projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ipv4: fix suspicious RCU usage in fib_dump_info_fnhe()
[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) {
8d0b94af
MKL		 179 rt->dst.dev = loopback_dev;
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
221 return ip6_neigh_lookup(&rt->rt6i_gateway, dst->dev, skb, daddr);
f83c7790
DM		 222}
223
63fca65d
JA		 224static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
225{
226 struct net_device *dev = dst->dev;
227 struct rt6_info *rt = (struct rt6_info *)dst;
228
f8a1b43b 229 daddr = choose_neigh_daddr(&rt->rt6i_gateway, NULL, daddr);
63fca65d
JA		 230 if (!daddr)
231 return;
232 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
233 return;
234 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
235 return;
236 __ipv6_confirm_neigh(dev, daddr);
237}
238
9a7ec3a9 239static struct dst_ops ip6_dst_ops_template = {
1da177e4 240 .family = AF_INET6,
1da177e4
LT		 241 .gc = ip6_dst_gc,
242 .gc_thresh = 1024,
243 .check = ip6_dst_check,
0dbaee3b 244 .default_advmss = ip6_default_advmss,
ebb762f2 245 .mtu = ip6_mtu,
d4ead6b3 246 .cow_metrics = dst_cow_metrics_generic,
1da177e4
LT		 247 .destroy = ip6_dst_destroy,
248 .ifdown = ip6_dst_ifdown,
249 .negative_advice = ip6_negative_advice,
250 .link_failure = ip6_link_failure,
251 .update_pmtu = ip6_rt_update_pmtu,
6e157b6a 252 .redirect = rt6_do_redirect,
9f8955cc 253 .local_out = __ip6_local_out,
f8a1b43b 254 .neigh_lookup = ip6_dst_neigh_lookup,
63fca65d 255 .confirm_neigh = ip6_confirm_neigh,
1da177e4
LT		 256};
257
ebb762f2 258static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
ec831ea7 259{
618f9bc7
SK		 260 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
261
262 return mtu ? : dst->dev->mtu;
ec831ea7
RD		 263}
264
6700c270
DM		 265static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
266 struct sk_buff *skb, u32 mtu)
14e50e57
DM		 267{
268}
269
6700c270
DM		 270static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
271 struct sk_buff *skb)
b587ee3b
DM		 272{
273}
274
14e50e57
DM		 275static struct dst_ops ip6_dst_blackhole_ops = {
276 .family = AF_INET6,
14e50e57
DM		 277 .destroy = ip6_dst_destroy,
278 .check = ip6_dst_check,
ebb762f2 279 .mtu = ip6_blackhole_mtu,
214f45c9 280 .default_advmss = ip6_default_advmss,
14e50e57 281 .update_pmtu = ip6_rt_blackhole_update_pmtu,
b587ee3b 282 .redirect = ip6_rt_blackhole_redirect,
0a1f5962 283 .cow_metrics = dst_cow_metrics_generic,
f8a1b43b 284 .neigh_lookup = ip6_dst_neigh_lookup,
14e50e57
DM		 285};
286
62fa8a84 287static const u32 ip6_template_metrics[RTAX_MAX] = {
14edd87d 288 [RTAX_HOPLIMIT - 1] = 0,
62fa8a84
DM		 289};
290
8d1c802b 291static const struct fib6_info fib6_null_entry_template = {
93c2fb25
DA		 292 .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
293 .fib6_protocol = RTPROT_KERNEL,
294 .fib6_metric = ~(u32)0,
f05713e0 295 .fib6_ref = REFCOUNT_INIT(1),
421842ed
DA		 296 .fib6_type = RTN_UNREACHABLE,
297 .fib6_metrics = (struct dst_metrics *)&dst_default_metrics,
298};
299
fb0af4c7 300static const struct rt6_info ip6_null_entry_template = {
d8d1f30b
CG		 301 .dst = {
302 .__refcnt = ATOMIC_INIT(1),
303 .__use = 1,
2c20cbd7 304 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 305 .error = -ENETUNREACH,
d8d1f30b
CG		 306 .input = ip6_pkt_discard,
307 .output = ip6_pkt_discard_out,
1da177e4
LT		 308 },
309 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
1da177e4
LT		 310};
311
101367c2
TG		 312#ifdef CONFIG_IPV6_MULTIPLE_TABLES
313
fb0af4c7 314static const struct rt6_info ip6_prohibit_entry_template = {
d8d1f30b
CG		 315 .dst = {
316 .__refcnt = ATOMIC_INIT(1),
317 .__use = 1,
2c20cbd7 318 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 319 .error = -EACCES,
d8d1f30b
CG		 320 .input = ip6_pkt_prohibit,
321 .output = ip6_pkt_prohibit_out,
101367c2
TG		 322 },
323 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 324};
325
fb0af4c7 326static const struct rt6_info ip6_blk_hole_entry_template = {
d8d1f30b
CG		 327 .dst = {
328 .__refcnt = ATOMIC_INIT(1),
329 .__use = 1,
2c20cbd7 330 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 331 .error = -EINVAL,
d8d1f30b 332 .input = dst_discard,
ede2059d 333 .output = dst_discard_out,
101367c2
TG		 334 },
335 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 336};
337
338#endif
339
ebfa45f0
MKL		 340static void rt6_info_init(struct rt6_info *rt)
341{
342 struct dst_entry *dst = &rt->dst;
343
344 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
ebfa45f0
MKL		 345 INIT_LIST_HEAD(&rt->rt6i_uncached);
346}
347
1da177e4 348/* allocate dst with ip6_dst_ops */
93531c67
DA		 349struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
350 int flags)
1da177e4 351{
97bab73f 352 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
b2a9c0ed 353 1, DST_OBSOLETE_FORCE_CHK, flags);
cf911662 354
81eb8447 355 if (rt) {
ebfa45f0 356 rt6_info_init(rt);
81eb8447
WW		 357 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
358 }
8104891b 359
cf911662 360 return rt;
1da177e4 361}
9ab179d8 362EXPORT_SYMBOL(ip6_dst_alloc);
d52d3997 363
1da177e4
LT		 364static void ip6_dst_destroy(struct dst_entry *dst)
365{
366 struct rt6_info *rt = (struct rt6_info *)dst;
a68886a6 367 struct fib6_info *from;
8d0b94af 368 struct inet6_dev *idev;
1da177e4 369
1620a336 370 ip_dst_metrics_put(dst);
8d0b94af
MKL		 371 rt6_uncached_list_del(rt);
372
373 idev = rt->rt6i_idev;
38308473 374 if (idev) {
1da177e4
LT		 375 rt->rt6i_idev = NULL;
376 in6_dev_put(idev);
1ab1457c 377 }
1716a961 378
0e233874 379 from = xchg((__force struct fib6_info **)&rt->from, NULL);
93531c67 380 fib6_info_release(from);
b3419363
DM		 381}
382
1da177e4
LT		 383static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
384 int how)
385{
386 struct rt6_info *rt = (struct rt6_info *)dst;
387 struct inet6_dev *idev = rt->rt6i_idev;
5a3e55d6 388 struct net_device *loopback_dev =
c346dca1 389 dev_net(dev)->loopback_dev;
1da177e4 390
e5645f51
WW		 391 if (idev && idev->dev != loopback_dev) {
392 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
393 if (loopback_idev) {
394 rt->rt6i_idev = loopback_idev;
395 in6_dev_put(idev);
97cac082 396 }
1da177e4
LT		 397 }
398}
399
5973fb1e
MKL		 400static bool __rt6_check_expired(const struct rt6_info *rt)
401{
402 if (rt->rt6i_flags & RTF_EXPIRES)
403 return time_after(jiffies, rt->dst.expires);
404 else
405 return false;
406}
407
a50feda5 408static bool rt6_check_expired(const struct rt6_info *rt)
1da177e4 409{
a68886a6
DA		 410 struct fib6_info *from;
411
412 from = rcu_dereference(rt->from);
413
1716a961
G		 414 if (rt->rt6i_flags & RTF_EXPIRES) {
415 if (time_after(jiffies, rt->dst.expires))
a50feda5 416 return true;
a68886a6 417 } else if (from) {
1e2ea8ad 418 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
a68886a6 419 fib6_check_expired(from);
1716a961 420 }
a50feda5 421 return false;
1da177e4
LT		 422}
423
b1d40991
DA		 424void fib6_select_path(const struct net *net, struct fib6_result *res,
425 struct flowi6 *fl6, int oif, bool have_oif_match,
426 const struct sk_buff *skb, int strict)
51ebd318 427{
8d1c802b 428 struct fib6_info *sibling, *next_sibling;
b1d40991
DA		 429 struct fib6_info *match = res->f6i;
430
f88d8ea6 431 if ((!match->fib6_nsiblings && !match->nh) || have_oif_match)
b1d40991 432 goto out;
51ebd318 433
b673d6cc
JS		 434 /* We might have already computed the hash for ICMPv6 errors. In such
435 * case it will always be non-zero. Otherwise now is the time to do it.
436 */
f88d8ea6
DA		 437 if (!fl6->mp_hash &&
438 (!match->nh || nexthop_is_multipath(match->nh)))
b4bac172 439 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
b673d6cc 440
f88d8ea6
DA		 441 if (unlikely(match->nh)) {
442 nexthop_path_fib6_result(res, fl6->mp_hash);
443 return;
444 }
445
1cf844c7 446 if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
b1d40991 447 goto out;
3d709f69 448
93c2fb25
DA		 449 list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
450 fib6_siblings) {
1cf844c7 451 const struct fib6_nh *nh = sibling->fib6_nh;
5e670d84
DA		 452 int nh_upper_bound;
453
702cea56 454 nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
5e670d84 455 if (fl6->mp_hash > nh_upper_bound)
3d709f69 456 continue;
702cea56 457 if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
3d709f69
IS		 458 break;
459 match = sibling;
460 break;
461 }
462
b1d40991
DA		 463out:
464 res->f6i = match;
1cf844c7 465 res->nh = match->fib6_nh;
51ebd318
ND		 466}
467
1da177e4 468/*
66f5d6ce 469 * Route lookup. rcu_read_lock() should be held.
1da177e4
LT		 470 */
471
0c59d006
DA		 472static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
473 const struct in6_addr *saddr, int oif, int flags)
474{
475 const struct net_device *dev;
476
477 if (nh->fib_nh_flags & RTNH_F_DEAD)
478 return false;
479
480 dev = nh->fib_nh_dev;
481 if (oif) {
482 if (dev->ifindex == oif)
483 return true;
484 } else {
485 if (ipv6_chk_addr(net, saddr, dev,
486 flags & RT6_LOOKUP_F_IFACE))
487 return true;
488 }
489
490 return false;
491}
492
962b6803
DA		 493struct fib6_nh_dm_arg {
494 struct net *net;
495 const struct in6_addr *saddr;
496 int oif;
497 int flags;
498 struct fib6_nh *nh;
499};
500
501static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
502{
503 struct fib6_nh_dm_arg *arg = _arg;
504
505 arg->nh = nh;
506 return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
507 arg->flags);
508}
509
510/* returns fib6_nh from nexthop or NULL */
511static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
512 struct fib6_result *res,
513 const struct in6_addr *saddr,
514 int oif, int flags)
515{
516 struct fib6_nh_dm_arg arg = {
517 .net = net,
518 .saddr = saddr,
519 .oif = oif,
520 .flags = flags,
521 };
522
523 if (nexthop_is_blackhole(nh))
524 return NULL;
525
526 if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
527 return arg.nh;
528
529 return NULL;
530}
531
75ef7389
DA		 532static void rt6_device_match(struct net *net, struct fib6_result *res,
533 const struct in6_addr *saddr, int oif, int flags)
1da177e4 534{
75ef7389
DA		 535 struct fib6_info *f6i = res->f6i;
536 struct fib6_info *spf6i;
537 struct fib6_nh *nh;
1da177e4 538
75ef7389 539 if (!oif && ipv6_addr_any(saddr)) {
f88d8ea6
DA		 540 if (unlikely(f6i->nh)) {
541 nh = nexthop_fib6_nh(f6i->nh);
542 if (nexthop_is_blackhole(f6i->nh))
543 goto out_blackhole;
544 } else {
545 nh = f6i->fib6_nh;
546 }
7d21fec9
DA		 547 if (!(nh->fib_nh_flags & RTNH_F_DEAD))
548 goto out;
75ef7389 549 }
dd3abc4e 550
75ef7389 551 for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
962b6803
DA		 552 bool matched = false;
553
554 if (unlikely(spf6i->nh)) {
555 nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
556 oif, flags);
557 if (nh)
558 matched = true;
559 } else {
560 nh = spf6i->fib6_nh;
561 if (__rt6_device_match(net, nh, saddr, oif, flags))
562 matched = true;
563 }
564 if (matched) {
75ef7389 565 res->f6i = spf6i;
7d21fec9 566 goto out;
75ef7389 567 }
dd3abc4e 568 }
1da177e4 569
75ef7389
DA		 570 if (oif && flags & RT6_LOOKUP_F_IFACE) {
571 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 572 nh = res->f6i->fib6_nh;
7d21fec9 573 goto out;
75ef7389 574 }
8067bb8c 575
f88d8ea6
DA		 576 if (unlikely(f6i->nh)) {
577 nh = nexthop_fib6_nh(f6i->nh);
578 if (nexthop_is_blackhole(f6i->nh))
579 goto out_blackhole;
580 } else {
581 nh = f6i->fib6_nh;
582 }
583
7d21fec9 584 if (nh->fib_nh_flags & RTNH_F_DEAD) {
75ef7389 585 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 586 nh = res->f6i->fib6_nh;
75ef7389 587 }
7d21fec9
DA		 588out:
589 res->nh = nh;
590 res->fib6_type = res->f6i->fib6_type;
591 res->fib6_flags = res->f6i->fib6_flags;
f88d8ea6
DA		 592 return;
593
594out_blackhole:
595 res->fib6_flags |= RTF_REJECT;
596 res->fib6_type = RTN_BLACKHOLE;
597 res->nh = nh;
1da177e4
LT		 598}
599
27097255 600#ifdef CONFIG_IPV6_ROUTER_PREF
c2f17e82
HFS		 601struct __rt6_probe_work {
602 struct work_struct work;
603 struct in6_addr target;
604 struct net_device *dev;
605};
606
607static void rt6_probe_deferred(struct work_struct *w)
608{
609 struct in6_addr mcaddr;
610 struct __rt6_probe_work *work =
611 container_of(w, struct __rt6_probe_work, work);
612
613 addrconf_addr_solict_mult(&work->target, &mcaddr);
adc176c5 614 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
c2f17e82 615 dev_put(work->dev);
662f5533 616 kfree(work);
c2f17e82
HFS		 617}
618
cc3a86c8 619static void rt6_probe(struct fib6_nh *fib6_nh)
27097255 620{
f547fac6 621 struct __rt6_probe_work *work = NULL;
5e670d84 622 const struct in6_addr *nh_gw;
f2c31e32 623 struct neighbour *neigh;
5e670d84 624 struct net_device *dev;
f547fac6 625 struct inet6_dev *idev;
5e670d84 626
27097255
YH		 627 /*
628 * Okay, this does not seem to be appropriate
629 * for now, however, we need to check if it
630 * is really so; aka Router Reachability Probing.
631 *
632 * Router Reachability Probe MUST be rate-limited
633 * to no more than one per minute.
634 */
cc3a86c8 635 if (fib6_nh->fib_nh_gw_family)
7ff74a59 636 return;
5e670d84 637
cc3a86c8
DA		 638 nh_gw = &fib6_nh->fib_nh_gw6;
639 dev = fib6_nh->fib_nh_dev;
2152caea 640 rcu_read_lock_bh();
f547fac6 641 idev = __in6_dev_get(dev);
5e670d84 642 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
2152caea 643 if (neigh) {
8d6c31bf
MKL		 644 if (neigh->nud_state & NUD_VALID)
645 goto out;
646
2152caea 647 write_lock(&neigh->lock);
990edb42
MKL		 648 if (!(neigh->nud_state & NUD_VALID) &&
649 time_after(jiffies,
dcd1f572 650 neigh->updated + idev->cnf.rtr_probe_interval)) {
990edb42
MKL		 651 work = kmalloc(sizeof(*work), GFP_ATOMIC);
652 if (work)
653 __neigh_set_probe_once(neigh);
c2f17e82 654 }
2152caea 655 write_unlock(&neigh->lock);
cc3a86c8 656 } else if (time_after(jiffies, fib6_nh->last_probe +
f547fac6 657 idev->cnf.rtr_probe_interval)) {
990edb42 658 work = kmalloc(sizeof(*work), GFP_ATOMIC);
f2c31e32 659 }
990edb42
MKL		 660
661 if (work) {
cc3a86c8 662 fib6_nh->last_probe = jiffies;
990edb42 663 INIT_WORK(&work->work, rt6_probe_deferred);
5e670d84
DA		 664 work->target = *nh_gw;
665 dev_hold(dev);
666 work->dev = dev;
990edb42
MKL		 667 schedule_work(&work->work);
668 }
669
8d6c31bf 670out:
2152caea 671 rcu_read_unlock_bh();
27097255
YH		 672}
673#else
cc3a86c8 674static inline void rt6_probe(struct fib6_nh *fib6_nh)
27097255 675{
27097255
YH		 676}
677#endif
678
1da177e4 679/*
554cfb7e 680 * Default Router Selection (RFC 2461 6.3.6)
1da177e4 681 */
1ba9a895 682static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
1da177e4 683{
afc154e9 684 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
5e670d84 685 struct neighbour *neigh;
f2c31e32 686
145a3621 687 rcu_read_lock_bh();
1ba9a895
DA		 688 neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
689 &fib6_nh->fib_nh_gw6);
145a3621
YH		 690 if (neigh) {
691 read_lock(&neigh->lock);
554cfb7e 692 if (neigh->nud_state & NUD_VALID)
afc154e9 693 ret = RT6_NUD_SUCCEED;
398bcbeb 694#ifdef CONFIG_IPV6_ROUTER_PREF
a5a81f0b 695 else if (!(neigh->nud_state & NUD_FAILED))
afc154e9 696 ret = RT6_NUD_SUCCEED;
7e980569
JB		 697 else
698 ret = RT6_NUD_FAIL_PROBE;
398bcbeb 699#endif
145a3621 700 read_unlock(&neigh->lock);
afc154e9
HFS		 701 } else {
702 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
7e980569 703 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
a5a81f0b 704 }
145a3621
YH		 705 rcu_read_unlock_bh();
706
a5a81f0b 707 return ret;
1da177e4
LT		 708}
709
702cea56
DA		 710static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
711 int strict)
1da177e4 712{
6e1809a5
DA		 713 int m = 0;
714
715 if (!oif || nh->fib_nh_dev->ifindex == oif)
716 m = 2;
1ab1457c 717
77d16f45 718 if (!m && (strict & RT6_LOOKUP_F_IFACE))
afc154e9 719 return RT6_NUD_FAIL_HARD;
ebacaaa0 720#ifdef CONFIG_IPV6_ROUTER_PREF
702cea56 721 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
ebacaaa0 722#endif
1ba9a895 723 if ((strict & RT6_LOOKUP_F_REACHABLE) &&
702cea56 724 !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
1ba9a895 725 int n = rt6_check_neigh(nh);
afc154e9
HFS		 726 if (n < 0)
727 return n;
728 }
554cfb7e
YH		 729 return m;
730}
731
28679ed1
DA		 732static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
733 int oif, int strict, int *mpri, bool *do_rr)
554cfb7e 734{
afc154e9 735 bool match_do_rr = false;
28679ed1
DA		 736 bool rc = false;
737 int m;
35103d11 738
28679ed1 739 if (nh->fib_nh_flags & RTNH_F_DEAD)
8067bb8c
IS		 740 goto out;
741
28679ed1
DA		 742 if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
743 nh->fib_nh_flags & RTNH_F_LINKDOWN &&
d5d32e4b 744 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
35103d11 745 goto out;
f11e6659 746
28679ed1 747 m = rt6_score_route(nh, fib6_flags, oif, strict);
7e980569 748 if (m == RT6_NUD_FAIL_DO_RR) {
afc154e9
HFS		 749 match_do_rr = true;
750 m = 0; /* lowest valid score */
7e980569 751 } else if (m == RT6_NUD_FAIL_HARD) {
f11e6659 752 goto out;
afc154e9
HFS		 753 }
754
755 if (strict & RT6_LOOKUP_F_REACHABLE)
28679ed1 756 rt6_probe(nh);
f11e6659 757
7e980569 758 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
f11e6659 759 if (m > *mpri) {
afc154e9 760 *do_rr = match_do_rr;
f11e6659 761 *mpri = m;
28679ed1 762 rc = true;
f11e6659 763 }
f11e6659 764out:
28679ed1 765 return rc;
f11e6659
DM		 766}
767
17a5984e
DA		 768struct fib6_nh_frl_arg {
769 u32 flags;
770 int oif;
771 int strict;
772 int *mpri;
773 bool *do_rr;
774 struct fib6_nh *nh;
775};
776
777static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
778{
779 struct fib6_nh_frl_arg *arg = _arg;
780
781 arg->nh = nh;
782 return find_match(nh, arg->flags, arg->oif, arg->strict,
783 arg->mpri, arg->do_rr);
784}
785
b7bc4b6a 786static void __find_rr_leaf(struct fib6_info *f6i_start,
30c15f03 787 struct fib6_info *nomatch, u32 metric,
b7bc4b6a 788 struct fib6_result *res, struct fib6_info **cont,
30c15f03 789 int oif, int strict, bool *do_rr, int *mpri)
f11e6659 790{
b7bc4b6a 791 struct fib6_info *f6i;
1da177e4 792
b7bc4b6a
DA		 793 for (f6i = f6i_start;
794 f6i && f6i != nomatch;
795 f6i = rcu_dereference(f6i->fib6_next)) {
17a5984e 796 bool matched = false;
30c15f03
DA		 797 struct fib6_nh *nh;
798
b7bc4b6a
DA		 799 if (cont && f6i->fib6_metric != metric) {
800 *cont = f6i;
30c15f03 801 return;
9fbdcfaf
SK		 802 }
803
b7bc4b6a 804 if (fib6_check_expired(f6i))
28679ed1
DA		 805 continue;
806
17a5984e
DA		 807 if (unlikely(f6i->nh)) {
808 struct fib6_nh_frl_arg arg = {
809 .flags = f6i->fib6_flags,
810 .oif = oif,
811 .strict = strict,
812 .mpri = mpri,
813 .do_rr = do_rr
814 };
815
816 if (nexthop_is_blackhole(f6i->nh)) {
817 res->fib6_flags = RTF_REJECT;
818 res->fib6_type = RTN_BLACKHOLE;
819 res->f6i = f6i;
820 res->nh = nexthop_fib6_nh(f6i->nh);
821 return;
822 }
823 if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
824 &arg)) {
825 matched = true;
826 nh = arg.nh;
827 }
828 } else {
829 nh = f6i->fib6_nh;
830 if (find_match(nh, f6i->fib6_flags, oif, strict,
831 mpri, do_rr))
832 matched = true;
833 }
834 if (matched) {
b7bc4b6a
DA		 835 res->f6i = f6i;
836 res->nh = nh;
7d21fec9
DA		 837 res->fib6_flags = f6i->fib6_flags;
838 res->fib6_type = f6i->fib6_type;
b7bc4b6a 839 }
9fbdcfaf 840 }
30c15f03 841}
9fbdcfaf 842
b7bc4b6a
DA		 843static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
844 struct fib6_info *rr_head, int oif, int strict,
845 bool *do_rr, struct fib6_result *res)
30c15f03 846{
b7bc4b6a
DA		 847 u32 metric = rr_head->fib6_metric;
848 struct fib6_info *cont = NULL;
30c15f03 849 int mpri = -1;
9fbdcfaf 850
b7bc4b6a 851 __find_rr_leaf(rr_head, NULL, metric, res, &cont,
30c15f03 852 oif, strict, do_rr, &mpri);
28679ed1 853
b7bc4b6a 854 __find_rr_leaf(leaf, rr_head, metric, res, &cont,
30c15f03 855 oif, strict, do_rr, &mpri);
9fbdcfaf 856
b7bc4b6a
DA		 857 if (res->f6i || !cont)
858 return;
9fbdcfaf 859
b7bc4b6a 860 __find_rr_leaf(cont, NULL, metric, res, NULL,
30c15f03 861 oif, strict, do_rr, &mpri);
f11e6659 862}
1da177e4 863
b7bc4b6a
DA		 864static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
865 struct fib6_result *res, int strict)
f11e6659 866{
8d1c802b 867 struct fib6_info *leaf = rcu_dereference(fn->leaf);
b7bc4b6a 868 struct fib6_info *rt0;
afc154e9 869 bool do_rr = false;
17ecf590 870 int key_plen;
1da177e4 871
b7bc4b6a
DA		 872 /* make sure this function or its helpers sets f6i */
873 res->f6i = NULL;
874
421842ed 875 if (!leaf || leaf == net->ipv6.fib6_null_entry)
b7bc4b6a 876 goto out;
8d1040e8 877
66f5d6ce 878 rt0 = rcu_dereference(fn->rr_ptr);
f11e6659 879 if (!rt0)
66f5d6ce 880 rt0 = leaf;
1da177e4 881
17ecf590
WW		 882 /* Double check to make sure fn is not an intermediate node
883 * and fn->leaf does not points to its child's leaf
884 * (This might happen if all routes under fn are deleted from
885 * the tree and fib6_repair_tree() is called on the node.)
