2 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
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.
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.
24 * Fixed routing subtrees.
27 #define pr_fmt(fmt) "IPv6: " fmt
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/export.h>
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>
40 #include <linux/mroute6.h>
41 #include <linux/init.h>
42 #include <linux/if_arp.h>
43 #include <linux/proc_fs.h>
44 #include <linux/seq_file.h>
45 #include <linux/nsproxy.h>
46 #include <linux/slab.h>
47 #include <net/net_namespace.h>
50 #include <net/ip6_fib.h>
51 #include <net/ip6_route.h>
52 #include <net/ndisc.h>
53 #include <net/addrconf.h>
55 #include <linux/rtnetlink.h>
58 #include <net/netevent.h>
59 #include <net/netlink.h>
60 #include <net/nexthop.h>
62 #include <asm/uaccess.h>
65 #include <linux/sysctl.h>
68 static struct rt6_info
*ip6_rt_copy(struct rt6_info
*ort
,
69 const struct in6_addr
*dest
);
70 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
);
71 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
);
72 static unsigned int ip6_mtu(const struct dst_entry
*dst
);
73 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*);
74 static void ip6_dst_destroy(struct dst_entry
*);
75 static void ip6_dst_ifdown(struct dst_entry
*,
76 struct net_device
*dev
, int how
);
77 static int ip6_dst_gc(struct dst_ops
*ops
);
79 static int ip6_pkt_discard(struct sk_buff
*skb
);
80 static int ip6_pkt_discard_out(struct sk_buff
*skb
);
81 static void ip6_link_failure(struct sk_buff
*skb
);
82 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
83 struct sk_buff
*skb
, u32 mtu
);
84 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
,
86 static int rt6_score_route(struct rt6_info
*rt
, int oif
, int strict
);
88 #ifdef CONFIG_IPV6_ROUTE_INFO
89 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
90 const struct in6_addr
*prefix
, int prefixlen
,
91 const struct in6_addr
*gwaddr
, int ifindex
,
93 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
94 const struct in6_addr
*prefix
, int prefixlen
,
95 const struct in6_addr
*gwaddr
, int ifindex
);
98 static u32
*ipv6_cow_metrics(struct dst_entry
*dst
, unsigned long old
)
100 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
101 struct inet_peer
*peer
;
104 if (!(rt
->dst
.flags
& DST_HOST
))
107 peer
= rt6_get_peer_create(rt
);
109 u32
*old_p
= __DST_METRICS_PTR(old
);
110 unsigned long prev
, new;
113 if (inet_metrics_new(peer
))
114 memcpy(p
, old_p
, sizeof(u32
) * RTAX_MAX
);
116 new = (unsigned long) p
;
117 prev
= cmpxchg(&dst
->_metrics
, old
, new);
120 p
= __DST_METRICS_PTR(prev
);
121 if (prev
& DST_METRICS_READ_ONLY
)
128 static inline const void *choose_neigh_daddr(struct rt6_info
*rt
,
132 struct in6_addr
*p
= &rt
->rt6i_gateway
;
134 if (!ipv6_addr_any(p
))
135 return (const void *) p
;
137 return &ipv6_hdr(skb
)->daddr
;
141 static struct neighbour
*ip6_neigh_lookup(const struct dst_entry
*dst
,
145 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
148 daddr
= choose_neigh_daddr(rt
, skb
, daddr
);
149 n
= __ipv6_neigh_lookup(dst
->dev
, daddr
);
