projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branch '1GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next...
[linux-block.git] / net / ipv6 / route.c
CommitLineData
2874c5fd 1// SPDX-License-Identifier: GPL-2.0-or-later
1da177e4
LT		 2/*
3 * Linux INET6 implementation
4 * FIB front-end.
5 *
6 * Authors:
1ab1457c 7 * Pedro Roque <roque@di.fc.ul.pt>
1da177e4
LT		 8 */
9
10/* Changes:
11 *
12 * YOSHIFUJI Hideaki @USAGI
13 * reworked default router selection.
14 * - respect outgoing interface
15 * - select from (probably) reachable routers (i.e.
16 * routers in REACHABLE, STALE, DELAY or PROBE states).
17 * - always select the same router if it is (probably)
18 * reachable. otherwise, round-robin the list.
c0bece9f
YH		 19 * Ville Nuorvala
20 * Fixed routing subtrees.
1da177e4
LT		 21 */
22
f3213831
JP		 23#define pr_fmt(fmt) "IPv6: " fmt
24
4fc268d2 25#include <linux/capability.h>
1da177e4 26#include <linux/errno.h>
bc3b2d7f 27#include <linux/export.h>
1da177e4
LT		 28#include <linux/types.h>
29#include <linux/times.h>
30#include <linux/socket.h>
31#include <linux/sockios.h>
32#include <linux/net.h>
33#include <linux/route.h>
34#include <linux/netdevice.h>
35#include <linux/in6.h>
7bc570c8 36#include <linux/mroute6.h>
1da177e4 37#include <linux/init.h>
1da177e4 38#include <linux/if_arp.h>
1da177e4
LT		 39#include <linux/proc_fs.h>
40#include <linux/seq_file.h>
5b7c931d 41#include <linux/nsproxy.h>
5a0e3ad6 42#include <linux/slab.h>
35732d01 43#include <linux/jhash.h>
457c4cbc 44#include <net/net_namespace.h>
1da177e4
LT		 45#include <net/snmp.h>
46#include <net/ipv6.h>
47#include <net/ip6_fib.h>
48#include <net/ip6_route.h>
49#include <net/ndisc.h>
50#include <net/addrconf.h>
51#include <net/tcp.h>
52#include <linux/rtnetlink.h>
53#include <net/dst.h>
904af04d 54#include <net/dst_metadata.h>
1da177e4 55#include <net/xfrm.h>
8d71740c 56#include <net/netevent.h>
21713ebc 57#include <net/netlink.h>
3c618c1d 58#include <net/rtnh.h>
19e42e45 59#include <net/lwtunnel.h>
904af04d 60#include <net/ip_tunnels.h>
ca254490 61#include <net/l3mdev.h>
eacb9384 62#include <net/ip.h>
7c0f6ba6 63#include <linux/uaccess.h>
1da177e4
LT		 64
65#ifdef CONFIG_SYSCTL
66#include <linux/sysctl.h>
67#endif
68
30d444d3
DA		 69static int ip6_rt_type_to_error(u8 fib6_type);
70
71#define CREATE_TRACE_POINTS
72#include <trace/events/fib6.h>
73EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
74#undef CREATE_TRACE_POINTS
75
afc154e9 76enum rt6_nud_state {
7e980569
JB		 77 RT6_NUD_FAIL_HARD = -3,
78 RT6_NUD_FAIL_PROBE = -2,
79 RT6_NUD_FAIL_DO_RR = -1,
afc154e9
HFS		 80 RT6_NUD_SUCCEED = 1
81};
82
1da177e4 83static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
0dbaee3b 84static unsigned int ip6_default_advmss(const struct dst_entry *dst);
ebb762f2 85static unsigned int ip6_mtu(const struct dst_entry *dst);
1da177e4
LT		 86static struct dst_entry *ip6_negative_advice(struct dst_entry *);
87static void ip6_dst_destroy(struct dst_entry *);
88static void ip6_dst_ifdown(struct dst_entry *,
89 struct net_device *dev, int how);
569d3645 90static int ip6_dst_gc(struct dst_ops *ops);
1da177e4
LT		 91
92static int ip6_pkt_discard(struct sk_buff *skb);
ede2059d 93static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
7150aede 94static int ip6_pkt_prohibit(struct sk_buff *skb);
ede2059d 95static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1da177e4 96static void ip6_link_failure(struct sk_buff *skb);
6700c270 97static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
bd085ef6
HL		 98 struct sk_buff *skb, u32 mtu,
99 bool confirm_neigh);
6700c270
DM		 100static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
101 struct sk_buff *skb);
702cea56
DA		 102static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
103 int strict);
a1b7a1f0 104static size_t rt6_nlmsg_size(struct fib6_info *f6i);
d4ead6b3 105static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 106 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 107 struct in6_addr *dest, struct in6_addr *src,
16a16cd3
DA		 108 int iif, int type, u32 portid, u32 seq,
109 unsigned int flags);
7e4b5128 110static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
510e2ced
WW		 111 const struct in6_addr *daddr,
112 const struct in6_addr *saddr);
1da177e4 113
70ceb4f5 114#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 115static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 116 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 117 const struct in6_addr *gwaddr,
118 struct net_device *dev,
95c96174 119 unsigned int pref);
8d1c802b 120static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 121 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 122 const struct in6_addr *gwaddr,
123 struct net_device *dev);
70ceb4f5
YH		 124#endif
125
8d0b94af
MKL		 126struct uncached_list {
127 spinlock_t lock;
128 struct list_head head;
129};
130
131static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
132
510c321b 133void rt6_uncached_list_add(struct rt6_info *rt)
8d0b94af
MKL		 134{
135 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
136
8d0b94af
MKL		 137 rt->rt6i_uncached_list = ul;
138
139 spin_lock_bh(&ul->lock);
140 list_add_tail(&rt->rt6i_uncached, &ul->head);
141 spin_unlock_bh(&ul->lock);
142}
143
510c321b 144void rt6_uncached_list_del(struct rt6_info *rt)
8d0b94af
MKL		 145{
146 if (!list_empty(&rt->rt6i_uncached)) {
147 struct uncached_list *ul = rt->rt6i_uncached_list;
81eb8447 148 struct net *net = dev_net(rt->dst.dev);
8d0b94af
MKL		 149
150 spin_lock_bh(&ul->lock);
151 list_del(&rt->rt6i_uncached);
81eb8447 152 atomic_dec(&net->ipv6.rt6_stats->fib_rt_uncache);
8d0b94af
MKL		 153 spin_unlock_bh(&ul->lock);
154 }
155}
156
157static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
158{
159 struct net_device *loopback_dev = net->loopback_dev;
160 int cpu;
161
e332bc67
EB		 162 if (dev == loopback_dev)
163 return;
164
8d0b94af
MKL		 165 for_each_possible_cpu(cpu) {
166 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
167 struct rt6_info *rt;
168
169 spin_lock_bh(&ul->lock);
170 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
171 struct inet6_dev *rt_idev = rt->rt6i_idev;
172 struct net_device *rt_dev = rt->dst.dev;
173
e332bc67 174 if (rt_idev->dev == dev) {
8d0b94af
MKL		 175 rt->rt6i_idev = in6_dev_get(loopback_dev);
176 in6_dev_put(rt_idev);
177 }
178
e332bc67 179 if (rt_dev == dev) {
8d7017fd 180 rt->dst.dev = blackhole_netdev;
8d0b94af
MKL		 181 dev_hold(rt->dst.dev);
182 dev_put(rt_dev);
183 }
184 }
185 spin_unlock_bh(&ul->lock);
186 }
187}
188
f8a1b43b 189static inline const void *choose_neigh_daddr(const struct in6_addr *p,
f894cbf8
DM		 190 struct sk_buff *skb,
191 const void *daddr)
39232973 192{
a7563f34 193 if (!ipv6_addr_any(p))
39232973 194 return (const void *) p;
f894cbf8
DM		 195 else if (skb)
196 return &ipv6_hdr(skb)->daddr;
39232973
DM		 197 return daddr;
198}
199
f8a1b43b
DA		 200struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
201 struct net_device *dev,
202 struct sk_buff *skb,
203 const void *daddr)
d3aaeb38 204{
39232973
DM		 205 struct neighbour *n;
206
f8a1b43b
DA		 207 daddr = choose_neigh_daddr(gw, skb, daddr);
208 n = __ipv6_neigh_lookup(dev, daddr);
f83c7790
DM		 209 if (n)
210 return n;
7adf3246
SB		 211
212 n = neigh_create(&nd_tbl, daddr, dev);
213 return IS_ERR(n) ? NULL : n;
f8a1b43b
DA		 214}
215
216static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
217 struct sk_buff *skb,
218 const void *daddr)
219{
220 const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
221
2c6b55f4
ND		 222 return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
223 dst->dev, skb, daddr);
f83c7790
DM		 224}
225
63fca65d
JA		 226static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
227{
228 struct net_device *dev = dst->dev;
229 struct rt6_info *rt = (struct rt6_info *)dst;
230
cbfd6891 231 daddr = choose_neigh_daddr(rt6_nexthop(rt, &in6addr_any), NULL, daddr);
63fca65d
JA		 232 if (!daddr)
233 return;
234 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
235 return;
236 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
237 return;
238 __ipv6_confirm_neigh(dev, daddr);
239}
240
9a7ec3a9 241static struct dst_ops ip6_dst_ops_template = {
1da177e4 242 .family = AF_INET6,
1da177e4
LT		 243 .gc = ip6_dst_gc,
244 .gc_thresh = 1024,
245 .check = ip6_dst_check,
0dbaee3b 246 .default_advmss = ip6_default_advmss,
ebb762f2 247 .mtu = ip6_mtu,
d4ead6b3 248 .cow_metrics = dst_cow_metrics_generic,
1da177e4
LT		 249 .destroy = ip6_dst_destroy,
250 .ifdown = ip6_dst_ifdown,
251 .negative_advice = ip6_negative_advice,
252 .link_failure = ip6_link_failure,
253 .update_pmtu = ip6_rt_update_pmtu,
6e157b6a 254 .redirect = rt6_do_redirect,
9f8955cc 255 .local_out = __ip6_local_out,
f8a1b43b 256 .neigh_lookup = ip6_dst_neigh_lookup,
63fca65d 257 .confirm_neigh = ip6_confirm_neigh,
1da177e4
LT		 258};
259
ebb762f2 260static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
ec831ea7 261{
618f9bc7
SK		 262 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
263
264 return mtu ? : dst->dev->mtu;
ec831ea7
RD		 265}
266
6700c270 267static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
bd085ef6
HL		 268 struct sk_buff *skb, u32 mtu,
269 bool confirm_neigh)
14e50e57
DM		 270{
271}
272
6700c270
DM		 273static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
274 struct sk_buff *skb)
b587ee3b
DM		 275{
276}
277
14e50e57
DM		 278static struct dst_ops ip6_dst_blackhole_ops = {
279 .family = AF_INET6,
14e50e57
DM		 280 .destroy = ip6_dst_destroy,
281 .check = ip6_dst_check,
ebb762f2 282 .mtu = ip6_blackhole_mtu,
214f45c9 283 .default_advmss = ip6_default_advmss,
14e50e57 284 .update_pmtu = ip6_rt_blackhole_update_pmtu,
b587ee3b 285 .redirect = ip6_rt_blackhole_redirect,
0a1f5962 286 .cow_metrics = dst_cow_metrics_generic,
f8a1b43b 287 .neigh_lookup = ip6_dst_neigh_lookup,
14e50e57
DM		 288};
289
62fa8a84 290static const u32 ip6_template_metrics[RTAX_MAX] = {
14edd87d 291 [RTAX_HOPLIMIT - 1] = 0,
62fa8a84
DM		 292};
293
8d1c802b 294static const struct fib6_info fib6_null_entry_template = {
93c2fb25
DA		 295 .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
296 .fib6_protocol = RTPROT_KERNEL,
297 .fib6_metric = ~(u32)0,
f05713e0 298 .fib6_ref = REFCOUNT_INIT(1),
421842ed
DA		 299 .fib6_type = RTN_UNREACHABLE,
300 .fib6_metrics = (struct dst_metrics *)&dst_default_metrics,
301};
302
fb0af4c7 303static const struct rt6_info ip6_null_entry_template = {
d8d1f30b
CG		 304 .dst = {
305 .__refcnt = ATOMIC_INIT(1),
306 .__use = 1,
2c20cbd7 307 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 308 .error = -ENETUNREACH,
d8d1f30b
CG		 309 .input = ip6_pkt_discard,
310 .output = ip6_pkt_discard_out,
1da177e4
LT		 311 },
312 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
1da177e4
LT		 313};
314
101367c2
TG		 315#ifdef CONFIG_IPV6_MULTIPLE_TABLES
316
fb0af4c7 317static const struct rt6_info ip6_prohibit_entry_template = {
d8d1f30b
CG		 318 .dst = {
319 .__refcnt = ATOMIC_INIT(1),
320 .__use = 1,
2c20cbd7 321 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 322 .error = -EACCES,
d8d1f30b
CG		 323 .input = ip6_pkt_prohibit,
324 .output = ip6_pkt_prohibit_out,
101367c2
TG		 325 },
326 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 327};
328
fb0af4c7 329static const struct rt6_info ip6_blk_hole_entry_template = {
d8d1f30b
CG		 330 .dst = {
331 .__refcnt = ATOMIC_INIT(1),
332 .__use = 1,
2c20cbd7 333 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 334 .error = -EINVAL,
d8d1f30b 335 .input = dst_discard,
ede2059d 336 .output = dst_discard_out,
101367c2
TG		 337 },
338 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 339};
340
341#endif
342
ebfa45f0
MKL		 343static void rt6_info_init(struct rt6_info *rt)
344{
345 struct dst_entry *dst = &rt->dst;
346
347 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
ebfa45f0
MKL		 348 INIT_LIST_HEAD(&rt->rt6i_uncached);
349}
350
1da177e4 351/* allocate dst with ip6_dst_ops */
93531c67
DA		 352struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
353 int flags)
1da177e4 354{
97bab73f 355 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
b2a9c0ed 356 1, DST_OBSOLETE_FORCE_CHK, flags);
cf911662 357
81eb8447 358 if (rt) {
ebfa45f0 359 rt6_info_init(rt);
81eb8447
WW		 360 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
361 }
8104891b 362
cf911662 363 return rt;
1da177e4 364}
9ab179d8 365EXPORT_SYMBOL(ip6_dst_alloc);
d52d3997 366
1da177e4
LT		 367static void ip6_dst_destroy(struct dst_entry *dst)
368{
369 struct rt6_info *rt = (struct rt6_info *)dst;
a68886a6 370 struct fib6_info *from;
8d0b94af 371 struct inet6_dev *idev;
1da177e4 372
1620a336 373 ip_dst_metrics_put(dst);
8d0b94af
MKL		 374 rt6_uncached_list_del(rt);
375
376 idev = rt->rt6i_idev;
38308473 377 if (idev) {
1da177e4
LT		 378 rt->rt6i_idev = NULL;
379 in6_dev_put(idev);
1ab1457c 380 }
1716a961 381
0e233874 382 from = xchg((__force struct fib6_info **)&rt->from, NULL);
93531c67 383 fib6_info_release(from);
b3419363
DM		 384}
385
1da177e4
LT		 386static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
387 int how)
388{
389 struct rt6_info *rt = (struct rt6_info *)dst;
390 struct inet6_dev *idev = rt->rt6i_idev;
5a3e55d6 391 struct net_device *loopback_dev =
c346dca1 392 dev_net(dev)->loopback_dev;
1da177e4 393
e5645f51
WW		 394 if (idev && idev->dev != loopback_dev) {
395 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
396 if (loopback_idev) {
397 rt->rt6i_idev = loopback_idev;
398 in6_dev_put(idev);
97cac082 399 }
1da177e4
LT		 400 }
401}
402
5973fb1e
MKL		 403static bool __rt6_check_expired(const struct rt6_info *rt)
404{
405 if (rt->rt6i_flags & RTF_EXPIRES)
406 return time_after(jiffies, rt->dst.expires);
407 else
408 return false;
409}
410
a50feda5 411static bool rt6_check_expired(const struct rt6_info *rt)
1da177e4 412{
a68886a6
DA		 413 struct fib6_info *from;
414
415 from = rcu_dereference(rt->from);
416
1716a961
G		 417 if (rt->rt6i_flags & RTF_EXPIRES) {
418 if (time_after(jiffies, rt->dst.expires))
a50feda5 419 return true;
a68886a6 420 } else if (from) {
1e2ea8ad 421 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
a68886a6 422 fib6_check_expired(from);
1716a961 423 }
a50feda5 424 return false;
1da177e4
LT		 425}
426
b1d40991
DA		 427void fib6_select_path(const struct net *net, struct fib6_result *res,
428 struct flowi6 *fl6, int oif, bool have_oif_match,
429 const struct sk_buff *skb, int strict)
51ebd318 430{
8d1c802b 431 struct fib6_info *sibling, *next_sibling;
b1d40991
DA		 432 struct fib6_info *match = res->f6i;
433
f88d8ea6 434 if ((!match->fib6_nsiblings && !match->nh) || have_oif_match)
b1d40991 435 goto out;
51ebd318 436
b673d6cc
JS		 437 /* We might have already computed the hash for ICMPv6 errors. In such
438 * case it will always be non-zero. Otherwise now is the time to do it.
439 */
f88d8ea6
DA		 440 if (!fl6->mp_hash &&
441 (!match->nh || nexthop_is_multipath(match->nh)))
b4bac172 442 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
b673d6cc 443
f88d8ea6
DA		 444 if (unlikely(match->nh)) {
445 nexthop_path_fib6_result(res, fl6->mp_hash);
446 return;
447 }
448
1cf844c7 449 if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
b1d40991 450 goto out;
3d709f69 451
93c2fb25
DA		 452 list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
453 fib6_siblings) {
1cf844c7 454 const struct fib6_nh *nh = sibling->fib6_nh;
5e670d84
DA		 455 int nh_upper_bound;
456
702cea56 457 nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
5e670d84 458 if (fl6->mp_hash > nh_upper_bound)
3d709f69 459 continue;
702cea56 460 if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
3d709f69
IS		 461 break;
462 match = sibling;
463 break;
464 }
465
b1d40991
DA		 466out:
467 res->f6i = match;
1cf844c7 468 res->nh = match->fib6_nh;
51ebd318
ND		 469}
470
1da177e4 471/*
66f5d6ce 472 * Route lookup. rcu_read_lock() should be held.
1da177e4
LT		 473 */
474
0c59d006
DA		 475static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
476 const struct in6_addr *saddr, int oif, int flags)
477{
478 const struct net_device *dev;
479
480 if (nh->fib_nh_flags & RTNH_F_DEAD)
481 return false;
482
483 dev = nh->fib_nh_dev;
484 if (oif) {
485 if (dev->ifindex == oif)
486 return true;
487 } else {
488 if (ipv6_chk_addr(net, saddr, dev,
489 flags & RT6_LOOKUP_F_IFACE))
490 return true;
491 }
492
493 return false;
494}
495
962b6803
DA		 496struct fib6_nh_dm_arg {
497 struct net *net;
498 const struct in6_addr *saddr;
499 int oif;
500 int flags;
501 struct fib6_nh *nh;
502};
503
504static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
505{
506 struct fib6_nh_dm_arg *arg = _arg;
507
508 arg->nh = nh;
509 return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
510 arg->flags);
511}
512
513/* returns fib6_nh from nexthop or NULL */
514static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
515 struct fib6_result *res,
516 const struct in6_addr *saddr,
517 int oif, int flags)
518{
519 struct fib6_nh_dm_arg arg = {
520 .net = net,
521 .saddr = saddr,
522 .oif = oif,
523 .flags = flags,
524 };
525
526 if (nexthop_is_blackhole(nh))
527 return NULL;
528
529 if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
530 return arg.nh;
531
532 return NULL;
533}
534
75ef7389
DA		 535static void rt6_device_match(struct net *net, struct fib6_result *res,
536 const struct in6_addr *saddr, int oif, int flags)
1da177e4 537{
75ef7389
DA		 538 struct fib6_info *f6i = res->f6i;
539 struct fib6_info *spf6i;
540 struct fib6_nh *nh;
1da177e4 541
75ef7389 542 if (!oif && ipv6_addr_any(saddr)) {
f88d8ea6
DA		 543 if (unlikely(f6i->nh)) {
544 nh = nexthop_fib6_nh(f6i->nh);
545 if (nexthop_is_blackhole(f6i->nh))
546 goto out_blackhole;
547 } else {
548 nh = f6i->fib6_nh;
549 }
7d21fec9
DA		 550 if (!(nh->fib_nh_flags & RTNH_F_DEAD))
551 goto out;
75ef7389 552 }
dd3abc4e 553
75ef7389 554 for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
962b6803
DA		 555 bool matched = false;
556
557 if (unlikely(spf6i->nh)) {
558 nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
559 oif, flags);
560 if (nh)
561 matched = true;
562 } else {
563 nh = spf6i->fib6_nh;
564 if (__rt6_device_match(net, nh, saddr, oif, flags))
565 matched = true;
566 }
567 if (matched) {
75ef7389 568 res->f6i = spf6i;
7d21fec9 569 goto out;
75ef7389 570 }
dd3abc4e 571 }
1da177e4 572
75ef7389
DA		 573 if (oif && flags & RT6_LOOKUP_F_IFACE) {
574 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 575 nh = res->f6i->fib6_nh;
7d21fec9 576 goto out;
75ef7389 577 }
8067bb8c 578
f88d8ea6
DA		 579 if (unlikely(f6i->nh)) {
580 nh = nexthop_fib6_nh(f6i->nh);
581 if (nexthop_is_blackhole(f6i->nh))
582 goto out_blackhole;
583 } else {
584 nh = f6i->fib6_nh;
585 }
586
7d21fec9 587 if (nh->fib_nh_flags & RTNH_F_DEAD) {
75ef7389 588 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 589 nh = res->f6i->fib6_nh;
75ef7389 590 }
7d21fec9
DA		 591out:
592 res->nh = nh;
593 res->fib6_type = res->f6i->fib6_type;
594 res->fib6_flags = res->f6i->fib6_flags;
f88d8ea6
DA		 595 return;
596
597out_blackhole:
598 res->fib6_flags |= RTF_REJECT;
599 res->fib6_type = RTN_BLACKHOLE;
600 res->nh = nh;
1da177e4
LT		 601}
602
27097255 603#ifdef CONFIG_IPV6_ROUTER_PREF
c2f17e82
HFS		 604struct __rt6_probe_work {
605 struct work_struct work;
606 struct in6_addr target;
607 struct net_device *dev;
608};
609
610static void rt6_probe_deferred(struct work_struct *w)
611{
612 struct in6_addr mcaddr;
613 struct __rt6_probe_work *work =
614 container_of(w, struct __rt6_probe_work, work);
615
616 addrconf_addr_solict_mult(&work->target, &mcaddr);
adc176c5 617 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
c2f17e82 618 dev_put(work->dev);
662f5533 619 kfree(work);
c2f17e82
HFS		 620}
621
cc3a86c8 622static void rt6_probe(struct fib6_nh *fib6_nh)
27097255 623{
f547fac6 624 struct __rt6_probe_work *work = NULL;
5e670d84 625 const struct in6_addr *nh_gw;
1bef4c22 626 unsigned long last_probe;
f2c31e32 627 struct neighbour *neigh;
5e670d84 628 struct net_device *dev;
f547fac6 629 struct inet6_dev *idev;
5e670d84 630
27097255
YH		 631 /*
632 * Okay, this does not seem to be appropriate
633 * for now, however, we need to check if it
634 * is really so; aka Router Reachability Probing.
635 *
636 * Router Reachability Probe MUST be rate-limited
637 * to no more than one per minute.
638 */
004b3942 639 if (!fib6_nh->fib_nh_gw_family)
7ff74a59 640 return;
5e670d84 641
cc3a86c8
DA		 642 nh_gw = &fib6_nh->fib_nh_gw6;
643 dev = fib6_nh->fib_nh_dev;
2152caea 644 rcu_read_lock_bh();
1bef4c22 645 last_probe = READ_ONCE(fib6_nh->last_probe);
f547fac6 646 idev = __in6_dev_get(dev);
5e670d84 647 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
2152caea 648 if (neigh) {
8d6c31bf
MKL		 649 if (neigh->nud_state & NUD_VALID)
650 goto out;
651
2152caea 652 write_lock(&neigh->lock);
990edb42
MKL		 653 if (!(neigh->nud_state & NUD_VALID) &&
654 time_after(jiffies,
dcd1f572 655 neigh->updated + idev->cnf.rtr_probe_interval)) {
990edb42
MKL		 656 work = kmalloc(sizeof(*work), GFP_ATOMIC);
657 if (work)
658 __neigh_set_probe_once(neigh);
c2f17e82 659 }
2152caea 660 write_unlock(&neigh->lock);
1bef4c22 661 } else if (time_after(jiffies, last_probe +
f547fac6 662 idev->cnf.rtr_probe_interval)) {
990edb42 663 work = kmalloc(sizeof(*work), GFP_ATOMIC);
f2c31e32 664 }
990edb42 665
1bef4c22
ED		 666 if (!work || cmpxchg(&fib6_nh->last_probe,
667 last_probe, jiffies) != last_probe) {
668 kfree(work);
669 } else {
990edb42 670 INIT_WORK(&work->work, rt6_probe_deferred);
5e670d84
DA		 671 work->target = *nh_gw;
672 dev_hold(dev);
673 work->dev = dev;
990edb42
MKL		 674 schedule_work(&work->work);
675 }
676
8d6c31bf 677out:
2152caea 678 rcu_read_unlock_bh();
27097255
YH		 679}
680#else
cc3a86c8 681static inline void rt6_probe(struct fib6_nh *fib6_nh)
27097255 682{
27097255
YH		 683}
684#endif
685
1da177e4 686/*
554cfb7e 687 * Default Router Selection (RFC 2461 6.3.6)
1da177e4 688 */
1ba9a895 689static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
1da177e4 690{
afc154e9 691 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
5e670d84 692 struct neighbour *neigh;
f2c31e32 693
145a3621 694 rcu_read_lock_bh();
1ba9a895
DA		 695 neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
696 &fib6_nh->fib_nh_gw6);
145a3621
YH		 697 if (neigh) {
698 read_lock(&neigh->lock);
554cfb7e 699 if (neigh->nud_state & NUD_VALID)
afc154e9 700 ret = RT6_NUD_SUCCEED;
398bcbeb 701#ifdef CONFIG_IPV6_ROUTER_PREF
a5a81f0b 702 else if (!(neigh->nud_state & NUD_FAILED))
afc154e9 703 ret = RT6_NUD_SUCCEED;
7e980569
JB		 704 else
705 ret = RT6_NUD_FAIL_PROBE;
398bcbeb 706#endif
145a3621 707 read_unlock(&neigh->lock);
afc154e9
HFS		 708 } else {
709 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
7e980569 710 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
a5a81f0b 711 }
145a3621
YH		 712 rcu_read_unlock_bh();
713
a5a81f0b 714 return ret;
1da177e4
LT		 715}
716
702cea56
DA		 717static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
718 int strict)
1da177e4 719{
6e1809a5
DA		 720 int m = 0;
721
722 if (!oif || nh->fib_nh_dev->ifindex == oif)
723 m = 2;
1ab1457c 724
77d16f45 725 if (!m && (strict & RT6_LOOKUP_F_IFACE))
afc154e9 726 return RT6_NUD_FAIL_HARD;
ebacaaa0 727#ifdef CONFIG_IPV6_ROUTER_PREF
702cea56 728 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
ebacaaa0 729#endif
1ba9a895 730 if ((strict & RT6_LOOKUP_F_REACHABLE) &&
702cea56 731 !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
1ba9a895 732 int n = rt6_check_neigh(nh);
afc154e9
HFS		 733 if (n < 0)
734 return n;
735 }
554cfb7e
YH		 736 return m;
737}
738
28679ed1
DA		 739static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
740 int oif, int strict, int *mpri, bool *do_rr)
554cfb7e 741{
afc154e9 742 bool match_do_rr = false;
28679ed1
DA		 743 bool rc = false;
744 int m;
35103d11 745
28679ed1 746 if (nh->fib_nh_flags & RTNH_F_DEAD)
8067bb8c
IS		 747 goto out;
748
28679ed1
DA		 749 if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
750 nh->fib_nh_flags & RTNH_F_LINKDOWN &&
d5d32e4b 751 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
35103d11 752 goto out;
f11e6659 753
28679ed1 754 m = rt6_score_route(nh, fib6_flags, oif, strict);
7e980569 755 if (m == RT6_NUD_FAIL_DO_RR) {
afc154e9
HFS		 756 match_do_rr = true;
757 m = 0; /* lowest valid score */
7e980569 758 } else if (m == RT6_NUD_FAIL_HARD) {
f11e6659 759 goto out;
afc154e9
HFS		 760 }
761
762 if (strict & RT6_LOOKUP_F_REACHABLE)
28679ed1 763 rt6_probe(nh);
f11e6659 764
7e980569 765 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
f11e6659 766 if (m > *mpri) {
afc154e9 767 *do_rr = match_do_rr;
f11e6659 768 *mpri = m;
28679ed1 769 rc = true;
f11e6659 770 }
f11e6659 771out:
28679ed1 772 return rc;
f11e6659
DM		 773}
774
17a5984e
DA		 775struct fib6_nh_frl_arg {
776 u32 flags;
777 int oif;
778 int strict;
779 int *mpri;
780 bool *do_rr;
781 struct fib6_nh *nh;
782};
783
784static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
785{
786 struct fib6_nh_frl_arg *arg = _arg;
787
788 arg->nh = nh;
789 return find_match(nh, arg->flags, arg->oif, arg->strict,
790 arg->mpri, arg->do_rr);
791}
792
b7bc4b6a 793static void __find_rr_leaf(struct fib6_info *f6i_start,
30c15f03 794 struct fib6_info *nomatch, u32 metric,
b7bc4b6a 795 struct fib6_result *res, struct fib6_info **cont,
30c15f03 796 int oif, int strict, bool *do_rr, int *mpri)
f11e6659 797{
b7bc4b6a 798 struct fib6_info *f6i;
1da177e4 799
b7bc4b6a
DA		 800 for (f6i = f6i_start;
801 f6i && f6i != nomatch;
802 f6i = rcu_dereference(f6i->fib6_next)) {
17a5984e 803 bool matched = false;
30c15f03
DA		 804 struct fib6_nh *nh;
805
b7bc4b6a
DA		 806 if (cont && f6i->fib6_metric != metric) {
807 *cont = f6i;
30c15f03 808 return;
9fbdcfaf
SK		 809 }
810
b7bc4b6a 811 if (fib6_check_expired(f6i))
28679ed1
DA		 812 continue;
813
17a5984e
DA		 814 if (unlikely(f6i->nh)) {
815 struct fib6_nh_frl_arg arg = {
816 .flags = f6i->fib6_flags,
817 .oif = oif,
818 .strict = strict,
819 .mpri = mpri,
820 .do_rr = do_rr
821 };
822
823 if (nexthop_is_blackhole(f6i->nh)) {
824 res->fib6_flags = RTF_REJECT;
825 res->fib6_type = RTN_BLACKHOLE;
826 res->f6i = f6i;
827 res->nh = nexthop_fib6_nh(f6i->nh);
828 return;
829 }
830 if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
831 &arg)) {
832 matched = true;
833 nh = arg.nh;
834 }
835 } else {
836 nh = f6i->fib6_nh;
837 if (find_match(nh, f6i->fib6_flags, oif, strict,
838 mpri, do_rr))
839 matched = true;
840 }
841 if (matched) {
b7bc4b6a
DA		 842 res->f6i = f6i;
843 res->nh = nh;
7d21fec9
DA		 844 res->fib6_flags = f6i->fib6_flags;
845 res->fib6_type = f6i->fib6_type;
b7bc4b6a 846 }
9fbdcfaf 847 }
30c15f03 848}
9fbdcfaf 849
b7bc4b6a
DA		 850static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
851 struct fib6_info *rr_head, int oif, int strict,
852 bool *do_rr, struct fib6_result *res)
30c15f03 853{
b7bc4b6a
DA		 854 u32 metric = rr_head->fib6_metric;
855 struct fib6_info *cont = NULL;
30c15f03 856 int mpri = -1;
9fbdcfaf 857
b7bc4b6a 858 __find_rr_leaf(rr_head, NULL, metric, res, &cont,
30c15f03 859 oif, strict, do_rr, &mpri);
28679ed1 860
b7bc4b6a 861 __find_rr_leaf(leaf, rr_head, metric, res, &cont,
30c15f03 862 oif, strict, do_rr, &mpri);
9fbdcfaf 863
b7bc4b6a
DA		 864 if (res->f6i || !cont)
865 return;
9fbdcfaf 866
b7bc4b6a 867 __find_rr_leaf(cont, NULL, metric, res, NULL,
30c15f03 868 oif, strict, do_rr, &mpri);
f11e6659 869}
1da177e4 870
b7bc4b6a
DA		 871static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
872 struct fib6_result *res, int strict)
f11e6659 873{
8d1c802b 874 struct fib6_info *leaf = rcu_dereference(fn->leaf);
b7bc4b6a 875 struct fib6_info *rt0;
afc154e9 876 bool do_rr = false;
17ecf590 877 int key_plen;
1da177e4 878
b7bc4b6a
DA		 879 /* make sure this function or its helpers sets f6i */
880 res->f6i = NULL;
881
421842ed 882 if (!leaf || leaf == net->ipv6.fib6_null_entry)
b7bc4b6a 883 goto out;
8d1040e8 884
66f5d6ce 885 rt0 = rcu_dereference(fn->rr_ptr);
f11e6659 886 if (!rt0)
66f5d6ce 887 rt0 = leaf;
1da177e4 888
17ecf590
WW		 889 /* Double check to make sure fn is not an intermediate node
890 * and fn->leaf does not points to its child's leaf
891 * (This might happen if all routes under fn are deleted from
892 * the tree and fib6_repair_tree() is called on the node.)
