projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'mac80211-for-davem-2017-05-08' of git://git.kernel.org/pub/scm/linux/kerne...
[linux-block.git] / net / ipv6 / route.c
CommitLineData
1da177e4
LT		 1/*
2 * Linux INET6 implementation
3 * FIB front-end.
4 *
5 * Authors:
1ab1457c 6 * Pedro Roque <roque@di.fc.ul.pt>
1da177e4 7 *
1da177e4
LT		 8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14/* Changes:
15 *
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
c0bece9f
YH		 23 * Ville Nuorvala
24 * Fixed routing subtrees.
1da177e4
LT		 25 */
26
f3213831
JP		 27#define pr_fmt(fmt) "IPv6: " fmt
28
4fc268d2 29#include <linux/capability.h>
1da177e4 30#include <linux/errno.h>
bc3b2d7f 31#include <linux/export.h>
1da177e4
LT		 32#include <linux/types.h>
33#include <linux/times.h>
34#include <linux/socket.h>
35#include <linux/sockios.h>
36#include <linux/net.h>
37#include <linux/route.h>
38#include <linux/netdevice.h>
39#include <linux/in6.h>
7bc570c8 40#include <linux/mroute6.h>
1da177e4 41#include <linux/init.h>
1da177e4 42#include <linux/if_arp.h>
1da177e4
LT		 43#include <linux/proc_fs.h>
44#include <linux/seq_file.h>
5b7c931d 45#include <linux/nsproxy.h>
5a0e3ad6 46#include <linux/slab.h>
457c4cbc 47#include <net/net_namespace.h>
1da177e4
LT		 48#include <net/snmp.h>
49#include <net/ipv6.h>
50#include <net/ip6_fib.h>
51#include <net/ip6_route.h>
52#include <net/ndisc.h>
53#include <net/addrconf.h>
54#include <net/tcp.h>
55#include <linux/rtnetlink.h>
56#include <net/dst.h>
904af04d 57#include <net/dst_metadata.h>
1da177e4 58#include <net/xfrm.h>
8d71740c 59#include <net/netevent.h>
21713ebc 60#include <net/netlink.h>
51ebd318 61#include <net/nexthop.h>
19e42e45 62#include <net/lwtunnel.h>
904af04d 63#include <net/ip_tunnels.h>
ca254490 64#include <net/l3mdev.h>
b811580d 65#include <trace/events/fib6.h>
1da177e4 66
7c0f6ba6 67#include <linux/uaccess.h>
1da177e4
LT		 68
69#ifdef CONFIG_SYSCTL
70#include <linux/sysctl.h>
71#endif
72
afc154e9 73enum rt6_nud_state {
7e980569
JB		 74 RT6_NUD_FAIL_HARD = -3,
75 RT6_NUD_FAIL_PROBE = -2,
76 RT6_NUD_FAIL_DO_RR = -1,
afc154e9
HFS		 77 RT6_NUD_SUCCEED = 1
78};
79
83a09abd 80static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort);
1da177e4 81static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
0dbaee3b 82static unsigned int ip6_default_advmss(const struct dst_entry *dst);
ebb762f2 83static unsigned int ip6_mtu(const struct dst_entry *dst);
1da177e4
LT		 84static struct dst_entry *ip6_negative_advice(struct dst_entry *);
85static void ip6_dst_destroy(struct dst_entry *);
86static void ip6_dst_ifdown(struct dst_entry *,
87 struct net_device *dev, int how);
569d3645 88static int ip6_dst_gc(struct dst_ops *ops);
1da177e4
LT		 89
90static int ip6_pkt_discard(struct sk_buff *skb);
ede2059d 91static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
7150aede 92static int ip6_pkt_prohibit(struct sk_buff *skb);
ede2059d 93static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1da177e4 94static void ip6_link_failure(struct sk_buff *skb);
6700c270
DM		 95static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
96 struct sk_buff *skb, u32 mtu);
97static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
98 struct sk_buff *skb);
4b32b5ad 99static void rt6_dst_from_metrics_check(struct rt6_info *rt);
52bd4c0c 100static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
16a16cd3
DA		 101static size_t rt6_nlmsg_size(struct rt6_info *rt);
102static int rt6_fill_node(struct net *net,
103 struct sk_buff *skb, struct rt6_info *rt,
104 struct in6_addr *dst, struct in6_addr *src,
105 int iif, int type, u32 portid, u32 seq,
106 unsigned int flags);
1da177e4 107
70ceb4f5 108#ifdef CONFIG_IPV6_ROUTE_INFO
efa2cea0 109static struct rt6_info *rt6_add_route_info(struct net *net,
b71d1d42 110 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 111 const struct in6_addr *gwaddr,
112 struct net_device *dev,
95c96174 113 unsigned int pref);
efa2cea0 114static struct rt6_info *rt6_get_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);
70ceb4f5
YH		 118#endif
119
8d0b94af
MKL		 120struct uncached_list {
121 spinlock_t lock;
122 struct list_head head;
123};
124
125static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
126
127static void rt6_uncached_list_add(struct rt6_info *rt)
128{
129 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
130
131 rt->dst.flags |= DST_NOCACHE;
132 rt->rt6i_uncached_list = ul;
133
134 spin_lock_bh(&ul->lock);
135 list_add_tail(&rt->rt6i_uncached, &ul->head);
136 spin_unlock_bh(&ul->lock);
137}
138
139static void rt6_uncached_list_del(struct rt6_info *rt)
140{
141 if (!list_empty(&rt->rt6i_uncached)) {
142 struct uncached_list *ul = rt->rt6i_uncached_list;
143
144 spin_lock_bh(&ul->lock);
145 list_del(&rt->rt6i_uncached);
146 spin_unlock_bh(&ul->lock);
147 }
148}
149
150static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
151{
152 struct net_device *loopback_dev = net->loopback_dev;
153 int cpu;
154
e332bc67
EB		 155 if (dev == loopback_dev)
156 return;
157
8d0b94af
MKL		 158 for_each_possible_cpu(cpu) {
159 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
160 struct rt6_info *rt;
161
162 spin_lock_bh(&ul->lock);
163 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
164 struct inet6_dev *rt_idev = rt->rt6i_idev;
165 struct net_device *rt_dev = rt->dst.dev;
166
e332bc67 167 if (rt_idev->dev == dev) {
8d0b94af
MKL		 168 rt->rt6i_idev = in6_dev_get(loopback_dev);
169 in6_dev_put(rt_idev);
170 }
171
e332bc67 172 if (rt_dev == dev) {
8d0b94af
MKL		 173 rt->dst.dev = loopback_dev;
174 dev_hold(rt->dst.dev);
175 dev_put(rt_dev);
176 }
177 }
178 spin_unlock_bh(&ul->lock);
179 }
180}
181
d52d3997
MKL		 182static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
183{
184 return dst_metrics_write_ptr(rt->dst.from);
185}
186
06582540
DM		 187static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
188{
4b32b5ad 189 struct rt6_info *rt = (struct rt6_info *)dst;
06582540 190
d52d3997
MKL		 191 if (rt->rt6i_flags & RTF_PCPU)
192 return rt6_pcpu_cow_metrics(rt);
193 else if (rt->rt6i_flags & RTF_CACHE)
4b32b5ad
MKL		 194 return NULL;
195 else
3b471175 196 return dst_cow_metrics_generic(dst, old);
06582540
DM		 197}
198
f894cbf8
DM		 199static inline const void *choose_neigh_daddr(struct rt6_info *rt,
200 struct sk_buff *skb,
201 const void *daddr)
39232973
DM		 202{
203 struct in6_addr *p = &rt->rt6i_gateway;
204
a7563f34 205 if (!ipv6_addr_any(p))
39232973 206 return (const void *) p;
f894cbf8
DM		 207 else if (skb)
208 return &ipv6_hdr(skb)->daddr;
39232973
DM		 209 return daddr;
210}
211
f894cbf8
DM		 212static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
213 struct sk_buff *skb,
214 const void *daddr)
d3aaeb38 215{
39232973
DM		 216 struct rt6_info *rt = (struct rt6_info *) dst;
217 struct neighbour *n;
218
f894cbf8 219 daddr = choose_neigh_daddr(rt, skb, daddr);
8e022ee6 220 n = __ipv6_neigh_lookup(dst->dev, daddr);
f83c7790
DM		 221 if (n)
222 return n;
223 return neigh_create(&nd_tbl, daddr, dst->dev);
224}
225
63fca65d
JA		 226static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
227{
228 struct net_device *dev = dst->dev;
229 struct rt6_info *rt = (struct rt6_info *)dst;
230
231 daddr = choose_neigh_daddr(rt, NULL, daddr);
232 if (!daddr)
233 return;
234 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
235 return;
236 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
237 return;
238 __ipv6_confirm_neigh(dev, daddr);
239}
240
9a7ec3a9 241static struct dst_ops ip6_dst_ops_template = {
1da177e4 242 .family = AF_INET6,
1da177e4
LT		 243 .gc = ip6_dst_gc,
244 .gc_thresh = 1024,
245 .check = ip6_dst_check,
0dbaee3b 246 .default_advmss = ip6_default_advmss,
ebb762f2 247 .mtu = ip6_mtu,
06582540 248 .cow_metrics = ipv6_cow_metrics,
1da177e4
LT		 249 .destroy = ip6_dst_destroy,
250 .ifdown = ip6_dst_ifdown,
251 .negative_advice = ip6_negative_advice,
252 .link_failure = ip6_link_failure,
253 .update_pmtu = ip6_rt_update_pmtu,
6e157b6a 254 .redirect = rt6_do_redirect,
9f8955cc 255 .local_out = __ip6_local_out,
d3aaeb38 256 .neigh_lookup = ip6_neigh_lookup,
63fca65d 257 .confirm_neigh = ip6_confirm_neigh,
1da177e4
LT		 258};
259
ebb762f2 260static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
ec831ea7 261{
618f9bc7
SK		 262 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
263
264 return mtu ? : dst->dev->mtu;
ec831ea7
RD		 265}
266
6700c270
DM		 267static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
268 struct sk_buff *skb, u32 mtu)
14e50e57
DM		 269{
270}
271
6700c270
DM		 272static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
273 struct sk_buff *skb)
b587ee3b
DM		 274{
275}
276
14e50e57
DM		 277static struct dst_ops ip6_dst_blackhole_ops = {
278 .family = AF_INET6,
14e50e57
DM		 279 .destroy = ip6_dst_destroy,
280 .check = ip6_dst_check,
ebb762f2 281 .mtu = ip6_blackhole_mtu,
214f45c9 282 .default_advmss = ip6_default_advmss,
14e50e57 283 .update_pmtu = ip6_rt_blackhole_update_pmtu,
b587ee3b 284 .redirect = ip6_rt_blackhole_redirect,
0a1f5962 285 .cow_metrics = dst_cow_metrics_generic,
d3aaeb38 286 .neigh_lookup = ip6_neigh_lookup,
14e50e57
DM		 287};
288
62fa8a84 289static const u32 ip6_template_metrics[RTAX_MAX] = {
14edd87d 290 [RTAX_HOPLIMIT - 1] = 0,
62fa8a84
DM		 291};
292
fb0af4c7 293static const struct rt6_info ip6_null_entry_template = {
d8d1f30b
CG		 294 .dst = {
295 .__refcnt = ATOMIC_INIT(1),
296 .__use = 1,
2c20cbd7 297 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 298 .error = -ENETUNREACH,
d8d1f30b
CG		 299 .input = ip6_pkt_discard,
300 .output = ip6_pkt_discard_out,
1da177e4
LT		 301 },
302 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
4f724279 303 .rt6i_protocol = RTPROT_KERNEL,
1da177e4
LT		 304 .rt6i_metric = ~(u32) 0,
305 .rt6i_ref = ATOMIC_INIT(1),
306};
307
101367c2
TG		 308#ifdef CONFIG_IPV6_MULTIPLE_TABLES
309
fb0af4c7 310static const struct rt6_info ip6_prohibit_entry_template = {
d8d1f30b
CG		 311 .dst = {
312 .__refcnt = ATOMIC_INIT(1),
313 .__use = 1,
2c20cbd7 314 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 315 .error = -EACCES,
d8d1f30b
CG		 316 .input = ip6_pkt_prohibit,
317 .output = ip6_pkt_prohibit_out,
101367c2
TG		 318 },
319 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
4f724279 320 .rt6i_protocol = RTPROT_KERNEL,
101367c2
TG		 321 .rt6i_metric = ~(u32) 0,
322 .rt6i_ref = ATOMIC_INIT(1),
323};
324
fb0af4c7 325static const struct rt6_info ip6_blk_hole_entry_template = {
d8d1f30b
CG		 326 .dst = {
327 .__refcnt = ATOMIC_INIT(1),
328 .__use = 1,
2c20cbd7 329 .obsolete = DST_OBSOLETE_FORCE_CHK,
d8d1f30b 330 .error = -EINVAL,
d8d1f30b 331 .input = dst_discard,
ede2059d 332 .output = dst_discard_out,
101367c2
TG		 333 },
334 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
4f724279 335 .rt6i_protocol = RTPROT_KERNEL,
101367c2
TG		 336 .rt6i_metric = ~(u32) 0,
337 .rt6i_ref = ATOMIC_INIT(1),
338};
339
340#endif
341
ebfa45f0
MKL		 342static void rt6_info_init(struct rt6_info *rt)
343{
344 struct dst_entry *dst = &rt->dst;
345
346 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
347 INIT_LIST_HEAD(&rt->rt6i_siblings);
348 INIT_LIST_HEAD(&rt->rt6i_uncached);
349}
350
1da177e4 351/* allocate dst with ip6_dst_ops */
d52d3997
MKL		 352static struct rt6_info *__ip6_dst_alloc(struct net *net,
353 struct net_device *dev,
ad706862 354 int flags)
1da177e4 355{
97bab73f 356 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
6f3118b5 357 0, DST_OBSOLETE_FORCE_CHK, flags);
cf911662 358
ebfa45f0
MKL		 359 if (rt)
360 rt6_info_init(rt);
8104891b 361
cf911662 362 return rt;
1da177e4
LT		 363}
364
9ab179d8
DA		 365struct rt6_info *ip6_dst_alloc(struct net *net,
366 struct net_device *dev,
367 int flags)
d52d3997 368{
ad706862 369 struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
d52d3997
MKL		 370
371 if (rt) {
372 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
373 if (rt->rt6i_pcpu) {
374 int cpu;
375
376 for_each_possible_cpu(cpu) {
377 struct rt6_info **p;
378
379 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
380 /* no one shares rt */
381 *p = NULL;
382 }
383 } else {
384 dst_destroy((struct dst_entry *)rt);
385 return NULL;
386 }
387 }
388
389 return rt;
390}
9ab179d8 391EXPORT_SYMBOL(ip6_dst_alloc);
d52d3997 392
1da177e4
LT		 393static void ip6_dst_destroy(struct dst_entry *dst)
394{
395 struct rt6_info *rt = (struct rt6_info *)dst;
ecd98837 396 struct dst_entry *from = dst->from;
8d0b94af 397 struct inet6_dev *idev;
1da177e4 398
4b32b5ad 399 dst_destroy_metrics_generic(dst);
87775312 400 free_percpu(rt->rt6i_pcpu);
8d0b94af
MKL		 401 rt6_uncached_list_del(rt);
402
403 idev = rt->rt6i_idev;
38308473 404 if (idev) {
1da177e4
LT		 405 rt->rt6i_idev = NULL;
406 in6_dev_put(idev);
1ab1457c 407 }
1716a961 408
ecd98837
YH		 409 dst->from = NULL;
410 dst_release(from);
b3419363
DM		 411}
412
1da177e4
LT		 413static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
414 int how)
415{
416 struct rt6_info *rt = (struct rt6_info *)dst;
417 struct inet6_dev *idev = rt->rt6i_idev;
5a3e55d6 418 struct net_device *loopback_dev =
c346dca1 419 dev_net(dev)->loopback_dev;
1da177e4 420
97cac082
DM		 421 if (dev != loopback_dev) {
422 if (idev && idev->dev == dev) {
423 struct inet6_dev *loopback_idev =
424 in6_dev_get(loopback_dev);
425 if (loopback_idev) {
426 rt->rt6i_idev = loopback_idev;
427 in6_dev_put(idev);
428 }
429 }
1da177e4
LT		 430 }
431}
432
5973fb1e
MKL		 433static bool __rt6_check_expired(const struct rt6_info *rt)
434{
435 if (rt->rt6i_flags & RTF_EXPIRES)
436 return time_after(jiffies, rt->dst.expires);
437 else
438 return false;
439}
440
a50feda5 441static bool rt6_check_expired(const struct rt6_info *rt)
1da177e4 442{
1716a961
G		 443 if (rt->rt6i_flags & RTF_EXPIRES) {
444 if (time_after(jiffies, rt->dst.expires))
a50feda5 445 return true;
1716a961 446 } else if (rt->dst.from) {
3fd91fb3 447 return rt6_check_expired((struct rt6_info *) rt->dst.from);
1716a961 448 }
a50feda5 449 return false;
1da177e4
LT		 450}
451
51ebd318
ND		 452/* Multipath route selection:
453 * Hash based function using packet header and flowlabel.
454 * Adapted from fib_info_hashfn()
455 */
456static int rt6_info_hash_nhsfn(unsigned int candidate_count,
457 const struct flowi6 *fl6)
458{
644d0e65 459 return get_hash_from_flowi6(fl6) % candidate_count;
51ebd318
ND		 460}
461
462static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
52bd4c0c
ND		 463 struct flowi6 *fl6, int oif,
464 int strict)
51ebd318
ND		 465{
466 struct rt6_info *sibling, *next_sibling;
467 int route_choosen;
468
469 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
470 /* Don't change the route, if route_choosen == 0
471 * (siblings does not include ourself)
472 */
473 if (route_choosen)
474 list_for_each_entry_safe(sibling, next_sibling,
475 &match->rt6i_siblings, rt6i_siblings) {
476 route_choosen--;
477 if (route_choosen == 0) {
52bd4c0c
ND		 478 if (rt6_score_route(sibling, oif, strict) < 0)
479 break;
51ebd318
ND		 480 match = sibling;
481 break;
482 }
483 }
484 return match;
485}
486
1da177e4 487/*
c71099ac 488 * Route lookup. Any table->tb6_lock is implied.
