projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net: mvneta: Don't advertise 2.5G modes
[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) {
ad65a2f0 2235 if (dst_hold_safe(&rt->dst))
93531c67 2236 rt6_remove_exception_rt(rt);
c5cff856 2237 } else {
a68886a6 2238 struct fib6_info *from;
c5cff856
WW		 2239 struct fib6_node *fn;
2240
a68886a6
DA		 2241 from = rcu_dereference(rt->from);
2242 if (from) {
2243 fn = rcu_dereference(from->fib6_node);
2244 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2245 fn->fn_sernum = -1;
2246 }
1eb4f758 2247 }
8a14e46f 2248 rcu_read_unlock();
1da177e4
LT		 2249 }
2250}
2251
6a3e030f
DA		 2252static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2253{
a68886a6
DA		 2254 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2255 struct fib6_info *from;
2256
2257 rcu_read_lock();
2258 from = rcu_dereference(rt0->from);
2259 if (from)
2260 rt0->dst.expires = from->expires;
2261 rcu_read_unlock();
2262 }
6a3e030f
DA		 2263
2264 dst_set_expires(&rt0->dst, timeout);
2265 rt0->rt6i_flags |= RTF_EXPIRES;
2266}
2267
45e4fd26
MKL		 2268static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2269{
2270 struct net *net = dev_net(rt->dst.dev);
2271
d4ead6b3 2272 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
45e4fd26 2273 rt->rt6i_flags |= RTF_MODIFIED;
45e4fd26
MKL		 2274 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2275}
2276
0d3f6d29
MKL		 2277static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2278{
a68886a6
DA		 2279 bool from_set;
2280
2281 rcu_read_lock();
2282 from_set = !!rcu_dereference(rt->from);
2283 rcu_read_unlock();
2284
0d3f6d29 2285 return !(rt->rt6i_flags & RTF_CACHE) &&
a68886a6 2286 (rt->rt6i_flags & RTF_PCPU || from_set);
0d3f6d29
MKL		 2287}
2288
45e4fd26
MKL		 2289static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2290 const struct ipv6hdr *iph, u32 mtu)
1da177e4 2291{
0dec879f 2292 const struct in6_addr *daddr, *saddr;
67ba4152 2293 struct rt6_info *rt6 = (struct rt6_info *)dst;
1da177e4 2294
19bda36c
XL		 2295 if (dst_metric_locked(dst, RTAX_MTU))
2296 return;
2297
0dec879f
JA		 2298 if (iph) {
2299 daddr = &iph->daddr;
2300 saddr = &iph->saddr;
2301 } else if (sk) {
2302 daddr = &sk->sk_v6_daddr;
2303 saddr = &inet6_sk(sk)->saddr;
2304 } else {
2305 daddr = NULL;
2306 saddr = NULL;
2307 }
2308 dst_confirm_neigh(dst, daddr);
45e4fd26
MKL		 2309 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
2310 if (mtu >= dst_mtu(dst))
2311 return;
9d289715 2312
0d3f6d29 2313 if (!rt6_cache_allowed_for_pmtu(rt6)) {
45e4fd26 2314 rt6_do_update_pmtu(rt6, mtu);
2b760fcf
WW		 2315 /* update rt6_ex->stamp for cache */
2316 if (rt6->rt6i_flags & RTF_CACHE)
2317 rt6_update_exception_stamp_rt(rt6);
0dec879f 2318 } else if (daddr) {
a68886a6 2319 struct fib6_info *from;
45e4fd26
MKL		 2320 struct rt6_info *nrt6;
2321
4d85cd0c 2322 rcu_read_lock();
a68886a6
DA		 2323 from = rcu_dereference(rt6->from);
2324 nrt6 = ip6_rt_cache_alloc(from, daddr, saddr);
45e4fd26
MKL		 2325 if (nrt6) {
2326 rt6_do_update_pmtu(nrt6, mtu);
a68886a6 2327 if (rt6_insert_exception(nrt6, from))
2b760fcf 2328 dst_release_immediate(&nrt6->dst);
45e4fd26 2329 }
a68886a6 2330 rcu_read_unlock();
1da177e4
LT		 2331 }
2332}
2333
45e4fd26
MKL		 2334static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2335 struct sk_buff *skb, u32 mtu)
2336{
2337 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
2338}
2339
42ae66c8 2340void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
e2d118a1 2341 int oif, u32 mark, kuid_t uid)
81aded24
DM		 2342{
2343 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2344 struct dst_entry *dst;
dc92095d
MŻ		 2345 struct flowi6 fl6 = {
2346 .flowi6_oif = oif,
2347 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2348 .daddr = iph->daddr,
2349 .saddr = iph->saddr,
2350 .flowlabel = ip6_flowinfo(iph),
2351 .flowi6_uid = uid,
2352 };
81aded24
DM		 2353
2354 dst = ip6_route_output(net, NULL, &fl6);
2355 if (!dst->error)
45e4fd26 2356 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
81aded24
DM		 2357 dst_release(dst);
2358}
2359EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2360
2361void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2362{
33c162a9
MKL		 2363 struct dst_entry *dst;
2364
81aded24 2365 ip6_update_pmtu(skb, sock_net(sk), mtu,
e2d118a1 2366 sk->sk_bound_dev_if, sk->sk_mark, sk->sk_uid);
33c162a9
MKL		 2367
2368 dst = __sk_dst_get(sk);
2369 if (!dst || !dst->obsolete ||
2370 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2371 return;
2372
2373 bh_lock_sock(sk);
2374 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2375 ip6_datagram_dst_update(sk, false);
2376 bh_unlock_sock(sk);
81aded24
DM		 2377}
2378EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2379
7d6850f7
AK		 2380void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2381 const struct flowi6 *fl6)
2382{
2383#ifdef CONFIG_IPV6_SUBTREES
2384 struct ipv6_pinfo *np = inet6_sk(sk);
2385#endif
2386
2387 ip6_dst_store(sk, dst,
2388 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2389 &sk->sk_v6_daddr : NULL,
2390#ifdef CONFIG_IPV6_SUBTREES
2391 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2392 &np->saddr :
2393#endif
2394 NULL);
2395}
2396
b55b76b2
DJ		 2397/* Handle redirects */
2398struct ip6rd_flowi {
2399 struct flowi6 fl6;
2400 struct in6_addr gateway;
2401};
2402
2403static struct rt6_info *__ip6_route_redirect(struct net *net,
2404 struct fib6_table *table,
2405 struct flowi6 *fl6,
b75cc8f9 2406 const struct sk_buff *skb,
b55b76b2
DJ		 2407 int flags)
2408{
2409 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
23fb93a4 2410 struct rt6_info *ret = NULL, *rt_cache;
8d1c802b 2411 struct fib6_info *rt;
b55b76b2
DJ		 2412 struct fib6_node *fn;
2413
2414 /* Get the "current" route for this destination and
67c408cf 2415 * check if the redirect has come from appropriate router.
b55b76b2
DJ		 2416 *
2417 * RFC 4861 specifies that redirects should only be
2418 * accepted if they come from the nexthop to the target.
2419 * Due to the way the routes are chosen, this notion
2420 * is a bit fuzzy and one might need to check all possible
2421 * routes.
2422 */
2423
66f5d6ce 2424 rcu_read_lock();
6454743b 2425 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
b55b76b2 2426restart:
66f5d6ce 2427 for_each_fib6_node_rt_rcu(fn) {
5e670d84 2428 if (rt->fib6_nh.nh_flags & RTNH_F_DEAD)
8067bb8c 2429 continue;
14895687 2430 if (fib6_check_expired(rt))
b55b76b2 2431 continue;
93c2fb25 2432 if (rt->fib6_flags & RTF_REJECT)
b55b76b2 2433 break;
93c2fb25 2434 if (!(rt->fib6_flags & RTF_GATEWAY))
b55b76b2 2435 continue;
5e670d84 2436 if (fl6->flowi6_oif != rt->fib6_nh.nh_dev->ifindex)
b55b76b2 2437 continue;
2b760fcf
WW		 2438 /* rt_cache's gateway might be different from its 'parent'
2439 * in the case of an ip redirect.
2440 * So we keep searching in the exception table if the gateway
2441 * is different.
2442 */
5e670d84 2443 if (!ipv6_addr_equal(&rdfl->gateway, &rt->fib6_nh.nh_gw)) {
2b760fcf
WW		 2444 rt_cache = rt6_find_cached_rt(rt,
2445 &fl6->daddr,
2446 &fl6->saddr);
2447 if (rt_cache &&
2448 ipv6_addr_equal(&rdfl->gateway,
2449 &rt_cache->rt6i_gateway)) {
23fb93a4 2450 ret = rt_cache;
2b760fcf
WW		 2451 break;
2452 }
b55b76b2 2453 continue;
2b760fcf 2454 }
b55b76b2
DJ		 2455 break;
2456 }
2457
2458 if (!rt)
421842ed 2459 rt = net->ipv6.fib6_null_entry;
93c2fb25 2460 else if (rt->fib6_flags & RTF_REJECT) {
23fb93a4 2461 ret = net->ipv6.ip6_null_entry;
b0a1ba59
MKL		 2462 goto out;
2463 }
2464
421842ed 2465 if (rt == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2466 fn = fib6_backtrack(fn, &fl6->saddr);
2467 if (fn)
2468 goto restart;
b55b76b2 2469 }
a3c00e46 2470
b0a1ba59 2471out:
23fb93a4 2472 if (ret)
e873e4b9 2473 ip6_hold_safe(net, &ret, true);
23fb93a4
DA		 2474 else
2475 ret = ip6_create_rt_rcu(rt);
b55b76b2 2476
66f5d6ce 2477 rcu_read_unlock();
b55b76b2 2478
b65f164d 2479 trace_fib6_table_lookup(net, rt, table, fl6);
23fb93a4 2480 return ret;
b55b76b2
DJ		 2481};
2482
2483static struct dst_entry *ip6_route_redirect(struct net *net,
b75cc8f9
DA		 2484 const struct flowi6 *fl6,
2485 const struct sk_buff *skb,
2486 const struct in6_addr *gateway)
b55b76b2
DJ		 2487{
2488 int flags = RT6_LOOKUP_F_HAS_SADDR;
2489 struct ip6rd_flowi rdfl;
2490
2491 rdfl.fl6 = *fl6;
2492 rdfl.gateway = *gateway;
2493
b75cc8f9 2494 return fib6_rule_lookup(net, &rdfl.fl6, skb,
b55b76b2
DJ		 2495 flags, __ip6_route_redirect);
2496}
2497
e2d118a1
LC		 2498void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
2499 kuid_t uid)
3a5ad2ee
DM		 2500{
2501 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2502 struct dst_entry *dst;
1f7f10ac
MŻ		 2503 struct flowi6 fl6 = {
2504 .flowi6_iif = LOOPBACK_IFINDEX,
2505 .flowi6_oif = oif,
2506 .flowi6_mark = mark,
2507 .daddr = iph->daddr,
2508 .saddr = iph->saddr,
2509 .flowlabel = ip6_flowinfo(iph),
2510 .flowi6_uid = uid,
2511 };
3a5ad2ee 2512
b75cc8f9 2513 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
b55b76b2 2514 rt6_do_redirect(dst, NULL, skb);
3a5ad2ee
DM		 2515 dst_release(dst);
2516}
2517EXPORT_SYMBOL_GPL(ip6_redirect);
2518
d456336d 2519void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
c92a59ec
DJ		 2520{
2521 const struct ipv6hdr *iph = ipv6_hdr(skb);
2522 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
2523 struct dst_entry *dst;
0b26fb17
MŻ		 2524 struct flowi6 fl6 = {
2525 .flowi6_iif = LOOPBACK_IFINDEX,
2526 .flowi6_oif = oif,
0b26fb17
MŻ		 2527 .daddr = msg->dest,
2528 .saddr = iph->daddr,
2529 .flowi6_uid = sock_net_uid(net, NULL),
2530 };
c92a59ec 2531
b75cc8f9 2532 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
b55b76b2 2533 rt6_do_redirect(dst, NULL, skb);
c92a59ec
DJ		 2534 dst_release(dst);
2535}
2536
3a5ad2ee
DM		 2537void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
2538{
e2d118a1
LC		 2539 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
2540 sk->sk_uid);
3a5ad2ee
DM		 2541}
2542EXPORT_SYMBOL_GPL(ip6_sk_redirect);
2543
0dbaee3b 2544static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1da177e4 2545{
0dbaee3b
DM		 2546 struct net_device *dev = dst->dev;
2547 unsigned int mtu = dst_mtu(dst);
2548 struct net *net = dev_net(dev);
2549
1da177e4
LT		 2550 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
2551
5578689a
DL		 2552 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
2553 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1da177e4
LT		 2554
2555 /*
1ab1457c
YH		 2556 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
2557 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
2558 * IPV6_MAXPLEN is also valid and means: "any MSS,
1da177e4
LT		 2559 * rely only on pmtu discovery"
2560 */
2561 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
2562 mtu = IPV6_MAXPLEN;
2563 return mtu;
2564}
2565
ebb762f2 2566static unsigned int ip6_mtu(const struct dst_entry *dst)
d33e4553 2567{
d33e4553 2568 struct inet6_dev *idev;
d4ead6b3 2569 unsigned int mtu;
4b32b5ad
MKL		 2570
2571 mtu = dst_metric_raw(dst, RTAX_MTU);
618f9bc7 2572 if (mtu)
30f78d8e 2573 goto out;
618f9bc7
SK		 2574
2575 mtu = IPV6_MIN_MTU;
d33e4553
DM		 2576
2577 rcu_read_lock();
2578 idev = __in6_dev_get(dst->dev);
2579 if (idev)
2580 mtu = idev->cnf.mtu6;
2581 rcu_read_unlock();
2582
30f78d8e 2583out:
14972cbd
RP		 2584 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
2585
2586 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
d33e4553
DM		 2587}
2588
901731b8
DA		 2589/* MTU selection:
2590 * 1. mtu on route is locked - use it
2591 * 2. mtu from nexthop exception
2592 * 3. mtu from egress device
2593 *
2594 * based on ip6_dst_mtu_forward and exception logic of
2595 * rt6_find_cached_rt; called with rcu_read_lock
2596 */
2597u32 ip6_mtu_from_fib6(struct fib6_info *f6i, struct in6_addr *daddr,
2598 struct in6_addr *saddr)
2599{
2600 struct rt6_exception_bucket *bucket;
2601 struct rt6_exception *rt6_ex;
2602 struct in6_addr *src_key;
2603 struct inet6_dev *idev;
2604 u32 mtu = 0;
2605
2606 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
2607 mtu = f6i->fib6_pmtu;
2608 if (mtu)
2609 goto out;
2610 }
2611
2612 src_key = NULL;
2613#ifdef CONFIG_IPV6_SUBTREES
2614 if (f6i->fib6_src.plen)
2615 src_key = saddr;
2616#endif
2617
2618 bucket = rcu_dereference(f6i->rt6i_exception_bucket);
