projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net: fddi: skfp: Remove unused macros 'PNMI_GET_ID' and 'PNMI_SET_ID'
[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{
990edb42 520 struct __rt6_probe_work *work;
5e670d84 521 const struct in6_addr *nh_gw;
f2c31e32 522 struct neighbour *neigh;
5e670d84
DA		 523 struct net_device *dev;
524
27097255
YH		 525 /*
526 * Okay, this does not seem to be appropriate
527 * for now, however, we need to check if it
528 * is really so; aka Router Reachability Probing.
529 *
530 * Router Reachability Probe MUST be rate-limited
531 * to no more than one per minute.
532 */
93c2fb25 533 if (!rt || !(rt->fib6_flags & RTF_GATEWAY))
7ff74a59 534 return;
5e670d84
DA		 535
536 nh_gw = &rt->fib6_nh.nh_gw;
537 dev = rt->fib6_nh.nh_dev;
2152caea 538 rcu_read_lock_bh();
5e670d84 539 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
2152caea 540 if (neigh) {
dcd1f572
DA		 541 struct inet6_dev *idev;
542
8d6c31bf
MKL		 543 if (neigh->nud_state & NUD_VALID)
544 goto out;
545
dcd1f572 546 idev = __in6_dev_get(dev);
990edb42 547 work = NULL;
2152caea 548 write_lock(&neigh->lock);
990edb42
MKL		 549 if (!(neigh->nud_state & NUD_VALID) &&
550 time_after(jiffies,
dcd1f572 551 neigh->updated + idev->cnf.rtr_probe_interval)) {
990edb42
MKL		 552 work = kmalloc(sizeof(*work), GFP_ATOMIC);
553 if (work)
554 __neigh_set_probe_once(neigh);
c2f17e82 555 }
2152caea 556 write_unlock(&neigh->lock);
990edb42
MKL		 557 } else {
558 work = kmalloc(sizeof(*work), GFP_ATOMIC);
f2c31e32 559 }
990edb42
MKL		 560
561 if (work) {
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
DA		 2747 if (!grt->dst.error &&
2748 (grt->rt6i_flags & flags || dev != grt->dst.dev)) {
44750f84
DA		 2749 NL_SET_ERR_MSG(extack,
2750 "Nexthop has invalid gateway or device mismatch");
fc1e64e1
DA		 2751 err = -EINVAL;
2752 }
2753
2754 ip6_rt_put(grt);
2755 }
2756
2757 return err;
2758}
2759
1edce99f
DA		 2760static int ip6_route_check_nh(struct net *net,
2761 struct fib6_config *cfg,
2762 struct net_device **_dev,
2763 struct inet6_dev **idev)
2764{
2765 const struct in6_addr *gw_addr = &cfg->fc_gateway;
2766 struct net_device *dev = _dev ? *_dev : NULL;
2767 struct rt6_info *grt = NULL;
2768 int err = -EHOSTUNREACH;
2769
2770 if (cfg->fc_table) {
f4797b33
DA		 2771 int flags = RT6_LOOKUP_F_IFACE;
2772
2773 grt = ip6_nh_lookup_table(net, cfg, gw_addr,
2774 cfg->fc_table, flags);
1edce99f
DA		 2775 if (grt) {
2776 if (grt->rt6i_flags & RTF_GATEWAY ||
2777 (dev && dev != grt->dst.dev)) {
2778 ip6_rt_put(grt);
2779 grt = NULL;
2780 }
2781 }
2782 }
2783
2784 if (!grt)
b75cc8f9 2785 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, NULL, 1);
1edce99f
DA		 2786
2787 if (!grt)
2788 goto out;
2789
2790 if (dev) {
2791 if (dev != grt->dst.dev) {
2792 ip6_rt_put(grt);
2793 goto out;
2794 }
2795 } else {
2796 *_dev = dev = grt->dst.dev;
2797 *idev = grt->rt6i_idev;
2798 dev_hold(dev);
2799 in6_dev_hold(grt->rt6i_idev);
2800 }
2801
2802 if (!(grt->rt6i_flags & RTF_GATEWAY))
2803 err = 0;
2804
2805 ip6_rt_put(grt);
2806
2807out:
2808 return err;
2809}
2810
9fbb704c
DA		 2811static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
2812 struct net_device **_dev, struct inet6_dev **idev,
2813 struct netlink_ext_ack *extack)
2814{
2815 const struct in6_addr *gw_addr = &cfg->fc_gateway;
2816 int gwa_type = ipv6_addr_type(gw_addr);
232378e8 2817 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
9fbb704c 2818 const struct net_device *dev = *_dev;
232378e8 2819 bool need_addr_check = !dev;
9fbb704c
DA		 2820 int err = -EINVAL;
2821
2822 /* if gw_addr is local we will fail to detect this in case
2823 * address is still TENTATIVE (DAD in progress). rt6_lookup()
2824 * will return already-added prefix route via interface that
2825 * prefix route was assigned to, which might be non-loopback.
2826 */
232378e8
DA		 2827 if (dev &&
2828 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
2829 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
9fbb704c
DA		 2830 goto out;
2831 }
2832
2833 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
2834 /* IPv6 strictly inhibits using not link-local
2835 * addresses as nexthop address.
2836 * Otherwise, router will not able to send redirects.
2837 * It is very good, but in some (rare!) circumstances
2838 * (SIT, PtP, NBMA NOARP links) it is handy to allow
2839 * some exceptions. --ANK
2840 * We allow IPv4-mapped nexthops to support RFC4798-type
2841 * addressing
2842 */
2843 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
2844 NL_SET_ERR_MSG(extack, "Invalid gateway address");
2845 goto out;
2846 }
2847
2848 if (cfg->fc_flags & RTNH_F_ONLINK)
2849 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
2850 else
2851 err = ip6_route_check_nh(net, cfg, _dev, idev);
2852
2853 if (err)
2854 goto out;
2855 }
2856
2857 /* reload in case device was changed */
2858 dev = *_dev;
2859
2860 err = -EINVAL;
2861 if (!dev) {
2862 NL_SET_ERR_MSG(extack, "Egress device not specified");
2863 goto out;
2864 } else if (dev->flags & IFF_LOOPBACK) {
2865 NL_SET_ERR_MSG(extack,
2866 "Egress device can not be loopback device for this route");
2867 goto out;
2868 }
232378e8
DA		 2869
2870 /* if we did not check gw_addr above, do so now that the
2871 * egress device has been resolved.
2872 */
2873 if (need_addr_check &&
2874 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
2875 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
2876 goto out;
2877 }
2878
9fbb704c
DA		 2879 err = 0;
2880out:
2881 return err;
2882}
2883
8d1c802b 2884static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
acb54e3c 2885 gfp_t gfp_flags,
333c4301 2886 struct netlink_ext_ack *extack)
1da177e4 2887{
5578689a 2888 struct net *net = cfg->fc_nlinfo.nl_net;
8d1c802b 2889 struct fib6_info *rt = NULL;
1da177e4
LT		 2890 struct net_device *dev = NULL;
2891 struct inet6_dev *idev = NULL;
c71099ac 2892 struct fib6_table *table;
1da177e4 2893 int addr_type;
8c5b83f0 2894 int err = -EINVAL;
1da177e4 2895
557c44be 2896 /* RTF_PCPU is an internal flag; can not be set by userspace */
d5d531cb
DA		 2897 if (cfg->fc_flags & RTF_PCPU) {
2898 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
557c44be 2899 goto out;
d5d531cb 2900 }
557c44be 2901
2ea2352e
WW		 2902 /* RTF_CACHE is an internal flag; can not be set by userspace */
2903 if (cfg->fc_flags & RTF_CACHE) {
2904 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
2905 goto out;
2906 }
2907
e8478e80
DA		 2908 if (cfg->fc_type > RTN_MAX) {
2909 NL_SET_ERR_MSG(extack, "Invalid route type");
2910 goto out;
2911 }
2912
d5d531cb
DA		 2913 if (cfg->fc_dst_len > 128) {
2914 NL_SET_ERR_MSG(extack, "Invalid prefix length");
2915 goto out;
2916 }
2917 if (cfg->fc_src_len > 128) {
2918 NL_SET_ERR_MSG(extack, "Invalid source address length");
8c5b83f0 2919 goto out;
d5d531cb 2920 }
1da177e4 2921#ifndef CONFIG_IPV6_SUBTREES
d5d531cb
DA		 2922 if (cfg->fc_src_len) {
2923 NL_SET_ERR_MSG(extack,
2924 "Specifying source address requires IPV6_SUBTREES to be enabled");
8c5b83f0 2925 goto out;
d5d531cb 2926 }
1da177e4 2927#endif
86872cb5 2928 if (cfg->fc_ifindex) {
1da177e4 2929 err = -ENODEV;
5578689a 2930 dev = dev_get_by_index(net, cfg->fc_ifindex);
1da177e4
LT		 2931 if (!dev)
2932 goto out;
2933 idev = in6_dev_get(dev);
2934 if (!idev)
2935 goto out;
2936 }
2937
86872cb5
TG		 2938 if (cfg->fc_metric == 0)
2939 cfg->fc_metric = IP6_RT_PRIO_USER;
1da177e4 2940
fc1e64e1
DA		 2941 if (cfg->fc_flags & RTNH_F_ONLINK) {
2942 if (!dev) {
2943 NL_SET_ERR_MSG(extack,
2944 "Nexthop device required for onlink");
2945 err = -ENODEV;
2946 goto out;
2947 }
2948
2949 if (!(dev->flags & IFF_UP)) {
2950 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
2951 err = -ENETDOWN;
2952 goto out;
2953 }
2954 }
2955
d71314b4 2956 err = -ENOBUFS;
38308473
DM		 2957 if (cfg->fc_nlinfo.nlh &&
2958 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
d71314b4 2959 table = fib6_get_table(net, cfg->fc_table);
38308473 2960 if (!table) {
f3213831 2961 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
d71314b4
MV		 2962 table = fib6_new_table(net, cfg->fc_table);
2963 }
2964 } else {
2965 table = fib6_new_table(net, cfg->fc_table);
2966 }
38308473
DM		 2967
2968 if (!table)
c71099ac 2969 goto out;
c71099ac 2970
93531c67
DA		 2971 err = -ENOMEM;
2972 rt = fib6_info_alloc(gfp_flags);
2973 if (!rt)
1da177e4 2974 goto out;
93531c67 2975
767a2217
DA		 2976 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len);
2977 if (IS_ERR(rt->fib6_metrics)) {
2978 err = PTR_ERR(rt->fib6_metrics);
fda21d46
ED		 2979 /* Do not leave garbage there. */
2980 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
767a2217
DA		 2981 goto out;
2982 }
2983
93531c67
DA		 2984 if (cfg->fc_flags & RTF_ADDRCONF)
2985 rt->dst_nocount = true;
1da177e4 2986
1716a961 2987 if (cfg->fc_flags & RTF_EXPIRES)
14895687 2988 fib6_set_expires(rt, jiffies +
1716a961
G		 2989 clock_t_to_jiffies(cfg->fc_expires));
2990 else
14895687 2991 fib6_clean_expires(rt);
1da177e4 2992
86872cb5
TG		 2993 if (cfg->fc_protocol == RTPROT_UNSPEC)
2994 cfg->fc_protocol = RTPROT_BOOT;
93c2fb25 2995 rt->fib6_protocol = cfg->fc_protocol;
86872cb5
TG		 2996
2997 addr_type = ipv6_addr_type(&cfg->fc_dst);
1da177e4 2998
19e42e45
RP		 2999 if (cfg->fc_encap) {
3000 struct lwtunnel_state *lwtstate;
3001
30357d7d 3002 err = lwtunnel_build_state(cfg->fc_encap_type,
127eb7cd 3003 cfg->fc_encap, AF_INET6, cfg,
9ae28727 3004 &lwtstate, extack);
19e42e45
RP		 3005 if (err)
3006 goto out;
5e670d84 3007 rt->fib6_nh.nh_lwtstate = lwtstate_get(lwtstate);
19e42e45
RP		 3008 }
3009
93c2fb25
DA		 3010 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3011 rt->fib6_dst.plen = cfg->fc_dst_len;
3012 if (rt->fib6_dst.plen == 128)
3b6761d1 3013 rt->dst_host = true;
e5fd387a 3014
1da177e4 3015#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 3016 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3017 rt->fib6_src.plen = cfg->fc_src_len;
1da177e4
LT		 3018#endif
3019
93c2fb25 3020 rt->fib6_metric = cfg->fc_metric;
5e670d84 3021 rt->fib6_nh.nh_weight = 1;
1da177e4 3022
e8478e80 3023 rt->fib6_type = cfg->fc_type;
1da177e4
LT		 3024
3025 /* We cannot add true routes via loopback here,
3026 they would result in kernel looping; promote them to reject routes
3027 */
86872cb5 3028 if ((cfg->fc_flags & RTF_REJECT) ||
38308473
DM		 3029 (dev && (dev->flags & IFF_LOOPBACK) &&
3030 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3031 !(cfg->fc_flags & RTF_LOCAL))) {
1da177e4 3032 /* hold loopback dev/idev if we haven't done so. */
5578689a 3033 if (dev != net->loopback_dev) {
1da177e4
LT		 3034 if (dev) {
3035 dev_put(dev);
3036 in6_dev_put(idev);
3037 }
5578689a 3038 dev = net->loopback_dev;
1da177e4
LT		 3039 dev_hold(dev);
3040 idev = in6_dev_get(dev);
