projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ipv6: Add IPv6 multipath notifications for add / replace
[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
d4ead6b3 1394 if (pcpu_rt)
10585b43 1395 ip6_hold_safe(NULL, &pcpu_rt);
d3843fe5 1396
a73e4195
MKL		 1397 return pcpu_rt;
1398}
1399
afb1d4b5 1400static struct rt6_info *rt6_make_pcpu_route(struct net *net,
db3fedee 1401 const struct fib6_result *res)
a73e4195
MKL		 1402{
1403 struct rt6_info *pcpu_rt, *prev, **p;
d52d3997 1404
db3fedee 1405 pcpu_rt = ip6_rt_pcpu_alloc(res);
d52d3997 1406 if (!pcpu_rt) {
9c7370a1
MKL		 1407 dst_hold(&net->ipv6.ip6_null_entry->dst);
1408 return net->ipv6.ip6_null_entry;
d52d3997
MKL		 1409 }
1410
a94b9367 1411 dst_hold(&pcpu_rt->dst);
f40b6ae2 1412 p = this_cpu_ptr(res->nh->rt6i_pcpu);
a94b9367 1413 prev = cmpxchg(p, NULL, pcpu_rt);
951f788a 1414 BUG_ON(prev);
a94b9367 1415
61fb0d01
ED		 1416 if (res->f6i->fib6_destroying) {
1417 struct fib6_info *from;
1418
1419 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1420 fib6_info_release(from);
1421 }
1422
d52d3997
MKL		 1423 return pcpu_rt;
1424}
1425
35732d01
WW		 1426/* exception hash table implementation
1427 */
1428static DEFINE_SPINLOCK(rt6_exception_lock);
1429
1430/* Remove rt6_ex from hash table and free the memory
1431 * Caller must hold rt6_exception_lock
1432 */
1433static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1434 struct rt6_exception *rt6_ex)
1435{
f5b51fe8 1436 struct fib6_info *from;
b2427e67 1437 struct net *net;
81eb8447 1438
35732d01
WW		 1439 if (!bucket || !rt6_ex)
1440 return;
b2427e67
CIK		 1441
1442 net = dev_net(rt6_ex->rt6i->dst.dev);
f5b51fe8
PA		 1443 net->ipv6.rt6_stats->fib_rt_cache--;
1444
1445 /* purge completely the exception to allow releasing the held resources:
1446 * some [sk] cache may keep the dst around for unlimited time
1447 */
0e233874 1448 from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
f5b51fe8
PA		 1449 fib6_info_release(from);
1450 dst_dev_put(&rt6_ex->rt6i->dst);
1451
35732d01 1452 hlist_del_rcu(&rt6_ex->hlist);
77634cc6 1453 dst_release(&rt6_ex->rt6i->dst);
35732d01
WW		 1454 kfree_rcu(rt6_ex, rcu);
1455 WARN_ON_ONCE(!bucket->depth);
1456 bucket->depth--;
1457}
1458
1459/* Remove oldest rt6_ex in bucket and free the memory
1460 * Caller must hold rt6_exception_lock
1461 */
1462static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1463{
1464 struct rt6_exception *rt6_ex, *oldest = NULL;
1465
1466 if (!bucket)
1467 return;
1468
1469 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1470 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1471 oldest = rt6_ex;
1472 }
1473 rt6_remove_exception(bucket, oldest);
1474}
1475
1476static u32 rt6_exception_hash(const struct in6_addr *dst,
1477 const struct in6_addr *src)
1478{
1479 static u32 seed __read_mostly;
1480 u32 val;
1481
1482 net_get_random_once(&seed, sizeof(seed));
1483 val = jhash(dst, sizeof(*dst), seed);
1484
1485#ifdef CONFIG_IPV6_SUBTREES
1486 if (src)
1487 val = jhash(src, sizeof(*src), val);
1488#endif
1489 return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1490}
1491
1492/* Helper function to find the cached rt in the hash table
1493 * and update bucket pointer to point to the bucket for this
1494 * (daddr, saddr) pair
1495 * Caller must hold rt6_exception_lock
1496 */
1497static struct rt6_exception *
1498__rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1499 const struct in6_addr *daddr,
1500 const struct in6_addr *saddr)
1501{
1502 struct rt6_exception *rt6_ex;
1503 u32 hval;
1504
1505 if (!(*bucket) || !daddr)
1506 return NULL;
1507
1508 hval = rt6_exception_hash(daddr, saddr);
1509 *bucket += hval;
1510
1511 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1512 struct rt6_info *rt6 = rt6_ex->rt6i;
1513 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1514
1515#ifdef CONFIG_IPV6_SUBTREES
1516 if (matched && saddr)
1517 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1518#endif
1519 if (matched)
1520 return rt6_ex;
1521 }
1522 return NULL;
1523}
1524
1525/* Helper function to find the cached rt in the hash table
1526 * and update bucket pointer to point to the bucket for this
1527 * (daddr, saddr) pair
1528 * Caller must hold rcu_read_lock()
1529 */
1530static struct rt6_exception *
1531__rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1532 const struct in6_addr *daddr,
1533 const struct in6_addr *saddr)
1534{
1535 struct rt6_exception *rt6_ex;
1536 u32 hval;
1537
1538 WARN_ON_ONCE(!rcu_read_lock_held());
1539
1540 if (!(*bucket) || !daddr)
1541 return NULL;
1542
1543 hval = rt6_exception_hash(daddr, saddr);
1544 *bucket += hval;
1545
1546 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1547 struct rt6_info *rt6 = rt6_ex->rt6i;
1548 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1549
1550#ifdef CONFIG_IPV6_SUBTREES
1551 if (matched && saddr)
1552 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1553#endif
1554 if (matched)
1555 return rt6_ex;
1556 }
1557 return NULL;
1558}
1559
b748f260 1560static unsigned int fib6_mtu(const struct fib6_result *res)
d4ead6b3 1561{
b748f260 1562 const struct fib6_nh *nh = res->nh;
d4ead6b3
DA		 1563 unsigned int mtu;
1564
b748f260
DA		 1565 if (res->f6i->fib6_pmtu) {
1566 mtu = res->f6i->fib6_pmtu;
dcd1f572 1567 } else {
b748f260 1568 struct net_device *dev = nh->fib_nh_dev;
dcd1f572
DA		 1569 struct inet6_dev *idev;
1570
1571 rcu_read_lock();
1572 idev = __in6_dev_get(dev);
1573 mtu = idev->cnf.mtu6;
1574 rcu_read_unlock();
1575 }
1576
d4ead6b3
DA		 1577 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1578
b748f260 1579 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
d4ead6b3
DA		 1580}
1581
cc5c073a
DA		 1582#define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1583
1584/* used when the flushed bit is not relevant, only access to the bucket
1585 * (ie., all bucket users except rt6_insert_exception);
1586 *
1587 * called under rcu lock; sometimes called with rt6_exception_lock held
1588 */
1589static
1590struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1591 spinlock_t *lock)
1592{
1593 struct rt6_exception_bucket *bucket;
1594
1595 if (lock)
1596 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1597 lockdep_is_held(lock));
1598 else
1599 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1600
1601 /* remove bucket flushed bit if set */
1602 if (bucket) {
1603 unsigned long p = (unsigned long)bucket;
1604
1605 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1606 bucket = (struct rt6_exception_bucket *)p;
1607 }
1608
1609 return bucket;
1610}
1611
1612static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1613{
1614 unsigned long p = (unsigned long)bucket;
1615
1616 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1617}
1618
1619/* called with rt6_exception_lock held */
1620static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1621 spinlock_t *lock)
1622{
1623 struct rt6_exception_bucket *bucket;
1624 unsigned long p;
1625
1626 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1627 lockdep_is_held(lock));
1628
1629 p = (unsigned long)bucket;
1630 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1631 bucket = (struct rt6_exception_bucket *)p;
1632 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1633}
1634
35732d01 1635static int rt6_insert_exception(struct rt6_info *nrt,
5012f0a5 1636 const struct fib6_result *res)
35732d01 1637{
5e670d84 1638 struct net *net = dev_net(nrt->dst.dev);
35732d01 1639 struct rt6_exception_bucket *bucket;
cc5c073a 1640 struct fib6_info *f6i = res->f6i;
35732d01
WW		 1641 struct in6_addr *src_key = NULL;
1642 struct rt6_exception *rt6_ex;
cc5c073a 1643 struct fib6_nh *nh = res->nh;
35732d01
WW		 1644 int err = 0;
1645
35732d01
WW		 1646 spin_lock_bh(&rt6_exception_lock);
1647
cc5c073a
DA		 1648 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1649 lockdep_is_held(&rt6_exception_lock));
35732d01
WW		 1650 if (!bucket) {
1651 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1652 GFP_ATOMIC);
1653 if (!bucket) {
1654 err = -ENOMEM;
1655 goto out;
1656 }
cc5c073a
DA		 1657 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1658 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1659 err = -EINVAL;
1660 goto out;
35732d01
WW		 1661 }
1662
1663#ifdef CONFIG_IPV6_SUBTREES
5012f0a5 1664 /* fib6_src.plen != 0 indicates f6i is in subtree
35732d01 1665 * and exception table is indexed by a hash of
5012f0a5 1666 * both fib6_dst and fib6_src.
35732d01 1667 * Otherwise, the exception table is indexed by
5012f0a5 1668 * a hash of only fib6_dst.
35732d01 1669 */
5012f0a5 1670 if (f6i->fib6_src.plen)
35732d01
WW		 1671 src_key = &nrt->rt6i_src.addr;
1672#endif
5012f0a5 1673 /* rt6_mtu_change() might lower mtu on f6i.
f5bbe7ee 1674 * Only insert this exception route if its mtu
5012f0a5 1675 * is less than f6i's mtu value.
f5bbe7ee 1676 */
b748f260 1677 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
f5bbe7ee
WW		 1678 err = -EINVAL;
1679 goto out;
1680 }
60006a48 1681
35732d01
WW		 1682 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1683 src_key);
1684 if (rt6_ex)
1685 rt6_remove_exception(bucket, rt6_ex);
1686
1687 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1688 if (!rt6_ex) {
1689 err = -ENOMEM;
1690 goto out;
1691 }
1692 rt6_ex->rt6i = nrt;
1693 rt6_ex->stamp = jiffies;
35732d01
WW		 1694 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1695 bucket->depth++;
81eb8447 1696 net->ipv6.rt6_stats->fib_rt_cache++;
35732d01
WW		 1697
1698 if (bucket->depth > FIB6_MAX_DEPTH)
1699 rt6_exception_remove_oldest(bucket);
1700
1701out:
1702 spin_unlock_bh(&rt6_exception_lock);
1703
1704 /* Update fn->fn_sernum to invalidate all cached dst */
b886d5f2 1705 if (!err) {
5012f0a5
DA		 1706 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1707 fib6_update_sernum(net, f6i);
1708 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
b886d5f2
PA		 1709 fib6_force_start_gc(net);
1710 }
35732d01
WW		 1711
1712 return err;
1713}
1714
c0b220cf 1715static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
35732d01
WW		 1716{
1717 struct rt6_exception_bucket *bucket;
1718 struct rt6_exception *rt6_ex;
1719 struct hlist_node *tmp;
1720 int i;
1721
1722 spin_lock_bh(&rt6_exception_lock);
35732d01 1723
cc5c073a 1724 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
35732d01
WW		 1725 if (!bucket)
1726 goto out;
1727
cc5c073a
DA		 1728 /* Prevent rt6_insert_exception() to recreate the bucket list */
1729 if (!from)
1730 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1731
35732d01 1732 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
cc5c073a
DA		 1733 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1734 if (!from ||
1735 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1736 rt6_remove_exception(bucket, rt6_ex);
1737 }
1738 WARN_ON_ONCE(!from && bucket->depth);
35732d01
WW		 1739 bucket++;
1740 }
35732d01
WW		 1741out:
1742 spin_unlock_bh(&rt6_exception_lock);
1743}
1744
e659ba31
DA		 1745static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1746{
1747 struct fib6_info *f6i = arg;
1748
1749 fib6_nh_flush_exceptions(nh, f6i);
1750
1751 return 0;
1752}
1753
c0b220cf
DA		 1754void rt6_flush_exceptions(struct fib6_info *f6i)
1755{
e659ba31
DA		 1756 if (f6i->nh)
1757 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1758 f6i);
1759 else
1760 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
c0b220cf
DA		 1761}
1762
35732d01
WW		 1763/* Find cached rt in the hash table inside passed in rt
1764 * Caller has to hold rcu_read_lock()
1765 */
7e4b5128 1766static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
510e2ced
WW		 1767 const struct in6_addr *daddr,
1768 const struct in6_addr *saddr)
35732d01 1769{
510e2ced 1770 const struct in6_addr *src_key = NULL;
35732d01 1771 struct rt6_exception_bucket *bucket;
35732d01 1772 struct rt6_exception *rt6_ex;
7e4b5128 1773 struct rt6_info *ret = NULL;
35732d01 1774
35732d01 1775#ifdef CONFIG_IPV6_SUBTREES
7e4b5128 1776 /* fib6i_src.plen != 0 indicates f6i is in subtree
35732d01 1777 * and exception table is indexed by a hash of
7e4b5128 1778 * both fib6_dst and fib6_src.
510e2ced
WW		 1779 * However, the src addr used to create the hash
1780 * might not be exactly the passed in saddr which
1781 * is a /128 addr from the flow.
1782 * So we need to use f6i->fib6_src to redo lookup
1783 * if the passed in saddr does not find anything.
1784 * (See the logic in ip6_rt_cache_alloc() on how
1785 * rt->rt6i_src is updated.)
35732d01 1786 */
7e4b5128 1787 if (res->f6i->fib6_src.plen)
35732d01 1788 src_key = saddr;
510e2ced 1789find_ex:
35732d01 1790#endif
cc5c073a 1791 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
35732d01
WW		 1792 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1793
1794 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
7e4b5128 1795 ret = rt6_ex->rt6i;
35732d01 1796
510e2ced
WW		 1797#ifdef CONFIG_IPV6_SUBTREES
1798 /* Use fib6_src as src_key and redo lookup */
1799 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1800 src_key = &res->f6i->fib6_src.addr;
1801 goto find_ex;
1802 }
1803#endif
1804
7e4b5128 1805 return ret;
35732d01
WW		 1806}
1807
1808/* Remove the passed in cached rt from the hash table that contains it */
cc5c073a 1809static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1810 const struct rt6_info *rt)
35732d01 1811{
c0b220cf 1812 const struct in6_addr *src_key = NULL;
35732d01 1813 struct rt6_exception_bucket *bucket;
35732d01
WW		 1814 struct rt6_exception *rt6_ex;
1815 int err;
1816
cc5c073a 1817 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
35732d01
WW		 1818 return -ENOENT;
1819
1820 spin_lock_bh(&rt6_exception_lock);
cc5c073a
DA		 1821 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1822
35732d01 1823#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1824 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1825 * and exception table is indexed by a hash of
1826 * both rt6i_dst and rt6i_src.
35732d01
WW		 1827 * Otherwise, the exception table is indexed by
1828 * a hash of only rt6i_dst.
1829 */
c0b220cf 1830 if (plen)
35732d01
WW		 1831 src_key = &rt->rt6i_src.addr;
1832#endif
1833 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1834 &rt->rt6i_dst.addr,
1835 src_key);
1836 if (rt6_ex) {
1837 rt6_remove_exception(bucket, rt6_ex);
1838 err = 0;
1839 } else {
1840 err = -ENOENT;
1841 }
1842
1843 spin_unlock_bh(&rt6_exception_lock);
1844 return err;
1845}
1846
e659ba31
DA		 1847struct fib6_nh_excptn_arg {
1848 struct rt6_info *rt;
1849 int plen;
1850};
1851
1852static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1853{
1854 struct fib6_nh_excptn_arg *arg = _arg;
1855 int err;
1856
1857 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1858 if (err == 0)
1859 return 1;
1860
1861 return 0;
1862}
1863
c0b220cf
DA		 1864static int rt6_remove_exception_rt(struct rt6_info *rt)
1865{
1866 struct fib6_info *from;
1867
1868 from = rcu_dereference(rt->from);
cc5c073a 1869 if (!from || !(rt->rt6i_flags & RTF_CACHE))
c0b220cf
DA		 1870 return -EINVAL;
1871
e659ba31
DA		 1872 if (from->nh) {
1873 struct fib6_nh_excptn_arg arg = {
1874 .rt = rt,
1875 .plen = from->fib6_src.plen
1876 };
1877 int rc;
1878
1879 /* rc = 1 means an entry was found */
1880 rc = nexthop_for_each_fib6_nh(from->nh,
1881 rt6_nh_remove_exception_rt,
1882 &arg);
1883 return rc ? 0 : -ENOENT;
1884 }
1885
1cf844c7 1886 return fib6_nh_remove_exception(from->fib6_nh,
cc5c073a 1887 from->fib6_src.plen, rt);
c0b220cf
DA		 1888}
1889
35732d01
WW		 1890/* Find rt6_ex which contains the passed in rt cache and
1891 * refresh its stamp
1892 */
cc5c073a 1893static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1894 const struct rt6_info *rt)
35732d01 1895{
c0b220cf 1896 const struct in6_addr *src_key = NULL;
35732d01 1897 struct rt6_exception_bucket *bucket;
35732d01 1898 struct rt6_exception *rt6_ex;
193f3685 1899
cc5c073a 1900 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
35732d01 1901#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1902 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1903 * and exception table is indexed by a hash of
1904 * both rt6i_dst and rt6i_src.
35732d01
WW		 1905 * Otherwise, the exception table is indexed by
1906 * a hash of only rt6i_dst.
1907 */
c0b220cf 1908 if (plen)
35732d01
WW		 1909 src_key = &rt->rt6i_src.addr;
1910#endif
cc5c073a 1911 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
35732d01
WW		 1912 if (rt6_ex)
1913 rt6_ex->stamp = jiffies;
c0b220cf
DA		 1914}
1915
e659ba31
DA		 1916struct fib6_nh_match_arg {
1917 const struct net_device *dev;
1918 const struct in6_addr *gw;
1919 struct fib6_nh *match;
1920};
1921
1922/* determine if fib6_nh has given device and gateway */
1923static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1924{
1925 struct fib6_nh_match_arg *arg = _arg;
1926
1927 if (arg->dev != nh->fib_nh_dev ||
1928 (arg->gw && !nh->fib_nh_gw_family) ||
1929 (!arg->gw && nh->fib_nh_gw_family) ||
1930 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1931 return 0;
1932
1933 arg->match = nh;
1934
1935 /* found a match, break the loop */
1936 return 1;
1937}
1938
c0b220cf
DA		 1939static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1940{
1941 struct fib6_info *from;
e659ba31 1942 struct fib6_nh *fib6_nh;
35732d01 1943
c0b220cf
DA		 1944 rcu_read_lock();
1945
1946 from = rcu_dereference(rt->from);
1947 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1948 goto unlock;
1949
e659ba31
DA		 1950 if (from->nh) {
1951 struct fib6_nh_match_arg arg = {
1952 .dev = rt->dst.dev,
1953 .gw = &rt->rt6i_gateway,
1954 };
1955
1956 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1957
1958 if (!arg.match)
1959 return;
1960 fib6_nh = arg.match;
1961 } else {
1962 fib6_nh = from->fib6_nh;
1963 }
1964 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
193f3685 1965unlock:
35732d01
WW		 1966 rcu_read_unlock();
1967}
1968
e9fa1495
SB		 1969static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1970 struct rt6_info *rt, int mtu)
1971{
1972 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1973 * lowest MTU in the path: always allow updating the route PMTU to
1974 * reflect PMTU decreases.
1975 *
1976 * If the new MTU is higher, and the route PMTU is equal to the local
1977 * MTU, this means the old MTU is the lowest in the path, so allow
1978 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1979 * handle this.
1980 */
1981
1982 if (dst_mtu(&rt->dst) >= mtu)
1983 return true;
1984
1985 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
1986 return true;
1987
1988 return false;
1989}
1990
1991static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
cc5c073a 1992 const struct fib6_nh *nh, int mtu)
f5bbe7ee
WW		 1993{
1994 struct rt6_exception_bucket *bucket;
1995 struct rt6_exception *rt6_ex;
1996 int i;
1997
cc5c073a 1998 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
e9fa1495
SB		 1999 if (!bucket)
2000 return;
2001
2002 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2003 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2004 struct rt6_info *entry = rt6_ex->rt6i;
2005
2006 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
d4ead6b3 2007 * route), the metrics of its rt->from have already
e9fa1495
SB		 2008 * been updated.
2009 */
d4ead6b3 2010 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
e9fa1495 2011 rt6_mtu_change_route_allowed(idev, entry, mtu))
d4ead6b3 2012 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
f5bbe7ee 2013 }
e9fa1495 2014 bucket++;
f5bbe7ee
WW		 2015 }
2016}
2017
b16cb459
WW		 2018#define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2019
cc5c073a
DA		 2020static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2021 const struct in6_addr *gateway)
b16cb459
WW		 2022{
2023 struct rt6_exception_bucket *bucket;
2024 struct rt6_exception *rt6_ex;
2025 struct hlist_node *tmp;
2026 int i;
2027
cc5c073a 2028 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
b16cb459
WW		 2029 return;
2030
2031 spin_lock_bh(&rt6_exception_lock);
cc5c073a 2032 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
b16cb459
WW		 2033 if (bucket) {
2034 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2035 hlist_for_each_entry_safe(rt6_ex, tmp,
2036 &bucket->chain, hlist) {
2037 struct rt6_info *entry = rt6_ex->rt6i;
2038
2039 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2040 RTF_CACHE_GATEWAY &&
2041 ipv6_addr_equal(gateway,
2042 &entry->rt6i_gateway)) {
2043 rt6_remove_exception(bucket, rt6_ex);
2044 }
2045 }
2046 bucket++;
2047 }
2048 }
2049
2050 spin_unlock_bh(&rt6_exception_lock);
2051}
2052
c757faa8
WW		 2053static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2054 struct rt6_exception *rt6_ex,
2055 struct fib6_gc_args *gc_args,
2056 unsigned long now)
2057{
2058 struct rt6_info *rt = rt6_ex->rt6i;
2059
1859bac0
PA		 2060 /* we are pruning and obsoleting aged-out and non gateway exceptions
2061 * even if others have still references to them, so that on next
2062 * dst_check() such references can be dropped.
