projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ipv6: convert rx data path to not take refcnt on dst
[linux-block.git] / net / ipv6 / route.c
CommitLineData
2874c5fd 1// SPDX-License-Identifier: GPL-2.0-or-later
1da177e4
LT		 2/*
3 * Linux INET6 implementation
4 * FIB front-end.
5 *
6 * Authors:
1ab1457c 7 * Pedro Roque <roque@di.fc.ul.pt>
1da177e4
LT		 8 */
9
10/* Changes:
11 *
12 * YOSHIFUJI Hideaki @USAGI
13 * reworked default router selection.
14 * - respect outgoing interface
15 * - select from (probably) reachable routers (i.e.
16 * routers in REACHABLE, STALE, DELAY or PROBE states).
17 * - always select the same router if it is (probably)
18 * reachable. otherwise, round-robin the list.
c0bece9f
YH		 19 * Ville Nuorvala
20 * Fixed routing subtrees.
1da177e4
LT		 21 */
22
f3213831
JP		 23#define pr_fmt(fmt) "IPv6: " fmt
24
4fc268d2 25#include <linux/capability.h>
1da177e4 26#include <linux/errno.h>
bc3b2d7f 27#include <linux/export.h>
1da177e4
LT		 28#include <linux/types.h>
29#include <linux/times.h>
30#include <linux/socket.h>
31#include <linux/sockios.h>
32#include <linux/net.h>
33#include <linux/route.h>
34#include <linux/netdevice.h>
35#include <linux/in6.h>
7bc570c8 36#include <linux/mroute6.h>
1da177e4 37#include <linux/init.h>
1da177e4 38#include <linux/if_arp.h>
1da177e4
LT		 39#include <linux/proc_fs.h>
40#include <linux/seq_file.h>
5b7c931d 41#include <linux/nsproxy.h>
5a0e3ad6 42#include <linux/slab.h>
35732d01 43#include <linux/jhash.h>
457c4cbc 44#include <net/net_namespace.h>
1da177e4
LT		 45#include <net/snmp.h>
46#include <net/ipv6.h>
47#include <net/ip6_fib.h>
48#include <net/ip6_route.h>
49#include <net/ndisc.h>
50#include <net/addrconf.h>
51#include <net/tcp.h>
52#include <linux/rtnetlink.h>
53#include <net/dst.h>
904af04d 54#include <net/dst_metadata.h>
1da177e4 55#include <net/xfrm.h>
8d71740c 56#include <net/netevent.h>
21713ebc 57#include <net/netlink.h>
3c618c1d 58#include <net/rtnh.h>
19e42e45 59#include <net/lwtunnel.h>
904af04d 60#include <net/ip_tunnels.h>
ca254490 61#include <net/l3mdev.h>
eacb9384 62#include <net/ip.h>
7c0f6ba6 63#include <linux/uaccess.h>
1da177e4
LT		 64
65#ifdef CONFIG_SYSCTL
66#include <linux/sysctl.h>
67#endif
68
30d444d3
DA		 69static int ip6_rt_type_to_error(u8 fib6_type);
70
71#define CREATE_TRACE_POINTS
72#include <trace/events/fib6.h>
73EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
74#undef CREATE_TRACE_POINTS
75
afc154e9 76enum rt6_nud_state {
7e980569
JB		 77 RT6_NUD_FAIL_HARD = -3,
78 RT6_NUD_FAIL_PROBE = -2,
79 RT6_NUD_FAIL_DO_RR = -1,
afc154e9
HFS		 80 RT6_NUD_SUCCEED = 1
81};
82
1da177e4 83static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
0dbaee3b 84static unsigned int ip6_default_advmss(const struct dst_entry *dst);
ebb762f2 85static unsigned int ip6_mtu(const struct dst_entry *dst);
1da177e4
LT		 86static struct dst_entry *ip6_negative_advice(struct dst_entry *);
87static void ip6_dst_destroy(struct dst_entry *);
88static void ip6_dst_ifdown(struct dst_entry *,
89 struct net_device *dev, int how);
569d3645 90static int ip6_dst_gc(struct dst_ops *ops);
1da177e4
LT		 91
92static int ip6_pkt_discard(struct sk_buff *skb);
ede2059d 93static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
7150aede 94static int ip6_pkt_prohibit(struct sk_buff *skb);
ede2059d 95static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1da177e4 96static void ip6_link_failure(struct sk_buff *skb);
6700c270
DM		 97static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
98 struct sk_buff *skb, u32 mtu);
99static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
100 struct sk_buff *skb);
702cea56
DA		 101static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
102 int strict);
a1b7a1f0 103static size_t rt6_nlmsg_size(struct fib6_info *f6i);
d4ead6b3 104static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 105 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 106 struct in6_addr *dest, struct in6_addr *src,
16a16cd3
DA		 107 int iif, int type, u32 portid, u32 seq,
108 unsigned int flags);
7e4b5128 109static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
510e2ced
WW		 110 const struct in6_addr *daddr,
111 const struct in6_addr *saddr);
1da177e4 112
70ceb4f5 113#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 114static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 115 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 116 const struct in6_addr *gwaddr,
117 struct net_device *dev,
95c96174 118 unsigned int pref);
8d1c802b 119static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 120 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 121 const struct in6_addr *gwaddr,
122 struct net_device *dev);
70ceb4f5
YH		 123#endif
124
8d0b94af
MKL		 125struct uncached_list {
126 spinlock_t lock;
127 struct list_head head;
128};
129
130static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
131
510c321b 132void rt6_uncached_list_add(struct rt6_info *rt)
8d0b94af
MKL		 133{
134 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
135
8d0b94af
MKL		 136 rt->rt6i_uncached_list = ul;
137
138 spin_lock_bh(&ul->lock);
139 list_add_tail(&rt->rt6i_uncached, &ul->head);
140 spin_unlock_bh(&ul->lock);
141}
142
510c321b 143void rt6_uncached_list_del(struct rt6_info *rt)
8d0b94af
MKL		 144{
145 if (!list_empty(&rt->rt6i_uncached)) {
146 struct uncached_list *ul = rt->rt6i_uncached_list;
81eb8447 147 struct net *net = dev_net(rt->dst.dev);
8d0b94af
MKL		 148
149 spin_lock_bh(&ul->lock);
150 list_del(&rt->rt6i_uncached);
81eb8447 151 atomic_dec(&net->ipv6.rt6_stats->fib_rt_uncache);
8d0b94af
MKL		 152 spin_unlock_bh(&ul->lock);
153 }
154}
155
156static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
157{
158 struct net_device *loopback_dev = net->loopback_dev;
159 int cpu;
160
e332bc67
EB		 161 if (dev == loopback_dev)
162 return;
163
8d0b94af
MKL		 164 for_each_possible_cpu(cpu) {
165 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
166 struct rt6_info *rt;
167
168 spin_lock_bh(&ul->lock);
169 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
170 struct inet6_dev *rt_idev = rt->rt6i_idev;
171 struct net_device *rt_dev = rt->dst.dev;
172
e332bc67 173 if (rt_idev->dev == dev) {
8d0b94af
MKL		 174 rt->rt6i_idev = in6_dev_get(loopback_dev);
175 in6_dev_put(rt_idev);
176 }
177
e332bc67 178 if (rt_dev == dev) {
8d0b94af
MKL		 179 rt->dst.dev = loopback_dev;
180 dev_hold(rt->dst.dev);
181 dev_put(rt_dev);
182 }
183 }
184 spin_unlock_bh(&ul->lock);
185 }
186}
187
f8a1b43b 188static inline const void *choose_neigh_daddr(const struct in6_addr *p,
f894cbf8
DM		 189 struct sk_buff *skb,
190 const void *daddr)
39232973 191{
a7563f34 192 if (!ipv6_addr_any(p))
39232973 193 return (const void *) p;
f894cbf8
DM		 194 else if (skb)
195 return &ipv6_hdr(skb)->daddr;
39232973
DM		 196 return daddr;
197}
198
f8a1b43b
DA		 199struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
200 struct net_device *dev,
201 struct sk_buff *skb,
202 const void *daddr)
d3aaeb38 203{
39232973
DM		 204 struct neighbour *n;
205
f8a1b43b
DA		 206 daddr = choose_neigh_daddr(gw, skb, daddr);
207 n = __ipv6_neigh_lookup(dev, daddr);
f83c7790
DM		 208 if (n)
209 return n;
7adf3246
SB		 210
211 n = neigh_create(&nd_tbl, daddr, dev);
212 return IS_ERR(n) ? NULL : n;
f8a1b43b
DA		 213}
214
215static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
216 struct sk_buff *skb,
217 const void *daddr)
218{
219 const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
220
221 return ip6_neigh_lookup(&rt->rt6i_gateway, dst->dev, skb, daddr);
f83c7790
DM		 222}
223
63fca65d
JA		 224static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
225{
226 struct net_device *dev = dst->dev;
227 struct rt6_info *rt = (struct rt6_info *)dst;
228
f8a1b43b 229 daddr = choose_neigh_daddr(&rt->rt6i_gateway, NULL, daddr);
63fca65d
JA		 230 if (!daddr)
231 return;
232 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
233 return;
234 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
235 return;
236 __ipv6_confirm_neigh(dev, daddr);
237}
238
9a7ec3a9 239static struct dst_ops ip6_dst_ops_template = {
1da177e4 240 .family = AF_INET6,
1da177e4
LT		 241 .gc = ip6_dst_gc,
242 .gc_thresh = 1024,
243 .check = ip6_dst_check,
0dbaee3b 244 .default_advmss = ip6_default_advmss,
ebb762f2 245 .mtu = ip6_mtu,
d4ead6b3 246 .cow_metrics = dst_cow_metrics_generic,
1da177e4
LT		 247 .destroy = ip6_dst_destroy,
248 .ifdown = ip6_dst_ifdown,
249 .negative_advice = ip6_negative_advice,
250 .link_failure = ip6_link_failure,
251 .update_pmtu = ip6_rt_update_pmtu,
6e157b6a 252 .redirect = rt6_do_redirect,
9f8955cc 253 .local_out = __ip6_local_out,
f8a1b43b 254 .neigh_lookup = ip6_dst_neigh_lookup,
63fca65d 255 .confirm_neigh = ip6_confirm_neigh,
1da177e4
LT		 256};
257
ebb762f2 258static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
ec831ea7 259{
618f9bc7
SK		 260 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
261
262 return mtu ? : dst->dev->mtu;
ec831ea7
RD		 263}
264
6700c270
DM		 265static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
266 struct sk_buff *skb, u32 mtu)
14e50e57
DM		 267{
268}
269
6700c270
DM		 270static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
271 struct sk_buff *skb)
b587ee3b
DM		 272{
273}
274
14e50e57
DM		 275static struct dst_ops ip6_dst_blackhole_ops = {
276 .family = AF_INET6,
14e50e57
DM		 277 .destroy = ip6_dst_destroy,
278 .check = ip6_dst_check,
ebb762f2 279 .mtu = ip6_blackhole_mtu,
214f45c9 280 .default_advmss = ip6_default_advmss,
14e50e57 281 .update_pmtu = ip6_rt_blackhole_update_pmtu,
b587ee3b 282 .redirect = ip6_rt_blackhole_redirect,
0a1f5962 283 .cow_metrics = dst_cow_metrics_generic,
f8a1b43b 284 .neigh_lookup = ip6_dst_neigh_lookup,
14e50e57
DM		 285};
286
62fa8a84 287static const u32 ip6_template_metrics[RTAX_MAX] = {
14edd87d 288 [RTAX_HOPLIMIT - 1] = 0,
62fa8a84
DM		 289};
290
8d1c802b 291static const struct fib6_info fib6_null_entry_template = {
93c2fb25
DA		 292 .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
293 .fib6_protocol = RTPROT_KERNEL,
294 .fib6_metric = ~(u32)0,
f05713e0 295 .fib6_ref = REFCOUNT_INIT(1),
421842ed
DA		 296 .fib6_type = RTN_UNREACHABLE,
297 .fib6_metrics = (struct dst_metrics *)&dst_default_metrics,
298};
299
fb0af4c7 300static const struct rt6_info ip6_null_entry_template = {
d8d1f30b
CG		 301 .dst = {
302 .__refcnt = ATOMIC_INIT(1),
303 .__use = 1,
2c20cbd7 304 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 305 .error = -ENETUNREACH,
d8d1f30b
CG		 306 .input = ip6_pkt_discard,
307 .output = ip6_pkt_discard_out,
1da177e4
LT		 308 },
309 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
1da177e4
LT		 310};
311
101367c2
TG		 312#ifdef CONFIG_IPV6_MULTIPLE_TABLES
313
fb0af4c7 314static const struct rt6_info ip6_prohibit_entry_template = {
d8d1f30b
CG		 315 .dst = {
316 .__refcnt = ATOMIC_INIT(1),
317 .__use = 1,
2c20cbd7 318 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 319 .error = -EACCES,
d8d1f30b
CG		 320 .input = ip6_pkt_prohibit,
321 .output = ip6_pkt_prohibit_out,
101367c2
TG		 322 },
323 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 324};
325
fb0af4c7 326static const struct rt6_info ip6_blk_hole_entry_template = {
d8d1f30b
CG		 327 .dst = {
328 .__refcnt = ATOMIC_INIT(1),
329 .__use = 1,
2c20cbd7 330 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 331 .error = -EINVAL,
d8d1f30b 332 .input = dst_discard,
ede2059d 333 .output = dst_discard_out,
101367c2
TG		 334 },
335 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 336};
337
338#endif
339
ebfa45f0
MKL		 340static void rt6_info_init(struct rt6_info *rt)
341{
342 struct dst_entry *dst = &rt->dst;
343
344 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
ebfa45f0
MKL		 345 INIT_LIST_HEAD(&rt->rt6i_uncached);
346}
347
1da177e4 348/* allocate dst with ip6_dst_ops */
93531c67
DA		 349struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
350 int flags)
1da177e4 351{
97bab73f 352 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
b2a9c0ed 353 1, DST_OBSOLETE_FORCE_CHK, flags);
cf911662 354
81eb8447 355 if (rt) {
ebfa45f0 356 rt6_info_init(rt);
81eb8447
WW		 357 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
358 }
8104891b 359
cf911662 360 return rt;
1da177e4 361}
9ab179d8 362EXPORT_SYMBOL(ip6_dst_alloc);
d52d3997 363
1da177e4
LT		 364static void ip6_dst_destroy(struct dst_entry *dst)
365{
366 struct rt6_info *rt = (struct rt6_info *)dst;
a68886a6 367 struct fib6_info *from;
8d0b94af 368 struct inet6_dev *idev;
1da177e4 369
1620a336 370 ip_dst_metrics_put(dst);
8d0b94af
MKL		 371 rt6_uncached_list_del(rt);
372
373 idev = rt->rt6i_idev;
38308473 374 if (idev) {
1da177e4
LT		 375 rt->rt6i_idev = NULL;
376 in6_dev_put(idev);
1ab1457c 377 }
1716a961 378
0e233874 379 from = xchg((__force struct fib6_info **)&rt->from, NULL);
93531c67 380 fib6_info_release(from);
b3419363
DM		 381}
382
1da177e4
LT		 383static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
384 int how)
385{
386 struct rt6_info *rt = (struct rt6_info *)dst;
387 struct inet6_dev *idev = rt->rt6i_idev;
5a3e55d6 388 struct net_device *loopback_dev =
c346dca1 389 dev_net(dev)->loopback_dev;
1da177e4 390
e5645f51
WW		 391 if (idev && idev->dev != loopback_dev) {
392 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
393 if (loopback_idev) {
394 rt->rt6i_idev = loopback_idev;
395 in6_dev_put(idev);
97cac082 396 }
1da177e4
LT		 397 }
398}
399
5973fb1e
MKL		 400static bool __rt6_check_expired(const struct rt6_info *rt)
401{
402 if (rt->rt6i_flags & RTF_EXPIRES)
403 return time_after(jiffies, rt->dst.expires);
404 else
405 return false;
406}
407
a50feda5 408static bool rt6_check_expired(const struct rt6_info *rt)
1da177e4 409{
a68886a6
DA		 410 struct fib6_info *from;
411
412 from = rcu_dereference(rt->from);
413
1716a961
G		 414 if (rt->rt6i_flags & RTF_EXPIRES) {
415 if (time_after(jiffies, rt->dst.expires))
a50feda5 416 return true;
a68886a6 417 } else if (from) {
1e2ea8ad 418 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
a68886a6 419 fib6_check_expired(from);
1716a961 420 }
a50feda5 421 return false;
1da177e4
LT		 422}
423
b1d40991
DA		 424void fib6_select_path(const struct net *net, struct fib6_result *res,
425 struct flowi6 *fl6, int oif, bool have_oif_match,
426 const struct sk_buff *skb, int strict)
51ebd318 427{
8d1c802b 428 struct fib6_info *sibling, *next_sibling;
b1d40991
DA		 429 struct fib6_info *match = res->f6i;
430
f88d8ea6 431 if ((!match->fib6_nsiblings && !match->nh) || have_oif_match)
b1d40991 432 goto out;
51ebd318 433
b673d6cc
JS		 434 /* We might have already computed the hash for ICMPv6 errors. In such
435 * case it will always be non-zero. Otherwise now is the time to do it.
436 */
f88d8ea6
DA		 437 if (!fl6->mp_hash &&
438 (!match->nh || nexthop_is_multipath(match->nh)))
b4bac172 439 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
b673d6cc 440
f88d8ea6
DA		 441 if (unlikely(match->nh)) {
442 nexthop_path_fib6_result(res, fl6->mp_hash);
443 return;
444 }
445
1cf844c7 446 if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
b1d40991 447 goto out;
3d709f69 448
93c2fb25
DA		 449 list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
450 fib6_siblings) {
1cf844c7 451 const struct fib6_nh *nh = sibling->fib6_nh;
5e670d84
DA		 452 int nh_upper_bound;
453
702cea56 454 nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
5e670d84 455 if (fl6->mp_hash > nh_upper_bound)
3d709f69 456 continue;
702cea56 457 if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
3d709f69
IS		 458 break;
459 match = sibling;
460 break;
461 }
462
b1d40991
DA		 463out:
464 res->f6i = match;
1cf844c7 465 res->nh = match->fib6_nh;
51ebd318
ND		 466}
467
1da177e4 468/*
66f5d6ce 469 * Route lookup. rcu_read_lock() should be held.
1da177e4
LT		 470 */
471
0c59d006
DA		 472static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
473 const struct in6_addr *saddr, int oif, int flags)
474{
475 const struct net_device *dev;
476
477 if (nh->fib_nh_flags & RTNH_F_DEAD)
478 return false;
479
480 dev = nh->fib_nh_dev;
481 if (oif) {
482 if (dev->ifindex == oif)
483 return true;
484 } else {
485 if (ipv6_chk_addr(net, saddr, dev,
486 flags & RT6_LOOKUP_F_IFACE))
487 return true;
488 }
489
490 return false;
491}
492
962b6803
DA		 493struct fib6_nh_dm_arg {
494 struct net *net;
495 const struct in6_addr *saddr;
496 int oif;
497 int flags;
498 struct fib6_nh *nh;
499};
500
501static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
502{
503 struct fib6_nh_dm_arg *arg = _arg;
504
505 arg->nh = nh;
506 return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
507 arg->flags);
508}
509
510/* returns fib6_nh from nexthop or NULL */
511static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
512 struct fib6_result *res,
513 const struct in6_addr *saddr,
514 int oif, int flags)
515{
516 struct fib6_nh_dm_arg arg = {
517 .net = net,
518 .saddr = saddr,
519 .oif = oif,
520 .flags = flags,
521 };
522
523 if (nexthop_is_blackhole(nh))
524 return NULL;
525
526 if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
527 return arg.nh;
528
529 return NULL;
530}
531
75ef7389
DA		 532static void rt6_device_match(struct net *net, struct fib6_result *res,
533 const struct in6_addr *saddr, int oif, int flags)
1da177e4 534{
75ef7389
DA		 535 struct fib6_info *f6i = res->f6i;
536 struct fib6_info *spf6i;
537 struct fib6_nh *nh;
1da177e4 538
75ef7389 539 if (!oif && ipv6_addr_any(saddr)) {
f88d8ea6
DA		 540 if (unlikely(f6i->nh)) {
541 nh = nexthop_fib6_nh(f6i->nh);
542 if (nexthop_is_blackhole(f6i->nh))
543 goto out_blackhole;
544 } else {
545 nh = f6i->fib6_nh;
546 }
7d21fec9
DA		 547 if (!(nh->fib_nh_flags & RTNH_F_DEAD))
548 goto out;
75ef7389 549 }
dd3abc4e 550
75ef7389 551 for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
962b6803
DA		 552 bool matched = false;
553
554 if (unlikely(spf6i->nh)) {
555 nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
556 oif, flags);
557 if (nh)
558 matched = true;
559 } else {
560 nh = spf6i->fib6_nh;
561 if (__rt6_device_match(net, nh, saddr, oif, flags))
562 matched = true;
563 }
564 if (matched) {
75ef7389 565 res->f6i = spf6i;
7d21fec9 566 goto out;
75ef7389 567 }
dd3abc4e 568 }
1da177e4 569
75ef7389
DA		 570 if (oif && flags & RT6_LOOKUP_F_IFACE) {
571 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 572 nh = res->f6i->fib6_nh;
7d21fec9 573 goto out;
75ef7389 574 }
8067bb8c 575
f88d8ea6
DA		 576 if (unlikely(f6i->nh)) {
577 nh = nexthop_fib6_nh(f6i->nh);
578 if (nexthop_is_blackhole(f6i->nh))
579 goto out_blackhole;
580 } else {
581 nh = f6i->fib6_nh;
582 }
583
7d21fec9 584 if (nh->fib_nh_flags & RTNH_F_DEAD) {
75ef7389 585 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 586 nh = res->f6i->fib6_nh;
75ef7389 587 }
7d21fec9
DA		 588out:
589 res->nh = nh;
590 res->fib6_type = res->f6i->fib6_type;
591 res->fib6_flags = res->f6i->fib6_flags;
f88d8ea6
DA		 592 return;
593
594out_blackhole:
595 res->fib6_flags |= RTF_REJECT;
596 res->fib6_type = RTN_BLACKHOLE;
597 res->nh = nh;
1da177e4
LT		 598}
599
27097255 600#ifdef CONFIG_IPV6_ROUTER_PREF
c2f17e82
HFS		 601struct __rt6_probe_work {
602 struct work_struct work;
603 struct in6_addr target;
604 struct net_device *dev;
605};
606
607static void rt6_probe_deferred(struct work_struct *w)
608{
609 struct in6_addr mcaddr;
610 struct __rt6_probe_work *work =
611 container_of(w, struct __rt6_probe_work, work);
612
613 addrconf_addr_solict_mult(&work->target, &mcaddr);
adc176c5 614 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
c2f17e82 615 dev_put(work->dev);
662f5533 616 kfree(work);
c2f17e82
HFS		 617}
618
cc3a86c8 619static void rt6_probe(struct fib6_nh *fib6_nh)
27097255 620{
f547fac6 621 struct __rt6_probe_work *work = NULL;
5e670d84 622 const struct in6_addr *nh_gw;
f2c31e32 623 struct neighbour *neigh;
5e670d84 624 struct net_device *dev;
f547fac6 625 struct inet6_dev *idev;
5e670d84 626
27097255
YH		 627 /*
628 * Okay, this does not seem to be appropriate
629 * for now, however, we need to check if it
630 * is really so; aka Router Reachability Probing.
631 *
632 * Router Reachability Probe MUST be rate-limited
633 * to no more than one per minute.