152 return neigh_create(&nd_tbl
, daddr
, dst
->dev
);
155 static struct dst_ops ip6_dst_ops_template
= {
157 .protocol
= cpu_to_be16(ETH_P_IPV6
),
160 .check
= ip6_dst_check
,
161 .default_advmss
= ip6_default_advmss
,
163 .cow_metrics
= ipv6_cow_metrics
,
164 .destroy
= ip6_dst_destroy
,
165 .ifdown
= ip6_dst_ifdown
,
166 .negative_advice
= ip6_negative_advice
,
167 .link_failure
= ip6_link_failure
,
168 .update_pmtu
= ip6_rt_update_pmtu
,
169 .redirect
= rt6_do_redirect
,
170 .local_out
= __ip6_local_out
,
171 .neigh_lookup
= ip6_neigh_lookup
,
174 static unsigned int ip6_blackhole_mtu(const struct dst_entry
*dst
)
176 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
178 return mtu
? : dst
->dev
->mtu
;
181 static void ip6_rt_blackhole_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
182 struct sk_buff
*skb
, u32 mtu
)
186 static void ip6_rt_blackhole_redirect(struct dst_entry
*dst
, struct sock
*sk
,
191 static u32
*ip6_rt_blackhole_cow_metrics(struct dst_entry
*dst
,
197 static struct dst_ops ip6_dst_blackhole_ops
= {
199 .protocol
= cpu_to_be16(ETH_P_IPV6
),
200 .destroy
= ip6_dst_destroy
,
201 .check
= ip6_dst_check
,
202 .mtu
= ip6_blackhole_mtu
,
203 .default_advmss
= ip6_default_advmss
,
204 .update_pmtu
= ip6_rt_blackhole_update_pmtu
,
205 .redirect
= ip6_rt_blackhole_redirect
,
206 .cow_metrics
= ip6_rt_blackhole_cow_metrics
,
207 .neigh_lookup
= ip6_neigh_lookup
,
210 static const u32 ip6_template_metrics
[RTAX_MAX
] = {
211 [RTAX_HOPLIMIT
- 1] = 0,
214 static const struct rt6_info ip6_null_entry_template
= {
216 .__refcnt
= ATOMIC_INIT(1),
218 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
219 .error
= -ENETUNREACH
,
220 .input
= ip6_pkt_discard
,
221 .output
= ip6_pkt_discard_out
,
223 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
224 .rt6i_protocol
= RTPROT_KERNEL
,
225 .rt6i_metric
= ~(u32
) 0,
226 .rt6i_ref
= ATOMIC_INIT(1),
229 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
231 static int ip6_pkt_prohibit(struct sk_buff
*skb
);
232 static int ip6_pkt_prohibit_out(struct sk_buff
*skb
);
234 static const struct rt6_info ip6_prohibit_entry_template
= {
236 .__refcnt
= ATOMIC_INIT(1),
238 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
240 .input
= ip6_pkt_prohibit
,
241 .output
= ip6_pkt_prohibit_out
,
243 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
244 .rt6i_protocol
= RTPROT_KERNEL
,
245 .rt6i_metric
= ~(u32
) 0,
246 .rt6i_ref
= ATOMIC_INIT(1),
249 static const struct rt6_info ip6_blk_hole_entry_template
= {
251 .__refcnt
= ATOMIC_INIT(1),
253 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
255 .input
= dst_discard
,
256 .output
= dst_discard
,
258 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
259 .rt6i_protocol
= RTPROT_KERNEL
,
260 .rt6i_metric
= ~(u32
) 0,
261 .rt6i_ref
= ATOMIC_INIT(1),
266 /* allocate dst with ip6_dst_ops */
267 static inline struct rt6_info
*ip6_dst_alloc(struct net
*net
,
268 struct net_device
*dev
,
270 struct fib6_table
*table
)
272 struct rt6_info
*rt
= dst_alloc(&net
->ipv6
.ip6_dst_ops
, dev
,
273 0, DST_OBSOLETE_FORCE_CHK
, flags
);
276 struct dst_entry
*dst
= &rt
->dst
;
