mtu,
104 bool confirm_neigh);
105 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
106 struct sk_buff *skb);
107 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
108 int strict);
109 static size_t rt6_nlmsg_size(struct fib6_info *f6i);
110 static 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);
115 static 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
120 static 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);
125 static 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
131 struct uncached_list {
132 spinlock_t lock;
133 struct list_head head;
134 };
135
136 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
137
rt6_uncached_list_add(struct rt6_info * rt)138 void 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
rt6_uncached_list_del(struct rt6_info * rt)149 void 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
rt6_uncached_list_flush_dev(struct net_device * dev)160 static 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
choose_neigh_daddr(const struct in6_addr * p,struct sk_buff * skb,const void * daddr)197 static 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
ip6_neigh_lookup(const struct in6_addr * gw,struct net_device * dev,struct sk_buff * skb,const void * daddr)208 struct 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
ip6_dst_neigh_lookup(const struct dst_entry * dst,struct sk_buff * skb,const void * daddr)224 static 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
ip6_confirm_neigh(const struct dst_entry * dst,const void * daddr)234 static 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
249 static 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
268 static 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
280 static const u32 ip6_template_metrics[RTAX_MAX] = {
281 [RTAX_HOPLIMIT - 1] = 0,
282 };
283
284 static 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
293 static 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
307 static 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
319 static 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
rt6_info_init(struct rt6_info * rt)333 static void rt6_info_init(struct rt6_info *rt)
334 {
335 memset_after(rt, 0, dst);
336 }
337
338 /* allocate dst with ip6_dst_ops */
ip6_dst_alloc(struct net * net,struct net_device * dev,int flags)339 struct 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 }
352 EXPORT_SYMBOL(ip6_dst_alloc);
353
ip6_dst_destroy(struct dst_entry * dst)354 static 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
ip6_dst_ifdown(struct dst_entry * dst,struct net_device * dev)373 static 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
__rt6_check_expired(const struct rt6_info * rt)391 static 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
rt6_check_expired(const struct rt6_info * rt)399 static 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
fib6_select_path(const struct net * net,struct fib6_result * res,struct flowi6 * fl6,int oif,bool have_oif_match,const struct sk_buff * skb,int strict)415 void 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
460 out:
461 res->f6i = match;
462 res->nh = match->fib6_nh;
463 }
464
465 /*
466 * Route lookup. rcu_read_lock() should be held.
467 */
468
__rt6_device_match(struct net * net,const struct fib6_nh * nh,const struct in6_addr * saddr,int oif,int flags)469 static 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
490 struct 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
__rt6_nh_dev_match(struct fib6_nh * nh,void * _arg)498 static 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 */
rt6_nh_dev_match(struct net * net,struct nexthop * nh,struct fib6_result * res,const struct in6_addr * saddr,int oif,int flags)508 static 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
rt6_device_match(struct net * net,struct fib6_result * res,const struct in6_addr * saddr,int oif,int flags)529 static 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 }
585 out:
586 res->nh = nh;
587 res->fib6_type = res->f6i->fib6_type;
588 res->fib6_flags = res->f6i->fib6_flags;
589 return;
590
591 out_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
598 struct __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
rt6_probe_deferred(struct work_struct * w)605 static 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
rt6_probe(struct fib6_nh * fib6_nh)617 static 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
675 out:
676 rcu_read_unlock();
677 }
678 #else
rt6_probe(struct fib6_nh * fib6_nh)679 static 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 */
rt6_check_neigh(const struct fib6_nh * fib6_nh)687 static 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
rt6_score_route(const struct fib6_nh * nh,u32 fib6_flags,int oif,int strict)715 static 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
find_match(struct fib6_nh * nh,u32 fib6_flags,int oif,int strict,int * mpri,bool * do_rr)737 static 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 }
769 out:
770 return rc;
771 }
772
773 struct 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
rt6_nh_find_match(struct fib6_nh * nh,void * _arg)782 static 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
__find_rr_leaf(struct fib6_info * f6i_start,struct fib6_info * nomatch,u32 metric,struct fib6_result * res,struct fib6_info ** cont,int oif,int strict,bool * do_rr,int * mpri)791 static 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
find_rr_leaf(struct fib6_node * fn,struct fib6_info * leaf,struct fib6_info * rr_head,int oif,int strict,bool * do_rr,struct fib6_result * res)848 static 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
rt6_select(struct net * net,struct fib6_node * fn,int oif,struct fib6_result * res,int strict)869 static 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
917 out:
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
rt6_is_gw_or_nonexthop(const struct fib6_result * res)926 static 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
rt6_route_rcv(struct net_device * dev,u8 * opt,int len,const struct in6_addr * gwaddr)933 int 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 */
ip6_rt_get_dev_rcu(const struct fib6_result * res)1022 static 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
1044 static 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
ip6_rt_type_to_error(u8 fib6_type)1059 static int ip6_rt_type_to_error(u8 fib6_type)
1060 {
1061 return fib6_prop[fib6_type];
1062 }
1063
fib6_info_dst_flags(struct fib6_info * rt)1064 static 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
ip6_rt_init_dst_reject(struct rt6_info * rt,u8 fib6_type)1076 static 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
ip6_rt_init_dst(struct rt6_info * rt,const struct fib6_result * res)1098 static 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 */
rt6_set_from(struct rt6_info * rt,struct fib6_info * from)1127 static 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 */
ip6_rt_copy_init(struct rt6_info * rt,const struct fib6_result * res)1135 static 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
fib6_backtrack(struct fib6_node * fn,struct in6_addr * saddr)1156 static 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
ip6_hold_safe(struct net * net,struct rt6_info ** prt)1174 static 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 */
ip6_create_rt_rcu(const struct fib6_result * res)1191 static 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
1211 fallback:
1212 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1213 dst_hold(&nrt->dst);
1214 return nrt;
1215 }
1216
ip6_pol_route_lookup(struct net * net,struct fib6_table * table,struct flowi6 * fl6,const struct sk_buff * skb,int flags)1217 INDIRECT_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);
1229 restart:
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 {
1258 do_create:
1259 rt = ip6_create_rt_rcu(&res);
1260 }
1261
1262 out:
1263 trace_fib6_table_lookup(net, &res, table, fl6);
1264
1265 rcu_read_unlock();
1266
1267 return rt;
1268 }
1269
ip6_route_lookup(struct net * net,struct flowi6 * fl6,const struct sk_buff * skb,int flags)1270 struct 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 }
1275 EXPORT_SYMBOL_GPL(ip6_route_lookup);
1276
rt6_lookup(struct net * net,const struct in6_addr * daddr,const struct in6_addr * saddr,int oif,const struct sk_buff * skb,int strict)1277 struct 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 }
1301 EXPORT_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
__ip6_ins_rt(struct fib6_info * rt,struct nl_info * info,struct netlink_ext_ack * extack)1309 static 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
ip6_ins_rt(struct net * net,struct fib6_info * rt)1323 int 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
ip6_rt_cache_alloc(const struct fib6_result * res,const struct in6_addr * daddr,const struct in6_addr * saddr)1330 static 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
ip6_rt_pcpu_alloc(const struct fib6_result * res)1372 static 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
rt6_is_valid(const struct rt6_info * rt6)1399 static 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 */
rt6_get_pcpu_route(const struct fib6_result * res)1405 static 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
rt6_make_pcpu_route(struct net * net,const struct fib6_result * res)1428 static 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 */
1453 static 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 */
rt6_remove_exception(struct rt6_exception_bucket * bucket,struct rt6_exception * rt6_ex)1458 static 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 */
rt6_exception_remove_oldest(struct rt6_exception_bucket * bucket)1484 static 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
rt6_exception_hash(const struct in6_addr * dst,const struct in6_addr * src)1498 static 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 */
1526 static struct rt6_exception *
__rt6_find_exception_spinlock(struct rt6_exception_bucket ** bucket,const struct in6_addr * daddr,const struct in6_addr * saddr)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 */
1559 static struct rt6_exception *
__rt6_find_exception_rcu(struct rt6_exception_bucket ** bucket,const struct in6_addr * daddr,const struct in6_addr * saddr)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
fib6_mtu(const struct fib6_result * res)1589 static 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 */
1618 static
fib6_nh_get_excptn_bucket(const struct fib6_nh * nh,spinlock_t * lock)1619 struct 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
fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket * bucket)1641 static 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 */
fib6_nh_excptn_bucket_set_flushed(struct fib6_nh * nh,spinlock_t * lock)1649 static 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
rt6_insert_exception(struct rt6_info * nrt,const struct fib6_result * res)1664 static 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
1733 out:
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
fib6_nh_flush_exceptions(struct fib6_nh * nh,struct fib6_info * from)1747 static 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 }
1773 out:
1774 spin_unlock_bh(&rt6_exception_lock);
1775 }
1776
rt6_nh_flush_exceptions(struct fib6_nh * nh,void * arg)1777 static 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
rt6_flush_exceptions(struct fib6_info * f6i)1786 void 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 */
rt6_find_cached_rt(const struct fib6_result * res,const struct in6_addr * daddr,const struct in6_addr * saddr)1798 static 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;
1821 find_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 */
fib6_nh_remove_exception(const struct fib6_nh * nh,int plen,const struct rt6_info * rt)1841 static 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
1879 struct fib6_nh_excptn_arg {
1880 struct rt6_info *rt;
1881 int plen;
1882 };
1883
rt6_nh_remove_exception_rt(struct fib6_nh * nh,void * _arg)1884 static 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
rt6_remove_exception_rt(struct rt6_info * rt)1896 static 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 */
fib6_nh_update_exception(const struct fib6_nh * nh,int plen,const struct rt6_info * rt)1925 static 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
1948 struct 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 */
fib6_nh_find_match(struct fib6_nh * nh,void * _arg)1955 static 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
rt6_update_exception_stamp_rt(struct rt6_info * rt)1971 static 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);
1997 unlock:
1998 rcu_read_unlock();
1999 }
2000
rt6_mtu_change_route_allowed(struct inet6_dev * idev,struct rt6_info * rt,int mtu)2001 static 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
rt6_exceptions_update_pmtu(struct inet6_dev * idev,const struct fib6_nh * nh,int mtu)2023 static 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
fib6_nh_exceptions_clean_tohost(const struct fib6_nh * nh,const struct in6_addr * gateway)2052 static 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
rt6_age_examine_exception(struct rt6_exception_bucket * bucket,struct rt6_exception * rt6_ex,struct fib6_gc_args * gc_args,unsigned long now)2085 static 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
fib6_nh_age_exceptions(const struct fib6_nh * nh,struct fib6_gc_args * gc_args,unsigned long now)2126 static 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
2155 struct fib6_nh_age_excptn_arg {
2156 struct fib6_gc_args *gc_args;
2157 unsigned long now;
2158 };
2159
rt6_nh_age_exceptions(struct fib6_nh * nh,void * _arg)2160 static 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
rt6_age_exceptions(struct fib6_info * f6i,struct fib6_gc_args * gc_args,unsigned long now)2168 void 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 */
fib6_table_lookup(struct net * net,struct fib6_table * table,int oif,struct flowi6 * fl6,struct fib6_result * res,int strict)2186 int 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
2194 redo_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
ip6_pol_route(struct net * net,struct fib6_table * table,int oif,struct flowi6 * fl6,const struct sk_buff * skb,int flags)2213 struct 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 }
2271 out:
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 }
2280 EXPORT_SYMBOL_GPL(ip6_pol_route);
2281
ip6_pol_route_input(struct net * net,struct fib6_table * table,struct flowi6 * fl6,const struct sk_buff * skb,int flags)2282 INDIRECT_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
ip6_route_input_lookup(struct net * net,struct net_device * dev,struct flowi6 * fl6,const struct sk_buff * skb,int flags)2291 struct 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 }
2302 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
2303
ip6_multipath_l3_keys(const struct sk_buff * skb,struct flow_keys * keys,struct flow_keys * flkeys)2304 static 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;
2335 out:
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
rt6_multipath_custom_hash_outer(const struct net * net,const struct sk_buff * skb,bool * p_has_inner)2349 static 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
rt6_multipath_custom_hash_inner(const struct net * net,const struct sk_buff * skb,bool has_inner)2380 static 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
rt6_multipath_custom_hash_skb(const struct net * net,const struct sk_buff * skb)2429 static 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
rt6_multipath_custom_hash_fl6(const struct net * net,const struct flowi6 * fl6)2441 static 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 */
rt6_multipath_hash(const struct net * net,const struct flowi6 * fl6,const struct sk_buff * skb,struct flow_keys * flkeys)2469 u32 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 */
ip6_route_input(struct sk_buff * skb)2570 void 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
ip6_pol_route_output(struct net * net,struct fib6_table * table,struct flowi6 * fl6,const struct sk_buff * skb,int flags)2600 INDIRECT_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
ip6_route_output_flags_noref(struct net * net,const struct sock * sk,struct flowi6 * fl6,int flags)2609 static 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
ip6_route_output_flags(struct net * net,const struct sock * sk,struct flowi6 * fl6,int flags)2642 struct 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 }
