projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net: add EXPORT_INDIRECT_CALLABLE wrapper
[linux-block.git] / net / ipv6 / route.c
CommitLineData
2874c5fd 1// SPDX-License-Identifier: GPL-2.0-or-later
1da177e4
LT		 2/*
3 * Linux INET6 implementation
4 * FIB front-end.
5 *
6 * Authors:
1ab1457c 7 * Pedro Roque <roque@di.fc.ul.pt>
1da177e4
LT		 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.
c0bece9f
YH		 19 * Ville Nuorvala
20 * Fixed routing subtrees.
1da177e4
LT		 21 */
22
f3213831
JP		 23#define pr_fmt(fmt) "IPv6: " fmt
24
4fc268d2 25#include <linux/capability.h>
1da177e4 26#include <linux/errno.h>
bc3b2d7f 27#include <linux/export.h>
1da177e4
LT		 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>
7bc570c8 36#include <linux/mroute6.h>
1da177e4 37#include <linux/init.h>
1da177e4 38#include <linux/if_arp.h>
1da177e4
LT		 39#include <linux/proc_fs.h>
40#include <linux/seq_file.h>
5b7c931d 41#include <linux/nsproxy.h>
5a0e3ad6 42#include <linux/slab.h>
35732d01 43#include <linux/jhash.h>
457c4cbc 44#include <net/net_namespace.h>
1da177e4
LT		 45#include <net/snmp.h>
46#include <net/ipv6.h>
47#include <net/ip6_fib.h>
48#include <net/ip6_route.h>
49#include <net/ndisc.h>
50#include <net/addrconf.h>
51#include <net/tcp.h>
52#include <linux/rtnetlink.h>
53#include <net/dst.h>
904af04d 54#include <net/dst_metadata.h>
1da177e4 55#include <net/xfrm.h>
8d71740c 56#include <net/netevent.h>
21713ebc 57#include <net/netlink.h>
3c618c1d 58#include <net/rtnh.h>
19e42e45 59#include <net/lwtunnel.h>
904af04d 60#include <net/ip_tunnels.h>
ca254490 61#include <net/l3mdev.h>
eacb9384 62#include <net/ip.h>
7c0f6ba6 63#include <linux/uaccess.h>
951cf368 64#include <linux/btf_ids.h>
1da177e4
LT		 65
66#ifdef CONFIG_SYSCTL
67#include <linux/sysctl.h>
68#endif
69
30d444d3
DA		 70static int ip6_rt_type_to_error(u8 fib6_type);
71
72#define CREATE_TRACE_POINTS
73#include <trace/events/fib6.h>
74EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
75#undef CREATE_TRACE_POINTS
76
afc154e9 77enum rt6_nud_state {
7e980569
JB		 78 RT6_NUD_FAIL_HARD = -3,
79 RT6_NUD_FAIL_PROBE = -2,
80 RT6_NUD_FAIL_DO_RR = -1,
afc154e9
HFS		 81 RT6_NUD_SUCCEED = 1
82};
83
bbd807df
BV		 84INDIRECT_CALLABLE_SCOPE
85struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
0dbaee3b 86static unsigned int ip6_default_advmss(const struct dst_entry *dst);
f67fbeae
BV		 87INDIRECT_CALLABLE_SCOPE
88unsigned int ip6_mtu(const struct dst_entry *dst);
1da177e4
LT		 89static struct dst_entry *ip6_negative_advice(struct dst_entry *);
90static void ip6_dst_destroy(struct dst_entry *);
91static void ip6_dst_ifdown(struct dst_entry *,
92 struct net_device *dev, int how);
569d3645 93static int ip6_dst_gc(struct dst_ops *ops);
1da177e4
LT		 94
95static int ip6_pkt_discard(struct sk_buff *skb);
ede2059d 96static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
7150aede 97static int ip6_pkt_prohibit(struct sk_buff *skb);
ede2059d 98static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1da177e4 99static void ip6_link_failure(struct sk_buff *skb);
6700c270 100static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
bd085ef6
HL		 101 struct sk_buff *skb, u32 mtu,
102 bool confirm_neigh);
6700c270
DM		 103static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
104 struct sk_buff *skb);
702cea56
DA		 105static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
106 int strict);
a1b7a1f0 107static size_t rt6_nlmsg_size(struct fib6_info *f6i);
d4ead6b3 108static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 109 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 110 struct in6_addr *dest, struct in6_addr *src,
16a16cd3
DA		 111 int iif, int type, u32 portid, u32 seq,
112 unsigned int flags);
7e4b5128 113static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
510e2ced
WW		 114 const struct in6_addr *daddr,
115 const struct in6_addr *saddr);
1da177e4 116
70ceb4f5 117#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 118static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 119 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 120 const struct in6_addr *gwaddr,
121 struct net_device *dev,
95c96174 122 unsigned int pref);
8d1c802b 123static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 124 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 125 const struct in6_addr *gwaddr,
126 struct net_device *dev);
70ceb4f5
YH		 127#endif
128
8d0b94af
MKL		 129struct uncached_list {
130 spinlock_t lock;
131 struct list_head head;
132};
133
134static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
135
510c321b 136void rt6_uncached_list_add(struct rt6_info *rt)
8d0b94af
MKL		 137{
138 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
139
8d0b94af
MKL		 140 rt->rt6i_uncached_list = ul;
141
142 spin_lock_bh(&ul->lock);
143 list_add_tail(&rt->rt6i_uncached, &ul->head);
144 spin_unlock_bh(&ul->lock);
145}
146
510c321b 147void rt6_uncached_list_del(struct rt6_info *rt)
8d0b94af
MKL		 148{
149 if (!list_empty(&rt->rt6i_uncached)) {
150 struct uncached_list *ul = rt->rt6i_uncached_list;
81eb8447 151 struct net *net = dev_net(rt->dst.dev);
8d0b94af
MKL		 152
153 spin_lock_bh(&ul->lock);
154 list_del(&rt->rt6i_uncached);
81eb8447 155 atomic_dec(&net->ipv6.rt6_stats->fib_rt_uncache);
8d0b94af
MKL		 156 spin_unlock_bh(&ul->lock);
157 }
158}
159
160static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
161{
162 struct net_device *loopback_dev = net->loopback_dev;
163 int cpu;
164
e332bc67
EB		 165 if (dev == loopback_dev)
166 return;
167
8d0b94af
MKL		 168 for_each_possible_cpu(cpu) {
169 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
170 struct rt6_info *rt;
171
172 spin_lock_bh(&ul->lock);
173 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
174 struct inet6_dev *rt_idev = rt->rt6i_idev;
175 struct net_device *rt_dev = rt->dst.dev;
176
e332bc67 177 if (rt_idev->dev == dev) {
8d0b94af
MKL		 178 rt->rt6i_idev = in6_dev_get(loopback_dev);
179 in6_dev_put(rt_idev);
180 }
181
e332bc67 182 if (rt_dev == dev) {
8d7017fd 183 rt->dst.dev = blackhole_netdev;
8d0b94af
MKL		 184 dev_hold(rt->dst.dev);
185 dev_put(rt_dev);
186 }
187 }
188 spin_unlock_bh(&ul->lock);
189 }
190}
191
f8a1b43b 192static inline const void *choose_neigh_daddr(const struct in6_addr *p,
f894cbf8
DM		 193 struct sk_buff *skb,
194 const void *daddr)
39232973 195{
a7563f34 196 if (!ipv6_addr_any(p))
39232973 197 return (const void *) p;
f894cbf8
DM		 198 else if (skb)
199 return &ipv6_hdr(skb)->daddr;
39232973
DM		 200 return daddr;
201}
202
f8a1b43b
DA		 203struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
204 struct net_device *dev,
205 struct sk_buff *skb,
206 const void *daddr)
d3aaeb38 207{
39232973
DM		 208 struct neighbour *n;
209
f8a1b43b
DA		 210 daddr = choose_neigh_daddr(gw, skb, daddr);
211 n = __ipv6_neigh_lookup(dev, daddr);
f83c7790
DM		 212 if (n)
213 return n;
7adf3246
SB		 214
215 n = neigh_create(&nd_tbl, daddr, dev);
216 return IS_ERR(n) ? NULL : n;
f8a1b43b
DA		 217}
218
219static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
220 struct sk_buff *skb,
221 const void *daddr)
222{
223 const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
224
2c6b55f4
ND		 225 return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
226 dst->dev, skb, daddr);
f83c7790
DM		 227}
228
63fca65d
JA		 229static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
230{
231 struct net_device *dev = dst->dev;
232 struct rt6_info *rt = (struct rt6_info *)dst;
233
cbfd6891 234 daddr = choose_neigh_daddr(rt6_nexthop(rt, &in6addr_any), NULL, daddr);
63fca65d
JA		 235 if (!daddr)
236 return;
237 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
238 return;
239 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
240 return;
241 __ipv6_confirm_neigh(dev, daddr);
242}
243
9a7ec3a9 244static struct dst_ops ip6_dst_ops_template = {
1da177e4 245 .family = AF_INET6,
1da177e4
LT		 246 .gc = ip6_dst_gc,
247 .gc_thresh = 1024,
248 .check = ip6_dst_check,
0dbaee3b 249 .default_advmss = ip6_default_advmss,
ebb762f2 250 .mtu = ip6_mtu,
d4ead6b3 251 .cow_metrics = dst_cow_metrics_generic,
1da177e4
LT		 252 .destroy = ip6_dst_destroy,
253 .ifdown = ip6_dst_ifdown,
254 .negative_advice = ip6_negative_advice,
255 .link_failure = ip6_link_failure,
256 .update_pmtu = ip6_rt_update_pmtu,
6e157b6a 257 .redirect = rt6_do_redirect,
9f8955cc 258 .local_out = __ip6_local_out,
f8a1b43b 259 .neigh_lookup = ip6_dst_neigh_lookup,
63fca65d 260 .confirm_neigh = ip6_confirm_neigh,
1da177e4
LT		 261};
262
ebb762f2 263static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
ec831ea7 264{
618f9bc7
SK		 265 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
266
267 return mtu ? : dst->dev->mtu;
ec831ea7
RD		 268}
269
6700c270 270static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
bd085ef6
HL		 271 struct sk_buff *skb, u32 mtu,
272 bool confirm_neigh)
14e50e57
DM		 273{
274}
275
6700c270
DM		 276static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
277 struct sk_buff *skb)
b587ee3b
DM		 278{
279}
280
14e50e57
DM		 281static struct dst_ops ip6_dst_blackhole_ops = {
282 .family = AF_INET6,
14e50e57
DM		 283 .destroy = ip6_dst_destroy,
284 .check = ip6_dst_check,
ebb762f2 285 .mtu = ip6_blackhole_mtu,
214f45c9 286 .default_advmss = ip6_default_advmss,
14e50e57 287 .update_pmtu = ip6_rt_blackhole_update_pmtu,
b587ee3b 288 .redirect = ip6_rt_blackhole_redirect,
0a1f5962 289 .cow_metrics = dst_cow_metrics_generic,
f8a1b43b 290 .neigh_lookup = ip6_dst_neigh_lookup,
14e50e57
DM		 291};
292
62fa8a84 293static const u32 ip6_template_metrics[RTAX_MAX] = {
14edd87d 294 [RTAX_HOPLIMIT - 1] = 0,
62fa8a84
DM		 295};
296
8d1c802b 297static const struct fib6_info fib6_null_entry_template = {
93c2fb25
DA		 298 .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
299 .fib6_protocol = RTPROT_KERNEL,
300 .fib6_metric = ~(u32)0,
f05713e0 301 .fib6_ref = REFCOUNT_INIT(1),
421842ed
DA		 302 .fib6_type = RTN_UNREACHABLE,
303 .fib6_metrics = (struct dst_metrics *)&dst_default_metrics,
304};
305
fb0af4c7 306static const struct rt6_info ip6_null_entry_template = {
d8d1f30b
CG		 307 .dst = {
308 .__refcnt = ATOMIC_INIT(1),
309 .__use = 1,
2c20cbd7 310 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 311 .error = -ENETUNREACH,
d8d1f30b
CG		 312 .input = ip6_pkt_discard,
313 .output = ip6_pkt_discard_out,
1da177e4
LT		 314 },
315 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
1da177e4
LT		 316};
317
101367c2
TG		 318#ifdef CONFIG_IPV6_MULTIPLE_TABLES
319
fb0af4c7 320static const struct rt6_info ip6_prohibit_entry_template = {
d8d1f30b
CG		 321 .dst = {
322 .__refcnt = ATOMIC_INIT(1),
323 .__use = 1,
2c20cbd7 324 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 325 .error = -EACCES,
d8d1f30b
CG		 326 .input = ip6_pkt_prohibit,
327 .output = ip6_pkt_prohibit_out,
101367c2
TG		 328 },
329 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 330};
331
fb0af4c7 332static const struct rt6_info ip6_blk_hole_entry_template = {
d8d1f30b
CG		 333 .dst = {
334 .__refcnt = ATOMIC_INIT(1),
335 .__use = 1,
2c20cbd7 336 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 337 .error = -EINVAL,
d8d1f30b 338 .input = dst_discard,
ede2059d 339 .output = dst_discard_out,
101367c2
TG		 340 },
341 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 342};
343
344#endif
345
ebfa45f0
MKL		 346static void rt6_info_init(struct rt6_info *rt)
347{
348 struct dst_entry *dst = &rt->dst;
349
350 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
ebfa45f0
MKL		 351 INIT_LIST_HEAD(&rt->rt6i_uncached);
352}
353
1da177e4 354/* allocate dst with ip6_dst_ops */
93531c67
DA		 355struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
356 int flags)
1da177e4 357{
97bab73f 358 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
b2a9c0ed 359 1, DST_OBSOLETE_FORCE_CHK, flags);
cf911662 360
81eb8447 361 if (rt) {
ebfa45f0 362 rt6_info_init(rt);
81eb8447
WW		 363 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
364 }
8104891b 365
cf911662 366 return rt;
1da177e4 367}
9ab179d8 368EXPORT_SYMBOL(ip6_dst_alloc);
d52d3997 369
1da177e4
LT		 370static void ip6_dst_destroy(struct dst_entry *dst)
371{
372 struct rt6_info *rt = (struct rt6_info *)dst;
a68886a6 373 struct fib6_info *from;
8d0b94af 374 struct inet6_dev *idev;
1da177e4 375
1620a336 376 ip_dst_metrics_put(dst);
8d0b94af
MKL		 377 rt6_uncached_list_del(rt);
378
379 idev = rt->rt6i_idev;
38308473 380 if (idev) {
1da177e4
LT		 381 rt->rt6i_idev = NULL;
382 in6_dev_put(idev);
1ab1457c 383 }
1716a961 384
0e233874 385 from = xchg((__force struct fib6_info **)&rt->from, NULL);
93531c67 386 fib6_info_release(from);
b3419363
DM		 387}
388
1da177e4
LT		 389static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
390 int how)
391{
392 struct rt6_info *rt = (struct rt6_info *)dst;
393 struct inet6_dev *idev = rt->rt6i_idev;
5a3e55d6 394 struct net_device *loopback_dev =
c346dca1 395 dev_net(dev)->loopback_dev;
1da177e4 396
e5645f51
WW		 397 if (idev && idev->dev != loopback_dev) {
398 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
399 if (loopback_idev) {
400 rt->rt6i_idev = loopback_idev;
401 in6_dev_put(idev);
97cac082 402 }
1da177e4
LT		 403 }
404}
405
5973fb1e
MKL		 406static bool __rt6_check_expired(const struct rt6_info *rt)
407{
408 if (rt->rt6i_flags & RTF_EXPIRES)
409 return time_after(jiffies, rt->dst.expires);
410 else
411 return false;
412}
413
a50feda5 414static bool rt6_check_expired(const struct rt6_info *rt)
1da177e4 415{
a68886a6
DA		 416 struct fib6_info *from;
417
418 from = rcu_dereference(rt->from);
419
1716a961
G		 420 if (rt->rt6i_flags & RTF_EXPIRES) {
421 if (time_after(jiffies, rt->dst.expires))
a50feda5 422 return true;
a68886a6 423 } else if (from) {
1e2ea8ad 424 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
a68886a6 425 fib6_check_expired(from);
1716a961 426 }
a50feda5 427 return false;
1da177e4
LT		 428}
429
b1d40991
DA		 430void fib6_select_path(const struct net *net, struct fib6_result *res,
431 struct flowi6 *fl6, int oif, bool have_oif_match,
432 const struct sk_buff *skb, int strict)
51ebd318 433{
8d1c802b 434 struct fib6_info *sibling, *next_sibling;
b1d40991
DA		 435 struct fib6_info *match = res->f6i;
436
34fe5a1c 437 if (!match->nh && (!match->fib6_nsiblings || have_oif_match))
b1d40991 438 goto out;
51ebd318 439
34fe5a1c
DA		 440 if (match->nh && have_oif_match && res->nh)
441 return;
442
b673d6cc
JS		 443 /* We might have already computed the hash for ICMPv6 errors. In such
444 * case it will always be non-zero. Otherwise now is the time to do it.
445 */
f88d8ea6
DA		 446 if (!fl6->mp_hash &&
447 (!match->nh || nexthop_is_multipath(match->nh)))
b4bac172 448 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
b673d6cc 449
f88d8ea6
DA		 450 if (unlikely(match->nh)) {
451 nexthop_path_fib6_result(res, fl6->mp_hash);
452 return;
453 }
454
1cf844c7 455 if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
b1d40991 456 goto out;
3d709f69 457
93c2fb25
DA		 458 list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
459 fib6_siblings) {
1cf844c7 460 const struct fib6_nh *nh = sibling->fib6_nh;
5e670d84
DA		 461 int nh_upper_bound;
462
702cea56 463 nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
5e670d84 464 if (fl6->mp_hash > nh_upper_bound)
3d709f69 465 continue;
702cea56 466 if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
3d709f69
IS		 467 break;
468 match = sibling;
469 break;
470 }
471
b1d40991
DA		 472out:
473 res->f6i = match;
1cf844c7 474 res->nh = match->fib6_nh;
51ebd318
ND		 475}
476
1da177e4 477/*
66f5d6ce 478 * Route lookup. rcu_read_lock() should be held.
1da177e4
LT		 479 */
480
0c59d006
DA		 481static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
482 const struct in6_addr *saddr, int oif, int flags)
483{
484 const struct net_device *dev;
485
486 if (nh->fib_nh_flags & RTNH_F_DEAD)
487 return false;
488
489 dev = nh->fib_nh_dev;
490 if (oif) {
491 if (dev->ifindex == oif)
492 return true;
493 } else {
494 if (ipv6_chk_addr(net, saddr, dev,
495 flags & RT6_LOOKUP_F_IFACE))
496 return true;
497 }
498
499 return false;
500}
501
962b6803
DA		 502struct fib6_nh_dm_arg {
503 struct net *net;
504 const struct in6_addr *saddr;
505 int oif;
506 int flags;
507 struct fib6_nh *nh;
508};
509
510static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
511{
512 struct fib6_nh_dm_arg *arg = _arg;
513
514 arg->nh = nh;
515 return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
516 arg->flags);
517}
518
519/* returns fib6_nh from nexthop or NULL */
520static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
521 struct fib6_result *res,
522 const struct in6_addr *saddr,
523 int oif, int flags)
524{
525 struct fib6_nh_dm_arg arg = {
526 .net = net,
527 .saddr = saddr,
528 .oif = oif,
529 .flags = flags,
530 };
531
532 if (nexthop_is_blackhole(nh))
533 return NULL;
534
535 if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
536 return arg.nh;
537
538 return NULL;
539}
540
75ef7389
DA		 541static void rt6_device_match(struct net *net, struct fib6_result *res,
542 const struct in6_addr *saddr, int oif, int flags)
1da177e4 543{
75ef7389
DA		 544 struct fib6_info *f6i = res->f6i;
545 struct fib6_info *spf6i;
546 struct fib6_nh *nh;
1da177e4 547
75ef7389 548 if (!oif && ipv6_addr_any(saddr)) {
f88d8ea6
DA		 549 if (unlikely(f6i->nh)) {
550 nh = nexthop_fib6_nh(f6i->nh);
551 if (nexthop_is_blackhole(f6i->nh))
552 goto out_blackhole;
553 } else {
554 nh = f6i->fib6_nh;
555 }
7d21fec9
DA		 556 if (!(nh->fib_nh_flags & RTNH_F_DEAD))
557 goto out;
75ef7389 558 }
dd3abc4e 559
75ef7389 560 for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
962b6803
DA		 561 bool matched = false;
562
563 if (unlikely(spf6i->nh)) {
564 nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
565 oif, flags);
566 if (nh)
567 matched = true;
568 } else {
569 nh = spf6i->fib6_nh;
570 if (__rt6_device_match(net, nh, saddr, oif, flags))
571 matched = true;
572 }
573 if (matched) {
75ef7389 574 res->f6i = spf6i;
7d21fec9 575 goto out;
75ef7389 576 }
dd3abc4e 577 }
1da177e4 578
75ef7389
DA		 579 if (oif && flags & RT6_LOOKUP_F_IFACE) {
580 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 581 nh = res->f6i->fib6_nh;
7d21fec9 582 goto out;
75ef7389 583 }
8067bb8c 584
f88d8ea6
DA		 585 if (unlikely(f6i->nh)) {
586 nh = nexthop_fib6_nh(f6i->nh);
587 if (nexthop_is_blackhole(f6i->nh))
588 goto out_blackhole;
589 } else {
590 nh = f6i->fib6_nh;
591 }
592
7d21fec9 593 if (nh->fib_nh_flags & RTNH_F_DEAD) {
75ef7389 594 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 595 nh = res->f6i->fib6_nh;
75ef7389 596 }
7d21fec9
DA		 597out:
598 res->nh = nh;
599 res->fib6_type = res->f6i->fib6_type;
600 res->fib6_flags = res->f6i->fib6_flags;
f88d8ea6
DA		 601 return;
602
603out_blackhole:
604 res->fib6_flags |= RTF_REJECT;
605 res->fib6_type = RTN_BLACKHOLE;
606 res->nh = nh;
1da177e4
LT		 607}
608
27097255 609#ifdef CONFIG_IPV6_ROUTER_PREF
c2f17e82
HFS		 610struct __rt6_probe_work {
611 struct work_struct work;
612 struct in6_addr target;
613 struct net_device *dev;
614};
615
616static void rt6_probe_deferred(struct work_struct *w)
617{
618 struct in6_addr mcaddr;
619 struct __rt6_probe_work *work =
620 container_of(w, struct __rt6_probe_work, work);
621
622 addrconf_addr_solict_mult(&work->target, &mcaddr);
adc176c5 623 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
c2f17e82 624 dev_put(work->dev);
662f5533 625 kfree(work);
c2f17e82
HFS		 626}
627
cc3a86c8 628static void rt6_probe(struct fib6_nh *fib6_nh)
27097255 629{
f547fac6 630 struct __rt6_probe_work *work = NULL;
5e670d84 631 const struct in6_addr *nh_gw;
1bef4c22 632 unsigned long last_probe;
f2c31e32 633 struct neighbour *neigh;
5e670d84 634 struct net_device *dev;
f547fac6 635 struct inet6_dev *idev;
5e670d84 636
27097255
YH		 637 /*
638 * Okay, this does not seem to be appropriate
639 * for now, however, we need to check if it
640 * is really so; aka Router Reachability Probing.
641 *
642 * Router Reachability Probe MUST be rate-limited
643 * to no more than one per minute.
644 */
004b3942 645 if (!fib6_nh->fib_nh_gw_family)
7ff74a59 646 return;
5e670d84 647
cc3a86c8
DA		 648 nh_gw = &fib6_nh->fib_nh_gw6;
649 dev = fib6_nh->fib_nh_dev;
2152caea 650 rcu_read_lock_bh();
1bef4c22 651 last_probe = READ_ONCE(fib6_nh->last_probe);
f547fac6 652 idev = __in6_dev_get(dev);
5e670d84 653 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
2152caea 654 if (neigh) {
8d6c31bf
MKL		 655 if (neigh->nud_state & NUD_VALID)
656 goto out;
657
2152caea 658 write_lock(&neigh->lock);
990edb42
MKL		 659 if (!(neigh->nud_state & NUD_VALID) &&
660 time_after(jiffies,
dcd1f572 661 neigh->updated + idev->cnf.rtr_probe_interval)) {
990edb42
MKL		 662 work = kmalloc(sizeof(*work), GFP_ATOMIC);
663 if (work)
664 __neigh_set_probe_once(neigh);
c2f17e82 665 }
2152caea 666 write_unlock(&neigh->lock);
1bef4c22 667 } else if (time_after(jiffies, last_probe +
f547fac6 668 idev->cnf.rtr_probe_interval)) {
990edb42 669 work = kmalloc(sizeof(*work), GFP_ATOMIC);
f2c31e32 670 }
990edb42 671
1bef4c22
ED		 672 if (!work || cmpxchg(&fib6_nh->last_probe,
673 last_probe, jiffies) != last_probe) {
674 kfree(work);
675 } else {
990edb42 676 INIT_WORK(&work->work, rt6_probe_deferred);
5e670d84
DA		 677 work->target = *nh_gw;
678 dev_hold(dev);
679 work->dev = dev;
990edb42
MKL		 680 schedule_work(&work->work);
681 }
682
8d6c31bf 683out:
2152caea 684 rcu_read_unlock_bh();
27097255
YH		 685}
686#else
cc3a86c8 687static inline void rt6_probe(struct fib6_nh *fib6_nh)
27097255 688{
27097255
YH		 689}
690#endif
691
1da177e4 692/*
554cfb7e 693 * Default Router Selection (RFC 2461 6.3.6)
1da177e4 694 */
1ba9a895 695static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
1da177e4 696{
afc154e9 697 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
5e670d84 698 struct neighbour *neigh;
f2c31e32 699
145a3621 700 rcu_read_lock_bh();
1ba9a895
DA		 701 neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
702 &fib6_nh->fib_nh_gw6);
145a3621
YH		 703 if (neigh) {
704 read_lock(&neigh->lock);
554cfb7e 705 if (neigh->nud_state & NUD_VALID)
afc154e9 706 ret = RT6_NUD_SUCCEED;
398bcbeb 707#ifdef CONFIG_IPV6_ROUTER_PREF
a5a81f0b 708 else if (!(neigh->nud_state & NUD_FAILED))
afc154e9 709 ret = RT6_NUD_SUCCEED;
7e980569
JB		 710 else
711 ret = RT6_NUD_FAIL_PROBE;
398bcbeb 712#endif
145a3621 713 read_unlock(&neigh->lock);
afc154e9
HFS		 714 } else {
715 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
7e980569 716 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
a5a81f0b 717 }
145a3621
YH		 718 rcu_read_unlock_bh();
719
a5a81f0b 720 return ret;
1da177e4
LT		 721}
722
702cea56
DA		 723static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
724 int strict)
1da177e4 725{
6e1809a5
DA		 726 int m = 0;
727
728 if (!oif || nh->fib_nh_dev->ifindex == oif)
729 m = 2;
1ab1457c 730
77d16f45 731 if (!m && (strict & RT6_LOOKUP_F_IFACE))
afc154e9 732 return RT6_NUD_FAIL_HARD;
ebacaaa0 733#ifdef CONFIG_IPV6_ROUTER_PREF
702cea56 734 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
ebacaaa0 735#endif
1ba9a895 736 if ((strict & RT6_LOOKUP_F_REACHABLE) &&
702cea56 737 !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
1ba9a895 738 int n = rt6_check_neigh(nh);
afc154e9
HFS		 739 if (n < 0)
740 return n;
741 }
554cfb7e
YH		 742 return m;
743}
744
28679ed1
DA		 745static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
746 int oif, int strict, int *mpri, bool *do_rr)
554cfb7e 747{
afc154e9 748 bool match_do_rr = false;
28679ed1
DA		 749 bool rc = false;
750 int m;
35103d11 751
28679ed1 752 if (nh->fib_nh_flags & RTNH_F_DEAD)
8067bb8c
IS		 753 goto out;
754
28679ed1
DA		 755 if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
756 nh->fib_nh_flags & RTNH_F_LINKDOWN &&
d5d32e4b 757 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
35103d11 758 goto out;
f11e6659 759
28679ed1 760 m = rt6_score_route(nh, fib6_flags, oif, strict);
7e980569 761 if (m == RT6_NUD_FAIL_DO_RR) {
afc154e9
HFS		 762 match_do_rr = true;
763 m = 0; /* lowest valid score */
7e980569 764 } else if (m == RT6_NUD_FAIL_HARD) {
f11e6659 765 goto out;
afc154e9
HFS		 766 }
767
768 if (strict & RT6_LOOKUP_F_REACHABLE)
28679ed1 769 rt6_probe(nh);
f11e6659 770
7e980569 771 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
f11e6659 772 if (m > *mpri) {
afc154e9 773 *do_rr = match_do_rr;
f11e6659 774 *mpri = m;
28679ed1 775 rc = true;
f11e6659 776 }
f11e6659 777out:
28679ed1 778 return rc;
f11e6659
DM		 779}
780
17a5984e
DA		 781struct fib6_nh_frl_arg {
782 u32 flags;
783 int oif;
784 int strict;
785 int *mpri;
786 bool *do_rr;
787 struct fib6_nh *nh;
788};
789
790static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
791{
792 struct fib6_nh_frl_arg *arg = _arg;
793
794 arg->nh = nh;
795 return find_match(nh, arg->flags, arg->oif, arg->strict,
796 arg->mpri, arg->do_rr);
797}
798
b7bc4b6a 799static void __find_rr_leaf(struct fib6_info *f6i_start,
30c15f03 800 struct fib6_info *nomatch, u32 metric,
b7bc4b6a 801 struct fib6_result *res, struct fib6_info **cont,
30c15f03 802 int oif, int strict, bool *do_rr, int *mpri)
f11e6659 803{
b7bc4b6a 804 struct fib6_info *f6i;
1da177e4 805
b7bc4b6a
DA		 806 for (f6i = f6i_start;
807 f6i && f6i != nomatch;
808 f6i = rcu_dereference(f6i->fib6_next)) {
17a5984e 809 bool matched = false;
30c15f03
DA		 810 struct fib6_nh *nh;
811
b7bc4b6a
DA		 812 if (cont && f6i->fib6_metric != metric) {
813 *cont = f6i;
30c15f03 814 return;
9fbdcfaf
SK		 815 }
816
b7bc4b6a 817 if (fib6_check_expired(f6i))
28679ed1
DA		 818 continue;
819
17a5984e
DA		 820 if (unlikely(f6i->nh)) {
821 struct fib6_nh_frl_arg arg = {
822 .flags = f6i->fib6_flags,
823 .oif = oif,
824 .strict = strict,
825 .mpri = mpri,
826 .do_rr = do_rr
827 };
828
829 if (nexthop_is_blackhole(f6i->nh)) {
830 res->fib6_flags = RTF_REJECT;
831 res->fib6_type = RTN_BLACKHOLE;
832 res->f6i = f6i;
833 res->nh = nexthop_fib6_nh(f6i->nh);
834 return;
835 }
836 if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
837 &arg)) {
838 matched = true;
839 nh = arg.nh;
840 }
841 } else {
842 nh = f6i->fib6_nh;
843 if (find_match(nh, f6i->fib6_flags, oif, strict,
844 mpri, do_rr))
845 matched = true;
846 }
847 if (matched) {
b7bc4b6a
DA		 848 res->f6i = f6i;
849 res->nh = nh;
7d21fec9
DA		 850 res->fib6_flags = f6i->fib6_flags;
851 res->fib6_type = f6i->fib6_type;
b7bc4b6a 852 }
9fbdcfaf 853 }
30c15f03 854}
9fbdcfaf 855
b7bc4b6a
DA		 856static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
857 struct fib6_info *rr_head, int oif, int strict,
858 bool *do_rr, struct fib6_result *res)
30c15f03 859{
b7bc4b6a
DA		 860 u32 metric = rr_head->fib6_metric;
861 struct fib6_info *cont = NULL;
30c15f03 862 int mpri = -1;
9fbdcfaf 863
b7bc4b6a 864 __find_rr_leaf(rr_head, NULL, metric, res, &cont,
30c15f03 865 oif, strict, do_rr, &mpri);
28679ed1 866
b7bc4b6a 867 __find_rr_leaf(leaf, rr_head, metric, res, &cont,
30c15f03 868 oif, strict, do_rr, &mpri);
9fbdcfaf 869
b7bc4b6a
DA		 870 if (res->f6i || !cont)
871 return;
9fbdcfaf 872
b7bc4b6a 873 __find_rr_leaf(cont, NULL, metric, res, NULL,
30c15f03 874 oif, strict, do_rr, &mpri);
f11e6659 875}
1da177e4 876
b7bc4b6a
DA		 877static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
878 struct fib6_result *res, int strict)
f11e6659 879{
8d1c802b 880 struct fib6_info *leaf = rcu_dereference(fn->leaf);
b7bc4b6a 881 struct fib6_info *rt0;
afc154e9 882 bool do_rr = false;
17ecf590 883 int key_plen;
1da177e4 884
b7bc4b6a
DA		 885 /* make sure this function or its helpers sets f6i */
886 res->f6i = NULL;
887
421842ed 888 if (!leaf || leaf == net->ipv6.fib6_null_entry)
b7bc4b6a 889 goto out;
8d1040e8 890
66f5d6ce 891 rt0 = rcu_dereference(fn->rr_ptr);
f11e6659 892 if (!rt0)
66f5d6ce 893 rt0 = leaf;
1da177e4 894
17ecf590
WW		 895 /* Double check to make sure fn is not an intermediate node
896 * and fn->leaf does not points to its child's leaf
897 * (This might happen if all routes under fn are deleted from
898 * the tree and fib6_repair_tree() is called on the node.)