886 */
93c2fb25 887 key_plen = rt0->fib6_dst.plen;
17ecf590 888#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 889 if (rt0->fib6_src.plen)
890 key_plen = rt0->fib6_src.plen;
17ecf590
WW		 891#endif
892 if (fn->fn_bit != key_plen)
b7bc4b6a 893 goto out;
1da177e4 894
b7bc4b6a 895 find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
afc154e9 896 if (do_rr) {
8fb11a9a 897 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
f11e6659 898
554cfb7e 899 /* no entries matched; do round-robin */
93c2fb25 900 if (!next || next->fib6_metric != rt0->fib6_metric)
8d1040e8 901 next = leaf;
f11e6659 902
66f5d6ce 903 if (next != rt0) {
93c2fb25 904 spin_lock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 905 /* make sure next is not being deleted from the tree */
93c2fb25 906 if (next->fib6_node)
66f5d6ce 907 rcu_assign_pointer(fn->rr_ptr, next);
93c2fb25 908 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 909 }
1da177e4 910 }
1da177e4 911
b7bc4b6a
DA		 912out:
913 if (!res->f6i) {
914 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 915 res->nh = res->f6i->fib6_nh;
7d21fec9
DA		 916 res->fib6_flags = res->f6i->fib6_flags;
917 res->fib6_type = res->f6i->fib6_type;
b7bc4b6a 918 }
1da177e4
LT		 919}
920
85bd05de 921static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
8b9df265 922{
85bd05de
DA		 923 return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
924 res->nh->fib_nh_gw_family;
8b9df265
MKL		 925}
926
70ceb4f5
YH		 927#ifdef CONFIG_IPV6_ROUTE_INFO
928int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
b71d1d42 929 const struct in6_addr *gwaddr)
70ceb4f5 930{
c346dca1 931 struct net *net = dev_net(dev);
70ceb4f5
YH		 932 struct route_info *rinfo = (struct route_info *) opt;
933 struct in6_addr prefix_buf, *prefix;
934 unsigned int pref;
4bed72e4 935 unsigned long lifetime;
8d1c802b 936 struct fib6_info *rt;
70ceb4f5
YH		 937
938 if (len < sizeof(struct route_info)) {
939 return -EINVAL;
940 }
941
942 /* Sanity check for prefix_len and length */
943 if (rinfo->length > 3) {
944 return -EINVAL;
945 } else if (rinfo->prefix_len > 128) {
946 return -EINVAL;
947 } else if (rinfo->prefix_len > 64) {
948 if (rinfo->length < 2) {
949 return -EINVAL;
950 }
951 } else if (rinfo->prefix_len > 0) {
952 if (rinfo->length < 1) {
953 return -EINVAL;
954 }
955 }
956
957 pref = rinfo->route_pref;
958 if (pref == ICMPV6_ROUTER_PREF_INVALID)
3933fc95 959 return -EINVAL;
70ceb4f5 960
4bed72e4 961 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
70ceb4f5
YH		 962
963 if (rinfo->length == 3)
964 prefix = (struct in6_addr *)rinfo->prefix;
965 else {
966 /* this function is safe */
967 ipv6_addr_prefix(&prefix_buf,
968 (struct in6_addr *)rinfo->prefix,
969 rinfo->prefix_len);
970 prefix = &prefix_buf;
971 }
972
f104a567 973 if (rinfo->prefix_len == 0)
afb1d4b5 974 rt = rt6_get_dflt_router(net, gwaddr, dev);
f104a567
DJ		 975 else
976 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
830218c1 977 gwaddr, dev);
70ceb4f5
YH		 978
979 if (rt && !lifetime) {
afb1d4b5 980 ip6_del_rt(net, rt);
70ceb4f5
YH		 981 rt = NULL;
982 }
983
984 if (!rt && lifetime)
830218c1
DA		 985 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
986 dev, pref);
70ceb4f5 987 else if (rt)
93c2fb25
DA		 988 rt->fib6_flags = RTF_ROUTEINFO |
989 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
70ceb4f5
YH		 990
991 if (rt) {
1716a961 992 if (!addrconf_finite_timeout(lifetime))
14895687 993 fib6_clean_expires(rt);
1716a961 994 else
14895687 995 fib6_set_expires(rt, jiffies + HZ * lifetime);
1716a961 996
93531c67 997 fib6_info_release(rt);
70ceb4f5
YH		 998 }
999 return 0;
1000}
1001#endif
1002
ae90d867
DA		 1003/*
1004 * Misc support functions
1005 */
1006
1007/* called with rcu_lock held */
0d161581 1008static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
ae90d867 1009{
0d161581 1010 struct net_device *dev = res->nh->fib_nh_dev;
ae90d867 1011
7d21fec9 1012 if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
ae90d867
DA		 1013 /* for copies of local routes, dst->dev needs to be the
1014 * device if it is a master device, the master device if
1015 * device is enslaved, and the loopback as the default
1016 */
1017 if (netif_is_l3_slave(dev) &&
7d21fec9 1018 !rt6_need_strict(&res->f6i->fib6_dst.addr))
ae90d867
DA		 1019 dev = l3mdev_master_dev_rcu(dev);
1020 else if (!netif_is_l3_master(dev))
1021 dev = dev_net(dev)->loopback_dev;
1022 /* last case is netif_is_l3_master(dev) is true in which
1023 * case we want dev returned to be dev
1024 */
1025 }
1026
1027 return dev;
1028}
1029
6edb3c96
DA		 1030static const int fib6_prop[RTN_MAX + 1] = {
1031 [RTN_UNSPEC] = 0,
1032 [RTN_UNICAST] = 0,
1033 [RTN_LOCAL] = 0,
1034 [RTN_BROADCAST] = 0,
1035 [RTN_ANYCAST] = 0,
1036 [RTN_MULTICAST] = 0,
1037 [RTN_BLACKHOLE] = -EINVAL,
1038 [RTN_UNREACHABLE] = -EHOSTUNREACH,
1039 [RTN_PROHIBIT] = -EACCES,
1040 [RTN_THROW] = -EAGAIN,
1041 [RTN_NAT] = -EINVAL,
1042 [RTN_XRESOLVE] = -EINVAL,
1043};
1044
1045static int ip6_rt_type_to_error(u8 fib6_type)
1046{
1047 return fib6_prop[fib6_type];
1048}
1049
8d1c802b 1050static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
3b6761d1
DA		 1051{
1052 unsigned short flags = 0;
1053
1054 if (rt->dst_nocount)
1055 flags |= DST_NOCOUNT;
1056 if (rt->dst_nopolicy)
1057 flags |= DST_NOPOLICY;
1058 if (rt->dst_host)
1059 flags |= DST_HOST;
1060
1061 return flags;
1062}
1063
7d21fec9 1064static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
6edb3c96 1065{
7d21fec9 1066 rt->dst.error = ip6_rt_type_to_error(fib6_type);
6edb3c96 1067
7d21fec9 1068 switch (fib6_type) {
6edb3c96
DA		 1069 case RTN_BLACKHOLE:
1070 rt->dst.output = dst_discard_out;
1071 rt->dst.input = dst_discard;
1072 break;
1073 case RTN_PROHIBIT:
1074 rt->dst.output = ip6_pkt_prohibit_out;
1075 rt->dst.input = ip6_pkt_prohibit;
1076 break;
1077 case RTN_THROW:
1078 case RTN_UNREACHABLE:
1079 default:
1080 rt->dst.output = ip6_pkt_discard_out;
1081 rt->dst.input = ip6_pkt_discard;
1082 break;
1083 }
1084}
1085
0d161581 1086static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
6edb3c96 1087{
7d21fec9 1088 struct fib6_info *f6i = res->f6i;
0d161581 1089
7d21fec9
DA		 1090 if (res->fib6_flags & RTF_REJECT) {
1091 ip6_rt_init_dst_reject(rt, res->fib6_type);
6edb3c96
DA		 1092 return;
1093 }
1094
1095 rt->dst.error = 0;
1096 rt->dst.output = ip6_output;
1097
7d21fec9 1098 if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
6edb3c96 1099 rt->dst.input = ip6_input;
7d21fec9 1100 } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
6edb3c96
DA		 1101 rt->dst.input = ip6_mc_input;
1102 } else {
1103 rt->dst.input = ip6_forward;
1104 }
1105
0d161581
DA		 1106 if (res->nh->fib_nh_lws) {
1107 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
6edb3c96
DA		 1108 lwtunnel_set_redirect(&rt->dst);
1109 }
1110
1111 rt->dst.lastuse = jiffies;
1112}
1113
e873e4b9 1114/* Caller must already hold reference to @from */
8d1c802b 1115static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
ae90d867 1116{
ae90d867 1117 rt->rt6i_flags &= ~RTF_EXPIRES;
a68886a6 1118 rcu_assign_pointer(rt->from, from);
e1255ed4 1119 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
ae90d867
DA		 1120}
1121
0d161581
DA		 1122/* Caller must already hold reference to f6i in result */
1123static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
ae90d867 1124{
0d161581
DA		 1125 const struct fib6_nh *nh = res->nh;
1126 const struct net_device *dev = nh->fib_nh_dev;
1127 struct fib6_info *f6i = res->f6i;
dcd1f572 1128
0d161581 1129 ip6_rt_init_dst(rt, res);
6edb3c96 1130
0d161581 1131 rt->rt6i_dst = f6i->fib6_dst;
dcd1f572 1132 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
7d21fec9 1133 rt->rt6i_flags = res->fib6_flags;
0d161581
DA		 1134 if (nh->fib_nh_gw_family) {
1135 rt->rt6i_gateway = nh->fib_nh_gw6;
2b2450ca
DA		 1136 rt->rt6i_flags |= RTF_GATEWAY;
1137 }
0d161581 1138 rt6_set_from(rt, f6i);
ae90d867 1139#ifdef CONFIG_IPV6_SUBTREES
0d161581 1140 rt->rt6i_src = f6i->fib6_src;
ae90d867 1141#endif
ae90d867
DA		 1142}
1143
a3c00e46
MKL		 1144static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1145 struct in6_addr *saddr)
1146{
66f5d6ce 1147 struct fib6_node *pn, *sn;
a3c00e46
MKL		 1148 while (1) {
1149 if (fn->fn_flags & RTN_TL_ROOT)
1150 return NULL;
66f5d6ce
WW		 1151 pn = rcu_dereference(fn->parent);
1152 sn = FIB6_SUBTREE(pn);
1153 if (sn && sn != fn)
6454743b 1154 fn = fib6_node_lookup(sn, NULL, saddr);
a3c00e46
MKL		 1155 else
1156 fn = pn;
1157 if (fn->fn_flags & RTN_RTINFO)
1158 return fn;
1159 }
1160}
c71099ac 1161
10585b43 1162static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
d3843fe5
WW		 1163{
1164 struct rt6_info *rt = *prt;
1165
1166 if (dst_hold_safe(&rt->dst))
1167 return true;
10585b43 1168 if (net) {
d3843fe5
WW		 1169 rt = net->ipv6.ip6_null_entry;
1170 dst_hold(&rt->dst);
1171 } else {
1172 rt = NULL;
1173 }
1174 *prt = rt;
1175 return false;
1176}
1177
dec9b0e2 1178/* called with rcu_lock held */
9b6b35ab 1179static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
dec9b0e2 1180{
9b6b35ab
DA		 1181 struct net_device *dev = res->nh->fib_nh_dev;
1182 struct fib6_info *f6i = res->f6i;
1183 unsigned short flags;
dec9b0e2
DA		 1184 struct rt6_info *nrt;
1185
9b6b35ab 1186 if (!fib6_info_hold_safe(f6i))
1c87e79a 1187 goto fallback;
e873e4b9 1188
9b6b35ab 1189 flags = fib6_info_dst_flags(f6i);
93531c67 1190 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1c87e79a 1191 if (!nrt) {
9b6b35ab 1192 fib6_info_release(f6i);
1c87e79a
XL		 1193 goto fallback;
1194 }
dec9b0e2 1195
0d161581 1196 ip6_rt_copy_init(nrt, res);
1c87e79a
XL		 1197 return nrt;
1198
1199fallback:
1200 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1201 dst_hold(&nrt->dst);
dec9b0e2
DA		 1202 return nrt;
1203}
1204
8ed67789
DL		 1205static struct rt6_info *ip6_pol_route_lookup(struct net *net,
1206 struct fib6_table *table,
b75cc8f9
DA		 1207 struct flowi6 *fl6,
1208 const struct sk_buff *skb,
1209 int flags)
1da177e4 1210{
b1d40991 1211 struct fib6_result res = {};
1da177e4 1212 struct fib6_node *fn;
23fb93a4 1213 struct rt6_info *rt;
1da177e4 1214
b6cdbc85
DA		 1215 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1216 flags &= ~RT6_LOOKUP_F_IFACE;
1217
66f5d6ce 1218 rcu_read_lock();
6454743b 1219 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
c71099ac 1220restart:
b1d40991
DA		 1221 res.f6i = rcu_dereference(fn->leaf);
1222 if (!res.f6i)
1223 res.f6i = net->ipv6.fib6_null_entry;
af52a52c 1224 else
75ef7389
DA		 1225 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1226 flags);
af52a52c 1227
b1d40991 1228 if (res.f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 1229 fn = fib6_backtrack(fn, &fl6->saddr);
1230 if (fn)
1231 goto restart;
2b760fcf 1232
af52a52c
DA		 1233 rt = net->ipv6.ip6_null_entry;
1234 dst_hold(&rt->dst);
1235 goto out;
f88d8ea6
DA		 1236 } else if (res.fib6_flags & RTF_REJECT) {
1237 goto do_create;
af52a52c 1238 }
d3843fe5 1239
b1d40991
DA		 1240 fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1241 fl6->flowi6_oif != 0, skb, flags);
1242
2b760fcf 1243 /* Search through exception table */
7e4b5128 1244 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
23fb93a4 1245 if (rt) {
10585b43 1246 if (ip6_hold_safe(net, &rt))
dec9b0e2 1247 dst_use_noref(&rt->dst, jiffies);
23fb93a4 1248 } else {
f88d8ea6 1249do_create:
9b6b35ab 1250 rt = ip6_create_rt_rcu(&res);
dec9b0e2 1251 }
b811580d 1252
af52a52c 1253out:
8ff2e5b2 1254 trace_fib6_table_lookup(net, &res, table, fl6);
af52a52c 1255
66f5d6ce 1256 rcu_read_unlock();
b811580d 1257
c71099ac 1258 return rt;
c71099ac
TG		 1259}
1260
67ba4152 1261struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
b75cc8f9 1262 const struct sk_buff *skb, int flags)
ea6e574e 1263{
b75cc8f9 1264 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
ea6e574e
FW		 1265}
1266EXPORT_SYMBOL_GPL(ip6_route_lookup);
1267
9acd9f3a 1268struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
b75cc8f9
DA		 1269 const struct in6_addr *saddr, int oif,
1270 const struct sk_buff *skb, int strict)
c71099ac 1271{
4c9483b2
DM		 1272 struct flowi6 fl6 = {
1273 .flowi6_oif = oif,
1274 .daddr = *daddr,
c71099ac
TG		 1275 };
1276 struct dst_entry *dst;
77d16f45 1277 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
c71099ac 1278
adaa70bb 1279 if (saddr) {
4c9483b2 1280 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
adaa70bb
TG		 1281 flags |= RT6_LOOKUP_F_HAS_SADDR;
1282 }
1283
b75cc8f9 1284 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
c71099ac
TG		 1285 if (dst->error == 0)
1286 return (struct rt6_info *) dst;
1287
1288 dst_release(dst);
1289
1da177e4
LT		 1290 return NULL;
1291}
7159039a
YH		 1292EXPORT_SYMBOL(rt6_lookup);
1293
c71099ac 1294/* ip6_ins_rt is called with FREE table->tb6_lock.
1cfb71ee
WW		 1295 * It takes new route entry, the addition fails by any reason the
1296 * route is released.
1297 * Caller must hold dst before calling it.
1da177e4
LT		 1298 */
1299
8d1c802b 1300static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
333c4301 1301 struct netlink_ext_ack *extack)
1da177e4
LT		 1302{
1303 int err;
c71099ac 1304 struct fib6_table *table;
1da177e4 1305
93c2fb25 1306 table = rt->fib6_table;
66f5d6ce 1307 spin_lock_bh(&table->tb6_lock);
d4ead6b3 1308 err = fib6_add(&table->tb6_root, rt, info, extack);
66f5d6ce 1309 spin_unlock_bh(&table->tb6_lock);
1da177e4
LT		 1310
1311 return err;
1312}
1313
8d1c802b 1314int ip6_ins_rt(struct net *net, struct fib6_info *rt)
40e22e8f 1315{
afb1d4b5 1316 struct nl_info info = { .nl_net = net, };
e715b6d3 1317
d4ead6b3 1318 return __ip6_ins_rt(rt, &info, NULL);
40e22e8f
TG		 1319}
1320
85bd05de 1321static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
8b9df265
MKL		 1322 const struct in6_addr *daddr,
1323 const struct in6_addr *saddr)
1da177e4 1324{
85bd05de 1325 struct fib6_info *f6i = res->f6i;
4832c30d 1326 struct net_device *dev;
1da177e4
LT		 1327 struct rt6_info *rt;
1328
1329 /*
1330 * Clone the route.
1331 */
1332
85bd05de 1333 if (!fib6_info_hold_safe(f6i))
e873e4b9
WW		 1334 return NULL;
1335
0d161581 1336 dev = ip6_rt_get_dev_rcu(res);
93531c67 1337 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
e873e4b9 1338 if (!rt) {
85bd05de 1339 fib6_info_release(f6i);
83a09abd 1340 return NULL;
e873e4b9 1341 }
83a09abd 1342
0d161581 1343 ip6_rt_copy_init(rt, res);
83a09abd 1344 rt->rt6i_flags |= RTF_CACHE;
83a09abd
MKL		 1345 rt->dst.flags |= DST_HOST;
1346 rt->rt6i_dst.addr = *daddr;
1347 rt->rt6i_dst.plen = 128;
1da177e4 1348
85bd05de
DA		 1349 if (!rt6_is_gw_or_nonexthop(res)) {
1350 if (f6i->fib6_dst.plen != 128 &&
1351 ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
83a09abd 1352 rt->rt6i_flags |= RTF_ANYCAST;
1da177e4 1353#ifdef CONFIG_IPV6_SUBTREES
83a09abd
MKL		 1354 if (rt->rt6i_src.plen && saddr) {
1355 rt->rt6i_src.addr = *saddr;
1356 rt->rt6i_src.plen = 128;
8b9df265 1357 }
83a09abd 1358#endif
95a9a5ba 1359 }
1da177e4 1360
95a9a5ba
YH		 1361 return rt;
1362}
1da177e4 1363
db3fedee 1364static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
d52d3997 1365{
db3fedee
DA		 1366 struct fib6_info *f6i = res->f6i;
1367 unsigned short flags = fib6_info_dst_flags(f6i);
4832c30d 1368 struct net_device *dev;
d52d3997
MKL		 1369 struct rt6_info *pcpu_rt;
1370
db3fedee 1371 if (!fib6_info_hold_safe(f6i))
e873e4b9
WW		 1372 return NULL;
1373
4832c30d 1374 rcu_read_lock();
0d161581 1375 dev = ip6_rt_get_dev_rcu(res);
93531c67 1376 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags);
4832c30d 1377 rcu_read_unlock();
e873e4b9 1378 if (!pcpu_rt) {
db3fedee 1379 fib6_info_release(f6i);
d52d3997 1380 return NULL;
e873e4b9 1381 }
0d161581 1382 ip6_rt_copy_init(pcpu_rt, res);
d52d3997
MKL		 1383 pcpu_rt->rt6i_flags |= RTF_PCPU;
1384 return pcpu_rt;
1385}
1386
66f5d6ce 1387/* It should be called with rcu_read_lock() acquired */
db3fedee 1388static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
d52d3997 1389{
c353071a 1390 struct rt6_info *pcpu_rt;
d52d3997 1391
c353071a 1392 pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
d52d3997 1393
a73e4195
MKL		 1394 return pcpu_rt;
1395}
1396
afb1d4b5 1397static struct rt6_info *rt6_make_pcpu_route(struct net *net,
db3fedee 1398 const struct fib6_result *res)
a73e4195
MKL		 1399{
1400 struct rt6_info *pcpu_rt, *prev, **p;
d52d3997 1401
db3fedee 1402 pcpu_rt = ip6_rt_pcpu_alloc(res);
0e09edcc
WW		 1403 if (!pcpu_rt)
1404 return NULL;
d52d3997 1405
f40b6ae2 1406 p = this_cpu_ptr(res->nh->rt6i_pcpu);
a94b9367 1407 prev = cmpxchg(p, NULL, pcpu_rt);
951f788a 1408 BUG_ON(prev);
a94b9367 1409
61fb0d01
ED		 1410 if (res->f6i->fib6_destroying) {
1411 struct fib6_info *from;
1412
1413 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1414 fib6_info_release(from);
1415 }
1416
d52d3997
MKL		 1417 return pcpu_rt;
1418}
1419
35732d01
WW		 1420/* exception hash table implementation
1421 */
1422static DEFINE_SPINLOCK(rt6_exception_lock);
1423
1424/* Remove rt6_ex from hash table and free the memory
1425 * Caller must hold rt6_exception_lock
1426 */
1427static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1428 struct rt6_exception *rt6_ex)
1429{
f5b51fe8 1430 struct fib6_info *from;
b2427e67 1431 struct net *net;
81eb8447 1432
35732d01
WW		 1433 if (!bucket || !rt6_ex)
1434 return;
b2427e67
CIK		 1435
1436 net = dev_net(rt6_ex->rt6i->dst.dev);
f5b51fe8
PA		 1437 net->ipv6.rt6_stats->fib_rt_cache--;
1438
1439 /* purge completely the exception to allow releasing the held resources:
1440 * some [sk] cache may keep the dst around for unlimited time
1441 */
0e233874 1442 from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
f5b51fe8
PA		 1443 fib6_info_release(from);
1444 dst_dev_put(&rt6_ex->rt6i->dst);
1445
35732d01 1446 hlist_del_rcu(&rt6_ex->hlist);
77634cc6 1447 dst_release(&rt6_ex->rt6i->dst);
35732d01
WW		 1448 kfree_rcu(rt6_ex, rcu);
1449 WARN_ON_ONCE(!bucket->depth);
1450 bucket->depth--;
1451}
1452
1453/* Remove oldest rt6_ex in bucket and free the memory
1454 * Caller must hold rt6_exception_lock
1455 */
1456static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1457{
1458 struct rt6_exception *rt6_ex, *oldest = NULL;
1459
1460 if (!bucket)
1461 return;
1462
1463 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1464 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1465 oldest = rt6_ex;
1466 }
1467 rt6_remove_exception(bucket, oldest);
1468}
1469
1470static u32 rt6_exception_hash(const struct in6_addr *dst,
1471 const struct in6_addr *src)
1472{
1473 static u32 seed __read_mostly;
1474 u32 val;
1475
1476 net_get_random_once(&seed, sizeof(seed));
1477 val = jhash(dst, sizeof(*dst), seed);
1478
1479#ifdef CONFIG_IPV6_SUBTREES
1480 if (src)
1481 val = jhash(src, sizeof(*src), val);
1482#endif
1483 return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1484}
1485
1486/* Helper function to find the cached rt in the hash table
1487 * and update bucket pointer to point to the bucket for this
1488 * (daddr, saddr) pair
1489 * Caller must hold rt6_exception_lock
1490 */
1491static struct rt6_exception *
1492__rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1493 const struct in6_addr *daddr,
1494 const struct in6_addr *saddr)
1495{
1496 struct rt6_exception *rt6_ex;
1497 u32 hval;
1498
1499 if (!(*bucket) || !daddr)
1500 return NULL;
1501
1502 hval = rt6_exception_hash(daddr, saddr);
1503 *bucket += hval;
1504
1505 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1506 struct rt6_info *rt6 = rt6_ex->rt6i;
1507 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1508
1509#ifdef CONFIG_IPV6_SUBTREES
1510 if (matched && saddr)
1511 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1512#endif
1513 if (matched)
1514 return rt6_ex;
1515 }
1516 return NULL;
1517}
1518
1519/* Helper function to find the cached rt in the hash table
1520 * and update bucket pointer to point to the bucket for this
1521 * (daddr, saddr) pair
1522 * Caller must hold rcu_read_lock()
1523 */
1524static struct rt6_exception *
1525__rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1526 const struct in6_addr *daddr,
1527 const struct in6_addr *saddr)
1528{
1529 struct rt6_exception *rt6_ex;
1530 u32 hval;
1531
1532 WARN_ON_ONCE(!rcu_read_lock_held());
1533
1534 if (!(*bucket) || !daddr)
1535 return NULL;
1536
1537 hval = rt6_exception_hash(daddr, saddr);
1538 *bucket += hval;
1539
1540 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1541 struct rt6_info *rt6 = rt6_ex->rt6i;
1542 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1543
1544#ifdef CONFIG_IPV6_SUBTREES
1545 if (matched && saddr)
1546 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1547#endif
1548 if (matched)
1549 return rt6_ex;
1550 }
1551 return NULL;
1552}
1553
b748f260 1554static unsigned int fib6_mtu(const struct fib6_result *res)
d4ead6b3 1555{
b748f260 1556 const struct fib6_nh *nh = res->nh;
d4ead6b3
DA		 1557 unsigned int mtu;
1558
b748f260
DA		 1559 if (res->f6i->fib6_pmtu) {
1560 mtu = res->f6i->fib6_pmtu;
dcd1f572 1561 } else {
b748f260 1562 struct net_device *dev = nh->fib_nh_dev;
dcd1f572
DA		 1563 struct inet6_dev *idev;
1564
1565 rcu_read_lock();
1566 idev = __in6_dev_get(dev);
1567 mtu = idev->cnf.mtu6;
1568 rcu_read_unlock();
1569 }
1570
d4ead6b3
DA		 1571 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1572
b748f260 1573 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
d4ead6b3
DA		 1574}
1575
cc5c073a
DA		 1576#define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1577
1578/* used when the flushed bit is not relevant, only access to the bucket
1579 * (ie., all bucket users except rt6_insert_exception);
1580 *
1581 * called under rcu lock; sometimes called with rt6_exception_lock held
1582 */
1583static
1584struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1585 spinlock_t *lock)
1586{
1587 struct rt6_exception_bucket *bucket;
1588
1589 if (lock)
1590 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1591 lockdep_is_held(lock));
1592 else
1593 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1594
1595 /* remove bucket flushed bit if set */
1596 if (bucket) {
1597 unsigned long p = (unsigned long)bucket;
1598
1599 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1600 bucket = (struct rt6_exception_bucket *)p;
1601 }
1602
1603 return bucket;
1604}
1605
1606static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1607{
1608 unsigned long p = (unsigned long)bucket;
1609
1610 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1611}
1612
1613/* called with rt6_exception_lock held */
1614static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1615 spinlock_t *lock)
1616{
1617 struct rt6_exception_bucket *bucket;
1618 unsigned long p;
1619
1620 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1621 lockdep_is_held(lock));
1622
1623 p = (unsigned long)bucket;
1624 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1625 bucket = (struct rt6_exception_bucket *)p;
1626 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1627}
1628
35732d01 1629static int rt6_insert_exception(struct rt6_info *nrt,
5012f0a5 1630 const struct fib6_result *res)
35732d01 1631{
5e670d84 1632 struct net *net = dev_net(nrt->dst.dev);
35732d01 1633 struct rt6_exception_bucket *bucket;
cc5c073a 1634 struct fib6_info *f6i = res->f6i;
35732d01
WW		 1635 struct in6_addr *src_key = NULL;
1636 struct rt6_exception *rt6_ex;
cc5c073a 1637 struct fib6_nh *nh = res->nh;
35732d01
WW		 1638 int err = 0;
1639
35732d01
WW		 1640 spin_lock_bh(&rt6_exception_lock);
1641
cc5c073a
DA		 1642 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1643 lockdep_is_held(&rt6_exception_lock));
35732d01
WW		 1644 if (!bucket) {
1645 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1646 GFP_ATOMIC);
1647 if (!bucket) {
1648 err = -ENOMEM;
1649 goto out;
1650 }
cc5c073a
DA		 1651 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1652 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1653 err = -EINVAL;
1654 goto out;
35732d01
WW		 1655 }
1656
1657#ifdef CONFIG_IPV6_SUBTREES
5012f0a5 1658 /* fib6_src.plen != 0 indicates f6i is in subtree
35732d01 1659 * and exception table is indexed by a hash of
5012f0a5 1660 * both fib6_dst and fib6_src.
35732d01 1661 * Otherwise, the exception table is indexed by
5012f0a5 1662 * a hash of only fib6_dst.
35732d01 1663 */
5012f0a5 1664 if (f6i->fib6_src.plen)
35732d01
WW		 1665 src_key = &nrt->rt6i_src.addr;
1666#endif
5012f0a5 1667 /* rt6_mtu_change() might lower mtu on f6i.
f5bbe7ee 1668 * Only insert this exception route if its mtu
5012f0a5 1669 * is less than f6i's mtu value.
f5bbe7ee 1670 */
b748f260 1671 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
f5bbe7ee
WW		 1672 err = -EINVAL;
1673 goto out;
1674 }
60006a48 1675
35732d01
WW		 1676 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1677 src_key);
1678 if (rt6_ex)
1679 rt6_remove_exception(bucket, rt6_ex);
1680
1681 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1682 if (!rt6_ex) {
1683 err = -ENOMEM;
1684 goto out;
1685 }
1686 rt6_ex->rt6i = nrt;
1687 rt6_ex->stamp = jiffies;
35732d01
WW		 1688 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1689 bucket->depth++;
81eb8447 1690 net->ipv6.rt6_stats->fib_rt_cache++;
35732d01
WW		 1691
1692 if (bucket->depth > FIB6_MAX_DEPTH)
1693 rt6_exception_remove_oldest(bucket);
1694
1695out:
1696 spin_unlock_bh(&rt6_exception_lock);
1697
1698 /* Update fn->fn_sernum to invalidate all cached dst */
b886d5f2 1699 if (!err) {
5012f0a5
DA		 1700 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1701 fib6_update_sernum(net, f6i);
1702 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
b886d5f2
PA		 1703 fib6_force_start_gc(net);
1704 }
35732d01
WW		 1705
1706 return err;
1707}
1708
c0b220cf 1709static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
35732d01
WW		 1710{
1711 struct rt6_exception_bucket *bucket;
1712 struct rt6_exception *rt6_ex;
1713 struct hlist_node *tmp;
1714 int i;
1715
1716 spin_lock_bh(&rt6_exception_lock);
35732d01 1717
cc5c073a 1718 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
35732d01
WW		 1719 if (!bucket)
1720 goto out;
1721
cc5c073a
DA		 1722 /* Prevent rt6_insert_exception() to recreate the bucket list */
1723 if (!from)
1724 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1725
35732d01 1726 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
cc5c073a
DA		 1727 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1728 if (!from ||
1729 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1730 rt6_remove_exception(bucket, rt6_ex);
1731 }
1732 WARN_ON_ONCE(!from && bucket->depth);
35732d01
WW		 1733 bucket++;
1734 }
35732d01
WW		 1735out:
1736 spin_unlock_bh(&rt6_exception_lock);
1737}
1738
e659ba31
DA		 1739static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1740{
1741 struct fib6_info *f6i = arg;
1742
1743 fib6_nh_flush_exceptions(nh, f6i);
1744
1745 return 0;
1746}
1747
c0b220cf
DA		 1748void rt6_flush_exceptions(struct fib6_info *f6i)
1749{
e659ba31
DA		 1750 if (f6i->nh)
1751 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1752 f6i);
1753 else
1754 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
c0b220cf
DA		 1755}
1756
35732d01
WW		 1757/* Find cached rt in the hash table inside passed in rt
1758 * Caller has to hold rcu_read_lock()
1759 */
7e4b5128 1760static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
510e2ced
WW		 1761 const struct in6_addr *daddr,
1762 const struct in6_addr *saddr)
35732d01 1763{
510e2ced 1764 const struct in6_addr *src_key = NULL;
35732d01 1765 struct rt6_exception_bucket *bucket;
35732d01 1766 struct rt6_exception *rt6_ex;
7e4b5128 1767 struct rt6_info *ret = NULL;
35732d01 1768
35732d01 1769#ifdef CONFIG_IPV6_SUBTREES
7e4b5128 1770 /* fib6i_src.plen != 0 indicates f6i is in subtree
35732d01 1771 * and exception table is indexed by a hash of
7e4b5128 1772 * both fib6_dst and fib6_src.
