projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ipv6: Handle all fib6_nh in a nexthop in __ip6_route_redirect
[linux-block.git] / net / ipv6 / route.c
CommitLineData
2874c5fd 1// SPDX-License-Identifier: GPL-2.0-or-later
1da177e4
LT		 2/*
3 * Linux INET6 implementation
4 * FIB front-end.
5 *
6 * Authors:
1ab1457c 7 * Pedro Roque <roque@di.fc.ul.pt>
1da177e4
LT		 8 */
9
10/* Changes:
11 *
12 * YOSHIFUJI Hideaki @USAGI
13 * reworked default router selection.
14 * - respect outgoing interface
15 * - select from (probably) reachable routers (i.e.
16 * routers in REACHABLE, STALE, DELAY or PROBE states).
17 * - always select the same router if it is (probably)
18 * reachable. otherwise, round-robin the list.
c0bece9f
YH		 19 * Ville Nuorvala
20 * Fixed routing subtrees.
1da177e4
LT		 21 */
22
f3213831
JP		 23#define pr_fmt(fmt) "IPv6: " fmt
24
4fc268d2 25#include <linux/capability.h>
1da177e4 26#include <linux/errno.h>
bc3b2d7f 27#include <linux/export.h>
1da177e4
LT		 28#include <linux/types.h>
29#include <linux/times.h>
30#include <linux/socket.h>
31#include <linux/sockios.h>
32#include <linux/net.h>
33#include <linux/route.h>
34#include <linux/netdevice.h>
35#include <linux/in6.h>
7bc570c8 36#include <linux/mroute6.h>
1da177e4 37#include <linux/init.h>
1da177e4 38#include <linux/if_arp.h>
1da177e4
LT		 39#include <linux/proc_fs.h>
40#include <linux/seq_file.h>
5b7c931d 41#include <linux/nsproxy.h>
5a0e3ad6 42#include <linux/slab.h>
35732d01 43#include <linux/jhash.h>
457c4cbc 44#include <net/net_namespace.h>
1da177e4
LT		 45#include <net/snmp.h>
46#include <net/ipv6.h>
47#include <net/ip6_fib.h>
48#include <net/ip6_route.h>
49#include <net/ndisc.h>
50#include <net/addrconf.h>
51#include <net/tcp.h>
52#include <linux/rtnetlink.h>
53#include <net/dst.h>
904af04d 54#include <net/dst_metadata.h>
1da177e4 55#include <net/xfrm.h>
8d71740c 56#include <net/netevent.h>
21713ebc 57#include <net/netlink.h>
3c618c1d 58#include <net/rtnh.h>
19e42e45 59#include <net/lwtunnel.h>
904af04d 60#include <net/ip_tunnels.h>
ca254490 61#include <net/l3mdev.h>
eacb9384 62#include <net/ip.h>
7c0f6ba6 63#include <linux/uaccess.h>
1da177e4
LT		 64
65#ifdef CONFIG_SYSCTL
66#include <linux/sysctl.h>
67#endif
68
30d444d3
DA		 69static int ip6_rt_type_to_error(u8 fib6_type);
70
71#define CREATE_TRACE_POINTS
72#include <trace/events/fib6.h>
73EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
74#undef CREATE_TRACE_POINTS
75
afc154e9 76enum rt6_nud_state {
7e980569
JB		 77 RT6_NUD_FAIL_HARD = -3,
78 RT6_NUD_FAIL_PROBE = -2,
79 RT6_NUD_FAIL_DO_RR = -1,
afc154e9
HFS		 80 RT6_NUD_SUCCEED = 1
81};
82
1da177e4 83static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
0dbaee3b 84static unsigned int ip6_default_advmss(const struct dst_entry *dst);
ebb762f2 85static unsigned int ip6_mtu(const struct dst_entry *dst);
1da177e4
LT		 86static struct dst_entry *ip6_negative_advice(struct dst_entry *);
87static void ip6_dst_destroy(struct dst_entry *);
88static void ip6_dst_ifdown(struct dst_entry *,
89 struct net_device *dev, int how);
569d3645 90static int ip6_dst_gc(struct dst_ops *ops);
1da177e4
LT		 91
92static int ip6_pkt_discard(struct sk_buff *skb);
ede2059d 93static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
7150aede 94static int ip6_pkt_prohibit(struct sk_buff *skb);
ede2059d 95static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1da177e4 96static void ip6_link_failure(struct sk_buff *skb);
6700c270
DM		 97static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
98 struct sk_buff *skb, u32 mtu);
99static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
100 struct sk_buff *skb);
702cea56
DA		 101static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
102 int strict);
a1b7a1f0 103static size_t rt6_nlmsg_size(struct fib6_info *f6i);
d4ead6b3 104static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 105 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 106 struct in6_addr *dest, struct in6_addr *src,
16a16cd3
DA		 107 int iif, int type, u32 portid, u32 seq,
108 unsigned int flags);
7e4b5128 109static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
510e2ced
WW		 110 const struct in6_addr *daddr,
111 const struct in6_addr *saddr);
1da177e4 112
70ceb4f5 113#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 114static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 115 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 116 const struct in6_addr *gwaddr,
117 struct net_device *dev,
95c96174 118 unsigned int pref);
8d1c802b 119static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 120 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 121 const struct in6_addr *gwaddr,
122 struct net_device *dev);
70ceb4f5
YH		 123#endif
124
8d0b94af
MKL		 125struct uncached_list {
126 spinlock_t lock;
127 struct list_head head;
128};
129
130static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
131
510c321b 132void rt6_uncached_list_add(struct rt6_info *rt)
8d0b94af
MKL		 133{
134 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
135
8d0b94af
MKL		 136 rt->rt6i_uncached_list = ul;
137
138 spin_lock_bh(&ul->lock);
139 list_add_tail(&rt->rt6i_uncached, &ul->head);
140 spin_unlock_bh(&ul->lock);
141}
142
510c321b 143void rt6_uncached_list_del(struct rt6_info *rt)
8d0b94af
MKL		 144{
145 if (!list_empty(&rt->rt6i_uncached)) {
146 struct uncached_list *ul = rt->rt6i_uncached_list;
81eb8447 147 struct net *net = dev_net(rt->dst.dev);
8d0b94af
MKL		 148
149 spin_lock_bh(&ul->lock);
150 list_del(&rt->rt6i_uncached);
81eb8447 151 atomic_dec(&net->ipv6.rt6_stats->fib_rt_uncache);
8d0b94af
MKL		 152 spin_unlock_bh(&ul->lock);
153 }
154}
155
156static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
157{
158 struct net_device *loopback_dev = net->loopback_dev;
159 int cpu;
160
e332bc67
EB		 161 if (dev == loopback_dev)
162 return;
163
8d0b94af
MKL		 164 for_each_possible_cpu(cpu) {
165 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
166 struct rt6_info *rt;
167
168 spin_lock_bh(&ul->lock);
169 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
170 struct inet6_dev *rt_idev = rt->rt6i_idev;
171 struct net_device *rt_dev = rt->dst.dev;
172
e332bc67 173 if (rt_idev->dev == dev) {
8d0b94af
MKL		 174 rt->rt6i_idev = in6_dev_get(loopback_dev);
175 in6_dev_put(rt_idev);
176 }
177
e332bc67 178 if (rt_dev == dev) {
8d0b94af
MKL		 179 rt->dst.dev = loopback_dev;
180 dev_hold(rt->dst.dev);
181 dev_put(rt_dev);
182 }
183 }
184 spin_unlock_bh(&ul->lock);
185 }
186}
187
f8a1b43b 188static inline const void *choose_neigh_daddr(const struct in6_addr *p,
f894cbf8
DM		 189 struct sk_buff *skb,
190 const void *daddr)
39232973 191{
a7563f34 192 if (!ipv6_addr_any(p))
39232973 193 return (const void *) p;
f894cbf8
DM		 194 else if (skb)
195 return &ipv6_hdr(skb)->daddr;
39232973
DM		 196 return daddr;
197}
198
f8a1b43b
DA		 199struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
200 struct net_device *dev,
201 struct sk_buff *skb,
202 const void *daddr)
d3aaeb38 203{
39232973
DM		 204 struct neighbour *n;
205
f8a1b43b
DA		 206 daddr = choose_neigh_daddr(gw, skb, daddr);
207 n = __ipv6_neigh_lookup(dev, daddr);
f83c7790
DM		 208 if (n)
209 return n;
7adf3246
SB		 210
211 n = neigh_create(&nd_tbl, daddr, dev);
212 return IS_ERR(n) ? NULL : n;
f8a1b43b
DA		 213}
214
215static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
216 struct sk_buff *skb,
217 const void *daddr)
218{
219 const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
220
221 return ip6_neigh_lookup(&rt->rt6i_gateway, dst->dev, skb, daddr);
f83c7790
DM		 222}
223
63fca65d
JA		 224static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
225{
226 struct net_device *dev = dst->dev;
227 struct rt6_info *rt = (struct rt6_info *)dst;
228
f8a1b43b 229 daddr = choose_neigh_daddr(&rt->rt6i_gateway, NULL, daddr);
63fca65d
JA		 230 if (!daddr)
231 return;
232 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
233 return;
234 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
235 return;
236 __ipv6_confirm_neigh(dev, daddr);
237}
238
9a7ec3a9 239static struct dst_ops ip6_dst_ops_template = {
1da177e4 240 .family = AF_INET6,
1da177e4
LT		 241 .gc = ip6_dst_gc,
242 .gc_thresh = 1024,
243 .check = ip6_dst_check,
0dbaee3b 244 .default_advmss = ip6_default_advmss,
ebb762f2 245 .mtu = ip6_mtu,
d4ead6b3 246 .cow_metrics = dst_cow_metrics_generic,
1da177e4
LT		 247 .destroy = ip6_dst_destroy,
248 .ifdown = ip6_dst_ifdown,
249 .negative_advice = ip6_negative_advice,
250 .link_failure = ip6_link_failure,
251 .update_pmtu = ip6_rt_update_pmtu,
6e157b6a 252 .redirect = rt6_do_redirect,
9f8955cc 253 .local_out = __ip6_local_out,
f8a1b43b 254 .neigh_lookup = ip6_dst_neigh_lookup,
63fca65d 255 .confirm_neigh = ip6_confirm_neigh,
1da177e4
LT		 256};
257
ebb762f2 258static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
ec831ea7 259{
618f9bc7
SK		 260 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
261
262 return mtu ? : dst->dev->mtu;
ec831ea7
RD		 263}
264
6700c270
DM		 265static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
266 struct sk_buff *skb, u32 mtu)
14e50e57
DM		 267{
268}
269
6700c270
DM		 270static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
271 struct sk_buff *skb)
b587ee3b
DM		 272{
273}
274
14e50e57
DM		 275static struct dst_ops ip6_dst_blackhole_ops = {
276 .family = AF_INET6,
14e50e57
DM		 277 .destroy = ip6_dst_destroy,
278 .check = ip6_dst_check,
ebb762f2 279 .mtu = ip6_blackhole_mtu,
214f45c9 280 .default_advmss = ip6_default_advmss,
14e50e57 281 .update_pmtu = ip6_rt_blackhole_update_pmtu,
b587ee3b 282 .redirect = ip6_rt_blackhole_redirect,
0a1f5962 283 .cow_metrics = dst_cow_metrics_generic,
f8a1b43b 284 .neigh_lookup = ip6_dst_neigh_lookup,
14e50e57
DM		 285};
286
62fa8a84 287static const u32 ip6_template_metrics[RTAX_MAX] = {
14edd87d 288 [RTAX_HOPLIMIT - 1] = 0,
62fa8a84
DM		 289};
290
8d1c802b 291static const struct fib6_info fib6_null_entry_template = {
93c2fb25
DA		 292 .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
293 .fib6_protocol = RTPROT_KERNEL,
294 .fib6_metric = ~(u32)0,
f05713e0 295 .fib6_ref = REFCOUNT_INIT(1),
421842ed
DA		 296 .fib6_type = RTN_UNREACHABLE,
297 .fib6_metrics = (struct dst_metrics *)&dst_default_metrics,
298};
299
fb0af4c7 300static const struct rt6_info ip6_null_entry_template = {
d8d1f30b
CG		 301 .dst = {
302 .__refcnt = ATOMIC_INIT(1),
303 .__use = 1,
2c20cbd7 304 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 305 .error = -ENETUNREACH,
d8d1f30b
CG		 306 .input = ip6_pkt_discard,
307 .output = ip6_pkt_discard_out,
1da177e4
LT		 308 },
309 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
1da177e4
LT		 310};
311
101367c2
TG		 312#ifdef CONFIG_IPV6_MULTIPLE_TABLES
313
fb0af4c7 314static const struct rt6_info ip6_prohibit_entry_template = {
d8d1f30b
CG		 315 .dst = {
316 .__refcnt = ATOMIC_INIT(1),
317 .__use = 1,
2c20cbd7 318 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 319 .error = -EACCES,
d8d1f30b
CG		 320 .input = ip6_pkt_prohibit,
321 .output = ip6_pkt_prohibit_out,
101367c2
TG		 322 },
323 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 324};
325
fb0af4c7 326static const struct rt6_info ip6_blk_hole_entry_template = {
d8d1f30b
CG		 327 .dst = {
328 .__refcnt = ATOMIC_INIT(1),
329 .__use = 1,
2c20cbd7 330 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 331 .error = -EINVAL,
d8d1f30b 332 .input = dst_discard,
ede2059d 333 .output = dst_discard_out,
101367c2
TG		 334 },
335 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
101367c2
TG		 336};
337
338#endif
339
ebfa45f0
MKL		 340static void rt6_info_init(struct rt6_info *rt)
341{
342 struct dst_entry *dst = &rt->dst;
343
344 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
ebfa45f0
MKL		 345 INIT_LIST_HEAD(&rt->rt6i_uncached);
346}
347
1da177e4 348/* allocate dst with ip6_dst_ops */
93531c67
DA		 349struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
350 int flags)
1da177e4 351{
97bab73f 352 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
b2a9c0ed 353 1, DST_OBSOLETE_FORCE_CHK, flags);
cf911662 354
81eb8447 355 if (rt) {
ebfa45f0 356 rt6_info_init(rt);
81eb8447
WW		 357 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
358 }
8104891b 359
cf911662 360 return rt;
1da177e4 361}
9ab179d8 362EXPORT_SYMBOL(ip6_dst_alloc);
d52d3997 363
1da177e4
LT		 364static void ip6_dst_destroy(struct dst_entry *dst)
365{
366 struct rt6_info *rt = (struct rt6_info *)dst;
a68886a6 367 struct fib6_info *from;
8d0b94af 368 struct inet6_dev *idev;
1da177e4 369
1620a336 370 ip_dst_metrics_put(dst);
8d0b94af
MKL		 371 rt6_uncached_list_del(rt);
372
373 idev = rt->rt6i_idev;
38308473 374 if (idev) {
1da177e4
LT		 375 rt->rt6i_idev = NULL;
376 in6_dev_put(idev);
1ab1457c 377 }
1716a961 378
0e233874 379 from = xchg((__force struct fib6_info **)&rt->from, NULL);
93531c67 380 fib6_info_release(from);
b3419363
DM		 381}
382
1da177e4
LT		 383static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
384 int how)
385{
386 struct rt6_info *rt = (struct rt6_info *)dst;
387 struct inet6_dev *idev = rt->rt6i_idev;
5a3e55d6 388 struct net_device *loopback_dev =
c346dca1 389 dev_net(dev)->loopback_dev;
1da177e4 390
e5645f51
WW		 391 if (idev && idev->dev != loopback_dev) {
392 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
393 if (loopback_idev) {
394 rt->rt6i_idev = loopback_idev;
395 in6_dev_put(idev);
97cac082 396 }
1da177e4
LT		 397 }
398}
399
5973fb1e
MKL		 400static bool __rt6_check_expired(const struct rt6_info *rt)
401{
402 if (rt->rt6i_flags & RTF_EXPIRES)
403 return time_after(jiffies, rt->dst.expires);
404 else
405 return false;
406}
407
a50feda5 408static bool rt6_check_expired(const struct rt6_info *rt)
1da177e4 409{
a68886a6
DA		 410 struct fib6_info *from;
411
412 from = rcu_dereference(rt->from);
413
1716a961
G		 414 if (rt->rt6i_flags & RTF_EXPIRES) {
415 if (time_after(jiffies, rt->dst.expires))
a50feda5 416 return true;
a68886a6 417 } else if (from) {
1e2ea8ad 418 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
a68886a6 419 fib6_check_expired(from);
1716a961 420 }
a50feda5 421 return false;
1da177e4
LT		 422}
423
b1d40991
DA		 424void fib6_select_path(const struct net *net, struct fib6_result *res,
425 struct flowi6 *fl6, int oif, bool have_oif_match,
426 const struct sk_buff *skb, int strict)
51ebd318 427{
8d1c802b 428 struct fib6_info *sibling, *next_sibling;
b1d40991
DA		 429 struct fib6_info *match = res->f6i;
430
f88d8ea6 431 if ((!match->fib6_nsiblings && !match->nh) || have_oif_match)
b1d40991 432 goto out;
51ebd318 433
b673d6cc
JS		 434 /* We might have already computed the hash for ICMPv6 errors. In such
435 * case it will always be non-zero. Otherwise now is the time to do it.
436 */
f88d8ea6
DA		 437 if (!fl6->mp_hash &&
438 (!match->nh || nexthop_is_multipath(match->nh)))
b4bac172 439 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
b673d6cc 440
f88d8ea6
DA		 441 if (unlikely(match->nh)) {
442 nexthop_path_fib6_result(res, fl6->mp_hash);
443 return;
444 }
445
1cf844c7 446 if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
b1d40991 447 goto out;
3d709f69 448
93c2fb25
DA		 449 list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
450 fib6_siblings) {
1cf844c7 451 const struct fib6_nh *nh = sibling->fib6_nh;
5e670d84
DA		 452 int nh_upper_bound;
453
702cea56 454 nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
5e670d84 455 if (fl6->mp_hash > nh_upper_bound)
3d709f69 456 continue;
702cea56 457 if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
3d709f69
IS		 458 break;
459 match = sibling;
460 break;
461 }
462
b1d40991
DA		 463out:
464 res->f6i = match;
1cf844c7 465 res->nh = match->fib6_nh;
51ebd318
ND		 466}
467
1da177e4 468/*
66f5d6ce 469 * Route lookup. rcu_read_lock() should be held.
1da177e4
LT		 470 */
471
0c59d006
DA		 472static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
473 const struct in6_addr *saddr, int oif, int flags)
474{
475 const struct net_device *dev;
476
477 if (nh->fib_nh_flags & RTNH_F_DEAD)
478 return false;
479
480 dev = nh->fib_nh_dev;
481 if (oif) {
482 if (dev->ifindex == oif)
483 return true;
484 } else {
485 if (ipv6_chk_addr(net, saddr, dev,
486 flags & RT6_LOOKUP_F_IFACE))
487 return true;
488 }
489
490 return false;
491}
492
962b6803
DA		 493struct fib6_nh_dm_arg {
494 struct net *net;
495 const struct in6_addr *saddr;
496 int oif;
497 int flags;
498 struct fib6_nh *nh;
499};
500
501static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
502{
503 struct fib6_nh_dm_arg *arg = _arg;
504
505 arg->nh = nh;
506 return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
507 arg->flags);
508}
509
510/* returns fib6_nh from nexthop or NULL */
511static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
512 struct fib6_result *res,
513 const struct in6_addr *saddr,
514 int oif, int flags)
515{
516 struct fib6_nh_dm_arg arg = {
517 .net = net,
518 .saddr = saddr,
519 .oif = oif,
520 .flags = flags,
521 };
522
523 if (nexthop_is_blackhole(nh))
524 return NULL;
525
526 if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
527 return arg.nh;
528
529 return NULL;
530}
531
75ef7389
DA		 532static void rt6_device_match(struct net *net, struct fib6_result *res,
533 const struct in6_addr *saddr, int oif, int flags)
1da177e4 534{
75ef7389
DA		 535 struct fib6_info *f6i = res->f6i;
536 struct fib6_info *spf6i;
537 struct fib6_nh *nh;
1da177e4 538
75ef7389 539 if (!oif && ipv6_addr_any(saddr)) {
f88d8ea6
DA		 540 if (unlikely(f6i->nh)) {
541 nh = nexthop_fib6_nh(f6i->nh);
542 if (nexthop_is_blackhole(f6i->nh))
543 goto out_blackhole;
544 } else {
545 nh = f6i->fib6_nh;
546 }
7d21fec9
DA		 547 if (!(nh->fib_nh_flags & RTNH_F_DEAD))
548 goto out;
75ef7389 549 }
dd3abc4e 550
75ef7389 551 for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
962b6803
DA		 552 bool matched = false;
553
554 if (unlikely(spf6i->nh)) {
555 nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
556 oif, flags);
557 if (nh)
558 matched = true;
559 } else {
560 nh = spf6i->fib6_nh;
561 if (__rt6_device_match(net, nh, saddr, oif, flags))
562 matched = true;
563 }
564 if (matched) {
75ef7389 565 res->f6i = spf6i;
7d21fec9 566 goto out;
75ef7389 567 }
dd3abc4e 568 }
1da177e4 569
75ef7389
DA		 570 if (oif && flags & RT6_LOOKUP_F_IFACE) {
571 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 572 nh = res->f6i->fib6_nh;
7d21fec9 573 goto out;
75ef7389 574 }
8067bb8c 575
f88d8ea6
DA		 576 if (unlikely(f6i->nh)) {
577 nh = nexthop_fib6_nh(f6i->nh);
578 if (nexthop_is_blackhole(f6i->nh))
579 goto out_blackhole;
580 } else {
581 nh = f6i->fib6_nh;
582 }
583
7d21fec9 584 if (nh->fib_nh_flags & RTNH_F_DEAD) {
75ef7389 585 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 586 nh = res->f6i->fib6_nh;
75ef7389 587 }
7d21fec9
DA		 588out:
589 res->nh = nh;
590 res->fib6_type = res->f6i->fib6_type;
591 res->fib6_flags = res->f6i->fib6_flags;
f88d8ea6
DA		 592 return;
593
594out_blackhole:
595 res->fib6_flags |= RTF_REJECT;
596 res->fib6_type = RTN_BLACKHOLE;
597 res->nh = nh;
1da177e4
LT		 598}
599
27097255 600#ifdef CONFIG_IPV6_ROUTER_PREF
c2f17e82
HFS		 601struct __rt6_probe_work {
602 struct work_struct work;
603 struct in6_addr target;
604 struct net_device *dev;
605};
606
607static void rt6_probe_deferred(struct work_struct *w)
608{
609 struct in6_addr mcaddr;
610 struct __rt6_probe_work *work =
611 container_of(w, struct __rt6_probe_work, work);
612
613 addrconf_addr_solict_mult(&work->target, &mcaddr);
adc176c5 614 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
c2f17e82 615 dev_put(work->dev);
662f5533 616 kfree(work);
c2f17e82
HFS		 617}
618
cc3a86c8 619static void rt6_probe(struct fib6_nh *fib6_nh)
27097255 620{
f547fac6 621 struct __rt6_probe_work *work = NULL;
5e670d84 622 const struct in6_addr *nh_gw;
f2c31e32 623 struct neighbour *neigh;
5e670d84 624 struct net_device *dev;
f547fac6 625 struct inet6_dev *idev;
5e670d84 626
27097255
YH		 627 /*
628 * Okay, this does not seem to be appropriate
629 * for now, however, we need to check if it
630 * is really so; aka Router Reachability Probing.
631 *
632 * Router Reachability Probe MUST be rate-limited
633 * to no more than one per minute.