899 */
93c2fb25 900 key_plen = rt0->fib6_dst.plen;
17ecf590 901#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 902 if (rt0->fib6_src.plen)
903 key_plen = rt0->fib6_src.plen;
17ecf590
WW		 904#endif
905 if (fn->fn_bit != key_plen)
b7bc4b6a 906 goto out;
1da177e4 907
b7bc4b6a 908 find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
afc154e9 909 if (do_rr) {
8fb11a9a 910 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
f11e6659 911
554cfb7e 912 /* no entries matched; do round-robin */
93c2fb25 913 if (!next || next->fib6_metric != rt0->fib6_metric)
8d1040e8 914 next = leaf;
f11e6659 915
66f5d6ce 916 if (next != rt0) {
93c2fb25 917 spin_lock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 918 /* make sure next is not being deleted from the tree */
93c2fb25 919 if (next->fib6_node)
66f5d6ce 920 rcu_assign_pointer(fn->rr_ptr, next);
93c2fb25 921 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 922 }
1da177e4 923 }
1da177e4 924
b7bc4b6a
DA		 925out:
926 if (!res->f6i) {
927 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 928 res->nh = res->f6i->fib6_nh;
7d21fec9
DA		 929 res->fib6_flags = res->f6i->fib6_flags;
930 res->fib6_type = res->f6i->fib6_type;
b7bc4b6a 931 }
1da177e4
LT		 932}
933
85bd05de 934static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
8b9df265 935{
85bd05de
DA		 936 return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
937 res->nh->fib_nh_gw_family;
8b9df265
MKL		 938}
939
70ceb4f5
YH		 940#ifdef CONFIG_IPV6_ROUTE_INFO
941int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
b71d1d42 942 const struct in6_addr *gwaddr)
70ceb4f5 943{
c346dca1 944 struct net *net = dev_net(dev);
70ceb4f5
YH		 945 struct route_info *rinfo = (struct route_info *) opt;
946 struct in6_addr prefix_buf, *prefix;
947 unsigned int pref;
4bed72e4 948 unsigned long lifetime;
8d1c802b 949 struct fib6_info *rt;
70ceb4f5
YH		 950
951 if (len < sizeof(struct route_info)) {
952 return -EINVAL;
953 }
954
955 /* Sanity check for prefix_len and length */
956 if (rinfo->length > 3) {
957 return -EINVAL;
958 } else if (rinfo->prefix_len > 128) {
959 return -EINVAL;
960 } else if (rinfo->prefix_len > 64) {
961 if (rinfo->length < 2) {
962 return -EINVAL;
963 }
964 } else if (rinfo->prefix_len > 0) {
965 if (rinfo->length < 1) {
966 return -EINVAL;
967 }
968 }
969
970 pref = rinfo->route_pref;
971 if (pref == ICMPV6_ROUTER_PREF_INVALID)
3933fc95 972 return -EINVAL;
70ceb4f5 973
4bed72e4 974 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
70ceb4f5
YH		 975
976 if (rinfo->length == 3)
977 prefix = (struct in6_addr *)rinfo->prefix;
978 else {
979 /* this function is safe */
980 ipv6_addr_prefix(&prefix_buf,
981 (struct in6_addr *)rinfo->prefix,
982 rinfo->prefix_len);
983 prefix = &prefix_buf;
984 }
985
f104a567 986 if (rinfo->prefix_len == 0)
afb1d4b5 987 rt = rt6_get_dflt_router(net, gwaddr, dev);
f104a567
DJ		 988 else
989 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
830218c1 990 gwaddr, dev);
70ceb4f5
YH		 991
992 if (rt && !lifetime) {
11dd74b3 993 ip6_del_rt(net, rt, false);
70ceb4f5
YH		 994 rt = NULL;
995 }
996
997 if (!rt && lifetime)
830218c1
DA		 998 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
999 dev, pref);
70ceb4f5 1000 else if (rt)
93c2fb25
DA		 1001 rt->fib6_flags = RTF_ROUTEINFO |
1002 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
70ceb4f5
YH		 1003
1004 if (rt) {
1716a961 1005 if (!addrconf_finite_timeout(lifetime))
14895687 1006 fib6_clean_expires(rt);
1716a961 1007 else
14895687 1008 fib6_set_expires(rt, jiffies + HZ * lifetime);
1716a961 1009
93531c67 1010 fib6_info_release(rt);
70ceb4f5
YH		 1011 }
1012 return 0;
1013}
1014#endif
1015
ae90d867
DA		 1016/*
1017 * Misc support functions
1018 */
1019
1020/* called with rcu_lock held */
0d161581 1021static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
ae90d867 1022{
0d161581 1023 struct net_device *dev = res->nh->fib_nh_dev;
ae90d867 1024
7d21fec9 1025 if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
ae90d867
DA		 1026 /* for copies of local routes, dst->dev needs to be the
1027 * device if it is a master device, the master device if
1028 * device is enslaved, and the loopback as the default
1029 */
1030 if (netif_is_l3_slave(dev) &&
7d21fec9 1031 !rt6_need_strict(&res->f6i->fib6_dst.addr))
ae90d867
DA		 1032 dev = l3mdev_master_dev_rcu(dev);
1033 else if (!netif_is_l3_master(dev))
1034 dev = dev_net(dev)->loopback_dev;
1035 /* last case is netif_is_l3_master(dev) is true in which
1036 * case we want dev returned to be dev
1037 */
1038 }
1039
1040 return dev;
1041}
1042
6edb3c96
DA		 1043static const int fib6_prop[RTN_MAX + 1] = {
1044 [RTN_UNSPEC] = 0,
1045 [RTN_UNICAST] = 0,
1046 [RTN_LOCAL] = 0,
1047 [RTN_BROADCAST] = 0,
1048 [RTN_ANYCAST] = 0,
1049 [RTN_MULTICAST] = 0,
1050 [RTN_BLACKHOLE] = -EINVAL,
1051 [RTN_UNREACHABLE] = -EHOSTUNREACH,
1052 [RTN_PROHIBIT] = -EACCES,
1053 [RTN_THROW] = -EAGAIN,
1054 [RTN_NAT] = -EINVAL,
1055 [RTN_XRESOLVE] = -EINVAL,
1056};
1057
1058static int ip6_rt_type_to_error(u8 fib6_type)
1059{
1060 return fib6_prop[fib6_type];
1061}
1062
8d1c802b 1063static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
3b6761d1
DA		 1064{
1065 unsigned short flags = 0;
1066
1067 if (rt->dst_nocount)
1068 flags |= DST_NOCOUNT;
1069 if (rt->dst_nopolicy)
1070 flags |= DST_NOPOLICY;
3b6761d1
DA		 1071
1072 return flags;
1073}
1074
7d21fec9 1075static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
6edb3c96 1076{
7d21fec9 1077 rt->dst.error = ip6_rt_type_to_error(fib6_type);
6edb3c96 1078
7d21fec9 1079 switch (fib6_type) {
6edb3c96
DA		 1080 case RTN_BLACKHOLE:
1081 rt->dst.output = dst_discard_out;
1082 rt->dst.input = dst_discard;
1083 break;
1084 case RTN_PROHIBIT:
1085 rt->dst.output = ip6_pkt_prohibit_out;
1086 rt->dst.input = ip6_pkt_prohibit;
1087 break;
1088 case RTN_THROW:
1089 case RTN_UNREACHABLE:
1090 default:
1091 rt->dst.output = ip6_pkt_discard_out;
1092 rt->dst.input = ip6_pkt_discard;
1093 break;
1094 }
1095}
1096
0d161581 1097static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
6edb3c96 1098{
7d21fec9 1099 struct fib6_info *f6i = res->f6i;
0d161581 1100
7d21fec9
DA		 1101 if (res->fib6_flags & RTF_REJECT) {
1102 ip6_rt_init_dst_reject(rt, res->fib6_type);
6edb3c96
DA		 1103 return;
1104 }
1105
1106 rt->dst.error = 0;
1107 rt->dst.output = ip6_output;
1108
7d21fec9 1109 if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
6edb3c96 1110 rt->dst.input = ip6_input;
7d21fec9 1111 } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
6edb3c96
DA		 1112 rt->dst.input = ip6_mc_input;
1113 } else {
1114 rt->dst.input = ip6_forward;
1115 }
1116
0d161581
DA		 1117 if (res->nh->fib_nh_lws) {
1118 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
6edb3c96
DA		 1119 lwtunnel_set_redirect(&rt->dst);
1120 }
1121
1122 rt->dst.lastuse = jiffies;
1123}
1124
e873e4b9 1125/* Caller must already hold reference to @from */
8d1c802b 1126static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
ae90d867 1127{
ae90d867 1128 rt->rt6i_flags &= ~RTF_EXPIRES;
a68886a6 1129 rcu_assign_pointer(rt->from, from);
e1255ed4 1130 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
ae90d867
DA		 1131}
1132
0d161581
DA		 1133/* Caller must already hold reference to f6i in result */
1134static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
ae90d867 1135{
0d161581
DA		 1136 const struct fib6_nh *nh = res->nh;
1137 const struct net_device *dev = nh->fib_nh_dev;
1138 struct fib6_info *f6i = res->f6i;
dcd1f572 1139
0d161581 1140 ip6_rt_init_dst(rt, res);
6edb3c96 1141
0d161581 1142 rt->rt6i_dst = f6i->fib6_dst;
dcd1f572 1143 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
7d21fec9 1144 rt->rt6i_flags = res->fib6_flags;
0d161581
DA		 1145 if (nh->fib_nh_gw_family) {
1146 rt->rt6i_gateway = nh->fib_nh_gw6;
2b2450ca
DA		 1147 rt->rt6i_flags |= RTF_GATEWAY;
1148 }
0d161581 1149 rt6_set_from(rt, f6i);
ae90d867 1150#ifdef CONFIG_IPV6_SUBTREES
0d161581 1151 rt->rt6i_src = f6i->fib6_src;
ae90d867 1152#endif
ae90d867
DA		 1153}
1154
a3c00e46
MKL		 1155static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1156 struct in6_addr *saddr)
1157{
66f5d6ce 1158 struct fib6_node *pn, *sn;
a3c00e46
MKL		 1159 while (1) {
1160 if (fn->fn_flags & RTN_TL_ROOT)
1161 return NULL;
66f5d6ce
WW		 1162 pn = rcu_dereference(fn->parent);
1163 sn = FIB6_SUBTREE(pn);
1164 if (sn && sn != fn)
6454743b 1165 fn = fib6_node_lookup(sn, NULL, saddr);
a3c00e46
MKL		 1166 else
1167 fn = pn;
1168 if (fn->fn_flags & RTN_RTINFO)
1169 return fn;
1170 }
1171}
c71099ac 1172
10585b43 1173static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
d3843fe5
WW		 1174{
1175 struct rt6_info *rt = *prt;
1176
1177 if (dst_hold_safe(&rt->dst))
1178 return true;
10585b43 1179 if (net) {
d3843fe5
WW		 1180 rt = net->ipv6.ip6_null_entry;
1181 dst_hold(&rt->dst);
1182 } else {
1183 rt = NULL;
1184 }
1185 *prt = rt;
1186 return false;
1187}
1188
dec9b0e2 1189/* called with rcu_lock held */
9b6b35ab 1190static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
dec9b0e2 1191{
9b6b35ab
DA		 1192 struct net_device *dev = res->nh->fib_nh_dev;
1193 struct fib6_info *f6i = res->f6i;
1194 unsigned short flags;
dec9b0e2
DA		 1195 struct rt6_info *nrt;
1196
9b6b35ab 1197 if (!fib6_info_hold_safe(f6i))
1c87e79a 1198 goto fallback;
e873e4b9 1199
9b6b35ab 1200 flags = fib6_info_dst_flags(f6i);
93531c67 1201 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1c87e79a 1202 if (!nrt) {
9b6b35ab 1203 fib6_info_release(f6i);
1c87e79a
XL		 1204 goto fallback;
1205 }
dec9b0e2 1206
0d161581 1207 ip6_rt_copy_init(nrt, res);
1c87e79a
XL		 1208 return nrt;
1209
1210fallback:
1211 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1212 dst_hold(&nrt->dst);
dec9b0e2
DA		 1213 return nrt;
1214}
1215
55cced4f 1216INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_lookup(struct net *net,
8ed67789 1217 struct fib6_table *table,
b75cc8f9
DA		 1218 struct flowi6 *fl6,
1219 const struct sk_buff *skb,
1220 int flags)
1da177e4 1221{
b1d40991 1222 struct fib6_result res = {};
1da177e4 1223 struct fib6_node *fn;
23fb93a4 1224 struct rt6_info *rt;
1da177e4 1225
b6cdbc85
DA		 1226 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1227 flags &= ~RT6_LOOKUP_F_IFACE;
1228
66f5d6ce 1229 rcu_read_lock();
6454743b 1230 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
c71099ac 1231restart:
b1d40991
DA		 1232 res.f6i = rcu_dereference(fn->leaf);
1233 if (!res.f6i)
1234 res.f6i = net->ipv6.fib6_null_entry;
af52a52c 1235 else
75ef7389
DA		 1236 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1237 flags);
af52a52c 1238
b1d40991 1239 if (res.f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 1240 fn = fib6_backtrack(fn, &fl6->saddr);
1241 if (fn)
1242 goto restart;
2b760fcf 1243
af52a52c
DA		 1244 rt = net->ipv6.ip6_null_entry;
1245 dst_hold(&rt->dst);
1246 goto out;
f88d8ea6
DA		 1247 } else if (res.fib6_flags & RTF_REJECT) {
1248 goto do_create;
af52a52c 1249 }
d3843fe5 1250
b1d40991
DA		 1251 fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1252 fl6->flowi6_oif != 0, skb, flags);
1253
2b760fcf 1254 /* Search through exception table */
7e4b5128 1255 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
23fb93a4 1256 if (rt) {
10585b43 1257 if (ip6_hold_safe(net, &rt))
dec9b0e2 1258 dst_use_noref(&rt->dst, jiffies);
23fb93a4 1259 } else {
f88d8ea6 1260do_create:
9b6b35ab 1261 rt = ip6_create_rt_rcu(&res);
dec9b0e2 1262 }
b811580d 1263
af52a52c 1264out:
8ff2e5b2 1265 trace_fib6_table_lookup(net, &res, table, fl6);
af52a52c 1266
66f5d6ce 1267 rcu_read_unlock();
b811580d 1268
c71099ac 1269 return rt;
c71099ac
TG		 1270}
1271
67ba4152 1272struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
b75cc8f9 1273 const struct sk_buff *skb, int flags)
ea6e574e 1274{
b75cc8f9 1275 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
ea6e574e
FW		 1276}
1277EXPORT_SYMBOL_GPL(ip6_route_lookup);
1278
9acd9f3a 1279struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
b75cc8f9
DA		 1280 const struct in6_addr *saddr, int oif,
1281 const struct sk_buff *skb, int strict)
c71099ac 1282{
4c9483b2
DM		 1283 struct flowi6 fl6 = {
1284 .flowi6_oif = oif,
1285 .daddr = *daddr,
c71099ac
TG		 1286 };
1287 struct dst_entry *dst;
77d16f45 1288 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
c71099ac 1289
adaa70bb 1290 if (saddr) {
4c9483b2 1291 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
adaa70bb
TG		 1292 flags |= RT6_LOOKUP_F_HAS_SADDR;
1293 }
1294
b75cc8f9 1295 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
c71099ac
TG		 1296 if (dst->error == 0)
1297 return (struct rt6_info *) dst;
1298
1299 dst_release(dst);
1300
1da177e4
LT		 1301 return NULL;
1302}
7159039a
YH		 1303EXPORT_SYMBOL(rt6_lookup);
1304
c71099ac 1305/* ip6_ins_rt is called with FREE table->tb6_lock.
1cfb71ee
WW		 1306 * It takes new route entry, the addition fails by any reason the
1307 * route is released.
1308 * Caller must hold dst before calling it.
1da177e4
LT		 1309 */
1310
8d1c802b 1311static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
333c4301 1312 struct netlink_ext_ack *extack)
1da177e4
LT		 1313{
1314 int err;
c71099ac 1315 struct fib6_table *table;
1da177e4 1316
93c2fb25 1317 table = rt->fib6_table;
66f5d6ce 1318 spin_lock_bh(&table->tb6_lock);
d4ead6b3 1319 err = fib6_add(&table->tb6_root, rt, info, extack);
66f5d6ce 1320 spin_unlock_bh(&table->tb6_lock);
1da177e4
LT		 1321
1322 return err;
1323}
1324
8d1c802b 1325int ip6_ins_rt(struct net *net, struct fib6_info *rt)
40e22e8f 1326{
afb1d4b5 1327 struct nl_info info = { .nl_net = net, };
e715b6d3 1328
d4ead6b3 1329 return __ip6_ins_rt(rt, &info, NULL);
40e22e8f
TG		 1330}
1331
85bd05de 1332static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
8b9df265
MKL		 1333 const struct in6_addr *daddr,
1334 const struct in6_addr *saddr)
1da177e4 1335{
85bd05de 1336 struct fib6_info *f6i = res->f6i;
4832c30d 1337 struct net_device *dev;
1da177e4
LT		 1338 struct rt6_info *rt;
1339
1340 /*
1341 * Clone the route.
1342 */
1343
85bd05de 1344 if (!fib6_info_hold_safe(f6i))
e873e4b9
WW		 1345 return NULL;
1346
0d161581 1347 dev = ip6_rt_get_dev_rcu(res);
93531c67 1348 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
e873e4b9 1349 if (!rt) {
85bd05de 1350 fib6_info_release(f6i);
83a09abd 1351 return NULL;
e873e4b9 1352 }
83a09abd 1353
0d161581 1354 ip6_rt_copy_init(rt, res);
83a09abd 1355 rt->rt6i_flags |= RTF_CACHE;
83a09abd
MKL		 1356 rt->rt6i_dst.addr = *daddr;
1357 rt->rt6i_dst.plen = 128;
1da177e4 1358
85bd05de
DA		 1359 if (!rt6_is_gw_or_nonexthop(res)) {
1360 if (f6i->fib6_dst.plen != 128 &&
1361 ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
83a09abd 1362 rt->rt6i_flags |= RTF_ANYCAST;
1da177e4 1363#ifdef CONFIG_IPV6_SUBTREES
83a09abd
MKL		 1364 if (rt->rt6i_src.plen && saddr) {
1365 rt->rt6i_src.addr = *saddr;
1366 rt->rt6i_src.plen = 128;
8b9df265 1367 }
83a09abd 1368#endif
95a9a5ba 1369 }
1da177e4 1370
95a9a5ba
YH		 1371 return rt;
1372}
1da177e4 1373
db3fedee 1374static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
d52d3997 1375{
db3fedee
DA		 1376 struct fib6_info *f6i = res->f6i;
1377 unsigned short flags = fib6_info_dst_flags(f6i);
4832c30d 1378 struct net_device *dev;
d52d3997
MKL		 1379 struct rt6_info *pcpu_rt;
1380
db3fedee 1381 if (!fib6_info_hold_safe(f6i))
e873e4b9
WW		 1382 return NULL;
1383
4832c30d 1384 rcu_read_lock();
0d161581 1385 dev = ip6_rt_get_dev_rcu(res);
d8882935 1386 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT);
4832c30d 1387 rcu_read_unlock();
e873e4b9 1388 if (!pcpu_rt) {
db3fedee 1389 fib6_info_release(f6i);
d52d3997 1390 return NULL;
e873e4b9 1391 }
0d161581 1392 ip6_rt_copy_init(pcpu_rt, res);
d52d3997 1393 pcpu_rt->rt6i_flags |= RTF_PCPU;
8f34e53b
DA		 1394
1395 if (f6i->nh)
1396 pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1397
d52d3997
MKL		 1398 return pcpu_rt;
1399}
1400
8f34e53b
DA		 1401static bool rt6_is_valid(const struct rt6_info *rt6)
1402{
1403 return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1404}
1405
66f5d6ce 1406/* It should be called with rcu_read_lock() acquired */
db3fedee 1407static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
d52d3997 1408{
c353071a 1409 struct rt6_info *pcpu_rt;
d52d3997 1410
c353071a 1411 pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
d52d3997 1412
8f34e53b
DA		 1413 if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1414 struct rt6_info *prev, **p;
1415
1416 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1417 prev = xchg(p, NULL);
1418 if (prev) {
1419 dst_dev_put(&prev->dst);
1420 dst_release(&prev->dst);
1421 }
1422
1423 pcpu_rt = NULL;
1424 }
1425
a73e4195
MKL		 1426 return pcpu_rt;
1427}
1428
afb1d4b5 1429static struct rt6_info *rt6_make_pcpu_route(struct net *net,
db3fedee 1430 const struct fib6_result *res)
a73e4195
MKL		 1431{
1432 struct rt6_info *pcpu_rt, *prev, **p;
d52d3997 1433
db3fedee 1434 pcpu_rt = ip6_rt_pcpu_alloc(res);
0e09edcc
WW		 1435 if (!pcpu_rt)
1436 return NULL;
d52d3997 1437
f40b6ae2 1438 p = this_cpu_ptr(res->nh->rt6i_pcpu);
a94b9367 1439 prev = cmpxchg(p, NULL, pcpu_rt);
951f788a 1440 BUG_ON(prev);
a94b9367 1441
61fb0d01
ED		 1442 if (res->f6i->fib6_destroying) {
1443 struct fib6_info *from;
1444
1445 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1446 fib6_info_release(from);
1447 }
1448
d52d3997
MKL		 1449 return pcpu_rt;
1450}
1451
35732d01
WW		 1452/* exception hash table implementation
1453 */
1454static DEFINE_SPINLOCK(rt6_exception_lock);
1455
1456/* Remove rt6_ex from hash table and free the memory
1457 * Caller must hold rt6_exception_lock
1458 */
1459static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1460 struct rt6_exception *rt6_ex)
1461{
f5b51fe8 1462 struct fib6_info *from;
b2427e67 1463 struct net *net;
81eb8447 1464
35732d01
WW		 1465 if (!bucket || !rt6_ex)
1466 return;
b2427e67
CIK		 1467
1468 net = dev_net(rt6_ex->rt6i->dst.dev);
f5b51fe8
PA		 1469 net->ipv6.rt6_stats->fib_rt_cache--;
1470
1471 /* purge completely the exception to allow releasing the held resources:
1472 * some [sk] cache may keep the dst around for unlimited time
1473 */
0e233874 1474 from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
f5b51fe8
PA		 1475 fib6_info_release(from);
1476 dst_dev_put(&rt6_ex->rt6i->dst);
1477
35732d01 1478 hlist_del_rcu(&rt6_ex->hlist);
77634cc6 1479 dst_release(&rt6_ex->rt6i->dst);
35732d01
WW		 1480 kfree_rcu(rt6_ex, rcu);
1481 WARN_ON_ONCE(!bucket->depth);
1482 bucket->depth--;
1483}
1484
1485/* Remove oldest rt6_ex in bucket and free the memory
1486 * Caller must hold rt6_exception_lock
1487 */
1488static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1489{
1490 struct rt6_exception *rt6_ex, *oldest = NULL;
1491
1492 if (!bucket)
1493 return;
1494
1495 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1496 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1497 oldest = rt6_ex;
1498 }
1499 rt6_remove_exception(bucket, oldest);
1500}
1501
1502static u32 rt6_exception_hash(const struct in6_addr *dst,
1503 const struct in6_addr *src)
1504{
1505 static u32 seed __read_mostly;
1506 u32 val;
1507
1508 net_get_random_once(&seed, sizeof(seed));
b6b556af 1509 val = jhash2((const u32 *)dst, sizeof(*dst)/sizeof(u32), seed);
35732d01
WW		 1510
1511#ifdef CONFIG_IPV6_SUBTREES
1512 if (src)
b6b556af 1513 val = jhash2((const u32 *)src, sizeof(*src)/sizeof(u32), val);
35732d01
WW		 1514#endif
1515 return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1516}
1517
1518/* Helper function to find the cached rt in the hash table
1519 * and update bucket pointer to point to the bucket for this
1520 * (daddr, saddr) pair
1521 * Caller must hold rt6_exception_lock
1522 */
1523static struct rt6_exception *
1524__rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1525 const struct in6_addr *daddr,
1526 const struct in6_addr *saddr)
1527{
1528 struct rt6_exception *rt6_ex;
1529 u32 hval;
1530
1531 if (!(*bucket) || !daddr)
1532 return NULL;
1533
1534 hval = rt6_exception_hash(daddr, saddr);
1535 *bucket += hval;
1536
1537 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1538 struct rt6_info *rt6 = rt6_ex->rt6i;
1539 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1540
1541#ifdef CONFIG_IPV6_SUBTREES
1542 if (matched && saddr)
1543 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1544#endif
1545 if (matched)
1546 return rt6_ex;
1547 }
1548 return NULL;
1549}
1550
1551/* Helper function to find the cached rt in the hash table
1552 * and update bucket pointer to point to the bucket for this
1553 * (daddr, saddr) pair
1554 * Caller must hold rcu_read_lock()
1555 */
1556static struct rt6_exception *
1557__rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1558 const struct in6_addr *daddr,
1559 const struct in6_addr *saddr)
1560{
1561 struct rt6_exception *rt6_ex;
1562 u32 hval;
1563
1564 WARN_ON_ONCE(!rcu_read_lock_held());
1565
1566 if (!(*bucket) || !daddr)
1567 return NULL;
1568
1569 hval = rt6_exception_hash(daddr, saddr);
1570 *bucket += hval;
1571
1572 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1573 struct rt6_info *rt6 = rt6_ex->rt6i;
1574 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1575
1576#ifdef CONFIG_IPV6_SUBTREES
1577 if (matched && saddr)
1578 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1579#endif
1580 if (matched)
1581 return rt6_ex;
1582 }
1583 return NULL;
1584}
1585
b748f260 1586static unsigned int fib6_mtu(const struct fib6_result *res)
d4ead6b3 1587{
b748f260 1588 const struct fib6_nh *nh = res->nh;
d4ead6b3
DA		 1589 unsigned int mtu;
1590
b748f260
DA		 1591 if (res->f6i->fib6_pmtu) {
1592 mtu = res->f6i->fib6_pmtu;
dcd1f572 1593 } else {
b748f260 1594 struct net_device *dev = nh->fib_nh_dev;
dcd1f572
DA		 1595 struct inet6_dev *idev;
1596
1597 rcu_read_lock();
1598 idev = __in6_dev_get(dev);
1599 mtu = idev->cnf.mtu6;
1600 rcu_read_unlock();
1601 }
1602
d4ead6b3
DA		 1603 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1604
b748f260 1605 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
d4ead6b3
DA		 1606}
1607
cc5c073a
DA		 1608#define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1609
1610/* used when the flushed bit is not relevant, only access to the bucket
1611 * (ie., all bucket users except rt6_insert_exception);
1612 *
1613 * called under rcu lock; sometimes called with rt6_exception_lock held
1614 */
1615static
1616struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1617 spinlock_t *lock)
1618{
1619 struct rt6_exception_bucket *bucket;
1620
1621 if (lock)
1622 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1623 lockdep_is_held(lock));
1624 else
1625 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1626
1627 /* remove bucket flushed bit if set */
1628 if (bucket) {
1629 unsigned long p = (unsigned long)bucket;
1630
1631 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1632 bucket = (struct rt6_exception_bucket *)p;
1633 }
1634
1635 return bucket;
1636}
1637
1638static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1639{
1640 unsigned long p = (unsigned long)bucket;
1641
1642 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1643}
1644
1645/* called with rt6_exception_lock held */
1646static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1647 spinlock_t *lock)
1648{
1649 struct rt6_exception_bucket *bucket;
1650 unsigned long p;
1651
1652 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1653 lockdep_is_held(lock));
1654
1655 p = (unsigned long)bucket;
1656 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1657 bucket = (struct rt6_exception_bucket *)p;
1658 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1659}
1660
35732d01 1661static int rt6_insert_exception(struct rt6_info *nrt,
5012f0a5 1662 const struct fib6_result *res)
35732d01 1663{
5e670d84 1664 struct net *net = dev_net(nrt->dst.dev);
35732d01 1665 struct rt6_exception_bucket *bucket;
cc5c073a 1666 struct fib6_info *f6i = res->f6i;
35732d01
WW		 1667 struct in6_addr *src_key = NULL;
1668 struct rt6_exception *rt6_ex;
cc5c073a 1669 struct fib6_nh *nh = res->nh;
35732d01
WW		 1670 int err = 0;
1671
35732d01
WW		 1672 spin_lock_bh(&rt6_exception_lock);
1673
cc5c073a
DA		 1674 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1675 lockdep_is_held(&rt6_exception_lock));
35732d01
WW		 1676 if (!bucket) {
1677 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1678 GFP_ATOMIC);
1679 if (!bucket) {
1680 err = -ENOMEM;
1681 goto out;
1682 }
cc5c073a
DA		 1683 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1684 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1685 err = -EINVAL;
1686 goto out;
35732d01
WW		 1687 }
1688
1689#ifdef CONFIG_IPV6_SUBTREES
5012f0a5 1690 /* fib6_src.plen != 0 indicates f6i is in subtree
35732d01 1691 * and exception table is indexed by a hash of
5012f0a5 1692 * both fib6_dst and fib6_src.