1da177e4
LT		 489 */
490
8ed67789
DL		 491static inline struct rt6_info *rt6_device_match(struct net *net,
492 struct rt6_info *rt,
b71d1d42 493 const struct in6_addr *saddr,
1da177e4 494 int oif,
d420895e 495 int flags)
1da177e4
LT		 496{
497 struct rt6_info *local = NULL;
498 struct rt6_info *sprt;
499
dd3abc4e
YH		 500 if (!oif && ipv6_addr_any(saddr))
501 goto out;
502
d8d1f30b 503 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
d1918542 504 struct net_device *dev = sprt->dst.dev;
dd3abc4e
YH		 505
506 if (oif) {
1da177e4
LT		 507 if (dev->ifindex == oif)
508 return sprt;
509 if (dev->flags & IFF_LOOPBACK) {
38308473 510 if (!sprt->rt6i_idev ||
1da177e4 511 sprt->rt6i_idev->dev->ifindex != oif) {
17fb0b2b 512 if (flags & RT6_LOOKUP_F_IFACE)
1da177e4 513 continue;
17fb0b2b
DA		 514 if (local &&
515 local->rt6i_idev->dev->ifindex == oif)
1da177e4
LT		 516 continue;
517 }
518 local = sprt;
519 }
dd3abc4e
YH		 520 } else {
521 if (ipv6_chk_addr(net, saddr, dev,
522 flags & RT6_LOOKUP_F_IFACE))
523 return sprt;
1da177e4 524 }
dd3abc4e 525 }
1da177e4 526
dd3abc4e 527 if (oif) {
1da177e4
LT		 528 if (local)
529 return local;
530
d420895e 531 if (flags & RT6_LOOKUP_F_IFACE)
8ed67789 532 return net->ipv6.ip6_null_entry;
1da177e4 533 }
dd3abc4e 534out:
1da177e4
LT		 535 return rt;
536}
537
27097255 538#ifdef CONFIG_IPV6_ROUTER_PREF
c2f17e82
HFS		 539struct __rt6_probe_work {
540 struct work_struct work;
541 struct in6_addr target;
542 struct net_device *dev;
543};
544
545static void rt6_probe_deferred(struct work_struct *w)
546{
547 struct in6_addr mcaddr;
548 struct __rt6_probe_work *work =
549 container_of(w, struct __rt6_probe_work, work);
550
551 addrconf_addr_solict_mult(&work->target, &mcaddr);
adc176c5 552 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
c2f17e82 553 dev_put(work->dev);
662f5533 554 kfree(work);
c2f17e82
HFS		 555}
556
27097255
YH		 557static void rt6_probe(struct rt6_info *rt)
558{
990edb42 559 struct __rt6_probe_work *work;
f2c31e32 560 struct neighbour *neigh;
27097255
YH		 561 /*
562 * Okay, this does not seem to be appropriate
563 * for now, however, we need to check if it
564 * is really so; aka Router Reachability Probing.
565 *
566 * Router Reachability Probe MUST be rate-limited
567 * to no more than one per minute.
568 */
2152caea 569 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
7ff74a59 570 return;
2152caea
YH		 571 rcu_read_lock_bh();
572 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
573 if (neigh) {
8d6c31bf
MKL		 574 if (neigh->nud_state & NUD_VALID)
575 goto out;
576
990edb42 577 work = NULL;
2152caea 578 write_lock(&neigh->lock);
990edb42
MKL		 579 if (!(neigh->nud_state & NUD_VALID) &&
580 time_after(jiffies,
581 neigh->updated +
582 rt->rt6i_idev->cnf.rtr_probe_interval)) {
583 work = kmalloc(sizeof(*work), GFP_ATOMIC);
584 if (work)
585 __neigh_set_probe_once(neigh);
c2f17e82 586 }
2152caea 587 write_unlock(&neigh->lock);
990edb42
MKL		 588 } else {
589 work = kmalloc(sizeof(*work), GFP_ATOMIC);
f2c31e32 590 }
990edb42
MKL		 591
592 if (work) {
593 INIT_WORK(&work->work, rt6_probe_deferred);
594 work->target = rt->rt6i_gateway;
595 dev_hold(rt->dst.dev);
596 work->dev = rt->dst.dev;
597 schedule_work(&work->work);
598 }
599
8d6c31bf 600out:
2152caea 601 rcu_read_unlock_bh();
27097255
YH		 602}
603#else
604static inline void rt6_probe(struct rt6_info *rt)
605{
27097255
YH		 606}
607#endif
608
1da177e4 609/*
554cfb7e 610 * Default Router Selection (RFC 2461 6.3.6)
1da177e4 611 */
b6f99a21 612static inline int rt6_check_dev(struct rt6_info *rt, int oif)
554cfb7e 613{
d1918542 614 struct net_device *dev = rt->dst.dev;
161980f4 615 if (!oif || dev->ifindex == oif)
554cfb7e 616 return 2;
161980f4
DM		 617 if ((dev->flags & IFF_LOOPBACK) &&
618 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
619 return 1;
620 return 0;
554cfb7e 621}
1da177e4 622
afc154e9 623static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
1da177e4 624{
f2c31e32 625 struct neighbour *neigh;
afc154e9 626 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
f2c31e32 627
4d0c5911
YH		 628 if (rt->rt6i_flags & RTF_NONEXTHOP ||
629 !(rt->rt6i_flags & RTF_GATEWAY))
afc154e9 630 return RT6_NUD_SUCCEED;
145a3621
YH		 631
632 rcu_read_lock_bh();
633 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
634 if (neigh) {
635 read_lock(&neigh->lock);
554cfb7e 636 if (neigh->nud_state & NUD_VALID)
afc154e9 637 ret = RT6_NUD_SUCCEED;
398bcbeb 638#ifdef CONFIG_IPV6_ROUTER_PREF
a5a81f0b 639 else if (!(neigh->nud_state & NUD_FAILED))
afc154e9 640 ret = RT6_NUD_SUCCEED;
7e980569
JB		 641 else
642 ret = RT6_NUD_FAIL_PROBE;
398bcbeb 643#endif
145a3621 644 read_unlock(&neigh->lock);
afc154e9
HFS		 645 } else {
646 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
7e980569 647 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
a5a81f0b 648 }
145a3621
YH		 649 rcu_read_unlock_bh();
650
a5a81f0b 651 return ret;
1da177e4
LT		 652}
653
554cfb7e
YH		 654static int rt6_score_route(struct rt6_info *rt, int oif,
655 int strict)
1da177e4 656{
a5a81f0b 657 int m;
1ab1457c 658
4d0c5911 659 m = rt6_check_dev(rt, oif);
77d16f45 660 if (!m && (strict & RT6_LOOKUP_F_IFACE))
afc154e9 661 return RT6_NUD_FAIL_HARD;
ebacaaa0
YH		 662#ifdef CONFIG_IPV6_ROUTER_PREF
663 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
664#endif
afc154e9
HFS		 665 if (strict & RT6_LOOKUP_F_REACHABLE) {
666 int n = rt6_check_neigh(rt);
667 if (n < 0)
668 return n;
669 }
554cfb7e
YH		 670 return m;
671}
672
f11e6659 673static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
afc154e9
HFS		 674 int *mpri, struct rt6_info *match,
675 bool *do_rr)
554cfb7e 676{
f11e6659 677 int m;
afc154e9 678 bool match_do_rr = false;
35103d11
AG		 679 struct inet6_dev *idev = rt->rt6i_idev;
680 struct net_device *dev = rt->dst.dev;
681
682 if (dev && !netif_carrier_ok(dev) &&
d5d32e4b
DA		 683 idev->cnf.ignore_routes_with_linkdown &&
684 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
35103d11 685 goto out;
f11e6659
DM		 686
687 if (rt6_check_expired(rt))
688 goto out;
689
690 m = rt6_score_route(rt, oif, strict);
7e980569 691 if (m == RT6_NUD_FAIL_DO_RR) {
afc154e9
HFS		 692 match_do_rr = true;
693 m = 0; /* lowest valid score */
7e980569 694 } else if (m == RT6_NUD_FAIL_HARD) {
f11e6659 695 goto out;
afc154e9
HFS		 696 }
697
698 if (strict & RT6_LOOKUP_F_REACHABLE)
699 rt6_probe(rt);
f11e6659 700
7e980569 701 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
f11e6659 702 if (m > *mpri) {
afc154e9 703 *do_rr = match_do_rr;
f11e6659
DM		 704 *mpri = m;
705 match = rt;
f11e6659 706 }
f11e6659
DM		 707out:
708 return match;
709}
710
711static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
712 struct rt6_info *rr_head,
afc154e9
HFS		 713 u32 metric, int oif, int strict,
714 bool *do_rr)
f11e6659 715{
9fbdcfaf 716 struct rt6_info *rt, *match, *cont;
554cfb7e 717 int mpri = -1;
1da177e4 718
f11e6659 719 match = NULL;
9fbdcfaf
SK		 720 cont = NULL;
721 for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
722 if (rt->rt6i_metric != metric) {
723 cont = rt;
724 break;
725 }
726
727 match = find_match(rt, oif, strict, &mpri, match, do_rr);
728 }
729
730 for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
731 if (rt->rt6i_metric != metric) {
732 cont = rt;
733 break;
734 }
735
afc154e9 736 match = find_match(rt, oif, strict, &mpri, match, do_rr);
9fbdcfaf
SK		 737 }
738
739 if (match || !cont)
740 return match;
741
742 for (rt = cont; rt; rt = rt->dst.rt6_next)
afc154e9 743 match = find_match(rt, oif, strict, &mpri, match, do_rr);
1da177e4 744
f11e6659
DM		 745 return match;
746}
1da177e4 747
f11e6659
DM		 748static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
749{
750 struct rt6_info *match, *rt0;
8ed67789 751 struct net *net;
afc154e9 752 bool do_rr = false;
1da177e4 753
f11e6659
DM		 754 rt0 = fn->rr_ptr;
755 if (!rt0)
756 fn->rr_ptr = rt0 = fn->leaf;
1da177e4 757
afc154e9
HFS		 758 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
759 &do_rr);
1da177e4 760
afc154e9 761 if (do_rr) {
d8d1f30b 762 struct rt6_info *next = rt0->dst.rt6_next;
f11e6659 763
554cfb7e 764 /* no entries matched; do round-robin */
f11e6659
DM		 765 if (!next || next->rt6i_metric != rt0->rt6i_metric)
766 next = fn->leaf;
767
768 if (next != rt0)
769 fn->rr_ptr = next;
1da177e4 770 }
1da177e4 771
d1918542 772 net = dev_net(rt0->dst.dev);
a02cec21 773 return match ? match : net->ipv6.ip6_null_entry;
1da177e4
LT		 774}
775
8b9df265
MKL		 776static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
777{
778 return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
779}
780
70ceb4f5
YH		 781#ifdef CONFIG_IPV6_ROUTE_INFO
782int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
b71d1d42 783 const struct in6_addr *gwaddr)
70ceb4f5 784{
c346dca1 785 struct net *net = dev_net(dev);
70ceb4f5
YH		 786 struct route_info *rinfo = (struct route_info *) opt;
787 struct in6_addr prefix_buf, *prefix;
788 unsigned int pref;
4bed72e4 789 unsigned long lifetime;
70ceb4f5
YH		 790 struct rt6_info *rt;
791
792 if (len < sizeof(struct route_info)) {
793 return -EINVAL;
794 }
795
796 /* Sanity check for prefix_len and length */
797 if (rinfo->length > 3) {
798 return -EINVAL;
799 } else if (rinfo->prefix_len > 128) {
800 return -EINVAL;
801 } else if (rinfo->prefix_len > 64) {
802 if (rinfo->length < 2) {
803 return -EINVAL;
804 }
805 } else if (rinfo->prefix_len > 0) {
806 if (rinfo->length < 1) {
807 return -EINVAL;
808 }
809 }
810
811 pref = rinfo->route_pref;
812 if (pref == ICMPV6_ROUTER_PREF_INVALID)
3933fc95 813 return -EINVAL;
70ceb4f5 814
4bed72e4 815 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
70ceb4f5
YH		 816
817 if (rinfo->length == 3)
818 prefix = (struct in6_addr *)rinfo->prefix;
819 else {
820 /* this function is safe */
821 ipv6_addr_prefix(&prefix_buf,
822 (struct in6_addr *)rinfo->prefix,
823 rinfo->prefix_len);
824 prefix = &prefix_buf;
825 }
826
f104a567
DJ		 827 if (rinfo->prefix_len == 0)
828 rt = rt6_get_dflt_router(gwaddr, dev);
829 else
830 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
830218c1 831 gwaddr, dev);
70ceb4f5
YH		 832
833 if (rt && !lifetime) {
e0a1ad73 834 ip6_del_rt(rt);
70ceb4f5
YH		 835 rt = NULL;
836 }
837
838 if (!rt && lifetime)
830218c1
DA		 839 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
840 dev, pref);
70ceb4f5
YH		 841 else if (rt)
842 rt->rt6i_flags = RTF_ROUTEINFO |
843 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
844
845 if (rt) {
1716a961
G		 846 if (!addrconf_finite_timeout(lifetime))
847 rt6_clean_expires(rt);
848 else
849 rt6_set_expires(rt, jiffies + HZ * lifetime);
850
94e187c0 851 ip6_rt_put(rt);
70ceb4f5
YH		 852 }
853 return 0;
854}
855#endif
856
a3c00e46
MKL		 857static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
858 struct in6_addr *saddr)
859{
860 struct fib6_node *pn;
861 while (1) {
862 if (fn->fn_flags & RTN_TL_ROOT)
863 return NULL;
864 pn = fn->parent;
865 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
866 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
867 else
868 fn = pn;
869 if (fn->fn_flags & RTN_RTINFO)
870 return fn;
871 }
872}
c71099ac 873
8ed67789
DL		 874static struct rt6_info *ip6_pol_route_lookup(struct net *net,
875 struct fib6_table *table,
4c9483b2 876 struct flowi6 *fl6, int flags)
1da177e4
LT		 877{
878 struct fib6_node *fn;
879 struct rt6_info *rt;
880
c71099ac 881 read_lock_bh(&table->tb6_lock);
4c9483b2 882 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
c71099ac
TG		 883restart:
884 rt = fn->leaf;
4c9483b2 885 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
51ebd318 886 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
52bd4c0c 887 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
a3c00e46
MKL		 888 if (rt == net->ipv6.ip6_null_entry) {
889 fn = fib6_backtrack(fn, &fl6->saddr);
890 if (fn)
891 goto restart;
892 }
d8d1f30b 893 dst_use(&rt->dst, jiffies);
c71099ac 894 read_unlock_bh(&table->tb6_lock);
b811580d
DA		 895
896 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
897
c71099ac
TG		 898 return rt;
899
900}
901
67ba4152 902struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
ea6e574e
FW		 903 int flags)
904{
905 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
906}
907EXPORT_SYMBOL_GPL(ip6_route_lookup);
908
9acd9f3a
YH		 909struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
910 const struct in6_addr *saddr, int oif, int strict)
c71099ac 911{
4c9483b2
DM		 912 struct flowi6 fl6 = {
913 .flowi6_oif = oif,
914 .daddr = *daddr,
c71099ac
TG		 915 };
916 struct dst_entry *dst;
77d16f45 917 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
c71099ac 918
adaa70bb 919 if (saddr) {
4c9483b2 920 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
adaa70bb
TG		 921 flags |= RT6_LOOKUP_F_HAS_SADDR;
922 }
923
4c9483b2 924 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
c71099ac
TG		 925 if (dst->error == 0)
926 return (struct rt6_info *) dst;
927
928 dst_release(dst);
929
1da177e4
LT		 930 return NULL;
931}
7159039a
YH		 932EXPORT_SYMBOL(rt6_lookup);
933
c71099ac 934/* ip6_ins_rt is called with FREE table->tb6_lock.
1da177e4
LT		 935 It takes new route entry, the addition fails by any reason the
936 route is freed. In any case, if caller does not hold it, it may
937 be destroyed.
938 */
939
e5fd387a 940static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
e715b6d3 941 struct mx6_config *mxc)
1da177e4
LT		 942{
943 int err;
c71099ac 944 struct fib6_table *table;
1da177e4 945
c71099ac
TG		 946 table = rt->rt6i_table;
947 write_lock_bh(&table->tb6_lock);
e715b6d3 948 err = fib6_add(&table->tb6_root, rt, info, mxc);
c71099ac 949 write_unlock_bh(&table->tb6_lock);
1da177e4
LT		 950
951 return err;
952}
953
40e22e8f
TG		 954int ip6_ins_rt(struct rt6_info *rt)
955{
e715b6d3
FW		 956 struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
957 struct mx6_config mxc = { .mx = NULL, };
958
959 return __ip6_ins_rt(rt, &info, &mxc);
40e22e8f
TG		 960}
961
8b9df265
MKL		 962static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
963 const struct in6_addr *daddr,
964 const struct in6_addr *saddr)
1da177e4 965{
1da177e4
LT		 966 struct rt6_info *rt;
967
968 /*
969 * Clone the route.
970 */
971
d52d3997 972 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
83a09abd 973 ort = (struct rt6_info *)ort->dst.from;
1da177e4 974
ad706862 975 rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
83a09abd
MKL		 976
977 if (!rt)
978 return NULL;
979
980 ip6_rt_copy_init(rt, ort);
981 rt->rt6i_flags |= RTF_CACHE;
982 rt->rt6i_metric = 0;
983 rt->dst.flags |= DST_HOST;
984 rt->rt6i_dst.addr = *daddr;
985 rt->rt6i_dst.plen = 128;
1da177e4 986
83a09abd
MKL		 987 if (!rt6_is_gw_or_nonexthop(ort)) {
988 if (ort->rt6i_dst.plen != 128 &&
989 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
990 rt->rt6i_flags |= RTF_ANYCAST;
1da177e4 991#ifdef CONFIG_IPV6_SUBTREES
83a09abd
MKL		 992 if (rt->rt6i_src.plen && saddr) {
993 rt->rt6i_src.addr = *saddr;
994 rt->rt6i_src.plen = 128;
8b9df265 995 }
83a09abd 996#endif
95a9a5ba 997 }
1da177e4 998
95a9a5ba
YH		 999 return rt;
1000}
1da177e4 1001
d52d3997
MKL		 1002static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
1003{
1004 struct rt6_info *pcpu_rt;
1005
1006 pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
ad706862 1007 rt->dst.dev, rt->dst.flags);
d52d3997
MKL		 1008
1009 if (!pcpu_rt)
1010 return NULL;
1011 ip6_rt_copy_init(pcpu_rt, rt);
1012 pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
1013 pcpu_rt->rt6i_flags |= RTF_PCPU;
1014 return pcpu_rt;
1015}
1016
1017/* It should be called with read_lock_bh(&tb6_lock) acquired */
1018static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
1019{
a73e4195 1020 struct rt6_info *pcpu_rt, **p;
d52d3997
MKL		 1021
1022 p = this_cpu_ptr(rt->rt6i_pcpu);
1023 pcpu_rt = *p;
1024
a73e4195
MKL		 1025 if (pcpu_rt) {
1026 dst_hold(&pcpu_rt->dst);
1027 rt6_dst_from_metrics_check(pcpu_rt);
1028 }
1029 return pcpu_rt;
1030}
1031
1032static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1033{
9c7370a1 1034 struct fib6_table *table = rt->rt6i_table;
a73e4195 1035 struct rt6_info *pcpu_rt, *prev, **p;
d52d3997
MKL		 1036
1037 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1038 if (!pcpu_rt) {
1039 struct net *net = dev_net(rt->dst.dev);
1040
9c7370a1
MKL		 1041 dst_hold(&net->ipv6.ip6_null_entry->dst);
1042 return net->ipv6.ip6_null_entry;
d52d3997
MKL		 1043 }
1044
9c7370a1
MKL		 1045 read_lock_bh(&table->tb6_lock);
1046 if (rt->rt6i_pcpu) {
1047 p = this_cpu_ptr(rt->rt6i_pcpu);
1048 prev = cmpxchg(p, NULL, pcpu_rt);
1049 if (prev) {
1050 /* If someone did it before us, return prev instead */
1051 dst_destroy(&pcpu_rt->dst);
1052 pcpu_rt = prev;
1053 }
1054 } else {
1055 /* rt has been removed from the fib6 tree
1056 * before we have a chance to acquire the read_lock.
1057 * In this case, don't brother to create a pcpu rt
1058 * since rt is going away anyway. The next
1059 * dst_check() will trigger a re-lookup.
