projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
mlxsw: spectrum_router: Start using new IPv6 route notifications
[linux-block.git] / net / ipv6 / route.c
CommitLineData
2874c5fd 1// SPDX-License-Identifier: GPL-2.0-or-later
1da177e4
LT		 2/*
3 * Linux INET6 implementation
4 * FIB front-end.
5 *
6 * Authors:
1ab1457c 7 * Pedro Roque <roque@di.fc.ul.pt>
1da177e4
LT		 8 */
9
10/* Changes:
11 *
12 * YOSHIFUJI Hideaki @USAGI
13 * reworked default router selection.
14 * - respect outgoing interface
15 * - select from (probably) reachable routers (i.e.
16 * routers in REACHABLE, STALE, DELAY or PROBE states).
17 * - always select the same router if it is (probably)
18 * reachable. otherwise, round-robin the list.
c0bece9f
YH		 19 * Ville Nuorvala
20 * Fixed routing subtrees.
1da177e4
LT		 21 */
22
f3213831
JP		 23#define pr_fmt(fmt) "IPv6: " fmt
24
4fc268d2 25#include <linux/capability.h>
1da177e4 26#include <linux/errno.h>
bc3b2d7f 27#include <linux/export.h>
1da177e4
LT		 28#include <linux/types.h>
29#include <linux/times.h>
30#include <linux/socket.h>
31#include <linux/sockios.h>
32#include <linux/net.h>
33#include <linux/route.h>
34#include <linux/netdevice.h>
35#include <linux/in6.h>
7bc570c8 36#include <linux/mroute6.h>
1da177e4 37#include <linux/init.h>
1da177e4 38#include <linux/if_arp.h>
1da177e4
LT		 39#include <linux/proc_fs.h>
40#include <linux/seq_file.h>
5b7c931d 41#include <linux/nsproxy.h>
5a0e3ad6 42#include <linux/slab.h>
35732d01 43#include <linux/jhash.h>
457c4cbc 44#include <net/net_namespace.h>
1da177e4
LT		 45#include <net/snmp.h>
46#include <net/ipv6.h>
47#include <net/ip6_fib.h>
48#include <net/ip6_route.h>
49#include <net/ndisc.h>
50#include <net/addrconf.h>
51#include <net/tcp.h>
52#include <linux/rtnetlink.h>
53#include <net/dst.h>
904af04d 54#include <net/dst_metadata.h>
1da177e4 55#include <net/xfrm.h>
8d71740c 56#include <net/netevent.h>
21713ebc 57#include <net/netlink.h>
3c618c1d 58#include <net/rtnh.h>
19e42e45 59#include <net/lwtunnel.h>
904af04d 60#include <net/ip_tunnels.h>
ca254490 61#include <net/l3mdev.h>
eacb9384 62#include <net/ip.h>
7c0f6ba6 63#include <linux/uaccess.h>
1da177e4
LT		 64
65#ifdef CONFIG_SYSCTL
66#include <linux/sysctl.h>
67#endif
68
30d444d3
DA		 69static int ip6_rt_type_to_error(u8 fib6_type);
70
71#define CREATE_TRACE_POINTS
72#include <trace/events/fib6.h>
73EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
74#undef CREATE_TRACE_POINTS
75
afc154e9 76enum rt6_nud_state {
7e980569
JB		 77 RT6_NUD_FAIL_HARD = -3,
78 RT6_NUD_FAIL_PROBE = -2,
79 RT6_NUD_FAIL_DO_RR = -1,
afc154e9
HFS		 80 RT6_NUD_SUCCEED = 1
81};
82
1da177e4 83static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
0dbaee3b 84static unsigned int ip6_default_advmss(const struct dst_entry *dst);
ebb762f2 85static unsigned int ip6_mtu(const struct dst_entry *dst);
1da177e4
LT		 86static struct dst_entry *ip6_negative_advice(struct dst_entry *);
87static void ip6_dst_destroy(struct dst_entry *);
88static void ip6_dst_ifdown(struct dst_entry *,
89 struct net_device *dev, int how);
569d3645 90static int ip6_dst_gc(struct dst_ops *ops);
1da177e4
LT		 91
92static int ip6_pkt_discard(struct sk_buff *skb);
ede2059d 93static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
7150aede 94static int ip6_pkt_prohibit(struct sk_buff *skb);
ede2059d 95static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1da177e4 96static void ip6_link_failure(struct sk_buff *skb);
6700c270
DM		 97static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
98 struct sk_buff *skb, u32 mtu);
99static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
100 struct sk_buff *skb);
702cea56
DA		 101static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
102 int strict);
a1b7a1f0 103static size_t rt6_nlmsg_size(struct fib6_info *f6i);
d4ead6b3 104static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 105 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 106 struct in6_addr *dest, struct in6_addr *src,
16a16cd3
DA		 107 int iif, int type, u32 portid, u32 seq,
108 unsigned int flags);
7e4b5128 109static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
510e2ced
WW		 110 const struct in6_addr *daddr,
111 const struct in6_addr *saddr);
1da177e4 112
70ceb4f5 113#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 114static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 115 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 116 const struct in6_addr *gwaddr,
117 struct net_device *dev,
95c96174 118 unsigned int pref);
8d1c802b 119static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 120 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 121 const struct in6_addr *gwaddr,
122 struct net_device *dev);
70ceb4f5
YH		 123#endif
124
8d0b94af
MKL		 125struct uncached_list {
126 spinlock_t lock;
127 struct list_head head;
128};
129
130static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
131
510c321b 132void rt6_uncached_list_add(struct rt6_info *rt)
8d0b94af
MKL		 133{
134 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
135
8d0b94af
MKL		 136 rt->rt6i_uncached_list = ul;
137
138 spin_lock_bh(&ul->lock);
139 list_add_tail(&rt->rt6i_uncached, &ul->head);
140 spin_unlock_bh(&ul->lock);
141}
142
510c321b 143void rt6_uncached_list_del(struct rt6_info *rt)
8d0b94af
MKL		 144{
145 if (!list_empty(&rt->rt6i_uncached)) {
146 struct uncached_list *ul = rt->rt6i_uncached_list;
81eb8447 147 struct net *net = dev_net(rt->dst.dev);
8d0b94af
MKL		 148
149 spin_lock_bh(&ul->lock);
150 list_del(&rt->rt6i_uncached);
81eb8447 151 atomic_dec(&net->ipv6.rt6_stats->fib_rt_uncache);
8d0b94af
MKL		 152 spin_unlock_bh(&ul->lock);
153 }
154}
155
156static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
157{
158 struct net_device *loopback_dev = net->loopback_dev;
159 int cpu;
160
e332bc67
EB		 161 if (dev == loopback_dev)
162 return;
163
8d0b94af
MKL		 164 for_each_possible_cpu(cpu) {
165 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
166 struct rt6_info *rt;
167
168 spin_lock_bh(&ul->lock);
169 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
170 struct inet6_dev *rt_idev = rt->rt6i_idev;
171 struct net_device *rt_dev = rt->dst.dev;
172
e332bc67 173 if (rt_idev->dev == dev) {
8d0b94af
MKL		 174 rt->rt6i_idev = in6_dev_get(loopback_dev);
175 in6_dev_put(rt_idev);
176 }
177
e332bc67 178 if (rt_dev == dev) {
8d7017fd 179 rt->dst.dev = blackhole_netdev;
8d0b94af
MKL		 180 dev_hold(rt->dst.dev);
181 dev_put(rt_dev);
182 }
183 }
184 spin_unlock_bh(&ul->lock);
185 }
186}
187
f8a1b43b 188static inline const void *choose_neigh_daddr(const struct in6_addr *p,
f894cbf8
DM		 189 struct sk_buff *skb,
190 const void *daddr)
39232973 191{
a7563f34 192 if (!ipv6_addr_any(p))
39232973 193 return (const void *) p;
f894cbf8
DM		 194 else if (skb)
195 return &ipv6_hdr(skb)->daddr;
39232973
DM		 196 return daddr;
197}
198
f8a1b43b
DA		 199struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
200 struct net_device *dev,
201 struct sk_buff *skb,
202 const void *daddr)
d3aaeb38 203{
39232973
DM		 204 struct neighbour *n;
205
f8a1b43b
DA		 206 daddr = choose_neigh_daddr(gw, skb, daddr);
207 n = __ipv6_neigh_lookup(dev, daddr);
f83c7790
DM		 208 if (n)
209 return n;
7adf3246
SB		 210
211 n = neigh_create(&nd_tbl, daddr, dev);
212 return IS_ERR(n) ? NULL : n;
f8a1b43b
DA		 213}
214
215static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
216 struct sk_buff *skb,
217 const void *daddr)
218{
219 const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
220
2c6b55f4
ND		 221 return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
222 dst->dev, skb, daddr);
f83c7790
DM		 223}
224
63fca65d
JA		 225static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
226{
227 struct net_device *dev = dst->dev;
228 struct rt6_info *rt = (struct rt6_info *)dst;
229
cbfd6891 230 daddr = choose_neigh_daddr(rt6_nexthop(rt, &in6addr_any), NULL, daddr);
63fca65d
JA		 231 if (!daddr)
232 return;
233 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
234 return;
235 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
236 return;
237 __ipv6_confirm_neigh(dev, daddr);
238}
239
9a7ec3a9 240static struct dst_ops ip6_dst_ops_template = {
1da177e4 241 .family = AF_INET6,
1da177e4
LT		 242 .gc = ip6_dst_gc,
243 .gc_thresh = 1024,
244 .check = ip6_dst_check,
0dbaee3b 245 .default_advmss = ip6_default_advmss,
ebb762f2 246 .mtu = ip6_mtu,
d4ead6b3 247 .cow_metrics = dst_cow_metrics_generic,
1da177e4
LT		 248 .destroy = ip6_dst_destroy,
249 .ifdown = ip6_dst_ifdown,
250 .negative_advice = ip6_negative_advice,
251 .link_failure = ip6_link_failure,
252 .update_pmtu = ip6_rt_update_pmtu,
6e157b6a 253 .redirect = rt6_do_redirect,
9f8955cc 254 .local_out = __ip6_local_out,
f8a1b43b 255 .neigh_lookup = ip6_dst_neigh_lookup,
63fca65d 256 .confirm_neigh = ip6_confirm_neigh,
1da177e4
LT		 257};
258
ebb762f2 259static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
ec831ea7 260{
618f9bc7
SK		 261 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
262
263 return mtu ? : dst->dev->mtu;
ec831ea7
RD		 264}
265
6700c270
DM		 266static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
267 struct sk_buff *skb, u32 mtu)
14e50e57
DM		 268{
269}
270
6700c270
DM		 271static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
272 struct sk_buff *skb)
b587ee3b
DM		 273{
274}
275
14e50e57
DM		 276static struct dst_ops ip6_dst_blackhole_ops = {
277 .family = AF_INET6,
14e50e57
DM		 278 .destroy = ip6_dst_destroy,
279 .check = ip6_dst_check,
ebb762f2 280 .mtu = ip6_blackhole_mtu,
214f45c9 281 .default_advmss = ip6_default_advmss,
14e50e57 282 .update_pmtu = ip6_rt_blackhole_update_pmtu,
b587ee3b 283 .redirect = ip6_rt_blackhole_redirect,
0a1f5962 284 .cow_metrics = dst_cow_metrics_generic,
f8a1b43b 285 .neigh_lookup = ip6_dst_neigh_lookup,
14e50e57
DM		 286};
287
62fa8a84 288static const u32 ip6_template_metrics[RTAX_MAX] = {
14edd87d 289 [RTAX_HOPLIMIT - 1] = 0,
62fa8a84
DM		 290};
291
8d1c802b 292static const struct fib6_info fib6_null_entry_template = {
93c2fb25
DA		 293 .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
294 .fib6_protocol = RTPROT_KERNEL,
295 .fib6_metric = ~(u32)0,
f05713e0 296 .fib6_ref = REFCOUNT_INIT(1),
421842ed
DA		 297 .fib6_type = RTN_UNREACHABLE,
298 .fib6_metrics = (struct dst_metrics *)&dst_default_metrics,
299};
300
fb0af4c7 301static const struct rt6_info ip6_null_entry_template = {
d8d1f30b
CG		 302 .dst = {
303 .__refcnt = ATOMIC_INIT(1),
304 .__use = 1,
2c20cbd7 305 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 306 .error = -ENETUNREACH,
d8d1f30b
CG		 307 .input = ip6_pkt_discard,
308 .output = ip6_pkt_discard_out,
1da177e4
LT		 309 },
310 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
1da177e4
LT		 311};
312
101367c2
TG		 313#ifdef CONFIG_IPV6_MULTIPLE_TABLES
314
fb0af4c7 315static const struct rt6_info ip6_prohibit_entry_template = {
d8d1f30b
CG		 316 .dst = {
317 .__refcnt = ATOMIC_INIT(1),
318 .__use = 1,
2c20cbd7 319 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 320 .error = -EACCES,
d8d1f30b
CG		 321 .input = ip6_pkt_prohibit,
322 .output = ip6_pkt_prohibit_out,
101367c2
TG		 323 },
324 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 325};
326
fb0af4c7 327static const struct rt6_info ip6_blk_hole_entry_template = {
d8d1f30b
CG		 328 .dst = {
329 .__refcnt = ATOMIC_INIT(1),
330 .__use = 1,
2c20cbd7 331 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 332 .error = -EINVAL,
d8d1f30b 333 .input = dst_discard,
ede2059d 334 .output = dst_discard_out,
101367c2
TG		 335 },
336 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 337};
338
339#endif
340
ebfa45f0
MKL		 341static void rt6_info_init(struct rt6_info *rt)
342{
343 struct dst_entry *dst = &rt->dst;
344
345 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
ebfa45f0
MKL		 346 INIT_LIST_HEAD(&rt->rt6i_uncached);
347}
348
1da177e4 349/* allocate dst with ip6_dst_ops */
93531c67
DA		 350struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
351 int flags)
1da177e4 352{
97bab73f 353 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
b2a9c0ed 354 1, DST_OBSOLETE_FORCE_CHK, flags);
cf911662 355
81eb8447 356 if (rt) {
ebfa45f0 357 rt6_info_init(rt);
81eb8447
WW		 358 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
359 }
8104891b 360
cf911662 361 return rt;
1da177e4 362}
9ab179d8 363EXPORT_SYMBOL(ip6_dst_alloc);
d52d3997 364
1da177e4
LT		 365static void ip6_dst_destroy(struct dst_entry *dst)
366{
367 struct rt6_info *rt = (struct rt6_info *)dst;
a68886a6 368 struct fib6_info *from;
8d0b94af 369 struct inet6_dev *idev;
1da177e4 370
1620a336 371 ip_dst_metrics_put(dst);
8d0b94af
MKL		 372 rt6_uncached_list_del(rt);
373
374 idev = rt->rt6i_idev;
38308473 375 if (idev) {
1da177e4
LT		 376 rt->rt6i_idev = NULL;
377 in6_dev_put(idev);
1ab1457c 378 }
1716a961 379
0e233874 380 from = xchg((__force struct fib6_info **)&rt->from, NULL);
93531c67 381 fib6_info_release(from);
b3419363
DM		 382}
383
1da177e4
LT		 384static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
385 int how)
386{
387 struct rt6_info *rt = (struct rt6_info *)dst;
388 struct inet6_dev *idev = rt->rt6i_idev;
5a3e55d6 389 struct net_device *loopback_dev =
c346dca1 390 dev_net(dev)->loopback_dev;
1da177e4 391
e5645f51
WW		 392 if (idev && idev->dev != loopback_dev) {
393 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
394 if (loopback_idev) {
395 rt->rt6i_idev = loopback_idev;
396 in6_dev_put(idev);
97cac082 397 }
1da177e4
LT		 398 }
399}
400
5973fb1e
MKL		 401static bool __rt6_check_expired(const struct rt6_info *rt)
402{
403 if (rt->rt6i_flags & RTF_EXPIRES)
404 return time_after(jiffies, rt->dst.expires);
405 else
406 return false;
407}
408
a50feda5 409static bool rt6_check_expired(const struct rt6_info *rt)
1da177e4 410{
a68886a6
DA		 411 struct fib6_info *from;
412
413 from = rcu_dereference(rt->from);
414
1716a961
G		 415 if (rt->rt6i_flags & RTF_EXPIRES) {
416 if (time_after(jiffies, rt->dst.expires))
a50feda5 417 return true;
a68886a6 418 } else if (from) {
1e2ea8ad 419 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
a68886a6 420 fib6_check_expired(from);
1716a961 421 }
a50feda5 422 return false;
1da177e4
LT		 423}
424
b1d40991
DA		 425void fib6_select_path(const struct net *net, struct fib6_result *res,
426 struct flowi6 *fl6, int oif, bool have_oif_match,
427 const struct sk_buff *skb, int strict)
51ebd318 428{
8d1c802b 429 struct fib6_info *sibling, *next_sibling;
b1d40991
DA		 430 struct fib6_info *match = res->f6i;
431
f88d8ea6 432 if ((!match->fib6_nsiblings && !match->nh) || have_oif_match)
b1d40991 433 goto out;
51ebd318 434
b673d6cc
JS		 435 /* We might have already computed the hash for ICMPv6 errors. In such
436 * case it will always be non-zero. Otherwise now is the time to do it.
437 */
f88d8ea6
DA		 438 if (!fl6->mp_hash &&
439 (!match->nh || nexthop_is_multipath(match->nh)))
b4bac172 440 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
b673d6cc 441
f88d8ea6
DA		 442 if (unlikely(match->nh)) {
443 nexthop_path_fib6_result(res, fl6->mp_hash);
444 return;
445 }
446
1cf844c7 447 if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
b1d40991 448 goto out;
3d709f69 449
93c2fb25
DA		 450 list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
451 fib6_siblings) {
1cf844c7 452 const struct fib6_nh *nh = sibling->fib6_nh;
5e670d84
DA		 453 int nh_upper_bound;
454
702cea56 455 nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
5e670d84 456 if (fl6->mp_hash > nh_upper_bound)
3d709f69 457 continue;
702cea56 458 if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
3d709f69
IS		 459 break;
460 match = sibling;
461 break;
462 }
463
b1d40991
DA		 464out:
465 res->f6i = match;
1cf844c7 466 res->nh = match->fib6_nh;
51ebd318
ND		 467}
468
1da177e4 469/*
66f5d6ce 470 * Route lookup. rcu_read_lock() should be held.
1da177e4
LT		 471 */
472
0c59d006
DA		 473static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
474 const struct in6_addr *saddr, int oif, int flags)
475{
476 const struct net_device *dev;
477
478 if (nh->fib_nh_flags & RTNH_F_DEAD)
479 return false;
480
481 dev = nh->fib_nh_dev;
482 if (oif) {
483 if (dev->ifindex == oif)
484 return true;
485 } else {
486 if (ipv6_chk_addr(net, saddr, dev,
487 flags & RT6_LOOKUP_F_IFACE))
488 return true;
489 }
490
491 return false;
492}
493
962b6803
DA		 494struct fib6_nh_dm_arg {
495 struct net *net;
496 const struct in6_addr *saddr;
497 int oif;
498 int flags;
499 struct fib6_nh *nh;
500};
501
502static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
503{
504 struct fib6_nh_dm_arg *arg = _arg;
505
506 arg->nh = nh;
507 return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
508 arg->flags);
509}
510
511/* returns fib6_nh from nexthop or NULL */
512static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
513 struct fib6_result *res,
514 const struct in6_addr *saddr,
515 int oif, int flags)
516{
517 struct fib6_nh_dm_arg arg = {
518 .net = net,
519 .saddr = saddr,
520 .oif = oif,
521 .flags = flags,
522 };
523
524 if (nexthop_is_blackhole(nh))
525 return NULL;
526
527 if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
528 return arg.nh;
529
530 return NULL;
531}
532
75ef7389
DA		 533static void rt6_device_match(struct net *net, struct fib6_result *res,
534 const struct in6_addr *saddr, int oif, int flags)
1da177e4 535{
75ef7389
DA		 536 struct fib6_info *f6i = res->f6i;
537 struct fib6_info *spf6i;
538 struct fib6_nh *nh;
1da177e4 539
75ef7389 540 if (!oif && ipv6_addr_any(saddr)) {
f88d8ea6
DA		 541 if (unlikely(f6i->nh)) {
542 nh = nexthop_fib6_nh(f6i->nh);
543 if (nexthop_is_blackhole(f6i->nh))
544 goto out_blackhole;
545 } else {
546 nh = f6i->fib6_nh;
547 }
7d21fec9
DA		 548 if (!(nh->fib_nh_flags & RTNH_F_DEAD))
549 goto out;
75ef7389 550 }
dd3abc4e 551
75ef7389 552 for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
962b6803
DA		 553 bool matched = false;
554
555 if (unlikely(spf6i->nh)) {
556 nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
557 oif, flags);
558 if (nh)
559 matched = true;
560 } else {
561 nh = spf6i->fib6_nh;
562 if (__rt6_device_match(net, nh, saddr, oif, flags))
563 matched = true;
564 }
565 if (matched) {
75ef7389 566 res->f6i = spf6i;
7d21fec9 567 goto out;
75ef7389 568 }
dd3abc4e 569 }
1da177e4 570
75ef7389
DA		 571 if (oif && flags & RT6_LOOKUP_F_IFACE) {
572 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 573 nh = res->f6i->fib6_nh;
7d21fec9 574 goto out;
75ef7389 575 }
8067bb8c 576
f88d8ea6
DA		 577 if (unlikely(f6i->nh)) {
578 nh = nexthop_fib6_nh(f6i->nh);
579 if (nexthop_is_blackhole(f6i->nh))
580 goto out_blackhole;
581 } else {
582 nh = f6i->fib6_nh;
583 }
584
7d21fec9 585 if (nh->fib_nh_flags & RTNH_F_DEAD) {
75ef7389 586 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 587 nh = res->f6i->fib6_nh;
75ef7389 588 }
7d21fec9
DA		 589out:
590 res->nh = nh;
591 res->fib6_type = res->f6i->fib6_type;
592 res->fib6_flags = res->f6i->fib6_flags;
f88d8ea6
DA		 593 return;
594
595out_blackhole:
596 res->fib6_flags |= RTF_REJECT;
597 res->fib6_type = RTN_BLACKHOLE;
598 res->nh = nh;
1da177e4
LT		 599}
600
27097255 601#ifdef CONFIG_IPV6_ROUTER_PREF
c2f17e82
HFS		 602struct __rt6_probe_work {
603 struct work_struct work;
604 struct in6_addr target;
605 struct net_device *dev;
606};
607
608static void rt6_probe_deferred(struct work_struct *w)
609{
610 struct in6_addr mcaddr;
611 struct __rt6_probe_work *work =
612 container_of(w, struct __rt6_probe_work, work);
613
614 addrconf_addr_solict_mult(&work->target, &mcaddr);
adc176c5 615 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
c2f17e82 616 dev_put(work->dev);
662f5533 617 kfree(work);
c2f17e82
HFS		 618}
619
cc3a86c8 620static void rt6_probe(struct fib6_nh *fib6_nh)
27097255 621{
f547fac6 622 struct __rt6_probe_work *work = NULL;
5e670d84 623 const struct in6_addr *nh_gw;
1bef4c22 624 unsigned long last_probe;
f2c31e32 625 struct neighbour *neigh;
5e670d84 626 struct net_device *dev;
f547fac6 627 struct inet6_dev *idev;
5e670d84 628
27097255
YH		 629 /*
630 * Okay, this does not seem to be appropriate
631 * for now, however, we need to check if it
632 * is really so; aka Router Reachability Probing.
633 *
634 * Router Reachability Probe MUST be rate-limited
635 * to no more than one per minute.
636 */
004b3942 637 if (!fib6_nh->fib_nh_gw_family)
7ff74a59 638 return;
5e670d84 639
cc3a86c8
DA		 640 nh_gw = &fib6_nh->fib_nh_gw6;
641 dev = fib6_nh->fib_nh_dev;
2152caea 642 rcu_read_lock_bh();
1bef4c22 643 last_probe = READ_ONCE(fib6_nh->last_probe);
f547fac6 644 idev = __in6_dev_get(dev);
5e670d84 645 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
2152caea 646 if (neigh) {
8d6c31bf
MKL		 647 if (neigh->nud_state & NUD_VALID)
648 goto out;
649
2152caea 650 write_lock(&neigh->lock);
990edb42
MKL		 651 if (!(neigh->nud_state & NUD_VALID) &&
652 time_after(jiffies,
dcd1f572 653 neigh->updated + idev->cnf.rtr_probe_interval)) {
990edb42
MKL		 654 work = kmalloc(sizeof(*work), GFP_ATOMIC);
655 if (work)
656 __neigh_set_probe_once(neigh);
c2f17e82 657 }
2152caea 658 write_unlock(&neigh->lock);
1bef4c22 659 } else if (time_after(jiffies, last_probe +
f547fac6 660 idev->cnf.rtr_probe_interval)) {
990edb42 661 work = kmalloc(sizeof(*work), GFP_ATOMIC);
f2c31e32 662 }
990edb42 663
1bef4c22
ED		 664 if (!work || cmpxchg(&fib6_nh->last_probe,
665 last_probe, jiffies) != last_probe) {
666 kfree(work);
667 } else {
990edb42 668 INIT_WORK(&work->work, rt6_probe_deferred);
5e670d84
DA		 669 work->target = *nh_gw;
670 dev_hold(dev);
671 work->dev = dev;
990edb42
MKL		 672 schedule_work(&work->work);
673 }
674
8d6c31bf 675out:
2152caea 676 rcu_read_unlock_bh();
27097255
YH		 677}
678#else
cc3a86c8 679static inline void rt6_probe(struct fib6_nh *fib6_nh)
27097255 680{
27097255
YH		 681}
682#endif
683
1da177e4 684/*
554cfb7e 685 * Default Router Selection (RFC 2461 6.3.6)
1da177e4 686 */
1ba9a895 687static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
1da177e4 688{
afc154e9 689 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
5e670d84 690 struct neighbour *neigh;
f2c31e32 691
145a3621 692 rcu_read_lock_bh();
1ba9a895
DA		 693 neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
694 &fib6_nh->fib_nh_gw6);
145a3621
YH		 695 if (neigh) {
696 read_lock(&neigh->lock);
554cfb7e 697 if (neigh->nud_state & NUD_VALID)
afc154e9 698 ret = RT6_NUD_SUCCEED;
398bcbeb 699#ifdef CONFIG_IPV6_ROUTER_PREF
a5a81f0b 700 else if (!(neigh->nud_state & NUD_FAILED))
afc154e9 701 ret = RT6_NUD_SUCCEED;
7e980569
JB		 702 else
703 ret = RT6_NUD_FAIL_PROBE;
398bcbeb 704#endif
145a3621 705 read_unlock(&neigh->lock);
afc154e9
HFS		 706 } else {
707 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
7e980569 708 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
a5a81f0b 709 }
145a3621
YH		 710 rcu_read_unlock_bh();
711
a5a81f0b 712 return ret;
1da177e4
LT		 713}
714
702cea56
DA		 715static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
716 int strict)
1da177e4 717{
6e1809a5
DA		 718 int m = 0;
719
720 if (!oif || nh->fib_nh_dev->ifindex == oif)
721 m = 2;
1ab1457c 722
77d16f45 723 if (!m && (strict & RT6_LOOKUP_F_IFACE))
afc154e9 724 return RT6_NUD_FAIL_HARD;
ebacaaa0 725#ifdef CONFIG_IPV6_ROUTER_PREF
702cea56 726 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
ebacaaa0 727#endif
1ba9a895 728 if ((strict & RT6_LOOKUP_F_REACHABLE) &&
702cea56 729 !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
1ba9a895 730 int n = rt6_check_neigh(nh);
afc154e9
HFS		 731 if (n < 0)
732 return n;
733 }
554cfb7e
YH		 734 return m;
735}
736
28679ed1
DA		 737static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
738 int oif, int strict, int *mpri, bool *do_rr)
554cfb7e 739{
afc154e9 740 bool match_do_rr = false;
28679ed1
DA		 741 bool rc = false;
742 int m;
35103d11 743
28679ed1 744 if (nh->fib_nh_flags & RTNH_F_DEAD)
8067bb8c
IS		 745 goto out;
746
28679ed1
DA		 747 if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
748 nh->fib_nh_flags & RTNH_F_LINKDOWN &&
d5d32e4b 749 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
35103d11 750 goto out;
f11e6659 751
28679ed1 752 m = rt6_score_route(nh, fib6_flags, oif, strict);
7e980569 753 if (m == RT6_NUD_FAIL_DO_RR) {
afc154e9
HFS		 754 match_do_rr = true;
755 m = 0; /* lowest valid score */
7e980569 756 } else if (m == RT6_NUD_FAIL_HARD) {
f11e6659 757 goto out;
afc154e9
HFS		 758 }
759
760 if (strict & RT6_LOOKUP_F_REACHABLE)
28679ed1 761 rt6_probe(nh);
f11e6659 762
7e980569 763 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
f11e6659 764 if (m > *mpri) {
afc154e9 765 *do_rr = match_do_rr;
f11e6659 766 *mpri = m;
28679ed1 767 rc = true;
f11e6659 768 }
f11e6659 769out:
28679ed1 770 return rc;
f11e6659
DM		 771}
772
17a5984e
DA		 773struct fib6_nh_frl_arg {
774 u32 flags;
775 int oif;
776 int strict;
777 int *mpri;
778 bool *do_rr;
779 struct fib6_nh *nh;
780};
781
782static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
783{
784 struct fib6_nh_frl_arg *arg = _arg;
785
786 arg->nh = nh;
787 return find_match(nh, arg->flags, arg->oif, arg->strict,
788 arg->mpri, arg->do_rr);
789}
790
b7bc4b6a 791static void __find_rr_leaf(struct fib6_info *f6i_start,
30c15f03 792 struct fib6_info *nomatch, u32 metric,
b7bc4b6a 793 struct fib6_result *res, struct fib6_info **cont,
30c15f03 794 int oif, int strict, bool *do_rr, int *mpri)
f11e6659 795{
b7bc4b6a 796 struct fib6_info *f6i;
1da177e4 797
b7bc4b6a
DA		 798 for (f6i = f6i_start;
799 f6i && f6i != nomatch;
800 f6i = rcu_dereference(f6i->fib6_next)) {
17a5984e 801 bool matched = false;
30c15f03
DA		 802 struct fib6_nh *nh;
803
b7bc4b6a
DA		 804 if (cont && f6i->fib6_metric != metric) {
805 *cont = f6i;
30c15f03 806 return;
9fbdcfaf
SK		 807 }
808
b7bc4b6a 809 if (fib6_check_expired(f6i))
28679ed1
DA		 810 continue;
811
17a5984e
DA		 812 if (unlikely(f6i->nh)) {
813 struct fib6_nh_frl_arg arg = {
814 .flags = f6i->fib6_flags,
815 .oif = oif,
816 .strict = strict,
817 .mpri = mpri,
818 .do_rr = do_rr
819 };
820
821 if (nexthop_is_blackhole(f6i->nh)) {
822 res->fib6_flags = RTF_REJECT;
823 res->fib6_type = RTN_BLACKHOLE;
824 res->f6i = f6i;
825 res->nh = nexthop_fib6_nh(f6i->nh);
826 return;
827 }
828 if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
829 &arg)) {
830 matched = true;
831 nh = arg.nh;
832 }
833 } else {
834 nh = f6i->fib6_nh;
835 if (find_match(nh, f6i->fib6_flags, oif, strict,
836 mpri, do_rr))
837 matched = true;
838 }
839 if (matched) {
b7bc4b6a
DA		 840 res->f6i = f6i;
841 res->nh = nh;
7d21fec9
DA		 842 res->fib6_flags = f6i->fib6_flags;
843 res->fib6_type = f6i->fib6_type;
b7bc4b6a 844 }
9fbdcfaf 845 }
30c15f03 846}
9fbdcfaf 847
b7bc4b6a
DA		 848static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
849 struct fib6_info *rr_head, int oif, int strict,
850 bool *do_rr, struct fib6_result *res)
30c15f03 851{
b7bc4b6a
DA		 852 u32 metric = rr_head->fib6_metric;
853 struct fib6_info *cont = NULL;
30c15f03 854 int mpri = -1;
9fbdcfaf 855
b7bc4b6a 856 __find_rr_leaf(rr_head, NULL, metric, res, &cont,
30c15f03 857 oif, strict, do_rr, &mpri);
28679ed1 858
b7bc4b6a 859 __find_rr_leaf(leaf, rr_head, metric, res, &cont,
30c15f03 860 oif, strict, do_rr, &mpri);
9fbdcfaf 861
b7bc4b6a
DA		 862 if (res->f6i || !cont)
863 return;
9fbdcfaf 864
b7bc4b6a 865 __find_rr_leaf(cont, NULL, metric, res, NULL,
30c15f03 866 oif, strict, do_rr, &mpri);
f11e6659 867}
1da177e4 868
b7bc4b6a
DA		 869static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
870 struct fib6_result *res, int strict)
f11e6659 871{
8d1c802b 872 struct fib6_info *leaf = rcu_dereference(fn->leaf);
b7bc4b6a 873 struct fib6_info *rt0;
afc154e9 874 bool do_rr = false;
17ecf590 875 int key_plen;
1da177e4 876
b7bc4b6a
DA		 877 /* make sure this function or its helpers sets f6i */
878 res->f6i = NULL;
879
421842ed 880 if (!leaf || leaf == net->ipv6.fib6_null_entry)
b7bc4b6a 881 goto out;
8d1040e8 882
66f5d6ce 883 rt0 = rcu_dereference(fn->rr_ptr);
f11e6659 884 if (!rt0)
66f5d6ce 885 rt0 = leaf;
1da177e4 886
17ecf590
WW		 887 /* Double check to make sure fn is not an intermediate node
888 * and fn->leaf does not points to its child's leaf
889 * (This might happen if all routes under fn are deleted from
890 * the tree and fib6_repair_tree() is called on the node.)