634 */
cc3a86c8 635 if (fib6_nh->fib_nh_gw_family)
7ff74a59 636 return;
5e670d84 637
cc3a86c8
DA		 638 nh_gw = &fib6_nh->fib_nh_gw6;
639 dev = fib6_nh->fib_nh_dev;
2152caea 640 rcu_read_lock_bh();
f547fac6 641 idev = __in6_dev_get(dev);
5e670d84 642 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
2152caea 643 if (neigh) {
8d6c31bf
MKL		 644 if (neigh->nud_state & NUD_VALID)
645 goto out;
646
2152caea 647 write_lock(&neigh->lock);
990edb42
MKL		 648 if (!(neigh->nud_state & NUD_VALID) &&
649 time_after(jiffies,
dcd1f572 650 neigh->updated + idev->cnf.rtr_probe_interval)) {
990edb42
MKL		 651 work = kmalloc(sizeof(*work), GFP_ATOMIC);
652 if (work)
653 __neigh_set_probe_once(neigh);
c2f17e82 654 }
2152caea 655 write_unlock(&neigh->lock);
cc3a86c8 656 } else if (time_after(jiffies, fib6_nh->last_probe +
f547fac6 657 idev->cnf.rtr_probe_interval)) {
990edb42 658 work = kmalloc(sizeof(*work), GFP_ATOMIC);
f2c31e32 659 }
990edb42
MKL		 660
661 if (work) {
cc3a86c8 662 fib6_nh->last_probe = jiffies;
990edb42 663 INIT_WORK(&work->work, rt6_probe_deferred);
5e670d84
DA		 664 work->target = *nh_gw;
665 dev_hold(dev);
666 work->dev = dev;
990edb42
MKL		 667 schedule_work(&work->work);
668 }
669
8d6c31bf 670out:
2152caea 671 rcu_read_unlock_bh();
27097255
YH		 672}
673#else
cc3a86c8 674static inline void rt6_probe(struct fib6_nh *fib6_nh)
27097255 675{
27097255
YH		 676}
677#endif
678
1da177e4 679/*
554cfb7e 680 * Default Router Selection (RFC 2461 6.3.6)
1da177e4 681 */
1ba9a895 682static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
1da177e4 683{
afc154e9 684 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
5e670d84 685 struct neighbour *neigh;
f2c31e32 686
145a3621 687 rcu_read_lock_bh();
1ba9a895
DA		 688 neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
689 &fib6_nh->fib_nh_gw6);
145a3621
YH		 690 if (neigh) {
691 read_lock(&neigh->lock);
554cfb7e 692 if (neigh->nud_state & NUD_VALID)
afc154e9 693 ret = RT6_NUD_SUCCEED;
398bcbeb 694#ifdef CONFIG_IPV6_ROUTER_PREF
a5a81f0b 695 else if (!(neigh->nud_state & NUD_FAILED))
afc154e9 696 ret = RT6_NUD_SUCCEED;
7e980569
JB		 697 else
698 ret = RT6_NUD_FAIL_PROBE;
398bcbeb 699#endif
145a3621 700 read_unlock(&neigh->lock);
afc154e9
HFS		 701 } else {
702 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
7e980569 703 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
a5a81f0b 704 }
145a3621
YH		 705 rcu_read_unlock_bh();
706
a5a81f0b 707 return ret;
1da177e4
LT		 708}
709
702cea56
DA		 710static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
711 int strict)
1da177e4 712{
6e1809a5
DA		 713 int m = 0;
714
715 if (!oif || nh->fib_nh_dev->ifindex == oif)
716 m = 2;
1ab1457c 717
77d16f45 718 if (!m && (strict & RT6_LOOKUP_F_IFACE))
afc154e9 719 return RT6_NUD_FAIL_HARD;
ebacaaa0 720#ifdef CONFIG_IPV6_ROUTER_PREF
702cea56 721 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
ebacaaa0 722#endif
1ba9a895 723 if ((strict & RT6_LOOKUP_F_REACHABLE) &&
702cea56 724 !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
1ba9a895 725 int n = rt6_check_neigh(nh);
afc154e9
HFS		 726 if (n < 0)
727 return n;
728 }
554cfb7e
YH		 729 return m;
730}
731
28679ed1
DA		 732static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
733 int oif, int strict, int *mpri, bool *do_rr)
554cfb7e 734{
afc154e9 735 bool match_do_rr = false;
28679ed1
DA		 736 bool rc = false;
737 int m;
35103d11 738
28679ed1 739 if (nh->fib_nh_flags & RTNH_F_DEAD)
8067bb8c
IS		 740 goto out;
741
28679ed1
DA		 742 if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
743 nh->fib_nh_flags & RTNH_F_LINKDOWN &&
d5d32e4b 744 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
35103d11 745 goto out;
f11e6659 746
28679ed1 747 m = rt6_score_route(nh, fib6_flags, oif, strict);
7e980569 748 if (m == RT6_NUD_FAIL_DO_RR) {
afc154e9
HFS		 749 match_do_rr = true;
750 m = 0; /* lowest valid score */
7e980569 751 } else if (m == RT6_NUD_FAIL_HARD) {
f11e6659 752 goto out;
afc154e9
HFS		 753 }
754
755 if (strict & RT6_LOOKUP_F_REACHABLE)
28679ed1 756 rt6_probe(nh);
f11e6659 757
7e980569 758 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
f11e6659 759 if (m > *mpri) {
afc154e9 760 *do_rr = match_do_rr;
f11e6659 761 *mpri = m;
28679ed1 762 rc = true;
f11e6659 763 }
f11e6659 764out:
28679ed1 765 return rc;
f11e6659
DM		 766}
767
17a5984e
DA		 768struct fib6_nh_frl_arg {
769 u32 flags;
770 int oif;
771 int strict;
772 int *mpri;
773 bool *do_rr;
774 struct fib6_nh *nh;
775};
776
777static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
778{
779 struct fib6_nh_frl_arg *arg = _arg;
780
781 arg->nh = nh;
782 return find_match(nh, arg->flags, arg->oif, arg->strict,
783 arg->mpri, arg->do_rr);
784}
785
b7bc4b6a 786static void __find_rr_leaf(struct fib6_info *f6i_start,
30c15f03 787 struct fib6_info *nomatch, u32 metric,
b7bc4b6a 788 struct fib6_result *res, struct fib6_info **cont,
30c15f03 789 int oif, int strict, bool *do_rr, int *mpri)
f11e6659 790{
b7bc4b6a 791 struct fib6_info *f6i;
1da177e4 792
b7bc4b6a
DA		 793 for (f6i = f6i_start;
794 f6i && f6i != nomatch;
795 f6i = rcu_dereference(f6i->fib6_next)) {
17a5984e 796 bool matched = false;
30c15f03
DA		 797 struct fib6_nh *nh;
798
b7bc4b6a
DA		 799 if (cont && f6i->fib6_metric != metric) {
800 *cont = f6i;
30c15f03 801 return;
9fbdcfaf
SK		 802 }
803
b7bc4b6a 804 if (fib6_check_expired(f6i))
28679ed1
DA		 805 continue;
806
17a5984e
DA		 807 if (unlikely(f6i->nh)) {
808 struct fib6_nh_frl_arg arg = {
809 .flags = f6i->fib6_flags,
810 .oif = oif,
811 .strict = strict,
812 .mpri = mpri,
813 .do_rr = do_rr
814 };
815
816 if (nexthop_is_blackhole(f6i->nh)) {
817 res->fib6_flags = RTF_REJECT;
818 res->fib6_type = RTN_BLACKHOLE;
819 res->f6i = f6i;
820 res->nh = nexthop_fib6_nh(f6i->nh);
821 return;
822 }
823 if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
824 &arg)) {
825 matched = true;
826 nh = arg.nh;
827 }
828 } else {
829 nh = f6i->fib6_nh;
830 if (find_match(nh, f6i->fib6_flags, oif, strict,
831 mpri, do_rr))
832 matched = true;
833 }
834 if (matched) {
b7bc4b6a
DA		 835 res->f6i = f6i;
836 res->nh = nh;
7d21fec9
DA		 837 res->fib6_flags = f6i->fib6_flags;
838 res->fib6_type = f6i->fib6_type;
b7bc4b6a 839 }
9fbdcfaf 840 }
30c15f03 841}
9fbdcfaf 842
b7bc4b6a
DA		 843static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
844 struct fib6_info *rr_head, int oif, int strict,
845 bool *do_rr, struct fib6_result *res)
30c15f03 846{
b7bc4b6a
DA		 847 u32 metric = rr_head->fib6_metric;
848 struct fib6_info *cont = NULL;
30c15f03 849 int mpri = -1;
9fbdcfaf 850
b7bc4b6a 851 __find_rr_leaf(rr_head, NULL, metric, res, &cont,
30c15f03 852 oif, strict, do_rr, &mpri);
28679ed1 853
b7bc4b6a 854 __find_rr_leaf(leaf, rr_head, metric, res, &cont,
30c15f03 855 oif, strict, do_rr, &mpri);
9fbdcfaf 856
b7bc4b6a
DA		 857 if (res->f6i || !cont)
858 return;
9fbdcfaf 859
b7bc4b6a 860 __find_rr_leaf(cont, NULL, metric, res, NULL,
30c15f03 861 oif, strict, do_rr, &mpri);
f11e6659 862}
1da177e4 863
b7bc4b6a
DA		 864static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
865 struct fib6_result *res, int strict)
f11e6659 866{
8d1c802b 867 struct fib6_info *leaf = rcu_dereference(fn->leaf);
b7bc4b6a 868 struct fib6_info *rt0;
afc154e9 869 bool do_rr = false;
17ecf590 870 int key_plen;
1da177e4 871
b7bc4b6a
DA		 872 /* make sure this function or its helpers sets f6i */
873 res->f6i = NULL;
874
421842ed 875 if (!leaf || leaf == net->ipv6.fib6_null_entry)
b7bc4b6a 876 goto out;
8d1040e8 877
66f5d6ce 878 rt0 = rcu_dereference(fn->rr_ptr);
f11e6659 879 if (!rt0)
66f5d6ce 880 rt0 = leaf;
1da177e4 881
17ecf590
WW		 882 /* Double check to make sure fn is not an intermediate node
883 * and fn->leaf does not points to its child's leaf
884 * (This might happen if all routes under fn are deleted from
885 * the tree and fib6_repair_tree() is called on the node.)
886 */
93c2fb25 887 key_plen = rt0->fib6_dst.plen;
17ecf590 888#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 889 if (rt0->fib6_src.plen)
890 key_plen = rt0->fib6_src.plen;
17ecf590
WW		 891#endif
892 if (fn->fn_bit != key_plen)
b7bc4b6a 893 goto out;
1da177e4 894
b7bc4b6a 895 find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
afc154e9 896 if (do_rr) {
8fb11a9a 897 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
f11e6659 898
554cfb7e 899 /* no entries matched; do round-robin */
93c2fb25 900 if (!next || next->fib6_metric != rt0->fib6_metric)
8d1040e8 901 next = leaf;
f11e6659 902
66f5d6ce 903 if (next != rt0) {
93c2fb25 904 spin_lock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 905 /* make sure next is not being deleted from the tree */
93c2fb25 906 if (next->fib6_node)
66f5d6ce 907 rcu_assign_pointer(fn->rr_ptr, next);
93c2fb25 908 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 909 }
1da177e4 910 }
1da177e4 911
b7bc4b6a
DA		 912out:
913 if (!res->f6i) {
914 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 915 res->nh = res->f6i->fib6_nh;
7d21fec9
DA		 916 res->fib6_flags = res->f6i->fib6_flags;
917 res->fib6_type = res->f6i->fib6_type;
b7bc4b6a 918 }
1da177e4
LT		 919}
920
85bd05de 921static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
8b9df265 922{
85bd05de
DA		 923 return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
924 res->nh->fib_nh_gw_family;
8b9df265
MKL		 925}
926
70ceb4f5
YH		 927#ifdef CONFIG_IPV6_ROUTE_INFO
928int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
b71d1d42 929 const struct in6_addr *gwaddr)
70ceb4f5 930{
c346dca1 931 struct net *net = dev_net(dev);
70ceb4f5
YH		 932 struct route_info *rinfo = (struct route_info *) opt;
933 struct in6_addr prefix_buf, *prefix;
934 unsigned int pref;
4bed72e4 935 unsigned long lifetime;
8d1c802b 936 struct fib6_info *rt;
70ceb4f5
YH		 937
938 if (len < sizeof(struct route_info)) {
939 return -EINVAL;
940 }
941
942 /* Sanity check for prefix_len and length */
943 if (rinfo->length > 3) {
944 return -EINVAL;
945 } else if (rinfo->prefix_len > 128) {
946 return -EINVAL;
947 } else if (rinfo->prefix_len > 64) {
948 if (rinfo->length < 2) {
949 return -EINVAL;
950 }
951 } else if (rinfo->prefix_len > 0) {
952 if (rinfo->length < 1) {
953 return -EINVAL;
954 }
955 }
956
957 pref = rinfo->route_pref;
958 if (pref == ICMPV6_ROUTER_PREF_INVALID)
3933fc95 959 return -EINVAL;
70ceb4f5 960
4bed72e4 961 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
70ceb4f5
YH		 962
963 if (rinfo->length == 3)
964 prefix = (struct in6_addr *)rinfo->prefix;
965 else {
966 /* this function is safe */
967 ipv6_addr_prefix(&prefix_buf,
968 (struct in6_addr *)rinfo->prefix,
969 rinfo->prefix_len);
970 prefix = &prefix_buf;
971 }
972
f104a567 973 if (rinfo->prefix_len == 0)
afb1d4b5 974 rt = rt6_get_dflt_router(net, gwaddr, dev);
f104a567
DJ		 975 else
976 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
830218c1 977 gwaddr, dev);
70ceb4f5
YH		 978
979 if (rt && !lifetime) {
afb1d4b5 980 ip6_del_rt(net, rt);
70ceb4f5
YH		 981 rt = NULL;
982 }
983
984 if (!rt && lifetime)
830218c1
DA		 985 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
986 dev, pref);
70ceb4f5 987 else if (rt)
93c2fb25
DA		 988 rt->fib6_flags = RTF_ROUTEINFO |
989 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
70ceb4f5
YH		 990
991 if (rt) {
1716a961 992 if (!addrconf_finite_timeout(lifetime))
14895687 993 fib6_clean_expires(rt);
1716a961 994 else
14895687 995 fib6_set_expires(rt, jiffies + HZ * lifetime);
1716a961 996
93531c67 997 fib6_info_release(rt);
70ceb4f5
YH		 998 }
999 return 0;
1000}
1001#endif
1002
ae90d867
DA		 1003/*
1004 * Misc support functions
1005 */
1006
1007/* called with rcu_lock held */
0d161581 1008static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
ae90d867 1009{
0d161581 1010 struct net_device *dev = res->nh->fib_nh_dev;
ae90d867 1011
7d21fec9 1012 if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
ae90d867
DA		 1013 /* for copies of local routes, dst->dev needs to be the
1014 * device if it is a master device, the master device if
1015 * device is enslaved, and the loopback as the default
1016 */
1017 if (netif_is_l3_slave(dev) &&
7d21fec9 1018 !rt6_need_strict(&res->f6i->fib6_dst.addr))
ae90d867
DA		 1019 dev = l3mdev_master_dev_rcu(dev);
1020 else if (!netif_is_l3_master(dev))
1021 dev = dev_net(dev)->loopback_dev;
1022 /* last case is netif_is_l3_master(dev) is true in which
1023 * case we want dev returned to be dev
1024 */
1025 }
1026
1027 return dev;
1028}
1029
6edb3c96
DA		 1030static const int fib6_prop[RTN_MAX + 1] = {
1031 [RTN_UNSPEC] = 0,
1032 [RTN_UNICAST] = 0,
1033 [RTN_LOCAL] = 0,
1034 [RTN_BROADCAST] = 0,
1035 [RTN_ANYCAST] = 0,
1036 [RTN_MULTICAST] = 0,
1037 [RTN_BLACKHOLE] = -EINVAL,
1038 [RTN_UNREACHABLE] = -EHOSTUNREACH,
1039 [RTN_PROHIBIT] = -EACCES,
1040 [RTN_THROW] = -EAGAIN,
1041 [RTN_NAT] = -EINVAL,
1042 [RTN_XRESOLVE] = -EINVAL,
1043};
1044
1045static int ip6_rt_type_to_error(u8 fib6_type)
1046{
1047 return fib6_prop[fib6_type];
1048}
1049
8d1c802b 1050static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
3b6761d1
DA		 1051{
1052 unsigned short flags = 0;
1053
1054 if (rt->dst_nocount)
1055 flags |= DST_NOCOUNT;
1056 if (rt->dst_nopolicy)
1057 flags |= DST_NOPOLICY;
1058 if (rt->dst_host)
1059 flags |= DST_HOST;
1060
1061 return flags;
1062}
1063
7d21fec9 1064static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
6edb3c96 1065{
7d21fec9 1066 rt->dst.error = ip6_rt_type_to_error(fib6_type);
6edb3c96 1067
7d21fec9 1068 switch (fib6_type) {
6edb3c96
DA		 1069 case RTN_BLACKHOLE:
1070 rt->dst.output = dst_discard_out;
1071 rt->dst.input = dst_discard;
1072 break;
1073 case RTN_PROHIBIT:
1074 rt->dst.output = ip6_pkt_prohibit_out;
1075 rt->dst.input = ip6_pkt_prohibit;
1076 break;
1077 case RTN_THROW:
1078 case RTN_UNREACHABLE:
1079 default:
1080 rt->dst.output = ip6_pkt_discard_out;
1081 rt->dst.input = ip6_pkt_discard;
1082 break;
1083 }
1084}
1085
0d161581 1086static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
6edb3c96 1087{
7d21fec9 1088 struct fib6_info *f6i = res->f6i;
0d161581 1089
7d21fec9
DA		 1090 if (res->fib6_flags & RTF_REJECT) {
1091 ip6_rt_init_dst_reject(rt, res->fib6_type);
6edb3c96
DA		 1092 return;
1093 }
1094
1095 rt->dst.error = 0;
1096 rt->dst.output = ip6_output;
1097
7d21fec9 1098 if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
6edb3c96 1099 rt->dst.input = ip6_input;
7d21fec9 1100 } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
6edb3c96
DA		 1101 rt->dst.input = ip6_mc_input;
1102 } else {
1103 rt->dst.input = ip6_forward;
1104 }
1105
0d161581
DA		 1106 if (res->nh->fib_nh_lws) {
1107 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
6edb3c96
DA		 1108 lwtunnel_set_redirect(&rt->dst);
1109 }
1110
1111 rt->dst.lastuse = jiffies;
1112}
1113
e873e4b9 1114/* Caller must already hold reference to @from */
8d1c802b 1115static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
ae90d867 1116{
ae90d867 1117 rt->rt6i_flags &= ~RTF_EXPIRES;
a68886a6 1118 rcu_assign_pointer(rt->from, from);
e1255ed4 1119 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
ae90d867
DA		 1120}
1121
0d161581
DA		 1122/* Caller must already hold reference to f6i in result */
1123static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
ae90d867 1124{
0d161581
DA		 1125 const struct fib6_nh *nh = res->nh;
1126 const struct net_device *dev = nh->fib_nh_dev;
1127 struct fib6_info *f6i = res->f6i;
dcd1f572 1128
0d161581 1129 ip6_rt_init_dst(rt, res);
6edb3c96 1130
0d161581 1131 rt->rt6i_dst = f6i->fib6_dst;
dcd1f572 1132 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
7d21fec9 1133 rt->rt6i_flags = res->fib6_flags;
0d161581
DA		 1134 if (nh->fib_nh_gw_family) {
1135 rt->rt6i_gateway = nh->fib_nh_gw6;
2b2450ca
DA		 1136 rt->rt6i_flags |= RTF_GATEWAY;
1137 }
0d161581 1138 rt6_set_from(rt, f6i);
ae90d867 1139#ifdef CONFIG_IPV6_SUBTREES
0d161581 1140 rt->rt6i_src = f6i->fib6_src;
ae90d867 1141#endif
ae90d867
DA		 1142}
1143
a3c00e46
MKL		 1144static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1145 struct in6_addr *saddr)
1146{
66f5d6ce 1147 struct fib6_node *pn, *sn;
a3c00e46
MKL		 1148 while (1) {
1149 if (fn->fn_flags & RTN_TL_ROOT)
1150 return NULL;
66f5d6ce
WW		 1151 pn = rcu_dereference(fn->parent);
1152 sn = FIB6_SUBTREE(pn);
1153 if (sn && sn != fn)
6454743b 1154 fn = fib6_node_lookup(sn, NULL, saddr);
a3c00e46
MKL		 1155 else
1156 fn = pn;
1157 if (fn->fn_flags & RTN_RTINFO)
1158 return fn;
1159 }
1160}
c71099ac 1161
10585b43 1162static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
d3843fe5
WW		 1163{
1164 struct rt6_info *rt = *prt;
1165
1166 if (dst_hold_safe(&rt->dst))
1167 return true;
10585b43 1168 if (net) {
d3843fe5
WW		 1169 rt = net->ipv6.ip6_null_entry;
1170 dst_hold(&rt->dst);
1171 } else {
1172 rt = NULL;
1173 }
1174 *prt = rt;
1175 return false;
1176}
1177
dec9b0e2 1178/* called with rcu_lock held */
9b6b35ab 1179static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
dec9b0e2 1180{
9b6b35ab
DA		 1181 struct net_device *dev = res->nh->fib_nh_dev;
1182 struct fib6_info *f6i = res->f6i;
1183 unsigned short flags;
dec9b0e2
DA		 1184 struct rt6_info *nrt;
1185
9b6b35ab 1186 if (!fib6_info_hold_safe(f6i))
1c87e79a 1187 goto fallback;
e873e4b9 1188
9b6b35ab 1189 flags = fib6_info_dst_flags(f6i);
93531c67 1190 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1c87e79a 1191 if (!nrt) {
9b6b35ab 1192 fib6_info_release(f6i);
1c87e79a
XL		 1193 goto fallback;
1194 }
dec9b0e2 1195
0d161581 1196 ip6_rt_copy_init(nrt, res);
1c87e79a
XL		 1197 return nrt;
1198
1199fallback:
1200 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1201 dst_hold(&nrt->dst);
dec9b0e2
DA		 1202 return nrt;
1203}
1204
8ed67789
DL		 1205static struct rt6_info *ip6_pol_route_lookup(struct net *net,
1206 struct fib6_table *table,
b75cc8f9
DA		 1207 struct flowi6 *fl6,
1208 const struct sk_buff *skb,
1209 int flags)
1da177e4 1210{
b1d40991 1211 struct fib6_result res = {};
1da177e4 1212 struct fib6_node *fn;
23fb93a4 1213 struct rt6_info *rt;
1da177e4 1214
b6cdbc85
DA		 1215 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1216 flags &= ~RT6_LOOKUP_F_IFACE;
1217
66f5d6ce 1218 rcu_read_lock();
6454743b 1219 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
c71099ac 1220restart:
b1d40991
DA		 1221 res.f6i = rcu_dereference(fn->leaf);
1222 if (!res.f6i)
1223 res.f6i = net->ipv6.fib6_null_entry;
af52a52c 1224 else
75ef7389
DA		 1225 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1226 flags);
af52a52c 1227
b1d40991 1228 if (res.f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 1229 fn = fib6_backtrack(fn, &fl6->saddr);
1230 if (fn)
1231 goto restart;
2b760fcf 1232
af52a52c
DA		 1233 rt = net->ipv6.ip6_null_entry;
1234 dst_hold(&rt->dst);
1235 goto out;
f88d8ea6
DA		 1236 } else if (res.fib6_flags & RTF_REJECT) {
1237 goto do_create;
af52a52c 1238 }
d3843fe5 1239
b1d40991
DA		 1240 fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1241 fl6->flowi6_oif != 0, skb, flags);
1242
2b760fcf 1243 /* Search through exception table */
7e4b5128 1244 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
23fb93a4 1245 if (rt) {
10585b43 1246 if (ip6_hold_safe(net, &rt))
dec9b0e2 1247 dst_use_noref(&rt->dst, jiffies);
23fb93a4 1248 } else {
f88d8ea6 1249do_create:
9b6b35ab 1250 rt = ip6_create_rt_rcu(&res);
dec9b0e2 1251 }
b811580d 1252
af52a52c 1253out:
8ff2e5b2 1254 trace_fib6_table_lookup(net, &res, table, fl6);
af52a52c 1255
66f5d6ce 1256 rcu_read_unlock();
b811580d 1257
c71099ac 1258 return rt;
c71099ac
TG		 1259}
1260
67ba4152 1261struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
b75cc8f9 1262 const struct sk_buff *skb, int flags)
ea6e574e 1263{
b75cc8f9 1264 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
ea6e574e
FW		 1265}
1266EXPORT_SYMBOL_GPL(ip6_route_lookup);
1267
9acd9f3a 1268struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
b75cc8f9
DA		 1269 const struct in6_addr *saddr, int oif,
1270 const struct sk_buff *skb, int strict)
c71099ac 1271{
4c9483b2
DM		 1272 struct flowi6 fl6 = {
1273 .flowi6_oif = oif,
1274 .daddr = *daddr,
c71099ac
TG		 1275 };
1276 struct dst_entry *dst;
77d16f45 1277 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
c71099ac 1278
adaa70bb 1279 if (saddr) {
4c9483b2 1280 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
adaa70bb
TG		 1281 flags |= RT6_LOOKUP_F_HAS_SADDR;
1282 }
1283
b75cc8f9 1284 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
c71099ac
TG		 1285 if (dst->error == 0)
1286 return (struct rt6_info *) dst;
1287
1288 dst_release(dst);
1289
1da177e4
LT		 1290 return NULL;
1291}
7159039a
YH		 1292EXPORT_SYMBOL(rt6_lookup);
1293
c71099ac 1294/* ip6_ins_rt is called with FREE table->tb6_lock.