3041 if (!idev) {
3042 err = -ENODEV;
3043 goto out;
3044 }
3045 }
93c2fb25 3046 rt->fib6_flags = RTF_REJECT|RTF_NONEXTHOP;
1da177e4
LT		 3047 goto install_route;
3048 }
3049
86872cb5 3050 if (cfg->fc_flags & RTF_GATEWAY) {
9fbb704c
DA		 3051 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3052 if (err)
48ed7b26 3053 goto out;
1da177e4 3054
93531c67 3055 rt->fib6_nh.nh_gw = cfg->fc_gateway;
1da177e4
LT		 3056 }
3057
3058 err = -ENODEV;
38308473 3059 if (!dev)
1da177e4
LT		 3060 goto out;
3061
428604fb
LB		 3062 if (idev->cnf.disable_ipv6) {
3063 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3064 err = -EACCES;
3065 goto out;
3066 }
3067
955ec4cb
DA		 3068 if (!(dev->flags & IFF_UP)) {
3069 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3070 err = -ENETDOWN;
3071 goto out;
3072 }
3073
c3968a85
DW		 3074 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3075 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
d5d531cb 3076 NL_SET_ERR_MSG(extack, "Invalid source address");
c3968a85
DW		 3077 err = -EINVAL;
3078 goto out;
3079 }
93c2fb25
DA		 3080 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3081 rt->fib6_prefsrc.plen = 128;
c3968a85 3082 } else
93c2fb25 3083 rt->fib6_prefsrc.plen = 0;
c3968a85 3084
93c2fb25 3085 rt->fib6_flags = cfg->fc_flags;
1da177e4
LT		 3086
3087install_route:
93c2fb25 3088 if (!(rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
5609b80a 3089 !netif_carrier_ok(dev))
5e670d84
DA		 3090 rt->fib6_nh.nh_flags |= RTNH_F_LINKDOWN;
3091 rt->fib6_nh.nh_flags |= (cfg->fc_flags & RTNH_F_ONLINK);
93531c67 3092 rt->fib6_nh.nh_dev = dev;
93c2fb25 3093 rt->fib6_table = table;
63152fc0 3094
dcd1f572
DA		 3095 if (idev)
3096 in6_dev_put(idev);
3097
8c5b83f0 3098 return rt;
6b9ea5a6
RP		 3099out:
3100 if (dev)
3101 dev_put(dev);
3102 if (idev)
3103 in6_dev_put(idev);
6b9ea5a6 3104
93531c67 3105 fib6_info_release(rt);
8c5b83f0 3106 return ERR_PTR(err);
6b9ea5a6
RP		 3107}
3108
acb54e3c 3109int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
333c4301 3110 struct netlink_ext_ack *extack)
6b9ea5a6 3111{
8d1c802b 3112 struct fib6_info *rt;
6b9ea5a6
RP		 3113 int err;
3114
acb54e3c 3115 rt = ip6_route_info_create(cfg, gfp_flags, extack);
d4ead6b3
DA		 3116 if (IS_ERR(rt))
3117 return PTR_ERR(rt);
6b9ea5a6 3118
d4ead6b3 3119 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
93531c67 3120 fib6_info_release(rt);
6b9ea5a6 3121
1da177e4
LT		 3122 return err;
3123}
3124
8d1c802b 3125static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
1da177e4 3126{
afb1d4b5 3127 struct net *net = info->nl_net;
c71099ac 3128 struct fib6_table *table;
afb1d4b5 3129 int err;
1da177e4 3130
421842ed 3131 if (rt == net->ipv6.fib6_null_entry) {
6825a26c
G		 3132 err = -ENOENT;
3133 goto out;
3134 }
6c813a72 3135
93c2fb25 3136 table = rt->fib6_table;
66f5d6ce 3137 spin_lock_bh(&table->tb6_lock);
86872cb5 3138 err = fib6_del(rt, info);
66f5d6ce 3139 spin_unlock_bh(&table->tb6_lock);
1da177e4 3140
6825a26c 3141out:
93531c67 3142 fib6_info_release(rt);
1da177e4
LT		 3143 return err;
3144}
3145
8d1c802b 3146int ip6_del_rt(struct net *net, struct fib6_info *rt)
e0a1ad73 3147{
afb1d4b5
DA		 3148 struct nl_info info = { .nl_net = net };
3149
528c4ceb 3150 return __ip6_del_rt(rt, &info);
e0a1ad73
TG		 3151}
3152
8d1c802b 3153static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
0ae81335
DA		 3154{
3155 struct nl_info *info = &cfg->fc_nlinfo;
e3330039 3156 struct net *net = info->nl_net;
16a16cd3 3157 struct sk_buff *skb = NULL;
0ae81335 3158 struct fib6_table *table;
e3330039 3159 int err = -ENOENT;
0ae81335 3160
421842ed 3161 if (rt == net->ipv6.fib6_null_entry)
e3330039 3162 goto out_put;
93c2fb25 3163 table = rt->fib6_table;
66f5d6ce 3164 spin_lock_bh(&table->tb6_lock);
0ae81335 3165
93c2fb25 3166 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
8d1c802b 3167 struct fib6_info *sibling, *next_sibling;
0ae81335 3168
16a16cd3
DA		 3169 /* prefer to send a single notification with all hops */
3170 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3171 if (skb) {
3172 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3173
d4ead6b3 3174 if (rt6_fill_node(net, skb, rt, NULL,
16a16cd3
DA		 3175 NULL, NULL, 0, RTM_DELROUTE,
3176 info->portid, seq, 0) < 0) {
3177 kfree_skb(skb);
3178 skb = NULL;
3179 } else
3180 info->skip_notify = 1;
3181 }
3182
0ae81335 3183 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25
DA		 3184 &rt->fib6_siblings,
3185 fib6_siblings) {
0ae81335
DA		 3186 err = fib6_del(sibling, info);
3187 if (err)
e3330039 3188 goto out_unlock;
0ae81335
DA		 3189 }
3190 }
3191
3192 err = fib6_del(rt, info);
e3330039 3193out_unlock:
66f5d6ce 3194 spin_unlock_bh(&table->tb6_lock);
e3330039 3195out_put:
93531c67 3196 fib6_info_release(rt);
16a16cd3
DA		 3197
3198 if (skb) {
e3330039 3199 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
16a16cd3
DA		 3200 info->nlh, gfp_any());
3201 }
0ae81335
DA		 3202 return err;
3203}
3204
23fb93a4
DA		 3205static int ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3206{
3207 int rc = -ESRCH;
3208
3209 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3210 goto out;
3211
3212 if (cfg->fc_flags & RTF_GATEWAY &&
3213 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3214 goto out;
3215 if (dst_hold_safe(&rt->dst))
3216 rc = rt6_remove_exception_rt(rt);
3217out:
3218 return rc;
3219}
3220
333c4301
DA		 3221static int ip6_route_del(struct fib6_config *cfg,
3222 struct netlink_ext_ack *extack)
1da177e4 3223{
8d1c802b 3224 struct rt6_info *rt_cache;
c71099ac 3225 struct fib6_table *table;
8d1c802b 3226 struct fib6_info *rt;
1da177e4 3227 struct fib6_node *fn;
1da177e4
LT		 3228 int err = -ESRCH;
3229
5578689a 3230 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
d5d531cb
DA		 3231 if (!table) {
3232 NL_SET_ERR_MSG(extack, "FIB table does not exist");
c71099ac 3233 return err;
d5d531cb 3234 }
c71099ac 3235
66f5d6ce 3236 rcu_read_lock();
1da177e4 3237
c71099ac 3238 fn = fib6_locate(&table->tb6_root,
86872cb5 3239 &cfg->fc_dst, cfg->fc_dst_len,
38fbeeee 3240 &cfg->fc_src, cfg->fc_src_len,
2b760fcf 3241 !(cfg->fc_flags & RTF_CACHE));
1ab1457c 3242
1da177e4 3243 if (fn) {
66f5d6ce 3244 for_each_fib6_node_rt_rcu(fn) {
2b760fcf 3245 if (cfg->fc_flags & RTF_CACHE) {
23fb93a4
DA		 3246 int rc;
3247
2b760fcf
WW		 3248 rt_cache = rt6_find_cached_rt(rt, &cfg->fc_dst,
3249 &cfg->fc_src);
23fb93a4
DA		 3250 if (rt_cache) {
3251 rc = ip6_del_cached_rt(rt_cache, cfg);
9e575010
ED		 3252 if (rc != -ESRCH) {
3253 rcu_read_unlock();
23fb93a4 3254 return rc;
9e575010 3255 }
23fb93a4
DA		 3256 }
3257 continue;
2b760fcf 3258 }
86872cb5 3259 if (cfg->fc_ifindex &&
5e670d84
DA		 3260 (!rt->fib6_nh.nh_dev ||
3261 rt->fib6_nh.nh_dev->ifindex != cfg->fc_ifindex))
1da177e4 3262 continue;
86872cb5 3263 if (cfg->fc_flags & RTF_GATEWAY &&
5e670d84 3264 !ipv6_addr_equal(&cfg->fc_gateway, &rt->fib6_nh.nh_gw))
1da177e4 3265 continue;
93c2fb25 3266 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
1da177e4 3267 continue;
93c2fb25 3268 if (cfg->fc_protocol && cfg->fc_protocol != rt->fib6_protocol)
c2ed1880 3269 continue;
e873e4b9
WW		 3270 if (!fib6_info_hold_safe(rt))
3271 continue;
66f5d6ce 3272 rcu_read_unlock();
1da177e4 3273
0ae81335
DA		 3274 /* if gateway was specified only delete the one hop */
3275 if (cfg->fc_flags & RTF_GATEWAY)
3276 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3277
3278 return __ip6_del_rt_siblings(rt, cfg);
1da177e4
LT		 3279 }
3280 }
66f5d6ce 3281 rcu_read_unlock();
1da177e4
LT		 3282
3283 return err;
3284}
3285
6700c270 3286static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
a6279458 3287{
a6279458 3288 struct netevent_redirect netevent;
e8599ff4 3289 struct rt6_info *rt, *nrt = NULL;
e8599ff4
DM		 3290 struct ndisc_options ndopts;
3291 struct inet6_dev *in6_dev;
3292 struct neighbour *neigh;
a68886a6 3293 struct fib6_info *from;
71bcdba0 3294 struct rd_msg *msg;
6e157b6a
DM		 3295 int optlen, on_link;
3296 u8 *lladdr;
e8599ff4 3297
29a3cad5 3298 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
71bcdba0 3299 optlen -= sizeof(*msg);
e8599ff4
DM		 3300
3301 if (optlen < 0) {
6e157b6a 3302 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
e8599ff4
DM		 3303 return;
3304 }
3305
71bcdba0 3306 msg = (struct rd_msg *)icmp6_hdr(skb);
e8599ff4 3307
71bcdba0 3308 if (ipv6_addr_is_multicast(&msg->dest)) {
6e157b6a 3309 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
e8599ff4
DM		 3310 return;
3311 }
3312
6e157b6a 3313 on_link = 0;
71bcdba0 3314 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
e8599ff4 3315 on_link = 1;
71bcdba0 3316 } else if (ipv6_addr_type(&msg->target) !=
e8599ff4 3317 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
6e157b6a 3318 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
e8599ff4
DM		 3319 return;
3320 }
3321
3322 in6_dev = __in6_dev_get(skb->dev);
3323 if (!in6_dev)
3324 return;
3325 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
3326 return;
3327
3328 /* RFC2461 8.1:
3329 * The IP source address of the Redirect MUST be the same as the current
3330 * first-hop router for the specified ICMP Destination Address.
3331 */
3332
f997c55c 3333 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
e8599ff4
DM		 3334 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
3335 return;
3336 }
6e157b6a
DM		 3337
3338 lladdr = NULL;
e8599ff4
DM		 3339 if (ndopts.nd_opts_tgt_lladdr) {
3340 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
3341 skb->dev);
3342 if (!lladdr) {
3343 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
3344 return;
3345 }
3346 }
3347
6e157b6a 3348 rt = (struct rt6_info *) dst;
ec13ad1d 3349 if (rt->rt6i_flags & RTF_REJECT) {
6e157b6a 3350 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
e8599ff4 3351 return;
6e157b6a 3352 }
e8599ff4 3353
6e157b6a
DM		 3354 /* Redirect received -> path was valid.
3355 * Look, redirects are sent only in response to data packets,
3356 * so that this nexthop apparently is reachable. --ANK
3357 */
0dec879f 3358 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
a6279458 3359
71bcdba0 3360 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
6e157b6a
DM		 3361 if (!neigh)
3362 return;
a6279458 3363
1da177e4
LT		 3364 /*
3365 * We have finally decided to accept it.
3366 */
3367
f997c55c 3368 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1da177e4
LT		 3369 NEIGH_UPDATE_F_WEAK_OVERRIDE|
3370 NEIGH_UPDATE_F_OVERRIDE|
3371 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
f997c55c
AA		 3372 NEIGH_UPDATE_F_ISROUTER)),
3373 NDISC_REDIRECT, &ndopts);
1da177e4 3374
4d85cd0c 3375 rcu_read_lock();
a68886a6 3376 from = rcu_dereference(rt->from);
e873e4b9
WW		 3377 /* This fib6_info_hold() is safe here because we hold reference to rt
3378 * and rt already holds reference to fib6_info.