2619 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
2620 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
2621 mtu = dst_metric_raw(&rt6_ex->rt6i->dst, RTAX_MTU);
2622
2623 if (likely(!mtu)) {
2624 struct net_device *dev = fib6_info_nh_dev(f6i);
2625
2626 mtu = IPV6_MIN_MTU;
2627 idev = __in6_dev_get(dev);
2628 if (idev && idev->cnf.mtu6 > mtu)
2629 mtu = idev->cnf.mtu6;
2630 }
2631
2632 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
2633out:
2634 return mtu - lwtunnel_headroom(fib6_info_nh_lwt(f6i), mtu);
2635}
2636
3b00944c 2637struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
87a11578 2638 struct flowi6 *fl6)
1da177e4 2639{
87a11578 2640 struct dst_entry *dst;
1da177e4
LT		 2641 struct rt6_info *rt;
2642 struct inet6_dev *idev = in6_dev_get(dev);
c346dca1 2643 struct net *net = dev_net(dev);
1da177e4 2644
38308473 2645 if (unlikely(!idev))
122bdf67 2646 return ERR_PTR(-ENODEV);
1da177e4 2647
ad706862 2648 rt = ip6_dst_alloc(net, dev, 0);
38308473 2649 if (unlikely(!rt)) {
1da177e4 2650 in6_dev_put(idev);
87a11578 2651 dst = ERR_PTR(-ENOMEM);
1da177e4
LT		 2652 goto out;
2653 }
2654
8e2ec639 2655 rt->dst.flags |= DST_HOST;
588753f1 2656 rt->dst.input = ip6_input;
8e2ec639 2657 rt->dst.output = ip6_output;
550bab42 2658 rt->rt6i_gateway = fl6->daddr;
87a11578 2659 rt->rt6i_dst.addr = fl6->daddr;
8e2ec639
YZ		 2660 rt->rt6i_dst.plen = 128;
2661 rt->rt6i_idev = idev;
14edd87d 2662 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1da177e4 2663
4c981e28 2664 /* Add this dst into uncached_list so that rt6_disable_ip() can
587fea74
WW		 2665 * do proper release of the net_device
2666 */
2667 rt6_uncached_list_add(rt);
81eb8447 2668 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
1da177e4 2669
87a11578
DM		 2670 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
2671
1da177e4 2672out:
87a11578 2673 return dst;
1da177e4
LT		 2674}
2675
569d3645 2676static int ip6_dst_gc(struct dst_ops *ops)
1da177e4 2677{
86393e52 2678 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
7019b78e
DL		 2679 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
2680 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
2681 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
2682 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
2683 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
fc66f95c 2684 int entries;
7019b78e 2685
fc66f95c 2686 entries = dst_entries_get_fast(ops);
49a18d86 2687 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
fc66f95c 2688 entries <= rt_max_size)
1da177e4
LT		 2689 goto out;
2690
6891a346 2691 net->ipv6.ip6_rt_gc_expire++;
14956643 2692 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
fc66f95c
ED		 2693 entries = dst_entries_get_slow(ops);
2694 if (entries < ops->gc_thresh)
7019b78e 2695 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1da177e4 2696out:
7019b78e 2697 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
fc66f95c 2698 return entries > rt_max_size;
1da177e4
LT		 2699}
2700
8c14586f
DA		 2701static struct rt6_info *ip6_nh_lookup_table(struct net *net,
2702 struct fib6_config *cfg,
f4797b33
DA		 2703 const struct in6_addr *gw_addr,
2704 u32 tbid, int flags)
8c14586f
DA		 2705{
2706 struct flowi6 fl6 = {
2707 .flowi6_oif = cfg->fc_ifindex,
2708 .daddr = *gw_addr,
2709 .saddr = cfg->fc_prefsrc,
2710 };
2711 struct fib6_table *table;
2712 struct rt6_info *rt;
8c14586f 2713
f4797b33 2714 table = fib6_get_table(net, tbid);
8c14586f
DA		 2715 if (!table)
2716 return NULL;
2717
2718 if (!ipv6_addr_any(&cfg->fc_prefsrc))
2719 flags |= RT6_LOOKUP_F_HAS_SADDR;
2720
f4797b33 2721 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
b75cc8f9 2722 rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, NULL, flags);
8c14586f
DA		 2723
2724 /* if table lookup failed, fall back to full lookup */
2725 if (rt == net->ipv6.ip6_null_entry) {
2726 ip6_rt_put(rt);
2727 rt = NULL;
2728 }
2729
2730 return rt;
2731}
2732
fc1e64e1
DA		 2733static int ip6_route_check_nh_onlink(struct net *net,
2734 struct fib6_config *cfg,
9fbb704c 2735 const struct net_device *dev,
fc1e64e1
DA		 2736 struct netlink_ext_ack *extack)
2737{
44750f84 2738 u32 tbid = l3mdev_fib_table(dev) ? : RT_TABLE_MAIN;
fc1e64e1
DA		 2739 const struct in6_addr *gw_addr = &cfg->fc_gateway;
2740 u32 flags = RTF_LOCAL | RTF_ANYCAST | RTF_REJECT;
2741 struct rt6_info *grt;
2742 int err;
2743
2744 err = 0;
2745 grt = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0);
2746 if (grt) {
58e354c0 2747 if (!grt->dst.error &&
4ed591c8
DA		 2748 /* ignore match if it is the default route */
2749 grt->from && !ipv6_addr_any(&grt->from->fib6_dst.addr) &&
58e354c0 2750 (grt->rt6i_flags & flags || dev != grt->dst.dev)) {
44750f84
DA		 2751 NL_SET_ERR_MSG(extack,
2752 "Nexthop has invalid gateway or device mismatch");
fc1e64e1
DA		 2753 err = -EINVAL;
2754 }
2755
2756 ip6_rt_put(grt);
2757 }
2758
2759 return err;
2760}
2761
1edce99f
DA		 2762static int ip6_route_check_nh(struct net *net,
2763 struct fib6_config *cfg,
2764 struct net_device **_dev,
2765 struct inet6_dev **idev)
2766{
2767 const struct in6_addr *gw_addr = &cfg->fc_gateway;
2768 struct net_device *dev = _dev ? *_dev : NULL;
2769 struct rt6_info *grt = NULL;
2770 int err = -EHOSTUNREACH;
2771
2772 if (cfg->fc_table) {
f4797b33
DA		 2773 int flags = RT6_LOOKUP_F_IFACE;
2774
2775 grt = ip6_nh_lookup_table(net, cfg, gw_addr,
2776 cfg->fc_table, flags);
1edce99f
DA		 2777 if (grt) {
2778 if (grt->rt6i_flags & RTF_GATEWAY ||
2779 (dev && dev != grt->dst.dev)) {
2780 ip6_rt_put(grt);
2781 grt = NULL;
2782 }
2783 }
2784 }
2785
2786 if (!grt)
b75cc8f9 2787 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, NULL, 1);
1edce99f
DA		 2788
2789 if (!grt)
2790 goto out;
2791
2792 if (dev) {
2793 if (dev != grt->dst.dev) {
2794 ip6_rt_put(grt);
2795 goto out;
2796 }
2797 } else {
2798 *_dev = dev = grt->dst.dev;
2799 *idev = grt->rt6i_idev;
2800 dev_hold(dev);
2801 in6_dev_hold(grt->rt6i_idev);
2802 }
2803
2804 if (!(grt->rt6i_flags & RTF_GATEWAY))
2805 err = 0;
2806
2807 ip6_rt_put(grt);
2808
2809out:
2810 return err;
2811}
2812
9fbb704c
DA		 2813static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
2814 struct net_device **_dev, struct inet6_dev **idev,
2815 struct netlink_ext_ack *extack)
2816{
2817 const struct in6_addr *gw_addr = &cfg->fc_gateway;
2818 int gwa_type = ipv6_addr_type(gw_addr);
232378e8 2819 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
9fbb704c 2820 const struct net_device *dev = *_dev;
232378e8 2821 bool need_addr_check = !dev;
9fbb704c
DA		 2822 int err = -EINVAL;
2823
2824 /* if gw_addr is local we will fail to detect this in case
2825 * address is still TENTATIVE (DAD in progress). rt6_lookup()
2826 * will return already-added prefix route via interface that
2827 * prefix route was assigned to, which might be non-loopback.
2828 */
232378e8
DA		 2829 if (dev &&
2830 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
2831 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
9fbb704c
DA		 2832 goto out;
2833 }
2834
2835 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
2836 /* IPv6 strictly inhibits using not link-local
2837 * addresses as nexthop address.
2838 * Otherwise, router will not able to send redirects.
2839 * It is very good, but in some (rare!) circumstances
2840 * (SIT, PtP, NBMA NOARP links) it is handy to allow
2841 * some exceptions. --ANK
2842 * We allow IPv4-mapped nexthops to support RFC4798-type
2843 * addressing
2844 */
2845 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
2846 NL_SET_ERR_MSG(extack, "Invalid gateway address");
2847 goto out;
2848 }
2849
2850 if (cfg->fc_flags & RTNH_F_ONLINK)
2851 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
2852 else
2853 err = ip6_route_check_nh(net, cfg, _dev, idev);
2854
2855 if (err)
2856 goto out;
2857 }
2858
2859 /* reload in case device was changed */
2860 dev = *_dev;
2861
2862 err = -EINVAL;
2863 if (!dev) {
2864 NL_SET_ERR_MSG(extack, "Egress device not specified");
2865 goto out;
2866 } else if (dev->flags & IFF_LOOPBACK) {
2867 NL_SET_ERR_MSG(extack,
2868 "Egress device can not be loopback device for this route");
2869 goto out;
2870 }
232378e8
DA		 2871
2872 /* if we did not check gw_addr above, do so now that the
2873 * egress device has been resolved.
2874 */
2875 if (need_addr_check &&
2876 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
2877 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
2878 goto out;
2879 }
2880
9fbb704c
DA		 2881 err = 0;
2882out:
2883 return err;
2884}
2885
8d1c802b 2886static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
acb54e3c 2887 gfp_t gfp_flags,
333c4301 2888 struct netlink_ext_ack *extack)
1da177e4 2889{
5578689a 2890 struct net *net = cfg->fc_nlinfo.nl_net;
8d1c802b 2891 struct fib6_info *rt = NULL;
1da177e4
LT		 2892 struct net_device *dev = NULL;
2893 struct inet6_dev *idev = NULL;
c71099ac 2894 struct fib6_table *table;
1da177e4 2895 int addr_type;
8c5b83f0 2896 int err = -EINVAL;
1da177e4 2897
557c44be 2898 /* RTF_PCPU is an internal flag; can not be set by userspace */
d5d531cb
DA		 2899 if (cfg->fc_flags & RTF_PCPU) {
2900 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
557c44be 2901 goto out;
d5d531cb 2902 }
557c44be 2903
2ea2352e
WW		 2904 /* RTF_CACHE is an internal flag; can not be set by userspace */
2905 if (cfg->fc_flags & RTF_CACHE) {
2906 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
2907 goto out;
2908 }
2909
e8478e80
DA		 2910 if (cfg->fc_type > RTN_MAX) {
2911 NL_SET_ERR_MSG(extack, "Invalid route type");
2912 goto out;
2913 }
2914
d5d531cb
DA		 2915 if (cfg->fc_dst_len > 128) {
2916 NL_SET_ERR_MSG(extack, "Invalid prefix length");
2917 goto out;
2918 }
2919 if (cfg->fc_src_len > 128) {
2920 NL_SET_ERR_MSG(extack, "Invalid source address length");
8c5b83f0 2921 goto out;
d5d531cb 2922 }
1da177e4 2923#ifndef CONFIG_IPV6_SUBTREES
d5d531cb
DA		 2924 if (cfg->fc_src_len) {
2925 NL_SET_ERR_MSG(extack,
2926 "Specifying source address requires IPV6_SUBTREES to be enabled");
8c5b83f0 2927 goto out;
d5d531cb 2928 }
1da177e4 2929#endif
86872cb5 2930 if (cfg->fc_ifindex) {
1da177e4 2931 err = -ENODEV;
5578689a 2932 dev = dev_get_by_index(net, cfg->fc_ifindex);
1da177e4
LT		 2933 if (!dev)
2934 goto out;
2935 idev = in6_dev_get(dev);
2936 if (!idev)
2937 goto out;
2938 }
2939
86872cb5
TG		 2940 if (cfg->fc_metric == 0)
2941 cfg->fc_metric = IP6_RT_PRIO_USER;
1da177e4 2942
fc1e64e1
DA		 2943 if (cfg->fc_flags & RTNH_F_ONLINK) {
2944 if (!dev) {
2945 NL_SET_ERR_MSG(extack,
2946 "Nexthop device required for onlink");
2947 err = -ENODEV;
2948 goto out;
2949 }
2950
2951 if (!(dev->flags & IFF_UP)) {
2952 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
2953 err = -ENETDOWN;
2954 goto out;
2955 }
2956 }
2957
d71314b4 2958 err = -ENOBUFS;
38308473
DM		 2959 if (cfg->fc_nlinfo.nlh &&
2960 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
d71314b4 2961 table = fib6_get_table(net, cfg->fc_table);
38308473 2962 if (!table) {
f3213831 2963 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
d71314b4
MV		 2964 table = fib6_new_table(net, cfg->fc_table);
2965 }
2966 } else {
2967 table = fib6_new_table(net, cfg->fc_table);
2968 }
38308473
DM		 2969
2970 if (!table)
c71099ac 2971 goto out;
c71099ac 2972
93531c67
DA		 2973 err = -ENOMEM;
2974 rt = fib6_info_alloc(gfp_flags);
2975 if (!rt)
1da177e4 2976 goto out;
93531c67 2977
767a2217
DA		 2978 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len);
2979 if (IS_ERR(rt->fib6_metrics)) {
2980 err = PTR_ERR(rt->fib6_metrics);
fda21d46
ED		 2981 /* Do not leave garbage there. */
2982 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
767a2217
DA		 2983 goto out;
2984 }
2985
93531c67
DA		 2986 if (cfg->fc_flags & RTF_ADDRCONF)
2987 rt->dst_nocount = true;
1da177e4 2988
1716a961 2989 if (cfg->fc_flags & RTF_EXPIRES)
14895687 2990 fib6_set_expires(rt, jiffies +
1716a961
G		 2991 clock_t_to_jiffies(cfg->fc_expires));
2992 else
14895687 2993 fib6_clean_expires(rt);
1da177e4 2994
86872cb5
TG		 2995 if (cfg->fc_protocol == RTPROT_UNSPEC)
2996 cfg->fc_protocol = RTPROT_BOOT;
93c2fb25 2997 rt->fib6_protocol = cfg->fc_protocol;
86872cb5
TG		 2998
2999 addr_type = ipv6_addr_type(&cfg->fc_dst);
1da177e4 3000
19e42e45
RP		 3001 if (cfg->fc_encap) {