1060 */
d52d3997 1061 dst_destroy(&pcpu_rt->dst);
9c7370a1 1062 pcpu_rt = rt;
d52d3997 1063 }
d52d3997
MKL		 1064 dst_hold(&pcpu_rt->dst);
1065 rt6_dst_from_metrics_check(pcpu_rt);
9c7370a1 1066 read_unlock_bh(&table->tb6_lock);
d52d3997
MKL		 1067 return pcpu_rt;
1068}
1069
9ff74384
DA		 1070struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
1071 int oif, struct flowi6 *fl6, int flags)
1da177e4 1072{
367efcb9 1073 struct fib6_node *fn, *saved_fn;
45e4fd26 1074 struct rt6_info *rt;
c71099ac 1075 int strict = 0;
1da177e4 1076
77d16f45 1077 strict |= flags & RT6_LOOKUP_F_IFACE;
d5d32e4b 1078 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
367efcb9
MKL		 1079 if (net->ipv6.devconf_all->forwarding == 0)
1080 strict |= RT6_LOOKUP_F_REACHABLE;
1da177e4 1081
c71099ac 1082 read_lock_bh(&table->tb6_lock);
1da177e4 1083
4c9483b2 1084 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
367efcb9 1085 saved_fn = fn;
1da177e4 1086
ca254490
DA		 1087 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1088 oif = 0;
1089
a3c00e46 1090redo_rt6_select:
367efcb9 1091 rt = rt6_select(fn, oif, strict);
52bd4c0c 1092 if (rt->rt6i_nsiblings)
367efcb9 1093 rt = rt6_multipath_select(rt, fl6, oif, strict);
a3c00e46
MKL		 1094 if (rt == net->ipv6.ip6_null_entry) {
1095 fn = fib6_backtrack(fn, &fl6->saddr);
1096 if (fn)
1097 goto redo_rt6_select;
367efcb9
MKL		 1098 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1099 /* also consider unreachable route */
1100 strict &= ~RT6_LOOKUP_F_REACHABLE;
1101 fn = saved_fn;
1102 goto redo_rt6_select;
367efcb9 1103 }
a3c00e46
MKL		 1104 }
1105
fb9de91e 1106
3da59bd9 1107 if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
d52d3997
MKL		 1108 dst_use(&rt->dst, jiffies);
1109 read_unlock_bh(&table->tb6_lock);
1110
1111 rt6_dst_from_metrics_check(rt);
b811580d
DA		 1112
1113 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
d52d3997 1114 return rt;
3da59bd9
MKL		 1115 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1116 !(rt->rt6i_flags & RTF_GATEWAY))) {
1117 /* Create a RTF_CACHE clone which will not be
1118 * owned by the fib6 tree. It is for the special case where
1119 * the daddr in the skb during the neighbor look-up is different
1120 * from the fl6->daddr used to look-up route here.
1121 */
1122
1123 struct rt6_info *uncached_rt;
1124
d52d3997
MKL		 1125 dst_use(&rt->dst, jiffies);
1126 read_unlock_bh(&table->tb6_lock);
1127
3da59bd9
MKL		 1128 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1129 dst_release(&rt->dst);
c71099ac 1130
3da59bd9 1131 if (uncached_rt)
8d0b94af 1132 rt6_uncached_list_add(uncached_rt);
3da59bd9
MKL		 1133 else
1134 uncached_rt = net->ipv6.ip6_null_entry;
d52d3997 1135
3da59bd9 1136 dst_hold(&uncached_rt->dst);
b811580d
DA		 1137
1138 trace_fib6_table_lookup(net, uncached_rt, table->tb6_id, fl6);
3da59bd9 1139 return uncached_rt;
3da59bd9 1140
d52d3997
MKL		 1141 } else {
1142 /* Get a percpu copy */
1143
1144 struct rt6_info *pcpu_rt;
1145
1146 rt->dst.lastuse = jiffies;
1147 rt->dst.__use++;
1148 pcpu_rt = rt6_get_pcpu_route(rt);
d52d3997 1149
9c7370a1
MKL		 1150 if (pcpu_rt) {
1151 read_unlock_bh(&table->tb6_lock);
1152 } else {
1153 /* We have to do the read_unlock first
1154 * because rt6_make_pcpu_route() may trigger
1155 * ip6_dst_gc() which will take the write_lock.
1156 */
1157 dst_hold(&rt->dst);
1158 read_unlock_bh(&table->tb6_lock);
a73e4195 1159 pcpu_rt = rt6_make_pcpu_route(rt);
9c7370a1
MKL		 1160 dst_release(&rt->dst);
1161 }
d52d3997 1162
b811580d 1163 trace_fib6_table_lookup(net, pcpu_rt, table->tb6_id, fl6);
d52d3997 1164 return pcpu_rt;
9c7370a1 1165
d52d3997 1166 }
1da177e4 1167}
9ff74384 1168EXPORT_SYMBOL_GPL(ip6_pol_route);
1da177e4 1169
8ed67789 1170static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
4c9483b2 1171 struct flowi6 *fl6, int flags)
4acad72d 1172{
4c9483b2 1173 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
4acad72d
PE		 1174}
1175
d409b847
MB		 1176struct dst_entry *ip6_route_input_lookup(struct net *net,
1177 struct net_device *dev,
1178 struct flowi6 *fl6, int flags)
72331bc0
SL		 1179{
1180 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1181 flags |= RT6_LOOKUP_F_IFACE;
1182
1183 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1184}
d409b847 1185EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
72331bc0 1186
c71099ac
TG		 1187void ip6_route_input(struct sk_buff *skb)
1188{
b71d1d42 1189 const struct ipv6hdr *iph = ipv6_hdr(skb);
c346dca1 1190 struct net *net = dev_net(skb->dev);
adaa70bb 1191 int flags = RT6_LOOKUP_F_HAS_SADDR;
904af04d 1192 struct ip_tunnel_info *tun_info;
4c9483b2 1193 struct flowi6 fl6 = {
e0d56fdd 1194 .flowi6_iif = skb->dev->ifindex,
4c9483b2
DM		 1195 .daddr = iph->daddr,
1196 .saddr = iph->saddr,
6502ca52 1197 .flowlabel = ip6_flowinfo(iph),
4c9483b2
DM		 1198 .flowi6_mark = skb->mark,
1199 .flowi6_proto = iph->nexthdr,
c71099ac 1200 };
adaa70bb 1201
904af04d 1202 tun_info = skb_tunnel_info(skb);
46fa062a 1203 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
904af04d 1204 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
06e9d040 1205 skb_dst_drop(skb);
72331bc0 1206 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
c71099ac
TG		 1207}
1208
8ed67789 1209static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
4c9483b2 1210 struct flowi6 *fl6, int flags)
1da177e4 1211{
4c9483b2 1212 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
c71099ac
TG		 1213}
1214
6f21c96a
PA		 1215struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1216 struct flowi6 *fl6, int flags)
c71099ac 1217{
d46a9d67 1218 bool any_src;
c71099ac 1219
4c1feac5
DA		 1220 if (rt6_need_strict(&fl6->daddr)) {
1221 struct dst_entry *dst;
1222
1223 dst = l3mdev_link_scope_lookup(net, fl6);
1224 if (dst)
1225 return dst;
1226 }
ca254490 1227
1fb9489b 1228 fl6->flowi6_iif = LOOPBACK_IFINDEX;
4dc27d1c 1229
d46a9d67 1230 any_src = ipv6_addr_any(&fl6->saddr);
741a11d9 1231 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
d46a9d67 1232 (fl6->flowi6_oif && any_src))
77d16f45 1233 flags |= RT6_LOOKUP_F_IFACE;
c71099ac 1234
d46a9d67 1235 if (!any_src)
adaa70bb 1236 flags |= RT6_LOOKUP_F_HAS_SADDR;
0c9a2ac1
YH		 1237 else if (sk)
1238 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
adaa70bb 1239
4c9483b2 1240 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1da177e4 1241}
6f21c96a 1242EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1da177e4 1243
2774c131 1244struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
14e50e57 1245{
5c1e6aa3 1246 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
14e50e57
DM		 1247 struct dst_entry *new = NULL;
1248
f5b0a874 1249 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
14e50e57 1250 if (rt) {
0a1f5962 1251 rt6_info_init(rt);
8104891b 1252
0a1f5962 1253 new = &rt->dst;
14e50e57 1254 new->__use = 1;
352e512c 1255 new->input = dst_discard;
ede2059d 1256 new->output = dst_discard_out;
14e50e57 1257
0a1f5962 1258 dst_copy_metrics(new, &ort->dst);
14e50e57
DM		 1259 rt->rt6i_idev = ort->rt6i_idev;
1260 if (rt->rt6i_idev)
1261 in6_dev_hold(rt->rt6i_idev);
14e50e57 1262
4e3fd7a0 1263 rt->rt6i_gateway = ort->rt6i_gateway;
0a1f5962 1264 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
14e50e57
DM		 1265 rt->rt6i_metric = 0;
1266
1267 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1268#ifdef CONFIG_IPV6_SUBTREES
1269 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1270#endif
1271
1272 dst_free(new);
1273 }
1274
69ead7af
DM		 1275 dst_release(dst_orig);
1276 return new ? new : ERR_PTR(-ENOMEM);
14e50e57 1277}
14e50e57 1278
1da177e4
LT		 1279/*
1280 * Destination cache support functions
1281 */
1282
4b32b5ad
MKL		 1283static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1284{
1285 if (rt->dst.from &&
1286 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1287 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1288}
1289
3da59bd9
MKL		 1290static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1291{
1292 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1293 return NULL;
1294
1295 if (rt6_check_expired(rt))
1296 return NULL;
1297
1298 return &rt->dst;
1299}
1300
1301static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1302{
5973fb1e
MKL		 1303 if (!__rt6_check_expired(rt) &&
1304 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
3da59bd9
MKL		 1305 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1306 return &rt->dst;
1307 else
1308 return NULL;
1309}
1310
1da177e4
LT		 1311static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1312{
1313 struct rt6_info *rt;
1314
1315 rt = (struct rt6_info *) dst;
1316
6f3118b5
ND		 1317 /* All IPV6 dsts are created with ->obsolete set to the value
1318 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1319 * into this function always.
1320 */
e3bc10bd 1321
4b32b5ad
MKL		 1322 rt6_dst_from_metrics_check(rt);
1323
02bcf4e0
MKL		 1324 if (rt->rt6i_flags & RTF_PCPU ||
1325 (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
3da59bd9
MKL		 1326 return rt6_dst_from_check(rt, cookie);
1327 else
1328 return rt6_check(rt, cookie);
1da177e4
LT		 1329}
1330
1331static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1332{
1333 struct rt6_info *rt = (struct rt6_info *) dst;
1334
1335 if (rt) {
54c1a859
YH		 1336 if (rt->rt6i_flags & RTF_CACHE) {
1337 if (rt6_check_expired(rt)) {
1338 ip6_del_rt(rt);
1339 dst = NULL;
1340 }
1341 } else {
1da177e4 1342 dst_release(dst);
54c1a859
YH		 1343 dst = NULL;
1344 }
1da177e4 1345 }
54c1a859 1346 return dst;
1da177e4
LT		 1347}
1348
1349static void ip6_link_failure(struct sk_buff *skb)
1350{
1351 struct rt6_info *rt;
1352
3ffe533c 1353 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1da177e4 1354
adf30907 1355 rt = (struct rt6_info *) skb_dst(skb);
1da177e4 1356 if (rt) {
1eb4f758
HFS		 1357 if (rt->rt6i_flags & RTF_CACHE) {
1358 dst_hold(&rt->dst);
8e3d5be7 1359 ip6_del_rt(rt);
1eb4f758 1360 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1da177e4 1361 rt->rt6i_node->fn_sernum = -1;
1eb4f758 1362 }
1da177e4
LT		 1363 }
1364}
1365
45e4fd26
MKL		 1366static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1367{
1368 struct net *net = dev_net(rt->dst.dev);
1369
1370 rt->rt6i_flags |= RTF_MODIFIED;
1371 rt->rt6i_pmtu = mtu;
1372 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1373}
1374
0d3f6d29
MKL		 1375static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1376{
1377 return !(rt->rt6i_flags & RTF_CACHE) &&
1378 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1379}
1380
45e4fd26
MKL		 1381static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1382 const struct ipv6hdr *iph, u32 mtu)
1da177e4 1383{
0dec879f 1384 const struct in6_addr *daddr, *saddr;
67ba4152 1385 struct rt6_info *rt6 = (struct rt6_info *)dst;
1da177e4 1386
45e4fd26
MKL		 1387 if (rt6->rt6i_flags & RTF_LOCAL)
1388 return;
81aded24 1389
19bda36c
XL		 1390 if (dst_metric_locked(dst, RTAX_MTU))
1391 return;
1392
0dec879f
JA		 1393 if (iph) {
1394 daddr = &iph->daddr;
1395 saddr = &iph->saddr;
1396 } else if (sk) {
1397 daddr = &sk->sk_v6_daddr;
1398 saddr = &inet6_sk(sk)->saddr;
1399 } else {
1400 daddr = NULL;
1401 saddr = NULL;
1402 }
1403 dst_confirm_neigh(dst, daddr);
45e4fd26
MKL		 1404 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1405 if (mtu >= dst_mtu(dst))
1406 return;
9d289715 1407
0d3f6d29 1408 if (!rt6_cache_allowed_for_pmtu(rt6)) {
45e4fd26 1409 rt6_do_update_pmtu(rt6, mtu);
0dec879f 1410 } else if (daddr) {
45e4fd26
MKL		 1411 struct rt6_info *nrt6;
1412
45e4fd26
MKL		 1413 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1414 if (nrt6) {
1415 rt6_do_update_pmtu(nrt6, mtu);
1416
1417 /* ip6_ins_rt(nrt6) will bump the
1418 * rt6->rt6i_node->fn_sernum
1419 * which will fail the next rt6_check() and
1420 * invalidate the sk->sk_dst_cache.
1421 */
1422 ip6_ins_rt(nrt6);
1423 }
1da177e4
LT		 1424 }
1425}
1426
45e4fd26
MKL		 1427static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1428 struct sk_buff *skb, u32 mtu)
1429{
1430 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1431}
1432
42ae66c8 1433void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
e2d118a1 1434 int oif, u32 mark, kuid_t uid)
81aded24
DM		 1435{
1436 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1437 struct dst_entry *dst;
1438 struct flowi6 fl6;
1439
1440 memset(&fl6, 0, sizeof(fl6));
1441 fl6.flowi6_oif = oif;
1b3c61dc 1442 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
81aded24
DM		 1443 fl6.daddr = iph->daddr;
1444 fl6.saddr = iph->saddr;
6502ca52 1445 fl6.flowlabel = ip6_flowinfo(iph);
e2d118a1 1446 fl6.flowi6_uid = uid;
81aded24
DM		 1447
1448 dst = ip6_route_output(net, NULL, &fl6);
1449 if (!dst->error)
45e4fd26 1450 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
81aded24
DM		 1451 dst_release(dst);
1452}
1453EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1454
1455void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1456{
33c162a9
MKL		 1457 struct dst_entry *dst;
1458
81aded24 1459 ip6_update_pmtu(skb, sock_net(sk), mtu,
e2d118a1 1460 sk->sk_bound_dev_if, sk->sk_mark, sk->sk_uid);
33c162a9
MKL		 1461
1462 dst = __sk_dst_get(sk);
1463 if (!dst || !dst->obsolete ||
1464 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
1465 return;
1466
1467 bh_lock_sock(sk);
1468 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
1469 ip6_datagram_dst_update(sk, false);
1470 bh_unlock_sock(sk);
81aded24
DM		 1471}
1472EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1473
b55b76b2
DJ		 1474/* Handle redirects */
1475struct ip6rd_flowi {
1476 struct flowi6 fl6;
1477 struct in6_addr gateway;
1478};
1479
1480static struct rt6_info *__ip6_route_redirect(struct net *net,
1481 struct fib6_table *table,
1482 struct flowi6 *fl6,
1483 int flags)
1484{
1485 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1486 struct rt6_info *rt;
1487 struct fib6_node *fn;
1488
1489 /* Get the "current" route for this destination and
67c408cf 1490 * check if the redirect has come from appropriate router.
b55b76b2
DJ		 1491 *
1492 * RFC 4861 specifies that redirects should only be
1493 * accepted if they come from the nexthop to the target.
1494 * Due to the way the routes are chosen, this notion
1495 * is a bit fuzzy and one might need to check all possible
1496 * routes.
