projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'linux-can-fixes-for-5.14-20210826' of git://git.kernel.org/pub/scm/linux...
[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>
951cf368 64#include <linux/btf_ids.h>
1da177e4
LT		 65
66#ifdef CONFIG_SYSCTL
67#include <linux/sysctl.h>
68#endif
69
30d444d3
DA		 70static int ip6_rt_type_to_error(u8 fib6_type);
71
72#define CREATE_TRACE_POINTS
73#include <trace/events/fib6.h>
74EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
75#undef CREATE_TRACE_POINTS
76
afc154e9 77enum rt6_nud_state {
7e980569
JB		 78 RT6_NUD_FAIL_HARD = -3,
79 RT6_NUD_FAIL_PROBE = -2,
80 RT6_NUD_FAIL_DO_RR = -1,
afc154e9
HFS		 81 RT6_NUD_SUCCEED = 1
82};
83
bbd807df
BV		 84INDIRECT_CALLABLE_SCOPE
85struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
0dbaee3b 86static unsigned int ip6_default_advmss(const struct dst_entry *dst);
f67fbeae
BV		 87INDIRECT_CALLABLE_SCOPE
88unsigned int ip6_mtu(const struct dst_entry *dst);
1da177e4
LT		 89static struct dst_entry *ip6_negative_advice(struct dst_entry *);
90static void ip6_dst_destroy(struct dst_entry *);
91static void ip6_dst_ifdown(struct dst_entry *,
92 struct net_device *dev, int how);
569d3645 93static int ip6_dst_gc(struct dst_ops *ops);
1da177e4
LT		 94
95static int ip6_pkt_discard(struct sk_buff *skb);
ede2059d 96static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
7150aede 97static int ip6_pkt_prohibit(struct sk_buff *skb);
ede2059d 98static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1da177e4 99static void ip6_link_failure(struct sk_buff *skb);
6700c270 100static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
bd085ef6
HL		 101 struct sk_buff *skb, u32 mtu,
102 bool confirm_neigh);
6700c270
DM		 103static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
104 struct sk_buff *skb);
702cea56
DA		 105static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
106 int strict);
a1b7a1f0 107static size_t rt6_nlmsg_size(struct fib6_info *f6i);
d4ead6b3 108static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 109 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 110 struct in6_addr *dest, struct in6_addr *src,
16a16cd3
DA		 111 int iif, int type, u32 portid, u32 seq,
112 unsigned int flags);
7e4b5128 113static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
510e2ced
WW		 114 const struct in6_addr *daddr,
115 const struct in6_addr *saddr);
1da177e4 116
70ceb4f5 117#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 118static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 119 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 120 const struct in6_addr *gwaddr,
121 struct net_device *dev,
95c96174 122 unsigned int pref);
8d1c802b 123static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 124 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 125 const struct in6_addr *gwaddr,
126 struct net_device *dev);
70ceb4f5
YH		 127#endif
128
8d0b94af
MKL		 129struct uncached_list {
130 spinlock_t lock;
131 struct list_head head;
132};
133
134static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
135
510c321b 136void rt6_uncached_list_add(struct rt6_info *rt)
8d0b94af
MKL		 137{
138 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
139
8d0b94af
MKL		 140 rt->rt6i_uncached_list = ul;
141
142 spin_lock_bh(&ul->lock);
143 list_add_tail(&rt->rt6i_uncached, &ul->head);
144 spin_unlock_bh(&ul->lock);
145}
146
510c321b 147void rt6_uncached_list_del(struct rt6_info *rt)
8d0b94af
MKL		 148{
149 if (!list_empty(&rt->rt6i_uncached)) {
150 struct uncached_list *ul = rt->rt6i_uncached_list;
81eb8447 151 struct net *net = dev_net(rt->dst.dev);
8d0b94af
MKL		 152
153 spin_lock_bh(&ul->lock);
154 list_del(&rt->rt6i_uncached);
81eb8447 155 atomic_dec(&net->ipv6.rt6_stats->fib_rt_uncache);
8d0b94af
MKL		 156 spin_unlock_bh(&ul->lock);
157 }
158}
159
160static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
161{
162 struct net_device *loopback_dev = net->loopback_dev;
163 int cpu;
164
e332bc67
EB		 165 if (dev == loopback_dev)
166 return;
167
8d0b94af
MKL		 168 for_each_possible_cpu(cpu) {
169 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
170 struct rt6_info *rt;
171
172 spin_lock_bh(&ul->lock);
173 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
174 struct inet6_dev *rt_idev = rt->rt6i_idev;
175 struct net_device *rt_dev = rt->dst.dev;
176
e332bc67 177 if (rt_idev->dev == dev) {
8d0b94af
MKL		 178 rt->rt6i_idev = in6_dev_get(loopback_dev);
179 in6_dev_put(rt_idev);
180 }
181
e332bc67 182 if (rt_dev == dev) {
8d7017fd 183 rt->dst.dev = blackhole_netdev;
8d0b94af
MKL		 184 dev_hold(rt->dst.dev);
185 dev_put(rt_dev);
186 }
187 }
188 spin_unlock_bh(&ul->lock);
189 }
190}
191
f8a1b43b 192static inline const void *choose_neigh_daddr(const struct in6_addr *p,
f894cbf8
DM		 193 struct sk_buff *skb,
194 const void *daddr)
39232973 195{
a7563f34 196 if (!ipv6_addr_any(p))
39232973 197 return (const void *) p;
f894cbf8
DM		 198 else if (skb)
199 return &ipv6_hdr(skb)->daddr;
39232973
DM		 200 return daddr;
201}
202
f8a1b43b
DA		 203struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
204 struct net_device *dev,
205 struct sk_buff *skb,
206 const void *daddr)
d3aaeb38 207{
39232973
DM		 208 struct neighbour *n;
209
f8a1b43b
DA		 210 daddr = choose_neigh_daddr(gw, skb, daddr);
211 n = __ipv6_neigh_lookup(dev, daddr);
f83c7790
DM		 212 if (n)
213 return n;
7adf3246
SB		 214
215 n = neigh_create(&nd_tbl, daddr, dev);
216 return IS_ERR(n) ? NULL : n;
f8a1b43b
DA		 217}
218
219static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
220 struct sk_buff *skb,
221 const void *daddr)
222{
223 const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
224
2c6b55f4
ND		 225 return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
226 dst->dev, skb, daddr);
f83c7790
DM		 227}
228
63fca65d
JA		 229static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
230{
231 struct net_device *dev = dst->dev;
232 struct rt6_info *rt = (struct rt6_info *)dst;
233
cbfd6891 234 daddr = choose_neigh_daddr(rt6_nexthop(rt, &in6addr_any), NULL, daddr);
63fca65d
JA		 235 if (!daddr)
236 return;
237 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
238 return;
239 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
240 return;
241 __ipv6_confirm_neigh(dev, daddr);
242}
243
9a7ec3a9 244static struct dst_ops ip6_dst_ops_template = {
1da177e4 245 .family = AF_INET6,
1da177e4
LT		 246 .gc = ip6_dst_gc,
247 .gc_thresh = 1024,
248 .check = ip6_dst_check,
0dbaee3b 249 .default_advmss = ip6_default_advmss,
ebb762f2 250 .mtu = ip6_mtu,
d4ead6b3 251 .cow_metrics = dst_cow_metrics_generic,
1da177e4
LT		 252 .destroy = ip6_dst_destroy,
253 .ifdown = ip6_dst_ifdown,
254 .negative_advice = ip6_negative_advice,
255 .link_failure = ip6_link_failure,
256 .update_pmtu = ip6_rt_update_pmtu,
6e157b6a 257 .redirect = rt6_do_redirect,
9f8955cc 258 .local_out = __ip6_local_out,
f8a1b43b 259 .neigh_lookup = ip6_dst_neigh_lookup,
63fca65d 260 .confirm_neigh = ip6_confirm_neigh,
1da177e4
LT		 261};
262
14e50e57 263static struct dst_ops ip6_dst_blackhole_ops = {
c4c877b2
DB		 264 .family = AF_INET6,
265 .default_advmss = ip6_default_advmss,
266 .neigh_lookup = ip6_dst_neigh_lookup,
267 .check = ip6_dst_check,
268 .destroy = ip6_dst_destroy,
269 .cow_metrics = dst_cow_metrics_generic,
270 .update_pmtu = dst_blackhole_update_pmtu,
271 .redirect = dst_blackhole_redirect,
272 .mtu = dst_blackhole_mtu,
14e50e57
DM		 273};
274
62fa8a84 275static const u32 ip6_template_metrics[RTAX_MAX] = {
14edd87d 276 [RTAX_HOPLIMIT - 1] = 0,
62fa8a84
DM		 277};
278
8d1c802b 279static const struct fib6_info fib6_null_entry_template = {
93c2fb25
DA		 280 .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
281 .fib6_protocol = RTPROT_KERNEL,
282 .fib6_metric = ~(u32)0,
f05713e0 283 .fib6_ref = REFCOUNT_INIT(1),
421842ed
DA		 284 .fib6_type = RTN_UNREACHABLE,
285 .fib6_metrics = (struct dst_metrics *)&dst_default_metrics,
286};
287
fb0af4c7 288static const struct rt6_info ip6_null_entry_template = {
d8d1f30b
CG		 289 .dst = {
290 .__refcnt = ATOMIC_INIT(1),
291 .__use = 1,
2c20cbd7 292 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 293 .error = -ENETUNREACH,
d8d1f30b
CG		 294 .input = ip6_pkt_discard,
295 .output = ip6_pkt_discard_out,
1da177e4
LT		 296 },
297 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
1da177e4
LT		 298};
299
101367c2
TG		 300#ifdef CONFIG_IPV6_MULTIPLE_TABLES
301
fb0af4c7 302static const struct rt6_info ip6_prohibit_entry_template = {
d8d1f30b
CG		 303 .dst = {
304 .__refcnt = ATOMIC_INIT(1),
305 .__use = 1,
2c20cbd7 306 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 307 .error = -EACCES,
d8d1f30b
CG		 308 .input = ip6_pkt_prohibit,
309 .output = ip6_pkt_prohibit_out,
101367c2
TG		 310 },
311 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 312};
313
fb0af4c7 314static const struct rt6_info ip6_blk_hole_entry_template = {
d8d1f30b
CG		 315 .dst = {
316 .__refcnt = ATOMIC_INIT(1),
317 .__use = 1,
2c20cbd7 318 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 319 .error = -EINVAL,
d8d1f30b 320 .input = dst_discard,
ede2059d 321 .output = dst_discard_out,
101367c2
TG		 322 },
323 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 324};
325
326#endif
327
ebfa45f0
MKL		 328static void rt6_info_init(struct rt6_info *rt)
329{
330 struct dst_entry *dst = &rt->dst;
331
332 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
ebfa45f0
MKL		 333 INIT_LIST_HEAD(&rt->rt6i_uncached);
334}
335
1da177e4 336/* allocate dst with ip6_dst_ops */
93531c67
DA		 337struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
338 int flags)
1da177e4 339{
97bab73f 340 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
b2a9c0ed 341 1, DST_OBSOLETE_FORCE_CHK, flags);
cf911662 342
81eb8447 343 if (rt) {
ebfa45f0 344 rt6_info_init(rt);
81eb8447
WW		 345 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
346 }
8104891b 347
cf911662 348 return rt;
1da177e4 349}
9ab179d8 350EXPORT_SYMBOL(ip6_dst_alloc);
d52d3997 351
1da177e4
LT		 352static void ip6_dst_destroy(struct dst_entry *dst)
353{
354 struct rt6_info *rt = (struct rt6_info *)dst;
a68886a6 355 struct fib6_info *from;
8d0b94af 356 struct inet6_dev *idev;
1da177e4 357
1620a336 358 ip_dst_metrics_put(dst);
8d0b94af
MKL		 359 rt6_uncached_list_del(rt);
360
361 idev = rt->rt6i_idev;
38308473 362 if (idev) {
1da177e4
LT		 363 rt->rt6i_idev = NULL;
364 in6_dev_put(idev);
1ab1457c 365 }
1716a961 366
0e233874 367 from = xchg((__force struct fib6_info **)&rt->from, NULL);
93531c67 368 fib6_info_release(from);
b3419363
DM		 369}
370
1da177e4
LT		 371static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
372 int how)
373{
374 struct rt6_info *rt = (struct rt6_info *)dst;
375 struct inet6_dev *idev = rt->rt6i_idev;
5a3e55d6 376 struct net_device *loopback_dev =
c346dca1 377 dev_net(dev)->loopback_dev;
1da177e4 378
e5645f51
WW		 379 if (idev && idev->dev != loopback_dev) {
380 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
381 if (loopback_idev) {
382 rt->rt6i_idev = loopback_idev;
383 in6_dev_put(idev);
97cac082 384 }
1da177e4
LT		 385 }
386}
387
5973fb1e
MKL		 388static bool __rt6_check_expired(const struct rt6_info *rt)
389{
390 if (rt->rt6i_flags & RTF_EXPIRES)
391 return time_after(jiffies, rt->dst.expires);
392 else
393 return false;
394}
395
a50feda5 396static bool rt6_check_expired(const struct rt6_info *rt)
1da177e4 397{
a68886a6
DA		 398 struct fib6_info *from;
399
400 from = rcu_dereference(rt->from);
401
1716a961
G		 402 if (rt->rt6i_flags & RTF_EXPIRES) {
403 if (time_after(jiffies, rt->dst.expires))
a50feda5 404 return true;
a68886a6 405 } else if (from) {
1e2ea8ad 406 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
a68886a6 407 fib6_check_expired(from);
1716a961 408 }
a50feda5 409 return false;
1da177e4
LT		 410}
411
b1d40991
DA		 412void fib6_select_path(const struct net *net, struct fib6_result *res,
413 struct flowi6 *fl6, int oif, bool have_oif_match,
414 const struct sk_buff *skb, int strict)
51ebd318 415{
8d1c802b 416 struct fib6_info *sibling, *next_sibling;
b1d40991
DA		 417 struct fib6_info *match = res->f6i;
418
34fe5a1c 419 if (!match->nh && (!match->fib6_nsiblings || have_oif_match))
b1d40991 420 goto out;
51ebd318 421
34fe5a1c
DA		 422 if (match->nh && have_oif_match && res->nh)
423 return;
424
b673d6cc
JS		 425 /* We might have already computed the hash for ICMPv6 errors. In such
426 * case it will always be non-zero. Otherwise now is the time to do it.
427 */
f88d8ea6
DA		 428 if (!fl6->mp_hash &&
429 (!match->nh || nexthop_is_multipath(match->nh)))
b4bac172 430 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
b673d6cc 431
f88d8ea6
DA		 432 if (unlikely(match->nh)) {
433 nexthop_path_fib6_result(res, fl6->mp_hash);
434 return;
435 }
436
1cf844c7 437 if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
b1d40991 438 goto out;
3d709f69 439
93c2fb25
DA		 440 list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
441 fib6_siblings) {
1cf844c7 442 const struct fib6_nh *nh = sibling->fib6_nh;
5e670d84
DA		 443 int nh_upper_bound;
444
702cea56 445 nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
5e670d84 446 if (fl6->mp_hash > nh_upper_bound)
3d709f69 447 continue;
702cea56 448 if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
3d709f69
IS		 449 break;
450 match = sibling;
451 break;
452 }
453
b1d40991
DA		 454out:
455 res->f6i = match;
1cf844c7 456 res->nh = match->fib6_nh;
51ebd318
ND		 457}
458
1da177e4 459/*
66f5d6ce 460 * Route lookup. rcu_read_lock() should be held.
1da177e4
LT		 461 */
462
0c59d006
DA		 463static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
464 const struct in6_addr *saddr, int oif, int flags)
465{
466 const struct net_device *dev;
467
468 if (nh->fib_nh_flags & RTNH_F_DEAD)
469 return false;
470
471 dev = nh->fib_nh_dev;
472 if (oif) {
473 if (dev->ifindex == oif)
474 return true;
475 } else {
476 if (ipv6_chk_addr(net, saddr, dev,
477 flags & RT6_LOOKUP_F_IFACE))
478 return true;
479 }
480
481 return false;
482}
483
962b6803
DA		 484struct fib6_nh_dm_arg {
485 struct net *net;
486 const struct in6_addr *saddr;
487 int oif;
488 int flags;
489 struct fib6_nh *nh;
490};
491
492static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
493{
494 struct fib6_nh_dm_arg *arg = _arg;
495
496 arg->nh = nh;
497 return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
498 arg->flags);
499}
500
501/* returns fib6_nh from nexthop or NULL */
502static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
503 struct fib6_result *res,
504 const struct in6_addr *saddr,
505 int oif, int flags)
506{
507 struct fib6_nh_dm_arg arg = {
508 .net = net,
509 .saddr = saddr,
510 .oif = oif,
511 .flags = flags,
512 };
513
514 if (nexthop_is_blackhole(nh))
515 return NULL;
516
517 if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
518 return arg.nh;
519
520 return NULL;
521}
522
75ef7389
DA		 523static void rt6_device_match(struct net *net, struct fib6_result *res,
524 const struct in6_addr *saddr, int oif, int flags)
1da177e4 525{
75ef7389
DA		 526 struct fib6_info *f6i = res->f6i;
527 struct fib6_info *spf6i;
528 struct fib6_nh *nh;
1da177e4 529
75ef7389 530 if (!oif && ipv6_addr_any(saddr)) {
f88d8ea6
DA		 531 if (unlikely(f6i->nh)) {
532 nh = nexthop_fib6_nh(f6i->nh);
533 if (nexthop_is_blackhole(f6i->nh))
534 goto out_blackhole;
535 } else {
536 nh = f6i->fib6_nh;
537 }
7d21fec9
DA		 538 if (!(nh->fib_nh_flags & RTNH_F_DEAD))
539 goto out;
75ef7389 540 }
dd3abc4e 541
75ef7389 542 for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
962b6803
DA		 543 bool matched = false;
544
545 if (unlikely(spf6i->nh)) {
546 nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
547 oif, flags);
548 if (nh)
549 matched = true;
550 } else {
551 nh = spf6i->fib6_nh;
552 if (__rt6_device_match(net, nh, saddr, oif, flags))
553 matched = true;
554 }
555 if (matched) {
75ef7389 556 res->f6i = spf6i;
7d21fec9 557 goto out;
75ef7389 558 }
dd3abc4e 559 }
1da177e4 560
75ef7389
DA		 561 if (oif && flags & RT6_LOOKUP_F_IFACE) {
562 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 563 nh = res->f6i->fib6_nh;
7d21fec9 564 goto out;
75ef7389 565 }
8067bb8c 566
f88d8ea6
DA		 567 if (unlikely(f6i->nh)) {
568 nh = nexthop_fib6_nh(f6i->nh);
569 if (nexthop_is_blackhole(f6i->nh))
570 goto out_blackhole;
571 } else {
572 nh = f6i->fib6_nh;
573 }
574
7d21fec9 575 if (nh->fib_nh_flags & RTNH_F_DEAD) {
75ef7389 576 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 577 nh = res->f6i->fib6_nh;
75ef7389 578 }
7d21fec9
DA		 579out:
580 res->nh = nh;
581 res->fib6_type = res->f6i->fib6_type;
582 res->fib6_flags = res->f6i->fib6_flags;
f88d8ea6
DA		 583 return;
584
585out_blackhole:
586 res->fib6_flags |= RTF_REJECT;
587 res->fib6_type = RTN_BLACKHOLE;
588 res->nh = nh;
1da177e4
LT		 589}
590
27097255 591#ifdef CONFIG_IPV6_ROUTER_PREF
c2f17e82
HFS		 592struct __rt6_probe_work {
593 struct work_struct work;
594 struct in6_addr target;
595 struct net_device *dev;
596};
597
598static void rt6_probe_deferred(struct work_struct *w)
599{
600 struct in6_addr mcaddr;
601 struct __rt6_probe_work *work =
602 container_of(w, struct __rt6_probe_work, work);
603
604 addrconf_addr_solict_mult(&work->target, &mcaddr);
adc176c5 605 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
c2f17e82 606 dev_put(work->dev);
662f5533 607 kfree(work);
c2f17e82
HFS		 608}
609
cc3a86c8 610static void rt6_probe(struct fib6_nh *fib6_nh)
27097255 611{
f547fac6 612 struct __rt6_probe_work *work = NULL;
5e670d84 613 const struct in6_addr *nh_gw;
1bef4c22 614 unsigned long last_probe;
f2c31e32 615 struct neighbour *neigh;
5e670d84 616 struct net_device *dev;
f547fac6 617 struct inet6_dev *idev;
5e670d84 618
27097255
YH		 619 /*
620 * Okay, this does not seem to be appropriate
621 * for now, however, we need to check if it
622 * is really so; aka Router Reachability Probing.
623 *
624 * Router Reachability Probe MUST be rate-limited
625 * to no more than one per minute.
626 */
004b3942 627 if (!fib6_nh->fib_nh_gw_family)
7ff74a59 628 return;
5e670d84 629
cc3a86c8
DA		 630 nh_gw = &fib6_nh->fib_nh_gw6;
631 dev = fib6_nh->fib_nh_dev;
2152caea 632 rcu_read_lock_bh();
1bef4c22 633 last_probe = READ_ONCE(fib6_nh->last_probe);
f547fac6 634 idev = __in6_dev_get(dev);
5e670d84 635 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
2152caea 636 if (neigh) {
8d6c31bf
MKL		 637 if (neigh->nud_state & NUD_VALID)
638 goto out;
639
2152caea 640 write_lock(&neigh->lock);
990edb42
MKL		 641 if (!(neigh->nud_state & NUD_VALID) &&
642 time_after(jiffies,
dcd1f572 643 neigh->updated + idev->cnf.rtr_probe_interval)) {
990edb42
MKL		 644 work = kmalloc(sizeof(*work), GFP_ATOMIC);
645 if (work)
646 __neigh_set_probe_once(neigh);
c2f17e82 647 }
2152caea 648 write_unlock(&neigh->lock);
1bef4c22 649 } else if (time_after(jiffies, last_probe +
f547fac6 650 idev->cnf.rtr_probe_interval)) {
990edb42 651 work = kmalloc(sizeof(*work), GFP_ATOMIC);
f2c31e32 652 }
990edb42 653
1bef4c22
ED		 654 if (!work || cmpxchg(&fib6_nh->last_probe,
655 last_probe, jiffies) != last_probe) {
656 kfree(work);
657 } else {
990edb42 658 INIT_WORK(&work->work, rt6_probe_deferred);
5e670d84
DA		 659 work->target = *nh_gw;
660 dev_hold(dev);
661 work->dev = dev;
990edb42
MKL		 662 schedule_work(&work->work);
663 }
664
8d6c31bf 665out:
2152caea 666 rcu_read_unlock_bh();
27097255
YH		 667}
668#else
cc3a86c8 669static inline void rt6_probe(struct fib6_nh *fib6_nh)
27097255 670{
27097255
YH		 671}
672#endif
673
1da177e4 674/*
554cfb7e 675 * Default Router Selection (RFC 2461 6.3.6)
1da177e4 676 */
1ba9a895 677static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
1da177e4 678{
afc154e9 679 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
5e670d84 680 struct neighbour *neigh;
f2c31e32 681
145a3621 682 rcu_read_lock_bh();
1ba9a895
DA		 683 neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
684 &fib6_nh->fib_nh_gw6);
145a3621
YH		 685 if (neigh) {
686 read_lock(&neigh->lock);
554cfb7e 687 if (neigh->nud_state & NUD_VALID)
afc154e9 688 ret = RT6_NUD_SUCCEED;
398bcbeb 689#ifdef CONFIG_IPV6_ROUTER_PREF
a5a81f0b 690 else if (!(neigh->nud_state & NUD_FAILED))
afc154e9 691 ret = RT6_NUD_SUCCEED;
7e980569
JB		 692 else
693 ret = RT6_NUD_FAIL_PROBE;
398bcbeb 694#endif
145a3621 695 read_unlock(&neigh->lock);
afc154e9
HFS		 696 } else {
697 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
7e980569 698 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
a5a81f0b 699 }
145a3621
YH		 700 rcu_read_unlock_bh();
701
a5a81f0b 702 return ret;
1da177e4
LT		 703}
704
702cea56
DA		 705static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
706 int strict)
1da177e4 707{
6e1809a5
DA		 708 int m = 0;
709
710 if (!oif || nh->fib_nh_dev->ifindex == oif)
711 m = 2;
1ab1457c 712
77d16f45 713 if (!m && (strict & RT6_LOOKUP_F_IFACE))
afc154e9 714 return RT6_NUD_FAIL_HARD;
ebacaaa0 715#ifdef CONFIG_IPV6_ROUTER_PREF
702cea56 716 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
ebacaaa0 717#endif
1ba9a895 718 if ((strict & RT6_LOOKUP_F_REACHABLE) &&
702cea56 719 !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
1ba9a895 720 int n = rt6_check_neigh(nh);
afc154e9
HFS		 721 if (n < 0)
722 return n;
723 }
554cfb7e
YH		 724 return m;
725}
726
28679ed1
DA		 727static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
728 int oif, int strict, int *mpri, bool *do_rr)
554cfb7e 729{
afc154e9 730 bool match_do_rr = false;
28679ed1
DA		 731 bool rc = false;
732 int m;
35103d11 733
28679ed1 734 if (nh->fib_nh_flags & RTNH_F_DEAD)
8067bb8c
IS		 735 goto out;
736
28679ed1
DA		 737 if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
738 nh->fib_nh_flags & RTNH_F_LINKDOWN &&
d5d32e4b 739 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
35103d11 740 goto out;
f11e6659 741
28679ed1 742 m = rt6_score_route(nh, fib6_flags, oif, strict);
7e980569 743 if (m == RT6_NUD_FAIL_DO_RR) {
afc154e9
HFS		 744 match_do_rr = true;
745 m = 0; /* lowest valid score */
7e980569 746 } else if (m == RT6_NUD_FAIL_HARD) {
f11e6659 747 goto out;
afc154e9
HFS		 748 }
749
750 if (strict & RT6_LOOKUP_F_REACHABLE)
28679ed1 751 rt6_probe(nh);
f11e6659 752
7e980569 753 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
f11e6659 754 if (m > *mpri) {
afc154e9 755 *do_rr = match_do_rr;
f11e6659 756 *mpri = m;
28679ed1 757 rc = true;
f11e6659 758 }
f11e6659 759out:
28679ed1 760 return rc;
f11e6659
DM		 761}
762
17a5984e
DA		 763struct fib6_nh_frl_arg {
764 u32 flags;
765 int oif;
766 int strict;
767 int *mpri;
768 bool *do_rr;
769 struct fib6_nh *nh;
770};
771
772static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
773{
774 struct fib6_nh_frl_arg *arg = _arg;
775
776 arg->nh = nh;
777 return find_match(nh, arg->flags, arg->oif, arg->strict,
778 arg->mpri, arg->do_rr);
779}
780
b7bc4b6a 781static void __find_rr_leaf(struct fib6_info *f6i_start,
30c15f03 782 struct fib6_info *nomatch, u32 metric,
b7bc4b6a 783 struct fib6_result *res, struct fib6_info **cont,
30c15f03 784 int oif, int strict, bool *do_rr, int *mpri)
f11e6659 785{
b7bc4b6a 786 struct fib6_info *f6i;
1da177e4 787
b7bc4b6a
DA		 788 for (f6i = f6i_start;
789 f6i && f6i != nomatch;
790 f6i = rcu_dereference(f6i->fib6_next)) {
17a5984e 791 bool matched = false;
30c15f03
DA		 792 struct fib6_nh *nh;
793
b7bc4b6a
DA		 794 if (cont && f6i->fib6_metric != metric) {
795 *cont = f6i;
30c15f03 796 return;
9fbdcfaf
SK		 797 }
798
b7bc4b6a 799 if (fib6_check_expired(f6i))
28679ed1
DA		 800 continue;
801
17a5984e
DA		 802 if (unlikely(f6i->nh)) {
803 struct fib6_nh_frl_arg arg = {
804 .flags = f6i->fib6_flags,
805 .oif = oif,
806 .strict = strict,
807 .mpri = mpri,
808 .do_rr = do_rr
809 };
810
811 if (nexthop_is_blackhole(f6i->nh)) {
812 res->fib6_flags = RTF_REJECT;
813 res->fib6_type = RTN_BLACKHOLE;
814 res->f6i = f6i;
815 res->nh = nexthop_fib6_nh(f6i->nh);
816 return;
817 }
818 if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
819 &arg)) {
820 matched = true;
821 nh = arg.nh;
822 }
823 } else {
824 nh = f6i->fib6_nh;
825 if (find_match(nh, f6i->fib6_flags, oif, strict,
826 mpri, do_rr))
827 matched = true;
828 }
829 if (matched) {
b7bc4b6a
DA		 830 res->f6i = f6i;
831 res->nh = nh;
7d21fec9
DA		 832 res->fib6_flags = f6i->fib6_flags;
833 res->fib6_type = f6i->fib6_type;
b7bc4b6a 834 }
9fbdcfaf 835 }
30c15f03 836}
9fbdcfaf 837
b7bc4b6a
DA		 838static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
839 struct fib6_info *rr_head, int oif, int strict,
840 bool *do_rr, struct fib6_result *res)
30c15f03 841{
b7bc4b6a
DA		 842 u32 metric = rr_head->fib6_metric;
843 struct fib6_info *cont = NULL;
30c15f03 844 int mpri = -1;
9fbdcfaf 845
b7bc4b6a 846 __find_rr_leaf(rr_head, NULL, metric, res, &cont,
30c15f03 847 oif, strict, do_rr, &mpri);
28679ed1 848
b7bc4b6a 849 __find_rr_leaf(leaf, rr_head, metric, res, &cont,
30c15f03 850 oif, strict, do_rr, &mpri);
9fbdcfaf 851
b7bc4b6a
DA		 852 if (res->f6i || !cont)
853 return;
9fbdcfaf 854
b7bc4b6a 855 __find_rr_leaf(cont, NULL, metric, res, NULL,
30c15f03 856 oif, strict, do_rr, &mpri);
f11e6659 857}
1da177e4 858
b7bc4b6a
DA		 859static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
860 struct fib6_result *res, int strict)
f11e6659 861{
8d1c802b 862 struct fib6_info *leaf = rcu_dereference(fn->leaf);
b7bc4b6a 863 struct fib6_info *rt0;
afc154e9 864 bool do_rr = false;
17ecf590 865 int key_plen;
1da177e4 866
b7bc4b6a
DA		 867 /* make sure this function or its helpers sets f6i */
868 res->f6i = NULL;
869
421842ed 870 if (!leaf || leaf == net->ipv6.fib6_null_entry)
b7bc4b6a 871 goto out;
8d1040e8 872
66f5d6ce 873 rt0 = rcu_dereference(fn->rr_ptr);
f11e6659 874 if (!rt0)
66f5d6ce 875 rt0 = leaf;
1da177e4 876
17ecf590
WW		 877 /* Double check to make sure fn is not an intermediate node
878 * and fn->leaf does not points to its child's leaf
879 * (This might happen if all routes under fn are deleted from
880 * the tree and fib6_repair_tree() is called on the node.)