634 */
cc3a86c8 635 if (fib6_nh->fib_nh_gw_family)
7ff74a59 636 return;
5e670d84 637
cc3a86c8
DA		 638 nh_gw = &fib6_nh->fib_nh_gw6;
639 dev = fib6_nh->fib_nh_dev;
2152caea 640 rcu_read_lock_bh();
f547fac6 641 idev = __in6_dev_get(dev);
5e670d84 642 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
2152caea 643 if (neigh) {
8d6c31bf
MKL		 644 if (neigh->nud_state & NUD_VALID)
645 goto out;
646
2152caea 647 write_lock(&neigh->lock);
990edb42
MKL		 648 if (!(neigh->nud_state & NUD_VALID) &&
649 time_after(jiffies,
dcd1f572 650 neigh->updated + idev->cnf.rtr_probe_interval)) {
990edb42
MKL		 651 work = kmalloc(sizeof(*work), GFP_ATOMIC);
652 if (work)
653 __neigh_set_probe_once(neigh);
c2f17e82 654 }
2152caea 655 write_unlock(&neigh->lock);
cc3a86c8 656 } else if (time_after(jiffies, fib6_nh->last_probe +
f547fac6 657 idev->cnf.rtr_probe_interval)) {
990edb42 658 work = kmalloc(sizeof(*work), GFP_ATOMIC);
f2c31e32 659 }
990edb42
MKL		 660
661 if (work) {
cc3a86c8 662 fib6_nh->last_probe = jiffies;
990edb42 663 INIT_WORK(&work->work, rt6_probe_deferred);
5e670d84
DA		 664 work->target = *nh_gw;
665 dev_hold(dev);
666 work->dev = dev;
990edb42
MKL		 667 schedule_work(&work->work);
668 }
669
8d6c31bf 670out:
2152caea 671 rcu_read_unlock_bh();
27097255
YH		 672}
673#else
cc3a86c8 674static inline void rt6_probe(struct fib6_nh *fib6_nh)
27097255 675{
27097255
YH		 676}
677#endif
678
1da177e4 679/*
554cfb7e 680 * Default Router Selection (RFC 2461 6.3.6)
1da177e4 681 */
1ba9a895 682static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
1da177e4 683{
afc154e9 684 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
5e670d84 685 struct neighbour *neigh;
f2c31e32 686
145a3621 687 rcu_read_lock_bh();
1ba9a895
DA		 688 neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
689 &fib6_nh->fib_nh_gw6);
145a3621
YH		 690 if (neigh) {
691 read_lock(&neigh->lock);
554cfb7e 692 if (neigh->nud_state & NUD_VALID)
afc154e9 693 ret = RT6_NUD_SUCCEED;
398bcbeb 694#ifdef CONFIG_IPV6_ROUTER_PREF
a5a81f0b 695 else if (!(neigh->nud_state & NUD_FAILED))
afc154e9 696 ret = RT6_NUD_SUCCEED;
7e980569
JB		 697 else
698 ret = RT6_NUD_FAIL_PROBE;
398bcbeb 699#endif
145a3621 700 read_unlock(&neigh->lock);
afc154e9
HFS		 701 } else {
702 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
7e980569 703 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
a5a81f0b 704 }
145a3621
YH		 705 rcu_read_unlock_bh();
706
a5a81f0b 707 return ret;
1da177e4
LT		 708}
709
702cea56
DA		 710static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
711 int strict)
1da177e4 712{
6e1809a5
DA		 713 int m = 0;
714
715 if (!oif || nh->fib_nh_dev->ifindex == oif)
716 m = 2;
1ab1457c 717
77d16f45 718 if (!m && (strict & RT6_LOOKUP_F_IFACE))
afc154e9 719 return RT6_NUD_FAIL_HARD;
ebacaaa0 720#ifdef CONFIG_IPV6_ROUTER_PREF
702cea56 721 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
ebacaaa0 722#endif
1ba9a895 723 if ((strict & RT6_LOOKUP_F_REACHABLE) &&
702cea56 724 !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
1ba9a895 725 int n = rt6_check_neigh(nh);
afc154e9
HFS		 726 if (n < 0)
727 return n;
728 }
554cfb7e
YH		 729 return m;
730}
731
28679ed1
DA		 732static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
733 int oif, int strict, int *mpri, bool *do_rr)
554cfb7e 734{
afc154e9 735 bool match_do_rr = false;
28679ed1
DA		 736 bool rc = false;
737 int m;
35103d11 738
28679ed1 739 if (nh->fib_nh_flags & RTNH_F_DEAD)
8067bb8c
IS		 740 goto out;
741
28679ed1
DA		 742 if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
743 nh->fib_nh_flags & RTNH_F_LINKDOWN &&
d5d32e4b 744 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
35103d11 745 goto out;
f11e6659 746
28679ed1 747 m = rt6_score_route(nh, fib6_flags, oif, strict);
7e980569 748 if (m == RT6_NUD_FAIL_DO_RR) {
afc154e9
HFS		 749 match_do_rr = true;
750 m = 0; /* lowest valid score */
7e980569 751 } else if (m == RT6_NUD_FAIL_HARD) {
f11e6659 752 goto out;
afc154e9
HFS		 753 }
754
755 if (strict & RT6_LOOKUP_F_REACHABLE)
28679ed1 756 rt6_probe(nh);
f11e6659 757
7e980569 758 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
f11e6659 759 if (m > *mpri) {
afc154e9 760 *do_rr = match_do_rr;
f11e6659 761 *mpri = m;
28679ed1 762 rc = true;
f11e6659 763 }
f11e6659 764out:
28679ed1 765 return rc;
f11e6659
DM		 766}
767
17a5984e
DA		 768struct fib6_nh_frl_arg {
769 u32 flags;
770 int oif;
771 int strict;
772 int *mpri;
773 bool *do_rr;
774 struct fib6_nh *nh;
775};
776
777static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
778{
779 struct fib6_nh_frl_arg *arg = _arg;
780
781 arg->nh = nh;
782 return find_match(nh, arg->flags, arg->oif, arg->strict,
783 arg->mpri, arg->do_rr);
784}
785
b7bc4b6a 786static void __find_rr_leaf(struct fib6_info *f6i_start,
30c15f03 787 struct fib6_info *nomatch, u32 metric,
b7bc4b6a 788 struct fib6_result *res, struct fib6_info **cont,
30c15f03 789 int oif, int strict, bool *do_rr, int *mpri)
f11e6659 790{
b7bc4b6a 791 struct fib6_info *f6i;
1da177e4 792
b7bc4b6a
DA		 793 for (f6i = f6i_start;
794 f6i && f6i != nomatch;
795 f6i = rcu_dereference(f6i->fib6_next)) {
17a5984e 796 bool matched = false;
30c15f03
DA		 797 struct fib6_nh *nh;
798
b7bc4b6a
DA		 799 if (cont && f6i->fib6_metric != metric) {
800 *cont = f6i;
30c15f03 801 return;
9fbdcfaf
SK		 802 }
803
b7bc4b6a 804 if (fib6_check_expired(f6i))
28679ed1
DA		 805 continue;
806
17a5984e
DA		 807 if (unlikely(f6i->nh)) {
808 struct fib6_nh_frl_arg arg = {
809 .flags = f6i->fib6_flags,
810 .oif = oif,
811 .strict = strict,
812 .mpri = mpri,
813 .do_rr = do_rr
814 };
815
816 if (nexthop_is_blackhole(f6i->nh)) {
817 res->fib6_flags = RTF_REJECT;
818 res->fib6_type = RTN_BLACKHOLE;
819 res->f6i = f6i;
820 res->nh = nexthop_fib6_nh(f6i->nh);
821 return;
822 }
823 if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
824 &arg)) {
825 matched = true;
826 nh = arg.nh;
827 }
828 } else {
829 nh = f6i->fib6_nh;
830 if (find_match(nh, f6i->fib6_flags, oif, strict,
831 mpri, do_rr))
832 matched = true;
833 }
834 if (matched) {
b7bc4b6a
DA		 835 res->f6i = f6i;
836 res->nh = nh;
7d21fec9
DA		 837 res->fib6_flags = f6i->fib6_flags;
838 res->fib6_type = f6i->fib6_type;
b7bc4b6a 839 }
9fbdcfaf 840 }
30c15f03 841}
9fbdcfaf 842
b7bc4b6a
DA		 843static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
844 struct fib6_info *rr_head, int oif, int strict,
845 bool *do_rr, struct fib6_result *res)
30c15f03 846{
b7bc4b6a
DA		 847 u32 metric = rr_head->fib6_metric;
848 struct fib6_info *cont = NULL;
30c15f03 849 int mpri = -1;
9fbdcfaf 850
b7bc4b6a 851 __find_rr_leaf(rr_head, NULL, metric, res, &cont,
30c15f03 852 oif, strict, do_rr, &mpri);
28679ed1 853
b7bc4b6a 854 __find_rr_leaf(leaf, rr_head, metric, res, &cont,
30c15f03 855 oif, strict, do_rr, &mpri);
9fbdcfaf 856
b7bc4b6a
DA		 857 if (res->f6i || !cont)
858 return;
9fbdcfaf 859
b7bc4b6a 860 __find_rr_leaf(cont, NULL, metric, res, NULL,
30c15f03 861 oif, strict, do_rr, &mpri);
f11e6659 862}
1da177e4 863
b7bc4b6a
DA		 864static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
865 struct fib6_result *res, int strict)
f11e6659 866{
8d1c802b 867 struct fib6_info *leaf = rcu_dereference(fn->leaf);
b7bc4b6a 868 struct fib6_info *rt0;
afc154e9 869 bool do_rr = false;
17ecf590 870 int key_plen;
1da177e4 871
b7bc4b6a
DA		 872 /* make sure this function or its helpers sets f6i */
873 res->f6i = NULL;
874
421842ed 875 if (!leaf || leaf == net->ipv6.fib6_null_entry)
b7bc4b6a 876 goto out;
8d1040e8 877
66f5d6ce 878 rt0 = rcu_dereference(fn->rr_ptr);
f11e6659 879 if (!rt0)
66f5d6ce 880 rt0 = leaf;
1da177e4 881
17ecf590
WW		 882 /* Double check to make sure fn is not an intermediate node
883 * and fn->leaf does not points to its child's leaf
884 * (This might happen if all routes under fn are deleted from
885 * the tree and fib6_repair_tree() is called on the node.)
886 */
93c2fb25 887 key_plen = rt0->fib6_dst.plen;
17ecf590 888#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 889 if (rt0->fib6_src.plen)
890 key_plen = rt0->fib6_src.plen;
17ecf590
WW		 891#endif
892 if (fn->fn_bit != key_plen)
b7bc4b6a 893 goto out;
1da177e4 894
b7bc4b6a 895 find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
afc154e9 896 if (do_rr) {
8fb11a9a 897 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
f11e6659 898
554cfb7e 899 /* no entries matched; do round-robin */
93c2fb25 900 if (!next || next->fib6_metric != rt0->fib6_metric)
8d1040e8 901 next = leaf;
f11e6659 902
66f5d6ce 903 if (next != rt0) {
93c2fb25 904 spin_lock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 905 /* make sure next is not being deleted from the tree */
93c2fb25 906 if (next->fib6_node)
66f5d6ce 907 rcu_assign_pointer(fn->rr_ptr, next);
93c2fb25 908 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
66f5d6ce 909 }
1da177e4 910 }
1da177e4 911
b7bc4b6a
DA		 912out:
913 if (!res->f6i) {
914 res->f6i = net->ipv6.fib6_null_entry;
1cf844c7 915 res->nh = res->f6i->fib6_nh;
7d21fec9
DA		 916 res->fib6_flags = res->f6i->fib6_flags;
917 res->fib6_type = res->f6i->fib6_type;
b7bc4b6a 918 }
1da177e4
LT		 919}
920
85bd05de 921static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
8b9df265 922{
85bd05de
DA		 923 return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
924 res->nh->fib_nh_gw_family;
8b9df265
MKL		 925}
926
70ceb4f5
YH		 927#ifdef CONFIG_IPV6_ROUTE_INFO
928int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
b71d1d42 929 const struct in6_addr *gwaddr)
70ceb4f5 930{
c346dca1 931 struct net *net = dev_net(dev);
70ceb4f5
YH		 932 struct route_info *rinfo = (struct route_info *) opt;
933 struct in6_addr prefix_buf, *prefix;
934 unsigned int pref;
4bed72e4 935 unsigned long lifetime;
8d1c802b 936 struct fib6_info *rt;
70ceb4f5
YH		 937
938 if (len < sizeof(struct route_info)) {
939 return -EINVAL;
940 }
941
942 /* Sanity check for prefix_len and length */
943 if (rinfo->length > 3) {
944 return -EINVAL;
945 } else if (rinfo->prefix_len > 128) {
946 return -EINVAL;
947 } else if (rinfo->prefix_len > 64) {
948 if (rinfo->length < 2) {
949 return -EINVAL;
950 }
951 } else if (rinfo->prefix_len > 0) {
952 if (rinfo->length < 1) {
953 return -EINVAL;
954 }
955 }
956
957 pref = rinfo->route_pref;
958 if (pref == ICMPV6_ROUTER_PREF_INVALID)
3933fc95 959 return -EINVAL;
70ceb4f5 960
4bed72e4 961 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
70ceb4f5
YH		 962
963 if (rinfo->length == 3)
964 prefix = (struct in6_addr *)rinfo->prefix;
965 else {
966 /* this function is safe */
967 ipv6_addr_prefix(&prefix_buf,
968 (struct in6_addr *)rinfo->prefix,
969 rinfo->prefix_len);
970 prefix = &prefix_buf;
971 }
972
f104a567 973 if (rinfo->prefix_len == 0)
afb1d4b5 974 rt = rt6_get_dflt_router(net, gwaddr, dev);
f104a567
DJ		 975 else
976 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
830218c1 977 gwaddr, dev);
70ceb4f5
YH		 978
979 if (rt && !lifetime) {
afb1d4b5 980 ip6_del_rt(net, rt);
70ceb4f5
YH		 981 rt = NULL;
982 }
983
984 if (!rt && lifetime)
830218c1
DA		 985 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
986 dev, pref);
70ceb4f5 987 else if (rt)
93c2fb25
DA		 988 rt->fib6_flags = RTF_ROUTEINFO |
989 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
70ceb4f5
YH		 990
991 if (rt) {
1716a961 992 if (!addrconf_finite_timeout(lifetime))
14895687 993 fib6_clean_expires(rt);
1716a961 994 else
14895687 995 fib6_set_expires(rt, jiffies + HZ * lifetime);
1716a961 996
93531c67 997 fib6_info_release(rt);
70ceb4f5
YH		 998 }
999 return 0;
1000}
1001#endif
1002
ae90d867
DA		 1003/*
1004 * Misc support functions
1005 */
1006
1007/* called with rcu_lock held */
0d161581 1008static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
ae90d867 1009{
0d161581 1010 struct net_device *dev = res->nh->fib_nh_dev;
ae90d867 1011
7d21fec9 1012 if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
ae90d867
DA		 1013 /* for copies of local routes, dst->dev needs to be the
1014 * device if it is a master device, the master device if
1015 * device is enslaved, and the loopback as the default
1016 */
1017 if (netif_is_l3_slave(dev) &&
7d21fec9 1018 !rt6_need_strict(&res->f6i->fib6_dst.addr))
ae90d867
DA		 1019 dev = l3mdev_master_dev_rcu(dev);
1020 else if (!netif_is_l3_master(dev))
1021 dev = dev_net(dev)->loopback_dev;
1022 /* last case is netif_is_l3_master(dev) is true in which
1023 * case we want dev returned to be dev
1024 */
1025 }
1026
1027 return dev;
1028}
1029
6edb3c96
DA		 1030static const int fib6_prop[RTN_MAX + 1] = {
1031 [RTN_UNSPEC] = 0,
1032 [RTN_UNICAST] = 0,
1033 [RTN_LOCAL] = 0,
1034 [RTN_BROADCAST] = 0,
1035 [RTN_ANYCAST] = 0,
1036 [RTN_MULTICAST] = 0,
1037 [RTN_BLACKHOLE] = -EINVAL,
1038 [RTN_UNREACHABLE] = -EHOSTUNREACH,
1039 [RTN_PROHIBIT] = -EACCES,
1040 [RTN_THROW] = -EAGAIN,
1041 [RTN_NAT] = -EINVAL,
1042 [RTN_XRESOLVE] = -EINVAL,
1043};
1044
1045static int ip6_rt_type_to_error(u8 fib6_type)
1046{
1047 return fib6_prop[fib6_type];
1048}
1049
8d1c802b 1050static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
3b6761d1
DA		 1051{
1052 unsigned short flags = 0;
1053
1054 if (rt->dst_nocount)
1055 flags |= DST_NOCOUNT;
1056 if (rt->dst_nopolicy)
1057 flags |= DST_NOPOLICY;
1058 if (rt->dst_host)
1059 flags |= DST_HOST;
1060
1061 return flags;
1062}
1063
7d21fec9 1064static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
6edb3c96 1065{
7d21fec9 1066 rt->dst.error = ip6_rt_type_to_error(fib6_type);
6edb3c96 1067
7d21fec9 1068 switch (fib6_type) {
6edb3c96
DA		 1069 case RTN_BLACKHOLE:
1070 rt->dst.output = dst_discard_out;
1071 rt->dst.input = dst_discard;
1072 break;
1073 case RTN_PROHIBIT:
1074 rt->dst.output = ip6_pkt_prohibit_out;
1075 rt->dst.input = ip6_pkt_prohibit;
1076 break;
1077 case RTN_THROW:
1078 case RTN_UNREACHABLE:
1079 default:
1080 rt->dst.output = ip6_pkt_discard_out;
1081 rt->dst.input = ip6_pkt_discard;
1082 break;
1083 }
1084}
1085
0d161581 1086static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
6edb3c96 1087{
7d21fec9 1088 struct fib6_info *f6i = res->f6i;
0d161581 1089
7d21fec9
DA		 1090 if (res->fib6_flags & RTF_REJECT) {
1091 ip6_rt_init_dst_reject(rt, res->fib6_type);
6edb3c96
DA		 1092 return;
1093 }
1094
1095 rt->dst.error = 0;
1096 rt->dst.output = ip6_output;
1097
7d21fec9 1098 if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
6edb3c96 1099 rt->dst.input = ip6_input;
7d21fec9 1100 } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
6edb3c96
DA		 1101 rt->dst.input = ip6_mc_input;
1102 } else {
1103 rt->dst.input = ip6_forward;
1104 }
1105
0d161581
DA		 1106 if (res->nh->fib_nh_lws) {
1107 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
6edb3c96
DA		 1108 lwtunnel_set_redirect(&rt->dst);
1109 }
1110
1111 rt->dst.lastuse = jiffies;
1112}
1113
e873e4b9 1114/* Caller must already hold reference to @from */
8d1c802b 1115static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
ae90d867 1116{
ae90d867 1117 rt->rt6i_flags &= ~RTF_EXPIRES;
a68886a6 1118 rcu_assign_pointer(rt->from, from);
e1255ed4 1119 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
ae90d867
DA		 1120}
1121
0d161581
DA		 1122/* Caller must already hold reference to f6i in result */
1123static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
ae90d867 1124{
0d161581
DA		 1125 const struct fib6_nh *nh = res->nh;
1126 const struct net_device *dev = nh->fib_nh_dev;
1127 struct fib6_info *f6i = res->f6i;
dcd1f572 1128
0d161581 1129 ip6_rt_init_dst(rt, res);
6edb3c96 1130
0d161581 1131 rt->rt6i_dst = f6i->fib6_dst;
dcd1f572 1132 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
7d21fec9 1133 rt->rt6i_flags = res->fib6_flags;
0d161581
DA		 1134 if (nh->fib_nh_gw_family) {
1135 rt->rt6i_gateway = nh->fib_nh_gw6;
2b2450ca
DA		 1136 rt->rt6i_flags |= RTF_GATEWAY;
1137 }
0d161581 1138 rt6_set_from(rt, f6i);
ae90d867 1139#ifdef CONFIG_IPV6_SUBTREES
0d161581 1140 rt->rt6i_src = f6i->fib6_src;
ae90d867 1141#endif
ae90d867
DA		 1142}
1143
a3c00e46
MKL		 1144static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1145 struct in6_addr *saddr)
1146{
66f5d6ce 1147 struct fib6_node *pn, *sn;
a3c00e46
MKL		 1148 while (1) {
1149 if (fn->fn_flags & RTN_TL_ROOT)
1150 return NULL;
66f5d6ce
WW		 1151 pn = rcu_dereference(fn->parent);
1152 sn = FIB6_SUBTREE(pn);
1153 if (sn && sn != fn)
6454743b 1154 fn = fib6_node_lookup(sn, NULL, saddr);
a3c00e46
MKL		 1155 else
1156 fn = pn;
1157 if (fn->fn_flags & RTN_RTINFO)
1158 return fn;
1159 }
1160}
c71099ac 1161
10585b43 1162static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
d3843fe5
WW		 1163{
1164 struct rt6_info *rt = *prt;
1165
1166 if (dst_hold_safe(&rt->dst))
1167 return true;
10585b43 1168 if (net) {
d3843fe5
WW		 1169 rt = net->ipv6.ip6_null_entry;
1170 dst_hold(&rt->dst);
1171 } else {
1172 rt = NULL;
1173 }
1174 *prt = rt;
1175 return false;
1176}
1177
dec9b0e2 1178/* called with rcu_lock held */
9b6b35ab 1179static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
dec9b0e2 1180{
9b6b35ab
DA		 1181 struct net_device *dev = res->nh->fib_nh_dev;
1182 struct fib6_info *f6i = res->f6i;
1183 unsigned short flags;
dec9b0e2
DA		 1184 struct rt6_info *nrt;
1185
9b6b35ab 1186 if (!fib6_info_hold_safe(f6i))
1c87e79a 1187 goto fallback;
e873e4b9 1188
9b6b35ab 1189 flags = fib6_info_dst_flags(f6i);
93531c67 1190 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1c87e79a 1191 if (!nrt) {
9b6b35ab 1192 fib6_info_release(f6i);
1c87e79a
XL		 1193 goto fallback;
1194 }
dec9b0e2 1195
0d161581 1196 ip6_rt_copy_init(nrt, res);
1c87e79a
XL		 1197 return nrt;
1198
1199fallback:
1200 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1201 dst_hold(&nrt->dst);
dec9b0e2
DA		 1202 return nrt;
1203}
1204
8ed67789
DL		 1205static struct rt6_info *ip6_pol_route_lookup(struct net *net,
1206 struct fib6_table *table,
b75cc8f9
DA		 1207 struct flowi6 *fl6,
1208 const struct sk_buff *skb,
1209 int flags)
1da177e4 1210{
b1d40991 1211 struct fib6_result res = {};
1da177e4 1212 struct fib6_node *fn;
23fb93a4 1213 struct rt6_info *rt;
1da177e4 1214
b6cdbc85
DA		 1215 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1216 flags &= ~RT6_LOOKUP_F_IFACE;
1217
66f5d6ce 1218 rcu_read_lock();
6454743b 1219 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
c71099ac 1220restart:
b1d40991
DA		 1221 res.f6i = rcu_dereference(fn->leaf);
1222 if (!res.f6i)
1223 res.f6i = net->ipv6.fib6_null_entry;
af52a52c 1224 else
75ef7389
DA		 1225 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1226 flags);
af52a52c 1227
b1d40991 1228 if (res.f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 1229 fn = fib6_backtrack(fn, &fl6->saddr);
1230 if (fn)
1231 goto restart;
2b760fcf 1232
af52a52c
DA		 1233 rt = net->ipv6.ip6_null_entry;
1234 dst_hold(&rt->dst);
1235 goto out;
f88d8ea6
DA		 1236 } else if (res.fib6_flags & RTF_REJECT) {
1237 goto do_create;
af52a52c 1238 }
d3843fe5 1239
b1d40991
DA		 1240 fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1241 fl6->flowi6_oif != 0, skb, flags);
1242
2b760fcf 1243 /* Search through exception table */
7e4b5128 1244 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
23fb93a4 1245 if (rt) {
10585b43 1246 if (ip6_hold_safe(net, &rt))
dec9b0e2 1247 dst_use_noref(&rt->dst, jiffies);
23fb93a4 1248 } else {
f88d8ea6 1249do_create:
9b6b35ab 1250 rt = ip6_create_rt_rcu(&res);
dec9b0e2 1251 }
b811580d 1252
af52a52c 1253out:
8ff2e5b2 1254 trace_fib6_table_lookup(net, &res, table, fl6);
af52a52c 1255
66f5d6ce 1256 rcu_read_unlock();
b811580d 1257
c71099ac 1258 return rt;
c71099ac
TG		 1259}
1260
67ba4152 1261struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
b75cc8f9 1262 const struct sk_buff *skb, int flags)
ea6e574e 1263{
b75cc8f9 1264 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
ea6e574e
FW		 1265}
1266EXPORT_SYMBOL_GPL(ip6_route_lookup);
1267
9acd9f3a 1268struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
b75cc8f9
DA		 1269 const struct in6_addr *saddr, int oif,
1270 const struct sk_buff *skb, int strict)
c71099ac 1271{
4c9483b2
DM		 1272 struct flowi6 fl6 = {
1273 .flowi6_oif = oif,
1274 .daddr = *daddr,
c71099ac
TG		 1275 };
1276 struct dst_entry *dst;
77d16f45 1277 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
c71099ac 1278
adaa70bb 1279 if (saddr) {
4c9483b2 1280 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
adaa70bb
TG		 1281 flags |= RT6_LOOKUP_F_HAS_SADDR;
1282 }
1283
b75cc8f9 1284 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
c71099ac
TG		 1285 if (dst->error == 0)
1286 return (struct rt6_info *) dst;
1287
1288 dst_release(dst);
1289
1da177e4
LT		 1290 return NULL;
1291}
7159039a
YH		 1292EXPORT_SYMBOL(rt6_lookup);
1293
c71099ac 1294/* ip6_ins_rt is called with FREE table->tb6_lock.
1cfb71ee
WW		 1295 * It takes new route entry, the addition fails by any reason the
1296 * route is released.
1297 * Caller must hold dst before calling it.
1da177e4
LT		 1298 */
1299
8d1c802b 1300static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
333c4301 1301 struct netlink_ext_ack *extack)
1da177e4
LT		 1302{
1303 int err;
c71099ac 1304 struct fib6_table *table;
1da177e4 1305
93c2fb25 1306 table = rt->fib6_table;
66f5d6ce 1307 spin_lock_bh(&table->tb6_lock);
d4ead6b3 1308 err = fib6_add(&table->tb6_root, rt, info, extack);
66f5d6ce 1309 spin_unlock_bh(&table->tb6_lock);
1da177e4
LT		 1310
1311 return err;
1312}
1313
8d1c802b 1314int ip6_ins_rt(struct net *net, struct fib6_info *rt)
40e22e8f 1315{
afb1d4b5 1316 struct nl_info info = { .nl_net = net, };
e715b6d3 1317
d4ead6b3 1318 return __ip6_ins_rt(rt, &info, NULL);
40e22e8f
TG		 1319}
1320
85bd05de 1321static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
8b9df265
MKL		 1322 const struct in6_addr *daddr,
1323 const struct in6_addr *saddr)
1da177e4 1324{
85bd05de 1325 struct fib6_info *f6i = res->f6i;
4832c30d 1326 struct net_device *dev;
1da177e4
LT		 1327 struct rt6_info *rt;
1328
1329 /*
1330 * Clone the route.
1331 */
1332
85bd05de 1333 if (!fib6_info_hold_safe(f6i))
e873e4b9
WW		 1334 return NULL;
1335
0d161581 1336 dev = ip6_rt_get_dev_rcu(res);
93531c67 1337 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
e873e4b9 1338 if (!rt) {
85bd05de 1339 fib6_info_release(f6i);
83a09abd 1340 return NULL;
e873e4b9 1341 }
83a09abd 1342
0d161581 1343 ip6_rt_copy_init(rt, res);
83a09abd 1344 rt->rt6i_flags |= RTF_CACHE;
83a09abd
MKL		 1345 rt->dst.flags |= DST_HOST;
1346 rt->rt6i_dst.addr = *daddr;
1347 rt->rt6i_dst.plen = 128;
1da177e4 1348
85bd05de
DA		 1349 if (!rt6_is_gw_or_nonexthop(res)) {
1350 if (f6i->fib6_dst.plen != 128 &&
1351 ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
83a09abd 1352 rt->rt6i_flags |= RTF_ANYCAST;
1da177e4 1353#ifdef CONFIG_IPV6_SUBTREES
83a09abd
MKL		 1354 if (rt->rt6i_src.plen && saddr) {
1355 rt->rt6i_src.addr = *saddr;
1356 rt->rt6i_src.plen = 128;
8b9df265 1357 }
83a09abd 1358#endif
95a9a5ba 1359 }
1da177e4 1360
95a9a5ba
YH		 1361 return rt;
1362}
1da177e4 1363
db3fedee 1364static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
d52d3997 1365{
db3fedee
DA		 1366 struct fib6_info *f6i = res->f6i;
1367 unsigned short flags = fib6_info_dst_flags(f6i);
4832c30d 1368 struct net_device *dev;
d52d3997
MKL		 1369 struct rt6_info *pcpu_rt;
1370
db3fedee 1371 if (!fib6_info_hold_safe(f6i))
e873e4b9
WW		 1372 return NULL;
1373
4832c30d 1374 rcu_read_lock();
0d161581 1375 dev = ip6_rt_get_dev_rcu(res);
93531c67 1376 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags);
4832c30d 1377 rcu_read_unlock();
e873e4b9 1378 if (!pcpu_rt) {
db3fedee 1379 fib6_info_release(f6i);
d52d3997 1380 return NULL;
e873e4b9 1381 }
0d161581 1382 ip6_rt_copy_init(pcpu_rt, res);
d52d3997
MKL		 1383 pcpu_rt->rt6i_flags |= RTF_PCPU;
1384 return pcpu_rt;
1385}
1386
66f5d6ce 1387/* It should be called with rcu_read_lock() acquired */
db3fedee 1388static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
d52d3997 1389{
c353071a 1390 struct rt6_info *pcpu_rt;
d52d3997 1391
c353071a 1392 pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
d52d3997 1393
d4ead6b3 1394 if (pcpu_rt)
10585b43 1395 ip6_hold_safe(NULL, &pcpu_rt);
d3843fe5 1396
a73e4195
MKL		 1397 return pcpu_rt;
1398}
1399
afb1d4b5 1400static struct rt6_info *rt6_make_pcpu_route(struct net *net,
db3fedee 1401 const struct fib6_result *res)
a73e4195
MKL		 1402{
1403 struct rt6_info *pcpu_rt, *prev, **p;
d52d3997 1404
db3fedee 1405 pcpu_rt = ip6_rt_pcpu_alloc(res);
d52d3997 1406 if (!pcpu_rt) {
9c7370a1
MKL		 1407 dst_hold(&net->ipv6.ip6_null_entry->dst);
1408 return net->ipv6.ip6_null_entry;
d52d3997
MKL		 1409 }
1410
a94b9367 1411 dst_hold(&pcpu_rt->dst);
f40b6ae2 1412 p = this_cpu_ptr(res->nh->rt6i_pcpu);
a94b9367 1413 prev = cmpxchg(p, NULL, pcpu_rt);
951f788a 1414 BUG_ON(prev);
a94b9367 1415
61fb0d01
ED		 1416 if (res->f6i->fib6_destroying) {
1417 struct fib6_info *from;
1418
1419 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1420 fib6_info_release(from);
1421 }
1422
d52d3997
MKL		 1423 return pcpu_rt;
1424}
1425
35732d01
WW		 1426/* exception hash table implementation
1427 */
1428static DEFINE_SPINLOCK(rt6_exception_lock);
1429
1430/* Remove rt6_ex from hash table and free the memory
1431 * Caller must hold rt6_exception_lock
1432 */
1433static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1434 struct rt6_exception *rt6_ex)
1435{
f5b51fe8 1436 struct fib6_info *from;
b2427e67 1437 struct net *net;
81eb8447 1438
35732d01
WW		 1439 if (!bucket || !rt6_ex)
1440 return;
b2427e67
CIK		 1441
1442 net = dev_net(rt6_ex->rt6i->dst.dev);
f5b51fe8
PA		 1443 net->ipv6.rt6_stats->fib_rt_cache--;
1444
1445 /* purge completely the exception to allow releasing the held resources:
1446 * some [sk] cache may keep the dst around for unlimited time
1447 */
0e233874 1448 from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
f5b51fe8
PA		 1449 fib6_info_release(from);
1450 dst_dev_put(&rt6_ex->rt6i->dst);
1451
35732d01 1452 hlist_del_rcu(&rt6_ex->hlist);
77634cc6 1453 dst_release(&rt6_ex->rt6i->dst);
35732d01
WW		 1454 kfree_rcu(rt6_ex, rcu);
1455 WARN_ON_ONCE(!bucket->depth);
1456 bucket->depth--;
1457}
1458
1459/* Remove oldest rt6_ex in bucket and free the memory
1460 * Caller must hold rt6_exception_lock
1461 */
1462static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1463{
1464 struct rt6_exception *rt6_ex, *oldest = NULL;
1465
1466 if (!bucket)
1467 return;
1468
1469 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1470 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1471 oldest = rt6_ex;
1472 }
1473 rt6_remove_exception(bucket, oldest);
1474}
1475
1476static u32 rt6_exception_hash(const struct in6_addr *dst,
1477 const struct in6_addr *src)
1478{
1479 static u32 seed __read_mostly;
1480 u32 val;
1481
1482 net_get_random_once(&seed, sizeof(seed));
1483 val = jhash(dst, sizeof(*dst), seed);
1484
1485#ifdef CONFIG_IPV6_SUBTREES
1486 if (src)
1487 val = jhash(src, sizeof(*src), val);
1488#endif
1489 return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1490}
1491
1492/* Helper function to find the cached rt in the hash table
1493 * and update bucket pointer to point to the bucket for this
1494 * (daddr, saddr) pair
1495 * Caller must hold rt6_exception_lock
1496 */
1497static struct rt6_exception *
1498__rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1499 const struct in6_addr *daddr,
1500 const struct in6_addr *saddr)
1501{
1502 struct rt6_exception *rt6_ex;
1503 u32 hval;
1504
1505 if (!(*bucket) || !daddr)
1506 return NULL;
1507
1508 hval = rt6_exception_hash(daddr, saddr);
1509 *bucket += hval;
1510
1511 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1512 struct rt6_info *rt6 = rt6_ex->rt6i;
1513 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1514
1515#ifdef CONFIG_IPV6_SUBTREES
1516 if (matched && saddr)
1517 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1518#endif
1519 if (matched)
1520 return rt6_ex;
1521 }
1522 return NULL;
1523}
1524
1525/* Helper function to find the cached rt in the hash table
1526 * and update bucket pointer to point to the bucket for this
1527 * (daddr, saddr) pair
1528 * Caller must hold rcu_read_lock()
1529 */
1530static struct rt6_exception *
1531__rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1532 const struct in6_addr *daddr,
1533 const struct in6_addr *saddr)
1534{
1535 struct rt6_exception *rt6_ex;
1536 u32 hval;
1537
1538 WARN_ON_ONCE(!rcu_read_lock_held());
1539
1540 if (!(*bucket) || !daddr)
1541 return NULL;
1542
1543 hval = rt6_exception_hash(daddr, saddr);
1544 *bucket += hval;
1545
1546 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1547 struct rt6_info *rt6 = rt6_ex->rt6i;
1548 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1549
1550#ifdef CONFIG_IPV6_SUBTREES
1551 if (matched && saddr)
1552 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1553#endif
1554 if (matched)
1555 return rt6_ex;
1556 }
1557 return NULL;
1558}
1559
b748f260 1560static unsigned int fib6_mtu(const struct fib6_result *res)
d4ead6b3 1561{
b748f260 1562 const struct fib6_nh *nh = res->nh;
d4ead6b3
DA		 1563 unsigned int mtu;
1564
b748f260
DA		 1565 if (res->f6i->fib6_pmtu) {
1566 mtu = res->f6i->fib6_pmtu;
dcd1f572 1567 } else {
b748f260 1568 struct net_device *dev = nh->fib_nh_dev;
dcd1f572
DA		 1569 struct inet6_dev *idev;
1570
1571 rcu_read_lock();
1572 idev = __in6_dev_get(dev);
1573 mtu = idev->cnf.mtu6;
1574 rcu_read_unlock();
1575 }
1576
d4ead6b3
DA		 1577 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1578
b748f260 1579 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
d4ead6b3
DA		 1580}
1581
cc5c073a
DA		 1582#define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1583
1584/* used when the flushed bit is not relevant, only access to the bucket
1585 * (ie., all bucket users except rt6_insert_exception);
1586 *
1587 * called under rcu lock; sometimes called with rt6_exception_lock held
1588 */
1589static
1590struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1591 spinlock_t *lock)
1592{
1593 struct rt6_exception_bucket *bucket;
1594
1595 if (lock)
1596 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1597 lockdep_is_held(lock));
1598 else
1599 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1600
1601 /* remove bucket flushed bit if set */
1602 if (bucket) {
1603 unsigned long p = (unsigned long)bucket;
1604
1605 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1606 bucket = (struct rt6_exception_bucket *)p;
1607 }
1608
1609 return bucket;
1610}
1611
1612static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1613{
1614 unsigned long p = (unsigned long)bucket;
1615
1616 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1617}
1618
1619/* called with rt6_exception_lock held */
1620static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1621 spinlock_t *lock)
1622{
1623 struct rt6_exception_bucket *bucket;
1624 unsigned long p;
1625
1626 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1627 lockdep_is_held(lock));
1628
1629 p = (unsigned long)bucket;
1630 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1631 bucket = (struct rt6_exception_bucket *)p;
1632 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1633}
1634
35732d01 1635static int rt6_insert_exception(struct rt6_info *nrt,
5012f0a5 1636 const struct fib6_result *res)
35732d01 1637{
5e670d84 1638 struct net *net = dev_net(nrt->dst.dev);
35732d01 1639 struct rt6_exception_bucket *bucket;
cc5c073a 1640 struct fib6_info *f6i = res->f6i;
35732d01
WW		 1641 struct in6_addr *src_key = NULL;
1642 struct rt6_exception *rt6_ex;
cc5c073a 1643 struct fib6_nh *nh = res->nh;
35732d01
WW		 1644 int err = 0;
1645
35732d01
WW		 1646 spin_lock_bh(&rt6_exception_lock);
1647
cc5c073a
DA		 1648 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1649 lockdep_is_held(&rt6_exception_lock));
35732d01
WW		 1650 if (!bucket) {
1651 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1652 GFP_ATOMIC);
1653 if (!bucket) {
1654 err = -ENOMEM;
1655 goto out;
1656 }
cc5c073a
DA		 1657 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1658 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1659 err = -EINVAL;
1660 goto out;
35732d01
WW		 1661 }
1662
1663#ifdef CONFIG_IPV6_SUBTREES
5012f0a5 1664 /* fib6_src.plen != 0 indicates f6i is in subtree
35732d01 1665 * and exception table is indexed by a hash of
5012f0a5 1666 * both fib6_dst and fib6_src.