278 memset(dst
+ 1, 0, sizeof(*rt
) - sizeof(*dst
));
279 rt6_init_peer(rt
, table
? &table
->tb6_peers
: net
->ipv6
.peers
);
280 rt
->rt6i_genid
= rt_genid(net
);
281 INIT_LIST_HEAD(&rt
->rt6i_siblings
);
282 rt
->rt6i_nsiblings
= 0;
287 static void ip6_dst_destroy(struct dst_entry
*dst
)
289 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
290 struct inet6_dev
*idev
= rt
->rt6i_idev
;
291 struct dst_entry
*from
= dst
->from
;
293 if (!(rt
->dst
.flags
& DST_HOST
))
294 dst_destroy_metrics_generic(dst
);
297 rt
->rt6i_idev
= NULL
;
304 if (rt6_has_peer(rt
)) {
305 struct inet_peer
*peer
= rt6_peer_ptr(rt
);
310 void rt6_bind_peer(struct rt6_info
*rt
, int create
)
312 struct inet_peer_base
*base
;
313 struct inet_peer
*peer
;
315 base
= inetpeer_base_ptr(rt
->_rt6i_peer
);
319 peer
= inet_getpeer_v6(base
, &rt
->rt6i_dst
.addr
, create
);
321 if (!rt6_set_peer(rt
, peer
))
326 static void ip6_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
,
329 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
330 struct inet6_dev
*idev
= rt
->rt6i_idev
;
331 struct net_device
*loopback_dev
=
332 dev_net(dev
)->loopback_dev
;
334 if (dev
!= loopback_dev
) {
335 if (idev
&& idev
->dev
== dev
) {
336 struct inet6_dev
*loopback_idev
=
337 in6_dev_get(loopback_dev
);
339 rt
->rt6i_idev
= loopback_idev
;
346 static bool rt6_check_expired(const struct rt6_info
*rt
)
348 if (rt
->rt6i_flags
& RTF_EXPIRES
) {
349 if (time_after(jiffies
, rt
->dst
.expires
))
351 } else if (rt
->dst
.from
) {
352 return rt6_check_expired((struct rt6_info
*) rt
->dst
.from
);
357 static bool rt6_need_strict(const struct in6_addr
*daddr
)
359 return ipv6_addr_type(daddr
) &
360 (IPV6_ADDR_MULTICAST
| IPV6_ADDR_LINKLOCAL
| IPV6_ADDR_LOOPBACK
);
363 /* Multipath route selection:
364 * Hash based function using packet header and flowlabel.
365 * Adapted from fib_info_hashfn()
367 static int rt6_info_hash_nhsfn(unsigned int candidate_count
,
368 const struct flowi6
*fl6
)
370 unsigned int val
= fl6
->flowi6_proto
;
372 val
^= ipv6_addr_hash(&fl6
->daddr
);
373 val
^= ipv6_addr_hash(&fl6
->saddr
);
375 /* Work only if this not encapsulated */
376 switch (fl6
->flowi6_proto
) {
380 val
^= (__force u16
)fl6
->fl6_sport
;
381 val
^= (__force u16
)fl6
->fl6_dport
;
385 val
^= (__force u16
)fl6
->fl6_icmp_type
;
386 val
^= (__force u16
)fl6
->fl6_icmp_code
;
389 /* RFC6438 recommands to use flowlabel */
390 val
^= (__force u32
)fl6
->flowlabel
;
392 /* Perhaps, we need to tune, this function? */
393 val
= val
^ (val
>> 7) ^ (val
>> 12);
394 return val
% candidate_count
;
397 static struct rt6_info
*rt6_multipath_select(struct rt6_info
*match
,
398 struct flowi6
*fl6
, int oif
,
401 struct rt6_info
*sibling
, *next_sibling
;
404 route_choosen
= rt6_info_hash_nhsfn(match
->rt6i_nsiblings
+ 1, fl6
);
405 /* Don't change the route, if route_choosen == 0
406 * (siblings does not include ourself)
409 list_for_each_entry_safe(sibling
, next_sibling
,
410 &match
->rt6i_siblings
, rt6i_siblings
) {
412 if (route_choosen
== 0) {
413 if (rt6_score_route(sibling
, oif
, strict
) < 0)
423 * Route lookup. Any table->tb6_lock is implied.