3379 */
8a14e46f 3380 fib6_info_hold(from);
4d85cd0c 3381 rcu_read_unlock();
8a14e46f
DA		 3382
3383 nrt = ip6_rt_cache_alloc(from, &msg->dest, NULL);
38308473 3384 if (!nrt)
1da177e4
LT		 3385 goto out;
3386
3387 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
3388 if (on_link)
3389 nrt->rt6i_flags &= ~RTF_GATEWAY;
3390
4e3fd7a0 3391 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1da177e4 3392
2b760fcf
WW		 3393 /* No need to remove rt from the exception table if rt is
3394 * a cached route because rt6_insert_exception() will
3395 * takes care of it
3396 */
8a14e46f 3397 if (rt6_insert_exception(nrt, from)) {
2b760fcf
WW		 3398 dst_release_immediate(&nrt->dst);
3399 goto out;
3400 }
1da177e4 3401
d8d1f30b
CG		 3402 netevent.old = &rt->dst;
3403 netevent.new = &nrt->dst;
71bcdba0 3404 netevent.daddr = &msg->dest;
60592833 3405 netevent.neigh = neigh;
8d71740c
TT		 3406 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
3407
1da177e4 3408out:
8a14e46f 3409 fib6_info_release(from);
e8599ff4 3410 neigh_release(neigh);
6e157b6a
DM		 3411}
3412
70ceb4f5 3413#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 3414static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 3415 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 3416 const struct in6_addr *gwaddr,
3417 struct net_device *dev)
70ceb4f5 3418{
830218c1
DA		 3419 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
3420 int ifindex = dev->ifindex;
70ceb4f5 3421 struct fib6_node *fn;
8d1c802b 3422 struct fib6_info *rt = NULL;
c71099ac
TG		 3423 struct fib6_table *table;
3424
830218c1 3425 table = fib6_get_table(net, tb_id);
38308473 3426 if (!table)
c71099ac 3427 return NULL;
70ceb4f5 3428
66f5d6ce 3429 rcu_read_lock();
38fbeeee 3430 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
70ceb4f5
YH		 3431 if (!fn)
3432 goto out;
3433
66f5d6ce 3434 for_each_fib6_node_rt_rcu(fn) {
5e670d84 3435 if (rt->fib6_nh.nh_dev->ifindex != ifindex)
70ceb4f5 3436 continue;
93c2fb25 3437 if ((rt->fib6_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
70ceb4f5 3438 continue;
5e670d84 3439 if (!ipv6_addr_equal(&rt->fib6_nh.nh_gw, gwaddr))
70ceb4f5 3440 continue;
e873e4b9
WW		 3441 if (!fib6_info_hold_safe(rt))
3442 continue;
70ceb4f5
YH		 3443 break;
3444 }
3445out:
66f5d6ce 3446 rcu_read_unlock();
70ceb4f5
YH		 3447 return rt;
3448}
3449
8d1c802b 3450static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 3451 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 3452 const struct in6_addr *gwaddr,
3453 struct net_device *dev,
95c96174 3454 unsigned int pref)
70ceb4f5 3455{
86872cb5 3456 struct fib6_config cfg = {
238fc7ea 3457 .fc_metric = IP6_RT_PRIO_USER,
830218c1 3458 .fc_ifindex = dev->ifindex,
86872cb5
TG		 3459 .fc_dst_len = prefixlen,
3460 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
3461 RTF_UP | RTF_PREF(pref),
b91d5329 3462 .fc_protocol = RTPROT_RA,
e8478e80 3463 .fc_type = RTN_UNICAST,
15e47304 3464 .fc_nlinfo.portid = 0,
efa2cea0
DL		 3465 .fc_nlinfo.nlh = NULL,
3466 .fc_nlinfo.nl_net = net,
86872cb5
TG		 3467 };
3468
830218c1 3469 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
4e3fd7a0
AD		 3470 cfg.fc_dst = *prefix;
3471 cfg.fc_gateway = *gwaddr;
70ceb4f5 3472
e317da96
YH		 3473 /* We should treat it as a default route if prefix length is 0. */
3474 if (!prefixlen)
86872cb5 3475 cfg.fc_flags |= RTF_DEFAULT;
70ceb4f5 3476
acb54e3c 3477 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
70ceb4f5 3478
830218c1 3479 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
70ceb4f5
YH		 3480}
3481#endif
3482
8d1c802b 3483struct fib6_info *rt6_get_dflt_router(struct net *net,
afb1d4b5
DA		 3484 const struct in6_addr *addr,
3485 struct net_device *dev)
1ab1457c 3486{
830218c1 3487 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
8d1c802b 3488 struct fib6_info *rt;
c71099ac 3489 struct fib6_table *table;
1da177e4 3490
afb1d4b5 3491 table = fib6_get_table(net, tb_id);
38308473 3492 if (!table)
c71099ac 3493 return NULL;
1da177e4 3494
66f5d6ce
WW		 3495 rcu_read_lock();
3496 for_each_fib6_node_rt_rcu(&table->tb6_root) {
5e670d84 3497 if (dev == rt->fib6_nh.nh_dev &&
93c2fb25 3498 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
5e670d84 3499 ipv6_addr_equal(&rt->fib6_nh.nh_gw, addr))
1da177e4
LT		 3500 break;
3501 }
e873e4b9
WW		 3502 if (rt && !fib6_info_hold_safe(rt))
3503 rt = NULL;
66f5d6ce 3504 rcu_read_unlock();
1da177e4
LT		 3505 return rt;
3506}
3507
8d1c802b 3508struct fib6_info *rt6_add_dflt_router(struct net *net,
afb1d4b5 3509 const struct in6_addr *gwaddr,
ebacaaa0
YH		 3510 struct net_device *dev,
3511 unsigned int pref)
1da177e4 3512{
86872cb5 3513 struct fib6_config cfg = {
ca254490 3514 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
238fc7ea 3515 .fc_metric = IP6_RT_PRIO_USER,
86872cb5
TG		 3516 .fc_ifindex = dev->ifindex,
3517 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
3518 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
b91d5329 3519 .fc_protocol = RTPROT_RA,
e8478e80 3520 .fc_type = RTN_UNICAST,
15e47304 3521 .fc_nlinfo.portid = 0,
5578689a 3522 .fc_nlinfo.nlh = NULL,
afb1d4b5 3523 .fc_nlinfo.nl_net = net,
86872cb5 3524 };
1da177e4 3525
4e3fd7a0 3526 cfg.fc_gateway = *gwaddr;
1da177e4 3527
acb54e3c 3528 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
830218c1
DA		 3529 struct fib6_table *table;
3530
3531 table = fib6_get_table(dev_net(dev), cfg.fc_table);
3532 if (table)
3533 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
3534 }
1da177e4 3535
afb1d4b5 3536 return rt6_get_dflt_router(net, gwaddr, dev);
1da177e4
LT		 3537}
3538
afb1d4b5
DA		 3539static void __rt6_purge_dflt_routers(struct net *net,
3540 struct fib6_table *table)
1da177e4 3541{
8d1c802b 3542 struct fib6_info *rt;
1da177e4
LT		 3543
3544restart:
66f5d6ce
WW		 3545 rcu_read_lock();
3546 for_each_fib6_node_rt_rcu(&table->tb6_root) {
dcd1f572
DA		 3547 struct net_device *dev = fib6_info_nh_dev(rt);
3548 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
3549
93c2fb25 3550 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
e873e4b9
WW		 3551 (!idev || idev->cnf.accept_ra != 2) &&
3552 fib6_info_hold_safe(rt)) {
93531c67
DA		 3553 rcu_read_unlock();
3554 ip6_del_rt(net, rt);
1da177e4
LT		 3555 goto restart;
3556 }
3557 }
66f5d6ce 3558 rcu_read_unlock();
830218c1
DA		 3559
3560 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
3561}
3562
3563void rt6_purge_dflt_routers(struct net *net)
3564{
3565 struct fib6_table *table;
3566 struct hlist_head *head;
3567 unsigned int h;
3568
3569 rcu_read_lock();
3570
3571 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
3572 head = &net->ipv6.fib_table_hash[h];
3573 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
3574 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
afb1d4b5 3575 __rt6_purge_dflt_routers(net, table);
830218c1
DA		 3576 }
3577 }
3578
3579 rcu_read_unlock();
1da177e4
LT		 3580}
3581
5578689a
DL		 3582static void rtmsg_to_fib6_config(struct net *net,
3583 struct in6_rtmsg *rtmsg,
86872cb5
TG		 3584 struct fib6_config *cfg)
3585{
8823a3ac
MŻ		 3586 *cfg = (struct fib6_config){
3587 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
3588 : RT6_TABLE_MAIN,
3589 .fc_ifindex = rtmsg->rtmsg_ifindex,
3590 .fc_metric = rtmsg->rtmsg_metric,
3591 .fc_expires = rtmsg->rtmsg_info,
3592 .fc_dst_len = rtmsg->rtmsg_dst_len,
3593 .fc_src_len = rtmsg->rtmsg_src_len,
3594 .fc_flags = rtmsg->rtmsg_flags,
3595 .fc_type = rtmsg->rtmsg_type,
3596
3597 .fc_nlinfo.nl_net = net,
3598
3599 .fc_dst = rtmsg->rtmsg_dst,
3600 .fc_src = rtmsg->rtmsg_src,
3601 .fc_gateway = rtmsg->rtmsg_gateway,
3602 };
86872cb5
TG		 3603}
3604
5578689a 3605int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1da177e4 3606{
86872cb5 3607 struct fib6_config cfg;
1da177e4
LT		 3608 struct in6_rtmsg rtmsg;
3609 int err;
3610
67ba4152 3611 switch (cmd) {
1da177e4
LT		 3612 case SIOCADDRT: /* Add a route */
3613 case SIOCDELRT: /* Delete a route */
af31f412 3614 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4
LT		 3615 return -EPERM;
3616 err = copy_from_user(&rtmsg, arg,
3617 sizeof(struct in6_rtmsg));
3618 if (err)
3619 return -EFAULT;
86872cb5 3620
5578689a 3621 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
86872cb5 3622
1da177e4
LT		 3623 rtnl_lock();
3624 switch (cmd) {
3625 case SIOCADDRT:
acb54e3c 3626 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
1da177e4
LT		 3627 break;
3628 case SIOCDELRT:
333c4301 3629 err = ip6_route_del(&cfg, NULL);
1da177e4
LT		 3630 break;
3631 default:
3632 err = -EINVAL;
3633 }
3634 rtnl_unlock();
3635
3636 return err;
3ff50b79 3637 }
1da177e4
LT		 3638
3639 return -EINVAL;
3640}
3641
3642/*
3643 * Drop the packet on the floor
3644 */
3645
d5fdd6ba 3646static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1da177e4 3647{
612f09e8 3648 int type;
adf30907 3649 struct dst_entry *dst = skb_dst(skb);
612f09e8
YH		 3650 switch (ipstats_mib_noroutes) {
3651 case IPSTATS_MIB_INNOROUTES:
0660e03f 3652 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
45bb0060 3653 if (type == IPV6_ADDR_ANY) {
bdb7cc64
SS		 3654 IP6_INC_STATS(dev_net(dst->dev),
3655 __in6_dev_get_safely(skb->dev),
3bd653c8 3656 IPSTATS_MIB_INADDRERRORS);
612f09e8
YH		 3657 break;
3658 }
3659 /* FALLTHROUGH */
3660 case IPSTATS_MIB_OUTNOROUTES:
3bd653c8
DL		 3661 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
3662 ipstats_mib_noroutes);
612f09e8
YH		 3663 break;
3664 }
3ffe533c 3665 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1da177e4
LT		 3666 kfree_skb(skb);
3667 return 0;
3668}
3669
9ce8ade0
TG		 3670static int ip6_pkt_discard(struct sk_buff *skb)
3671{
612f09e8 3672 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 3673}
3674
ede2059d 3675static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
1da177e4 3676{
adf30907 3677 skb->dev = skb_dst(skb)->dev;
612f09e8 3678 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1da177e4
LT		 3679}
3680
9ce8ade0
TG		 3681static int ip6_pkt_prohibit(struct sk_buff *skb)
3682{
612f09e8 3683 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 3684}
3685
ede2059d 3686static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
9ce8ade0 3687{
adf30907 3688 skb->dev = skb_dst(skb)->dev;
612f09e8 3689 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
9ce8ade0
TG		 3690}
3691
1da177e4
LT		 3692/*
3693 * Allocate a dst for local (unicast / anycast) address.