1cfb71ee
WW		 1295 * It takes new route entry, the addition fails by any reason the
1296 * route is released.
1297 * Caller must hold dst before calling it.
1da177e4
LT		 1298 */
1299
8d1c802b 1300static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
333c4301 1301 struct netlink_ext_ack *extack)
1da177e4
LT		 1302{
1303 int err;
c71099ac 1304 struct fib6_table *table;
1da177e4 1305
93c2fb25 1306 table = rt->fib6_table;
66f5d6ce 1307 spin_lock_bh(&table->tb6_lock);
d4ead6b3 1308 err = fib6_add(&table->tb6_root, rt, info, extack);
66f5d6ce 1309 spin_unlock_bh(&table->tb6_lock);
1da177e4
LT		 1310
1311 return err;
1312}
1313
8d1c802b 1314int ip6_ins_rt(struct net *net, struct fib6_info *rt)
40e22e8f 1315{
afb1d4b5 1316 struct nl_info info = { .nl_net = net, };
e715b6d3 1317
d4ead6b3 1318 return __ip6_ins_rt(rt, &info, NULL);
40e22e8f
TG		 1319}
1320
85bd05de 1321static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
8b9df265
MKL		 1322 const struct in6_addr *daddr,
1323 const struct in6_addr *saddr)
1da177e4 1324{
85bd05de 1325 struct fib6_info *f6i = res->f6i;
4832c30d 1326 struct net_device *dev;
1da177e4
LT		 1327 struct rt6_info *rt;
1328
1329 /*
1330 * Clone the route.
1331 */
1332
85bd05de 1333 if (!fib6_info_hold_safe(f6i))
e873e4b9
WW		 1334 return NULL;
1335
0d161581 1336 dev = ip6_rt_get_dev_rcu(res);
93531c67 1337 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
e873e4b9 1338 if (!rt) {
85bd05de 1339 fib6_info_release(f6i);
83a09abd 1340 return NULL;
e873e4b9 1341 }
83a09abd 1342
0d161581 1343 ip6_rt_copy_init(rt, res);
83a09abd 1344 rt->rt6i_flags |= RTF_CACHE;
83a09abd
MKL		 1345 rt->dst.flags |= DST_HOST;
1346 rt->rt6i_dst.addr = *daddr;
1347 rt->rt6i_dst.plen = 128;
1da177e4 1348
85bd05de
DA		 1349 if (!rt6_is_gw_or_nonexthop(res)) {
1350 if (f6i->fib6_dst.plen != 128 &&
1351 ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
83a09abd 1352 rt->rt6i_flags |= RTF_ANYCAST;
1da177e4 1353#ifdef CONFIG_IPV6_SUBTREES
83a09abd
MKL		 1354 if (rt->rt6i_src.plen && saddr) {
1355 rt->rt6i_src.addr = *saddr;
1356 rt->rt6i_src.plen = 128;
8b9df265 1357 }
83a09abd 1358#endif
95a9a5ba 1359 }
1da177e4 1360
95a9a5ba
YH		 1361 return rt;
1362}
1da177e4 1363
db3fedee 1364static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
d52d3997 1365{
db3fedee
DA		 1366 struct fib6_info *f6i = res->f6i;
1367 unsigned short flags = fib6_info_dst_flags(f6i);
4832c30d 1368 struct net_device *dev;
d52d3997
MKL		 1369 struct rt6_info *pcpu_rt;
1370
db3fedee 1371 if (!fib6_info_hold_safe(f6i))
e873e4b9
WW		 1372 return NULL;
1373
4832c30d 1374 rcu_read_lock();
0d161581 1375 dev = ip6_rt_get_dev_rcu(res);
93531c67 1376 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags);
4832c30d 1377 rcu_read_unlock();
e873e4b9 1378 if (!pcpu_rt) {
db3fedee 1379 fib6_info_release(f6i);
d52d3997 1380 return NULL;
e873e4b9 1381 }
0d161581 1382 ip6_rt_copy_init(pcpu_rt, res);
d52d3997
MKL		 1383 pcpu_rt->rt6i_flags |= RTF_PCPU;
1384 return pcpu_rt;
1385}
1386
66f5d6ce 1387/* It should be called with rcu_read_lock() acquired */
db3fedee 1388static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
d52d3997 1389{
c353071a 1390 struct rt6_info *pcpu_rt;
d52d3997 1391
c353071a 1392 pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
d52d3997 1393
a73e4195
MKL		 1394 return pcpu_rt;
1395}
1396
afb1d4b5 1397static struct rt6_info *rt6_make_pcpu_route(struct net *net,
db3fedee 1398 const struct fib6_result *res)
a73e4195
MKL		 1399{
1400 struct rt6_info *pcpu_rt, *prev, **p;
d52d3997 1401
db3fedee 1402 pcpu_rt = ip6_rt_pcpu_alloc(res);
0e09edcc
WW		 1403 if (!pcpu_rt)
1404 return NULL;
d52d3997 1405
f40b6ae2 1406 p = this_cpu_ptr(res->nh->rt6i_pcpu);
a94b9367 1407 prev = cmpxchg(p, NULL, pcpu_rt);
951f788a 1408 BUG_ON(prev);
a94b9367 1409
61fb0d01
ED		 1410 if (res->f6i->fib6_destroying) {
1411 struct fib6_info *from;
1412
1413 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1414 fib6_info_release(from);
1415 }
1416
d52d3997
MKL		 1417 return pcpu_rt;
1418}
1419
35732d01
WW		 1420/* exception hash table implementation
1421 */
1422static DEFINE_SPINLOCK(rt6_exception_lock);
1423
1424/* Remove rt6_ex from hash table and free the memory
1425 * Caller must hold rt6_exception_lock
1426 */
1427static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1428 struct rt6_exception *rt6_ex)
1429{
f5b51fe8 1430 struct fib6_info *from;
b2427e67 1431 struct net *net;
81eb8447 1432
35732d01
WW		 1433 if (!bucket || !rt6_ex)
1434 return;
b2427e67
CIK		 1435
1436 net = dev_net(rt6_ex->rt6i->dst.dev);
f5b51fe8
PA		 1437 net->ipv6.rt6_stats->fib_rt_cache--;
1438
1439 /* purge completely the exception to allow releasing the held resources:
1440 * some [sk] cache may keep the dst around for unlimited time
1441 */
0e233874 1442 from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
f5b51fe8
PA		 1443 fib6_info_release(from);
1444 dst_dev_put(&rt6_ex->rt6i->dst);
1445
35732d01 1446 hlist_del_rcu(&rt6_ex->hlist);
77634cc6 1447 dst_release(&rt6_ex->rt6i->dst);
35732d01
WW		 1448 kfree_rcu(rt6_ex, rcu);
1449 WARN_ON_ONCE(!bucket->depth);
1450 bucket->depth--;
1451}
1452
1453/* Remove oldest rt6_ex in bucket and free the memory
1454 * Caller must hold rt6_exception_lock
1455 */
1456static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1457{
1458 struct rt6_exception *rt6_ex, *oldest = NULL;
1459
1460 if (!bucket)
1461 return;
1462
1463 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1464 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1465 oldest = rt6_ex;
1466 }
1467 rt6_remove_exception(bucket, oldest);
1468}
1469
1470static u32 rt6_exception_hash(const struct in6_addr *dst,
1471 const struct in6_addr *src)
1472{
1473 static u32 seed __read_mostly;
1474 u32 val;
1475
1476 net_get_random_once(&seed, sizeof(seed));
1477 val = jhash(dst, sizeof(*dst), seed);
1478
1479#ifdef CONFIG_IPV6_SUBTREES
1480 if (src)
1481 val = jhash(src, sizeof(*src), val);
1482#endif
1483 return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1484}
1485
1486/* Helper function to find the cached rt in the hash table
1487 * and update bucket pointer to point to the bucket for this
1488 * (daddr, saddr) pair
1489 * Caller must hold rt6_exception_lock
1490 */
1491static struct rt6_exception *
1492__rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1493 const struct in6_addr *daddr,
1494 const struct in6_addr *saddr)
1495{
1496 struct rt6_exception *rt6_ex;
1497 u32 hval;
1498
1499 if (!(*bucket) || !daddr)
1500 return NULL;
1501
1502 hval = rt6_exception_hash(daddr, saddr);
1503 *bucket += hval;
1504
1505 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1506 struct rt6_info *rt6 = rt6_ex->rt6i;
1507 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1508
1509#ifdef CONFIG_IPV6_SUBTREES
1510 if (matched && saddr)
1511 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1512#endif
1513 if (matched)
1514 return rt6_ex;
1515 }
1516 return NULL;
1517}
1518
1519/* Helper function to find the cached rt in the hash table
1520 * and update bucket pointer to point to the bucket for this
1521 * (daddr, saddr) pair
1522 * Caller must hold rcu_read_lock()
1523 */
1524static struct rt6_exception *
1525__rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1526 const struct in6_addr *daddr,
1527 const struct in6_addr *saddr)
1528{
1529 struct rt6_exception *rt6_ex;
1530 u32 hval;
1531
1532 WARN_ON_ONCE(!rcu_read_lock_held());
1533
1534 if (!(*bucket) || !daddr)
1535 return NULL;
1536
1537 hval = rt6_exception_hash(daddr, saddr);
1538 *bucket += hval;
1539
1540 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1541 struct rt6_info *rt6 = rt6_ex->rt6i;
1542 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1543
1544#ifdef CONFIG_IPV6_SUBTREES
1545 if (matched && saddr)
1546 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1547#endif
1548 if (matched)
1549 return rt6_ex;
1550 }
1551 return NULL;
1552}
1553
b748f260 1554static unsigned int fib6_mtu(const struct fib6_result *res)
d4ead6b3 1555{
b748f260 1556 const struct fib6_nh *nh = res->nh;
d4ead6b3
DA		 1557 unsigned int mtu;
1558
b748f260
DA		 1559 if (res->f6i->fib6_pmtu) {
1560 mtu = res->f6i->fib6_pmtu;
dcd1f572 1561 } else {
b748f260 1562 struct net_device *dev = nh->fib_nh_dev;
dcd1f572
DA		 1563 struct inet6_dev *idev;
1564
1565 rcu_read_lock();
1566 idev = __in6_dev_get(dev);
1567 mtu = idev->cnf.mtu6;
1568 rcu_read_unlock();
1569 }
1570
d4ead6b3
DA		 1571 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1572
b748f260 1573 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
d4ead6b3
DA		 1574}
1575
cc5c073a
DA		 1576#define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1577
1578/* used when the flushed bit is not relevant, only access to the bucket
1579 * (ie., all bucket users except rt6_insert_exception);
1580 *
1581 * called under rcu lock; sometimes called with rt6_exception_lock held
1582 */
1583static
1584struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1585 spinlock_t *lock)
1586{
1587 struct rt6_exception_bucket *bucket;
1588
1589 if (lock)
1590 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1591 lockdep_is_held(lock));
1592 else
1593 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1594
1595 /* remove bucket flushed bit if set */
1596 if (bucket) {
1597 unsigned long p = (unsigned long)bucket;
1598
1599 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1600 bucket = (struct rt6_exception_bucket *)p;
1601 }
1602
1603 return bucket;
1604}
1605
1606static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1607{
1608 unsigned long p = (unsigned long)bucket;
1609
1610 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1611}
1612
1613/* called with rt6_exception_lock held */
1614static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1615 spinlock_t *lock)
1616{
1617 struct rt6_exception_bucket *bucket;
1618 unsigned long p;
1619
1620 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1621 lockdep_is_held(lock));
1622
1623 p = (unsigned long)bucket;
1624 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1625 bucket = (struct rt6_exception_bucket *)p;
1626 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1627}
1628
35732d01 1629static int rt6_insert_exception(struct rt6_info *nrt,
5012f0a5 1630 const struct fib6_result *res)
35732d01 1631{
5e670d84 1632 struct net *net = dev_net(nrt->dst.dev);
35732d01 1633 struct rt6_exception_bucket *bucket;
cc5c073a 1634 struct fib6_info *f6i = res->f6i;
35732d01
WW		 1635 struct in6_addr *src_key = NULL;
1636 struct rt6_exception *rt6_ex;
cc5c073a 1637 struct fib6_nh *nh = res->nh;
35732d01
WW		 1638 int err = 0;
1639
35732d01
WW		 1640 spin_lock_bh(&rt6_exception_lock);
1641
cc5c073a
DA		 1642 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1643 lockdep_is_held(&rt6_exception_lock));
35732d01
WW		 1644 if (!bucket) {
1645 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1646 GFP_ATOMIC);
1647 if (!bucket) {
1648 err = -ENOMEM;
1649 goto out;
1650 }
cc5c073a
DA		 1651 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1652 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1653 err = -EINVAL;
1654 goto out;
35732d01
WW		 1655 }
1656
1657#ifdef CONFIG_IPV6_SUBTREES
5012f0a5 1658 /* fib6_src.plen != 0 indicates f6i is in subtree
35732d01 1659 * and exception table is indexed by a hash of
5012f0a5 1660 * both fib6_dst and fib6_src.
35732d01 1661 * Otherwise, the exception table is indexed by
5012f0a5 1662 * a hash of only fib6_dst.
35732d01 1663 */
5012f0a5 1664 if (f6i->fib6_src.plen)
35732d01
WW		 1665 src_key = &nrt->rt6i_src.addr;
1666#endif
5012f0a5 1667 /* rt6_mtu_change() might lower mtu on f6i.
f5bbe7ee 1668 * Only insert this exception route if its mtu
5012f0a5 1669 * is less than f6i's mtu value.
f5bbe7ee 1670 */
b748f260 1671 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
f5bbe7ee
WW		 1672 err = -EINVAL;
1673 goto out;
1674 }
60006a48 1675
35732d01
WW		 1676 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1677 src_key);
1678 if (rt6_ex)
1679 rt6_remove_exception(bucket, rt6_ex);
1680
1681 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1682 if (!rt6_ex) {
1683 err = -ENOMEM;
1684 goto out;
1685 }
1686 rt6_ex->rt6i = nrt;
1687 rt6_ex->stamp = jiffies;
35732d01
WW		 1688 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1689 bucket->depth++;
81eb8447 1690 net->ipv6.rt6_stats->fib_rt_cache++;
35732d01
WW		 1691
1692 if (bucket->depth > FIB6_MAX_DEPTH)
1693 rt6_exception_remove_oldest(bucket);
1694
1695out:
1696 spin_unlock_bh(&rt6_exception_lock);
1697
1698 /* Update fn->fn_sernum to invalidate all cached dst */
b886d5f2 1699 if (!err) {
5012f0a5
DA		 1700 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1701 fib6_update_sernum(net, f6i);
1702 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
b886d5f2
PA		 1703 fib6_force_start_gc(net);
1704 }
35732d01
WW		 1705
1706 return err;
1707}
1708
c0b220cf 1709static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
35732d01
WW		 1710{
1711 struct rt6_exception_bucket *bucket;
1712 struct rt6_exception *rt6_ex;
1713 struct hlist_node *tmp;
1714 int i;
1715
1716 spin_lock_bh(&rt6_exception_lock);
35732d01 1717
cc5c073a 1718 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
35732d01
WW		 1719 if (!bucket)
1720 goto out;
1721
cc5c073a
DA		 1722 /* Prevent rt6_insert_exception() to recreate the bucket list */
1723 if (!from)
1724 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1725
35732d01 1726 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
cc5c073a
DA		 1727 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1728 if (!from ||
1729 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1730 rt6_remove_exception(bucket, rt6_ex);
1731 }
1732 WARN_ON_ONCE(!from && bucket->depth);
35732d01
WW		 1733 bucket++;
1734 }
35732d01
WW		 1735out:
1736 spin_unlock_bh(&rt6_exception_lock);
1737}
1738
e659ba31
DA		 1739static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1740{
1741 struct fib6_info *f6i = arg;
1742
1743 fib6_nh_flush_exceptions(nh, f6i);
1744
1745 return 0;
1746}
1747
c0b220cf
DA		 1748void rt6_flush_exceptions(struct fib6_info *f6i)
1749{
e659ba31
DA		 1750 if (f6i->nh)
1751 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1752 f6i);
1753 else
1754 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
c0b220cf
DA		 1755}
1756
35732d01
WW		 1757/* Find cached rt in the hash table inside passed in rt
1758 * Caller has to hold rcu_read_lock()
1759 */
7e4b5128 1760static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
510e2ced
WW		 1761 const struct in6_addr *daddr,
1762 const struct in6_addr *saddr)
35732d01 1763{
510e2ced 1764 const struct in6_addr *src_key = NULL;
35732d01 1765 struct rt6_exception_bucket *bucket;
35732d01 1766 struct rt6_exception *rt6_ex;
7e4b5128 1767 struct rt6_info *ret = NULL;
35732d01 1768
35732d01 1769#ifdef CONFIG_IPV6_SUBTREES
7e4b5128 1770 /* fib6i_src.plen != 0 indicates f6i is in subtree
35732d01 1771 * and exception table is indexed by a hash of
7e4b5128 1772 * both fib6_dst and fib6_src.
510e2ced
WW		 1773 * However, the src addr used to create the hash
1774 * might not be exactly the passed in saddr which
1775 * is a /128 addr from the flow.
1776 * So we need to use f6i->fib6_src to redo lookup
1777 * if the passed in saddr does not find anything.
1778 * (See the logic in ip6_rt_cache_alloc() on how
1779 * rt->rt6i_src is updated.)
35732d01 1780 */
7e4b5128 1781 if (res->f6i->fib6_src.plen)
35732d01 1782 src_key = saddr;
510e2ced 1783find_ex:
35732d01 1784#endif
cc5c073a 1785 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
35732d01
WW		 1786 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1787
1788 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
7e4b5128 1789 ret = rt6_ex->rt6i;
35732d01 1790
510e2ced
WW		 1791#ifdef CONFIG_IPV6_SUBTREES
1792 /* Use fib6_src as src_key and redo lookup */
1793 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1794 src_key = &res->f6i->fib6_src.addr;
1795 goto find_ex;
1796 }
1797#endif
1798
7e4b5128 1799 return ret;
35732d01
WW		 1800}
1801
1802/* Remove the passed in cached rt from the hash table that contains it */
cc5c073a 1803static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1804 const struct rt6_info *rt)
35732d01 1805{
c0b220cf 1806 const struct in6_addr *src_key = NULL;
35732d01 1807 struct rt6_exception_bucket *bucket;
35732d01
WW		 1808 struct rt6_exception *rt6_ex;
1809 int err;
1810
cc5c073a 1811 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
35732d01
WW		 1812 return -ENOENT;
1813
1814 spin_lock_bh(&rt6_exception_lock);
cc5c073a
DA		 1815 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1816
35732d01 1817#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1818 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1819 * and exception table is indexed by a hash of
1820 * both rt6i_dst and rt6i_src.
35732d01
WW		 1821 * Otherwise, the exception table is indexed by
1822 * a hash of only rt6i_dst.
1823 */
c0b220cf 1824 if (plen)
35732d01
WW		 1825 src_key = &rt->rt6i_src.addr;
1826#endif
1827 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1828 &rt->rt6i_dst.addr,
1829 src_key);
1830 if (rt6_ex) {
1831 rt6_remove_exception(bucket, rt6_ex);
1832 err = 0;
1833 } else {
1834 err = -ENOENT;
1835 }
1836
1837 spin_unlock_bh(&rt6_exception_lock);
1838 return err;
1839}
1840
e659ba31
DA		 1841struct fib6_nh_excptn_arg {
1842 struct rt6_info *rt;
1843 int plen;
1844};
1845
1846static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1847{
1848 struct fib6_nh_excptn_arg *arg = _arg;
1849 int err;
1850
1851 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1852 if (err == 0)
1853 return 1;
1854
1855 return 0;
1856}
1857
c0b220cf
DA		 1858static int rt6_remove_exception_rt(struct rt6_info *rt)
1859{
1860 struct fib6_info *from;
1861
1862 from = rcu_dereference(rt->from);
cc5c073a 1863 if (!from || !(rt->rt6i_flags & RTF_CACHE))
c0b220cf
DA		 1864 return -EINVAL;
1865
e659ba31
DA		 1866 if (from->nh) {
1867 struct fib6_nh_excptn_arg arg = {
1868 .rt = rt,
1869 .plen = from->fib6_src.plen
1870 };
1871 int rc;
1872
1873 /* rc = 1 means an entry was found */
1874 rc = nexthop_for_each_fib6_nh(from->nh,
1875 rt6_nh_remove_exception_rt,
1876 &arg);
1877 return rc ? 0 : -ENOENT;
1878 }
1879
1cf844c7 1880 return fib6_nh_remove_exception(from->fib6_nh,
cc5c073a 1881 from->fib6_src.plen, rt);
c0b220cf
DA		 1882}
1883
35732d01
WW		 1884/* Find rt6_ex which contains the passed in rt cache and
1885 * refresh its stamp
1886 */
cc5c073a 1887static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1888 const struct rt6_info *rt)
35732d01 1889{
c0b220cf 1890 const struct in6_addr *src_key = NULL;
35732d01 1891 struct rt6_exception_bucket *bucket;
35732d01 1892 struct rt6_exception *rt6_ex;
193f3685 1893
cc5c073a 1894 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
35732d01 1895#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1896 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1897 * and exception table is indexed by a hash of
1898 * both rt6i_dst and rt6i_src.
35732d01
WW		 1899 * Otherwise, the exception table is indexed by
1900 * a hash of only rt6i_dst.
1901 */
c0b220cf 1902 if (plen)
35732d01
WW		 1903 src_key = &rt->rt6i_src.addr;
1904#endif
cc5c073a 1905 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
35732d01
WW		 1906 if (rt6_ex)
1907 rt6_ex->stamp = jiffies;
c0b220cf
DA		 1908}
1909
e659ba31
DA		 1910struct fib6_nh_match_arg {
1911 const struct net_device *dev;
1912 const struct in6_addr *gw;
1913 struct fib6_nh *match;
1914};
1915
1916/* determine if fib6_nh has given device and gateway */
1917static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1918{
1919 struct fib6_nh_match_arg *arg = _arg;
1920
1921 if (arg->dev != nh->fib_nh_dev ||
1922 (arg->gw && !nh->fib_nh_gw_family) ||
1923 (!arg->gw && nh->fib_nh_gw_family) ||
1924 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1925 return 0;
1926
1927 arg->match = nh;
1928
1929 /* found a match, break the loop */
1930 return 1;
1931}
1932
c0b220cf
DA		 1933static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1934{
1935 struct fib6_info *from;
e659ba31 1936 struct fib6_nh *fib6_nh;
35732d01 1937
c0b220cf
DA		 1938 rcu_read_lock();
1939
1940 from = rcu_dereference(rt->from);
1941 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1942 goto unlock;
1943
e659ba31
DA		 1944 if (from->nh) {
1945 struct fib6_nh_match_arg arg = {
1946 .dev = rt->dst.dev,
1947 .gw = &rt->rt6i_gateway,
1948 };
1949
1950 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1951
1952 if (!arg.match)
1953 return;
1954 fib6_nh = arg.match;
1955 } else {
1956 fib6_nh = from->fib6_nh;
1957 }
1958 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
193f3685 1959unlock:
35732d01
WW		 1960 rcu_read_unlock();
1961}
1962
e9fa1495
SB		 1963static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1964 struct rt6_info *rt, int mtu)
1965{
1966 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1967 * lowest MTU in the path: always allow updating the route PMTU to
1968 * reflect PMTU decreases.
1969 *
1970 * If the new MTU is higher, and the route PMTU is equal to the local
1971 * MTU, this means the old MTU is the lowest in the path, so allow
1972 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1973 * handle this.