881 */
93c2fb25 882 key_plen = rt0->fib6_dst.plen;
17ecf590 883#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 884 if (rt0->fib6_src.plen)
885 key_plen = rt0->fib6_src.plen;
17ecf590
WW		 886#endif
887 if (fn->fn_bit != key_plen)
b7bc4b6a 888 goto out;
1da177e4 889
b7bc4b6a 890 find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
afc154e9 891 if (do_rr) {
8fb11a9a 892 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
f11e6659 893
554cfb7e 894 /* no entries matched; do round-robin */
93c2fb25 895 if (!next || next->fib6_metric != rt0->fib6_metric)
8d1040e8 896 next = leaf;
f11e6659 897
66f5d6ce 898 if (next != rt0) {
93c2fb25 899 spin_lock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 900 /* make sure next is not being deleted from the tree */
93c2fb25 901 if (next->fib6_node)
66f5d6ce 902 rcu_assign_pointer(fn->rr_ptr, next);
93c2fb25 903 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 904 }
1da177e4 905 }
1da177e4 906
b7bc4b6a
DA		 907out:
908 if (!res->f6i) {
909 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 910 res->nh = res->f6i->fib6_nh;
7d21fec9
DA		 911 res->fib6_flags = res->f6i->fib6_flags;
912 res->fib6_type = res->f6i->fib6_type;
b7bc4b6a 913 }
1da177e4
LT		 914}
915
85bd05de 916static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
8b9df265 917{
85bd05de
DA		 918 return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
919 res->nh->fib_nh_gw_family;
8b9df265
MKL		 920}
921
70ceb4f5
YH		 922#ifdef CONFIG_IPV6_ROUTE_INFO
923int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
b71d1d42 924 const struct in6_addr *gwaddr)
70ceb4f5 925{
c346dca1 926 struct net *net = dev_net(dev);
70ceb4f5
YH		 927 struct route_info *rinfo = (struct route_info *) opt;
928 struct in6_addr prefix_buf, *prefix;
929 unsigned int pref;
4bed72e4 930 unsigned long lifetime;
8d1c802b 931 struct fib6_info *rt;
70ceb4f5
YH		 932
933 if (len < sizeof(struct route_info)) {
934 return -EINVAL;
935 }
936
937 /* Sanity check for prefix_len and length */
938 if (rinfo->length > 3) {
939 return -EINVAL;
940 } else if (rinfo->prefix_len > 128) {
941 return -EINVAL;
942 } else if (rinfo->prefix_len > 64) {
943 if (rinfo->length < 2) {
944 return -EINVAL;
945 }
946 } else if (rinfo->prefix_len > 0) {
947 if (rinfo->length < 1) {
948 return -EINVAL;
949 }
950 }
951
952 pref = rinfo->route_pref;
953 if (pref == ICMPV6_ROUTER_PREF_INVALID)
3933fc95 954 return -EINVAL;
70ceb4f5 955
4bed72e4 956 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
70ceb4f5
YH		 957
958 if (rinfo->length == 3)
959 prefix = (struct in6_addr *)rinfo->prefix;
960 else {
961 /* this function is safe */
962 ipv6_addr_prefix(&prefix_buf,
963 (struct in6_addr *)rinfo->prefix,
964 rinfo->prefix_len);
965 prefix = &prefix_buf;
966 }
967
f104a567 968 if (rinfo->prefix_len == 0)
afb1d4b5 969 rt = rt6_get_dflt_router(net, gwaddr, dev);
f104a567
DJ		 970 else
971 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
830218c1 972 gwaddr, dev);
70ceb4f5
YH		 973
974 if (rt && !lifetime) {
11dd74b3 975 ip6_del_rt(net, rt, false);
70ceb4f5
YH		 976 rt = NULL;
977 }
978
979 if (!rt && lifetime)
830218c1
DA		 980 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
981 dev, pref);
70ceb4f5 982 else if (rt)
93c2fb25
DA		 983 rt->fib6_flags = RTF_ROUTEINFO |
984 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
70ceb4f5
YH		 985
986 if (rt) {
1716a961 987 if (!addrconf_finite_timeout(lifetime))
14895687 988 fib6_clean_expires(rt);
1716a961 989 else
14895687 990 fib6_set_expires(rt, jiffies + HZ * lifetime);
1716a961 991
93531c67 992 fib6_info_release(rt);
70ceb4f5
YH		 993 }
994 return 0;
995}
996#endif
997
ae90d867
DA		 998/*
999 * Misc support functions
1000 */
1001
1002/* called with rcu_lock held */
0d161581 1003static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
ae90d867 1004{
0d161581 1005 struct net_device *dev = res->nh->fib_nh_dev;
ae90d867 1006
7d21fec9 1007 if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
ae90d867
DA		 1008 /* for copies of local routes, dst->dev needs to be the
1009 * device if it is a master device, the master device if
1010 * device is enslaved, and the loopback as the default
1011 */
1012 if (netif_is_l3_slave(dev) &&
7d21fec9 1013 !rt6_need_strict(&res->f6i->fib6_dst.addr))
ae90d867
DA		 1014 dev = l3mdev_master_dev_rcu(dev);
1015 else if (!netif_is_l3_master(dev))
1016 dev = dev_net(dev)->loopback_dev;
1017 /* last case is netif_is_l3_master(dev) is true in which
1018 * case we want dev returned to be dev
1019 */
1020 }
1021
1022 return dev;
1023}
1024
6edb3c96
DA		 1025static const int fib6_prop[RTN_MAX + 1] = {
1026 [RTN_UNSPEC] = 0,
1027 [RTN_UNICAST] = 0,
1028 [RTN_LOCAL] = 0,
1029 [RTN_BROADCAST] = 0,
1030 [RTN_ANYCAST] = 0,
1031 [RTN_MULTICAST] = 0,
1032 [RTN_BLACKHOLE] = -EINVAL,
1033 [RTN_UNREACHABLE] = -EHOSTUNREACH,
1034 [RTN_PROHIBIT] = -EACCES,
1035 [RTN_THROW] = -EAGAIN,
1036 [RTN_NAT] = -EINVAL,
1037 [RTN_XRESOLVE] = -EINVAL,
1038};
1039
1040static int ip6_rt_type_to_error(u8 fib6_type)
1041{
1042 return fib6_prop[fib6_type];
1043}
1044
8d1c802b 1045static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
3b6761d1
DA		 1046{
1047 unsigned short flags = 0;
1048
1049 if (rt->dst_nocount)
1050 flags |= DST_NOCOUNT;
1051 if (rt->dst_nopolicy)
1052 flags |= DST_NOPOLICY;
3b6761d1
DA		 1053
1054 return flags;
1055}
1056
7d21fec9 1057static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
6edb3c96 1058{
7d21fec9 1059 rt->dst.error = ip6_rt_type_to_error(fib6_type);
6edb3c96 1060
7d21fec9 1061 switch (fib6_type) {
6edb3c96
DA		 1062 case RTN_BLACKHOLE:
1063 rt->dst.output = dst_discard_out;
1064 rt->dst.input = dst_discard;
1065 break;
1066 case RTN_PROHIBIT:
1067 rt->dst.output = ip6_pkt_prohibit_out;
1068 rt->dst.input = ip6_pkt_prohibit;
1069 break;
1070 case RTN_THROW:
1071 case RTN_UNREACHABLE:
1072 default:
1073 rt->dst.output = ip6_pkt_discard_out;
1074 rt->dst.input = ip6_pkt_discard;
1075 break;
1076 }
1077}
1078
0d161581 1079static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
6edb3c96 1080{
7d21fec9 1081 struct fib6_info *f6i = res->f6i;
0d161581 1082
7d21fec9
DA		 1083 if (res->fib6_flags & RTF_REJECT) {
1084 ip6_rt_init_dst_reject(rt, res->fib6_type);
6edb3c96
DA		 1085 return;
1086 }
1087
1088 rt->dst.error = 0;
1089 rt->dst.output = ip6_output;
1090
7d21fec9 1091 if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
6edb3c96 1092 rt->dst.input = ip6_input;
7d21fec9 1093 } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
6edb3c96
DA		 1094 rt->dst.input = ip6_mc_input;
1095 } else {
1096 rt->dst.input = ip6_forward;
1097 }
1098
0d161581
DA		 1099 if (res->nh->fib_nh_lws) {
1100 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
6edb3c96
DA		 1101 lwtunnel_set_redirect(&rt->dst);
1102 }
1103
1104 rt->dst.lastuse = jiffies;
1105}
1106
e873e4b9 1107/* Caller must already hold reference to @from */
8d1c802b 1108static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
ae90d867 1109{
ae90d867 1110 rt->rt6i_flags &= ~RTF_EXPIRES;
a68886a6 1111 rcu_assign_pointer(rt->from, from);
e1255ed4 1112 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
ae90d867
DA		 1113}
1114
0d161581
DA		 1115/* Caller must already hold reference to f6i in result */
1116static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
ae90d867 1117{
0d161581
DA		 1118 const struct fib6_nh *nh = res->nh;
1119 const struct net_device *dev = nh->fib_nh_dev;
1120 struct fib6_info *f6i = res->f6i;
dcd1f572 1121
0d161581 1122 ip6_rt_init_dst(rt, res);
6edb3c96 1123
0d161581 1124 rt->rt6i_dst = f6i->fib6_dst;
dcd1f572 1125 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
7d21fec9 1126 rt->rt6i_flags = res->fib6_flags;
0d161581
DA		 1127 if (nh->fib_nh_gw_family) {
1128 rt->rt6i_gateway = nh->fib_nh_gw6;
2b2450ca
DA		 1129 rt->rt6i_flags |= RTF_GATEWAY;
1130 }
0d161581 1131 rt6_set_from(rt, f6i);
ae90d867 1132#ifdef CONFIG_IPV6_SUBTREES
0d161581 1133 rt->rt6i_src = f6i->fib6_src;
ae90d867 1134#endif
ae90d867
DA		 1135}
1136
a3c00e46
MKL		 1137static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1138 struct in6_addr *saddr)
1139{
66f5d6ce 1140 struct fib6_node *pn, *sn;
a3c00e46
MKL		 1141 while (1) {
1142 if (fn->fn_flags & RTN_TL_ROOT)
1143 return NULL;
66f5d6ce
WW		 1144 pn = rcu_dereference(fn->parent);
1145 sn = FIB6_SUBTREE(pn);
1146 if (sn && sn != fn)
6454743b 1147 fn = fib6_node_lookup(sn, NULL, saddr);
a3c00e46
MKL		 1148 else
1149 fn = pn;
1150 if (fn->fn_flags & RTN_RTINFO)
1151 return fn;
1152 }
1153}
c71099ac 1154
10585b43 1155static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
d3843fe5
WW		 1156{
1157 struct rt6_info *rt = *prt;
1158
1159 if (dst_hold_safe(&rt->dst))
1160 return true;
10585b43 1161 if (net) {
d3843fe5
WW		 1162 rt = net->ipv6.ip6_null_entry;
1163 dst_hold(&rt->dst);
1164 } else {
1165 rt = NULL;
1166 }
1167 *prt = rt;
1168 return false;
1169}
1170
dec9b0e2 1171/* called with rcu_lock held */
9b6b35ab 1172static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
dec9b0e2 1173{
9b6b35ab
DA		 1174 struct net_device *dev = res->nh->fib_nh_dev;
1175 struct fib6_info *f6i = res->f6i;
1176 unsigned short flags;
dec9b0e2
DA		 1177 struct rt6_info *nrt;
1178
9b6b35ab 1179 if (!fib6_info_hold_safe(f6i))
1c87e79a 1180 goto fallback;
e873e4b9 1181
9b6b35ab 1182 flags = fib6_info_dst_flags(f6i);
93531c67 1183 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1c87e79a 1184 if (!nrt) {
9b6b35ab 1185 fib6_info_release(f6i);
1c87e79a
XL		 1186 goto fallback;
1187 }
dec9b0e2 1188
0d161581 1189 ip6_rt_copy_init(nrt, res);
1c87e79a
XL		 1190 return nrt;
1191
1192fallback:
1193 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1194 dst_hold(&nrt->dst);
dec9b0e2
DA		 1195 return nrt;
1196}
1197
55cced4f 1198INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_lookup(struct net *net,
8ed67789 1199 struct fib6_table *table,
b75cc8f9
DA		 1200 struct flowi6 *fl6,
1201 const struct sk_buff *skb,
1202 int flags)
1da177e4 1203{
b1d40991 1204 struct fib6_result res = {};
1da177e4 1205 struct fib6_node *fn;
23fb93a4 1206 struct rt6_info *rt;
1da177e4 1207
b6cdbc85
DA		 1208 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1209 flags &= ~RT6_LOOKUP_F_IFACE;
1210
66f5d6ce 1211 rcu_read_lock();
6454743b 1212 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
c71099ac 1213restart:
b1d40991
DA		 1214 res.f6i = rcu_dereference(fn->leaf);
1215 if (!res.f6i)
1216 res.f6i = net->ipv6.fib6_null_entry;
af52a52c 1217 else
75ef7389
DA		 1218 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1219 flags);
af52a52c 1220
b1d40991 1221 if (res.f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 1222 fn = fib6_backtrack(fn, &fl6->saddr);
1223 if (fn)
1224 goto restart;
2b760fcf 1225
af52a52c
DA		 1226 rt = net->ipv6.ip6_null_entry;
1227 dst_hold(&rt->dst);
1228 goto out;
f88d8ea6
DA		 1229 } else if (res.fib6_flags & RTF_REJECT) {
1230 goto do_create;
af52a52c 1231 }
d3843fe5 1232
b1d40991
DA		 1233 fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1234 fl6->flowi6_oif != 0, skb, flags);
1235
2b760fcf 1236 /* Search through exception table */
7e4b5128 1237 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
23fb93a4 1238 if (rt) {
10585b43 1239 if (ip6_hold_safe(net, &rt))
dec9b0e2 1240 dst_use_noref(&rt->dst, jiffies);
23fb93a4 1241 } else {
f88d8ea6 1242do_create:
9b6b35ab 1243 rt = ip6_create_rt_rcu(&res);
dec9b0e2 1244 }
b811580d 1245
af52a52c 1246out:
8ff2e5b2 1247 trace_fib6_table_lookup(net, &res, table, fl6);
af52a52c 1248
66f5d6ce 1249 rcu_read_unlock();
b811580d 1250
c71099ac 1251 return rt;
c71099ac
TG		 1252}
1253
67ba4152 1254struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
b75cc8f9 1255 const struct sk_buff *skb, int flags)
ea6e574e 1256{
b75cc8f9 1257 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
ea6e574e
FW		 1258}
1259EXPORT_SYMBOL_GPL(ip6_route_lookup);
1260
9acd9f3a 1261struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
b75cc8f9
DA		 1262 const struct in6_addr *saddr, int oif,
1263 const struct sk_buff *skb, int strict)
c71099ac 1264{
4c9483b2
DM		 1265 struct flowi6 fl6 = {
1266 .flowi6_oif = oif,
1267 .daddr = *daddr,
c71099ac
TG		 1268 };
1269 struct dst_entry *dst;
77d16f45 1270 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
c71099ac 1271
adaa70bb 1272 if (saddr) {
4c9483b2 1273 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
adaa70bb
TG		 1274 flags |= RT6_LOOKUP_F_HAS_SADDR;
1275 }
1276
b75cc8f9 1277 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
c71099ac
TG		 1278 if (dst->error == 0)
1279 return (struct rt6_info *) dst;
1280
1281 dst_release(dst);
1282
1da177e4
LT		 1283 return NULL;
1284}
7159039a
YH		 1285EXPORT_SYMBOL(rt6_lookup);
1286
c71099ac 1287/* ip6_ins_rt is called with FREE table->tb6_lock.
1cfb71ee
WW		 1288 * It takes new route entry, the addition fails by any reason the
1289 * route is released.
1290 * Caller must hold dst before calling it.
1da177e4
LT		 1291 */
1292
8d1c802b 1293static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
333c4301 1294 struct netlink_ext_ack *extack)
1da177e4
LT		 1295{
1296 int err;
c71099ac 1297 struct fib6_table *table;
1da177e4 1298
93c2fb25 1299 table = rt->fib6_table;
66f5d6ce 1300 spin_lock_bh(&table->tb6_lock);
d4ead6b3 1301 err = fib6_add(&table->tb6_root, rt, info, extack);
66f5d6ce 1302 spin_unlock_bh(&table->tb6_lock);
1da177e4
LT		 1303
1304 return err;
1305}
1306
8d1c802b 1307int ip6_ins_rt(struct net *net, struct fib6_info *rt)
40e22e8f 1308{
afb1d4b5 1309 struct nl_info info = { .nl_net = net, };
e715b6d3 1310
d4ead6b3 1311 return __ip6_ins_rt(rt, &info, NULL);
40e22e8f
TG		 1312}
1313
85bd05de 1314static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
8b9df265
MKL		 1315 const struct in6_addr *daddr,
1316 const struct in6_addr *saddr)
1da177e4 1317{
85bd05de 1318 struct fib6_info *f6i = res->f6i;
4832c30d 1319 struct net_device *dev;
1da177e4
LT		 1320 struct rt6_info *rt;
1321
1322 /*
1323 * Clone the route.
1324 */
1325
85bd05de 1326 if (!fib6_info_hold_safe(f6i))
e873e4b9
WW		 1327 return NULL;
1328
0d161581 1329 dev = ip6_rt_get_dev_rcu(res);
93531c67 1330 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
e873e4b9 1331 if (!rt) {
85bd05de 1332 fib6_info_release(f6i);
83a09abd 1333 return NULL;
e873e4b9 1334 }
83a09abd 1335
0d161581 1336 ip6_rt_copy_init(rt, res);
83a09abd 1337 rt->rt6i_flags |= RTF_CACHE;
83a09abd
MKL		 1338 rt->rt6i_dst.addr = *daddr;
1339 rt->rt6i_dst.plen = 128;
1da177e4 1340
85bd05de
DA		 1341 if (!rt6_is_gw_or_nonexthop(res)) {
1342 if (f6i->fib6_dst.plen != 128 &&
1343 ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
83a09abd 1344 rt->rt6i_flags |= RTF_ANYCAST;
1da177e4 1345#ifdef CONFIG_IPV6_SUBTREES
83a09abd
MKL		 1346 if (rt->rt6i_src.plen && saddr) {
1347 rt->rt6i_src.addr = *saddr;
1348 rt->rt6i_src.plen = 128;
8b9df265 1349 }
83a09abd 1350#endif
95a9a5ba 1351 }
1da177e4 1352
95a9a5ba
YH		 1353 return rt;
1354}
1da177e4 1355
db3fedee 1356static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
d52d3997 1357{
db3fedee
DA		 1358 struct fib6_info *f6i = res->f6i;
1359 unsigned short flags = fib6_info_dst_flags(f6i);
4832c30d 1360 struct net_device *dev;
d52d3997
MKL		 1361 struct rt6_info *pcpu_rt;
1362
db3fedee 1363 if (!fib6_info_hold_safe(f6i))
e873e4b9
WW		 1364 return NULL;
1365
4832c30d 1366 rcu_read_lock();
0d161581 1367 dev = ip6_rt_get_dev_rcu(res);
d8882935 1368 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT);
4832c30d 1369 rcu_read_unlock();
e873e4b9 1370 if (!pcpu_rt) {
db3fedee 1371 fib6_info_release(f6i);
d52d3997 1372 return NULL;
e873e4b9 1373 }
0d161581 1374 ip6_rt_copy_init(pcpu_rt, res);
d52d3997 1375 pcpu_rt->rt6i_flags |= RTF_PCPU;
8f34e53b
DA		 1376
1377 if (f6i->nh)
1378 pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1379
d52d3997
MKL		 1380 return pcpu_rt;
1381}
1382
8f34e53b
DA		 1383static bool rt6_is_valid(const struct rt6_info *rt6)
1384{
1385 return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1386}
1387
66f5d6ce 1388/* It should be called with rcu_read_lock() acquired */
db3fedee 1389static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
d52d3997 1390{
c353071a 1391 struct rt6_info *pcpu_rt;
d52d3997 1392
c353071a 1393 pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
d52d3997 1394
8f34e53b
DA		 1395 if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1396 struct rt6_info *prev, **p;
1397
1398 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1399 prev = xchg(p, NULL);
1400 if (prev) {
1401 dst_dev_put(&prev->dst);
1402 dst_release(&prev->dst);
1403 }
1404
1405 pcpu_rt = NULL;
1406 }
1407
a73e4195
MKL		 1408 return pcpu_rt;
1409}
1410
afb1d4b5 1411static struct rt6_info *rt6_make_pcpu_route(struct net *net,
db3fedee 1412 const struct fib6_result *res)
a73e4195
MKL		 1413{
1414 struct rt6_info *pcpu_rt, *prev, **p;
d52d3997 1415
db3fedee 1416 pcpu_rt = ip6_rt_pcpu_alloc(res);
0e09edcc
WW		 1417 if (!pcpu_rt)
1418 return NULL;
d52d3997 1419
f40b6ae2 1420 p = this_cpu_ptr(res->nh->rt6i_pcpu);
a94b9367 1421 prev = cmpxchg(p, NULL, pcpu_rt);
951f788a 1422 BUG_ON(prev);
a94b9367 1423
61fb0d01
ED		 1424 if (res->f6i->fib6_destroying) {
1425 struct fib6_info *from;
1426
1427 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1428 fib6_info_release(from);
1429 }
1430
d52d3997
MKL		 1431 return pcpu_rt;
1432}
1433
35732d01
WW		 1434/* exception hash table implementation
1435 */
1436static DEFINE_SPINLOCK(rt6_exception_lock);
1437
1438/* Remove rt6_ex from hash table and free the memory
1439 * Caller must hold rt6_exception_lock
1440 */
1441static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1442 struct rt6_exception *rt6_ex)
1443{
f5b51fe8 1444 struct fib6_info *from;
b2427e67 1445 struct net *net;
81eb8447 1446
35732d01
WW		 1447 if (!bucket || !rt6_ex)
1448 return;
b2427e67
CIK		 1449
1450 net = dev_net(rt6_ex->rt6i->dst.dev);
f5b51fe8
PA		 1451 net->ipv6.rt6_stats->fib_rt_cache--;
1452
1453 /* purge completely the exception to allow releasing the held resources:
1454 * some [sk] cache may keep the dst around for unlimited time
1455 */
0e233874 1456 from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
f5b51fe8
PA		 1457 fib6_info_release(from);
1458 dst_dev_put(&rt6_ex->rt6i->dst);
1459
35732d01 1460 hlist_del_rcu(&rt6_ex->hlist);
77634cc6 1461 dst_release(&rt6_ex->rt6i->dst);
35732d01
WW		 1462 kfree_rcu(rt6_ex, rcu);
1463 WARN_ON_ONCE(!bucket->depth);
1464 bucket->depth--;
1465}
1466
1467/* Remove oldest rt6_ex in bucket and free the memory
1468 * Caller must hold rt6_exception_lock
1469 */
1470static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1471{
1472 struct rt6_exception *rt6_ex, *oldest = NULL;
1473
1474 if (!bucket)
1475 return;
1476
1477 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1478 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1479 oldest = rt6_ex;
1480 }
1481 rt6_remove_exception(bucket, oldest);
1482}
1483
1484static u32 rt6_exception_hash(const struct in6_addr *dst,
1485 const struct in6_addr *src)
1486{
1487 static u32 seed __read_mostly;
1488 u32 val;
1489
1490 net_get_random_once(&seed, sizeof(seed));
b6b556af 1491 val = jhash2((const u32 *)dst, sizeof(*dst)/sizeof(u32), seed);
35732d01
WW		 1492
1493#ifdef CONFIG_IPV6_SUBTREES
1494 if (src)
b6b556af 1495 val = jhash2((const u32 *)src, sizeof(*src)/sizeof(u32), val);
35732d01
WW		 1496#endif
1497 return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1498}
1499
1500/* Helper function to find the cached rt in the hash table
1501 * and update bucket pointer to point to the bucket for this
1502 * (daddr, saddr) pair
1503 * Caller must hold rt6_exception_lock
1504 */
1505static struct rt6_exception *
1506__rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1507 const struct in6_addr *daddr,
1508 const struct in6_addr *saddr)
1509{
1510 struct rt6_exception *rt6_ex;
1511 u32 hval;
1512
1513 if (!(*bucket) || !daddr)
1514 return NULL;
1515
1516 hval = rt6_exception_hash(daddr, saddr);
1517 *bucket += hval;
1518
1519 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1520 struct rt6_info *rt6 = rt6_ex->rt6i;
1521 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1522
1523#ifdef CONFIG_IPV6_SUBTREES
1524 if (matched && saddr)
1525 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1526#endif
1527 if (matched)
1528 return rt6_ex;
1529 }
1530 return NULL;
1531}
1532
1533/* Helper function to find the cached rt in the hash table
1534 * and update bucket pointer to point to the bucket for this
1535 * (daddr, saddr) pair
1536 * Caller must hold rcu_read_lock()
1537 */
1538static struct rt6_exception *
1539__rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1540 const struct in6_addr *daddr,
1541 const struct in6_addr *saddr)
1542{
1543 struct rt6_exception *rt6_ex;
1544 u32 hval;
1545
1546 WARN_ON_ONCE(!rcu_read_lock_held());
1547
1548 if (!(*bucket) || !daddr)
1549 return NULL;
1550
1551 hval = rt6_exception_hash(daddr, saddr);
1552 *bucket += hval;
1553
1554 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1555 struct rt6_info *rt6 = rt6_ex->rt6i;
1556 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1557
1558#ifdef CONFIG_IPV6_SUBTREES
1559 if (matched && saddr)
1560 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1561#endif
1562 if (matched)
1563 return rt6_ex;
1564 }
1565 return NULL;
1566}
1567
b748f260 1568static unsigned int fib6_mtu(const struct fib6_result *res)
d4ead6b3 1569{
b748f260 1570 const struct fib6_nh *nh = res->nh;
d4ead6b3
DA		 1571 unsigned int mtu;
1572
b748f260
DA		 1573 if (res->f6i->fib6_pmtu) {
1574 mtu = res->f6i->fib6_pmtu;
dcd1f572 1575 } else {
b748f260 1576 struct net_device *dev = nh->fib_nh_dev;
dcd1f572
DA		 1577 struct inet6_dev *idev;
1578
1579 rcu_read_lock();
1580 idev = __in6_dev_get(dev);
1581 mtu = idev->cnf.mtu6;
1582 rcu_read_unlock();
1583 }
1584
d4ead6b3
DA		 1585 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1586
b748f260 1587 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
d4ead6b3
DA		 1588}
1589
cc5c073a
DA		 1590#define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1591
1592/* used when the flushed bit is not relevant, only access to the bucket
1593 * (ie., all bucket users except rt6_insert_exception);
1594 *
1595 * called under rcu lock; sometimes called with rt6_exception_lock held
1596 */
1597static
1598struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1599 spinlock_t *lock)
1600{
1601 struct rt6_exception_bucket *bucket;
1602
1603 if (lock)
1604 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1605 lockdep_is_held(lock));
1606 else
1607 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1608
1609 /* remove bucket flushed bit if set */
1610 if (bucket) {
1611 unsigned long p = (unsigned long)bucket;
1612
1613 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1614 bucket = (struct rt6_exception_bucket *)p;
1615 }
1616
1617 return bucket;
1618}
1619
1620static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1621{
1622 unsigned long p = (unsigned long)bucket;
1623
1624 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1625}
1626
1627/* called with rt6_exception_lock held */
1628static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1629 spinlock_t *lock)
1630{
1631 struct rt6_exception_bucket *bucket;
1632 unsigned long p;
1633
1634 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1635 lockdep_is_held(lock));
1636
1637 p = (unsigned long)bucket;
1638 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1639 bucket = (struct rt6_exception_bucket *)p;
1640 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1641}
1642
35732d01 1643static int rt6_insert_exception(struct rt6_info *nrt,
5012f0a5 1644 const struct fib6_result *res)
35732d01 1645{
5e670d84 1646 struct net *net = dev_net(nrt->dst.dev);
35732d01 1647 struct rt6_exception_bucket *bucket;
cc5c073a 1648 struct fib6_info *f6i = res->f6i;
35732d01
WW		 1649 struct in6_addr *src_key = NULL;
1650 struct rt6_exception *rt6_ex;
cc5c073a 1651 struct fib6_nh *nh = res->nh;
35732d01
WW		 1652 int err = 0;
1653
35732d01
WW		 1654 spin_lock_bh(&rt6_exception_lock);
1655
cc5c073a
DA		 1656 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1657 lockdep_is_held(&rt6_exception_lock));
35732d01
WW		 1658 if (!bucket) {
1659 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1660 GFP_ATOMIC);
1661 if (!bucket) {
1662 err = -ENOMEM;
1663 goto out;
1664 }
cc5c073a
DA		 1665 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1666 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1667 err = -EINVAL;
1668 goto out;
35732d01
WW		 1669 }
1670
1671#ifdef CONFIG_IPV6_SUBTREES
5012f0a5 1672 /* fib6_src.plen != 0 indicates f6i is in subtree
35732d01 1673 * and exception table is indexed by a hash of
5012f0a5 1674 * both fib6_dst and fib6_src.
35732d01 1675 * Otherwise, the exception table is indexed by
5012f0a5 1676 * a hash of only fib6_dst.