35732d01 1693 * Otherwise, the exception table is indexed by
5012f0a5 1694 * a hash of only fib6_dst.
35732d01 1695 */
5012f0a5 1696 if (f6i->fib6_src.plen)
35732d01
WW		 1697 src_key = &nrt->rt6i_src.addr;
1698#endif
5012f0a5 1699 /* rt6_mtu_change() might lower mtu on f6i.
f5bbe7ee 1700 * Only insert this exception route if its mtu
5012f0a5 1701 * is less than f6i's mtu value.
f5bbe7ee 1702 */
b748f260 1703 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
f5bbe7ee
WW		 1704 err = -EINVAL;
1705 goto out;
1706 }
60006a48 1707
35732d01
WW		 1708 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1709 src_key);
1710 if (rt6_ex)
1711 rt6_remove_exception(bucket, rt6_ex);
1712
1713 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1714 if (!rt6_ex) {
1715 err = -ENOMEM;
1716 goto out;
1717 }
1718 rt6_ex->rt6i = nrt;
1719 rt6_ex->stamp = jiffies;
35732d01
WW		 1720 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1721 bucket->depth++;
81eb8447 1722 net->ipv6.rt6_stats->fib_rt_cache++;
35732d01
WW		 1723
1724 if (bucket->depth > FIB6_MAX_DEPTH)
1725 rt6_exception_remove_oldest(bucket);
1726
1727out:
1728 spin_unlock_bh(&rt6_exception_lock);
1729
1730 /* Update fn->fn_sernum to invalidate all cached dst */
b886d5f2 1731 if (!err) {
5012f0a5
DA		 1732 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1733 fib6_update_sernum(net, f6i);
1734 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
b886d5f2
PA		 1735 fib6_force_start_gc(net);
1736 }
35732d01
WW		 1737
1738 return err;
1739}
1740
c0b220cf 1741static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
35732d01
WW		 1742{
1743 struct rt6_exception_bucket *bucket;
1744 struct rt6_exception *rt6_ex;
1745 struct hlist_node *tmp;
1746 int i;
1747
1748 spin_lock_bh(&rt6_exception_lock);
35732d01 1749
cc5c073a 1750 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
35732d01
WW		 1751 if (!bucket)
1752 goto out;
1753
cc5c073a
DA		 1754 /* Prevent rt6_insert_exception() to recreate the bucket list */
1755 if (!from)
1756 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1757
35732d01 1758 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
cc5c073a
DA		 1759 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1760 if (!from ||
1761 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1762 rt6_remove_exception(bucket, rt6_ex);
1763 }
1764 WARN_ON_ONCE(!from && bucket->depth);
35732d01
WW		 1765 bucket++;
1766 }
35732d01
WW		 1767out:
1768 spin_unlock_bh(&rt6_exception_lock);
1769}
1770
e659ba31
DA		 1771static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1772{
1773 struct fib6_info *f6i = arg;
1774
1775 fib6_nh_flush_exceptions(nh, f6i);
1776
1777 return 0;
1778}
1779
c0b220cf
DA		 1780void rt6_flush_exceptions(struct fib6_info *f6i)
1781{
e659ba31
DA		 1782 if (f6i->nh)
1783 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1784 f6i);
1785 else
1786 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
c0b220cf
DA		 1787}
1788
35732d01
WW		 1789/* Find cached rt in the hash table inside passed in rt
1790 * Caller has to hold rcu_read_lock()
1791 */
7e4b5128 1792static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
510e2ced
WW		 1793 const struct in6_addr *daddr,
1794 const struct in6_addr *saddr)
35732d01 1795{
510e2ced 1796 const struct in6_addr *src_key = NULL;
35732d01 1797 struct rt6_exception_bucket *bucket;
35732d01 1798 struct rt6_exception *rt6_ex;
7e4b5128 1799 struct rt6_info *ret = NULL;
35732d01 1800
35732d01 1801#ifdef CONFIG_IPV6_SUBTREES
7e4b5128 1802 /* fib6i_src.plen != 0 indicates f6i is in subtree
35732d01 1803 * and exception table is indexed by a hash of
7e4b5128 1804 * both fib6_dst and fib6_src.
510e2ced
WW		 1805 * However, the src addr used to create the hash
1806 * might not be exactly the passed in saddr which
1807 * is a /128 addr from the flow.
1808 * So we need to use f6i->fib6_src to redo lookup
1809 * if the passed in saddr does not find anything.
1810 * (See the logic in ip6_rt_cache_alloc() on how
1811 * rt->rt6i_src is updated.)
35732d01 1812 */
7e4b5128 1813 if (res->f6i->fib6_src.plen)
35732d01 1814 src_key = saddr;
510e2ced 1815find_ex:
35732d01 1816#endif
cc5c073a 1817 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
35732d01
WW		 1818 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1819
1820 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
7e4b5128 1821 ret = rt6_ex->rt6i;
35732d01 1822
510e2ced
WW		 1823#ifdef CONFIG_IPV6_SUBTREES
1824 /* Use fib6_src as src_key and redo lookup */
1825 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1826 src_key = &res->f6i->fib6_src.addr;
1827 goto find_ex;
1828 }
1829#endif
1830
7e4b5128 1831 return ret;
35732d01
WW		 1832}
1833
1834/* Remove the passed in cached rt from the hash table that contains it */
cc5c073a 1835static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1836 const struct rt6_info *rt)
35732d01 1837{
c0b220cf 1838 const struct in6_addr *src_key = NULL;
35732d01 1839 struct rt6_exception_bucket *bucket;
35732d01
WW		 1840 struct rt6_exception *rt6_ex;
1841 int err;
1842
cc5c073a 1843 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
35732d01
WW		 1844 return -ENOENT;
1845
1846 spin_lock_bh(&rt6_exception_lock);
cc5c073a
DA		 1847 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1848
35732d01 1849#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1850 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1851 * and exception table is indexed by a hash of
1852 * both rt6i_dst and rt6i_src.
35732d01
WW		 1853 * Otherwise, the exception table is indexed by
1854 * a hash of only rt6i_dst.
1855 */
c0b220cf 1856 if (plen)
35732d01
WW		 1857 src_key = &rt->rt6i_src.addr;
1858#endif
1859 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1860 &rt->rt6i_dst.addr,
1861 src_key);
1862 if (rt6_ex) {
1863 rt6_remove_exception(bucket, rt6_ex);
1864 err = 0;
1865 } else {
1866 err = -ENOENT;
1867 }
1868
1869 spin_unlock_bh(&rt6_exception_lock);
1870 return err;
1871}
1872
e659ba31
DA		 1873struct fib6_nh_excptn_arg {
1874 struct rt6_info *rt;
1875 int plen;
1876};
1877
1878static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1879{
1880 struct fib6_nh_excptn_arg *arg = _arg;
1881 int err;
1882
1883 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1884 if (err == 0)
1885 return 1;
1886
1887 return 0;
1888}
1889
c0b220cf
DA		 1890static int rt6_remove_exception_rt(struct rt6_info *rt)
1891{
1892 struct fib6_info *from;
1893
1894 from = rcu_dereference(rt->from);
cc5c073a 1895 if (!from || !(rt->rt6i_flags & RTF_CACHE))
c0b220cf
DA		 1896 return -EINVAL;
1897
e659ba31
DA		 1898 if (from->nh) {
1899 struct fib6_nh_excptn_arg arg = {
1900 .rt = rt,
1901 .plen = from->fib6_src.plen
1902 };
1903 int rc;
1904
1905 /* rc = 1 means an entry was found */
1906 rc = nexthop_for_each_fib6_nh(from->nh,
1907 rt6_nh_remove_exception_rt,
1908 &arg);
1909 return rc ? 0 : -ENOENT;
1910 }
1911
1cf844c7 1912 return fib6_nh_remove_exception(from->fib6_nh,
cc5c073a 1913 from->fib6_src.plen, rt);
c0b220cf
DA		 1914}
1915
35732d01
WW		 1916/* Find rt6_ex which contains the passed in rt cache and
1917 * refresh its stamp
1918 */
cc5c073a 1919static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1920 const struct rt6_info *rt)
35732d01 1921{
c0b220cf 1922 const struct in6_addr *src_key = NULL;
35732d01 1923 struct rt6_exception_bucket *bucket;
35732d01 1924 struct rt6_exception *rt6_ex;
193f3685 1925
cc5c073a 1926 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
35732d01 1927#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1928 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1929 * and exception table is indexed by a hash of
1930 * both rt6i_dst and rt6i_src.
35732d01
WW		 1931 * Otherwise, the exception table is indexed by
1932 * a hash of only rt6i_dst.
1933 */
c0b220cf 1934 if (plen)
35732d01
WW		 1935 src_key = &rt->rt6i_src.addr;
1936#endif
cc5c073a 1937 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
35732d01
WW		 1938 if (rt6_ex)
1939 rt6_ex->stamp = jiffies;
c0b220cf
DA		 1940}
1941
e659ba31
DA		 1942struct fib6_nh_match_arg {
1943 const struct net_device *dev;
1944 const struct in6_addr *gw;
1945 struct fib6_nh *match;
1946};
1947
1948/* determine if fib6_nh has given device and gateway */
1949static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1950{
1951 struct fib6_nh_match_arg *arg = _arg;
1952
1953 if (arg->dev != nh->fib_nh_dev ||
1954 (arg->gw && !nh->fib_nh_gw_family) ||
1955 (!arg->gw && nh->fib_nh_gw_family) ||
1956 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1957 return 0;
1958
1959 arg->match = nh;
1960
1961 /* found a match, break the loop */
1962 return 1;
1963}
1964
c0b220cf
DA		 1965static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1966{
1967 struct fib6_info *from;
e659ba31 1968 struct fib6_nh *fib6_nh;
35732d01 1969
c0b220cf
DA		 1970 rcu_read_lock();
1971
1972 from = rcu_dereference(rt->from);
1973 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1974 goto unlock;
1975
e659ba31
DA		 1976 if (from->nh) {
1977 struct fib6_nh_match_arg arg = {
1978 .dev = rt->dst.dev,
1979 .gw = &rt->rt6i_gateway,
1980 };
1981
1982 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1983
1984 if (!arg.match)
cff6a327 1985 goto unlock;
e659ba31
DA		 1986 fib6_nh = arg.match;
1987 } else {
1988 fib6_nh = from->fib6_nh;
1989 }
1990 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
193f3685 1991unlock:
35732d01
WW		 1992 rcu_read_unlock();
1993}
1994
e9fa1495
SB		 1995static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1996 struct rt6_info *rt, int mtu)
1997{
1998 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1999 * lowest MTU in the path: always allow updating the route PMTU to
2000 * reflect PMTU decreases.
2001 *
2002 * If the new MTU is higher, and the route PMTU is equal to the local
2003 * MTU, this means the old MTU is the lowest in the path, so allow
2004 * updating it: if other nodes now have lower MTUs, PMTU discovery will
2005 * handle this.
2006 */
2007
2008 if (dst_mtu(&rt->dst) >= mtu)
2009 return true;
2010
2011 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
2012 return true;
2013
2014 return false;
2015}
2016
2017static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
cc5c073a 2018 const struct fib6_nh *nh, int mtu)
f5bbe7ee
WW		 2019{
2020 struct rt6_exception_bucket *bucket;
2021 struct rt6_exception *rt6_ex;
2022 int i;
2023
cc5c073a 2024 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
e9fa1495
SB		 2025 if (!bucket)
2026 return;
2027
2028 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2029 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2030 struct rt6_info *entry = rt6_ex->rt6i;
2031
2032 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
d4ead6b3 2033 * route), the metrics of its rt->from have already
e9fa1495
SB		 2034 * been updated.
2035 */
d4ead6b3 2036 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
e9fa1495 2037 rt6_mtu_change_route_allowed(idev, entry, mtu))
d4ead6b3 2038 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
f5bbe7ee 2039 }
e9fa1495 2040 bucket++;
f5bbe7ee
WW		 2041 }
2042}
2043
b16cb459
WW		 2044#define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2045
cc5c073a
DA		 2046static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2047 const struct in6_addr *gateway)
b16cb459
WW		 2048{
2049 struct rt6_exception_bucket *bucket;
2050 struct rt6_exception *rt6_ex;
2051 struct hlist_node *tmp;
2052 int i;
2053
cc5c073a 2054 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
b16cb459
WW		 2055 return;
2056
2057 spin_lock_bh(&rt6_exception_lock);
cc5c073a 2058 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
b16cb459
WW		 2059 if (bucket) {
2060 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2061 hlist_for_each_entry_safe(rt6_ex, tmp,
2062 &bucket->chain, hlist) {
2063 struct rt6_info *entry = rt6_ex->rt6i;
2064
2065 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2066 RTF_CACHE_GATEWAY &&
2067 ipv6_addr_equal(gateway,
2068 &entry->rt6i_gateway)) {
2069 rt6_remove_exception(bucket, rt6_ex);
2070 }
2071 }
2072 bucket++;
2073 }
2074 }
2075
2076 spin_unlock_bh(&rt6_exception_lock);
2077}
2078
c757faa8
WW		 2079static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2080 struct rt6_exception *rt6_ex,
2081 struct fib6_gc_args *gc_args,
2082 unsigned long now)
2083{
2084 struct rt6_info *rt = rt6_ex->rt6i;
2085
1859bac0
PA		 2086 /* we are pruning and obsoleting aged-out and non gateway exceptions
2087 * even if others have still references to them, so that on next
2088 * dst_check() such references can be dropped.
2089 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2090 * expired, independently from their aging, as per RFC 8201 section 4
2091 */
31afeb42
WW		 2092 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2093 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2094 RT6_TRACE("aging clone %p\n", rt);
2095 rt6_remove_exception(bucket, rt6_ex);
2096 return;
2097 }
2098 } else if (time_after(jiffies, rt->dst.expires)) {
2099 RT6_TRACE("purging expired route %p\n", rt);
c757faa8
WW		 2100 rt6_remove_exception(bucket, rt6_ex);
2101 return;
31afeb42
WW		 2102 }
2103
2104 if (rt->rt6i_flags & RTF_GATEWAY) {
c757faa8
WW		 2105 struct neighbour *neigh;
2106 __u8 neigh_flags = 0;
2107
1bfa26ff
ED		 2108 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2109 if (neigh)
c757faa8 2110 neigh_flags = neigh->flags;
1bfa26ff 2111
c757faa8
WW		 2112 if (!(neigh_flags & NTF_ROUTER)) {
2113 RT6_TRACE("purging route %p via non-router but gateway\n",
2114 rt);
2115 rt6_remove_exception(bucket, rt6_ex);
2116 return;
2117 }
2118 }
31afeb42 2119
c757faa8
WW		 2120 gc_args->more++;
2121}
2122
cc5c073a 2123static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
c0b220cf
DA		 2124 struct fib6_gc_args *gc_args,
2125 unsigned long now)
c757faa8
WW		 2126{
2127 struct rt6_exception_bucket *bucket;
2128 struct rt6_exception *rt6_ex;
2129 struct hlist_node *tmp;
2130 int i;
2131
cc5c073a 2132 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
c757faa8
WW		 2133 return;
2134
1bfa26ff
ED		 2135 rcu_read_lock_bh();
2136 spin_lock(&rt6_exception_lock);
cc5c073a 2137 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
c757faa8
WW		 2138 if (bucket) {
2139 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2140 hlist_for_each_entry_safe(rt6_ex, tmp,
2141 &bucket->chain, hlist) {
2142 rt6_age_examine_exception(bucket, rt6_ex,
2143 gc_args, now);
2144 }
2145 bucket++;
2146 }
2147 }
1bfa26ff
ED		 2148 spin_unlock(&rt6_exception_lock);
2149 rcu_read_unlock_bh();
c757faa8
WW		 2150}
2151
e659ba31
DA		 2152struct fib6_nh_age_excptn_arg {
2153 struct fib6_gc_args *gc_args;
2154 unsigned long now;
2155};
2156
2157static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2158{
2159 struct fib6_nh_age_excptn_arg *arg = _arg;
2160
2161 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2162 return 0;
2163}
2164
cc5c073a 2165void rt6_age_exceptions(struct fib6_info *f6i,
c0b220cf
DA		 2166 struct fib6_gc_args *gc_args,
2167 unsigned long now)
2168{
e659ba31
DA		 2169 if (f6i->nh) {
2170 struct fib6_nh_age_excptn_arg arg = {
2171 .gc_args = gc_args,
2172 .now = now
2173 };
2174
2175 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2176 &arg);
2177 } else {
2178 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2179 }
c0b220cf
DA		 2180}
2181
1d053da9 2182/* must be called with rcu lock held */
effda4dd
DA		 2183int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2184 struct flowi6 *fl6, struct fib6_result *res, int strict)
1da177e4 2185{
367efcb9 2186 struct fib6_node *fn, *saved_fn;
1da177e4 2187
6454743b 2188 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
367efcb9 2189 saved_fn = fn;
1da177e4 2190
ca254490
DA		 2191 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2192 oif = 0;
2193
a3c00e46 2194redo_rt6_select:
effda4dd
DA		 2195 rt6_select(net, fn, oif, res, strict);
2196 if (res->f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2197 fn = fib6_backtrack(fn, &fl6->saddr);
2198 if (fn)
2199 goto redo_rt6_select;
367efcb9
MKL		 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;
367efcb9 2205 }
a3c00e46
MKL		 2206 }
2207
effda4dd 2208 trace_fib6_table_lookup(net, res, table, fl6);
fb9de91e 2209
effda4dd 2210 return 0;
1d053da9
DA		 2211}
2212
2213struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2214 int oif, struct flowi6 *fl6,
2215 const struct sk_buff *skb, int flags)
2216{
b1d40991 2217 struct fib6_result res = {};
0e09edcc 2218 struct rt6_info *rt = NULL;
1d053da9
DA		 2219 int strict = 0;
2220
0e09edcc
WW		 2221 WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2222 !rcu_read_lock_held());
2223
1d053da9
DA		 2224 strict |= flags & RT6_LOOKUP_F_IFACE;
2225 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2226 if (net->ipv6.devconf_all->forwarding == 0)
2227 strict |= RT6_LOOKUP_F_REACHABLE;
2228
2229 rcu_read_lock();
2230
effda4dd 2231 fib6_table_lookup(net, table, oif, fl6, &res, strict);
0e09edcc
WW		 2232 if (res.f6i == net->ipv6.fib6_null_entry)
2233 goto out;
23fb93a4 2234
b1d40991 2235 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
d83009d4 2236
23fb93a4 2237 /*Search through exception table */
7e4b5128 2238 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
23fb93a4 2239 if (rt) {
0e09edcc 2240 goto out;
3da59bd9 2241 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
b1d40991 2242 !res.nh->fib_nh_gw_family)) {
3da59bd9
MKL		 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 */
0e09edcc 2248 rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
c71099ac 2249
0e09edcc
WW		 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.
1cfb71ee 2255 */
0e09edcc 2256 rt6_uncached_list_add(rt);
81eb8447 2257 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
0e09edcc 2258 rcu_read_unlock();
b811580d 2259
0e09edcc
WW		 2260 return rt;
2261 }
d52d3997
MKL		 2262 } else {
2263 /* Get a percpu copy */
951f788a 2264 local_bh_disable();
0e09edcc 2265 rt = rt6_get_pcpu_route(&res);
d52d3997 2266
0e09edcc
WW		 2267 if (!rt)
2268 rt = rt6_make_pcpu_route(net, &res);
93531c67 2269
951f788a 2270 local_bh_enable();
d52d3997 2271 }
0e09edcc
WW		 2272out:
2273 if (!rt)
2274 rt = net->ipv6.ip6_null_entry;
2275 if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2276 ip6_hold_safe(net, &rt);
2277 rcu_read_unlock();
2278
2279 return rt;
1da177e4 2280}
9ff74384 2281EXPORT_SYMBOL_GPL(ip6_pol_route);
1da177e4 2282
55cced4f 2283INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_input(struct net *net,
b75cc8f9
DA		 2284 struct fib6_table *table,
2285 struct flowi6 *fl6,
2286 const struct sk_buff *skb,
2287 int flags)
4acad72d 2288{
b75cc8f9 2289 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
4acad72d
PE		 2290}
2291
d409b847
MB		 2292struct dst_entry *ip6_route_input_lookup(struct net *net,
2293 struct net_device *dev,
b75cc8f9
DA		 2294 struct flowi6 *fl6,
2295 const struct sk_buff *skb,
2296 int flags)
72331bc0
SL		 2297{
2298 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2299 flags |= RT6_LOOKUP_F_IFACE;
2300
b75cc8f9 2301 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
72331bc0 2302}
d409b847 2303EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
72331bc0 2304
23aebdac 2305static void ip6_multipath_l3_keys(const struct sk_buff *skb,
5e5d6fed
RP		 2306 struct flow_keys *keys,
2307 struct flow_keys *flkeys)
23aebdac
JS		 2308{
2309 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2310 const struct ipv6hdr *key_iph = outer_iph;
5e5d6fed 2311 struct flow_keys *_flkeys = flkeys;
23aebdac
JS		 2312 const struct ipv6hdr *inner_iph;
2313 const struct icmp6hdr *icmph;
2314 struct ipv6hdr _inner_iph;
cea67a2d 2315 struct icmp6hdr _icmph;
23aebdac
JS		 2316
2317 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2318 goto out;
2319
cea67a2d
ED		 2320 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2321 sizeof(_icmph), &_icmph);
2322 if (!icmph)
2323 goto out;
2324
54074f1d 2325 if (!icmpv6_is_err(icmph->icmp6_type))
23aebdac
JS		 2326 goto out;
2327
2328 inner_iph = skb_header_pointer(skb,
2329 skb_transport_offset(skb) + sizeof(*icmph),
2330 sizeof(_inner_iph), &_inner_iph);
2331 if (!inner_iph)
2332 goto out;
2333
2334 key_iph = inner_iph;
5e5d6fed 2335 _flkeys = NULL;
23aebdac 2336out:
5e5d6fed
RP		 2337 if (_flkeys) {
2338 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2339 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2340 keys->tags.flow_label = _flkeys->tags.flow_label;
2341 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2342 } else {
2343 keys->addrs.v6addrs.src = key_iph->saddr;
2344 keys->addrs.v6addrs.dst = key_iph->daddr;
fa1be7e0 2345 keys->tags.flow_label = ip6_flowlabel(key_iph);
5e5d6fed
RP		 2346 keys->basic.ip_proto = key_iph->nexthdr;
2347 }
23aebdac
JS		 2348}
2349
2350/* if skb is set it will be used and fl6 can be NULL */
b4bac172
DA		 2351u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2352 const struct sk_buff *skb, struct flow_keys *flkeys)
23aebdac
JS		 2353{
2354 struct flow_keys hash_keys;
9a2a537a 2355 u32 mhash;
23aebdac 2356
bbfa047a 2357 switch (ip6_multipath_hash_policy(net)) {
b4bac172
DA		 2358 case 0:
2359 memset(&hash_keys, 0, sizeof(hash_keys));
2360 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2361 if (skb) {
2362 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2363 } else {
2364 hash_keys.addrs.v6addrs.src = fl6->saddr;
2365 hash_keys.addrs.v6addrs.dst = fl6->daddr;
fa1be7e0 2366 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
b4bac172
DA		 2367 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2368 }
2369 break;
2370 case 1:
2371 if (skb) {
2372 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2373 struct flow_keys keys;
2374
2375 /* short-circuit if we already have L4 hash present */
2376 if (skb->l4_hash)
2377 return skb_get_hash_raw(skb) >> 1;
2378
2379 memset(&hash_keys, 0, sizeof(hash_keys));
2380
2381 if (!flkeys) {
2382 skb_flow_dissect_flow_keys(skb, &keys, flag);
2383 flkeys = &keys;
2384 }
2385 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2386 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2387 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2388 hash_keys.ports.src = flkeys->ports.src;
2389 hash_keys.ports.dst = flkeys->ports.dst;
2390 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2391 } else {
2392 memset(&hash_keys, 0, sizeof(hash_keys));
2393 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2394 hash_keys.addrs.v6addrs.src = fl6->saddr;
2395 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2396 hash_keys.ports.src = fl6->fl6_sport;
2397 hash_keys.ports.dst = fl6->fl6_dport;
2398 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2399 }
2400 break;
d8f74f09
SS		 2401 case 2:
2402 memset(&hash_keys, 0, sizeof(hash_keys));
2403 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2404 if (skb) {
2405 struct flow_keys keys;
2406
2407 if (!flkeys) {
2408 skb_flow_dissect_flow_keys(skb, &keys, 0);
2409 flkeys = &keys;
2410 }
2411
2412 /* Inner can be v4 or v6 */
2413 if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2414 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2415 hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2416 hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2417 } else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2418 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2419 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2420 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2421 hash_keys.tags.flow_label = flkeys->tags.flow_label;
2422 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2423 } else {
2424 /* Same as case 0 */
2425 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2426 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2427 }
2428 } else {
2429 /* Same as case 0 */
2430 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2431 hash_keys.addrs.v6addrs.src = fl6->saddr;
2432 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2433 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2434 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2435 }
2436 break;
23aebdac 2437 }
9a2a537a 2438 mhash = flow_hash_from_keys(&hash_keys);
23aebdac 2439
9a2a537a 2440 return mhash >> 1;
23aebdac
JS		 2441}
2442
67f415dd 2443/* Called with rcu held */
c71099ac
TG		 2444void ip6_route_input(struct sk_buff *skb)
2445{
b71d1d42 2446 const struct ipv6hdr *iph = ipv6_hdr(skb);
c346dca1 2447 struct net *net = dev_net(skb->dev);
67f415dd 2448 int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
904af04d 2449 struct ip_tunnel_info *tun_info;
4c9483b2 2450 struct flowi6 fl6 = {
e0d56fdd 2451 .flowi6_iif = skb->dev->ifindex,
4c9483b2
DM		 2452 .daddr = iph->daddr,
2453 .saddr = iph->saddr,
6502ca52 2454 .flowlabel = ip6_flowinfo(iph),
4c9483b2
DM		 2455 .flowi6_mark = skb->mark,
2456 .flowi6_proto = iph->nexthdr,
c71099ac 2457 };
5e5d6fed 2458 struct flow_keys *flkeys = NULL, _flkeys;
adaa70bb 2459
904af04d 2460 tun_info = skb_tunnel_info(skb);
46fa062a 2461 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
904af04d 2462 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
5e5d6fed
RP		 2463
2464 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2465 flkeys = &_flkeys;
2466
23aebdac 2467 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
b4bac172 2468 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
06e9d040 2469 skb_dst_drop(skb);
67f415dd
WW		 2470 skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2471 &fl6, skb, flags));
c71099ac
TG		 2472}
2473
55cced4f 2474INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_output(struct net *net,
b75cc8f9
DA		 2475 struct fib6_table *table,
2476 struct flowi6 *fl6,
2477 const struct sk_buff *skb,
2478 int flags)
1da177e4 2479{
b75cc8f9 2480 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
c71099ac
TG		 2481}
2482
7d9e5f42
WW		 2483struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2484 const struct sock *sk,
2485 struct flowi6 *fl6, int flags)
c71099ac 2486{
d46a9d67 2487 bool any_src;
c71099ac 2488
3ede0bbc
RS		 2489 if (ipv6_addr_type(&fl6->daddr) &
2490 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
4c1feac5
DA		 2491 struct dst_entry *dst;
2492
7d9e5f42 2493 /* This function does not take refcnt on the dst */
4c1feac5
DA		 2494 dst = l3mdev_link_scope_lookup(net, fl6);
2495 if (dst)
2496 return dst;
2497 }
ca254490 2498
1fb9489b 2499 fl6->flowi6_iif = LOOPBACK_IFINDEX;
4dc27d1c 2500
7d9e5f42 2501 flags |= RT6_LOOKUP_F_DST_NOREF;
d46a9d67 2502 any_src = ipv6_addr_any(&fl6->saddr);
741a11d9 2503 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
d46a9d67 2504 (fl6->flowi6_oif && any_src))
77d16f45 2505 flags |= RT6_LOOKUP_F_IFACE;
c71099ac 2506
d46a9d67 2507 if (!any_src)
adaa70bb 2508 flags |= RT6_LOOKUP_F_HAS_SADDR;
0c9a2ac1
YH		 2509 else if (sk)
2510 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
adaa70bb 2511
b75cc8f9 2512 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
1da177e4 2513}
7d9e5f42
WW		 2514EXPORT_SYMBOL_GPL(ip6_route_output_flags_noref);
2515
2516struct dst_entry *ip6_route_output_flags(struct net *net,
2517 const struct sock *sk,
2518 struct flowi6 *fl6,
2519 int flags)
2520{
2521 struct dst_entry *dst;
2522 struct rt6_info *rt6;
2523
2524 rcu_read_lock();
2525 dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2526 rt6 = (struct rt6_info *)dst;
2527 /* For dst cached in uncached_list, refcnt is already taken. */
2528 if (list_empty(&rt6->rt6i_uncached) && !dst_hold_safe(dst)) {
2529 dst = &net->ipv6.ip6_null_entry->dst;
2530 dst_hold(dst);
2531 }
2532 rcu_read_unlock();
2533
2534 return dst;
2535}
6f21c96a 2536EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1da177e4 2537
2774c131 2538struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
14e50e57 2539{
5c1e6aa3 2540 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1dbe3252 2541 struct net_device *loopback_dev = net->loopback_dev;
14e50e57
DM		 2542 struct dst_entry *new = NULL;
2543
1dbe3252 2544 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
62cf27e5 2545 DST_OBSOLETE_DEAD, 0);
14e50e57 2546 if (rt) {
0a1f5962 2547 rt6_info_init(rt);
81eb8447 2548 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
8104891b 2549
0a1f5962 2550 new = &rt->dst;
14e50e57 2551 new->__use = 1;
352e512c 2552 new->input = dst_discard;
ede2059d 2553 new->output = dst_discard_out;
14e50e57 2554
0a1f5962 2555 dst_copy_metrics(new, &ort->dst);
14e50e57 2556
1dbe3252 2557 rt->rt6i_idev = in6_dev_get(loopback_dev);
4e3fd7a0 2558 rt->rt6i_gateway = ort->rt6i_gateway;
0a1f5962 2559 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
14e50e57
DM		 2560
2561 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2562#ifdef CONFIG_IPV6_SUBTREES
2563 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2564#endif
14e50e57
DM		 2565 }
2566
69ead7af
DM		 2567 dst_release(dst_orig);
2568 return new ? new : ERR_PTR(-ENOMEM);
14e50e57 2569}
14e50e57 2570
1da177e4
LT		 2571/*
2572 * Destination cache support functions
2573 */
2574
8d1c802b 2575static bool fib6_check(struct fib6_info *f6i, u32 cookie)
4b32b5ad 2576{
93531c67
DA		 2577 u32 rt_cookie = 0;
2578
8ae86971 2579 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
93531c67
DA		 2580 return false;
2581
2582 if (fib6_check_expired(f6i))
2583 return false;
2584
2585 return true;
4b32b5ad
MKL		 2586}
2587
a68886a6
DA		 2588static struct dst_entry *rt6_check(struct rt6_info *rt,
2589 struct fib6_info *from,
2590 u32 cookie)
3da59bd9 2591{
36143645 2592 u32 rt_cookie = 0;
c5cff856 2593
49d05fe2 2594 if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
93531c67 2595 rt_cookie != cookie)
3da59bd9
MKL		 2596 return NULL;
2597
2598 if (rt6_check_expired(rt))
2599 return NULL;
2600
2601 return &rt->dst;
2602}
2603
a68886a6
DA		 2604static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2605 struct fib6_info *from,
2606 u32 cookie)
3da59bd9 2607{
5973fb1e
MKL		 2608 if (!__rt6_check_expired(rt) &&
2609 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
a68886a6 2610 fib6_check(from, cookie))
3da59bd9
MKL		 2611 return &rt->dst;
2612 else
2613 return NULL;
2614}
2615
bbd807df
BV		 2616INDIRECT_CALLABLE_SCOPE struct dst_entry *ip6_dst_check(struct dst_entry *dst,
2617 u32 cookie)
1da177e4 2618{
a87b7dc9 2619 struct dst_entry *dst_ret;
a68886a6 2620 struct fib6_info *from;
1da177e4
LT		 2621 struct rt6_info *rt;
2622
a87b7dc9
DA		 2623 rt = container_of(dst, struct rt6_info, dst);
2624
8f34e53b
DA		 2625 if (rt->sernum)
2626 return rt6_is_valid(rt) ? dst : NULL;
2627
a87b7dc9 2628 rcu_read_lock();
1da177e4 2629
6f3118b5
ND		 2630 /* All IPV6 dsts are created with ->obsolete set to the value
2631 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2632 * into this function always.
