projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
tipc: fix a double free in tipc_enable_bearer()
[linux-block.git] / net / ipv6 / route.c
CommitLineData
1da177e4
LT		 1/*
2 * Linux INET6 implementation
3 * FIB front-end.
4 *
5 * Authors:
1ab1457c 6 * Pedro Roque <roque@di.fc.ul.pt>
1da177e4 7 *
1da177e4
LT		 8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14/* Changes:
15 *
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
c0bece9f
YH		 23 * Ville Nuorvala
24 * Fixed routing subtrees.
1da177e4
LT		 25 */
26
f3213831
JP		 27#define pr_fmt(fmt) "IPv6: " fmt
28
4fc268d2 29#include <linux/capability.h>
1da177e4 30#include <linux/errno.h>
bc3b2d7f 31#include <linux/export.h>
1da177e4
LT		 32#include <linux/types.h>
33#include <linux/times.h>
34#include <linux/socket.h>
35#include <linux/sockios.h>
36#include <linux/net.h>
37#include <linux/route.h>
38#include <linux/netdevice.h>
39#include <linux/in6.h>
7bc570c8 40#include <linux/mroute6.h>
1da177e4 41#include <linux/init.h>
1da177e4 42#include <linux/if_arp.h>
1da177e4
LT		 43#include <linux/proc_fs.h>
44#include <linux/seq_file.h>
5b7c931d 45#include <linux/nsproxy.h>
5a0e3ad6 46#include <linux/slab.h>
35732d01 47#include <linux/jhash.h>
457c4cbc 48#include <net/net_namespace.h>
1da177e4
LT		 49#include <net/snmp.h>
50#include <net/ipv6.h>
51#include <net/ip6_fib.h>
52#include <net/ip6_route.h>
53#include <net/ndisc.h>
54#include <net/addrconf.h>
55#include <net/tcp.h>
56#include <linux/rtnetlink.h>
57#include <net/dst.h>
904af04d 58#include <net/dst_metadata.h>
1da177e4 59#include <net/xfrm.h>
8d71740c 60#include <net/netevent.h>
21713ebc 61#include <net/netlink.h>
51ebd318 62#include <net/nexthop.h>
19e42e45 63#include <net/lwtunnel.h>
904af04d 64#include <net/ip_tunnels.h>
ca254490 65#include <net/l3mdev.h>
eacb9384 66#include <net/ip.h>
7c0f6ba6 67#include <linux/uaccess.h>
1da177e4
LT		 68
69#ifdef CONFIG_SYSCTL
70#include <linux/sysctl.h>
71#endif
72
30d444d3
DA		 73static int ip6_rt_type_to_error(u8 fib6_type);
74
75#define CREATE_TRACE_POINTS
76#include <trace/events/fib6.h>
77EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
78#undef CREATE_TRACE_POINTS
79
afc154e9 80enum rt6_nud_state {
7e980569
JB		 81 RT6_NUD_FAIL_HARD = -3,
82 RT6_NUD_FAIL_PROBE = -2,
83 RT6_NUD_FAIL_DO_RR = -1,
afc154e9
HFS		 84 RT6_NUD_SUCCEED = 1
85};
86
1da177e4 87static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
0dbaee3b 88static unsigned int ip6_default_advmss(const struct dst_entry *dst);
ebb762f2 89static unsigned int ip6_mtu(const struct dst_entry *dst);
1da177e4
LT		 90static struct dst_entry *ip6_negative_advice(struct dst_entry *);
91static void ip6_dst_destroy(struct dst_entry *);
92static void ip6_dst_ifdown(struct dst_entry *,
93 struct net_device *dev, int how);
569d3645 94static int ip6_dst_gc(struct dst_ops *ops);
1da177e4
LT		 95
96static int ip6_pkt_discard(struct sk_buff *skb);
ede2059d 97static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
7150aede 98static int ip6_pkt_prohibit(struct sk_buff *skb);
ede2059d 99static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1da177e4 100static void ip6_link_failure(struct sk_buff *skb);
6700c270
DM		 101static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
102 struct sk_buff *skb, u32 mtu);
103static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
104 struct sk_buff *skb);
8d1c802b
DA		 105static int rt6_score_route(struct fib6_info *rt, int oif, int strict);
106static size_t rt6_nlmsg_size(struct fib6_info *rt);
d4ead6b3 107static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 108 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 109 struct in6_addr *dest, struct in6_addr *src,
16a16cd3
DA		 110 int iif, int type, u32 portid, u32 seq,
111 unsigned int flags);
8d1c802b 112static struct rt6_info *rt6_find_cached_rt(struct fib6_info *rt,
35732d01
WW		 113 struct in6_addr *daddr,
114 struct in6_addr *saddr);
1da177e4 115
70ceb4f5 116#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 117static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 118 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 119 const struct in6_addr *gwaddr,
120 struct net_device *dev,
95c96174 121 unsigned int pref);
8d1c802b 122static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 123 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 124 const struct in6_addr *gwaddr,
125 struct net_device *dev);
70ceb4f5
YH		 126#endif
127
8d0b94af
MKL		 128struct uncached_list {
129 spinlock_t lock;
130 struct list_head head;
131};
132
133static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
134
510c321b 135void rt6_uncached_list_add(struct rt6_info *rt)
8d0b94af
MKL		 136{
137 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
138
8d0b94af
MKL		 139 rt->rt6i_uncached_list = ul;
140
141 spin_lock_bh(&ul->lock);
142 list_add_tail(&rt->rt6i_uncached, &ul->head);
143 spin_unlock_bh(&ul->lock);
144}
145
510c321b 146void rt6_uncached_list_del(struct rt6_info *rt)
8d0b94af
MKL		 147{
148 if (!list_empty(&rt->rt6i_uncached)) {
149 struct uncached_list *ul = rt->rt6i_uncached_list;
81eb8447 150 struct net *net = dev_net(rt->dst.dev);
8d0b94af
MKL		 151
152 spin_lock_bh(&ul->lock);
153 list_del(&rt->rt6i_uncached);
81eb8447 154 atomic_dec(&net->ipv6.rt6_stats->fib_rt_uncache);
8d0b94af
MKL		 155 spin_unlock_bh(&ul->lock);
156 }
157}
158
159static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
160{
161 struct net_device *loopback_dev = net->loopback_dev;
162 int cpu;
163
e332bc67
EB		 164 if (dev == loopback_dev)
165 return;
166
8d0b94af
MKL		 167 for_each_possible_cpu(cpu) {
168 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
169 struct rt6_info *rt;
170
171 spin_lock_bh(&ul->lock);
172 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
173 struct inet6_dev *rt_idev = rt->rt6i_idev;
174 struct net_device *rt_dev = rt->dst.dev;
175
e332bc67 176 if (rt_idev->dev == dev) {
8d0b94af
MKL		 177 rt->rt6i_idev = in6_dev_get(loopback_dev);
178 in6_dev_put(rt_idev);
179 }
180
e332bc67 181 if (rt_dev == dev) {
8d0b94af
MKL		 182 rt->dst.dev = loopback_dev;
183 dev_hold(rt->dst.dev);
184 dev_put(rt_dev);
185 }
186 }
187 spin_unlock_bh(&ul->lock);
188 }
189}
190
f8a1b43b 191static inline const void *choose_neigh_daddr(const struct in6_addr *p,
f894cbf8
DM		 192 struct sk_buff *skb,
193 const void *daddr)
39232973 194{
a7563f34 195 if (!ipv6_addr_any(p))
39232973 196 return (const void *) p;
f894cbf8
DM		 197 else if (skb)
198 return &ipv6_hdr(skb)->daddr;
39232973
DM		 199 return daddr;
200}
201
f8a1b43b
DA		 202struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
203 struct net_device *dev,
204 struct sk_buff *skb,
205 const void *daddr)
d3aaeb38 206{
39232973
DM		 207 struct neighbour *n;
208
f8a1b43b
DA		 209 daddr = choose_neigh_daddr(gw, skb, daddr);
210 n = __ipv6_neigh_lookup(dev, daddr);
f83c7790
DM		 211 if (n)
212 return n;
f8a1b43b
DA		 213 return neigh_create(&nd_tbl, daddr, dev);
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
222 return ip6_neigh_lookup(&rt->rt6i_gateway, dst->dev, skb, daddr);
f83c7790
DM		 223}
224
63fca65d
JA		 225static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
226{
227 struct net_device *dev = dst->dev;
228 struct rt6_info *rt = (struct rt6_info *)dst;
229
f8a1b43b 230 daddr = choose_neigh_daddr(&rt->rt6i_gateway, NULL, daddr);
63fca65d
JA		 231 if (!daddr)
232 return;
233 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
234 return;
235 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
236 return;
237 __ipv6_confirm_neigh(dev, daddr);
238}
239
9a7ec3a9 240static struct dst_ops ip6_dst_ops_template = {
1da177e4 241 .family = AF_INET6,
1da177e4
LT		 242 .gc = ip6_dst_gc,
243 .gc_thresh = 1024,
244 .check = ip6_dst_check,
0dbaee3b 245 .default_advmss = ip6_default_advmss,
ebb762f2 246 .mtu = ip6_mtu,
d4ead6b3 247 .cow_metrics = dst_cow_metrics_generic,
1da177e4
LT		 248 .destroy = ip6_dst_destroy,
249 .ifdown = ip6_dst_ifdown,
250 .negative_advice = ip6_negative_advice,
251 .link_failure = ip6_link_failure,
252 .update_pmtu = ip6_rt_update_pmtu,
6e157b6a 253 .redirect = rt6_do_redirect,
9f8955cc 254 .local_out = __ip6_local_out,
f8a1b43b 255 .neigh_lookup = ip6_dst_neigh_lookup,
63fca65d 256 .confirm_neigh = ip6_confirm_neigh,
1da177e4
LT		 257};
258
ebb762f2 259static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
ec831ea7 260{
618f9bc7
SK		 261 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
262
263 return mtu ? : dst->dev->mtu;
ec831ea7
RD		 264}
265
6700c270
DM		 266static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
267 struct sk_buff *skb, u32 mtu)
14e50e57
DM		 268{
269}
270
6700c270
DM		 271static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
272 struct sk_buff *skb)
b587ee3b
DM		 273{
274}
275
14e50e57
DM		 276static struct dst_ops ip6_dst_blackhole_ops = {
277 .family = AF_INET6,
14e50e57
DM		 278 .destroy = ip6_dst_destroy,
279 .check = ip6_dst_check,
ebb762f2 280 .mtu = ip6_blackhole_mtu,
214f45c9 281 .default_advmss = ip6_default_advmss,
14e50e57 282 .update_pmtu = ip6_rt_blackhole_update_pmtu,
b587ee3b 283 .redirect = ip6_rt_blackhole_redirect,
0a1f5962 284 .cow_metrics = dst_cow_metrics_generic,
f8a1b43b 285 .neigh_lookup = ip6_dst_neigh_lookup,
14e50e57
DM		 286};
287
62fa8a84 288static const u32 ip6_template_metrics[RTAX_MAX] = {
14edd87d 289 [RTAX_HOPLIMIT - 1] = 0,
62fa8a84
DM		 290};
291
8d1c802b 292static const struct fib6_info fib6_null_entry_template = {
93c2fb25
DA		 293 .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
294 .fib6_protocol = RTPROT_KERNEL,
295 .fib6_metric = ~(u32)0,
296 .fib6_ref = ATOMIC_INIT(1),
421842ed
DA		 297 .fib6_type = RTN_UNREACHABLE,
298 .fib6_metrics = (struct dst_metrics *)&dst_default_metrics,
299};
300
fb0af4c7 301static const struct rt6_info ip6_null_entry_template = {
d8d1f30b
CG		 302 .dst = {
303 .__refcnt = ATOMIC_INIT(1),
304 .__use = 1,
2c20cbd7 305 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 306 .error = -ENETUNREACH,
d8d1f30b
CG		 307 .input = ip6_pkt_discard,
308 .output = ip6_pkt_discard_out,
1da177e4
LT		 309 },
310 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
1da177e4
LT		 311};
312
101367c2
TG		 313#ifdef CONFIG_IPV6_MULTIPLE_TABLES
314
fb0af4c7 315static const struct rt6_info ip6_prohibit_entry_template = {
d8d1f30b
CG		 316 .dst = {
317 .__refcnt = ATOMIC_INIT(1),
318 .__use = 1,
2c20cbd7 319 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 320 .error = -EACCES,
d8d1f30b
CG		 321 .input = ip6_pkt_prohibit,
322 .output = ip6_pkt_prohibit_out,
101367c2
TG		 323 },
324 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 325};
326
fb0af4c7 327static const struct rt6_info ip6_blk_hole_entry_template = {
d8d1f30b
CG		 328 .dst = {
329 .__refcnt = ATOMIC_INIT(1),
330 .__use = 1,
2c20cbd7 331 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 332 .error = -EINVAL,
d8d1f30b 333 .input = dst_discard,
ede2059d 334 .output = dst_discard_out,
101367c2
TG		 335 },
336 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 337};
338
339#endif
340
ebfa45f0
MKL		 341static void rt6_info_init(struct rt6_info *rt)
342{
343 struct dst_entry *dst = &rt->dst;
344
345 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
ebfa45f0
MKL		 346 INIT_LIST_HEAD(&rt->rt6i_uncached);
347}
348
1da177e4 349/* allocate dst with ip6_dst_ops */
93531c67
DA		 350struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
351 int flags)
1da177e4 352{
97bab73f 353 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
b2a9c0ed 354 1, DST_OBSOLETE_FORCE_CHK, flags);
cf911662 355
81eb8447 356 if (rt) {
ebfa45f0 357 rt6_info_init(rt);
81eb8447
WW		 358 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
359 }
8104891b 360
cf911662 361 return rt;
1da177e4 362}
9ab179d8 363EXPORT_SYMBOL(ip6_dst_alloc);
d52d3997 364
1da177e4
LT		 365static void ip6_dst_destroy(struct dst_entry *dst)
366{
367 struct rt6_info *rt = (struct rt6_info *)dst;
a68886a6 368 struct fib6_info *from;
8d0b94af 369 struct inet6_dev *idev;
1da177e4 370
1620a336 371 ip_dst_metrics_put(dst);
8d0b94af
MKL		 372 rt6_uncached_list_del(rt);
373
374 idev = rt->rt6i_idev;
38308473 375 if (idev) {
1da177e4
LT		 376 rt->rt6i_idev = NULL;
377 in6_dev_put(idev);
1ab1457c 378 }
1716a961 379
a68886a6
DA		 380 rcu_read_lock();
381 from = rcu_dereference(rt->from);
382 rcu_assign_pointer(rt->from, NULL);
93531c67 383 fib6_info_release(from);
a68886a6 384 rcu_read_unlock();
b3419363
DM		 385}
386
1da177e4
LT		 387static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
388 int how)
389{
390 struct rt6_info *rt = (struct rt6_info *)dst;
391 struct inet6_dev *idev = rt->rt6i_idev;
5a3e55d6 392 struct net_device *loopback_dev =
c346dca1 393 dev_net(dev)->loopback_dev;
1da177e4 394
e5645f51
WW		 395 if (idev && idev->dev != loopback_dev) {
396 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
397 if (loopback_idev) {
398 rt->rt6i_idev = loopback_idev;
399 in6_dev_put(idev);
97cac082 400 }
1da177e4
LT		 401 }
402}
403
5973fb1e
MKL		 404static bool __rt6_check_expired(const struct rt6_info *rt)
405{
406 if (rt->rt6i_flags & RTF_EXPIRES)
407 return time_after(jiffies, rt->dst.expires);
408 else
409 return false;
410}
411
a50feda5 412static bool rt6_check_expired(const struct rt6_info *rt)
1da177e4 413{
a68886a6
DA		 414 struct fib6_info *from;
415
416 from = rcu_dereference(rt->from);
417
1716a961
G		 418 if (rt->rt6i_flags & RTF_EXPIRES) {
419 if (time_after(jiffies, rt->dst.expires))
a50feda5 420 return true;
a68886a6 421 } else if (from) {
1e2ea8ad 422 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
a68886a6 423 fib6_check_expired(from);
1716a961 424 }
a50feda5 425 return false;
1da177e4
LT		 426}
427
3b290a31
DA		 428struct fib6_info *fib6_multipath_select(const struct net *net,
429 struct fib6_info *match,
430 struct flowi6 *fl6, int oif,
431 const struct sk_buff *skb,
432 int strict)
51ebd318 433{
8d1c802b 434 struct fib6_info *sibling, *next_sibling;
51ebd318 435
b673d6cc
JS		 436 /* We might have already computed the hash for ICMPv6 errors. In such
437 * case it will always be non-zero. Otherwise now is the time to do it.
438 */
439 if (!fl6->mp_hash)
b4bac172 440 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
b673d6cc 441
5e670d84 442 if (fl6->mp_hash <= atomic_read(&match->fib6_nh.nh_upper_bound))
3d709f69
IS		 443 return match;
444
93c2fb25
DA		 445 list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
446 fib6_siblings) {
5e670d84
DA		 447 int nh_upper_bound;
448
449 nh_upper_bound = atomic_read(&sibling->fib6_nh.nh_upper_bound);
450 if (fl6->mp_hash > nh_upper_bound)
3d709f69
IS		 451 continue;
452 if (rt6_score_route(sibling, oif, strict) < 0)
453 break;
454 match = sibling;
455 break;
456 }
457
51ebd318
ND		 458 return match;
459}
460
1da177e4 461/*
66f5d6ce 462 * Route lookup. rcu_read_lock() should be held.
1da177e4
LT		 463 */
464
8d1c802b
DA		 465static inline struct fib6_info *rt6_device_match(struct net *net,
466 struct fib6_info *rt,
b71d1d42 467 const struct in6_addr *saddr,
1da177e4 468 int oif,
d420895e 469 int flags)
1da177e4 470{
8d1c802b 471 struct fib6_info *sprt;
1da177e4 472
5e670d84
DA		 473 if (!oif && ipv6_addr_any(saddr) &&
474 !(rt->fib6_nh.nh_flags & RTNH_F_DEAD))
8067bb8c 475 return rt;
dd3abc4e 476
8fb11a9a 477 for (sprt = rt; sprt; sprt = rcu_dereference(sprt->fib6_next)) {
5e670d84 478 const struct net_device *dev = sprt->fib6_nh.nh_dev;
dd3abc4e 479
5e670d84 480 if (sprt->fib6_nh.nh_flags & RTNH_F_DEAD)
8067bb8c
IS		 481 continue;
482
dd3abc4e 483 if (oif) {
1da177e4
LT		 484 if (dev->ifindex == oif)
485 return sprt;
dd3abc4e
YH		 486 } else {
487 if (ipv6_chk_addr(net, saddr, dev,
488 flags & RT6_LOOKUP_F_IFACE))
489 return sprt;
1da177e4 490 }
dd3abc4e 491 }
1da177e4 492
eea68cd3
DA		 493 if (oif && flags & RT6_LOOKUP_F_IFACE)
494 return net->ipv6.fib6_null_entry;
8067bb8c 495
421842ed 496 return rt->fib6_nh.nh_flags & RTNH_F_DEAD ? net->ipv6.fib6_null_entry : rt;
1da177e4
LT		 497}
498
27097255 499#ifdef CONFIG_IPV6_ROUTER_PREF
c2f17e82
HFS		 500struct __rt6_probe_work {
501 struct work_struct work;
502 struct in6_addr target;
503 struct net_device *dev;
504};
505
506static void rt6_probe_deferred(struct work_struct *w)
507{
508 struct in6_addr mcaddr;
509 struct __rt6_probe_work *work =
510 container_of(w, struct __rt6_probe_work, work);
511
512 addrconf_addr_solict_mult(&work->target, &mcaddr);
adc176c5 513 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
c2f17e82 514 dev_put(work->dev);
662f5533 515 kfree(work);
c2f17e82
HFS		 516}
517
8d1c802b 518static void rt6_probe(struct fib6_info *rt)
27097255 519{
f547fac6 520 struct __rt6_probe_work *work = NULL;
5e670d84 521 const struct in6_addr *nh_gw;
f2c31e32 522 struct neighbour *neigh;
5e670d84 523 struct net_device *dev;
f547fac6 524 struct inet6_dev *idev;
5e670d84 525
27097255
YH		 526 /*
527 * Okay, this does not seem to be appropriate
528 * for now, however, we need to check if it
529 * is really so; aka Router Reachability Probing.
530 *
531 * Router Reachability Probe MUST be rate-limited
532 * to no more than one per minute.
533 */
93c2fb25 534 if (!rt || !(rt->fib6_flags & RTF_GATEWAY))
7ff74a59 535 return;
5e670d84
DA		 536
537 nh_gw = &rt->fib6_nh.nh_gw;
538 dev = rt->fib6_nh.nh_dev;
2152caea 539 rcu_read_lock_bh();
f547fac6 540 idev = __in6_dev_get(dev);
5e670d84 541 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
2152caea 542 if (neigh) {
8d6c31bf
MKL		 543 if (neigh->nud_state & NUD_VALID)
544 goto out;
545
2152caea 546 write_lock(&neigh->lock);
990edb42
MKL		 547 if (!(neigh->nud_state & NUD_VALID) &&
548 time_after(jiffies,
dcd1f572 549 neigh->updated + idev->cnf.rtr_probe_interval)) {
990edb42
MKL		 550 work = kmalloc(sizeof(*work), GFP_ATOMIC);
551 if (work)
552 __neigh_set_probe_once(neigh);
c2f17e82 553 }
2152caea 554 write_unlock(&neigh->lock);
f547fac6
SD		 555 } else if (time_after(jiffies, rt->last_probe +
556 idev->cnf.rtr_probe_interval)) {
990edb42 557 work = kmalloc(sizeof(*work), GFP_ATOMIC);
f2c31e32 558 }
990edb42
MKL		 559
560 if (work) {
f547fac6 561 rt->last_probe = jiffies;
990edb42 562 INIT_WORK(&work->work, rt6_probe_deferred);
5e670d84
DA		 563 work->target = *nh_gw;
564 dev_hold(dev);
565 work->dev = dev;
990edb42
MKL		 566 schedule_work(&work->work);
567 }
568
8d6c31bf 569out:
2152caea 570 rcu_read_unlock_bh();
27097255
YH		 571}
572#else
8d1c802b 573static inline void rt6_probe(struct fib6_info *rt)
27097255 574{
27097255
YH		 575}
576#endif
577
1da177e4 578/*
554cfb7e 579 * Default Router Selection (RFC 2461 6.3.6)
1da177e4 580 */
8d1c802b 581static inline int rt6_check_dev(struct fib6_info *rt, int oif)
554cfb7e 582{
5e670d84
DA		 583 const struct net_device *dev = rt->fib6_nh.nh_dev;
584
161980f4 585 if (!oif || dev->ifindex == oif)
554cfb7e 586 return 2;
161980f4 587 return 0;
554cfb7e 588}
1da177e4 589
8d1c802b 590static inline enum rt6_nud_state rt6_check_neigh(struct fib6_info *rt)
1da177e4 591{
afc154e9 592 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
5e670d84 593 struct neighbour *neigh;
f2c31e32 594
93c2fb25
DA		 595 if (rt->fib6_flags & RTF_NONEXTHOP ||
596 !(rt->fib6_flags & RTF_GATEWAY))
afc154e9 597 return RT6_NUD_SUCCEED;
145a3621
YH		 598
599 rcu_read_lock_bh();
5e670d84
DA		 600 neigh = __ipv6_neigh_lookup_noref(rt->fib6_nh.nh_dev,
601 &rt->fib6_nh.nh_gw);
145a3621
YH		 602 if (neigh) {
603 read_lock(&neigh->lock);
554cfb7e 604 if (neigh->nud_state & NUD_VALID)
afc154e9 605 ret = RT6_NUD_SUCCEED;
398bcbeb 606#ifdef CONFIG_IPV6_ROUTER_PREF
a5a81f0b 607 else if (!(neigh->nud_state & NUD_FAILED))
afc154e9 608 ret = RT6_NUD_SUCCEED;
7e980569
JB		 609 else
610 ret = RT6_NUD_FAIL_PROBE;
398bcbeb 611#endif
145a3621 612 read_unlock(&neigh->lock);
afc154e9
HFS		 613 } else {
614 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
7e980569 615 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
a5a81f0b 616 }
145a3621
YH		 617 rcu_read_unlock_bh();
618
a5a81f0b 619 return ret;
1da177e4
LT		 620}
621
8d1c802b 622static int rt6_score_route(struct fib6_info *rt, int oif, int strict)
1da177e4 623{
a5a81f0b 624 int m;
1ab1457c 625
4d0c5911 626 m = rt6_check_dev(rt, oif);
77d16f45 627 if (!m && (strict & RT6_LOOKUP_F_IFACE))
afc154e9 628 return RT6_NUD_FAIL_HARD;
ebacaaa0 629#ifdef CONFIG_IPV6_ROUTER_PREF
93c2fb25 630 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->fib6_flags)) << 2;
ebacaaa0 631#endif
afc154e9
HFS		 632 if (strict & RT6_LOOKUP_F_REACHABLE) {
633 int n = rt6_check_neigh(rt);
634 if (n < 0)
635 return n;
636 }
554cfb7e
YH		 637 return m;
638}
639
dcd1f572
DA		 640/* called with rc_read_lock held */
641static inline bool fib6_ignore_linkdown(const struct fib6_info *f6i)
642{
643 const struct net_device *dev = fib6_info_nh_dev(f6i);
644 bool rc = false;
645
646 if (dev) {
647 const struct inet6_dev *idev = __in6_dev_get(dev);
648
649 rc = !!idev->cnf.ignore_routes_with_linkdown;
650 }
651
652 return rc;
653}
654
8d1c802b
DA		 655static struct fib6_info *find_match(struct fib6_info *rt, int oif, int strict,
656 int *mpri, struct fib6_info *match,
afc154e9 657 bool *do_rr)
554cfb7e 658{
f11e6659 659 int m;
afc154e9 660 bool match_do_rr = false;
35103d11 661
5e670d84 662 if (rt->fib6_nh.nh_flags & RTNH_F_DEAD)
8067bb8c
IS		 663 goto out;
664
dcd1f572 665 if (fib6_ignore_linkdown(rt) &&
5e670d84 666 rt->fib6_nh.nh_flags & RTNH_F_LINKDOWN &&
d5d32e4b 667 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
35103d11 668 goto out;
f11e6659 669
14895687 670 if (fib6_check_expired(rt))
f11e6659
DM		 671 goto out;
672
673 m = rt6_score_route(rt, oif, strict);
7e980569 674 if (m == RT6_NUD_FAIL_DO_RR) {
afc154e9
HFS		 675 match_do_rr = true;
676 m = 0; /* lowest valid score */
7e980569 677 } else if (m == RT6_NUD_FAIL_HARD) {
f11e6659 678 goto out;
afc154e9
HFS		 679 }
680
681 if (strict & RT6_LOOKUP_F_REACHABLE)
682 rt6_probe(rt);
f11e6659 683
7e980569 684 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
f11e6659 685 if (m > *mpri) {
afc154e9 686 *do_rr = match_do_rr;
f11e6659
DM		 687 *mpri = m;
688 match = rt;
f11e6659 689 }
f11e6659
DM		 690out:
691 return match;
692}
693
8d1c802b
DA		 694static struct fib6_info *find_rr_leaf(struct fib6_node *fn,
695 struct fib6_info *leaf,
696 struct fib6_info *rr_head,
afc154e9
HFS		 697 u32 metric, int oif, int strict,
698 bool *do_rr)
f11e6659 699{
8d1c802b 700 struct fib6_info *rt, *match, *cont;
554cfb7e 701 int mpri = -1;
1da177e4 702
f11e6659 703 match = NULL;
9fbdcfaf 704 cont = NULL;
8fb11a9a 705 for (rt = rr_head; rt; rt = rcu_dereference(rt->fib6_next)) {
93c2fb25 706 if (rt->fib6_metric != metric) {
9fbdcfaf
SK		 707 cont = rt;
708 break;
709 }
710
711 match = find_match(rt, oif, strict, &mpri, match, do_rr);
712 }
713
66f5d6ce 714 for (rt = leaf; rt && rt != rr_head;
8fb11a9a 715 rt = rcu_dereference(rt->fib6_next)) {
93c2fb25 716 if (rt->fib6_metric != metric) {
9fbdcfaf
SK		 717 cont = rt;
718 break;
719 }
720
afc154e9 721 match = find_match(rt, oif, strict, &mpri, match, do_rr);
9fbdcfaf
SK		 722 }
723
724 if (match || !cont)
725 return match;
726
8fb11a9a 727 for (rt = cont; rt; rt = rcu_dereference(rt->fib6_next))
afc154e9 728 match = find_match(rt, oif, strict, &mpri, match, do_rr);
1da177e4 729
f11e6659
DM		 730 return match;
731}
1da177e4 732
8d1c802b 733static struct fib6_info *rt6_select(struct net *net, struct fib6_node *fn,
8d1040e8 734 int oif, int strict)
f11e6659 735{
8d1c802b
DA		 736 struct fib6_info *leaf = rcu_dereference(fn->leaf);
737 struct fib6_info *match, *rt0;
afc154e9 738 bool do_rr = false;
17ecf590 739 int key_plen;
1da177e4 740
421842ed
DA		 741 if (!leaf || leaf == net->ipv6.fib6_null_entry)
742 return net->ipv6.fib6_null_entry;
8d1040e8 743
66f5d6ce 744 rt0 = rcu_dereference(fn->rr_ptr);
f11e6659 745 if (!rt0)
66f5d6ce 746 rt0 = leaf;
1da177e4 747
17ecf590
WW		 748 /* Double check to make sure fn is not an intermediate node
749 * and fn->leaf does not points to its child's leaf
750 * (This might happen if all routes under fn are deleted from
751 * the tree and fib6_repair_tree() is called on the node.)