35732d01 1667 * Otherwise, the exception table is indexed by
5012f0a5 1668 * a hash of only fib6_dst.
35732d01 1669 */
5012f0a5 1670 if (f6i->fib6_src.plen)
35732d01
WW		 1671 src_key = &nrt->rt6i_src.addr;
1672#endif
5012f0a5 1673 /* rt6_mtu_change() might lower mtu on f6i.
f5bbe7ee 1674 * Only insert this exception route if its mtu
5012f0a5 1675 * is less than f6i's mtu value.
f5bbe7ee 1676 */
b748f260 1677 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
f5bbe7ee
WW		 1678 err = -EINVAL;
1679 goto out;
1680 }
60006a48 1681
35732d01
WW		 1682 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1683 src_key);
1684 if (rt6_ex)
1685 rt6_remove_exception(bucket, rt6_ex);
1686
1687 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1688 if (!rt6_ex) {
1689 err = -ENOMEM;
1690 goto out;
1691 }
1692 rt6_ex->rt6i = nrt;
1693 rt6_ex->stamp = jiffies;
35732d01
WW		 1694 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1695 bucket->depth++;
81eb8447 1696 net->ipv6.rt6_stats->fib_rt_cache++;
35732d01
WW		 1697
1698 if (bucket->depth > FIB6_MAX_DEPTH)
1699 rt6_exception_remove_oldest(bucket);
1700
1701out:
1702 spin_unlock_bh(&rt6_exception_lock);
1703
1704 /* Update fn->fn_sernum to invalidate all cached dst */
b886d5f2 1705 if (!err) {
5012f0a5
DA		 1706 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1707 fib6_update_sernum(net, f6i);
1708 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
b886d5f2
PA		 1709 fib6_force_start_gc(net);
1710 }
35732d01
WW		 1711
1712 return err;
1713}
1714
c0b220cf 1715static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
35732d01
WW		 1716{
1717 struct rt6_exception_bucket *bucket;
1718 struct rt6_exception *rt6_ex;
1719 struct hlist_node *tmp;
1720 int i;
1721
1722 spin_lock_bh(&rt6_exception_lock);
35732d01 1723
cc5c073a 1724 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
35732d01
WW		 1725 if (!bucket)
1726 goto out;
1727
cc5c073a
DA		 1728 /* Prevent rt6_insert_exception() to recreate the bucket list */
1729 if (!from)
1730 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1731
35732d01 1732 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
cc5c073a
DA		 1733 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1734 if (!from ||
1735 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1736 rt6_remove_exception(bucket, rt6_ex);
1737 }
1738 WARN_ON_ONCE(!from && bucket->depth);
35732d01
WW		 1739 bucket++;
1740 }
35732d01
WW		 1741out:
1742 spin_unlock_bh(&rt6_exception_lock);
1743}
1744
e659ba31
DA		 1745static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1746{
1747 struct fib6_info *f6i = arg;
1748
1749 fib6_nh_flush_exceptions(nh, f6i);
1750
1751 return 0;
1752}
1753
c0b220cf
DA		 1754void rt6_flush_exceptions(struct fib6_info *f6i)
1755{
e659ba31
DA		 1756 if (f6i->nh)
1757 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1758 f6i);
1759 else
1760 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
c0b220cf
DA		 1761}
1762
35732d01
WW		 1763/* Find cached rt in the hash table inside passed in rt
1764 * Caller has to hold rcu_read_lock()
1765 */
7e4b5128 1766static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
510e2ced
WW		 1767 const struct in6_addr *daddr,
1768 const struct in6_addr *saddr)
35732d01 1769{
510e2ced 1770 const struct in6_addr *src_key = NULL;
35732d01 1771 struct rt6_exception_bucket *bucket;
35732d01 1772 struct rt6_exception *rt6_ex;
7e4b5128 1773 struct rt6_info *ret = NULL;
35732d01 1774
35732d01 1775#ifdef CONFIG_IPV6_SUBTREES
7e4b5128 1776 /* fib6i_src.plen != 0 indicates f6i is in subtree
35732d01 1777 * and exception table is indexed by a hash of
7e4b5128 1778 * both fib6_dst and fib6_src.
510e2ced
WW		 1779 * However, the src addr used to create the hash
1780 * might not be exactly the passed in saddr which
1781 * is a /128 addr from the flow.
1782 * So we need to use f6i->fib6_src to redo lookup
1783 * if the passed in saddr does not find anything.
1784 * (See the logic in ip6_rt_cache_alloc() on how
1785 * rt->rt6i_src is updated.)
35732d01 1786 */
7e4b5128 1787 if (res->f6i->fib6_src.plen)
35732d01 1788 src_key = saddr;
510e2ced 1789find_ex:
35732d01 1790#endif
cc5c073a 1791 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
35732d01
WW		 1792 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1793
1794 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
7e4b5128 1795 ret = rt6_ex->rt6i;
35732d01 1796
510e2ced
WW		 1797#ifdef CONFIG_IPV6_SUBTREES
1798 /* Use fib6_src as src_key and redo lookup */
1799 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1800 src_key = &res->f6i->fib6_src.addr;
1801 goto find_ex;
1802 }
1803#endif
1804
7e4b5128 1805 return ret;
35732d01
WW		 1806}
1807
1808/* Remove the passed in cached rt from the hash table that contains it */
cc5c073a 1809static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1810 const struct rt6_info *rt)
35732d01 1811{
c0b220cf 1812 const struct in6_addr *src_key = NULL;
35732d01 1813 struct rt6_exception_bucket *bucket;
35732d01
WW		 1814 struct rt6_exception *rt6_ex;
1815 int err;
1816
cc5c073a 1817 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
35732d01
WW		 1818 return -ENOENT;
1819
1820 spin_lock_bh(&rt6_exception_lock);
cc5c073a
DA		 1821 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1822
35732d01 1823#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1824 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1825 * and exception table is indexed by a hash of
1826 * both rt6i_dst and rt6i_src.
35732d01
WW		 1827 * Otherwise, the exception table is indexed by
1828 * a hash of only rt6i_dst.
1829 */
c0b220cf 1830 if (plen)
35732d01
WW		 1831 src_key = &rt->rt6i_src.addr;
1832#endif
1833 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1834 &rt->rt6i_dst.addr,
1835 src_key);
1836 if (rt6_ex) {
1837 rt6_remove_exception(bucket, rt6_ex);
1838 err = 0;
1839 } else {
1840 err = -ENOENT;
1841 }
1842
1843 spin_unlock_bh(&rt6_exception_lock);
1844 return err;
1845}
1846
e659ba31
DA		 1847struct fib6_nh_excptn_arg {
1848 struct rt6_info *rt;
1849 int plen;
1850};
1851
1852static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1853{
1854 struct fib6_nh_excptn_arg *arg = _arg;
1855 int err;
1856
1857 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1858 if (err == 0)
1859 return 1;
1860
1861 return 0;
1862}
1863
c0b220cf
DA		 1864static int rt6_remove_exception_rt(struct rt6_info *rt)
1865{
1866 struct fib6_info *from;
1867
1868 from = rcu_dereference(rt->from);
cc5c073a 1869 if (!from || !(rt->rt6i_flags & RTF_CACHE))
c0b220cf
DA		 1870 return -EINVAL;
1871
e659ba31
DA		 1872 if (from->nh) {
1873 struct fib6_nh_excptn_arg arg = {
1874 .rt = rt,
1875 .plen = from->fib6_src.plen
1876 };
1877 int rc;
1878
1879 /* rc = 1 means an entry was found */
1880 rc = nexthop_for_each_fib6_nh(from->nh,
1881 rt6_nh_remove_exception_rt,
1882 &arg);
1883 return rc ? 0 : -ENOENT;
1884 }
1885
1cf844c7 1886 return fib6_nh_remove_exception(from->fib6_nh,
cc5c073a 1887 from->fib6_src.plen, rt);
c0b220cf
DA		 1888}
1889
35732d01
WW		 1890/* Find rt6_ex which contains the passed in rt cache and
1891 * refresh its stamp
1892 */
cc5c073a 1893static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
c0b220cf 1894 const struct rt6_info *rt)
35732d01 1895{
c0b220cf 1896 const struct in6_addr *src_key = NULL;
35732d01 1897 struct rt6_exception_bucket *bucket;
35732d01 1898 struct rt6_exception *rt6_ex;
193f3685 1899
cc5c073a 1900 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
35732d01 1901#ifdef CONFIG_IPV6_SUBTREES
cc5c073a
DA		 1902 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1903 * and exception table is indexed by a hash of
1904 * both rt6i_dst and rt6i_src.
35732d01
WW		 1905 * Otherwise, the exception table is indexed by
1906 * a hash of only rt6i_dst.
1907 */
c0b220cf 1908 if (plen)
35732d01
WW		 1909 src_key = &rt->rt6i_src.addr;
1910#endif
cc5c073a 1911 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
35732d01
WW		 1912 if (rt6_ex)
1913 rt6_ex->stamp = jiffies;
c0b220cf
DA		 1914}
1915
e659ba31
DA		 1916struct fib6_nh_match_arg {
1917 const struct net_device *dev;
1918 const struct in6_addr *gw;
1919 struct fib6_nh *match;
1920};
1921
1922/* determine if fib6_nh has given device and gateway */
1923static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1924{
1925 struct fib6_nh_match_arg *arg = _arg;
1926
1927 if (arg->dev != nh->fib_nh_dev ||
1928 (arg->gw && !nh->fib_nh_gw_family) ||
1929 (!arg->gw && nh->fib_nh_gw_family) ||
1930 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1931 return 0;
1932
1933 arg->match = nh;
1934
1935 /* found a match, break the loop */
1936 return 1;
1937}
1938
c0b220cf
DA		 1939static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1940{
1941 struct fib6_info *from;
e659ba31 1942 struct fib6_nh *fib6_nh;
35732d01 1943
c0b220cf
DA		 1944 rcu_read_lock();
1945
1946 from = rcu_dereference(rt->from);
1947 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1948 goto unlock;
1949
e659ba31
DA		 1950 if (from->nh) {
1951 struct fib6_nh_match_arg arg = {
1952 .dev = rt->dst.dev,
1953 .gw = &rt->rt6i_gateway,
1954 };
1955
1956 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1957
1958 if (!arg.match)
1959 return;
1960 fib6_nh = arg.match;
1961 } else {
1962 fib6_nh = from->fib6_nh;
1963 }
1964 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
193f3685 1965unlock:
35732d01
WW		 1966 rcu_read_unlock();
1967}
1968
e9fa1495
SB		 1969static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1970 struct rt6_info *rt, int mtu)
1971{
1972 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1973 * lowest MTU in the path: always allow updating the route PMTU to
1974 * reflect PMTU decreases.
1975 *
1976 * If the new MTU is higher, and the route PMTU is equal to the local
1977 * MTU, this means the old MTU is the lowest in the path, so allow
1978 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1979 * handle this.
1980 */
1981
1982 if (dst_mtu(&rt->dst) >= mtu)
1983 return true;
1984
1985 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
1986 return true;
1987
1988 return false;
1989}
1990
1991static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
cc5c073a 1992 const struct fib6_nh *nh, int mtu)
f5bbe7ee
WW		 1993{
1994 struct rt6_exception_bucket *bucket;
1995 struct rt6_exception *rt6_ex;
1996 int i;
1997
cc5c073a 1998 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
e9fa1495
SB		 1999 if (!bucket)
2000 return;
2001
2002 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2003 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2004 struct rt6_info *entry = rt6_ex->rt6i;
2005
2006 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
d4ead6b3 2007 * route), the metrics of its rt->from have already
e9fa1495
SB		 2008 * been updated.
2009 */
d4ead6b3 2010 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
e9fa1495 2011 rt6_mtu_change_route_allowed(idev, entry, mtu))
d4ead6b3 2012 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
f5bbe7ee 2013 }
e9fa1495 2014 bucket++;
f5bbe7ee
WW		 2015 }
2016}
2017
b16cb459
WW		 2018#define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2019
cc5c073a
DA		 2020static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2021 const struct in6_addr *gateway)
b16cb459
WW		 2022{
2023 struct rt6_exception_bucket *bucket;
2024 struct rt6_exception *rt6_ex;
2025 struct hlist_node *tmp;
2026 int i;
2027
cc5c073a 2028 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
b16cb459
WW		 2029 return;
2030
2031 spin_lock_bh(&rt6_exception_lock);
cc5c073a 2032 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
b16cb459
WW		 2033 if (bucket) {
2034 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2035 hlist_for_each_entry_safe(rt6_ex, tmp,
2036 &bucket->chain, hlist) {
2037 struct rt6_info *entry = rt6_ex->rt6i;
2038
2039 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2040 RTF_CACHE_GATEWAY &&
2041 ipv6_addr_equal(gateway,
2042 &entry->rt6i_gateway)) {
2043 rt6_remove_exception(bucket, rt6_ex);
2044 }
2045 }
2046 bucket++;
2047 }
2048 }
2049
2050 spin_unlock_bh(&rt6_exception_lock);
2051}
2052
c757faa8
WW		 2053static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2054 struct rt6_exception *rt6_ex,
2055 struct fib6_gc_args *gc_args,
2056 unsigned long now)
2057{
2058 struct rt6_info *rt = rt6_ex->rt6i;
2059
1859bac0
PA		 2060 /* we are pruning and obsoleting aged-out and non gateway exceptions
2061 * even if others have still references to them, so that on next
2062 * dst_check() such references can be dropped.
2063 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2064 * expired, independently from their aging, as per RFC 8201 section 4
2065 */
31afeb42
WW		 2066 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2067 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2068 RT6_TRACE("aging clone %p\n", rt);
2069 rt6_remove_exception(bucket, rt6_ex);
2070 return;
2071 }
2072 } else if (time_after(jiffies, rt->dst.expires)) {
2073 RT6_TRACE("purging expired route %p\n", rt);
c757faa8
WW		 2074 rt6_remove_exception(bucket, rt6_ex);
2075 return;
31afeb42
WW		 2076 }
2077
2078 if (rt->rt6i_flags & RTF_GATEWAY) {
c757faa8
WW		 2079 struct neighbour *neigh;
2080 __u8 neigh_flags = 0;
2081
1bfa26ff
ED		 2082 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2083 if (neigh)
c757faa8 2084 neigh_flags = neigh->flags;
1bfa26ff 2085
c757faa8
WW		 2086 if (!(neigh_flags & NTF_ROUTER)) {
2087 RT6_TRACE("purging route %p via non-router but gateway\n",
2088 rt);
2089 rt6_remove_exception(bucket, rt6_ex);
2090 return;
2091 }
2092 }
31afeb42 2093
c757faa8
WW		 2094 gc_args->more++;
2095}
2096
cc5c073a 2097static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
c0b220cf
DA		 2098 struct fib6_gc_args *gc_args,
2099 unsigned long now)
c757faa8
WW		 2100{
2101 struct rt6_exception_bucket *bucket;
2102 struct rt6_exception *rt6_ex;
2103 struct hlist_node *tmp;
2104 int i;
2105
cc5c073a 2106 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
c757faa8
WW		 2107 return;
2108
1bfa26ff
ED		 2109 rcu_read_lock_bh();
2110 spin_lock(&rt6_exception_lock);
cc5c073a 2111 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
c757faa8
WW		 2112 if (bucket) {
2113 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2114 hlist_for_each_entry_safe(rt6_ex, tmp,
2115 &bucket->chain, hlist) {
2116 rt6_age_examine_exception(bucket, rt6_ex,
2117 gc_args, now);
2118 }
2119 bucket++;
2120 }
2121 }
1bfa26ff
ED		 2122 spin_unlock(&rt6_exception_lock);
2123 rcu_read_unlock_bh();
c757faa8
WW		 2124}
2125
e659ba31
DA		 2126struct fib6_nh_age_excptn_arg {
2127 struct fib6_gc_args *gc_args;
2128 unsigned long now;
2129};
2130
2131static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2132{
2133 struct fib6_nh_age_excptn_arg *arg = _arg;
2134
2135 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2136 return 0;
2137}
2138
cc5c073a 2139void rt6_age_exceptions(struct fib6_info *f6i,
c0b220cf
DA		 2140 struct fib6_gc_args *gc_args,
2141 unsigned long now)
2142{
e659ba31
DA		 2143 if (f6i->nh) {
2144 struct fib6_nh_age_excptn_arg arg = {
2145 .gc_args = gc_args,
2146 .now = now
2147 };
2148
2149 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2150 &arg);
2151 } else {
2152 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2153 }
c0b220cf
DA		 2154}
2155
1d053da9 2156/* must be called with rcu lock held */
effda4dd
DA		 2157int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2158 struct flowi6 *fl6, struct fib6_result *res, int strict)
1da177e4 2159{
367efcb9 2160 struct fib6_node *fn, *saved_fn;
1da177e4 2161
6454743b 2162 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
367efcb9 2163 saved_fn = fn;
1da177e4 2164
ca254490
DA		 2165 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2166 oif = 0;
2167
a3c00e46 2168redo_rt6_select:
effda4dd
DA		 2169 rt6_select(net, fn, oif, res, strict);
2170 if (res->f6i == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2171 fn = fib6_backtrack(fn, &fl6->saddr);
2172 if (fn)
2173 goto redo_rt6_select;
367efcb9
MKL		 2174 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2175 /* also consider unreachable route */
2176 strict &= ~RT6_LOOKUP_F_REACHABLE;
2177 fn = saved_fn;
2178 goto redo_rt6_select;
367efcb9 2179 }
a3c00e46
MKL		 2180 }
2181
effda4dd 2182 trace_fib6_table_lookup(net, res, table, fl6);
fb9de91e 2183
effda4dd 2184 return 0;
1d053da9
DA		 2185}
2186
2187struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2188 int oif, struct flowi6 *fl6,
2189 const struct sk_buff *skb, int flags)
2190{
b1d40991 2191 struct fib6_result res = {};
1d053da9
DA		 2192 struct rt6_info *rt;
2193 int strict = 0;
2194
2195 strict |= flags & RT6_LOOKUP_F_IFACE;
2196 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2197 if (net->ipv6.devconf_all->forwarding == 0)
2198 strict |= RT6_LOOKUP_F_REACHABLE;
2199
2200 rcu_read_lock();
2201
effda4dd 2202 fib6_table_lookup(net, table, oif, fl6, &res, strict);
b1d40991 2203 if (res.f6i == net->ipv6.fib6_null_entry) {
421842ed 2204 rt = net->ipv6.ip6_null_entry;
66f5d6ce 2205 rcu_read_unlock();
d3843fe5 2206 dst_hold(&rt->dst);
d3843fe5 2207 return rt;
23fb93a4
DA		 2208 }
2209
b1d40991 2210 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
d83009d4 2211
23fb93a4 2212 /*Search through exception table */
7e4b5128 2213 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
23fb93a4 2214 if (rt) {
10585b43 2215 if (ip6_hold_safe(net, &rt))
d3843fe5 2216 dst_use_noref(&rt->dst, jiffies);
d4ead6b3 2217
66f5d6ce 2218 rcu_read_unlock();
d52d3997 2219 return rt;
3da59bd9 2220 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
b1d40991 2221 !res.nh->fib_nh_gw_family)) {
3da59bd9
MKL		 2222 /* Create a RTF_CACHE clone which will not be
2223 * owned by the fib6 tree. It is for the special case where
2224 * the daddr in the skb during the neighbor look-up is different
2225 * from the fl6->daddr used to look-up route here.