891 */
93c2fb25 892 key_plen = rt0->fib6_dst.plen;
17ecf590 893#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 894 if (rt0->fib6_src.plen)
895 key_plen = rt0->fib6_src.plen;
17ecf590
WW		 896#endif
897 if (fn->fn_bit != key_plen)
b7bc4b6a 898 goto out;
1da177e4 899
b7bc4b6a 900 find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
afc154e9 901 if (do_rr) {
8fb11a9a 902 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
f11e6659 903
554cfb7e 904 /* no entries matched; do round-robin */
93c2fb25 905 if (!next || next->fib6_metric != rt0->fib6_metric)
8d1040e8 906 next = leaf;
f11e6659 907
66f5d6ce 908 if (next != rt0) {
93c2fb25 909 spin_lock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 910 /* make sure next is not being deleted from the tree */
93c2fb25 911 if (next->fib6_node)
66f5d6ce 912 rcu_assign_pointer(fn->rr_ptr, next);
93c2fb25 913 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 914 }
1da177e4 915 }
1da177e4 916
b7bc4b6a
DA		 917out:
918 if (!res->f6i) {
919 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 920 res->nh = res->f6i->fib6_nh;
7d21fec9
DA		 921 res->fib6_flags = res->f6i->fib6_flags;
922 res->fib6_type = res->f6i->fib6_type;
b7bc4b6a 923 }
1da177e4
LT		 924}
925
85bd05de 926static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
8b9df265 927{
85bd05de
DA		 928 return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
929 res->nh->fib_nh_gw_family;
8b9df265
MKL		 930}
931
70ceb4f5
YH		 932#ifdef CONFIG_IPV6_ROUTE_INFO
933int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
b71d1d42 934 const struct in6_addr *gwaddr)
70ceb4f5 935{
c346dca1 936 struct net *net = dev_net(dev);
70ceb4f5
YH		 937 struct route_info *rinfo = (struct route_info *) opt;
938 struct in6_addr prefix_buf, *prefix;
939 unsigned int pref;
4bed72e4 940 unsigned long lifetime;
8d1c802b 941 struct fib6_info *rt;
70ceb4f5
YH		 942
943 if (len < sizeof(struct route_info)) {
944 return -EINVAL;
945 }
946
947 /* Sanity check for prefix_len and length */
948 if (rinfo->length > 3) {
949 return -EINVAL;
950 } else if (rinfo->prefix_len > 128) {
951 return -EINVAL;
952 } else if (rinfo->prefix_len > 64) {
953 if (rinfo->length < 2) {
954 return -EINVAL;
955 }
956 } else if (rinfo->prefix_len > 0) {
957 if (rinfo->length < 1) {
958 return -EINVAL;
959 }
960 }
961
962 pref = rinfo->route_pref;
963 if (pref == ICMPV6_ROUTER_PREF_INVALID)
3933fc95 964 return -EINVAL;
70ceb4f5 965
4bed72e4 966 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
70ceb4f5
YH		 967
968 if (rinfo->length == 3)
969 prefix = (struct in6_addr *)rinfo->prefix;
970 else {
971 /* this function is safe */
972 ipv6_addr_prefix(&prefix_buf,
973 (struct in6_addr *)rinfo->prefix,
974 rinfo->prefix_len);
975 prefix = &prefix_buf;
976 }
977
f104a567 978 if (rinfo->prefix_len == 0)
afb1d4b5 979 rt = rt6_get_dflt_router(net, gwaddr, dev);
f104a567
DJ		 980 else
981 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
830218c1 982 gwaddr, dev);
70ceb4f5
YH		 983
984 if (rt && !lifetime) {
afb1d4b5 985 ip6_del_rt(net, rt);
70ceb4f5
YH		 986 rt = NULL;
987 }
988
989 if (!rt && lifetime)
830218c1
DA		 990 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
991 dev, pref);
70ceb4f5 992 else if (rt)
93c2fb25
DA		 993 rt->fib6_flags = RTF_ROUTEINFO |
994 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
70ceb4f5
YH		 995
996 if (rt) {
1716a961 997 if (!addrconf_finite_timeout(lifetime))
14895687 998 fib6_clean_expires(rt);
1716a961 999 else
14895687 1000 fib6_set_expires(rt, jiffies + HZ * lifetime);
1716a961 1001
93531c67 1002 fib6_info_release(rt);
70ceb4f5
YH		 1003 }
1004 return 0;
1005}
1006#endif
1007
ae90d867
DA		 1008/*
1009 * Misc support functions
1010 */
1011
1012/* called with rcu_lock held */
0d161581 1013static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
ae90d867 1014{
0d161581 1015 struct net_device *dev = res->nh->fib_nh_dev;
ae90d867 1016
7d21fec9 1017 if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
ae90d867
DA		 1018 /* for copies of local routes, dst->dev needs to be the
1019 * device if it is a master device, the master device if
1020 * device is enslaved, and the loopback as the default
1021 */
1022 if (netif_is_l3_slave(dev) &&
7d21fec9 1023 !rt6_need_strict(&res->f6i->fib6_dst.addr))
ae90d867
DA		 1024 dev = l3mdev_master_dev_rcu(dev);
1025 else if (!netif_is_l3_master(dev))
1026 dev = dev_net(dev)->loopback_dev;
1027 /* last case is netif_is_l3_master(dev) is true in which
1028 * case we want dev returned to be dev
1029 */
1030 }
1031
1032 return dev;
1033}
1034
6edb3c96
DA		 1035static const int fib6_prop[RTN_MAX + 1] = {
1036 [RTN_UNSPEC] = 0,
1037 [RTN_UNICAST] = 0,
1038 [RTN_LOCAL] = 0,
1039 [RTN_BROADCAST] = 0,
1040 [RTN_ANYCAST] = 0,
1041 [RTN_MULTICAST] = 0,
1042 [RTN_BLACKHOLE] = -EINVAL,
1043 [RTN_UNREACHABLE] = -EHOSTUNREACH,
1044 [RTN_PROHIBIT] = -EACCES,
1045 [RTN_THROW] = -EAGAIN,
1046 [RTN_NAT] = -EINVAL,
1047 [RTN_XRESOLVE] = -EINVAL,
1048};
1049
1050static int ip6_rt_type_to_error(u8 fib6_type)
1051{
1052 return fib6_prop[fib6_type];
1053}
1054
8d1c802b 1055static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
3b6761d1
DA		 1056{
1057 unsigned short flags = 0;
1058
1059 if (rt->dst_nocount)
1060 flags |= DST_NOCOUNT;
1061 if (rt->dst_nopolicy)
1062 flags |= DST_NOPOLICY;
1063 if (rt->dst_host)
1064 flags |= DST_HOST;
1065
1066 return flags;
1067}
1068
7d21fec9 1069static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
6edb3c96 1070{
7d21fec9 1071 rt->dst.error = ip6_rt_type_to_error(fib6_type);
6edb3c96 1072
7d21fec9 1073 switch (fib6_type) {
6edb3c96
DA		 1074 case RTN_BLACKHOLE:
1075 rt->dst.output = dst_discard_out;
1076 rt->dst.input = dst_discard;
1077 break;
1078 case RTN_PROHIBIT:
1079 rt->dst.output = ip6_pkt_prohibit_out;
1080 rt->dst.input = ip6_pkt_prohibit;
1081 break;
1082 case RTN_THROW:
1083 case RTN_UNREACHABLE:
1084 default:
1085 rt->dst.output = ip6_pkt_discard_out;
1086 rt->dst.input = ip6_pkt_discard;
1087 break;
1088 }
1089}
1090
0d161581 1091static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
6edb3c96 1092{
7d21fec9 1093 struct fib6_info *f6i = res->f6i;
0d161581 1094
7d21fec9
DA		 1095 if (res->fib6_flags & RTF_REJECT) {
1096 ip6_rt_init_dst_reject(rt, res->fib6_type);
6edb3c96
DA		 1097 return;
1098 }
1099
1100 rt->dst.error = 0;
1101 rt->dst.output = ip6_output;
1102
7d21fec9 1103 if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
6edb3c96 1104 rt->dst.input = ip6_input;
7d21fec9 1105 } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
6edb3c96
DA		 1106 rt->dst.input = ip6_mc_input;
1107 } else {
1108 rt->dst.input = ip6_forward;
1109 }
1110
0d161581
DA		 1111 if (res->nh->fib_nh_lws) {
1112 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
6edb3c96
DA		 1113 lwtunnel_set_redirect(&rt->dst);
1114 }
1115
1116 rt->dst.lastuse = jiffies;
1117}
1118
e873e4b9 1119/* Caller must already hold reference to @from */
8d1c802b 1120static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
ae90d867 1121{
ae90d867 1122 rt->rt6i_flags &= ~RTF_EXPIRES;
a68886a6 1123 rcu_assign_pointer(rt->from, from);
e1255ed4 1124 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
ae90d867
DA		 1125}
1126
0d161581
DA		 1127/* Caller must already hold reference to f6i in result */
1128static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
ae90d867 1129{
0d161581
DA		 1130 const struct fib6_nh *nh = res->nh;
1131 const struct net_device *dev = nh->fib_nh_dev;
1132 struct fib6_info *f6i = res->f6i;
dcd1f572 1133
0d161581 1134 ip6_rt_init_dst(rt, res);
6edb3c96 1135
0d161581 1136 rt->rt6i_dst = f6i->fib6_dst;
dcd1f572 1137 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
7d21fec9 1138 rt->rt6i_flags = res->fib6_flags;
0d161581
DA		 1139 if (nh->fib_nh_gw_family) {
1140 rt->rt6i_gateway = nh->fib_nh_gw6;
2b2450ca
DA		 1141 rt->rt6i_flags |= RTF_GATEWAY;
1142 }
0d161581 1143 rt6_set_from(rt, f6i);
ae90d867 1144#ifdef CONFIG_IPV6_SUBTREES
0d161581 1145 rt->rt6i_src = f6i->fib6_src;
ae90d867 1146#endif
ae90d867
DA		 1147}
1148
a3c00e46
MKL		 1149static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1150 struct in6_addr *saddr)
1151{
66f5d6ce 1152 struct fib6_node *pn, *sn;
a3c00e46
MKL		 1153 while (1) {
1154 if (fn->fn_flags & RTN_TL_ROOT)
1155 return NULL;
66f5d6ce
WW		 1156 pn = rcu_dereference(fn->parent);
1157 sn = FIB6_SUBTREE(pn);
1158 if (sn && sn != fn)
6454743b 1159 fn = fib6_node_lookup(sn, NULL, saddr);
a3c00e46
MKL		 1160 else
1161 fn = pn;
1162 if (fn->fn_flags & RTN_RTINFO)
1163 return fn;
1164 }
1165}
c71099ac 1166
10585b43 1167static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
d3843fe5
WW		 1168{
1169 struct rt6_info *rt = *prt;
1170
1171 if (dst_hold_safe(&rt->dst))
1172 return true;
10585b43 1173 if (net) {
d3843fe5
WW		 1174 rt = net->ipv6.ip6_null_entry;
1175 dst_hold(&rt->dst);
1176 } else {
1177 rt = NULL;
1178 }
1179 *prt = rt;
1180 return false;
1181}
1182
dec9b0e2 1183/* called with rcu_lock held */
9b6b35ab 1184static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
dec9b0e2 1185{
9b6b35ab
DA		 1186 struct net_device *dev = res->nh->fib_nh_dev;
1187 struct fib6_info *f6i = res->f6i;
1188 unsigned short flags;
dec9b0e2
DA		 1189 struct rt6_info *nrt;
1190
9b6b35ab 1191 if (!fib6_info_hold_safe(f6i))
1c87e79a 1192 goto fallback;
e873e4b9 1193
9b6b35ab 1194 flags = fib6_info_dst_flags(f6i);
93531c67 1195 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1c87e79a 1196 if (!nrt) {
9b6b35ab 1197 fib6_info_release(f6i);
1c87e79a
XL		 1198 goto fallback;
1199 }
dec9b0e2 1200
0d161581 1201 ip6_rt_copy_init(nrt, res);
1c87e79a
XL		 1202 return nrt;
1203
1204fallback:
1205 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1206 dst_hold(&nrt->dst);
dec9b0e2
DA		 1207 return nrt;
1208}
1209
8ed67789
DL		 1210static struct rt6_info *ip6_pol_route_lookup(struct net *net,
1211 struct fib6_table *table,
b75cc8f9
DA		 1212 struct flowi6 *fl6,
1213 const struct sk_buff *skb,
1214 int flags)
1da177e4 1215{
b1d40991 1216 struct fib6_result res = {};
1da177e4 1217 struct fib6_node *fn;
23fb93a4 1218 struct rt6_info *rt;
1da177e4 1219
b6cdbc85
DA		 1220 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1221 flags &= ~RT6_LOOKUP_F_IFACE;
1222
66f5d6ce 1223 rcu_read_lock();
6454743b 1224 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
c71099ac 1225restart:
b1d40991
DA		 1226 res.f6i = rcu_dereference(fn->leaf);
1227 if (!res.f6i)
1228 res.f6i = net->ipv6.fib6_null_entry;
af52a52c 1229 else
75ef7389
DA		 1230 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1231 flags);
af52a52c 1232
b1d40991 1233 if (res.f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 1234 fn = fib6_backtrack(fn, &fl6->saddr);
1235 if (fn)
1236 goto restart;
2b760fcf 1237
af52a52c
DA		 1238 rt = net->ipv6.ip6_null_entry;
1239 dst_hold(&rt->dst);
1240 goto out;
f88d8ea6
DA		 1241 } else if (res.fib6_flags & RTF_REJECT) {
1242 goto do_create;
af52a52c 1243 }
d3843fe5 1244
b1d40991
DA		 1245 fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1246 fl6->flowi6_oif != 0, skb, flags);
1247
2b760fcf 1248 /* Search through exception table */
7e4b5128 1249 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
23fb93a4 1250 if (rt) {
10585b43 1251 if (ip6_hold_safe(net, &rt))
dec9b0e2 1252 dst_use_noref(&rt->dst, jiffies);
23fb93a4 1253 } else {
f88d8ea6 1254do_create:
9b6b35ab 1255 rt = ip6_create_rt_rcu(&res);
dec9b0e2 1256 }
b811580d 1257
af52a52c 1258out:
8ff2e5b2 1259 trace_fib6_table_lookup(net, &res, table, fl6);
af52a52c 1260
66f5d6ce 1261 rcu_read_unlock();
b811580d 1262
c71099ac 1263 return rt;
c71099ac
TG		 1264}
1265
67ba4152 1266struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
b75cc8f9 1267 const struct sk_buff *skb, int flags)
ea6e574e 1268{
b75cc8f9 1269 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
ea6e574e
FW		 1270}
1271EXPORT_SYMBOL_GPL(ip6_route_lookup);
1272
9acd9f3a 1273struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
b75cc8f9
DA		 1274 const struct in6_addr *saddr, int oif,
1275 const struct sk_buff *skb, int strict)
c71099ac 1276{
4c9483b2
DM		 1277 struct flowi6 fl6 = {
1278 .flowi6_oif = oif,
1279 .daddr = *daddr,
c71099ac
TG		 1280 };
1281 struct dst_entry *dst;
77d16f45 1282 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
c71099ac 1283
adaa70bb 1284 if (saddr) {
4c9483b2 1285 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
adaa70bb
TG		 1286 flags |= RT6_LOOKUP_F_HAS_SADDR;
1287 }
1288
b75cc8f9 1289 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
c71099ac
TG		 1290 if (dst->error == 0)
1291 return (struct rt6_info *) dst;
1292
1293 dst_release(dst);
1294
1da177e4
LT		 1295 return NULL;
1296}
7159039a
YH		 1297EXPORT_SYMBOL(rt6_lookup);
1298
c71099ac 1299/* ip6_ins_rt is called with FREE table->tb6_lock.
1cfb71ee
WW		 1300 * It takes new route entry, the addition fails by any reason the
1301 * route is released.
1302 * Caller must hold dst before calling it.
1da177e4
LT		 1303 */
1304
8d1c802b 1305static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
333c4301 1306 struct netlink_ext_ack *extack)
1da177e4
LT		 1307{
1308 int err;
c71099ac 1309 struct fib6_table *table;
1da177e4 1310
93c2fb25 1311 table = rt->fib6_table;
66f5d6ce 1312 spin_lock_bh(&table->tb6_lock);
d4ead6b3 1313 err = fib6_add(&table->tb6_root, rt, info, extack);
66f5d6ce 1314 spin_unlock_bh(&table->tb6_lock);
1da177e4
LT		 1315
1316 return err;
1317}
1318
8d1c802b 1319int ip6_ins_rt(struct net *net, struct fib6_info *rt)
40e22e8f 1320{
afb1d4b5 1321 struct nl_info info = { .nl_net = net, };
e715b6d3 1322
d4ead6b3 1323 return __ip6_ins_rt(rt, &info, NULL);
40e22e8f
TG		 1324}
1325
85bd05de 1326static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
8b9df265
MKL		 1327 const struct in6_addr *daddr,
1328 const struct in6_addr *saddr)
1da177e4 1329{
85bd05de 1330 struct fib6_info *f6i = res->f6i;
4832c30d 1331 struct net_device *dev;
1da177e4
LT		 1332 struct rt6_info *rt;
1333
1334 /*
1335 * Clone the route.
1336 */
1337
85bd05de 1338 if (!fib6_info_hold_safe(f6i))
e873e4b9
WW		 1339 return NULL;
1340
0d161581 1341 dev = ip6_rt_get_dev_rcu(res);
93531c67 1342 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
e873e4b9 1343 if (!rt) {
85bd05de 1344 fib6_info_release(f6i);
83a09abd 1345 return NULL;
e873e4b9 1346 }
83a09abd 1347
0d161581 1348 ip6_rt_copy_init(rt, res);
83a09abd 1349 rt->rt6i_flags |= RTF_CACHE;
83a09abd
MKL		 1350 rt->dst.flags |= DST_HOST;
1351 rt->rt6i_dst.addr = *daddr;
1352 rt->rt6i_dst.plen = 128;
1da177e4 1353
85bd05de
DA		 1354 if (!rt6_is_gw_or_nonexthop(res)) {
1355 if (f6i->fib6_dst.plen != 128 &&
1356 ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
83a09abd 1357 rt->rt6i_flags |= RTF_ANYCAST;
1da177e4 1358#ifdef CONFIG_IPV6_SUBTREES
83a09abd
MKL		 1359 if (rt->rt6i_src.plen && saddr) {
1360 rt->rt6i_src.addr = *saddr;
1361 rt->rt6i_src.plen = 128;
8b9df265 1362 }
83a09abd 1363#endif
95a9a5ba 1364 }
1da177e4 1365
95a9a5ba
YH		 1366 return rt;
1367}
1da177e4 1368
db3fedee 1369static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
d52d3997 1370{
db3fedee
DA		 1371 struct fib6_info *f6i = res->f6i;
1372 unsigned short flags = fib6_info_dst_flags(f6i);
4832c30d 1373 struct net_device *dev;
d52d3997
MKL		 1374 struct rt6_info *pcpu_rt;
1375
db3fedee 1376 if (!fib6_info_hold_safe(f6i))
e873e4b9
WW		 1377 return NULL;
1378
4832c30d 1379 rcu_read_lock();
0d161581 1380 dev = ip6_rt_get_dev_rcu(res);
93531c67 1381 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags);
4832c30d 1382 rcu_read_unlock();
e873e4b9 1383 if (!pcpu_rt) {
db3fedee 1384 fib6_info_release(f6i);
d52d3997 1385 return NULL;
e873e4b9 1386 }
0d161581 1387 ip6_rt_copy_init(pcpu_rt, res);
d52d3997
MKL		 1388 pcpu_rt->rt6i_flags |= RTF_PCPU;
1389 return pcpu_rt;
1390}
1391
66f5d6ce 1392/* It should be called with rcu_read_lock() acquired */
db3fedee 1393static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
d52d3997 1394{
c353071a 1395 struct rt6_info *pcpu_rt;
d52d3997 1396
c353071a 1397 pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
d52d3997 1398
a73e4195
MKL		 1399 return pcpu_rt;
1400}
1401
afb1d4b5 1402static struct rt6_info *rt6_make_pcpu_route(struct net *net,
db3fedee 1403 const struct fib6_result *res)
a73e4195
MKL		 1404{
1405 struct rt6_info *pcpu_rt, *prev, **p;
d52d3997 1406
db3fedee 1407 pcpu_rt = ip6_rt_pcpu_alloc(res);
0e09edcc
WW		 1408 if (!pcpu_rt)
1409 return NULL;
d52d3997 1410
f40b6ae2 1411 p = this_cpu_ptr(res->nh->rt6i_pcpu);
a94b9367 1412 prev = cmpxchg(p, NULL, pcpu_rt);
951f788a 1413 BUG_ON(prev);
a94b9367 1414
61fb0d01
ED		 1415 if (res->f6i->fib6_destroying) {
1416 struct fib6_info *from;
1417
1418 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1419 fib6_info_release(from);
1420 }
1421
d52d3997
MKL		 1422 return pcpu_rt;
1423}
1424
35732d01
WW		 1425/* exception hash table implementation
1426 */
1427static DEFINE_SPINLOCK(rt6_exception_lock);
1428
1429/* Remove rt6_ex from hash table and free the memory
1430 * Caller must hold rt6_exception_lock
1431 */
1432static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1433 struct rt6_exception *rt6_ex)
1434{
f5b51fe8 1435 struct fib6_info *from;
b2427e67 1436 struct net *net;
81eb8447 1437
35732d01
WW		 1438 if (!bucket || !rt6_ex)
1439 return;
b2427e67
CIK		 1440
1441 net = dev_net(rt6_ex->rt6i->dst.dev);
f5b51fe8
PA		 1442 net->ipv6.rt6_stats->fib_rt_cache--;
1443
1444 /* purge completely the exception to allow releasing the held resources:
1445 * some [sk] cache may keep the dst around for unlimited time
1446 */
0e233874 1447 from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
f5b51fe8
PA		 1448 fib6_info_release(from);
1449 dst_dev_put(&rt6_ex->rt6i->dst);
1450
35732d01 1451 hlist_del_rcu(&rt6_ex->hlist);
77634cc6 1452 dst_release(&rt6_ex->rt6i->dst);
35732d01
WW		 1453 kfree_rcu(rt6_ex, rcu);
1454 WARN_ON_ONCE(!bucket->depth);
1455 bucket->depth--;
1456}
1457
1458/* Remove oldest rt6_ex in bucket and free the memory
1459 * Caller must hold rt6_exception_lock
1460 */
1461static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1462{
1463 struct rt6_exception *rt6_ex, *oldest = NULL;
1464
1465 if (!bucket)
1466 return;
1467
1468 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1469 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1470 oldest = rt6_ex;
1471 }
1472 rt6_remove_exception(bucket, oldest);
1473}
1474
1475static u32 rt6_exception_hash(const struct in6_addr *dst,
1476 const struct in6_addr *src)
1477{
1478 static u32 seed __read_mostly;
1479 u32 val;
1480
1481 net_get_random_once(&seed, sizeof(seed));
b6b556af 1482 val = jhash2((const u32 *)dst, sizeof(*dst)/sizeof(u32), seed);
35732d01
WW		 1483
1484#ifdef CONFIG_IPV6_SUBTREES
1485 if (src)
b6b556af 1486 val = jhash2((const u32 *)src, sizeof(*src)/sizeof(u32), val);
35732d01
WW		 1487#endif
1488 return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1489}
1490
1491/* Helper function to find the cached rt in the hash table
1492 * and update bucket pointer to point to the bucket for this
1493 * (daddr, saddr) pair
1494 * Caller must hold rt6_exception_lock
1495 */
1496static struct rt6_exception *
1497__rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1498 const struct in6_addr *daddr,
1499 const struct in6_addr *saddr)
1500{
1501 struct rt6_exception *rt6_ex;
1502 u32 hval;
1503
1504 if (!(*bucket) || !daddr)
1505 return NULL;
1506
1507 hval = rt6_exception_hash(daddr, saddr);
1508 *bucket += hval;
1509
1510 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1511 struct rt6_info *rt6 = rt6_ex->rt6i;
1512 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1513
1514#ifdef CONFIG_IPV6_SUBTREES
1515 if (matched && saddr)
1516 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1517#endif
1518 if (matched)
1519 return rt6_ex;
1520 }
1521 return NULL;
1522}
1523
1524/* Helper function to find the cached rt in the hash table
1525 * and update bucket pointer to point to the bucket for this
1526 * (daddr, saddr) pair
1527 * Caller must hold rcu_read_lock()
1528 */
1529static struct rt6_exception *
1530__rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1531 const struct in6_addr *daddr,
1532 const struct in6_addr *saddr)
1533{
1534 struct rt6_exception *rt6_ex;
1535 u32 hval;
1536
1537 WARN_ON_ONCE(!rcu_read_lock_held());
1538
1539 if (!(*bucket) || !daddr)
1540 return NULL;
1541
1542 hval = rt6_exception_hash(daddr, saddr);
1543 *bucket += hval;
1544
1545 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1546 struct rt6_info *rt6 = rt6_ex->rt6i;
1547 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1548
1549#ifdef CONFIG_IPV6_SUBTREES
1550 if (matched && saddr)
1551 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1552#endif
1553 if (matched)
1554 return rt6_ex;
1555 }
1556 return NULL;
1557}
1558
b748f260 1559static unsigned int fib6_mtu(const struct fib6_result *res)
d4ead6b3 1560{
b748f260 1561 const struct fib6_nh *nh = res->nh;
d4ead6b3
DA		 1562 unsigned int mtu;
1563
b748f260
DA		 1564 if (res->f6i->fib6_pmtu) {
1565 mtu = res->f6i->fib6_pmtu;
dcd1f572 1566 } else {
b748f260 1567 struct net_device *dev = nh->fib_nh_dev;
dcd1f572
DA		 1568 struct inet6_dev *idev;
1569
1570 rcu_read_lock();
1571 idev = __in6_dev_get(dev);
1572 mtu = idev->cnf.mtu6;
1573 rcu_read_unlock();
1574 }
1575
d4ead6b3
DA		 1576 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1577
b748f260 1578 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
d4ead6b3
DA		 1579}
1580
cc5c073a
DA		 1581#define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1582
1583/* used when the flushed bit is not relevant, only access to the bucket
1584 * (ie., all bucket users except rt6_insert_exception);
1585 *
1586 * called under rcu lock; sometimes called with rt6_exception_lock held
1587 */
1588static
1589struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1590 spinlock_t *lock)
1591{
1592 struct rt6_exception_bucket *bucket;
1593
1594 if (lock)
1595 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1596 lockdep_is_held(lock));
1597 else
1598 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1599
1600 /* remove bucket flushed bit if set */
1601 if (bucket) {
1602 unsigned long p = (unsigned long)bucket;
1603
1604 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1605 bucket = (struct rt6_exception_bucket *)p;
1606 }
1607
1608 return bucket;
1609}
1610
1611static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1612{
1613 unsigned long p = (unsigned long)bucket;
1614
1615 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1616}
1617
1618/* called with rt6_exception_lock held */
1619static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1620 spinlock_t *lock)
1621{
1622 struct rt6_exception_bucket *bucket;
1623 unsigned long p;
1624
1625 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1626 lockdep_is_held(lock));
1627
1628 p = (unsigned long)bucket;
1629 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1630 bucket = (struct rt6_exception_bucket *)p;
1631 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1632}
1633
35732d01 1634static int rt6_insert_exception(struct rt6_info *nrt,
5012f0a5 1635 const struct fib6_result *res)
35732d01 1636{
5e670d84 1637 struct net *net = dev_net(nrt->dst.dev);
35732d01 1638 struct rt6_exception_bucket *bucket;
cc5c073a 1639 struct fib6_info *f6i = res->f6i;
35732d01
WW		 1640 struct in6_addr *src_key = NULL;
1641 struct rt6_exception *rt6_ex;
cc5c073a 1642 struct fib6_nh *nh = res->nh;
35732d01
WW		 1643 int err = 0;
1644
35732d01
WW		 1645 spin_lock_bh(&rt6_exception_lock);
1646
cc5c073a
DA		 1647 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1648 lockdep_is_held(&rt6_exception_lock));
35732d01
WW		 1649 if (!bucket) {
1650 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1651 GFP_ATOMIC);
1652 if (!bucket) {
1653 err = -ENOMEM;
1654 goto out;
1655 }
cc5c073a
DA		 1656 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1657 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1658 err = -EINVAL;
1659 goto out;
35732d01
WW		 1660 }
1661
1662#ifdef CONFIG_IPV6_SUBTREES
5012f0a5 1663 /* fib6_src.plen != 0 indicates f6i is in subtree
35732d01 1664 * and exception table is indexed by a hash of
5012f0a5 1665 * both fib6_dst and fib6_src.