1974 */
1975
1976 if (dst_mtu(&rt->dst) >= mtu)
1977 return true;
1978
1979 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
1980 return true;
1981
1982 return false;
1983}
1984
1985static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
cc5c073a 1986 const struct fib6_nh *nh, int mtu)
f5bbe7ee
WW		 1987{
1988 struct rt6_exception_bucket *bucket;
1989 struct rt6_exception *rt6_ex;
1990 int i;
1991
cc5c073a 1992 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
e9fa1495
SB		 1993 if (!bucket)
1994 return;
1995
1996 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1997 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1998 struct rt6_info *entry = rt6_ex->rt6i;
1999
2000 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
d4ead6b3 2001 * route), the metrics of its rt->from have already
e9fa1495
SB		 2002 * been updated.
2003 */
d4ead6b3 2004 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
e9fa1495 2005 rt6_mtu_change_route_allowed(idev, entry, mtu))
d4ead6b3 2006 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
f5bbe7ee 2007 }
e9fa1495 2008 bucket++;
f5bbe7ee
WW		 2009 }
2010}
2011
b16cb459
WW		 2012#define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2013
cc5c073a
DA		 2014static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2015 const struct in6_addr *gateway)
b16cb459
WW		 2016{
2017 struct rt6_exception_bucket *bucket;
2018 struct rt6_exception *rt6_ex;
2019 struct hlist_node *tmp;
2020 int i;
2021
cc5c073a 2022 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
b16cb459
WW		 2023 return;
2024
2025 spin_lock_bh(&rt6_exception_lock);
cc5c073a 2026 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
b16cb459
WW		 2027 if (bucket) {
2028 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2029 hlist_for_each_entry_safe(rt6_ex, tmp,
2030 &bucket->chain, hlist) {
2031 struct rt6_info *entry = rt6_ex->rt6i;
2032
2033 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2034 RTF_CACHE_GATEWAY &&
2035 ipv6_addr_equal(gateway,
2036 &entry->rt6i_gateway)) {
2037 rt6_remove_exception(bucket, rt6_ex);
2038 }
2039 }
2040 bucket++;
2041 }
2042 }
2043
2044 spin_unlock_bh(&rt6_exception_lock);
2045}
2046
c757faa8
WW		 2047static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2048 struct rt6_exception *rt6_ex,
2049 struct fib6_gc_args *gc_args,
2050 unsigned long now)
2051{
2052 struct rt6_info *rt = rt6_ex->rt6i;
2053
1859bac0
PA		 2054 /* we are pruning and obsoleting aged-out and non gateway exceptions
2055 * even if others have still references to them, so that on next
2056 * dst_check() such references can be dropped.
2057 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2058 * expired, independently from their aging, as per RFC 8201 section 4
2059 */
31afeb42
WW		 2060 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2061 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2062 RT6_TRACE("aging clone %p\n", rt);
2063 rt6_remove_exception(bucket, rt6_ex);
2064 return;
2065 }
2066 } else if (time_after(jiffies, rt->dst.expires)) {
2067 RT6_TRACE("purging expired route %p\n", rt);
c757faa8
WW		 2068 rt6_remove_exception(bucket, rt6_ex);
2069 return;
31afeb42
WW		 2070 }
2071
2072 if (rt->rt6i_flags & RTF_GATEWAY) {
c757faa8
WW		 2073 struct neighbour *neigh;
2074 __u8 neigh_flags = 0;
2075
1bfa26ff
ED		 2076 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2077 if (neigh)
c757faa8 2078 neigh_flags = neigh->flags;
1bfa26ff 2079
c757faa8
WW		 2080 if (!(neigh_flags & NTF_ROUTER)) {
2081 RT6_TRACE("purging route %p via non-router but gateway\n",
2082 rt);
2083 rt6_remove_exception(bucket, rt6_ex);
2084 return;
2085 }
2086 }
31afeb42 2087
c757faa8
WW		 2088 gc_args->more++;
2089}
2090
cc5c073a 2091static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
c0b220cf
DA		 2092 struct fib6_gc_args *gc_args,
2093 unsigned long now)
c757faa8
WW		 2094{
2095 struct rt6_exception_bucket *bucket;
2096 struct rt6_exception *rt6_ex;
2097 struct hlist_node *tmp;
2098 int i;
2099
cc5c073a 2100 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
c757faa8
WW		 2101 return;
2102
1bfa26ff
ED		 2103 rcu_read_lock_bh();
2104 spin_lock(&rt6_exception_lock);
cc5c073a 2105 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
c757faa8
WW		 2106 if (bucket) {
2107 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2108 hlist_for_each_entry_safe(rt6_ex, tmp,
2109 &bucket->chain, hlist) {
2110 rt6_age_examine_exception(bucket, rt6_ex,
2111 gc_args, now);
2112 }
2113 bucket++;
2114 }
2115 }
1bfa26ff
ED		 2116 spin_unlock(&rt6_exception_lock);
2117 rcu_read_unlock_bh();
c757faa8
WW		 2118}
2119
e659ba31
DA		 2120struct fib6_nh_age_excptn_arg {
2121 struct fib6_gc_args *gc_args;
2122 unsigned long now;
2123};
2124
2125static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2126{
2127 struct fib6_nh_age_excptn_arg *arg = _arg;
2128
2129 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2130 return 0;
2131}
2132
cc5c073a 2133void rt6_age_exceptions(struct fib6_info *f6i,
c0b220cf
DA		 2134 struct fib6_gc_args *gc_args,
2135 unsigned long now)
2136{
e659ba31
DA		 2137 if (f6i->nh) {
2138 struct fib6_nh_age_excptn_arg arg = {
2139 .gc_args = gc_args,
2140 .now = now
2141 };
2142
2143 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2144 &arg);
2145 } else {
2146 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2147 }
c0b220cf
DA		 2148}
2149
1d053da9 2150/* must be called with rcu lock held */
effda4dd
DA		 2151int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2152 struct flowi6 *fl6, struct fib6_result *res, int strict)
1da177e4 2153{
367efcb9 2154 struct fib6_node *fn, *saved_fn;
1da177e4 2155
6454743b 2156 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
367efcb9 2157 saved_fn = fn;
1da177e4 2158
ca254490
DA		 2159 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2160 oif = 0;
2161
a3c00e46 2162redo_rt6_select:
effda4dd
DA		 2163 rt6_select(net, fn, oif, res, strict);
2164 if (res->f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2165 fn = fib6_backtrack(fn, &fl6->saddr);
2166 if (fn)
2167 goto redo_rt6_select;
367efcb9
MKL		 2168 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2169 /* also consider unreachable route */
2170 strict &= ~RT6_LOOKUP_F_REACHABLE;
2171 fn = saved_fn;
2172 goto redo_rt6_select;
367efcb9 2173 }
a3c00e46
MKL		 2174 }
2175
effda4dd 2176 trace_fib6_table_lookup(net, res, table, fl6);
fb9de91e 2177
effda4dd 2178 return 0;
1d053da9
DA		 2179}
2180
2181struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2182 int oif, struct flowi6 *fl6,
2183 const struct sk_buff *skb, int flags)
2184{
b1d40991 2185 struct fib6_result res = {};
0e09edcc 2186 struct rt6_info *rt = NULL;
1d053da9
DA		 2187 int strict = 0;
2188
0e09edcc
WW		 2189 WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2190 !rcu_read_lock_held());
2191
1d053da9
DA		 2192 strict |= flags & RT6_LOOKUP_F_IFACE;
2193 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2194 if (net->ipv6.devconf_all->forwarding == 0)
2195 strict |= RT6_LOOKUP_F_REACHABLE;
2196
2197 rcu_read_lock();
2198
effda4dd 2199 fib6_table_lookup(net, table, oif, fl6, &res, strict);
0e09edcc
WW		 2200 if (res.f6i == net->ipv6.fib6_null_entry)
2201 goto out;
23fb93a4 2202
b1d40991 2203 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
d83009d4 2204
23fb93a4 2205 /*Search through exception table */
7e4b5128 2206 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
23fb93a4 2207 if (rt) {
0e09edcc 2208 goto out;
3da59bd9 2209 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
b1d40991 2210 !res.nh->fib_nh_gw_family)) {
3da59bd9
MKL		 2211 /* Create a RTF_CACHE clone which will not be
2212 * owned by the fib6 tree. It is for the special case where
2213 * the daddr in the skb during the neighbor look-up is different
2214 * from the fl6->daddr used to look-up route here.
2215 */
0e09edcc 2216 rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
c71099ac 2217
0e09edcc
WW		 2218 if (rt) {
2219 /* 1 refcnt is taken during ip6_rt_cache_alloc().
2220 * As rt6_uncached_list_add() does not consume refcnt,
2221 * this refcnt is always returned to the caller even
2222 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
1cfb71ee 2223 */
0e09edcc 2224 rt6_uncached_list_add(rt);
81eb8447 2225 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
0e09edcc 2226 rcu_read_unlock();
b811580d 2227
0e09edcc
WW		 2228 return rt;
2229 }
d52d3997
MKL		 2230 } else {
2231 /* Get a percpu copy */
951f788a 2232 local_bh_disable();
0e09edcc 2233 rt = rt6_get_pcpu_route(&res);
d52d3997 2234
0e09edcc
WW		 2235 if (!rt)
2236 rt = rt6_make_pcpu_route(net, &res);
93531c67 2237
951f788a 2238 local_bh_enable();
d52d3997 2239 }
0e09edcc
WW		 2240out:
2241 if (!rt)
2242 rt = net->ipv6.ip6_null_entry;
2243 if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2244 ip6_hold_safe(net, &rt);
2245 rcu_read_unlock();
2246
2247 return rt;
1da177e4 2248}
9ff74384 2249EXPORT_SYMBOL_GPL(ip6_pol_route);
1da177e4 2250
b75cc8f9
DA		 2251static struct rt6_info *ip6_pol_route_input(struct net *net,
2252 struct fib6_table *table,
2253 struct flowi6 *fl6,
2254 const struct sk_buff *skb,
2255 int flags)
4acad72d 2256{
b75cc8f9 2257 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
4acad72d
PE		 2258}
2259
d409b847
MB		 2260struct dst_entry *ip6_route_input_lookup(struct net *net,
2261 struct net_device *dev,
b75cc8f9
DA		 2262 struct flowi6 *fl6,
2263 const struct sk_buff *skb,
2264 int flags)
72331bc0
SL		 2265{
2266 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2267 flags |= RT6_LOOKUP_F_IFACE;
2268
b75cc8f9 2269 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
72331bc0 2270}
d409b847 2271EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
72331bc0 2272
23aebdac 2273static void ip6_multipath_l3_keys(const struct sk_buff *skb,
5e5d6fed
RP		 2274 struct flow_keys *keys,
2275 struct flow_keys *flkeys)
23aebdac
JS		 2276{
2277 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2278 const struct ipv6hdr *key_iph = outer_iph;
5e5d6fed 2279 struct flow_keys *_flkeys = flkeys;
23aebdac
JS		 2280 const struct ipv6hdr *inner_iph;
2281 const struct icmp6hdr *icmph;
2282 struct ipv6hdr _inner_iph;
cea67a2d 2283 struct icmp6hdr _icmph;
23aebdac
JS		 2284
2285 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2286 goto out;
2287
cea67a2d
ED		 2288 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2289 sizeof(_icmph), &_icmph);
2290 if (!icmph)
2291 goto out;
2292
23aebdac
JS		 2293 if (icmph->icmp6_type != ICMPV6_DEST_UNREACH &&
2294 icmph->icmp6_type != ICMPV6_PKT_TOOBIG &&
2295 icmph->icmp6_type != ICMPV6_TIME_EXCEED &&
2296 icmph->icmp6_type != ICMPV6_PARAMPROB)
2297 goto out;
2298
2299 inner_iph = skb_header_pointer(skb,
2300 skb_transport_offset(skb) + sizeof(*icmph),
2301 sizeof(_inner_iph), &_inner_iph);
2302 if (!inner_iph)
2303 goto out;
2304
2305 key_iph = inner_iph;
5e5d6fed 2306 _flkeys = NULL;
23aebdac 2307out:
5e5d6fed
RP		 2308 if (_flkeys) {
2309 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2310 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2311 keys->tags.flow_label = _flkeys->tags.flow_label;
2312 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2313 } else {
2314 keys->addrs.v6addrs.src = key_iph->saddr;
2315 keys->addrs.v6addrs.dst = key_iph->daddr;
fa1be7e0 2316 keys->tags.flow_label = ip6_flowlabel(key_iph);
5e5d6fed
RP		 2317 keys->basic.ip_proto = key_iph->nexthdr;
2318 }
23aebdac
JS		 2319}
2320
2321/* if skb is set it will be used and fl6 can be NULL */
b4bac172
DA		 2322u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2323 const struct sk_buff *skb, struct flow_keys *flkeys)
23aebdac
JS		 2324{
2325 struct flow_keys hash_keys;
9a2a537a 2326 u32 mhash;
23aebdac 2327
bbfa047a 2328 switch (ip6_multipath_hash_policy(net)) {
b4bac172
DA		 2329 case 0:
2330 memset(&hash_keys, 0, sizeof(hash_keys));
2331 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2332 if (skb) {
2333 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2334 } else {
2335 hash_keys.addrs.v6addrs.src = fl6->saddr;
2336 hash_keys.addrs.v6addrs.dst = fl6->daddr;
fa1be7e0 2337 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
b4bac172
DA		 2338 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2339 }
2340 break;
2341 case 1:
2342 if (skb) {
2343 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2344 struct flow_keys keys;
2345
2346 /* short-circuit if we already have L4 hash present */
2347 if (skb->l4_hash)
2348 return skb_get_hash_raw(skb) >> 1;
2349
2350 memset(&hash_keys, 0, sizeof(hash_keys));
2351
2352 if (!flkeys) {
2353 skb_flow_dissect_flow_keys(skb, &keys, flag);
2354 flkeys = &keys;
2355 }
2356 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2357 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2358 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2359 hash_keys.ports.src = flkeys->ports.src;
2360 hash_keys.ports.dst = flkeys->ports.dst;
2361 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2362 } else {
2363 memset(&hash_keys, 0, sizeof(hash_keys));
2364 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2365 hash_keys.addrs.v6addrs.src = fl6->saddr;
2366 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2367 hash_keys.ports.src = fl6->fl6_sport;
2368 hash_keys.ports.dst = fl6->fl6_dport;
2369 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2370 }
2371 break;
23aebdac 2372 }
9a2a537a 2373 mhash = flow_hash_from_keys(&hash_keys);
23aebdac 2374
9a2a537a 2375 return mhash >> 1;
23aebdac
JS		 2376}
2377
67f415dd 2378/* Called with rcu held */
c71099ac
TG		 2379void ip6_route_input(struct sk_buff *skb)
2380{
b71d1d42 2381 const struct ipv6hdr *iph = ipv6_hdr(skb);
c346dca1 2382 struct net *net = dev_net(skb->dev);
67f415dd 2383 int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
904af04d 2384 struct ip_tunnel_info *tun_info;
4c9483b2 2385 struct flowi6 fl6 = {
e0d56fdd 2386 .flowi6_iif = skb->dev->ifindex,
4c9483b2
DM		 2387 .daddr = iph->daddr,
2388 .saddr = iph->saddr,
6502ca52 2389 .flowlabel = ip6_flowinfo(iph),
4c9483b2
DM		 2390 .flowi6_mark = skb->mark,
2391 .flowi6_proto = iph->nexthdr,
c71099ac 2392 };
5e5d6fed 2393 struct flow_keys *flkeys = NULL, _flkeys;
adaa70bb 2394
904af04d 2395 tun_info = skb_tunnel_info(skb);
46fa062a 2396 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
904af04d 2397 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
5e5d6fed
RP		 2398
2399 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2400 flkeys = &_flkeys;
2401
23aebdac 2402 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
b4bac172 2403 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
06e9d040 2404 skb_dst_drop(skb);
67f415dd
WW		 2405 skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2406 &fl6, skb, flags));
c71099ac
TG		 2407}
2408
b75cc8f9
DA		 2409static struct rt6_info *ip6_pol_route_output(struct net *net,
2410 struct fib6_table *table,
2411 struct flowi6 *fl6,
2412 const struct sk_buff *skb,
2413 int flags)
1da177e4 2414{
b75cc8f9 2415 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
c71099ac
TG		 2416}
2417
6f21c96a
PA		 2418struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
2419 struct flowi6 *fl6, int flags)
c71099ac 2420{
d46a9d67 2421 bool any_src;
c71099ac 2422
3ede0bbc
RS		 2423 if (ipv6_addr_type(&fl6->daddr) &
2424 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
4c1feac5
DA		 2425 struct dst_entry *dst;
2426
2427 dst = l3mdev_link_scope_lookup(net, fl6);
2428 if (dst)
2429 return dst;
2430 }
ca254490 2431
1fb9489b 2432 fl6->flowi6_iif = LOOPBACK_IFINDEX;
4dc27d1c 2433
d46a9d67 2434 any_src = ipv6_addr_any(&fl6->saddr);
741a11d9 2435 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
d46a9d67 2436 (fl6->flowi6_oif && any_src))
77d16f45 2437 flags |= RT6_LOOKUP_F_IFACE;
c71099ac 2438
d46a9d67 2439 if (!any_src)
adaa70bb 2440 flags |= RT6_LOOKUP_F_HAS_SADDR;
0c9a2ac1
YH		 2441 else if (sk)
2442 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
adaa70bb 2443
b75cc8f9 2444 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
1da177e4 2445}
6f21c96a 2446EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1da177e4 2447
2774c131 2448struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
14e50e57 2449{
5c1e6aa3 2450 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1dbe3252 2451 struct net_device *loopback_dev = net->loopback_dev;
14e50e57
DM		 2452 struct dst_entry *new = NULL;
2453
1dbe3252 2454 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
62cf27e5 2455 DST_OBSOLETE_DEAD, 0);
14e50e57 2456 if (rt) {
0a1f5962 2457 rt6_info_init(rt);
81eb8447 2458 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
8104891b 2459
0a1f5962 2460 new = &rt->dst;
14e50e57 2461 new->__use = 1;
352e512c 2462 new->input = dst_discard;
ede2059d 2463 new->output = dst_discard_out;
14e50e57 2464
0a1f5962 2465 dst_copy_metrics(new, &ort->dst);
14e50e57 2466
1dbe3252 2467 rt->rt6i_idev = in6_dev_get(loopback_dev);
4e3fd7a0 2468 rt->rt6i_gateway = ort->rt6i_gateway;
0a1f5962 2469 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
14e50e57
DM		 2470
2471 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2472#ifdef CONFIG_IPV6_SUBTREES
2473 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2474#endif
14e50e57
DM		 2475 }
2476
69ead7af
DM		 2477 dst_release(dst_orig);
2478 return new ? new : ERR_PTR(-ENOMEM);
14e50e57 2479}
14e50e57 2480
1da177e4
LT		 2481/*
2482 * Destination cache support functions
2483 */
2484
8d1c802b 2485static bool fib6_check(struct fib6_info *f6i, u32 cookie)
4b32b5ad 2486{
93531c67
DA		 2487 u32 rt_cookie = 0;
2488
8ae86971 2489 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
93531c67
DA		 2490 return false;
2491
2492 if (fib6_check_expired(f6i))
2493 return false;
2494
2495 return true;
4b32b5ad
MKL		 2496}
2497
a68886a6
DA		 2498static struct dst_entry *rt6_check(struct rt6_info *rt,
2499 struct fib6_info *from,
2500 u32 cookie)
3da59bd9 2501{
36143645 2502 u32 rt_cookie = 0;
c5cff856 2503
a68886a6 2504 if ((from && !fib6_get_cookie_safe(from, &rt_cookie)) ||
93531c67 2505 rt_cookie != cookie)
3da59bd9
MKL		 2506 return NULL;
2507
2508 if (rt6_check_expired(rt))
2509 return NULL;
2510
2511 return &rt->dst;
2512}
2513
a68886a6
DA		 2514static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2515 struct fib6_info *from,
2516 u32 cookie)
3da59bd9 2517{
5973fb1e
MKL		 2518 if (!__rt6_check_expired(rt) &&
2519 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
a68886a6 2520 fib6_check(from, cookie))
3da59bd9
MKL		 2521 return &rt->dst;
2522 else
2523 return NULL;
2524}
2525
1da177e4
LT		 2526static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
2527{
a87b7dc9 2528 struct dst_entry *dst_ret;
a68886a6 2529 struct fib6_info *from;
1da177e4
LT		 2530 struct rt6_info *rt;
2531
a87b7dc9
DA		 2532 rt = container_of(dst, struct rt6_info, dst);
2533
2534 rcu_read_lock();
1da177e4 2535
6f3118b5
ND		 2536 /* All IPV6 dsts are created with ->obsolete set to the value
2537 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2538 * into this function always.
2539 */
e3bc10bd 2540
a68886a6 2541 from = rcu_dereference(rt->from);
4b32b5ad 2542
a68886a6
DA		 2543 if (from && (rt->rt6i_flags & RTF_PCPU ||
2544 unlikely(!list_empty(&rt->rt6i_uncached))))
2545 dst_ret = rt6_dst_from_check(rt, from, cookie);
3da59bd9 2546 else
a68886a6 2547 dst_ret = rt6_check(rt, from, cookie);
a87b7dc9
DA		 2548
2549 rcu_read_unlock();
2550
2551 return dst_ret;
1da177e4
LT		 2552}
2553
2554static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2555{
2556 struct rt6_info *rt = (struct rt6_info *) dst;
2557
2558 if (rt) {
54c1a859 2559 if (rt->rt6i_flags & RTF_CACHE) {
c3c14da0 2560 rcu_read_lock();
54c1a859 2561 if (rt6_check_expired(rt)) {
93531c67 2562 rt6_remove_exception_rt(rt);
54c1a859
YH		 2563 dst = NULL;
2564 }
c3c14da0 2565 rcu_read_unlock();
54c1a859 2566 } else {
1da177e4 2567 dst_release(dst);
54c1a859
YH		 2568 dst = NULL;
2569 }
1da177e4 2570 }
54c1a859 2571 return dst;
1da177e4
LT		 2572}
2573
2574static void ip6_link_failure(struct sk_buff *skb)
2575{
2576 struct rt6_info *rt;
2577
3ffe533c 2578 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1da177e4 2579
adf30907 2580 rt = (struct rt6_info *) skb_dst(skb);
1da177e4 2581 if (rt) {
8a14e46f 2582 rcu_read_lock();
1eb4f758 2583 if (rt->rt6i_flags & RTF_CACHE) {
761f6026 2584 rt6_remove_exception_rt(rt);
c5cff856 2585 } else {
a68886a6 2586 struct fib6_info *from;
c5cff856
WW		 2587 struct fib6_node *fn;
2588
a68886a6
DA		 2589 from = rcu_dereference(rt->from);
2590 if (from) {
2591 fn = rcu_dereference(from->fib6_node);
2592 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2593 fn->fn_sernum = -1;
2594 }
1eb4f758 2595 }
8a14e46f 2596 rcu_read_unlock();
1da177e4
LT		 2597 }
2598}
2599
6a3e030f
DA		 2600static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2601{
a68886a6
DA		 2602 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2603 struct fib6_info *from;
2604
2605 rcu_read_lock();
2606 from = rcu_dereference(rt0->from);
2607 if (from)
2608 rt0->dst.expires = from->expires;
2609 rcu_read_unlock();
2610 }
6a3e030f
DA		 2611
2612 dst_set_expires(&rt0->dst, timeout);
2613 rt0->rt6i_flags |= RTF_EXPIRES;
2614}
2615
45e4fd26
MKL		 2616static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2617{
2618 struct net *net = dev_net(rt->dst.dev);
2619
d4ead6b3 2620 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
45e4fd26 2621 rt->rt6i_flags |= RTF_MODIFIED;
45e4fd26
MKL		 2622 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2623}
2624
0d3f6d29
MKL		 2625static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2626{
2627 return !(rt->rt6i_flags & RTF_CACHE) &&
1490ed2a 2628 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
0d3f6d29
MKL		 2629}
2630
45e4fd26
MKL		 2631static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2632 const struct ipv6hdr *iph, u32 mtu)
1da177e4 2633{
0dec879f 2634 const struct in6_addr *daddr, *saddr;
67ba4152 2635 struct rt6_info *rt6 = (struct rt6_info *)dst;
1da177e4 2636
19bda36c
XL		 2637 if (dst_metric_locked(dst, RTAX_MTU))
2638 return;
2639
0dec879f
JA		 2640 if (iph) {
2641 daddr = &iph->daddr;
2642 saddr = &iph->saddr;
2643 } else if (sk) {
2644 daddr = &sk->sk_v6_daddr;
2645 saddr = &inet6_sk(sk)->saddr;
2646 } else {
2647 daddr = NULL;
2648 saddr = NULL;
2649 }
2650 dst_confirm_neigh(dst, daddr);
45e4fd26
MKL		 2651 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
2652 if (mtu >= dst_mtu(dst))
2653 return;
9d289715 2654
0d3f6d29 2655 if (!rt6_cache_allowed_for_pmtu(rt6)) {
45e4fd26 2656 rt6_do_update_pmtu(rt6, mtu);
2b760fcf
WW		 2657 /* update rt6_ex->stamp for cache */
2658 if (rt6->rt6i_flags & RTF_CACHE)
2659 rt6_update_exception_stamp_rt(rt6);
0dec879f 2660 } else if (daddr) {
85bd05de 2661 struct fib6_result res = {};
45e4fd26
MKL		 2662 struct rt6_info *nrt6;
2663
4d85cd0c 2664 rcu_read_lock();
85bd05de
DA		 2665 res.f6i = rcu_dereference(rt6->from);
2666 if (!res.f6i) {
9c69a132
JL		 2667 rcu_read_unlock();
2668 return;
2669 }
7d21fec9
DA		 2670 res.fib6_flags = res.f6i->fib6_flags;
2671 res.fib6_type = res.f6i->fib6_type;
2672
2d44234b
DA		 2673 if (res.f6i->nh) {
2674 struct fib6_nh_match_arg arg = {
2675 .dev = dst->dev,
2676 .gw = &rt6->rt6i_gateway,
2677 };
2678
2679 nexthop_for_each_fib6_nh(res.f6i->nh,
2680 fib6_nh_find_match, &arg);
2681
2682 /* fib6_info uses a nexthop that does not have fib6_nh
2683 * using the dst->dev + gw. Should be impossible.