2226 */
3da59bd9
MKL		 2227 struct rt6_info *uncached_rt;
2228
85bd05de 2229 uncached_rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
d52d3997 2230
4d85cd0c 2231 rcu_read_unlock();
c71099ac 2232
1cfb71ee
WW		 2233 if (uncached_rt) {
2234 /* Uncached_rt's refcnt is taken during ip6_rt_cache_alloc()
2235 * No need for another dst_hold()
2236 */
8d0b94af 2237 rt6_uncached_list_add(uncached_rt);
81eb8447 2238 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
1cfb71ee 2239 } else {
3da59bd9 2240 uncached_rt = net->ipv6.ip6_null_entry;
1cfb71ee
WW		 2241 dst_hold(&uncached_rt->dst);
2242 }
b811580d 2243
3da59bd9 2244 return uncached_rt;
d52d3997
MKL		 2245 } else {
2246 /* Get a percpu copy */
2247
2248 struct rt6_info *pcpu_rt;
2249
951f788a 2250 local_bh_disable();
db3fedee 2251 pcpu_rt = rt6_get_pcpu_route(&res);
d52d3997 2252
93531c67 2253 if (!pcpu_rt)
db3fedee 2254 pcpu_rt = rt6_make_pcpu_route(net, &res);
93531c67 2255
951f788a
ED		 2256 local_bh_enable();
2257 rcu_read_unlock();
d4bea421 2258
d52d3997
MKL		 2259 return pcpu_rt;
2260 }
1da177e4 2261}
9ff74384 2262EXPORT_SYMBOL_GPL(ip6_pol_route);
1da177e4 2263
b75cc8f9
DA		 2264static struct rt6_info *ip6_pol_route_input(struct net *net,
2265 struct fib6_table *table,
2266 struct flowi6 *fl6,
2267 const struct sk_buff *skb,
2268 int flags)
4acad72d 2269{
b75cc8f9 2270 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
4acad72d
PE		 2271}
2272
d409b847
MB		 2273struct dst_entry *ip6_route_input_lookup(struct net *net,
2274 struct net_device *dev,
b75cc8f9
DA		 2275 struct flowi6 *fl6,
2276 const struct sk_buff *skb,
2277 int flags)
72331bc0
SL		 2278{
2279 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2280 flags |= RT6_LOOKUP_F_IFACE;
2281
b75cc8f9 2282 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
72331bc0 2283}
d409b847 2284EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
72331bc0 2285
23aebdac 2286static void ip6_multipath_l3_keys(const struct sk_buff *skb,
5e5d6fed
RP		 2287 struct flow_keys *keys,
2288 struct flow_keys *flkeys)
23aebdac
JS		 2289{
2290 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2291 const struct ipv6hdr *key_iph = outer_iph;
5e5d6fed 2292 struct flow_keys *_flkeys = flkeys;
23aebdac
JS		 2293 const struct ipv6hdr *inner_iph;
2294 const struct icmp6hdr *icmph;
2295 struct ipv6hdr _inner_iph;
cea67a2d 2296 struct icmp6hdr _icmph;
23aebdac
JS		 2297
2298 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2299 goto out;
2300
cea67a2d
ED		 2301 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2302 sizeof(_icmph), &_icmph);
2303 if (!icmph)
2304 goto out;
2305
23aebdac
JS		 2306 if (icmph->icmp6_type != ICMPV6_DEST_UNREACH &&
2307 icmph->icmp6_type != ICMPV6_PKT_TOOBIG &&
2308 icmph->icmp6_type != ICMPV6_TIME_EXCEED &&
2309 icmph->icmp6_type != ICMPV6_PARAMPROB)
2310 goto out;
2311
2312 inner_iph = skb_header_pointer(skb,
2313 skb_transport_offset(skb) + sizeof(*icmph),
2314 sizeof(_inner_iph), &_inner_iph);
2315 if (!inner_iph)
2316 goto out;
2317
2318 key_iph = inner_iph;
5e5d6fed 2319 _flkeys = NULL;
23aebdac 2320out:
5e5d6fed
RP		 2321 if (_flkeys) {
2322 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2323 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2324 keys->tags.flow_label = _flkeys->tags.flow_label;
2325 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2326 } else {
2327 keys->addrs.v6addrs.src = key_iph->saddr;
2328 keys->addrs.v6addrs.dst = key_iph->daddr;
fa1be7e0 2329 keys->tags.flow_label = ip6_flowlabel(key_iph);
5e5d6fed
RP		 2330 keys->basic.ip_proto = key_iph->nexthdr;
2331 }
23aebdac
JS		 2332}
2333
2334/* if skb is set it will be used and fl6 can be NULL */
b4bac172
DA		 2335u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2336 const struct sk_buff *skb, struct flow_keys *flkeys)
23aebdac
JS		 2337{
2338 struct flow_keys hash_keys;
9a2a537a 2339 u32 mhash;
23aebdac 2340
bbfa047a 2341 switch (ip6_multipath_hash_policy(net)) {
b4bac172
DA		 2342 case 0:
2343 memset(&hash_keys, 0, sizeof(hash_keys));
2344 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2345 if (skb) {
2346 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2347 } else {
2348 hash_keys.addrs.v6addrs.src = fl6->saddr;
2349 hash_keys.addrs.v6addrs.dst = fl6->daddr;
fa1be7e0 2350 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
b4bac172
DA		 2351 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2352 }
2353 break;
2354 case 1:
2355 if (skb) {
2356 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2357 struct flow_keys keys;
2358
2359 /* short-circuit if we already have L4 hash present */
2360 if (skb->l4_hash)
2361 return skb_get_hash_raw(skb) >> 1;
2362
2363 memset(&hash_keys, 0, sizeof(hash_keys));
2364
2365 if (!flkeys) {
2366 skb_flow_dissect_flow_keys(skb, &keys, flag);
2367 flkeys = &keys;
2368 }
2369 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2370 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2371 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2372 hash_keys.ports.src = flkeys->ports.src;
2373 hash_keys.ports.dst = flkeys->ports.dst;
2374 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2375 } else {
2376 memset(&hash_keys, 0, sizeof(hash_keys));
2377 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2378 hash_keys.addrs.v6addrs.src = fl6->saddr;
2379 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2380 hash_keys.ports.src = fl6->fl6_sport;
2381 hash_keys.ports.dst = fl6->fl6_dport;
2382 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2383 }
2384 break;
23aebdac 2385 }
9a2a537a 2386 mhash = flow_hash_from_keys(&hash_keys);
23aebdac 2387
9a2a537a 2388 return mhash >> 1;
23aebdac
JS		 2389}
2390
c71099ac
TG		 2391void ip6_route_input(struct sk_buff *skb)
2392{
b71d1d42 2393 const struct ipv6hdr *iph = ipv6_hdr(skb);
c346dca1 2394 struct net *net = dev_net(skb->dev);
adaa70bb 2395 int flags = RT6_LOOKUP_F_HAS_SADDR;
904af04d 2396 struct ip_tunnel_info *tun_info;
4c9483b2 2397 struct flowi6 fl6 = {
e0d56fdd 2398 .flowi6_iif = skb->dev->ifindex,
4c9483b2
DM		 2399 .daddr = iph->daddr,
2400 .saddr = iph->saddr,
6502ca52 2401 .flowlabel = ip6_flowinfo(iph),
4c9483b2
DM		 2402 .flowi6_mark = skb->mark,
2403 .flowi6_proto = iph->nexthdr,
c71099ac 2404 };
5e5d6fed 2405 struct flow_keys *flkeys = NULL, _flkeys;
adaa70bb 2406
904af04d 2407 tun_info = skb_tunnel_info(skb);
46fa062a 2408 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
904af04d 2409 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
5e5d6fed
RP		 2410
2411 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2412 flkeys = &_flkeys;
2413
23aebdac 2414 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
b4bac172 2415 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
06e9d040 2416 skb_dst_drop(skb);
b75cc8f9
DA		 2417 skb_dst_set(skb,
2418 ip6_route_input_lookup(net, skb->dev, &fl6, skb, flags));
c71099ac
TG		 2419}
2420
b75cc8f9
DA		 2421static struct rt6_info *ip6_pol_route_output(struct net *net,
2422 struct fib6_table *table,
2423 struct flowi6 *fl6,
2424 const struct sk_buff *skb,
2425 int flags)
1da177e4 2426{
b75cc8f9 2427 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
c71099ac
TG		 2428}
2429
6f21c96a
PA		 2430struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
2431 struct flowi6 *fl6, int flags)
c71099ac 2432{
d46a9d67 2433 bool any_src;
c71099ac 2434
3ede0bbc
RS		 2435 if (ipv6_addr_type(&fl6->daddr) &
2436 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
4c1feac5
DA		 2437 struct dst_entry *dst;
2438
2439 dst = l3mdev_link_scope_lookup(net, fl6);
2440 if (dst)
2441 return dst;
2442 }
ca254490 2443
1fb9489b 2444 fl6->flowi6_iif = LOOPBACK_IFINDEX;
4dc27d1c 2445
d46a9d67 2446 any_src = ipv6_addr_any(&fl6->saddr);
741a11d9 2447 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
d46a9d67 2448 (fl6->flowi6_oif && any_src))
77d16f45 2449 flags |= RT6_LOOKUP_F_IFACE;
c71099ac 2450
d46a9d67 2451 if (!any_src)
adaa70bb 2452 flags |= RT6_LOOKUP_F_HAS_SADDR;
0c9a2ac1
YH		 2453 else if (sk)
2454 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
adaa70bb 2455
b75cc8f9 2456 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
1da177e4 2457}
6f21c96a 2458EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1da177e4 2459
2774c131 2460struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
14e50e57 2461{
5c1e6aa3 2462 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1dbe3252 2463 struct net_device *loopback_dev = net->loopback_dev;
14e50e57
DM		 2464 struct dst_entry *new = NULL;
2465
1dbe3252 2466 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
62cf27e5 2467 DST_OBSOLETE_DEAD, 0);
14e50e57 2468 if (rt) {
0a1f5962 2469 rt6_info_init(rt);
81eb8447 2470 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
8104891b 2471
0a1f5962 2472 new = &rt->dst;
14e50e57 2473 new->__use = 1;
352e512c 2474 new->input = dst_discard;
ede2059d 2475 new->output = dst_discard_out;
14e50e57 2476
0a1f5962 2477 dst_copy_metrics(new, &ort->dst);
14e50e57 2478
1dbe3252 2479 rt->rt6i_idev = in6_dev_get(loopback_dev);
4e3fd7a0 2480 rt->rt6i_gateway = ort->rt6i_gateway;
0a1f5962 2481 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
14e50e57
DM		 2482
2483 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2484#ifdef CONFIG_IPV6_SUBTREES
2485 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2486#endif
14e50e57
DM		 2487 }
2488
69ead7af
DM		 2489 dst_release(dst_orig);
2490 return new ? new : ERR_PTR(-ENOMEM);
14e50e57 2491}
14e50e57 2492
1da177e4
LT		 2493/*
2494 * Destination cache support functions
2495 */
2496
8d1c802b 2497static bool fib6_check(struct fib6_info *f6i, u32 cookie)
4b32b5ad 2498{
93531c67
DA		 2499 u32 rt_cookie = 0;
2500
8ae86971 2501 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
93531c67
DA		 2502 return false;
2503
2504 if (fib6_check_expired(f6i))
2505 return false;
2506
2507 return true;
4b32b5ad
MKL		 2508}
2509
a68886a6
DA		 2510static struct dst_entry *rt6_check(struct rt6_info *rt,
2511 struct fib6_info *from,
2512 u32 cookie)
3da59bd9 2513{
36143645 2514 u32 rt_cookie = 0;
c5cff856 2515
a68886a6 2516 if ((from && !fib6_get_cookie_safe(from, &rt_cookie)) ||
93531c67 2517 rt_cookie != cookie)
3da59bd9
MKL		 2518 return NULL;
2519
2520 if (rt6_check_expired(rt))
2521 return NULL;
2522
2523 return &rt->dst;
2524}
2525
a68886a6
DA		 2526static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2527 struct fib6_info *from,
2528 u32 cookie)
3da59bd9 2529{
5973fb1e
MKL		 2530 if (!__rt6_check_expired(rt) &&
2531 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
a68886a6 2532 fib6_check(from, cookie))
3da59bd9
MKL		 2533 return &rt->dst;
2534 else
2535 return NULL;
2536}
2537
1da177e4
LT		 2538static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
2539{
a87b7dc9 2540 struct dst_entry *dst_ret;
a68886a6 2541 struct fib6_info *from;
1da177e4
LT		 2542 struct rt6_info *rt;
2543
a87b7dc9
DA		 2544 rt = container_of(dst, struct rt6_info, dst);
2545
2546 rcu_read_lock();
1da177e4 2547
6f3118b5
ND		 2548 /* All IPV6 dsts are created with ->obsolete set to the value
2549 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2550 * into this function always.
2551 */
e3bc10bd 2552
a68886a6 2553 from = rcu_dereference(rt->from);
4b32b5ad 2554
a68886a6
DA		 2555 if (from && (rt->rt6i_flags & RTF_PCPU ||
2556 unlikely(!list_empty(&rt->rt6i_uncached))))
2557 dst_ret = rt6_dst_from_check(rt, from, cookie);
3da59bd9 2558 else
a68886a6 2559 dst_ret = rt6_check(rt, from, cookie);
a87b7dc9
DA		 2560
2561 rcu_read_unlock();
2562
2563 return dst_ret;
1da177e4
LT		 2564}
2565
2566static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2567{
2568 struct rt6_info *rt = (struct rt6_info *) dst;
2569
2570 if (rt) {
54c1a859 2571 if (rt->rt6i_flags & RTF_CACHE) {
c3c14da0 2572 rcu_read_lock();
54c1a859 2573 if (rt6_check_expired(rt)) {
93531c67 2574 rt6_remove_exception_rt(rt);
54c1a859
YH		 2575 dst = NULL;
2576 }
c3c14da0 2577 rcu_read_unlock();
54c1a859 2578 } else {
1da177e4 2579 dst_release(dst);
54c1a859
YH		 2580 dst = NULL;
2581 }
1da177e4 2582 }
54c1a859 2583 return dst;
1da177e4
LT		 2584}
2585
2586static void ip6_link_failure(struct sk_buff *skb)
2587{
2588 struct rt6_info *rt;
2589
3ffe533c 2590 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1da177e4 2591
adf30907 2592 rt = (struct rt6_info *) skb_dst(skb);
1da177e4 2593 if (rt) {
8a14e46f 2594 rcu_read_lock();
1eb4f758 2595 if (rt->rt6i_flags & RTF_CACHE) {
761f6026 2596 rt6_remove_exception_rt(rt);
c5cff856 2597 } else {
a68886a6 2598 struct fib6_info *from;
c5cff856
WW		 2599 struct fib6_node *fn;
2600
a68886a6
DA		 2601 from = rcu_dereference(rt->from);
2602 if (from) {
2603 fn = rcu_dereference(from->fib6_node);
2604 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2605 fn->fn_sernum = -1;
2606 }
1eb4f758 2607 }
8a14e46f 2608 rcu_read_unlock();
1da177e4
LT		 2609 }
2610}
2611
6a3e030f
DA		 2612static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2613{
a68886a6
DA		 2614 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2615 struct fib6_info *from;
2616
2617 rcu_read_lock();
2618 from = rcu_dereference(rt0->from);
2619 if (from)
2620 rt0->dst.expires = from->expires;
2621 rcu_read_unlock();
2622 }
6a3e030f
DA		 2623
2624 dst_set_expires(&rt0->dst, timeout);
2625 rt0->rt6i_flags |= RTF_EXPIRES;
2626}
2627
45e4fd26
MKL		 2628static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2629{
2630 struct net *net = dev_net(rt->dst.dev);
2631
d4ead6b3 2632 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
45e4fd26 2633 rt->rt6i_flags |= RTF_MODIFIED;
45e4fd26
MKL		 2634 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2635}
2636
0d3f6d29
MKL		 2637static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2638{
2639 return !(rt->rt6i_flags & RTF_CACHE) &&
1490ed2a 2640 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
0d3f6d29
MKL		 2641}
2642
45e4fd26
MKL		 2643static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2644 const struct ipv6hdr *iph, u32 mtu)
1da177e4 2645{
0dec879f 2646 const struct in6_addr *daddr, *saddr;
67ba4152 2647 struct rt6_info *rt6 = (struct rt6_info *)dst;
1da177e4 2648
19bda36c
XL		 2649 if (dst_metric_locked(dst, RTAX_MTU))
2650 return;
2651
0dec879f
JA		 2652 if (iph) {
2653 daddr = &iph->daddr;
2654 saddr = &iph->saddr;
2655 } else if (sk) {
2656 daddr = &sk->sk_v6_daddr;
2657 saddr = &inet6_sk(sk)->saddr;
2658 } else {
2659 daddr = NULL;
2660 saddr = NULL;
2661 }
2662 dst_confirm_neigh(dst, daddr);
45e4fd26
MKL		 2663 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
2664 if (mtu >= dst_mtu(dst))
2665 return;
9d289715 2666
0d3f6d29 2667 if (!rt6_cache_allowed_for_pmtu(rt6)) {
45e4fd26 2668 rt6_do_update_pmtu(rt6, mtu);
2b760fcf
WW		 2669 /* update rt6_ex->stamp for cache */
2670 if (rt6->rt6i_flags & RTF_CACHE)
2671 rt6_update_exception_stamp_rt(rt6);
0dec879f 2672 } else if (daddr) {
85bd05de 2673 struct fib6_result res = {};
45e4fd26
MKL		 2674 struct rt6_info *nrt6;
2675
4d85cd0c 2676 rcu_read_lock();
85bd05de
DA		 2677 res.f6i = rcu_dereference(rt6->from);
2678 if (!res.f6i) {
9c69a132
JL		 2679 rcu_read_unlock();
2680 return;
2681 }
1cf844c7 2682 res.nh = res.f6i->fib6_nh;
7d21fec9
DA		 2683 res.fib6_flags = res.f6i->fib6_flags;
2684 res.fib6_type = res.f6i->fib6_type;
2685
85bd05de 2686 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
45e4fd26
MKL		 2687 if (nrt6) {
2688 rt6_do_update_pmtu(nrt6, mtu);
5012f0a5 2689 if (rt6_insert_exception(nrt6, &res))
2b760fcf 2690 dst_release_immediate(&nrt6->dst);
45e4fd26 2691 }
a68886a6 2692 rcu_read_unlock();
1da177e4
LT		 2693 }
2694}
2695
45e4fd26
MKL		 2696static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2697 struct sk_buff *skb, u32 mtu)
2698{
2699 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
2700}
2701
42ae66c8 2702void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
e2d118a1 2703 int oif, u32 mark, kuid_t uid)
81aded24
DM		 2704{
2705 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2706 struct dst_entry *dst;
dc92095d
MŻ		 2707 struct flowi6 fl6 = {
2708 .flowi6_oif = oif,
2709 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2710 .daddr = iph->daddr,
2711 .saddr = iph->saddr,
2712 .flowlabel = ip6_flowinfo(iph),
2713 .flowi6_uid = uid,
2714 };
81aded24
DM		 2715
2716 dst = ip6_route_output(net, NULL, &fl6);
2717 if (!dst->error)
45e4fd26 2718 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
81aded24
DM		 2719 dst_release(dst);
2720}
2721EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2722
2723void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2724{
7ddacfa5 2725 int oif = sk->sk_bound_dev_if;
33c162a9
MKL		 2726 struct dst_entry *dst;
2727
7ddacfa5
DA		 2728 if (!oif && skb->dev)
2729 oif = l3mdev_master_ifindex(skb->dev);
2730
2731 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
33c162a9
MKL		 2732
2733 dst = __sk_dst_get(sk);
2734 if (!dst || !dst->obsolete ||
2735 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2736 return;
2737
2738 bh_lock_sock(sk);
2739 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2740 ip6_datagram_dst_update(sk, false);
2741 bh_unlock_sock(sk);
81aded24
DM		 2742}
2743EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2744
7d6850f7
AK		 2745void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2746 const struct flowi6 *fl6)
2747{
2748#ifdef CONFIG_IPV6_SUBTREES
2749 struct ipv6_pinfo *np = inet6_sk(sk);
2750#endif
2751
2752 ip6_dst_store(sk, dst,
2753 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2754 &sk->sk_v6_daddr : NULL,
2755#ifdef CONFIG_IPV6_SUBTREES
2756 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2757 &np->saddr :
2758#endif
2759 NULL);
2760}
2761
9b6b35ab 2762static bool ip6_redirect_nh_match(const struct fib6_result *res,
0b34eb00
DA		 2763 struct flowi6 *fl6,
2764 const struct in6_addr *gw,
2765 struct rt6_info **ret)
2766{
9b6b35ab
DA		 2767 const struct fib6_nh *nh = res->nh;
2768
0b34eb00
DA		 2769 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2770 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2771 return false;
2772
2773 /* rt_cache's gateway might be different from its 'parent'
2774 * in the case of an ip redirect.
2775 * So we keep searching in the exception table if the gateway
2776 * is different.
2777 */
2778 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
2779 struct rt6_info *rt_cache;
2780
9b6b35ab 2781 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
0b34eb00
DA		 2782 if (rt_cache &&
2783 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
2784 *ret = rt_cache;
2785 return true;
2786 }
2787 return false;
2788 }
2789 return true;
2790}
2791
c55c8988
DA		 2792struct fib6_nh_rd_arg {
2793 struct fib6_result *res;
2794 struct flowi6 *fl6;
2795 const struct in6_addr *gw;
2796 struct rt6_info **ret;
2797};
2798
2799static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
2800{
2801 struct fib6_nh_rd_arg *arg = _arg;
2802
2803 arg->res->nh = nh;
2804 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
2805}
2806
b55b76b2
DJ		 2807/* Handle redirects */
2808struct ip6rd_flowi {
2809 struct flowi6 fl6;
2810 struct in6_addr gateway;
2811};
2812
2813static struct rt6_info *__ip6_route_redirect(struct net *net,
2814 struct fib6_table *table,
2815 struct flowi6 *fl6,
b75cc8f9 2816 const struct sk_buff *skb,
b55b76b2
DJ		 2817 int flags)
2818{
2819 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
0b34eb00 2820 struct rt6_info *ret = NULL;
9b6b35ab 2821 struct fib6_result res = {};
c55c8988
DA		 2822 struct fib6_nh_rd_arg arg = {
2823 .res = &res,
2824 .fl6 = fl6,
2825 .gw = &rdfl->gateway,
2826 .ret = &ret
2827 };
8d1c802b 2828 struct fib6_info *rt;
b55b76b2
DJ		 2829 struct fib6_node *fn;
2830
31680ac2
DA		 2831 /* l3mdev_update_flow overrides oif if the device is enslaved; in
2832 * this case we must match on the real ingress device, so reset it
2833 */
2834 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2835 fl6->flowi6_oif = skb->dev->ifindex;
2836
b55b76b2 2837 /* Get the "current" route for this destination and
67c408cf 2838 * check if the redirect has come from appropriate router.
b55b76b2
DJ		 2839 *
2840 * RFC 4861 specifies that redirects should only be
2841 * accepted if they come from the nexthop to the target.
2842 * Due to the way the routes are chosen, this notion
2843 * is a bit fuzzy and one might need to check all possible
2844 * routes.
2845 */
2846
66f5d6ce 2847 rcu_read_lock();
6454743b 2848 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
b55b76b2 2849restart:
66f5d6ce 2850 for_each_fib6_node_rt_rcu(fn) {
9b6b35ab 2851 res.f6i = rt;
14895687 2852 if (fib6_check_expired(rt))
b55b76b2 2853 continue;
93c2fb25 2854 if (rt->fib6_flags & RTF_REJECT)
b55b76b2 2855 break;
c55c8988
DA		 2856 if (unlikely(rt->nh)) {
2857 if (nexthop_is_blackhole(rt->nh))
2858 continue;
2859 /* on match, res->nh is filled in and potentially ret */
2860 if (nexthop_for_each_fib6_nh(rt->nh,
2861 fib6_nh_redirect_match,
2862 &arg))
2863 goto out;
2864 } else {
2865 res.nh = rt->fib6_nh;
2866 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
2867 &ret))
2868 goto out;
2869 }
b55b76b2
DJ		 2870 }
2871
2872 if (!rt)
421842ed 2873 rt = net->ipv6.fib6_null_entry;
93c2fb25 2874 else if (rt->fib6_flags & RTF_REJECT) {
23fb93a4 2875 ret = net->ipv6.ip6_null_entry;
b0a1ba59
MKL		 2876 goto out;
2877 }
2878
421842ed 2879 if (rt == net->ipv6.fib6_null_entry) {
a3c00e46
MKL		 2880 fn = fib6_backtrack(fn, &fl6->saddr);
2881 if (fn)
2882 goto restart;
b55b76b2 2883 }
a3c00e46 2884
9b6b35ab 2885 res.f6i = rt;
1cf844c7 2886 res.nh = rt->fib6_nh;
b0a1ba59 2887out:
7d21fec9 2888 if (ret) {
10585b43 2889 ip6_hold_safe(net, &ret);
7d21fec9
DA		 2890 } else {
2891 res.fib6_flags = res.f6i->fib6_flags;
2892 res.fib6_type = res.f6i->fib6_type;
9b6b35ab 2893 ret = ip6_create_rt_rcu(&res);
7d21fec9 2894 }
b55b76b2 2895
66f5d6ce 2896 rcu_read_unlock();
b55b76b2 2897
8ff2e5b2 2898 trace_fib6_table_lookup(net, &res, table, fl6);
23fb93a4 2899 return ret;
b55b76b2
DJ		 2900};
2901
2902static struct dst_entry *ip6_route_redirect(struct net *net,
b75cc8f9
DA		 2903 const struct flowi6 *fl6,
2904 const struct sk_buff *skb,
2905 const struct in6_addr *gateway)
b55b76b2
DJ		 2906{
2907 int flags = RT6_LOOKUP_F_HAS_SADDR;
2908 struct ip6rd_flowi rdfl;
2909
2910 rdfl.fl6 = *fl6;
2911 rdfl.gateway = *gateway;
2912
b75cc8f9 2913 return fib6_rule_lookup(net, &rdfl.fl6, skb,
b55b76b2
DJ		 2914 flags, __ip6_route_redirect);
2915}
2916
e2d118a1
LC		 2917void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
2918 kuid_t uid)
3a5ad2ee
DM		 2919{
2920 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2921 struct dst_entry *dst;
1f7f10ac
MŻ		 2922 struct flowi6 fl6 = {
2923 .flowi6_iif = LOOPBACK_IFINDEX,
2924 .flowi6_oif = oif,
2925 .flowi6_mark = mark,
2926 .daddr = iph->daddr,
2927 .saddr = iph->saddr,
2928 .flowlabel = ip6_flowinfo(iph),
2929 .flowi6_uid = uid,
2930 };
3a5ad2ee 2931
b75cc8f9 2932 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
b55b76b2 2933 rt6_do_redirect(dst, NULL, skb);
3a5ad2ee
DM		 2934 dst_release(dst);
2935}
2936EXPORT_SYMBOL_GPL(ip6_redirect);
2937
d456336d 2938void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
c92a59ec
DJ		 2939{
2940 const struct ipv6hdr *iph = ipv6_hdr(skb);
2941 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
2942 struct dst_entry *dst;
0b26fb17
MŻ		 2943 struct flowi6 fl6 = {
2944 .flowi6_iif = LOOPBACK_IFINDEX,
2945 .flowi6_oif = oif,
0b26fb17
MŻ		 2946 .daddr = msg->dest,
2947 .saddr = iph->daddr,
2948 .flowi6_uid = sock_net_uid(net, NULL),
2949 };
c92a59ec 2950
b75cc8f9 2951 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
b55b76b2 2952 rt6_do_redirect(dst, NULL, skb);
c92a59ec
DJ		 2953 dst_release(dst);
2954}
2955
3a5ad2ee
DM		 2956void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
2957{
e2d118a1
LC		 2958 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
2959 sk->sk_uid);
3a5ad2ee
DM		 2960}
2961EXPORT_SYMBOL_GPL(ip6_sk_redirect);
2962
0dbaee3b 2963static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1da177e4 2964{
0dbaee3b
DM		 2965 struct net_device *dev = dst->dev;
2966 unsigned int mtu = dst_mtu(dst);
2967 struct net *net = dev_net(dev);
2968
1da177e4
LT		 2969 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
2970
5578689a
DL		 2971 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
2972 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1da177e4
LT		 2973
2974 /*
1ab1457c
YH		 2975 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
2976 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
2977 * IPV6_MAXPLEN is also valid and means: "any MSS,
1da177e4
LT		 2978 * rely only on pmtu discovery"
2979 */
2980 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
2981 mtu = IPV6_MAXPLEN;
2982 return mtu;
2983}
2984
ebb762f2 2985static unsigned int ip6_mtu(const struct dst_entry *dst)
d33e4553 2986{
d33e4553 2987 struct inet6_dev *idev;
d4ead6b3 2988 unsigned int mtu;
4b32b5ad
MKL		 2989
2990 mtu = dst_metric_raw(dst, RTAX_MTU);
618f9bc7 2991 if (mtu)
30f78d8e 2992 goto out;
618f9bc7
SK		 2993
2994 mtu = IPV6_MIN_MTU;
d33e4553
DM		 2995
2996 rcu_read_lock();
2997 idev = __in6_dev_get(dst->dev);
2998 if (idev)
2999 mtu = idev->cnf.mtu6;
3000 rcu_read_unlock();
3001
30f78d8e 3002out:
14972cbd
RP		 3003 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3004
3005 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
d33e4553
DM		 3006}
3007
901731b8
DA		 3008/* MTU selection:
3009 * 1. mtu on route is locked - use it
3010 * 2. mtu from nexthop exception
3011 * 3. mtu from egress device
3012 *
3013 * based on ip6_dst_mtu_forward and exception logic of
3014 * rt6_find_cached_rt; called with rcu_read_lock
3015 */
b748f260
DA		 3016u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3017 const struct in6_addr *daddr,
3018 const struct in6_addr *saddr)
901731b8 3019{
b748f260
DA		 3020 const struct fib6_nh *nh = res->nh;
3021 struct fib6_info *f6i = res->f6i;
901731b8 3022 struct inet6_dev *idev;
510e2ced 3023 struct rt6_info *rt;
901731b8
DA		 3024 u32 mtu = 0;
3025
3026 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3027 mtu = f6i->fib6_pmtu;
3028 if (mtu)
3029 goto out;
3030 }
3031
510e2ced
WW		 3032 rt = rt6_find_cached_rt(res, daddr, saddr);
3033 if (unlikely(rt)) {
3034 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3035 } else {
b748f260 3036 struct net_device *dev = nh->fib_nh_dev;
901731b8
DA		 3037
3038 mtu = IPV6_MIN_MTU;
3039 idev = __in6_dev_get(dev);
3040 if (idev && idev->cnf.mtu6 > mtu)
3041 mtu = idev->cnf.mtu6;
3042 }
3043
3044 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3045out:
b748f260 3046 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
901731b8
DA		 3047}
3048
3b00944c 3049struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
87a11578 3050 struct flowi6 *fl6)
1da177e4 3051{
87a11578 3052 struct dst_entry *dst;
1da177e4
LT		 3053 struct rt6_info *rt;
3054 struct inet6_dev *idev = in6_dev_get(dev);
c346dca1 3055 struct net *net = dev_net(dev);
1da177e4 3056
38308473 3057 if (unlikely(!idev))
122bdf67 3058 return ERR_PTR(-ENODEV);
1da177e4 3059
ad706862 3060 rt = ip6_dst_alloc(net, dev, 0);
38308473 3061 if (unlikely(!rt)) {
1da177e4 3062 in6_dev_put(idev);
87a11578 3063 dst = ERR_PTR(-ENOMEM);
1da177e4
LT		 3064 goto out;
3065 }
3066
8e2ec639 3067 rt->dst.flags |= DST_HOST;
588753f1 3068 rt->dst.input = ip6_input;
8e2ec639 3069 rt->dst.output = ip6_output;
550bab42 3070 rt->rt6i_gateway = fl6->daddr;
87a11578 3071 rt->rt6i_dst.addr = fl6->daddr;
8e2ec639
YZ		 3072 rt->rt6i_dst.plen = 128;
3073 rt->rt6i_idev = idev;
14edd87d 3074 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1da177e4 3075
4c981e28 3076 /* Add this dst into uncached_list so that rt6_disable_ip() can
587fea74
WW		 3077 * do proper release of the net_device
3078 */
3079 rt6_uncached_list_add(rt);
81eb8447 3080 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
1da177e4 3081
87a11578
DM		 3082 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3083
1da177e4 3084out:
87a11578 3085 return dst;
1da177e4
LT		 3086}
3087
569d3645 3088static int ip6_dst_gc(struct dst_ops *ops)
1da177e4 3089{
86393e52 3090 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
7019b78e
DL		 3091 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3092 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3093 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3094 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3095 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
fc66f95c 3096 int entries;
7019b78e 3097
fc66f95c 3098 entries = dst_entries_get_fast(ops);
49a18d86 3099 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
fc66f95c 3100 entries <= rt_max_size)
1da177e4
LT		 3101 goto out;
3102
6891a346 3103 net->ipv6.ip6_rt_gc_expire++;
14956643 3104 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
fc66f95c
ED		 3105 entries = dst_entries_get_slow(ops);
3106 if (entries < ops->gc_thresh)
7019b78e 3107 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1da177e4 3108out:
7019b78e 3109 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
fc66f95c 3110 return entries > rt_max_size;
1da177e4
LT		 3111}
3112
8c14586f
DA		 3113static struct rt6_info *ip6_nh_lookup_table(struct net *net,
3114 struct fib6_config *cfg,
f4797b33
DA		 3115 const struct in6_addr *gw_addr,
3116 u32 tbid, int flags)
8c14586f
DA		 3117{
3118 struct flowi6 fl6 = {
3119 .flowi6_oif = cfg->fc_ifindex,
3120 .daddr = *gw_addr,
3121 .saddr = cfg->fc_prefsrc,
3122 };
3123 struct fib6_table *table;
3124 struct rt6_info *rt;
8c14586f 3125
f4797b33 3126 table = fib6_get_table(net, tbid);
8c14586f
DA		 3127 if (!table)
3128 return NULL;
3129
3130 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3131 flags |= RT6_LOOKUP_F_HAS_SADDR;
3132
f4797b33 3133 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
b75cc8f9 3134 rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, NULL, flags);
8c14586f
DA		 3135
3136 /* if table lookup failed, fall back to full lookup */
3137 if (rt == net->ipv6.ip6_null_entry) {
3138 ip6_rt_put(rt);
3139 rt = NULL;
3140 }
3141
3142 return rt;
3143}
3144
fc1e64e1
DA		 3145static int ip6_route_check_nh_onlink(struct net *net,
3146 struct fib6_config *cfg,
9fbb704c 3147 const struct net_device *dev,
fc1e64e1
DA		 3148 struct netlink_ext_ack *extack)
3149{
44750f84 3150 u32 tbid = l3mdev_fib_table(dev) ? : RT_TABLE_MAIN;
fc1e64e1
DA		 3151 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3152 u32 flags = RTF_LOCAL | RTF_ANYCAST | RTF_REJECT;
bf1dc8ba 3153 struct fib6_info *from;
fc1e64e1
DA		 3154 struct rt6_info *grt;
3155 int err;
3156
3157 err = 0;
3158 grt = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0);
3159 if (grt) {
bf1dc8ba
PA		 3160 rcu_read_lock();
3161 from = rcu_dereference(grt->from);
58e354c0 3162 if (!grt->dst.error &&
4ed591c8 3163 /* ignore match if it is the default route */
bf1dc8ba 3164 from && !ipv6_addr_any(&from->fib6_dst.addr) &&
58e354c0 3165 (grt->rt6i_flags & flags || dev != grt->dst.dev)) {
44750f84
DA		 3166 NL_SET_ERR_MSG(extack,
3167 "Nexthop has invalid gateway or device mismatch");
fc1e64e1
DA		 3168 err = -EINVAL;
3169 }
bf1dc8ba 3170 rcu_read_unlock();
fc1e64e1
DA		 3171
3172 ip6_rt_put(grt);
3173 }
3174
3175 return err;
3176}
3177
1edce99f
DA		 3178static int ip6_route_check_nh(struct net *net,
3179 struct fib6_config *cfg,
3180 struct net_device **_dev,
3181 struct inet6_dev **idev)
3182{
3183 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3184 struct net_device *dev = _dev ? *_dev : NULL;
3185 struct rt6_info *grt = NULL;
3186 int err = -EHOSTUNREACH;
3187
3188 if (cfg->fc_table) {
f4797b33
DA		 3189 int flags = RT6_LOOKUP_F_IFACE;
3190
3191 grt = ip6_nh_lookup_table(net, cfg, gw_addr,
3192 cfg->fc_table, flags);
1edce99f
DA		 3193 if (grt) {
3194 if (grt->rt6i_flags & RTF_GATEWAY ||
3195 (dev && dev != grt->dst.dev)) {
3196 ip6_rt_put(grt);
3197 grt = NULL;
3198 }
3199 }
3200 }
3201
3202 if (!grt)
b75cc8f9 3203 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, NULL, 1);
1edce99f
DA		 3204
3205 if (!grt)
3206 goto out;
3207
3208 if (dev) {
3209 if (dev != grt->dst.dev) {
3210 ip6_rt_put(grt);
3211 goto out;
3212 }
3213 } else {
3214 *_dev = dev = grt->dst.dev;
3215 *idev = grt->rt6i_idev;
3216 dev_hold(dev);
3217 in6_dev_hold(grt->rt6i_idev);
3218 }
3219
3220 if (!(grt->rt6i_flags & RTF_GATEWAY))
3221 err = 0;
3222
3223 ip6_rt_put(grt);
3224
3225out:
3226 return err;
3227}
3228
9fbb704c
DA		 3229static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3230 struct net_device **_dev, struct inet6_dev **idev,
3231 struct netlink_ext_ack *extack)
3232{
3233 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3234 int gwa_type = ipv6_addr_type(gw_addr);
232378e8 3235 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
9fbb704c 3236 const struct net_device *dev = *_dev;
232378e8 3237 bool need_addr_check = !dev;
9fbb704c
DA		 3238 int err = -EINVAL;
3239
3240 /* if gw_addr is local we will fail to detect this in case
3241 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3242 * will return already-added prefix route via interface that
3243 * prefix route was assigned to, which might be non-loopback.