893 */
93c2fb25 894 key_plen = rt0->fib6_dst.plen;
17ecf590 895#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 896 if (rt0->fib6_src.plen)
897 key_plen = rt0->fib6_src.plen;
17ecf590
WW		 898#endif
899 if (fn->fn_bit != key_plen)
b7bc4b6a 900 goto out;
1da177e4 901
b7bc4b6a 902 find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
afc154e9 903 if (do_rr) {
8fb11a9a 904 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
f11e6659 905
554cfb7e 906 /* no entries matched; do round-robin */
93c2fb25 907 if (!next || next->fib6_metric != rt0->fib6_metric)
8d1040e8 908 next = leaf;
f11e6659 909
66f5d6ce 910 if (next != rt0) {
93c2fb25 911 spin_lock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 912 /* make sure next is not being deleted from the tree */
93c2fb25 913 if (next->fib6_node)
66f5d6ce 914 rcu_assign_pointer(fn->rr_ptr, next);
93c2fb25 915 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 916 }
1da177e4 917 }
1da177e4 918
b7bc4b6a
DA		 919out:
920 if (!res->f6i) {
921 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 922 res->nh = res->f6i->fib6_nh;
7d21fec9
DA		 923 res->fib6_flags = res->f6i->fib6_flags;
924 res->fib6_type = res->f6i->fib6_type;
b7bc4b6a 925 }
1da177e4
LT		 926}
927
85bd05de 928static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
8b9df265 929{
85bd05de
DA		 930 return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
931 res->nh->fib_nh_gw_family;
8b9df265
MKL		 932}
933
70ceb4f5
YH		 934#ifdef CONFIG_IPV6_ROUTE_INFO
935int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
b71d1d42 936 const struct in6_addr *gwaddr)
70ceb4f5 937{
c346dca1 938 struct net *net = dev_net(dev);
70ceb4f5
YH		 939 struct route_info *rinfo = (struct route_info *) opt;
940 struct in6_addr prefix_buf, *prefix;
941 unsigned int pref;
4bed72e4 942 unsigned long lifetime;
8d1c802b 943 struct fib6_info *rt;
70ceb4f5
YH		 944
945 if (len < sizeof(struct route_info)) {
946 return -EINVAL;
947 }
948
949 /* Sanity check for prefix_len and length */
950 if (rinfo->length > 3) {
951 return -EINVAL;
952 } else if (rinfo->prefix_len > 128) {
953 return -EINVAL;
954 } else if (rinfo->prefix_len > 64) {
955 if (rinfo->length < 2) {
956 return -EINVAL;
957 }
958 } else if (rinfo->prefix_len > 0) {
959 if (rinfo->length < 1) {
960 return -EINVAL;
961 }
962 }
963
964 pref = rinfo->route_pref;
965 if (pref == ICMPV6_ROUTER_PREF_INVALID)
3933fc95 966 return -EINVAL;
70ceb4f5 967
4bed72e4 968 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
70ceb4f5
YH		 969
970 if (rinfo->length == 3)
971 prefix = (struct in6_addr *)rinfo->prefix;
972 else {
973 /* this function is safe */
974 ipv6_addr_prefix(&prefix_buf,
975 (struct in6_addr *)rinfo->prefix,
976 rinfo->prefix_len);
977 prefix = &prefix_buf;
978 }
979
f104a567 980 if (rinfo->prefix_len == 0)
afb1d4b5 981 rt = rt6_get_dflt_router(net, gwaddr, dev);
f104a567
DJ		 982 else
983 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
830218c1 984 gwaddr, dev);
70ceb4f5
YH		 985
986 if (rt && !lifetime) {
11dd74b3 987 ip6_del_rt(net, rt, false);
70ceb4f5
YH		 988 rt = NULL;
989 }
990
991 if (!rt && lifetime)
830218c1
DA		 992 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
993 dev, pref);
70ceb4f5 994 else if (rt)
93c2fb25
DA		 995 rt->fib6_flags = RTF_ROUTEINFO |
996 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
70ceb4f5
YH		 997
998 if (rt) {
1716a961 999 if (!addrconf_finite_timeout(lifetime))
14895687 1000 fib6_clean_expires(rt);
1716a961 1001 else
14895687 1002 fib6_set_expires(rt, jiffies + HZ * lifetime);
1716a961 1003
93531c67 1004 fib6_info_release(rt);
70ceb4f5
YH		 1005 }
1006 return 0;
1007}
1008#endif
1009
ae90d867
DA		 1010/*
1011 * Misc support functions
1012 */
1013
1014/* called with rcu_lock held */
0d161581 1015static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
ae90d867 1016{
0d161581 1017 struct net_device *dev = res->nh->fib_nh_dev;
ae90d867 1018
7d21fec9 1019 if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
ae90d867
DA		 1020 /* for copies of local routes, dst->dev needs to be the
1021 * device if it is a master device, the master device if
1022 * device is enslaved, and the loopback as the default
1023 */
1024 if (netif_is_l3_slave(dev) &&
7d21fec9 1025 !rt6_need_strict(&res->f6i->fib6_dst.addr))
ae90d867
DA		 1026 dev = l3mdev_master_dev_rcu(dev);
1027 else if (!netif_is_l3_master(dev))
1028 dev = dev_net(dev)->loopback_dev;
1029 /* last case is netif_is_l3_master(dev) is true in which
1030 * case we want dev returned to be dev
1031 */
1032 }
1033
1034 return dev;
1035}
1036
6edb3c96
DA		 1037static const int fib6_prop[RTN_MAX + 1] = {
1038 [RTN_UNSPEC] = 0,
1039 [RTN_UNICAST] = 0,
1040 [RTN_LOCAL] = 0,
1041 [RTN_BROADCAST] = 0,
1042 [RTN_ANYCAST] = 0,
1043 [RTN_MULTICAST] = 0,
1044 [RTN_BLACKHOLE] = -EINVAL,
1045 [RTN_UNREACHABLE] = -EHOSTUNREACH,
1046 [RTN_PROHIBIT] = -EACCES,
1047 [RTN_THROW] = -EAGAIN,
1048 [RTN_NAT] = -EINVAL,
1049 [RTN_XRESOLVE] = -EINVAL,
1050};
1051
1052static int ip6_rt_type_to_error(u8 fib6_type)
1053{
1054 return fib6_prop[fib6_type];
1055}
1056
8d1c802b 1057static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
3b6761d1
DA		 1058{
1059 unsigned short flags = 0;
1060
1061 if (rt->dst_nocount)
1062 flags |= DST_NOCOUNT;
1063 if (rt->dst_nopolicy)
1064 flags |= DST_NOPOLICY;
3b6761d1
DA		 1065
1066 return flags;
1067}
1068
7d21fec9 1069static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
6edb3c96 1070{
7d21fec9 1071 rt->dst.error = ip6_rt_type_to_error(fib6_type);
6edb3c96 1072
7d21fec9 1073 switch (fib6_type) {
6edb3c96
DA		 1074 case RTN_BLACKHOLE:
1075 rt->dst.output = dst_discard_out;
1076 rt->dst.input = dst_discard;
1077 break;
1078 case RTN_PROHIBIT:
1079 rt->dst.output = ip6_pkt_prohibit_out;
1080 rt->dst.input = ip6_pkt_prohibit;
1081 break;
1082 case RTN_THROW:
1083 case RTN_UNREACHABLE:
1084 default:
1085 rt->dst.output = ip6_pkt_discard_out;
1086 rt->dst.input = ip6_pkt_discard;
1087 break;
1088 }
1089}
1090
0d161581 1091static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
6edb3c96 1092{
7d21fec9 1093 struct fib6_info *f6i = res->f6i;
0d161581 1094
7d21fec9
DA		 1095 if (res->fib6_flags & RTF_REJECT) {
1096 ip6_rt_init_dst_reject(rt, res->fib6_type);
6edb3c96
DA		 1097 return;
1098 }
1099
1100 rt->dst.error = 0;
1101 rt->dst.output = ip6_output;
1102
7d21fec9 1103 if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
6edb3c96 1104 rt->dst.input = ip6_input;
7d21fec9 1105 } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
6edb3c96
DA		 1106 rt->dst.input = ip6_mc_input;
1107 } else {
1108 rt->dst.input = ip6_forward;
1109 }
1110
0d161581
DA		 1111 if (res->nh->fib_nh_lws) {
1112 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
6edb3c96
DA		 1113 lwtunnel_set_redirect(&rt->dst);
1114 }
1115
1116 rt->dst.lastuse = jiffies;
1117}
1118
e873e4b9 1119/* Caller must already hold reference to @from */
8d1c802b 1120static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
ae90d867 1121{
ae90d867 1122 rt->rt6i_flags &= ~RTF_EXPIRES;
a68886a6 1123 rcu_assign_pointer(rt->from, from);
e1255ed4 1124 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
ae90d867
DA		 1125}
1126
0d161581
DA		 1127/* Caller must already hold reference to f6i in result */
1128static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
ae90d867 1129{
0d161581
DA		 1130 const struct fib6_nh *nh = res->nh;
1131 const struct net_device *dev = nh->fib_nh_dev;
1132 struct fib6_info *f6i = res->f6i;
dcd1f572 1133
0d161581 1134 ip6_rt_init_dst(rt, res);
6edb3c96 1135
0d161581 1136 rt->rt6i_dst = f6i->fib6_dst;
dcd1f572 1137 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
7d21fec9 1138 rt->rt6i_flags = res->fib6_flags;
0d161581
DA		 1139 if (nh->fib_nh_gw_family) {
1140 rt->rt6i_gateway = nh->fib_nh_gw6;
2b2450ca
DA		 1141 rt->rt6i_flags |= RTF_GATEWAY;
1142 }
0d161581 1143 rt6_set_from(rt, f6i);
ae90d867 1144#ifdef CONFIG_IPV6_SUBTREES
0d161581 1145 rt->rt6i_src = f6i->fib6_src;
ae90d867 1146#endif
ae90d867
DA		 1147}
1148
a3c00e46
MKL		 1149static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1150 struct in6_addr *saddr)
1151{
66f5d6ce 1152 struct fib6_node *pn, *sn;
a3c00e46
MKL		 1153 while (1) {
1154 if (fn->fn_flags & RTN_TL_ROOT)
1155 return NULL;
66f5d6ce
WW		 1156 pn = rcu_dereference(fn->parent);
1157 sn = FIB6_SUBTREE(pn);
1158 if (sn && sn != fn)
6454743b 1159 fn = fib6_node_lookup(sn, NULL, saddr);
a3c00e46
MKL		 1160 else
1161 fn = pn;
1162 if (fn->fn_flags & RTN_RTINFO)
1163 return fn;
1164 }
1165}
c71099ac 1166
10585b43 1167static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
d3843fe5
WW		 1168{
1169 struct rt6_info *rt = *prt;
1170
1171 if (dst_hold_safe(&rt->dst))
1172 return true;
10585b43 1173 if (net) {
d3843fe5
WW		 1174 rt = net->ipv6.ip6_null_entry;
1175 dst_hold(&rt->dst);
1176 } else {
1177 rt = NULL;
1178 }
1179 *prt = rt;
1180 return false;
1181}
1182
dec9b0e2 1183/* called with rcu_lock held */
9b6b35ab 1184static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
dec9b0e2 1185{
9b6b35ab
DA		 1186 struct net_device *dev = res->nh->fib_nh_dev;
1187 struct fib6_info *f6i = res->f6i;
1188 unsigned short flags;
dec9b0e2
DA		 1189 struct rt6_info *nrt;
1190
9b6b35ab 1191 if (!fib6_info_hold_safe(f6i))
1c87e79a 1192 goto fallback;
e873e4b9 1193
9b6b35ab 1194 flags = fib6_info_dst_flags(f6i);
93531c67 1195 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1c87e79a 1196 if (!nrt) {
9b6b35ab 1197 fib6_info_release(f6i);
1c87e79a
XL		 1198 goto fallback;
1199 }
dec9b0e2 1200
0d161581 1201 ip6_rt_copy_init(nrt, res);
1c87e79a
XL		 1202 return nrt;
1203
1204fallback:
1205 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1206 dst_hold(&nrt->dst);
dec9b0e2
DA		 1207 return nrt;
1208}
1209
8ed67789
DL		 1210static struct rt6_info *ip6_pol_route_lookup(struct net *net,
1211 struct fib6_table *table,
b75cc8f9
DA		 1212 struct flowi6 *fl6,
1213 const struct sk_buff *skb,
1214 int flags)
1da177e4 1215{
b1d40991 1216 struct fib6_result res = {};
1da177e4 1217 struct fib6_node *fn;
23fb93a4 1218 struct rt6_info *rt;
1da177e4 1219
b6cdbc85
DA		 1220 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1221 flags &= ~RT6_LOOKUP_F_IFACE;
1222
66f5d6ce 1223 rcu_read_lock();
6454743b 1224 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
c71099ac 1225restart:
b1d40991
DA		 1226 res.f6i = rcu_dereference(fn->leaf);
1227 if (!res.f6i)
1228 res.f6i = net->ipv6.fib6_null_entry;
af52a52c 1229 else
75ef7389
DA		 1230 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1231 flags);
af52a52c 1232
b1d40991 1233 if (res.f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 1234 fn = fib6_backtrack(fn, &fl6->saddr);
1235 if (fn)
1236 goto restart;
2b760fcf 1237
af52a52c
DA		 1238 rt = net->ipv6.ip6_null_entry;
1239 dst_hold(&rt->dst);
1240 goto out;
f88d8ea6
DA		 1241 } else if (res.fib6_flags & RTF_REJECT) {
1242 goto do_create;
af52a52c 1243 }
d3843fe5 1244
b1d40991
DA		 1245 fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1246 fl6->flowi6_oif != 0, skb, flags);
1247
2b760fcf 1248 /* Search through exception table */
7e4b5128 1249 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
23fb93a4 1250 if (rt) {
10585b43 1251 if (ip6_hold_safe(net, &rt))
dec9b0e2 1252 dst_use_noref(&rt->dst, jiffies);
23fb93a4 1253 } else {
f88d8ea6 1254do_create:
9b6b35ab 1255 rt = ip6_create_rt_rcu(&res);
dec9b0e2 1256 }
b811580d 1257
af52a52c 1258out:
8ff2e5b2 1259 trace_fib6_table_lookup(net, &res, table, fl6);
af52a52c 1260
66f5d6ce 1261 rcu_read_unlock();
b811580d 1262
c71099ac 1263 return rt;
c71099ac
TG		 1264}
1265
67ba4152 1266struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
b75cc8f9 1267 const struct sk_buff *skb, int flags)
ea6e574e 1268{
b75cc8f9 1269 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
ea6e574e
FW		 1270}
1271EXPORT_SYMBOL_GPL(ip6_route_lookup);
1272
9acd9f3a 1273struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
b75cc8f9
DA		 1274 const struct in6_addr *saddr, int oif,
1275 const struct sk_buff *skb, int strict)
c71099ac 1276{
4c9483b2
DM		 1277 struct flowi6 fl6 = {
1278 .flowi6_oif = oif,
1279 .daddr = *daddr,
c71099ac
TG		 1280 };
1281 struct dst_entry *dst;
77d16f45 1282 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
c71099ac 1283
adaa70bb 1284 if (saddr) {
4c9483b2 1285 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
adaa70bb
TG		 1286 flags |= RT6_LOOKUP_F_HAS_SADDR;
1287 }
1288
b75cc8f9 1289 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
c71099ac
TG		 1290 if (dst->error == 0)
1291 return (struct rt6_info *) dst;
1292
1293 dst_release(dst);
1294
1da177e4
LT		 1295 return NULL;
1296}
7159039a
YH		 1297EXPORT_SYMBOL(rt6_lookup);
1298
c71099ac 1299/* ip6_ins_rt is called with FREE table->tb6_lock.
1cfb71ee
WW		 1300 * It takes new route entry, the addition fails by any reason the
1301 * route is released.
1302 * Caller must hold dst before calling it.
1da177e4
LT		 1303 */
1304
8d1c802b 1305static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
333c4301 1306 struct netlink_ext_ack *extack)
1da177e4
LT		 1307{
1308 int err;
c71099ac 1309 struct fib6_table *table;
1da177e4 1310
93c2fb25 1311 table = rt->fib6_table;
66f5d6ce 1312 spin_lock_bh(&table->tb6_lock);
d4ead6b3 1313 err = fib6_add(&table->tb6_root, rt, info, extack);
66f5d6ce 1314 spin_unlock_bh(&table->tb6_lock);
1da177e4
LT		 1315
1316 return err;
1317}
1318
8d1c802b 1319int ip6_ins_rt(struct net *net, struct fib6_info *rt)
40e22e8f 1320{
afb1d4b5 1321 struct nl_info info = { .nl_net = net, };
e715b6d3 1322
d4ead6b3 1323 return __ip6_ins_rt(rt, &info, NULL);
40e22e8f
TG		 1324}
1325
85bd05de 1326static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
8b9df265
MKL		 1327 const struct in6_addr *daddr,
1328 const struct in6_addr *saddr)
1da177e4 1329{
85bd05de 1330 struct fib6_info *f6i = res->f6i;
4832c30d 1331 struct net_device *dev;
1da177e4
LT		 1332 struct rt6_info *rt;
1333
1334 /*
1335 * Clone the route.
1336 */
1337
85bd05de 1338 if (!fib6_info_hold_safe(f6i))
e873e4b9
WW		 1339 return NULL;
1340
0d161581 1341 dev = ip6_rt_get_dev_rcu(res);
93531c67 1342 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
e873e4b9 1343 if (!rt) {
85bd05de 1344 fib6_info_release(f6i);
83a09abd 1345 return NULL;
e873e4b9 1346 }
83a09abd 1347
0d161581 1348 ip6_rt_copy_init(rt, res);
83a09abd 1349 rt->rt6i_flags |= RTF_CACHE;
83a09abd
MKL		 1350 rt->rt6i_dst.addr = *daddr;
1351 rt->rt6i_dst.plen = 128;
1da177e4 1352
85bd05de
DA		 1353 if (!rt6_is_gw_or_nonexthop(res)) {
1354 if (f6i->fib6_dst.plen != 128 &&
1355 ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
83a09abd 1356 rt->rt6i_flags |= RTF_ANYCAST;
1da177e4 1357#ifdef CONFIG_IPV6_SUBTREES
83a09abd
MKL		 1358 if (rt->rt6i_src.plen && saddr) {
1359 rt->rt6i_src.addr = *saddr;
1360 rt->rt6i_src.plen = 128;
8b9df265 1361 }
83a09abd 1362#endif
95a9a5ba 1363 }
1da177e4 1364
95a9a5ba
YH		 1365 return rt;
1366}
1da177e4 1367
db3fedee 1368static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
d52d3997 1369{
db3fedee
DA		 1370 struct fib6_info *f6i = res->f6i;
1371 unsigned short flags = fib6_info_dst_flags(f6i);
4832c30d 1372 struct net_device *dev;
d52d3997
MKL		 1373 struct rt6_info *pcpu_rt;
1374
db3fedee 1375 if (!fib6_info_hold_safe(f6i))
e873e4b9
WW		 1376 return NULL;
1377
4832c30d 1378 rcu_read_lock();
0d161581 1379 dev = ip6_rt_get_dev_rcu(res);
d8882935 1380 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT);
4832c30d 1381 rcu_read_unlock();
e873e4b9 1382 if (!pcpu_rt) {
db3fedee 1383 fib6_info_release(f6i);
d52d3997 1384 return NULL;
e873e4b9 1385 }
0d161581 1386 ip6_rt_copy_init(pcpu_rt, res);
d52d3997 1387 pcpu_rt->rt6i_flags |= RTF_PCPU;
8f34e53b
DA		 1388
1389 if (f6i->nh)
1390 pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1391
d52d3997
MKL		 1392 return pcpu_rt;
1393}
1394
8f34e53b
DA		 1395static bool rt6_is_valid(const struct rt6_info *rt6)
1396{
1397 return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1398}
1399
66f5d6ce 1400/* It should be called with rcu_read_lock() acquired */
db3fedee 1401static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
d52d3997 1402{
c353071a 1403 struct rt6_info *pcpu_rt;
d52d3997 1404
c353071a 1405 pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
d52d3997 1406
8f34e53b
DA		 1407 if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1408 struct rt6_info *prev, **p;
1409
1410 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1411 prev = xchg(p, NULL);
1412 if (prev) {
1413 dst_dev_put(&prev->dst);
1414 dst_release(&prev->dst);
1415 }
1416
1417 pcpu_rt = NULL;
1418 }
1419
a73e4195
MKL		 1420 return pcpu_rt;
1421}
1422
afb1d4b5 1423static struct rt6_info *rt6_make_pcpu_route(struct net *net,
db3fedee 1424 const struct fib6_result *res)
a73e4195
MKL		 1425{
1426 struct rt6_info *pcpu_rt, *prev, **p;
d52d3997 1427
db3fedee 1428 pcpu_rt = ip6_rt_pcpu_alloc(res);
0e09edcc
WW		 1429 if (!pcpu_rt)
1430 return NULL;
d52d3997 1431
f40b6ae2 1432 p = this_cpu_ptr(res->nh->rt6i_pcpu);
a94b9367 1433 prev = cmpxchg(p, NULL, pcpu_rt);
951f788a 1434 BUG_ON(prev);
a94b9367 1435
61fb0d01
ED		 1436 if (res->f6i->fib6_destroying) {
1437 struct fib6_info *from;
1438
1439 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1440 fib6_info_release(from);
1441 }
1442
d52d3997
MKL		 1443 return pcpu_rt;
1444}
1445
35732d01
WW		 1446/* exception hash table implementation
1447 */
1448static DEFINE_SPINLOCK(rt6_exception_lock);
1449
1450/* Remove rt6_ex from hash table and free the memory
1451 * Caller must hold rt6_exception_lock
1452 */
1453static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1454 struct rt6_exception *rt6_ex)
1455{
f5b51fe8 1456 struct fib6_info *from;
b2427e67 1457 struct net *net;
81eb8447 1458
35732d01
WW		 1459 if (!bucket || !rt6_ex)
1460 return;
b2427e67
CIK		 1461
1462 net = dev_net(rt6_ex->rt6i->dst.dev);
f5b51fe8
PA		 1463 net->ipv6.rt6_stats->fib_rt_cache--;
1464
1465 /* purge completely the exception to allow releasing the held resources:
1466 * some [sk] cache may keep the dst around for unlimited time
1467 */
0e233874 1468 from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
f5b51fe8
PA		 1469 fib6_info_release(from);
1470 dst_dev_put(&rt6_ex->rt6i->dst);
1471
35732d01 1472 hlist_del_rcu(&rt6_ex->hlist);
77634cc6 1473 dst_release(&rt6_ex->rt6i->dst);
35732d01
WW		 1474 kfree_rcu(rt6_ex, rcu);
1475 WARN_ON_ONCE(!bucket->depth);
1476 bucket->depth--;
1477}
1478
1479/* Remove oldest rt6_ex in bucket and free the memory
1480 * Caller must hold rt6_exception_lock
1481 */
1482static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1483{
1484 struct rt6_exception *rt6_ex, *oldest = NULL;
1485
1486 if (!bucket)
1487 return;
1488
1489 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1490 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1491 oldest = rt6_ex;
1492 }
1493 rt6_remove_exception(bucket, oldest);
1494}
1495
1496static u32 rt6_exception_hash(const struct in6_addr *dst,
1497 const struct in6_addr *src)
1498{
1499 static u32 seed __read_mostly;
1500 u32 val;
1501
1502 net_get_random_once(&seed, sizeof(seed));
b6b556af 1503 val = jhash2((const u32 *)dst, sizeof(*dst)/sizeof(u32), seed);
35732d01
WW		 1504
1505#ifdef CONFIG_IPV6_SUBTREES
1506 if (src)
b6b556af 1507 val = jhash2((const u32 *)src, sizeof(*src)/sizeof(u32), val);
35732d01
WW		 1508#endif
1509 return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1510}
1511
1512/* Helper function to find the cached rt in the hash table
1513 * and update bucket pointer to point to the bucket for this
1514 * (daddr, saddr) pair
1515 * Caller must hold rt6_exception_lock
1516 */
1517static struct rt6_exception *
1518__rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1519 const struct in6_addr *daddr,
1520 const struct in6_addr *saddr)
1521{
1522 struct rt6_exception *rt6_ex;
1523 u32 hval;
1524
1525 if (!(*bucket) || !daddr)
1526 return NULL;
1527
1528 hval = rt6_exception_hash(daddr, saddr);
1529 *bucket += hval;
1530
1531 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1532 struct rt6_info *rt6 = rt6_ex->rt6i;
1533 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1534
1535#ifdef CONFIG_IPV6_SUBTREES
1536 if (matched && saddr)
1537 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1538#endif
1539 if (matched)
1540 return rt6_ex;
1541 }
1542 return NULL;
1543}
1544
1545/* Helper function to find the cached rt in the hash table
1546 * and update bucket pointer to point to the bucket for this
1547 * (daddr, saddr) pair
1548 * Caller must hold rcu_read_lock()
1549 */
1550static struct rt6_exception *
1551__rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1552 const struct in6_addr *daddr,
1553 const struct in6_addr *saddr)
1554{
1555 struct rt6_exception *rt6_ex;
1556 u32 hval;
1557
1558 WARN_ON_ONCE(!rcu_read_lock_held());
1559
1560 if (!(*bucket) || !daddr)
1561 return NULL;
1562
1563 hval = rt6_exception_hash(daddr, saddr);
1564 *bucket += hval;
1565
1566 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1567 struct rt6_info *rt6 = rt6_ex->rt6i;
1568 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1569
1570#ifdef CONFIG_IPV6_SUBTREES
1571 if (matched && saddr)
1572 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1573#endif
1574 if (matched)
1575 return rt6_ex;
1576 }
1577 return NULL;
1578}
1579
b748f260 1580static unsigned int fib6_mtu(const struct fib6_result *res)
d4ead6b3 1581{
b748f260 1582 const struct fib6_nh *nh = res->nh;
d4ead6b3
DA		 1583 unsigned int mtu;
1584
b748f260
DA		 1585 if (res->f6i->fib6_pmtu) {
1586 mtu = res->f6i->fib6_pmtu;
dcd1f572 1587 } else {
b748f260 1588 struct net_device *dev = nh->fib_nh_dev;
dcd1f572
DA		 1589 struct inet6_dev *idev;
1590
1591 rcu_read_lock();
1592 idev = __in6_dev_get(dev);
1593 mtu = idev->cnf.mtu6;
1594 rcu_read_unlock();
1595 }
1596
d4ead6b3
DA		 1597 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1598
b748f260 1599 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
d4ead6b3
DA		 1600}
1601
cc5c073a
DA		 1602#define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1603
1604/* used when the flushed bit is not relevant, only access to the bucket
1605 * (ie., all bucket users except rt6_insert_exception);
1606 *
1607 * called under rcu lock; sometimes called with rt6_exception_lock held
1608 */
1609static
1610struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1611 spinlock_t *lock)
1612{
1613 struct rt6_exception_bucket *bucket;
1614
1615 if (lock)
1616 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1617 lockdep_is_held(lock));
1618 else
1619 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1620
1621 /* remove bucket flushed bit if set */
1622 if (bucket) {
1623 unsigned long p = (unsigned long)bucket;
1624
1625 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1626 bucket = (struct rt6_exception_bucket *)p;
1627 }
1628
1629 return bucket;
1630}
1631
1632static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1633{
1634 unsigned long p = (unsigned long)bucket;
1635
1636 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1637}
1638
1639/* called with rt6_exception_lock held */
1640static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1641 spinlock_t *lock)
1642{
1643 struct rt6_exception_bucket *bucket;
1644 unsigned long p;
1645
1646 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1647 lockdep_is_held(lock));
1648
1649 p = (unsigned long)bucket;
1650 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1651 bucket = (struct rt6_exception_bucket *)p;
1652 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1653}
1654
35732d01 1655static int rt6_insert_exception(struct rt6_info *nrt,
5012f0a5 1656 const struct fib6_result *res)
35732d01 1657{
5e670d84 1658 struct net *net = dev_net(nrt->dst.dev);
35732d01 1659 struct rt6_exception_bucket *bucket;
cc5c073a 1660 struct fib6_info *f6i = res->f6i;
35732d01
WW		 1661 struct in6_addr *src_key = NULL;
1662 struct rt6_exception *rt6_ex;
cc5c073a 1663 struct fib6_nh *nh = res->nh;
35732d01
WW		 1664 int err = 0;
1665
35732d01
WW		 1666 spin_lock_bh(&rt6_exception_lock);
1667
cc5c073a
DA		 1668 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1669 lockdep_is_held(&rt6_exception_lock));
35732d01
WW		 1670 if (!bucket) {
1671 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1672 GFP_ATOMIC);
1673 if (!bucket) {
1674 err = -ENOMEM;
1675 goto out;
1676 }
cc5c073a
DA		 1677 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1678 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1679 err = -EINVAL;
1680 goto out;
35732d01
WW		 1681 }
1682
1683#ifdef CONFIG_IPV6_SUBTREES
5012f0a5 1684 /* fib6_src.plen != 0 indicates f6i is in subtree
35732d01 1685 * and exception table is indexed by a hash of
5012f0a5 1686 * both fib6_dst and fib6_src.