2633 */
e3bc10bd 2634
a68886a6 2635 from = rcu_dereference(rt->from);
4b32b5ad 2636
a68886a6
DA		 2637 if (from && (rt->rt6i_flags & RTF_PCPU ||
2638 unlikely(!list_empty(&rt->rt6i_uncached))))
2639 dst_ret = rt6_dst_from_check(rt, from, cookie);
3da59bd9 2640 else
a68886a6 2641 dst_ret = rt6_check(rt, from, cookie);
a87b7dc9
DA		 2642
2643 rcu_read_unlock();
2644
2645 return dst_ret;
1da177e4 2646}
bbd807df 2647EXPORT_SYMBOL(ip6_dst_check);
1da177e4
LT		 2648
2649static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2650{
2651 struct rt6_info *rt = (struct rt6_info *) dst;
2652
2653 if (rt) {
54c1a859 2654 if (rt->rt6i_flags & RTF_CACHE) {
c3c14da0 2655 rcu_read_lock();
54c1a859 2656 if (rt6_check_expired(rt)) {
93531c67 2657 rt6_remove_exception_rt(rt);
54c1a859
YH		 2658 dst = NULL;
2659 }
c3c14da0 2660 rcu_read_unlock();
54c1a859 2661 } else {
1da177e4 2662 dst_release(dst);
54c1a859
YH		 2663 dst = NULL;
2664 }
1da177e4 2665 }
54c1a859 2666 return dst;
1da177e4
LT		 2667}
2668
2669static void ip6_link_failure(struct sk_buff *skb)
2670{
2671 struct rt6_info *rt;
2672
3ffe533c 2673 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1da177e4 2674
adf30907 2675 rt = (struct rt6_info *) skb_dst(skb);
1da177e4 2676 if (rt) {
8a14e46f 2677 rcu_read_lock();
1eb4f758 2678 if (rt->rt6i_flags & RTF_CACHE) {
761f6026 2679 rt6_remove_exception_rt(rt);
c5cff856 2680 } else {
a68886a6 2681 struct fib6_info *from;
c5cff856
WW		 2682 struct fib6_node *fn;
2683
a68886a6
DA		 2684 from = rcu_dereference(rt->from);
2685 if (from) {
2686 fn = rcu_dereference(from->fib6_node);
2687 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2688 fn->fn_sernum = -1;
2689 }
1eb4f758 2690 }
8a14e46f 2691 rcu_read_unlock();
1da177e4
LT		 2692 }
2693}
2694
6a3e030f
DA		 2695static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2696{
a68886a6
DA		 2697 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2698 struct fib6_info *from;
2699
2700 rcu_read_lock();
2701 from = rcu_dereference(rt0->from);
2702 if (from)
2703 rt0->dst.expires = from->expires;
2704 rcu_read_unlock();
2705 }
6a3e030f
DA		 2706
2707 dst_set_expires(&rt0->dst, timeout);
2708 rt0->rt6i_flags |= RTF_EXPIRES;
2709}
2710
45e4fd26
MKL		 2711static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2712{
2713 struct net *net = dev_net(rt->dst.dev);
2714
d4ead6b3 2715 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
45e4fd26 2716 rt->rt6i_flags |= RTF_MODIFIED;
45e4fd26
MKL		 2717 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2718}
2719
0d3f6d29
MKL		 2720static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2721{
2722 return !(rt->rt6i_flags & RTF_CACHE) &&
1490ed2a 2723 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
0d3f6d29
MKL		 2724}
2725
45e4fd26 2726static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
bd085ef6
HL		 2727 const struct ipv6hdr *iph, u32 mtu,
2728 bool confirm_neigh)
1da177e4 2729{
0dec879f 2730 const struct in6_addr *daddr, *saddr;
67ba4152 2731 struct rt6_info *rt6 = (struct rt6_info *)dst;
1da177e4 2732
09454fd0
MŻ		 2733 /* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU)
2734 * IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it.
2735 * [see also comment in rt6_mtu_change_route()]
2736 */
19bda36c 2737
0dec879f
JA		 2738 if (iph) {
2739 daddr = &iph->daddr;
2740 saddr = &iph->saddr;
2741 } else if (sk) {
2742 daddr = &sk->sk_v6_daddr;
2743 saddr = &inet6_sk(sk)->saddr;
2744 } else {
2745 daddr = NULL;
2746 saddr = NULL;
2747 }
bd085ef6
HL		 2748
2749 if (confirm_neigh)
2750 dst_confirm_neigh(dst, daddr);
2751
4a65dff8
GK		 2752 if (mtu < IPV6_MIN_MTU)
2753 return;
45e4fd26
MKL		 2754 if (mtu >= dst_mtu(dst))
2755 return;
9d289715 2756
0d3f6d29 2757 if (!rt6_cache_allowed_for_pmtu(rt6)) {
45e4fd26 2758 rt6_do_update_pmtu(rt6, mtu);
2b760fcf
WW		 2759 /* update rt6_ex->stamp for cache */
2760 if (rt6->rt6i_flags & RTF_CACHE)
2761 rt6_update_exception_stamp_rt(rt6);
0dec879f 2762 } else if (daddr) {
85bd05de 2763 struct fib6_result res = {};
45e4fd26
MKL		 2764 struct rt6_info *nrt6;
2765
4d85cd0c 2766 rcu_read_lock();
85bd05de 2767 res.f6i = rcu_dereference(rt6->from);
43a4b60d
DA		 2768 if (!res.f6i)
2769 goto out_unlock;
2770
7d21fec9
DA		 2771 res.fib6_flags = res.f6i->fib6_flags;
2772 res.fib6_type = res.f6i->fib6_type;
2773
2d44234b
DA		 2774 if (res.f6i->nh) {
2775 struct fib6_nh_match_arg arg = {
2776 .dev = dst->dev,
2777 .gw = &rt6->rt6i_gateway,
2778 };
2779
2780 nexthop_for_each_fib6_nh(res.f6i->nh,
2781 fib6_nh_find_match, &arg);
2782
2783 /* fib6_info uses a nexthop that does not have fib6_nh
2784 * using the dst->dev + gw. Should be impossible.
2785 */
43a4b60d
DA		 2786 if (!arg.match)
2787 goto out_unlock;
2d44234b
DA		 2788
2789 res.nh = arg.match;
2790 } else {
2791 res.nh = res.f6i->fib6_nh;
2792 }
2793
85bd05de 2794 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
45e4fd26
MKL		 2795 if (nrt6) {
2796 rt6_do_update_pmtu(nrt6, mtu);
5012f0a5 2797 if (rt6_insert_exception(nrt6, &res))
2b760fcf 2798 dst_release_immediate(&nrt6->dst);
45e4fd26 2799 }
43a4b60d 2800out_unlock:
a68886a6 2801 rcu_read_unlock();
1da177e4
LT		 2802 }
2803}
2804
45e4fd26 2805static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
bd085ef6
HL		 2806 struct sk_buff *skb, u32 mtu,
2807 bool confirm_neigh)
45e4fd26 2808{
bd085ef6
HL		 2809 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2810 confirm_neigh);
45e4fd26
MKL		 2811}
2812
42ae66c8 2813void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
e2d118a1 2814 int oif, u32 mark, kuid_t uid)
81aded24
DM		 2815{
2816 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2817 struct dst_entry *dst;
dc92095d
MŻ		 2818 struct flowi6 fl6 = {
2819 .flowi6_oif = oif,
2820 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2821 .daddr = iph->daddr,
2822 .saddr = iph->saddr,
2823 .flowlabel = ip6_flowinfo(iph),
2824 .flowi6_uid = uid,
2825 };
81aded24
DM		 2826
2827 dst = ip6_route_output(net, NULL, &fl6);
2828 if (!dst->error)
bd085ef6 2829 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
81aded24
DM		 2830 dst_release(dst);
2831}
2832EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2833
2834void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2835{
7ddacfa5 2836 int oif = sk->sk_bound_dev_if;
33c162a9
MKL		 2837 struct dst_entry *dst;
2838
7ddacfa5
DA		 2839 if (!oif && skb->dev)
2840 oif = l3mdev_master_ifindex(skb->dev);
2841
2842 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
33c162a9
MKL		 2843
2844 dst = __sk_dst_get(sk);
2845 if (!dst || !dst->obsolete ||
2846 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2847 return;
2848
2849 bh_lock_sock(sk);
2850 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2851 ip6_datagram_dst_update(sk, false);
2852 bh_unlock_sock(sk);
81aded24
DM		 2853}
2854EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2855
7d6850f7
AK		 2856void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2857 const struct flowi6 *fl6)
2858{
2859#ifdef CONFIG_IPV6_SUBTREES
2860 struct ipv6_pinfo *np = inet6_sk(sk);
2861#endif
2862
2863 ip6_dst_store(sk, dst,
2864 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2865 &sk->sk_v6_daddr : NULL,
2866#ifdef CONFIG_IPV6_SUBTREES
2867 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2868 &np->saddr :
2869#endif
2870 NULL);
2871}
2872
9b6b35ab 2873static bool ip6_redirect_nh_match(const struct fib6_result *res,
0b34eb00
DA		 2874 struct flowi6 *fl6,
2875 const struct in6_addr *gw,
2876 struct rt6_info **ret)
2877{
9b6b35ab
DA		 2878 const struct fib6_nh *nh = res->nh;
2879
0b34eb00
DA		 2880 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2881 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2882 return false;
2883
2884 /* rt_cache's gateway might be different from its 'parent'
2885 * in the case of an ip redirect.
2886 * So we keep searching in the exception table if the gateway
2887 * is different.
2888 */
2889 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
2890 struct rt6_info *rt_cache;
2891
9b6b35ab 2892 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
0b34eb00
DA		 2893 if (rt_cache &&
2894 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
2895 *ret = rt_cache;
2896 return true;
2897 }
2898 return false;
2899 }
2900 return true;
2901}
2902
c55c8988
DA		 2903struct fib6_nh_rd_arg {
2904 struct fib6_result *res;
2905 struct flowi6 *fl6;
2906 const struct in6_addr *gw;
2907 struct rt6_info **ret;
2908};
2909
2910static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
2911{
2912 struct fib6_nh_rd_arg *arg = _arg;
2913
2914 arg->res->nh = nh;
2915 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
2916}
2917
b55b76b2
DJ		 2918/* Handle redirects */
2919struct ip6rd_flowi {
2920 struct flowi6 fl6;
2921 struct in6_addr gateway;
2922};
2923
55cced4f 2924INDIRECT_CALLABLE_SCOPE struct rt6_info *__ip6_route_redirect(struct net *net,
b55b76b2
DJ		 2925 struct fib6_table *table,
2926 struct flowi6 *fl6,
b75cc8f9 2927 const struct sk_buff *skb,
b55b76b2
DJ		 2928 int flags)
2929{
2930 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
0b34eb00 2931 struct rt6_info *ret = NULL;
9b6b35ab 2932 struct fib6_result res = {};
c55c8988
DA		 2933 struct fib6_nh_rd_arg arg = {
2934 .res = &res,
2935 .fl6 = fl6,
2936 .gw = &rdfl->gateway,
2937 .ret = &ret
2938 };
8d1c802b 2939 struct fib6_info *rt;
b55b76b2
DJ		 2940 struct fib6_node *fn;
2941
31680ac2
DA		 2942 /* l3mdev_update_flow overrides oif if the device is enslaved; in
2943 * this case we must match on the real ingress device, so reset it
2944 */
2945 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2946 fl6->flowi6_oif = skb->dev->ifindex;
2947
b55b76b2 2948 /* Get the "current" route for this destination and
67c408cf 2949 * check if the redirect has come from appropriate router.
b55b76b2
DJ		 2950 *
2951 * RFC 4861 specifies that redirects should only be
2952 * accepted if they come from the nexthop to the target.
2953 * Due to the way the routes are chosen, this notion
2954 * is a bit fuzzy and one might need to check all possible
2955 * routes.
2956 */
2957
66f5d6ce 2958 rcu_read_lock();
6454743b 2959 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
b55b76b2 2960restart:
66f5d6ce 2961 for_each_fib6_node_rt_rcu(fn) {
9b6b35ab 2962 res.f6i = rt;
14895687 2963 if (fib6_check_expired(rt))
b55b76b2 2964 continue;
93c2fb25 2965 if (rt->fib6_flags & RTF_REJECT)
b55b76b2 2966 break;
c55c8988
DA		 2967 if (unlikely(rt->nh)) {
2968 if (nexthop_is_blackhole(rt->nh))
2969 continue;
2970 /* on match, res->nh is filled in and potentially ret */
2971 if (nexthop_for_each_fib6_nh(rt->nh,
2972 fib6_nh_redirect_match,
2973 &arg))
2974 goto out;
2975 } else {
2976 res.nh = rt->fib6_nh;
2977 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
2978 &ret))
2979 goto out;
2980 }
b55b76b2
DJ		 2981 }
2982
2983 if (!rt)
421842ed 2984 rt = net->ipv6.fib6_null_entry;
93c2fb25 2985 else if (rt->fib6_flags & RTF_REJECT) {
23fb93a4 2986 ret = net->ipv6.ip6_null_entry;
b0a1ba59
MKL		 2987 goto out;
2988 }
2989
421842ed 2990 if (rt == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2991 fn = fib6_backtrack(fn, &fl6->saddr);
2992 if (fn)
2993 goto restart;
b55b76b2 2994 }
a3c00e46 2995
9b6b35ab 2996 res.f6i = rt;
1cf844c7 2997 res.nh = rt->fib6_nh;
b0a1ba59 2998out:
7d21fec9 2999 if (ret) {
10585b43 3000 ip6_hold_safe(net, &ret);
7d21fec9
DA		 3001 } else {
3002 res.fib6_flags = res.f6i->fib6_flags;
3003 res.fib6_type = res.f6i->fib6_type;
9b6b35ab 3004 ret = ip6_create_rt_rcu(&res);
7d21fec9 3005 }
b55b76b2 3006
66f5d6ce 3007 rcu_read_unlock();
b55b76b2 3008
8ff2e5b2 3009 trace_fib6_table_lookup(net, &res, table, fl6);
23fb93a4 3010 return ret;
b55b76b2
DJ		 3011};
3012
3013static struct dst_entry *ip6_route_redirect(struct net *net,
b75cc8f9
DA		 3014 const struct flowi6 *fl6,
3015 const struct sk_buff *skb,
3016 const struct in6_addr *gateway)
b55b76b2
DJ		 3017{
3018 int flags = RT6_LOOKUP_F_HAS_SADDR;
3019 struct ip6rd_flowi rdfl;
3020
3021 rdfl.fl6 = *fl6;
3022 rdfl.gateway = *gateway;
3023
b75cc8f9 3024 return fib6_rule_lookup(net, &rdfl.fl6, skb,
b55b76b2
DJ		 3025 flags, __ip6_route_redirect);
3026}
3027
e2d118a1
LC		 3028void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
3029 kuid_t uid)
3a5ad2ee
DM		 3030{
3031 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
3032 struct dst_entry *dst;
1f7f10ac
MŻ		 3033 struct flowi6 fl6 = {
3034 .flowi6_iif = LOOPBACK_IFINDEX,
3035 .flowi6_oif = oif,
3036 .flowi6_mark = mark,
3037 .daddr = iph->daddr,
3038 .saddr = iph->saddr,
3039 .flowlabel = ip6_flowinfo(iph),
3040 .flowi6_uid = uid,
3041 };
3a5ad2ee 3042
b75cc8f9 3043 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
b55b76b2 3044 rt6_do_redirect(dst, NULL, skb);
3a5ad2ee
DM		 3045 dst_release(dst);
3046}
3047EXPORT_SYMBOL_GPL(ip6_redirect);
3048
d456336d 3049void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
c92a59ec
DJ		 3050{
3051 const struct ipv6hdr *iph = ipv6_hdr(skb);
3052 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3053 struct dst_entry *dst;
0b26fb17
MŻ		 3054 struct flowi6 fl6 = {
3055 .flowi6_iif = LOOPBACK_IFINDEX,
3056 .flowi6_oif = oif,
0b26fb17
MŻ		 3057 .daddr = msg->dest,
3058 .saddr = iph->daddr,
3059 .flowi6_uid = sock_net_uid(net, NULL),
3060 };
c92a59ec 3061
b75cc8f9 3062 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
b55b76b2 3063 rt6_do_redirect(dst, NULL, skb);
c92a59ec
DJ		 3064 dst_release(dst);
3065}
3066
3a5ad2ee
DM		 3067void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3068{
e2d118a1
LC		 3069 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
3070 sk->sk_uid);
3a5ad2ee
DM		 3071}
3072EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3073
0dbaee3b 3074static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1da177e4 3075{
0dbaee3b
DM		 3076 struct net_device *dev = dst->dev;
3077 unsigned int mtu = dst_mtu(dst);
3078 struct net *net = dev_net(dev);
3079
1da177e4
LT		 3080 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3081
5578689a
DL		 3082 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3083 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1da177e4
LT		 3084
3085 /*
1ab1457c
YH		 3086 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3087 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3088 * IPV6_MAXPLEN is also valid and means: "any MSS,
1da177e4
LT		 3089 * rely only on pmtu discovery"
3090 */
3091 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3092 mtu = IPV6_MAXPLEN;
3093 return mtu;
3094}
3095
f67fbeae 3096INDIRECT_CALLABLE_SCOPE unsigned int ip6_mtu(const struct dst_entry *dst)
d33e4553 3097{
d33e4553 3098 struct inet6_dev *idev;
d4ead6b3 3099 unsigned int mtu;
4b32b5ad
MKL		 3100
3101 mtu = dst_metric_raw(dst, RTAX_MTU);
618f9bc7 3102 if (mtu)
30f78d8e 3103 goto out;
618f9bc7
SK		 3104
3105 mtu = IPV6_MIN_MTU;
d33e4553
DM		 3106
3107 rcu_read_lock();
3108 idev = __in6_dev_get(dst->dev);
3109 if (idev)
3110 mtu = idev->cnf.mtu6;
3111 rcu_read_unlock();
3112
30f78d8e 3113out:
14972cbd
RP		 3114 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3115
3116 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
d33e4553 3117}
f67fbeae 3118EXPORT_SYMBOL(ip6_mtu);
d33e4553 3119
901731b8
DA		 3120/* MTU selection:
3121 * 1. mtu on route is locked - use it
3122 * 2. mtu from nexthop exception
3123 * 3. mtu from egress device
3124 *
3125 * based on ip6_dst_mtu_forward and exception logic of
3126 * rt6_find_cached_rt; called with rcu_read_lock
3127 */
b748f260
DA		 3128u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3129 const struct in6_addr *daddr,
3130 const struct in6_addr *saddr)
901731b8 3131{
b748f260
DA		 3132 const struct fib6_nh *nh = res->nh;
3133 struct fib6_info *f6i = res->f6i;
901731b8 3134 struct inet6_dev *idev;
510e2ced 3135 struct rt6_info *rt;
901731b8
DA		 3136 u32 mtu = 0;
3137
3138 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3139 mtu = f6i->fib6_pmtu;
3140 if (mtu)
3141 goto out;
3142 }
3143
510e2ced
WW		 3144 rt = rt6_find_cached_rt(res, daddr, saddr);
3145 if (unlikely(rt)) {
3146 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3147 } else {
b748f260 3148 struct net_device *dev = nh->fib_nh_dev;
901731b8
DA		 3149
3150 mtu = IPV6_MIN_MTU;
3151 idev = __in6_dev_get(dev);
3152 if (idev && idev->cnf.mtu6 > mtu)
3153 mtu = idev->cnf.mtu6;
3154 }
3155
3156 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3157out:
b748f260 3158 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
901731b8
DA		 3159}
3160
3b00944c 3161struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
87a11578 3162 struct flowi6 *fl6)
1da177e4 3163{
87a11578 3164 struct dst_entry *dst;
1da177e4
LT		 3165 struct rt6_info *rt;
3166 struct inet6_dev *idev = in6_dev_get(dev);
c346dca1 3167 struct net *net = dev_net(dev);
1da177e4 3168
38308473 3169 if (unlikely(!idev))
122bdf67 3170 return ERR_PTR(-ENODEV);
1da177e4 3171
ad706862 3172 rt = ip6_dst_alloc(net, dev, 0);
38308473 3173 if (unlikely(!rt)) {
1da177e4 3174 in6_dev_put(idev);
87a11578 3175 dst = ERR_PTR(-ENOMEM);
1da177e4
LT		 3176 goto out;
3177 }
3178
588753f1 3179 rt->dst.input = ip6_input;
8e2ec639 3180 rt->dst.output = ip6_output;
550bab42 3181 rt->rt6i_gateway = fl6->daddr;
87a11578 3182 rt->rt6i_dst.addr = fl6->daddr;
8e2ec639
YZ		 3183 rt->rt6i_dst.plen = 128;
3184 rt->rt6i_idev = idev;
14edd87d 3185 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1da177e4 3186
4c981e28 3187 /* Add this dst into uncached_list so that rt6_disable_ip() can
587fea74
WW		 3188 * do proper release of the net_device
3189 */
3190 rt6_uncached_list_add(rt);
81eb8447 3191 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
1da177e4 3192
87a11578
DM		 3193 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3194
1da177e4 3195out:
87a11578 3196 return dst;
1da177e4
LT		 3197}
3198
569d3645 3199static int ip6_dst_gc(struct dst_ops *ops)
1da177e4 3200{
86393e52 3201 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
7019b78e
DL		 3202 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3203 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3204 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3205 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3206 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
fc66f95c 3207 int entries;
7019b78e 3208
fc66f95c 3209 entries = dst_entries_get_fast(ops);
cf86a086
ED		 3210 if (entries > rt_max_size)
3211 entries = dst_entries_get_slow(ops);
3212
49a18d86 3213 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
fc66f95c 3214 entries <= rt_max_size)
1da177e4
LT		 3215 goto out;
3216
6891a346 3217 net->ipv6.ip6_rt_gc_expire++;
14956643 3218 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
fc66f95c
ED		 3219 entries = dst_entries_get_slow(ops);
3220 if (entries < ops->gc_thresh)
7019b78e 3221 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1da177e4 3222out:
7019b78e 3223 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
fc66f95c 3224 return entries > rt_max_size;
1da177e4
LT		 3225}
3226
b2c709cc
DA		 3227static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3228 const struct in6_addr *gw_addr, u32 tbid,
3229 int flags, struct fib6_result *res)
8c14586f
DA		 3230{
3231 struct flowi6 fl6 = {
3232 .flowi6_oif = cfg->fc_ifindex,
3233 .daddr = *gw_addr,
3234 .saddr = cfg->fc_prefsrc,
3235 };
3236 struct fib6_table *table;
b2c709cc 3237 int err;
8c14586f 3238
f4797b33 3239 table = fib6_get_table(net, tbid);
8c14586f 3240 if (!table)
b2c709cc 3241 return -EINVAL;
8c14586f
DA		 3242
3243 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3244 flags |= RT6_LOOKUP_F_HAS_SADDR;
3245
f4797b33 3246 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
8c14586f 3247
b2c709cc
DA		 3248 err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3249 if (!err && res->f6i != net->ipv6.fib6_null_entry)
3250 fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3251 cfg->fc_ifindex != 0, NULL, flags);
8c14586f 3252
b2c709cc 3253 return err;
8c14586f
DA		 3254}
3255
fc1e64e1
DA		 3256static int ip6_route_check_nh_onlink(struct net *net,
3257 struct fib6_config *cfg,
9fbb704c 3258 const struct net_device *dev,
fc1e64e1
DA		 3259 struct netlink_ext_ack *extack)
3260{
b2c709cc 3261 u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
fc1e64e1 3262 const struct in6_addr *gw_addr = &cfg->fc_gateway;
b2c709cc 3263 struct fib6_result res = {};
fc1e64e1
DA		 3264 int err;
3265
b2c709cc
DA		 3266 err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3267 if (!err && !(res.fib6_flags & RTF_REJECT) &&
3268 /* ignore match if it is the default route */
3269 !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3270 (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3271 NL_SET_ERR_MSG(extack,
3272 "Nexthop has invalid gateway or device mismatch");
3273 err = -EINVAL;
fc1e64e1
DA		 3274 }
3275
3276 return err;
3277}
3278
1edce99f
DA		 3279static int ip6_route_check_nh(struct net *net,
3280 struct fib6_config *cfg,
3281 struct net_device **_dev,
3282 struct inet6_dev **idev)
3283{
3284 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3285 struct net_device *dev = _dev ? *_dev : NULL;
b2c709cc
DA		 3286 int flags = RT6_LOOKUP_F_IFACE;
3287 struct fib6_result res = {};
1edce99f
DA		 3288 int err = -EHOSTUNREACH;
3289
3290 if (cfg->fc_table) {
b2c709cc
DA		 3291 err = ip6_nh_lookup_table(net, cfg, gw_addr,
3292 cfg->fc_table, flags, &res);
3293 /* gw_addr can not require a gateway or resolve to a reject
3294 * route. If a device is given, it must match the result.