2063 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2064 * expired, independently from their aging, as per RFC 8201 section 4
2065 */
31afeb42
WW		 2066 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2067 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2068 RT6_TRACE("aging clone %p\n", rt);
2069 rt6_remove_exception(bucket, rt6_ex);
2070 return;
2071 }
2072 } else if (time_after(jiffies, rt->dst.expires)) {
2073 RT6_TRACE("purging expired route %p\n", rt);
c757faa8
WW		 2074 rt6_remove_exception(bucket, rt6_ex);
2075 return;
31afeb42
WW		 2076 }
2077
2078 if (rt->rt6i_flags & RTF_GATEWAY) {
c757faa8
WW		 2079 struct neighbour *neigh;
2080 __u8 neigh_flags = 0;
2081
1bfa26ff
ED		 2082 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2083 if (neigh)
c757faa8 2084 neigh_flags = neigh->flags;
1bfa26ff 2085
c757faa8
WW		 2086 if (!(neigh_flags & NTF_ROUTER)) {
2087 RT6_TRACE("purging route %p via non-router but gateway\n",
2088 rt);
2089 rt6_remove_exception(bucket, rt6_ex);
2090 return;
2091 }
2092 }
31afeb42 2093
c757faa8
WW		 2094 gc_args->more++;
2095}
2096
cc5c073a 2097static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
c0b220cf
DA		 2098 struct fib6_gc_args *gc_args,
2099 unsigned long now)
c757faa8
WW		 2100{
2101 struct rt6_exception_bucket *bucket;
2102 struct rt6_exception *rt6_ex;
2103 struct hlist_node *tmp;
2104 int i;
2105
cc5c073a 2106 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
c757faa8
WW		 2107 return;
2108
1bfa26ff
ED		 2109 rcu_read_lock_bh();
2110 spin_lock(&rt6_exception_lock);
cc5c073a 2111 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
c757faa8
WW		 2112 if (bucket) {
2113 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2114 hlist_for_each_entry_safe(rt6_ex, tmp,
2115 &bucket->chain, hlist) {
2116 rt6_age_examine_exception(bucket, rt6_ex,
2117 gc_args, now);
2118 }
2119 bucket++;
2120 }
2121 }
1bfa26ff
ED		 2122 spin_unlock(&rt6_exception_lock);
2123 rcu_read_unlock_bh();
c757faa8
WW		 2124}
2125
e659ba31
DA		 2126struct fib6_nh_age_excptn_arg {
2127 struct fib6_gc_args *gc_args;
2128 unsigned long now;
2129};
2130
2131static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2132{
2133 struct fib6_nh_age_excptn_arg *arg = _arg;
2134
2135 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2136 return 0;
2137}
2138
cc5c073a 2139void rt6_age_exceptions(struct fib6_info *f6i,
c0b220cf
DA		 2140 struct fib6_gc_args *gc_args,
2141 unsigned long now)
2142{
e659ba31
DA		 2143 if (f6i->nh) {
2144 struct fib6_nh_age_excptn_arg arg = {
2145 .gc_args = gc_args,
2146 .now = now
2147 };
2148
2149 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2150 &arg);
2151 } else {
2152 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2153 }
c0b220cf
DA		 2154}
2155
1d053da9 2156/* must be called with rcu lock held */
effda4dd
DA		 2157int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2158 struct flowi6 *fl6, struct fib6_result *res, int strict)
1da177e4 2159{
367efcb9 2160 struct fib6_node *fn, *saved_fn;
1da177e4 2161
6454743b 2162 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
367efcb9 2163 saved_fn = fn;
1da177e4 2164
ca254490
DA		 2165 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2166 oif = 0;
2167
a3c00e46 2168redo_rt6_select:
effda4dd
DA		 2169 rt6_select(net, fn, oif, res, strict);
2170 if (res->f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2171 fn = fib6_backtrack(fn, &fl6->saddr);
2172 if (fn)
2173 goto redo_rt6_select;
367efcb9
MKL		 2174 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2175 /* also consider unreachable route */
2176 strict &= ~RT6_LOOKUP_F_REACHABLE;
2177 fn = saved_fn;
2178 goto redo_rt6_select;
367efcb9 2179 }
a3c00e46
MKL		 2180 }
2181
effda4dd 2182 trace_fib6_table_lookup(net, res, table, fl6);
fb9de91e 2183
effda4dd 2184 return 0;
1d053da9
DA		 2185}
2186
2187struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2188 int oif, struct flowi6 *fl6,
2189 const struct sk_buff *skb, int flags)
2190{
b1d40991 2191 struct fib6_result res = {};
1d053da9
DA		 2192 struct rt6_info *rt;
2193 int strict = 0;
2194
2195 strict |= flags & RT6_LOOKUP_F_IFACE;
2196 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2197 if (net->ipv6.devconf_all->forwarding == 0)
2198 strict |= RT6_LOOKUP_F_REACHABLE;
2199
2200 rcu_read_lock();
2201
effda4dd 2202 fib6_table_lookup(net, table, oif, fl6, &res, strict);
b1d40991 2203 if (res.f6i == net->ipv6.fib6_null_entry) {
421842ed 2204 rt = net->ipv6.ip6_null_entry;
66f5d6ce 2205 rcu_read_unlock();
d3843fe5 2206 dst_hold(&rt->dst);
d3843fe5 2207 return rt;
23fb93a4
DA		 2208 }
2209
b1d40991 2210 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
d83009d4 2211
23fb93a4 2212 /*Search through exception table */
7e4b5128 2213 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
23fb93a4 2214 if (rt) {
10585b43 2215 if (ip6_hold_safe(net, &rt))
d3843fe5 2216 dst_use_noref(&rt->dst, jiffies);
d4ead6b3 2217
66f5d6ce 2218 rcu_read_unlock();
d52d3997 2219 return rt;
3da59bd9 2220 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
b1d40991 2221 !res.nh->fib_nh_gw_family)) {
3da59bd9
MKL		 2222 /* Create a RTF_CACHE clone which will not be
2223 * owned by the fib6 tree. It is for the special case where
2224 * the daddr in the skb during the neighbor look-up is different
2225 * from the fl6->daddr used to look-up route here.
2226 */
3da59bd9
MKL		 2227 struct rt6_info *uncached_rt;
2228
85bd05de 2229 uncached_rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
d52d3997 2230
4d85cd0c 2231 rcu_read_unlock();
c71099ac 2232
1cfb71ee
WW		 2233 if (uncached_rt) {
2234 /* Uncached_rt's refcnt is taken during ip6_rt_cache_alloc()
2235 * No need for another dst_hold()
2236 */
8d0b94af 2237 rt6_uncached_list_add(uncached_rt);
81eb8447 2238 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
1cfb71ee 2239 } else {
3da59bd9 2240 uncached_rt = net->ipv6.ip6_null_entry;
1cfb71ee
WW		 2241 dst_hold(&uncached_rt->dst);
2242 }
b811580d 2243
3da59bd9 2244 return uncached_rt;
d52d3997
MKL		 2245 } else {
2246 /* Get a percpu copy */
2247
2248 struct rt6_info *pcpu_rt;
2249
951f788a 2250 local_bh_disable();
db3fedee 2251 pcpu_rt = rt6_get_pcpu_route(&res);
d52d3997 2252
93531c67 2253 if (!pcpu_rt)
db3fedee 2254 pcpu_rt = rt6_make_pcpu_route(net, &res);
93531c67 2255
951f788a
ED		 2256 local_bh_enable();
2257 rcu_read_unlock();
d4bea421 2258
d52d3997
MKL		 2259 return pcpu_rt;
2260 }
1da177e4 2261}
9ff74384 2262EXPORT_SYMBOL_GPL(ip6_pol_route);
1da177e4 2263
b75cc8f9
DA		 2264static struct rt6_info *ip6_pol_route_input(struct net *net,
2265 struct fib6_table *table,
2266 struct flowi6 *fl6,
2267 const struct sk_buff *skb,
2268 int flags)
4acad72d 2269{
b75cc8f9 2270 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
4acad72d
PE		 2271}
2272
d409b847
MB		 2273struct dst_entry *ip6_route_input_lookup(struct net *net,
2274 struct net_device *dev,
b75cc8f9
DA		 2275 struct flowi6 *fl6,
2276 const struct sk_buff *skb,
2277 int flags)
72331bc0
SL		 2278{
2279 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2280 flags |= RT6_LOOKUP_F_IFACE;
2281
b75cc8f9 2282 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
72331bc0 2283}
d409b847 2284EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
72331bc0 2285
23aebdac 2286static void ip6_multipath_l3_keys(const struct sk_buff *skb,
5e5d6fed
RP		 2287 struct flow_keys *keys,
2288 struct flow_keys *flkeys)
23aebdac
JS		 2289{
2290 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2291 const struct ipv6hdr *key_iph = outer_iph;
5e5d6fed 2292 struct flow_keys *_flkeys = flkeys;
23aebdac
JS		 2293 const struct ipv6hdr *inner_iph;
2294 const struct icmp6hdr *icmph;
2295 struct ipv6hdr _inner_iph;
cea67a2d 2296 struct icmp6hdr _icmph;
23aebdac
JS		 2297
2298 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2299 goto out;
2300
cea67a2d
ED		 2301 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2302 sizeof(_icmph), &_icmph);
2303 if (!icmph)
2304 goto out;
2305
23aebdac
JS		 2306 if (icmph->icmp6_type != ICMPV6_DEST_UNREACH &&
2307 icmph->icmp6_type != ICMPV6_PKT_TOOBIG &&
2308 icmph->icmp6_type != ICMPV6_TIME_EXCEED &&
2309 icmph->icmp6_type != ICMPV6_PARAMPROB)
2310 goto out;
2311
2312 inner_iph = skb_header_pointer(skb,
2313 skb_transport_offset(skb) + sizeof(*icmph),
2314 sizeof(_inner_iph), &_inner_iph);
2315 if (!inner_iph)
2316 goto out;
2317
2318 key_iph = inner_iph;
5e5d6fed 2319 _flkeys = NULL;
23aebdac 2320out:
5e5d6fed
RP		 2321 if (_flkeys) {
2322 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2323 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2324 keys->tags.flow_label = _flkeys->tags.flow_label;
2325 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2326 } else {
2327 keys->addrs.v6addrs.src = key_iph->saddr;
2328 keys->addrs.v6addrs.dst = key_iph->daddr;
fa1be7e0 2329 keys->tags.flow_label = ip6_flowlabel(key_iph);
5e5d6fed
RP		 2330 keys->basic.ip_proto = key_iph->nexthdr;
2331 }
23aebdac
JS		 2332}
2333
2334/* if skb is set it will be used and fl6 can be NULL */
b4bac172
DA		 2335u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2336 const struct sk_buff *skb, struct flow_keys *flkeys)
23aebdac
JS		 2337{
2338 struct flow_keys hash_keys;
9a2a537a 2339 u32 mhash;
23aebdac 2340
bbfa047a 2341 switch (ip6_multipath_hash_policy(net)) {
b4bac172
DA		 2342 case 0:
2343 memset(&hash_keys, 0, sizeof(hash_keys));
2344 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2345 if (skb) {
2346 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2347 } else {
2348 hash_keys.addrs.v6addrs.src = fl6->saddr;
2349 hash_keys.addrs.v6addrs.dst = fl6->daddr;
fa1be7e0 2350 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
b4bac172
DA		 2351 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2352 }
2353 break;
2354 case 1:
2355 if (skb) {
2356 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2357 struct flow_keys keys;
2358
2359 /* short-circuit if we already have L4 hash present */
2360 if (skb->l4_hash)
2361 return skb_get_hash_raw(skb) >> 1;
2362
2363 memset(&hash_keys, 0, sizeof(hash_keys));
2364
2365 if (!flkeys) {
2366 skb_flow_dissect_flow_keys(skb, &keys, flag);
2367 flkeys = &keys;
2368 }
2369 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2370 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2371 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2372 hash_keys.ports.src = flkeys->ports.src;
2373 hash_keys.ports.dst = flkeys->ports.dst;
2374 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2375 } else {
2376 memset(&hash_keys, 0, sizeof(hash_keys));
2377 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2378 hash_keys.addrs.v6addrs.src = fl6->saddr;
2379 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2380 hash_keys.ports.src = fl6->fl6_sport;
2381 hash_keys.ports.dst = fl6->fl6_dport;
2382 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2383 }
2384 break;
23aebdac 2385 }
9a2a537a 2386 mhash = flow_hash_from_keys(&hash_keys);
23aebdac 2387
9a2a537a 2388 return mhash >> 1;
23aebdac
JS		 2389}
2390
c71099ac
TG		 2391void ip6_route_input(struct sk_buff *skb)
2392{
b71d1d42 2393 const struct ipv6hdr *iph = ipv6_hdr(skb);
c346dca1 2394 struct net *net = dev_net(skb->dev);
adaa70bb 2395 int flags = RT6_LOOKUP_F_HAS_SADDR;
904af04d 2396 struct ip_tunnel_info *tun_info;
4c9483b2 2397 struct flowi6 fl6 = {
e0d56fdd 2398 .flowi6_iif = skb->dev->ifindex,
4c9483b2
DM		 2399 .daddr = iph->daddr,
2400 .saddr = iph->saddr,
6502ca52 2401 .flowlabel = ip6_flowinfo(iph),
4c9483b2
DM		 2402 .flowi6_mark = skb->mark,
2403 .flowi6_proto = iph->nexthdr,
c71099ac 2404 };
5e5d6fed 2405 struct flow_keys *flkeys = NULL, _flkeys;
adaa70bb 2406
904af04d 2407 tun_info = skb_tunnel_info(skb);
46fa062a 2408 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
904af04d 2409 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
5e5d6fed
RP		 2410
2411 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2412 flkeys = &_flkeys;
2413
23aebdac 2414 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
b4bac172 2415 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
06e9d040 2416 skb_dst_drop(skb);
b75cc8f9
DA		 2417 skb_dst_set(skb,
2418 ip6_route_input_lookup(net, skb->dev, &fl6, skb, flags));
c71099ac
TG		 2419}
2420
b75cc8f9
DA		 2421static struct rt6_info *ip6_pol_route_output(struct net *net,
2422 struct fib6_table *table,
2423 struct flowi6 *fl6,
2424 const struct sk_buff *skb,
2425 int flags)
1da177e4 2426{
b75cc8f9 2427 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
c71099ac
TG		 2428}
2429
6f21c96a
PA		 2430struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
2431 struct flowi6 *fl6, int flags)
c71099ac 2432{
d46a9d67 2433 bool any_src;
c71099ac 2434
3ede0bbc
RS		 2435 if (ipv6_addr_type(&fl6->daddr) &
2436 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
4c1feac5
DA		 2437 struct dst_entry *dst;
2438
2439 dst = l3mdev_link_scope_lookup(net, fl6);
2440 if (dst)
2441 return dst;
2442 }
ca254490 2443
1fb9489b 2444 fl6->flowi6_iif = LOOPBACK_IFINDEX;
4dc27d1c 2445
d46a9d67 2446 any_src = ipv6_addr_any(&fl6->saddr);
741a11d9 2447 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
d46a9d67 2448 (fl6->flowi6_oif && any_src))
77d16f45 2449 flags |= RT6_LOOKUP_F_IFACE;
c71099ac 2450
d46a9d67 2451 if (!any_src)
adaa70bb 2452 flags |= RT6_LOOKUP_F_HAS_SADDR;
0c9a2ac1
YH		 2453 else if (sk)
2454 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
adaa70bb 2455
b75cc8f9 2456 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
1da177e4 2457}
6f21c96a 2458EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1da177e4 2459
2774c131 2460struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
14e50e57 2461{
5c1e6aa3 2462 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1dbe3252 2463 struct net_device *loopback_dev = net->loopback_dev;
14e50e57
DM		 2464 struct dst_entry *new = NULL;
2465
1dbe3252 2466 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
62cf27e5 2467 DST_OBSOLETE_DEAD, 0);
14e50e57 2468 if (rt) {
0a1f5962 2469 rt6_info_init(rt);
81eb8447 2470 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
8104891b 2471
0a1f5962 2472 new = &rt->dst;
14e50e57 2473 new->__use = 1;
352e512c 2474 new->input = dst_discard;
ede2059d 2475 new->output = dst_discard_out;
14e50e57 2476
0a1f5962 2477 dst_copy_metrics(new, &ort->dst);
14e50e57 2478
1dbe3252 2479 rt->rt6i_idev = in6_dev_get(loopback_dev);
4e3fd7a0 2480 rt->rt6i_gateway = ort->rt6i_gateway;
0a1f5962 2481 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
14e50e57
DM		 2482
2483 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2484#ifdef CONFIG_IPV6_SUBTREES
2485 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2486#endif
14e50e57
DM		 2487 }
2488
69ead7af
DM		 2489 dst_release(dst_orig);
2490 return new ? new : ERR_PTR(-ENOMEM);
14e50e57 2491}
14e50e57 2492
1da177e4
LT		 2493/*
2494 * Destination cache support functions
2495 */
2496
8d1c802b 2497static bool fib6_check(struct fib6_info *f6i, u32 cookie)
4b32b5ad 2498{
93531c67
DA		 2499 u32 rt_cookie = 0;
2500
8ae86971 2501 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
93531c67
DA		 2502 return false;
2503
2504 if (fib6_check_expired(f6i))
2505 return false;
2506
2507 return true;
4b32b5ad
MKL		 2508}
2509
a68886a6
DA		 2510static struct dst_entry *rt6_check(struct rt6_info *rt,
2511 struct fib6_info *from,
2512 u32 cookie)
3da59bd9 2513{
36143645 2514 u32 rt_cookie = 0;
c5cff856 2515
a68886a6 2516 if ((from && !fib6_get_cookie_safe(from, &rt_cookie)) ||
93531c67 2517 rt_cookie != cookie)
3da59bd9
MKL		 2518 return NULL;
2519
2520 if (rt6_check_expired(rt))
2521 return NULL;
2522
2523 return &rt->dst;
2524}
2525
a68886a6
DA		 2526static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2527 struct fib6_info *from,
2528 u32 cookie)
3da59bd9 2529{
5973fb1e
MKL		 2530 if (!__rt6_check_expired(rt) &&
2531 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
a68886a6 2532 fib6_check(from, cookie))
3da59bd9
MKL		 2533 return &rt->dst;
2534 else
2535 return NULL;
2536}
2537
1da177e4
LT		 2538static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
2539{
a87b7dc9 2540 struct dst_entry *dst_ret;
a68886a6 2541 struct fib6_info *from;
1da177e4
LT		 2542 struct rt6_info *rt;
2543
a87b7dc9
DA		 2544 rt = container_of(dst, struct rt6_info, dst);
2545
2546 rcu_read_lock();
1da177e4 2547
6f3118b5
ND		 2548 /* All IPV6 dsts are created with ->obsolete set to the value
2549 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2550 * into this function always.
2551 */
e3bc10bd 2552
a68886a6 2553 from = rcu_dereference(rt->from);
4b32b5ad 2554
a68886a6
DA		 2555 if (from && (rt->rt6i_flags & RTF_PCPU ||
2556 unlikely(!list_empty(&rt->rt6i_uncached))))
2557 dst_ret = rt6_dst_from_check(rt, from, cookie);
3da59bd9 2558 else
a68886a6 2559 dst_ret = rt6_check(rt, from, cookie);
a87b7dc9
DA		 2560
2561 rcu_read_unlock();
2562
2563 return dst_ret;
1da177e4
LT		 2564}
2565
2566static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2567{
2568 struct rt6_info *rt = (struct rt6_info *) dst;
2569
2570 if (rt) {
54c1a859 2571 if (rt->rt6i_flags & RTF_CACHE) {
c3c14da0 2572 rcu_read_lock();
54c1a859 2573 if (rt6_check_expired(rt)) {
93531c67 2574 rt6_remove_exception_rt(rt);
54c1a859
YH		 2575 dst = NULL;
2576 }
c3c14da0 2577 rcu_read_unlock();
54c1a859 2578 } else {
1da177e4 2579 dst_release(dst);
54c1a859
YH		 2580 dst = NULL;
2581 }
1da177e4 2582 }
54c1a859 2583 return dst;
1da177e4
LT		 2584}
2585
2586static void ip6_link_failure(struct sk_buff *skb)
2587{
2588 struct rt6_info *rt;
2589
3ffe533c 2590 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1da177e4 2591
adf30907 2592 rt = (struct rt6_info *) skb_dst(skb);
1da177e4 2593 if (rt) {
8a14e46f 2594 rcu_read_lock();
1eb4f758 2595 if (rt->rt6i_flags & RTF_CACHE) {
761f6026 2596 rt6_remove_exception_rt(rt);
c5cff856 2597 } else {
a68886a6 2598 struct fib6_info *from;
c5cff856
WW		 2599 struct fib6_node *fn;
2600
a68886a6
DA		 2601 from = rcu_dereference(rt->from);
2602 if (from) {
2603 fn = rcu_dereference(from->fib6_node);
2604 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2605 fn->fn_sernum = -1;
2606 }
1eb4f758 2607 }
8a14e46f 2608 rcu_read_unlock();
1da177e4
LT		 2609 }
2610}
2611
6a3e030f
DA		 2612static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2613{
a68886a6
DA		 2614 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2615 struct fib6_info *from;
2616
2617 rcu_read_lock();
2618 from = rcu_dereference(rt0->from);
2619 if (from)
2620 rt0->dst.expires = from->expires;
2621 rcu_read_unlock();
2622 }
6a3e030f
DA		 2623
2624 dst_set_expires(&rt0->dst, timeout);
2625 rt0->rt6i_flags |= RTF_EXPIRES;
2626}
2627
45e4fd26
MKL		 2628static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2629{
2630 struct net *net = dev_net(rt->dst.dev);
2631
d4ead6b3 2632 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
45e4fd26 2633 rt->rt6i_flags |= RTF_MODIFIED;
45e4fd26
MKL		 2634 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2635}
2636
0d3f6d29
MKL		 2637static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2638{
2639 return !(rt->rt6i_flags & RTF_CACHE) &&
1490ed2a 2640 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
0d3f6d29
MKL		 2641}
2642
45e4fd26
MKL		 2643static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2644 const struct ipv6hdr *iph, u32 mtu)
1da177e4 2645{
0dec879f 2646 const struct in6_addr *daddr, *saddr;
67ba4152 2647 struct rt6_info *rt6 = (struct rt6_info *)dst;
1da177e4 2648
19bda36c
XL		 2649 if (dst_metric_locked(dst, RTAX_MTU))
2650 return;
2651
0dec879f
JA		 2652 if (iph) {
2653 daddr = &iph->daddr;
2654 saddr = &iph->saddr;
2655 } else if (sk) {
2656 daddr = &sk->sk_v6_daddr;
2657 saddr = &inet6_sk(sk)->saddr;
2658 } else {
2659 daddr = NULL;
2660 saddr = NULL;
2661 }
2662 dst_confirm_neigh(dst, daddr);
45e4fd26
MKL		 2663 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
2664 if (mtu >= dst_mtu(dst))
2665 return;
9d289715 2666
0d3f6d29 2667 if (!rt6_cache_allowed_for_pmtu(rt6)) {
45e4fd26 2668 rt6_do_update_pmtu(rt6, mtu);
2b760fcf
WW		 2669 /* update rt6_ex->stamp for cache */
2670 if (rt6->rt6i_flags & RTF_CACHE)
2671 rt6_update_exception_stamp_rt(rt6);
0dec879f 2672 } else if (daddr) {
85bd05de 2673 struct fib6_result res = {};
45e4fd26
MKL		 2674 struct rt6_info *nrt6;
2675
4d85cd0c 2676 rcu_read_lock();
85bd05de
DA		 2677 res.f6i = rcu_dereference(rt6->from);
2678 if (!res.f6i) {
9c69a132
JL		 2679 rcu_read_unlock();
2680 return;
2681 }
7d21fec9
DA		 2682 res.fib6_flags = res.f6i->fib6_flags;
2683 res.fib6_type = res.f6i->fib6_type;
2684
2d44234b
DA		 2685 if (res.f6i->nh) {
2686 struct fib6_nh_match_arg arg = {
2687 .dev = dst->dev,
2688 .gw = &rt6->rt6i_gateway,
2689 };
2690
2691 nexthop_for_each_fib6_nh(res.f6i->nh,
2692 fib6_nh_find_match, &arg);
2693
2694 /* fib6_info uses a nexthop that does not have fib6_nh
2695 * using the dst->dev + gw. Should be impossible.