35732d01 1687 * Otherwise, the exception table is indexed by
5012f0a5 1688 * a hash of only fib6_dst.
35732d01 1689 */
5012f0a5 1690 if (f6i->fib6_src.plen)
35732d01
WW		 1691 src_key = &nrt->rt6i_src.addr;
1692#endif
5012f0a5 1693 /* rt6_mtu_change() might lower mtu on f6i.
f5bbe7ee 1694 * Only insert this exception route if its mtu
5012f0a5 1695 * is less than f6i's mtu value.
f5bbe7ee 1696 */
b748f260 1697 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
f5bbe7ee
WW		 1698 err = -EINVAL;
1699 goto out;
1700 }
60006a48 1701
35732d01
WW		 1702 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1703 src_key);
1704 if (rt6_ex)
1705 rt6_remove_exception(bucket, rt6_ex);
1706
1707 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1708 if (!rt6_ex) {
1709 err = -ENOMEM;
1710 goto out;
1711 }
1712 rt6_ex->rt6i = nrt;
1713 rt6_ex->stamp = jiffies;
35732d01
WW		 1714 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1715 bucket->depth++;
81eb8447 1716 net->ipv6.rt6_stats->fib_rt_cache++;
35732d01
WW		 1717
1718 if (bucket->depth > FIB6_MAX_DEPTH)
1719 rt6_exception_remove_oldest(bucket);
1720
1721out:
1722 spin_unlock_bh(&rt6_exception_lock);
1723
1724 /* Update fn->fn_sernum to invalidate all cached dst */
b886d5f2 1725 if (!err) {
5012f0a5
DA		 1726 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1727 fib6_update_sernum(net, f6i);
1728 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
b886d5f2
PA		 1729 fib6_force_start_gc(net);
1730 }
35732d01
WW		 1731
1732 return err;
1733}
1734
c0b220cf 1735static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
35732d01
WW		 1736{
1737 struct rt6_exception_bucket *bucket;
1738 struct rt6_exception *rt6_ex;
1739 struct hlist_node *tmp;
1740 int i;
1741
1742 spin_lock_bh(&rt6_exception_lock);
35732d01 1743
cc5c073a 1744 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
35732d01
WW		 1745 if (!bucket)
1746 goto out;
1747
cc5c073a
DA		 1748 /* Prevent rt6_insert_exception() to recreate the bucket list */
1749 if (!from)
1750 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1751
35732d01 1752 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
cc5c073a
DA		 1753 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1754 if (!from ||
1755 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1756 rt6_remove_exception(bucket, rt6_ex);
1757 }
1758 WARN_ON_ONCE(!from && bucket->depth);
35732d01
WW		 1759 bucket++;
1760 }
35732d01
WW		 1761out:
1762 spin_unlock_bh(&rt6_exception_lock);
1763}
1764
e659ba31
DA		 1765static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1766{
1767 struct fib6_info *f6i = arg;
1768
1769 fib6_nh_flush_exceptions(nh, f6i);
1770
1771 return 0;
1772}
1773
c0b220cf
DA		 1774void rt6_flush_exceptions(struct fib6_info *f6i)
1775{
e659ba31
DA		 1776 if (f6i->nh)
1777 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1778 f6i);
1779 else
1780 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
c0b220cf
DA		 1781}
1782
35732d01
WW		 1783/* Find cached rt in the hash table inside passed in rt
1784 * Caller has to hold rcu_read_lock()
1785 */
7e4b5128 1786static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
510e2ced
WW		 1787 const struct in6_addr *daddr,
1788 const struct in6_addr *saddr)
35732d01 1789{
510e2ced 1790 const struct in6_addr *src_key = NULL;
35732d01 1791 struct rt6_exception_bucket *bucket;
35732d01 1792 struct rt6_exception *rt6_ex;
7e4b5128 1793 struct rt6_info *ret = NULL;
35732d01 1794
35732d01 1795#ifdef CONFIG_IPV6_SUBTREES
7e4b5128 1796 /* fib6i_src.plen != 0 indicates f6i is in subtree
35732d01 1797 * and exception table is indexed by a hash of
7e4b5128 1798 * both fib6_dst and fib6_src.
510e2ced
WW		 1799 * However, the src addr used to create the hash
1800 * might not be exactly the passed in saddr which
1801 * is a /128 addr from the flow.
1802 * So we need to use f6i->fib6_src to redo lookup
1803 * if the passed in saddr does not find anything.
1804 * (See the logic in ip6_rt_cache_alloc() on how
1805 * rt->rt6i_src is updated.)
35732d01 1806 */
7e4b5128 1807 if (res->f6i->fib6_src.plen)
35732d01 1808 src_key = saddr;
510e2ced 1809find_ex:
35732d01 1810#endif
cc5c073a 1811 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
35732d01
WW		 1812 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1813
1814 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
7e4b5128 1815 ret = rt6_ex->rt6i;
35732d01 1816
510e2ced
WW		 1817#ifdef CONFIG_IPV6_SUBTREES
1818 /* Use fib6_src as src_key and redo lookup */
1819 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1820 src_key = &res->f6i->fib6_src.addr;
1821 goto find_ex;
1822 }
1823#endif
1824
7e4b5128 1825 return ret;
35732d01
WW		 1826}
1827
1828/* Remove the passed in cached rt from the hash table that contains it */
cc5c073a 1829static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1830 const struct rt6_info *rt)
35732d01 1831{
c0b220cf 1832 const struct in6_addr *src_key = NULL;
35732d01 1833 struct rt6_exception_bucket *bucket;
35732d01
WW		 1834 struct rt6_exception *rt6_ex;
1835 int err;
1836
cc5c073a 1837 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
35732d01
WW		 1838 return -ENOENT;
1839
1840 spin_lock_bh(&rt6_exception_lock);
cc5c073a
DA		 1841 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1842
35732d01 1843#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1844 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1845 * and exception table is indexed by a hash of
1846 * both rt6i_dst and rt6i_src.
35732d01
WW		 1847 * Otherwise, the exception table is indexed by
1848 * a hash of only rt6i_dst.
1849 */
c0b220cf 1850 if (plen)
35732d01
WW		 1851 src_key = &rt->rt6i_src.addr;
1852#endif
1853 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1854 &rt->rt6i_dst.addr,
1855 src_key);
1856 if (rt6_ex) {
1857 rt6_remove_exception(bucket, rt6_ex);
1858 err = 0;
1859 } else {
1860 err = -ENOENT;
1861 }
1862
1863 spin_unlock_bh(&rt6_exception_lock);
1864 return err;
1865}
1866
e659ba31
DA		 1867struct fib6_nh_excptn_arg {
1868 struct rt6_info *rt;
1869 int plen;
1870};
1871
1872static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1873{
1874 struct fib6_nh_excptn_arg *arg = _arg;
1875 int err;
1876
1877 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1878 if (err == 0)
1879 return 1;
1880
1881 return 0;
1882}
1883
c0b220cf
DA		 1884static int rt6_remove_exception_rt(struct rt6_info *rt)
1885{
1886 struct fib6_info *from;
1887
1888 from = rcu_dereference(rt->from);
cc5c073a 1889 if (!from || !(rt->rt6i_flags & RTF_CACHE))
c0b220cf
DA		 1890 return -EINVAL;
1891
e659ba31
DA		 1892 if (from->nh) {
1893 struct fib6_nh_excptn_arg arg = {
1894 .rt = rt,
1895 .plen = from->fib6_src.plen
1896 };
1897 int rc;
1898
1899 /* rc = 1 means an entry was found */
1900 rc = nexthop_for_each_fib6_nh(from->nh,
1901 rt6_nh_remove_exception_rt,
1902 &arg);
1903 return rc ? 0 : -ENOENT;
1904 }
1905
1cf844c7 1906 return fib6_nh_remove_exception(from->fib6_nh,
cc5c073a 1907 from->fib6_src.plen, rt);
c0b220cf
DA		 1908}
1909
35732d01
WW		 1910/* Find rt6_ex which contains the passed in rt cache and
1911 * refresh its stamp
1912 */
cc5c073a 1913static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1914 const struct rt6_info *rt)
35732d01 1915{
c0b220cf 1916 const struct in6_addr *src_key = NULL;
35732d01 1917 struct rt6_exception_bucket *bucket;
35732d01 1918 struct rt6_exception *rt6_ex;
193f3685 1919
cc5c073a 1920 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
35732d01 1921#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1922 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1923 * and exception table is indexed by a hash of
1924 * both rt6i_dst and rt6i_src.
35732d01
WW		 1925 * Otherwise, the exception table is indexed by
1926 * a hash of only rt6i_dst.
1927 */
c0b220cf 1928 if (plen)
35732d01
WW		 1929 src_key = &rt->rt6i_src.addr;
1930#endif
cc5c073a 1931 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
35732d01
WW		 1932 if (rt6_ex)
1933 rt6_ex->stamp = jiffies;
c0b220cf
DA		 1934}
1935
e659ba31
DA		 1936struct fib6_nh_match_arg {
1937 const struct net_device *dev;
1938 const struct in6_addr *gw;
1939 struct fib6_nh *match;
1940};
1941
1942/* determine if fib6_nh has given device and gateway */
1943static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1944{
1945 struct fib6_nh_match_arg *arg = _arg;
1946
1947 if (arg->dev != nh->fib_nh_dev ||
1948 (arg->gw && !nh->fib_nh_gw_family) ||
1949 (!arg->gw && nh->fib_nh_gw_family) ||
1950 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1951 return 0;
1952
1953 arg->match = nh;
1954
1955 /* found a match, break the loop */
1956 return 1;
1957}
1958
c0b220cf
DA		 1959static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1960{
1961 struct fib6_info *from;
e659ba31 1962 struct fib6_nh *fib6_nh;
35732d01 1963
c0b220cf
DA		 1964 rcu_read_lock();
1965
1966 from = rcu_dereference(rt->from);
1967 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1968 goto unlock;
1969
e659ba31
DA		 1970 if (from->nh) {
1971 struct fib6_nh_match_arg arg = {
1972 .dev = rt->dst.dev,
1973 .gw = &rt->rt6i_gateway,
1974 };
1975
1976 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1977
1978 if (!arg.match)
cff6a327 1979 goto unlock;
e659ba31
DA		 1980 fib6_nh = arg.match;
1981 } else {
1982 fib6_nh = from->fib6_nh;
1983 }
1984 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
193f3685 1985unlock:
35732d01
WW		 1986 rcu_read_unlock();
1987}
1988
e9fa1495
SB		 1989static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1990 struct rt6_info *rt, int mtu)
1991{
1992 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1993 * lowest MTU in the path: always allow updating the route PMTU to
1994 * reflect PMTU decreases.
1995 *
1996 * If the new MTU is higher, and the route PMTU is equal to the local
1997 * MTU, this means the old MTU is the lowest in the path, so allow
1998 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1999 * handle this.
2000 */
2001
2002 if (dst_mtu(&rt->dst) >= mtu)
2003 return true;
2004
2005 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
2006 return true;
2007
2008 return false;
2009}
2010
2011static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
cc5c073a 2012 const struct fib6_nh *nh, int mtu)
f5bbe7ee
WW		 2013{
2014 struct rt6_exception_bucket *bucket;
2015 struct rt6_exception *rt6_ex;
2016 int i;
2017
cc5c073a 2018 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
e9fa1495
SB		 2019 if (!bucket)
2020 return;
2021
2022 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2023 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2024 struct rt6_info *entry = rt6_ex->rt6i;
2025
2026 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
d4ead6b3 2027 * route), the metrics of its rt->from have already
e9fa1495
SB		 2028 * been updated.
2029 */
d4ead6b3 2030 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
e9fa1495 2031 rt6_mtu_change_route_allowed(idev, entry, mtu))
d4ead6b3 2032 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
f5bbe7ee 2033 }
e9fa1495 2034 bucket++;
f5bbe7ee
WW		 2035 }
2036}
2037
b16cb459
WW		 2038#define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2039
cc5c073a
DA		 2040static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2041 const struct in6_addr *gateway)
b16cb459
WW		 2042{
2043 struct rt6_exception_bucket *bucket;
2044 struct rt6_exception *rt6_ex;
2045 struct hlist_node *tmp;
2046 int i;
2047
cc5c073a 2048 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
b16cb459
WW		 2049 return;
2050
2051 spin_lock_bh(&rt6_exception_lock);
cc5c073a 2052 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
b16cb459
WW		 2053 if (bucket) {
2054 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2055 hlist_for_each_entry_safe(rt6_ex, tmp,
2056 &bucket->chain, hlist) {
2057 struct rt6_info *entry = rt6_ex->rt6i;
2058
2059 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2060 RTF_CACHE_GATEWAY &&
2061 ipv6_addr_equal(gateway,
2062 &entry->rt6i_gateway)) {
2063 rt6_remove_exception(bucket, rt6_ex);
2064 }
2065 }
2066 bucket++;
2067 }
2068 }
2069
2070 spin_unlock_bh(&rt6_exception_lock);
2071}
2072
c757faa8
WW		 2073static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2074 struct rt6_exception *rt6_ex,
2075 struct fib6_gc_args *gc_args,
2076 unsigned long now)
2077{
2078 struct rt6_info *rt = rt6_ex->rt6i;
2079
1859bac0
PA		 2080 /* we are pruning and obsoleting aged-out and non gateway exceptions
2081 * even if others have still references to them, so that on next
2082 * dst_check() such references can be dropped.
2083 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2084 * expired, independently from their aging, as per RFC 8201 section 4
2085 */
31afeb42
WW		 2086 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2087 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2088 RT6_TRACE("aging clone %p\n", rt);
2089 rt6_remove_exception(bucket, rt6_ex);
2090 return;
2091 }
2092 } else if (time_after(jiffies, rt->dst.expires)) {
2093 RT6_TRACE("purging expired route %p\n", rt);
c757faa8
WW		 2094 rt6_remove_exception(bucket, rt6_ex);
2095 return;
31afeb42
WW		 2096 }
2097
2098 if (rt->rt6i_flags & RTF_GATEWAY) {
c757faa8
WW		 2099 struct neighbour *neigh;
2100 __u8 neigh_flags = 0;
2101
1bfa26ff
ED		 2102 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2103 if (neigh)
c757faa8 2104 neigh_flags = neigh->flags;
1bfa26ff 2105
c757faa8
WW		 2106 if (!(neigh_flags & NTF_ROUTER)) {
2107 RT6_TRACE("purging route %p via non-router but gateway\n",
2108 rt);
2109 rt6_remove_exception(bucket, rt6_ex);
2110 return;
2111 }
2112 }
31afeb42 2113
c757faa8
WW		 2114 gc_args->more++;
2115}
2116
cc5c073a 2117static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
c0b220cf
DA		 2118 struct fib6_gc_args *gc_args,
2119 unsigned long now)
c757faa8
WW		 2120{
2121 struct rt6_exception_bucket *bucket;
2122 struct rt6_exception *rt6_ex;
2123 struct hlist_node *tmp;
2124 int i;
2125
cc5c073a 2126 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
c757faa8
WW		 2127 return;
2128
1bfa26ff
ED		 2129 rcu_read_lock_bh();
2130 spin_lock(&rt6_exception_lock);
cc5c073a 2131 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
c757faa8
WW		 2132 if (bucket) {
2133 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2134 hlist_for_each_entry_safe(rt6_ex, tmp,
2135 &bucket->chain, hlist) {
2136 rt6_age_examine_exception(bucket, rt6_ex,
2137 gc_args, now);
2138 }
2139 bucket++;
2140 }
2141 }
1bfa26ff
ED		 2142 spin_unlock(&rt6_exception_lock);
2143 rcu_read_unlock_bh();
c757faa8
WW		 2144}
2145
e659ba31
DA		 2146struct fib6_nh_age_excptn_arg {
2147 struct fib6_gc_args *gc_args;
2148 unsigned long now;
2149};
2150
2151static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2152{
2153 struct fib6_nh_age_excptn_arg *arg = _arg;
2154
2155 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2156 return 0;
2157}
2158
cc5c073a 2159void rt6_age_exceptions(struct fib6_info *f6i,
c0b220cf
DA		 2160 struct fib6_gc_args *gc_args,
2161 unsigned long now)
2162{
e659ba31
DA		 2163 if (f6i->nh) {
2164 struct fib6_nh_age_excptn_arg arg = {
2165 .gc_args = gc_args,
2166 .now = now
2167 };
2168
2169 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2170 &arg);
2171 } else {
2172 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2173 }
c0b220cf
DA		 2174}
2175
1d053da9 2176/* must be called with rcu lock held */
effda4dd
DA		 2177int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2178 struct flowi6 *fl6, struct fib6_result *res, int strict)
1da177e4 2179{
367efcb9 2180 struct fib6_node *fn, *saved_fn;
1da177e4 2181
6454743b 2182 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
367efcb9 2183 saved_fn = fn;
1da177e4 2184
ca254490
DA		 2185 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2186 oif = 0;
2187
a3c00e46 2188redo_rt6_select:
effda4dd
DA		 2189 rt6_select(net, fn, oif, res, strict);
2190 if (res->f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2191 fn = fib6_backtrack(fn, &fl6->saddr);
2192 if (fn)
2193 goto redo_rt6_select;
367efcb9
MKL		 2194 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2195 /* also consider unreachable route */
2196 strict &= ~RT6_LOOKUP_F_REACHABLE;
2197 fn = saved_fn;
2198 goto redo_rt6_select;
367efcb9 2199 }
a3c00e46
MKL		 2200 }
2201
effda4dd 2202 trace_fib6_table_lookup(net, res, table, fl6);
fb9de91e 2203
effda4dd 2204 return 0;
1d053da9
DA		 2205}
2206
2207struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2208 int oif, struct flowi6 *fl6,
2209 const struct sk_buff *skb, int flags)
2210{
b1d40991 2211 struct fib6_result res = {};
0e09edcc 2212 struct rt6_info *rt = NULL;
1d053da9
DA		 2213 int strict = 0;
2214
0e09edcc
WW		 2215 WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2216 !rcu_read_lock_held());
2217
1d053da9
DA		 2218 strict |= flags & RT6_LOOKUP_F_IFACE;
2219 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2220 if (net->ipv6.devconf_all->forwarding == 0)
2221 strict |= RT6_LOOKUP_F_REACHABLE;
2222
2223 rcu_read_lock();
2224
effda4dd 2225 fib6_table_lookup(net, table, oif, fl6, &res, strict);
0e09edcc
WW		 2226 if (res.f6i == net->ipv6.fib6_null_entry)
2227 goto out;
23fb93a4 2228
b1d40991 2229 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
d83009d4 2230
23fb93a4 2231 /*Search through exception table */
7e4b5128 2232 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
23fb93a4 2233 if (rt) {
0e09edcc 2234 goto out;
3da59bd9 2235 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
b1d40991 2236 !res.nh->fib_nh_gw_family)) {
3da59bd9
MKL		 2237 /* Create a RTF_CACHE clone which will not be
2238 * owned by the fib6 tree. It is for the special case where
2239 * the daddr in the skb during the neighbor look-up is different
2240 * from the fl6->daddr used to look-up route here.
2241 */
0e09edcc 2242 rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
c71099ac 2243
0e09edcc
WW		 2244 if (rt) {
2245 /* 1 refcnt is taken during ip6_rt_cache_alloc().
2246 * As rt6_uncached_list_add() does not consume refcnt,
2247 * this refcnt is always returned to the caller even
2248 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
1cfb71ee 2249 */
0e09edcc 2250 rt6_uncached_list_add(rt);
81eb8447 2251 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
0e09edcc 2252 rcu_read_unlock();
b811580d 2253
0e09edcc
WW		 2254 return rt;
2255 }
d52d3997
MKL		 2256 } else {
2257 /* Get a percpu copy */
951f788a 2258 local_bh_disable();
0e09edcc 2259 rt = rt6_get_pcpu_route(&res);
d52d3997 2260
0e09edcc
WW		 2261 if (!rt)
2262 rt = rt6_make_pcpu_route(net, &res);
93531c67 2263
951f788a 2264 local_bh_enable();
d52d3997 2265 }
0e09edcc
WW		 2266out:
2267 if (!rt)
2268 rt = net->ipv6.ip6_null_entry;
2269 if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2270 ip6_hold_safe(net, &rt);
2271 rcu_read_unlock();
2272
2273 return rt;
1da177e4 2274}
9ff74384 2275EXPORT_SYMBOL_GPL(ip6_pol_route);
1da177e4 2276
b75cc8f9
DA		 2277static struct rt6_info *ip6_pol_route_input(struct net *net,
2278 struct fib6_table *table,
2279 struct flowi6 *fl6,
2280 const struct sk_buff *skb,
2281 int flags)
4acad72d 2282{
b75cc8f9 2283 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
4acad72d
PE		 2284}
2285
d409b847
MB		 2286struct dst_entry *ip6_route_input_lookup(struct net *net,
2287 struct net_device *dev,
b75cc8f9
DA		 2288 struct flowi6 *fl6,
2289 const struct sk_buff *skb,
2290 int flags)
72331bc0
SL		 2291{
2292 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2293 flags |= RT6_LOOKUP_F_IFACE;
2294
b75cc8f9 2295 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
72331bc0 2296}
d409b847 2297EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
72331bc0 2298
23aebdac 2299static void ip6_multipath_l3_keys(const struct sk_buff *skb,
5e5d6fed
RP		 2300 struct flow_keys *keys,
2301 struct flow_keys *flkeys)
23aebdac
JS		 2302{
2303 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2304 const struct ipv6hdr *key_iph = outer_iph;
5e5d6fed 2305 struct flow_keys *_flkeys = flkeys;
23aebdac
JS		 2306 const struct ipv6hdr *inner_iph;
2307 const struct icmp6hdr *icmph;
2308 struct ipv6hdr _inner_iph;
cea67a2d 2309 struct icmp6hdr _icmph;
23aebdac
JS		 2310
2311 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2312 goto out;
2313
cea67a2d
ED		 2314 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2315 sizeof(_icmph), &_icmph);
2316 if (!icmph)
2317 goto out;
2318
54074f1d 2319 if (!icmpv6_is_err(icmph->icmp6_type))
23aebdac
JS		 2320 goto out;
2321
2322 inner_iph = skb_header_pointer(skb,
2323 skb_transport_offset(skb) + sizeof(*icmph),
2324 sizeof(_inner_iph), &_inner_iph);
2325 if (!inner_iph)
2326 goto out;
2327
2328 key_iph = inner_iph;
5e5d6fed 2329 _flkeys = NULL;
23aebdac 2330out:
5e5d6fed
RP		 2331 if (_flkeys) {
2332 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2333 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2334 keys->tags.flow_label = _flkeys->tags.flow_label;
2335 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2336 } else {
2337 keys->addrs.v6addrs.src = key_iph->saddr;
2338 keys->addrs.v6addrs.dst = key_iph->daddr;
fa1be7e0 2339 keys->tags.flow_label = ip6_flowlabel(key_iph);
5e5d6fed
RP		 2340 keys->basic.ip_proto = key_iph->nexthdr;
2341 }
23aebdac
JS		 2342}
2343
2344/* if skb is set it will be used and fl6 can be NULL */
b4bac172
DA		 2345u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2346 const struct sk_buff *skb, struct flow_keys *flkeys)
23aebdac
JS		 2347{
2348 struct flow_keys hash_keys;
9a2a537a 2349 u32 mhash;
23aebdac 2350
bbfa047a 2351 switch (ip6_multipath_hash_policy(net)) {
b4bac172
DA		 2352 case 0:
2353 memset(&hash_keys, 0, sizeof(hash_keys));
2354 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2355 if (skb) {
2356 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2357 } else {
2358 hash_keys.addrs.v6addrs.src = fl6->saddr;
2359 hash_keys.addrs.v6addrs.dst = fl6->daddr;
fa1be7e0 2360 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
b4bac172
DA		 2361 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2362 }
2363 break;
2364 case 1:
2365 if (skb) {
2366 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2367 struct flow_keys keys;
2368
2369 /* short-circuit if we already have L4 hash present */
2370 if (skb->l4_hash)
2371 return skb_get_hash_raw(skb) >> 1;
2372
2373 memset(&hash_keys, 0, sizeof(hash_keys));
2374
2375 if (!flkeys) {
2376 skb_flow_dissect_flow_keys(skb, &keys, flag);
2377 flkeys = &keys;
2378 }
2379 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2380 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2381 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2382 hash_keys.ports.src = flkeys->ports.src;
2383 hash_keys.ports.dst = flkeys->ports.dst;
2384 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2385 } else {
2386 memset(&hash_keys, 0, sizeof(hash_keys));
2387 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2388 hash_keys.addrs.v6addrs.src = fl6->saddr;
2389 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2390 hash_keys.ports.src = fl6->fl6_sport;
2391 hash_keys.ports.dst = fl6->fl6_dport;
2392 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2393 }
2394 break;
d8f74f09
SS		 2395 case 2:
2396 memset(&hash_keys, 0, sizeof(hash_keys));
2397 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2398 if (skb) {
2399 struct flow_keys keys;
2400
2401 if (!flkeys) {
2402 skb_flow_dissect_flow_keys(skb, &keys, 0);
2403 flkeys = &keys;
2404 }
2405
2406 /* Inner can be v4 or v6 */
2407 if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2408 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2409 hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2410 hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2411 } else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2412 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2413 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2414 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2415 hash_keys.tags.flow_label = flkeys->tags.flow_label;
2416 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2417 } else {
2418 /* Same as case 0 */
2419 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2420 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2421 }
2422 } else {
2423 /* Same as case 0 */
2424 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2425 hash_keys.addrs.v6addrs.src = fl6->saddr;
2426 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2427 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2428 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2429 }
2430 break;
23aebdac 2431 }
9a2a537a 2432 mhash = flow_hash_from_keys(&hash_keys);
23aebdac 2433
9a2a537a 2434 return mhash >> 1;
23aebdac
JS		 2435}
2436
67f415dd 2437/* Called with rcu held */
c71099ac
TG		 2438void ip6_route_input(struct sk_buff *skb)
2439{
b71d1d42 2440 const struct ipv6hdr *iph = ipv6_hdr(skb);
c346dca1 2441 struct net *net = dev_net(skb->dev);
67f415dd 2442 int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
904af04d 2443 struct ip_tunnel_info *tun_info;
4c9483b2 2444 struct flowi6 fl6 = {
e0d56fdd 2445 .flowi6_iif = skb->dev->ifindex,
4c9483b2
DM		 2446 .daddr = iph->daddr,
2447 .saddr = iph->saddr,
6502ca52 2448 .flowlabel = ip6_flowinfo(iph),
4c9483b2
DM		 2449 .flowi6_mark = skb->mark,
2450 .flowi6_proto = iph->nexthdr,
c71099ac 2451 };
5e5d6fed 2452 struct flow_keys *flkeys = NULL, _flkeys;
adaa70bb 2453
904af04d 2454 tun_info = skb_tunnel_info(skb);
46fa062a 2455 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
904af04d 2456 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
5e5d6fed
RP		 2457
2458 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2459 flkeys = &_flkeys;
2460
23aebdac 2461 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
b4bac172 2462 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
06e9d040 2463 skb_dst_drop(skb);
67f415dd
WW		 2464 skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2465 &fl6, skb, flags));
c71099ac
TG		 2466}
2467
b75cc8f9
DA		 2468static struct rt6_info *ip6_pol_route_output(struct net *net,
2469 struct fib6_table *table,
2470 struct flowi6 *fl6,
2471 const struct sk_buff *skb,
2472 int flags)
1da177e4 2473{
b75cc8f9 2474 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
c71099ac
TG		 2475}
2476
7d9e5f42
WW		 2477struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2478 const struct sock *sk,
2479 struct flowi6 *fl6, int flags)
c71099ac 2480{
d46a9d67 2481 bool any_src;
c71099ac 2482
3ede0bbc
RS		 2483 if (ipv6_addr_type(&fl6->daddr) &
2484 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
4c1feac5
DA		 2485 struct dst_entry *dst;
2486
7d9e5f42 2487 /* This function does not take refcnt on the dst */
4c1feac5
DA		 2488 dst = l3mdev_link_scope_lookup(net, fl6);
2489 if (dst)
2490 return dst;
2491 }
ca254490 2492
1fb9489b 2493 fl6->flowi6_iif = LOOPBACK_IFINDEX;
4dc27d1c 2494
7d9e5f42 2495 flags |= RT6_LOOKUP_F_DST_NOREF;
d46a9d67 2496 any_src = ipv6_addr_any(&fl6->saddr);
741a11d9 2497 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
d46a9d67 2498 (fl6->flowi6_oif && any_src))
77d16f45 2499 flags |= RT6_LOOKUP_F_IFACE;
c71099ac 2500
d46a9d67 2501 if (!any_src)
adaa70bb 2502 flags |= RT6_LOOKUP_F_HAS_SADDR;
0c9a2ac1
YH		 2503 else if (sk)
2504 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
adaa70bb 2505
b75cc8f9 2506 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
1da177e4 2507}
7d9e5f42
WW		 2508EXPORT_SYMBOL_GPL(ip6_route_output_flags_noref);
2509
2510struct dst_entry *ip6_route_output_flags(struct net *net,
2511 const struct sock *sk,
2512 struct flowi6 *fl6,
2513 int flags)
2514{
2515 struct dst_entry *dst;
2516 struct rt6_info *rt6;
2517
2518 rcu_read_lock();
2519 dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2520 rt6 = (struct rt6_info *)dst;
2521 /* For dst cached in uncached_list, refcnt is already taken. */
2522 if (list_empty(&rt6->rt6i_uncached) && !dst_hold_safe(dst)) {
2523 dst = &net->ipv6.ip6_null_entry->dst;
2524 dst_hold(dst);
2525 }
2526 rcu_read_unlock();
2527
2528 return dst;
2529}
6f21c96a 2530EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1da177e4 2531
2774c131 2532struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
14e50e57 2533{
5c1e6aa3 2534 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1dbe3252 2535 struct net_device *loopback_dev = net->loopback_dev;
14e50e57
DM		 2536 struct dst_entry *new = NULL;
2537
1dbe3252 2538 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
62cf27e5 2539 DST_OBSOLETE_DEAD, 0);
14e50e57 2540 if (rt) {
0a1f5962 2541 rt6_info_init(rt);
81eb8447 2542 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
8104891b 2543
0a1f5962 2544 new = &rt->dst;
14e50e57 2545 new->__use = 1;
352e512c 2546 new->input = dst_discard;
ede2059d 2547 new->output = dst_discard_out;
14e50e57 2548
0a1f5962 2549 dst_copy_metrics(new, &ort->dst);
14e50e57 2550
1dbe3252 2551 rt->rt6i_idev = in6_dev_get(loopback_dev);
4e3fd7a0 2552 rt->rt6i_gateway = ort->rt6i_gateway;
0a1f5962 2553 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
14e50e57
DM		 2554
2555 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2556#ifdef CONFIG_IPV6_SUBTREES
2557 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2558#endif
14e50e57
DM		 2559 }
2560
69ead7af
DM		 2561 dst_release(dst_orig);
2562 return new ? new : ERR_PTR(-ENOMEM);
14e50e57 2563}
14e50e57 2564
1da177e4
LT		 2565/*
2566 * Destination cache support functions
2567 */
2568
8d1c802b 2569static bool fib6_check(struct fib6_info *f6i, u32 cookie)
4b32b5ad 2570{
93531c67
DA		 2571 u32 rt_cookie = 0;
2572
8ae86971 2573 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
93531c67
DA		 2574 return false;
2575
2576 if (fib6_check_expired(f6i))
2577 return false;
2578
2579 return true;
4b32b5ad
MKL		 2580}
2581
a68886a6
DA		 2582static struct dst_entry *rt6_check(struct rt6_info *rt,
2583 struct fib6_info *from,
2584 u32 cookie)
3da59bd9 2585{
36143645 2586 u32 rt_cookie = 0;
c5cff856 2587
49d05fe2 2588 if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
93531c67 2589 rt_cookie != cookie)
3da59bd9
MKL		 2590 return NULL;
2591
2592 if (rt6_check_expired(rt))
2593 return NULL;
2594
2595 return &rt->dst;
2596}
2597
a68886a6
DA		 2598static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2599 struct fib6_info *from,
2600 u32 cookie)
3da59bd9 2601{
5973fb1e
MKL		 2602 if (!__rt6_check_expired(rt) &&
2603 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
a68886a6 2604 fib6_check(from, cookie))
3da59bd9
MKL		 2605 return &rt->dst;
2606 else
2607 return NULL;
2608}
2609
1da177e4
LT		 2610static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
2611{
a87b7dc9 2612 struct dst_entry *dst_ret;
a68886a6 2613 struct fib6_info *from;
1da177e4
LT		 2614 struct rt6_info *rt;
2615
a87b7dc9
DA		 2616 rt = container_of(dst, struct rt6_info, dst);
2617
8f34e53b
DA		 2618 if (rt->sernum)
2619 return rt6_is_valid(rt) ? dst : NULL;
2620
a87b7dc9 2621 rcu_read_lock();
1da177e4 2622
6f3118b5
ND		 2623 /* All IPV6 dsts are created with ->obsolete set to the value
2624 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2625 * into this function always.