510e2ced
WW		 1773 * However, the src addr used to create the hash
1774 * might not be exactly the passed in saddr which
1775 * is a /128 addr from the flow.
1776 * So we need to use f6i->fib6_src to redo lookup
1777 * if the passed in saddr does not find anything.
1778 * (See the logic in ip6_rt_cache_alloc() on how
1779 * rt->rt6i_src is updated.)
35732d01 1780 */
7e4b5128 1781 if (res->f6i->fib6_src.plen)
35732d01 1782 src_key = saddr;
510e2ced 1783find_ex:
35732d01 1784#endif
cc5c073a 1785 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
35732d01
WW		 1786 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1787
1788 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
7e4b5128 1789 ret = rt6_ex->rt6i;
35732d01 1790
510e2ced
WW		 1791#ifdef CONFIG_IPV6_SUBTREES
1792 /* Use fib6_src as src_key and redo lookup */
1793 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1794 src_key = &res->f6i->fib6_src.addr;
1795 goto find_ex;
1796 }
1797#endif
1798
7e4b5128 1799 return ret;
35732d01
WW		 1800}
1801
1802/* Remove the passed in cached rt from the hash table that contains it */
cc5c073a 1803static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1804 const struct rt6_info *rt)
35732d01 1805{
c0b220cf 1806 const struct in6_addr *src_key = NULL;
35732d01 1807 struct rt6_exception_bucket *bucket;
35732d01
WW		 1808 struct rt6_exception *rt6_ex;
1809 int err;
1810
cc5c073a 1811 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
35732d01
WW		 1812 return -ENOENT;
1813
1814 spin_lock_bh(&rt6_exception_lock);
cc5c073a
DA		 1815 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1816
35732d01 1817#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1818 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1819 * and exception table is indexed by a hash of
1820 * both rt6i_dst and rt6i_src.
35732d01
WW		 1821 * Otherwise, the exception table is indexed by
1822 * a hash of only rt6i_dst.
1823 */
c0b220cf 1824 if (plen)
35732d01
WW		 1825 src_key = &rt->rt6i_src.addr;
1826#endif
1827 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1828 &rt->rt6i_dst.addr,
1829 src_key);
1830 if (rt6_ex) {
1831 rt6_remove_exception(bucket, rt6_ex);
1832 err = 0;
1833 } else {
1834 err = -ENOENT;
1835 }
1836
1837 spin_unlock_bh(&rt6_exception_lock);
1838 return err;
1839}
1840
e659ba31
DA		 1841struct fib6_nh_excptn_arg {
1842 struct rt6_info *rt;
1843 int plen;
1844};
1845
1846static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1847{
1848 struct fib6_nh_excptn_arg *arg = _arg;
1849 int err;
1850
1851 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1852 if (err == 0)
1853 return 1;
1854
1855 return 0;
1856}
1857
c0b220cf
DA		 1858static int rt6_remove_exception_rt(struct rt6_info *rt)
1859{
1860 struct fib6_info *from;
1861
1862 from = rcu_dereference(rt->from);
cc5c073a 1863 if (!from || !(rt->rt6i_flags & RTF_CACHE))
c0b220cf
DA		 1864 return -EINVAL;
1865
e659ba31
DA		 1866 if (from->nh) {
1867 struct fib6_nh_excptn_arg arg = {
1868 .rt = rt,
1869 .plen = from->fib6_src.plen
1870 };
1871 int rc;
1872
1873 /* rc = 1 means an entry was found */
1874 rc = nexthop_for_each_fib6_nh(from->nh,
1875 rt6_nh_remove_exception_rt,
1876 &arg);
1877 return rc ? 0 : -ENOENT;
1878 }
1879
1cf844c7 1880 return fib6_nh_remove_exception(from->fib6_nh,
cc5c073a 1881 from->fib6_src.plen, rt);
c0b220cf
DA		 1882}
1883
35732d01
WW		 1884/* Find rt6_ex which contains the passed in rt cache and
1885 * refresh its stamp
1886 */
cc5c073a 1887static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1888 const struct rt6_info *rt)
35732d01 1889{
c0b220cf 1890 const struct in6_addr *src_key = NULL;
35732d01 1891 struct rt6_exception_bucket *bucket;
35732d01 1892 struct rt6_exception *rt6_ex;
193f3685 1893
cc5c073a 1894 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
35732d01 1895#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1896 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1897 * and exception table is indexed by a hash of
1898 * both rt6i_dst and rt6i_src.
35732d01
WW		 1899 * Otherwise, the exception table is indexed by
1900 * a hash of only rt6i_dst.
1901 */
c0b220cf 1902 if (plen)
35732d01
WW		 1903 src_key = &rt->rt6i_src.addr;
1904#endif
cc5c073a 1905 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
35732d01
WW		 1906 if (rt6_ex)
1907 rt6_ex->stamp = jiffies;
c0b220cf
DA		 1908}
1909
e659ba31
DA		 1910struct fib6_nh_match_arg {
1911 const struct net_device *dev;
1912 const struct in6_addr *gw;
1913 struct fib6_nh *match;
1914};
1915
1916/* determine if fib6_nh has given device and gateway */
1917static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1918{
1919 struct fib6_nh_match_arg *arg = _arg;
1920
1921 if (arg->dev != nh->fib_nh_dev ||
1922 (arg->gw && !nh->fib_nh_gw_family) ||
1923 (!arg->gw && nh->fib_nh_gw_family) ||
1924 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1925 return 0;
1926
1927 arg->match = nh;
1928
1929 /* found a match, break the loop */
1930 return 1;
1931}
1932
c0b220cf
DA		 1933static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1934{
1935 struct fib6_info *from;
e659ba31 1936 struct fib6_nh *fib6_nh;
35732d01 1937
c0b220cf
DA		 1938 rcu_read_lock();
1939
1940 from = rcu_dereference(rt->from);
1941 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1942 goto unlock;
1943
e659ba31
DA		 1944 if (from->nh) {
1945 struct fib6_nh_match_arg arg = {
1946 .dev = rt->dst.dev,
1947 .gw = &rt->rt6i_gateway,
1948 };
1949
1950 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1951
1952 if (!arg.match)
1953 return;
1954 fib6_nh = arg.match;
1955 } else {
1956 fib6_nh = from->fib6_nh;
1957 }
1958 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
193f3685 1959unlock:
35732d01
WW		 1960 rcu_read_unlock();
1961}
1962
e9fa1495
SB		 1963static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1964 struct rt6_info *rt, int mtu)
1965{
1966 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1967 * lowest MTU in the path: always allow updating the route PMTU to
1968 * reflect PMTU decreases.
1969 *
1970 * If the new MTU is higher, and the route PMTU is equal to the local
1971 * MTU, this means the old MTU is the lowest in the path, so allow
1972 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1973 * handle this.
1974 */
1975
1976 if (dst_mtu(&rt->dst) >= mtu)
1977 return true;
1978
1979 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
1980 return true;
1981
1982 return false;
1983}
1984
1985static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
cc5c073a 1986 const struct fib6_nh *nh, int mtu)
f5bbe7ee
WW		 1987{
1988 struct rt6_exception_bucket *bucket;
1989 struct rt6_exception *rt6_ex;
1990 int i;
1991
cc5c073a 1992 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
e9fa1495
SB		 1993 if (!bucket)
1994 return;
1995
1996 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1997 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1998 struct rt6_info *entry = rt6_ex->rt6i;
1999
2000 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
d4ead6b3 2001 * route), the metrics of its rt->from have already
e9fa1495
SB		 2002 * been updated.
2003 */
d4ead6b3 2004 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
e9fa1495 2005 rt6_mtu_change_route_allowed(idev, entry, mtu))
d4ead6b3 2006 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
f5bbe7ee 2007 }
e9fa1495 2008 bucket++;
f5bbe7ee
WW		 2009 }
2010}
2011
b16cb459
WW		 2012#define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2013
cc5c073a
DA		 2014static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2015 const struct in6_addr *gateway)
b16cb459
WW		 2016{
2017 struct rt6_exception_bucket *bucket;
2018 struct rt6_exception *rt6_ex;
2019 struct hlist_node *tmp;
2020 int i;
2021
cc5c073a 2022 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
b16cb459
WW		 2023 return;
2024
2025 spin_lock_bh(&rt6_exception_lock);
cc5c073a 2026 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
b16cb459
WW		 2027 if (bucket) {
2028 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2029 hlist_for_each_entry_safe(rt6_ex, tmp,
2030 &bucket->chain, hlist) {
2031 struct rt6_info *entry = rt6_ex->rt6i;
2032
2033 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2034 RTF_CACHE_GATEWAY &&
2035 ipv6_addr_equal(gateway,
2036 &entry->rt6i_gateway)) {
2037 rt6_remove_exception(bucket, rt6_ex);
2038 }
2039 }
2040 bucket++;
2041 }
2042 }
2043
2044 spin_unlock_bh(&rt6_exception_lock);
2045}
2046
c757faa8
WW		 2047static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2048 struct rt6_exception *rt6_ex,
2049 struct fib6_gc_args *gc_args,
2050 unsigned long now)
2051{
2052 struct rt6_info *rt = rt6_ex->rt6i;
2053
1859bac0
PA		 2054 /* we are pruning and obsoleting aged-out and non gateway exceptions
2055 * even if others have still references to them, so that on next
2056 * dst_check() such references can be dropped.
2057 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2058 * expired, independently from their aging, as per RFC 8201 section 4
2059 */
31afeb42
WW		 2060 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2061 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2062 RT6_TRACE("aging clone %p\n", rt);
2063 rt6_remove_exception(bucket, rt6_ex);
2064 return;
2065 }
2066 } else if (time_after(jiffies, rt->dst.expires)) {
2067 RT6_TRACE("purging expired route %p\n", rt);
c757faa8
WW		 2068 rt6_remove_exception(bucket, rt6_ex);
2069 return;
31afeb42
WW		 2070 }
2071
2072 if (rt->rt6i_flags & RTF_GATEWAY) {
c757faa8
WW		 2073 struct neighbour *neigh;
2074 __u8 neigh_flags = 0;
2075
1bfa26ff
ED		 2076 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2077 if (neigh)
c757faa8 2078 neigh_flags = neigh->flags;
1bfa26ff 2079
c757faa8
WW		 2080 if (!(neigh_flags & NTF_ROUTER)) {
2081 RT6_TRACE("purging route %p via non-router but gateway\n",
2082 rt);
2083 rt6_remove_exception(bucket, rt6_ex);
2084 return;
2085 }
2086 }
31afeb42 2087
c757faa8
WW		 2088 gc_args->more++;
2089}
2090
cc5c073a 2091static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
c0b220cf
DA		 2092 struct fib6_gc_args *gc_args,
2093 unsigned long now)
c757faa8
WW		 2094{
2095 struct rt6_exception_bucket *bucket;
2096 struct rt6_exception *rt6_ex;
2097 struct hlist_node *tmp;
2098 int i;
2099
cc5c073a 2100 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
c757faa8
WW		 2101 return;
2102
1bfa26ff
ED		 2103 rcu_read_lock_bh();
2104 spin_lock(&rt6_exception_lock);
cc5c073a 2105 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
c757faa8
WW		 2106 if (bucket) {
2107 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2108 hlist_for_each_entry_safe(rt6_ex, tmp,
2109 &bucket->chain, hlist) {
2110 rt6_age_examine_exception(bucket, rt6_ex,
2111 gc_args, now);
2112 }
2113 bucket++;
2114 }
2115 }
1bfa26ff
ED		 2116 spin_unlock(&rt6_exception_lock);
2117 rcu_read_unlock_bh();
c757faa8
WW		 2118}
2119
e659ba31
DA		 2120struct fib6_nh_age_excptn_arg {
2121 struct fib6_gc_args *gc_args;
2122 unsigned long now;
2123};
2124
2125static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2126{
2127 struct fib6_nh_age_excptn_arg *arg = _arg;
2128
2129 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2130 return 0;
2131}
2132
cc5c073a 2133void rt6_age_exceptions(struct fib6_info *f6i,
c0b220cf
DA		 2134 struct fib6_gc_args *gc_args,
2135 unsigned long now)
2136{
e659ba31
DA		 2137 if (f6i->nh) {
2138 struct fib6_nh_age_excptn_arg arg = {
2139 .gc_args = gc_args,
2140 .now = now
2141 };
2142
2143 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2144 &arg);
2145 } else {
2146 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2147 }
c0b220cf
DA		 2148}
2149
1d053da9 2150/* must be called with rcu lock held */
effda4dd
DA		 2151int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2152 struct flowi6 *fl6, struct fib6_result *res, int strict)
1da177e4 2153{
367efcb9 2154 struct fib6_node *fn, *saved_fn;
1da177e4 2155
6454743b 2156 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
367efcb9 2157 saved_fn = fn;
1da177e4 2158
ca254490
DA		 2159 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2160 oif = 0;
2161
a3c00e46 2162redo_rt6_select:
effda4dd
DA		 2163 rt6_select(net, fn, oif, res, strict);
2164 if (res->f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2165 fn = fib6_backtrack(fn, &fl6->saddr);
2166 if (fn)
2167 goto redo_rt6_select;
367efcb9
MKL		 2168 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2169 /* also consider unreachable route */
2170 strict &= ~RT6_LOOKUP_F_REACHABLE;
2171 fn = saved_fn;
2172 goto redo_rt6_select;
367efcb9 2173 }
a3c00e46
MKL		 2174 }
2175
effda4dd 2176 trace_fib6_table_lookup(net, res, table, fl6);
fb9de91e 2177
effda4dd 2178 return 0;
1d053da9
DA		 2179}
2180
2181struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2182 int oif, struct flowi6 *fl6,
2183 const struct sk_buff *skb, int flags)
2184{
b1d40991 2185 struct fib6_result res = {};
0e09edcc 2186 struct rt6_info *rt = NULL;
1d053da9
DA		 2187 int strict = 0;
2188
0e09edcc
WW		 2189 WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2190 !rcu_read_lock_held());
2191
1d053da9
DA		 2192 strict |= flags & RT6_LOOKUP_F_IFACE;
2193 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2194 if (net->ipv6.devconf_all->forwarding == 0)
2195 strict |= RT6_LOOKUP_F_REACHABLE;
2196
2197 rcu_read_lock();
2198
effda4dd 2199 fib6_table_lookup(net, table, oif, fl6, &res, strict);
0e09edcc
WW		 2200 if (res.f6i == net->ipv6.fib6_null_entry)
2201 goto out;
23fb93a4 2202
b1d40991 2203 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
d83009d4 2204
23fb93a4 2205 /*Search through exception table */
7e4b5128 2206 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
23fb93a4 2207 if (rt) {
0e09edcc 2208 goto out;
3da59bd9 2209 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
b1d40991 2210 !res.nh->fib_nh_gw_family)) {
3da59bd9
MKL		 2211 /* Create a RTF_CACHE clone which will not be
2212 * owned by the fib6 tree. It is for the special case where
2213 * the daddr in the skb during the neighbor look-up is different
2214 * from the fl6->daddr used to look-up route here.
2215 */
0e09edcc 2216 rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
c71099ac 2217
0e09edcc
WW		 2218 if (rt) {
2219 /* 1 refcnt is taken during ip6_rt_cache_alloc().
2220 * As rt6_uncached_list_add() does not consume refcnt,
2221 * this refcnt is always returned to the caller even
2222 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
1cfb71ee 2223 */
0e09edcc 2224 rt6_uncached_list_add(rt);
81eb8447 2225 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
0e09edcc 2226 rcu_read_unlock();
b811580d 2227
0e09edcc
WW		 2228 return rt;
2229 }
d52d3997
MKL		 2230 } else {
2231 /* Get a percpu copy */
951f788a 2232 local_bh_disable();
0e09edcc 2233 rt = rt6_get_pcpu_route(&res);
d52d3997 2234
0e09edcc
WW		 2235 if (!rt)
2236 rt = rt6_make_pcpu_route(net, &res);
93531c67 2237
951f788a 2238 local_bh_enable();
d52d3997 2239 }
0e09edcc
WW		 2240out:
2241 if (!rt)
2242 rt = net->ipv6.ip6_null_entry;
2243 if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2244 ip6_hold_safe(net, &rt);
2245 rcu_read_unlock();
2246
2247 return rt;
1da177e4 2248}
9ff74384 2249EXPORT_SYMBOL_GPL(ip6_pol_route);
1da177e4 2250
b75cc8f9
DA		 2251static struct rt6_info *ip6_pol_route_input(struct net *net,
2252 struct fib6_table *table,
2253 struct flowi6 *fl6,
2254 const struct sk_buff *skb,
2255 int flags)
4acad72d 2256{
b75cc8f9 2257 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
4acad72d
PE		 2258}
2259
d409b847
MB		 2260struct dst_entry *ip6_route_input_lookup(struct net *net,
2261 struct net_device *dev,
b75cc8f9
DA		 2262 struct flowi6 *fl6,
2263 const struct sk_buff *skb,
2264 int flags)
72331bc0
SL		 2265{
2266 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2267 flags |= RT6_LOOKUP_F_IFACE;
2268
b75cc8f9 2269 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
72331bc0 2270}
d409b847 2271EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
72331bc0 2272
23aebdac 2273static void ip6_multipath_l3_keys(const struct sk_buff *skb,
5e5d6fed
RP		 2274 struct flow_keys *keys,
2275 struct flow_keys *flkeys)
23aebdac
JS		 2276{
2277 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2278 const struct ipv6hdr *key_iph = outer_iph;
5e5d6fed 2279 struct flow_keys *_flkeys = flkeys;
23aebdac
JS		 2280 const struct ipv6hdr *inner_iph;
2281 const struct icmp6hdr *icmph;
2282 struct ipv6hdr _inner_iph;
cea67a2d 2283 struct icmp6hdr _icmph;
23aebdac
JS		 2284
2285 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2286 goto out;
2287
cea67a2d
ED		 2288 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2289 sizeof(_icmph), &_icmph);
2290 if (!icmph)
2291 goto out;
2292
23aebdac
JS		 2293 if (icmph->icmp6_type != ICMPV6_DEST_UNREACH &&
2294 icmph->icmp6_type != ICMPV6_PKT_TOOBIG &&
2295 icmph->icmp6_type != ICMPV6_TIME_EXCEED &&
2296 icmph->icmp6_type != ICMPV6_PARAMPROB)
2297 goto out;
2298
2299 inner_iph = skb_header_pointer(skb,
2300 skb_transport_offset(skb) + sizeof(*icmph),
2301 sizeof(_inner_iph), &_inner_iph);
2302 if (!inner_iph)
2303 goto out;
2304
2305 key_iph = inner_iph;
5e5d6fed 2306 _flkeys = NULL;
23aebdac 2307out:
5e5d6fed
RP		 2308 if (_flkeys) {
2309 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2310 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2311 keys->tags.flow_label = _flkeys->tags.flow_label;
2312 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2313 } else {
2314 keys->addrs.v6addrs.src = key_iph->saddr;
2315 keys->addrs.v6addrs.dst = key_iph->daddr;
fa1be7e0 2316 keys->tags.flow_label = ip6_flowlabel(key_iph);
5e5d6fed
RP		 2317 keys->basic.ip_proto = key_iph->nexthdr;
2318 }
23aebdac
JS		 2319}
2320
2321/* if skb is set it will be used and fl6 can be NULL */
b4bac172
DA		 2322u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2323 const struct sk_buff *skb, struct flow_keys *flkeys)
23aebdac
JS		 2324{
2325 struct flow_keys hash_keys;
9a2a537a 2326 u32 mhash;
23aebdac 2327
bbfa047a 2328 switch (ip6_multipath_hash_policy(net)) {
b4bac172
DA		 2329 case 0:
2330 memset(&hash_keys, 0, sizeof(hash_keys));
2331 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2332 if (skb) {
2333 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2334 } else {
2335 hash_keys.addrs.v6addrs.src = fl6->saddr;
2336 hash_keys.addrs.v6addrs.dst = fl6->daddr;
fa1be7e0 2337 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
b4bac172
DA		 2338 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2339 }
2340 break;
2341 case 1:
2342 if (skb) {
2343 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2344 struct flow_keys keys;
2345
2346 /* short-circuit if we already have L4 hash present */
2347 if (skb->l4_hash)
2348 return skb_get_hash_raw(skb) >> 1;
2349
2350 memset(&hash_keys, 0, sizeof(hash_keys));
2351
2352 if (!flkeys) {
2353 skb_flow_dissect_flow_keys(skb, &keys, flag);
2354 flkeys = &keys;
2355 }
2356 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2357 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2358 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2359 hash_keys.ports.src = flkeys->ports.src;
2360 hash_keys.ports.dst = flkeys->ports.dst;
2361 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2362 } else {
2363 memset(&hash_keys, 0, sizeof(hash_keys));
2364 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2365 hash_keys.addrs.v6addrs.src = fl6->saddr;
2366 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2367 hash_keys.ports.src = fl6->fl6_sport;
2368 hash_keys.ports.dst = fl6->fl6_dport;
2369 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2370 }
2371 break;
23aebdac 2372 }
9a2a537a 2373 mhash = flow_hash_from_keys(&hash_keys);
23aebdac 2374
9a2a537a 2375 return mhash >> 1;
23aebdac
JS		 2376}
2377
67f415dd 2378/* Called with rcu held */
c71099ac
TG		 2379void ip6_route_input(struct sk_buff *skb)
2380{
b71d1d42 2381 const struct ipv6hdr *iph = ipv6_hdr(skb);
c346dca1 2382 struct net *net = dev_net(skb->dev);
67f415dd 2383 int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
904af04d 2384 struct ip_tunnel_info *tun_info;
4c9483b2 2385 struct flowi6 fl6 = {
e0d56fdd 2386 .flowi6_iif = skb->dev->ifindex,
4c9483b2
DM		 2387 .daddr = iph->daddr,
2388 .saddr = iph->saddr,
6502ca52 2389 .flowlabel = ip6_flowinfo(iph),
4c9483b2
DM		 2390 .flowi6_mark = skb->mark,
2391 .flowi6_proto = iph->nexthdr,
c71099ac 2392 };
5e5d6fed 2393 struct flow_keys *flkeys = NULL, _flkeys;
adaa70bb 2394
904af04d 2395 tun_info = skb_tunnel_info(skb);
46fa062a 2396 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
904af04d 2397 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
5e5d6fed
RP		 2398
2399 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2400 flkeys = &_flkeys;
2401
23aebdac 2402 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
b4bac172 2403 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
06e9d040 2404 skb_dst_drop(skb);
67f415dd
WW		 2405 skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2406 &fl6, skb, flags));
c71099ac
TG		 2407}
2408
b75cc8f9
DA		 2409static struct rt6_info *ip6_pol_route_output(struct net *net,
2410 struct fib6_table *table,
2411 struct flowi6 *fl6,
2412 const struct sk_buff *skb,
2413 int flags)
1da177e4 2414{
b75cc8f9 2415 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
c71099ac
TG		 2416}
2417
7d9e5f42
WW		 2418struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2419 const struct sock *sk,
2420 struct flowi6 *fl6, int flags)
c71099ac 2421{
d46a9d67 2422 bool any_src;
c71099ac 2423
3ede0bbc
RS		 2424 if (ipv6_addr_type(&fl6->daddr) &
2425 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
4c1feac5
DA		 2426 struct dst_entry *dst;
2427
7d9e5f42 2428 /* This function does not take refcnt on the dst */
4c1feac5
DA		 2429 dst = l3mdev_link_scope_lookup(net, fl6);
2430 if (dst)
2431 return dst;
2432 }
ca254490 2433
1fb9489b 2434 fl6->flowi6_iif = LOOPBACK_IFINDEX;
4dc27d1c 2435
7d9e5f42 2436 flags |= RT6_LOOKUP_F_DST_NOREF;
d46a9d67 2437 any_src = ipv6_addr_any(&fl6->saddr);
741a11d9 2438 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
d46a9d67 2439 (fl6->flowi6_oif && any_src))
77d16f45 2440 flags |= RT6_LOOKUP_F_IFACE;
c71099ac 2441
d46a9d67 2442 if (!any_src)
adaa70bb 2443 flags |= RT6_LOOKUP_F_HAS_SADDR;
0c9a2ac1
YH		 2444 else if (sk)
2445 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
adaa70bb 2446
b75cc8f9 2447 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
1da177e4 2448}
7d9e5f42
WW		 2449EXPORT_SYMBOL_GPL(ip6_route_output_flags_noref);
2450
2451struct dst_entry *ip6_route_output_flags(struct net *net,
2452 const struct sock *sk,
2453 struct flowi6 *fl6,
2454 int flags)
2455{
2456 struct dst_entry *dst;
2457 struct rt6_info *rt6;
2458
2459 rcu_read_lock();
2460 dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2461 rt6 = (struct rt6_info *)dst;
2462 /* For dst cached in uncached_list, refcnt is already taken. */
2463 if (list_empty(&rt6->rt6i_uncached) && !dst_hold_safe(dst)) {
2464 dst = &net->ipv6.ip6_null_entry->dst;
2465 dst_hold(dst);
2466 }
2467 rcu_read_unlock();
2468
2469 return dst;
2470}
6f21c96a 2471EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1da177e4 2472
2774c131 2473struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
14e50e57 2474{
5c1e6aa3 2475 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1dbe3252 2476 struct net_device *loopback_dev = net->loopback_dev;
14e50e57
DM		 2477 struct dst_entry *new = NULL;
2478
1dbe3252 2479 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
62cf27e5 2480 DST_OBSOLETE_DEAD, 0);
14e50e57 2481 if (rt) {
0a1f5962 2482 rt6_info_init(rt);
81eb8447 2483 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
8104891b 2484
0a1f5962 2485 new = &rt->dst;
14e50e57 2486 new->__use = 1;
352e512c 2487 new->input = dst_discard;
ede2059d 2488 new->output = dst_discard_out;
14e50e57 2489
0a1f5962 2490 dst_copy_metrics(new, &ort->dst);
14e50e57 2491
1dbe3252 2492 rt->rt6i_idev = in6_dev_get(loopback_dev);
4e3fd7a0 2493 rt->rt6i_gateway = ort->rt6i_gateway;
0a1f5962 2494 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
14e50e57
DM		 2495
2496 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2497#ifdef CONFIG_IPV6_SUBTREES
2498 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2499#endif
14e50e57
DM		 2500 }
2501
69ead7af
DM		 2502 dst_release(dst_orig);
2503 return new ? new : ERR_PTR(-ENOMEM);
14e50e57 2504}
14e50e57 2505
1da177e4
LT		 2506/*
2507 * Destination cache support functions
2508 */
2509
8d1c802b 2510static bool fib6_check(struct fib6_info *f6i, u32 cookie)
4b32b5ad 2511{
93531c67
DA		 2512 u32 rt_cookie = 0;
2513
8ae86971 2514 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
93531c67
DA		 2515 return false;
2516
2517 if (fib6_check_expired(f6i))
2518 return false;
2519
2520 return true;
4b32b5ad
MKL		 2521}
2522
a68886a6
DA		 2523static struct dst_entry *rt6_check(struct rt6_info *rt,
2524 struct fib6_info *from,
2525 u32 cookie)
3da59bd9 2526{
36143645 2527 u32 rt_cookie = 0;
c5cff856 2528
a68886a6 2529 if ((from && !fib6_get_cookie_safe(from, &rt_cookie)) ||
93531c67 2530 rt_cookie != cookie)
3da59bd9
MKL		 2531 return NULL;
2532
2533 if (rt6_check_expired(rt))
2534 return NULL;
2535
2536 return &rt->dst;
2537}
2538
a68886a6
DA		 2539static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2540 struct fib6_info *from,
2541 u32 cookie)
3da59bd9 2542{
5973fb1e
MKL		 2543 if (!__rt6_check_expired(rt) &&
2544 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
a68886a6 2545 fib6_check(from, cookie))
3da59bd9
MKL		 2546 return &rt->dst;
2547 else
2548 return NULL;
2549}
2550
1da177e4
LT		 2551static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
2552{
a87b7dc9 2553 struct dst_entry *dst_ret;
a68886a6 2554 struct fib6_info *from;
1da177e4
LT		 2555 struct rt6_info *rt;
2556
a87b7dc9
DA		 2557 rt = container_of(dst, struct rt6_info, dst);
2558
2559 rcu_read_lock();
1da177e4 2560
6f3118b5
ND		 2561 /* All IPV6 dsts are created with ->obsolete set to the value
2562 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2563 * into this function always.