35732d01 1666 * Otherwise, the exception table is indexed by
5012f0a5 1667 * a hash of only fib6_dst.
35732d01 1668 */
5012f0a5 1669 if (f6i->fib6_src.plen)
35732d01
WW		 1670 src_key = &nrt->rt6i_src.addr;
1671#endif
5012f0a5 1672 /* rt6_mtu_change() might lower mtu on f6i.
f5bbe7ee 1673 * Only insert this exception route if its mtu
5012f0a5 1674 * is less than f6i's mtu value.
f5bbe7ee 1675 */
b748f260 1676 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
f5bbe7ee
WW		 1677 err = -EINVAL;
1678 goto out;
1679 }
60006a48 1680
35732d01
WW		 1681 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1682 src_key);
1683 if (rt6_ex)
1684 rt6_remove_exception(bucket, rt6_ex);
1685
1686 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1687 if (!rt6_ex) {
1688 err = -ENOMEM;
1689 goto out;
1690 }
1691 rt6_ex->rt6i = nrt;
1692 rt6_ex->stamp = jiffies;
35732d01
WW		 1693 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1694 bucket->depth++;
81eb8447 1695 net->ipv6.rt6_stats->fib_rt_cache++;
35732d01
WW		 1696
1697 if (bucket->depth > FIB6_MAX_DEPTH)
1698 rt6_exception_remove_oldest(bucket);
1699
1700out:
1701 spin_unlock_bh(&rt6_exception_lock);
1702
1703 /* Update fn->fn_sernum to invalidate all cached dst */
b886d5f2 1704 if (!err) {
5012f0a5
DA		 1705 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1706 fib6_update_sernum(net, f6i);
1707 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
b886d5f2
PA		 1708 fib6_force_start_gc(net);
1709 }
35732d01
WW		 1710
1711 return err;
1712}
1713
c0b220cf 1714static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
35732d01
WW		 1715{
1716 struct rt6_exception_bucket *bucket;
1717 struct rt6_exception *rt6_ex;
1718 struct hlist_node *tmp;
1719 int i;
1720
1721 spin_lock_bh(&rt6_exception_lock);
35732d01 1722
cc5c073a 1723 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
35732d01
WW		 1724 if (!bucket)
1725 goto out;
1726
cc5c073a
DA		 1727 /* Prevent rt6_insert_exception() to recreate the bucket list */
1728 if (!from)
1729 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1730
35732d01 1731 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
cc5c073a
DA		 1732 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1733 if (!from ||
1734 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1735 rt6_remove_exception(bucket, rt6_ex);
1736 }
1737 WARN_ON_ONCE(!from && bucket->depth);
35732d01
WW		 1738 bucket++;
1739 }
35732d01
WW		 1740out:
1741 spin_unlock_bh(&rt6_exception_lock);
1742}
1743
e659ba31
DA		 1744static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1745{
1746 struct fib6_info *f6i = arg;
1747
1748 fib6_nh_flush_exceptions(nh, f6i);
1749
1750 return 0;
1751}
1752
c0b220cf
DA		 1753void rt6_flush_exceptions(struct fib6_info *f6i)
1754{
e659ba31
DA		 1755 if (f6i->nh)
1756 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1757 f6i);
1758 else
1759 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
c0b220cf
DA		 1760}
1761
35732d01
WW		 1762/* Find cached rt in the hash table inside passed in rt
1763 * Caller has to hold rcu_read_lock()
1764 */
7e4b5128 1765static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
510e2ced
WW		 1766 const struct in6_addr *daddr,
1767 const struct in6_addr *saddr)
35732d01 1768{
510e2ced 1769 const struct in6_addr *src_key = NULL;
35732d01 1770 struct rt6_exception_bucket *bucket;
35732d01 1771 struct rt6_exception *rt6_ex;
7e4b5128 1772 struct rt6_info *ret = NULL;
35732d01 1773
35732d01 1774#ifdef CONFIG_IPV6_SUBTREES
7e4b5128 1775 /* fib6i_src.plen != 0 indicates f6i is in subtree
35732d01 1776 * and exception table is indexed by a hash of
7e4b5128 1777 * both fib6_dst and fib6_src.
510e2ced
WW		 1778 * However, the src addr used to create the hash
1779 * might not be exactly the passed in saddr which
1780 * is a /128 addr from the flow.
1781 * So we need to use f6i->fib6_src to redo lookup
1782 * if the passed in saddr does not find anything.
1783 * (See the logic in ip6_rt_cache_alloc() on how
1784 * rt->rt6i_src is updated.)
35732d01 1785 */
7e4b5128 1786 if (res->f6i->fib6_src.plen)
35732d01 1787 src_key = saddr;
510e2ced 1788find_ex:
35732d01 1789#endif
cc5c073a 1790 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
35732d01
WW		 1791 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1792
1793 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
7e4b5128 1794 ret = rt6_ex->rt6i;
35732d01 1795
510e2ced
WW		 1796#ifdef CONFIG_IPV6_SUBTREES
1797 /* Use fib6_src as src_key and redo lookup */
1798 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1799 src_key = &res->f6i->fib6_src.addr;
1800 goto find_ex;
1801 }
1802#endif
1803
7e4b5128 1804 return ret;
35732d01
WW		 1805}
1806
1807/* Remove the passed in cached rt from the hash table that contains it */
cc5c073a 1808static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1809 const struct rt6_info *rt)
35732d01 1810{
c0b220cf 1811 const struct in6_addr *src_key = NULL;
35732d01 1812 struct rt6_exception_bucket *bucket;
35732d01
WW		 1813 struct rt6_exception *rt6_ex;
1814 int err;
1815
cc5c073a 1816 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
35732d01
WW		 1817 return -ENOENT;
1818
1819 spin_lock_bh(&rt6_exception_lock);
cc5c073a
DA		 1820 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1821
35732d01 1822#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1823 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1824 * and exception table is indexed by a hash of
1825 * both rt6i_dst and rt6i_src.
35732d01
WW		 1826 * Otherwise, the exception table is indexed by
1827 * a hash of only rt6i_dst.
1828 */
c0b220cf 1829 if (plen)
35732d01
WW		 1830 src_key = &rt->rt6i_src.addr;
1831#endif
1832 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1833 &rt->rt6i_dst.addr,
1834 src_key);
1835 if (rt6_ex) {
1836 rt6_remove_exception(bucket, rt6_ex);
1837 err = 0;
1838 } else {
1839 err = -ENOENT;
1840 }
1841
1842 spin_unlock_bh(&rt6_exception_lock);
1843 return err;
1844}
1845
e659ba31
DA		 1846struct fib6_nh_excptn_arg {
1847 struct rt6_info *rt;
1848 int plen;
1849};
1850
1851static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1852{
1853 struct fib6_nh_excptn_arg *arg = _arg;
1854 int err;
1855
1856 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1857 if (err == 0)
1858 return 1;
1859
1860 return 0;
1861}
1862
c0b220cf
DA		 1863static int rt6_remove_exception_rt(struct rt6_info *rt)
1864{
1865 struct fib6_info *from;
1866
1867 from = rcu_dereference(rt->from);
cc5c073a 1868 if (!from || !(rt->rt6i_flags & RTF_CACHE))
c0b220cf
DA		 1869 return -EINVAL;
1870
e659ba31
DA		 1871 if (from->nh) {
1872 struct fib6_nh_excptn_arg arg = {
1873 .rt = rt,
1874 .plen = from->fib6_src.plen
1875 };
1876 int rc;
1877
1878 /* rc = 1 means an entry was found */
1879 rc = nexthop_for_each_fib6_nh(from->nh,
1880 rt6_nh_remove_exception_rt,
1881 &arg);
1882 return rc ? 0 : -ENOENT;
1883 }
1884
1cf844c7 1885 return fib6_nh_remove_exception(from->fib6_nh,
cc5c073a 1886 from->fib6_src.plen, rt);
c0b220cf
DA		 1887}
1888
35732d01
WW		 1889/* Find rt6_ex which contains the passed in rt cache and
1890 * refresh its stamp
1891 */
cc5c073a 1892static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1893 const struct rt6_info *rt)
35732d01 1894{
c0b220cf 1895 const struct in6_addr *src_key = NULL;
35732d01 1896 struct rt6_exception_bucket *bucket;
35732d01 1897 struct rt6_exception *rt6_ex;
193f3685 1898
cc5c073a 1899 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
35732d01 1900#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1901 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1902 * and exception table is indexed by a hash of
1903 * both rt6i_dst and rt6i_src.
35732d01
WW		 1904 * Otherwise, the exception table is indexed by
1905 * a hash of only rt6i_dst.
1906 */
c0b220cf 1907 if (plen)
35732d01
WW		 1908 src_key = &rt->rt6i_src.addr;
1909#endif
cc5c073a 1910 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
35732d01
WW		 1911 if (rt6_ex)
1912 rt6_ex->stamp = jiffies;
c0b220cf
DA		 1913}
1914
e659ba31
DA		 1915struct fib6_nh_match_arg {
1916 const struct net_device *dev;
1917 const struct in6_addr *gw;
1918 struct fib6_nh *match;
1919};
1920
1921/* determine if fib6_nh has given device and gateway */
1922static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1923{
1924 struct fib6_nh_match_arg *arg = _arg;
1925
1926 if (arg->dev != nh->fib_nh_dev ||
1927 (arg->gw && !nh->fib_nh_gw_family) ||
1928 (!arg->gw && nh->fib_nh_gw_family) ||
1929 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1930 return 0;
1931
1932 arg->match = nh;
1933
1934 /* found a match, break the loop */
1935 return 1;
1936}
1937
c0b220cf
DA		 1938static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1939{
1940 struct fib6_info *from;
e659ba31 1941 struct fib6_nh *fib6_nh;
35732d01 1942
c0b220cf
DA		 1943 rcu_read_lock();
1944
1945 from = rcu_dereference(rt->from);
1946 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1947 goto unlock;
1948
e659ba31
DA		 1949 if (from->nh) {
1950 struct fib6_nh_match_arg arg = {
1951 .dev = rt->dst.dev,
1952 .gw = &rt->rt6i_gateway,
1953 };
1954
1955 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1956
1957 if (!arg.match)
cff6a327 1958 goto unlock;
e659ba31
DA		 1959 fib6_nh = arg.match;
1960 } else {
1961 fib6_nh = from->fib6_nh;
1962 }
1963 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
193f3685 1964unlock:
35732d01
WW		 1965 rcu_read_unlock();
1966}
1967
e9fa1495
SB		 1968static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1969 struct rt6_info *rt, int mtu)
1970{
1971 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1972 * lowest MTU in the path: always allow updating the route PMTU to
1973 * reflect PMTU decreases.
1974 *
1975 * If the new MTU is higher, and the route PMTU is equal to the local
1976 * MTU, this means the old MTU is the lowest in the path, so allow
1977 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1978 * handle this.
1979 */
1980
1981 if (dst_mtu(&rt->dst) >= mtu)
1982 return true;
1983
1984 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
1985 return true;
1986
1987 return false;
1988}
1989
1990static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
cc5c073a 1991 const struct fib6_nh *nh, int mtu)
f5bbe7ee
WW		 1992{
1993 struct rt6_exception_bucket *bucket;
1994 struct rt6_exception *rt6_ex;
1995 int i;
1996
cc5c073a 1997 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
e9fa1495
SB		 1998 if (!bucket)
1999 return;
2000
2001 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2002 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2003 struct rt6_info *entry = rt6_ex->rt6i;
2004
2005 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
d4ead6b3 2006 * route), the metrics of its rt->from have already
e9fa1495
SB		 2007 * been updated.
2008 */
d4ead6b3 2009 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
e9fa1495 2010 rt6_mtu_change_route_allowed(idev, entry, mtu))
d4ead6b3 2011 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
f5bbe7ee 2012 }
e9fa1495 2013 bucket++;
f5bbe7ee
WW		 2014 }
2015}
2016
b16cb459
WW		 2017#define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2018
cc5c073a
DA		 2019static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2020 const struct in6_addr *gateway)
b16cb459
WW		 2021{
2022 struct rt6_exception_bucket *bucket;
2023 struct rt6_exception *rt6_ex;
2024 struct hlist_node *tmp;
2025 int i;
2026
cc5c073a 2027 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
b16cb459
WW		 2028 return;
2029
2030 spin_lock_bh(&rt6_exception_lock);
cc5c073a 2031 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
b16cb459
WW		 2032 if (bucket) {
2033 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2034 hlist_for_each_entry_safe(rt6_ex, tmp,
2035 &bucket->chain, hlist) {
2036 struct rt6_info *entry = rt6_ex->rt6i;
2037
2038 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2039 RTF_CACHE_GATEWAY &&
2040 ipv6_addr_equal(gateway,
2041 &entry->rt6i_gateway)) {
2042 rt6_remove_exception(bucket, rt6_ex);
2043 }
2044 }
2045 bucket++;
2046 }
2047 }
2048
2049 spin_unlock_bh(&rt6_exception_lock);
2050}
2051
c757faa8
WW		 2052static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2053 struct rt6_exception *rt6_ex,
2054 struct fib6_gc_args *gc_args,
2055 unsigned long now)
2056{
2057 struct rt6_info *rt = rt6_ex->rt6i;
2058
1859bac0
PA		 2059 /* we are pruning and obsoleting aged-out and non gateway exceptions
2060 * even if others have still references to them, so that on next
2061 * dst_check() such references can be dropped.
2062 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2063 * expired, independently from their aging, as per RFC 8201 section 4
2064 */
31afeb42
WW		 2065 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2066 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2067 RT6_TRACE("aging clone %p\n", rt);
2068 rt6_remove_exception(bucket, rt6_ex);
2069 return;
2070 }
2071 } else if (time_after(jiffies, rt->dst.expires)) {
2072 RT6_TRACE("purging expired route %p\n", rt);
c757faa8
WW		 2073 rt6_remove_exception(bucket, rt6_ex);
2074 return;
31afeb42
WW		 2075 }
2076
2077 if (rt->rt6i_flags & RTF_GATEWAY) {
c757faa8
WW		 2078 struct neighbour *neigh;
2079 __u8 neigh_flags = 0;
2080
1bfa26ff
ED		 2081 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2082 if (neigh)
c757faa8 2083 neigh_flags = neigh->flags;
1bfa26ff 2084
c757faa8
WW		 2085 if (!(neigh_flags & NTF_ROUTER)) {
2086 RT6_TRACE("purging route %p via non-router but gateway\n",
2087 rt);
2088 rt6_remove_exception(bucket, rt6_ex);
2089 return;
2090 }
2091 }
31afeb42 2092
c757faa8
WW		 2093 gc_args->more++;
2094}
2095
cc5c073a 2096static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
c0b220cf
DA		 2097 struct fib6_gc_args *gc_args,
2098 unsigned long now)
c757faa8
WW		 2099{
2100 struct rt6_exception_bucket *bucket;
2101 struct rt6_exception *rt6_ex;
2102 struct hlist_node *tmp;
2103 int i;
2104
cc5c073a 2105 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
c757faa8
WW		 2106 return;
2107
1bfa26ff
ED		 2108 rcu_read_lock_bh();
2109 spin_lock(&rt6_exception_lock);
cc5c073a 2110 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
c757faa8
WW		 2111 if (bucket) {
2112 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2113 hlist_for_each_entry_safe(rt6_ex, tmp,
2114 &bucket->chain, hlist) {
2115 rt6_age_examine_exception(bucket, rt6_ex,
2116 gc_args, now);
2117 }
2118 bucket++;
2119 }
2120 }
1bfa26ff
ED		 2121 spin_unlock(&rt6_exception_lock);
2122 rcu_read_unlock_bh();
c757faa8
WW		 2123}
2124
e659ba31
DA		 2125struct fib6_nh_age_excptn_arg {
2126 struct fib6_gc_args *gc_args;
2127 unsigned long now;
2128};
2129
2130static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2131{
2132 struct fib6_nh_age_excptn_arg *arg = _arg;
2133
2134 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2135 return 0;
2136}
2137
cc5c073a 2138void rt6_age_exceptions(struct fib6_info *f6i,
c0b220cf
DA		 2139 struct fib6_gc_args *gc_args,
2140 unsigned long now)
2141{
e659ba31
DA		 2142 if (f6i->nh) {
2143 struct fib6_nh_age_excptn_arg arg = {
2144 .gc_args = gc_args,
2145 .now = now
2146 };
2147
2148 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2149 &arg);
2150 } else {
2151 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2152 }
c0b220cf
DA		 2153}
2154
1d053da9 2155/* must be called with rcu lock held */
effda4dd
DA		 2156int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2157 struct flowi6 *fl6, struct fib6_result *res, int strict)
1da177e4 2158{
367efcb9 2159 struct fib6_node *fn, *saved_fn;
1da177e4 2160
6454743b 2161 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
367efcb9 2162 saved_fn = fn;
1da177e4 2163
ca254490
DA		 2164 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2165 oif = 0;
2166
a3c00e46 2167redo_rt6_select:
effda4dd
DA		 2168 rt6_select(net, fn, oif, res, strict);
2169 if (res->f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2170 fn = fib6_backtrack(fn, &fl6->saddr);
2171 if (fn)
2172 goto redo_rt6_select;
367efcb9
MKL		 2173 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2174 /* also consider unreachable route */
2175 strict &= ~RT6_LOOKUP_F_REACHABLE;
2176 fn = saved_fn;
2177 goto redo_rt6_select;
367efcb9 2178 }
a3c00e46
MKL		 2179 }
2180
effda4dd 2181 trace_fib6_table_lookup(net, res, table, fl6);
fb9de91e 2182
effda4dd 2183 return 0;
1d053da9
DA		 2184}
2185
2186struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2187 int oif, struct flowi6 *fl6,
2188 const struct sk_buff *skb, int flags)
2189{
b1d40991 2190 struct fib6_result res = {};
0e09edcc 2191 struct rt6_info *rt = NULL;
1d053da9
DA		 2192 int strict = 0;
2193
0e09edcc
WW		 2194 WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2195 !rcu_read_lock_held());
2196
1d053da9
DA		 2197 strict |= flags & RT6_LOOKUP_F_IFACE;
2198 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2199 if (net->ipv6.devconf_all->forwarding == 0)
2200 strict |= RT6_LOOKUP_F_REACHABLE;
2201
2202 rcu_read_lock();
2203
effda4dd 2204 fib6_table_lookup(net, table, oif, fl6, &res, strict);
0e09edcc
WW		 2205 if (res.f6i == net->ipv6.fib6_null_entry)
2206 goto out;
23fb93a4 2207
b1d40991 2208 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
d83009d4 2209
23fb93a4 2210 /*Search through exception table */
7e4b5128 2211 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
23fb93a4 2212 if (rt) {
0e09edcc 2213 goto out;
3da59bd9 2214 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
b1d40991 2215 !res.nh->fib_nh_gw_family)) {
3da59bd9
MKL		 2216 /* Create a RTF_CACHE clone which will not be
2217 * owned by the fib6 tree. It is for the special case where
2218 * the daddr in the skb during the neighbor look-up is different
2219 * from the fl6->daddr used to look-up route here.
2220 */
0e09edcc 2221 rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
c71099ac 2222
0e09edcc
WW		 2223 if (rt) {
2224 /* 1 refcnt is taken during ip6_rt_cache_alloc().
2225 * As rt6_uncached_list_add() does not consume refcnt,
2226 * this refcnt is always returned to the caller even
2227 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
1cfb71ee 2228 */
0e09edcc 2229 rt6_uncached_list_add(rt);
81eb8447 2230 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
0e09edcc 2231 rcu_read_unlock();
b811580d 2232
0e09edcc
WW		 2233 return rt;
2234 }
d52d3997
MKL		 2235 } else {
2236 /* Get a percpu copy */
951f788a 2237 local_bh_disable();
0e09edcc 2238 rt = rt6_get_pcpu_route(&res);
d52d3997 2239
0e09edcc
WW		 2240 if (!rt)
2241 rt = rt6_make_pcpu_route(net, &res);
93531c67 2242
951f788a 2243 local_bh_enable();
d52d3997 2244 }
0e09edcc
WW		 2245out:
2246 if (!rt)
2247 rt = net->ipv6.ip6_null_entry;
2248 if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2249 ip6_hold_safe(net, &rt);
2250 rcu_read_unlock();
2251
2252 return rt;
1da177e4 2253}
9ff74384 2254EXPORT_SYMBOL_GPL(ip6_pol_route);
1da177e4 2255
b75cc8f9
DA		 2256static struct rt6_info *ip6_pol_route_input(struct net *net,
2257 struct fib6_table *table,
2258 struct flowi6 *fl6,
2259 const struct sk_buff *skb,
2260 int flags)
4acad72d 2261{
b75cc8f9 2262 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
4acad72d
PE		 2263}
2264
d409b847
MB		 2265struct dst_entry *ip6_route_input_lookup(struct net *net,
2266 struct net_device *dev,
b75cc8f9
DA		 2267 struct flowi6 *fl6,
2268 const struct sk_buff *skb,
2269 int flags)
72331bc0
SL		 2270{
2271 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2272 flags |= RT6_LOOKUP_F_IFACE;
2273
b75cc8f9 2274 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
72331bc0 2275}
d409b847 2276EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
72331bc0 2277
23aebdac 2278static void ip6_multipath_l3_keys(const struct sk_buff *skb,
5e5d6fed
RP		 2279 struct flow_keys *keys,
2280 struct flow_keys *flkeys)
23aebdac
JS		 2281{
2282 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2283 const struct ipv6hdr *key_iph = outer_iph;
5e5d6fed 2284 struct flow_keys *_flkeys = flkeys;
23aebdac
JS		 2285 const struct ipv6hdr *inner_iph;
2286 const struct icmp6hdr *icmph;
2287 struct ipv6hdr _inner_iph;
cea67a2d 2288 struct icmp6hdr _icmph;
23aebdac
JS		 2289
2290 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2291 goto out;
2292
cea67a2d
ED		 2293 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2294 sizeof(_icmph), &_icmph);
2295 if (!icmph)
2296 goto out;
2297
54074f1d 2298 if (!icmpv6_is_err(icmph->icmp6_type))
23aebdac
JS		 2299 goto out;
2300
2301 inner_iph = skb_header_pointer(skb,
2302 skb_transport_offset(skb) + sizeof(*icmph),
2303 sizeof(_inner_iph), &_inner_iph);
2304 if (!inner_iph)
2305 goto out;
2306
2307 key_iph = inner_iph;
5e5d6fed 2308 _flkeys = NULL;
23aebdac 2309out:
5e5d6fed
RP		 2310 if (_flkeys) {
2311 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2312 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2313 keys->tags.flow_label = _flkeys->tags.flow_label;
2314 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2315 } else {
2316 keys->addrs.v6addrs.src = key_iph->saddr;
2317 keys->addrs.v6addrs.dst = key_iph->daddr;
fa1be7e0 2318 keys->tags.flow_label = ip6_flowlabel(key_iph);
5e5d6fed
RP		 2319 keys->basic.ip_proto = key_iph->nexthdr;
2320 }
23aebdac
JS		 2321}
2322
2323/* if skb is set it will be used and fl6 can be NULL */
b4bac172
DA		 2324u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2325 const struct sk_buff *skb, struct flow_keys *flkeys)
23aebdac
JS		 2326{
2327 struct flow_keys hash_keys;
9a2a537a 2328 u32 mhash;
23aebdac 2329
bbfa047a 2330 switch (ip6_multipath_hash_policy(net)) {
b4bac172
DA		 2331 case 0:
2332 memset(&hash_keys, 0, sizeof(hash_keys));
2333 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2334 if (skb) {
2335 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2336 } else {
2337 hash_keys.addrs.v6addrs.src = fl6->saddr;
2338 hash_keys.addrs.v6addrs.dst = fl6->daddr;
fa1be7e0 2339 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
b4bac172
DA		 2340 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2341 }
2342 break;
2343 case 1:
2344 if (skb) {
2345 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2346 struct flow_keys keys;
2347
2348 /* short-circuit if we already have L4 hash present */
2349 if (skb->l4_hash)
2350 return skb_get_hash_raw(skb) >> 1;
2351
2352 memset(&hash_keys, 0, sizeof(hash_keys));
2353
2354 if (!flkeys) {
2355 skb_flow_dissect_flow_keys(skb, &keys, flag);
2356 flkeys = &keys;
2357 }
2358 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2359 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2360 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2361 hash_keys.ports.src = flkeys->ports.src;
2362 hash_keys.ports.dst = flkeys->ports.dst;
2363 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2364 } else {
2365 memset(&hash_keys, 0, sizeof(hash_keys));
2366 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2367 hash_keys.addrs.v6addrs.src = fl6->saddr;
2368 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2369 hash_keys.ports.src = fl6->fl6_sport;
2370 hash_keys.ports.dst = fl6->fl6_dport;
2371 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2372 }
2373 break;
d8f74f09
SS		 2374 case 2:
2375 memset(&hash_keys, 0, sizeof(hash_keys));
2376 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2377 if (skb) {
2378 struct flow_keys keys;
2379
2380 if (!flkeys) {
2381 skb_flow_dissect_flow_keys(skb, &keys, 0);
2382 flkeys = &keys;
2383 }
2384
2385 /* Inner can be v4 or v6 */
2386 if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2387 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2388 hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2389 hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2390 } else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2391 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2392 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2393 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2394 hash_keys.tags.flow_label = flkeys->tags.flow_label;
2395 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2396 } else {
2397 /* Same as case 0 */
2398 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2399 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2400 }
2401 } else {
2402 /* Same as case 0 */
2403 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2404 hash_keys.addrs.v6addrs.src = fl6->saddr;
2405 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2406 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2407 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2408 }
2409 break;
23aebdac 2410 }
9a2a537a 2411 mhash = flow_hash_from_keys(&hash_keys);
23aebdac 2412
9a2a537a 2413 return mhash >> 1;
23aebdac
JS		 2414}
2415
67f415dd 2416/* Called with rcu held */
c71099ac
TG		 2417void ip6_route_input(struct sk_buff *skb)
2418{
b71d1d42 2419 const struct ipv6hdr *iph = ipv6_hdr(skb);
c346dca1 2420 struct net *net = dev_net(skb->dev);
67f415dd 2421 int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
904af04d 2422 struct ip_tunnel_info *tun_info;
4c9483b2 2423 struct flowi6 fl6 = {
e0d56fdd 2424 .flowi6_iif = skb->dev->ifindex,
4c9483b2
DM		 2425 .daddr = iph->daddr,
2426 .saddr = iph->saddr,
6502ca52 2427 .flowlabel = ip6_flowinfo(iph),
4c9483b2
DM		 2428 .flowi6_mark = skb->mark,
2429 .flowi6_proto = iph->nexthdr,
c71099ac 2430 };
5e5d6fed 2431 struct flow_keys *flkeys = NULL, _flkeys;
adaa70bb 2432
904af04d 2433 tun_info = skb_tunnel_info(skb);
46fa062a 2434 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
904af04d 2435 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
5e5d6fed
RP		 2436
2437 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2438 flkeys = &_flkeys;
2439
23aebdac 2440 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
b4bac172 2441 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
06e9d040 2442 skb_dst_drop(skb);
67f415dd
WW		 2443 skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2444 &fl6, skb, flags));
c71099ac
TG		 2445}
2446
b75cc8f9
DA		 2447static struct rt6_info *ip6_pol_route_output(struct net *net,
2448 struct fib6_table *table,
2449 struct flowi6 *fl6,
2450 const struct sk_buff *skb,
2451 int flags)
1da177e4 2452{
b75cc8f9 2453 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
c71099ac
TG		 2454}
2455
7d9e5f42
WW		 2456struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2457 const struct sock *sk,
2458 struct flowi6 *fl6, int flags)
c71099ac 2459{
d46a9d67 2460 bool any_src;
c71099ac 2461
3ede0bbc
RS		 2462 if (ipv6_addr_type(&fl6->daddr) &
2463 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
4c1feac5
DA		 2464 struct dst_entry *dst;
2465
7d9e5f42 2466 /* This function does not take refcnt on the dst */
4c1feac5
DA		 2467 dst = l3mdev_link_scope_lookup(net, fl6);
2468 if (dst)
2469 return dst;
2470 }
ca254490 2471
1fb9489b 2472 fl6->flowi6_iif = LOOPBACK_IFINDEX;
4dc27d1c 2473
7d9e5f42 2474 flags |= RT6_LOOKUP_F_DST_NOREF;
d46a9d67 2475 any_src = ipv6_addr_any(&fl6->saddr);
741a11d9 2476 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
d46a9d67 2477 (fl6->flowi6_oif && any_src))
77d16f45 2478 flags |= RT6_LOOKUP_F_IFACE;
c71099ac 2479
d46a9d67 2480 if (!any_src)
adaa70bb 2481 flags |= RT6_LOOKUP_F_HAS_SADDR;
0c9a2ac1
YH		 2482 else if (sk)
2483 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
adaa70bb 2484
b75cc8f9 2485 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
1da177e4 2486}
7d9e5f42
WW		 2487EXPORT_SYMBOL_GPL(ip6_route_output_flags_noref);
2488
2489struct dst_entry *ip6_route_output_flags(struct net *net,
2490 const struct sock *sk,
2491 struct flowi6 *fl6,
2492 int flags)
2493{
2494 struct dst_entry *dst;
2495 struct rt6_info *rt6;
2496
2497 rcu_read_lock();
2498 dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2499 rt6 = (struct rt6_info *)dst;
2500 /* For dst cached in uncached_list, refcnt is already taken. */
2501 if (list_empty(&rt6->rt6i_uncached) && !dst_hold_safe(dst)) {
2502 dst = &net->ipv6.ip6_null_entry->dst;
2503 dst_hold(dst);
2504 }
2505 rcu_read_unlock();
2506
2507 return dst;
2508}
6f21c96a 2509EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1da177e4 2510
2774c131 2511struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
14e50e57 2512{
5c1e6aa3 2513 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1dbe3252 2514 struct net_device *loopback_dev = net->loopback_dev;
14e50e57
DM		 2515 struct dst_entry *new = NULL;
2516
1dbe3252 2517 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
62cf27e5 2518 DST_OBSOLETE_DEAD, 0);
14e50e57 2519 if (rt) {
0a1f5962 2520 rt6_info_init(rt);
81eb8447 2521 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
8104891b 2522
0a1f5962 2523 new = &rt->dst;
14e50e57 2524 new->__use = 1;
352e512c 2525 new->input = dst_discard;
ede2059d 2526 new->output = dst_discard_out;
14e50e57 2527
0a1f5962 2528 dst_copy_metrics(new, &ort->dst);
14e50e57 2529
1dbe3252 2530 rt->rt6i_idev = in6_dev_get(loopback_dev);
4e3fd7a0 2531 rt->rt6i_gateway = ort->rt6i_gateway;
0a1f5962 2532 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
14e50e57
DM		 2533
2534 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2535#ifdef CONFIG_IPV6_SUBTREES
2536 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2537#endif
14e50e57
DM		 2538 }
2539
69ead7af
DM		 2540 dst_release(dst_orig);
2541 return new ? new : ERR_PTR(-ENOMEM);
14e50e57 2542}
14e50e57 2543
1da177e4
LT		 2544/*
2545 * Destination cache support functions
2546 */
2547
8d1c802b 2548static bool fib6_check(struct fib6_info *f6i, u32 cookie)
4b32b5ad 2549{
93531c67
DA		 2550 u32 rt_cookie = 0;
2551
8ae86971 2552 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
93531c67
DA		 2553 return false;
2554
2555 if (fib6_check_expired(f6i))
2556 return false;
2557
2558 return true;
4b32b5ad
MKL		 2559}
2560
a68886a6
DA		 2561static struct dst_entry *rt6_check(struct rt6_info *rt,
2562 struct fib6_info *from,
2563 u32 cookie)
3da59bd9 2564{
36143645 2565 u32 rt_cookie = 0;
c5cff856 2566
49d05fe2 2567 if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
93531c67 2568 rt_cookie != cookie)
3da59bd9
MKL		 2569 return NULL;
2570
2571 if (rt6_check_expired(rt))
2572 return NULL;
2573
2574 return &rt->dst;
2575}
2576
a68886a6
DA		 2577static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2578 struct fib6_info *from,
2579 u32 cookie)
3da59bd9 2580{
5973fb1e
MKL		 2581 if (!__rt6_check_expired(rt) &&
2582 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
a68886a6 2583 fib6_check(from, cookie))
3da59bd9
MKL		 2584 return &rt->dst;
2585 else
2586 return NULL;
2587}
2588
1da177e4
LT		 2589static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
2590{
a87b7dc9 2591 struct dst_entry *dst_ret;
a68886a6 2592 struct fib6_info *from;
1da177e4
LT		 2593 struct rt6_info *rt;
2594
a87b7dc9
DA		 2595 rt = container_of(dst, struct rt6_info, dst);
2596
2597 rcu_read_lock();
1da177e4 2598
6f3118b5
ND		 2599 /* All IPV6 dsts are created with ->obsolete set to the value
2600 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2601 * into this function always.