2684 */
2685 if (!arg.match) {
2686 rcu_read_unlock();
2687 return;
2688 }
2689
2690 res.nh = arg.match;
2691 } else {
2692 res.nh = res.f6i->fib6_nh;
2693 }
2694
85bd05de 2695 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
45e4fd26
MKL		 2696 if (nrt6) {
2697 rt6_do_update_pmtu(nrt6, mtu);
5012f0a5 2698 if (rt6_insert_exception(nrt6, &res))
2b760fcf 2699 dst_release_immediate(&nrt6->dst);
45e4fd26 2700 }
a68886a6 2701 rcu_read_unlock();
1da177e4
LT		 2702 }
2703}
2704
45e4fd26
MKL		 2705static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2706 struct sk_buff *skb, u32 mtu)
2707{
2708 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
2709}
2710
42ae66c8 2711void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
e2d118a1 2712 int oif, u32 mark, kuid_t uid)
81aded24
DM		 2713{
2714 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2715 struct dst_entry *dst;
dc92095d
MŻ		 2716 struct flowi6 fl6 = {
2717 .flowi6_oif = oif,
2718 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2719 .daddr = iph->daddr,
2720 .saddr = iph->saddr,
2721 .flowlabel = ip6_flowinfo(iph),
2722 .flowi6_uid = uid,
2723 };
81aded24
DM		 2724
2725 dst = ip6_route_output(net, NULL, &fl6);
2726 if (!dst->error)
45e4fd26 2727 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
81aded24
DM		 2728 dst_release(dst);
2729}
2730EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2731
2732void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2733{
7ddacfa5 2734 int oif = sk->sk_bound_dev_if;
33c162a9
MKL		 2735 struct dst_entry *dst;
2736
7ddacfa5
DA		 2737 if (!oif && skb->dev)
2738 oif = l3mdev_master_ifindex(skb->dev);
2739
2740 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
33c162a9
MKL		 2741
2742 dst = __sk_dst_get(sk);
2743 if (!dst || !dst->obsolete ||
2744 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2745 return;
2746
2747 bh_lock_sock(sk);
2748 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2749 ip6_datagram_dst_update(sk, false);
2750 bh_unlock_sock(sk);
81aded24
DM		 2751}
2752EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2753
7d6850f7
AK		 2754void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2755 const struct flowi6 *fl6)
2756{
2757#ifdef CONFIG_IPV6_SUBTREES
2758 struct ipv6_pinfo *np = inet6_sk(sk);
2759#endif
2760
2761 ip6_dst_store(sk, dst,
2762 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2763 &sk->sk_v6_daddr : NULL,
2764#ifdef CONFIG_IPV6_SUBTREES
2765 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2766 &np->saddr :
2767#endif
2768 NULL);
2769}
2770
9b6b35ab 2771static bool ip6_redirect_nh_match(const struct fib6_result *res,
0b34eb00
DA		 2772 struct flowi6 *fl6,
2773 const struct in6_addr *gw,
2774 struct rt6_info **ret)
2775{
9b6b35ab
DA		 2776 const struct fib6_nh *nh = res->nh;
2777
0b34eb00
DA		 2778 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2779 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2780 return false;
2781
2782 /* rt_cache's gateway might be different from its 'parent'
2783 * in the case of an ip redirect.
2784 * So we keep searching in the exception table if the gateway
2785 * is different.
2786 */
2787 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
2788 struct rt6_info *rt_cache;
2789
9b6b35ab 2790 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
0b34eb00
DA		 2791 if (rt_cache &&
2792 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
2793 *ret = rt_cache;
2794 return true;
2795 }
2796 return false;
2797 }
2798 return true;
2799}
2800
c55c8988
DA		 2801struct fib6_nh_rd_arg {
2802 struct fib6_result *res;
2803 struct flowi6 *fl6;
2804 const struct in6_addr *gw;
2805 struct rt6_info **ret;
2806};
2807
2808static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
2809{
2810 struct fib6_nh_rd_arg *arg = _arg;
2811
2812 arg->res->nh = nh;
2813 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
2814}
2815
b55b76b2
DJ		 2816/* Handle redirects */
2817struct ip6rd_flowi {
2818 struct flowi6 fl6;
2819 struct in6_addr gateway;
2820};
2821
2822static struct rt6_info *__ip6_route_redirect(struct net *net,
2823 struct fib6_table *table,
2824 struct flowi6 *fl6,
b75cc8f9 2825 const struct sk_buff *skb,
b55b76b2
DJ		 2826 int flags)
2827{
2828 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
0b34eb00 2829 struct rt6_info *ret = NULL;
9b6b35ab 2830 struct fib6_result res = {};
c55c8988
DA		 2831 struct fib6_nh_rd_arg arg = {
2832 .res = &res,
2833 .fl6 = fl6,
2834 .gw = &rdfl->gateway,
2835 .ret = &ret
2836 };
8d1c802b 2837 struct fib6_info *rt;
b55b76b2
DJ		 2838 struct fib6_node *fn;
2839
31680ac2
DA		 2840 /* l3mdev_update_flow overrides oif if the device is enslaved; in
2841 * this case we must match on the real ingress device, so reset it
2842 */
2843 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2844 fl6->flowi6_oif = skb->dev->ifindex;
2845
b55b76b2 2846 /* Get the "current" route for this destination and
67c408cf 2847 * check if the redirect has come from appropriate router.
b55b76b2
DJ		 2848 *
2849 * RFC 4861 specifies that redirects should only be
2850 * accepted if they come from the nexthop to the target.
2851 * Due to the way the routes are chosen, this notion
2852 * is a bit fuzzy and one might need to check all possible
2853 * routes.
2854 */
2855
66f5d6ce 2856 rcu_read_lock();
6454743b 2857 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
b55b76b2 2858restart:
66f5d6ce 2859 for_each_fib6_node_rt_rcu(fn) {
9b6b35ab 2860 res.f6i = rt;
14895687 2861 if (fib6_check_expired(rt))
b55b76b2 2862 continue;
93c2fb25 2863 if (rt->fib6_flags & RTF_REJECT)
b55b76b2 2864 break;
c55c8988
DA		 2865 if (unlikely(rt->nh)) {
2866 if (nexthop_is_blackhole(rt->nh))
2867 continue;
2868 /* on match, res->nh is filled in and potentially ret */
2869 if (nexthop_for_each_fib6_nh(rt->nh,
2870 fib6_nh_redirect_match,
2871 &arg))
2872 goto out;
2873 } else {
2874 res.nh = rt->fib6_nh;
2875 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
2876 &ret))
2877 goto out;
2878 }
b55b76b2
DJ		 2879 }
2880
2881 if (!rt)
421842ed 2882 rt = net->ipv6.fib6_null_entry;
93c2fb25 2883 else if (rt->fib6_flags & RTF_REJECT) {
23fb93a4 2884 ret = net->ipv6.ip6_null_entry;
b0a1ba59
MKL		 2885 goto out;
2886 }
2887
421842ed 2888 if (rt == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2889 fn = fib6_backtrack(fn, &fl6->saddr);
2890 if (fn)
2891 goto restart;
b55b76b2 2892 }
a3c00e46 2893
9b6b35ab 2894 res.f6i = rt;
1cf844c7 2895 res.nh = rt->fib6_nh;
b0a1ba59 2896out:
7d21fec9 2897 if (ret) {
10585b43 2898 ip6_hold_safe(net, &ret);
7d21fec9
DA		 2899 } else {
2900 res.fib6_flags = res.f6i->fib6_flags;
2901 res.fib6_type = res.f6i->fib6_type;
9b6b35ab 2902 ret = ip6_create_rt_rcu(&res);
7d21fec9 2903 }
b55b76b2 2904
66f5d6ce 2905 rcu_read_unlock();
b55b76b2 2906
8ff2e5b2 2907 trace_fib6_table_lookup(net, &res, table, fl6);
23fb93a4 2908 return ret;
b55b76b2
DJ		 2909};
2910
2911static struct dst_entry *ip6_route_redirect(struct net *net,
b75cc8f9
DA		 2912 const struct flowi6 *fl6,
2913 const struct sk_buff *skb,
2914 const struct in6_addr *gateway)
b55b76b2
DJ		 2915{
2916 int flags = RT6_LOOKUP_F_HAS_SADDR;
2917 struct ip6rd_flowi rdfl;
2918
2919 rdfl.fl6 = *fl6;
2920 rdfl.gateway = *gateway;
2921
b75cc8f9 2922 return fib6_rule_lookup(net, &rdfl.fl6, skb,
b55b76b2
DJ		 2923 flags, __ip6_route_redirect);
2924}
2925
e2d118a1
LC		 2926void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
2927 kuid_t uid)
3a5ad2ee
DM		 2928{
2929 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2930 struct dst_entry *dst;
1f7f10ac
MŻ		 2931 struct flowi6 fl6 = {
2932 .flowi6_iif = LOOPBACK_IFINDEX,
2933 .flowi6_oif = oif,
2934 .flowi6_mark = mark,
2935 .daddr = iph->daddr,
2936 .saddr = iph->saddr,
2937 .flowlabel = ip6_flowinfo(iph),
2938 .flowi6_uid = uid,
2939 };
3a5ad2ee 2940
b75cc8f9 2941 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
b55b76b2 2942 rt6_do_redirect(dst, NULL, skb);
3a5ad2ee
DM		 2943 dst_release(dst);
2944}
2945EXPORT_SYMBOL_GPL(ip6_redirect);
2946
d456336d 2947void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
c92a59ec
DJ		 2948{
2949 const struct ipv6hdr *iph = ipv6_hdr(skb);
2950 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
2951 struct dst_entry *dst;
0b26fb17
MŻ		 2952 struct flowi6 fl6 = {
2953 .flowi6_iif = LOOPBACK_IFINDEX,
2954 .flowi6_oif = oif,
0b26fb17
MŻ		 2955 .daddr = msg->dest,
2956 .saddr = iph->daddr,
2957 .flowi6_uid = sock_net_uid(net, NULL),
2958 };
c92a59ec 2959
b75cc8f9 2960 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
b55b76b2 2961 rt6_do_redirect(dst, NULL, skb);
c92a59ec
DJ		 2962 dst_release(dst);
2963}
2964
3a5ad2ee
DM		 2965void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
2966{
e2d118a1
LC		 2967 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
2968 sk->sk_uid);
3a5ad2ee
DM		 2969}
2970EXPORT_SYMBOL_GPL(ip6_sk_redirect);
2971
0dbaee3b 2972static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1da177e4 2973{
0dbaee3b
DM		 2974 struct net_device *dev = dst->dev;
2975 unsigned int mtu = dst_mtu(dst);
2976 struct net *net = dev_net(dev);
2977
1da177e4
LT		 2978 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
2979
5578689a
DL		 2980 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
2981 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1da177e4
LT		 2982
2983 /*
1ab1457c
YH		 2984 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
2985 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
2986 * IPV6_MAXPLEN is also valid and means: "any MSS,
1da177e4
LT		 2987 * rely only on pmtu discovery"
2988 */
2989 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
2990 mtu = IPV6_MAXPLEN;
2991 return mtu;
2992}
2993
ebb762f2 2994static unsigned int ip6_mtu(const struct dst_entry *dst)
d33e4553 2995{
d33e4553 2996 struct inet6_dev *idev;
d4ead6b3 2997 unsigned int mtu;
4b32b5ad
MKL		 2998
2999 mtu = dst_metric_raw(dst, RTAX_MTU);
618f9bc7 3000 if (mtu)
30f78d8e 3001 goto out;
618f9bc7
SK		 3002
3003 mtu = IPV6_MIN_MTU;
d33e4553
DM		 3004
3005 rcu_read_lock();
3006 idev = __in6_dev_get(dst->dev);
3007 if (idev)
3008 mtu = idev->cnf.mtu6;
3009 rcu_read_unlock();
3010
30f78d8e 3011out:
14972cbd
RP		 3012 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3013
3014 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
d33e4553
DM		 3015}
3016
901731b8
DA		 3017/* MTU selection:
3018 * 1. mtu on route is locked - use it
3019 * 2. mtu from nexthop exception
3020 * 3. mtu from egress device
3021 *
3022 * based on ip6_dst_mtu_forward and exception logic of
3023 * rt6_find_cached_rt; called with rcu_read_lock
3024 */
b748f260
DA		 3025u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3026 const struct in6_addr *daddr,
3027 const struct in6_addr *saddr)
901731b8 3028{
b748f260
DA		 3029 const struct fib6_nh *nh = res->nh;
3030 struct fib6_info *f6i = res->f6i;
901731b8 3031 struct inet6_dev *idev;
510e2ced 3032 struct rt6_info *rt;
901731b8
DA		 3033 u32 mtu = 0;
3034
3035 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3036 mtu = f6i->fib6_pmtu;
3037 if (mtu)
3038 goto out;
3039 }
3040
510e2ced
WW		 3041 rt = rt6_find_cached_rt(res, daddr, saddr);
3042 if (unlikely(rt)) {
3043 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3044 } else {
b748f260 3045 struct net_device *dev = nh->fib_nh_dev;
901731b8
DA		 3046
3047 mtu = IPV6_MIN_MTU;
3048 idev = __in6_dev_get(dev);
3049 if (idev && idev->cnf.mtu6 > mtu)
3050 mtu = idev->cnf.mtu6;
3051 }
3052
3053 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3054out:
b748f260 3055 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
901731b8
DA		 3056}
3057
3b00944c 3058struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
87a11578 3059 struct flowi6 *fl6)
1da177e4 3060{
87a11578 3061 struct dst_entry *dst;
1da177e4
LT		 3062 struct rt6_info *rt;
3063 struct inet6_dev *idev = in6_dev_get(dev);
c346dca1 3064 struct net *net = dev_net(dev);
1da177e4 3065
38308473 3066 if (unlikely(!idev))
122bdf67 3067 return ERR_PTR(-ENODEV);
1da177e4 3068
ad706862 3069 rt = ip6_dst_alloc(net, dev, 0);
38308473 3070 if (unlikely(!rt)) {
1da177e4 3071 in6_dev_put(idev);
87a11578 3072 dst = ERR_PTR(-ENOMEM);
1da177e4
LT		 3073 goto out;
3074 }
3075
8e2ec639 3076 rt->dst.flags |= DST_HOST;
588753f1 3077 rt->dst.input = ip6_input;
8e2ec639 3078 rt->dst.output = ip6_output;
550bab42 3079 rt->rt6i_gateway = fl6->daddr;
87a11578 3080 rt->rt6i_dst.addr = fl6->daddr;
8e2ec639
YZ		 3081 rt->rt6i_dst.plen = 128;
3082 rt->rt6i_idev = idev;
14edd87d 3083 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1da177e4 3084
4c981e28 3085 /* Add this dst into uncached_list so that rt6_disable_ip() can
587fea74
WW		 3086 * do proper release of the net_device
3087 */
3088 rt6_uncached_list_add(rt);
81eb8447 3089 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
1da177e4 3090
87a11578
DM		 3091 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3092
1da177e4 3093out:
87a11578 3094 return dst;
1da177e4
LT		 3095}
3096
569d3645 3097static int ip6_dst_gc(struct dst_ops *ops)
1da177e4 3098{
86393e52 3099 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
7019b78e
DL		 3100 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3101 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3102 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3103 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3104 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
fc66f95c 3105 int entries;
7019b78e 3106
fc66f95c 3107 entries = dst_entries_get_fast(ops);
49a18d86 3108 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
fc66f95c 3109 entries <= rt_max_size)
1da177e4
LT		 3110 goto out;
3111
6891a346 3112 net->ipv6.ip6_rt_gc_expire++;
14956643 3113 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
fc66f95c
ED		 3114 entries = dst_entries_get_slow(ops);
3115 if (entries < ops->gc_thresh)
7019b78e 3116 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1da177e4 3117out:
7019b78e 3118 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
fc66f95c 3119 return entries > rt_max_size;
1da177e4
LT		 3120}
3121
8c14586f
DA		 3122static struct rt6_info *ip6_nh_lookup_table(struct net *net,
3123 struct fib6_config *cfg,
f4797b33
DA		 3124 const struct in6_addr *gw_addr,
3125 u32 tbid, int flags)
8c14586f
DA		 3126{
3127 struct flowi6 fl6 = {
3128 .flowi6_oif = cfg->fc_ifindex,
3129 .daddr = *gw_addr,
3130 .saddr = cfg->fc_prefsrc,
3131 };
3132 struct fib6_table *table;
3133 struct rt6_info *rt;
8c14586f 3134
f4797b33 3135 table = fib6_get_table(net, tbid);
8c14586f
DA		 3136 if (!table)
3137 return NULL;
3138
3139 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3140 flags |= RT6_LOOKUP_F_HAS_SADDR;
3141
f4797b33 3142 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
b75cc8f9 3143 rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, NULL, flags);
8c14586f
DA		 3144
3145 /* if table lookup failed, fall back to full lookup */
3146 if (rt == net->ipv6.ip6_null_entry) {
3147 ip6_rt_put(rt);
3148 rt = NULL;
3149 }
3150
3151 return rt;
3152}
3153
fc1e64e1
DA		 3154static int ip6_route_check_nh_onlink(struct net *net,
3155 struct fib6_config *cfg,
9fbb704c 3156 const struct net_device *dev,
fc1e64e1
DA		 3157 struct netlink_ext_ack *extack)
3158{
44750f84 3159 u32 tbid = l3mdev_fib_table(dev) ? : RT_TABLE_MAIN;
fc1e64e1
DA		 3160 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3161 u32 flags = RTF_LOCAL | RTF_ANYCAST | RTF_REJECT;
bf1dc8ba 3162 struct fib6_info *from;
fc1e64e1
DA		 3163 struct rt6_info *grt;
3164 int err;
3165
3166 err = 0;
3167 grt = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0);
3168 if (grt) {
bf1dc8ba
PA		 3169 rcu_read_lock();
3170 from = rcu_dereference(grt->from);
58e354c0 3171 if (!grt->dst.error &&
4ed591c8 3172 /* ignore match if it is the default route */
bf1dc8ba 3173 from && !ipv6_addr_any(&from->fib6_dst.addr) &&
58e354c0 3174 (grt->rt6i_flags & flags || dev != grt->dst.dev)) {
44750f84
DA		 3175 NL_SET_ERR_MSG(extack,
3176 "Nexthop has invalid gateway or device mismatch");
fc1e64e1
DA		 3177 err = -EINVAL;
3178 }
bf1dc8ba 3179 rcu_read_unlock();
fc1e64e1
DA		 3180
3181 ip6_rt_put(grt);
3182 }
3183
3184 return err;
3185}
3186
1edce99f
DA		 3187static int ip6_route_check_nh(struct net *net,
3188 struct fib6_config *cfg,
3189 struct net_device **_dev,
3190 struct inet6_dev **idev)
3191{
3192 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3193 struct net_device *dev = _dev ? *_dev : NULL;
3194 struct rt6_info *grt = NULL;
3195 int err = -EHOSTUNREACH;
3196
3197 if (cfg->fc_table) {
f4797b33
DA		 3198 int flags = RT6_LOOKUP_F_IFACE;
3199
3200 grt = ip6_nh_lookup_table(net, cfg, gw_addr,
3201 cfg->fc_table, flags);
1edce99f
DA		 3202 if (grt) {
3203 if (grt->rt6i_flags & RTF_GATEWAY ||
3204 (dev && dev != grt->dst.dev)) {
3205 ip6_rt_put(grt);
3206 grt = NULL;
3207 }
3208 }
3209 }
3210
3211 if (!grt)
b75cc8f9 3212 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, NULL, 1);
1edce99f
DA		 3213
3214 if (!grt)
3215 goto out;
3216
3217 if (dev) {
3218 if (dev != grt->dst.dev) {
3219 ip6_rt_put(grt);
3220 goto out;
3221 }
3222 } else {
3223 *_dev = dev = grt->dst.dev;
3224 *idev = grt->rt6i_idev;
3225 dev_hold(dev);
3226 in6_dev_hold(grt->rt6i_idev);
3227 }
3228
3229 if (!(grt->rt6i_flags & RTF_GATEWAY))
3230 err = 0;
3231
3232 ip6_rt_put(grt);
3233
3234out:
3235 return err;
3236}
3237
9fbb704c
DA		 3238static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3239 struct net_device **_dev, struct inet6_dev **idev,
3240 struct netlink_ext_ack *extack)
3241{
3242 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3243 int gwa_type = ipv6_addr_type(gw_addr);
232378e8 3244 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
9fbb704c 3245 const struct net_device *dev = *_dev;
232378e8 3246 bool need_addr_check = !dev;
9fbb704c
DA		 3247 int err = -EINVAL;
3248
3249 /* if gw_addr is local we will fail to detect this in case
3250 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3251 * will return already-added prefix route via interface that
3252 * prefix route was assigned to, which might be non-loopback.
3253 */
232378e8
DA		 3254 if (dev &&
3255 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3256 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
9fbb704c
DA		 3257 goto out;
3258 }
3259
3260 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3261 /* IPv6 strictly inhibits using not link-local
3262 * addresses as nexthop address.
3263 * Otherwise, router will not able to send redirects.
3264 * It is very good, but in some (rare!) circumstances
3265 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3266 * some exceptions. --ANK
3267 * We allow IPv4-mapped nexthops to support RFC4798-type
3268 * addressing
3269 */
3270 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3271 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3272 goto out;
3273 }
3274
3275 if (cfg->fc_flags & RTNH_F_ONLINK)
3276 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3277 else
3278 err = ip6_route_check_nh(net, cfg, _dev, idev);
3279
3280 if (err)
3281 goto out;
3282 }
3283
3284 /* reload in case device was changed */
3285 dev = *_dev;
3286
3287 err = -EINVAL;
3288 if (!dev) {
3289 NL_SET_ERR_MSG(extack, "Egress device not specified");
3290 goto out;
3291 } else if (dev->flags & IFF_LOOPBACK) {
3292 NL_SET_ERR_MSG(extack,
3293 "Egress device can not be loopback device for this route");
3294 goto out;
3295 }
232378e8
DA		 3296
3297 /* if we did not check gw_addr above, do so now that the
3298 * egress device has been resolved.
