projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
bpf: Make btf_sock_ids global
[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>
1da177e4
LT		 64
65#ifdef CONFIG_SYSCTL
66#include <linux/sysctl.h>
67#endif
68
30d444d3
DA		 69static int ip6_rt_type_to_error(u8 fib6_type);
70
71#define CREATE_TRACE_POINTS
72#include <trace/events/fib6.h>
73EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
74#undef CREATE_TRACE_POINTS
75
afc154e9 76enum rt6_nud_state {
7e980569
JB		 77 RT6_NUD_FAIL_HARD = -3,
78 RT6_NUD_FAIL_PROBE = -2,
79 RT6_NUD_FAIL_DO_RR = -1,
afc154e9
HFS		 80 RT6_NUD_SUCCEED = 1
81};
82
1da177e4 83static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
0dbaee3b 84static unsigned int ip6_default_advmss(const struct dst_entry *dst);
ebb762f2 85static unsigned int ip6_mtu(const struct dst_entry *dst);
1da177e4
LT		 86static struct dst_entry *ip6_negative_advice(struct dst_entry *);
87static void ip6_dst_destroy(struct dst_entry *);
88static void ip6_dst_ifdown(struct dst_entry *,
89 struct net_device *dev, int how);
569d3645 90static int ip6_dst_gc(struct dst_ops *ops);
1da177e4
LT		 91
92static int ip6_pkt_discard(struct sk_buff *skb);
ede2059d 93static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
7150aede 94static int ip6_pkt_prohibit(struct sk_buff *skb);
ede2059d 95static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1da177e4 96static void ip6_link_failure(struct sk_buff *skb);
6700c270 97static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
bd085ef6
HL		 98 struct sk_buff *skb, u32 mtu,
99 bool confirm_neigh);
6700c270
DM		 100static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
101 struct sk_buff *skb);
702cea56
DA		 102static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
103 int strict);
a1b7a1f0 104static size_t rt6_nlmsg_size(struct fib6_info *f6i);
d4ead6b3 105static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 106 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 107 struct in6_addr *dest, struct in6_addr *src,
16a16cd3
DA		 108 int iif, int type, u32 portid, u32 seq,
109 unsigned int flags);
7e4b5128 110static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
510e2ced
WW		 111 const struct in6_addr *daddr,
112 const struct in6_addr *saddr);
1da177e4 113
70ceb4f5 114#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 115static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 116 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 117 const struct in6_addr *gwaddr,
118 struct net_device *dev,
95c96174 119 unsigned int pref);
8d1c802b 120static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 121 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 122 const struct in6_addr *gwaddr,
123 struct net_device *dev);
70ceb4f5
YH		 124#endif
125
8d0b94af
MKL		 126struct uncached_list {
127 spinlock_t lock;
128 struct list_head head;
129};
130
131static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
132
510c321b 133void rt6_uncached_list_add(struct rt6_info *rt)
8d0b94af
MKL		 134{
135 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
136
8d0b94af
MKL		 137 rt->rt6i_uncached_list = ul;
138
139 spin_lock_bh(&ul->lock);
140 list_add_tail(&rt->rt6i_uncached, &ul->head);
141 spin_unlock_bh(&ul->lock);
142}
143
510c321b 144void rt6_uncached_list_del(struct rt6_info *rt)
8d0b94af
MKL		 145{
146 if (!list_empty(&rt->rt6i_uncached)) {
147 struct uncached_list *ul = rt->rt6i_uncached_list;
81eb8447 148 struct net *net = dev_net(rt->dst.dev);
8d0b94af
MKL		 149
150 spin_lock_bh(&ul->lock);
151 list_del(&rt->rt6i_uncached);
81eb8447 152 atomic_dec(&net->ipv6.rt6_stats->fib_rt_uncache);
8d0b94af
MKL		 153 spin_unlock_bh(&ul->lock);
154 }
155}
156
157static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
158{
159 struct net_device *loopback_dev = net->loopback_dev;
160 int cpu;
161
e332bc67
EB		 162 if (dev == loopback_dev)
163 return;
164
8d0b94af
MKL		 165 for_each_possible_cpu(cpu) {
166 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
167 struct rt6_info *rt;
168
169 spin_lock_bh(&ul->lock);
170 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
171 struct inet6_dev *rt_idev = rt->rt6i_idev;
172 struct net_device *rt_dev = rt->dst.dev;
173
e332bc67 174 if (rt_idev->dev == dev) {
8d0b94af
MKL		 175 rt->rt6i_idev = in6_dev_get(loopback_dev);
176 in6_dev_put(rt_idev);
177 }
178
e332bc67 179 if (rt_dev == dev) {
8d7017fd 180 rt->dst.dev = blackhole_netdev;
8d0b94af
MKL		 181 dev_hold(rt->dst.dev);
182 dev_put(rt_dev);
183 }
184 }
185 spin_unlock_bh(&ul->lock);
186 }
187}
188
f8a1b43b 189static inline const void *choose_neigh_daddr(const struct in6_addr *p,
f894cbf8
DM		 190 struct sk_buff *skb,
191 const void *daddr)
39232973 192{
a7563f34 193 if (!ipv6_addr_any(p))
39232973 194 return (const void *) p;
f894cbf8
DM		 195 else if (skb)
196 return &ipv6_hdr(skb)->daddr;
39232973
DM		 197 return daddr;
198}
199
f8a1b43b
DA		 200struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
201 struct net_device *dev,
202 struct sk_buff *skb,
203 const void *daddr)
d3aaeb38 204{
39232973
DM		 205 struct neighbour *n;
206
f8a1b43b
DA		 207 daddr = choose_neigh_daddr(gw, skb, daddr);
208 n = __ipv6_neigh_lookup(dev, daddr);
f83c7790
DM		 209 if (n)
210 return n;
7adf3246
SB		 211
212 n = neigh_create(&nd_tbl, daddr, dev);
213 return IS_ERR(n) ? NULL : n;
f8a1b43b
DA		 214}
215
216static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
217 struct sk_buff *skb,
218 const void *daddr)
219{
220 const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
221
2c6b55f4
ND		 222 return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
223 dst->dev, skb, daddr);
f83c7790
DM		 224}
225
63fca65d
JA		 226static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
227{
228 struct net_device *dev = dst->dev;
229 struct rt6_info *rt = (struct rt6_info *)dst;
230
cbfd6891 231 daddr = choose_neigh_daddr(rt6_nexthop(rt, &in6addr_any), NULL, daddr);
63fca65d
JA		 232 if (!daddr)
233 return;
234 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
235 return;
236 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
237 return;
238 __ipv6_confirm_neigh(dev, daddr);
239}
240
9a7ec3a9 241static struct dst_ops ip6_dst_ops_template = {
1da177e4 242 .family = AF_INET6,
1da177e4
LT		 243 .gc = ip6_dst_gc,
244 .gc_thresh = 1024,
245 .check = ip6_dst_check,
0dbaee3b 246 .default_advmss = ip6_default_advmss,
ebb762f2 247 .mtu = ip6_mtu,
d4ead6b3 248 .cow_metrics = dst_cow_metrics_generic,
1da177e4
LT		 249 .destroy = ip6_dst_destroy,
250 .ifdown = ip6_dst_ifdown,
251 .negative_advice = ip6_negative_advice,
252 .link_failure = ip6_link_failure,
253 .update_pmtu = ip6_rt_update_pmtu,
6e157b6a 254 .redirect = rt6_do_redirect,
9f8955cc 255 .local_out = __ip6_local_out,
f8a1b43b 256 .neigh_lookup = ip6_dst_neigh_lookup,
63fca65d 257 .confirm_neigh = ip6_confirm_neigh,
1da177e4
LT		 258};
259
ebb762f2 260static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
ec831ea7 261{
618f9bc7
SK		 262 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
263
264 return mtu ? : dst->dev->mtu;
ec831ea7
RD		 265}
266
6700c270 267static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
bd085ef6
HL		 268 struct sk_buff *skb, u32 mtu,
269 bool confirm_neigh)
14e50e57
DM		 270{
271}
272
6700c270
DM		 273static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
274 struct sk_buff *skb)
b587ee3b
DM		 275{
276}
277
14e50e57
DM		 278static struct dst_ops ip6_dst_blackhole_ops = {
279 .family = AF_INET6,
14e50e57
DM		 280 .destroy = ip6_dst_destroy,
281 .check = ip6_dst_check,
ebb762f2 282 .mtu = ip6_blackhole_mtu,
214f45c9 283 .default_advmss = ip6_default_advmss,
14e50e57 284 .update_pmtu = ip6_rt_blackhole_update_pmtu,
b587ee3b 285 .redirect = ip6_rt_blackhole_redirect,
0a1f5962 286 .cow_metrics = dst_cow_metrics_generic,
f8a1b43b 287 .neigh_lookup = ip6_dst_neigh_lookup,
14e50e57
DM		 288};
289
62fa8a84 290static const u32 ip6_template_metrics[RTAX_MAX] = {
14edd87d 291 [RTAX_HOPLIMIT - 1] = 0,
62fa8a84
DM		 292};
293
8d1c802b 294static const struct fib6_info fib6_null_entry_template = {
93c2fb25
DA		 295 .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
296 .fib6_protocol = RTPROT_KERNEL,
297 .fib6_metric = ~(u32)0,
f05713e0 298 .fib6_ref = REFCOUNT_INIT(1),
421842ed
DA		 299 .fib6_type = RTN_UNREACHABLE,
300 .fib6_metrics = (struct dst_metrics *)&dst_default_metrics,
301};
302
fb0af4c7 303static const struct rt6_info ip6_null_entry_template = {
d8d1f30b
CG		 304 .dst = {
305 .__refcnt = ATOMIC_INIT(1),
306 .__use = 1,
2c20cbd7 307 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 308 .error = -ENETUNREACH,
d8d1f30b
CG		 309 .input = ip6_pkt_discard,
310 .output = ip6_pkt_discard_out,
1da177e4
LT		 311 },
312 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
1da177e4
LT		 313};
314
101367c2
TG		 315#ifdef CONFIG_IPV6_MULTIPLE_TABLES
316
fb0af4c7 317static const struct rt6_info ip6_prohibit_entry_template = {
d8d1f30b
CG		 318 .dst = {
319 .__refcnt = ATOMIC_INIT(1),
320 .__use = 1,
2c20cbd7 321 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 322 .error = -EACCES,
d8d1f30b
CG		 323 .input = ip6_pkt_prohibit,
324 .output = ip6_pkt_prohibit_out,
101367c2
TG		 325 },
326 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 327};
328
fb0af4c7 329static const struct rt6_info ip6_blk_hole_entry_template = {
d8d1f30b
CG		 330 .dst = {
331 .__refcnt = ATOMIC_INIT(1),
332 .__use = 1,
2c20cbd7 333 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 334 .error = -EINVAL,
d8d1f30b 335 .input = dst_discard,
ede2059d 336 .output = dst_discard_out,
101367c2
TG		 337 },
338 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 339};
340
341#endif
342
ebfa45f0
MKL		 343static void rt6_info_init(struct rt6_info *rt)
344{
345 struct dst_entry *dst = &rt->dst;
346
347 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
ebfa45f0
MKL		 348 INIT_LIST_HEAD(&rt->rt6i_uncached);
349}
350
1da177e4 351/* allocate dst with ip6_dst_ops */
93531c67
DA		 352struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
353 int flags)
1da177e4 354{
97bab73f 355 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
b2a9c0ed 356 1, DST_OBSOLETE_FORCE_CHK, flags);
cf911662 357
81eb8447 358 if (rt) {
ebfa45f0 359 rt6_info_init(rt);
81eb8447
WW		 360 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
361 }
8104891b 362
cf911662 363 return rt;
1da177e4 364}
9ab179d8 365EXPORT_SYMBOL(ip6_dst_alloc);
d52d3997 366
1da177e4
LT		 367static void ip6_dst_destroy(struct dst_entry *dst)
368{
369 struct rt6_info *rt = (struct rt6_info *)dst;
a68886a6 370 struct fib6_info *from;
8d0b94af 371 struct inet6_dev *idev;
1da177e4 372
1620a336 373 ip_dst_metrics_put(dst);
8d0b94af
MKL		 374 rt6_uncached_list_del(rt);
375
376 idev = rt->rt6i_idev;
38308473 377 if (idev) {
1da177e4
LT		 378 rt->rt6i_idev = NULL;
379 in6_dev_put(idev);
1ab1457c 380 }
1716a961 381
0e233874 382 from = xchg((__force struct fib6_info **)&rt->from, NULL);
93531c67 383 fib6_info_release(from);
b3419363
DM		 384}
385
1da177e4
LT		 386static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
387 int how)
388{
389 struct rt6_info *rt = (struct rt6_info *)dst;
390 struct inet6_dev *idev = rt->rt6i_idev;
5a3e55d6 391 struct net_device *loopback_dev =
c346dca1 392 dev_net(dev)->loopback_dev;
1da177e4 393
e5645f51
WW		 394 if (idev && idev->dev != loopback_dev) {
395 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
396 if (loopback_idev) {
397 rt->rt6i_idev = loopback_idev;
398 in6_dev_put(idev);
97cac082 399 }
1da177e4
LT		 400 }
401}
402
5973fb1e
MKL		 403static bool __rt6_check_expired(const struct rt6_info *rt)
404{
405 if (rt->rt6i_flags & RTF_EXPIRES)
406 return time_after(jiffies, rt->dst.expires);
407 else
408 return false;
409}
410
a50feda5 411static bool rt6_check_expired(const struct rt6_info *rt)
1da177e4 412{
a68886a6
DA		 413 struct fib6_info *from;
414
415 from = rcu_dereference(rt->from);
416
1716a961
G		 417 if (rt->rt6i_flags & RTF_EXPIRES) {
418 if (time_after(jiffies, rt->dst.expires))
a50feda5 419 return true;
a68886a6 420 } else if (from) {
1e2ea8ad 421 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
a68886a6 422 fib6_check_expired(from);
1716a961 423 }
a50feda5 424 return false;
1da177e4
LT		 425}
426
b1d40991
DA		 427void fib6_select_path(const struct net *net, struct fib6_result *res,
428 struct flowi6 *fl6, int oif, bool have_oif_match,
429 const struct sk_buff *skb, int strict)
51ebd318 430{
8d1c802b 431 struct fib6_info *sibling, *next_sibling;
b1d40991
DA		 432 struct fib6_info *match = res->f6i;
433
34fe5a1c 434 if (!match->nh && (!match->fib6_nsiblings || have_oif_match))
b1d40991 435 goto out;
51ebd318 436
34fe5a1c
DA		 437 if (match->nh && have_oif_match && res->nh)
438 return;
439
b673d6cc
JS		 440 /* We might have already computed the hash for ICMPv6 errors. In such
441 * case it will always be non-zero. Otherwise now is the time to do it.
442 */
f88d8ea6
DA		 443 if (!fl6->mp_hash &&
444 (!match->nh || nexthop_is_multipath(match->nh)))
b4bac172 445 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
b673d6cc 446
f88d8ea6
DA		 447 if (unlikely(match->nh)) {
448 nexthop_path_fib6_result(res, fl6->mp_hash);
449 return;
450 }
451
1cf844c7 452 if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
b1d40991 453 goto out;
3d709f69 454
93c2fb25
DA		 455 list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
456 fib6_siblings) {
1cf844c7 457 const struct fib6_nh *nh = sibling->fib6_nh;
5e670d84
DA		 458 int nh_upper_bound;
459
702cea56 460 nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
5e670d84 461 if (fl6->mp_hash > nh_upper_bound)
3d709f69 462 continue;
702cea56 463 if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
3d709f69
IS		 464 break;
465 match = sibling;
466 break;
467 }
468
b1d40991
DA		 469out:
470 res->f6i = match;
1cf844c7 471 res->nh = match->fib6_nh;
51ebd318
ND		 472}
473
1da177e4 474/*
66f5d6ce 475 * Route lookup. rcu_read_lock() should be held.
1da177e4
LT		 476 */
477
0c59d006
DA		 478static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
479 const struct in6_addr *saddr, int oif, int flags)
480{
481 const struct net_device *dev;
482
483 if (nh->fib_nh_flags & RTNH_F_DEAD)
484 return false;
485
486 dev = nh->fib_nh_dev;
487 if (oif) {
488 if (dev->ifindex == oif)
489 return true;
490 } else {
491 if (ipv6_chk_addr(net, saddr, dev,
492 flags & RT6_LOOKUP_F_IFACE))
493 return true;
494 }
495
496 return false;
497}
498
962b6803
DA		 499struct fib6_nh_dm_arg {
500 struct net *net;
501 const struct in6_addr *saddr;
502 int oif;
503 int flags;
504 struct fib6_nh *nh;
505};
506
507static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
508{
509 struct fib6_nh_dm_arg *arg = _arg;
510
511 arg->nh = nh;
512 return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
513 arg->flags);
514}
515
516/* returns fib6_nh from nexthop or NULL */
517static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
518 struct fib6_result *res,
519 const struct in6_addr *saddr,
520 int oif, int flags)
521{
522 struct fib6_nh_dm_arg arg = {
523 .net = net,
524 .saddr = saddr,
525 .oif = oif,
526 .flags = flags,
527 };
528
529 if (nexthop_is_blackhole(nh))
530 return NULL;
531
532 if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
533 return arg.nh;
534
535 return NULL;
536}
537
75ef7389
DA		 538static void rt6_device_match(struct net *net, struct fib6_result *res,
539 const struct in6_addr *saddr, int oif, int flags)
1da177e4 540{
75ef7389
DA		 541 struct fib6_info *f6i = res->f6i;
542 struct fib6_info *spf6i;
543 struct fib6_nh *nh;
1da177e4 544
75ef7389 545 if (!oif && ipv6_addr_any(saddr)) {
f88d8ea6
DA		 546 if (unlikely(f6i->nh)) {
547 nh = nexthop_fib6_nh(f6i->nh);
548 if (nexthop_is_blackhole(f6i->nh))
549 goto out_blackhole;
550 } else {
551 nh = f6i->fib6_nh;
552 }
7d21fec9
DA		 553 if (!(nh->fib_nh_flags & RTNH_F_DEAD))
554 goto out;
75ef7389 555 }
dd3abc4e 556
75ef7389 557 for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
962b6803
DA		 558 bool matched = false;
559
560 if (unlikely(spf6i->nh)) {
561 nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
562 oif, flags);
563 if (nh)
564 matched = true;
565 } else {
566 nh = spf6i->fib6_nh;
567 if (__rt6_device_match(net, nh, saddr, oif, flags))
568 matched = true;
569 }
570 if (matched) {
75ef7389 571 res->f6i = spf6i;
7d21fec9 572 goto out;
75ef7389 573 }
dd3abc4e 574 }
1da177e4 575
75ef7389
DA		 576 if (oif && flags & RT6_LOOKUP_F_IFACE) {
577 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 578 nh = res->f6i->fib6_nh;
7d21fec9 579 goto out;
75ef7389 580 }
8067bb8c 581
f88d8ea6
DA		 582 if (unlikely(f6i->nh)) {
583 nh = nexthop_fib6_nh(f6i->nh);
584 if (nexthop_is_blackhole(f6i->nh))
585 goto out_blackhole;
586 } else {
587 nh = f6i->fib6_nh;
588 }
589
7d21fec9 590 if (nh->fib_nh_flags & RTNH_F_DEAD) {
75ef7389 591 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 592 nh = res->f6i->fib6_nh;
75ef7389 593 }
7d21fec9
DA		 594out:
595 res->nh = nh;
596 res->fib6_type = res->f6i->fib6_type;
597 res->fib6_flags = res->f6i->fib6_flags;
f88d8ea6
DA		 598 return;
599
600out_blackhole:
601 res->fib6_flags |= RTF_REJECT;
602 res->fib6_type = RTN_BLACKHOLE;
603 res->nh = nh;
1da177e4
LT		 604}
605
27097255 606#ifdef CONFIG_IPV6_ROUTER_PREF
c2f17e82
HFS		 607struct __rt6_probe_work {
608 struct work_struct work;
609 struct in6_addr target;
610 struct net_device *dev;
611};
612
613static void rt6_probe_deferred(struct work_struct *w)
614{
615 struct in6_addr mcaddr;
616 struct __rt6_probe_work *work =
617 container_of(w, struct __rt6_probe_work, work);
618
619 addrconf_addr_solict_mult(&work->target, &mcaddr);
adc176c5 620 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
c2f17e82 621 dev_put(work->dev);
662f5533 622 kfree(work);
c2f17e82
HFS		 623}
624
cc3a86c8 625static void rt6_probe(struct fib6_nh *fib6_nh)
27097255 626{
f547fac6 627 struct __rt6_probe_work *work = NULL;
5e670d84 628 const struct in6_addr *nh_gw;
1bef4c22 629 unsigned long last_probe;
f2c31e32 630 struct neighbour *neigh;
5e670d84 631 struct net_device *dev;
f547fac6 632 struct inet6_dev *idev;
5e670d84 633
27097255
YH		 634 /*
635 * Okay, this does not seem to be appropriate
636 * for now, however, we need to check if it
637 * is really so; aka Router Reachability Probing.
638 *
639 * Router Reachability Probe MUST be rate-limited
640 * to no more than one per minute.
641 */
004b3942 642 if (!fib6_nh->fib_nh_gw_family)
7ff74a59 643 return;
5e670d84 644
cc3a86c8
DA		 645 nh_gw = &fib6_nh->fib_nh_gw6;
646 dev = fib6_nh->fib_nh_dev;
2152caea 647 rcu_read_lock_bh();
1bef4c22 648 last_probe = READ_ONCE(fib6_nh->last_probe);
f547fac6 649 idev = __in6_dev_get(dev);
5e670d84 650 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
2152caea 651 if (neigh) {
8d6c31bf
MKL		 652 if (neigh->nud_state & NUD_VALID)
653 goto out;
654
2152caea 655 write_lock(&neigh->lock);
990edb42
MKL		 656 if (!(neigh->nud_state & NUD_VALID) &&
657 time_after(jiffies,
dcd1f572 658 neigh->updated + idev->cnf.rtr_probe_interval)) {
990edb42
MKL		 659 work = kmalloc(sizeof(*work), GFP_ATOMIC);
660 if (work)
661 __neigh_set_probe_once(neigh);
c2f17e82 662 }
2152caea 663 write_unlock(&neigh->lock);
1bef4c22 664 } else if (time_after(jiffies, last_probe +
f547fac6 665 idev->cnf.rtr_probe_interval)) {
990edb42 666 work = kmalloc(sizeof(*work), GFP_ATOMIC);
f2c31e32 667 }
990edb42 668
1bef4c22
ED		 669 if (!work || cmpxchg(&fib6_nh->last_probe,
670 last_probe, jiffies) != last_probe) {
671 kfree(work);
672 } else {
990edb42 673 INIT_WORK(&work->work, rt6_probe_deferred);
5e670d84
DA		 674 work->target = *nh_gw;
675 dev_hold(dev);
676 work->dev = dev;
990edb42
MKL		 677 schedule_work(&work->work);
678 }
679
8d6c31bf 680out:
2152caea 681 rcu_read_unlock_bh();
27097255
YH		 682}
683#else
cc3a86c8 684static inline void rt6_probe(struct fib6_nh *fib6_nh)
27097255 685{
27097255
YH		 686}
687#endif
688
1da177e4 689/*
554cfb7e 690 * Default Router Selection (RFC 2461 6.3.6)
1da177e4 691 */
1ba9a895 692static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
1da177e4 693{
afc154e9 694 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
5e670d84 695 struct neighbour *neigh;
f2c31e32 696
145a3621 697 rcu_read_lock_bh();
1ba9a895
DA		 698 neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
699 &fib6_nh->fib_nh_gw6);
145a3621
YH		 700 if (neigh) {
701 read_lock(&neigh->lock);
554cfb7e 702 if (neigh->nud_state & NUD_VALID)
afc154e9 703 ret = RT6_NUD_SUCCEED;
398bcbeb 704#ifdef CONFIG_IPV6_ROUTER_PREF
a5a81f0b 705 else if (!(neigh->nud_state & NUD_FAILED))
afc154e9 706 ret = RT6_NUD_SUCCEED;
7e980569
JB		 707 else
708 ret = RT6_NUD_FAIL_PROBE;
398bcbeb 709#endif
145a3621 710 read_unlock(&neigh->lock);
afc154e9
HFS		 711 } else {
712 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
7e980569 713 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
a5a81f0b 714 }
145a3621
YH		 715 rcu_read_unlock_bh();
716
a5a81f0b 717 return ret;
1da177e4
LT		 718}
719
702cea56
DA		 720static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
721 int strict)
1da177e4 722{
6e1809a5
DA		 723 int m = 0;
724
725 if (!oif || nh->fib_nh_dev->ifindex == oif)
726 m = 2;
1ab1457c 727
77d16f45 728 if (!m && (strict & RT6_LOOKUP_F_IFACE))
afc154e9 729 return RT6_NUD_FAIL_HARD;
ebacaaa0 730#ifdef CONFIG_IPV6_ROUTER_PREF
702cea56 731 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
ebacaaa0 732#endif
1ba9a895 733 if ((strict & RT6_LOOKUP_F_REACHABLE) &&
702cea56 734 !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
1ba9a895 735 int n = rt6_check_neigh(nh);
afc154e9
HFS		 736 if (n < 0)
737 return n;
738 }
554cfb7e
YH		 739 return m;
740}
741
28679ed1
DA		 742static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
743 int oif, int strict, int *mpri, bool *do_rr)
554cfb7e 744{
afc154e9 745 bool match_do_rr = false;
28679ed1
DA		 746 bool rc = false;
747 int m;
35103d11 748
28679ed1 749 if (nh->fib_nh_flags & RTNH_F_DEAD)
8067bb8c
IS		 750 goto out;
751
28679ed1
DA		 752 if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
753 nh->fib_nh_flags & RTNH_F_LINKDOWN &&
d5d32e4b 754 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
35103d11 755 goto out;
f11e6659 756
28679ed1 757 m = rt6_score_route(nh, fib6_flags, oif, strict);
7e980569 758 if (m == RT6_NUD_FAIL_DO_RR) {
afc154e9
HFS		 759 match_do_rr = true;
760 m = 0; /* lowest valid score */
7e980569 761 } else if (m == RT6_NUD_FAIL_HARD) {
f11e6659 762 goto out;
afc154e9
HFS		 763 }
764
765 if (strict & RT6_LOOKUP_F_REACHABLE)
28679ed1 766 rt6_probe(nh);
f11e6659 767
7e980569 768 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
f11e6659 769 if (m > *mpri) {
afc154e9 770 *do_rr = match_do_rr;
f11e6659 771 *mpri = m;
28679ed1 772 rc = true;
f11e6659 773 }
f11e6659 774out:
28679ed1 775 return rc;
f11e6659
DM		 776}
777
17a5984e
DA		 778struct fib6_nh_frl_arg {
779 u32 flags;
780 int oif;
781 int strict;
782 int *mpri;
783 bool *do_rr;
784 struct fib6_nh *nh;
785};
786
787static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
788{
789 struct fib6_nh_frl_arg *arg = _arg;
790
791 arg->nh = nh;
792 return find_match(nh, arg->flags, arg->oif, arg->strict,
793 arg->mpri, arg->do_rr);
794}
795
b7bc4b6a 796static void __find_rr_leaf(struct fib6_info *f6i_start,
30c15f03 797 struct fib6_info *nomatch, u32 metric,
b7bc4b6a 798 struct fib6_result *res, struct fib6_info **cont,
30c15f03 799 int oif, int strict, bool *do_rr, int *mpri)
f11e6659 800{
b7bc4b6a 801 struct fib6_info *f6i;
1da177e4 802
b7bc4b6a
DA		 803 for (f6i = f6i_start;
804 f6i && f6i != nomatch;
805 f6i = rcu_dereference(f6i->fib6_next)) {
17a5984e 806 bool matched = false;
30c15f03
DA		 807 struct fib6_nh *nh;
808
b7bc4b6a
DA		 809 if (cont && f6i->fib6_metric != metric) {
810 *cont = f6i;
30c15f03 811 return;
9fbdcfaf
SK		 812 }
813
b7bc4b6a 814 if (fib6_check_expired(f6i))
28679ed1
DA		 815 continue;
816
17a5984e
DA		 817 if (unlikely(f6i->nh)) {
818 struct fib6_nh_frl_arg arg = {
819 .flags = f6i->fib6_flags,
820 .oif = oif,
821 .strict = strict,
822 .mpri = mpri,
823 .do_rr = do_rr
824 };
825
826 if (nexthop_is_blackhole(f6i->nh)) {
827 res->fib6_flags = RTF_REJECT;
828 res->fib6_type = RTN_BLACKHOLE;
829 res->f6i = f6i;
830 res->nh = nexthop_fib6_nh(f6i->nh);
831 return;
832 }
833 if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
834 &arg)) {
835 matched = true;
836 nh = arg.nh;
837 }
838 } else {
839 nh = f6i->fib6_nh;
840 if (find_match(nh, f6i->fib6_flags, oif, strict,
841 mpri, do_rr))
842 matched = true;
843 }
844 if (matched) {
b7bc4b6a
DA		 845 res->f6i = f6i;
846 res->nh = nh;
7d21fec9
DA		 847 res->fib6_flags = f6i->fib6_flags;
848 res->fib6_type = f6i->fib6_type;
b7bc4b6a 849 }
9fbdcfaf 850 }
30c15f03 851}
9fbdcfaf 852
b7bc4b6a
DA		 853static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
854 struct fib6_info *rr_head, int oif, int strict,
855 bool *do_rr, struct fib6_result *res)
30c15f03 856{
b7bc4b6a
DA		 857 u32 metric = rr_head->fib6_metric;
858 struct fib6_info *cont = NULL;
30c15f03 859 int mpri = -1;
9fbdcfaf 860
b7bc4b6a 861 __find_rr_leaf(rr_head, NULL, metric, res, &cont,
30c15f03 862 oif, strict, do_rr, &mpri);
28679ed1 863
b7bc4b6a 864 __find_rr_leaf(leaf, rr_head, metric, res, &cont,
30c15f03 865 oif, strict, do_rr, &mpri);
9fbdcfaf 866
b7bc4b6a
DA		 867 if (res->f6i || !cont)
868 return;
9fbdcfaf 869
b7bc4b6a 870 __find_rr_leaf(cont, NULL, metric, res, NULL,
30c15f03 871 oif, strict, do_rr, &mpri);
f11e6659 872}
1da177e4 873
b7bc4b6a
DA		 874static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
875 struct fib6_result *res, int strict)
f11e6659 876{
8d1c802b 877 struct fib6_info *leaf = rcu_dereference(fn->leaf);
b7bc4b6a 878 struct fib6_info *rt0;
afc154e9 879 bool do_rr = false;
17ecf590 880 int key_plen;
1da177e4 881
b7bc4b6a
DA		 882 /* make sure this function or its helpers sets f6i */
883 res->f6i = NULL;
884
421842ed 885 if (!leaf || leaf == net->ipv6.fib6_null_entry)
b7bc4b6a 886 goto out;
8d1040e8 887
66f5d6ce 888 rt0 = rcu_dereference(fn->rr_ptr);
f11e6659 889 if (!rt0)
66f5d6ce 890 rt0 = leaf;
1da177e4 891
17ecf590
WW		 892 /* Double check to make sure fn is not an intermediate node
893 * and fn->leaf does not points to its child's leaf
894 * (This might happen if all routes under fn are deleted from
895 * the tree and fib6_repair_tree() is called on the node.)