752 */
93c2fb25 753 key_plen = rt0->fib6_dst.plen;
17ecf590 754#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 755 if (rt0->fib6_src.plen)
756 key_plen = rt0->fib6_src.plen;
17ecf590
WW		 757#endif
758 if (fn->fn_bit != key_plen)
421842ed 759 return net->ipv6.fib6_null_entry;
17ecf590 760
93c2fb25 761 match = find_rr_leaf(fn, leaf, rt0, rt0->fib6_metric, oif, strict,
afc154e9 762 &do_rr);
1da177e4 763
afc154e9 764 if (do_rr) {
8fb11a9a 765 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
f11e6659 766
554cfb7e 767 /* no entries matched; do round-robin */
93c2fb25 768 if (!next || next->fib6_metric != rt0->fib6_metric)
8d1040e8 769 next = leaf;
f11e6659 770
66f5d6ce 771 if (next != rt0) {
93c2fb25 772 spin_lock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 773 /* make sure next is not being deleted from the tree */
93c2fb25 774 if (next->fib6_node)
66f5d6ce 775 rcu_assign_pointer(fn->rr_ptr, next);
93c2fb25 776 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 777 }
1da177e4 778 }
1da177e4 779
421842ed 780 return match ? match : net->ipv6.fib6_null_entry;
1da177e4
LT		 781}
782
8d1c802b 783static bool rt6_is_gw_or_nonexthop(const struct fib6_info *rt)
8b9df265 784{
93c2fb25 785 return (rt->fib6_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
8b9df265
MKL		 786}
787
70ceb4f5
YH		 788#ifdef CONFIG_IPV6_ROUTE_INFO
789int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
b71d1d42 790 const struct in6_addr *gwaddr)
70ceb4f5 791{
c346dca1 792 struct net *net = dev_net(dev);
70ceb4f5
YH		 793 struct route_info *rinfo = (struct route_info *) opt;
794 struct in6_addr prefix_buf, *prefix;
795 unsigned int pref;
4bed72e4 796 unsigned long lifetime;
8d1c802b 797 struct fib6_info *rt;
70ceb4f5
YH		 798
799 if (len < sizeof(struct route_info)) {
800 return -EINVAL;
801 }
802
803 /* Sanity check for prefix_len and length */
804 if (rinfo->length > 3) {
805 return -EINVAL;
806 } else if (rinfo->prefix_len > 128) {
807 return -EINVAL;
808 } else if (rinfo->prefix_len > 64) {
809 if (rinfo->length < 2) {
810 return -EINVAL;
811 }
812 } else if (rinfo->prefix_len > 0) {
813 if (rinfo->length < 1) {
814 return -EINVAL;
815 }
816 }
817
818 pref = rinfo->route_pref;
819 if (pref == ICMPV6_ROUTER_PREF_INVALID)
3933fc95 820 return -EINVAL;
70ceb4f5 821
4bed72e4 822 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
70ceb4f5
YH		 823
824 if (rinfo->length == 3)
825 prefix = (struct in6_addr *)rinfo->prefix;
826 else {
827 /* this function is safe */
828 ipv6_addr_prefix(&prefix_buf,
829 (struct in6_addr *)rinfo->prefix,
830 rinfo->prefix_len);
831 prefix = &prefix_buf;
832 }
833
f104a567 834 if (rinfo->prefix_len == 0)
afb1d4b5 835 rt = rt6_get_dflt_router(net, gwaddr, dev);
f104a567
DJ		 836 else
837 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
830218c1 838 gwaddr, dev);
70ceb4f5
YH		 839
840 if (rt && !lifetime) {
afb1d4b5 841 ip6_del_rt(net, rt);
70ceb4f5
YH		 842 rt = NULL;
843 }
844
845 if (!rt && lifetime)
830218c1
DA		 846 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
847 dev, pref);
70ceb4f5 848 else if (rt)
93c2fb25
DA		 849 rt->fib6_flags = RTF_ROUTEINFO |
850 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
70ceb4f5
YH		 851
852 if (rt) {
1716a961 853 if (!addrconf_finite_timeout(lifetime))
14895687 854 fib6_clean_expires(rt);
1716a961 855 else
14895687 856 fib6_set_expires(rt, jiffies + HZ * lifetime);
1716a961 857
93531c67 858 fib6_info_release(rt);
70ceb4f5
YH		 859 }
860 return 0;
861}
862#endif
863
ae90d867
DA		 864/*
865 * Misc support functions
866 */
867
868/* called with rcu_lock held */
8d1c802b 869static struct net_device *ip6_rt_get_dev_rcu(struct fib6_info *rt)
ae90d867 870{
5e670d84 871 struct net_device *dev = rt->fib6_nh.nh_dev;
ae90d867 872
93c2fb25 873 if (rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
ae90d867
DA		 874 /* for copies of local routes, dst->dev needs to be the
875 * device if it is a master device, the master device if
876 * device is enslaved, and the loopback as the default
877 */
878 if (netif_is_l3_slave(dev) &&
93c2fb25 879 !rt6_need_strict(&rt->fib6_dst.addr))
ae90d867
DA		 880 dev = l3mdev_master_dev_rcu(dev);
881 else if (!netif_is_l3_master(dev))
882 dev = dev_net(dev)->loopback_dev;
883 /* last case is netif_is_l3_master(dev) is true in which
884 * case we want dev returned to be dev
885 */
886 }
887
888 return dev;
889}
890
6edb3c96
DA		 891static const int fib6_prop[RTN_MAX + 1] = {
892 [RTN_UNSPEC] = 0,
893 [RTN_UNICAST] = 0,
894 [RTN_LOCAL] = 0,
895 [RTN_BROADCAST] = 0,
896 [RTN_ANYCAST] = 0,
897 [RTN_MULTICAST] = 0,
898 [RTN_BLACKHOLE] = -EINVAL,
899 [RTN_UNREACHABLE] = -EHOSTUNREACH,
900 [RTN_PROHIBIT] = -EACCES,
901 [RTN_THROW] = -EAGAIN,
902 [RTN_NAT] = -EINVAL,
903 [RTN_XRESOLVE] = -EINVAL,
904};
905
906static int ip6_rt_type_to_error(u8 fib6_type)
907{
908 return fib6_prop[fib6_type];
909}
910
8d1c802b 911static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
3b6761d1
DA		 912{
913 unsigned short flags = 0;
914
915 if (rt->dst_nocount)
916 flags |= DST_NOCOUNT;
917 if (rt->dst_nopolicy)
918 flags |= DST_NOPOLICY;
919 if (rt->dst_host)
920 flags |= DST_HOST;
921
922 return flags;
923}
924
8d1c802b 925static void ip6_rt_init_dst_reject(struct rt6_info *rt, struct fib6_info *ort)
6edb3c96
DA		 926{
927 rt->dst.error = ip6_rt_type_to_error(ort->fib6_type);
928
929 switch (ort->fib6_type) {
930 case RTN_BLACKHOLE:
931 rt->dst.output = dst_discard_out;
932 rt->dst.input = dst_discard;
933 break;
934 case RTN_PROHIBIT:
935 rt->dst.output = ip6_pkt_prohibit_out;
936 rt->dst.input = ip6_pkt_prohibit;
937 break;
938 case RTN_THROW:
939 case RTN_UNREACHABLE:
940 default:
941 rt->dst.output = ip6_pkt_discard_out;
942 rt->dst.input = ip6_pkt_discard;
943 break;
944 }
945}
946
8d1c802b 947static void ip6_rt_init_dst(struct rt6_info *rt, struct fib6_info *ort)
6edb3c96 948{
93c2fb25 949 if (ort->fib6_flags & RTF_REJECT) {
6edb3c96
DA		 950 ip6_rt_init_dst_reject(rt, ort);
951 return;
952 }
953
954 rt->dst.error = 0;
955 rt->dst.output = ip6_output;
956
d23c4b63 957 if (ort->fib6_type == RTN_LOCAL || ort->fib6_type == RTN_ANYCAST) {
6edb3c96 958 rt->dst.input = ip6_input;
93c2fb25 959 } else if (ipv6_addr_type(&ort->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
6edb3c96
DA		 960 rt->dst.input = ip6_mc_input;
961 } else {
962 rt->dst.input = ip6_forward;
963 }
964
965 if (ort->fib6_nh.nh_lwtstate) {
966 rt->dst.lwtstate = lwtstate_get(ort->fib6_nh.nh_lwtstate);
967 lwtunnel_set_redirect(&rt->dst);
968 }
969
970 rt->dst.lastuse = jiffies;
971}
972
e873e4b9 973/* Caller must already hold reference to @from */
8d1c802b 974static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
ae90d867 975{
ae90d867 976 rt->rt6i_flags &= ~RTF_EXPIRES;
a68886a6 977 rcu_assign_pointer(rt->from, from);
e1255ed4 978 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
ae90d867
DA		 979}
980
e873e4b9 981/* Caller must already hold reference to @ort */
8d1c802b 982static void ip6_rt_copy_init(struct rt6_info *rt, struct fib6_info *ort)
ae90d867 983{
dcd1f572
DA		 984 struct net_device *dev = fib6_info_nh_dev(ort);
985
6edb3c96
DA		 986 ip6_rt_init_dst(rt, ort);
987
93c2fb25 988 rt->rt6i_dst = ort->fib6_dst;
dcd1f572 989 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
5e670d84 990 rt->rt6i_gateway = ort->fib6_nh.nh_gw;
93c2fb25 991 rt->rt6i_flags = ort->fib6_flags;
ae90d867 992 rt6_set_from(rt, ort);
ae90d867 993#ifdef CONFIG_IPV6_SUBTREES
93c2fb25 994 rt->rt6i_src = ort->fib6_src;
ae90d867 995#endif
ae90d867
DA		 996}
997
a3c00e46
MKL		 998static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
999 struct in6_addr *saddr)
1000{
66f5d6ce 1001 struct fib6_node *pn, *sn;
a3c00e46
MKL		 1002 while (1) {
1003 if (fn->fn_flags & RTN_TL_ROOT)
1004 return NULL;
66f5d6ce
WW		 1005 pn = rcu_dereference(fn->parent);
1006 sn = FIB6_SUBTREE(pn);
1007 if (sn && sn != fn)
6454743b 1008 fn = fib6_node_lookup(sn, NULL, saddr);
a3c00e46
MKL		 1009 else
1010 fn = pn;
1011 if (fn->fn_flags & RTN_RTINFO)
1012 return fn;
1013 }
1014}
c71099ac 1015
d3843fe5
WW		 1016static bool ip6_hold_safe(struct net *net, struct rt6_info **prt,
1017 bool null_fallback)
1018{
1019 struct rt6_info *rt = *prt;
1020
1021 if (dst_hold_safe(&rt->dst))
1022 return true;
1023 if (null_fallback) {
1024 rt = net->ipv6.ip6_null_entry;
1025 dst_hold(&rt->dst);
1026 } else {
1027 rt = NULL;
1028 }
1029 *prt = rt;
1030 return false;
1031}
1032
dec9b0e2 1033/* called with rcu_lock held */
8d1c802b 1034static struct rt6_info *ip6_create_rt_rcu(struct fib6_info *rt)
dec9b0e2 1035{
3b6761d1 1036 unsigned short flags = fib6_info_dst_flags(rt);
dec9b0e2
DA		 1037 struct net_device *dev = rt->fib6_nh.nh_dev;
1038 struct rt6_info *nrt;
1039
e873e4b9
WW		 1040 if (!fib6_info_hold_safe(rt))
1041 return NULL;
1042
93531c67 1043 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
dec9b0e2
DA		 1044 if (nrt)
1045 ip6_rt_copy_init(nrt, rt);
e873e4b9
WW		 1046 else
1047 fib6_info_release(rt);
dec9b0e2
DA		 1048
1049 return nrt;
1050}
1051
8ed67789
DL		 1052static struct rt6_info *ip6_pol_route_lookup(struct net *net,
1053 struct fib6_table *table,
b75cc8f9
DA		 1054 struct flowi6 *fl6,
1055 const struct sk_buff *skb,
1056 int flags)
1da177e4 1057{
8d1c802b 1058 struct fib6_info *f6i;
1da177e4 1059 struct fib6_node *fn;
23fb93a4 1060 struct rt6_info *rt;
1da177e4 1061
b6cdbc85
DA		 1062 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1063 flags &= ~RT6_LOOKUP_F_IFACE;
1064
66f5d6ce 1065 rcu_read_lock();
6454743b 1066 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
c71099ac 1067restart:
23fb93a4
DA		 1068 f6i = rcu_dereference(fn->leaf);
1069 if (!f6i) {
1070 f6i = net->ipv6.fib6_null_entry;
66f5d6ce 1071 } else {
23fb93a4 1072 f6i = rt6_device_match(net, f6i, &fl6->saddr,
66f5d6ce 1073 fl6->flowi6_oif, flags);
93c2fb25 1074 if (f6i->fib6_nsiblings && fl6->flowi6_oif == 0)
3b290a31
DA		 1075 f6i = fib6_multipath_select(net, f6i, fl6,
1076 fl6->flowi6_oif, skb,
1077 flags);
66f5d6ce 1078 }
23fb93a4 1079 if (f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 1080 fn = fib6_backtrack(fn, &fl6->saddr);
1081 if (fn)
1082 goto restart;
1083 }
2b760fcf 1084
d4bea421 1085 trace_fib6_table_lookup(net, f6i, table, fl6);
d3843fe5 1086
2b760fcf 1087 /* Search through exception table */
23fb93a4
DA		 1088 rt = rt6_find_cached_rt(f6i, &fl6->daddr, &fl6->saddr);
1089 if (rt) {
dec9b0e2
DA		 1090 if (ip6_hold_safe(net, &rt, true))
1091 dst_use_noref(&rt->dst, jiffies);
23fb93a4 1092 } else if (f6i == net->ipv6.fib6_null_entry) {
dec9b0e2
DA		 1093 rt = net->ipv6.ip6_null_entry;
1094 dst_hold(&rt->dst);
23fb93a4
DA		 1095 } else {
1096 rt = ip6_create_rt_rcu(f6i);
1097 if (!rt) {
1098 rt = net->ipv6.ip6_null_entry;
1099 dst_hold(&rt->dst);
1100 }
dec9b0e2 1101 }
b811580d 1102
66f5d6ce 1103 rcu_read_unlock();
b811580d 1104
c71099ac 1105 return rt;
c71099ac
TG		 1106}
1107
67ba4152 1108struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
b75cc8f9 1109 const struct sk_buff *skb, int flags)
ea6e574e 1110{
b75cc8f9 1111 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
ea6e574e
FW		 1112}
1113EXPORT_SYMBOL_GPL(ip6_route_lookup);
1114
9acd9f3a 1115struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
b75cc8f9
DA		 1116 const struct in6_addr *saddr, int oif,
1117 const struct sk_buff *skb, int strict)
c71099ac 1118{
4c9483b2
DM		 1119 struct flowi6 fl6 = {
1120 .flowi6_oif = oif,
1121 .daddr = *daddr,
c71099ac
TG		 1122 };
1123 struct dst_entry *dst;
77d16f45 1124 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
c71099ac 1125
adaa70bb 1126 if (saddr) {
4c9483b2 1127 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
adaa70bb
TG		 1128 flags |= RT6_LOOKUP_F_HAS_SADDR;
1129 }
1130
b75cc8f9 1131 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
c71099ac
TG		 1132 if (dst->error == 0)
1133 return (struct rt6_info *) dst;
1134
1135 dst_release(dst);
1136
1da177e4
LT		 1137 return NULL;
1138}
7159039a
YH		 1139EXPORT_SYMBOL(rt6_lookup);
1140
c71099ac 1141/* ip6_ins_rt is called with FREE table->tb6_lock.
1cfb71ee
WW		 1142 * It takes new route entry, the addition fails by any reason the
1143 * route is released.
1144 * Caller must hold dst before calling it.
1da177e4
LT		 1145 */
1146
8d1c802b 1147static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
333c4301 1148 struct netlink_ext_ack *extack)
1da177e4
LT		 1149{
1150 int err;
c71099ac 1151 struct fib6_table *table;
1da177e4 1152
93c2fb25 1153 table = rt->fib6_table;
66f5d6ce 1154 spin_lock_bh(&table->tb6_lock);
d4ead6b3 1155 err = fib6_add(&table->tb6_root, rt, info, extack);
66f5d6ce 1156 spin_unlock_bh(&table->tb6_lock);
1da177e4
LT		 1157
1158 return err;
1159}
1160
8d1c802b 1161int ip6_ins_rt(struct net *net, struct fib6_info *rt)
40e22e8f 1162{
afb1d4b5 1163 struct nl_info info = { .nl_net = net, };
e715b6d3 1164
d4ead6b3 1165 return __ip6_ins_rt(rt, &info, NULL);
40e22e8f
TG		 1166}
1167
8d1c802b 1168static struct rt6_info *ip6_rt_cache_alloc(struct fib6_info *ort,
8b9df265
MKL		 1169 const struct in6_addr *daddr,
1170 const struct in6_addr *saddr)
1da177e4 1171{
4832c30d 1172 struct net_device *dev;
1da177e4
LT		 1173 struct rt6_info *rt;
1174
1175 /*
1176 * Clone the route.
1177 */
1178
e873e4b9
WW		 1179 if (!fib6_info_hold_safe(ort))
1180 return NULL;
1181
4832c30d 1182 dev = ip6_rt_get_dev_rcu(ort);
93531c67 1183 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
e873e4b9
WW		 1184 if (!rt) {
1185 fib6_info_release(ort);
83a09abd 1186 return NULL;
e873e4b9 1187 }
83a09abd
MKL		 1188
1189 ip6_rt_copy_init(rt, ort);
1190 rt->rt6i_flags |= RTF_CACHE;
83a09abd
MKL		 1191 rt->dst.flags |= DST_HOST;
1192 rt->rt6i_dst.addr = *daddr;
1193 rt->rt6i_dst.plen = 128;
1da177e4 1194
83a09abd 1195 if (!rt6_is_gw_or_nonexthop(ort)) {
93c2fb25
DA		 1196 if (ort->fib6_dst.plen != 128 &&
1197 ipv6_addr_equal(&ort->fib6_dst.addr, daddr))
83a09abd 1198 rt->rt6i_flags |= RTF_ANYCAST;
1da177e4 1199#ifdef CONFIG_IPV6_SUBTREES
83a09abd
MKL		 1200 if (rt->rt6i_src.plen && saddr) {
1201 rt->rt6i_src.addr = *saddr;
1202 rt->rt6i_src.plen = 128;
8b9df265 1203 }
83a09abd 1204#endif
95a9a5ba 1205 }
1da177e4 1206
95a9a5ba
YH		 1207 return rt;
1208}
1da177e4 1209
8d1c802b 1210static struct rt6_info *ip6_rt_pcpu_alloc(struct fib6_info *rt)
d52d3997 1211{
3b6761d1 1212 unsigned short flags = fib6_info_dst_flags(rt);
4832c30d 1213 struct net_device *dev;
d52d3997
MKL		 1214 struct rt6_info *pcpu_rt;
1215
e873e4b9
WW		 1216 if (!fib6_info_hold_safe(rt))
1217 return NULL;
1218
4832c30d
DA		 1219 rcu_read_lock();
1220 dev = ip6_rt_get_dev_rcu(rt);
93531c67 1221 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags);
4832c30d 1222 rcu_read_unlock();
e873e4b9
WW		 1223 if (!pcpu_rt) {
1224 fib6_info_release(rt);
d52d3997 1225 return NULL;
e873e4b9 1226 }
d52d3997 1227 ip6_rt_copy_init(pcpu_rt, rt);
d52d3997
MKL		 1228 pcpu_rt->rt6i_flags |= RTF_PCPU;
1229 return pcpu_rt;
1230}
1231
66f5d6ce 1232/* It should be called with rcu_read_lock() acquired */
8d1c802b 1233static struct rt6_info *rt6_get_pcpu_route(struct fib6_info *rt)
d52d3997 1234{
a73e4195 1235 struct rt6_info *pcpu_rt, **p;
d52d3997
MKL		 1236
1237 p = this_cpu_ptr(rt->rt6i_pcpu);
1238 pcpu_rt = *p;
1239
d4ead6b3
DA		 1240 if (pcpu_rt)
1241 ip6_hold_safe(NULL, &pcpu_rt, false);
d3843fe5 1242
a73e4195
MKL		 1243 return pcpu_rt;
1244}
1245
afb1d4b5 1246static struct rt6_info *rt6_make_pcpu_route(struct net *net,
8d1c802b 1247 struct fib6_info *rt)
a73e4195
MKL		 1248{
1249 struct rt6_info *pcpu_rt, *prev, **p;
d52d3997
MKL		 1250
1251 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1252 if (!pcpu_rt) {
9c7370a1
MKL		 1253 dst_hold(&net->ipv6.ip6_null_entry->dst);
1254 return net->ipv6.ip6_null_entry;
d52d3997
MKL		 1255 }
1256
a94b9367
WW		 1257 dst_hold(&pcpu_rt->dst);
1258 p = this_cpu_ptr(rt->rt6i_pcpu);
1259 prev = cmpxchg(p, NULL, pcpu_rt);
951f788a 1260 BUG_ON(prev);
a94b9367 1261
d52d3997
MKL		 1262 return pcpu_rt;
1263}
1264
35732d01
WW		 1265/* exception hash table implementation
1266 */
1267static DEFINE_SPINLOCK(rt6_exception_lock);
1268
1269/* Remove rt6_ex from hash table and free the memory
1270 * Caller must hold rt6_exception_lock
1271 */
1272static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1273 struct rt6_exception *rt6_ex)
1274{
b2427e67 1275 struct net *net;
81eb8447 1276
35732d01
WW		 1277 if (!bucket || !rt6_ex)
1278 return;
b2427e67
CIK		 1279
1280 net = dev_net(rt6_ex->rt6i->dst.dev);
35732d01 1281 hlist_del_rcu(&rt6_ex->hlist);
77634cc6 1282 dst_release(&rt6_ex->rt6i->dst);
35732d01
WW		 1283 kfree_rcu(rt6_ex, rcu);
1284 WARN_ON_ONCE(!bucket->depth);
1285 bucket->depth--;
81eb8447 1286 net->ipv6.rt6_stats->fib_rt_cache--;
35732d01
WW		 1287}
1288
1289/* Remove oldest rt6_ex in bucket and free the memory
1290 * Caller must hold rt6_exception_lock
1291 */
1292static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1293{
1294 struct rt6_exception *rt6_ex, *oldest = NULL;
1295
1296 if (!bucket)
1297 return;
1298
1299 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1300 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1301 oldest = rt6_ex;
1302 }
1303 rt6_remove_exception(bucket, oldest);
1304}
1305
1306static u32 rt6_exception_hash(const struct in6_addr *dst,
1307 const struct in6_addr *src)
1308{
1309 static u32 seed __read_mostly;
1310 u32 val;
1311
1312 net_get_random_once(&seed, sizeof(seed));
1313 val = jhash(dst, sizeof(*dst), seed);
1314
1315#ifdef CONFIG_IPV6_SUBTREES
1316 if (src)
1317 val = jhash(src, sizeof(*src), val);
1318#endif
1319 return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1320}
1321
1322/* Helper function to find the cached rt in the hash table
1323 * and update bucket pointer to point to the bucket for this
1324 * (daddr, saddr) pair
1325 * Caller must hold rt6_exception_lock
1326 */
1327static struct rt6_exception *
1328__rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1329 const struct in6_addr *daddr,
1330 const struct in6_addr *saddr)
1331{
1332 struct rt6_exception *rt6_ex;
1333 u32 hval;
1334
1335 if (!(*bucket) || !daddr)
1336 return NULL;
1337
1338 hval = rt6_exception_hash(daddr, saddr);
1339 *bucket += hval;
1340
1341 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1342 struct rt6_info *rt6 = rt6_ex->rt6i;
1343 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1344
1345#ifdef CONFIG_IPV6_SUBTREES
1346 if (matched && saddr)
1347 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1348#endif
1349 if (matched)
1350 return rt6_ex;
1351 }
1352 return NULL;
1353}
1354
1355/* Helper function to find the cached rt in the hash table
1356 * and update bucket pointer to point to the bucket for this
1357 * (daddr, saddr) pair
1358 * Caller must hold rcu_read_lock()
1359 */
1360static struct rt6_exception *
1361__rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1362 const struct in6_addr *daddr,
1363 const struct in6_addr *saddr)
1364{
1365 struct rt6_exception *rt6_ex;
1366 u32 hval;
1367
1368 WARN_ON_ONCE(!rcu_read_lock_held());
1369
1370 if (!(*bucket) || !daddr)
1371 return NULL;
1372
1373 hval = rt6_exception_hash(daddr, saddr);
1374 *bucket += hval;
1375
1376 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1377 struct rt6_info *rt6 = rt6_ex->rt6i;
1378 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1379
1380#ifdef CONFIG_IPV6_SUBTREES
1381 if (matched && saddr)
1382 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1383#endif
1384 if (matched)
1385 return rt6_ex;
1386 }
1387 return NULL;
1388}
1389
8d1c802b 1390static unsigned int fib6_mtu(const struct fib6_info *rt)
d4ead6b3
DA		 1391{
1392 unsigned int mtu;
1393
dcd1f572
DA		 1394 if (rt->fib6_pmtu) {
1395 mtu = rt->fib6_pmtu;
1396 } else {
1397 struct net_device *dev = fib6_info_nh_dev(rt);
1398 struct inet6_dev *idev;
1399
1400 rcu_read_lock();
1401 idev = __in6_dev_get(dev);
1402 mtu = idev->cnf.mtu6;
1403 rcu_read_unlock();
1404 }
1405
d4ead6b3
DA		 1406 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1407
1408 return mtu - lwtunnel_headroom(rt->fib6_nh.nh_lwtstate, mtu);
1409}
1410
35732d01 1411static int rt6_insert_exception(struct rt6_info *nrt,
8d1c802b 1412 struct fib6_info *ort)
35732d01 1413{
5e670d84 1414 struct net *net = dev_net(nrt->dst.dev);
35732d01
WW		 1415 struct rt6_exception_bucket *bucket;
1416 struct in6_addr *src_key = NULL;
1417 struct rt6_exception *rt6_ex;
1418 int err = 0;
1419
35732d01
WW		 1420 spin_lock_bh(&rt6_exception_lock);
1421
1422 if (ort->exception_bucket_flushed) {
1423 err = -EINVAL;
1424 goto out;
1425 }
1426
1427 bucket = rcu_dereference_protected(ort->rt6i_exception_bucket,
1428 lockdep_is_held(&rt6_exception_lock));
1429 if (!bucket) {
1430 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1431 GFP_ATOMIC);
1432 if (!bucket) {
1433 err = -ENOMEM;
1434 goto out;
1435 }
1436 rcu_assign_pointer(ort->rt6i_exception_bucket, bucket);
1437 }
1438
1439#ifdef CONFIG_IPV6_SUBTREES
1440 /* rt6i_src.plen != 0 indicates ort is in subtree
1441 * and exception table is indexed by a hash of
1442 * both rt6i_dst and rt6i_src.