3002 struct lwtunnel_state *lwtstate;
3003
30357d7d 3004 err = lwtunnel_build_state(cfg->fc_encap_type,
127eb7cd 3005 cfg->fc_encap, AF_INET6, cfg,
9ae28727 3006 &lwtstate, extack);
19e42e45
RP		 3007 if (err)
3008 goto out;
5e670d84 3009 rt->fib6_nh.nh_lwtstate = lwtstate_get(lwtstate);
19e42e45
RP		 3010 }
3011
93c2fb25
DA		 3012 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3013 rt->fib6_dst.plen = cfg->fc_dst_len;
3014 if (rt->fib6_dst.plen == 128)
3b6761d1 3015 rt->dst_host = true;
e5fd387a 3016
1da177e4 3017#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 3018 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3019 rt->fib6_src.plen = cfg->fc_src_len;
1da177e4
LT		 3020#endif
3021
93c2fb25 3022 rt->fib6_metric = cfg->fc_metric;
5e670d84 3023 rt->fib6_nh.nh_weight = 1;
1da177e4 3024
e8478e80 3025 rt->fib6_type = cfg->fc_type;
1da177e4
LT		 3026
3027 /* We cannot add true routes via loopback here,
3028 they would result in kernel looping; promote them to reject routes
3029 */
86872cb5 3030 if ((cfg->fc_flags & RTF_REJECT) ||
38308473
DM		 3031 (dev && (dev->flags & IFF_LOOPBACK) &&
3032 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3033 !(cfg->fc_flags & RTF_LOCAL))) {
1da177e4 3034 /* hold loopback dev/idev if we haven't done so. */
5578689a 3035 if (dev != net->loopback_dev) {
1da177e4
LT		 3036 if (dev) {
3037 dev_put(dev);
3038 in6_dev_put(idev);
3039 }
5578689a 3040 dev = net->loopback_dev;
1da177e4
LT		 3041 dev_hold(dev);
3042 idev = in6_dev_get(dev);
3043 if (!idev) {
3044 err = -ENODEV;
3045 goto out;
3046 }
3047 }
93c2fb25 3048 rt->fib6_flags = RTF_REJECT|RTF_NONEXTHOP;
1da177e4
LT		 3049 goto install_route;
3050 }
3051
86872cb5 3052 if (cfg->fc_flags & RTF_GATEWAY) {
9fbb704c
DA		 3053 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3054 if (err)
48ed7b26 3055 goto out;
1da177e4 3056
93531c67 3057 rt->fib6_nh.nh_gw = cfg->fc_gateway;
1da177e4
LT		 3058 }
3059
3060 err = -ENODEV;
38308473 3061 if (!dev)
1da177e4
LT		 3062 goto out;
3063
428604fb
LB		 3064 if (idev->cnf.disable_ipv6) {
3065 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3066 err = -EACCES;
3067 goto out;
3068 }
3069
955ec4cb
DA		 3070 if (!(dev->flags & IFF_UP)) {
3071 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3072 err = -ENETDOWN;
3073 goto out;
3074 }
3075
c3968a85
DW		 3076 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3077 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
d5d531cb 3078 NL_SET_ERR_MSG(extack, "Invalid source address");
c3968a85
DW		 3079 err = -EINVAL;
3080 goto out;
3081 }
93c2fb25
DA		 3082 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3083 rt->fib6_prefsrc.plen = 128;
c3968a85 3084 } else
93c2fb25 3085 rt->fib6_prefsrc.plen = 0;
c3968a85 3086
93c2fb25 3087 rt->fib6_flags = cfg->fc_flags;
1da177e4
LT		 3088
3089install_route:
93c2fb25 3090 if (!(rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
5609b80a 3091 !netif_carrier_ok(dev))
5e670d84
DA		 3092 rt->fib6_nh.nh_flags |= RTNH_F_LINKDOWN;
3093 rt->fib6_nh.nh_flags |= (cfg->fc_flags & RTNH_F_ONLINK);
93531c67 3094 rt->fib6_nh.nh_dev = dev;
93c2fb25 3095 rt->fib6_table = table;
63152fc0 3096
dcd1f572
DA		 3097 if (idev)
3098 in6_dev_put(idev);
3099
8c5b83f0 3100 return rt;
6b9ea5a6
RP		 3101out:
3102 if (dev)
3103 dev_put(dev);
3104 if (idev)
3105 in6_dev_put(idev);
6b9ea5a6 3106
93531c67 3107 fib6_info_release(rt);
8c5b83f0 3108 return ERR_PTR(err);
6b9ea5a6
RP		 3109}
3110
acb54e3c 3111int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
333c4301 3112 struct netlink_ext_ack *extack)
6b9ea5a6 3113{
8d1c802b 3114 struct fib6_info *rt;
6b9ea5a6
RP		 3115 int err;
3116
acb54e3c 3117 rt = ip6_route_info_create(cfg, gfp_flags, extack);
d4ead6b3
DA		 3118 if (IS_ERR(rt))
3119 return PTR_ERR(rt);
6b9ea5a6 3120
d4ead6b3 3121 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
93531c67 3122 fib6_info_release(rt);
6b9ea5a6 3123
1da177e4
LT		 3124 return err;
3125}
3126
8d1c802b 3127static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
1da177e4 3128{
afb1d4b5 3129 struct net *net = info->nl_net;
c71099ac 3130 struct fib6_table *table;
afb1d4b5 3131 int err;
1da177e4 3132
421842ed 3133 if (rt == net->ipv6.fib6_null_entry) {
6825a26c
G		 3134 err = -ENOENT;
3135 goto out;
3136 }
6c813a72 3137
93c2fb25 3138 table = rt->fib6_table;
66f5d6ce 3139 spin_lock_bh(&table->tb6_lock);
86872cb5 3140 err = fib6_del(rt, info);
66f5d6ce 3141 spin_unlock_bh(&table->tb6_lock);
1da177e4 3142
6825a26c 3143out:
93531c67 3144 fib6_info_release(rt);
1da177e4
LT		 3145 return err;
3146}
3147
8d1c802b 3148int ip6_del_rt(struct net *net, struct fib6_info *rt)
e0a1ad73 3149{
afb1d4b5
DA		 3150 struct nl_info info = { .nl_net = net };
3151
528c4ceb 3152 return __ip6_del_rt(rt, &info);
e0a1ad73
TG		 3153}
3154
8d1c802b 3155static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
0ae81335
DA		 3156{
3157 struct nl_info *info = &cfg->fc_nlinfo;
e3330039 3158 struct net *net = info->nl_net;
16a16cd3 3159 struct sk_buff *skb = NULL;
0ae81335 3160 struct fib6_table *table;
e3330039 3161 int err = -ENOENT;
0ae81335 3162
421842ed 3163 if (rt == net->ipv6.fib6_null_entry)
e3330039 3164 goto out_put;
93c2fb25 3165 table = rt->fib6_table;
66f5d6ce 3166 spin_lock_bh(&table->tb6_lock);
0ae81335 3167
93c2fb25 3168 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
8d1c802b 3169 struct fib6_info *sibling, *next_sibling;
0ae81335 3170
16a16cd3
DA		 3171 /* prefer to send a single notification with all hops */
3172 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3173 if (skb) {
3174 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3175
d4ead6b3 3176 if (rt6_fill_node(net, skb, rt, NULL,
16a16cd3
DA		 3177 NULL, NULL, 0, RTM_DELROUTE,
3178 info->portid, seq, 0) < 0) {
3179 kfree_skb(skb);
3180 skb = NULL;
3181 } else
3182 info->skip_notify = 1;
3183 }
3184
0ae81335 3185 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25
DA		 3186 &rt->fib6_siblings,
3187 fib6_siblings) {
0ae81335
DA		 3188 err = fib6_del(sibling, info);
3189 if (err)
e3330039 3190 goto out_unlock;
0ae81335
DA		 3191 }
3192 }
3193
3194 err = fib6_del(rt, info);
e3330039 3195out_unlock:
66f5d6ce 3196 spin_unlock_bh(&table->tb6_lock);
e3330039 3197out_put:
93531c67 3198 fib6_info_release(rt);
16a16cd3
DA		 3199
3200 if (skb) {
e3330039 3201 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
16a16cd3
DA		 3202 info->nlh, gfp_any());
3203 }
0ae81335
DA		 3204 return err;
3205}
3206
23fb93a4
DA		 3207static int ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3208{
3209 int rc = -ESRCH;
3210
3211 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3212 goto out;
3213
3214 if (cfg->fc_flags & RTF_GATEWAY &&
3215 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3216 goto out;
3217 if (dst_hold_safe(&rt->dst))
3218 rc = rt6_remove_exception_rt(rt);
3219out:
3220 return rc;
3221}
3222
333c4301
DA		 3223static int ip6_route_del(struct fib6_config *cfg,
3224 struct netlink_ext_ack *extack)
1da177e4 3225{
8d1c802b 3226 struct rt6_info *rt_cache;
c71099ac 3227 struct fib6_table *table;
8d1c802b 3228 struct fib6_info *rt;
1da177e4 3229 struct fib6_node *fn;
1da177e4
LT		 3230 int err = -ESRCH;
3231
5578689a 3232 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
d5d531cb
DA		 3233 if (!table) {
3234 NL_SET_ERR_MSG(extack, "FIB table does not exist");
c71099ac 3235 return err;
d5d531cb 3236 }
c71099ac 3237
66f5d6ce 3238 rcu_read_lock();
1da177e4 3239
c71099ac 3240 fn = fib6_locate(&table->tb6_root,
86872cb5 3241 &cfg->fc_dst, cfg->fc_dst_len,
38fbeeee 3242 &cfg->fc_src, cfg->fc_src_len,
2b760fcf 3243 !(cfg->fc_flags & RTF_CACHE));
1ab1457c 3244
1da177e4 3245 if (fn) {
66f5d6ce 3246 for_each_fib6_node_rt_rcu(fn) {
2b760fcf 3247 if (cfg->fc_flags & RTF_CACHE) {
23fb93a4
DA		 3248 int rc;
3249
2b760fcf
WW		 3250 rt_cache = rt6_find_cached_rt(rt, &cfg->fc_dst,
3251 &cfg->fc_src);
23fb93a4
DA		 3252 if (rt_cache) {
3253 rc = ip6_del_cached_rt(rt_cache, cfg);
9e575010
ED		 3254 if (rc != -ESRCH) {
3255 rcu_read_unlock();
23fb93a4 3256 return rc;
9e575010 3257 }
23fb93a4
DA		 3258 }
3259 continue;
2b760fcf 3260 }
86872cb5 3261 if (cfg->fc_ifindex &&
5e670d84
DA		 3262 (!rt->fib6_nh.nh_dev ||
3263 rt->fib6_nh.nh_dev->ifindex != cfg->fc_ifindex))
1da177e4 3264 continue;
86872cb5 3265 if (cfg->fc_flags & RTF_GATEWAY &&
5e670d84 3266 !ipv6_addr_equal(&cfg->fc_gateway, &rt->fib6_nh.nh_gw))
1da177e4 3267 continue;
93c2fb25 3268 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
1da177e4 3269 continue;
93c2fb25 3270 if (cfg->fc_protocol && cfg->fc_protocol != rt->fib6_protocol)
c2ed1880 3271 continue;
e873e4b9
WW		 3272 if (!fib6_info_hold_safe(rt))
3273 continue;
66f5d6ce 3274 rcu_read_unlock();
1da177e4 3275
0ae81335
DA		 3276 /* if gateway was specified only delete the one hop */
3277 if (cfg->fc_flags & RTF_GATEWAY)
3278 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3279
3280 return __ip6_del_rt_siblings(rt, cfg);
1da177e4
LT		 3281 }
3282 }
66f5d6ce 3283 rcu_read_unlock();
1da177e4
LT		 3284
3285 return err;
3286}
3287
6700c270 3288static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
a6279458 3289{
a6279458 3290 struct netevent_redirect netevent;
e8599ff4 3291 struct rt6_info *rt, *nrt = NULL;
e8599ff4
DM		 3292 struct ndisc_options ndopts;
3293 struct inet6_dev *in6_dev;
3294 struct neighbour *neigh;
a68886a6 3295 struct fib6_info *from;
71bcdba0 3296 struct rd_msg *msg;
6e157b6a
DM		 3297 int optlen, on_link;
3298 u8 *lladdr;
e8599ff4 3299
29a3cad5 3300 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
71bcdba0 3301 optlen -= sizeof(*msg);
e8599ff4
DM		 3302
3303 if (optlen < 0) {
6e157b6a 3304 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
e8599ff4
DM		 3305 return;
3306 }
3307
71bcdba0 3308 msg = (struct rd_msg *)icmp6_hdr(skb);
e8599ff4 3309
71bcdba0 3310 if (ipv6_addr_is_multicast(&msg->dest)) {
6e157b6a 3311 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
e8599ff4
DM		 3312 return;
3313 }
3314
6e157b6a 3315 on_link = 0;
71bcdba0 3316 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
e8599ff4 3317 on_link = 1;
71bcdba0 3318 } else if (ipv6_addr_type(&msg->target) !=
e8599ff4 3319 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
6e157b6a 3320 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
e8599ff4
DM		 3321 return;
3322 }
3323
3324 in6_dev = __in6_dev_get(skb->dev);
3325 if (!in6_dev)
3326 return;
3327 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
3328 return;
3329
3330 /* RFC2461 8.1:
3331 * The IP source address of the Redirect MUST be the same as the current
3332 * first-hop router for the specified ICMP Destination Address.
3333 */
3334
f997c55c 3335 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
e8599ff4
DM		 3336 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
3337 return;
3338 }
6e157b6a
DM		 3339
3340 lladdr = NULL;
e8599ff4
DM		 3341 if (ndopts.nd_opts_tgt_lladdr) {
3342 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
3343 skb->dev);
3344 if (!lladdr) {
3345 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
3346 return;
3347 }
3348 }
3349
6e157b6a 3350 rt = (struct rt6_info *) dst;
ec13ad1d 3351 if (rt->rt6i_flags & RTF_REJECT) {
6e157b6a 3352 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
e8599ff4 3353 return;
6e157b6a 3354 }
e8599ff4 3355
6e157b6a
DM		 3356 /* Redirect received -> path was valid.
3357 * Look, redirects are sent only in response to data packets,
3358 * so that this nexthop apparently is reachable. --ANK
3359 */
0dec879f 3360 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
a6279458 3361
71bcdba0 3362 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
6e157b6a
DM		 3363 if (!neigh)
3364 return;
a6279458 3365
1da177e4
LT		 3366 /*
3367 * We have finally decided to accept it.
3368 */
3369
f997c55c 3370 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1da177e4
LT		 3371 NEIGH_UPDATE_F_WEAK_OVERRIDE|
3372 NEIGH_UPDATE_F_OVERRIDE|
3373 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
f997c55c
AA		 3374 NEIGH_UPDATE_F_ISROUTER)),
3375 NDISC_REDIRECT, &ndopts);
1da177e4 3376
4d85cd0c 3377 rcu_read_lock();
a68886a6 3378 from = rcu_dereference(rt->from);
e873e4b9
WW		 3379 /* This fib6_info_hold() is safe here because we hold reference to rt
3380 * and rt already holds reference to fib6_info.