2662 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2663
ip6_blackhole_route(struct net * net,struct dst_entry * dst_orig)2664 struct 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
fib6_check(struct fib6_info * f6i,u32 cookie)2701 static 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
rt6_check(struct rt6_info * rt,struct fib6_info * from,u32 cookie)2714 static 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
rt6_dst_from_check(struct rt6_info * rt,struct fib6_info * from,u32 cookie)2730 static 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
ip6_dst_check(struct dst_entry * dst,u32 cookie)2742 INDIRECT_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 }
2773 EXPORT_INDIRECT_CALLABLE(ip6_dst_check);
2774
ip6_negative_advice(struct sock * sk,struct dst_entry * dst)2775 static 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
ip6_link_failure(struct sk_buff * skb)2795 static 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
rt6_update_expires(struct rt6_info * rt0,int timeout)2821 static 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
rt6_do_update_pmtu(struct rt6_info * rt,u32 mtu)2837 static 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
rt6_cache_allowed_for_pmtu(const struct rt6_info * rt)2846 static 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
__ip6_rt_update_pmtu(struct dst_entry * dst,const struct sock * sk,const struct ipv6hdr * iph,u32 mtu,bool confirm_neigh)2852 static 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 }
2926 out_unlock:
2927 rcu_read_unlock();
2928 }
2929 }
2930
ip6_rt_update_pmtu(struct dst_entry * dst,struct sock * sk,struct sk_buff * skb,u32 mtu,bool confirm_neigh)2931 static 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
ip6_update_pmtu(struct sk_buff * skb,struct net * net,__be32 mtu,int oif,u32 mark,kuid_t uid)2939 void 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 }
2958 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2959
ip6_sk_update_pmtu(struct sk_buff * skb,struct sock * sk,__be32 mtu)2960 void 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 }
2981 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2982
ip6_sk_dst_store_flow(struct sock * sk,struct dst_entry * dst,const struct flowi6 * fl6)2983 void 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
ip6_redirect_nh_match(const struct fib6_result * res,struct flowi6 * fl6,const struct in6_addr * gw,struct rt6_info ** ret)3000 static 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
3030 struct 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
fib6_nh_redirect_match(struct fib6_nh * nh,void * _arg)3037 static 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 */
3046 struct ip6rd_flowi {
3047 struct flowi6 fl6;
3048 struct in6_addr gateway;
3049 };
3050
__ip6_route_redirect(struct net * net,struct fib6_table * table,struct flowi6 * fl6,const struct sk_buff * skb,int flags)3051 INDIRECT_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);
3081 restart:
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;
3119 out:
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
ip6_route_redirect(struct net * net,const struct flowi6 * fl6,const struct sk_buff * skb,const struct in6_addr * gateway)3134 static 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
ip6_redirect(struct sk_buff * skb,struct net * net,int oif,u32 mark,kuid_t uid)3149 void 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 }
3168 EXPORT_SYMBOL_GPL(ip6_redirect);
3169
ip6_redirect_no_header(struct sk_buff * skb,struct net * net,int oif)3170 void 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
ip6_sk_redirect(struct sk_buff * skb,struct sock * sk)3188 void 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 }
3193 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3194
ip6_default_advmss(const struct dst_entry * dst)3195 static 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 = dev_net(dev);
3200
3201 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3202
3203 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3204 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3205
3206 /*
3207 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3208 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3209 * IPV6_MAXPLEN is also valid and means: "any MSS,
3210 * rely only on pmtu discovery"
3211 */
3212 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3213 mtu = IPV6_MAXPLEN;
3214 return mtu;
3215 }
3216
ip6_mtu(const struct dst_entry * dst)3217 INDIRECT_CALLABLE_SCOPE unsigned int ip6_mtu(const struct dst_entry *dst)
3218 {
3219 return ip6_dst_mtu_maybe_forward(dst, false);
3220 }
3221 EXPORT_INDIRECT_CALLABLE(ip6_mtu);
3222
3223 /* MTU selection:
3224 * 1. mtu on route is locked - use it
3225 * 2. mtu from nexthop exception
3226 * 3. mtu from egress device
3227 *
3228 * based on ip6_dst_mtu_forward and exception logic of
3229 * rt6_find_cached_rt; called with rcu_read_lock
3230 */
ip6_mtu_from_fib6(const struct fib6_result * res,const struct in6_addr * daddr,const struct in6_addr * saddr)3231 u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3232 const struct in6_addr *daddr,
3233 const struct in6_addr *saddr)
3234 {
3235 const struct fib6_nh *nh = res->nh;
3236 struct fib6_info *f6i = res->f6i;
3237 struct inet6_dev *idev;
3238 struct rt6_info *rt;
3239 u32 mtu = 0;
3240
3241 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3242 mtu = f6i->fib6_pmtu;
3243 if (mtu)
3244 goto out;
3245 }
3246
3247 rt = rt6_find_cached_rt(res, daddr, saddr);
3248 if (unlikely(rt)) {
3249 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3250 } else {
3251 struct net_device *dev = nh->fib_nh_dev;
3252
3253 mtu = IPV6_MIN_MTU;
3254 idev = __in6_dev_get(dev);
3255 if (idev)
3256 mtu = max_t(u32, mtu, READ_ONCE(idev->cnf.mtu6));
3257 }
3258
3259 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3260 out:
3261 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3262 }
3263
icmp6_dst_alloc(struct net_device * dev,struct flowi6 * fl6)3264 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3265 struct flowi6 *fl6)
3266 {
3267 struct dst_entry *dst;
3268 struct rt6_info *rt;
3269 struct inet6_dev *idev = in6_dev_get(dev);
3270 struct net *net = dev_net(dev);
3271
3272 if (unlikely(!idev))
3273 return ERR_PTR(-ENODEV);
3274
3275 rt = ip6_dst_alloc(net, dev, 0);
3276 if (unlikely(!rt)) {
3277 in6_dev_put(idev);
3278 dst = ERR_PTR(-ENOMEM);
3279 goto out;
3280 }
3281
3282 rt->dst.input = ip6_input;
3283 rt->dst.output = ip6_output;
3284 rt->rt6i_gateway = fl6->daddr;
3285 rt->rt6i_dst.addr = fl6->daddr;
3286 rt->rt6i_dst.plen = 128;
3287 rt->rt6i_idev = idev;
3288 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3289
3290 /* Add this dst into uncached_list so that rt6_disable_ip() can
3291 * do proper release of the net_device
3292 */
3293 rt6_uncached_list_add(rt);
3294
3295 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3296
3297 out:
3298 return dst;
3299 }
3300
ip6_dst_gc(struct dst_ops * ops)3301 static void ip6_dst_gc(struct dst_ops *ops)
3302 {
3303 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3304 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3305 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3306 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3307 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3308 unsigned int val;
3309 int entries;
3310
3311 if (time_after(rt_last_gc + rt_min_interval, jiffies))
3312 goto out;
3313
3314 fib6_run_gc(atomic_inc_return(&net->ipv6.ip6_rt_gc_expire), net, true);
3315 entries = dst_entries_get_slow(ops);
3316 if (entries < ops->gc_thresh)
3317 atomic_set(&net->ipv6.ip6_rt_gc_expire, rt_gc_timeout >> 1);
3318 out:
3319 val = atomic_read(&net->ipv6.ip6_rt_gc_expire);
3320 atomic_set(&net->ipv6.ip6_rt_gc_expire, val - (val >> rt_elasticity));
3321 }
3322
ip6_nh_lookup_table(struct net * net,struct fib6_config * cfg,const struct in6_addr * gw_addr,u32 tbid,int flags,struct fib6_result * res)3323 static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3324 const struct in6_addr *gw_addr, u32 tbid,
3325 int flags, struct fib6_result *res)
3326 {
3327 struct flowi6 fl6 = {
3328 .flowi6_oif = cfg->fc_ifindex,
3329 .daddr = *gw_addr,
3330 .saddr = cfg->fc_prefsrc,
3331 };
3332 struct fib6_table *table;
3333 int err;
3334
3335 table = fib6_get_table(net, tbid);
3336 if (!table)
3337 return -EINVAL;
3338
3339 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3340 flags |= RT6_LOOKUP_F_HAS_SADDR;
3341
3342 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3343
3344 err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3345 if (!err && res->f6i != net->ipv6.fib6_null_entry)
3346 fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3347 cfg->fc_ifindex != 0, NULL, flags);
3348
3349 return err;
3350 }
3351
ip6_route_check_nh_onlink(struct net * net,struct fib6_config * cfg,const struct net_device * dev,struct netlink_ext_ack * extack)3352 static int ip6_route_check_nh_onlink(struct net *net,
3353 struct fib6_config *cfg,
3354 const struct net_device *dev,
3355 struct netlink_ext_ack *extack)
3356 {
3357 u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3358 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3359 struct fib6_result res = {};
3360 int err;
3361
3362 err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3363 if (!err && !(res.fib6_flags & RTF_REJECT) &&
3364 /* ignore match if it is the default route */
3365 !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3366 (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3367 NL_SET_ERR_MSG(extack,
3368 "Nexthop has invalid gateway or device mismatch");
3369 err = -EINVAL;
3370 }
3371
3372 return err;
3373 }
3374
ip6_route_check_nh(struct net * net,struct fib6_config * cfg,struct net_device ** _dev,netdevice_tracker * dev_tracker,struct inet6_dev ** idev)3375 static int ip6_route_check_nh(struct net *net,
3376 struct fib6_config *cfg,
3377 struct net_device **_dev,
3378 netdevice_tracker *dev_tracker,
3379 struct inet6_dev **idev)
3380 {
3381 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3382 struct net_device *dev = _dev ? *_dev : NULL;
3383 int flags = RT6_LOOKUP_F_IFACE;
3384 struct fib6_result res = {};
3385 int err = -EHOSTUNREACH;
3386
3387 if (cfg->fc_table) {
3388 err = ip6_nh_lookup_table(net, cfg, gw_addr,
3389 cfg->fc_table, flags, &res);
3390 /* gw_addr can not require a gateway or resolve to a reject
3391 * route. If a device is given, it must match the result.
3392 */
3393 if (err || res.fib6_flags & RTF_REJECT ||
3394 res.nh->fib_nh_gw_family ||
3395 (dev && dev != res.nh->fib_nh_dev))
3396 err = -EHOSTUNREACH;
3397 }
3398
3399 if (err < 0) {
3400 struct flowi6 fl6 = {
3401 .flowi6_oif = cfg->fc_ifindex,
3402 .daddr = *gw_addr,
3403 };
3404
3405 err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3406 if (err || res.fib6_flags & RTF_REJECT ||
3407 res.nh->fib_nh_gw_family)
3408 err = -EHOSTUNREACH;
3409
3410 if (err)
3411 return err;
3412
3413 fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3414 cfg->fc_ifindex != 0, NULL, flags);
3415 }
3416
3417 err = 0;
3418 if (dev) {
3419 if (dev != res.nh->fib_nh_dev)
3420 err = -EHOSTUNREACH;
3421 } else {
3422 *_dev = dev = res.nh->fib_nh_dev;
3423 netdev_hold(dev, dev_tracker, GFP_ATOMIC);
3424 *idev = in6_dev_get(dev);
3425 }
3426
3427 return err;
3428 }
3429
ip6_validate_gw(struct net * net,struct fib6_config * cfg,struct net_device ** _dev,netdevice_tracker * dev_tracker,struct inet6_dev ** idev,struct netlink_ext_ack * extack)3430 static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3431 struct net_device **_dev,
3432 netdevice_tracker *dev_tracker,
3433 struct inet6_dev **idev,
3434 struct netlink_ext_ack *extack)
3435 {
3436 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3437 int gwa_type = ipv6_addr_type(gw_addr);
3438 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3439 const struct net_device *dev = *_dev;
3440 bool need_addr_check = !dev;
3441 int err = -EINVAL;
3442
3443 /* if gw_addr is local we will fail to detect this in case
3444 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3445 * will return already-added prefix route via interface that
3446 * prefix route was assigned to, which might be non-loopback.
3447 */
3448 if (dev &&
3449 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3450 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3451 goto out;
3452 }
3453
3454 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3455 /* IPv6 strictly inhibits using not link-local
3456 * addresses as nexthop address.
3457 * Otherwise, router will not able to send redirects.
3458 * It is very good, but in some (rare!) circumstances
3459 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3460 * some exceptions. --ANK
3461 * We allow IPv4-mapped nexthops to support RFC4798-type
3462 * addressing
3463 */
3464 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3465 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3466 goto out;
3467 }
3468
3469 rcu_read_lock();
3470
3471 if (cfg->fc_flags & RTNH_F_ONLINK)
3472 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3473 else
3474 err = ip6_route_check_nh(net, cfg, _dev, dev_tracker,
3475 idev);
3476
3477 rcu_read_unlock();
3478
3479 if (err)
3480 goto out;
3481 }
3482
3483 /* reload in case device was changed */
3484 dev = *_dev;
3485
3486 err = -EINVAL;
3487 if (!dev) {
3488 NL_SET_ERR_MSG(extack, "Egress device not specified");
3489 goto out;
3490 } else if (dev->flags & IFF_LOOPBACK) {
3491 NL_SET_ERR_MSG(extack,
3492 "Egress device can not be loopback device for this route");
3493 goto out;
3494 }
3495
3496 /* if we did not check gw_addr above, do so now that the
3497 * egress device has been resolved.
3498 */
3499 if (need_addr_check &&
3500 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3501 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3502 goto out;
3503 }
3504
3505 err = 0;
3506 out:
3507 return err;
3508 }
3509
fib6_is_reject(u32 flags,struct net_device * dev,int addr_type)3510 static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3511 {
3512 if ((flags & RTF_REJECT) ||
3513 (dev && (dev->flags & IFF_LOOPBACK) &&
3514 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3515 !(flags & (RTF_ANYCAST | RTF_LOCAL))))
3516 return true;
3517
3518 return false;
3519 }
3520
fib6_nh_init(struct net * net,struct fib6_nh * fib6_nh,struct fib6_config * cfg,gfp_t gfp_flags,struct netlink_ext_ack * extack)3521 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3522 struct fib6_config *cfg, gfp_t gfp_flags,
3523 struct netlink_ext_ack *extack)
3524 {
3525 netdevice_tracker *dev_tracker = &fib6_nh->fib_nh_dev_tracker;
3526 struct net_device *dev = NULL;
3527 struct inet6_dev *idev = NULL;
3528 int addr_type;
3529 int err;
3530
3531 fib6_nh->fib_nh_family = AF_INET6;
3532 #ifdef CONFIG_IPV6_ROUTER_PREF
3533 fib6_nh->last_probe = jiffies;
3534 #endif
3535 if (cfg->fc_is_fdb) {
3536 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3537 fib6_nh->fib_nh_gw_family = AF_INET6;
3538 return 0;
3539 }
3540
3541 err = -ENODEV;
3542 if (cfg->fc_ifindex) {
3543 dev = netdev_get_by_index(net, cfg->fc_ifindex,
3544 dev_tracker, gfp_flags);
3545 if (!dev)
3546 goto out;
3547 idev = in6_dev_get(dev);
3548 if (!idev)
3549 goto out;
3550 }
3551
3552 if (cfg->fc_flags & RTNH_F_ONLINK) {
3553 if (!dev) {
3554 NL_SET_ERR_MSG(extack,
3555 "Nexthop device required for onlink");
3556 goto out;
3557 }
3558
3559 if (!(dev->flags & IFF_UP)) {
3560 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3561 err = -ENETDOWN;
3562 goto out;
3563 }
3564
3565 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3566 }
3567
3568 fib6_nh->fib_nh_weight = 1;
3569
3570 /* We cannot add true routes via loopback here,
3571 * they would result in kernel looping; promote them to reject routes
3572 */
3573 addr_type = ipv6_addr_type(&cfg->fc_dst);
3574 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3575 /* hold loopback dev/idev if we haven't done so. */
3576 if (dev != net->loopback_dev) {
3577 if (dev) {
3578 netdev_put(dev, dev_tracker);
3579 in6_dev_put(idev);
3580 }
3581 dev = net->loopback_dev;
3582 netdev_hold(dev, dev_tracker, gfp_flags);
3583 idev = in6_dev_get(dev);
3584 if (!idev) {
3585 err = -ENODEV;
3586 goto out;
3587 }
3588 }
3589 goto pcpu_alloc;
3590 }
3591
3592 if (cfg->fc_flags & RTF_GATEWAY) {
3593 err = ip6_validate_gw(net, cfg, &dev, dev_tracker,
3594 &idev, extack);
3595 if (err)
3596 goto out;
3597