1443 * Otherwise, the exception table is indexed by
1444 * a hash of only rt6i_dst.
1445 */
93c2fb25 1446 if (ort->fib6_src.plen)
35732d01
WW		 1447 src_key = &nrt->rt6i_src.addr;
1448#endif
f5bbe7ee
WW		 1449 /* rt6_mtu_change() might lower mtu on ort.
1450 * Only insert this exception route if its mtu
1451 * is less than ort's mtu value.
1452 */
d4ead6b3 1453 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(ort)) {
f5bbe7ee
WW		 1454 err = -EINVAL;
1455 goto out;
1456 }
60006a48 1457
35732d01
WW		 1458 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1459 src_key);
1460 if (rt6_ex)
1461 rt6_remove_exception(bucket, rt6_ex);
1462
1463 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1464 if (!rt6_ex) {
1465 err = -ENOMEM;
1466 goto out;
1467 }
1468 rt6_ex->rt6i = nrt;
1469 rt6_ex->stamp = jiffies;
35732d01
WW		 1470 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1471 bucket->depth++;
81eb8447 1472 net->ipv6.rt6_stats->fib_rt_cache++;
35732d01
WW		 1473
1474 if (bucket->depth > FIB6_MAX_DEPTH)
1475 rt6_exception_remove_oldest(bucket);
1476
1477out:
1478 spin_unlock_bh(&rt6_exception_lock);
1479
1480 /* Update fn->fn_sernum to invalidate all cached dst */
b886d5f2 1481 if (!err) {
93c2fb25 1482 spin_lock_bh(&ort->fib6_table->tb6_lock);
7aef6859 1483 fib6_update_sernum(net, ort);
93c2fb25 1484 spin_unlock_bh(&ort->fib6_table->tb6_lock);
b886d5f2
PA		 1485 fib6_force_start_gc(net);
1486 }
35732d01
WW		 1487
1488 return err;
1489}
1490
8d1c802b 1491void rt6_flush_exceptions(struct fib6_info *rt)
35732d01
WW		 1492{
1493 struct rt6_exception_bucket *bucket;
1494 struct rt6_exception *rt6_ex;
1495 struct hlist_node *tmp;
1496 int i;
1497
1498 spin_lock_bh(&rt6_exception_lock);
1499 /* Prevent rt6_insert_exception() to recreate the bucket list */
1500 rt->exception_bucket_flushed = 1;
1501
1502 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1503 lockdep_is_held(&rt6_exception_lock));
1504 if (!bucket)
1505 goto out;
1506
1507 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1508 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist)
1509 rt6_remove_exception(bucket, rt6_ex);
1510 WARN_ON_ONCE(bucket->depth);
1511 bucket++;
1512 }
1513
1514out:
1515 spin_unlock_bh(&rt6_exception_lock);
1516}
1517
1518/* Find cached rt in the hash table inside passed in rt
1519 * Caller has to hold rcu_read_lock()
1520 */
8d1c802b 1521static struct rt6_info *rt6_find_cached_rt(struct fib6_info *rt,
35732d01
WW		 1522 struct in6_addr *daddr,
1523 struct in6_addr *saddr)
1524{
1525 struct rt6_exception_bucket *bucket;
1526 struct in6_addr *src_key = NULL;
1527 struct rt6_exception *rt6_ex;
1528 struct rt6_info *res = NULL;
1529
1530 bucket = rcu_dereference(rt->rt6i_exception_bucket);
1531
1532#ifdef CONFIG_IPV6_SUBTREES
1533 /* rt6i_src.plen != 0 indicates rt is in subtree
1534 * and exception table is indexed by a hash of
1535 * both rt6i_dst and rt6i_src.
1536 * Otherwise, the exception table is indexed by
1537 * a hash of only rt6i_dst.
1538 */
93c2fb25 1539 if (rt->fib6_src.plen)
35732d01
WW		 1540 src_key = saddr;
1541#endif
1542 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1543
1544 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1545 res = rt6_ex->rt6i;
1546
1547 return res;
1548}
1549
1550/* Remove the passed in cached rt from the hash table that contains it */
23fb93a4 1551static int rt6_remove_exception_rt(struct rt6_info *rt)
35732d01 1552{
35732d01
WW		 1553 struct rt6_exception_bucket *bucket;
1554 struct in6_addr *src_key = NULL;
1555 struct rt6_exception *rt6_ex;
8a14e46f 1556 struct fib6_info *from;
35732d01
WW		 1557 int err;
1558
091311de 1559 from = rcu_dereference(rt->from);
35732d01 1560 if (!from ||
442d713b 1561 !(rt->rt6i_flags & RTF_CACHE))
35732d01
WW		 1562 return -EINVAL;
1563
1564 if (!rcu_access_pointer(from->rt6i_exception_bucket))
1565 return -ENOENT;
1566
1567 spin_lock_bh(&rt6_exception_lock);
1568 bucket = rcu_dereference_protected(from->rt6i_exception_bucket,
1569 lockdep_is_held(&rt6_exception_lock));
1570#ifdef CONFIG_IPV6_SUBTREES
1571 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1572 * and exception table is indexed by a hash of
1573 * both rt6i_dst and rt6i_src.
1574 * Otherwise, the exception table is indexed by
1575 * a hash of only rt6i_dst.
1576 */
93c2fb25 1577 if (from->fib6_src.plen)
35732d01
WW		 1578 src_key = &rt->rt6i_src.addr;
1579#endif
1580 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1581 &rt->rt6i_dst.addr,
1582 src_key);
1583 if (rt6_ex) {
1584 rt6_remove_exception(bucket, rt6_ex);
1585 err = 0;
1586 } else {
1587 err = -ENOENT;
1588 }
1589
1590 spin_unlock_bh(&rt6_exception_lock);
1591 return err;
1592}
1593
1594/* Find rt6_ex which contains the passed in rt cache and
1595 * refresh its stamp
1596 */
1597static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1598{
35732d01 1599 struct rt6_exception_bucket *bucket;
8d1c802b 1600 struct fib6_info *from = rt->from;
35732d01
WW		 1601 struct in6_addr *src_key = NULL;
1602 struct rt6_exception *rt6_ex;
1603
1604 if (!from ||
442d713b 1605 !(rt->rt6i_flags & RTF_CACHE))
35732d01
WW		 1606 return;
1607
1608 rcu_read_lock();
1609 bucket = rcu_dereference(from->rt6i_exception_bucket);
1610
1611#ifdef CONFIG_IPV6_SUBTREES
1612 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1613 * and exception table is indexed by a hash of
1614 * both rt6i_dst and rt6i_src.
1615 * Otherwise, the exception table is indexed by
1616 * a hash of only rt6i_dst.
1617 */
93c2fb25 1618 if (from->fib6_src.plen)
35732d01
WW		 1619 src_key = &rt->rt6i_src.addr;
1620#endif
1621 rt6_ex = __rt6_find_exception_rcu(&bucket,
1622 &rt->rt6i_dst.addr,
1623 src_key);
1624 if (rt6_ex)
1625 rt6_ex->stamp = jiffies;
1626
1627 rcu_read_unlock();
1628}
1629
e9fa1495
SB		 1630static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1631 struct rt6_info *rt, int mtu)
1632{
1633 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1634 * lowest MTU in the path: always allow updating the route PMTU to
1635 * reflect PMTU decreases.
1636 *
1637 * If the new MTU is higher, and the route PMTU is equal to the local
1638 * MTU, this means the old MTU is the lowest in the path, so allow
1639 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1640 * handle this.
1641 */
1642
1643 if (dst_mtu(&rt->dst) >= mtu)
1644 return true;
1645
1646 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
1647 return true;
1648
1649 return false;
1650}
1651
1652static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
8d1c802b 1653 struct fib6_info *rt, int mtu)
f5bbe7ee
WW		 1654{
1655 struct rt6_exception_bucket *bucket;
1656 struct rt6_exception *rt6_ex;
1657 int i;
1658
1659 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1660 lockdep_is_held(&rt6_exception_lock));
1661
e9fa1495
SB		 1662 if (!bucket)
1663 return;
1664
1665 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1666 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1667 struct rt6_info *entry = rt6_ex->rt6i;
1668
1669 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
d4ead6b3 1670 * route), the metrics of its rt->from have already
e9fa1495
SB		 1671 * been updated.
1672 */
d4ead6b3 1673 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
e9fa1495 1674 rt6_mtu_change_route_allowed(idev, entry, mtu))
d4ead6b3 1675 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
f5bbe7ee 1676 }
e9fa1495 1677 bucket++;
f5bbe7ee
WW		 1678 }
1679}
1680
b16cb459
WW		 1681#define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
1682
8d1c802b 1683static void rt6_exceptions_clean_tohost(struct fib6_info *rt,
b16cb459
WW		 1684 struct in6_addr *gateway)
1685{
1686 struct rt6_exception_bucket *bucket;
1687 struct rt6_exception *rt6_ex;
1688 struct hlist_node *tmp;
1689 int i;
1690
1691 if (!rcu_access_pointer(rt->rt6i_exception_bucket))
1692 return;
1693
1694 spin_lock_bh(&rt6_exception_lock);
1695 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1696 lockdep_is_held(&rt6_exception_lock));
1697
1698 if (bucket) {
1699 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1700 hlist_for_each_entry_safe(rt6_ex, tmp,
1701 &bucket->chain, hlist) {
1702 struct rt6_info *entry = rt6_ex->rt6i;
1703
1704 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
1705 RTF_CACHE_GATEWAY &&
1706 ipv6_addr_equal(gateway,
1707 &entry->rt6i_gateway)) {
1708 rt6_remove_exception(bucket, rt6_ex);
1709 }
1710 }
1711 bucket++;
1712 }
1713 }
1714
1715 spin_unlock_bh(&rt6_exception_lock);
1716}
1717
c757faa8
WW		 1718static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
1719 struct rt6_exception *rt6_ex,
1720 struct fib6_gc_args *gc_args,
1721 unsigned long now)
1722{
1723 struct rt6_info *rt = rt6_ex->rt6i;
1724
1859bac0
PA		 1725 /* we are pruning and obsoleting aged-out and non gateway exceptions
1726 * even if others have still references to them, so that on next
1727 * dst_check() such references can be dropped.
1728 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
1729 * expired, independently from their aging, as per RFC 8201 section 4
1730 */
31afeb42
WW		 1731 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1732 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
1733 RT6_TRACE("aging clone %p\n", rt);
1734 rt6_remove_exception(bucket, rt6_ex);
1735 return;
1736 }
1737 } else if (time_after(jiffies, rt->dst.expires)) {
1738 RT6_TRACE("purging expired route %p\n", rt);
c757faa8
WW		 1739 rt6_remove_exception(bucket, rt6_ex);
1740 return;
31afeb42
WW		 1741 }
1742
1743 if (rt->rt6i_flags & RTF_GATEWAY) {
c757faa8
WW		 1744 struct neighbour *neigh;
1745 __u8 neigh_flags = 0;
1746
1bfa26ff
ED		 1747 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
1748 if (neigh)
c757faa8 1749 neigh_flags = neigh->flags;
1bfa26ff 1750
c757faa8
WW		 1751 if (!(neigh_flags & NTF_ROUTER)) {
1752 RT6_TRACE("purging route %p via non-router but gateway\n",
1753 rt);
1754 rt6_remove_exception(bucket, rt6_ex);
1755 return;
1756 }
1757 }
31afeb42 1758
c757faa8
WW		 1759 gc_args->more++;
1760}
1761
8d1c802b 1762void rt6_age_exceptions(struct fib6_info *rt,
c757faa8
WW		 1763 struct fib6_gc_args *gc_args,
1764 unsigned long now)
1765{
1766 struct rt6_exception_bucket *bucket;
1767 struct rt6_exception *rt6_ex;
1768 struct hlist_node *tmp;
1769 int i;
1770
1771 if (!rcu_access_pointer(rt->rt6i_exception_bucket))
1772 return;
1773
1bfa26ff
ED		 1774 rcu_read_lock_bh();
1775 spin_lock(&rt6_exception_lock);
c757faa8
WW		 1776 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1777 lockdep_is_held(&rt6_exception_lock));
1778
1779 if (bucket) {
1780 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1781 hlist_for_each_entry_safe(rt6_ex, tmp,
1782 &bucket->chain, hlist) {
1783 rt6_age_examine_exception(bucket, rt6_ex,
1784 gc_args, now);
1785 }
1786 bucket++;
1787 }
1788 }
1bfa26ff
ED		 1789 spin_unlock(&rt6_exception_lock);
1790 rcu_read_unlock_bh();
c757faa8
WW		 1791}
1792
1d053da9
DA		 1793/* must be called with rcu lock held */
1794struct fib6_info *fib6_table_lookup(struct net *net, struct fib6_table *table,
1795 int oif, struct flowi6 *fl6, int strict)
1da177e4 1796{
367efcb9 1797 struct fib6_node *fn, *saved_fn;
8d1c802b 1798 struct fib6_info *f6i;
1da177e4 1799
6454743b 1800 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
367efcb9 1801 saved_fn = fn;
1da177e4 1802
ca254490
DA		 1803 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1804 oif = 0;
1805
a3c00e46 1806redo_rt6_select:
23fb93a4 1807 f6i = rt6_select(net, fn, oif, strict);
23fb93a4 1808 if (f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 1809 fn = fib6_backtrack(fn, &fl6->saddr);
1810 if (fn)
1811 goto redo_rt6_select;
367efcb9
MKL		 1812 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1813 /* also consider unreachable route */
1814 strict &= ~RT6_LOOKUP_F_REACHABLE;
1815 fn = saved_fn;
1816 goto redo_rt6_select;
367efcb9 1817 }
a3c00e46
MKL		 1818 }
1819
d4bea421 1820 trace_fib6_table_lookup(net, f6i, table, fl6);
fb9de91e 1821
1d053da9
DA		 1822 return f6i;
1823}
1824
1825struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
1826 int oif, struct flowi6 *fl6,
1827 const struct sk_buff *skb, int flags)
1828{
1829 struct fib6_info *f6i;
1830 struct rt6_info *rt;
1831 int strict = 0;
1832
1833 strict |= flags & RT6_LOOKUP_F_IFACE;
1834 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
1835 if (net->ipv6.devconf_all->forwarding == 0)
1836 strict |= RT6_LOOKUP_F_REACHABLE;
1837
1838 rcu_read_lock();
1839
1840 f6i = fib6_table_lookup(net, table, oif, fl6, strict);
1841 if (f6i->fib6_nsiblings)
1842 f6i = fib6_multipath_select(net, f6i, fl6, oif, skb, strict);
1843
23fb93a4 1844 if (f6i == net->ipv6.fib6_null_entry) {
421842ed 1845 rt = net->ipv6.ip6_null_entry;
66f5d6ce 1846 rcu_read_unlock();
d3843fe5 1847 dst_hold(&rt->dst);
d3843fe5 1848 return rt;
23fb93a4
DA		 1849 }
1850
1851 /*Search through exception table */
1852 rt = rt6_find_cached_rt(f6i, &fl6->daddr, &fl6->saddr);
1853 if (rt) {
d4ead6b3 1854 if (ip6_hold_safe(net, &rt, true))
d3843fe5 1855 dst_use_noref(&rt->dst, jiffies);
d4ead6b3 1856
66f5d6ce 1857 rcu_read_unlock();
d52d3997 1858 return rt;
3da59bd9 1859 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
93c2fb25 1860 !(f6i->fib6_flags & RTF_GATEWAY))) {
3da59bd9
MKL		 1861 /* Create a RTF_CACHE clone which will not be
1862 * owned by the fib6 tree. It is for the special case where
1863 * the daddr in the skb during the neighbor look-up is different
1864 * from the fl6->daddr used to look-up route here.
1865 */
3da59bd9
MKL		 1866 struct rt6_info *uncached_rt;
1867
23fb93a4 1868 uncached_rt = ip6_rt_cache_alloc(f6i, &fl6->daddr, NULL);
d52d3997 1869
4d85cd0c 1870 rcu_read_unlock();
c71099ac 1871
1cfb71ee
WW		 1872 if (uncached_rt) {
1873 /* Uncached_rt's refcnt is taken during ip6_rt_cache_alloc()
1874 * No need for another dst_hold()
1875 */
8d0b94af 1876 rt6_uncached_list_add(uncached_rt);
81eb8447 1877 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
1cfb71ee 1878 } else {
3da59bd9 1879 uncached_rt = net->ipv6.ip6_null_entry;
1cfb71ee
WW		 1880 dst_hold(&uncached_rt->dst);
1881 }
b811580d 1882
3da59bd9 1883 return uncached_rt;
d52d3997
MKL		 1884 } else {
1885 /* Get a percpu copy */
1886
1887 struct rt6_info *pcpu_rt;
1888
951f788a 1889 local_bh_disable();
23fb93a4 1890 pcpu_rt = rt6_get_pcpu_route(f6i);
d52d3997 1891
93531c67
DA		 1892 if (!pcpu_rt)
1893 pcpu_rt = rt6_make_pcpu_route(net, f6i);
1894
951f788a
ED		 1895 local_bh_enable();
1896 rcu_read_unlock();
d4bea421 1897
d52d3997
MKL		 1898 return pcpu_rt;
1899 }
1da177e4 1900}
9ff74384 1901EXPORT_SYMBOL_GPL(ip6_pol_route);
1da177e4 1902
b75cc8f9
DA		 1903static struct rt6_info *ip6_pol_route_input(struct net *net,
1904 struct fib6_table *table,
1905 struct flowi6 *fl6,
1906 const struct sk_buff *skb,
1907 int flags)
4acad72d 1908{
b75cc8f9 1909 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
4acad72d
PE		 1910}
1911
d409b847
MB		 1912struct dst_entry *ip6_route_input_lookup(struct net *net,
1913 struct net_device *dev,
b75cc8f9
DA		 1914 struct flowi6 *fl6,
1915 const struct sk_buff *skb,
1916 int flags)
72331bc0
SL		 1917{
1918 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1919 flags |= RT6_LOOKUP_F_IFACE;
1920
b75cc8f9 1921 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
72331bc0 1922}
d409b847 1923EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
72331bc0 1924
23aebdac 1925static void ip6_multipath_l3_keys(const struct sk_buff *skb,
5e5d6fed
RP		 1926 struct flow_keys *keys,
1927 struct flow_keys *flkeys)
23aebdac
JS		 1928{
1929 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
1930 const struct ipv6hdr *key_iph = outer_iph;
5e5d6fed 1931 struct flow_keys *_flkeys = flkeys;
23aebdac
JS		 1932 const struct ipv6hdr *inner_iph;
1933 const struct icmp6hdr *icmph;
1934 struct ipv6hdr _inner_iph;
cea67a2d 1935 struct icmp6hdr _icmph;
23aebdac
JS		 1936
1937 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
1938 goto out;
1939
cea67a2d
ED		 1940 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
1941 sizeof(_icmph), &_icmph);
1942 if (!icmph)
1943 goto out;
1944
23aebdac
JS		 1945 if (icmph->icmp6_type != ICMPV6_DEST_UNREACH &&
1946 icmph->icmp6_type != ICMPV6_PKT_TOOBIG &&
1947 icmph->icmp6_type != ICMPV6_TIME_EXCEED &&
1948 icmph->icmp6_type != ICMPV6_PARAMPROB)
1949 goto out;
1950
1951 inner_iph = skb_header_pointer(skb,
1952 skb_transport_offset(skb) + sizeof(*icmph),
1953 sizeof(_inner_iph), &_inner_iph);
1954 if (!inner_iph)
1955 goto out;
1956
1957 key_iph = inner_iph;
5e5d6fed 1958 _flkeys = NULL;
23aebdac 1959out:
5e5d6fed
RP		 1960 if (_flkeys) {
1961 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
1962 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
1963 keys->tags.flow_label = _flkeys->tags.flow_label;
1964 keys->basic.ip_proto = _flkeys->basic.ip_proto;
1965 } else {
1966 keys->addrs.v6addrs.src = key_iph->saddr;
1967 keys->addrs.v6addrs.dst = key_iph->daddr;
fa1be7e0 1968 keys->tags.flow_label = ip6_flowlabel(key_iph);
5e5d6fed
RP		 1969 keys->basic.ip_proto = key_iph->nexthdr;
1970 }
23aebdac
JS		 1971}
1972
1973/* if skb is set it will be used and fl6 can be NULL */
b4bac172
DA		 1974u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
1975 const struct sk_buff *skb, struct flow_keys *flkeys)
23aebdac
JS		 1976{
1977 struct flow_keys hash_keys;
9a2a537a 1978 u32 mhash;
23aebdac 1979
bbfa047a 1980 switch (ip6_multipath_hash_policy(net)) {
b4bac172
DA		 1981 case 0:
1982 memset(&hash_keys, 0, sizeof(hash_keys));
1983 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1984 if (skb) {
1985 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
1986 } else {
1987 hash_keys.addrs.v6addrs.src = fl6->saddr;
1988 hash_keys.addrs.v6addrs.dst = fl6->daddr;
fa1be7e0 1989 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
b4bac172
DA		 1990 hash_keys.basic.ip_proto = fl6->flowi6_proto;
1991 }
1992 break;
1993 case 1:
1994 if (skb) {
1995 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
1996 struct flow_keys keys;
1997
1998 /* short-circuit if we already have L4 hash present */
1999 if (skb->l4_hash)
2000 return skb_get_hash_raw(skb) >> 1;
2001
2002 memset(&hash_keys, 0, sizeof(hash_keys));
2003
2004 if (!flkeys) {
2005 skb_flow_dissect_flow_keys(skb, &keys, flag);
2006 flkeys = &keys;
2007 }
2008 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2009 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2010 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2011 hash_keys.ports.src = flkeys->ports.src;
2012 hash_keys.ports.dst = flkeys->ports.dst;
2013 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2014 } else {
2015 memset(&hash_keys, 0, sizeof(hash_keys));
2016 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2017 hash_keys.addrs.v6addrs.src = fl6->saddr;
2018 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2019 hash_keys.ports.src = fl6->fl6_sport;
2020 hash_keys.ports.dst = fl6->fl6_dport;
2021 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2022 }
2023 break;
23aebdac 2024 }
9a2a537a 2025 mhash = flow_hash_from_keys(&hash_keys);
23aebdac 2026
9a2a537a 2027 return mhash >> 1;
23aebdac
JS		 2028}
2029
c71099ac
TG		 2030void ip6_route_input(struct sk_buff *skb)
2031{
b71d1d42 2032 const struct ipv6hdr *iph = ipv6_hdr(skb);
c346dca1 2033 struct net *net = dev_net(skb->dev);
adaa70bb 2034 int flags = RT6_LOOKUP_F_HAS_SADDR;
904af04d 2035 struct ip_tunnel_info *tun_info;
4c9483b2 2036 struct flowi6 fl6 = {
e0d56fdd 2037 .flowi6_iif = skb->dev->ifindex,
4c9483b2
DM		 2038 .daddr = iph->daddr,
2039 .saddr = iph->saddr,
6502ca52 2040 .flowlabel = ip6_flowinfo(iph),
4c9483b2
DM		 2041 .flowi6_mark = skb->mark,
2042 .flowi6_proto = iph->nexthdr,
c71099ac 2043 };
5e5d6fed 2044 struct flow_keys *flkeys = NULL, _flkeys;
adaa70bb 2045
904af04d 2046 tun_info = skb_tunnel_info(skb);
46fa062a 2047 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
904af04d 2048 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
5e5d6fed
RP		 2049
2050 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2051 flkeys = &_flkeys;
2052
23aebdac 2053 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
b4bac172 2054 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
06e9d040 2055 skb_dst_drop(skb);
b75cc8f9
DA		 2056 skb_dst_set(skb,
2057 ip6_route_input_lookup(net, skb->dev, &fl6, skb, flags));
c71099ac
TG		 2058}
2059
b75cc8f9
DA		 2060static struct rt6_info *ip6_pol_route_output(struct net *net,
2061 struct fib6_table *table,
2062 struct flowi6 *fl6,
2063 const struct sk_buff *skb,
2064 int flags)
1da177e4 2065{
b75cc8f9 2066 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
c71099ac
TG		 2067}
2068
6f21c96a
PA		 2069struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
2070 struct flowi6 *fl6, int flags)
c71099ac 2071{
d46a9d67 2072 bool any_src;
c71099ac 2073
3ede0bbc
RS		 2074 if (ipv6_addr_type(&fl6->daddr) &
2075 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
4c1feac5
DA		 2076 struct dst_entry *dst;
2077
2078 dst = l3mdev_link_scope_lookup(net, fl6);
2079 if (dst)
2080 return dst;
2081 }
ca254490 2082
1fb9489b 2083 fl6->flowi6_iif = LOOPBACK_IFINDEX;
4dc27d1c 2084
d46a9d67 2085 any_src = ipv6_addr_any(&fl6->saddr);
741a11d9 2086 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
d46a9d67 2087 (fl6->flowi6_oif && any_src))
77d16f45 2088 flags |= RT6_LOOKUP_F_IFACE;
c71099ac 2089
d46a9d67 2090 if (!any_src)
adaa70bb 2091 flags |= RT6_LOOKUP_F_HAS_SADDR;
0c9a2ac1
YH		 2092 else if (sk)
2093 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
adaa70bb 2094
b75cc8f9 2095 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
1da177e4 2096}
6f21c96a 2097EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1da177e4 2098
2774c131 2099struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
14e50e57 2100{
5c1e6aa3 2101 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1dbe3252 2102 struct net_device *loopback_dev = net->loopback_dev;
14e50e57
DM		 2103 struct dst_entry *new = NULL;
2104
1dbe3252 2105 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
62cf27e5 2106 DST_OBSOLETE_DEAD, 0);
14e50e57 2107 if (rt) {
0a1f5962 2108 rt6_info_init(rt);
81eb8447 2109 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
8104891b 2110
0a1f5962 2111 new = &rt->dst;
14e50e57 2112 new->__use = 1;
352e512c 2113 new->input = dst_discard;
ede2059d 2114 new->output = dst_discard_out;
14e50e57 2115
0a1f5962 2116 dst_copy_metrics(new, &ort->dst);
14e50e57 2117
1dbe3252 2118 rt->rt6i_idev = in6_dev_get(loopback_dev);
4e3fd7a0 2119 rt->rt6i_gateway = ort->rt6i_gateway;
0a1f5962 2120 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
14e50e57
DM		 2121
2122 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2123#ifdef CONFIG_IPV6_SUBTREES
2124 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2125#endif
14e50e57
DM		 2126 }
2127
69ead7af
DM		 2128 dst_release(dst_orig);
2129 return new ? new : ERR_PTR(-ENOMEM);
14e50e57 2130}
14e50e57 2131
1da177e4
LT		 2132/*
2133 * Destination cache support functions
2134 */
2135
8d1c802b 2136static bool fib6_check(struct fib6_info *f6i, u32 cookie)
4b32b5ad 2137{
93531c67
DA		 2138 u32 rt_cookie = 0;
2139
8ae86971 2140 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
93531c67
DA		 2141 return false;
2142
2143 if (fib6_check_expired(f6i))
2144 return false;
2145
2146 return true;
4b32b5ad
MKL		 2147}
2148
a68886a6
DA		 2149static struct dst_entry *rt6_check(struct rt6_info *rt,
2150 struct fib6_info *from,
2151 u32 cookie)
3da59bd9 2152{
36143645 2153 u32 rt_cookie = 0;
c5cff856 2154
a68886a6 2155 if ((from && !fib6_get_cookie_safe(from, &rt_cookie)) ||
93531c67 2156 rt_cookie != cookie)
3da59bd9
MKL		 2157 return NULL;
2158
2159 if (rt6_check_expired(rt))
2160 return NULL;
2161
2162 return &rt->dst;
2163}
2164
a68886a6
DA		 2165static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2166 struct fib6_info *from,
2167 u32 cookie)
3da59bd9 2168{
5973fb1e
MKL		 2169 if (!__rt6_check_expired(rt) &&
2170 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
a68886a6 2171 fib6_check(from, cookie))
3da59bd9
MKL		 2172 return &rt->dst;
2173 else
2174 return NULL;
2175}
2176
1da177e4
LT		 2177static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
2178{
a87b7dc9 2179 struct dst_entry *dst_ret;
a68886a6 2180 struct fib6_info *from;
1da177e4
LT		 2181 struct rt6_info *rt;
2182
a87b7dc9
DA		 2183 rt = container_of(dst, struct rt6_info, dst);
2184
2185 rcu_read_lock();
1da177e4 2186
6f3118b5
ND		 2187 /* All IPV6 dsts are created with ->obsolete set to the value
2188 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2189 * into this function always.