2564 */
e3bc10bd 2565
a68886a6 2566 from = rcu_dereference(rt->from);
4b32b5ad 2567
a68886a6
DA		 2568 if (from && (rt->rt6i_flags & RTF_PCPU ||
2569 unlikely(!list_empty(&rt->rt6i_uncached))))
2570 dst_ret = rt6_dst_from_check(rt, from, cookie);
3da59bd9 2571 else
a68886a6 2572 dst_ret = rt6_check(rt, from, cookie);
a87b7dc9
DA		 2573
2574 rcu_read_unlock();
2575
2576 return dst_ret;
1da177e4
LT		 2577}
2578
2579static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2580{
2581 struct rt6_info *rt = (struct rt6_info *) dst;
2582
2583 if (rt) {
54c1a859 2584 if (rt->rt6i_flags & RTF_CACHE) {
c3c14da0 2585 rcu_read_lock();
54c1a859 2586 if (rt6_check_expired(rt)) {
93531c67 2587 rt6_remove_exception_rt(rt);
54c1a859
YH		 2588 dst = NULL;
2589 }
c3c14da0 2590 rcu_read_unlock();
54c1a859 2591 } else {
1da177e4 2592 dst_release(dst);
54c1a859
YH		 2593 dst = NULL;
2594 }
1da177e4 2595 }
54c1a859 2596 return dst;
1da177e4
LT		 2597}
2598
2599static void ip6_link_failure(struct sk_buff *skb)
2600{
2601 struct rt6_info *rt;
2602
3ffe533c 2603 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1da177e4 2604
adf30907 2605 rt = (struct rt6_info *) skb_dst(skb);
1da177e4 2606 if (rt) {
8a14e46f 2607 rcu_read_lock();
1eb4f758 2608 if (rt->rt6i_flags & RTF_CACHE) {
761f6026 2609 rt6_remove_exception_rt(rt);
c5cff856 2610 } else {
a68886a6 2611 struct fib6_info *from;
c5cff856
WW		 2612 struct fib6_node *fn;
2613
a68886a6
DA		 2614 from = rcu_dereference(rt->from);
2615 if (from) {
2616 fn = rcu_dereference(from->fib6_node);
2617 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2618 fn->fn_sernum = -1;
2619 }
1eb4f758 2620 }
8a14e46f 2621 rcu_read_unlock();
1da177e4
LT		 2622 }
2623}
2624
6a3e030f
DA		 2625static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2626{
a68886a6
DA		 2627 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2628 struct fib6_info *from;
2629
2630 rcu_read_lock();
2631 from = rcu_dereference(rt0->from);
2632 if (from)
2633 rt0->dst.expires = from->expires;
2634 rcu_read_unlock();
2635 }
6a3e030f
DA		 2636
2637 dst_set_expires(&rt0->dst, timeout);
2638 rt0->rt6i_flags |= RTF_EXPIRES;
2639}
2640
45e4fd26
MKL		 2641static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2642{
2643 struct net *net = dev_net(rt->dst.dev);
2644
d4ead6b3 2645 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
45e4fd26 2646 rt->rt6i_flags |= RTF_MODIFIED;
45e4fd26
MKL		 2647 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2648}
2649
0d3f6d29
MKL		 2650static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2651{
2652 return !(rt->rt6i_flags & RTF_CACHE) &&
1490ed2a 2653 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
0d3f6d29
MKL		 2654}
2655
45e4fd26
MKL		 2656static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2657 const struct ipv6hdr *iph, u32 mtu)
1da177e4 2658{
0dec879f 2659 const struct in6_addr *daddr, *saddr;
67ba4152 2660 struct rt6_info *rt6 = (struct rt6_info *)dst;
1da177e4 2661
19bda36c
XL		 2662 if (dst_metric_locked(dst, RTAX_MTU))
2663 return;
2664
0dec879f
JA		 2665 if (iph) {
2666 daddr = &iph->daddr;
2667 saddr = &iph->saddr;
2668 } else if (sk) {
2669 daddr = &sk->sk_v6_daddr;
2670 saddr = &inet6_sk(sk)->saddr;
2671 } else {
2672 daddr = NULL;
2673 saddr = NULL;
2674 }
2675 dst_confirm_neigh(dst, daddr);
45e4fd26
MKL		 2676 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
2677 if (mtu >= dst_mtu(dst))
2678 return;
9d289715 2679
0d3f6d29 2680 if (!rt6_cache_allowed_for_pmtu(rt6)) {
45e4fd26 2681 rt6_do_update_pmtu(rt6, mtu);
2b760fcf
WW		 2682 /* update rt6_ex->stamp for cache */
2683 if (rt6->rt6i_flags & RTF_CACHE)
2684 rt6_update_exception_stamp_rt(rt6);
0dec879f 2685 } else if (daddr) {
85bd05de 2686 struct fib6_result res = {};
45e4fd26
MKL		 2687 struct rt6_info *nrt6;
2688
4d85cd0c 2689 rcu_read_lock();
85bd05de
DA		 2690 res.f6i = rcu_dereference(rt6->from);
2691 if (!res.f6i) {
9c69a132
JL		 2692 rcu_read_unlock();
2693 return;
2694 }
7d21fec9
DA		 2695 res.fib6_flags = res.f6i->fib6_flags;
2696 res.fib6_type = res.f6i->fib6_type;
2697
2d44234b
DA		 2698 if (res.f6i->nh) {
2699 struct fib6_nh_match_arg arg = {
2700 .dev = dst->dev,
2701 .gw = &rt6->rt6i_gateway,
2702 };
2703
2704 nexthop_for_each_fib6_nh(res.f6i->nh,
2705 fib6_nh_find_match, &arg);
2706
2707 /* fib6_info uses a nexthop that does not have fib6_nh
2708 * using the dst->dev + gw. Should be impossible.
2709 */
2710 if (!arg.match) {
2711 rcu_read_unlock();
2712 return;
2713 }
2714
2715 res.nh = arg.match;
2716 } else {
2717 res.nh = res.f6i->fib6_nh;
2718 }
2719
85bd05de 2720 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
45e4fd26
MKL		 2721 if (nrt6) {
2722 rt6_do_update_pmtu(nrt6, mtu);
5012f0a5 2723 if (rt6_insert_exception(nrt6, &res))
2b760fcf 2724 dst_release_immediate(&nrt6->dst);
45e4fd26 2725 }
a68886a6 2726 rcu_read_unlock();
1da177e4
LT		 2727 }
2728}
2729
45e4fd26
MKL		 2730static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2731 struct sk_buff *skb, u32 mtu)
2732{
2733 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
2734}
2735
42ae66c8 2736void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
e2d118a1 2737 int oif, u32 mark, kuid_t uid)
81aded24
DM		 2738{
2739 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2740 struct dst_entry *dst;
dc92095d
MŻ		 2741 struct flowi6 fl6 = {
2742 .flowi6_oif = oif,
2743 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2744 .daddr = iph->daddr,
2745 .saddr = iph->saddr,
2746 .flowlabel = ip6_flowinfo(iph),
2747 .flowi6_uid = uid,
2748 };
81aded24
DM		 2749
2750 dst = ip6_route_output(net, NULL, &fl6);
2751 if (!dst->error)
45e4fd26 2752 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
81aded24
DM		 2753 dst_release(dst);
2754}
2755EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2756
2757void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2758{
7ddacfa5 2759 int oif = sk->sk_bound_dev_if;
33c162a9
MKL		 2760 struct dst_entry *dst;
2761
7ddacfa5
DA		 2762 if (!oif && skb->dev)
2763 oif = l3mdev_master_ifindex(skb->dev);
2764
2765 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
33c162a9
MKL		 2766
2767 dst = __sk_dst_get(sk);
2768 if (!dst || !dst->obsolete ||
2769 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2770 return;
2771
2772 bh_lock_sock(sk);
2773 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2774 ip6_datagram_dst_update(sk, false);
2775 bh_unlock_sock(sk);
81aded24
DM		 2776}
2777EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2778
7d6850f7
AK		 2779void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2780 const struct flowi6 *fl6)
2781{
2782#ifdef CONFIG_IPV6_SUBTREES
2783 struct ipv6_pinfo *np = inet6_sk(sk);
2784#endif
2785
2786 ip6_dst_store(sk, dst,
2787 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2788 &sk->sk_v6_daddr : NULL,
2789#ifdef CONFIG_IPV6_SUBTREES
2790 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2791 &np->saddr :
2792#endif
2793 NULL);
2794}
2795
9b6b35ab 2796static bool ip6_redirect_nh_match(const struct fib6_result *res,
0b34eb00
DA		 2797 struct flowi6 *fl6,
2798 const struct in6_addr *gw,
2799 struct rt6_info **ret)
2800{
9b6b35ab
DA		 2801 const struct fib6_nh *nh = res->nh;
2802
0b34eb00
DA		 2803 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2804 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2805 return false;
2806
2807 /* rt_cache's gateway might be different from its 'parent'
2808 * in the case of an ip redirect.
2809 * So we keep searching in the exception table if the gateway
2810 * is different.
2811 */
2812 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
2813 struct rt6_info *rt_cache;
2814
9b6b35ab 2815 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
0b34eb00
DA		 2816 if (rt_cache &&
2817 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
2818 *ret = rt_cache;
2819 return true;
2820 }
2821 return false;
2822 }
2823 return true;
2824}
2825
c55c8988
DA		 2826struct fib6_nh_rd_arg {
2827 struct fib6_result *res;
2828 struct flowi6 *fl6;
2829 const struct in6_addr *gw;
2830 struct rt6_info **ret;
2831};
2832
2833static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
2834{
2835 struct fib6_nh_rd_arg *arg = _arg;
2836
2837 arg->res->nh = nh;
2838 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
2839}
2840
b55b76b2
DJ		 2841/* Handle redirects */
2842struct ip6rd_flowi {
2843 struct flowi6 fl6;
2844 struct in6_addr gateway;
2845};
2846
2847static struct rt6_info *__ip6_route_redirect(struct net *net,
2848 struct fib6_table *table,
2849 struct flowi6 *fl6,
b75cc8f9 2850 const struct sk_buff *skb,
b55b76b2
DJ		 2851 int flags)
2852{
2853 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
0b34eb00 2854 struct rt6_info *ret = NULL;
9b6b35ab 2855 struct fib6_result res = {};
c55c8988
DA		 2856 struct fib6_nh_rd_arg arg = {
2857 .res = &res,
2858 .fl6 = fl6,
2859 .gw = &rdfl->gateway,
2860 .ret = &ret
2861 };
8d1c802b 2862 struct fib6_info *rt;
b55b76b2
DJ		 2863 struct fib6_node *fn;
2864
31680ac2
DA		 2865 /* l3mdev_update_flow overrides oif if the device is enslaved; in
2866 * this case we must match on the real ingress device, so reset it
2867 */
2868 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2869 fl6->flowi6_oif = skb->dev->ifindex;
2870
b55b76b2 2871 /* Get the "current" route for this destination and
67c408cf 2872 * check if the redirect has come from appropriate router.
b55b76b2
DJ		 2873 *
2874 * RFC 4861 specifies that redirects should only be
2875 * accepted if they come from the nexthop to the target.
2876 * Due to the way the routes are chosen, this notion
2877 * is a bit fuzzy and one might need to check all possible
2878 * routes.
2879 */
2880
66f5d6ce 2881 rcu_read_lock();
6454743b 2882 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
b55b76b2 2883restart:
66f5d6ce 2884 for_each_fib6_node_rt_rcu(fn) {
9b6b35ab 2885 res.f6i = rt;
14895687 2886 if (fib6_check_expired(rt))
b55b76b2 2887 continue;
93c2fb25 2888 if (rt->fib6_flags & RTF_REJECT)
b55b76b2 2889 break;
c55c8988
DA		 2890 if (unlikely(rt->nh)) {
2891 if (nexthop_is_blackhole(rt->nh))
2892 continue;
2893 /* on match, res->nh is filled in and potentially ret */
2894 if (nexthop_for_each_fib6_nh(rt->nh,
2895 fib6_nh_redirect_match,
2896 &arg))
2897 goto out;
2898 } else {
2899 res.nh = rt->fib6_nh;
2900 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
2901 &ret))
2902 goto out;
2903 }
b55b76b2
DJ		 2904 }
2905
2906 if (!rt)
421842ed 2907 rt = net->ipv6.fib6_null_entry;
93c2fb25 2908 else if (rt->fib6_flags & RTF_REJECT) {
23fb93a4 2909 ret = net->ipv6.ip6_null_entry;
b0a1ba59
MKL		 2910 goto out;
2911 }
2912
421842ed 2913 if (rt == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2914 fn = fib6_backtrack(fn, &fl6->saddr);
2915 if (fn)
2916 goto restart;
b55b76b2 2917 }
a3c00e46 2918
9b6b35ab 2919 res.f6i = rt;
1cf844c7 2920 res.nh = rt->fib6_nh;
b0a1ba59 2921out:
7d21fec9 2922 if (ret) {
10585b43 2923 ip6_hold_safe(net, &ret);
7d21fec9
DA		 2924 } else {
2925 res.fib6_flags = res.f6i->fib6_flags;
2926 res.fib6_type = res.f6i->fib6_type;
9b6b35ab 2927 ret = ip6_create_rt_rcu(&res);
7d21fec9 2928 }
b55b76b2 2929
66f5d6ce 2930 rcu_read_unlock();
b55b76b2 2931
8ff2e5b2 2932 trace_fib6_table_lookup(net, &res, table, fl6);
23fb93a4 2933 return ret;
b55b76b2
DJ		 2934};
2935
2936static struct dst_entry *ip6_route_redirect(struct net *net,
b75cc8f9
DA		 2937 const struct flowi6 *fl6,
2938 const struct sk_buff *skb,
2939 const struct in6_addr *gateway)
b55b76b2
DJ		 2940{
2941 int flags = RT6_LOOKUP_F_HAS_SADDR;
2942 struct ip6rd_flowi rdfl;
2943
2944 rdfl.fl6 = *fl6;
2945 rdfl.gateway = *gateway;
2946
b75cc8f9 2947 return fib6_rule_lookup(net, &rdfl.fl6, skb,
b55b76b2
DJ		 2948 flags, __ip6_route_redirect);
2949}
2950
e2d118a1
LC		 2951void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
2952 kuid_t uid)
3a5ad2ee
DM		 2953{
2954 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2955 struct dst_entry *dst;
1f7f10ac
MŻ		 2956 struct flowi6 fl6 = {
2957 .flowi6_iif = LOOPBACK_IFINDEX,
2958 .flowi6_oif = oif,
2959 .flowi6_mark = mark,
2960 .daddr = iph->daddr,
2961 .saddr = iph->saddr,
2962 .flowlabel = ip6_flowinfo(iph),
2963 .flowi6_uid = uid,
2964 };
3a5ad2ee 2965
b75cc8f9 2966 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
b55b76b2 2967 rt6_do_redirect(dst, NULL, skb);
3a5ad2ee
DM		 2968 dst_release(dst);
2969}
2970EXPORT_SYMBOL_GPL(ip6_redirect);
2971
d456336d 2972void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
c92a59ec
DJ		 2973{
2974 const struct ipv6hdr *iph = ipv6_hdr(skb);
2975 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
2976 struct dst_entry *dst;
0b26fb17
MŻ		 2977 struct flowi6 fl6 = {
2978 .flowi6_iif = LOOPBACK_IFINDEX,
2979 .flowi6_oif = oif,
0b26fb17
MŻ		 2980 .daddr = msg->dest,
2981 .saddr = iph->daddr,
2982 .flowi6_uid = sock_net_uid(net, NULL),
2983 };
c92a59ec 2984
b75cc8f9 2985 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
b55b76b2 2986 rt6_do_redirect(dst, NULL, skb);
c92a59ec
DJ		 2987 dst_release(dst);
2988}
2989
3a5ad2ee
DM		 2990void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
2991{
e2d118a1
LC		 2992 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
2993 sk->sk_uid);
3a5ad2ee
DM		 2994}
2995EXPORT_SYMBOL_GPL(ip6_sk_redirect);
2996
0dbaee3b 2997static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1da177e4 2998{
0dbaee3b
DM		 2999 struct net_device *dev = dst->dev;
3000 unsigned int mtu = dst_mtu(dst);
3001 struct net *net = dev_net(dev);
3002
1da177e4
LT		 3003 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3004
5578689a
DL		 3005 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3006 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1da177e4
LT		 3007
3008 /*
1ab1457c
YH		 3009 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3010 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3011 * IPV6_MAXPLEN is also valid and means: "any MSS,
1da177e4
LT		 3012 * rely only on pmtu discovery"
3013 */
3014 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3015 mtu = IPV6_MAXPLEN;
3016 return mtu;
3017}
3018
ebb762f2 3019static unsigned int ip6_mtu(const struct dst_entry *dst)
d33e4553 3020{
d33e4553 3021 struct inet6_dev *idev;
d4ead6b3 3022 unsigned int mtu;
4b32b5ad
MKL		 3023
3024 mtu = dst_metric_raw(dst, RTAX_MTU);
618f9bc7 3025 if (mtu)
30f78d8e 3026 goto out;
618f9bc7
SK		 3027
3028 mtu = IPV6_MIN_MTU;
d33e4553
DM		 3029
3030 rcu_read_lock();
3031 idev = __in6_dev_get(dst->dev);
3032 if (idev)
3033 mtu = idev->cnf.mtu6;
3034 rcu_read_unlock();
3035
30f78d8e 3036out:
14972cbd
RP		 3037 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3038
3039 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
d33e4553
DM		 3040}
3041
901731b8
DA		 3042/* MTU selection:
3043 * 1. mtu on route is locked - use it
3044 * 2. mtu from nexthop exception
3045 * 3. mtu from egress device
3046 *
3047 * based on ip6_dst_mtu_forward and exception logic of
3048 * rt6_find_cached_rt; called with rcu_read_lock
3049 */
b748f260
DA		 3050u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3051 const struct in6_addr *daddr,
3052 const struct in6_addr *saddr)
901731b8 3053{
b748f260
DA		 3054 const struct fib6_nh *nh = res->nh;
3055 struct fib6_info *f6i = res->f6i;
901731b8 3056 struct inet6_dev *idev;
510e2ced 3057 struct rt6_info *rt;
901731b8
DA		 3058 u32 mtu = 0;
3059
3060 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3061 mtu = f6i->fib6_pmtu;
3062 if (mtu)
3063 goto out;
3064 }
3065
510e2ced
WW		 3066 rt = rt6_find_cached_rt(res, daddr, saddr);
3067 if (unlikely(rt)) {
3068 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3069 } else {
b748f260 3070 struct net_device *dev = nh->fib_nh_dev;
901731b8
DA		 3071
3072 mtu = IPV6_MIN_MTU;
3073 idev = __in6_dev_get(dev);
3074 if (idev && idev->cnf.mtu6 > mtu)
3075 mtu = idev->cnf.mtu6;
3076 }
3077
3078 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3079out:
b748f260 3080 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
901731b8
DA		 3081}
3082
3b00944c 3083struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
87a11578 3084 struct flowi6 *fl6)
1da177e4 3085{
87a11578 3086 struct dst_entry *dst;
1da177e4
LT		 3087 struct rt6_info *rt;
3088 struct inet6_dev *idev = in6_dev_get(dev);
c346dca1 3089 struct net *net = dev_net(dev);
1da177e4 3090
38308473 3091 if (unlikely(!idev))
122bdf67 3092 return ERR_PTR(-ENODEV);
1da177e4 3093
ad706862 3094 rt = ip6_dst_alloc(net, dev, 0);
38308473 3095 if (unlikely(!rt)) {
1da177e4 3096 in6_dev_put(idev);
87a11578 3097 dst = ERR_PTR(-ENOMEM);
1da177e4
LT		 3098 goto out;
3099 }
3100
8e2ec639 3101 rt->dst.flags |= DST_HOST;
588753f1 3102 rt->dst.input = ip6_input;
8e2ec639 3103 rt->dst.output = ip6_output;
550bab42 3104 rt->rt6i_gateway = fl6->daddr;
87a11578 3105 rt->rt6i_dst.addr = fl6->daddr;
8e2ec639
YZ		 3106 rt->rt6i_dst.plen = 128;
3107 rt->rt6i_idev = idev;
14edd87d 3108 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1da177e4 3109
4c981e28 3110 /* Add this dst into uncached_list so that rt6_disable_ip() can
587fea74
WW		 3111 * do proper release of the net_device
3112 */
3113 rt6_uncached_list_add(rt);
81eb8447 3114 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
1da177e4 3115
87a11578
DM		 3116 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3117
1da177e4 3118out:
87a11578 3119 return dst;
1da177e4
LT		 3120}
3121
569d3645 3122static int ip6_dst_gc(struct dst_ops *ops)
1da177e4 3123{
86393e52 3124 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
7019b78e
DL		 3125 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3126 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3127 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3128 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3129 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
fc66f95c 3130 int entries;
7019b78e 3131
fc66f95c 3132 entries = dst_entries_get_fast(ops);
49a18d86 3133 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
fc66f95c 3134 entries <= rt_max_size)
1da177e4
LT		 3135 goto out;
3136
6891a346 3137 net->ipv6.ip6_rt_gc_expire++;
14956643 3138 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
fc66f95c
ED		 3139 entries = dst_entries_get_slow(ops);
3140 if (entries < ops->gc_thresh)
7019b78e 3141 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1da177e4 3142out:
7019b78e 3143 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
fc66f95c 3144 return entries > rt_max_size;
1da177e4
LT		 3145}
3146
8c14586f
DA		 3147static struct rt6_info *ip6_nh_lookup_table(struct net *net,
3148 struct fib6_config *cfg,
f4797b33
DA		 3149 const struct in6_addr *gw_addr,
3150 u32 tbid, int flags)
8c14586f
DA		 3151{
3152 struct flowi6 fl6 = {
3153 .flowi6_oif = cfg->fc_ifindex,
3154 .daddr = *gw_addr,
3155 .saddr = cfg->fc_prefsrc,
3156 };
3157 struct fib6_table *table;
3158 struct rt6_info *rt;
8c14586f 3159
f4797b33 3160 table = fib6_get_table(net, tbid);
8c14586f
DA		 3161 if (!table)
3162 return NULL;
3163
3164 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3165 flags |= RT6_LOOKUP_F_HAS_SADDR;
3166
f4797b33 3167 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
b75cc8f9 3168 rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, NULL, flags);
8c14586f
DA		 3169
3170 /* if table lookup failed, fall back to full lookup */
3171 if (rt == net->ipv6.ip6_null_entry) {
3172 ip6_rt_put(rt);
3173 rt = NULL;
3174 }
3175
3176 return rt;
3177}
3178
fc1e64e1
DA		 3179static int ip6_route_check_nh_onlink(struct net *net,
3180 struct fib6_config *cfg,
9fbb704c 3181 const struct net_device *dev,
fc1e64e1
DA		 3182 struct netlink_ext_ack *extack)
3183{
44750f84 3184 u32 tbid = l3mdev_fib_table(dev) ? : RT_TABLE_MAIN;
fc1e64e1
DA		 3185 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3186 u32 flags = RTF_LOCAL | RTF_ANYCAST | RTF_REJECT;
bf1dc8ba 3187 struct fib6_info *from;
fc1e64e1
DA		 3188 struct rt6_info *grt;
3189 int err;
3190
3191 err = 0;
3192 grt = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0);
3193 if (grt) {
bf1dc8ba
PA		 3194 rcu_read_lock();
3195 from = rcu_dereference(grt->from);
58e354c0 3196 if (!grt->dst.error &&
4ed591c8 3197 /* ignore match if it is the default route */
bf1dc8ba 3198 from && !ipv6_addr_any(&from->fib6_dst.addr) &&
58e354c0 3199 (grt->rt6i_flags & flags || dev != grt->dst.dev)) {
44750f84
DA		 3200 NL_SET_ERR_MSG(extack,
3201 "Nexthop has invalid gateway or device mismatch");
fc1e64e1
DA		 3202 err = -EINVAL;
3203 }
bf1dc8ba 3204 rcu_read_unlock();
fc1e64e1
DA		 3205
3206 ip6_rt_put(grt);
3207 }
3208
3209 return err;
3210}
3211
1edce99f
DA		 3212static int ip6_route_check_nh(struct net *net,
3213 struct fib6_config *cfg,
3214 struct net_device **_dev,
3215 struct inet6_dev **idev)
3216{
3217 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3218 struct net_device *dev = _dev ? *_dev : NULL;
3219 struct rt6_info *grt = NULL;
3220 int err = -EHOSTUNREACH;
3221
3222 if (cfg->fc_table) {
f4797b33
DA		 3223 int flags = RT6_LOOKUP_F_IFACE;
3224
3225 grt = ip6_nh_lookup_table(net, cfg, gw_addr,
3226 cfg->fc_table, flags);
1edce99f
DA		 3227 if (grt) {
3228 if (grt->rt6i_flags & RTF_GATEWAY ||
3229 (dev && dev != grt->dst.dev)) {
3230 ip6_rt_put(grt);
3231 grt = NULL;
3232 }
3233 }
3234 }
3235
3236 if (!grt)
b75cc8f9 3237 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, NULL, 1);
1edce99f
DA		 3238
3239 if (!grt)
3240 goto out;
3241
3242 if (dev) {
3243 if (dev != grt->dst.dev) {
3244 ip6_rt_put(grt);
3245 goto out;
3246 }
3247 } else {
3248 *_dev = dev = grt->dst.dev;
3249 *idev = grt->rt6i_idev;
3250 dev_hold(dev);
3251 in6_dev_hold(grt->rt6i_idev);
3252 }
3253
3254 if (!(grt->rt6i_flags & RTF_GATEWAY))
3255 err = 0;
3256
3257 ip6_rt_put(grt);
3258
3259out:
3260 return err;
3261}
3262
9fbb704c
DA		 3263static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3264 struct net_device **_dev, struct inet6_dev **idev,
3265 struct netlink_ext_ack *extack)
3266{
3267 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3268 int gwa_type = ipv6_addr_type(gw_addr);
232378e8 3269 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
9fbb704c 3270 const struct net_device *dev = *_dev;
232378e8 3271 bool need_addr_check = !dev;
9fbb704c
DA		 3272 int err = -EINVAL;
3273
3274 /* if gw_addr is local we will fail to detect this in case
3275 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3276 * will return already-added prefix route via interface that
3277 * prefix route was assigned to, which might be non-loopback.