3244 */
232378e8
DA		 3245 if (dev &&
3246 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3247 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
9fbb704c
DA		 3248 goto out;
3249 }
3250
3251 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3252 /* IPv6 strictly inhibits using not link-local
3253 * addresses as nexthop address.
3254 * Otherwise, router will not able to send redirects.
3255 * It is very good, but in some (rare!) circumstances
3256 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3257 * some exceptions. --ANK
3258 * We allow IPv4-mapped nexthops to support RFC4798-type
3259 * addressing
3260 */
3261 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3262 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3263 goto out;
3264 }
3265
3266 if (cfg->fc_flags & RTNH_F_ONLINK)
3267 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3268 else
3269 err = ip6_route_check_nh(net, cfg, _dev, idev);
3270
3271 if (err)
3272 goto out;
3273 }
3274
3275 /* reload in case device was changed */
3276 dev = *_dev;
3277
3278 err = -EINVAL;
3279 if (!dev) {
3280 NL_SET_ERR_MSG(extack, "Egress device not specified");
3281 goto out;
3282 } else if (dev->flags & IFF_LOOPBACK) {
3283 NL_SET_ERR_MSG(extack,
3284 "Egress device can not be loopback device for this route");
3285 goto out;
3286 }
232378e8
DA		 3287
3288 /* if we did not check gw_addr above, do so now that the
3289 * egress device has been resolved.
3290 */
3291 if (need_addr_check &&
3292 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3293 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3294 goto out;
3295 }
3296
9fbb704c
DA		 3297 err = 0;
3298out:
3299 return err;
3300}
3301
83c44251
DA		 3302static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3303{
3304 if ((flags & RTF_REJECT) ||
3305 (dev && (dev->flags & IFF_LOOPBACK) &&
3306 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3307 !(flags & RTF_LOCAL)))
3308 return true;
3309
3310 return false;
3311}
3312
3313int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3314 struct fib6_config *cfg, gfp_t gfp_flags,
3315 struct netlink_ext_ack *extack)
3316{
3317 struct net_device *dev = NULL;
3318 struct inet6_dev *idev = NULL;
3319 int addr_type;
3320 int err;
3321
f1741730
DA		 3322 fib6_nh->fib_nh_family = AF_INET6;
3323
83c44251
DA		 3324 err = -ENODEV;
3325 if (cfg->fc_ifindex) {
3326 dev = dev_get_by_index(net, cfg->fc_ifindex);
3327 if (!dev)
3328 goto out;
3329 idev = in6_dev_get(dev);
3330 if (!idev)
3331 goto out;
3332 }
3333
3334 if (cfg->fc_flags & RTNH_F_ONLINK) {
3335 if (!dev) {
3336 NL_SET_ERR_MSG(extack,
3337 "Nexthop device required for onlink");
3338 goto out;
3339 }
3340
3341 if (!(dev->flags & IFF_UP)) {
3342 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3343 err = -ENETDOWN;
3344 goto out;
3345 }
3346
ad1601ae 3347 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
83c44251
DA		 3348 }
3349
ad1601ae 3350 fib6_nh->fib_nh_weight = 1;
83c44251
DA		 3351
3352 /* We cannot add true routes via loopback here,
3353 * they would result in kernel looping; promote them to reject routes
3354 */
3355 addr_type = ipv6_addr_type(&cfg->fc_dst);
3356 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3357 /* hold loopback dev/idev if we haven't done so. */
3358 if (dev != net->loopback_dev) {
3359 if (dev) {
3360 dev_put(dev);
3361 in6_dev_put(idev);
3362 }
3363 dev = net->loopback_dev;
3364 dev_hold(dev);
3365 idev = in6_dev_get(dev);
3366 if (!idev) {
3367 err = -ENODEV;
3368 goto out;
3369 }
3370 }
7dd73168 3371 goto pcpu_alloc;
83c44251
DA		 3372 }
3373
3374 if (cfg->fc_flags & RTF_GATEWAY) {
3375 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3376 if (err)
3377 goto out;
3378
ad1601ae 3379 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
bdf00467 3380 fib6_nh->fib_nh_gw_family = AF_INET6;
83c44251
DA		 3381 }
3382
3383 err = -ENODEV;
3384 if (!dev)
3385 goto out;
3386
3387 if (idev->cnf.disable_ipv6) {
3388 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3389 err = -EACCES;
3390 goto out;
3391 }
3392
3393 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3394 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3395 err = -ENETDOWN;
3396 goto out;
3397 }
3398
3399 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3400 !netif_carrier_ok(dev))
ad1601ae 3401 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
83c44251 3402
7dd73168
DA		 3403 err = fib_nh_common_init(&fib6_nh->nh_common, cfg->fc_encap,
3404 cfg->fc_encap_type, cfg, gfp_flags, extack);
3405 if (err)
3406 goto out;
3407
3408pcpu_alloc:
f40b6ae2
DA		 3409 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3410 if (!fib6_nh->rt6i_pcpu) {
3411 err = -ENOMEM;
3412 goto out;
3413 }
3414
ad1601ae 3415 fib6_nh->fib_nh_dev = dev;
f1741730 3416 fib6_nh->fib_nh_oif = dev->ifindex;
83c44251
DA		 3417 err = 0;
3418out:
3419 if (idev)
3420 in6_dev_put(idev);
3421
3422 if (err) {
ad1601ae
DA		 3423 lwtstate_put(fib6_nh->fib_nh_lws);
3424 fib6_nh->fib_nh_lws = NULL;
83c44251
DA		 3425 if (dev)
3426 dev_put(dev);
3427 }
3428
3429 return err;
3430}
3431
dac7d0f2
DA		 3432void fib6_nh_release(struct fib6_nh *fib6_nh)
3433{
cc5c073a
DA		 3434 struct rt6_exception_bucket *bucket;
3435
3436 rcu_read_lock();
3437
3438 fib6_nh_flush_exceptions(fib6_nh, NULL);
3439 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3440 if (bucket) {
3441 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3442 kfree(bucket);
3443 }
3444
3445 rcu_read_unlock();
3446
f40b6ae2
DA		 3447 if (fib6_nh->rt6i_pcpu) {
3448 int cpu;
3449
3450 for_each_possible_cpu(cpu) {
3451 struct rt6_info **ppcpu_rt;
3452 struct rt6_info *pcpu_rt;
3453
3454 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3455 pcpu_rt = *ppcpu_rt;
3456 if (pcpu_rt) {
3457 dst_dev_put(&pcpu_rt->dst);
3458 dst_release(&pcpu_rt->dst);
3459 *ppcpu_rt = NULL;
3460 }
3461 }
3462
3463 free_percpu(fib6_nh->rt6i_pcpu);
3464 }
3465
979e276e 3466 fib_nh_common_release(&fib6_nh->nh_common);
dac7d0f2
DA		 3467}
3468
8d1c802b 3469static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
acb54e3c 3470 gfp_t gfp_flags,
333c4301 3471 struct netlink_ext_ack *extack)
1da177e4 3472{
5578689a 3473 struct net *net = cfg->fc_nlinfo.nl_net;
8d1c802b 3474 struct fib6_info *rt = NULL;
f88d8ea6 3475 struct nexthop *nh = NULL;
c71099ac 3476 struct fib6_table *table;
f88d8ea6 3477 struct fib6_nh *fib6_nh;
8c5b83f0 3478 int err = -EINVAL;
83c44251 3479 int addr_type;
1da177e4 3480
557c44be 3481 /* RTF_PCPU is an internal flag; can not be set by userspace */
d5d531cb
DA		 3482 if (cfg->fc_flags & RTF_PCPU) {
3483 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
557c44be 3484 goto out;
d5d531cb 3485 }
557c44be 3486
2ea2352e
WW		 3487 /* RTF_CACHE is an internal flag; can not be set by userspace */
3488 if (cfg->fc_flags & RTF_CACHE) {
3489 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3490 goto out;
3491 }
3492
e8478e80
DA		 3493 if (cfg->fc_type > RTN_MAX) {
3494 NL_SET_ERR_MSG(extack, "Invalid route type");
3495 goto out;
3496 }
3497
d5d531cb
DA		 3498 if (cfg->fc_dst_len > 128) {
3499 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3500 goto out;
3501 }
3502 if (cfg->fc_src_len > 128) {
3503 NL_SET_ERR_MSG(extack, "Invalid source address length");
8c5b83f0 3504 goto out;
d5d531cb 3505 }
1da177e4 3506#ifndef CONFIG_IPV6_SUBTREES
d5d531cb
DA		 3507 if (cfg->fc_src_len) {
3508 NL_SET_ERR_MSG(extack,
3509 "Specifying source address requires IPV6_SUBTREES to be enabled");
8c5b83f0 3510 goto out;
d5d531cb 3511 }
1da177e4 3512#endif
fc1e64e1 3513
d71314b4 3514 err = -ENOBUFS;
38308473
DM		 3515 if (cfg->fc_nlinfo.nlh &&
3516 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
d71314b4 3517 table = fib6_get_table(net, cfg->fc_table);
38308473 3518 if (!table) {
f3213831 3519 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
d71314b4
MV		 3520 table = fib6_new_table(net, cfg->fc_table);
3521 }
3522 } else {
3523 table = fib6_new_table(net, cfg->fc_table);
3524 }
38308473
DM		 3525
3526 if (!table)
c71099ac 3527 goto out;
c71099ac 3528
93531c67 3529 err = -ENOMEM;
f88d8ea6 3530 rt = fib6_info_alloc(gfp_flags, !nh);
93531c67 3531 if (!rt)
1da177e4 3532 goto out;
93531c67 3533
d7e774f3
DA		 3534 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3535 extack);
767a2217
DA		 3536 if (IS_ERR(rt->fib6_metrics)) {
3537 err = PTR_ERR(rt->fib6_metrics);
fda21d46
ED		 3538 /* Do not leave garbage there. */
3539 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
767a2217
DA		 3540 goto out;
3541 }
3542
93531c67
DA		 3543 if (cfg->fc_flags & RTF_ADDRCONF)
3544 rt->dst_nocount = true;
1da177e4 3545
1716a961 3546 if (cfg->fc_flags & RTF_EXPIRES)
14895687 3547 fib6_set_expires(rt, jiffies +
1716a961
G		 3548 clock_t_to_jiffies(cfg->fc_expires));
3549 else
14895687 3550 fib6_clean_expires(rt);
1da177e4 3551
86872cb5
TG		 3552 if (cfg->fc_protocol == RTPROT_UNSPEC)
3553 cfg->fc_protocol = RTPROT_BOOT;
93c2fb25 3554 rt->fib6_protocol = cfg->fc_protocol;
86872cb5 3555
83c44251
DA		 3556 rt->fib6_table = table;
3557 rt->fib6_metric = cfg->fc_metric;
3558 rt->fib6_type = cfg->fc_type;
2b2450ca 3559 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
19e42e45 3560
93c2fb25
DA		 3561 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3562 rt->fib6_dst.plen = cfg->fc_dst_len;
3563 if (rt->fib6_dst.plen == 128)
3b6761d1 3564 rt->dst_host = true;
e5fd387a 3565
1da177e4 3566#ifdef CONFIG_IPV6_SUBTREES
93c2fb25
DA		 3567 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3568 rt->fib6_src.plen = cfg->fc_src_len;
1da177e4 3569#endif
f88d8ea6
DA		 3570 if (nh) {
3571 if (!nexthop_get(nh)) {
3572 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3573 goto out;
3574 }
3575 if (rt->fib6_src.plen) {
4daa95af 3576 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
f88d8ea6
DA		 3577 goto out;
3578 }
3579 rt->nh = nh;
3580 fib6_nh = nexthop_fib6_nh(rt->nh);
3581 } else {
3582 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3583 if (err)
3584 goto out;
1da177e4 3585
f88d8ea6
DA		 3586 fib6_nh = rt->fib6_nh;
3587
3588 /* We cannot add true routes via loopback here, they would
3589 * result in kernel looping; promote them to reject routes
3590 */
3591 addr_type = ipv6_addr_type(&cfg->fc_dst);
3592 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3593 addr_type))
3594 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3595 }
955ec4cb 3596
c3968a85 3597 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
f88d8ea6 3598 struct net_device *dev = fib6_nh->fib_nh_dev;
83c44251 3599
c3968a85 3600 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
d5d531cb 3601 NL_SET_ERR_MSG(extack, "Invalid source address");
c3968a85
DW		 3602 err = -EINVAL;
3603 goto out;
3604 }
93c2fb25
DA		 3605 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3606 rt->fib6_prefsrc.plen = 128;
c3968a85 3607 } else
93c2fb25 3608 rt->fib6_prefsrc.plen = 0;
c3968a85 3609
8c5b83f0 3610 return rt;
6b9ea5a6 3611out:
93531c67 3612 fib6_info_release(rt);
8c5b83f0 3613 return ERR_PTR(err);
6b9ea5a6
RP		 3614}
3615
acb54e3c 3616int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
333c4301 3617 struct netlink_ext_ack *extack)
6b9ea5a6 3618{
8d1c802b 3619 struct fib6_info *rt;
6b9ea5a6
RP		 3620 int err;
3621
acb54e3c 3622 rt = ip6_route_info_create(cfg, gfp_flags, extack);
d4ead6b3
DA		 3623 if (IS_ERR(rt))
3624 return PTR_ERR(rt);
6b9ea5a6 3625
d4ead6b3 3626 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
93531c67 3627 fib6_info_release(rt);
6b9ea5a6 3628
1da177e4
LT		 3629 return err;
3630}
3631
8d1c802b 3632static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
1da177e4 3633{
afb1d4b5 3634 struct net *net = info->nl_net;
c71099ac 3635 struct fib6_table *table;
afb1d4b5 3636 int err;
1da177e4 3637
421842ed 3638 if (rt == net->ipv6.fib6_null_entry) {
6825a26c
G		 3639 err = -ENOENT;
3640 goto out;
3641 }
6c813a72 3642
93c2fb25 3643 table = rt->fib6_table;
66f5d6ce 3644 spin_lock_bh(&table->tb6_lock);
86872cb5 3645 err = fib6_del(rt, info);
66f5d6ce 3646 spin_unlock_bh(&table->tb6_lock);
1da177e4 3647
6825a26c 3648out:
93531c67 3649 fib6_info_release(rt);
1da177e4
LT		 3650 return err;
3651}
3652
8d1c802b 3653int ip6_del_rt(struct net *net, struct fib6_info *rt)
e0a1ad73 3654{
afb1d4b5
DA		 3655 struct nl_info info = { .nl_net = net };
3656
528c4ceb 3657 return __ip6_del_rt(rt, &info);
e0a1ad73
TG		 3658}
3659
8d1c802b 3660static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
0ae81335
DA		 3661{
3662 struct nl_info *info = &cfg->fc_nlinfo;
e3330039 3663 struct net *net = info->nl_net;
16a16cd3 3664 struct sk_buff *skb = NULL;
0ae81335 3665 struct fib6_table *table;
e3330039 3666 int err = -ENOENT;
0ae81335 3667
421842ed 3668 if (rt == net->ipv6.fib6_null_entry)
e3330039 3669 goto out_put;
93c2fb25 3670 table = rt->fib6_table;
66f5d6ce 3671 spin_lock_bh(&table->tb6_lock);
0ae81335 3672
93c2fb25 3673 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
8d1c802b 3674 struct fib6_info *sibling, *next_sibling;
0ae81335 3675
16a16cd3
DA		 3676 /* prefer to send a single notification with all hops */
3677 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3678 if (skb) {
3679 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3680
d4ead6b3 3681 if (rt6_fill_node(net, skb, rt, NULL,
16a16cd3
DA		 3682 NULL, NULL, 0, RTM_DELROUTE,
3683 info->portid, seq, 0) < 0) {
3684 kfree_skb(skb);
3685 skb = NULL;
3686 } else
3687 info->skip_notify = 1;
3688 }
3689
0ae81335 3690 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25
DA		 3691 &rt->fib6_siblings,
3692 fib6_siblings) {
0ae81335
DA		 3693 err = fib6_del(sibling, info);
3694 if (err)
e3330039 3695 goto out_unlock;
0ae81335
DA		 3696 }
3697 }
3698
3699 err = fib6_del(rt, info);
e3330039 3700out_unlock:
66f5d6ce 3701 spin_unlock_bh(&table->tb6_lock);
e3330039 3702out_put:
93531c67 3703 fib6_info_release(rt);
16a16cd3
DA		 3704
3705 if (skb) {
e3330039 3706 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
16a16cd3
DA		 3707 info->nlh, gfp_any());
3708 }
0ae81335
DA		 3709 return err;
3710}
3711
0fa6efc5 3712static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
23fb93a4
DA		 3713{
3714 int rc = -ESRCH;
3715
3716 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3717 goto out;
3718
3719 if (cfg->fc_flags & RTF_GATEWAY &&
3720 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3721 goto out;
761f6026
XL		 3722
3723 rc = rt6_remove_exception_rt(rt);
23fb93a4
DA		 3724out:
3725 return rc;
3726}
3727
0fa6efc5
DA		 3728static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3729 struct fib6_nh *nh)
3730{
3731 struct fib6_result res = {
3732 .f6i = rt,
3733 .nh = nh,
3734 };
3735 struct rt6_info *rt_cache;
3736
3737 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3738 if (rt_cache)
3739 return __ip6_del_cached_rt(rt_cache, cfg);
3740
3741 return 0;
3742}
3743
333c4301
DA		 3744static int ip6_route_del(struct fib6_config *cfg,
3745 struct netlink_ext_ack *extack)
1da177e4 3746{
c71099ac 3747 struct fib6_table *table;
8d1c802b 3748 struct fib6_info *rt;
1da177e4 3749 struct fib6_node *fn;
1da177e4
LT		 3750 int err = -ESRCH;
3751
5578689a 3752 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
d5d531cb
DA		 3753 if (!table) {
3754 NL_SET_ERR_MSG(extack, "FIB table does not exist");
c71099ac 3755 return err;
d5d531cb 3756 }
c71099ac 3757
66f5d6ce 3758 rcu_read_lock();
1da177e4 3759
c71099ac 3760 fn = fib6_locate(&table->tb6_root,
86872cb5 3761 &cfg->fc_dst, cfg->fc_dst_len,
38fbeeee 3762 &cfg->fc_src, cfg->fc_src_len,
2b760fcf 3763 !(cfg->fc_flags & RTF_CACHE));
1ab1457c 3764
1da177e4 3765 if (fn) {
66f5d6ce 3766 for_each_fib6_node_rt_rcu(fn) {
ad1601ae
DA		 3767 struct fib6_nh *nh;
3768
1cf844c7 3769 nh = rt->fib6_nh;
2b760fcf 3770 if (cfg->fc_flags & RTF_CACHE) {
23fb93a4
DA		 3771 int rc;
3772
0fa6efc5
DA		 3773 rc = ip6_del_cached_rt(cfg, rt, nh);
3774 if (rc != -ESRCH) {
3775 rcu_read_unlock();
3776 return rc;
23fb93a4
DA		 3777 }
3778 continue;
2b760fcf 3779 }
ad1601ae 3780
86872cb5 3781 if (cfg->fc_ifindex &&
ad1601ae
DA		 3782 (!nh->fib_nh_dev ||
3783 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
1da177e4 3784 continue;
86872cb5 3785 if (cfg->fc_flags & RTF_GATEWAY &&
ad1601ae 3786 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
1da177e4 3787 continue;
93c2fb25 3788 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
1da177e4 3789 continue;
93c2fb25 3790 if (cfg->fc_protocol && cfg->fc_protocol != rt->fib6_protocol)
c2ed1880 3791 continue;
e873e4b9
WW		 3792 if (!fib6_info_hold_safe(rt))
3793 continue;
66f5d6ce 3794 rcu_read_unlock();
1da177e4 3795
0ae81335
DA		 3796 /* if gateway was specified only delete the one hop */
3797 if (cfg->fc_flags & RTF_GATEWAY)
3798 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3799
3800 return __ip6_del_rt_siblings(rt, cfg);
1da177e4
LT		 3801 }
3802 }
66f5d6ce 3803 rcu_read_unlock();
1da177e4
LT		 3804
3805 return err;
3806}
3807
6700c270 3808static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
a6279458 3809{
a6279458 3810 struct netevent_redirect netevent;
e8599ff4 3811 struct rt6_info *rt, *nrt = NULL;
85bd05de 3812 struct fib6_result res = {};
e8599ff4
DM		 3813 struct ndisc_options ndopts;
3814 struct inet6_dev *in6_dev;
3815 struct neighbour *neigh;
71bcdba0 3816 struct rd_msg *msg;
6e157b6a
DM		 3817 int optlen, on_link;
3818 u8 *lladdr;
e8599ff4 3819
29a3cad5 3820 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
71bcdba0 3821 optlen -= sizeof(*msg);
e8599ff4
DM		 3822
3823 if (optlen < 0) {
6e157b6a 3824 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
e8599ff4
DM		 3825 return;
3826 }
3827
71bcdba0 3828 msg = (struct rd_msg *)icmp6_hdr(skb);
e8599ff4 3829
71bcdba0 3830 if (ipv6_addr_is_multicast(&msg->dest)) {
6e157b6a 3831 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
e8599ff4
DM		 3832 return;
3833 }
3834
6e157b6a 3835 on_link = 0;
71bcdba0 3836 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
e8599ff4 3837 on_link = 1;
71bcdba0 3838 } else if (ipv6_addr_type(&msg->target) !=
e8599ff4 3839 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
6e157b6a 3840 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
e8599ff4
DM		 3841 return;
3842 }
3843
3844 in6_dev = __in6_dev_get(skb->dev);
3845 if (!in6_dev)
3846 return;
3847 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
3848 return;
3849
3850 /* RFC2461 8.1:
3851 * The IP source address of the Redirect MUST be the same as the current
3852 * first-hop router for the specified ICMP Destination Address.