3295 */
3296 if (err || res.fib6_flags & RTF_REJECT ||
3297 res.nh->fib_nh_gw_family ||
3298 (dev && dev != res.nh->fib_nh_dev))
3299 err = -EHOSTUNREACH;
1edce99f
DA		 3300 }
3301
b2c709cc
DA		 3302 if (err < 0) {
3303 struct flowi6 fl6 = {
3304 .flowi6_oif = cfg->fc_ifindex,
3305 .daddr = *gw_addr,
3306 };
1edce99f 3307
b2c709cc
DA		 3308 err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3309 if (err || res.fib6_flags & RTF_REJECT ||
3310 res.nh->fib_nh_gw_family)
3311 err = -EHOSTUNREACH;
3312
3313 if (err)
3314 return err;
3315
3316 fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3317 cfg->fc_ifindex != 0, NULL, flags);
3318 }
1edce99f 3319
b2c709cc 3320 err = 0;
1edce99f 3321 if (dev) {
b2c709cc
DA		 3322 if (dev != res.nh->fib_nh_dev)
3323 err = -EHOSTUNREACH;
1edce99f 3324 } else {
b2c709cc 3325 *_dev = dev = res.nh->fib_nh_dev;
1edce99f 3326 dev_hold(dev);
b2c709cc 3327 *idev = in6_dev_get(dev);
1edce99f
DA		 3328 }
3329
1edce99f
DA		 3330 return err;
3331}
3332
9fbb704c
DA		 3333static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3334 struct net_device **_dev, struct inet6_dev **idev,
3335 struct netlink_ext_ack *extack)
3336{
3337 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3338 int gwa_type = ipv6_addr_type(gw_addr);
232378e8 3339 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
9fbb704c 3340 const struct net_device *dev = *_dev;
232378e8 3341 bool need_addr_check = !dev;
9fbb704c
DA		 3342 int err = -EINVAL;
3343
3344 /* if gw_addr is local we will fail to detect this in case
3345 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3346 * will return already-added prefix route via interface that
3347 * prefix route was assigned to, which might be non-loopback.
3348 */
232378e8
DA		 3349 if (dev &&
3350 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3351 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
9fbb704c
DA		 3352 goto out;
3353 }
3354
3355 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3356 /* IPv6 strictly inhibits using not link-local
3357 * addresses as nexthop address.
3358 * Otherwise, router will not able to send redirects.
3359 * It is very good, but in some (rare!) circumstances
3360 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3361 * some exceptions. --ANK
3362 * We allow IPv4-mapped nexthops to support RFC4798-type
3363 * addressing
3364 */
3365 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3366 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3367 goto out;
3368 }
3369
b2c709cc
DA		 3370 rcu_read_lock();
3371
9fbb704c
DA		 3372 if (cfg->fc_flags & RTNH_F_ONLINK)
3373 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3374 else
3375 err = ip6_route_check_nh(net, cfg, _dev, idev);
3376
b2c709cc
DA		 3377 rcu_read_unlock();
3378
9fbb704c
DA		 3379 if (err)
3380 goto out;
3381 }
3382
3383 /* reload in case device was changed */
3384 dev = *_dev;
3385
3386 err = -EINVAL;
3387 if (!dev) {
3388 NL_SET_ERR_MSG(extack, "Egress device not specified");
3389 goto out;
3390 } else if (dev->flags & IFF_LOOPBACK) {
3391 NL_SET_ERR_MSG(extack,
3392 "Egress device can not be loopback device for this route");
3393 goto out;
3394 }
232378e8
DA		 3395
3396 /* if we did not check gw_addr above, do so now that the
3397 * egress device has been resolved.
3398 */
3399 if (need_addr_check &&
3400 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3401 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3402 goto out;
3403 }
3404
9fbb704c
DA		 3405 err = 0;
3406out:
3407 return err;
3408}
3409
83c44251
DA		 3410static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3411{
3412 if ((flags & RTF_REJECT) ||
3413 (dev && (dev->flags & IFF_LOOPBACK) &&
3414 !(addr_type & IPV6_ADDR_LOOPBACK) &&
aea23c32 3415 !(flags & (RTF_ANYCAST | RTF_LOCAL))))
83c44251
DA		 3416 return true;
3417
3418 return false;
3419}
3420
3421int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3422 struct fib6_config *cfg, gfp_t gfp_flags,
3423 struct netlink_ext_ack *extack)
3424{
3425 struct net_device *dev = NULL;
3426 struct inet6_dev *idev = NULL;
3427 int addr_type;
3428 int err;
3429
f1741730 3430 fib6_nh->fib_nh_family = AF_INET6;
1bef4c22
ED		 3431#ifdef CONFIG_IPV6_ROUTER_PREF
3432 fib6_nh->last_probe = jiffies;
3433#endif
38428d68
RP		 3434 if (cfg->fc_is_fdb) {
3435 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3436 fib6_nh->fib_nh_gw_family = AF_INET6;
3437 return 0;
3438 }
f1741730 3439
83c44251
DA		 3440 err = -ENODEV;
3441 if (cfg->fc_ifindex) {
3442 dev = dev_get_by_index(net, cfg->fc_ifindex);
3443 if (!dev)
3444 goto out;
3445 idev = in6_dev_get(dev);
3446 if (!idev)
3447 goto out;
3448 }
3449
3450 if (cfg->fc_flags & RTNH_F_ONLINK) {
3451 if (!dev) {
3452 NL_SET_ERR_MSG(extack,
3453 "Nexthop device required for onlink");
3454 goto out;
3455 }
3456
3457 if (!(dev->flags & IFF_UP)) {
3458 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3459 err = -ENETDOWN;
3460 goto out;
3461 }
3462
ad1601ae 3463 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
83c44251
DA		 3464 }
3465
ad1601ae 3466 fib6_nh->fib_nh_weight = 1;
83c44251
DA		 3467
3468 /* We cannot add true routes via loopback here,
3469 * they would result in kernel looping; promote them to reject routes
3470 */
3471 addr_type = ipv6_addr_type(&cfg->fc_dst);
3472 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3473 /* hold loopback dev/idev if we haven't done so. */
3474 if (dev != net->loopback_dev) {
3475 if (dev) {
3476 dev_put(dev);
3477 in6_dev_put(idev);
3478 }
3479 dev = net->loopback_dev;
3480 dev_hold(dev);
3481 idev = in6_dev_get(dev);
3482 if (!idev) {
3483 err = -ENODEV;
3484 goto out;
3485 }
3486 }
7dd73168 3487 goto pcpu_alloc;
83c44251
DA		 3488 }
3489
3490 if (cfg->fc_flags & RTF_GATEWAY) {
3491 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3492 if (err)
3493 goto out;
3494
ad1601ae 3495 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
bdf00467 3496 fib6_nh->fib_nh_gw_family = AF_INET6;
83c44251
DA		 3497 }
3498
3499 err = -ENODEV;
3500 if (!dev)
3501 goto out;
3502
3503 if (idev->cnf.disable_ipv6) {
3504 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3505 err = -EACCES;
3506 goto out;
3507 }
3508
3509 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3510 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3511 err = -ENETDOWN;
3512 goto out;
3513 }
3514
3515 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3516 !netif_carrier_ok(dev))
ad1601ae 3517 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
83c44251 3518
faee6769 3519 err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
7dd73168
DA		 3520 cfg->fc_encap_type, cfg, gfp_flags, extack);
3521 if (err)
3522 goto out;
3523
3524pcpu_alloc:
f40b6ae2
DA		 3525 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3526 if (!fib6_nh->rt6i_pcpu) {
3527 err = -ENOMEM;
3528 goto out;
3529 }
3530
ad1601ae 3531 fib6_nh->fib_nh_dev = dev;
f1741730 3532 fib6_nh->fib_nh_oif = dev->ifindex;
83c44251
DA		 3533 err = 0;
3534out:
3535 if (idev)
3536 in6_dev_put(idev);
3537
3538 if (err) {
ad1601ae
DA		 3539 lwtstate_put(fib6_nh->fib_nh_lws);
3540 fib6_nh->fib_nh_lws = NULL;
83c44251
DA		 3541 if (dev)
3542 dev_put(dev);
3543 }
3544
3545 return err;
3546}
3547
dac7d0f2
DA		 3548void fib6_nh_release(struct fib6_nh *fib6_nh)
3549{
cc5c073a
DA		 3550 struct rt6_exception_bucket *bucket;
3551
3552 rcu_read_lock();
3553
3554 fib6_nh_flush_exceptions(fib6_nh, NULL);
3555 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3556 if (bucket) {
3557 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3558 kfree(bucket);
3559 }
3560
3561 rcu_read_unlock();
3562
f40b6ae2
DA		 3563 if (fib6_nh->rt6i_pcpu) {
3564 int cpu;
3565
3566 for_each_possible_cpu(cpu) {
3567 struct rt6_info **ppcpu_rt;
3568 struct rt6_info *pcpu_rt;
3569
3570 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3571 pcpu_rt = *ppcpu_rt;
3572 if (pcpu_rt) {
3573 dst_dev_put(&pcpu_rt->dst);
3574 dst_release(&pcpu_rt->dst);
3575 *ppcpu_rt = NULL;
3576 }
3577 }
3578
3579 free_percpu(fib6_nh->rt6i_pcpu);
3580 }
3581
979e276e 3582 fib_nh_common_release(&fib6_nh->nh_common);
dac7d0f2
DA		 3583}
3584
8d1c802b 3585static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
acb54e3c 3586 gfp_t gfp_flags,
333c4301 3587 struct netlink_ext_ack *extack)
1da177e4 3588{
5578689a 3589 struct net *net = cfg->fc_nlinfo.nl_net;
8d1c802b 3590 struct fib6_info *rt = NULL;
f88d8ea6 3591 struct nexthop *nh = NULL;
c71099ac 3592 struct fib6_table *table;
f88d8ea6 3593 struct fib6_nh *fib6_nh;
8c5b83f0 3594 int err = -EINVAL;
83c44251 3595 int addr_type;
1da177e4 3596
557c44be 3597 /* RTF_PCPU is an internal flag; can not be set by userspace */
d5d531cb
DA		 3598 if (cfg->fc_flags & RTF_PCPU) {
3599 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
557c44be 3600 goto out;
d5d531cb 3601 }
557c44be 3602
2ea2352e
WW		 3603 /* RTF_CACHE is an internal flag; can not be set by userspace */
3604 if (cfg->fc_flags & RTF_CACHE) {
3605 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3606 goto out;
3607 }
3608
e8478e80
DA		 3609 if (cfg->fc_type > RTN_MAX) {
3610 NL_SET_ERR_MSG(extack, "Invalid route type");
3611 goto out;
3612 }
3613
d5d531cb
DA		 3614 if (cfg->fc_dst_len > 128) {
3615 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3616 goto out;
3617 }
3618 if (cfg->fc_src_len > 128) {
3619 NL_SET_ERR_MSG(extack, "Invalid source address length");
8c5b83f0 3620 goto out;
d5d531cb 3621 }
1da177e4 3622#ifndef CONFIG_IPV6_SUBTREES
d5d531cb
DA		 3623 if (cfg->fc_src_len) {
3624 NL_SET_ERR_MSG(extack,
3625 "Specifying source address requires IPV6_SUBTREES to be enabled");
8c5b83f0 3626 goto out;
d5d531cb 3627 }
1da177e4 3628#endif
5b98324e
DA		 3629 if (cfg->fc_nh_id) {
3630 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3631 if (!nh) {
3632 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3633 goto out;
3634 }
3635 err = fib6_check_nexthop(nh, cfg, extack);
3636 if (err)
3637 goto out;
3638 }
fc1e64e1 3639
d71314b4 3640 err = -ENOBUFS;
38308473
DM		 3641 if (cfg->fc_nlinfo.nlh &&
3642 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
d71314b4 3643 table = fib6_get_table(net, cfg->fc_table);
38308473 3644 if (!table) {
f3213831 3645 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
d71314b4
MV		 3646 table = fib6_new_table(net, cfg->fc_table);
3647 }
3648 } else {
3649 table = fib6_new_table(net, cfg->fc_table);
3650 }
38308473
DM		 3651
3652 if (!table)
c71099ac 3653 goto out;
c71099ac 3654
93531c67 3655 err = -ENOMEM;
f88d8ea6 3656 rt = fib6_info_alloc(gfp_flags, !nh);
93531c67 3657 if (!rt)
1da177e4 3658 goto out;
93531c67 3659
d7e774f3
DA		 3660 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3661 extack);
767a2217
DA		 3662 if (IS_ERR(rt->fib6_metrics)) {
3663 err = PTR_ERR(rt->fib6_metrics);
fda21d46
ED		 3664 /* Do not leave garbage there. */
3665 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
767a2217
DA		 3666 goto out;
3667 }
3668
93531c67
DA		 3669 if (cfg->fc_flags & RTF_ADDRCONF)
3670 rt->dst_nocount = true;
1da177e4 3671
1716a961 3672 if (cfg->fc_flags & RTF_EXPIRES)
14895687 3673 fib6_set_expires(rt, jiffies +
1716a961
G		 3674 clock_t_to_jiffies(cfg->fc_expires));
3675 else
14895687 3676 fib6_clean_expires(rt);
1da177e4 3677
86872cb5
TG		 3678 if (cfg->fc_protocol == RTPROT_UNSPEC)
3679 cfg->fc_protocol = RTPROT_BOOT;
93c2fb25 3680 rt->fib6_protocol = cfg->fc_protocol;
86872cb5 3681
83c44251
DA		 3682 rt->fib6_table = table;
3683 rt->fib6_metric = cfg->fc_metric;
c7036d97 3684 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
2b2450ca 3685 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
19e42e45 3686
93c2fb25
DA		 3687 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3688 rt->fib6_dst.plen = cfg->fc_dst_len;
e5fd387a 3689
1da177e4 3690#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 3691 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3692 rt->fib6_src.plen = cfg->fc_src_len;
1da177e4 3693#endif
f88d8ea6 3694 if (nh) {
f88d8ea6 3695 if (rt->fib6_src.plen) {
4daa95af 3696 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
f88d8ea6
DA		 3697 goto out;
3698 }
706ec919
XY		 3699 if (!nexthop_get(nh)) {
3700 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3701 goto out;
3702 }
f88d8ea6
DA		 3703 rt->nh = nh;
3704 fib6_nh = nexthop_fib6_nh(rt->nh);
3705 } else {
3706 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3707 if (err)
3708 goto out;
1da177e4 3709
f88d8ea6
DA		 3710 fib6_nh = rt->fib6_nh;
3711
3712 /* We cannot add true routes via loopback here, they would
3713 * result in kernel looping; promote them to reject routes
3714 */
3715 addr_type = ipv6_addr_type(&cfg->fc_dst);
3716 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3717 addr_type))
3718 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3719 }
955ec4cb 3720
c3968a85 3721 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
f88d8ea6 3722 struct net_device *dev = fib6_nh->fib_nh_dev;
83c44251 3723
c3968a85 3724 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
d5d531cb 3725 NL_SET_ERR_MSG(extack, "Invalid source address");
c3968a85
DW		 3726 err = -EINVAL;
3727 goto out;
3728 }
93c2fb25
DA		 3729 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3730 rt->fib6_prefsrc.plen = 128;
c3968a85 3731 } else
93c2fb25 3732 rt->fib6_prefsrc.plen = 0;
c3968a85 3733
8c5b83f0 3734 return rt;
6b9ea5a6 3735out:
93531c67 3736 fib6_info_release(rt);
8c5b83f0 3737 return ERR_PTR(err);
6b9ea5a6
RP		 3738}
3739
acb54e3c 3740int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
333c4301 3741 struct netlink_ext_ack *extack)
6b9ea5a6 3742{
8d1c802b 3743 struct fib6_info *rt;
6b9ea5a6
RP		 3744 int err;
3745
acb54e3c 3746 rt = ip6_route_info_create(cfg, gfp_flags, extack);
d4ead6b3
DA		 3747 if (IS_ERR(rt))
3748 return PTR_ERR(rt);
6b9ea5a6 3749
d4ead6b3 3750 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
93531c67 3751 fib6_info_release(rt);
6b9ea5a6 3752
1da177e4
LT		 3753 return err;
3754}
3755
8d1c802b 3756static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
1da177e4 3757{
afb1d4b5 3758 struct net *net = info->nl_net;
c71099ac 3759 struct fib6_table *table;
afb1d4b5 3760 int err;
1da177e4 3761
421842ed 3762 if (rt == net->ipv6.fib6_null_entry) {
6825a26c
G		 3763 err = -ENOENT;
3764 goto out;
3765 }
6c813a72 3766
93c2fb25 3767 table = rt->fib6_table;
66f5d6ce 3768 spin_lock_bh(&table->tb6_lock);
86872cb5 3769 err = fib6_del(rt, info);
66f5d6ce 3770 spin_unlock_bh(&table->tb6_lock);
1da177e4 3771
6825a26c 3772out:
93531c67 3773 fib6_info_release(rt);
1da177e4
LT		 3774 return err;
3775}
3776
11dd74b3 3777int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
e0a1ad73 3778{
11dd74b3
RP		 3779 struct nl_info info = {
3780 .nl_net = net,
3781 .skip_notify = skip_notify
3782 };
afb1d4b5 3783
528c4ceb 3784 return __ip6_del_rt(rt, &info);
e0a1ad73
TG		 3785}
3786
8d1c802b 3787static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
0ae81335
DA		 3788{
3789 struct nl_info *info = &cfg->fc_nlinfo;
e3330039 3790 struct net *net = info->nl_net;
16a16cd3 3791 struct sk_buff *skb = NULL;
0ae81335 3792 struct fib6_table *table;
e3330039 3793 int err = -ENOENT;
0ae81335 3794
421842ed 3795 if (rt == net->ipv6.fib6_null_entry)
e3330039 3796 goto out_put;
93c2fb25 3797 table = rt->fib6_table;
66f5d6ce 3798 spin_lock_bh(&table->tb6_lock);
0ae81335 3799
93c2fb25 3800 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
8d1c802b 3801 struct fib6_info *sibling, *next_sibling;
0284696b 3802 struct fib6_node *fn;
0ae81335 3803
16a16cd3
DA		 3804 /* prefer to send a single notification with all hops */
3805 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3806 if (skb) {
3807 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3808
d4ead6b3 3809 if (rt6_fill_node(net, skb, rt, NULL,
16a16cd3
DA		 3810 NULL, NULL, 0, RTM_DELROUTE,
3811 info->portid, seq, 0) < 0) {
3812 kfree_skb(skb);
3813 skb = NULL;
3814 } else
3815 info->skip_notify = 1;
3816 }
3817
0284696b
IS		 3818 /* 'rt' points to the first sibling route. If it is not the
3819 * leaf, then we do not need to send a notification. Otherwise,
3820 * we need to check if the last sibling has a next route or not
3821 * and emit a replace or delete notification, respectively.
3822 */
2881fd61 3823 info->skip_notify_kernel = 1;
0284696b
IS		 3824 fn = rcu_dereference_protected(rt->fib6_node,
3825 lockdep_is_held(&table->tb6_lock));
3826 if (rcu_access_pointer(fn->leaf) == rt) {
3827 struct fib6_info *last_sibling, *replace_rt;
3828
3829 last_sibling = list_last_entry(&rt->fib6_siblings,
3830 struct fib6_info,
3831 fib6_siblings);
3832 replace_rt = rcu_dereference_protected(
3833 last_sibling->fib6_next,
3834 lockdep_is_held(&table->tb6_lock));
3835 if (replace_rt)
3836 call_fib6_entry_notifiers_replace(net,
3837 replace_rt);
3838 else
3839 call_fib6_multipath_entry_notifiers(net,
caafb250 3840 FIB_EVENT_ENTRY_DEL,
0284696b
IS		 3841 rt, rt->fib6_nsiblings,
3842 NULL);
3843 }
0ae81335 3844 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25
DA		 3845 &rt->fib6_siblings,
3846 fib6_siblings) {
0ae81335
DA		 3847 err = fib6_del(sibling, info);
3848 if (err)
e3330039 3849 goto out_unlock;
0ae81335
DA		 3850 }
3851 }
3852
3853 err = fib6_del(rt, info);
e3330039 3854out_unlock:
66f5d6ce 3855 spin_unlock_bh(&table->tb6_lock);
e3330039 3856out_put:
93531c67 3857 fib6_info_release(rt);
16a16cd3
DA		 3858
3859 if (skb) {
e3330039 3860 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
16a16cd3
DA		 3861 info->nlh, gfp_any());
3862 }
0ae81335
DA		 3863 return err;
3864}
3865
0fa6efc5 3866static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
23fb93a4
DA		 3867{
3868 int rc = -ESRCH;
3869
3870 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3871 goto out;
3872
3873 if (cfg->fc_flags & RTF_GATEWAY &&
3874 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3875 goto out;
761f6026
XL		 3876
3877 rc = rt6_remove_exception_rt(rt);
23fb93a4
DA		 3878out:
3879 return rc;
3880}
3881
0fa6efc5
DA		 3882static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3883 struct fib6_nh *nh)
3884{
3885 struct fib6_result res = {
3886 .f6i = rt,
3887 .nh = nh,
3888 };
3889 struct rt6_info *rt_cache;
3890
3891 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3892 if (rt_cache)
3893 return __ip6_del_cached_rt(rt_cache, cfg);
3894
3895 return 0;
3896}
3897
5b98324e
DA		 3898struct fib6_nh_del_cached_rt_arg {
3899 struct fib6_config *cfg;
3900 struct fib6_info *f6i;
3901};
3902
3903static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
3904{
3905 struct fib6_nh_del_cached_rt_arg *arg = _arg;
3906 int rc;
3907
3908 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
3909 return rc != -ESRCH ? rc : 0;
3910}
3911
3912static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
3913{
3914 struct fib6_nh_del_cached_rt_arg arg = {
3915 .cfg = cfg,
3916 .f6i = f6i
3917 };
3918
3919 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
3920}
3921
333c4301
DA		 3922static int ip6_route_del(struct fib6_config *cfg,
3923 struct netlink_ext_ack *extack)
1da177e4 3924{
c71099ac 3925 struct fib6_table *table;
8d1c802b 3926 struct fib6_info *rt;
1da177e4 3927 struct fib6_node *fn;
1da177e4
LT		 3928 int err = -ESRCH;
3929
5578689a 3930 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
d5d531cb
DA		 3931 if (!table) {
3932 NL_SET_ERR_MSG(extack, "FIB table does not exist");
c71099ac 3933 return err;
d5d531cb 3934 }
c71099ac 3935
66f5d6ce 3936 rcu_read_lock();
1da177e4 3937
c71099ac 3938 fn = fib6_locate(&table->tb6_root,
86872cb5 3939 &cfg->fc_dst, cfg->fc_dst_len,
38fbeeee 3940 &cfg->fc_src, cfg->fc_src_len,
2b760fcf 3941 !(cfg->fc_flags & RTF_CACHE));
1ab1457c 3942
1da177e4 3943 if (fn) {
66f5d6ce 3944 for_each_fib6_node_rt_rcu(fn) {
ad1601ae
DA		 3945 struct fib6_nh *nh;
3946
3401bfb1
SB		 3947 if (rt->nh && cfg->fc_nh_id &&
3948 rt->nh->id != cfg->fc_nh_id)
5b98324e 3949 continue;
23fb93a4 3950
5b98324e
DA		 3951 if (cfg->fc_flags & RTF_CACHE) {
3952 int rc = 0;
3953
3954 if (rt->nh) {
3955 rc = ip6_del_cached_rt_nh(cfg, rt);
3956 } else if (cfg->fc_nh_id) {
3957 continue;
3958 } else {
3959 nh = rt->fib6_nh;
3960 rc = ip6_del_cached_rt(cfg, rt, nh);
3961 }
0fa6efc5
DA		 3962 if (rc != -ESRCH) {
3963 rcu_read_unlock();
3964 return rc;
23fb93a4
DA		 3965 }
3966 continue;
2b760fcf 3967 }
ad1601ae 3968
5b98324e
DA		 3969 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
3970 continue;
3971 if (cfg->fc_protocol &&
3972 cfg->fc_protocol != rt->fib6_protocol)
3973 continue;
3974
3975 if (rt->nh) {
3976 if (!fib6_info_hold_safe(rt))
3977 continue;
3978 rcu_read_unlock();
3979
3980 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3981 }
3982 if (cfg->fc_nh_id)
3983 continue;
3984
3985 nh = rt->fib6_nh;
86872cb5 3986 if (cfg->fc_ifindex &&
ad1601ae
DA		 3987 (!nh->fib_nh_dev ||
3988 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
1da177e4 3989 continue;
86872cb5 3990 if (cfg->fc_flags & RTF_GATEWAY &&
ad1601ae 3991 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
1da177e4 3992 continue;
e873e4b9
WW		 3993 if (!fib6_info_hold_safe(rt))
3994 continue;
66f5d6ce 3995 rcu_read_unlock();
1da177e4 3996
0ae81335
DA		 3997 /* if gateway was specified only delete the one hop */
3998 if (cfg->fc_flags & RTF_GATEWAY)
3999 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4000
4001 return __ip6_del_rt_siblings(rt, cfg);
1da177e4
LT		 4002 }
4003 }
66f5d6ce 4004 rcu_read_unlock();
1da177e4
LT		 4005
4006 return err;
4007}
4008
6700c270 4009static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
a6279458 4010{
a6279458 4011 struct netevent_redirect netevent;
e8599ff4 4012 struct rt6_info *rt, *nrt = NULL;
85bd05de 4013 struct fib6_result res = {};
e8599ff4
DM		 4014 struct ndisc_options ndopts;
4015 struct inet6_dev *in6_dev;
4016 struct neighbour *neigh;
71bcdba0 4017 struct rd_msg *msg;
6e157b6a
DM		 4018 int optlen, on_link;
4019 u8 *lladdr;
e8599ff4 4020
29a3cad5 4021 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
71bcdba0 4022 optlen -= sizeof(*msg);
e8599ff4
DM		 4023
4024 if (optlen < 0) {
6e157b6a 4025 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
e8599ff4
DM		 4026 return;
4027 }
4028
71bcdba0 4029 msg = (struct rd_msg *)icmp6_hdr(skb);
e8599ff4 4030
71bcdba0 4031 if (ipv6_addr_is_multicast(&msg->dest)) {
6e157b6a 4032 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
e8599ff4
DM		 4033 return;
4034 }
4035
6e157b6a 4036 on_link = 0;
71bcdba0 4037 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
e8599ff4 4038 on_link = 1;
71bcdba0 4039 } else if (ipv6_addr_type(&msg->target) !=
e8599ff4 4040 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
6e157b6a 4041 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
e8599ff4
DM		 4042 return;
4043 }
4044
4045 in6_dev = __in6_dev_get(skb->dev);
4046 if (!in6_dev)
4047 return;
4048 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
4049 return;
4050
4051 /* RFC2461 8.1:
4052 * The IP source address of the Redirect MUST be the same as the current
4053 * first-hop router for the specified ICMP Destination Address.
4054 */
4055
f997c55c 4056 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
e8599ff4
DM		 4057 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4058 return;
4059 }
6e157b6a
DM		 4060
4061 lladdr = NULL;
e8599ff4
DM		 4062 if (ndopts.nd_opts_tgt_lladdr) {
4063 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4064 skb->dev);
4065 if (!lladdr) {
4066 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4067 return;
4068 }
4069 }
4070
6e157b6a 4071 rt = (struct rt6_info *) dst;
ec13ad1d 4072 if (rt->rt6i_flags & RTF_REJECT) {
6e157b6a 4073 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
e8599ff4 4074 return;
6e157b6a 4075 }
e8599ff4 4076
6e157b6a
DM		 4077 /* Redirect received -> path was valid.
4078 * Look, redirects are sent only in response to data packets,
4079 * so that this nexthop apparently is reachable. --ANK
4080 */
0dec879f 4081 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
a6279458 4082
71bcdba0 4083 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
6e157b6a
DM		 4084 if (!neigh)
4085 return;
a6279458 4086
1da177e4
LT		 4087 /*
4088 * We have finally decided to accept it.