35732d01 1677 */
5012f0a5 1678 if (f6i->fib6_src.plen)
35732d01
WW		 1679 src_key = &nrt->rt6i_src.addr;
1680#endif
5012f0a5 1681 /* rt6_mtu_change() might lower mtu on f6i.
f5bbe7ee 1682 * Only insert this exception route if its mtu
5012f0a5 1683 * is less than f6i's mtu value.
f5bbe7ee 1684 */
b748f260 1685 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
f5bbe7ee
WW		 1686 err = -EINVAL;
1687 goto out;
1688 }
60006a48 1689
35732d01
WW		 1690 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1691 src_key);
1692 if (rt6_ex)
1693 rt6_remove_exception(bucket, rt6_ex);
1694
1695 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1696 if (!rt6_ex) {
1697 err = -ENOMEM;
1698 goto out;
1699 }
1700 rt6_ex->rt6i = nrt;
1701 rt6_ex->stamp = jiffies;
35732d01
WW		 1702 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1703 bucket->depth++;
81eb8447 1704 net->ipv6.rt6_stats->fib_rt_cache++;
35732d01
WW		 1705
1706 if (bucket->depth > FIB6_MAX_DEPTH)
1707 rt6_exception_remove_oldest(bucket);
1708
1709out:
1710 spin_unlock_bh(&rt6_exception_lock);
1711
1712 /* Update fn->fn_sernum to invalidate all cached dst */
b886d5f2 1713 if (!err) {
5012f0a5
DA		 1714 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1715 fib6_update_sernum(net, f6i);
1716 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
b886d5f2
PA		 1717 fib6_force_start_gc(net);
1718 }
35732d01
WW		 1719
1720 return err;
1721}
1722
c0b220cf 1723static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
35732d01
WW		 1724{
1725 struct rt6_exception_bucket *bucket;
1726 struct rt6_exception *rt6_ex;
1727 struct hlist_node *tmp;
1728 int i;
1729
1730 spin_lock_bh(&rt6_exception_lock);
35732d01 1731
cc5c073a 1732 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
35732d01
WW		 1733 if (!bucket)
1734 goto out;
1735
cc5c073a
DA		 1736 /* Prevent rt6_insert_exception() to recreate the bucket list */
1737 if (!from)
1738 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1739
35732d01 1740 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
cc5c073a
DA		 1741 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1742 if (!from ||
1743 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1744 rt6_remove_exception(bucket, rt6_ex);
1745 }
1746 WARN_ON_ONCE(!from && bucket->depth);
35732d01
WW		 1747 bucket++;
1748 }
35732d01
WW		 1749out:
1750 spin_unlock_bh(&rt6_exception_lock);
1751}
1752
e659ba31
DA		 1753static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1754{
1755 struct fib6_info *f6i = arg;
1756
1757 fib6_nh_flush_exceptions(nh, f6i);
1758
1759 return 0;
1760}
1761
c0b220cf
DA		 1762void rt6_flush_exceptions(struct fib6_info *f6i)
1763{
e659ba31
DA		 1764 if (f6i->nh)
1765 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1766 f6i);
1767 else
1768 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
c0b220cf
DA		 1769}
1770
35732d01
WW		 1771/* Find cached rt in the hash table inside passed in rt
1772 * Caller has to hold rcu_read_lock()
1773 */
7e4b5128 1774static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
510e2ced
WW		 1775 const struct in6_addr *daddr,
1776 const struct in6_addr *saddr)
35732d01 1777{
510e2ced 1778 const struct in6_addr *src_key = NULL;
35732d01 1779 struct rt6_exception_bucket *bucket;
35732d01 1780 struct rt6_exception *rt6_ex;
7e4b5128 1781 struct rt6_info *ret = NULL;
35732d01 1782
35732d01 1783#ifdef CONFIG_IPV6_SUBTREES
7e4b5128 1784 /* fib6i_src.plen != 0 indicates f6i is in subtree
35732d01 1785 * and exception table is indexed by a hash of
7e4b5128 1786 * both fib6_dst and fib6_src.
510e2ced
WW		 1787 * However, the src addr used to create the hash
1788 * might not be exactly the passed in saddr which
1789 * is a /128 addr from the flow.
1790 * So we need to use f6i->fib6_src to redo lookup
1791 * if the passed in saddr does not find anything.
1792 * (See the logic in ip6_rt_cache_alloc() on how
1793 * rt->rt6i_src is updated.)
35732d01 1794 */
7e4b5128 1795 if (res->f6i->fib6_src.plen)
35732d01 1796 src_key = saddr;
510e2ced 1797find_ex:
35732d01 1798#endif
cc5c073a 1799 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
35732d01
WW		 1800 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1801
1802 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
7e4b5128 1803 ret = rt6_ex->rt6i;
35732d01 1804
510e2ced
WW		 1805#ifdef CONFIG_IPV6_SUBTREES
1806 /* Use fib6_src as src_key and redo lookup */
1807 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1808 src_key = &res->f6i->fib6_src.addr;
1809 goto find_ex;
1810 }
1811#endif
1812
7e4b5128 1813 return ret;
35732d01
WW		 1814}
1815
1816/* Remove the passed in cached rt from the hash table that contains it */
cc5c073a 1817static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1818 const struct rt6_info *rt)
35732d01 1819{
c0b220cf 1820 const struct in6_addr *src_key = NULL;
35732d01 1821 struct rt6_exception_bucket *bucket;
35732d01
WW		 1822 struct rt6_exception *rt6_ex;
1823 int err;
1824
cc5c073a 1825 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
35732d01
WW		 1826 return -ENOENT;
1827
1828 spin_lock_bh(&rt6_exception_lock);
cc5c073a
DA		 1829 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1830
35732d01 1831#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1832 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1833 * and exception table is indexed by a hash of
1834 * both rt6i_dst and rt6i_src.
35732d01
WW		 1835 * Otherwise, the exception table is indexed by
1836 * a hash of only rt6i_dst.
1837 */
c0b220cf 1838 if (plen)
35732d01
WW		 1839 src_key = &rt->rt6i_src.addr;
1840#endif
1841 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1842 &rt->rt6i_dst.addr,
1843 src_key);
1844 if (rt6_ex) {
1845 rt6_remove_exception(bucket, rt6_ex);
1846 err = 0;
1847 } else {
1848 err = -ENOENT;
1849 }
1850
1851 spin_unlock_bh(&rt6_exception_lock);
1852 return err;
1853}
1854
e659ba31
DA		 1855struct fib6_nh_excptn_arg {
1856 struct rt6_info *rt;
1857 int plen;
1858};
1859
1860static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1861{
1862 struct fib6_nh_excptn_arg *arg = _arg;
1863 int err;
1864
1865 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1866 if (err == 0)
1867 return 1;
1868
1869 return 0;
1870}
1871
c0b220cf
DA		 1872static int rt6_remove_exception_rt(struct rt6_info *rt)
1873{
1874 struct fib6_info *from;
1875
1876 from = rcu_dereference(rt->from);
cc5c073a 1877 if (!from || !(rt->rt6i_flags & RTF_CACHE))
c0b220cf
DA		 1878 return -EINVAL;
1879
e659ba31
DA		 1880 if (from->nh) {
1881 struct fib6_nh_excptn_arg arg = {
1882 .rt = rt,
1883 .plen = from->fib6_src.plen
1884 };
1885 int rc;
1886
1887 /* rc = 1 means an entry was found */
1888 rc = nexthop_for_each_fib6_nh(from->nh,
1889 rt6_nh_remove_exception_rt,
1890 &arg);
1891 return rc ? 0 : -ENOENT;
1892 }
1893
1cf844c7 1894 return fib6_nh_remove_exception(from->fib6_nh,
cc5c073a 1895 from->fib6_src.plen, rt);
c0b220cf
DA		 1896}
1897
35732d01
WW		 1898/* Find rt6_ex which contains the passed in rt cache and
1899 * refresh its stamp
1900 */
cc5c073a 1901static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1902 const struct rt6_info *rt)
35732d01 1903{
c0b220cf 1904 const struct in6_addr *src_key = NULL;
35732d01 1905 struct rt6_exception_bucket *bucket;
35732d01 1906 struct rt6_exception *rt6_ex;
193f3685 1907
cc5c073a 1908 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
35732d01 1909#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1910 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1911 * and exception table is indexed by a hash of
1912 * both rt6i_dst and rt6i_src.
35732d01
WW		 1913 * Otherwise, the exception table is indexed by
1914 * a hash of only rt6i_dst.
1915 */
c0b220cf 1916 if (plen)
35732d01
WW		 1917 src_key = &rt->rt6i_src.addr;
1918#endif
cc5c073a 1919 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
35732d01
WW		 1920 if (rt6_ex)
1921 rt6_ex->stamp = jiffies;
c0b220cf
DA		 1922}
1923
e659ba31
DA		 1924struct fib6_nh_match_arg {
1925 const struct net_device *dev;
1926 const struct in6_addr *gw;
1927 struct fib6_nh *match;
1928};
1929
1930/* determine if fib6_nh has given device and gateway */
1931static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1932{
1933 struct fib6_nh_match_arg *arg = _arg;
1934
1935 if (arg->dev != nh->fib_nh_dev ||
1936 (arg->gw && !nh->fib_nh_gw_family) ||
1937 (!arg->gw && nh->fib_nh_gw_family) ||
1938 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1939 return 0;
1940
1941 arg->match = nh;
1942
1943 /* found a match, break the loop */
1944 return 1;
1945}
1946
c0b220cf
DA		 1947static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1948{
1949 struct fib6_info *from;
e659ba31 1950 struct fib6_nh *fib6_nh;
35732d01 1951
c0b220cf
DA		 1952 rcu_read_lock();
1953
1954 from = rcu_dereference(rt->from);
1955 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1956 goto unlock;
1957
e659ba31
DA		 1958 if (from->nh) {
1959 struct fib6_nh_match_arg arg = {
1960 .dev = rt->dst.dev,
1961 .gw = &rt->rt6i_gateway,
1962 };
1963
1964 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1965
1966 if (!arg.match)
cff6a327 1967 goto unlock;
e659ba31
DA		 1968 fib6_nh = arg.match;
1969 } else {
1970 fib6_nh = from->fib6_nh;
1971 }
1972 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
193f3685 1973unlock:
35732d01
WW		 1974 rcu_read_unlock();
1975}
1976
e9fa1495
SB		 1977static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1978 struct rt6_info *rt, int mtu)
1979{
1980 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1981 * lowest MTU in the path: always allow updating the route PMTU to
1982 * reflect PMTU decreases.
1983 *
1984 * If the new MTU is higher, and the route PMTU is equal to the local
1985 * MTU, this means the old MTU is the lowest in the path, so allow
1986 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1987 * handle this.
1988 */
1989
1990 if (dst_mtu(&rt->dst) >= mtu)
1991 return true;
1992
1993 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
1994 return true;
1995
1996 return false;
1997}
1998
1999static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
cc5c073a 2000 const struct fib6_nh *nh, int mtu)
f5bbe7ee
WW		 2001{
2002 struct rt6_exception_bucket *bucket;
2003 struct rt6_exception *rt6_ex;
2004 int i;
2005
cc5c073a 2006 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
e9fa1495
SB		 2007 if (!bucket)
2008 return;
2009
2010 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2011 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2012 struct rt6_info *entry = rt6_ex->rt6i;
2013
2014 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
d4ead6b3 2015 * route), the metrics of its rt->from have already
e9fa1495
SB		 2016 * been updated.
2017 */
d4ead6b3 2018 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
e9fa1495 2019 rt6_mtu_change_route_allowed(idev, entry, mtu))
d4ead6b3 2020 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
f5bbe7ee 2021 }
e9fa1495 2022 bucket++;
f5bbe7ee
WW		 2023 }
2024}
2025
b16cb459
WW		 2026#define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2027
cc5c073a
DA		 2028static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2029 const struct in6_addr *gateway)
b16cb459
WW		 2030{
2031 struct rt6_exception_bucket *bucket;
2032 struct rt6_exception *rt6_ex;
2033 struct hlist_node *tmp;
2034 int i;
2035
cc5c073a 2036 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
b16cb459
WW		 2037 return;
2038
2039 spin_lock_bh(&rt6_exception_lock);
cc5c073a 2040 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
b16cb459
WW		 2041 if (bucket) {
2042 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2043 hlist_for_each_entry_safe(rt6_ex, tmp,
2044 &bucket->chain, hlist) {
2045 struct rt6_info *entry = rt6_ex->rt6i;
2046
2047 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2048 RTF_CACHE_GATEWAY &&
2049 ipv6_addr_equal(gateway,
2050 &entry->rt6i_gateway)) {
2051 rt6_remove_exception(bucket, rt6_ex);
2052 }
2053 }
2054 bucket++;
2055 }
2056 }
2057
2058 spin_unlock_bh(&rt6_exception_lock);
2059}
2060
c757faa8
WW		 2061static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2062 struct rt6_exception *rt6_ex,
2063 struct fib6_gc_args *gc_args,
2064 unsigned long now)
2065{
2066 struct rt6_info *rt = rt6_ex->rt6i;
2067
1859bac0
PA		 2068 /* we are pruning and obsoleting aged-out and non gateway exceptions
2069 * even if others have still references to them, so that on next
2070 * dst_check() such references can be dropped.
2071 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2072 * expired, independently from their aging, as per RFC 8201 section 4
2073 */
31afeb42
WW		 2074 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2075 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2076 RT6_TRACE("aging clone %p\n", rt);
2077 rt6_remove_exception(bucket, rt6_ex);
2078 return;
2079 }
2080 } else if (time_after(jiffies, rt->dst.expires)) {
2081 RT6_TRACE("purging expired route %p\n", rt);
c757faa8
WW		 2082 rt6_remove_exception(bucket, rt6_ex);
2083 return;
31afeb42
WW		 2084 }
2085
2086 if (rt->rt6i_flags & RTF_GATEWAY) {
c757faa8 2087 struct neighbour *neigh;
c757faa8 2088
1bfa26ff 2089 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
1bfa26ff 2090
b7a320c3 2091 if (!(neigh && (neigh->flags & NTF_ROUTER))) {
c757faa8
WW		 2092 RT6_TRACE("purging route %p via non-router but gateway\n",
2093 rt);
2094 rt6_remove_exception(bucket, rt6_ex);
2095 return;
2096 }
2097 }
31afeb42 2098
c757faa8
WW		 2099 gc_args->more++;
2100}
2101
cc5c073a 2102static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
c0b220cf
DA		 2103 struct fib6_gc_args *gc_args,
2104 unsigned long now)
c757faa8
WW		 2105{
2106 struct rt6_exception_bucket *bucket;
2107 struct rt6_exception *rt6_ex;
2108 struct hlist_node *tmp;
2109 int i;
2110
cc5c073a 2111 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
c757faa8
WW		 2112 return;
2113
1bfa26ff
ED		 2114 rcu_read_lock_bh();
2115 spin_lock(&rt6_exception_lock);
cc5c073a 2116 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
c757faa8
WW		 2117 if (bucket) {
2118 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2119 hlist_for_each_entry_safe(rt6_ex, tmp,
2120 &bucket->chain, hlist) {
2121 rt6_age_examine_exception(bucket, rt6_ex,
2122 gc_args, now);
2123 }
2124 bucket++;
2125 }
2126 }
1bfa26ff
ED		 2127 spin_unlock(&rt6_exception_lock);
2128 rcu_read_unlock_bh();
c757faa8
WW		 2129}
2130
e659ba31
DA		 2131struct fib6_nh_age_excptn_arg {
2132 struct fib6_gc_args *gc_args;
2133 unsigned long now;
2134};
2135
2136static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2137{
2138 struct fib6_nh_age_excptn_arg *arg = _arg;
2139
2140 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2141 return 0;
2142}
2143
cc5c073a 2144void rt6_age_exceptions(struct fib6_info *f6i,
c0b220cf
DA		 2145 struct fib6_gc_args *gc_args,
2146 unsigned long now)
2147{
e659ba31
DA		 2148 if (f6i->nh) {
2149 struct fib6_nh_age_excptn_arg arg = {
2150 .gc_args = gc_args,
2151 .now = now
2152 };
2153
2154 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2155 &arg);
2156 } else {
2157 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2158 }
c0b220cf
DA		 2159}
2160
1d053da9 2161/* must be called with rcu lock held */
effda4dd
DA		 2162int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2163 struct flowi6 *fl6, struct fib6_result *res, int strict)
1da177e4 2164{
367efcb9 2165 struct fib6_node *fn, *saved_fn;
1da177e4 2166
6454743b 2167 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
367efcb9 2168 saved_fn = fn;
1da177e4 2169
ca254490
DA		 2170 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2171 oif = 0;
2172
a3c00e46 2173redo_rt6_select:
effda4dd
DA		 2174 rt6_select(net, fn, oif, res, strict);
2175 if (res->f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2176 fn = fib6_backtrack(fn, &fl6->saddr);
2177 if (fn)
2178 goto redo_rt6_select;
367efcb9
MKL		 2179 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2180 /* also consider unreachable route */
2181 strict &= ~RT6_LOOKUP_F_REACHABLE;
2182 fn = saved_fn;
2183 goto redo_rt6_select;
367efcb9 2184 }
a3c00e46
MKL		 2185 }
2186
effda4dd 2187 trace_fib6_table_lookup(net, res, table, fl6);
fb9de91e 2188
effda4dd 2189 return 0;
1d053da9
DA		 2190}
2191
2192struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2193 int oif, struct flowi6 *fl6,
2194 const struct sk_buff *skb, int flags)
2195{
b1d40991 2196 struct fib6_result res = {};
0e09edcc 2197 struct rt6_info *rt = NULL;
1d053da9
DA		 2198 int strict = 0;
2199
0e09edcc
WW		 2200 WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2201 !rcu_read_lock_held());
2202
1d053da9
DA		 2203 strict |= flags & RT6_LOOKUP_F_IFACE;
2204 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2205 if (net->ipv6.devconf_all->forwarding == 0)
2206 strict |= RT6_LOOKUP_F_REACHABLE;
2207
2208 rcu_read_lock();
2209
effda4dd 2210 fib6_table_lookup(net, table, oif, fl6, &res, strict);
0e09edcc
WW		 2211 if (res.f6i == net->ipv6.fib6_null_entry)
2212 goto out;
23fb93a4 2213
b1d40991 2214 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
d83009d4 2215
23fb93a4 2216 /*Search through exception table */
7e4b5128 2217 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
23fb93a4 2218 if (rt) {
0e09edcc 2219 goto out;
3da59bd9 2220 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
b1d40991 2221 !res.nh->fib_nh_gw_family)) {
3da59bd9
MKL		 2222 /* Create a RTF_CACHE clone which will not be
2223 * owned by the fib6 tree. It is for the special case where
2224 * the daddr in the skb during the neighbor look-up is different
2225 * from the fl6->daddr used to look-up route here.
2226 */
0e09edcc 2227 rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
c71099ac 2228
0e09edcc
WW		 2229 if (rt) {
2230 /* 1 refcnt is taken during ip6_rt_cache_alloc().
2231 * As rt6_uncached_list_add() does not consume refcnt,
2232 * this refcnt is always returned to the caller even
2233 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
1cfb71ee 2234 */
0e09edcc 2235 rt6_uncached_list_add(rt);
81eb8447 2236 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
0e09edcc 2237 rcu_read_unlock();
b811580d 2238
0e09edcc
WW		 2239 return rt;
2240 }
d52d3997
MKL		 2241 } else {
2242 /* Get a percpu copy */
951f788a 2243 local_bh_disable();
0e09edcc 2244 rt = rt6_get_pcpu_route(&res);
d52d3997 2245
0e09edcc
WW		 2246 if (!rt)
2247 rt = rt6_make_pcpu_route(net, &res);
93531c67 2248
951f788a 2249 local_bh_enable();
d52d3997 2250 }
0e09edcc
WW		 2251out:
2252 if (!rt)
2253 rt = net->ipv6.ip6_null_entry;
2254 if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2255 ip6_hold_safe(net, &rt);
2256 rcu_read_unlock();
2257
2258 return rt;
1da177e4 2259}
9ff74384 2260EXPORT_SYMBOL_GPL(ip6_pol_route);
1da177e4 2261
55cced4f 2262INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_input(struct net *net,
b75cc8f9
DA		 2263 struct fib6_table *table,
2264 struct flowi6 *fl6,
2265 const struct sk_buff *skb,
2266 int flags)
4acad72d 2267{
b75cc8f9 2268 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
4acad72d
PE		 2269}
2270
d409b847
MB		 2271struct dst_entry *ip6_route_input_lookup(struct net *net,
2272 struct net_device *dev,
b75cc8f9
DA		 2273 struct flowi6 *fl6,
2274 const struct sk_buff *skb,
2275 int flags)
72331bc0
SL		 2276{
2277 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2278 flags |= RT6_LOOKUP_F_IFACE;
2279
b75cc8f9 2280 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
72331bc0 2281}
d409b847 2282EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
72331bc0 2283
23aebdac 2284static void ip6_multipath_l3_keys(const struct sk_buff *skb,
5e5d6fed
RP		 2285 struct flow_keys *keys,
2286 struct flow_keys *flkeys)
23aebdac
JS		 2287{
2288 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2289 const struct ipv6hdr *key_iph = outer_iph;
5e5d6fed 2290 struct flow_keys *_flkeys = flkeys;
23aebdac
JS		 2291 const struct ipv6hdr *inner_iph;
2292 const struct icmp6hdr *icmph;
2293 struct ipv6hdr _inner_iph;
cea67a2d 2294 struct icmp6hdr _icmph;
23aebdac
JS		 2295
2296 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2297 goto out;
2298
cea67a2d
ED		 2299 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2300 sizeof(_icmph), &_icmph);
2301 if (!icmph)
2302 goto out;
2303
54074f1d 2304 if (!icmpv6_is_err(icmph->icmp6_type))
23aebdac
JS		 2305 goto out;
2306
2307 inner_iph = skb_header_pointer(skb,
2308 skb_transport_offset(skb) + sizeof(*icmph),
2309 sizeof(_inner_iph), &_inner_iph);
2310 if (!inner_iph)
2311 goto out;
2312
2313 key_iph = inner_iph;
5e5d6fed 2314 _flkeys = NULL;
23aebdac 2315out:
5e5d6fed
RP		 2316 if (_flkeys) {
2317 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2318 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2319 keys->tags.flow_label = _flkeys->tags.flow_label;
2320 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2321 } else {
2322 keys->addrs.v6addrs.src = key_iph->saddr;
2323 keys->addrs.v6addrs.dst = key_iph->daddr;
fa1be7e0 2324 keys->tags.flow_label = ip6_flowlabel(key_iph);
5e5d6fed
RP		 2325 keys->basic.ip_proto = key_iph->nexthdr;
2326 }
23aebdac
JS		 2327}
2328
73c2c5cb
IS		 2329static u32 rt6_multipath_custom_hash_outer(const struct net *net,
2330 const struct sk_buff *skb,
2331 bool *p_has_inner)
2332{
2333 u32 hash_fields = ip6_multipath_hash_fields(net);
2334 struct flow_keys keys, hash_keys;
2335
2336 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2337 return 0;
2338
2339 memset(&hash_keys, 0, sizeof(hash_keys));
2340 skb_flow_dissect_flow_keys(skb, &keys, FLOW_DISSECTOR_F_STOP_AT_ENCAP);
2341
2342 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2343 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2344 hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2345 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2346 hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2347 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2348 hash_keys.basic.ip_proto = keys.basic.ip_proto;
2349 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2350 hash_keys.tags.flow_label = keys.tags.flow_label;
2351 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2352 hash_keys.ports.src = keys.ports.src;
2353 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2354 hash_keys.ports.dst = keys.ports.dst;
2355
2356 *p_has_inner = !!(keys.control.flags & FLOW_DIS_ENCAPSULATION);
2357 return flow_hash_from_keys(&hash_keys);
2358}
2359
2360static u32 rt6_multipath_custom_hash_inner(const struct net *net,
2361 const struct sk_buff *skb,
2362 bool has_inner)
2363{
2364 u32 hash_fields = ip6_multipath_hash_fields(net);
2365 struct flow_keys keys, hash_keys;
2366
2367 /* We assume the packet carries an encapsulation, but if none was
2368 * encountered during dissection of the outer flow, then there is no
2369 * point in calling the flow dissector again.
2370 */
2371 if (!has_inner)
2372 return 0;
2373
2374 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_MASK))
2375 return 0;
2376
2377 memset(&hash_keys, 0, sizeof(hash_keys));
2378 skb_flow_dissect_flow_keys(skb, &keys, 0);
2379
2380 if (!(keys.control.flags & FLOW_DIS_ENCAPSULATION))
2381 return 0;
2382
2383 if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2384 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2385 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2386 hash_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src;
2387 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2388 hash_keys.addrs.v4addrs.dst = keys.addrs.v4addrs.dst;
2389 } else if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2390 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2391 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2392 hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2393 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2394 hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2395 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL)
2396 hash_keys.tags.flow_label = keys.tags.flow_label;
2397 }
2398
2399 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO)
2400 hash_keys.basic.ip_proto = keys.basic.ip_proto;
2401 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT)
2402 hash_keys.ports.src = keys.ports.src;
2403 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT)
2404 hash_keys.ports.dst = keys.ports.dst;
2405
2406 return flow_hash_from_keys(&hash_keys);
2407}
2408
2409static u32 rt6_multipath_custom_hash_skb(const struct net *net,
2410 const struct sk_buff *skb)
2411{
2412 u32 mhash, mhash_inner;
2413 bool has_inner = true;
2414
2415 mhash = rt6_multipath_custom_hash_outer(net, skb, &has_inner);
2416 mhash_inner = rt6_multipath_custom_hash_inner(net, skb, has_inner);
2417
2418 return jhash_2words(mhash, mhash_inner, 0);
2419}
2420
2421static u32 rt6_multipath_custom_hash_fl6(const struct net *net,
2422 const struct flowi6 *fl6)
2423{
2424 u32 hash_fields = ip6_multipath_hash_fields(net);
2425 struct flow_keys hash_keys;
2426
2427 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2428 return 0;
2429
2430 memset(&hash_keys, 0, sizeof(hash_keys));
2431 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2432 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2433 hash_keys.addrs.v6addrs.src = fl6->saddr;
2434 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2435 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2436 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2437 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2438 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2439 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2440 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2441 hash_keys.ports.src = fl6->fl6_sport;
2442 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2443 hash_keys.ports.dst = fl6->fl6_dport;
2444
2445 return flow_hash_from_keys(&hash_keys);
2446}
2447
23aebdac 2448/* if skb is set it will be used and fl6 can be NULL */
b4bac172
DA		 2449u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2450 const struct sk_buff *skb, struct flow_keys *flkeys)
23aebdac
JS		 2451{
2452 struct flow_keys hash_keys;
b95b6e07 2453 u32 mhash = 0;
23aebdac 2454
bbfa047a 2455 switch (ip6_multipath_hash_policy(net)) {
b4bac172
DA		 2456 case 0:
2457 memset(&hash_keys, 0, sizeof(hash_keys));
2458 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2459 if (skb) {
2460 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2461 } else {
2462 hash_keys.addrs.v6addrs.src = fl6->saddr;
2463 hash_keys.addrs.v6addrs.dst = fl6->daddr;
fa1be7e0 2464 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
b4bac172
DA		 2465 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2466 }
b95b6e07 2467 mhash = flow_hash_from_keys(&hash_keys);
b4bac172
DA		 2468 break;
2469 case 1:
2470 if (skb) {
2471 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2472 struct flow_keys keys;
2473
2474 /* short-circuit if we already have L4 hash present */
2475 if (skb->l4_hash)
2476 return skb_get_hash_raw(skb) >> 1;
2477
2478 memset(&hash_keys, 0, sizeof(hash_keys));
2479
13fdb940 2480 if (!flkeys) {
b4bac172
DA		 2481 skb_flow_dissect_flow_keys(skb, &keys, flag);
2482 flkeys = &keys;
2483 }
2484 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2485 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2486 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2487 hash_keys.ports.src = flkeys->ports.src;
2488 hash_keys.ports.dst = flkeys->ports.dst;
2489 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2490 } else {
2491 memset(&hash_keys, 0, sizeof(hash_keys));
2492 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2493 hash_keys.addrs.v6addrs.src = fl6->saddr;
2494 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2495 hash_keys.ports.src = fl6->fl6_sport;
2496 hash_keys.ports.dst = fl6->fl6_dport;
2497 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2498 }
b95b6e07 2499 mhash = flow_hash_from_keys(&hash_keys);
b4bac172 2500 break;
d8f74f09
SS		 2501 case 2:
2502 memset(&hash_keys, 0, sizeof(hash_keys));
2503 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2504 if (skb) {
2505 struct flow_keys keys;
2506
2507 if (!flkeys) {
2508 skb_flow_dissect_flow_keys(skb, &keys, 0);
2509 flkeys = &keys;
2510 }
2511
2512 /* Inner can be v4 or v6 */
2513 if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2514 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2515 hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2516 hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2517 } else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2518 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2519 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2520 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2521 hash_keys.tags.flow_label = flkeys->tags.flow_label;
2522 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2523 } else {
2524 /* Same as case 0 */
2525 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2526 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2527 }
2528 } else {
2529 /* Same as case 0 */
2530 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2531 hash_keys.addrs.v6addrs.src = fl6->saddr;
2532 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2533 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2534 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2535 }
b95b6e07 2536 mhash = flow_hash_from_keys(&hash_keys);
d8f74f09 2537 break;
73c2c5cb
IS		 2538 case 3:
2539 if (skb)
2540 mhash = rt6_multipath_custom_hash_skb(net, skb);
2541 else
2542 mhash = rt6_multipath_custom_hash_fl6(net, fl6);
2543 break;
23aebdac
JS		 2544 }
2545
9a2a537a 2546 return mhash >> 1;
23aebdac
JS		 2547}
2548
67f415dd 2549/* Called with rcu held */
c71099ac
TG		 2550void ip6_route_input(struct sk_buff *skb)
2551{
b71d1d42 2552 const struct ipv6hdr *iph = ipv6_hdr(skb);
c346dca1 2553 struct net *net = dev_net(skb->dev);
67f415dd 2554 int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
904af04d 2555 struct ip_tunnel_info *tun_info;
4c9483b2 2556 struct flowi6 fl6 = {
e0d56fdd 2557 .flowi6_iif = skb->dev->ifindex,
4c9483b2
DM		 2558 .daddr = iph->daddr,
2559 .saddr = iph->saddr,
6502ca52 2560 .flowlabel = ip6_flowinfo(iph),
4c9483b2
DM		 2561 .flowi6_mark = skb->mark,
2562 .flowi6_proto = iph->nexthdr,
c71099ac 2563 };
5e5d6fed 2564 struct flow_keys *flkeys = NULL, _flkeys;
adaa70bb 2565
904af04d 2566 tun_info = skb_tunnel_info(skb);
46fa062a 2567 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
904af04d 2568 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
5e5d6fed
RP		 2569
2570 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2571 flkeys = &_flkeys;
2572
23aebdac 2573 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
b4bac172 2574 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
06e9d040 2575 skb_dst_drop(skb);
67f415dd
WW		 2576 skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2577 &fl6, skb, flags));
c71099ac
TG		 2578}
2579
55cced4f 2580INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_output(struct net *net,
b75cc8f9
DA		 2581 struct fib6_table *table,
2582 struct flowi6 *fl6,
2583 const struct sk_buff *skb,
2584 int flags)
1da177e4 2585{
b75cc8f9 2586 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
c71099ac
TG		 2587}
2588
7d9e5f42
WW		 2589struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2590 const struct sock *sk,
2591 struct flowi6 *fl6, int flags)
c71099ac 2592{
d46a9d67 2593 bool any_src;
c71099ac 2594
3ede0bbc
RS		 2595 if (ipv6_addr_type(&fl6->daddr) &
2596 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
4c1feac5
DA		 2597 struct dst_entry *dst;
2598
7d9e5f42 2599 /* This function does not take refcnt on the dst */
4c1feac5
DA		 2600 dst = l3mdev_link_scope_lookup(net, fl6);
2601 if (dst)
2602 return dst;
2603 }
ca254490 2604
1fb9489b 2605 fl6->flowi6_iif = LOOPBACK_IFINDEX;
4dc27d1c 2606
7d9e5f42 2607 flags |= RT6_LOOKUP_F_DST_NOREF;
d46a9d67 2608 any_src = ipv6_addr_any(&fl6->saddr);
741a11d9 2609 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
d46a9d67 2610 (fl6->flowi6_oif && any_src))
77d16f45 2611 flags |= RT6_LOOKUP_F_IFACE;
c71099ac 2612
d46a9d67 2613 if (!any_src)
adaa70bb 2614 flags |= RT6_LOOKUP_F_HAS_SADDR;
0c9a2ac1
YH		 2615 else if (sk)
2616 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
adaa70bb 2617
b75cc8f9 2618 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
1da177e4 2619}
7d9e5f42
WW		 2620EXPORT_SYMBOL_GPL(ip6_route_output_flags_noref);
2621
2622struct dst_entry *ip6_route_output_flags(struct net *net,
2623 const struct sock *sk,
2624 struct flowi6 *fl6,
2625 int flags)
2626{
13fdb940
SK		 2627 struct dst_entry *dst;
2628 struct rt6_info *rt6;
7d9e5f42 2629
13fdb940
SK		 2630 rcu_read_lock();
2631 dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2632 rt6 = (struct rt6_info *)dst;
2633 /* For dst cached in uncached_list, refcnt is already taken. */
2634 if (list_empty(&rt6->rt6i_uncached) && !dst_hold_safe(dst)) {
2635 dst = &net->ipv6.ip6_null_entry->dst;
2636 dst_hold(dst);
2637 }
2638 rcu_read_unlock();
7d9e5f42 2639
13fdb940 2640 return dst;
7d9e5f42 2641}
6f21c96a 2642EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1da177e4 2643
2774c131 2644struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
14e50e57 2645{
5c1e6aa3 2646 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1dbe3252 2647 struct net_device *loopback_dev = net->loopback_dev;
14e50e57
DM		 2648 struct dst_entry *new = NULL;
2649
1dbe3252 2650 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
62cf27e5 2651 DST_OBSOLETE_DEAD, 0);
14e50e57 2652 if (rt) {
0a1f5962 2653 rt6_info_init(rt);
81eb8447 2654 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
8104891b 2655
0a1f5962 2656 new = &rt->dst;
14e50e57 2657 new->__use = 1;
352e512c 2658 new->input = dst_discard;
ede2059d 2659 new->output = dst_discard_out;
14e50e57 2660
0a1f5962 2661 dst_copy_metrics(new, &ort->dst);
14e50e57 2662
1dbe3252 2663 rt->rt6i_idev = in6_dev_get(loopback_dev);
4e3fd7a0 2664 rt->rt6i_gateway = ort->rt6i_gateway;
0a1f5962 2665 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
14e50e57
DM		 2666
2667 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2668#ifdef CONFIG_IPV6_SUBTREES
2669 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2670#endif
14e50e57
DM		 2671 }
2672
69ead7af
DM		 2673 dst_release(dst_orig);
2674 return new ? new : ERR_PTR(-ENOMEM);
14e50e57 2675}
14e50e57 2676
1da177e4
LT		 2677/*
2678 * Destination cache support functions
2679 */
2680
8d1c802b 2681static bool fib6_check(struct fib6_info *f6i, u32 cookie)
4b32b5ad 2682{
93531c67
DA		 2683 u32 rt_cookie = 0;
2684
8ae86971 2685 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
93531c67
DA		 2686 return false;
2687
2688 if (fib6_check_expired(f6i))
2689 return false;
2690
2691 return true;
4b32b5ad
MKL		 2692}
2693
a68886a6
DA		 2694static struct dst_entry *rt6_check(struct rt6_info *rt,
2695 struct fib6_info *from,
2696 u32 cookie)
3da59bd9 2697{
36143645 2698 u32 rt_cookie = 0;
c5cff856 2699
49d05fe2 2700 if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
93531c67 2701 rt_cookie != cookie)
3da59bd9
MKL		 2702 return NULL;
2703
2704 if (rt6_check_expired(rt))
2705 return NULL;
2706
2707 return &rt->dst;
2708}
2709
a68886a6
DA		 2710static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2711 struct fib6_info *from,
2712 u32 cookie)
3da59bd9 2713{
5973fb1e
MKL		 2714 if (!__rt6_check_expired(rt) &&
2715 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
a68886a6 2716 fib6_check(from, cookie))
3da59bd9
MKL		 2717 return &rt->dst;
2718 else
2719 return NULL;
2720}
2721
bbd807df
BV		 2722INDIRECT_CALLABLE_SCOPE struct dst_entry *ip6_dst_check(struct dst_entry *dst,
2723 u32 cookie)
1da177e4 2724{
a87b7dc9 2725 struct dst_entry *dst_ret;
a68886a6 2726 struct fib6_info *from;
1da177e4
LT		 2727 struct rt6_info *rt;
2728
a87b7dc9
DA		 2729 rt = container_of(dst, struct rt6_info, dst);
2730
8f34e53b
DA		 2731 if (rt->sernum)
2732 return rt6_is_valid(rt) ? dst : NULL;
2733
a87b7dc9 2734 rcu_read_lock();
1da177e4 2735
6f3118b5
ND		 2736 /* All IPV6 dsts are created with ->obsolete set to the value
2737 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2738 * into this function always.