2696 */
2697 if (!arg.match) {
2698 rcu_read_unlock();
2699 return;
2700 }
2701
2702 res.nh = arg.match;
2703 } else {
2704 res.nh = res.f6i->fib6_nh;
2705 }
2706
85bd05de 2707 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
45e4fd26
MKL		 2708 if (nrt6) {
2709 rt6_do_update_pmtu(nrt6, mtu);
5012f0a5 2710 if (rt6_insert_exception(nrt6, &res))
2b760fcf 2711 dst_release_immediate(&nrt6->dst);
45e4fd26 2712 }
a68886a6 2713 rcu_read_unlock();
1da177e4
LT		 2714 }
2715}
2716
45e4fd26
MKL		 2717static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2718 struct sk_buff *skb, u32 mtu)
2719{
2720 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
2721}
2722
42ae66c8 2723void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
e2d118a1 2724 int oif, u32 mark, kuid_t uid)
81aded24
DM		 2725{
2726 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2727 struct dst_entry *dst;
dc92095d
MŻ		 2728 struct flowi6 fl6 = {
2729 .flowi6_oif = oif,
2730 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2731 .daddr = iph->daddr,
2732 .saddr = iph->saddr,
2733 .flowlabel = ip6_flowinfo(iph),
2734 .flowi6_uid = uid,
2735 };
81aded24
DM		 2736
2737 dst = ip6_route_output(net, NULL, &fl6);
2738 if (!dst->error)
45e4fd26 2739 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
81aded24
DM		 2740 dst_release(dst);
2741}
2742EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2743
2744void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2745{
7ddacfa5 2746 int oif = sk->sk_bound_dev_if;
33c162a9
MKL		 2747 struct dst_entry *dst;
2748
7ddacfa5
DA		 2749 if (!oif && skb->dev)
2750 oif = l3mdev_master_ifindex(skb->dev);
2751
2752 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
33c162a9
MKL		 2753
2754 dst = __sk_dst_get(sk);
2755 if (!dst || !dst->obsolete ||
2756 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2757 return;
2758
2759 bh_lock_sock(sk);
2760 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2761 ip6_datagram_dst_update(sk, false);
2762 bh_unlock_sock(sk);
81aded24
DM		 2763}
2764EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2765
7d6850f7
AK		 2766void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2767 const struct flowi6 *fl6)
2768{
2769#ifdef CONFIG_IPV6_SUBTREES
2770 struct ipv6_pinfo *np = inet6_sk(sk);
2771#endif
2772
2773 ip6_dst_store(sk, dst,
2774 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2775 &sk->sk_v6_daddr : NULL,
2776#ifdef CONFIG_IPV6_SUBTREES
2777 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2778 &np->saddr :
2779#endif
2780 NULL);
2781}
2782
9b6b35ab 2783static bool ip6_redirect_nh_match(const struct fib6_result *res,
0b34eb00
DA		 2784 struct flowi6 *fl6,
2785 const struct in6_addr *gw,
2786 struct rt6_info **ret)
2787{
9b6b35ab
DA		 2788 const struct fib6_nh *nh = res->nh;
2789
0b34eb00
DA		 2790 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2791 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2792 return false;
2793
2794 /* rt_cache's gateway might be different from its 'parent'
2795 * in the case of an ip redirect.
2796 * So we keep searching in the exception table if the gateway
2797 * is different.
2798 */
2799 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
2800 struct rt6_info *rt_cache;
2801
9b6b35ab 2802 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
0b34eb00
DA		 2803 if (rt_cache &&
2804 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
2805 *ret = rt_cache;
2806 return true;
2807 }
2808 return false;
2809 }
2810 return true;
2811}
2812
c55c8988
DA		 2813struct fib6_nh_rd_arg {
2814 struct fib6_result *res;
2815 struct flowi6 *fl6;
2816 const struct in6_addr *gw;
2817 struct rt6_info **ret;
2818};
2819
2820static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
2821{
2822 struct fib6_nh_rd_arg *arg = _arg;
2823
2824 arg->res->nh = nh;
2825 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
2826}
2827
b55b76b2
DJ		 2828/* Handle redirects */
2829struct ip6rd_flowi {
2830 struct flowi6 fl6;
2831 struct in6_addr gateway;
2832};
2833
2834static struct rt6_info *__ip6_route_redirect(struct net *net,
2835 struct fib6_table *table,
2836 struct flowi6 *fl6,
b75cc8f9 2837 const struct sk_buff *skb,
b55b76b2
DJ		 2838 int flags)
2839{
2840 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
0b34eb00 2841 struct rt6_info *ret = NULL;
9b6b35ab 2842 struct fib6_result res = {};
c55c8988
DA		 2843 struct fib6_nh_rd_arg arg = {
2844 .res = &res,
2845 .fl6 = fl6,
2846 .gw = &rdfl->gateway,
2847 .ret = &ret
2848 };
8d1c802b 2849 struct fib6_info *rt;
b55b76b2
DJ		 2850 struct fib6_node *fn;
2851
31680ac2
DA		 2852 /* l3mdev_update_flow overrides oif if the device is enslaved; in
2853 * this case we must match on the real ingress device, so reset it
2854 */
2855 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2856 fl6->flowi6_oif = skb->dev->ifindex;
2857
b55b76b2 2858 /* Get the "current" route for this destination and
67c408cf 2859 * check if the redirect has come from appropriate router.
b55b76b2
DJ		 2860 *
2861 * RFC 4861 specifies that redirects should only be
2862 * accepted if they come from the nexthop to the target.
2863 * Due to the way the routes are chosen, this notion
2864 * is a bit fuzzy and one might need to check all possible
2865 * routes.
2866 */
2867
66f5d6ce 2868 rcu_read_lock();
6454743b 2869 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
b55b76b2 2870restart:
66f5d6ce 2871 for_each_fib6_node_rt_rcu(fn) {
9b6b35ab 2872 res.f6i = rt;
14895687 2873 if (fib6_check_expired(rt))
b55b76b2 2874 continue;
93c2fb25 2875 if (rt->fib6_flags & RTF_REJECT)
b55b76b2 2876 break;
c55c8988
DA		 2877 if (unlikely(rt->nh)) {
2878 if (nexthop_is_blackhole(rt->nh))
2879 continue;
2880 /* on match, res->nh is filled in and potentially ret */
2881 if (nexthop_for_each_fib6_nh(rt->nh,
2882 fib6_nh_redirect_match,
2883 &arg))
2884 goto out;
2885 } else {
2886 res.nh = rt->fib6_nh;
2887 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
2888 &ret))
2889 goto out;
2890 }
b55b76b2
DJ		 2891 }
2892
2893 if (!rt)
421842ed 2894 rt = net->ipv6.fib6_null_entry;
93c2fb25 2895 else if (rt->fib6_flags & RTF_REJECT) {
23fb93a4 2896 ret = net->ipv6.ip6_null_entry;
b0a1ba59
MKL		 2897 goto out;
2898 }
2899
421842ed 2900 if (rt == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2901 fn = fib6_backtrack(fn, &fl6->saddr);
2902 if (fn)
2903 goto restart;
b55b76b2 2904 }
a3c00e46 2905
9b6b35ab 2906 res.f6i = rt;
1cf844c7 2907 res.nh = rt->fib6_nh;
b0a1ba59 2908out:
7d21fec9 2909 if (ret) {
10585b43 2910 ip6_hold_safe(net, &ret);
7d21fec9
DA		 2911 } else {
2912 res.fib6_flags = res.f6i->fib6_flags;
2913 res.fib6_type = res.f6i->fib6_type;
9b6b35ab 2914 ret = ip6_create_rt_rcu(&res);
7d21fec9 2915 }
b55b76b2 2916
66f5d6ce 2917 rcu_read_unlock();
b55b76b2 2918
8ff2e5b2 2919 trace_fib6_table_lookup(net, &res, table, fl6);
23fb93a4 2920 return ret;
b55b76b2
DJ		 2921};
2922
2923static struct dst_entry *ip6_route_redirect(struct net *net,
b75cc8f9
DA		 2924 const struct flowi6 *fl6,
2925 const struct sk_buff *skb,
2926 const struct in6_addr *gateway)
b55b76b2
DJ		 2927{
2928 int flags = RT6_LOOKUP_F_HAS_SADDR;
2929 struct ip6rd_flowi rdfl;
2930
2931 rdfl.fl6 = *fl6;
2932 rdfl.gateway = *gateway;
2933
b75cc8f9 2934 return fib6_rule_lookup(net, &rdfl.fl6, skb,
b55b76b2
DJ		 2935 flags, __ip6_route_redirect);
2936}
2937
e2d118a1
LC		 2938void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
2939 kuid_t uid)
3a5ad2ee
DM		 2940{
2941 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2942 struct dst_entry *dst;
1f7f10ac
MŻ		 2943 struct flowi6 fl6 = {
2944 .flowi6_iif = LOOPBACK_IFINDEX,
2945 .flowi6_oif = oif,
2946 .flowi6_mark = mark,
2947 .daddr = iph->daddr,
2948 .saddr = iph->saddr,
2949 .flowlabel = ip6_flowinfo(iph),
2950 .flowi6_uid = uid,
2951 };
3a5ad2ee 2952
b75cc8f9 2953 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
b55b76b2 2954 rt6_do_redirect(dst, NULL, skb);
3a5ad2ee
DM		 2955 dst_release(dst);
2956}
2957EXPORT_SYMBOL_GPL(ip6_redirect);
2958
d456336d 2959void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
c92a59ec
DJ		 2960{
2961 const struct ipv6hdr *iph = ipv6_hdr(skb);
2962 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
2963 struct dst_entry *dst;
0b26fb17
MŻ		 2964 struct flowi6 fl6 = {
2965 .flowi6_iif = LOOPBACK_IFINDEX,
2966 .flowi6_oif = oif,
0b26fb17
MŻ		 2967 .daddr = msg->dest,
2968 .saddr = iph->daddr,
2969 .flowi6_uid = sock_net_uid(net, NULL),
2970 };
c92a59ec 2971
b75cc8f9 2972 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
b55b76b2 2973 rt6_do_redirect(dst, NULL, skb);
c92a59ec
DJ		 2974 dst_release(dst);
2975}
2976
3a5ad2ee
DM		 2977void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
2978{
e2d118a1
LC		 2979 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
2980 sk->sk_uid);
3a5ad2ee
DM		 2981}
2982EXPORT_SYMBOL_GPL(ip6_sk_redirect);
2983
0dbaee3b 2984static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1da177e4 2985{
0dbaee3b
DM		 2986 struct net_device *dev = dst->dev;
2987 unsigned int mtu = dst_mtu(dst);
2988 struct net *net = dev_net(dev);
2989
1da177e4
LT		 2990 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
2991
5578689a
DL		 2992 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
2993 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1da177e4
LT		 2994
2995 /*
1ab1457c
YH		 2996 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
2997 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
2998 * IPV6_MAXPLEN is also valid and means: "any MSS,
1da177e4
LT		 2999 * rely only on pmtu discovery"
3000 */
3001 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3002 mtu = IPV6_MAXPLEN;
3003 return mtu;
3004}
3005
ebb762f2 3006static unsigned int ip6_mtu(const struct dst_entry *dst)
d33e4553 3007{
d33e4553 3008 struct inet6_dev *idev;
d4ead6b3 3009 unsigned int mtu;
4b32b5ad
MKL		 3010
3011 mtu = dst_metric_raw(dst, RTAX_MTU);
618f9bc7 3012 if (mtu)
30f78d8e 3013 goto out;
618f9bc7
SK		 3014
3015 mtu = IPV6_MIN_MTU;
d33e4553
DM		 3016
3017 rcu_read_lock();
3018 idev = __in6_dev_get(dst->dev);
3019 if (idev)
3020 mtu = idev->cnf.mtu6;
3021 rcu_read_unlock();
3022
30f78d8e 3023out:
14972cbd
RP		 3024 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3025
3026 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
d33e4553
DM		 3027}
3028
901731b8
DA		 3029/* MTU selection:
3030 * 1. mtu on route is locked - use it
3031 * 2. mtu from nexthop exception
3032 * 3. mtu from egress device
3033 *
3034 * based on ip6_dst_mtu_forward and exception logic of
3035 * rt6_find_cached_rt; called with rcu_read_lock
3036 */
b748f260
DA		 3037u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3038 const struct in6_addr *daddr,
3039 const struct in6_addr *saddr)
901731b8 3040{
b748f260
DA		 3041 const struct fib6_nh *nh = res->nh;
3042 struct fib6_info *f6i = res->f6i;
901731b8 3043 struct inet6_dev *idev;
510e2ced 3044 struct rt6_info *rt;
901731b8
DA		 3045 u32 mtu = 0;
3046
3047 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3048 mtu = f6i->fib6_pmtu;
3049 if (mtu)
3050 goto out;
3051 }
3052
510e2ced
WW		 3053 rt = rt6_find_cached_rt(res, daddr, saddr);
3054 if (unlikely(rt)) {
3055 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3056 } else {
b748f260 3057 struct net_device *dev = nh->fib_nh_dev;
901731b8
DA		 3058
3059 mtu = IPV6_MIN_MTU;
3060 idev = __in6_dev_get(dev);
3061 if (idev && idev->cnf.mtu6 > mtu)
3062 mtu = idev->cnf.mtu6;
3063 }
3064
3065 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3066out:
b748f260 3067 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
901731b8
DA		 3068}
3069
3b00944c 3070struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
87a11578 3071 struct flowi6 *fl6)
1da177e4 3072{
87a11578 3073 struct dst_entry *dst;
1da177e4
LT		 3074 struct rt6_info *rt;
3075 struct inet6_dev *idev = in6_dev_get(dev);
c346dca1 3076 struct net *net = dev_net(dev);
1da177e4 3077
38308473 3078 if (unlikely(!idev))
122bdf67 3079 return ERR_PTR(-ENODEV);
1da177e4 3080
ad706862 3081 rt = ip6_dst_alloc(net, dev, 0);
38308473 3082 if (unlikely(!rt)) {
1da177e4 3083 in6_dev_put(idev);
87a11578 3084 dst = ERR_PTR(-ENOMEM);
1da177e4
LT		 3085 goto out;
3086 }
3087
8e2ec639 3088 rt->dst.flags |= DST_HOST;
588753f1 3089 rt->dst.input = ip6_input;
8e2ec639 3090 rt->dst.output = ip6_output;
550bab42 3091 rt->rt6i_gateway = fl6->daddr;
87a11578 3092 rt->rt6i_dst.addr = fl6->daddr;
8e2ec639
YZ		 3093 rt->rt6i_dst.plen = 128;
3094 rt->rt6i_idev = idev;
14edd87d 3095 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1da177e4 3096
4c981e28 3097 /* Add this dst into uncached_list so that rt6_disable_ip() can
587fea74
WW		 3098 * do proper release of the net_device
3099 */
3100 rt6_uncached_list_add(rt);
81eb8447 3101 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
1da177e4 3102
87a11578
DM		 3103 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3104
1da177e4 3105out:
87a11578 3106 return dst;
1da177e4
LT		 3107}
3108
569d3645 3109static int ip6_dst_gc(struct dst_ops *ops)
1da177e4 3110{
86393e52 3111 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
7019b78e
DL		 3112 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3113 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3114 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3115 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3116 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
fc66f95c 3117 int entries;
7019b78e 3118
fc66f95c 3119 entries = dst_entries_get_fast(ops);
49a18d86 3120 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
fc66f95c 3121 entries <= rt_max_size)
1da177e4
LT		 3122 goto out;
3123
6891a346 3124 net->ipv6.ip6_rt_gc_expire++;
14956643 3125 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
fc66f95c
ED		 3126 entries = dst_entries_get_slow(ops);
3127 if (entries < ops->gc_thresh)
7019b78e 3128 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1da177e4 3129out:
7019b78e 3130 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
fc66f95c 3131 return entries > rt_max_size;
1da177e4
LT		 3132}
3133
8c14586f
DA		 3134static struct rt6_info *ip6_nh_lookup_table(struct net *net,
3135 struct fib6_config *cfg,
f4797b33
DA		 3136 const struct in6_addr *gw_addr,
3137 u32 tbid, int flags)
8c14586f
DA		 3138{
3139 struct flowi6 fl6 = {
3140 .flowi6_oif = cfg->fc_ifindex,
3141 .daddr = *gw_addr,
3142 .saddr = cfg->fc_prefsrc,
3143 };
3144 struct fib6_table *table;
3145 struct rt6_info *rt;
8c14586f 3146
f4797b33 3147 table = fib6_get_table(net, tbid);
8c14586f
DA		 3148 if (!table)
3149 return NULL;
3150
3151 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3152 flags |= RT6_LOOKUP_F_HAS_SADDR;
3153
f4797b33 3154 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
b75cc8f9 3155 rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, NULL, flags);
8c14586f
DA		 3156
3157 /* if table lookup failed, fall back to full lookup */
3158 if (rt == net->ipv6.ip6_null_entry) {
3159 ip6_rt_put(rt);
3160 rt = NULL;
3161 }
3162
3163 return rt;
3164}
3165
fc1e64e1
DA		 3166static int ip6_route_check_nh_onlink(struct net *net,
3167 struct fib6_config *cfg,
9fbb704c 3168 const struct net_device *dev,
fc1e64e1
DA		 3169 struct netlink_ext_ack *extack)
3170{
44750f84 3171 u32 tbid = l3mdev_fib_table(dev) ? : RT_TABLE_MAIN;
fc1e64e1
DA		 3172 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3173 u32 flags = RTF_LOCAL | RTF_ANYCAST | RTF_REJECT;
bf1dc8ba 3174 struct fib6_info *from;
fc1e64e1
DA		 3175 struct rt6_info *grt;
3176 int err;
3177
3178 err = 0;
3179 grt = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0);
3180 if (grt) {
bf1dc8ba
PA		 3181 rcu_read_lock();
3182 from = rcu_dereference(grt->from);
58e354c0 3183 if (!grt->dst.error &&
4ed591c8 3184 /* ignore match if it is the default route */
bf1dc8ba 3185 from && !ipv6_addr_any(&from->fib6_dst.addr) &&
58e354c0 3186 (grt->rt6i_flags & flags || dev != grt->dst.dev)) {
44750f84
DA		 3187 NL_SET_ERR_MSG(extack,
3188 "Nexthop has invalid gateway or device mismatch");
fc1e64e1
DA		 3189 err = -EINVAL;
3190 }
bf1dc8ba 3191 rcu_read_unlock();
fc1e64e1
DA		 3192
3193 ip6_rt_put(grt);
3194 }
3195
3196 return err;
3197}
3198
1edce99f
DA		 3199static int ip6_route_check_nh(struct net *net,
3200 struct fib6_config *cfg,
3201 struct net_device **_dev,
3202 struct inet6_dev **idev)
3203{
3204 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3205 struct net_device *dev = _dev ? *_dev : NULL;
3206 struct rt6_info *grt = NULL;
3207 int err = -EHOSTUNREACH;
3208
3209 if (cfg->fc_table) {
f4797b33
DA		 3210 int flags = RT6_LOOKUP_F_IFACE;
3211
3212 grt = ip6_nh_lookup_table(net, cfg, gw_addr,
3213 cfg->fc_table, flags);
1edce99f
DA		 3214 if (grt) {
3215 if (grt->rt6i_flags & RTF_GATEWAY ||
3216 (dev && dev != grt->dst.dev)) {
3217 ip6_rt_put(grt);
3218 grt = NULL;
3219 }
3220 }
3221 }
3222
3223 if (!grt)
b75cc8f9 3224 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, NULL, 1);
1edce99f
DA		 3225
3226 if (!grt)
3227 goto out;
3228
3229 if (dev) {
3230 if (dev != grt->dst.dev) {
3231 ip6_rt_put(grt);
3232 goto out;
3233 }
3234 } else {
3235 *_dev = dev = grt->dst.dev;
3236 *idev = grt->rt6i_idev;
3237 dev_hold(dev);
3238 in6_dev_hold(grt->rt6i_idev);
3239 }
3240
3241 if (!(grt->rt6i_flags & RTF_GATEWAY))
3242 err = 0;
3243
3244 ip6_rt_put(grt);
3245
3246out:
3247 return err;
3248}
3249
9fbb704c
DA		 3250static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3251 struct net_device **_dev, struct inet6_dev **idev,
3252 struct netlink_ext_ack *extack)
3253{
3254 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3255 int gwa_type = ipv6_addr_type(gw_addr);
232378e8 3256 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
9fbb704c 3257 const struct net_device *dev = *_dev;
232378e8 3258 bool need_addr_check = !dev;
9fbb704c
DA		 3259 int err = -EINVAL;
3260
3261 /* if gw_addr is local we will fail to detect this in case
3262 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3263 * will return already-added prefix route via interface that
3264 * prefix route was assigned to, which might be non-loopback.