4089 */
4090
f997c55c 4091 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1da177e4
LT		 4092 NEIGH_UPDATE_F_WEAK_OVERRIDE|
4093 NEIGH_UPDATE_F_OVERRIDE|
4094 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
f997c55c
AA		 4095 NEIGH_UPDATE_F_ISROUTER)),
4096 NDISC_REDIRECT, &ndopts);
1da177e4 4097
4d85cd0c 4098 rcu_read_lock();
85bd05de 4099 res.f6i = rcu_dereference(rt->from);
ff24e498 4100 if (!res.f6i)
886b7a50 4101 goto out;
8a14e46f 4102
49d5b8ef
DA		 4103 if (res.f6i->nh) {
4104 struct fib6_nh_match_arg arg = {
4105 .dev = dst->dev,
4106 .gw = &rt->rt6i_gateway,
4107 };
4108
4109 nexthop_for_each_fib6_nh(res.f6i->nh,
4110 fib6_nh_find_match, &arg);
4111
4112 /* fib6_info uses a nexthop that does not have fib6_nh
4113 * using the dst->dev. Should be impossible
4114 */
4115 if (!arg.match)
4116 goto out;
4117 res.nh = arg.match;
4118 } else {
4119 res.nh = res.f6i->fib6_nh;
4120 }
4121
7d21fec9
DA		 4122 res.fib6_flags = res.f6i->fib6_flags;
4123 res.fib6_type = res.f6i->fib6_type;
85bd05de 4124 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
38308473 4125 if (!nrt)
1da177e4
LT		 4126 goto out;
4127
4128 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4129 if (on_link)
4130 nrt->rt6i_flags &= ~RTF_GATEWAY;
4131
4e3fd7a0 4132 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1da177e4 4133
886b7a50 4134 /* rt6_insert_exception() will take care of duplicated exceptions */
5012f0a5 4135 if (rt6_insert_exception(nrt, &res)) {
2b760fcf
WW		 4136 dst_release_immediate(&nrt->dst);
4137 goto out;
4138 }
1da177e4 4139
d8d1f30b
CG		 4140 netevent.old = &rt->dst;
4141 netevent.new = &nrt->dst;
71bcdba0 4142 netevent.daddr = &msg->dest;
60592833 4143 netevent.neigh = neigh;
8d71740c
TT		 4144 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4145
1da177e4 4146out:
886b7a50 4147 rcu_read_unlock();
e8599ff4 4148 neigh_release(neigh);
6e157b6a
DM		 4149}
4150
70ceb4f5 4151#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 4152static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 4153 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 4154 const struct in6_addr *gwaddr,
4155 struct net_device *dev)
70ceb4f5 4156{
830218c1
DA		 4157 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4158 int ifindex = dev->ifindex;
70ceb4f5 4159 struct fib6_node *fn;
8d1c802b 4160 struct fib6_info *rt = NULL;
c71099ac
TG		 4161 struct fib6_table *table;
4162
830218c1 4163 table = fib6_get_table(net, tb_id);
38308473 4164 if (!table)
c71099ac 4165 return NULL;
70ceb4f5 4166
66f5d6ce 4167 rcu_read_lock();
38fbeeee 4168 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
70ceb4f5
YH		 4169 if (!fn)
4170 goto out;
4171
66f5d6ce 4172 for_each_fib6_node_rt_rcu(fn) {
f88d8ea6
DA		 4173 /* these routes do not use nexthops */
4174 if (rt->nh)
4175 continue;
1cf844c7 4176 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
70ceb4f5 4177 continue;
2b2450ca 4178 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
1cf844c7 4179 !rt->fib6_nh->fib_nh_gw_family)
70ceb4f5 4180 continue;
1cf844c7 4181 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
70ceb4f5 4182 continue;
e873e4b9
WW		 4183 if (!fib6_info_hold_safe(rt))
4184 continue;
70ceb4f5
YH		 4185 break;
4186 }
4187out:
66f5d6ce 4188 rcu_read_unlock();
70ceb4f5
YH		 4189 return rt;
4190}
4191
8d1c802b 4192static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 4193 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 4194 const struct in6_addr *gwaddr,
4195 struct net_device *dev,
95c96174 4196 unsigned int pref)
70ceb4f5 4197{
86872cb5 4198 struct fib6_config cfg = {
238fc7ea 4199 .fc_metric = IP6_RT_PRIO_USER,
830218c1 4200 .fc_ifindex = dev->ifindex,
86872cb5
TG		 4201 .fc_dst_len = prefixlen,
4202 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4203 RTF_UP | RTF_PREF(pref),
b91d5329 4204 .fc_protocol = RTPROT_RA,
e8478e80 4205 .fc_type = RTN_UNICAST,
15e47304 4206 .fc_nlinfo.portid = 0,
efa2cea0
DL		 4207 .fc_nlinfo.nlh = NULL,
4208 .fc_nlinfo.nl_net = net,
86872cb5
TG		 4209 };
4210
91b2c9a0 4211 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4e3fd7a0
AD		 4212 cfg.fc_dst = *prefix;
4213 cfg.fc_gateway = *gwaddr;
70ceb4f5 4214
e317da96
YH		 4215 /* We should treat it as a default route if prefix length is 0. */
4216 if (!prefixlen)
86872cb5 4217 cfg.fc_flags |= RTF_DEFAULT;
70ceb4f5 4218
acb54e3c 4219 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
70ceb4f5 4220
830218c1 4221 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
70ceb4f5
YH		 4222}
4223#endif
4224
8d1c802b 4225struct fib6_info *rt6_get_dflt_router(struct net *net,
afb1d4b5
DA		 4226 const struct in6_addr *addr,
4227 struct net_device *dev)
1ab1457c 4228{
830218c1 4229 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
8d1c802b 4230 struct fib6_info *rt;
c71099ac 4231 struct fib6_table *table;
1da177e4 4232
afb1d4b5 4233 table = fib6_get_table(net, tb_id);
38308473 4234 if (!table)
c71099ac 4235 return NULL;
1da177e4 4236
66f5d6ce
WW		 4237 rcu_read_lock();
4238 for_each_fib6_node_rt_rcu(&table->tb6_root) {
f88d8ea6 4239 struct fib6_nh *nh;
ad1601ae 4240
f88d8ea6
DA		 4241 /* RA routes do not use nexthops */
4242 if (rt->nh)
4243 continue;
4244
4245 nh = rt->fib6_nh;
ad1601ae 4246 if (dev == nh->fib_nh_dev &&
93c2fb25 4247 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
ad1601ae 4248 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
1da177e4
LT		 4249 break;
4250 }
e873e4b9
WW		 4251 if (rt && !fib6_info_hold_safe(rt))
4252 rt = NULL;
66f5d6ce 4253 rcu_read_unlock();
1da177e4
LT		 4254 return rt;
4255}
4256
8d1c802b 4257struct fib6_info *rt6_add_dflt_router(struct net *net,
afb1d4b5 4258 const struct in6_addr *gwaddr,
ebacaaa0 4259 struct net_device *dev,
6b2e04bc
PC		 4260 unsigned int pref,
4261 u32 defrtr_usr_metric)
1da177e4 4262{
86872cb5 4263 struct fib6_config cfg = {
ca254490 4264 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
6b2e04bc 4265 .fc_metric = defrtr_usr_metric,
86872cb5
TG		 4266 .fc_ifindex = dev->ifindex,
4267 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4268 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
b91d5329 4269 .fc_protocol = RTPROT_RA,
e8478e80 4270 .fc_type = RTN_UNICAST,
15e47304 4271 .fc_nlinfo.portid = 0,
5578689a 4272 .fc_nlinfo.nlh = NULL,
afb1d4b5 4273 .fc_nlinfo.nl_net = net,
86872cb5 4274 };
1da177e4 4275
4e3fd7a0 4276 cfg.fc_gateway = *gwaddr;
1da177e4 4277
acb54e3c 4278 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
830218c1
DA		 4279 struct fib6_table *table;
4280
4281 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4282 if (table)
4283 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4284 }
1da177e4 4285
afb1d4b5 4286 return rt6_get_dflt_router(net, gwaddr, dev);
1da177e4
LT		 4287}
4288
afb1d4b5
DA		 4289static void __rt6_purge_dflt_routers(struct net *net,
4290 struct fib6_table *table)
1da177e4 4291{
8d1c802b 4292 struct fib6_info *rt;
1da177e4
LT		 4293
4294restart:
66f5d6ce
WW		 4295 rcu_read_lock();
4296 for_each_fib6_node_rt_rcu(&table->tb6_root) {
dcd1f572
DA		 4297 struct net_device *dev = fib6_info_nh_dev(rt);
4298 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4299
93c2fb25 4300 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
e873e4b9
WW		 4301 (!idev || idev->cnf.accept_ra != 2) &&
4302 fib6_info_hold_safe(rt)) {
93531c67 4303 rcu_read_unlock();
11dd74b3 4304 ip6_del_rt(net, rt, false);
1da177e4
LT		 4305 goto restart;
4306 }
4307 }
66f5d6ce 4308 rcu_read_unlock();
830218c1
DA		 4309
4310 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4311}
4312
4313void rt6_purge_dflt_routers(struct net *net)
4314{
4315 struct fib6_table *table;
4316 struct hlist_head *head;
4317 unsigned int h;
4318
4319 rcu_read_lock();
4320
4321 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4322 head = &net->ipv6.fib_table_hash[h];
4323 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4324 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
afb1d4b5 4325 __rt6_purge_dflt_routers(net, table);
830218c1
DA		 4326 }
4327 }
4328
4329 rcu_read_unlock();
1da177e4
LT		 4330}
4331
5578689a
DL		 4332static void rtmsg_to_fib6_config(struct net *net,
4333 struct in6_rtmsg *rtmsg,
86872cb5
TG		 4334 struct fib6_config *cfg)
4335{
8823a3ac
MŻ		 4336 *cfg = (struct fib6_config){
4337 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4338 : RT6_TABLE_MAIN,
4339 .fc_ifindex = rtmsg->rtmsg_ifindex,
67f69513 4340 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
8823a3ac
MŻ		 4341 .fc_expires = rtmsg->rtmsg_info,
4342 .fc_dst_len = rtmsg->rtmsg_dst_len,
4343 .fc_src_len = rtmsg->rtmsg_src_len,
4344 .fc_flags = rtmsg->rtmsg_flags,
4345 .fc_type = rtmsg->rtmsg_type,
4346
4347 .fc_nlinfo.nl_net = net,
4348
4349 .fc_dst = rtmsg->rtmsg_dst,
4350 .fc_src = rtmsg->rtmsg_src,
4351 .fc_gateway = rtmsg->rtmsg_gateway,
4352 };
86872cb5
TG		 4353}
4354
7c1552da 4355int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
1da177e4 4356{
86872cb5 4357 struct fib6_config cfg;
1da177e4
LT		 4358 int err;
4359
7c1552da
CH		 4360 if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4361 return -EINVAL;
4362 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4363 return -EPERM;
86872cb5 4364
7c1552da 4365 rtmsg_to_fib6_config(net, rtmsg, &cfg);
1da177e4 4366
7c1552da
CH		 4367 rtnl_lock();
4368 switch (cmd) {
4369 case SIOCADDRT:
4370 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4371 break;
4372 case SIOCDELRT:
4373 err = ip6_route_del(&cfg, NULL);
4374 break;
3ff50b79 4375 }
7c1552da
CH		 4376 rtnl_unlock();
4377 return err;
1da177e4
LT		 4378}
4379
4380/*
4381 * Drop the packet on the floor
4382 */
4383
d5fdd6ba 4384static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1da177e4 4385{
adf30907 4386 struct dst_entry *dst = skb_dst(skb);
1d3fd8a1
SS		 4387 struct net *net = dev_net(dst->dev);
4388 struct inet6_dev *idev;
4389 int type;
4390
4391 if (netif_is_l3_master(skb->dev) &&
4392 dst->dev == net->loopback_dev)
4393 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4394 else
4395 idev = ip6_dst_idev(dst);
4396
612f09e8
YH		 4397 switch (ipstats_mib_noroutes) {
4398 case IPSTATS_MIB_INNOROUTES:
0660e03f 4399 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
45bb0060 4400 if (type == IPV6_ADDR_ANY) {
1d3fd8a1 4401 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
612f09e8
YH		 4402 break;
4403 }
a8eceea8 4404 fallthrough;
612f09e8 4405 case IPSTATS_MIB_OUTNOROUTES:
1d3fd8a1 4406 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
612f09e8
YH		 4407 break;
4408 }
1d3fd8a1
SS		 4409
4410 /* Start over by dropping the dst for l3mdev case */
4411 if (netif_is_l3_master(skb->dev))
4412 skb_dst_drop(skb);
4413
3ffe533c 4414 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1da177e4
LT		 4415 kfree_skb(skb);
4416 return 0;
4417}
4418
9ce8ade0
TG		 4419static int ip6_pkt_discard(struct sk_buff *skb)
4420{
612f09e8 4421 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4422}
4423
ede2059d 4424static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
1da177e4 4425{
adf30907 4426 skb->dev = skb_dst(skb)->dev;
612f09e8 4427 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1da177e4
LT		 4428}
4429
9ce8ade0
TG		 4430static int ip6_pkt_prohibit(struct sk_buff *skb)
4431{
612f09e8 4432 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4433}
4434
ede2059d 4435static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
9ce8ade0 4436{
adf30907 4437 skb->dev = skb_dst(skb)->dev;
612f09e8 4438 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
9ce8ade0
TG		 4439}
4440
1da177e4
LT		 4441/*
4442 * Allocate a dst for local (unicast / anycast) address.
4443 */
4444
360a9887
DA		 4445struct fib6_info *addrconf_f6i_alloc(struct net *net,
4446 struct inet6_dev *idev,
4447 const struct in6_addr *addr,
4448 bool anycast, gfp_t gfp_flags)
1da177e4 4449{
c7a1ce39
DA		 4450 struct fib6_config cfg = {
4451 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4452 .fc_ifindex = idev->dev->ifindex,
d55a2e37 4453 .fc_flags = RTF_UP | RTF_NONEXTHOP,
c7a1ce39
DA		 4454 .fc_dst = *addr,
4455 .fc_dst_len = 128,
4456 .fc_protocol = RTPROT_KERNEL,
4457 .fc_nlinfo.nl_net = net,
4458 .fc_ignore_dev_down = true,
4459 };
d55a2e37 4460 struct fib6_info *f6i;
1da177e4 4461
e8478e80 4462 if (anycast) {
c7a1ce39
DA		 4463 cfg.fc_type = RTN_ANYCAST;
4464 cfg.fc_flags |= RTF_ANYCAST;
e8478e80 4465 } else {
c7a1ce39
DA		 4466 cfg.fc_type = RTN_LOCAL;
4467 cfg.fc_flags |= RTF_LOCAL;
e8478e80 4468 }
1da177e4 4469
d55a2e37 4470 f6i = ip6_route_info_create(&cfg, gfp_flags, NULL);
8652f17c 4471 if (!IS_ERR(f6i))
d55a2e37
MŻ		 4472 f6i->dst_nocount = true;
4473 return f6i;
1da177e4
LT		 4474}
4475
c3968a85
DW		 4476/* remove deleted ip from prefsrc entries */
4477struct arg_dev_net_ip {
4478 struct net_device *dev;
4479 struct net *net;
4480 struct in6_addr *addr;
4481};
4482
8d1c802b 4483static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
c3968a85
DW		 4484{
4485 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4486 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4487 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4488
f88d8ea6
DA		 4489 if (!rt->nh &&
4490 ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
421842ed 4491 rt != net->ipv6.fib6_null_entry &&
93c2fb25 4492 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
60006a48 4493 spin_lock_bh(&rt6_exception_lock);
c3968a85 4494 /* remove prefsrc entry */
93c2fb25 4495 rt->fib6_prefsrc.plen = 0;
60006a48 4496 spin_unlock_bh(&rt6_exception_lock);
c3968a85
DW		 4497 }
4498 return 0;
4499}
4500
4501void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4502{
4503 struct net *net = dev_net(ifp->idev->dev);
4504 struct arg_dev_net_ip adni = {
4505 .dev = ifp->idev->dev,
4506 .net = net,
4507 .addr = &ifp->addr,
4508 };
0c3584d5 4509 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
c3968a85
DW		 4510}
4511
2b2450ca 4512#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
be7a010d
DJ		 4513
4514/* Remove routers and update dst entries when gateway turn into host. */
8d1c802b 4515static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
be7a010d
DJ		 4516{
4517 struct in6_addr *gateway = (struct in6_addr *)arg;
f88d8ea6
DA		 4518 struct fib6_nh *nh;
4519
4520 /* RA routes do not use nexthops */
4521 if (rt->nh)
4522 return 0;
be7a010d 4523
f88d8ea6 4524 nh = rt->fib6_nh;
93c2fb25 4525 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
cc5c073a 4526 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
be7a010d 4527 return -1;
b16cb459
WW		 4528
4529 /* Further clean up cached routes in exception table.
4530 * This is needed because cached route may have a different
4531 * gateway than its 'parent' in the case of an ip redirect.
4532 */
cc5c073a 4533 fib6_nh_exceptions_clean_tohost(nh, gateway);
b16cb459 4534
be7a010d
DJ		 4535 return 0;
4536}
4537
4538void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4539{
4540 fib6_clean_all(net, fib6_clean_tohost, gateway);
4541}
4542
2127d95a
IS		 4543struct arg_netdev_event {
4544 const struct net_device *dev;
4c981e28 4545 union {
ecc5663c 4546 unsigned char nh_flags;
4c981e28
IS		 4547 unsigned long event;
4548 };
2127d95a
IS		 4549};
4550
8d1c802b 4551static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
d7dedee1 4552{
8d1c802b 4553 struct fib6_info *iter;
d7dedee1
IS		 4554 struct fib6_node *fn;
4555
93c2fb25
DA		 4556 fn = rcu_dereference_protected(rt->fib6_node,
4557 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4558 iter = rcu_dereference_protected(fn->leaf,
93c2fb25 4559 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4560 while (iter) {
93c2fb25 4561 if (iter->fib6_metric == rt->fib6_metric &&
33bd5ac5 4562 rt6_qualify_for_ecmp(iter))
d7dedee1 4563 return iter;
8fb11a9a 4564 iter = rcu_dereference_protected(iter->fib6_next,
93c2fb25 4565 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1
IS		 4566 }
4567
4568 return NULL;
4569}
4570
f88d8ea6 4571/* only called for fib entries with builtin fib6_nh */
8d1c802b 4572static bool rt6_is_dead(const struct fib6_info *rt)
d7dedee1 4573{
1cf844c7
DA		 4574 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4575 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4576 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
d7dedee1
IS		 4577 return true;
4578
4579 return false;
4580}
4581
8d1c802b 4582static int rt6_multipath_total_weight(const struct fib6_info *rt)
d7dedee1 4583{
8d1c802b 4584 struct fib6_info *iter;
d7dedee1
IS		 4585 int total = 0;
4586
4587 if (!rt6_is_dead(rt))
1cf844c7 4588 total += rt->fib6_nh->fib_nh_weight;
d7dedee1 4589
93c2fb25 4590 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
d7dedee1 4591 if (!rt6_is_dead(iter))
1cf844c7 4592 total += iter->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4593 }
4594
4595 return total;
4596}
4597
8d1c802b 4598static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
d7dedee1
IS		 4599{
4600 int upper_bound = -1;
4601
4602 if (!rt6_is_dead(rt)) {
1cf844c7 4603 *weight += rt->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4604 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4605 total) - 1;
4606 }
1cf844c7 4607 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
d7dedee1
IS		 4608}
4609
8d1c802b 4610static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
d7dedee1 4611{
8d1c802b 4612 struct fib6_info *iter;
d7dedee1
IS		 4613 int weight = 0;
4614
4615 rt6_upper_bound_set(rt, &weight, total);
4616
93c2fb25 4617 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
d7dedee1
IS		 4618 rt6_upper_bound_set(iter, &weight, total);
4619}
4620
8d1c802b 4621void rt6_multipath_rebalance(struct fib6_info *rt)
d7dedee1 4622{
8d1c802b 4623 struct fib6_info *first;
d7dedee1
IS		 4624 int total;
4625
4626 /* In case the entire multipath route was marked for flushing,
4627 * then there is no need to rebalance upon the removal of every
4628 * sibling route.
4629 */
93c2fb25 4630 if (!rt->fib6_nsiblings || rt->should_flush)
d7dedee1
IS		 4631 return;
4632
4633 /* During lookup routes are evaluated in order, so we need to
4634 * make sure upper bounds are assigned from the first sibling
4635 * onwards.
4636 */
4637 first = rt6_multipath_first_sibling(rt);
4638 if (WARN_ON_ONCE(!first))
4639 return;
4640
4641 total = rt6_multipath_total_weight(first);
4642 rt6_multipath_upper_bound_set(first, total);
4643}
4644
8d1c802b 4645static int fib6_ifup(struct fib6_info *rt, void *p_arg)
2127d95a
IS		 4646{
4647 const struct arg_netdev_event *arg = p_arg;
7aef6859 4648 struct net *net = dev_net(arg->dev);
2127d95a 4649
f88d8ea6 4650 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
1cf844c7
DA		 4651 rt->fib6_nh->fib_nh_dev == arg->dev) {
4652 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
7aef6859 4653 fib6_update_sernum_upto_root(net, rt);
d7dedee1 4654 rt6_multipath_rebalance(rt);
1de178ed 4655 }
2127d95a
IS		 4656
4657 return 0;
4658}
4659
ecc5663c 4660void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
2127d95a
IS		 4661{
4662 struct arg_netdev_event arg = {
4663 .dev = dev,
6802f3ad
IS		 4664 {
4665 .nh_flags = nh_flags,
4666 },
2127d95a
IS		 4667 };
4668
4669 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4670 arg.nh_flags |= RTNH_F_LINKDOWN;
4671
4672 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4673}
4674
f88d8ea6 4675/* only called for fib entries with inline fib6_nh */
8d1c802b 4676static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
1de178ed
IS		 4677 const struct net_device *dev)
4678{
8d1c802b 4679 struct fib6_info *iter;
1de178ed 4680
1cf844c7 4681 if (rt->fib6_nh->fib_nh_dev == dev)
1de178ed 4682 return true;
93c2fb25 4683 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7 4684 if (iter->fib6_nh->fib_nh_dev == dev)
1de178ed
IS		 4685 return true;
4686
4687 return false;
4688}
4689
8d1c802b 4690static void rt6_multipath_flush(struct fib6_info *rt)
1de178ed 4691{
8d1c802b 4692 struct fib6_info *iter;
1de178ed
IS		 4693
4694 rt->should_flush = 1;
93c2fb25 4695 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1de178ed
IS		 4696 iter->should_flush = 1;
4697}
4698
8d1c802b 4699static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
1de178ed
IS		 4700 const struct net_device *down_dev)
4701{
8d1c802b 4702 struct fib6_info *iter;
1de178ed
IS		 4703 unsigned int dead = 0;
4704
1cf844c7
DA		 4705 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4706 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed 4707 dead++;
93c2fb25 4708 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4709 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4710 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed
IS		 4711 dead++;
4712
4713 return dead;
4714}
4715
8d1c802b 4716static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
1de178ed 4717 const struct net_device *dev,
ecc5663c 4718 unsigned char nh_flags)
1de178ed 4719{
8d1c802b 4720 struct fib6_info *iter;
1de178ed 4721
1cf844c7
DA		 4722 if (rt->fib6_nh->fib_nh_dev == dev)
4723 rt->fib6_nh->fib_nh_flags |= nh_flags;
93c2fb25 4724 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4725 if (iter->fib6_nh->fib_nh_dev == dev)
4726 iter->fib6_nh->fib_nh_flags |= nh_flags;
1de178ed
IS		 4727}
4728
a1a22c12 4729/* called with write lock held for table with rt */
8d1c802b 4730static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
1da177e4 4731{
4c981e28
IS		 4732 const struct arg_netdev_event *arg = p_arg;
4733 const struct net_device *dev = arg->dev;
7aef6859 4734 struct net *net = dev_net(dev);
8ed67789 4735
f88d8ea6 4736 if (rt == net->ipv6.fib6_null_entry || rt->nh)
27c6fa73
IS		 4737 return 0;
4738
4739 switch (arg->event) {
4740 case NETDEV_UNREGISTER:
1cf844c7 4741 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
27c6fa73 4742 case NETDEV_DOWN:
1de178ed 4743 if (rt->should_flush)
27c6fa73 4744 return -1;
93c2fb25 4745 if (!rt->fib6_nsiblings)
1cf844c7 4746 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
1de178ed
IS		 4747 if (rt6_multipath_uses_dev(rt, dev)) {
4748 unsigned int count;
4749
4750 count = rt6_multipath_dead_count(rt, dev);
93c2fb25 4751 if (rt->fib6_nsiblings + 1 == count) {
1de178ed
IS		 4752 rt6_multipath_flush(rt);
4753 return -1;
4754 }
4755 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4756 RTNH_F_LINKDOWN);
7aef6859 4757 fib6_update_sernum(net, rt);
d7dedee1 4758 rt6_multipath_rebalance(rt);
1de178ed
IS		 4759 }
4760 return -2;
27c6fa73 4761 case NETDEV_CHANGE:
1cf844c7 4762 if (rt->fib6_nh->fib_nh_dev != dev ||
93c2fb25 4763 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
27c6fa73 4764 break;
1cf844c7 4765 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
d7dedee1 4766 rt6_multipath_rebalance(rt);
27c6fa73 4767 break;
2b241361 4768 }
c159d30c 4769
1da177e4
LT		 4770 return 0;
4771}
4772
27c6fa73 4773void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
1da177e4 4774{
4c981e28 4775 struct arg_netdev_event arg = {
8ed67789 4776 .dev = dev,
6802f3ad
IS		 4777 {
4778 .event = event,
4779 },
8ed67789 4780 };
7c6bb7d2 4781 struct net *net = dev_net(dev);
8ed67789 4782
7c6bb7d2
DA		 4783 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4784 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4785 else
4786 fib6_clean_all(net, fib6_ifdown, &arg);
4c981e28
IS		 4787}
4788
4789void rt6_disable_ip(struct net_device *dev, unsigned long event)
4790{
4791 rt6_sync_down_dev(dev, event);
4792 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4793 neigh_ifdown(&nd_tbl, dev);
1da177e4
LT		 4794}
4795
95c96174 4796struct rt6_mtu_change_arg {
1da177e4 4797 struct net_device *dev;
95c96174 4798 unsigned int mtu;
c0b220cf 4799 struct fib6_info *f6i;
1da177e4
LT		 4800};
4801
cc5c073a 4802static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
c0b220cf
DA		 4803{
4804 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
cc5c073a 4805 struct fib6_info *f6i = arg->f6i;
c0b220cf
DA		 4806
4807 /* For administrative MTU increase, there is no way to discover
4808 * IPv6 PMTU increase, so PMTU increase should be updated here.
4809 * Since RFC 1981 doesn't include administrative MTU increase
4810 * update PMTU increase is a MUST. (i.e. jumbo frame)
4811 */
4812 if (nh->fib_nh_dev == arg->dev) {
4813 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4814 u32 mtu = f6i->fib6_pmtu;
4815
4816 if (mtu >= arg->mtu ||
4817 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4818 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4819
4820 spin_lock_bh(&rt6_exception_lock);
cc5c073a 4821 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
c0b220cf
DA		 4822 spin_unlock_bh(&rt6_exception_lock);
4823 }
4824
4825 return 0;
4826}
4827
4828static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
1da177e4
LT		 4829{
4830 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4831 struct inet6_dev *idev;
4832
4833 /* In IPv6 pmtu discovery is not optional,
4834 so that RTAX_MTU lock cannot disable it.
4835 We still use this lock to block changes
4836 caused by addrconf/ndisc.