2190 */
e3bc10bd 2191
a68886a6 2192 from = rcu_dereference(rt->from);
4b32b5ad 2193
a68886a6
DA		 2194 if (from && (rt->rt6i_flags & RTF_PCPU ||
2195 unlikely(!list_empty(&rt->rt6i_uncached))))
2196 dst_ret = rt6_dst_from_check(rt, from, cookie);
3da59bd9 2197 else
a68886a6 2198 dst_ret = rt6_check(rt, from, cookie);
a87b7dc9
DA		 2199
2200 rcu_read_unlock();
2201
2202 return dst_ret;
1da177e4
LT		 2203}
2204
2205static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2206{
2207 struct rt6_info *rt = (struct rt6_info *) dst;
2208
2209 if (rt) {
54c1a859 2210 if (rt->rt6i_flags & RTF_CACHE) {
c3c14da0 2211 rcu_read_lock();
54c1a859 2212 if (rt6_check_expired(rt)) {
93531c67 2213 rt6_remove_exception_rt(rt);
54c1a859
YH		 2214 dst = NULL;
2215 }
c3c14da0 2216 rcu_read_unlock();
54c1a859 2217 } else {
1da177e4 2218 dst_release(dst);
54c1a859
YH		 2219 dst = NULL;
2220 }
1da177e4 2221 }
54c1a859 2222 return dst;
1da177e4
LT		 2223}
2224
2225static void ip6_link_failure(struct sk_buff *skb)
2226{
2227 struct rt6_info *rt;
2228
3ffe533c 2229 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1da177e4 2230
adf30907 2231 rt = (struct rt6_info *) skb_dst(skb);
1da177e4 2232 if (rt) {
8a14e46f 2233 rcu_read_lock();
1eb4f758 2234 if (rt->rt6i_flags & RTF_CACHE) {
761f6026 2235 rt6_remove_exception_rt(rt);
c5cff856 2236 } else {
a68886a6 2237 struct fib6_info *from;
c5cff856
WW		 2238 struct fib6_node *fn;
2239
a68886a6
DA		 2240 from = rcu_dereference(rt->from);
2241 if (from) {
2242 fn = rcu_dereference(from->fib6_node);
2243 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2244 fn->fn_sernum = -1;
2245 }
1eb4f758 2246 }
8a14e46f 2247 rcu_read_unlock();
1da177e4
LT		 2248 }
2249}
2250
6a3e030f
DA		 2251static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2252{
a68886a6
DA		 2253 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2254 struct fib6_info *from;
2255
2256 rcu_read_lock();
2257 from = rcu_dereference(rt0->from);
2258 if (from)
2259 rt0->dst.expires = from->expires;
2260 rcu_read_unlock();
2261 }
6a3e030f
DA		 2262
2263 dst_set_expires(&rt0->dst, timeout);
2264 rt0->rt6i_flags |= RTF_EXPIRES;
2265}
2266
45e4fd26
MKL		 2267static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2268{
2269 struct net *net = dev_net(rt->dst.dev);
2270
d4ead6b3 2271 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
45e4fd26 2272 rt->rt6i_flags |= RTF_MODIFIED;
45e4fd26
MKL		 2273 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2274}
2275
0d3f6d29
MKL		 2276static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2277{
a68886a6
DA		 2278 bool from_set;
2279
2280 rcu_read_lock();
2281 from_set = !!rcu_dereference(rt->from);
2282 rcu_read_unlock();
2283
0d3f6d29 2284 return !(rt->rt6i_flags & RTF_CACHE) &&
a68886a6 2285 (rt->rt6i_flags & RTF_PCPU || from_set);
0d3f6d29
MKL		 2286}
2287
45e4fd26
MKL		 2288static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2289 const struct ipv6hdr *iph, u32 mtu)
1da177e4 2290{
0dec879f 2291 const struct in6_addr *daddr, *saddr;
67ba4152 2292 struct rt6_info *rt6 = (struct rt6_info *)dst;
1da177e4 2293
19bda36c
XL		 2294 if (dst_metric_locked(dst, RTAX_MTU))
2295 return;
2296
0dec879f
JA		 2297 if (iph) {
2298 daddr = &iph->daddr;
2299 saddr = &iph->saddr;
2300 } else if (sk) {
2301 daddr = &sk->sk_v6_daddr;
2302 saddr = &inet6_sk(sk)->saddr;
2303 } else {
2304 daddr = NULL;
2305 saddr = NULL;
2306 }
2307 dst_confirm_neigh(dst, daddr);
45e4fd26
MKL		 2308 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
2309 if (mtu >= dst_mtu(dst))
2310 return;
9d289715 2311
0d3f6d29 2312 if (!rt6_cache_allowed_for_pmtu(rt6)) {
45e4fd26 2313 rt6_do_update_pmtu(rt6, mtu);
2b760fcf
WW		 2314 /* update rt6_ex->stamp for cache */
2315 if (rt6->rt6i_flags & RTF_CACHE)
2316 rt6_update_exception_stamp_rt(rt6);
0dec879f 2317 } else if (daddr) {
a68886a6 2318 struct fib6_info *from;
45e4fd26
MKL		 2319 struct rt6_info *nrt6;
2320
4d85cd0c 2321 rcu_read_lock();
a68886a6
DA		 2322 from = rcu_dereference(rt6->from);
2323 nrt6 = ip6_rt_cache_alloc(from, daddr, saddr);
45e4fd26
MKL		 2324 if (nrt6) {
2325 rt6_do_update_pmtu(nrt6, mtu);
a68886a6 2326 if (rt6_insert_exception(nrt6, from))
2b760fcf 2327 dst_release_immediate(&nrt6->dst);
45e4fd26 2328 }
a68886a6 2329 rcu_read_unlock();
1da177e4
LT		 2330 }
2331}
2332
45e4fd26
MKL		 2333static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2334 struct sk_buff *skb, u32 mtu)
2335{
2336 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
2337}
2338
42ae66c8 2339void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
e2d118a1 2340 int oif, u32 mark, kuid_t uid)
81aded24
DM		 2341{
2342 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2343 struct dst_entry *dst;
dc92095d
MŻ		 2344 struct flowi6 fl6 = {
2345 .flowi6_oif = oif,
2346 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2347 .daddr = iph->daddr,
2348 .saddr = iph->saddr,
2349 .flowlabel = ip6_flowinfo(iph),
2350 .flowi6_uid = uid,
2351 };
81aded24
DM		 2352
2353 dst = ip6_route_output(net, NULL, &fl6);
2354 if (!dst->error)
45e4fd26 2355 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
81aded24
DM		 2356 dst_release(dst);
2357}
2358EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2359
2360void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2361{
7ddacfa5 2362 int oif = sk->sk_bound_dev_if;
33c162a9
MKL		 2363 struct dst_entry *dst;
2364
7ddacfa5
DA		 2365 if (!oif && skb->dev)
2366 oif = l3mdev_master_ifindex(skb->dev);
2367
2368 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
33c162a9
MKL		 2369
2370 dst = __sk_dst_get(sk);
2371 if (!dst || !dst->obsolete ||
2372 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2373 return;
2374
2375 bh_lock_sock(sk);
2376 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2377 ip6_datagram_dst_update(sk, false);
2378 bh_unlock_sock(sk);
81aded24
DM		 2379}
2380EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2381
7d6850f7
AK		 2382void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2383 const struct flowi6 *fl6)
2384{
2385#ifdef CONFIG_IPV6_SUBTREES
2386 struct ipv6_pinfo *np = inet6_sk(sk);
2387#endif
2388
2389 ip6_dst_store(sk, dst,
2390 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2391 &sk->sk_v6_daddr : NULL,
2392#ifdef CONFIG_IPV6_SUBTREES
2393 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2394 &np->saddr :
2395#endif
2396 NULL);
2397}
2398
b55b76b2
DJ		 2399/* Handle redirects */
2400struct ip6rd_flowi {
2401 struct flowi6 fl6;
2402 struct in6_addr gateway;
2403};
2404
2405static struct rt6_info *__ip6_route_redirect(struct net *net,
2406 struct fib6_table *table,
2407 struct flowi6 *fl6,
b75cc8f9 2408 const struct sk_buff *skb,
b55b76b2
DJ		 2409 int flags)
2410{
2411 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
23fb93a4 2412 struct rt6_info *ret = NULL, *rt_cache;
8d1c802b 2413 struct fib6_info *rt;
b55b76b2
DJ		 2414 struct fib6_node *fn;
2415
2416 /* Get the "current" route for this destination and
67c408cf 2417 * check if the redirect has come from appropriate router.
b55b76b2
DJ		 2418 *
2419 * RFC 4861 specifies that redirects should only be
2420 * accepted if they come from the nexthop to the target.
2421 * Due to the way the routes are chosen, this notion
2422 * is a bit fuzzy and one might need to check all possible
2423 * routes.
2424 */
2425
66f5d6ce 2426 rcu_read_lock();
6454743b 2427 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
b55b76b2 2428restart:
66f5d6ce 2429 for_each_fib6_node_rt_rcu(fn) {
5e670d84 2430 if (rt->fib6_nh.nh_flags & RTNH_F_DEAD)
8067bb8c 2431 continue;
14895687 2432 if (fib6_check_expired(rt))
b55b76b2 2433 continue;
93c2fb25 2434 if (rt->fib6_flags & RTF_REJECT)
b55b76b2 2435 break;
93c2fb25 2436 if (!(rt->fib6_flags & RTF_GATEWAY))
b55b76b2 2437 continue;
5e670d84 2438 if (fl6->flowi6_oif != rt->fib6_nh.nh_dev->ifindex)
b55b76b2 2439 continue;
2b760fcf
WW		 2440 /* rt_cache's gateway might be different from its 'parent'
2441 * in the case of an ip redirect.
2442 * So we keep searching in the exception table if the gateway
2443 * is different.
2444 */
5e670d84 2445 if (!ipv6_addr_equal(&rdfl->gateway, &rt->fib6_nh.nh_gw)) {
2b760fcf
WW		 2446 rt_cache = rt6_find_cached_rt(rt,
2447 &fl6->daddr,
2448 &fl6->saddr);
2449 if (rt_cache &&
2450 ipv6_addr_equal(&rdfl->gateway,
2451 &rt_cache->rt6i_gateway)) {
23fb93a4 2452 ret = rt_cache;
2b760fcf
WW		 2453 break;
2454 }
b55b76b2 2455 continue;
2b760fcf 2456 }
b55b76b2
DJ		 2457 break;
2458 }
2459
2460 if (!rt)
421842ed 2461 rt = net->ipv6.fib6_null_entry;
93c2fb25 2462 else if (rt->fib6_flags & RTF_REJECT) {
23fb93a4 2463 ret = net->ipv6.ip6_null_entry;
b0a1ba59
MKL		 2464 goto out;
2465 }
2466
421842ed 2467 if (rt == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2468 fn = fib6_backtrack(fn, &fl6->saddr);
2469 if (fn)
2470 goto restart;
b55b76b2 2471 }
a3c00e46 2472
b0a1ba59 2473out:
23fb93a4 2474 if (ret)
e873e4b9 2475 ip6_hold_safe(net, &ret, true);
23fb93a4
DA		 2476 else
2477 ret = ip6_create_rt_rcu(rt);
b55b76b2 2478
66f5d6ce 2479 rcu_read_unlock();
b55b76b2 2480
b65f164d 2481 trace_fib6_table_lookup(net, rt, table, fl6);
23fb93a4 2482 return ret;
b55b76b2
DJ		 2483};
2484
2485static struct dst_entry *ip6_route_redirect(struct net *net,
b75cc8f9
DA		 2486 const struct flowi6 *fl6,
2487 const struct sk_buff *skb,
2488 const struct in6_addr *gateway)
b55b76b2
DJ		 2489{
2490 int flags = RT6_LOOKUP_F_HAS_SADDR;
2491 struct ip6rd_flowi rdfl;
2492
2493 rdfl.fl6 = *fl6;
2494 rdfl.gateway = *gateway;
2495
b75cc8f9 2496 return fib6_rule_lookup(net, &rdfl.fl6, skb,
b55b76b2
DJ		 2497 flags, __ip6_route_redirect);
2498}
2499
e2d118a1
LC		 2500void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
2501 kuid_t uid)
3a5ad2ee
DM		 2502{
2503 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2504 struct dst_entry *dst;
1f7f10ac
MŻ		 2505 struct flowi6 fl6 = {
2506 .flowi6_iif = LOOPBACK_IFINDEX,
2507 .flowi6_oif = oif,
2508 .flowi6_mark = mark,
2509 .daddr = iph->daddr,
2510 .saddr = iph->saddr,
2511 .flowlabel = ip6_flowinfo(iph),
2512 .flowi6_uid = uid,
2513 };
3a5ad2ee 2514
b75cc8f9 2515 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
b55b76b2 2516 rt6_do_redirect(dst, NULL, skb);
3a5ad2ee
DM		 2517 dst_release(dst);
2518}
2519EXPORT_SYMBOL_GPL(ip6_redirect);
2520
d456336d 2521void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
c92a59ec
DJ		 2522{
2523 const struct ipv6hdr *iph = ipv6_hdr(skb);
2524 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
2525 struct dst_entry *dst;
0b26fb17
MŻ		 2526 struct flowi6 fl6 = {
2527 .flowi6_iif = LOOPBACK_IFINDEX,
2528 .flowi6_oif = oif,
0b26fb17
MŻ		 2529 .daddr = msg->dest,
2530 .saddr = iph->daddr,
2531 .flowi6_uid = sock_net_uid(net, NULL),
2532 };
c92a59ec 2533
b75cc8f9 2534 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
b55b76b2 2535 rt6_do_redirect(dst, NULL, skb);
c92a59ec
DJ		 2536 dst_release(dst);
2537}
2538
3a5ad2ee
DM		 2539void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
2540{
e2d118a1
LC		 2541 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
2542 sk->sk_uid);
3a5ad2ee
DM		 2543}
2544EXPORT_SYMBOL_GPL(ip6_sk_redirect);
2545
0dbaee3b 2546static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1da177e4 2547{
0dbaee3b
DM		 2548 struct net_device *dev = dst->dev;
2549 unsigned int mtu = dst_mtu(dst);
2550 struct net *net = dev_net(dev);
2551
1da177e4
LT		 2552 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
2553
5578689a
DL		 2554 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
2555 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1da177e4
LT		 2556
2557 /*
1ab1457c
YH		 2558 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
2559 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
2560 * IPV6_MAXPLEN is also valid and means: "any MSS,
1da177e4
LT		 2561 * rely only on pmtu discovery"
2562 */
2563 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
2564 mtu = IPV6_MAXPLEN;
2565 return mtu;
2566}
2567
ebb762f2 2568static unsigned int ip6_mtu(const struct dst_entry *dst)
d33e4553 2569{
d33e4553 2570 struct inet6_dev *idev;
d4ead6b3 2571 unsigned int mtu;
4b32b5ad
MKL		 2572
2573 mtu = dst_metric_raw(dst, RTAX_MTU);
618f9bc7 2574 if (mtu)
30f78d8e 2575 goto out;
618f9bc7
SK		 2576
2577 mtu = IPV6_MIN_MTU;
d33e4553
DM		 2578
2579 rcu_read_lock();
2580 idev = __in6_dev_get(dst->dev);
2581 if (idev)
2582 mtu = idev->cnf.mtu6;
2583 rcu_read_unlock();
2584
30f78d8e 2585out:
14972cbd
RP		 2586 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
2587
2588 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
d33e4553
DM		 2589}
2590
901731b8
DA		 2591/* MTU selection:
2592 * 1. mtu on route is locked - use it
2593 * 2. mtu from nexthop exception
2594 * 3. mtu from egress device
2595 *
2596 * based on ip6_dst_mtu_forward and exception logic of
2597 * rt6_find_cached_rt; called with rcu_read_lock
2598 */
2599u32 ip6_mtu_from_fib6(struct fib6_info *f6i, struct in6_addr *daddr,
2600 struct in6_addr *saddr)
2601{
2602 struct rt6_exception_bucket *bucket;
2603 struct rt6_exception *rt6_ex;
2604 struct in6_addr *src_key;
2605 struct inet6_dev *idev;
2606 u32 mtu = 0;
2607
2608 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
2609 mtu = f6i->fib6_pmtu;
2610 if (mtu)
2611 goto out;
2612 }
2613
2614 src_key = NULL;
2615#ifdef CONFIG_IPV6_SUBTREES
2616 if (f6i->fib6_src.plen)
2617 src_key = saddr;
2618#endif
2619
2620 bucket = rcu_dereference(f6i->rt6i_exception_bucket);
2621 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
2622 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
2623 mtu = dst_metric_raw(&rt6_ex->rt6i->dst, RTAX_MTU);
2624
2625 if (likely(!mtu)) {
2626 struct net_device *dev = fib6_info_nh_dev(f6i);
2627
2628 mtu = IPV6_MIN_MTU;
2629 idev = __in6_dev_get(dev);
2630 if (idev && idev->cnf.mtu6 > mtu)
2631 mtu = idev->cnf.mtu6;
2632 }
2633
2634 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
2635out:
2636 return mtu - lwtunnel_headroom(fib6_info_nh_lwt(f6i), mtu);
2637}
2638
3b00944c 2639struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
87a11578 2640 struct flowi6 *fl6)
1da177e4 2641{
87a11578 2642 struct dst_entry *dst;
1da177e4
LT		 2643 struct rt6_info *rt;
2644 struct inet6_dev *idev = in6_dev_get(dev);
c346dca1 2645 struct net *net = dev_net(dev);
1da177e4 2646
38308473 2647 if (unlikely(!idev))
122bdf67 2648 return ERR_PTR(-ENODEV);
1da177e4 2649
ad706862 2650 rt = ip6_dst_alloc(net, dev, 0);
38308473 2651 if (unlikely(!rt)) {
1da177e4 2652 in6_dev_put(idev);
87a11578 2653 dst = ERR_PTR(-ENOMEM);
1da177e4
LT		 2654 goto out;
2655 }
2656
8e2ec639 2657 rt->dst.flags |= DST_HOST;
588753f1 2658 rt->dst.input = ip6_input;
8e2ec639 2659 rt->dst.output = ip6_output;
550bab42 2660 rt->rt6i_gateway = fl6->daddr;
87a11578 2661 rt->rt6i_dst.addr = fl6->daddr;
8e2ec639
YZ		 2662 rt->rt6i_dst.plen = 128;
2663 rt->rt6i_idev = idev;
14edd87d 2664 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1da177e4 2665
4c981e28 2666 /* Add this dst into uncached_list so that rt6_disable_ip() can
587fea74
WW		 2667 * do proper release of the net_device
2668 */
2669 rt6_uncached_list_add(rt);
81eb8447 2670 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
1da177e4 2671
87a11578
DM		 2672 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
2673
1da177e4 2674out:
87a11578 2675 return dst;
1da177e4
LT		 2676}
2677
569d3645 2678static int ip6_dst_gc(struct dst_ops *ops)
1da177e4 2679{
86393e52 2680 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
7019b78e
DL		 2681 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
2682 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
2683 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
2684 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
2685 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
fc66f95c 2686 int entries;
7019b78e 2687
fc66f95c 2688 entries = dst_entries_get_fast(ops);
49a18d86 2689 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
fc66f95c 2690 entries <= rt_max_size)
1da177e4
LT		 2691 goto out;
2692
6891a346 2693 net->ipv6.ip6_rt_gc_expire++;
14956643 2694 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
fc66f95c
ED		 2695 entries = dst_entries_get_slow(ops);
2696 if (entries < ops->gc_thresh)
7019b78e 2697 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1da177e4 2698out:
7019b78e 2699 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
fc66f95c 2700 return entries > rt_max_size;
1da177e4
LT		 2701}
2702
8c14586f
DA		 2703static struct rt6_info *ip6_nh_lookup_table(struct net *net,
2704 struct fib6_config *cfg,
f4797b33
DA		 2705 const struct in6_addr *gw_addr,
2706 u32 tbid, int flags)
8c14586f
DA		 2707{
2708 struct flowi6 fl6 = {
2709 .flowi6_oif = cfg->fc_ifindex,
2710 .daddr = *gw_addr,
2711 .saddr = cfg->fc_prefsrc,
2712 };
2713 struct fib6_table *table;
2714 struct rt6_info *rt;
8c14586f 2715
f4797b33 2716 table = fib6_get_table(net, tbid);
8c14586f
DA		 2717 if (!table)
2718 return NULL;
2719
2720 if (!ipv6_addr_any(&cfg->fc_prefsrc))
2721 flags |= RT6_LOOKUP_F_HAS_SADDR;
2722
f4797b33 2723 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
b75cc8f9 2724 rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, NULL, flags);
8c14586f
DA		 2725
2726 /* if table lookup failed, fall back to full lookup */
2727 if (rt == net->ipv6.ip6_null_entry) {
2728 ip6_rt_put(rt);
2729 rt = NULL;
2730 }
2731
2732 return rt;
2733}
2734
fc1e64e1
DA		 2735static int ip6_route_check_nh_onlink(struct net *net,
2736 struct fib6_config *cfg,
9fbb704c 2737 const struct net_device *dev,
fc1e64e1
DA		 2738 struct netlink_ext_ack *extack)
2739{
44750f84 2740 u32 tbid = l3mdev_fib_table(dev) ? : RT_TABLE_MAIN;
fc1e64e1
DA		 2741 const struct in6_addr *gw_addr = &cfg->fc_gateway;
2742 u32 flags = RTF_LOCAL | RTF_ANYCAST | RTF_REJECT;
2743 struct rt6_info *grt;
2744 int err;
2745
2746 err = 0;
2747 grt = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0);
2748 if (grt) {
58e354c0 2749 if (!grt->dst.error &&
4ed591c8
DA		 2750 /* ignore match if it is the default route */
2751 grt->from && !ipv6_addr_any(&grt->from->fib6_dst.addr) &&
58e354c0 2752 (grt->rt6i_flags & flags || dev != grt->dst.dev)) {
44750f84
DA		 2753 NL_SET_ERR_MSG(extack,
2754 "Nexthop has invalid gateway or device mismatch");
fc1e64e1
DA		 2755 err = -EINVAL;
2756 }
2757
2758 ip6_rt_put(grt);
2759 }
2760
2761 return err;
2762}
2763
1edce99f
DA		 2764static int ip6_route_check_nh(struct net *net,
2765 struct fib6_config *cfg,
2766 struct net_device **_dev,
2767 struct inet6_dev **idev)
2768{
2769 const struct in6_addr *gw_addr = &cfg->fc_gateway;
2770 struct net_device *dev = _dev ? *_dev : NULL;
2771 struct rt6_info *grt = NULL;
2772 int err = -EHOSTUNREACH;
2773
2774 if (cfg->fc_table) {
f4797b33
DA		 2775 int flags = RT6_LOOKUP_F_IFACE;
2776
2777 grt = ip6_nh_lookup_table(net, cfg, gw_addr,
2778 cfg->fc_table, flags);
1edce99f
DA		 2779 if (grt) {
2780 if (grt->rt6i_flags & RTF_GATEWAY ||
2781 (dev && dev != grt->dst.dev)) {
2782 ip6_rt_put(grt);
2783 grt = NULL;
2784 }
2785 }
2786 }
2787
2788 if (!grt)
b75cc8f9 2789 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, NULL, 1);
1edce99f
DA		 2790
2791 if (!grt)
2792 goto out;
2793
2794 if (dev) {
2795 if (dev != grt->dst.dev) {
2796 ip6_rt_put(grt);
2797 goto out;
2798 }
2799 } else {
2800 *_dev = dev = grt->dst.dev;
2801 *idev = grt->rt6i_idev;
2802 dev_hold(dev);
2803 in6_dev_hold(grt->rt6i_idev);
2804 }
2805
2806 if (!(grt->rt6i_flags & RTF_GATEWAY))
2807 err = 0;
2808
2809 ip6_rt_put(grt);
2810
2811out:
2812 return err;
2813}
2814
9fbb704c
DA		 2815static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
2816 struct net_device **_dev, struct inet6_dev **idev,
2817 struct netlink_ext_ack *extack)
2818{
2819 const struct in6_addr *gw_addr = &cfg->fc_gateway;
2820 int gwa_type = ipv6_addr_type(gw_addr);
232378e8 2821 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
9fbb704c 2822 const struct net_device *dev = *_dev;
232378e8 2823 bool need_addr_check = !dev;
9fbb704c
DA		 2824 int err = -EINVAL;
2825
2826 /* if gw_addr is local we will fail to detect this in case
2827 * address is still TENTATIVE (DAD in progress). rt6_lookup()
2828 * will return already-added prefix route via interface that
2829 * prefix route was assigned to, which might be non-loopback.