3381 */
8a14e46f 3382 fib6_info_hold(from);
4d85cd0c 3383 rcu_read_unlock();
8a14e46f
DA		 3384
3385 nrt = ip6_rt_cache_alloc(from, &msg->dest, NULL);
38308473 3386 if (!nrt)
1da177e4
LT		 3387 goto out;
3388
3389 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
3390 if (on_link)
3391 nrt->rt6i_flags &= ~RTF_GATEWAY;
3392
4e3fd7a0 3393 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1da177e4 3394
2b760fcf
WW		 3395 /* No need to remove rt from the exception table if rt is
3396 * a cached route because rt6_insert_exception() will
3397 * takes care of it
3398 */
8a14e46f 3399 if (rt6_insert_exception(nrt, from)) {
2b760fcf
WW		 3400 dst_release_immediate(&nrt->dst);
3401 goto out;
3402 }
1da177e4 3403
d8d1f30b
CG		 3404 netevent.old = &rt->dst;
3405 netevent.new = &nrt->dst;
71bcdba0 3406 netevent.daddr = &msg->dest;
60592833 3407 netevent.neigh = neigh;
8d71740c
TT		 3408 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
3409
1da177e4 3410out:
8a14e46f 3411 fib6_info_release(from);
e8599ff4 3412 neigh_release(neigh);
6e157b6a
DM		 3413}
3414
70ceb4f5 3415#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 3416static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 3417 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 3418 const struct in6_addr *gwaddr,
3419 struct net_device *dev)
70ceb4f5 3420{
830218c1
DA		 3421 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
3422 int ifindex = dev->ifindex;
70ceb4f5 3423 struct fib6_node *fn;
8d1c802b 3424 struct fib6_info *rt = NULL;
c71099ac
TG		 3425 struct fib6_table *table;
3426
830218c1 3427 table = fib6_get_table(net, tb_id);
38308473 3428 if (!table)
c71099ac 3429 return NULL;
70ceb4f5 3430
66f5d6ce 3431 rcu_read_lock();
38fbeeee 3432 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
70ceb4f5
YH		 3433 if (!fn)
3434 goto out;
3435
66f5d6ce 3436 for_each_fib6_node_rt_rcu(fn) {
5e670d84 3437 if (rt->fib6_nh.nh_dev->ifindex != ifindex)
70ceb4f5 3438 continue;
93c2fb25 3439 if ((rt->fib6_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
70ceb4f5 3440 continue;
5e670d84 3441 if (!ipv6_addr_equal(&rt->fib6_nh.nh_gw, gwaddr))
70ceb4f5 3442 continue;
e873e4b9
WW		 3443 if (!fib6_info_hold_safe(rt))
3444 continue;
70ceb4f5
YH		 3445 break;
3446 }
3447out:
66f5d6ce 3448 rcu_read_unlock();
70ceb4f5
YH		 3449 return rt;
3450}
3451
8d1c802b 3452static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 3453 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 3454 const struct in6_addr *gwaddr,
3455 struct net_device *dev,
95c96174 3456 unsigned int pref)
70ceb4f5 3457{
86872cb5 3458 struct fib6_config cfg = {
238fc7ea 3459 .fc_metric = IP6_RT_PRIO_USER,
830218c1 3460 .fc_ifindex = dev->ifindex,
86872cb5
TG		 3461 .fc_dst_len = prefixlen,
3462 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
3463 RTF_UP | RTF_PREF(pref),
b91d5329 3464 .fc_protocol = RTPROT_RA,
e8478e80 3465 .fc_type = RTN_UNICAST,
15e47304 3466 .fc_nlinfo.portid = 0,
efa2cea0
DL		 3467 .fc_nlinfo.nlh = NULL,
3468 .fc_nlinfo.nl_net = net,
86872cb5
TG		 3469 };
3470
830218c1 3471 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
4e3fd7a0
AD		 3472 cfg.fc_dst = *prefix;
3473 cfg.fc_gateway = *gwaddr;
70ceb4f5 3474
e317da96
YH		 3475 /* We should treat it as a default route if prefix length is 0. */
3476 if (!prefixlen)
86872cb5 3477 cfg.fc_flags |= RTF_DEFAULT;
70ceb4f5 3478
acb54e3c 3479 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
70ceb4f5 3480
830218c1 3481 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
70ceb4f5
YH		 3482}
3483#endif
3484
8d1c802b 3485struct fib6_info *rt6_get_dflt_router(struct net *net,
afb1d4b5
DA		 3486 const struct in6_addr *addr,
3487 struct net_device *dev)
1ab1457c 3488{
830218c1 3489 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
8d1c802b 3490 struct fib6_info *rt;
c71099ac 3491 struct fib6_table *table;
1da177e4 3492
afb1d4b5 3493 table = fib6_get_table(net, tb_id);
38308473 3494 if (!table)
c71099ac 3495 return NULL;
1da177e4 3496
66f5d6ce
WW		 3497 rcu_read_lock();
3498 for_each_fib6_node_rt_rcu(&table->tb6_root) {
5e670d84 3499 if (dev == rt->fib6_nh.nh_dev &&
93c2fb25 3500 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
5e670d84 3501 ipv6_addr_equal(&rt->fib6_nh.nh_gw, addr))
1da177e4
LT		 3502 break;
3503 }
e873e4b9
WW		 3504 if (rt && !fib6_info_hold_safe(rt))
3505 rt = NULL;
66f5d6ce 3506 rcu_read_unlock();
1da177e4
LT		 3507 return rt;
3508}
3509
8d1c802b 3510struct fib6_info *rt6_add_dflt_router(struct net *net,
afb1d4b5 3511 const struct in6_addr *gwaddr,
ebacaaa0
YH		 3512 struct net_device *dev,
3513 unsigned int pref)
1da177e4 3514{
86872cb5 3515 struct fib6_config cfg = {
ca254490 3516 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
238fc7ea 3517 .fc_metric = IP6_RT_PRIO_USER,
86872cb5
TG		 3518 .fc_ifindex = dev->ifindex,
3519 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
3520 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
b91d5329 3521 .fc_protocol = RTPROT_RA,
e8478e80 3522 .fc_type = RTN_UNICAST,
15e47304 3523 .fc_nlinfo.portid = 0,
5578689a 3524 .fc_nlinfo.nlh = NULL,
afb1d4b5 3525 .fc_nlinfo.nl_net = net,
86872cb5 3526 };
1da177e4 3527
4e3fd7a0 3528 cfg.fc_gateway = *gwaddr;
1da177e4 3529
acb54e3c 3530 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
830218c1
DA		 3531 struct fib6_table *table;
3532
3533 table = fib6_get_table(dev_net(dev), cfg.fc_table);
3534 if (table)
3535 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
3536 }
1da177e4 3537
afb1d4b5 3538 return rt6_get_dflt_router(net, gwaddr, dev);
1da177e4
LT		 3539}
3540
afb1d4b5
DA		 3541static void __rt6_purge_dflt_routers(struct net *net,
3542 struct fib6_table *table)
1da177e4 3543{
8d1c802b 3544 struct fib6_info *rt;
1da177e4
LT		 3545
3546restart:
66f5d6ce
WW		 3547 rcu_read_lock();
3548 for_each_fib6_node_rt_rcu(&table->tb6_root) {
dcd1f572
DA		 3549 struct net_device *dev = fib6_info_nh_dev(rt);
3550 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
3551
93c2fb25 3552 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
e873e4b9
WW		 3553 (!idev || idev->cnf.accept_ra != 2) &&
3554 fib6_info_hold_safe(rt)) {
93531c67
DA		 3555 rcu_read_unlock();
3556 ip6_del_rt(net, rt);
1da177e4
LT		 3557 goto restart;
3558 }
3559 }
66f5d6ce 3560 rcu_read_unlock();
830218c1
DA		 3561
3562 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
3563}
3564
3565void rt6_purge_dflt_routers(struct net *net)
3566{
3567 struct fib6_table *table;
3568 struct hlist_head *head;
3569 unsigned int h;
3570
3571 rcu_read_lock();
3572
3573 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
3574 head = &net->ipv6.fib_table_hash[h];
3575 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
3576 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
afb1d4b5 3577 __rt6_purge_dflt_routers(net, table);
830218c1
DA		 3578 }
3579 }
3580
3581 rcu_read_unlock();
1da177e4
LT		 3582}
3583
5578689a
DL		 3584static void rtmsg_to_fib6_config(struct net *net,
3585 struct in6_rtmsg *rtmsg,
86872cb5
TG		 3586 struct fib6_config *cfg)
3587{
8823a3ac
MŻ		 3588 *cfg = (struct fib6_config){
3589 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
3590 : RT6_TABLE_MAIN,
3591 .fc_ifindex = rtmsg->rtmsg_ifindex,
3592 .fc_metric = rtmsg->rtmsg_metric,
3593 .fc_expires = rtmsg->rtmsg_info,
3594 .fc_dst_len = rtmsg->rtmsg_dst_len,
3595 .fc_src_len = rtmsg->rtmsg_src_len,
3596 .fc_flags = rtmsg->rtmsg_flags,
3597 .fc_type = rtmsg->rtmsg_type,
3598
3599 .fc_nlinfo.nl_net = net,
3600
3601 .fc_dst = rtmsg->rtmsg_dst,
3602 .fc_src = rtmsg->rtmsg_src,
3603 .fc_gateway = rtmsg->rtmsg_gateway,
3604 };
86872cb5
TG		 3605}
3606
5578689a 3607int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1da177e4 3608{
86872cb5 3609 struct fib6_config cfg;
1da177e4
LT		 3610 struct in6_rtmsg rtmsg;
3611 int err;
3612
67ba4152 3613 switch (cmd) {
1da177e4
LT		 3614 case SIOCADDRT: /* Add a route */
3615 case SIOCDELRT: /* Delete a route */
af31f412 3616 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4
LT		 3617 return -EPERM;
3618 err = copy_from_user(&rtmsg, arg,
3619 sizeof(struct in6_rtmsg));
3620 if (err)
3621 return -EFAULT;
86872cb5 3622
5578689a 3623 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
86872cb5 3624
1da177e4
LT		 3625 rtnl_lock();
3626 switch (cmd) {
3627 case SIOCADDRT:
acb54e3c 3628 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
1da177e4
LT		 3629 break;
3630 case SIOCDELRT:
333c4301 3631 err = ip6_route_del(&cfg, NULL);
1da177e4
LT		 3632 break;
3633 default:
3634 err = -EINVAL;
3635 }
3636 rtnl_unlock();
3637
3638 return err;
3ff50b79 3639 }
1da177e4
LT		 3640
3641 return -EINVAL;
3642}
3643
3644/*
3645 * Drop the packet on the floor
3646 */
3647
d5fdd6ba 3648static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1da177e4 3649{
612f09e8 3650 int type;
adf30907 3651 struct dst_entry *dst = skb_dst(skb);
612f09e8
YH		 3652 switch (ipstats_mib_noroutes) {
3653 case IPSTATS_MIB_INNOROUTES:
0660e03f 3654 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
45bb0060 3655 if (type == IPV6_ADDR_ANY) {
bdb7cc64
SS		 3656 IP6_INC_STATS(dev_net(dst->dev),
3657 __in6_dev_get_safely(skb->dev),
3bd653c8 3658 IPSTATS_MIB_INADDRERRORS);
612f09e8
YH		 3659 break;
3660 }
3661 /* FALLTHROUGH */
3662 case IPSTATS_MIB_OUTNOROUTES:
3bd653c8
DL		 3663 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
3664 ipstats_mib_noroutes);
612f09e8
YH		 3665 break;
3666 }
3ffe533c 3667 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1da177e4
LT		 3668 kfree_skb(skb);
3669 return 0;
3670}
3671
9ce8ade0
TG		 3672static int ip6_pkt_discard(struct sk_buff *skb)
3673{
612f09e8 3674 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 3675}
3676
ede2059d 3677static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
1da177e4 3678{
adf30907 3679 skb->dev = skb_dst(skb)->dev;
612f09e8 3680 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1da177e4
LT		 3681}
3682
9ce8ade0
TG		 3683static int ip6_pkt_prohibit(struct sk_buff *skb)
3684{
612f09e8 3685 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 3686}
3687
ede2059d 3688static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
9ce8ade0 3689{
adf30907 3690 skb->dev = skb_dst(skb)->dev;
612f09e8 3691 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
9ce8ade0
TG		 3692}
3693
1da177e4
LT		 3694/*
3695 * Allocate a dst for local (unicast / anycast) address.
3696 */
3697
360a9887
DA		 3698struct fib6_info *addrconf_f6i_alloc(struct net *net,
3699 struct inet6_dev *idev,
3700 const struct in6_addr *addr,
3701 bool anycast, gfp_t gfp_flags)
1da177e4 3702{
ca254490 3703 u32 tb_id;
4832c30d 3704 struct net_device *dev = idev->dev;
360a9887 3705 struct fib6_info *f6i;
5f02ce24 3706
360a9887
DA		 3707 f6i = fib6_info_alloc(gfp_flags);
3708 if (!f6i)
1da177e4
LT		 3709 return ERR_PTR(-ENOMEM);
3710
767a2217 3711 f6i->fib6_metrics = ip_fib_metrics_init(net, NULL, 0);
360a9887 3712 f6i->dst_nocount = true;
360a9887
DA		 3713 f6i->dst_host = true;
3714 f6i->fib6_protocol = RTPROT_KERNEL;
3715 f6i->fib6_flags = RTF_UP | RTF_NONEXTHOP;
e8478e80 3716 if (anycast) {
360a9887
DA		 3717 f6i->fib6_type = RTN_ANYCAST;
3718 f6i->fib6_flags |= RTF_ANYCAST;
e8478e80 3719 } else {
360a9887
DA		 3720 f6i->fib6_type = RTN_LOCAL;
3721 f6i->fib6_flags |= RTF_LOCAL;
e8478e80 3722 }
1da177e4 3723
360a9887 3724 f6i->fib6_nh.nh_gw = *addr;
93531c67 3725 dev_hold(dev);
360a9887
DA		 3726 f6i->fib6_nh.nh_dev = dev;
3727 f6i->fib6_dst.addr = *addr;
3728 f6i->fib6_dst.plen = 128;
ca254490 3729 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
360a9887 3730 f6i->fib6_table = fib6_get_table(net, tb_id);
1da177e4 3731
360a9887 3732 return f6i;
1da177e4
LT		 3733}
3734
c3968a85
DW		 3735/* remove deleted ip from prefsrc entries */
3736struct arg_dev_net_ip {
3737 struct net_device *dev;
3738 struct net *net;
3739 struct in6_addr *addr;
3740};
3741
8d1c802b 3742static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
c3968a85
DW		 3743{
3744 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
3745 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
3746 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
3747
5e670d84 3748 if (((void *)rt->fib6_nh.nh_dev == dev || !dev) &&
421842ed 3749 rt != net->ipv6.fib6_null_entry &&
93c2fb25 3750 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
60006a48 3751 spin_lock_bh(&rt6_exception_lock);
c3968a85 3752 /* remove prefsrc entry */
93c2fb25 3753 rt->fib6_prefsrc.plen = 0;
60006a48 3754 spin_unlock_bh(&rt6_exception_lock);
c3968a85
DW		 3755 }
3756 return 0;
3757}
3758
3759void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
3760{
3761 struct net *net = dev_net(ifp->idev->dev);
3762 struct arg_dev_net_ip adni = {
3763 .dev = ifp->idev->dev,
3764 .net = net,
3765 .addr = &ifp->addr,
3766 };
0c3584d5 3767 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
c3968a85
DW		 3768}
3769
be7a010d 3770#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
be7a010d
DJ		 3771
3772/* Remove routers and update dst entries when gateway turn into host. */
8d1c802b 3773static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
be7a010d
DJ		 3774{
3775 struct in6_addr *gateway = (struct in6_addr *)arg;
3776
93c2fb25 3777 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
5e670d84 3778 ipv6_addr_equal(gateway, &rt->fib6_nh.nh_gw)) {
be7a010d
DJ		 3779 return -1;
3780 }
b16cb459
WW		 3781
3782 /* Further clean up cached routes in exception table.
3783 * This is needed because cached route may have a different
3784 * gateway than its 'parent' in the case of an ip redirect.
3785 */
3786 rt6_exceptions_clean_tohost(rt, gateway);
3787
be7a010d
DJ		 3788 return 0;
3789}
3790
3791void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
3792{
3793 fib6_clean_all(net, fib6_clean_tohost, gateway);
3794}
3795
2127d95a
IS		 3796struct arg_netdev_event {
3797 const struct net_device *dev;
4c981e28
IS		 3798 union {
3799 unsigned int nh_flags;
3800 unsigned long event;
3801 };
2127d95a
IS		 3802};
3803
8d1c802b 3804static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
d7dedee1 3805{
8d1c802b 3806 struct fib6_info *iter;
d7dedee1
IS		 3807 struct fib6_node *fn;
3808
93c2fb25
DA		 3809 fn = rcu_dereference_protected(rt->fib6_node,
3810 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 3811 iter = rcu_dereference_protected(fn->leaf,
93c2fb25 3812 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 3813 while (iter) {
93c2fb25 3814 if (iter->fib6_metric == rt->fib6_metric &&
33bd5ac5 3815 rt6_qualify_for_ecmp(iter))
d7dedee1 3816 return iter;
8fb11a9a 3817 iter = rcu_dereference_protected(iter->fib6_next,
93c2fb25 3818 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1
IS		 3819 }
3820
3821 return NULL;
3822}
3823
8d1c802b 3824static bool rt6_is_dead(const struct fib6_info *rt)
d7dedee1 3825{
5e670d84
DA		 3826 if (rt->fib6_nh.nh_flags & RTNH_F_DEAD ||
3827 (rt->fib6_nh.nh_flags & RTNH_F_LINKDOWN &&
dcd1f572 3828 fib6_ignore_linkdown(rt)))
d7dedee1
IS		 3829 return true;
3830
3831 return false;
3832}
3833
8d1c802b 3834static int rt6_multipath_total_weight(const struct fib6_info *rt)
d7dedee1 3835{
8d1c802b 3836 struct fib6_info *iter;
d7dedee1
IS		 3837 int total = 0;
3838
3839 if (!rt6_is_dead(rt))
5e670d84 3840 total += rt->fib6_nh.nh_weight;
d7dedee1 3841
93c2fb25 3842 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
d7dedee1 3843 if (!rt6_is_dead(iter))
5e670d84 3844 total += iter->fib6_nh.nh_weight;
d7dedee1
IS		 3845 }
3846
3847 return total;
3848}
3849
8d1c802b 3850static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
d7dedee1
IS		 3851{
3852 int upper_bound = -1;
3853
3854 if (!rt6_is_dead(rt)) {
5e670d84 3855 *weight += rt->fib6_nh.nh_weight;
d7dedee1
IS		 3856 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
3857 total) - 1;
3858 }
5e670d84 3859 atomic_set(&rt->fib6_nh.nh_upper_bound, upper_bound);
d7dedee1
IS		 3860}
3861
8d1c802b 3862static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
d7dedee1 3863{
8d1c802b 3864 struct fib6_info *iter;
d7dedee1
IS		 3865 int weight = 0;
3866
3867 rt6_upper_bound_set(rt, &weight, total);
3868
93c2fb25 3869 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
d7dedee1
IS		 3870 rt6_upper_bound_set(iter, &weight, total);
3871}
3872
8d1c802b 3873void rt6_multipath_rebalance(struct fib6_info *rt)
d7dedee1 3874{
8d1c802b 3875 struct fib6_info *first;
d7dedee1
IS		 3876 int total;
3877
3878 /* In case the entire multipath route was marked for flushing,
3879 * then there is no need to rebalance upon the removal of every
3880 * sibling route.