2626 */
e3bc10bd 2627
a68886a6 2628 from = rcu_dereference(rt->from);
4b32b5ad 2629
a68886a6
DA		 2630 if (from && (rt->rt6i_flags & RTF_PCPU ||
2631 unlikely(!list_empty(&rt->rt6i_uncached))))
2632 dst_ret = rt6_dst_from_check(rt, from, cookie);
3da59bd9 2633 else
a68886a6 2634 dst_ret = rt6_check(rt, from, cookie);
a87b7dc9
DA		 2635
2636 rcu_read_unlock();
2637
2638 return dst_ret;
1da177e4
LT		 2639}
2640
2641static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2642{
2643 struct rt6_info *rt = (struct rt6_info *) dst;
2644
2645 if (rt) {
54c1a859 2646 if (rt->rt6i_flags & RTF_CACHE) {
c3c14da0 2647 rcu_read_lock();
54c1a859 2648 if (rt6_check_expired(rt)) {
93531c67 2649 rt6_remove_exception_rt(rt);
54c1a859
YH		 2650 dst = NULL;
2651 }
c3c14da0 2652 rcu_read_unlock();
54c1a859 2653 } else {
1da177e4 2654 dst_release(dst);
54c1a859
YH		 2655 dst = NULL;
2656 }
1da177e4 2657 }
54c1a859 2658 return dst;
1da177e4
LT		 2659}
2660
2661static void ip6_link_failure(struct sk_buff *skb)
2662{
2663 struct rt6_info *rt;
2664
3ffe533c 2665 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1da177e4 2666
adf30907 2667 rt = (struct rt6_info *) skb_dst(skb);
1da177e4 2668 if (rt) {
8a14e46f 2669 rcu_read_lock();
1eb4f758 2670 if (rt->rt6i_flags & RTF_CACHE) {
761f6026 2671 rt6_remove_exception_rt(rt);
c5cff856 2672 } else {
a68886a6 2673 struct fib6_info *from;
c5cff856
WW		 2674 struct fib6_node *fn;
2675
a68886a6
DA		 2676 from = rcu_dereference(rt->from);
2677 if (from) {
2678 fn = rcu_dereference(from->fib6_node);
2679 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2680 fn->fn_sernum = -1;
2681 }
1eb4f758 2682 }
8a14e46f 2683 rcu_read_unlock();
1da177e4
LT		 2684 }
2685}
2686
6a3e030f
DA		 2687static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2688{
a68886a6
DA		 2689 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2690 struct fib6_info *from;
2691
2692 rcu_read_lock();
2693 from = rcu_dereference(rt0->from);
2694 if (from)
2695 rt0->dst.expires = from->expires;
2696 rcu_read_unlock();
2697 }
6a3e030f
DA		 2698
2699 dst_set_expires(&rt0->dst, timeout);
2700 rt0->rt6i_flags |= RTF_EXPIRES;
2701}
2702
45e4fd26
MKL		 2703static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2704{
2705 struct net *net = dev_net(rt->dst.dev);
2706
d4ead6b3 2707 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
45e4fd26 2708 rt->rt6i_flags |= RTF_MODIFIED;
45e4fd26
MKL		 2709 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2710}
2711
0d3f6d29
MKL		 2712static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2713{
2714 return !(rt->rt6i_flags & RTF_CACHE) &&
1490ed2a 2715 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
0d3f6d29
MKL		 2716}
2717
45e4fd26 2718static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
bd085ef6
HL		 2719 const struct ipv6hdr *iph, u32 mtu,
2720 bool confirm_neigh)
1da177e4 2721{
0dec879f 2722 const struct in6_addr *daddr, *saddr;
67ba4152 2723 struct rt6_info *rt6 = (struct rt6_info *)dst;
1da177e4 2724
09454fd0
MŻ		 2725 /* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU)
2726 * IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it.
2727 * [see also comment in rt6_mtu_change_route()]
2728 */
19bda36c 2729
0dec879f
JA		 2730 if (iph) {
2731 daddr = &iph->daddr;
2732 saddr = &iph->saddr;
2733 } else if (sk) {
2734 daddr = &sk->sk_v6_daddr;
2735 saddr = &inet6_sk(sk)->saddr;
2736 } else {
2737 daddr = NULL;
2738 saddr = NULL;
2739 }
bd085ef6
HL		 2740
2741 if (confirm_neigh)
2742 dst_confirm_neigh(dst, daddr);
2743
45e4fd26
MKL		 2744 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
2745 if (mtu >= dst_mtu(dst))
2746 return;
9d289715 2747
0d3f6d29 2748 if (!rt6_cache_allowed_for_pmtu(rt6)) {
45e4fd26 2749 rt6_do_update_pmtu(rt6, mtu);
2b760fcf
WW		 2750 /* update rt6_ex->stamp for cache */
2751 if (rt6->rt6i_flags & RTF_CACHE)
2752 rt6_update_exception_stamp_rt(rt6);
0dec879f 2753 } else if (daddr) {
85bd05de 2754 struct fib6_result res = {};
45e4fd26
MKL		 2755 struct rt6_info *nrt6;
2756
4d85cd0c 2757 rcu_read_lock();
85bd05de 2758 res.f6i = rcu_dereference(rt6->from);
43a4b60d
DA		 2759 if (!res.f6i)
2760 goto out_unlock;
2761
7d21fec9
DA		 2762 res.fib6_flags = res.f6i->fib6_flags;
2763 res.fib6_type = res.f6i->fib6_type;
2764
2d44234b
DA		 2765 if (res.f6i->nh) {
2766 struct fib6_nh_match_arg arg = {
2767 .dev = dst->dev,
2768 .gw = &rt6->rt6i_gateway,
2769 };
2770
2771 nexthop_for_each_fib6_nh(res.f6i->nh,
2772 fib6_nh_find_match, &arg);
2773
2774 /* fib6_info uses a nexthop that does not have fib6_nh
2775 * using the dst->dev + gw. Should be impossible.
2776 */
43a4b60d
DA		 2777 if (!arg.match)
2778 goto out_unlock;
2d44234b
DA		 2779
2780 res.nh = arg.match;
2781 } else {
2782 res.nh = res.f6i->fib6_nh;
2783 }
2784
85bd05de 2785 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
45e4fd26
MKL		 2786 if (nrt6) {
2787 rt6_do_update_pmtu(nrt6, mtu);
5012f0a5 2788 if (rt6_insert_exception(nrt6, &res))
2b760fcf 2789 dst_release_immediate(&nrt6->dst);
45e4fd26 2790 }
43a4b60d 2791out_unlock:
a68886a6 2792 rcu_read_unlock();
1da177e4
LT		 2793 }
2794}
2795
45e4fd26 2796static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
bd085ef6
HL		 2797 struct sk_buff *skb, u32 mtu,
2798 bool confirm_neigh)
45e4fd26 2799{
bd085ef6
HL		 2800 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2801 confirm_neigh);
45e4fd26
MKL		 2802}
2803
42ae66c8 2804void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
e2d118a1 2805 int oif, u32 mark, kuid_t uid)
81aded24
DM		 2806{
2807 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2808 struct dst_entry *dst;
dc92095d
MŻ		 2809 struct flowi6 fl6 = {
2810 .flowi6_oif = oif,
2811 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2812 .daddr = iph->daddr,
2813 .saddr = iph->saddr,
2814 .flowlabel = ip6_flowinfo(iph),
2815 .flowi6_uid = uid,
2816 };
81aded24
DM		 2817
2818 dst = ip6_route_output(net, NULL, &fl6);
2819 if (!dst->error)
bd085ef6 2820 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
81aded24
DM		 2821 dst_release(dst);
2822}
2823EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2824
2825void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2826{
7ddacfa5 2827 int oif = sk->sk_bound_dev_if;
33c162a9
MKL		 2828 struct dst_entry *dst;
2829
7ddacfa5
DA		 2830 if (!oif && skb->dev)
2831 oif = l3mdev_master_ifindex(skb->dev);
2832
2833 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
33c162a9
MKL		 2834
2835 dst = __sk_dst_get(sk);
2836 if (!dst || !dst->obsolete ||
2837 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2838 return;
2839
2840 bh_lock_sock(sk);
2841 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2842 ip6_datagram_dst_update(sk, false);
2843 bh_unlock_sock(sk);
81aded24
DM		 2844}
2845EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2846
7d6850f7
AK		 2847void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2848 const struct flowi6 *fl6)
2849{
2850#ifdef CONFIG_IPV6_SUBTREES
2851 struct ipv6_pinfo *np = inet6_sk(sk);
2852#endif
2853
2854 ip6_dst_store(sk, dst,
2855 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2856 &sk->sk_v6_daddr : NULL,
2857#ifdef CONFIG_IPV6_SUBTREES
2858 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2859 &np->saddr :
2860#endif
2861 NULL);
2862}
2863
9b6b35ab 2864static bool ip6_redirect_nh_match(const struct fib6_result *res,
0b34eb00
DA		 2865 struct flowi6 *fl6,
2866 const struct in6_addr *gw,
2867 struct rt6_info **ret)
2868{
9b6b35ab
DA		 2869 const struct fib6_nh *nh = res->nh;
2870
0b34eb00
DA		 2871 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2872 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2873 return false;
2874
2875 /* rt_cache's gateway might be different from its 'parent'
2876 * in the case of an ip redirect.
2877 * So we keep searching in the exception table if the gateway
2878 * is different.
2879 */
2880 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
2881 struct rt6_info *rt_cache;
2882
9b6b35ab 2883 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
0b34eb00
DA		 2884 if (rt_cache &&
2885 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
2886 *ret = rt_cache;
2887 return true;
2888 }
2889 return false;
2890 }
2891 return true;
2892}
2893
c55c8988
DA		 2894struct fib6_nh_rd_arg {
2895 struct fib6_result *res;
2896 struct flowi6 *fl6;
2897 const struct in6_addr *gw;
2898 struct rt6_info **ret;
2899};
2900
2901static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
2902{
2903 struct fib6_nh_rd_arg *arg = _arg;
2904
2905 arg->res->nh = nh;
2906 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
2907}
2908
b55b76b2
DJ		 2909/* Handle redirects */
2910struct ip6rd_flowi {
2911 struct flowi6 fl6;
2912 struct in6_addr gateway;
2913};
2914
2915static struct rt6_info *__ip6_route_redirect(struct net *net,
2916 struct fib6_table *table,
2917 struct flowi6 *fl6,
b75cc8f9 2918 const struct sk_buff *skb,
b55b76b2
DJ		 2919 int flags)
2920{
2921 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
0b34eb00 2922 struct rt6_info *ret = NULL;
9b6b35ab 2923 struct fib6_result res = {};
c55c8988
DA		 2924 struct fib6_nh_rd_arg arg = {
2925 .res = &res,
2926 .fl6 = fl6,
2927 .gw = &rdfl->gateway,
2928 .ret = &ret
2929 };
8d1c802b 2930 struct fib6_info *rt;
b55b76b2
DJ		 2931 struct fib6_node *fn;
2932
31680ac2
DA		 2933 /* l3mdev_update_flow overrides oif if the device is enslaved; in
2934 * this case we must match on the real ingress device, so reset it
2935 */
2936 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2937 fl6->flowi6_oif = skb->dev->ifindex;
2938
b55b76b2 2939 /* Get the "current" route for this destination and
67c408cf 2940 * check if the redirect has come from appropriate router.
b55b76b2
DJ		 2941 *
2942 * RFC 4861 specifies that redirects should only be
2943 * accepted if they come from the nexthop to the target.
2944 * Due to the way the routes are chosen, this notion
2945 * is a bit fuzzy and one might need to check all possible
2946 * routes.
2947 */
2948
66f5d6ce 2949 rcu_read_lock();
6454743b 2950 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
b55b76b2 2951restart:
66f5d6ce 2952 for_each_fib6_node_rt_rcu(fn) {
9b6b35ab 2953 res.f6i = rt;
14895687 2954 if (fib6_check_expired(rt))
b55b76b2 2955 continue;
93c2fb25 2956 if (rt->fib6_flags & RTF_REJECT)
b55b76b2 2957 break;
c55c8988
DA		 2958 if (unlikely(rt->nh)) {
2959 if (nexthop_is_blackhole(rt->nh))
2960 continue;
2961 /* on match, res->nh is filled in and potentially ret */
2962 if (nexthop_for_each_fib6_nh(rt->nh,
2963 fib6_nh_redirect_match,
2964 &arg))
2965 goto out;
2966 } else {
2967 res.nh = rt->fib6_nh;
2968 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
2969 &ret))
2970 goto out;
2971 }
b55b76b2
DJ		 2972 }
2973
2974 if (!rt)
421842ed 2975 rt = net->ipv6.fib6_null_entry;
93c2fb25 2976 else if (rt->fib6_flags & RTF_REJECT) {
23fb93a4 2977 ret = net->ipv6.ip6_null_entry;
b0a1ba59
MKL		 2978 goto out;
2979 }
2980
421842ed 2981 if (rt == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2982 fn = fib6_backtrack(fn, &fl6->saddr);
2983 if (fn)
2984 goto restart;
b55b76b2 2985 }
a3c00e46 2986
9b6b35ab 2987 res.f6i = rt;
1cf844c7 2988 res.nh = rt->fib6_nh;
b0a1ba59 2989out:
7d21fec9 2990 if (ret) {
10585b43 2991 ip6_hold_safe(net, &ret);
7d21fec9
DA		 2992 } else {
2993 res.fib6_flags = res.f6i->fib6_flags;
2994 res.fib6_type = res.f6i->fib6_type;
9b6b35ab 2995 ret = ip6_create_rt_rcu(&res);
7d21fec9 2996 }
b55b76b2 2997
66f5d6ce 2998 rcu_read_unlock();
b55b76b2 2999
8ff2e5b2 3000 trace_fib6_table_lookup(net, &res, table, fl6);
23fb93a4 3001 return ret;
b55b76b2
DJ		 3002};
3003
3004static struct dst_entry *ip6_route_redirect(struct net *net,
b75cc8f9
DA		 3005 const struct flowi6 *fl6,
3006 const struct sk_buff *skb,
3007 const struct in6_addr *gateway)
b55b76b2
DJ		 3008{
3009 int flags = RT6_LOOKUP_F_HAS_SADDR;
3010 struct ip6rd_flowi rdfl;
3011
3012 rdfl.fl6 = *fl6;
3013 rdfl.gateway = *gateway;
3014
b75cc8f9 3015 return fib6_rule_lookup(net, &rdfl.fl6, skb,
b55b76b2
DJ		 3016 flags, __ip6_route_redirect);
3017}
3018
e2d118a1
LC		 3019void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
3020 kuid_t uid)
3a5ad2ee
DM		 3021{
3022 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
3023 struct dst_entry *dst;
1f7f10ac
MŻ		 3024 struct flowi6 fl6 = {
3025 .flowi6_iif = LOOPBACK_IFINDEX,
3026 .flowi6_oif = oif,
3027 .flowi6_mark = mark,
3028 .daddr = iph->daddr,
3029 .saddr = iph->saddr,
3030 .flowlabel = ip6_flowinfo(iph),
3031 .flowi6_uid = uid,
3032 };
3a5ad2ee 3033
b75cc8f9 3034 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
b55b76b2 3035 rt6_do_redirect(dst, NULL, skb);
3a5ad2ee
DM		 3036 dst_release(dst);
3037}
3038EXPORT_SYMBOL_GPL(ip6_redirect);
3039
d456336d 3040void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
c92a59ec
DJ		 3041{
3042 const struct ipv6hdr *iph = ipv6_hdr(skb);
3043 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3044 struct dst_entry *dst;
0b26fb17
MŻ		 3045 struct flowi6 fl6 = {
3046 .flowi6_iif = LOOPBACK_IFINDEX,
3047 .flowi6_oif = oif,
0b26fb17
MŻ		 3048 .daddr = msg->dest,
3049 .saddr = iph->daddr,
3050 .flowi6_uid = sock_net_uid(net, NULL),
3051 };
c92a59ec 3052
b75cc8f9 3053 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
b55b76b2 3054 rt6_do_redirect(dst, NULL, skb);
c92a59ec
DJ		 3055 dst_release(dst);
3056}
3057
3a5ad2ee
DM		 3058void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3059{
e2d118a1
LC		 3060 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
3061 sk->sk_uid);
3a5ad2ee
DM		 3062}
3063EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3064
0dbaee3b 3065static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1da177e4 3066{
0dbaee3b
DM		 3067 struct net_device *dev = dst->dev;
3068 unsigned int mtu = dst_mtu(dst);
3069 struct net *net = dev_net(dev);
3070
1da177e4
LT		 3071 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3072
5578689a
DL		 3073 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3074 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1da177e4
LT		 3075
3076 /*
1ab1457c
YH		 3077 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3078 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3079 * IPV6_MAXPLEN is also valid and means: "any MSS,
1da177e4
LT		 3080 * rely only on pmtu discovery"
3081 */
3082 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3083 mtu = IPV6_MAXPLEN;
3084 return mtu;
3085}
3086
ebb762f2 3087static unsigned int ip6_mtu(const struct dst_entry *dst)
d33e4553 3088{
d33e4553 3089 struct inet6_dev *idev;
d4ead6b3 3090 unsigned int mtu;
4b32b5ad
MKL		 3091
3092 mtu = dst_metric_raw(dst, RTAX_MTU);
618f9bc7 3093 if (mtu)
30f78d8e 3094 goto out;
618f9bc7
SK		 3095
3096 mtu = IPV6_MIN_MTU;
d33e4553
DM		 3097
3098 rcu_read_lock();
3099 idev = __in6_dev_get(dst->dev);
3100 if (idev)
3101 mtu = idev->cnf.mtu6;
3102 rcu_read_unlock();
3103
30f78d8e 3104out:
14972cbd
RP		 3105 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3106
3107 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
d33e4553
DM		 3108}
3109
901731b8
DA		 3110/* MTU selection:
3111 * 1. mtu on route is locked - use it
3112 * 2. mtu from nexthop exception
3113 * 3. mtu from egress device
3114 *
3115 * based on ip6_dst_mtu_forward and exception logic of
3116 * rt6_find_cached_rt; called with rcu_read_lock
3117 */
b748f260
DA		 3118u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3119 const struct in6_addr *daddr,
3120 const struct in6_addr *saddr)
901731b8 3121{
b748f260
DA		 3122 const struct fib6_nh *nh = res->nh;
3123 struct fib6_info *f6i = res->f6i;
901731b8 3124 struct inet6_dev *idev;
510e2ced 3125 struct rt6_info *rt;
901731b8
DA		 3126 u32 mtu = 0;
3127
3128 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3129 mtu = f6i->fib6_pmtu;
3130 if (mtu)
3131 goto out;
3132 }
3133
510e2ced
WW		 3134 rt = rt6_find_cached_rt(res, daddr, saddr);
3135 if (unlikely(rt)) {
3136 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3137 } else {
b748f260 3138 struct net_device *dev = nh->fib_nh_dev;
901731b8
DA		 3139
3140 mtu = IPV6_MIN_MTU;
3141 idev = __in6_dev_get(dev);
3142 if (idev && idev->cnf.mtu6 > mtu)
3143 mtu = idev->cnf.mtu6;
3144 }
3145
3146 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3147out:
b748f260 3148 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
901731b8
DA		 3149}
3150
3b00944c 3151struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
87a11578 3152 struct flowi6 *fl6)
1da177e4 3153{
87a11578 3154 struct dst_entry *dst;
1da177e4
LT		 3155 struct rt6_info *rt;
3156 struct inet6_dev *idev = in6_dev_get(dev);
c346dca1 3157 struct net *net = dev_net(dev);
1da177e4 3158
38308473 3159 if (unlikely(!idev))
122bdf67 3160 return ERR_PTR(-ENODEV);
1da177e4 3161
ad706862 3162 rt = ip6_dst_alloc(net, dev, 0);
38308473 3163 if (unlikely(!rt)) {
1da177e4 3164 in6_dev_put(idev);
87a11578 3165 dst = ERR_PTR(-ENOMEM);
1da177e4
LT		 3166 goto out;
3167 }
3168
588753f1 3169 rt->dst.input = ip6_input;
8e2ec639 3170 rt->dst.output = ip6_output;
550bab42 3171 rt->rt6i_gateway = fl6->daddr;
87a11578 3172 rt->rt6i_dst.addr = fl6->daddr;
8e2ec639
YZ		 3173 rt->rt6i_dst.plen = 128;
3174 rt->rt6i_idev = idev;
14edd87d 3175 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1da177e4 3176
4c981e28 3177 /* Add this dst into uncached_list so that rt6_disable_ip() can
587fea74
WW		 3178 * do proper release of the net_device
3179 */
3180 rt6_uncached_list_add(rt);
81eb8447 3181 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
1da177e4 3182
87a11578
DM		 3183 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3184
1da177e4 3185out:
87a11578 3186 return dst;
1da177e4
LT		 3187}
3188
569d3645 3189static int ip6_dst_gc(struct dst_ops *ops)
1da177e4 3190{
86393e52 3191 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
7019b78e
DL		 3192 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3193 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3194 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3195 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3196 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
fc66f95c 3197 int entries;
7019b78e 3198
fc66f95c 3199 entries = dst_entries_get_fast(ops);
cf86a086
ED		 3200 if (entries > rt_max_size)
3201 entries = dst_entries_get_slow(ops);
3202
49a18d86 3203 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
fc66f95c 3204 entries <= rt_max_size)
1da177e4
LT		 3205 goto out;
3206
6891a346 3207 net->ipv6.ip6_rt_gc_expire++;
14956643 3208 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
fc66f95c
ED		 3209 entries = dst_entries_get_slow(ops);
3210 if (entries < ops->gc_thresh)
7019b78e 3211 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1da177e4 3212out:
7019b78e 3213 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
fc66f95c 3214 return entries > rt_max_size;
1da177e4
LT		 3215}
3216
b2c709cc
DA		 3217static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3218 const struct in6_addr *gw_addr, u32 tbid,
3219 int flags, struct fib6_result *res)
8c14586f
DA		 3220{
3221 struct flowi6 fl6 = {
3222 .flowi6_oif = cfg->fc_ifindex,
3223 .daddr = *gw_addr,
3224 .saddr = cfg->fc_prefsrc,
3225 };
3226 struct fib6_table *table;
b2c709cc 3227 int err;
8c14586f 3228
f4797b33 3229 table = fib6_get_table(net, tbid);
8c14586f 3230 if (!table)
b2c709cc 3231 return -EINVAL;
8c14586f
DA		 3232
3233 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3234 flags |= RT6_LOOKUP_F_HAS_SADDR;
3235
f4797b33 3236 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
8c14586f 3237
b2c709cc
DA		 3238 err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3239 if (!err && res->f6i != net->ipv6.fib6_null_entry)
3240 fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3241 cfg->fc_ifindex != 0, NULL, flags);
8c14586f 3242
b2c709cc 3243 return err;
8c14586f
DA		 3244}
3245
fc1e64e1
DA		 3246static int ip6_route_check_nh_onlink(struct net *net,
3247 struct fib6_config *cfg,
9fbb704c 3248 const struct net_device *dev,
fc1e64e1
DA		 3249 struct netlink_ext_ack *extack)
3250{
b2c709cc 3251 u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
fc1e64e1 3252 const struct in6_addr *gw_addr = &cfg->fc_gateway;
b2c709cc 3253 struct fib6_result res = {};
fc1e64e1
DA		 3254 int err;
3255
b2c709cc
DA		 3256 err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3257 if (!err && !(res.fib6_flags & RTF_REJECT) &&
3258 /* ignore match if it is the default route */
3259 !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3260 (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3261 NL_SET_ERR_MSG(extack,
3262 "Nexthop has invalid gateway or device mismatch");
3263 err = -EINVAL;
fc1e64e1
DA		 3264 }
3265
3266 return err;
3267}
3268
1edce99f
DA		 3269static int ip6_route_check_nh(struct net *net,
3270 struct fib6_config *cfg,
3271 struct net_device **_dev,
3272 struct inet6_dev **idev)
3273{
3274 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3275 struct net_device *dev = _dev ? *_dev : NULL;
b2c709cc
DA		 3276 int flags = RT6_LOOKUP_F_IFACE;
3277 struct fib6_result res = {};
1edce99f
DA		 3278 int err = -EHOSTUNREACH;
3279
3280 if (cfg->fc_table) {
b2c709cc
DA		 3281 err = ip6_nh_lookup_table(net, cfg, gw_addr,
3282 cfg->fc_table, flags, &res);
3283 /* gw_addr can not require a gateway or resolve to a reject
3284 * route. If a device is given, it must match the result.