2830 */
232378e8
DA		 2831 if (dev &&
2832 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
2833 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
9fbb704c
DA		 2834 goto out;
2835 }
2836
2837 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
2838 /* IPv6 strictly inhibits using not link-local
2839 * addresses as nexthop address.
2840 * Otherwise, router will not able to send redirects.
2841 * It is very good, but in some (rare!) circumstances
2842 * (SIT, PtP, NBMA NOARP links) it is handy to allow
2843 * some exceptions. --ANK
2844 * We allow IPv4-mapped nexthops to support RFC4798-type
2845 * addressing
2846 */
2847 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
2848 NL_SET_ERR_MSG(extack, "Invalid gateway address");
2849 goto out;
2850 }
2851
2852 if (cfg->fc_flags & RTNH_F_ONLINK)
2853 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
2854 else
2855 err = ip6_route_check_nh(net, cfg, _dev, idev);
2856
2857 if (err)
2858 goto out;
2859 }
2860
2861 /* reload in case device was changed */
2862 dev = *_dev;
2863
2864 err = -EINVAL;
2865 if (!dev) {
2866 NL_SET_ERR_MSG(extack, "Egress device not specified");
2867 goto out;
2868 } else if (dev->flags & IFF_LOOPBACK) {
2869 NL_SET_ERR_MSG(extack,
2870 "Egress device can not be loopback device for this route");
2871 goto out;
2872 }
232378e8
DA		 2873
2874 /* if we did not check gw_addr above, do so now that the
2875 * egress device has been resolved.
2876 */
2877 if (need_addr_check &&
2878 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
2879 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
2880 goto out;
2881 }
2882
9fbb704c
DA		 2883 err = 0;
2884out:
2885 return err;
2886}
2887
8d1c802b 2888static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
acb54e3c 2889 gfp_t gfp_flags,
333c4301 2890 struct netlink_ext_ack *extack)
1da177e4 2891{
5578689a 2892 struct net *net = cfg->fc_nlinfo.nl_net;
8d1c802b 2893 struct fib6_info *rt = NULL;
1da177e4
LT		 2894 struct net_device *dev = NULL;
2895 struct inet6_dev *idev = NULL;
c71099ac 2896 struct fib6_table *table;
1da177e4 2897 int addr_type;
8c5b83f0 2898 int err = -EINVAL;
1da177e4 2899
557c44be 2900 /* RTF_PCPU is an internal flag; can not be set by userspace */
d5d531cb
DA		 2901 if (cfg->fc_flags & RTF_PCPU) {
2902 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
557c44be 2903 goto out;
d5d531cb 2904 }
557c44be 2905
2ea2352e
WW		 2906 /* RTF_CACHE is an internal flag; can not be set by userspace */
2907 if (cfg->fc_flags & RTF_CACHE) {
2908 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
2909 goto out;
2910 }
2911
e8478e80
DA		 2912 if (cfg->fc_type > RTN_MAX) {
2913 NL_SET_ERR_MSG(extack, "Invalid route type");
2914 goto out;
2915 }
2916
d5d531cb
DA		 2917 if (cfg->fc_dst_len > 128) {
2918 NL_SET_ERR_MSG(extack, "Invalid prefix length");
2919 goto out;
2920 }
2921 if (cfg->fc_src_len > 128) {
2922 NL_SET_ERR_MSG(extack, "Invalid source address length");
8c5b83f0 2923 goto out;
d5d531cb 2924 }
1da177e4 2925#ifndef CONFIG_IPV6_SUBTREES
d5d531cb
DA		 2926 if (cfg->fc_src_len) {
2927 NL_SET_ERR_MSG(extack,
2928 "Specifying source address requires IPV6_SUBTREES to be enabled");
8c5b83f0 2929 goto out;
d5d531cb 2930 }
1da177e4 2931#endif
86872cb5 2932 if (cfg->fc_ifindex) {
1da177e4 2933 err = -ENODEV;
5578689a 2934 dev = dev_get_by_index(net, cfg->fc_ifindex);
1da177e4
LT		 2935 if (!dev)
2936 goto out;
2937 idev = in6_dev_get(dev);
2938 if (!idev)
2939 goto out;
2940 }
2941
86872cb5
TG		 2942 if (cfg->fc_metric == 0)
2943 cfg->fc_metric = IP6_RT_PRIO_USER;
1da177e4 2944
fc1e64e1
DA		 2945 if (cfg->fc_flags & RTNH_F_ONLINK) {
2946 if (!dev) {
2947 NL_SET_ERR_MSG(extack,
2948 "Nexthop device required for onlink");
2949 err = -ENODEV;
2950 goto out;
2951 }
2952
2953 if (!(dev->flags & IFF_UP)) {
2954 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
2955 err = -ENETDOWN;
2956 goto out;
2957 }
2958 }
2959
d71314b4 2960 err = -ENOBUFS;
38308473
DM		 2961 if (cfg->fc_nlinfo.nlh &&
2962 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
d71314b4 2963 table = fib6_get_table(net, cfg->fc_table);
38308473 2964 if (!table) {
f3213831 2965 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
d71314b4
MV		 2966 table = fib6_new_table(net, cfg->fc_table);
2967 }
2968 } else {
2969 table = fib6_new_table(net, cfg->fc_table);
2970 }
38308473
DM		 2971
2972 if (!table)
c71099ac 2973 goto out;
c71099ac 2974
93531c67
DA		 2975 err = -ENOMEM;
2976 rt = fib6_info_alloc(gfp_flags);
2977 if (!rt)
1da177e4 2978 goto out;
93531c67 2979
d7e774f3
DA		 2980 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
2981 extack);
767a2217
DA		 2982 if (IS_ERR(rt->fib6_metrics)) {
2983 err = PTR_ERR(rt->fib6_metrics);
fda21d46
ED		 2984 /* Do not leave garbage there. */
2985 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
767a2217
DA		 2986 goto out;
2987 }
2988
93531c67
DA		 2989 if (cfg->fc_flags & RTF_ADDRCONF)
2990 rt->dst_nocount = true;
1da177e4 2991
1716a961 2992 if (cfg->fc_flags & RTF_EXPIRES)
14895687 2993 fib6_set_expires(rt, jiffies +
1716a961
G		 2994 clock_t_to_jiffies(cfg->fc_expires));
2995 else
14895687 2996 fib6_clean_expires(rt);
1da177e4 2997
86872cb5
TG		 2998 if (cfg->fc_protocol == RTPROT_UNSPEC)
2999 cfg->fc_protocol = RTPROT_BOOT;
93c2fb25 3000 rt->fib6_protocol = cfg->fc_protocol;
86872cb5
TG		 3001
3002 addr_type = ipv6_addr_type(&cfg->fc_dst);
1da177e4 3003
19e42e45
RP		 3004 if (cfg->fc_encap) {
3005 struct lwtunnel_state *lwtstate;
3006
30357d7d 3007 err = lwtunnel_build_state(cfg->fc_encap_type,
127eb7cd 3008 cfg->fc_encap, AF_INET6, cfg,
9ae28727 3009 &lwtstate, extack);
19e42e45
RP		 3010 if (err)
3011 goto out;
5e670d84 3012 rt->fib6_nh.nh_lwtstate = lwtstate_get(lwtstate);
19e42e45
RP		 3013 }
3014
93c2fb25
DA		 3015 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3016 rt->fib6_dst.plen = cfg->fc_dst_len;
3017 if (rt->fib6_dst.plen == 128)
3b6761d1 3018 rt->dst_host = true;
e5fd387a 3019
1da177e4 3020#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 3021 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3022 rt->fib6_src.plen = cfg->fc_src_len;
1da177e4
LT		 3023#endif
3024
93c2fb25 3025 rt->fib6_metric = cfg->fc_metric;
5e670d84 3026 rt->fib6_nh.nh_weight = 1;
1da177e4 3027
e8478e80 3028 rt->fib6_type = cfg->fc_type;
1da177e4
LT		 3029
3030 /* We cannot add true routes via loopback here,
3031 they would result in kernel looping; promote them to reject routes
3032 */
86872cb5 3033 if ((cfg->fc_flags & RTF_REJECT) ||
38308473
DM		 3034 (dev && (dev->flags & IFF_LOOPBACK) &&
3035 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3036 !(cfg->fc_flags & RTF_LOCAL))) {
1da177e4 3037 /* hold loopback dev/idev if we haven't done so. */
5578689a 3038 if (dev != net->loopback_dev) {
1da177e4
LT		 3039 if (dev) {
3040 dev_put(dev);
3041 in6_dev_put(idev);
3042 }
5578689a 3043 dev = net->loopback_dev;
1da177e4
LT		 3044 dev_hold(dev);
3045 idev = in6_dev_get(dev);
3046 if (!idev) {
3047 err = -ENODEV;
3048 goto out;
3049 }
3050 }
93c2fb25 3051 rt->fib6_flags = RTF_REJECT|RTF_NONEXTHOP;
1da177e4
LT		 3052 goto install_route;
3053 }
3054
86872cb5 3055 if (cfg->fc_flags & RTF_GATEWAY) {
9fbb704c
DA		 3056 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3057 if (err)
48ed7b26 3058 goto out;
1da177e4 3059
93531c67 3060 rt->fib6_nh.nh_gw = cfg->fc_gateway;
1da177e4
LT		 3061 }
3062
3063 err = -ENODEV;
38308473 3064 if (!dev)
1da177e4
LT		 3065 goto out;
3066
428604fb
LB		 3067 if (idev->cnf.disable_ipv6) {
3068 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3069 err = -EACCES;
3070 goto out;
3071 }
3072
955ec4cb
DA		 3073 if (!(dev->flags & IFF_UP)) {
3074 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3075 err = -ENETDOWN;
3076 goto out;
3077 }
3078
c3968a85
DW		 3079 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3080 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
d5d531cb 3081 NL_SET_ERR_MSG(extack, "Invalid source address");
c3968a85
DW		 3082 err = -EINVAL;
3083 goto out;
3084 }
93c2fb25
DA		 3085 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3086 rt->fib6_prefsrc.plen = 128;
c3968a85 3087 } else
93c2fb25 3088 rt->fib6_prefsrc.plen = 0;
c3968a85 3089
93c2fb25 3090 rt->fib6_flags = cfg->fc_flags;
1da177e4
LT		 3091
3092install_route:
93c2fb25 3093 if (!(rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
5609b80a 3094 !netif_carrier_ok(dev))
5e670d84
DA		 3095 rt->fib6_nh.nh_flags |= RTNH_F_LINKDOWN;
3096 rt->fib6_nh.nh_flags |= (cfg->fc_flags & RTNH_F_ONLINK);
93531c67 3097 rt->fib6_nh.nh_dev = dev;
93c2fb25 3098 rt->fib6_table = table;
63152fc0 3099
dcd1f572
DA		 3100 if (idev)
3101 in6_dev_put(idev);
3102
8c5b83f0 3103 return rt;
6b9ea5a6
RP		 3104out:
3105 if (dev)
3106 dev_put(dev);
3107 if (idev)
3108 in6_dev_put(idev);
6b9ea5a6 3109
93531c67 3110 fib6_info_release(rt);
8c5b83f0 3111 return ERR_PTR(err);
6b9ea5a6
RP		 3112}
3113
acb54e3c 3114int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
333c4301 3115 struct netlink_ext_ack *extack)
6b9ea5a6 3116{
8d1c802b 3117 struct fib6_info *rt;
6b9ea5a6
RP		 3118 int err;
3119
acb54e3c 3120 rt = ip6_route_info_create(cfg, gfp_flags, extack);
d4ead6b3
DA		 3121 if (IS_ERR(rt))
3122 return PTR_ERR(rt);
6b9ea5a6 3123
d4ead6b3 3124 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
93531c67 3125 fib6_info_release(rt);
6b9ea5a6 3126
1da177e4
LT		 3127 return err;
3128}
3129
8d1c802b 3130static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
1da177e4 3131{
afb1d4b5 3132 struct net *net = info->nl_net;
c71099ac 3133 struct fib6_table *table;
afb1d4b5 3134 int err;
1da177e4 3135
421842ed 3136 if (rt == net->ipv6.fib6_null_entry) {
6825a26c
G		 3137 err = -ENOENT;
3138 goto out;
3139 }
6c813a72 3140
93c2fb25 3141 table = rt->fib6_table;
66f5d6ce 3142 spin_lock_bh(&table->tb6_lock);
86872cb5 3143 err = fib6_del(rt, info);
66f5d6ce 3144 spin_unlock_bh(&table->tb6_lock);
1da177e4 3145
6825a26c 3146out:
93531c67 3147 fib6_info_release(rt);
1da177e4
LT		 3148 return err;
3149}
3150
8d1c802b 3151int ip6_del_rt(struct net *net, struct fib6_info *rt)
e0a1ad73 3152{
afb1d4b5
DA		 3153 struct nl_info info = { .nl_net = net };
3154
528c4ceb 3155 return __ip6_del_rt(rt, &info);
e0a1ad73
TG		 3156}
3157
8d1c802b 3158static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
0ae81335
DA		 3159{
3160 struct nl_info *info = &cfg->fc_nlinfo;
e3330039 3161 struct net *net = info->nl_net;
16a16cd3 3162 struct sk_buff *skb = NULL;
0ae81335 3163 struct fib6_table *table;
e3330039 3164 int err = -ENOENT;
0ae81335 3165
421842ed 3166 if (rt == net->ipv6.fib6_null_entry)
e3330039 3167 goto out_put;
93c2fb25 3168 table = rt->fib6_table;
66f5d6ce 3169 spin_lock_bh(&table->tb6_lock);
0ae81335 3170
93c2fb25 3171 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
8d1c802b 3172 struct fib6_info *sibling, *next_sibling;
0ae81335 3173
16a16cd3
DA		 3174 /* prefer to send a single notification with all hops */
3175 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3176 if (skb) {
3177 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3178
d4ead6b3 3179 if (rt6_fill_node(net, skb, rt, NULL,
16a16cd3
DA		 3180 NULL, NULL, 0, RTM_DELROUTE,
3181 info->portid, seq, 0) < 0) {
3182 kfree_skb(skb);
3183 skb = NULL;
3184 } else
3185 info->skip_notify = 1;
3186 }
3187
0ae81335 3188 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25
DA		 3189 &rt->fib6_siblings,
3190 fib6_siblings) {
0ae81335
DA		 3191 err = fib6_del(sibling, info);
3192 if (err)
e3330039 3193 goto out_unlock;
0ae81335
DA		 3194 }
3195 }
3196
3197 err = fib6_del(rt, info);
e3330039 3198out_unlock:
66f5d6ce 3199 spin_unlock_bh(&table->tb6_lock);
e3330039 3200out_put:
93531c67 3201 fib6_info_release(rt);
16a16cd3
DA		 3202
3203 if (skb) {
e3330039 3204 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
16a16cd3
DA		 3205 info->nlh, gfp_any());
3206 }
0ae81335
DA		 3207 return err;
3208}
3209
23fb93a4
DA		 3210static int ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3211{
3212 int rc = -ESRCH;
3213
3214 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3215 goto out;
3216
3217 if (cfg->fc_flags & RTF_GATEWAY &&
3218 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3219 goto out;
761f6026
XL		 3220
3221 rc = rt6_remove_exception_rt(rt);
23fb93a4
DA		 3222out:
3223 return rc;
3224}
3225
333c4301
DA		 3226static int ip6_route_del(struct fib6_config *cfg,
3227 struct netlink_ext_ack *extack)
1da177e4 3228{
8d1c802b 3229 struct rt6_info *rt_cache;
c71099ac 3230 struct fib6_table *table;
8d1c802b 3231 struct fib6_info *rt;
1da177e4 3232 struct fib6_node *fn;
1da177e4
LT		 3233 int err = -ESRCH;
3234
5578689a 3235 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
d5d531cb
DA		 3236 if (!table) {
3237 NL_SET_ERR_MSG(extack, "FIB table does not exist");
c71099ac 3238 return err;
d5d531cb 3239 }
c71099ac 3240
66f5d6ce 3241 rcu_read_lock();
1da177e4 3242
c71099ac 3243 fn = fib6_locate(&table->tb6_root,
86872cb5 3244 &cfg->fc_dst, cfg->fc_dst_len,
38fbeeee 3245 &cfg->fc_src, cfg->fc_src_len,
2b760fcf 3246 !(cfg->fc_flags & RTF_CACHE));
1ab1457c 3247
1da177e4 3248 if (fn) {
66f5d6ce 3249 for_each_fib6_node_rt_rcu(fn) {
2b760fcf 3250 if (cfg->fc_flags & RTF_CACHE) {
23fb93a4
DA		 3251 int rc;
3252
2b760fcf
WW		 3253 rt_cache = rt6_find_cached_rt(rt, &cfg->fc_dst,
3254 &cfg->fc_src);
23fb93a4
DA		 3255 if (rt_cache) {
3256 rc = ip6_del_cached_rt(rt_cache, cfg);
9e575010
ED		 3257 if (rc != -ESRCH) {
3258 rcu_read_unlock();
23fb93a4 3259 return rc;
9e575010 3260 }
23fb93a4
DA		 3261 }
3262 continue;
2b760fcf 3263 }
86872cb5 3264 if (cfg->fc_ifindex &&
5e670d84
DA		 3265 (!rt->fib6_nh.nh_dev ||
3266 rt->fib6_nh.nh_dev->ifindex != cfg->fc_ifindex))
1da177e4 3267 continue;
86872cb5 3268 if (cfg->fc_flags & RTF_GATEWAY &&
5e670d84 3269 !ipv6_addr_equal(&cfg->fc_gateway, &rt->fib6_nh.nh_gw))
1da177e4 3270 continue;
93c2fb25 3271 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
1da177e4 3272 continue;
93c2fb25 3273 if (cfg->fc_protocol && cfg->fc_protocol != rt->fib6_protocol)
c2ed1880 3274 continue;
e873e4b9
WW		 3275 if (!fib6_info_hold_safe(rt))
3276 continue;
66f5d6ce 3277 rcu_read_unlock();
1da177e4 3278
0ae81335
DA		 3279 /* if gateway was specified only delete the one hop */
3280 if (cfg->fc_flags & RTF_GATEWAY)
3281 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3282
3283 return __ip6_del_rt_siblings(rt, cfg);
1da177e4
LT		 3284 }
3285 }
66f5d6ce 3286 rcu_read_unlock();
1da177e4
LT		 3287
3288 return err;
3289}
3290
6700c270 3291static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
a6279458 3292{
a6279458 3293 struct netevent_redirect netevent;
e8599ff4 3294 struct rt6_info *rt, *nrt = NULL;
e8599ff4
DM		 3295 struct ndisc_options ndopts;
3296 struct inet6_dev *in6_dev;
3297 struct neighbour *neigh;
a68886a6 3298 struct fib6_info *from;
71bcdba0 3299 struct rd_msg *msg;
6e157b6a
DM		 3300 int optlen, on_link;
3301 u8 *lladdr;
e8599ff4 3302
29a3cad5 3303 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
71bcdba0 3304 optlen -= sizeof(*msg);
e8599ff4
DM		 3305
3306 if (optlen < 0) {
6e157b6a 3307 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
e8599ff4
DM		 3308 return;
3309 }
3310
71bcdba0 3311 msg = (struct rd_msg *)icmp6_hdr(skb);
e8599ff4 3312
71bcdba0 3313 if (ipv6_addr_is_multicast(&msg->dest)) {
6e157b6a 3314 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
e8599ff4
DM		 3315 return;
3316 }
3317
6e157b6a 3318 on_link = 0;
71bcdba0 3319 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
e8599ff4 3320 on_link = 1;
71bcdba0 3321 } else if (ipv6_addr_type(&msg->target) !=
e8599ff4 3322 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
6e157b6a 3323 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
e8599ff4
DM		 3324 return;
3325 }
3326
3327 in6_dev = __in6_dev_get(skb->dev);
3328 if (!in6_dev)
3329 return;
3330 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
3331 return;
3332
3333 /* RFC2461 8.1:
3334 * The IP source address of the Redirect MUST be the same as the current
3335 * first-hop router for the specified ICMP Destination Address.
3336 */
3337
f997c55c 3338 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
e8599ff4
DM		 3339 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
3340 return;
3341 }
6e157b6a
DM		 3342
3343 lladdr = NULL;
e8599ff4
DM		 3344 if (ndopts.nd_opts_tgt_lladdr) {
3345 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
3346 skb->dev);
3347 if (!lladdr) {
3348 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
3349 return;
3350 }
3351 }
3352
6e157b6a 3353 rt = (struct rt6_info *) dst;
ec13ad1d 3354 if (rt->rt6i_flags & RTF_REJECT) {
6e157b6a 3355 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
e8599ff4 3356 return;
6e157b6a 3357 }
e8599ff4 3358
6e157b6a
DM		 3359 /* Redirect received -> path was valid.
3360 * Look, redirects are sent only in response to data packets,
3361 * so that this nexthop apparently is reachable. --ANK
3362 */
0dec879f 3363 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
a6279458 3364
71bcdba0 3365 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
6e157b6a
DM		 3366 if (!neigh)
3367 return;
a6279458 3368
1da177e4
LT		 3369 /*
3370 * We have finally decided to accept it.
3371 */
3372
f997c55c 3373 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1da177e4
LT		 3374 NEIGH_UPDATE_F_WEAK_OVERRIDE|
3375 NEIGH_UPDATE_F_OVERRIDE|
3376 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
f997c55c
AA		 3377 NEIGH_UPDATE_F_ISROUTER)),
3378 NDISC_REDIRECT, &ndopts);
1da177e4 3379
4d85cd0c 3380 rcu_read_lock();
a68886a6 3381 from = rcu_dereference(rt->from);
e873e4b9
WW		 3382 /* This fib6_info_hold() is safe here because we hold reference to rt
3383 * and rt already holds reference to fib6_info.