3265 */
232378e8
DA		 3266 if (dev &&
3267 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3268 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
9fbb704c
DA		 3269 goto out;
3270 }
3271
3272 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3273 /* IPv6 strictly inhibits using not link-local
3274 * addresses as nexthop address.
3275 * Otherwise, router will not able to send redirects.
3276 * It is very good, but in some (rare!) circumstances
3277 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3278 * some exceptions. --ANK
3279 * We allow IPv4-mapped nexthops to support RFC4798-type
3280 * addressing
3281 */
3282 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3283 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3284 goto out;
3285 }
3286
3287 if (cfg->fc_flags & RTNH_F_ONLINK)
3288 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3289 else
3290 err = ip6_route_check_nh(net, cfg, _dev, idev);
3291
3292 if (err)
3293 goto out;
3294 }
3295
3296 /* reload in case device was changed */
3297 dev = *_dev;
3298
3299 err = -EINVAL;
3300 if (!dev) {
3301 NL_SET_ERR_MSG(extack, "Egress device not specified");
3302 goto out;
3303 } else if (dev->flags & IFF_LOOPBACK) {
3304 NL_SET_ERR_MSG(extack,
3305 "Egress device can not be loopback device for this route");
3306 goto out;
3307 }
232378e8
DA		 3308
3309 /* if we did not check gw_addr above, do so now that the
3310 * egress device has been resolved.
3311 */
3312 if (need_addr_check &&
3313 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3314 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3315 goto out;
3316 }
3317
9fbb704c
DA		 3318 err = 0;
3319out:
3320 return err;
3321}
3322
83c44251
DA		 3323static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3324{
3325 if ((flags & RTF_REJECT) ||
3326 (dev && (dev->flags & IFF_LOOPBACK) &&
3327 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3328 !(flags & RTF_LOCAL)))
3329 return true;
3330
3331 return false;
3332}
3333
3334int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3335 struct fib6_config *cfg, gfp_t gfp_flags,
3336 struct netlink_ext_ack *extack)
3337{
3338 struct net_device *dev = NULL;
3339 struct inet6_dev *idev = NULL;
3340 int addr_type;
3341 int err;
3342
f1741730
DA		 3343 fib6_nh->fib_nh_family = AF_INET6;
3344
83c44251
DA		 3345 err = -ENODEV;
3346 if (cfg->fc_ifindex) {
3347 dev = dev_get_by_index(net, cfg->fc_ifindex);
3348 if (!dev)
3349 goto out;
3350 idev = in6_dev_get(dev);
3351 if (!idev)
3352 goto out;
3353 }
3354
3355 if (cfg->fc_flags & RTNH_F_ONLINK) {
3356 if (!dev) {
3357 NL_SET_ERR_MSG(extack,
3358 "Nexthop device required for onlink");
3359 goto out;
3360 }
3361
3362 if (!(dev->flags & IFF_UP)) {
3363 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3364 err = -ENETDOWN;
3365 goto out;
3366 }
3367
ad1601ae 3368 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
83c44251
DA		 3369 }
3370
ad1601ae 3371 fib6_nh->fib_nh_weight = 1;
83c44251
DA		 3372
3373 /* We cannot add true routes via loopback here,
3374 * they would result in kernel looping; promote them to reject routes
3375 */
3376 addr_type = ipv6_addr_type(&cfg->fc_dst);
3377 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3378 /* hold loopback dev/idev if we haven't done so. */
3379 if (dev != net->loopback_dev) {
3380 if (dev) {
3381 dev_put(dev);
3382 in6_dev_put(idev);
3383 }
3384 dev = net->loopback_dev;
3385 dev_hold(dev);
3386 idev = in6_dev_get(dev);
3387 if (!idev) {
3388 err = -ENODEV;
3389 goto out;
3390 }
3391 }
7dd73168 3392 goto pcpu_alloc;
83c44251
DA		 3393 }
3394
3395 if (cfg->fc_flags & RTF_GATEWAY) {
3396 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3397 if (err)
3398 goto out;
3399
ad1601ae 3400 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
bdf00467 3401 fib6_nh->fib_nh_gw_family = AF_INET6;
83c44251
DA		 3402 }
3403
3404 err = -ENODEV;
3405 if (!dev)
3406 goto out;
3407
3408 if (idev->cnf.disable_ipv6) {
3409 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3410 err = -EACCES;
3411 goto out;
3412 }
3413
3414 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3415 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3416 err = -ENETDOWN;
3417 goto out;
3418 }
3419
3420 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3421 !netif_carrier_ok(dev))
ad1601ae 3422 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
83c44251 3423
7dd73168
DA		 3424 err = fib_nh_common_init(&fib6_nh->nh_common, cfg->fc_encap,
3425 cfg->fc_encap_type, cfg, gfp_flags, extack);
3426 if (err)
3427 goto out;
3428
3429pcpu_alloc:
f40b6ae2
DA		 3430 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3431 if (!fib6_nh->rt6i_pcpu) {
3432 err = -ENOMEM;
3433 goto out;
3434 }
3435
ad1601ae 3436 fib6_nh->fib_nh_dev = dev;
f1741730 3437 fib6_nh->fib_nh_oif = dev->ifindex;
83c44251
DA		 3438 err = 0;
3439out:
3440 if (idev)
3441 in6_dev_put(idev);
3442
3443 if (err) {
ad1601ae
DA		 3444 lwtstate_put(fib6_nh->fib_nh_lws);
3445 fib6_nh->fib_nh_lws = NULL;
83c44251
DA		 3446 if (dev)
3447 dev_put(dev);
3448 }
3449
3450 return err;
3451}
3452
dac7d0f2
DA		 3453void fib6_nh_release(struct fib6_nh *fib6_nh)
3454{
cc5c073a
DA		 3455 struct rt6_exception_bucket *bucket;
3456
3457 rcu_read_lock();
3458
3459 fib6_nh_flush_exceptions(fib6_nh, NULL);
3460 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3461 if (bucket) {
3462 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3463 kfree(bucket);
3464 }
3465
3466 rcu_read_unlock();
3467
f40b6ae2
DA		 3468 if (fib6_nh->rt6i_pcpu) {
3469 int cpu;
3470
3471 for_each_possible_cpu(cpu) {
3472 struct rt6_info **ppcpu_rt;
3473 struct rt6_info *pcpu_rt;
3474
3475 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3476 pcpu_rt = *ppcpu_rt;
3477 if (pcpu_rt) {
3478 dst_dev_put(&pcpu_rt->dst);
3479 dst_release(&pcpu_rt->dst);
3480 *ppcpu_rt = NULL;
3481 }
3482 }
3483
3484 free_percpu(fib6_nh->rt6i_pcpu);
3485 }
3486
979e276e 3487 fib_nh_common_release(&fib6_nh->nh_common);
dac7d0f2
DA		 3488}
3489
8d1c802b 3490static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
acb54e3c 3491 gfp_t gfp_flags,
333c4301 3492 struct netlink_ext_ack *extack)
1da177e4 3493{
5578689a 3494 struct net *net = cfg->fc_nlinfo.nl_net;
8d1c802b 3495 struct fib6_info *rt = NULL;
f88d8ea6 3496 struct nexthop *nh = NULL;
c71099ac 3497 struct fib6_table *table;
f88d8ea6 3498 struct fib6_nh *fib6_nh;
8c5b83f0 3499 int err = -EINVAL;
83c44251 3500 int addr_type;
1da177e4 3501
557c44be 3502 /* RTF_PCPU is an internal flag; can not be set by userspace */
d5d531cb
DA		 3503 if (cfg->fc_flags & RTF_PCPU) {
3504 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
557c44be 3505 goto out;
d5d531cb 3506 }
557c44be 3507
2ea2352e
WW		 3508 /* RTF_CACHE is an internal flag; can not be set by userspace */
3509 if (cfg->fc_flags & RTF_CACHE) {
3510 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3511 goto out;
3512 }
3513
e8478e80
DA		 3514 if (cfg->fc_type > RTN_MAX) {
3515 NL_SET_ERR_MSG(extack, "Invalid route type");
3516 goto out;
3517 }
3518
d5d531cb
DA		 3519 if (cfg->fc_dst_len > 128) {
3520 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3521 goto out;
3522 }
3523 if (cfg->fc_src_len > 128) {
3524 NL_SET_ERR_MSG(extack, "Invalid source address length");
8c5b83f0 3525 goto out;
d5d531cb 3526 }
1da177e4 3527#ifndef CONFIG_IPV6_SUBTREES
d5d531cb
DA		 3528 if (cfg->fc_src_len) {
3529 NL_SET_ERR_MSG(extack,
3530 "Specifying source address requires IPV6_SUBTREES to be enabled");
8c5b83f0 3531 goto out;
d5d531cb 3532 }
1da177e4 3533#endif
5b98324e
DA		 3534 if (cfg->fc_nh_id) {
3535 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3536 if (!nh) {
3537 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3538 goto out;
3539 }
3540 err = fib6_check_nexthop(nh, cfg, extack);
3541 if (err)
3542 goto out;
3543 }
fc1e64e1 3544
d71314b4 3545 err = -ENOBUFS;
38308473
DM		 3546 if (cfg->fc_nlinfo.nlh &&
3547 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
d71314b4 3548 table = fib6_get_table(net, cfg->fc_table);
38308473 3549 if (!table) {
f3213831 3550 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
d71314b4
MV		 3551 table = fib6_new_table(net, cfg->fc_table);
3552 }
3553 } else {
3554 table = fib6_new_table(net, cfg->fc_table);
3555 }
38308473
DM		 3556
3557 if (!table)
c71099ac 3558 goto out;
c71099ac 3559
93531c67 3560 err = -ENOMEM;
f88d8ea6 3561 rt = fib6_info_alloc(gfp_flags, !nh);
93531c67 3562 if (!rt)
1da177e4 3563 goto out;
93531c67 3564
d7e774f3
DA		 3565 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3566 extack);
767a2217
DA		 3567 if (IS_ERR(rt->fib6_metrics)) {
3568 err = PTR_ERR(rt->fib6_metrics);
fda21d46
ED		 3569 /* Do not leave garbage there. */
3570 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
767a2217
DA		 3571 goto out;
3572 }
3573
93531c67
DA		 3574 if (cfg->fc_flags & RTF_ADDRCONF)
3575 rt->dst_nocount = true;
1da177e4 3576
1716a961 3577 if (cfg->fc_flags & RTF_EXPIRES)
14895687 3578 fib6_set_expires(rt, jiffies +
1716a961
G		 3579 clock_t_to_jiffies(cfg->fc_expires));
3580 else
14895687 3581 fib6_clean_expires(rt);
1da177e4 3582
86872cb5
TG		 3583 if (cfg->fc_protocol == RTPROT_UNSPEC)
3584 cfg->fc_protocol = RTPROT_BOOT;
93c2fb25 3585 rt->fib6_protocol = cfg->fc_protocol;
86872cb5 3586
83c44251
DA		 3587 rt->fib6_table = table;
3588 rt->fib6_metric = cfg->fc_metric;
3589 rt->fib6_type = cfg->fc_type;
2b2450ca 3590 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
19e42e45 3591
93c2fb25
DA		 3592 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3593 rt->fib6_dst.plen = cfg->fc_dst_len;
3594 if (rt->fib6_dst.plen == 128)
3b6761d1 3595 rt->dst_host = true;
e5fd387a 3596
1da177e4 3597#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 3598 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3599 rt->fib6_src.plen = cfg->fc_src_len;
1da177e4 3600#endif
f88d8ea6
DA		 3601 if (nh) {
3602 if (!nexthop_get(nh)) {
3603 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3604 goto out;
3605 }
3606 if (rt->fib6_src.plen) {
4daa95af 3607 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
f88d8ea6
DA		 3608 goto out;
3609 }
3610 rt->nh = nh;
3611 fib6_nh = nexthop_fib6_nh(rt->nh);
3612 } else {
3613 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3614 if (err)
3615 goto out;
1da177e4 3616
f88d8ea6
DA		 3617 fib6_nh = rt->fib6_nh;
3618
3619 /* We cannot add true routes via loopback here, they would
3620 * result in kernel looping; promote them to reject routes
3621 */
3622 addr_type = ipv6_addr_type(&cfg->fc_dst);
3623 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3624 addr_type))
3625 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3626 }
955ec4cb 3627
c3968a85 3628 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
f88d8ea6 3629 struct net_device *dev = fib6_nh->fib_nh_dev;
83c44251 3630
c3968a85 3631 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
d5d531cb 3632 NL_SET_ERR_MSG(extack, "Invalid source address");
c3968a85
DW		 3633 err = -EINVAL;
3634 goto out;
3635 }
93c2fb25
DA		 3636 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3637 rt->fib6_prefsrc.plen = 128;
c3968a85 3638 } else
93c2fb25 3639 rt->fib6_prefsrc.plen = 0;
c3968a85 3640
8c5b83f0 3641 return rt;
6b9ea5a6 3642out:
93531c67 3643 fib6_info_release(rt);
8c5b83f0 3644 return ERR_PTR(err);
6b9ea5a6
RP		 3645}
3646
acb54e3c 3647int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
333c4301 3648 struct netlink_ext_ack *extack)
6b9ea5a6 3649{
8d1c802b 3650 struct fib6_info *rt;
6b9ea5a6
RP		 3651 int err;
3652
acb54e3c 3653 rt = ip6_route_info_create(cfg, gfp_flags, extack);
d4ead6b3
DA		 3654 if (IS_ERR(rt))
3655 return PTR_ERR(rt);
6b9ea5a6 3656
d4ead6b3 3657 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
93531c67 3658 fib6_info_release(rt);
6b9ea5a6 3659
1da177e4
LT		 3660 return err;
3661}
3662
8d1c802b 3663static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
1da177e4 3664{
afb1d4b5 3665 struct net *net = info->nl_net;
c71099ac 3666 struct fib6_table *table;
afb1d4b5 3667 int err;
1da177e4 3668
421842ed 3669 if (rt == net->ipv6.fib6_null_entry) {
6825a26c
G		 3670 err = -ENOENT;
3671 goto out;
3672 }
6c813a72 3673
93c2fb25 3674 table = rt->fib6_table;
66f5d6ce 3675 spin_lock_bh(&table->tb6_lock);
86872cb5 3676 err = fib6_del(rt, info);
66f5d6ce 3677 spin_unlock_bh(&table->tb6_lock);
1da177e4 3678
6825a26c 3679out:
93531c67 3680 fib6_info_release(rt);
1da177e4
LT		 3681 return err;
3682}
3683
8d1c802b 3684int ip6_del_rt(struct net *net, struct fib6_info *rt)
e0a1ad73 3685{
afb1d4b5
DA		 3686 struct nl_info info = { .nl_net = net };
3687
528c4ceb 3688 return __ip6_del_rt(rt, &info);
e0a1ad73
TG		 3689}
3690
8d1c802b 3691static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
0ae81335
DA		 3692{
3693 struct nl_info *info = &cfg->fc_nlinfo;
e3330039 3694 struct net *net = info->nl_net;
16a16cd3 3695 struct sk_buff *skb = NULL;
0ae81335 3696 struct fib6_table *table;
e3330039 3697 int err = -ENOENT;
0ae81335 3698
421842ed 3699 if (rt == net->ipv6.fib6_null_entry)
e3330039 3700 goto out_put;
93c2fb25 3701 table = rt->fib6_table;
66f5d6ce 3702 spin_lock_bh(&table->tb6_lock);
0ae81335 3703
93c2fb25 3704 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
8d1c802b 3705 struct fib6_info *sibling, *next_sibling;
0ae81335 3706
16a16cd3
DA		 3707 /* prefer to send a single notification with all hops */
3708 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3709 if (skb) {
3710 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3711
d4ead6b3 3712 if (rt6_fill_node(net, skb, rt, NULL,
16a16cd3
DA		 3713 NULL, NULL, 0, RTM_DELROUTE,
3714 info->portid, seq, 0) < 0) {
3715 kfree_skb(skb);
3716 skb = NULL;
3717 } else
3718 info->skip_notify = 1;
3719 }
3720
0ae81335 3721 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25
DA		 3722 &rt->fib6_siblings,
3723 fib6_siblings) {
0ae81335
DA		 3724 err = fib6_del(sibling, info);
3725 if (err)
e3330039 3726 goto out_unlock;
0ae81335
DA		 3727 }
3728 }
3729
3730 err = fib6_del(rt, info);
e3330039 3731out_unlock:
66f5d6ce 3732 spin_unlock_bh(&table->tb6_lock);
e3330039 3733out_put:
93531c67 3734 fib6_info_release(rt);
16a16cd3
DA		 3735
3736 if (skb) {
e3330039 3737 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
16a16cd3
DA		 3738 info->nlh, gfp_any());
3739 }
0ae81335
DA		 3740 return err;
3741}
3742
0fa6efc5 3743static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
23fb93a4
DA		 3744{
3745 int rc = -ESRCH;
3746
3747 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3748 goto out;
3749
3750 if (cfg->fc_flags & RTF_GATEWAY &&
3751 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3752 goto out;
761f6026
XL		 3753
3754 rc = rt6_remove_exception_rt(rt);
23fb93a4
DA		 3755out:
3756 return rc;
3757}
3758
0fa6efc5
DA		 3759static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3760 struct fib6_nh *nh)
3761{
3762 struct fib6_result res = {
3763 .f6i = rt,
3764 .nh = nh,
3765 };
3766 struct rt6_info *rt_cache;
3767
3768 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3769 if (rt_cache)
3770 return __ip6_del_cached_rt(rt_cache, cfg);
3771
3772 return 0;
3773}
3774
5b98324e
DA		 3775struct fib6_nh_del_cached_rt_arg {
3776 struct fib6_config *cfg;
3777 struct fib6_info *f6i;
3778};
3779
3780static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
3781{
3782 struct fib6_nh_del_cached_rt_arg *arg = _arg;
3783 int rc;
3784
3785 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
3786 return rc != -ESRCH ? rc : 0;
3787}
3788
3789static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
3790{
3791 struct fib6_nh_del_cached_rt_arg arg = {
3792 .cfg = cfg,
3793 .f6i = f6i
3794 };
3795
3796 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
3797}
3798
333c4301
DA		 3799static int ip6_route_del(struct fib6_config *cfg,
3800 struct netlink_ext_ack *extack)
1da177e4 3801{
c71099ac 3802 struct fib6_table *table;
8d1c802b 3803 struct fib6_info *rt;
1da177e4 3804 struct fib6_node *fn;
1da177e4
LT		 3805 int err = -ESRCH;
3806
5578689a 3807 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
d5d531cb
DA		 3808 if (!table) {
3809 NL_SET_ERR_MSG(extack, "FIB table does not exist");
c71099ac 3810 return err;
d5d531cb 3811 }
c71099ac 3812
66f5d6ce 3813 rcu_read_lock();
1da177e4 3814
c71099ac 3815 fn = fib6_locate(&table->tb6_root,
86872cb5 3816 &cfg->fc_dst, cfg->fc_dst_len,
38fbeeee 3817 &cfg->fc_src, cfg->fc_src_len,
2b760fcf 3818 !(cfg->fc_flags & RTF_CACHE));
1ab1457c 3819
1da177e4 3820 if (fn) {
66f5d6ce 3821 for_each_fib6_node_rt_rcu(fn) {
ad1601ae
DA		 3822 struct fib6_nh *nh;
3823
5b98324e
DA		 3824 if (rt->nh && rt->nh->id != cfg->fc_nh_id)
3825 continue;
23fb93a4 3826
5b98324e
DA		 3827 if (cfg->fc_flags & RTF_CACHE) {
3828 int rc = 0;
3829
3830 if (rt->nh) {
3831 rc = ip6_del_cached_rt_nh(cfg, rt);
3832 } else if (cfg->fc_nh_id) {
3833 continue;
3834 } else {
3835 nh = rt->fib6_nh;
3836 rc = ip6_del_cached_rt(cfg, rt, nh);
3837 }
0fa6efc5
DA		 3838 if (rc != -ESRCH) {
3839 rcu_read_unlock();
3840 return rc;
23fb93a4
DA		 3841 }
3842 continue;
2b760fcf 3843 }
ad1601ae 3844
5b98324e
DA		 3845 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
3846 continue;
3847 if (cfg->fc_protocol &&
3848 cfg->fc_protocol != rt->fib6_protocol)
3849 continue;
3850
3851 if (rt->nh) {
3852 if (!fib6_info_hold_safe(rt))
3853 continue;
3854 rcu_read_unlock();
3855
3856 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3857 }
3858 if (cfg->fc_nh_id)
3859 continue;
3860
3861 nh = rt->fib6_nh;
86872cb5 3862 if (cfg->fc_ifindex &&
ad1601ae
DA		 3863 (!nh->fib_nh_dev ||
3864 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
1da177e4 3865 continue;
86872cb5 3866 if (cfg->fc_flags & RTF_GATEWAY &&
ad1601ae 3867 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
1da177e4 3868 continue;
e873e4b9
WW		 3869 if (!fib6_info_hold_safe(rt))
3870 continue;
66f5d6ce 3871 rcu_read_unlock();
1da177e4 3872
0ae81335
DA		 3873 /* if gateway was specified only delete the one hop */
3874 if (cfg->fc_flags & RTF_GATEWAY)
3875 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3876
3877 return __ip6_del_rt_siblings(rt, cfg);
1da177e4
LT		 3878 }
3879 }
66f5d6ce 3880 rcu_read_unlock();
1da177e4
LT		 3881
3882 return err;
3883}
3884
6700c270 3885static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
a6279458 3886{
a6279458 3887 struct netevent_redirect netevent;
e8599ff4 3888 struct rt6_info *rt, *nrt = NULL;
85bd05de 3889 struct fib6_result res = {};
e8599ff4
DM		 3890 struct ndisc_options ndopts;
3891 struct inet6_dev *in6_dev;
3892 struct neighbour *neigh;
71bcdba0 3893 struct rd_msg *msg;
6e157b6a
DM		 3894 int optlen, on_link;
3895 u8 *lladdr;
e8599ff4 3896
29a3cad5 3897 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
71bcdba0 3898 optlen -= sizeof(*msg);
e8599ff4
DM		 3899
3900 if (optlen < 0) {
6e157b6a 3901 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
e8599ff4
DM		 3902 return;
3903 }
3904
71bcdba0 3905 msg = (struct rd_msg *)icmp6_hdr(skb);
e8599ff4 3906
71bcdba0 3907 if (ipv6_addr_is_multicast(&msg->dest)) {
6e157b6a 3908 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
e8599ff4
DM		 3909 return;
3910 }
3911
6e157b6a 3912 on_link = 0;
71bcdba0 3913 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
e8599ff4 3914 on_link = 1;
71bcdba0 3915 } else if (ipv6_addr_type(&msg->target) !=
e8599ff4 3916 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
6e157b6a 3917 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
e8599ff4
DM		 3918 return;
3919 }
3920
3921 in6_dev = __in6_dev_get(skb->dev);
3922 if (!in6_dev)
3923 return;
3924 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
3925 return;
3926
3927 /* RFC2461 8.1:
3928 * The IP source address of the Redirect MUST be the same as the current
3929 * first-hop router for the specified ICMP Destination Address.