896 */
93c2fb25 897 key_plen = rt0->fib6_dst.plen;
17ecf590 898#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 899 if (rt0->fib6_src.plen)
900 key_plen = rt0->fib6_src.plen;
17ecf590
WW		 901#endif
902 if (fn->fn_bit != key_plen)
b7bc4b6a 903 goto out;
1da177e4 904
b7bc4b6a 905 find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
afc154e9 906 if (do_rr) {
8fb11a9a 907 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
f11e6659 908
554cfb7e 909 /* no entries matched; do round-robin */
93c2fb25 910 if (!next || next->fib6_metric != rt0->fib6_metric)
8d1040e8 911 next = leaf;
f11e6659 912
66f5d6ce 913 if (next != rt0) {
93c2fb25 914 spin_lock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 915 /* make sure next is not being deleted from the tree */
93c2fb25 916 if (next->fib6_node)
66f5d6ce 917 rcu_assign_pointer(fn->rr_ptr, next);
93c2fb25 918 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 919 }
1da177e4 920 }
1da177e4 921
b7bc4b6a
DA		 922out:
923 if (!res->f6i) {
924 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 925 res->nh = res->f6i->fib6_nh;
7d21fec9
DA		 926 res->fib6_flags = res->f6i->fib6_flags;
927 res->fib6_type = res->f6i->fib6_type;
b7bc4b6a 928 }
1da177e4
LT		 929}
930
85bd05de 931static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
8b9df265 932{
85bd05de
DA		 933 return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
934 res->nh->fib_nh_gw_family;
8b9df265
MKL		 935}
936
70ceb4f5
YH		 937#ifdef CONFIG_IPV6_ROUTE_INFO
938int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
b71d1d42 939 const struct in6_addr *gwaddr)
70ceb4f5 940{
c346dca1 941 struct net *net = dev_net(dev);
70ceb4f5
YH		 942 struct route_info *rinfo = (struct route_info *) opt;
943 struct in6_addr prefix_buf, *prefix;
944 unsigned int pref;
4bed72e4 945 unsigned long lifetime;
8d1c802b 946 struct fib6_info *rt;
70ceb4f5
YH		 947
948 if (len < sizeof(struct route_info)) {
949 return -EINVAL;
950 }
951
952 /* Sanity check for prefix_len and length */
953 if (rinfo->length > 3) {
954 return -EINVAL;
955 } else if (rinfo->prefix_len > 128) {
956 return -EINVAL;
957 } else if (rinfo->prefix_len > 64) {
958 if (rinfo->length < 2) {
959 return -EINVAL;
960 }
961 } else if (rinfo->prefix_len > 0) {
962 if (rinfo->length < 1) {
963 return -EINVAL;
964 }
965 }
966
967 pref = rinfo->route_pref;
968 if (pref == ICMPV6_ROUTER_PREF_INVALID)
3933fc95 969 return -EINVAL;
70ceb4f5 970
4bed72e4 971 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
70ceb4f5
YH		 972
973 if (rinfo->length == 3)
974 prefix = (struct in6_addr *)rinfo->prefix;
975 else {
976 /* this function is safe */
977 ipv6_addr_prefix(&prefix_buf,
978 (struct in6_addr *)rinfo->prefix,
979 rinfo->prefix_len);
980 prefix = &prefix_buf;
981 }
982
f104a567 983 if (rinfo->prefix_len == 0)
afb1d4b5 984 rt = rt6_get_dflt_router(net, gwaddr, dev);
f104a567
DJ		 985 else
986 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
830218c1 987 gwaddr, dev);
70ceb4f5
YH		 988
989 if (rt && !lifetime) {
11dd74b3 990 ip6_del_rt(net, rt, false);
70ceb4f5
YH		 991 rt = NULL;
992 }
993
994 if (!rt && lifetime)
830218c1
DA		 995 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
996 dev, pref);
70ceb4f5 997 else if (rt)
93c2fb25
DA		 998 rt->fib6_flags = RTF_ROUTEINFO |
999 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
70ceb4f5
YH		 1000
1001 if (rt) {
1716a961 1002 if (!addrconf_finite_timeout(lifetime))
14895687 1003 fib6_clean_expires(rt);
1716a961 1004 else
14895687 1005 fib6_set_expires(rt, jiffies + HZ * lifetime);
1716a961 1006
93531c67 1007 fib6_info_release(rt);
70ceb4f5
YH		 1008 }
1009 return 0;
1010}
1011#endif
1012
ae90d867
DA		 1013/*
1014 * Misc support functions
1015 */
1016
1017/* called with rcu_lock held */
0d161581 1018static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
ae90d867 1019{
0d161581 1020 struct net_device *dev = res->nh->fib_nh_dev;
ae90d867 1021
7d21fec9 1022 if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
ae90d867
DA		 1023 /* for copies of local routes, dst->dev needs to be the
1024 * device if it is a master device, the master device if
1025 * device is enslaved, and the loopback as the default
1026 */
1027 if (netif_is_l3_slave(dev) &&
7d21fec9 1028 !rt6_need_strict(&res->f6i->fib6_dst.addr))
ae90d867
DA		 1029 dev = l3mdev_master_dev_rcu(dev);
1030 else if (!netif_is_l3_master(dev))
1031 dev = dev_net(dev)->loopback_dev;
1032 /* last case is netif_is_l3_master(dev) is true in which
1033 * case we want dev returned to be dev
1034 */
1035 }
1036
1037 return dev;
1038}
1039
6edb3c96
DA		 1040static const int fib6_prop[RTN_MAX + 1] = {
1041 [RTN_UNSPEC] = 0,
1042 [RTN_UNICAST] = 0,
1043 [RTN_LOCAL] = 0,
1044 [RTN_BROADCAST] = 0,
1045 [RTN_ANYCAST] = 0,
1046 [RTN_MULTICAST] = 0,
1047 [RTN_BLACKHOLE] = -EINVAL,
1048 [RTN_UNREACHABLE] = -EHOSTUNREACH,
1049 [RTN_PROHIBIT] = -EACCES,
1050 [RTN_THROW] = -EAGAIN,
1051 [RTN_NAT] = -EINVAL,
1052 [RTN_XRESOLVE] = -EINVAL,
1053};
1054
1055static int ip6_rt_type_to_error(u8 fib6_type)
1056{
1057 return fib6_prop[fib6_type];
1058}
1059
8d1c802b 1060static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
3b6761d1
DA		 1061{
1062 unsigned short flags = 0;
1063
1064 if (rt->dst_nocount)
1065 flags |= DST_NOCOUNT;
1066 if (rt->dst_nopolicy)
1067 flags |= DST_NOPOLICY;
3b6761d1
DA		 1068
1069 return flags;
1070}
1071
7d21fec9 1072static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
6edb3c96 1073{
7d21fec9 1074 rt->dst.error = ip6_rt_type_to_error(fib6_type);
6edb3c96 1075
7d21fec9 1076 switch (fib6_type) {
6edb3c96
DA		 1077 case RTN_BLACKHOLE:
1078 rt->dst.output = dst_discard_out;
1079 rt->dst.input = dst_discard;
1080 break;
1081 case RTN_PROHIBIT:
1082 rt->dst.output = ip6_pkt_prohibit_out;
1083 rt->dst.input = ip6_pkt_prohibit;
1084 break;
1085 case RTN_THROW:
1086 case RTN_UNREACHABLE:
1087 default:
1088 rt->dst.output = ip6_pkt_discard_out;
1089 rt->dst.input = ip6_pkt_discard;
1090 break;
1091 }
1092}
1093
0d161581 1094static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
6edb3c96 1095{
7d21fec9 1096 struct fib6_info *f6i = res->f6i;
0d161581 1097
7d21fec9
DA		 1098 if (res->fib6_flags & RTF_REJECT) {
1099 ip6_rt_init_dst_reject(rt, res->fib6_type);
6edb3c96
DA		 1100 return;
1101 }
1102
1103 rt->dst.error = 0;
1104 rt->dst.output = ip6_output;
1105
7d21fec9 1106 if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
6edb3c96 1107 rt->dst.input = ip6_input;
7d21fec9 1108 } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
6edb3c96
DA		 1109 rt->dst.input = ip6_mc_input;
1110 } else {
1111 rt->dst.input = ip6_forward;
1112 }
1113
0d161581
DA		 1114 if (res->nh->fib_nh_lws) {
1115 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
6edb3c96
DA		 1116 lwtunnel_set_redirect(&rt->dst);
1117 }
1118
1119 rt->dst.lastuse = jiffies;
1120}
1121
e873e4b9 1122/* Caller must already hold reference to @from */
8d1c802b 1123static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
ae90d867 1124{
ae90d867 1125 rt->rt6i_flags &= ~RTF_EXPIRES;
a68886a6 1126 rcu_assign_pointer(rt->from, from);
e1255ed4 1127 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
ae90d867
DA		 1128}
1129
0d161581
DA		 1130/* Caller must already hold reference to f6i in result */
1131static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
ae90d867 1132{
0d161581
DA		 1133 const struct fib6_nh *nh = res->nh;
1134 const struct net_device *dev = nh->fib_nh_dev;
1135 struct fib6_info *f6i = res->f6i;
dcd1f572 1136
0d161581 1137 ip6_rt_init_dst(rt, res);
6edb3c96 1138
0d161581 1139 rt->rt6i_dst = f6i->fib6_dst;
dcd1f572 1140 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
7d21fec9 1141 rt->rt6i_flags = res->fib6_flags;
0d161581
DA		 1142 if (nh->fib_nh_gw_family) {
1143 rt->rt6i_gateway = nh->fib_nh_gw6;
2b2450ca
DA		 1144 rt->rt6i_flags |= RTF_GATEWAY;
1145 }
0d161581 1146 rt6_set_from(rt, f6i);
ae90d867 1147#ifdef CONFIG_IPV6_SUBTREES
0d161581 1148 rt->rt6i_src = f6i->fib6_src;
ae90d867 1149#endif
ae90d867
DA		 1150}
1151
a3c00e46
MKL		 1152static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1153 struct in6_addr *saddr)
1154{
66f5d6ce 1155 struct fib6_node *pn, *sn;
a3c00e46
MKL		 1156 while (1) {
1157 if (fn->fn_flags & RTN_TL_ROOT)
1158 return NULL;
66f5d6ce
WW		 1159 pn = rcu_dereference(fn->parent);
1160 sn = FIB6_SUBTREE(pn);
1161 if (sn && sn != fn)
6454743b 1162 fn = fib6_node_lookup(sn, NULL, saddr);
a3c00e46
MKL		 1163 else
1164 fn = pn;
1165 if (fn->fn_flags & RTN_RTINFO)
1166 return fn;
1167 }
1168}
c71099ac 1169
10585b43 1170static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
d3843fe5
WW		 1171{
1172 struct rt6_info *rt = *prt;
1173
1174 if (dst_hold_safe(&rt->dst))
1175 return true;
10585b43 1176 if (net) {
d3843fe5
WW		 1177 rt = net->ipv6.ip6_null_entry;
1178 dst_hold(&rt->dst);
1179 } else {
1180 rt = NULL;
1181 }
1182 *prt = rt;
1183 return false;
1184}
1185
dec9b0e2 1186/* called with rcu_lock held */
9b6b35ab 1187static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
dec9b0e2 1188{
9b6b35ab
DA		 1189 struct net_device *dev = res->nh->fib_nh_dev;
1190 struct fib6_info *f6i = res->f6i;
1191 unsigned short flags;
dec9b0e2
DA		 1192 struct rt6_info *nrt;
1193
9b6b35ab 1194 if (!fib6_info_hold_safe(f6i))
1c87e79a 1195 goto fallback;
e873e4b9 1196
9b6b35ab 1197 flags = fib6_info_dst_flags(f6i);
93531c67 1198 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1c87e79a 1199 if (!nrt) {
9b6b35ab 1200 fib6_info_release(f6i);
1c87e79a
XL		 1201 goto fallback;
1202 }
dec9b0e2 1203
0d161581 1204 ip6_rt_copy_init(nrt, res);
1c87e79a
XL		 1205 return nrt;
1206
1207fallback:
1208 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1209 dst_hold(&nrt->dst);
dec9b0e2
DA		 1210 return nrt;
1211}
1212
55cced4f 1213INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_lookup(struct net *net,
8ed67789 1214 struct fib6_table *table,
b75cc8f9
DA		 1215 struct flowi6 *fl6,
1216 const struct sk_buff *skb,
1217 int flags)
1da177e4 1218{
b1d40991 1219 struct fib6_result res = {};
1da177e4 1220 struct fib6_node *fn;
23fb93a4 1221 struct rt6_info *rt;
1da177e4 1222
b6cdbc85
DA		 1223 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1224 flags &= ~RT6_LOOKUP_F_IFACE;
1225
66f5d6ce 1226 rcu_read_lock();
6454743b 1227 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
c71099ac 1228restart:
b1d40991
DA		 1229 res.f6i = rcu_dereference(fn->leaf);
1230 if (!res.f6i)
1231 res.f6i = net->ipv6.fib6_null_entry;
af52a52c 1232 else
75ef7389
DA		 1233 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1234 flags);
af52a52c 1235
b1d40991 1236 if (res.f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 1237 fn = fib6_backtrack(fn, &fl6->saddr);
1238 if (fn)
1239 goto restart;
2b760fcf 1240
af52a52c
DA		 1241 rt = net->ipv6.ip6_null_entry;
1242 dst_hold(&rt->dst);
1243 goto out;
f88d8ea6
DA		 1244 } else if (res.fib6_flags & RTF_REJECT) {
1245 goto do_create;
af52a52c 1246 }
d3843fe5 1247
b1d40991
DA		 1248 fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1249 fl6->flowi6_oif != 0, skb, flags);
1250
2b760fcf 1251 /* Search through exception table */
7e4b5128 1252 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
23fb93a4 1253 if (rt) {
10585b43 1254 if (ip6_hold_safe(net, &rt))
dec9b0e2 1255 dst_use_noref(&rt->dst, jiffies);
23fb93a4 1256 } else {
f88d8ea6 1257do_create:
9b6b35ab 1258 rt = ip6_create_rt_rcu(&res);
dec9b0e2 1259 }
b811580d 1260
af52a52c 1261out:
8ff2e5b2 1262 trace_fib6_table_lookup(net, &res, table, fl6);
af52a52c 1263
66f5d6ce 1264 rcu_read_unlock();
b811580d 1265
c71099ac 1266 return rt;
c71099ac
TG		 1267}
1268
67ba4152 1269struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
b75cc8f9 1270 const struct sk_buff *skb, int flags)
ea6e574e 1271{
b75cc8f9 1272 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
ea6e574e
FW		 1273}
1274EXPORT_SYMBOL_GPL(ip6_route_lookup);
1275
9acd9f3a 1276struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
b75cc8f9
DA		 1277 const struct in6_addr *saddr, int oif,
1278 const struct sk_buff *skb, int strict)
c71099ac 1279{
4c9483b2
DM		 1280 struct flowi6 fl6 = {
1281 .flowi6_oif = oif,
1282 .daddr = *daddr,
c71099ac
TG		 1283 };
1284 struct dst_entry *dst;
77d16f45 1285 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
c71099ac 1286
adaa70bb 1287 if (saddr) {
4c9483b2 1288 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
adaa70bb
TG		 1289 flags |= RT6_LOOKUP_F_HAS_SADDR;
1290 }
1291
b75cc8f9 1292 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
c71099ac
TG		 1293 if (dst->error == 0)
1294 return (struct rt6_info *) dst;
1295
1296 dst_release(dst);
1297
1da177e4
LT		 1298 return NULL;
1299}
7159039a
YH		 1300EXPORT_SYMBOL(rt6_lookup);
1301
c71099ac 1302/* ip6_ins_rt is called with FREE table->tb6_lock.
1cfb71ee
WW		 1303 * It takes new route entry, the addition fails by any reason the
1304 * route is released.
1305 * Caller must hold dst before calling it.
1da177e4
LT		 1306 */
1307
8d1c802b 1308static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
333c4301 1309 struct netlink_ext_ack *extack)
1da177e4
LT		 1310{
1311 int err;
c71099ac 1312 struct fib6_table *table;
1da177e4 1313
93c2fb25 1314 table = rt->fib6_table;
66f5d6ce 1315 spin_lock_bh(&table->tb6_lock);
d4ead6b3 1316 err = fib6_add(&table->tb6_root, rt, info, extack);
66f5d6ce 1317 spin_unlock_bh(&table->tb6_lock);
1da177e4
LT		 1318
1319 return err;
1320}
1321
8d1c802b 1322int ip6_ins_rt(struct net *net, struct fib6_info *rt)
40e22e8f 1323{
afb1d4b5 1324 struct nl_info info = { .nl_net = net, };
e715b6d3 1325
d4ead6b3 1326 return __ip6_ins_rt(rt, &info, NULL);
40e22e8f
TG		 1327}
1328
85bd05de 1329static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
8b9df265
MKL		 1330 const struct in6_addr *daddr,
1331 const struct in6_addr *saddr)
1da177e4 1332{
85bd05de 1333 struct fib6_info *f6i = res->f6i;
4832c30d 1334 struct net_device *dev;
1da177e4
LT		 1335 struct rt6_info *rt;
1336
1337 /*
1338 * Clone the route.
1339 */
1340
85bd05de 1341 if (!fib6_info_hold_safe(f6i))
e873e4b9
WW		 1342 return NULL;
1343
0d161581 1344 dev = ip6_rt_get_dev_rcu(res);
93531c67 1345 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
e873e4b9 1346 if (!rt) {
85bd05de 1347 fib6_info_release(f6i);
83a09abd 1348 return NULL;
e873e4b9 1349 }
83a09abd 1350
0d161581 1351 ip6_rt_copy_init(rt, res);
83a09abd 1352 rt->rt6i_flags |= RTF_CACHE;
83a09abd
MKL		 1353 rt->rt6i_dst.addr = *daddr;
1354 rt->rt6i_dst.plen = 128;
1da177e4 1355
85bd05de
DA		 1356 if (!rt6_is_gw_or_nonexthop(res)) {
1357 if (f6i->fib6_dst.plen != 128 &&
1358 ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
83a09abd 1359 rt->rt6i_flags |= RTF_ANYCAST;
1da177e4 1360#ifdef CONFIG_IPV6_SUBTREES
83a09abd
MKL		 1361 if (rt->rt6i_src.plen && saddr) {
1362 rt->rt6i_src.addr = *saddr;
1363 rt->rt6i_src.plen = 128;
8b9df265 1364 }
83a09abd 1365#endif
95a9a5ba 1366 }
1da177e4 1367
95a9a5ba
YH		 1368 return rt;
1369}
1da177e4 1370
db3fedee 1371static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
d52d3997 1372{
db3fedee
DA		 1373 struct fib6_info *f6i = res->f6i;
1374 unsigned short flags = fib6_info_dst_flags(f6i);
4832c30d 1375 struct net_device *dev;
d52d3997
MKL		 1376 struct rt6_info *pcpu_rt;
1377
db3fedee 1378 if (!fib6_info_hold_safe(f6i))
e873e4b9
WW		 1379 return NULL;
1380
4832c30d 1381 rcu_read_lock();
0d161581 1382 dev = ip6_rt_get_dev_rcu(res);
d8882935 1383 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT);
4832c30d 1384 rcu_read_unlock();
e873e4b9 1385 if (!pcpu_rt) {
db3fedee 1386 fib6_info_release(f6i);
d52d3997 1387 return NULL;
e873e4b9 1388 }
0d161581 1389 ip6_rt_copy_init(pcpu_rt, res);
d52d3997 1390 pcpu_rt->rt6i_flags |= RTF_PCPU;
8f34e53b
DA		 1391
1392 if (f6i->nh)
1393 pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1394
d52d3997
MKL		 1395 return pcpu_rt;
1396}
1397
8f34e53b
DA		 1398static bool rt6_is_valid(const struct rt6_info *rt6)
1399{
1400 return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1401}
1402
66f5d6ce 1403/* It should be called with rcu_read_lock() acquired */
db3fedee 1404static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
d52d3997 1405{
c353071a 1406 struct rt6_info *pcpu_rt;
d52d3997 1407
c353071a 1408 pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
d52d3997 1409
8f34e53b
DA		 1410 if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1411 struct rt6_info *prev, **p;
1412
1413 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1414 prev = xchg(p, NULL);
1415 if (prev) {
1416 dst_dev_put(&prev->dst);
1417 dst_release(&prev->dst);
1418 }
1419
1420 pcpu_rt = NULL;
1421 }
1422
a73e4195
MKL		 1423 return pcpu_rt;
1424}
1425
afb1d4b5 1426static struct rt6_info *rt6_make_pcpu_route(struct net *net,
db3fedee 1427 const struct fib6_result *res)
a73e4195
MKL		 1428{
1429 struct rt6_info *pcpu_rt, *prev, **p;
d52d3997 1430
db3fedee 1431 pcpu_rt = ip6_rt_pcpu_alloc(res);
0e09edcc
WW		 1432 if (!pcpu_rt)
1433 return NULL;
d52d3997 1434
f40b6ae2 1435 p = this_cpu_ptr(res->nh->rt6i_pcpu);
a94b9367 1436 prev = cmpxchg(p, NULL, pcpu_rt);
951f788a 1437 BUG_ON(prev);
a94b9367 1438
61fb0d01
ED		 1439 if (res->f6i->fib6_destroying) {
1440 struct fib6_info *from;
1441
1442 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1443 fib6_info_release(from);
1444 }
1445
d52d3997
MKL		 1446 return pcpu_rt;
1447}
1448
35732d01
WW		 1449/* exception hash table implementation
1450 */
1451static DEFINE_SPINLOCK(rt6_exception_lock);
1452
1453/* Remove rt6_ex from hash table and free the memory
1454 * Caller must hold rt6_exception_lock
1455 */
1456static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1457 struct rt6_exception *rt6_ex)
1458{
f5b51fe8 1459 struct fib6_info *from;
b2427e67 1460 struct net *net;
81eb8447 1461
35732d01
WW		 1462 if (!bucket || !rt6_ex)
1463 return;
b2427e67
CIK		 1464
1465 net = dev_net(rt6_ex->rt6i->dst.dev);
f5b51fe8
PA		 1466 net->ipv6.rt6_stats->fib_rt_cache--;
1467
1468 /* purge completely the exception to allow releasing the held resources:
1469 * some [sk] cache may keep the dst around for unlimited time
1470 */
0e233874 1471 from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
f5b51fe8
PA		 1472 fib6_info_release(from);
1473 dst_dev_put(&rt6_ex->rt6i->dst);
1474
35732d01 1475 hlist_del_rcu(&rt6_ex->hlist);
77634cc6 1476 dst_release(&rt6_ex->rt6i->dst);
35732d01
WW		 1477 kfree_rcu(rt6_ex, rcu);
1478 WARN_ON_ONCE(!bucket->depth);
1479 bucket->depth--;
1480}
1481
1482/* Remove oldest rt6_ex in bucket and free the memory
1483 * Caller must hold rt6_exception_lock
1484 */
1485static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1486{
1487 struct rt6_exception *rt6_ex, *oldest = NULL;
1488
1489 if (!bucket)
1490 return;
1491
1492 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1493 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1494 oldest = rt6_ex;
1495 }
1496 rt6_remove_exception(bucket, oldest);
1497}
1498
1499static u32 rt6_exception_hash(const struct in6_addr *dst,
1500 const struct in6_addr *src)
1501{
1502 static u32 seed __read_mostly;
1503 u32 val;
1504
1505 net_get_random_once(&seed, sizeof(seed));
b6b556af 1506 val = jhash2((const u32 *)dst, sizeof(*dst)/sizeof(u32), seed);
35732d01
WW		 1507
1508#ifdef CONFIG_IPV6_SUBTREES
1509 if (src)
b6b556af 1510 val = jhash2((const u32 *)src, sizeof(*src)/sizeof(u32), val);
35732d01
WW		 1511#endif
1512 return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1513}
1514
1515/* Helper function to find the cached rt in the hash table
1516 * and update bucket pointer to point to the bucket for this
1517 * (daddr, saddr) pair
1518 * Caller must hold rt6_exception_lock
1519 */
1520static struct rt6_exception *
1521__rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1522 const struct in6_addr *daddr,
1523 const struct in6_addr *saddr)
1524{
1525 struct rt6_exception *rt6_ex;
1526 u32 hval;
1527
1528 if (!(*bucket) || !daddr)
1529 return NULL;
1530
1531 hval = rt6_exception_hash(daddr, saddr);
1532 *bucket += hval;
1533
1534 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1535 struct rt6_info *rt6 = rt6_ex->rt6i;
1536 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1537
1538#ifdef CONFIG_IPV6_SUBTREES
1539 if (matched && saddr)
1540 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1541#endif
1542 if (matched)
1543 return rt6_ex;
1544 }
1545 return NULL;
1546}
1547
1548/* Helper function to find the cached rt in the hash table
1549 * and update bucket pointer to point to the bucket for this
1550 * (daddr, saddr) pair
1551 * Caller must hold rcu_read_lock()
1552 */
1553static struct rt6_exception *
1554__rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1555 const struct in6_addr *daddr,
1556 const struct in6_addr *saddr)
1557{
1558 struct rt6_exception *rt6_ex;
1559 u32 hval;
1560
1561 WARN_ON_ONCE(!rcu_read_lock_held());
1562
1563 if (!(*bucket) || !daddr)
1564 return NULL;
1565
1566 hval = rt6_exception_hash(daddr, saddr);
1567 *bucket += hval;
1568
1569 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1570 struct rt6_info *rt6 = rt6_ex->rt6i;
1571 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1572
1573#ifdef CONFIG_IPV6_SUBTREES
1574 if (matched && saddr)
1575 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1576#endif
1577 if (matched)
1578 return rt6_ex;
1579 }
1580 return NULL;
1581}
1582
b748f260 1583static unsigned int fib6_mtu(const struct fib6_result *res)
d4ead6b3 1584{
b748f260 1585 const struct fib6_nh *nh = res->nh;
d4ead6b3
DA		 1586 unsigned int mtu;
1587
b748f260
DA		 1588 if (res->f6i->fib6_pmtu) {
1589 mtu = res->f6i->fib6_pmtu;
dcd1f572 1590 } else {
b748f260 1591 struct net_device *dev = nh->fib_nh_dev;
dcd1f572
DA		 1592 struct inet6_dev *idev;
1593
1594 rcu_read_lock();
1595 idev = __in6_dev_get(dev);
1596 mtu = idev->cnf.mtu6;
1597 rcu_read_unlock();
1598 }
1599
d4ead6b3
DA		 1600 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1601
b748f260 1602 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
d4ead6b3
DA		 1603}
1604
cc5c073a
DA		 1605#define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1606
1607/* used when the flushed bit is not relevant, only access to the bucket
1608 * (ie., all bucket users except rt6_insert_exception);
1609 *
1610 * called under rcu lock; sometimes called with rt6_exception_lock held
1611 */
1612static
1613struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1614 spinlock_t *lock)
1615{
1616 struct rt6_exception_bucket *bucket;
1617
1618 if (lock)
1619 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1620 lockdep_is_held(lock));
1621 else
1622 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1623
1624 /* remove bucket flushed bit if set */
1625 if (bucket) {
1626 unsigned long p = (unsigned long)bucket;
1627
1628 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1629 bucket = (struct rt6_exception_bucket *)p;
1630 }
1631
1632 return bucket;
1633}
1634
1635static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1636{
1637 unsigned long p = (unsigned long)bucket;
1638
1639 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1640}
1641
1642/* called with rt6_exception_lock held */
1643static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1644 spinlock_t *lock)
1645{
1646 struct rt6_exception_bucket *bucket;
1647 unsigned long p;
1648
1649 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1650 lockdep_is_held(lock));
1651
1652 p = (unsigned long)bucket;
1653 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1654 bucket = (struct rt6_exception_bucket *)p;
1655 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1656}
1657
35732d01 1658static int rt6_insert_exception(struct rt6_info *nrt,
5012f0a5 1659 const struct fib6_result *res)
35732d01 1660{
5e670d84 1661 struct net *net = dev_net(nrt->dst.dev);
35732d01 1662 struct rt6_exception_bucket *bucket;
cc5c073a 1663 struct fib6_info *f6i = res->f6i;
35732d01
WW		 1664 struct in6_addr *src_key = NULL;
1665 struct rt6_exception *rt6_ex;
cc5c073a 1666 struct fib6_nh *nh = res->nh;
35732d01
WW		 1667 int err = 0;
1668
35732d01
WW		 1669 spin_lock_bh(&rt6_exception_lock);
1670
cc5c073a
DA		 1671 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1672 lockdep_is_held(&rt6_exception_lock));
35732d01
WW		 1673 if (!bucket) {
1674 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1675 GFP_ATOMIC);
1676 if (!bucket) {
1677 err = -ENOMEM;
1678 goto out;
1679 }
cc5c073a
DA		 1680 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1681 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1682 err = -EINVAL;
1683 goto out;
35732d01
WW		 1684 }
1685
1686#ifdef CONFIG_IPV6_SUBTREES
5012f0a5 1687 /* fib6_src.plen != 0 indicates f6i is in subtree
35732d01 1688 * and exception table is indexed by a hash of
5012f0a5 1689 * both fib6_dst and fib6_src.