3299 */
3300 if (need_addr_check &&
3301 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3302 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3303 goto out;
3304 }
3305
9fbb704c
DA		 3306 err = 0;
3307out:
3308 return err;
3309}
3310
83c44251
DA		 3311static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3312{
3313 if ((flags & RTF_REJECT) ||
3314 (dev && (dev->flags & IFF_LOOPBACK) &&
3315 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3316 !(flags & RTF_LOCAL)))
3317 return true;
3318
3319 return false;
3320}
3321
3322int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3323 struct fib6_config *cfg, gfp_t gfp_flags,
3324 struct netlink_ext_ack *extack)
3325{
3326 struct net_device *dev = NULL;
3327 struct inet6_dev *idev = NULL;
3328 int addr_type;
3329 int err;
3330
f1741730
DA		 3331 fib6_nh->fib_nh_family = AF_INET6;
3332
83c44251
DA		 3333 err = -ENODEV;
3334 if (cfg->fc_ifindex) {
3335 dev = dev_get_by_index(net, cfg->fc_ifindex);
3336 if (!dev)
3337 goto out;
3338 idev = in6_dev_get(dev);
3339 if (!idev)
3340 goto out;
3341 }
3342
3343 if (cfg->fc_flags & RTNH_F_ONLINK) {
3344 if (!dev) {
3345 NL_SET_ERR_MSG(extack,
3346 "Nexthop device required for onlink");
3347 goto out;
3348 }
3349
3350 if (!(dev->flags & IFF_UP)) {
3351 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3352 err = -ENETDOWN;
3353 goto out;
3354 }
3355
ad1601ae 3356 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
83c44251
DA		 3357 }
3358
ad1601ae 3359 fib6_nh->fib_nh_weight = 1;
83c44251
DA		 3360
3361 /* We cannot add true routes via loopback here,
3362 * they would result in kernel looping; promote them to reject routes
3363 */
3364 addr_type = ipv6_addr_type(&cfg->fc_dst);
3365 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3366 /* hold loopback dev/idev if we haven't done so. */
3367 if (dev != net->loopback_dev) {
3368 if (dev) {
3369 dev_put(dev);
3370 in6_dev_put(idev);
3371 }
3372 dev = net->loopback_dev;
3373 dev_hold(dev);
3374 idev = in6_dev_get(dev);
3375 if (!idev) {
3376 err = -ENODEV;
3377 goto out;
3378 }
3379 }
7dd73168 3380 goto pcpu_alloc;
83c44251
DA		 3381 }
3382
3383 if (cfg->fc_flags & RTF_GATEWAY) {
3384 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3385 if (err)
3386 goto out;
3387
ad1601ae 3388 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
bdf00467 3389 fib6_nh->fib_nh_gw_family = AF_INET6;
83c44251
DA		 3390 }
3391
3392 err = -ENODEV;
3393 if (!dev)
3394 goto out;
3395
3396 if (idev->cnf.disable_ipv6) {
3397 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3398 err = -EACCES;
3399 goto out;
3400 }
3401
3402 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3403 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3404 err = -ENETDOWN;
3405 goto out;
3406 }
3407
3408 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3409 !netif_carrier_ok(dev))
ad1601ae 3410 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
83c44251 3411
7dd73168
DA		 3412 err = fib_nh_common_init(&fib6_nh->nh_common, cfg->fc_encap,
3413 cfg->fc_encap_type, cfg, gfp_flags, extack);
3414 if (err)
3415 goto out;
3416
3417pcpu_alloc:
f40b6ae2
DA		 3418 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3419 if (!fib6_nh->rt6i_pcpu) {
3420 err = -ENOMEM;
3421 goto out;
3422 }
3423
ad1601ae 3424 fib6_nh->fib_nh_dev = dev;
f1741730 3425 fib6_nh->fib_nh_oif = dev->ifindex;
83c44251
DA		 3426 err = 0;
3427out:
3428 if (idev)
3429 in6_dev_put(idev);
3430
3431 if (err) {
ad1601ae
DA		 3432 lwtstate_put(fib6_nh->fib_nh_lws);
3433 fib6_nh->fib_nh_lws = NULL;
83c44251
DA		 3434 if (dev)
3435 dev_put(dev);
3436 }
3437
3438 return err;
3439}
3440
dac7d0f2
DA		 3441void fib6_nh_release(struct fib6_nh *fib6_nh)
3442{
cc5c073a
DA		 3443 struct rt6_exception_bucket *bucket;
3444
3445 rcu_read_lock();
3446
3447 fib6_nh_flush_exceptions(fib6_nh, NULL);
3448 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3449 if (bucket) {
3450 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3451 kfree(bucket);
3452 }
3453
3454 rcu_read_unlock();
3455
f40b6ae2
DA		 3456 if (fib6_nh->rt6i_pcpu) {
3457 int cpu;
3458
3459 for_each_possible_cpu(cpu) {
3460 struct rt6_info **ppcpu_rt;
3461 struct rt6_info *pcpu_rt;
3462
3463 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3464 pcpu_rt = *ppcpu_rt;
3465 if (pcpu_rt) {
3466 dst_dev_put(&pcpu_rt->dst);
3467 dst_release(&pcpu_rt->dst);
3468 *ppcpu_rt = NULL;
3469 }
3470 }
3471
3472 free_percpu(fib6_nh->rt6i_pcpu);
3473 }
3474
979e276e 3475 fib_nh_common_release(&fib6_nh->nh_common);
dac7d0f2
DA		 3476}
3477
8d1c802b 3478static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
acb54e3c 3479 gfp_t gfp_flags,
333c4301 3480 struct netlink_ext_ack *extack)
1da177e4 3481{
5578689a 3482 struct net *net = cfg->fc_nlinfo.nl_net;
8d1c802b 3483 struct fib6_info *rt = NULL;
f88d8ea6 3484 struct nexthop *nh = NULL;
c71099ac 3485 struct fib6_table *table;
f88d8ea6 3486 struct fib6_nh *fib6_nh;
8c5b83f0 3487 int err = -EINVAL;
83c44251 3488 int addr_type;
1da177e4 3489
557c44be 3490 /* RTF_PCPU is an internal flag; can not be set by userspace */
d5d531cb
DA		 3491 if (cfg->fc_flags & RTF_PCPU) {
3492 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
557c44be 3493 goto out;
d5d531cb 3494 }
557c44be 3495
2ea2352e
WW		 3496 /* RTF_CACHE is an internal flag; can not be set by userspace */
3497 if (cfg->fc_flags & RTF_CACHE) {
3498 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3499 goto out;
3500 }
3501
e8478e80
DA		 3502 if (cfg->fc_type > RTN_MAX) {
3503 NL_SET_ERR_MSG(extack, "Invalid route type");
3504 goto out;
3505 }
3506
d5d531cb
DA		 3507 if (cfg->fc_dst_len > 128) {
3508 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3509 goto out;
3510 }
3511 if (cfg->fc_src_len > 128) {
3512 NL_SET_ERR_MSG(extack, "Invalid source address length");
8c5b83f0 3513 goto out;
d5d531cb 3514 }
1da177e4 3515#ifndef CONFIG_IPV6_SUBTREES
d5d531cb
DA		 3516 if (cfg->fc_src_len) {
3517 NL_SET_ERR_MSG(extack,
3518 "Specifying source address requires IPV6_SUBTREES to be enabled");
8c5b83f0 3519 goto out;
d5d531cb 3520 }
1da177e4 3521#endif
5b98324e
DA		 3522 if (cfg->fc_nh_id) {
3523 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3524 if (!nh) {
3525 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3526 goto out;
3527 }
3528 err = fib6_check_nexthop(nh, cfg, extack);
3529 if (err)
3530 goto out;
3531 }
fc1e64e1 3532
d71314b4 3533 err = -ENOBUFS;
38308473
DM		 3534 if (cfg->fc_nlinfo.nlh &&
3535 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
d71314b4 3536 table = fib6_get_table(net, cfg->fc_table);
38308473 3537 if (!table) {
f3213831 3538 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
d71314b4
MV		 3539 table = fib6_new_table(net, cfg->fc_table);
3540 }
3541 } else {
3542 table = fib6_new_table(net, cfg->fc_table);
3543 }
38308473
DM		 3544
3545 if (!table)
c71099ac 3546 goto out;
c71099ac 3547
93531c67 3548 err = -ENOMEM;
f88d8ea6 3549 rt = fib6_info_alloc(gfp_flags, !nh);
93531c67 3550 if (!rt)
1da177e4 3551 goto out;
93531c67 3552
d7e774f3
DA		 3553 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3554 extack);
767a2217
DA		 3555 if (IS_ERR(rt->fib6_metrics)) {
3556 err = PTR_ERR(rt->fib6_metrics);
fda21d46
ED		 3557 /* Do not leave garbage there. */
3558 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
767a2217
DA		 3559 goto out;
3560 }
3561
93531c67
DA		 3562 if (cfg->fc_flags & RTF_ADDRCONF)
3563 rt->dst_nocount = true;
1da177e4 3564
1716a961 3565 if (cfg->fc_flags & RTF_EXPIRES)
14895687 3566 fib6_set_expires(rt, jiffies +
1716a961
G		 3567 clock_t_to_jiffies(cfg->fc_expires));
3568 else
14895687 3569 fib6_clean_expires(rt);
1da177e4 3570
86872cb5
TG		 3571 if (cfg->fc_protocol == RTPROT_UNSPEC)
3572 cfg->fc_protocol = RTPROT_BOOT;
93c2fb25 3573 rt->fib6_protocol = cfg->fc_protocol;
86872cb5 3574
83c44251
DA		 3575 rt->fib6_table = table;
3576 rt->fib6_metric = cfg->fc_metric;
c7036d97 3577 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
2b2450ca 3578 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
19e42e45 3579
93c2fb25
DA		 3580 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3581 rt->fib6_dst.plen = cfg->fc_dst_len;
3582 if (rt->fib6_dst.plen == 128)
3b6761d1 3583 rt->dst_host = true;
e5fd387a 3584
1da177e4 3585#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 3586 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3587 rt->fib6_src.plen = cfg->fc_src_len;
1da177e4 3588#endif
f88d8ea6
DA		 3589 if (nh) {
3590 if (!nexthop_get(nh)) {
3591 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3592 goto out;
3593 }
3594 if (rt->fib6_src.plen) {
4daa95af 3595 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
f88d8ea6
DA		 3596 goto out;
3597 }
3598 rt->nh = nh;
3599 fib6_nh = nexthop_fib6_nh(rt->nh);
3600 } else {
3601 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3602 if (err)
3603 goto out;
1da177e4 3604
f88d8ea6
DA		 3605 fib6_nh = rt->fib6_nh;
3606
3607 /* We cannot add true routes via loopback here, they would
3608 * result in kernel looping; promote them to reject routes
3609 */
3610 addr_type = ipv6_addr_type(&cfg->fc_dst);
3611 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3612 addr_type))
3613 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3614 }
955ec4cb 3615
c3968a85 3616 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
f88d8ea6 3617 struct net_device *dev = fib6_nh->fib_nh_dev;
83c44251 3618
c3968a85 3619 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
d5d531cb 3620 NL_SET_ERR_MSG(extack, "Invalid source address");
c3968a85
DW		 3621 err = -EINVAL;
3622 goto out;
3623 }
93c2fb25
DA		 3624 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3625 rt->fib6_prefsrc.plen = 128;
c3968a85 3626 } else
93c2fb25 3627 rt->fib6_prefsrc.plen = 0;
c3968a85 3628
8c5b83f0 3629 return rt;
6b9ea5a6 3630out:
93531c67 3631 fib6_info_release(rt);
8c5b83f0 3632 return ERR_PTR(err);
6b9ea5a6
RP		 3633}
3634
acb54e3c 3635int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
333c4301 3636 struct netlink_ext_ack *extack)
6b9ea5a6 3637{
8d1c802b 3638 struct fib6_info *rt;
6b9ea5a6
RP		 3639 int err;
3640
acb54e3c 3641 rt = ip6_route_info_create(cfg, gfp_flags, extack);
d4ead6b3
DA		 3642 if (IS_ERR(rt))
3643 return PTR_ERR(rt);
6b9ea5a6 3644
d4ead6b3 3645 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
93531c67 3646 fib6_info_release(rt);
6b9ea5a6 3647
1da177e4
LT		 3648 return err;
3649}
3650
8d1c802b 3651static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
1da177e4 3652{
afb1d4b5 3653 struct net *net = info->nl_net;
c71099ac 3654 struct fib6_table *table;
afb1d4b5 3655 int err;
1da177e4 3656
421842ed 3657 if (rt == net->ipv6.fib6_null_entry) {
6825a26c
G		 3658 err = -ENOENT;
3659 goto out;
3660 }
6c813a72 3661
93c2fb25 3662 table = rt->fib6_table;
66f5d6ce 3663 spin_lock_bh(&table->tb6_lock);
86872cb5 3664 err = fib6_del(rt, info);
66f5d6ce 3665 spin_unlock_bh(&table->tb6_lock);
1da177e4 3666
6825a26c 3667out:
93531c67 3668 fib6_info_release(rt);
1da177e4
LT		 3669 return err;
3670}
3671
8d1c802b 3672int ip6_del_rt(struct net *net, struct fib6_info *rt)
e0a1ad73 3673{
afb1d4b5
DA		 3674 struct nl_info info = { .nl_net = net };
3675
528c4ceb 3676 return __ip6_del_rt(rt, &info);
e0a1ad73
TG		 3677}
3678
8d1c802b 3679static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
0ae81335
DA		 3680{
3681 struct nl_info *info = &cfg->fc_nlinfo;
e3330039 3682 struct net *net = info->nl_net;
16a16cd3 3683 struct sk_buff *skb = NULL;
0ae81335 3684 struct fib6_table *table;
e3330039 3685 int err = -ENOENT;
0ae81335 3686
421842ed 3687 if (rt == net->ipv6.fib6_null_entry)
e3330039 3688 goto out_put;
93c2fb25 3689 table = rt->fib6_table;
66f5d6ce 3690 spin_lock_bh(&table->tb6_lock);
0ae81335 3691
93c2fb25 3692 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
8d1c802b 3693 struct fib6_info *sibling, *next_sibling;
0ae81335 3694
16a16cd3
DA		 3695 /* prefer to send a single notification with all hops */
3696 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3697 if (skb) {
3698 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3699
d4ead6b3 3700 if (rt6_fill_node(net, skb, rt, NULL,
16a16cd3
DA		 3701 NULL, NULL, 0, RTM_DELROUTE,
3702 info->portid, seq, 0) < 0) {
3703 kfree_skb(skb);
3704 skb = NULL;
3705 } else
3706 info->skip_notify = 1;
3707 }
3708
2881fd61
IS		 3709 info->skip_notify_kernel = 1;
3710 call_fib6_multipath_entry_notifiers(net,
3711 FIB_EVENT_ENTRY_DEL,
3712 rt,
3713 rt->fib6_nsiblings,
3714 NULL);
0ae81335 3715 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25
DA		 3716 &rt->fib6_siblings,
3717 fib6_siblings) {
0ae81335
DA		 3718 err = fib6_del(sibling, info);
3719 if (err)
e3330039 3720 goto out_unlock;
0ae81335
DA		 3721 }
3722 }
3723
3724 err = fib6_del(rt, info);
e3330039 3725out_unlock:
66f5d6ce 3726 spin_unlock_bh(&table->tb6_lock);
e3330039 3727out_put:
93531c67 3728 fib6_info_release(rt);
16a16cd3
DA		 3729
3730 if (skb) {
e3330039 3731 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
16a16cd3
DA		 3732 info->nlh, gfp_any());
3733 }
0ae81335
DA		 3734 return err;
3735}
3736
0fa6efc5 3737static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
23fb93a4
DA		 3738{
3739 int rc = -ESRCH;
3740
3741 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3742 goto out;
3743
3744 if (cfg->fc_flags & RTF_GATEWAY &&
3745 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3746 goto out;
761f6026
XL		 3747
3748 rc = rt6_remove_exception_rt(rt);
23fb93a4
DA		 3749out:
3750 return rc;
3751}
3752
0fa6efc5
DA		 3753static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3754 struct fib6_nh *nh)
3755{
3756 struct fib6_result res = {
3757 .f6i = rt,
3758 .nh = nh,
3759 };
3760 struct rt6_info *rt_cache;
3761
3762 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3763 if (rt_cache)
3764 return __ip6_del_cached_rt(rt_cache, cfg);
3765
3766 return 0;
3767}
3768
5b98324e
DA		 3769struct fib6_nh_del_cached_rt_arg {
3770 struct fib6_config *cfg;
3771 struct fib6_info *f6i;
3772};
3773
3774static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
3775{
3776 struct fib6_nh_del_cached_rt_arg *arg = _arg;
3777 int rc;
3778
3779 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
3780 return rc != -ESRCH ? rc : 0;
3781}
3782
3783static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
3784{
3785 struct fib6_nh_del_cached_rt_arg arg = {
3786 .cfg = cfg,
3787 .f6i = f6i
3788 };
3789
3790 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
3791}
3792
333c4301
DA		 3793static int ip6_route_del(struct fib6_config *cfg,
3794 struct netlink_ext_ack *extack)
1da177e4 3795{
c71099ac 3796 struct fib6_table *table;
8d1c802b 3797 struct fib6_info *rt;
1da177e4 3798 struct fib6_node *fn;
1da177e4
LT		 3799 int err = -ESRCH;
3800
5578689a 3801 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
d5d531cb
DA		 3802 if (!table) {
3803 NL_SET_ERR_MSG(extack, "FIB table does not exist");
c71099ac 3804 return err;
d5d531cb 3805 }
c71099ac 3806
66f5d6ce 3807 rcu_read_lock();
1da177e4 3808
c71099ac 3809 fn = fib6_locate(&table->tb6_root,
86872cb5 3810 &cfg->fc_dst, cfg->fc_dst_len,
38fbeeee 3811 &cfg->fc_src, cfg->fc_src_len,
2b760fcf 3812 !(cfg->fc_flags & RTF_CACHE));
1ab1457c 3813
1da177e4 3814 if (fn) {
66f5d6ce 3815 for_each_fib6_node_rt_rcu(fn) {
ad1601ae
DA		 3816 struct fib6_nh *nh;
3817
5b98324e
DA		 3818 if (rt->nh && rt->nh->id != cfg->fc_nh_id)
3819 continue;
23fb93a4 3820
5b98324e
DA		 3821 if (cfg->fc_flags & RTF_CACHE) {
3822 int rc = 0;
3823
3824 if (rt->nh) {
3825 rc = ip6_del_cached_rt_nh(cfg, rt);
3826 } else if (cfg->fc_nh_id) {
3827 continue;
3828 } else {
3829 nh = rt->fib6_nh;
3830 rc = ip6_del_cached_rt(cfg, rt, nh);
3831 }
0fa6efc5
DA		 3832 if (rc != -ESRCH) {
3833 rcu_read_unlock();
3834 return rc;
23fb93a4
DA		 3835 }
3836 continue;
2b760fcf 3837 }
ad1601ae 3838
5b98324e
DA		 3839 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
3840 continue;
3841 if (cfg->fc_protocol &&
3842 cfg->fc_protocol != rt->fib6_protocol)
3843 continue;
3844
3845 if (rt->nh) {
3846 if (!fib6_info_hold_safe(rt))
3847 continue;
3848 rcu_read_unlock();
3849
3850 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3851 }
3852 if (cfg->fc_nh_id)
3853 continue;
3854
3855 nh = rt->fib6_nh;
86872cb5 3856 if (cfg->fc_ifindex &&
ad1601ae
DA		 3857 (!nh->fib_nh_dev ||
3858 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
1da177e4 3859 continue;
86872cb5 3860 if (cfg->fc_flags & RTF_GATEWAY &&
ad1601ae 3861 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
1da177e4 3862 continue;
e873e4b9
WW		 3863 if (!fib6_info_hold_safe(rt))
3864 continue;
66f5d6ce 3865 rcu_read_unlock();
1da177e4 3866
0ae81335
DA		 3867 /* if gateway was specified only delete the one hop */
3868 if (cfg->fc_flags & RTF_GATEWAY)
3869 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3870
3871 return __ip6_del_rt_siblings(rt, cfg);
1da177e4
LT		 3872 }
3873 }
66f5d6ce 3874 rcu_read_unlock();
1da177e4
LT		 3875
3876 return err;
3877}
3878
6700c270 3879static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
a6279458 3880{
a6279458 3881 struct netevent_redirect netevent;
e8599ff4 3882 struct rt6_info *rt, *nrt = NULL;
85bd05de 3883 struct fib6_result res = {};
e8599ff4
DM		 3884 struct ndisc_options ndopts;
3885 struct inet6_dev *in6_dev;
3886 struct neighbour *neigh;
71bcdba0 3887 struct rd_msg *msg;
6e157b6a
DM		 3888 int optlen, on_link;
3889 u8 *lladdr;
e8599ff4 3890
29a3cad5 3891 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
71bcdba0 3892 optlen -= sizeof(*msg);
e8599ff4
DM		 3893
3894 if (optlen < 0) {
6e157b6a 3895 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
e8599ff4
DM		 3896 return;
3897 }
3898
71bcdba0 3899 msg = (struct rd_msg *)icmp6_hdr(skb);
e8599ff4 3900
71bcdba0 3901 if (ipv6_addr_is_multicast(&msg->dest)) {
6e157b6a 3902 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
e8599ff4
DM		 3903 return;
3904 }
3905
6e157b6a 3906 on_link = 0;
71bcdba0 3907 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
e8599ff4 3908 on_link = 1;
71bcdba0 3909 } else if (ipv6_addr_type(&msg->target) !=
e8599ff4 3910 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
6e157b6a 3911 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
e8599ff4
DM		 3912 return;
3913 }
3914
3915 in6_dev = __in6_dev_get(skb->dev);
3916 if (!in6_dev)
3917 return;
3918 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
3919 return;
3920
3921 /* RFC2461 8.1:
3922 * The IP source address of the Redirect MUST be the same as the current
3923 * first-hop router for the specified ICMP Destination Address.
3924 */
3925
f997c55c 3926 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
e8599ff4
DM		 3927 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
3928 return;
3929 }
6e157b6a
DM		 3930
3931 lladdr = NULL;
e8599ff4
DM		 3932 if (ndopts.nd_opts_tgt_lladdr) {
3933 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
3934 skb->dev);
3935 if (!lladdr) {
3936 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
3937 return;
3938 }
3939 }
3940
6e157b6a 3941 rt = (struct rt6_info *) dst;
ec13ad1d 3942 if (rt->rt6i_flags & RTF_REJECT) {
6e157b6a 3943 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
e8599ff4 3944 return;
6e157b6a 3945 }
e8599ff4 3946
6e157b6a
DM		 3947 /* Redirect received -> path was valid.
3948 * Look, redirects are sent only in response to data packets,
3949 * so that this nexthop apparently is reachable. --ANK
3950 */
0dec879f 3951 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
a6279458 3952
71bcdba0 3953 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
6e157b6a
DM		 3954 if (!neigh)
3955 return;
a6279458 3956
1da177e4
LT		 3957 /*
3958 * We have finally decided to accept it.