3384 */
8a14e46f 3385 fib6_info_hold(from);
4d85cd0c 3386 rcu_read_unlock();
8a14e46f
DA		 3387
3388 nrt = ip6_rt_cache_alloc(from, &msg->dest, NULL);
38308473 3389 if (!nrt)
1da177e4
LT		 3390 goto out;
3391
3392 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
3393 if (on_link)
3394 nrt->rt6i_flags &= ~RTF_GATEWAY;
3395
4e3fd7a0 3396 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1da177e4 3397
2b760fcf
WW		 3398 /* No need to remove rt from the exception table if rt is
3399 * a cached route because rt6_insert_exception() will
3400 * takes care of it
3401 */
8a14e46f 3402 if (rt6_insert_exception(nrt, from)) {
2b760fcf
WW		 3403 dst_release_immediate(&nrt->dst);
3404 goto out;
3405 }
1da177e4 3406
d8d1f30b
CG		 3407 netevent.old = &rt->dst;
3408 netevent.new = &nrt->dst;
71bcdba0 3409 netevent.daddr = &msg->dest;
60592833 3410 netevent.neigh = neigh;
8d71740c
TT		 3411 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
3412
1da177e4 3413out:
8a14e46f 3414 fib6_info_release(from);
e8599ff4 3415 neigh_release(neigh);
6e157b6a
DM		 3416}
3417
70ceb4f5 3418#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 3419static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 3420 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 3421 const struct in6_addr *gwaddr,
3422 struct net_device *dev)
70ceb4f5 3423{
830218c1
DA		 3424 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
3425 int ifindex = dev->ifindex;
70ceb4f5 3426 struct fib6_node *fn;
8d1c802b 3427 struct fib6_info *rt = NULL;
c71099ac
TG		 3428 struct fib6_table *table;
3429
830218c1 3430 table = fib6_get_table(net, tb_id);
38308473 3431 if (!table)
c71099ac 3432 return NULL;
70ceb4f5 3433
66f5d6ce 3434 rcu_read_lock();
38fbeeee 3435 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
70ceb4f5
YH		 3436 if (!fn)
3437 goto out;
3438
66f5d6ce 3439 for_each_fib6_node_rt_rcu(fn) {
5e670d84 3440 if (rt->fib6_nh.nh_dev->ifindex != ifindex)
70ceb4f5 3441 continue;
93c2fb25 3442 if ((rt->fib6_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
70ceb4f5 3443 continue;
5e670d84 3444 if (!ipv6_addr_equal(&rt->fib6_nh.nh_gw, gwaddr))
70ceb4f5 3445 continue;
e873e4b9
WW		 3446 if (!fib6_info_hold_safe(rt))
3447 continue;
70ceb4f5
YH		 3448 break;
3449 }
3450out:
66f5d6ce 3451 rcu_read_unlock();
70ceb4f5
YH		 3452 return rt;
3453}
3454
8d1c802b 3455static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 3456 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 3457 const struct in6_addr *gwaddr,
3458 struct net_device *dev,
95c96174 3459 unsigned int pref)
70ceb4f5 3460{
86872cb5 3461 struct fib6_config cfg = {
238fc7ea 3462 .fc_metric = IP6_RT_PRIO_USER,
830218c1 3463 .fc_ifindex = dev->ifindex,
86872cb5
TG		 3464 .fc_dst_len = prefixlen,
3465 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
3466 RTF_UP | RTF_PREF(pref),
b91d5329 3467 .fc_protocol = RTPROT_RA,
e8478e80 3468 .fc_type = RTN_UNICAST,
15e47304 3469 .fc_nlinfo.portid = 0,
efa2cea0
DL		 3470 .fc_nlinfo.nlh = NULL,
3471 .fc_nlinfo.nl_net = net,
86872cb5
TG		 3472 };
3473
830218c1 3474 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
4e3fd7a0
AD		 3475 cfg.fc_dst = *prefix;
3476 cfg.fc_gateway = *gwaddr;
70ceb4f5 3477
e317da96
YH		 3478 /* We should treat it as a default route if prefix length is 0. */
3479 if (!prefixlen)
86872cb5 3480 cfg.fc_flags |= RTF_DEFAULT;
70ceb4f5 3481
acb54e3c 3482 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
70ceb4f5 3483
830218c1 3484 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
70ceb4f5
YH		 3485}
3486#endif
3487
8d1c802b 3488struct fib6_info *rt6_get_dflt_router(struct net *net,
afb1d4b5
DA		 3489 const struct in6_addr *addr,
3490 struct net_device *dev)
1ab1457c 3491{
830218c1 3492 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
8d1c802b 3493 struct fib6_info *rt;
c71099ac 3494 struct fib6_table *table;
1da177e4 3495
afb1d4b5 3496 table = fib6_get_table(net, tb_id);
38308473 3497 if (!table)
c71099ac 3498 return NULL;
1da177e4 3499
66f5d6ce
WW		 3500 rcu_read_lock();
3501 for_each_fib6_node_rt_rcu(&table->tb6_root) {
5e670d84 3502 if (dev == rt->fib6_nh.nh_dev &&
93c2fb25 3503 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
5e670d84 3504 ipv6_addr_equal(&rt->fib6_nh.nh_gw, addr))
1da177e4
LT		 3505 break;
3506 }
e873e4b9
WW		 3507 if (rt && !fib6_info_hold_safe(rt))
3508 rt = NULL;
66f5d6ce 3509 rcu_read_unlock();
1da177e4
LT		 3510 return rt;
3511}
3512
8d1c802b 3513struct fib6_info *rt6_add_dflt_router(struct net *net,
afb1d4b5 3514 const struct in6_addr *gwaddr,
ebacaaa0
YH		 3515 struct net_device *dev,
3516 unsigned int pref)
1da177e4 3517{
86872cb5 3518 struct fib6_config cfg = {
ca254490 3519 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
238fc7ea 3520 .fc_metric = IP6_RT_PRIO_USER,
86872cb5
TG		 3521 .fc_ifindex = dev->ifindex,
3522 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
3523 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
b91d5329 3524 .fc_protocol = RTPROT_RA,
e8478e80 3525 .fc_type = RTN_UNICAST,
15e47304 3526 .fc_nlinfo.portid = 0,
5578689a 3527 .fc_nlinfo.nlh = NULL,
afb1d4b5 3528 .fc_nlinfo.nl_net = net,
86872cb5 3529 };
1da177e4 3530
4e3fd7a0 3531 cfg.fc_gateway = *gwaddr;
1da177e4 3532
acb54e3c 3533 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
830218c1
DA		 3534 struct fib6_table *table;
3535
3536 table = fib6_get_table(dev_net(dev), cfg.fc_table);
3537 if (table)
3538 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
3539 }
1da177e4 3540
afb1d4b5 3541 return rt6_get_dflt_router(net, gwaddr, dev);
1da177e4
LT		 3542}
3543
afb1d4b5
DA		 3544static void __rt6_purge_dflt_routers(struct net *net,
3545 struct fib6_table *table)
1da177e4 3546{
8d1c802b 3547 struct fib6_info *rt;
1da177e4
LT		 3548
3549restart:
66f5d6ce
WW		 3550 rcu_read_lock();
3551 for_each_fib6_node_rt_rcu(&table->tb6_root) {
dcd1f572
DA		 3552 struct net_device *dev = fib6_info_nh_dev(rt);
3553 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
3554
93c2fb25 3555 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
e873e4b9
WW		 3556 (!idev || idev->cnf.accept_ra != 2) &&
3557 fib6_info_hold_safe(rt)) {
93531c67
DA		 3558 rcu_read_unlock();
3559 ip6_del_rt(net, rt);
1da177e4
LT		 3560 goto restart;
3561 }
3562 }
66f5d6ce 3563 rcu_read_unlock();
830218c1
DA		 3564
3565 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
3566}
3567
3568void rt6_purge_dflt_routers(struct net *net)
3569{
3570 struct fib6_table *table;
3571 struct hlist_head *head;
3572 unsigned int h;
3573
3574 rcu_read_lock();
3575
3576 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
3577 head = &net->ipv6.fib_table_hash[h];
3578 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
3579 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
afb1d4b5 3580 __rt6_purge_dflt_routers(net, table);
830218c1
DA		 3581 }
3582 }
3583
3584 rcu_read_unlock();
1da177e4
LT		 3585}
3586
5578689a
DL		 3587static void rtmsg_to_fib6_config(struct net *net,
3588 struct in6_rtmsg *rtmsg,
86872cb5
TG		 3589 struct fib6_config *cfg)
3590{
8823a3ac
MŻ		 3591 *cfg = (struct fib6_config){
3592 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
3593 : RT6_TABLE_MAIN,
3594 .fc_ifindex = rtmsg->rtmsg_ifindex,
3595 .fc_metric = rtmsg->rtmsg_metric,
3596 .fc_expires = rtmsg->rtmsg_info,
3597 .fc_dst_len = rtmsg->rtmsg_dst_len,
3598 .fc_src_len = rtmsg->rtmsg_src_len,
3599 .fc_flags = rtmsg->rtmsg_flags,
3600 .fc_type = rtmsg->rtmsg_type,
3601
3602 .fc_nlinfo.nl_net = net,
3603
3604 .fc_dst = rtmsg->rtmsg_dst,
3605 .fc_src = rtmsg->rtmsg_src,
3606 .fc_gateway = rtmsg->rtmsg_gateway,
3607 };
86872cb5
TG		 3608}
3609
5578689a 3610int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1da177e4 3611{
86872cb5 3612 struct fib6_config cfg;
1da177e4
LT		 3613 struct in6_rtmsg rtmsg;
3614 int err;
3615
67ba4152 3616 switch (cmd) {
1da177e4
LT		 3617 case SIOCADDRT: /* Add a route */
3618 case SIOCDELRT: /* Delete a route */
af31f412 3619 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4
LT		 3620 return -EPERM;
3621 err = copy_from_user(&rtmsg, arg,
3622 sizeof(struct in6_rtmsg));
3623 if (err)
3624 return -EFAULT;
86872cb5 3625
5578689a 3626 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
86872cb5 3627
1da177e4
LT		 3628 rtnl_lock();
3629 switch (cmd) {
3630 case SIOCADDRT:
acb54e3c 3631 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
1da177e4
LT		 3632 break;
3633 case SIOCDELRT:
333c4301 3634 err = ip6_route_del(&cfg, NULL);
1da177e4
LT		 3635 break;
3636 default:
3637 err = -EINVAL;
3638 }
3639 rtnl_unlock();
3640
3641 return err;
3ff50b79 3642 }
1da177e4
LT		 3643
3644 return -EINVAL;
3645}
3646
3647/*
3648 * Drop the packet on the floor
3649 */
3650
d5fdd6ba 3651static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1da177e4 3652{
612f09e8 3653 int type;
adf30907 3654 struct dst_entry *dst = skb_dst(skb);
612f09e8
YH		 3655 switch (ipstats_mib_noroutes) {
3656 case IPSTATS_MIB_INNOROUTES:
0660e03f 3657 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
45bb0060 3658 if (type == IPV6_ADDR_ANY) {
bdb7cc64
SS		 3659 IP6_INC_STATS(dev_net(dst->dev),
3660 __in6_dev_get_safely(skb->dev),
3bd653c8 3661 IPSTATS_MIB_INADDRERRORS);
612f09e8
YH		 3662 break;
3663 }
3664 /* FALLTHROUGH */
3665 case IPSTATS_MIB_OUTNOROUTES:
3bd653c8
DL		 3666 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
3667 ipstats_mib_noroutes);
612f09e8
YH		 3668 break;
3669 }
3ffe533c 3670 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1da177e4
LT		 3671 kfree_skb(skb);
3672 return 0;
3673}
3674
9ce8ade0
TG		 3675static int ip6_pkt_discard(struct sk_buff *skb)
3676{
612f09e8 3677 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 3678}
3679
ede2059d 3680static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
1da177e4 3681{
adf30907 3682 skb->dev = skb_dst(skb)->dev;
612f09e8 3683 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1da177e4
LT		 3684}
3685
9ce8ade0
TG		 3686static int ip6_pkt_prohibit(struct sk_buff *skb)
3687{
612f09e8 3688 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 3689}
3690
ede2059d 3691static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
9ce8ade0 3692{
adf30907 3693 skb->dev = skb_dst(skb)->dev;
612f09e8 3694 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
9ce8ade0
TG		 3695}
3696
1da177e4
LT		 3697/*
3698 * Allocate a dst for local (unicast / anycast) address.
3699 */
3700
360a9887
DA		 3701struct fib6_info *addrconf_f6i_alloc(struct net *net,
3702 struct inet6_dev *idev,
3703 const struct in6_addr *addr,
3704 bool anycast, gfp_t gfp_flags)
1da177e4 3705{
ca254490 3706 u32 tb_id;
4832c30d 3707 struct net_device *dev = idev->dev;
360a9887 3708 struct fib6_info *f6i;
5f02ce24 3709
360a9887
DA		 3710 f6i = fib6_info_alloc(gfp_flags);
3711 if (!f6i)
1da177e4
LT		 3712 return ERR_PTR(-ENOMEM);
3713
d7e774f3 3714 f6i->fib6_metrics = ip_fib_metrics_init(net, NULL, 0, NULL);
360a9887 3715 f6i->dst_nocount = true;
360a9887
DA		 3716 f6i->dst_host = true;
3717 f6i->fib6_protocol = RTPROT_KERNEL;
3718 f6i->fib6_flags = RTF_UP | RTF_NONEXTHOP;
e8478e80 3719 if (anycast) {
360a9887
DA		 3720 f6i->fib6_type = RTN_ANYCAST;
3721 f6i->fib6_flags |= RTF_ANYCAST;
e8478e80 3722 } else {
360a9887
DA		 3723 f6i->fib6_type = RTN_LOCAL;
3724 f6i->fib6_flags |= RTF_LOCAL;
e8478e80 3725 }
1da177e4 3726
360a9887 3727 f6i->fib6_nh.nh_gw = *addr;
93531c67 3728 dev_hold(dev);
360a9887
DA		 3729 f6i->fib6_nh.nh_dev = dev;
3730 f6i->fib6_dst.addr = *addr;
3731 f6i->fib6_dst.plen = 128;
ca254490 3732 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
360a9887 3733 f6i->fib6_table = fib6_get_table(net, tb_id);
1da177e4 3734
360a9887 3735 return f6i;
1da177e4
LT		 3736}
3737
c3968a85
DW		 3738/* remove deleted ip from prefsrc entries */
3739struct arg_dev_net_ip {
3740 struct net_device *dev;
3741 struct net *net;
3742 struct in6_addr *addr;
3743};
3744
8d1c802b 3745static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
c3968a85
DW		 3746{
3747 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
3748 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
3749 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
3750
5e670d84 3751 if (((void *)rt->fib6_nh.nh_dev == dev || !dev) &&
421842ed 3752 rt != net->ipv6.fib6_null_entry &&
93c2fb25 3753 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
60006a48 3754 spin_lock_bh(&rt6_exception_lock);
c3968a85 3755 /* remove prefsrc entry */
93c2fb25 3756 rt->fib6_prefsrc.plen = 0;
60006a48 3757 spin_unlock_bh(&rt6_exception_lock);
c3968a85
DW		 3758 }
3759 return 0;
3760}
3761
3762void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
3763{
3764 struct net *net = dev_net(ifp->idev->dev);
3765 struct arg_dev_net_ip adni = {
3766 .dev = ifp->idev->dev,
3767 .net = net,
3768 .addr = &ifp->addr,
3769 };
0c3584d5 3770 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
c3968a85
DW		 3771}
3772
be7a010d 3773#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
be7a010d
DJ		 3774
3775/* Remove routers and update dst entries when gateway turn into host. */
8d1c802b 3776static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
be7a010d
DJ		 3777{
3778 struct in6_addr *gateway = (struct in6_addr *)arg;
3779
93c2fb25 3780 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
5e670d84 3781 ipv6_addr_equal(gateway, &rt->fib6_nh.nh_gw)) {
be7a010d
DJ		 3782 return -1;
3783 }
b16cb459
WW		 3784
3785 /* Further clean up cached routes in exception table.
3786 * This is needed because cached route may have a different
3787 * gateway than its 'parent' in the case of an ip redirect.
3788 */
3789 rt6_exceptions_clean_tohost(rt, gateway);
3790
be7a010d
DJ		 3791 return 0;
3792}
3793
3794void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
3795{
3796 fib6_clean_all(net, fib6_clean_tohost, gateway);
3797}
3798
2127d95a
IS		 3799struct arg_netdev_event {
3800 const struct net_device *dev;
4c981e28
IS		 3801 union {
3802 unsigned int nh_flags;
3803 unsigned long event;
3804 };
2127d95a
IS		 3805};
3806
8d1c802b 3807static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
d7dedee1 3808{
8d1c802b 3809 struct fib6_info *iter;
d7dedee1
IS		 3810 struct fib6_node *fn;
3811
93c2fb25
DA		 3812 fn = rcu_dereference_protected(rt->fib6_node,
3813 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 3814 iter = rcu_dereference_protected(fn->leaf,
93c2fb25 3815 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 3816 while (iter) {
93c2fb25 3817 if (iter->fib6_metric == rt->fib6_metric &&
33bd5ac5 3818 rt6_qualify_for_ecmp(iter))
d7dedee1 3819 return iter;
8fb11a9a 3820 iter = rcu_dereference_protected(iter->fib6_next,
93c2fb25 3821 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1
IS		 3822 }
3823
3824 return NULL;
3825}
3826
8d1c802b 3827static bool rt6_is_dead(const struct fib6_info *rt)
d7dedee1 3828{
5e670d84
DA		 3829 if (rt->fib6_nh.nh_flags & RTNH_F_DEAD ||
3830 (rt->fib6_nh.nh_flags & RTNH_F_LINKDOWN &&
dcd1f572 3831 fib6_ignore_linkdown(rt)))
d7dedee1
IS		 3832 return true;
3833
3834 return false;
3835}
3836
8d1c802b 3837static int rt6_multipath_total_weight(const struct fib6_info *rt)
d7dedee1 3838{
8d1c802b 3839 struct fib6_info *iter;
d7dedee1
IS		 3840 int total = 0;
3841
3842 if (!rt6_is_dead(rt))
5e670d84 3843 total += rt->fib6_nh.nh_weight;
d7dedee1 3844
93c2fb25 3845 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
d7dedee1 3846 if (!rt6_is_dead(iter))
5e670d84 3847 total += iter->fib6_nh.nh_weight;
d7dedee1
IS		 3848 }
3849
3850 return total;
3851}
3852
8d1c802b 3853static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
d7dedee1
IS		 3854{
3855 int upper_bound = -1;
3856
3857 if (!rt6_is_dead(rt)) {
5e670d84 3858 *weight += rt->fib6_nh.nh_weight;
d7dedee1
IS		 3859 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
3860 total) - 1;
3861 }
5e670d84 3862 atomic_set(&rt->fib6_nh.nh_upper_bound, upper_bound);
d7dedee1
IS		 3863}
3864
8d1c802b 3865static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
d7dedee1 3866{
8d1c802b 3867 struct fib6_info *iter;
d7dedee1
IS		 3868 int weight = 0;
3869
3870 rt6_upper_bound_set(rt, &weight, total);
3871
93c2fb25 3872 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
d7dedee1
IS		 3873 rt6_upper_bound_set(iter, &weight, total);
3874}
3875
8d1c802b 3876void rt6_multipath_rebalance(struct fib6_info *rt)
d7dedee1 3877{
8d1c802b 3878 struct fib6_info *first;
d7dedee1
IS		 3879 int total;
3880
3881 /* In case the entire multipath route was marked for flushing,
3882 * then there is no need to rebalance upon the removal of every
3883 * sibling route.
3884 */
93c2fb25 3885 if (!rt->fib6_nsiblings || rt->should_flush)
d7dedee1
IS		 3886 return;
3887
3888 /* During lookup routes are evaluated in order, so we need to
3889 * make sure upper bounds are assigned from the first sibling
3890 * onwards.
3891 */
3892 first = rt6_multipath_first_sibling(rt);
3893 if (WARN_ON_ONCE(!first))
3894 return;
3895
3896 total = rt6_multipath_total_weight(first);
3897 rt6_multipath_upper_bound_set(first, total);
3898}
3899
8d1c802b 3900static int fib6_ifup(struct fib6_info *rt, void *p_arg)
2127d95a
IS		 3901{
3902 const struct arg_netdev_event *arg = p_arg;
7aef6859 3903 struct net *net = dev_net(arg->dev);
2127d95a 3904
421842ed 3905 if (rt != net->ipv6.fib6_null_entry && rt->fib6_nh.nh_dev == arg->dev) {
5e670d84 3906 rt->fib6_nh.nh_flags &= ~arg->nh_flags;
7aef6859 3907 fib6_update_sernum_upto_root(net, rt);
d7dedee1 3908 rt6_multipath_rebalance(rt);
1de178ed 3909 }
2127d95a
IS		 3910
3911 return 0;
3912}
3913
3914void rt6_sync_up(struct net_device *dev, unsigned int nh_flags)
3915{
3916 struct arg_netdev_event arg = {
3917 .dev = dev,
6802f3ad
IS		 3918 {
3919 .nh_flags = nh_flags,
3920 },
2127d95a
IS		 3921 };
3922
3923 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
3924 arg.nh_flags |= RTNH_F_LINKDOWN;
3925
3926 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
3927}
3928
8d1c802b 3929static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
1de178ed
IS		 3930 const struct net_device *dev)
3931{
8d1c802b 3932 struct fib6_info *iter;
1de178ed 3933
5e670d84 3934 if (rt->fib6_nh.nh_dev == dev)
1de178ed 3935 return true;
93c2fb25 3936 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
5e670d84 3937 if (iter->fib6_nh.nh_dev == dev)
1de178ed
IS		 3938 return true;
3939
3940 return false;
3941}
3942
8d1c802b 3943static void rt6_multipath_flush(struct fib6_info *rt)
1de178ed 3944{
8d1c802b 3945 struct fib6_info *iter;
1de178ed
IS		 3946
3947 rt->should_flush = 1;
93c2fb25 3948 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1de178ed
IS		 3949 iter->should_flush = 1;
3950}
3951
8d1c802b 3952static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
1de178ed
IS		 3953 const struct net_device *down_dev)
3954{
8d1c802b 3955 struct fib6_info *iter;
1de178ed
IS		 3956 unsigned int dead = 0;
3957
5e670d84
DA		 3958 if (rt->fib6_nh.nh_dev == down_dev ||
3959 rt->fib6_nh.nh_flags & RTNH_F_DEAD)
1de178ed 3960 dead++;
93c2fb25 3961 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
5e670d84
DA		 3962 if (iter->fib6_nh.nh_dev == down_dev ||
3963 iter->fib6_nh.nh_flags & RTNH_F_DEAD)
1de178ed
IS		 3964 dead++;
3965
3966 return dead;
3967}
3968
8d1c802b 3969static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
1de178ed
IS		 3970 const struct net_device *dev,
3971 unsigned int nh_flags)
3972{
8d1c802b 3973 struct fib6_info *iter;
1de178ed 3974
5e670d84
DA		 3975 if (rt->fib6_nh.nh_dev == dev)
3976 rt->fib6_nh.nh_flags |= nh_flags;
93c2fb25 3977 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
5e670d84
DA		 3978 if (iter->fib6_nh.nh_dev == dev)
3979 iter->fib6_nh.nh_flags |= nh_flags;
1de178ed
IS		 3980}
3981
a1a22c12 3982/* called with write lock held for table with rt */
8d1c802b 3983static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
1da177e4 3984{
4c981e28
IS		 3985 const struct arg_netdev_event *arg = p_arg;
3986 const struct net_device *dev = arg->dev;
7aef6859 3987 struct net *net = dev_net(dev);
8ed67789 3988
421842ed 3989 if (rt == net->ipv6.fib6_null_entry)
27c6fa73
IS		 3990 return 0;
3991
3992 switch (arg->event) {
3993 case NETDEV_UNREGISTER:
5e670d84 3994 return rt->fib6_nh.nh_dev == dev ? -1 : 0;
27c6fa73 3995 case NETDEV_DOWN:
1de178ed 3996 if (rt->should_flush)
27c6fa73 3997 return -1;
93c2fb25 3998 if (!rt->fib6_nsiblings)
5e670d84 3999 return rt->fib6_nh.nh_dev == dev ? -1 : 0;
1de178ed
IS		 4000 if (rt6_multipath_uses_dev(rt, dev)) {
4001 unsigned int count;
4002
4003 count = rt6_multipath_dead_count(rt, dev);
93c2fb25 4004 if (rt->fib6_nsiblings + 1 == count) {
1de178ed
IS		 4005 rt6_multipath_flush(rt);
4006 return -1;
4007 }
4008 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4009 RTNH_F_LINKDOWN);
7aef6859 4010 fib6_update_sernum(net, rt);
d7dedee1 4011 rt6_multipath_rebalance(rt);
1de178ed
IS		 4012 }
4013 return -2;
27c6fa73 4014 case NETDEV_CHANGE:
5e670d84 4015 if (rt->fib6_nh.nh_dev != dev ||
93c2fb25 4016 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
27c6fa73 4017 break;
5e670d84 4018 rt->fib6_nh.nh_flags |= RTNH_F_LINKDOWN;
d7dedee1 4019 rt6_multipath_rebalance(rt);
27c6fa73 4020 break;
2b241361 4021 }
c159d30c 4022
1da177e4
LT		 4023 return 0;
4024}
4025
27c6fa73 4026void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
1da177e4 4027{
4c981e28 4028 struct arg_netdev_event arg = {
8ed67789 4029 .dev = dev,
6802f3ad
IS		 4030 {
4031 .event = event,
4032 },
8ed67789 4033 };
7c6bb7d2 4034 struct net *net = dev_net(dev);
8ed67789 4035
7c6bb7d2
DA		 4036 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4037 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4038 else
4039 fib6_clean_all(net, fib6_ifdown, &arg);
4c981e28
IS		 4040}
4041
4042void rt6_disable_ip(struct net_device *dev, unsigned long event)
4043{
4044 rt6_sync_down_dev(dev, event);
4045 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4046 neigh_ifdown(&nd_tbl, dev);
1da177e4
LT		 4047}
4048
95c96174 4049struct rt6_mtu_change_arg {
1da177e4 4050 struct net_device *dev;
95c96174 4051 unsigned int mtu;
1da177e4
LT		 4052};
4053
8d1c802b 4054static int rt6_mtu_change_route(struct fib6_info *rt, void *p_arg)
1da177e4
LT		 4055{
4056 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4057 struct inet6_dev *idev;
4058
4059 /* In IPv6 pmtu discovery is not optional,
4060 so that RTAX_MTU lock cannot disable it.
4061 We still use this lock to block changes
4062 caused by addrconf/ndisc.