426 static inline struct rt6_info
*rt6_device_match(struct net
*net
,
428 const struct in6_addr
*saddr
,
432 struct rt6_info
*local
= NULL
;
433 struct rt6_info
*sprt
;
435 if (!oif
&& ipv6_addr_any(saddr
))
438 for (sprt
= rt
; sprt
; sprt
= sprt
->dst
.rt6_next
) {
439 struct net_device
*dev
= sprt
->dst
.dev
;
442 if (dev
->ifindex
== oif
)
444 if (dev
->flags
& IFF_LOOPBACK
) {
445 if (!sprt
->rt6i_idev
||
446 sprt
->rt6i_idev
->dev
->ifindex
!= oif
) {
447 if (flags
& RT6_LOOKUP_F_IFACE
&& oif
)
449 if (local
&& (!oif
||
450 local
->rt6i_idev
->dev
->ifindex
== oif
))
456 if (ipv6_chk_addr(net
, saddr
, dev
,
457 flags
& RT6_LOOKUP_F_IFACE
))
466 if (flags
& RT6_LOOKUP_F_IFACE
)
467 return net
->ipv6
.ip6_null_entry
;
473 #ifdef CONFIG_IPV6_ROUTER_PREF
474 static void rt6_probe(struct rt6_info
*rt
)
476 struct neighbour
*neigh
;
478 * Okay, this does not seem to be appropriate
479 * for now, however, we need to check if it
480 * is really so; aka Router Reachability Probing.
482 * Router Reachability Probe MUST be rate-limited
483 * to no more than one per minute.
485 if (!rt
|| !(rt
->rt6i_flags
& RTF_GATEWAY
))
488 neigh
= __ipv6_neigh_lookup_noref(rt
->dst
.dev
, &rt
->rt6i_gateway
);
490 write_lock(&neigh
->lock
);
491 if (neigh
->nud_state
& NUD_VALID
)
496 time_after(jiffies
, neigh
->updated
+ rt
->rt6i_idev
->cnf
.rtr_probe_interval
)) {
497 struct in6_addr mcaddr
;
498 struct in6_addr
*target
;
501 neigh
->updated
= jiffies
;
502 write_unlock(&neigh
->lock
);
505 target
= (struct in6_addr
*)&rt
->rt6i_gateway
;
506 addrconf_addr_solict_mult(target
, &mcaddr
);
507 ndisc_send_ns(rt
->dst
.dev
, NULL
, target
, &mcaddr
, NULL
);
510 write_unlock(&neigh
->lock
);
512 rcu_read_unlock_bh();
515 static inline void rt6_probe(struct rt6_info
*rt
)
521 * Default Router Selection (RFC 2461 6.3.6)
523 static inline int rt6_check_dev(struct rt6_info
*rt
, int oif
)
525 struct net_device
*dev
= rt
->dst
.dev
;
526 if (!oif
|| dev
->ifindex
== oif
)
528 if ((dev
->flags
& IFF_LOOPBACK
) &&
529 rt
->rt6i_idev
&& rt
->rt6i_idev
->dev
->ifindex
== oif
)
534 static inline bool rt6_check_neigh(struct rt6_info
*rt
)
536 struct neighbour
*neigh
;
539 if (rt
->rt6i_flags
& RTF_NONEXTHOP
||
540 !(rt
->rt6i_flags
& RTF_GATEWAY
))
544 neigh
= __ipv6_neigh_lookup_noref(rt
->dst
.dev
, &rt
->rt6i_gateway
);
546 read_lock(&neigh
->lock
);
547 if (neigh
->nud_state
& NUD_VALID
)
549 #ifdef CONFIG_IPV6_ROUTER_PREF
550 else if (!(neigh
->nud_state
& NUD_FAILED
))
553 read_unlock(&neigh
->lock
);
554 } else if (IS_ENABLED(CONFIG_IPV6_ROUTER_PREF
)) {
557 rcu_read_unlock_bh();
562 static int rt6_score_route(struct rt6_info
*rt
, int oif
,
567 m
= rt6_check_dev(rt
, oif
);
568 if (!m
&& (strict
& RT6_LOOKUP_F_IFACE
))
570 #ifdef CONFIG_IPV6_ROUTER_PREF
571 m
|= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt
->rt6i_flags
)) << 2;
573 if (!rt6_check_neigh(rt
) && (strict
& RT6_LOOKUP_F_REACHABLE
))
578 static struct rt6_info
*find_match(struct rt6_info
*rt
, int oif
, int strict
,
579 int *mpri
, struct rt6_info
*match
)
583 if (rt6_check_expired(rt
))
586 m
= rt6_score_route(rt
, oif
, strict
);
591 if (strict
& RT6_LOOKUP_F_REACHABLE
)
595 } else if (strict
& RT6_LOOKUP_F_REACHABLE