3598 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3599 fib6_nh->fib_nh_gw_family = AF_INET6;
3600 }
3601
3602 err = -ENODEV;
3603 if (!dev)
3604 goto out;
3605
3606 if (!idev || idev->cnf.disable_ipv6) {
3607 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3608 err = -EACCES;
3609 goto out;
3610 }
3611
3612 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3613 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3614 err = -ENETDOWN;
3615 goto out;
3616 }
3617
3618 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3619 !netif_carrier_ok(dev))
3620 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3621
3622 err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
3623 cfg->fc_encap_type, cfg, gfp_flags, extack);
3624 if (err)
3625 goto out;
3626
3627 pcpu_alloc:
3628 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3629 if (!fib6_nh->rt6i_pcpu) {
3630 err = -ENOMEM;
3631 goto out;
3632 }
3633
3634 fib6_nh->fib_nh_dev = dev;
3635 fib6_nh->fib_nh_oif = dev->ifindex;
3636 err = 0;
3637 out:
3638 if (idev)
3639 in6_dev_put(idev);
3640
3641 if (err) {
3642 lwtstate_put(fib6_nh->fib_nh_lws);
3643 fib6_nh->fib_nh_lws = NULL;
3644 netdev_put(dev, dev_tracker);
3645 }
3646
3647 return err;
3648 }
3649
fib6_nh_release(struct fib6_nh * fib6_nh)3650 void fib6_nh_release(struct fib6_nh *fib6_nh)
3651 {
3652 struct rt6_exception_bucket *bucket;
3653
3654 rcu_read_lock();
3655
3656 fib6_nh_flush_exceptions(fib6_nh, NULL);
3657 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3658 if (bucket) {
3659 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3660 kfree(bucket);
3661 }
3662
3663 rcu_read_unlock();
3664
3665 fib6_nh_release_dsts(fib6_nh);
3666 free_percpu(fib6_nh->rt6i_pcpu);
3667
3668 fib_nh_common_release(&fib6_nh->nh_common);
3669 }
3670
fib6_nh_release_dsts(struct fib6_nh * fib6_nh)3671 void fib6_nh_release_dsts(struct fib6_nh *fib6_nh)
3672 {
3673 int cpu;
3674
3675 if (!fib6_nh->rt6i_pcpu)
3676 return;
3677
3678 for_each_possible_cpu(cpu) {
3679 struct rt6_info *pcpu_rt, **ppcpu_rt;
3680
3681 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3682 pcpu_rt = xchg(ppcpu_rt, NULL);
3683 if (pcpu_rt) {
3684 dst_dev_put(&pcpu_rt->dst);
3685 dst_release(&pcpu_rt->dst);
3686 }
3687 }
3688 }
3689
ip6_route_info_create(struct fib6_config * cfg,gfp_t gfp_flags,struct netlink_ext_ack * extack)3690 static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3691 gfp_t gfp_flags,
3692 struct netlink_ext_ack *extack)
3693 {
3694 struct net *net = cfg->fc_nlinfo.nl_net;
3695 struct fib6_info *rt = NULL;
3696 struct nexthop *nh = NULL;
3697 struct fib6_table *table;
3698 struct fib6_nh *fib6_nh;
3699 int err = -EINVAL;
3700 int addr_type;
3701
3702 /* RTF_PCPU is an internal flag; can not be set by userspace */
3703 if (cfg->fc_flags & RTF_PCPU) {
3704 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3705 goto out;
3706 }
3707
3708 /* RTF_CACHE is an internal flag; can not be set by userspace */
3709 if (cfg->fc_flags & RTF_CACHE) {
3710 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3711 goto out;
3712 }
3713
3714 if (cfg->fc_type > RTN_MAX) {
3715 NL_SET_ERR_MSG(extack, "Invalid route type");
3716 goto out;
3717 }
3718
3719 if (cfg->fc_dst_len > 128) {
3720 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3721 goto out;
3722 }
3723 if (cfg->fc_src_len > 128) {
3724 NL_SET_ERR_MSG(extack, "Invalid source address length");
3725 goto out;
3726 }
3727 #ifndef CONFIG_IPV6_SUBTREES
3728 if (cfg->fc_src_len) {
3729 NL_SET_ERR_MSG(extack,
3730 "Specifying source address requires IPV6_SUBTREES to be enabled");
3731 goto out;
3732 }
3733 #endif
3734 if (cfg->fc_nh_id) {
3735 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3736 if (!nh) {
3737 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3738 goto out;
3739 }
3740 err = fib6_check_nexthop(nh, cfg, extack);
3741 if (err)
3742 goto out;
3743 }
3744
3745 err = -ENOBUFS;
3746 if (cfg->fc_nlinfo.nlh &&
3747 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3748 table = fib6_get_table(net, cfg->fc_table);
3749 if (!table) {
3750 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3751 table = fib6_new_table(net, cfg->fc_table);
3752 }
3753 } else {
3754 table = fib6_new_table(net, cfg->fc_table);
3755 }
3756
3757 if (!table)
3758 goto out;
3759
3760 err = -ENOMEM;
3761 rt = fib6_info_alloc(gfp_flags, !nh);
3762 if (!rt)
3763 goto out;
3764
3765 rt->fib6_metrics = ip_fib_metrics_init(cfg->fc_mx, cfg->fc_mx_len,
3766 extack);
3767 if (IS_ERR(rt->fib6_metrics)) {
3768 err = PTR_ERR(rt->fib6_metrics);
3769 /* Do not leave garbage there. */
3770 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3771 goto out_free;
3772 }
3773
3774 if (cfg->fc_flags & RTF_ADDRCONF)
3775 rt->dst_nocount = true;
3776
3777 if (cfg->fc_flags & RTF_EXPIRES)
3778 fib6_set_expires(rt, jiffies +
3779 clock_t_to_jiffies(cfg->fc_expires));
3780
3781 if (cfg->fc_protocol == RTPROT_UNSPEC)
3782 cfg->fc_protocol = RTPROT_BOOT;
3783 rt->fib6_protocol = cfg->fc_protocol;
3784
3785 rt->fib6_table = table;
3786 rt->fib6_metric = cfg->fc_metric;
3787 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3788 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3789
3790 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3791 rt->fib6_dst.plen = cfg->fc_dst_len;
3792
3793 #ifdef CONFIG_IPV6_SUBTREES
3794 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3795 rt->fib6_src.plen = cfg->fc_src_len;
3796 #endif
3797 if (nh) {
3798 if (rt->fib6_src.plen) {
3799 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3800 goto out_free;
3801 }
3802 if (!nexthop_get(nh)) {
3803 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3804 goto out_free;
3805 }
3806 rt->nh = nh;
3807 fib6_nh = nexthop_fib6_nh(rt->nh);
3808 } else {
3809 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3810 if (err)
3811 goto out;
3812
3813 fib6_nh = rt->fib6_nh;
3814
3815 /* We cannot add true routes via loopback here, they would
3816 * result in kernel looping; promote them to reject routes
3817 */
3818 addr_type = ipv6_addr_type(&cfg->fc_dst);
3819 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3820 addr_type))
3821 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3822 }
3823
3824 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3825 struct net_device *dev = fib6_nh->fib_nh_dev;
3826
3827 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3828 NL_SET_ERR_MSG(extack, "Invalid source address");
3829 err = -EINVAL;
3830 goto out;
3831 }
3832 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3833 rt->fib6_prefsrc.plen = 128;
3834 } else
3835 rt->fib6_prefsrc.plen = 0;
3836
3837 return rt;
3838 out:
3839 fib6_info_release(rt);
3840 return ERR_PTR(err);
3841 out_free:
3842 ip_fib_metrics_put(rt->fib6_metrics);
3843 kfree(rt);
3844 return ERR_PTR(err);
3845 }
3846
ip6_route_add(struct fib6_config * cfg,gfp_t gfp_flags,struct netlink_ext_ack * extack)3847 int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3848 struct netlink_ext_ack *extack)
3849 {
3850 struct fib6_info *rt;
3851 int err;
3852
3853 rt = ip6_route_info_create(cfg, gfp_flags, extack);
3854 if (IS_ERR(rt))
3855 return PTR_ERR(rt);
3856
3857 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3858 fib6_info_release(rt);
3859
3860 return err;
3861 }
3862
__ip6_del_rt(struct fib6_info * rt,struct nl_info * info)3863 static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3864 {
3865 struct net *net = info->nl_net;
3866 struct fib6_table *table;
3867 int err;
3868
3869 if (rt == net->ipv6.fib6_null_entry) {
3870 err = -ENOENT;
3871 goto out;
3872 }
3873
3874 table = rt->fib6_table;
3875 spin_lock_bh(&table->tb6_lock);
3876 err = fib6_del(rt, info);
3877 spin_unlock_bh(&table->tb6_lock);
3878
3879 out:
3880 fib6_info_release(rt);
3881 return err;
3882 }
3883
ip6_del_rt(struct net * net,struct fib6_info * rt,bool skip_notify)3884 int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
3885 {
3886 struct nl_info info = {
3887 .nl_net = net,
3888 .skip_notify = skip_notify
3889 };
3890
3891 return __ip6_del_rt(rt, &info);
3892 }
3893
__ip6_del_rt_siblings(struct fib6_info * rt,struct fib6_config * cfg)3894 static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3895 {
3896 struct nl_info *info = &cfg->fc_nlinfo;
3897 struct net *net = info->nl_net;
3898 struct sk_buff *skb = NULL;
3899 struct fib6_table *table;
3900 int err = -ENOENT;
3901
3902 if (rt == net->ipv6.fib6_null_entry)
3903 goto out_put;
3904 table = rt->fib6_table;
3905 spin_lock_bh(&table->tb6_lock);
3906
3907 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3908 struct fib6_info *sibling, *next_sibling;
3909 struct fib6_node *fn;
3910
3911 /* prefer to send a single notification with all hops */
3912 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3913 if (skb) {
3914 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3915
3916 if (rt6_fill_node(net, skb, rt, NULL,
3917 NULL, NULL, 0, RTM_DELROUTE,
3918 info->portid, seq, 0) < 0) {
3919 kfree_skb(skb);
3920 skb = NULL;
3921 } else
3922 info->skip_notify = 1;
3923 }
3924
3925 /* 'rt' points to the first sibling route. If it is not the
3926 * leaf, then we do not need to send a notification. Otherwise,
3927 * we need to check if the last sibling has a next route or not
3928 * and emit a replace or delete notification, respectively.
3929 */
3930 info->skip_notify_kernel = 1;
3931 fn = rcu_dereference_protected(rt->fib6_node,
3932 lockdep_is_held(&table->tb6_lock));
3933 if (rcu_access_pointer(fn->leaf) == rt) {
3934 struct fib6_info *last_sibling, *replace_rt;
3935
3936 last_sibling = list_last_entry(&rt->fib6_siblings,
3937 struct fib6_info,
3938 fib6_siblings);
3939 replace_rt = rcu_dereference_protected(
3940 last_sibling->fib6_next,
3941 lockdep_is_held(&table->tb6_lock));
3942 if (replace_rt)
3943 call_fib6_entry_notifiers_replace(net,
3944 replace_rt);
3945 else
3946 call_fib6_multipath_entry_notifiers(net,
3947 FIB_EVENT_ENTRY_DEL,
3948 rt, rt->fib6_nsiblings,
3949 NULL);
3950 }
3951 list_for_each_entry_safe(sibling, next_sibling,
3952 &rt->fib6_siblings,
3953 fib6_siblings) {
3954 err = fib6_del(sibling, info);
3955 if (err)
3956 goto out_unlock;
3957 }
3958 }
3959
3960 err = fib6_del(rt, info);
3961 out_unlock:
3962 spin_unlock_bh(&table->tb6_lock);
3963 out_put:
3964 fib6_info_release(rt);
3965
3966 if (skb) {
3967 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3968 info->nlh, gfp_any());
3969 }
3970 return err;
3971 }
3972
__ip6_del_cached_rt(struct rt6_info * rt,struct fib6_config * cfg)3973 static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3974 {
3975 int rc = -ESRCH;
3976
3977 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3978 goto out;
3979
3980 if (cfg->fc_flags & RTF_GATEWAY &&
3981 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3982 goto out;
3983
3984 rc = rt6_remove_exception_rt(rt);
3985 out:
3986 return rc;
3987 }
3988
ip6_del_cached_rt(struct fib6_config * cfg,struct fib6_info * rt,struct fib6_nh * nh)3989 static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3990 struct fib6_nh *nh)
3991 {
3992 struct fib6_result res = {
3993 .f6i = rt,
3994 .nh = nh,
3995 };
3996 struct rt6_info *rt_cache;
3997
3998 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3999 if (rt_cache)
4000 return __ip6_del_cached_rt(rt_cache, cfg);
4001
4002 return 0;
4003 }
4004
4005 struct fib6_nh_del_cached_rt_arg {
4006 struct fib6_config *cfg;
4007 struct fib6_info *f6i;
4008 };
4009
fib6_nh_del_cached_rt(struct fib6_nh * nh,void * _arg)4010 static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
4011 {
4012 struct fib6_nh_del_cached_rt_arg *arg = _arg;
4013 int rc;
4014
4015 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
4016 return rc != -ESRCH ? rc : 0;
4017 }
4018
ip6_del_cached_rt_nh(struct fib6_config * cfg,struct fib6_info * f6i)4019 static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
4020 {
4021 struct fib6_nh_del_cached_rt_arg arg = {
4022 .cfg = cfg,
4023 .f6i = f6i
4024 };
4025
4026 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
4027 }
4028
ip6_route_del(struct fib6_config * cfg,struct netlink_ext_ack * extack)4029 static int ip6_route_del(struct fib6_config *cfg,
4030 struct netlink_ext_ack *extack)
4031 {
4032 struct fib6_table *table;
4033 struct fib6_info *rt;
4034 struct fib6_node *fn;
4035 int err = -ESRCH;
4036
4037 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
4038 if (!table) {
4039 NL_SET_ERR_MSG(extack, "FIB table does not exist");
4040 return err;
4041 }
4042
4043 rcu_read_lock();
4044
4045 fn = fib6_locate(&table->tb6_root,
4046 &cfg->fc_dst, cfg->fc_dst_len,
4047 &cfg->fc_src, cfg->fc_src_len,
4048 !(cfg->fc_flags & RTF_CACHE));
4049
4050 if (fn) {
4051 for_each_fib6_node_rt_rcu(fn) {
4052 struct fib6_nh *nh;
4053
4054 if (rt->nh && cfg->fc_nh_id &&
4055 rt->nh->id != cfg->fc_nh_id)
4056 continue;
4057
4058 if (cfg->fc_flags & RTF_CACHE) {
4059 int rc = 0;
4060
4061 if (rt->nh) {
4062 rc = ip6_del_cached_rt_nh(cfg, rt);
4063 } else if (cfg->fc_nh_id) {
4064 continue;
4065 } else {
4066 nh = rt->fib6_nh;
4067 rc = ip6_del_cached_rt(cfg, rt, nh);
4068 }
4069 if (rc != -ESRCH) {
4070 rcu_read_unlock();
4071 return rc;
4072 }
4073 continue;
4074 }
4075
4076 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
4077 continue;
4078 if (cfg->fc_protocol &&
4079 cfg->fc_protocol != rt->fib6_protocol)
4080 continue;
4081
4082 if (rt->nh) {
4083 if (!fib6_info_hold_safe(rt))
4084 continue;
4085 rcu_read_unlock();
4086
4087 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4088 }
4089 if (cfg->fc_nh_id)
4090 continue;
4091
4092 nh = rt->fib6_nh;
4093 if (cfg->fc_ifindex &&
4094 (!nh->fib_nh_dev ||
4095 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
4096 continue;
4097 if (cfg->fc_flags & RTF_GATEWAY &&
4098 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
4099 continue;
4100 if (!fib6_info_hold_safe(rt))
4101 continue;
4102 rcu_read_unlock();
4103
4104 /* if gateway was specified only delete the one hop */
4105 if (cfg->fc_flags & RTF_GATEWAY)
4106 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4107
4108 return __ip6_del_rt_siblings(rt, cfg);
4109 }
4110 }
4111 rcu_read_unlock();
4112
4113 return err;
4114 }
4115
rt6_do_redirect(struct dst_entry * dst,struct sock * sk,struct sk_buff * skb)4116 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
4117 {
4118 struct netevent_redirect netevent;
4119 struct rt6_info *rt, *nrt = NULL;
4120 struct fib6_result res = {};
4121 struct ndisc_options ndopts;
4122 struct inet6_dev *in6_dev;
4123 struct neighbour *neigh;
4124 struct rd_msg *msg;
4125 int optlen, on_link;
4126 u8 *lladdr;
4127
4128 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
4129 optlen -= sizeof(*msg);
4130
4131 if (optlen < 0) {
4132 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
4133 return;
4134 }
4135
4136 msg = (struct rd_msg *)icmp6_hdr(skb);
4137
4138 if (ipv6_addr_is_multicast(&msg->dest)) {
4139 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
4140 return;
4141 }
4142
4143 on_link = 0;
4144 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
4145 on_link = 1;
4146 } else if (ipv6_addr_type(&msg->target) !=
4147 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
4148 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
4149 return;
4150 }
4151
4152 in6_dev = __in6_dev_get(skb->dev);
4153 if (!in6_dev)
4154 return;
4155 if (READ_ONCE(in6_dev->cnf.forwarding) ||
4156 !READ_ONCE(in6_dev->cnf.accept_redirects))
4157 return;
4158
4159 /* RFC2461 8.1:
4160 * The IP source address of the Redirect MUST be the same as the current
4161 * first-hop router for the specified ICMP Destination Address.
4162 */
4163
4164 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
4165 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4166 return;
4167 }
4168
4169 lladdr = NULL;
4170 if (ndopts.nd_opts_tgt_lladdr) {
4171 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4172 skb->dev);
4173 if (!lladdr) {
4174 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4175 return;
4176 }
4177 }
4178
4179 rt = dst_rt6_info(dst);
4180 if (rt->rt6i_flags & RTF_REJECT) {
4181 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
4182 return;
4183 }
4184
4185 /* Redirect received -> path was valid.
4186 * Look, redirects are sent only in response to data packets,
4187 * so that this nexthop apparently is reachable. --ANK
4188 */
4189 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
4190
4191 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
4192 if (!neigh)
4193 return;
4194
4195 /*
4196 * We have finally decided to accept it.