4063 */
4064
4065 idev = __in6_dev_get(arg->dev);
38308473 4066 if (!idev)
1da177e4
LT		 4067 return 0;
4068
4069 /* For administrative MTU increase, there is no way to discover
4070 IPv6 PMTU increase, so PMTU increase should be updated here.
4071 Since RFC 1981 doesn't include administrative MTU increase
4072 update PMTU increase is a MUST. (i.e. jumbo frame)
4073 */
5e670d84 4074 if (rt->fib6_nh.nh_dev == arg->dev &&
d4ead6b3
DA		 4075 !fib6_metric_locked(rt, RTAX_MTU)) {
4076 u32 mtu = rt->fib6_pmtu;
4077
4078 if (mtu >= arg->mtu ||
4079 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4080 fib6_metric_set(rt, RTAX_MTU, arg->mtu);
4081
f5bbe7ee 4082 spin_lock_bh(&rt6_exception_lock);
e9fa1495 4083 rt6_exceptions_update_pmtu(idev, rt, arg->mtu);
f5bbe7ee 4084 spin_unlock_bh(&rt6_exception_lock);
566cfd8f 4085 }
1da177e4
LT		 4086 return 0;
4087}
4088
95c96174 4089void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
1da177e4 4090{
c71099ac
TG		 4091 struct rt6_mtu_change_arg arg = {
4092 .dev = dev,
4093 .mtu = mtu,
4094 };
1da177e4 4095
0c3584d5 4096 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
1da177e4
LT		 4097}
4098
ef7c79ed 4099static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
5176f91e 4100 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
aa8f8778 4101 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
86872cb5 4102 [RTA_OIF] = { .type = NLA_U32 },
ab364a6f 4103 [RTA_IIF] = { .type = NLA_U32 },
86872cb5
TG		 4104 [RTA_PRIORITY] = { .type = NLA_U32 },
4105 [RTA_METRICS] = { .type = NLA_NESTED },
51ebd318 4106 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
c78ba6d6 4107 [RTA_PREF] = { .type = NLA_U8 },
19e42e45
RP		 4108 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4109 [RTA_ENCAP] = { .type = NLA_NESTED },
32bc201e 4110 [RTA_EXPIRES] = { .type = NLA_U32 },
622ec2c9 4111 [RTA_UID] = { .type = NLA_U32 },
3b45a410 4112 [RTA_MARK] = { .type = NLA_U32 },
aa8f8778 4113 [RTA_TABLE] = { .type = NLA_U32 },
eacb9384
RP		 4114 [RTA_IP_PROTO] = { .type = NLA_U8 },
4115 [RTA_SPORT] = { .type = NLA_U16 },
4116 [RTA_DPORT] = { .type = NLA_U16 },
86872cb5
TG		 4117};
4118
4119static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
333c4301
DA		 4120 struct fib6_config *cfg,
4121 struct netlink_ext_ack *extack)
1da177e4 4122{
86872cb5
TG		 4123 struct rtmsg *rtm;
4124 struct nlattr *tb[RTA_MAX+1];
c78ba6d6 4125 unsigned int pref;
86872cb5 4126 int err;
1da177e4 4127
fceb6435 4128 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
dac9c979 4129 extack);
86872cb5
TG		 4130 if (err < 0)
4131 goto errout;
1da177e4 4132
86872cb5
TG		 4133 err = -EINVAL;
4134 rtm = nlmsg_data(nlh);
86872cb5 4135
84db8407
MŻ		 4136 *cfg = (struct fib6_config){
4137 .fc_table = rtm->rtm_table,
4138 .fc_dst_len = rtm->rtm_dst_len,
4139 .fc_src_len = rtm->rtm_src_len,
4140 .fc_flags = RTF_UP,
4141 .fc_protocol = rtm->rtm_protocol,
4142 .fc_type = rtm->rtm_type,
4143
4144 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
4145 .fc_nlinfo.nlh = nlh,
4146 .fc_nlinfo.nl_net = sock_net(skb->sk),
4147 };
86872cb5 4148
ef2c7d7b
ND		 4149 if (rtm->rtm_type == RTN_UNREACHABLE ||
4150 rtm->rtm_type == RTN_BLACKHOLE ||
b4949ab2
ND		 4151 rtm->rtm_type == RTN_PROHIBIT ||
4152 rtm->rtm_type == RTN_THROW)
86872cb5
TG		 4153 cfg->fc_flags |= RTF_REJECT;
4154
ab79ad14
MŻ		 4155 if (rtm->rtm_type == RTN_LOCAL)
4156 cfg->fc_flags |= RTF_LOCAL;
4157
1f56a01f
MKL		 4158 if (rtm->rtm_flags & RTM_F_CLONED)
4159 cfg->fc_flags |= RTF_CACHE;
4160
fc1e64e1
DA		 4161 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4162
86872cb5 4163 if (tb[RTA_GATEWAY]) {
67b61f6c 4164 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
86872cb5 4165 cfg->fc_flags |= RTF_GATEWAY;
1da177e4 4166 }
86872cb5
TG		 4167
4168 if (tb[RTA_DST]) {
4169 int plen = (rtm->rtm_dst_len + 7) >> 3;
4170
4171 if (nla_len(tb[RTA_DST]) < plen)
4172 goto errout;
4173
4174 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1da177e4 4175 }
86872cb5
TG		 4176
4177 if (tb[RTA_SRC]) {
4178 int plen = (rtm->rtm_src_len + 7) >> 3;
4179
4180 if (nla_len(tb[RTA_SRC]) < plen)
4181 goto errout;
4182
4183 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1da177e4 4184 }
86872cb5 4185
c3968a85 4186 if (tb[RTA_PREFSRC])
67b61f6c 4187 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
c3968a85 4188
86872cb5
TG		 4189 if (tb[RTA_OIF])
4190 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4191
4192 if (tb[RTA_PRIORITY])
4193 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4194
4195 if (tb[RTA_METRICS]) {
4196 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4197 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
1da177e4 4198 }
86872cb5
TG		 4199
4200 if (tb[RTA_TABLE])
4201 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4202
51ebd318
ND		 4203 if (tb[RTA_MULTIPATH]) {
4204 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4205 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
9ed59592
DA		 4206
4207 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
c255bd68 4208 cfg->fc_mp_len, extack);
9ed59592
DA		 4209 if (err < 0)
4210 goto errout;
51ebd318
ND		 4211 }
4212
c78ba6d6
LR		 4213 if (tb[RTA_PREF]) {
4214 pref = nla_get_u8(tb[RTA_PREF]);
4215 if (pref != ICMPV6_ROUTER_PREF_LOW &&
4216 pref != ICMPV6_ROUTER_PREF_HIGH)
4217 pref = ICMPV6_ROUTER_PREF_MEDIUM;
4218 cfg->fc_flags |= RTF_PREF(pref);
4219 }
4220
19e42e45
RP		 4221 if (tb[RTA_ENCAP])
4222 cfg->fc_encap = tb[RTA_ENCAP];
4223
9ed59592 4224 if (tb[RTA_ENCAP_TYPE]) {
19e42e45
RP		 4225 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
4226
c255bd68 4227 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
9ed59592
DA		 4228 if (err < 0)
4229 goto errout;
4230 }
4231
32bc201e
XL		 4232 if (tb[RTA_EXPIRES]) {
4233 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
4234
4235 if (addrconf_finite_timeout(timeout)) {
4236 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
4237 cfg->fc_flags |= RTF_EXPIRES;
4238 }
4239 }
4240
86872cb5
TG		 4241 err = 0;
4242errout:
4243 return err;
1da177e4
LT		 4244}
4245
6b9ea5a6 4246struct rt6_nh {
8d1c802b 4247 struct fib6_info *fib6_info;
6b9ea5a6 4248 struct fib6_config r_cfg;
6b9ea5a6
RP		 4249 struct list_head next;
4250};
4251
4252static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
4253{
4254 struct rt6_nh *nh;
4255
4256 list_for_each_entry(nh, rt6_nh_list, next) {
7d4d5065 4257 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6c nexthop %pI6c ifi %d\n",
6b9ea5a6
RP		 4258 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
4259 nh->r_cfg.fc_ifindex);
4260 }
4261}
4262
d4ead6b3
DA		 4263static int ip6_route_info_append(struct net *net,
4264 struct list_head *rt6_nh_list,
8d1c802b
DA		 4265 struct fib6_info *rt,
4266 struct fib6_config *r_cfg)
6b9ea5a6
RP		 4267{
4268 struct rt6_nh *nh;
6b9ea5a6
RP		 4269 int err = -EEXIST;
4270
4271 list_for_each_entry(nh, rt6_nh_list, next) {
8d1c802b
DA		 4272 /* check if fib6_info already exists */
4273 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
6b9ea5a6
RP		 4274 return err;
4275 }
4276
4277 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
4278 if (!nh)
4279 return -ENOMEM;
8d1c802b 4280 nh->fib6_info = rt;
6b9ea5a6
RP		 4281 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
4282 list_add_tail(&nh->next, rt6_nh_list);
4283
4284 return 0;
4285}
4286
8d1c802b
DA		 4287static void ip6_route_mpath_notify(struct fib6_info *rt,
4288 struct fib6_info *rt_last,
3b1137fe
DA		 4289 struct nl_info *info,
4290 __u16 nlflags)
4291{
4292 /* if this is an APPEND route, then rt points to the first route
4293 * inserted and rt_last points to last route inserted. Userspace
4294 * wants a consistent dump of the route which starts at the first
4295 * nexthop. Since sibling routes are always added at the end of
4296 * the list, find the first sibling of the last route appended
4297 */
93c2fb25
DA		 4298 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
4299 rt = list_first_entry(&rt_last->fib6_siblings,
8d1c802b 4300 struct fib6_info,
93c2fb25 4301 fib6_siblings);
3b1137fe
DA		 4302 }
4303
4304 if (rt)
4305 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
4306}
4307
333c4301
DA		 4308static int ip6_route_multipath_add(struct fib6_config *cfg,
4309 struct netlink_ext_ack *extack)
51ebd318 4310{
8d1c802b 4311 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
3b1137fe 4312 struct nl_info *info = &cfg->fc_nlinfo;
51ebd318
ND		 4313 struct fib6_config r_cfg;
4314 struct rtnexthop *rtnh;
8d1c802b 4315 struct fib6_info *rt;
6b9ea5a6
RP		 4316 struct rt6_nh *err_nh;
4317 struct rt6_nh *nh, *nh_safe;
3b1137fe 4318 __u16 nlflags;
51ebd318
ND		 4319 int remaining;
4320 int attrlen;
6b9ea5a6
RP		 4321 int err = 1;
4322 int nhn = 0;
4323 int replace = (cfg->fc_nlinfo.nlh &&
4324 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
4325 LIST_HEAD(rt6_nh_list);
51ebd318 4326
3b1137fe
DA		 4327 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
4328 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
4329 nlflags |= NLM_F_APPEND;
4330
35f1b4e9 4331 remaining = cfg->fc_mp_len;
51ebd318 4332 rtnh = (struct rtnexthop *)cfg->fc_mp;
51ebd318 4333
6b9ea5a6 4334 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
8d1c802b 4335 * fib6_info structs per nexthop
6b9ea5a6 4336 */
51ebd318
ND		 4337 while (rtnh_ok(rtnh, remaining)) {
4338 memcpy(&r_cfg, cfg, sizeof(*cfg));
4339 if (rtnh->rtnh_ifindex)
4340 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
4341
4342 attrlen = rtnh_attrlen(rtnh);
4343 if (attrlen > 0) {
4344 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
4345
4346 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
4347 if (nla) {
67b61f6c 4348 r_cfg.fc_gateway = nla_get_in6_addr(nla);
51ebd318
ND		 4349 r_cfg.fc_flags |= RTF_GATEWAY;
4350 }
19e42e45
RP		 4351 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
4352 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
4353 if (nla)
4354 r_cfg.fc_encap_type = nla_get_u16(nla);
51ebd318 4355 }
6b9ea5a6 4356
68e2ffde 4357 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
acb54e3c 4358 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
8c5b83f0
RP		 4359 if (IS_ERR(rt)) {
4360 err = PTR_ERR(rt);
4361 rt = NULL;
6b9ea5a6 4362 goto cleanup;
8c5b83f0 4363 }
b5d2d75e
DA		 4364 if (!rt6_qualify_for_ecmp(rt)) {
4365 err = -EINVAL;
4366 NL_SET_ERR_MSG(extack,
4367 "Device only routes can not be added for IPv6 using the multipath API.");
4368 fib6_info_release(rt);
4369 goto cleanup;
4370 }
6b9ea5a6 4371
5e670d84 4372 rt->fib6_nh.nh_weight = rtnh->rtnh_hops + 1;
398958ae 4373
d4ead6b3
DA		 4374 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
4375 rt, &r_cfg);
51ebd318 4376 if (err) {
93531c67 4377 fib6_info_release(rt);
6b9ea5a6
RP		 4378 goto cleanup;
4379 }
4380
4381 rtnh = rtnh_next(rtnh, &remaining);
4382 }
4383
3b1137fe
DA		 4384 /* for add and replace send one notification with all nexthops.
4385 * Skip the notification in fib6_add_rt2node and send one with
4386 * the full route when done
4387 */
4388 info->skip_notify = 1;
4389
6b9ea5a6
RP		 4390 err_nh = NULL;
4391 list_for_each_entry(nh, &rt6_nh_list, next) {
8d1c802b
DA		 4392 err = __ip6_ins_rt(nh->fib6_info, info, extack);
4393 fib6_info_release(nh->fib6_info);
93531c67 4394
f7225172
DA		 4395 if (!err) {
4396 /* save reference to last route successfully inserted */
4397 rt_last = nh->fib6_info;
4398
4399 /* save reference to first route for notification */
4400 if (!rt_notif)
4401 rt_notif = nh->fib6_info;
4402 }
3b1137fe 4403
8d1c802b
DA		 4404 /* nh->fib6_info is used or freed at this point, reset to NULL*/
4405 nh->fib6_info = NULL;
6b9ea5a6
RP		 4406 if (err) {
4407 if (replace && nhn)
4408 ip6_print_replace_route_err(&rt6_nh_list);
4409 err_nh = nh;
4410 goto add_errout;
51ebd318 4411 }
6b9ea5a6 4412
1a72418b 4413 /* Because each route is added like a single route we remove
27596472
MK		 4414 * these flags after the first nexthop: if there is a collision,
4415 * we have already failed to add the first nexthop:
4416 * fib6_add_rt2node() has rejected it; when replacing, old
4417 * nexthops have been replaced by first new, the rest should
4418 * be added to it.
1a72418b 4419 */
27596472
MK		 4420 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
4421 NLM_F_REPLACE);
6b9ea5a6
RP		 4422 nhn++;
4423 }
4424
3b1137fe
DA		 4425 /* success ... tell user about new route */
4426 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
6b9ea5a6
RP		 4427 goto cleanup;
4428
4429add_errout:
3b1137fe
DA		 4430 /* send notification for routes that were added so that
4431 * the delete notifications sent by ip6_route_del are
4432 * coherent
4433 */
4434 if (rt_notif)
4435 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
4436
6b9ea5a6
RP		 4437 /* Delete routes that were already added */
4438 list_for_each_entry(nh, &rt6_nh_list, next) {
4439 if (err_nh == nh)
4440 break;
333c4301 4441 ip6_route_del(&nh->r_cfg, extack);
6b9ea5a6
RP		 4442 }
4443
4444cleanup:
4445 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
8d1c802b
DA		 4446 if (nh->fib6_info)
4447 fib6_info_release(nh->fib6_info);
6b9ea5a6
RP		 4448 list_del(&nh->next);
4449 kfree(nh);
4450 }
4451
4452 return err;
4453}
4454
333c4301
DA		 4455static int ip6_route_multipath_del(struct fib6_config *cfg,
4456 struct netlink_ext_ack *extack)
6b9ea5a6
RP		 4457{
4458 struct fib6_config r_cfg;
4459 struct rtnexthop *rtnh;
4460 int remaining;
4461 int attrlen;
4462 int err = 1, last_err = 0;
4463
4464 remaining = cfg->fc_mp_len;
4465 rtnh = (struct rtnexthop *)cfg->fc_mp;
4466
4467 /* Parse a Multipath Entry */
4468 while (rtnh_ok(rtnh, remaining)) {
4469 memcpy(&r_cfg, cfg, sizeof(*cfg));
4470 if (rtnh->rtnh_ifindex)
4471 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
4472
4473 attrlen = rtnh_attrlen(rtnh);
4474 if (attrlen > 0) {
4475 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
4476
4477 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
4478 if (nla) {
4479 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
4480 r_cfg.fc_flags |= RTF_GATEWAY;
4481 }
4482 }
333c4301 4483 err = ip6_route_del(&r_cfg, extack);
6b9ea5a6
RP		 4484 if (err)
4485 last_err = err;
4486
51ebd318
ND		 4487 rtnh = rtnh_next(rtnh, &remaining);
4488 }
4489
4490 return last_err;
4491}
4492
c21ef3e3
DA		 4493static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
4494 struct netlink_ext_ack *extack)
1da177e4 4495{
86872cb5
TG		 4496 struct fib6_config cfg;
4497 int err;
1da177e4 4498
333c4301 4499 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 4500 if (err < 0)
4501 return err;
4502
51ebd318 4503 if (cfg.fc_mp)
333c4301 4504 return ip6_route_multipath_del(&cfg, extack);
0ae81335
DA		 4505 else {
4506 cfg.fc_delete_all_nh = 1;
333c4301 4507 return ip6_route_del(&cfg, extack);
0ae81335 4508 }
1da177e4
LT		 4509}
4510
c21ef3e3
DA		 4511static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
4512 struct netlink_ext_ack *extack)
1da177e4 4513{
86872cb5
TG		 4514 struct fib6_config cfg;
4515 int err;
1da177e4 4516
333c4301 4517 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 4518 if (err < 0)
4519 return err;
4520
51ebd318 4521 if (cfg.fc_mp)
333c4301 4522 return ip6_route_multipath_add(&cfg, extack);
51ebd318 4523 else
acb54e3c 4524 return ip6_route_add(&cfg, GFP_KERNEL, extack);
1da177e4
LT		 4525}
4526
8d1c802b 4527static size_t rt6_nlmsg_size(struct fib6_info *rt)
339bf98f 4528{
beb1afac
DA		 4529 int nexthop_len = 0;
4530
93c2fb25 4531 if (rt->fib6_nsiblings) {
beb1afac
DA		 4532 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
4533 + NLA_ALIGN(sizeof(struct rtnexthop))
4534 + nla_total_size(16) /* RTA_GATEWAY */
5e670d84 4535 + lwtunnel_get_encap_size(rt->fib6_nh.nh_lwtstate);
beb1afac 4536
93c2fb25 4537 nexthop_len *= rt->fib6_nsiblings;
beb1afac
DA		 4538 }
4539
339bf98f
TG		 4540 return NLMSG_ALIGN(sizeof(struct rtmsg))
4541 + nla_total_size(16) /* RTA_SRC */
4542 + nla_total_size(16) /* RTA_DST */
4543 + nla_total_size(16) /* RTA_GATEWAY */
4544 + nla_total_size(16) /* RTA_PREFSRC */
4545 + nla_total_size(4) /* RTA_TABLE */
4546 + nla_total_size(4) /* RTA_IIF */
4547 + nla_total_size(4) /* RTA_OIF */
4548 + nla_total_size(4) /* RTA_PRIORITY */
6a2b9ce0 4549 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
ea697639 4550 + nla_total_size(sizeof(struct rta_cacheinfo))
c78ba6d6 4551 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
19e42e45 4552 + nla_total_size(1) /* RTA_PREF */
5e670d84 4553 + lwtunnel_get_encap_size(rt->fib6_nh.nh_lwtstate)
beb1afac
DA		 4554 + nexthop_len;
4555}
4556
8d1c802b 4557static int rt6_nexthop_info(struct sk_buff *skb, struct fib6_info *rt,
5be083ce 4558 unsigned int *flags, bool skip_oif)
beb1afac 4559{
5e670d84 4560 if (rt->fib6_nh.nh_flags & RTNH_F_DEAD)
f9d882ea
IS		 4561 *flags |= RTNH_F_DEAD;
4562
5e670d84 4563 if (rt->fib6_nh.nh_flags & RTNH_F_LINKDOWN) {
beb1afac 4564 *flags |= RTNH_F_LINKDOWN;
dcd1f572
DA		 4565
4566 rcu_read_lock();
4567 if (fib6_ignore_linkdown(rt))
beb1afac 4568 *flags |= RTNH_F_DEAD;
dcd1f572 4569 rcu_read_unlock();
beb1afac
DA		 4570 }
4571
93c2fb25 4572 if (rt->fib6_flags & RTF_GATEWAY) {
5e670d84 4573 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->fib6_nh.nh_gw) < 0)
beb1afac
DA		 4574 goto nla_put_failure;
4575 }
4576
5e670d84
DA		 4577 *flags |= (rt->fib6_nh.nh_flags & RTNH_F_ONLINK);
4578 if (rt->fib6_nh.nh_flags & RTNH_F_OFFLOAD)
61e4d01e
IS		 4579 *flags |= RTNH_F_OFFLOAD;
4580
5be083ce 4581 /* not needed for multipath encoding b/c it has a rtnexthop struct */
5e670d84
DA		 4582 if (!skip_oif && rt->fib6_nh.nh_dev &&
4583 nla_put_u32(skb, RTA_OIF, rt->fib6_nh.nh_dev->ifindex))
beb1afac
DA		 4584 goto nla_put_failure;
4585
5e670d84
DA		 4586 if (rt->fib6_nh.nh_lwtstate &&
4587 lwtunnel_fill_encap(skb, rt->fib6_nh.nh_lwtstate) < 0)
beb1afac
DA		 4588 goto nla_put_failure;
4589
4590 return 0;
4591
4592nla_put_failure:
4593 return -EMSGSIZE;
4594}
4595
5be083ce 4596/* add multipath next hop */
8d1c802b 4597static int rt6_add_nexthop(struct sk_buff *skb, struct fib6_info *rt)
beb1afac 4598{
5e670d84 4599 const struct net_device *dev = rt->fib6_nh.nh_dev;
beb1afac
DA		 4600 struct rtnexthop *rtnh;
4601 unsigned int flags = 0;
4602
4603 rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
4604 if (!rtnh)
4605 goto nla_put_failure;
4606
5e670d84
DA		 4607 rtnh->rtnh_hops = rt->fib6_nh.nh_weight - 1;
4608 rtnh->rtnh_ifindex = dev ? dev->ifindex : 0;
beb1afac 4609
5be083ce 4610 if (rt6_nexthop_info(skb, rt, &flags, true) < 0)
beb1afac
DA		 4611 goto nla_put_failure;
4612
4613 rtnh->rtnh_flags = flags;
4614
4615 /* length of rtnetlink header + attributes */
4616 rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
4617
4618 return 0;
4619
4620nla_put_failure:
4621 return -EMSGSIZE;
339bf98f
TG		 4622}
4623
d4ead6b3 4624static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 4625 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 4626 struct in6_addr *dest, struct in6_addr *src,
15e47304 4627 int iif, int type, u32 portid, u32 seq,
f8cfe2ce 4628 unsigned int flags)
1da177e4 4629{
22d0bd82
XL		 4630 struct rt6_info *rt6 = (struct rt6_info *)dst;
4631 struct rt6key *rt6_dst, *rt6_src;
4632 u32 *pmetrics, table, rt6_flags;
2d7202bf 4633 struct nlmsghdr *nlh;
22d0bd82 4634 struct rtmsg *rtm;
d4ead6b3 4635 long expires = 0;
1da177e4 4636
15e47304 4637 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
38308473 4638 if (!nlh)
26932566 4639 return -EMSGSIZE;
2d7202bf 4640
22d0bd82
XL		 4641 if (rt6) {
4642 rt6_dst = &rt6->rt6i_dst;
4643 rt6_src = &rt6->rt6i_src;
4644 rt6_flags = rt6->rt6i_flags;
4645 } else {
4646 rt6_dst = &rt->fib6_dst;
4647 rt6_src = &rt->fib6_src;
4648 rt6_flags = rt->fib6_flags;
4649 }
4650
2d7202bf 4651 rtm = nlmsg_data(nlh);
1da177e4 4652 rtm->rtm_family = AF_INET6;
22d0bd82
XL		 4653 rtm->rtm_dst_len = rt6_dst->plen;
4654 rtm->rtm_src_len = rt6_src->plen;
1da177e4 4655 rtm->rtm_tos = 0;
93c2fb25
DA		 4656 if (rt->fib6_table)
4657 table = rt->fib6_table->tb6_id;
c71099ac 4658 else
9e762a4a
PM		 4659 table = RT6_TABLE_UNSPEC;
4660 rtm->rtm_table = table;
c78679e8
DM		 4661 if (nla_put_u32(skb, RTA_TABLE, table))
4662 goto nla_put_failure;
e8478e80
DA		 4663
4664 rtm->rtm_type = rt->fib6_type;
1da177e4
LT		 4665 rtm->rtm_flags = 0;
4666 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
93c2fb25 4667 rtm->rtm_protocol = rt->fib6_protocol;
1da177e4 4668
22d0bd82 4669 if (rt6_flags & RTF_CACHE)
1da177e4
LT		 4670 rtm->rtm_flags |= RTM_F_CLONED;
4671
d4ead6b3
DA		 4672 if (dest) {
4673 if (nla_put_in6_addr(skb, RTA_DST, dest))
c78679e8 4674 goto nla_put_failure;
1ab1457c 4675 rtm->rtm_dst_len = 128;
1da177e4 4676 } else if (rtm->rtm_dst_len)
22d0bd82 4677 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
c78679e8 4678 goto nla_put_failure;
1da177e4
LT		 4679#ifdef CONFIG_IPV6_SUBTREES
4680 if (src) {
930345ea 4681 if (nla_put_in6_addr(skb, RTA_SRC, src))
c78679e8 4682 goto nla_put_failure;
1ab1457c 4683 rtm->rtm_src_len = 128;
c78679e8 4684 } else if (rtm->rtm_src_len &&
22d0bd82 4685 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
c78679e8 4686 goto nla_put_failure;
1da177e4 4687#endif
7bc570c8
YH		 4688 if (iif) {
4689#ifdef CONFIG_IPV6_MROUTE
22d0bd82 4690 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
fd61c6ba
DA		 4691 int err = ip6mr_get_route(net, skb, rtm, portid);
4692
4693 if (err == 0)
4694 return 0;
4695 if (err < 0)
4696 goto nla_put_failure;
7bc570c8
YH		 4697 } else
4698#endif
c78679e8
DM		 4699 if (nla_put_u32(skb, RTA_IIF, iif))
4700 goto nla_put_failure;
d4ead6b3 4701 } else if (dest) {
1da177e4 4702 struct in6_addr saddr_buf;
d4ead6b3 4703 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
930345ea 4704 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 4705 goto nla_put_failure;
1da177e4 4706 }
2d7202bf 4707
93c2fb25 4708 if (rt->fib6_prefsrc.plen) {
c3968a85 4709 struct in6_addr saddr_buf;
93c2fb25 4710 saddr_buf = rt->fib6_prefsrc.addr;
930345ea 4711 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 4712 goto nla_put_failure;
c3968a85
DW		 4713 }
4714
d4ead6b3
DA		 4715 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
4716 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
2d7202bf
TG		 4717 goto nla_put_failure;
4718
93c2fb25 4719 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
c78679e8 4720 goto nla_put_failure;
8253947e 4721
beb1afac
DA		 4722 /* For multipath routes, walk the siblings list and add
4723 * each as a nexthop within RTA_MULTIPATH.