3278 */
232378e8
DA		 3279 if (dev &&
3280 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3281 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
9fbb704c
DA		 3282 goto out;
3283 }
3284
3285 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3286 /* IPv6 strictly inhibits using not link-local
3287 * addresses as nexthop address.
3288 * Otherwise, router will not able to send redirects.
3289 * It is very good, but in some (rare!) circumstances
3290 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3291 * some exceptions. --ANK
3292 * We allow IPv4-mapped nexthops to support RFC4798-type
3293 * addressing
3294 */
3295 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3296 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3297 goto out;
3298 }
3299
3300 if (cfg->fc_flags & RTNH_F_ONLINK)
3301 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3302 else
3303 err = ip6_route_check_nh(net, cfg, _dev, idev);
3304
3305 if (err)
3306 goto out;
3307 }
3308
3309 /* reload in case device was changed */
3310 dev = *_dev;
3311
3312 err = -EINVAL;
3313 if (!dev) {
3314 NL_SET_ERR_MSG(extack, "Egress device not specified");
3315 goto out;
3316 } else if (dev->flags & IFF_LOOPBACK) {
3317 NL_SET_ERR_MSG(extack,
3318 "Egress device can not be loopback device for this route");
3319 goto out;
3320 }
232378e8
DA		 3321
3322 /* if we did not check gw_addr above, do so now that the
3323 * egress device has been resolved.
3324 */
3325 if (need_addr_check &&
3326 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3327 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3328 goto out;
3329 }
3330
9fbb704c
DA		 3331 err = 0;
3332out:
3333 return err;
3334}
3335
83c44251
DA		 3336static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3337{
3338 if ((flags & RTF_REJECT) ||
3339 (dev && (dev->flags & IFF_LOOPBACK) &&
3340 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3341 !(flags & RTF_LOCAL)))
3342 return true;
3343
3344 return false;
3345}
3346
3347int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3348 struct fib6_config *cfg, gfp_t gfp_flags,
3349 struct netlink_ext_ack *extack)
3350{
3351 struct net_device *dev = NULL;
3352 struct inet6_dev *idev = NULL;
3353 int addr_type;
3354 int err;
3355
f1741730
DA		 3356 fib6_nh->fib_nh_family = AF_INET6;
3357
83c44251
DA		 3358 err = -ENODEV;
3359 if (cfg->fc_ifindex) {
3360 dev = dev_get_by_index(net, cfg->fc_ifindex);
3361 if (!dev)
3362 goto out;
3363 idev = in6_dev_get(dev);
3364 if (!idev)
3365 goto out;
3366 }
3367
3368 if (cfg->fc_flags & RTNH_F_ONLINK) {
3369 if (!dev) {
3370 NL_SET_ERR_MSG(extack,
3371 "Nexthop device required for onlink");
3372 goto out;
3373 }
3374
3375 if (!(dev->flags & IFF_UP)) {
3376 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3377 err = -ENETDOWN;
3378 goto out;
3379 }
3380
ad1601ae 3381 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
83c44251
DA		 3382 }
3383
ad1601ae 3384 fib6_nh->fib_nh_weight = 1;
83c44251
DA		 3385
3386 /* We cannot add true routes via loopback here,
3387 * they would result in kernel looping; promote them to reject routes
3388 */
3389 addr_type = ipv6_addr_type(&cfg->fc_dst);
3390 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3391 /* hold loopback dev/idev if we haven't done so. */
3392 if (dev != net->loopback_dev) {
3393 if (dev) {
3394 dev_put(dev);
3395 in6_dev_put(idev);
3396 }
3397 dev = net->loopback_dev;
3398 dev_hold(dev);
3399 idev = in6_dev_get(dev);
3400 if (!idev) {
3401 err = -ENODEV;
3402 goto out;
3403 }
3404 }
7dd73168 3405 goto pcpu_alloc;
83c44251
DA		 3406 }
3407
3408 if (cfg->fc_flags & RTF_GATEWAY) {
3409 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3410 if (err)
3411 goto out;
3412
ad1601ae 3413 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
bdf00467 3414 fib6_nh->fib_nh_gw_family = AF_INET6;
83c44251
DA		 3415 }
3416
3417 err = -ENODEV;
3418 if (!dev)
3419 goto out;
3420
3421 if (idev->cnf.disable_ipv6) {
3422 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3423 err = -EACCES;
3424 goto out;
3425 }
3426
3427 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3428 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3429 err = -ENETDOWN;
3430 goto out;
3431 }
3432
3433 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3434 !netif_carrier_ok(dev))
ad1601ae 3435 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
83c44251 3436
7dd73168
DA		 3437 err = fib_nh_common_init(&fib6_nh->nh_common, cfg->fc_encap,
3438 cfg->fc_encap_type, cfg, gfp_flags, extack);
3439 if (err)
3440 goto out;
3441
3442pcpu_alloc:
f40b6ae2
DA		 3443 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3444 if (!fib6_nh->rt6i_pcpu) {
3445 err = -ENOMEM;
3446 goto out;
3447 }
3448
ad1601ae 3449 fib6_nh->fib_nh_dev = dev;
f1741730 3450 fib6_nh->fib_nh_oif = dev->ifindex;
83c44251
DA		 3451 err = 0;
3452out:
3453 if (idev)
3454 in6_dev_put(idev);
3455
3456 if (err) {
ad1601ae
DA		 3457 lwtstate_put(fib6_nh->fib_nh_lws);
3458 fib6_nh->fib_nh_lws = NULL;
83c44251
DA		 3459 if (dev)
3460 dev_put(dev);
3461 }
3462
3463 return err;
3464}
3465
dac7d0f2
DA		 3466void fib6_nh_release(struct fib6_nh *fib6_nh)
3467{
cc5c073a
DA		 3468 struct rt6_exception_bucket *bucket;
3469
3470 rcu_read_lock();
3471
3472 fib6_nh_flush_exceptions(fib6_nh, NULL);
3473 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3474 if (bucket) {
3475 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3476 kfree(bucket);
3477 }
3478
3479 rcu_read_unlock();
3480
f40b6ae2
DA		 3481 if (fib6_nh->rt6i_pcpu) {
3482 int cpu;
3483
3484 for_each_possible_cpu(cpu) {
3485 struct rt6_info **ppcpu_rt;
3486 struct rt6_info *pcpu_rt;
3487
3488 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3489 pcpu_rt = *ppcpu_rt;
3490 if (pcpu_rt) {
3491 dst_dev_put(&pcpu_rt->dst);
3492 dst_release(&pcpu_rt->dst);
3493 *ppcpu_rt = NULL;
3494 }
3495 }
3496
3497 free_percpu(fib6_nh->rt6i_pcpu);
3498 }
3499
979e276e 3500 fib_nh_common_release(&fib6_nh->nh_common);
dac7d0f2
DA		 3501}
3502
8d1c802b 3503static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
acb54e3c 3504 gfp_t gfp_flags,
333c4301 3505 struct netlink_ext_ack *extack)
1da177e4 3506{
5578689a 3507 struct net *net = cfg->fc_nlinfo.nl_net;
8d1c802b 3508 struct fib6_info *rt = NULL;
f88d8ea6 3509 struct nexthop *nh = NULL;
c71099ac 3510 struct fib6_table *table;
f88d8ea6 3511 struct fib6_nh *fib6_nh;
8c5b83f0 3512 int err = -EINVAL;
83c44251 3513 int addr_type;
1da177e4 3514
557c44be 3515 /* RTF_PCPU is an internal flag; can not be set by userspace */
d5d531cb
DA		 3516 if (cfg->fc_flags & RTF_PCPU) {
3517 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
557c44be 3518 goto out;
d5d531cb 3519 }
557c44be 3520
2ea2352e
WW		 3521 /* RTF_CACHE is an internal flag; can not be set by userspace */
3522 if (cfg->fc_flags & RTF_CACHE) {
3523 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3524 goto out;
3525 }
3526
e8478e80
DA		 3527 if (cfg->fc_type > RTN_MAX) {
3528 NL_SET_ERR_MSG(extack, "Invalid route type");
3529 goto out;
3530 }
3531
d5d531cb
DA		 3532 if (cfg->fc_dst_len > 128) {
3533 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3534 goto out;
3535 }
3536 if (cfg->fc_src_len > 128) {
3537 NL_SET_ERR_MSG(extack, "Invalid source address length");
8c5b83f0 3538 goto out;
d5d531cb 3539 }
1da177e4 3540#ifndef CONFIG_IPV6_SUBTREES
d5d531cb
DA		 3541 if (cfg->fc_src_len) {
3542 NL_SET_ERR_MSG(extack,
3543 "Specifying source address requires IPV6_SUBTREES to be enabled");
8c5b83f0 3544 goto out;
d5d531cb 3545 }
1da177e4 3546#endif
5b98324e
DA		 3547 if (cfg->fc_nh_id) {
3548 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3549 if (!nh) {
3550 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3551 goto out;
3552 }
3553 err = fib6_check_nexthop(nh, cfg, extack);
3554 if (err)
3555 goto out;
3556 }
fc1e64e1 3557
d71314b4 3558 err = -ENOBUFS;
38308473
DM		 3559 if (cfg->fc_nlinfo.nlh &&
3560 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
d71314b4 3561 table = fib6_get_table(net, cfg->fc_table);
38308473 3562 if (!table) {
f3213831 3563 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
d71314b4
MV		 3564 table = fib6_new_table(net, cfg->fc_table);
3565 }
3566 } else {
3567 table = fib6_new_table(net, cfg->fc_table);
3568 }
38308473
DM		 3569
3570 if (!table)
c71099ac 3571 goto out;
c71099ac 3572
93531c67 3573 err = -ENOMEM;
f88d8ea6 3574 rt = fib6_info_alloc(gfp_flags, !nh);
93531c67 3575 if (!rt)
1da177e4 3576 goto out;
93531c67 3577
d7e774f3
DA		 3578 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3579 extack);
767a2217
DA		 3580 if (IS_ERR(rt->fib6_metrics)) {
3581 err = PTR_ERR(rt->fib6_metrics);
fda21d46
ED		 3582 /* Do not leave garbage there. */
3583 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
767a2217
DA		 3584 goto out;
3585 }
3586
93531c67
DA		 3587 if (cfg->fc_flags & RTF_ADDRCONF)
3588 rt->dst_nocount = true;
1da177e4 3589
1716a961 3590 if (cfg->fc_flags & RTF_EXPIRES)
14895687 3591 fib6_set_expires(rt, jiffies +
1716a961
G		 3592 clock_t_to_jiffies(cfg->fc_expires));
3593 else
14895687 3594 fib6_clean_expires(rt);
1da177e4 3595
86872cb5
TG		 3596 if (cfg->fc_protocol == RTPROT_UNSPEC)
3597 cfg->fc_protocol = RTPROT_BOOT;
93c2fb25 3598 rt->fib6_protocol = cfg->fc_protocol;
86872cb5 3599
83c44251
DA		 3600 rt->fib6_table = table;
3601 rt->fib6_metric = cfg->fc_metric;
c7036d97 3602 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
2b2450ca 3603 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
19e42e45 3604
93c2fb25
DA		 3605 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3606 rt->fib6_dst.plen = cfg->fc_dst_len;
3607 if (rt->fib6_dst.plen == 128)
3b6761d1 3608 rt->dst_host = true;
e5fd387a 3609
1da177e4 3610#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 3611 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3612 rt->fib6_src.plen = cfg->fc_src_len;
1da177e4 3613#endif
f88d8ea6
DA		 3614 if (nh) {
3615 if (!nexthop_get(nh)) {
3616 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3617 goto out;
3618 }
3619 if (rt->fib6_src.plen) {
4daa95af 3620 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
f88d8ea6
DA		 3621 goto out;
3622 }
3623 rt->nh = nh;
3624 fib6_nh = nexthop_fib6_nh(rt->nh);
3625 } else {
3626 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3627 if (err)
3628 goto out;
1da177e4 3629
f88d8ea6
DA		 3630 fib6_nh = rt->fib6_nh;
3631
3632 /* We cannot add true routes via loopback here, they would
3633 * result in kernel looping; promote them to reject routes
3634 */
3635 addr_type = ipv6_addr_type(&cfg->fc_dst);
3636 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3637 addr_type))
3638 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3639 }
955ec4cb 3640
c3968a85 3641 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
f88d8ea6 3642 struct net_device *dev = fib6_nh->fib_nh_dev;
83c44251 3643
c3968a85 3644 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
d5d531cb 3645 NL_SET_ERR_MSG(extack, "Invalid source address");
c3968a85
DW		 3646 err = -EINVAL;
3647 goto out;
3648 }
93c2fb25
DA		 3649 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3650 rt->fib6_prefsrc.plen = 128;
c3968a85 3651 } else
93c2fb25 3652 rt->fib6_prefsrc.plen = 0;
c3968a85 3653
8c5b83f0 3654 return rt;
6b9ea5a6 3655out:
93531c67 3656 fib6_info_release(rt);
8c5b83f0 3657 return ERR_PTR(err);
6b9ea5a6
RP		 3658}
3659
acb54e3c 3660int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
333c4301 3661 struct netlink_ext_ack *extack)
6b9ea5a6 3662{
8d1c802b 3663 struct fib6_info *rt;
6b9ea5a6
RP		 3664 int err;
3665
acb54e3c 3666 rt = ip6_route_info_create(cfg, gfp_flags, extack);
d4ead6b3
DA		 3667 if (IS_ERR(rt))
3668 return PTR_ERR(rt);
6b9ea5a6 3669
d4ead6b3 3670 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
93531c67 3671 fib6_info_release(rt);
6b9ea5a6 3672
1da177e4
LT		 3673 return err;
3674}
3675
8d1c802b 3676static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
1da177e4 3677{
afb1d4b5 3678 struct net *net = info->nl_net;
c71099ac 3679 struct fib6_table *table;
afb1d4b5 3680 int err;
1da177e4 3681
421842ed 3682 if (rt == net->ipv6.fib6_null_entry) {
6825a26c
G		 3683 err = -ENOENT;
3684 goto out;
3685 }
6c813a72 3686
93c2fb25 3687 table = rt->fib6_table;
66f5d6ce 3688 spin_lock_bh(&table->tb6_lock);
86872cb5 3689 err = fib6_del(rt, info);
66f5d6ce 3690 spin_unlock_bh(&table->tb6_lock);
1da177e4 3691
6825a26c 3692out:
93531c67 3693 fib6_info_release(rt);
1da177e4
LT		 3694 return err;
3695}
3696
8d1c802b 3697int ip6_del_rt(struct net *net, struct fib6_info *rt)
e0a1ad73 3698{
afb1d4b5
DA		 3699 struct nl_info info = { .nl_net = net };
3700
528c4ceb 3701 return __ip6_del_rt(rt, &info);
e0a1ad73
TG		 3702}
3703
8d1c802b 3704static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
0ae81335
DA		 3705{
3706 struct nl_info *info = &cfg->fc_nlinfo;
e3330039 3707 struct net *net = info->nl_net;
16a16cd3 3708 struct sk_buff *skb = NULL;
0ae81335 3709 struct fib6_table *table;
e3330039 3710 int err = -ENOENT;
0ae81335 3711
421842ed 3712 if (rt == net->ipv6.fib6_null_entry)
e3330039 3713 goto out_put;
93c2fb25 3714 table = rt->fib6_table;
66f5d6ce 3715 spin_lock_bh(&table->tb6_lock);
0ae81335 3716
93c2fb25 3717 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
8d1c802b 3718 struct fib6_info *sibling, *next_sibling;
0ae81335 3719
16a16cd3
DA		 3720 /* prefer to send a single notification with all hops */
3721 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3722 if (skb) {
3723 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3724
d4ead6b3 3725 if (rt6_fill_node(net, skb, rt, NULL,
16a16cd3
DA		 3726 NULL, NULL, 0, RTM_DELROUTE,
3727 info->portid, seq, 0) < 0) {
3728 kfree_skb(skb);
3729 skb = NULL;
3730 } else
3731 info->skip_notify = 1;
3732 }
3733
2881fd61
IS		 3734 info->skip_notify_kernel = 1;
3735 call_fib6_multipath_entry_notifiers(net,
3736 FIB_EVENT_ENTRY_DEL,
3737 rt,
3738 rt->fib6_nsiblings,
3739 NULL);
0ae81335 3740 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25
DA		 3741 &rt->fib6_siblings,
3742 fib6_siblings) {
0ae81335
DA		 3743 err = fib6_del(sibling, info);
3744 if (err)
e3330039 3745 goto out_unlock;
0ae81335
DA		 3746 }
3747 }
3748
3749 err = fib6_del(rt, info);
e3330039 3750out_unlock:
66f5d6ce 3751 spin_unlock_bh(&table->tb6_lock);
e3330039 3752out_put:
93531c67 3753 fib6_info_release(rt);
16a16cd3
DA		 3754
3755 if (skb) {
e3330039 3756 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
16a16cd3
DA		 3757 info->nlh, gfp_any());
3758 }
0ae81335
DA		 3759 return err;
3760}
3761
0fa6efc5 3762static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
23fb93a4
DA		 3763{
3764 int rc = -ESRCH;
3765
3766 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3767 goto out;
3768
3769 if (cfg->fc_flags & RTF_GATEWAY &&
3770 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3771 goto out;
761f6026
XL		 3772
3773 rc = rt6_remove_exception_rt(rt);
23fb93a4
DA		 3774out:
3775 return rc;
3776}
3777
0fa6efc5
DA		 3778static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3779 struct fib6_nh *nh)
3780{
3781 struct fib6_result res = {
3782 .f6i = rt,
3783 .nh = nh,
3784 };
3785 struct rt6_info *rt_cache;
3786
3787 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3788 if (rt_cache)
3789 return __ip6_del_cached_rt(rt_cache, cfg);
3790
3791 return 0;
3792}
3793
5b98324e
DA		 3794struct fib6_nh_del_cached_rt_arg {
3795 struct fib6_config *cfg;
3796 struct fib6_info *f6i;
3797};
3798
3799static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
3800{
3801 struct fib6_nh_del_cached_rt_arg *arg = _arg;
3802 int rc;
3803
3804 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
3805 return rc != -ESRCH ? rc : 0;
3806}
3807
3808static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
3809{
3810 struct fib6_nh_del_cached_rt_arg arg = {
3811 .cfg = cfg,
3812 .f6i = f6i
3813 };
3814
3815 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
3816}
3817
333c4301
DA		 3818static int ip6_route_del(struct fib6_config *cfg,
3819 struct netlink_ext_ack *extack)
1da177e4 3820{
c71099ac 3821 struct fib6_table *table;
8d1c802b 3822 struct fib6_info *rt;
1da177e4 3823 struct fib6_node *fn;
1da177e4
LT		 3824 int err = -ESRCH;
3825
5578689a 3826 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
d5d531cb
DA		 3827 if (!table) {
3828 NL_SET_ERR_MSG(extack, "FIB table does not exist");
c71099ac 3829 return err;
d5d531cb 3830 }
c71099ac 3831
66f5d6ce 3832 rcu_read_lock();
1da177e4 3833
c71099ac 3834 fn = fib6_locate(&table->tb6_root,
86872cb5 3835 &cfg->fc_dst, cfg->fc_dst_len,
38fbeeee 3836 &cfg->fc_src, cfg->fc_src_len,
2b760fcf 3837 !(cfg->fc_flags & RTF_CACHE));
1ab1457c 3838
1da177e4 3839 if (fn) {
66f5d6ce 3840 for_each_fib6_node_rt_rcu(fn) {
ad1601ae
DA		 3841 struct fib6_nh *nh;
3842
3401bfb1
SB		 3843 if (rt->nh && cfg->fc_nh_id &&
3844 rt->nh->id != cfg->fc_nh_id)
5b98324e 3845 continue;
23fb93a4 3846
5b98324e
DA		 3847 if (cfg->fc_flags & RTF_CACHE) {
3848 int rc = 0;
3849
3850 if (rt->nh) {
3851 rc = ip6_del_cached_rt_nh(cfg, rt);
3852 } else if (cfg->fc_nh_id) {
3853 continue;
3854 } else {
3855 nh = rt->fib6_nh;
3856 rc = ip6_del_cached_rt(cfg, rt, nh);
3857 }
0fa6efc5
DA		 3858 if (rc != -ESRCH) {
3859 rcu_read_unlock();
3860 return rc;
23fb93a4
DA		 3861 }
3862 continue;
2b760fcf 3863 }
ad1601ae 3864
5b98324e
DA		 3865 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
3866 continue;
3867 if (cfg->fc_protocol &&
3868 cfg->fc_protocol != rt->fib6_protocol)
3869 continue;
3870
3871 if (rt->nh) {
3872 if (!fib6_info_hold_safe(rt))
3873 continue;
3874 rcu_read_unlock();
3875
3876 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3877 }
3878 if (cfg->fc_nh_id)
3879 continue;
3880
3881 nh = rt->fib6_nh;
86872cb5 3882 if (cfg->fc_ifindex &&
ad1601ae
DA		 3883 (!nh->fib_nh_dev ||
3884 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
1da177e4 3885 continue;
86872cb5 3886 if (cfg->fc_flags & RTF_GATEWAY &&
ad1601ae 3887 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
1da177e4 3888 continue;
e873e4b9
WW		 3889 if (!fib6_info_hold_safe(rt))
3890 continue;
66f5d6ce 3891 rcu_read_unlock();
1da177e4 3892
0ae81335
DA		 3893 /* if gateway was specified only delete the one hop */
3894 if (cfg->fc_flags & RTF_GATEWAY)
3895 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3896
3897 return __ip6_del_rt_siblings(rt, cfg);
1da177e4
LT		 3898 }
3899 }
66f5d6ce 3900 rcu_read_unlock();
1da177e4
LT		 3901
3902 return err;
3903}
3904
6700c270 3905static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
a6279458 3906{
a6279458 3907 struct netevent_redirect netevent;
e8599ff4 3908 struct rt6_info *rt, *nrt = NULL;
85bd05de 3909 struct fib6_result res = {};
e8599ff4
DM		 3910 struct ndisc_options ndopts;
3911 struct inet6_dev *in6_dev;
3912 struct neighbour *neigh;
71bcdba0 3913 struct rd_msg *msg;
6e157b6a
DM		 3914 int optlen, on_link;
3915 u8 *lladdr;
e8599ff4 3916
29a3cad5 3917 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
71bcdba0 3918 optlen -= sizeof(*msg);
e8599ff4
DM		 3919
3920 if (optlen < 0) {
6e157b6a 3921 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
e8599ff4
DM		 3922 return;
3923 }
3924
71bcdba0 3925 msg = (struct rd_msg *)icmp6_hdr(skb);
e8599ff4 3926
71bcdba0 3927 if (ipv6_addr_is_multicast(&msg->dest)) {
6e157b6a 3928 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
e8599ff4
DM		 3929 return;
3930 }
3931
6e157b6a 3932 on_link = 0;
71bcdba0 3933 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
e8599ff4 3934 on_link = 1;
71bcdba0 3935 } else if (ipv6_addr_type(&msg->target) !=
e8599ff4 3936 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
6e157b6a 3937 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
e8599ff4
DM		 3938 return;
3939 }
3940
3941 in6_dev = __in6_dev_get(skb->dev);
3942 if (!in6_dev)
3943 return;
3944 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
3945 return;
3946
3947 /* RFC2461 8.1:
3948 * The IP source address of the Redirect MUST be the same as the current
3949 * first-hop router for the specified ICMP Destination Address.
3950 */
3951
f997c55c 3952 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
e8599ff4
DM		 3953 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
3954 return;
3955 }
6e157b6a
DM		 3956
3957 lladdr = NULL;
e8599ff4
DM		 3958 if (ndopts.nd_opts_tgt_lladdr) {
3959 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
3960 skb->dev);
3961 if (!lladdr) {
3962 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
3963 return;
3964 }
3965 }
3966
6e157b6a 3967 rt = (struct rt6_info *) dst;
ec13ad1d 3968 if (rt->rt6i_flags & RTF_REJECT) {
6e157b6a 3969 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
e8599ff4 3970 return;
6e157b6a 3971 }
e8599ff4 3972
6e157b6a
DM		 3973 /* Redirect received -> path was valid.
3974 * Look, redirects are sent only in response to data packets,
3975 * so that this nexthop apparently is reachable. --ANK
3976 */
0dec879f 3977 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
a6279458 3978
71bcdba0 3979 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
6e157b6a
DM		 3980 if (!neigh)
3981 return;
a6279458 3982
1da177e4
LT		 3983 /*
3984 * We have finally decided to accept it.