3930 */
3931
f997c55c 3932 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
e8599ff4
DM		 3933 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
3934 return;
3935 }
6e157b6a
DM		 3936
3937 lladdr = NULL;
e8599ff4
DM		 3938 if (ndopts.nd_opts_tgt_lladdr) {
3939 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
3940 skb->dev);
3941 if (!lladdr) {
3942 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
3943 return;
3944 }
3945 }
3946
6e157b6a 3947 rt = (struct rt6_info *) dst;
ec13ad1d 3948 if (rt->rt6i_flags & RTF_REJECT) {
6e157b6a 3949 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
e8599ff4 3950 return;
6e157b6a 3951 }
e8599ff4 3952
6e157b6a
DM		 3953 /* Redirect received -> path was valid.
3954 * Look, redirects are sent only in response to data packets,
3955 * so that this nexthop apparently is reachable. --ANK
3956 */
0dec879f 3957 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
a6279458 3958
71bcdba0 3959 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
6e157b6a
DM		 3960 if (!neigh)
3961 return;
a6279458 3962
1da177e4
LT		 3963 /*
3964 * We have finally decided to accept it.
3965 */
3966
f997c55c 3967 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1da177e4
LT		 3968 NEIGH_UPDATE_F_WEAK_OVERRIDE|
3969 NEIGH_UPDATE_F_OVERRIDE|
3970 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
f997c55c
AA		 3971 NEIGH_UPDATE_F_ISROUTER)),
3972 NDISC_REDIRECT, &ndopts);
1da177e4 3973
4d85cd0c 3974 rcu_read_lock();
85bd05de 3975 res.f6i = rcu_dereference(rt->from);
ff24e498 3976 if (!res.f6i)
886b7a50 3977 goto out;
8a14e46f 3978
49d5b8ef
DA		 3979 if (res.f6i->nh) {
3980 struct fib6_nh_match_arg arg = {
3981 .dev = dst->dev,
3982 .gw = &rt->rt6i_gateway,
3983 };
3984
3985 nexthop_for_each_fib6_nh(res.f6i->nh,
3986 fib6_nh_find_match, &arg);
3987
3988 /* fib6_info uses a nexthop that does not have fib6_nh
3989 * using the dst->dev. Should be impossible
3990 */
3991 if (!arg.match)
3992 goto out;
3993 res.nh = arg.match;
3994 } else {
3995 res.nh = res.f6i->fib6_nh;
3996 }
3997
7d21fec9
DA		 3998 res.fib6_flags = res.f6i->fib6_flags;
3999 res.fib6_type = res.f6i->fib6_type;
85bd05de 4000 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
38308473 4001 if (!nrt)
1da177e4
LT		 4002 goto out;
4003
4004 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4005 if (on_link)
4006 nrt->rt6i_flags &= ~RTF_GATEWAY;
4007
4e3fd7a0 4008 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1da177e4 4009
886b7a50 4010 /* rt6_insert_exception() will take care of duplicated exceptions */
5012f0a5 4011 if (rt6_insert_exception(nrt, &res)) {
2b760fcf
WW		 4012 dst_release_immediate(&nrt->dst);
4013 goto out;
4014 }
1da177e4 4015
d8d1f30b
CG		 4016 netevent.old = &rt->dst;
4017 netevent.new = &nrt->dst;
71bcdba0 4018 netevent.daddr = &msg->dest;
60592833 4019 netevent.neigh = neigh;
8d71740c
TT		 4020 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4021
1da177e4 4022out:
886b7a50 4023 rcu_read_unlock();
e8599ff4 4024 neigh_release(neigh);
6e157b6a
DM		 4025}
4026
70ceb4f5 4027#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 4028static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 4029 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 4030 const struct in6_addr *gwaddr,
4031 struct net_device *dev)
70ceb4f5 4032{
830218c1
DA		 4033 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4034 int ifindex = dev->ifindex;
70ceb4f5 4035 struct fib6_node *fn;
8d1c802b 4036 struct fib6_info *rt = NULL;
c71099ac
TG		 4037 struct fib6_table *table;
4038
830218c1 4039 table = fib6_get_table(net, tb_id);
38308473 4040 if (!table)
c71099ac 4041 return NULL;
70ceb4f5 4042
66f5d6ce 4043 rcu_read_lock();
38fbeeee 4044 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
70ceb4f5
YH		 4045 if (!fn)
4046 goto out;
4047
66f5d6ce 4048 for_each_fib6_node_rt_rcu(fn) {
f88d8ea6
DA		 4049 /* these routes do not use nexthops */
4050 if (rt->nh)
4051 continue;
1cf844c7 4052 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
70ceb4f5 4053 continue;
2b2450ca 4054 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
1cf844c7 4055 !rt->fib6_nh->fib_nh_gw_family)
70ceb4f5 4056 continue;
1cf844c7 4057 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
70ceb4f5 4058 continue;
e873e4b9
WW		 4059 if (!fib6_info_hold_safe(rt))
4060 continue;
70ceb4f5
YH		 4061 break;
4062 }
4063out:
66f5d6ce 4064 rcu_read_unlock();
70ceb4f5
YH		 4065 return rt;
4066}
4067
8d1c802b 4068static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 4069 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 4070 const struct in6_addr *gwaddr,
4071 struct net_device *dev,
95c96174 4072 unsigned int pref)
70ceb4f5 4073{
86872cb5 4074 struct fib6_config cfg = {
238fc7ea 4075 .fc_metric = IP6_RT_PRIO_USER,
830218c1 4076 .fc_ifindex = dev->ifindex,
86872cb5
TG		 4077 .fc_dst_len = prefixlen,
4078 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4079 RTF_UP | RTF_PREF(pref),
b91d5329 4080 .fc_protocol = RTPROT_RA,
e8478e80 4081 .fc_type = RTN_UNICAST,
15e47304 4082 .fc_nlinfo.portid = 0,
efa2cea0
DL		 4083 .fc_nlinfo.nlh = NULL,
4084 .fc_nlinfo.nl_net = net,
86872cb5
TG		 4085 };
4086
830218c1 4087 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
4e3fd7a0
AD		 4088 cfg.fc_dst = *prefix;
4089 cfg.fc_gateway = *gwaddr;
70ceb4f5 4090
e317da96
YH		 4091 /* We should treat it as a default route if prefix length is 0. */
4092 if (!prefixlen)
86872cb5 4093 cfg.fc_flags |= RTF_DEFAULT;
70ceb4f5 4094
acb54e3c 4095 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
70ceb4f5 4096
830218c1 4097 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
70ceb4f5
YH		 4098}
4099#endif
4100
8d1c802b 4101struct fib6_info *rt6_get_dflt_router(struct net *net,
afb1d4b5
DA		 4102 const struct in6_addr *addr,
4103 struct net_device *dev)
1ab1457c 4104{
830218c1 4105 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
8d1c802b 4106 struct fib6_info *rt;
c71099ac 4107 struct fib6_table *table;
1da177e4 4108
afb1d4b5 4109 table = fib6_get_table(net, tb_id);
38308473 4110 if (!table)
c71099ac 4111 return NULL;
1da177e4 4112
66f5d6ce
WW		 4113 rcu_read_lock();
4114 for_each_fib6_node_rt_rcu(&table->tb6_root) {
f88d8ea6 4115 struct fib6_nh *nh;
ad1601ae 4116
f88d8ea6
DA		 4117 /* RA routes do not use nexthops */
4118 if (rt->nh)
4119 continue;
4120
4121 nh = rt->fib6_nh;
ad1601ae 4122 if (dev == nh->fib_nh_dev &&
93c2fb25 4123 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
ad1601ae 4124 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
1da177e4
LT		 4125 break;
4126 }
e873e4b9
WW		 4127 if (rt && !fib6_info_hold_safe(rt))
4128 rt = NULL;
66f5d6ce 4129 rcu_read_unlock();
1da177e4
LT		 4130 return rt;
4131}
4132
8d1c802b 4133struct fib6_info *rt6_add_dflt_router(struct net *net,
afb1d4b5 4134 const struct in6_addr *gwaddr,
ebacaaa0
YH		 4135 struct net_device *dev,
4136 unsigned int pref)
1da177e4 4137{
86872cb5 4138 struct fib6_config cfg = {
ca254490 4139 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
238fc7ea 4140 .fc_metric = IP6_RT_PRIO_USER,
86872cb5
TG		 4141 .fc_ifindex = dev->ifindex,
4142 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4143 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
b91d5329 4144 .fc_protocol = RTPROT_RA,
e8478e80 4145 .fc_type = RTN_UNICAST,
15e47304 4146 .fc_nlinfo.portid = 0,
5578689a 4147 .fc_nlinfo.nlh = NULL,
afb1d4b5 4148 .fc_nlinfo.nl_net = net,
86872cb5 4149 };
1da177e4 4150
4e3fd7a0 4151 cfg.fc_gateway = *gwaddr;
1da177e4 4152
acb54e3c 4153 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
830218c1
DA		 4154 struct fib6_table *table;
4155
4156 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4157 if (table)
4158 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4159 }
1da177e4 4160
afb1d4b5 4161 return rt6_get_dflt_router(net, gwaddr, dev);
1da177e4
LT		 4162}
4163
afb1d4b5
DA		 4164static void __rt6_purge_dflt_routers(struct net *net,
4165 struct fib6_table *table)
1da177e4 4166{
8d1c802b 4167 struct fib6_info *rt;
1da177e4
LT		 4168
4169restart:
66f5d6ce
WW		 4170 rcu_read_lock();
4171 for_each_fib6_node_rt_rcu(&table->tb6_root) {
dcd1f572
DA		 4172 struct net_device *dev = fib6_info_nh_dev(rt);
4173 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4174
93c2fb25 4175 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
e873e4b9
WW		 4176 (!idev || idev->cnf.accept_ra != 2) &&
4177 fib6_info_hold_safe(rt)) {
93531c67
DA		 4178 rcu_read_unlock();
4179 ip6_del_rt(net, rt);
1da177e4
LT		 4180 goto restart;
4181 }
4182 }
66f5d6ce 4183 rcu_read_unlock();
830218c1
DA		 4184
4185 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4186}
4187
4188void rt6_purge_dflt_routers(struct net *net)
4189{
4190 struct fib6_table *table;
4191 struct hlist_head *head;
4192 unsigned int h;
4193
4194 rcu_read_lock();
4195
4196 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4197 head = &net->ipv6.fib_table_hash[h];
4198 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4199 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
afb1d4b5 4200 __rt6_purge_dflt_routers(net, table);
830218c1
DA		 4201 }
4202 }
4203
4204 rcu_read_unlock();
1da177e4
LT		 4205}
4206
5578689a
DL		 4207static void rtmsg_to_fib6_config(struct net *net,
4208 struct in6_rtmsg *rtmsg,
86872cb5
TG		 4209 struct fib6_config *cfg)
4210{
8823a3ac
MŻ		 4211 *cfg = (struct fib6_config){
4212 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4213 : RT6_TABLE_MAIN,
4214 .fc_ifindex = rtmsg->rtmsg_ifindex,
67f69513 4215 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
8823a3ac
MŻ		 4216 .fc_expires = rtmsg->rtmsg_info,
4217 .fc_dst_len = rtmsg->rtmsg_dst_len,
4218 .fc_src_len = rtmsg->rtmsg_src_len,
4219 .fc_flags = rtmsg->rtmsg_flags,
4220 .fc_type = rtmsg->rtmsg_type,
4221
4222 .fc_nlinfo.nl_net = net,
4223
4224 .fc_dst = rtmsg->rtmsg_dst,
4225 .fc_src = rtmsg->rtmsg_src,
4226 .fc_gateway = rtmsg->rtmsg_gateway,
4227 };
86872cb5
TG		 4228}
4229
5578689a 4230int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1da177e4 4231{
86872cb5 4232 struct fib6_config cfg;
1da177e4
LT		 4233 struct in6_rtmsg rtmsg;
4234 int err;
4235
67ba4152 4236 switch (cmd) {
1da177e4
LT		 4237 case SIOCADDRT: /* Add a route */
4238 case SIOCDELRT: /* Delete a route */
af31f412 4239 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4
LT		 4240 return -EPERM;
4241 err = copy_from_user(&rtmsg, arg,
4242 sizeof(struct in6_rtmsg));
4243 if (err)
4244 return -EFAULT;
86872cb5 4245
5578689a 4246 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
86872cb5 4247
1da177e4
LT		 4248 rtnl_lock();
4249 switch (cmd) {
4250 case SIOCADDRT:
acb54e3c 4251 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
1da177e4
LT		 4252 break;
4253 case SIOCDELRT:
333c4301 4254 err = ip6_route_del(&cfg, NULL);
1da177e4
LT		 4255 break;
4256 default:
4257 err = -EINVAL;
4258 }
4259 rtnl_unlock();
4260
4261 return err;
3ff50b79 4262 }
1da177e4
LT		 4263
4264 return -EINVAL;
4265}
4266
4267/*
4268 * Drop the packet on the floor
4269 */
4270
d5fdd6ba 4271static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1da177e4 4272{
adf30907 4273 struct dst_entry *dst = skb_dst(skb);
1d3fd8a1
SS		 4274 struct net *net = dev_net(dst->dev);
4275 struct inet6_dev *idev;
4276 int type;
4277
4278 if (netif_is_l3_master(skb->dev) &&
4279 dst->dev == net->loopback_dev)
4280 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4281 else
4282 idev = ip6_dst_idev(dst);
4283
612f09e8
YH		 4284 switch (ipstats_mib_noroutes) {
4285 case IPSTATS_MIB_INNOROUTES:
0660e03f 4286 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
45bb0060 4287 if (type == IPV6_ADDR_ANY) {
1d3fd8a1 4288 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
612f09e8
YH		 4289 break;
4290 }
4291 /* FALLTHROUGH */
4292 case IPSTATS_MIB_OUTNOROUTES:
1d3fd8a1 4293 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
612f09e8
YH		 4294 break;
4295 }
1d3fd8a1
SS		 4296
4297 /* Start over by dropping the dst for l3mdev case */
4298 if (netif_is_l3_master(skb->dev))
4299 skb_dst_drop(skb);
4300
3ffe533c 4301 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1da177e4
LT		 4302 kfree_skb(skb);
4303 return 0;
4304}
4305
9ce8ade0
TG		 4306static int ip6_pkt_discard(struct sk_buff *skb)
4307{
612f09e8 4308 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4309}
4310
ede2059d 4311static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
1da177e4 4312{
adf30907 4313 skb->dev = skb_dst(skb)->dev;
612f09e8 4314 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1da177e4
LT		 4315}
4316
9ce8ade0
TG		 4317static int ip6_pkt_prohibit(struct sk_buff *skb)
4318{
612f09e8 4319 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4320}
4321
ede2059d 4322static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
9ce8ade0 4323{
adf30907 4324 skb->dev = skb_dst(skb)->dev;
612f09e8 4325 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
9ce8ade0
TG		 4326}
4327
1da177e4
LT		 4328/*
4329 * Allocate a dst for local (unicast / anycast) address.
4330 */
4331
360a9887
DA		 4332struct fib6_info *addrconf_f6i_alloc(struct net *net,
4333 struct inet6_dev *idev,
4334 const struct in6_addr *addr,
4335 bool anycast, gfp_t gfp_flags)
1da177e4 4336{
c7a1ce39
DA		 4337 struct fib6_config cfg = {
4338 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4339 .fc_ifindex = idev->dev->ifindex,
4340 .fc_flags = RTF_UP | RTF_ADDRCONF | RTF_NONEXTHOP,
4341 .fc_dst = *addr,
4342 .fc_dst_len = 128,
4343 .fc_protocol = RTPROT_KERNEL,
4344 .fc_nlinfo.nl_net = net,
4345 .fc_ignore_dev_down = true,
4346 };
1da177e4 4347
e8478e80 4348 if (anycast) {
c7a1ce39
DA		 4349 cfg.fc_type = RTN_ANYCAST;
4350 cfg.fc_flags |= RTF_ANYCAST;
e8478e80 4351 } else {
c7a1ce39
DA		 4352 cfg.fc_type = RTN_LOCAL;
4353 cfg.fc_flags |= RTF_LOCAL;
e8478e80 4354 }
1da177e4 4355
c7a1ce39 4356 return ip6_route_info_create(&cfg, gfp_flags, NULL);
1da177e4
LT		 4357}
4358
c3968a85
DW		 4359/* remove deleted ip from prefsrc entries */
4360struct arg_dev_net_ip {
4361 struct net_device *dev;
4362 struct net *net;
4363 struct in6_addr *addr;
4364};
4365
8d1c802b 4366static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
c3968a85
DW		 4367{
4368 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4369 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4370 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4371
f88d8ea6
DA		 4372 if (!rt->nh &&
4373 ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
421842ed 4374 rt != net->ipv6.fib6_null_entry &&
93c2fb25 4375 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
60006a48 4376 spin_lock_bh(&rt6_exception_lock);
c3968a85 4377 /* remove prefsrc entry */
93c2fb25 4378 rt->fib6_prefsrc.plen = 0;
60006a48 4379 spin_unlock_bh(&rt6_exception_lock);
c3968a85
DW		 4380 }
4381 return 0;
4382}
4383
4384void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4385{
4386 struct net *net = dev_net(ifp->idev->dev);
4387 struct arg_dev_net_ip adni = {
4388 .dev = ifp->idev->dev,
4389 .net = net,
4390 .addr = &ifp->addr,
4391 };
0c3584d5 4392 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
c3968a85
DW		 4393}
4394
2b2450ca 4395#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
be7a010d
DJ		 4396
4397/* Remove routers and update dst entries when gateway turn into host. */
8d1c802b 4398static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
be7a010d
DJ		 4399{
4400 struct in6_addr *gateway = (struct in6_addr *)arg;
f88d8ea6
DA		 4401 struct fib6_nh *nh;
4402
4403 /* RA routes do not use nexthops */
4404 if (rt->nh)
4405 return 0;
be7a010d 4406
f88d8ea6 4407 nh = rt->fib6_nh;
93c2fb25 4408 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
cc5c073a 4409 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
be7a010d 4410 return -1;
b16cb459
WW		 4411
4412 /* Further clean up cached routes in exception table.
4413 * This is needed because cached route may have a different
4414 * gateway than its 'parent' in the case of an ip redirect.
4415 */
cc5c073a 4416 fib6_nh_exceptions_clean_tohost(nh, gateway);
b16cb459 4417
be7a010d
DJ		 4418 return 0;
4419}
4420
4421void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4422{
4423 fib6_clean_all(net, fib6_clean_tohost, gateway);
4424}
4425
2127d95a
IS		 4426struct arg_netdev_event {
4427 const struct net_device *dev;
4c981e28 4428 union {
ecc5663c 4429 unsigned char nh_flags;
4c981e28
IS		 4430 unsigned long event;
4431 };
2127d95a
IS		 4432};
4433
8d1c802b 4434static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
d7dedee1 4435{
8d1c802b 4436 struct fib6_info *iter;
d7dedee1
IS		 4437 struct fib6_node *fn;
4438
93c2fb25
DA		 4439 fn = rcu_dereference_protected(rt->fib6_node,
4440 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4441 iter = rcu_dereference_protected(fn->leaf,
93c2fb25 4442 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4443 while (iter) {
93c2fb25 4444 if (iter->fib6_metric == rt->fib6_metric &&
33bd5ac5 4445 rt6_qualify_for_ecmp(iter))
d7dedee1 4446 return iter;
8fb11a9a 4447 iter = rcu_dereference_protected(iter->fib6_next,
93c2fb25 4448 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1
IS		 4449 }
4450
4451 return NULL;
4452}
4453
f88d8ea6 4454/* only called for fib entries with builtin fib6_nh */
8d1c802b 4455static bool rt6_is_dead(const struct fib6_info *rt)
d7dedee1 4456{
1cf844c7
DA		 4457 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4458 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4459 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
d7dedee1
IS		 4460 return true;
4461
4462 return false;
4463}
4464
8d1c802b 4465static int rt6_multipath_total_weight(const struct fib6_info *rt)
d7dedee1 4466{
8d1c802b 4467 struct fib6_info *iter;
d7dedee1
IS		 4468 int total = 0;
4469
4470 if (!rt6_is_dead(rt))
1cf844c7 4471 total += rt->fib6_nh->fib_nh_weight;
d7dedee1 4472
93c2fb25 4473 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
d7dedee1 4474 if (!rt6_is_dead(iter))
1cf844c7 4475 total += iter->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4476 }
4477
4478 return total;
4479}
4480
8d1c802b 4481static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
d7dedee1
IS		 4482{
4483 int upper_bound = -1;
4484
4485 if (!rt6_is_dead(rt)) {
1cf844c7 4486 *weight += rt->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4487 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4488 total) - 1;
4489 }
1cf844c7 4490 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
d7dedee1
IS		 4491}
4492
8d1c802b 4493static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
d7dedee1 4494{
8d1c802b 4495 struct fib6_info *iter;
d7dedee1
IS		 4496 int weight = 0;
4497
4498 rt6_upper_bound_set(rt, &weight, total);
4499
93c2fb25 4500 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
d7dedee1
IS		 4501 rt6_upper_bound_set(iter, &weight, total);
4502}
4503
8d1c802b 4504void rt6_multipath_rebalance(struct fib6_info *rt)
d7dedee1 4505{
8d1c802b 4506 struct fib6_info *first;
d7dedee1
IS		 4507 int total;
4508
4509 /* In case the entire multipath route was marked for flushing,
4510 * then there is no need to rebalance upon the removal of every
4511 * sibling route.