2739 */
e3bc10bd 2740
a68886a6 2741 from = rcu_dereference(rt->from);
4b32b5ad 2742
a68886a6
DA		 2743 if (from && (rt->rt6i_flags & RTF_PCPU ||
2744 unlikely(!list_empty(&rt->rt6i_uncached))))
2745 dst_ret = rt6_dst_from_check(rt, from, cookie);
3da59bd9 2746 else
a68886a6 2747 dst_ret = rt6_check(rt, from, cookie);
a87b7dc9
DA		 2748
2749 rcu_read_unlock();
2750
2751 return dst_ret;
1da177e4 2752}
9c97921a 2753EXPORT_INDIRECT_CALLABLE(ip6_dst_check);
1da177e4
LT		 2754
2755static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2756{
2757 struct rt6_info *rt = (struct rt6_info *) dst;
2758
2759 if (rt) {
54c1a859 2760 if (rt->rt6i_flags & RTF_CACHE) {
c3c14da0 2761 rcu_read_lock();
54c1a859 2762 if (rt6_check_expired(rt)) {
93531c67 2763 rt6_remove_exception_rt(rt);
54c1a859
YH		 2764 dst = NULL;
2765 }
c3c14da0 2766 rcu_read_unlock();
54c1a859 2767 } else {
1da177e4 2768 dst_release(dst);
54c1a859
YH		 2769 dst = NULL;
2770 }
1da177e4 2771 }
54c1a859 2772 return dst;
1da177e4
LT		 2773}
2774
2775static void ip6_link_failure(struct sk_buff *skb)
2776{
2777 struct rt6_info *rt;
2778
3ffe533c 2779 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1da177e4 2780
adf30907 2781 rt = (struct rt6_info *) skb_dst(skb);
1da177e4 2782 if (rt) {
8a14e46f 2783 rcu_read_lock();
1eb4f758 2784 if (rt->rt6i_flags & RTF_CACHE) {
761f6026 2785 rt6_remove_exception_rt(rt);
c5cff856 2786 } else {
a68886a6 2787 struct fib6_info *from;
c5cff856
WW		 2788 struct fib6_node *fn;
2789
a68886a6
DA		 2790 from = rcu_dereference(rt->from);
2791 if (from) {
2792 fn = rcu_dereference(from->fib6_node);
2793 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2794 fn->fn_sernum = -1;
2795 }
1eb4f758 2796 }
8a14e46f 2797 rcu_read_unlock();
1da177e4
LT		 2798 }
2799}
2800
6a3e030f
DA		 2801static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2802{
a68886a6
DA		 2803 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2804 struct fib6_info *from;
2805
2806 rcu_read_lock();
2807 from = rcu_dereference(rt0->from);
2808 if (from)
2809 rt0->dst.expires = from->expires;
2810 rcu_read_unlock();
2811 }
6a3e030f
DA		 2812
2813 dst_set_expires(&rt0->dst, timeout);
2814 rt0->rt6i_flags |= RTF_EXPIRES;
2815}
2816
45e4fd26
MKL		 2817static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2818{
2819 struct net *net = dev_net(rt->dst.dev);
2820
d4ead6b3 2821 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
45e4fd26 2822 rt->rt6i_flags |= RTF_MODIFIED;
45e4fd26
MKL		 2823 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2824}
2825
0d3f6d29
MKL		 2826static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2827{
2828 return !(rt->rt6i_flags & RTF_CACHE) &&
1490ed2a 2829 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
0d3f6d29
MKL		 2830}
2831
45e4fd26 2832static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
bd085ef6
HL		 2833 const struct ipv6hdr *iph, u32 mtu,
2834 bool confirm_neigh)
1da177e4 2835{
0dec879f 2836 const struct in6_addr *daddr, *saddr;
67ba4152 2837 struct rt6_info *rt6 = (struct rt6_info *)dst;
1da177e4 2838
09454fd0
MŻ		 2839 /* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU)
2840 * IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it.
2841 * [see also comment in rt6_mtu_change_route()]
2842 */
19bda36c 2843
0dec879f
JA		 2844 if (iph) {
2845 daddr = &iph->daddr;
2846 saddr = &iph->saddr;
2847 } else if (sk) {
2848 daddr = &sk->sk_v6_daddr;
2849 saddr = &inet6_sk(sk)->saddr;
2850 } else {
2851 daddr = NULL;
2852 saddr = NULL;
2853 }
bd085ef6
HL		 2854
2855 if (confirm_neigh)
2856 dst_confirm_neigh(dst, daddr);
2857
4a65dff8
GK		 2858 if (mtu < IPV6_MIN_MTU)
2859 return;
45e4fd26
MKL		 2860 if (mtu >= dst_mtu(dst))
2861 return;
9d289715 2862
0d3f6d29 2863 if (!rt6_cache_allowed_for_pmtu(rt6)) {
45e4fd26 2864 rt6_do_update_pmtu(rt6, mtu);
2b760fcf
WW		 2865 /* update rt6_ex->stamp for cache */
2866 if (rt6->rt6i_flags & RTF_CACHE)
2867 rt6_update_exception_stamp_rt(rt6);
0dec879f 2868 } else if (daddr) {
85bd05de 2869 struct fib6_result res = {};
45e4fd26
MKL		 2870 struct rt6_info *nrt6;
2871
4d85cd0c 2872 rcu_read_lock();
85bd05de 2873 res.f6i = rcu_dereference(rt6->from);
43a4b60d
DA		 2874 if (!res.f6i)
2875 goto out_unlock;
2876
7d21fec9
DA		 2877 res.fib6_flags = res.f6i->fib6_flags;
2878 res.fib6_type = res.f6i->fib6_type;
2879
2d44234b
DA		 2880 if (res.f6i->nh) {
2881 struct fib6_nh_match_arg arg = {
2882 .dev = dst->dev,
2883 .gw = &rt6->rt6i_gateway,
2884 };
2885
2886 nexthop_for_each_fib6_nh(res.f6i->nh,
2887 fib6_nh_find_match, &arg);
2888
2889 /* fib6_info uses a nexthop that does not have fib6_nh
2890 * using the dst->dev + gw. Should be impossible.
2891 */
43a4b60d
DA		 2892 if (!arg.match)
2893 goto out_unlock;
2d44234b
DA		 2894
2895 res.nh = arg.match;
2896 } else {
2897 res.nh = res.f6i->fib6_nh;
2898 }
2899
85bd05de 2900 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
45e4fd26
MKL		 2901 if (nrt6) {
2902 rt6_do_update_pmtu(nrt6, mtu);
5012f0a5 2903 if (rt6_insert_exception(nrt6, &res))
2b760fcf 2904 dst_release_immediate(&nrt6->dst);
45e4fd26 2905 }
43a4b60d 2906out_unlock:
a68886a6 2907 rcu_read_unlock();
1da177e4
LT		 2908 }
2909}
2910
45e4fd26 2911static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
bd085ef6
HL		 2912 struct sk_buff *skb, u32 mtu,
2913 bool confirm_neigh)
45e4fd26 2914{
bd085ef6
HL		 2915 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2916 confirm_neigh);
45e4fd26
MKL		 2917}
2918
42ae66c8 2919void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
e2d118a1 2920 int oif, u32 mark, kuid_t uid)
81aded24
DM		 2921{
2922 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2923 struct dst_entry *dst;
dc92095d
MŻ		 2924 struct flowi6 fl6 = {
2925 .flowi6_oif = oif,
2926 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2927 .daddr = iph->daddr,
2928 .saddr = iph->saddr,
2929 .flowlabel = ip6_flowinfo(iph),
2930 .flowi6_uid = uid,
2931 };
81aded24
DM		 2932
2933 dst = ip6_route_output(net, NULL, &fl6);
2934 if (!dst->error)
bd085ef6 2935 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
81aded24
DM		 2936 dst_release(dst);
2937}
2938EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2939
2940void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2941{
7ddacfa5 2942 int oif = sk->sk_bound_dev_if;
33c162a9
MKL		 2943 struct dst_entry *dst;
2944
7ddacfa5
DA		 2945 if (!oif && skb->dev)
2946 oif = l3mdev_master_ifindex(skb->dev);
2947
2948 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
33c162a9
MKL		 2949
2950 dst = __sk_dst_get(sk);
2951 if (!dst || !dst->obsolete ||
2952 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2953 return;
2954
2955 bh_lock_sock(sk);
2956 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2957 ip6_datagram_dst_update(sk, false);
2958 bh_unlock_sock(sk);
81aded24
DM		 2959}
2960EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2961
7d6850f7
AK		 2962void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2963 const struct flowi6 *fl6)
2964{
2965#ifdef CONFIG_IPV6_SUBTREES
2966 struct ipv6_pinfo *np = inet6_sk(sk);
2967#endif
2968
2969 ip6_dst_store(sk, dst,
2970 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2971 &sk->sk_v6_daddr : NULL,
2972#ifdef CONFIG_IPV6_SUBTREES
2973 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2974 &np->saddr :
2975#endif
2976 NULL);
2977}
2978
9b6b35ab 2979static bool ip6_redirect_nh_match(const struct fib6_result *res,
0b34eb00
DA		 2980 struct flowi6 *fl6,
2981 const struct in6_addr *gw,
2982 struct rt6_info **ret)
2983{
9b6b35ab
DA		 2984 const struct fib6_nh *nh = res->nh;
2985
0b34eb00
DA		 2986 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2987 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2988 return false;
2989
2990 /* rt_cache's gateway might be different from its 'parent'
2991 * in the case of an ip redirect.
2992 * So we keep searching in the exception table if the gateway
2993 * is different.
2994 */
2995 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
2996 struct rt6_info *rt_cache;
2997
9b6b35ab 2998 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
0b34eb00
DA		 2999 if (rt_cache &&
3000 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
3001 *ret = rt_cache;
3002 return true;
3003 }
3004 return false;
3005 }
3006 return true;
3007}
3008
c55c8988
DA		 3009struct fib6_nh_rd_arg {
3010 struct fib6_result *res;
3011 struct flowi6 *fl6;
3012 const struct in6_addr *gw;
3013 struct rt6_info **ret;
3014};
3015
3016static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
3017{
3018 struct fib6_nh_rd_arg *arg = _arg;
3019
3020 arg->res->nh = nh;
3021 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
3022}
3023
b55b76b2
DJ		 3024/* Handle redirects */
3025struct ip6rd_flowi {
3026 struct flowi6 fl6;
3027 struct in6_addr gateway;
3028};
3029
55cced4f 3030INDIRECT_CALLABLE_SCOPE struct rt6_info *__ip6_route_redirect(struct net *net,
b55b76b2
DJ		 3031 struct fib6_table *table,
3032 struct flowi6 *fl6,
b75cc8f9 3033 const struct sk_buff *skb,
b55b76b2
DJ		 3034 int flags)
3035{
3036 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
0b34eb00 3037 struct rt6_info *ret = NULL;
9b6b35ab 3038 struct fib6_result res = {};
c55c8988
DA		 3039 struct fib6_nh_rd_arg arg = {
3040 .res = &res,
3041 .fl6 = fl6,
3042 .gw = &rdfl->gateway,
3043 .ret = &ret
3044 };
8d1c802b 3045 struct fib6_info *rt;
b55b76b2
DJ		 3046 struct fib6_node *fn;
3047
31680ac2
DA		 3048 /* l3mdev_update_flow overrides oif if the device is enslaved; in
3049 * this case we must match on the real ingress device, so reset it
3050 */
3051 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
3052 fl6->flowi6_oif = skb->dev->ifindex;
3053
b55b76b2 3054 /* Get the "current" route for this destination and
67c408cf 3055 * check if the redirect has come from appropriate router.
b55b76b2
DJ		 3056 *
3057 * RFC 4861 specifies that redirects should only be
3058 * accepted if they come from the nexthop to the target.
3059 * Due to the way the routes are chosen, this notion
3060 * is a bit fuzzy and one might need to check all possible
3061 * routes.
3062 */
3063
66f5d6ce 3064 rcu_read_lock();
6454743b 3065 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
b55b76b2 3066restart:
66f5d6ce 3067 for_each_fib6_node_rt_rcu(fn) {
9b6b35ab 3068 res.f6i = rt;
14895687 3069 if (fib6_check_expired(rt))
b55b76b2 3070 continue;
93c2fb25 3071 if (rt->fib6_flags & RTF_REJECT)
b55b76b2 3072 break;
c55c8988
DA		 3073 if (unlikely(rt->nh)) {
3074 if (nexthop_is_blackhole(rt->nh))
3075 continue;
3076 /* on match, res->nh is filled in and potentially ret */
3077 if (nexthop_for_each_fib6_nh(rt->nh,
3078 fib6_nh_redirect_match,
3079 &arg))
3080 goto out;
3081 } else {
3082 res.nh = rt->fib6_nh;
3083 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
3084 &ret))
3085 goto out;
3086 }
b55b76b2
DJ		 3087 }
3088
3089 if (!rt)
421842ed 3090 rt = net->ipv6.fib6_null_entry;
93c2fb25 3091 else if (rt->fib6_flags & RTF_REJECT) {
23fb93a4 3092 ret = net->ipv6.ip6_null_entry;
b0a1ba59
MKL		 3093 goto out;
3094 }
3095
421842ed 3096 if (rt == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 3097 fn = fib6_backtrack(fn, &fl6->saddr);
3098 if (fn)
3099 goto restart;
b55b76b2 3100 }
a3c00e46 3101
9b6b35ab 3102 res.f6i = rt;
1cf844c7 3103 res.nh = rt->fib6_nh;
b0a1ba59 3104out:
7d21fec9 3105 if (ret) {
10585b43 3106 ip6_hold_safe(net, &ret);
7d21fec9
DA		 3107 } else {
3108 res.fib6_flags = res.f6i->fib6_flags;
3109 res.fib6_type = res.f6i->fib6_type;
9b6b35ab 3110 ret = ip6_create_rt_rcu(&res);
7d21fec9 3111 }
b55b76b2 3112
66f5d6ce 3113 rcu_read_unlock();
b55b76b2 3114
8ff2e5b2 3115 trace_fib6_table_lookup(net, &res, table, fl6);
23fb93a4 3116 return ret;
b55b76b2
DJ		 3117};
3118
3119static struct dst_entry *ip6_route_redirect(struct net *net,
b75cc8f9
DA		 3120 const struct flowi6 *fl6,
3121 const struct sk_buff *skb,
3122 const struct in6_addr *gateway)
b55b76b2
DJ		 3123{
3124 int flags = RT6_LOOKUP_F_HAS_SADDR;
3125 struct ip6rd_flowi rdfl;
3126
3127 rdfl.fl6 = *fl6;
3128 rdfl.gateway = *gateway;
3129
b75cc8f9 3130 return fib6_rule_lookup(net, &rdfl.fl6, skb,
b55b76b2
DJ		 3131 flags, __ip6_route_redirect);
3132}
3133
e2d118a1
LC		 3134void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
3135 kuid_t uid)
3a5ad2ee
DM		 3136{
3137 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
3138 struct dst_entry *dst;
1f7f10ac
MŻ		 3139 struct flowi6 fl6 = {
3140 .flowi6_iif = LOOPBACK_IFINDEX,
3141 .flowi6_oif = oif,
3142 .flowi6_mark = mark,
3143 .daddr = iph->daddr,
3144 .saddr = iph->saddr,
3145 .flowlabel = ip6_flowinfo(iph),
3146 .flowi6_uid = uid,
3147 };
3a5ad2ee 3148
b75cc8f9 3149 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
b55b76b2 3150 rt6_do_redirect(dst, NULL, skb);
3a5ad2ee
DM		 3151 dst_release(dst);
3152}
3153EXPORT_SYMBOL_GPL(ip6_redirect);
3154
d456336d 3155void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
c92a59ec
DJ		 3156{
3157 const struct ipv6hdr *iph = ipv6_hdr(skb);
3158 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3159 struct dst_entry *dst;
0b26fb17
MŻ		 3160 struct flowi6 fl6 = {
3161 .flowi6_iif = LOOPBACK_IFINDEX,
3162 .flowi6_oif = oif,
0b26fb17
MŻ		 3163 .daddr = msg->dest,
3164 .saddr = iph->daddr,
3165 .flowi6_uid = sock_net_uid(net, NULL),
3166 };
c92a59ec 3167
b75cc8f9 3168 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
b55b76b2 3169 rt6_do_redirect(dst, NULL, skb);
c92a59ec
DJ		 3170 dst_release(dst);
3171}
3172
3a5ad2ee
DM		 3173void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3174{
e2d118a1
LC		 3175 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
3176 sk->sk_uid);
3a5ad2ee
DM		 3177}
3178EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3179
0dbaee3b 3180static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1da177e4 3181{
0dbaee3b
DM		 3182 struct net_device *dev = dst->dev;
3183 unsigned int mtu = dst_mtu(dst);
3184 struct net *net = dev_net(dev);
3185
1da177e4
LT		 3186 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3187
5578689a
DL		 3188 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3189 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1da177e4
LT		 3190
3191 /*
1ab1457c
YH		 3192 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3193 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3194 * IPV6_MAXPLEN is also valid and means: "any MSS,
1da177e4
LT		 3195 * rely only on pmtu discovery"
3196 */
3197 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3198 mtu = IPV6_MAXPLEN;
3199 return mtu;
3200}
3201
f67fbeae 3202INDIRECT_CALLABLE_SCOPE unsigned int ip6_mtu(const struct dst_entry *dst)
d33e4553 3203{
d33e4553 3204 struct inet6_dev *idev;
d4ead6b3 3205 unsigned int mtu;
4b32b5ad
MKL		 3206
3207 mtu = dst_metric_raw(dst, RTAX_MTU);
618f9bc7 3208 if (mtu)
30f78d8e 3209 goto out;
618f9bc7
SK		 3210
3211 mtu = IPV6_MIN_MTU;
d33e4553
DM		 3212
3213 rcu_read_lock();
3214 idev = __in6_dev_get(dst->dev);
3215 if (idev)
3216 mtu = idev->cnf.mtu6;
3217 rcu_read_unlock();
3218
30f78d8e 3219out:
14972cbd
RP		 3220 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3221
3222 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
d33e4553 3223}
9c97921a 3224EXPORT_INDIRECT_CALLABLE(ip6_mtu);
d33e4553 3225
901731b8
DA		 3226/* MTU selection:
3227 * 1. mtu on route is locked - use it
3228 * 2. mtu from nexthop exception
3229 * 3. mtu from egress device
3230 *
3231 * based on ip6_dst_mtu_forward and exception logic of
3232 * rt6_find_cached_rt; called with rcu_read_lock
3233 */
b748f260
DA		 3234u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3235 const struct in6_addr *daddr,
3236 const struct in6_addr *saddr)
901731b8 3237{
b748f260
DA		 3238 const struct fib6_nh *nh = res->nh;
3239 struct fib6_info *f6i = res->f6i;
901731b8 3240 struct inet6_dev *idev;
510e2ced 3241 struct rt6_info *rt;
901731b8
DA		 3242 u32 mtu = 0;
3243
3244 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3245 mtu = f6i->fib6_pmtu;
3246 if (mtu)
3247 goto out;
3248 }
3249
510e2ced
WW		 3250 rt = rt6_find_cached_rt(res, daddr, saddr);
3251 if (unlikely(rt)) {
3252 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3253 } else {
b748f260 3254 struct net_device *dev = nh->fib_nh_dev;
901731b8
DA		 3255
3256 mtu = IPV6_MIN_MTU;
3257 idev = __in6_dev_get(dev);
3258 if (idev && idev->cnf.mtu6 > mtu)
3259 mtu = idev->cnf.mtu6;
3260 }
3261
3262 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3263out:
b748f260 3264 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
901731b8
DA		 3265}
3266
3b00944c 3267struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
87a11578 3268 struct flowi6 *fl6)
1da177e4 3269{
87a11578 3270 struct dst_entry *dst;
1da177e4
LT		 3271 struct rt6_info *rt;
3272 struct inet6_dev *idev = in6_dev_get(dev);
c346dca1 3273 struct net *net = dev_net(dev);
1da177e4 3274
38308473 3275 if (unlikely(!idev))
122bdf67 3276 return ERR_PTR(-ENODEV);
1da177e4 3277
ad706862 3278 rt = ip6_dst_alloc(net, dev, 0);
38308473 3279 if (unlikely(!rt)) {
1da177e4 3280 in6_dev_put(idev);
87a11578 3281 dst = ERR_PTR(-ENOMEM);
1da177e4
LT		 3282 goto out;
3283 }
3284
588753f1 3285 rt->dst.input = ip6_input;
8e2ec639 3286 rt->dst.output = ip6_output;
550bab42 3287 rt->rt6i_gateway = fl6->daddr;
87a11578 3288 rt->rt6i_dst.addr = fl6->daddr;
8e2ec639
YZ		 3289 rt->rt6i_dst.plen = 128;
3290 rt->rt6i_idev = idev;
14edd87d 3291 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1da177e4 3292
4c981e28 3293 /* Add this dst into uncached_list so that rt6_disable_ip() can
587fea74
WW		 3294 * do proper release of the net_device
3295 */
3296 rt6_uncached_list_add(rt);
81eb8447 3297 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
1da177e4 3298
87a11578
DM		 3299 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3300
1da177e4 3301out:
87a11578 3302 return dst;
1da177e4
LT		 3303}
3304
569d3645 3305static int ip6_dst_gc(struct dst_ops *ops)
1da177e4 3306{
86393e52 3307 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
7019b78e
DL		 3308 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3309 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3310 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3311 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3312 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
fc66f95c 3313 int entries;
7019b78e 3314
fc66f95c 3315 entries = dst_entries_get_fast(ops);
cf86a086
ED		 3316 if (entries > rt_max_size)
3317 entries = dst_entries_get_slow(ops);
3318
49a18d86 3319 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
fc66f95c 3320 entries <= rt_max_size)
1da177e4
LT		 3321 goto out;
3322
6891a346 3323 net->ipv6.ip6_rt_gc_expire++;
14956643 3324 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
fc66f95c
ED		 3325 entries = dst_entries_get_slow(ops);
3326 if (entries < ops->gc_thresh)
7019b78e 3327 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1da177e4 3328out:
7019b78e 3329 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
fc66f95c 3330 return entries > rt_max_size;
1da177e4
LT		 3331}
3332
b2c709cc
DA		 3333static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3334 const struct in6_addr *gw_addr, u32 tbid,
3335 int flags, struct fib6_result *res)
8c14586f
DA		 3336{
3337 struct flowi6 fl6 = {
3338 .flowi6_oif = cfg->fc_ifindex,
3339 .daddr = *gw_addr,
3340 .saddr = cfg->fc_prefsrc,
3341 };
3342 struct fib6_table *table;
b2c709cc 3343 int err;
8c14586f 3344
f4797b33 3345 table = fib6_get_table(net, tbid);
8c14586f 3346 if (!table)
b2c709cc 3347 return -EINVAL;
8c14586f
DA		 3348
3349 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3350 flags |= RT6_LOOKUP_F_HAS_SADDR;
3351
f4797b33 3352 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
8c14586f 3353
b2c709cc
DA		 3354 err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3355 if (!err && res->f6i != net->ipv6.fib6_null_entry)
3356 fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3357 cfg->fc_ifindex != 0, NULL, flags);
8c14586f 3358
b2c709cc 3359 return err;
8c14586f
DA		 3360}
3361
fc1e64e1
DA		 3362static int ip6_route_check_nh_onlink(struct net *net,
3363 struct fib6_config *cfg,
9fbb704c 3364 const struct net_device *dev,
fc1e64e1
DA		 3365 struct netlink_ext_ack *extack)
3366{
b2c709cc 3367 u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
fc1e64e1 3368 const struct in6_addr *gw_addr = &cfg->fc_gateway;
b2c709cc 3369 struct fib6_result res = {};
fc1e64e1
DA		 3370 int err;
3371
b2c709cc
DA		 3372 err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3373 if (!err && !(res.fib6_flags & RTF_REJECT) &&
3374 /* ignore match if it is the default route */
3375 !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3376 (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3377 NL_SET_ERR_MSG(extack,
3378 "Nexthop has invalid gateway or device mismatch");
3379 err = -EINVAL;
fc1e64e1
DA		 3380 }
3381
3382 return err;
3383}
3384
1edce99f
DA		 3385static int ip6_route_check_nh(struct net *net,
3386 struct fib6_config *cfg,
3387 struct net_device **_dev,
3388 struct inet6_dev **idev)
3389{
3390 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3391 struct net_device *dev = _dev ? *_dev : NULL;
b2c709cc
DA		 3392 int flags = RT6_LOOKUP_F_IFACE;
3393 struct fib6_result res = {};
1edce99f
DA		 3394 int err = -EHOSTUNREACH;
3395
3396 if (cfg->fc_table) {
b2c709cc
DA		 3397 err = ip6_nh_lookup_table(net, cfg, gw_addr,
3398 cfg->fc_table, flags, &res);
3399 /* gw_addr can not require a gateway or resolve to a reject
3400 * route. If a device is given, it must match the result.