2602 */
e3bc10bd 2603
a68886a6 2604 from = rcu_dereference(rt->from);
4b32b5ad 2605
a68886a6
DA		 2606 if (from && (rt->rt6i_flags & RTF_PCPU ||
2607 unlikely(!list_empty(&rt->rt6i_uncached))))
2608 dst_ret = rt6_dst_from_check(rt, from, cookie);
3da59bd9 2609 else
a68886a6 2610 dst_ret = rt6_check(rt, from, cookie);
a87b7dc9
DA		 2611
2612 rcu_read_unlock();
2613
2614 return dst_ret;
1da177e4
LT		 2615}
2616
2617static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2618{
2619 struct rt6_info *rt = (struct rt6_info *) dst;
2620
2621 if (rt) {
54c1a859 2622 if (rt->rt6i_flags & RTF_CACHE) {
c3c14da0 2623 rcu_read_lock();
54c1a859 2624 if (rt6_check_expired(rt)) {
93531c67 2625 rt6_remove_exception_rt(rt);
54c1a859
YH		 2626 dst = NULL;
2627 }
c3c14da0 2628 rcu_read_unlock();
54c1a859 2629 } else {
1da177e4 2630 dst_release(dst);
54c1a859
YH		 2631 dst = NULL;
2632 }
1da177e4 2633 }
54c1a859 2634 return dst;
1da177e4
LT		 2635}
2636
2637static void ip6_link_failure(struct sk_buff *skb)
2638{
2639 struct rt6_info *rt;
2640
3ffe533c 2641 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1da177e4 2642
adf30907 2643 rt = (struct rt6_info *) skb_dst(skb);
1da177e4 2644 if (rt) {
8a14e46f 2645 rcu_read_lock();
1eb4f758 2646 if (rt->rt6i_flags & RTF_CACHE) {
761f6026 2647 rt6_remove_exception_rt(rt);
c5cff856 2648 } else {
a68886a6 2649 struct fib6_info *from;
c5cff856
WW		 2650 struct fib6_node *fn;
2651
a68886a6
DA		 2652 from = rcu_dereference(rt->from);
2653 if (from) {
2654 fn = rcu_dereference(from->fib6_node);
2655 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2656 fn->fn_sernum = -1;
2657 }
1eb4f758 2658 }
8a14e46f 2659 rcu_read_unlock();
1da177e4
LT		 2660 }
2661}
2662
6a3e030f
DA		 2663static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2664{
a68886a6
DA		 2665 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2666 struct fib6_info *from;
2667
2668 rcu_read_lock();
2669 from = rcu_dereference(rt0->from);
2670 if (from)
2671 rt0->dst.expires = from->expires;
2672 rcu_read_unlock();
2673 }
6a3e030f
DA		 2674
2675 dst_set_expires(&rt0->dst, timeout);
2676 rt0->rt6i_flags |= RTF_EXPIRES;
2677}
2678
45e4fd26
MKL		 2679static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2680{
2681 struct net *net = dev_net(rt->dst.dev);
2682
d4ead6b3 2683 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
45e4fd26 2684 rt->rt6i_flags |= RTF_MODIFIED;
45e4fd26
MKL		 2685 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2686}
2687
0d3f6d29
MKL		 2688static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2689{
2690 return !(rt->rt6i_flags & RTF_CACHE) &&
1490ed2a 2691 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
0d3f6d29
MKL		 2692}
2693
45e4fd26
MKL		 2694static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2695 const struct ipv6hdr *iph, u32 mtu)
1da177e4 2696{
0dec879f 2697 const struct in6_addr *daddr, *saddr;
67ba4152 2698 struct rt6_info *rt6 = (struct rt6_info *)dst;
1da177e4 2699
19bda36c
XL		 2700 if (dst_metric_locked(dst, RTAX_MTU))
2701 return;
2702
0dec879f
JA		 2703 if (iph) {
2704 daddr = &iph->daddr;
2705 saddr = &iph->saddr;
2706 } else if (sk) {
2707 daddr = &sk->sk_v6_daddr;
2708 saddr = &inet6_sk(sk)->saddr;
2709 } else {
2710 daddr = NULL;
2711 saddr = NULL;
2712 }
2713 dst_confirm_neigh(dst, daddr);
45e4fd26
MKL		 2714 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
2715 if (mtu >= dst_mtu(dst))
2716 return;
9d289715 2717
0d3f6d29 2718 if (!rt6_cache_allowed_for_pmtu(rt6)) {
45e4fd26 2719 rt6_do_update_pmtu(rt6, mtu);
2b760fcf
WW		 2720 /* update rt6_ex->stamp for cache */
2721 if (rt6->rt6i_flags & RTF_CACHE)
2722 rt6_update_exception_stamp_rt(rt6);
0dec879f 2723 } else if (daddr) {
85bd05de 2724 struct fib6_result res = {};
45e4fd26
MKL		 2725 struct rt6_info *nrt6;
2726
4d85cd0c 2727 rcu_read_lock();
85bd05de 2728 res.f6i = rcu_dereference(rt6->from);
43a4b60d
DA		 2729 if (!res.f6i)
2730 goto out_unlock;
2731
7d21fec9
DA		 2732 res.fib6_flags = res.f6i->fib6_flags;
2733 res.fib6_type = res.f6i->fib6_type;
2734
2d44234b
DA		 2735 if (res.f6i->nh) {
2736 struct fib6_nh_match_arg arg = {
2737 .dev = dst->dev,
2738 .gw = &rt6->rt6i_gateway,
2739 };
2740
2741 nexthop_for_each_fib6_nh(res.f6i->nh,
2742 fib6_nh_find_match, &arg);
2743
2744 /* fib6_info uses a nexthop that does not have fib6_nh
2745 * using the dst->dev + gw. Should be impossible.
2746 */
43a4b60d
DA		 2747 if (!arg.match)
2748 goto out_unlock;
2d44234b
DA		 2749
2750 res.nh = arg.match;
2751 } else {
2752 res.nh = res.f6i->fib6_nh;
2753 }
2754
85bd05de 2755 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
45e4fd26
MKL		 2756 if (nrt6) {
2757 rt6_do_update_pmtu(nrt6, mtu);
5012f0a5 2758 if (rt6_insert_exception(nrt6, &res))
2b760fcf 2759 dst_release_immediate(&nrt6->dst);
45e4fd26 2760 }
43a4b60d 2761out_unlock:
a68886a6 2762 rcu_read_unlock();
1da177e4
LT		 2763 }
2764}
2765
45e4fd26
MKL		 2766static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2767 struct sk_buff *skb, u32 mtu)
2768{
2769 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
2770}
2771
42ae66c8 2772void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
e2d118a1 2773 int oif, u32 mark, kuid_t uid)
81aded24
DM		 2774{
2775 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2776 struct dst_entry *dst;
dc92095d
MŻ		 2777 struct flowi6 fl6 = {
2778 .flowi6_oif = oif,
2779 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2780 .daddr = iph->daddr,
2781 .saddr = iph->saddr,
2782 .flowlabel = ip6_flowinfo(iph),
2783 .flowi6_uid = uid,
2784 };
81aded24
DM		 2785
2786 dst = ip6_route_output(net, NULL, &fl6);
2787 if (!dst->error)
45e4fd26 2788 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
81aded24
DM		 2789 dst_release(dst);
2790}
2791EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2792
2793void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2794{
7ddacfa5 2795 int oif = sk->sk_bound_dev_if;
33c162a9
MKL		 2796 struct dst_entry *dst;
2797
7ddacfa5
DA		 2798 if (!oif && skb->dev)
2799 oif = l3mdev_master_ifindex(skb->dev);
2800
2801 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
33c162a9
MKL		 2802
2803 dst = __sk_dst_get(sk);
2804 if (!dst || !dst->obsolete ||
2805 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2806 return;
2807
2808 bh_lock_sock(sk);
2809 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2810 ip6_datagram_dst_update(sk, false);
2811 bh_unlock_sock(sk);
81aded24
DM		 2812}
2813EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2814
7d6850f7
AK		 2815void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2816 const struct flowi6 *fl6)
2817{
2818#ifdef CONFIG_IPV6_SUBTREES
2819 struct ipv6_pinfo *np = inet6_sk(sk);
2820#endif
2821
2822 ip6_dst_store(sk, dst,
2823 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2824 &sk->sk_v6_daddr : NULL,
2825#ifdef CONFIG_IPV6_SUBTREES
2826 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2827 &np->saddr :
2828#endif
2829 NULL);
2830}
2831
9b6b35ab 2832static bool ip6_redirect_nh_match(const struct fib6_result *res,
0b34eb00
DA		 2833 struct flowi6 *fl6,
2834 const struct in6_addr *gw,
2835 struct rt6_info **ret)
2836{
9b6b35ab
DA		 2837 const struct fib6_nh *nh = res->nh;
2838
0b34eb00
DA		 2839 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2840 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2841 return false;
2842
2843 /* rt_cache's gateway might be different from its 'parent'
2844 * in the case of an ip redirect.
2845 * So we keep searching in the exception table if the gateway
2846 * is different.
2847 */
2848 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
2849 struct rt6_info *rt_cache;
2850
9b6b35ab 2851 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
0b34eb00
DA		 2852 if (rt_cache &&
2853 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
2854 *ret = rt_cache;
2855 return true;
2856 }
2857 return false;
2858 }
2859 return true;
2860}
2861
c55c8988
DA		 2862struct fib6_nh_rd_arg {
2863 struct fib6_result *res;
2864 struct flowi6 *fl6;
2865 const struct in6_addr *gw;
2866 struct rt6_info **ret;
2867};
2868
2869static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
2870{
2871 struct fib6_nh_rd_arg *arg = _arg;
2872
2873 arg->res->nh = nh;
2874 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
2875}
2876
b55b76b2
DJ		 2877/* Handle redirects */
2878struct ip6rd_flowi {
2879 struct flowi6 fl6;
2880 struct in6_addr gateway;
2881};
2882
2883static struct rt6_info *__ip6_route_redirect(struct net *net,
2884 struct fib6_table *table,
2885 struct flowi6 *fl6,
b75cc8f9 2886 const struct sk_buff *skb,
b55b76b2
DJ		 2887 int flags)
2888{
2889 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
0b34eb00 2890 struct rt6_info *ret = NULL;
9b6b35ab 2891 struct fib6_result res = {};
c55c8988
DA		 2892 struct fib6_nh_rd_arg arg = {
2893 .res = &res,
2894 .fl6 = fl6,
2895 .gw = &rdfl->gateway,
2896 .ret = &ret
2897 };
8d1c802b 2898 struct fib6_info *rt;
b55b76b2
DJ		 2899 struct fib6_node *fn;
2900
31680ac2
DA		 2901 /* l3mdev_update_flow overrides oif if the device is enslaved; in
2902 * this case we must match on the real ingress device, so reset it
2903 */
2904 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2905 fl6->flowi6_oif = skb->dev->ifindex;
2906
b55b76b2 2907 /* Get the "current" route for this destination and
67c408cf 2908 * check if the redirect has come from appropriate router.
b55b76b2
DJ		 2909 *
2910 * RFC 4861 specifies that redirects should only be
2911 * accepted if they come from the nexthop to the target.
2912 * Due to the way the routes are chosen, this notion
2913 * is a bit fuzzy and one might need to check all possible
2914 * routes.
2915 */
2916
66f5d6ce 2917 rcu_read_lock();
6454743b 2918 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
b55b76b2 2919restart:
66f5d6ce 2920 for_each_fib6_node_rt_rcu(fn) {
9b6b35ab 2921 res.f6i = rt;
14895687 2922 if (fib6_check_expired(rt))
b55b76b2 2923 continue;
93c2fb25 2924 if (rt->fib6_flags & RTF_REJECT)
b55b76b2 2925 break;
c55c8988
DA		 2926 if (unlikely(rt->nh)) {
2927 if (nexthop_is_blackhole(rt->nh))
2928 continue;
2929 /* on match, res->nh is filled in and potentially ret */
2930 if (nexthop_for_each_fib6_nh(rt->nh,
2931 fib6_nh_redirect_match,
2932 &arg))
2933 goto out;
2934 } else {
2935 res.nh = rt->fib6_nh;
2936 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
2937 &ret))
2938 goto out;
2939 }
b55b76b2
DJ		 2940 }
2941
2942 if (!rt)
421842ed 2943 rt = net->ipv6.fib6_null_entry;
93c2fb25 2944 else if (rt->fib6_flags & RTF_REJECT) {
23fb93a4 2945 ret = net->ipv6.ip6_null_entry;
b0a1ba59
MKL		 2946 goto out;
2947 }
2948
421842ed 2949 if (rt == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2950 fn = fib6_backtrack(fn, &fl6->saddr);
2951 if (fn)
2952 goto restart;
b55b76b2 2953 }
a3c00e46 2954
9b6b35ab 2955 res.f6i = rt;
1cf844c7 2956 res.nh = rt->fib6_nh;
b0a1ba59 2957out:
7d21fec9 2958 if (ret) {
10585b43 2959 ip6_hold_safe(net, &ret);
7d21fec9
DA		 2960 } else {
2961 res.fib6_flags = res.f6i->fib6_flags;
2962 res.fib6_type = res.f6i->fib6_type;
9b6b35ab 2963 ret = ip6_create_rt_rcu(&res);
7d21fec9 2964 }
b55b76b2 2965
66f5d6ce 2966 rcu_read_unlock();
b55b76b2 2967
8ff2e5b2 2968 trace_fib6_table_lookup(net, &res, table, fl6);
23fb93a4 2969 return ret;
b55b76b2
DJ		 2970};
2971
2972static struct dst_entry *ip6_route_redirect(struct net *net,
b75cc8f9
DA		 2973 const struct flowi6 *fl6,
2974 const struct sk_buff *skb,
2975 const struct in6_addr *gateway)
b55b76b2
DJ		 2976{
2977 int flags = RT6_LOOKUP_F_HAS_SADDR;
2978 struct ip6rd_flowi rdfl;
2979
2980 rdfl.fl6 = *fl6;
2981 rdfl.gateway = *gateway;
2982
b75cc8f9 2983 return fib6_rule_lookup(net, &rdfl.fl6, skb,
b55b76b2
DJ		 2984 flags, __ip6_route_redirect);
2985}
2986
e2d118a1
LC		 2987void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
2988 kuid_t uid)
3a5ad2ee
DM		 2989{
2990 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2991 struct dst_entry *dst;
1f7f10ac
MŻ		 2992 struct flowi6 fl6 = {
2993 .flowi6_iif = LOOPBACK_IFINDEX,
2994 .flowi6_oif = oif,
2995 .flowi6_mark = mark,
2996 .daddr = iph->daddr,
2997 .saddr = iph->saddr,
2998 .flowlabel = ip6_flowinfo(iph),
2999 .flowi6_uid = uid,
3000 };
3a5ad2ee 3001
b75cc8f9 3002 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
b55b76b2 3003 rt6_do_redirect(dst, NULL, skb);
3a5ad2ee
DM		 3004 dst_release(dst);
3005}
3006EXPORT_SYMBOL_GPL(ip6_redirect);
3007
d456336d 3008void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
c92a59ec
DJ		 3009{
3010 const struct ipv6hdr *iph = ipv6_hdr(skb);
3011 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3012 struct dst_entry *dst;
0b26fb17
MŻ		 3013 struct flowi6 fl6 = {
3014 .flowi6_iif = LOOPBACK_IFINDEX,
3015 .flowi6_oif = oif,
0b26fb17
MŻ		 3016 .daddr = msg->dest,
3017 .saddr = iph->daddr,
3018 .flowi6_uid = sock_net_uid(net, NULL),
3019 };
c92a59ec 3020
b75cc8f9 3021 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
b55b76b2 3022 rt6_do_redirect(dst, NULL, skb);
c92a59ec
DJ		 3023 dst_release(dst);
3024}
3025
3a5ad2ee
DM		 3026void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3027{
e2d118a1
LC		 3028 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
3029 sk->sk_uid);
3a5ad2ee
DM		 3030}
3031EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3032
0dbaee3b 3033static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1da177e4 3034{
0dbaee3b
DM		 3035 struct net_device *dev = dst->dev;
3036 unsigned int mtu = dst_mtu(dst);
3037 struct net *net = dev_net(dev);
3038
1da177e4
LT		 3039 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3040
5578689a
DL		 3041 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3042 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1da177e4
LT		 3043
3044 /*
1ab1457c
YH		 3045 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3046 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3047 * IPV6_MAXPLEN is also valid and means: "any MSS,
1da177e4
LT		 3048 * rely only on pmtu discovery"
3049 */
3050 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3051 mtu = IPV6_MAXPLEN;
3052 return mtu;
3053}
3054
ebb762f2 3055static unsigned int ip6_mtu(const struct dst_entry *dst)
d33e4553 3056{
d33e4553 3057 struct inet6_dev *idev;
d4ead6b3 3058 unsigned int mtu;
4b32b5ad
MKL		 3059
3060 mtu = dst_metric_raw(dst, RTAX_MTU);
618f9bc7 3061 if (mtu)
30f78d8e 3062 goto out;
618f9bc7
SK		 3063
3064 mtu = IPV6_MIN_MTU;
d33e4553
DM		 3065
3066 rcu_read_lock();
3067 idev = __in6_dev_get(dst->dev);
3068 if (idev)
3069 mtu = idev->cnf.mtu6;
3070 rcu_read_unlock();
3071
30f78d8e 3072out:
14972cbd
RP		 3073 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3074
3075 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
d33e4553
DM		 3076}
3077
901731b8
DA		 3078/* MTU selection:
3079 * 1. mtu on route is locked - use it
3080 * 2. mtu from nexthop exception
3081 * 3. mtu from egress device
3082 *
3083 * based on ip6_dst_mtu_forward and exception logic of
3084 * rt6_find_cached_rt; called with rcu_read_lock
3085 */
b748f260
DA		 3086u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3087 const struct in6_addr *daddr,
3088 const struct in6_addr *saddr)
901731b8 3089{
b748f260
DA		 3090 const struct fib6_nh *nh = res->nh;
3091 struct fib6_info *f6i = res->f6i;
901731b8 3092 struct inet6_dev *idev;
510e2ced 3093 struct rt6_info *rt;
901731b8
DA		 3094 u32 mtu = 0;
3095
3096 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3097 mtu = f6i->fib6_pmtu;
3098 if (mtu)
3099 goto out;
3100 }
3101
510e2ced
WW		 3102 rt = rt6_find_cached_rt(res, daddr, saddr);
3103 if (unlikely(rt)) {
3104 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3105 } else {
b748f260 3106 struct net_device *dev = nh->fib_nh_dev;
901731b8
DA		 3107
3108 mtu = IPV6_MIN_MTU;
3109 idev = __in6_dev_get(dev);
3110 if (idev && idev->cnf.mtu6 > mtu)
3111 mtu = idev->cnf.mtu6;
3112 }
3113
3114 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3115out:
b748f260 3116 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
901731b8
DA		 3117}
3118
3b00944c 3119struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
87a11578 3120 struct flowi6 *fl6)
1da177e4 3121{
87a11578 3122 struct dst_entry *dst;
1da177e4
LT		 3123 struct rt6_info *rt;
3124 struct inet6_dev *idev = in6_dev_get(dev);
c346dca1 3125 struct net *net = dev_net(dev);
1da177e4 3126
38308473 3127 if (unlikely(!idev))
122bdf67 3128 return ERR_PTR(-ENODEV);
1da177e4 3129
ad706862 3130 rt = ip6_dst_alloc(net, dev, 0);
38308473 3131 if (unlikely(!rt)) {
1da177e4 3132 in6_dev_put(idev);
87a11578 3133 dst = ERR_PTR(-ENOMEM);
1da177e4
LT		 3134 goto out;
3135 }
3136
8e2ec639 3137 rt->dst.flags |= DST_HOST;
588753f1 3138 rt->dst.input = ip6_input;
8e2ec639 3139 rt->dst.output = ip6_output;
550bab42 3140 rt->rt6i_gateway = fl6->daddr;
87a11578 3141 rt->rt6i_dst.addr = fl6->daddr;
8e2ec639
YZ		 3142 rt->rt6i_dst.plen = 128;
3143 rt->rt6i_idev = idev;
14edd87d 3144 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1da177e4 3145
4c981e28 3146 /* Add this dst into uncached_list so that rt6_disable_ip() can
587fea74
WW		 3147 * do proper release of the net_device
3148 */
3149 rt6_uncached_list_add(rt);
81eb8447 3150 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
1da177e4 3151
87a11578
DM		 3152 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3153
1da177e4 3154out:
87a11578 3155 return dst;
1da177e4
LT		 3156}
3157
569d3645 3158static int ip6_dst_gc(struct dst_ops *ops)
1da177e4 3159{
86393e52 3160 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
7019b78e
DL		 3161 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3162 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3163 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3164 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3165 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
fc66f95c 3166 int entries;
7019b78e 3167
fc66f95c 3168 entries = dst_entries_get_fast(ops);
49a18d86 3169 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
fc66f95c 3170 entries <= rt_max_size)
1da177e4
LT		 3171 goto out;
3172
6891a346 3173 net->ipv6.ip6_rt_gc_expire++;
14956643 3174 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
fc66f95c
ED		 3175 entries = dst_entries_get_slow(ops);
3176 if (entries < ops->gc_thresh)
7019b78e 3177 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1da177e4 3178out:
7019b78e 3179 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
fc66f95c 3180 return entries > rt_max_size;
1da177e4
LT		 3181}
3182
b2c709cc
DA		 3183static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3184 const struct in6_addr *gw_addr, u32 tbid,
3185 int flags, struct fib6_result *res)
8c14586f
DA		 3186{
3187 struct flowi6 fl6 = {
3188 .flowi6_oif = cfg->fc_ifindex,
3189 .daddr = *gw_addr,
3190 .saddr = cfg->fc_prefsrc,
3191 };
3192 struct fib6_table *table;
b2c709cc 3193 int err;
8c14586f 3194
f4797b33 3195 table = fib6_get_table(net, tbid);
8c14586f 3196 if (!table)
b2c709cc 3197 return -EINVAL;
8c14586f
DA		 3198
3199 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3200 flags |= RT6_LOOKUP_F_HAS_SADDR;
3201
f4797b33 3202 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
8c14586f 3203
b2c709cc
DA		 3204 err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3205 if (!err && res->f6i != net->ipv6.fib6_null_entry)
3206 fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3207 cfg->fc_ifindex != 0, NULL, flags);
8c14586f 3208
b2c709cc 3209 return err;
8c14586f
DA		 3210}
3211
fc1e64e1
DA		 3212static int ip6_route_check_nh_onlink(struct net *net,
3213 struct fib6_config *cfg,
9fbb704c 3214 const struct net_device *dev,
fc1e64e1
DA		 3215 struct netlink_ext_ack *extack)
3216{
b2c709cc 3217 u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
fc1e64e1 3218 const struct in6_addr *gw_addr = &cfg->fc_gateway;
b2c709cc 3219 struct fib6_result res = {};
fc1e64e1
DA		 3220 int err;
3221
b2c709cc
DA		 3222 err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3223 if (!err && !(res.fib6_flags & RTF_REJECT) &&
3224 /* ignore match if it is the default route */
3225 !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3226 (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3227 NL_SET_ERR_MSG(extack,
3228 "Nexthop has invalid gateway or device mismatch");
3229 err = -EINVAL;
fc1e64e1
DA		 3230 }
3231
3232 return err;
3233}
3234
1edce99f
DA		 3235static int ip6_route_check_nh(struct net *net,
3236 struct fib6_config *cfg,
3237 struct net_device **_dev,
3238 struct inet6_dev **idev)
3239{
3240 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3241 struct net_device *dev = _dev ? *_dev : NULL;
b2c709cc
DA		 3242 int flags = RT6_LOOKUP_F_IFACE;
3243 struct fib6_result res = {};
1edce99f
DA		 3244 int err = -EHOSTUNREACH;
3245
3246 if (cfg->fc_table) {
b2c709cc
DA		 3247 err = ip6_nh_lookup_table(net, cfg, gw_addr,
3248 cfg->fc_table, flags, &res);
3249 /* gw_addr can not require a gateway or resolve to a reject
3250 * route. If a device is given, it must match the result.