35732d01 1690 * Otherwise, the exception table is indexed by
5012f0a5 1691 * a hash of only fib6_dst.
35732d01 1692 */
5012f0a5 1693 if (f6i->fib6_src.plen)
35732d01
WW		 1694 src_key = &nrt->rt6i_src.addr;
1695#endif
5012f0a5 1696 /* rt6_mtu_change() might lower mtu on f6i.
f5bbe7ee 1697 * Only insert this exception route if its mtu
5012f0a5 1698 * is less than f6i's mtu value.
f5bbe7ee 1699 */
b748f260 1700 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
f5bbe7ee
WW		 1701 err = -EINVAL;
1702 goto out;
1703 }
60006a48 1704
35732d01
WW		 1705 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1706 src_key);
1707 if (rt6_ex)
1708 rt6_remove_exception(bucket, rt6_ex);
1709
1710 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1711 if (!rt6_ex) {
1712 err = -ENOMEM;
1713 goto out;
1714 }
1715 rt6_ex->rt6i = nrt;
1716 rt6_ex->stamp = jiffies;
35732d01
WW		 1717 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1718 bucket->depth++;
81eb8447 1719 net->ipv6.rt6_stats->fib_rt_cache++;
35732d01
WW		 1720
1721 if (bucket->depth > FIB6_MAX_DEPTH)
1722 rt6_exception_remove_oldest(bucket);
1723
1724out:
1725 spin_unlock_bh(&rt6_exception_lock);
1726
1727 /* Update fn->fn_sernum to invalidate all cached dst */
b886d5f2 1728 if (!err) {
5012f0a5
DA		 1729 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1730 fib6_update_sernum(net, f6i);
1731 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
b886d5f2
PA		 1732 fib6_force_start_gc(net);
1733 }
35732d01
WW		 1734
1735 return err;
1736}
1737
c0b220cf 1738static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
35732d01
WW		 1739{
1740 struct rt6_exception_bucket *bucket;
1741 struct rt6_exception *rt6_ex;
1742 struct hlist_node *tmp;
1743 int i;
1744
1745 spin_lock_bh(&rt6_exception_lock);
35732d01 1746
cc5c073a 1747 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
35732d01
WW		 1748 if (!bucket)
1749 goto out;
1750
cc5c073a
DA		 1751 /* Prevent rt6_insert_exception() to recreate the bucket list */
1752 if (!from)
1753 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1754
35732d01 1755 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
cc5c073a
DA		 1756 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1757 if (!from ||
1758 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1759 rt6_remove_exception(bucket, rt6_ex);
1760 }
1761 WARN_ON_ONCE(!from && bucket->depth);
35732d01
WW		 1762 bucket++;
1763 }
35732d01
WW		 1764out:
1765 spin_unlock_bh(&rt6_exception_lock);
1766}
1767
e659ba31
DA		 1768static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1769{
1770 struct fib6_info *f6i = arg;
1771
1772 fib6_nh_flush_exceptions(nh, f6i);
1773
1774 return 0;
1775}
1776
c0b220cf
DA		 1777void rt6_flush_exceptions(struct fib6_info *f6i)
1778{
e659ba31
DA		 1779 if (f6i->nh)
1780 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1781 f6i);
1782 else
1783 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
c0b220cf
DA		 1784}
1785
35732d01
WW		 1786/* Find cached rt in the hash table inside passed in rt
1787 * Caller has to hold rcu_read_lock()
1788 */
7e4b5128 1789static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
510e2ced
WW		 1790 const struct in6_addr *daddr,
1791 const struct in6_addr *saddr)
35732d01 1792{
510e2ced 1793 const struct in6_addr *src_key = NULL;
35732d01 1794 struct rt6_exception_bucket *bucket;
35732d01 1795 struct rt6_exception *rt6_ex;
7e4b5128 1796 struct rt6_info *ret = NULL;
35732d01 1797
35732d01 1798#ifdef CONFIG_IPV6_SUBTREES
7e4b5128 1799 /* fib6i_src.plen != 0 indicates f6i is in subtree
35732d01 1800 * and exception table is indexed by a hash of
7e4b5128 1801 * both fib6_dst and fib6_src.
510e2ced
WW		 1802 * However, the src addr used to create the hash
1803 * might not be exactly the passed in saddr which
1804 * is a /128 addr from the flow.
1805 * So we need to use f6i->fib6_src to redo lookup
1806 * if the passed in saddr does not find anything.
1807 * (See the logic in ip6_rt_cache_alloc() on how
1808 * rt->rt6i_src is updated.)
35732d01 1809 */
7e4b5128 1810 if (res->f6i->fib6_src.plen)
35732d01 1811 src_key = saddr;
510e2ced 1812find_ex:
35732d01 1813#endif
cc5c073a 1814 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
35732d01
WW		 1815 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1816
1817 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
7e4b5128 1818 ret = rt6_ex->rt6i;
35732d01 1819
510e2ced
WW		 1820#ifdef CONFIG_IPV6_SUBTREES
1821 /* Use fib6_src as src_key and redo lookup */
1822 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1823 src_key = &res->f6i->fib6_src.addr;
1824 goto find_ex;
1825 }
1826#endif
1827
7e4b5128 1828 return ret;
35732d01
WW		 1829}
1830
1831/* Remove the passed in cached rt from the hash table that contains it */
cc5c073a 1832static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1833 const struct rt6_info *rt)
35732d01 1834{
c0b220cf 1835 const struct in6_addr *src_key = NULL;
35732d01 1836 struct rt6_exception_bucket *bucket;
35732d01
WW		 1837 struct rt6_exception *rt6_ex;
1838 int err;
1839
cc5c073a 1840 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
35732d01
WW		 1841 return -ENOENT;
1842
1843 spin_lock_bh(&rt6_exception_lock);
cc5c073a
DA		 1844 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1845
35732d01 1846#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1847 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1848 * and exception table is indexed by a hash of
1849 * both rt6i_dst and rt6i_src.
35732d01
WW		 1850 * Otherwise, the exception table is indexed by
1851 * a hash of only rt6i_dst.
1852 */
c0b220cf 1853 if (plen)
35732d01
WW		 1854 src_key = &rt->rt6i_src.addr;
1855#endif
1856 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1857 &rt->rt6i_dst.addr,
1858 src_key);
1859 if (rt6_ex) {
1860 rt6_remove_exception(bucket, rt6_ex);
1861 err = 0;
1862 } else {
1863 err = -ENOENT;
1864 }
1865
1866 spin_unlock_bh(&rt6_exception_lock);
1867 return err;
1868}
1869
e659ba31
DA		 1870struct fib6_nh_excptn_arg {
1871 struct rt6_info *rt;
1872 int plen;
1873};
1874
1875static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1876{
1877 struct fib6_nh_excptn_arg *arg = _arg;
1878 int err;
1879
1880 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1881 if (err == 0)
1882 return 1;
1883
1884 return 0;
1885}
1886
c0b220cf
DA		 1887static int rt6_remove_exception_rt(struct rt6_info *rt)
1888{
1889 struct fib6_info *from;
1890
1891 from = rcu_dereference(rt->from);
cc5c073a 1892 if (!from || !(rt->rt6i_flags & RTF_CACHE))
c0b220cf
DA		 1893 return -EINVAL;
1894
e659ba31
DA		 1895 if (from->nh) {
1896 struct fib6_nh_excptn_arg arg = {
1897 .rt = rt,
1898 .plen = from->fib6_src.plen
1899 };
1900 int rc;
1901
1902 /* rc = 1 means an entry was found */
1903 rc = nexthop_for_each_fib6_nh(from->nh,
1904 rt6_nh_remove_exception_rt,
1905 &arg);
1906 return rc ? 0 : -ENOENT;
1907 }
1908
1cf844c7 1909 return fib6_nh_remove_exception(from->fib6_nh,
cc5c073a 1910 from->fib6_src.plen, rt);
c0b220cf
DA		 1911}
1912
35732d01
WW		 1913/* Find rt6_ex which contains the passed in rt cache and
1914 * refresh its stamp
1915 */
cc5c073a 1916static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1917 const struct rt6_info *rt)
35732d01 1918{
c0b220cf 1919 const struct in6_addr *src_key = NULL;
35732d01 1920 struct rt6_exception_bucket *bucket;
35732d01 1921 struct rt6_exception *rt6_ex;
193f3685 1922
cc5c073a 1923 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
35732d01 1924#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1925 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1926 * and exception table is indexed by a hash of
1927 * both rt6i_dst and rt6i_src.
35732d01
WW		 1928 * Otherwise, the exception table is indexed by
1929 * a hash of only rt6i_dst.
1930 */
c0b220cf 1931 if (plen)
35732d01
WW		 1932 src_key = &rt->rt6i_src.addr;
1933#endif
cc5c073a 1934 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
35732d01
WW		 1935 if (rt6_ex)
1936 rt6_ex->stamp = jiffies;
c0b220cf
DA		 1937}
1938
e659ba31
DA		 1939struct fib6_nh_match_arg {
1940 const struct net_device *dev;
1941 const struct in6_addr *gw;
1942 struct fib6_nh *match;
1943};
1944
1945/* determine if fib6_nh has given device and gateway */
1946static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1947{
1948 struct fib6_nh_match_arg *arg = _arg;
1949
1950 if (arg->dev != nh->fib_nh_dev ||
1951 (arg->gw && !nh->fib_nh_gw_family) ||
1952 (!arg->gw && nh->fib_nh_gw_family) ||
1953 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1954 return 0;
1955
1956 arg->match = nh;
1957
1958 /* found a match, break the loop */
1959 return 1;
1960}
1961
c0b220cf
DA		 1962static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1963{
1964 struct fib6_info *from;
e659ba31 1965 struct fib6_nh *fib6_nh;
35732d01 1966
c0b220cf
DA		 1967 rcu_read_lock();
1968
1969 from = rcu_dereference(rt->from);
1970 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1971 goto unlock;
1972
e659ba31
DA		 1973 if (from->nh) {
1974 struct fib6_nh_match_arg arg = {
1975 .dev = rt->dst.dev,
1976 .gw = &rt->rt6i_gateway,
1977 };
1978
1979 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1980
1981 if (!arg.match)
cff6a327 1982 goto unlock;
e659ba31
DA		 1983 fib6_nh = arg.match;
1984 } else {
1985 fib6_nh = from->fib6_nh;
1986 }
1987 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
193f3685 1988unlock:
35732d01
WW		 1989 rcu_read_unlock();
1990}
1991
e9fa1495
SB		 1992static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1993 struct rt6_info *rt, int mtu)
1994{
1995 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1996 * lowest MTU in the path: always allow updating the route PMTU to
1997 * reflect PMTU decreases.
1998 *
1999 * If the new MTU is higher, and the route PMTU is equal to the local
2000 * MTU, this means the old MTU is the lowest in the path, so allow
2001 * updating it: if other nodes now have lower MTUs, PMTU discovery will
2002 * handle this.
2003 */
2004
2005 if (dst_mtu(&rt->dst) >= mtu)
2006 return true;
2007
2008 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
2009 return true;
2010
2011 return false;
2012}
2013
2014static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
cc5c073a 2015 const struct fib6_nh *nh, int mtu)
f5bbe7ee
WW		 2016{
2017 struct rt6_exception_bucket *bucket;
2018 struct rt6_exception *rt6_ex;
2019 int i;
2020
cc5c073a 2021 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
e9fa1495
SB		 2022 if (!bucket)
2023 return;
2024
2025 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2026 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2027 struct rt6_info *entry = rt6_ex->rt6i;
2028
2029 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
d4ead6b3 2030 * route), the metrics of its rt->from have already
e9fa1495
SB		 2031 * been updated.
2032 */
d4ead6b3 2033 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
e9fa1495 2034 rt6_mtu_change_route_allowed(idev, entry, mtu))
d4ead6b3 2035 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
f5bbe7ee 2036 }
e9fa1495 2037 bucket++;
f5bbe7ee
WW		 2038 }
2039}
2040
b16cb459
WW		 2041#define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2042
cc5c073a
DA		 2043static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2044 const struct in6_addr *gateway)
b16cb459
WW		 2045{
2046 struct rt6_exception_bucket *bucket;
2047 struct rt6_exception *rt6_ex;
2048 struct hlist_node *tmp;
2049 int i;
2050
cc5c073a 2051 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
b16cb459
WW		 2052 return;
2053
2054 spin_lock_bh(&rt6_exception_lock);
cc5c073a 2055 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
b16cb459
WW		 2056 if (bucket) {
2057 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2058 hlist_for_each_entry_safe(rt6_ex, tmp,
2059 &bucket->chain, hlist) {
2060 struct rt6_info *entry = rt6_ex->rt6i;
2061
2062 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2063 RTF_CACHE_GATEWAY &&
2064 ipv6_addr_equal(gateway,
2065 &entry->rt6i_gateway)) {
2066 rt6_remove_exception(bucket, rt6_ex);
2067 }
2068 }
2069 bucket++;
2070 }
2071 }
2072
2073 spin_unlock_bh(&rt6_exception_lock);
2074}
2075
c757faa8
WW		 2076static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2077 struct rt6_exception *rt6_ex,
2078 struct fib6_gc_args *gc_args,
2079 unsigned long now)
2080{
2081 struct rt6_info *rt = rt6_ex->rt6i;
2082
1859bac0
PA		 2083 /* we are pruning and obsoleting aged-out and non gateway exceptions
2084 * even if others have still references to them, so that on next
2085 * dst_check() such references can be dropped.
2086 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2087 * expired, independently from their aging, as per RFC 8201 section 4
2088 */
31afeb42
WW		 2089 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2090 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2091 RT6_TRACE("aging clone %p\n", rt);
2092 rt6_remove_exception(bucket, rt6_ex);
2093 return;
2094 }
2095 } else if (time_after(jiffies, rt->dst.expires)) {
2096 RT6_TRACE("purging expired route %p\n", rt);
c757faa8
WW		 2097 rt6_remove_exception(bucket, rt6_ex);
2098 return;
31afeb42
WW		 2099 }
2100
2101 if (rt->rt6i_flags & RTF_GATEWAY) {
c757faa8
WW		 2102 struct neighbour *neigh;
2103 __u8 neigh_flags = 0;
2104
1bfa26ff
ED		 2105 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2106 if (neigh)
c757faa8 2107 neigh_flags = neigh->flags;
1bfa26ff 2108
c757faa8
WW		 2109 if (!(neigh_flags & NTF_ROUTER)) {
2110 RT6_TRACE("purging route %p via non-router but gateway\n",
2111 rt);
2112 rt6_remove_exception(bucket, rt6_ex);
2113 return;
2114 }
2115 }
31afeb42 2116
c757faa8
WW		 2117 gc_args->more++;
2118}
2119
cc5c073a 2120static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
c0b220cf
DA		 2121 struct fib6_gc_args *gc_args,
2122 unsigned long now)
c757faa8
WW		 2123{
2124 struct rt6_exception_bucket *bucket;
2125 struct rt6_exception *rt6_ex;
2126 struct hlist_node *tmp;
2127 int i;
2128
cc5c073a 2129 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
c757faa8
WW		 2130 return;
2131
1bfa26ff
ED		 2132 rcu_read_lock_bh();
2133 spin_lock(&rt6_exception_lock);
cc5c073a 2134 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
c757faa8
WW		 2135 if (bucket) {
2136 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2137 hlist_for_each_entry_safe(rt6_ex, tmp,
2138 &bucket->chain, hlist) {
2139 rt6_age_examine_exception(bucket, rt6_ex,
2140 gc_args, now);
2141 }
2142 bucket++;
2143 }
2144 }
1bfa26ff
ED		 2145 spin_unlock(&rt6_exception_lock);
2146 rcu_read_unlock_bh();
c757faa8
WW		 2147}
2148
e659ba31
DA		 2149struct fib6_nh_age_excptn_arg {
2150 struct fib6_gc_args *gc_args;
2151 unsigned long now;
2152};
2153
2154static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2155{
2156 struct fib6_nh_age_excptn_arg *arg = _arg;
2157
2158 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2159 return 0;
2160}
2161
cc5c073a 2162void rt6_age_exceptions(struct fib6_info *f6i,
c0b220cf
DA		 2163 struct fib6_gc_args *gc_args,
2164 unsigned long now)
2165{
e659ba31
DA		 2166 if (f6i->nh) {
2167 struct fib6_nh_age_excptn_arg arg = {
2168 .gc_args = gc_args,
2169 .now = now
2170 };
2171
2172 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2173 &arg);
2174 } else {
2175 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2176 }
c0b220cf
DA		 2177}
2178
1d053da9 2179/* must be called with rcu lock held */
effda4dd
DA		 2180int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2181 struct flowi6 *fl6, struct fib6_result *res, int strict)
1da177e4 2182{
367efcb9 2183 struct fib6_node *fn, *saved_fn;
1da177e4 2184
6454743b 2185 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
367efcb9 2186 saved_fn = fn;
1da177e4 2187
ca254490
DA		 2188 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2189 oif = 0;
2190
a3c00e46 2191redo_rt6_select:
effda4dd
DA		 2192 rt6_select(net, fn, oif, res, strict);
2193 if (res->f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2194 fn = fib6_backtrack(fn, &fl6->saddr);
2195 if (fn)
2196 goto redo_rt6_select;
367efcb9
MKL		 2197 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2198 /* also consider unreachable route */
2199 strict &= ~RT6_LOOKUP_F_REACHABLE;
2200 fn = saved_fn;
2201 goto redo_rt6_select;
367efcb9 2202 }
a3c00e46
MKL		 2203 }
2204
effda4dd 2205 trace_fib6_table_lookup(net, res, table, fl6);
fb9de91e 2206
effda4dd 2207 return 0;
1d053da9
DA		 2208}
2209
2210struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2211 int oif, struct flowi6 *fl6,
2212 const struct sk_buff *skb, int flags)
2213{
b1d40991 2214 struct fib6_result res = {};
0e09edcc 2215 struct rt6_info *rt = NULL;
1d053da9
DA		 2216 int strict = 0;
2217
0e09edcc
WW		 2218 WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2219 !rcu_read_lock_held());
2220
1d053da9
DA		 2221 strict |= flags & RT6_LOOKUP_F_IFACE;
2222 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2223 if (net->ipv6.devconf_all->forwarding == 0)
2224 strict |= RT6_LOOKUP_F_REACHABLE;
2225
2226 rcu_read_lock();
2227
effda4dd 2228 fib6_table_lookup(net, table, oif, fl6, &res, strict);
0e09edcc
WW		 2229 if (res.f6i == net->ipv6.fib6_null_entry)
2230 goto out;
23fb93a4 2231
b1d40991 2232 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
d83009d4 2233
23fb93a4 2234 /*Search through exception table */
7e4b5128 2235 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
23fb93a4 2236 if (rt) {
0e09edcc 2237 goto out;
3da59bd9 2238 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
b1d40991 2239 !res.nh->fib_nh_gw_family)) {
3da59bd9
MKL		 2240 /* Create a RTF_CACHE clone which will not be
2241 * owned by the fib6 tree. It is for the special case where
2242 * the daddr in the skb during the neighbor look-up is different
2243 * from the fl6->daddr used to look-up route here.
2244 */
0e09edcc 2245 rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
c71099ac 2246
0e09edcc
WW		 2247 if (rt) {
2248 /* 1 refcnt is taken during ip6_rt_cache_alloc().
2249 * As rt6_uncached_list_add() does not consume refcnt,
2250 * this refcnt is always returned to the caller even
2251 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
1cfb71ee 2252 */
0e09edcc 2253 rt6_uncached_list_add(rt);
81eb8447 2254 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
0e09edcc 2255 rcu_read_unlock();
b811580d 2256
0e09edcc
WW		 2257 return rt;
2258 }
d52d3997
MKL		 2259 } else {
2260 /* Get a percpu copy */
951f788a 2261 local_bh_disable();
0e09edcc 2262 rt = rt6_get_pcpu_route(&res);
d52d3997 2263
0e09edcc
WW		 2264 if (!rt)
2265 rt = rt6_make_pcpu_route(net, &res);
93531c67 2266
951f788a 2267 local_bh_enable();
d52d3997 2268 }
0e09edcc
WW		 2269out:
2270 if (!rt)
2271 rt = net->ipv6.ip6_null_entry;
2272 if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2273 ip6_hold_safe(net, &rt);
2274 rcu_read_unlock();
2275
2276 return rt;
1da177e4 2277}
9ff74384 2278EXPORT_SYMBOL_GPL(ip6_pol_route);
1da177e4 2279
55cced4f 2280INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_input(struct net *net,
b75cc8f9
DA		 2281 struct fib6_table *table,
2282 struct flowi6 *fl6,
2283 const struct sk_buff *skb,
2284 int flags)
4acad72d 2285{
b75cc8f9 2286 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
4acad72d
PE		 2287}
2288
d409b847
MB		 2289struct dst_entry *ip6_route_input_lookup(struct net *net,
2290 struct net_device *dev,
b75cc8f9
DA		 2291 struct flowi6 *fl6,
2292 const struct sk_buff *skb,
2293 int flags)
72331bc0
SL		 2294{
2295 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2296 flags |= RT6_LOOKUP_F_IFACE;
2297
b75cc8f9 2298 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
72331bc0 2299}
d409b847 2300EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
72331bc0 2301
23aebdac 2302static void ip6_multipath_l3_keys(const struct sk_buff *skb,
5e5d6fed
RP		 2303 struct flow_keys *keys,
2304 struct flow_keys *flkeys)
23aebdac
JS		 2305{
2306 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2307 const struct ipv6hdr *key_iph = outer_iph;
5e5d6fed 2308 struct flow_keys *_flkeys = flkeys;
23aebdac
JS		 2309 const struct ipv6hdr *inner_iph;
2310 const struct icmp6hdr *icmph;
2311 struct ipv6hdr _inner_iph;
cea67a2d 2312 struct icmp6hdr _icmph;
23aebdac
JS		 2313
2314 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2315 goto out;
2316
cea67a2d
ED		 2317 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2318 sizeof(_icmph), &_icmph);
2319 if (!icmph)
2320 goto out;
2321
54074f1d 2322 if (!icmpv6_is_err(icmph->icmp6_type))
23aebdac
JS		 2323 goto out;
2324
2325 inner_iph = skb_header_pointer(skb,
2326 skb_transport_offset(skb) + sizeof(*icmph),
2327 sizeof(_inner_iph), &_inner_iph);
2328 if (!inner_iph)
2329 goto out;
2330
2331 key_iph = inner_iph;
5e5d6fed 2332 _flkeys = NULL;
23aebdac 2333out:
5e5d6fed
RP		 2334 if (_flkeys) {
2335 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2336 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2337 keys->tags.flow_label = _flkeys->tags.flow_label;
2338 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2339 } else {
2340 keys->addrs.v6addrs.src = key_iph->saddr;
2341 keys->addrs.v6addrs.dst = key_iph->daddr;
fa1be7e0 2342 keys->tags.flow_label = ip6_flowlabel(key_iph);
5e5d6fed
RP		 2343 keys->basic.ip_proto = key_iph->nexthdr;
2344 }
23aebdac
JS		 2345}
2346
2347/* if skb is set it will be used and fl6 can be NULL */
b4bac172
DA		 2348u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2349 const struct sk_buff *skb, struct flow_keys *flkeys)
23aebdac
JS		 2350{
2351 struct flow_keys hash_keys;
9a2a537a 2352 u32 mhash;
23aebdac 2353
bbfa047a 2354 switch (ip6_multipath_hash_policy(net)) {
b4bac172
DA		 2355 case 0:
2356 memset(&hash_keys, 0, sizeof(hash_keys));
2357 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2358 if (skb) {
2359 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2360 } else {
2361 hash_keys.addrs.v6addrs.src = fl6->saddr;
2362 hash_keys.addrs.v6addrs.dst = fl6->daddr;
fa1be7e0 2363 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
b4bac172
DA		 2364 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2365 }
2366 break;
2367 case 1:
2368 if (skb) {
2369 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2370 struct flow_keys keys;
2371
2372 /* short-circuit if we already have L4 hash present */
2373 if (skb->l4_hash)
2374 return skb_get_hash_raw(skb) >> 1;
2375
2376 memset(&hash_keys, 0, sizeof(hash_keys));
2377
2378 if (!flkeys) {
2379 skb_flow_dissect_flow_keys(skb, &keys, flag);
2380 flkeys = &keys;
2381 }
2382 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2383 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2384 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2385 hash_keys.ports.src = flkeys->ports.src;
2386 hash_keys.ports.dst = flkeys->ports.dst;
2387 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2388 } else {
2389 memset(&hash_keys, 0, sizeof(hash_keys));
2390 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2391 hash_keys.addrs.v6addrs.src = fl6->saddr;
2392 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2393 hash_keys.ports.src = fl6->fl6_sport;
2394 hash_keys.ports.dst = fl6->fl6_dport;
2395 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2396 }
2397 break;
d8f74f09
SS		 2398 case 2:
2399 memset(&hash_keys, 0, sizeof(hash_keys));
2400 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2401 if (skb) {
2402 struct flow_keys keys;
2403
2404 if (!flkeys) {
2405 skb_flow_dissect_flow_keys(skb, &keys, 0);
2406 flkeys = &keys;
2407 }
2408
2409 /* Inner can be v4 or v6 */
2410 if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2411 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2412 hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2413 hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2414 } else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2415 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2416 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2417 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2418 hash_keys.tags.flow_label = flkeys->tags.flow_label;
2419 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2420 } else {
2421 /* Same as case 0 */
2422 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2423 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2424 }
2425 } else {
2426 /* Same as case 0 */
2427 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2428 hash_keys.addrs.v6addrs.src = fl6->saddr;
2429 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2430 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2431 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2432 }
2433 break;
23aebdac 2434 }
9a2a537a 2435 mhash = flow_hash_from_keys(&hash_keys);
23aebdac 2436
9a2a537a 2437 return mhash >> 1;
23aebdac
JS		 2438}
2439
67f415dd 2440/* Called with rcu held */
c71099ac
TG		 2441void ip6_route_input(struct sk_buff *skb)
2442{
b71d1d42 2443 const struct ipv6hdr *iph = ipv6_hdr(skb);
c346dca1 2444 struct net *net = dev_net(skb->dev);
67f415dd 2445 int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
904af04d 2446 struct ip_tunnel_info *tun_info;
4c9483b2 2447 struct flowi6 fl6 = {
e0d56fdd 2448 .flowi6_iif = skb->dev->ifindex,
4c9483b2
DM		 2449 .daddr = iph->daddr,
2450 .saddr = iph->saddr,
6502ca52 2451 .flowlabel = ip6_flowinfo(iph),
4c9483b2
DM		 2452 .flowi6_mark = skb->mark,
2453 .flowi6_proto = iph->nexthdr,
c71099ac 2454 };
5e5d6fed 2455 struct flow_keys *flkeys = NULL, _flkeys;
adaa70bb 2456
904af04d 2457 tun_info = skb_tunnel_info(skb);
46fa062a 2458 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
904af04d 2459 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
5e5d6fed
RP		 2460
2461 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2462 flkeys = &_flkeys;
2463
23aebdac 2464 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
b4bac172 2465 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
06e9d040 2466 skb_dst_drop(skb);
67f415dd
WW		 2467 skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2468 &fl6, skb, flags));
c71099ac
TG		 2469}
2470
55cced4f 2471INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_output(struct net *net,
b75cc8f9
DA		 2472 struct fib6_table *table,
2473 struct flowi6 *fl6,
2474 const struct sk_buff *skb,
2475 int flags)
1da177e4 2476{
b75cc8f9 2477 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
c71099ac
TG		 2478}
2479
7d9e5f42
WW		 2480struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2481 const struct sock *sk,
2482 struct flowi6 *fl6, int flags)
c71099ac 2483{
d46a9d67 2484 bool any_src;
c71099ac 2485
3ede0bbc
RS		 2486 if (ipv6_addr_type(&fl6->daddr) &
2487 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
4c1feac5
DA		 2488 struct dst_entry *dst;
2489
7d9e5f42 2490 /* This function does not take refcnt on the dst */
4c1feac5
DA		 2491 dst = l3mdev_link_scope_lookup(net, fl6);
2492 if (dst)
2493 return dst;
2494 }
ca254490 2495
1fb9489b 2496 fl6->flowi6_iif = LOOPBACK_IFINDEX;
4dc27d1c 2497
7d9e5f42 2498 flags |= RT6_LOOKUP_F_DST_NOREF;
d46a9d67 2499 any_src = ipv6_addr_any(&fl6->saddr);
741a11d9 2500 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
d46a9d67 2501 (fl6->flowi6_oif && any_src))
77d16f45 2502 flags |= RT6_LOOKUP_F_IFACE;
c71099ac 2503
d46a9d67 2504 if (!any_src)
adaa70bb 2505 flags |= RT6_LOOKUP_F_HAS_SADDR;
0c9a2ac1
YH		 2506 else if (sk)
2507 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
adaa70bb 2508
b75cc8f9 2509 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
1da177e4 2510}
7d9e5f42
WW		 2511EXPORT_SYMBOL_GPL(ip6_route_output_flags_noref);
2512
2513struct dst_entry *ip6_route_output_flags(struct net *net,
2514 const struct sock *sk,
2515 struct flowi6 *fl6,
2516 int flags)
2517{
2518 struct dst_entry *dst;
2519 struct rt6_info *rt6;
2520
2521 rcu_read_lock();
2522 dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2523 rt6 = (struct rt6_info *)dst;
2524 /* For dst cached in uncached_list, refcnt is already taken. */
2525 if (list_empty(&rt6->rt6i_uncached) && !dst_hold_safe(dst)) {
2526 dst = &net->ipv6.ip6_null_entry->dst;
2527 dst_hold(dst);
2528 }
2529 rcu_read_unlock();
2530
2531 return dst;
2532}
6f21c96a 2533EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1da177e4 2534
2774c131 2535struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
14e50e57 2536{
5c1e6aa3 2537 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1dbe3252 2538 struct net_device *loopback_dev = net->loopback_dev;
14e50e57
DM		 2539 struct dst_entry *new = NULL;
2540
1dbe3252 2541 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
62cf27e5 2542 DST_OBSOLETE_DEAD, 0);
14e50e57 2543 if (rt) {
0a1f5962 2544 rt6_info_init(rt);
81eb8447 2545 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
8104891b 2546
0a1f5962 2547 new = &rt->dst;
14e50e57 2548 new->__use = 1;
352e512c 2549 new->input = dst_discard;
ede2059d 2550 new->output = dst_discard_out;
14e50e57 2551
0a1f5962 2552 dst_copy_metrics(new, &ort->dst);
14e50e57 2553
1dbe3252 2554 rt->rt6i_idev = in6_dev_get(loopback_dev);
4e3fd7a0 2555 rt->rt6i_gateway = ort->rt6i_gateway;
0a1f5962 2556 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
14e50e57
DM		 2557
2558 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2559#ifdef CONFIG_IPV6_SUBTREES
2560 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2561#endif
14e50e57
DM		 2562 }
2563
69ead7af
DM		 2564 dst_release(dst_orig);
2565 return new ? new : ERR_PTR(-ENOMEM);
14e50e57 2566}
14e50e57 2567
1da177e4
LT		 2568/*
2569 * Destination cache support functions
2570 */
2571
8d1c802b 2572static bool fib6_check(struct fib6_info *f6i, u32 cookie)
4b32b5ad 2573{
93531c67
DA		 2574 u32 rt_cookie = 0;
2575
8ae86971 2576 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
93531c67
DA		 2577 return false;
2578
2579 if (fib6_check_expired(f6i))
2580 return false;
2581
2582 return true;
4b32b5ad
MKL		 2583}
2584
a68886a6
DA		 2585static struct dst_entry *rt6_check(struct rt6_info *rt,
2586 struct fib6_info *from,
2587 u32 cookie)
3da59bd9 2588{
36143645 2589 u32 rt_cookie = 0;
c5cff856 2590
49d05fe2 2591 if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
93531c67 2592 rt_cookie != cookie)
3da59bd9
MKL		 2593 return NULL;
2594
2595 if (rt6_check_expired(rt))
2596 return NULL;
2597
2598 return &rt->dst;
2599}
2600
a68886a6
DA		 2601static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2602 struct fib6_info *from,
2603 u32 cookie)
3da59bd9 2604{
5973fb1e
MKL		 2605 if (!__rt6_check_expired(rt) &&
2606 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
a68886a6 2607 fib6_check(from, cookie))
3da59bd9
MKL		 2608 return &rt->dst;
2609 else
2610 return NULL;
2611}
2612
1da177e4
LT		 2613static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
2614{
a87b7dc9 2615 struct dst_entry *dst_ret;
a68886a6 2616 struct fib6_info *from;
1da177e4
LT		 2617 struct rt6_info *rt;
2618
a87b7dc9
DA		 2619 rt = container_of(dst, struct rt6_info, dst);
2620
8f34e53b
DA		 2621 if (rt->sernum)
2622 return rt6_is_valid(rt) ? dst : NULL;
2623
a87b7dc9 2624 rcu_read_lock();
1da177e4 2625
6f3118b5
ND		 2626 /* All IPV6 dsts are created with ->obsolete set to the value
2627 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2628 * into this function always.