3853 */
3854
f997c55c 3855 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
e8599ff4
DM		 3856 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
3857 return;
3858 }
6e157b6a
DM		 3859
3860 lladdr = NULL;
e8599ff4
DM		 3861 if (ndopts.nd_opts_tgt_lladdr) {
3862 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
3863 skb->dev);
3864 if (!lladdr) {
3865 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
3866 return;
3867 }
3868 }
3869
6e157b6a 3870 rt = (struct rt6_info *) dst;
ec13ad1d 3871 if (rt->rt6i_flags & RTF_REJECT) {
6e157b6a 3872 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
e8599ff4 3873 return;
6e157b6a 3874 }
e8599ff4 3875
6e157b6a
DM		 3876 /* Redirect received -> path was valid.
3877 * Look, redirects are sent only in response to data packets,
3878 * so that this nexthop apparently is reachable. --ANK
3879 */
0dec879f 3880 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
a6279458 3881
71bcdba0 3882 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
6e157b6a
DM		 3883 if (!neigh)
3884 return;
a6279458 3885
1da177e4
LT		 3886 /*
3887 * We have finally decided to accept it.
3888 */
3889
f997c55c 3890 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1da177e4
LT		 3891 NEIGH_UPDATE_F_WEAK_OVERRIDE|
3892 NEIGH_UPDATE_F_OVERRIDE|
3893 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
f997c55c
AA		 3894 NEIGH_UPDATE_F_ISROUTER)),
3895 NDISC_REDIRECT, &ndopts);
1da177e4 3896
4d85cd0c 3897 rcu_read_lock();
85bd05de 3898 res.f6i = rcu_dereference(rt->from);
ff24e498 3899 if (!res.f6i)
886b7a50 3900 goto out;
8a14e46f 3901
1cf844c7 3902 res.nh = res.f6i->fib6_nh;
7d21fec9
DA		 3903 res.fib6_flags = res.f6i->fib6_flags;
3904 res.fib6_type = res.f6i->fib6_type;
85bd05de 3905 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
38308473 3906 if (!nrt)
1da177e4
LT		 3907 goto out;
3908
3909 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
3910 if (on_link)
3911 nrt->rt6i_flags &= ~RTF_GATEWAY;
3912
4e3fd7a0 3913 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1da177e4 3914
886b7a50 3915 /* rt6_insert_exception() will take care of duplicated exceptions */
5012f0a5 3916 if (rt6_insert_exception(nrt, &res)) {
2b760fcf
WW		 3917 dst_release_immediate(&nrt->dst);
3918 goto out;
3919 }
1da177e4 3920
d8d1f30b
CG		 3921 netevent.old = &rt->dst;
3922 netevent.new = &nrt->dst;
71bcdba0 3923 netevent.daddr = &msg->dest;
60592833 3924 netevent.neigh = neigh;
8d71740c
TT		 3925 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
3926
1da177e4 3927out:
886b7a50 3928 rcu_read_unlock();
e8599ff4 3929 neigh_release(neigh);
6e157b6a
DM		 3930}
3931
70ceb4f5 3932#ifdef CONFIG_IPV6_ROUTE_INFO
8d1c802b 3933static struct fib6_info *rt6_get_route_info(struct net *net,
b71d1d42 3934 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 3935 const struct in6_addr *gwaddr,
3936 struct net_device *dev)
70ceb4f5 3937{
830218c1
DA		 3938 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
3939 int ifindex = dev->ifindex;
70ceb4f5 3940 struct fib6_node *fn;
8d1c802b 3941 struct fib6_info *rt = NULL;
c71099ac
TG		 3942 struct fib6_table *table;
3943
830218c1 3944 table = fib6_get_table(net, tb_id);
38308473 3945 if (!table)
c71099ac 3946 return NULL;
70ceb4f5 3947
66f5d6ce 3948 rcu_read_lock();
38fbeeee 3949 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
70ceb4f5
YH		 3950 if (!fn)
3951 goto out;
3952
66f5d6ce 3953 for_each_fib6_node_rt_rcu(fn) {
f88d8ea6
DA		 3954 /* these routes do not use nexthops */
3955 if (rt->nh)
3956 continue;
1cf844c7 3957 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
70ceb4f5 3958 continue;
2b2450ca 3959 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
1cf844c7 3960 !rt->fib6_nh->fib_nh_gw_family)
70ceb4f5 3961 continue;
1cf844c7 3962 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
70ceb4f5 3963 continue;
e873e4b9
WW		 3964 if (!fib6_info_hold_safe(rt))
3965 continue;
70ceb4f5
YH		 3966 break;
3967 }
3968out:
66f5d6ce 3969 rcu_read_unlock();
70ceb4f5
YH		 3970 return rt;
3971}
3972
8d1c802b 3973static struct fib6_info *rt6_add_route_info(struct net *net,
b71d1d42 3974 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 3975 const struct in6_addr *gwaddr,
3976 struct net_device *dev,
95c96174 3977 unsigned int pref)
70ceb4f5 3978{
86872cb5 3979 struct fib6_config cfg = {
238fc7ea 3980 .fc_metric = IP6_RT_PRIO_USER,
830218c1 3981 .fc_ifindex = dev->ifindex,
86872cb5
TG		 3982 .fc_dst_len = prefixlen,
3983 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
3984 RTF_UP | RTF_PREF(pref),
b91d5329 3985 .fc_protocol = RTPROT_RA,
e8478e80 3986 .fc_type = RTN_UNICAST,
15e47304 3987 .fc_nlinfo.portid = 0,
efa2cea0
DL		 3988 .fc_nlinfo.nlh = NULL,
3989 .fc_nlinfo.nl_net = net,
86872cb5
TG		 3990 };
3991
830218c1 3992 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
4e3fd7a0
AD		 3993 cfg.fc_dst = *prefix;
3994 cfg.fc_gateway = *gwaddr;
70ceb4f5 3995
e317da96
YH		 3996 /* We should treat it as a default route if prefix length is 0. */
3997 if (!prefixlen)
86872cb5 3998 cfg.fc_flags |= RTF_DEFAULT;
70ceb4f5 3999
acb54e3c 4000 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
70ceb4f5 4001
830218c1 4002 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
70ceb4f5
YH		 4003}
4004#endif
4005
8d1c802b 4006struct fib6_info *rt6_get_dflt_router(struct net *net,
afb1d4b5
DA		 4007 const struct in6_addr *addr,
4008 struct net_device *dev)
1ab1457c 4009{
830218c1 4010 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
8d1c802b 4011 struct fib6_info *rt;
c71099ac 4012 struct fib6_table *table;
1da177e4 4013
afb1d4b5 4014 table = fib6_get_table(net, tb_id);
38308473 4015 if (!table)
c71099ac 4016 return NULL;
1da177e4 4017
66f5d6ce
WW		 4018 rcu_read_lock();
4019 for_each_fib6_node_rt_rcu(&table->tb6_root) {
f88d8ea6 4020 struct fib6_nh *nh;
ad1601ae 4021
f88d8ea6
DA		 4022 /* RA routes do not use nexthops */
4023 if (rt->nh)
4024 continue;
4025
4026 nh = rt->fib6_nh;
ad1601ae 4027 if (dev == nh->fib_nh_dev &&
93c2fb25 4028 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
ad1601ae 4029 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
1da177e4
LT		 4030 break;
4031 }
e873e4b9
WW		 4032 if (rt && !fib6_info_hold_safe(rt))
4033 rt = NULL;
66f5d6ce 4034 rcu_read_unlock();
1da177e4
LT		 4035 return rt;
4036}
4037
8d1c802b 4038struct fib6_info *rt6_add_dflt_router(struct net *net,
afb1d4b5 4039 const struct in6_addr *gwaddr,
ebacaaa0
YH		 4040 struct net_device *dev,
4041 unsigned int pref)
1da177e4 4042{
86872cb5 4043 struct fib6_config cfg = {
ca254490 4044 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
238fc7ea 4045 .fc_metric = IP6_RT_PRIO_USER,
86872cb5
TG		 4046 .fc_ifindex = dev->ifindex,
4047 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4048 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
b91d5329 4049 .fc_protocol = RTPROT_RA,
e8478e80 4050 .fc_type = RTN_UNICAST,
15e47304 4051 .fc_nlinfo.portid = 0,
5578689a 4052 .fc_nlinfo.nlh = NULL,
afb1d4b5 4053 .fc_nlinfo.nl_net = net,
86872cb5 4054 };
1da177e4 4055
4e3fd7a0 4056 cfg.fc_gateway = *gwaddr;
1da177e4 4057
acb54e3c 4058 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
830218c1
DA		 4059 struct fib6_table *table;
4060
4061 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4062 if (table)
4063 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4064 }
1da177e4 4065
afb1d4b5 4066 return rt6_get_dflt_router(net, gwaddr, dev);
1da177e4
LT		 4067}
4068
afb1d4b5
DA		 4069static void __rt6_purge_dflt_routers(struct net *net,
4070 struct fib6_table *table)
1da177e4 4071{
8d1c802b 4072 struct fib6_info *rt;
1da177e4
LT		 4073
4074restart:
66f5d6ce
WW		 4075 rcu_read_lock();
4076 for_each_fib6_node_rt_rcu(&table->tb6_root) {
dcd1f572
DA		 4077 struct net_device *dev = fib6_info_nh_dev(rt);
4078 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4079
93c2fb25 4080 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
e873e4b9
WW		 4081 (!idev || idev->cnf.accept_ra != 2) &&
4082 fib6_info_hold_safe(rt)) {
93531c67
DA		 4083 rcu_read_unlock();
4084 ip6_del_rt(net, rt);
1da177e4
LT		 4085 goto restart;
4086 }
4087 }
66f5d6ce 4088 rcu_read_unlock();
830218c1
DA		 4089
4090 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4091}
4092
4093void rt6_purge_dflt_routers(struct net *net)
4094{
4095 struct fib6_table *table;
4096 struct hlist_head *head;
4097 unsigned int h;
4098
4099 rcu_read_lock();
4100
4101 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4102 head = &net->ipv6.fib_table_hash[h];
4103 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4104 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
afb1d4b5 4105 __rt6_purge_dflt_routers(net, table);
830218c1
DA		 4106 }
4107 }
4108
4109 rcu_read_unlock();
1da177e4
LT		 4110}
4111
5578689a
DL		 4112static void rtmsg_to_fib6_config(struct net *net,
4113 struct in6_rtmsg *rtmsg,
86872cb5
TG		 4114 struct fib6_config *cfg)
4115{
8823a3ac
MŻ		 4116 *cfg = (struct fib6_config){
4117 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4118 : RT6_TABLE_MAIN,
4119 .fc_ifindex = rtmsg->rtmsg_ifindex,
67f69513 4120 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
8823a3ac
MŻ		 4121 .fc_expires = rtmsg->rtmsg_info,
4122 .fc_dst_len = rtmsg->rtmsg_dst_len,
4123 .fc_src_len = rtmsg->rtmsg_src_len,
4124 .fc_flags = rtmsg->rtmsg_flags,
4125 .fc_type = rtmsg->rtmsg_type,
4126
4127 .fc_nlinfo.nl_net = net,
4128
4129 .fc_dst = rtmsg->rtmsg_dst,
4130 .fc_src = rtmsg->rtmsg_src,
4131 .fc_gateway = rtmsg->rtmsg_gateway,
4132 };
86872cb5
TG		 4133}
4134
5578689a 4135int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1da177e4 4136{
86872cb5 4137 struct fib6_config cfg;
1da177e4
LT		 4138 struct in6_rtmsg rtmsg;
4139 int err;
4140
67ba4152 4141 switch (cmd) {
1da177e4
LT		 4142 case SIOCADDRT: /* Add a route */
4143 case SIOCDELRT: /* Delete a route */
af31f412 4144 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4
LT		 4145 return -EPERM;
4146 err = copy_from_user(&rtmsg, arg,
4147 sizeof(struct in6_rtmsg));
4148 if (err)
4149 return -EFAULT;
86872cb5 4150
5578689a 4151 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
86872cb5 4152
1da177e4
LT		 4153 rtnl_lock();
4154 switch (cmd) {
4155 case SIOCADDRT:
acb54e3c 4156 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
1da177e4
LT		 4157 break;
4158 case SIOCDELRT:
333c4301 4159 err = ip6_route_del(&cfg, NULL);
1da177e4
LT		 4160 break;
4161 default:
4162 err = -EINVAL;
4163 }
4164 rtnl_unlock();
4165
4166 return err;
3ff50b79 4167 }
1da177e4
LT		 4168
4169 return -EINVAL;
4170}
4171
4172/*
4173 * Drop the packet on the floor
4174 */
4175
d5fdd6ba 4176static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1da177e4 4177{
adf30907 4178 struct dst_entry *dst = skb_dst(skb);
1d3fd8a1
SS		 4179 struct net *net = dev_net(dst->dev);
4180 struct inet6_dev *idev;
4181 int type;
4182
4183 if (netif_is_l3_master(skb->dev) &&
4184 dst->dev == net->loopback_dev)
4185 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4186 else
4187 idev = ip6_dst_idev(dst);
4188
612f09e8
YH		 4189 switch (ipstats_mib_noroutes) {
4190 case IPSTATS_MIB_INNOROUTES:
0660e03f 4191 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
45bb0060 4192 if (type == IPV6_ADDR_ANY) {
1d3fd8a1 4193 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
612f09e8
YH		 4194 break;
4195 }
4196 /* FALLTHROUGH */
4197 case IPSTATS_MIB_OUTNOROUTES:
1d3fd8a1 4198 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
612f09e8
YH		 4199 break;
4200 }
1d3fd8a1
SS		 4201
4202 /* Start over by dropping the dst for l3mdev case */
4203 if (netif_is_l3_master(skb->dev))
4204 skb_dst_drop(skb);
4205
3ffe533c 4206 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1da177e4
LT		 4207 kfree_skb(skb);
4208 return 0;
4209}
4210
9ce8ade0
TG		 4211static int ip6_pkt_discard(struct sk_buff *skb)
4212{
612f09e8 4213 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4214}
4215
ede2059d 4216static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
1da177e4 4217{
adf30907 4218 skb->dev = skb_dst(skb)->dev;
612f09e8 4219 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1da177e4
LT		 4220}
4221
9ce8ade0
TG		 4222static int ip6_pkt_prohibit(struct sk_buff *skb)
4223{
612f09e8 4224 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 4225}
4226
ede2059d 4227static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
9ce8ade0 4228{
adf30907 4229 skb->dev = skb_dst(skb)->dev;
612f09e8 4230 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
9ce8ade0
TG		 4231}
4232
1da177e4
LT		 4233/*
4234 * Allocate a dst for local (unicast / anycast) address.
4235 */
4236
360a9887
DA		 4237struct fib6_info *addrconf_f6i_alloc(struct net *net,
4238 struct inet6_dev *idev,
4239 const struct in6_addr *addr,
4240 bool anycast, gfp_t gfp_flags)
1da177e4 4241{
c7a1ce39
DA		 4242 struct fib6_config cfg = {
4243 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4244 .fc_ifindex = idev->dev->ifindex,
4245 .fc_flags = RTF_UP | RTF_ADDRCONF | RTF_NONEXTHOP,
4246 .fc_dst = *addr,
4247 .fc_dst_len = 128,
4248 .fc_protocol = RTPROT_KERNEL,
4249 .fc_nlinfo.nl_net = net,
4250 .fc_ignore_dev_down = true,
4251 };
1da177e4 4252
e8478e80 4253 if (anycast) {
c7a1ce39
DA		 4254 cfg.fc_type = RTN_ANYCAST;
4255 cfg.fc_flags |= RTF_ANYCAST;
e8478e80 4256 } else {
c7a1ce39
DA		 4257 cfg.fc_type = RTN_LOCAL;
4258 cfg.fc_flags |= RTF_LOCAL;
e8478e80 4259 }
1da177e4 4260
c7a1ce39 4261 return ip6_route_info_create(&cfg, gfp_flags, NULL);
1da177e4
LT		 4262}
4263
c3968a85
DW		 4264/* remove deleted ip from prefsrc entries */
4265struct arg_dev_net_ip {
4266 struct net_device *dev;
4267 struct net *net;
4268 struct in6_addr *addr;
4269};
4270
8d1c802b 4271static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
c3968a85
DW		 4272{
4273 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4274 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4275 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4276
f88d8ea6
DA		 4277 if (!rt->nh &&
4278 ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
421842ed 4279 rt != net->ipv6.fib6_null_entry &&
93c2fb25 4280 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
60006a48 4281 spin_lock_bh(&rt6_exception_lock);
c3968a85 4282 /* remove prefsrc entry */
93c2fb25 4283 rt->fib6_prefsrc.plen = 0;
60006a48 4284 spin_unlock_bh(&rt6_exception_lock);
c3968a85
DW		 4285 }
4286 return 0;
4287}
4288
4289void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4290{
4291 struct net *net = dev_net(ifp->idev->dev);
4292 struct arg_dev_net_ip adni = {
4293 .dev = ifp->idev->dev,
4294 .net = net,
4295 .addr = &ifp->addr,
4296 };
0c3584d5 4297 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
c3968a85
DW		 4298}
4299
2b2450ca 4300#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
be7a010d
DJ		 4301
4302/* Remove routers and update dst entries when gateway turn into host. */
8d1c802b 4303static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
be7a010d
DJ		 4304{
4305 struct in6_addr *gateway = (struct in6_addr *)arg;
f88d8ea6
DA		 4306 struct fib6_nh *nh;
4307
4308 /* RA routes do not use nexthops */
4309 if (rt->nh)
4310 return 0;
be7a010d 4311
f88d8ea6 4312 nh = rt->fib6_nh;
93c2fb25 4313 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
cc5c073a 4314 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
be7a010d 4315 return -1;
b16cb459
WW		 4316
4317 /* Further clean up cached routes in exception table.
4318 * This is needed because cached route may have a different
4319 * gateway than its 'parent' in the case of an ip redirect.
4320 */
cc5c073a 4321 fib6_nh_exceptions_clean_tohost(nh, gateway);
b16cb459 4322
be7a010d
DJ		 4323 return 0;
4324}
4325
4326void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4327{
4328 fib6_clean_all(net, fib6_clean_tohost, gateway);
4329}
4330
2127d95a
IS		 4331struct arg_netdev_event {
4332 const struct net_device *dev;
4c981e28 4333 union {
ecc5663c 4334 unsigned char nh_flags;
4c981e28
IS		 4335 unsigned long event;
4336 };
2127d95a
IS		 4337};
4338
8d1c802b 4339static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
d7dedee1 4340{
8d1c802b 4341 struct fib6_info *iter;
d7dedee1
IS		 4342 struct fib6_node *fn;
4343
93c2fb25
DA		 4344 fn = rcu_dereference_protected(rt->fib6_node,
4345 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4346 iter = rcu_dereference_protected(fn->leaf,
93c2fb25 4347 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1 4348 while (iter) {
93c2fb25 4349 if (iter->fib6_metric == rt->fib6_metric &&
33bd5ac5 4350 rt6_qualify_for_ecmp(iter))
d7dedee1 4351 return iter;
8fb11a9a 4352 iter = rcu_dereference_protected(iter->fib6_next,
93c2fb25 4353 lockdep_is_held(&rt->fib6_table->tb6_lock));
d7dedee1
IS		 4354 }
4355
4356 return NULL;
4357}
4358
f88d8ea6 4359/* only called for fib entries with builtin fib6_nh */
8d1c802b 4360static bool rt6_is_dead(const struct fib6_info *rt)
d7dedee1 4361{
1cf844c7
DA		 4362 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4363 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4364 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
d7dedee1
IS		 4365 return true;
4366
4367 return false;
4368}
4369
8d1c802b 4370static int rt6_multipath_total_weight(const struct fib6_info *rt)
d7dedee1 4371{
8d1c802b 4372 struct fib6_info *iter;
d7dedee1
IS		 4373 int total = 0;
4374
4375 if (!rt6_is_dead(rt))
1cf844c7 4376 total += rt->fib6_nh->fib_nh_weight;
d7dedee1 4377
93c2fb25 4378 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
d7dedee1 4379 if (!rt6_is_dead(iter))
1cf844c7 4380 total += iter->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4381 }
4382
4383 return total;
4384}
4385
8d1c802b 4386static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
d7dedee1
IS		 4387{
4388 int upper_bound = -1;
4389
4390 if (!rt6_is_dead(rt)) {
1cf844c7 4391 *weight += rt->fib6_nh->fib_nh_weight;
d7dedee1
IS		 4392 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4393 total) - 1;
4394 }
1cf844c7 4395 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
d7dedee1
IS		 4396}
4397
8d1c802b 4398static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
d7dedee1 4399{
8d1c802b 4400 struct fib6_info *iter;
d7dedee1
IS		 4401 int weight = 0;
4402
4403 rt6_upper_bound_set(rt, &weight, total);
4404
93c2fb25 4405 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
d7dedee1
IS		 4406 rt6_upper_bound_set(iter, &weight, total);
4407}
4408
8d1c802b 4409void rt6_multipath_rebalance(struct fib6_info *rt)
d7dedee1 4410{
8d1c802b 4411 struct fib6_info *first;
d7dedee1
IS		 4412 int total;
4413
4414 /* In case the entire multipath route was marked for flushing,
4415 * then there is no need to rebalance upon the removal of every
4416 * sibling route.
4417 */
93c2fb25 4418 if (!rt->fib6_nsiblings || rt->should_flush)
d7dedee1
IS		 4419 return;
4420
4421 /* During lookup routes are evaluated in order, so we need to
4422 * make sure upper bounds are assigned from the first sibling
4423 * onwards.
4424 */
4425 first = rt6_multipath_first_sibling(rt);
4426 if (WARN_ON_ONCE(!first))
4427 return;
4428
4429 total = rt6_multipath_total_weight(first);
4430 rt6_multipath_upper_bound_set(first, total);
4431}
4432
8d1c802b 4433static int fib6_ifup(struct fib6_info *rt, void *p_arg)
2127d95a
IS		 4434{
4435 const struct arg_netdev_event *arg = p_arg;
7aef6859 4436 struct net *net = dev_net(arg->dev);
2127d95a 4437
f88d8ea6 4438 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
1cf844c7
DA		 4439 rt->fib6_nh->fib_nh_dev == arg->dev) {
4440 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
7aef6859 4441 fib6_update_sernum_upto_root(net, rt);
d7dedee1 4442 rt6_multipath_rebalance(rt);
1de178ed 4443 }
2127d95a
IS		 4444
4445 return 0;
4446}
4447
ecc5663c 4448void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
2127d95a
IS		 4449{
4450 struct arg_netdev_event arg = {
4451 .dev = dev,
6802f3ad
IS		 4452 {
4453 .nh_flags = nh_flags,
4454 },
2127d95a
IS		 4455 };
4456
4457 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4458 arg.nh_flags |= RTNH_F_LINKDOWN;
4459
4460 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4461}
4462
f88d8ea6 4463/* only called for fib entries with inline fib6_nh */
8d1c802b 4464static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
1de178ed
IS		 4465 const struct net_device *dev)
4466{
8d1c802b 4467 struct fib6_info *iter;
1de178ed 4468
1cf844c7 4469 if (rt->fib6_nh->fib_nh_dev == dev)
1de178ed 4470 return true;
93c2fb25 4471 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7 4472 if (iter->fib6_nh->fib_nh_dev == dev)
1de178ed
IS		 4473 return true;
4474
4475 return false;
4476}
4477
8d1c802b 4478static void rt6_multipath_flush(struct fib6_info *rt)
1de178ed 4479{
8d1c802b 4480 struct fib6_info *iter;
1de178ed
IS		 4481
4482 rt->should_flush = 1;
93c2fb25 4483 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1de178ed
IS		 4484 iter->should_flush = 1;
4485}
4486
8d1c802b 4487static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
1de178ed
IS		 4488 const struct net_device *down_dev)
4489{
8d1c802b 4490 struct fib6_info *iter;
1de178ed
IS		 4491 unsigned int dead = 0;
4492
1cf844c7
DA		 4493 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4494 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed 4495 dead++;
93c2fb25 4496 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4497 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4498 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
1de178ed
IS		 4499 dead++;
4500
4501 return dead;
4502}
4503
8d1c802b 4504static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
1de178ed 4505 const struct net_device *dev,
ecc5663c 4506 unsigned char nh_flags)
1de178ed 4507{
8d1c802b 4508 struct fib6_info *iter;
1de178ed 4509
1cf844c7
DA		 4510 if (rt->fib6_nh->fib_nh_dev == dev)
4511 rt->fib6_nh->fib_nh_flags |= nh_flags;
93c2fb25 4512 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
1cf844c7
DA		 4513 if (iter->fib6_nh->fib_nh_dev == dev)
4514 iter->fib6_nh->fib_nh_flags |= nh_flags;
1de178ed
IS		 4515}
4516
a1a22c12 4517/* called with write lock held for table with rt */
8d1c802b 4518static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
1da177e4 4519{
4c981e28
IS		 4520 const struct arg_netdev_event *arg = p_arg;
4521 const struct net_device *dev = arg->dev;
7aef6859 4522 struct net *net = dev_net(dev);
8ed67789 4523
f88d8ea6 4524 if (rt == net->ipv6.fib6_null_entry || rt->nh)
27c6fa73
IS		 4525 return 0;
4526
4527 switch (arg->event) {
4528 case NETDEV_UNREGISTER:
1cf844c7 4529 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
27c6fa73 4530 case NETDEV_DOWN:
1de178ed 4531 if (rt->should_flush)
27c6fa73 4532 return -1;
93c2fb25 4533 if (!rt->fib6_nsiblings)
1cf844c7 4534 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
1de178ed
IS		 4535 if (rt6_multipath_uses_dev(rt, dev)) {
4536 unsigned int count;
4537
4538 count = rt6_multipath_dead_count(rt, dev);
93c2fb25 4539 if (rt->fib6_nsiblings + 1 == count) {
1de178ed
IS		 4540 rt6_multipath_flush(rt);
4541 return -1;
4542 }
4543 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4544 RTNH_F_LINKDOWN);
7aef6859 4545 fib6_update_sernum(net, rt);
d7dedee1 4546 rt6_multipath_rebalance(rt);
1de178ed
IS		 4547 }
4548 return -2;
27c6fa73 4549 case NETDEV_CHANGE:
1cf844c7 4550 if (rt->fib6_nh->fib_nh_dev != dev ||
93c2fb25 4551 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
27c6fa73 4552 break;
1cf844c7 4553 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
d7dedee1 4554 rt6_multipath_rebalance(rt);
27c6fa73 4555 break;
2b241361 4556 }
c159d30c 4557
1da177e4
LT		 4558 return 0;
4559}
4560
27c6fa73 4561void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
1da177e4 4562{
4c981e28 4563 struct arg_netdev_event arg = {
8ed67789 4564 .dev = dev,
6802f3ad
IS		 4565 {
4566 .event = event,
4567 },
8ed67789 4568 };
7c6bb7d2 4569 struct net *net = dev_net(dev);
8ed67789 4570
7c6bb7d2
DA		 4571 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4572 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4573 else
4574 fib6_clean_all(net, fib6_ifdown, &arg);
4c981e28
IS		 4575}
4576
4577void rt6_disable_ip(struct net_device *dev, unsigned long event)
4578{
4579 rt6_sync_down_dev(dev, event);
4580 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4581 neigh_ifdown(&nd_tbl, dev);
1da177e4
LT		 4582}
4583
95c96174 4584struct rt6_mtu_change_arg {
1da177e4 4585 struct net_device *dev;
95c96174 4586 unsigned int mtu;
c0b220cf 4587 struct fib6_info *f6i;
1da177e4
LT		 4588};
4589
cc5c073a 4590static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
c0b220cf
DA		 4591{
4592 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
cc5c073a 4593 struct fib6_info *f6i = arg->f6i;
c0b220cf
DA		 4594
4595 /* For administrative MTU increase, there is no way to discover
4596 * IPv6 PMTU increase, so PMTU increase should be updated here.
4597 * Since RFC 1981 doesn't include administrative MTU increase
4598 * update PMTU increase is a MUST. (i.e. jumbo frame)
4599 */
4600 if (nh->fib_nh_dev == arg->dev) {
4601 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4602 u32 mtu = f6i->fib6_pmtu;
4603
4604 if (mtu >= arg->mtu ||
4605 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4606 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4607
4608 spin_lock_bh(&rt6_exception_lock);
cc5c073a 4609 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
c0b220cf
DA		 4610 spin_unlock_bh(&rt6_exception_lock);
4611 }
4612
4613 return 0;
4614}
4615
4616static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
1da177e4
LT		 4617{
4618 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4619 struct inet6_dev *idev;
4620
4621 /* In IPv6 pmtu discovery is not optional,
4622 so that RTAX_MTU lock cannot disable it.