3285 */
3286 if (err || res.fib6_flags & RTF_REJECT ||
3287 res.nh->fib_nh_gw_family ||
3288 (dev && dev != res.nh->fib_nh_dev))
3289 err = -EHOSTUNREACH;
1edce99f
DA		 3290 }
3291
b2c709cc
DA		 3292 if (err < 0) {
3293 struct flowi6 fl6 = {
3294 .flowi6_oif = cfg->fc_ifindex,
3295 .daddr = *gw_addr,
3296 };
1edce99f 3297
b2c709cc
DA		 3298 err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3299 if (err || res.fib6_flags & RTF_REJECT ||
3300 res.nh->fib_nh_gw_family)
3301 err = -EHOSTUNREACH;
3302
3303 if (err)
3304 return err;
3305
3306 fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3307 cfg->fc_ifindex != 0, NULL, flags);
3308 }
1edce99f 3309
b2c709cc 3310 err = 0;
1edce99f 3311 if (dev) {
b2c709cc
DA		 3312 if (dev != res.nh->fib_nh_dev)
3313 err = -EHOSTUNREACH;
1edce99f 3314 } else {
b2c709cc 3315 *_dev = dev = res.nh->fib_nh_dev;
1edce99f 3316 dev_hold(dev);
b2c709cc 3317 *idev = in6_dev_get(dev);
1edce99f
DA		 3318 }
3319
1edce99f
DA		 3320 return err;
3321}
3322
9fbb704c
DA		 3323static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3324 struct net_device **_dev, struct inet6_dev **idev,
3325 struct netlink_ext_ack *extack)
3326{
3327 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3328 int gwa_type = ipv6_addr_type(gw_addr);
232378e8 3329 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
9fbb704c 3330 const struct net_device *dev = *_dev;
232378e8 3331 bool need_addr_check = !dev;
9fbb704c
DA		 3332 int err = -EINVAL;
3333
3334 /* if gw_addr is local we will fail to detect this in case
3335 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3336 * will return already-added prefix route via interface that
3337 * prefix route was assigned to, which might be non-loopback.
3338 */
232378e8
DA		 3339 if (dev &&
3340 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3341 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
9fbb704c
DA		 3342 goto out;
3343 }
3344
3345 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3346 /* IPv6 strictly inhibits using not link-local
3347 * addresses as nexthop address.
3348 * Otherwise, router will not able to send redirects.
3349 * It is very good, but in some (rare!) circumstances
3350 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3351 * some exceptions. --ANK
3352 * We allow IPv4-mapped nexthops to support RFC4798-type
3353 * addressing
3354 */
3355 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3356 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3357 goto out;
3358 }
3359
b2c709cc
DA		 3360 rcu_read_lock();
3361
9fbb704c
DA		 3362 if (cfg->fc_flags & RTNH_F_ONLINK)
3363 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3364 else
3365 err = ip6_route_check_nh(net, cfg, _dev, idev);
3366
b2c709cc
DA		 3367 rcu_read_unlock();
3368
9fbb704c
DA		 3369 if (err)
3370 goto out;
3371 }
3372
3373 /* reload in case device was changed */
3374 dev = *_dev;
3375
3376 err = -EINVAL;
3377 if (!dev) {
3378 NL_SET_ERR_MSG(extack, "Egress device not specified");
3379 goto out;
3380 } else if (dev->flags & IFF_LOOPBACK) {
3381 NL_SET_ERR_MSG(extack,
3382 "Egress device can not be loopback device for this route");
3383 goto out;
3384 }
232378e8
DA		 3385
3386 /* if we did not check gw_addr above, do so now that the
3387 * egress device has been resolved.
3388 */
3389 if (need_addr_check &&
3390 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3391 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3392 goto out;
3393 }
3394
9fbb704c
DA		 3395 err = 0;
3396out:
3397 return err;
3398}
3399
83c44251
DA		 3400static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3401{
3402 if ((flags & RTF_REJECT) ||
3403 (dev && (dev->flags & IFF_LOOPBACK) &&
3404 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3405 !(flags & RTF_LOCAL)))
3406 return true;
3407
3408 return false;
3409}
3410
3411int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3412 struct fib6_config *cfg, gfp_t gfp_flags,
3413 struct netlink_ext_ack *extack)
3414{
3415 struct net_device *dev = NULL;
3416 struct inet6_dev *idev = NULL;
3417 int addr_type;
3418 int err;
3419
f1741730 3420 fib6_nh->fib_nh_family = AF_INET6;
1bef4c22
ED		 3421#ifdef CONFIG_IPV6_ROUTER_PREF
3422 fib6_nh->last_probe = jiffies;
3423#endif
f1741730 3424
83c44251
DA		 3425 err = -ENODEV;
3426 if (cfg->fc_ifindex) {
3427 dev = dev_get_by_index(net, cfg->fc_ifindex);
3428 if (!dev)
3429 goto out;
3430 idev = in6_dev_get(dev);
3431 if (!idev)
3432 goto out;
3433 }
3434
3435 if (cfg->fc_flags & RTNH_F_ONLINK) {
3436 if (!dev) {
3437 NL_SET_ERR_MSG(extack,
3438 "Nexthop device required for onlink");
3439 goto out;
3440 }
3441
3442 if (!(dev->flags & IFF_UP)) {
3443 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3444 err = -ENETDOWN;
3445 goto out;
3446 }
3447
ad1601ae 3448 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
83c44251
DA		 3449 }
3450
ad1601ae 3451 fib6_nh->fib_nh_weight = 1;
83c44251
DA		 3452
3453 /* We cannot add true routes via loopback here,
3454 * they would result in kernel looping; promote them to reject routes
3455 */
3456 addr_type = ipv6_addr_type(&cfg->fc_dst);
3457 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3458 /* hold loopback dev/idev if we haven't done so. */
3459 if (dev != net->loopback_dev) {
3460 if (dev) {
3461 dev_put(dev);
3462 in6_dev_put(idev);
3463 }
3464 dev = net->loopback_dev;
3465 dev_hold(dev);
3466 idev = in6_dev_get(dev);
3467 if (!idev) {
3468 err = -ENODEV;
3469 goto out;
3470 }
3471 }
7dd73168 3472 goto pcpu_alloc;
83c44251
DA		 3473 }
3474
3475 if (cfg->fc_flags & RTF_GATEWAY) {
3476 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3477 if (err)
3478 goto out;
3479
ad1601ae 3480 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
bdf00467 3481 fib6_nh->fib_nh_gw_family = AF_INET6;
83c44251
DA		 3482 }
3483
3484 err = -ENODEV;
3485 if (!dev)
3486 goto out;
3487
3488 if (idev->cnf.disable_ipv6) {
3489 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3490 err = -EACCES;
3491 goto out;
3492 }
3493
3494 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3495 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3496 err = -ENETDOWN;
3497 goto out;
3498 }
3499
3500 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3501 !netif_carrier_ok(dev))
ad1601ae 3502 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
83c44251 3503
faee6769 3504 err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
7dd73168
DA		 3505 cfg->fc_encap_type, cfg, gfp_flags, extack);
3506 if (err)
3507 goto out;
3508
3509pcpu_alloc:
f40b6ae2
DA		 3510 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3511 if (!fib6_nh->rt6i_pcpu) {
3512 err = -ENOMEM;
3513 goto out;
3514 }
3515
ad1601ae 3516 fib6_nh->fib_nh_dev = dev;
f1741730 3517 fib6_nh->fib_nh_oif = dev->ifindex;
83c44251
DA		 3518 err = 0;
3519out:
3520 if (idev)
3521 in6_dev_put(idev);
3522
3523 if (err) {
ad1601ae
DA		 3524 lwtstate_put(fib6_nh->fib_nh_lws);
3525 fib6_nh->fib_nh_lws = NULL;
83c44251
DA		 3526 if (dev)
3527 dev_put(dev);
3528 }
3529
3530 return err;
3531}
3532
dac7d0f2
DA		 3533void fib6_nh_release(struct fib6_nh *fib6_nh)
3534{
cc5c073a
DA		 3535 struct rt6_exception_bucket *bucket;
3536
3537 rcu_read_lock();
3538
3539 fib6_nh_flush_exceptions(fib6_nh, NULL);
3540 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3541 if (bucket) {
3542 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3543 kfree(bucket);
3544 }
3545
3546 rcu_read_unlock();
3547
f40b6ae2
DA		 3548 if (fib6_nh->rt6i_pcpu) {
3549 int cpu;
3550
3551 for_each_possible_cpu(cpu) {
3552 struct rt6_info **ppcpu_rt;
3553 struct rt6_info *pcpu_rt;
3554
3555 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3556 pcpu_rt = *ppcpu_rt;
3557 if (pcpu_rt) {
3558 dst_dev_put(&pcpu_rt->dst);
3559 dst_release(&pcpu_rt->dst);
3560 *ppcpu_rt = NULL;
3561 }
3562 }
3563
3564 free_percpu(fib6_nh->rt6i_pcpu);
3565 }
3566
979e276e 3567 fib_nh_common_release(&fib6_nh->nh_common);
dac7d0f2
DA		 3568}
3569
8d1c802b 3570static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
acb54e3c 3571 gfp_t gfp_flags,
333c4301 3572 struct netlink_ext_ack *extack)
1da177e4 3573{
5578689a 3574 struct net *net = cfg->fc_nlinfo.nl_net;
8d1c802b 3575 struct fib6_info *rt = NULL;
f88d8ea6 3576 struct nexthop *nh = NULL;
c71099ac 3577 struct fib6_table *table;
f88d8ea6 3578 struct fib6_nh *fib6_nh;
8c5b83f0 3579 int err = -EINVAL;
83c44251 3580 int addr_type;
1da177e4 3581
557c44be 3582 /* RTF_PCPU is an internal flag; can not be set by userspace */
d5d531cb
DA		 3583 if (cfg->fc_flags & RTF_PCPU) {
3584 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
557c44be 3585 goto out;
d5d531cb 3586 }
557c44be 3587
2ea2352e
WW		 3588 /* RTF_CACHE is an internal flag; can not be set by userspace */
3589 if (cfg->fc_flags & RTF_CACHE) {
3590 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3591 goto out;
3592 }
3593
e8478e80
DA		 3594 if (cfg->fc_type > RTN_MAX) {
3595 NL_SET_ERR_MSG(extack, "Invalid route type");
3596 goto out;
3597 }
3598
d5d531cb
DA		 3599 if (cfg->fc_dst_len > 128) {
3600 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3601 goto out;
3602 }
3603 if (cfg->fc_src_len > 128) {
3604 NL_SET_ERR_MSG(extack, "Invalid source address length");
8c5b83f0 3605 goto out;
d5d531cb 3606 }
1da177e4 3607#ifndef CONFIG_IPV6_SUBTREES
d5d531cb
DA		 3608 if (cfg->fc_src_len) {
3609 NL_SET_ERR_MSG(extack,
3610 "Specifying source address requires IPV6_SUBTREES to be enabled");
8c5b83f0 3611 goto out;
d5d531cb 3612 }
1da177e4 3613#endif
5b98324e
DA		 3614 if (cfg->fc_nh_id) {
3615 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3616 if (!nh) {
3617 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3618 goto out;
3619 }
3620 err = fib6_check_nexthop(nh, cfg, extack);
3621 if (err)
3622 goto out;
3623 }
fc1e64e1 3624
d71314b4 3625 err = -ENOBUFS;
38308473
DM		 3626 if (cfg->fc_nlinfo.nlh &&
3627 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
d71314b4 3628 table = fib6_get_table(net, cfg->fc_table);
38308473 3629 if (!table) {
f3213831 3630 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
d71314b4
MV		 3631 table = fib6_new_table(net, cfg->fc_table);
3632 }
3633 } else {
3634 table = fib6_new_table(net, cfg->fc_table);
3635 }
38308473
DM		 3636
3637 if (!table)
c71099ac 3638 goto out;
c71099ac 3639
93531c67 3640 err = -ENOMEM;
f88d8ea6 3641 rt = fib6_info_alloc(gfp_flags, !nh);
93531c67 3642 if (!rt)
1da177e4 3643 goto out;
93531c67 3644
d7e774f3
DA		 3645 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3646 extack);
767a2217
DA		 3647 if (IS_ERR(rt->fib6_metrics)) {
3648 err = PTR_ERR(rt->fib6_metrics);
fda21d46
ED		 3649 /* Do not leave garbage there. */
3650 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
767a2217
DA		 3651 goto out;
3652 }
3653
93531c67
DA		 3654 if (cfg->fc_flags & RTF_ADDRCONF)
3655 rt->dst_nocount = true;
1da177e4 3656
1716a961 3657 if (cfg->fc_flags & RTF_EXPIRES)
14895687 3658 fib6_set_expires(rt, jiffies +
1716a961
G		 3659 clock_t_to_jiffies(cfg->fc_expires));
3660 else
14895687 3661 fib6_clean_expires(rt);
1da177e4 3662
86872cb5
TG		 3663 if (cfg->fc_protocol == RTPROT_UNSPEC)
3664 cfg->fc_protocol = RTPROT_BOOT;
93c2fb25 3665 rt->fib6_protocol = cfg->fc_protocol;
86872cb5 3666
83c44251
DA		 3667 rt->fib6_table = table;
3668 rt->fib6_metric = cfg->fc_metric;
c7036d97 3669 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
2b2450ca 3670 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
19e42e45 3671
93c2fb25
DA		 3672 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3673 rt->fib6_dst.plen = cfg->fc_dst_len;
e5fd387a 3674
1da177e4 3675#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 3676 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3677 rt->fib6_src.plen = cfg->fc_src_len;
1da177e4 3678#endif
f88d8ea6
DA		 3679 if (nh) {
3680 if (!nexthop_get(nh)) {
3681 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3682 goto out;
3683 }
3684 if (rt->fib6_src.plen) {
4daa95af 3685 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
f88d8ea6
DA		 3686 goto out;
3687 }
3688 rt->nh = nh;
3689 fib6_nh = nexthop_fib6_nh(rt->nh);
3690 } else {
3691 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3692 if (err)
3693 goto out;
1da177e4 3694
f88d8ea6
DA		 3695 fib6_nh = rt->fib6_nh;
3696
3697 /* We cannot add true routes via loopback here, they would
3698 * result in kernel looping; promote them to reject routes
3699 */
3700 addr_type = ipv6_addr_type(&cfg->fc_dst);
3701 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3702 addr_type))
3703 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3704 }
955ec4cb 3705
c3968a85 3706 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
f88d8ea6 3707 struct net_device *dev = fib6_nh->fib_nh_dev;
83c44251 3708
c3968a85 3709 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
d5d531cb 3710 NL_SET_ERR_MSG(extack, "Invalid source address");
c3968a85
DW		 3711 err = -EINVAL;
3712 goto out;
3713 }
93c2fb25
DA		 3714 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3715 rt->fib6_prefsrc.plen = 128;
c3968a85 3716 } else
93c2fb25 3717 rt->fib6_prefsrc.plen = 0;
c3968a85 3718
8c5b83f0 3719 return rt;
6b9ea5a6 3720out:
93531c67 3721 fib6_info_release(rt);
8c5b83f0 3722 return ERR_PTR(err);
6b9ea5a6
RP		 3723}
3724
acb54e3c 3725int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
333c4301 3726 struct netlink_ext_ack *extack)
6b9ea5a6 3727{
8d1c802b 3728 struct fib6_info *rt;
6b9ea5a6
RP		 3729 int err;
3730
acb54e3c 3731 rt = ip6_route_info_create(cfg, gfp_flags, extack);
d4ead6b3
DA		 3732 if (IS_ERR(rt))
3733 return PTR_ERR(rt);
6b9ea5a6 3734
d4ead6b3 3735 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
93531c67 3736 fib6_info_release(rt);
6b9ea5a6 3737
1da177e4
LT		 3738 return err;
3739}
3740
8d1c802b 3741static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
1da177e4 3742{
afb1d4b5 3743 struct net *net = info->nl_net;
c71099ac 3744 struct fib6_table *table;
afb1d4b5 3745 int err;
1da177e4 3746
421842ed 3747 if (rt == net->ipv6.fib6_null_entry) {
6825a26c
G		 3748 err = -ENOENT;
3749 goto out;
3750 }
6c813a72 3751
93c2fb25 3752 table = rt->fib6_table;
66f5d6ce 3753 spin_lock_bh(&table->tb6_lock);
86872cb5 3754 err = fib6_del(rt, info);
66f5d6ce 3755 spin_unlock_bh(&table->tb6_lock);
1da177e4 3756
6825a26c 3757out:
93531c67 3758 fib6_info_release(rt);
1da177e4
LT		 3759 return err;
3760}
3761
11dd74b3 3762int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
e0a1ad73 3763{
11dd74b3
RP		 3764 struct nl_info info = {
3765 .nl_net = net,
3766 .skip_notify = skip_notify
3767 };
afb1d4b5 3768
528c4ceb 3769 return __ip6_del_rt(rt, &info);
e0a1ad73
TG		 3770}
3771
8d1c802b 3772static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
0ae81335
DA		 3773{
3774 struct nl_info *info = &cfg->fc_nlinfo;
e3330039 3775 struct net *net = info->nl_net;
16a16cd3 3776 struct sk_buff *skb = NULL;
0ae81335 3777 struct fib6_table *table;
e3330039 3778 int err = -ENOENT;
0ae81335 3779
421842ed 3780 if (rt == net->ipv6.fib6_null_entry)
e3330039 3781 goto out_put;
93c2fb25 3782 table = rt->fib6_table;
66f5d6ce 3783 spin_lock_bh(&table->tb6_lock);
0ae81335 3784
93c2fb25 3785 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
8d1c802b 3786 struct fib6_info *sibling, *next_sibling;
0284696b 3787 struct fib6_node *fn;
0ae81335 3788
16a16cd3
DA		 3789 /* prefer to send a single notification with all hops */
3790 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3791 if (skb) {
3792 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3793
d4ead6b3 3794 if (rt6_fill_node(net, skb, rt, NULL,
16a16cd3
DA		 3795 NULL, NULL, 0, RTM_DELROUTE,
3796 info->portid, seq, 0) < 0) {
3797 kfree_skb(skb);
3798 skb = NULL;
3799 } else
3800 info->skip_notify = 1;
3801 }
3802
0284696b
IS		 3803 /* 'rt' points to the first sibling route. If it is not the
3804 * leaf, then we do not need to send a notification. Otherwise,
3805 * we need to check if the last sibling has a next route or not
3806 * and emit a replace or delete notification, respectively.
3807 */
2881fd61 3808 info->skip_notify_kernel = 1;
0284696b
IS		 3809 fn = rcu_dereference_protected(rt->fib6_node,
3810 lockdep_is_held(&table->tb6_lock));
3811 if (rcu_access_pointer(fn->leaf) == rt) {
3812 struct fib6_info *last_sibling, *replace_rt;
3813
3814 last_sibling = list_last_entry(&rt->fib6_siblings,
3815 struct fib6_info,
3816 fib6_siblings);
3817 replace_rt = rcu_dereference_protected(
3818 last_sibling->fib6_next,
3819 lockdep_is_held(&table->tb6_lock));
3820 if (replace_rt)
3821 call_fib6_entry_notifiers_replace(net,
3822 replace_rt);
3823 else
3824 call_fib6_multipath_entry_notifiers(net,
caafb250 3825 FIB_EVENT_ENTRY_DEL,
0284696b
IS		 3826 rt, rt->fib6_nsiblings,
3827 NULL);
3828 }
0ae81335 3829 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25
DA		 3830 &rt->fib6_siblings,
3831 fib6_siblings) {
0ae81335
DA		 3832 err = fib6_del(sibling, info);
3833 if (err)
e3330039 3834 goto out_unlock;
0ae81335
DA		 3835 }
3836 }
3837
3838 err = fib6_del(rt, info);
e3330039 3839out_unlock:
66f5d6ce 3840 spin_unlock_bh(&table->tb6_lock);
e3330039 3841out_put:
93531c67 3842 fib6_info_release(rt);
16a16cd3
DA		 3843
3844 if (skb) {
e3330039 3845 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
16a16cd3
DA		 3846 info->nlh, gfp_any());
3847 }
0ae81335
DA		 3848 return err;
3849}
3850
0fa6efc5 3851static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
23fb93a4
DA		 3852{
3853 int rc = -ESRCH;
3854
3855 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3856 goto out;
3857
3858 if (cfg->fc_flags & RTF_GATEWAY &&
3859 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3860 goto out;
761f6026
XL		 3861
3862 rc = rt6_remove_exception_rt(rt);
23fb93a4
DA		 3863out:
3864 return rc;
3865}
3866
0fa6efc5
DA		 3867static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3868 struct fib6_nh *nh)
3869{
3870 struct fib6_result res = {
3871 .f6i = rt,
3872 .nh = nh,
3873 };
3874 struct rt6_info *rt_cache;
3875
3876 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3877 if (rt_cache)
3878 return __ip6_del_cached_rt(rt_cache, cfg);
3879
3880 return 0;
3881}
3882
5b98324e
DA		 3883struct fib6_nh_del_cached_rt_arg {
3884 struct fib6_config *cfg;
3885 struct fib6_info *f6i;
3886};
3887
3888static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
3889{
3890 struct fib6_nh_del_cached_rt_arg *arg = _arg;
3891 int rc;
3892
3893 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
3894 return rc != -ESRCH ? rc : 0;
3895}
3896
3897static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
3898{
3899 struct fib6_nh_del_cached_rt_arg arg = {
3900 .cfg = cfg,
3901 .f6i = f6i
3902 };
3903
3904 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
3905}
3906
333c4301
DA		 3907static int ip6_route_del(struct fib6_config *cfg,
3908 struct netlink_ext_ack *extack)
1da177e4 3909{
c71099ac 3910 struct fib6_table *table;
8d1c802b 3911 struct fib6_info *rt;
1da177e4 3912 struct fib6_node *fn;
1da177e4
LT		 3913 int err = -ESRCH;
3914
5578689a 3915 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
d5d531cb
DA		 3916 if (!table) {
3917 NL_SET_ERR_MSG(extack, "FIB table does not exist");
c71099ac 3918 return err;
d5d531cb 3919 }
c71099ac 3920
66f5d6ce 3921 rcu_read_lock();
1da177e4 3922
c71099ac 3923 fn = fib6_locate(&table->tb6_root,
86872cb5 3924 &cfg->fc_dst, cfg->fc_dst_len,
38fbeeee 3925 &cfg->fc_src, cfg->fc_src_len,
2b760fcf 3926 !(cfg->fc_flags & RTF_CACHE));
1ab1457c 3927
1da177e4 3928 if (fn) {
66f5d6ce 3929 for_each_fib6_node_rt_rcu(fn) {
ad1601ae
DA		 3930 struct fib6_nh *nh;
3931
3401bfb1
SB		 3932 if (rt->nh && cfg->fc_nh_id &&
3933 rt->nh->id != cfg->fc_nh_id)
5b98324e 3934 continue;
23fb93a4 3935
5b98324e
DA		 3936 if (cfg->fc_flags & RTF_CACHE) {
3937 int rc = 0;
3938
3939 if (rt->nh) {
3940 rc = ip6_del_cached_rt_nh(cfg, rt);
3941 } else if (cfg->fc_nh_id) {
3942 continue;
3943 } else {
3944 nh = rt->fib6_nh;
3945 rc = ip6_del_cached_rt(cfg, rt, nh);
3946 }
0fa6efc5
DA		 3947 if (rc != -ESRCH) {
3948 rcu_read_unlock();
3949 return rc;
23fb93a4
DA		 3950 }
3951 continue;
2b760fcf 3952 }
ad1601ae 3953
5b98324e
DA		 3954 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
3955 continue;
3956 if (cfg->fc_protocol &&
3957 cfg->fc_protocol != rt->fib6_protocol)
3958 continue;
3959
3960 if (rt->nh) {
3961 if (!fib6_info_hold_safe(rt))
3962 continue;
3963 rcu_read_unlock();
3964
3965 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3966 }
3967 if (cfg->fc_nh_id)
3968 continue;
3969
3970 nh = rt->fib6_nh;
86872cb5 3971 if (cfg->fc_ifindex &&
ad1601ae
DA		 3972 (!nh->fib_nh_dev ||
3973 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
1da177e4 3974 continue;
86872cb5 3975 if (cfg->fc_flags & RTF_GATEWAY &&
ad1601ae 3976 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
1da177e4 3977 continue;
e873e4b9
WW		 3978 if (!fib6_info_hold_safe(rt))
3979 continue;
66f5d6ce 3980 rcu_read_unlock();
1da177e4 3981
0ae81335
DA		 3982 /* if gateway was specified only delete the one hop */
3983 if (cfg->fc_flags & RTF_GATEWAY)
3984 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3985
3986 return __ip6_del_rt_siblings(rt, cfg);
1da177e4
LT		 3987 }
3988 }
66f5d6ce 3989 rcu_read_unlock();
1da177e4
LT		 3990
3991 return err;
3992}
3993
6700c270 3994static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
a6279458 3995{
a6279458 3996 struct netevent_redirect netevent;
e8599ff4 3997 struct rt6_info *rt, *nrt = NULL;
85bd05de 3998 struct fib6_result res = {};
e8599ff4
DM		 3999 struct ndisc_options ndopts;
4000 struct inet6_dev *in6_dev;
4001 struct neighbour *neigh;
71bcdba0 4002 struct rd_msg *msg;
6e157b6a
DM		 4003 int optlen, on_link;
4004 u8 *lladdr;
e8599ff4 4005
29a3cad5 4006 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
71bcdba0 4007 optlen -= sizeof(*msg);
e8599ff4
DM		 4008
4009 if (optlen < 0) {
6e157b6a 4010 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
e8599ff4
DM		 4011 return;
4012 }
4013
71bcdba0 4014 msg = (struct rd_msg *)icmp6_hdr(skb);
e8599ff4 4015
71bcdba0 4016 if (ipv6_addr_is_multicast(&msg->dest)) {
6e157b6a 4017 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
e8599ff4
DM		 4018 return;
4019 }
4020
6e157b6a 4021 on_link = 0;
71bcdba0 4022 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
e8599ff4 4023 on_link = 1;
71bcdba0 4024 } else if (ipv6_addr_type(&msg->target) !=
e8599ff4 4025 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
6e157b6a 4026 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
e8599ff4
DM		 4027 return;
4028 }
4029
4030 in6_dev = __in6_dev_get(skb->dev);
4031 if (!in6_dev)
4032 return;
4033 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
4034 return;
4035
4036 /* RFC2461 8.1:
4037 * The IP source address of the Redirect MUST be the same as the current
4038 * first-hop router for the specified ICMP Destination Address.
4039 */
4040
f997c55c 4041 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
e8599ff4
DM		 4042 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4043 return;
4044 }
6e157b6a
DM		 4045
4046 lladdr = NULL;
e8599ff4
DM		 4047 if (ndopts.nd_opts_tgt_lladdr) {
4048 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4049 skb->dev);
4050 if (!lladdr) {
4051 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4052 return;
4053 }
4054 }
4055
6e157b6a 4056 rt = (struct rt6_info *) dst;
ec13ad1d 4057 if (rt->rt6i_flags & RTF_REJECT) {
6e157b6a 4058 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
e8599ff4 4059 return;
6e157b6a 4060 }
e8599ff4 4061
6e157b6a
DM		 4062 /* Redirect received -> path was valid.
4063 * Look, redirects are sent only in response to data packets,
4064 * so that this nexthop apparently is reachable. --ANK
4065 */
0dec879f 4066 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
a6279458 4067
71bcdba0 4068 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
6e157b6a
DM		 4069 if (!neigh)
4070 return;
a6279458 4071
1da177e4
LT		 4072 /*
4073 * We have finally decided to accept it.