2629 */
e3bc10bd 2630
a68886a6 2631 from = rcu_dereference(rt->from);
4b32b5ad 2632
a68886a6
DA		 2633 if (from && (rt->rt6i_flags & RTF_PCPU ||
2634 unlikely(!list_empty(&rt->rt6i_uncached))))
2635 dst_ret = rt6_dst_from_check(rt, from, cookie);
3da59bd9 2636 else
a68886a6 2637 dst_ret = rt6_check(rt, from, cookie);
a87b7dc9
DA		 2638
2639 rcu_read_unlock();
2640
2641 return dst_ret;
1da177e4
LT		 2642}
2643
2644static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2645{
2646 struct rt6_info *rt = (struct rt6_info *) dst;
2647
2648 if (rt) {
54c1a859 2649 if (rt->rt6i_flags & RTF_CACHE) {
c3c14da0 2650 rcu_read_lock();
54c1a859 2651 if (rt6_check_expired(rt)) {
93531c67 2652 rt6_remove_exception_rt(rt);
54c1a859
YH		 2653 dst = NULL;
2654 }
c3c14da0 2655 rcu_read_unlock();
54c1a859 2656 } else {
1da177e4 2657 dst_release(dst);
54c1a859
YH		 2658 dst = NULL;
2659 }
1da177e4 2660 }
54c1a859 2661 return dst;
1da177e4
LT		 2662}
2663
2664static void ip6_link_failure(struct sk_buff *skb)
2665{
2666 struct rt6_info *rt;
2667
3ffe533c 2668 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1da177e4 2669
adf30907 2670 rt = (struct rt6_info *) skb_dst(skb);
1da177e4 2671 if (rt) {
8a14e46f 2672 rcu_read_lock();
1eb4f758 2673 if (rt->rt6i_flags & RTF_CACHE) {
761f6026 2674 rt6_remove_exception_rt(rt);
c5cff856 2675 } else {
a68886a6 2676 struct fib6_info *from;
c5cff856
WW		 2677 struct fib6_node *fn;
2678
a68886a6
DA		 2679 from = rcu_dereference(rt->from);
2680 if (from) {
2681 fn = rcu_dereference(from->fib6_node);
2682 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2683 fn->fn_sernum = -1;
2684 }
1eb4f758 2685 }
8a14e46f 2686 rcu_read_unlock();
1da177e4
LT		 2687 }
2688}
2689
6a3e030f
DA		 2690static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2691{
a68886a6
DA		 2692 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2693 struct fib6_info *from;
2694
2695 rcu_read_lock();
2696 from = rcu_dereference(rt0->from);
2697 if (from)
2698 rt0->dst.expires = from->expires;
2699 rcu_read_unlock();
2700 }
6a3e030f
DA		 2701
2702 dst_set_expires(&rt0->dst, timeout);
2703 rt0->rt6i_flags |= RTF_EXPIRES;
2704}
2705
45e4fd26
MKL		 2706static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2707{
2708 struct net *net = dev_net(rt->dst.dev);
2709
d4ead6b3 2710 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
45e4fd26 2711 rt->rt6i_flags |= RTF_MODIFIED;
45e4fd26
MKL		 2712 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2713}
2714
0d3f6d29
MKL		 2715static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2716{
2717 return !(rt->rt6i_flags & RTF_CACHE) &&
1490ed2a 2718 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
0d3f6d29
MKL		 2719}
2720
45e4fd26 2721static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
bd085ef6
HL		 2722 const struct ipv6hdr *iph, u32 mtu,
2723 bool confirm_neigh)
1da177e4 2724{
0dec879f 2725 const struct in6_addr *daddr, *saddr;
67ba4152 2726 struct rt6_info *rt6 = (struct rt6_info *)dst;
1da177e4 2727
09454fd0
MŻ		 2728 /* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU)
2729 * IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it.
2730 * [see also comment in rt6_mtu_change_route()]
2731 */
19bda36c 2732
0dec879f
JA		 2733 if (iph) {
2734 daddr = &iph->daddr;
2735 saddr = &iph->saddr;
2736 } else if (sk) {
2737 daddr = &sk->sk_v6_daddr;
2738 saddr = &inet6_sk(sk)->saddr;
2739 } else {
2740 daddr = NULL;
2741 saddr = NULL;
2742 }
bd085ef6
HL		 2743
2744 if (confirm_neigh)
2745 dst_confirm_neigh(dst, daddr);
2746
45e4fd26
MKL		 2747 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
2748 if (mtu >= dst_mtu(dst))
2749 return;
9d289715 2750
0d3f6d29 2751 if (!rt6_cache_allowed_for_pmtu(rt6)) {
45e4fd26 2752 rt6_do_update_pmtu(rt6, mtu);
2b760fcf
WW		 2753 /* update rt6_ex->stamp for cache */
2754 if (rt6->rt6i_flags & RTF_CACHE)
2755 rt6_update_exception_stamp_rt(rt6);
0dec879f 2756 } else if (daddr) {
85bd05de 2757 struct fib6_result res = {};
45e4fd26
MKL		 2758 struct rt6_info *nrt6;
2759
4d85cd0c 2760 rcu_read_lock();
85bd05de 2761 res.f6i = rcu_dereference(rt6->from);
43a4b60d
DA		 2762 if (!res.f6i)
2763 goto out_unlock;
2764
7d21fec9
DA		 2765 res.fib6_flags = res.f6i->fib6_flags;
2766 res.fib6_type = res.f6i->fib6_type;
2767
2d44234b
DA		 2768 if (res.f6i->nh) {
2769 struct fib6_nh_match_arg arg = {
2770 .dev = dst->dev,
2771 .gw = &rt6->rt6i_gateway,
2772 };
2773
2774 nexthop_for_each_fib6_nh(res.f6i->nh,
2775 fib6_nh_find_match, &arg);
2776
2777 /* fib6_info uses a nexthop that does not have fib6_nh
2778 * using the dst->dev + gw. Should be impossible.
2779 */
43a4b60d
DA		 2780 if (!arg.match)
2781 goto out_unlock;
2d44234b
DA		 2782
2783 res.nh = arg.match;
2784 } else {
2785 res.nh = res.f6i->fib6_nh;
2786 }
2787
85bd05de 2788 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
45e4fd26
MKL		 2789 if (nrt6) {
2790 rt6_do_update_pmtu(nrt6, mtu);
5012f0a5 2791 if (rt6_insert_exception(nrt6, &res))
2b760fcf 2792 dst_release_immediate(&nrt6->dst);
45e4fd26 2793 }
43a4b60d 2794out_unlock:
a68886a6 2795 rcu_read_unlock();
1da177e4
LT		 2796 }
2797}
2798
45e4fd26 2799static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
bd085ef6
HL		 2800 struct sk_buff *skb, u32 mtu,
2801 bool confirm_neigh)
45e4fd26 2802{
bd085ef6
HL		 2803 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2804 confirm_neigh);
45e4fd26
MKL		 2805}
2806
42ae66c8 2807void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
e2d118a1 2808 int oif, u32 mark, kuid_t uid)
81aded24
DM		 2809{
2810 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2811 struct dst_entry *dst;
dc92095d
MŻ		 2812 struct flowi6 fl6 = {
2813 .flowi6_oif = oif,
2814 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2815 .daddr = iph->daddr,
2816 .saddr = iph->saddr,
2817 .flowlabel = ip6_flowinfo(iph),
2818 .flowi6_uid = uid,
2819 };
81aded24
DM		 2820
2821 dst = ip6_route_output(net, NULL, &fl6);
2822 if (!dst->error)
bd085ef6 2823 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
81aded24
DM		 2824 dst_release(dst);
2825}
2826EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2827
2828void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2829{
7ddacfa5 2830 int oif = sk->sk_bound_dev_if;
33c162a9
MKL		 2831 struct dst_entry *dst;
2832
7ddacfa5
DA		 2833 if (!oif && skb->dev)
2834 oif = l3mdev_master_ifindex(skb->dev);
2835
2836 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
33c162a9
MKL		 2837
2838 dst = __sk_dst_get(sk);
2839 if (!dst || !dst->obsolete ||
2840 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2841 return;
2842
2843 bh_lock_sock(sk);
2844 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2845 ip6_datagram_dst_update(sk, false);
2846 bh_unlock_sock(sk);
81aded24
DM		 2847}
2848EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2849
7d6850f7
AK		 2850void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2851 const struct flowi6 *fl6)
2852{
2853#ifdef CONFIG_IPV6_SUBTREES
2854 struct ipv6_pinfo *np = inet6_sk(sk);
2855#endif
2856
2857 ip6_dst_store(sk, dst,
2858 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2859 &sk->sk_v6_daddr : NULL,
2860#ifdef CONFIG_IPV6_SUBTREES
2861 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2862 &np->saddr :
2863#endif
2864 NULL);
2865}
2866
9b6b35ab 2867static bool ip6_redirect_nh_match(const struct fib6_result *res,
0b34eb00
DA		 2868 struct flowi6 *fl6,
2869 const struct in6_addr *gw,
2870 struct rt6_info **ret)
2871{
9b6b35ab
DA		 2872 const struct fib6_nh *nh = res->nh;
2873
0b34eb00
DA		 2874 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2875 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2876 return false;
2877
2878 /* rt_cache's gateway might be different from its 'parent'
2879 * in the case of an ip redirect.
2880 * So we keep searching in the exception table if the gateway
2881 * is different.
2882 */
2883 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
2884 struct rt6_info *rt_cache;
2885
9b6b35ab 2886 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
0b34eb00
DA		 2887 if (rt_cache &&
2888 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
2889 *ret = rt_cache;
2890 return true;
2891 }
2892 return false;
2893 }
2894 return true;
2895}
2896
c55c8988
DA		 2897struct fib6_nh_rd_arg {
2898 struct fib6_result *res;
2899 struct flowi6 *fl6;
2900 const struct in6_addr *gw;
2901 struct rt6_info **ret;
2902};
2903
2904static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
2905{
2906 struct fib6_nh_rd_arg *arg = _arg;
2907
2908 arg->res->nh = nh;
2909 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
2910}
2911
b55b76b2
DJ		 2912/* Handle redirects */
2913struct ip6rd_flowi {
2914 struct flowi6 fl6;
2915 struct in6_addr gateway;
2916};
2917
55cced4f 2918INDIRECT_CALLABLE_SCOPE struct rt6_info *__ip6_route_redirect(struct net *net,
b55b76b2
DJ		 2919 struct fib6_table *table,
2920 struct flowi6 *fl6,
b75cc8f9 2921 const struct sk_buff *skb,
b55b76b2
DJ		 2922 int flags)
2923{
2924 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
0b34eb00 2925 struct rt6_info *ret = NULL;
9b6b35ab 2926 struct fib6_result res = {};
c55c8988
DA		 2927 struct fib6_nh_rd_arg arg = {
2928 .res = &res,
2929 .fl6 = fl6,
2930 .gw = &rdfl->gateway,
2931 .ret = &ret
2932 };
8d1c802b 2933 struct fib6_info *rt;
b55b76b2
DJ		 2934 struct fib6_node *fn;
2935
31680ac2
DA		 2936 /* l3mdev_update_flow overrides oif if the device is enslaved; in
2937 * this case we must match on the real ingress device, so reset it
2938 */
2939 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2940 fl6->flowi6_oif = skb->dev->ifindex;
2941
b55b76b2 2942 /* Get the "current" route for this destination and
67c408cf 2943 * check if the redirect has come from appropriate router.
b55b76b2
DJ		 2944 *
2945 * RFC 4861 specifies that redirects should only be
2946 * accepted if they come from the nexthop to the target.
2947 * Due to the way the routes are chosen, this notion
2948 * is a bit fuzzy and one might need to check all possible
2949 * routes.
2950 */
2951
66f5d6ce 2952 rcu_read_lock();
6454743b 2953 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
b55b76b2 2954restart:
66f5d6ce 2955 for_each_fib6_node_rt_rcu(fn) {
9b6b35ab 2956 res.f6i = rt;
14895687 2957 if (fib6_check_expired(rt))
b55b76b2 2958 continue;
93c2fb25 2959 if (rt->fib6_flags & RTF_REJECT)
b55b76b2 2960 break;
c55c8988
DA		 2961 if (unlikely(rt->nh)) {
2962 if (nexthop_is_blackhole(rt->nh))
2963 continue;
2964 /* on match, res->nh is filled in and potentially ret */
2965 if (nexthop_for_each_fib6_nh(rt->nh,
2966 fib6_nh_redirect_match,
2967 &arg))
2968 goto out;
2969 } else {
2970 res.nh = rt->fib6_nh;
2971 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
2972 &ret))
2973 goto out;
2974 }
b55b76b2
DJ		 2975 }
2976
2977 if (!rt)
421842ed 2978 rt = net->ipv6.fib6_null_entry;
93c2fb25 2979 else if (rt->fib6_flags & RTF_REJECT) {
23fb93a4 2980 ret = net->ipv6.ip6_null_entry;
b0a1ba59
MKL		 2981 goto out;
2982 }
2983
421842ed 2984 if (rt == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2985 fn = fib6_backtrack(fn, &fl6->saddr);
2986 if (fn)
2987 goto restart;
b55b76b2 2988 }
a3c00e46 2989
9b6b35ab 2990 res.f6i = rt;
1cf844c7 2991 res.nh = rt->fib6_nh;
b0a1ba59 2992out:
7d21fec9 2993 if (ret) {
10585b43 2994 ip6_hold_safe(net, &ret);
7d21fec9
DA		 2995 } else {
2996 res.fib6_flags = res.f6i->fib6_flags;
2997 res.fib6_type = res.f6i->fib6_type;
9b6b35ab 2998 ret = ip6_create_rt_rcu(&res);
7d21fec9 2999 }
b55b76b2 3000
66f5d6ce 3001 rcu_read_unlock();
b55b76b2 3002
8ff2e5b2 3003 trace_fib6_table_lookup(net, &res, table, fl6);
23fb93a4 3004 return ret;
b55b76b2
DJ		 3005};
3006
3007static struct dst_entry *ip6_route_redirect(struct net *net,
b75cc8f9
DA		 3008 const struct flowi6 *fl6,
3009 const struct sk_buff *skb,
3010 const struct in6_addr *gateway)
b55b76b2
DJ		 3011{
3012 int flags = RT6_LOOKUP_F_HAS_SADDR;
3013 struct ip6rd_flowi rdfl;
3014
3015 rdfl.fl6 = *fl6;
3016 rdfl.gateway = *gateway;
3017
b75cc8f9 3018 return fib6_rule_lookup(net, &rdfl.fl6, skb,
b55b76b2
DJ		 3019 flags, __ip6_route_redirect);
3020}
3021
e2d118a1
LC		 3022void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
3023 kuid_t uid)
3a5ad2ee
DM		 3024{
3025 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
3026 struct dst_entry *dst;
1f7f10ac
MŻ		 3027 struct flowi6 fl6 = {
3028 .flowi6_iif = LOOPBACK_IFINDEX,
3029 .flowi6_oif = oif,
3030 .flowi6_mark = mark,
3031 .daddr = iph->daddr,
3032 .saddr = iph->saddr,
3033 .flowlabel = ip6_flowinfo(iph),
3034 .flowi6_uid = uid,
3035 };
3a5ad2ee 3036
b75cc8f9 3037 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
b55b76b2 3038 rt6_do_redirect(dst, NULL, skb);
3a5ad2ee
DM		 3039 dst_release(dst);
3040}
3041EXPORT_SYMBOL_GPL(ip6_redirect);
3042
d456336d 3043void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
c92a59ec
DJ		 3044{
3045 const struct ipv6hdr *iph = ipv6_hdr(skb);
3046 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3047 struct dst_entry *dst;
0b26fb17
MŻ		 3048 struct flowi6 fl6 = {
3049 .flowi6_iif = LOOPBACK_IFINDEX,
3050 .flowi6_oif = oif,
0b26fb17
MŻ		 3051 .daddr = msg->dest,
3052 .saddr = iph->daddr,
3053 .flowi6_uid = sock_net_uid(net, NULL),
3054 };
c92a59ec 3055
b75cc8f9 3056 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
b55b76b2 3057 rt6_do_redirect(dst, NULL, skb);
c92a59ec
DJ		 3058 dst_release(dst);
3059}
3060
3a5ad2ee
DM		 3061void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3062{
e2d118a1
LC		 3063 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
3064 sk->sk_uid);
3a5ad2ee
DM		 3065}
3066EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3067
0dbaee3b 3068static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1da177e4 3069{
0dbaee3b
DM		 3070 struct net_device *dev = dst->dev;
3071 unsigned int mtu = dst_mtu(dst);
3072 struct net *net = dev_net(dev);
3073
1da177e4
LT		 3074 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3075
5578689a
DL		 3076 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3077 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1da177e4
LT		 3078
3079 /*
1ab1457c
YH		 3080 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3081 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3082 * IPV6_MAXPLEN is also valid and means: "any MSS,
1da177e4
LT		 3083 * rely only on pmtu discovery"
3084 */
3085 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3086 mtu = IPV6_MAXPLEN;
3087 return mtu;
3088}
3089
ebb762f2 3090static unsigned int ip6_mtu(const struct dst_entry *dst)
d33e4553 3091{
d33e4553 3092 struct inet6_dev *idev;
d4ead6b3 3093 unsigned int mtu;
4b32b5ad
MKL		 3094
3095 mtu = dst_metric_raw(dst, RTAX_MTU);
618f9bc7 3096 if (mtu)
30f78d8e 3097 goto out;
618f9bc7
SK		 3098
3099 mtu = IPV6_MIN_MTU;
d33e4553
DM		 3100
3101 rcu_read_lock();
3102 idev = __in6_dev_get(dst->dev);
3103 if (idev)
3104 mtu = idev->cnf.mtu6;
3105 rcu_read_unlock();
3106
30f78d8e 3107out:
14972cbd
RP		 3108 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3109
3110 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
d33e4553
DM		 3111}
3112
901731b8
DA		 3113/* MTU selection:
3114 * 1. mtu on route is locked - use it
3115 * 2. mtu from nexthop exception
3116 * 3. mtu from egress device
3117 *
3118 * based on ip6_dst_mtu_forward and exception logic of
3119 * rt6_find_cached_rt; called with rcu_read_lock
3120 */
b748f260
DA		 3121u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3122 const struct in6_addr *daddr,
3123 const struct in6_addr *saddr)
901731b8 3124{
b748f260
DA		 3125 const struct fib6_nh *nh = res->nh;
3126 struct fib6_info *f6i = res->f6i;
901731b8 3127 struct inet6_dev *idev;
510e2ced 3128 struct rt6_info *rt;
901731b8
DA		 3129 u32 mtu = 0;
3130
3131 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3132 mtu = f6i->fib6_pmtu;
3133 if (mtu)
3134 goto out;
3135 }
3136
510e2ced
WW		 3137 rt = rt6_find_cached_rt(res, daddr, saddr);
3138 if (unlikely(rt)) {
3139 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3140 } else {
b748f260 3141 struct net_device *dev = nh->fib_nh_dev;
901731b8
DA		 3142
3143 mtu = IPV6_MIN_MTU;
3144 idev = __in6_dev_get(dev);
3145 if (idev && idev->cnf.mtu6 > mtu)
3146 mtu = idev->cnf.mtu6;
3147 }
3148
3149 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3150out:
b748f260 3151 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
901731b8
DA		 3152}
3153
3b00944c 3154struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
87a11578 3155 struct flowi6 *fl6)
1da177e4 3156{
87a11578 3157 struct dst_entry *dst;
1da177e4
LT		 3158 struct rt6_info *rt;
3159 struct inet6_dev *idev = in6_dev_get(dev);
c346dca1 3160 struct net *net = dev_net(dev);
1da177e4 3161
38308473 3162 if (unlikely(!idev))
122bdf67 3163 return ERR_PTR(-ENODEV);
1da177e4 3164
ad706862 3165 rt = ip6_dst_alloc(net, dev, 0);
38308473 3166 if (unlikely(!rt)) {
1da177e4 3167 in6_dev_put(idev);
87a11578 3168 dst = ERR_PTR(-ENOMEM);
1da177e4
LT		 3169 goto out;
3170 }
3171
588753f1 3172 rt->dst.input = ip6_input;
8e2ec639 3173 rt->dst.output = ip6_output;
550bab42 3174 rt->rt6i_gateway = fl6->daddr;
87a11578 3175 rt->rt6i_dst.addr = fl6->daddr;
8e2ec639
YZ		 3176 rt->rt6i_dst.plen = 128;
3177 rt->rt6i_idev = idev;
14edd87d 3178 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1da177e4 3179
4c981e28 3180 /* Add this dst into uncached_list so that rt6_disable_ip() can
587fea74
WW		 3181 * do proper release of the net_device
3182 */
3183 rt6_uncached_list_add(rt);
81eb8447 3184 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
1da177e4 3185
87a11578
DM		 3186 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3187
1da177e4 3188out:
87a11578 3189 return dst;
1da177e4
LT		 3190}
3191
569d3645 3192static int ip6_dst_gc(struct dst_ops *ops)
1da177e4 3193{
86393e52 3194 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
7019b78e
DL		 3195 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3196 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3197 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3198 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3199 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
fc66f95c 3200 int entries;
7019b78e 3201
fc66f95c 3202 entries = dst_entries_get_fast(ops);
cf86a086
ED		 3203 if (entries > rt_max_size)
3204 entries = dst_entries_get_slow(ops);
3205
49a18d86 3206 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
fc66f95c 3207 entries <= rt_max_size)
1da177e4
LT		 3208 goto out;
3209
6891a346 3210 net->ipv6.ip6_rt_gc_expire++;
14956643 3211 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
fc66f95c
ED		 3212 entries = dst_entries_get_slow(ops);
3213 if (entries < ops->gc_thresh)
7019b78e 3214 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1da177e4 3215out:
7019b78e 3216 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
fc66f95c 3217 return entries > rt_max_size;
1da177e4
LT		 3218}
3219
b2c709cc
DA		 3220static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3221 const struct in6_addr *gw_addr, u32 tbid,
3222 int flags, struct fib6_result *res)
8c14586f
DA		 3223{
3224 struct flowi6 fl6 = {
3225 .flowi6_oif = cfg->fc_ifindex,
3226 .daddr = *gw_addr,
3227 .saddr = cfg->fc_prefsrc,
3228 };
3229 struct fib6_table *table;
b2c709cc 3230 int err;
8c14586f 3231
f4797b33 3232 table = fib6_get_table(net, tbid);
8c14586f 3233 if (!table)
b2c709cc 3234 return -EINVAL;
8c14586f
DA		 3235
3236 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3237 flags |= RT6_LOOKUP_F_HAS_SADDR;
3238
f4797b33 3239 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
8c14586f 3240
b2c709cc
DA		 3241 err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3242 if (!err && res->f6i != net->ipv6.fib6_null_entry)
3243 fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3244 cfg->fc_ifindex != 0, NULL, flags);
8c14586f 3245
b2c709cc 3246 return err;
8c14586f
DA		 3247}
3248
fc1e64e1
DA		 3249static int ip6_route_check_nh_onlink(struct net *net,
3250 struct fib6_config *cfg,
9fbb704c 3251 const struct net_device *dev,
fc1e64e1
DA		 3252 struct netlink_ext_ack *extack)
3253{
b2c709cc 3254 u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
fc1e64e1 3255 const struct in6_addr *gw_addr = &cfg->fc_gateway;
b2c709cc 3256 struct fib6_result res = {};
fc1e64e1
DA		 3257 int err;
3258
b2c709cc
DA		 3259 err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3260 if (!err && !(res.fib6_flags & RTF_REJECT) &&
3261 /* ignore match if it is the default route */
3262 !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3263 (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3264 NL_SET_ERR_MSG(extack,
3265 "Nexthop has invalid gateway or device mismatch");
3266 err = -EINVAL;
fc1e64e1
DA		 3267 }
3268
3269 return err;
3270}
3271
1edce99f
DA		 3272static int ip6_route_check_nh(struct net *net,
3273 struct fib6_config *cfg,
3274 struct net_device **_dev,
3275 struct inet6_dev **idev)
3276{
3277 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3278 struct net_device *dev = _dev ? *_dev : NULL;
b2c709cc
DA		 3279 int flags = RT6_LOOKUP_F_IFACE;
3280 struct fib6_result res = {};
1edce99f
DA		 3281 int err = -EHOSTUNREACH;
3282
3283 if (cfg->fc_table) {
b2c709cc
DA		 3284 err = ip6_nh_lookup_table(net, cfg, gw_addr,
3285 cfg->fc_table, flags, &res);
3286 /* gw_addr can not require a gateway or resolve to a reject
3287 * route. If a device is given, it must match the result.