4197 */
4198
4199 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
4200 NEIGH_UPDATE_F_WEAK_OVERRIDE|
4201 NEIGH_UPDATE_F_OVERRIDE|
4202 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
4203 NEIGH_UPDATE_F_ISROUTER)),
4204 NDISC_REDIRECT, &ndopts);
4205
4206 rcu_read_lock();
4207 res.f6i = rcu_dereference(rt->from);
4208 if (!res.f6i)
4209 goto out;
4210
4211 if (res.f6i->nh) {
4212 struct fib6_nh_match_arg arg = {
4213 .dev = dst->dev,
4214 .gw = &rt->rt6i_gateway,
4215 };
4216
4217 nexthop_for_each_fib6_nh(res.f6i->nh,
4218 fib6_nh_find_match, &arg);
4219
4220 /* fib6_info uses a nexthop that does not have fib6_nh
4221 * using the dst->dev. Should be impossible
4222 */
4223 if (!arg.match)
4224 goto out;
4225 res.nh = arg.match;
4226 } else {
4227 res.nh = res.f6i->fib6_nh;
4228 }
4229
4230 res.fib6_flags = res.f6i->fib6_flags;
4231 res.fib6_type = res.f6i->fib6_type;
4232 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4233 if (!nrt)
4234 goto out;
4235
4236 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4237 if (on_link)
4238 nrt->rt6i_flags &= ~RTF_GATEWAY;
4239
4240 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4241
4242 /* rt6_insert_exception() will take care of duplicated exceptions */
4243 if (rt6_insert_exception(nrt, &res)) {
4244 dst_release_immediate(&nrt->dst);
4245 goto out;
4246 }
4247
4248 netevent.old = &rt->dst;
4249 netevent.new = &nrt->dst;
4250 netevent.daddr = &msg->dest;
4251 netevent.neigh = neigh;
4252 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4253
4254 out:
4255 rcu_read_unlock();
4256 neigh_release(neigh);
4257 }
4258
4259 #ifdef CONFIG_IPV6_ROUTE_INFO
rt6_get_route_info(struct net * net,const struct in6_addr * prefix,int prefixlen,const struct in6_addr * gwaddr,struct net_device * dev)4260 static struct fib6_info *rt6_get_route_info(struct net *net,
4261 const struct in6_addr *prefix, int prefixlen,
4262 const struct in6_addr *gwaddr,
4263 struct net_device *dev)
4264 {
4265 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4266 int ifindex = dev->ifindex;
4267 struct fib6_node *fn;
4268 struct fib6_info *rt = NULL;
4269 struct fib6_table *table;
4270
4271 table = fib6_get_table(net, tb_id);
4272 if (!table)
4273 return NULL;
4274
4275 rcu_read_lock();
4276 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4277 if (!fn)
4278 goto out;
4279
4280 for_each_fib6_node_rt_rcu(fn) {
4281 /* these routes do not use nexthops */
4282 if (rt->nh)
4283 continue;
4284 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4285 continue;
4286 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4287 !rt->fib6_nh->fib_nh_gw_family)
4288 continue;
4289 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4290 continue;
4291 if (!fib6_info_hold_safe(rt))
4292 continue;
4293 break;
4294 }
4295 out:
4296 rcu_read_unlock();
4297 return rt;
4298 }
4299
rt6_add_route_info(struct net * net,const struct in6_addr * prefix,int prefixlen,const struct in6_addr * gwaddr,struct net_device * dev,unsigned int pref)4300 static struct fib6_info *rt6_add_route_info(struct net *net,
4301 const struct in6_addr *prefix, int prefixlen,
4302 const struct in6_addr *gwaddr,
4303 struct net_device *dev,
4304 unsigned int pref)
4305 {
4306 struct fib6_config cfg = {
4307 .fc_metric = IP6_RT_PRIO_USER,
4308 .fc_ifindex = dev->ifindex,
4309 .fc_dst_len = prefixlen,
4310 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4311 RTF_UP | RTF_PREF(pref),
4312 .fc_protocol = RTPROT_RA,
4313 .fc_type = RTN_UNICAST,
4314 .fc_nlinfo.portid = 0,
4315 .fc_nlinfo.nlh = NULL,
4316 .fc_nlinfo.nl_net = net,
4317 };
4318
4319 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4320 cfg.fc_dst = *prefix;
4321 cfg.fc_gateway = *gwaddr;
4322
4323 /* We should treat it as a default route if prefix length is 0. */
4324 if (!prefixlen)
4325 cfg.fc_flags |= RTF_DEFAULT;
4326
4327 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4328
4329 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4330 }
4331 #endif
4332
rt6_get_dflt_router(struct net * net,const struct in6_addr * addr,struct net_device * dev)4333 struct fib6_info *rt6_get_dflt_router(struct net *net,
4334 const struct in6_addr *addr,
4335 struct net_device *dev)
4336 {
4337 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4338 struct fib6_info *rt;
4339 struct fib6_table *table;
4340
4341 table = fib6_get_table(net, tb_id);
4342 if (!table)
4343 return NULL;
4344
4345 rcu_read_lock();
4346 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4347 struct fib6_nh *nh;
4348
4349 /* RA routes do not use nexthops */
4350 if (rt->nh)
4351 continue;
4352
4353 nh = rt->fib6_nh;
4354 if (dev == nh->fib_nh_dev &&
4355 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4356 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4357 break;
4358 }
4359 if (rt && !fib6_info_hold_safe(rt))
4360 rt = NULL;
4361 rcu_read_unlock();
4362 return rt;
4363 }
4364
rt6_add_dflt_router(struct net * net,const struct in6_addr * gwaddr,struct net_device * dev,unsigned int pref,u32 defrtr_usr_metric,int lifetime)4365 struct fib6_info *rt6_add_dflt_router(struct net *net,
4366 const struct in6_addr *gwaddr,
4367 struct net_device *dev,
4368 unsigned int pref,
4369 u32 defrtr_usr_metric,
4370 int lifetime)
4371 {
4372 struct fib6_config cfg = {
4373 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4374 .fc_metric = defrtr_usr_metric,
4375 .fc_ifindex = dev->ifindex,
4376 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4377 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4378 .fc_protocol = RTPROT_RA,
4379 .fc_type = RTN_UNICAST,
4380 .fc_nlinfo.portid = 0,
4381 .fc_nlinfo.nlh = NULL,
4382 .fc_nlinfo.nl_net = net,
4383 .fc_expires = jiffies_to_clock_t(lifetime * HZ),
4384 };
4385
4386 cfg.fc_gateway = *gwaddr;
4387
4388 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4389 struct fib6_table *table;
4390
4391 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4392 if (table)
4393 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4394 }
4395
4396 return rt6_get_dflt_router(net, gwaddr, dev);
4397 }
4398
__rt6_purge_dflt_routers(struct net * net,struct fib6_table * table)4399 static void __rt6_purge_dflt_routers(struct net *net,
4400 struct fib6_table *table)
4401 {
4402 struct fib6_info *rt;
4403
4404 restart:
4405 rcu_read_lock();
4406 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4407 struct net_device *dev = fib6_info_nh_dev(rt);
4408 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4409
4410 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4411 (!idev || idev->cnf.accept_ra != 2) &&
4412 fib6_info_hold_safe(rt)) {
4413 rcu_read_unlock();
4414 ip6_del_rt(net, rt, false);
4415 goto restart;
4416 }
4417 }
4418 rcu_read_unlock();
4419
4420 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4421 }
4422
rt6_purge_dflt_routers(struct net * net)4423 void rt6_purge_dflt_routers(struct net *net)
4424 {
4425 struct fib6_table *table;
4426 struct hlist_head *head;
4427 unsigned int h;
4428
4429 rcu_read_lock();
4430
4431 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4432 head = &net->ipv6.fib_table_hash[h];
4433 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4434 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4435 __rt6_purge_dflt_routers(net, table);
4436 }
4437 }
4438
4439 rcu_read_unlock();
4440 }
4441
rtmsg_to_fib6_config(struct net * net,struct in6_rtmsg * rtmsg,struct fib6_config * cfg)4442 static void rtmsg_to_fib6_config(struct net *net,
4443 struct in6_rtmsg *rtmsg,
4444 struct fib6_config *cfg)
4445 {
4446 *cfg = (struct fib6_config){
4447 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4448 : RT6_TABLE_MAIN,
4449 .fc_ifindex = rtmsg->rtmsg_ifindex,
4450 .fc_metric = rtmsg->rtmsg_metric,
4451 .fc_expires = rtmsg->rtmsg_info,
4452 .fc_dst_len = rtmsg->rtmsg_dst_len,
4453 .fc_src_len = rtmsg->rtmsg_src_len,
4454 .fc_flags = rtmsg->rtmsg_flags,
4455 .fc_type = rtmsg->rtmsg_type,
4456
4457 .fc_nlinfo.nl_net = net,
4458
4459 .fc_dst = rtmsg->rtmsg_dst,
4460 .fc_src = rtmsg->rtmsg_src,
4461 .fc_gateway = rtmsg->rtmsg_gateway,
4462 };
4463 }
4464
ipv6_route_ioctl(struct net * net,unsigned int cmd,struct in6_rtmsg * rtmsg)4465 int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
4466 {
4467 struct fib6_config cfg;
4468 int err;
4469
4470 if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4471 return -EINVAL;
4472 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4473 return -EPERM;
4474
4475 rtmsg_to_fib6_config(net, rtmsg, &cfg);
4476
4477 rtnl_lock();
4478 switch (cmd) {
4479 case SIOCADDRT:
4480 /* Only do the default setting of fc_metric in route adding */
4481 if (cfg.fc_metric == 0)
4482 cfg.fc_metric = IP6_RT_PRIO_USER;
4483 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4484 break;
4485 case SIOCDELRT:
4486 err = ip6_route_del(&cfg, NULL);
4487 break;
4488 }
4489 rtnl_unlock();
4490 return err;
4491 }
4492
4493 /*
4494 * Drop the packet on the floor
4495 */
4496
ip6_pkt_drop(struct sk_buff * skb,u8 code,int ipstats_mib_noroutes)4497 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4498 {
4499 struct dst_entry *dst = skb_dst(skb);
4500 struct net *net = dev_net(dst->dev);
4501 struct inet6_dev *idev;
4502 SKB_DR(reason);
4503 int type;
4504
4505 if (netif_is_l3_master(skb->dev) ||
4506 dst->dev == net->loopback_dev)
4507 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4508 else
4509 idev = ip6_dst_idev(dst);
4510
4511 switch (ipstats_mib_noroutes) {
4512 case IPSTATS_MIB_INNOROUTES:
4513 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4514 if (type == IPV6_ADDR_ANY) {
4515 SKB_DR_SET(reason, IP_INADDRERRORS);
4516 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4517 break;
4518 }
4519 SKB_DR_SET(reason, IP_INNOROUTES);
4520 fallthrough;
4521 case IPSTATS_MIB_OUTNOROUTES:
4522 SKB_DR_OR(reason, IP_OUTNOROUTES);
4523 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4524 break;
4525 }
4526
4527 /* Start over by dropping the dst for l3mdev case */
4528 if (netif_is_l3_master(skb->dev))
4529 skb_dst_drop(skb);
4530
4531 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4532 kfree_skb_reason(skb, reason);
4533 return 0;
4534 }
4535
ip6_pkt_discard(struct sk_buff * skb)4536 static int ip6_pkt_discard(struct sk_buff *skb)
4537 {
4538 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4539 }
4540
ip6_pkt_discard_out(struct net * net,struct sock * sk,struct sk_buff * skb)4541 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4542 {
4543 skb->dev = skb_dst(skb)->dev;
4544 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4545 }
4546
ip6_pkt_prohibit(struct sk_buff * skb)4547 static int ip6_pkt_prohibit(struct sk_buff *skb)
4548 {
4549 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4550 }
4551
ip6_pkt_prohibit_out(struct net * net,struct sock * sk,struct sk_buff * skb)4552 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4553 {
4554 skb->dev = skb_dst(skb)->dev;
4555 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4556 }
4557
4558 /*
4559 * Allocate a dst for local (unicast / anycast) address.
4560 */
4561
addrconf_f6i_alloc(struct net * net,struct inet6_dev * idev,const struct in6_addr * addr,bool anycast,gfp_t gfp_flags,struct netlink_ext_ack * extack)4562 struct fib6_info *addrconf_f6i_alloc(struct net *net,
4563 struct inet6_dev *idev,
4564 const struct in6_addr *addr,
4565 bool anycast, gfp_t gfp_flags,
4566 struct netlink_ext_ack *extack)
4567 {
4568 struct fib6_config cfg = {
4569 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4570 .fc_ifindex = idev->dev->ifindex,
4571 .fc_flags = RTF_UP | RTF_NONEXTHOP,
4572 .fc_dst = *addr,
4573 .fc_dst_len = 128,
4574 .fc_protocol = RTPROT_KERNEL,
4575 .fc_nlinfo.nl_net = net,
4576 .fc_ignore_dev_down = true,
4577 };
4578 struct fib6_info *f6i;
4579
4580 if (anycast) {
4581 cfg.fc_type = RTN_ANYCAST;
4582 cfg.fc_flags |= RTF_ANYCAST;
4583 } else {
4584 cfg.fc_type = RTN_LOCAL;
4585 cfg.fc_flags |= RTF_LOCAL;
4586 }
4587
4588 f6i = ip6_route_info_create(&cfg, gfp_flags, extack);
4589 if (!IS_ERR(f6i)) {
4590 f6i->dst_nocount = true;
4591
4592 if (!anycast &&
4593 (READ_ONCE(net->ipv6.devconf_all->disable_policy) ||
4594 READ_ONCE(idev->cnf.disable_policy)))
4595 f6i->dst_nopolicy = true;
4596 }
4597
4598 return f6i;
4599 }
4600
4601 /* remove deleted ip from prefsrc entries */
4602 struct arg_dev_net_ip {
4603 struct net *net;
4604 struct in6_addr *addr;
4605 };
4606
fib6_remove_prefsrc(struct fib6_info * rt,void * arg)4607 static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4608 {
4609 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4610 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4611
4612 if (!rt->nh &&
4613 rt != net->ipv6.fib6_null_entry &&
4614 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr) &&
4615 !ipv6_chk_addr(net, addr, rt->fib6_nh->fib_nh_dev, 0)) {
4616 spin_lock_bh(&rt6_exception_lock);
4617 /* remove prefsrc entry */
4618 rt->fib6_prefsrc.plen = 0;
4619 spin_unlock_bh(&rt6_exception_lock);
4620 }
4621 return 0;
4622 }
4623
rt6_remove_prefsrc(struct inet6_ifaddr * ifp)4624 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4625 {
4626 struct net *net = dev_net(ifp->idev->dev);
4627 struct arg_dev_net_ip adni = {
4628 .net = net,
4629 .addr = &ifp->addr,
4630 };
4631 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4632 }
4633
4634 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
4635
4636 /* Remove routers and update dst entries when gateway turn into host. */
fib6_clean_tohost(struct fib6_info * rt,void * arg)4637 static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4638 {
4639 struct in6_addr *gateway = (struct in6_addr *)arg;
4640 struct fib6_nh *nh;
4641
4642 /* RA routes do not use nexthops */
4643 if (rt->nh)
4644 return 0;
4645
4646 nh = rt->fib6_nh;
4647 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4648 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4649 return -1;
4650
4651 /* Further clean up cached routes in exception table.
4652 * This is needed because cached route may have a different
4653 * gateway than its 'parent' in the case of an ip redirect.
4654 */
4655 fib6_nh_exceptions_clean_tohost(nh, gateway);
4656
4657 return 0;
4658 }
4659
rt6_clean_tohost(struct net * net,struct in6_addr * gateway)4660 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4661 {
4662 fib6_clean_all(net, fib6_clean_tohost, gateway);
4663 }
4664
4665 struct arg_netdev_event {
4666 const struct net_device *dev;
4667 union {
4668 unsigned char nh_flags;
4669 unsigned long event;
4670 };
4671 };
4672
rt6_multipath_first_sibling(const struct fib6_info * rt)4673 static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4674 {
4675 struct fib6_info *iter;
4676 struct fib6_node *fn;
4677
4678 fn = rcu_dereference_protected(rt->fib6_node,
4679 lockdep_is_held(&rt->fib6_table->tb6_lock));
4680 iter = rcu_dereference_protected(fn->leaf,
4681 lockdep_is_held(&rt->fib6_table->tb6_lock));
4682 while (iter) {
4683 if (iter->fib6_metric == rt->fib6_metric &&
4684 rt6_qualify_for_ecmp(iter))
4685 return iter;
4686 iter = rcu_dereference_protected(iter->fib6_next,
4687 lockdep_is_held(&rt->fib6_table->tb6_lock));
4688 }
4689
4690 return NULL;
4691 }
4692
4693 /* only called for fib entries with builtin fib6_nh */
rt6_is_dead(const struct fib6_info * rt)4694 static bool rt6_is_dead(const struct fib6_info *rt)
4695 {
4696 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4697 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4698 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4699 return true;
4700
4701 return false;
4702 }
4703
rt6_multipath_total_weight(const struct fib6_info * rt)4704 static int rt6_multipath_total_weight(const struct fib6_info *rt)
4705 {
4706 struct fib6_info *iter;
4707 int total = 0;
4708
4709 if (!rt6_is_dead(rt))
4710 total += rt->fib6_nh->fib_nh_weight;
4711
4712 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4713 if (!rt6_is_dead(iter))
4714 total += iter->fib6_nh->fib_nh_weight;
4715 }
4716
4717 return total;
4718 }
4719
rt6_upper_bound_set(struct fib6_info * rt,int * weight,int total)4720 static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4721 {
4722 int upper_bound = -1;
4723
4724 if (!rt6_is_dead(rt)) {
4725 *weight += rt->fib6_nh->fib_nh_weight;
4726 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4727 total) - 1;
4728 }
4729 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4730 }
4731
rt6_multipath_upper_bound_set(struct fib6_info * rt,int total)4732 static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4733 {
4734 struct fib6_info *iter;
4735 int weight = 0;
4736
4737 rt6_upper_bound_set(rt, &weight, total);
4738
4739 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4740 rt6_upper_bound_set(iter, &weight, total);
4741 }
4742
rt6_multipath_rebalance(struct fib6_info * rt)4743 void rt6_multipath_rebalance(struct fib6_info *rt)
4744 {
4745 struct fib6_info *first;
4746 int total;
4747
4748 /* In case the entire multipath route was marked for flushing,
4749 * then there is no need to rebalance upon the removal of every
4750 * sibling route.