4074 */
4075
f997c55c 4076 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1da177e4
LT		 4077 NEIGH_UPDATE_F_WEAK_OVERRIDE|
4078 NEIGH_UPDATE_F_OVERRIDE|
4079 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
f997c55c
AA		 4080 NEIGH_UPDATE_F_ISROUTER)),
4081 NDISC_REDIRECT, &ndopts);
1da177e4 4082
4d85cd0c 4083 rcu_read_lock();
85bd05de 4084 res.f6i = rcu_dereference(rt->from);
ff24e498 4085 if (!res.f6i)
886b7a50 4086 goto out;
8a14e46f 4087
49d5b8ef
DA		 4088 if (res.f6i->nh) {
4089 struct fib6_nh_match_arg arg = {
4090 .dev = dst->dev,
4091 .gw = &rt->rt6i_gateway,
4092 };
4093
4094 nexthop_for_each_fib6_nh(res.f6i->nh,
4095 fib6_nh_find_match, &arg);
4096
4097 /* fib6_info uses a nexthop that does not have fib6_nh
4098 * using the dst->dev. Should be impossible
4099 */
4100 if (!arg.match)
4101 goto out;
4102 res.nh = arg.match;
4103 } else {
4104 res.nh = res.f6i->fib6_nh;
4105 }
4106
7d21fec9
DA		 4107 res.fib6_flags = res.f6i->fib6_flags;
4108 res.fib6_type = res.f6i->fib6_type;
85bd05de 4109 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
38308473 4110 if (!nrt)
1da177e4
LT		 4111 goto out;
4112
4113 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4114 if (on_link)
4115 nrt->rt6i_flags &= ~RTF_GATEWAY;
4116
4e3fd7a0 4117 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1da177e4 4118
886b7a50 4119 /* rt6_insert_exception() will take care of duplicated exceptions */
5012f0a5 4120 if (rt6_insert_exception(nrt, &res)) {
2b760fcf
WW		 4121 dst_release_immediate(&nrt->dst);
4122 goto out;
4123 }
1da177e4 4124
d8d1f30b
CG		 4125 netevent.old = &rt->dst;
4126 netevent.new = &nrt->dst;
71bcdba0 4127 netevent.daddr = &msg->dest;
60592833 4128 netevent.neigh = neigh;
8d71740c
TT		 4129 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4130
1da177e4 4131out:
886b7a50 4132 rcu_read_unlock();
e8599ff4 4133 neigh_release(neigh);
6e157b6a
DM		 4134}
4135
70ceb4f5 4136#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 4137static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 4138 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 4139 const struct in6_addr *gwaddr,
4140 struct net_device *dev)
70ceb4f5 4141{
830218c1
DA		 4142 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4143 int ifindex = dev->ifindex;
70ceb4f5 4144 struct fib6_node *fn;
8d1c802b 4145 struct fib6_info *rt = NULL;
c71099ac
TG		 4146 struct fib6_table *table;
4147
830218c1 4148 table = fib6_get_table(net, tb_id);
38308473 4149 if (!table)
c71099ac 4150 return NULL;
70ceb4f5 4151
66f5d6ce 4152 rcu_read_lock();
38fbeeee 4153 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
70ceb4f5
YH		 4154 if (!fn)
4155 goto out;
4156
66f5d6ce 4157 for_each_fib6_node_rt_rcu(fn) {
f88d8ea6
DA		 4158 /* these routes do not use nexthops */
4159 if (rt->nh)
4160 continue;
1cf844c7 4161 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
70ceb4f5 4162 continue;
2b2450ca 4163 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
1cf844c7 4164 !rt->fib6_nh->fib_nh_gw_family)
70ceb4f5 4165 continue;
1cf844c7 4166 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
70ceb4f5 4167 continue;
e873e4b9
WW		 4168 if (!fib6_info_hold_safe(rt))
4169 continue;
70ceb4f5
YH		 4170 break;
4171 }
4172out:
66f5d6ce 4173 rcu_read_unlock();
70ceb4f5
YH		 4174 return rt;
4175}
4176
8d1c802b 4177static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 4178 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 4179 const struct in6_addr *gwaddr,
4180 struct net_device *dev,
95c96174 4181 unsigned int pref)
70ceb4f5 4182{
86872cb5 4183 struct fib6_config cfg = {
238fc7ea 4184 .fc_metric = IP6_RT_PRIO_USER,
830218c1 4185 .fc_ifindex = dev->ifindex,
86872cb5
TG		 4186 .fc_dst_len = prefixlen,
4187 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4188 RTF_UP | RTF_PREF(pref),
b91d5329 4189 .fc_protocol = RTPROT_RA,
e8478e80 4190 .fc_type = RTN_UNICAST,
15e47304 4191 .fc_nlinfo.portid = 0,
efa2cea0
DL		 4192 .fc_nlinfo.nlh = NULL,
4193 .fc_nlinfo.nl_net = net,
86872cb5
TG		 4194 };
4195
830218c1 4196 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
4e3fd7a0
AD		 4197 cfg.fc_dst = *prefix;
4198 cfg.fc_gateway = *gwaddr;
70ceb4f5 4199
e317da96
YH		 4200 /* We should treat it as a default route if prefix length is 0. */
4201 if (!prefixlen)
86872cb5 4202 cfg.fc_flags |= RTF_DEFAULT;
70ceb4f5 4203
acb54e3c 4204 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
70ceb4f5 4205
830218c1 4206 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
70ceb4f5
YH		 4207}
4208#endif
4209
8d1c802b 4210struct fib6_info *rt6_get_dflt_router(struct net *net,
afb1d4b5
DA		 4211 const struct in6_addr *addr,
4212 struct net_device *dev)
1ab1457c 4213{
830218c1 4214 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
8d1c802b 4215 struct fib6_info *rt;
c71099ac 4216 struct fib6_table *table;
1da177e4 4217
afb1d4b5 4218 table = fib6_get_table(net, tb_id);
38308473 4219 if (!table)
c71099ac 4220 return NULL;
1da177e4 4221
66f5d6ce
WW		 4222 rcu_read_lock();
4223 for_each_fib6_node_rt_rcu(&table->tb6_root) {
f88d8ea6 4224 struct fib6_nh *nh;
ad1601ae 4225
f88d8ea6
DA		 4226 /* RA routes do not use nexthops */
4227 if (rt->nh)
4228 continue;
4229
4230 nh = rt->fib6_nh;
ad1601ae 4231 if (dev == nh->fib_nh_dev &&
93c2fb25 4232 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
ad1601ae 4233 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
1da177e4
LT		 4234 break;
4235 }
e873e4b9
WW		 4236 if (rt && !fib6_info_hold_safe(rt))
4237 rt = NULL;
66f5d6ce 4238 rcu_read_unlock();
1da177e4
LT		 4239 return rt;
4240}
4241
8d1c802b 4242struct fib6_info *rt6_add_dflt_router(struct net *net,
afb1d4b5 4243 const struct in6_addr *gwaddr,
ebacaaa0
YH		 4244 struct net_device *dev,
4245 unsigned int pref)
1da177e4 4246{
86872cb5 4247 struct fib6_config cfg = {
ca254490 4248 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
238fc7ea 4249 .fc_metric = IP6_RT_PRIO_USER,
86872cb5
TG		 4250 .fc_ifindex = dev->ifindex,
4251 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4252 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
b91d5329 4253 .fc_protocol = RTPROT_RA,
e8478e80 4254 .fc_type = RTN_UNICAST,
15e47304 4255 .fc_nlinfo.portid = 0,
5578689a 4256 .fc_nlinfo.nlh = NULL,
afb1d4b5 4257 .fc_nlinfo.nl_net = net,
86872cb5 4258 };
1da177e4 4259
4e3fd7a0 4260 cfg.fc_gateway = *gwaddr;
1da177e4 4261
acb54e3c 4262 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
830218c1
DA		 4263 struct fib6_table *table;
4264
4265 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4266 if (table)
4267 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4268 }
1da177e4 4269
afb1d4b5 4270 return rt6_get_dflt_router(net, gwaddr, dev);
1da177e4
LT		 4271}
4272
afb1d4b5
DA		 4273static void __rt6_purge_dflt_routers(struct net *net,
4274 struct fib6_table *table)
1da177e4 4275{
8d1c802b 4276 struct fib6_info *rt;
1da177e4
LT		 4277
4278restart:
66f5d6ce
WW		 4279 rcu_read_lock();
4280 for_each_fib6_node_rt_rcu(&table->tb6_root) {
dcd1f572
DA		 4281 struct net_device *dev = fib6_info_nh_dev(rt);
4282 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4283
93c2fb25 4284 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
e873e4b9
WW		 4285 (!idev || idev->cnf.accept_ra != 2) &&
4286 fib6_info_hold_safe(rt)) {
93531c67 4287 rcu_read_unlock();
11dd74b3 4288 ip6_del_rt(net, rt, false);
1da177e4
LT		 4289 goto restart;
4290 }
4291 }
66f5d6ce 4292 rcu_read_unlock();
830218c1
DA		 4293
4294 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4295}
4296
4297void rt6_purge_dflt_routers(struct net *net)
4298{
4299 struct fib6_table *table;
4300 struct hlist_head *head;
4301 unsigned int h;
4302
4303 rcu_read_lock();
4304
4305 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4306 head = &net->ipv6.fib_table_hash[h];
4307 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4308 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
afb1d4b5 4309 __rt6_purge_dflt_routers(net, table);
830218c1
DA		 4310 }
4311 }
4312
4313 rcu_read_unlock();
1da177e4
LT		 4314}
4315
5578689a
DL		 4316static void rtmsg_to_fib6_config(struct net *net,
4317 struct in6_rtmsg *rtmsg,
86872cb5
TG		 4318 struct fib6_config *cfg)
4319{
8823a3ac
MŻ		 4320 *cfg = (struct fib6_config){
4321 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4322 : RT6_TABLE_MAIN,
4323 .fc_ifindex = rtmsg->rtmsg_ifindex,
67f69513 4324 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
8823a3ac
MŻ		 4325 .fc_expires = rtmsg->rtmsg_info,
4326 .fc_dst_len = rtmsg->rtmsg_dst_len,
4327 .fc_src_len = rtmsg->rtmsg_src_len,
4328 .fc_flags = rtmsg->rtmsg_flags,
4329 .fc_type = rtmsg->rtmsg_type,
4330
4331 .fc_nlinfo.nl_net = net,
4332
4333 .fc_dst = rtmsg->rtmsg_dst,
4334 .fc_src = rtmsg->rtmsg_src,
4335 .fc_gateway = rtmsg->rtmsg_gateway,
4336 };
86872cb5
TG		 4337}
4338
7c1552da 4339int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
1da177e4 4340{
86872cb5 4341 struct fib6_config cfg;
1da177e4
LT		 4342 int err;
4343
7c1552da
CH		 4344 if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4345 return -EINVAL;
4346 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4347 return -EPERM;
86872cb5 4348
7c1552da 4349 rtmsg_to_fib6_config(net, rtmsg, &cfg);
1da177e4 4350
7c1552da
CH		 4351 rtnl_lock();
4352 switch (cmd) {
4353 case SIOCADDRT:
4354 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4355 break;
4356 case SIOCDELRT:
4357 err = ip6_route_del(&cfg, NULL);
4358 break;
3ff50b79 4359 }
7c1552da
CH		 4360 rtnl_unlock();
4361 return err;
1da177e4
LT		 4362}
4363
4364/*
4365 * Drop the packet on the floor
4366 */
4367
d5fdd6ba 4368static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1da177e4 4369{
adf30907 4370 struct dst_entry *dst = skb_dst(skb);
1d3fd8a1
SS		 4371 struct net *net = dev_net(dst->dev);
4372 struct inet6_dev *idev;
4373 int type;
4374
4375 if (netif_is_l3_master(skb->dev) &&
4376 dst->dev == net->loopback_dev)
4377 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4378 else
4379 idev = ip6_dst_idev(dst);
4380
612f09e8
YH		 4381 switch (ipstats_mib_noroutes) {
4382 case IPSTATS_MIB_INNOROUTES:
0660e03f 4383 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
45bb0060 4384 if (type == IPV6_ADDR_ANY) {
1d3fd8a1 4385 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
612f09e8
YH		 4386 break;
4387 }
a8eceea8 4388 fallthrough;
612f09e8 4389 case IPSTATS_MIB_OUTNOROUTES:
1d3fd8a1 4390 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
612f09e8
YH		 4391 break;
4392 }
1d3fd8a1
SS		 4393
4394 /* Start over by dropping the dst for l3mdev case */
4395 if (netif_is_l3_master(skb->dev))
4396 skb_dst_drop(skb);
4397
3ffe533c 4398 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1da177e4
LT		 4399 kfree_skb(skb);
4400 return 0;
4401}
4402
9ce8ade0
TG		 4403static int ip6_pkt_discard(struct sk_buff *skb)
4404{
612f09e8 4405 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4406}
4407
ede2059d 4408static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
1da177e4 4409{
adf30907 4410 skb->dev = skb_dst(skb)->dev;
612f09e8 4411 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1da177e4
LT		 4412}
4413
9ce8ade0
TG		 4414static int ip6_pkt_prohibit(struct sk_buff *skb)
4415{
612f09e8 4416 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4417}
4418
ede2059d 4419static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
9ce8ade0 4420{
adf30907 4421 skb->dev = skb_dst(skb)->dev;
612f09e8 4422 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
9ce8ade0
TG		 4423}
4424
1da177e4
LT		 4425/*
4426 * Allocate a dst for local (unicast / anycast) address.
4427 */
4428
360a9887
DA		 4429struct fib6_info *addrconf_f6i_alloc(struct net *net,
4430 struct inet6_dev *idev,
4431 const struct in6_addr *addr,
4432 bool anycast, gfp_t gfp_flags)
1da177e4 4433{
c7a1ce39
DA		 4434 struct fib6_config cfg = {
4435 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4436 .fc_ifindex = idev->dev->ifindex,
d55a2e37 4437 .fc_flags = RTF_UP | RTF_NONEXTHOP,
c7a1ce39
DA		 4438 .fc_dst = *addr,
4439 .fc_dst_len = 128,
4440 .fc_protocol = RTPROT_KERNEL,
4441 .fc_nlinfo.nl_net = net,
4442 .fc_ignore_dev_down = true,
4443 };
d55a2e37 4444 struct fib6_info *f6i;
1da177e4 4445
e8478e80 4446 if (anycast) {
c7a1ce39
DA		 4447 cfg.fc_type = RTN_ANYCAST;
4448 cfg.fc_flags |= RTF_ANYCAST;
e8478e80 4449 } else {
c7a1ce39
DA		 4450 cfg.fc_type = RTN_LOCAL;
4451 cfg.fc_flags |= RTF_LOCAL;
e8478e80 4452 }
1da177e4 4453
d55a2e37 4454 f6i = ip6_route_info_create(&cfg, gfp_flags, NULL);
8652f17c 4455 if (!IS_ERR(f6i))
d55a2e37
MŻ		 4456 f6i->dst_nocount = true;
4457 return f6i;
1da177e4
LT		 4458}
4459
c3968a85
DW		 4460/* remove deleted ip from prefsrc entries */
4461struct arg_dev_net_ip {
4462 struct net_device *dev;
4463 struct net *net;
4464 struct in6_addr *addr;
4465};
4466
8d1c802b 4467static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
c3968a85
DW		 4468{
4469 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4470 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4471 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4472
f88d8ea6
DA		 4473 if (!rt->nh &&
4474 ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
421842ed 4475 rt != net->ipv6.fib6_null_entry &&
93c2fb25 4476 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
60006a48 4477 spin_lock_bh(&rt6_exception_lock);
c3968a85 4478 /* remove prefsrc entry */
93c2fb25 4479 rt->fib6_prefsrc.plen = 0;
60006a48 4480 spin_unlock_bh(&rt6_exception_lock);
c3968a85
DW		 4481 }
4482 return 0;
4483}
4484
4485void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4486{
4487 struct net *net = dev_net(ifp->idev->dev);
4488 struct arg_dev_net_ip adni = {
4489 .dev = ifp->idev->dev,
4490 .net = net,
4491 .addr = &ifp->addr,
4492 };
0c3584d5 4493 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
c3968a85
DW		 4494}
4495
2b2450ca 4496#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
be7a010d
DJ		 4497
4498/* Remove routers and update dst entries when gateway turn into host. */
8d1c802b 4499static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
be7a010d
DJ		 4500{
4501 struct in6_addr *gateway = (struct in6_addr *)arg;
f88d8ea6
DA		 4502 struct fib6_nh *nh;
4503
4504 /* RA routes do not use nexthops */
4505 if (rt->nh)
4506 return 0;
be7a010d 4507
f88d8ea6 4508 nh = rt->fib6_nh;
93c2fb25 4509 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
cc5c073a 4510 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
be7a010d 4511 return -1;
b16cb459
WW		 4512
4513 /* Further clean up cached routes in exception table.
4514 * This is needed because cached route may have a different
4515 * gateway than its 'parent' in the case of an ip redirect.
4516 */
cc5c073a 4517 fib6_nh_exceptions_clean_tohost(nh, gateway);
b16cb459 4518
be7a010d
DJ		 4519 return 0;
4520}
4521
4522void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4523{
4524 fib6_clean_all(net, fib6_clean_tohost, gateway);
4525}
4526
2127d95a
IS		 4527struct arg_netdev_event {
4528 const struct net_device *dev;
4c981e28 4529 union {
ecc5663c 4530 unsigned char nh_flags;
4c981e28
IS		 4531 unsigned long event;
4532 };
2127d95a
IS		 4533};
4534
8d1c802b 4535static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
d7dedee1 4536{
8d1c802b 4537 struct fib6_info *iter;
d7dedee1
IS		 4538 struct fib6_node *fn;
4539
93c2fb25
DA		 4540 fn = rcu_dereference_protected(rt->fib6_node,
4541 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4542 iter = rcu_dereference_protected(fn->leaf,
93c2fb25 4543 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4544 while (iter) {
93c2fb25 4545 if (iter->fib6_metric == rt->fib6_metric &&
33bd5ac5 4546 rt6_qualify_for_ecmp(iter))
d7dedee1 4547 return iter;
8fb11a9a 4548 iter = rcu_dereference_protected(iter->fib6_next,
93c2fb25 4549 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1
IS		 4550 }
4551
4552 return NULL;
4553}
4554
f88d8ea6 4555/* only called for fib entries with builtin fib6_nh */
8d1c802b 4556static bool rt6_is_dead(const struct fib6_info *rt)
d7dedee1 4557{
1cf844c7
DA		 4558 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4559 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4560 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
d7dedee1
IS		 4561 return true;
4562
4563 return false;
4564}
4565
8d1c802b 4566static int rt6_multipath_total_weight(const struct fib6_info *rt)
d7dedee1 4567{
8d1c802b 4568 struct fib6_info *iter;
d7dedee1
IS		 4569 int total = 0;
4570
4571 if (!rt6_is_dead(rt))
1cf844c7 4572 total += rt->fib6_nh->fib_nh_weight;
d7dedee1 4573
93c2fb25 4574 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
d7dedee1 4575 if (!rt6_is_dead(iter))
1cf844c7 4576 total += iter->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4577 }
4578
4579 return total;
4580}
4581
8d1c802b 4582static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
d7dedee1
IS		 4583{
4584 int upper_bound = -1;
4585
4586 if (!rt6_is_dead(rt)) {
1cf844c7 4587 *weight += rt->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4588 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4589 total) - 1;
4590 }
1cf844c7 4591 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
d7dedee1
IS		 4592}
4593
8d1c802b 4594static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
d7dedee1 4595{
8d1c802b 4596 struct fib6_info *iter;
d7dedee1
IS		 4597 int weight = 0;
4598
4599 rt6_upper_bound_set(rt, &weight, total);
4600
93c2fb25 4601 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
d7dedee1
IS		 4602 rt6_upper_bound_set(iter, &weight, total);
4603}
4604
8d1c802b 4605void rt6_multipath_rebalance(struct fib6_info *rt)
d7dedee1 4606{
8d1c802b 4607 struct fib6_info *first;
d7dedee1
IS		 4608 int total;
4609
4610 /* In case the entire multipath route was marked for flushing,
4611 * then there is no need to rebalance upon the removal of every
4612 * sibling route.
4613 */
93c2fb25 4614 if (!rt->fib6_nsiblings || rt->should_flush)
d7dedee1
IS		 4615 return;
4616
4617 /* During lookup routes are evaluated in order, so we need to
4618 * make sure upper bounds are assigned from the first sibling
4619 * onwards.
4620 */
4621 first = rt6_multipath_first_sibling(rt);
4622 if (WARN_ON_ONCE(!first))
4623 return;
4624
4625 total = rt6_multipath_total_weight(first);
4626 rt6_multipath_upper_bound_set(first, total);
4627}
4628
8d1c802b 4629static int fib6_ifup(struct fib6_info *rt, void *p_arg)
2127d95a
IS		 4630{
4631 const struct arg_netdev_event *arg = p_arg;
7aef6859 4632 struct net *net = dev_net(arg->dev);
2127d95a 4633
f88d8ea6 4634 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
1cf844c7
DA		 4635 rt->fib6_nh->fib_nh_dev == arg->dev) {
4636 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
7aef6859 4637 fib6_update_sernum_upto_root(net, rt);
d7dedee1 4638 rt6_multipath_rebalance(rt);
1de178ed 4639 }
2127d95a
IS		 4640
4641 return 0;
4642}
4643
ecc5663c 4644void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
2127d95a
IS		 4645{
4646 struct arg_netdev_event arg = {
4647 .dev = dev,
6802f3ad
IS		 4648 {
4649 .nh_flags = nh_flags,
4650 },
2127d95a
IS		 4651 };
4652
4653 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4654 arg.nh_flags |= RTNH_F_LINKDOWN;
4655
4656 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4657}
4658
f88d8ea6 4659/* only called for fib entries with inline fib6_nh */
8d1c802b 4660static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
1de178ed
IS		 4661 const struct net_device *dev)
4662{
8d1c802b 4663 struct fib6_info *iter;
1de178ed 4664
1cf844c7 4665 if (rt->fib6_nh->fib_nh_dev == dev)
1de178ed 4666 return true;
93c2fb25 4667 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7 4668 if (iter->fib6_nh->fib_nh_dev == dev)
1de178ed
IS		 4669 return true;
4670
4671 return false;
4672}
4673
8d1c802b 4674static void rt6_multipath_flush(struct fib6_info *rt)
1de178ed 4675{
8d1c802b 4676 struct fib6_info *iter;
1de178ed
IS		 4677
4678 rt->should_flush = 1;
93c2fb25 4679 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1de178ed
IS		 4680 iter->should_flush = 1;
4681}
4682
8d1c802b 4683static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
1de178ed
IS		 4684 const struct net_device *down_dev)
4685{
8d1c802b 4686 struct fib6_info *iter;
1de178ed
IS		 4687 unsigned int dead = 0;
4688
1cf844c7
DA		 4689 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4690 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed 4691 dead++;
93c2fb25 4692 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4693 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4694 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed
IS		 4695 dead++;
4696
4697 return dead;
4698}
4699
8d1c802b 4700static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
1de178ed 4701 const struct net_device *dev,
ecc5663c 4702 unsigned char nh_flags)
1de178ed 4703{
8d1c802b 4704 struct fib6_info *iter;
1de178ed 4705
1cf844c7
DA		 4706 if (rt->fib6_nh->fib_nh_dev == dev)
4707 rt->fib6_nh->fib_nh_flags |= nh_flags;
93c2fb25 4708 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4709 if (iter->fib6_nh->fib_nh_dev == dev)
4710 iter->fib6_nh->fib_nh_flags |= nh_flags;
1de178ed
IS		 4711}
4712
a1a22c12 4713/* called with write lock held for table with rt */
8d1c802b 4714static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
1da177e4 4715{
4c981e28
IS		 4716 const struct arg_netdev_event *arg = p_arg;
4717 const struct net_device *dev = arg->dev;
7aef6859 4718 struct net *net = dev_net(dev);
8ed67789 4719
f88d8ea6 4720 if (rt == net->ipv6.fib6_null_entry || rt->nh)
27c6fa73
IS		 4721 return 0;
4722
4723 switch (arg->event) {
4724 case NETDEV_UNREGISTER:
1cf844c7 4725 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
27c6fa73 4726 case NETDEV_DOWN:
1de178ed 4727 if (rt->should_flush)
27c6fa73 4728 return -1;
93c2fb25 4729 if (!rt->fib6_nsiblings)
1cf844c7 4730 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
1de178ed
IS		 4731 if (rt6_multipath_uses_dev(rt, dev)) {
4732 unsigned int count;
4733
4734 count = rt6_multipath_dead_count(rt, dev);
93c2fb25 4735 if (rt->fib6_nsiblings + 1 == count) {
1de178ed
IS		 4736 rt6_multipath_flush(rt);
4737 return -1;
4738 }
4739 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4740 RTNH_F_LINKDOWN);
7aef6859 4741 fib6_update_sernum(net, rt);
d7dedee1 4742 rt6_multipath_rebalance(rt);
1de178ed
IS		 4743 }
4744 return -2;
27c6fa73 4745 case NETDEV_CHANGE:
1cf844c7 4746 if (rt->fib6_nh->fib_nh_dev != dev ||
93c2fb25 4747 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
27c6fa73 4748 break;
1cf844c7 4749 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
d7dedee1 4750 rt6_multipath_rebalance(rt);
27c6fa73 4751 break;
2b241361 4752 }
c159d30c 4753
1da177e4
LT		 4754 return 0;
4755}
4756
27c6fa73 4757void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
1da177e4 4758{
4c981e28 4759 struct arg_netdev_event arg = {
8ed67789 4760 .dev = dev,
6802f3ad
IS		 4761 {
4762 .event = event,
4763 },
8ed67789 4764 };
7c6bb7d2 4765 struct net *net = dev_net(dev);
8ed67789 4766
7c6bb7d2
DA		 4767 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4768 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4769 else
4770 fib6_clean_all(net, fib6_ifdown, &arg);
4c981e28
IS		 4771}
4772
4773void rt6_disable_ip(struct net_device *dev, unsigned long event)
4774{
4775 rt6_sync_down_dev(dev, event);
4776 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4777 neigh_ifdown(&nd_tbl, dev);
1da177e4
LT		 4778}
4779
95c96174 4780struct rt6_mtu_change_arg {
1da177e4 4781 struct net_device *dev;
95c96174 4782 unsigned int mtu;
c0b220cf 4783 struct fib6_info *f6i;
1da177e4
LT		 4784};
4785
cc5c073a 4786static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
c0b220cf
DA		 4787{
4788 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
cc5c073a 4789 struct fib6_info *f6i = arg->f6i;
c0b220cf
DA		 4790
4791 /* For administrative MTU increase, there is no way to discover
4792 * IPv6 PMTU increase, so PMTU increase should be updated here.
4793 * Since RFC 1981 doesn't include administrative MTU increase
4794 * update PMTU increase is a MUST. (i.e. jumbo frame)
4795 */
4796 if (nh->fib_nh_dev == arg->dev) {
4797 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4798 u32 mtu = f6i->fib6_pmtu;
4799
4800 if (mtu >= arg->mtu ||
4801 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4802 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4803
4804 spin_lock_bh(&rt6_exception_lock);
cc5c073a 4805 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
c0b220cf
DA		 4806 spin_unlock_bh(&rt6_exception_lock);
4807 }
4808
4809 return 0;
4810}
4811
4812static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
1da177e4
LT		 4813{
4814 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4815 struct inet6_dev *idev;
4816
4817 /* In IPv6 pmtu discovery is not optional,
4818 so that RTAX_MTU lock cannot disable it.
4819 We still use this lock to block changes
4820 caused by addrconf/ndisc.