3401 */
3402 if (err || res.fib6_flags & RTF_REJECT ||
3403 res.nh->fib_nh_gw_family ||
3404 (dev && dev != res.nh->fib_nh_dev))
3405 err = -EHOSTUNREACH;
1edce99f
DA		 3406 }
3407
b2c709cc
DA		 3408 if (err < 0) {
3409 struct flowi6 fl6 = {
3410 .flowi6_oif = cfg->fc_ifindex,
3411 .daddr = *gw_addr,
3412 };
1edce99f 3413
b2c709cc
DA		 3414 err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3415 if (err || res.fib6_flags & RTF_REJECT ||
3416 res.nh->fib_nh_gw_family)
3417 err = -EHOSTUNREACH;
3418
3419 if (err)
3420 return err;
3421
3422 fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3423 cfg->fc_ifindex != 0, NULL, flags);
3424 }
1edce99f 3425
b2c709cc 3426 err = 0;
1edce99f 3427 if (dev) {
b2c709cc
DA		 3428 if (dev != res.nh->fib_nh_dev)
3429 err = -EHOSTUNREACH;
1edce99f 3430 } else {
b2c709cc 3431 *_dev = dev = res.nh->fib_nh_dev;
1edce99f 3432 dev_hold(dev);
b2c709cc 3433 *idev = in6_dev_get(dev);
1edce99f
DA		 3434 }
3435
1edce99f
DA		 3436 return err;
3437}
3438
9fbb704c
DA		 3439static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3440 struct net_device **_dev, struct inet6_dev **idev,
3441 struct netlink_ext_ack *extack)
3442{
3443 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3444 int gwa_type = ipv6_addr_type(gw_addr);
232378e8 3445 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
9fbb704c 3446 const struct net_device *dev = *_dev;
232378e8 3447 bool need_addr_check = !dev;
9fbb704c
DA		 3448 int err = -EINVAL;
3449
3450 /* if gw_addr is local we will fail to detect this in case
3451 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3452 * will return already-added prefix route via interface that
3453 * prefix route was assigned to, which might be non-loopback.
3454 */
232378e8
DA		 3455 if (dev &&
3456 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3457 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
9fbb704c
DA		 3458 goto out;
3459 }
3460
3461 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3462 /* IPv6 strictly inhibits using not link-local
3463 * addresses as nexthop address.
3464 * Otherwise, router will not able to send redirects.
3465 * It is very good, but in some (rare!) circumstances
3466 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3467 * some exceptions. --ANK
3468 * We allow IPv4-mapped nexthops to support RFC4798-type
3469 * addressing
3470 */
3471 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3472 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3473 goto out;
3474 }
3475
b2c709cc
DA		 3476 rcu_read_lock();
3477
9fbb704c
DA		 3478 if (cfg->fc_flags & RTNH_F_ONLINK)
3479 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3480 else
3481 err = ip6_route_check_nh(net, cfg, _dev, idev);
3482
b2c709cc
DA		 3483 rcu_read_unlock();
3484
9fbb704c
DA		 3485 if (err)
3486 goto out;
3487 }
3488
3489 /* reload in case device was changed */
3490 dev = *_dev;
3491
3492 err = -EINVAL;
3493 if (!dev) {
3494 NL_SET_ERR_MSG(extack, "Egress device not specified");
3495 goto out;
3496 } else if (dev->flags & IFF_LOOPBACK) {
3497 NL_SET_ERR_MSG(extack,
3498 "Egress device can not be loopback device for this route");
3499 goto out;
3500 }
232378e8
DA		 3501
3502 /* if we did not check gw_addr above, do so now that the
3503 * egress device has been resolved.
3504 */
3505 if (need_addr_check &&
3506 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3507 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3508 goto out;
3509 }
3510
9fbb704c
DA		 3511 err = 0;
3512out:
3513 return err;
3514}
3515
83c44251
DA		 3516static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3517{
3518 if ((flags & RTF_REJECT) ||
3519 (dev && (dev->flags & IFF_LOOPBACK) &&
3520 !(addr_type & IPV6_ADDR_LOOPBACK) &&
aea23c32 3521 !(flags & (RTF_ANYCAST | RTF_LOCAL))))
83c44251
DA		 3522 return true;
3523
3524 return false;
3525}
3526
3527int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3528 struct fib6_config *cfg, gfp_t gfp_flags,
3529 struct netlink_ext_ack *extack)
3530{
3531 struct net_device *dev = NULL;
3532 struct inet6_dev *idev = NULL;
3533 int addr_type;
3534 int err;
3535
f1741730 3536 fib6_nh->fib_nh_family = AF_INET6;
1bef4c22
ED		 3537#ifdef CONFIG_IPV6_ROUTER_PREF
3538 fib6_nh->last_probe = jiffies;
3539#endif
38428d68
RP		 3540 if (cfg->fc_is_fdb) {
3541 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3542 fib6_nh->fib_nh_gw_family = AF_INET6;
3543 return 0;
3544 }
f1741730 3545
83c44251
DA		 3546 err = -ENODEV;
3547 if (cfg->fc_ifindex) {
3548 dev = dev_get_by_index(net, cfg->fc_ifindex);
3549 if (!dev)
3550 goto out;
3551 idev = in6_dev_get(dev);
3552 if (!idev)
3553 goto out;
3554 }
3555
3556 if (cfg->fc_flags & RTNH_F_ONLINK) {
3557 if (!dev) {
3558 NL_SET_ERR_MSG(extack,
3559 "Nexthop device required for onlink");
3560 goto out;
3561 }
3562
3563 if (!(dev->flags & IFF_UP)) {
3564 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3565 err = -ENETDOWN;
3566 goto out;
3567 }
3568
ad1601ae 3569 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
83c44251
DA		 3570 }
3571
ad1601ae 3572 fib6_nh->fib_nh_weight = 1;
83c44251
DA		 3573
3574 /* We cannot add true routes via loopback here,
3575 * they would result in kernel looping; promote them to reject routes
3576 */
3577 addr_type = ipv6_addr_type(&cfg->fc_dst);
3578 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3579 /* hold loopback dev/idev if we haven't done so. */
3580 if (dev != net->loopback_dev) {
3581 if (dev) {
3582 dev_put(dev);
3583 in6_dev_put(idev);
3584 }
3585 dev = net->loopback_dev;
3586 dev_hold(dev);
3587 idev = in6_dev_get(dev);
3588 if (!idev) {
3589 err = -ENODEV;
3590 goto out;
3591 }
3592 }
7dd73168 3593 goto pcpu_alloc;
83c44251
DA		 3594 }
3595
3596 if (cfg->fc_flags & RTF_GATEWAY) {
3597 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3598 if (err)
3599 goto out;
3600
ad1601ae 3601 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
bdf00467 3602 fib6_nh->fib_nh_gw_family = AF_INET6;
83c44251
DA		 3603 }
3604
3605 err = -ENODEV;
3606 if (!dev)
3607 goto out;
3608
3609 if (idev->cnf.disable_ipv6) {
3610 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3611 err = -EACCES;
3612 goto out;
3613 }
3614
3615 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3616 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3617 err = -ENETDOWN;
3618 goto out;
3619 }
3620
3621 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3622 !netif_carrier_ok(dev))
ad1601ae 3623 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
83c44251 3624
faee6769 3625 err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
7dd73168
DA		 3626 cfg->fc_encap_type, cfg, gfp_flags, extack);
3627 if (err)
3628 goto out;
3629
3630pcpu_alloc:
f40b6ae2
DA		 3631 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3632 if (!fib6_nh->rt6i_pcpu) {
3633 err = -ENOMEM;
3634 goto out;
3635 }
3636
ad1601ae 3637 fib6_nh->fib_nh_dev = dev;
f1741730 3638 fib6_nh->fib_nh_oif = dev->ifindex;
83c44251
DA		 3639 err = 0;
3640out:
3641 if (idev)
3642 in6_dev_put(idev);
3643
3644 if (err) {
ad1601ae
DA		 3645 lwtstate_put(fib6_nh->fib_nh_lws);
3646 fib6_nh->fib_nh_lws = NULL;
83c44251
DA		 3647 if (dev)
3648 dev_put(dev);
3649 }
3650
3651 return err;
3652}
3653
dac7d0f2
DA		 3654void fib6_nh_release(struct fib6_nh *fib6_nh)
3655{
cc5c073a
DA		 3656 struct rt6_exception_bucket *bucket;
3657
3658 rcu_read_lock();
3659
3660 fib6_nh_flush_exceptions(fib6_nh, NULL);
3661 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3662 if (bucket) {
3663 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3664 kfree(bucket);
3665 }
3666
3667 rcu_read_unlock();
3668
f40b6ae2
DA		 3669 if (fib6_nh->rt6i_pcpu) {
3670 int cpu;
3671
3672 for_each_possible_cpu(cpu) {
3673 struct rt6_info **ppcpu_rt;
3674 struct rt6_info *pcpu_rt;
3675
3676 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3677 pcpu_rt = *ppcpu_rt;
3678 if (pcpu_rt) {
3679 dst_dev_put(&pcpu_rt->dst);
3680 dst_release(&pcpu_rt->dst);
3681 *ppcpu_rt = NULL;
3682 }
3683 }
3684
3685 free_percpu(fib6_nh->rt6i_pcpu);
3686 }
3687
979e276e 3688 fib_nh_common_release(&fib6_nh->nh_common);
dac7d0f2
DA		 3689}
3690
8d1c802b 3691static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
acb54e3c 3692 gfp_t gfp_flags,
333c4301 3693 struct netlink_ext_ack *extack)
1da177e4 3694{
5578689a 3695 struct net *net = cfg->fc_nlinfo.nl_net;
8d1c802b 3696 struct fib6_info *rt = NULL;
f88d8ea6 3697 struct nexthop *nh = NULL;
c71099ac 3698 struct fib6_table *table;
f88d8ea6 3699 struct fib6_nh *fib6_nh;
8c5b83f0 3700 int err = -EINVAL;
83c44251 3701 int addr_type;
1da177e4 3702
557c44be 3703 /* RTF_PCPU is an internal flag; can not be set by userspace */
d5d531cb
DA		 3704 if (cfg->fc_flags & RTF_PCPU) {
3705 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
557c44be 3706 goto out;
d5d531cb 3707 }
557c44be 3708
2ea2352e
WW		 3709 /* RTF_CACHE is an internal flag; can not be set by userspace */
3710 if (cfg->fc_flags & RTF_CACHE) {
3711 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3712 goto out;
3713 }
3714
e8478e80
DA		 3715 if (cfg->fc_type > RTN_MAX) {
3716 NL_SET_ERR_MSG(extack, "Invalid route type");
3717 goto out;
3718 }
3719
d5d531cb
DA		 3720 if (cfg->fc_dst_len > 128) {
3721 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3722 goto out;
3723 }
3724 if (cfg->fc_src_len > 128) {
3725 NL_SET_ERR_MSG(extack, "Invalid source address length");
8c5b83f0 3726 goto out;
d5d531cb 3727 }
1da177e4 3728#ifndef CONFIG_IPV6_SUBTREES
d5d531cb
DA		 3729 if (cfg->fc_src_len) {
3730 NL_SET_ERR_MSG(extack,
3731 "Specifying source address requires IPV6_SUBTREES to be enabled");
8c5b83f0 3732 goto out;
d5d531cb 3733 }
1da177e4 3734#endif
5b98324e
DA		 3735 if (cfg->fc_nh_id) {
3736 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3737 if (!nh) {
3738 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3739 goto out;
3740 }
3741 err = fib6_check_nexthop(nh, cfg, extack);
3742 if (err)
3743 goto out;
3744 }
fc1e64e1 3745
d71314b4 3746 err = -ENOBUFS;
38308473
DM		 3747 if (cfg->fc_nlinfo.nlh &&
3748 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
d71314b4 3749 table = fib6_get_table(net, cfg->fc_table);
38308473 3750 if (!table) {
f3213831 3751 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
d71314b4
MV		 3752 table = fib6_new_table(net, cfg->fc_table);
3753 }
3754 } else {
3755 table = fib6_new_table(net, cfg->fc_table);
3756 }
38308473
DM		 3757
3758 if (!table)
c71099ac 3759 goto out;
c71099ac 3760
93531c67 3761 err = -ENOMEM;
f88d8ea6 3762 rt = fib6_info_alloc(gfp_flags, !nh);
93531c67 3763 if (!rt)
1da177e4 3764 goto out;
93531c67 3765
d7e774f3
DA		 3766 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3767 extack);
767a2217
DA		 3768 if (IS_ERR(rt->fib6_metrics)) {
3769 err = PTR_ERR(rt->fib6_metrics);
fda21d46
ED		 3770 /* Do not leave garbage there. */
3771 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
8fb4792f 3772 goto out_free;
767a2217
DA		 3773 }
3774
93531c67
DA		 3775 if (cfg->fc_flags & RTF_ADDRCONF)
3776 rt->dst_nocount = true;
1da177e4 3777
1716a961 3778 if (cfg->fc_flags & RTF_EXPIRES)
14895687 3779 fib6_set_expires(rt, jiffies +
1716a961
G		 3780 clock_t_to_jiffies(cfg->fc_expires));
3781 else
14895687 3782 fib6_clean_expires(rt);
1da177e4 3783
86872cb5
TG		 3784 if (cfg->fc_protocol == RTPROT_UNSPEC)
3785 cfg->fc_protocol = RTPROT_BOOT;
93c2fb25 3786 rt->fib6_protocol = cfg->fc_protocol;
86872cb5 3787
83c44251
DA		 3788 rt->fib6_table = table;
3789 rt->fib6_metric = cfg->fc_metric;
c7036d97 3790 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
2b2450ca 3791 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
19e42e45 3792
93c2fb25
DA		 3793 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3794 rt->fib6_dst.plen = cfg->fc_dst_len;
e5fd387a 3795
1da177e4 3796#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 3797 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3798 rt->fib6_src.plen = cfg->fc_src_len;
1da177e4 3799#endif
f88d8ea6 3800 if (nh) {
f88d8ea6 3801 if (rt->fib6_src.plen) {
4daa95af 3802 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
821bbf79 3803 goto out_free;
f88d8ea6 3804 }
706ec919
XY		 3805 if (!nexthop_get(nh)) {
3806 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
821bbf79 3807 goto out_free;
706ec919 3808 }
f88d8ea6
DA		 3809 rt->nh = nh;
3810 fib6_nh = nexthop_fib6_nh(rt->nh);
3811 } else {
3812 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3813 if (err)
3814 goto out;
1da177e4 3815
f88d8ea6
DA		 3816 fib6_nh = rt->fib6_nh;
3817
3818 /* We cannot add true routes via loopback here, they would
3819 * result in kernel looping; promote them to reject routes
3820 */
3821 addr_type = ipv6_addr_type(&cfg->fc_dst);
3822 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3823 addr_type))
3824 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3825 }
955ec4cb 3826
c3968a85 3827 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
f88d8ea6 3828 struct net_device *dev = fib6_nh->fib_nh_dev;
83c44251 3829
c3968a85 3830 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
d5d531cb 3831 NL_SET_ERR_MSG(extack, "Invalid source address");
c3968a85
DW		 3832 err = -EINVAL;
3833 goto out;
3834 }
93c2fb25
DA		 3835 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3836 rt->fib6_prefsrc.plen = 128;
c3968a85 3837 } else
93c2fb25 3838 rt->fib6_prefsrc.plen = 0;
c3968a85 3839
8c5b83f0 3840 return rt;
6b9ea5a6 3841out:
93531c67 3842 fib6_info_release(rt);
8c5b83f0 3843 return ERR_PTR(err);
821bbf79
CL		 3844out_free:
3845 ip_fib_metrics_put(rt->fib6_metrics);
3846 kfree(rt);
3847 return ERR_PTR(err);
6b9ea5a6
RP		 3848}
3849
acb54e3c 3850int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
333c4301 3851 struct netlink_ext_ack *extack)
6b9ea5a6 3852{
8d1c802b 3853 struct fib6_info *rt;
6b9ea5a6
RP		 3854 int err;
3855
acb54e3c 3856 rt = ip6_route_info_create(cfg, gfp_flags, extack);
d4ead6b3
DA		 3857 if (IS_ERR(rt))
3858 return PTR_ERR(rt);
6b9ea5a6 3859
d4ead6b3 3860 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
93531c67 3861 fib6_info_release(rt);
6b9ea5a6 3862
1da177e4
LT		 3863 return err;
3864}
3865
8d1c802b 3866static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
1da177e4 3867{
afb1d4b5 3868 struct net *net = info->nl_net;
c71099ac 3869 struct fib6_table *table;
afb1d4b5 3870 int err;
1da177e4 3871
421842ed 3872 if (rt == net->ipv6.fib6_null_entry) {
6825a26c
G		 3873 err = -ENOENT;
3874 goto out;
3875 }
6c813a72 3876
93c2fb25 3877 table = rt->fib6_table;
66f5d6ce 3878 spin_lock_bh(&table->tb6_lock);
86872cb5 3879 err = fib6_del(rt, info);
66f5d6ce 3880 spin_unlock_bh(&table->tb6_lock);
1da177e4 3881
6825a26c 3882out:
93531c67 3883 fib6_info_release(rt);
1da177e4
LT		 3884 return err;
3885}
3886
11dd74b3 3887int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
e0a1ad73 3888{
11dd74b3
RP		 3889 struct nl_info info = {
3890 .nl_net = net,
3891 .skip_notify = skip_notify
3892 };
afb1d4b5 3893
528c4ceb 3894 return __ip6_del_rt(rt, &info);
e0a1ad73
TG		 3895}
3896
8d1c802b 3897static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
0ae81335
DA		 3898{
3899 struct nl_info *info = &cfg->fc_nlinfo;
e3330039 3900 struct net *net = info->nl_net;
16a16cd3 3901 struct sk_buff *skb = NULL;
0ae81335 3902 struct fib6_table *table;
e3330039 3903 int err = -ENOENT;
0ae81335 3904
421842ed 3905 if (rt == net->ipv6.fib6_null_entry)
e3330039 3906 goto out_put;
93c2fb25 3907 table = rt->fib6_table;
66f5d6ce 3908 spin_lock_bh(&table->tb6_lock);
0ae81335 3909
93c2fb25 3910 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
8d1c802b 3911 struct fib6_info *sibling, *next_sibling;
0284696b 3912 struct fib6_node *fn;
0ae81335 3913
16a16cd3
DA		 3914 /* prefer to send a single notification with all hops */
3915 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3916 if (skb) {
3917 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3918
d4ead6b3 3919 if (rt6_fill_node(net, skb, rt, NULL,
16a16cd3
DA		 3920 NULL, NULL, 0, RTM_DELROUTE,
3921 info->portid, seq, 0) < 0) {
3922 kfree_skb(skb);
3923 skb = NULL;
3924 } else
3925 info->skip_notify = 1;
3926 }
3927
0284696b
IS		 3928 /* 'rt' points to the first sibling route. If it is not the
3929 * leaf, then we do not need to send a notification. Otherwise,
3930 * we need to check if the last sibling has a next route or not
3931 * and emit a replace or delete notification, respectively.
3932 */
2881fd61 3933 info->skip_notify_kernel = 1;
0284696b
IS		 3934 fn = rcu_dereference_protected(rt->fib6_node,
3935 lockdep_is_held(&table->tb6_lock));
3936 if (rcu_access_pointer(fn->leaf) == rt) {
3937 struct fib6_info *last_sibling, *replace_rt;
3938
3939 last_sibling = list_last_entry(&rt->fib6_siblings,
3940 struct fib6_info,
3941 fib6_siblings);
3942 replace_rt = rcu_dereference_protected(
3943 last_sibling->fib6_next,
3944 lockdep_is_held(&table->tb6_lock));
3945 if (replace_rt)
3946 call_fib6_entry_notifiers_replace(net,
3947 replace_rt);
3948 else
3949 call_fib6_multipath_entry_notifiers(net,
caafb250 3950 FIB_EVENT_ENTRY_DEL,
0284696b
IS		 3951 rt, rt->fib6_nsiblings,
3952 NULL);
3953 }
0ae81335 3954 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25
DA		 3955 &rt->fib6_siblings,
3956 fib6_siblings) {
0ae81335
DA		 3957 err = fib6_del(sibling, info);
3958 if (err)
e3330039 3959 goto out_unlock;
0ae81335
DA		 3960 }
3961 }
3962
3963 err = fib6_del(rt, info);
e3330039 3964out_unlock:
66f5d6ce 3965 spin_unlock_bh(&table->tb6_lock);
e3330039 3966out_put:
93531c67 3967 fib6_info_release(rt);
16a16cd3
DA		 3968
3969 if (skb) {
e3330039 3970 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
16a16cd3
DA		 3971 info->nlh, gfp_any());
3972 }
0ae81335
DA		 3973 return err;
3974}
3975
0fa6efc5 3976static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
23fb93a4
DA		 3977{
3978 int rc = -ESRCH;
3979
3980 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3981 goto out;
3982
3983 if (cfg->fc_flags & RTF_GATEWAY &&
3984 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3985 goto out;
761f6026
XL		 3986
3987 rc = rt6_remove_exception_rt(rt);
23fb93a4
DA		 3988out:
3989 return rc;
3990}
3991
0fa6efc5
DA		 3992static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3993 struct fib6_nh *nh)
3994{
3995 struct fib6_result res = {
3996 .f6i = rt,
3997 .nh = nh,
3998 };
3999 struct rt6_info *rt_cache;
4000
4001 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
4002 if (rt_cache)
4003 return __ip6_del_cached_rt(rt_cache, cfg);
4004
4005 return 0;
4006}
4007
5b98324e
DA		 4008struct fib6_nh_del_cached_rt_arg {
4009 struct fib6_config *cfg;
4010 struct fib6_info *f6i;
4011};
4012
4013static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
4014{
4015 struct fib6_nh_del_cached_rt_arg *arg = _arg;
4016 int rc;
4017
4018 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
4019 return rc != -ESRCH ? rc : 0;
4020}
4021
4022static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
4023{
4024 struct fib6_nh_del_cached_rt_arg arg = {
4025 .cfg = cfg,
4026 .f6i = f6i
4027 };
4028
4029 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
4030}
4031
333c4301
DA		 4032static int ip6_route_del(struct fib6_config *cfg,
4033 struct netlink_ext_ack *extack)
1da177e4 4034{
c71099ac 4035 struct fib6_table *table;
8d1c802b 4036 struct fib6_info *rt;
1da177e4 4037 struct fib6_node *fn;
1da177e4
LT		 4038 int err = -ESRCH;
4039
5578689a 4040 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
d5d531cb
DA		 4041 if (!table) {
4042 NL_SET_ERR_MSG(extack, "FIB table does not exist");
c71099ac 4043 return err;
d5d531cb 4044 }
c71099ac 4045
66f5d6ce 4046 rcu_read_lock();
1da177e4 4047
c71099ac 4048 fn = fib6_locate(&table->tb6_root,
86872cb5 4049 &cfg->fc_dst, cfg->fc_dst_len,
38fbeeee 4050 &cfg->fc_src, cfg->fc_src_len,
2b760fcf 4051 !(cfg->fc_flags & RTF_CACHE));
1ab1457c 4052
1da177e4 4053 if (fn) {
66f5d6ce 4054 for_each_fib6_node_rt_rcu(fn) {
ad1601ae
DA		 4055 struct fib6_nh *nh;
4056
3401bfb1
SB		 4057 if (rt->nh && cfg->fc_nh_id &&
4058 rt->nh->id != cfg->fc_nh_id)
5b98324e 4059 continue;
23fb93a4 4060
5b98324e
DA		 4061 if (cfg->fc_flags & RTF_CACHE) {
4062 int rc = 0;
4063
4064 if (rt->nh) {
4065 rc = ip6_del_cached_rt_nh(cfg, rt);
4066 } else if (cfg->fc_nh_id) {
4067 continue;
4068 } else {
4069 nh = rt->fib6_nh;
4070 rc = ip6_del_cached_rt(cfg, rt, nh);
4071 }
0fa6efc5
DA		 4072 if (rc != -ESRCH) {
4073 rcu_read_unlock();
4074 return rc;
23fb93a4
DA		 4075 }
4076 continue;
2b760fcf 4077 }
ad1601ae 4078
5b98324e
DA		 4079 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
4080 continue;
4081 if (cfg->fc_protocol &&
4082 cfg->fc_protocol != rt->fib6_protocol)
4083 continue;
4084
4085 if (rt->nh) {
4086 if (!fib6_info_hold_safe(rt))
4087 continue;
4088 rcu_read_unlock();
4089
4090 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4091 }
4092 if (cfg->fc_nh_id)
4093 continue;
4094
4095 nh = rt->fib6_nh;
86872cb5 4096 if (cfg->fc_ifindex &&
ad1601ae
DA		 4097 (!nh->fib_nh_dev ||
4098 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
1da177e4 4099 continue;
86872cb5 4100 if (cfg->fc_flags & RTF_GATEWAY &&
ad1601ae 4101 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
1da177e4 4102 continue;
e873e4b9
WW		 4103 if (!fib6_info_hold_safe(rt))
4104 continue;
66f5d6ce 4105 rcu_read_unlock();
1da177e4 4106
0ae81335
DA		 4107 /* if gateway was specified only delete the one hop */
4108 if (cfg->fc_flags & RTF_GATEWAY)
4109 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4110
4111 return __ip6_del_rt_siblings(rt, cfg);
1da177e4
LT		 4112 }
4113 }
66f5d6ce 4114 rcu_read_unlock();
1da177e4
LT		 4115
4116 return err;
4117}
4118
6700c270 4119static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
a6279458 4120{
a6279458 4121 struct netevent_redirect netevent;
e8599ff4 4122 struct rt6_info *rt, *nrt = NULL;
85bd05de 4123 struct fib6_result res = {};
e8599ff4
DM		 4124 struct ndisc_options ndopts;
4125 struct inet6_dev *in6_dev;
4126 struct neighbour *neigh;
71bcdba0 4127 struct rd_msg *msg;
6e157b6a
DM		 4128 int optlen, on_link;
4129 u8 *lladdr;
e8599ff4 4130
29a3cad5 4131 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
71bcdba0 4132 optlen -= sizeof(*msg);
e8599ff4
DM		 4133
4134 if (optlen < 0) {
6e157b6a 4135 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
e8599ff4
DM		 4136 return;
4137 }
4138
71bcdba0 4139 msg = (struct rd_msg *)icmp6_hdr(skb);
e8599ff4 4140
71bcdba0 4141 if (ipv6_addr_is_multicast(&msg->dest)) {
6e157b6a 4142 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
e8599ff4
DM		 4143 return;
4144 }
4145
6e157b6a 4146 on_link = 0;
71bcdba0 4147 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
e8599ff4 4148 on_link = 1;
71bcdba0 4149 } else if (ipv6_addr_type(&msg->target) !=
e8599ff4 4150 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
6e157b6a 4151 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
e8599ff4
DM		 4152 return;
4153 }
4154
4155 in6_dev = __in6_dev_get(skb->dev);
4156 if (!in6_dev)
4157 return;
4158 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
4159 return;
4160
4161 /* RFC2461 8.1:
4162 * The IP source address of the Redirect MUST be the same as the current
4163 * first-hop router for the specified ICMP Destination Address.
4164 */
4165
f997c55c 4166 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
e8599ff4
DM		 4167 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4168 return;
4169 }
6e157b6a
DM		 4170
4171 lladdr = NULL;
e8599ff4
DM		 4172 if (ndopts.nd_opts_tgt_lladdr) {
4173 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4174 skb->dev);
4175 if (!lladdr) {
4176 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4177 return;
4178 }
4179 }
4180
6e157b6a 4181 rt = (struct rt6_info *) dst;
ec13ad1d 4182 if (rt->rt6i_flags & RTF_REJECT) {
6e157b6a 4183 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
e8599ff4 4184 return;
6e157b6a 4185 }
e8599ff4 4186
6e157b6a
DM		 4187 /* Redirect received -> path was valid.
4188 * Look, redirects are sent only in response to data packets,
4189 * so that this nexthop apparently is reachable. --ANK
4190 */
0dec879f 4191 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
a6279458 4192
71bcdba0 4193 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
6e157b6a
DM		 4194 if (!neigh)
4195 return;
a6279458 4196
1da177e4
LT		 4197 /*
4198 * We have finally decided to accept it.