4751 */
4752 if (!rt->fib6_nsiblings || rt->should_flush)
4753 return;
4754
4755 /* During lookup routes are evaluated in order, so we need to
4756 * make sure upper bounds are assigned from the first sibling
4757 * onwards.
4758 */
4759 first = rt6_multipath_first_sibling(rt);
4760 if (WARN_ON_ONCE(!first))
4761 return;
4762
4763 total = rt6_multipath_total_weight(first);
4764 rt6_multipath_upper_bound_set(first, total);
4765 }
4766
fib6_ifup(struct fib6_info * rt,void * p_arg)4767 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4768 {
4769 const struct arg_netdev_event *arg = p_arg;
4770 struct net *net = dev_net(arg->dev);
4771
4772 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4773 rt->fib6_nh->fib_nh_dev == arg->dev) {
4774 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4775 fib6_update_sernum_upto_root(net, rt);
4776 rt6_multipath_rebalance(rt);
4777 }
4778
4779 return 0;
4780 }
4781
rt6_sync_up(struct net_device * dev,unsigned char nh_flags)4782 void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4783 {
4784 struct arg_netdev_event arg = {
4785 .dev = dev,
4786 {
4787 .nh_flags = nh_flags,
4788 },
4789 };
4790
4791 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4792 arg.nh_flags |= RTNH_F_LINKDOWN;
4793
4794 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4795 }
4796
4797 /* only called for fib entries with inline fib6_nh */
rt6_multipath_uses_dev(const struct fib6_info * rt,const struct net_device * dev)4798 static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4799 const struct net_device *dev)
4800 {
4801 struct fib6_info *iter;
4802
4803 if (rt->fib6_nh->fib_nh_dev == dev)
4804 return true;
4805 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4806 if (iter->fib6_nh->fib_nh_dev == dev)
4807 return true;
4808
4809 return false;
4810 }
4811
rt6_multipath_flush(struct fib6_info * rt)4812 static void rt6_multipath_flush(struct fib6_info *rt)
4813 {
4814 struct fib6_info *iter;
4815
4816 rt->should_flush = 1;
4817 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4818 iter->should_flush = 1;
4819 }
4820
rt6_multipath_dead_count(const struct fib6_info * rt,const struct net_device * down_dev)4821 static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4822 const struct net_device *down_dev)
4823 {
4824 struct fib6_info *iter;
4825 unsigned int dead = 0;
4826
4827 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4828 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4829 dead++;
4830 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4831 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4832 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4833 dead++;
4834
4835 return dead;
4836 }
4837
rt6_multipath_nh_flags_set(struct fib6_info * rt,const struct net_device * dev,unsigned char nh_flags)4838 static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4839 const struct net_device *dev,
4840 unsigned char nh_flags)
4841 {
4842 struct fib6_info *iter;
4843
4844 if (rt->fib6_nh->fib_nh_dev == dev)
4845 rt->fib6_nh->fib_nh_flags |= nh_flags;
4846 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4847 if (iter->fib6_nh->fib_nh_dev == dev)
4848 iter->fib6_nh->fib_nh_flags |= nh_flags;
4849 }
4850
4851 /* called with write lock held for table with rt */
fib6_ifdown(struct fib6_info * rt,void * p_arg)4852 static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4853 {
4854 const struct arg_netdev_event *arg = p_arg;
4855 const struct net_device *dev = arg->dev;
4856 struct net *net = dev_net(dev);
4857
4858 if (rt == net->ipv6.fib6_null_entry || rt->nh)
4859 return 0;
4860
4861 switch (arg->event) {
4862 case NETDEV_UNREGISTER:
4863 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4864 case NETDEV_DOWN:
4865 if (rt->should_flush)
4866 return -1;
4867 if (!rt->fib6_nsiblings)
4868 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4869 if (rt6_multipath_uses_dev(rt, dev)) {
4870 unsigned int count;
4871
4872 count = rt6_multipath_dead_count(rt, dev);
4873 if (rt->fib6_nsiblings + 1 == count) {
4874 rt6_multipath_flush(rt);
4875 return -1;
4876 }
4877 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4878 RTNH_F_LINKDOWN);
4879 fib6_update_sernum(net, rt);
4880 rt6_multipath_rebalance(rt);
4881 }
4882 return -2;
4883 case NETDEV_CHANGE:
4884 if (rt->fib6_nh->fib_nh_dev != dev ||
4885 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4886 break;
4887 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4888 rt6_multipath_rebalance(rt);
4889 break;
4890 }
4891
4892 return 0;
4893 }
4894
rt6_sync_down_dev(struct net_device * dev,unsigned long event)4895 void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4896 {
4897 struct arg_netdev_event arg = {
4898 .dev = dev,
4899 {
4900 .event = event,
4901 },
4902 };
4903 struct net *net = dev_net(dev);
4904
4905 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4906 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4907 else
4908 fib6_clean_all(net, fib6_ifdown, &arg);
4909 }
4910
rt6_disable_ip(struct net_device * dev,unsigned long event)4911 void rt6_disable_ip(struct net_device *dev, unsigned long event)
4912 {
4913 rt6_sync_down_dev(dev, event);
4914 rt6_uncached_list_flush_dev(dev);
4915 neigh_ifdown(&nd_tbl, dev);
4916 }
4917
4918 struct rt6_mtu_change_arg {
4919 struct net_device *dev;
4920 unsigned int mtu;
4921 struct fib6_info *f6i;
4922 };
4923
fib6_nh_mtu_change(struct fib6_nh * nh,void * _arg)4924 static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4925 {
4926 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4927 struct fib6_info *f6i = arg->f6i;
4928
4929 /* For administrative MTU increase, there is no way to discover
4930 * IPv6 PMTU increase, so PMTU increase should be updated here.
4931 * Since RFC 1981 doesn't include administrative MTU increase
4932 * update PMTU increase is a MUST. (i.e. jumbo frame)
4933 */
4934 if (nh->fib_nh_dev == arg->dev) {
4935 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4936 u32 mtu = f6i->fib6_pmtu;
4937
4938 if (mtu >= arg->mtu ||
4939 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4940 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4941
4942 spin_lock_bh(&rt6_exception_lock);
4943 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4944 spin_unlock_bh(&rt6_exception_lock);
4945 }
4946
4947 return 0;
4948 }
4949
rt6_mtu_change_route(struct fib6_info * f6i,void * p_arg)4950 static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4951 {
4952 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4953 struct inet6_dev *idev;
4954
4955 /* In IPv6 pmtu discovery is not optional,
4956 so that RTAX_MTU lock cannot disable it.
4957 We still use this lock to block changes
4958 caused by addrconf/ndisc.
4959 */
4960
4961 idev = __in6_dev_get(arg->dev);
4962 if (!idev)
4963 return 0;
4964
4965 if (fib6_metric_locked(f6i, RTAX_MTU))
4966 return 0;
4967
4968 arg->f6i = f6i;
4969 if (f6i->nh) {
4970 /* fib6_nh_mtu_change only returns 0, so this is safe */
4971 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4972 arg);
4973 }
4974
4975 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4976 }
4977
rt6_mtu_change(struct net_device * dev,unsigned int mtu)4978 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4979 {
4980 struct rt6_mtu_change_arg arg = {
4981 .dev = dev,
4982 .mtu = mtu,
4983 };
4984
4985 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4986 }
4987
4988 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4989 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
4990 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
4991 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
4992 [RTA_OIF] = { .type = NLA_U32 },
4993 [RTA_IIF] = { .type = NLA_U32 },
4994 [RTA_PRIORITY] = { .type = NLA_U32 },
4995 [RTA_METRICS] = { .type = NLA_NESTED },
4996 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
4997 [RTA_PREF] = { .type = NLA_U8 },
4998 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4999 [RTA_ENCAP] = { .type = NLA_NESTED },
5000 [RTA_EXPIRES] = { .type = NLA_U32 },
5001 [RTA_UID] = { .type = NLA_U32 },
5002 [RTA_MARK] = { .type = NLA_U32 },
5003 [RTA_TABLE] = { .type = NLA_U32 },
5004 [RTA_IP_PROTO] = { .type = NLA_U8 },
5005 [RTA_SPORT] = { .type = NLA_U16 },
5006 [RTA_DPORT] = { .type = NLA_U16 },
5007 [RTA_NH_ID] = { .type = NLA_U32 },
5008 };
5009
rtm_to_fib6_config(struct sk_buff * skb,struct nlmsghdr * nlh,struct fib6_config * cfg,struct netlink_ext_ack * extack)5010 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
5011 struct fib6_config *cfg,
5012 struct netlink_ext_ack *extack)
5013 {
5014 struct rtmsg *rtm;
5015 struct nlattr *tb[RTA_MAX+1];
5016 unsigned int pref;
5017 int err;
5018
5019 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5020 rtm_ipv6_policy, extack);
5021 if (err < 0)
5022 goto errout;
5023
5024 err = -EINVAL;
5025 rtm = nlmsg_data(nlh);
5026
5027 if (rtm->rtm_tos) {
5028 NL_SET_ERR_MSG(extack,
5029 "Invalid dsfield (tos): option not available for IPv6");
5030 goto errout;
5031 }
5032
5033 *cfg = (struct fib6_config){
5034 .fc_table = rtm->rtm_table,
5035 .fc_dst_len = rtm->rtm_dst_len,
5036 .fc_src_len = rtm->rtm_src_len,
5037 .fc_flags = RTF_UP,
5038 .fc_protocol = rtm->rtm_protocol,
5039 .fc_type = rtm->rtm_type,
5040
5041 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
5042 .fc_nlinfo.nlh = nlh,
5043 .fc_nlinfo.nl_net = sock_net(skb->sk),
5044 };
5045
5046 if (rtm->rtm_type == RTN_UNREACHABLE ||
5047 rtm->rtm_type == RTN_BLACKHOLE ||
5048 rtm->rtm_type == RTN_PROHIBIT ||
5049 rtm->rtm_type == RTN_THROW)
5050 cfg->fc_flags |= RTF_REJECT;
5051
5052 if (rtm->rtm_type == RTN_LOCAL)
5053 cfg->fc_flags |= RTF_LOCAL;
5054
5055 if (rtm->rtm_flags & RTM_F_CLONED)
5056 cfg->fc_flags |= RTF_CACHE;
5057
5058 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
5059
5060 if (tb[RTA_NH_ID]) {
5061 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
5062 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
5063 NL_SET_ERR_MSG(extack,
5064 "Nexthop specification and nexthop id are mutually exclusive");
5065 goto errout;
5066 }
5067 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
5068 }
5069
5070 if (tb[RTA_GATEWAY]) {
5071 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
5072 cfg->fc_flags |= RTF_GATEWAY;
5073 }
5074 if (tb[RTA_VIA]) {
5075 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
5076 goto errout;
5077 }
5078
5079 if (tb[RTA_DST]) {
5080 int plen = (rtm->rtm_dst_len + 7) >> 3;
5081
5082 if (nla_len(tb[RTA_DST]) < plen)
5083 goto errout;
5084
5085 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
5086 }
5087
5088 if (tb[RTA_SRC]) {
5089 int plen = (rtm->rtm_src_len + 7) >> 3;
5090
5091 if (nla_len(tb[RTA_SRC]) < plen)
5092 goto errout;
5093
5094 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
5095 }
5096
5097 if (tb[RTA_PREFSRC])
5098 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
5099
5100 if (tb[RTA_OIF])
5101 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
5102
5103 if (tb[RTA_PRIORITY])
5104 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
5105
5106 if (tb[RTA_METRICS]) {
5107 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
5108 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
5109 }
5110
5111 if (tb[RTA_TABLE])
5112 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
5113
5114 if (tb[RTA_MULTIPATH]) {
5115 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
5116 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
5117
5118 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
5119 cfg->fc_mp_len, extack);
5120 if (err < 0)
5121 goto errout;
5122 }
5123
5124 if (tb[RTA_PREF]) {
5125 pref = nla_get_u8(tb[RTA_PREF]);
5126 if (pref != ICMPV6_ROUTER_PREF_LOW &&
5127 pref != ICMPV6_ROUTER_PREF_HIGH)
5128 pref = ICMPV6_ROUTER_PREF_MEDIUM;
5129 cfg->fc_flags |= RTF_PREF(pref);
5130 }
5131
5132 if (tb[RTA_ENCAP])
5133 cfg->fc_encap = tb[RTA_ENCAP];
5134
5135 if (tb[RTA_ENCAP_TYPE]) {
5136 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5137
5138 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
5139 if (err < 0)
5140 goto errout;
5141 }
5142
5143 if (tb[RTA_EXPIRES]) {
5144 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5145
5146 if (addrconf_finite_timeout(timeout)) {
5147 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5148 cfg->fc_flags |= RTF_EXPIRES;
5149 }
5150 }
5151
5152 err = 0;
5153 errout:
5154 return err;
5155 }
5156
5157 struct rt6_nh {
5158 struct fib6_info *fib6_info;
5159 struct fib6_config r_cfg;
5160 struct list_head next;
5161 };
5162
ip6_route_info_append(struct net * net,struct list_head * rt6_nh_list,struct fib6_info * rt,struct fib6_config * r_cfg)5163 static int ip6_route_info_append(struct net *net,
5164 struct list_head *rt6_nh_list,
5165 struct fib6_info *rt,
5166 struct fib6_config *r_cfg)
5167 {
5168 struct rt6_nh *nh;
5169 int err = -EEXIST;
5170
5171 list_for_each_entry(nh, rt6_nh_list, next) {
5172 /* check if fib6_info already exists */
5173 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
5174 return err;
5175 }
5176
5177 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5178 if (!nh)
5179 return -ENOMEM;
5180 nh->fib6_info = rt;
5181 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5182 list_add_tail(&nh->next, rt6_nh_list);
5183
5184 return 0;
5185 }
5186
ip6_route_mpath_notify(struct fib6_info * rt,struct fib6_info * rt_last,struct nl_info * info,__u16 nlflags)5187 static void ip6_route_mpath_notify(struct fib6_info *rt,
5188 struct fib6_info *rt_last,
5189 struct nl_info *info,
5190 __u16 nlflags)
5191 {
5192 /* if this is an APPEND route, then rt points to the first route
5193 * inserted and rt_last points to last route inserted. Userspace
5194 * wants a consistent dump of the route which starts at the first
5195 * nexthop. Since sibling routes are always added at the end of
5196 * the list, find the first sibling of the last route appended
5197 */
5198 rcu_read_lock();
5199
5200 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5201 rt = list_first_or_null_rcu(&rt_last->fib6_siblings,
5202 struct fib6_info,
5203 fib6_siblings);
5204 }
5205
5206 if (rt)
5207 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5208
5209 rcu_read_unlock();
5210 }
5211
ip6_route_mpath_should_notify(const struct fib6_info * rt)5212 static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5213 {
5214 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5215 bool should_notify = false;
5216 struct fib6_info *leaf;
5217 struct fib6_node *fn;
5218
5219 rcu_read_lock();
5220 fn = rcu_dereference(rt->fib6_node);
5221 if (!fn)
5222 goto out;
5223
5224 leaf = rcu_dereference(fn->leaf);
5225 if (!leaf)
5226 goto out;
5227
5228 if (rt == leaf ||
5229 (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5230 rt6_qualify_for_ecmp(leaf)))
5231 should_notify = true;
5232 out:
5233 rcu_read_unlock();
5234
5235 return should_notify;
5236 }
5237
fib6_gw_from_attr(struct in6_addr * gw,struct nlattr * nla,struct netlink_ext_ack * extack)5238 static int fib6_gw_from_attr(struct in6_addr *gw, struct nlattr *nla,
5239 struct netlink_ext_ack *extack)
5240 {
5241 if (nla_len(nla) < sizeof(*gw)) {
5242 NL_SET_ERR_MSG(extack, "Invalid IPv6 address in RTA_GATEWAY");
5243 return -EINVAL;
5244 }
5245
5246 *gw = nla_get_in6_addr(nla);
5247
5248 return 0;
5249 }
5250
ip6_route_multipath_add(struct fib6_config * cfg,struct netlink_ext_ack * extack)5251 static int ip6_route_multipath_add(struct fib6_config *cfg,
5252 struct netlink_ext_ack *extack)
5253 {
5254 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5255 struct nl_info *info = &cfg->fc_nlinfo;
5256 struct fib6_config r_cfg;
5257 struct rtnexthop *rtnh;
5258 struct fib6_info *rt;
5259 struct rt6_nh *err_nh;
5260 struct rt6_nh *nh, *nh_safe;
5261 __u16 nlflags;
5262 int remaining;
5263 int attrlen;
5264 int err = 1;
5265 int nhn = 0;
5266 int replace = (cfg->fc_nlinfo.nlh &&
5267 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5268 LIST_HEAD(rt6_nh_list);
5269
5270 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5271 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5272 nlflags |= NLM_F_APPEND;
5273
5274 remaining = cfg->fc_mp_len;
5275 rtnh = (struct rtnexthop *)cfg->fc_mp;
5276
5277 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
5278 * fib6_info structs per nexthop
5279 */
5280 while (rtnh_ok(rtnh, remaining)) {
5281 memcpy(&r_cfg, cfg, sizeof(*cfg));
5282 if (rtnh->rtnh_ifindex)
5283 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5284
5285 attrlen = rtnh_attrlen(rtnh);
5286 if (attrlen > 0) {
5287 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5288
5289 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5290 if (nla) {
5291 err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5292 extack);
5293 if (err)
5294 goto cleanup;
5295
5296 r_cfg.fc_flags |= RTF_GATEWAY;
5297 }
5298 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5299
5300 /* RTA_ENCAP_TYPE length checked in
5301 * lwtunnel_valid_encap_type_attr
5302 */
5303 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5304 if (nla)
5305 r_cfg.fc_encap_type = nla_get_u16(nla);
5306 }
5307
5308 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5309 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5310 if (IS_ERR(rt)) {
5311 err = PTR_ERR(rt);
5312 rt = NULL;
5313 goto cleanup;
5314 }
5315 if (!rt6_qualify_for_ecmp(rt)) {
5316 err = -EINVAL;
5317 NL_SET_ERR_MSG(extack,
5318 "Device only routes can not be added for IPv6 using the multipath API.");
5319 fib6_info_release(rt);
5320 goto cleanup;
5321 }
5322
5323 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5324
5325 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5326 rt, &r_cfg);
5327 if (err) {
5328 fib6_info_release(rt);
5329 goto cleanup;
5330 }
5331
5332 rtnh = rtnh_next(rtnh, &remaining);
5333 }
5334
5335 if (list_empty(&rt6_nh_list)) {
5336 NL_SET_ERR_MSG(extack,
5337 "Invalid nexthop configuration - no valid nexthops");
5338 return -EINVAL;
5339 }
5340
5341 /* for add and replace send one notification with all nexthops.