3694 */
3695
360a9887
DA		 3696struct fib6_info *addrconf_f6i_alloc(struct net *net,
3697 struct inet6_dev *idev,
3698 const struct in6_addr *addr,
3699 bool anycast, gfp_t gfp_flags)
1da177e4 3700{
ca254490 3701 u32 tb_id;
4832c30d 3702 struct net_device *dev = idev->dev;
360a9887 3703 struct fib6_info *f6i;
5f02ce24 3704
360a9887
DA		 3705 f6i = fib6_info_alloc(gfp_flags);
3706 if (!f6i)
1da177e4
LT		 3707 return ERR_PTR(-ENOMEM);
3708
767a2217 3709 f6i->fib6_metrics = ip_fib_metrics_init(net, NULL, 0);
360a9887 3710 f6i->dst_nocount = true;
360a9887
DA		 3711 f6i->dst_host = true;
3712 f6i->fib6_protocol = RTPROT_KERNEL;
3713 f6i->fib6_flags = RTF_UP | RTF_NONEXTHOP;
e8478e80 3714 if (anycast) {
360a9887
DA		 3715 f6i->fib6_type = RTN_ANYCAST;
3716 f6i->fib6_flags |= RTF_ANYCAST;
e8478e80 3717 } else {
360a9887
DA		 3718 f6i->fib6_type = RTN_LOCAL;
3719 f6i->fib6_flags |= RTF_LOCAL;
e8478e80 3720 }
1da177e4 3721
360a9887 3722 f6i->fib6_nh.nh_gw = *addr;
93531c67 3723 dev_hold(dev);
360a9887
DA		 3724 f6i->fib6_nh.nh_dev = dev;
3725 f6i->fib6_dst.addr = *addr;
3726 f6i->fib6_dst.plen = 128;
ca254490 3727 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
360a9887 3728 f6i->fib6_table = fib6_get_table(net, tb_id);
1da177e4 3729
360a9887 3730 return f6i;
1da177e4
LT		 3731}
3732
c3968a85
DW		 3733/* remove deleted ip from prefsrc entries */
3734struct arg_dev_net_ip {
3735 struct net_device *dev;
3736 struct net *net;
3737 struct in6_addr *addr;
3738};
3739
8d1c802b 3740static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
c3968a85
DW		 3741{
3742 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
3743 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
3744 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
3745
5e670d84 3746 if (((void *)rt->fib6_nh.nh_dev == dev || !dev) &&
421842ed 3747 rt != net->ipv6.fib6_null_entry &&
93c2fb25 3748 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
60006a48 3749 spin_lock_bh(&rt6_exception_lock);
c3968a85 3750 /* remove prefsrc entry */
93c2fb25 3751 rt->fib6_prefsrc.plen = 0;
60006a48 3752 spin_unlock_bh(&rt6_exception_lock);
c3968a85
DW		 3753 }
3754 return 0;
3755}
3756
3757void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
3758{
3759 struct net *net = dev_net(ifp->idev->dev);
3760 struct arg_dev_net_ip adni = {
3761 .dev = ifp->idev->dev,
3762 .net = net,
3763 .addr = &ifp->addr,
3764 };
0c3584d5 3765 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
c3968a85
DW		 3766}
3767
be7a010d 3768#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
be7a010d
DJ		 3769
3770/* Remove routers and update dst entries when gateway turn into host. */
8d1c802b 3771static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
be7a010d
DJ		 3772{
3773 struct in6_addr *gateway = (struct in6_addr *)arg;
3774
93c2fb25 3775 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
5e670d84 3776 ipv6_addr_equal(gateway, &rt->fib6_nh.nh_gw)) {
be7a010d
DJ		 3777 return -1;
3778 }
b16cb459
WW		 3779
3780 /* Further clean up cached routes in exception table.
3781 * This is needed because cached route may have a different
3782 * gateway than its 'parent' in the case of an ip redirect.
3783 */
3784 rt6_exceptions_clean_tohost(rt, gateway);
3785
be7a010d
DJ		 3786 return 0;
3787}
3788
3789void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
3790{
3791 fib6_clean_all(net, fib6_clean_tohost, gateway);
3792}
3793
2127d95a
IS		 3794struct arg_netdev_event {
3795 const struct net_device *dev;
4c981e28
IS		 3796 union {
3797 unsigned int nh_flags;
3798 unsigned long event;
3799 };
2127d95a
IS		 3800};
3801
8d1c802b 3802static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
d7dedee1 3803{
8d1c802b 3804 struct fib6_info *iter;
d7dedee1
IS		 3805 struct fib6_node *fn;
3806
93c2fb25
DA		 3807 fn = rcu_dereference_protected(rt->fib6_node,
3808 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 3809 iter = rcu_dereference_protected(fn->leaf,
93c2fb25 3810 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 3811 while (iter) {
93c2fb25 3812 if (iter->fib6_metric == rt->fib6_metric &&
33bd5ac5 3813 rt6_qualify_for_ecmp(iter))
d7dedee1 3814 return iter;
8fb11a9a 3815 iter = rcu_dereference_protected(iter->fib6_next,
93c2fb25 3816 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1
IS		 3817 }
3818
3819 return NULL;
3820}
3821
8d1c802b 3822static bool rt6_is_dead(const struct fib6_info *rt)
d7dedee1 3823{
5e670d84
DA		 3824 if (rt->fib6_nh.nh_flags & RTNH_F_DEAD ||
3825 (rt->fib6_nh.nh_flags & RTNH_F_LINKDOWN &&
dcd1f572 3826 fib6_ignore_linkdown(rt)))
d7dedee1
IS		 3827 return true;
3828
3829 return false;
3830}
3831
8d1c802b 3832static int rt6_multipath_total_weight(const struct fib6_info *rt)
d7dedee1 3833{
8d1c802b 3834 struct fib6_info *iter;
d7dedee1
IS		 3835 int total = 0;
3836
3837 if (!rt6_is_dead(rt))
5e670d84 3838 total += rt->fib6_nh.nh_weight;
d7dedee1 3839
93c2fb25 3840 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
d7dedee1 3841 if (!rt6_is_dead(iter))
5e670d84 3842 total += iter->fib6_nh.nh_weight;
d7dedee1
IS		 3843 }
3844
3845 return total;
3846}
3847
8d1c802b 3848static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
d7dedee1
IS		 3849{
3850 int upper_bound = -1;
3851
3852 if (!rt6_is_dead(rt)) {
5e670d84 3853 *weight += rt->fib6_nh.nh_weight;
d7dedee1
IS		 3854 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
3855 total) - 1;
3856 }
5e670d84 3857 atomic_set(&rt->fib6_nh.nh_upper_bound, upper_bound);
d7dedee1
IS		 3858}
3859
8d1c802b 3860static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
d7dedee1 3861{
8d1c802b 3862 struct fib6_info *iter;
d7dedee1
IS		 3863 int weight = 0;
3864
3865 rt6_upper_bound_set(rt, &weight, total);
3866
93c2fb25 3867 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
d7dedee1
IS		 3868 rt6_upper_bound_set(iter, &weight, total);
3869}
3870
8d1c802b 3871void rt6_multipath_rebalance(struct fib6_info *rt)
d7dedee1 3872{
8d1c802b 3873 struct fib6_info *first;
d7dedee1
IS		 3874 int total;
3875
3876 /* In case the entire multipath route was marked for flushing,
3877 * then there is no need to rebalance upon the removal of every
3878 * sibling route.
3879 */
93c2fb25 3880 if (!rt->fib6_nsiblings || rt->should_flush)
d7dedee1
IS		 3881 return;
3882
3883 /* During lookup routes are evaluated in order, so we need to
3884 * make sure upper bounds are assigned from the first sibling
3885 * onwards.
3886 */
3887 first = rt6_multipath_first_sibling(rt);
3888 if (WARN_ON_ONCE(!first))
3889 return;
3890
3891 total = rt6_multipath_total_weight(first);
3892 rt6_multipath_upper_bound_set(first, total);
3893}
3894
8d1c802b 3895static int fib6_ifup(struct fib6_info *rt, void *p_arg)
2127d95a
IS		 3896{
3897 const struct arg_netdev_event *arg = p_arg;
7aef6859 3898 struct net *net = dev_net(arg->dev);
2127d95a 3899
421842ed 3900 if (rt != net->ipv6.fib6_null_entry && rt->fib6_nh.nh_dev == arg->dev) {
5e670d84 3901 rt->fib6_nh.nh_flags &= ~arg->nh_flags;
7aef6859 3902 fib6_update_sernum_upto_root(net, rt);
d7dedee1 3903 rt6_multipath_rebalance(rt);
1de178ed 3904 }
2127d95a
IS		 3905
3906 return 0;
3907}
3908
3909void rt6_sync_up(struct net_device *dev, unsigned int nh_flags)
3910{
3911 struct arg_netdev_event arg = {
3912 .dev = dev,
6802f3ad
IS		 3913 {
3914 .nh_flags = nh_flags,
3915 },
2127d95a
IS		 3916 };
3917
3918 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
3919 arg.nh_flags |= RTNH_F_LINKDOWN;
3920
3921 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
3922}
3923
8d1c802b 3924static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
1de178ed
IS		 3925 const struct net_device *dev)
3926{
8d1c802b 3927 struct fib6_info *iter;
1de178ed 3928
5e670d84 3929 if (rt->fib6_nh.nh_dev == dev)
1de178ed 3930 return true;
93c2fb25 3931 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
5e670d84 3932 if (iter->fib6_nh.nh_dev == dev)
1de178ed
IS		 3933 return true;
3934
3935 return false;
3936}
3937
8d1c802b 3938static void rt6_multipath_flush(struct fib6_info *rt)
1de178ed 3939{
8d1c802b 3940 struct fib6_info *iter;
1de178ed
IS		 3941
3942 rt->should_flush = 1;
93c2fb25 3943 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1de178ed
IS		 3944 iter->should_flush = 1;
3945}
3946
8d1c802b 3947static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
1de178ed
IS		 3948 const struct net_device *down_dev)
3949{
8d1c802b 3950 struct fib6_info *iter;
1de178ed
IS		 3951 unsigned int dead = 0;
3952
5e670d84
DA		 3953 if (rt->fib6_nh.nh_dev == down_dev ||
3954 rt->fib6_nh.nh_flags & RTNH_F_DEAD)
1de178ed 3955 dead++;
93c2fb25 3956 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
5e670d84
DA		 3957 if (iter->fib6_nh.nh_dev == down_dev ||
3958 iter->fib6_nh.nh_flags & RTNH_F_DEAD)
1de178ed
IS		 3959 dead++;
3960
3961 return dead;
3962}
3963
8d1c802b 3964static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
1de178ed
IS		 3965 const struct net_device *dev,
3966 unsigned int nh_flags)
3967{
8d1c802b 3968 struct fib6_info *iter;
1de178ed 3969
5e670d84
DA		 3970 if (rt->fib6_nh.nh_dev == dev)
3971 rt->fib6_nh.nh_flags |= nh_flags;
93c2fb25 3972 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
5e670d84
DA		 3973 if (iter->fib6_nh.nh_dev == dev)
3974 iter->fib6_nh.nh_flags |= nh_flags;
1de178ed
IS		 3975}
3976
a1a22c12 3977/* called with write lock held for table with rt */
8d1c802b 3978static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
1da177e4 3979{
4c981e28
IS		 3980 const struct arg_netdev_event *arg = p_arg;
3981 const struct net_device *dev = arg->dev;
7aef6859 3982 struct net *net = dev_net(dev);
8ed67789 3983
421842ed 3984 if (rt == net->ipv6.fib6_null_entry)
27c6fa73
IS		 3985 return 0;
3986
3987 switch (arg->event) {
3988 case NETDEV_UNREGISTER:
5e670d84 3989 return rt->fib6_nh.nh_dev == dev ? -1 : 0;
27c6fa73 3990 case NETDEV_DOWN:
1de178ed 3991 if (rt->should_flush)
27c6fa73 3992 return -1;
93c2fb25 3993 if (!rt->fib6_nsiblings)
5e670d84 3994 return rt->fib6_nh.nh_dev == dev ? -1 : 0;
1de178ed
IS		 3995 if (rt6_multipath_uses_dev(rt, dev)) {
3996 unsigned int count;
3997
3998 count = rt6_multipath_dead_count(rt, dev);
93c2fb25 3999 if (rt->fib6_nsiblings + 1 == count) {
1de178ed
IS		 4000 rt6_multipath_flush(rt);
4001 return -1;
4002 }
4003 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4004 RTNH_F_LINKDOWN);
7aef6859 4005 fib6_update_sernum(net, rt);
d7dedee1 4006 rt6_multipath_rebalance(rt);
1de178ed
IS		 4007 }
4008 return -2;
27c6fa73 4009 case NETDEV_CHANGE:
5e670d84 4010 if (rt->fib6_nh.nh_dev != dev ||
93c2fb25 4011 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
27c6fa73 4012 break;
5e670d84 4013 rt->fib6_nh.nh_flags |= RTNH_F_LINKDOWN;
d7dedee1 4014 rt6_multipath_rebalance(rt);
27c6fa73 4015 break;
2b241361 4016 }
c159d30c 4017
1da177e4
LT		 4018 return 0;
4019}
4020
27c6fa73 4021void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
1da177e4 4022{
4c981e28 4023 struct arg_netdev_event arg = {
8ed67789 4024 .dev = dev,
6802f3ad
IS		 4025 {
4026 .event = event,
4027 },
8ed67789
DL		 4028 };
4029
4c981e28
IS		 4030 fib6_clean_all(dev_net(dev), fib6_ifdown, &arg);
4031}
4032
4033void rt6_disable_ip(struct net_device *dev, unsigned long event)
4034{
4035 rt6_sync_down_dev(dev, event);
4036 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4037 neigh_ifdown(&nd_tbl, dev);
1da177e4
LT		 4038}
4039
95c96174 4040struct rt6_mtu_change_arg {
1da177e4 4041 struct net_device *dev;
95c96174 4042 unsigned int mtu;
1da177e4
LT		 4043};
4044
8d1c802b 4045static int rt6_mtu_change_route(struct fib6_info *rt, void *p_arg)
1da177e4
LT		 4046{
4047 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4048 struct inet6_dev *idev;
4049
4050 /* In IPv6 pmtu discovery is not optional,
4051 so that RTAX_MTU lock cannot disable it.
4052 We still use this lock to block changes
4053 caused by addrconf/ndisc.