4199 */
4200
f997c55c 4201 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1da177e4
LT		 4202 NEIGH_UPDATE_F_WEAK_OVERRIDE|
4203 NEIGH_UPDATE_F_OVERRIDE|
4204 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
f997c55c
AA		 4205 NEIGH_UPDATE_F_ISROUTER)),
4206 NDISC_REDIRECT, &ndopts);
1da177e4 4207
4d85cd0c 4208 rcu_read_lock();
85bd05de 4209 res.f6i = rcu_dereference(rt->from);
ff24e498 4210 if (!res.f6i)
886b7a50 4211 goto out;
8a14e46f 4212
49d5b8ef
DA		 4213 if (res.f6i->nh) {
4214 struct fib6_nh_match_arg arg = {
4215 .dev = dst->dev,
4216 .gw = &rt->rt6i_gateway,
4217 };
4218
4219 nexthop_for_each_fib6_nh(res.f6i->nh,
4220 fib6_nh_find_match, &arg);
4221
4222 /* fib6_info uses a nexthop that does not have fib6_nh
4223 * using the dst->dev. Should be impossible
4224 */
4225 if (!arg.match)
4226 goto out;
4227 res.nh = arg.match;
4228 } else {
4229 res.nh = res.f6i->fib6_nh;
4230 }
4231
7d21fec9
DA		 4232 res.fib6_flags = res.f6i->fib6_flags;
4233 res.fib6_type = res.f6i->fib6_type;
85bd05de 4234 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
38308473 4235 if (!nrt)
1da177e4
LT		 4236 goto out;
4237
4238 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4239 if (on_link)
4240 nrt->rt6i_flags &= ~RTF_GATEWAY;
4241
4e3fd7a0 4242 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1da177e4 4243
886b7a50 4244 /* rt6_insert_exception() will take care of duplicated exceptions */
5012f0a5 4245 if (rt6_insert_exception(nrt, &res)) {
2b760fcf
WW		 4246 dst_release_immediate(&nrt->dst);
4247 goto out;
4248 }
1da177e4 4249
d8d1f30b
CG		 4250 netevent.old = &rt->dst;
4251 netevent.new = &nrt->dst;
71bcdba0 4252 netevent.daddr = &msg->dest;
60592833 4253 netevent.neigh = neigh;
8d71740c
TT		 4254 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4255
1da177e4 4256out:
886b7a50 4257 rcu_read_unlock();
e8599ff4 4258 neigh_release(neigh);
6e157b6a
DM		 4259}
4260
70ceb4f5 4261#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 4262static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 4263 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 4264 const struct in6_addr *gwaddr,
4265 struct net_device *dev)
70ceb4f5 4266{
830218c1
DA		 4267 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4268 int ifindex = dev->ifindex;
70ceb4f5 4269 struct fib6_node *fn;
8d1c802b 4270 struct fib6_info *rt = NULL;
c71099ac
TG		 4271 struct fib6_table *table;
4272
830218c1 4273 table = fib6_get_table(net, tb_id);
38308473 4274 if (!table)
c71099ac 4275 return NULL;
70ceb4f5 4276
66f5d6ce 4277 rcu_read_lock();
38fbeeee 4278 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
70ceb4f5
YH		 4279 if (!fn)
4280 goto out;
4281
66f5d6ce 4282 for_each_fib6_node_rt_rcu(fn) {
f88d8ea6
DA		 4283 /* these routes do not use nexthops */
4284 if (rt->nh)
4285 continue;
1cf844c7 4286 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
70ceb4f5 4287 continue;
2b2450ca 4288 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
1cf844c7 4289 !rt->fib6_nh->fib_nh_gw_family)
70ceb4f5 4290 continue;
1cf844c7 4291 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
70ceb4f5 4292 continue;
e873e4b9
WW		 4293 if (!fib6_info_hold_safe(rt))
4294 continue;
70ceb4f5
YH		 4295 break;
4296 }
4297out:
66f5d6ce 4298 rcu_read_unlock();
70ceb4f5
YH		 4299 return rt;
4300}
4301
8d1c802b 4302static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 4303 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 4304 const struct in6_addr *gwaddr,
4305 struct net_device *dev,
95c96174 4306 unsigned int pref)
70ceb4f5 4307{
86872cb5 4308 struct fib6_config cfg = {
238fc7ea 4309 .fc_metric = IP6_RT_PRIO_USER,
830218c1 4310 .fc_ifindex = dev->ifindex,
86872cb5
TG		 4311 .fc_dst_len = prefixlen,
4312 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4313 RTF_UP | RTF_PREF(pref),
b91d5329 4314 .fc_protocol = RTPROT_RA,
e8478e80 4315 .fc_type = RTN_UNICAST,
15e47304 4316 .fc_nlinfo.portid = 0,
efa2cea0
DL		 4317 .fc_nlinfo.nlh = NULL,
4318 .fc_nlinfo.nl_net = net,
86872cb5
TG		 4319 };
4320
91b2c9a0 4321 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4e3fd7a0
AD		 4322 cfg.fc_dst = *prefix;
4323 cfg.fc_gateway = *gwaddr;
70ceb4f5 4324
e317da96
YH		 4325 /* We should treat it as a default route if prefix length is 0. */
4326 if (!prefixlen)
86872cb5 4327 cfg.fc_flags |= RTF_DEFAULT;
70ceb4f5 4328
acb54e3c 4329 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
70ceb4f5 4330
830218c1 4331 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
70ceb4f5
YH		 4332}
4333#endif
4334
8d1c802b 4335struct fib6_info *rt6_get_dflt_router(struct net *net,
afb1d4b5
DA		 4336 const struct in6_addr *addr,
4337 struct net_device *dev)
1ab1457c 4338{
830218c1 4339 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
8d1c802b 4340 struct fib6_info *rt;
c71099ac 4341 struct fib6_table *table;
1da177e4 4342
afb1d4b5 4343 table = fib6_get_table(net, tb_id);
38308473 4344 if (!table)
c71099ac 4345 return NULL;
1da177e4 4346
66f5d6ce
WW		 4347 rcu_read_lock();
4348 for_each_fib6_node_rt_rcu(&table->tb6_root) {
f88d8ea6 4349 struct fib6_nh *nh;
ad1601ae 4350
f88d8ea6
DA		 4351 /* RA routes do not use nexthops */
4352 if (rt->nh)
4353 continue;
4354
4355 nh = rt->fib6_nh;
ad1601ae 4356 if (dev == nh->fib_nh_dev &&
93c2fb25 4357 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
ad1601ae 4358 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
1da177e4
LT		 4359 break;
4360 }
e873e4b9
WW		 4361 if (rt && !fib6_info_hold_safe(rt))
4362 rt = NULL;
66f5d6ce 4363 rcu_read_unlock();
1da177e4
LT		 4364 return rt;
4365}
4366
8d1c802b 4367struct fib6_info *rt6_add_dflt_router(struct net *net,
afb1d4b5 4368 const struct in6_addr *gwaddr,
ebacaaa0 4369 struct net_device *dev,
6b2e04bc
PC		 4370 unsigned int pref,
4371 u32 defrtr_usr_metric)
1da177e4 4372{
86872cb5 4373 struct fib6_config cfg = {
ca254490 4374 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
6b2e04bc 4375 .fc_metric = defrtr_usr_metric,
86872cb5
TG		 4376 .fc_ifindex = dev->ifindex,
4377 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4378 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
b91d5329 4379 .fc_protocol = RTPROT_RA,
e8478e80 4380 .fc_type = RTN_UNICAST,
15e47304 4381 .fc_nlinfo.portid = 0,
5578689a 4382 .fc_nlinfo.nlh = NULL,
afb1d4b5 4383 .fc_nlinfo.nl_net = net,
86872cb5 4384 };
1da177e4 4385
4e3fd7a0 4386 cfg.fc_gateway = *gwaddr;
1da177e4 4387
acb54e3c 4388 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
830218c1
DA		 4389 struct fib6_table *table;
4390
4391 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4392 if (table)
4393 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4394 }
1da177e4 4395
afb1d4b5 4396 return rt6_get_dflt_router(net, gwaddr, dev);
1da177e4
LT		 4397}
4398
afb1d4b5
DA		 4399static void __rt6_purge_dflt_routers(struct net *net,
4400 struct fib6_table *table)
1da177e4 4401{
8d1c802b 4402 struct fib6_info *rt;
1da177e4
LT		 4403
4404restart:
66f5d6ce
WW		 4405 rcu_read_lock();
4406 for_each_fib6_node_rt_rcu(&table->tb6_root) {
dcd1f572
DA		 4407 struct net_device *dev = fib6_info_nh_dev(rt);
4408 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4409
93c2fb25 4410 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
e873e4b9
WW		 4411 (!idev || idev->cnf.accept_ra != 2) &&
4412 fib6_info_hold_safe(rt)) {
93531c67 4413 rcu_read_unlock();
11dd74b3 4414 ip6_del_rt(net, rt, false);
1da177e4
LT		 4415 goto restart;
4416 }
4417 }
66f5d6ce 4418 rcu_read_unlock();
830218c1
DA		 4419
4420 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4421}
4422
4423void rt6_purge_dflt_routers(struct net *net)
4424{
4425 struct fib6_table *table;
4426 struct hlist_head *head;
4427 unsigned int h;
4428
4429 rcu_read_lock();
4430
4431 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4432 head = &net->ipv6.fib_table_hash[h];
4433 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4434 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
afb1d4b5 4435 __rt6_purge_dflt_routers(net, table);
830218c1
DA		 4436 }
4437 }
4438
4439 rcu_read_unlock();
1da177e4
LT		 4440}
4441
5578689a
DL		 4442static void rtmsg_to_fib6_config(struct net *net,
4443 struct in6_rtmsg *rtmsg,
86872cb5
TG		 4444 struct fib6_config *cfg)
4445{
8823a3ac
MŻ		 4446 *cfg = (struct fib6_config){
4447 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4448 : RT6_TABLE_MAIN,
4449 .fc_ifindex = rtmsg->rtmsg_ifindex,
67f69513 4450 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
8823a3ac
MŻ		 4451 .fc_expires = rtmsg->rtmsg_info,
4452 .fc_dst_len = rtmsg->rtmsg_dst_len,
4453 .fc_src_len = rtmsg->rtmsg_src_len,
4454 .fc_flags = rtmsg->rtmsg_flags,
4455 .fc_type = rtmsg->rtmsg_type,
4456
4457 .fc_nlinfo.nl_net = net,
4458
4459 .fc_dst = rtmsg->rtmsg_dst,
4460 .fc_src = rtmsg->rtmsg_src,
4461 .fc_gateway = rtmsg->rtmsg_gateway,
4462 };
86872cb5
TG		 4463}
4464
7c1552da 4465int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
1da177e4 4466{
86872cb5 4467 struct fib6_config cfg;
1da177e4
LT		 4468 int err;
4469
7c1552da
CH		 4470 if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4471 return -EINVAL;
4472 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4473 return -EPERM;
86872cb5 4474
7c1552da 4475 rtmsg_to_fib6_config(net, rtmsg, &cfg);
1da177e4 4476
7c1552da
CH		 4477 rtnl_lock();
4478 switch (cmd) {
4479 case SIOCADDRT:
4480 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4481 break;
4482 case SIOCDELRT:
4483 err = ip6_route_del(&cfg, NULL);
4484 break;
3ff50b79 4485 }
7c1552da
CH		 4486 rtnl_unlock();
4487 return err;
1da177e4
LT		 4488}
4489
4490/*
4491 * Drop the packet on the floor
4492 */
4493
d5fdd6ba 4494static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1da177e4 4495{
adf30907 4496 struct dst_entry *dst = skb_dst(skb);
1d3fd8a1
SS		 4497 struct net *net = dev_net(dst->dev);
4498 struct inet6_dev *idev;
4499 int type;
4500
4501 if (netif_is_l3_master(skb->dev) &&
4502 dst->dev == net->loopback_dev)
4503 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4504 else
4505 idev = ip6_dst_idev(dst);
4506
612f09e8
YH		 4507 switch (ipstats_mib_noroutes) {
4508 case IPSTATS_MIB_INNOROUTES:
0660e03f 4509 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
45bb0060 4510 if (type == IPV6_ADDR_ANY) {
1d3fd8a1 4511 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
612f09e8
YH		 4512 break;
4513 }
a8eceea8 4514 fallthrough;
612f09e8 4515 case IPSTATS_MIB_OUTNOROUTES:
1d3fd8a1 4516 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
612f09e8
YH		 4517 break;
4518 }
1d3fd8a1
SS		 4519
4520 /* Start over by dropping the dst for l3mdev case */
4521 if (netif_is_l3_master(skb->dev))
4522 skb_dst_drop(skb);
4523
3ffe533c 4524 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1da177e4
LT		 4525 kfree_skb(skb);
4526 return 0;
4527}
4528
9ce8ade0
TG		 4529static int ip6_pkt_discard(struct sk_buff *skb)
4530{
612f09e8 4531 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4532}
4533
ede2059d 4534static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
1da177e4 4535{
adf30907 4536 skb->dev = skb_dst(skb)->dev;
612f09e8 4537 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1da177e4
LT		 4538}
4539
9ce8ade0
TG		 4540static int ip6_pkt_prohibit(struct sk_buff *skb)
4541{
612f09e8 4542 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4543}
4544
ede2059d 4545static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
9ce8ade0 4546{
adf30907 4547 skb->dev = skb_dst(skb)->dev;
612f09e8 4548 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
9ce8ade0
TG		 4549}
4550
1da177e4
LT		 4551/*
4552 * Allocate a dst for local (unicast / anycast) address.
4553 */
4554
360a9887
DA		 4555struct fib6_info *addrconf_f6i_alloc(struct net *net,
4556 struct inet6_dev *idev,
4557 const struct in6_addr *addr,
4558 bool anycast, gfp_t gfp_flags)
1da177e4 4559{
c7a1ce39
DA		 4560 struct fib6_config cfg = {
4561 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4562 .fc_ifindex = idev->dev->ifindex,
d55a2e37 4563 .fc_flags = RTF_UP | RTF_NONEXTHOP,
c7a1ce39
DA		 4564 .fc_dst = *addr,
4565 .fc_dst_len = 128,
4566 .fc_protocol = RTPROT_KERNEL,
4567 .fc_nlinfo.nl_net = net,
4568 .fc_ignore_dev_down = true,
4569 };
d55a2e37 4570 struct fib6_info *f6i;
1da177e4 4571
e8478e80 4572 if (anycast) {
c7a1ce39
DA		 4573 cfg.fc_type = RTN_ANYCAST;
4574 cfg.fc_flags |= RTF_ANYCAST;
e8478e80 4575 } else {
c7a1ce39
DA		 4576 cfg.fc_type = RTN_LOCAL;
4577 cfg.fc_flags |= RTF_LOCAL;
e8478e80 4578 }
1da177e4 4579
d55a2e37 4580 f6i = ip6_route_info_create(&cfg, gfp_flags, NULL);
8652f17c 4581 if (!IS_ERR(f6i))
d55a2e37
MŻ		 4582 f6i->dst_nocount = true;
4583 return f6i;
1da177e4
LT		 4584}
4585
c3968a85
DW		 4586/* remove deleted ip from prefsrc entries */
4587struct arg_dev_net_ip {
4588 struct net_device *dev;
4589 struct net *net;
4590 struct in6_addr *addr;
4591};
4592
8d1c802b 4593static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
c3968a85
DW		 4594{
4595 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4596 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4597 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4598
f88d8ea6
DA		 4599 if (!rt->nh &&
4600 ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
421842ed 4601 rt != net->ipv6.fib6_null_entry &&
93c2fb25 4602 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
60006a48 4603 spin_lock_bh(&rt6_exception_lock);
c3968a85 4604 /* remove prefsrc entry */
93c2fb25 4605 rt->fib6_prefsrc.plen = 0;
60006a48 4606 spin_unlock_bh(&rt6_exception_lock);
c3968a85
DW		 4607 }
4608 return 0;
4609}
4610
4611void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4612{
4613 struct net *net = dev_net(ifp->idev->dev);
4614 struct arg_dev_net_ip adni = {
4615 .dev = ifp->idev->dev,
4616 .net = net,
4617 .addr = &ifp->addr,
4618 };
0c3584d5 4619 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
c3968a85
DW		 4620}
4621
2b2450ca 4622#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
be7a010d
DJ		 4623
4624/* Remove routers and update dst entries when gateway turn into host. */
8d1c802b 4625static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
be7a010d
DJ		 4626{
4627 struct in6_addr *gateway = (struct in6_addr *)arg;
f88d8ea6
DA		 4628 struct fib6_nh *nh;
4629
4630 /* RA routes do not use nexthops */
4631 if (rt->nh)
4632 return 0;
be7a010d 4633
f88d8ea6 4634 nh = rt->fib6_nh;
93c2fb25 4635 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
cc5c073a 4636 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
be7a010d 4637 return -1;
b16cb459
WW		 4638
4639 /* Further clean up cached routes in exception table.
4640 * This is needed because cached route may have a different
4641 * gateway than its 'parent' in the case of an ip redirect.
4642 */
cc5c073a 4643 fib6_nh_exceptions_clean_tohost(nh, gateway);
b16cb459 4644
be7a010d
DJ		 4645 return 0;
4646}
4647
4648void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4649{
4650 fib6_clean_all(net, fib6_clean_tohost, gateway);
4651}
4652
2127d95a
IS		 4653struct arg_netdev_event {
4654 const struct net_device *dev;
4c981e28 4655 union {
ecc5663c 4656 unsigned char nh_flags;
4c981e28
IS		 4657 unsigned long event;
4658 };
2127d95a
IS		 4659};
4660
8d1c802b 4661static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
d7dedee1 4662{
8d1c802b 4663 struct fib6_info *iter;
d7dedee1
IS		 4664 struct fib6_node *fn;
4665
93c2fb25
DA		 4666 fn = rcu_dereference_protected(rt->fib6_node,
4667 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4668 iter = rcu_dereference_protected(fn->leaf,
93c2fb25 4669 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4670 while (iter) {
93c2fb25 4671 if (iter->fib6_metric == rt->fib6_metric &&
33bd5ac5 4672 rt6_qualify_for_ecmp(iter))
d7dedee1 4673 return iter;
8fb11a9a 4674 iter = rcu_dereference_protected(iter->fib6_next,
93c2fb25 4675 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1
IS		 4676 }
4677
4678 return NULL;
4679}
4680
f88d8ea6 4681/* only called for fib entries with builtin fib6_nh */
8d1c802b 4682static bool rt6_is_dead(const struct fib6_info *rt)
d7dedee1 4683{
1cf844c7
DA		 4684 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4685 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4686 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
d7dedee1
IS		 4687 return true;
4688
4689 return false;
4690}
4691
8d1c802b 4692static int rt6_multipath_total_weight(const struct fib6_info *rt)
d7dedee1 4693{
8d1c802b 4694 struct fib6_info *iter;
d7dedee1
IS		 4695 int total = 0;
4696
4697 if (!rt6_is_dead(rt))
1cf844c7 4698 total += rt->fib6_nh->fib_nh_weight;
d7dedee1 4699
93c2fb25 4700 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
d7dedee1 4701 if (!rt6_is_dead(iter))
1cf844c7 4702 total += iter->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4703 }
4704
4705 return total;
4706}
4707
8d1c802b 4708static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
d7dedee1
IS		 4709{
4710 int upper_bound = -1;
4711
4712 if (!rt6_is_dead(rt)) {
1cf844c7 4713 *weight += rt->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4714 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4715 total) - 1;
4716 }
1cf844c7 4717 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
d7dedee1
IS		 4718}
4719
8d1c802b 4720static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
d7dedee1 4721{
8d1c802b 4722 struct fib6_info *iter;
d7dedee1
IS		 4723 int weight = 0;
4724
4725 rt6_upper_bound_set(rt, &weight, total);
4726
93c2fb25 4727 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
d7dedee1
IS		 4728 rt6_upper_bound_set(iter, &weight, total);
4729}
4730
8d1c802b 4731void rt6_multipath_rebalance(struct fib6_info *rt)
d7dedee1 4732{
8d1c802b 4733 struct fib6_info *first;
d7dedee1
IS		 4734 int total;
4735
4736 /* In case the entire multipath route was marked for flushing,
4737 * then there is no need to rebalance upon the removal of every
4738 * sibling route.
4739 */
93c2fb25 4740 if (!rt->fib6_nsiblings || rt->should_flush)
d7dedee1
IS		 4741 return;
4742
4743 /* During lookup routes are evaluated in order, so we need to
4744 * make sure upper bounds are assigned from the first sibling
4745 * onwards.
4746 */
4747 first = rt6_multipath_first_sibling(rt);
4748 if (WARN_ON_ONCE(!first))
4749 return;
4750
4751 total = rt6_multipath_total_weight(first);
4752 rt6_multipath_upper_bound_set(first, total);
4753}
4754
8d1c802b 4755static int fib6_ifup(struct fib6_info *rt, void *p_arg)
2127d95a
IS		 4756{
4757 const struct arg_netdev_event *arg = p_arg;
7aef6859 4758 struct net *net = dev_net(arg->dev);
2127d95a 4759
f88d8ea6 4760 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
1cf844c7
DA		 4761 rt->fib6_nh->fib_nh_dev == arg->dev) {
4762 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
7aef6859 4763 fib6_update_sernum_upto_root(net, rt);
d7dedee1 4764 rt6_multipath_rebalance(rt);
1de178ed 4765 }
2127d95a
IS		 4766
4767 return 0;
4768}
4769
ecc5663c 4770void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
2127d95a
IS		 4771{
4772 struct arg_netdev_event arg = {
4773 .dev = dev,
6802f3ad
IS		 4774 {
4775 .nh_flags = nh_flags,
4776 },
2127d95a
IS		 4777 };
4778
4779 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4780 arg.nh_flags |= RTNH_F_LINKDOWN;
4781
4782 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4783}
4784
f88d8ea6 4785/* only called for fib entries with inline fib6_nh */
8d1c802b 4786static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
1de178ed
IS		 4787 const struct net_device *dev)
4788{
8d1c802b 4789 struct fib6_info *iter;
1de178ed 4790
1cf844c7 4791 if (rt->fib6_nh->fib_nh_dev == dev)
1de178ed 4792 return true;
93c2fb25 4793 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7 4794 if (iter->fib6_nh->fib_nh_dev == dev)
1de178ed
IS		 4795 return true;
4796
4797 return false;
4798}
4799
8d1c802b 4800static void rt6_multipath_flush(struct fib6_info *rt)
1de178ed 4801{
8d1c802b 4802 struct fib6_info *iter;
1de178ed
IS		 4803
4804 rt->should_flush = 1;
93c2fb25 4805 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1de178ed
IS		 4806 iter->should_flush = 1;
4807}
4808
8d1c802b 4809static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
1de178ed
IS		 4810 const struct net_device *down_dev)
4811{
8d1c802b 4812 struct fib6_info *iter;
1de178ed
IS		 4813 unsigned int dead = 0;
4814
1cf844c7
DA		 4815 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4816 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed 4817 dead++;
93c2fb25 4818 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4819 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4820 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed
IS		 4821 dead++;
4822
4823 return dead;
4824}
4825
8d1c802b 4826static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
1de178ed 4827 const struct net_device *dev,
ecc5663c 4828 unsigned char nh_flags)
1de178ed 4829{
8d1c802b 4830 struct fib6_info *iter;
1de178ed 4831
1cf844c7
DA		 4832 if (rt->fib6_nh->fib_nh_dev == dev)
4833 rt->fib6_nh->fib_nh_flags |= nh_flags;
93c2fb25 4834 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4835 if (iter->fib6_nh->fib_nh_dev == dev)
4836 iter->fib6_nh->fib_nh_flags |= nh_flags;
1de178ed
IS		 4837}
4838
a1a22c12 4839/* called with write lock held for table with rt */
8d1c802b 4840static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
1da177e4 4841{
4c981e28
IS		 4842 const struct arg_netdev_event *arg = p_arg;
4843 const struct net_device *dev = arg->dev;
7aef6859 4844 struct net *net = dev_net(dev);
8ed67789 4845
f88d8ea6 4846 if (rt == net->ipv6.fib6_null_entry || rt->nh)
27c6fa73
IS		 4847 return 0;
4848
4849 switch (arg->event) {
4850 case NETDEV_UNREGISTER:
1cf844c7 4851 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
27c6fa73 4852 case NETDEV_DOWN:
1de178ed 4853 if (rt->should_flush)
27c6fa73 4854 return -1;
93c2fb25 4855 if (!rt->fib6_nsiblings)
1cf844c7 4856 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
1de178ed
IS		 4857 if (rt6_multipath_uses_dev(rt, dev)) {
4858 unsigned int count;
4859
4860 count = rt6_multipath_dead_count(rt, dev);
93c2fb25 4861 if (rt->fib6_nsiblings + 1 == count) {
1de178ed
IS		 4862 rt6_multipath_flush(rt);
4863 return -1;
4864 }
4865 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4866 RTNH_F_LINKDOWN);
7aef6859 4867 fib6_update_sernum(net, rt);
d7dedee1 4868 rt6_multipath_rebalance(rt);
1de178ed
IS		 4869 }
4870 return -2;
27c6fa73 4871 case NETDEV_CHANGE:
1cf844c7 4872 if (rt->fib6_nh->fib_nh_dev != dev ||
93c2fb25 4873 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
27c6fa73 4874 break;
1cf844c7 4875 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
d7dedee1 4876 rt6_multipath_rebalance(rt);
27c6fa73 4877 break;
2b241361 4878 }
c159d30c 4879
1da177e4
LT		 4880 return 0;
4881}
4882
27c6fa73 4883void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
1da177e4 4884{
4c981e28 4885 struct arg_netdev_event arg = {
8ed67789 4886 .dev = dev,
6802f3ad
IS		 4887 {
4888 .event = event,
4889 },
8ed67789 4890 };
7c6bb7d2 4891 struct net *net = dev_net(dev);
8ed67789 4892
7c6bb7d2
DA		 4893 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4894 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4895 else
4896 fib6_clean_all(net, fib6_ifdown, &arg);
4c981e28
IS		 4897}
4898
4899void rt6_disable_ip(struct net_device *dev, unsigned long event)
4900{
4901 rt6_sync_down_dev(dev, event);
4902 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4903 neigh_ifdown(&nd_tbl, dev);
1da177e4
LT		 4904}
4905
95c96174 4906struct rt6_mtu_change_arg {
1da177e4 4907 struct net_device *dev;
95c96174 4908 unsigned int mtu;
c0b220cf 4909 struct fib6_info *f6i;
1da177e4
LT		 4910};
4911
cc5c073a 4912static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
c0b220cf
DA		 4913{
4914 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
cc5c073a 4915 struct fib6_info *f6i = arg->f6i;
c0b220cf
DA		 4916
4917 /* For administrative MTU increase, there is no way to discover
4918 * IPv6 PMTU increase, so PMTU increase should be updated here.
4919 * Since RFC 1981 doesn't include administrative MTU increase
4920 * update PMTU increase is a MUST. (i.e. jumbo frame)
4921 */
4922 if (nh->fib_nh_dev == arg->dev) {
4923 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4924 u32 mtu = f6i->fib6_pmtu;
4925
4926 if (mtu >= arg->mtu ||
4927 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4928 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4929
4930 spin_lock_bh(&rt6_exception_lock);
cc5c073a 4931 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
c0b220cf
DA		 4932 spin_unlock_bh(&rt6_exception_lock);
4933 }
4934
4935 return 0;
4936}
4937
4938static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
1da177e4
LT		 4939{
4940 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4941 struct inet6_dev *idev;
4942
4943 /* In IPv6 pmtu discovery is not optional,
4944 so that RTAX_MTU lock cannot disable it.
4945 We still use this lock to block changes
4946 caused by addrconf/ndisc.