5342 * Skip the notification in fib6_add_rt2node and send one with
5343 * the full route when done
5344 */
5345 info->skip_notify = 1;
5346
5347 /* For add and replace, send one notification with all nexthops. For
5348 * append, send one notification with all appended nexthops.
5349 */
5350 info->skip_notify_kernel = 1;
5351
5352 err_nh = NULL;
5353 list_for_each_entry(nh, &rt6_nh_list, next) {
5354 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5355
5356 if (err) {
5357 if (replace && nhn)
5358 NL_SET_ERR_MSG_MOD(extack,
5359 "multipath route replace failed (check consistency of installed routes)");
5360 err_nh = nh;
5361 goto add_errout;
5362 }
5363 /* save reference to last route successfully inserted */
5364 rt_last = nh->fib6_info;
5365
5366 /* save reference to first route for notification */
5367 if (!rt_notif)
5368 rt_notif = nh->fib6_info;
5369
5370 /* Because each route is added like a single route we remove
5371 * these flags after the first nexthop: if there is a collision,
5372 * we have already failed to add the first nexthop:
5373 * fib6_add_rt2node() has rejected it; when replacing, old
5374 * nexthops have been replaced by first new, the rest should
5375 * be added to it.
5376 */
5377 if (cfg->fc_nlinfo.nlh) {
5378 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5379 NLM_F_REPLACE);
5380 cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
5381 }
5382 nhn++;
5383 }
5384
5385 /* An in-kernel notification should only be sent in case the new
5386 * multipath route is added as the first route in the node, or if
5387 * it was appended to it. We pass 'rt_notif' since it is the first
5388 * sibling and might allow us to skip some checks in the replace case.
5389 */
5390 if (ip6_route_mpath_should_notify(rt_notif)) {
5391 enum fib_event_type fib_event;
5392
5393 if (rt_notif->fib6_nsiblings != nhn - 1)
5394 fib_event = FIB_EVENT_ENTRY_APPEND;
5395 else
5396 fib_event = FIB_EVENT_ENTRY_REPLACE;
5397
5398 err = call_fib6_multipath_entry_notifiers(info->nl_net,
5399 fib_event, rt_notif,
5400 nhn - 1, extack);
5401 if (err) {
5402 /* Delete all the siblings that were just added */
5403 err_nh = NULL;
5404 goto add_errout;
5405 }
5406 }
5407
5408 /* success ... tell user about new route */
5409 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5410 goto cleanup;
5411
5412 add_errout:
5413 /* send notification for routes that were added so that
5414 * the delete notifications sent by ip6_route_del are
5415 * coherent
5416 */
5417 if (rt_notif)
5418 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5419
5420 /* Delete routes that were already added */
5421 list_for_each_entry(nh, &rt6_nh_list, next) {
5422 if (err_nh == nh)
5423 break;
5424 ip6_route_del(&nh->r_cfg, extack);
5425 }
5426
5427 cleanup:
5428 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5429 fib6_info_release(nh->fib6_info);
5430 list_del(&nh->next);
5431 kfree(nh);
5432 }
5433
5434 return err;
5435 }
5436
ip6_route_multipath_del(struct fib6_config * cfg,struct netlink_ext_ack * extack)5437 static int ip6_route_multipath_del(struct fib6_config *cfg,
5438 struct netlink_ext_ack *extack)
5439 {
5440 struct fib6_config r_cfg;
5441 struct rtnexthop *rtnh;
5442 int last_err = 0;
5443 int remaining;
5444 int attrlen;
5445 int err;
5446
5447 remaining = cfg->fc_mp_len;
5448 rtnh = (struct rtnexthop *)cfg->fc_mp;
5449
5450 /* Parse a Multipath Entry */
5451 while (rtnh_ok(rtnh, remaining)) {
5452 memcpy(&r_cfg, cfg, sizeof(*cfg));
5453 if (rtnh->rtnh_ifindex)
5454 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5455
5456 attrlen = rtnh_attrlen(rtnh);
5457 if (attrlen > 0) {
5458 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5459
5460 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5461 if (nla) {
5462 err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5463 extack);
5464 if (err) {
5465 last_err = err;
5466 goto next_rtnh;
5467 }
5468
5469 r_cfg.fc_flags |= RTF_GATEWAY;
5470 }
5471 }
5472 err = ip6_route_del(&r_cfg, extack);
5473 if (err)
5474 last_err = err;
5475
5476 next_rtnh:
5477 rtnh = rtnh_next(rtnh, &remaining);
5478 }
5479
5480 return last_err;
5481 }
5482
inet6_rtm_delroute(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)5483 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5484 struct netlink_ext_ack *extack)
5485 {
5486 struct fib6_config cfg;
5487 int err;
5488
5489 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5490 if (err < 0)
5491 return err;
5492
5493 if (cfg.fc_nh_id &&
5494 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5495 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5496 return -EINVAL;
5497 }
5498
5499 if (cfg.fc_mp)
5500 return ip6_route_multipath_del(&cfg, extack);
5501 else {
5502 cfg.fc_delete_all_nh = 1;
5503 return ip6_route_del(&cfg, extack);
5504 }
5505 }
5506
inet6_rtm_newroute(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)5507 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5508 struct netlink_ext_ack *extack)
5509 {
5510 struct fib6_config cfg;
5511 int err;
5512
5513 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5514 if (err < 0)
5515 return err;
5516
5517 if (cfg.fc_metric == 0)
5518 cfg.fc_metric = IP6_RT_PRIO_USER;
5519
5520 if (cfg.fc_mp)
5521 return ip6_route_multipath_add(&cfg, extack);
5522 else
5523 return ip6_route_add(&cfg, GFP_KERNEL, extack);
5524 }
5525
5526 /* add the overhead of this fib6_nh to nexthop_len */
rt6_nh_nlmsg_size(struct fib6_nh * nh,void * arg)5527 static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5528 {
5529 int *nexthop_len = arg;
5530
5531 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5532 + NLA_ALIGN(sizeof(struct rtnexthop))
5533 + nla_total_size(16); /* RTA_GATEWAY */
5534
5535 if (nh->fib_nh_lws) {
5536 /* RTA_ENCAP_TYPE */
5537 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5538 /* RTA_ENCAP */
5539 *nexthop_len += nla_total_size(2);
5540 }
5541
5542 return 0;
5543 }
5544
rt6_nlmsg_size(struct fib6_info * f6i)5545 static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5546 {
5547 int nexthop_len;
5548
5549 if (f6i->nh) {
5550 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5551 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5552 &nexthop_len);
5553 } else {
5554 struct fib6_nh *nh = f6i->fib6_nh;
5555 struct fib6_info *sibling;
5556
5557 nexthop_len = 0;
5558 if (f6i->fib6_nsiblings) {
5559 rt6_nh_nlmsg_size(nh, &nexthop_len);
5560
5561 rcu_read_lock();
5562
5563 list_for_each_entry_rcu(sibling, &f6i->fib6_siblings,
5564 fib6_siblings) {
5565 rt6_nh_nlmsg_size(sibling->fib6_nh, &nexthop_len);
5566 }
5567
5568 rcu_read_unlock();
5569 }
5570 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5571 }
5572
5573 return NLMSG_ALIGN(sizeof(struct rtmsg))
5574 + nla_total_size(16) /* RTA_SRC */
5575 + nla_total_size(16) /* RTA_DST */
5576 + nla_total_size(16) /* RTA_GATEWAY */
5577 + nla_total_size(16) /* RTA_PREFSRC */
5578 + nla_total_size(4) /* RTA_TABLE */
5579 + nla_total_size(4) /* RTA_IIF */
5580 + nla_total_size(4) /* RTA_OIF */
5581 + nla_total_size(4) /* RTA_PRIORITY */
5582 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5583 + nla_total_size(sizeof(struct rta_cacheinfo))
5584 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5585 + nla_total_size(1) /* RTA_PREF */
5586 + nexthop_len;
5587 }
5588
rt6_fill_node_nexthop(struct sk_buff * skb,struct nexthop * nh,unsigned char * flags)5589 static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5590 unsigned char *flags)
5591 {
5592 if (nexthop_is_multipath(nh)) {
5593 struct nlattr *mp;
5594
5595 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5596 if (!mp)
5597 goto nla_put_failure;
5598
5599 if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
5600 goto nla_put_failure;
5601
5602 nla_nest_end(skb, mp);
5603 } else {
5604 struct fib6_nh *fib6_nh;
5605
5606 fib6_nh = nexthop_fib6_nh(nh);
5607 if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
5608 flags, false) < 0)
5609 goto nla_put_failure;
5610 }
5611
5612 return 0;
5613
5614 nla_put_failure:
5615 return -EMSGSIZE;
5616 }
5617
rt6_fill_node(struct net * net,struct sk_buff * skb,struct fib6_info * rt,struct dst_entry * dst,struct in6_addr * dest,struct in6_addr * src,int iif,int type,u32 portid,u32 seq,unsigned int flags)5618 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5619 struct fib6_info *rt, struct dst_entry *dst,
5620 struct in6_addr *dest, struct in6_addr *src,
5621 int iif, int type, u32 portid, u32 seq,
5622 unsigned int flags)
5623 {
5624 struct rt6_info *rt6 = dst_rt6_info(dst);
5625 struct rt6key *rt6_dst, *rt6_src;
5626 u32 *pmetrics, table, rt6_flags;
5627 unsigned char nh_flags = 0;
5628 struct nlmsghdr *nlh;
5629 struct rtmsg *rtm;
5630 long expires = 0;
5631
5632 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5633 if (!nlh)
5634 return -EMSGSIZE;
5635
5636 if (rt6) {
5637 rt6_dst = &rt6->rt6i_dst;
5638 rt6_src = &rt6->rt6i_src;
5639 rt6_flags = rt6->rt6i_flags;
5640 } else {
5641 rt6_dst = &rt->fib6_dst;
5642 rt6_src = &rt->fib6_src;
5643 rt6_flags = rt->fib6_flags;
5644 }
5645
5646 rtm = nlmsg_data(nlh);
5647 rtm->rtm_family = AF_INET6;
5648 rtm->rtm_dst_len = rt6_dst->plen;
5649 rtm->rtm_src_len = rt6_src->plen;
5650 rtm->rtm_tos = 0;
5651 if (rt->fib6_table)
5652 table = rt->fib6_table->tb6_id;
5653 else
5654 table = RT6_TABLE_UNSPEC;
5655 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5656 if (nla_put_u32(skb, RTA_TABLE, table))
5657 goto nla_put_failure;
5658
5659 rtm->rtm_type = rt->fib6_type;
5660 rtm->rtm_flags = 0;
5661 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5662 rtm->rtm_protocol = rt->fib6_protocol;
5663
5664 if (rt6_flags & RTF_CACHE)
5665 rtm->rtm_flags |= RTM_F_CLONED;
5666
5667 if (dest) {
5668 if (nla_put_in6_addr(skb, RTA_DST, dest))
5669 goto nla_put_failure;
5670 rtm->rtm_dst_len = 128;
5671 } else if (rtm->rtm_dst_len)
5672 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5673 goto nla_put_failure;
5674 #ifdef CONFIG_IPV6_SUBTREES
5675 if (src) {
5676 if (nla_put_in6_addr(skb, RTA_SRC, src))
5677 goto nla_put_failure;
5678 rtm->rtm_src_len = 128;
5679 } else if (rtm->rtm_src_len &&
5680 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5681 goto nla_put_failure;
5682 #endif
5683 if (iif) {
5684 #ifdef CONFIG_IPV6_MROUTE
5685 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5686 int err = ip6mr_get_route(net, skb, rtm, portid);
5687
5688 if (err == 0)
5689 return 0;
5690 if (err < 0)
5691 goto nla_put_failure;
5692 } else
5693 #endif
5694 if (nla_put_u32(skb, RTA_IIF, iif))
5695 goto nla_put_failure;
5696 } else if (dest) {
5697 struct in6_addr saddr_buf;
5698 if (ip6_route_get_saddr(net, rt, dest, 0, 0, &saddr_buf) == 0 &&
5699 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5700 goto nla_put_failure;
5701 }
5702
5703 if (rt->fib6_prefsrc.plen) {
5704 struct in6_addr saddr_buf;
5705 saddr_buf = rt->fib6_prefsrc.addr;
5706 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5707 goto nla_put_failure;
5708 }
5709
5710 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5711 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5712 goto nla_put_failure;
5713
5714 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5715 goto nla_put_failure;
5716
5717 /* For multipath routes, walk the siblings list and add
5718 * each as a nexthop within RTA_MULTIPATH.