3881 */
93c2fb25 3882 if (!rt->fib6_nsiblings || rt->should_flush)
d7dedee1
IS		 3883 return;
3884
3885 /* During lookup routes are evaluated in order, so we need to
3886 * make sure upper bounds are assigned from the first sibling
3887 * onwards.
3888 */
3889 first = rt6_multipath_first_sibling(rt);
3890 if (WARN_ON_ONCE(!first))
3891 return;
3892
3893 total = rt6_multipath_total_weight(first);
3894 rt6_multipath_upper_bound_set(first, total);
3895}
3896
8d1c802b 3897static int fib6_ifup(struct fib6_info *rt, void *p_arg)
2127d95a
IS		 3898{
3899 const struct arg_netdev_event *arg = p_arg;
7aef6859 3900 struct net *net = dev_net(arg->dev);
2127d95a 3901
421842ed 3902 if (rt != net->ipv6.fib6_null_entry && rt->fib6_nh.nh_dev == arg->dev) {
5e670d84 3903 rt->fib6_nh.nh_flags &= ~arg->nh_flags;
7aef6859 3904 fib6_update_sernum_upto_root(net, rt);
d7dedee1 3905 rt6_multipath_rebalance(rt);
1de178ed 3906 }
2127d95a
IS		 3907
3908 return 0;
3909}
3910
3911void rt6_sync_up(struct net_device *dev, unsigned int nh_flags)
3912{
3913 struct arg_netdev_event arg = {
3914 .dev = dev,
6802f3ad
IS		 3915 {
3916 .nh_flags = nh_flags,
3917 },
2127d95a
IS		 3918 };
3919
3920 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
3921 arg.nh_flags |= RTNH_F_LINKDOWN;
3922
3923 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
3924}
3925
8d1c802b 3926static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
1de178ed
IS		 3927 const struct net_device *dev)
3928{
8d1c802b 3929 struct fib6_info *iter;
1de178ed 3930
5e670d84 3931 if (rt->fib6_nh.nh_dev == dev)
1de178ed 3932 return true;
93c2fb25 3933 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
5e670d84 3934 if (iter->fib6_nh.nh_dev == dev)
1de178ed
IS		 3935 return true;
3936
3937 return false;
3938}
3939
8d1c802b 3940static void rt6_multipath_flush(struct fib6_info *rt)
1de178ed 3941{
8d1c802b 3942 struct fib6_info *iter;
1de178ed
IS		 3943
3944 rt->should_flush = 1;
93c2fb25 3945 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1de178ed
IS		 3946 iter->should_flush = 1;
3947}
3948
8d1c802b 3949static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
1de178ed
IS		 3950 const struct net_device *down_dev)
3951{
8d1c802b 3952 struct fib6_info *iter;
1de178ed
IS		 3953 unsigned int dead = 0;
3954
5e670d84
DA		 3955 if (rt->fib6_nh.nh_dev == down_dev ||
3956 rt->fib6_nh.nh_flags & RTNH_F_DEAD)
1de178ed 3957 dead++;
93c2fb25 3958 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
5e670d84
DA		 3959 if (iter->fib6_nh.nh_dev == down_dev ||
3960 iter->fib6_nh.nh_flags & RTNH_F_DEAD)
1de178ed
IS		 3961 dead++;
3962
3963 return dead;
3964}
3965
8d1c802b 3966static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
1de178ed
IS		 3967 const struct net_device *dev,
3968 unsigned int nh_flags)
3969{
8d1c802b 3970 struct fib6_info *iter;
1de178ed 3971
5e670d84
DA		 3972 if (rt->fib6_nh.nh_dev == dev)
3973 rt->fib6_nh.nh_flags |= nh_flags;
93c2fb25 3974 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
5e670d84
DA		 3975 if (iter->fib6_nh.nh_dev == dev)
3976 iter->fib6_nh.nh_flags |= nh_flags;
1de178ed
IS		 3977}
3978
a1a22c12 3979/* called with write lock held for table with rt */
8d1c802b 3980static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
1da177e4 3981{
4c981e28
IS		 3982 const struct arg_netdev_event *arg = p_arg;
3983 const struct net_device *dev = arg->dev;
7aef6859 3984 struct net *net = dev_net(dev);
8ed67789 3985
421842ed 3986 if (rt == net->ipv6.fib6_null_entry)
27c6fa73
IS		 3987 return 0;
3988
3989 switch (arg->event) {
3990 case NETDEV_UNREGISTER:
5e670d84 3991 return rt->fib6_nh.nh_dev == dev ? -1 : 0;
27c6fa73 3992 case NETDEV_DOWN:
1de178ed 3993 if (rt->should_flush)
27c6fa73 3994 return -1;
93c2fb25 3995 if (!rt->fib6_nsiblings)
5e670d84 3996 return rt->fib6_nh.nh_dev == dev ? -1 : 0;
1de178ed
IS		 3997 if (rt6_multipath_uses_dev(rt, dev)) {
3998 unsigned int count;
3999
4000 count = rt6_multipath_dead_count(rt, dev);
93c2fb25 4001 if (rt->fib6_nsiblings + 1 == count) {
1de178ed
IS		 4002 rt6_multipath_flush(rt);
4003 return -1;
4004 }
4005 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4006 RTNH_F_LINKDOWN);
7aef6859 4007 fib6_update_sernum(net, rt);
d7dedee1 4008 rt6_multipath_rebalance(rt);
1de178ed
IS		 4009 }
4010 return -2;
27c6fa73 4011 case NETDEV_CHANGE:
5e670d84 4012 if (rt->fib6_nh.nh_dev != dev ||
93c2fb25 4013 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
27c6fa73 4014 break;
5e670d84 4015 rt->fib6_nh.nh_flags |= RTNH_F_LINKDOWN;
d7dedee1 4016 rt6_multipath_rebalance(rt);
27c6fa73 4017 break;
2b241361 4018 }
c159d30c 4019
1da177e4
LT		 4020 return 0;
4021}
4022
27c6fa73 4023void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
1da177e4 4024{
4c981e28 4025 struct arg_netdev_event arg = {
8ed67789 4026 .dev = dev,
6802f3ad
IS		 4027 {
4028 .event = event,
4029 },
8ed67789 4030 };
7c6bb7d2 4031 struct net *net = dev_net(dev);
8ed67789 4032
7c6bb7d2
DA		 4033 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4034 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4035 else
4036 fib6_clean_all(net, fib6_ifdown, &arg);
4c981e28
IS		 4037}
4038
4039void rt6_disable_ip(struct net_device *dev, unsigned long event)
4040{
4041 rt6_sync_down_dev(dev, event);
4042 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4043 neigh_ifdown(&nd_tbl, dev);
1da177e4
LT		 4044}
4045
95c96174 4046struct rt6_mtu_change_arg {
1da177e4 4047 struct net_device *dev;
95c96174 4048 unsigned int mtu;
1da177e4
LT		 4049};
4050
8d1c802b 4051static int rt6_mtu_change_route(struct fib6_info *rt, void *p_arg)
1da177e4
LT		 4052{
4053 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4054 struct inet6_dev *idev;
4055
4056 /* In IPv6 pmtu discovery is not optional,
4057 so that RTAX_MTU lock cannot disable it.
4058 We still use this lock to block changes
4059 caused by addrconf/ndisc.
4060 */
4061
4062 idev = __in6_dev_get(arg->dev);
38308473 4063 if (!idev)
1da177e4
LT		 4064 return 0;
4065
4066 /* For administrative MTU increase, there is no way to discover
4067 IPv6 PMTU increase, so PMTU increase should be updated here.
4068 Since RFC 1981 doesn't include administrative MTU increase
4069 update PMTU increase is a MUST. (i.e. jumbo frame)
4070 */
5e670d84 4071 if (rt->fib6_nh.nh_dev == arg->dev &&
d4ead6b3
DA		 4072 !fib6_metric_locked(rt, RTAX_MTU)) {
4073 u32 mtu = rt->fib6_pmtu;
4074
4075 if (mtu >= arg->mtu ||
4076 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4077 fib6_metric_set(rt, RTAX_MTU, arg->mtu);
4078
f5bbe7ee 4079 spin_lock_bh(&rt6_exception_lock);
e9fa1495 4080 rt6_exceptions_update_pmtu(idev, rt, arg->mtu);
f5bbe7ee 4081 spin_unlock_bh(&rt6_exception_lock);
566cfd8f 4082 }
1da177e4
LT		 4083 return 0;
4084}
4085
95c96174 4086void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
1da177e4 4087{
c71099ac
TG		 4088 struct rt6_mtu_change_arg arg = {
4089 .dev = dev,
4090 .mtu = mtu,
4091 };
1da177e4 4092
0c3584d5 4093 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
1da177e4
LT		 4094}
4095
ef7c79ed 4096static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
5176f91e 4097 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
aa8f8778 4098 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
86872cb5 4099 [RTA_OIF] = { .type = NLA_U32 },
ab364a6f 4100 [RTA_IIF] = { .type = NLA_U32 },
86872cb5
TG		 4101 [RTA_PRIORITY] = { .type = NLA_U32 },
4102 [RTA_METRICS] = { .type = NLA_NESTED },
51ebd318 4103 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
c78ba6d6 4104 [RTA_PREF] = { .type = NLA_U8 },
19e42e45
RP		 4105 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4106 [RTA_ENCAP] = { .type = NLA_NESTED },
32bc201e 4107 [RTA_EXPIRES] = { .type = NLA_U32 },
622ec2c9 4108 [RTA_UID] = { .type = NLA_U32 },
3b45a410 4109 [RTA_MARK] = { .type = NLA_U32 },
aa8f8778 4110 [RTA_TABLE] = { .type = NLA_U32 },
eacb9384
RP		 4111 [RTA_IP_PROTO] = { .type = NLA_U8 },
4112 [RTA_SPORT] = { .type = NLA_U16 },
4113 [RTA_DPORT] = { .type = NLA_U16 },
86872cb5
TG		 4114};
4115
4116static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
333c4301
DA		 4117 struct fib6_config *cfg,
4118 struct netlink_ext_ack *extack)
1da177e4 4119{
86872cb5
TG		 4120 struct rtmsg *rtm;
4121 struct nlattr *tb[RTA_MAX+1];
c78ba6d6 4122 unsigned int pref;
86872cb5 4123 int err;
1da177e4 4124
fceb6435 4125 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
dac9c979 4126 extack);
86872cb5
TG		 4127 if (err < 0)
4128 goto errout;
1da177e4 4129
86872cb5
TG		 4130 err = -EINVAL;
4131 rtm = nlmsg_data(nlh);
86872cb5 4132
84db8407
MŻ		 4133 *cfg = (struct fib6_config){
4134 .fc_table = rtm->rtm_table,
4135 .fc_dst_len = rtm->rtm_dst_len,
4136 .fc_src_len = rtm->rtm_src_len,
4137 .fc_flags = RTF_UP,
4138 .fc_protocol = rtm->rtm_protocol,
4139 .fc_type = rtm->rtm_type,
4140
4141 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
4142 .fc_nlinfo.nlh = nlh,
4143 .fc_nlinfo.nl_net = sock_net(skb->sk),
4144 };
86872cb5 4145
ef2c7d7b
ND		 4146 if (rtm->rtm_type == RTN_UNREACHABLE ||
4147 rtm->rtm_type == RTN_BLACKHOLE ||
b4949ab2
ND		 4148 rtm->rtm_type == RTN_PROHIBIT ||
4149 rtm->rtm_type == RTN_THROW)
86872cb5
TG		 4150 cfg->fc_flags |= RTF_REJECT;
4151
ab79ad14
MŻ		 4152 if (rtm->rtm_type == RTN_LOCAL)
4153 cfg->fc_flags |= RTF_LOCAL;
4154
1f56a01f
MKL		 4155 if (rtm->rtm_flags & RTM_F_CLONED)
4156 cfg->fc_flags |= RTF_CACHE;
4157
fc1e64e1
DA		 4158 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4159
86872cb5 4160 if (tb[RTA_GATEWAY]) {
67b61f6c 4161 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
86872cb5 4162 cfg->fc_flags |= RTF_GATEWAY;
1da177e4 4163 }
86872cb5
TG		 4164
4165 if (tb[RTA_DST]) {
4166 int plen = (rtm->rtm_dst_len + 7) >> 3;
4167
4168 if (nla_len(tb[RTA_DST]) < plen)
4169 goto errout;
4170
4171 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1da177e4 4172 }
86872cb5
TG		 4173
4174 if (tb[RTA_SRC]) {
4175 int plen = (rtm->rtm_src_len + 7) >> 3;
4176
4177 if (nla_len(tb[RTA_SRC]) < plen)
4178 goto errout;
4179
4180 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1da177e4 4181 }
86872cb5 4182
c3968a85 4183 if (tb[RTA_PREFSRC])
67b61f6c 4184 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
c3968a85 4185
86872cb5
TG		 4186 if (tb[RTA_OIF])
4187 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4188
4189 if (tb[RTA_PRIORITY])
4190 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4191
4192 if (tb[RTA_METRICS]) {
4193 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4194 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
1da177e4 4195 }
86872cb5
TG		 4196
4197 if (tb[RTA_TABLE])
4198 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4199
51ebd318
ND		 4200 if (tb[RTA_MULTIPATH]) {
4201 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4202 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
9ed59592
DA		 4203
4204 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
c255bd68 4205 cfg->fc_mp_len, extack);
9ed59592
DA		 4206 if (err < 0)
4207 goto errout;
51ebd318
ND		 4208 }
4209
c78ba6d6
LR		 4210 if (tb[RTA_PREF]) {
4211 pref = nla_get_u8(tb[RTA_PREF]);
4212 if (pref != ICMPV6_ROUTER_PREF_LOW &&
4213 pref != ICMPV6_ROUTER_PREF_HIGH)
4214 pref = ICMPV6_ROUTER_PREF_MEDIUM;
4215 cfg->fc_flags |= RTF_PREF(pref);
4216 }
4217
19e42e45
RP		 4218 if (tb[RTA_ENCAP])
4219 cfg->fc_encap = tb[RTA_ENCAP];
4220
9ed59592 4221 if (tb[RTA_ENCAP_TYPE]) {
19e42e45
RP		 4222 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
4223
c255bd68 4224 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
9ed59592
DA		 4225 if (err < 0)
4226 goto errout;
4227 }
4228
32bc201e
XL		 4229 if (tb[RTA_EXPIRES]) {
4230 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
4231
4232 if (addrconf_finite_timeout(timeout)) {
4233 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
4234 cfg->fc_flags |= RTF_EXPIRES;
4235 }
4236 }
4237
86872cb5
TG		 4238 err = 0;
4239errout:
4240 return err;
1da177e4
LT		 4241}
4242
6b9ea5a6 4243struct rt6_nh {
8d1c802b 4244 struct fib6_info *fib6_info;
6b9ea5a6 4245 struct fib6_config r_cfg;
6b9ea5a6
RP		 4246 struct list_head next;
4247};
4248
4249static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
4250{
4251 struct rt6_nh *nh;
4252
4253 list_for_each_entry(nh, rt6_nh_list, next) {
7d4d5065 4254 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6c nexthop %pI6c ifi %d\n",
6b9ea5a6
RP		 4255 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
4256 nh->r_cfg.fc_ifindex);
4257 }
4258}
4259
d4ead6b3
DA		 4260static int ip6_route_info_append(struct net *net,
4261 struct list_head *rt6_nh_list,
8d1c802b
DA		 4262 struct fib6_info *rt,
4263 struct fib6_config *r_cfg)
6b9ea5a6
RP		 4264{
4265 struct rt6_nh *nh;
6b9ea5a6
RP		 4266 int err = -EEXIST;
4267
4268 list_for_each_entry(nh, rt6_nh_list, next) {
8d1c802b
DA		 4269 /* check if fib6_info already exists */
4270 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
6b9ea5a6
RP		 4271 return err;
4272 }
4273
4274 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
4275 if (!nh)
4276 return -ENOMEM;
8d1c802b 4277 nh->fib6_info = rt;
6b9ea5a6
RP		 4278 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
4279 list_add_tail(&nh->next, rt6_nh_list);
4280
4281 return 0;
4282}
4283
8d1c802b
DA		 4284static void ip6_route_mpath_notify(struct fib6_info *rt,
4285 struct fib6_info *rt_last,
3b1137fe
DA		 4286 struct nl_info *info,
4287 __u16 nlflags)
4288{
4289 /* if this is an APPEND route, then rt points to the first route
4290 * inserted and rt_last points to last route inserted. Userspace
4291 * wants a consistent dump of the route which starts at the first
4292 * nexthop. Since sibling routes are always added at the end of
4293 * the list, find the first sibling of the last route appended
4294 */
93c2fb25
DA		 4295 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
4296 rt = list_first_entry(&rt_last->fib6_siblings,
8d1c802b 4297 struct fib6_info,
93c2fb25 4298 fib6_siblings);
3b1137fe
DA		 4299 }
4300
4301 if (rt)
4302 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
4303}
4304
333c4301
DA		 4305static int ip6_route_multipath_add(struct fib6_config *cfg,
4306 struct netlink_ext_ack *extack)
51ebd318 4307{