1497 */
1498
1499 read_lock_bh(&table->tb6_lock);
1500 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1501restart:
1502 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1503 if (rt6_check_expired(rt))
1504 continue;
1505 if (rt->dst.error)
1506 break;
1507 if (!(rt->rt6i_flags & RTF_GATEWAY))
1508 continue;
1509 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1510 continue;
1511 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1512 continue;
1513 break;
1514 }
1515
1516 if (!rt)
1517 rt = net->ipv6.ip6_null_entry;
1518 else if (rt->dst.error) {
1519 rt = net->ipv6.ip6_null_entry;
b0a1ba59
MKL		 1520 goto out;
1521 }
1522
1523 if (rt == net->ipv6.ip6_null_entry) {
a3c00e46
MKL		 1524 fn = fib6_backtrack(fn, &fl6->saddr);
1525 if (fn)
1526 goto restart;
b55b76b2 1527 }
a3c00e46 1528
b0a1ba59 1529out:
b55b76b2
DJ		 1530 dst_hold(&rt->dst);
1531
1532 read_unlock_bh(&table->tb6_lock);
1533
b811580d 1534 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
b55b76b2
DJ		 1535 return rt;
1536};
1537
1538static struct dst_entry *ip6_route_redirect(struct net *net,
1539 const struct flowi6 *fl6,
1540 const struct in6_addr *gateway)
1541{
1542 int flags = RT6_LOOKUP_F_HAS_SADDR;
1543 struct ip6rd_flowi rdfl;
1544
1545 rdfl.fl6 = *fl6;
1546 rdfl.gateway = *gateway;
1547
1548 return fib6_rule_lookup(net, &rdfl.fl6,
1549 flags, __ip6_route_redirect);
1550}
1551
e2d118a1
LC		 1552void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
1553 kuid_t uid)
3a5ad2ee
DM		 1554{
1555 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1556 struct dst_entry *dst;
1557 struct flowi6 fl6;
1558
1559 memset(&fl6, 0, sizeof(fl6));
e374c618 1560 fl6.flowi6_iif = LOOPBACK_IFINDEX;
3a5ad2ee
DM		 1561 fl6.flowi6_oif = oif;
1562 fl6.flowi6_mark = mark;
3a5ad2ee
DM		 1563 fl6.daddr = iph->daddr;
1564 fl6.saddr = iph->saddr;
6502ca52 1565 fl6.flowlabel = ip6_flowinfo(iph);
e2d118a1 1566 fl6.flowi6_uid = uid;
3a5ad2ee 1567
b55b76b2
DJ		 1568 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1569 rt6_do_redirect(dst, NULL, skb);
3a5ad2ee
DM		 1570 dst_release(dst);
1571}
1572EXPORT_SYMBOL_GPL(ip6_redirect);
1573
c92a59ec
DJ		 1574void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1575 u32 mark)
1576{
1577 const struct ipv6hdr *iph = ipv6_hdr(skb);
1578 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1579 struct dst_entry *dst;
1580 struct flowi6 fl6;
1581
1582 memset(&fl6, 0, sizeof(fl6));
e374c618 1583 fl6.flowi6_iif = LOOPBACK_IFINDEX;
c92a59ec
DJ		 1584 fl6.flowi6_oif = oif;
1585 fl6.flowi6_mark = mark;
c92a59ec
DJ		 1586 fl6.daddr = msg->dest;
1587 fl6.saddr = iph->daddr;
e2d118a1 1588 fl6.flowi6_uid = sock_net_uid(net, NULL);
c92a59ec 1589
b55b76b2
DJ		 1590 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1591 rt6_do_redirect(dst, NULL, skb);
c92a59ec
DJ		 1592 dst_release(dst);
1593}
1594
3a5ad2ee
DM		 1595void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1596{
e2d118a1
LC		 1597 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
1598 sk->sk_uid);
3a5ad2ee
DM		 1599}
1600EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1601
0dbaee3b 1602static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1da177e4 1603{
0dbaee3b
DM		 1604 struct net_device *dev = dst->dev;
1605 unsigned int mtu = dst_mtu(dst);
1606 struct net *net = dev_net(dev);
1607
1da177e4
LT		 1608 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1609
5578689a
DL		 1610 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1611 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1da177e4
LT		 1612
1613 /*
1ab1457c
YH		 1614 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1615 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1616 * IPV6_MAXPLEN is also valid and means: "any MSS,
1da177e4
LT		 1617 * rely only on pmtu discovery"
1618 */
1619 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1620 mtu = IPV6_MAXPLEN;
1621 return mtu;
1622}
1623
ebb762f2 1624static unsigned int ip6_mtu(const struct dst_entry *dst)
d33e4553 1625{
4b32b5ad
MKL		 1626 const struct rt6_info *rt = (const struct rt6_info *)dst;
1627 unsigned int mtu = rt->rt6i_pmtu;
d33e4553 1628 struct inet6_dev *idev;
618f9bc7 1629
4b32b5ad
MKL		 1630 if (mtu)
1631 goto out;
1632
1633 mtu = dst_metric_raw(dst, RTAX_MTU);
618f9bc7 1634 if (mtu)
30f78d8e 1635 goto out;
618f9bc7
SK		 1636
1637 mtu = IPV6_MIN_MTU;
d33e4553
DM		 1638
1639 rcu_read_lock();
1640 idev = __in6_dev_get(dst->dev);
1641 if (idev)
1642 mtu = idev->cnf.mtu6;
1643 rcu_read_unlock();
1644
30f78d8e 1645out:
14972cbd
RP		 1646 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1647
1648 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
d33e4553
DM		 1649}
1650
3b00944c
YH		 1651static struct dst_entry *icmp6_dst_gc_list;
1652static DEFINE_SPINLOCK(icmp6_dst_lock);
5d0bbeeb 1653
3b00944c 1654struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
87a11578 1655 struct flowi6 *fl6)
1da177e4 1656{
87a11578 1657 struct dst_entry *dst;
1da177e4
LT		 1658 struct rt6_info *rt;
1659 struct inet6_dev *idev = in6_dev_get(dev);
c346dca1 1660 struct net *net = dev_net(dev);
1da177e4 1661
38308473 1662 if (unlikely(!idev))
122bdf67 1663 return ERR_PTR(-ENODEV);
1da177e4 1664
ad706862 1665 rt = ip6_dst_alloc(net, dev, 0);
38308473 1666 if (unlikely(!rt)) {
1da177e4 1667 in6_dev_put(idev);
87a11578 1668 dst = ERR_PTR(-ENOMEM);
1da177e4
LT		 1669 goto out;
1670 }
1671
8e2ec639
YZ		 1672 rt->dst.flags |= DST_HOST;
1673 rt->dst.output = ip6_output;
d8d1f30b 1674 atomic_set(&rt->dst.__refcnt, 1);
550bab42 1675 rt->rt6i_gateway = fl6->daddr;
87a11578 1676 rt->rt6i_dst.addr = fl6->daddr;
8e2ec639
YZ		 1677 rt->rt6i_dst.plen = 128;
1678 rt->rt6i_idev = idev;
14edd87d 1679 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1da177e4 1680
3b00944c 1681 spin_lock_bh(&icmp6_dst_lock);
d8d1f30b
CG		 1682 rt->dst.next = icmp6_dst_gc_list;
1683 icmp6_dst_gc_list = &rt->dst;
3b00944c 1684 spin_unlock_bh(&icmp6_dst_lock);
1da177e4 1685
5578689a 1686 fib6_force_start_gc(net);
1da177e4 1687
87a11578
DM		 1688 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1689
1da177e4 1690out:
87a11578 1691 return dst;
1da177e4
LT		 1692}
1693
3d0f24a7 1694int icmp6_dst_gc(void)
1da177e4 1695{
e9476e95 1696 struct dst_entry *dst, **pprev;
3d0f24a7 1697 int more = 0;
1da177e4 1698
3b00944c
YH		 1699 spin_lock_bh(&icmp6_dst_lock);
1700 pprev = &icmp6_dst_gc_list;
5d0bbeeb 1701
1da177e4
LT		 1702 while ((dst = *pprev) != NULL) {
1703 if (!atomic_read(&dst->__refcnt)) {
1704 *pprev = dst->next;
1705 dst_free(dst);
1da177e4
LT		 1706 } else {
1707 pprev = &dst->next;
3d0f24a7 1708 ++more;
1da177e4
LT		 1709 }
1710 }
1711
3b00944c 1712 spin_unlock_bh(&icmp6_dst_lock);
5d0bbeeb 1713
3d0f24a7 1714 return more;
1da177e4
LT		 1715}
1716
1e493d19
DM		 1717static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1718 void *arg)
1719{
1720 struct dst_entry *dst, **pprev;
1721
1722 spin_lock_bh(&icmp6_dst_lock);
1723 pprev = &icmp6_dst_gc_list;
1724 while ((dst = *pprev) != NULL) {
1725 struct rt6_info *rt = (struct rt6_info *) dst;
1726 if (func(rt, arg)) {
1727 *pprev = dst->next;
1728 dst_free(dst);
1729 } else {
1730 pprev = &dst->next;
1731 }
1732 }
1733 spin_unlock_bh(&icmp6_dst_lock);
1734}
1735
569d3645 1736static int ip6_dst_gc(struct dst_ops *ops)
1da177e4 1737{
86393e52 1738 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
7019b78e
DL		 1739 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1740 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1741 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1742 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1743 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
fc66f95c 1744 int entries;
7019b78e 1745
fc66f95c 1746 entries = dst_entries_get_fast(ops);
49a18d86 1747 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
fc66f95c 1748 entries <= rt_max_size)
1da177e4
LT		 1749 goto out;
1750
6891a346 1751 net->ipv6.ip6_rt_gc_expire++;
14956643 1752 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
fc66f95c
ED		 1753 entries = dst_entries_get_slow(ops);
1754 if (entries < ops->gc_thresh)
7019b78e 1755 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1da177e4 1756out:
7019b78e 1757 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
fc66f95c 1758 return entries > rt_max_size;
1da177e4
LT		 1759}
1760
e715b6d3
FW		 1761static int ip6_convert_metrics(struct mx6_config *mxc,
1762 const struct fib6_config *cfg)
1763{
c3a8d947 1764 bool ecn_ca = false;
e715b6d3
FW		 1765 struct nlattr *nla;
1766 int remaining;
1767 u32 *mp;
1768
63159f29 1769 if (!cfg->fc_mx)
e715b6d3
FW		 1770 return 0;
1771
1772 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1773 if (unlikely(!mp))
1774 return -ENOMEM;
1775
1776 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1777 int type = nla_type(nla);
1bb14807 1778 u32 val;
e715b6d3 1779
1bb14807
DB		 1780 if (!type)
1781 continue;
1782 if (unlikely(type > RTAX_MAX))
1783 goto err;
ea697639 1784
1bb14807
DB		 1785 if (type == RTAX_CC_ALGO) {
1786 char tmp[TCP_CA_NAME_MAX];
e715b6d3 1787
1bb14807 1788 nla_strlcpy(tmp, nla, sizeof(tmp));
c3a8d947 1789 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1bb14807
DB		 1790 if (val == TCP_CA_UNSPEC)
1791 goto err;
1792 } else {
1793 val = nla_get_u32(nla);
e715b6d3 1794 }
626abd59
PA		 1795 if (type == RTAX_HOPLIMIT && val > 255)
1796 val = 255;
b8d3e416
DB		 1797 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1798 goto err;
1bb14807
DB		 1799
1800 mp[type - 1] = val;
1801 __set_bit(type - 1, mxc->mx_valid);
e715b6d3
FW		 1802 }
1803
c3a8d947
DB		 1804 if (ecn_ca) {
1805 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1806 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1807 }
e715b6d3 1808
c3a8d947 1809 mxc->mx = mp;
e715b6d3
FW		 1810 return 0;
1811 err:
1812 kfree(mp);
1813 return -EINVAL;
1814}
1da177e4 1815
8c14586f
DA		 1816static struct rt6_info *ip6_nh_lookup_table(struct net *net,
1817 struct fib6_config *cfg,
1818 const struct in6_addr *gw_addr)
1819{
1820 struct flowi6 fl6 = {
1821 .flowi6_oif = cfg->fc_ifindex,
1822 .daddr = *gw_addr,
1823 .saddr = cfg->fc_prefsrc,
1824 };
1825 struct fib6_table *table;
1826 struct rt6_info *rt;
d5d32e4b 1827 int flags = RT6_LOOKUP_F_IFACE | RT6_LOOKUP_F_IGNORE_LINKSTATE;
8c14586f
DA		 1828
1829 table = fib6_get_table(net, cfg->fc_table);
1830 if (!table)
1831 return NULL;
1832
1833 if (!ipv6_addr_any(&cfg->fc_prefsrc))
1834 flags |= RT6_LOOKUP_F_HAS_SADDR;
1835
1836 rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, flags);
1837
1838 /* if table lookup failed, fall back to full lookup */
1839 if (rt == net->ipv6.ip6_null_entry) {
1840 ip6_rt_put(rt);
1841 rt = NULL;
1842 }
1843
1844 return rt;
1845}
1846
8c5b83f0 1847static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1da177e4 1848{
5578689a 1849 struct net *net = cfg->fc_nlinfo.nl_net;
1da177e4
LT		 1850 struct rt6_info *rt = NULL;
1851 struct net_device *dev = NULL;
1852 struct inet6_dev *idev = NULL;
c71099ac 1853 struct fib6_table *table;
1da177e4 1854 int addr_type;
8c5b83f0 1855 int err = -EINVAL;
1da177e4 1856
557c44be
DA		 1857 /* RTF_PCPU is an internal flag; can not be set by userspace */
1858 if (cfg->fc_flags & RTF_PCPU)
1859 goto out;
1860
86872cb5 1861 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
8c5b83f0 1862 goto out;
1da177e4 1863#ifndef CONFIG_IPV6_SUBTREES
86872cb5 1864 if (cfg->fc_src_len)
8c5b83f0 1865 goto out;
1da177e4 1866#endif
86872cb5 1867 if (cfg->fc_ifindex) {
1da177e4 1868 err = -ENODEV;
5578689a 1869 dev = dev_get_by_index(net, cfg->fc_ifindex);
1da177e4
LT		 1870 if (!dev)
1871 goto out;
1872 idev = in6_dev_get(dev);
1873 if (!idev)
1874 goto out;
1875 }
1876
86872cb5
TG		 1877 if (cfg->fc_metric == 0)
1878 cfg->fc_metric = IP6_RT_PRIO_USER;
1da177e4 1879
d71314b4 1880 err = -ENOBUFS;
38308473
DM		 1881 if (cfg->fc_nlinfo.nlh &&
1882 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
d71314b4 1883 table = fib6_get_table(net, cfg->fc_table);
38308473 1884 if (!table) {
f3213831 1885 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
d71314b4
MV		 1886 table = fib6_new_table(net, cfg->fc_table);
1887 }
1888 } else {
1889 table = fib6_new_table(net, cfg->fc_table);
1890 }
38308473
DM		 1891
1892 if (!table)
c71099ac 1893 goto out;
c71099ac 1894
ad706862
MKL		 1895 rt = ip6_dst_alloc(net, NULL,
1896 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1da177e4 1897
38308473 1898 if (!rt) {
1da177e4
LT		 1899 err = -ENOMEM;
1900 goto out;
1901 }
1902
1716a961
G		 1903 if (cfg->fc_flags & RTF_EXPIRES)
1904 rt6_set_expires(rt, jiffies +
1905 clock_t_to_jiffies(cfg->fc_expires));
1906 else
1907 rt6_clean_expires(rt);
1da177e4 1908
86872cb5
TG		 1909 if (cfg->fc_protocol == RTPROT_UNSPEC)
1910 cfg->fc_protocol = RTPROT_BOOT;
1911 rt->rt6i_protocol = cfg->fc_protocol;
1912
1913 addr_type = ipv6_addr_type(&cfg->fc_dst);
1da177e4
LT		 1914
1915 if (addr_type & IPV6_ADDR_MULTICAST)
d8d1f30b 1916 rt->dst.input = ip6_mc_input;
ab79ad14
MŻ		 1917 else if (cfg->fc_flags & RTF_LOCAL)
1918 rt->dst.input = ip6_input;
1da177e4 1919 else
d8d1f30b 1920 rt->dst.input = ip6_forward;
1da177e4 1921
d8d1f30b 1922 rt->dst.output = ip6_output;
1da177e4 1923
19e42e45
RP		 1924 if (cfg->fc_encap) {
1925 struct lwtunnel_state *lwtstate;
1926
30357d7d 1927 err = lwtunnel_build_state(cfg->fc_encap_type,
127eb7cd
TH		 1928 cfg->fc_encap, AF_INET6, cfg,
1929 &lwtstate);
19e42e45
RP		 1930 if (err)
1931 goto out;
61adedf3
JB		 1932 rt->dst.lwtstate = lwtstate_get(lwtstate);
1933 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1934 rt->dst.lwtstate->orig_output = rt->dst.output;
1935 rt->dst.output = lwtunnel_output;
25368623 1936 }
61adedf3
JB		 1937 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1938 rt->dst.lwtstate->orig_input = rt->dst.input;
1939 rt->dst.input = lwtunnel_input;
25368623 1940 }
19e42e45
RP		 1941 }
1942
86872cb5
TG		 1943 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1944 rt->rt6i_dst.plen = cfg->fc_dst_len;
afc4eef8 1945 if (rt->rt6i_dst.plen == 128)
e5fd387a 1946 rt->dst.flags |= DST_HOST;
e5fd387a 1947
1da177e4 1948#ifdef CONFIG_IPV6_SUBTREES
86872cb5
TG		 1949 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1950 rt->rt6i_src.plen = cfg->fc_src_len;
1da177e4
LT		 1951#endif
1952
86872cb5 1953 rt->rt6i_metric = cfg->fc_metric;
1da177e4
LT		 1954
1955 /* We cannot add true routes via loopback here,
1956 they would result in kernel looping; promote them to reject routes
1957 */
86872cb5 1958 if ((cfg->fc_flags & RTF_REJECT) ||
38308473
DM		 1959 (dev && (dev->flags & IFF_LOOPBACK) &&
1960 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1961 !(cfg->fc_flags & RTF_LOCAL))) {
1da177e4 1962 /* hold loopback dev/idev if we haven't done so. */
5578689a 1963 if (dev != net->loopback_dev) {
1da177e4
LT		 1964 if (dev) {
1965 dev_put(dev);
1966 in6_dev_put(idev);
1967 }
5578689a 1968 dev = net->loopback_dev;
1da177e4
LT		 1969 dev_hold(dev);
1970 idev = in6_dev_get(dev);
1971 if (!idev) {
1972 err = -ENODEV;
1973 goto out;
1974 }
1975 }
1da177e4 1976 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
ef2c7d7b
ND		 1977 switch (cfg->fc_type) {
1978 case RTN_BLACKHOLE:
1979 rt->dst.error = -EINVAL;
ede2059d 1980 rt->dst.output = dst_discard_out;
7150aede 1981 rt->dst.input = dst_discard;
ef2c7d7b
ND		 1982 break;
1983 case RTN_PROHIBIT:
1984 rt->dst.error = -EACCES;
7150aede
K		 1985 rt->dst.output = ip6_pkt_prohibit_out;
1986 rt->dst.input = ip6_pkt_prohibit;
ef2c7d7b 1987 break;
b4949ab2 1988 case RTN_THROW:
0315e382 1989 case RTN_UNREACHABLE:
ef2c7d7b 1990 default:
7150aede 1991 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
0315e382
NF		 1992 : (cfg->fc_type == RTN_UNREACHABLE)
1993 ? -EHOSTUNREACH : -ENETUNREACH;
7150aede
K		 1994 rt->dst.output = ip6_pkt_discard_out;
1995 rt->dst.input = ip6_pkt_discard;
ef2c7d7b
ND		 1996 break;
1997 }
1da177e4
LT		 1998 goto install_route;
1999 }
2000
86872cb5 2001 if (cfg->fc_flags & RTF_GATEWAY) {
b71d1d42 2002 const struct in6_addr *gw_addr;
1da177e4
LT		 2003 int gwa_type;
2004
86872cb5 2005 gw_addr = &cfg->fc_gateway;
330567b7 2006 gwa_type = ipv6_addr_type(gw_addr);
48ed7b26
FW		 2007
2008 /* if gw_addr is local we will fail to detect this in case
2009 * address is still TENTATIVE (DAD in progress). rt6_lookup()
2010 * will return already-added prefix route via interface that
2011 * prefix route was assigned to, which might be non-loopback.
2012 */
2013 err = -EINVAL;
330567b7
FW		 2014 if (ipv6_chk_addr_and_flags(net, gw_addr,
2015 gwa_type & IPV6_ADDR_LINKLOCAL ?
2016 dev : NULL, 0, 0))
48ed7b26
FW		 2017 goto out;
2018
4e3fd7a0 2019 rt->rt6i_gateway = *gw_addr;
1da177e4
LT		 2020
2021 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
8c14586f 2022 struct rt6_info *grt = NULL;
1da177e4
LT		 2023
2024 /* IPv6 strictly inhibits using not link-local
2025 addresses as nexthop address.