) {
603 static struct rt6_info
*find_rr_leaf(struct fib6_node
*fn
,
604 struct rt6_info
*rr_head
,
605 u32 metric
, int oif
, int strict
)
607 struct rt6_info
*rt
, *match
;
611 for (rt
= rr_head
; rt
&& rt
->rt6i_metric
== metric
;
612 rt
= rt
->dst
.rt6_next
)
613 match
= find_match(rt
, oif
, strict
, &mpri
, match
);
614 for (rt
= fn
->leaf
; rt
&& rt
!= rr_head
&& rt
->rt6i_metric
== metric
;
615 rt
= rt
->dst
.rt6_next
)
616 match
= find_match(rt
, oif
, strict
, &mpri
, match
);
621 static struct rt6_info
*rt6_select(struct fib6_node
*fn
, int oif
, int strict
)
623 struct rt6_info
*match
, *rt0
;
628 fn
->rr_ptr
= rt0
= fn
->leaf
;
630 match
= find_rr_leaf(fn
, rt0
, rt0
->rt6i_metric
, oif
, strict
);
633 (strict
& RT6_LOOKUP_F_REACHABLE
)) {
634 struct rt6_info
*next
= rt0
->dst
.rt6_next
;
636 /* no entries matched; do round-robin */
637 if (!next
|| next
->rt6i_metric
!= rt0
->rt6i_metric
)
644 net
= dev_net(rt0
->dst
.dev
);
645 return match
? match
: net
->ipv6
.ip6_null_entry
;
648 #ifdef CONFIG_IPV6_ROUTE_INFO
649 int rt6_route_rcv(struct net_device
*dev
, u8
*opt
, int len
,
650 const struct in6_addr
*gwaddr
)
652 struct net
*net
= dev_net(dev
);
653 struct route_info
*rinfo
= (struct route_info
*) opt
;
654 struct in6_addr prefix_buf
, *prefix
;
656 unsigned long lifetime
;
659 if (len
< sizeof(struct route_info
)) {
663 /* Sanity check for prefix_len and length */
664 if (rinfo
->length
> 3) {
666 } else if (rinfo
->prefix_len
> 128) {
668 } else if (rinfo
->prefix_len
> 64) {
669 if (rinfo
->length
< 2) {
672 } else if (rinfo
->prefix_len
> 0) {
673 if (rinfo
->length
< 1) {
678 pref
= rinfo
->route_pref
;
679 if (pref
== ICMPV6_ROUTER_PREF_INVALID
)
682 lifetime
= addrconf_timeout_fixup(ntohl(rinfo
->lifetime
), HZ
);
684 if (rinfo
->length
== 3)
685 prefix
= (struct in6_addr
*)rinfo
->prefix
;
687 /* this function is safe */
688 ipv6_addr_prefix(&prefix_buf
,
689 (struct in6_addr
*)rinfo
->prefix
,
691 prefix
= &prefix_buf
;
694 rt
= rt6_get_route_info(net
, prefix
, rinfo
->prefix_len
, gwaddr
,
697 if (rt
&& !lifetime
) {
703 rt
= rt6_add_route_info(net
, prefix
, rinfo
->prefix_len
, gwaddr
, dev
->ifindex
,
706 rt
->rt6i_flags
= RTF_ROUTEINFO
|
707 (rt
->rt6i_flags
& ~RTF_PREF_MASK
) | RTF_PREF(pref
);
710 if (!addrconf_finite_timeout(lifetime
))
711 rt6_clean_expires(rt
);
713 rt6_set_expires(rt
, jiffies
+ HZ
* lifetime
);
721 #define BACKTRACK(__net, saddr) \
723 if (rt == __net->ipv6.ip6_null_entry) { \
724 struct fib6_node *pn; \
726 if (fn->fn_flags & RTN_TL_ROOT) \
729 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
730 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
733 if (fn->fn_flags & RTN_RTINFO) \
739 static struct rt6_info
*ip6_pol_route_lookup(struct net
*net
,
740 struct fib6_table
*table
,
741 struct flowi6
*fl6
, int flags
)
743 struct fib6_node
*fn
;
746 read_lock_bh(&table
->tb6_lock
);
747 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
750 rt
= rt6_device_match(net
, rt
, &fl6
->saddr
, fl6
->flowi6_oif
, flags
);
751 if (rt
->rt6i_nsiblings
&& fl6
->flowi6_oif
== 0)
752 rt
= rt6_multipath_select(rt
, fl6
, fl6
->flowi6_oif
, flags
);
753 BACKTRACK(net
, &fl6
->saddr
);
755 dst_use(&rt