3985 */
3986
f997c55c 3987 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1da177e4
LT		 3988 NEIGH_UPDATE_F_WEAK_OVERRIDE|
3989 NEIGH_UPDATE_F_OVERRIDE|
3990 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
f997c55c
AA		 3991 NEIGH_UPDATE_F_ISROUTER)),
3992 NDISC_REDIRECT, &ndopts);
1da177e4 3993
4d85cd0c 3994 rcu_read_lock();
85bd05de 3995 res.f6i = rcu_dereference(rt->from);
ff24e498 3996 if (!res.f6i)
886b7a50 3997 goto out;
8a14e46f 3998
49d5b8ef
DA		 3999 if (res.f6i->nh) {
4000 struct fib6_nh_match_arg arg = {
4001 .dev = dst->dev,
4002 .gw = &rt->rt6i_gateway,
4003 };
4004
4005 nexthop_for_each_fib6_nh(res.f6i->nh,
4006 fib6_nh_find_match, &arg);
4007
4008 /* fib6_info uses a nexthop that does not have fib6_nh
4009 * using the dst->dev. Should be impossible
4010 */
4011 if (!arg.match)
4012 goto out;
4013 res.nh = arg.match;
4014 } else {
4015 res.nh = res.f6i->fib6_nh;
4016 }
4017
7d21fec9
DA		 4018 res.fib6_flags = res.f6i->fib6_flags;
4019 res.fib6_type = res.f6i->fib6_type;
85bd05de 4020 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
38308473 4021 if (!nrt)
1da177e4
LT		 4022 goto out;
4023
4024 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4025 if (on_link)
4026 nrt->rt6i_flags &= ~RTF_GATEWAY;
4027
4e3fd7a0 4028 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1da177e4 4029
886b7a50 4030 /* rt6_insert_exception() will take care of duplicated exceptions */
5012f0a5 4031 if (rt6_insert_exception(nrt, &res)) {
2b760fcf
WW		 4032 dst_release_immediate(&nrt->dst);
4033 goto out;
4034 }
1da177e4 4035
d8d1f30b
CG		 4036 netevent.old = &rt->dst;
4037 netevent.new = &nrt->dst;
71bcdba0 4038 netevent.daddr = &msg->dest;
60592833 4039 netevent.neigh = neigh;
8d71740c
TT		 4040 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4041
1da177e4 4042out:
886b7a50 4043 rcu_read_unlock();
e8599ff4 4044 neigh_release(neigh);
6e157b6a
DM		 4045}
4046
70ceb4f5 4047#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 4048static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 4049 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 4050 const struct in6_addr *gwaddr,
4051 struct net_device *dev)
70ceb4f5 4052{
830218c1
DA		 4053 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4054 int ifindex = dev->ifindex;
70ceb4f5 4055 struct fib6_node *fn;
8d1c802b 4056 struct fib6_info *rt = NULL;
c71099ac
TG		 4057 struct fib6_table *table;
4058
830218c1 4059 table = fib6_get_table(net, tb_id);
38308473 4060 if (!table)
c71099ac 4061 return NULL;
70ceb4f5 4062
66f5d6ce 4063 rcu_read_lock();
38fbeeee 4064 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
70ceb4f5
YH		 4065 if (!fn)
4066 goto out;
4067
66f5d6ce 4068 for_each_fib6_node_rt_rcu(fn) {
f88d8ea6
DA		 4069 /* these routes do not use nexthops */
4070 if (rt->nh)
4071 continue;
1cf844c7 4072 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
70ceb4f5 4073 continue;
2b2450ca 4074 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
1cf844c7 4075 !rt->fib6_nh->fib_nh_gw_family)
70ceb4f5 4076 continue;
1cf844c7 4077 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
70ceb4f5 4078 continue;
e873e4b9
WW		 4079 if (!fib6_info_hold_safe(rt))
4080 continue;
70ceb4f5
YH		 4081 break;
4082 }
4083out:
66f5d6ce 4084 rcu_read_unlock();
70ceb4f5
YH		 4085 return rt;
4086}
4087
8d1c802b 4088static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 4089 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 4090 const struct in6_addr *gwaddr,
4091 struct net_device *dev,
95c96174 4092 unsigned int pref)
70ceb4f5 4093{
86872cb5 4094 struct fib6_config cfg = {
238fc7ea 4095 .fc_metric = IP6_RT_PRIO_USER,
830218c1 4096 .fc_ifindex = dev->ifindex,
86872cb5
TG		 4097 .fc_dst_len = prefixlen,
4098 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4099 RTF_UP | RTF_PREF(pref),
b91d5329 4100 .fc_protocol = RTPROT_RA,
e8478e80 4101 .fc_type = RTN_UNICAST,
15e47304 4102 .fc_nlinfo.portid = 0,
efa2cea0
DL		 4103 .fc_nlinfo.nlh = NULL,
4104 .fc_nlinfo.nl_net = net,
86872cb5
TG		 4105 };
4106
830218c1 4107 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
4e3fd7a0
AD		 4108 cfg.fc_dst = *prefix;
4109 cfg.fc_gateway = *gwaddr;
70ceb4f5 4110
e317da96
YH		 4111 /* We should treat it as a default route if prefix length is 0. */
4112 if (!prefixlen)
86872cb5 4113 cfg.fc_flags |= RTF_DEFAULT;
70ceb4f5 4114
acb54e3c 4115 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
70ceb4f5 4116
830218c1 4117 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
70ceb4f5
YH		 4118}
4119#endif
4120
8d1c802b 4121struct fib6_info *rt6_get_dflt_router(struct net *net,
afb1d4b5
DA		 4122 const struct in6_addr *addr,
4123 struct net_device *dev)
1ab1457c 4124{
830218c1 4125 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
8d1c802b 4126 struct fib6_info *rt;
c71099ac 4127 struct fib6_table *table;
1da177e4 4128
afb1d4b5 4129 table = fib6_get_table(net, tb_id);
38308473 4130 if (!table)
c71099ac 4131 return NULL;
1da177e4 4132
66f5d6ce
WW		 4133 rcu_read_lock();
4134 for_each_fib6_node_rt_rcu(&table->tb6_root) {
f88d8ea6 4135 struct fib6_nh *nh;
ad1601ae 4136
f88d8ea6
DA		 4137 /* RA routes do not use nexthops */
4138 if (rt->nh)
4139 continue;
4140
4141 nh = rt->fib6_nh;
ad1601ae 4142 if (dev == nh->fib_nh_dev &&
93c2fb25 4143 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
ad1601ae 4144 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
1da177e4
LT		 4145 break;
4146 }
e873e4b9
WW		 4147 if (rt && !fib6_info_hold_safe(rt))
4148 rt = NULL;
66f5d6ce 4149 rcu_read_unlock();
1da177e4
LT		 4150 return rt;
4151}
4152
8d1c802b 4153struct fib6_info *rt6_add_dflt_router(struct net *net,
afb1d4b5 4154 const struct in6_addr *gwaddr,
ebacaaa0
YH		 4155 struct net_device *dev,
4156 unsigned int pref)
1da177e4 4157{
86872cb5 4158 struct fib6_config cfg = {
ca254490 4159 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
238fc7ea 4160 .fc_metric = IP6_RT_PRIO_USER,
86872cb5
TG		 4161 .fc_ifindex = dev->ifindex,
4162 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4163 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
b91d5329 4164 .fc_protocol = RTPROT_RA,
e8478e80 4165 .fc_type = RTN_UNICAST,
15e47304 4166 .fc_nlinfo.portid = 0,
5578689a 4167 .fc_nlinfo.nlh = NULL,
afb1d4b5 4168 .fc_nlinfo.nl_net = net,
86872cb5 4169 };
1da177e4 4170
4e3fd7a0 4171 cfg.fc_gateway = *gwaddr;
1da177e4 4172
acb54e3c 4173 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
830218c1
DA		 4174 struct fib6_table *table;
4175
4176 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4177 if (table)
4178 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4179 }
1da177e4 4180
afb1d4b5 4181 return rt6_get_dflt_router(net, gwaddr, dev);
1da177e4
LT		 4182}
4183
afb1d4b5
DA		 4184static void __rt6_purge_dflt_routers(struct net *net,
4185 struct fib6_table *table)
1da177e4 4186{
8d1c802b 4187 struct fib6_info *rt;
1da177e4
LT		 4188
4189restart:
66f5d6ce
WW		 4190 rcu_read_lock();
4191 for_each_fib6_node_rt_rcu(&table->tb6_root) {
dcd1f572
DA		 4192 struct net_device *dev = fib6_info_nh_dev(rt);
4193 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4194
93c2fb25 4195 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
e873e4b9
WW		 4196 (!idev || idev->cnf.accept_ra != 2) &&
4197 fib6_info_hold_safe(rt)) {
93531c67
DA		 4198 rcu_read_unlock();
4199 ip6_del_rt(net, rt);
1da177e4
LT		 4200 goto restart;
4201 }
4202 }
66f5d6ce 4203 rcu_read_unlock();
830218c1
DA		 4204
4205 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4206}
4207
4208void rt6_purge_dflt_routers(struct net *net)
4209{
4210 struct fib6_table *table;
4211 struct hlist_head *head;
4212 unsigned int h;
4213
4214 rcu_read_lock();
4215
4216 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4217 head = &net->ipv6.fib_table_hash[h];
4218 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4219 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
afb1d4b5 4220 __rt6_purge_dflt_routers(net, table);
830218c1
DA		 4221 }
4222 }
4223
4224 rcu_read_unlock();
1da177e4
LT		 4225}
4226
5578689a
DL		 4227static void rtmsg_to_fib6_config(struct net *net,
4228 struct in6_rtmsg *rtmsg,
86872cb5
TG		 4229 struct fib6_config *cfg)
4230{
8823a3ac
MŻ		 4231 *cfg = (struct fib6_config){
4232 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4233 : RT6_TABLE_MAIN,
4234 .fc_ifindex = rtmsg->rtmsg_ifindex,
67f69513 4235 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
8823a3ac
MŻ		 4236 .fc_expires = rtmsg->rtmsg_info,
4237 .fc_dst_len = rtmsg->rtmsg_dst_len,
4238 .fc_src_len = rtmsg->rtmsg_src_len,
4239 .fc_flags = rtmsg->rtmsg_flags,
4240 .fc_type = rtmsg->rtmsg_type,
4241
4242 .fc_nlinfo.nl_net = net,
4243
4244 .fc_dst = rtmsg->rtmsg_dst,
4245 .fc_src = rtmsg->rtmsg_src,
4246 .fc_gateway = rtmsg->rtmsg_gateway,
4247 };
86872cb5
TG		 4248}
4249
5578689a 4250int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1da177e4 4251{
86872cb5 4252 struct fib6_config cfg;
1da177e4
LT		 4253 struct in6_rtmsg rtmsg;
4254 int err;
4255
67ba4152 4256 switch (cmd) {
1da177e4
LT		 4257 case SIOCADDRT: /* Add a route */
4258 case SIOCDELRT: /* Delete a route */
af31f412 4259 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4
LT		 4260 return -EPERM;
4261 err = copy_from_user(&rtmsg, arg,
4262 sizeof(struct in6_rtmsg));
4263 if (err)
4264 return -EFAULT;
86872cb5 4265
5578689a 4266 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
86872cb5 4267
1da177e4
LT		 4268 rtnl_lock();
4269 switch (cmd) {
4270 case SIOCADDRT:
acb54e3c 4271 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
1da177e4
LT		 4272 break;
4273 case SIOCDELRT:
333c4301 4274 err = ip6_route_del(&cfg, NULL);
1da177e4
LT		 4275 break;
4276 default:
4277 err = -EINVAL;
4278 }
4279 rtnl_unlock();
4280
4281 return err;
3ff50b79 4282 }
1da177e4
LT		 4283
4284 return -EINVAL;
4285}
4286
4287/*
4288 * Drop the packet on the floor
4289 */
4290
d5fdd6ba 4291static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1da177e4 4292{
adf30907 4293 struct dst_entry *dst = skb_dst(skb);
1d3fd8a1
SS		 4294 struct net *net = dev_net(dst->dev);
4295 struct inet6_dev *idev;
4296 int type;
4297
4298 if (netif_is_l3_master(skb->dev) &&
4299 dst->dev == net->loopback_dev)
4300 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4301 else
4302 idev = ip6_dst_idev(dst);
4303
612f09e8
YH		 4304 switch (ipstats_mib_noroutes) {
4305 case IPSTATS_MIB_INNOROUTES:
0660e03f 4306 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
45bb0060 4307 if (type == IPV6_ADDR_ANY) {
1d3fd8a1 4308 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
612f09e8
YH		 4309 break;
4310 }
4311 /* FALLTHROUGH */
4312 case IPSTATS_MIB_OUTNOROUTES:
1d3fd8a1 4313 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
612f09e8
YH		 4314 break;
4315 }
1d3fd8a1
SS		 4316
4317 /* Start over by dropping the dst for l3mdev case */
4318 if (netif_is_l3_master(skb->dev))
4319 skb_dst_drop(skb);
4320
3ffe533c 4321 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1da177e4
LT		 4322 kfree_skb(skb);
4323 return 0;
4324}
4325
9ce8ade0
TG		 4326static int ip6_pkt_discard(struct sk_buff *skb)
4327{
612f09e8 4328 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4329}
4330
ede2059d 4331static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
1da177e4 4332{
adf30907 4333 skb->dev = skb_dst(skb)->dev;
612f09e8 4334 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1da177e4
LT		 4335}
4336
9ce8ade0
TG		 4337static int ip6_pkt_prohibit(struct sk_buff *skb)
4338{
612f09e8 4339 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4340}
4341
ede2059d 4342static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
9ce8ade0 4343{
adf30907 4344 skb->dev = skb_dst(skb)->dev;
612f09e8 4345 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
9ce8ade0
TG		 4346}
4347
1da177e4
LT		 4348/*
4349 * Allocate a dst for local (unicast / anycast) address.
4350 */
4351
360a9887
DA		 4352struct fib6_info *addrconf_f6i_alloc(struct net *net,
4353 struct inet6_dev *idev,
4354 const struct in6_addr *addr,
4355 bool anycast, gfp_t gfp_flags)
1da177e4 4356{
c7a1ce39
DA		 4357 struct fib6_config cfg = {
4358 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4359 .fc_ifindex = idev->dev->ifindex,
4360 .fc_flags = RTF_UP | RTF_ADDRCONF | RTF_NONEXTHOP,
4361 .fc_dst = *addr,
4362 .fc_dst_len = 128,
4363 .fc_protocol = RTPROT_KERNEL,
4364 .fc_nlinfo.nl_net = net,
4365 .fc_ignore_dev_down = true,
4366 };
1da177e4 4367
e8478e80 4368 if (anycast) {
c7a1ce39
DA		 4369 cfg.fc_type = RTN_ANYCAST;
4370 cfg.fc_flags |= RTF_ANYCAST;
e8478e80 4371 } else {
c7a1ce39
DA		 4372 cfg.fc_type = RTN_LOCAL;
4373 cfg.fc_flags |= RTF_LOCAL;
e8478e80 4374 }
1da177e4 4375
c7a1ce39 4376 return ip6_route_info_create(&cfg, gfp_flags, NULL);
1da177e4
LT		 4377}
4378
c3968a85
DW		 4379/* remove deleted ip from prefsrc entries */
4380struct arg_dev_net_ip {
4381 struct net_device *dev;
4382 struct net *net;
4383 struct in6_addr *addr;
4384};
4385
8d1c802b 4386static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
c3968a85
DW		 4387{
4388 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4389 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4390 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4391
f88d8ea6
DA		 4392 if (!rt->nh &&
4393 ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
421842ed 4394 rt != net->ipv6.fib6_null_entry &&
93c2fb25 4395 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
60006a48 4396 spin_lock_bh(&rt6_exception_lock);
c3968a85 4397 /* remove prefsrc entry */
93c2fb25 4398 rt->fib6_prefsrc.plen = 0;
60006a48 4399 spin_unlock_bh(&rt6_exception_lock);
c3968a85
DW		 4400 }
4401 return 0;
4402}
4403
4404void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4405{
4406 struct net *net = dev_net(ifp->idev->dev);
4407 struct arg_dev_net_ip adni = {
4408 .dev = ifp->idev->dev,
4409 .net = net,
4410 .addr = &ifp->addr,
4411 };
0c3584d5 4412 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
c3968a85
DW		 4413}
4414
2b2450ca 4415#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
be7a010d
DJ		 4416
4417/* Remove routers and update dst entries when gateway turn into host. */
8d1c802b 4418static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
be7a010d
DJ		 4419{
4420 struct in6_addr *gateway = (struct in6_addr *)arg;
f88d8ea6
DA		 4421 struct fib6_nh *nh;
4422
4423 /* RA routes do not use nexthops */
4424 if (rt->nh)
4425 return 0;
be7a010d 4426
f88d8ea6 4427 nh = rt->fib6_nh;
93c2fb25 4428 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
cc5c073a 4429 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
be7a010d 4430 return -1;
b16cb459
WW		 4431
4432 /* Further clean up cached routes in exception table.
4433 * This is needed because cached route may have a different
4434 * gateway than its 'parent' in the case of an ip redirect.
4435 */
cc5c073a 4436 fib6_nh_exceptions_clean_tohost(nh, gateway);
b16cb459 4437
be7a010d
DJ		 4438 return 0;
4439}
4440
4441void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4442{
4443 fib6_clean_all(net, fib6_clean_tohost, gateway);
4444}
4445
2127d95a
IS		 4446struct arg_netdev_event {
4447 const struct net_device *dev;
4c981e28 4448 union {
ecc5663c 4449 unsigned char nh_flags;
4c981e28
IS		 4450 unsigned long event;
4451 };
2127d95a
IS		 4452};
4453
8d1c802b 4454static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
d7dedee1 4455{
8d1c802b 4456 struct fib6_info *iter;
d7dedee1
IS		 4457 struct fib6_node *fn;
4458
93c2fb25
DA		 4459 fn = rcu_dereference_protected(rt->fib6_node,
4460 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4461 iter = rcu_dereference_protected(fn->leaf,
93c2fb25 4462 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4463 while (iter) {
93c2fb25 4464 if (iter->fib6_metric == rt->fib6_metric &&
33bd5ac5 4465 rt6_qualify_for_ecmp(iter))
d7dedee1 4466 return iter;
8fb11a9a 4467 iter = rcu_dereference_protected(iter->fib6_next,
93c2fb25 4468 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1
IS		 4469 }
4470
4471 return NULL;
4472}
4473
f88d8ea6 4474/* only called for fib entries with builtin fib6_nh */
8d1c802b 4475static bool rt6_is_dead(const struct fib6_info *rt)
d7dedee1 4476{
1cf844c7
DA		 4477 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4478 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4479 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
d7dedee1
IS		 4480 return true;
4481
4482 return false;
4483}
4484
8d1c802b 4485static int rt6_multipath_total_weight(const struct fib6_info *rt)
d7dedee1 4486{
8d1c802b 4487 struct fib6_info *iter;
d7dedee1
IS		 4488 int total = 0;
4489
4490 if (!rt6_is_dead(rt))
1cf844c7 4491 total += rt->fib6_nh->fib_nh_weight;
d7dedee1 4492
93c2fb25 4493 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
d7dedee1 4494 if (!rt6_is_dead(iter))
1cf844c7 4495 total += iter->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4496 }
4497
4498 return total;
4499}
4500
8d1c802b 4501static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
d7dedee1
IS		 4502{
4503 int upper_bound = -1;
4504
4505 if (!rt6_is_dead(rt)) {
1cf844c7 4506 *weight += rt->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4507 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4508 total) - 1;
4509 }
1cf844c7 4510 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
d7dedee1
IS		 4511}
4512
8d1c802b 4513static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
d7dedee1 4514{
8d1c802b 4515 struct fib6_info *iter;
d7dedee1
IS		 4516 int weight = 0;
4517
4518 rt6_upper_bound_set(rt, &weight, total);
4519
93c2fb25 4520 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
d7dedee1
IS		 4521 rt6_upper_bound_set(iter, &weight, total);
4522}
4523
8d1c802b 4524void rt6_multipath_rebalance(struct fib6_info *rt)
d7dedee1 4525{
8d1c802b 4526 struct fib6_info *first;
d7dedee1
IS		 4527 int total;
4528
4529 /* In case the entire multipath route was marked for flushing,
4530 * then there is no need to rebalance upon the removal of every
4531 * sibling route.
4532 */
93c2fb25 4533 if (!rt->fib6_nsiblings || rt->should_flush)
d7dedee1
IS		 4534 return;
4535
4536 /* During lookup routes are evaluated in order, so we need to
4537 * make sure upper bounds are assigned from the first sibling
4538 * onwards.
4539 */
4540 first = rt6_multipath_first_sibling(rt);
4541 if (WARN_ON_ONCE(!first))
4542 return;
4543
4544 total = rt6_multipath_total_weight(first);
4545 rt6_multipath_upper_bound_set(first, total);
4546}
4547
8d1c802b 4548static int fib6_ifup(struct fib6_info *rt, void *p_arg)
2127d95a
IS		 4549{
4550 const struct arg_netdev_event *arg = p_arg;
7aef6859 4551 struct net *net = dev_net(arg->dev);
2127d95a 4552
f88d8ea6 4553 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
1cf844c7
DA		 4554 rt->fib6_nh->fib_nh_dev == arg->dev) {
4555 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
7aef6859 4556 fib6_update_sernum_upto_root(net, rt);
d7dedee1 4557 rt6_multipath_rebalance(rt);
1de178ed 4558 }
2127d95a
IS		 4559
4560 return 0;
4561}
4562
ecc5663c 4563void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
2127d95a
IS		 4564{
4565 struct arg_netdev_event arg = {
4566 .dev = dev,
6802f3ad
IS		 4567 {
4568 .nh_flags = nh_flags,
4569 },
2127d95a
IS		 4570 };
4571
4572 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4573 arg.nh_flags |= RTNH_F_LINKDOWN;
4574
4575 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4576}
4577
f88d8ea6 4578/* only called for fib entries with inline fib6_nh */
8d1c802b 4579static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
1de178ed
IS		 4580 const struct net_device *dev)
4581{
8d1c802b 4582 struct fib6_info *iter;
1de178ed 4583
1cf844c7 4584 if (rt->fib6_nh->fib_nh_dev == dev)
1de178ed 4585 return true;
93c2fb25 4586 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7 4587 if (iter->fib6_nh->fib_nh_dev == dev)
1de178ed
IS		 4588 return true;
4589
4590 return false;
4591}
4592
8d1c802b 4593static void rt6_multipath_flush(struct fib6_info *rt)
1de178ed 4594{
8d1c802b 4595 struct fib6_info *iter;
1de178ed
IS		 4596
4597 rt->should_flush = 1;
93c2fb25 4598 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1de178ed
IS		 4599 iter->should_flush = 1;
4600}
4601
8d1c802b 4602static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
1de178ed
IS		 4603 const struct net_device *down_dev)
4604{
8d1c802b 4605 struct fib6_info *iter;
1de178ed
IS		 4606 unsigned int dead = 0;
4607
1cf844c7
DA		 4608 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4609 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed 4610 dead++;
93c2fb25 4611 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4612 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4613 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed
IS		 4614 dead++;
4615
4616 return dead;
4617}
4618
8d1c802b 4619static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
1de178ed 4620 const struct net_device *dev,
ecc5663c 4621 unsigned char nh_flags)
1de178ed 4622{
8d1c802b 4623 struct fib6_info *iter;
1de178ed 4624
1cf844c7
DA		 4625 if (rt->fib6_nh->fib_nh_dev == dev)
4626 rt->fib6_nh->fib_nh_flags |= nh_flags;
93c2fb25 4627 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4628 if (iter->fib6_nh->fib_nh_dev == dev)
4629 iter->fib6_nh->fib_nh_flags |= nh_flags;
1de178ed
IS		 4630}
4631
a1a22c12 4632/* called with write lock held for table with rt */
8d1c802b 4633static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
1da177e4 4634{
4c981e28
IS		 4635 const struct arg_netdev_event *arg = p_arg;
4636 const struct net_device *dev = arg->dev;
7aef6859 4637 struct net *net = dev_net(dev);
8ed67789 4638
f88d8ea6 4639 if (rt == net->ipv6.fib6_null_entry || rt->nh)
27c6fa73
IS		 4640 return 0;
4641
4642 switch (arg->event) {
4643 case NETDEV_UNREGISTER:
1cf844c7 4644 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
27c6fa73 4645 case NETDEV_DOWN:
1de178ed 4646 if (rt->should_flush)
27c6fa73 4647 return -1;
93c2fb25 4648 if (!rt->fib6_nsiblings)
1cf844c7 4649 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
1de178ed
IS		 4650 if (rt6_multipath_uses_dev(rt, dev)) {
4651 unsigned int count;
4652
4653 count = rt6_multipath_dead_count(rt, dev);
93c2fb25 4654 if (rt->fib6_nsiblings + 1 == count) {
1de178ed
IS		 4655 rt6_multipath_flush(rt);
4656 return -1;
4657 }
4658 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4659 RTNH_F_LINKDOWN);
7aef6859 4660 fib6_update_sernum(net, rt);
d7dedee1 4661 rt6_multipath_rebalance(rt);
1de178ed
IS		 4662 }
4663 return -2;
27c6fa73 4664 case NETDEV_CHANGE:
1cf844c7 4665 if (rt->fib6_nh->fib_nh_dev != dev ||
93c2fb25 4666 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
27c6fa73 4667 break;
1cf844c7 4668 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
d7dedee1 4669 rt6_multipath_rebalance(rt);
27c6fa73 4670 break;
2b241361 4671 }
c159d30c 4672
1da177e4
LT		 4673 return 0;
4674}
4675
27c6fa73 4676void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
1da177e4 4677{
4c981e28 4678 struct arg_netdev_event arg = {
8ed67789 4679 .dev = dev,
6802f3ad
IS		 4680 {
4681 .event = event,
4682 },
8ed67789 4683 };
7c6bb7d2 4684 struct net *net = dev_net(dev);
8ed67789 4685
7c6bb7d2
DA		 4686 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4687 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4688 else
4689 fib6_clean_all(net, fib6_ifdown, &arg);
4c981e28
IS		 4690}
4691
4692void rt6_disable_ip(struct net_device *dev, unsigned long event)
4693{
4694 rt6_sync_down_dev(dev, event);
4695 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4696 neigh_ifdown(&nd_tbl, dev);
1da177e4
LT		 4697}
4698
95c96174 4699struct rt6_mtu_change_arg {
1da177e4 4700 struct net_device *dev;
95c96174 4701 unsigned int mtu;
c0b220cf 4702 struct fib6_info *f6i;
1da177e4
LT		 4703};
4704
cc5c073a 4705static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
c0b220cf
DA		 4706{
4707 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
cc5c073a 4708 struct fib6_info *f6i = arg->f6i;
c0b220cf
DA		 4709
4710 /* For administrative MTU increase, there is no way to discover
4711 * IPv6 PMTU increase, so PMTU increase should be updated here.
4712 * Since RFC 1981 doesn't include administrative MTU increase
4713 * update PMTU increase is a MUST. (i.e. jumbo frame)
4714 */
4715 if (nh->fib_nh_dev == arg->dev) {
4716 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4717 u32 mtu = f6i->fib6_pmtu;
4718
4719 if (mtu >= arg->mtu ||
4720 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4721 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4722
4723 spin_lock_bh(&rt6_exception_lock);
cc5c073a 4724 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
c0b220cf
DA		 4725 spin_unlock_bh(&rt6_exception_lock);
4726 }
4727
4728 return 0;
4729}
4730
4731static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
1da177e4
LT		 4732{
4733 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4734 struct inet6_dev *idev;
4735
4736 /* In IPv6 pmtu discovery is not optional,
4737 so that RTAX_MTU lock cannot disable it.
4738 We still use this lock to block changes
4739 caused by addrconf/ndisc.