2026 Otherwise, router will not able to send redirects.
2027 It is very good, but in some (rare!) circumstances
2028 (SIT, PtP, NBMA NOARP links) it is handy to allow
2029 some exceptions. --ANK
96d5822c
EN		 2030 We allow IPv4-mapped nexthops to support RFC4798-type
2031 addressing
1da177e4 2032 */
96d5822c
EN		 2033 if (!(gwa_type & (IPV6_ADDR_UNICAST |
2034 IPV6_ADDR_MAPPED)))
1da177e4
LT		 2035 goto out;
2036
a435a07f 2037 if (cfg->fc_table) {
8c14586f
DA		 2038 grt = ip6_nh_lookup_table(net, cfg, gw_addr);
2039
a435a07f
VB		 2040 if (grt) {
2041 if (grt->rt6i_flags & RTF_GATEWAY ||
2042 (dev && dev != grt->dst.dev)) {
2043 ip6_rt_put(grt);
2044 grt = NULL;
2045 }
2046 }
2047 }
2048
8c14586f
DA		 2049 if (!grt)
2050 grt = rt6_lookup(net, gw_addr, NULL,
2051 cfg->fc_ifindex, 1);
1da177e4
LT		 2052
2053 err = -EHOSTUNREACH;
38308473 2054 if (!grt)
1da177e4
LT		 2055 goto out;
2056 if (dev) {
d1918542 2057 if (dev != grt->dst.dev) {
94e187c0 2058 ip6_rt_put(grt);
1da177e4
LT		 2059 goto out;
2060 }
2061 } else {
d1918542 2062 dev = grt->dst.dev;
1da177e4
LT		 2063 idev = grt->rt6i_idev;
2064 dev_hold(dev);
2065 in6_dev_hold(grt->rt6i_idev);
2066 }
38308473 2067 if (!(grt->rt6i_flags & RTF_GATEWAY))
1da177e4 2068 err = 0;
94e187c0 2069 ip6_rt_put(grt);
1da177e4
LT		 2070
2071 if (err)
2072 goto out;
2073 }
2074 err = -EINVAL;
38308473 2075 if (!dev || (dev->flags & IFF_LOOPBACK))
1da177e4
LT		 2076 goto out;
2077 }
2078
2079 err = -ENODEV;
38308473 2080 if (!dev)
1da177e4
LT		 2081 goto out;
2082
c3968a85
DW		 2083 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
2084 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
2085 err = -EINVAL;
2086 goto out;
2087 }
4e3fd7a0 2088 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
c3968a85
DW		 2089 rt->rt6i_prefsrc.plen = 128;
2090 } else
2091 rt->rt6i_prefsrc.plen = 0;
2092
86872cb5 2093 rt->rt6i_flags = cfg->fc_flags;
1da177e4
LT		 2094
2095install_route:
d8d1f30b 2096 rt->dst.dev = dev;
1da177e4 2097 rt->rt6i_idev = idev;
c71099ac 2098 rt->rt6i_table = table;
63152fc0 2099
c346dca1 2100 cfg->fc_nlinfo.nl_net = dev_net(dev);
63152fc0 2101
8c5b83f0 2102 return rt;
6b9ea5a6
RP		 2103out:
2104 if (dev)
2105 dev_put(dev);
2106 if (idev)
2107 in6_dev_put(idev);
2108 if (rt)
2109 dst_free(&rt->dst);
2110
8c5b83f0 2111 return ERR_PTR(err);
6b9ea5a6
RP		 2112}
2113
2114int ip6_route_add(struct fib6_config *cfg)
2115{
2116 struct mx6_config mxc = { .mx = NULL, };
8c5b83f0 2117 struct rt6_info *rt;
6b9ea5a6
RP		 2118 int err;
2119
8c5b83f0
RP		 2120 rt = ip6_route_info_create(cfg);
2121 if (IS_ERR(rt)) {
2122 err = PTR_ERR(rt);
2123 rt = NULL;
6b9ea5a6 2124 goto out;
8c5b83f0 2125 }
6b9ea5a6 2126
e715b6d3
FW		 2127 err = ip6_convert_metrics(&mxc, cfg);
2128 if (err)
2129 goto out;
1da177e4 2130
e715b6d3
FW		 2131 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2132
2133 kfree(mxc.mx);
6b9ea5a6 2134
e715b6d3 2135 return err;
1da177e4 2136out:
1da177e4 2137 if (rt)
d8d1f30b 2138 dst_free(&rt->dst);
6b9ea5a6 2139
1da177e4
LT		 2140 return err;
2141}
2142
86872cb5 2143static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1da177e4
LT		 2144{
2145 int err;
c71099ac 2146 struct fib6_table *table;
d1918542 2147 struct net *net = dev_net(rt->dst.dev);
1da177e4 2148
8e3d5be7
MKL		 2149 if (rt == net->ipv6.ip6_null_entry ||
2150 rt->dst.flags & DST_NOCACHE) {
6825a26c
G		 2151 err = -ENOENT;
2152 goto out;
2153 }
6c813a72 2154
c71099ac
TG		 2155 table = rt->rt6i_table;
2156 write_lock_bh(&table->tb6_lock);
86872cb5 2157 err = fib6_del(rt, info);
c71099ac 2158 write_unlock_bh(&table->tb6_lock);
1da177e4 2159
6825a26c 2160out:
94e187c0 2161 ip6_rt_put(rt);
1da177e4
LT		 2162 return err;
2163}
2164
e0a1ad73
TG		 2165int ip6_del_rt(struct rt6_info *rt)
2166{
4d1169c1 2167 struct nl_info info = {
d1918542 2168 .nl_net = dev_net(rt->dst.dev),
4d1169c1 2169 };
528c4ceb 2170 return __ip6_del_rt(rt, &info);
e0a1ad73
TG		 2171}
2172
0ae81335
DA		 2173static int __ip6_del_rt_siblings(struct rt6_info *rt, struct fib6_config *cfg)
2174{
2175 struct nl_info *info = &cfg->fc_nlinfo;
e3330039 2176 struct net *net = info->nl_net;
16a16cd3 2177 struct sk_buff *skb = NULL;
0ae81335 2178 struct fib6_table *table;
e3330039 2179 int err = -ENOENT;
0ae81335 2180
e3330039
WC		 2181 if (rt == net->ipv6.ip6_null_entry)
2182 goto out_put;
0ae81335
DA		 2183 table = rt->rt6i_table;
2184 write_lock_bh(&table->tb6_lock);
2185
2186 if (rt->rt6i_nsiblings && cfg->fc_delete_all_nh) {
2187 struct rt6_info *sibling, *next_sibling;
2188
16a16cd3
DA		 2189 /* prefer to send a single notification with all hops */
2190 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
2191 if (skb) {
2192 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2193
e3330039 2194 if (rt6_fill_node(net, skb, rt,
16a16cd3
DA		 2195 NULL, NULL, 0, RTM_DELROUTE,
2196 info->portid, seq, 0) < 0) {
2197 kfree_skb(skb);
2198 skb = NULL;
2199 } else
2200 info->skip_notify = 1;
2201 }
2202
0ae81335
DA		 2203 list_for_each_entry_safe(sibling, next_sibling,
2204 &rt->rt6i_siblings,
2205 rt6i_siblings) {
2206 err = fib6_del(sibling, info);
2207 if (err)
e3330039 2208 goto out_unlock;
0ae81335
DA		 2209 }
2210 }
2211
2212 err = fib6_del(rt, info);
e3330039 2213out_unlock:
0ae81335 2214 write_unlock_bh(&table->tb6_lock);
e3330039 2215out_put:
0ae81335 2216 ip6_rt_put(rt);
16a16cd3
DA		 2217
2218 if (skb) {
e3330039 2219 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
16a16cd3
DA		 2220 info->nlh, gfp_any());
2221 }
0ae81335
DA		 2222 return err;
2223}
2224
86872cb5 2225static int ip6_route_del(struct fib6_config *cfg)
1da177e4 2226{
c71099ac 2227 struct fib6_table *table;
1da177e4
LT		 2228 struct fib6_node *fn;
2229 struct rt6_info *rt;
2230 int err = -ESRCH;
2231
5578689a 2232 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
38308473 2233 if (!table)
c71099ac
TG		 2234 return err;
2235
2236 read_lock_bh(&table->tb6_lock);
1da177e4 2237
c71099ac 2238 fn = fib6_locate(&table->tb6_root,
86872cb5
TG		 2239 &cfg->fc_dst, cfg->fc_dst_len,
2240 &cfg->fc_src, cfg->fc_src_len);
1ab1457c 2241
1da177e4 2242 if (fn) {
d8d1f30b 2243 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1f56a01f
MKL		 2244 if ((rt->rt6i_flags & RTF_CACHE) &&
2245 !(cfg->fc_flags & RTF_CACHE))
2246 continue;
86872cb5 2247 if (cfg->fc_ifindex &&
d1918542
DM		 2248 (!rt->dst.dev ||
2249 rt->dst.dev->ifindex != cfg->fc_ifindex))
1da177e4 2250 continue;
86872cb5
TG		 2251 if (cfg->fc_flags & RTF_GATEWAY &&
2252 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1da177e4 2253 continue;
86872cb5 2254 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1da177e4 2255 continue;
c2ed1880
M		 2256 if (cfg->fc_protocol && cfg->fc_protocol != rt->rt6i_protocol)
2257 continue;
d8d1f30b 2258 dst_hold(&rt->dst);
c71099ac 2259 read_unlock_bh(&table->tb6_lock);
1da177e4 2260
0ae81335
DA		 2261 /* if gateway was specified only delete the one hop */
2262 if (cfg->fc_flags & RTF_GATEWAY)
2263 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2264
2265 return __ip6_del_rt_siblings(rt, cfg);
1da177e4
LT		 2266 }
2267 }
c71099ac 2268 read_unlock_bh(&table->tb6_lock);
1da177e4
LT		 2269
2270 return err;
2271}
2272
6700c270 2273static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
a6279458 2274{
a6279458 2275 struct netevent_redirect netevent;
e8599ff4 2276 struct rt6_info *rt, *nrt = NULL;
e8599ff4
DM		 2277 struct ndisc_options ndopts;
2278 struct inet6_dev *in6_dev;
2279 struct neighbour *neigh;
71bcdba0 2280 struct rd_msg *msg;
6e157b6a
DM		 2281 int optlen, on_link;
2282 u8 *lladdr;
e8599ff4 2283
29a3cad5 2284 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
71bcdba0 2285 optlen -= sizeof(*msg);
e8599ff4
DM		 2286
2287 if (optlen < 0) {
6e157b6a 2288 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
e8599ff4
DM		 2289 return;
2290 }
2291
71bcdba0 2292 msg = (struct rd_msg *)icmp6_hdr(skb);
e8599ff4 2293
71bcdba0 2294 if (ipv6_addr_is_multicast(&msg->dest)) {
6e157b6a 2295 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
e8599ff4
DM		 2296 return;
2297 }
2298
6e157b6a 2299 on_link = 0;
71bcdba0 2300 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
e8599ff4 2301 on_link = 1;
71bcdba0 2302 } else if (ipv6_addr_type(&msg->target) !=
e8599ff4 2303 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
6e157b6a 2304 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
e8599ff4
DM		 2305 return;
2306 }
2307
2308 in6_dev = __in6_dev_get(skb->dev);
2309 if (!in6_dev)
2310 return;
2311 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2312 return;
2313
2314 /* RFC2461 8.1:
2315 * The IP source address of the Redirect MUST be the same as the current
2316 * first-hop router for the specified ICMP Destination Address.
2317 */
2318
f997c55c 2319 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
e8599ff4
DM		 2320 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2321 return;
2322 }
6e157b6a
DM		 2323
2324 lladdr = NULL;
e8599ff4
DM		 2325 if (ndopts.nd_opts_tgt_lladdr) {
2326 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2327 skb->dev);
2328 if (!lladdr) {
2329 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2330 return;
2331 }
2332 }
2333
6e157b6a 2334 rt = (struct rt6_info *) dst;
ec13ad1d 2335 if (rt->rt6i_flags & RTF_REJECT) {
6e157b6a 2336 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
e8599ff4 2337 return;
6e157b6a 2338 }
e8599ff4 2339
6e157b6a
DM		 2340 /* Redirect received -> path was valid.
2341 * Look, redirects are sent only in response to data packets,
2342 * so that this nexthop apparently is reachable. --ANK
2343 */
0dec879f 2344 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
a6279458 2345
71bcdba0 2346 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
6e157b6a
DM		 2347 if (!neigh)
2348 return;
a6279458 2349
1da177e4
LT		 2350 /*
2351 * We have finally decided to accept it.
2352 */
2353
f997c55c 2354 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
1da177e4
LT		 2355 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2356 NEIGH_UPDATE_F_OVERRIDE|
2357 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
f997c55c
AA		 2358 NEIGH_UPDATE_F_ISROUTER)),
2359 NDISC_REDIRECT, &ndopts);
1da177e4 2360
83a09abd 2361 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
38308473 2362 if (!nrt)
1da177e4
LT		 2363 goto out;
2364
2365 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2366 if (on_link)
2367 nrt->rt6i_flags &= ~RTF_GATEWAY;
2368
4e3fd7a0 2369 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1da177e4 2370
40e22e8f 2371 if (ip6_ins_rt(nrt))
1da177e4
LT		 2372 goto out;
2373
d8d1f30b
CG		 2374 netevent.old = &rt->dst;
2375 netevent.new = &nrt->dst;
71bcdba0 2376 netevent.daddr = &msg->dest;
60592833 2377 netevent.neigh = neigh;
8d71740c
TT		 2378 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2379
38308473 2380 if (rt->rt6i_flags & RTF_CACHE) {
6e157b6a 2381 rt = (struct rt6_info *) dst_clone(&rt->dst);
e0a1ad73 2382 ip6_del_rt(rt);
1da177e4
LT		 2383 }
2384
2385out:
e8599ff4 2386 neigh_release(neigh);
6e157b6a
DM		 2387}
2388
1da177e4
LT		 2389/*
2390 * Misc support functions
2391 */
2392
4b32b5ad
MKL		 2393static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2394{
2395 BUG_ON(from->dst.from);
2396
2397 rt->rt6i_flags &= ~RTF_EXPIRES;
2398 dst_hold(&from->dst);
2399 rt->dst.from = &from->dst;
2400 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2401}
2402
83a09abd
MKL		 2403static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2404{
2405 rt->dst.input = ort->dst.input;
2406 rt->dst.output = ort->dst.output;
2407 rt->rt6i_dst = ort->rt6i_dst;
2408 rt->dst.error = ort->dst.error;
2409 rt->rt6i_idev = ort->rt6i_idev;
2410 if (rt->rt6i_idev)
2411 in6_dev_hold(rt->rt6i_idev);
2412 rt->dst.lastuse = jiffies;
2413 rt->rt6i_gateway = ort->rt6i_gateway;
2414 rt->rt6i_flags = ort->rt6i_flags;
2415 rt6_set_from(rt, ort);
2416 rt->rt6i_metric = ort->rt6i_metric;
1da177e4 2417#ifdef CONFIG_IPV6_SUBTREES
83a09abd 2418 rt->rt6i_src = ort->rt6i_src;
1da177e4 2419#endif
83a09abd
MKL		 2420 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2421 rt->rt6i_table = ort->rt6i_table;
61adedf3 2422 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
1da177e4
LT		 2423}
2424
70ceb4f5 2425#ifdef CONFIG_IPV6_ROUTE_INFO
efa2cea0 2426static struct rt6_info *rt6_get_route_info(struct net *net,
b71d1d42 2427 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 2428 const struct in6_addr *gwaddr,
2429 struct net_device *dev)
70ceb4f5 2430{
830218c1
DA		 2431 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
2432 int ifindex = dev->ifindex;
70ceb4f5
YH		 2433 struct fib6_node *fn;
2434 struct rt6_info *rt = NULL;
c71099ac
TG		 2435 struct fib6_table *table;
2436
830218c1 2437 table = fib6_get_table(net, tb_id);
38308473 2438 if (!table)
c71099ac 2439 return NULL;
70ceb4f5 2440
5744dd9b 2441 read_lock_bh(&table->tb6_lock);
67ba4152 2442 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
70ceb4f5
YH		 2443 if (!fn)
2444 goto out;
2445
d8d1f30b 2446 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
d1918542 2447 if (rt->dst.dev->ifindex != ifindex)
70ceb4f5
YH		 2448 continue;
2449 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2450 continue;
2451 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2452 continue;
d8d1f30b 2453 dst_hold(&rt->dst);
70ceb4f5
YH		 2454 break;
2455 }
2456out:
5744dd9b 2457 read_unlock_bh(&table->tb6_lock);
70ceb4f5
YH		 2458 return rt;
2459}
2460
efa2cea0 2461static struct rt6_info *rt6_add_route_info(struct net *net,
b71d1d42 2462 const struct in6_addr *prefix, int prefixlen,
830218c1
DA		 2463 const struct in6_addr *gwaddr,
2464 struct net_device *dev,
95c96174 2465 unsigned int pref)
70ceb4f5 2466{
86872cb5 2467 struct fib6_config cfg = {
238fc7ea 2468 .fc_metric = IP6_RT_PRIO_USER,
830218c1 2469 .fc_ifindex = dev->ifindex,
86872cb5
TG		 2470 .fc_dst_len = prefixlen,
2471 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2472 RTF_UP | RTF_PREF(pref),
15e47304 2473 .fc_nlinfo.portid = 0,
efa2cea0
DL		 2474 .fc_nlinfo.nlh = NULL,
2475 .fc_nlinfo.nl_net = net,
86872cb5
TG		 2476 };
2477
830218c1 2478 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
4e3fd7a0
AD		 2479 cfg.fc_dst = *prefix;
2480 cfg.fc_gateway = *gwaddr;
70ceb4f5 2481
e317da96
YH		 2482 /* We should treat it as a default route if prefix length is 0. */
2483 if (!prefixlen)
86872cb5 2484 cfg.fc_flags |= RTF_DEFAULT;
70ceb4f5 2485
86872cb5 2486 ip6_route_add(&cfg);
70ceb4f5 2487
830218c1 2488 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
70ceb4f5
YH		 2489}
2490#endif
2491
b71d1d42 2492struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1ab1457c 2493{
830218c1 2494 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
1da177e4 2495 struct rt6_info *rt;
c71099ac 2496 struct fib6_table *table;
1da177e4 2497
830218c1 2498 table = fib6_get_table(dev_net(dev), tb_id);
38308473 2499 if (!table)
c71099ac 2500 return NULL;
1da177e4 2501
5744dd9b 2502 read_lock_bh(&table->tb6_lock);
67ba4152 2503 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
d1918542 2504 if (dev == rt->dst.dev &&
045927ff 2505 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1da177e4
LT		 2506 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2507 break;
2508 }
2509 if (rt)
d8d1f30b 2510 dst_hold(&rt->dst);
5744dd9b 2511 read_unlock_bh(&table->tb6_lock);
1da177e4
LT		 2512 return rt;
2513}
2514
b71d1d42 2515struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
ebacaaa0
YH		 2516 struct net_device *dev,
2517 unsigned int pref)
1da177e4 2518{
86872cb5 2519 struct fib6_config cfg = {
ca254490 2520 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
238fc7ea 2521 .fc_metric = IP6_RT_PRIO_USER,
86872cb5
TG		 2522 .fc_ifindex = dev->ifindex,
2523 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2524 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
15e47304 2525 .fc_nlinfo.portid = 0,
5578689a 2526 .fc_nlinfo.nlh = NULL,
c346dca1 2527 .fc_nlinfo.nl_net = dev_net(dev),
86872cb5 2528 };
1da177e4 2529
4e3fd7a0 2530 cfg.fc_gateway = *gwaddr;
1da177e4 2531
830218c1
DA		 2532 if (!ip6_route_add(&cfg)) {
2533 struct fib6_table *table;
2534
2535 table = fib6_get_table(dev_net(dev), cfg.fc_table);
2536 if (table)
2537 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
2538 }
1da177e4 2539
1da177e4
LT		 2540 return rt6_get_dflt_router(gwaddr, dev);
2541}
2542
830218c1 2543static void __rt6_purge_dflt_routers(struct fib6_table *table)
1da177e4
LT		 2544{
2545 struct rt6_info *rt;
2546
2547restart:
c71099ac 2548 read_lock_bh(&table->tb6_lock);
d8d1f30b 2549 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
3e8b0ac3
LC		 2550 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2551 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
d8d1f30b 2552 dst_hold(&rt->dst);
c71099ac 2553 read_unlock_bh(&table->tb6_lock);
e0a1ad73 2554 ip6_del_rt(rt);
1da177e4
LT		 2555 goto restart;
2556 }
2557 }
c71099ac 2558 read_unlock_bh(&table->tb6_lock);
830218c1
DA		 2559
2560 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
2561}
2562
2563void rt6_purge_dflt_routers(struct net *net)
2564{
2565 struct fib6_table *table;
2566 struct hlist_head *head;
2567 unsigned int h;
2568
2569 rcu_read_lock();
2570
2571 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
2572 head = &net->ipv6.fib_table_hash[h];
2573 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
2574 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
2575 __rt6_purge_dflt_routers(table);
2576 }
2577 }
2578
2579 rcu_read_unlock();
1da177e4
LT		 2580}
2581
5578689a
DL		 2582static void rtmsg_to_fib6_config(struct net *net,
2583 struct in6_rtmsg *rtmsg,
86872cb5
TG		 2584 struct fib6_config *cfg)
2585{
2586 memset(cfg, 0, sizeof(*cfg));
2587
ca254490
DA		 2588 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2589 : RT6_TABLE_MAIN;
86872cb5
TG		 2590 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2591 cfg->fc_metric = rtmsg->rtmsg_metric;
2592 cfg->fc_expires = rtmsg->rtmsg_info;
2593 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2594 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2595 cfg->fc_flags = rtmsg->rtmsg_flags;
2596
5578689a 2597 cfg->fc_nlinfo.nl_net = net;
f1243c2d 2598
4e3fd7a0
AD		 2599 cfg->fc_dst = rtmsg->rtmsg_dst;
2600 cfg->fc_src = rtmsg->rtmsg_src;
2601 cfg->fc_gateway = rtmsg->rtmsg_gateway;
86872cb5
TG		 2602}
2603
5578689a 2604int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1da177e4 2605{
86872cb5 2606 struct fib6_config cfg;
1da177e4
LT		 2607 struct in6_rtmsg rtmsg;
2608 int err;
2609
67ba4152 2610 switch (cmd) {
1da177e4
LT		 2611 case SIOCADDRT: /* Add a route */
2612 case SIOCDELRT: /* Delete a route */
af31f412 2613 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4
LT		 2614 return -EPERM;
2615 err = copy_from_user(&rtmsg, arg,
2616 sizeof(struct in6_rtmsg));
2617 if (err)
2618 return -EFAULT;
86872cb5 2619
5578689a 2620 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
86872cb5 2621
1da177e4
LT		 2622 rtnl_lock();
2623 switch (cmd) {
2624 case SIOCADDRT:
86872cb5 2625 err = ip6_route_add(&cfg);
1da177e4
LT		 2626 break;
2627 case SIOCDELRT:
86872cb5 2628 err = ip6_route_del(&cfg);
1da177e4
LT		 2629 break;
2630 default:
2631 err = -EINVAL;
2632 }
2633 rtnl_unlock();
2634
2635 return err;
3ff50b79 2636 }
1da177e4
LT		 2637
2638 return -EINVAL;
2639}
2640
2641/*
2642 * Drop the packet on the floor
2643 */
2644
d5fdd6ba 2645static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1da177e4 2646{
612f09e8 2647 int type;
adf30907 2648 struct dst_entry *dst = skb_dst(skb);
612f09e8
YH		 2649 switch (ipstats_mib_noroutes) {
2650 case IPSTATS_MIB_INNOROUTES:
0660e03f 2651 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
45bb0060 2652 if (type == IPV6_ADDR_ANY) {
3bd653c8
DL		 2653 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2654 IPSTATS_MIB_INADDRERRORS);
612f09e8
YH		 2655 break;
2656 }
2657 /* FALLTHROUGH */
2658 case IPSTATS_MIB_OUTNOROUTES:
3bd653c8
DL		 2659 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2660 ipstats_mib_noroutes);
612f09e8
YH		 2661 break;
2662 }
3ffe533c 2663 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1da177e4
LT		 2664 kfree_skb(skb);
2665 return 0;
2666}
2667
9ce8ade0
TG		 2668static int ip6_pkt_discard(struct sk_buff *skb)
2669{
612f09e8 2670 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 2671}
2672
ede2059d 2673static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
1da177e4 2674{
adf30907 2675 skb->dev = skb_dst(skb)->dev;
612f09e8 2676 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1da177e4
LT		 2677}
2678
9ce8ade0
TG		 2679static int ip6_pkt_prohibit(struct sk_buff *skb)
2680{
612f09e8 2681 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
9ce8ade0
TG		 2682}
2683
ede2059d 2684static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
9ce8ade0 2685{
adf30907 2686 skb->dev = skb_dst(skb)->dev;
612f09e8 2687 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
9ce8ade0
TG		 2688}
2689
1da177e4
LT		 2690/*
2691 * Allocate a dst for local (unicast / anycast) address.