3251 */
3252 if (err || res.fib6_flags & RTF_REJECT ||
3253 res.nh->fib_nh_gw_family ||
3254 (dev && dev != res.nh->fib_nh_dev))
3255 err = -EHOSTUNREACH;
1edce99f
DA		 3256 }
3257
b2c709cc
DA		 3258 if (err < 0) {
3259 struct flowi6 fl6 = {
3260 .flowi6_oif = cfg->fc_ifindex,
3261 .daddr = *gw_addr,
3262 };
1edce99f 3263
b2c709cc
DA		 3264 err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3265 if (err || res.fib6_flags & RTF_REJECT ||
3266 res.nh->fib_nh_gw_family)
3267 err = -EHOSTUNREACH;
3268
3269 if (err)
3270 return err;
3271
3272 fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3273 cfg->fc_ifindex != 0, NULL, flags);
3274 }
1edce99f 3275
b2c709cc 3276 err = 0;
1edce99f 3277 if (dev) {
b2c709cc
DA		 3278 if (dev != res.nh->fib_nh_dev)
3279 err = -EHOSTUNREACH;
1edce99f 3280 } else {
b2c709cc 3281 *_dev = dev = res.nh->fib_nh_dev;
1edce99f 3282 dev_hold(dev);
b2c709cc 3283 *idev = in6_dev_get(dev);
1edce99f
DA		 3284 }
3285
1edce99f
DA		 3286 return err;
3287}
3288
9fbb704c
DA		 3289static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3290 struct net_device **_dev, struct inet6_dev **idev,
3291 struct netlink_ext_ack *extack)
3292{
3293 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3294 int gwa_type = ipv6_addr_type(gw_addr);
232378e8 3295 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
9fbb704c 3296 const struct net_device *dev = *_dev;
232378e8 3297 bool need_addr_check = !dev;
9fbb704c
DA		 3298 int err = -EINVAL;
3299
3300 /* if gw_addr is local we will fail to detect this in case
3301 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3302 * will return already-added prefix route via interface that
3303 * prefix route was assigned to, which might be non-loopback.
3304 */
232378e8
DA		 3305 if (dev &&
3306 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3307 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
9fbb704c
DA		 3308 goto out;
3309 }
3310
3311 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3312 /* IPv6 strictly inhibits using not link-local
3313 * addresses as nexthop address.
3314 * Otherwise, router will not able to send redirects.
3315 * It is very good, but in some (rare!) circumstances
3316 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3317 * some exceptions. --ANK
3318 * We allow IPv4-mapped nexthops to support RFC4798-type
3319 * addressing
3320 */
3321 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3322 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3323 goto out;
3324 }
3325
b2c709cc
DA		 3326 rcu_read_lock();
3327
9fbb704c
DA		 3328 if (cfg->fc_flags & RTNH_F_ONLINK)
3329 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3330 else
3331 err = ip6_route_check_nh(net, cfg, _dev, idev);
3332
b2c709cc
DA		 3333 rcu_read_unlock();
3334
9fbb704c
DA		 3335 if (err)
3336 goto out;
3337 }
3338
3339 /* reload in case device was changed */
3340 dev = *_dev;
3341
3342 err = -EINVAL;
3343 if (!dev) {
3344 NL_SET_ERR_MSG(extack, "Egress device not specified");
3345 goto out;
3346 } else if (dev->flags & IFF_LOOPBACK) {
3347 NL_SET_ERR_MSG(extack,
3348 "Egress device can not be loopback device for this route");
3349 goto out;
3350 }
232378e8
DA		 3351
3352 /* if we did not check gw_addr above, do so now that the
3353 * egress device has been resolved.
3354 */
3355 if (need_addr_check &&
3356 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3357 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3358 goto out;
3359 }
3360
9fbb704c
DA		 3361 err = 0;
3362out:
3363 return err;
3364}
3365
83c44251
DA		 3366static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3367{
3368 if ((flags & RTF_REJECT) ||
3369 (dev && (dev->flags & IFF_LOOPBACK) &&
3370 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3371 !(flags & RTF_LOCAL)))
3372 return true;
3373
3374 return false;
3375}
3376
3377int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3378 struct fib6_config *cfg, gfp_t gfp_flags,
3379 struct netlink_ext_ack *extack)
3380{
3381 struct net_device *dev = NULL;
3382 struct inet6_dev *idev = NULL;
3383 int addr_type;
3384 int err;
3385
f1741730 3386 fib6_nh->fib_nh_family = AF_INET6;
1bef4c22
ED		 3387#ifdef CONFIG_IPV6_ROUTER_PREF
3388 fib6_nh->last_probe = jiffies;
3389#endif
f1741730 3390
83c44251
DA		 3391 err = -ENODEV;
3392 if (cfg->fc_ifindex) {
3393 dev = dev_get_by_index(net, cfg->fc_ifindex);
3394 if (!dev)
3395 goto out;
3396 idev = in6_dev_get(dev);
3397 if (!idev)
3398 goto out;
3399 }
3400
3401 if (cfg->fc_flags & RTNH_F_ONLINK) {
3402 if (!dev) {
3403 NL_SET_ERR_MSG(extack,
3404 "Nexthop device required for onlink");
3405 goto out;
3406 }
3407
3408 if (!(dev->flags & IFF_UP)) {
3409 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3410 err = -ENETDOWN;
3411 goto out;
3412 }
3413
ad1601ae 3414 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
83c44251
DA		 3415 }
3416
ad1601ae 3417 fib6_nh->fib_nh_weight = 1;
83c44251
DA		 3418
3419 /* We cannot add true routes via loopback here,
3420 * they would result in kernel looping; promote them to reject routes
3421 */
3422 addr_type = ipv6_addr_type(&cfg->fc_dst);
3423 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3424 /* hold loopback dev/idev if we haven't done so. */
3425 if (dev != net->loopback_dev) {
3426 if (dev) {
3427 dev_put(dev);
3428 in6_dev_put(idev);
3429 }
3430 dev = net->loopback_dev;
3431 dev_hold(dev);
3432 idev = in6_dev_get(dev);
3433 if (!idev) {
3434 err = -ENODEV;
3435 goto out;
3436 }
3437 }
7dd73168 3438 goto pcpu_alloc;
83c44251
DA		 3439 }
3440
3441 if (cfg->fc_flags & RTF_GATEWAY) {
3442 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3443 if (err)
3444 goto out;
3445
ad1601ae 3446 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
bdf00467 3447 fib6_nh->fib_nh_gw_family = AF_INET6;
83c44251
DA		 3448 }
3449
3450 err = -ENODEV;
3451 if (!dev)
3452 goto out;
3453
3454 if (idev->cnf.disable_ipv6) {
3455 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3456 err = -EACCES;
3457 goto out;
3458 }
3459
3460 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3461 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3462 err = -ENETDOWN;
3463 goto out;
3464 }
3465
3466 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3467 !netif_carrier_ok(dev))
ad1601ae 3468 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
83c44251 3469
7dd73168
DA		 3470 err = fib_nh_common_init(&fib6_nh->nh_common, cfg->fc_encap,
3471 cfg->fc_encap_type, cfg, gfp_flags, extack);
3472 if (err)
3473 goto out;
3474
3475pcpu_alloc:
f40b6ae2
DA		 3476 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3477 if (!fib6_nh->rt6i_pcpu) {
3478 err = -ENOMEM;
3479 goto out;
3480 }
3481
ad1601ae 3482 fib6_nh->fib_nh_dev = dev;
f1741730 3483 fib6_nh->fib_nh_oif = dev->ifindex;
83c44251
DA		 3484 err = 0;
3485out:
3486 if (idev)
3487 in6_dev_put(idev);
3488
3489 if (err) {
ad1601ae
DA		 3490 lwtstate_put(fib6_nh->fib_nh_lws);
3491 fib6_nh->fib_nh_lws = NULL;
83c44251
DA		 3492 if (dev)
3493 dev_put(dev);
3494 }
3495
3496 return err;
3497}
3498
dac7d0f2
DA		 3499void fib6_nh_release(struct fib6_nh *fib6_nh)
3500{
cc5c073a
DA		 3501 struct rt6_exception_bucket *bucket;
3502
3503 rcu_read_lock();
3504
3505 fib6_nh_flush_exceptions(fib6_nh, NULL);
3506 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3507 if (bucket) {
3508 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3509 kfree(bucket);
3510 }
3511
3512 rcu_read_unlock();
3513
f40b6ae2
DA		 3514 if (fib6_nh->rt6i_pcpu) {
3515 int cpu;
3516
3517 for_each_possible_cpu(cpu) {
3518 struct rt6_info **ppcpu_rt;
3519 struct rt6_info *pcpu_rt;
3520
3521 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3522 pcpu_rt = *ppcpu_rt;
3523 if (pcpu_rt) {
3524 dst_dev_put(&pcpu_rt->dst);
3525 dst_release(&pcpu_rt->dst);
3526 *ppcpu_rt = NULL;
3527 }
3528 }
3529
3530 free_percpu(fib6_nh->rt6i_pcpu);
3531 }
3532
979e276e 3533 fib_nh_common_release(&fib6_nh->nh_common);
dac7d0f2
DA		 3534}
3535
8d1c802b 3536static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
acb54e3c 3537 gfp_t gfp_flags,
333c4301 3538 struct netlink_ext_ack *extack)
1da177e4 3539{
5578689a 3540 struct net *net = cfg->fc_nlinfo.nl_net;
8d1c802b 3541 struct fib6_info *rt = NULL;
f88d8ea6 3542 struct nexthop *nh = NULL;
c71099ac 3543 struct fib6_table *table;
f88d8ea6 3544 struct fib6_nh *fib6_nh;
8c5b83f0 3545 int err = -EINVAL;
83c44251 3546 int addr_type;
1da177e4 3547
557c44be 3548 /* RTF_PCPU is an internal flag; can not be set by userspace */
d5d531cb
DA		 3549 if (cfg->fc_flags & RTF_PCPU) {
3550 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
557c44be 3551 goto out;
d5d531cb 3552 }
557c44be 3553
2ea2352e
WW		 3554 /* RTF_CACHE is an internal flag; can not be set by userspace */
3555 if (cfg->fc_flags & RTF_CACHE) {
3556 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3557 goto out;
3558 }
3559
e8478e80
DA		 3560 if (cfg->fc_type > RTN_MAX) {
3561 NL_SET_ERR_MSG(extack, "Invalid route type");
3562 goto out;
3563 }
3564
d5d531cb
DA		 3565 if (cfg->fc_dst_len > 128) {
3566 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3567 goto out;
3568 }
3569 if (cfg->fc_src_len > 128) {
3570 NL_SET_ERR_MSG(extack, "Invalid source address length");
8c5b83f0 3571 goto out;
d5d531cb 3572 }
1da177e4 3573#ifndef CONFIG_IPV6_SUBTREES
d5d531cb
DA		 3574 if (cfg->fc_src_len) {
3575 NL_SET_ERR_MSG(extack,
3576 "Specifying source address requires IPV6_SUBTREES to be enabled");
8c5b83f0 3577 goto out;
d5d531cb 3578 }
1da177e4 3579#endif
5b98324e
DA		 3580 if (cfg->fc_nh_id) {
3581 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3582 if (!nh) {
3583 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3584 goto out;
3585 }
3586 err = fib6_check_nexthop(nh, cfg, extack);
3587 if (err)
3588 goto out;
3589 }
fc1e64e1 3590
d71314b4 3591 err = -ENOBUFS;
38308473
DM		 3592 if (cfg->fc_nlinfo.nlh &&
3593 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
d71314b4 3594 table = fib6_get_table(net, cfg->fc_table);
38308473 3595 if (!table) {
f3213831 3596 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
d71314b4
MV		 3597 table = fib6_new_table(net, cfg->fc_table);
3598 }
3599 } else {
3600 table = fib6_new_table(net, cfg->fc_table);
3601 }
38308473
DM		 3602
3603 if (!table)
c71099ac 3604 goto out;
c71099ac 3605
93531c67 3606 err = -ENOMEM;
f88d8ea6 3607 rt = fib6_info_alloc(gfp_flags, !nh);
93531c67 3608 if (!rt)
1da177e4 3609 goto out;
93531c67 3610
d7e774f3
DA		 3611 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3612 extack);
767a2217
DA		 3613 if (IS_ERR(rt->fib6_metrics)) {
3614 err = PTR_ERR(rt->fib6_metrics);
fda21d46
ED		 3615 /* Do not leave garbage there. */
3616 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
767a2217
DA		 3617 goto out;
3618 }
3619
93531c67
DA		 3620 if (cfg->fc_flags & RTF_ADDRCONF)
3621 rt->dst_nocount = true;
1da177e4 3622
1716a961 3623 if (cfg->fc_flags & RTF_EXPIRES)
14895687 3624 fib6_set_expires(rt, jiffies +
1716a961
G		 3625 clock_t_to_jiffies(cfg->fc_expires));
3626 else
14895687 3627 fib6_clean_expires(rt);
1da177e4 3628
86872cb5
TG		 3629 if (cfg->fc_protocol == RTPROT_UNSPEC)
3630 cfg->fc_protocol = RTPROT_BOOT;
93c2fb25 3631 rt->fib6_protocol = cfg->fc_protocol;
86872cb5 3632
83c44251
DA		 3633 rt->fib6_table = table;
3634 rt->fib6_metric = cfg->fc_metric;
c7036d97 3635 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
2b2450ca 3636 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
19e42e45 3637
93c2fb25
DA		 3638 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3639 rt->fib6_dst.plen = cfg->fc_dst_len;
3640 if (rt->fib6_dst.plen == 128)
3b6761d1 3641 rt->dst_host = true;
e5fd387a 3642
1da177e4 3643#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 3644 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3645 rt->fib6_src.plen = cfg->fc_src_len;
1da177e4 3646#endif
f88d8ea6
DA		 3647 if (nh) {
3648 if (!nexthop_get(nh)) {
3649 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3650 goto out;
3651 }
3652 if (rt->fib6_src.plen) {
4daa95af 3653 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
f88d8ea6
DA		 3654 goto out;
3655 }
3656 rt->nh = nh;
3657 fib6_nh = nexthop_fib6_nh(rt->nh);
3658 } else {
3659 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3660 if (err)
3661 goto out;
1da177e4 3662
f88d8ea6
DA		 3663 fib6_nh = rt->fib6_nh;
3664
3665 /* We cannot add true routes via loopback here, they would
3666 * result in kernel looping; promote them to reject routes
3667 */
3668 addr_type = ipv6_addr_type(&cfg->fc_dst);
3669 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3670 addr_type))
3671 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3672 }
955ec4cb 3673
c3968a85 3674 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
f88d8ea6 3675 struct net_device *dev = fib6_nh->fib_nh_dev;
83c44251 3676
c3968a85 3677 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
d5d531cb 3678 NL_SET_ERR_MSG(extack, "Invalid source address");
c3968a85
DW		 3679 err = -EINVAL;
3680 goto out;
3681 }
93c2fb25
DA		 3682 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3683 rt->fib6_prefsrc.plen = 128;
c3968a85 3684 } else
93c2fb25 3685 rt->fib6_prefsrc.plen = 0;
c3968a85 3686
8c5b83f0 3687 return rt;
6b9ea5a6 3688out:
93531c67 3689 fib6_info_release(rt);
8c5b83f0 3690 return ERR_PTR(err);
6b9ea5a6
RP		 3691}
3692
acb54e3c 3693int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
333c4301 3694 struct netlink_ext_ack *extack)
6b9ea5a6 3695{
8d1c802b 3696 struct fib6_info *rt;
6b9ea5a6
RP		 3697 int err;
3698
acb54e3c 3699 rt = ip6_route_info_create(cfg, gfp_flags, extack);
d4ead6b3
DA		 3700 if (IS_ERR(rt))
3701 return PTR_ERR(rt);
6b9ea5a6 3702
d4ead6b3 3703 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
93531c67 3704 fib6_info_release(rt);
6b9ea5a6 3705
1da177e4
LT		 3706 return err;
3707}
3708
8d1c802b 3709static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
1da177e4 3710{
afb1d4b5 3711 struct net *net = info->nl_net;
c71099ac 3712 struct fib6_table *table;
afb1d4b5 3713 int err;
1da177e4 3714
421842ed 3715 if (rt == net->ipv6.fib6_null_entry) {
6825a26c
G		 3716 err = -ENOENT;
3717 goto out;
3718 }
6c813a72 3719
93c2fb25 3720 table = rt->fib6_table;
66f5d6ce 3721 spin_lock_bh(&table->tb6_lock);
86872cb5 3722 err = fib6_del(rt, info);
66f5d6ce 3723 spin_unlock_bh(&table->tb6_lock);
1da177e4 3724
6825a26c 3725out:
93531c67 3726 fib6_info_release(rt);
1da177e4
LT		 3727 return err;
3728}
3729
8d1c802b 3730int ip6_del_rt(struct net *net, struct fib6_info *rt)
e0a1ad73 3731{
afb1d4b5
DA		 3732 struct nl_info info = { .nl_net = net };
3733
528c4ceb 3734 return __ip6_del_rt(rt, &info);
e0a1ad73
TG		 3735}
3736
8d1c802b 3737static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
0ae81335
DA		 3738{
3739 struct nl_info *info = &cfg->fc_nlinfo;
e3330039 3740 struct net *net = info->nl_net;
16a16cd3 3741 struct sk_buff *skb = NULL;
0ae81335 3742 struct fib6_table *table;
e3330039 3743 int err = -ENOENT;
0ae81335 3744
421842ed 3745 if (rt == net->ipv6.fib6_null_entry)
e3330039 3746 goto out_put;
93c2fb25 3747 table = rt->fib6_table;
66f5d6ce 3748 spin_lock_bh(&table->tb6_lock);
0ae81335 3749
93c2fb25 3750 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
8d1c802b 3751 struct fib6_info *sibling, *next_sibling;
0284696b 3752 struct fib6_node *fn;
0ae81335 3753
16a16cd3
DA		 3754 /* prefer to send a single notification with all hops */
3755 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3756 if (skb) {
3757 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3758
d4ead6b3 3759 if (rt6_fill_node(net, skb, rt, NULL,
16a16cd3
DA		 3760 NULL, NULL, 0, RTM_DELROUTE,
3761 info->portid, seq, 0) < 0) {
3762 kfree_skb(skb);
3763 skb = NULL;
3764 } else
3765 info->skip_notify = 1;
3766 }
3767
0284696b
IS		 3768 /* 'rt' points to the first sibling route. If it is not the
3769 * leaf, then we do not need to send a notification. Otherwise,
3770 * we need to check if the last sibling has a next route or not
3771 * and emit a replace or delete notification, respectively.
3772 */
2881fd61 3773 info->skip_notify_kernel = 1;
0284696b
IS		 3774 fn = rcu_dereference_protected(rt->fib6_node,
3775 lockdep_is_held(&table->tb6_lock));
3776 if (rcu_access_pointer(fn->leaf) == rt) {
3777 struct fib6_info *last_sibling, *replace_rt;
3778
3779 last_sibling = list_last_entry(&rt->fib6_siblings,
3780 struct fib6_info,
3781 fib6_siblings);
3782 replace_rt = rcu_dereference_protected(
3783 last_sibling->fib6_next,
3784 lockdep_is_held(&table->tb6_lock));
3785 if (replace_rt)
3786 call_fib6_entry_notifiers_replace(net,
3787 replace_rt);
3788 else
3789 call_fib6_multipath_entry_notifiers(net,
3790 FIB_EVENT_ENTRY_DEL_TMP,
3791 rt, rt->fib6_nsiblings,
3792 NULL);
3793 }
2881fd61
IS		 3794 call_fib6_multipath_entry_notifiers(net,
3795 FIB_EVENT_ENTRY_DEL,
3796 rt,
3797 rt->fib6_nsiblings,
3798 NULL);
0ae81335 3799 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25
DA		 3800 &rt->fib6_siblings,
3801 fib6_siblings) {
0ae81335
DA		 3802 err = fib6_del(sibling, info);
3803 if (err)
e3330039 3804 goto out_unlock;
0ae81335
DA		 3805 }
3806 }
3807
3808 err = fib6_del(rt, info);
e3330039 3809out_unlock:
66f5d6ce 3810 spin_unlock_bh(&table->tb6_lock);
e3330039 3811out_put:
93531c67 3812 fib6_info_release(rt);
16a16cd3
DA		 3813
3814 if (skb) {
e3330039 3815 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
16a16cd3
DA		 3816 info->nlh, gfp_any());
3817 }
0ae81335
DA		 3818 return err;
3819}
3820
0fa6efc5 3821static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
23fb93a4
DA		 3822{
3823 int rc = -ESRCH;
3824
3825 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3826 goto out;
3827
3828 if (cfg->fc_flags & RTF_GATEWAY &&
3829 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3830 goto out;
761f6026
XL		 3831
3832 rc = rt6_remove_exception_rt(rt);
23fb93a4
DA		 3833out:
3834 return rc;
3835}
3836
0fa6efc5
DA		 3837static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3838 struct fib6_nh *nh)
3839{
3840 struct fib6_result res = {
3841 .f6i = rt,
3842 .nh = nh,
3843 };
3844 struct rt6_info *rt_cache;
3845
3846 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3847 if (rt_cache)
3848 return __ip6_del_cached_rt(rt_cache, cfg);
3849
3850 return 0;
3851}
3852
5b98324e
DA		 3853struct fib6_nh_del_cached_rt_arg {
3854 struct fib6_config *cfg;
3855 struct fib6_info *f6i;
3856};
3857
3858static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
3859{
3860 struct fib6_nh_del_cached_rt_arg *arg = _arg;
3861 int rc;
3862
3863 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
3864 return rc != -ESRCH ? rc : 0;
3865}
3866
3867static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
3868{
3869 struct fib6_nh_del_cached_rt_arg arg = {
3870 .cfg = cfg,
3871 .f6i = f6i
3872 };
3873
3874 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
3875}
3876
333c4301
DA		 3877static int ip6_route_del(struct fib6_config *cfg,
3878 struct netlink_ext_ack *extack)
1da177e4 3879{
c71099ac 3880 struct fib6_table *table;
8d1c802b 3881 struct fib6_info *rt;
1da177e4 3882 struct fib6_node *fn;
1da177e4
LT		 3883 int err = -ESRCH;
3884
5578689a 3885 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
d5d531cb
DA		 3886 if (!table) {
3887 NL_SET_ERR_MSG(extack, "FIB table does not exist");
c71099ac 3888 return err;
d5d531cb 3889 }
c71099ac 3890
66f5d6ce 3891 rcu_read_lock();
1da177e4 3892
c71099ac 3893 fn = fib6_locate(&table->tb6_root,
86872cb5 3894 &cfg->fc_dst, cfg->fc_dst_len,
38fbeeee 3895 &cfg->fc_src, cfg->fc_src_len,
2b760fcf 3896 !(cfg->fc_flags & RTF_CACHE));
1ab1457c 3897
1da177e4 3898 if (fn) {
66f5d6ce 3899 for_each_fib6_node_rt_rcu(fn) {
ad1601ae
DA		 3900 struct fib6_nh *nh;
3901
3401bfb1
SB		 3902 if (rt->nh && cfg->fc_nh_id &&
3903 rt->nh->id != cfg->fc_nh_id)
5b98324e 3904 continue;
23fb93a4 3905
5b98324e
DA		 3906 if (cfg->fc_flags & RTF_CACHE) {
3907 int rc = 0;
3908
3909 if (rt->nh) {
3910 rc = ip6_del_cached_rt_nh(cfg, rt);
3911 } else if (cfg->fc_nh_id) {
3912 continue;
3913 } else {
3914 nh = rt->fib6_nh;
3915 rc = ip6_del_cached_rt(cfg, rt, nh);
3916 }
0fa6efc5
DA		 3917 if (rc != -ESRCH) {
3918 rcu_read_unlock();
3919 return rc;
23fb93a4
DA		 3920 }
3921 continue;
2b760fcf 3922 }
ad1601ae 3923
5b98324e
DA		 3924 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
3925 continue;
3926 if (cfg->fc_protocol &&
3927 cfg->fc_protocol != rt->fib6_protocol)
3928 continue;
3929
3930 if (rt->nh) {
3931 if (!fib6_info_hold_safe(rt))
3932 continue;
3933 rcu_read_unlock();
3934
3935 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3936 }
3937 if (cfg->fc_nh_id)
3938 continue;
3939
3940 nh = rt->fib6_nh;
86872cb5 3941 if (cfg->fc_ifindex &&
ad1601ae
DA		 3942 (!nh->fib_nh_dev ||
3943 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
1da177e4 3944 continue;
86872cb5 3945 if (cfg->fc_flags & RTF_GATEWAY &&
ad1601ae 3946 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
1da177e4 3947 continue;
e873e4b9
WW		 3948 if (!fib6_info_hold_safe(rt))
3949 continue;
66f5d6ce 3950 rcu_read_unlock();
1da177e4 3951
0ae81335
DA		 3952 /* if gateway was specified only delete the one hop */
3953 if (cfg->fc_flags & RTF_GATEWAY)
3954 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3955
3956 return __ip6_del_rt_siblings(rt, cfg);
1da177e4
LT		 3957 }
3958 }
66f5d6ce 3959 rcu_read_unlock();
1da177e4
LT		 3960
3961 return err;
3962}
3963
6700c270 3964static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
a6279458 3965{
a6279458 3966 struct netevent_redirect netevent;
e8599ff4 3967 struct rt6_info *rt, *nrt = NULL;
85bd05de 3968 struct fib6_result res = {};
e8599ff4
DM		 3969 struct ndisc_options ndopts;
3970 struct inet6_dev *in6_dev;
3971 struct neighbour *neigh;
71bcdba0 3972 struct rd_msg *msg;
6e157b6a
DM		 3973 int optlen, on_link;
3974 u8 *lladdr;
e8599ff4 3975
29a3cad5 3976 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
71bcdba0 3977 optlen -= sizeof(*msg);
e8599ff4
DM		 3978
3979 if (optlen < 0) {
6e157b6a 3980 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
e8599ff4
DM		 3981 return;
3982 }
3983
71bcdba0 3984 msg = (struct rd_msg *)icmp6_hdr(skb);
e8599ff4 3985
71bcdba0 3986 if (ipv6_addr_is_multicast(&msg->dest)) {
6e157b6a 3987 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
e8599ff4
DM		 3988 return;
3989 }
3990
6e157b6a 3991 on_link = 0;
71bcdba0 3992 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
e8599ff4 3993 on_link = 1;
71bcdba0 3994 } else if (ipv6_addr_type(&msg->target) !=
e8599ff4 3995 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
6e157b6a 3996 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
e8599ff4
DM		 3997 return;
3998 }
3999
4000 in6_dev = __in6_dev_get(skb->dev);
4001 if (!in6_dev)
4002 return;
4003 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
4004 return;
4005
4006 /* RFC2461 8.1:
4007 * The IP source address of the Redirect MUST be the same as the current
4008 * first-hop router for the specified ICMP Destination Address.
4009 */
4010
f997c55c 4011 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
e8599ff4
DM		 4012 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4013 return;
4014 }
6e157b6a
DM		 4015
4016 lladdr = NULL;
e8599ff4
DM		 4017 if (ndopts.nd_opts_tgt_lladdr) {
4018 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4019 skb->dev);
4020 if (!lladdr) {
4021 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4022 return;
4023 }
4024 }
4025
6e157b6a 4026 rt = (struct rt6_info *) dst;
ec13ad1d 4027 if (rt->rt6i_flags & RTF_REJECT) {
6e157b6a 4028 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
e8599ff4 4029 return;
6e157b6a 4030 }
e8599ff4 4031
6e157b6a
DM		 4032 /* Redirect received -> path was valid.
4033 * Look, redirects are sent only in response to data packets,
4034 * so that this nexthop apparently is reachable. --ANK
4035 */
0dec879f 4036 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
a6279458 4037
71bcdba0 4038 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
6e157b6a
DM		 4039 if (!neigh)
4040 return;
a6279458 4041
1da177e4
LT		 4042 /*
4043 * We have finally decided to accept it.