4837 */
4838
4839 idev = __in6_dev_get(arg->dev);
38308473 4840 if (!idev)
1da177e4
LT		 4841 return 0;
4842
c0b220cf
DA		 4843 if (fib6_metric_locked(f6i, RTAX_MTU))
4844 return 0;
d4ead6b3 4845
c0b220cf 4846 arg->f6i = f6i;
2d44234b
DA		 4847 if (f6i->nh) {
4848 /* fib6_nh_mtu_change only returns 0, so this is safe */
4849 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4850 arg);
4851 }
4852
1cf844c7 4853 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
1da177e4
LT		 4854}
4855
95c96174 4856void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
1da177e4 4857{
c71099ac
TG		 4858 struct rt6_mtu_change_arg arg = {
4859 .dev = dev,
4860 .mtu = mtu,
4861 };
1da177e4 4862
0c3584d5 4863 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
1da177e4
LT		 4864}
4865
ef7c79ed 4866static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
75425657 4867 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
5176f91e 4868 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
aa8f8778 4869 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
86872cb5 4870 [RTA_OIF] = { .type = NLA_U32 },
ab364a6f 4871 [RTA_IIF] = { .type = NLA_U32 },
86872cb5
TG		 4872 [RTA_PRIORITY] = { .type = NLA_U32 },
4873 [RTA_METRICS] = { .type = NLA_NESTED },
51ebd318 4874 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
c78ba6d6 4875 [RTA_PREF] = { .type = NLA_U8 },
19e42e45
RP		 4876 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4877 [RTA_ENCAP] = { .type = NLA_NESTED },
32bc201e 4878 [RTA_EXPIRES] = { .type = NLA_U32 },
622ec2c9 4879 [RTA_UID] = { .type = NLA_U32 },
3b45a410 4880 [RTA_MARK] = { .type = NLA_U32 },
aa8f8778 4881 [RTA_TABLE] = { .type = NLA_U32 },
eacb9384
RP		 4882 [RTA_IP_PROTO] = { .type = NLA_U8 },
4883 [RTA_SPORT] = { .type = NLA_U16 },
4884 [RTA_DPORT] = { .type = NLA_U16 },
5b98324e 4885 [RTA_NH_ID] = { .type = NLA_U32 },
86872cb5
TG		 4886};
4887
4888static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
333c4301
DA		 4889 struct fib6_config *cfg,
4890 struct netlink_ext_ack *extack)
1da177e4 4891{
86872cb5
TG		 4892 struct rtmsg *rtm;
4893 struct nlattr *tb[RTA_MAX+1];
c78ba6d6 4894 unsigned int pref;
86872cb5 4895 int err;
1da177e4 4896
8cb08174
JB		 4897 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
4898 rtm_ipv6_policy, extack);
86872cb5
TG		 4899 if (err < 0)
4900 goto errout;
1da177e4 4901
86872cb5
TG		 4902 err = -EINVAL;
4903 rtm = nlmsg_data(nlh);
86872cb5 4904
84db8407
MŻ		 4905 *cfg = (struct fib6_config){
4906 .fc_table = rtm->rtm_table,
4907 .fc_dst_len = rtm->rtm_dst_len,
4908 .fc_src_len = rtm->rtm_src_len,
4909 .fc_flags = RTF_UP,
4910 .fc_protocol = rtm->rtm_protocol,
4911 .fc_type = rtm->rtm_type,
4912
4913 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
4914 .fc_nlinfo.nlh = nlh,
4915 .fc_nlinfo.nl_net = sock_net(skb->sk),
4916 };
86872cb5 4917
ef2c7d7b
ND		 4918 if (rtm->rtm_type == RTN_UNREACHABLE ||
4919 rtm->rtm_type == RTN_BLACKHOLE ||
b4949ab2
ND		 4920 rtm->rtm_type == RTN_PROHIBIT ||
4921 rtm->rtm_type == RTN_THROW)
86872cb5
TG		 4922 cfg->fc_flags |= RTF_REJECT;
4923
ab79ad14
MŻ		 4924 if (rtm->rtm_type == RTN_LOCAL)
4925 cfg->fc_flags |= RTF_LOCAL;
4926
1f56a01f
MKL		 4927 if (rtm->rtm_flags & RTM_F_CLONED)
4928 cfg->fc_flags |= RTF_CACHE;
4929
fc1e64e1
DA		 4930 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4931
5b98324e
DA		 4932 if (tb[RTA_NH_ID]) {
4933 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
4934 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
4935 NL_SET_ERR_MSG(extack,
4936 "Nexthop specification and nexthop id are mutually exclusive");
4937 goto errout;
4938 }
4939 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
4940 }
4941
86872cb5 4942 if (tb[RTA_GATEWAY]) {
67b61f6c 4943 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
86872cb5 4944 cfg->fc_flags |= RTF_GATEWAY;
1da177e4 4945 }
e3818541
DA		 4946 if (tb[RTA_VIA]) {
4947 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
4948 goto errout;
4949 }
86872cb5
TG		 4950
4951 if (tb[RTA_DST]) {
4952 int plen = (rtm->rtm_dst_len + 7) >> 3;
4953
4954 if (nla_len(tb[RTA_DST]) < plen)
4955 goto errout;
4956
4957 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1da177e4 4958 }
86872cb5
TG		 4959
4960 if (tb[RTA_SRC]) {
4961 int plen = (rtm->rtm_src_len + 7) >> 3;
4962
4963 if (nla_len(tb[RTA_SRC]) < plen)
4964 goto errout;
4965
4966 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1da177e4 4967 }
86872cb5 4968
c3968a85 4969 if (tb[RTA_PREFSRC])
67b61f6c 4970 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
c3968a85 4971
86872cb5
TG		 4972 if (tb[RTA_OIF])
4973 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4974
4975 if (tb[RTA_PRIORITY])
4976 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4977
4978 if (tb[RTA_METRICS]) {
4979 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4980 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
1da177e4 4981 }
86872cb5
TG		 4982
4983 if (tb[RTA_TABLE])
4984 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4985
51ebd318
ND		 4986 if (tb[RTA_MULTIPATH]) {
4987 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4988 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
9ed59592
DA		 4989
4990 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
c255bd68 4991 cfg->fc_mp_len, extack);
9ed59592
DA		 4992 if (err < 0)
4993 goto errout;
51ebd318
ND		 4994 }
4995
c78ba6d6
LR		 4996 if (tb[RTA_PREF]) {
4997 pref = nla_get_u8(tb[RTA_PREF]);
4998 if (pref != ICMPV6_ROUTER_PREF_LOW &&
4999 pref != ICMPV6_ROUTER_PREF_HIGH)
5000 pref = ICMPV6_ROUTER_PREF_MEDIUM;
5001 cfg->fc_flags |= RTF_PREF(pref);
5002 }
5003
19e42e45
RP		 5004 if (tb[RTA_ENCAP])
5005 cfg->fc_encap = tb[RTA_ENCAP];
5006
9ed59592 5007 if (tb[RTA_ENCAP_TYPE]) {
19e42e45
RP		 5008 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5009
c255bd68 5010 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
9ed59592
DA		 5011 if (err < 0)
5012 goto errout;
5013 }
5014
32bc201e
XL		 5015 if (tb[RTA_EXPIRES]) {
5016 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5017
5018 if (addrconf_finite_timeout(timeout)) {
5019 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5020 cfg->fc_flags |= RTF_EXPIRES;
5021 }
5022 }
5023
86872cb5
TG		 5024 err = 0;
5025errout:
5026 return err;
1da177e4
LT		 5027}
5028
6b9ea5a6 5029struct rt6_nh {
8d1c802b 5030 struct fib6_info *fib6_info;
6b9ea5a6 5031 struct fib6_config r_cfg;
6b9ea5a6
RP		 5032 struct list_head next;
5033};
5034
d4ead6b3
DA		 5035static int ip6_route_info_append(struct net *net,
5036 struct list_head *rt6_nh_list,
8d1c802b
DA		 5037 struct fib6_info *rt,
5038 struct fib6_config *r_cfg)
6b9ea5a6
RP		 5039{
5040 struct rt6_nh *nh;
6b9ea5a6
RP		 5041 int err = -EEXIST;
5042
5043 list_for_each_entry(nh, rt6_nh_list, next) {
8d1c802b
DA		 5044 /* check if fib6_info already exists */
5045 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
6b9ea5a6
RP		 5046 return err;
5047 }
5048
5049 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5050 if (!nh)
5051 return -ENOMEM;
8d1c802b 5052 nh->fib6_info = rt;
6b9ea5a6
RP		 5053 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5054 list_add_tail(&nh->next, rt6_nh_list);
5055
5056 return 0;
5057}
5058
8d1c802b
DA		 5059static void ip6_route_mpath_notify(struct fib6_info *rt,
5060 struct fib6_info *rt_last,
3b1137fe
DA		 5061 struct nl_info *info,
5062 __u16 nlflags)
5063{
5064 /* if this is an APPEND route, then rt points to the first route
5065 * inserted and rt_last points to last route inserted. Userspace
5066 * wants a consistent dump of the route which starts at the first
5067 * nexthop. Since sibling routes are always added at the end of
5068 * the list, find the first sibling of the last route appended
5069 */
93c2fb25
DA		 5070 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5071 rt = list_first_entry(&rt_last->fib6_siblings,
8d1c802b 5072 struct fib6_info,
93c2fb25 5073 fib6_siblings);
3b1137fe
DA		 5074 }
5075
5076 if (rt)
5077 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5078}
5079
0ee0f47c
IS		 5080static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5081{
5082 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5083 bool should_notify = false;
5084 struct fib6_info *leaf;
5085 struct fib6_node *fn;
5086
5087 rcu_read_lock();
5088 fn = rcu_dereference(rt->fib6_node);
5089 if (!fn)
5090 goto out;
5091
5092 leaf = rcu_dereference(fn->leaf);
5093 if (!leaf)
5094 goto out;
5095
5096 if (rt == leaf ||
5097 (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5098 rt6_qualify_for_ecmp(leaf)))
5099 should_notify = true;
5100out:
5101 rcu_read_unlock();
5102
5103 return should_notify;
5104}
5105
333c4301
DA		 5106static int ip6_route_multipath_add(struct fib6_config *cfg,
5107 struct netlink_ext_ack *extack)
51ebd318 5108{
8d1c802b 5109 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
3b1137fe 5110 struct nl_info *info = &cfg->fc_nlinfo;
51ebd318
ND		 5111 struct fib6_config r_cfg;
5112 struct rtnexthop *rtnh;
8d1c802b 5113 struct fib6_info *rt;
6b9ea5a6
RP		 5114 struct rt6_nh *err_nh;
5115 struct rt6_nh *nh, *nh_safe;
3b1137fe 5116 __u16 nlflags;
51ebd318
ND		 5117 int remaining;
5118 int attrlen;
6b9ea5a6
RP		 5119 int err = 1;
5120 int nhn = 0;
5121 int replace = (cfg->fc_nlinfo.nlh &&
5122 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5123 LIST_HEAD(rt6_nh_list);
51ebd318 5124
3b1137fe
DA		 5125 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5126 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5127 nlflags |= NLM_F_APPEND;
5128
35f1b4e9 5129 remaining = cfg->fc_mp_len;
51ebd318 5130 rtnh = (struct rtnexthop *)cfg->fc_mp;
51ebd318 5131
6b9ea5a6 5132 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
8d1c802b 5133 * fib6_info structs per nexthop
6b9ea5a6 5134 */
51ebd318
ND		 5135 while (rtnh_ok(rtnh, remaining)) {
5136 memcpy(&r_cfg, cfg, sizeof(*cfg));
5137 if (rtnh->rtnh_ifindex)
5138 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5139
5140 attrlen = rtnh_attrlen(rtnh);
5141 if (attrlen > 0) {
5142 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5143
5144 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5145 if (nla) {
67b61f6c 5146 r_cfg.fc_gateway = nla_get_in6_addr(nla);
51ebd318
ND		 5147 r_cfg.fc_flags |= RTF_GATEWAY;
5148 }
19e42e45
RP		 5149 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5150 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5151 if (nla)
5152 r_cfg.fc_encap_type = nla_get_u16(nla);
51ebd318 5153 }
6b9ea5a6 5154
68e2ffde 5155 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
acb54e3c 5156 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
8c5b83f0
RP		 5157 if (IS_ERR(rt)) {
5158 err = PTR_ERR(rt);
5159 rt = NULL;
6b9ea5a6 5160 goto cleanup;
8c5b83f0 5161 }
b5d2d75e
DA		 5162 if (!rt6_qualify_for_ecmp(rt)) {
5163 err = -EINVAL;
5164 NL_SET_ERR_MSG(extack,
5165 "Device only routes can not be added for IPv6 using the multipath API.");
5166 fib6_info_release(rt);
5167 goto cleanup;
5168 }
6b9ea5a6 5169
1cf844c7 5170 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
398958ae 5171
d4ead6b3
DA		 5172 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5173 rt, &r_cfg);
51ebd318 5174 if (err) {
93531c67 5175 fib6_info_release(rt);
6b9ea5a6
RP		 5176 goto cleanup;
5177 }
5178
5179 rtnh = rtnh_next(rtnh, &remaining);
5180 }
5181
9eee3b49
IS		 5182 if (list_empty(&rt6_nh_list)) {
5183 NL_SET_ERR_MSG(extack,
5184 "Invalid nexthop configuration - no valid nexthops");
5185 return -EINVAL;
5186 }
5187
3b1137fe
DA		 5188 /* for add and replace send one notification with all nexthops.
5189 * Skip the notification in fib6_add_rt2node and send one with
5190 * the full route when done
5191 */
5192 info->skip_notify = 1;
5193
ebee3cad
IS		 5194 /* For add and replace, send one notification with all nexthops. For
5195 * append, send one notification with all appended nexthops.
5196 */
5197 info->skip_notify_kernel = 1;
5198
6b9ea5a6
RP		 5199 err_nh = NULL;
5200 list_for_each_entry(nh, &rt6_nh_list, next) {
8d1c802b
DA		 5201 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5202 fib6_info_release(nh->fib6_info);
93531c67 5203
f7225172
DA		 5204 if (!err) {
5205 /* save reference to last route successfully inserted */
5206 rt_last = nh->fib6_info;
5207
5208 /* save reference to first route for notification */
5209 if (!rt_notif)
5210 rt_notif = nh->fib6_info;
5211 }
3b1137fe 5212
8d1c802b
DA		 5213 /* nh->fib6_info is used or freed at this point, reset to NULL*/
5214 nh->fib6_info = NULL;
6b9ea5a6
RP		 5215 if (err) {
5216 if (replace && nhn)
a5a82d84
JK		 5217 NL_SET_ERR_MSG_MOD(extack,
5218 "multipath route replace failed (check consistency of installed routes)");
6b9ea5a6
RP		 5219 err_nh = nh;
5220 goto add_errout;
51ebd318 5221 }
6b9ea5a6 5222
1a72418b 5223 /* Because each route is added like a single route we remove
27596472
MK		 5224 * these flags after the first nexthop: if there is a collision,
5225 * we have already failed to add the first nexthop:
5226 * fib6_add_rt2node() has rejected it; when replacing, old
5227 * nexthops have been replaced by first new, the rest should
5228 * be added to it.
1a72418b 5229 */
27596472
MK		 5230 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5231 NLM_F_REPLACE);
afecdb37 5232 cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
6b9ea5a6
RP		 5233 nhn++;
5234 }
5235
0ee0f47c
IS		 5236 /* An in-kernel notification should only be sent in case the new
5237 * multipath route is added as the first route in the node, or if
5238 * it was appended to it. We pass 'rt_notif' since it is the first
5239 * sibling and might allow us to skip some checks in the replace case.
5240 */
5241 if (ip6_route_mpath_should_notify(rt_notif)) {
5242 enum fib_event_type fib_event;
5243
5244 if (rt_notif->fib6_nsiblings != nhn - 1)
5245 fib_event = FIB_EVENT_ENTRY_APPEND;
5246 else
caafb250 5247 fib_event = FIB_EVENT_ENTRY_REPLACE;
0ee0f47c
IS		 5248
5249 err = call_fib6_multipath_entry_notifiers(info->nl_net,
5250 fib_event, rt_notif,
5251 nhn - 1, extack);
5252 if (err) {
5253 /* Delete all the siblings that were just added */
5254 err_nh = NULL;
5255 goto add_errout;
5256 }
5257 }
ebee3cad 5258
3b1137fe
DA		 5259 /* success ... tell user about new route */
5260 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
6b9ea5a6
RP		 5261 goto cleanup;
5262
5263add_errout:
3b1137fe
DA		 5264 /* send notification for routes that were added so that
5265 * the delete notifications sent by ip6_route_del are
5266 * coherent
5267 */
5268 if (rt_notif)
5269 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5270
6b9ea5a6
RP		 5271 /* Delete routes that were already added */
5272 list_for_each_entry(nh, &rt6_nh_list, next) {
5273 if (err_nh == nh)
5274 break;
333c4301 5275 ip6_route_del(&nh->r_cfg, extack);
6b9ea5a6
RP		 5276 }
5277
5278cleanup:
5279 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
8d1c802b
DA		 5280 if (nh->fib6_info)
5281 fib6_info_release(nh->fib6_info);
6b9ea5a6
RP		 5282 list_del(&nh->next);
5283 kfree(nh);
5284 }
5285
5286 return err;
5287}
5288
333c4301
DA		 5289static int ip6_route_multipath_del(struct fib6_config *cfg,
5290 struct netlink_ext_ack *extack)
6b9ea5a6
RP		 5291{
5292 struct fib6_config r_cfg;
5293 struct rtnexthop *rtnh;
2291267e 5294 int last_err = 0;
6b9ea5a6
RP		 5295 int remaining;
5296 int attrlen;
2291267e 5297 int err;
6b9ea5a6
RP		 5298
5299 remaining = cfg->fc_mp_len;
5300 rtnh = (struct rtnexthop *)cfg->fc_mp;
5301
5302 /* Parse a Multipath Entry */
5303 while (rtnh_ok(rtnh, remaining)) {
5304 memcpy(&r_cfg, cfg, sizeof(*cfg));
5305 if (rtnh->rtnh_ifindex)
5306 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5307
5308 attrlen = rtnh_attrlen(rtnh);
5309 if (attrlen > 0) {
5310 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5311
5312 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5313 if (nla) {
5314 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
5315 r_cfg.fc_flags |= RTF_GATEWAY;
5316 }
5317 }
333c4301 5318 err = ip6_route_del(&r_cfg, extack);
6b9ea5a6
RP		 5319 if (err)
5320 last_err = err;
5321
51ebd318
ND		 5322 rtnh = rtnh_next(rtnh, &remaining);
5323 }
5324
5325 return last_err;
5326}
5327
c21ef3e3
DA		 5328static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5329 struct netlink_ext_ack *extack)
1da177e4 5330{
86872cb5
TG		 5331 struct fib6_config cfg;
5332 int err;
1da177e4 5333
333c4301 5334 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 5335 if (err < 0)
5336 return err;
5337
5b98324e
DA		 5338 if (cfg.fc_nh_id &&
5339 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5340 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5341 return -EINVAL;
5342 }
5343
51ebd318 5344 if (cfg.fc_mp)
333c4301 5345 return ip6_route_multipath_del(&cfg, extack);
0ae81335
DA		 5346 else {
5347 cfg.fc_delete_all_nh = 1;
333c4301 5348 return ip6_route_del(&cfg, extack);
0ae81335 5349 }
1da177e4
LT		 5350}
5351
c21ef3e3
DA		 5352static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5353 struct netlink_ext_ack *extack)
1da177e4 5354{
86872cb5
TG		 5355 struct fib6_config cfg;
5356 int err;
1da177e4 5357
333c4301 5358 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 5359 if (err < 0)
5360 return err;
5361
67f69513
DA		 5362 if (cfg.fc_metric == 0)
5363 cfg.fc_metric = IP6_RT_PRIO_USER;
5364
51ebd318 5365 if (cfg.fc_mp)
333c4301 5366 return ip6_route_multipath_add(&cfg, extack);
51ebd318 5367 else
acb54e3c 5368 return ip6_route_add(&cfg, GFP_KERNEL, extack);
1da177e4
LT		 5369}
5370
a1b7a1f0
DA		 5371/* add the overhead of this fib6_nh to nexthop_len */
5372static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
339bf98f 5373{
a1b7a1f0 5374 int *nexthop_len = arg;
beb1afac 5375
a1b7a1f0
DA		 5376 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5377 + NLA_ALIGN(sizeof(struct rtnexthop))
5378 + nla_total_size(16); /* RTA_GATEWAY */
5379
5380 if (nh->fib_nh_lws) {
5381 /* RTA_ENCAP_TYPE */
5382 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5383 /* RTA_ENCAP */
5384 *nexthop_len += nla_total_size(2);
5385 }
f88d8ea6 5386
a1b7a1f0
DA		 5387 return 0;
5388}
beb1afac 5389
a1b7a1f0
DA		 5390static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5391{
5392 int nexthop_len;
5393
5394 if (f6i->nh) {
5395 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5396 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5397 &nexthop_len);
5398 } else {
5399 struct fib6_nh *nh = f6i->fib6_nh;
5400
5401 nexthop_len = 0;
5402 if (f6i->fib6_nsiblings) {
5403 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
5404 + NLA_ALIGN(sizeof(struct rtnexthop))
5405 + nla_total_size(16) /* RTA_GATEWAY */
5406 + lwtunnel_get_encap_size(nh->fib_nh_lws);
5407
5408 nexthop_len *= f6i->fib6_nsiblings;
5409 }
5410 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
beb1afac
DA		 5411 }
5412
339bf98f
TG		 5413 return NLMSG_ALIGN(sizeof(struct rtmsg))
5414 + nla_total_size(16) /* RTA_SRC */
5415 + nla_total_size(16) /* RTA_DST */
5416 + nla_total_size(16) /* RTA_GATEWAY */
5417 + nla_total_size(16) /* RTA_PREFSRC */
5418 + nla_total_size(4) /* RTA_TABLE */
5419 + nla_total_size(4) /* RTA_IIF */
5420 + nla_total_size(4) /* RTA_OIF */
5421 + nla_total_size(4) /* RTA_PRIORITY */
6a2b9ce0 5422 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
ea697639 5423 + nla_total_size(sizeof(struct rta_cacheinfo))
c78ba6d6 5424 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
19e42e45 5425 + nla_total_size(1) /* RTA_PREF */
beb1afac
DA		 5426 + nexthop_len;
5427}
5428
f88d8ea6
DA		 5429static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5430 unsigned char *flags)
5431{
5432 if (nexthop_is_multipath(nh)) {
5433 struct nlattr *mp;
5434
4255ff05 5435 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
f88d8ea6
DA		 5436 if (!mp)
5437 goto nla_put_failure;
5438
7bdf4de1 5439 if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
f88d8ea6
DA		 5440 goto nla_put_failure;
5441
5442 nla_nest_end(skb, mp);
5443 } else {
5444 struct fib6_nh *fib6_nh;
5445
5446 fib6_nh = nexthop_fib6_nh(nh);
7bdf4de1 5447 if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
f88d8ea6
DA		 5448 flags, false) < 0)
5449 goto nla_put_failure;
5450 }
5451
5452 return 0;
5453
5454nla_put_failure:
5455 return -EMSGSIZE;
5456}
5457
d4ead6b3 5458static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 5459 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 5460 struct in6_addr *dest, struct in6_addr *src,
15e47304 5461 int iif, int type, u32 portid, u32 seq,
f8cfe2ce 5462 unsigned int flags)
1da177e4 5463{
22d0bd82
XL		 5464 struct rt6_info *rt6 = (struct rt6_info *)dst;
5465 struct rt6key *rt6_dst, *rt6_src;
5466 u32 *pmetrics, table, rt6_flags;
f88d8ea6 5467 unsigned char nh_flags = 0;
2d7202bf 5468 struct nlmsghdr *nlh;
22d0bd82 5469 struct rtmsg *rtm;
d4ead6b3 5470 long expires = 0;
1da177e4 5471
15e47304 5472 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
38308473 5473 if (!nlh)
26932566 5474 return -EMSGSIZE;
2d7202bf 5475
22d0bd82
XL		 5476 if (rt6) {
5477 rt6_dst = &rt6->rt6i_dst;
5478 rt6_src = &rt6->rt6i_src;
5479 rt6_flags = rt6->rt6i_flags;
5480 } else {
5481 rt6_dst = &rt->fib6_dst;
5482 rt6_src = &rt->fib6_src;
5483 rt6_flags = rt->fib6_flags;
5484 }
5485
2d7202bf 5486 rtm = nlmsg_data(nlh);
1da177e4 5487 rtm->rtm_family = AF_INET6;
22d0bd82
XL		 5488 rtm->rtm_dst_len = rt6_dst->plen;
5489 rtm->rtm_src_len = rt6_src->plen;
1da177e4 5490 rtm->rtm_tos = 0;
93c2fb25
DA		 5491 if (rt->fib6_table)
5492 table = rt->fib6_table->tb6_id;
c71099ac 5493 else
9e762a4a 5494 table = RT6_TABLE_UNSPEC;
97f0082a 5495 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
c78679e8
DM		 5496 if (nla_put_u32(skb, RTA_TABLE, table))
5497 goto nla_put_failure;
e8478e80
DA		 5498
5499 rtm->rtm_type = rt->fib6_type;
1da177e4
LT		 5500 rtm->rtm_flags = 0;
5501 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
93c2fb25 5502 rtm->rtm_protocol = rt->fib6_protocol;
1da177e4 5503
22d0bd82 5504 if (rt6_flags & RTF_CACHE)
1da177e4
LT		 5505 rtm->rtm_flags |= RTM_F_CLONED;
5506
d4ead6b3
DA		 5507 if (dest) {
5508 if (nla_put_in6_addr(skb, RTA_DST, dest))
c78679e8 5509 goto nla_put_failure;
1ab1457c 5510 rtm->rtm_dst_len = 128;
1da177e4 5511 } else if (rtm->rtm_dst_len)
22d0bd82 5512 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
c78679e8 5513 goto nla_put_failure;
1da177e4
LT		 5514#ifdef CONFIG_IPV6_SUBTREES
5515 if (src) {
930345ea 5516 if (nla_put_in6_addr(skb, RTA_SRC, src))
c78679e8 5517 goto nla_put_failure;
1ab1457c 5518 rtm->rtm_src_len = 128;
c78679e8 5519 } else if (rtm->rtm_src_len &&
22d0bd82 5520 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
c78679e8 5521 goto nla_put_failure;
1da177e4 5522#endif
7bc570c8
YH		 5523 if (iif) {
5524#ifdef CONFIG_IPV6_MROUTE
22d0bd82 5525 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
fd61c6ba
DA		 5526 int err = ip6mr_get_route(net, skb, rtm, portid);
5527
5528 if (err == 0)
5529 return 0;
5530 if (err < 0)
5531 goto nla_put_failure;
7bc570c8
YH		 5532 } else
5533#endif
c78679e8
DM		 5534 if (nla_put_u32(skb, RTA_IIF, iif))
5535 goto nla_put_failure;
d4ead6b3 5536 } else if (dest) {
1da177e4 5537 struct in6_addr saddr_buf;
d4ead6b3 5538 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
930345ea 5539 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 5540 goto nla_put_failure;
1da177e4 5541 }
2d7202bf 5542
93c2fb25 5543 if (rt->fib6_prefsrc.plen) {
c3968a85 5544 struct in6_addr saddr_buf;
93c2fb25 5545 saddr_buf = rt->fib6_prefsrc.addr;
930345ea 5546 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 5547 goto nla_put_failure;
c3968a85
DW		 5548 }
5549
d4ead6b3
DA		 5550 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5551 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
2d7202bf
TG		 5552 goto nla_put_failure;
5553
93c2fb25 5554 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
c78679e8 5555 goto nla_put_failure;
8253947e 5556
beb1afac
DA		 5557 /* For multipath routes, walk the siblings list and add
5558 * each as a nexthop within RTA_MULTIPATH.
5559 */
22d0bd82
XL		 5560 if (rt6) {
5561 if (rt6_flags & RTF_GATEWAY &&
5562 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5563 goto nla_put_failure;
5564
5565 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5566 goto nla_put_failure;
6b13d8f7
OH		 5567
5568 if (dst->lwtstate &&
5569 lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
5570 goto nla_put_failure;
22d0bd82 5571 } else if (rt->fib6_nsiblings) {
8d1c802b 5572 struct fib6_info *sibling, *next_sibling;
beb1afac
DA		 5573 struct nlattr *mp;
5574
ae0be8de 5575 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
beb1afac
DA		 5576 if (!mp)
5577 goto nla_put_failure;
5578
1cf844c7 5579 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
7bdf4de1 5580 rt->fib6_nh->fib_nh_weight, AF_INET6) < 0)
beb1afac
DA		 5581 goto nla_put_failure;
5582
5583 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25 5584 &rt->fib6_siblings, fib6_siblings) {
1cf844c7 5585 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
7bdf4de1
DS		 5586 sibling->fib6_nh->fib_nh_weight,
5587 AF_INET6) < 0)
beb1afac
DA		 5588 goto nla_put_failure;
5589 }
5590
5591 nla_nest_end(skb, mp);
f88d8ea6
DA		 5592 } else if (rt->nh) {
5593 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5594 goto nla_put_failure;
ecc5663c 5595
f88d8ea6
DA		 5596 if (nexthop_is_blackhole(rt->nh))
5597 rtm->rtm_type = RTN_BLACKHOLE;
5598
4f80116d
RP		 5599 if (net->ipv4.sysctl_nexthop_compat_mode &&
5600 rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
f88d8ea6
DA		 5601 goto nla_put_failure;
5602
5603 rtm->rtm_flags |= nh_flags;
5604 } else {
7bdf4de1 5605 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
ecc5663c 5606 &nh_flags, false) < 0)
beb1afac 5607 goto nla_put_failure;
ecc5663c
DA		 5608
5609 rtm->rtm_flags |= nh_flags;
beb1afac
DA		 5610 }
5611
22d0bd82 5612 if (rt6_flags & RTF_EXPIRES) {
14895687
DA		 5613 expires = dst ? dst->expires : rt->expires;
5614 expires -= jiffies;
5615 }
69cdf8f9 5616
bb3c4ab9
IS		 5617 if (!dst) {
5618 if (rt->offload)
5619 rtm->rtm_flags |= RTM_F_OFFLOAD;
5620 if (rt->trap)
5621 rtm->rtm_flags |= RTM_F_TRAP;
5622 }
5623
d4ead6b3 5624 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
e3703b3d 5625 goto nla_put_failure;
2d7202bf 5626
22d0bd82 5627 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
c78ba6d6
LR		 5628 goto nla_put_failure;
5629
19e42e45 5630
053c095a
JB		 5631 nlmsg_end(skb, nlh);
5632 return 0;
2d7202bf
TG		 5633
5634nla_put_failure:
26932566
PM		 5635 nlmsg_cancel(skb, nlh);
5636 return -EMSGSIZE;
1da177e4
LT		 5637}
5638
2c170e07
DA		 5639static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5640{
5641 const struct net_device *dev = arg;
5642
5643 if (nh->fib_nh_dev == dev)
5644 return 1;
5645
5646 return 0;
5647}
5648
13e38901
DA		 5649static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5650 const struct net_device *dev)
5651{
2c170e07
DA		 5652 if (f6i->nh) {
5653 struct net_device *_dev = (struct net_device *)dev;
5654
5655 return !!nexthop_for_each_fib6_nh(f6i->nh,
5656 fib6_info_nh_uses_dev,
5657 _dev);
5658 }
5659
1cf844c7 5660 if (f6i->fib6_nh->fib_nh_dev == dev)
13e38901
DA		 5661 return true;
5662
5663 if (f6i->fib6_nsiblings) {
5664 struct fib6_info *sibling, *next_sibling;
5665
5666 list_for_each_entry_safe(sibling, next_sibling,
5667 &f6i->fib6_siblings, fib6_siblings) {
1cf844c7 5668 if (sibling->fib6_nh->fib_nh_dev == dev)
13e38901
DA		 5669 return true;
5670 }
5671 }
5672
5673 return false;
5674}
5675
1e47b483
SB		 5676struct fib6_nh_exception_dump_walker {
5677 struct rt6_rtnl_dump_arg *dump;
5678 struct fib6_info *rt;
5679 unsigned int flags;
5680 unsigned int skip;
5681 unsigned int count;
5682};
5683
5684static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5685{
5686 struct fib6_nh_exception_dump_walker *w = arg;
5687 struct rt6_rtnl_dump_arg *dump = w->dump;
5688 struct rt6_exception_bucket *bucket;
5689 struct rt6_exception *rt6_ex;
5690 int i, err;
5691
5692 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5693 if (!bucket)
5694 return 0;
5695
5696 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5697 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5698 if (w->skip) {
5699 w->skip--;
5700 continue;
5701 }
5702
5703 /* Expiration of entries doesn't bump sernum, insertion
5704 * does. Removal is triggered by insertion, so we can
5705 * rely on the fact that if entries change between two
5706 * partial dumps, this node is scanned again completely,
5707 * see rt6_insert_exception() and fib6_dump_table().
5708 *
5709 * Count expired entries we go through as handled
5710 * entries that we'll skip next time, in case of partial
5711 * node dump. Otherwise, if entries expire meanwhile,
5712 * we'll skip the wrong amount.