8d1c802b 4308 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
3b1137fe 4309 struct nl_info *info = &cfg->fc_nlinfo;
51ebd318
ND		 4310 struct fib6_config r_cfg;
4311 struct rtnexthop *rtnh;
8d1c802b 4312 struct fib6_info *rt;
6b9ea5a6
RP		 4313 struct rt6_nh *err_nh;
4314 struct rt6_nh *nh, *nh_safe;
3b1137fe 4315 __u16 nlflags;
51ebd318
ND		 4316 int remaining;
4317 int attrlen;
6b9ea5a6
RP		 4318 int err = 1;
4319 int nhn = 0;
4320 int replace = (cfg->fc_nlinfo.nlh &&
4321 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
4322 LIST_HEAD(rt6_nh_list);
51ebd318 4323
3b1137fe
DA		 4324 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
4325 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
4326 nlflags |= NLM_F_APPEND;
4327
35f1b4e9 4328 remaining = cfg->fc_mp_len;
51ebd318 4329 rtnh = (struct rtnexthop *)cfg->fc_mp;
51ebd318 4330
6b9ea5a6 4331 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
8d1c802b 4332 * fib6_info structs per nexthop
6b9ea5a6 4333 */
51ebd318
ND		 4334 while (rtnh_ok(rtnh, remaining)) {
4335 memcpy(&r_cfg, cfg, sizeof(*cfg));
4336 if (rtnh->rtnh_ifindex)
4337 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
4338
4339 attrlen = rtnh_attrlen(rtnh);
4340 if (attrlen > 0) {
4341 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
4342
4343 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
4344 if (nla) {
67b61f6c 4345 r_cfg.fc_gateway = nla_get_in6_addr(nla);
51ebd318
ND		 4346 r_cfg.fc_flags |= RTF_GATEWAY;
4347 }
19e42e45
RP		 4348 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
4349 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
4350 if (nla)
4351 r_cfg.fc_encap_type = nla_get_u16(nla);
51ebd318 4352 }
6b9ea5a6 4353
68e2ffde 4354 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
acb54e3c 4355 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
8c5b83f0
RP		 4356 if (IS_ERR(rt)) {
4357 err = PTR_ERR(rt);
4358 rt = NULL;
6b9ea5a6 4359 goto cleanup;
8c5b83f0 4360 }
b5d2d75e
DA		 4361 if (!rt6_qualify_for_ecmp(rt)) {
4362 err = -EINVAL;
4363 NL_SET_ERR_MSG(extack,
4364 "Device only routes can not be added for IPv6 using the multipath API.");
4365 fib6_info_release(rt);
4366 goto cleanup;
4367 }
6b9ea5a6 4368
5e670d84 4369 rt->fib6_nh.nh_weight = rtnh->rtnh_hops + 1;
398958ae 4370
d4ead6b3
DA		 4371 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
4372 rt, &r_cfg);
51ebd318 4373 if (err) {
93531c67 4374 fib6_info_release(rt);
6b9ea5a6
RP		 4375 goto cleanup;
4376 }
4377
4378 rtnh = rtnh_next(rtnh, &remaining);
4379 }
4380
3b1137fe
DA		 4381 /* for add and replace send one notification with all nexthops.
4382 * Skip the notification in fib6_add_rt2node and send one with
4383 * the full route when done
4384 */
4385 info->skip_notify = 1;
4386
6b9ea5a6
RP		 4387 err_nh = NULL;
4388 list_for_each_entry(nh, &rt6_nh_list, next) {
8d1c802b
DA		 4389 err = __ip6_ins_rt(nh->fib6_info, info, extack);
4390 fib6_info_release(nh->fib6_info);
93531c67 4391
f7225172
DA		 4392 if (!err) {
4393 /* save reference to last route successfully inserted */
4394 rt_last = nh->fib6_info;
4395
4396 /* save reference to first route for notification */
4397 if (!rt_notif)
4398 rt_notif = nh->fib6_info;
4399 }
3b1137fe 4400
8d1c802b
DA		 4401 /* nh->fib6_info is used or freed at this point, reset to NULL*/
4402 nh->fib6_info = NULL;
6b9ea5a6
RP		 4403 if (err) {
4404 if (replace && nhn)
4405 ip6_print_replace_route_err(&rt6_nh_list);
4406 err_nh = nh;
4407 goto add_errout;
51ebd318 4408 }
6b9ea5a6 4409
1a72418b 4410 /* Because each route is added like a single route we remove
27596472
MK		 4411 * these flags after the first nexthop: if there is a collision,
4412 * we have already failed to add the first nexthop:
4413 * fib6_add_rt2node() has rejected it; when replacing, old
4414 * nexthops have been replaced by first new, the rest should
4415 * be added to it.
1a72418b 4416 */
27596472
MK		 4417 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
4418 NLM_F_REPLACE);
6b9ea5a6
RP		 4419 nhn++;
4420 }
4421
3b1137fe
DA		 4422 /* success ... tell user about new route */
4423 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
6b9ea5a6
RP		 4424 goto cleanup;
4425
4426add_errout:
3b1137fe
DA		 4427 /* send notification for routes that were added so that
4428 * the delete notifications sent by ip6_route_del are
4429 * coherent
4430 */
4431 if (rt_notif)
4432 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
4433
6b9ea5a6
RP		 4434 /* Delete routes that were already added */
4435 list_for_each_entry(nh, &rt6_nh_list, next) {
4436 if (err_nh == nh)
4437 break;
333c4301 4438 ip6_route_del(&nh->r_cfg, extack);
6b9ea5a6
RP		 4439 }
4440
4441cleanup:
4442 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
8d1c802b
DA		 4443 if (nh->fib6_info)
4444 fib6_info_release(nh->fib6_info);
6b9ea5a6
RP		 4445 list_del(&nh->next);
4446 kfree(nh);
4447 }
4448
4449 return err;
4450}
4451
333c4301
DA		 4452static int ip6_route_multipath_del(struct fib6_config *cfg,
4453 struct netlink_ext_ack *extack)
6b9ea5a6
RP		 4454{
4455 struct fib6_config r_cfg;
4456 struct rtnexthop *rtnh;
4457 int remaining;
4458 int attrlen;
4459 int err = 1, last_err = 0;
4460
4461 remaining = cfg->fc_mp_len;
4462 rtnh = (struct rtnexthop *)cfg->fc_mp;
4463
4464 /* Parse a Multipath Entry */
4465 while (rtnh_ok(rtnh, remaining)) {
4466 memcpy(&r_cfg, cfg, sizeof(*cfg));
4467 if (rtnh->rtnh_ifindex)
4468 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
4469
4470 attrlen = rtnh_attrlen(rtnh);
4471 if (attrlen > 0) {
4472 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
4473
4474 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
4475 if (nla) {
4476 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
4477 r_cfg.fc_flags |= RTF_GATEWAY;
4478 }
4479 }
333c4301 4480 err = ip6_route_del(&r_cfg, extack);
6b9ea5a6
RP		 4481 if (err)
4482 last_err = err;
4483
51ebd318
ND		 4484 rtnh = rtnh_next(rtnh, &remaining);
4485 }
4486
4487 return last_err;
4488}
4489
c21ef3e3
DA		 4490static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
4491 struct netlink_ext_ack *extack)
1da177e4 4492{
86872cb5
TG		 4493 struct fib6_config cfg;
4494 int err;
1da177e4 4495
333c4301 4496 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 4497 if (err < 0)
4498 return err;
4499
51ebd318 4500 if (cfg.fc_mp)
333c4301 4501 return ip6_route_multipath_del(&cfg, extack);
0ae81335
DA		 4502 else {
4503 cfg.fc_delete_all_nh = 1;
333c4301 4504 return ip6_route_del(&cfg, extack);
0ae81335 4505 }
1da177e4
LT		 4506}
4507
c21ef3e3
DA		 4508static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
4509 struct netlink_ext_ack *extack)
1da177e4 4510{
86872cb5
TG		 4511 struct fib6_config cfg;
4512 int err;
1da177e4 4513
333c4301 4514 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 4515 if (err < 0)
4516 return err;
4517
51ebd318 4518 if (cfg.fc_mp)
333c4301 4519 return ip6_route_multipath_add(&cfg, extack);
51ebd318 4520 else
acb54e3c 4521 return ip6_route_add(&cfg, GFP_KERNEL, extack);
1da177e4
LT		 4522}
4523
8d1c802b 4524static size_t rt6_nlmsg_size(struct fib6_info *rt)
339bf98f 4525{
beb1afac
DA		 4526 int nexthop_len = 0;
4527
93c2fb25 4528 if (rt->fib6_nsiblings) {
beb1afac
DA		 4529 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
4530 + NLA_ALIGN(sizeof(struct rtnexthop))
4531 + nla_total_size(16) /* RTA_GATEWAY */
5e670d84 4532 + lwtunnel_get_encap_size(rt->fib6_nh.nh_lwtstate);
beb1afac 4533
93c2fb25 4534 nexthop_len *= rt->fib6_nsiblings;
beb1afac
DA		 4535 }
4536
339bf98f
TG		 4537 return NLMSG_ALIGN(sizeof(struct rtmsg))
4538 + nla_total_size(16) /* RTA_SRC */
4539 + nla_total_size(16) /* RTA_DST */
4540 + nla_total_size(16) /* RTA_GATEWAY */
4541 + nla_total_size(16) /* RTA_PREFSRC */
4542 + nla_total_size(4) /* RTA_TABLE */
4543 + nla_total_size(4) /* RTA_IIF */
4544 + nla_total_size(4) /* RTA_OIF */
4545 + nla_total_size(4) /* RTA_PRIORITY */
6a2b9ce0 4546 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
ea697639 4547 + nla_total_size(sizeof(struct rta_cacheinfo))
c78ba6d6 4548 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
19e42e45 4549 + nla_total_size(1) /* RTA_PREF */
5e670d84 4550 + lwtunnel_get_encap_size(rt->fib6_nh.nh_lwtstate)
beb1afac
DA		 4551 + nexthop_len;
4552}
4553
8d1c802b 4554static int rt6_nexthop_info(struct sk_buff *skb, struct fib6_info *rt,
5be083ce 4555 unsigned int *flags, bool skip_oif)
beb1afac 4556{
5e670d84 4557 if (rt->fib6_nh.nh_flags & RTNH_F_DEAD)
f9d882ea
IS		 4558 *flags |= RTNH_F_DEAD;
4559
5e670d84 4560 if (rt->fib6_nh.nh_flags & RTNH_F_LINKDOWN) {
beb1afac 4561 *flags |= RTNH_F_LINKDOWN;
dcd1f572
DA		 4562
4563 rcu_read_lock();
4564 if (fib6_ignore_linkdown(rt))
beb1afac 4565 *flags |= RTNH_F_DEAD;
dcd1f572 4566 rcu_read_unlock();
beb1afac
DA		 4567 }
4568
93c2fb25 4569 if (rt->fib6_flags & RTF_GATEWAY) {
5e670d84 4570 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->fib6_nh.nh_gw) < 0)
beb1afac
DA		 4571 goto nla_put_failure;
4572 }
4573
5e670d84
DA		 4574 *flags |= (rt->fib6_nh.nh_flags & RTNH_F_ONLINK);
4575 if (rt->fib6_nh.nh_flags & RTNH_F_OFFLOAD)
61e4d01e
IS		 4576 *flags |= RTNH_F_OFFLOAD;
4577
5be083ce 4578 /* not needed for multipath encoding b/c it has a rtnexthop struct */
5e670d84
DA		 4579 if (!skip_oif && rt->fib6_nh.nh_dev &&
4580 nla_put_u32(skb, RTA_OIF, rt->fib6_nh.nh_dev->ifindex))
beb1afac
DA		 4581 goto nla_put_failure;
4582
5e670d84
DA		 4583 if (rt->fib6_nh.nh_lwtstate &&
4584 lwtunnel_fill_encap(skb, rt->fib6_nh.nh_lwtstate) < 0)
beb1afac
DA		 4585 goto nla_put_failure;
4586
4587 return 0;
4588
4589nla_put_failure:
4590 return -EMSGSIZE;
4591}
4592
5be083ce 4593/* add multipath next hop */
8d1c802b 4594static int rt6_add_nexthop(struct sk_buff *skb, struct fib6_info *rt)
beb1afac 4595{
5e670d84 4596 const struct net_device *dev = rt->fib6_nh.nh_dev;
beb1afac
DA		 4597 struct rtnexthop *rtnh;
4598 unsigned int flags = 0;
4599
4600 rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
4601 if (!rtnh)
4602 goto nla_put_failure;
4603
5e670d84
DA		 4604 rtnh->rtnh_hops = rt->fib6_nh.nh_weight - 1;
4605 rtnh->rtnh_ifindex = dev ? dev->ifindex : 0;
beb1afac 4606
5be083ce 4607 if (rt6_nexthop_info(skb, rt, &flags, true) < 0)
beb1afac
DA		 4608 goto nla_put_failure;
4609
4610 rtnh->rtnh_flags = flags;
4611
4612 /* length of rtnetlink header + attributes */
4613 rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
4614
4615 return 0;
4616
4617nla_put_failure:
4618 return -EMSGSIZE;
339bf98f
TG		 4619}
4620
d4ead6b3 4621static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 4622 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 4623 struct in6_addr *dest, struct in6_addr *src,
15e47304 4624 int iif, int type, u32 portid, u32 seq,
f8cfe2ce 4625 unsigned int flags)
1da177e4 4626{
22d0bd82
XL		 4627 struct rt6_info *rt6 = (struct rt6_info *)dst;
4628 struct rt6key *rt6_dst, *rt6_src;
4629 u32 *pmetrics, table, rt6_flags;
2d7202bf 4630 struct nlmsghdr *nlh;
22d0bd82 4631 struct rtmsg *rtm;
d4ead6b3 4632 long expires = 0;
1da177e4 4633
15e47304 4634 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
38308473 4635 if (!nlh)
26932566 4636 return -EMSGSIZE;
2d7202bf 4637
22d0bd82
XL		 4638 if (rt6) {
4639 rt6_dst = &rt6->rt6i_dst;
4640 rt6_src = &rt6->rt6i_src;
4641 rt6_flags = rt6->rt6i_flags;
4642 } else {
4643 rt6_dst = &rt->fib6_dst;
4644 rt6_src = &rt->fib6_src;
4645 rt6_flags = rt->fib6_flags;
4646 }
4647
2d7202bf 4648 rtm = nlmsg_data(nlh);
1da177e4 4649 rtm->rtm_family = AF_INET6;
22d0bd82
XL		 4650 rtm->rtm_dst_len = rt6_dst->plen;
4651 rtm->rtm_src_len = rt6_src->plen;
1da177e4 4652 rtm->rtm_tos = 0;
93c2fb25
DA		 4653 if (rt->fib6_table)
4654 table = rt->fib6_table->tb6_id;
c71099ac 4655 else
9e762a4a
PM		 4656 table = RT6_TABLE_UNSPEC;
4657 rtm->rtm_table = table;
c78679e8
DM		 4658 if (nla_put_u32(skb, RTA_TABLE, table))
4659 goto nla_put_failure;
e8478e80
DA		 4660
4661 rtm->rtm_type = rt->fib6_type;
1da177e4
LT		 4662 rtm->rtm_flags = 0;
4663 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
93c2fb25 4664 rtm->rtm_protocol = rt->fib6_protocol;
1da177e4 4665
22d0bd82 4666 if (rt6_flags & RTF_CACHE)
1da177e4
LT		 4667 rtm->rtm_flags |= RTM_F_CLONED;
4668
d4ead6b3
DA		 4669 if (dest) {
4670 if (nla_put_in6_addr(skb, RTA_DST, dest))
c78679e8 4671 goto nla_put_failure;
1ab1457c 4672 rtm->rtm_dst_len = 128;
1da177e4 4673 } else if (rtm->rtm_dst_len)
22d0bd82 4674 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
c78679e8 4675 goto nla_put_failure;
1da177e4
LT		 4676#ifdef CONFIG_IPV6_SUBTREES
4677 if (src) {
930345ea 4678 if (nla_put_in6_addr(skb, RTA_SRC, src))
c78679e8 4679 goto nla_put_failure;
1ab1457c 4680 rtm->rtm_src_len = 128;
c78679e8 4681 } else if (rtm->rtm_src_len &&
22d0bd82 4682 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
c78679e8 4683 goto nla_put_failure;
1da177e4 4684#endif
7bc570c8
YH		 4685 if (iif) {
4686#ifdef CONFIG_IPV6_MROUTE
22d0bd82 4687 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
fd61c6ba
DA		 4688 int err = ip6mr_get_route(net, skb, rtm, portid);
4689
4690 if (err == 0)
4691 return 0;
4692 if (err < 0)
4693 goto nla_put_failure;
7bc570c8
YH		 4694 } else
4695#endif
c78679e8
DM		 4696 if (nla_put_u32(skb, RTA_IIF, iif))
4697 goto nla_put_failure;
d4ead6b3 4698 } else if (dest) {
1da177e4 4699 struct in6_addr saddr_buf;
d4ead6b3 4700 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
930345ea 4701 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 4702 goto nla_put_failure;
1da177e4 4703 }
2d7202bf 4704
93c2fb25 4705 if (rt->fib6_prefsrc.plen) {
c3968a85 4706 struct in6_addr saddr_buf;
93c2fb25 4707 saddr_buf = rt->fib6_prefsrc.addr;
930345ea 4708 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 4709 goto nla_put_failure;
c3968a85
DW		 4710 }
4711
d4ead6b3
DA		 4712 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
4713 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
2d7202bf
TG		 4714 goto nla_put_failure;
4715
93c2fb25 4716 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
c78679e8 4717 goto nla_put_failure;
8253947e 4718
beb1afac
DA		 4719 /* For multipath routes, walk the siblings list and add
4720 * each as a nexthop within RTA_MULTIPATH.