4740 */
4741
4742 idev = __in6_dev_get(arg->dev);
38308473 4743 if (!idev)
1da177e4
LT		 4744 return 0;
4745
c0b220cf
DA		 4746 if (fib6_metric_locked(f6i, RTAX_MTU))
4747 return 0;
d4ead6b3 4748
c0b220cf 4749 arg->f6i = f6i;
2d44234b
DA		 4750 if (f6i->nh) {
4751 /* fib6_nh_mtu_change only returns 0, so this is safe */
4752 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4753 arg);
4754 }
4755
1cf844c7 4756 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
1da177e4
LT		 4757}
4758
95c96174 4759void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
1da177e4 4760{
c71099ac
TG		 4761 struct rt6_mtu_change_arg arg = {
4762 .dev = dev,
4763 .mtu = mtu,
4764 };
1da177e4 4765
0c3584d5 4766 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
1da177e4
LT		 4767}
4768
ef7c79ed 4769static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
75425657 4770 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
5176f91e 4771 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
aa8f8778 4772 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
86872cb5 4773 [RTA_OIF] = { .type = NLA_U32 },
ab364a6f 4774 [RTA_IIF] = { .type = NLA_U32 },
86872cb5
TG		 4775 [RTA_PRIORITY] = { .type = NLA_U32 },
4776 [RTA_METRICS] = { .type = NLA_NESTED },
51ebd318 4777 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
c78ba6d6 4778 [RTA_PREF] = { .type = NLA_U8 },
19e42e45
RP		 4779 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4780 [RTA_ENCAP] = { .type = NLA_NESTED },
32bc201e 4781 [RTA_EXPIRES] = { .type = NLA_U32 },
622ec2c9 4782 [RTA_UID] = { .type = NLA_U32 },
3b45a410 4783 [RTA_MARK] = { .type = NLA_U32 },
aa8f8778 4784 [RTA_TABLE] = { .type = NLA_U32 },
eacb9384
RP		 4785 [RTA_IP_PROTO] = { .type = NLA_U8 },
4786 [RTA_SPORT] = { .type = NLA_U16 },
4787 [RTA_DPORT] = { .type = NLA_U16 },
5b98324e 4788 [RTA_NH_ID] = { .type = NLA_U32 },
86872cb5
TG		 4789};
4790
4791static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
333c4301
DA		 4792 struct fib6_config *cfg,
4793 struct netlink_ext_ack *extack)
1da177e4 4794{
86872cb5
TG		 4795 struct rtmsg *rtm;
4796 struct nlattr *tb[RTA_MAX+1];
c78ba6d6 4797 unsigned int pref;
86872cb5 4798 int err;
1da177e4 4799
8cb08174
JB		 4800 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
4801 rtm_ipv6_policy, extack);
86872cb5
TG		 4802 if (err < 0)
4803 goto errout;
1da177e4 4804
86872cb5
TG		 4805 err = -EINVAL;
4806 rtm = nlmsg_data(nlh);
86872cb5 4807
84db8407
MŻ		 4808 *cfg = (struct fib6_config){
4809 .fc_table = rtm->rtm_table,
4810 .fc_dst_len = rtm->rtm_dst_len,
4811 .fc_src_len = rtm->rtm_src_len,
4812 .fc_flags = RTF_UP,
4813 .fc_protocol = rtm->rtm_protocol,
4814 .fc_type = rtm->rtm_type,
4815
4816 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
4817 .fc_nlinfo.nlh = nlh,
4818 .fc_nlinfo.nl_net = sock_net(skb->sk),
4819 };
86872cb5 4820
ef2c7d7b
ND		 4821 if (rtm->rtm_type == RTN_UNREACHABLE ||
4822 rtm->rtm_type == RTN_BLACKHOLE ||
b4949ab2
ND		 4823 rtm->rtm_type == RTN_PROHIBIT ||
4824 rtm->rtm_type == RTN_THROW)
86872cb5
TG		 4825 cfg->fc_flags |= RTF_REJECT;
4826
ab79ad14
MŻ		 4827 if (rtm->rtm_type == RTN_LOCAL)
4828 cfg->fc_flags |= RTF_LOCAL;
4829
1f56a01f
MKL		 4830 if (rtm->rtm_flags & RTM_F_CLONED)
4831 cfg->fc_flags |= RTF_CACHE;
4832
fc1e64e1
DA		 4833 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4834
5b98324e
DA		 4835 if (tb[RTA_NH_ID]) {
4836 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
4837 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
4838 NL_SET_ERR_MSG(extack,
4839 "Nexthop specification and nexthop id are mutually exclusive");
4840 goto errout;
4841 }
4842 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
4843 }
4844
86872cb5 4845 if (tb[RTA_GATEWAY]) {
67b61f6c 4846 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
86872cb5 4847 cfg->fc_flags |= RTF_GATEWAY;
1da177e4 4848 }
e3818541
DA		 4849 if (tb[RTA_VIA]) {
4850 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
4851 goto errout;
4852 }
86872cb5
TG		 4853
4854 if (tb[RTA_DST]) {
4855 int plen = (rtm->rtm_dst_len + 7) >> 3;
4856
4857 if (nla_len(tb[RTA_DST]) < plen)
4858 goto errout;
4859
4860 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1da177e4 4861 }
86872cb5
TG		 4862
4863 if (tb[RTA_SRC]) {
4864 int plen = (rtm->rtm_src_len + 7) >> 3;
4865
4866 if (nla_len(tb[RTA_SRC]) < plen)
4867 goto errout;
4868
4869 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1da177e4 4870 }
86872cb5 4871
c3968a85 4872 if (tb[RTA_PREFSRC])
67b61f6c 4873 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
c3968a85 4874
86872cb5
TG		 4875 if (tb[RTA_OIF])
4876 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4877
4878 if (tb[RTA_PRIORITY])
4879 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4880
4881 if (tb[RTA_METRICS]) {
4882 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4883 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
1da177e4 4884 }
86872cb5
TG		 4885
4886 if (tb[RTA_TABLE])
4887 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4888
51ebd318
ND		 4889 if (tb[RTA_MULTIPATH]) {
4890 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4891 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
9ed59592
DA		 4892
4893 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
c255bd68 4894 cfg->fc_mp_len, extack);
9ed59592
DA		 4895 if (err < 0)
4896 goto errout;
51ebd318
ND		 4897 }
4898
c78ba6d6
LR		 4899 if (tb[RTA_PREF]) {
4900 pref = nla_get_u8(tb[RTA_PREF]);
4901 if (pref != ICMPV6_ROUTER_PREF_LOW &&
4902 pref != ICMPV6_ROUTER_PREF_HIGH)
4903 pref = ICMPV6_ROUTER_PREF_MEDIUM;
4904 cfg->fc_flags |= RTF_PREF(pref);
4905 }
4906
19e42e45
RP		 4907 if (tb[RTA_ENCAP])
4908 cfg->fc_encap = tb[RTA_ENCAP];
4909
9ed59592 4910 if (tb[RTA_ENCAP_TYPE]) {
19e42e45
RP		 4911 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
4912
c255bd68 4913 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
9ed59592
DA		 4914 if (err < 0)
4915 goto errout;
4916 }
4917
32bc201e
XL		 4918 if (tb[RTA_EXPIRES]) {
4919 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
4920
4921 if (addrconf_finite_timeout(timeout)) {
4922 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
4923 cfg->fc_flags |= RTF_EXPIRES;
4924 }
4925 }
4926
86872cb5
TG		 4927 err = 0;
4928errout:
4929 return err;
1da177e4
LT		 4930}
4931
6b9ea5a6 4932struct rt6_nh {
8d1c802b 4933 struct fib6_info *fib6_info;
6b9ea5a6 4934 struct fib6_config r_cfg;
6b9ea5a6
RP		 4935 struct list_head next;
4936};
4937
d4ead6b3
DA		 4938static int ip6_route_info_append(struct net *net,
4939 struct list_head *rt6_nh_list,
8d1c802b
DA		 4940 struct fib6_info *rt,
4941 struct fib6_config *r_cfg)
6b9ea5a6
RP		 4942{
4943 struct rt6_nh *nh;
6b9ea5a6
RP		 4944 int err = -EEXIST;
4945
4946 list_for_each_entry(nh, rt6_nh_list, next) {
8d1c802b
DA		 4947 /* check if fib6_info already exists */
4948 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
6b9ea5a6
RP		 4949 return err;
4950 }
4951
4952 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
4953 if (!nh)
4954 return -ENOMEM;
8d1c802b 4955 nh->fib6_info = rt;
6b9ea5a6
RP		 4956 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
4957 list_add_tail(&nh->next, rt6_nh_list);
4958
4959 return 0;
4960}
4961
8d1c802b
DA		 4962static void ip6_route_mpath_notify(struct fib6_info *rt,
4963 struct fib6_info *rt_last,
3b1137fe
DA		 4964 struct nl_info *info,
4965 __u16 nlflags)
4966{
4967 /* if this is an APPEND route, then rt points to the first route
4968 * inserted and rt_last points to last route inserted. Userspace
4969 * wants a consistent dump of the route which starts at the first
4970 * nexthop. Since sibling routes are always added at the end of
4971 * the list, find the first sibling of the last route appended
4972 */
93c2fb25
DA		 4973 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
4974 rt = list_first_entry(&rt_last->fib6_siblings,
8d1c802b 4975 struct fib6_info,
93c2fb25 4976 fib6_siblings);
3b1137fe
DA		 4977 }
4978
4979 if (rt)
4980 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
4981}
4982
333c4301
DA		 4983static int ip6_route_multipath_add(struct fib6_config *cfg,
4984 struct netlink_ext_ack *extack)
51ebd318 4985{
8d1c802b 4986 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
3b1137fe 4987 struct nl_info *info = &cfg->fc_nlinfo;
ebee3cad 4988 enum fib_event_type event_type;
51ebd318
ND		 4989 struct fib6_config r_cfg;
4990 struct rtnexthop *rtnh;
8d1c802b 4991 struct fib6_info *rt;
6b9ea5a6
RP		 4992 struct rt6_nh *err_nh;
4993 struct rt6_nh *nh, *nh_safe;
3b1137fe 4994 __u16 nlflags;
51ebd318
ND		 4995 int remaining;
4996 int attrlen;
6b9ea5a6
RP		 4997 int err = 1;
4998 int nhn = 0;
4999 int replace = (cfg->fc_nlinfo.nlh &&
5000 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5001 LIST_HEAD(rt6_nh_list);
51ebd318 5002
3b1137fe
DA		 5003 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5004 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5005 nlflags |= NLM_F_APPEND;
5006
35f1b4e9 5007 remaining = cfg->fc_mp_len;
51ebd318 5008 rtnh = (struct rtnexthop *)cfg->fc_mp;
51ebd318 5009
6b9ea5a6 5010 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
8d1c802b 5011 * fib6_info structs per nexthop
6b9ea5a6 5012 */
51ebd318
ND		 5013 while (rtnh_ok(rtnh, remaining)) {
5014 memcpy(&r_cfg, cfg, sizeof(*cfg));
5015 if (rtnh->rtnh_ifindex)
5016 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5017
5018 attrlen = rtnh_attrlen(rtnh);
5019 if (attrlen > 0) {
5020 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5021
5022 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5023 if (nla) {
67b61f6c 5024 r_cfg.fc_gateway = nla_get_in6_addr(nla);
51ebd318
ND		 5025 r_cfg.fc_flags |= RTF_GATEWAY;
5026 }
19e42e45
RP		 5027 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5028 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5029 if (nla)
5030 r_cfg.fc_encap_type = nla_get_u16(nla);
51ebd318 5031 }
6b9ea5a6 5032
68e2ffde 5033 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
acb54e3c 5034 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
8c5b83f0
RP		 5035 if (IS_ERR(rt)) {
5036 err = PTR_ERR(rt);
5037 rt = NULL;
6b9ea5a6 5038 goto cleanup;
8c5b83f0 5039 }
b5d2d75e
DA		 5040 if (!rt6_qualify_for_ecmp(rt)) {
5041 err = -EINVAL;
5042 NL_SET_ERR_MSG(extack,
5043 "Device only routes can not be added for IPv6 using the multipath API.");
5044 fib6_info_release(rt);
5045 goto cleanup;
5046 }
6b9ea5a6 5047
1cf844c7 5048 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
398958ae 5049
d4ead6b3
DA		 5050 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5051 rt, &r_cfg);
51ebd318 5052 if (err) {
93531c67 5053 fib6_info_release(rt);
6b9ea5a6
RP		 5054 goto cleanup;
5055 }
5056
5057 rtnh = rtnh_next(rtnh, &remaining);
5058 }
5059
9eee3b49
IS		 5060 if (list_empty(&rt6_nh_list)) {
5061 NL_SET_ERR_MSG(extack,
5062 "Invalid nexthop configuration - no valid nexthops");
5063 return -EINVAL;
5064 }
5065
3b1137fe
DA		 5066 /* for add and replace send one notification with all nexthops.
5067 * Skip the notification in fib6_add_rt2node and send one with
5068 * the full route when done
5069 */
5070 info->skip_notify = 1;
5071
ebee3cad
IS		 5072 /* For add and replace, send one notification with all nexthops. For
5073 * append, send one notification with all appended nexthops.
5074 */
5075 info->skip_notify_kernel = 1;
5076
6b9ea5a6
RP		 5077 err_nh = NULL;
5078 list_for_each_entry(nh, &rt6_nh_list, next) {
8d1c802b
DA		 5079 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5080 fib6_info_release(nh->fib6_info);
93531c67 5081
f7225172
DA		 5082 if (!err) {
5083 /* save reference to last route successfully inserted */
5084 rt_last = nh->fib6_info;
5085
5086 /* save reference to first route for notification */
5087 if (!rt_notif)
5088 rt_notif = nh->fib6_info;
5089 }
3b1137fe 5090
8d1c802b
DA		 5091 /* nh->fib6_info is used or freed at this point, reset to NULL*/
5092 nh->fib6_info = NULL;
6b9ea5a6
RP		 5093 if (err) {
5094 if (replace && nhn)
a5a82d84
JK		 5095 NL_SET_ERR_MSG_MOD(extack,
5096 "multipath route replace failed (check consistency of installed routes)");
6b9ea5a6
RP		 5097 err_nh = nh;
5098 goto add_errout;
51ebd318 5099 }
6b9ea5a6 5100
1a72418b 5101 /* Because each route is added like a single route we remove
27596472
MK		 5102 * these flags after the first nexthop: if there is a collision,
5103 * we have already failed to add the first nexthop:
5104 * fib6_add_rt2node() has rejected it; when replacing, old
5105 * nexthops have been replaced by first new, the rest should
5106 * be added to it.
1a72418b 5107 */
27596472
MK		 5108 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5109 NLM_F_REPLACE);
6b9ea5a6
RP		 5110 nhn++;
5111 }
5112
ebee3cad
IS		 5113 event_type = replace ? FIB_EVENT_ENTRY_REPLACE : FIB_EVENT_ENTRY_ADD;
5114 err = call_fib6_multipath_entry_notifiers(info->nl_net, event_type,
5115 rt_notif, nhn - 1, extack);
5116 if (err) {
5117 /* Delete all the siblings that were just added */
5118 err_nh = NULL;
5119 goto add_errout;
5120 }
5121
3b1137fe
DA		 5122 /* success ... tell user about new route */
5123 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
6b9ea5a6
RP		 5124 goto cleanup;
5125
5126add_errout:
3b1137fe
DA		 5127 /* send notification for routes that were added so that
5128 * the delete notifications sent by ip6_route_del are
5129 * coherent
5130 */
5131 if (rt_notif)
5132 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5133
6b9ea5a6
RP		 5134 /* Delete routes that were already added */
5135 list_for_each_entry(nh, &rt6_nh_list, next) {
5136 if (err_nh == nh)
5137 break;
333c4301 5138 ip6_route_del(&nh->r_cfg, extack);
6b9ea5a6
RP		 5139 }
5140
5141cleanup:
5142 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
8d1c802b
DA		 5143 if (nh->fib6_info)
5144 fib6_info_release(nh->fib6_info);
6b9ea5a6
RP		 5145 list_del(&nh->next);
5146 kfree(nh);
5147 }
5148
5149 return err;
5150}
5151
333c4301
DA		 5152static int ip6_route_multipath_del(struct fib6_config *cfg,
5153 struct netlink_ext_ack *extack)
6b9ea5a6
RP		 5154{
5155 struct fib6_config r_cfg;
5156 struct rtnexthop *rtnh;
5157 int remaining;
5158 int attrlen;
5159 int err = 1, last_err = 0;
5160
5161 remaining = cfg->fc_mp_len;
5162 rtnh = (struct rtnexthop *)cfg->fc_mp;
5163
5164 /* Parse a Multipath Entry */
5165 while (rtnh_ok(rtnh, remaining)) {
5166 memcpy(&r_cfg, cfg, sizeof(*cfg));
5167 if (rtnh->rtnh_ifindex)
5168 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5169
5170 attrlen = rtnh_attrlen(rtnh);
5171 if (attrlen > 0) {
5172 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5173
5174 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5175 if (nla) {
5176 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
5177 r_cfg.fc_flags |= RTF_GATEWAY;
5178 }
5179 }
333c4301 5180 err = ip6_route_del(&r_cfg, extack);
6b9ea5a6
RP		 5181 if (err)
5182 last_err = err;
5183
51ebd318
ND		 5184 rtnh = rtnh_next(rtnh, &remaining);
5185 }
5186
5187 return last_err;
5188}
5189
c21ef3e3
DA		 5190static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5191 struct netlink_ext_ack *extack)
1da177e4 5192{
86872cb5
TG		 5193 struct fib6_config cfg;
5194 int err;
1da177e4 5195
333c4301 5196 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 5197 if (err < 0)
5198 return err;
5199
5b98324e
DA		 5200 if (cfg.fc_nh_id &&
5201 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5202 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5203 return -EINVAL;
5204 }
5205
51ebd318 5206 if (cfg.fc_mp)
333c4301 5207 return ip6_route_multipath_del(&cfg, extack);
0ae81335
DA		 5208 else {
5209 cfg.fc_delete_all_nh = 1;
333c4301 5210 return ip6_route_del(&cfg, extack);
0ae81335 5211 }
1da177e4
LT		 5212}
5213
c21ef3e3
DA		 5214static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5215 struct netlink_ext_ack *extack)
1da177e4 5216{
86872cb5
TG		 5217 struct fib6_config cfg;
5218 int err;
1da177e4 5219
333c4301 5220 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 5221 if (err < 0)
5222 return err;
5223
67f69513
DA		 5224 if (cfg.fc_metric == 0)
5225 cfg.fc_metric = IP6_RT_PRIO_USER;
5226
51ebd318 5227 if (cfg.fc_mp)
333c4301 5228 return ip6_route_multipath_add(&cfg, extack);
51ebd318 5229 else
acb54e3c 5230 return ip6_route_add(&cfg, GFP_KERNEL, extack);
1da177e4
LT		 5231}
5232
a1b7a1f0
DA		 5233/* add the overhead of this fib6_nh to nexthop_len */
5234static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
339bf98f 5235{
a1b7a1f0 5236 int *nexthop_len = arg;
beb1afac 5237
a1b7a1f0
DA		 5238 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5239 + NLA_ALIGN(sizeof(struct rtnexthop))
5240 + nla_total_size(16); /* RTA_GATEWAY */
5241
5242 if (nh->fib_nh_lws) {
5243 /* RTA_ENCAP_TYPE */
5244 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5245 /* RTA_ENCAP */
5246 *nexthop_len += nla_total_size(2);
5247 }
f88d8ea6 5248
a1b7a1f0
DA		 5249 return 0;
5250}
beb1afac 5251
a1b7a1f0
DA		 5252static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5253{
5254 int nexthop_len;
5255
5256 if (f6i->nh) {
5257 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5258 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5259 &nexthop_len);
5260 } else {
5261 struct fib6_nh *nh = f6i->fib6_nh;
5262
5263 nexthop_len = 0;
5264 if (f6i->fib6_nsiblings) {
5265 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
5266 + NLA_ALIGN(sizeof(struct rtnexthop))
5267 + nla_total_size(16) /* RTA_GATEWAY */
5268 + lwtunnel_get_encap_size(nh->fib_nh_lws);
5269
5270 nexthop_len *= f6i->fib6_nsiblings;
5271 }
5272 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
beb1afac
DA		 5273 }
5274
339bf98f
TG		 5275 return NLMSG_ALIGN(sizeof(struct rtmsg))
5276 + nla_total_size(16) /* RTA_SRC */
5277 + nla_total_size(16) /* RTA_DST */
5278 + nla_total_size(16) /* RTA_GATEWAY */
5279 + nla_total_size(16) /* RTA_PREFSRC */
5280 + nla_total_size(4) /* RTA_TABLE */
5281 + nla_total_size(4) /* RTA_IIF */
5282 + nla_total_size(4) /* RTA_OIF */
5283 + nla_total_size(4) /* RTA_PRIORITY */
6a2b9ce0 5284 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
ea697639 5285 + nla_total_size(sizeof(struct rta_cacheinfo))
c78ba6d6 5286 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
19e42e45 5287 + nla_total_size(1) /* RTA_PREF */
beb1afac
DA		 5288 + nexthop_len;
5289}
5290
f88d8ea6
DA		 5291static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5292 unsigned char *flags)
5293{
5294 if (nexthop_is_multipath(nh)) {
5295 struct nlattr *mp;
5296
5297 mp = nla_nest_start(skb, RTA_MULTIPATH);
5298 if (!mp)
5299 goto nla_put_failure;
5300
5301 if (nexthop_mpath_fill_node(skb, nh))
5302 goto nla_put_failure;
5303
5304 nla_nest_end(skb, mp);
5305 } else {
5306 struct fib6_nh *fib6_nh;
5307
5308 fib6_nh = nexthop_fib6_nh(nh);
5309 if (fib_nexthop_info(skb, &fib6_nh->nh_common,
5310 flags, false) < 0)
5311 goto nla_put_failure;
5312 }
5313
5314 return 0;
5315
5316nla_put_failure:
5317 return -EMSGSIZE;
5318}
5319
d4ead6b3 5320static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 5321 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 5322 struct in6_addr *dest, struct in6_addr *src,
15e47304 5323 int iif, int type, u32 portid, u32 seq,
f8cfe2ce 5324 unsigned int flags)
1da177e4 5325{
22d0bd82
XL		 5326 struct rt6_info *rt6 = (struct rt6_info *)dst;
5327 struct rt6key *rt6_dst, *rt6_src;
5328 u32 *pmetrics, table, rt6_flags;
f88d8ea6 5329 unsigned char nh_flags = 0;
2d7202bf 5330 struct nlmsghdr *nlh;
22d0bd82 5331 struct rtmsg *rtm;
d4ead6b3 5332 long expires = 0;
1da177e4 5333
15e47304 5334 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
38308473 5335 if (!nlh)
26932566 5336 return -EMSGSIZE;
2d7202bf 5337
22d0bd82
XL		 5338 if (rt6) {
5339 rt6_dst = &rt6->rt6i_dst;
5340 rt6_src = &rt6->rt6i_src;
5341 rt6_flags = rt6->rt6i_flags;
5342 } else {
5343 rt6_dst = &rt->fib6_dst;
5344 rt6_src = &rt->fib6_src;
5345 rt6_flags = rt->fib6_flags;
5346 }
5347
2d7202bf 5348 rtm = nlmsg_data(nlh);
1da177e4 5349 rtm->rtm_family = AF_INET6;
22d0bd82
XL		 5350 rtm->rtm_dst_len = rt6_dst->plen;
5351 rtm->rtm_src_len = rt6_src->plen;
1da177e4 5352 rtm->rtm_tos = 0;
93c2fb25
DA		 5353 if (rt->fib6_table)
5354 table = rt->fib6_table->tb6_id;
c71099ac 5355 else
9e762a4a 5356 table = RT6_TABLE_UNSPEC;
97f0082a 5357 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
c78679e8
DM		 5358 if (nla_put_u32(skb, RTA_TABLE, table))
5359 goto nla_put_failure;
e8478e80
DA		 5360
5361 rtm->rtm_type = rt->fib6_type;
1da177e4
LT		 5362 rtm->rtm_flags = 0;
5363 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
93c2fb25 5364 rtm->rtm_protocol = rt->fib6_protocol;
1da177e4 5365
22d0bd82 5366 if (rt6_flags & RTF_CACHE)
1da177e4
LT		 5367 rtm->rtm_flags |= RTM_F_CLONED;
5368
d4ead6b3
DA		 5369 if (dest) {
5370 if (nla_put_in6_addr(skb, RTA_DST, dest))
c78679e8 5371 goto nla_put_failure;
1ab1457c 5372 rtm->rtm_dst_len = 128;
1da177e4 5373 } else if (rtm->rtm_dst_len)
22d0bd82 5374 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
c78679e8 5375 goto nla_put_failure;
1da177e4
LT		 5376#ifdef CONFIG_IPV6_SUBTREES
5377 if (src) {
930345ea 5378 if (nla_put_in6_addr(skb, RTA_SRC, src))
c78679e8 5379 goto nla_put_failure;
1ab1457c 5380 rtm->rtm_src_len = 128;
c78679e8 5381 } else if (rtm->rtm_src_len &&
22d0bd82 5382 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
c78679e8 5383 goto nla_put_failure;
1da177e4 5384#endif
7bc570c8
YH		 5385 if (iif) {
5386#ifdef CONFIG_IPV6_MROUTE
22d0bd82 5387 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
fd61c6ba
DA		 5388 int err = ip6mr_get_route(net, skb, rtm, portid);
5389
5390 if (err == 0)
5391 return 0;
5392 if (err < 0)
5393 goto nla_put_failure;
7bc570c8
YH		 5394 } else
5395#endif
c78679e8
DM		 5396 if (nla_put_u32(skb, RTA_IIF, iif))
5397 goto nla_put_failure;
d4ead6b3 5398 } else if (dest) {
1da177e4 5399 struct in6_addr saddr_buf;
d4ead6b3 5400 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
930345ea 5401 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 5402 goto nla_put_failure;
1da177e4 5403 }
2d7202bf 5404
93c2fb25 5405 if (rt->fib6_prefsrc.plen) {
c3968a85 5406 struct in6_addr saddr_buf;
93c2fb25 5407 saddr_buf = rt->fib6_prefsrc.addr;
930345ea 5408 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 5409 goto nla_put_failure;
c3968a85
DW		 5410 }
5411
d4ead6b3
DA		 5412 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5413 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
2d7202bf
TG		 5414 goto nla_put_failure;
5415
93c2fb25 5416 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
c78679e8 5417 goto nla_put_failure;
8253947e 5418
beb1afac
DA		 5419 /* For multipath routes, walk the siblings list and add
5420 * each as a nexthop within RTA_MULTIPATH.
5421 */
22d0bd82
XL		 5422 if (rt6) {
5423 if (rt6_flags & RTF_GATEWAY &&
5424 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5425 goto nla_put_failure;
5426
5427 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5428 goto nla_put_failure;
5429 } else if (rt->fib6_nsiblings) {
8d1c802b 5430 struct fib6_info *sibling, *next_sibling;
beb1afac
DA		 5431 struct nlattr *mp;
5432
ae0be8de 5433 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
beb1afac
DA		 5434 if (!mp)
5435 goto nla_put_failure;
5436
1cf844c7
DA		 5437 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5438 rt->fib6_nh->fib_nh_weight) < 0)
beb1afac
DA		 5439 goto nla_put_failure;
5440
5441 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25 5442 &rt->fib6_siblings, fib6_siblings) {
1cf844c7
DA		 5443 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5444 sibling->fib6_nh->fib_nh_weight) < 0)
beb1afac
DA		 5445 goto nla_put_failure;
5446 }
5447
5448 nla_nest_end(skb, mp);
f88d8ea6
DA		 5449 } else if (rt->nh) {
5450 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5451 goto nla_put_failure;
ecc5663c 5452
f88d8ea6
DA		 5453 if (nexthop_is_blackhole(rt->nh))
5454 rtm->rtm_type = RTN_BLACKHOLE;
5455
5456 if (rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5457 goto nla_put_failure;
5458
5459 rtm->rtm_flags |= nh_flags;
5460 } else {
1cf844c7 5461 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common,
ecc5663c 5462 &nh_flags, false) < 0)
beb1afac 5463 goto nla_put_failure;
ecc5663c
DA		 5464
5465 rtm->rtm_flags |= nh_flags;
beb1afac
DA		 5466 }
5467
22d0bd82 5468 if (rt6_flags & RTF_EXPIRES) {
14895687
DA		 5469 expires = dst ? dst->expires : rt->expires;
5470 expires -= jiffies;
5471 }
69cdf8f9 5472
d4ead6b3 5473 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
e3703b3d 5474 goto nla_put_failure;
2d7202bf 5475
22d0bd82 5476 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
c78ba6d6
LR		 5477 goto nla_put_failure;
5478
19e42e45 5479
053c095a
JB		 5480 nlmsg_end(skb, nlh);
5481 return 0;
2d7202bf
TG		 5482
5483nla_put_failure:
26932566
PM		 5484 nlmsg_cancel(skb, nlh);
5485 return -EMSGSIZE;
1da177e4
LT		 5486}
5487
2c170e07
DA		 5488static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5489{
5490 const struct net_device *dev = arg;
5491
5492 if (nh->fib_nh_dev == dev)
5493 return 1;
5494
5495 return 0;
5496}
5497
13e38901
DA		 5498static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5499 const struct net_device *dev)
5500{
2c170e07
DA		 5501 if (f6i->nh) {
5502 struct net_device *_dev = (struct net_device *)dev;
5503
5504 return !!nexthop_for_each_fib6_nh(f6i->nh,
5505 fib6_info_nh_uses_dev,
5506 _dev);
5507 }
5508
1cf844c7 5509 if (f6i->fib6_nh->fib_nh_dev == dev)
13e38901
DA		 5510 return true;
5511
5512 if (f6i->fib6_nsiblings) {
5513 struct fib6_info *sibling, *next_sibling;
5514
5515 list_for_each_entry_safe(sibling, next_sibling,
5516 &f6i->fib6_siblings, fib6_siblings) {
1cf844c7 5517 if (sibling->fib6_nh->fib_nh_dev == dev)
13e38901
DA		 5518 return true;
5519 }
5520 }
5521
5522 return false;
5523}
5524
1e47b483
SB		 5525struct fib6_nh_exception_dump_walker {
5526 struct rt6_rtnl_dump_arg *dump;
5527 struct fib6_info *rt;
5528 unsigned int flags;
5529 unsigned int skip;
5530 unsigned int count;
5531};
5532
5533static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5534{
5535 struct fib6_nh_exception_dump_walker *w = arg;
5536 struct rt6_rtnl_dump_arg *dump = w->dump;
5537 struct rt6_exception_bucket *bucket;
5538 struct rt6_exception *rt6_ex;
5539 int i, err;
5540
5541 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5542 if (!bucket)
5543 return 0;
5544
5545 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5546 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5547 if (w->skip) {
5548 w->skip--;
5549 continue;
5550 }
5551
5552 /* Expiration of entries doesn't bump sernum, insertion
5553 * does. Removal is triggered by insertion, so we can
5554 * rely on the fact that if entries change between two
5555 * partial dumps, this node is scanned again completely,
5556 * see rt6_insert_exception() and fib6_dump_table().
5557 *
5558 * Count expired entries we go through as handled
5559 * entries that we'll skip next time, in case of partial
5560 * node dump. Otherwise, if entries expire meanwhile,
5561 * we'll skip the wrong amount.