4947 */
4948
4949 idev = __in6_dev_get(arg->dev);
38308473 4950 if (!idev)
1da177e4
LT		 4951 return 0;
4952
c0b220cf
DA		 4953 if (fib6_metric_locked(f6i, RTAX_MTU))
4954 return 0;
d4ead6b3 4955
c0b220cf 4956 arg->f6i = f6i;
2d44234b
DA		 4957 if (f6i->nh) {
4958 /* fib6_nh_mtu_change only returns 0, so this is safe */
4959 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4960 arg);
4961 }
4962
1cf844c7 4963 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
1da177e4
LT		 4964}
4965
95c96174 4966void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
1da177e4 4967{
c71099ac
TG		 4968 struct rt6_mtu_change_arg arg = {
4969 .dev = dev,
4970 .mtu = mtu,
4971 };
1da177e4 4972
0c3584d5 4973 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
1da177e4
LT		 4974}
4975
ef7c79ed 4976static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
75425657 4977 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
5176f91e 4978 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
aa8f8778 4979 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
86872cb5 4980 [RTA_OIF] = { .type = NLA_U32 },
ab364a6f 4981 [RTA_IIF] = { .type = NLA_U32 },
86872cb5
TG		 4982 [RTA_PRIORITY] = { .type = NLA_U32 },
4983 [RTA_METRICS] = { .type = NLA_NESTED },
51ebd318 4984 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
c78ba6d6 4985 [RTA_PREF] = { .type = NLA_U8 },
19e42e45
RP		 4986 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4987 [RTA_ENCAP] = { .type = NLA_NESTED },
32bc201e 4988 [RTA_EXPIRES] = { .type = NLA_U32 },
622ec2c9 4989 [RTA_UID] = { .type = NLA_U32 },
3b45a410 4990 [RTA_MARK] = { .type = NLA_U32 },
aa8f8778 4991 [RTA_TABLE] = { .type = NLA_U32 },
eacb9384
RP		 4992 [RTA_IP_PROTO] = { .type = NLA_U8 },
4993 [RTA_SPORT] = { .type = NLA_U16 },
4994 [RTA_DPORT] = { .type = NLA_U16 },
5b98324e 4995 [RTA_NH_ID] = { .type = NLA_U32 },
86872cb5
TG		 4996};
4997
4998static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
333c4301
DA		 4999 struct fib6_config *cfg,
5000 struct netlink_ext_ack *extack)
1da177e4 5001{
86872cb5
TG		 5002 struct rtmsg *rtm;
5003 struct nlattr *tb[RTA_MAX+1];
c78ba6d6 5004 unsigned int pref;
86872cb5 5005 int err;
1da177e4 5006
8cb08174
JB		 5007 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5008 rtm_ipv6_policy, extack);
86872cb5
TG		 5009 if (err < 0)
5010 goto errout;
1da177e4 5011
86872cb5
TG		 5012 err = -EINVAL;
5013 rtm = nlmsg_data(nlh);
86872cb5 5014
84db8407
MŻ		 5015 *cfg = (struct fib6_config){
5016 .fc_table = rtm->rtm_table,
5017 .fc_dst_len = rtm->rtm_dst_len,
5018 .fc_src_len = rtm->rtm_src_len,
5019 .fc_flags = RTF_UP,
5020 .fc_protocol = rtm->rtm_protocol,
5021 .fc_type = rtm->rtm_type,
5022
5023 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
5024 .fc_nlinfo.nlh = nlh,
5025 .fc_nlinfo.nl_net = sock_net(skb->sk),
5026 };
86872cb5 5027
ef2c7d7b
ND		 5028 if (rtm->rtm_type == RTN_UNREACHABLE ||
5029 rtm->rtm_type == RTN_BLACKHOLE ||
b4949ab2
ND		 5030 rtm->rtm_type == RTN_PROHIBIT ||
5031 rtm->rtm_type == RTN_THROW)
86872cb5
TG		 5032 cfg->fc_flags |= RTF_REJECT;
5033
ab79ad14
MŻ		 5034 if (rtm->rtm_type == RTN_LOCAL)
5035 cfg->fc_flags |= RTF_LOCAL;
5036
1f56a01f
MKL		 5037 if (rtm->rtm_flags & RTM_F_CLONED)
5038 cfg->fc_flags |= RTF_CACHE;
5039
fc1e64e1
DA		 5040 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
5041
5b98324e
DA		 5042 if (tb[RTA_NH_ID]) {
5043 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
5044 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
5045 NL_SET_ERR_MSG(extack,
5046 "Nexthop specification and nexthop id are mutually exclusive");
5047 goto errout;
5048 }
5049 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
5050 }
5051
86872cb5 5052 if (tb[RTA_GATEWAY]) {
67b61f6c 5053 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
86872cb5 5054 cfg->fc_flags |= RTF_GATEWAY;
1da177e4 5055 }
e3818541
DA		 5056 if (tb[RTA_VIA]) {
5057 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
5058 goto errout;
5059 }
86872cb5
TG		 5060
5061 if (tb[RTA_DST]) {
5062 int plen = (rtm->rtm_dst_len + 7) >> 3;
5063
5064 if (nla_len(tb[RTA_DST]) < plen)
5065 goto errout;
5066
5067 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1da177e4 5068 }
86872cb5
TG		 5069
5070 if (tb[RTA_SRC]) {
5071 int plen = (rtm->rtm_src_len + 7) >> 3;
5072
5073 if (nla_len(tb[RTA_SRC]) < plen)
5074 goto errout;
5075
5076 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1da177e4 5077 }
86872cb5 5078
c3968a85 5079 if (tb[RTA_PREFSRC])
67b61f6c 5080 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
c3968a85 5081
86872cb5
TG		 5082 if (tb[RTA_OIF])
5083 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
5084
5085 if (tb[RTA_PRIORITY])
5086 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
5087
5088 if (tb[RTA_METRICS]) {
5089 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
5090 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
1da177e4 5091 }
86872cb5
TG		 5092
5093 if (tb[RTA_TABLE])
5094 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
5095
51ebd318
ND		 5096 if (tb[RTA_MULTIPATH]) {
5097 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
5098 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
9ed59592
DA		 5099
5100 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
c255bd68 5101 cfg->fc_mp_len, extack);
9ed59592
DA		 5102 if (err < 0)
5103 goto errout;
51ebd318
ND		 5104 }
5105
c78ba6d6
LR		 5106 if (tb[RTA_PREF]) {
5107 pref = nla_get_u8(tb[RTA_PREF]);
5108 if (pref != ICMPV6_ROUTER_PREF_LOW &&
5109 pref != ICMPV6_ROUTER_PREF_HIGH)
5110 pref = ICMPV6_ROUTER_PREF_MEDIUM;
5111 cfg->fc_flags |= RTF_PREF(pref);
5112 }
5113
19e42e45
RP		 5114 if (tb[RTA_ENCAP])
5115 cfg->fc_encap = tb[RTA_ENCAP];
5116
9ed59592 5117 if (tb[RTA_ENCAP_TYPE]) {
19e42e45
RP		 5118 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5119
c255bd68 5120 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
9ed59592
DA		 5121 if (err < 0)
5122 goto errout;
5123 }
5124
32bc201e
XL		 5125 if (tb[RTA_EXPIRES]) {
5126 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5127
5128 if (addrconf_finite_timeout(timeout)) {
5129 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5130 cfg->fc_flags |= RTF_EXPIRES;
5131 }
5132 }
5133
86872cb5
TG		 5134 err = 0;
5135errout:
5136 return err;
1da177e4
LT		 5137}
5138
6b9ea5a6 5139struct rt6_nh {
8d1c802b 5140 struct fib6_info *fib6_info;
6b9ea5a6 5141 struct fib6_config r_cfg;
6b9ea5a6
RP		 5142 struct list_head next;
5143};
5144
d4ead6b3
DA		 5145static int ip6_route_info_append(struct net *net,
5146 struct list_head *rt6_nh_list,
8d1c802b
DA		 5147 struct fib6_info *rt,
5148 struct fib6_config *r_cfg)
6b9ea5a6
RP		 5149{
5150 struct rt6_nh *nh;
6b9ea5a6
RP		 5151 int err = -EEXIST;
5152
5153 list_for_each_entry(nh, rt6_nh_list, next) {
8d1c802b
DA		 5154 /* check if fib6_info already exists */
5155 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
6b9ea5a6
RP		 5156 return err;
5157 }
5158
5159 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5160 if (!nh)
5161 return -ENOMEM;
8d1c802b 5162 nh->fib6_info = rt;
6b9ea5a6
RP		 5163 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5164 list_add_tail(&nh->next, rt6_nh_list);
5165
5166 return 0;
5167}
5168
8d1c802b
DA		 5169static void ip6_route_mpath_notify(struct fib6_info *rt,
5170 struct fib6_info *rt_last,
3b1137fe
DA		 5171 struct nl_info *info,
5172 __u16 nlflags)
5173{
5174 /* if this is an APPEND route, then rt points to the first route
5175 * inserted and rt_last points to last route inserted. Userspace
5176 * wants a consistent dump of the route which starts at the first
5177 * nexthop. Since sibling routes are always added at the end of
5178 * the list, find the first sibling of the last route appended
5179 */
93c2fb25
DA		 5180 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5181 rt = list_first_entry(&rt_last->fib6_siblings,
8d1c802b 5182 struct fib6_info,
93c2fb25 5183 fib6_siblings);
3b1137fe
DA		 5184 }
5185
5186 if (rt)
5187 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5188}
5189
0ee0f47c
IS		 5190static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5191{
5192 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5193 bool should_notify = false;
5194 struct fib6_info *leaf;
5195 struct fib6_node *fn;
5196
5197 rcu_read_lock();
5198 fn = rcu_dereference(rt->fib6_node);
5199 if (!fn)
5200 goto out;
5201
5202 leaf = rcu_dereference(fn->leaf);
5203 if (!leaf)
5204 goto out;
5205
5206 if (rt == leaf ||
5207 (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5208 rt6_qualify_for_ecmp(leaf)))
5209 should_notify = true;
5210out:
5211 rcu_read_unlock();
5212
5213 return should_notify;
5214}
5215
333c4301
DA		 5216static int ip6_route_multipath_add(struct fib6_config *cfg,
5217 struct netlink_ext_ack *extack)
51ebd318 5218{
8d1c802b 5219 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
3b1137fe 5220 struct nl_info *info = &cfg->fc_nlinfo;
51ebd318
ND		 5221 struct fib6_config r_cfg;
5222 struct rtnexthop *rtnh;
8d1c802b 5223 struct fib6_info *rt;
6b9ea5a6
RP		 5224 struct rt6_nh *err_nh;
5225 struct rt6_nh *nh, *nh_safe;
3b1137fe 5226 __u16 nlflags;
51ebd318
ND		 5227 int remaining;
5228 int attrlen;
6b9ea5a6
RP		 5229 int err = 1;
5230 int nhn = 0;
5231 int replace = (cfg->fc_nlinfo.nlh &&
5232 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5233 LIST_HEAD(rt6_nh_list);
51ebd318 5234
3b1137fe
DA		 5235 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5236 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5237 nlflags |= NLM_F_APPEND;
5238
35f1b4e9 5239 remaining = cfg->fc_mp_len;
51ebd318 5240 rtnh = (struct rtnexthop *)cfg->fc_mp;
51ebd318 5241
6b9ea5a6 5242 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
8d1c802b 5243 * fib6_info structs per nexthop
6b9ea5a6 5244 */
51ebd318
ND		 5245 while (rtnh_ok(rtnh, remaining)) {
5246 memcpy(&r_cfg, cfg, sizeof(*cfg));
5247 if (rtnh->rtnh_ifindex)
5248 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5249
5250 attrlen = rtnh_attrlen(rtnh);
5251 if (attrlen > 0) {
5252 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5253
5254 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5255 if (nla) {
67b61f6c 5256 r_cfg.fc_gateway = nla_get_in6_addr(nla);
51ebd318
ND		 5257 r_cfg.fc_flags |= RTF_GATEWAY;
5258 }
19e42e45
RP		 5259 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5260 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5261 if (nla)
5262 r_cfg.fc_encap_type = nla_get_u16(nla);
51ebd318 5263 }
6b9ea5a6 5264
68e2ffde 5265 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
acb54e3c 5266 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
8c5b83f0
RP		 5267 if (IS_ERR(rt)) {
5268 err = PTR_ERR(rt);
5269 rt = NULL;
6b9ea5a6 5270 goto cleanup;
8c5b83f0 5271 }
b5d2d75e
DA		 5272 if (!rt6_qualify_for_ecmp(rt)) {
5273 err = -EINVAL;
5274 NL_SET_ERR_MSG(extack,
5275 "Device only routes can not be added for IPv6 using the multipath API.");
5276 fib6_info_release(rt);
5277 goto cleanup;
5278 }
6b9ea5a6 5279
1cf844c7 5280 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
398958ae 5281
d4ead6b3
DA		 5282 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5283 rt, &r_cfg);
51ebd318 5284 if (err) {
93531c67 5285 fib6_info_release(rt);
6b9ea5a6
RP		 5286 goto cleanup;
5287 }
5288
5289 rtnh = rtnh_next(rtnh, &remaining);
5290 }
5291
9eee3b49
IS		 5292 if (list_empty(&rt6_nh_list)) {
5293 NL_SET_ERR_MSG(extack,
5294 "Invalid nexthop configuration - no valid nexthops");
5295 return -EINVAL;
5296 }
5297
3b1137fe
DA		 5298 /* for add and replace send one notification with all nexthops.
5299 * Skip the notification in fib6_add_rt2node and send one with
5300 * the full route when done
5301 */
5302 info->skip_notify = 1;
5303
ebee3cad
IS		 5304 /* For add and replace, send one notification with all nexthops. For
5305 * append, send one notification with all appended nexthops.
5306 */
5307 info->skip_notify_kernel = 1;
5308
6b9ea5a6
RP		 5309 err_nh = NULL;
5310 list_for_each_entry(nh, &rt6_nh_list, next) {
8d1c802b
DA		 5311 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5312 fib6_info_release(nh->fib6_info);
93531c67 5313
f7225172
DA		 5314 if (!err) {
5315 /* save reference to last route successfully inserted */
5316 rt_last = nh->fib6_info;
5317
5318 /* save reference to first route for notification */
5319 if (!rt_notif)
5320 rt_notif = nh->fib6_info;
5321 }
3b1137fe 5322
8d1c802b
DA		 5323 /* nh->fib6_info is used or freed at this point, reset to NULL*/
5324 nh->fib6_info = NULL;
6b9ea5a6
RP		 5325 if (err) {
5326 if (replace && nhn)
a5a82d84
JK		 5327 NL_SET_ERR_MSG_MOD(extack,
5328 "multipath route replace failed (check consistency of installed routes)");
6b9ea5a6
RP		 5329 err_nh = nh;
5330 goto add_errout;
51ebd318 5331 }
6b9ea5a6 5332
1a72418b 5333 /* Because each route is added like a single route we remove
27596472
MK		 5334 * these flags after the first nexthop: if there is a collision,
5335 * we have already failed to add the first nexthop:
5336 * fib6_add_rt2node() has rejected it; when replacing, old
5337 * nexthops have been replaced by first new, the rest should
5338 * be added to it.
1a72418b 5339 */
864db232
MUA		 5340 if (cfg->fc_nlinfo.nlh) {
5341 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5342 NLM_F_REPLACE);
5343 cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
5344 }
6b9ea5a6
RP		 5345 nhn++;
5346 }
5347
0ee0f47c
IS		 5348 /* An in-kernel notification should only be sent in case the new
5349 * multipath route is added as the first route in the node, or if
5350 * it was appended to it. We pass 'rt_notif' since it is the first
5351 * sibling and might allow us to skip some checks in the replace case.
5352 */
5353 if (ip6_route_mpath_should_notify(rt_notif)) {
5354 enum fib_event_type fib_event;
5355
5356 if (rt_notif->fib6_nsiblings != nhn - 1)
5357 fib_event = FIB_EVENT_ENTRY_APPEND;
5358 else
caafb250 5359 fib_event = FIB_EVENT_ENTRY_REPLACE;
0ee0f47c
IS		 5360
5361 err = call_fib6_multipath_entry_notifiers(info->nl_net,
5362 fib_event, rt_notif,
5363 nhn - 1, extack);
5364 if (err) {
5365 /* Delete all the siblings that were just added */
5366 err_nh = NULL;
5367 goto add_errout;
5368 }
5369 }
ebee3cad 5370
3b1137fe
DA		 5371 /* success ... tell user about new route */
5372 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
6b9ea5a6
RP		 5373 goto cleanup;
5374
5375add_errout:
3b1137fe
DA		 5376 /* send notification for routes that were added so that
5377 * the delete notifications sent by ip6_route_del are
5378 * coherent
5379 */
5380 if (rt_notif)
5381 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5382
6b9ea5a6
RP		 5383 /* Delete routes that were already added */
5384 list_for_each_entry(nh, &rt6_nh_list, next) {
5385 if (err_nh == nh)
5386 break;
333c4301 5387 ip6_route_del(&nh->r_cfg, extack);
6b9ea5a6
RP		 5388 }
5389
5390cleanup:
5391 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
8d1c802b
DA		 5392 if (nh->fib6_info)
5393 fib6_info_release(nh->fib6_info);
6b9ea5a6
RP		 5394 list_del(&nh->next);
5395 kfree(nh);
5396 }
5397
5398 return err;
5399}
5400
333c4301
DA		 5401static int ip6_route_multipath_del(struct fib6_config *cfg,
5402 struct netlink_ext_ack *extack)
6b9ea5a6
RP		 5403{
5404 struct fib6_config r_cfg;
5405 struct rtnexthop *rtnh;
2291267e 5406 int last_err = 0;
6b9ea5a6
RP		 5407 int remaining;
5408 int attrlen;
2291267e 5409 int err;
6b9ea5a6
RP		 5410
5411 remaining = cfg->fc_mp_len;
5412 rtnh = (struct rtnexthop *)cfg->fc_mp;
5413
5414 /* Parse a Multipath Entry */
5415 while (rtnh_ok(rtnh, remaining)) {
5416 memcpy(&r_cfg, cfg, sizeof(*cfg));
5417 if (rtnh->rtnh_ifindex)
5418 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5419
5420 attrlen = rtnh_attrlen(rtnh);
5421 if (attrlen > 0) {
5422 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5423
5424 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5425 if (nla) {
5426 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
5427 r_cfg.fc_flags |= RTF_GATEWAY;
5428 }
5429 }
333c4301 5430 err = ip6_route_del(&r_cfg, extack);
6b9ea5a6
RP		 5431 if (err)
5432 last_err = err;
5433
51ebd318
ND		 5434 rtnh = rtnh_next(rtnh, &remaining);
5435 }
5436
5437 return last_err;
5438}
5439
c21ef3e3
DA		 5440static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5441 struct netlink_ext_ack *extack)
1da177e4 5442{
86872cb5
TG		 5443 struct fib6_config cfg;
5444 int err;
1da177e4 5445
333c4301 5446 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 5447 if (err < 0)
5448 return err;
5449
5b98324e
DA		 5450 if (cfg.fc_nh_id &&
5451 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5452 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5453 return -EINVAL;
5454 }
5455
51ebd318 5456 if (cfg.fc_mp)
333c4301 5457 return ip6_route_multipath_del(&cfg, extack);
0ae81335
DA		 5458 else {
5459 cfg.fc_delete_all_nh = 1;
333c4301 5460 return ip6_route_del(&cfg, extack);
0ae81335 5461 }
1da177e4
LT		 5462}
5463
c21ef3e3
DA		 5464static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5465 struct netlink_ext_ack *extack)
1da177e4 5466{
86872cb5
TG		 5467 struct fib6_config cfg;
5468 int err;
1da177e4 5469
333c4301 5470 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 5471 if (err < 0)
5472 return err;
5473
67f69513
DA		 5474 if (cfg.fc_metric == 0)
5475 cfg.fc_metric = IP6_RT_PRIO_USER;
5476
51ebd318 5477 if (cfg.fc_mp)
333c4301 5478 return ip6_route_multipath_add(&cfg, extack);
51ebd318 5479 else
acb54e3c 5480 return ip6_route_add(&cfg, GFP_KERNEL, extack);
1da177e4
LT		 5481}
5482
a1b7a1f0
DA		 5483/* add the overhead of this fib6_nh to nexthop_len */
5484static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
339bf98f 5485{
a1b7a1f0 5486 int *nexthop_len = arg;
beb1afac 5487
a1b7a1f0
DA		 5488 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5489 + NLA_ALIGN(sizeof(struct rtnexthop))
5490 + nla_total_size(16); /* RTA_GATEWAY */
5491
5492 if (nh->fib_nh_lws) {
5493 /* RTA_ENCAP_TYPE */
5494 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5495 /* RTA_ENCAP */
5496 *nexthop_len += nla_total_size(2);
5497 }
f88d8ea6 5498
a1b7a1f0
DA		 5499 return 0;
5500}
beb1afac 5501
a1b7a1f0
DA		 5502static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5503{
5504 int nexthop_len;
5505
5506 if (f6i->nh) {
5507 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5508 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5509 &nexthop_len);
5510 } else {
5511 struct fib6_nh *nh = f6i->fib6_nh;
5512
5513 nexthop_len = 0;
5514 if (f6i->fib6_nsiblings) {
5515 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
5516 + NLA_ALIGN(sizeof(struct rtnexthop))
5517 + nla_total_size(16) /* RTA_GATEWAY */
5518 + lwtunnel_get_encap_size(nh->fib_nh_lws);
5519
5520 nexthop_len *= f6i->fib6_nsiblings;
5521 }
5522 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
beb1afac
DA		 5523 }
5524
339bf98f
TG		 5525 return NLMSG_ALIGN(sizeof(struct rtmsg))
5526 + nla_total_size(16) /* RTA_SRC */
5527 + nla_total_size(16) /* RTA_DST */
5528 + nla_total_size(16) /* RTA_GATEWAY */
5529 + nla_total_size(16) /* RTA_PREFSRC */
5530 + nla_total_size(4) /* RTA_TABLE */
5531 + nla_total_size(4) /* RTA_IIF */
5532 + nla_total_size(4) /* RTA_OIF */
5533 + nla_total_size(4) /* RTA_PRIORITY */
6a2b9ce0 5534 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
ea697639 5535 + nla_total_size(sizeof(struct rta_cacheinfo))
c78ba6d6 5536 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
19e42e45 5537 + nla_total_size(1) /* RTA_PREF */
beb1afac
DA		 5538 + nexthop_len;
5539}
5540
f88d8ea6
DA		 5541static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5542 unsigned char *flags)
5543{
5544 if (nexthop_is_multipath(nh)) {
5545 struct nlattr *mp;
5546
4255ff05 5547 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
f88d8ea6
DA		 5548 if (!mp)
5549 goto nla_put_failure;
5550
7bdf4de1 5551 if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
f88d8ea6
DA		 5552 goto nla_put_failure;
5553
5554 nla_nest_end(skb, mp);
5555 } else {
5556 struct fib6_nh *fib6_nh;
5557
5558 fib6_nh = nexthop_fib6_nh(nh);
7bdf4de1 5559 if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
f88d8ea6
DA		 5560 flags, false) < 0)
5561 goto nla_put_failure;
5562 }
5563
5564 return 0;
5565
5566nla_put_failure:
5567 return -EMSGSIZE;
5568}
5569
d4ead6b3 5570static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 5571 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 5572 struct in6_addr *dest, struct in6_addr *src,
15e47304 5573 int iif, int type, u32 portid, u32 seq,
f8cfe2ce 5574 unsigned int flags)
1da177e4 5575{
22d0bd82
XL		 5576 struct rt6_info *rt6 = (struct rt6_info *)dst;
5577 struct rt6key *rt6_dst, *rt6_src;
5578 u32 *pmetrics, table, rt6_flags;
f88d8ea6 5579 unsigned char nh_flags = 0;
2d7202bf 5580 struct nlmsghdr *nlh;
22d0bd82 5581 struct rtmsg *rtm;
d4ead6b3 5582 long expires = 0;
1da177e4 5583
15e47304 5584 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
38308473 5585 if (!nlh)
26932566 5586 return -EMSGSIZE;
2d7202bf 5587
22d0bd82
XL		 5588 if (rt6) {
5589 rt6_dst = &rt6->rt6i_dst;
5590 rt6_src = &rt6->rt6i_src;
5591 rt6_flags = rt6->rt6i_flags;
5592 } else {
5593 rt6_dst = &rt->fib6_dst;
5594 rt6_src = &rt->fib6_src;
5595 rt6_flags = rt->fib6_flags;
5596 }
5597
2d7202bf 5598 rtm = nlmsg_data(nlh);
1da177e4 5599 rtm->rtm_family = AF_INET6;
22d0bd82
XL		 5600 rtm->rtm_dst_len = rt6_dst->plen;
5601 rtm->rtm_src_len = rt6_src->plen;
1da177e4 5602 rtm->rtm_tos = 0;
93c2fb25
DA		 5603 if (rt->fib6_table)
5604 table = rt->fib6_table->tb6_id;
c71099ac 5605 else
9e762a4a 5606 table = RT6_TABLE_UNSPEC;
97f0082a 5607 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
c78679e8
DM		 5608 if (nla_put_u32(skb, RTA_TABLE, table))
5609 goto nla_put_failure;
e8478e80
DA		 5610
5611 rtm->rtm_type = rt->fib6_type;
1da177e4
LT		 5612 rtm->rtm_flags = 0;
5613 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
93c2fb25 5614 rtm->rtm_protocol = rt->fib6_protocol;
1da177e4 5615
22d0bd82 5616 if (rt6_flags & RTF_CACHE)
1da177e4
LT		 5617 rtm->rtm_flags |= RTM_F_CLONED;
5618
d4ead6b3
DA		 5619 if (dest) {
5620 if (nla_put_in6_addr(skb, RTA_DST, dest))
c78679e8 5621 goto nla_put_failure;
1ab1457c 5622 rtm->rtm_dst_len = 128;
1da177e4 5623 } else if (rtm->rtm_dst_len)
22d0bd82 5624 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
c78679e8 5625 goto nla_put_failure;
1da177e4
LT		 5626#ifdef CONFIG_IPV6_SUBTREES
5627 if (src) {
930345ea 5628 if (nla_put_in6_addr(skb, RTA_SRC, src))
c78679e8 5629 goto nla_put_failure;
1ab1457c 5630 rtm->rtm_src_len = 128;
c78679e8 5631 } else if (rtm->rtm_src_len &&
22d0bd82 5632 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
c78679e8 5633 goto nla_put_failure;
1da177e4 5634#endif
7bc570c8
YH		 5635 if (iif) {
5636#ifdef CONFIG_IPV6_MROUTE
22d0bd82 5637 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
fd61c6ba
DA		 5638 int err = ip6mr_get_route(net, skb, rtm, portid);
5639
5640 if (err == 0)
5641 return 0;
5642 if (err < 0)
5643 goto nla_put_failure;
7bc570c8
YH		 5644 } else
5645#endif
c78679e8
DM		 5646 if (nla_put_u32(skb, RTA_IIF, iif))
5647 goto nla_put_failure;
d4ead6b3 5648 } else if (dest) {
1da177e4 5649 struct in6_addr saddr_buf;
d4ead6b3 5650 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
930345ea 5651 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 5652 goto nla_put_failure;
1da177e4 5653 }
2d7202bf 5654
93c2fb25 5655 if (rt->fib6_prefsrc.plen) {
c3968a85 5656 struct in6_addr saddr_buf;
93c2fb25 5657 saddr_buf = rt->fib6_prefsrc.addr;
930345ea 5658 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 5659 goto nla_put_failure;
c3968a85
DW		 5660 }
5661
d4ead6b3
DA		 5662 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5663 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
2d7202bf
TG		 5664 goto nla_put_failure;
5665
93c2fb25 5666 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
c78679e8 5667 goto nla_put_failure;
8253947e 5668
beb1afac
DA		 5669 /* For multipath routes, walk the siblings list and add
5670 * each as a nexthop within RTA_MULTIPATH.
5671 */
22d0bd82
XL		 5672 if (rt6) {
5673 if (rt6_flags & RTF_GATEWAY &&
5674 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5675 goto nla_put_failure;
5676
5677 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5678 goto nla_put_failure;
6b13d8f7
OH		 5679
5680 if (dst->lwtstate &&
5681 lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
5682 goto nla_put_failure;
22d0bd82 5683 } else if (rt->fib6_nsiblings) {
8d1c802b 5684 struct fib6_info *sibling, *next_sibling;
beb1afac
DA		 5685 struct nlattr *mp;
5686
ae0be8de 5687 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
beb1afac
DA		 5688 if (!mp)
5689 goto nla_put_failure;
5690
1cf844c7 5691 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
7bdf4de1 5692 rt->fib6_nh->fib_nh_weight, AF_INET6) < 0)
beb1afac
DA		 5693 goto nla_put_failure;
5694
5695 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25 5696 &rt->fib6_siblings, fib6_siblings) {
1cf844c7 5697 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
7bdf4de1
DS		 5698 sibling->fib6_nh->fib_nh_weight,
5699 AF_INET6) < 0)
beb1afac
DA		 5700 goto nla_put_failure;
5701 }
5702
5703 nla_nest_end(skb, mp);
f88d8ea6
DA		 5704 } else if (rt->nh) {
5705 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5706 goto nla_put_failure;
ecc5663c 5707
f88d8ea6
DA		 5708 if (nexthop_is_blackhole(rt->nh))
5709 rtm->rtm_type = RTN_BLACKHOLE;
5710
4f80116d
RP		 5711 if (net->ipv4.sysctl_nexthop_compat_mode &&
5712 rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
f88d8ea6
DA		 5713 goto nla_put_failure;
5714
5715 rtm->rtm_flags |= nh_flags;
5716 } else {
7bdf4de1 5717 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
ecc5663c 5718 &nh_flags, false) < 0)
beb1afac 5719 goto nla_put_failure;
ecc5663c
DA		 5720
5721 rtm->rtm_flags |= nh_flags;
beb1afac
DA		 5722 }
5723
22d0bd82 5724 if (rt6_flags & RTF_EXPIRES) {
14895687
DA		 5725 expires = dst ? dst->expires : rt->expires;
5726 expires -= jiffies;
5727 }
69cdf8f9 5728
bb3c4ab9
IS		 5729 if (!dst) {
5730 if (rt->offload)
5731 rtm->rtm_flags |= RTM_F_OFFLOAD;
5732 if (rt->trap)
5733 rtm->rtm_flags |= RTM_F_TRAP;
0c5fcf9e
AC		 5734 if (rt->offload_failed)
5735 rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
bb3c4ab9
IS		 5736 }
5737
d4ead6b3 5738 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
e3703b3d 5739 goto nla_put_failure;
2d7202bf 5740
22d0bd82 5741 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
c78ba6d6
LR		 5742 goto nla_put_failure;
5743
19e42e45 5744
053c095a
JB		 5745 nlmsg_end(skb, nlh);
5746 return 0;
2d7202bf
TG		 5747
5748nla_put_failure:
26932566
PM		 5749 nlmsg_cancel(skb, nlh);
5750 return -EMSGSIZE;
1da177e4
LT		 5751}
5752
2c170e07
DA		 5753static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5754{
5755 const struct net_device *dev = arg;
5756
5757 if (nh->fib_nh_dev == dev)
5758 return 1;
5759
5760 return 0;
5761}
5762
13e38901
DA		 5763static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5764 const struct net_device *dev)
5765{
2c170e07
DA		 5766 if (f6i->nh) {
5767 struct net_device *_dev = (struct net_device *)dev;
5768
5769 return !!nexthop_for_each_fib6_nh(f6i->nh,
5770 fib6_info_nh_uses_dev,
5771 _dev);
5772 }
5773
1cf844c7 5774 if (f6i->fib6_nh->fib_nh_dev == dev)
13e38901
DA		 5775 return true;
5776
5777 if (f6i->fib6_nsiblings) {
5778 struct fib6_info *sibling, *next_sibling;
5779
5780 list_for_each_entry_safe(sibling, next_sibling,
5781 &f6i->fib6_siblings, fib6_siblings) {
1cf844c7 5782 if (sibling->fib6_nh->fib_nh_dev == dev)
13e38901
DA		 5783 return true;
5784 }
5785 }
5786
5787 return false;
5788}
5789
1e47b483
SB		 5790struct fib6_nh_exception_dump_walker {
5791 struct rt6_rtnl_dump_arg *dump;
5792 struct fib6_info *rt;
5793 unsigned int flags;
5794 unsigned int skip;
5795 unsigned int count;
5796};
5797
5798static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5799{
5800 struct fib6_nh_exception_dump_walker *w = arg;
5801 struct rt6_rtnl_dump_arg *dump = w->dump;
5802 struct rt6_exception_bucket *bucket;
5803 struct rt6_exception *rt6_ex;
5804 int i, err;
5805
5806 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5807 if (!bucket)
5808 return 0;
5809
5810 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5811 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5812 if (w->skip) {
5813 w->skip--;
5814 continue;
5815 }
5816
5817 /* Expiration of entries doesn't bump sernum, insertion
5818 * does. Removal is triggered by insertion, so we can
5819 * rely on the fact that if entries change between two
5820 * partial dumps, this node is scanned again completely,
5821 * see rt6_insert_exception() and fib6_dump_table().
5822 *
5823 * Count expired entries we go through as handled
5824 * entries that we'll skip next time, in case of partial
5825 * node dump. Otherwise, if entries expire meanwhile,
5826 * we'll skip the wrong amount.