4721 */
22d0bd82
XL		 4722 if (rt6) {
4723 if (rt6_flags & RTF_GATEWAY &&
4724 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
4725 goto nla_put_failure;
4726
4727 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
4728 goto nla_put_failure;
4729 } else if (rt->fib6_nsiblings) {
8d1c802b 4730 struct fib6_info *sibling, *next_sibling;
beb1afac
DA		 4731 struct nlattr *mp;
4732
4733 mp = nla_nest_start(skb, RTA_MULTIPATH);
4734 if (!mp)
4735 goto nla_put_failure;
4736
4737 if (rt6_add_nexthop(skb, rt) < 0)
4738 goto nla_put_failure;
4739
4740 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25 4741 &rt->fib6_siblings, fib6_siblings) {
beb1afac
DA		 4742 if (rt6_add_nexthop(skb, sibling) < 0)
4743 goto nla_put_failure;
4744 }
4745
4746 nla_nest_end(skb, mp);
4747 } else {
5be083ce 4748 if (rt6_nexthop_info(skb, rt, &rtm->rtm_flags, false) < 0)
beb1afac
DA		 4749 goto nla_put_failure;
4750 }
4751
22d0bd82 4752 if (rt6_flags & RTF_EXPIRES) {
14895687
DA		 4753 expires = dst ? dst->expires : rt->expires;
4754 expires -= jiffies;
4755 }
69cdf8f9 4756
d4ead6b3 4757 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
e3703b3d 4758 goto nla_put_failure;
2d7202bf 4759
22d0bd82 4760 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
c78ba6d6
LR		 4761 goto nla_put_failure;
4762
19e42e45 4763
053c095a
JB		 4764 nlmsg_end(skb, nlh);
4765 return 0;
2d7202bf
TG		 4766
4767nla_put_failure:
26932566
PM		 4768 nlmsg_cancel(skb, nlh);
4769 return -EMSGSIZE;
1da177e4
LT		 4770}
4771
13e38901
DA		 4772static bool fib6_info_uses_dev(const struct fib6_info *f6i,
4773 const struct net_device *dev)
4774{
4775 if (f6i->fib6_nh.nh_dev == dev)
4776 return true;
4777
4778 if (f6i->fib6_nsiblings) {
4779 struct fib6_info *sibling, *next_sibling;
4780
4781 list_for_each_entry_safe(sibling, next_sibling,
4782 &f6i->fib6_siblings, fib6_siblings) {
4783 if (sibling->fib6_nh.nh_dev == dev)
4784 return true;
4785 }
4786 }
4787
4788 return false;
4789}
4790
8d1c802b 4791int rt6_dump_route(struct fib6_info *rt, void *p_arg)
1da177e4
LT		 4792{
4793 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
13e38901
DA		 4794 struct fib_dump_filter *filter = &arg->filter;
4795 unsigned int flags = NLM_F_MULTI;
1f17e2f2
DA		 4796 struct net *net = arg->net;
4797
421842ed 4798 if (rt == net->ipv6.fib6_null_entry)
1f17e2f2 4799 return 0;
1da177e4 4800
13e38901
DA		 4801 if ((filter->flags & RTM_F_PREFIX) &&
4802 !(rt->fib6_flags & RTF_PREFIX_RT)) {
4803 /* success since this is not a prefix route */
4804 return 1;
4805 }
4806 if (filter->filter_set) {
4807 if ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
4808 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
4809 (filter->protocol && rt->fib6_protocol != filter->protocol)) {
f8cfe2ce
DA		 4810 return 1;
4811 }
13e38901 4812 flags |= NLM_F_DUMP_FILTERED;
f8cfe2ce 4813 }
1da177e4 4814
d4ead6b3
DA		 4815 return rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL, 0,
4816 RTM_NEWROUTE, NETLINK_CB(arg->cb->skb).portid,
13e38901 4817 arg->cb->nlh->nlmsg_seq, flags);
1da177e4
LT		 4818}
4819
c21ef3e3
DA		 4820static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4821 struct netlink_ext_ack *extack)
1da177e4 4822{
3b1e0a65 4823 struct net *net = sock_net(in_skb->sk);
ab364a6f 4824 struct nlattr *tb[RTA_MAX+1];
18c3a61c 4825 int err, iif = 0, oif = 0;
a68886a6 4826 struct fib6_info *from;
18c3a61c 4827 struct dst_entry *dst;
ab364a6f 4828 struct rt6_info *rt;
1da177e4 4829 struct sk_buff *skb;
ab364a6f 4830 struct rtmsg *rtm;
744486d4 4831 struct flowi6 fl6 = {};
18c3a61c 4832 bool fibmatch;
1da177e4 4833
fceb6435 4834 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
c21ef3e3 4835 extack);
ab364a6f
TG		 4836 if (err < 0)
4837 goto errout;
1da177e4 4838
ab364a6f 4839 err = -EINVAL;
38b7097b
HFS		 4840 rtm = nlmsg_data(nlh);
4841 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
18c3a61c 4842 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
1da177e4 4843
ab364a6f
TG		 4844 if (tb[RTA_SRC]) {
4845 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
4846 goto errout;
4847
4e3fd7a0 4848 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
ab364a6f
TG		 4849 }
4850
4851 if (tb[RTA_DST]) {
4852 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
4853 goto errout;
4854
4e3fd7a0 4855 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
ab364a6f
TG		 4856 }
4857
4858 if (tb[RTA_IIF])
4859 iif = nla_get_u32(tb[RTA_IIF]);
4860
4861 if (tb[RTA_OIF])
72331bc0 4862 oif = nla_get_u32(tb[RTA_OIF]);
1da177e4 4863
2e47b291
LC		 4864 if (tb[RTA_MARK])
4865 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
4866
622ec2c9
LC		 4867 if (tb[RTA_UID])
4868 fl6.flowi6_uid = make_kuid(current_user_ns(),
4869 nla_get_u32(tb[RTA_UID]));
4870 else
4871 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
4872
eacb9384
RP		 4873 if (tb[RTA_SPORT])
4874 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
4875
4876 if (tb[RTA_DPORT])
4877 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
4878
4879 if (tb[RTA_IP_PROTO]) {
4880 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
4881 &fl6.flowi6_proto, extack);
4882 if (err)
4883 goto errout;
4884 }
4885
1da177e4
LT		 4886 if (iif) {
4887 struct net_device *dev;
72331bc0
SL		 4888 int flags = 0;
4889
121622db
FW		 4890 rcu_read_lock();
4891
4892 dev = dev_get_by_index_rcu(net, iif);
1da177e4 4893 if (!dev) {
121622db 4894 rcu_read_unlock();
1da177e4 4895 err = -ENODEV;
ab364a6f 4896 goto errout;
1da177e4 4897 }
72331bc0
SL		 4898
4899 fl6.flowi6_iif = iif;
4900
4901 if (!ipv6_addr_any(&fl6.saddr))
4902 flags |= RT6_LOOKUP_F_HAS_SADDR;
4903
b75cc8f9 4904 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
121622db
FW		 4905
4906 rcu_read_unlock();
72331bc0
SL		 4907 } else {
4908 fl6.flowi6_oif = oif;
4909
58acfd71 4910 dst = ip6_route_output(net, NULL, &fl6);
18c3a61c
RP		 4911 }
4912
18c3a61c
RP		 4913
4914 rt = container_of(dst, struct rt6_info, dst);
4915 if (rt->dst.error) {
4916 err = rt->dst.error;
4917 ip6_rt_put(rt);
4918 goto errout;
1da177e4
LT		 4919 }
4920
9d6acb3b
WC		 4921 if (rt == net->ipv6.ip6_null_entry) {
4922 err = rt->dst.error;
4923 ip6_rt_put(rt);
4924 goto errout;
4925 }
4926
ab364a6f 4927 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
38308473 4928 if (!skb) {
94e187c0 4929 ip6_rt_put(rt);
ab364a6f
TG		 4930 err = -ENOBUFS;
4931 goto errout;
4932 }
1da177e4 4933
d8d1f30b 4934 skb_dst_set(skb, &rt->dst);
a68886a6
DA		 4935
4936 rcu_read_lock();
4937 from = rcu_dereference(rt->from);
4938
18c3a61c 4939 if (fibmatch)
a68886a6 4940 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL, iif,
18c3a61c
RP		 4941 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
4942 nlh->nlmsg_seq, 0);
4943 else
a68886a6
DA		 4944 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
4945 &fl6.saddr, iif, RTM_NEWROUTE,
d4ead6b3
DA		 4946 NETLINK_CB(in_skb).portid, nlh->nlmsg_seq,
4947 0);
a68886a6
DA		 4948 rcu_read_unlock();
4949
1da177e4 4950 if (err < 0) {
ab364a6f
TG		 4951 kfree_skb(skb);
4952 goto errout;
1da177e4
LT		 4953 }
4954
15e47304 4955 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
ab364a6f 4956errout:
1da177e4 4957 return err;
1da177e4
LT		 4958}
4959
8d1c802b 4960void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
37a1d361 4961 unsigned int nlm_flags)
1da177e4
LT		 4962{
4963 struct sk_buff *skb;
5578689a 4964 struct net *net = info->nl_net;
528c4ceb
DL		 4965 u32 seq;
4966 int err;
4967
4968 err = -ENOBUFS;
38308473 4969 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
86872cb5 4970
19e42e45 4971 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
38308473 4972 if (!skb)
21713ebc
TG		 4973 goto errout;
4974
d4ead6b3
DA		 4975 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
4976 event, info->portid, seq, nlm_flags);
26932566
PM		 4977 if (err < 0) {
4978 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
4979 WARN_ON(err == -EMSGSIZE);
4980 kfree_skb(skb);
4981 goto errout;
4982 }
15e47304 4983 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
1ce85fe4
PNA		 4984 info->nlh, gfp_any());
4985 return;
21713ebc
TG		 4986errout:
4987 if (err < 0)
5578689a 4988 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
1da177e4
LT		 4989}
4990
8ed67789 4991static int ip6_route_dev_notify(struct notifier_block *this,
351638e7 4992 unsigned long event, void *ptr)
8ed67789 4993{
351638e7 4994 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
c346dca1 4995 struct net *net = dev_net(dev);
8ed67789 4996
242d3a49
WC		 4997 if (!(dev->flags & IFF_LOOPBACK))
4998 return NOTIFY_OK;
4999
5000 if (event == NETDEV_REGISTER) {
421842ed 5001 net->ipv6.fib6_null_entry->fib6_nh.nh_dev = dev;
d8d1f30b 5002 net->ipv6.ip6_null_entry->dst.dev = dev;
8ed67789
DL		 5003 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
5004#ifdef CONFIG_IPV6_MULTIPLE_TABLES
d8d1f30b 5005 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
8ed67789 5006 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
d8d1f30b 5007 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
8ed67789 5008 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
242d3a49 5009#endif
76da0704
WC		 5010 } else if (event == NETDEV_UNREGISTER &&
5011 dev->reg_state != NETREG_UNREGISTERED) {
5012 /* NETDEV_UNREGISTER could be fired for multiple times by
5013 * netdev_wait_allrefs(). Make sure we only call this once.