5713 */
5714 if (rt6_check_expired(rt6_ex->rt6i)) {
5715 w->count++;
5716 continue;
5717 }
5718
5719 err = rt6_fill_node(dump->net, dump->skb, w->rt,
5720 &rt6_ex->rt6i->dst, NULL, NULL, 0,
5721 RTM_NEWROUTE,
5722 NETLINK_CB(dump->cb->skb).portid,
5723 dump->cb->nlh->nlmsg_seq, w->flags);
5724 if (err)
5725 return err;
5726
5727 w->count++;
5728 }
5729 bucket++;
5730 }
5731
5732 return 0;
5733}
5734
bf9a8a06 5735/* Return -1 if done with node, number of handled routes on partial dump */
1e47b483 5736int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
1da177e4
LT		 5737{
5738 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
13e38901
DA		 5739 struct fib_dump_filter *filter = &arg->filter;
5740 unsigned int flags = NLM_F_MULTI;
1f17e2f2 5741 struct net *net = arg->net;
1e47b483 5742 int count = 0;
1f17e2f2 5743
421842ed 5744 if (rt == net->ipv6.fib6_null_entry)
bf9a8a06 5745 return -1;
1da177e4 5746
13e38901
DA		 5747 if ((filter->flags & RTM_F_PREFIX) &&
5748 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5749 /* success since this is not a prefix route */
bf9a8a06 5750 return -1;
13e38901 5751 }
1e47b483
SB		 5752 if (filter->filter_set &&
5753 ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5754 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5755 (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5756 return -1;
5757 }
5758
5759 if (filter->filter_set ||
5760 !filter->dump_routes || !filter->dump_exceptions) {
13e38901 5761 flags |= NLM_F_DUMP_FILTERED;
f8cfe2ce 5762 }
1da177e4 5763
1e47b483
SB		 5764 if (filter->dump_routes) {
5765 if (skip) {
5766 skip--;
5767 } else {
5768 if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5769 0, RTM_NEWROUTE,
5770 NETLINK_CB(arg->cb->skb).portid,
5771 arg->cb->nlh->nlmsg_seq, flags)) {
5772 return 0;
5773 }
5774 count++;
5775 }
5776 }
5777
5778 if (filter->dump_exceptions) {
5779 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5780 .rt = rt,
5781 .flags = flags,
5782 .skip = skip,
5783 .count = 0 };
5784 int err;
5785
3b525691 5786 rcu_read_lock();
1e47b483
SB		 5787 if (rt->nh) {
5788 err = nexthop_for_each_fib6_nh(rt->nh,
5789 rt6_nh_dump_exceptions,
5790 &w);
5791 } else {
5792 err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5793 }
3b525691 5794 rcu_read_unlock();
1e47b483
SB		 5795
5796 if (err)
5797 return count += w.count;
5798 }
bf9a8a06
SB		 5799
5800 return -1;
1da177e4
LT		 5801}
5802
0eff0a27
JK		 5803static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5804 const struct nlmsghdr *nlh,
5805 struct nlattr **tb,
5806 struct netlink_ext_ack *extack)
5807{
5808 struct rtmsg *rtm;
5809 int i, err;
5810
5811 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5812 NL_SET_ERR_MSG_MOD(extack,
5813 "Invalid header for get route request");
5814 return -EINVAL;
5815 }
5816
5817 if (!netlink_strict_get_check(skb))
8cb08174
JB		 5818 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5819 rtm_ipv6_policy, extack);
0eff0a27
JK		 5820
5821 rtm = nlmsg_data(nlh);
5822 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5823 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5824 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5825 rtm->rtm_type) {
5826 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5827 return -EINVAL;
5828 }
5829 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5830 NL_SET_ERR_MSG_MOD(extack,
5831 "Invalid flags for get route request");
5832 return -EINVAL;
5833 }
5834
8cb08174
JB		 5835 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5836 rtm_ipv6_policy, extack);
0eff0a27
JK		 5837 if (err)
5838 return err;
5839
5840 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5841 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5842 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5843 return -EINVAL;
5844 }
5845
5846 for (i = 0; i <= RTA_MAX; i++) {
5847 if (!tb[i])
5848 continue;
5849
5850 switch (i) {
5851 case RTA_SRC:
5852 case RTA_DST:
5853 case RTA_IIF:
5854 case RTA_OIF:
5855 case RTA_MARK:
5856 case RTA_UID:
5857 case RTA_SPORT:
5858 case RTA_DPORT:
5859 case RTA_IP_PROTO:
5860 break;
5861 default:
5862 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
5863 return -EINVAL;
5864 }
5865 }
5866
5867 return 0;
5868}
5869
c21ef3e3
DA		 5870static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5871 struct netlink_ext_ack *extack)
1da177e4 5872{
3b1e0a65 5873 struct net *net = sock_net(in_skb->sk);
ab364a6f 5874 struct nlattr *tb[RTA_MAX+1];
18c3a61c 5875 int err, iif = 0, oif = 0;
a68886a6 5876 struct fib6_info *from;
18c3a61c 5877 struct dst_entry *dst;
ab364a6f 5878 struct rt6_info *rt;
1da177e4 5879 struct sk_buff *skb;
ab364a6f 5880 struct rtmsg *rtm;
744486d4 5881 struct flowi6 fl6 = {};
18c3a61c 5882 bool fibmatch;
1da177e4 5883
0eff0a27 5884 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
ab364a6f
TG		 5885 if (err < 0)
5886 goto errout;
1da177e4 5887
ab364a6f 5888 err = -EINVAL;
38b7097b
HFS		 5889 rtm = nlmsg_data(nlh);
5890 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
18c3a61c 5891 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
1da177e4 5892
ab364a6f
TG		 5893 if (tb[RTA_SRC]) {
5894 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
5895 goto errout;
5896
4e3fd7a0 5897 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
ab364a6f
TG		 5898 }
5899
5900 if (tb[RTA_DST]) {
5901 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
5902 goto errout;
5903
4e3fd7a0 5904 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
ab364a6f
TG		 5905 }
5906
5907 if (tb[RTA_IIF])
5908 iif = nla_get_u32(tb[RTA_IIF]);
5909
5910 if (tb[RTA_OIF])
72331bc0 5911 oif = nla_get_u32(tb[RTA_OIF]);
1da177e4 5912
2e47b291
LC		 5913 if (tb[RTA_MARK])
5914 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
5915
622ec2c9
LC		 5916 if (tb[RTA_UID])
5917 fl6.flowi6_uid = make_kuid(current_user_ns(),
5918 nla_get_u32(tb[RTA_UID]));
5919 else
5920 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
5921
eacb9384
RP		 5922 if (tb[RTA_SPORT])
5923 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
5924
5925 if (tb[RTA_DPORT])
5926 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
5927
5928 if (tb[RTA_IP_PROTO]) {
5929 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
5e1a99ea
HL		 5930 &fl6.flowi6_proto, AF_INET6,
5931 extack);
eacb9384
RP		 5932 if (err)
5933 goto errout;
5934 }
5935
1da177e4
LT		 5936 if (iif) {
5937 struct net_device *dev;
72331bc0
SL		 5938 int flags = 0;
5939
121622db
FW		 5940 rcu_read_lock();
5941
5942 dev = dev_get_by_index_rcu(net, iif);
1da177e4 5943 if (!dev) {
121622db 5944 rcu_read_unlock();
1da177e4 5945 err = -ENODEV;
ab364a6f 5946 goto errout;
1da177e4 5947 }
72331bc0
SL		 5948
5949 fl6.flowi6_iif = iif;
5950
5951 if (!ipv6_addr_any(&fl6.saddr))
5952 flags |= RT6_LOOKUP_F_HAS_SADDR;
5953
b75cc8f9 5954 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
121622db
FW		 5955
5956 rcu_read_unlock();
72331bc0
SL		 5957 } else {
5958 fl6.flowi6_oif = oif;
5959
58acfd71 5960 dst = ip6_route_output(net, NULL, &fl6);
18c3a61c
RP		 5961 }
5962
18c3a61c
RP		 5963
5964 rt = container_of(dst, struct rt6_info, dst);
5965 if (rt->dst.error) {
5966 err = rt->dst.error;
5967 ip6_rt_put(rt);
5968 goto errout;
1da177e4
LT		 5969 }
5970
9d6acb3b
WC		 5971 if (rt == net->ipv6.ip6_null_entry) {
5972 err = rt->dst.error;
5973 ip6_rt_put(rt);
5974 goto errout;
5975 }
5976
ab364a6f 5977 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
38308473 5978 if (!skb) {
94e187c0 5979 ip6_rt_put(rt);
ab364a6f
TG		 5980 err = -ENOBUFS;
5981 goto errout;
5982 }
1da177e4 5983
d8d1f30b 5984 skb_dst_set(skb, &rt->dst);
a68886a6
DA		 5985
5986 rcu_read_lock();
5987 from = rcu_dereference(rt->from);
886b7a50
MKL		 5988 if (from) {
5989 if (fibmatch)
5990 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
5991 iif, RTM_NEWROUTE,
5992 NETLINK_CB(in_skb).portid,
5993 nlh->nlmsg_seq, 0);
5994 else
5995 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
5996 &fl6.saddr, iif, RTM_NEWROUTE,
5997 NETLINK_CB(in_skb).portid,
5998 nlh->nlmsg_seq, 0);
5999 } else {
6000 err = -ENETUNREACH;
6001 }
a68886a6
DA		 6002 rcu_read_unlock();
6003
1da177e4 6004 if (err < 0) {
ab364a6f
TG		 6005 kfree_skb(skb);
6006 goto errout;
1da177e4
LT		 6007 }
6008
15e47304 6009 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
ab364a6f 6010errout:
1da177e4 6011 return err;
1da177e4
LT		 6012}
6013
8d1c802b 6014void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
37a1d361 6015 unsigned int nlm_flags)
1da177e4
LT		 6016{
6017 struct sk_buff *skb;
5578689a 6018 struct net *net = info->nl_net;
528c4ceb
DL		 6019 u32 seq;
6020 int err;
6021
6022 err = -ENOBUFS;
38308473 6023 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
86872cb5 6024
19e42e45 6025 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
38308473 6026 if (!skb)
21713ebc
TG		 6027 goto errout;
6028
d4ead6b3
DA		 6029 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6030 event, info->portid, seq, nlm_flags);
26932566
PM		 6031 if (err < 0) {
6032 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6033 WARN_ON(err == -EMSGSIZE);
6034 kfree_skb(skb);
6035 goto errout;
6036 }
15e47304 6037 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
1ce85fe4
PNA		 6038 info->nlh, gfp_any());
6039 return;
21713ebc
TG		 6040errout:
6041 if (err < 0)
5578689a 6042 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
1da177e4
LT		 6043}
6044
19a3b7ee
DA		 6045void fib6_rt_update(struct net *net, struct fib6_info *rt,
6046 struct nl_info *info)
6047{
6048 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6049 struct sk_buff *skb;
6050 int err = -ENOBUFS;
6051
19a3b7ee
DA		 6052 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6053 if (!skb)
6054 goto errout;
6055
6056 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6057 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6058 if (err < 0) {
6059 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6060 WARN_ON(err == -EMSGSIZE);
6061 kfree_skb(skb);
6062 goto errout;
6063 }
6064 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6065 info->nlh, gfp_any());
6066 return;
6067errout:
6068 if (err < 0)
6069 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6070}
6071
907eea48
AC		 6072void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
6073 bool offload, bool trap)
6074{
6075 struct sk_buff *skb;
6076 int err;
6077
6078 if (f6i->offload == offload && f6i->trap == trap)
6079 return;
6080
6081 f6i->offload = offload;
6082 f6i->trap = trap;
6083
6084 if (!rcu_access_pointer(f6i->fib6_node))
6085 /* The route was removed from the tree, do not send
6086 * notfication.
6087 */
6088 return;
6089
6090 if (!net->ipv6.sysctl.fib_notify_on_flag_change)
6091 return;
6092
6093 skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
6094 if (!skb) {
6095 err = -ENOBUFS;
6096 goto errout;
6097 }
6098
6099 err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
6100 0, 0);
6101 if (err < 0) {
6102 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6103 WARN_ON(err == -EMSGSIZE);
6104 kfree_skb(skb);
6105 goto errout;
6106 }
6107
6108 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
6109 return;
6110
6111errout:
6112 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6113}
6114EXPORT_SYMBOL(fib6_info_hw_flags_set);
6115
8ed67789 6116static int ip6_route_dev_notify(struct notifier_block *this,
351638e7 6117 unsigned long event, void *ptr)
8ed67789 6118{
351638e7 6119 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
c346dca1 6120 struct net *net = dev_net(dev);
8ed67789 6121
242d3a49
WC		 6122 if (!(dev->flags & IFF_LOOPBACK))
6123 return NOTIFY_OK;
6124
6125 if (event == NETDEV_REGISTER) {
1cf844c7 6126 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
d8d1f30b 6127 net->ipv6.ip6_null_entry->dst.dev = dev;
8ed67789
DL		 6128 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6129#ifdef CONFIG_IPV6_MULTIPLE_TABLES
d8d1f30b 6130 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
8ed67789 6131 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
d8d1f30b 6132 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
8ed67789 6133 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
242d3a49 6134#endif
76da0704
WC		 6135 } else if (event == NETDEV_UNREGISTER &&
6136 dev->reg_state != NETREG_UNREGISTERED) {
6137 /* NETDEV_UNREGISTER could be fired for multiple times by
6138 * netdev_wait_allrefs(). Make sure we only call this once.
6139 */
12d94a80 6140 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
242d3a49 6141#ifdef CONFIG_IPV6_MULTIPLE_TABLES
12d94a80
ED		 6142 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6143 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
8ed67789
DL		 6144#endif
6145 }
6146
6147 return NOTIFY_OK;
6148}
6149
1da177e4
LT		 6150/*
6151 * /proc
6152 */
6153
6154#ifdef CONFIG_PROC_FS
1da177e4
LT		 6155static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6156{
69ddb805 6157 struct net *net = (struct net *)seq->private;
1da177e4 6158 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
69ddb805
DL		 6159 net->ipv6.rt6_stats->fib_nodes,
6160 net->ipv6.rt6_stats->fib_route_nodes,
81eb8447 6161 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
69ddb805
DL		 6162 net->ipv6.rt6_stats->fib_rt_entries,
6163 net->ipv6.rt6_stats->fib_rt_cache,
fc66f95c 6164 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
69ddb805 6165 net->ipv6.rt6_stats->fib_discarded_routes);
1da177e4
LT		 6166
6167 return 0;
6168}
1da177e4
LT		 6169#endif /* CONFIG_PROC_FS */
6170
6171#ifdef CONFIG_SYSCTL
6172
32927393
CH		 6173static int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
6174 void *buffer, size_t *lenp, loff_t *ppos)
1da177e4 6175{
c486da34
LAG		 6176 struct net *net;
6177 int delay;
f0fb9b28 6178 int ret;
c486da34 6179 if (!write)
1da177e4 6180 return -EINVAL;
c486da34
LAG		 6181
6182 net = (struct net *)ctl->extra1;
6183 delay = net->ipv6.sysctl.flush_delay;
f0fb9b28
AP		 6184 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6185 if (ret)
6186 return ret;
6187
2ac3ac8f 6188 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
c486da34 6189 return 0;
1da177e4
LT		 6190}
6191
ed792e28 6192static struct ctl_table ipv6_route_table_template[] = {
1ab1457c 6193 {
1da177e4 6194 .procname = "flush",
4990509f 6195 .data = &init_net.ipv6.sysctl.flush_delay,
1da177e4 6196 .maxlen = sizeof(int),
89c8b3a1 6197 .mode = 0200,
6d9f239a 6198 .proc_handler = ipv6_sysctl_rtcache_flush
1da177e4
LT		 6199 },
6200 {
1da177e4 6201 .procname = "gc_thresh",
9a7ec3a9 6202 .data = &ip6_dst_ops_template.gc_thresh,
1da177e4
LT		 6203 .maxlen = sizeof(int),
6204 .mode = 0644,
6d9f239a 6205 .proc_handler = proc_dointvec,
1da177e4
LT		 6206 },
6207 {
1da177e4 6208 .procname = "max_size",
4990509f 6209 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
1da177e4
LT		 6210 .maxlen = sizeof(int),
6211 .mode = 0644,
6d9f239a 6212 .proc_handler = proc_dointvec,
1da177e4
LT		 6213 },
6214 {
1da177e4 6215 .procname = "gc_min_interval",
4990509f 6216 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 6217 .maxlen = sizeof(int),
6218 .mode = 0644,
6d9f239a 6219 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6220 },
6221 {
1da177e4 6222 .procname = "gc_timeout",
4990509f 6223 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
1da177e4
LT		 6224 .maxlen = sizeof(int),
6225 .mode = 0644,
6d9f239a 6226 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6227 },
6228 {
1da177e4 6229 .procname = "gc_interval",
4990509f 6230 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
1da177e4
LT		 6231 .maxlen = sizeof(int),
6232 .mode = 0644,
6d9f239a 6233 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6234 },
6235 {
1da177e4 6236 .procname = "gc_elasticity",
4990509f 6237 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
1da177e4
LT		 6238 .maxlen = sizeof(int),
6239 .mode = 0644,
f3d3f616 6240 .proc_handler = proc_dointvec,
1da177e4
LT		 6241 },
6242 {
1da177e4 6243 .procname = "mtu_expires",
4990509f 6244 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
1da177e4
LT		 6245 .maxlen = sizeof(int),
6246 .mode = 0644,
6d9f239a 6247 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6248 },
6249 {
1da177e4 6250 .procname = "min_adv_mss",
4990509f 6251 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
1da177e4
LT		 6252 .maxlen = sizeof(int),
6253 .mode = 0644,
f3d3f616 6254 .proc_handler = proc_dointvec,
1da177e4
LT		 6255 },
6256 {
1da177e4 6257 .procname = "gc_min_interval_ms",
4990509f 6258 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 6259 .maxlen = sizeof(int),
6260 .mode = 0644,
6d9f239a 6261 .proc_handler = proc_dointvec_ms_jiffies,
1da177e4 6262 },
7c6bb7d2
DA		 6263 {
6264 .procname = "skip_notify_on_dev_down",
6265 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6266 .maxlen = sizeof(int),
6267 .mode = 0644,
b8e8a863 6268 .proc_handler = proc_dointvec_minmax,
eec4844f
MC		 6269 .extra1 = SYSCTL_ZERO,
6270 .extra2 = SYSCTL_ONE,
7c6bb7d2 6271 },
f8572d8f 6272 { }
1da177e4
LT		 6273};
6274
2c8c1e72 6275struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
760f2d01
DL		 6276{
6277 struct ctl_table *table;
6278
6279 table = kmemdup(ipv6_route_table_template,
6280 sizeof(ipv6_route_table_template),
6281 GFP_KERNEL);
5ee09105
YH		 6282
6283 if (table) {
6284 table[0].data = &net->ipv6.sysctl.flush_delay;
c486da34 6285 table[0].extra1 = net;
86393e52 6286 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
5ee09105
YH		 6287 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
6288 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6289 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6290 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6291 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6292 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6293 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
9c69fabe 6294 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
7c6bb7d2 6295 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
464dc801
EB		 6296
6297 /* Don't export sysctls to unprivileged users */
6298 if (net->user_ns != &init_user_ns)
6299 table[0].procname = NULL;
5ee09105
YH		 6300 }
6301
760f2d01
DL		 6302 return table;
6303}
1da177e4
LT		 6304#endif
6305
2c8c1e72 6306static int __net_init ip6_route_net_init(struct net *net)
cdb18761 6307{
633d424b 6308 int ret = -ENOMEM;
8ed67789 6309
86393e52
AD		 6310 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6311 sizeof(net->ipv6.ip6_dst_ops));
f2fc6a54 6312
fc66f95c
ED		 6313 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6314 goto out_ip6_dst_ops;
6315
1cf844c7 6316 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
421842ed
DA		 6317 if (!net->ipv6.fib6_null_entry)
6318 goto out_ip6_dst_entries;
1cf844c7
DA		 6319 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6320 sizeof(*net->ipv6.fib6_null_entry));
421842ed 6321
8ed67789
DL		 6322 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6323 sizeof(*net->ipv6.ip6_null_entry),
6324 GFP_KERNEL);
6325 if (!net->ipv6.ip6_null_entry)
421842ed 6326 goto out_fib6_null_entry;
d8d1f30b 6327 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6328 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6329 ip6_template_metrics, true);
74109218 6330 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->rt6i_uncached);
8ed67789
DL		 6331
6332#ifdef CONFIG_IPV6_MULTIPLE_TABLES
feca7d8c 6333 net->ipv6.fib6_has_custom_rules = false;
8ed67789
DL		 6334 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6335 sizeof(*net->ipv6.ip6_prohibit_entry),
6336 GFP_KERNEL);
68fffc67
PZ		 6337 if (!net->ipv6.ip6_prohibit_entry)
6338 goto out_ip6_null_entry;
d8d1f30b 6339 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6340 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6341 ip6_template_metrics, true);
74109218 6342 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->rt6i_uncached);
8ed67789
DL		 6343
6344 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6345 sizeof(*net->ipv6.ip6_blk_hole_entry),
6346 GFP_KERNEL);
68fffc67
PZ		 6347 if (!net->ipv6.ip6_blk_hole_entry)
6348 goto out_ip6_prohibit_entry;
d8d1f30b 6349 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6350 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6351 ip6_template_metrics, true);
74109218 6352 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->rt6i_uncached);
b9b33e7c
PA		 6353#ifdef CONFIG_IPV6_SUBTREES
6354 net->ipv6.fib6_routes_require_src = 0;
6355#endif
8ed67789
DL		 6356#endif
6357
b339a47c
PZ		 6358 net->ipv6.sysctl.flush_delay = 0;
6359 net->ipv6.sysctl.ip6_rt_max_size = 4096;
6360 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6361 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6362 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6363 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6364 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6365 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
7c6bb7d2 6366 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
b339a47c 6367
6891a346
BT		 6368 net->ipv6.ip6_rt_gc_expire = 30*HZ;
6369
8ed67789
DL		 6370 ret = 0;
6371out:
6372 return ret;
f2fc6a54 6373
68fffc67
PZ		 6374#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6375out_ip6_prohibit_entry:
6376 kfree(net->ipv6.ip6_prohibit_entry);
6377out_ip6_null_entry:
6378 kfree(net->ipv6.ip6_null_entry);
6379#endif
421842ed
DA		 6380out_fib6_null_entry:
6381 kfree(net->ipv6.fib6_null_entry);
fc66f95c
ED		 6382out_ip6_dst_entries:
6383 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
f2fc6a54 6384out_ip6_dst_ops:
f2fc6a54 6385 goto out;
cdb18761
DL		 6386}
6387
2c8c1e72 6388static void __net_exit ip6_route_net_exit(struct net *net)
cdb18761 6389{
421842ed 6390 kfree(net->ipv6.fib6_null_entry);
8ed67789
DL		 6391 kfree(net->ipv6.ip6_null_entry);
6392#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6393 kfree(net->ipv6.ip6_prohibit_entry);
6394 kfree(net->ipv6.ip6_blk_hole_entry);
6395#endif
41bb78b4 6396 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
cdb18761
DL		 6397}
6398
d189634e
TG		 6399static int __net_init ip6_route_net_init_late(struct net *net)
6400{
6401#ifdef CONFIG_PROC_FS
c3506372
CH		 6402 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6403 sizeof(struct ipv6_route_iter));
3617d949
CH		 6404 proc_create_net_single("rt6_stats", 0444, net->proc_net,
6405 rt6_stats_seq_show, NULL);
d189634e
TG		 6406#endif
6407 return 0;
6408}
6409
6410static void __net_exit ip6_route_net_exit_late(struct net *net)
6411{
6412#ifdef CONFIG_PROC_FS
ece31ffd
G		 6413 remove_proc_entry("ipv6_route", net->proc_net);
6414 remove_proc_entry("rt6_stats", net->proc_net);
d189634e
TG		 6415#endif
6416}
6417
cdb18761
DL		 6418static struct pernet_operations ip6_route_net_ops = {
6419 .init = ip6_route_net_init,
6420 .exit = ip6_route_net_exit,
6421};
6422
c3426b47
DM		 6423static int __net_init ipv6_inetpeer_init(struct net *net)
6424{
6425 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6426
6427 if (!bp)
6428 return -ENOMEM;
6429 inet_peer_base_init(bp);
6430 net->ipv6.peers = bp;
6431 return 0;
6432}
6433
6434static void __net_exit ipv6_inetpeer_exit(struct net *net)
6435{
6436 struct inet_peer_base *bp = net->ipv6.peers;
6437
6438 net->ipv6.peers = NULL;
56a6b248 6439 inetpeer_invalidate_tree(bp);
c3426b47
DM		 6440 kfree(bp);
6441}
6442
2b823f72 6443static struct pernet_operations ipv6_inetpeer_ops = {
c3426b47
DM		 6444 .init = ipv6_inetpeer_init,
6445 .exit = ipv6_inetpeer_exit,
6446};
6447
d189634e
TG		 6448static struct pernet_operations ip6_route_net_late_ops = {
6449 .init = ip6_route_net_init_late,
6450 .exit = ip6_route_net_exit_late,
6451};
6452
8ed67789
DL		 6453static struct notifier_block ip6_route_dev_notifier = {
6454 .notifier_call = ip6_route_dev_notify,
242d3a49 6455 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
8ed67789
DL		 6456};
6457
2f460933
WC		 6458void __init ip6_route_init_special_entries(void)
6459{
6460 /* Registering of the loopback is done before this portion of code,
6461 * the loopback reference in rt6_info will not be taken, do it
6462 * manually for init_net */
1cf844c7 6463 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
2f460933
WC		 6464 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6465 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6466 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6467 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6468 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6469 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6470 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6471 #endif
6472}
6473
138d0be3
YS		 6474#if IS_BUILTIN(CONFIG_IPV6)
6475#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6476DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6477
951cf368
YS		 6478BTF_ID_LIST(btf_fib6_info_id)
6479BTF_ID(struct, fib6_info)
6480
14fc6bd6 6481static const struct bpf_iter_seq_info ipv6_route_seq_info = {
15172a46
YS		 6482 .seq_ops = &ipv6_route_seq_ops,
6483 .init_seq_private = bpf_iter_init_seq_net,
6484 .fini_seq_private = bpf_iter_fini_seq_net,
6485 .seq_priv_size = sizeof(struct ipv6_route_iter),
14fc6bd6
YS		 6486};
6487
6488static struct bpf_iter_reg ipv6_route_reg_info = {
6489 .target = "ipv6_route",
3c32cc1b
YS		 6490 .ctx_arg_info_size = 1,
6491 .ctx_arg_info = {
6492 { offsetof(struct bpf_iter__ipv6_route, rt),
6493 PTR_TO_BTF_ID_OR_NULL },
6494 },
14fc6bd6 6495 .seq_info = &ipv6_route_seq_info,
15172a46
YS		 6496};
6497
138d0be3
YS		 6498static int __init bpf_iter_register(void)
6499{
951cf368 6500 ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id;
15172a46 6501 return bpf_iter_reg_target(&ipv6_route_reg_info);
138d0be3
YS		 6502}
6503
6504static void bpf_iter_unregister(void)
6505{
ab2ee4fc 6506 bpf_iter_unreg_target(&ipv6_route_reg_info);
138d0be3
YS		 6507}
6508#endif
6509#endif
6510
433d49c3 6511int __init ip6_route_init(void)
1da177e4 6512{
433d49c3 6513 int ret;
8d0b94af 6514 int cpu;
433d49c3 6515
9a7ec3a9
DL		 6516 ret = -ENOMEM;
6517 ip6_dst_ops_template.kmem_cachep =
e5d679f3 6518 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
f845ab6b 6519 SLAB_HWCACHE_ALIGN, NULL);
9a7ec3a9 6520 if (!ip6_dst_ops_template.kmem_cachep)
c19a28e1 6521 goto out;
14e50e57 6522
fc66f95c 6523 ret = dst_entries_init(&ip6_dst_blackhole_ops);
8ed67789 6524 if (ret)
bdb3289f 6525 goto out_kmem_cache;
bdb3289f 6526
c3426b47
DM		 6527 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6528 if (ret)
e8803b6c 6529 goto out_dst_entries;
2a0c451a 6530
7e52b33b
DM		 6531 ret = register_pernet_subsys(&ip6_route_net_ops);
6532 if (ret)
6533 goto out_register_inetpeer;
c3426b47 6534
5dc121e9
AE		 6535 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6536
e8803b6c 6537 ret = fib6_init();
433d49c3 6538 if (ret)
8ed67789 6539 goto out_register_subsys;
433d49c3 6540
433d49c3
DL		 6541 ret = xfrm6_init();
6542 if (ret)
e8803b6c 6543 goto out_fib6_init;
c35b7e72 6544
433d49c3
DL		 6545 ret = fib6_rules_init();
6546 if (ret)
6547 goto xfrm6_init;
7e5449c2 6548
d189634e
TG		 6549 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6550 if (ret)
6551 goto fib6_rules_init;
6552
16feebcf
FW		 6553 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6554 inet6_rtm_newroute, NULL, 0);
6555 if (ret < 0)
6556 goto out_register_late_subsys;
6557
6558 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6559 inet6_rtm_delroute, NULL, 0);
6560 if (ret < 0)
6561 goto out_register_late_subsys;
6562
6563 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6564 inet6_rtm_getroute, NULL,
6565 RTNL_FLAG_DOIT_UNLOCKED);
6566 if (ret < 0)
d189634e 6567 goto out_register_late_subsys;
c127ea2c 6568
8ed67789 6569 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
cdb18761 6570 if (ret)
d189634e 6571 goto out_register_late_subsys;
8ed67789 6572
138d0be3
YS		 6573#if IS_BUILTIN(CONFIG_IPV6)
6574#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6575 ret = bpf_iter_register();
6576 if (ret)
6577 goto out_register_late_subsys;
6578#endif
6579#endif
6580
8d0b94af
MKL		 6581 for_each_possible_cpu(cpu) {
6582 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6583
6584 INIT_LIST_HEAD(&ul->head);
6585 spin_lock_init(&ul->lock);
6586 }
6587
433d49c3
DL		 6588out:
6589 return ret;
6590
d189634e 6591out_register_late_subsys:
16feebcf 6592 rtnl_unregister_all(PF_INET6);
d189634e 6593 unregister_pernet_subsys(&ip6_route_net_late_ops);
433d49c3 6594fib6_rules_init:
433d49c3
DL		 6595 fib6_rules_cleanup();
6596xfrm6_init:
433d49c3 6597 xfrm6_fini();
2a0c451a
TG		 6598out_fib6_init:
6599 fib6_gc_cleanup();
8ed67789
DL		 6600out_register_subsys:
6601 unregister_pernet_subsys(&ip6_route_net_ops);
7e52b33b
DM		 6602out_register_inetpeer:
6603 unregister_pernet_subsys(&ipv6_inetpeer_ops);
fc66f95c
ED		 6604out_dst_entries:
6605 dst_entries_destroy(&ip6_dst_blackhole_ops);
433d49c3 6606out_kmem_cache:
f2fc6a54 6607 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
433d49c3 6608 goto out;
1da177e4
LT		 6609}
6610
6611void ip6_route_cleanup(void)
6612{
138d0be3
YS		 6613#if IS_BUILTIN(CONFIG_IPV6)
6614#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6615 bpf_iter_unregister();
6616#endif
6617#endif
8ed67789 6618 unregister_netdevice_notifier(&ip6_route_dev_notifier);
d189634e 6619 unregister_pernet_subsys(&ip6_route_net_late_ops);
101367c2 6620 fib6_rules_cleanup();
1da177e4 6621 xfrm6_fini();
1da177e4 6622 fib6_gc_cleanup();
c3426b47 6623 unregister_pernet_subsys(&ipv6_inetpeer_ops);
8ed67789 6624 unregister_pernet_subsys(&ip6_route_net_ops);
41bb78b4 6625 dst_entries_destroy(&ip6_dst_blackhole_ops);
f2fc6a54 6626 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
1da177e4 6627}
Linux 6.x block layer and io_uring tree(s)