->dst
, jiffies
);
756 read_unlock_bh(&table
->tb6_lock
);
761 struct dst_entry
* ip6_route_lookup(struct net
*net
, struct flowi6
*fl6
,
764 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_lookup
);
766 EXPORT_SYMBOL_GPL(ip6_route_lookup
);
768 struct rt6_info
*rt6_lookup(struct net
*net
, const struct in6_addr
*daddr
,
769 const struct in6_addr
*saddr
, int oif
, int strict
)
771 struct flowi6 fl6
= {
775 struct dst_entry
*dst
;
776 int flags
= strict
? RT6_LOOKUP_F_IFACE
: 0;
779 memcpy(&fl6
.saddr
, saddr
, sizeof(*saddr
));
780 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
783 dst
= fib6_rule_lookup(net
, &fl6
, flags
, ip6_pol_route_lookup
);
785 return (struct rt6_info
*) dst
;
792 EXPORT_SYMBOL(rt6_lookup
);
794 /* ip6_ins_rt is called with FREE table->tb6_lock.
795 It takes new route entry, the addition fails by any reason the
796 route is freed. In any case, if caller does not hold it, it may
800 static int __ip6_ins_rt(struct rt6_info
*rt
, struct nl_info
*info
)
803 struct fib6_table
*table
;
805 table
= rt
->rt6i_table
;
806 write_lock_bh(&table
->tb6_lock
);
807 err
= fib6_add(&table
->tb6_root
, rt
, info
);
808 write_unlock_bh(&table
->tb6_lock
);
813 int ip6_ins_rt(struct rt6_info
*rt
)
815 struct nl_info info
= {
816 .nl_net
= dev_net(rt
->dst
.dev
),
818 return __ip6_ins_rt(rt
, &info
);
821 static struct rt6_info
*rt6_alloc_cow(struct rt6_info
*ort
,
822 const struct in6_addr
*daddr
,
823 const struct in6_addr
*saddr
)
831 rt
= ip6_rt_copy(ort
, daddr
);
834 if (!(rt
->rt6i_flags
& RTF_GATEWAY
)) {
835 if (ort
->rt6i_dst
.plen
!= 128 &&
836 ipv6_addr_equal(&ort
->rt6i_dst
.addr
, daddr
))
837 rt
->rt6i_flags
|= RTF_ANYCAST
;
838 rt
->rt6i_gateway
= *daddr
;
841 rt
->rt6i_flags
|= RTF_CACHE
;
843 #ifdef CONFIG_IPV6_SUBTREES
844 if (rt
->rt6i_src
.plen
&& saddr
) {
845 rt
->rt6i_src
.addr
= *saddr
;
846 rt
->rt6i_src
.plen
= 128;
854 static struct rt6_info
*rt6_alloc_clone(struct rt6_info
*ort
,
855 const struct in6_addr
*daddr
)
857 struct rt6_info
*rt
= ip6_rt_copy(ort
, daddr
);
860 rt
->rt6i_flags
|= RTF_CACHE
;
864 static struct rt6_info
*ip6_pol_route(struct net
*net
, struct fib6_table
*table
, int oif
,
865 struct flowi6
*fl6
, int flags
)
867 struct fib6_node
*fn
;
868 struct rt6_info
*rt
, *nrt
;
872 int reachable
= net
->ipv6
.devconf_all
->forwarding
? 0 : RT6_LOOKUP_F_REACHABLE
;
874 strict
|= flags
& RT6_LOOKUP_F_IFACE
;
877 read_lock_bh(&table
->tb6_lock
);
880 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
883 rt
= rt6_select(fn
, oif
, strict
| reachable
);
884 if (rt
->rt6i_nsiblings
)
885 rt
= rt6_multipath_select(rt
, fl6
, oif
, strict
| reachable
);
886 BACKTRACK(net
, &fl6
->saddr
);
887 if (rt
== net
->ipv6
.ip6_null_entry
||
888 rt
->rt6i_flags
& RTF_CACHE
)
892 read_unlock_bh(&table
->tb6_lock
);
894 if (!(rt
->rt6i_flags
& (RTF_NONEXTHOP
| RTF_GATEWAY
)))
895 nrt
= rt6_alloc_cow(rt
, &fl6
->daddr
, &fl6
->saddr
);
896 else if (!(rt
->dst
.flags
& DST_HOST
))
897 nrt
= rt6_alloc_clone(rt
, &fl6
->daddr
);
902 rt
= nrt
? : net
->ipv6
.ip6_null_entry
;
906 err
= ip6_ins_rt(nrt
);
915 * Race condition! In the gap, when table->tb6_lock was
916 * released someone could insert this route. Relookup.