4044 */
4045
f997c55c 4046 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1da177e4
LT		 4047 NEIGH_UPDATE_F_WEAK_OVERRIDE|
4048 NEIGH_UPDATE_F_OVERRIDE|
4049 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
f997c55c
AA		 4050 NEIGH_UPDATE_F_ISROUTER)),
4051 NDISC_REDIRECT, &ndopts);
1da177e4 4052
4d85cd0c 4053 rcu_read_lock();
85bd05de 4054 res.f6i = rcu_dereference(rt->from);
ff24e498 4055 if (!res.f6i)
886b7a50 4056 goto out;
8a14e46f 4057
49d5b8ef
DA		 4058 if (res.f6i->nh) {
4059 struct fib6_nh_match_arg arg = {
4060 .dev = dst->dev,
4061 .gw = &rt->rt6i_gateway,
4062 };
4063
4064 nexthop_for_each_fib6_nh(res.f6i->nh,
4065 fib6_nh_find_match, &arg);
4066
4067 /* fib6_info uses a nexthop that does not have fib6_nh
4068 * using the dst->dev. Should be impossible
4069 */
4070 if (!arg.match)
4071 goto out;
4072 res.nh = arg.match;
4073 } else {
4074 res.nh = res.f6i->fib6_nh;
4075 }
4076
7d21fec9
DA		 4077 res.fib6_flags = res.f6i->fib6_flags;
4078 res.fib6_type = res.f6i->fib6_type;
85bd05de 4079 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
38308473 4080 if (!nrt)
1da177e4
LT		 4081 goto out;
4082
4083 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4084 if (on_link)
4085 nrt->rt6i_flags &= ~RTF_GATEWAY;
4086
4e3fd7a0 4087 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1da177e4 4088
886b7a50 4089 /* rt6_insert_exception() will take care of duplicated exceptions */
5012f0a5 4090 if (rt6_insert_exception(nrt, &res)) {
2b760fcf
WW		 4091 dst_release_immediate(&nrt->dst);
4092 goto out;
4093 }
1da177e4 4094
d8d1f30b
CG		 4095 netevent.old = &rt->dst;
4096 netevent.new = &nrt->dst;
71bcdba0 4097 netevent.daddr = &msg->dest;
60592833 4098 netevent.neigh = neigh;
8d71740c
TT		 4099 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4100
1da177e4 4101out:
886b7a50 4102 rcu_read_unlock();
e8599ff4 4103 neigh_release(neigh);
6e157b6a
DM		 4104}
4105
70ceb4f5 4106#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 4107static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 4108 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 4109 const struct in6_addr *gwaddr,
4110 struct net_device *dev)
70ceb4f5 4111{
830218c1
DA		 4112 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4113 int ifindex = dev->ifindex;
70ceb4f5 4114 struct fib6_node *fn;
8d1c802b 4115 struct fib6_info *rt = NULL;
c71099ac
TG		 4116 struct fib6_table *table;
4117
830218c1 4118 table = fib6_get_table(net, tb_id);
38308473 4119 if (!table)
c71099ac 4120 return NULL;
70ceb4f5 4121
66f5d6ce 4122 rcu_read_lock();
38fbeeee 4123 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
70ceb4f5
YH		 4124 if (!fn)
4125 goto out;
4126
66f5d6ce 4127 for_each_fib6_node_rt_rcu(fn) {
f88d8ea6
DA		 4128 /* these routes do not use nexthops */
4129 if (rt->nh)
4130 continue;
1cf844c7 4131 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
70ceb4f5 4132 continue;
2b2450ca 4133 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
1cf844c7 4134 !rt->fib6_nh->fib_nh_gw_family)
70ceb4f5 4135 continue;
1cf844c7 4136 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
70ceb4f5 4137 continue;
e873e4b9
WW		 4138 if (!fib6_info_hold_safe(rt))
4139 continue;
70ceb4f5
YH		 4140 break;
4141 }
4142out:
66f5d6ce 4143 rcu_read_unlock();
70ceb4f5
YH		 4144 return rt;
4145}
4146
8d1c802b 4147static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 4148 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 4149 const struct in6_addr *gwaddr,
4150 struct net_device *dev,
95c96174 4151 unsigned int pref)
70ceb4f5 4152{
86872cb5 4153 struct fib6_config cfg = {
238fc7ea 4154 .fc_metric = IP6_RT_PRIO_USER,
830218c1 4155 .fc_ifindex = dev->ifindex,
86872cb5
TG		 4156 .fc_dst_len = prefixlen,
4157 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4158 RTF_UP | RTF_PREF(pref),
b91d5329 4159 .fc_protocol = RTPROT_RA,
e8478e80 4160 .fc_type = RTN_UNICAST,
15e47304 4161 .fc_nlinfo.portid = 0,
efa2cea0
DL		 4162 .fc_nlinfo.nlh = NULL,
4163 .fc_nlinfo.nl_net = net,
86872cb5
TG		 4164 };
4165
830218c1 4166 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
4e3fd7a0
AD		 4167 cfg.fc_dst = *prefix;
4168 cfg.fc_gateway = *gwaddr;
70ceb4f5 4169
e317da96
YH		 4170 /* We should treat it as a default route if prefix length is 0. */
4171 if (!prefixlen)
86872cb5 4172 cfg.fc_flags |= RTF_DEFAULT;
70ceb4f5 4173
acb54e3c 4174 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
70ceb4f5 4175
830218c1 4176 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
70ceb4f5
YH		 4177}
4178#endif
4179
8d1c802b 4180struct fib6_info *rt6_get_dflt_router(struct net *net,
afb1d4b5
DA		 4181 const struct in6_addr *addr,
4182 struct net_device *dev)
1ab1457c 4183{
830218c1 4184 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
8d1c802b 4185 struct fib6_info *rt;
c71099ac 4186 struct fib6_table *table;
1da177e4 4187
afb1d4b5 4188 table = fib6_get_table(net, tb_id);
38308473 4189 if (!table)
c71099ac 4190 return NULL;
1da177e4 4191
66f5d6ce
WW		 4192 rcu_read_lock();
4193 for_each_fib6_node_rt_rcu(&table->tb6_root) {
f88d8ea6 4194 struct fib6_nh *nh;
ad1601ae 4195
f88d8ea6
DA		 4196 /* RA routes do not use nexthops */
4197 if (rt->nh)
4198 continue;
4199
4200 nh = rt->fib6_nh;
ad1601ae 4201 if (dev == nh->fib_nh_dev &&
93c2fb25 4202 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
ad1601ae 4203 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
1da177e4
LT		 4204 break;
4205 }
e873e4b9
WW		 4206 if (rt && !fib6_info_hold_safe(rt))
4207 rt = NULL;
66f5d6ce 4208 rcu_read_unlock();
1da177e4
LT		 4209 return rt;
4210}
4211
8d1c802b 4212struct fib6_info *rt6_add_dflt_router(struct net *net,
afb1d4b5 4213 const struct in6_addr *gwaddr,
ebacaaa0
YH		 4214 struct net_device *dev,
4215 unsigned int pref)
1da177e4 4216{
86872cb5 4217 struct fib6_config cfg = {
ca254490 4218 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
238fc7ea 4219 .fc_metric = IP6_RT_PRIO_USER,
86872cb5
TG		 4220 .fc_ifindex = dev->ifindex,
4221 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4222 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
b91d5329 4223 .fc_protocol = RTPROT_RA,
e8478e80 4224 .fc_type = RTN_UNICAST,
15e47304 4225 .fc_nlinfo.portid = 0,
5578689a 4226 .fc_nlinfo.nlh = NULL,
afb1d4b5 4227 .fc_nlinfo.nl_net = net,
86872cb5 4228 };
1da177e4 4229
4e3fd7a0 4230 cfg.fc_gateway = *gwaddr;
1da177e4 4231
acb54e3c 4232 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
830218c1
DA		 4233 struct fib6_table *table;
4234
4235 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4236 if (table)
4237 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4238 }
1da177e4 4239
afb1d4b5 4240 return rt6_get_dflt_router(net, gwaddr, dev);
1da177e4
LT		 4241}
4242
afb1d4b5
DA		 4243static void __rt6_purge_dflt_routers(struct net *net,
4244 struct fib6_table *table)
1da177e4 4245{
8d1c802b 4246 struct fib6_info *rt;
1da177e4
LT		 4247
4248restart:
66f5d6ce
WW		 4249 rcu_read_lock();
4250 for_each_fib6_node_rt_rcu(&table->tb6_root) {
dcd1f572
DA		 4251 struct net_device *dev = fib6_info_nh_dev(rt);
4252 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4253
93c2fb25 4254 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
e873e4b9
WW		 4255 (!idev || idev->cnf.accept_ra != 2) &&
4256 fib6_info_hold_safe(rt)) {
93531c67
DA		 4257 rcu_read_unlock();
4258 ip6_del_rt(net, rt);
1da177e4
LT		 4259 goto restart;
4260 }
4261 }
66f5d6ce 4262 rcu_read_unlock();
830218c1
DA		 4263
4264 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4265}
4266
4267void rt6_purge_dflt_routers(struct net *net)
4268{
4269 struct fib6_table *table;
4270 struct hlist_head *head;
4271 unsigned int h;
4272
4273 rcu_read_lock();
4274
4275 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4276 head = &net->ipv6.fib_table_hash[h];
4277 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4278 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
afb1d4b5 4279 __rt6_purge_dflt_routers(net, table);
830218c1
DA		 4280 }
4281 }
4282
4283 rcu_read_unlock();
1da177e4
LT		 4284}
4285
5578689a
DL		 4286static void rtmsg_to_fib6_config(struct net *net,
4287 struct in6_rtmsg *rtmsg,
86872cb5
TG		 4288 struct fib6_config *cfg)
4289{
8823a3ac
MŻ		 4290 *cfg = (struct fib6_config){
4291 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4292 : RT6_TABLE_MAIN,
4293 .fc_ifindex = rtmsg->rtmsg_ifindex,
67f69513 4294 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
8823a3ac
MŻ		 4295 .fc_expires = rtmsg->rtmsg_info,
4296 .fc_dst_len = rtmsg->rtmsg_dst_len,
4297 .fc_src_len = rtmsg->rtmsg_src_len,
4298 .fc_flags = rtmsg->rtmsg_flags,
4299 .fc_type = rtmsg->rtmsg_type,
4300
4301 .fc_nlinfo.nl_net = net,
4302
4303 .fc_dst = rtmsg->rtmsg_dst,
4304 .fc_src = rtmsg->rtmsg_src,
4305 .fc_gateway = rtmsg->rtmsg_gateway,
4306 };
86872cb5
TG		 4307}
4308
5578689a 4309int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1da177e4 4310{
86872cb5 4311 struct fib6_config cfg;
1da177e4
LT		 4312 struct in6_rtmsg rtmsg;
4313 int err;
4314
67ba4152 4315 switch (cmd) {
1da177e4
LT		 4316 case SIOCADDRT: /* Add a route */
4317 case SIOCDELRT: /* Delete a route */
af31f412 4318 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4
LT		 4319 return -EPERM;
4320 err = copy_from_user(&rtmsg, arg,
4321 sizeof(struct in6_rtmsg));
4322 if (err)
4323 return -EFAULT;
86872cb5 4324
5578689a 4325 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
86872cb5 4326
1da177e4
LT		 4327 rtnl_lock();
4328 switch (cmd) {
4329 case SIOCADDRT:
acb54e3c 4330 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
1da177e4
LT		 4331 break;
4332 case SIOCDELRT:
333c4301 4333 err = ip6_route_del(&cfg, NULL);
1da177e4
LT		 4334 break;
4335 default:
4336 err = -EINVAL;
4337 }
4338 rtnl_unlock();
4339
4340 return err;
3ff50b79 4341 }
1da177e4
LT		 4342
4343 return -EINVAL;
4344}
4345
4346/*
4347 * Drop the packet on the floor
4348 */
4349
d5fdd6ba 4350static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1da177e4 4351{
adf30907 4352 struct dst_entry *dst = skb_dst(skb);
1d3fd8a1
SS		 4353 struct net *net = dev_net(dst->dev);
4354 struct inet6_dev *idev;
4355 int type;
4356
4357 if (netif_is_l3_master(skb->dev) &&
4358 dst->dev == net->loopback_dev)
4359 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4360 else
4361 idev = ip6_dst_idev(dst);
4362
612f09e8
YH		 4363 switch (ipstats_mib_noroutes) {
4364 case IPSTATS_MIB_INNOROUTES:
0660e03f 4365 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
45bb0060 4366 if (type == IPV6_ADDR_ANY) {
1d3fd8a1 4367 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
612f09e8
YH		 4368 break;
4369 }
4370 /* FALLTHROUGH */
4371 case IPSTATS_MIB_OUTNOROUTES:
1d3fd8a1 4372 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
612f09e8
YH		 4373 break;
4374 }
1d3fd8a1
SS		 4375
4376 /* Start over by dropping the dst for l3mdev case */
4377 if (netif_is_l3_master(skb->dev))
4378 skb_dst_drop(skb);
4379
3ffe533c 4380 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1da177e4
LT		 4381 kfree_skb(skb);
4382 return 0;
4383}
4384
9ce8ade0
TG		 4385static int ip6_pkt_discard(struct sk_buff *skb)
4386{
612f09e8 4387 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4388}
4389
ede2059d 4390static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
1da177e4 4391{
adf30907 4392 skb->dev = skb_dst(skb)->dev;
612f09e8 4393 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1da177e4
LT		 4394}
4395
9ce8ade0
TG		 4396static int ip6_pkt_prohibit(struct sk_buff *skb)
4397{
612f09e8 4398 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4399}
4400
ede2059d 4401static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
9ce8ade0 4402{
adf30907 4403 skb->dev = skb_dst(skb)->dev;
612f09e8 4404 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
9ce8ade0
TG		 4405}
4406
1da177e4
LT		 4407/*
4408 * Allocate a dst for local (unicast / anycast) address.
4409 */
4410
360a9887
DA		 4411struct fib6_info *addrconf_f6i_alloc(struct net *net,
4412 struct inet6_dev *idev,
4413 const struct in6_addr *addr,
4414 bool anycast, gfp_t gfp_flags)
1da177e4 4415{
c7a1ce39
DA		 4416 struct fib6_config cfg = {
4417 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4418 .fc_ifindex = idev->dev->ifindex,
d55a2e37 4419 .fc_flags = RTF_UP | RTF_NONEXTHOP,
c7a1ce39
DA		 4420 .fc_dst = *addr,
4421 .fc_dst_len = 128,
4422 .fc_protocol = RTPROT_KERNEL,
4423 .fc_nlinfo.nl_net = net,
4424 .fc_ignore_dev_down = true,
4425 };
d55a2e37 4426 struct fib6_info *f6i;
1da177e4 4427
e8478e80 4428 if (anycast) {
c7a1ce39
DA		 4429 cfg.fc_type = RTN_ANYCAST;
4430 cfg.fc_flags |= RTF_ANYCAST;
e8478e80 4431 } else {
c7a1ce39
DA		 4432 cfg.fc_type = RTN_LOCAL;
4433 cfg.fc_flags |= RTF_LOCAL;
e8478e80 4434 }
1da177e4 4435
d55a2e37 4436 f6i = ip6_route_info_create(&cfg, gfp_flags, NULL);
8652f17c 4437 if (!IS_ERR(f6i))
d55a2e37
MŻ		 4438 f6i->dst_nocount = true;
4439 return f6i;
1da177e4
LT		 4440}
4441
c3968a85
DW		 4442/* remove deleted ip from prefsrc entries */
4443struct arg_dev_net_ip {
4444 struct net_device *dev;
4445 struct net *net;
4446 struct in6_addr *addr;
4447};
4448
8d1c802b 4449static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
c3968a85
DW		 4450{
4451 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4452 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4453 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4454
f88d8ea6
DA		 4455 if (!rt->nh &&
4456 ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
421842ed 4457 rt != net->ipv6.fib6_null_entry &&
93c2fb25 4458 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
60006a48 4459 spin_lock_bh(&rt6_exception_lock);
c3968a85 4460 /* remove prefsrc entry */
93c2fb25 4461 rt->fib6_prefsrc.plen = 0;
60006a48 4462 spin_unlock_bh(&rt6_exception_lock);
c3968a85
DW		 4463 }
4464 return 0;
4465}
4466
4467void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4468{
4469 struct net *net = dev_net(ifp->idev->dev);
4470 struct arg_dev_net_ip adni = {
4471 .dev = ifp->idev->dev,
4472 .net = net,
4473 .addr = &ifp->addr,
4474 };
0c3584d5 4475 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
c3968a85
DW		 4476}
4477
2b2450ca 4478#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
be7a010d
DJ		 4479
4480/* Remove routers and update dst entries when gateway turn into host. */
8d1c802b 4481static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
be7a010d
DJ		 4482{
4483 struct in6_addr *gateway = (struct in6_addr *)arg;
f88d8ea6
DA		 4484 struct fib6_nh *nh;
4485
4486 /* RA routes do not use nexthops */
4487 if (rt->nh)
4488 return 0;
be7a010d 4489
f88d8ea6 4490 nh = rt->fib6_nh;
93c2fb25 4491 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
cc5c073a 4492 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
be7a010d 4493 return -1;
b16cb459
WW		 4494
4495 /* Further clean up cached routes in exception table.
4496 * This is needed because cached route may have a different
4497 * gateway than its 'parent' in the case of an ip redirect.
4498 */
cc5c073a 4499 fib6_nh_exceptions_clean_tohost(nh, gateway);
b16cb459 4500
be7a010d
DJ		 4501 return 0;
4502}
4503
4504void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4505{
4506 fib6_clean_all(net, fib6_clean_tohost, gateway);
4507}
4508
2127d95a
IS		 4509struct arg_netdev_event {
4510 const struct net_device *dev;
4c981e28 4511 union {
ecc5663c 4512 unsigned char nh_flags;
4c981e28
IS		 4513 unsigned long event;
4514 };
2127d95a
IS		 4515};
4516
8d1c802b 4517static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
d7dedee1 4518{
8d1c802b 4519 struct fib6_info *iter;
d7dedee1
IS		 4520 struct fib6_node *fn;
4521
93c2fb25
DA		 4522 fn = rcu_dereference_protected(rt->fib6_node,
4523 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4524 iter = rcu_dereference_protected(fn->leaf,
93c2fb25 4525 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4526 while (iter) {
93c2fb25 4527 if (iter->fib6_metric == rt->fib6_metric &&
33bd5ac5 4528 rt6_qualify_for_ecmp(iter))
d7dedee1 4529 return iter;
8fb11a9a 4530 iter = rcu_dereference_protected(iter->fib6_next,
93c2fb25 4531 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1
IS		 4532 }
4533
4534 return NULL;
4535}
4536
f88d8ea6 4537/* only called for fib entries with builtin fib6_nh */
8d1c802b 4538static bool rt6_is_dead(const struct fib6_info *rt)
d7dedee1 4539{
1cf844c7
DA		 4540 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4541 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4542 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
d7dedee1
IS		 4543 return true;
4544
4545 return false;
4546}
4547
8d1c802b 4548static int rt6_multipath_total_weight(const struct fib6_info *rt)
d7dedee1 4549{
8d1c802b 4550 struct fib6_info *iter;
d7dedee1
IS		 4551 int total = 0;
4552
4553 if (!rt6_is_dead(rt))
1cf844c7 4554 total += rt->fib6_nh->fib_nh_weight;
d7dedee1 4555
93c2fb25 4556 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
d7dedee1 4557 if (!rt6_is_dead(iter))
1cf844c7 4558 total += iter->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4559 }
4560
4561 return total;
4562}
4563
8d1c802b 4564static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
d7dedee1
IS		 4565{
4566 int upper_bound = -1;
4567
4568 if (!rt6_is_dead(rt)) {
1cf844c7 4569 *weight += rt->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4570 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4571 total) - 1;
4572 }
1cf844c7 4573 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
d7dedee1
IS		 4574}
4575
8d1c802b 4576static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
d7dedee1 4577{
8d1c802b 4578 struct fib6_info *iter;
d7dedee1
IS		 4579 int weight = 0;
4580
4581 rt6_upper_bound_set(rt, &weight, total);
4582
93c2fb25 4583 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
d7dedee1
IS		 4584 rt6_upper_bound_set(iter, &weight, total);
4585}
4586
8d1c802b 4587void rt6_multipath_rebalance(struct fib6_info *rt)
d7dedee1 4588{
8d1c802b 4589 struct fib6_info *first;
d7dedee1
IS		 4590 int total;
4591
4592 /* In case the entire multipath route was marked for flushing,
4593 * then there is no need to rebalance upon the removal of every
4594 * sibling route.
4595 */
93c2fb25 4596 if (!rt->fib6_nsiblings || rt->should_flush)
d7dedee1
IS		 4597 return;
4598
4599 /* During lookup routes are evaluated in order, so we need to
4600 * make sure upper bounds are assigned from the first sibling
4601 * onwards.
4602 */
4603 first = rt6_multipath_first_sibling(rt);
4604 if (WARN_ON_ONCE(!first))
4605 return;
4606
4607 total = rt6_multipath_total_weight(first);
4608 rt6_multipath_upper_bound_set(first, total);
4609}
4610
8d1c802b 4611static int fib6_ifup(struct fib6_info *rt, void *p_arg)
2127d95a
IS		 4612{
4613 const struct arg_netdev_event *arg = p_arg;
7aef6859 4614 struct net *net = dev_net(arg->dev);
2127d95a 4615
f88d8ea6 4616 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
1cf844c7
DA		 4617 rt->fib6_nh->fib_nh_dev == arg->dev) {
4618 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
7aef6859 4619 fib6_update_sernum_upto_root(net, rt);
d7dedee1 4620 rt6_multipath_rebalance(rt);
1de178ed 4621 }
2127d95a
IS		 4622
4623 return 0;
4624}
4625
ecc5663c 4626void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
2127d95a
IS		 4627{
4628 struct arg_netdev_event arg = {
4629 .dev = dev,
6802f3ad
IS		 4630 {
4631 .nh_flags = nh_flags,
4632 },
2127d95a
IS		 4633 };
4634
4635 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4636 arg.nh_flags |= RTNH_F_LINKDOWN;
4637
4638 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4639}
4640
f88d8ea6 4641/* only called for fib entries with inline fib6_nh */
8d1c802b 4642static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
1de178ed
IS		 4643 const struct net_device *dev)
4644{
8d1c802b 4645 struct fib6_info *iter;
1de178ed 4646
1cf844c7 4647 if (rt->fib6_nh->fib_nh_dev == dev)
1de178ed 4648 return true;
93c2fb25 4649 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7 4650 if (iter->fib6_nh->fib_nh_dev == dev)
1de178ed
IS		 4651 return true;
4652
4653 return false;
4654}
4655
8d1c802b 4656static void rt6_multipath_flush(struct fib6_info *rt)
1de178ed 4657{
8d1c802b 4658 struct fib6_info *iter;
1de178ed
IS		 4659
4660 rt->should_flush = 1;
93c2fb25 4661 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1de178ed
IS		 4662 iter->should_flush = 1;
4663}
4664
8d1c802b 4665static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
1de178ed
IS		 4666 const struct net_device *down_dev)
4667{
8d1c802b 4668 struct fib6_info *iter;
1de178ed
IS		 4669 unsigned int dead = 0;
4670
1cf844c7
DA		 4671 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4672 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed 4673 dead++;
93c2fb25 4674 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4675 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4676 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed
IS		 4677 dead++;
4678
4679 return dead;
4680}
4681
8d1c802b 4682static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
1de178ed 4683 const struct net_device *dev,
ecc5663c 4684 unsigned char nh_flags)
1de178ed 4685{
8d1c802b 4686 struct fib6_info *iter;
1de178ed 4687
1cf844c7
DA		 4688 if (rt->fib6_nh->fib_nh_dev == dev)
4689 rt->fib6_nh->fib_nh_flags |= nh_flags;
93c2fb25 4690 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4691 if (iter->fib6_nh->fib_nh_dev == dev)
4692 iter->fib6_nh->fib_nh_flags |= nh_flags;
1de178ed
IS		 4693}
4694
a1a22c12 4695/* called with write lock held for table with rt */
8d1c802b 4696static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
1da177e4 4697{
4c981e28
IS		 4698 const struct arg_netdev_event *arg = p_arg;
4699 const struct net_device *dev = arg->dev;
7aef6859 4700 struct net *net = dev_net(dev);
8ed67789 4701
f88d8ea6 4702 if (rt == net->ipv6.fib6_null_entry || rt->nh)
27c6fa73
IS		 4703 return 0;
4704
4705 switch (arg->event) {
4706 case NETDEV_UNREGISTER:
1cf844c7 4707 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
27c6fa73 4708 case NETDEV_DOWN:
1de178ed 4709 if (rt->should_flush)
27c6fa73 4710 return -1;
93c2fb25 4711 if (!rt->fib6_nsiblings)
1cf844c7 4712 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
1de178ed
IS		 4713 if (rt6_multipath_uses_dev(rt, dev)) {
4714 unsigned int count;
4715
4716 count = rt6_multipath_dead_count(rt, dev);
93c2fb25 4717 if (rt->fib6_nsiblings + 1 == count) {
1de178ed
IS		 4718 rt6_multipath_flush(rt);
4719 return -1;
4720 }
4721 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4722 RTNH_F_LINKDOWN);
7aef6859 4723 fib6_update_sernum(net, rt);
d7dedee1 4724 rt6_multipath_rebalance(rt);
1de178ed
IS		 4725 }
4726 return -2;
27c6fa73 4727 case NETDEV_CHANGE:
1cf844c7 4728 if (rt->fib6_nh->fib_nh_dev != dev ||
93c2fb25 4729 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
27c6fa73 4730 break;
1cf844c7 4731 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
d7dedee1 4732 rt6_multipath_rebalance(rt);
27c6fa73 4733 break;
2b241361 4734 }
c159d30c 4735
1da177e4
LT		 4736 return 0;
4737}
4738
27c6fa73 4739void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
1da177e4 4740{
4c981e28 4741 struct arg_netdev_event arg = {
8ed67789 4742 .dev = dev,
6802f3ad
IS		 4743 {
4744 .event = event,
4745 },
8ed67789 4746 };
7c6bb7d2 4747 struct net *net = dev_net(dev);
8ed67789 4748
7c6bb7d2
DA		 4749 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4750 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4751 else
4752 fib6_clean_all(net, fib6_ifdown, &arg);
4c981e28
IS		 4753}
4754
4755void rt6_disable_ip(struct net_device *dev, unsigned long event)
4756{
4757 rt6_sync_down_dev(dev, event);
4758 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4759 neigh_ifdown(&nd_tbl, dev);
1da177e4
LT		 4760}
4761
95c96174 4762struct rt6_mtu_change_arg {
1da177e4 4763 struct net_device *dev;
95c96174 4764 unsigned int mtu;
c0b220cf 4765 struct fib6_info *f6i;
1da177e4
LT		 4766};
4767
cc5c073a 4768static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
c0b220cf
DA		 4769{
4770 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
cc5c073a 4771 struct fib6_info *f6i = arg->f6i;
c0b220cf
DA		 4772
4773 /* For administrative MTU increase, there is no way to discover
4774 * IPv6 PMTU increase, so PMTU increase should be updated here.
4775 * Since RFC 1981 doesn't include administrative MTU increase
4776 * update PMTU increase is a MUST. (i.e. jumbo frame)
4777 */
4778 if (nh->fib_nh_dev == arg->dev) {
4779 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4780 u32 mtu = f6i->fib6_pmtu;
4781
4782 if (mtu >= arg->mtu ||
4783 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4784 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4785
4786 spin_lock_bh(&rt6_exception_lock);
cc5c073a 4787 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
c0b220cf
DA		 4788 spin_unlock_bh(&rt6_exception_lock);
4789 }
4790
4791 return 0;
4792}
4793
4794static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
1da177e4
LT		 4795{
4796 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4797 struct inet6_dev *idev;
4798
4799 /* In IPv6 pmtu discovery is not optional,
4800 so that RTAX_MTU lock cannot disable it.
4801 We still use this lock to block changes
4802 caused by addrconf/ndisc.