4512 */
93c2fb25 4513 if (!rt->fib6_nsiblings || rt->should_flush)
d7dedee1
IS		 4514 return;
4515
4516 /* During lookup routes are evaluated in order, so we need to
4517 * make sure upper bounds are assigned from the first sibling
4518 * onwards.
4519 */
4520 first = rt6_multipath_first_sibling(rt);
4521 if (WARN_ON_ONCE(!first))
4522 return;
4523
4524 total = rt6_multipath_total_weight(first);
4525 rt6_multipath_upper_bound_set(first, total);
4526}
4527
8d1c802b 4528static int fib6_ifup(struct fib6_info *rt, void *p_arg)
2127d95a
IS		 4529{
4530 const struct arg_netdev_event *arg = p_arg;
7aef6859 4531 struct net *net = dev_net(arg->dev);
2127d95a 4532
f88d8ea6 4533 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
1cf844c7
DA		 4534 rt->fib6_nh->fib_nh_dev == arg->dev) {
4535 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
7aef6859 4536 fib6_update_sernum_upto_root(net, rt);
d7dedee1 4537 rt6_multipath_rebalance(rt);
1de178ed 4538 }
2127d95a
IS		 4539
4540 return 0;
4541}
4542
ecc5663c 4543void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
2127d95a
IS		 4544{
4545 struct arg_netdev_event arg = {
4546 .dev = dev,
6802f3ad
IS		 4547 {
4548 .nh_flags = nh_flags,
4549 },
2127d95a
IS		 4550 };
4551
4552 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4553 arg.nh_flags |= RTNH_F_LINKDOWN;
4554
4555 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4556}
4557
f88d8ea6 4558/* only called for fib entries with inline fib6_nh */
8d1c802b 4559static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
1de178ed
IS		 4560 const struct net_device *dev)
4561{
8d1c802b 4562 struct fib6_info *iter;
1de178ed 4563
1cf844c7 4564 if (rt->fib6_nh->fib_nh_dev == dev)
1de178ed 4565 return true;
93c2fb25 4566 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7 4567 if (iter->fib6_nh->fib_nh_dev == dev)
1de178ed
IS		 4568 return true;
4569
4570 return false;
4571}
4572
8d1c802b 4573static void rt6_multipath_flush(struct fib6_info *rt)
1de178ed 4574{
8d1c802b 4575 struct fib6_info *iter;
1de178ed
IS		 4576
4577 rt->should_flush = 1;
93c2fb25 4578 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1de178ed
IS		 4579 iter->should_flush = 1;
4580}
4581
8d1c802b 4582static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
1de178ed
IS		 4583 const struct net_device *down_dev)
4584{
8d1c802b 4585 struct fib6_info *iter;
1de178ed
IS		 4586 unsigned int dead = 0;
4587
1cf844c7
DA		 4588 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4589 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed 4590 dead++;
93c2fb25 4591 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4592 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4593 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed
IS		 4594 dead++;
4595
4596 return dead;
4597}
4598
8d1c802b 4599static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
1de178ed 4600 const struct net_device *dev,
ecc5663c 4601 unsigned char nh_flags)
1de178ed 4602{
8d1c802b 4603 struct fib6_info *iter;
1de178ed 4604
1cf844c7
DA		 4605 if (rt->fib6_nh->fib_nh_dev == dev)
4606 rt->fib6_nh->fib_nh_flags |= nh_flags;
93c2fb25 4607 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4608 if (iter->fib6_nh->fib_nh_dev == dev)
4609 iter->fib6_nh->fib_nh_flags |= nh_flags;
1de178ed
IS		 4610}
4611
a1a22c12 4612/* called with write lock held for table with rt */
8d1c802b 4613static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
1da177e4 4614{
4c981e28
IS		 4615 const struct arg_netdev_event *arg = p_arg;
4616 const struct net_device *dev = arg->dev;
7aef6859 4617 struct net *net = dev_net(dev);
8ed67789 4618
f88d8ea6 4619 if (rt == net->ipv6.fib6_null_entry || rt->nh)
27c6fa73
IS		 4620 return 0;
4621
4622 switch (arg->event) {
4623 case NETDEV_UNREGISTER:
1cf844c7 4624 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
27c6fa73 4625 case NETDEV_DOWN:
1de178ed 4626 if (rt->should_flush)
27c6fa73 4627 return -1;
93c2fb25 4628 if (!rt->fib6_nsiblings)
1cf844c7 4629 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
1de178ed
IS		 4630 if (rt6_multipath_uses_dev(rt, dev)) {
4631 unsigned int count;
4632
4633 count = rt6_multipath_dead_count(rt, dev);
93c2fb25 4634 if (rt->fib6_nsiblings + 1 == count) {
1de178ed
IS		 4635 rt6_multipath_flush(rt);
4636 return -1;
4637 }
4638 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4639 RTNH_F_LINKDOWN);
7aef6859 4640 fib6_update_sernum(net, rt);
d7dedee1 4641 rt6_multipath_rebalance(rt);
1de178ed
IS		 4642 }
4643 return -2;
27c6fa73 4644 case NETDEV_CHANGE:
1cf844c7 4645 if (rt->fib6_nh->fib_nh_dev != dev ||
93c2fb25 4646 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
27c6fa73 4647 break;
1cf844c7 4648 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
d7dedee1 4649 rt6_multipath_rebalance(rt);
27c6fa73 4650 break;
2b241361 4651 }
c159d30c 4652
1da177e4
LT		 4653 return 0;
4654}
4655
27c6fa73 4656void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
1da177e4 4657{
4c981e28 4658 struct arg_netdev_event arg = {
8ed67789 4659 .dev = dev,
6802f3ad
IS		 4660 {
4661 .event = event,
4662 },
8ed67789 4663 };
7c6bb7d2 4664 struct net *net = dev_net(dev);
8ed67789 4665
7c6bb7d2
DA		 4666 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4667 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4668 else
4669 fib6_clean_all(net, fib6_ifdown, &arg);
4c981e28
IS		 4670}
4671
4672void rt6_disable_ip(struct net_device *dev, unsigned long event)
4673{
4674 rt6_sync_down_dev(dev, event);
4675 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4676 neigh_ifdown(&nd_tbl, dev);
1da177e4
LT		 4677}
4678
95c96174 4679struct rt6_mtu_change_arg {
1da177e4 4680 struct net_device *dev;
95c96174 4681 unsigned int mtu;
c0b220cf 4682 struct fib6_info *f6i;
1da177e4
LT		 4683};
4684
cc5c073a 4685static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
c0b220cf
DA		 4686{
4687 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
cc5c073a 4688 struct fib6_info *f6i = arg->f6i;
c0b220cf
DA		 4689
4690 /* For administrative MTU increase, there is no way to discover
4691 * IPv6 PMTU increase, so PMTU increase should be updated here.
4692 * Since RFC 1981 doesn't include administrative MTU increase
4693 * update PMTU increase is a MUST. (i.e. jumbo frame)
4694 */
4695 if (nh->fib_nh_dev == arg->dev) {
4696 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4697 u32 mtu = f6i->fib6_pmtu;
4698
4699 if (mtu >= arg->mtu ||
4700 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4701 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4702
4703 spin_lock_bh(&rt6_exception_lock);
cc5c073a 4704 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
c0b220cf
DA		 4705 spin_unlock_bh(&rt6_exception_lock);
4706 }
4707
4708 return 0;
4709}
4710
4711static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
1da177e4
LT		 4712{
4713 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4714 struct inet6_dev *idev;
4715
4716 /* In IPv6 pmtu discovery is not optional,
4717 so that RTAX_MTU lock cannot disable it.
4718 We still use this lock to block changes
4719 caused by addrconf/ndisc.
4720 */
4721
4722 idev = __in6_dev_get(arg->dev);
38308473 4723 if (!idev)
1da177e4
LT		 4724 return 0;
4725
c0b220cf
DA		 4726 if (fib6_metric_locked(f6i, RTAX_MTU))
4727 return 0;
d4ead6b3 4728
c0b220cf 4729 arg->f6i = f6i;
2d44234b
DA		 4730 if (f6i->nh) {
4731 /* fib6_nh_mtu_change only returns 0, so this is safe */
4732 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4733 arg);
4734 }
4735
1cf844c7 4736 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
1da177e4
LT		 4737}
4738
95c96174 4739void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
1da177e4 4740{
c71099ac
TG		 4741 struct rt6_mtu_change_arg arg = {
4742 .dev = dev,
4743 .mtu = mtu,
4744 };
1da177e4 4745
0c3584d5 4746 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
1da177e4
LT		 4747}
4748
ef7c79ed 4749static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
75425657 4750 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
5176f91e 4751 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
aa8f8778 4752 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
86872cb5 4753 [RTA_OIF] = { .type = NLA_U32 },
ab364a6f 4754 [RTA_IIF] = { .type = NLA_U32 },
86872cb5
TG		 4755 [RTA_PRIORITY] = { .type = NLA_U32 },
4756 [RTA_METRICS] = { .type = NLA_NESTED },
51ebd318 4757 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
c78ba6d6 4758 [RTA_PREF] = { .type = NLA_U8 },
19e42e45
RP		 4759 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4760 [RTA_ENCAP] = { .type = NLA_NESTED },
32bc201e 4761 [RTA_EXPIRES] = { .type = NLA_U32 },
622ec2c9 4762 [RTA_UID] = { .type = NLA_U32 },
3b45a410 4763 [RTA_MARK] = { .type = NLA_U32 },
aa8f8778 4764 [RTA_TABLE] = { .type = NLA_U32 },
eacb9384
RP		 4765 [RTA_IP_PROTO] = { .type = NLA_U8 },
4766 [RTA_SPORT] = { .type = NLA_U16 },
4767 [RTA_DPORT] = { .type = NLA_U16 },
5b98324e 4768 [RTA_NH_ID] = { .type = NLA_U32 },
86872cb5
TG		 4769};
4770
4771static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
333c4301
DA		 4772 struct fib6_config *cfg,
4773 struct netlink_ext_ack *extack)
1da177e4 4774{
86872cb5
TG		 4775 struct rtmsg *rtm;
4776 struct nlattr *tb[RTA_MAX+1];
c78ba6d6 4777 unsigned int pref;
86872cb5 4778 int err;
1da177e4 4779
8cb08174
JB		 4780 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
4781 rtm_ipv6_policy, extack);
86872cb5
TG		 4782 if (err < 0)
4783 goto errout;
1da177e4 4784
86872cb5
TG		 4785 err = -EINVAL;
4786 rtm = nlmsg_data(nlh);
86872cb5 4787
84db8407
MŻ		 4788 *cfg = (struct fib6_config){
4789 .fc_table = rtm->rtm_table,
4790 .fc_dst_len = rtm->rtm_dst_len,
4791 .fc_src_len = rtm->rtm_src_len,
4792 .fc_flags = RTF_UP,
4793 .fc_protocol = rtm->rtm_protocol,
4794 .fc_type = rtm->rtm_type,
4795
4796 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
4797 .fc_nlinfo.nlh = nlh,
4798 .fc_nlinfo.nl_net = sock_net(skb->sk),
4799 };
86872cb5 4800
ef2c7d7b
ND		 4801 if (rtm->rtm_type == RTN_UNREACHABLE ||
4802 rtm->rtm_type == RTN_BLACKHOLE ||
b4949ab2
ND		 4803 rtm->rtm_type == RTN_PROHIBIT ||
4804 rtm->rtm_type == RTN_THROW)
86872cb5
TG		 4805 cfg->fc_flags |= RTF_REJECT;
4806
ab79ad14
MŻ		 4807 if (rtm->rtm_type == RTN_LOCAL)
4808 cfg->fc_flags |= RTF_LOCAL;
4809
1f56a01f
MKL		 4810 if (rtm->rtm_flags & RTM_F_CLONED)
4811 cfg->fc_flags |= RTF_CACHE;
4812
fc1e64e1
DA		 4813 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4814
5b98324e
DA		 4815 if (tb[RTA_NH_ID]) {
4816 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
4817 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
4818 NL_SET_ERR_MSG(extack,
4819 "Nexthop specification and nexthop id are mutually exclusive");
4820 goto errout;
4821 }
4822 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
4823 }
4824
86872cb5 4825 if (tb[RTA_GATEWAY]) {
67b61f6c 4826 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
86872cb5 4827 cfg->fc_flags |= RTF_GATEWAY;
1da177e4 4828 }
e3818541
DA		 4829 if (tb[RTA_VIA]) {
4830 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
4831 goto errout;
4832 }
86872cb5
TG		 4833
4834 if (tb[RTA_DST]) {
4835 int plen = (rtm->rtm_dst_len + 7) >> 3;
4836
4837 if (nla_len(tb[RTA_DST]) < plen)
4838 goto errout;
4839
4840 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1da177e4 4841 }
86872cb5
TG		 4842
4843 if (tb[RTA_SRC]) {
4844 int plen = (rtm->rtm_src_len + 7) >> 3;
4845
4846 if (nla_len(tb[RTA_SRC]) < plen)
4847 goto errout;
4848
4849 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1da177e4 4850 }
86872cb5 4851
c3968a85 4852 if (tb[RTA_PREFSRC])
67b61f6c 4853 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
c3968a85 4854
86872cb5
TG		 4855 if (tb[RTA_OIF])
4856 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4857
4858 if (tb[RTA_PRIORITY])
4859 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4860
4861 if (tb[RTA_METRICS]) {
4862 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4863 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
1da177e4 4864 }
86872cb5
TG		 4865
4866 if (tb[RTA_TABLE])
4867 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4868
51ebd318
ND		 4869 if (tb[RTA_MULTIPATH]) {
4870 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4871 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
9ed59592
DA		 4872
4873 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
c255bd68 4874 cfg->fc_mp_len, extack);
9ed59592
DA		 4875 if (err < 0)
4876 goto errout;
51ebd318
ND		 4877 }
4878
c78ba6d6
LR		 4879 if (tb[RTA_PREF]) {
4880 pref = nla_get_u8(tb[RTA_PREF]);
4881 if (pref != ICMPV6_ROUTER_PREF_LOW &&
4882 pref != ICMPV6_ROUTER_PREF_HIGH)
4883 pref = ICMPV6_ROUTER_PREF_MEDIUM;
4884 cfg->fc_flags |= RTF_PREF(pref);
4885 }
4886
19e42e45
RP		 4887 if (tb[RTA_ENCAP])
4888 cfg->fc_encap = tb[RTA_ENCAP];
4889
9ed59592 4890 if (tb[RTA_ENCAP_TYPE]) {
19e42e45
RP		 4891 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
4892
c255bd68 4893 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
9ed59592
DA		 4894 if (err < 0)
4895 goto errout;
4896 }
4897
32bc201e
XL		 4898 if (tb[RTA_EXPIRES]) {
4899 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
4900
4901 if (addrconf_finite_timeout(timeout)) {
4902 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
4903 cfg->fc_flags |= RTF_EXPIRES;
4904 }
4905 }
4906
86872cb5
TG		 4907 err = 0;
4908errout:
4909 return err;
1da177e4
LT		 4910}
4911
6b9ea5a6 4912struct rt6_nh {
8d1c802b 4913 struct fib6_info *fib6_info;
6b9ea5a6 4914 struct fib6_config r_cfg;
6b9ea5a6
RP		 4915 struct list_head next;
4916};
4917
d4ead6b3
DA		 4918static int ip6_route_info_append(struct net *net,
4919 struct list_head *rt6_nh_list,
8d1c802b
DA		 4920 struct fib6_info *rt,
4921 struct fib6_config *r_cfg)
6b9ea5a6
RP		 4922{
4923 struct rt6_nh *nh;
6b9ea5a6
RP		 4924 int err = -EEXIST;
4925
4926 list_for_each_entry(nh, rt6_nh_list, next) {
8d1c802b
DA		 4927 /* check if fib6_info already exists */
4928 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
6b9ea5a6
RP		 4929 return err;
4930 }
4931
4932 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
4933 if (!nh)
4934 return -ENOMEM;
8d1c802b 4935 nh->fib6_info = rt;
6b9ea5a6
RP		 4936 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
4937 list_add_tail(&nh->next, rt6_nh_list);
4938
4939 return 0;
4940}
4941
8d1c802b
DA		 4942static void ip6_route_mpath_notify(struct fib6_info *rt,
4943 struct fib6_info *rt_last,
3b1137fe
DA		 4944 struct nl_info *info,
4945 __u16 nlflags)
4946{
4947 /* if this is an APPEND route, then rt points to the first route
4948 * inserted and rt_last points to last route inserted. Userspace
4949 * wants a consistent dump of the route which starts at the first
4950 * nexthop. Since sibling routes are always added at the end of
4951 * the list, find the first sibling of the last route appended
4952 */
93c2fb25
DA		 4953 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
4954 rt = list_first_entry(&rt_last->fib6_siblings,
8d1c802b 4955 struct fib6_info,
93c2fb25 4956 fib6_siblings);
3b1137fe
DA		 4957 }
4958
4959 if (rt)
4960 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
4961}
4962
333c4301
DA		 4963static int ip6_route_multipath_add(struct fib6_config *cfg,
4964 struct netlink_ext_ack *extack)
51ebd318 4965{
8d1c802b 4966 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
3b1137fe 4967 struct nl_info *info = &cfg->fc_nlinfo;
ebee3cad 4968 enum fib_event_type event_type;
51ebd318
ND		 4969 struct fib6_config r_cfg;
4970 struct rtnexthop *rtnh;
8d1c802b 4971 struct fib6_info *rt;
6b9ea5a6
RP		 4972 struct rt6_nh *err_nh;
4973 struct rt6_nh *nh, *nh_safe;
3b1137fe 4974 __u16 nlflags;
51ebd318
ND		 4975 int remaining;
4976 int attrlen;
6b9ea5a6
RP		 4977 int err = 1;
4978 int nhn = 0;
4979 int replace = (cfg->fc_nlinfo.nlh &&
4980 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
4981 LIST_HEAD(rt6_nh_list);
51ebd318 4982
3b1137fe
DA		 4983 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
4984 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
4985 nlflags |= NLM_F_APPEND;
4986
35f1b4e9 4987 remaining = cfg->fc_mp_len;
51ebd318 4988 rtnh = (struct rtnexthop *)cfg->fc_mp;
51ebd318 4989
6b9ea5a6 4990 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
8d1c802b 4991 * fib6_info structs per nexthop
6b9ea5a6 4992 */
51ebd318
ND		 4993 while (rtnh_ok(rtnh, remaining)) {
4994 memcpy(&r_cfg, cfg, sizeof(*cfg));
4995 if (rtnh->rtnh_ifindex)
4996 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
4997
4998 attrlen = rtnh_attrlen(rtnh);
4999 if (attrlen > 0) {
5000 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5001
5002 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5003 if (nla) {
67b61f6c 5004 r_cfg.fc_gateway = nla_get_in6_addr(nla);
51ebd318
ND		 5005 r_cfg.fc_flags |= RTF_GATEWAY;
5006 }
19e42e45
RP		 5007 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5008 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5009 if (nla)
5010 r_cfg.fc_encap_type = nla_get_u16(nla);
51ebd318 5011 }
6b9ea5a6 5012
68e2ffde 5013 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
acb54e3c 5014 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
8c5b83f0
RP		 5015 if (IS_ERR(rt)) {
5016 err = PTR_ERR(rt);
5017 rt = NULL;
6b9ea5a6 5018 goto cleanup;
8c5b83f0 5019 }
b5d2d75e
DA		 5020 if (!rt6_qualify_for_ecmp(rt)) {
5021 err = -EINVAL;
5022 NL_SET_ERR_MSG(extack,
5023 "Device only routes can not be added for IPv6 using the multipath API.");
5024 fib6_info_release(rt);
5025 goto cleanup;
5026 }
6b9ea5a6 5027
1cf844c7 5028 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
398958ae 5029
d4ead6b3
DA		 5030 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5031 rt, &r_cfg);
51ebd318 5032 if (err) {
93531c67 5033 fib6_info_release(rt);
6b9ea5a6
RP		 5034 goto cleanup;
5035 }
5036
5037 rtnh = rtnh_next(rtnh, &remaining);
5038 }
5039
3b1137fe
DA		 5040 /* for add and replace send one notification with all nexthops.
5041 * Skip the notification in fib6_add_rt2node and send one with
5042 * the full route when done
5043 */
5044 info->skip_notify = 1;
5045
ebee3cad
IS		 5046 /* For add and replace, send one notification with all nexthops. For
5047 * append, send one notification with all appended nexthops.