4821 */
4822
4823 idev = __in6_dev_get(arg->dev);
38308473 4824 if (!idev)
1da177e4
LT		 4825 return 0;
4826
c0b220cf
DA		 4827 if (fib6_metric_locked(f6i, RTAX_MTU))
4828 return 0;
d4ead6b3 4829
c0b220cf 4830 arg->f6i = f6i;
2d44234b
DA		 4831 if (f6i->nh) {
4832 /* fib6_nh_mtu_change only returns 0, so this is safe */
4833 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4834 arg);
4835 }
4836
1cf844c7 4837 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
1da177e4
LT		 4838}
4839
95c96174 4840void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
1da177e4 4841{
c71099ac
TG		 4842 struct rt6_mtu_change_arg arg = {
4843 .dev = dev,
4844 .mtu = mtu,
4845 };
1da177e4 4846
0c3584d5 4847 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
1da177e4
LT		 4848}
4849
ef7c79ed 4850static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
75425657 4851 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
5176f91e 4852 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
aa8f8778 4853 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
86872cb5 4854 [RTA_OIF] = { .type = NLA_U32 },
ab364a6f 4855 [RTA_IIF] = { .type = NLA_U32 },
86872cb5
TG		 4856 [RTA_PRIORITY] = { .type = NLA_U32 },
4857 [RTA_METRICS] = { .type = NLA_NESTED },
51ebd318 4858 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
c78ba6d6 4859 [RTA_PREF] = { .type = NLA_U8 },
19e42e45
RP		 4860 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4861 [RTA_ENCAP] = { .type = NLA_NESTED },
32bc201e 4862 [RTA_EXPIRES] = { .type = NLA_U32 },
622ec2c9 4863 [RTA_UID] = { .type = NLA_U32 },
3b45a410 4864 [RTA_MARK] = { .type = NLA_U32 },
aa8f8778 4865 [RTA_TABLE] = { .type = NLA_U32 },
eacb9384
RP		 4866 [RTA_IP_PROTO] = { .type = NLA_U8 },
4867 [RTA_SPORT] = { .type = NLA_U16 },
4868 [RTA_DPORT] = { .type = NLA_U16 },
5b98324e 4869 [RTA_NH_ID] = { .type = NLA_U32 },
86872cb5
TG		 4870};
4871
4872static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
333c4301
DA		 4873 struct fib6_config *cfg,
4874 struct netlink_ext_ack *extack)
1da177e4 4875{
86872cb5
TG		 4876 struct rtmsg *rtm;
4877 struct nlattr *tb[RTA_MAX+1];
c78ba6d6 4878 unsigned int pref;
86872cb5 4879 int err;
1da177e4 4880
8cb08174
JB		 4881 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
4882 rtm_ipv6_policy, extack);
86872cb5
TG		 4883 if (err < 0)
4884 goto errout;
1da177e4 4885
86872cb5
TG		 4886 err = -EINVAL;
4887 rtm = nlmsg_data(nlh);
86872cb5 4888
84db8407
MŻ		 4889 *cfg = (struct fib6_config){
4890 .fc_table = rtm->rtm_table,
4891 .fc_dst_len = rtm->rtm_dst_len,
4892 .fc_src_len = rtm->rtm_src_len,
4893 .fc_flags = RTF_UP,
4894 .fc_protocol = rtm->rtm_protocol,
4895 .fc_type = rtm->rtm_type,
4896
4897 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
4898 .fc_nlinfo.nlh = nlh,
4899 .fc_nlinfo.nl_net = sock_net(skb->sk),
4900 };
86872cb5 4901
ef2c7d7b
ND		 4902 if (rtm->rtm_type == RTN_UNREACHABLE ||
4903 rtm->rtm_type == RTN_BLACKHOLE ||
b4949ab2
ND		 4904 rtm->rtm_type == RTN_PROHIBIT ||
4905 rtm->rtm_type == RTN_THROW)
86872cb5
TG		 4906 cfg->fc_flags |= RTF_REJECT;
4907
ab79ad14
MŻ		 4908 if (rtm->rtm_type == RTN_LOCAL)
4909 cfg->fc_flags |= RTF_LOCAL;
4910
1f56a01f
MKL		 4911 if (rtm->rtm_flags & RTM_F_CLONED)
4912 cfg->fc_flags |= RTF_CACHE;
4913
fc1e64e1
DA		 4914 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4915
5b98324e
DA		 4916 if (tb[RTA_NH_ID]) {
4917 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
4918 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
4919 NL_SET_ERR_MSG(extack,
4920 "Nexthop specification and nexthop id are mutually exclusive");
4921 goto errout;
4922 }
4923 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
4924 }
4925
86872cb5 4926 if (tb[RTA_GATEWAY]) {
67b61f6c 4927 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
86872cb5 4928 cfg->fc_flags |= RTF_GATEWAY;
1da177e4 4929 }
e3818541
DA		 4930 if (tb[RTA_VIA]) {
4931 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
4932 goto errout;
4933 }
86872cb5
TG		 4934
4935 if (tb[RTA_DST]) {
4936 int plen = (rtm->rtm_dst_len + 7) >> 3;
4937
4938 if (nla_len(tb[RTA_DST]) < plen)
4939 goto errout;
4940
4941 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1da177e4 4942 }
86872cb5
TG		 4943
4944 if (tb[RTA_SRC]) {
4945 int plen = (rtm->rtm_src_len + 7) >> 3;
4946
4947 if (nla_len(tb[RTA_SRC]) < plen)
4948 goto errout;
4949
4950 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1da177e4 4951 }
86872cb5 4952
c3968a85 4953 if (tb[RTA_PREFSRC])
67b61f6c 4954 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
c3968a85 4955
86872cb5
TG		 4956 if (tb[RTA_OIF])
4957 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4958
4959 if (tb[RTA_PRIORITY])
4960 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4961
4962 if (tb[RTA_METRICS]) {
4963 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4964 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
1da177e4 4965 }
86872cb5
TG		 4966
4967 if (tb[RTA_TABLE])
4968 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4969
51ebd318
ND		 4970 if (tb[RTA_MULTIPATH]) {
4971 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4972 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
9ed59592
DA		 4973
4974 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
c255bd68 4975 cfg->fc_mp_len, extack);
9ed59592
DA		 4976 if (err < 0)
4977 goto errout;
51ebd318
ND		 4978 }
4979
c78ba6d6
LR		 4980 if (tb[RTA_PREF]) {
4981 pref = nla_get_u8(tb[RTA_PREF]);
4982 if (pref != ICMPV6_ROUTER_PREF_LOW &&
4983 pref != ICMPV6_ROUTER_PREF_HIGH)
4984 pref = ICMPV6_ROUTER_PREF_MEDIUM;
4985 cfg->fc_flags |= RTF_PREF(pref);
4986 }
4987
19e42e45
RP		 4988 if (tb[RTA_ENCAP])
4989 cfg->fc_encap = tb[RTA_ENCAP];
4990
9ed59592 4991 if (tb[RTA_ENCAP_TYPE]) {
19e42e45
RP		 4992 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
4993
c255bd68 4994 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
9ed59592
DA		 4995 if (err < 0)
4996 goto errout;
4997 }
4998
32bc201e
XL		 4999 if (tb[RTA_EXPIRES]) {
5000 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5001
5002 if (addrconf_finite_timeout(timeout)) {
5003 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5004 cfg->fc_flags |= RTF_EXPIRES;
5005 }
5006 }
5007
86872cb5
TG		 5008 err = 0;
5009errout:
5010 return err;
1da177e4
LT		 5011}
5012
6b9ea5a6 5013struct rt6_nh {
8d1c802b 5014 struct fib6_info *fib6_info;
6b9ea5a6 5015 struct fib6_config r_cfg;
6b9ea5a6
RP		 5016 struct list_head next;
5017};
5018
d4ead6b3
DA		 5019static int ip6_route_info_append(struct net *net,
5020 struct list_head *rt6_nh_list,
8d1c802b
DA		 5021 struct fib6_info *rt,
5022 struct fib6_config *r_cfg)
6b9ea5a6
RP		 5023{
5024 struct rt6_nh *nh;
6b9ea5a6
RP		 5025 int err = -EEXIST;
5026
5027 list_for_each_entry(nh, rt6_nh_list, next) {
8d1c802b
DA		 5028 /* check if fib6_info already exists */
5029 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
6b9ea5a6
RP		 5030 return err;
5031 }
5032
5033 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5034 if (!nh)
5035 return -ENOMEM;
8d1c802b 5036 nh->fib6_info = rt;
6b9ea5a6
RP		 5037 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5038 list_add_tail(&nh->next, rt6_nh_list);
5039
5040 return 0;
5041}
5042
8d1c802b
DA		 5043static void ip6_route_mpath_notify(struct fib6_info *rt,
5044 struct fib6_info *rt_last,
3b1137fe
DA		 5045 struct nl_info *info,
5046 __u16 nlflags)
5047{
5048 /* if this is an APPEND route, then rt points to the first route
5049 * inserted and rt_last points to last route inserted. Userspace
5050 * wants a consistent dump of the route which starts at the first
5051 * nexthop. Since sibling routes are always added at the end of
5052 * the list, find the first sibling of the last route appended
5053 */
93c2fb25
DA		 5054 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5055 rt = list_first_entry(&rt_last->fib6_siblings,
8d1c802b 5056 struct fib6_info,
93c2fb25 5057 fib6_siblings);
3b1137fe
DA		 5058 }
5059
5060 if (rt)
5061 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5062}
5063
0ee0f47c
IS		 5064static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5065{
5066 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5067 bool should_notify = false;
5068 struct fib6_info *leaf;
5069 struct fib6_node *fn;
5070
5071 rcu_read_lock();
5072 fn = rcu_dereference(rt->fib6_node);
5073 if (!fn)
5074 goto out;
5075
5076 leaf = rcu_dereference(fn->leaf);
5077 if (!leaf)
5078 goto out;
5079
5080 if (rt == leaf ||
5081 (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5082 rt6_qualify_for_ecmp(leaf)))
5083 should_notify = true;
5084out:
5085 rcu_read_unlock();
5086
5087 return should_notify;
5088}
5089
333c4301
DA		 5090static int ip6_route_multipath_add(struct fib6_config *cfg,
5091 struct netlink_ext_ack *extack)
51ebd318 5092{
8d1c802b 5093 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
3b1137fe 5094 struct nl_info *info = &cfg->fc_nlinfo;
51ebd318
ND		 5095 struct fib6_config r_cfg;
5096 struct rtnexthop *rtnh;
8d1c802b 5097 struct fib6_info *rt;
6b9ea5a6
RP		 5098 struct rt6_nh *err_nh;
5099 struct rt6_nh *nh, *nh_safe;
3b1137fe 5100 __u16 nlflags;
51ebd318
ND		 5101 int remaining;
5102 int attrlen;
6b9ea5a6
RP		 5103 int err = 1;
5104 int nhn = 0;
5105 int replace = (cfg->fc_nlinfo.nlh &&
5106 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5107 LIST_HEAD(rt6_nh_list);
51ebd318 5108
3b1137fe
DA		 5109 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5110 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5111 nlflags |= NLM_F_APPEND;
5112
35f1b4e9 5113 remaining = cfg->fc_mp_len;
51ebd318 5114 rtnh = (struct rtnexthop *)cfg->fc_mp;
51ebd318 5115
6b9ea5a6 5116 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
8d1c802b 5117 * fib6_info structs per nexthop
6b9ea5a6 5118 */
51ebd318
ND		 5119 while (rtnh_ok(rtnh, remaining)) {
5120 memcpy(&r_cfg, cfg, sizeof(*cfg));
5121 if (rtnh->rtnh_ifindex)
5122 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5123
5124 attrlen = rtnh_attrlen(rtnh);
5125 if (attrlen > 0) {
5126 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5127
5128 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5129 if (nla) {
67b61f6c 5130 r_cfg.fc_gateway = nla_get_in6_addr(nla);
51ebd318
ND		 5131 r_cfg.fc_flags |= RTF_GATEWAY;
5132 }
19e42e45
RP		 5133 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5134 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5135 if (nla)
5136 r_cfg.fc_encap_type = nla_get_u16(nla);
51ebd318 5137 }
6b9ea5a6 5138
68e2ffde 5139 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
acb54e3c 5140 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
8c5b83f0
RP		 5141 if (IS_ERR(rt)) {
5142 err = PTR_ERR(rt);
5143 rt = NULL;
6b9ea5a6 5144 goto cleanup;
8c5b83f0 5145 }
b5d2d75e
DA		 5146 if (!rt6_qualify_for_ecmp(rt)) {
5147 err = -EINVAL;
5148 NL_SET_ERR_MSG(extack,
5149 "Device only routes can not be added for IPv6 using the multipath API.");
5150 fib6_info_release(rt);
5151 goto cleanup;
5152 }
6b9ea5a6 5153
1cf844c7 5154 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
398958ae 5155
d4ead6b3
DA		 5156 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5157 rt, &r_cfg);
51ebd318 5158 if (err) {
93531c67 5159 fib6_info_release(rt);
6b9ea5a6
RP		 5160 goto cleanup;
5161 }
5162
5163 rtnh = rtnh_next(rtnh, &remaining);
5164 }
5165
9eee3b49
IS		 5166 if (list_empty(&rt6_nh_list)) {
5167 NL_SET_ERR_MSG(extack,
5168 "Invalid nexthop configuration - no valid nexthops");
5169 return -EINVAL;
5170 }
5171
3b1137fe
DA		 5172 /* for add and replace send one notification with all nexthops.
5173 * Skip the notification in fib6_add_rt2node and send one with
5174 * the full route when done
5175 */
5176 info->skip_notify = 1;
5177
ebee3cad
IS		 5178 /* For add and replace, send one notification with all nexthops. For
5179 * append, send one notification with all appended nexthops.
5180 */
5181 info->skip_notify_kernel = 1;
5182
6b9ea5a6
RP		 5183 err_nh = NULL;
5184 list_for_each_entry(nh, &rt6_nh_list, next) {
8d1c802b
DA		 5185 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5186 fib6_info_release(nh->fib6_info);
93531c67 5187
f7225172
DA		 5188 if (!err) {
5189 /* save reference to last route successfully inserted */
5190 rt_last = nh->fib6_info;
5191
5192 /* save reference to first route for notification */
5193 if (!rt_notif)
5194 rt_notif = nh->fib6_info;
5195 }
3b1137fe 5196
8d1c802b
DA		 5197 /* nh->fib6_info is used or freed at this point, reset to NULL*/
5198 nh->fib6_info = NULL;
6b9ea5a6
RP		 5199 if (err) {
5200 if (replace && nhn)
a5a82d84
JK		 5201 NL_SET_ERR_MSG_MOD(extack,
5202 "multipath route replace failed (check consistency of installed routes)");
6b9ea5a6
RP		 5203 err_nh = nh;
5204 goto add_errout;
51ebd318 5205 }
6b9ea5a6 5206
1a72418b 5207 /* Because each route is added like a single route we remove
27596472
MK		 5208 * these flags after the first nexthop: if there is a collision,
5209 * we have already failed to add the first nexthop:
5210 * fib6_add_rt2node() has rejected it; when replacing, old
5211 * nexthops have been replaced by first new, the rest should
5212 * be added to it.
1a72418b 5213 */
27596472
MK		 5214 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5215 NLM_F_REPLACE);
afecdb37 5216 cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
6b9ea5a6
RP		 5217 nhn++;
5218 }
5219
0ee0f47c
IS		 5220 /* An in-kernel notification should only be sent in case the new
5221 * multipath route is added as the first route in the node, or if
5222 * it was appended to it. We pass 'rt_notif' since it is the first
5223 * sibling and might allow us to skip some checks in the replace case.
5224 */
5225 if (ip6_route_mpath_should_notify(rt_notif)) {
5226 enum fib_event_type fib_event;
5227
5228 if (rt_notif->fib6_nsiblings != nhn - 1)
5229 fib_event = FIB_EVENT_ENTRY_APPEND;
5230 else
caafb250 5231 fib_event = FIB_EVENT_ENTRY_REPLACE;
0ee0f47c
IS		 5232
5233 err = call_fib6_multipath_entry_notifiers(info->nl_net,
5234 fib_event, rt_notif,
5235 nhn - 1, extack);
5236 if (err) {
5237 /* Delete all the siblings that were just added */
5238 err_nh = NULL;
5239 goto add_errout;
5240 }
5241 }
ebee3cad 5242
3b1137fe
DA		 5243 /* success ... tell user about new route */
5244 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
6b9ea5a6
RP		 5245 goto cleanup;
5246
5247add_errout:
3b1137fe
DA		 5248 /* send notification for routes that were added so that
5249 * the delete notifications sent by ip6_route_del are
5250 * coherent
5251 */
5252 if (rt_notif)
5253 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5254
6b9ea5a6
RP		 5255 /* Delete routes that were already added */
5256 list_for_each_entry(nh, &rt6_nh_list, next) {
5257 if (err_nh == nh)
5258 break;
333c4301 5259 ip6_route_del(&nh->r_cfg, extack);
6b9ea5a6
RP		 5260 }
5261
5262cleanup:
5263 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
8d1c802b
DA		 5264 if (nh->fib6_info)
5265 fib6_info_release(nh->fib6_info);
6b9ea5a6
RP		 5266 list_del(&nh->next);
5267 kfree(nh);
5268 }
5269
5270 return err;
5271}
5272
333c4301
DA		 5273static int ip6_route_multipath_del(struct fib6_config *cfg,
5274 struct netlink_ext_ack *extack)
6b9ea5a6
RP		 5275{
5276 struct fib6_config r_cfg;
5277 struct rtnexthop *rtnh;
5278 int remaining;
5279 int attrlen;
5280 int err = 1, last_err = 0;
5281
5282 remaining = cfg->fc_mp_len;
5283 rtnh = (struct rtnexthop *)cfg->fc_mp;
5284
5285 /* Parse a Multipath Entry */
5286 while (rtnh_ok(rtnh, remaining)) {
5287 memcpy(&r_cfg, cfg, sizeof(*cfg));
5288 if (rtnh->rtnh_ifindex)
5289 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5290
5291 attrlen = rtnh_attrlen(rtnh);
5292 if (attrlen > 0) {
5293 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5294
5295 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5296 if (nla) {
5297 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
5298 r_cfg.fc_flags |= RTF_GATEWAY;
5299 }
5300 }
333c4301 5301 err = ip6_route_del(&r_cfg, extack);
6b9ea5a6
RP		 5302 if (err)
5303 last_err = err;
5304
51ebd318
ND		 5305 rtnh = rtnh_next(rtnh, &remaining);
5306 }
5307
5308 return last_err;
5309}
5310
c21ef3e3
DA		 5311static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5312 struct netlink_ext_ack *extack)
1da177e4 5313{
86872cb5
TG		 5314 struct fib6_config cfg;
5315 int err;
1da177e4 5316
333c4301 5317 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 5318 if (err < 0)
5319 return err;
5320
5b98324e
DA		 5321 if (cfg.fc_nh_id &&
5322 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5323 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5324 return -EINVAL;
5325 }
5326
51ebd318 5327 if (cfg.fc_mp)
333c4301 5328 return ip6_route_multipath_del(&cfg, extack);
0ae81335
DA		 5329 else {
5330 cfg.fc_delete_all_nh = 1;
333c4301 5331 return ip6_route_del(&cfg, extack);
0ae81335 5332 }
1da177e4
LT		 5333}
5334
c21ef3e3
DA		 5335static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5336 struct netlink_ext_ack *extack)
1da177e4 5337{
86872cb5
TG		 5338 struct fib6_config cfg;
5339 int err;
1da177e4 5340
333c4301 5341 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 5342 if (err < 0)
5343 return err;
5344
67f69513
DA		 5345 if (cfg.fc_metric == 0)
5346 cfg.fc_metric = IP6_RT_PRIO_USER;
5347
51ebd318 5348 if (cfg.fc_mp)
333c4301 5349 return ip6_route_multipath_add(&cfg, extack);
51ebd318 5350 else
acb54e3c 5351 return ip6_route_add(&cfg, GFP_KERNEL, extack);
1da177e4
LT		 5352}
5353
a1b7a1f0
DA		 5354/* add the overhead of this fib6_nh to nexthop_len */
5355static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
339bf98f 5356{
a1b7a1f0 5357 int *nexthop_len = arg;
beb1afac 5358
a1b7a1f0
DA		 5359 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5360 + NLA_ALIGN(sizeof(struct rtnexthop))
5361 + nla_total_size(16); /* RTA_GATEWAY */
5362
5363 if (nh->fib_nh_lws) {
5364 /* RTA_ENCAP_TYPE */
5365 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5366 /* RTA_ENCAP */
5367 *nexthop_len += nla_total_size(2);
5368 }
f88d8ea6 5369
a1b7a1f0
DA		 5370 return 0;
5371}
beb1afac 5372
a1b7a1f0
DA		 5373static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5374{
5375 int nexthop_len;
5376
5377 if (f6i->nh) {
5378 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5379 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5380 &nexthop_len);
5381 } else {
5382 struct fib6_nh *nh = f6i->fib6_nh;
5383
5384 nexthop_len = 0;
5385 if (f6i->fib6_nsiblings) {
5386 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
5387 + NLA_ALIGN(sizeof(struct rtnexthop))
5388 + nla_total_size(16) /* RTA_GATEWAY */
5389 + lwtunnel_get_encap_size(nh->fib_nh_lws);
5390
5391 nexthop_len *= f6i->fib6_nsiblings;
5392 }
5393 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
beb1afac
DA		 5394 }
5395
339bf98f
TG		 5396 return NLMSG_ALIGN(sizeof(struct rtmsg))
5397 + nla_total_size(16) /* RTA_SRC */
5398 + nla_total_size(16) /* RTA_DST */
5399 + nla_total_size(16) /* RTA_GATEWAY */
5400 + nla_total_size(16) /* RTA_PREFSRC */
5401 + nla_total_size(4) /* RTA_TABLE */
5402 + nla_total_size(4) /* RTA_IIF */
5403 + nla_total_size(4) /* RTA_OIF */
5404 + nla_total_size(4) /* RTA_PRIORITY */
6a2b9ce0 5405 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
ea697639 5406 + nla_total_size(sizeof(struct rta_cacheinfo))
c78ba6d6 5407 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
19e42e45 5408 + nla_total_size(1) /* RTA_PREF */
beb1afac
DA		 5409 + nexthop_len;
5410}
5411
f88d8ea6
DA		 5412static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5413 unsigned char *flags)
5414{
5415 if (nexthop_is_multipath(nh)) {
5416 struct nlattr *mp;
5417
4255ff05 5418 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
f88d8ea6
DA		 5419 if (!mp)
5420 goto nla_put_failure;
5421
7bdf4de1 5422 if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
f88d8ea6
DA		 5423 goto nla_put_failure;
5424
5425 nla_nest_end(skb, mp);
5426 } else {
5427 struct fib6_nh *fib6_nh;
5428
5429 fib6_nh = nexthop_fib6_nh(nh);
7bdf4de1 5430 if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
f88d8ea6
DA		 5431 flags, false) < 0)
5432 goto nla_put_failure;
5433 }
5434
5435 return 0;
5436
5437nla_put_failure:
5438 return -EMSGSIZE;
5439}
5440
d4ead6b3 5441static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 5442 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 5443 struct in6_addr *dest, struct in6_addr *src,
15e47304 5444 int iif, int type, u32 portid, u32 seq,
f8cfe2ce 5445 unsigned int flags)
1da177e4 5446{
22d0bd82
XL		 5447 struct rt6_info *rt6 = (struct rt6_info *)dst;
5448 struct rt6key *rt6_dst, *rt6_src;
5449 u32 *pmetrics, table, rt6_flags;
f88d8ea6 5450 unsigned char nh_flags = 0;
2d7202bf 5451 struct nlmsghdr *nlh;
22d0bd82 5452 struct rtmsg *rtm;
d4ead6b3 5453 long expires = 0;
1da177e4 5454
15e47304 5455 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
38308473 5456 if (!nlh)
26932566 5457 return -EMSGSIZE;
2d7202bf 5458
22d0bd82
XL		 5459 if (rt6) {
5460 rt6_dst = &rt6->rt6i_dst;
5461 rt6_src = &rt6->rt6i_src;
5462 rt6_flags = rt6->rt6i_flags;
5463 } else {
5464 rt6_dst = &rt->fib6_dst;
5465 rt6_src = &rt->fib6_src;
5466 rt6_flags = rt->fib6_flags;
5467 }
5468
2d7202bf 5469 rtm = nlmsg_data(nlh);
1da177e4 5470 rtm->rtm_family = AF_INET6;
22d0bd82
XL		 5471 rtm->rtm_dst_len = rt6_dst->plen;
5472 rtm->rtm_src_len = rt6_src->plen;
1da177e4 5473 rtm->rtm_tos = 0;
93c2fb25
DA		 5474 if (rt->fib6_table)
5475 table = rt->fib6_table->tb6_id;
c71099ac 5476 else
9e762a4a 5477 table = RT6_TABLE_UNSPEC;
97f0082a 5478 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
c78679e8
DM		 5479 if (nla_put_u32(skb, RTA_TABLE, table))
5480 goto nla_put_failure;
e8478e80
DA		 5481
5482 rtm->rtm_type = rt->fib6_type;
1da177e4
LT		 5483 rtm->rtm_flags = 0;
5484 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
93c2fb25 5485 rtm->rtm_protocol = rt->fib6_protocol;
1da177e4 5486
22d0bd82 5487 if (rt6_flags & RTF_CACHE)
1da177e4
LT		 5488 rtm->rtm_flags |= RTM_F_CLONED;
5489
d4ead6b3
DA		 5490 if (dest) {
5491 if (nla_put_in6_addr(skb, RTA_DST, dest))
c78679e8 5492 goto nla_put_failure;
1ab1457c 5493 rtm->rtm_dst_len = 128;
1da177e4 5494 } else if (rtm->rtm_dst_len)
22d0bd82 5495 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
c78679e8 5496 goto nla_put_failure;
1da177e4
LT		 5497#ifdef CONFIG_IPV6_SUBTREES
5498 if (src) {
930345ea 5499 if (nla_put_in6_addr(skb, RTA_SRC, src))
c78679e8 5500 goto nla_put_failure;
1ab1457c 5501 rtm->rtm_src_len = 128;
c78679e8 5502 } else if (rtm->rtm_src_len &&
22d0bd82 5503 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
c78679e8 5504 goto nla_put_failure;
1da177e4 5505#endif
7bc570c8
YH		 5506 if (iif) {
5507#ifdef CONFIG_IPV6_MROUTE
22d0bd82 5508 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
fd61c6ba
DA		 5509 int err = ip6mr_get_route(net, skb, rtm, portid);
5510
5511 if (err == 0)
5512 return 0;
5513 if (err < 0)
5514 goto nla_put_failure;
7bc570c8
YH		 5515 } else
5516#endif
c78679e8
DM		 5517 if (nla_put_u32(skb, RTA_IIF, iif))
5518 goto nla_put_failure;
d4ead6b3 5519 } else if (dest) {
1da177e4 5520 struct in6_addr saddr_buf;
d4ead6b3 5521 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
930345ea 5522 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 5523 goto nla_put_failure;
1da177e4 5524 }
2d7202bf 5525
93c2fb25 5526 if (rt->fib6_prefsrc.plen) {
c3968a85 5527 struct in6_addr saddr_buf;
93c2fb25 5528 saddr_buf = rt->fib6_prefsrc.addr;
930345ea 5529 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 5530 goto nla_put_failure;
c3968a85
DW		 5531 }
5532
d4ead6b3
DA		 5533 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5534 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
2d7202bf
TG		 5535 goto nla_put_failure;
5536
93c2fb25 5537 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
c78679e8 5538 goto nla_put_failure;
8253947e 5539
beb1afac
DA		 5540 /* For multipath routes, walk the siblings list and add
5541 * each as a nexthop within RTA_MULTIPATH.
5542 */
22d0bd82
XL		 5543 if (rt6) {
5544 if (rt6_flags & RTF_GATEWAY &&
5545 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5546 goto nla_put_failure;
5547
5548 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5549 goto nla_put_failure;
5550 } else if (rt->fib6_nsiblings) {
8d1c802b 5551 struct fib6_info *sibling, *next_sibling;
beb1afac
DA		 5552 struct nlattr *mp;
5553
ae0be8de 5554 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
beb1afac
DA		 5555 if (!mp)
5556 goto nla_put_failure;
5557
1cf844c7 5558 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
7bdf4de1 5559 rt->fib6_nh->fib_nh_weight, AF_INET6) < 0)
beb1afac
DA		 5560 goto nla_put_failure;
5561
5562 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25 5563 &rt->fib6_siblings, fib6_siblings) {
1cf844c7 5564 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
7bdf4de1
DS		 5565 sibling->fib6_nh->fib_nh_weight,
5566 AF_INET6) < 0)
beb1afac
DA		 5567 goto nla_put_failure;
5568 }
5569
5570 nla_nest_end(skb, mp);
f88d8ea6
DA		 5571 } else if (rt->nh) {
5572 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5573 goto nla_put_failure;
ecc5663c 5574
f88d8ea6
DA		 5575 if (nexthop_is_blackhole(rt->nh))
5576 rtm->rtm_type = RTN_BLACKHOLE;
5577
4f80116d
RP		 5578 if (net->ipv4.sysctl_nexthop_compat_mode &&
5579 rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
f88d8ea6
DA		 5580 goto nla_put_failure;
5581
5582 rtm->rtm_flags |= nh_flags;
5583 } else {
7bdf4de1 5584 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
ecc5663c 5585 &nh_flags, false) < 0)
beb1afac 5586 goto nla_put_failure;
ecc5663c
DA		 5587
5588 rtm->rtm_flags |= nh_flags;
beb1afac
DA		 5589 }
5590
22d0bd82 5591 if (rt6_flags & RTF_EXPIRES) {
14895687
DA		 5592 expires = dst ? dst->expires : rt->expires;
5593 expires -= jiffies;
5594 }
69cdf8f9 5595
bb3c4ab9
IS		 5596 if (!dst) {
5597 if (rt->offload)
5598 rtm->rtm_flags |= RTM_F_OFFLOAD;
5599 if (rt->trap)
5600 rtm->rtm_flags |= RTM_F_TRAP;
5601 }
5602
d4ead6b3 5603 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
e3703b3d 5604 goto nla_put_failure;
2d7202bf 5605
22d0bd82 5606 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
c78ba6d6
LR		 5607 goto nla_put_failure;
5608
19e42e45 5609
053c095a
JB		 5610 nlmsg_end(skb, nlh);
5611 return 0;
2d7202bf
TG		 5612
5613nla_put_failure:
26932566
PM		 5614 nlmsg_cancel(skb, nlh);
5615 return -EMSGSIZE;
1da177e4
LT		 5616}
5617
2c170e07
DA		 5618static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5619{
5620 const struct net_device *dev = arg;
5621
5622 if (nh->fib_nh_dev == dev)
5623 return 1;
5624
5625 return 0;
5626}
5627
13e38901
DA		 5628static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5629 const struct net_device *dev)
5630{
2c170e07
DA		 5631 if (f6i->nh) {
5632 struct net_device *_dev = (struct net_device *)dev;
5633
5634 return !!nexthop_for_each_fib6_nh(f6i->nh,
5635 fib6_info_nh_uses_dev,
5636 _dev);
5637 }
5638
1cf844c7 5639 if (f6i->fib6_nh->fib_nh_dev == dev)
13e38901
DA		 5640 return true;
5641
5642 if (f6i->fib6_nsiblings) {
5643 struct fib6_info *sibling, *next_sibling;
5644
5645 list_for_each_entry_safe(sibling, next_sibling,
5646 &f6i->fib6_siblings, fib6_siblings) {
1cf844c7 5647 if (sibling->fib6_nh->fib_nh_dev == dev)
13e38901
DA		 5648 return true;
5649 }
5650 }
5651
5652 return false;
5653}
5654
1e47b483
SB		 5655struct fib6_nh_exception_dump_walker {
5656 struct rt6_rtnl_dump_arg *dump;
5657 struct fib6_info *rt;
5658 unsigned int flags;
5659 unsigned int skip;
5660 unsigned int count;
5661};
5662
5663static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5664{
5665 struct fib6_nh_exception_dump_walker *w = arg;
5666 struct rt6_rtnl_dump_arg *dump = w->dump;
5667 struct rt6_exception_bucket *bucket;
5668 struct rt6_exception *rt6_ex;
5669 int i, err;
5670
5671 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5672 if (!bucket)
5673 return 0;
5674
5675 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5676 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5677 if (w->skip) {
5678 w->skip--;
5679 continue;
5680 }
5681
5682 /* Expiration of entries doesn't bump sernum, insertion
5683 * does. Removal is triggered by insertion, so we can
5684 * rely on the fact that if entries change between two
5685 * partial dumps, this node is scanned again completely,
5686 * see rt6_insert_exception() and fib6_dump_table().
5687 *
5688 * Count expired entries we go through as handled
5689 * entries that we'll skip next time, in case of partial
5690 * node dump. Otherwise, if entries expire meanwhile,
5691 * we'll skip the wrong amount.