4803 */
4804
4805 idev = __in6_dev_get(arg->dev);
38308473 4806 if (!idev)
1da177e4
LT		 4807 return 0;
4808
c0b220cf
DA		 4809 if (fib6_metric_locked(f6i, RTAX_MTU))
4810 return 0;
d4ead6b3 4811
c0b220cf 4812 arg->f6i = f6i;
2d44234b
DA		 4813 if (f6i->nh) {
4814 /* fib6_nh_mtu_change only returns 0, so this is safe */
4815 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4816 arg);
4817 }
4818
1cf844c7 4819 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
1da177e4
LT		 4820}
4821
95c96174 4822void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
1da177e4 4823{
c71099ac
TG		 4824 struct rt6_mtu_change_arg arg = {
4825 .dev = dev,
4826 .mtu = mtu,
4827 };
1da177e4 4828
0c3584d5 4829 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
1da177e4
LT		 4830}
4831
ef7c79ed 4832static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
75425657 4833 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
5176f91e 4834 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
aa8f8778 4835 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
86872cb5 4836 [RTA_OIF] = { .type = NLA_U32 },
ab364a6f 4837 [RTA_IIF] = { .type = NLA_U32 },
86872cb5
TG		 4838 [RTA_PRIORITY] = { .type = NLA_U32 },
4839 [RTA_METRICS] = { .type = NLA_NESTED },
51ebd318 4840 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
c78ba6d6 4841 [RTA_PREF] = { .type = NLA_U8 },
19e42e45
RP		 4842 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4843 [RTA_ENCAP] = { .type = NLA_NESTED },
32bc201e 4844 [RTA_EXPIRES] = { .type = NLA_U32 },
622ec2c9 4845 [RTA_UID] = { .type = NLA_U32 },
3b45a410 4846 [RTA_MARK] = { .type = NLA_U32 },
aa8f8778 4847 [RTA_TABLE] = { .type = NLA_U32 },
eacb9384
RP		 4848 [RTA_IP_PROTO] = { .type = NLA_U8 },
4849 [RTA_SPORT] = { .type = NLA_U16 },
4850 [RTA_DPORT] = { .type = NLA_U16 },
5b98324e 4851 [RTA_NH_ID] = { .type = NLA_U32 },
86872cb5
TG		 4852};
4853
4854static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
333c4301
DA		 4855 struct fib6_config *cfg,
4856 struct netlink_ext_ack *extack)
1da177e4 4857{
86872cb5
TG		 4858 struct rtmsg *rtm;
4859 struct nlattr *tb[RTA_MAX+1];
c78ba6d6 4860 unsigned int pref;
86872cb5 4861 int err;
1da177e4 4862
8cb08174
JB		 4863 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
4864 rtm_ipv6_policy, extack);
86872cb5
TG		 4865 if (err < 0)
4866 goto errout;
1da177e4 4867
86872cb5
TG		 4868 err = -EINVAL;
4869 rtm = nlmsg_data(nlh);
86872cb5 4870
84db8407
MŻ		 4871 *cfg = (struct fib6_config){
4872 .fc_table = rtm->rtm_table,
4873 .fc_dst_len = rtm->rtm_dst_len,
4874 .fc_src_len = rtm->rtm_src_len,
4875 .fc_flags = RTF_UP,
4876 .fc_protocol = rtm->rtm_protocol,
4877 .fc_type = rtm->rtm_type,
4878
4879 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
4880 .fc_nlinfo.nlh = nlh,
4881 .fc_nlinfo.nl_net = sock_net(skb->sk),
4882 };
86872cb5 4883
ef2c7d7b
ND		 4884 if (rtm->rtm_type == RTN_UNREACHABLE ||
4885 rtm->rtm_type == RTN_BLACKHOLE ||
b4949ab2
ND		 4886 rtm->rtm_type == RTN_PROHIBIT ||
4887 rtm->rtm_type == RTN_THROW)
86872cb5
TG		 4888 cfg->fc_flags |= RTF_REJECT;
4889
ab79ad14
MŻ		 4890 if (rtm->rtm_type == RTN_LOCAL)
4891 cfg->fc_flags |= RTF_LOCAL;
4892
1f56a01f
MKL		 4893 if (rtm->rtm_flags & RTM_F_CLONED)
4894 cfg->fc_flags |= RTF_CACHE;
4895
fc1e64e1
DA		 4896 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4897
5b98324e
DA		 4898 if (tb[RTA_NH_ID]) {
4899 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
4900 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
4901 NL_SET_ERR_MSG(extack,
4902 "Nexthop specification and nexthop id are mutually exclusive");
4903 goto errout;
4904 }
4905 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
4906 }
4907
86872cb5 4908 if (tb[RTA_GATEWAY]) {
67b61f6c 4909 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
86872cb5 4910 cfg->fc_flags |= RTF_GATEWAY;
1da177e4 4911 }
e3818541
DA		 4912 if (tb[RTA_VIA]) {
4913 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
4914 goto errout;
4915 }
86872cb5
TG		 4916
4917 if (tb[RTA_DST]) {
4918 int plen = (rtm->rtm_dst_len + 7) >> 3;
4919
4920 if (nla_len(tb[RTA_DST]) < plen)
4921 goto errout;
4922
4923 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1da177e4 4924 }
86872cb5
TG		 4925
4926 if (tb[RTA_SRC]) {
4927 int plen = (rtm->rtm_src_len + 7) >> 3;
4928
4929 if (nla_len(tb[RTA_SRC]) < plen)
4930 goto errout;
4931
4932 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1da177e4 4933 }
86872cb5 4934
c3968a85 4935 if (tb[RTA_PREFSRC])
67b61f6c 4936 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
c3968a85 4937
86872cb5
TG		 4938 if (tb[RTA_OIF])
4939 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4940
4941 if (tb[RTA_PRIORITY])
4942 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4943
4944 if (tb[RTA_METRICS]) {
4945 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4946 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
1da177e4 4947 }
86872cb5
TG		 4948
4949 if (tb[RTA_TABLE])
4950 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4951
51ebd318
ND		 4952 if (tb[RTA_MULTIPATH]) {
4953 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4954 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
9ed59592
DA		 4955
4956 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
c255bd68 4957 cfg->fc_mp_len, extack);
9ed59592
DA		 4958 if (err < 0)
4959 goto errout;
51ebd318
ND		 4960 }
4961
c78ba6d6
LR		 4962 if (tb[RTA_PREF]) {
4963 pref = nla_get_u8(tb[RTA_PREF]);
4964 if (pref != ICMPV6_ROUTER_PREF_LOW &&
4965 pref != ICMPV6_ROUTER_PREF_HIGH)
4966 pref = ICMPV6_ROUTER_PREF_MEDIUM;
4967 cfg->fc_flags |= RTF_PREF(pref);
4968 }
4969
19e42e45
RP		 4970 if (tb[RTA_ENCAP])
4971 cfg->fc_encap = tb[RTA_ENCAP];
4972
9ed59592 4973 if (tb[RTA_ENCAP_TYPE]) {
19e42e45
RP		 4974 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
4975
c255bd68 4976 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
9ed59592
DA		 4977 if (err < 0)
4978 goto errout;
4979 }
4980
32bc201e
XL		 4981 if (tb[RTA_EXPIRES]) {
4982 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
4983
4984 if (addrconf_finite_timeout(timeout)) {
4985 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
4986 cfg->fc_flags |= RTF_EXPIRES;
4987 }
4988 }
4989
86872cb5
TG		 4990 err = 0;
4991errout:
4992 return err;
1da177e4
LT		 4993}
4994
6b9ea5a6 4995struct rt6_nh {
8d1c802b 4996 struct fib6_info *fib6_info;
6b9ea5a6 4997 struct fib6_config r_cfg;
6b9ea5a6
RP		 4998 struct list_head next;
4999};
5000
d4ead6b3
DA		 5001static int ip6_route_info_append(struct net *net,
5002 struct list_head *rt6_nh_list,
8d1c802b
DA		 5003 struct fib6_info *rt,
5004 struct fib6_config *r_cfg)
6b9ea5a6
RP		 5005{
5006 struct rt6_nh *nh;
6b9ea5a6
RP		 5007 int err = -EEXIST;
5008
5009 list_for_each_entry(nh, rt6_nh_list, next) {
8d1c802b
DA		 5010 /* check if fib6_info already exists */
5011 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
6b9ea5a6
RP		 5012 return err;
5013 }
5014
5015 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5016 if (!nh)
5017 return -ENOMEM;
8d1c802b 5018 nh->fib6_info = rt;
6b9ea5a6
RP		 5019 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5020 list_add_tail(&nh->next, rt6_nh_list);
5021
5022 return 0;
5023}
5024
8d1c802b
DA		 5025static void ip6_route_mpath_notify(struct fib6_info *rt,
5026 struct fib6_info *rt_last,
3b1137fe
DA		 5027 struct nl_info *info,
5028 __u16 nlflags)
5029{
5030 /* if this is an APPEND route, then rt points to the first route
5031 * inserted and rt_last points to last route inserted. Userspace
5032 * wants a consistent dump of the route which starts at the first
5033 * nexthop. Since sibling routes are always added at the end of
5034 * the list, find the first sibling of the last route appended
5035 */
93c2fb25
DA		 5036 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5037 rt = list_first_entry(&rt_last->fib6_siblings,
8d1c802b 5038 struct fib6_info,
93c2fb25 5039 fib6_siblings);
3b1137fe
DA		 5040 }
5041
5042 if (rt)
5043 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5044}
5045
0ee0f47c
IS		 5046static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5047{
5048 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5049 bool should_notify = false;
5050 struct fib6_info *leaf;
5051 struct fib6_node *fn;
5052
5053 rcu_read_lock();
5054 fn = rcu_dereference(rt->fib6_node);
5055 if (!fn)
5056 goto out;
5057
5058 leaf = rcu_dereference(fn->leaf);
5059 if (!leaf)
5060 goto out;
5061
5062 if (rt == leaf ||
5063 (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5064 rt6_qualify_for_ecmp(leaf)))
5065 should_notify = true;
5066out:
5067 rcu_read_unlock();
5068
5069 return should_notify;
5070}
5071
333c4301
DA		 5072static int ip6_route_multipath_add(struct fib6_config *cfg,
5073 struct netlink_ext_ack *extack)
51ebd318 5074{
8d1c802b 5075 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
3b1137fe 5076 struct nl_info *info = &cfg->fc_nlinfo;
ebee3cad 5077 enum fib_event_type event_type;
51ebd318
ND		 5078 struct fib6_config r_cfg;
5079 struct rtnexthop *rtnh;
8d1c802b 5080 struct fib6_info *rt;
6b9ea5a6
RP		 5081 struct rt6_nh *err_nh;
5082 struct rt6_nh *nh, *nh_safe;
3b1137fe 5083 __u16 nlflags;
51ebd318
ND		 5084 int remaining;
5085 int attrlen;
6b9ea5a6
RP		 5086 int err = 1;
5087 int nhn = 0;
5088 int replace = (cfg->fc_nlinfo.nlh &&
5089 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5090 LIST_HEAD(rt6_nh_list);
51ebd318 5091
3b1137fe
DA		 5092 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5093 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5094 nlflags |= NLM_F_APPEND;
5095
35f1b4e9 5096 remaining = cfg->fc_mp_len;
51ebd318 5097 rtnh = (struct rtnexthop *)cfg->fc_mp;
51ebd318 5098
6b9ea5a6 5099 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
8d1c802b 5100 * fib6_info structs per nexthop
6b9ea5a6 5101 */
51ebd318
ND		 5102 while (rtnh_ok(rtnh, remaining)) {
5103 memcpy(&r_cfg, cfg, sizeof(*cfg));
5104 if (rtnh->rtnh_ifindex)
5105 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5106
5107 attrlen = rtnh_attrlen(rtnh);
5108 if (attrlen > 0) {
5109 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5110
5111 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5112 if (nla) {
67b61f6c 5113 r_cfg.fc_gateway = nla_get_in6_addr(nla);
51ebd318
ND		 5114 r_cfg.fc_flags |= RTF_GATEWAY;
5115 }
19e42e45
RP		 5116 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5117 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5118 if (nla)
5119 r_cfg.fc_encap_type = nla_get_u16(nla);
51ebd318 5120 }
6b9ea5a6 5121
68e2ffde 5122 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
acb54e3c 5123 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
8c5b83f0
RP		 5124 if (IS_ERR(rt)) {
5125 err = PTR_ERR(rt);
5126 rt = NULL;
6b9ea5a6 5127 goto cleanup;
8c5b83f0 5128 }
b5d2d75e
DA		 5129 if (!rt6_qualify_for_ecmp(rt)) {
5130 err = -EINVAL;
5131 NL_SET_ERR_MSG(extack,
5132 "Device only routes can not be added for IPv6 using the multipath API.");
5133 fib6_info_release(rt);
5134 goto cleanup;
5135 }
6b9ea5a6 5136
1cf844c7 5137 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
398958ae 5138
d4ead6b3
DA		 5139 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5140 rt, &r_cfg);
51ebd318 5141 if (err) {
93531c67 5142 fib6_info_release(rt);
6b9ea5a6
RP		 5143 goto cleanup;
5144 }
5145
5146 rtnh = rtnh_next(rtnh, &remaining);
5147 }
5148
9eee3b49
IS		 5149 if (list_empty(&rt6_nh_list)) {
5150 NL_SET_ERR_MSG(extack,
5151 "Invalid nexthop configuration - no valid nexthops");
5152 return -EINVAL;
5153 }
5154
3b1137fe
DA		 5155 /* for add and replace send one notification with all nexthops.
5156 * Skip the notification in fib6_add_rt2node and send one with
5157 * the full route when done
5158 */
5159 info->skip_notify = 1;
5160
ebee3cad
IS		 5161 /* For add and replace, send one notification with all nexthops. For
5162 * append, send one notification with all appended nexthops.
5163 */
5164 info->skip_notify_kernel = 1;
5165
6b9ea5a6
RP		 5166 err_nh = NULL;
5167 list_for_each_entry(nh, &rt6_nh_list, next) {
8d1c802b
DA		 5168 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5169 fib6_info_release(nh->fib6_info);
93531c67 5170
f7225172
DA		 5171 if (!err) {
5172 /* save reference to last route successfully inserted */
5173 rt_last = nh->fib6_info;
5174
5175 /* save reference to first route for notification */
5176 if (!rt_notif)
5177 rt_notif = nh->fib6_info;
5178 }
3b1137fe 5179
8d1c802b
DA		 5180 /* nh->fib6_info is used or freed at this point, reset to NULL*/
5181 nh->fib6_info = NULL;
6b9ea5a6
RP		 5182 if (err) {
5183 if (replace && nhn)
a5a82d84
JK		 5184 NL_SET_ERR_MSG_MOD(extack,
5185 "multipath route replace failed (check consistency of installed routes)");
6b9ea5a6
RP		 5186 err_nh = nh;
5187 goto add_errout;
51ebd318 5188 }
6b9ea5a6 5189
1a72418b 5190 /* Because each route is added like a single route we remove
27596472
MK		 5191 * these flags after the first nexthop: if there is a collision,
5192 * we have already failed to add the first nexthop:
5193 * fib6_add_rt2node() has rejected it; when replacing, old
5194 * nexthops have been replaced by first new, the rest should
5195 * be added to it.
1a72418b 5196 */
27596472
MK		 5197 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5198 NLM_F_REPLACE);
6b9ea5a6
RP		 5199 nhn++;
5200 }
5201
0ee0f47c
IS		 5202 /* An in-kernel notification should only be sent in case the new
5203 * multipath route is added as the first route in the node, or if
5204 * it was appended to it. We pass 'rt_notif' since it is the first
5205 * sibling and might allow us to skip some checks in the replace case.
5206 */
5207 if (ip6_route_mpath_should_notify(rt_notif)) {
5208 enum fib_event_type fib_event;
5209
5210 if (rt_notif->fib6_nsiblings != nhn - 1)
5211 fib_event = FIB_EVENT_ENTRY_APPEND;
5212 else
5213 fib_event = FIB_EVENT_ENTRY_REPLACE_TMP;
5214
5215 err = call_fib6_multipath_entry_notifiers(info->nl_net,
5216 fib_event, rt_notif,
5217 nhn - 1, extack);
5218 if (err) {
5219 /* Delete all the siblings that were just added */
5220 err_nh = NULL;
5221 goto add_errout;
5222 }
5223 }
ebee3cad
IS		 5224 event_type = replace ? FIB_EVENT_ENTRY_REPLACE : FIB_EVENT_ENTRY_ADD;
5225 err = call_fib6_multipath_entry_notifiers(info->nl_net, event_type,
5226 rt_notif, nhn - 1, extack);
5227 if (err) {
5228 /* Delete all the siblings that were just added */
5229 err_nh = NULL;
5230 goto add_errout;
5231 }
5232
3b1137fe
DA		 5233 /* success ... tell user about new route */
5234 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
6b9ea5a6
RP		 5235 goto cleanup;
5236
5237add_errout:
3b1137fe
DA		 5238 /* send notification for routes that were added so that
5239 * the delete notifications sent by ip6_route_del are
5240 * coherent
5241 */
5242 if (rt_notif)
5243 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5244
6b9ea5a6
RP		 5245 /* Delete routes that were already added */
5246 list_for_each_entry(nh, &rt6_nh_list, next) {
5247 if (err_nh == nh)
5248 break;
333c4301 5249 ip6_route_del(&nh->r_cfg, extack);
6b9ea5a6
RP		 5250 }
5251
5252cleanup:
5253 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
8d1c802b
DA		 5254 if (nh->fib6_info)
5255 fib6_info_release(nh->fib6_info);
6b9ea5a6
RP		 5256 list_del(&nh->next);
5257 kfree(nh);
5258 }
5259
5260 return err;
5261}
5262
333c4301
DA		 5263static int ip6_route_multipath_del(struct fib6_config *cfg,
5264 struct netlink_ext_ack *extack)
6b9ea5a6
RP		 5265{
5266 struct fib6_config r_cfg;
5267 struct rtnexthop *rtnh;
5268 int remaining;
5269 int attrlen;
5270 int err = 1, last_err = 0;
5271
5272 remaining = cfg->fc_mp_len;
5273 rtnh = (struct rtnexthop *)cfg->fc_mp;
5274
5275 /* Parse a Multipath Entry */
5276 while (rtnh_ok(rtnh, remaining)) {
5277 memcpy(&r_cfg, cfg, sizeof(*cfg));
5278 if (rtnh->rtnh_ifindex)
5279 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5280
5281 attrlen = rtnh_attrlen(rtnh);
5282 if (attrlen > 0) {
5283 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5284
5285 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5286 if (nla) {
5287 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
5288 r_cfg.fc_flags |= RTF_GATEWAY;
5289 }
5290 }
333c4301 5291 err = ip6_route_del(&r_cfg, extack);
6b9ea5a6
RP		 5292 if (err)
5293 last_err = err;
5294
51ebd318
ND		 5295 rtnh = rtnh_next(rtnh, &remaining);
5296 }
5297
5298 return last_err;
5299}
5300
c21ef3e3
DA		 5301static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5302 struct netlink_ext_ack *extack)
1da177e4 5303{
86872cb5
TG		 5304 struct fib6_config cfg;
5305 int err;
1da177e4 5306
333c4301 5307 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 5308 if (err < 0)
5309 return err;
5310
5b98324e
DA		 5311 if (cfg.fc_nh_id &&
5312 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5313 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5314 return -EINVAL;
5315 }
5316
51ebd318 5317 if (cfg.fc_mp)
333c4301 5318 return ip6_route_multipath_del(&cfg, extack);
0ae81335
DA		 5319 else {
5320 cfg.fc_delete_all_nh = 1;
333c4301 5321 return ip6_route_del(&cfg, extack);
0ae81335 5322 }
1da177e4
LT		 5323}
5324
c21ef3e3
DA		 5325static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5326 struct netlink_ext_ack *extack)
1da177e4 5327{
86872cb5
TG		 5328 struct fib6_config cfg;
5329 int err;
1da177e4 5330
333c4301 5331 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 5332 if (err < 0)
5333 return err;
5334
67f69513
DA		 5335 if (cfg.fc_metric == 0)
5336 cfg.fc_metric = IP6_RT_PRIO_USER;
5337
51ebd318 5338 if (cfg.fc_mp)
333c4301 5339 return ip6_route_multipath_add(&cfg, extack);
51ebd318 5340 else
acb54e3c 5341 return ip6_route_add(&cfg, GFP_KERNEL, extack);
1da177e4
LT		 5342}
5343
a1b7a1f0
DA		 5344/* add the overhead of this fib6_nh to nexthop_len */
5345static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
339bf98f 5346{
a1b7a1f0 5347 int *nexthop_len = arg;
beb1afac 5348
a1b7a1f0
DA		 5349 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5350 + NLA_ALIGN(sizeof(struct rtnexthop))
5351 + nla_total_size(16); /* RTA_GATEWAY */
5352
5353 if (nh->fib_nh_lws) {
5354 /* RTA_ENCAP_TYPE */
5355 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5356 /* RTA_ENCAP */
5357 *nexthop_len += nla_total_size(2);
5358 }
f88d8ea6 5359
a1b7a1f0
DA		 5360 return 0;
5361}
beb1afac 5362
a1b7a1f0
DA		 5363static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5364{
5365 int nexthop_len;
5366
5367 if (f6i->nh) {
5368 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5369 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5370 &nexthop_len);
5371 } else {
5372 struct fib6_nh *nh = f6i->fib6_nh;
5373
5374 nexthop_len = 0;
5375 if (f6i->fib6_nsiblings) {
5376 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
5377 + NLA_ALIGN(sizeof(struct rtnexthop))
5378 + nla_total_size(16) /* RTA_GATEWAY */
5379 + lwtunnel_get_encap_size(nh->fib_nh_lws);
5380
5381 nexthop_len *= f6i->fib6_nsiblings;
5382 }
5383 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
beb1afac
DA		 5384 }
5385
339bf98f
TG		 5386 return NLMSG_ALIGN(sizeof(struct rtmsg))
5387 + nla_total_size(16) /* RTA_SRC */
5388 + nla_total_size(16) /* RTA_DST */
5389 + nla_total_size(16) /* RTA_GATEWAY */
5390 + nla_total_size(16) /* RTA_PREFSRC */
5391 + nla_total_size(4) /* RTA_TABLE */
5392 + nla_total_size(4) /* RTA_IIF */
5393 + nla_total_size(4) /* RTA_OIF */
5394 + nla_total_size(4) /* RTA_PRIORITY */
6a2b9ce0 5395 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
ea697639 5396 + nla_total_size(sizeof(struct rta_cacheinfo))
c78ba6d6 5397 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
19e42e45 5398 + nla_total_size(1) /* RTA_PREF */
beb1afac
DA		 5399 + nexthop_len;
5400}
5401
f88d8ea6
DA		 5402static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5403 unsigned char *flags)
5404{
5405 if (nexthop_is_multipath(nh)) {
5406 struct nlattr *mp;
5407
4255ff05 5408 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
f88d8ea6
DA		 5409 if (!mp)
5410 goto nla_put_failure;
5411
7bdf4de1 5412 if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
f88d8ea6
DA		 5413 goto nla_put_failure;
5414
5415 nla_nest_end(skb, mp);
5416 } else {
5417 struct fib6_nh *fib6_nh;
5418
5419 fib6_nh = nexthop_fib6_nh(nh);
7bdf4de1 5420 if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
f88d8ea6
DA		 5421 flags, false) < 0)
5422 goto nla_put_failure;
5423 }
5424
5425 return 0;
5426
5427nla_put_failure:
5428 return -EMSGSIZE;
5429}
5430
d4ead6b3 5431static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 5432 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 5433 struct in6_addr *dest, struct in6_addr *src,
15e47304 5434 int iif, int type, u32 portid, u32 seq,
f8cfe2ce 5435 unsigned int flags)
1da177e4 5436{
22d0bd82
XL		 5437 struct rt6_info *rt6 = (struct rt6_info *)dst;
5438 struct rt6key *rt6_dst, *rt6_src;
5439 u32 *pmetrics, table, rt6_flags;
f88d8ea6 5440 unsigned char nh_flags = 0;
2d7202bf 5441 struct nlmsghdr *nlh;
22d0bd82 5442 struct rtmsg *rtm;
d4ead6b3 5443 long expires = 0;
1da177e4 5444
15e47304 5445 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
38308473 5446 if (!nlh)
26932566 5447 return -EMSGSIZE;
2d7202bf 5448
22d0bd82
XL		 5449 if (rt6) {
5450 rt6_dst = &rt6->rt6i_dst;
5451 rt6_src = &rt6->rt6i_src;
5452 rt6_flags = rt6->rt6i_flags;
5453 } else {
5454 rt6_dst = &rt->fib6_dst;
5455 rt6_src = &rt->fib6_src;
5456 rt6_flags = rt->fib6_flags;
5457 }
5458
2d7202bf 5459 rtm = nlmsg_data(nlh);
1da177e4 5460 rtm->rtm_family = AF_INET6;
22d0bd82
XL		 5461 rtm->rtm_dst_len = rt6_dst->plen;
5462 rtm->rtm_src_len = rt6_src->plen;
1da177e4 5463 rtm->rtm_tos = 0;
93c2fb25
DA		 5464 if (rt->fib6_table)
5465 table = rt->fib6_table->tb6_id;
c71099ac 5466 else
9e762a4a 5467 table = RT6_TABLE_UNSPEC;
97f0082a 5468 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
c78679e8
DM		 5469 if (nla_put_u32(skb, RTA_TABLE, table))
5470 goto nla_put_failure;
e8478e80
DA		 5471
5472 rtm->rtm_type = rt->fib6_type;
1da177e4
LT		 5473 rtm->rtm_flags = 0;
5474 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
93c2fb25 5475 rtm->rtm_protocol = rt->fib6_protocol;
1da177e4 5476
22d0bd82 5477 if (rt6_flags & RTF_CACHE)
1da177e4
LT		 5478 rtm->rtm_flags |= RTM_F_CLONED;
5479
d4ead6b3
DA		 5480 if (dest) {
5481 if (nla_put_in6_addr(skb, RTA_DST, dest))
c78679e8 5482 goto nla_put_failure;
1ab1457c 5483 rtm->rtm_dst_len = 128;
1da177e4 5484 } else if (rtm->rtm_dst_len)
22d0bd82 5485 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
c78679e8 5486 goto nla_put_failure;
1da177e4
LT		 5487#ifdef CONFIG_IPV6_SUBTREES
5488 if (src) {
930345ea 5489 if (nla_put_in6_addr(skb, RTA_SRC, src))
c78679e8 5490 goto nla_put_failure;
1ab1457c 5491 rtm->rtm_src_len = 128;
c78679e8 5492 } else if (rtm->rtm_src_len &&
22d0bd82 5493 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
c78679e8 5494 goto nla_put_failure;
1da177e4 5495#endif
7bc570c8
YH		 5496 if (iif) {
5497#ifdef CONFIG_IPV6_MROUTE
22d0bd82 5498 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
fd61c6ba
DA		 5499 int err = ip6mr_get_route(net, skb, rtm, portid);
5500
5501 if (err == 0)
5502 return 0;
5503 if (err < 0)
5504 goto nla_put_failure;
7bc570c8
YH		 5505 } else
5506#endif
c78679e8
DM		 5507 if (nla_put_u32(skb, RTA_IIF, iif))
5508 goto nla_put_failure;
d4ead6b3 5509 } else if (dest) {
1da177e4 5510 struct in6_addr saddr_buf;
d4ead6b3 5511 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
930345ea 5512 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 5513 goto nla_put_failure;
1da177e4 5514 }
2d7202bf 5515
93c2fb25 5516 if (rt->fib6_prefsrc.plen) {
c3968a85 5517 struct in6_addr saddr_buf;
93c2fb25 5518 saddr_buf = rt->fib6_prefsrc.addr;
930345ea 5519 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 5520 goto nla_put_failure;
c3968a85
DW		 5521 }
5522
d4ead6b3
DA		 5523 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5524 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
2d7202bf
TG		 5525 goto nla_put_failure;
5526
93c2fb25 5527 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
c78679e8 5528 goto nla_put_failure;
8253947e 5529
beb1afac
DA		 5530 /* For multipath routes, walk the siblings list and add
5531 * each as a nexthop within RTA_MULTIPATH.
5532 */
22d0bd82
XL		 5533 if (rt6) {
5534 if (rt6_flags & RTF_GATEWAY &&
5535 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5536 goto nla_put_failure;
5537
5538 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5539 goto nla_put_failure;
5540 } else if (rt->fib6_nsiblings) {
8d1c802b 5541 struct fib6_info *sibling, *next_sibling;
beb1afac
DA		 5542 struct nlattr *mp;
5543
ae0be8de 5544 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
beb1afac
DA		 5545 if (!mp)
5546 goto nla_put_failure;
5547
1cf844c7 5548 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
7bdf4de1 5549 rt->fib6_nh->fib_nh_weight, AF_INET6) < 0)
beb1afac
DA		 5550 goto nla_put_failure;
5551
5552 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25 5553 &rt->fib6_siblings, fib6_siblings) {
1cf844c7 5554 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
7bdf4de1
DS		 5555 sibling->fib6_nh->fib_nh_weight,
5556 AF_INET6) < 0)
beb1afac
DA		 5557 goto nla_put_failure;
5558 }
5559
5560 nla_nest_end(skb, mp);
f88d8ea6
DA		 5561 } else if (rt->nh) {
5562 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5563 goto nla_put_failure;
ecc5663c 5564
f88d8ea6
DA		 5565 if (nexthop_is_blackhole(rt->nh))
5566 rtm->rtm_type = RTN_BLACKHOLE;
5567
5568 if (rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5569 goto nla_put_failure;
5570
5571 rtm->rtm_flags |= nh_flags;
5572 } else {
7bdf4de1 5573 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
ecc5663c 5574 &nh_flags, false) < 0)
beb1afac 5575 goto nla_put_failure;
ecc5663c
DA		 5576
5577 rtm->rtm_flags |= nh_flags;
beb1afac
DA		 5578 }
5579
22d0bd82 5580 if (rt6_flags & RTF_EXPIRES) {
14895687
DA		 5581 expires = dst ? dst->expires : rt->expires;
5582 expires -= jiffies;
5583 }
69cdf8f9 5584
d4ead6b3 5585 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
e3703b3d 5586 goto nla_put_failure;
2d7202bf 5587
22d0bd82 5588 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
c78ba6d6
LR		 5589 goto nla_put_failure;
5590
19e42e45 5591
053c095a
JB		 5592 nlmsg_end(skb, nlh);
5593 return 0;
2d7202bf
TG		 5594
5595nla_put_failure:
26932566
PM		 5596 nlmsg_cancel(skb, nlh);
5597 return -EMSGSIZE;
1da177e4
LT		 5598}
5599
2c170e07
DA		 5600static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5601{
5602 const struct net_device *dev = arg;
5603
5604 if (nh->fib_nh_dev == dev)
5605 return 1;
5606
5607 return 0;
5608}
5609
13e38901
DA		 5610static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5611 const struct net_device *dev)
5612{
2c170e07
DA		 5613 if (f6i->nh) {
5614 struct net_device *_dev = (struct net_device *)dev;
5615
5616 return !!nexthop_for_each_fib6_nh(f6i->nh,
5617 fib6_info_nh_uses_dev,
5618 _dev);
5619 }
5620
1cf844c7 5621 if (f6i->fib6_nh->fib_nh_dev == dev)
13e38901
DA		 5622 return true;
5623
5624 if (f6i->fib6_nsiblings) {
5625 struct fib6_info *sibling, *next_sibling;
5626
5627 list_for_each_entry_safe(sibling, next_sibling,
5628 &f6i->fib6_siblings, fib6_siblings) {
1cf844c7 5629 if (sibling->fib6_nh->fib_nh_dev == dev)
13e38901
DA		 5630 return true;
5631 }
5632 }
5633
5634 return false;
5635}
5636
1e47b483
SB		 5637struct fib6_nh_exception_dump_walker {
5638 struct rt6_rtnl_dump_arg *dump;
5639 struct fib6_info *rt;
5640 unsigned int flags;
5641 unsigned int skip;
5642 unsigned int count;
5643};
5644
5645static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5646{
5647 struct fib6_nh_exception_dump_walker *w = arg;
5648 struct rt6_rtnl_dump_arg *dump = w->dump;
5649 struct rt6_exception_bucket *bucket;
5650 struct rt6_exception *rt6_ex;
5651 int i, err;
5652
5653 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5654 if (!bucket)
5655 return 0;
5656
5657 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5658 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5659 if (w->skip) {
5660 w->skip--;
5661 continue;
5662 }
5663
5664 /* Expiration of entries doesn't bump sernum, insertion
5665 * does. Removal is triggered by insertion, so we can
5666 * rely on the fact that if entries change between two
5667 * partial dumps, this node is scanned again completely,
5668 * see rt6_insert_exception() and fib6_dump_table().