3959 */
3960
f997c55c 3961 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1da177e4
LT		 3962 NEIGH_UPDATE_F_WEAK_OVERRIDE|
3963 NEIGH_UPDATE_F_OVERRIDE|
3964 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
f997c55c
AA		 3965 NEIGH_UPDATE_F_ISROUTER)),
3966 NDISC_REDIRECT, &ndopts);
1da177e4 3967
4d85cd0c 3968 rcu_read_lock();
85bd05de 3969 res.f6i = rcu_dereference(rt->from);
ff24e498 3970 if (!res.f6i)
886b7a50 3971 goto out;
8a14e46f 3972
49d5b8ef
DA		 3973 if (res.f6i->nh) {
3974 struct fib6_nh_match_arg arg = {
3975 .dev = dst->dev,
3976 .gw = &rt->rt6i_gateway,
3977 };
3978
3979 nexthop_for_each_fib6_nh(res.f6i->nh,
3980 fib6_nh_find_match, &arg);
3981
3982 /* fib6_info uses a nexthop that does not have fib6_nh
3983 * using the dst->dev. Should be impossible
3984 */
3985 if (!arg.match)
3986 goto out;
3987 res.nh = arg.match;
3988 } else {
3989 res.nh = res.f6i->fib6_nh;
3990 }
3991
7d21fec9
DA		 3992 res.fib6_flags = res.f6i->fib6_flags;
3993 res.fib6_type = res.f6i->fib6_type;
85bd05de 3994 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
38308473 3995 if (!nrt)
1da177e4
LT		 3996 goto out;
3997
3998 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
3999 if (on_link)
4000 nrt->rt6i_flags &= ~RTF_GATEWAY;
4001
4e3fd7a0 4002 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1da177e4 4003
886b7a50 4004 /* rt6_insert_exception() will take care of duplicated exceptions */
5012f0a5 4005 if (rt6_insert_exception(nrt, &res)) {
2b760fcf
WW		 4006 dst_release_immediate(&nrt->dst);
4007 goto out;
4008 }
1da177e4 4009
d8d1f30b
CG		 4010 netevent.old = &rt->dst;
4011 netevent.new = &nrt->dst;
71bcdba0 4012 netevent.daddr = &msg->dest;
60592833 4013 netevent.neigh = neigh;
8d71740c
TT		 4014 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4015
1da177e4 4016out:
886b7a50 4017 rcu_read_unlock();
e8599ff4 4018 neigh_release(neigh);
6e157b6a
DM		 4019}
4020
70ceb4f5 4021#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 4022static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 4023 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 4024 const struct in6_addr *gwaddr,
4025 struct net_device *dev)
70ceb4f5 4026{
830218c1
DA		 4027 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4028 int ifindex = dev->ifindex;
70ceb4f5 4029 struct fib6_node *fn;
8d1c802b 4030 struct fib6_info *rt = NULL;
c71099ac
TG		 4031 struct fib6_table *table;
4032
830218c1 4033 table = fib6_get_table(net, tb_id);
38308473 4034 if (!table)
c71099ac 4035 return NULL;
70ceb4f5 4036
66f5d6ce 4037 rcu_read_lock();
38fbeeee 4038 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
70ceb4f5
YH		 4039 if (!fn)
4040 goto out;
4041
66f5d6ce 4042 for_each_fib6_node_rt_rcu(fn) {
f88d8ea6
DA		 4043 /* these routes do not use nexthops */
4044 if (rt->nh)
4045 continue;
1cf844c7 4046 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
70ceb4f5 4047 continue;
2b2450ca 4048 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
1cf844c7 4049 !rt->fib6_nh->fib_nh_gw_family)
70ceb4f5 4050 continue;
1cf844c7 4051 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
70ceb4f5 4052 continue;
e873e4b9
WW		 4053 if (!fib6_info_hold_safe(rt))
4054 continue;
70ceb4f5
YH		 4055 break;
4056 }
4057out:
66f5d6ce 4058 rcu_read_unlock();
70ceb4f5
YH		 4059 return rt;
4060}
4061
8d1c802b 4062static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 4063 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 4064 const struct in6_addr *gwaddr,
4065 struct net_device *dev,
95c96174 4066 unsigned int pref)
70ceb4f5 4067{
86872cb5 4068 struct fib6_config cfg = {
238fc7ea 4069 .fc_metric = IP6_RT_PRIO_USER,
830218c1 4070 .fc_ifindex = dev->ifindex,
86872cb5
TG		 4071 .fc_dst_len = prefixlen,
4072 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4073 RTF_UP | RTF_PREF(pref),
b91d5329 4074 .fc_protocol = RTPROT_RA,
e8478e80 4075 .fc_type = RTN_UNICAST,
15e47304 4076 .fc_nlinfo.portid = 0,
efa2cea0
DL		 4077 .fc_nlinfo.nlh = NULL,
4078 .fc_nlinfo.nl_net = net,
86872cb5
TG		 4079 };
4080
830218c1 4081 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
4e3fd7a0
AD		 4082 cfg.fc_dst = *prefix;
4083 cfg.fc_gateway = *gwaddr;
70ceb4f5 4084
e317da96
YH		 4085 /* We should treat it as a default route if prefix length is 0. */
4086 if (!prefixlen)
86872cb5 4087 cfg.fc_flags |= RTF_DEFAULT;
70ceb4f5 4088
acb54e3c 4089 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
70ceb4f5 4090
830218c1 4091 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
70ceb4f5
YH		 4092}
4093#endif
4094
8d1c802b 4095struct fib6_info *rt6_get_dflt_router(struct net *net,
afb1d4b5
DA		 4096 const struct in6_addr *addr,
4097 struct net_device *dev)
1ab1457c 4098{
830218c1 4099 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
8d1c802b 4100 struct fib6_info *rt;
c71099ac 4101 struct fib6_table *table;
1da177e4 4102
afb1d4b5 4103 table = fib6_get_table(net, tb_id);
38308473 4104 if (!table)
c71099ac 4105 return NULL;
1da177e4 4106
66f5d6ce
WW		 4107 rcu_read_lock();
4108 for_each_fib6_node_rt_rcu(&table->tb6_root) {
f88d8ea6 4109 struct fib6_nh *nh;
ad1601ae 4110
f88d8ea6
DA		 4111 /* RA routes do not use nexthops */
4112 if (rt->nh)
4113 continue;
4114
4115 nh = rt->fib6_nh;
ad1601ae 4116 if (dev == nh->fib_nh_dev &&
93c2fb25 4117 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
ad1601ae 4118 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
1da177e4
LT		 4119 break;
4120 }
e873e4b9
WW		 4121 if (rt && !fib6_info_hold_safe(rt))
4122 rt = NULL;
66f5d6ce 4123 rcu_read_unlock();
1da177e4
LT		 4124 return rt;
4125}
4126
8d1c802b 4127struct fib6_info *rt6_add_dflt_router(struct net *net,
afb1d4b5 4128 const struct in6_addr *gwaddr,
ebacaaa0
YH		 4129 struct net_device *dev,
4130 unsigned int pref)
1da177e4 4131{
86872cb5 4132 struct fib6_config cfg = {
ca254490 4133 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
238fc7ea 4134 .fc_metric = IP6_RT_PRIO_USER,
86872cb5
TG		 4135 .fc_ifindex = dev->ifindex,
4136 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4137 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
b91d5329 4138 .fc_protocol = RTPROT_RA,
e8478e80 4139 .fc_type = RTN_UNICAST,
15e47304 4140 .fc_nlinfo.portid = 0,
5578689a 4141 .fc_nlinfo.nlh = NULL,
afb1d4b5 4142 .fc_nlinfo.nl_net = net,
86872cb5 4143 };
1da177e4 4144
4e3fd7a0 4145 cfg.fc_gateway = *gwaddr;
1da177e4 4146
acb54e3c 4147 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
830218c1
DA		 4148 struct fib6_table *table;
4149
4150 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4151 if (table)
4152 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4153 }
1da177e4 4154
afb1d4b5 4155 return rt6_get_dflt_router(net, gwaddr, dev);
1da177e4
LT		 4156}
4157
afb1d4b5
DA		 4158static void __rt6_purge_dflt_routers(struct net *net,
4159 struct fib6_table *table)
1da177e4 4160{
8d1c802b 4161 struct fib6_info *rt;
1da177e4
LT		 4162
4163restart:
66f5d6ce
WW		 4164 rcu_read_lock();
4165 for_each_fib6_node_rt_rcu(&table->tb6_root) {
dcd1f572
DA		 4166 struct net_device *dev = fib6_info_nh_dev(rt);
4167 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4168
93c2fb25 4169 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
e873e4b9
WW		 4170 (!idev || idev->cnf.accept_ra != 2) &&
4171 fib6_info_hold_safe(rt)) {
93531c67
DA		 4172 rcu_read_unlock();
4173 ip6_del_rt(net, rt);
1da177e4
LT		 4174 goto restart;
4175 }
4176 }
66f5d6ce 4177 rcu_read_unlock();
830218c1
DA		 4178
4179 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4180}
4181
4182void rt6_purge_dflt_routers(struct net *net)
4183{
4184 struct fib6_table *table;
4185 struct hlist_head *head;
4186 unsigned int h;
4187
4188 rcu_read_lock();
4189
4190 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4191 head = &net->ipv6.fib_table_hash[h];
4192 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4193 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
afb1d4b5 4194 __rt6_purge_dflt_routers(net, table);
830218c1
DA		 4195 }
4196 }
4197
4198 rcu_read_unlock();
1da177e4
LT		 4199}
4200
5578689a
DL		 4201static void rtmsg_to_fib6_config(struct net *net,
4202 struct in6_rtmsg *rtmsg,
86872cb5
TG		 4203 struct fib6_config *cfg)
4204{
8823a3ac
MŻ		 4205 *cfg = (struct fib6_config){
4206 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4207 : RT6_TABLE_MAIN,
4208 .fc_ifindex = rtmsg->rtmsg_ifindex,
67f69513 4209 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
8823a3ac
MŻ		 4210 .fc_expires = rtmsg->rtmsg_info,
4211 .fc_dst_len = rtmsg->rtmsg_dst_len,
4212 .fc_src_len = rtmsg->rtmsg_src_len,
4213 .fc_flags = rtmsg->rtmsg_flags,
4214 .fc_type = rtmsg->rtmsg_type,
4215
4216 .fc_nlinfo.nl_net = net,
4217
4218 .fc_dst = rtmsg->rtmsg_dst,
4219 .fc_src = rtmsg->rtmsg_src,
4220 .fc_gateway = rtmsg->rtmsg_gateway,
4221 };
86872cb5
TG		 4222}
4223
5578689a 4224int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1da177e4 4225{
86872cb5 4226 struct fib6_config cfg;
1da177e4
LT		 4227 struct in6_rtmsg rtmsg;
4228 int err;
4229
67ba4152 4230 switch (cmd) {
1da177e4
LT		 4231 case SIOCADDRT: /* Add a route */
4232 case SIOCDELRT: /* Delete a route */
af31f412 4233 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4
LT		 4234 return -EPERM;
4235 err = copy_from_user(&rtmsg, arg,
4236 sizeof(struct in6_rtmsg));
4237 if (err)
4238 return -EFAULT;
86872cb5 4239
5578689a 4240 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
86872cb5 4241
1da177e4
LT		 4242 rtnl_lock();
4243 switch (cmd) {
4244 case SIOCADDRT:
acb54e3c 4245 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
1da177e4
LT		 4246 break;
4247 case SIOCDELRT:
333c4301 4248 err = ip6_route_del(&cfg, NULL);
1da177e4
LT		 4249 break;
4250 default:
4251 err = -EINVAL;
4252 }
4253 rtnl_unlock();
4254
4255 return err;
3ff50b79 4256 }
1da177e4
LT		 4257
4258 return -EINVAL;
4259}
4260
4261/*
4262 * Drop the packet on the floor
4263 */
4264
d5fdd6ba 4265static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1da177e4 4266{
adf30907 4267 struct dst_entry *dst = skb_dst(skb);
1d3fd8a1
SS		 4268 struct net *net = dev_net(dst->dev);
4269 struct inet6_dev *idev;
4270 int type;
4271
4272 if (netif_is_l3_master(skb->dev) &&
4273 dst->dev == net->loopback_dev)
4274 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4275 else
4276 idev = ip6_dst_idev(dst);
4277
612f09e8
YH		 4278 switch (ipstats_mib_noroutes) {
4279 case IPSTATS_MIB_INNOROUTES:
0660e03f 4280 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
45bb0060 4281 if (type == IPV6_ADDR_ANY) {
1d3fd8a1 4282 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
612f09e8
YH		 4283 break;
4284 }
4285 /* FALLTHROUGH */
4286 case IPSTATS_MIB_OUTNOROUTES:
1d3fd8a1 4287 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
612f09e8
YH		 4288 break;
4289 }
1d3fd8a1
SS		 4290
4291 /* Start over by dropping the dst for l3mdev case */
4292 if (netif_is_l3_master(skb->dev))
4293 skb_dst_drop(skb);
4294
3ffe533c 4295 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1da177e4
LT		 4296 kfree_skb(skb);
4297 return 0;
4298}
4299
9ce8ade0
TG		 4300static int ip6_pkt_discard(struct sk_buff *skb)
4301{
612f09e8 4302 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4303}
4304
ede2059d 4305static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
1da177e4 4306{
adf30907 4307 skb->dev = skb_dst(skb)->dev;
612f09e8 4308 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1da177e4
LT		 4309}
4310
9ce8ade0
TG		 4311static int ip6_pkt_prohibit(struct sk_buff *skb)
4312{
612f09e8 4313 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4314}
4315
ede2059d 4316static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
9ce8ade0 4317{
adf30907 4318 skb->dev = skb_dst(skb)->dev;
612f09e8 4319 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
9ce8ade0
TG		 4320}
4321
1da177e4
LT		 4322/*
4323 * Allocate a dst for local (unicast / anycast) address.
4324 */
4325
360a9887
DA		 4326struct fib6_info *addrconf_f6i_alloc(struct net *net,
4327 struct inet6_dev *idev,
4328 const struct in6_addr *addr,
4329 bool anycast, gfp_t gfp_flags)
1da177e4 4330{
c7a1ce39
DA		 4331 struct fib6_config cfg = {
4332 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4333 .fc_ifindex = idev->dev->ifindex,
4334 .fc_flags = RTF_UP | RTF_ADDRCONF | RTF_NONEXTHOP,
4335 .fc_dst = *addr,
4336 .fc_dst_len = 128,
4337 .fc_protocol = RTPROT_KERNEL,
4338 .fc_nlinfo.nl_net = net,
4339 .fc_ignore_dev_down = true,
4340 };
1da177e4 4341
e8478e80 4342 if (anycast) {
c7a1ce39
DA		 4343 cfg.fc_type = RTN_ANYCAST;
4344 cfg.fc_flags |= RTF_ANYCAST;
e8478e80 4345 } else {
c7a1ce39
DA		 4346 cfg.fc_type = RTN_LOCAL;
4347 cfg.fc_flags |= RTF_LOCAL;
e8478e80 4348 }
1da177e4 4349
c7a1ce39 4350 return ip6_route_info_create(&cfg, gfp_flags, NULL);
1da177e4
LT		 4351}
4352
c3968a85
DW		 4353/* remove deleted ip from prefsrc entries */
4354struct arg_dev_net_ip {
4355 struct net_device *dev;
4356 struct net *net;
4357 struct in6_addr *addr;
4358};
4359
8d1c802b 4360static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
c3968a85
DW		 4361{
4362 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4363 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4364 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4365
f88d8ea6
DA		 4366 if (!rt->nh &&
4367 ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
421842ed 4368 rt != net->ipv6.fib6_null_entry &&
93c2fb25 4369 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
60006a48 4370 spin_lock_bh(&rt6_exception_lock);
c3968a85 4371 /* remove prefsrc entry */
93c2fb25 4372 rt->fib6_prefsrc.plen = 0;
60006a48 4373 spin_unlock_bh(&rt6_exception_lock);
c3968a85
DW		 4374 }
4375 return 0;
4376}
4377
4378void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4379{
4380 struct net *net = dev_net(ifp->idev->dev);
4381 struct arg_dev_net_ip adni = {
4382 .dev = ifp->idev->dev,
4383 .net = net,
4384 .addr = &ifp->addr,
4385 };
0c3584d5 4386 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
c3968a85
DW		 4387}
4388
2b2450ca 4389#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
be7a010d
DJ		 4390
4391/* Remove routers and update dst entries when gateway turn into host. */
8d1c802b 4392static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
be7a010d
DJ		 4393{
4394 struct in6_addr *gateway = (struct in6_addr *)arg;
f88d8ea6
DA		 4395 struct fib6_nh *nh;
4396
4397 /* RA routes do not use nexthops */
4398 if (rt->nh)
4399 return 0;
be7a010d 4400
f88d8ea6 4401 nh = rt->fib6_nh;
93c2fb25 4402 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
cc5c073a 4403 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
be7a010d 4404 return -1;
b16cb459
WW		 4405
4406 /* Further clean up cached routes in exception table.
4407 * This is needed because cached route may have a different
4408 * gateway than its 'parent' in the case of an ip redirect.
4409 */
cc5c073a 4410 fib6_nh_exceptions_clean_tohost(nh, gateway);
b16cb459 4411
be7a010d
DJ		 4412 return 0;
4413}
4414
4415void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4416{
4417 fib6_clean_all(net, fib6_clean_tohost, gateway);
4418}
4419
2127d95a
IS		 4420struct arg_netdev_event {
4421 const struct net_device *dev;
4c981e28 4422 union {
ecc5663c 4423 unsigned char nh_flags;
4c981e28
IS		 4424 unsigned long event;
4425 };
2127d95a
IS		 4426};
4427
8d1c802b 4428static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
d7dedee1 4429{
8d1c802b 4430 struct fib6_info *iter;
d7dedee1
IS		 4431 struct fib6_node *fn;
4432
93c2fb25
DA		 4433 fn = rcu_dereference_protected(rt->fib6_node,
4434 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4435 iter = rcu_dereference_protected(fn->leaf,
93c2fb25 4436 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4437 while (iter) {
93c2fb25 4438 if (iter->fib6_metric == rt->fib6_metric &&
33bd5ac5 4439 rt6_qualify_for_ecmp(iter))
d7dedee1 4440 return iter;
8fb11a9a 4441 iter = rcu_dereference_protected(iter->fib6_next,
93c2fb25 4442 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1
IS		 4443 }
4444
4445 return NULL;
4446}
4447
f88d8ea6 4448/* only called for fib entries with builtin fib6_nh */
8d1c802b 4449static bool rt6_is_dead(const struct fib6_info *rt)
d7dedee1 4450{
1cf844c7
DA		 4451 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4452 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4453 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
d7dedee1
IS		 4454 return true;
4455
4456 return false;
4457}
4458
8d1c802b 4459static int rt6_multipath_total_weight(const struct fib6_info *rt)
d7dedee1 4460{
8d1c802b 4461 struct fib6_info *iter;
d7dedee1
IS		 4462 int total = 0;
4463
4464 if (!rt6_is_dead(rt))
1cf844c7 4465 total += rt->fib6_nh->fib_nh_weight;
d7dedee1 4466
93c2fb25 4467 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
d7dedee1 4468 if (!rt6_is_dead(iter))
1cf844c7 4469 total += iter->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4470 }
4471
4472 return total;
4473}
4474
8d1c802b 4475static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
d7dedee1
IS		 4476{
4477 int upper_bound = -1;
4478
4479 if (!rt6_is_dead(rt)) {
1cf844c7 4480 *weight += rt->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4481 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4482 total) - 1;
4483 }
1cf844c7 4484 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
d7dedee1
IS		 4485}
4486
8d1c802b 4487static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
d7dedee1 4488{
8d1c802b 4489 struct fib6_info *iter;
d7dedee1
IS		 4490 int weight = 0;
4491
4492 rt6_upper_bound_set(rt, &weight, total);
4493
93c2fb25 4494 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
d7dedee1
IS		 4495 rt6_upper_bound_set(iter, &weight, total);
4496}
4497
8d1c802b 4498void rt6_multipath_rebalance(struct fib6_info *rt)
d7dedee1 4499{
8d1c802b 4500 struct fib6_info *first;
d7dedee1
IS		 4501 int total;
4502
4503 /* In case the entire multipath route was marked for flushing,
4504 * then there is no need to rebalance upon the removal of every
4505 * sibling route.
4506 */
93c2fb25 4507 if (!rt->fib6_nsiblings || rt->should_flush)
d7dedee1
IS		 4508 return;
4509
4510 /* During lookup routes are evaluated in order, so we need to
4511 * make sure upper bounds are assigned from the first sibling
4512 * onwards.
4513 */
4514 first = rt6_multipath_first_sibling(rt);
4515 if (WARN_ON_ONCE(!first))
4516 return;
4517
4518 total = rt6_multipath_total_weight(first);
4519 rt6_multipath_upper_bound_set(first, total);
4520}
4521
8d1c802b 4522static int fib6_ifup(struct fib6_info *rt, void *p_arg)
2127d95a
IS		 4523{
4524 const struct arg_netdev_event *arg = p_arg;
7aef6859 4525 struct net *net = dev_net(arg->dev);
2127d95a 4526
f88d8ea6 4527 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
1cf844c7
DA		 4528 rt->fib6_nh->fib_nh_dev == arg->dev) {
4529 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
7aef6859 4530 fib6_update_sernum_upto_root(net, rt);
d7dedee1 4531 rt6_multipath_rebalance(rt);
1de178ed 4532 }
2127d95a
IS		 4533
4534 return 0;
4535}
4536
ecc5663c 4537void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
2127d95a
IS		 4538{
4539 struct arg_netdev_event arg = {
4540 .dev = dev,
6802f3ad
IS		 4541 {
4542 .nh_flags = nh_flags,
4543 },
2127d95a
IS		 4544 };
4545
4546 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4547 arg.nh_flags |= RTNH_F_LINKDOWN;
4548
4549 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4550}
4551
f88d8ea6 4552/* only called for fib entries with inline fib6_nh */
8d1c802b 4553static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
1de178ed
IS		 4554 const struct net_device *dev)
4555{
8d1c802b 4556 struct fib6_info *iter;
1de178ed 4557
1cf844c7 4558 if (rt->fib6_nh->fib_nh_dev == dev)
1de178ed 4559 return true;
93c2fb25 4560 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7 4561 if (iter->fib6_nh->fib_nh_dev == dev)
1de178ed
IS		 4562 return true;
4563
4564 return false;
4565}
4566
8d1c802b 4567static void rt6_multipath_flush(struct fib6_info *rt)
1de178ed 4568{
8d1c802b 4569 struct fib6_info *iter;
1de178ed
IS		 4570
4571 rt->should_flush = 1;
93c2fb25 4572 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1de178ed
IS		 4573 iter->should_flush = 1;
4574}
4575
8d1c802b 4576static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
1de178ed
IS		 4577 const struct net_device *down_dev)
4578{
8d1c802b 4579 struct fib6_info *iter;
1de178ed
IS		 4580 unsigned int dead = 0;
4581
1cf844c7
DA		 4582 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4583 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed 4584 dead++;
93c2fb25 4585 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4586 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4587 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed
IS		 4588 dead++;
4589
4590 return dead;
4591}
4592
8d1c802b 4593static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
1de178ed 4594 const struct net_device *dev,
ecc5663c 4595 unsigned char nh_flags)
1de178ed 4596{
8d1c802b 4597 struct fib6_info *iter;
1de178ed 4598
1cf844c7
DA		 4599 if (rt->fib6_nh->fib_nh_dev == dev)
4600 rt->fib6_nh->fib_nh_flags |= nh_flags;
93c2fb25 4601 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4602 if (iter->fib6_nh->fib_nh_dev == dev)
4603 iter->fib6_nh->fib_nh_flags |= nh_flags;
1de178ed
IS		 4604}
4605
a1a22c12 4606/* called with write lock held for table with rt */
8d1c802b 4607static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
1da177e4 4608{
4c981e28
IS		 4609 const struct arg_netdev_event *arg = p_arg;
4610 const struct net_device *dev = arg->dev;
7aef6859 4611 struct net *net = dev_net(dev);
8ed67789 4612
f88d8ea6 4613 if (rt == net->ipv6.fib6_null_entry || rt->nh)
27c6fa73
IS		 4614 return 0;
4615
4616 switch (arg->event) {
4617 case NETDEV_UNREGISTER:
1cf844c7 4618 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
27c6fa73 4619 case NETDEV_DOWN:
1de178ed 4620 if (rt->should_flush)
27c6fa73 4621 return -1;
93c2fb25 4622 if (!rt->fib6_nsiblings)
1cf844c7 4623 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
1de178ed
IS		 4624 if (rt6_multipath_uses_dev(rt, dev)) {
4625 unsigned int count;
4626
4627 count = rt6_multipath_dead_count(rt, dev);
93c2fb25 4628 if (rt->fib6_nsiblings + 1 == count) {
1de178ed
IS		 4629 rt6_multipath_flush(rt);
4630 return -1;
4631 }
4632 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4633 RTNH_F_LINKDOWN);
7aef6859 4634 fib6_update_sernum(net, rt);
d7dedee1 4635 rt6_multipath_rebalance(rt);
1de178ed
IS		 4636 }
4637 return -2;
27c6fa73 4638 case NETDEV_CHANGE:
1cf844c7 4639 if (rt->fib6_nh->fib_nh_dev != dev ||
93c2fb25 4640 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
27c6fa73 4641 break;
1cf844c7 4642 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
d7dedee1 4643 rt6_multipath_rebalance(rt);
27c6fa73 4644 break;
2b241361 4645 }
c159d30c 4646
1da177e4
LT		 4647 return 0;
4648}
4649
27c6fa73 4650void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
1da177e4 4651{
4c981e28 4652 struct arg_netdev_event arg = {
8ed67789 4653 .dev = dev,
6802f3ad
IS		 4654 {
4655 .event = event,
4656 },
8ed67789 4657 };
7c6bb7d2 4658 struct net *net = dev_net(dev);
8ed67789 4659
7c6bb7d2
DA		 4660 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4661 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4662 else
4663 fib6_clean_all(net, fib6_ifdown, &arg);
4c981e28
IS		 4664}
4665
4666void rt6_disable_ip(struct net_device *dev, unsigned long event)
4667{
4668 rt6_sync_down_dev(dev, event);
4669 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4670 neigh_ifdown(&nd_tbl, dev);
1da177e4
LT		 4671}
4672
95c96174 4673struct rt6_mtu_change_arg {
1da177e4 4674 struct net_device *dev;
95c96174 4675 unsigned int mtu;
c0b220cf 4676 struct fib6_info *f6i;
1da177e4
LT		 4677};
4678
cc5c073a 4679static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
c0b220cf
DA		 4680{
4681 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
cc5c073a 4682 struct fib6_info *f6i = arg->f6i;
c0b220cf
DA		 4683
4684 /* For administrative MTU increase, there is no way to discover
4685 * IPv6 PMTU increase, so PMTU increase should be updated here.