2692 */
2693
2694struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2695 const struct in6_addr *addr,
8f031519 2696 bool anycast)
1da177e4 2697{
ca254490 2698 u32 tb_id;
c346dca1 2699 struct net *net = dev_net(idev->dev);
5f02ce24
DA		 2700 struct net_device *dev = net->loopback_dev;
2701 struct rt6_info *rt;
2702
2703 /* use L3 Master device as loopback for host routes if device
2704 * is enslaved and address is not link local or multicast
2705 */
2706 if (!rt6_need_strict(addr))
2707 dev = l3mdev_master_dev_rcu(idev->dev) ? : dev;
2708
2709 rt = ip6_dst_alloc(net, dev, DST_NOCOUNT);
a3300ef4 2710 if (!rt)
1da177e4
LT		 2711 return ERR_PTR(-ENOMEM);
2712
1da177e4
LT		 2713 in6_dev_hold(idev);
2714
11d53b49 2715 rt->dst.flags |= DST_HOST;
d8d1f30b
CG		 2716 rt->dst.input = ip6_input;
2717 rt->dst.output = ip6_output;
1da177e4 2718 rt->rt6i_idev = idev;
1da177e4 2719
94b5e0f9 2720 rt->rt6i_protocol = RTPROT_KERNEL;
1da177e4 2721 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
58c4fb86
YH		 2722 if (anycast)
2723 rt->rt6i_flags |= RTF_ANYCAST;
2724 else
1da177e4 2725 rt->rt6i_flags |= RTF_LOCAL;
1da177e4 2726
550bab42 2727 rt->rt6i_gateway = *addr;
4e3fd7a0 2728 rt->rt6i_dst.addr = *addr;
1da177e4 2729 rt->rt6i_dst.plen = 128;
ca254490
DA		 2730 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2731 rt->rt6i_table = fib6_get_table(net, tb_id);
8e3d5be7 2732 rt->dst.flags |= DST_NOCACHE;
1da177e4 2733
d8d1f30b 2734 atomic_set(&rt->dst.__refcnt, 1);
1da177e4
LT		 2735
2736 return rt;
2737}
2738
c3968a85
DW		 2739/* remove deleted ip from prefsrc entries */
2740struct arg_dev_net_ip {
2741 struct net_device *dev;
2742 struct net *net;
2743 struct in6_addr *addr;
2744};
2745
2746static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2747{
2748 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2749 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2750 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2751
d1918542 2752 if (((void *)rt->dst.dev == dev || !dev) &&
c3968a85
DW		 2753 rt != net->ipv6.ip6_null_entry &&
2754 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2755 /* remove prefsrc entry */
2756 rt->rt6i_prefsrc.plen = 0;
2757 }
2758 return 0;
2759}
2760
2761void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2762{
2763 struct net *net = dev_net(ifp->idev->dev);
2764 struct arg_dev_net_ip adni = {
2765 .dev = ifp->idev->dev,
2766 .net = net,
2767 .addr = &ifp->addr,
2768 };
0c3584d5 2769 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
c3968a85
DW		 2770}
2771
be7a010d
DJ		 2772#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2773#define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2774
2775/* Remove routers and update dst entries when gateway turn into host. */
2776static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2777{
2778 struct in6_addr *gateway = (struct in6_addr *)arg;
2779
2780 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2781 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2782 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2783 return -1;
2784 }
2785 return 0;
2786}
2787
2788void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2789{
2790 fib6_clean_all(net, fib6_clean_tohost, gateway);
2791}
2792
8ed67789
DL		 2793struct arg_dev_net {
2794 struct net_device *dev;
2795 struct net *net;
2796};
2797
a1a22c12 2798/* called with write lock held for table with rt */
1da177e4
LT		 2799static int fib6_ifdown(struct rt6_info *rt, void *arg)
2800{
bc3ef660 2801 const struct arg_dev_net *adn = arg;
2802 const struct net_device *dev = adn->dev;
8ed67789 2803
d1918542 2804 if ((rt->dst.dev == dev || !dev) &&
a1a22c12
DA		 2805 rt != adn->net->ipv6.ip6_null_entry &&
2806 (rt->rt6i_nsiblings == 0 ||
2807 !rt->rt6i_idev->cnf.ignore_routes_with_linkdown))
1da177e4 2808 return -1;
c159d30c 2809
1da177e4
LT		 2810 return 0;
2811}
2812
f3db4851 2813void rt6_ifdown(struct net *net, struct net_device *dev)
1da177e4 2814{
8ed67789
DL		 2815 struct arg_dev_net adn = {
2816 .dev = dev,
2817 .net = net,
2818 };
2819
0c3584d5 2820 fib6_clean_all(net, fib6_ifdown, &adn);
1e493d19 2821 icmp6_clean_all(fib6_ifdown, &adn);
e332bc67
EB		 2822 if (dev)
2823 rt6_uncached_list_flush_dev(net, dev);
1da177e4
LT		 2824}
2825
95c96174 2826struct rt6_mtu_change_arg {
1da177e4 2827 struct net_device *dev;
95c96174 2828 unsigned int mtu;
1da177e4
LT		 2829};
2830
2831static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2832{
2833 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2834 struct inet6_dev *idev;
2835
2836 /* In IPv6 pmtu discovery is not optional,
2837 so that RTAX_MTU lock cannot disable it.
2838 We still use this lock to block changes
2839 caused by addrconf/ndisc.
2840 */
2841
2842 idev = __in6_dev_get(arg->dev);
38308473 2843 if (!idev)
1da177e4
LT		 2844 return 0;
2845
2846 /* For administrative MTU increase, there is no way to discover
2847 IPv6 PMTU increase, so PMTU increase should be updated here.
2848 Since RFC 1981 doesn't include administrative MTU increase
2849 update PMTU increase is a MUST. (i.e. jumbo frame)
2850 */
2851 /*
2852 If new MTU is less than route PMTU, this new MTU will be the
2853 lowest MTU in the path, update the route PMTU to reflect PMTU
2854 decreases; if new MTU is greater than route PMTU, and the
2855 old MTU is the lowest MTU in the path, update the route PMTU
2856 to reflect the increase. In this case if the other nodes' MTU
2857 also have the lowest MTU, TOO BIG MESSAGE will be lead to
67c408cf 2858 PMTU discovery.
1da177e4 2859 */
d1918542 2860 if (rt->dst.dev == arg->dev &&
fb56be83 2861 dst_metric_raw(&rt->dst, RTAX_MTU) &&
4b32b5ad
MKL		 2862 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2863 if (rt->rt6i_flags & RTF_CACHE) {
2864 /* For RTF_CACHE with rt6i_pmtu == 0
2865 * (i.e. a redirected route),
2866 * the metrics of its rt->dst.from has already
2867 * been updated.
2868 */
2869 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2870 rt->rt6i_pmtu = arg->mtu;
2871 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2872 (dst_mtu(&rt->dst) < arg->mtu &&
2873 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2874 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2875 }
566cfd8f 2876 }
1da177e4
LT		 2877 return 0;
2878}
2879
95c96174 2880void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
1da177e4 2881{
c71099ac
TG		 2882 struct rt6_mtu_change_arg arg = {
2883 .dev = dev,
2884 .mtu = mtu,
2885 };
1da177e4 2886
0c3584d5 2887 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
1da177e4
LT		 2888}
2889
ef7c79ed 2890static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
5176f91e 2891 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
86872cb5 2892 [RTA_OIF] = { .type = NLA_U32 },
ab364a6f 2893 [RTA_IIF] = { .type = NLA_U32 },
86872cb5
TG		 2894 [RTA_PRIORITY] = { .type = NLA_U32 },
2895 [RTA_METRICS] = { .type = NLA_NESTED },
51ebd318 2896 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
c78ba6d6 2897 [RTA_PREF] = { .type = NLA_U8 },
19e42e45
RP		 2898 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2899 [RTA_ENCAP] = { .type = NLA_NESTED },
32bc201e 2900 [RTA_EXPIRES] = { .type = NLA_U32 },
622ec2c9 2901 [RTA_UID] = { .type = NLA_U32 },
3b45a410 2902 [RTA_MARK] = { .type = NLA_U32 },
86872cb5
TG		 2903};
2904
2905static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2906 struct fib6_config *cfg)
1da177e4 2907{
86872cb5
TG		 2908 struct rtmsg *rtm;
2909 struct nlattr *tb[RTA_MAX+1];
c78ba6d6 2910 unsigned int pref;
86872cb5 2911 int err;
1da177e4 2912
fceb6435
JB		 2913 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
2914 NULL);
86872cb5
TG		 2915 if (err < 0)
2916 goto errout;
1da177e4 2917
86872cb5
TG		 2918 err = -EINVAL;
2919 rtm = nlmsg_data(nlh);
2920 memset(cfg, 0, sizeof(*cfg));
2921
2922 cfg->fc_table = rtm->rtm_table;
2923 cfg->fc_dst_len = rtm->rtm_dst_len;
2924 cfg->fc_src_len = rtm->rtm_src_len;
2925 cfg->fc_flags = RTF_UP;
2926 cfg->fc_protocol = rtm->rtm_protocol;
ef2c7d7b 2927 cfg->fc_type = rtm->rtm_type;
86872cb5 2928
ef2c7d7b
ND		 2929 if (rtm->rtm_type == RTN_UNREACHABLE ||
2930 rtm->rtm_type == RTN_BLACKHOLE ||
b4949ab2
ND		 2931 rtm->rtm_type == RTN_PROHIBIT ||
2932 rtm->rtm_type == RTN_THROW)
86872cb5
TG		 2933 cfg->fc_flags |= RTF_REJECT;
2934
ab79ad14
MŻ		 2935 if (rtm->rtm_type == RTN_LOCAL)
2936 cfg->fc_flags |= RTF_LOCAL;
2937
1f56a01f
MKL		 2938 if (rtm->rtm_flags & RTM_F_CLONED)
2939 cfg->fc_flags |= RTF_CACHE;
2940
15e47304 2941 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
86872cb5 2942 cfg->fc_nlinfo.nlh = nlh;
3b1e0a65 2943 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
86872cb5
TG		 2944
2945 if (tb[RTA_GATEWAY]) {
67b61f6c 2946 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
86872cb5 2947 cfg->fc_flags |= RTF_GATEWAY;
1da177e4 2948 }
86872cb5
TG		 2949
2950 if (tb[RTA_DST]) {
2951 int plen = (rtm->rtm_dst_len + 7) >> 3;
2952
2953 if (nla_len(tb[RTA_DST]) < plen)
2954 goto errout;
2955
2956 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1da177e4 2957 }
86872cb5
TG		 2958
2959 if (tb[RTA_SRC]) {
2960 int plen = (rtm->rtm_src_len + 7) >> 3;
2961
2962 if (nla_len(tb[RTA_SRC]) < plen)
2963 goto errout;
2964
2965 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1da177e4 2966 }
86872cb5 2967
c3968a85 2968 if (tb[RTA_PREFSRC])
67b61f6c 2969 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
c3968a85 2970
86872cb5
TG		 2971 if (tb[RTA_OIF])
2972 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2973
2974 if (tb[RTA_PRIORITY])
2975 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2976
2977 if (tb[RTA_METRICS]) {
2978 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2979 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
1da177e4 2980 }
86872cb5
TG		 2981
2982 if (tb[RTA_TABLE])
2983 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2984
51ebd318
ND		 2985 if (tb[RTA_MULTIPATH]) {
2986 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2987 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
9ed59592
DA		 2988
2989 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
2990 cfg->fc_mp_len);
2991 if (err < 0)
2992 goto errout;
51ebd318
ND		 2993 }
2994
c78ba6d6
LR		 2995 if (tb[RTA_PREF]) {
2996 pref = nla_get_u8(tb[RTA_PREF]);
2997 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2998 pref != ICMPV6_ROUTER_PREF_HIGH)
2999 pref = ICMPV6_ROUTER_PREF_MEDIUM;
3000 cfg->fc_flags |= RTF_PREF(pref);
3001 }
3002
19e42e45
RP		 3003 if (tb[RTA_ENCAP])
3004 cfg->fc_encap = tb[RTA_ENCAP];
3005
9ed59592 3006 if (tb[RTA_ENCAP_TYPE]) {
19e42e45
RP		 3007 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
3008
9ed59592
DA		 3009 err = lwtunnel_valid_encap_type(cfg->fc_encap_type);
3010 if (err < 0)
3011 goto errout;
3012 }
3013
32bc201e
XL		 3014 if (tb[RTA_EXPIRES]) {
3015 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
3016
3017 if (addrconf_finite_timeout(timeout)) {
3018 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
3019 cfg->fc_flags |= RTF_EXPIRES;
3020 }
3021 }
3022
86872cb5
TG		 3023 err = 0;
3024errout:
3025 return err;
1da177e4
LT		 3026}
3027
6b9ea5a6
RP		 3028struct rt6_nh {
3029 struct rt6_info *rt6_info;
3030 struct fib6_config r_cfg;
3031 struct mx6_config mxc;
3032 struct list_head next;
3033};
3034
3035static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
3036{
3037 struct rt6_nh *nh;
3038
3039 list_for_each_entry(nh, rt6_nh_list, next) {
7d4d5065 3040 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6c nexthop %pI6c ifi %d\n",
6b9ea5a6
RP		 3041 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
3042 nh->r_cfg.fc_ifindex);
3043 }
3044}
3045
3046static int ip6_route_info_append(struct list_head *rt6_nh_list,
3047 struct rt6_info *rt, struct fib6_config *r_cfg)
3048{
3049 struct rt6_nh *nh;
3050 struct rt6_info *rtnh;
3051 int err = -EEXIST;
3052
3053 list_for_each_entry(nh, rt6_nh_list, next) {
3054 /* check if rt6_info already exists */
3055 rtnh = nh->rt6_info;
3056
3057 if (rtnh->dst.dev == rt->dst.dev &&
3058 rtnh->rt6i_idev == rt->rt6i_idev &&
3059 ipv6_addr_equal(&rtnh->rt6i_gateway,
3060 &rt->rt6i_gateway))
3061 return err;
3062 }
3063
3064 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
3065 if (!nh)
3066 return -ENOMEM;
3067 nh->rt6_info = rt;
3068 err = ip6_convert_metrics(&nh->mxc, r_cfg);
3069 if (err) {
3070 kfree(nh);
3071 return err;
3072 }
3073 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
3074 list_add_tail(&nh->next, rt6_nh_list);
3075
3076 return 0;
3077}
3078
3b1137fe
DA		 3079static void ip6_route_mpath_notify(struct rt6_info *rt,
3080 struct rt6_info *rt_last,
3081 struct nl_info *info,
3082 __u16 nlflags)
3083{
3084 /* if this is an APPEND route, then rt points to the first route
3085 * inserted and rt_last points to last route inserted. Userspace
3086 * wants a consistent dump of the route which starts at the first
3087 * nexthop. Since sibling routes are always added at the end of
3088 * the list, find the first sibling of the last route appended
3089 */
3090 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->rt6i_nsiblings) {
3091 rt = list_first_entry(&rt_last->rt6i_siblings,
3092 struct rt6_info,
3093 rt6i_siblings);
3094 }
3095
3096 if (rt)
3097 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
3098}
3099
6b9ea5a6 3100static int ip6_route_multipath_add(struct fib6_config *cfg)
51ebd318 3101{
3b1137fe
DA		 3102 struct rt6_info *rt_notif = NULL, *rt_last = NULL;
3103 struct nl_info *info = &cfg->fc_nlinfo;
51ebd318
ND		 3104 struct fib6_config r_cfg;
3105 struct rtnexthop *rtnh;
6b9ea5a6
RP		 3106 struct rt6_info *rt;
3107 struct rt6_nh *err_nh;
3108 struct rt6_nh *nh, *nh_safe;
3b1137fe 3109 __u16 nlflags;
51ebd318
ND		 3110 int remaining;
3111 int attrlen;
6b9ea5a6
RP		 3112 int err = 1;
3113 int nhn = 0;
3114 int replace = (cfg->fc_nlinfo.nlh &&
3115 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
3116 LIST_HEAD(rt6_nh_list);
51ebd318 3117
3b1137fe
DA		 3118 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
3119 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
3120 nlflags |= NLM_F_APPEND;
3121
35f1b4e9 3122 remaining = cfg->fc_mp_len;
51ebd318 3123 rtnh = (struct rtnexthop *)cfg->fc_mp;
51ebd318 3124
6b9ea5a6
RP		 3125 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
3126 * rt6_info structs per nexthop
3127 */
51ebd318
ND		 3128 while (rtnh_ok(rtnh, remaining)) {
3129 memcpy(&r_cfg, cfg, sizeof(*cfg));
3130 if (rtnh->rtnh_ifindex)
3131 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3132
3133 attrlen = rtnh_attrlen(rtnh);
3134 if (attrlen > 0) {
3135 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3136
3137 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3138 if (nla) {
67b61f6c 3139 r_cfg.fc_gateway = nla_get_in6_addr(nla);
51ebd318
ND		 3140 r_cfg.fc_flags |= RTF_GATEWAY;
3141 }
19e42e45
RP		 3142 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
3143 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
3144 if (nla)
3145 r_cfg.fc_encap_type = nla_get_u16(nla);
51ebd318 3146 }
6b9ea5a6 3147
8c5b83f0
RP		 3148 rt = ip6_route_info_create(&r_cfg);
3149 if (IS_ERR(rt)) {
3150 err = PTR_ERR(rt);
3151 rt = NULL;
6b9ea5a6 3152 goto cleanup;
8c5b83f0 3153 }
6b9ea5a6
RP		 3154
3155 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
51ebd318 3156 if (err) {
6b9ea5a6
RP		 3157 dst_free(&rt->dst);
3158 goto cleanup;
3159 }
3160
3161 rtnh = rtnh_next(rtnh, &remaining);
3162 }
3163
3b1137fe
DA		 3164 /* for add and replace send one notification with all nexthops.