5692 */
5693 if (rt6_check_expired(rt6_ex->rt6i)) {
5694 w->count++;
5695 continue;
5696 }
5697
5698 err = rt6_fill_node(dump->net, dump->skb, w->rt,
5699 &rt6_ex->rt6i->dst, NULL, NULL, 0,
5700 RTM_NEWROUTE,
5701 NETLINK_CB(dump->cb->skb).portid,
5702 dump->cb->nlh->nlmsg_seq, w->flags);
5703 if (err)
5704 return err;
5705
5706 w->count++;
5707 }
5708 bucket++;
5709 }
5710
5711 return 0;
5712}
5713
bf9a8a06 5714/* Return -1 if done with node, number of handled routes on partial dump */
1e47b483 5715int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
1da177e4
LT		 5716{
5717 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
13e38901
DA		 5718 struct fib_dump_filter *filter = &arg->filter;
5719 unsigned int flags = NLM_F_MULTI;
1f17e2f2 5720 struct net *net = arg->net;
1e47b483 5721 int count = 0;
1f17e2f2 5722
421842ed 5723 if (rt == net->ipv6.fib6_null_entry)
bf9a8a06 5724 return -1;
1da177e4 5725
13e38901
DA		 5726 if ((filter->flags & RTM_F_PREFIX) &&
5727 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5728 /* success since this is not a prefix route */
bf9a8a06 5729 return -1;
13e38901 5730 }
1e47b483
SB		 5731 if (filter->filter_set &&
5732 ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5733 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5734 (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5735 return -1;
5736 }
5737
5738 if (filter->filter_set ||
5739 !filter->dump_routes || !filter->dump_exceptions) {
13e38901 5740 flags |= NLM_F_DUMP_FILTERED;
f8cfe2ce 5741 }
1da177e4 5742
1e47b483
SB		 5743 if (filter->dump_routes) {
5744 if (skip) {
5745 skip--;
5746 } else {
5747 if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5748 0, RTM_NEWROUTE,
5749 NETLINK_CB(arg->cb->skb).portid,
5750 arg->cb->nlh->nlmsg_seq, flags)) {
5751 return 0;
5752 }
5753 count++;
5754 }
5755 }
5756
5757 if (filter->dump_exceptions) {
5758 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5759 .rt = rt,
5760 .flags = flags,
5761 .skip = skip,
5762 .count = 0 };
5763 int err;
5764
3b525691 5765 rcu_read_lock();
1e47b483
SB		 5766 if (rt->nh) {
5767 err = nexthop_for_each_fib6_nh(rt->nh,
5768 rt6_nh_dump_exceptions,
5769 &w);
5770 } else {
5771 err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5772 }
3b525691 5773 rcu_read_unlock();
1e47b483
SB		 5774
5775 if (err)
5776 return count += w.count;
5777 }
bf9a8a06
SB		 5778
5779 return -1;
1da177e4
LT		 5780}
5781
0eff0a27
JK		 5782static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5783 const struct nlmsghdr *nlh,
5784 struct nlattr **tb,
5785 struct netlink_ext_ack *extack)
5786{
5787 struct rtmsg *rtm;
5788 int i, err;
5789
5790 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5791 NL_SET_ERR_MSG_MOD(extack,
5792 "Invalid header for get route request");
5793 return -EINVAL;
5794 }
5795
5796 if (!netlink_strict_get_check(skb))
8cb08174
JB		 5797 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5798 rtm_ipv6_policy, extack);
0eff0a27
JK		 5799
5800 rtm = nlmsg_data(nlh);
5801 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5802 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5803 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5804 rtm->rtm_type) {
5805 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5806 return -EINVAL;
5807 }
5808 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5809 NL_SET_ERR_MSG_MOD(extack,
5810 "Invalid flags for get route request");
5811 return -EINVAL;
5812 }
5813
8cb08174
JB		 5814 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5815 rtm_ipv6_policy, extack);
0eff0a27
JK		 5816 if (err)
5817 return err;
5818
5819 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5820 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5821 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5822 return -EINVAL;
5823 }
5824
5825 for (i = 0; i <= RTA_MAX; i++) {
5826 if (!tb[i])
5827 continue;
5828
5829 switch (i) {
5830 case RTA_SRC:
5831 case RTA_DST:
5832 case RTA_IIF:
5833 case RTA_OIF:
5834 case RTA_MARK:
5835 case RTA_UID:
5836 case RTA_SPORT:
5837 case RTA_DPORT:
5838 case RTA_IP_PROTO:
5839 break;
5840 default:
5841 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
5842 return -EINVAL;
5843 }
5844 }
5845
5846 return 0;
5847}
5848
c21ef3e3
DA		 5849static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5850 struct netlink_ext_ack *extack)
1da177e4 5851{
3b1e0a65 5852 struct net *net = sock_net(in_skb->sk);
ab364a6f 5853 struct nlattr *tb[RTA_MAX+1];
18c3a61c 5854 int err, iif = 0, oif = 0;
a68886a6 5855 struct fib6_info *from;
18c3a61c 5856 struct dst_entry *dst;
ab364a6f 5857 struct rt6_info *rt;
1da177e4 5858 struct sk_buff *skb;
ab364a6f 5859 struct rtmsg *rtm;
744486d4 5860 struct flowi6 fl6 = {};
18c3a61c 5861 bool fibmatch;
1da177e4 5862
0eff0a27 5863 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
ab364a6f
TG		 5864 if (err < 0)
5865 goto errout;
1da177e4 5866
ab364a6f 5867 err = -EINVAL;
38b7097b
HFS		 5868 rtm = nlmsg_data(nlh);
5869 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
18c3a61c 5870 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
1da177e4 5871
ab364a6f
TG		 5872 if (tb[RTA_SRC]) {
5873 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
5874 goto errout;
5875
4e3fd7a0 5876 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
ab364a6f
TG		 5877 }
5878
5879 if (tb[RTA_DST]) {
5880 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
5881 goto errout;
5882
4e3fd7a0 5883 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
ab364a6f
TG		 5884 }
5885
5886 if (tb[RTA_IIF])
5887 iif = nla_get_u32(tb[RTA_IIF]);
5888
5889 if (tb[RTA_OIF])
72331bc0 5890 oif = nla_get_u32(tb[RTA_OIF]);
1da177e4 5891
2e47b291
LC		 5892 if (tb[RTA_MARK])
5893 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
5894
622ec2c9
LC		 5895 if (tb[RTA_UID])
5896 fl6.flowi6_uid = make_kuid(current_user_ns(),
5897 nla_get_u32(tb[RTA_UID]));
5898 else
5899 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
5900
eacb9384
RP		 5901 if (tb[RTA_SPORT])
5902 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
5903
5904 if (tb[RTA_DPORT])
5905 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
5906
5907 if (tb[RTA_IP_PROTO]) {
5908 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
5e1a99ea
HL		 5909 &fl6.flowi6_proto, AF_INET6,
5910 extack);
eacb9384
RP		 5911 if (err)
5912 goto errout;
5913 }
5914
1da177e4
LT		 5915 if (iif) {
5916 struct net_device *dev;
72331bc0
SL		 5917 int flags = 0;
5918
121622db
FW		 5919 rcu_read_lock();
5920
5921 dev = dev_get_by_index_rcu(net, iif);
1da177e4 5922 if (!dev) {
121622db 5923 rcu_read_unlock();
1da177e4 5924 err = -ENODEV;
ab364a6f 5925 goto errout;
1da177e4 5926 }
72331bc0
SL		 5927
5928 fl6.flowi6_iif = iif;
5929
5930 if (!ipv6_addr_any(&fl6.saddr))
5931 flags |= RT6_LOOKUP_F_HAS_SADDR;
5932
b75cc8f9 5933 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
121622db
FW		 5934
5935 rcu_read_unlock();
72331bc0
SL		 5936 } else {
5937 fl6.flowi6_oif = oif;
5938
58acfd71 5939 dst = ip6_route_output(net, NULL, &fl6);
18c3a61c
RP		 5940 }
5941
18c3a61c
RP		 5942
5943 rt = container_of(dst, struct rt6_info, dst);
5944 if (rt->dst.error) {
5945 err = rt->dst.error;
5946 ip6_rt_put(rt);
5947 goto errout;
1da177e4
LT		 5948 }
5949
9d6acb3b
WC		 5950 if (rt == net->ipv6.ip6_null_entry) {
5951 err = rt->dst.error;
5952 ip6_rt_put(rt);
5953 goto errout;
5954 }
5955
ab364a6f 5956 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
38308473 5957 if (!skb) {
94e187c0 5958 ip6_rt_put(rt);
ab364a6f
TG		 5959 err = -ENOBUFS;
5960 goto errout;
5961 }
1da177e4 5962
d8d1f30b 5963 skb_dst_set(skb, &rt->dst);
a68886a6
DA		 5964
5965 rcu_read_lock();
5966 from = rcu_dereference(rt->from);
886b7a50
MKL		 5967 if (from) {
5968 if (fibmatch)
5969 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
5970 iif, RTM_NEWROUTE,
5971 NETLINK_CB(in_skb).portid,
5972 nlh->nlmsg_seq, 0);
5973 else
5974 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
5975 &fl6.saddr, iif, RTM_NEWROUTE,
5976 NETLINK_CB(in_skb).portid,
5977 nlh->nlmsg_seq, 0);
5978 } else {
5979 err = -ENETUNREACH;
5980 }
a68886a6
DA		 5981 rcu_read_unlock();
5982
1da177e4 5983 if (err < 0) {
ab364a6f
TG		 5984 kfree_skb(skb);
5985 goto errout;
1da177e4
LT		 5986 }
5987
15e47304 5988 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
ab364a6f 5989errout:
1da177e4 5990 return err;
1da177e4
LT		 5991}
5992
8d1c802b 5993void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
37a1d361 5994 unsigned int nlm_flags)
1da177e4
LT		 5995{
5996 struct sk_buff *skb;
5578689a 5997 struct net *net = info->nl_net;
528c4ceb
DL		 5998 u32 seq;
5999 int err;
6000
6001 err = -ENOBUFS;
38308473 6002 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
86872cb5 6003
19e42e45 6004 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
38308473 6005 if (!skb)
21713ebc
TG		 6006 goto errout;
6007
d4ead6b3
DA		 6008 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6009 event, info->portid, seq, nlm_flags);
26932566
PM		 6010 if (err < 0) {
6011 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6012 WARN_ON(err == -EMSGSIZE);
6013 kfree_skb(skb);
6014 goto errout;
6015 }
15e47304 6016 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
1ce85fe4
PNA		 6017 info->nlh, gfp_any());
6018 return;
21713ebc
TG		 6019errout:
6020 if (err < 0)
5578689a 6021 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
1da177e4
LT		 6022}
6023
19a3b7ee
DA		 6024void fib6_rt_update(struct net *net, struct fib6_info *rt,
6025 struct nl_info *info)
6026{
6027 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6028 struct sk_buff *skb;
6029 int err = -ENOBUFS;
6030
6031 /* call_fib6_entry_notifiers will be removed when in-kernel notifier
6032 * is implemented and supported for nexthop objects
6033 */
6034 call_fib6_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, rt, NULL);
6035
6036 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6037 if (!skb)
6038 goto errout;
6039
6040 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6041 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6042 if (err < 0) {
6043 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6044 WARN_ON(err == -EMSGSIZE);
6045 kfree_skb(skb);
6046 goto errout;
6047 }
6048 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6049 info->nlh, gfp_any());
6050 return;
6051errout:
6052 if (err < 0)
6053 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6054}
6055
8ed67789 6056static int ip6_route_dev_notify(struct notifier_block *this,
351638e7 6057 unsigned long event, void *ptr)
8ed67789 6058{
351638e7 6059 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
c346dca1 6060 struct net *net = dev_net(dev);
8ed67789 6061
242d3a49
WC		 6062 if (!(dev->flags & IFF_LOOPBACK))
6063 return NOTIFY_OK;
6064
6065 if (event == NETDEV_REGISTER) {
1cf844c7 6066 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
d8d1f30b 6067 net->ipv6.ip6_null_entry->dst.dev = dev;
8ed67789
DL		 6068 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6069#ifdef CONFIG_IPV6_MULTIPLE_TABLES
d8d1f30b 6070 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
8ed67789 6071 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
d8d1f30b 6072 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
8ed67789 6073 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
242d3a49 6074#endif
76da0704
WC		 6075 } else if (event == NETDEV_UNREGISTER &&
6076 dev->reg_state != NETREG_UNREGISTERED) {
6077 /* NETDEV_UNREGISTER could be fired for multiple times by
6078 * netdev_wait_allrefs(). Make sure we only call this once.
6079 */
12d94a80 6080 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
242d3a49 6081#ifdef CONFIG_IPV6_MULTIPLE_TABLES
12d94a80
ED		 6082 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6083 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
8ed67789
DL		 6084#endif
6085 }
6086
6087 return NOTIFY_OK;
6088}
6089
1da177e4
LT		 6090/*
6091 * /proc
6092 */
6093
6094#ifdef CONFIG_PROC_FS
1da177e4
LT		 6095static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6096{
69ddb805 6097 struct net *net = (struct net *)seq->private;
1da177e4 6098 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
69ddb805
DL		 6099 net->ipv6.rt6_stats->fib_nodes,
6100 net->ipv6.rt6_stats->fib_route_nodes,
81eb8447 6101 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
69ddb805
DL		 6102 net->ipv6.rt6_stats->fib_rt_entries,
6103 net->ipv6.rt6_stats->fib_rt_cache,
fc66f95c 6104 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
69ddb805 6105 net->ipv6.rt6_stats->fib_discarded_routes);
1da177e4
LT		 6106
6107 return 0;
6108}
1da177e4
LT		 6109#endif /* CONFIG_PROC_FS */
6110
6111#ifdef CONFIG_SYSCTL
6112
32927393
CH		 6113static int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
6114 void *buffer, size_t *lenp, loff_t *ppos)
1da177e4 6115{
c486da34
LAG		 6116 struct net *net;
6117 int delay;
f0fb9b28 6118 int ret;
c486da34 6119 if (!write)
1da177e4 6120 return -EINVAL;
c486da34
LAG		 6121
6122 net = (struct net *)ctl->extra1;
6123 delay = net->ipv6.sysctl.flush_delay;
f0fb9b28
AP		 6124 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6125 if (ret)
6126 return ret;
6127
2ac3ac8f 6128 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
c486da34 6129 return 0;
1da177e4
LT		 6130}
6131
ed792e28 6132static struct ctl_table ipv6_route_table_template[] = {
1ab1457c 6133 {
1da177e4 6134 .procname = "flush",
4990509f 6135 .data = &init_net.ipv6.sysctl.flush_delay,
1da177e4 6136 .maxlen = sizeof(int),
89c8b3a1 6137 .mode = 0200,
6d9f239a 6138 .proc_handler = ipv6_sysctl_rtcache_flush
1da177e4
LT		 6139 },
6140 {
1da177e4 6141 .procname = "gc_thresh",
9a7ec3a9 6142 .data = &ip6_dst_ops_template.gc_thresh,
1da177e4
LT		 6143 .maxlen = sizeof(int),
6144 .mode = 0644,
6d9f239a 6145 .proc_handler = proc_dointvec,
1da177e4
LT		 6146 },
6147 {
1da177e4 6148 .procname = "max_size",
4990509f 6149 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
1da177e4
LT		 6150 .maxlen = sizeof(int),
6151 .mode = 0644,
6d9f239a 6152 .proc_handler = proc_dointvec,
1da177e4
LT		 6153 },
6154 {
1da177e4 6155 .procname = "gc_min_interval",
4990509f 6156 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 6157 .maxlen = sizeof(int),
6158 .mode = 0644,
6d9f239a 6159 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6160 },
6161 {
1da177e4 6162 .procname = "gc_timeout",
4990509f 6163 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
1da177e4
LT		 6164 .maxlen = sizeof(int),
6165 .mode = 0644,
6d9f239a 6166 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6167 },
6168 {
1da177e4 6169 .procname = "gc_interval",
4990509f 6170 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
1da177e4
LT		 6171 .maxlen = sizeof(int),
6172 .mode = 0644,
6d9f239a 6173 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6174 },
6175 {
1da177e4 6176 .procname = "gc_elasticity",
4990509f 6177 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
1da177e4
LT		 6178 .maxlen = sizeof(int),
6179 .mode = 0644,
f3d3f616 6180 .proc_handler = proc_dointvec,
1da177e4
LT		 6181 },
6182 {
1da177e4 6183 .procname = "mtu_expires",
4990509f 6184 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
1da177e4
LT		 6185 .maxlen = sizeof(int),
6186 .mode = 0644,
6d9f239a 6187 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6188 },
6189 {
1da177e4 6190 .procname = "min_adv_mss",
4990509f 6191 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
1da177e4
LT		 6192 .maxlen = sizeof(int),
6193 .mode = 0644,
f3d3f616 6194 .proc_handler = proc_dointvec,
1da177e4
LT		 6195 },
6196 {
1da177e4 6197 .procname = "gc_min_interval_ms",
4990509f 6198 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 6199 .maxlen = sizeof(int),
6200 .mode = 0644,
6d9f239a 6201 .proc_handler = proc_dointvec_ms_jiffies,
1da177e4 6202 },
7c6bb7d2
DA		 6203 {
6204 .procname = "skip_notify_on_dev_down",
6205 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6206 .maxlen = sizeof(int),
6207 .mode = 0644,
b8e8a863 6208 .proc_handler = proc_dointvec_minmax,
eec4844f
MC		 6209 .extra1 = SYSCTL_ZERO,
6210 .extra2 = SYSCTL_ONE,
7c6bb7d2 6211 },
f8572d8f 6212 { }
1da177e4
LT		 6213};
6214
2c8c1e72 6215struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
760f2d01
DL		 6216{
6217 struct ctl_table *table;
6218
6219 table = kmemdup(ipv6_route_table_template,
6220 sizeof(ipv6_route_table_template),
6221 GFP_KERNEL);
5ee09105
YH		 6222
6223 if (table) {
6224 table[0].data = &net->ipv6.sysctl.flush_delay;
c486da34 6225 table[0].extra1 = net;
86393e52 6226 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
5ee09105
YH		 6227 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
6228 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6229 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6230 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6231 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6232 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6233 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
9c69fabe 6234 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
7c6bb7d2 6235 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
464dc801
EB		 6236
6237 /* Don't export sysctls to unprivileged users */
6238 if (net->user_ns != &init_user_ns)
6239 table[0].procname = NULL;
5ee09105
YH		 6240 }
6241
760f2d01
DL		 6242 return table;
6243}
1da177e4
LT		 6244#endif
6245
2c8c1e72 6246static int __net_init ip6_route_net_init(struct net *net)
cdb18761 6247{
633d424b 6248 int ret = -ENOMEM;
8ed67789 6249
86393e52
AD		 6250 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6251 sizeof(net->ipv6.ip6_dst_ops));
f2fc6a54 6252
fc66f95c
ED		 6253 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6254 goto out_ip6_dst_ops;
6255
1cf844c7 6256 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
421842ed
DA		 6257 if (!net->ipv6.fib6_null_entry)
6258 goto out_ip6_dst_entries;
1cf844c7
DA		 6259 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6260 sizeof(*net->ipv6.fib6_null_entry));
421842ed 6261
8ed67789
DL		 6262 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6263 sizeof(*net->ipv6.ip6_null_entry),
6264 GFP_KERNEL);
6265 if (!net->ipv6.ip6_null_entry)
421842ed 6266 goto out_fib6_null_entry;
d8d1f30b 6267 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6268 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6269 ip6_template_metrics, true);
74109218 6270 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->rt6i_uncached);
8ed67789
DL		 6271
6272#ifdef CONFIG_IPV6_MULTIPLE_TABLES
feca7d8c 6273 net->ipv6.fib6_has_custom_rules = false;
8ed67789
DL		 6274 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6275 sizeof(*net->ipv6.ip6_prohibit_entry),
6276 GFP_KERNEL);
68fffc67
PZ		 6277 if (!net->ipv6.ip6_prohibit_entry)
6278 goto out_ip6_null_entry;
d8d1f30b 6279 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6280 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6281 ip6_template_metrics, true);
74109218 6282 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->rt6i_uncached);
8ed67789
DL		 6283
6284 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6285 sizeof(*net->ipv6.ip6_blk_hole_entry),
6286 GFP_KERNEL);
68fffc67
PZ		 6287 if (!net->ipv6.ip6_blk_hole_entry)
6288 goto out_ip6_prohibit_entry;
d8d1f30b 6289 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6290 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6291 ip6_template_metrics, true);
74109218 6292 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->rt6i_uncached);
b9b33e7c
PA		 6293#ifdef CONFIG_IPV6_SUBTREES
6294 net->ipv6.fib6_routes_require_src = 0;
6295#endif
8ed67789
DL		 6296#endif
6297
b339a47c
PZ		 6298 net->ipv6.sysctl.flush_delay = 0;
6299 net->ipv6.sysctl.ip6_rt_max_size = 4096;
6300 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6301 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6302 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6303 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6304 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6305 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
7c6bb7d2 6306 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
b339a47c 6307
6891a346
BT		 6308 net->ipv6.ip6_rt_gc_expire = 30*HZ;
6309
8ed67789
DL		 6310 ret = 0;
6311out:
6312 return ret;
f2fc6a54 6313
68fffc67
PZ		 6314#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6315out_ip6_prohibit_entry:
6316 kfree(net->ipv6.ip6_prohibit_entry);
6317out_ip6_null_entry:
6318 kfree(net->ipv6.ip6_null_entry);
6319#endif
421842ed
DA		 6320out_fib6_null_entry:
6321 kfree(net->ipv6.fib6_null_entry);
fc66f95c
ED		 6322out_ip6_dst_entries:
6323 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
f2fc6a54 6324out_ip6_dst_ops:
f2fc6a54 6325 goto out;
cdb18761
DL		 6326}
6327
2c8c1e72 6328static void __net_exit ip6_route_net_exit(struct net *net)
cdb18761 6329{
421842ed 6330 kfree(net->ipv6.fib6_null_entry);
8ed67789
DL		 6331 kfree(net->ipv6.ip6_null_entry);
6332#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6333 kfree(net->ipv6.ip6_prohibit_entry);
6334 kfree(net->ipv6.ip6_blk_hole_entry);
6335#endif
41bb78b4 6336 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
cdb18761
DL		 6337}
6338
d189634e
TG		 6339static int __net_init ip6_route_net_init_late(struct net *net)
6340{
6341#ifdef CONFIG_PROC_FS
c3506372
CH		 6342 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6343 sizeof(struct ipv6_route_iter));
3617d949
CH		 6344 proc_create_net_single("rt6_stats", 0444, net->proc_net,
6345 rt6_stats_seq_show, NULL);
d189634e
TG		 6346#endif
6347 return 0;
6348}
6349
6350static void __net_exit ip6_route_net_exit_late(struct net *net)
6351{
6352#ifdef CONFIG_PROC_FS
ece31ffd
G		 6353 remove_proc_entry("ipv6_route", net->proc_net);
6354 remove_proc_entry("rt6_stats", net->proc_net);
d189634e
TG		 6355#endif
6356}
6357
cdb18761
DL		 6358static struct pernet_operations ip6_route_net_ops = {
6359 .init = ip6_route_net_init,
6360 .exit = ip6_route_net_exit,
6361};
6362
c3426b47
DM		 6363static int __net_init ipv6_inetpeer_init(struct net *net)
6364{
6365 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6366
6367 if (!bp)
6368 return -ENOMEM;
6369 inet_peer_base_init(bp);
6370 net->ipv6.peers = bp;
6371 return 0;
6372}
6373
6374static void __net_exit ipv6_inetpeer_exit(struct net *net)
6375{
6376 struct inet_peer_base *bp = net->ipv6.peers;
6377
6378 net->ipv6.peers = NULL;
56a6b248 6379 inetpeer_invalidate_tree(bp);
c3426b47
DM		 6380 kfree(bp);
6381}
6382
2b823f72 6383static struct pernet_operations ipv6_inetpeer_ops = {
c3426b47
DM		 6384 .init = ipv6_inetpeer_init,
6385 .exit = ipv6_inetpeer_exit,
6386};
6387
d189634e
TG		 6388static struct pernet_operations ip6_route_net_late_ops = {
6389 .init = ip6_route_net_init_late,
6390 .exit = ip6_route_net_exit_late,
6391};
6392
8ed67789
DL		 6393static struct notifier_block ip6_route_dev_notifier = {
6394 .notifier_call = ip6_route_dev_notify,
242d3a49 6395 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
8ed67789
DL		 6396};
6397
2f460933
WC		 6398void __init ip6_route_init_special_entries(void)
6399{
6400 /* Registering of the loopback is done before this portion of code,
6401 * the loopback reference in rt6_info will not be taken, do it
6402 * manually for init_net */
1cf844c7 6403 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
2f460933
WC		 6404 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6405 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6406 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6407 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6408 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6409 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6410 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6411 #endif
6412}
6413
138d0be3
YS		 6414#if IS_BUILTIN(CONFIG_IPV6)
6415#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6416DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6417
15172a46
YS		 6418static const struct bpf_iter_reg ipv6_route_reg_info = {
6419 .target = "ipv6_route",
6420 .seq_ops = &ipv6_route_seq_ops,
6421 .init_seq_private = bpf_iter_init_seq_net,
6422 .fini_seq_private = bpf_iter_fini_seq_net,
6423 .seq_priv_size = sizeof(struct ipv6_route_iter),
3c32cc1b
YS		 6424 .ctx_arg_info_size = 1,
6425 .ctx_arg_info = {
6426 { offsetof(struct bpf_iter__ipv6_route, rt),
6427 PTR_TO_BTF_ID_OR_NULL },
6428 },
15172a46
YS		 6429};
6430
138d0be3
YS		 6431static int __init bpf_iter_register(void)
6432{
15172a46 6433 return bpf_iter_reg_target(&ipv6_route_reg_info);
138d0be3
YS		 6434}
6435
6436static void bpf_iter_unregister(void)
6437{
ab2ee4fc 6438 bpf_iter_unreg_target(&ipv6_route_reg_info);
138d0be3
YS		 6439}
6440#endif
6441#endif
6442
433d49c3 6443int __init ip6_route_init(void)
1da177e4 6444{
433d49c3 6445 int ret;
8d0b94af 6446 int cpu;
433d49c3 6447
9a7ec3a9
DL		 6448 ret = -ENOMEM;
6449 ip6_dst_ops_template.kmem_cachep =
e5d679f3 6450 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
f845ab6b 6451 SLAB_HWCACHE_ALIGN, NULL);
9a7ec3a9 6452 if (!ip6_dst_ops_template.kmem_cachep)
c19a28e1 6453 goto out;
14e50e57 6454
fc66f95c 6455 ret = dst_entries_init(&ip6_dst_blackhole_ops);
8ed67789 6456 if (ret)
bdb3289f 6457 goto out_kmem_cache;
bdb3289f 6458
c3426b47
DM		 6459 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6460 if (ret)
e8803b6c 6461 goto out_dst_entries;
2a0c451a 6462
7e52b33b
DM		 6463 ret = register_pernet_subsys(&ip6_route_net_ops);
6464 if (ret)
6465 goto out_register_inetpeer;
c3426b47 6466
5dc121e9
AE		 6467 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6468
e8803b6c 6469 ret = fib6_init();
433d49c3 6470 if (ret)
8ed67789 6471 goto out_register_subsys;
433d49c3 6472
433d49c3
DL		 6473 ret = xfrm6_init();
6474 if (ret)
e8803b6c 6475 goto out_fib6_init;
c35b7e72 6476
433d49c3
DL		 6477 ret = fib6_rules_init();
6478 if (ret)
6479 goto xfrm6_init;
7e5449c2 6480
d189634e
TG		 6481 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6482 if (ret)
6483 goto fib6_rules_init;
6484
16feebcf
FW		 6485 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6486 inet6_rtm_newroute, NULL, 0);
6487 if (ret < 0)
6488 goto out_register_late_subsys;
6489
6490 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6491 inet6_rtm_delroute, NULL, 0);
6492 if (ret < 0)
6493 goto out_register_late_subsys;
6494
6495 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6496 inet6_rtm_getroute, NULL,
6497 RTNL_FLAG_DOIT_UNLOCKED);
6498 if (ret < 0)
d189634e 6499 goto out_register_late_subsys;
c127ea2c 6500
8ed67789 6501 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
cdb18761 6502 if (ret)
d189634e 6503 goto out_register_late_subsys;
8ed67789 6504
138d0be3
YS		 6505#if IS_BUILTIN(CONFIG_IPV6)
6506#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6507 ret = bpf_iter_register();
6508 if (ret)
6509 goto out_register_late_subsys;
6510#endif
6511#endif
6512
8d0b94af
MKL		 6513 for_each_possible_cpu(cpu) {
6514 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6515
6516 INIT_LIST_HEAD(&ul->head);
6517 spin_lock_init(&ul->lock);
6518 }
6519
433d49c3
DL		 6520out:
6521 return ret;
6522
d189634e 6523out_register_late_subsys:
16feebcf 6524 rtnl_unregister_all(PF_INET6);
d189634e 6525 unregister_pernet_subsys(&ip6_route_net_late_ops);
433d49c3 6526fib6_rules_init:
433d49c3
DL		 6527 fib6_rules_cleanup();
6528xfrm6_init:
433d49c3 6529 xfrm6_fini();
2a0c451a
TG		 6530out_fib6_init:
6531 fib6_gc_cleanup();
8ed67789
DL		 6532out_register_subsys:
6533 unregister_pernet_subsys(&ip6_route_net_ops);
7e52b33b
DM		 6534out_register_inetpeer:
6535 unregister_pernet_subsys(&ipv6_inetpeer_ops);
fc66f95c
ED		 6536out_dst_entries:
6537 dst_entries_destroy(&ip6_dst_blackhole_ops);
433d49c3 6538out_kmem_cache:
f2fc6a54 6539 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
433d49c3 6540 goto out;
1da177e4
LT		 6541}
6542
6543void ip6_route_cleanup(void)
6544{
138d0be3
YS		 6545#if IS_BUILTIN(CONFIG_IPV6)
6546#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6547 bpf_iter_unregister();
6548#endif
6549#endif
8ed67789 6550 unregister_netdevice_notifier(&ip6_route_dev_notifier);
d189634e 6551 unregister_pernet_subsys(&ip6_route_net_late_ops);
101367c2 6552 fib6_rules_cleanup();
1da177e4 6553 xfrm6_fini();
1da177e4 6554 fib6_gc_cleanup();
c3426b47 6555 unregister_pernet_subsys(&ipv6_inetpeer_ops);
8ed67789 6556 unregister_pernet_subsys(&ip6_route_net_ops);
41bb78b4 6557 dst_entries_destroy(&ip6_dst_blackhole_ops);
f2fc6a54 6558 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
1da177e4 6559}
Linux 6.x block layer and io_uring tree(s)