4686 * Since RFC 1981 doesn't include administrative MTU increase
4687 * update PMTU increase is a MUST. (i.e. jumbo frame)
4688 */
4689 if (nh->fib_nh_dev == arg->dev) {
4690 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4691 u32 mtu = f6i->fib6_pmtu;
4692
4693 if (mtu >= arg->mtu ||
4694 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4695 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4696
4697 spin_lock_bh(&rt6_exception_lock);
cc5c073a 4698 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
c0b220cf
DA		 4699 spin_unlock_bh(&rt6_exception_lock);
4700 }
4701
4702 return 0;
4703}
4704
4705static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
1da177e4
LT		 4706{
4707 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4708 struct inet6_dev *idev;
4709
4710 /* In IPv6 pmtu discovery is not optional,
4711 so that RTAX_MTU lock cannot disable it.
4712 We still use this lock to block changes
4713 caused by addrconf/ndisc.
4714 */
4715
4716 idev = __in6_dev_get(arg->dev);
38308473 4717 if (!idev)
1da177e4
LT		 4718 return 0;
4719
c0b220cf
DA		 4720 if (fib6_metric_locked(f6i, RTAX_MTU))
4721 return 0;
d4ead6b3 4722
c0b220cf 4723 arg->f6i = f6i;
2d44234b
DA		 4724 if (f6i->nh) {
4725 /* fib6_nh_mtu_change only returns 0, so this is safe */
4726 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4727 arg);
4728 }
4729
1cf844c7 4730 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
1da177e4
LT		 4731}
4732
95c96174 4733void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
1da177e4 4734{
c71099ac
TG		 4735 struct rt6_mtu_change_arg arg = {
4736 .dev = dev,
4737 .mtu = mtu,
4738 };
1da177e4 4739
0c3584d5 4740 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
1da177e4
LT		 4741}
4742
ef7c79ed 4743static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
75425657 4744 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
5176f91e 4745 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
aa8f8778 4746 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
86872cb5 4747 [RTA_OIF] = { .type = NLA_U32 },
ab364a6f 4748 [RTA_IIF] = { .type = NLA_U32 },
86872cb5
TG		 4749 [RTA_PRIORITY] = { .type = NLA_U32 },
4750 [RTA_METRICS] = { .type = NLA_NESTED },
51ebd318 4751 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
c78ba6d6 4752 [RTA_PREF] = { .type = NLA_U8 },
19e42e45
RP		 4753 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4754 [RTA_ENCAP] = { .type = NLA_NESTED },
32bc201e 4755 [RTA_EXPIRES] = { .type = NLA_U32 },
622ec2c9 4756 [RTA_UID] = { .type = NLA_U32 },
3b45a410 4757 [RTA_MARK] = { .type = NLA_U32 },
aa8f8778 4758 [RTA_TABLE] = { .type = NLA_U32 },
eacb9384
RP		 4759 [RTA_IP_PROTO] = { .type = NLA_U8 },
4760 [RTA_SPORT] = { .type = NLA_U16 },
4761 [RTA_DPORT] = { .type = NLA_U16 },
5b98324e 4762 [RTA_NH_ID] = { .type = NLA_U32 },
86872cb5
TG		 4763};
4764
4765static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
333c4301
DA		 4766 struct fib6_config *cfg,
4767 struct netlink_ext_ack *extack)
1da177e4 4768{
86872cb5
TG		 4769 struct rtmsg *rtm;
4770 struct nlattr *tb[RTA_MAX+1];
c78ba6d6 4771 unsigned int pref;
86872cb5 4772 int err;
1da177e4 4773
8cb08174
JB		 4774 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
4775 rtm_ipv6_policy, extack);
86872cb5
TG		 4776 if (err < 0)
4777 goto errout;
1da177e4 4778
86872cb5
TG		 4779 err = -EINVAL;
4780 rtm = nlmsg_data(nlh);
86872cb5 4781
84db8407
MŻ		 4782 *cfg = (struct fib6_config){
4783 .fc_table = rtm->rtm_table,
4784 .fc_dst_len = rtm->rtm_dst_len,
4785 .fc_src_len = rtm->rtm_src_len,
4786 .fc_flags = RTF_UP,
4787 .fc_protocol = rtm->rtm_protocol,
4788 .fc_type = rtm->rtm_type,
4789
4790 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
4791 .fc_nlinfo.nlh = nlh,
4792 .fc_nlinfo.nl_net = sock_net(skb->sk),
4793 };
86872cb5 4794
ef2c7d7b
ND		 4795 if (rtm->rtm_type == RTN_UNREACHABLE ||
4796 rtm->rtm_type == RTN_BLACKHOLE ||
b4949ab2
ND		 4797 rtm->rtm_type == RTN_PROHIBIT ||
4798 rtm->rtm_type == RTN_THROW)
86872cb5
TG		 4799 cfg->fc_flags |= RTF_REJECT;
4800
ab79ad14
MŻ		 4801 if (rtm->rtm_type == RTN_LOCAL)
4802 cfg->fc_flags |= RTF_LOCAL;
4803
1f56a01f
MKL		 4804 if (rtm->rtm_flags & RTM_F_CLONED)
4805 cfg->fc_flags |= RTF_CACHE;
4806
fc1e64e1
DA		 4807 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4808
5b98324e
DA		 4809 if (tb[RTA_NH_ID]) {
4810 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
4811 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
4812 NL_SET_ERR_MSG(extack,
4813 "Nexthop specification and nexthop id are mutually exclusive");
4814 goto errout;
4815 }
4816 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
4817 }
4818
86872cb5 4819 if (tb[RTA_GATEWAY]) {
67b61f6c 4820 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
86872cb5 4821 cfg->fc_flags |= RTF_GATEWAY;
1da177e4 4822 }
e3818541
DA		 4823 if (tb[RTA_VIA]) {
4824 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
4825 goto errout;
4826 }
86872cb5
TG		 4827
4828 if (tb[RTA_DST]) {
4829 int plen = (rtm->rtm_dst_len + 7) >> 3;
4830
4831 if (nla_len(tb[RTA_DST]) < plen)
4832 goto errout;
4833
4834 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1da177e4 4835 }
86872cb5
TG		 4836
4837 if (tb[RTA_SRC]) {
4838 int plen = (rtm->rtm_src_len + 7) >> 3;
4839
4840 if (nla_len(tb[RTA_SRC]) < plen)
4841 goto errout;
4842
4843 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1da177e4 4844 }
86872cb5 4845
c3968a85 4846 if (tb[RTA_PREFSRC])
67b61f6c 4847 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
c3968a85 4848
86872cb5
TG		 4849 if (tb[RTA_OIF])
4850 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4851
4852 if (tb[RTA_PRIORITY])
4853 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4854
4855 if (tb[RTA_METRICS]) {
4856 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4857 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
1da177e4 4858 }
86872cb5
TG		 4859
4860 if (tb[RTA_TABLE])
4861 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4862
51ebd318
ND		 4863 if (tb[RTA_MULTIPATH]) {
4864 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4865 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
9ed59592
DA		 4866
4867 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
c255bd68 4868 cfg->fc_mp_len, extack);
9ed59592
DA		 4869 if (err < 0)
4870 goto errout;
51ebd318
ND		 4871 }
4872
c78ba6d6
LR		 4873 if (tb[RTA_PREF]) {
4874 pref = nla_get_u8(tb[RTA_PREF]);
4875 if (pref != ICMPV6_ROUTER_PREF_LOW &&
4876 pref != ICMPV6_ROUTER_PREF_HIGH)
4877 pref = ICMPV6_ROUTER_PREF_MEDIUM;
4878 cfg->fc_flags |= RTF_PREF(pref);
4879 }
4880
19e42e45
RP		 4881 if (tb[RTA_ENCAP])
4882 cfg->fc_encap = tb[RTA_ENCAP];
4883
9ed59592 4884 if (tb[RTA_ENCAP_TYPE]) {
19e42e45
RP		 4885 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
4886
c255bd68 4887 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
9ed59592
DA		 4888 if (err < 0)
4889 goto errout;
4890 }
4891
32bc201e
XL		 4892 if (tb[RTA_EXPIRES]) {
4893 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
4894
4895 if (addrconf_finite_timeout(timeout)) {
4896 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
4897 cfg->fc_flags |= RTF_EXPIRES;
4898 }
4899 }
4900
86872cb5
TG		 4901 err = 0;
4902errout:
4903 return err;
1da177e4
LT		 4904}
4905
6b9ea5a6 4906struct rt6_nh {
8d1c802b 4907 struct fib6_info *fib6_info;
6b9ea5a6 4908 struct fib6_config r_cfg;
6b9ea5a6
RP		 4909 struct list_head next;
4910};
4911
d4ead6b3
DA		 4912static int ip6_route_info_append(struct net *net,
4913 struct list_head *rt6_nh_list,
8d1c802b
DA		 4914 struct fib6_info *rt,
4915 struct fib6_config *r_cfg)
6b9ea5a6
RP		 4916{
4917 struct rt6_nh *nh;
6b9ea5a6
RP		 4918 int err = -EEXIST;
4919
4920 list_for_each_entry(nh, rt6_nh_list, next) {
8d1c802b
DA		 4921 /* check if fib6_info already exists */
4922 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
6b9ea5a6
RP		 4923 return err;
4924 }
4925
4926 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
4927 if (!nh)
4928 return -ENOMEM;
8d1c802b 4929 nh->fib6_info = rt;
6b9ea5a6
RP		 4930 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
4931 list_add_tail(&nh->next, rt6_nh_list);
4932
4933 return 0;
4934}
4935
8d1c802b
DA		 4936static void ip6_route_mpath_notify(struct fib6_info *rt,
4937 struct fib6_info *rt_last,
3b1137fe
DA		 4938 struct nl_info *info,
4939 __u16 nlflags)
4940{
4941 /* if this is an APPEND route, then rt points to the first route
4942 * inserted and rt_last points to last route inserted. Userspace
4943 * wants a consistent dump of the route which starts at the first
4944 * nexthop. Since sibling routes are always added at the end of
4945 * the list, find the first sibling of the last route appended
4946 */
93c2fb25
DA		 4947 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
4948 rt = list_first_entry(&rt_last->fib6_siblings,
8d1c802b 4949 struct fib6_info,
93c2fb25 4950 fib6_siblings);
3b1137fe
DA		 4951 }
4952
4953 if (rt)
4954 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
4955}
4956
333c4301
DA		 4957static int ip6_route_multipath_add(struct fib6_config *cfg,
4958 struct netlink_ext_ack *extack)
51ebd318 4959{
8d1c802b 4960 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
3b1137fe 4961 struct nl_info *info = &cfg->fc_nlinfo;
ebee3cad 4962 enum fib_event_type event_type;
51ebd318
ND		 4963 struct fib6_config r_cfg;
4964 struct rtnexthop *rtnh;
8d1c802b 4965 struct fib6_info *rt;
6b9ea5a6
RP		 4966 struct rt6_nh *err_nh;
4967 struct rt6_nh *nh, *nh_safe;
3b1137fe 4968 __u16 nlflags;
51ebd318
ND		 4969 int remaining;
4970 int attrlen;
6b9ea5a6
RP		 4971 int err = 1;
4972 int nhn = 0;
4973 int replace = (cfg->fc_nlinfo.nlh &&
4974 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
4975 LIST_HEAD(rt6_nh_list);
51ebd318 4976
3b1137fe
DA		 4977 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
4978 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
4979 nlflags |= NLM_F_APPEND;
4980
35f1b4e9 4981 remaining = cfg->fc_mp_len;
51ebd318 4982 rtnh = (struct rtnexthop *)cfg->fc_mp;
51ebd318 4983
6b9ea5a6 4984 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
8d1c802b 4985 * fib6_info structs per nexthop
6b9ea5a6 4986 */
51ebd318
ND		 4987 while (rtnh_ok(rtnh, remaining)) {
4988 memcpy(&r_cfg, cfg, sizeof(*cfg));
4989 if (rtnh->rtnh_ifindex)
4990 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
4991
4992 attrlen = rtnh_attrlen(rtnh);
4993 if (attrlen > 0) {
4994 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
4995
4996 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
4997 if (nla) {
67b61f6c 4998 r_cfg.fc_gateway = nla_get_in6_addr(nla);
51ebd318
ND		 4999 r_cfg.fc_flags |= RTF_GATEWAY;
5000 }
19e42e45
RP		 5001 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5002 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5003 if (nla)
5004 r_cfg.fc_encap_type = nla_get_u16(nla);
51ebd318 5005 }
6b9ea5a6 5006
68e2ffde 5007 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
acb54e3c 5008 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
8c5b83f0
RP		 5009 if (IS_ERR(rt)) {
5010 err = PTR_ERR(rt);
5011 rt = NULL;
6b9ea5a6 5012 goto cleanup;
8c5b83f0 5013 }
b5d2d75e
DA		 5014 if (!rt6_qualify_for_ecmp(rt)) {
5015 err = -EINVAL;
5016 NL_SET_ERR_MSG(extack,
5017 "Device only routes can not be added for IPv6 using the multipath API.");
5018 fib6_info_release(rt);
5019 goto cleanup;
5020 }
6b9ea5a6 5021
1cf844c7 5022 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
398958ae 5023
d4ead6b3
DA		 5024 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5025 rt, &r_cfg);
51ebd318 5026 if (err) {
93531c67 5027 fib6_info_release(rt);
6b9ea5a6
RP		 5028 goto cleanup;
5029 }
5030
5031 rtnh = rtnh_next(rtnh, &remaining);
5032 }
5033
9eee3b49
IS		 5034 if (list_empty(&rt6_nh_list)) {
5035 NL_SET_ERR_MSG(extack,
5036 "Invalid nexthop configuration - no valid nexthops");
5037 return -EINVAL;
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);
74109218 6020 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->rt6i_uncached);
8ed67789
DL		 6021
6022#ifdef CONFIG_IPV6_MULTIPLE_TABLES
feca7d8c 6023 net->ipv6.fib6_has_custom_rules = false;
8ed67789
DL		 6024 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6025 sizeof(*net->ipv6.ip6_prohibit_entry),
6026 GFP_KERNEL);
68fffc67
PZ		 6027 if (!net->ipv6.ip6_prohibit_entry)
6028 goto out_ip6_null_entry;
d8d1f30b 6029 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6030 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6031 ip6_template_metrics, true);
74109218 6032 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->rt6i_uncached);
8ed67789
DL		 6033
6034 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6035 sizeof(*net->ipv6.ip6_blk_hole_entry),
6036 GFP_KERNEL);
68fffc67
PZ		 6037 if (!net->ipv6.ip6_blk_hole_entry)
6038 goto out_ip6_prohibit_entry;
d8d1f30b 6039 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 6040 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6041 ip6_template_metrics, true);
74109218 6042 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->rt6i_uncached);
8ed67789
DL		 6043#endif
6044
b339a47c
PZ		 6045 net->ipv6.sysctl.flush_delay = 0;
6046 net->ipv6.sysctl.ip6_rt_max_size = 4096;
6047 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6048 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6049 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6050 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6051 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6052 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
7c6bb7d2 6053 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
b339a47c 6054
6891a346
BT		 6055 net->ipv6.ip6_rt_gc_expire = 30*HZ;
6056
8ed67789
DL		 6057 ret = 0;
6058out:
6059 return ret;
f2fc6a54 6060
68fffc67
PZ		 6061#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6062out_ip6_prohibit_entry:
6063 kfree(net->ipv6.ip6_prohibit_entry);
6064out_ip6_null_entry:
6065 kfree(net->ipv6.ip6_null_entry);
6066#endif
421842ed
DA		 6067out_fib6_null_entry:
6068 kfree(net->ipv6.fib6_null_entry);
fc66f95c
ED		 6069out_ip6_dst_entries:
6070 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
f2fc6a54 6071out_ip6_dst_ops:
f2fc6a54 6072 goto out;
cdb18761
DL		 6073}
6074
2c8c1e72 6075static void __net_exit ip6_route_net_exit(struct net *net)
cdb18761 6076{
421842ed 6077 kfree(net->ipv6.fib6_null_entry);
8ed67789
DL		 6078 kfree(net->ipv6.ip6_null_entry);
6079#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6080 kfree(net->ipv6.ip6_prohibit_entry);
6081 kfree(net->ipv6.ip6_blk_hole_entry);
6082#endif
41bb78b4 6083 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
cdb18761
DL		 6084}
6085
d189634e
TG		 6086static int __net_init ip6_route_net_init_late(struct net *net)
6087{
6088#ifdef CONFIG_PROC_FS
c3506372
CH		 6089 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6090 sizeof(struct ipv6_route_iter));
3617d949
CH		 6091 proc_create_net_single("rt6_stats", 0444, net->proc_net,
6092 rt6_stats_seq_show, NULL);
d189634e
TG		 6093#endif
6094 return 0;
6095}
6096
6097static void __net_exit ip6_route_net_exit_late(struct net *net)
6098{
6099#ifdef CONFIG_PROC_FS
ece31ffd
G		 6100 remove_proc_entry("ipv6_route", net->proc_net);
6101 remove_proc_entry("rt6_stats", net->proc_net);
d189634e
TG		 6102#endif
6103}
6104
cdb18761
DL		 6105static struct pernet_operations ip6_route_net_ops = {
6106 .init = ip6_route_net_init,
6107 .exit = ip6_route_net_exit,
6108};
6109
c3426b47
DM		 6110static int __net_init ipv6_inetpeer_init(struct net *net)
6111{
6112 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6113
6114 if (!bp)
6115 return -ENOMEM;
6116 inet_peer_base_init(bp);
6117 net->ipv6.peers = bp;
6118 return 0;
6119}
6120
6121static void __net_exit ipv6_inetpeer_exit(struct net *net)
6122{
6123 struct inet_peer_base *bp = net->ipv6.peers;
6124
6125 net->ipv6.peers = NULL;
56a6b248 6126 inetpeer_invalidate_tree(bp);
c3426b47
DM		 6127 kfree(bp);
6128}
6129
2b823f72 6130static struct pernet_operations ipv6_inetpeer_ops = {
c3426b47
DM		 6131 .init = ipv6_inetpeer_init,
6132 .exit = ipv6_inetpeer_exit,
6133};
6134
d189634e
TG		 6135static struct pernet_operations ip6_route_net_late_ops = {
6136 .init = ip6_route_net_init_late,
6137 .exit = ip6_route_net_exit_late,
6138};
6139
8ed67789
DL		 6140static struct notifier_block ip6_route_dev_notifier = {
6141 .notifier_call = ip6_route_dev_notify,
242d3a49 6142 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
8ed67789
DL		 6143};
6144
2f460933
WC		 6145void __init ip6_route_init_special_entries(void)
6146{
6147 /* Registering of the loopback is done before this portion of code,
6148 * the loopback reference in rt6_info will not be taken, do it
6149 * manually for init_net */
1cf844c7 6150 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
2f460933
WC		 6151 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6152 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6153 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6154 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6155 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6156 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6157 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6158 #endif
6159}
6160
433d49c3 6161int __init ip6_route_init(void)
1da177e4 6162{
433d49c3 6163 int ret;
8d0b94af 6164 int cpu;
433d49c3 6165
9a7ec3a9
DL		 6166 ret = -ENOMEM;
6167 ip6_dst_ops_template.kmem_cachep =
e5d679f3 6168 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
f845ab6b 6169 SLAB_HWCACHE_ALIGN, NULL);
9a7ec3a9 6170 if (!ip6_dst_ops_template.kmem_cachep)
c19a28e1 6171 goto out;
14e50e57 6172
fc66f95c 6173 ret = dst_entries_init(&ip6_dst_blackhole_ops);
8ed67789 6174 if (ret)
bdb3289f 6175 goto out_kmem_cache;
bdb3289f 6176
c3426b47
DM		 6177 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6178 if (ret)
e8803b6c 6179 goto out_dst_entries;
2a0c451a 6180
7e52b33b
DM		 6181 ret = register_pernet_subsys(&ip6_route_net_ops);
6182 if (ret)
6183 goto out_register_inetpeer;
c3426b47 6184
5dc121e9
AE		 6185 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6186
e8803b6c 6187 ret = fib6_init();
433d49c3 6188 if (ret)
8ed67789 6189 goto out_register_subsys;
433d49c3 6190
433d49c3
DL		 6191 ret = xfrm6_init();
6192 if (ret)
e8803b6c 6193 goto out_fib6_init;
c35b7e72 6194
433d49c3
DL		 6195 ret = fib6_rules_init();
6196 if (ret)
6197 goto xfrm6_init;
7e5449c2 6198
d189634e
TG		 6199 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6200 if (ret)
6201 goto fib6_rules_init;
6202
16feebcf
FW		 6203 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6204 inet6_rtm_newroute, NULL, 0);
6205 if (ret < 0)
6206 goto out_register_late_subsys;
6207
6208 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6209 inet6_rtm_delroute, NULL, 0);
6210 if (ret < 0)
6211 goto out_register_late_subsys;
6212
6213 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6214 inet6_rtm_getroute, NULL,
6215 RTNL_FLAG_DOIT_UNLOCKED);
6216 if (ret < 0)
d189634e 6217 goto out_register_late_subsys;
c127ea2c 6218
8ed67789 6219 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
cdb18761 6220 if (ret)
d189634e 6221 goto out_register_late_subsys;
8ed67789 6222
8d0b94af
MKL		 6223 for_each_possible_cpu(cpu) {
6224 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6225
6226 INIT_LIST_HEAD(&ul->head);
6227 spin_lock_init(&ul->lock);
6228 }
6229
433d49c3
DL		 6230out:
6231 return ret;
6232
d189634e 6233out_register_late_subsys:
16feebcf 6234 rtnl_unregister_all(PF_INET6);
d189634e 6235 unregister_pernet_subsys(&ip6_route_net_late_ops);
433d49c3 6236fib6_rules_init:
433d49c3
DL		 6237 fib6_rules_cleanup();
6238xfrm6_init:
433d49c3 6239 xfrm6_fini();
2a0c451a
TG		 6240out_fib6_init:
6241 fib6_gc_cleanup();
8ed67789
DL		 6242out_register_subsys:
6243 unregister_pernet_subsys(&ip6_route_net_ops);
7e52b33b
DM		 6244out_register_inetpeer:
6245 unregister_pernet_subsys(&ipv6_inetpeer_ops);
fc66f95c
ED		 6246out_dst_entries:
6247 dst_entries_destroy(&ip6_dst_blackhole_ops);
433d49c3 6248out_kmem_cache:
f2fc6a54 6249 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
433d49c3 6250 goto out;
1da177e4
LT		 6251}
6252
6253void ip6_route_cleanup(void)
6254{
8ed67789 6255 unregister_netdevice_notifier(&ip6_route_dev_notifier);
d189634e 6256 unregister_pernet_subsys(&ip6_route_net_late_ops);
101367c2 6257 fib6_rules_cleanup();
1da177e4 6258 xfrm6_fini();
1da177e4 6259 fib6_gc_cleanup();
c3426b47 6260 unregister_pernet_subsys(&ipv6_inetpeer_ops);
8ed67789 6261 unregister_pernet_subsys(&ip6_route_net_ops);
41bb78b4 6262 dst_entries_destroy(&ip6_dst_blackhole_ops);
f2fc6a54 6263 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
1da177e4 6264}
Linux 6.x block layer and io_uring tree(s)