3165 * Skip the notification in fib6_add_rt2node and send one with
3166 * the full route when done
3167 */
3168 info->skip_notify = 1;
3169
6b9ea5a6
RP		 3170 err_nh = NULL;
3171 list_for_each_entry(nh, &rt6_nh_list, next) {
3b1137fe
DA		 3172 rt_last = nh->rt6_info;
3173 err = __ip6_ins_rt(nh->rt6_info, info, &nh->mxc);
3174 /* save reference to first route for notification */
3175 if (!rt_notif && !err)
3176 rt_notif = nh->rt6_info;
3177
6b9ea5a6
RP		 3178 /* nh->rt6_info is used or freed at this point, reset to NULL*/
3179 nh->rt6_info = NULL;
3180 if (err) {
3181 if (replace && nhn)
3182 ip6_print_replace_route_err(&rt6_nh_list);
3183 err_nh = nh;
3184 goto add_errout;
51ebd318 3185 }
6b9ea5a6 3186
1a72418b 3187 /* Because each route is added like a single route we remove
27596472
MK		 3188 * these flags after the first nexthop: if there is a collision,
3189 * we have already failed to add the first nexthop:
3190 * fib6_add_rt2node() has rejected it; when replacing, old
3191 * nexthops have been replaced by first new, the rest should
3192 * be added to it.
1a72418b 3193 */
27596472
MK		 3194 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
3195 NLM_F_REPLACE);
6b9ea5a6
RP		 3196 nhn++;
3197 }
3198
3b1137fe
DA		 3199 /* success ... tell user about new route */
3200 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
6b9ea5a6
RP		 3201 goto cleanup;
3202
3203add_errout:
3b1137fe
DA		 3204 /* send notification for routes that were added so that
3205 * the delete notifications sent by ip6_route_del are
3206 * coherent
3207 */
3208 if (rt_notif)
3209 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
3210
6b9ea5a6
RP		 3211 /* Delete routes that were already added */
3212 list_for_each_entry(nh, &rt6_nh_list, next) {
3213 if (err_nh == nh)
3214 break;
3215 ip6_route_del(&nh->r_cfg);
3216 }
3217
3218cleanup:
3219 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
3220 if (nh->rt6_info)
3221 dst_free(&nh->rt6_info->dst);
52fe51f8 3222 kfree(nh->mxc.mx);
6b9ea5a6
RP		 3223 list_del(&nh->next);
3224 kfree(nh);
3225 }
3226
3227 return err;
3228}
3229
3230static int ip6_route_multipath_del(struct fib6_config *cfg)
3231{
3232 struct fib6_config r_cfg;
3233 struct rtnexthop *rtnh;
3234 int remaining;
3235 int attrlen;
3236 int err = 1, last_err = 0;
3237
3238 remaining = cfg->fc_mp_len;
3239 rtnh = (struct rtnexthop *)cfg->fc_mp;
3240
3241 /* Parse a Multipath Entry */
3242 while (rtnh_ok(rtnh, remaining)) {
3243 memcpy(&r_cfg, cfg, sizeof(*cfg));
3244 if (rtnh->rtnh_ifindex)
3245 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3246
3247 attrlen = rtnh_attrlen(rtnh);
3248 if (attrlen > 0) {
3249 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3250
3251 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3252 if (nla) {
3253 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3254 r_cfg.fc_flags |= RTF_GATEWAY;
3255 }
3256 }
3257 err = ip6_route_del(&r_cfg);
3258 if (err)
3259 last_err = err;
3260
51ebd318
ND		 3261 rtnh = rtnh_next(rtnh, &remaining);
3262 }
3263
3264 return last_err;
3265}
3266
c21ef3e3
DA		 3267static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
3268 struct netlink_ext_ack *extack)
1da177e4 3269{
86872cb5
TG		 3270 struct fib6_config cfg;
3271 int err;
1da177e4 3272
86872cb5
TG		 3273 err = rtm_to_fib6_config(skb, nlh, &cfg);
3274 if (err < 0)
3275 return err;
3276
51ebd318 3277 if (cfg.fc_mp)
6b9ea5a6 3278 return ip6_route_multipath_del(&cfg);
0ae81335
DA		 3279 else {
3280 cfg.fc_delete_all_nh = 1;
51ebd318 3281 return ip6_route_del(&cfg);
0ae81335 3282 }
1da177e4
LT		 3283}
3284
c21ef3e3
DA		 3285static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
3286 struct netlink_ext_ack *extack)
1da177e4 3287{
86872cb5
TG		 3288 struct fib6_config cfg;
3289 int err;
1da177e4 3290
86872cb5
TG		 3291 err = rtm_to_fib6_config(skb, nlh, &cfg);
3292 if (err < 0)
3293 return err;
3294
51ebd318 3295 if (cfg.fc_mp)
6b9ea5a6 3296 return ip6_route_multipath_add(&cfg);
51ebd318
ND		 3297 else
3298 return ip6_route_add(&cfg);
1da177e4
LT		 3299}
3300
beb1afac 3301static size_t rt6_nlmsg_size(struct rt6_info *rt)
339bf98f 3302{
beb1afac
DA		 3303 int nexthop_len = 0;
3304
3305 if (rt->rt6i_nsiblings) {
3306 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
3307 + NLA_ALIGN(sizeof(struct rtnexthop))
3308 + nla_total_size(16) /* RTA_GATEWAY */
beb1afac
DA		 3309 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3310
3311 nexthop_len *= rt->rt6i_nsiblings;
3312 }
3313
339bf98f
TG		 3314 return NLMSG_ALIGN(sizeof(struct rtmsg))
3315 + nla_total_size(16) /* RTA_SRC */
3316 + nla_total_size(16) /* RTA_DST */
3317 + nla_total_size(16) /* RTA_GATEWAY */
3318 + nla_total_size(16) /* RTA_PREFSRC */
3319 + nla_total_size(4) /* RTA_TABLE */
3320 + nla_total_size(4) /* RTA_IIF */
3321 + nla_total_size(4) /* RTA_OIF */
3322 + nla_total_size(4) /* RTA_PRIORITY */
6a2b9ce0 3323 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
ea697639 3324 + nla_total_size(sizeof(struct rta_cacheinfo))
c78ba6d6 3325 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
19e42e45 3326 + nla_total_size(1) /* RTA_PREF */
beb1afac
DA		 3327 + lwtunnel_get_encap_size(rt->dst.lwtstate)
3328 + nexthop_len;
3329}
3330
3331static int rt6_nexthop_info(struct sk_buff *skb, struct rt6_info *rt,
5be083ce 3332 unsigned int *flags, bool skip_oif)
beb1afac
DA		 3333{
3334 if (!netif_running(rt->dst.dev) || !netif_carrier_ok(rt->dst.dev)) {
3335 *flags |= RTNH_F_LINKDOWN;
3336 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3337 *flags |= RTNH_F_DEAD;
3338 }
3339
3340 if (rt->rt6i_flags & RTF_GATEWAY) {
3341 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3342 goto nla_put_failure;
3343 }
3344
5be083ce
DA		 3345 /* not needed for multipath encoding b/c it has a rtnexthop struct */
3346 if (!skip_oif && rt->dst.dev &&
beb1afac
DA		 3347 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3348 goto nla_put_failure;
3349
3350 if (rt->dst.lwtstate &&
3351 lwtunnel_fill_encap(skb, rt->dst.lwtstate) < 0)
3352 goto nla_put_failure;
3353
3354 return 0;
3355
3356nla_put_failure:
3357 return -EMSGSIZE;
3358}
3359
5be083ce 3360/* add multipath next hop */
beb1afac
DA		 3361static int rt6_add_nexthop(struct sk_buff *skb, struct rt6_info *rt)
3362{
3363 struct rtnexthop *rtnh;
3364 unsigned int flags = 0;
3365
3366 rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
3367 if (!rtnh)
3368 goto nla_put_failure;
3369
3370 rtnh->rtnh_hops = 0;
3371 rtnh->rtnh_ifindex = rt->dst.dev ? rt->dst.dev->ifindex : 0;
3372
5be083ce 3373 if (rt6_nexthop_info(skb, rt, &flags, true) < 0)
beb1afac
DA		 3374 goto nla_put_failure;
3375
3376 rtnh->rtnh_flags = flags;
3377
3378 /* length of rtnetlink header + attributes */
3379 rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
3380
3381 return 0;
3382
3383nla_put_failure:
3384 return -EMSGSIZE;
339bf98f
TG		 3385}
3386
191cd582
BH		 3387static int rt6_fill_node(struct net *net,
3388 struct sk_buff *skb, struct rt6_info *rt,
0d51aa80 3389 struct in6_addr *dst, struct in6_addr *src,
15e47304 3390 int iif, int type, u32 portid, u32 seq,
f8cfe2ce 3391 unsigned int flags)
1da177e4 3392{
4b32b5ad 3393 u32 metrics[RTAX_MAX];
1da177e4 3394 struct rtmsg *rtm;
2d7202bf 3395 struct nlmsghdr *nlh;
e3703b3d 3396 long expires;
9e762a4a 3397 u32 table;
1da177e4 3398
15e47304 3399 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
38308473 3400 if (!nlh)
26932566 3401 return -EMSGSIZE;
2d7202bf
TG		 3402
3403 rtm = nlmsg_data(nlh);
1da177e4
LT		 3404 rtm->rtm_family = AF_INET6;
3405 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3406 rtm->rtm_src_len = rt->rt6i_src.plen;
3407 rtm->rtm_tos = 0;
c71099ac 3408 if (rt->rt6i_table)
9e762a4a 3409 table = rt->rt6i_table->tb6_id;
c71099ac 3410 else
9e762a4a
PM		 3411 table = RT6_TABLE_UNSPEC;
3412 rtm->rtm_table = table;
c78679e8
DM		 3413 if (nla_put_u32(skb, RTA_TABLE, table))
3414 goto nla_put_failure;
ef2c7d7b
ND		 3415 if (rt->rt6i_flags & RTF_REJECT) {
3416 switch (rt->dst.error) {
3417 case -EINVAL:
3418 rtm->rtm_type = RTN_BLACKHOLE;
3419 break;
3420 case -EACCES:
3421 rtm->rtm_type = RTN_PROHIBIT;
3422 break;
b4949ab2
ND		 3423 case -EAGAIN:
3424 rtm->rtm_type = RTN_THROW;
3425 break;
ef2c7d7b
ND		 3426 default:
3427 rtm->rtm_type = RTN_UNREACHABLE;
3428 break;
3429 }
3430 }
38308473 3431 else if (rt->rt6i_flags & RTF_LOCAL)
ab79ad14 3432 rtm->rtm_type = RTN_LOCAL;
4ee39733
DA		 3433 else if (rt->rt6i_flags & RTF_ANYCAST)
3434 rtm->rtm_type = RTN_ANYCAST;
d1918542 3435 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
1da177e4
LT		 3436 rtm->rtm_type = RTN_LOCAL;
3437 else
3438 rtm->rtm_type = RTN_UNICAST;
3439 rtm->rtm_flags = 0;
3440 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3441 rtm->rtm_protocol = rt->rt6i_protocol;
38308473 3442 if (rt->rt6i_flags & RTF_DYNAMIC)
1da177e4 3443 rtm->rtm_protocol = RTPROT_REDIRECT;
f0396f60
DO		 3444 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3445 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3446 rtm->rtm_protocol = RTPROT_RA;
3447 else
3448 rtm->rtm_protocol = RTPROT_KERNEL;
3449 }
1da177e4 3450
38308473 3451 if (rt->rt6i_flags & RTF_CACHE)
1da177e4
LT		 3452 rtm->rtm_flags |= RTM_F_CLONED;
3453
3454 if (dst) {
930345ea 3455 if (nla_put_in6_addr(skb, RTA_DST, dst))
c78679e8 3456 goto nla_put_failure;
1ab1457c 3457 rtm->rtm_dst_len = 128;
1da177e4 3458 } else if (rtm->rtm_dst_len)
930345ea 3459 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
c78679e8 3460 goto nla_put_failure;
1da177e4
LT		 3461#ifdef CONFIG_IPV6_SUBTREES
3462 if (src) {
930345ea 3463 if (nla_put_in6_addr(skb, RTA_SRC, src))
c78679e8 3464 goto nla_put_failure;
1ab1457c 3465 rtm->rtm_src_len = 128;
c78679e8 3466 } else if (rtm->rtm_src_len &&
930345ea 3467 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
c78679e8 3468 goto nla_put_failure;
1da177e4 3469#endif
7bc570c8
YH		 3470 if (iif) {
3471#ifdef CONFIG_IPV6_MROUTE
3472 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
fd61c6ba
DA		 3473 int err = ip6mr_get_route(net, skb, rtm, portid);
3474
3475 if (err == 0)
3476 return 0;
3477 if (err < 0)
3478 goto nla_put_failure;
7bc570c8
YH		 3479 } else
3480#endif
c78679e8
DM		 3481 if (nla_put_u32(skb, RTA_IIF, iif))
3482 goto nla_put_failure;
7bc570c8 3483 } else if (dst) {
1da177e4 3484 struct in6_addr saddr_buf;
c78679e8 3485 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
930345ea 3486 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 3487 goto nla_put_failure;
1da177e4 3488 }
2d7202bf 3489
c3968a85
DW		 3490 if (rt->rt6i_prefsrc.plen) {
3491 struct in6_addr saddr_buf;
4e3fd7a0 3492 saddr_buf = rt->rt6i_prefsrc.addr;
930345ea 3493 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
c78679e8 3494 goto nla_put_failure;
c3968a85
DW		 3495 }
3496
4b32b5ad
MKL		 3497 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3498 if (rt->rt6i_pmtu)
3499 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3500 if (rtnetlink_put_metrics(skb, metrics) < 0)
2d7202bf
TG		 3501 goto nla_put_failure;
3502
c78679e8
DM		 3503 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3504 goto nla_put_failure;
8253947e 3505
beb1afac
DA		 3506 /* For multipath routes, walk the siblings list and add
3507 * each as a nexthop within RTA_MULTIPATH.