4054 */
4055
4056 idev = __in6_dev_get(arg->dev);
38308473 4057 if (!idev)
1da177e4
LT		 4058 return 0;
4059
4060 /* For administrative MTU increase, there is no way to discover
4061 IPv6 PMTU increase, so PMTU increase should be updated here.
4062 Since RFC 1981 doesn't include administrative MTU increase
4063 update PMTU increase is a MUST. (i.e. jumbo frame)
4064 */
5e670d84 4065 if (rt->fib6_nh.nh_dev == arg->dev &&
d4ead6b3
DA		 4066 !fib6_metric_locked(rt, RTAX_MTU)) {
4067 u32 mtu = rt->fib6_pmtu;
4068
4069 if (mtu >= arg->mtu ||
4070 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4071 fib6_metric_set(rt, RTAX_MTU, arg->mtu);
4072
f5bbe7ee 4073 spin_lock_bh(&rt6_exception_lock);
e9fa1495 4074 rt6_exceptions_update_pmtu(idev, rt, arg->mtu);
f5bbe7ee 4075 spin_unlock_bh(&rt6_exception_lock);
566cfd8f 4076 }
1da177e4
LT		 4077 return 0;
4078}
4079
95c96174 4080void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
1da177e4 4081{
c71099ac
TG		 4082 struct rt6_mtu_change_arg arg = {
4083 .dev = dev,
4084 .mtu = mtu,
4085 };
1da177e4 4086
0c3584d5 4087 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
1da177e4
LT		 4088}
4089
ef7c79ed 4090static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
5176f91e 4091 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
aa8f8778 4092 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
86872cb5 4093 [RTA_OIF] = { .type = NLA_U32 },
ab364a6f 4094 [RTA_IIF] = { .type = NLA_U32 },
86872cb5
TG		 4095 [RTA_PRIORITY] = { .type = NLA_U32 },
4096 [RTA_METRICS] = { .type = NLA_NESTED },
51ebd318 4097 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
c78ba6d6 4098 [RTA_PREF] = { .type = NLA_U8 },
19e42e45
RP		 4099 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4100 [RTA_ENCAP] = { .type = NLA_NESTED },
32bc201e 4101 [RTA_EXPIRES] = { .type = NLA_U32 },
622ec2c9 4102 [RTA_UID] = { .type = NLA_U32 },
3b45a410 4103 [RTA_MARK] = { .type = NLA_U32 },
aa8f8778 4104 [RTA_TABLE] = { .type = NLA_U32 },
eacb9384
RP		 4105 [RTA_IP_PROTO] = { .type = NLA_U8 },
4106 [RTA_SPORT] = { .type = NLA_U16 },
4107 [RTA_DPORT] = { .type = NLA_U16 },
86872cb5
TG		 4108};
4109
4110static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
333c4301
DA		 4111 struct fib6_config *cfg,
4112 struct netlink_ext_ack *extack)
1da177e4 4113{
86872cb5
TG		 4114 struct rtmsg *rtm;
4115 struct nlattr *tb[RTA_MAX+1];
c78ba6d6 4116 unsigned int pref;
86872cb5 4117 int err;
1da177e4 4118
fceb6435 4119 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
dac9c979 4120 extack);
86872cb5
TG		 4121 if (err < 0)
4122 goto errout;
1da177e4 4123
86872cb5
TG		 4124 err = -EINVAL;
4125 rtm = nlmsg_data(nlh);
86872cb5 4126
84db8407
MŻ		 4127 *cfg = (struct fib6_config){
4128 .fc_table = rtm->rtm_table,
4129 .fc_dst_len = rtm->rtm_dst_len,
4130 .fc_src_len = rtm->rtm_src_len,
4131 .fc_flags = RTF_UP,
4132 .fc_protocol = rtm->rtm_protocol,
4133 .fc_type = rtm->rtm_type,
4134
4135 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
4136 .fc_nlinfo.nlh = nlh,
4137 .fc_nlinfo.nl_net = sock_net(skb->sk),
4138 };
86872cb5 4139
ef2c7d7b
ND		 4140 if (rtm->rtm_type == RTN_UNREACHABLE ||
4141 rtm->rtm_type == RTN_BLACKHOLE ||
b4949ab2
ND		 4142 rtm->rtm_type == RTN_PROHIBIT ||
4143 rtm->rtm_type == RTN_THROW)
86872cb5
TG		 4144 cfg->fc_flags |= RTF_REJECT;
4145
ab79ad14
MŻ		 4146 if (rtm->rtm_type == RTN_LOCAL)
4147 cfg->fc_flags |= RTF_LOCAL;
4148
1f56a01f
MKL		 4149 if (rtm->rtm_flags & RTM_F_CLONED)
4150 cfg->fc_flags |= RTF_CACHE;
4151
fc1e64e1
DA		 4152 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4153
86872cb5 4154 if (tb[RTA_GATEWAY]) {
67b61f6c 4155 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
86872cb5 4156 cfg->fc_flags |= RTF_GATEWAY;
1da177e4 4157 }
86872cb5
TG		 4158
4159 if (tb[RTA_DST]) {
4160 int plen = (rtm->rtm_dst_len + 7) >> 3;
4161
4162 if (nla_len(tb[RTA_DST]) < plen)
4163 goto errout;
4164
4165 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1da177e4 4166 }
86872cb5
TG		 4167
4168 if (tb[RTA_SRC]) {
4169 int plen = (rtm->rtm_src_len + 7) >> 3;
4170
4171 if (nla_len(tb[RTA_SRC]) < plen)
4172 goto errout;
4173
4174 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1da177e4 4175 }
86872cb5 4176
c3968a85 4177 if (tb[RTA_PREFSRC])
67b61f6c 4178 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
c3968a85 4179
86872cb5
TG		 4180 if (tb[RTA_OIF])
4181 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4182
4183 if (tb[RTA_PRIORITY])
4184 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4185
4186 if (tb[RTA_METRICS]) {
4187 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4188 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
1da177e4 4189 }
86872cb5
TG		 4190
4191 if (tb[RTA_TABLE])
4192 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4193
51ebd318
ND		 4194 if (tb[RTA_MULTIPATH]) {
4195 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4196 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
9ed59592
DA		 4197
4198 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
c255bd68 4199 cfg->fc_mp_len, extack);
9ed59592
DA		 4200 if (err < 0)
4201 goto errout;
51ebd318
ND		 4202 }
4203
c78ba6d6
LR		 4204 if (tb[RTA_PREF]) {
4205 pref = nla_get_u8(tb[RTA_PREF]);
4206 if (pref != ICMPV6_ROUTER_PREF_LOW &&
4207 pref != ICMPV6_ROUTER_PREF_HIGH)
4208 pref = ICMPV6_ROUTER_PREF_MEDIUM;
4209 cfg->fc_flags |= RTF_PREF(pref);
4210 }
4211
19e42e45
RP		 4212 if (tb[RTA_ENCAP])
4213 cfg->fc_encap = tb[RTA_ENCAP];
4214
9ed59592 4215 if (tb[RTA_ENCAP_TYPE]) {
19e42e45
RP		 4216 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
4217
c255bd68 4218 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
9ed59592
DA		 4219 if (err < 0)
4220 goto errout;
4221 }
4222
32bc201e
XL		 4223 if (tb[RTA_EXPIRES]) {
4224 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
4225
4226 if (addrconf_finite_timeout(timeout)) {
4227 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
4228 cfg->fc_flags |= RTF_EXPIRES;
4229 }
4230 }
4231
86872cb5
TG		 4232 err = 0;
4233errout:
4234 return err;
1da177e4
LT		 4235}
4236
6b9ea5a6 4237struct rt6_nh {
8d1c802b 4238 struct fib6_info *fib6_info;
6b9ea5a6 4239 struct fib6_config r_cfg;
6b9ea5a6
RP		 4240 struct list_head next;
4241};
4242
4243static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
4244{
4245 struct rt6_nh *nh;
4246
4247 list_for_each_entry(nh, rt6_nh_list, next) {
7d4d5065 4248 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6c nexthop %pI6c ifi %d\n",
6b9ea5a6
RP		 4249 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
4250 nh->r_cfg.fc_ifindex);
4251 }
4252}
4253
d4ead6b3
DA		 4254static int ip6_route_info_append(struct net *net,
4255 struct list_head *rt6_nh_list,
8d1c802b
DA		 4256 struct fib6_info *rt,
4257 struct fib6_config *r_cfg)
6b9ea5a6
RP		 4258{
4259 struct rt6_nh *nh;
6b9ea5a6
RP		 4260 int err = -EEXIST;
4261
4262 list_for_each_entry(nh, rt6_nh_list, next) {
8d1c802b
DA		 4263 /* check if fib6_info already exists */
4264 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
6b9ea5a6
RP		 4265 return err;
4266 }
4267
4268 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
4269 if (!nh)
4270 return -ENOMEM;
8d1c802b 4271 nh->fib6_info = rt;
6b9ea5a6
RP		 4272 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
4273 list_add_tail(&nh->next, rt6_nh_list);
4274
4275 return 0;
4276}
4277
8d1c802b
DA		 4278static void ip6_route_mpath_notify(struct fib6_info *rt,
4279 struct fib6_info *rt_last,
3b1137fe
DA		 4280 struct nl_info *info,
4281 __u16 nlflags)
4282{
4283 /* if this is an APPEND route, then rt points to the first route
4284 * inserted and rt_last points to last route inserted. Userspace
4285 * wants a consistent dump of the route which starts at the first
4286 * nexthop. Since sibling routes are always added at the end of
4287 * the list, find the first sibling of the last route appended
4288 */
93c2fb25
DA		 4289 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
4290 rt = list_first_entry(&rt_last->fib6_siblings,
8d1c802b 4291 struct fib6_info,
93c2fb25 4292 fib6_siblings);
3b1137fe
DA		 4293 }
4294
4295 if (rt)
4296 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
4297}
4298
333c4301
DA		 4299static int ip6_route_multipath_add(struct fib6_config *cfg,
4300 struct netlink_ext_ack *extack)
51ebd318 4301{
8d1c802b 4302 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
3b1137fe 4303 struct nl_info *info = &cfg->fc_nlinfo;
51ebd318
ND		 4304 struct fib6_config r_cfg;
4305 struct rtnexthop *rtnh;
8d1c802b 4306 struct fib6_info *rt;
6b9ea5a6
RP		 4307 struct rt6_nh *err_nh;
4308 struct rt6_nh *nh, *nh_safe;
3b1137fe 4309 __u16 nlflags;
51ebd318
ND		 4310 int remaining;
4311 int attrlen;
6b9ea5a6
RP		 4312 int err = 1;
4313 int nhn = 0;
4314 int replace = (cfg->fc_nlinfo.nlh &&
4315 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
4316 LIST_HEAD(rt6_nh_list);
51ebd318 4317
3b1137fe
DA		 4318 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
4319 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
4320 nlflags |= NLM_F_APPEND;
4321
35f1b4e9 4322 remaining = cfg->fc_mp_len;
51ebd318 4323 rtnh = (struct rtnexthop *)cfg->fc_mp;
51ebd318 4324
6b9ea5a6 4325 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
8d1c802b 4326 * fib6_info structs per nexthop
6b9ea5a6 4327 */
51ebd318
ND		 4328 while (rtnh_ok(rtnh, remaining)) {
4329 memcpy(&r_cfg, cfg, sizeof(*cfg));
4330 if (rtnh->rtnh_ifindex)
4331 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
4332
4333 attrlen = rtnh_attrlen(rtnh);
4334 if (attrlen > 0) {
4335 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
4336
4337 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
4338 if (nla) {
67b61f6c 4339 r_cfg.fc_gateway = nla_get_in6_addr(nla);
51ebd318
ND		 4340 r_cfg.fc_flags |= RTF_GATEWAY;
4341 }
19e42e45
RP		 4342 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
4343 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
4344 if (nla)
4345 r_cfg.fc_encap_type = nla_get_u16(nla);
51ebd318 4346 }
6b9ea5a6 4347
68e2ffde 4348 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
acb54e3c 4349 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
8c5b83f0
RP		 4350 if (IS_ERR(rt)) {
4351 err = PTR_ERR(rt);
4352 rt = NULL;
6b9ea5a6 4353 goto cleanup;
8c5b83f0 4354 }
b5d2d75e
DA		 4355 if (!rt6_qualify_for_ecmp(rt)) {
4356 err = -EINVAL;
4357 NL_SET_ERR_MSG(extack,
4358 "Device only routes can not be added for IPv6 using the multipath API.");
4359 fib6_info_release(rt);
4360 goto cleanup;
4361 }
6b9ea5a6 4362
5e670d84 4363 rt->fib6_nh.nh_weight = rtnh->rtnh_hops + 1;
398958ae 4364
d4ead6b3
DA		 4365 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
4366 rt, &r_cfg);
51ebd318 4367 if (err) {
93531c67 4368 fib6_info_release(rt);
6b9ea5a6
RP		 4369 goto cleanup;
4370 }
4371
4372 rtnh = rtnh_next(rtnh, &remaining);
4373 }
4374
3b1137fe
DA		 4375 /* for add and replace send one notification with all nexthops.
4376 * Skip the notification in fib6_add_rt2node and send one with
4377 * the full route when done
4378 */
4379 info->skip_notify = 1;
4380
6b9ea5a6
RP		 4381 err_nh = NULL;
4382 list_for_each_entry(nh, &rt6_nh_list, next) {
8d1c802b
DA		 4383 err = __ip6_ins_rt(nh->fib6_info, info, extack);
4384 fib6_info_release(nh->fib6_info);
93531c67 4385
f7225172
DA		 4386 if (!err) {
4387 /* save reference to last route successfully inserted */
4388 rt_last = nh->fib6_info;
4389
4390 /* save reference to first route for notification */
4391 if (!rt_notif)
4392 rt_notif = nh->fib6_info;
4393 }
3b1137fe 4394
8d1c802b
DA		 4395 /* nh->fib6_info is used or freed at this point, reset to NULL*/
4396 nh->fib6_info = NULL;
6b9ea5a6
RP		 4397 if (err) {
4398 if (replace && nhn)
4399 ip6_print_replace_route_err(&rt6_nh_list);
4400 err_nh = nh;
4401 goto add_errout;
51ebd318 4402 }
6b9ea5a6 4403
1a72418b 4404 /* Because each route is added like a single route we remove
27596472
MK		 4405 * these flags after the first nexthop: if there is a collision,
4406 * we have already failed to add the first nexthop:
4407 * fib6_add_rt2node() has rejected it; when replacing, old
4408 * nexthops have been replaced by first new, the rest should
4409 * be added to it.