5048 */
5049 info->skip_notify_kernel = 1;
5050
6b9ea5a6
RP		 5051 err_nh = NULL;
5052 list_for_each_entry(nh, &rt6_nh_list, next) {
8d1c802b
DA		 5053 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5054 fib6_info_release(nh->fib6_info);
93531c67 5055
f7225172
DA		 5056 if (!err) {
5057 /* save reference to last route successfully inserted */
5058 rt_last = nh->fib6_info;
5059
5060 /* save reference to first route for notification */
5061 if (!rt_notif)
5062 rt_notif = nh->fib6_info;
5063 }
3b1137fe 5064
8d1c802b
DA		 5065 /* nh->fib6_info is used or freed at this point, reset to NULL*/
5066 nh->fib6_info = NULL;
6b9ea5a6
RP		 5067 if (err) {
5068 if (replace && nhn)
a5a82d84
JK		 5069 NL_SET_ERR_MSG_MOD(extack,
5070 "multipath route replace failed (check consistency of installed routes)");
6b9ea5a6
RP		 5071 err_nh = nh;
5072 goto add_errout;
51ebd318 5073 }
6b9ea5a6 5074
1a72418b 5075 /* Because each route is added like a single route we remove
27596472
MK		 5076 * these flags after the first nexthop: if there is a collision,
5077 * we have already failed to add the first nexthop:
5078 * fib6_add_rt2node() has rejected it; when replacing, old
5079 * nexthops have been replaced by first new, the rest should
5080 * be added to it.
1a72418b 5081 */
27596472
MK		 5082 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5083 NLM_F_REPLACE);
6b9ea5a6
RP		 5084 nhn++;
5085 }
5086
ebee3cad
IS		 5087 event_type = replace ? FIB_EVENT_ENTRY_REPLACE : FIB_EVENT_ENTRY_ADD;
5088 err = call_fib6_multipath_entry_notifiers(info->nl_net, event_type,
5089 rt_notif, nhn - 1, extack);
5090 if (err) {
5091 /* Delete all the siblings that were just added */
5092 err_nh = NULL;
5093 goto add_errout;
5094 }
5095
3b1137fe
DA		 5096 /* success ... tell user about new route */
5097 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
6b9ea5a6
RP		 5098 goto cleanup;
5099
5100add_errout:
3b1137fe
DA		 5101 /* send notification for routes that were added so that
5102 * the delete notifications sent by ip6_route_del are
5103 * coherent
5104 */
5105 if (rt_notif)
5106 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5107
6b9ea5a6
RP		 5108 /* Delete routes that were already added */
5109 list_for_each_entry(nh, &rt6_nh_list, next) {
5110 if (err_nh == nh)
5111 break;
333c4301 5112 ip6_route_del(&nh->r_cfg, extack);
6b9ea5a6
RP		 5113 }
5114
5115cleanup:
5116 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
8d1c802b
DA		 5117 if (nh->fib6_info)
5118 fib6_info_release(nh->fib6_info);
6b9ea5a6
RP		 5119 list_del(&nh->next);
5120 kfree(nh);
5121 }
5122
5123 return err;
5124}
5125
333c4301
DA		 5126static int ip6_route_multipath_del(struct fib6_config *cfg,
5127 struct netlink_ext_ack *extack)
6b9ea5a6
RP		 5128{
5129 struct fib6_config r_cfg;
5130 struct rtnexthop *rtnh;
5131 int remaining;
5132 int attrlen;
5133 int err = 1, last_err = 0;
5134
5135 remaining = cfg->fc_mp_len;
5136 rtnh = (struct rtnexthop *)cfg->fc_mp;
5137
5138 /* Parse a Multipath Entry */
5139 while (rtnh_ok(rtnh, remaining)) {
5140 memcpy(&r_cfg, cfg, sizeof(*cfg));
5141 if (rtnh->rtnh_ifindex)
5142 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5143
5144 attrlen = rtnh_attrlen(rtnh);
5145 if (attrlen > 0) {
5146 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5147
5148 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5149 if (nla) {
5150 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
5151 r_cfg.fc_flags |= RTF_GATEWAY;
5152 }
5153 }
333c4301 5154 err = ip6_route_del(&r_cfg, extack);
6b9ea5a6
RP		 5155 if (err)
5156 last_err = err;
5157
51ebd318
ND		 5158 rtnh = rtnh_next(rtnh, &remaining);
5159 }
5160
5161 return last_err;
5162}
5163
c21ef3e3
DA		 5164static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5165 struct netlink_ext_ack *extack)
1da177e4 5166{
86872cb5
TG		 5167 struct fib6_config cfg;
5168 int err;
1da177e4 5169
333c4301 5170 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 5171 if (err < 0)
5172 return err;
5173
5b98324e
DA		 5174 if (cfg.fc_nh_id &&
5175 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5176 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5177 return -EINVAL;
5178 }
5179
51ebd318 5180 if (cfg.fc_mp)
333c4301 5181 return ip6_route_multipath_del(&cfg, extack);
0ae81335
DA		 5182 else {
5183 cfg.fc_delete_all_nh = 1;
333c4301 5184 return ip6_route_del(&cfg, extack);
0ae81335 5185 }
1da177e4
LT		 5186}
5187
c21ef3e3
DA		 5188static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5189 struct netlink_ext_ack *extack)
1da177e4 5190{
86872cb5
TG		 5191 struct fib6_config cfg;
5192 int err;
1da177e4 5193
333c4301 5194 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 5195 if (err < 0)
5196 return err;
5197
67f69513
DA		 5198 if (cfg.fc_metric == 0)
5199 cfg.fc_metric = IP6_RT_PRIO_USER;
5200
51ebd318 5201 if (cfg.fc_mp)
333c4301 5202 return ip6_route_multipath_add(&cfg, extack);
51ebd318 5203 else
acb54e3c 5204 return ip6_route_add(&cfg, GFP_KERNEL, extack);
1da177e4
LT		 5205}
5206
a1b7a1f0
DA		 5207/* add the overhead of this fib6_nh to nexthop_len */
5208static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
339bf98f 5209{
a1b7a1f0 5210 int *nexthop_len = arg;
beb1afac 5211
a1b7a1f0
DA		 5212 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5213 + NLA_ALIGN(sizeof(struct rtnexthop))
5214 + nla_total_size(16); /* RTA_GATEWAY */
5215
5216 if (nh->fib_nh_lws) {
5217 /* RTA_ENCAP_TYPE */
5218 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5219 /* RTA_ENCAP */
5220 *nexthop_len += nla_total_size(2);
5221 }
f88d8ea6 5222
a1b7a1f0
DA		 5223 return 0;
5224}
beb1afac 5225
a1b7a1f0
DA		 5226static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5227{
5228 int nexthop_len;
5229
5230 if (f6i->nh) {
5231 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5232 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5233 &nexthop_len);
5234 } else {
5235 struct fib6_nh *nh = f6i->fib6_nh;
5236
5237 nexthop_len = 0;
5238 if (f6i->fib6_nsiblings) {
5239 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
5240 + NLA_ALIGN(sizeof(struct rtnexthop))
5241 + nla_total_size(16) /* RTA_GATEWAY */
5242 + lwtunnel_get_encap_size(nh->fib_nh_lws);
5243
5244 nexthop_len *= f6i->fib6_nsiblings;
5245 }
5246 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
beb1afac
DA		 5247 }
5248
339bf98f
TG		 5249 return NLMSG_ALIGN(sizeof(struct rtmsg))
5250 + nla_total_size(16) /* RTA_SRC */
5251 + nla_total_size(16) /* RTA_DST */
5252 + nla_total_size(16) /* RTA_GATEWAY */
5253 + nla_total_size(16) /* RTA_PREFSRC */
5254 + nla_total_size(4) /* RTA_TABLE */
5255 + nla_total_size(4) /* RTA_IIF */
5256 + nla_total_size(4) /* RTA_OIF */
5257 + nla_total_size(4) /* RTA_PRIORITY */
6a2b9ce0 5258 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
ea697639 5259 + nla_total_size(sizeof(struct rta_cacheinfo))
c78ba6d6 5260 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
19e42e45 5261 + nla_total_size(1) /* RTA_PREF */
beb1afac
DA		 5262 + nexthop_len;
5263}
5264
f88d8ea6
DA		 5265static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5266 unsigned char *flags)
5267{
5268 if (nexthop_is_multipath(nh)) {
5269 struct nlattr *mp;
5270
5271 mp = nla_nest_start(skb, RTA_MULTIPATH);
5272 if (!mp)
5273 goto nla_put_failure;
5274
5275 if (nexthop_mpath_fill_node(skb, nh))
5276 goto nla_put_failure;
5277
5278 nla_nest_end(skb, mp);
5279 } else {
5280 struct fib6_nh *fib6_nh;
5281
5282 fib6_nh = nexthop_fib6_nh(nh);
5283 if (fib_nexthop_info(skb, &fib6_nh->nh_common,
5284 flags, false) < 0)
5285 goto nla_put_failure;
5286 }
5287
5288 return 0;
5289
5290nla_put_failure:
5291 return -EMSGSIZE;
5292}
5293
d4ead6b3 5294static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 5295 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 5296 struct in6_addr *dest, struct in6_addr *src,
15e47304 5297 int iif, int type, u32 portid, u32 seq,
f8cfe2ce 5298 unsigned int flags)
1da177e4 5299{
22d0bd82
XL		 5300 struct rt6_info *rt6 = (struct rt6_info *)dst;
5301 struct rt6key *rt6_dst, *rt6_src;
5302 u32 *pmetrics, table, rt6_flags;
f88d8ea6 5303 unsigned char nh_flags = 0;
2d7202bf 5304 struct nlmsghdr *nlh;
22d0bd82 5305 struct rtmsg *rtm;
d4ead6b3 5306 long expires = 0;
1da177e4 5307
15e47304 5308 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
38308473 5309 if (!nlh)
26932566 5310 return -EMSGSIZE;
2d7202bf 5311
22d0bd82
XL		 5312 if (rt6) {
5313 rt6_dst = &rt6->rt6i_dst;
5314 rt6_src = &rt6->rt6i_src;
5315 rt6_flags = rt6->rt6i_flags;
5316 } else {
5317 rt6_dst = &rt->fib6_dst;
5318 rt6_src = &rt->fib6_src;
5319 rt6_flags = rt->fib6_flags;
5320 }
5321
2d7202bf 5322 rtm = nlmsg_data(nlh);
1da177e4 5323 rtm->rtm_family = AF_INET6;
22d0bd82
XL		 5324 rtm->rtm_dst_len = rt6_dst->plen;
5325 rtm->rtm_src_len = rt6_src->plen;
1da177e4 5326 rtm->rtm_tos = 0;
93c2fb25
DA		 5327 if (rt->fib6_table)
5328 table = rt->fib6_table->tb6_id;
c71099ac 5329 else
9e762a4a 5330 table = RT6_TABLE_UNSPEC;
97f0082a 5331 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
c78679e8
DM		 5332 if (nla_put_u32(skb, RTA_TABLE, table))
5333 goto nla_put_failure;
e8478e80
DA		 5334
5335 rtm->rtm_type = rt->fib6_type;
1da177e4
LT		 5336 rtm->rtm_flags = 0;
5337 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
93c2fb25 5338 rtm->rtm_protocol = rt->fib6_protocol;
1da177e4 5339
22d0bd82 5340 if (rt6_flags & RTF_CACHE)
1da177e4
LT		 5341 rtm->rtm_flags |= RTM_F_CLONED;
5342
d4ead6b3
DA		 5343 if (dest) {
5344 if (nla_put_in6_addr(skb, RTA_DST, dest))
c78679e8 5345 goto nla_put_failure;
1ab1457c 5346 rtm->rtm_dst_len = 128;
1da177e4 5347 } else if (rtm->rtm_dst_len)
22d0bd82 5348 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
c78679e8 5349 goto nla_put_failure;
1da177e4
LT		 5350#ifdef CONFIG_IPV6_SUBTREES
5351 if (src) {
930345ea 5352 if (nla_put_in6_addr(skb, RTA_SRC, src))
c78679e8 5353 goto nla_put_failure;
1ab1457c 5354 rtm->rtm_src_len = 128;
c78679e8 5355 } else if (rtm->rtm_src_len &&
22d0bd82 5356 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
c78679e8 5357 goto nla_put_failure;
1da177e4 5358#endif
7bc570c8
YH		 5359 if (iif) {
5360#ifdef CONFIG_IPV6_MROUTE
22d0bd82 5361 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
fd61c6ba
DA		 5362 int err = ip6mr_get_route(net, skb, rtm, portid);
5363
5364 if (err == 0)
5365 return 0;
5366 if (err < 0)
5367 goto nla_put_failure;
7bc570c8
YH		 5368 } else
5369#endif
c78679e8
DM		 5370 if (nla_put_u32(skb, RTA_IIF, iif))
5371 goto nla_put_failure;
d4ead6b3 5372 } else if (dest) {
1da177e4 5373 struct in6_addr saddr_buf;
d4ead6b3 5374 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
930345ea 5375 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 5376 goto nla_put_failure;
1da177e4 5377 }
2d7202bf 5378
93c2fb25 5379 if (rt->fib6_prefsrc.plen) {
c3968a85 5380 struct in6_addr saddr_buf;
93c2fb25 5381 saddr_buf = rt->fib6_prefsrc.addr;
930345ea 5382 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 5383 goto nla_put_failure;
c3968a85
DW		 5384 }
5385
d4ead6b3
DA		 5386 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5387 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
2d7202bf
TG		 5388 goto nla_put_failure;
5389
93c2fb25 5390 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
c78679e8 5391 goto nla_put_failure;
8253947e 5392
beb1afac
DA		 5393 /* For multipath routes, walk the siblings list and add
5394 * each as a nexthop within RTA_MULTIPATH.
5395 */
22d0bd82
XL		 5396 if (rt6) {
5397 if (rt6_flags & RTF_GATEWAY &&
5398 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5399 goto nla_put_failure;
5400
5401 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5402 goto nla_put_failure;
5403 } else if (rt->fib6_nsiblings) {
8d1c802b 5404 struct fib6_info *sibling, *next_sibling;
beb1afac
DA		 5405 struct nlattr *mp;
5406
ae0be8de 5407 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
beb1afac
DA		 5408 if (!mp)
5409 goto nla_put_failure;
5410
1cf844c7
DA		 5411 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5412 rt->fib6_nh->fib_nh_weight) < 0)
beb1afac
DA		 5413 goto nla_put_failure;
5414
5415 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25 5416 &rt->fib6_siblings, fib6_siblings) {
1cf844c7
DA		 5417 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5418 sibling->fib6_nh->fib_nh_weight) < 0)
beb1afac
DA		 5419 goto nla_put_failure;
5420 }
5421
5422 nla_nest_end(skb, mp);
f88d8ea6
DA		 5423 } else if (rt->nh) {
5424 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5425 goto nla_put_failure;
ecc5663c 5426
f88d8ea6
DA		 5427 if (nexthop_is_blackhole(rt->nh))
5428 rtm->rtm_type = RTN_BLACKHOLE;
5429
5430 if (rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5431 goto nla_put_failure;
5432
5433 rtm->rtm_flags |= nh_flags;
5434 } else {
1cf844c7 5435 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common,
ecc5663c 5436 &nh_flags, false) < 0)
beb1afac 5437 goto nla_put_failure;
ecc5663c
DA		 5438
5439 rtm->rtm_flags |= nh_flags;
beb1afac
DA		 5440 }
5441
22d0bd82 5442 if (rt6_flags & RTF_EXPIRES) {
14895687
DA		 5443 expires = dst ? dst->expires : rt->expires;
5444 expires -= jiffies;
5445 }
69cdf8f9 5446
d4ead6b3 5447 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
e3703b3d 5448 goto nla_put_failure;
2d7202bf 5449
22d0bd82 5450 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
c78ba6d6
LR		 5451 goto nla_put_failure;
5452
19e42e45 5453
053c095a
JB		 5454 nlmsg_end(skb, nlh);
5455 return 0;
2d7202bf
TG		 5456
5457nla_put_failure:
26932566
PM		 5458 nlmsg_cancel(skb, nlh);
5459 return -EMSGSIZE;
1da177e4
LT		 5460}
5461
2c170e07
DA		 5462static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5463{
5464 const struct net_device *dev = arg;
5465
5466 if (nh->fib_nh_dev == dev)
5467 return 1;
5468
5469 return 0;
5470}
5471
13e38901
DA		 5472static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5473 const struct net_device *dev)
5474{
2c170e07
DA		 5475 if (f6i->nh) {
5476 struct net_device *_dev = (struct net_device *)dev;
5477
5478 return !!nexthop_for_each_fib6_nh(f6i->nh,
5479 fib6_info_nh_uses_dev,
5480 _dev);
5481 }
5482
1cf844c7 5483 if (f6i->fib6_nh->fib_nh_dev == dev)
13e38901
DA		 5484 return true;
5485
5486 if (f6i->fib6_nsiblings) {
5487 struct fib6_info *sibling, *next_sibling;
5488
5489 list_for_each_entry_safe(sibling, next_sibling,
5490 &f6i->fib6_siblings, fib6_siblings) {
1cf844c7 5491 if (sibling->fib6_nh->fib_nh_dev == dev)
13e38901
DA		 5492 return true;
5493 }
5494 }
5495
5496 return false;
5497}
5498
8d1c802b 5499int rt6_dump_route(struct fib6_info *rt, void *p_arg)
1da177e4
LT		 5500{
5501 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
13e38901
DA		 5502 struct fib_dump_filter *filter = &arg->filter;
5503 unsigned int flags = NLM_F_MULTI;
1f17e2f2
DA		 5504 struct net *net = arg->net;
5505
421842ed 5506 if (rt == net->ipv6.fib6_null_entry)
1f17e2f2 5507 return 0;
1da177e4 5508
13e38901
DA		 5509 if ((filter->flags & RTM_F_PREFIX) &&
5510 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5511 /* success since this is not a prefix route */
5512 return 1;
5513 }
5514 if (filter->filter_set) {
5515 if ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5516 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5517 (filter->protocol && rt->fib6_protocol != filter->protocol)) {
f8cfe2ce
DA		 5518 return 1;
5519 }
13e38901 5520 flags |= NLM_F_DUMP_FILTERED;
f8cfe2ce 5521 }
1da177e4 5522
d4ead6b3
DA		 5523 return rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL, 0,
5524 RTM_NEWROUTE, NETLINK_CB(arg->cb->skb).portid,
13e38901 5525 arg->cb->nlh->nlmsg_seq, flags);
1da177e4
LT		 5526}
5527
0eff0a27
JK		 5528static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5529 const struct nlmsghdr *nlh,
5530 struct nlattr **tb,
5531 struct netlink_ext_ack *extack)
5532{
5533 struct rtmsg *rtm;
5534 int i, err;
5535
5536 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5537 NL_SET_ERR_MSG_MOD(extack,
5538 "Invalid header for get route request");
5539 return -EINVAL;
5540 }
5541
5542 if (!netlink_strict_get_check(skb))
8cb08174
JB		 5543 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5544 rtm_ipv6_policy, extack);
0eff0a27
JK		 5545
5546 rtm = nlmsg_data(nlh);
5547 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5548 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5549 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5550 rtm->rtm_type) {
5551 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5552 return -EINVAL;
5553 }
5554 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5555 NL_SET_ERR_MSG_MOD(extack,
5556 "Invalid flags for get route request");
5557 return -EINVAL;
5558 }
5559
8cb08174
JB		 5560 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5561 rtm_ipv6_policy, extack);
0eff0a27
JK		 5562 if (err)
5563 return err;
5564
5565 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5566 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5567 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5568 return -EINVAL;
5569 }
5570
5571 for (i = 0; i <= RTA_MAX; i++) {
5572 if (!tb[i])
5573 continue;
5574
5575 switch (i) {
5576 case RTA_SRC:
5577 case RTA_DST:
5578 case RTA_IIF:
5579 case RTA_OIF:
5580 case RTA_MARK:
5581 case RTA_UID:
5582 case RTA_SPORT:
5583 case RTA_DPORT:
5584 case RTA_IP_PROTO:
5585 break;
5586 default:
5587 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
5588 return -EINVAL;
5589 }
5590 }
5591
5592 return 0;
5593}
5594
c21ef3e3
DA		 5595static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5596 struct netlink_ext_ack *extack)
1da177e4 5597{
3b1e0a65 5598 struct net *net = sock_net(in_skb->sk);
ab364a6f 5599 struct nlattr *tb[RTA_MAX+1];
18c3a61c 5600 int err, iif = 0, oif = 0;
a68886a6 5601 struct fib6_info *from;
18c3a61c 5602 struct dst_entry *dst;
ab364a6f 5603 struct rt6_info *rt;
1da177e4 5604 struct sk_buff *skb;
ab364a6f 5605 struct rtmsg *rtm;
744486d4 5606 struct flowi6 fl6 = {};
18c3a61c 5607 bool fibmatch;
1da177e4 5608
0eff0a27 5609 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
ab364a6f
TG		 5610 if (err < 0)
5611 goto errout;
1da177e4 5612
ab364a6f 5613 err = -EINVAL;
38b7097b
HFS		 5614 rtm = nlmsg_data(nlh);
5615 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
18c3a61c 5616 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
1da177e4 5617
ab364a6f
TG		 5618 if (tb[RTA_SRC]) {
5619 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
5620 goto errout;
5621
4e3fd7a0 5622 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
ab364a6f
TG		 5623 }
5624
5625 if (tb[RTA_DST]) {
5626 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
5627 goto errout;
5628
4e3fd7a0 5629 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
ab364a6f
TG		 5630 }
5631
5632 if (tb[RTA_IIF])
5633 iif = nla_get_u32(tb[RTA_IIF]);
5634
5635 if (tb[RTA_OIF])
72331bc0 5636 oif = nla_get_u32(tb[RTA_OIF]);
1da177e4 5637
2e47b291
LC		 5638 if (tb[RTA_MARK])
5639 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
5640
622ec2c9
LC		 5641 if (tb[RTA_UID])
5642 fl6.flowi6_uid = make_kuid(current_user_ns(),
5643 nla_get_u32(tb[RTA_UID]));
5644 else
5645 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
5646
eacb9384
RP		 5647 if (tb[RTA_SPORT])
5648 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
5649
5650 if (tb[RTA_DPORT])
5651 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
5652
5653 if (tb[RTA_IP_PROTO]) {
5654 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
5e1a99ea
HL		 5655 &fl6.flowi6_proto, AF_INET6,
5656 extack);
eacb9384
RP		 5657 if (err)
5658 goto errout;
5659 }
5660
1da177e4
LT		 5661 if (iif) {
5662 struct net_device *dev;
72331bc0
SL		 5663 int flags = 0;
5664
121622db
FW		 5665 rcu_read_lock();
5666
5667 dev = dev_get_by_index_rcu(net, iif);
1da177e4 5668 if (!dev) {
121622db 5669 rcu_read_unlock();
1da177e4 5670 err = -ENODEV;
ab364a6f 5671 goto errout;
1da177e4 5672 }
72331bc0
SL		 5673
5674 fl6.flowi6_iif = iif;
5675
5676 if (!ipv6_addr_any(&fl6.saddr))
5677 flags |= RT6_LOOKUP_F_HAS_SADDR;
5678
b75cc8f9 5679 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
121622db
FW		 5680
5681 rcu_read_unlock();
72331bc0
SL		 5682 } else {
5683 fl6.flowi6_oif = oif;
5684
58acfd71 5685 dst = ip6_route_output(net, NULL, &fl6);
18c3a61c
RP		 5686 }
5687
18c3a61c
RP		 5688
5689 rt = container_of(dst, struct rt6_info, dst);
5690 if (rt->dst.error) {
5691 err = rt->dst.error;
5692 ip6_rt_put(rt);
5693 goto errout;
1da177e4
LT		 5694 }
5695
9d6acb3b
WC		 5696 if (rt == net->ipv6.ip6_null_entry) {
5697 err = rt->dst.error;
5698 ip6_rt_put(rt);
5699 goto errout;
5700 }
5701
ab364a6f 5702 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
38308473 5703 if (!skb) {
94e187c0 5704 ip6_rt_put(rt);
ab364a6f
TG		 5705 err = -ENOBUFS;
5706 goto errout;
5707 }
1da177e4 5708
d8d1f30b 5709 skb_dst_set(skb, &rt->dst);
a68886a6
DA		 5710
5711 rcu_read_lock();
5712 from = rcu_dereference(rt->from);
886b7a50
MKL		 5713 if (from) {
5714 if (fibmatch)
5715 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
5716 iif, RTM_NEWROUTE,
5717 NETLINK_CB(in_skb).portid,
5718 nlh->nlmsg_seq, 0);
5719 else
5720 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
5721 &fl6.saddr, iif, RTM_NEWROUTE,
5722 NETLINK_CB(in_skb).portid,
5723 nlh->nlmsg_seq, 0);
5724 } else {
5725 err = -ENETUNREACH;
5726 }
a68886a6
DA		 5727 rcu_read_unlock();
5728
1da177e4 5729 if (err < 0) {
ab364a6f
TG		 5730 kfree_skb(skb);
5731 goto errout;
1da177e4
LT		 5732 }
5733
15e47304 5734 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
ab364a6f 5735errout:
1da177e4 5736 return err;
1da177e4
LT		 5737}
5738
8d1c802b 5739void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
37a1d361 5740 unsigned int nlm_flags)
1da177e4
LT		 5741{
5742 struct sk_buff *skb;
5578689a 5743 struct net *net = info->nl_net;
528c4ceb
DL		 5744 u32 seq;
5745 int err;
5746
5747 err = -ENOBUFS;
38308473 5748 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
86872cb5 5749
19e42e45 5750 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
38308473 5751 if (!skb)
21713ebc
TG		 5752 goto errout;
5753
d4ead6b3
DA		 5754 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
5755 event, info->portid, seq, nlm_flags);
26932566
PM		 5756 if (err < 0) {
5757 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
5758 WARN_ON(err == -EMSGSIZE);
5759 kfree_skb(skb);
5760 goto errout;
5761 }
15e47304 5762 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
1ce85fe4
PNA		 5763 info->nlh, gfp_any());
5764 return;
21713ebc
TG		 5765errout:
5766 if (err < 0)
5578689a 5767 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
1da177e4
LT		 5768}
5769
19a3b7ee
DA		 5770void fib6_rt_update(struct net *net, struct fib6_info *rt,
5771 struct nl_info *info)
5772{
5773 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
5774 struct sk_buff *skb;
5775 int err = -ENOBUFS;
5776
5777 /* call_fib6_entry_notifiers will be removed when in-kernel notifier
5778 * is implemented and supported for nexthop objects
5779 */
5780 call_fib6_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, rt, NULL);
5781
5782 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
5783 if (!skb)
5784 goto errout;
5785
5786 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
5787 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
5788 if (err < 0) {
5789 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
5790 WARN_ON(err == -EMSGSIZE);
5791 kfree_skb(skb);
5792 goto errout;
5793 }
5794 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
5795 info->nlh, gfp_any());
5796 return;
5797errout:
5798 if (err < 0)
5799 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
5800}
5801
8ed67789 5802static int ip6_route_dev_notify(struct notifier_block *this,
351638e7 5803 unsigned long event, void *ptr)
8ed67789 5804{
351638e7 5805 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
c346dca1 5806 struct net *net = dev_net(dev);
8ed67789 5807
242d3a49
WC		 5808 if (!(dev->flags & IFF_LOOPBACK))
5809 return NOTIFY_OK;
5810
5811 if (event == NETDEV_REGISTER) {
1cf844c7 5812 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
d8d1f30b 5813 net->ipv6.ip6_null_entry->dst.dev = dev;
8ed67789
DL		 5814 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
5815#ifdef CONFIG_IPV6_MULTIPLE_TABLES
d8d1f30b 5816 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
8ed67789 5817 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
d8d1f30b 5818 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
8ed67789 5819 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
242d3a49 5820#endif
76da0704
WC		 5821 } else if (event == NETDEV_UNREGISTER &&
5822 dev->reg_state != NETREG_UNREGISTERED) {
5823 /* NETDEV_UNREGISTER could be fired for multiple times by
5824 * netdev_wait_allrefs(). Make sure we only call this once.