3508 */
3509 if (rt->rt6i_nsiblings) {
3510 struct rt6_info *sibling, *next_sibling;
3511 struct nlattr *mp;
3512
3513 mp = nla_nest_start(skb, RTA_MULTIPATH);
3514 if (!mp)
3515 goto nla_put_failure;
3516
3517 if (rt6_add_nexthop(skb, rt) < 0)
3518 goto nla_put_failure;
3519
3520 list_for_each_entry_safe(sibling, next_sibling,
3521 &rt->rt6i_siblings, rt6i_siblings) {
3522 if (rt6_add_nexthop(skb, sibling) < 0)
3523 goto nla_put_failure;
3524 }
3525
3526 nla_nest_end(skb, mp);
3527 } else {
5be083ce 3528 if (rt6_nexthop_info(skb, rt, &rtm->rtm_flags, false) < 0)
beb1afac
DA		 3529 goto nla_put_failure;
3530 }
3531
8253947e 3532 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
69cdf8f9 3533
87a50699 3534 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
e3703b3d 3535 goto nla_put_failure;
2d7202bf 3536
c78ba6d6
LR		 3537 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3538 goto nla_put_failure;
3539
19e42e45 3540
053c095a
JB		 3541 nlmsg_end(skb, nlh);
3542 return 0;
2d7202bf
TG		 3543
3544nla_put_failure:
26932566
PM		 3545 nlmsg_cancel(skb, nlh);
3546 return -EMSGSIZE;
1da177e4
LT		 3547}
3548
1b43af54 3549int rt6_dump_route(struct rt6_info *rt, void *p_arg)
1da177e4
LT		 3550{
3551 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
1f17e2f2
DA		 3552 struct net *net = arg->net;
3553
3554 if (rt == net->ipv6.ip6_null_entry)
3555 return 0;
1da177e4 3556
2d7202bf
TG		 3557 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3558 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
f8cfe2ce
DA		 3559
3560 /* user wants prefix routes only */
3561 if (rtm->rtm_flags & RTM_F_PREFIX &&
3562 !(rt->rt6i_flags & RTF_PREFIX_RT)) {
3563 /* success since this is not a prefix route */
3564 return 1;
3565 }
3566 }
1da177e4 3567
1f17e2f2 3568 return rt6_fill_node(net,
191cd582 3569 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
15e47304 3570 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
f8cfe2ce 3571 NLM_F_MULTI);
1da177e4
LT		 3572}
3573
c21ef3e3
DA		 3574static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3575 struct netlink_ext_ack *extack)
1da177e4 3576{
3b1e0a65 3577 struct net *net = sock_net(in_skb->sk);
ab364a6f
TG		 3578 struct nlattr *tb[RTA_MAX+1];
3579 struct rt6_info *rt;
1da177e4 3580 struct sk_buff *skb;
ab364a6f 3581 struct rtmsg *rtm;
4c9483b2 3582 struct flowi6 fl6;
72331bc0 3583 int err, iif = 0, oif = 0;
1da177e4 3584
fceb6435 3585 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
c21ef3e3 3586 extack);
ab364a6f
TG		 3587 if (err < 0)
3588 goto errout;
1da177e4 3589
ab364a6f 3590 err = -EINVAL;
4c9483b2 3591 memset(&fl6, 0, sizeof(fl6));
38b7097b
HFS		 3592 rtm = nlmsg_data(nlh);
3593 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
1da177e4 3594
ab364a6f
TG		 3595 if (tb[RTA_SRC]) {
3596 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3597 goto errout;
3598
4e3fd7a0 3599 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
ab364a6f
TG		 3600 }
3601
3602 if (tb[RTA_DST]) {
3603 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3604 goto errout;
3605
4e3fd7a0 3606 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
ab364a6f
TG		 3607 }
3608
3609 if (tb[RTA_IIF])
3610 iif = nla_get_u32(tb[RTA_IIF]);
3611
3612 if (tb[RTA_OIF])
72331bc0 3613 oif = nla_get_u32(tb[RTA_OIF]);
1da177e4 3614
2e47b291
LC		 3615 if (tb[RTA_MARK])
3616 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3617
622ec2c9
LC		 3618 if (tb[RTA_UID])
3619 fl6.flowi6_uid = make_kuid(current_user_ns(),
3620 nla_get_u32(tb[RTA_UID]));
3621 else
3622 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
3623
1da177e4
LT		 3624 if (iif) {
3625 struct net_device *dev;
72331bc0
SL		 3626 int flags = 0;
3627
5578689a 3628 dev = __dev_get_by_index(net, iif);
1da177e4
LT		 3629 if (!dev) {
3630 err = -ENODEV;
ab364a6f 3631 goto errout;
1da177e4 3632 }
72331bc0
SL		 3633
3634 fl6.flowi6_iif = iif;
3635
3636 if (!ipv6_addr_any(&fl6.saddr))
3637 flags |= RT6_LOOKUP_F_HAS_SADDR;
3638
3639 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3640 flags);
3641 } else {
3642 fl6.flowi6_oif = oif;
3643
3644 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
1da177e4
LT		 3645 }
3646
9d6acb3b
WC		 3647 if (rt == net->ipv6.ip6_null_entry) {
3648 err = rt->dst.error;
3649 ip6_rt_put(rt);
3650 goto errout;
3651 }
3652
ab364a6f 3653 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
38308473 3654 if (!skb) {
94e187c0 3655 ip6_rt_put(rt);
ab364a6f
TG		 3656 err = -ENOBUFS;
3657 goto errout;
3658 }
1da177e4 3659
d8d1f30b 3660 skb_dst_set(skb, &rt->dst);
1da177e4 3661
4c9483b2 3662 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
15e47304 3663 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
f8cfe2ce 3664 nlh->nlmsg_seq, 0);
1da177e4 3665 if (err < 0) {
ab364a6f
TG		 3666 kfree_skb(skb);
3667 goto errout;
1da177e4
LT		 3668 }
3669
15e47304 3670 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
ab364a6f 3671errout:
1da177e4 3672 return err;
1da177e4
LT		 3673}
3674
37a1d361
RP		 3675void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3676 unsigned int nlm_flags)
1da177e4
LT		 3677{
3678 struct sk_buff *skb;
5578689a 3679 struct net *net = info->nl_net;
528c4ceb
DL		 3680 u32 seq;
3681 int err;
3682
3683 err = -ENOBUFS;
38308473 3684 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
86872cb5 3685
19e42e45 3686 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
38308473 3687 if (!skb)
21713ebc
TG		 3688 goto errout;
3689
191cd582 3690 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
f8cfe2ce 3691 event, info->portid, seq, nlm_flags);
26932566
PM		 3692 if (err < 0) {
3693 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3694 WARN_ON(err == -EMSGSIZE);
3695 kfree_skb(skb);
3696 goto errout;
3697 }
15e47304 3698 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
1ce85fe4
PNA		 3699 info->nlh, gfp_any());
3700 return;
21713ebc
TG		 3701errout:
3702 if (err < 0)
5578689a 3703 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
1da177e4
LT		 3704}
3705
8ed67789 3706static int ip6_route_dev_notify(struct notifier_block *this,
351638e7 3707 unsigned long event, void *ptr)
8ed67789 3708{
351638e7 3709 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
c346dca1 3710 struct net *net = dev_net(dev);
8ed67789
DL		 3711
3712 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
d8d1f30b 3713 net->ipv6.ip6_null_entry->dst.dev = dev;
8ed67789
DL		 3714 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3715#ifdef CONFIG_IPV6_MULTIPLE_TABLES
d8d1f30b 3716 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
8ed67789 3717 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
d8d1f30b 3718 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
8ed67789
DL		 3719 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3720#endif
3721 }
3722
3723 return NOTIFY_OK;
3724}
3725
1da177e4
LT		 3726/*
3727 * /proc
3728 */
3729
3730#ifdef CONFIG_PROC_FS
3731
33120b30
AD		 3732static const struct file_operations ipv6_route_proc_fops = {
3733 .owner = THIS_MODULE,
3734 .open = ipv6_route_open,
3735 .read = seq_read,
3736 .llseek = seq_lseek,
8d2ca1d7 3737 .release = seq_release_net,
33120b30
AD		 3738};
3739
1da177e4
LT		 3740static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3741{
69ddb805 3742 struct net *net = (struct net *)seq->private;
1da177e4 3743 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
69ddb805
DL		 3744 net->ipv6.rt6_stats->fib_nodes,
3745 net->ipv6.rt6_stats->fib_route_nodes,
3746 net->ipv6.rt6_stats->fib_rt_alloc,
3747 net->ipv6.rt6_stats->fib_rt_entries,
3748 net->ipv6.rt6_stats->fib_rt_cache,
fc66f95c 3749 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
69ddb805 3750 net->ipv6.rt6_stats->fib_discarded_routes);
1da177e4
LT		 3751
3752 return 0;
3753}
3754
3755static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3756{
de05c557 3757 return single_open_net(inode, file, rt6_stats_seq_show);
69ddb805
DL		 3758}
3759
9a32144e 3760static const struct file_operations rt6_stats_seq_fops = {
1da177e4
LT		 3761 .owner = THIS_MODULE,
3762 .open = rt6_stats_seq_open,
3763 .read = seq_read,
3764 .llseek = seq_lseek,
b6fcbdb4 3765 .release = single_release_net,
1da177e4
LT		 3766};
3767#endif /* CONFIG_PROC_FS */
3768
3769#ifdef CONFIG_SYSCTL
3770
1da177e4 3771static
fe2c6338 3772int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
1da177e4
LT		 3773 void __user *buffer, size_t *lenp, loff_t *ppos)
3774{
c486da34
LAG		 3775 struct net *net;
3776 int delay;
3777 if (!write)
1da177e4 3778 return -EINVAL;
c486da34
LAG		 3779
3780 net = (struct net *)ctl->extra1;
3781 delay = net->ipv6.sysctl.flush_delay;
3782 proc_dointvec(ctl, write, buffer, lenp, ppos);
2ac3ac8f 3783 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
c486da34 3784 return 0;
1da177e4
LT		 3785}
3786
fe2c6338 3787struct ctl_table ipv6_route_table_template[] = {
1ab1457c 3788 {
1da177e4 3789 .procname = "flush",
4990509f 3790 .data = &init_net.ipv6.sysctl.flush_delay,
1da177e4 3791 .maxlen = sizeof(int),
89c8b3a1 3792 .mode = 0200,
6d9f239a 3793 .proc_handler = ipv6_sysctl_rtcache_flush
1da177e4
LT		 3794 },
3795 {
1da177e4 3796 .procname = "gc_thresh",
9a7ec3a9 3797 .data = &ip6_dst_ops_template.gc_thresh,
1da177e4
LT		 3798 .maxlen = sizeof(int),
3799 .mode = 0644,
6d9f239a 3800 .proc_handler = proc_dointvec,
1da177e4
LT		 3801 },
3802 {
1da177e4 3803 .procname = "max_size",
4990509f 3804 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
1da177e4
LT		 3805 .maxlen = sizeof(int),
3806 .mode = 0644,
6d9f239a 3807 .proc_handler = proc_dointvec,
1da177e4
LT		 3808 },
3809 {
1da177e4 3810 .procname = "gc_min_interval",
4990509f 3811 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 3812 .maxlen = sizeof(int),
3813 .mode = 0644,
6d9f239a 3814 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 3815 },
3816 {
1da177e4 3817 .procname = "gc_timeout",
4990509f 3818 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
1da177e4
LT		 3819 .maxlen = sizeof(int),
3820 .mode = 0644,
6d9f239a 3821 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 3822 },
3823 {
1da177e4 3824 .procname = "gc_interval",
4990509f 3825 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
1da177e4
LT		 3826 .maxlen = sizeof(int),
3827 .mode = 0644,
6d9f239a 3828 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 3829 },
3830 {
1da177e4 3831 .procname = "gc_elasticity",
4990509f 3832 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
1da177e4
LT		 3833 .maxlen = sizeof(int),
3834 .mode = 0644,
f3d3f616 3835 .proc_handler = proc_dointvec,
1da177e4
LT		 3836 },
3837 {
1da177e4 3838 .procname = "mtu_expires",
4990509f 3839 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
1da177e4
LT		 3840 .maxlen = sizeof(int),
3841 .mode = 0644,
6d9f239a 3842 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 3843 },
3844 {
1da177e4 3845 .procname = "min_adv_mss",
4990509f 3846 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
1da177e4
LT		 3847 .maxlen = sizeof(int),
3848 .mode = 0644,
f3d3f616 3849 .proc_handler = proc_dointvec,
1da177e4
LT		 3850 },
3851 {
1da177e4 3852 .procname = "gc_min_interval_ms",
4990509f 3853 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
1da177e4
LT		 3854 .maxlen = sizeof(int),
3855 .mode = 0644,
6d9f239a 3856 .proc_handler = proc_dointvec_ms_jiffies,
1da177e4 3857 },
f8572d8f 3858 { }
1da177e4
LT		 3859};
3860
2c8c1e72 3861struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
760f2d01
DL		 3862{
3863 struct ctl_table *table;
3864
3865 table = kmemdup(ipv6_route_table_template,
3866 sizeof(ipv6_route_table_template),
3867 GFP_KERNEL);
5ee09105
YH		 3868
3869 if (table) {
3870 table[0].data = &net->ipv6.sysctl.flush_delay;
c486da34 3871 table[0].extra1 = net;
86393e52 3872 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
5ee09105
YH		 3873 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3874 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3875 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3876 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3877 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3878 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3879 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
9c69fabe 3880 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
464dc801
EB		 3881
3882 /* Don't export sysctls to unprivileged users */
3883 if (net->user_ns != &init_user_ns)
3884 table[0].procname = NULL;
5ee09105
YH		 3885 }
3886
760f2d01
DL		 3887 return table;
3888}
1da177e4
LT		 3889#endif
3890
2c8c1e72 3891static int __net_init ip6_route_net_init(struct net *net)
cdb18761 3892{
633d424b 3893 int ret = -ENOMEM;
8ed67789 3894
86393e52
AD		 3895 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3896 sizeof(net->ipv6.ip6_dst_ops));
f2fc6a54 3897
fc66f95c
ED		 3898 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3899 goto out_ip6_dst_ops;
3900
8ed67789
DL		 3901 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3902 sizeof(*net->ipv6.ip6_null_entry),
3903 GFP_KERNEL);
3904 if (!net->ipv6.ip6_null_entry)
fc66f95c 3905 goto out_ip6_dst_entries;
d8d1f30b 3906 net->ipv6.ip6_null_entry->dst.path =
8ed67789 3907 (struct dst_entry *)net->ipv6.ip6_null_entry;
d8d1f30b 3908 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 3909 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3910 ip6_template_metrics, true);
8ed67789
DL		 3911
3912#ifdef CONFIG_IPV6_MULTIPLE_TABLES
3913 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3914 sizeof(*net->ipv6.ip6_prohibit_entry),
3915 GFP_KERNEL);
68fffc67
PZ		 3916 if (!net->ipv6.ip6_prohibit_entry)
3917 goto out_ip6_null_entry;
d8d1f30b 3918 net->ipv6.ip6_prohibit_entry->dst.path =
8ed67789 3919 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
d8d1f30b 3920 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 3921 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3922 ip6_template_metrics, true);
8ed67789
DL		 3923
3924 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3925 sizeof(*net->ipv6.ip6_blk_hole_entry),
3926 GFP_KERNEL);
68fffc67
PZ		 3927 if (!net->ipv6.ip6_blk_hole_entry)
3928 goto out_ip6_prohibit_entry;
d8d1f30b 3929 net->ipv6.ip6_blk_hole_entry->dst.path =
8ed67789 3930 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
d8d1f30b 3931 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
62fa8a84
DM		 3932 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3933 ip6_template_metrics, true);
8ed67789
DL		 3934#endif
3935
b339a47c
PZ		 3936 net->ipv6.sysctl.flush_delay = 0;
3937 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3938 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3939 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3940 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3941 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3942 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3943 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3944
6891a346
BT		 3945 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3946
8ed67789
DL		 3947 ret = 0;
3948out:
3949 return ret;
f2fc6a54 3950
68fffc67
PZ		 3951#ifdef CONFIG_IPV6_MULTIPLE_TABLES
3952out_ip6_prohibit_entry:
3953 kfree(net->ipv6.ip6_prohibit_entry);
3954out_ip6_null_entry:
3955 kfree(net->ipv6.ip6_null_entry);
3956#endif
fc66f95c
ED		 3957out_ip6_dst_entries:
3958 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
f2fc6a54 3959out_ip6_dst_ops:
f2fc6a54 3960 goto out;
cdb18761
DL		 3961}
3962
2c8c1e72 3963static void __net_exit ip6_route_net_exit(struct net *net)
cdb18761 3964{
8ed67789
DL		 3965 kfree(net->ipv6.ip6_null_entry);
3966#ifdef CONFIG_IPV6_MULTIPLE_TABLES
3967 kfree(net->ipv6.ip6_prohibit_entry);
3968 kfree(net->ipv6.ip6_blk_hole_entry);
3969#endif
41bb78b4 3970 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
cdb18761
DL		 3971}
3972
d189634e
TG		 3973static int __net_init ip6_route_net_init_late(struct net *net)
3974{
3975#ifdef CONFIG_PROC_FS
d4beaa66
G		 3976 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3977 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
d189634e
TG		 3978#endif
3979 return 0;
3980}
3981
3982static void __net_exit ip6_route_net_exit_late(struct net *net)
3983{
3984#ifdef CONFIG_PROC_FS
ece31ffd
G		 3985 remove_proc_entry("ipv6_route", net->proc_net);
3986 remove_proc_entry("rt6_stats", net->proc_net);
d189634e
TG		 3987#endif
3988}
3989
cdb18761
DL		 3990static struct pernet_operations ip6_route_net_ops = {
3991 .init = ip6_route_net_init,
3992 .exit = ip6_route_net_exit,
3993};
3994
c3426b47
DM		 3995static int __net_init ipv6_inetpeer_init(struct net *net)
3996{
3997 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3998
3999 if (!bp)
4000 return -ENOMEM;
4001 inet_peer_base_init(bp);
4002 net->ipv6.peers = bp;
4003 return 0;
4004}
4005
4006static void __net_exit ipv6_inetpeer_exit(struct net *net)
4007{
4008 struct inet_peer_base *bp = net->ipv6.peers;
4009
4010 net->ipv6.peers = NULL;
56a6b248 4011 inetpeer_invalidate_tree(bp);
c3426b47
DM		 4012 kfree(bp);
4013}
4014
2b823f72 4015static struct pernet_operations ipv6_inetpeer_ops = {
c3426b47
DM		 4016 .init = ipv6_inetpeer_init,
4017 .exit = ipv6_inetpeer_exit,
4018};
4019
d189634e
TG		 4020static struct pernet_operations ip6_route_net_late_ops = {
4021 .init = ip6_route_net_init_late,
4022 .exit = ip6_route_net_exit_late,
4023};
4024
8ed67789
DL		 4025static struct notifier_block ip6_route_dev_notifier = {
4026 .notifier_call = ip6_route_dev_notify,
4027 .priority = 0,
4028};
4029
2f460933
WC		 4030void __init ip6_route_init_special_entries(void)
4031{
4032 /* Registering of the loopback is done before this portion of code,
4033 * the loopback reference in rt6_info will not be taken, do it
4034 * manually for init_net */
4035 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
4036 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4037 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4038 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
4039 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4040 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
4041 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4042 #endif
4043}
4044
433d49c3 4045int __init ip6_route_init(void)
1da177e4 4046{
433d49c3 4047 int ret;
8d0b94af 4048 int cpu;
433d49c3 4049
9a7ec3a9
DL		 4050 ret = -ENOMEM;
4051 ip6_dst_ops_template.kmem_cachep =
e5d679f3 4052 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
f845ab6b 4053 SLAB_HWCACHE_ALIGN, NULL);
9a7ec3a9 4054 if (!ip6_dst_ops_template.kmem_cachep)
c19a28e1 4055 goto out;
14e50e57 4056
fc66f95c 4057 ret = dst_entries_init(&ip6_dst_blackhole_ops);
8ed67789 4058 if (ret)
bdb3289f 4059 goto out_kmem_cache;
bdb3289f 4060
c3426b47
DM		 4061 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
4062 if (ret)
e8803b6c 4063 goto out_dst_entries;
2a0c451a 4064
7e52b33b
DM		 4065 ret = register_pernet_subsys(&ip6_route_net_ops);
4066 if (ret)
4067 goto out_register_inetpeer;
c3426b47 4068
5dc121e9
AE		 4069 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
4070
e8803b6c 4071 ret = fib6_init();
433d49c3 4072 if (ret)
8ed67789 4073 goto out_register_subsys;
433d49c3 4074
433d49c3
DL		 4075 ret = xfrm6_init();
4076 if (ret)
e8803b6c 4077 goto out_fib6_init;
c35b7e72 4078
433d49c3
DL		 4079 ret = fib6_rules_init();
4080 if (ret)
4081 goto xfrm6_init;
7e5449c2 4082
d189634e
TG		 4083 ret = register_pernet_subsys(&ip6_route_net_late_ops);
4084 if (ret)
4085 goto fib6_rules_init;
4086
433d49c3 4087 ret = -ENOBUFS;
c7ac8679
GR		 4088 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
4089 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
4090 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
d189634e 4091 goto out_register_late_subsys;
c127ea2c 4092
8ed67789 4093 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
cdb18761 4094 if (ret)
d189634e 4095 goto out_register_late_subsys;
8ed67789 4096
8d0b94af
MKL		 4097 for_each_possible_cpu(cpu) {
4098 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
4099
4100 INIT_LIST_HEAD(&ul->head);
4101 spin_lock_init(&ul->lock);
4102 }
4103
433d49c3
DL		 4104out:
4105 return ret;
4106
d189634e
TG		 4107out_register_late_subsys:
4108 unregister_pernet_subsys(&ip6_route_net_late_ops);
433d49c3 4109fib6_rules_init:
433d49c3
DL		 4110 fib6_rules_cleanup();
4111xfrm6_init:
433d49c3 4112 xfrm6_fini();
2a0c451a
TG		 4113out_fib6_init:
4114 fib6_gc_cleanup();
8ed67789
DL		 4115out_register_subsys:
4116 unregister_pernet_subsys(&ip6_route_net_ops);
7e52b33b
DM		 4117out_register_inetpeer:
4118 unregister_pernet_subsys(&ipv6_inetpeer_ops);
fc66f95c
ED		 4119out_dst_entries:
4120 dst_entries_destroy(&ip6_dst_blackhole_ops);
433d49c3 4121out_kmem_cache:
f2fc6a54 4122 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
433d49c3 4123 goto out;
1da177e4
LT		 4124}
4125
4126void ip6_route_cleanup(void)
4127{
8ed67789 4128 unregister_netdevice_notifier(&ip6_route_dev_notifier);
d189634e 4129 unregister_pernet_subsys(&ip6_route_net_late_ops);
101367c2 4130 fib6_rules_cleanup();
1da177e4 4131 xfrm6_fini();
1da177e4 4132 fib6_gc_cleanup();
c3426b47 4133 unregister_pernet_subsys(&ipv6_inetpeer_ops);
8ed67789 4134 unregister_pernet_subsys(&ip6_route_net_ops);
41bb78b4 4135 dst_entries_destroy(&ip6_dst_blackhole_ops);
f2fc6a54 4136 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
1da177e4 4137}
Linux 6.x block layer and io_uring tree(s)