3288 */
3289 if (err || res.fib6_flags & RTF_REJECT ||
3290 res.nh->fib_nh_gw_family ||
3291 (dev && dev != res.nh->fib_nh_dev))
3292 err = -EHOSTUNREACH;
1edce99f
DA		 3293 }
3294
b2c709cc
DA		 3295 if (err < 0) {
3296 struct flowi6 fl6 = {
3297 .flowi6_oif = cfg->fc_ifindex,
3298 .daddr = *gw_addr,
3299 };
1edce99f 3300
b2c709cc
DA		 3301 err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3302 if (err || res.fib6_flags & RTF_REJECT ||
3303 res.nh->fib_nh_gw_family)
3304 err = -EHOSTUNREACH;
3305
3306 if (err)
3307 return err;
3308
3309 fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3310 cfg->fc_ifindex != 0, NULL, flags);
3311 }
1edce99f 3312
b2c709cc 3313 err = 0;
1edce99f 3314 if (dev) {
b2c709cc
DA		 3315 if (dev != res.nh->fib_nh_dev)
3316 err = -EHOSTUNREACH;
1edce99f 3317 } else {
b2c709cc 3318 *_dev = dev = res.nh->fib_nh_dev;
1edce99f 3319 dev_hold(dev);
b2c709cc 3320 *idev = in6_dev_get(dev);
1edce99f
DA		 3321 }
3322
1edce99f
DA		 3323 return err;
3324}
3325
9fbb704c
DA		 3326static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3327 struct net_device **_dev, struct inet6_dev **idev,
3328 struct netlink_ext_ack *extack)
3329{
3330 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3331 int gwa_type = ipv6_addr_type(gw_addr);
232378e8 3332 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
9fbb704c 3333 const struct net_device *dev = *_dev;
232378e8 3334 bool need_addr_check = !dev;
9fbb704c
DA		 3335 int err = -EINVAL;
3336
3337 /* if gw_addr is local we will fail to detect this in case
3338 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3339 * will return already-added prefix route via interface that
3340 * prefix route was assigned to, which might be non-loopback.
3341 */
232378e8
DA		 3342 if (dev &&
3343 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3344 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
9fbb704c
DA		 3345 goto out;
3346 }
3347
3348 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3349 /* IPv6 strictly inhibits using not link-local
3350 * addresses as nexthop address.
3351 * Otherwise, router will not able to send redirects.
3352 * It is very good, but in some (rare!) circumstances
3353 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3354 * some exceptions. --ANK
3355 * We allow IPv4-mapped nexthops to support RFC4798-type
3356 * addressing
3357 */
3358 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3359 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3360 goto out;
3361 }
3362
b2c709cc
DA		 3363 rcu_read_lock();
3364
9fbb704c
DA		 3365 if (cfg->fc_flags & RTNH_F_ONLINK)
3366 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3367 else
3368 err = ip6_route_check_nh(net, cfg, _dev, idev);
3369
b2c709cc
DA		 3370 rcu_read_unlock();
3371
9fbb704c
DA		 3372 if (err)
3373 goto out;
3374 }
3375
3376 /* reload in case device was changed */
3377 dev = *_dev;
3378
3379 err = -EINVAL;
3380 if (!dev) {
3381 NL_SET_ERR_MSG(extack, "Egress device not specified");
3382 goto out;
3383 } else if (dev->flags & IFF_LOOPBACK) {
3384 NL_SET_ERR_MSG(extack,
3385 "Egress device can not be loopback device for this route");
3386 goto out;
3387 }
232378e8
DA		 3388
3389 /* if we did not check gw_addr above, do so now that the
3390 * egress device has been resolved.
3391 */
3392 if (need_addr_check &&
3393 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3394 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3395 goto out;
3396 }
3397
9fbb704c
DA		 3398 err = 0;
3399out:
3400 return err;
3401}
3402
83c44251
DA		 3403static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3404{
3405 if ((flags & RTF_REJECT) ||
3406 (dev && (dev->flags & IFF_LOOPBACK) &&
3407 !(addr_type & IPV6_ADDR_LOOPBACK) &&
aea23c32 3408 !(flags & (RTF_ANYCAST | RTF_LOCAL))))
83c44251
DA		 3409 return true;
3410
3411 return false;
3412}
3413
3414int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3415 struct fib6_config *cfg, gfp_t gfp_flags,
3416 struct netlink_ext_ack *extack)
3417{
3418 struct net_device *dev = NULL;
3419 struct inet6_dev *idev = NULL;
3420 int addr_type;
3421 int err;
3422
f1741730 3423 fib6_nh->fib_nh_family = AF_INET6;
1bef4c22
ED		 3424#ifdef CONFIG_IPV6_ROUTER_PREF
3425 fib6_nh->last_probe = jiffies;
3426#endif
38428d68
RP		 3427 if (cfg->fc_is_fdb) {
3428 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3429 fib6_nh->fib_nh_gw_family = AF_INET6;
3430 return 0;
3431 }
f1741730 3432
83c44251
DA		 3433 err = -ENODEV;
3434 if (cfg->fc_ifindex) {
3435 dev = dev_get_by_index(net, cfg->fc_ifindex);
3436 if (!dev)
3437 goto out;
3438 idev = in6_dev_get(dev);
3439 if (!idev)
3440 goto out;
3441 }
3442
3443 if (cfg->fc_flags & RTNH_F_ONLINK) {
3444 if (!dev) {
3445 NL_SET_ERR_MSG(extack,
3446 "Nexthop device required for onlink");
3447 goto out;
3448 }
3449
3450 if (!(dev->flags & IFF_UP)) {
3451 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3452 err = -ENETDOWN;
3453 goto out;
3454 }
3455
ad1601ae 3456 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
83c44251
DA		 3457 }
3458
ad1601ae 3459 fib6_nh->fib_nh_weight = 1;
83c44251
DA		 3460
3461 /* We cannot add true routes via loopback here,
3462 * they would result in kernel looping; promote them to reject routes
3463 */
3464 addr_type = ipv6_addr_type(&cfg->fc_dst);
3465 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3466 /* hold loopback dev/idev if we haven't done so. */
3467 if (dev != net->loopback_dev) {
3468 if (dev) {
3469 dev_put(dev);
3470 in6_dev_put(idev);
3471 }
3472 dev = net->loopback_dev;
3473 dev_hold(dev);
3474 idev = in6_dev_get(dev);
3475 if (!idev) {
3476 err = -ENODEV;
3477 goto out;
3478 }
3479 }
7dd73168 3480 goto pcpu_alloc;
83c44251
DA		 3481 }
3482
3483 if (cfg->fc_flags & RTF_GATEWAY) {
3484 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3485 if (err)
3486 goto out;
3487
ad1601ae 3488 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
bdf00467 3489 fib6_nh->fib_nh_gw_family = AF_INET6;
83c44251
DA		 3490 }
3491
3492 err = -ENODEV;
3493 if (!dev)
3494 goto out;
3495
3496 if (idev->cnf.disable_ipv6) {
3497 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3498 err = -EACCES;
3499 goto out;
3500 }
3501
3502 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3503 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3504 err = -ENETDOWN;
3505 goto out;
3506 }
3507
3508 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3509 !netif_carrier_ok(dev))
ad1601ae 3510 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
83c44251 3511
faee6769 3512 err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
7dd73168
DA		 3513 cfg->fc_encap_type, cfg, gfp_flags, extack);
3514 if (err)
3515 goto out;
3516
3517pcpu_alloc:
f40b6ae2
DA		 3518 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3519 if (!fib6_nh->rt6i_pcpu) {
3520 err = -ENOMEM;
3521 goto out;
3522 }
3523
ad1601ae 3524 fib6_nh->fib_nh_dev = dev;
f1741730 3525 fib6_nh->fib_nh_oif = dev->ifindex;
83c44251
DA		 3526 err = 0;
3527out:
3528 if (idev)
3529 in6_dev_put(idev);
3530
3531 if (err) {
ad1601ae
DA		 3532 lwtstate_put(fib6_nh->fib_nh_lws);
3533 fib6_nh->fib_nh_lws = NULL;
83c44251
DA		 3534 if (dev)
3535 dev_put(dev);
3536 }
3537
3538 return err;
3539}
3540
dac7d0f2
DA		 3541void fib6_nh_release(struct fib6_nh *fib6_nh)
3542{
cc5c073a
DA		 3543 struct rt6_exception_bucket *bucket;
3544
3545 rcu_read_lock();
3546
3547 fib6_nh_flush_exceptions(fib6_nh, NULL);
3548 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3549 if (bucket) {
3550 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3551 kfree(bucket);
3552 }
3553
3554 rcu_read_unlock();
3555
f40b6ae2
DA		 3556 if (fib6_nh->rt6i_pcpu) {
3557 int cpu;
3558
3559 for_each_possible_cpu(cpu) {
3560 struct rt6_info **ppcpu_rt;
3561 struct rt6_info *pcpu_rt;
3562
3563 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3564 pcpu_rt = *ppcpu_rt;
3565 if (pcpu_rt) {
3566 dst_dev_put(&pcpu_rt->dst);
3567 dst_release(&pcpu_rt->dst);
3568 *ppcpu_rt = NULL;
3569 }
3570 }
3571
3572 free_percpu(fib6_nh->rt6i_pcpu);
3573 }
3574
979e276e 3575 fib_nh_common_release(&fib6_nh->nh_common);
dac7d0f2
DA		 3576}
3577
8d1c802b 3578static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
acb54e3c 3579 gfp_t gfp_flags,
333c4301 3580 struct netlink_ext_ack *extack)
1da177e4 3581{
5578689a 3582 struct net *net = cfg->fc_nlinfo.nl_net;
8d1c802b 3583 struct fib6_info *rt = NULL;
f88d8ea6 3584 struct nexthop *nh = NULL;
c71099ac 3585 struct fib6_table *table;
f88d8ea6 3586 struct fib6_nh *fib6_nh;
8c5b83f0 3587 int err = -EINVAL;
83c44251 3588 int addr_type;
1da177e4 3589
557c44be 3590 /* RTF_PCPU is an internal flag; can not be set by userspace */
d5d531cb
DA		 3591 if (cfg->fc_flags & RTF_PCPU) {
3592 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
557c44be 3593 goto out;
d5d531cb 3594 }
557c44be 3595
2ea2352e
WW		 3596 /* RTF_CACHE is an internal flag; can not be set by userspace */
3597 if (cfg->fc_flags & RTF_CACHE) {
3598 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3599 goto out;
3600 }
3601
e8478e80
DA		 3602 if (cfg->fc_type > RTN_MAX) {
3603 NL_SET_ERR_MSG(extack, "Invalid route type");
3604 goto out;
3605 }
3606
d5d531cb
DA		 3607 if (cfg->fc_dst_len > 128) {
3608 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3609 goto out;
3610 }
3611 if (cfg->fc_src_len > 128) {
3612 NL_SET_ERR_MSG(extack, "Invalid source address length");
8c5b83f0 3613 goto out;
d5d531cb 3614 }
1da177e4 3615#ifndef CONFIG_IPV6_SUBTREES
d5d531cb
DA		 3616 if (cfg->fc_src_len) {
3617 NL_SET_ERR_MSG(extack,
3618 "Specifying source address requires IPV6_SUBTREES to be enabled");
8c5b83f0 3619 goto out;
d5d531cb 3620 }
1da177e4 3621#endif
5b98324e
DA		 3622 if (cfg->fc_nh_id) {
3623 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3624 if (!nh) {
3625 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3626 goto out;
3627 }
3628 err = fib6_check_nexthop(nh, cfg, extack);
3629 if (err)
3630 goto out;
3631 }
fc1e64e1 3632
d71314b4 3633 err = -ENOBUFS;
38308473
DM		 3634 if (cfg->fc_nlinfo.nlh &&
3635 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
d71314b4 3636 table = fib6_get_table(net, cfg->fc_table);
38308473 3637 if (!table) {
f3213831 3638 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
d71314b4
MV		 3639 table = fib6_new_table(net, cfg->fc_table);
3640 }
3641 } else {
3642 table = fib6_new_table(net, cfg->fc_table);
3643 }
38308473
DM		 3644
3645 if (!table)
c71099ac 3646 goto out;
c71099ac 3647
93531c67 3648 err = -ENOMEM;
f88d8ea6 3649 rt = fib6_info_alloc(gfp_flags, !nh);
93531c67 3650 if (!rt)
1da177e4 3651 goto out;
93531c67 3652
d7e774f3
DA		 3653 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3654 extack);
767a2217
DA		 3655 if (IS_ERR(rt->fib6_metrics)) {
3656 err = PTR_ERR(rt->fib6_metrics);
fda21d46
ED		 3657 /* Do not leave garbage there. */
3658 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
767a2217
DA		 3659 goto out;
3660 }
3661
93531c67
DA		 3662 if (cfg->fc_flags & RTF_ADDRCONF)
3663 rt->dst_nocount = true;
1da177e4 3664
1716a961 3665 if (cfg->fc_flags & RTF_EXPIRES)
14895687 3666 fib6_set_expires(rt, jiffies +
1716a961
G		 3667 clock_t_to_jiffies(cfg->fc_expires));
3668 else
14895687 3669 fib6_clean_expires(rt);
1da177e4 3670
86872cb5
TG		 3671 if (cfg->fc_protocol == RTPROT_UNSPEC)
3672 cfg->fc_protocol = RTPROT_BOOT;
93c2fb25 3673 rt->fib6_protocol = cfg->fc_protocol;
86872cb5 3674
83c44251
DA		 3675 rt->fib6_table = table;
3676 rt->fib6_metric = cfg->fc_metric;
c7036d97 3677 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
2b2450ca 3678 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
19e42e45 3679
93c2fb25
DA		 3680 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3681 rt->fib6_dst.plen = cfg->fc_dst_len;
e5fd387a 3682
1da177e4 3683#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 3684 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3685 rt->fib6_src.plen = cfg->fc_src_len;
1da177e4 3686#endif
f88d8ea6
DA		 3687 if (nh) {
3688 if (!nexthop_get(nh)) {
3689 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3690 goto out;
3691 }
3692 if (rt->fib6_src.plen) {
4daa95af 3693 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
f88d8ea6
DA		 3694 goto out;
3695 }
3696 rt->nh = nh;
3697 fib6_nh = nexthop_fib6_nh(rt->nh);
3698 } else {
3699 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3700 if (err)
3701 goto out;
1da177e4 3702
f88d8ea6
DA		 3703 fib6_nh = rt->fib6_nh;
3704
3705 /* We cannot add true routes via loopback here, they would
3706 * result in kernel looping; promote them to reject routes
3707 */
3708 addr_type = ipv6_addr_type(&cfg->fc_dst);
3709 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3710 addr_type))
3711 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3712 }
955ec4cb 3713
c3968a85 3714 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
f88d8ea6 3715 struct net_device *dev = fib6_nh->fib_nh_dev;
83c44251 3716
c3968a85 3717 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
d5d531cb 3718 NL_SET_ERR_MSG(extack, "Invalid source address");
c3968a85
DW		 3719 err = -EINVAL;
3720 goto out;
3721 }
93c2fb25
DA		 3722 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3723 rt->fib6_prefsrc.plen = 128;
c3968a85 3724 } else
93c2fb25 3725 rt->fib6_prefsrc.plen = 0;
c3968a85 3726
8c5b83f0 3727 return rt;
6b9ea5a6 3728out:
93531c67 3729 fib6_info_release(rt);
8c5b83f0 3730 return ERR_PTR(err);
6b9ea5a6
RP		 3731}
3732
acb54e3c 3733int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
333c4301 3734 struct netlink_ext_ack *extack)
6b9ea5a6 3735{
8d1c802b 3736 struct fib6_info *rt;
6b9ea5a6
RP		 3737 int err;
3738
acb54e3c 3739 rt = ip6_route_info_create(cfg, gfp_flags, extack);
d4ead6b3
DA		 3740 if (IS_ERR(rt))
3741 return PTR_ERR(rt);
6b9ea5a6 3742
d4ead6b3 3743 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
93531c67 3744 fib6_info_release(rt);
6b9ea5a6 3745
1da177e4
LT		 3746 return err;
3747}
3748
8d1c802b 3749static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
1da177e4 3750{
afb1d4b5 3751 struct net *net = info->nl_net;
c71099ac 3752 struct fib6_table *table;
afb1d4b5 3753 int err;
1da177e4 3754
421842ed 3755 if (rt == net->ipv6.fib6_null_entry) {
6825a26c
G		 3756 err = -ENOENT;
3757 goto out;
3758 }
6c813a72 3759
93c2fb25 3760 table = rt->fib6_table;
66f5d6ce 3761 spin_lock_bh(&table->tb6_lock);
86872cb5 3762 err = fib6_del(rt, info);
66f5d6ce 3763 spin_unlock_bh(&table->tb6_lock);
1da177e4 3764
6825a26c 3765out:
93531c67 3766 fib6_info_release(rt);
1da177e4
LT		 3767 return err;
3768}
3769
11dd74b3 3770int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
e0a1ad73 3771{
11dd74b3
RP		 3772 struct nl_info info = {
3773 .nl_net = net,
3774 .skip_notify = skip_notify
3775 };
afb1d4b5 3776
528c4ceb 3777 return __ip6_del_rt(rt, &info);
e0a1ad73
TG		 3778}
3779
8d1c802b 3780static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
0ae81335
DA		 3781{
3782 struct nl_info *info = &cfg->fc_nlinfo;
e3330039 3783 struct net *net = info->nl_net;
16a16cd3 3784 struct sk_buff *skb = NULL;
0ae81335 3785 struct fib6_table *table;
e3330039 3786 int err = -ENOENT;
0ae81335 3787
421842ed 3788 if (rt == net->ipv6.fib6_null_entry)
e3330039 3789 goto out_put;
93c2fb25 3790 table = rt->fib6_table;
66f5d6ce 3791 spin_lock_bh(&table->tb6_lock);
0ae81335 3792
93c2fb25 3793 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
8d1c802b 3794 struct fib6_info *sibling, *next_sibling;
0284696b 3795 struct fib6_node *fn;
0ae81335 3796
16a16cd3
DA		 3797 /* prefer to send a single notification with all hops */
3798 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3799 if (skb) {
3800 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3801
d4ead6b3 3802 if (rt6_fill_node(net, skb, rt, NULL,
16a16cd3
DA		 3803 NULL, NULL, 0, RTM_DELROUTE,
3804 info->portid, seq, 0) < 0) {
3805 kfree_skb(skb);
3806 skb = NULL;
3807 } else
3808 info->skip_notify = 1;
3809 }
3810
0284696b
IS		 3811 /* 'rt' points to the first sibling route. If it is not the
3812 * leaf, then we do not need to send a notification. Otherwise,
3813 * we need to check if the last sibling has a next route or not
3814 * and emit a replace or delete notification, respectively.
3815 */
2881fd61 3816 info->skip_notify_kernel = 1;
0284696b
IS		 3817 fn = rcu_dereference_protected(rt->fib6_node,
3818 lockdep_is_held(&table->tb6_lock));
3819 if (rcu_access_pointer(fn->leaf) == rt) {
3820 struct fib6_info *last_sibling, *replace_rt;
3821
3822 last_sibling = list_last_entry(&rt->fib6_siblings,
3823 struct fib6_info,
3824 fib6_siblings);
3825 replace_rt = rcu_dereference_protected(
3826 last_sibling->fib6_next,
3827 lockdep_is_held(&table->tb6_lock));
3828 if (replace_rt)
3829 call_fib6_entry_notifiers_replace(net,
3830 replace_rt);
3831 else
3832 call_fib6_multipath_entry_notifiers(net,
caafb250 3833 FIB_EVENT_ENTRY_DEL,
0284696b
IS		 3834 rt, rt->fib6_nsiblings,
3835 NULL);
3836 }
0ae81335 3837 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25
DA		 3838 &rt->fib6_siblings,
3839 fib6_siblings) {
0ae81335
DA		 3840 err = fib6_del(sibling, info);
3841 if (err)
e3330039 3842 goto out_unlock;
0ae81335
DA		 3843 }
3844 }
3845
3846 err = fib6_del(rt, info);
e3330039 3847out_unlock:
66f5d6ce 3848 spin_unlock_bh(&table->tb6_lock);
e3330039 3849out_put:
93531c67 3850 fib6_info_release(rt);
16a16cd3
DA		 3851
3852 if (skb) {
e3330039 3853 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
16a16cd3
DA		 3854 info->nlh, gfp_any());
3855 }
0ae81335
DA		 3856 return err;
3857}
3858
0fa6efc5 3859static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
23fb93a4
DA		 3860{
3861 int rc = -ESRCH;
3862
3863 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3864 goto out;
3865
3866 if (cfg->fc_flags & RTF_GATEWAY &&
3867 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3868 goto out;
761f6026
XL		 3869
3870 rc = rt6_remove_exception_rt(rt);
23fb93a4
DA		 3871out:
3872 return rc;
3873}
3874
0fa6efc5
DA		 3875static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3876 struct fib6_nh *nh)
3877{
3878 struct fib6_result res = {
3879 .f6i = rt,
3880 .nh = nh,
3881 };
3882 struct rt6_info *rt_cache;
3883
3884 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3885 if (rt_cache)
3886 return __ip6_del_cached_rt(rt_cache, cfg);
3887
3888 return 0;
3889}
3890
5b98324e
DA		 3891struct fib6_nh_del_cached_rt_arg {
3892 struct fib6_config *cfg;
3893 struct fib6_info *f6i;
3894};
3895
3896static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
3897{
3898 struct fib6_nh_del_cached_rt_arg *arg = _arg;
3899 int rc;
3900
3901 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
3902 return rc != -ESRCH ? rc : 0;
3903}
3904
3905static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
3906{
3907 struct fib6_nh_del_cached_rt_arg arg = {
3908 .cfg = cfg,
3909 .f6i = f6i
3910 };
3911
3912 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
3913}
3914
333c4301
DA		 3915static int ip6_route_del(struct fib6_config *cfg,
3916 struct netlink_ext_ack *extack)
1da177e4 3917{
c71099ac 3918 struct fib6_table *table;
8d1c802b 3919 struct fib6_info *rt;
1da177e4 3920 struct fib6_node *fn;
1da177e4
LT		 3921 int err = -ESRCH;
3922
5578689a 3923 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
d5d531cb
DA		 3924 if (!table) {
3925 NL_SET_ERR_MSG(extack, "FIB table does not exist");
c71099ac 3926 return err;
d5d531cb 3927 }
c71099ac 3928
66f5d6ce 3929 rcu_read_lock();
1da177e4 3930
c71099ac 3931 fn = fib6_locate(&table->tb6_root,
86872cb5 3932 &cfg->fc_dst, cfg->fc_dst_len,
38fbeeee 3933 &cfg->fc_src, cfg->fc_src_len,
2b760fcf 3934 !(cfg->fc_flags & RTF_CACHE));
1ab1457c 3935
1da177e4 3936 if (fn) {
66f5d6ce 3937 for_each_fib6_node_rt_rcu(fn) {
ad1601ae
DA		 3938 struct fib6_nh *nh;
3939
3401bfb1
SB		 3940 if (rt->nh && cfg->fc_nh_id &&
3941 rt->nh->id != cfg->fc_nh_id)
5b98324e 3942 continue;
23fb93a4 3943
5b98324e
DA		 3944 if (cfg->fc_flags & RTF_CACHE) {
3945 int rc = 0;
3946
3947 if (rt->nh) {
3948 rc = ip6_del_cached_rt_nh(cfg, rt);
3949 } else if (cfg->fc_nh_id) {
3950 continue;
3951 } else {
3952 nh = rt->fib6_nh;
3953 rc = ip6_del_cached_rt(cfg, rt, nh);
3954 }
0fa6efc5
DA		 3955 if (rc != -ESRCH) {
3956 rcu_read_unlock();
3957 return rc;
23fb93a4
DA		 3958 }
3959 continue;
2b760fcf 3960 }
ad1601ae 3961
5b98324e
DA		 3962 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
3963 continue;
3964 if (cfg->fc_protocol &&
3965 cfg->fc_protocol != rt->fib6_protocol)
3966 continue;
3967
3968 if (rt->nh) {
3969 if (!fib6_info_hold_safe(rt))
3970 continue;
3971 rcu_read_unlock();
3972
3973 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3974 }
3975 if (cfg->fc_nh_id)
3976 continue;
3977
3978 nh = rt->fib6_nh;
86872cb5 3979 if (cfg->fc_ifindex &&
ad1601ae
DA		 3980 (!nh->fib_nh_dev ||
3981 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
1da177e4 3982 continue;
86872cb5 3983 if (cfg->fc_flags & RTF_GATEWAY &&
ad1601ae 3984 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
1da177e4 3985 continue;
e873e4b9
WW		 3986 if (!fib6_info_hold_safe(rt))
3987 continue;
66f5d6ce 3988 rcu_read_unlock();
1da177e4 3989
0ae81335
DA		 3990 /* if gateway was specified only delete the one hop */
3991 if (cfg->fc_flags & RTF_GATEWAY)
3992 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3993
3994 return __ip6_del_rt_siblings(rt, cfg);
1da177e4
LT		 3995 }
3996 }
66f5d6ce 3997 rcu_read_unlock();
1da177e4
LT		 3998
3999 return err;
4000}
4001
6700c270 4002static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
a6279458 4003{
a6279458 4004 struct netevent_redirect netevent;
e8599ff4 4005 struct rt6_info *rt, *nrt = NULL;
85bd05de 4006 struct fib6_result res = {};
e8599ff4
DM		 4007 struct ndisc_options ndopts;
4008 struct inet6_dev *in6_dev;
4009 struct neighbour *neigh;
71bcdba0 4010 struct rd_msg *msg;
6e157b6a
DM		 4011 int optlen, on_link;
4012 u8 *lladdr;
e8599ff4 4013
29a3cad5 4014 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
71bcdba0 4015 optlen -= sizeof(*msg);
e8599ff4
DM		 4016
4017 if (optlen < 0) {
6e157b6a 4018 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
e8599ff4
DM		 4019 return;
4020 }
4021
71bcdba0 4022 msg = (struct rd_msg *)icmp6_hdr(skb);
e8599ff4 4023
71bcdba0 4024 if (ipv6_addr_is_multicast(&msg->dest)) {
6e157b6a 4025 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
e8599ff4
DM		 4026 return;
4027 }
4028
6e157b6a 4029 on_link = 0;
71bcdba0 4030 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
e8599ff4 4031 on_link = 1;
71bcdba0 4032 } else if (ipv6_addr_type(&msg->target) !=
e8599ff4 4033 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
6e157b6a 4034 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
e8599ff4
DM		 4035 return;
4036 }
4037
4038 in6_dev = __in6_dev_get(skb->dev);
4039 if (!in6_dev)
4040 return;
4041 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
4042 return;
4043
4044 /* RFC2461 8.1:
4045 * The IP source address of the Redirect MUST be the same as the current
4046 * first-hop router for the specified ICMP Destination Address.
4047 */
4048
f997c55c 4049 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
e8599ff4
DM		 4050 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4051 return;
4052 }
6e157b6a
DM		 4053
4054 lladdr = NULL;
e8599ff4
DM		 4055 if (ndopts.nd_opts_tgt_lladdr) {
4056 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4057 skb->dev);
4058 if (!lladdr) {
4059 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4060 return;
4061 }
4062 }
4063
6e157b6a 4064 rt = (struct rt6_info *) dst;
ec13ad1d 4065 if (rt->rt6i_flags & RTF_REJECT) {
6e157b6a 4066 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
e8599ff4 4067 return;
6e157b6a 4068 }
e8599ff4 4069
6e157b6a
DM		 4070 /* Redirect received -> path was valid.
4071 * Look, redirects are sent only in response to data packets,
4072 * so that this nexthop apparently is reachable. --ANK
4073 */
0dec879f 4074 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
a6279458 4075
71bcdba0 4076 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
6e157b6a
DM		 4077 if (!neigh)
4078 return;
a6279458 4079
1da177e4
LT		 4080 /*
4081 * We have finally decided to accept it.