927 read_unlock_bh(&table
->tb6_lock
);
929 rt
->dst
.lastuse
= jiffies
;
935 static struct rt6_info
*ip6_pol_route_input(struct net
*net
, struct fib6_table
*table
,
936 struct flowi6
*fl6
, int flags
)
938 return ip6_pol_route(net
, table
, fl6
->flowi6_iif
, fl6
, flags
);
941 static struct dst_entry
*ip6_route_input_lookup(struct net
*net
,
942 struct net_device
*dev
,
943 struct flowi6
*fl6
, int flags
)
945 if (rt6_need_strict(&fl6
->daddr
) && dev
->type
!= ARPHRD_PIMREG
)
946 flags
|= RT6_LOOKUP_F_IFACE
;
948 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_input
);
951 void ip6_route_input(struct sk_buff
*skb
)
953 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
954 struct net
*net
= dev_net(skb
->dev
);
955 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
956 struct flowi6 fl6
= {
957 .flowi6_iif
= skb
->dev
->ifindex
,
960 .flowlabel
= ip6_flowinfo(iph
),
961 .flowi6_mark
= skb
->mark
,
962 .flowi6_proto
= iph
->nexthdr
,
965 skb_dst_set(skb
, ip6_route_input_lookup(net
, skb
->dev
, &fl6
, flags
));
968 static struct rt6_info
*ip6_pol_route_output(struct net
*net
, struct fib6_table
*table
,
969 struct flowi6
*fl6
, int flags
)
971 return ip6_pol_route(net
, table
, fl6
->flowi6_oif
, fl6
, flags
);
974 struct dst_entry
* ip6_route_output(struct net
*net
, const struct sock
*sk
,
979 fl6
->flowi6_iif
= LOOPBACK_IFINDEX
;
981 if ((sk
&& sk
->sk_bound_dev_if
) || rt6_need_strict(&fl6
->daddr
))
982 flags
|= RT6_LOOKUP_F_IFACE
;
984 if (!ipv6_addr_any(&fl6
->saddr
))
985 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
987 flags
|= rt6_srcprefs2flags(inet6_sk(sk
)->srcprefs
);
989 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_output
);
992 EXPORT_SYMBOL(ip6_route_output
);
994 struct dst_entry
*ip6_blackhole_route(struct net
*net
, struct dst_entry
*dst_orig
)
996 struct rt6_info
*rt
, *ort
= (struct rt6_info
*) dst_orig
;
997 struct dst_entry
*new = NULL
;
999 rt
= dst_alloc(&ip6_dst_blackhole_ops
, ort
->dst
.dev
, 1, DST_OBSOLETE_NONE
, 0);
1003 memset(new + 1, 0, sizeof(*rt
) - sizeof(*new));
1004 rt6_init_peer(rt
, net
->ipv6
.peers
);
1007 new->input
= dst_discard
;
1008 new->output
= dst_discard
;
1010 if (dst_metrics_read_only(&ort
->dst
))
1011 new->_metrics
= ort
->dst
._metrics
;
1013 dst_copy_metrics(new, &ort
->dst
);
1014 rt
->rt6i_idev
= ort
->rt6i_idev
;
1016 in6_dev_hold(rt
->rt6i_idev
);
1018 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
1019 rt
->rt6i_flags
= ort
->rt6i_flags
;
1020 rt
->rt6i_metric
= 0;
1022 memcpy(&rt
->rt6i_dst
, &ort
->rt6i_dst
, sizeof(struct rt6key
));
1023 #ifdef CONFIG_IPV6_SUBTREES
1024 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1030 dst_release(dst_orig
);
1031 return new ? new : ERR_PTR(-ENOMEM
);
1035 * Destination cache support functions
1038 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
)
1040 struct rt6_info
*rt
;
1042 rt
= (struct rt6_info
*) dst
;
1044 /* All IPV6 dsts are created with ->obsolete set to the value
1045 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1046 * into this function always.