5827 */
5828 if (rt6_check_expired(rt6_ex->rt6i)) {
5829 w->count++;
5830 continue;
5831 }
5832
5833 err = rt6_fill_node(dump->net, dump->skb, w->rt,
5834 &rt6_ex->rt6i->dst, NULL, NULL, 0,
5835 RTM_NEWROUTE,
5836 NETLINK_CB(dump->cb->skb).portid,
5837 dump->cb->nlh->nlmsg_seq, w->flags);
5838 if (err)
5839 return err;
5840
5841 w->count++;
5842 }
5843 bucket++;
5844 }
5845
5846 return 0;
5847}
5848
bf9a8a06 5849/* Return -1 if done with node, number of handled routes on partial dump */
1e47b483 5850int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
1da177e4
LT		 5851{
5852 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
13e38901
DA		 5853 struct fib_dump_filter *filter = &arg->filter;
5854 unsigned int flags = NLM_F_MULTI;
1f17e2f2 5855 struct net *net = arg->net;
1e47b483 5856 int count = 0;
1f17e2f2 5857
421842ed 5858 if (rt == net->ipv6.fib6_null_entry)
bf9a8a06 5859 return -1;
1da177e4 5860
13e38901
DA		 5861 if ((filter->flags & RTM_F_PREFIX) &&
5862 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5863 /* success since this is not a prefix route */
bf9a8a06 5864 return -1;
13e38901 5865 }
1e47b483
SB		 5866 if (filter->filter_set &&
5867 ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5868 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5869 (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5870 return -1;
5871 }
5872
5873 if (filter->filter_set ||
5874 !filter->dump_routes || !filter->dump_exceptions) {
13e38901 5875 flags |= NLM_F_DUMP_FILTERED;
f8cfe2ce 5876 }
1da177e4 5877
1e47b483
SB		 5878 if (filter->dump_routes) {
5879 if (skip) {
5880 skip--;
5881 } else {
5882 if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5883 0, RTM_NEWROUTE,
5884 NETLINK_CB(arg->cb->skb).portid,
5885 arg->cb->nlh->nlmsg_seq, flags)) {
5886 return 0;
5887 }
5888 count++;
5889 }
5890 }
5891
5892 if (filter->dump_exceptions) {
5893 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5894 .rt = rt,
5895 .flags = flags,
5896 .skip = skip,
5897 .count = 0 };
5898 int err;
5899
3b525691 5900 rcu_read_lock();
1e47b483
SB		 5901 if (rt->nh) {
5902 err = nexthop_for_each_fib6_nh(rt->nh,
5903 rt6_nh_dump_exceptions,
5904 &w);
5905 } else {
5906 err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5907 }
3b525691 5908 rcu_read_unlock();
1e47b483
SB		 5909
5910 if (err)
5911 return count += w.count;
5912 }
bf9a8a06
SB		 5913
5914 return -1;
1da177e4
LT		 5915}
5916
0eff0a27
JK		 5917static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5918 const struct nlmsghdr *nlh,
5919 struct nlattr **tb,
5920 struct netlink_ext_ack *extack)
5921{
5922 struct rtmsg *rtm;
5923 int i, err;
5924
5925 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5926 NL_SET_ERR_MSG_MOD(extack,
5927 "Invalid header for get route request");
5928 return -EINVAL;
5929 }
5930
5931 if (!netlink_strict_get_check(skb))
8cb08174
JB		 5932 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5933 rtm_ipv6_policy, extack);
0eff0a27
JK		 5934
5935 rtm = nlmsg_data(nlh);
5936 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5937 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5938 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5939 rtm->rtm_type) {
5940 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5941 return -EINVAL;
5942 }
5943 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5944 NL_SET_ERR_MSG_MOD(extack,
5945 "Invalid flags for get route request");
5946 return -EINVAL;
5947 }
5948
8cb08174
JB		 5949 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5950 rtm_ipv6_policy, extack);
0eff0a27
JK		 5951 if (err)
5952 return err;
5953
5954 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5955 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5956 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5957 return -EINVAL;
5958 }
5959
5960 for (i = 0; i <= RTA_MAX; i++) {
5961 if (!tb[i])
5962 continue;
5963
5964 switch (i) {
5965 case RTA_SRC:
5966 case RTA_DST:
5967 case RTA_IIF:
5968 case RTA_OIF:
5969 case RTA_MARK:
5970 case RTA_UID:
5971 case RTA_SPORT:
5972 case RTA_DPORT:
5973 case RTA_IP_PROTO:
5974 break;
5975 default:
5976 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
5977 return -EINVAL;
5978 }
5979 }
5980
5981 return 0;
5982}
5983
c21ef3e3
DA		 5984static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5985 struct netlink_ext_ack *extack)
1da177e4 5986{
3b1e0a65 5987 struct net *net = sock_net(in_skb->sk);
ab364a6f 5988 struct nlattr *tb[RTA_MAX+1];
18c3a61c 5989 int err, iif = 0, oif = 0;
a68886a6 5990 struct fib6_info *from;
18c3a61c 5991 struct dst_entry *dst;
ab364a6f 5992 struct rt6_info *rt;
1da177e4 5993 struct sk_buff *skb;
ab364a6f 5994 struct rtmsg *rtm;
744486d4 5995 struct flowi6 fl6 = {};
18c3a61c 5996 bool fibmatch;
1da177e4 5997
0eff0a27 5998 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
ab364a6f
TG		 5999 if (err < 0)
6000 goto errout;
1da177e4 6001
ab364a6f 6002 err = -EINVAL;
38b7097b
HFS		 6003 rtm = nlmsg_data(nlh);
6004 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
18c3a61c 6005 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
1da177e4 6006
ab364a6f
TG		 6007 if (tb[RTA_SRC]) {
6008 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
6009 goto errout;
6010
4e3fd7a0 6011 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
ab364a6f
TG		 6012 }
6013
6014 if (tb[RTA_DST]) {
6015 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
6016 goto errout;
6017
4e3fd7a0 6018 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
ab364a6f
TG		 6019 }
6020
6021 if (tb[RTA_IIF])
6022 iif = nla_get_u32(tb[RTA_IIF]);
6023
6024 if (tb[RTA_OIF])
72331bc0 6025 oif = nla_get_u32(tb[RTA_OIF]);
1da177e4 6026
2e47b291
LC		 6027 if (tb[RTA_MARK])
6028 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
6029
622ec2c9
LC		 6030 if (tb[RTA_UID])
6031 fl6.flowi6_uid = make_kuid(current_user_ns(),
6032 nla_get_u32(tb[RTA_UID]));
6033 else
6034 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
6035
eacb9384
RP		 6036 if (tb[RTA_SPORT])
6037 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
6038
6039 if (tb[RTA_DPORT])
6040 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
6041
6042 if (tb[RTA_IP_PROTO]) {
6043 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
5e1a99ea
HL		 6044 &fl6.flowi6_proto, AF_INET6,
6045 extack);
eacb9384
RP		 6046 if (err)
6047 goto errout;
6048 }
6049
1da177e4
LT		 6050 if (iif) {
6051 struct net_device *dev;
72331bc0
SL		 6052 int flags = 0;
6053
121622db
FW		 6054 rcu_read_lock();
6055
6056 dev = dev_get_by_index_rcu(net, iif);
1da177e4 6057 if (!dev) {
121622db 6058 rcu_read_unlock();
1da177e4 6059 err = -ENODEV;
ab364a6f 6060 goto errout;
1da177e4 6061 }
72331bc0
SL		 6062
6063 fl6.flowi6_iif = iif;
6064
6065 if (!ipv6_addr_any(&fl6.saddr))
6066 flags |= RT6_LOOKUP_F_HAS_SADDR;
6067
b75cc8f9 6068 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
121622db
FW		 6069
6070 rcu_read_unlock();
72331bc0
SL		 6071 } else {
6072 fl6.flowi6_oif = oif;
6073
58acfd71 6074 dst = ip6_route_output(net, NULL, &fl6);
18c3a61c
RP		 6075 }
6076
18c3a61c
RP		 6077
6078 rt = container_of(dst, struct rt6_info, dst);
6079 if (rt->dst.error) {
6080 err = rt->dst.error;
6081 ip6_rt_put(rt);
6082 goto errout;
1da177e4
LT		 6083 }
6084
9d6acb3b
WC		 6085 if (rt == net->ipv6.ip6_null_entry) {
6086 err = rt->dst.error;
6087 ip6_rt_put(rt);
6088 goto errout;
6089 }
6090
ab364a6f 6091 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
38308473 6092 if (!skb) {
94e187c0 6093 ip6_rt_put(rt);
ab364a6f
TG		 6094 err = -ENOBUFS;
6095 goto errout;
6096 }
1da177e4 6097
d8d1f30b 6098 skb_dst_set(skb, &rt->dst);
a68886a6
DA		 6099
6100 rcu_read_lock();
6101 from = rcu_dereference(rt->from);
886b7a50
MKL		 6102 if (from) {
6103 if (fibmatch)
6104 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
6105 iif, RTM_NEWROUTE,
6106 NETLINK_CB(in_skb).portid,
6107 nlh->nlmsg_seq, 0);
6108 else
6109 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
6110 &fl6.saddr, iif, RTM_NEWROUTE,
6111 NETLINK_CB(in_skb).portid,
6112 nlh->nlmsg_seq, 0);
6113 } else {
6114 err = -ENETUNREACH;
6115 }
a68886a6
DA		 6116 rcu_read_unlock();
6117
1da177e4 6118 if (err < 0) {
ab364a6f
TG		 6119 kfree_skb(skb);
6120 goto errout;
1da177e4
LT		 6121 }
6122
15e47304 6123 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
ab364a6f 6124errout:
1da177e4 6125 return err;
1da177e4
LT		 6126}
6127
8d1c802b 6128void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
37a1d361 6129 unsigned int nlm_flags)
1da177e4
LT		 6130{
6131 struct sk_buff *skb;
5578689a 6132 struct net *net = info->nl_net;
528c4ceb
DL		 6133 u32 seq;
6134 int err;
6135
6136 err = -ENOBUFS;
38308473 6137 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
86872cb5 6138
19e42e45 6139 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
38308473 6140 if (!skb)
21713ebc
TG		 6141 goto errout;
6142
d4ead6b3
DA		 6143 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6144 event, info->portid, seq, nlm_flags);
26932566
PM		 6145 if (err < 0) {
6146 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6147 WARN_ON(err == -EMSGSIZE);
6148 kfree_skb(skb);
6149 goto errout;
6150 }
15e47304 6151 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
1ce85fe4
PNA		 6152 info->nlh, gfp_any());
6153 return;
21713ebc
TG		 6154errout:
6155 if (err < 0)
5578689a 6156 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
1da177e4
LT		 6157}
6158
19a3b7ee
DA		 6159void fib6_rt_update(struct net *net, struct fib6_info *rt,
6160 struct nl_info *info)
6161{
6162 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6163 struct sk_buff *skb;
6164 int err = -ENOBUFS;
6165
19a3b7ee
DA		 6166 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6167 if (!skb)
6168 goto errout;
6169
6170 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6171 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6172 if (err < 0) {
6173 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6174 WARN_ON(err == -EMSGSIZE);
6175 kfree_skb(skb);
6176 goto errout;
6177 }
6178 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6179 info->nlh, gfp_any());
6180 return;
6181errout:
6182 if (err < 0)
6183 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6184}
6185
907eea48 6186void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
0c5fcf9e 6187 bool offload, bool trap, bool offload_failed)
907eea48
AC		 6188{
6189 struct sk_buff *skb;
6190 int err;
6191
0c5fcf9e
AC		 6192 if (f6i->offload == offload && f6i->trap == trap &&
6193 f6i->offload_failed == offload_failed)
907eea48
AC		 6194 return;
6195
6196 f6i->offload = offload;
6197 f6i->trap = trap;
6fad361a
AC		 6198
6199 /* 2 means send notifications only if offload_failed was changed. */
6200 if (net->ipv6.sysctl.fib_notify_on_flag_change == 2 &&
6201 f6i->offload_failed == offload_failed)
6202 return;
6203
0c5fcf9e 6204 f6i->offload_failed = offload_failed;
907eea48
AC		 6205
6206 if (!rcu_access_pointer(f6i->fib6_node))
6207 /* The route was removed from the tree, do not send
89e8347f 6208 * notification.
907eea48
AC		 6209 */
6210 return;
6211
6212 if (!net->ipv6.sysctl.fib_notify_on_flag_change)
6213 return;
6214
6215 skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
6216 if (!skb) {
6217 err = -ENOBUFS;
6218 goto errout;
6219 }
6220
6221 err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
6222 0, 0);
6223 if (err < 0) {
6224 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6225 WARN_ON(err == -EMSGSIZE);
6226 kfree_skb(skb);
6227 goto errout;
6228 }
6229
6230 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
6231 return;
6232
6233errout:
6234 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6235}
6236EXPORT_SYMBOL(fib6_info_hw_flags_set);
6237
8ed67789 6238static int ip6_route_dev_notify(struct notifier_block *this,
351638e7 6239 unsigned long event, void *ptr)
8ed67789 6240{
351638e7 6241 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
c346dca1 6242 struct net *net = dev_net(dev);
8ed67789 6243
242d3a49
WC		 6244 if (!(dev->flags & IFF_LOOPBACK))
6245 return NOTIFY_OK;
6246
6247 if (event == NETDEV_REGISTER) {
1cf844c7 6248 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
d8d1f30b 6249 net->ipv6.ip6_null_entry->dst.dev = dev;
8ed67789
DL		 6250 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6251#ifdef CONFIG_IPV6_MULTIPLE_TABLES
d8d1f30b 6252 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
8ed67789 6253 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
d8d1f30b 6254 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
8ed67789 6255 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
242d3a49 6256#endif
76da0704
WC		 6257 } else if (event == NETDEV_UNREGISTER &&
6258 dev->reg_state != NETREG_UNREGISTERED) {
6259 /* NETDEV_UNREGISTER could be fired for multiple times by
6260 * netdev_wait_allrefs(). Make sure we only call this once.
6261 */
12d94a80 6262 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
242d3a49 6263#ifdef CONFIG_IPV6_MULTIPLE_TABLES
12d94a80
ED		 6264 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6265 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
8ed67789
DL		 6266#endif
6267 }
6268
6269 return NOTIFY_OK;
6270}
6271
1da177e4
LT		 6272/*
6273 * /proc
6274 */
6275
6276#ifdef CONFIG_PROC_FS
1da177e4
LT		 6277static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6278{
69ddb805 6279 struct net *net = (struct net *)seq->private;
1da177e4 6280 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
69ddb805
DL		 6281 net->ipv6.rt6_stats->fib_nodes,
6282 net->ipv6.rt6_stats->fib_route_nodes,
81eb8447 6283 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
69ddb805
DL		 6284 net->ipv6.rt6_stats->fib_rt_entries,
6285 net->ipv6.rt6_stats->fib_rt_cache,
fc66f95c 6286 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
69ddb805 6287 net->ipv6.rt6_stats->fib_discarded_routes);
1da177e4
LT		 6288
6289 return 0;
6290}
1da177e4
LT		 6291#endif /* CONFIG_PROC_FS */
6292
6293#ifdef CONFIG_SYSCTL
6294
32927393
CH		 6295static int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
6296 void *buffer, size_t *lenp, loff_t *ppos)
1da177e4 6297{
c486da34
LAG		 6298 struct net *net;
6299 int delay;
f0fb9b28 6300 int ret;
c486da34 6301 if (!write)
1da177e4 6302 return -EINVAL;
c486da34
LAG		 6303
6304 net = (struct net *)ctl->extra1;
6305 delay = net->ipv6.sysctl.flush_delay;
f0fb9b28
AP		 6306 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6307 if (ret)
6308 return ret;
6309
2ac3ac8f 6310 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
c486da34 6311 return 0;
1da177e4
LT		 6312}
6313
ed792e28 6314static struct ctl_table ipv6_route_table_template[] = {
1ab1457c 6315 {
1da177e4 6316 .procname = "flush",
4990509f 6317 .data = &init_net.ipv6.sysctl.flush_delay,
1da177e4 6318 .maxlen = sizeof(int),
89c8b3a1 6319 .mode = 0200,
6d9f239a 6320 .proc_handler = ipv6_sysctl_rtcache_flush
1da177e4
LT		 6321 },
6322 {
1da177e4 6323 .procname = "gc_thresh",
9a7ec3a9 6324 .data = &ip6_dst_ops_template.gc_thresh,
1da177e4
LT		 6325 .maxlen = sizeof(int),
6326 .mode = 0644,
6d9f239a 6327 .proc_handler = proc_dointvec,
1da177e4
LT		 6328 },
6329 {
1da177e4 6330 .procname = "max_size",
4990509f 6331 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
1da177e4
LT		 6332 .maxlen = sizeof(int),
6333 .mode = 0644,
6d9f239a 6334 .proc_handler = proc_dointvec,
1da177e4
LT		 6335 },
6336 {
1da177e4 6337 .procname = "gc_min_interval",
4990509f 6338 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 6339 .maxlen = sizeof(int),
6340 .mode = 0644,
6d9f239a 6341 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6342 },
6343 {
1da177e4 6344 .procname = "gc_timeout",
4990509f 6345 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
1da177e4
LT		 6346 .maxlen = sizeof(int),
6347 .mode = 0644,
6d9f239a 6348 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6349 },
6350 {
1da177e4 6351 .procname = "gc_interval",
4990509f 6352 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
1da177e4
LT		 6353 .maxlen = sizeof(int),
6354 .mode = 0644,
6d9f239a 6355 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6356 },
6357 {
1da177e4 6358 .procname = "gc_elasticity",
4990509f 6359 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
1da177e4
LT		 6360 .maxlen = sizeof(int),
6361 .mode = 0644,
f3d3f616 6362 .proc_handler = proc_dointvec,
1da177e4
LT		 6363 },
6364 {
1da177e4 6365 .procname = "mtu_expires",
4990509f 6366 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
1da177e4
LT		 6367 .maxlen = sizeof(int),
6368 .mode = 0644,
6d9f239a 6369 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6370 },
6371 {
1da177e4 6372 .procname = "min_adv_mss",
4990509f 6373 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
1da177e4
LT		 6374 .maxlen = sizeof(int),
6375 .mode = 0644,
f3d3f616 6376 .proc_handler = proc_dointvec,
1da177e4
LT		 6377 },
6378 {
1da177e4 6379 .procname = "gc_min_interval_ms",
4990509f 6380 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 6381 .maxlen = sizeof(int),
6382 .mode = 0644,
6d9f239a 6383 .proc_handler = proc_dointvec_ms_jiffies,
1da177e4 6384 },
7c6bb7d2
DA		 6385 {
6386 .procname = "skip_notify_on_dev_down",
6387 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6388 .maxlen = sizeof(int),
6389 .mode = 0644,
b8e8a863 6390 .proc_handler = proc_dointvec_minmax,
eec4844f
MC		 6391 .extra1 = SYSCTL_ZERO,
6392 .extra2 = SYSCTL_ONE,
7c6bb7d2 6393 },
f8572d8f 6394 { }
1da177e4
LT		 6395};
6396
2c8c1e72 6397struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
760f2d01
DL		 6398{
6399 struct ctl_table *table;
6400
6401 table = kmemdup(ipv6_route_table_template,
6402 sizeof(ipv6_route_table_template),
6403 GFP_KERNEL);
5ee09105
YH		 6404
6405 if (table) {
6406 table[0].data = &net->ipv6.sysctl.flush_delay;
c486da34 6407 table[0].extra1 = net;
86393e52 6408 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
5ee09105
YH		 6409 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
6410 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6411 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6412 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6413 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6414 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6415 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
9c69fabe 6416 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
7c6bb7d2 6417 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
464dc801
EB		 6418
6419 /* Don't export sysctls to unprivileged users */
6420 if (net->user_ns != &init_user_ns)
6421 table[0].procname = NULL;
5ee09105
YH		 6422 }
6423
760f2d01
DL		 6424 return table;
6425}
1da177e4
LT		 6426#endif
6427
2c8c1e72 6428static int __net_init ip6_route_net_init(struct net *net)
cdb18761 6429{
633d424b 6430 int ret = -ENOMEM;
8ed67789 6431
86393e52
AD		 6432 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6433 sizeof(net->ipv6.ip6_dst_ops));
f2fc6a54 6434
fc66f95c
ED		 6435 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6436 goto out_ip6_dst_ops;
6437
1cf844c7 6438 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
421842ed
DA		 6439 if (!net->ipv6.fib6_null_entry)
6440 goto out_ip6_dst_entries;
1cf844c7
DA		 6441 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6442 sizeof(*net->ipv6.fib6_null_entry));
421842ed 6443
8ed67789
DL		 6444 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6445 sizeof(*net->ipv6.ip6_null_entry),
6446 GFP_KERNEL);
6447 if (!net->ipv6.ip6_null_entry)
421842ed 6448 goto out_fib6_null_entry;
d8d1f30b 6449 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6450 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6451 ip6_template_metrics, true);
74109218 6452 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->rt6i_uncached);
8ed67789
DL		 6453
6454#ifdef CONFIG_IPV6_MULTIPLE_TABLES
feca7d8c 6455 net->ipv6.fib6_has_custom_rules = false;
8ed67789
DL		 6456 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6457 sizeof(*net->ipv6.ip6_prohibit_entry),
6458 GFP_KERNEL);
68fffc67
PZ		 6459 if (!net->ipv6.ip6_prohibit_entry)
6460 goto out_ip6_null_entry;
d8d1f30b 6461 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6462 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6463 ip6_template_metrics, true);
74109218 6464 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->rt6i_uncached);
8ed67789
DL		 6465
6466 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6467 sizeof(*net->ipv6.ip6_blk_hole_entry),
6468 GFP_KERNEL);
68fffc67
PZ		 6469 if (!net->ipv6.ip6_blk_hole_entry)
6470 goto out_ip6_prohibit_entry;
d8d1f30b 6471 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6472 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6473 ip6_template_metrics, true);
74109218 6474 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->rt6i_uncached);
b9b33e7c
PA		 6475#ifdef CONFIG_IPV6_SUBTREES
6476 net->ipv6.fib6_routes_require_src = 0;
6477#endif
8ed67789
DL		 6478#endif
6479
b339a47c
PZ		 6480 net->ipv6.sysctl.flush_delay = 0;
6481 net->ipv6.sysctl.ip6_rt_max_size = 4096;
6482 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6483 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6484 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6485 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6486 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6487 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
7c6bb7d2 6488 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
b339a47c 6489
6891a346
BT		 6490 net->ipv6.ip6_rt_gc_expire = 30*HZ;
6491
8ed67789
DL		 6492 ret = 0;
6493out:
6494 return ret;
f2fc6a54 6495
68fffc67
PZ		 6496#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6497out_ip6_prohibit_entry:
6498 kfree(net->ipv6.ip6_prohibit_entry);
6499out_ip6_null_entry:
6500 kfree(net->ipv6.ip6_null_entry);
6501#endif
421842ed
DA		 6502out_fib6_null_entry:
6503 kfree(net->ipv6.fib6_null_entry);
fc66f95c
ED		 6504out_ip6_dst_entries:
6505 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
f2fc6a54 6506out_ip6_dst_ops:
f2fc6a54 6507 goto out;
cdb18761
DL		 6508}
6509
2c8c1e72 6510static void __net_exit ip6_route_net_exit(struct net *net)
cdb18761 6511{
421842ed 6512 kfree(net->ipv6.fib6_null_entry);
8ed67789
DL		 6513 kfree(net->ipv6.ip6_null_entry);
6514#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6515 kfree(net->ipv6.ip6_prohibit_entry);
6516 kfree(net->ipv6.ip6_blk_hole_entry);
6517#endif
41bb78b4 6518 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
cdb18761
DL		 6519}
6520
d189634e
TG		 6521static int __net_init ip6_route_net_init_late(struct net *net)
6522{
6523#ifdef CONFIG_PROC_FS
c3506372
CH		 6524 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6525 sizeof(struct ipv6_route_iter));
3617d949
CH		 6526 proc_create_net_single("rt6_stats", 0444, net->proc_net,
6527 rt6_stats_seq_show, NULL);
d189634e
TG		 6528#endif
6529 return 0;
6530}
6531
6532static void __net_exit ip6_route_net_exit_late(struct net *net)
6533{
6534#ifdef CONFIG_PROC_FS
ece31ffd
G		 6535 remove_proc_entry("ipv6_route", net->proc_net);
6536 remove_proc_entry("rt6_stats", net->proc_net);
d189634e
TG		 6537#endif
6538}
6539
cdb18761
DL		 6540static struct pernet_operations ip6_route_net_ops = {
6541 .init = ip6_route_net_init,
6542 .exit = ip6_route_net_exit,
6543};
6544
c3426b47
DM		 6545static int __net_init ipv6_inetpeer_init(struct net *net)
6546{
6547 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6548
6549 if (!bp)
6550 return -ENOMEM;
6551 inet_peer_base_init(bp);
6552 net->ipv6.peers = bp;
6553 return 0;
6554}
6555
6556static void __net_exit ipv6_inetpeer_exit(struct net *net)
6557{
6558 struct inet_peer_base *bp = net->ipv6.peers;
6559
6560 net->ipv6.peers = NULL;
56a6b248 6561 inetpeer_invalidate_tree(bp);
c3426b47
DM		 6562 kfree(bp);
6563}
6564
2b823f72 6565static struct pernet_operations ipv6_inetpeer_ops = {
c3426b47
DM		 6566 .init = ipv6_inetpeer_init,
6567 .exit = ipv6_inetpeer_exit,
6568};
6569
d189634e
TG		 6570static struct pernet_operations ip6_route_net_late_ops = {
6571 .init = ip6_route_net_init_late,
6572 .exit = ip6_route_net_exit_late,
6573};
6574
8ed67789
DL		 6575static struct notifier_block ip6_route_dev_notifier = {
6576 .notifier_call = ip6_route_dev_notify,
242d3a49 6577 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
8ed67789
DL		 6578};
6579
2f460933
WC		 6580void __init ip6_route_init_special_entries(void)
6581{
6582 /* Registering of the loopback is done before this portion of code,
6583 * the loopback reference in rt6_info will not be taken, do it
6584 * manually for init_net */
1cf844c7 6585 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
2f460933
WC		 6586 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6587 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6588 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6589 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6590 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6591 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6592 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6593 #endif
6594}
6595
138d0be3
YS		 6596#if IS_BUILTIN(CONFIG_IPV6)
6597#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6598DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6599
951cf368
YS		 6600BTF_ID_LIST(btf_fib6_info_id)
6601BTF_ID(struct, fib6_info)
6602
14fc6bd6 6603static const struct bpf_iter_seq_info ipv6_route_seq_info = {
15172a46
YS		 6604 .seq_ops = &ipv6_route_seq_ops,
6605 .init_seq_private = bpf_iter_init_seq_net,
6606 .fini_seq_private = bpf_iter_fini_seq_net,
6607 .seq_priv_size = sizeof(struct ipv6_route_iter),
14fc6bd6
YS		 6608};
6609
6610static struct bpf_iter_reg ipv6_route_reg_info = {
6611 .target = "ipv6_route",
3c32cc1b
YS		 6612 .ctx_arg_info_size = 1,
6613 .ctx_arg_info = {
6614 { offsetof(struct bpf_iter__ipv6_route, rt),
6615 PTR_TO_BTF_ID_OR_NULL },
6616 },
14fc6bd6 6617 .seq_info = &ipv6_route_seq_info,
15172a46
YS		 6618};
6619
138d0be3
YS		 6620static int __init bpf_iter_register(void)
6621{
951cf368 6622 ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id;
15172a46 6623 return bpf_iter_reg_target(&ipv6_route_reg_info);
138d0be3
YS		 6624}
6625
6626static void bpf_iter_unregister(void)
6627{
ab2ee4fc 6628 bpf_iter_unreg_target(&ipv6_route_reg_info);
138d0be3
YS		 6629}
6630#endif
6631#endif
6632
433d49c3 6633int __init ip6_route_init(void)
1da177e4 6634{
433d49c3 6635 int ret;
8d0b94af 6636 int cpu;
433d49c3 6637
9a7ec3a9
DL		 6638 ret = -ENOMEM;
6639 ip6_dst_ops_template.kmem_cachep =
e5d679f3 6640 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
f845ab6b 6641 SLAB_HWCACHE_ALIGN, NULL);
9a7ec3a9 6642 if (!ip6_dst_ops_template.kmem_cachep)
c19a28e1 6643 goto out;
14e50e57 6644
fc66f95c 6645 ret = dst_entries_init(&ip6_dst_blackhole_ops);
8ed67789 6646 if (ret)
bdb3289f 6647 goto out_kmem_cache;
bdb3289f 6648
c3426b47
DM		 6649 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6650 if (ret)
e8803b6c 6651 goto out_dst_entries;
2a0c451a 6652
7e52b33b
DM		 6653 ret = register_pernet_subsys(&ip6_route_net_ops);
6654 if (ret)
6655 goto out_register_inetpeer;
c3426b47 6656
5dc121e9
AE		 6657 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6658
e8803b6c 6659 ret = fib6_init();
433d49c3 6660 if (ret)
8ed67789 6661 goto out_register_subsys;
433d49c3 6662
433d49c3
DL		 6663 ret = xfrm6_init();
6664 if (ret)
e8803b6c 6665 goto out_fib6_init;
c35b7e72 6666
433d49c3
DL		 6667 ret = fib6_rules_init();
6668 if (ret)
6669 goto xfrm6_init;
7e5449c2 6670
d189634e
TG		 6671 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6672 if (ret)
6673 goto fib6_rules_init;
6674
16feebcf
FW		 6675 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6676 inet6_rtm_newroute, NULL, 0);
6677 if (ret < 0)
6678 goto out_register_late_subsys;
6679
6680 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6681 inet6_rtm_delroute, NULL, 0);
6682 if (ret < 0)
6683 goto out_register_late_subsys;
6684
6685 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6686 inet6_rtm_getroute, NULL,
6687 RTNL_FLAG_DOIT_UNLOCKED);
6688 if (ret < 0)
d189634e 6689 goto out_register_late_subsys;
c127ea2c 6690
8ed67789 6691 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
cdb18761 6692 if (ret)
d189634e 6693 goto out_register_late_subsys;
8ed67789 6694
138d0be3
YS		 6695#if IS_BUILTIN(CONFIG_IPV6)
6696#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6697 ret = bpf_iter_register();
6698 if (ret)
6699 goto out_register_late_subsys;
6700#endif
6701#endif
6702
8d0b94af
MKL		 6703 for_each_possible_cpu(cpu) {
6704 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6705
6706 INIT_LIST_HEAD(&ul->head);
6707 spin_lock_init(&ul->lock);
6708 }
6709
433d49c3
DL		 6710out:
6711 return ret;
6712
d189634e 6713out_register_late_subsys:
16feebcf 6714 rtnl_unregister_all(PF_INET6);
d189634e 6715 unregister_pernet_subsys(&ip6_route_net_late_ops);
433d49c3 6716fib6_rules_init:
433d49c3
DL		 6717 fib6_rules_cleanup();
6718xfrm6_init:
433d49c3 6719 xfrm6_fini();
2a0c451a
TG		 6720out_fib6_init:
6721 fib6_gc_cleanup();
8ed67789
DL		 6722out_register_subsys:
6723 unregister_pernet_subsys(&ip6_route_net_ops);
7e52b33b
DM		 6724out_register_inetpeer:
6725 unregister_pernet_subsys(&ipv6_inetpeer_ops);
fc66f95c
ED		 6726out_dst_entries:
6727 dst_entries_destroy(&ip6_dst_blackhole_ops);
433d49c3 6728out_kmem_cache:
f2fc6a54 6729 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
433d49c3 6730 goto out;
1da177e4
LT		 6731}
6732
6733void ip6_route_cleanup(void)
6734{
138d0be3
YS		 6735#if IS_BUILTIN(CONFIG_IPV6)
6736#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6737 bpf_iter_unregister();
6738#endif
6739#endif
8ed67789 6740 unregister_netdevice_notifier(&ip6_route_dev_notifier);
d189634e 6741 unregister_pernet_subsys(&ip6_route_net_late_ops);
101367c2 6742 fib6_rules_cleanup();
1da177e4 6743 xfrm6_fini();
1da177e4 6744 fib6_gc_cleanup();
c3426b47 6745 unregister_pernet_subsys(&ipv6_inetpeer_ops);
8ed67789 6746 unregister_pernet_subsys(&ip6_route_net_ops);
41bb78b4 6747 dst_entries_destroy(&ip6_dst_blackhole_ops);
f2fc6a54 6748 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
1da177e4 6749}
Linux 6.x block layer and io_uring tree(s)