4082 */
4083
f997c55c 4084 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1da177e4
LT		 4085 NEIGH_UPDATE_F_WEAK_OVERRIDE|
4086 NEIGH_UPDATE_F_OVERRIDE|
4087 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
f997c55c
AA		 4088 NEIGH_UPDATE_F_ISROUTER)),
4089 NDISC_REDIRECT, &ndopts);
1da177e4 4090
4d85cd0c 4091 rcu_read_lock();
85bd05de 4092 res.f6i = rcu_dereference(rt->from);
ff24e498 4093 if (!res.f6i)
886b7a50 4094 goto out;
8a14e46f 4095
49d5b8ef
DA		 4096 if (res.f6i->nh) {
4097 struct fib6_nh_match_arg arg = {
4098 .dev = dst->dev,
4099 .gw = &rt->rt6i_gateway,
4100 };
4101
4102 nexthop_for_each_fib6_nh(res.f6i->nh,
4103 fib6_nh_find_match, &arg);
4104
4105 /* fib6_info uses a nexthop that does not have fib6_nh
4106 * using the dst->dev. Should be impossible
4107 */
4108 if (!arg.match)
4109 goto out;
4110 res.nh = arg.match;
4111 } else {
4112 res.nh = res.f6i->fib6_nh;
4113 }
4114
7d21fec9
DA		 4115 res.fib6_flags = res.f6i->fib6_flags;
4116 res.fib6_type = res.f6i->fib6_type;
85bd05de 4117 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
38308473 4118 if (!nrt)
1da177e4
LT		 4119 goto out;
4120
4121 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4122 if (on_link)
4123 nrt->rt6i_flags &= ~RTF_GATEWAY;
4124
4e3fd7a0 4125 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1da177e4 4126
886b7a50 4127 /* rt6_insert_exception() will take care of duplicated exceptions */
5012f0a5 4128 if (rt6_insert_exception(nrt, &res)) {
2b760fcf
WW		 4129 dst_release_immediate(&nrt->dst);
4130 goto out;
4131 }
1da177e4 4132
d8d1f30b
CG		 4133 netevent.old = &rt->dst;
4134 netevent.new = &nrt->dst;
71bcdba0 4135 netevent.daddr = &msg->dest;
60592833 4136 netevent.neigh = neigh;
8d71740c
TT		 4137 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4138
1da177e4 4139out:
886b7a50 4140 rcu_read_unlock();
e8599ff4 4141 neigh_release(neigh);
6e157b6a
DM		 4142}
4143
70ceb4f5 4144#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 4145static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 4146 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 4147 const struct in6_addr *gwaddr,
4148 struct net_device *dev)
70ceb4f5 4149{
830218c1
DA		 4150 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4151 int ifindex = dev->ifindex;
70ceb4f5 4152 struct fib6_node *fn;
8d1c802b 4153 struct fib6_info *rt = NULL;
c71099ac
TG		 4154 struct fib6_table *table;
4155
830218c1 4156 table = fib6_get_table(net, tb_id);
38308473 4157 if (!table)
c71099ac 4158 return NULL;
70ceb4f5 4159
66f5d6ce 4160 rcu_read_lock();
38fbeeee 4161 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
70ceb4f5
YH		 4162 if (!fn)
4163 goto out;
4164
66f5d6ce 4165 for_each_fib6_node_rt_rcu(fn) {
f88d8ea6
DA		 4166 /* these routes do not use nexthops */
4167 if (rt->nh)
4168 continue;
1cf844c7 4169 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
70ceb4f5 4170 continue;
2b2450ca 4171 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
1cf844c7 4172 !rt->fib6_nh->fib_nh_gw_family)
70ceb4f5 4173 continue;
1cf844c7 4174 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
70ceb4f5 4175 continue;
e873e4b9
WW		 4176 if (!fib6_info_hold_safe(rt))
4177 continue;
70ceb4f5
YH		 4178 break;
4179 }
4180out:
66f5d6ce 4181 rcu_read_unlock();
70ceb4f5
YH		 4182 return rt;
4183}
4184
8d1c802b 4185static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 4186 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 4187 const struct in6_addr *gwaddr,
4188 struct net_device *dev,
95c96174 4189 unsigned int pref)
70ceb4f5 4190{
86872cb5 4191 struct fib6_config cfg = {
238fc7ea 4192 .fc_metric = IP6_RT_PRIO_USER,
830218c1 4193 .fc_ifindex = dev->ifindex,
86872cb5
TG		 4194 .fc_dst_len = prefixlen,
4195 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4196 RTF_UP | RTF_PREF(pref),
b91d5329 4197 .fc_protocol = RTPROT_RA,
e8478e80 4198 .fc_type = RTN_UNICAST,
15e47304 4199 .fc_nlinfo.portid = 0,
efa2cea0
DL		 4200 .fc_nlinfo.nlh = NULL,
4201 .fc_nlinfo.nl_net = net,
86872cb5
TG		 4202 };
4203
830218c1 4204 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
4e3fd7a0
AD		 4205 cfg.fc_dst = *prefix;
4206 cfg.fc_gateway = *gwaddr;
70ceb4f5 4207
e317da96
YH		 4208 /* We should treat it as a default route if prefix length is 0. */
4209 if (!prefixlen)
86872cb5 4210 cfg.fc_flags |= RTF_DEFAULT;
70ceb4f5 4211
acb54e3c 4212 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
70ceb4f5 4213
830218c1 4214 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
70ceb4f5
YH		 4215}
4216#endif
4217
8d1c802b 4218struct fib6_info *rt6_get_dflt_router(struct net *net,
afb1d4b5
DA		 4219 const struct in6_addr *addr,
4220 struct net_device *dev)
1ab1457c 4221{
830218c1 4222 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
8d1c802b 4223 struct fib6_info *rt;
c71099ac 4224 struct fib6_table *table;
1da177e4 4225
afb1d4b5 4226 table = fib6_get_table(net, tb_id);
38308473 4227 if (!table)
c71099ac 4228 return NULL;
1da177e4 4229
66f5d6ce
WW		 4230 rcu_read_lock();
4231 for_each_fib6_node_rt_rcu(&table->tb6_root) {
f88d8ea6 4232 struct fib6_nh *nh;
ad1601ae 4233
f88d8ea6
DA		 4234 /* RA routes do not use nexthops */
4235 if (rt->nh)
4236 continue;
4237
4238 nh = rt->fib6_nh;
ad1601ae 4239 if (dev == nh->fib_nh_dev &&
93c2fb25 4240 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
ad1601ae 4241 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
1da177e4
LT		 4242 break;
4243 }
e873e4b9
WW		 4244 if (rt && !fib6_info_hold_safe(rt))
4245 rt = NULL;
66f5d6ce 4246 rcu_read_unlock();
1da177e4
LT		 4247 return rt;
4248}
4249
8d1c802b 4250struct fib6_info *rt6_add_dflt_router(struct net *net,
afb1d4b5 4251 const struct in6_addr *gwaddr,
ebacaaa0
YH		 4252 struct net_device *dev,
4253 unsigned int pref)
1da177e4 4254{
86872cb5 4255 struct fib6_config cfg = {
ca254490 4256 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
238fc7ea 4257 .fc_metric = IP6_RT_PRIO_USER,
86872cb5
TG		 4258 .fc_ifindex = dev->ifindex,
4259 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4260 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
b91d5329 4261 .fc_protocol = RTPROT_RA,
e8478e80 4262 .fc_type = RTN_UNICAST,
15e47304 4263 .fc_nlinfo.portid = 0,
5578689a 4264 .fc_nlinfo.nlh = NULL,
afb1d4b5 4265 .fc_nlinfo.nl_net = net,
86872cb5 4266 };
1da177e4 4267
4e3fd7a0 4268 cfg.fc_gateway = *gwaddr;
1da177e4 4269
acb54e3c 4270 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
830218c1
DA		 4271 struct fib6_table *table;
4272
4273 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4274 if (table)
4275 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4276 }
1da177e4 4277
afb1d4b5 4278 return rt6_get_dflt_router(net, gwaddr, dev);
1da177e4
LT		 4279}
4280
afb1d4b5
DA		 4281static void __rt6_purge_dflt_routers(struct net *net,
4282 struct fib6_table *table)
1da177e4 4283{
8d1c802b 4284 struct fib6_info *rt;
1da177e4
LT		 4285
4286restart:
66f5d6ce
WW		 4287 rcu_read_lock();
4288 for_each_fib6_node_rt_rcu(&table->tb6_root) {
dcd1f572
DA		 4289 struct net_device *dev = fib6_info_nh_dev(rt);
4290 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4291
93c2fb25 4292 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
e873e4b9
WW		 4293 (!idev || idev->cnf.accept_ra != 2) &&
4294 fib6_info_hold_safe(rt)) {
93531c67 4295 rcu_read_unlock();
11dd74b3 4296 ip6_del_rt(net, rt, false);
1da177e4
LT		 4297 goto restart;
4298 }
4299 }
66f5d6ce 4300 rcu_read_unlock();
830218c1
DA		 4301
4302 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4303}
4304
4305void rt6_purge_dflt_routers(struct net *net)
4306{
4307 struct fib6_table *table;
4308 struct hlist_head *head;
4309 unsigned int h;
4310
4311 rcu_read_lock();
4312
4313 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4314 head = &net->ipv6.fib_table_hash[h];
4315 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4316 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
afb1d4b5 4317 __rt6_purge_dflt_routers(net, table);
830218c1
DA		 4318 }
4319 }
4320
4321 rcu_read_unlock();
1da177e4
LT		 4322}
4323
5578689a
DL		 4324static void rtmsg_to_fib6_config(struct net *net,
4325 struct in6_rtmsg *rtmsg,
86872cb5
TG		 4326 struct fib6_config *cfg)
4327{
8823a3ac
MŻ		 4328 *cfg = (struct fib6_config){
4329 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4330 : RT6_TABLE_MAIN,
4331 .fc_ifindex = rtmsg->rtmsg_ifindex,
67f69513 4332 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
8823a3ac
MŻ		 4333 .fc_expires = rtmsg->rtmsg_info,
4334 .fc_dst_len = rtmsg->rtmsg_dst_len,
4335 .fc_src_len = rtmsg->rtmsg_src_len,
4336 .fc_flags = rtmsg->rtmsg_flags,
4337 .fc_type = rtmsg->rtmsg_type,
4338
4339 .fc_nlinfo.nl_net = net,
4340
4341 .fc_dst = rtmsg->rtmsg_dst,
4342 .fc_src = rtmsg->rtmsg_src,
4343 .fc_gateway = rtmsg->rtmsg_gateway,
4344 };
86872cb5
TG		 4345}
4346
7c1552da 4347int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
1da177e4 4348{
86872cb5 4349 struct fib6_config cfg;
1da177e4
LT		 4350 int err;
4351
7c1552da
CH		 4352 if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4353 return -EINVAL;
4354 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4355 return -EPERM;
86872cb5 4356
7c1552da 4357 rtmsg_to_fib6_config(net, rtmsg, &cfg);
1da177e4 4358
7c1552da
CH		 4359 rtnl_lock();
4360 switch (cmd) {
4361 case SIOCADDRT:
4362 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4363 break;
4364 case SIOCDELRT:
4365 err = ip6_route_del(&cfg, NULL);
4366 break;
3ff50b79 4367 }
7c1552da
CH		 4368 rtnl_unlock();
4369 return err;
1da177e4
LT		 4370}
4371
4372/*
4373 * Drop the packet on the floor
4374 */
4375
d5fdd6ba 4376static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1da177e4 4377{
adf30907 4378 struct dst_entry *dst = skb_dst(skb);
1d3fd8a1
SS		 4379 struct net *net = dev_net(dst->dev);
4380 struct inet6_dev *idev;
4381 int type;
4382
4383 if (netif_is_l3_master(skb->dev) &&
4384 dst->dev == net->loopback_dev)
4385 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4386 else
4387 idev = ip6_dst_idev(dst);
4388
612f09e8
YH		 4389 switch (ipstats_mib_noroutes) {
4390 case IPSTATS_MIB_INNOROUTES:
0660e03f 4391 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
45bb0060 4392 if (type == IPV6_ADDR_ANY) {
1d3fd8a1 4393 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
612f09e8
YH		 4394 break;
4395 }
a8eceea8 4396 fallthrough;
612f09e8 4397 case IPSTATS_MIB_OUTNOROUTES:
1d3fd8a1 4398 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
612f09e8
YH		 4399 break;
4400 }
1d3fd8a1
SS		 4401
4402 /* Start over by dropping the dst for l3mdev case */
4403 if (netif_is_l3_master(skb->dev))
4404 skb_dst_drop(skb);
4405
3ffe533c 4406 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1da177e4
LT		 4407 kfree_skb(skb);
4408 return 0;
4409}
4410
9ce8ade0
TG		 4411static int ip6_pkt_discard(struct sk_buff *skb)
4412{
612f09e8 4413 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4414}
4415
ede2059d 4416static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
1da177e4 4417{
adf30907 4418 skb->dev = skb_dst(skb)->dev;
612f09e8 4419 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1da177e4
LT		 4420}
4421
9ce8ade0
TG		 4422static int ip6_pkt_prohibit(struct sk_buff *skb)
4423{
612f09e8 4424 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4425}
4426
ede2059d 4427static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
9ce8ade0 4428{
adf30907 4429 skb->dev = skb_dst(skb)->dev;
612f09e8 4430 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
9ce8ade0
TG		 4431}
4432
1da177e4
LT		 4433/*
4434 * Allocate a dst for local (unicast / anycast) address.
4435 */
4436
360a9887
DA		 4437struct fib6_info *addrconf_f6i_alloc(struct net *net,
4438 struct inet6_dev *idev,
4439 const struct in6_addr *addr,
4440 bool anycast, gfp_t gfp_flags)
1da177e4 4441{
c7a1ce39
DA		 4442 struct fib6_config cfg = {
4443 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4444 .fc_ifindex = idev->dev->ifindex,
d55a2e37 4445 .fc_flags = RTF_UP | RTF_NONEXTHOP,
c7a1ce39
DA		 4446 .fc_dst = *addr,
4447 .fc_dst_len = 128,
4448 .fc_protocol = RTPROT_KERNEL,
4449 .fc_nlinfo.nl_net = net,
4450 .fc_ignore_dev_down = true,
4451 };
d55a2e37 4452 struct fib6_info *f6i;
1da177e4 4453
e8478e80 4454 if (anycast) {
c7a1ce39
DA		 4455 cfg.fc_type = RTN_ANYCAST;
4456 cfg.fc_flags |= RTF_ANYCAST;
e8478e80 4457 } else {
c7a1ce39
DA		 4458 cfg.fc_type = RTN_LOCAL;
4459 cfg.fc_flags |= RTF_LOCAL;
e8478e80 4460 }
1da177e4 4461
d55a2e37 4462 f6i = ip6_route_info_create(&cfg, gfp_flags, NULL);
8652f17c 4463 if (!IS_ERR(f6i))
d55a2e37
MŻ		 4464 f6i->dst_nocount = true;
4465 return f6i;
1da177e4
LT		 4466}
4467
c3968a85
DW		 4468/* remove deleted ip from prefsrc entries */
4469struct arg_dev_net_ip {
4470 struct net_device *dev;
4471 struct net *net;
4472 struct in6_addr *addr;
4473};
4474
8d1c802b 4475static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
c3968a85
DW		 4476{
4477 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4478 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4479 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4480
f88d8ea6
DA		 4481 if (!rt->nh &&
4482 ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
421842ed 4483 rt != net->ipv6.fib6_null_entry &&
93c2fb25 4484 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
60006a48 4485 spin_lock_bh(&rt6_exception_lock);
c3968a85 4486 /* remove prefsrc entry */
93c2fb25 4487 rt->fib6_prefsrc.plen = 0;
60006a48 4488 spin_unlock_bh(&rt6_exception_lock);
c3968a85
DW		 4489 }
4490 return 0;
4491}
4492
4493void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4494{
4495 struct net *net = dev_net(ifp->idev->dev);
4496 struct arg_dev_net_ip adni = {
4497 .dev = ifp->idev->dev,
4498 .net = net,
4499 .addr = &ifp->addr,
4500 };
0c3584d5 4501 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
c3968a85
DW		 4502}
4503
2b2450ca 4504#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
be7a010d
DJ		 4505
4506/* Remove routers and update dst entries when gateway turn into host. */
8d1c802b 4507static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
be7a010d
DJ		 4508{
4509 struct in6_addr *gateway = (struct in6_addr *)arg;
f88d8ea6
DA		 4510 struct fib6_nh *nh;
4511
4512 /* RA routes do not use nexthops */
4513 if (rt->nh)
4514 return 0;
be7a010d 4515
f88d8ea6 4516 nh = rt->fib6_nh;
93c2fb25 4517 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
cc5c073a 4518 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
be7a010d 4519 return -1;
b16cb459
WW		 4520
4521 /* Further clean up cached routes in exception table.
4522 * This is needed because cached route may have a different
4523 * gateway than its 'parent' in the case of an ip redirect.
4524 */
cc5c073a 4525 fib6_nh_exceptions_clean_tohost(nh, gateway);
b16cb459 4526
be7a010d
DJ		 4527 return 0;
4528}
4529
4530void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4531{
4532 fib6_clean_all(net, fib6_clean_tohost, gateway);
4533}
4534
2127d95a
IS		 4535struct arg_netdev_event {
4536 const struct net_device *dev;
4c981e28 4537 union {
ecc5663c 4538 unsigned char nh_flags;
4c981e28
IS		 4539 unsigned long event;
4540 };
2127d95a
IS		 4541};
4542
8d1c802b 4543static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
d7dedee1 4544{
8d1c802b 4545 struct fib6_info *iter;
d7dedee1
IS		 4546 struct fib6_node *fn;
4547
93c2fb25
DA		 4548 fn = rcu_dereference_protected(rt->fib6_node,
4549 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4550 iter = rcu_dereference_protected(fn->leaf,
93c2fb25 4551 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4552 while (iter) {
93c2fb25 4553 if (iter->fib6_metric == rt->fib6_metric &&
33bd5ac5 4554 rt6_qualify_for_ecmp(iter))
d7dedee1 4555 return iter;
8fb11a9a 4556 iter = rcu_dereference_protected(iter->fib6_next,
93c2fb25 4557 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1
IS		 4558 }
4559
4560 return NULL;
4561}
4562
f88d8ea6 4563/* only called for fib entries with builtin fib6_nh */
8d1c802b 4564static bool rt6_is_dead(const struct fib6_info *rt)
d7dedee1 4565{
1cf844c7
DA		 4566 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4567 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4568 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
d7dedee1
IS		 4569 return true;
4570
4571 return false;
4572}
4573
8d1c802b 4574static int rt6_multipath_total_weight(const struct fib6_info *rt)
d7dedee1 4575{
8d1c802b 4576 struct fib6_info *iter;
d7dedee1
IS		 4577 int total = 0;
4578
4579 if (!rt6_is_dead(rt))
1cf844c7 4580 total += rt->fib6_nh->fib_nh_weight;
d7dedee1 4581
93c2fb25 4582 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
d7dedee1 4583 if (!rt6_is_dead(iter))
1cf844c7 4584 total += iter->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4585 }
4586
4587 return total;
4588}
4589
8d1c802b 4590static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
d7dedee1
IS		 4591{
4592 int upper_bound = -1;
4593
4594 if (!rt6_is_dead(rt)) {
1cf844c7 4595 *weight += rt->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4596 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4597 total) - 1;
4598 }
1cf844c7 4599 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
d7dedee1
IS		 4600}
4601
8d1c802b 4602static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
d7dedee1 4603{
8d1c802b 4604 struct fib6_info *iter;
d7dedee1
IS		 4605 int weight = 0;
4606
4607 rt6_upper_bound_set(rt, &weight, total);
4608
93c2fb25 4609 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
d7dedee1
IS		 4610 rt6_upper_bound_set(iter, &weight, total);
4611}
4612
8d1c802b 4613void rt6_multipath_rebalance(struct fib6_info *rt)
d7dedee1 4614{
8d1c802b 4615 struct fib6_info *first;
d7dedee1
IS		 4616 int total;
4617
4618 /* In case the entire multipath route was marked for flushing,
4619 * then there is no need to rebalance upon the removal of every
4620 * sibling route.
4621 */
93c2fb25 4622 if (!rt->fib6_nsiblings || rt->should_flush)
d7dedee1
IS		 4623 return;
4624
4625 /* During lookup routes are evaluated in order, so we need to
4626 * make sure upper bounds are assigned from the first sibling
4627 * onwards.
4628 */
4629 first = rt6_multipath_first_sibling(rt);
4630 if (WARN_ON_ONCE(!first))
4631 return;
4632
4633 total = rt6_multipath_total_weight(first);
4634 rt6_multipath_upper_bound_set(first, total);
4635}
4636
8d1c802b 4637static int fib6_ifup(struct fib6_info *rt, void *p_arg)
2127d95a
IS		 4638{
4639 const struct arg_netdev_event *arg = p_arg;
7aef6859 4640 struct net *net = dev_net(arg->dev);
2127d95a 4641
f88d8ea6 4642 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
1cf844c7
DA		 4643 rt->fib6_nh->fib_nh_dev == arg->dev) {
4644 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
7aef6859 4645 fib6_update_sernum_upto_root(net, rt);
d7dedee1 4646 rt6_multipath_rebalance(rt);
1de178ed 4647 }
2127d95a
IS		 4648
4649 return 0;
4650}
4651
ecc5663c 4652void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
2127d95a
IS		 4653{
4654 struct arg_netdev_event arg = {
4655 .dev = dev,
6802f3ad
IS		 4656 {
4657 .nh_flags = nh_flags,
4658 },
2127d95a
IS		 4659 };
4660
4661 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4662 arg.nh_flags |= RTNH_F_LINKDOWN;
4663
4664 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4665}
4666
f88d8ea6 4667/* only called for fib entries with inline fib6_nh */
8d1c802b 4668static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
1de178ed
IS		 4669 const struct net_device *dev)
4670{
8d1c802b 4671 struct fib6_info *iter;
1de178ed 4672
1cf844c7 4673 if (rt->fib6_nh->fib_nh_dev == dev)
1de178ed 4674 return true;
93c2fb25 4675 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7 4676 if (iter->fib6_nh->fib_nh_dev == dev)
1de178ed
IS		 4677 return true;
4678
4679 return false;
4680}
4681
8d1c802b 4682static void rt6_multipath_flush(struct fib6_info *rt)
1de178ed 4683{
8d1c802b 4684 struct fib6_info *iter;
1de178ed
IS		 4685
4686 rt->should_flush = 1;
93c2fb25 4687 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1de178ed
IS		 4688 iter->should_flush = 1;
4689}
4690
8d1c802b 4691static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
1de178ed
IS		 4692 const struct net_device *down_dev)
4693{
8d1c802b 4694 struct fib6_info *iter;
1de178ed
IS		 4695 unsigned int dead = 0;
4696
1cf844c7
DA		 4697 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4698 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed 4699 dead++;
93c2fb25 4700 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4701 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4702 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed
IS		 4703 dead++;
4704
4705 return dead;
4706}
4707
8d1c802b 4708static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
1de178ed 4709 const struct net_device *dev,
ecc5663c 4710 unsigned char nh_flags)
1de178ed 4711{
8d1c802b 4712 struct fib6_info *iter;
1de178ed 4713
1cf844c7
DA		 4714 if (rt->fib6_nh->fib_nh_dev == dev)
4715 rt->fib6_nh->fib_nh_flags |= nh_flags;
93c2fb25 4716 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4717 if (iter->fib6_nh->fib_nh_dev == dev)
4718 iter->fib6_nh->fib_nh_flags |= nh_flags;
1de178ed
IS		 4719}
4720
a1a22c12 4721/* called with write lock held for table with rt */
8d1c802b 4722static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
1da177e4 4723{
4c981e28
IS		 4724 const struct arg_netdev_event *arg = p_arg;
4725 const struct net_device *dev = arg->dev;
7aef6859 4726 struct net *net = dev_net(dev);
8ed67789 4727
f88d8ea6 4728 if (rt == net->ipv6.fib6_null_entry || rt->nh)
27c6fa73
IS		 4729 return 0;
4730
4731 switch (arg->event) {
4732 case NETDEV_UNREGISTER:
1cf844c7 4733 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
27c6fa73 4734 case NETDEV_DOWN:
1de178ed 4735 if (rt->should_flush)
27c6fa73 4736 return -1;
93c2fb25 4737 if (!rt->fib6_nsiblings)
1cf844c7 4738 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
1de178ed
IS		 4739 if (rt6_multipath_uses_dev(rt, dev)) {
4740 unsigned int count;
4741
4742 count = rt6_multipath_dead_count(rt, dev);
93c2fb25 4743 if (rt->fib6_nsiblings + 1 == count) {
1de178ed
IS		 4744 rt6_multipath_flush(rt);
4745 return -1;
4746 }
4747 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4748 RTNH_F_LINKDOWN);
7aef6859 4749 fib6_update_sernum(net, rt);
d7dedee1 4750 rt6_multipath_rebalance(rt);
1de178ed
IS		 4751 }
4752 return -2;
27c6fa73 4753 case NETDEV_CHANGE:
1cf844c7 4754 if (rt->fib6_nh->fib_nh_dev != dev ||
93c2fb25 4755 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
27c6fa73 4756 break;
1cf844c7 4757 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
d7dedee1 4758 rt6_multipath_rebalance(rt);
27c6fa73 4759 break;
2b241361 4760 }
c159d30c 4761
1da177e4
LT		 4762 return 0;
4763}
4764
27c6fa73 4765void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
1da177e4 4766{
4c981e28 4767 struct arg_netdev_event arg = {
8ed67789 4768 .dev = dev,
6802f3ad
IS		 4769 {
4770 .event = event,
4771 },
8ed67789 4772 };
7c6bb7d2 4773 struct net *net = dev_net(dev);
8ed67789 4774
7c6bb7d2
DA		 4775 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4776 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4777 else
4778 fib6_clean_all(net, fib6_ifdown, &arg);
4c981e28
IS		 4779}
4780
4781void rt6_disable_ip(struct net_device *dev, unsigned long event)
4782{
4783 rt6_sync_down_dev(dev, event);
4784 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4785 neigh_ifdown(&nd_tbl, dev);
1da177e4
LT		 4786}
4787
95c96174 4788struct rt6_mtu_change_arg {
1da177e4 4789 struct net_device *dev;
95c96174 4790 unsigned int mtu;
c0b220cf 4791 struct fib6_info *f6i;
1da177e4
LT		 4792};
4793
cc5c073a 4794static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
c0b220cf
DA		 4795{
4796 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
cc5c073a 4797 struct fib6_info *f6i = arg->f6i;
c0b220cf
DA		 4798
4799 /* For administrative MTU increase, there is no way to discover
4800 * IPv6 PMTU increase, so PMTU increase should be updated here.
4801 * Since RFC 1981 doesn't include administrative MTU increase
4802 * update PMTU increase is a MUST. (i.e. jumbo frame)
4803 */
4804 if (nh->fib_nh_dev == arg->dev) {
4805 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4806 u32 mtu = f6i->fib6_pmtu;
4807
4808 if (mtu >= arg->mtu ||
4809 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4810 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4811
4812 spin_lock_bh(&rt6_exception_lock);
cc5c073a 4813 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
c0b220cf
DA		 4814 spin_unlock_bh(&rt6_exception_lock);
4815 }
4816
4817 return 0;
4818}
4819
4820static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
1da177e4
LT		 4821{
4822 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4823 struct inet6_dev *idev;
4824
4825 /* In IPv6 pmtu discovery is not optional,
4826 so that RTAX_MTU lock cannot disable it.
4827 We still use this lock to block changes
4828 caused by addrconf/ndisc.