4623 We still use this lock to block changes
4624 caused by addrconf/ndisc.
4625 */
4626
4627 idev = __in6_dev_get(arg->dev);
38308473 4628 if (!idev)
1da177e4
LT		 4629 return 0;
4630
c0b220cf
DA		 4631 if (fib6_metric_locked(f6i, RTAX_MTU))
4632 return 0;
d4ead6b3 4633
c0b220cf 4634 arg->f6i = f6i;
1cf844c7 4635 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
1da177e4
LT		 4636}
4637
95c96174 4638void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
1da177e4 4639{
c71099ac
TG		 4640 struct rt6_mtu_change_arg arg = {
4641 .dev = dev,
4642 .mtu = mtu,
4643 };
1da177e4 4644
0c3584d5 4645 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
1da177e4
LT		 4646}
4647
ef7c79ed 4648static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
75425657 4649 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
5176f91e 4650 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
aa8f8778 4651 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
86872cb5 4652 [RTA_OIF] = { .type = NLA_U32 },
ab364a6f 4653 [RTA_IIF] = { .type = NLA_U32 },
86872cb5
TG		 4654 [RTA_PRIORITY] = { .type = NLA_U32 },
4655 [RTA_METRICS] = { .type = NLA_NESTED },
51ebd318 4656 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
c78ba6d6 4657 [RTA_PREF] = { .type = NLA_U8 },
19e42e45
RP		 4658 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4659 [RTA_ENCAP] = { .type = NLA_NESTED },
32bc201e 4660 [RTA_EXPIRES] = { .type = NLA_U32 },
622ec2c9 4661 [RTA_UID] = { .type = NLA_U32 },
3b45a410 4662 [RTA_MARK] = { .type = NLA_U32 },
aa8f8778 4663 [RTA_TABLE] = { .type = NLA_U32 },
eacb9384
RP		 4664 [RTA_IP_PROTO] = { .type = NLA_U8 },
4665 [RTA_SPORT] = { .type = NLA_U16 },
4666 [RTA_DPORT] = { .type = NLA_U16 },
86872cb5
TG		 4667};
4668
4669static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
333c4301
DA		 4670 struct fib6_config *cfg,
4671 struct netlink_ext_ack *extack)
1da177e4 4672{
86872cb5
TG		 4673 struct rtmsg *rtm;
4674 struct nlattr *tb[RTA_MAX+1];
c78ba6d6 4675 unsigned int pref;
86872cb5 4676 int err;
1da177e4 4677
8cb08174
JB		 4678 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
4679 rtm_ipv6_policy, extack);
86872cb5
TG		 4680 if (err < 0)
4681 goto errout;
1da177e4 4682
86872cb5
TG		 4683 err = -EINVAL;
4684 rtm = nlmsg_data(nlh);
86872cb5 4685
84db8407
MŻ		 4686 *cfg = (struct fib6_config){
4687 .fc_table = rtm->rtm_table,
4688 .fc_dst_len = rtm->rtm_dst_len,
4689 .fc_src_len = rtm->rtm_src_len,
4690 .fc_flags = RTF_UP,
4691 .fc_protocol = rtm->rtm_protocol,
4692 .fc_type = rtm->rtm_type,
4693
4694 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
4695 .fc_nlinfo.nlh = nlh,
4696 .fc_nlinfo.nl_net = sock_net(skb->sk),
4697 };
86872cb5 4698
ef2c7d7b
ND		 4699 if (rtm->rtm_type == RTN_UNREACHABLE ||
4700 rtm->rtm_type == RTN_BLACKHOLE ||
b4949ab2
ND		 4701 rtm->rtm_type == RTN_PROHIBIT ||
4702 rtm->rtm_type == RTN_THROW)
86872cb5
TG		 4703 cfg->fc_flags |= RTF_REJECT;
4704
ab79ad14
MŻ		 4705 if (rtm->rtm_type == RTN_LOCAL)
4706 cfg->fc_flags |= RTF_LOCAL;
4707
1f56a01f
MKL		 4708 if (rtm->rtm_flags & RTM_F_CLONED)
4709 cfg->fc_flags |= RTF_CACHE;
4710
fc1e64e1
DA		 4711 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4712
86872cb5 4713 if (tb[RTA_GATEWAY]) {
67b61f6c 4714 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
86872cb5 4715 cfg->fc_flags |= RTF_GATEWAY;
1da177e4 4716 }
e3818541
DA		 4717 if (tb[RTA_VIA]) {
4718 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
4719 goto errout;
4720 }
86872cb5
TG		 4721
4722 if (tb[RTA_DST]) {
4723 int plen = (rtm->rtm_dst_len + 7) >> 3;
4724
4725 if (nla_len(tb[RTA_DST]) < plen)
4726 goto errout;
4727
4728 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1da177e4 4729 }
86872cb5
TG		 4730
4731 if (tb[RTA_SRC]) {
4732 int plen = (rtm->rtm_src_len + 7) >> 3;
4733
4734 if (nla_len(tb[RTA_SRC]) < plen)
4735 goto errout;
4736
4737 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1da177e4 4738 }
86872cb5 4739
c3968a85 4740 if (tb[RTA_PREFSRC])
67b61f6c 4741 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
c3968a85 4742
86872cb5
TG		 4743 if (tb[RTA_OIF])
4744 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4745
4746 if (tb[RTA_PRIORITY])
4747 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4748
4749 if (tb[RTA_METRICS]) {
4750 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4751 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
1da177e4 4752 }
86872cb5
TG		 4753
4754 if (tb[RTA_TABLE])
4755 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4756
51ebd318
ND		 4757 if (tb[RTA_MULTIPATH]) {
4758 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4759 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
9ed59592
DA		 4760
4761 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
c255bd68 4762 cfg->fc_mp_len, extack);
9ed59592
DA		 4763 if (err < 0)
4764 goto errout;
51ebd318
ND		 4765 }
4766
c78ba6d6
LR		 4767 if (tb[RTA_PREF]) {
4768 pref = nla_get_u8(tb[RTA_PREF]);
4769 if (pref != ICMPV6_ROUTER_PREF_LOW &&
4770 pref != ICMPV6_ROUTER_PREF_HIGH)
4771 pref = ICMPV6_ROUTER_PREF_MEDIUM;
4772 cfg->fc_flags |= RTF_PREF(pref);
4773 }
4774
19e42e45
RP		 4775 if (tb[RTA_ENCAP])
4776 cfg->fc_encap = tb[RTA_ENCAP];
4777
9ed59592 4778 if (tb[RTA_ENCAP_TYPE]) {
19e42e45
RP		 4779 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
4780
c255bd68 4781 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
9ed59592
DA		 4782 if (err < 0)
4783 goto errout;
4784 }
4785
32bc201e
XL		 4786 if (tb[RTA_EXPIRES]) {
4787 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
4788
4789 if (addrconf_finite_timeout(timeout)) {
4790 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
4791 cfg->fc_flags |= RTF_EXPIRES;
4792 }
4793 }
4794
86872cb5
TG		 4795 err = 0;
4796errout:
4797 return err;
1da177e4
LT		 4798}
4799
6b9ea5a6 4800struct rt6_nh {
8d1c802b 4801 struct fib6_info *fib6_info;
6b9ea5a6 4802 struct fib6_config r_cfg;
6b9ea5a6
RP		 4803 struct list_head next;
4804};
4805
d4ead6b3
DA		 4806static int ip6_route_info_append(struct net *net,
4807 struct list_head *rt6_nh_list,
8d1c802b
DA		 4808 struct fib6_info *rt,
4809 struct fib6_config *r_cfg)
6b9ea5a6
RP		 4810{
4811 struct rt6_nh *nh;
6b9ea5a6
RP		 4812 int err = -EEXIST;
4813
4814 list_for_each_entry(nh, rt6_nh_list, next) {
8d1c802b
DA		 4815 /* check if fib6_info already exists */
4816 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
6b9ea5a6
RP		 4817 return err;
4818 }
4819
4820 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
4821 if (!nh)
4822 return -ENOMEM;
8d1c802b 4823 nh->fib6_info = rt;
6b9ea5a6
RP		 4824 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
4825 list_add_tail(&nh->next, rt6_nh_list);
4826
4827 return 0;
4828}
4829
8d1c802b
DA		 4830static void ip6_route_mpath_notify(struct fib6_info *rt,
4831 struct fib6_info *rt_last,
3b1137fe
DA		 4832 struct nl_info *info,
4833 __u16 nlflags)
4834{
4835 /* if this is an APPEND route, then rt points to the first route
4836 * inserted and rt_last points to last route inserted. Userspace
4837 * wants a consistent dump of the route which starts at the first
4838 * nexthop. Since sibling routes are always added at the end of
4839 * the list, find the first sibling of the last route appended
4840 */
93c2fb25
DA		 4841 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
4842 rt = list_first_entry(&rt_last->fib6_siblings,
8d1c802b 4843 struct fib6_info,
93c2fb25 4844 fib6_siblings);
3b1137fe
DA		 4845 }
4846
4847 if (rt)
4848 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
4849}
4850
333c4301
DA		 4851static int ip6_route_multipath_add(struct fib6_config *cfg,
4852 struct netlink_ext_ack *extack)
51ebd318 4853{
8d1c802b 4854 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
3b1137fe 4855 struct nl_info *info = &cfg->fc_nlinfo;
51ebd318
ND		 4856 struct fib6_config r_cfg;
4857 struct rtnexthop *rtnh;
8d1c802b 4858 struct fib6_info *rt;
6b9ea5a6
RP		 4859 struct rt6_nh *err_nh;
4860 struct rt6_nh *nh, *nh_safe;
3b1137fe 4861 __u16 nlflags;
51ebd318
ND		 4862 int remaining;
4863 int attrlen;
6b9ea5a6
RP		 4864 int err = 1;
4865 int nhn = 0;
4866 int replace = (cfg->fc_nlinfo.nlh &&
4867 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
4868 LIST_HEAD(rt6_nh_list);
51ebd318 4869
3b1137fe
DA		 4870 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
4871 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
4872 nlflags |= NLM_F_APPEND;
4873
35f1b4e9 4874 remaining = cfg->fc_mp_len;
51ebd318 4875 rtnh = (struct rtnexthop *)cfg->fc_mp;
51ebd318 4876
6b9ea5a6 4877 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
8d1c802b 4878 * fib6_info structs per nexthop
6b9ea5a6 4879 */
51ebd318
ND		 4880 while (rtnh_ok(rtnh, remaining)) {
4881 memcpy(&r_cfg, cfg, sizeof(*cfg));
4882 if (rtnh->rtnh_ifindex)
4883 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
4884
4885 attrlen = rtnh_attrlen(rtnh);
4886 if (attrlen > 0) {
4887 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
4888
4889 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
4890 if (nla) {
67b61f6c 4891 r_cfg.fc_gateway = nla_get_in6_addr(nla);
51ebd318
ND		 4892 r_cfg.fc_flags |= RTF_GATEWAY;
4893 }
19e42e45
RP		 4894 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
4895 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
4896 if (nla)
4897 r_cfg.fc_encap_type = nla_get_u16(nla);
51ebd318 4898 }
6b9ea5a6 4899
68e2ffde 4900 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
acb54e3c 4901 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
8c5b83f0
RP		 4902 if (IS_ERR(rt)) {
4903 err = PTR_ERR(rt);
4904 rt = NULL;
6b9ea5a6 4905 goto cleanup;
8c5b83f0 4906 }
b5d2d75e
DA		 4907 if (!rt6_qualify_for_ecmp(rt)) {
4908 err = -EINVAL;
4909 NL_SET_ERR_MSG(extack,
4910 "Device only routes can not be added for IPv6 using the multipath API.");
4911 fib6_info_release(rt);
4912 goto cleanup;
4913 }
6b9ea5a6 4914
1cf844c7 4915 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
398958ae 4916
d4ead6b3
DA		 4917 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
4918 rt, &r_cfg);
51ebd318 4919 if (err) {
93531c67 4920 fib6_info_release(rt);
6b9ea5a6
RP		 4921 goto cleanup;
4922 }
4923
4924 rtnh = rtnh_next(rtnh, &remaining);
4925 }
4926
3b1137fe
DA		 4927 /* for add and replace send one notification with all nexthops.
4928 * Skip the notification in fib6_add_rt2node and send one with
4929 * the full route when done
4930 */
4931 info->skip_notify = 1;
4932
6b9ea5a6
RP		 4933 err_nh = NULL;
4934 list_for_each_entry(nh, &rt6_nh_list, next) {
8d1c802b
DA		 4935 err = __ip6_ins_rt(nh->fib6_info, info, extack);
4936 fib6_info_release(nh->fib6_info);
93531c67 4937
f7225172
DA		 4938 if (!err) {
4939 /* save reference to last route successfully inserted */
4940 rt_last = nh->fib6_info;
4941
4942 /* save reference to first route for notification */
4943 if (!rt_notif)
4944 rt_notif = nh->fib6_info;
4945 }
3b1137fe 4946
8d1c802b
DA		 4947 /* nh->fib6_info is used or freed at this point, reset to NULL*/
4948 nh->fib6_info = NULL;
6b9ea5a6
RP		 4949 if (err) {
4950 if (replace && nhn)
a5a82d84
JK		 4951 NL_SET_ERR_MSG_MOD(extack,
4952 "multipath route replace failed (check consistency of installed routes)");
6b9ea5a6
RP		 4953 err_nh = nh;
4954 goto add_errout;
51ebd318 4955 }
6b9ea5a6 4956
1a72418b 4957 /* Because each route is added like a single route we remove
27596472
MK		 4958 * these flags after the first nexthop: if there is a collision,
4959 * we have already failed to add the first nexthop:
4960 * fib6_add_rt2node() has rejected it; when replacing, old
4961 * nexthops have been replaced by first new, the rest should
4962 * be added to it.
1a72418b 4963 */
27596472
MK		 4964 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
4965 NLM_F_REPLACE);
6b9ea5a6
RP		 4966 nhn++;
4967 }
4968
3b1137fe
DA		 4969 /* success ... tell user about new route */
4970 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
6b9ea5a6
RP		 4971 goto cleanup;
4972
4973add_errout:
3b1137fe
DA		 4974 /* send notification for routes that were added so that
4975 * the delete notifications sent by ip6_route_del are
4976 * coherent
4977 */
4978 if (rt_notif)
4979 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
4980
6b9ea5a6
RP		 4981 /* Delete routes that were already added */
4982 list_for_each_entry(nh, &rt6_nh_list, next) {
4983 if (err_nh == nh)
4984 break;
333c4301 4985 ip6_route_del(&nh->r_cfg, extack);
6b9ea5a6
RP		 4986 }
4987
4988cleanup:
4989 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
8d1c802b
DA		 4990 if (nh->fib6_info)
4991 fib6_info_release(nh->fib6_info);
6b9ea5a6
RP		 4992 list_del(&nh->next);
4993 kfree(nh);
4994 }
4995
4996 return err;
4997}
4998
333c4301
DA		 4999static int ip6_route_multipath_del(struct fib6_config *cfg,
5000 struct netlink_ext_ack *extack)
6b9ea5a6
RP		 5001{
5002 struct fib6_config r_cfg;
5003 struct rtnexthop *rtnh;
5004 int remaining;
5005 int attrlen;
5006 int err = 1, last_err = 0;
5007
5008 remaining = cfg->fc_mp_len;
5009 rtnh = (struct rtnexthop *)cfg->fc_mp;
5010
5011 /* Parse a Multipath Entry */
5012 while (rtnh_ok(rtnh, remaining)) {
5013 memcpy(&r_cfg, cfg, sizeof(*cfg));
5014 if (rtnh->rtnh_ifindex)
5015 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5016
5017 attrlen = rtnh_attrlen(rtnh);
5018 if (attrlen > 0) {
5019 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5020
5021 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5022 if (nla) {
5023 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
5024 r_cfg.fc_flags |= RTF_GATEWAY;
5025 }
5026 }
333c4301 5027 err = ip6_route_del(&r_cfg, extack);
6b9ea5a6
RP		 5028 if (err)
5029 last_err = err;
5030
51ebd318
ND		 5031 rtnh = rtnh_next(rtnh, &remaining);
5032 }
5033
5034 return last_err;
5035}
5036
c21ef3e3
DA		 5037static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5038 struct netlink_ext_ack *extack)
1da177e4 5039{
86872cb5
TG		 5040 struct fib6_config cfg;
5041 int err;
1da177e4 5042
333c4301 5043 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 5044 if (err < 0)
5045 return err;
5046
51ebd318 5047 if (cfg.fc_mp)
333c4301 5048 return ip6_route_multipath_del(&cfg, extack);
0ae81335
DA		 5049 else {
5050 cfg.fc_delete_all_nh = 1;
333c4301 5051 return ip6_route_del(&cfg, extack);
0ae81335 5052 }
1da177e4
LT		 5053}
5054
c21ef3e3
DA		 5055static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5056 struct netlink_ext_ack *extack)
1da177e4 5057{
86872cb5
TG		 5058 struct fib6_config cfg;
5059 int err;
1da177e4 5060
333c4301 5061 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
86872cb5
TG		 5062 if (err < 0)
5063 return err;
5064
67f69513
DA		 5065 if (cfg.fc_metric == 0)
5066 cfg.fc_metric = IP6_RT_PRIO_USER;
5067
51ebd318 5068 if (cfg.fc_mp)
333c4301 5069 return ip6_route_multipath_add(&cfg, extack);
51ebd318 5070 else
acb54e3c 5071 return ip6_route_add(&cfg, GFP_KERNEL, extack);
1da177e4
LT		 5072}
5073
a1b7a1f0
DA		 5074/* add the overhead of this fib6_nh to nexthop_len */
5075static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
339bf98f 5076{
a1b7a1f0 5077 int *nexthop_len = arg;
beb1afac 5078
a1b7a1f0
DA		 5079 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5080 + NLA_ALIGN(sizeof(struct rtnexthop))
5081 + nla_total_size(16); /* RTA_GATEWAY */
5082
5083 if (nh->fib_nh_lws) {
5084 /* RTA_ENCAP_TYPE */
5085 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5086 /* RTA_ENCAP */
5087 *nexthop_len += nla_total_size(2);
5088 }
f88d8ea6 5089
a1b7a1f0
DA		 5090 return 0;
5091}
beb1afac 5092
a1b7a1f0
DA		 5093static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5094{
5095 int nexthop_len;
5096
5097 if (f6i->nh) {
5098 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5099 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5100 &nexthop_len);
5101 } else {
5102 struct fib6_nh *nh = f6i->fib6_nh;
5103
5104 nexthop_len = 0;
5105 if (f6i->fib6_nsiblings) {
5106 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
5107 + NLA_ALIGN(sizeof(struct rtnexthop))
5108 + nla_total_size(16) /* RTA_GATEWAY */
5109 + lwtunnel_get_encap_size(nh->fib_nh_lws);
5110
5111 nexthop_len *= f6i->fib6_nsiblings;
5112 }
5113 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
beb1afac
DA		 5114 }
5115
339bf98f
TG		 5116 return NLMSG_ALIGN(sizeof(struct rtmsg))
5117 + nla_total_size(16) /* RTA_SRC */
5118 + nla_total_size(16) /* RTA_DST */
5119 + nla_total_size(16) /* RTA_GATEWAY */
5120 + nla_total_size(16) /* RTA_PREFSRC */
5121 + nla_total_size(4) /* RTA_TABLE */
5122 + nla_total_size(4) /* RTA_IIF */
5123 + nla_total_size(4) /* RTA_OIF */
5124 + nla_total_size(4) /* RTA_PRIORITY */
6a2b9ce0 5125 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
ea697639 5126 + nla_total_size(sizeof(struct rta_cacheinfo))
c78ba6d6 5127 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
19e42e45 5128 + nla_total_size(1) /* RTA_PREF */
beb1afac
DA		 5129 + nexthop_len;
5130}
5131
f88d8ea6
DA		 5132static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5133 unsigned char *flags)
5134{
5135 if (nexthop_is_multipath(nh)) {
5136 struct nlattr *mp;
5137
5138 mp = nla_nest_start(skb, RTA_MULTIPATH);
5139 if (!mp)
5140 goto nla_put_failure;
5141
5142 if (nexthop_mpath_fill_node(skb, nh))
5143 goto nla_put_failure;
5144
5145 nla_nest_end(skb, mp);
5146 } else {
5147 struct fib6_nh *fib6_nh;
5148
5149 fib6_nh = nexthop_fib6_nh(nh);
5150 if (fib_nexthop_info(skb, &fib6_nh->nh_common,
5151 flags, false) < 0)
5152 goto nla_put_failure;
5153 }
5154
5155 return 0;
5156
5157nla_put_failure:
5158 return -EMSGSIZE;
5159}
5160
d4ead6b3 5161static int rt6_fill_node(struct net *net, struct sk_buff *skb,
8d1c802b 5162 struct fib6_info *rt, struct dst_entry *dst,
d4ead6b3 5163 struct in6_addr *dest, struct in6_addr *src,
15e47304 5164 int iif, int type, u32 portid, u32 seq,
f8cfe2ce 5165 unsigned int flags)
1da177e4 5166{
22d0bd82
XL		 5167 struct rt6_info *rt6 = (struct rt6_info *)dst;
5168 struct rt6key *rt6_dst, *rt6_src;
5169 u32 *pmetrics, table, rt6_flags;
f88d8ea6 5170 unsigned char nh_flags = 0;
2d7202bf 5171 struct nlmsghdr *nlh;
22d0bd82 5172 struct rtmsg *rtm;
d4ead6b3 5173 long expires = 0;
1da177e4 5174
15e47304 5175 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
38308473 5176 if (!nlh)
26932566 5177 return -EMSGSIZE;
2d7202bf 5178
22d0bd82
XL		 5179 if (rt6) {
5180 rt6_dst = &rt6->rt6i_dst;
5181 rt6_src = &rt6->rt6i_src;
5182 rt6_flags = rt6->rt6i_flags;
5183 } else {
5184 rt6_dst = &rt->fib6_dst;
5185 rt6_src = &rt->fib6_src;
5186 rt6_flags = rt->fib6_flags;
5187 }
5188
2d7202bf 5189 rtm = nlmsg_data(nlh);
1da177e4 5190 rtm->rtm_family = AF_INET6;
22d0bd82
XL		 5191 rtm->rtm_dst_len = rt6_dst->plen;
5192 rtm->rtm_src_len = rt6_src->plen;
1da177e4 5193 rtm->rtm_tos = 0;
93c2fb25
DA		 5194 if (rt->fib6_table)
5195 table = rt->fib6_table->tb6_id;
c71099ac 5196 else
9e762a4a 5197 table = RT6_TABLE_UNSPEC;
97f0082a 5198 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
c78679e8
DM		 5199 if (nla_put_u32(skb, RTA_TABLE, table))
5200 goto nla_put_failure;
e8478e80
DA		 5201
5202 rtm->rtm_type = rt->fib6_type;
1da177e4
LT		 5203 rtm->rtm_flags = 0;
5204 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
93c2fb25 5205 rtm->rtm_protocol = rt->fib6_protocol;
1da177e4 5206
22d0bd82 5207 if (rt6_flags & RTF_CACHE)
1da177e4
LT		 5208 rtm->rtm_flags |= RTM_F_CLONED;
5209
d4ead6b3
DA		 5210 if (dest) {
5211 if (nla_put_in6_addr(skb, RTA_DST, dest))
c78679e8 5212 goto nla_put_failure;
1ab1457c 5213 rtm->rtm_dst_len = 128;
1da177e4 5214 } else if (rtm->rtm_dst_len)
22d0bd82 5215 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
c78679e8 5216 goto nla_put_failure;
1da177e4
LT		 5217#ifdef CONFIG_IPV6_SUBTREES
5218 if (src) {
930345ea 5219 if (nla_put_in6_addr(skb, RTA_SRC, src))
c78679e8 5220 goto nla_put_failure;
1ab1457c 5221 rtm->rtm_src_len = 128;
c78679e8 5222 } else if (rtm->rtm_src_len &&
22d0bd82 5223 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
c78679e8 5224 goto nla_put_failure;
1da177e4 5225#endif
7bc570c8
YH		 5226 if (iif) {
5227#ifdef CONFIG_IPV6_MROUTE
22d0bd82 5228 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
fd61c6ba
DA		 5229 int err = ip6mr_get_route(net, skb, rtm, portid);
5230
5231 if (err == 0)
5232 return 0;
5233 if (err < 0)
5234 goto nla_put_failure;
7bc570c8
YH		 5235 } else
5236#endif
c78679e8
DM		 5237 if (nla_put_u32(skb, RTA_IIF, iif))
5238 goto nla_put_failure;
d4ead6b3 5239 } else if (dest) {
1da177e4 5240 struct in6_addr saddr_buf;
d4ead6b3 5241 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
930345ea 5242 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 5243 goto nla_put_failure;
1da177e4 5244 }
2d7202bf 5245
93c2fb25 5246 if (rt->fib6_prefsrc.plen) {
c3968a85 5247 struct in6_addr saddr_buf;
93c2fb25 5248 saddr_buf = rt->fib6_prefsrc.addr;
930345ea 5249 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 5250 goto nla_put_failure;
c3968a85
DW		 5251 }
5252
d4ead6b3
DA		 5253 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5254 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
2d7202bf
TG		 5255 goto nla_put_failure;
5256
93c2fb25 5257 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
c78679e8 5258 goto nla_put_failure;
8253947e 5259
beb1afac
DA		 5260 /* For multipath routes, walk the siblings list and add
5261 * each as a nexthop within RTA_MULTIPATH.