5562 */
5563 if (rt6_check_expired(rt6_ex->rt6i)) {
5564 w->count++;
5565 continue;
5566 }
5567
5568 err = rt6_fill_node(dump->net, dump->skb, w->rt,
5569 &rt6_ex->rt6i->dst, NULL, NULL, 0,
5570 RTM_NEWROUTE,
5571 NETLINK_CB(dump->cb->skb).portid,
5572 dump->cb->nlh->nlmsg_seq, w->flags);
5573 if (err)
5574 return err;
5575
5576 w->count++;
5577 }
5578 bucket++;
5579 }
5580
5581 return 0;
5582}
5583
bf9a8a06 5584/* Return -1 if done with node, number of handled routes on partial dump */
1e47b483 5585int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
1da177e4
LT		 5586{
5587 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
13e38901
DA		 5588 struct fib_dump_filter *filter = &arg->filter;
5589 unsigned int flags = NLM_F_MULTI;
1f17e2f2 5590 struct net *net = arg->net;
1e47b483 5591 int count = 0;
1f17e2f2 5592
421842ed 5593 if (rt == net->ipv6.fib6_null_entry)
bf9a8a06 5594 return -1;
1da177e4 5595
13e38901
DA		 5596 if ((filter->flags & RTM_F_PREFIX) &&
5597 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5598 /* success since this is not a prefix route */
bf9a8a06 5599 return -1;
13e38901 5600 }
1e47b483
SB		 5601 if (filter->filter_set &&
5602 ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5603 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5604 (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5605 return -1;
5606 }
5607
5608 if (filter->filter_set ||
5609 !filter->dump_routes || !filter->dump_exceptions) {
13e38901 5610 flags |= NLM_F_DUMP_FILTERED;
f8cfe2ce 5611 }
1da177e4 5612
1e47b483
SB		 5613 if (filter->dump_routes) {
5614 if (skip) {
5615 skip--;
5616 } else {
5617 if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5618 0, RTM_NEWROUTE,
5619 NETLINK_CB(arg->cb->skb).portid,
5620 arg->cb->nlh->nlmsg_seq, flags)) {
5621 return 0;
5622 }
5623 count++;
5624 }
5625 }
5626
5627 if (filter->dump_exceptions) {
5628 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5629 .rt = rt,
5630 .flags = flags,
5631 .skip = skip,
5632 .count = 0 };
5633 int err;
5634
5635 if (rt->nh) {
5636 err = nexthop_for_each_fib6_nh(rt->nh,
5637 rt6_nh_dump_exceptions,
5638 &w);
5639 } else {
5640 err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5641 }
5642
5643 if (err)
5644 return count += w.count;
5645 }
bf9a8a06
SB		 5646
5647 return -1;
1da177e4
LT		 5648}
5649
0eff0a27
JK		 5650static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5651 const struct nlmsghdr *nlh,
5652 struct nlattr **tb,
5653 struct netlink_ext_ack *extack)
5654{
5655 struct rtmsg *rtm;
5656 int i, err;
5657
5658 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5659 NL_SET_ERR_MSG_MOD(extack,
5660 "Invalid header for get route request");
5661 return -EINVAL;
5662 }
5663
5664 if (!netlink_strict_get_check(skb))
8cb08174
JB		 5665 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5666 rtm_ipv6_policy, extack);
0eff0a27
JK		 5667
5668 rtm = nlmsg_data(nlh);
5669 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5670 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5671 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5672 rtm->rtm_type) {
5673 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5674 return -EINVAL;
5675 }
5676 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5677 NL_SET_ERR_MSG_MOD(extack,
5678 "Invalid flags for get route request");
5679 return -EINVAL;
5680 }
5681
8cb08174
JB		 5682 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5683 rtm_ipv6_policy, extack);
0eff0a27
JK		 5684 if (err)
5685 return err;
5686
5687 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5688 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5689 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5690 return -EINVAL;
5691 }
5692
5693 for (i = 0; i <= RTA_MAX; i++) {
5694 if (!tb[i])
5695 continue;
5696
5697 switch (i) {
5698 case RTA_SRC:
5699 case RTA_DST:
5700 case RTA_IIF:
5701 case RTA_OIF:
5702 case RTA_MARK:
5703 case RTA_UID:
5704 case RTA_SPORT:
5705 case RTA_DPORT:
5706 case RTA_IP_PROTO:
5707 break;
5708 default:
5709 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
5710 return -EINVAL;
5711 }
5712 }
5713
5714 return 0;
5715}
5716
c21ef3e3
DA		 5717static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5718 struct netlink_ext_ack *extack)
1da177e4 5719{
3b1e0a65 5720 struct net *net = sock_net(in_skb->sk);
ab364a6f 5721 struct nlattr *tb[RTA_MAX+1];
18c3a61c 5722 int err, iif = 0, oif = 0;
a68886a6 5723 struct fib6_info *from;
18c3a61c 5724 struct dst_entry *dst;
ab364a6f 5725 struct rt6_info *rt;
1da177e4 5726 struct sk_buff *skb;
ab364a6f 5727 struct rtmsg *rtm;
744486d4 5728 struct flowi6 fl6 = {};
18c3a61c 5729 bool fibmatch;
1da177e4 5730
0eff0a27 5731 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
ab364a6f
TG		 5732 if (err < 0)
5733 goto errout;
1da177e4 5734
ab364a6f 5735 err = -EINVAL;
38b7097b
HFS		 5736 rtm = nlmsg_data(nlh);
5737 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
18c3a61c 5738 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
1da177e4 5739
ab364a6f
TG		 5740 if (tb[RTA_SRC]) {
5741 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
5742 goto errout;
5743
4e3fd7a0 5744 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
ab364a6f
TG		 5745 }
5746
5747 if (tb[RTA_DST]) {
5748 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
5749 goto errout;
5750
4e3fd7a0 5751 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
ab364a6f
TG		 5752 }
5753
5754 if (tb[RTA_IIF])
5755 iif = nla_get_u32(tb[RTA_IIF]);
5756
5757 if (tb[RTA_OIF])
72331bc0 5758 oif = nla_get_u32(tb[RTA_OIF]);
1da177e4 5759
2e47b291
LC		 5760 if (tb[RTA_MARK])
5761 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
5762
622ec2c9
LC		 5763 if (tb[RTA_UID])
5764 fl6.flowi6_uid = make_kuid(current_user_ns(),
5765 nla_get_u32(tb[RTA_UID]));
5766 else
5767 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
5768
eacb9384
RP		 5769 if (tb[RTA_SPORT])
5770 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
5771
5772 if (tb[RTA_DPORT])
5773 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
5774
5775 if (tb[RTA_IP_PROTO]) {
5776 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
5e1a99ea
HL		 5777 &fl6.flowi6_proto, AF_INET6,
5778 extack);
eacb9384
RP		 5779 if (err)
5780 goto errout;
5781 }
5782
1da177e4
LT		 5783 if (iif) {
5784 struct net_device *dev;
72331bc0
SL		 5785 int flags = 0;
5786
121622db
FW		 5787 rcu_read_lock();
5788
5789 dev = dev_get_by_index_rcu(net, iif);
1da177e4 5790 if (!dev) {
121622db 5791 rcu_read_unlock();
1da177e4 5792 err = -ENODEV;
ab364a6f 5793 goto errout;
1da177e4 5794 }
72331bc0
SL		 5795
5796 fl6.flowi6_iif = iif;
5797
5798 if (!ipv6_addr_any(&fl6.saddr))
5799 flags |= RT6_LOOKUP_F_HAS_SADDR;
5800
b75cc8f9 5801 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
121622db
FW		 5802
5803 rcu_read_unlock();
72331bc0
SL		 5804 } else {
5805 fl6.flowi6_oif = oif;
5806
58acfd71 5807 dst = ip6_route_output(net, NULL, &fl6);
18c3a61c
RP		 5808 }
5809
18c3a61c
RP		 5810
5811 rt = container_of(dst, struct rt6_info, dst);
5812 if (rt->dst.error) {
5813 err = rt->dst.error;
5814 ip6_rt_put(rt);
5815 goto errout;
1da177e4
LT		 5816 }
5817
9d6acb3b
WC		 5818 if (rt == net->ipv6.ip6_null_entry) {
5819 err = rt->dst.error;
5820 ip6_rt_put(rt);
5821 goto errout;
5822 }
5823
ab364a6f 5824 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
38308473 5825 if (!skb) {
94e187c0 5826 ip6_rt_put(rt);
ab364a6f
TG		 5827 err = -ENOBUFS;
5828 goto errout;
5829 }
1da177e4 5830
d8d1f30b 5831 skb_dst_set(skb, &rt->dst);
a68886a6
DA		 5832
5833 rcu_read_lock();
5834 from = rcu_dereference(rt->from);
886b7a50
MKL		 5835 if (from) {
5836 if (fibmatch)
5837 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
5838 iif, RTM_NEWROUTE,
5839 NETLINK_CB(in_skb).portid,
5840 nlh->nlmsg_seq, 0);
5841 else
5842 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
5843 &fl6.saddr, iif, RTM_NEWROUTE,
5844 NETLINK_CB(in_skb).portid,
5845 nlh->nlmsg_seq, 0);
5846 } else {
5847 err = -ENETUNREACH;
5848 }
a68886a6
DA		 5849 rcu_read_unlock();
5850
1da177e4 5851 if (err < 0) {
ab364a6f
TG		 5852 kfree_skb(skb);
5853 goto errout;
1da177e4
LT		 5854 }
5855
15e47304 5856 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
ab364a6f 5857errout:
1da177e4 5858 return err;
1da177e4
LT		 5859}
5860
8d1c802b 5861void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
37a1d361 5862 unsigned int nlm_flags)
1da177e4
LT		 5863{
5864 struct sk_buff *skb;
5578689a 5865 struct net *net = info->nl_net;
528c4ceb
DL		 5866 u32 seq;
5867 int err;
5868
5869 err = -ENOBUFS;
38308473 5870 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
86872cb5 5871
19e42e45 5872 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
38308473 5873 if (!skb)
21713ebc
TG		 5874 goto errout;
5875
d4ead6b3
DA		 5876 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
5877 event, info->portid, seq, nlm_flags);
26932566
PM		 5878 if (err < 0) {
5879 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
5880 WARN_ON(err == -EMSGSIZE);
5881 kfree_skb(skb);
5882 goto errout;
5883 }
15e47304 5884 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
1ce85fe4
PNA		 5885 info->nlh, gfp_any());
5886 return;
21713ebc
TG		 5887errout:
5888 if (err < 0)
5578689a 5889 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
1da177e4
LT		 5890}
5891
19a3b7ee
DA		 5892void fib6_rt_update(struct net *net, struct fib6_info *rt,
5893 struct nl_info *info)
5894{
5895 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
5896 struct sk_buff *skb;
5897 int err = -ENOBUFS;
5898
5899 /* call_fib6_entry_notifiers will be removed when in-kernel notifier
5900 * is implemented and supported for nexthop objects
5901 */
5902 call_fib6_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, rt, NULL);
5903
5904 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
5905 if (!skb)
5906 goto errout;
5907
5908 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
5909 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
5910 if (err < 0) {
5911 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
5912 WARN_ON(err == -EMSGSIZE);
5913 kfree_skb(skb);
5914 goto errout;
5915 }
5916 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
5917 info->nlh, gfp_any());
5918 return;
5919errout:
5920 if (err < 0)
5921 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
5922}
5923
8ed67789 5924static int ip6_route_dev_notify(struct notifier_block *this,
351638e7 5925 unsigned long event, void *ptr)
8ed67789 5926{
351638e7 5927 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
c346dca1 5928 struct net *net = dev_net(dev);
8ed67789 5929
242d3a49
WC		 5930 if (!(dev->flags & IFF_LOOPBACK))
5931 return NOTIFY_OK;
5932
5933 if (event == NETDEV_REGISTER) {
1cf844c7 5934 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
d8d1f30b 5935 net->ipv6.ip6_null_entry->dst.dev = dev;
8ed67789
DL		 5936 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
5937#ifdef CONFIG_IPV6_MULTIPLE_TABLES
d8d1f30b 5938 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
8ed67789 5939 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
d8d1f30b 5940 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
8ed67789 5941 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
242d3a49 5942#endif
76da0704
WC		 5943 } else if (event == NETDEV_UNREGISTER &&
5944 dev->reg_state != NETREG_UNREGISTERED) {
5945 /* NETDEV_UNREGISTER could be fired for multiple times by
5946 * netdev_wait_allrefs(). Make sure we only call this once.
5947 */
12d94a80 5948 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
242d3a49 5949#ifdef CONFIG_IPV6_MULTIPLE_TABLES
12d94a80
ED		 5950 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
5951 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
8ed67789
DL		 5952#endif
5953 }
5954
5955 return NOTIFY_OK;
5956}
5957
1da177e4
LT		 5958/*
5959 * /proc
5960 */
5961
5962#ifdef CONFIG_PROC_FS
1da177e4
LT		 5963static int rt6_stats_seq_show(struct seq_file *seq, void *v)
5964{
69ddb805 5965 struct net *net = (struct net *)seq->private;
1da177e4 5966 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
69ddb805
DL		 5967 net->ipv6.rt6_stats->fib_nodes,
5968 net->ipv6.rt6_stats->fib_route_nodes,
81eb8447 5969 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
69ddb805
DL		 5970 net->ipv6.rt6_stats->fib_rt_entries,
5971 net->ipv6.rt6_stats->fib_rt_cache,
fc66f95c 5972 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
69ddb805 5973 net->ipv6.rt6_stats->fib_discarded_routes);
1da177e4
LT		 5974
5975 return 0;
5976}
1da177e4
LT		 5977#endif /* CONFIG_PROC_FS */
5978
5979#ifdef CONFIG_SYSCTL
5980
1da177e4 5981static
fe2c6338 5982int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
1da177e4
LT		 5983 void __user *buffer, size_t *lenp, loff_t *ppos)
5984{
c486da34
LAG		 5985 struct net *net;
5986 int delay;
f0fb9b28 5987 int ret;
c486da34 5988 if (!write)
1da177e4 5989 return -EINVAL;
c486da34
LAG		 5990
5991 net = (struct net *)ctl->extra1;
5992 delay = net->ipv6.sysctl.flush_delay;
f0fb9b28
AP		 5993 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5994 if (ret)
5995 return ret;
5996
2ac3ac8f 5997 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
c486da34 5998 return 0;
1da177e4
LT		 5999}
6000
7c6bb7d2
DA		 6001static int zero;
6002static int one = 1;
6003
ed792e28 6004static struct ctl_table ipv6_route_table_template[] = {
1ab1457c 6005 {
1da177e4 6006 .procname = "flush",
4990509f 6007 .data = &init_net.ipv6.sysctl.flush_delay,
1da177e4 6008 .maxlen = sizeof(int),
89c8b3a1 6009 .mode = 0200,
6d9f239a 6010 .proc_handler = ipv6_sysctl_rtcache_flush
1da177e4
LT		 6011 },
6012 {
1da177e4 6013 .procname = "gc_thresh",
9a7ec3a9 6014 .data = &ip6_dst_ops_template.gc_thresh,
1da177e4
LT		 6015 .maxlen = sizeof(int),
6016 .mode = 0644,
6d9f239a 6017 .proc_handler = proc_dointvec,
1da177e4
LT		 6018 },
6019 {
1da177e4 6020 .procname = "max_size",
4990509f 6021 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
1da177e4
LT		 6022 .maxlen = sizeof(int),
6023 .mode = 0644,
6d9f239a 6024 .proc_handler = proc_dointvec,
1da177e4
LT		 6025 },
6026 {
1da177e4 6027 .procname = "gc_min_interval",
4990509f 6028 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 6029 .maxlen = sizeof(int),
6030 .mode = 0644,
6d9f239a 6031 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6032 },
6033 {
1da177e4 6034 .procname = "gc_timeout",
4990509f 6035 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
1da177e4
LT		 6036 .maxlen = sizeof(int),
6037 .mode = 0644,
6d9f239a 6038 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6039 },
6040 {
1da177e4 6041 .procname = "gc_interval",
4990509f 6042 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
1da177e4
LT		 6043 .maxlen = sizeof(int),
6044 .mode = 0644,
6d9f239a 6045 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6046 },
6047 {
1da177e4 6048 .procname = "gc_elasticity",
4990509f 6049 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
1da177e4
LT		 6050 .maxlen = sizeof(int),
6051 .mode = 0644,
f3d3f616 6052 .proc_handler = proc_dointvec,
1da177e4
LT		 6053 },
6054 {
1da177e4 6055 .procname = "mtu_expires",
4990509f 6056 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
1da177e4
LT		 6057 .maxlen = sizeof(int),
6058 .mode = 0644,
6d9f239a 6059 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 6060 },
6061 {
1da177e4 6062 .procname = "min_adv_mss",
4990509f 6063 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
1da177e4
LT		 6064 .maxlen = sizeof(int),
6065 .mode = 0644,
f3d3f616 6066 .proc_handler = proc_dointvec,
1da177e4
LT		 6067 },
6068 {
1da177e4 6069 .procname = "gc_min_interval_ms",
4990509f 6070 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 6071 .maxlen = sizeof(int),
6072 .mode = 0644,
6d9f239a 6073 .proc_handler = proc_dointvec_ms_jiffies,
1da177e4 6074 },
7c6bb7d2
DA		 6075 {
6076 .procname = "skip_notify_on_dev_down",
6077 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6078 .maxlen = sizeof(int),
6079 .mode = 0644,
6080 .proc_handler = proc_dointvec,
6081 .extra1 = &zero,
6082 .extra2 = &one,
6083 },
f8572d8f 6084 { }
1da177e4
LT		 6085};
6086
2c8c1e72 6087struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
760f2d01
DL		 6088{
6089 struct ctl_table *table;
6090
6091 table = kmemdup(ipv6_route_table_template,
6092 sizeof(ipv6_route_table_template),
6093 GFP_KERNEL);
5ee09105
YH		 6094
6095 if (table) {
6096 table[0].data = &net->ipv6.sysctl.flush_delay;
c486da34 6097 table[0].extra1 = net;
86393e52 6098 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
5ee09105
YH		 6099 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
6100 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6101 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6102 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6103 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6104 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6105 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
9c69fabe 6106 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
7c6bb7d2 6107 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
464dc801
EB		 6108
6109 /* Don't export sysctls to unprivileged users */
6110 if (net->user_ns != &init_user_ns)
6111 table[0].procname = NULL;
5ee09105
YH		 6112 }
6113
760f2d01
DL		 6114 return table;
6115}
1da177e4
LT		 6116#endif
6117
2c8c1e72 6118static int __net_init ip6_route_net_init(struct net *net)
cdb18761 6119{
633d424b 6120 int ret = -ENOMEM;
8ed67789 6121
86393e52
AD		 6122 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6123 sizeof(net->ipv6.ip6_dst_ops));
f2fc6a54 6124
fc66f95c
ED		 6125 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6126 goto out_ip6_dst_ops;
6127
1cf844c7 6128 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
421842ed
DA		 6129 if (!net->ipv6.fib6_null_entry)
6130 goto out_ip6_dst_entries;
1cf844c7
DA		 6131 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6132 sizeof(*net->ipv6.fib6_null_entry));
421842ed 6133
8ed67789
DL		 6134 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6135 sizeof(*net->ipv6.ip6_null_entry),
6136 GFP_KERNEL);
6137 if (!net->ipv6.ip6_null_entry)
421842ed 6138 goto out_fib6_null_entry;
d8d1f30b 6139 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6140 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6141 ip6_template_metrics, true);
74109218 6142 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->rt6i_uncached);
8ed67789
DL		 6143
6144#ifdef CONFIG_IPV6_MULTIPLE_TABLES
feca7d8c 6145 net->ipv6.fib6_has_custom_rules = false;
8ed67789
DL		 6146 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6147 sizeof(*net->ipv6.ip6_prohibit_entry),
6148 GFP_KERNEL);
68fffc67
PZ		 6149 if (!net->ipv6.ip6_prohibit_entry)
6150 goto out_ip6_null_entry;
d8d1f30b 6151 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6152 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6153 ip6_template_metrics, true);
74109218 6154 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->rt6i_uncached);
8ed67789
DL		 6155
6156 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6157 sizeof(*net->ipv6.ip6_blk_hole_entry),
6158 GFP_KERNEL);
68fffc67
PZ		 6159 if (!net->ipv6.ip6_blk_hole_entry)
6160 goto out_ip6_prohibit_entry;
d8d1f30b 6161 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6162 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6163 ip6_template_metrics, true);
74109218 6164 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->rt6i_uncached);
8ed67789
DL		 6165#endif
6166
b339a47c
PZ		 6167 net->ipv6.sysctl.flush_delay = 0;
6168 net->ipv6.sysctl.ip6_rt_max_size = 4096;
6169 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6170 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6171 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6172 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6173 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6174 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
7c6bb7d2 6175 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
b339a47c 6176
6891a346
BT		 6177 net->ipv6.ip6_rt_gc_expire = 30*HZ;
6178
8ed67789
DL		 6179 ret = 0;
6180out:
6181 return ret;
f2fc6a54 6182
68fffc67
PZ		 6183#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6184out_ip6_prohibit_entry:
6185 kfree(net->ipv6.ip6_prohibit_entry);
6186out_ip6_null_entry:
6187 kfree(net->ipv6.ip6_null_entry);
6188#endif
421842ed
DA		 6189out_fib6_null_entry:
6190 kfree(net->ipv6.fib6_null_entry);
fc66f95c
ED		 6191out_ip6_dst_entries:
6192 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
f2fc6a54 6193out_ip6_dst_ops:
f2fc6a54 6194 goto out;
cdb18761
DL		 6195}
6196
2c8c1e72 6197static void __net_exit ip6_route_net_exit(struct net *net)
cdb18761 6198{
421842ed 6199 kfree(net->ipv6.fib6_null_entry);
8ed67789
DL		 6200 kfree(net->ipv6.ip6_null_entry);
6201#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6202 kfree(net->ipv6.ip6_prohibit_entry);
6203 kfree(net->ipv6.ip6_blk_hole_entry);
6204#endif
41bb78b4 6205 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
cdb18761
DL		 6206}
6207
d189634e
TG		 6208static int __net_init ip6_route_net_init_late(struct net *net)
6209{
6210#ifdef CONFIG_PROC_FS
c3506372
CH		 6211 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6212 sizeof(struct ipv6_route_iter));
3617d949
CH		 6213 proc_create_net_single("rt6_stats", 0444, net->proc_net,
6214 rt6_stats_seq_show, NULL);
d189634e
TG		 6215#endif
6216 return 0;
6217}
6218
6219static void __net_exit ip6_route_net_exit_late(struct net *net)
6220{
6221#ifdef CONFIG_PROC_FS
ece31ffd
G		 6222 remove_proc_entry("ipv6_route", net->proc_net);
6223 remove_proc_entry("rt6_stats", net->proc_net);
d189634e
TG		 6224#endif
6225}
6226
cdb18761
DL		 6227static struct pernet_operations ip6_route_net_ops = {
6228 .init = ip6_route_net_init,
6229 .exit = ip6_route_net_exit,
6230};
6231
c3426b47
DM		 6232static int __net_init ipv6_inetpeer_init(struct net *net)
6233{
6234 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6235
6236 if (!bp)
6237 return -ENOMEM;
6238 inet_peer_base_init(bp);
6239 net->ipv6.peers = bp;
6240 return 0;
6241}
6242
6243static void __net_exit ipv6_inetpeer_exit(struct net *net)
6244{
6245 struct inet_peer_base *bp = net->ipv6.peers;
6246
6247 net->ipv6.peers = NULL;
56a6b248 6248 inetpeer_invalidate_tree(bp);
c3426b47
DM		 6249 kfree(bp);
6250}
6251
2b823f72 6252static struct pernet_operations ipv6_inetpeer_ops = {
c3426b47
DM		 6253 .init = ipv6_inetpeer_init,
6254 .exit = ipv6_inetpeer_exit,
6255};
6256
d189634e
TG		 6257static struct pernet_operations ip6_route_net_late_ops = {
6258 .init = ip6_route_net_init_late,
6259 .exit = ip6_route_net_exit_late,
6260};
6261
8ed67789
DL		 6262static struct notifier_block ip6_route_dev_notifier = {
6263 .notifier_call = ip6_route_dev_notify,
242d3a49 6264 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
8ed67789
DL		 6265};
6266
2f460933
WC		 6267void __init ip6_route_init_special_entries(void)
6268{
6269 /* Registering of the loopback is done before this portion of code,
6270 * the loopback reference in rt6_info will not be taken, do it
6271 * manually for init_net */
1cf844c7 6272 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
2f460933
WC		 6273 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6274 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6275 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6276 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6277 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6278 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6279 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6280 #endif
6281}
6282
433d49c3 6283int __init ip6_route_init(void)
1da177e4 6284{
433d49c3 6285 int ret;
8d0b94af 6286 int cpu;
433d49c3 6287
9a7ec3a9
DL		 6288 ret = -ENOMEM;
6289 ip6_dst_ops_template.kmem_cachep =
e5d679f3 6290 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
f845ab6b 6291 SLAB_HWCACHE_ALIGN, NULL);
9a7ec3a9 6292 if (!ip6_dst_ops_template.kmem_cachep)
c19a28e1 6293 goto out;
14e50e57 6294
fc66f95c 6295 ret = dst_entries_init(&ip6_dst_blackhole_ops);
8ed67789 6296 if (ret)
bdb3289f 6297 goto out_kmem_cache;
bdb3289f 6298
c3426b47
DM		 6299 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6300 if (ret)
e8803b6c 6301 goto out_dst_entries;
2a0c451a 6302
7e52b33b
DM		 6303 ret = register_pernet_subsys(&ip6_route_net_ops);
6304 if (ret)
6305 goto out_register_inetpeer;
c3426b47 6306
5dc121e9
AE		 6307 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6308
e8803b6c 6309 ret = fib6_init();
433d49c3 6310 if (ret)
8ed67789 6311 goto out_register_subsys;
433d49c3 6312
433d49c3
DL		 6313 ret = xfrm6_init();
6314 if (ret)
e8803b6c 6315 goto out_fib6_init;
c35b7e72 6316
433d49c3
DL		 6317 ret = fib6_rules_init();
6318 if (ret)
6319 goto xfrm6_init;
7e5449c2 6320
d189634e
TG		 6321 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6322 if (ret)
6323 goto fib6_rules_init;
6324
16feebcf
FW		 6325 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6326 inet6_rtm_newroute, NULL, 0);
6327 if (ret < 0)
6328 goto out_register_late_subsys;
6329
6330 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6331 inet6_rtm_delroute, NULL, 0);
6332 if (ret < 0)
6333 goto out_register_late_subsys;
6334
6335 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6336 inet6_rtm_getroute, NULL,
6337 RTNL_FLAG_DOIT_UNLOCKED);
6338 if (ret < 0)
d189634e 6339 goto out_register_late_subsys;
c127ea2c 6340
8ed67789 6341 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
cdb18761 6342 if (ret)
d189634e 6343 goto out_register_late_subsys;
8ed67789 6344
8d0b94af
MKL		 6345 for_each_possible_cpu(cpu) {
6346 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6347
6348 INIT_LIST_HEAD(&ul->head);
6349 spin_lock_init(&ul->lock);
6350 }
6351
433d49c3
DL		 6352out:
6353 return ret;
6354
d189634e 6355out_register_late_subsys:
16feebcf 6356 rtnl_unregister_all(PF_INET6);
d189634e 6357 unregister_pernet_subsys(&ip6_route_net_late_ops);
433d49c3 6358fib6_rules_init:
433d49c3
DL		 6359 fib6_rules_cleanup();
6360xfrm6_init:
433d49c3 6361 xfrm6_fini();
2a0c451a
TG		 6362out_fib6_init:
6363 fib6_gc_cleanup();
8ed67789
DL		 6364out_register_subsys:
6365 unregister_pernet_subsys(&ip6_route_net_ops);
7e52b33b
DM		 6366out_register_inetpeer:
6367 unregister_pernet_subsys(&ipv6_inetpeer_ops);
fc66f95c
ED		 6368out_dst_entries:
6369 dst_entries_destroy(&ip6_dst_blackhole_ops);
433d49c3 6370out_kmem_cache:
f2fc6a54 6371 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
433d49c3 6372 goto out;
1da177e4
LT		 6373}
6374
6375void ip6_route_cleanup(void)
6376{
8ed67789 6377 unregister_netdevice_notifier(&ip6_route_dev_notifier);
d189634e 6378 unregister_pernet_subsys(&ip6_route_net_late_ops);
101367c2 6379 fib6_rules_cleanup();
1da177e4 6380 xfrm6_fini();
1da177e4 6381 fib6_gc_cleanup();
c3426b47 6382 unregister_pernet_subsys(&ipv6_inetpeer_ops);
8ed67789 6383 unregister_pernet_subsys(&ip6_route_net_ops);
41bb78b4 6384 dst_entries_destroy(&ip6_dst_blackhole_ops);
f2fc6a54 6385 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
1da177e4 6386}
Linux 6.x block layer and io_uring tree(s)