4829 */
4830
4831 idev = __in6_dev_get(arg->dev);
38308473 4832 if (!idev)
1da177e4
LT		 4833 return 0;
4834
c0b220cf
DA		 4835 if (fib6_metric_locked(f6i, RTAX_MTU))
4836 return 0;
d4ead6b3 4837
c0b220cf 4838 arg->f6i = f6i;
2d44234b
DA		 4839 if (f6i->nh) {
4840 /* fib6_nh_mtu_change only returns 0, so this is safe */
4841 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4842 arg);
4843 }
4844
1cf844c7 4845 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
1da177e4
LT		 4846}
4847
95c96174 4848void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
1da177e4 4849{
c71099ac
TG		 4850 struct rt6_mtu_change_arg arg = {
4851 .dev = dev,
4852 .mtu = mtu,
4853 };
1da177e4 4854
0c3584d5 4855 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
1da177e4
LT		 4856}
4857
ef7c79ed 4858static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
75425657 4859 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
5176f91e 4860 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
aa8f8778 4861 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
86872cb5 4862 [RTA_OIF] = { .type = NLA_U32 },
ab364a6f 4863 [RTA_IIF] = { .type = NLA_U32 },
86872cb5
TG		 4864 [RTA_PRIORITY] = { .type = NLA_U32 },
4865 [RTA_METRICS] = { .type = NLA_NESTED },
51ebd318 4866 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
c78ba6d6 4867 [RTA_PREF] = { .type = NLA_U8 },
19e42e45
RP		 4868 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4869 [RTA_ENCAP] = { .type = NLA_NESTED },
32bc201e 4870 [RTA_EXPIRES] = { .type = NLA_U32 },
622ec2c9 4871 [RTA_UID] = { .type = NLA_U32 },
3b45a410 4872 [RTA_MARK] = { .type = NLA_U32 },
aa8f8778 4873 [RTA_TABLE] = { .type = NLA_U32 },
eacb9384
RP		 4874 [RTA_IP_PROTO] = { .type = NLA_U8 },
4875 [RTA_SPORT] = { .type = NLA_U16 },
4876 [RTA_DPORT] = { .type = NLA_U16 },
5b98324e 4877 [RTA_NH_ID] = { .type = NLA_U32 },
86872cb5
TG		 4878};
4879
4880static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
333c4301
DA		 4881 struct fib6_config *cfg,
4882 struct netlink_ext_ack *extack)
1da177e4 4883{
86872cb5
TG		 4884 struct rtmsg *rtm;
4885 struct nlattr *tb[RTA_MAX+1];
c78ba6d6 4886 unsigned int pref;
86872cb5 4887 int err;
1da177e4 4888
8cb08174
JB		 4889 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
4890 rtm_ipv6_policy, extack);
86872cb5
TG		 4891 if (err < 0)
4892 goto errout;
1da177e4 4893
86872cb5
TG		 4894 err = -EINVAL;
4895 rtm = nlmsg_data(nlh);
86872cb5 4896
84db8407
MŻ		 4897 *cfg = (struct fib6_config){
4898 .fc_table = rtm->rtm_table,
4899 .fc_dst_len = rtm->rtm_dst_len,
4900 .fc_src_len = rtm->rtm_src_len,
4901 .fc_flags = RTF_UP,
4902 .fc_protocol = rtm->rtm_protocol,
4903 .fc_type = rtm->rtm_type,
4904
4905 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
4906 .fc_nlinfo.nlh = nlh,
4907 .fc_nlinfo.nl_net = sock_net(skb->sk),
4908 };
86872cb5 4909
ef2c7d7b
ND		 4910 if (rtm->rtm_type == RTN_UNREACHABLE ||
4911 rtm->rtm_type == RTN_BLACKHOLE ||
b4949ab2
ND		 4912 rtm->rtm_type == RTN_PROHIBIT ||
4913 rtm->rtm_type == RTN_THROW)
86872cb5
TG		 4914 cfg->fc_flags |= RTF_REJECT;
4915
ab79ad14
MŻ		 4916 if (rtm->rtm_type == RTN_LOCAL)
4917 cfg->fc_flags |= RTF_LOCAL;
4918
1f56a01f
MKL		 4919 if (rtm->rtm_flags & RTM_F_CLONED)
4920 cfg->fc_flags |= RTF_CACHE;
4921
fc1e64e1
DA		 4922 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4923
5b98324e
DA		 4924 if (tb[RTA_NH_ID]) {
4925 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
4926 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
4927 NL_SET_ERR_MSG(extack,
4928 "Nexthop specification and nexthop id are mutually exclusive");
4929 goto errout;
4930 }
4931 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
4932 }
4933
86872cb5 4934 if (tb[RTA_GATEWAY]) {
67b61f6c 4935 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
86872cb5 4936 cfg->fc_flags |= RTF_GATEWAY;
1da177e4 4937 }
e3818541
DA		 4938 if (tb[RTA_VIA]) {
4939 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
4940 goto errout;
4941 }
86872cb5
TG		 4942
4943 if (tb[RTA_DST]) {
4944 int plen = (rtm->rtm_dst_len + 7) >> 3;
4945
4946 if (nla_len(tb[RTA_DST]) < plen)
4947 goto errout;
4948
4949 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1da177e4 4950 }
86872cb5
TG		 4951
4952 if (tb[RTA_SRC]) {
4953 int plen = (rtm->rtm_src_len + 7) >> 3;
4954
4955 if (nla_len(tb[RTA_SRC]) < plen)
4956 goto errout;
4957
4958 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1da177e4 4959 }
86872cb5 4960
c3968a85 4961 if (tb[RTA_PREFSRC])
67b61f6c 4962 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
c3968a85 4963
86872cb5
TG		 4964 if (tb[RTA_OIF])
4965 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4966
4967 if (tb[RTA_PRIORITY])
4968 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4969
4970 if (tb[RTA_METRICS]) {
4971 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4972 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
1da177e4 4973 }
86872cb5
TG		 4974
4975 if (tb[RTA_TABLE])
4976 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4977
51ebd318
ND		 4978 if (tb[RTA_MULTIPATH]) {
4979 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4980 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
9ed59592
DA		 4981
4982 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
c255bd68 4983 cfg->fc_mp_len, extack);
9ed59592
DA		 4984 if (err < 0)
4985 goto errout;
51ebd318
ND		 4986 }
4987
c78ba6d6
LR		 4988 if (tb[RTA_PREF]) {
4989 pref = nla_get_u8(tb[RTA_PREF]);
4990 if (pref != ICMPV6_ROUTER_PREF_LOW &&
4991 pref != ICMPV6_ROUTER_PREF_HIGH)
4992 pref = ICMPV6_ROUTER_PREF_MEDIUM;
4993 cfg->fc_flags |= RTF_PREF(pref);
4994 }
4995
19e42e45
RP		 4996 if (tb[RTA_ENCAP])
4997 cfg->fc_encap = tb[RTA_ENCAP];
4998
9ed59592 4999 if (tb[RTA_ENCAP_TYPE]) {
19e42e45
RP		 5000 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5001
c255bd68 5002 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
9ed59592
DA		 5003 if (err < 0)
5004 goto errout;
5005 }
5006
32bc201e
XL		 5007 if (tb[RTA_EXPIRES]) {
5008 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5009
5010 if (addrconf_finite_timeout(timeout)) {
5011 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5012 cfg->fc_flags |= RTF_EXPIRES;
5013 }
5014 }
5015
86872cb5
TG		 5016 err = 0;
5017errout:
5018 return err;
1da177e4
LT		 5019}
5020
6b9ea5a6 5021struct rt6_nh {
8d1c802b 5022 struct fib6_info *fib6_info;
6b9ea5a6 5023 struct fib6_config r_cfg;
6b9ea5a6
RP		 5024 struct list_head next;
5025};
5026
d4ead6b3
DA		 5027static int ip6_route_info_append(struct net *net,
5028 struct list_head *rt6_nh_list,
8d1c802b
DA		 5029 struct fib6_info *rt,
5030 struct fib6_config *r_cfg)
6b9ea5a6
RP		 5031{
5032 struct rt6_nh *nh;
6b9ea5a6
RP		 5033 int err = -EEXIST;
5034
5035 list_for_each_entry(nh, rt6_nh_list, next) {
8d1c802b
DA		 5036 /* check if fib6_info already exists */
5037 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
6b9ea5a6
RP		 5038 return err;
5039 }
5040
5041 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5042 if (!nh)
5043 return -ENOMEM;
8d1c802b 5044 nh->fib6_info = rt;
6b9ea5a6
RP		 5045 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5046 list_add_tail(&nh->next, rt6_nh_list);
5047
5048 return 0;
5049}
5050
8d1c802b
DA		 5051static void ip6_route_mpath_notify(struct fib6_info *rt,
5052 struct fib6_info *rt_last,
3b1137fe
DA		 5053 struct nl_info *info,
5054 __u16 nlflags)
5055{
5056 /* if this is an APPEND route, then rt points to the first route
5057 * inserted and rt_last points to last route inserted. Userspace
5058 * wants a consistent dump of the route which starts at the first
5059 * nexthop. Since sibling routes are always added at the end of
5060 * the list, find the first sibling of the last route appended
5061 */
93c2fb25
DA		 5062 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5063 rt = list_first_entry(&rt_last->fib6_siblings,
8d1c802b 5064 struct fib6_info,
93c2fb25 5065 fib6_siblings);
3b1137fe
DA		 5066 }
5067
5068 if (rt)
5069 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5070}
5071
0ee0f47c
IS		 5072static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5073{
5074 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5075 bool should_notify = false;
5076 struct fib6_info *leaf;
5077 struct fib6_node *fn;
5078
5079 rcu_read_lock();
5080 fn = rcu_dereference(rt->fib6_node);
5081 if (!fn)
5082 goto out;
5083
5084 leaf = rcu_dereference(fn->leaf);
5085 if (!leaf)
5086 goto out;
5087
5088 if (rt == leaf ||
5089 (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5090 rt6_qualify_for_ecmp(leaf)))
5091 should_notify = true;
5092out:
5093 rcu_read_unlock();
5094
5095 return should_notify;
5096}
5097
333c4301
DA		 5098static int ip6_route_multipath_add(struct fib6_config *cfg,
5099 struct netlink_ext_ack *extack)
51ebd318 5100{
8d1c802b 5101 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
3b1137fe 5102 struct nl_info *info = &cfg->fc_nlinfo;
51ebd318
ND		 5103 struct fib6_config r_cfg;
5104 struct rtnexthop *rtnh;
8d1c802b 5105 struct fib6_info *rt;
6b9ea5a6
RP		 5106 struct rt6_nh *err_nh;
5107 struct rt6_nh *nh, *nh_safe;
3b1137fe 5108 __u16 nlflags;
51ebd318
ND		 5109 int remaining;
5110 int attrlen;
6b9ea5a6
RP		 5111 int err = 1;
5112 int nhn = 0;
5113 int replace = (cfg->fc_nlinfo.nlh &&
5114 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5115 LIST_HEAD(rt6_nh_list);
51ebd318 5116
3b1137fe
DA		 5117 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5118 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5119 nlflags |= NLM_F_APPEND;
5120
35f1b4e9 5121 remaining = cfg->fc_mp_len;
51ebd318 5122 rtnh = (struct rtnexthop *)cfg->fc_mp;
51ebd318 5123
6b9ea5a6 5124 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
8d1c802b 5125 * fib6_info structs per nexthop
6b9ea5a6 5126 */
51ebd318
ND		 5127 while (rtnh_ok(rtnh, remaining)) {
5128 memcpy(&r_cfg, cfg, sizeof(*cfg));
5129 if (rtnh->rtnh_ifindex)
5130 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5131
5132 attrlen = rtnh_attrlen(rtnh);
5133 if (attrlen > 0) {
5134 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5135
5136 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5137 if (nla) {
67b61f6c 5138 r_cfg.fc_gateway = nla_get_in6_addr(nla);
51ebd318
ND		 5139 r_cfg.fc_flags |= RTF_GATEWAY;
5140 }
19e42e45
RP		 5141 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5142 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5143 if (nla)
5144 r_cfg.fc_encap_type = nla_get_u16(nla);
51ebd318 5145 }
6b9ea5a6 5146
68e2ffde 5147 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
acb54e3c 5148 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
8c5b83f0
RP		 5149 if (IS_ERR(rt)) {
5150 err = PTR_ERR(rt);
5151 rt = NULL;
6b9ea5a6 5152 goto cleanup;
8c5b83f0 5153 }
b5d2d75e
DA		 5154 if (!rt6_qualify_for_ecmp(rt)) {
5155 err = -EINVAL;
5156 NL_SET_ERR_MSG(extack,
5157 "Device only routes can not be added for IPv6 using the multipath API.");
5158 fib6_info_release(rt);
5159 goto cleanup;
5160 }
6b9ea5a6 5161
1cf844c7 5162 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
398958ae 5163
d4ead6b3
DA		 5164 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5165 rt, &r_cfg);
51ebd318 5166 if (err) {
93531c67 5167 fib6_info_release(rt);
6b9ea5a6
RP		 5168 goto cleanup;
5169 }
5170
5171 rtnh = rtnh_next(rtnh, &remaining);
5172 }
5173
9eee3b49
IS		 5174 if (list_empty(&rt6_nh_list)) {
5175 NL_SET_ERR_MSG(extack,
5176 "Invalid nexthop configuration - no valid nexthops");
5177 return -EINVAL;
5178 }
5179
3b1137fe
DA		 5180 /* for add and replace send one notification with all nexthops.
5181 * Skip the notification in fib6_add_rt2node and send one with
5182 * the full route when done
5183 */
5184 info->skip_notify = 1;
5185
ebee3cad
IS		 5186 /* For add and replace, send one notification with all nexthops. For
5187 * append, send one notification with all appended nexthops.
5188 */
5189 info->skip_notify_kernel = 1;
5190
6b9ea5a6
RP		 5191 err_nh = NULL;
5192 list_for_each_entry(nh, &rt6_nh_list, next) {
8d1c802b
DA		 5193 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5194 fib6_info_release(nh->fib6_info);
93531c67 5195
f7225172
DA		 5196 if (!err) {
5197 /* save reference to last route successfully inserted */
5198 rt_last = nh->fib6_info;
5199
5200 /* save reference to first route for notification */
5201 if (!rt_notif)
5202 rt_notif = nh->fib6_info;
5203 }
3b1137fe 5204
8d1c802b
DA		 5205 /* nh->fib6_info is used or freed at this point, reset to NULL*/
5206 nh->fib6_info = NULL;
6b9ea5a6
RP		 5207 if (err) {
5208 if (replace && nhn)
a5a82d84
JK		 5209 NL_SET_ERR_MSG_MOD(extack,
5210 "multipath route replace failed (check consistency of installed routes)");
6b9ea5a6
RP		 5211 err_nh = nh;
5212 goto add_errout;
51ebd318 5213 }
6b9ea5a6 5214
1a72418b 5215 /* Because each route is added like a single route we remove
27596472
MK		 5216 * these flags after the first nexthop: if there is a collision,
5217 * we have already failed to add the first nexthop:
5218 * fib6_add_rt2node() has rejected it; when replacing, old
5219 * nexthops have been replaced by first new, the rest should
5220 * be added to it.
1a72418b 5221 */
27596472
MK		 5222 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5223 NLM_F_REPLACE);
afecdb37 5224 cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
6b9ea5a6
RP		 5225 nhn++;
5226 }
5227
0ee0f47c
IS		 5228 /* An in-kernel notification should only be sent in case the new
5229 * multipath route is added as the first route in the node, or if
5230 * it was appended to it. We pass 'rt_notif' since it is the first
5231 * sibling and might allow us to skip some checks in the replace case.
5232 */
5233 if (ip6_route_mpath_should_notify(rt_notif)) {
5234 enum fib_event_type fib_event;
5235
5236 if (rt_notif->fib6_nsiblings != nhn - 1)
5237 fib_event = FIB_EVENT_ENTRY_APPEND;
5238 else
caafb250 5239 fib_event = FIB_EVENT_ENTRY_REPLACE;
0ee0f47c
IS		 5240
5241 err = call_fib6_multipath_entry_notifiers(info->nl_net,
5242 fib_event, rt_notif,
5243 nhn - 1, extack);
5244 if (err) {
5245 /* Delete all the siblings that were just added */
5246 err_nh = NULL;
5247 goto add_errout;
5248 }
5249 }
ebee3cad 5250
3b1137fe
DA		 5251 /* success ... tell user about new route */
5252 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
6b9ea5a6
RP		 5253 goto cleanup;
5254
5255add_errout:
3b1137fe
DA		 5256 /* send notification for routes that were added so that
5257 * the delete notifications sent by ip6_route_del are
5258 * coherent
5259 */
5260 if (rt_notif)
5261 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5262
6b9ea5a6
RP		 5263 /* Delete routes that were already added */
5264 list_for_each_entry(nh, &rt6_nh_list, next) {
5265 if (err_nh == nh)
5266 break;
333c4301 5267 ip6_route_del(&nh->r_cfg, extack);
6b9ea5a6
RP		 5268 }
5269
5270cleanup:
5271 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
8d1c802b
DA		 5272 if (nh->fib6_info)
5273 fib6_info_release(nh->fib6_info);
6b9ea5a6
RP		 5274 list_del(&nh->next);
5275 kfree(nh);
5276 }
5277
5278 return err;
5279}
5280
333c4301
DA		 5281static int ip6_route_multipath_del(struct fib6_config *cfg,
5282 struct netlink_ext_ack *extack)
6b9ea5a6
RP		 5283{
5284 struct fib6_config r_cfg;
5285 struct rtnexthop *rtnh;
5286 int remaining;
5287 int attrlen;
5288 int err = 1, last_err = 0;
5289
5290 remaining = cfg->fc_mp_len;
5291 rtnh = (struct rtnexthop *)cfg->fc_mp;
5292
5293 /* Parse a Multipath Entry */
5294 while (rtnh_ok(rtnh, remaining)) {
5295 memcpy(&r_cfg, cfg, sizeof(*cfg));
5296 if (rtnh->rtnh_ifindex)
5297 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5298
5299 attrlen = rtnh_attrlen(rtnh);
5300 if (attrlen > 0) {
5301 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5302
5303 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5304 if (nla) {
5305 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
5306 r_cfg.fc_flags |= RTF_GATEWAY;
5307 }
5308 }
333c4301 5309 err = ip6_route_del(&r_cfg, extack);
6b9ea5a6
RP		 5310 if (err)
5311 last_err = err;
5312
51ebd318
ND		 5313 rtnh = rtnh_next(rtnh, &remaining);
5314 }
5315
5316 return last_err;
5317}
5318
c21ef3e3
DA		 5319static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5320 struct netlink_ext_ack *extack)
1da177e4 5321{
86872cb5
TG		 5322 struct fib6_config cfg;
5323 int err;
1da177e4 5324
333c4301 5325 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 5326 if (err < 0)
5327 return err;
5328
5b98324e
DA		 5329 if (cfg.fc_nh_id &&
5330 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5331 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5332 return -EINVAL;
5333 }
5334
51ebd318 5335 if (cfg.fc_mp)
333c4301 5336 return ip6_route_multipath_del(&cfg, extack);
0ae81335
DA		 5337 else {
5338 cfg.fc_delete_all_nh = 1;
333c4301 5339 return ip6_route_del(&cfg, extack);
0ae81335 5340 }
1da177e4
LT		 5341}
5342
c21ef3e3
DA		 5343static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5344 struct netlink_ext_ack *extack)
1da177e4 5345{
86872cb5
TG		 5346 struct fib6_config cfg;
5347 int err;
1da177e4 5348
333c4301 5349 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 5350 if (err < 0)
5351 return err;
5352
67f69513
DA		 5353 if (cfg.fc_metric == 0)
5354 cfg.fc_metric = IP6_RT_PRIO_USER;
5355
51ebd318 5356 if (cfg.fc_mp)
333c4301 5357 return ip6_route_multipath_add(&cfg, extack);
51ebd318 5358 else
acb54e3c 5359 return ip6_route_add(&cfg, GFP_KERNEL, extack);
1da177e4
LT		 5360}
5361
a1b7a1f0
DA		 5362/* add the overhead of this fib6_nh to nexthop_len */
5363static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
339bf98f 5364{
a1b7a1f0 5365 int *nexthop_len = arg;
beb1afac 5366
a1b7a1f0
DA		 5367 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5368 + NLA_ALIGN(sizeof(struct rtnexthop))
5369 + nla_total_size(16); /* RTA_GATEWAY */
5370
5371 if (nh->fib_nh_lws) {
5372 /* RTA_ENCAP_TYPE */
5373 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5374 /* RTA_ENCAP */
5375 *nexthop_len += nla_total_size(2);
5376 }
f88d8ea6 5377
a1b7a1f0
DA		 5378 return 0;
5379}
beb1afac 5380
a1b7a1f0
DA		 5381static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5382{
5383 int nexthop_len;
5384
5385 if (f6i->nh) {
5386 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5387 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5388 &nexthop_len);
5389 } else {
5390 struct fib6_nh *nh = f6i->fib6_nh;
5391
5392 nexthop_len = 0;
5393 if (f6i->fib6_nsiblings) {
5394 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
5395 + NLA_ALIGN(sizeof(struct rtnexthop))
5396 + nla_total_size(16) /* RTA_GATEWAY */
5397 + lwtunnel_get_encap_size(nh->fib_nh_lws);
5398
5399 nexthop_len *= f6i->fib6_nsiblings;
5400 }
5401 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
beb1afac
DA		 5402 }
5403
339bf98f
TG		 5404 return NLMSG_ALIGN(sizeof(struct rtmsg))
5405 + nla_total_size(16) /* RTA_SRC */
5406 + nla_total_size(16) /* RTA_DST */
5407 + nla_total_size(16) /* RTA_GATEWAY */
5408 + nla_total_size(16) /* RTA_PREFSRC */
5409 + nla_total_size(4) /* RTA_TABLE */
5410 + nla_total_size(4) /* RTA_IIF */
5411 + nla_total_size(4) /* RTA_OIF */
5412 + nla_total_size(4) /* RTA_PRIORITY */
6a2b9ce0 5413 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
ea697639 5414 + nla_total_size(sizeof(struct rta_cacheinfo))
c78ba6d6 5415 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
19e42e45 5416 + nla_total_size(1) /* RTA_PREF */
beb1afac
DA		 5417 + nexthop_len;
5418}
5419
f88d8ea6
DA		 5420static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5421 unsigned char *flags)
5422{
5423 if (nexthop_is_multipath(nh)) {
5424 struct nlattr *mp;
5425
4255ff05 5426 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
f88d8ea6
DA		 5427 if (!mp)
5428 goto nla_put_failure;
5429
7bdf4de1 5430 if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
f88d8ea6
DA		 5431 goto nla_put_failure;
5432
5433 nla_nest_end(skb, mp);
5434 } else {
5435 struct fib6_nh *fib6_nh;
5436
5437 fib6_nh = nexthop_fib6_nh(nh);
7bdf4de1 5438 if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
f88d8ea6
DA		 5439 flags, false) < 0)
5440 goto nla_put_failure;
5441 }
5442
5443 return 0;
5444
5445nla_put_failure:
5446 return -EMSGSIZE;
5447}
5448
d4ead6b3 5449static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 5450 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 5451 struct in6_addr *dest, struct in6_addr *src,
15e47304 5452 int iif, int type, u32 portid, u32 seq,
f8cfe2ce 5453 unsigned int flags)
1da177e4 5454{
22d0bd82
XL		 5455 struct rt6_info *rt6 = (struct rt6_info *)dst;
5456 struct rt6key *rt6_dst, *rt6_src;
5457 u32 *pmetrics, table, rt6_flags;
f88d8ea6 5458 unsigned char nh_flags = 0;
2d7202bf 5459 struct nlmsghdr *nlh;
22d0bd82 5460 struct rtmsg *rtm;
d4ead6b3 5461 long expires = 0;
1da177e4 5462
15e47304 5463 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
38308473 5464 if (!nlh)
26932566 5465 return -EMSGSIZE;
2d7202bf 5466
22d0bd82
XL		 5467 if (rt6) {
5468 rt6_dst = &rt6->rt6i_dst;
5469 rt6_src = &rt6->rt6i_src;
5470 rt6_flags = rt6->rt6i_flags;
5471 } else {
5472 rt6_dst = &rt->fib6_dst;
5473 rt6_src = &rt->fib6_src;
5474 rt6_flags = rt->fib6_flags;
5475 }
5476
2d7202bf 5477 rtm = nlmsg_data(nlh);
1da177e4 5478 rtm->rtm_family = AF_INET6;
22d0bd82
XL		 5479 rtm->rtm_dst_len = rt6_dst->plen;
5480 rtm->rtm_src_len = rt6_src->plen;
1da177e4 5481 rtm->rtm_tos = 0;
93c2fb25
DA		 5482 if (rt->fib6_table)
5483 table = rt->fib6_table->tb6_id;
c71099ac 5484 else
9e762a4a 5485 table = RT6_TABLE_UNSPEC;
97f0082a 5486 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
c78679e8
DM		 5487 if (nla_put_u32(skb, RTA_TABLE, table))
5488 goto nla_put_failure;
e8478e80
DA		 5489
5490 rtm->rtm_type = rt->fib6_type;
1da177e4
LT		 5491 rtm->rtm_flags = 0;
5492 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
93c2fb25 5493 rtm->rtm_protocol = rt->fib6_protocol;
1da177e4 5494
22d0bd82 5495 if (rt6_flags & RTF_CACHE)
1da177e4
LT		 5496 rtm->rtm_flags |= RTM_F_CLONED;
5497
d4ead6b3
DA		 5498 if (dest) {
5499 if (nla_put_in6_addr(skb, RTA_DST, dest))
c78679e8 5500 goto nla_put_failure;
1ab1457c 5501 rtm->rtm_dst_len = 128;
1da177e4 5502 } else if (rtm->rtm_dst_len)
22d0bd82 5503 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
c78679e8 5504 goto nla_put_failure;
1da177e4
LT		 5505#ifdef CONFIG_IPV6_SUBTREES
5506 if (src) {
930345ea 5507 if (nla_put_in6_addr(skb, RTA_SRC, src))
c78679e8 5508 goto nla_put_failure;
1ab1457c 5509 rtm->rtm_src_len = 128;
c78679e8 5510 } else if (rtm->rtm_src_len &&
22d0bd82 5511 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
c78679e8 5512 goto nla_put_failure;
1da177e4 5513#endif
7bc570c8
YH		 5514 if (iif) {
5515#ifdef CONFIG_IPV6_MROUTE
22d0bd82 5516 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
fd61c6ba
DA		 5517 int err = ip6mr_get_route(net, skb, rtm, portid);
5518
5519 if (err == 0)
5520 return 0;
5521 if (err < 0)
5522 goto nla_put_failure;
7bc570c8
YH		 5523 } else
5524#endif
c78679e8
DM		 5525 if (nla_put_u32(skb, RTA_IIF, iif))
5526 goto nla_put_failure;
d4ead6b3 5527 } else if (dest) {
1da177e4 5528 struct in6_addr saddr_buf;
d4ead6b3 5529 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
930345ea 5530 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 5531 goto nla_put_failure;
1da177e4 5532 }
2d7202bf 5533
93c2fb25 5534 if (rt->fib6_prefsrc.plen) {
c3968a85 5535 struct in6_addr saddr_buf;
93c2fb25 5536 saddr_buf = rt->fib6_prefsrc.addr;
930345ea 5537 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 5538 goto nla_put_failure;
c3968a85
DW		 5539 }
5540
d4ead6b3
DA		 5541 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5542 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
2d7202bf
TG		 5543 goto nla_put_failure;
5544
93c2fb25 5545 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
c78679e8 5546 goto nla_put_failure;
8253947e 5547
beb1afac
DA		 5548 /* For multipath routes, walk the siblings list and add
5549 * each as a nexthop within RTA_MULTIPATH.
5550 */
22d0bd82
XL		 5551 if (rt6) {
5552 if (rt6_flags & RTF_GATEWAY &&
5553 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5554 goto nla_put_failure;
5555
5556 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5557 goto nla_put_failure;
5558 } else if (rt->fib6_nsiblings) {
8d1c802b 5559 struct fib6_info *sibling, *next_sibling;
beb1afac
DA		 5560 struct nlattr *mp;
5561
ae0be8de 5562 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
beb1afac
DA		 5563 if (!mp)
5564 goto nla_put_failure;
5565
1cf844c7 5566 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
7bdf4de1 5567 rt->fib6_nh->fib_nh_weight, AF_INET6) < 0)
beb1afac
DA		 5568 goto nla_put_failure;
5569
5570 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25 5571 &rt->fib6_siblings, fib6_siblings) {
1cf844c7 5572 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
7bdf4de1
DS		 5573 sibling->fib6_nh->fib_nh_weight,
5574 AF_INET6) < 0)
beb1afac
DA		 5575 goto nla_put_failure;
5576 }
5577
5578 nla_nest_end(skb, mp);
f88d8ea6
DA		 5579 } else if (rt->nh) {
5580 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5581 goto nla_put_failure;
ecc5663c 5582
f88d8ea6
DA		 5583 if (nexthop_is_blackhole(rt->nh))
5584 rtm->rtm_type = RTN_BLACKHOLE;
5585
4f80116d
RP		 5586 if (net->ipv4.sysctl_nexthop_compat_mode &&
5587 rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
f88d8ea6
DA		 5588 goto nla_put_failure;
5589
5590 rtm->rtm_flags |= nh_flags;
5591 } else {
7bdf4de1 5592 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
ecc5663c 5593 &nh_flags, false) < 0)
beb1afac 5594 goto nla_put_failure;
ecc5663c
DA		 5595
5596 rtm->rtm_flags |= nh_flags;
beb1afac
DA		 5597 }
5598
22d0bd82 5599 if (rt6_flags & RTF_EXPIRES) {
14895687
DA		 5600 expires = dst ? dst->expires : rt->expires;
5601 expires -= jiffies;
5602 }
69cdf8f9 5603
bb3c4ab9
IS		 5604 if (!dst) {
5605 if (rt->offload)
5606 rtm->rtm_flags |= RTM_F_OFFLOAD;
5607 if (rt->trap)
5608 rtm->rtm_flags |= RTM_F_TRAP;
5609 }
5610
d4ead6b3 5611 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
e3703b3d 5612 goto nla_put_failure;
2d7202bf 5613
22d0bd82 5614 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
c78ba6d6
LR		 5615 goto nla_put_failure;
5616
19e42e45 5617
053c095a
JB		 5618 nlmsg_end(skb, nlh);
5619 return 0;
2d7202bf
TG		 5620
5621nla_put_failure:
26932566
PM		 5622 nlmsg_cancel(skb, nlh);
5623 return -EMSGSIZE;
1da177e4
LT		 5624}
5625
2c170e07
DA		 5626static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5627{
5628 const struct net_device *dev = arg;
5629
5630 if (nh->fib_nh_dev == dev)
5631 return 1;
5632
5633 return 0;
5634}
5635
13e38901
DA		 5636static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5637 const struct net_device *dev)
5638{
2c170e07
DA		 5639 if (f6i->nh) {
5640 struct net_device *_dev = (struct net_device *)dev;
5641
5642 return !!nexthop_for_each_fib6_nh(f6i->nh,
5643 fib6_info_nh_uses_dev,
5644 _dev);
5645 }
5646
1cf844c7 5647 if (f6i->fib6_nh->fib_nh_dev == dev)
13e38901
DA		 5648 return true;
5649
5650 if (f6i->fib6_nsiblings) {
5651 struct fib6_info *sibling, *next_sibling;
5652
5653 list_for_each_entry_safe(sibling, next_sibling,
5654 &f6i->fib6_siblings, fib6_siblings) {
1cf844c7 5655 if (sibling->fib6_nh->fib_nh_dev == dev)
13e38901
DA		 5656 return true;
5657 }
5658 }
5659
5660 return false;
5661}
5662
1e47b483
SB		 5663struct fib6_nh_exception_dump_walker {
5664 struct rt6_rtnl_dump_arg *dump;
5665 struct fib6_info *rt;
5666 unsigned int flags;
5667 unsigned int skip;
5668 unsigned int count;
5669};
5670
5671static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5672{
5673 struct fib6_nh_exception_dump_walker *w = arg;
5674 struct rt6_rtnl_dump_arg *dump = w->dump;
5675 struct rt6_exception_bucket *bucket;
5676 struct rt6_exception *rt6_ex;
5677 int i, err;
5678
5679 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5680 if (!bucket)
5681 return 0;
5682
5683 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5684 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5685 if (w->skip) {
5686 w->skip--;
5687 continue;
5688 }
5689
5690 /* Expiration of entries doesn't bump sernum, insertion
5691 * does. Removal is triggered by insertion, so we can
5692 * rely on the fact that if entries change between two
5693 * partial dumps, this node is scanned again completely,
5694 * see rt6_insert_exception() and fib6_dump_table().
5695 *
5696 * Count expired entries we go through as handled
5697 * entries that we'll skip next time, in case of partial
5698 * node dump. Otherwise, if entries expire meanwhile,
5699 * we'll skip the wrong amount.