5014 */
12d94a80 5015 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
242d3a49 5016#ifdef CONFIG_IPV6_MULTIPLE_TABLES
12d94a80
ED		 5017 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
5018 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
8ed67789
DL		 5019#endif
5020 }
5021
5022 return NOTIFY_OK;
5023}
5024
1da177e4
LT		 5025/*
5026 * /proc
5027 */
5028
5029#ifdef CONFIG_PROC_FS
1da177e4
LT		 5030static int rt6_stats_seq_show(struct seq_file *seq, void *v)
5031{
69ddb805 5032 struct net *net = (struct net *)seq->private;
1da177e4 5033 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
69ddb805
DL		 5034 net->ipv6.rt6_stats->fib_nodes,
5035 net->ipv6.rt6_stats->fib_route_nodes,
81eb8447 5036 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
69ddb805
DL		 5037 net->ipv6.rt6_stats->fib_rt_entries,
5038 net->ipv6.rt6_stats->fib_rt_cache,
fc66f95c 5039 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
69ddb805 5040 net->ipv6.rt6_stats->fib_discarded_routes);
1da177e4
LT		 5041
5042 return 0;
5043}
1da177e4
LT		 5044#endif /* CONFIG_PROC_FS */
5045
5046#ifdef CONFIG_SYSCTL
5047
1da177e4 5048static
fe2c6338 5049int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
1da177e4
LT		 5050 void __user *buffer, size_t *lenp, loff_t *ppos)
5051{
c486da34
LAG		 5052 struct net *net;
5053 int delay;
5054 if (!write)
1da177e4 5055 return -EINVAL;
c486da34
LAG		 5056
5057 net = (struct net *)ctl->extra1;
5058 delay = net->ipv6.sysctl.flush_delay;
5059 proc_dointvec(ctl, write, buffer, lenp, ppos);
2ac3ac8f 5060 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
c486da34 5061 return 0;
1da177e4
LT		 5062}
5063
7c6bb7d2
DA		 5064static int zero;
5065static int one = 1;
5066
ed792e28 5067static struct ctl_table ipv6_route_table_template[] = {
1ab1457c 5068 {
1da177e4 5069 .procname = "flush",
4990509f 5070 .data = &init_net.ipv6.sysctl.flush_delay,
1da177e4 5071 .maxlen = sizeof(int),
89c8b3a1 5072 .mode = 0200,
6d9f239a 5073 .proc_handler = ipv6_sysctl_rtcache_flush
1da177e4
LT		 5074 },
5075 {
1da177e4 5076 .procname = "gc_thresh",
9a7ec3a9 5077 .data = &ip6_dst_ops_template.gc_thresh,
1da177e4
LT		 5078 .maxlen = sizeof(int),
5079 .mode = 0644,
6d9f239a 5080 .proc_handler = proc_dointvec,
1da177e4
LT		 5081 },
5082 {
1da177e4 5083 .procname = "max_size",
4990509f 5084 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
1da177e4
LT		 5085 .maxlen = sizeof(int),
5086 .mode = 0644,
6d9f239a 5087 .proc_handler = proc_dointvec,
1da177e4
LT		 5088 },
5089 {
1da177e4 5090 .procname = "gc_min_interval",
4990509f 5091 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 5092 .maxlen = sizeof(int),
5093 .mode = 0644,
6d9f239a 5094 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5095 },
5096 {
1da177e4 5097 .procname = "gc_timeout",
4990509f 5098 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
1da177e4
LT		 5099 .maxlen = sizeof(int),
5100 .mode = 0644,
6d9f239a 5101 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5102 },
5103 {
1da177e4 5104 .procname = "gc_interval",
4990509f 5105 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
1da177e4
LT		 5106 .maxlen = sizeof(int),
5107 .mode = 0644,
6d9f239a 5108 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5109 },
5110 {
1da177e4 5111 .procname = "gc_elasticity",
4990509f 5112 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
1da177e4
LT		 5113 .maxlen = sizeof(int),
5114 .mode = 0644,
f3d3f616 5115 .proc_handler = proc_dointvec,
1da177e4
LT		 5116 },
5117 {
1da177e4 5118 .procname = "mtu_expires",
4990509f 5119 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
1da177e4
LT		 5120 .maxlen = sizeof(int),
5121 .mode = 0644,
6d9f239a 5122 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5123 },
5124 {
1da177e4 5125 .procname = "min_adv_mss",
4990509f 5126 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
1da177e4
LT		 5127 .maxlen = sizeof(int),
5128 .mode = 0644,
f3d3f616 5129 .proc_handler = proc_dointvec,
1da177e4
LT		 5130 },
5131 {
1da177e4 5132 .procname = "gc_min_interval_ms",
4990509f 5133 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 5134 .maxlen = sizeof(int),
5135 .mode = 0644,
6d9f239a 5136 .proc_handler = proc_dointvec_ms_jiffies,
1da177e4 5137 },
7c6bb7d2
DA		 5138 {
5139 .procname = "skip_notify_on_dev_down",
5140 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
5141 .maxlen = sizeof(int),
5142 .mode = 0644,
5143 .proc_handler = proc_dointvec,
5144 .extra1 = &zero,
5145 .extra2 = &one,
5146 },
f8572d8f 5147 { }
1da177e4
LT		 5148};
5149
2c8c1e72 5150struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
760f2d01
DL		 5151{
5152 struct ctl_table *table;
5153
5154 table = kmemdup(ipv6_route_table_template,
5155 sizeof(ipv6_route_table_template),
5156 GFP_KERNEL);
5ee09105
YH		 5157
5158 if (table) {
5159 table[0].data = &net->ipv6.sysctl.flush_delay;
c486da34 5160 table[0].extra1 = net;
86393e52 5161 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
5ee09105
YH		 5162 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
5163 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
5164 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
5165 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
5166 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
5167 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
5168 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
9c69fabe 5169 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
7c6bb7d2 5170 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
464dc801
EB		 5171
5172 /* Don't export sysctls to unprivileged users */
5173 if (net->user_ns != &init_user_ns)
5174 table[0].procname = NULL;
5ee09105
YH		 5175 }
5176
760f2d01
DL		 5177 return table;
5178}
1da177e4
LT		 5179#endif
5180
2c8c1e72 5181static int __net_init ip6_route_net_init(struct net *net)
cdb18761 5182{
633d424b 5183 int ret = -ENOMEM;
8ed67789 5184
86393e52
AD		 5185 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
5186 sizeof(net->ipv6.ip6_dst_ops));
f2fc6a54 5187
fc66f95c
ED		 5188 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
5189 goto out_ip6_dst_ops;
5190
421842ed
DA		 5191 net->ipv6.fib6_null_entry = kmemdup(&fib6_null_entry_template,
5192 sizeof(*net->ipv6.fib6_null_entry),
5193 GFP_KERNEL);
5194 if (!net->ipv6.fib6_null_entry)
5195 goto out_ip6_dst_entries;
5196
8ed67789
DL		 5197 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
5198 sizeof(*net->ipv6.ip6_null_entry),
5199 GFP_KERNEL);
5200 if (!net->ipv6.ip6_null_entry)
421842ed 5201 goto out_fib6_null_entry;
d8d1f30b 5202 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 5203 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
5204 ip6_template_metrics, true);
8ed67789
DL		 5205
5206#ifdef CONFIG_IPV6_MULTIPLE_TABLES
feca7d8c 5207 net->ipv6.fib6_has_custom_rules = false;
8ed67789
DL		 5208 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
5209 sizeof(*net->ipv6.ip6_prohibit_entry),
5210 GFP_KERNEL);
68fffc67
PZ		 5211 if (!net->ipv6.ip6_prohibit_entry)
5212 goto out_ip6_null_entry;
d8d1f30b 5213 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 5214 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
5215 ip6_template_metrics, true);
8ed67789
DL		 5216
5217 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
5218 sizeof(*net->ipv6.ip6_blk_hole_entry),
5219 GFP_KERNEL);
68fffc67
PZ		 5220 if (!net->ipv6.ip6_blk_hole_entry)
5221 goto out_ip6_prohibit_entry;
d8d1f30b 5222 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 5223 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
5224 ip6_template_metrics, true);
8ed67789
DL		 5225#endif
5226
b339a47c
PZ		 5227 net->ipv6.sysctl.flush_delay = 0;
5228 net->ipv6.sysctl.ip6_rt_max_size = 4096;
5229 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
5230 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
5231 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
5232 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
5233 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
5234 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
7c6bb7d2 5235 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
b339a47c 5236
6891a346
BT		 5237 net->ipv6.ip6_rt_gc_expire = 30*HZ;
5238
8ed67789
DL		 5239 ret = 0;
5240out:
5241 return ret;
f2fc6a54 5242
68fffc67
PZ		 5243#ifdef CONFIG_IPV6_MULTIPLE_TABLES
5244out_ip6_prohibit_entry:
5245 kfree(net->ipv6.ip6_prohibit_entry);
5246out_ip6_null_entry:
5247 kfree(net->ipv6.ip6_null_entry);
5248#endif
421842ed
DA		 5249out_fib6_null_entry:
5250 kfree(net->ipv6.fib6_null_entry);
fc66f95c
ED		 5251out_ip6_dst_entries:
5252 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
f2fc6a54 5253out_ip6_dst_ops:
f2fc6a54 5254 goto out;
cdb18761
DL		 5255}
5256
2c8c1e72 5257static void __net_exit ip6_route_net_exit(struct net *net)
cdb18761 5258{
421842ed 5259 kfree(net->ipv6.fib6_null_entry);
8ed67789
DL		 5260 kfree(net->ipv6.ip6_null_entry);
5261#ifdef CONFIG_IPV6_MULTIPLE_TABLES
5262 kfree(net->ipv6.ip6_prohibit_entry);
5263 kfree(net->ipv6.ip6_blk_hole_entry);
5264#endif
41bb78b4 5265 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
cdb18761
DL		 5266}
5267
d189634e
TG		 5268static int __net_init ip6_route_net_init_late(struct net *net)
5269{
5270#ifdef CONFIG_PROC_FS
c3506372
CH		 5271 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
5272 sizeof(struct ipv6_route_iter));
3617d949
CH		 5273 proc_create_net_single("rt6_stats", 0444, net->proc_net,
5274 rt6_stats_seq_show, NULL);
d189634e
TG		 5275#endif
5276 return 0;
5277}
5278
5279static void __net_exit ip6_route_net_exit_late(struct net *net)
5280{
5281#ifdef CONFIG_PROC_FS
ece31ffd
G		 5282 remove_proc_entry("ipv6_route", net->proc_net);
5283 remove_proc_entry("rt6_stats", net->proc_net);
d189634e
TG		 5284#endif
5285}
5286
cdb18761
DL		 5287static struct pernet_operations ip6_route_net_ops = {
5288 .init = ip6_route_net_init,
5289 .exit = ip6_route_net_exit,
5290};
5291
c3426b47
DM		 5292static int __net_init ipv6_inetpeer_init(struct net *net)
5293{
5294 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
5295
5296 if (!bp)
5297 return -ENOMEM;
5298 inet_peer_base_init(bp);
5299 net->ipv6.peers = bp;
5300 return 0;
5301}
5302
5303static void __net_exit ipv6_inetpeer_exit(struct net *net)
5304{
5305 struct inet_peer_base *bp = net->ipv6.peers;
5306
5307 net->ipv6.peers = NULL;
56a6b248 5308 inetpeer_invalidate_tree(bp);
c3426b47
DM		 5309 kfree(bp);
5310}
5311
2b823f72 5312static struct pernet_operations ipv6_inetpeer_ops = {
c3426b47
DM		 5313 .init = ipv6_inetpeer_init,
5314 .exit = ipv6_inetpeer_exit,
5315};
5316
d189634e
TG		 5317static struct pernet_operations ip6_route_net_late_ops = {
5318 .init = ip6_route_net_init_late,
5319 .exit = ip6_route_net_exit_late,
5320};
5321
8ed67789
DL		 5322static struct notifier_block ip6_route_dev_notifier = {
5323 .notifier_call = ip6_route_dev_notify,
242d3a49 5324 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
8ed67789
DL		 5325};
5326
2f460933
WC		 5327void __init ip6_route_init_special_entries(void)
5328{
5329 /* Registering of the loopback is done before this portion of code,
5330 * the loopback reference in rt6_info will not be taken, do it
5331 * manually for init_net */
421842ed 5332 init_net.ipv6.fib6_null_entry->fib6_nh.nh_dev = init_net.loopback_dev;
2f460933
WC		 5333 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
5334 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
5335 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
5336 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
5337 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
5338 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
5339 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
5340 #endif
5341}
5342
433d49c3 5343int __init ip6_route_init(void)
1da177e4 5344{
433d49c3 5345 int ret;
8d0b94af 5346 int cpu;
433d49c3 5347
9a7ec3a9
DL		 5348 ret = -ENOMEM;
5349 ip6_dst_ops_template.kmem_cachep =
e5d679f3 5350 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
f845ab6b 5351 SLAB_HWCACHE_ALIGN, NULL);
9a7ec3a9 5352 if (!ip6_dst_ops_template.kmem_cachep)
c19a28e1 5353 goto out;
14e50e57 5354
fc66f95c 5355 ret = dst_entries_init(&ip6_dst_blackhole_ops);
8ed67789 5356 if (ret)
bdb3289f 5357 goto out_kmem_cache;
bdb3289f 5358
c3426b47
DM		 5359 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
5360 if (ret)
e8803b6c 5361 goto out_dst_entries;
2a0c451a 5362
7e52b33b
DM		 5363 ret = register_pernet_subsys(&ip6_route_net_ops);
5364 if (ret)
5365 goto out_register_inetpeer;
c3426b47 5366
5dc121e9
AE		 5367 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
5368
e8803b6c 5369 ret = fib6_init();
433d49c3 5370 if (ret)
8ed67789 5371 goto out_register_subsys;
433d49c3 5372
433d49c3
DL		 5373 ret = xfrm6_init();
5374 if (ret)
e8803b6c 5375 goto out_fib6_init;
c35b7e72 5376
433d49c3
DL		 5377 ret = fib6_rules_init();
5378 if (ret)
5379 goto xfrm6_init;
7e5449c2 5380
d189634e
TG		 5381 ret = register_pernet_subsys(&ip6_route_net_late_ops);
5382 if (ret)
5383 goto fib6_rules_init;
5384
16feebcf
FW		 5385 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
5386 inet6_rtm_newroute, NULL, 0);
5387 if (ret < 0)
5388 goto out_register_late_subsys;
5389
5390 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
5391 inet6_rtm_delroute, NULL, 0);
5392 if (ret < 0)
5393 goto out_register_late_subsys;
5394
5395 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
5396 inet6_rtm_getroute, NULL,
5397 RTNL_FLAG_DOIT_UNLOCKED);
5398 if (ret < 0)
d189634e 5399 goto out_register_late_subsys;
c127ea2c 5400
8ed67789 5401 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
cdb18761 5402 if (ret)
d189634e 5403 goto out_register_late_subsys;
8ed67789 5404
8d0b94af
MKL		 5405 for_each_possible_cpu(cpu) {
5406 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
5407
5408 INIT_LIST_HEAD(&ul->head);
5409 spin_lock_init(&ul->lock);
5410 }
5411
433d49c3
DL		 5412out:
5413 return ret;
5414
d189634e 5415out_register_late_subsys:
16feebcf 5416 rtnl_unregister_all(PF_INET6);
d189634e 5417 unregister_pernet_subsys(&ip6_route_net_late_ops);
433d49c3 5418fib6_rules_init:
433d49c3
DL		 5419 fib6_rules_cleanup();
5420xfrm6_init:
433d49c3 5421 xfrm6_fini();
2a0c451a
TG		 5422out_fib6_init:
5423 fib6_gc_cleanup();
8ed67789
DL		 5424out_register_subsys:
5425 unregister_pernet_subsys(&ip6_route_net_ops);
7e52b33b
DM		 5426out_register_inetpeer:
5427 unregister_pernet_subsys(&ipv6_inetpeer_ops);
fc66f95c
ED		 5428out_dst_entries:
5429 dst_entries_destroy(&ip6_dst_blackhole_ops);
433d49c3 5430out_kmem_cache:
f2fc6a54 5431 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
433d49c3 5432 goto out;
1da177e4
LT		 5433}
5434
5435void ip6_route_cleanup(void)
5436{
8ed67789 5437 unregister_netdevice_notifier(&ip6_route_dev_notifier);
d189634e 5438 unregister_pernet_subsys(&ip6_route_net_late_ops);
101367c2 5439 fib6_rules_cleanup();
1da177e4 5440 xfrm6_fini();
1da177e4 5441 fib6_gc_cleanup();
c3426b47 5442 unregister_pernet_subsys(&ipv6_inetpeer_ops);
8ed67789 5443 unregister_pernet_subsys(&ip6_route_net_ops);
41bb78b4 5444 dst_entries_destroy(&ip6_dst_blackhole_ops);
f2fc6a54 5445 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
1da177e4 5446}
Linux 6.x block layer and io_uring tree(s)