1a72418b 4410 */
27596472
MK		 4411 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
4412 NLM_F_REPLACE);
6b9ea5a6
RP		 4413 nhn++;
4414 }
4415
3b1137fe
DA		 4416 /* success ... tell user about new route */
4417 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
6b9ea5a6
RP		 4418 goto cleanup;
4419
4420add_errout:
3b1137fe
DA		 4421 /* send notification for routes that were added so that
4422 * the delete notifications sent by ip6_route_del are
4423 * coherent
4424 */
4425 if (rt_notif)
4426 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
4427
6b9ea5a6
RP		 4428 /* Delete routes that were already added */
4429 list_for_each_entry(nh, &rt6_nh_list, next) {
4430 if (err_nh == nh)
4431 break;
333c4301 4432 ip6_route_del(&nh->r_cfg, extack);
6b9ea5a6
RP		 4433 }
4434
4435cleanup:
4436 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
8d1c802b
DA		 4437 if (nh->fib6_info)
4438 fib6_info_release(nh->fib6_info);
6b9ea5a6
RP		 4439 list_del(&nh->next);
4440 kfree(nh);
4441 }
4442
4443 return err;
4444}
4445
333c4301
DA		 4446static int ip6_route_multipath_del(struct fib6_config *cfg,
4447 struct netlink_ext_ack *extack)
6b9ea5a6
RP		 4448{
4449 struct fib6_config r_cfg;
4450 struct rtnexthop *rtnh;
4451 int remaining;
4452 int attrlen;
4453 int err = 1, last_err = 0;
4454
4455 remaining = cfg->fc_mp_len;
4456 rtnh = (struct rtnexthop *)cfg->fc_mp;
4457
4458 /* Parse a Multipath Entry */
4459 while (rtnh_ok(rtnh, remaining)) {
4460 memcpy(&r_cfg, cfg, sizeof(*cfg));
4461 if (rtnh->rtnh_ifindex)
4462 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
4463
4464 attrlen = rtnh_attrlen(rtnh);
4465 if (attrlen > 0) {
4466 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
4467
4468 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
4469 if (nla) {
4470 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
4471 r_cfg.fc_flags |= RTF_GATEWAY;
4472 }
4473 }
333c4301 4474 err = ip6_route_del(&r_cfg, extack);
6b9ea5a6
RP		 4475 if (err)
4476 last_err = err;
4477
51ebd318
ND		 4478 rtnh = rtnh_next(rtnh, &remaining);
4479 }
4480
4481 return last_err;
4482}
4483
c21ef3e3
DA		 4484static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
4485 struct netlink_ext_ack *extack)
1da177e4 4486{
86872cb5
TG		 4487 struct fib6_config cfg;
4488 int err;
1da177e4 4489
333c4301 4490 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 4491 if (err < 0)
4492 return err;
4493
51ebd318 4494 if (cfg.fc_mp)
333c4301 4495 return ip6_route_multipath_del(&cfg, extack);
0ae81335
DA		 4496 else {
4497 cfg.fc_delete_all_nh = 1;
333c4301 4498 return ip6_route_del(&cfg, extack);
0ae81335 4499 }
1da177e4
LT		 4500}
4501
c21ef3e3
DA		 4502static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
4503 struct netlink_ext_ack *extack)
1da177e4 4504{
86872cb5
TG		 4505 struct fib6_config cfg;
4506 int err;
1da177e4 4507
333c4301 4508 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 4509 if (err < 0)
4510 return err;
4511
51ebd318 4512 if (cfg.fc_mp)
333c4301 4513 return ip6_route_multipath_add(&cfg, extack);
51ebd318 4514 else
acb54e3c 4515 return ip6_route_add(&cfg, GFP_KERNEL, extack);
1da177e4
LT		 4516}
4517
8d1c802b 4518static size_t rt6_nlmsg_size(struct fib6_info *rt)
339bf98f 4519{
beb1afac
DA		 4520 int nexthop_len = 0;
4521
93c2fb25 4522 if (rt->fib6_nsiblings) {
beb1afac
DA		 4523 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
4524 + NLA_ALIGN(sizeof(struct rtnexthop))
4525 + nla_total_size(16) /* RTA_GATEWAY */
5e670d84 4526 + lwtunnel_get_encap_size(rt->fib6_nh.nh_lwtstate);
beb1afac 4527
93c2fb25 4528 nexthop_len *= rt->fib6_nsiblings;
beb1afac
DA		 4529 }
4530
339bf98f
TG		 4531 return NLMSG_ALIGN(sizeof(struct rtmsg))
4532 + nla_total_size(16) /* RTA_SRC */
4533 + nla_total_size(16) /* RTA_DST */
4534 + nla_total_size(16) /* RTA_GATEWAY */
4535 + nla_total_size(16) /* RTA_PREFSRC */
4536 + nla_total_size(4) /* RTA_TABLE */
4537 + nla_total_size(4) /* RTA_IIF */
4538 + nla_total_size(4) /* RTA_OIF */
4539 + nla_total_size(4) /* RTA_PRIORITY */
6a2b9ce0 4540 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
ea697639 4541 + nla_total_size(sizeof(struct rta_cacheinfo))
c78ba6d6 4542 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
19e42e45 4543 + nla_total_size(1) /* RTA_PREF */
5e670d84 4544 + lwtunnel_get_encap_size(rt->fib6_nh.nh_lwtstate)
beb1afac
DA		 4545 + nexthop_len;
4546}
4547
8d1c802b 4548static int rt6_nexthop_info(struct sk_buff *skb, struct fib6_info *rt,
5be083ce 4549 unsigned int *flags, bool skip_oif)
beb1afac 4550{
5e670d84 4551 if (rt->fib6_nh.nh_flags & RTNH_F_DEAD)
f9d882ea
IS		 4552 *flags |= RTNH_F_DEAD;
4553
5e670d84 4554 if (rt->fib6_nh.nh_flags & RTNH_F_LINKDOWN) {
beb1afac 4555 *flags |= RTNH_F_LINKDOWN;
dcd1f572
DA		 4556
4557 rcu_read_lock();
4558 if (fib6_ignore_linkdown(rt))
beb1afac 4559 *flags |= RTNH_F_DEAD;
dcd1f572 4560 rcu_read_unlock();
beb1afac
DA		 4561 }
4562
93c2fb25 4563 if (rt->fib6_flags & RTF_GATEWAY) {
5e670d84 4564 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->fib6_nh.nh_gw) < 0)
beb1afac
DA		 4565 goto nla_put_failure;
4566 }
4567
5e670d84
DA		 4568 *flags |= (rt->fib6_nh.nh_flags & RTNH_F_ONLINK);
4569 if (rt->fib6_nh.nh_flags & RTNH_F_OFFLOAD)
61e4d01e
IS		 4570 *flags |= RTNH_F_OFFLOAD;
4571
5be083ce 4572 /* not needed for multipath encoding b/c it has a rtnexthop struct */
5e670d84
DA		 4573 if (!skip_oif && rt->fib6_nh.nh_dev &&
4574 nla_put_u32(skb, RTA_OIF, rt->fib6_nh.nh_dev->ifindex))
beb1afac
DA		 4575 goto nla_put_failure;
4576
5e670d84
DA		 4577 if (rt->fib6_nh.nh_lwtstate &&
4578 lwtunnel_fill_encap(skb, rt->fib6_nh.nh_lwtstate) < 0)
beb1afac
DA		 4579 goto nla_put_failure;
4580
4581 return 0;
4582
4583nla_put_failure:
4584 return -EMSGSIZE;
4585}
4586
5be083ce 4587/* add multipath next hop */
8d1c802b 4588static int rt6_add_nexthop(struct sk_buff *skb, struct fib6_info *rt)
beb1afac 4589{
5e670d84 4590 const struct net_device *dev = rt->fib6_nh.nh_dev;
beb1afac
DA		 4591 struct rtnexthop *rtnh;
4592 unsigned int flags = 0;
4593
4594 rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
4595 if (!rtnh)
4596 goto nla_put_failure;
4597
5e670d84
DA		 4598 rtnh->rtnh_hops = rt->fib6_nh.nh_weight - 1;
4599 rtnh->rtnh_ifindex = dev ? dev->ifindex : 0;
beb1afac 4600
5be083ce 4601 if (rt6_nexthop_info(skb, rt, &flags, true) < 0)
beb1afac
DA		 4602 goto nla_put_failure;
4603
4604 rtnh->rtnh_flags = flags;
4605
4606 /* length of rtnetlink header + attributes */
4607 rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
4608
4609 return 0;
4610
4611nla_put_failure:
4612 return -EMSGSIZE;
339bf98f
TG		 4613}
4614
d4ead6b3 4615static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 4616 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 4617 struct in6_addr *dest, struct in6_addr *src,
15e47304 4618 int iif, int type, u32 portid, u32 seq,
f8cfe2ce 4619 unsigned int flags)
1da177e4 4620{
22d0bd82
XL		 4621 struct rt6_info *rt6 = (struct rt6_info *)dst;
4622 struct rt6key *rt6_dst, *rt6_src;
4623 u32 *pmetrics, table, rt6_flags;
2d7202bf 4624 struct nlmsghdr *nlh;
22d0bd82 4625 struct rtmsg *rtm;
d4ead6b3 4626 long expires = 0;
1da177e4 4627
15e47304 4628 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
38308473 4629 if (!nlh)
26932566 4630 return -EMSGSIZE;
2d7202bf 4631
22d0bd82
XL		 4632 if (rt6) {
4633 rt6_dst = &rt6->rt6i_dst;
4634 rt6_src = &rt6->rt6i_src;
4635 rt6_flags = rt6->rt6i_flags;
4636 } else {
4637 rt6_dst = &rt->fib6_dst;
4638 rt6_src = &rt->fib6_src;
4639 rt6_flags = rt->fib6_flags;
4640 }
4641
2d7202bf 4642 rtm = nlmsg_data(nlh);
1da177e4 4643 rtm->rtm_family = AF_INET6;
22d0bd82
XL		 4644 rtm->rtm_dst_len = rt6_dst->plen;
4645 rtm->rtm_src_len = rt6_src->plen;
1da177e4 4646 rtm->rtm_tos = 0;
93c2fb25
DA		 4647 if (rt->fib6_table)
4648 table = rt->fib6_table->tb6_id;
c71099ac 4649 else
9e762a4a
PM		 4650 table = RT6_TABLE_UNSPEC;
4651 rtm->rtm_table = table;
c78679e8
DM		 4652 if (nla_put_u32(skb, RTA_TABLE, table))
4653 goto nla_put_failure;
e8478e80
DA		 4654
4655 rtm->rtm_type = rt->fib6_type;
1da177e4
LT		 4656 rtm->rtm_flags = 0;
4657 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
93c2fb25 4658 rtm->rtm_protocol = rt->fib6_protocol;
1da177e4 4659
22d0bd82 4660 if (rt6_flags & RTF_CACHE)
1da177e4
LT		 4661 rtm->rtm_flags |= RTM_F_CLONED;
4662
d4ead6b3
DA		 4663 if (dest) {
4664 if (nla_put_in6_addr(skb, RTA_DST, dest))
c78679e8 4665 goto nla_put_failure;
1ab1457c 4666 rtm->rtm_dst_len = 128;
1da177e4 4667 } else if (rtm->rtm_dst_len)
22d0bd82 4668 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
c78679e8 4669 goto nla_put_failure;
1da177e4
LT		 4670#ifdef CONFIG_IPV6_SUBTREES
4671 if (src) {
930345ea 4672 if (nla_put_in6_addr(skb, RTA_SRC, src))
c78679e8 4673 goto nla_put_failure;
1ab1457c 4674 rtm->rtm_src_len = 128;
c78679e8 4675 } else if (rtm->rtm_src_len &&
22d0bd82 4676 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
c78679e8 4677 goto nla_put_failure;
1da177e4 4678#endif
7bc570c8
YH		 4679 if (iif) {
4680#ifdef CONFIG_IPV6_MROUTE
22d0bd82 4681 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
fd61c6ba
DA		 4682 int err = ip6mr_get_route(net, skb, rtm, portid);
4683
4684 if (err == 0)
4685 return 0;
4686 if (err < 0)
4687 goto nla_put_failure;
7bc570c8
YH		 4688 } else
4689#endif
c78679e8
DM		 4690 if (nla_put_u32(skb, RTA_IIF, iif))
4691 goto nla_put_failure;
d4ead6b3 4692 } else if (dest) {
1da177e4 4693 struct in6_addr saddr_buf;
d4ead6b3 4694 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
930345ea 4695 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 4696 goto nla_put_failure;
1da177e4 4697 }
2d7202bf 4698
93c2fb25 4699 if (rt->fib6_prefsrc.plen) {
c3968a85 4700 struct in6_addr saddr_buf;
93c2fb25 4701 saddr_buf = rt->fib6_prefsrc.addr;
930345ea 4702 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 4703 goto nla_put_failure;
c3968a85
DW		 4704 }
4705
d4ead6b3
DA		 4706 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
4707 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
2d7202bf
TG		 4708 goto nla_put_failure;
4709
93c2fb25 4710 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
c78679e8 4711 goto nla_put_failure;
8253947e 4712
beb1afac
DA		 4713 /* For multipath routes, walk the siblings list and add
4714 * each as a nexthop within RTA_MULTIPATH.