5825 */
12d94a80 5826 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
242d3a49 5827#ifdef CONFIG_IPV6_MULTIPLE_TABLES
12d94a80
ED		 5828 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
5829 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
8ed67789
DL		 5830#endif
5831 }
5832
5833 return NOTIFY_OK;
5834}
5835
1da177e4
LT		 5836/*
5837 * /proc
5838 */
5839
5840#ifdef CONFIG_PROC_FS
1da177e4
LT		 5841static int rt6_stats_seq_show(struct seq_file *seq, void *v)
5842{
69ddb805 5843 struct net *net = (struct net *)seq->private;
1da177e4 5844 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
69ddb805
DL		 5845 net->ipv6.rt6_stats->fib_nodes,
5846 net->ipv6.rt6_stats->fib_route_nodes,
81eb8447 5847 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
69ddb805
DL		 5848 net->ipv6.rt6_stats->fib_rt_entries,
5849 net->ipv6.rt6_stats->fib_rt_cache,
fc66f95c 5850 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
69ddb805 5851 net->ipv6.rt6_stats->fib_discarded_routes);
1da177e4
LT		 5852
5853 return 0;
5854}
1da177e4
LT		 5855#endif /* CONFIG_PROC_FS */
5856
5857#ifdef CONFIG_SYSCTL
5858
1da177e4 5859static
fe2c6338 5860int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
1da177e4
LT		 5861 void __user *buffer, size_t *lenp, loff_t *ppos)
5862{
c486da34
LAG		 5863 struct net *net;
5864 int delay;
f0fb9b28 5865 int ret;
c486da34 5866 if (!write)
1da177e4 5867 return -EINVAL;
c486da34
LAG		 5868
5869 net = (struct net *)ctl->extra1;
5870 delay = net->ipv6.sysctl.flush_delay;
f0fb9b28
AP		 5871 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5872 if (ret)
5873 return ret;
5874
2ac3ac8f 5875 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
c486da34 5876 return 0;
1da177e4
LT		 5877}
5878
7c6bb7d2
DA		 5879static int zero;
5880static int one = 1;
5881
ed792e28 5882static struct ctl_table ipv6_route_table_template[] = {
1ab1457c 5883 {
1da177e4 5884 .procname = "flush",
4990509f 5885 .data = &init_net.ipv6.sysctl.flush_delay,
1da177e4 5886 .maxlen = sizeof(int),
89c8b3a1 5887 .mode = 0200,
6d9f239a 5888 .proc_handler = ipv6_sysctl_rtcache_flush
1da177e4
LT		 5889 },
5890 {
1da177e4 5891 .procname = "gc_thresh",
9a7ec3a9 5892 .data = &ip6_dst_ops_template.gc_thresh,
1da177e4
LT		 5893 .maxlen = sizeof(int),
5894 .mode = 0644,
6d9f239a 5895 .proc_handler = proc_dointvec,
1da177e4
LT		 5896 },
5897 {
1da177e4 5898 .procname = "max_size",
4990509f 5899 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
1da177e4
LT		 5900 .maxlen = sizeof(int),
5901 .mode = 0644,
6d9f239a 5902 .proc_handler = proc_dointvec,
1da177e4
LT		 5903 },
5904 {
1da177e4 5905 .procname = "gc_min_interval",
4990509f 5906 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 5907 .maxlen = sizeof(int),
5908 .mode = 0644,
6d9f239a 5909 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5910 },
5911 {
1da177e4 5912 .procname = "gc_timeout",
4990509f 5913 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
1da177e4
LT		 5914 .maxlen = sizeof(int),
5915 .mode = 0644,
6d9f239a 5916 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5917 },
5918 {
1da177e4 5919 .procname = "gc_interval",
4990509f 5920 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
1da177e4
LT		 5921 .maxlen = sizeof(int),
5922 .mode = 0644,
6d9f239a 5923 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5924 },
5925 {
1da177e4 5926 .procname = "gc_elasticity",
4990509f 5927 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
1da177e4
LT		 5928 .maxlen = sizeof(int),
5929 .mode = 0644,
f3d3f616 5930 .proc_handler = proc_dointvec,
1da177e4
LT		 5931 },
5932 {
1da177e4 5933 .procname = "mtu_expires",
4990509f 5934 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
1da177e4
LT		 5935 .maxlen = sizeof(int),
5936 .mode = 0644,
6d9f239a 5937 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5938 },
5939 {
1da177e4 5940 .procname = "min_adv_mss",
4990509f 5941 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
1da177e4
LT		 5942 .maxlen = sizeof(int),
5943 .mode = 0644,
f3d3f616 5944 .proc_handler = proc_dointvec,
1da177e4
LT		 5945 },
5946 {
1da177e4 5947 .procname = "gc_min_interval_ms",
4990509f 5948 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 5949 .maxlen = sizeof(int),
5950 .mode = 0644,
6d9f239a 5951 .proc_handler = proc_dointvec_ms_jiffies,
1da177e4 5952 },
7c6bb7d2
DA		 5953 {
5954 .procname = "skip_notify_on_dev_down",
5955 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
5956 .maxlen = sizeof(int),
5957 .mode = 0644,
5958 .proc_handler = proc_dointvec,
5959 .extra1 = &zero,
5960 .extra2 = &one,
5961 },
f8572d8f 5962 { }
1da177e4
LT		 5963};
5964
2c8c1e72 5965struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
760f2d01
DL		 5966{
5967 struct ctl_table *table;
5968
5969 table = kmemdup(ipv6_route_table_template,
5970 sizeof(ipv6_route_table_template),
5971 GFP_KERNEL);
5ee09105
YH		 5972
5973 if (table) {
5974 table[0].data = &net->ipv6.sysctl.flush_delay;
c486da34 5975 table[0].extra1 = net;
86393e52 5976 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
5ee09105
YH		 5977 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
5978 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
5979 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
5980 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
5981 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
5982 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
5983 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
9c69fabe 5984 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
7c6bb7d2 5985 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
464dc801
EB		 5986
5987 /* Don't export sysctls to unprivileged users */
5988 if (net->user_ns != &init_user_ns)
5989 table[0].procname = NULL;
5ee09105
YH		 5990 }
5991
760f2d01
DL		 5992 return table;
5993}
1da177e4
LT		 5994#endif
5995
2c8c1e72 5996static int __net_init ip6_route_net_init(struct net *net)
cdb18761 5997{
633d424b 5998 int ret = -ENOMEM;
8ed67789 5999
86393e52
AD		 6000 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6001 sizeof(net->ipv6.ip6_dst_ops));
f2fc6a54 6002
fc66f95c
ED		 6003 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6004 goto out_ip6_dst_ops;
6005
1cf844c7 6006 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
421842ed
DA		 6007 if (!net->ipv6.fib6_null_entry)
6008 goto out_ip6_dst_entries;
1cf844c7
DA		 6009 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6010 sizeof(*net->ipv6.fib6_null_entry));
421842ed 6011
8ed67789
DL		 6012 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6013 sizeof(*net->ipv6.ip6_null_entry),
6014 GFP_KERNEL);
6015 if (!net->ipv6.ip6_null_entry)
421842ed 6016 goto out_fib6_null_entry;
d8d1f30b 6017 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6018 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6019 ip6_template_metrics, true);
8ed67789
DL		 6020
6021#ifdef CONFIG_IPV6_MULTIPLE_TABLES
feca7d8c 6022 net->ipv6.fib6_has_custom_rules = false;
8ed67789
DL		 6023 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6024 sizeof(*net->ipv6.ip6_prohibit_entry),
6025 GFP_KERNEL);
68fffc67
PZ		 6026 if (!net->ipv6.ip6_prohibit_entry)
6027 goto out_ip6_null_entry;
d8d1f30b 6028 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6029 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6030 ip6_template_metrics, true);
8ed67789
DL		 6031
6032 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6033 sizeof(*net->ipv6.ip6_blk_hole_entry),
6034 GFP_KERNEL);
68fffc67
PZ		 6035 if (!net->ipv6.ip6_blk_hole_entry)
6036 goto out_ip6_prohibit_entry;
d8d1f30b 6037 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6038 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6039 ip6_template_metrics, true);
8ed67789
DL		 6040#endif
6041
b339a47c
PZ		 6042 net->ipv6.sysctl.flush_delay = 0;
6043 net->ipv6.sysctl.ip6_rt_max_size = 4096;
6044 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6045 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6046 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6047 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6048 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6049 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
7c6bb7d2 6050 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
b339a47c 6051
6891a346
BT		 6052 net->ipv6.ip6_rt_gc_expire = 30*HZ;
6053
8ed67789
DL		 6054 ret = 0;
6055out:
6056 return ret;
f2fc6a54 6057
68fffc67
PZ		 6058#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6059out_ip6_prohibit_entry:
6060 kfree(net->ipv6.ip6_prohibit_entry);
6061out_ip6_null_entry:
6062 kfree(net->ipv6.ip6_null_entry);
6063#endif
421842ed
DA		 6064out_fib6_null_entry:
6065 kfree(net->ipv6.fib6_null_entry);
fc66f95c
ED		 6066out_ip6_dst_entries:
6067 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
f2fc6a54 6068out_ip6_dst_ops:
f2fc6a54 6069 goto out;
cdb18761
DL		 6070}
6071
2c8c1e72 6072static void __net_exit ip6_route_net_exit(struct net *net)
cdb18761 6073{
421842ed 6074 kfree(net->ipv6.fib6_null_entry);
8ed67789
DL		 6075 kfree(net->ipv6.ip6_null_entry);
6076#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6077 kfree(net->ipv6.ip6_prohibit_entry);
6078 kfree(net->ipv6.ip6_blk_hole_entry);
6079#endif
41bb78b4 6080 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
cdb18761
DL		 6081}
6082
d189634e
TG		 6083static int __net_init ip6_route_net_init_late(struct net *net)
6084{
6085#ifdef CONFIG_PROC_FS
c3506372
CH		 6086 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6087 sizeof(struct ipv6_route_iter));
3617d949
CH		 6088 proc_create_net_single("rt6_stats", 0444, net->proc_net,
6089 rt6_stats_seq_show, NULL);
d189634e
TG		 6090#endif
6091 return 0;
6092}
6093
6094static void __net_exit ip6_route_net_exit_late(struct net *net)
6095{
6096#ifdef CONFIG_PROC_FS
ece31ffd
G		 6097 remove_proc_entry("ipv6_route", net->proc_net);
6098 remove_proc_entry("rt6_stats", net->proc_net);
d189634e
TG		 6099#endif
6100}
6101
cdb18761
DL		 6102static struct pernet_operations ip6_route_net_ops = {
6103 .init = ip6_route_net_init,
6104 .exit = ip6_route_net_exit,
6105};
6106
c3426b47
DM		 6107static int __net_init ipv6_inetpeer_init(struct net *net)
6108{
6109 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6110
6111 if (!bp)
6112 return -ENOMEM;
6113 inet_peer_base_init(bp);
6114 net->ipv6.peers = bp;
6115 return 0;
6116}
6117
6118static void __net_exit ipv6_inetpeer_exit(struct net *net)
6119{
6120 struct inet_peer_base *bp = net->ipv6.peers;
6121
6122 net->ipv6.peers = NULL;
56a6b248 6123 inetpeer_invalidate_tree(bp);
c3426b47
DM		 6124 kfree(bp);
6125}
6126
2b823f72 6127static struct pernet_operations ipv6_inetpeer_ops = {
c3426b47
DM		 6128 .init = ipv6_inetpeer_init,
6129 .exit = ipv6_inetpeer_exit,
6130};
6131
d189634e
TG		 6132static struct pernet_operations ip6_route_net_late_ops = {
6133 .init = ip6_route_net_init_late,
6134 .exit = ip6_route_net_exit_late,
6135};
6136
8ed67789
DL		 6137static struct notifier_block ip6_route_dev_notifier = {
6138 .notifier_call = ip6_route_dev_notify,
242d3a49 6139 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
8ed67789
DL		 6140};
6141
2f460933
WC		 6142void __init ip6_route_init_special_entries(void)
6143{
6144 /* Registering of the loopback is done before this portion of code,
6145 * the loopback reference in rt6_info will not be taken, do it
6146 * manually for init_net */
1cf844c7 6147 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
2f460933
WC		 6148 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6149 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6150 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6151 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6152 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6153 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6154 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6155 #endif
6156}
6157
433d49c3 6158int __init ip6_route_init(void)
1da177e4 6159{
433d49c3 6160 int ret;
8d0b94af 6161 int cpu;
433d49c3 6162
9a7ec3a9
DL		 6163 ret = -ENOMEM;
6164 ip6_dst_ops_template.kmem_cachep =
e5d679f3 6165 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
f845ab6b 6166 SLAB_HWCACHE_ALIGN, NULL);
9a7ec3a9 6167 if (!ip6_dst_ops_template.kmem_cachep)
c19a28e1 6168 goto out;
14e50e57 6169
fc66f95c 6170 ret = dst_entries_init(&ip6_dst_blackhole_ops);
8ed67789 6171 if (ret)
bdb3289f 6172 goto out_kmem_cache;
bdb3289f 6173
c3426b47
DM		 6174 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6175 if (ret)
e8803b6c 6176 goto out_dst_entries;
2a0c451a 6177
7e52b33b
DM		 6178 ret = register_pernet_subsys(&ip6_route_net_ops);
6179 if (ret)
6180 goto out_register_inetpeer;
c3426b47 6181
5dc121e9
AE		 6182 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6183
e8803b6c 6184 ret = fib6_init();
433d49c3 6185 if (ret)
8ed67789 6186 goto out_register_subsys;
433d49c3 6187
433d49c3
DL		 6188 ret = xfrm6_init();
6189 if (ret)
e8803b6c 6190 goto out_fib6_init;
c35b7e72 6191
433d49c3
DL		 6192 ret = fib6_rules_init();
6193 if (ret)
6194 goto xfrm6_init;
7e5449c2 6195
d189634e
TG		 6196 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6197 if (ret)
6198 goto fib6_rules_init;
6199
16feebcf
FW		 6200 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6201 inet6_rtm_newroute, NULL, 0);
6202 if (ret < 0)
6203 goto out_register_late_subsys;
6204
6205 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6206 inet6_rtm_delroute, NULL, 0);
6207 if (ret < 0)
6208 goto out_register_late_subsys;
6209
6210 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6211 inet6_rtm_getroute, NULL,
6212 RTNL_FLAG_DOIT_UNLOCKED);
6213 if (ret < 0)
d189634e 6214 goto out_register_late_subsys;
c127ea2c 6215
8ed67789 6216 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
cdb18761 6217 if (ret)
d189634e 6218 goto out_register_late_subsys;
8ed67789 6219
8d0b94af
MKL		 6220 for_each_possible_cpu(cpu) {
6221 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6222
6223 INIT_LIST_HEAD(&ul->head);
6224 spin_lock_init(&ul->lock);
6225 }
6226
433d49c3
DL		 6227out:
6228 return ret;
6229
d189634e 6230out_register_late_subsys:
16feebcf 6231 rtnl_unregister_all(PF_INET6);
d189634e 6232 unregister_pernet_subsys(&ip6_route_net_late_ops);
433d49c3 6233fib6_rules_init:
433d49c3
DL		 6234 fib6_rules_cleanup();
6235xfrm6_init:
433d49c3 6236 xfrm6_fini();
2a0c451a
TG		 6237out_fib6_init:
6238 fib6_gc_cleanup();
8ed67789
DL		 6239out_register_subsys:
6240 unregister_pernet_subsys(&ip6_route_net_ops);
7e52b33b
DM		 6241out_register_inetpeer:
6242 unregister_pernet_subsys(&ipv6_inetpeer_ops);
fc66f95c
ED		 6243out_dst_entries:
6244 dst_entries_destroy(&ip6_dst_blackhole_ops);
433d49c3 6245out_kmem_cache:
f2fc6a54 6246 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
433d49c3 6247 goto out;
1da177e4
LT		 6248}
6249
6250void ip6_route_cleanup(void)
6251{
8ed67789 6252 unregister_netdevice_notifier(&ip6_route_dev_notifier);
d189634e 6253 unregister_pernet_subsys(&ip6_route_net_late_ops);
101367c2 6254 fib6_rules_cleanup();
1da177e4 6255 xfrm6_fini();
1da177e4 6256 fib6_gc_cleanup();
c3426b47 6257 unregister_pernet_subsys(&ipv6_inetpeer_ops);
8ed67789 6258 unregister_pernet_subsys(&ip6_route_net_ops);
41bb78b4 6259 dst_entries_destroy(&ip6_dst_blackhole_ops);
f2fc6a54 6260 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
1da177e4 6261}
Linux 6.x block layer and io_uring tree(s)