5669 *
5670 * Count expired entries we go through as handled
5671 * entries that we'll skip next time, in case of partial
5672 * node dump. Otherwise, if entries expire meanwhile,
5673 * we'll skip the wrong amount.
5674 */
5675 if (rt6_check_expired(rt6_ex->rt6i)) {
5676 w->count++;
5677 continue;
5678 }
5679
5680 err = rt6_fill_node(dump->net, dump->skb, w->rt,
5681 &rt6_ex->rt6i->dst, NULL, NULL, 0,
5682 RTM_NEWROUTE,
5683 NETLINK_CB(dump->cb->skb).portid,
5684 dump->cb->nlh->nlmsg_seq, w->flags);
5685 if (err)
5686 return err;
5687
5688 w->count++;
5689 }
5690 bucket++;
5691 }
5692
5693 return 0;
5694}
5695
bf9a8a06 5696/* Return -1 if done with node, number of handled routes on partial dump */
1e47b483 5697int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
1da177e4
LT		 5698{
5699 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
13e38901
DA		 5700 struct fib_dump_filter *filter = &arg->filter;
5701 unsigned int flags = NLM_F_MULTI;
1f17e2f2 5702 struct net *net = arg->net;
1e47b483 5703 int count = 0;
1f17e2f2 5704
421842ed 5705 if (rt == net->ipv6.fib6_null_entry)
bf9a8a06 5706 return -1;
1da177e4 5707
13e38901
DA		 5708 if ((filter->flags & RTM_F_PREFIX) &&
5709 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5710 /* success since this is not a prefix route */
bf9a8a06 5711 return -1;
13e38901 5712 }
1e47b483
SB		 5713 if (filter->filter_set &&
5714 ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5715 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5716 (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5717 return -1;
5718 }
5719
5720 if (filter->filter_set ||
5721 !filter->dump_routes || !filter->dump_exceptions) {
13e38901 5722 flags |= NLM_F_DUMP_FILTERED;
f8cfe2ce 5723 }
1da177e4 5724
1e47b483
SB		 5725 if (filter->dump_routes) {
5726 if (skip) {
5727 skip--;
5728 } else {
5729 if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5730 0, RTM_NEWROUTE,
5731 NETLINK_CB(arg->cb->skb).portid,
5732 arg->cb->nlh->nlmsg_seq, flags)) {
5733 return 0;
5734 }
5735 count++;
5736 }
5737 }
5738
5739 if (filter->dump_exceptions) {
5740 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5741 .rt = rt,
5742 .flags = flags,
5743 .skip = skip,
5744 .count = 0 };
5745 int err;
5746
3b525691 5747 rcu_read_lock();
1e47b483
SB		 5748 if (rt->nh) {
5749 err = nexthop_for_each_fib6_nh(rt->nh,
5750 rt6_nh_dump_exceptions,
5751 &w);
5752 } else {
5753 err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5754 }
3b525691 5755 rcu_read_unlock();
1e47b483
SB		 5756
5757 if (err)
5758 return count += w.count;
5759 }
bf9a8a06
SB		 5760
5761 return -1;
1da177e4
LT		 5762}
5763
0eff0a27
JK		 5764static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5765 const struct nlmsghdr *nlh,
5766 struct nlattr **tb,
5767 struct netlink_ext_ack *extack)
5768{
5769 struct rtmsg *rtm;
5770 int i, err;
5771
5772 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5773 NL_SET_ERR_MSG_MOD(extack,
5774 "Invalid header for get route request");
5775 return -EINVAL;
5776 }
5777
5778 if (!netlink_strict_get_check(skb))
8cb08174
JB		 5779 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5780 rtm_ipv6_policy, extack);
0eff0a27
JK		 5781
5782 rtm = nlmsg_data(nlh);
5783 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5784 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5785 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5786 rtm->rtm_type) {
5787 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5788 return -EINVAL;
5789 }
5790 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5791 NL_SET_ERR_MSG_MOD(extack,
5792 "Invalid flags for get route request");
5793 return -EINVAL;
5794 }
5795
8cb08174
JB		 5796 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5797 rtm_ipv6_policy, extack);
0eff0a27
JK		 5798 if (err)
5799 return err;
5800
5801 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5802 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5803 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5804 return -EINVAL;
5805 }
5806
5807 for (i = 0; i <= RTA_MAX; i++) {
5808 if (!tb[i])
5809 continue;
5810
5811 switch (i) {
5812 case RTA_SRC:
5813 case RTA_DST:
5814 case RTA_IIF:
5815 case RTA_OIF:
5816 case RTA_MARK:
5817 case RTA_UID:
5818 case RTA_SPORT:
5819 case RTA_DPORT:
5820 case RTA_IP_PROTO:
5821 break;
5822 default:
5823 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
5824 return -EINVAL;
5825 }
5826 }
5827
5828 return 0;
5829}
5830
c21ef3e3
DA		 5831static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5832 struct netlink_ext_ack *extack)
1da177e4 5833{
3b1e0a65 5834 struct net *net = sock_net(in_skb->sk);
ab364a6f 5835 struct nlattr *tb[RTA_MAX+1];
18c3a61c 5836 int err, iif = 0, oif = 0;
a68886a6 5837 struct fib6_info *from;
18c3a61c 5838 struct dst_entry *dst;
ab364a6f 5839 struct rt6_info *rt;
1da177e4 5840 struct sk_buff *skb;
ab364a6f 5841 struct rtmsg *rtm;
744486d4 5842 struct flowi6 fl6 = {};
18c3a61c 5843 bool fibmatch;
1da177e4 5844
0eff0a27 5845 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
ab364a6f
TG		 5846 if (err < 0)
5847 goto errout;
1da177e4 5848
ab364a6f 5849 err = -EINVAL;
38b7097b
HFS		 5850 rtm = nlmsg_data(nlh);
5851 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
18c3a61c 5852 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
1da177e4 5853
ab364a6f
TG		 5854 if (tb[RTA_SRC]) {
5855 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
5856 goto errout;
5857
4e3fd7a0 5858 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
ab364a6f
TG		 5859 }
5860
5861 if (tb[RTA_DST]) {
5862 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
5863 goto errout;
5864
4e3fd7a0 5865 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
ab364a6f
TG		 5866 }
5867
5868 if (tb[RTA_IIF])
5869 iif = nla_get_u32(tb[RTA_IIF]);
5870
5871 if (tb[RTA_OIF])
72331bc0 5872 oif = nla_get_u32(tb[RTA_OIF]);
1da177e4 5873
2e47b291
LC		 5874 if (tb[RTA_MARK])
5875 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
5876
622ec2c9
LC		 5877 if (tb[RTA_UID])
5878 fl6.flowi6_uid = make_kuid(current_user_ns(),
5879 nla_get_u32(tb[RTA_UID]));
5880 else
5881 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
5882
eacb9384
RP		 5883 if (tb[RTA_SPORT])
5884 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
5885
5886 if (tb[RTA_DPORT])
5887 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
5888
5889 if (tb[RTA_IP_PROTO]) {
5890 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
5e1a99ea
HL		 5891 &fl6.flowi6_proto, AF_INET6,
5892 extack);
eacb9384
RP		 5893 if (err)
5894 goto errout;
5895 }
5896
1da177e4
LT		 5897 if (iif) {
5898 struct net_device *dev;
72331bc0
SL		 5899 int flags = 0;
5900
121622db
FW		 5901 rcu_read_lock();
5902
5903 dev = dev_get_by_index_rcu(net, iif);
1da177e4 5904 if (!dev) {
121622db 5905 rcu_read_unlock();
1da177e4 5906 err = -ENODEV;
ab364a6f 5907 goto errout;
1da177e4 5908 }
72331bc0
SL		 5909
5910 fl6.flowi6_iif = iif;
5911
5912 if (!ipv6_addr_any(&fl6.saddr))
5913 flags |= RT6_LOOKUP_F_HAS_SADDR;
5914
b75cc8f9 5915 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
121622db
FW		 5916
5917 rcu_read_unlock();
72331bc0
SL		 5918 } else {
5919 fl6.flowi6_oif = oif;
5920
58acfd71 5921 dst = ip6_route_output(net, NULL, &fl6);
18c3a61c
RP		 5922 }
5923
18c3a61c
RP		 5924
5925 rt = container_of(dst, struct rt6_info, dst);
5926 if (rt->dst.error) {
5927 err = rt->dst.error;
5928 ip6_rt_put(rt);
5929 goto errout;
1da177e4
LT		 5930 }
5931
9d6acb3b
WC		 5932 if (rt == net->ipv6.ip6_null_entry) {
5933 err = rt->dst.error;
5934 ip6_rt_put(rt);
5935 goto errout;
5936 }
5937
ab364a6f 5938 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
38308473 5939 if (!skb) {
94e187c0 5940 ip6_rt_put(rt);
ab364a6f
TG		 5941 err = -ENOBUFS;
5942 goto errout;
5943 }
1da177e4 5944
d8d1f30b 5945 skb_dst_set(skb, &rt->dst);
a68886a6
DA		 5946
5947 rcu_read_lock();
5948 from = rcu_dereference(rt->from);
886b7a50
MKL		 5949 if (from) {
5950 if (fibmatch)
5951 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
5952 iif, RTM_NEWROUTE,
5953 NETLINK_CB(in_skb).portid,
5954 nlh->nlmsg_seq, 0);
5955 else
5956 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
5957 &fl6.saddr, iif, RTM_NEWROUTE,
5958 NETLINK_CB(in_skb).portid,
5959 nlh->nlmsg_seq, 0);
5960 } else {
5961 err = -ENETUNREACH;
5962 }
a68886a6
DA		 5963 rcu_read_unlock();
5964
1da177e4 5965 if (err < 0) {
ab364a6f
TG		 5966 kfree_skb(skb);
5967 goto errout;
1da177e4
LT		 5968 }
5969
15e47304 5970 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
ab364a6f 5971errout:
1da177e4 5972 return err;
1da177e4
LT		 5973}
5974
8d1c802b 5975void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
37a1d361 5976 unsigned int nlm_flags)
1da177e4
LT		 5977{
5978 struct sk_buff *skb;
5578689a 5979 struct net *net = info->nl_net;
528c4ceb
DL		 5980 u32 seq;
5981 int err;
5982
5983 err = -ENOBUFS;
38308473 5984 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
86872cb5 5985
19e42e45 5986 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
38308473 5987 if (!skb)
21713ebc
TG		 5988 goto errout;
5989
d4ead6b3
DA		 5990 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
5991 event, info->portid, seq, nlm_flags);
26932566
PM		 5992 if (err < 0) {
5993 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
5994 WARN_ON(err == -EMSGSIZE);
5995 kfree_skb(skb);
5996 goto errout;
5997 }
15e47304 5998 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
1ce85fe4
PNA		 5999 info->nlh, gfp_any());
6000 return;
21713ebc
TG		 6001errout:
6002 if (err < 0)
5578689a 6003 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
1da177e4
LT		 6004}
6005
19a3b7ee
DA		 6006void fib6_rt_update(struct net *net, struct fib6_info *rt,
6007 struct nl_info *info)
6008{
6009 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6010 struct sk_buff *skb;
6011 int err = -ENOBUFS;
6012
6013 /* call_fib6_entry_notifiers will be removed when in-kernel notifier
6014 * is implemented and supported for nexthop objects
6015 */
6016 call_fib6_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, rt, NULL);
6017
6018 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6019 if (!skb)
6020 goto errout;
6021
6022 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6023 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6024 if (err < 0) {
6025 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6026 WARN_ON(err == -EMSGSIZE);
6027 kfree_skb(skb);
6028 goto errout;
6029 }
6030 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6031 info->nlh, gfp_any());
6032 return;
6033errout:
6034 if (err < 0)
6035 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6036}
6037
8ed67789 6038static int ip6_route_dev_notify(struct notifier_block *this,
351638e7 6039 unsigned long event, void *ptr)
8ed67789 6040{
351638e7 6041 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
c346dca1 6042 struct net *net = dev_net(dev);
8ed67789 6043
242d3a49
WC		 6044 if (!(dev->flags & IFF_LOOPBACK))
6045 return NOTIFY_OK;
6046
6047 if (event == NETDEV_REGISTER) {
1cf844c7 6048 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
d8d1f30b 6049 net->ipv6.ip6_null_entry->dst.dev = dev;
8ed67789
DL		 6050 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6051#ifdef CONFIG_IPV6_MULTIPLE_TABLES
d8d1f30b 6052 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
8ed67789 6053 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
d8d1f30b 6054 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
8ed67789 6055 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
242d3a49 6056#endif
76da0704
WC		 6057 } else if (event == NETDEV_UNREGISTER &&
6058 dev->reg_state != NETREG_UNREGISTERED) {
6059 /* NETDEV_UNREGISTER could be fired for multiple times by
6060 * netdev_wait_allrefs(). Make sure we only call this once.
6061 */
12d94a80 6062 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
242d3a49 6063#ifdef CONFIG_IPV6_MULTIPLE_TABLES
12d94a80
ED		 6064 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6065 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
8ed67789
DL		 6066#endif
6067 }
6068
6069 return NOTIFY_OK;
6070}
6071
1da177e4
LT		 6072/*
6073 * /proc
6074 */
6075
6076#ifdef CONFIG_PROC_FS
1da177e4
LT		 6077static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6078{
69ddb805 6079 struct net *net = (struct net *)seq->private;
1da177e4 6080 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
69ddb805
DL		 6081 net->ipv6.rt6_stats->fib_nodes,
6082 net->ipv6.rt6_stats->fib_route_nodes,
81eb8447 6083 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
69ddb805
DL		 6084 net->ipv6.rt6_stats->fib_rt_entries,
6085 net->ipv6.rt6_stats->fib_rt_cache,
fc66f95c 6086 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
69ddb805 6087 net->ipv6.rt6_stats->fib_discarded_routes);
1da177e4
LT		 6088
6089 return 0;
6090}
1da177e4
LT		 6091#endif /* CONFIG_PROC_FS */
6092
6093#ifdef CONFIG_SYSCTL
6094
1da177e4 6095static
fe2c6338 6096int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
1da177e4
LT		 6097 void __user *buffer, size_t *lenp, loff_t *ppos)
6098{
c486da34
LAG		 6099 struct net *net;
6100 int delay;
f0fb9b28 6101 int ret;
c486da34 6102 if (!write)
1da177e4 6103 return -EINVAL;
c486da34
LAG		 6104
6105 net = (struct net *)ctl->extra1;
6106 delay = net->ipv6.sysctl.flush_delay;
f0fb9b28
AP		 6107 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6108 if (ret)
6109 return ret;
6110
2ac3ac8f 6111 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
c486da34 6112 return 0;
1da177e4
LT		 6113}
6114
ed792e28 6115static struct ctl_table ipv6_route_table_template[] = {
1ab1457c 6116 {
1da177e4 6117 .procname = "flush",
4990509f 6118 .data = &init_net.ipv6.sysctl.flush_delay,
1da177e4 6119 .maxlen = sizeof(int),
89c8b3a1 6120 .mode = 0200,
6d9f239a 6121 .proc_handler = ipv6_sysctl_rtcache_flush
1da177e4
LT		 6122 },
6123 {
1da177e4 6124 .procname = "gc_thresh",
9a7ec3a9 6125 .data = &ip6_dst_ops_template.gc_thresh,
1da177e4
LT		 6126 .maxlen = sizeof(int),
6127 .mode = 0644,
6d9f239a 6128 .proc_handler = proc_dointvec,
1da177e4
LT		 6129 },
6130 {
1da177e4 6131 .procname = "max_size",
4990509f 6132 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
1da177e4
LT		 6133 .maxlen = sizeof(int),
6134 .mode = 0644,
6d9f239a 6135 .proc_handler = proc_dointvec,
1da177e4
LT		 6136 },
6137 {
1da177e4 6138 .procname = "gc_min_interval",
4990509f 6139 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 6140 .maxlen = sizeof(int),
6141 .mode = 0644,
6d9f239a 6142 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6143 },
6144 {
1da177e4 6145 .procname = "gc_timeout",
4990509f 6146 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
1da177e4
LT		 6147 .maxlen = sizeof(int),
6148 .mode = 0644,
6d9f239a 6149 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6150 },
6151 {
1da177e4 6152 .procname = "gc_interval",
4990509f 6153 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
1da177e4
LT		 6154 .maxlen = sizeof(int),
6155 .mode = 0644,
6d9f239a 6156 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6157 },
6158 {
1da177e4 6159 .procname = "gc_elasticity",
4990509f 6160 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
1da177e4
LT		 6161 .maxlen = sizeof(int),
6162 .mode = 0644,
f3d3f616 6163 .proc_handler = proc_dointvec,
1da177e4
LT		 6164 },
6165 {
1da177e4 6166 .procname = "mtu_expires",
4990509f 6167 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
1da177e4
LT		 6168 .maxlen = sizeof(int),
6169 .mode = 0644,
6d9f239a 6170 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6171 },
6172 {
1da177e4 6173 .procname = "min_adv_mss",
4990509f 6174 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
1da177e4
LT		 6175 .maxlen = sizeof(int),
6176 .mode = 0644,
f3d3f616 6177 .proc_handler = proc_dointvec,
1da177e4
LT		 6178 },
6179 {
1da177e4 6180 .procname = "gc_min_interval_ms",
4990509f 6181 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 6182 .maxlen = sizeof(int),
6183 .mode = 0644,
6d9f239a 6184 .proc_handler = proc_dointvec_ms_jiffies,
1da177e4 6185 },
7c6bb7d2
DA		 6186 {
6187 .procname = "skip_notify_on_dev_down",
6188 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6189 .maxlen = sizeof(int),
6190 .mode = 0644,
b8e8a863 6191 .proc_handler = proc_dointvec_minmax,
eec4844f
MC		 6192 .extra1 = SYSCTL_ZERO,
6193 .extra2 = SYSCTL_ONE,
7c6bb7d2 6194 },
f8572d8f 6195 { }
1da177e4
LT		 6196};
6197
2c8c1e72 6198struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
760f2d01
DL		 6199{
6200 struct ctl_table *table;
6201
6202 table = kmemdup(ipv6_route_table_template,
6203 sizeof(ipv6_route_table_template),
6204 GFP_KERNEL);
5ee09105
YH		 6205
6206 if (table) {
6207 table[0].data = &net->ipv6.sysctl.flush_delay;
c486da34 6208 table[0].extra1 = net;
86393e52 6209 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
5ee09105
YH		 6210 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
6211 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6212 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6213 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6214 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6215 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6216 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
9c69fabe 6217 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
7c6bb7d2 6218 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
464dc801
EB		 6219
6220 /* Don't export sysctls to unprivileged users */
6221 if (net->user_ns != &init_user_ns)
6222 table[0].procname = NULL;
5ee09105
YH		 6223 }
6224
760f2d01
DL		 6225 return table;
6226}
1da177e4
LT		 6227#endif
6228
2c8c1e72 6229static int __net_init ip6_route_net_init(struct net *net)
cdb18761 6230{
633d424b 6231 int ret = -ENOMEM;
8ed67789 6232
86393e52
AD		 6233 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6234 sizeof(net->ipv6.ip6_dst_ops));
f2fc6a54 6235
fc66f95c
ED		 6236 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6237 goto out_ip6_dst_ops;
6238
1cf844c7 6239 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
421842ed
DA		 6240 if (!net->ipv6.fib6_null_entry)
6241 goto out_ip6_dst_entries;
1cf844c7
DA		 6242 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6243 sizeof(*net->ipv6.fib6_null_entry));
421842ed 6244
8ed67789
DL		 6245 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6246 sizeof(*net->ipv6.ip6_null_entry),
6247 GFP_KERNEL);
6248 if (!net->ipv6.ip6_null_entry)
421842ed 6249 goto out_fib6_null_entry;
d8d1f30b 6250 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6251 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6252 ip6_template_metrics, true);
74109218 6253 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->rt6i_uncached);
8ed67789
DL		 6254
6255#ifdef CONFIG_IPV6_MULTIPLE_TABLES
feca7d8c 6256 net->ipv6.fib6_has_custom_rules = false;
8ed67789
DL		 6257 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6258 sizeof(*net->ipv6.ip6_prohibit_entry),
6259 GFP_KERNEL);
68fffc67
PZ		 6260 if (!net->ipv6.ip6_prohibit_entry)
6261 goto out_ip6_null_entry;
d8d1f30b 6262 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6263 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6264 ip6_template_metrics, true);
74109218 6265 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->rt6i_uncached);
8ed67789
DL		 6266
6267 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6268 sizeof(*net->ipv6.ip6_blk_hole_entry),
6269 GFP_KERNEL);
68fffc67
PZ		 6270 if (!net->ipv6.ip6_blk_hole_entry)
6271 goto out_ip6_prohibit_entry;
d8d1f30b 6272 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6273 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6274 ip6_template_metrics, true);
74109218 6275 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->rt6i_uncached);
b9b33e7c
PA		 6276#ifdef CONFIG_IPV6_SUBTREES
6277 net->ipv6.fib6_routes_require_src = 0;
6278#endif
8ed67789
DL		 6279#endif
6280
b339a47c
PZ		 6281 net->ipv6.sysctl.flush_delay = 0;
6282 net->ipv6.sysctl.ip6_rt_max_size = 4096;
6283 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6284 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6285 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6286 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6287 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6288 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
7c6bb7d2 6289 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
b339a47c 6290
6891a346
BT		 6291 net->ipv6.ip6_rt_gc_expire = 30*HZ;
6292
8ed67789
DL		 6293 ret = 0;
6294out:
6295 return ret;
f2fc6a54 6296
68fffc67
PZ		 6297#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6298out_ip6_prohibit_entry:
6299 kfree(net->ipv6.ip6_prohibit_entry);
6300out_ip6_null_entry:
6301 kfree(net->ipv6.ip6_null_entry);
6302#endif
421842ed
DA		 6303out_fib6_null_entry:
6304 kfree(net->ipv6.fib6_null_entry);
fc66f95c
ED		 6305out_ip6_dst_entries:
6306 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
f2fc6a54 6307out_ip6_dst_ops:
f2fc6a54 6308 goto out;
cdb18761
DL		 6309}
6310
2c8c1e72 6311static void __net_exit ip6_route_net_exit(struct net *net)
cdb18761 6312{
421842ed 6313 kfree(net->ipv6.fib6_null_entry);
8ed67789
DL		 6314 kfree(net->ipv6.ip6_null_entry);
6315#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6316 kfree(net->ipv6.ip6_prohibit_entry);
6317 kfree(net->ipv6.ip6_blk_hole_entry);
6318#endif
41bb78b4 6319 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
cdb18761
DL		 6320}
6321
d189634e
TG		 6322static int __net_init ip6_route_net_init_late(struct net *net)
6323{
6324#ifdef CONFIG_PROC_FS
c3506372
CH		 6325 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6326 sizeof(struct ipv6_route_iter));
3617d949
CH		 6327 proc_create_net_single("rt6_stats", 0444, net->proc_net,
6328 rt6_stats_seq_show, NULL);
d189634e
TG		 6329#endif
6330 return 0;
6331}
6332
6333static void __net_exit ip6_route_net_exit_late(struct net *net)
6334{
6335#ifdef CONFIG_PROC_FS
ece31ffd
G		 6336 remove_proc_entry("ipv6_route", net->proc_net);
6337 remove_proc_entry("rt6_stats", net->proc_net);
d189634e
TG		 6338#endif
6339}
6340
cdb18761
DL		 6341static struct pernet_operations ip6_route_net_ops = {
6342 .init = ip6_route_net_init,
6343 .exit = ip6_route_net_exit,
6344};
6345
c3426b47
DM		 6346static int __net_init ipv6_inetpeer_init(struct net *net)
6347{
6348 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6349
6350 if (!bp)
6351 return -ENOMEM;
6352 inet_peer_base_init(bp);
6353 net->ipv6.peers = bp;
6354 return 0;
6355}
6356
6357static void __net_exit ipv6_inetpeer_exit(struct net *net)
6358{
6359 struct inet_peer_base *bp = net->ipv6.peers;
6360
6361 net->ipv6.peers = NULL;
56a6b248 6362 inetpeer_invalidate_tree(bp);
c3426b47
DM		 6363 kfree(bp);
6364}
6365
2b823f72 6366static struct pernet_operations ipv6_inetpeer_ops = {
c3426b47
DM		 6367 .init = ipv6_inetpeer_init,
6368 .exit = ipv6_inetpeer_exit,
6369};
6370
d189634e
TG		 6371static struct pernet_operations ip6_route_net_late_ops = {
6372 .init = ip6_route_net_init_late,
6373 .exit = ip6_route_net_exit_late,
6374};
6375
8ed67789
DL		 6376static struct notifier_block ip6_route_dev_notifier = {
6377 .notifier_call = ip6_route_dev_notify,
242d3a49 6378 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
8ed67789
DL		 6379};
6380
2f460933
WC		 6381void __init ip6_route_init_special_entries(void)
6382{
6383 /* Registering of the loopback is done before this portion of code,
6384 * the loopback reference in rt6_info will not be taken, do it
6385 * manually for init_net */
1cf844c7 6386 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
2f460933
WC		 6387 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6388 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6389 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6390 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6391 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6392 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6393 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6394 #endif
6395}
6396
433d49c3 6397int __init ip6_route_init(void)
1da177e4 6398{
433d49c3 6399 int ret;
8d0b94af 6400 int cpu;
433d49c3 6401
9a7ec3a9
DL		 6402 ret = -ENOMEM;
6403 ip6_dst_ops_template.kmem_cachep =
e5d679f3 6404 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
f845ab6b 6405 SLAB_HWCACHE_ALIGN, NULL);
9a7ec3a9 6406 if (!ip6_dst_ops_template.kmem_cachep)
c19a28e1 6407 goto out;
14e50e57 6408
fc66f95c 6409 ret = dst_entries_init(&ip6_dst_blackhole_ops);
8ed67789 6410 if (ret)
bdb3289f 6411 goto out_kmem_cache;
bdb3289f 6412
c3426b47
DM		 6413 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6414 if (ret)
e8803b6c 6415 goto out_dst_entries;
2a0c451a 6416
7e52b33b
DM		 6417 ret = register_pernet_subsys(&ip6_route_net_ops);
6418 if (ret)
6419 goto out_register_inetpeer;
c3426b47 6420
5dc121e9
AE		 6421 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6422
e8803b6c 6423 ret = fib6_init();
433d49c3 6424 if (ret)
8ed67789 6425 goto out_register_subsys;
433d49c3 6426
433d49c3
DL		 6427 ret = xfrm6_init();
6428 if (ret)
e8803b6c 6429 goto out_fib6_init;
c35b7e72 6430
433d49c3
DL		 6431 ret = fib6_rules_init();
6432 if (ret)
6433 goto xfrm6_init;
7e5449c2 6434
d189634e
TG		 6435 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6436 if (ret)
6437 goto fib6_rules_init;
6438
16feebcf
FW		 6439 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6440 inet6_rtm_newroute, NULL, 0);
6441 if (ret < 0)
6442 goto out_register_late_subsys;
6443
6444 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6445 inet6_rtm_delroute, NULL, 0);
6446 if (ret < 0)
6447 goto out_register_late_subsys;
6448
6449 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6450 inet6_rtm_getroute, NULL,
6451 RTNL_FLAG_DOIT_UNLOCKED);
6452 if (ret < 0)
d189634e 6453 goto out_register_late_subsys;
c127ea2c 6454
8ed67789 6455 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
cdb18761 6456 if (ret)
d189634e 6457 goto out_register_late_subsys;
8ed67789 6458
8d0b94af
MKL		 6459 for_each_possible_cpu(cpu) {
6460 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6461
6462 INIT_LIST_HEAD(&ul->head);
6463 spin_lock_init(&ul->lock);
6464 }
6465
433d49c3
DL		 6466out:
6467 return ret;
6468
d189634e 6469out_register_late_subsys:
16feebcf 6470 rtnl_unregister_all(PF_INET6);
d189634e 6471 unregister_pernet_subsys(&ip6_route_net_late_ops);
433d49c3 6472fib6_rules_init:
433d49c3
DL		 6473 fib6_rules_cleanup();
6474xfrm6_init:
433d49c3 6475 xfrm6_fini();
2a0c451a
TG		 6476out_fib6_init:
6477 fib6_gc_cleanup();
8ed67789
DL		 6478out_register_subsys:
6479 unregister_pernet_subsys(&ip6_route_net_ops);
7e52b33b
DM		 6480out_register_inetpeer:
6481 unregister_pernet_subsys(&ipv6_inetpeer_ops);
fc66f95c
ED		 6482out_dst_entries:
6483 dst_entries_destroy(&ip6_dst_blackhole_ops);
433d49c3 6484out_kmem_cache:
f2fc6a54 6485 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
433d49c3 6486 goto out;
1da177e4
LT		 6487}
6488
6489void ip6_route_cleanup(void)
6490{
8ed67789 6491 unregister_netdevice_notifier(&ip6_route_dev_notifier);
d189634e 6492 unregister_pernet_subsys(&ip6_route_net_late_ops);
101367c2 6493 fib6_rules_cleanup();
1da177e4 6494 xfrm6_fini();
1da177e4 6495 fib6_gc_cleanup();
c3426b47 6496 unregister_pernet_subsys(&ipv6_inetpeer_ops);
8ed67789 6497 unregister_pernet_subsys(&ip6_route_net_ops);
41bb78b4 6498 dst_entries_destroy(&ip6_dst_blackhole_ops);
f2fc6a54 6499 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
1da177e4 6500}
Linux 6.x block layer and io_uring tree(s)