5719 */
5720 if (rt6) {
5721 if (rt6_flags & RTF_GATEWAY &&
5722 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5723 goto nla_put_failure;
5724
5725 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5726 goto nla_put_failure;
5727
5728 if (dst->lwtstate &&
5729 lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
5730 goto nla_put_failure;
5731 } else if (rt->fib6_nsiblings) {
5732 struct fib6_info *sibling;
5733 struct nlattr *mp;
5734
5735 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5736 if (!mp)
5737 goto nla_put_failure;
5738
5739 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5740 rt->fib6_nh->fib_nh_weight, AF_INET6,
5741 0) < 0)
5742 goto nla_put_failure;
5743
5744 rcu_read_lock();
5745
5746 list_for_each_entry_rcu(sibling, &rt->fib6_siblings,
5747 fib6_siblings) {
5748 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5749 sibling->fib6_nh->fib_nh_weight,
5750 AF_INET6, 0) < 0) {
5751 rcu_read_unlock();
5752
5753 goto nla_put_failure;
5754 }
5755 }
5756
5757 rcu_read_unlock();
5758
5759 nla_nest_end(skb, mp);
5760 } else if (rt->nh) {
5761 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5762 goto nla_put_failure;
5763
5764 if (nexthop_is_blackhole(rt->nh))
5765 rtm->rtm_type = RTN_BLACKHOLE;
5766
5767 if (READ_ONCE(net->ipv4.sysctl_nexthop_compat_mode) &&
5768 rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5769 goto nla_put_failure;
5770
5771 rtm->rtm_flags |= nh_flags;
5772 } else {
5773 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
5774 &nh_flags, false) < 0)
5775 goto nla_put_failure;
5776
5777 rtm->rtm_flags |= nh_flags;
5778 }
5779
5780 if (rt6_flags & RTF_EXPIRES) {
5781 expires = dst ? dst->expires : rt->expires;
5782 expires -= jiffies;
5783 }
5784
5785 if (!dst) {
5786 if (READ_ONCE(rt->offload))
5787 rtm->rtm_flags |= RTM_F_OFFLOAD;
5788 if (READ_ONCE(rt->trap))
5789 rtm->rtm_flags |= RTM_F_TRAP;
5790 if (READ_ONCE(rt->offload_failed))
5791 rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
5792 }
5793
5794 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5795 goto nla_put_failure;
5796
5797 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5798 goto nla_put_failure;
5799
5800
5801 nlmsg_end(skb, nlh);
5802 return 0;
5803
5804 nla_put_failure:
5805 nlmsg_cancel(skb, nlh);
5806 return -EMSGSIZE;
5807 }
5808
fib6_info_nh_uses_dev(struct fib6_nh * nh,void * arg)5809 static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5810 {
5811 const struct net_device *dev = arg;
5812
5813 if (nh->fib_nh_dev == dev)
5814 return 1;
5815
5816 return 0;
5817 }
5818
fib6_info_uses_dev(const struct fib6_info * f6i,const struct net_device * dev)5819 static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5820 const struct net_device *dev)
5821 {
5822 if (f6i->nh) {
5823 struct net_device *_dev = (struct net_device *)dev;
5824
5825 return !!nexthop_for_each_fib6_nh(f6i->nh,
5826 fib6_info_nh_uses_dev,
5827 _dev);
5828 }
5829
5830 if (f6i->fib6_nh->fib_nh_dev == dev)
5831 return true;
5832
5833 if (f6i->fib6_nsiblings) {
5834 struct fib6_info *sibling, *next_sibling;
5835
5836 list_for_each_entry_safe(sibling, next_sibling,
5837 &f6i->fib6_siblings, fib6_siblings) {
5838 if (sibling->fib6_nh->fib_nh_dev == dev)
5839 return true;
5840 }
5841 }
5842
5843 return false;
5844 }
5845
5846 struct fib6_nh_exception_dump_walker {
5847 struct rt6_rtnl_dump_arg *dump;
5848 struct fib6_info *rt;
5849 unsigned int flags;
5850 unsigned int skip;
5851 unsigned int count;
5852 };
5853
rt6_nh_dump_exceptions(struct fib6_nh * nh,void * arg)5854 static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5855 {
5856 struct fib6_nh_exception_dump_walker *w = arg;
5857 struct rt6_rtnl_dump_arg *dump = w->dump;
5858 struct rt6_exception_bucket *bucket;
5859 struct rt6_exception *rt6_ex;
5860 int i, err;
5861
5862 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5863 if (!bucket)
5864 return 0;
5865
5866 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5867 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5868 if (w->skip) {
5869 w->skip--;
5870 continue;
5871 }
5872
5873 /* Expiration of entries doesn't bump sernum, insertion
5874 * does. Removal is triggered by insertion, so we can
5875 * rely on the fact that if entries change between two
5876 * partial dumps, this node is scanned again completely,
5877 * see rt6_insert_exception() and fib6_dump_table().
5878 *
5879 * Count expired entries we go through as handled
5880 * entries that we'll skip next time, in case of partial
5881 * node dump. Otherwise, if entries expire meanwhile,
5882 * we'll skip the wrong amount.
5883 */
5884 if (rt6_check_expired(rt6_ex->rt6i)) {
5885 w->count++;
5886 continue;
5887 }
5888
5889 err = rt6_fill_node(dump->net, dump->skb, w->rt,
5890 &rt6_ex->rt6i->dst, NULL, NULL, 0,
5891 RTM_NEWROUTE,
5892 NETLINK_CB(dump->cb->skb).portid,
5893 dump->cb->nlh->nlmsg_seq, w->flags);
5894 if (err)
5895 return err;
5896
5897 w->count++;
5898 }
5899 bucket++;
5900 }
5901
5902 return 0;
5903 }
5904
5905 /* Return -1 if done with node, number of handled routes on partial dump */
rt6_dump_route(struct fib6_info * rt,void * p_arg,unsigned int skip)5906 int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5907 {
5908 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5909 struct fib_dump_filter *filter = &arg->filter;
5910 unsigned int flags = NLM_F_MULTI;
5911 struct net *net = arg->net;
5912 int count = 0;
5913
5914 if (rt == net->ipv6.fib6_null_entry)
5915 return -1;
5916
5917 if ((filter->flags & RTM_F_PREFIX) &&
5918 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5919 /* success since this is not a prefix route */
5920 return -1;
5921 }
5922 if (filter->filter_set &&
5923 ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5924 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5925 (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5926 return -1;
5927 }
5928
5929 if (filter->filter_set ||
5930 !filter->dump_routes || !filter->dump_exceptions) {
5931 flags |= NLM_F_DUMP_FILTERED;
5932 }
5933
5934 if (filter->dump_routes) {
5935 if (skip) {
5936 skip--;
5937 } else {
5938 if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5939 0, RTM_NEWROUTE,
5940 NETLINK_CB(arg->cb->skb).portid,
5941 arg->cb->nlh->nlmsg_seq, flags)) {
5942 return 0;
5943 }
5944 count++;
5945 }
5946 }
5947
5948 if (filter->dump_exceptions) {
5949 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5950 .rt = rt,
5951 .flags = flags,
5952 .skip = skip,
5953 .count = 0 };
5954 int err;
5955
5956 rcu_read_lock();
5957 if (rt->nh) {
5958 err = nexthop_for_each_fib6_nh(rt->nh,
5959 rt6_nh_dump_exceptions,
5960 &w);
5961 } else {
5962 err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5963 }
5964 rcu_read_unlock();
5965
5966 if (err)
5967 return count + w.count;
5968 }
5969
5970 return -1;
5971 }
5972
inet6_rtm_valid_getroute_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)5973 static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5974 const struct nlmsghdr *nlh,
5975 struct nlattr **tb,
5976 struct netlink_ext_ack *extack)
5977 {
5978 struct rtmsg *rtm;
5979 int i, err;
5980
5981 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5982 NL_SET_ERR_MSG_MOD(extack,
5983 "Invalid header for get route request");
5984 return -EINVAL;
5985 }
5986
5987 if (!netlink_strict_get_check(skb))
5988 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5989 rtm_ipv6_policy, extack);
5990
5991 rtm = nlmsg_data(nlh);
5992 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5993 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5994 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5995 rtm->rtm_type) {
5996 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5997 return -EINVAL;
5998 }
5999 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
6000 NL_SET_ERR_MSG_MOD(extack,
6001 "Invalid flags for get route request");
6002 return -EINVAL;
6003 }
6004
6005 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
6006 rtm_ipv6_policy, extack);
6007 if (err)
6008 return err;
6009
6010 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
6011 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
6012 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
6013 return -EINVAL;
6014 }
6015
6016 for (i = 0; i <= RTA_MAX; i++) {
6017 if (!tb[i])
6018 continue;
6019
6020 switch (i) {
6021 case RTA_SRC:
6022 case RTA_DST:
6023 case RTA_IIF:
6024 case RTA_OIF:
6025 case RTA_MARK:
6026 case RTA_UID:
6027 case RTA_SPORT:
6028 case RTA_DPORT:
6029 case RTA_IP_PROTO:
6030 break;
6031 default:
6032 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
6033 return -EINVAL;
6034 }
6035 }
6036
6037 return 0;
6038 }
6039
inet6_rtm_getroute(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)6040 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
6041 struct netlink_ext_ack *extack)
6042 {
6043 struct net *net = sock_net(in_skb->sk);
6044 struct nlattr *tb[RTA_MAX+1];
6045 int err, iif = 0, oif = 0;
6046 struct fib6_info *from;
6047 struct dst_entry *dst;
6048 struct rt6_info *rt;
6049 struct sk_buff *skb;
6050 struct rtmsg *rtm;
6051 struct flowi6 fl6 = {};
6052 bool fibmatch;
6053
6054 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
6055 if (err < 0)
6056 goto errout;
6057
6058 err = -EINVAL;
6059 rtm = nlmsg_data(nlh);
6060 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
6061 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
6062
6063 if (tb[RTA_SRC]) {
6064 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
6065 goto errout;
6066
6067 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
6068 }
6069
6070 if (tb[RTA_DST]) {
6071 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
6072 goto errout;
6073
6074 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
6075 }
6076
6077 if (tb[RTA_IIF])
6078 iif = nla_get_u32(tb[RTA_IIF]);
6079
6080 if (tb[RTA_OIF])
6081 oif = nla_get_u32(tb[RTA_OIF]);
6082
6083 if (tb[RTA_MARK])
6084 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
6085
6086 if (tb[RTA_UID])
6087 fl6.flowi6_uid = make_kuid(current_user_ns(),
6088 nla_get_u32(tb[RTA_UID]));
6089 else
6090 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
6091
6092 if (tb[RTA_SPORT])
6093 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
6094
6095 if (tb[RTA_DPORT])
6096 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
6097
6098 if (tb[RTA_IP_PROTO]) {
6099 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
6100 &fl6.flowi6_proto, AF_INET6,
6101 extack);
6102 if (err)
6103 goto errout;
6104 }
6105
6106 if (iif) {
6107 struct net_device *dev;
6108 int flags = 0;
6109
6110 rcu_read_lock();
6111
6112 dev = dev_get_by_index_rcu(net, iif);
6113 if (!dev) {
6114 rcu_read_unlock();
6115 err = -ENODEV;
6116 goto errout;
6117 }
6118
6119 fl6.flowi6_iif = iif;
6120
6121 if (!ipv6_addr_any(&fl6.saddr))
6122 flags |= RT6_LOOKUP_F_HAS_SADDR;
6123
6124 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
6125
6126 rcu_read_unlock();
6127 } else {
6128 fl6.flowi6_oif = oif;
6129
6130 dst = ip6_route_output(net, NULL, &fl6);
6131 }
6132
6133
6134 rt = dst_rt6_info(dst);
6135 if (rt->dst.error) {
6136 err = rt->dst.error;
6137 ip6_rt_put(rt);
6138 goto errout;
6139 }
6140
6141 if (rt == net->ipv6.ip6_null_entry) {
6142 err = rt->dst.error;
6143 ip6_rt_put(rt);
6144 goto errout;
6145 }
6146
6147 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
6148 if (!skb) {
6149 ip6_rt_put(rt);
6150 err = -ENOBUFS;
6151 goto errout;
6152 }
6153
6154 skb_dst_set(skb, &rt->dst);
6155
6156 rcu_read_lock();
6157 from = rcu_dereference(rt->from);
6158 if (from) {
6159 if (fibmatch)
6160 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
6161 iif, RTM_NEWROUTE,
6162 NETLINK_CB(in_skb).portid,
6163 nlh->nlmsg_seq, 0);
6164 else
6165 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
6166 &fl6.saddr, iif, RTM_NEWROUTE,
6167 NETLINK_CB(in_skb).portid,
6168 nlh->nlmsg_seq, 0);
6169 } else {
6170 err = -ENETUNREACH;
6171 }
6172 rcu_read_unlock();
6173
6174 if (err < 0) {
6175 kfree_skb(skb);
6176 goto errout;
6177 }
6178
6179 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
6180 errout:
6181 return err;
6182 }
6183
inet6_rt_notify(int event,struct fib6_info * rt,struct nl_info * info,unsigned int nlm_flags)6184 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
6185 unsigned int nlm_flags)
6186 {
6187 struct sk_buff *skb;
6188 struct net *net = info->nl_net;
6189 u32 seq;
6190 int err;
6191
6192 err = -ENOBUFS;
6193 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6194
6195 skb = nlmsg_new(rt6_nlmsg_size(rt), GFP_ATOMIC);
6196 if (!skb)
6197 goto errout;
6198
6199 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6200 event, info->portid, seq, nlm_flags);
6201 if (err < 0) {
6202 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6203 WARN_ON(err == -EMSGSIZE);
6204 kfree_skb(skb);
6205 goto errout;
6206 }
6207 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6208 info->nlh, GFP_ATOMIC);
6209 return;
6210 errout:
6211 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6212 }
6213
fib6_rt_update(struct net * net,struct fib6_info * rt,struct nl_info * info)6214 void fib6_rt_update(struct net *net, struct fib6_info *rt,
6215 struct nl_info *info)
6216 {
6217 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6218 struct sk_buff *skb;
6219 int err = -ENOBUFS;
6220
6221 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6222 if (!skb)
6223 goto errout;
6224
6225 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6226 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6227 if (err < 0) {
6228 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6229 WARN_ON(err == -EMSGSIZE);
6230 kfree_skb(skb);
6231 goto errout;
6232 }
6233 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6234 info->nlh, gfp_any());
6235 return;
6236 errout:
6237 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6238 }
6239
fib6_info_hw_flags_set(struct net * net,struct fib6_info * f6i,bool offload,bool trap,bool offload_failed)6240 void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
6241 bool offload, bool trap, bool offload_failed)
6242 {
6243 struct sk_buff *skb;
6244 int err;
6245
6246 if (READ_ONCE(f6i->offload) == offload &&
6247 READ_ONCE(f6i->trap) == trap &&
6248 READ_ONCE(f6i->offload_failed) == offload_failed)
6249 return;
6250
6251 WRITE_ONCE(f6i->offload, offload);
6252 WRITE_ONCE(f6i->trap, trap);
6253
6254 /* 2 means send notifications only if offload_failed was changed. */
6255 if (net->ipv6.sysctl.fib_notify_on_flag_change == 2 &&
6256 READ_ONCE(f6i->offload_failed) == offload_failed)
6257 return;
6258
6259 WRITE_ONCE(f6i->offload_failed, offload_failed);
6260
6261 if (!rcu_access_pointer(f6i->fib6_node))
6262 /* The route was removed from the tree, do not send
6263 * notification.
6264 */
6265 return;
6266
6267 if (!net->ipv6.sysctl.fib_notify_on_flag_change)
6268 return;
6269
6270 skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
6271 if (!skb) {
6272 err = -ENOBUFS;
6273 goto errout;
6274 }
6275
6276 err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
6277 0, 0);
6278 if (err < 0) {
6279 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6280 WARN_ON(err == -EMSGSIZE);
6281 kfree_skb(skb);
6282 goto errout;
6283 }
6284
6285 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
6286 return;
6287
6288 errout:
6289 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6290 }
6291 EXPORT_SYMBOL(fib6_info_hw_flags_set);
6292
ip6_route_dev_notify(struct notifier_block * this,unsigned long event,void * ptr)6293 static int ip6_route_dev_notify(struct notifier_block *this,
6294 unsigned long event, void *ptr)
6295 {
6296 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6297 struct net *net = dev_net(dev);
6298
6299 if (!(dev->flags & IFF_LOOPBACK))
6300 return NOTIFY_OK;
6301
6302 if (event == NETDEV_REGISTER) {
6303 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
6304 net->ipv6.ip6_null_entry->dst.dev = dev;
6305 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6306 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6307 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
6308 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
6309 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
6310 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
6311 #endif
6312 } else if (event == NETDEV_UNREGISTER &&
6313 dev->reg_state != NETREG_UNREGISTERED) {
6314 /* NETDEV_UNREGISTER could be fired for multiple times by
6315 * netdev_wait_allrefs(). Make sure we only call this once.