4724 */
22d0bd82
XL		 4725 if (rt6) {
4726 if (rt6_flags & RTF_GATEWAY &&
4727 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
4728 goto nla_put_failure;
4729
4730 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
4731 goto nla_put_failure;
4732 } else if (rt->fib6_nsiblings) {
8d1c802b 4733 struct fib6_info *sibling, *next_sibling;
beb1afac
DA		 4734 struct nlattr *mp;
4735
4736 mp = nla_nest_start(skb, RTA_MULTIPATH);
4737 if (!mp)
4738 goto nla_put_failure;
4739
4740 if (rt6_add_nexthop(skb, rt) < 0)
4741 goto nla_put_failure;
4742
4743 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25 4744 &rt->fib6_siblings, fib6_siblings) {
beb1afac
DA		 4745 if (rt6_add_nexthop(skb, sibling) < 0)
4746 goto nla_put_failure;
4747 }
4748
4749 nla_nest_end(skb, mp);
4750 } else {
5be083ce 4751 if (rt6_nexthop_info(skb, rt, &rtm->rtm_flags, false) < 0)
beb1afac
DA		 4752 goto nla_put_failure;
4753 }
4754
22d0bd82 4755 if (rt6_flags & RTF_EXPIRES) {
14895687
DA		 4756 expires = dst ? dst->expires : rt->expires;
4757 expires -= jiffies;
4758 }
69cdf8f9 4759
d4ead6b3 4760 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
e3703b3d 4761 goto nla_put_failure;
2d7202bf 4762
22d0bd82 4763 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
c78ba6d6
LR		 4764 goto nla_put_failure;
4765
19e42e45 4766
053c095a
JB		 4767 nlmsg_end(skb, nlh);
4768 return 0;
2d7202bf
TG		 4769
4770nla_put_failure:
26932566
PM		 4771 nlmsg_cancel(skb, nlh);
4772 return -EMSGSIZE;
1da177e4
LT		 4773}
4774
13e38901
DA		 4775static bool fib6_info_uses_dev(const struct fib6_info *f6i,
4776 const struct net_device *dev)
4777{
4778 if (f6i->fib6_nh.nh_dev == dev)
4779 return true;
4780
4781 if (f6i->fib6_nsiblings) {
4782 struct fib6_info *sibling, *next_sibling;
4783
4784 list_for_each_entry_safe(sibling, next_sibling,
4785 &f6i->fib6_siblings, fib6_siblings) {
4786 if (sibling->fib6_nh.nh_dev == dev)
4787 return true;
4788 }
4789 }
4790
4791 return false;
4792}
4793
8d1c802b 4794int rt6_dump_route(struct fib6_info *rt, void *p_arg)
1da177e4
LT		 4795{
4796 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
13e38901
DA		 4797 struct fib_dump_filter *filter = &arg->filter;
4798 unsigned int flags = NLM_F_MULTI;
1f17e2f2
DA		 4799 struct net *net = arg->net;
4800
421842ed 4801 if (rt == net->ipv6.fib6_null_entry)
1f17e2f2 4802 return 0;
1da177e4 4803
13e38901
DA		 4804 if ((filter->flags & RTM_F_PREFIX) &&
4805 !(rt->fib6_flags & RTF_PREFIX_RT)) {
4806 /* success since this is not a prefix route */
4807 return 1;
4808 }
4809 if (filter->filter_set) {
4810 if ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
4811 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
4812 (filter->protocol && rt->fib6_protocol != filter->protocol)) {
f8cfe2ce
DA		 4813 return 1;
4814 }
13e38901 4815 flags |= NLM_F_DUMP_FILTERED;
f8cfe2ce 4816 }
1da177e4 4817
d4ead6b3
DA		 4818 return rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL, 0,
4819 RTM_NEWROUTE, NETLINK_CB(arg->cb->skb).portid,
13e38901 4820 arg->cb->nlh->nlmsg_seq, flags);
1da177e4
LT		 4821}
4822
c21ef3e3
DA		 4823static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4824 struct netlink_ext_ack *extack)
1da177e4 4825{
3b1e0a65 4826 struct net *net = sock_net(in_skb->sk);
ab364a6f 4827 struct nlattr *tb[RTA_MAX+1];
18c3a61c 4828 int err, iif = 0, oif = 0;
a68886a6 4829 struct fib6_info *from;
18c3a61c 4830 struct dst_entry *dst;
ab364a6f 4831 struct rt6_info *rt;
1da177e4 4832 struct sk_buff *skb;
ab364a6f 4833 struct rtmsg *rtm;
744486d4 4834 struct flowi6 fl6 = {};
18c3a61c 4835 bool fibmatch;
1da177e4 4836
fceb6435 4837 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
c21ef3e3 4838 extack);
ab364a6f
TG		 4839 if (err < 0)
4840 goto errout;
1da177e4 4841
ab364a6f 4842 err = -EINVAL;
38b7097b
HFS		 4843 rtm = nlmsg_data(nlh);
4844 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
18c3a61c 4845 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
1da177e4 4846
ab364a6f
TG		 4847 if (tb[RTA_SRC]) {
4848 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
4849 goto errout;
4850
4e3fd7a0 4851 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
ab364a6f
TG		 4852 }
4853
4854 if (tb[RTA_DST]) {
4855 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
4856 goto errout;
4857
4e3fd7a0 4858 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
ab364a6f
TG		 4859 }
4860
4861 if (tb[RTA_IIF])
4862 iif = nla_get_u32(tb[RTA_IIF]);
4863
4864 if (tb[RTA_OIF])
72331bc0 4865 oif = nla_get_u32(tb[RTA_OIF]);
1da177e4 4866
2e47b291
LC		 4867 if (tb[RTA_MARK])
4868 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
4869
622ec2c9
LC		 4870 if (tb[RTA_UID])
4871 fl6.flowi6_uid = make_kuid(current_user_ns(),
4872 nla_get_u32(tb[RTA_UID]));
4873 else
4874 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
4875
eacb9384
RP		 4876 if (tb[RTA_SPORT])
4877 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
4878
4879 if (tb[RTA_DPORT])
4880 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
4881
4882 if (tb[RTA_IP_PROTO]) {
4883 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
4884 &fl6.flowi6_proto, extack);
4885 if (err)
4886 goto errout;
4887 }
4888
1da177e4
LT		 4889 if (iif) {
4890 struct net_device *dev;
72331bc0
SL		 4891 int flags = 0;
4892
121622db
FW		 4893 rcu_read_lock();
4894
4895 dev = dev_get_by_index_rcu(net, iif);
1da177e4 4896 if (!dev) {
121622db 4897 rcu_read_unlock();
1da177e4 4898 err = -ENODEV;
ab364a6f 4899 goto errout;
1da177e4 4900 }
72331bc0
SL		 4901
4902 fl6.flowi6_iif = iif;
4903
4904 if (!ipv6_addr_any(&fl6.saddr))
4905 flags |= RT6_LOOKUP_F_HAS_SADDR;
4906
b75cc8f9 4907 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
121622db
FW		 4908
4909 rcu_read_unlock();
72331bc0
SL		 4910 } else {
4911 fl6.flowi6_oif = oif;
4912
58acfd71 4913 dst = ip6_route_output(net, NULL, &fl6);
18c3a61c
RP		 4914 }
4915
18c3a61c
RP		 4916
4917 rt = container_of(dst, struct rt6_info, dst);
4918 if (rt->dst.error) {
4919 err = rt->dst.error;
4920 ip6_rt_put(rt);
4921 goto errout;
1da177e4
LT		 4922 }
4923
9d6acb3b
WC		 4924 if (rt == net->ipv6.ip6_null_entry) {
4925 err = rt->dst.error;
4926 ip6_rt_put(rt);
4927 goto errout;
4928 }
4929
ab364a6f 4930 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
38308473 4931 if (!skb) {
94e187c0 4932 ip6_rt_put(rt);
ab364a6f
TG		 4933 err = -ENOBUFS;
4934 goto errout;
4935 }
1da177e4 4936
d8d1f30b 4937 skb_dst_set(skb, &rt->dst);
a68886a6
DA		 4938
4939 rcu_read_lock();
4940 from = rcu_dereference(rt->from);
4941
18c3a61c 4942 if (fibmatch)
a68886a6 4943 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL, iif,
18c3a61c
RP		 4944 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
4945 nlh->nlmsg_seq, 0);
4946 else
a68886a6
DA		 4947 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
4948 &fl6.saddr, iif, RTM_NEWROUTE,
d4ead6b3
DA		 4949 NETLINK_CB(in_skb).portid, nlh->nlmsg_seq,
4950 0);
a68886a6
DA		 4951 rcu_read_unlock();
4952
1da177e4 4953 if (err < 0) {
ab364a6f
TG		 4954 kfree_skb(skb);
4955 goto errout;
1da177e4
LT		 4956 }
4957
15e47304 4958 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
ab364a6f 4959errout:
1da177e4 4960 return err;
1da177e4
LT		 4961}
4962
8d1c802b 4963void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
37a1d361 4964 unsigned int nlm_flags)
1da177e4
LT		 4965{
4966 struct sk_buff *skb;
5578689a 4967 struct net *net = info->nl_net;
528c4ceb
DL		 4968 u32 seq;
4969 int err;
4970
4971 err = -ENOBUFS;
38308473 4972 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
86872cb5 4973
19e42e45 4974 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
38308473 4975 if (!skb)
21713ebc
TG		 4976 goto errout;
4977
d4ead6b3
DA		 4978 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
4979 event, info->portid, seq, nlm_flags);
26932566
PM		 4980 if (err < 0) {
4981 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
4982 WARN_ON(err == -EMSGSIZE);
4983 kfree_skb(skb);
4984 goto errout;
4985 }
15e47304 4986 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
1ce85fe4
PNA		 4987 info->nlh, gfp_any());
4988 return;
21713ebc
TG		 4989errout:
4990 if (err < 0)
5578689a 4991 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
1da177e4
LT		 4992}
4993
8ed67789 4994static int ip6_route_dev_notify(struct notifier_block *this,
351638e7 4995 unsigned long event, void *ptr)
8ed67789 4996{
351638e7 4997 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
c346dca1 4998 struct net *net = dev_net(dev);
8ed67789 4999
242d3a49
WC		 5000 if (!(dev->flags & IFF_LOOPBACK))
5001 return NOTIFY_OK;
5002
5003 if (event == NETDEV_REGISTER) {
421842ed 5004 net->ipv6.fib6_null_entry->fib6_nh.nh_dev = dev;
d8d1f30b 5005 net->ipv6.ip6_null_entry->dst.dev = dev;
8ed67789
DL		 5006 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
5007#ifdef CONFIG_IPV6_MULTIPLE_TABLES
d8d1f30b 5008 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
8ed67789 5009 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
d8d1f30b 5010 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
8ed67789 5011 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
242d3a49 5012#endif
76da0704
WC		 5013 } else if (event == NETDEV_UNREGISTER &&
5014 dev->reg_state != NETREG_UNREGISTERED) {
5015 /* NETDEV_UNREGISTER could be fired for multiple times by
5016 * netdev_wait_allrefs(). Make sure we only call this once.
5017 */
12d94a80 5018 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
242d3a49 5019#ifdef CONFIG_IPV6_MULTIPLE_TABLES
12d94a80
ED		 5020 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
5021 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
8ed67789
DL		 5022#endif
5023 }
5024
5025 return NOTIFY_OK;
5026}
5027
1da177e4
LT		 5028/*
5029 * /proc
5030 */
5031
5032#ifdef CONFIG_PROC_FS
1da177e4
LT		 5033static int rt6_stats_seq_show(struct seq_file *seq, void *v)
5034{
69ddb805 5035 struct net *net = (struct net *)seq->private;
1da177e4 5036 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
69ddb805
DL		 5037 net->ipv6.rt6_stats->fib_nodes,
5038 net->ipv6.rt6_stats->fib_route_nodes,
81eb8447 5039 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
69ddb805
DL		 5040 net->ipv6.rt6_stats->fib_rt_entries,
5041 net->ipv6.rt6_stats->fib_rt_cache,
fc66f95c 5042 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
69ddb805 5043 net->ipv6.rt6_stats->fib_discarded_routes);
1da177e4
LT		 5044
5045 return 0;
5046}
1da177e4
LT		 5047#endif /* CONFIG_PROC_FS */
5048
5049#ifdef CONFIG_SYSCTL
5050
1da177e4 5051static
fe2c6338 5052int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
1da177e4
LT		 5053 void __user *buffer, size_t *lenp, loff_t *ppos)
5054{
c486da34
LAG		 5055 struct net *net;
5056 int delay;
5057 if (!write)
1da177e4 5058 return -EINVAL;
c486da34
LAG		 5059
5060 net = (struct net *)ctl->extra1;
5061 delay = net->ipv6.sysctl.flush_delay;
5062 proc_dointvec(ctl, write, buffer, lenp, ppos);
2ac3ac8f 5063 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
c486da34 5064 return 0;
1da177e4
LT		 5065}
5066
7c6bb7d2
DA		 5067static int zero;
5068static int one = 1;
5069
ed792e28 5070static struct ctl_table ipv6_route_table_template[] = {
1ab1457c 5071 {
1da177e4 5072 .procname = "flush",
4990509f 5073 .data = &init_net.ipv6.sysctl.flush_delay,
1da177e4 5074 .maxlen = sizeof(int),
89c8b3a1 5075 .mode = 0200,
6d9f239a 5076 .proc_handler = ipv6_sysctl_rtcache_flush
1da177e4
LT		 5077 },
5078 {
1da177e4 5079 .procname = "gc_thresh",
9a7ec3a9 5080 .data = &ip6_dst_ops_template.gc_thresh,
1da177e4
LT		 5081 .maxlen = sizeof(int),
5082 .mode = 0644,
6d9f239a 5083 .proc_handler = proc_dointvec,
1da177e4
LT		 5084 },
5085 {
1da177e4 5086 .procname = "max_size",
4990509f 5087 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
1da177e4
LT		 5088 .maxlen = sizeof(int),
5089 .mode = 0644,
6d9f239a 5090 .proc_handler = proc_dointvec,
1da177e4
LT		 5091 },
5092 {
1da177e4 5093 .procname = "gc_min_interval",
4990509f 5094 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 5095 .maxlen = sizeof(int),
5096 .mode = 0644,
6d9f239a 5097 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5098 },
5099 {
1da177e4 5100 .procname = "gc_timeout",
4990509f 5101 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
1da177e4
LT		 5102 .maxlen = sizeof(int),
5103 .mode = 0644,
6d9f239a 5104 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5105 },
5106 {
1da177e4 5107 .procname = "gc_interval",
4990509f 5108 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
1da177e4
LT		 5109 .maxlen = sizeof(int),
5110 .mode = 0644,
6d9f239a 5111 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5112 },
5113 {
1da177e4 5114 .procname = "gc_elasticity",
4990509f 5115 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
1da177e4
LT		 5116 .maxlen = sizeof(int),
5117 .mode = 0644,
f3d3f616 5118 .proc_handler = proc_dointvec,
1da177e4
LT		 5119 },
5120 {
1da177e4 5121 .procname = "mtu_expires",
4990509f 5122 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
1da177e4
LT		 5123 .maxlen = sizeof(int),
5124 .mode = 0644,
6d9f239a 5125 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5126 },
5127 {
1da177e4 5128 .procname = "min_adv_mss",
4990509f 5129 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
1da177e4
LT		 5130 .maxlen = sizeof(int),
5131 .mode = 0644,
f3d3f616 5132 .proc_handler = proc_dointvec,
1da177e4
LT		 5133 },
5134 {
1da177e4 5135 .procname = "gc_min_interval_ms",
4990509f 5136 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 5137 .maxlen = sizeof(int),
5138 .mode = 0644,
6d9f239a 5139 .proc_handler = proc_dointvec_ms_jiffies,
1da177e4 5140 },
7c6bb7d2
DA		 5141 {
5142 .procname = "skip_notify_on_dev_down",
5143 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
5144 .maxlen = sizeof(int),
5145 .mode = 0644,
5146 .proc_handler = proc_dointvec,
5147 .extra1 = &zero,
5148 .extra2 = &one,
5149 },
f8572d8f 5150 { }
1da177e4
LT		 5151};
5152
2c8c1e72 5153struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
760f2d01
DL		 5154{
5155 struct ctl_table *table;
5156
5157 table = kmemdup(ipv6_route_table_template,
5158 sizeof(ipv6_route_table_template),
5159 GFP_KERNEL);
5ee09105
YH		 5160
5161 if (table) {
5162 table[0].data = &net->ipv6.sysctl.flush_delay;
c486da34 5163 table[0].extra1 = net;
86393e52 5164 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
5ee09105
YH		 5165 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
5166 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
5167 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
5168 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
5169 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
5170 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
5171 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
9c69fabe 5172 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
7c6bb7d2 5173 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
464dc801
EB		 5174
5175 /* Don't export sysctls to unprivileged users */
5176 if (net->user_ns != &init_user_ns)
5177 table[0].procname = NULL;
5ee09105
YH		 5178 }
5179
760f2d01
DL		 5180 return table;
5181}
1da177e4
LT		 5182#endif
5183
2c8c1e72 5184static int __net_init ip6_route_net_init(struct net *net)
cdb18761 5185{
633d424b 5186 int ret = -ENOMEM;
8ed67789 5187
86393e52
AD		 5188 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
5189 sizeof(net->ipv6.ip6_dst_ops));
f2fc6a54 5190
fc66f95c
ED		 5191 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
5192 goto out_ip6_dst_ops;
5193
421842ed
DA		 5194 net->ipv6.fib6_null_entry = kmemdup(&fib6_null_entry_template,
5195 sizeof(*net->ipv6.fib6_null_entry),
5196 GFP_KERNEL);
5197 if (!net->ipv6.fib6_null_entry)
5198 goto out_ip6_dst_entries;
5199
8ed67789
DL		 5200 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
5201 sizeof(*net->ipv6.ip6_null_entry),
5202 GFP_KERNEL);
5203 if (!net->ipv6.ip6_null_entry)
421842ed 5204 goto out_fib6_null_entry;
d8d1f30b 5205 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 5206 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
5207 ip6_template_metrics, true);
8ed67789
DL		 5208
5209#ifdef CONFIG_IPV6_MULTIPLE_TABLES
feca7d8c 5210 net->ipv6.fib6_has_custom_rules = false;
8ed67789
DL		 5211 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
5212 sizeof(*net->ipv6.ip6_prohibit_entry),
5213 GFP_KERNEL);
68fffc67
PZ		 5214 if (!net->ipv6.ip6_prohibit_entry)
5215 goto out_ip6_null_entry;
d8d1f30b 5216 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 5217 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
5218 ip6_template_metrics, true);
8ed67789
DL		 5219
5220 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
5221 sizeof(*net->ipv6.ip6_blk_hole_entry),
5222 GFP_KERNEL);
68fffc67
PZ		 5223 if (!net->ipv6.ip6_blk_hole_entry)
5224 goto out_ip6_prohibit_entry;
d8d1f30b 5225 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 5226 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
5227 ip6_template_metrics, true);
8ed67789
DL		 5228#endif
5229
b339a47c
PZ		 5230 net->ipv6.sysctl.flush_delay = 0;
5231 net->ipv6.sysctl.ip6_rt_max_size = 4096;
5232 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
5233 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
5234 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
5235 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
5236 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
5237 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
7c6bb7d2 5238 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
b339a47c 5239
6891a346
BT		 5240 net->ipv6.ip6_rt_gc_expire = 30*HZ;
5241
8ed67789
DL		 5242 ret = 0;
5243out:
5244 return ret;
f2fc6a54 5245
68fffc67
PZ		 5246#ifdef CONFIG_IPV6_MULTIPLE_TABLES
5247out_ip6_prohibit_entry:
5248 kfree(net->ipv6.ip6_prohibit_entry);
5249out_ip6_null_entry:
5250 kfree(net->ipv6.ip6_null_entry);
5251#endif
421842ed
DA		 5252out_fib6_null_entry:
5253 kfree(net->ipv6.fib6_null_entry);
fc66f95c
ED		 5254out_ip6_dst_entries:
5255 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
f2fc6a54 5256out_ip6_dst_ops:
f2fc6a54 5257 goto out;
cdb18761
DL		 5258}
5259
2c8c1e72 5260static void __net_exit ip6_route_net_exit(struct net *net)
cdb18761 5261{
421842ed 5262 kfree(net->ipv6.fib6_null_entry);
8ed67789
DL		 5263 kfree(net->ipv6.ip6_null_entry);
5264#ifdef CONFIG_IPV6_MULTIPLE_TABLES
5265 kfree(net->ipv6.ip6_prohibit_entry);
5266 kfree(net->ipv6.ip6_blk_hole_entry);
5267#endif
41bb78b4 5268 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
cdb18761
DL		 5269}
5270
d189634e
TG		 5271static int __net_init ip6_route_net_init_late(struct net *net)
5272{
5273#ifdef CONFIG_PROC_FS
c3506372
CH		 5274 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
5275 sizeof(struct ipv6_route_iter));
3617d949
CH		 5276 proc_create_net_single("rt6_stats", 0444, net->proc_net,
5277 rt6_stats_seq_show, NULL);
d189634e
TG		 5278#endif
5279 return 0;
5280}
5281
5282static void __net_exit ip6_route_net_exit_late(struct net *net)
5283{
5284#ifdef CONFIG_PROC_FS
ece31ffd
G		 5285 remove_proc_entry("ipv6_route", net->proc_net);
5286 remove_proc_entry("rt6_stats", net->proc_net);
d189634e
TG		 5287#endif
5288}
5289
cdb18761
DL		 5290static struct pernet_operations ip6_route_net_ops = {
5291 .init = ip6_route_net_init,
5292 .exit = ip6_route_net_exit,
5293};
5294
c3426b47
DM		 5295static int __net_init ipv6_inetpeer_init(struct net *net)
5296{
5297 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
5298
5299 if (!bp)
5300 return -ENOMEM;
5301 inet_peer_base_init(bp);
5302 net->ipv6.peers = bp;
5303 return 0;
5304}
5305
5306static void __net_exit ipv6_inetpeer_exit(struct net *net)
5307{
5308 struct inet_peer_base *bp = net->ipv6.peers;
5309
5310 net->ipv6.peers = NULL;
56a6b248 5311 inetpeer_invalidate_tree(bp);
c3426b47
DM		 5312 kfree(bp);
5313}
5314
2b823f72 5315static struct pernet_operations ipv6_inetpeer_ops = {
c3426b47
DM		 5316 .init = ipv6_inetpeer_init,
5317 .exit = ipv6_inetpeer_exit,
5318};
5319
d189634e
TG		 5320static struct pernet_operations ip6_route_net_late_ops = {
5321 .init = ip6_route_net_init_late,
5322 .exit = ip6_route_net_exit_late,
5323};
5324
8ed67789
DL		 5325static struct notifier_block ip6_route_dev_notifier = {
5326 .notifier_call = ip6_route_dev_notify,
242d3a49 5327 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
8ed67789
DL		 5328};
5329
2f460933
WC		 5330void __init ip6_route_init_special_entries(void)
5331{
5332 /* Registering of the loopback is done before this portion of code,
5333 * the loopback reference in rt6_info will not be taken, do it
5334 * manually for init_net */
421842ed 5335 init_net.ipv6.fib6_null_entry->fib6_nh.nh_dev = init_net.loopback_dev;
2f460933
WC		 5336 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
5337 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
5338 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
5339 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
5340 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
5341 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
5342 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
5343 #endif
5344}
5345
433d49c3 5346int __init ip6_route_init(void)
1da177e4 5347{
433d49c3 5348 int ret;
8d0b94af 5349 int cpu;
433d49c3 5350
9a7ec3a9
DL		 5351 ret = -ENOMEM;
5352 ip6_dst_ops_template.kmem_cachep =
e5d679f3 5353 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
f845ab6b 5354 SLAB_HWCACHE_ALIGN, NULL);
9a7ec3a9 5355 if (!ip6_dst_ops_template.kmem_cachep)
c19a28e1 5356 goto out;
14e50e57 5357
fc66f95c 5358 ret = dst_entries_init(&ip6_dst_blackhole_ops);
8ed67789 5359 if (ret)
bdb3289f 5360 goto out_kmem_cache;
bdb3289f 5361
c3426b47
DM		 5362 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
5363 if (ret)
e8803b6c 5364 goto out_dst_entries;
2a0c451a 5365
7e52b33b
DM		 5366 ret = register_pernet_subsys(&ip6_route_net_ops);
5367 if (ret)
5368 goto out_register_inetpeer;
c3426b47 5369
5dc121e9
AE		 5370 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
5371
e8803b6c 5372 ret = fib6_init();
433d49c3 5373 if (ret)
8ed67789 5374 goto out_register_subsys;
433d49c3 5375
433d49c3
DL		 5376 ret = xfrm6_init();
5377 if (ret)
e8803b6c 5378 goto out_fib6_init;
c35b7e72 5379
433d49c3
DL		 5380 ret = fib6_rules_init();
5381 if (ret)
5382 goto xfrm6_init;
7e5449c2 5383
d189634e
TG		 5384 ret = register_pernet_subsys(&ip6_route_net_late_ops);
5385 if (ret)
5386 goto fib6_rules_init;
5387
16feebcf
FW		 5388 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
5389 inet6_rtm_newroute, NULL, 0);
5390 if (ret < 0)
5391 goto out_register_late_subsys;
5392
5393 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
5394 inet6_rtm_delroute, NULL, 0);
5395 if (ret < 0)
5396 goto out_register_late_subsys;
5397
5398 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
5399 inet6_rtm_getroute, NULL,
5400 RTNL_FLAG_DOIT_UNLOCKED);
5401 if (ret < 0)
d189634e 5402 goto out_register_late_subsys;
c127ea2c 5403
8ed67789 5404 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
cdb18761 5405 if (ret)
d189634e 5406 goto out_register_late_subsys;
8ed67789 5407
8d0b94af
MKL		 5408 for_each_possible_cpu(cpu) {
5409 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
5410
5411 INIT_LIST_HEAD(&ul->head);
5412 spin_lock_init(&ul->lock);
5413 }
5414
433d49c3
DL		 5415out:
5416 return ret;
5417
d189634e 5418out_register_late_subsys:
16feebcf 5419 rtnl_unregister_all(PF_INET6);
d189634e 5420 unregister_pernet_subsys(&ip6_route_net_late_ops);
433d49c3 5421fib6_rules_init:
433d49c3
DL		 5422 fib6_rules_cleanup();
5423xfrm6_init:
433d49c3 5424 xfrm6_fini();
2a0c451a
TG		 5425out_fib6_init:
5426 fib6_gc_cleanup();
8ed67789
DL		 5427out_register_subsys:
5428 unregister_pernet_subsys(&ip6_route_net_ops);
7e52b33b
DM		 5429out_register_inetpeer:
5430 unregister_pernet_subsys(&ipv6_inetpeer_ops);
fc66f95c
ED		 5431out_dst_entries:
5432 dst_entries_destroy(&ip6_dst_blackhole_ops);
433d49c3 5433out_kmem_cache:
f2fc6a54 5434 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
433d49c3 5435 goto out;
1da177e4
LT		 5436}
5437
5438void ip6_route_cleanup(void)
5439{
8ed67789 5440 unregister_netdevice_notifier(&ip6_route_dev_notifier);
d189634e 5441 unregister_pernet_subsys(&ip6_route_net_late_ops);
101367c2 5442 fib6_rules_cleanup();
1da177e4 5443 xfrm6_fini();
1da177e4 5444 fib6_gc_cleanup();
c3426b47 5445 unregister_pernet_subsys(&ipv6_inetpeer_ops);
8ed67789 5446 unregister_pernet_subsys(&ip6_route_net_ops);
41bb78b4 5447 dst_entries_destroy(&ip6_dst_blackhole_ops);
f2fc6a54 5448 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
1da177e4 5449}
Linux 6.x block layer and io_uring tree(s)