4715 */
22d0bd82
XL		 4716 if (rt6) {
4717 if (rt6_flags & RTF_GATEWAY &&
4718 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
4719 goto nla_put_failure;
4720
4721 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
4722 goto nla_put_failure;
4723 } else if (rt->fib6_nsiblings) {
8d1c802b 4724 struct fib6_info *sibling, *next_sibling;
beb1afac
DA		 4725 struct nlattr *mp;
4726
4727 mp = nla_nest_start(skb, RTA_MULTIPATH);
4728 if (!mp)
4729 goto nla_put_failure;
4730
4731 if (rt6_add_nexthop(skb, rt) < 0)
4732 goto nla_put_failure;
4733
4734 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25 4735 &rt->fib6_siblings, fib6_siblings) {
beb1afac
DA		 4736 if (rt6_add_nexthop(skb, sibling) < 0)
4737 goto nla_put_failure;
4738 }
4739
4740 nla_nest_end(skb, mp);
4741 } else {
5be083ce 4742 if (rt6_nexthop_info(skb, rt, &rtm->rtm_flags, false) < 0)
beb1afac
DA		 4743 goto nla_put_failure;
4744 }
4745
22d0bd82 4746 if (rt6_flags & RTF_EXPIRES) {
14895687
DA		 4747 expires = dst ? dst->expires : rt->expires;
4748 expires -= jiffies;
4749 }
69cdf8f9 4750
d4ead6b3 4751 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
e3703b3d 4752 goto nla_put_failure;
2d7202bf 4753
22d0bd82 4754 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
c78ba6d6
LR		 4755 goto nla_put_failure;
4756
19e42e45 4757
053c095a
JB		 4758 nlmsg_end(skb, nlh);
4759 return 0;
2d7202bf
TG		 4760
4761nla_put_failure:
26932566
PM		 4762 nlmsg_cancel(skb, nlh);
4763 return -EMSGSIZE;
1da177e4
LT		 4764}
4765
8d1c802b 4766int rt6_dump_route(struct fib6_info *rt, void *p_arg)
1da177e4
LT		 4767{
4768 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
1f17e2f2
DA		 4769 struct net *net = arg->net;
4770
421842ed 4771 if (rt == net->ipv6.fib6_null_entry)
1f17e2f2 4772 return 0;
1da177e4 4773
2d7202bf
TG		 4774 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
4775 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
f8cfe2ce
DA		 4776
4777 /* user wants prefix routes only */
4778 if (rtm->rtm_flags & RTM_F_PREFIX &&
93c2fb25 4779 !(rt->fib6_flags & RTF_PREFIX_RT)) {
f8cfe2ce
DA		 4780 /* success since this is not a prefix route */
4781 return 1;
4782 }
4783 }
1da177e4 4784
d4ead6b3
DA		 4785 return rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL, 0,
4786 RTM_NEWROUTE, NETLINK_CB(arg->cb->skb).portid,
4787 arg->cb->nlh->nlmsg_seq, NLM_F_MULTI);
1da177e4
LT		 4788}
4789
c21ef3e3
DA		 4790static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4791 struct netlink_ext_ack *extack)
1da177e4 4792{
3b1e0a65 4793 struct net *net = sock_net(in_skb->sk);
ab364a6f 4794 struct nlattr *tb[RTA_MAX+1];
18c3a61c 4795 int err, iif = 0, oif = 0;
a68886a6 4796 struct fib6_info *from;
18c3a61c 4797 struct dst_entry *dst;
ab364a6f 4798 struct rt6_info *rt;
1da177e4 4799 struct sk_buff *skb;
ab364a6f 4800 struct rtmsg *rtm;
744486d4 4801 struct flowi6 fl6 = {};
18c3a61c 4802 bool fibmatch;
1da177e4 4803
fceb6435 4804 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
c21ef3e3 4805 extack);
ab364a6f
TG		 4806 if (err < 0)
4807 goto errout;
1da177e4 4808
ab364a6f 4809 err = -EINVAL;
38b7097b
HFS		 4810 rtm = nlmsg_data(nlh);
4811 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
18c3a61c 4812 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
1da177e4 4813
ab364a6f
TG		 4814 if (tb[RTA_SRC]) {
4815 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
4816 goto errout;
4817
4e3fd7a0 4818 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
ab364a6f
TG		 4819 }
4820
4821 if (tb[RTA_DST]) {
4822 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
4823 goto errout;
4824
4e3fd7a0 4825 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
ab364a6f
TG		 4826 }
4827
4828 if (tb[RTA_IIF])
4829 iif = nla_get_u32(tb[RTA_IIF]);
4830
4831 if (tb[RTA_OIF])
72331bc0 4832 oif = nla_get_u32(tb[RTA_OIF]);
1da177e4 4833
2e47b291
LC		 4834 if (tb[RTA_MARK])
4835 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
4836
622ec2c9
LC		 4837 if (tb[RTA_UID])
4838 fl6.flowi6_uid = make_kuid(current_user_ns(),
4839 nla_get_u32(tb[RTA_UID]));
4840 else
4841 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
4842
eacb9384
RP		 4843 if (tb[RTA_SPORT])
4844 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
4845
4846 if (tb[RTA_DPORT])
4847 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
4848
4849 if (tb[RTA_IP_PROTO]) {
4850 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
4851 &fl6.flowi6_proto, extack);
4852 if (err)
4853 goto errout;
4854 }
4855
1da177e4
LT		 4856 if (iif) {
4857 struct net_device *dev;
72331bc0
SL		 4858 int flags = 0;
4859
121622db
FW		 4860 rcu_read_lock();
4861
4862 dev = dev_get_by_index_rcu(net, iif);
1da177e4 4863 if (!dev) {
121622db 4864 rcu_read_unlock();
1da177e4 4865 err = -ENODEV;
ab364a6f 4866 goto errout;
1da177e4 4867 }
72331bc0
SL		 4868
4869 fl6.flowi6_iif = iif;
4870
4871 if (!ipv6_addr_any(&fl6.saddr))
4872 flags |= RT6_LOOKUP_F_HAS_SADDR;
4873
b75cc8f9 4874 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
121622db
FW		 4875
4876 rcu_read_unlock();
72331bc0
SL		 4877 } else {
4878 fl6.flowi6_oif = oif;
4879
58acfd71 4880 dst = ip6_route_output(net, NULL, &fl6);
18c3a61c
RP		 4881 }
4882
18c3a61c
RP		 4883
4884 rt = container_of(dst, struct rt6_info, dst);
4885 if (rt->dst.error) {
4886 err = rt->dst.error;
4887 ip6_rt_put(rt);
4888 goto errout;
1da177e4
LT		 4889 }
4890
9d6acb3b
WC		 4891 if (rt == net->ipv6.ip6_null_entry) {
4892 err = rt->dst.error;
4893 ip6_rt_put(rt);
4894 goto errout;
4895 }
4896
ab364a6f 4897 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
38308473 4898 if (!skb) {
94e187c0 4899 ip6_rt_put(rt);
ab364a6f
TG		 4900 err = -ENOBUFS;
4901 goto errout;
4902 }
1da177e4 4903
d8d1f30b 4904 skb_dst_set(skb, &rt->dst);
a68886a6
DA		 4905
4906 rcu_read_lock();
4907 from = rcu_dereference(rt->from);
4908
18c3a61c 4909 if (fibmatch)
a68886a6 4910 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL, iif,
18c3a61c
RP		 4911 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
4912 nlh->nlmsg_seq, 0);
4913 else
a68886a6
DA		 4914 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
4915 &fl6.saddr, iif, RTM_NEWROUTE,
d4ead6b3
DA		 4916 NETLINK_CB(in_skb).portid, nlh->nlmsg_seq,
4917 0);
a68886a6
DA		 4918 rcu_read_unlock();
4919
1da177e4 4920 if (err < 0) {
ab364a6f
TG		 4921 kfree_skb(skb);
4922 goto errout;
1da177e4
LT		 4923 }
4924
15e47304 4925 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
ab364a6f 4926errout:
1da177e4 4927 return err;
1da177e4
LT		 4928}
4929
8d1c802b 4930void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
37a1d361 4931 unsigned int nlm_flags)
1da177e4
LT		 4932{
4933 struct sk_buff *skb;
5578689a 4934 struct net *net = info->nl_net;
528c4ceb
DL		 4935 u32 seq;
4936 int err;
4937
4938 err = -ENOBUFS;
38308473 4939 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
86872cb5 4940
19e42e45 4941 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
38308473 4942 if (!skb)
21713ebc
TG		 4943 goto errout;
4944
d4ead6b3
DA		 4945 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
4946 event, info->portid, seq, nlm_flags);
26932566
PM		 4947 if (err < 0) {
4948 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
4949 WARN_ON(err == -EMSGSIZE);
4950 kfree_skb(skb);
4951 goto errout;
4952 }
15e47304 4953 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
1ce85fe4
PNA		 4954 info->nlh, gfp_any());
4955 return;
21713ebc
TG		 4956errout:
4957 if (err < 0)
5578689a 4958 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
1da177e4
LT		 4959}
4960
8ed67789 4961static int ip6_route_dev_notify(struct notifier_block *this,
351638e7 4962 unsigned long event, void *ptr)
8ed67789 4963{
351638e7 4964 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
c346dca1 4965 struct net *net = dev_net(dev);
8ed67789 4966
242d3a49
WC		 4967 if (!(dev->flags & IFF_LOOPBACK))
4968 return NOTIFY_OK;
4969
4970 if (event == NETDEV_REGISTER) {
421842ed 4971 net->ipv6.fib6_null_entry->fib6_nh.nh_dev = dev;
d8d1f30b 4972 net->ipv6.ip6_null_entry->dst.dev = dev;
8ed67789
DL		 4973 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
4974#ifdef CONFIG_IPV6_MULTIPLE_TABLES
d8d1f30b 4975 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
8ed67789 4976 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
d8d1f30b 4977 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
8ed67789 4978 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
242d3a49 4979#endif
76da0704
WC		 4980 } else if (event == NETDEV_UNREGISTER &&
4981 dev->reg_state != NETREG_UNREGISTERED) {
4982 /* NETDEV_UNREGISTER could be fired for multiple times by
4983 * netdev_wait_allrefs(). Make sure we only call this once.