5262 */
22d0bd82
XL		 5263 if (rt6) {
5264 if (rt6_flags & RTF_GATEWAY &&
5265 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5266 goto nla_put_failure;
5267
5268 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5269 goto nla_put_failure;
5270 } else if (rt->fib6_nsiblings) {
8d1c802b 5271 struct fib6_info *sibling, *next_sibling;
beb1afac
DA		 5272 struct nlattr *mp;
5273
ae0be8de 5274 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
beb1afac
DA		 5275 if (!mp)
5276 goto nla_put_failure;
5277
1cf844c7
DA		 5278 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5279 rt->fib6_nh->fib_nh_weight) < 0)
beb1afac
DA		 5280 goto nla_put_failure;
5281
5282 list_for_each_entry_safe(sibling, next_sibling,
93c2fb25 5283 &rt->fib6_siblings, fib6_siblings) {
1cf844c7
DA		 5284 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5285 sibling->fib6_nh->fib_nh_weight) < 0)
beb1afac
DA		 5286 goto nla_put_failure;
5287 }
5288
5289 nla_nest_end(skb, mp);
f88d8ea6
DA		 5290 } else if (rt->nh) {
5291 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5292 goto nla_put_failure;
ecc5663c 5293
f88d8ea6
DA		 5294 if (nexthop_is_blackhole(rt->nh))
5295 rtm->rtm_type = RTN_BLACKHOLE;
5296
5297 if (rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5298 goto nla_put_failure;
5299
5300 rtm->rtm_flags |= nh_flags;
5301 } else {
1cf844c7 5302 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common,
ecc5663c 5303 &nh_flags, false) < 0)
beb1afac 5304 goto nla_put_failure;
ecc5663c
DA		 5305
5306 rtm->rtm_flags |= nh_flags;
beb1afac
DA		 5307 }
5308
22d0bd82 5309 if (rt6_flags & RTF_EXPIRES) {
14895687
DA		 5310 expires = dst ? dst->expires : rt->expires;
5311 expires -= jiffies;
5312 }
69cdf8f9 5313
d4ead6b3 5314 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
e3703b3d 5315 goto nla_put_failure;
2d7202bf 5316
22d0bd82 5317 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
c78ba6d6
LR		 5318 goto nla_put_failure;
5319
19e42e45 5320
053c095a
JB		 5321 nlmsg_end(skb, nlh);
5322 return 0;
2d7202bf
TG		 5323
5324nla_put_failure:
26932566
PM		 5325 nlmsg_cancel(skb, nlh);
5326 return -EMSGSIZE;
1da177e4
LT		 5327}
5328
2c170e07
DA		 5329static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5330{
5331 const struct net_device *dev = arg;
5332
5333 if (nh->fib_nh_dev == dev)
5334 return 1;
5335
5336 return 0;
5337}
5338
13e38901
DA		 5339static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5340 const struct net_device *dev)
5341{
2c170e07
DA		 5342 if (f6i->nh) {
5343 struct net_device *_dev = (struct net_device *)dev;
5344
5345 return !!nexthop_for_each_fib6_nh(f6i->nh,
5346 fib6_info_nh_uses_dev,
5347 _dev);
5348 }
5349
1cf844c7 5350 if (f6i->fib6_nh->fib_nh_dev == dev)
13e38901
DA		 5351 return true;
5352
5353 if (f6i->fib6_nsiblings) {
5354 struct fib6_info *sibling, *next_sibling;
5355
5356 list_for_each_entry_safe(sibling, next_sibling,
5357 &f6i->fib6_siblings, fib6_siblings) {
1cf844c7 5358 if (sibling->fib6_nh->fib_nh_dev == dev)
13e38901
DA		 5359 return true;
5360 }
5361 }
5362
5363 return false;
5364}
5365
8d1c802b 5366int rt6_dump_route(struct fib6_info *rt, void *p_arg)
1da177e4
LT		 5367{
5368 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
13e38901
DA		 5369 struct fib_dump_filter *filter = &arg->filter;
5370 unsigned int flags = NLM_F_MULTI;
1f17e2f2
DA		 5371 struct net *net = arg->net;
5372
421842ed 5373 if (rt == net->ipv6.fib6_null_entry)
1f17e2f2 5374 return 0;
1da177e4 5375
13e38901
DA		 5376 if ((filter->flags & RTM_F_PREFIX) &&
5377 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5378 /* success since this is not a prefix route */
5379 return 1;
5380 }
5381 if (filter->filter_set) {
5382 if ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5383 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5384 (filter->protocol && rt->fib6_protocol != filter->protocol)) {
f8cfe2ce
DA		 5385 return 1;
5386 }
13e38901 5387 flags |= NLM_F_DUMP_FILTERED;
f8cfe2ce 5388 }
1da177e4 5389
d4ead6b3
DA		 5390 return rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL, 0,
5391 RTM_NEWROUTE, NETLINK_CB(arg->cb->skb).portid,
13e38901 5392 arg->cb->nlh->nlmsg_seq, flags);
1da177e4
LT		 5393}
5394
0eff0a27
JK		 5395static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5396 const struct nlmsghdr *nlh,
5397 struct nlattr **tb,
5398 struct netlink_ext_ack *extack)
5399{
5400 struct rtmsg *rtm;
5401 int i, err;
5402
5403 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5404 NL_SET_ERR_MSG_MOD(extack,
5405 "Invalid header for get route request");
5406 return -EINVAL;
5407 }
5408
5409 if (!netlink_strict_get_check(skb))
8cb08174
JB		 5410 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5411 rtm_ipv6_policy, extack);
0eff0a27
JK		 5412
5413 rtm = nlmsg_data(nlh);
5414 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5415 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5416 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5417 rtm->rtm_type) {
5418 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5419 return -EINVAL;
5420 }
5421 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5422 NL_SET_ERR_MSG_MOD(extack,
5423 "Invalid flags for get route request");
5424 return -EINVAL;
5425 }
5426
8cb08174
JB		 5427 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5428 rtm_ipv6_policy, extack);
0eff0a27
JK		 5429 if (err)
5430 return err;
5431
5432 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5433 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5434 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5435 return -EINVAL;
5436 }
5437
5438 for (i = 0; i <= RTA_MAX; i++) {
5439 if (!tb[i])
5440 continue;
5441
5442 switch (i) {
5443 case RTA_SRC:
5444 case RTA_DST:
5445 case RTA_IIF:
5446 case RTA_OIF:
5447 case RTA_MARK:
5448 case RTA_UID:
5449 case RTA_SPORT:
5450 case RTA_DPORT:
5451 case RTA_IP_PROTO:
5452 break;
5453 default:
5454 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
5455 return -EINVAL;
5456 }
5457 }
5458
5459 return 0;
5460}
5461
c21ef3e3
DA		 5462static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5463 struct netlink_ext_ack *extack)
1da177e4 5464{
3b1e0a65 5465 struct net *net = sock_net(in_skb->sk);
ab364a6f 5466 struct nlattr *tb[RTA_MAX+1];
18c3a61c 5467 int err, iif = 0, oif = 0;
a68886a6 5468 struct fib6_info *from;
18c3a61c 5469 struct dst_entry *dst;
ab364a6f 5470 struct rt6_info *rt;
1da177e4 5471 struct sk_buff *skb;
ab364a6f 5472 struct rtmsg *rtm;
744486d4 5473 struct flowi6 fl6 = {};
18c3a61c 5474 bool fibmatch;
1da177e4 5475
0eff0a27 5476 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
ab364a6f
TG		 5477 if (err < 0)
5478 goto errout;
1da177e4 5479
ab364a6f 5480 err = -EINVAL;
38b7097b
HFS		 5481 rtm = nlmsg_data(nlh);
5482 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
18c3a61c 5483 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
1da177e4 5484
ab364a6f
TG		 5485 if (tb[RTA_SRC]) {
5486 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
5487 goto errout;
5488
4e3fd7a0 5489 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
ab364a6f
TG		 5490 }
5491
5492 if (tb[RTA_DST]) {
5493 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
5494 goto errout;
5495
4e3fd7a0 5496 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
ab364a6f
TG		 5497 }
5498
5499 if (tb[RTA_IIF])
5500 iif = nla_get_u32(tb[RTA_IIF]);
5501
5502 if (tb[RTA_OIF])
72331bc0 5503 oif = nla_get_u32(tb[RTA_OIF]);
1da177e4 5504
2e47b291
LC		 5505 if (tb[RTA_MARK])
5506 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
5507
622ec2c9
LC		 5508 if (tb[RTA_UID])
5509 fl6.flowi6_uid = make_kuid(current_user_ns(),
5510 nla_get_u32(tb[RTA_UID]));
5511 else
5512 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
5513
eacb9384
RP		 5514 if (tb[RTA_SPORT])
5515 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
5516
5517 if (tb[RTA_DPORT])
5518 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
5519
5520 if (tb[RTA_IP_PROTO]) {
5521 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
5e1a99ea
HL		 5522 &fl6.flowi6_proto, AF_INET6,
5523 extack);
eacb9384
RP		 5524 if (err)
5525 goto errout;
5526 }
5527
1da177e4
LT		 5528 if (iif) {
5529 struct net_device *dev;
72331bc0
SL		 5530 int flags = 0;
5531
121622db
FW		 5532 rcu_read_lock();
5533
5534 dev = dev_get_by_index_rcu(net, iif);
1da177e4 5535 if (!dev) {
121622db 5536 rcu_read_unlock();
1da177e4 5537 err = -ENODEV;
ab364a6f 5538 goto errout;
1da177e4 5539 }
72331bc0
SL		 5540
5541 fl6.flowi6_iif = iif;
5542
5543 if (!ipv6_addr_any(&fl6.saddr))
5544 flags |= RT6_LOOKUP_F_HAS_SADDR;
5545
b75cc8f9 5546 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
121622db
FW		 5547
5548 rcu_read_unlock();
72331bc0
SL		 5549 } else {
5550 fl6.flowi6_oif = oif;
5551
58acfd71 5552 dst = ip6_route_output(net, NULL, &fl6);
18c3a61c
RP		 5553 }
5554
18c3a61c
RP		 5555
5556 rt = container_of(dst, struct rt6_info, dst);
5557 if (rt->dst.error) {
5558 err = rt->dst.error;
5559 ip6_rt_put(rt);
5560 goto errout;
1da177e4
LT		 5561 }
5562
9d6acb3b
WC		 5563 if (rt == net->ipv6.ip6_null_entry) {
5564 err = rt->dst.error;
5565 ip6_rt_put(rt);
5566 goto errout;
5567 }
5568
ab364a6f 5569 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
38308473 5570 if (!skb) {
94e187c0 5571 ip6_rt_put(rt);
ab364a6f
TG		 5572 err = -ENOBUFS;
5573 goto errout;
5574 }
1da177e4 5575
d8d1f30b 5576 skb_dst_set(skb, &rt->dst);
a68886a6
DA		 5577
5578 rcu_read_lock();
5579 from = rcu_dereference(rt->from);
886b7a50
MKL		 5580 if (from) {
5581 if (fibmatch)
5582 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
5583 iif, RTM_NEWROUTE,
5584 NETLINK_CB(in_skb).portid,
5585 nlh->nlmsg_seq, 0);
5586 else
5587 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
5588 &fl6.saddr, iif, RTM_NEWROUTE,
5589 NETLINK_CB(in_skb).portid,
5590 nlh->nlmsg_seq, 0);
5591 } else {
5592 err = -ENETUNREACH;
5593 }
a68886a6
DA		 5594 rcu_read_unlock();
5595
1da177e4 5596 if (err < 0) {
ab364a6f
TG		 5597 kfree_skb(skb);
5598 goto errout;
1da177e4
LT		 5599 }
5600
15e47304 5601 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
ab364a6f 5602errout:
1da177e4 5603 return err;
1da177e4
LT		 5604}
5605
8d1c802b 5606void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
37a1d361 5607 unsigned int nlm_flags)
1da177e4
LT		 5608{
5609 struct sk_buff *skb;
5578689a 5610 struct net *net = info->nl_net;
528c4ceb
DL		 5611 u32 seq;
5612 int err;
5613
5614 err = -ENOBUFS;
38308473 5615 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
86872cb5 5616
19e42e45 5617 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
38308473 5618 if (!skb)
21713ebc
TG		 5619 goto errout;
5620
d4ead6b3
DA		 5621 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
5622 event, info->portid, seq, nlm_flags);
26932566
PM		 5623 if (err < 0) {
5624 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
5625 WARN_ON(err == -EMSGSIZE);
5626 kfree_skb(skb);
5627 goto errout;
5628 }
15e47304 5629 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
1ce85fe4
PNA		 5630 info->nlh, gfp_any());
5631 return;
21713ebc
TG		 5632errout:
5633 if (err < 0)
5578689a 5634 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
1da177e4
LT		 5635}
5636
19a3b7ee
DA		 5637void fib6_rt_update(struct net *net, struct fib6_info *rt,
5638 struct nl_info *info)
5639{
5640 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
5641 struct sk_buff *skb;
5642 int err = -ENOBUFS;
5643
5644 /* call_fib6_entry_notifiers will be removed when in-kernel notifier
5645 * is implemented and supported for nexthop objects
5646 */
5647 call_fib6_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, rt, NULL);
5648
5649 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
5650 if (!skb)
5651 goto errout;
5652
5653 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
5654 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
5655 if (err < 0) {
5656 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
5657 WARN_ON(err == -EMSGSIZE);
5658 kfree_skb(skb);
5659 goto errout;
5660 }
5661 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
5662 info->nlh, gfp_any());
5663 return;
5664errout:
5665 if (err < 0)
5666 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
5667}
5668
8ed67789 5669static int ip6_route_dev_notify(struct notifier_block *this,
351638e7 5670 unsigned long event, void *ptr)
8ed67789 5671{
351638e7 5672 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
c346dca1 5673 struct net *net = dev_net(dev);
8ed67789 5674
242d3a49
WC		 5675 if (!(dev->flags & IFF_LOOPBACK))
5676 return NOTIFY_OK;
5677
5678 if (event == NETDEV_REGISTER) {
1cf844c7 5679 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
d8d1f30b 5680 net->ipv6.ip6_null_entry->dst.dev = dev;
8ed67789
DL		 5681 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
5682#ifdef CONFIG_IPV6_MULTIPLE_TABLES
d8d1f30b 5683 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
8ed67789 5684 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
d8d1f30b 5685 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
8ed67789 5686 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
242d3a49 5687#endif
76da0704
WC		 5688 } else if (event == NETDEV_UNREGISTER &&
5689 dev->reg_state != NETREG_UNREGISTERED) {
5690 /* NETDEV_UNREGISTER could be fired for multiple times by
5691 * netdev_wait_allrefs(). Make sure we only call this once.
5692 */
12d94a80 5693 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
242d3a49 5694#ifdef CONFIG_IPV6_MULTIPLE_TABLES
12d94a80
ED		 5695 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
5696 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
8ed67789
DL		 5697#endif
5698 }
5699
5700 return NOTIFY_OK;
5701}
5702
1da177e4
LT		 5703/*
5704 * /proc
5705 */
5706
5707#ifdef CONFIG_PROC_FS
1da177e4
LT		 5708static int rt6_stats_seq_show(struct seq_file *seq, void *v)
5709{
69ddb805 5710 struct net *net = (struct net *)seq->private;
1da177e4 5711 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
69ddb805
DL		 5712 net->ipv6.rt6_stats->fib_nodes,
5713 net->ipv6.rt6_stats->fib_route_nodes,
81eb8447 5714 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
69ddb805
DL		 5715 net->ipv6.rt6_stats->fib_rt_entries,
5716 net->ipv6.rt6_stats->fib_rt_cache,
fc66f95c 5717 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
69ddb805 5718 net->ipv6.rt6_stats->fib_discarded_routes);
1da177e4
LT		 5719
5720 return 0;
5721}
1da177e4
LT		 5722#endif /* CONFIG_PROC_FS */
5723
5724#ifdef CONFIG_SYSCTL
5725
1da177e4 5726static
fe2c6338 5727int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
1da177e4
LT		 5728 void __user *buffer, size_t *lenp, loff_t *ppos)
5729{
c486da34
LAG		 5730 struct net *net;
5731 int delay;
f0fb9b28 5732 int ret;
c486da34 5733 if (!write)
1da177e4 5734 return -EINVAL;
c486da34
LAG		 5735
5736 net = (struct net *)ctl->extra1;
5737 delay = net->ipv6.sysctl.flush_delay;
f0fb9b28
AP		 5738 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5739 if (ret)
5740 return ret;
5741
2ac3ac8f 5742 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
c486da34 5743 return 0;
1da177e4
LT		 5744}
5745
7c6bb7d2
DA		 5746static int zero;
5747static int one = 1;
5748
ed792e28 5749static struct ctl_table ipv6_route_table_template[] = {
1ab1457c 5750 {
1da177e4 5751 .procname = "flush",
4990509f 5752 .data = &init_net.ipv6.sysctl.flush_delay,
1da177e4 5753 .maxlen = sizeof(int),
89c8b3a1 5754 .mode = 0200,
6d9f239a 5755 .proc_handler = ipv6_sysctl_rtcache_flush
1da177e4
LT		 5756 },
5757 {
1da177e4 5758 .procname = "gc_thresh",
9a7ec3a9 5759 .data = &ip6_dst_ops_template.gc_thresh,
1da177e4
LT		 5760 .maxlen = sizeof(int),
5761 .mode = 0644,
6d9f239a 5762 .proc_handler = proc_dointvec,
1da177e4
LT		 5763 },
5764 {
1da177e4 5765 .procname = "max_size",
4990509f 5766 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
1da177e4
LT		 5767 .maxlen = sizeof(int),
5768 .mode = 0644,
6d9f239a 5769 .proc_handler = proc_dointvec,
1da177e4
LT		 5770 },
5771 {
1da177e4 5772 .procname = "gc_min_interval",
4990509f 5773 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 5774 .maxlen = sizeof(int),
5775 .mode = 0644,
6d9f239a 5776 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5777 },
5778 {
1da177e4 5779 .procname = "gc_timeout",
4990509f 5780 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
1da177e4
LT		 5781 .maxlen = sizeof(int),
5782 .mode = 0644,
6d9f239a 5783 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5784 },
5785 {
1da177e4 5786 .procname = "gc_interval",
4990509f 5787 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
1da177e4
LT		 5788 .maxlen = sizeof(int),
5789 .mode = 0644,
6d9f239a 5790 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5791 },
5792 {
1da177e4 5793 .procname = "gc_elasticity",
4990509f 5794 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
1da177e4
LT		 5795 .maxlen = sizeof(int),
5796 .mode = 0644,
f3d3f616 5797 .proc_handler = proc_dointvec,
1da177e4
LT		 5798 },
5799 {
1da177e4 5800 .procname = "mtu_expires",
4990509f 5801 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
1da177e4
LT		 5802 .maxlen = sizeof(int),
5803 .mode = 0644,
6d9f239a 5804 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 5805 },
5806 {
1da177e4 5807 .procname = "min_adv_mss",
4990509f 5808 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
1da177e4
LT		 5809 .maxlen = sizeof(int),
5810 .mode = 0644,
f3d3f616 5811 .proc_handler = proc_dointvec,
1da177e4
LT		 5812 },
5813 {
1da177e4 5814 .procname = "gc_min_interval_ms",
4990509f 5815 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 5816 .maxlen = sizeof(int),
5817 .mode = 0644,
6d9f239a 5818 .proc_handler = proc_dointvec_ms_jiffies,
1da177e4 5819 },
7c6bb7d2
DA		 5820 {
5821 .procname = "skip_notify_on_dev_down",
5822 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
5823 .maxlen = sizeof(int),
5824 .mode = 0644,
5825 .proc_handler = proc_dointvec,
5826 .extra1 = &zero,
5827 .extra2 = &one,
5828 },
f8572d8f 5829 { }
1da177e4
LT		 5830};
5831
2c8c1e72 5832struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
760f2d01
DL		 5833{
5834 struct ctl_table *table;
5835
5836 table = kmemdup(ipv6_route_table_template,
5837 sizeof(ipv6_route_table_template),
5838 GFP_KERNEL);
5ee09105
YH		 5839
5840 if (table) {
5841 table[0].data = &net->ipv6.sysctl.flush_delay;
c486da34 5842 table[0].extra1 = net;
86393e52 5843 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
5ee09105
YH		 5844 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
5845 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
5846 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
5847 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
5848 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
5849 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
5850 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
9c69fabe 5851 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
7c6bb7d2 5852 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
464dc801
EB		 5853
5854 /* Don't export sysctls to unprivileged users */
5855 if (net->user_ns != &init_user_ns)
5856 table[0].procname = NULL;
5ee09105
YH		 5857 }
5858
760f2d01
DL		 5859 return table;
5860}
1da177e4
LT		 5861#endif
5862
2c8c1e72 5863static int __net_init ip6_route_net_init(struct net *net)
cdb18761 5864{
633d424b 5865 int ret = -ENOMEM;
8ed67789 5866
86393e52
AD		 5867 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
5868 sizeof(net->ipv6.ip6_dst_ops));
f2fc6a54 5869
fc66f95c
ED		 5870 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
5871 goto out_ip6_dst_ops;
5872
1cf844c7 5873 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
421842ed
DA		 5874 if (!net->ipv6.fib6_null_entry)
5875 goto out_ip6_dst_entries;
1cf844c7
DA		 5876 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
5877 sizeof(*net->ipv6.fib6_null_entry));
421842ed 5878
8ed67789
DL		 5879 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
5880 sizeof(*net->ipv6.ip6_null_entry),
5881 GFP_KERNEL);
5882 if (!net->ipv6.ip6_null_entry)
421842ed 5883 goto out_fib6_null_entry;
d8d1f30b 5884 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 5885 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
5886 ip6_template_metrics, true);
8ed67789
DL		 5887
5888#ifdef CONFIG_IPV6_MULTIPLE_TABLES
feca7d8c 5889 net->ipv6.fib6_has_custom_rules = false;
8ed67789
DL		 5890 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
5891 sizeof(*net->ipv6.ip6_prohibit_entry),
5892 GFP_KERNEL);
68fffc67
PZ		 5893 if (!net->ipv6.ip6_prohibit_entry)
5894 goto out_ip6_null_entry;
d8d1f30b 5895 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 5896 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
5897 ip6_template_metrics, true);
8ed67789
DL		 5898
5899 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
5900 sizeof(*net->ipv6.ip6_blk_hole_entry),
5901 GFP_KERNEL);
68fffc67
PZ		 5902 if (!net->ipv6.ip6_blk_hole_entry)
5903 goto out_ip6_prohibit_entry;
d8d1f30b 5904 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 5905 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
5906 ip6_template_metrics, true);
8ed67789
DL		 5907#endif
5908
b339a47c
PZ		 5909 net->ipv6.sysctl.flush_delay = 0;
5910 net->ipv6.sysctl.ip6_rt_max_size = 4096;
5911 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
5912 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
5913 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
5914 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
5915 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
5916 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
7c6bb7d2 5917 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
b339a47c 5918
6891a346
BT		 5919 net->ipv6.ip6_rt_gc_expire = 30*HZ;
5920
8ed67789
DL		 5921 ret = 0;
5922out:
5923 return ret;
f2fc6a54 5924
68fffc67
PZ		 5925#ifdef CONFIG_IPV6_MULTIPLE_TABLES
5926out_ip6_prohibit_entry:
5927 kfree(net->ipv6.ip6_prohibit_entry);
5928out_ip6_null_entry:
5929 kfree(net->ipv6.ip6_null_entry);
5930#endif
421842ed
DA		 5931out_fib6_null_entry:
5932 kfree(net->ipv6.fib6_null_entry);
fc66f95c
ED		 5933out_ip6_dst_entries:
5934 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
f2fc6a54 5935out_ip6_dst_ops:
f2fc6a54 5936 goto out;
cdb18761
DL		 5937}
5938
2c8c1e72 5939static void __net_exit ip6_route_net_exit(struct net *net)
cdb18761 5940{
421842ed 5941 kfree(net->ipv6.fib6_null_entry);
8ed67789
DL		 5942 kfree(net->ipv6.ip6_null_entry);
5943#ifdef CONFIG_IPV6_MULTIPLE_TABLES
5944 kfree(net->ipv6.ip6_prohibit_entry);
5945 kfree(net->ipv6.ip6_blk_hole_entry);
5946#endif
41bb78b4 5947 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
cdb18761
DL		 5948}
5949
d189634e
TG		 5950static int __net_init ip6_route_net_init_late(struct net *net)
5951{
5952#ifdef CONFIG_PROC_FS
c3506372
CH		 5953 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
5954 sizeof(struct ipv6_route_iter));
3617d949
CH		 5955 proc_create_net_single("rt6_stats", 0444, net->proc_net,
5956 rt6_stats_seq_show, NULL);
d189634e
TG		 5957#endif
5958 return 0;
5959}
5960
5961static void __net_exit ip6_route_net_exit_late(struct net *net)
5962{
5963#ifdef CONFIG_PROC_FS
ece31ffd
G		 5964 remove_proc_entry("ipv6_route", net->proc_net);
5965 remove_proc_entry("rt6_stats", net->proc_net);
d189634e
TG		 5966#endif
5967}
5968
cdb18761
DL		 5969static struct pernet_operations ip6_route_net_ops = {
5970 .init = ip6_route_net_init,
5971 .exit = ip6_route_net_exit,
5972};
5973
c3426b47
DM		 5974static int __net_init ipv6_inetpeer_init(struct net *net)
5975{
5976 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
5977
5978 if (!bp)
5979 return -ENOMEM;
5980 inet_peer_base_init(bp);
5981 net->ipv6.peers = bp;
5982 return 0;
5983}
5984
5985static void __net_exit ipv6_inetpeer_exit(struct net *net)
5986{
5987 struct inet_peer_base *bp = net->ipv6.peers;
5988
5989 net->ipv6.peers = NULL;
56a6b248 5990 inetpeer_invalidate_tree(bp);
c3426b47
DM		 5991 kfree(bp);
5992}
5993
2b823f72 5994static struct pernet_operations ipv6_inetpeer_ops = {
c3426b47
DM		 5995 .init = ipv6_inetpeer_init,
5996 .exit = ipv6_inetpeer_exit,
5997};
5998
d189634e
TG		 5999static struct pernet_operations ip6_route_net_late_ops = {
6000 .init = ip6_route_net_init_late,
6001 .exit = ip6_route_net_exit_late,
6002};
6003
8ed67789
DL		 6004static struct notifier_block ip6_route_dev_notifier = {
6005 .notifier_call = ip6_route_dev_notify,
242d3a49 6006 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
8ed67789
DL		 6007};
6008
2f460933
WC		 6009void __init ip6_route_init_special_entries(void)
6010{
6011 /* Registering of the loopback is done before this portion of code,
6012 * the loopback reference in rt6_info will not be taken, do it
6013 * manually for init_net */
1cf844c7 6014 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
2f460933
WC		 6015 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6016 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6017 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6018 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6019 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6020 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6021 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6022 #endif
6023}
6024
433d49c3 6025int __init ip6_route_init(void)
1da177e4 6026{
433d49c3 6027 int ret;
8d0b94af 6028 int cpu;
433d49c3 6029
9a7ec3a9
DL		 6030 ret = -ENOMEM;
6031 ip6_dst_ops_template.kmem_cachep =
e5d679f3 6032 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
f845ab6b 6033 SLAB_HWCACHE_ALIGN, NULL);
9a7ec3a9 6034 if (!ip6_dst_ops_template.kmem_cachep)
c19a28e1 6035 goto out;
14e50e57 6036
fc66f95c 6037 ret = dst_entries_init(&ip6_dst_blackhole_ops);
8ed67789 6038 if (ret)
bdb3289f 6039 goto out_kmem_cache;
bdb3289f 6040
c3426b47
DM		 6041 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6042 if (ret)
e8803b6c 6043 goto out_dst_entries;
2a0c451a 6044
7e52b33b
DM		 6045 ret = register_pernet_subsys(&ip6_route_net_ops);
6046 if (ret)
6047 goto out_register_inetpeer;
c3426b47 6048
5dc121e9
AE		 6049 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6050
e8803b6c 6051 ret = fib6_init();
433d49c3 6052 if (ret)
8ed67789 6053 goto out_register_subsys;
433d49c3 6054
433d49c3
DL		 6055 ret = xfrm6_init();
6056 if (ret)
e8803b6c 6057 goto out_fib6_init;
c35b7e72 6058
433d49c3
DL		 6059 ret = fib6_rules_init();
6060 if (ret)
6061 goto xfrm6_init;
7e5449c2 6062
d189634e
TG		 6063 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6064 if (ret)
6065 goto fib6_rules_init;
6066
16feebcf
FW		 6067 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6068 inet6_rtm_newroute, NULL, 0);
6069 if (ret < 0)
6070 goto out_register_late_subsys;
6071
6072 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6073 inet6_rtm_delroute, NULL, 0);
6074 if (ret < 0)
6075 goto out_register_late_subsys;
6076
6077 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6078 inet6_rtm_getroute, NULL,
6079 RTNL_FLAG_DOIT_UNLOCKED);
6080 if (ret < 0)
d189634e 6081 goto out_register_late_subsys;
c127ea2c 6082
8ed67789 6083 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
cdb18761 6084 if (ret)
d189634e 6085 goto out_register_late_subsys;
8ed67789 6086
8d0b94af
MKL		 6087 for_each_possible_cpu(cpu) {
6088 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6089
6090 INIT_LIST_HEAD(&ul->head);
6091 spin_lock_init(&ul->lock);
6092 }
6093
433d49c3
DL		 6094out:
6095 return ret;
6096
d189634e 6097out_register_late_subsys:
16feebcf 6098 rtnl_unregister_all(PF_INET6);
d189634e 6099 unregister_pernet_subsys(&ip6_route_net_late_ops);
433d49c3 6100fib6_rules_init:
433d49c3
DL		 6101 fib6_rules_cleanup();
6102xfrm6_init:
433d49c3 6103 xfrm6_fini();
2a0c451a
TG		 6104out_fib6_init:
6105 fib6_gc_cleanup();
8ed67789
DL		 6106out_register_subsys:
6107 unregister_pernet_subsys(&ip6_route_net_ops);
7e52b33b
DM		 6108out_register_inetpeer:
6109 unregister_pernet_subsys(&ipv6_inetpeer_ops);
fc66f95c
ED		 6110out_dst_entries:
6111 dst_entries_destroy(&ip6_dst_blackhole_ops);
433d49c3 6112out_kmem_cache:
f2fc6a54 6113 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
433d49c3 6114 goto out;
1da177e4
LT		 6115}
6116
6117void ip6_route_cleanup(void)
6118{
8ed67789 6119 unregister_netdevice_notifier(&ip6_route_dev_notifier);
d189634e 6120 unregister_pernet_subsys(&ip6_route_net_late_ops);
101367c2 6121 fib6_rules_cleanup();
1da177e4 6122 xfrm6_fini();
1da177e4 6123 fib6_gc_cleanup();
c3426b47 6124 unregister_pernet_subsys(&ipv6_inetpeer_ops);
8ed67789 6125 unregister_pernet_subsys(&ip6_route_net_ops);
41bb78b4 6126 dst_entries_destroy(&ip6_dst_blackhole_ops);
f2fc6a54 6127 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
1da177e4 6128}
Linux 6.x block layer and io_uring tree(s)