6316 */
6317 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
6318 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6319 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6320 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
6321 #endif
6322 }
6323
6324 return NOTIFY_OK;
6325 }
6326
6327 /*
6328 * /proc
6329 */
6330
6331 #ifdef CONFIG_PROC_FS
rt6_stats_seq_show(struct seq_file * seq,void * v)6332 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6333 {
6334 struct net *net = (struct net *)seq->private;
6335 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6336 net->ipv6.rt6_stats->fib_nodes,
6337 net->ipv6.rt6_stats->fib_route_nodes,
6338 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6339 net->ipv6.rt6_stats->fib_rt_entries,
6340 net->ipv6.rt6_stats->fib_rt_cache,
6341 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6342 net->ipv6.rt6_stats->fib_discarded_routes);
6343
6344 return 0;
6345 }
6346 #endif /* CONFIG_PROC_FS */
6347
6348 #ifdef CONFIG_SYSCTL
6349
ipv6_sysctl_rtcache_flush(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6350 static int ipv6_sysctl_rtcache_flush(const struct ctl_table *ctl, int write,
6351 void *buffer, size_t *lenp, loff_t *ppos)
6352 {
6353 struct net *net;
6354 int delay;
6355 int ret;
6356 if (!write)
6357 return -EINVAL;
6358
6359 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6360 if (ret)
6361 return ret;
6362
6363 net = (struct net *)ctl->extra1;
6364 delay = net->ipv6.sysctl.flush_delay;
6365 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6366 return 0;
6367 }
6368
6369 static struct ctl_table ipv6_route_table_template[] = {
6370 {
6371 .procname = "max_size",
6372 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
6373 .maxlen = sizeof(int),
6374 .mode = 0644,
6375 .proc_handler = proc_dointvec,
6376 },
6377 {
6378 .procname = "gc_thresh",
6379 .data = &ip6_dst_ops_template.gc_thresh,
6380 .maxlen = sizeof(int),
6381 .mode = 0644,
6382 .proc_handler = proc_dointvec,
6383 },
6384 {
6385 .procname = "flush",
6386 .data = &init_net.ipv6.sysctl.flush_delay,
6387 .maxlen = sizeof(int),
6388 .mode = 0200,
6389 .proc_handler = ipv6_sysctl_rtcache_flush
6390 },
6391 {
6392 .procname = "gc_min_interval",
6393 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6394 .maxlen = sizeof(int),
6395 .mode = 0644,
6396 .proc_handler = proc_dointvec_jiffies,
6397 },
6398 {
6399 .procname = "gc_timeout",
6400 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6401 .maxlen = sizeof(int),
6402 .mode = 0644,
6403 .proc_handler = proc_dointvec_jiffies,
6404 },
6405 {
6406 .procname = "gc_interval",
6407 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
6408 .maxlen = sizeof(int),
6409 .mode = 0644,
6410 .proc_handler = proc_dointvec_jiffies,
6411 },
6412 {
6413 .procname = "gc_elasticity",
6414 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6415 .maxlen = sizeof(int),
6416 .mode = 0644,
6417 .proc_handler = proc_dointvec,
6418 },
6419 {
6420 .procname = "mtu_expires",
6421 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6422 .maxlen = sizeof(int),
6423 .mode = 0644,
6424 .proc_handler = proc_dointvec_jiffies,
6425 },
6426 {
6427 .procname = "min_adv_mss",
6428 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
6429 .maxlen = sizeof(int),
6430 .mode = 0644,
6431 .proc_handler = proc_dointvec,
6432 },
6433 {
6434 .procname = "gc_min_interval_ms",
6435 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6436 .maxlen = sizeof(int),
6437 .mode = 0644,
6438 .proc_handler = proc_dointvec_ms_jiffies,
6439 },
6440 {
6441 .procname = "skip_notify_on_dev_down",
6442 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6443 .maxlen = sizeof(u8),
6444 .mode = 0644,
6445 .proc_handler = proc_dou8vec_minmax,
6446 .extra1 = SYSCTL_ZERO,
6447 .extra2 = SYSCTL_ONE,
6448 },
6449 };
6450
ipv6_route_sysctl_init(struct net * net)6451 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6452 {
6453 struct ctl_table *table;
6454
6455 table = kmemdup(ipv6_route_table_template,
6456 sizeof(ipv6_route_table_template),
6457 GFP_KERNEL);
6458
6459 if (table) {
6460 table[0].data = &net->ipv6.sysctl.ip6_rt_max_size;
6461 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6462 table[2].data = &net->ipv6.sysctl.flush_delay;
6463 table[2].extra1 = net;
6464 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6465 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6466 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6467 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6468 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6469 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6470 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6471 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6472 }
6473
6474 return table;
6475 }
6476
ipv6_route_sysctl_table_size(struct net * net)6477 size_t ipv6_route_sysctl_table_size(struct net *net)
6478 {
6479 /* Don't export sysctls to unprivileged users */
6480 if (net->user_ns != &init_user_ns)
6481 return 1;
6482
6483 return ARRAY_SIZE(ipv6_route_table_template);
6484 }
6485 #endif
6486
ip6_route_net_init(struct net * net)6487 static int __net_init ip6_route_net_init(struct net *net)
6488 {
6489 int ret = -ENOMEM;
6490
6491 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6492 sizeof(net->ipv6.ip6_dst_ops));
6493
6494 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6495 goto out_ip6_dst_ops;
6496
6497 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6498 if (!net->ipv6.fib6_null_entry)
6499 goto out_ip6_dst_entries;
6500 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6501 sizeof(*net->ipv6.fib6_null_entry));
6502
6503 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6504 sizeof(*net->ipv6.ip6_null_entry),
6505 GFP_KERNEL);
6506 if (!net->ipv6.ip6_null_entry)
6507 goto out_fib6_null_entry;
6508 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6509 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6510 ip6_template_metrics, true);
6511 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->dst.rt_uncached);
6512
6513 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6514 net->ipv6.fib6_has_custom_rules = false;
6515 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6516 sizeof(*net->ipv6.ip6_prohibit_entry),
6517 GFP_KERNEL);
6518 if (!net->ipv6.ip6_prohibit_entry)
6519 goto out_ip6_null_entry;
6520 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6521 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6522 ip6_template_metrics, true);
6523 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->dst.rt_uncached);
6524
6525 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6526 sizeof(*net->ipv6.ip6_blk_hole_entry),
6527 GFP_KERNEL);
6528 if (!net->ipv6.ip6_blk_hole_entry)
6529 goto out_ip6_prohibit_entry;
6530 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6531 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6532 ip6_template_metrics, true);
6533 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->dst.rt_uncached);
6534 #ifdef CONFIG_IPV6_SUBTREES
6535 net->ipv6.fib6_routes_require_src = 0;
6536 #endif
6537 #endif
6538
6539 net->ipv6.sysctl.flush_delay = 0;
6540 net->ipv6.sysctl.ip6_rt_max_size = INT_MAX;
6541 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6542 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6543 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6544 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6545 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6546 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6547 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6548
6549 atomic_set(&net->ipv6.ip6_rt_gc_expire, 30*HZ);
6550
6551 ret = 0;
6552 out:
6553 return ret;
6554
6555 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6556 out_ip6_prohibit_entry:
6557 kfree(net->ipv6.ip6_prohibit_entry);
6558 out_ip6_null_entry:
6559 kfree(net->ipv6.ip6_null_entry);
6560 #endif
6561 out_fib6_null_entry:
6562 kfree(net->ipv6.fib6_null_entry);
6563 out_ip6_dst_entries:
6564 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6565 out_ip6_dst_ops:
6566 goto out;
6567 }
6568
ip6_route_net_exit(struct net * net)6569 static void __net_exit ip6_route_net_exit(struct net *net)
6570 {
6571 kfree(net->ipv6.fib6_null_entry);
6572 kfree(net->ipv6.ip6_null_entry);
6573 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6574 kfree(net->ipv6.ip6_prohibit_entry);
6575 kfree(net->ipv6.ip6_blk_hole_entry);
6576 #endif
6577 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6578 }
6579
ip6_route_net_init_late(struct net * net)6580 static int __net_init ip6_route_net_init_late(struct net *net)
6581 {
6582 #ifdef CONFIG_PROC_FS
6583 if (!proc_create_net("ipv6_route", 0, net->proc_net,
6584 &ipv6_route_seq_ops,
6585 sizeof(struct ipv6_route_iter)))
6586 return -ENOMEM;
6587
6588 if (!proc_create_net_single("rt6_stats", 0444, net->proc_net,
6589 rt6_stats_seq_show, NULL)) {
6590 remove_proc_entry("ipv6_route", net->proc_net);
6591 return -ENOMEM;
6592 }
6593 #endif
6594 return 0;
6595 }
6596
ip6_route_net_exit_late(struct net * net)6597 static void __net_exit ip6_route_net_exit_late(struct net *net)
6598 {
6599 #ifdef CONFIG_PROC_FS
6600 remove_proc_entry("ipv6_route", net->proc_net);
6601 remove_proc_entry("rt6_stats", net->proc_net);
6602 #endif
6603 }
6604
6605 static struct pernet_operations ip6_route_net_ops = {
6606 .init = ip6_route_net_init,
6607 .exit = ip6_route_net_exit,
6608 };
6609
ipv6_inetpeer_init(struct net * net)6610 static int __net_init ipv6_inetpeer_init(struct net *net)
6611 {
6612 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6613
6614 if (!bp)
6615 return -ENOMEM;
6616 inet_peer_base_init(bp);
6617 net->ipv6.peers = bp;
6618 return 0;
6619 }
6620
ipv6_inetpeer_exit(struct net * net)6621 static void __net_exit ipv6_inetpeer_exit(struct net *net)
6622 {
6623 struct inet_peer_base *bp = net->ipv6.peers;
6624
6625 net->ipv6.peers = NULL;
6626 inetpeer_invalidate_tree(bp);
6627 kfree(bp);
6628 }
6629
6630 static struct pernet_operations ipv6_inetpeer_ops = {
6631 .init = ipv6_inetpeer_init,
6632 .exit = ipv6_inetpeer_exit,
6633 };
6634
6635 static struct pernet_operations ip6_route_net_late_ops = {
6636 .init = ip6_route_net_init_late,
6637 .exit = ip6_route_net_exit_late,
6638 };
6639
6640 static struct notifier_block ip6_route_dev_notifier = {
6641 .notifier_call = ip6_route_dev_notify,
6642 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6643 };
6644
ip6_route_init_special_entries(void)6645 void __init ip6_route_init_special_entries(void)
6646 {
6647 /* Registering of the loopback is done before this portion of code,
6648 * the loopback reference in rt6_info will not be taken, do it
6649 * manually for init_net */
6650 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6651 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6652 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6653 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6654 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6655 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6656 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6657 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6658 #endif
6659 }
6660
6661 #if IS_BUILTIN(CONFIG_IPV6)
6662 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6663 DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6664
6665 BTF_ID_LIST(btf_fib6_info_id)
6666 BTF_ID(struct, fib6_info)
6667
6668 static const struct bpf_iter_seq_info ipv6_route_seq_info = {
6669 .seq_ops = &ipv6_route_seq_ops,
6670 .init_seq_private = bpf_iter_init_seq_net,
6671 .fini_seq_private = bpf_iter_fini_seq_net,
6672 .seq_priv_size = sizeof(struct ipv6_route_iter),
6673 };
6674
6675 static struct bpf_iter_reg ipv6_route_reg_info = {
6676 .target = "ipv6_route",
6677 .ctx_arg_info_size = 1,
6678 .ctx_arg_info = {
6679 { offsetof(struct bpf_iter__ipv6_route, rt),
6680 PTR_TO_BTF_ID_OR_NULL },
6681 },
6682 .seq_info = &ipv6_route_seq_info,
6683 };
6684
bpf_iter_register(void)6685 static int __init bpf_iter_register(void)
6686 {
6687 ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id;
6688 return bpf_iter_reg_target(&ipv6_route_reg_info);
6689 }
6690
bpf_iter_unregister(void)6691 static void bpf_iter_unregister(void)
6692 {
6693 bpf_iter_unreg_target(&ipv6_route_reg_info);
6694 }
6695 #endif
6696 #endif
6697
6698 static const struct rtnl_msg_handler ip6_route_rtnl_msg_handlers[] __initconst_or_module = {
6699 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_NEWROUTE,
6700 .doit = inet6_rtm_newroute},
6701 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_DELROUTE,
6702 .doit = inet6_rtm_delroute},
6703 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETROUTE,
6704 .doit = inet6_rtm_getroute, .flags = RTNL_FLAG_DOIT_UNLOCKED},
6705 };
6706
ip6_route_init(void)6707 int __init ip6_route_init(void)
6708 {
6709 int ret;
6710 int cpu;
6711
6712 ret = -ENOMEM;
6713 ip6_dst_ops_template.kmem_cachep =
6714 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6715 SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT, NULL);
6716 if (!ip6_dst_ops_template.kmem_cachep)
6717 goto out;
6718
6719 ret = dst_entries_init(&ip6_dst_blackhole_ops);
6720 if (ret)
6721 goto out_kmem_cache;
6722
6723 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6724 if (ret)
6725 goto out_dst_entries;
6726
6727 ret = register_pernet_subsys(&ip6_route_net_ops);
6728 if (ret)
6729 goto out_register_inetpeer;
6730
6731 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6732
6733 ret = fib6_init();
6734 if (ret)
6735 goto out_register_subsys;
6736
6737 ret = xfrm6_init();
6738 if (ret)
6739 goto out_fib6_init;
6740
6741 ret = fib6_rules_init();
6742 if (ret)
6743 goto xfrm6_init;
6744
6745 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6746 if (ret)
6747 goto fib6_rules_init;
6748
6749 ret = rtnl_register_many(ip6_route_rtnl_msg_handlers);
6750 if (ret < 0)
6751 goto out_register_late_subsys;
6752
6753 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6754 if (ret)
6755 goto out_register_late_subsys;
6756
6757 #if IS_BUILTIN(CONFIG_IPV6)
6758 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6759 ret = bpf_iter_register();
6760 if (ret)
6761 goto out_register_late_subsys;
6762 #endif
6763 #endif
6764
6765 for_each_possible_cpu(cpu) {
6766 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6767
6768 INIT_LIST_HEAD(&ul->head);
6769 spin_lock_init(&ul->lock);
6770 }
6771
6772 out:
6773 return ret;
6774
6775 out_register_late_subsys:
6776 rtnl_unregister_all(PF_INET6);
6777 unregister_pernet_subsys(&ip6_route_net_late_ops);
6778 fib6_rules_init:
6779 fib6_rules_cleanup();
6780 xfrm6_init:
6781 xfrm6_fini();
6782 out_fib6_init:
6783 fib6_gc_cleanup();
6784 out_register_subsys:
6785 unregister_pernet_subsys(&ip6_route_net_ops);
6786 out_register_inetpeer:
6787 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6788 out_dst_entries:
6789 dst_entries_destroy(&ip6_dst_blackhole_ops);
6790 out_kmem_cache:
6791 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6792 goto out;
6793 }
6794
ip6_route_cleanup(void)6795 void ip6_route_cleanup(void)
6796 {
6797 #if IS_BUILTIN(CONFIG_IPV6)
6798 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6799 bpf_iter_unregister();
6800 #endif
6801 #endif
6802 unregister_netdevice_notifier(&ip6_route_dev_notifier);
6803 unregister_pernet_subsys(&ip6_route_net_late_ops);
6804 fib6_rules_cleanup();
6805 xfrm6_fini();
6806 fib6_gc_cleanup();
6807 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6808 unregister_pernet_subsys(&ip6_route_net_ops);
6809 dst_entries_destroy(&ip6_dst_blackhole_ops);
6810 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6811 }
6812