4984 */
12d94a80 4985 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
242d3a49 4986#ifdef CONFIG_IPV6_MULTIPLE_TABLES
12d94a80
ED		 4987 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
4988 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
8ed67789
DL		 4989#endif
4990 }
4991
4992 return NOTIFY_OK;
4993}
4994
1da177e4
LT		 4995/*
4996 * /proc
4997 */
4998
4999#ifdef CONFIG_PROC_FS
1da177e4
LT		 5000static int rt6_stats_seq_show(struct seq_file *seq, void *v)
5001{
69ddb805 5002 struct net *net = (struct net *)seq->private;
1da177e4 5003 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
69ddb805
DL		 5004 net->ipv6.rt6_stats->fib_nodes,
5005 net->ipv6.rt6_stats->fib_route_nodes,
81eb8447 5006 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
69ddb805
DL		 5007 net->ipv6.rt6_stats->fib_rt_entries,
5008 net->ipv6.rt6_stats->fib_rt_cache,
fc66f95c 5009 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
69ddb805 5010 net->ipv6.rt6_stats->fib_discarded_routes);
1da177e4
LT		 5011
5012 return 0;
5013}
1da177e4
LT		 5014#endif /* CONFIG_PROC_FS */
5015
5016#ifdef CONFIG_SYSCTL
5017
1da177e4 5018static
fe2c6338 5019int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
1da177e4
LT		 5020 void __user *buffer, size_t *lenp, loff_t *ppos)
5021{
c486da34
LAG		 5022 struct net *net;
5023 int delay;
5024 if (!write)
1da177e4 5025 return -EINVAL;
c486da34
LAG		 5026
5027 net = (struct net *)ctl->extra1;
5028 delay = net->ipv6.sysctl.flush_delay;
5029 proc_dointvec(ctl, write, buffer, lenp, ppos);
2ac3ac8f 5030 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
c486da34 5031 return 0;
1da177e4
LT		 5032}
5033
ed792e28 5034static struct ctl_table ipv6_route_table_template[] = {
1ab1457c 5035 {
1da177e4 5036 .procname = "flush",
4990509f 5037 .data = &init_net.ipv6.sysctl.flush_delay,
1da177e4 5038 .maxlen = sizeof(int),
89c8b3a1 5039 .mode = 0200,
6d9f239a 5040 .proc_handler = ipv6_sysctl_rtcache_flush
1da177e4
LT		 5041 },
5042 {
1da177e4 5043 .procname = "gc_thresh",
9a7ec3a9 5044 .data = &ip6_dst_ops_template.gc_thresh,
1da177e4
LT		 5045 .maxlen = sizeof(int),
5046 .mode = 0644,
6d9f239a 5047 .proc_handler = proc_dointvec,
1da177e4
LT		 5048 },
5049 {
1da177e4 5050 .procname = "max_size",
4990509f 5051 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
1da177e4
LT		 5052 .maxlen = sizeof(int),
5053 .mode = 0644,
6d9f239a 5054 .proc_handler = proc_dointvec,
1da177e4
LT		 5055 },
5056 {
1da177e4 5057 .procname = "gc_min_interval",
4990509f 5058 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 5059 .maxlen = sizeof(int),
5060 .mode = 0644,
6d9f239a 5061 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5062 },
5063 {
1da177e4 5064 .procname = "gc_timeout",
4990509f 5065 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
1da177e4
LT		 5066 .maxlen = sizeof(int),
5067 .mode = 0644,
6d9f239a 5068 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5069 },
5070 {
1da177e4 5071 .procname = "gc_interval",
4990509f 5072 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
1da177e4
LT		 5073 .maxlen = sizeof(int),
5074 .mode = 0644,
6d9f239a 5075 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5076 },
5077 {
1da177e4 5078 .procname = "gc_elasticity",
4990509f 5079 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
1da177e4
LT		 5080 .maxlen = sizeof(int),
5081 .mode = 0644,
f3d3f616 5082 .proc_handler = proc_dointvec,
1da177e4
LT		 5083 },
5084 {
1da177e4 5085 .procname = "mtu_expires",
4990509f 5086 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
1da177e4
LT		 5087 .maxlen = sizeof(int),
5088 .mode = 0644,
6d9f239a 5089 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5090 },
5091 {
1da177e4 5092 .procname = "min_adv_mss",
4990509f 5093 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
1da177e4
LT		 5094 .maxlen = sizeof(int),
5095 .mode = 0644,
f3d3f616 5096 .proc_handler = proc_dointvec,
1da177e4
LT		 5097 },
5098 {
1da177e4 5099 .procname = "gc_min_interval_ms",
4990509f 5100 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 5101 .maxlen = sizeof(int),
5102 .mode = 0644,
6d9f239a 5103 .proc_handler = proc_dointvec_ms_jiffies,
1da177e4 5104 },
f8572d8f 5105 { }
1da177e4
LT		 5106};
5107
2c8c1e72 5108struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
760f2d01
DL		 5109{
5110 struct ctl_table *table;
5111
5112 table = kmemdup(ipv6_route_table_template,
5113 sizeof(ipv6_route_table_template),
5114 GFP_KERNEL);
5ee09105
YH		 5115
5116 if (table) {
5117 table[0].data = &net->ipv6.sysctl.flush_delay;
c486da34 5118 table[0].extra1 = net;
86393e52 5119 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
5ee09105
YH		 5120 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
5121 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
5122 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
5123 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
5124 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
5125 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
5126 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
9c69fabe 5127 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
464dc801
EB		 5128
5129 /* Don't export sysctls to unprivileged users */
5130 if (net->user_ns != &init_user_ns)
5131 table[0].procname = NULL;
5ee09105
YH		 5132 }
5133
760f2d01
DL		 5134 return table;
5135}
1da177e4
LT		 5136#endif
5137
2c8c1e72 5138static int __net_init ip6_route_net_init(struct net *net)
cdb18761 5139{
633d424b 5140 int ret = -ENOMEM;
8ed67789 5141
86393e52
AD		 5142 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
5143 sizeof(net->ipv6.ip6_dst_ops));
f2fc6a54 5144
fc66f95c
ED		 5145 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
5146 goto out_ip6_dst_ops;
5147
421842ed
DA		 5148 net->ipv6.fib6_null_entry = kmemdup(&fib6_null_entry_template,
5149 sizeof(*net->ipv6.fib6_null_entry),
5150 GFP_KERNEL);
5151 if (!net->ipv6.fib6_null_entry)
5152 goto out_ip6_dst_entries;
5153
8ed67789
DL		 5154 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
5155 sizeof(*net->ipv6.ip6_null_entry),
5156 GFP_KERNEL);
5157 if (!net->ipv6.ip6_null_entry)
421842ed 5158 goto out_fib6_null_entry;
d8d1f30b 5159 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 5160 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
5161 ip6_template_metrics, true);
8ed67789
DL		 5162
5163#ifdef CONFIG_IPV6_MULTIPLE_TABLES
feca7d8c 5164 net->ipv6.fib6_has_custom_rules = false;
8ed67789
DL		 5165 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
5166 sizeof(*net->ipv6.ip6_prohibit_entry),
5167 GFP_KERNEL);
68fffc67
PZ		 5168 if (!net->ipv6.ip6_prohibit_entry)
5169 goto out_ip6_null_entry;
d8d1f30b 5170 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 5171 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
5172 ip6_template_metrics, true);
8ed67789
DL		 5173
5174 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
5175 sizeof(*net->ipv6.ip6_blk_hole_entry),
5176 GFP_KERNEL);
68fffc67
PZ		 5177 if (!net->ipv6.ip6_blk_hole_entry)
5178 goto out_ip6_prohibit_entry;
d8d1f30b 5179 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 5180 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
5181 ip6_template_metrics, true);
8ed67789
DL		 5182#endif
5183
b339a47c
PZ		 5184 net->ipv6.sysctl.flush_delay = 0;
5185 net->ipv6.sysctl.ip6_rt_max_size = 4096;
5186 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
5187 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
5188 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
5189 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
5190 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
5191 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
5192
6891a346
BT		 5193 net->ipv6.ip6_rt_gc_expire = 30*HZ;
5194
8ed67789
DL		 5195 ret = 0;
5196out:
5197 return ret;
f2fc6a54 5198
68fffc67
PZ		 5199#ifdef CONFIG_IPV6_MULTIPLE_TABLES
5200out_ip6_prohibit_entry:
5201 kfree(net->ipv6.ip6_prohibit_entry);
5202out_ip6_null_entry:
5203 kfree(net->ipv6.ip6_null_entry);
5204#endif
421842ed
DA		 5205out_fib6_null_entry:
5206 kfree(net->ipv6.fib6_null_entry);
fc66f95c
ED		 5207out_ip6_dst_entries:
5208 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
f2fc6a54 5209out_ip6_dst_ops:
f2fc6a54 5210 goto out;
cdb18761
DL		 5211}
5212
2c8c1e72 5213static void __net_exit ip6_route_net_exit(struct net *net)
cdb18761 5214{
421842ed 5215 kfree(net->ipv6.fib6_null_entry);
8ed67789
DL		 5216 kfree(net->ipv6.ip6_null_entry);
5217#ifdef CONFIG_IPV6_MULTIPLE_TABLES
5218 kfree(net->ipv6.ip6_prohibit_entry);
5219 kfree(net->ipv6.ip6_blk_hole_entry);
5220#endif
41bb78b4 5221 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
cdb18761
DL		 5222}
5223
d189634e
TG		 5224static int __net_init ip6_route_net_init_late(struct net *net)
5225{
5226#ifdef CONFIG_PROC_FS
c3506372
CH		 5227 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
5228 sizeof(struct ipv6_route_iter));
3617d949
CH		 5229 proc_create_net_single("rt6_stats", 0444, net->proc_net,
5230 rt6_stats_seq_show, NULL);
d189634e
TG		 5231#endif
5232 return 0;
5233}
5234
5235static void __net_exit ip6_route_net_exit_late(struct net *net)
5236{
5237#ifdef CONFIG_PROC_FS
ece31ffd
G		 5238 remove_proc_entry("ipv6_route", net->proc_net);
5239 remove_proc_entry("rt6_stats", net->proc_net);
d189634e
TG		 5240#endif
5241}
5242
cdb18761
DL		 5243static struct pernet_operations ip6_route_net_ops = {
5244 .init = ip6_route_net_init,
5245 .exit = ip6_route_net_exit,
5246};
5247
c3426b47
DM		 5248static int __net_init ipv6_inetpeer_init(struct net *net)
5249{
5250 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
5251
5252 if (!bp)
5253 return -ENOMEM;
5254 inet_peer_base_init(bp);
5255 net->ipv6.peers = bp;
5256 return 0;
5257}
5258
5259static void __net_exit ipv6_inetpeer_exit(struct net *net)
5260{
5261 struct inet_peer_base *bp = net->ipv6.peers;
5262
5263 net->ipv6.peers = NULL;
56a6b248 5264 inetpeer_invalidate_tree(bp);
c3426b47
DM		 5265 kfree(bp);
5266}
5267
2b823f72 5268static struct pernet_operations ipv6_inetpeer_ops = {
c3426b47
DM		 5269 .init = ipv6_inetpeer_init,
5270 .exit = ipv6_inetpeer_exit,
5271};
5272
d189634e
TG		 5273static struct pernet_operations ip6_route_net_late_ops = {
5274 .init = ip6_route_net_init_late,
5275 .exit = ip6_route_net_exit_late,
5276};
5277
8ed67789
DL		 5278static struct notifier_block ip6_route_dev_notifier = {
5279 .notifier_call = ip6_route_dev_notify,
242d3a49 5280 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
8ed67789
DL		 5281};
5282
2f460933
WC		 5283void __init ip6_route_init_special_entries(void)
5284{
5285 /* Registering of the loopback is done before this portion of code,
5286 * the loopback reference in rt6_info will not be taken, do it
5287 * manually for init_net */
421842ed 5288 init_net.ipv6.fib6_null_entry->fib6_nh.nh_dev = init_net.loopback_dev;
2f460933
WC		 5289 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
5290 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
5291 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
5292 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
5293 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
5294 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
5295 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
5296 #endif
5297}
5298
433d49c3 5299int __init ip6_route_init(void)
1da177e4 5300{
433d49c3 5301 int ret;
8d0b94af 5302 int cpu;
433d49c3 5303
9a7ec3a9
DL		 5304 ret = -ENOMEM;
5305 ip6_dst_ops_template.kmem_cachep =
e5d679f3 5306 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
f845ab6b 5307 SLAB_HWCACHE_ALIGN, NULL);
9a7ec3a9 5308 if (!ip6_dst_ops_template.kmem_cachep)
c19a28e1 5309 goto out;
14e50e57 5310
fc66f95c 5311 ret = dst_entries_init(&ip6_dst_blackhole_ops);
8ed67789 5312 if (ret)
bdb3289f 5313 goto out_kmem_cache;
bdb3289f 5314
c3426b47
DM		 5315 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
5316 if (ret)
e8803b6c 5317 goto out_dst_entries;
2a0c451a 5318
7e52b33b
DM		 5319 ret = register_pernet_subsys(&ip6_route_net_ops);
5320 if (ret)
5321 goto out_register_inetpeer;
c3426b47 5322
5dc121e9
AE		 5323 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
5324
e8803b6c 5325 ret = fib6_init();
433d49c3 5326 if (ret)
8ed67789 5327 goto out_register_subsys;
433d49c3 5328
433d49c3
DL		 5329 ret = xfrm6_init();
5330 if (ret)
e8803b6c 5331 goto out_fib6_init;
c35b7e72 5332
433d49c3
DL		 5333 ret = fib6_rules_init();
5334 if (ret)
5335 goto xfrm6_init;
7e5449c2 5336
d189634e
TG		 5337 ret = register_pernet_subsys(&ip6_route_net_late_ops);
5338 if (ret)
5339 goto fib6_rules_init;
5340
16feebcf
FW		 5341 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
5342 inet6_rtm_newroute, NULL, 0);
5343 if (ret < 0)
5344 goto out_register_late_subsys;
5345
5346 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
5347 inet6_rtm_delroute, NULL, 0);
5348 if (ret < 0)
5349 goto out_register_late_subsys;
5350
5351 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
5352 inet6_rtm_getroute, NULL,
5353 RTNL_FLAG_DOIT_UNLOCKED);
5354 if (ret < 0)
d189634e 5355 goto out_register_late_subsys;
c127ea2c 5356
8ed67789 5357 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
cdb18761 5358 if (ret)
d189634e 5359 goto out_register_late_subsys;
8ed67789 5360
8d0b94af
MKL		 5361 for_each_possible_cpu(cpu) {
5362 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
5363
5364 INIT_LIST_HEAD(&ul->head);
5365 spin_lock_init(&ul->lock);
5366 }
5367
433d49c3
DL		 5368out:
5369 return ret;
5370
d189634e 5371out_register_late_subsys:
16feebcf 5372 rtnl_unregister_all(PF_INET6);
d189634e 5373 unregister_pernet_subsys(&ip6_route_net_late_ops);
433d49c3 5374fib6_rules_init:
433d49c3
DL		 5375 fib6_rules_cleanup();
5376xfrm6_init:
433d49c3 5377 xfrm6_fini();
2a0c451a
TG		 5378out_fib6_init:
5379 fib6_gc_cleanup();
8ed67789
DL		 5380out_register_subsys:
5381 unregister_pernet_subsys(&ip6_route_net_ops);
7e52b33b
DM		 5382out_register_inetpeer:
5383 unregister_pernet_subsys(&ipv6_inetpeer_ops);
fc66f95c
ED		 5384out_dst_entries:
5385 dst_entries_destroy(&ip6_dst_blackhole_ops);
433d49c3 5386out_kmem_cache:
f2fc6a54 5387 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
433d49c3 5388 goto out;
1da177e4
LT		 5389}
5390
5391void ip6_route_cleanup(void)
5392{
8ed67789 5393 unregister_netdevice_notifier(&ip6_route_dev_notifier);
d189634e 5394 unregister_pernet_subsys(&ip6_route_net_late_ops);
101367c2 5395 fib6_rules_cleanup();
1da177e4 5396 xfrm6_fini();
1da177e4 5397 fib6_gc_cleanup();
c3426b47 5398 unregister_pernet_subsys(&ipv6_inetpeer_ops);
8ed67789 5399 unregister_pernet_subsys(&ip6_route_net_ops);
41bb78b4 5400 dst_entries_destroy(&ip6_dst_blackhole_ops);
f2fc6a54 5401 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
1da177e4 5402}
Linux 6.x block layer and io_uring tree(s)