5700 */
5701 if (rt6_check_expired(rt6_ex->rt6i)) {
5702 w->count++;
5703 continue;
5704 }
5705
5706 err = rt6_fill_node(dump->net, dump->skb, w->rt,
5707 &rt6_ex->rt6i->dst, NULL, NULL, 0,
5708 RTM_NEWROUTE,
5709 NETLINK_CB(dump->cb->skb).portid,
5710 dump->cb->nlh->nlmsg_seq, w->flags);
5711 if (err)
5712 return err;
5713
5714 w->count++;
5715 }
5716 bucket++;
5717 }
5718
5719 return 0;
5720}
5721
bf9a8a06 5722/* Return -1 if done with node, number of handled routes on partial dump */
1e47b483 5723int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
1da177e4
LT		 5724{
5725 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
13e38901
DA		 5726 struct fib_dump_filter *filter = &arg->filter;
5727 unsigned int flags = NLM_F_MULTI;
1f17e2f2 5728 struct net *net = arg->net;
1e47b483 5729 int count = 0;
1f17e2f2 5730
421842ed 5731 if (rt == net->ipv6.fib6_null_entry)
bf9a8a06 5732 return -1;
1da177e4 5733
13e38901
DA		 5734 if ((filter->flags & RTM_F_PREFIX) &&
5735 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5736 /* success since this is not a prefix route */
bf9a8a06 5737 return -1;
13e38901 5738 }
1e47b483
SB		 5739 if (filter->filter_set &&
5740 ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5741 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5742 (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5743 return -1;
5744 }
5745
5746 if (filter->filter_set ||
5747 !filter->dump_routes || !filter->dump_exceptions) {
13e38901 5748 flags |= NLM_F_DUMP_FILTERED;
f8cfe2ce 5749 }
1da177e4 5750
1e47b483
SB		 5751 if (filter->dump_routes) {
5752 if (skip) {
5753 skip--;
5754 } else {
5755 if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5756 0, RTM_NEWROUTE,
5757 NETLINK_CB(arg->cb->skb).portid,
5758 arg->cb->nlh->nlmsg_seq, flags)) {
5759 return 0;
5760 }
5761 count++;
5762 }
5763 }
5764
5765 if (filter->dump_exceptions) {
5766 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5767 .rt = rt,
5768 .flags = flags,
5769 .skip = skip,
5770 .count = 0 };
5771 int err;
5772
3b525691 5773 rcu_read_lock();
1e47b483
SB		 5774 if (rt->nh) {
5775 err = nexthop_for_each_fib6_nh(rt->nh,
5776 rt6_nh_dump_exceptions,
5777 &w);
5778 } else {
5779 err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5780 }
3b525691 5781 rcu_read_unlock();
1e47b483
SB		 5782
5783 if (err)
5784 return count += w.count;
5785 }
bf9a8a06
SB		 5786
5787 return -1;
1da177e4
LT		 5788}
5789
0eff0a27
JK		 5790static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5791 const struct nlmsghdr *nlh,
5792 struct nlattr **tb,
5793 struct netlink_ext_ack *extack)
5794{
5795 struct rtmsg *rtm;
5796 int i, err;
5797
5798 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5799 NL_SET_ERR_MSG_MOD(extack,
5800 "Invalid header for get route request");
5801 return -EINVAL;
5802 }
5803
5804 if (!netlink_strict_get_check(skb))
8cb08174
JB		 5805 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5806 rtm_ipv6_policy, extack);
0eff0a27
JK		 5807
5808 rtm = nlmsg_data(nlh);
5809 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5810 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5811 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5812 rtm->rtm_type) {
5813 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5814 return -EINVAL;
5815 }
5816 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5817 NL_SET_ERR_MSG_MOD(extack,
5818 "Invalid flags for get route request");
5819 return -EINVAL;
5820 }
5821
8cb08174
JB		 5822 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5823 rtm_ipv6_policy, extack);
0eff0a27
JK		 5824 if (err)
5825 return err;
5826
5827 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5828 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5829 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5830 return -EINVAL;
5831 }
5832
5833 for (i = 0; i <= RTA_MAX; i++) {
5834 if (!tb[i])
5835 continue;
5836
5837 switch (i) {
5838 case RTA_SRC:
5839 case RTA_DST:
5840 case RTA_IIF:
5841 case RTA_OIF:
5842 case RTA_MARK:
5843 case RTA_UID:
5844 case RTA_SPORT:
5845 case RTA_DPORT:
5846 case RTA_IP_PROTO:
5847 break;
5848 default:
5849 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
5850 return -EINVAL;
5851 }
5852 }
5853
5854 return 0;
5855}
5856
c21ef3e3
DA		 5857static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5858 struct netlink_ext_ack *extack)
1da177e4 5859{
3b1e0a65 5860 struct net *net = sock_net(in_skb->sk);
ab364a6f 5861 struct nlattr *tb[RTA_MAX+1];
18c3a61c 5862 int err, iif = 0, oif = 0;
a68886a6 5863 struct fib6_info *from;
18c3a61c 5864 struct dst_entry *dst;
ab364a6f 5865 struct rt6_info *rt;
1da177e4 5866 struct sk_buff *skb;
ab364a6f 5867 struct rtmsg *rtm;
744486d4 5868 struct flowi6 fl6 = {};
18c3a61c 5869 bool fibmatch;
1da177e4 5870
0eff0a27 5871 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
ab364a6f
TG		 5872 if (err < 0)
5873 goto errout;
1da177e4 5874
ab364a6f 5875 err = -EINVAL;
38b7097b
HFS		 5876 rtm = nlmsg_data(nlh);
5877 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
18c3a61c 5878 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
1da177e4 5879
ab364a6f
TG		 5880 if (tb[RTA_SRC]) {
5881 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
5882 goto errout;
5883
4e3fd7a0 5884 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
ab364a6f
TG		 5885 }
5886
5887 if (tb[RTA_DST]) {
5888 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
5889 goto errout;
5890
4e3fd7a0 5891 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
ab364a6f
TG		 5892 }
5893
5894 if (tb[RTA_IIF])
5895 iif = nla_get_u32(tb[RTA_IIF]);
5896
5897 if (tb[RTA_OIF])
72331bc0 5898 oif = nla_get_u32(tb[RTA_OIF]);
1da177e4 5899
2e47b291
LC		 5900 if (tb[RTA_MARK])
5901 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
5902
622ec2c9
LC		 5903 if (tb[RTA_UID])
5904 fl6.flowi6_uid = make_kuid(current_user_ns(),
5905 nla_get_u32(tb[RTA_UID]));
5906 else
5907 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
5908
eacb9384
RP		 5909 if (tb[RTA_SPORT])
5910 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
5911
5912 if (tb[RTA_DPORT])
5913 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
5914
5915 if (tb[RTA_IP_PROTO]) {
5916 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
5e1a99ea
HL		 5917 &fl6.flowi6_proto, AF_INET6,
5918 extack);
eacb9384
RP		 5919 if (err)
5920 goto errout;
5921 }
5922
1da177e4
LT		 5923 if (iif) {
5924 struct net_device *dev;
72331bc0
SL		 5925 int flags = 0;
5926
121622db
FW		 5927 rcu_read_lock();
5928
5929 dev = dev_get_by_index_rcu(net, iif);
1da177e4 5930 if (!dev) {
121622db 5931 rcu_read_unlock();
1da177e4 5932 err = -ENODEV;
ab364a6f 5933 goto errout;
1da177e4 5934 }
72331bc0
SL		 5935
5936 fl6.flowi6_iif = iif;
5937
5938 if (!ipv6_addr_any(&fl6.saddr))
5939 flags |= RT6_LOOKUP_F_HAS_SADDR;
5940
b75cc8f9 5941 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
121622db
FW		 5942
5943 rcu_read_unlock();
72331bc0
SL		 5944 } else {
5945 fl6.flowi6_oif = oif;
5946
58acfd71 5947 dst = ip6_route_output(net, NULL, &fl6);
18c3a61c
RP		 5948 }
5949
18c3a61c
RP		 5950
5951 rt = container_of(dst, struct rt6_info, dst);
5952 if (rt->dst.error) {
5953 err = rt->dst.error;
5954 ip6_rt_put(rt);
5955 goto errout;
1da177e4
LT		 5956 }
5957
9d6acb3b
WC		 5958 if (rt == net->ipv6.ip6_null_entry) {
5959 err = rt->dst.error;
5960 ip6_rt_put(rt);
5961 goto errout;
5962 }
5963
ab364a6f 5964 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
38308473 5965 if (!skb) {
94e187c0 5966 ip6_rt_put(rt);
ab364a6f
TG		 5967 err = -ENOBUFS;
5968 goto errout;
5969 }
1da177e4 5970
d8d1f30b 5971 skb_dst_set(skb, &rt->dst);
a68886a6
DA		 5972
5973 rcu_read_lock();
5974 from = rcu_dereference(rt->from);
886b7a50
MKL		 5975 if (from) {
5976 if (fibmatch)
5977 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
5978 iif, RTM_NEWROUTE,
5979 NETLINK_CB(in_skb).portid,
5980 nlh->nlmsg_seq, 0);
5981 else
5982 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
5983 &fl6.saddr, iif, RTM_NEWROUTE,
5984 NETLINK_CB(in_skb).portid,
5985 nlh->nlmsg_seq, 0);
5986 } else {
5987 err = -ENETUNREACH;
5988 }
a68886a6
DA		 5989 rcu_read_unlock();
5990
1da177e4 5991 if (err < 0) {
ab364a6f
TG		 5992 kfree_skb(skb);
5993 goto errout;
1da177e4
LT		 5994 }
5995
15e47304 5996 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
ab364a6f 5997errout:
1da177e4 5998 return err;
1da177e4
LT		 5999}
6000
8d1c802b 6001void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
37a1d361 6002 unsigned int nlm_flags)
1da177e4
LT		 6003{
6004 struct sk_buff *skb;
5578689a 6005 struct net *net = info->nl_net;
528c4ceb
DL		 6006 u32 seq;
6007 int err;
6008
6009 err = -ENOBUFS;
38308473 6010 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
86872cb5 6011
19e42e45 6012 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
38308473 6013 if (!skb)
21713ebc
TG		 6014 goto errout;
6015
d4ead6b3
DA		 6016 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6017 event, info->portid, seq, nlm_flags);
26932566
PM		 6018 if (err < 0) {
6019 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6020 WARN_ON(err == -EMSGSIZE);
6021 kfree_skb(skb);
6022 goto errout;
6023 }
15e47304 6024 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
1ce85fe4
PNA		 6025 info->nlh, gfp_any());
6026 return;
21713ebc
TG		 6027errout:
6028 if (err < 0)
5578689a 6029 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
1da177e4
LT		 6030}
6031
19a3b7ee
DA		 6032void fib6_rt_update(struct net *net, struct fib6_info *rt,
6033 struct nl_info *info)
6034{
6035 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6036 struct sk_buff *skb;
6037 int err = -ENOBUFS;
6038
6039 /* call_fib6_entry_notifiers will be removed when in-kernel notifier
6040 * is implemented and supported for nexthop objects
6041 */
6042 call_fib6_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, rt, NULL);
6043
6044 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6045 if (!skb)
6046 goto errout;
6047
6048 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6049 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6050 if (err < 0) {
6051 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6052 WARN_ON(err == -EMSGSIZE);
6053 kfree_skb(skb);
6054 goto errout;
6055 }
6056 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6057 info->nlh, gfp_any());
6058 return;
6059errout:
6060 if (err < 0)
6061 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6062}
6063
8ed67789 6064static int ip6_route_dev_notify(struct notifier_block *this,
351638e7 6065 unsigned long event, void *ptr)
8ed67789 6066{
351638e7 6067 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
c346dca1 6068 struct net *net = dev_net(dev);
8ed67789 6069
242d3a49
WC		 6070 if (!(dev->flags & IFF_LOOPBACK))
6071 return NOTIFY_OK;
6072
6073 if (event == NETDEV_REGISTER) {
1cf844c7 6074 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
d8d1f30b 6075 net->ipv6.ip6_null_entry->dst.dev = dev;
8ed67789
DL		 6076 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6077#ifdef CONFIG_IPV6_MULTIPLE_TABLES
d8d1f30b 6078 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
8ed67789 6079 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
d8d1f30b 6080 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
8ed67789 6081 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
242d3a49 6082#endif
76da0704
WC		 6083 } else if (event == NETDEV_UNREGISTER &&
6084 dev->reg_state != NETREG_UNREGISTERED) {
6085 /* NETDEV_UNREGISTER could be fired for multiple times by
6086 * netdev_wait_allrefs(). Make sure we only call this once.
6087 */
12d94a80 6088 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
242d3a49 6089#ifdef CONFIG_IPV6_MULTIPLE_TABLES
12d94a80
ED		 6090 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6091 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
8ed67789
DL		 6092#endif
6093 }
6094
6095 return NOTIFY_OK;
6096}
6097
1da177e4
LT		 6098/*
6099 * /proc
6100 */
6101
6102#ifdef CONFIG_PROC_FS
1da177e4
LT		 6103static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6104{
69ddb805 6105 struct net *net = (struct net *)seq->private;
1da177e4 6106 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
69ddb805
DL		 6107 net->ipv6.rt6_stats->fib_nodes,
6108 net->ipv6.rt6_stats->fib_route_nodes,
81eb8447 6109 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
69ddb805
DL		 6110 net->ipv6.rt6_stats->fib_rt_entries,
6111 net->ipv6.rt6_stats->fib_rt_cache,
fc66f95c 6112 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
69ddb805 6113 net->ipv6.rt6_stats->fib_discarded_routes);
1da177e4
LT		 6114
6115 return 0;
6116}
1da177e4
LT		 6117#endif /* CONFIG_PROC_FS */
6118
6119#ifdef CONFIG_SYSCTL
6120
32927393
CH		 6121static int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
6122 void *buffer, size_t *lenp, loff_t *ppos)
1da177e4 6123{
c486da34
LAG		 6124 struct net *net;
6125 int delay;
f0fb9b28 6126 int ret;
c486da34 6127 if (!write)
1da177e4 6128 return -EINVAL;
c486da34
LAG		 6129
6130 net = (struct net *)ctl->extra1;
6131 delay = net->ipv6.sysctl.flush_delay;
f0fb9b28
AP		 6132 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6133 if (ret)
6134 return ret;
6135
2ac3ac8f 6136 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
c486da34 6137 return 0;
1da177e4
LT		 6138}
6139
ed792e28 6140static struct ctl_table ipv6_route_table_template[] = {
1ab1457c 6141 {
1da177e4 6142 .procname = "flush",
4990509f 6143 .data = &init_net.ipv6.sysctl.flush_delay,
1da177e4 6144 .maxlen = sizeof(int),
89c8b3a1 6145 .mode = 0200,
6d9f239a 6146 .proc_handler = ipv6_sysctl_rtcache_flush
1da177e4
LT		 6147 },
6148 {
1da177e4 6149 .procname = "gc_thresh",
9a7ec3a9 6150 .data = &ip6_dst_ops_template.gc_thresh,
1da177e4
LT		 6151 .maxlen = sizeof(int),
6152 .mode = 0644,
6d9f239a 6153 .proc_handler = proc_dointvec,
1da177e4
LT		 6154 },
6155 {
1da177e4 6156 .procname = "max_size",
4990509f 6157 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
1da177e4
LT		 6158 .maxlen = sizeof(int),
6159 .mode = 0644,
6d9f239a 6160 .proc_handler = proc_dointvec,
1da177e4
LT		 6161 },
6162 {
1da177e4 6163 .procname = "gc_min_interval",
4990509f 6164 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 6165 .maxlen = sizeof(int),
6166 .mode = 0644,
6d9f239a 6167 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6168 },
6169 {
1da177e4 6170 .procname = "gc_timeout",
4990509f 6171 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
1da177e4
LT		 6172 .maxlen = sizeof(int),
6173 .mode = 0644,
6d9f239a 6174 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6175 },
6176 {
1da177e4 6177 .procname = "gc_interval",
4990509f 6178 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
1da177e4
LT		 6179 .maxlen = sizeof(int),
6180 .mode = 0644,
6d9f239a 6181 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6182 },
6183 {
1da177e4 6184 .procname = "gc_elasticity",
4990509f 6185 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
1da177e4
LT		 6186 .maxlen = sizeof(int),
6187 .mode = 0644,
f3d3f616 6188 .proc_handler = proc_dointvec,
1da177e4
LT		 6189 },
6190 {
1da177e4 6191 .procname = "mtu_expires",
4990509f 6192 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
1da177e4
LT		 6193 .maxlen = sizeof(int),
6194 .mode = 0644,
6d9f239a 6195 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6196 },
6197 {
1da177e4 6198 .procname = "min_adv_mss",
4990509f 6199 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
1da177e4
LT		 6200 .maxlen = sizeof(int),
6201 .mode = 0644,
f3d3f616 6202 .proc_handler = proc_dointvec,
1da177e4
LT		 6203 },
6204 {
1da177e4 6205 .procname = "gc_min_interval_ms",
4990509f 6206 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 6207 .maxlen = sizeof(int),
6208 .mode = 0644,
6d9f239a 6209 .proc_handler = proc_dointvec_ms_jiffies,
1da177e4 6210 },
7c6bb7d2
DA		 6211 {
6212 .procname = "skip_notify_on_dev_down",
6213 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6214 .maxlen = sizeof(int),
6215 .mode = 0644,
b8e8a863 6216 .proc_handler = proc_dointvec_minmax,
eec4844f
MC		 6217 .extra1 = SYSCTL_ZERO,
6218 .extra2 = SYSCTL_ONE,
7c6bb7d2 6219 },
f8572d8f 6220 { }
1da177e4
LT		 6221};
6222
2c8c1e72 6223struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
760f2d01
DL		 6224{
6225 struct ctl_table *table;
6226
6227 table = kmemdup(ipv6_route_table_template,
6228 sizeof(ipv6_route_table_template),
6229 GFP_KERNEL);
5ee09105
YH		 6230
6231 if (table) {
6232 table[0].data = &net->ipv6.sysctl.flush_delay;
c486da34 6233 table[0].extra1 = net;
86393e52 6234 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
5ee09105
YH		 6235 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
6236 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6237 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6238 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6239 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6240 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6241 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
9c69fabe 6242 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
7c6bb7d2 6243 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
464dc801
EB		 6244
6245 /* Don't export sysctls to unprivileged users */
6246 if (net->user_ns != &init_user_ns)
6247 table[0].procname = NULL;
5ee09105
YH		 6248 }
6249
760f2d01
DL		 6250 return table;
6251}
1da177e4
LT		 6252#endif
6253
2c8c1e72 6254static int __net_init ip6_route_net_init(struct net *net)
cdb18761 6255{
633d424b 6256 int ret = -ENOMEM;
8ed67789 6257
86393e52
AD		 6258 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6259 sizeof(net->ipv6.ip6_dst_ops));
f2fc6a54 6260
fc66f95c
ED		 6261 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6262 goto out_ip6_dst_ops;
6263
1cf844c7 6264 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
421842ed
DA		 6265 if (!net->ipv6.fib6_null_entry)
6266 goto out_ip6_dst_entries;
1cf844c7
DA		 6267 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6268 sizeof(*net->ipv6.fib6_null_entry));
421842ed 6269
8ed67789
DL		 6270 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6271 sizeof(*net->ipv6.ip6_null_entry),
6272 GFP_KERNEL);
6273 if (!net->ipv6.ip6_null_entry)
421842ed 6274 goto out_fib6_null_entry;
d8d1f30b 6275 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6276 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6277 ip6_template_metrics, true);
74109218 6278 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->rt6i_uncached);
8ed67789
DL		 6279
6280#ifdef CONFIG_IPV6_MULTIPLE_TABLES
feca7d8c 6281 net->ipv6.fib6_has_custom_rules = false;
8ed67789
DL		 6282 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6283 sizeof(*net->ipv6.ip6_prohibit_entry),
6284 GFP_KERNEL);
68fffc67
PZ		 6285 if (!net->ipv6.ip6_prohibit_entry)
6286 goto out_ip6_null_entry;
d8d1f30b 6287 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6288 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6289 ip6_template_metrics, true);
74109218 6290 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->rt6i_uncached);
8ed67789
DL		 6291
6292 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6293 sizeof(*net->ipv6.ip6_blk_hole_entry),
6294 GFP_KERNEL);
68fffc67
PZ		 6295 if (!net->ipv6.ip6_blk_hole_entry)
6296 goto out_ip6_prohibit_entry;
d8d1f30b 6297 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6298 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6299 ip6_template_metrics, true);
74109218 6300 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->rt6i_uncached);
b9b33e7c
PA		 6301#ifdef CONFIG_IPV6_SUBTREES
6302 net->ipv6.fib6_routes_require_src = 0;
6303#endif
8ed67789
DL		 6304#endif
6305
b339a47c
PZ		 6306 net->ipv6.sysctl.flush_delay = 0;
6307 net->ipv6.sysctl.ip6_rt_max_size = 4096;
6308 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6309 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6310 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6311 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6312 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6313 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
7c6bb7d2 6314 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
b339a47c 6315
6891a346
BT		 6316 net->ipv6.ip6_rt_gc_expire = 30*HZ;
6317
8ed67789
DL		 6318 ret = 0;
6319out:
6320 return ret;
f2fc6a54 6321
68fffc67
PZ		 6322#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6323out_ip6_prohibit_entry:
6324 kfree(net->ipv6.ip6_prohibit_entry);
6325out_ip6_null_entry:
6326 kfree(net->ipv6.ip6_null_entry);
6327#endif
421842ed
DA		 6328out_fib6_null_entry:
6329 kfree(net->ipv6.fib6_null_entry);
fc66f95c
ED		 6330out_ip6_dst_entries:
6331 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
f2fc6a54 6332out_ip6_dst_ops:
f2fc6a54 6333 goto out;
cdb18761
DL		 6334}
6335
2c8c1e72 6336static void __net_exit ip6_route_net_exit(struct net *net)
cdb18761 6337{
421842ed 6338 kfree(net->ipv6.fib6_null_entry);
8ed67789
DL		 6339 kfree(net->ipv6.ip6_null_entry);
6340#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6341 kfree(net->ipv6.ip6_prohibit_entry);
6342 kfree(net->ipv6.ip6_blk_hole_entry);
6343#endif
41bb78b4 6344 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
cdb18761
DL		 6345}
6346
d189634e
TG		 6347static int __net_init ip6_route_net_init_late(struct net *net)
6348{
6349#ifdef CONFIG_PROC_FS
c3506372
CH		 6350 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6351 sizeof(struct ipv6_route_iter));
3617d949
CH		 6352 proc_create_net_single("rt6_stats", 0444, net->proc_net,
6353 rt6_stats_seq_show, NULL);
d189634e
TG		 6354#endif
6355 return 0;
6356}
6357
6358static void __net_exit ip6_route_net_exit_late(struct net *net)
6359{
6360#ifdef CONFIG_PROC_FS
ece31ffd
G		 6361 remove_proc_entry("ipv6_route", net->proc_net);
6362 remove_proc_entry("rt6_stats", net->proc_net);
d189634e
TG		 6363#endif
6364}
6365
cdb18761
DL		 6366static struct pernet_operations ip6_route_net_ops = {
6367 .init = ip6_route_net_init,
6368 .exit = ip6_route_net_exit,
6369};
6370
c3426b47
DM		 6371static int __net_init ipv6_inetpeer_init(struct net *net)
6372{
6373 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6374
6375 if (!bp)
6376 return -ENOMEM;
6377 inet_peer_base_init(bp);
6378 net->ipv6.peers = bp;
6379 return 0;
6380}
6381
6382static void __net_exit ipv6_inetpeer_exit(struct net *net)
6383{
6384 struct inet_peer_base *bp = net->ipv6.peers;
6385
6386 net->ipv6.peers = NULL;
56a6b248 6387 inetpeer_invalidate_tree(bp);
c3426b47
DM		 6388 kfree(bp);
6389}
6390
2b823f72 6391static struct pernet_operations ipv6_inetpeer_ops = {
c3426b47
DM		 6392 .init = ipv6_inetpeer_init,
6393 .exit = ipv6_inetpeer_exit,
6394};
6395
d189634e
TG		 6396static struct pernet_operations ip6_route_net_late_ops = {
6397 .init = ip6_route_net_init_late,
6398 .exit = ip6_route_net_exit_late,
6399};
6400
8ed67789
DL		 6401static struct notifier_block ip6_route_dev_notifier = {
6402 .notifier_call = ip6_route_dev_notify,
242d3a49 6403 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
8ed67789
DL		 6404};
6405
2f460933
WC		 6406void __init ip6_route_init_special_entries(void)
6407{
6408 /* Registering of the loopback is done before this portion of code,
6409 * the loopback reference in rt6_info will not be taken, do it
6410 * manually for init_net */
1cf844c7 6411 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
2f460933
WC		 6412 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6413 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6414 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6415 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6416 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6417 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6418 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6419 #endif
6420}
6421
138d0be3
YS		 6422#if IS_BUILTIN(CONFIG_IPV6)
6423#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6424DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6425
15172a46
YS		 6426static const struct bpf_iter_reg ipv6_route_reg_info = {
6427 .target = "ipv6_route",
6428 .seq_ops = &ipv6_route_seq_ops,
6429 .init_seq_private = bpf_iter_init_seq_net,
6430 .fini_seq_private = bpf_iter_fini_seq_net,
6431 .seq_priv_size = sizeof(struct ipv6_route_iter),
3c32cc1b
YS		 6432 .ctx_arg_info_size = 1,
6433 .ctx_arg_info = {
6434 { offsetof(struct bpf_iter__ipv6_route, rt),
6435 PTR_TO_BTF_ID_OR_NULL },
6436 },
15172a46
YS		 6437};
6438
138d0be3
YS		 6439static int __init bpf_iter_register(void)
6440{
15172a46 6441 return bpf_iter_reg_target(&ipv6_route_reg_info);
138d0be3
YS		 6442}
6443
6444static void bpf_iter_unregister(void)
6445{
ab2ee4fc 6446 bpf_iter_unreg_target(&ipv6_route_reg_info);
138d0be3
YS		 6447}
6448#endif
6449#endif
6450
433d49c3 6451int __init ip6_route_init(void)
1da177e4 6452{
433d49c3 6453 int ret;
8d0b94af 6454 int cpu;
433d49c3 6455
9a7ec3a9
DL		 6456 ret = -ENOMEM;
6457 ip6_dst_ops_template.kmem_cachep =
e5d679f3 6458 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
f845ab6b 6459 SLAB_HWCACHE_ALIGN, NULL);
9a7ec3a9 6460 if (!ip6_dst_ops_template.kmem_cachep)
c19a28e1 6461 goto out;
14e50e57 6462
fc66f95c 6463 ret = dst_entries_init(&ip6_dst_blackhole_ops);
8ed67789 6464 if (ret)
bdb3289f 6465 goto out_kmem_cache;
bdb3289f 6466
c3426b47
DM		 6467 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6468 if (ret)
e8803b6c 6469 goto out_dst_entries;
2a0c451a 6470
7e52b33b
DM		 6471 ret = register_pernet_subsys(&ip6_route_net_ops);
6472 if (ret)
6473 goto out_register_inetpeer;
c3426b47 6474
5dc121e9
AE		 6475 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6476
e8803b6c 6477 ret = fib6_init();
433d49c3 6478 if (ret)
8ed67789 6479 goto out_register_subsys;
433d49c3 6480
433d49c3
DL		 6481 ret = xfrm6_init();
6482 if (ret)
e8803b6c 6483 goto out_fib6_init;
c35b7e72 6484
433d49c3
DL		 6485 ret = fib6_rules_init();
6486 if (ret)
6487 goto xfrm6_init;
7e5449c2 6488
d189634e
TG		 6489 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6490 if (ret)
6491 goto fib6_rules_init;
6492
16feebcf
FW		 6493 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6494 inet6_rtm_newroute, NULL, 0);
6495 if (ret < 0)
6496 goto out_register_late_subsys;
6497
6498 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6499 inet6_rtm_delroute, NULL, 0);
6500 if (ret < 0)
6501 goto out_register_late_subsys;
6502
6503 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6504 inet6_rtm_getroute, NULL,
6505 RTNL_FLAG_DOIT_UNLOCKED);
6506 if (ret < 0)
d189634e 6507 goto out_register_late_subsys;
c127ea2c 6508
8ed67789 6509 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
cdb18761 6510 if (ret)
d189634e 6511 goto out_register_late_subsys;
8ed67789 6512
138d0be3
YS		 6513#if IS_BUILTIN(CONFIG_IPV6)
6514#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6515 ret = bpf_iter_register();
6516 if (ret)
6517 goto out_register_late_subsys;
6518#endif
6519#endif
6520
8d0b94af
MKL		 6521 for_each_possible_cpu(cpu) {
6522 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6523
6524 INIT_LIST_HEAD(&ul->head);
6525 spin_lock_init(&ul->lock);
6526 }
6527
433d49c3
DL		 6528out:
6529 return ret;
6530
d189634e 6531out_register_late_subsys:
16feebcf 6532 rtnl_unregister_all(PF_INET6);
d189634e 6533 unregister_pernet_subsys(&ip6_route_net_late_ops);
433d49c3 6534fib6_rules_init:
433d49c3
DL		 6535 fib6_rules_cleanup();
6536xfrm6_init:
433d49c3 6537 xfrm6_fini();
2a0c451a
TG		 6538out_fib6_init:
6539 fib6_gc_cleanup();
8ed67789
DL		 6540out_register_subsys:
6541 unregister_pernet_subsys(&ip6_route_net_ops);
7e52b33b
DM		 6542out_register_inetpeer:
6543 unregister_pernet_subsys(&ipv6_inetpeer_ops);
fc66f95c
ED		 6544out_dst_entries:
6545 dst_entries_destroy(&ip6_dst_blackhole_ops);
433d49c3 6546out_kmem_cache:
f2fc6a54 6547 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
433d49c3 6548 goto out;
1da177e4
LT		 6549}
6550
6551void ip6_route_cleanup(void)
6552{
138d0be3
YS		 6553#if IS_BUILTIN(CONFIG_IPV6)
6554#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6555 bpf_iter_unregister();
6556#endif
6557#endif
8ed67789 6558 unregister_netdevice_notifier(&ip6_route_dev_notifier);
d189634e 6559 unregister_pernet_subsys(&ip6_route_net_late_ops);
101367c2 6560 fib6_rules_cleanup();
1da177e4 6561 xfrm6_fini();
1da177e4 6562 fib6_gc_cleanup();
c3426b47 6563 unregister_pernet_subsys(&ipv6_inetpeer_ops);
8ed67789 6564 unregister_pernet_subsys(&ip6_route_net_ops);
41bb78b4 6565 dst_entries_destroy(&ip6_dst_blackhole_ops);
f2fc6a54 6566 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
1da177e4 6567}
Linux 6.x block layer and io_uring tree(s)