1048 if (rt
->rt6i_genid
!= rt_genid(dev_net(rt
->dst
.dev
)))
1051 if (rt
->rt6i_node
&& (rt
->rt6i_node
->fn_sernum
== cookie
))
1057 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*dst
)
1059 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1062 if (rt
->rt6i_flags
& RTF_CACHE
) {
1063 if (rt6_check_expired(rt
)) {
1075 static void ip6_link_failure(struct sk_buff
*skb
)
1077 struct rt6_info
*rt
;
1079 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, ICMPV6_ADDR_UNREACH
, 0);
1081 rt
= (struct rt6_info
*) skb_dst(skb
);
1083 if (rt
->rt6i_flags
& RTF_CACHE
)
1084 rt6_update_expires(rt
, 0);
1085 else if (rt
->rt6i_node
&& (rt
->rt6i_flags
& RTF_DEFAULT
))
1086 rt
->rt6i_node
->fn_sernum
= -1;
1090 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
1091 struct sk_buff
*skb
, u32 mtu
)
1093 struct rt6_info
*rt6
= (struct rt6_info
*)dst
;
1096 if (mtu
< dst_mtu(dst
) && rt6
->rt6i_dst
.plen
== 128) {
1097 struct net
*net
= dev_net(dst
->dev
);
1099 rt6
->rt6i_flags
|= RTF_MODIFIED
;
1100 if (mtu
< IPV6_MIN_MTU
) {
1101 u32 features
= dst_metric(dst
, RTAX_FEATURES
);
1103 features
|= RTAX_FEATURE_ALLFRAG
;
1104 dst_metric_set(dst
, RTAX_FEATURES
, features
);
1106 dst_metric_set(dst
, RTAX_MTU
, mtu
);
1107 rt6_update_expires(rt6
, net
->ipv6
.sysctl
.ip6_rt_mtu_expires
);
1111 void ip6_update_pmtu(struct sk_buff
*skb
, struct net
*net
, __be32 mtu
,
1114 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1115 struct dst_entry
*dst
;
1118 memset(&fl6
, 0, sizeof(fl6
));
1119 fl6
.flowi6_oif
= oif
;
1120 fl6
.flowi6_mark
= mark
;
1121 fl6
.flowi6_flags
= 0;
1122 fl6
.daddr
= iph
->daddr
;
1123 fl6
.saddr
= iph
->saddr
;
1124 fl6
.flowlabel
= ip6_flowinfo(iph
);
1126 dst
= ip6_route_output(net
, NULL
, &fl6
);
1128 ip6_rt_update_pmtu(dst
, NULL
, skb
, ntohl(mtu
));
1131 EXPORT_SYMBOL_GPL(ip6_update_pmtu
);
1133 void ip6_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, __be32 mtu
)
1135 ip6_update_pmtu(skb
, sock_net(sk
), mtu
,
1136 sk
->sk_bound_dev_if
, sk
->sk_mark
);
1138 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu
);
1140 void ip6_redirect(struct sk_buff
*skb
, struct net
*net
, int oif
, u32 mark
)
1142 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1143 struct dst_entry
*dst
;
1146 memset(&fl6
, 0, sizeof(fl6
));
1147 fl6
.flowi6_oif
= oif
;
1148 fl6
.flowi6_mark
= mark
;
1149 fl6
.flowi6_flags
= 0;
1150 fl6
.daddr
= iph
->daddr
;
1151 fl6
.saddr
= iph
->saddr
;
1152 fl6
.flowlabel
= ip6_flowinfo(iph
);
1154 dst
= ip6_route_output(net
, NULL
, &fl6
);
1156 rt6_do_redirect(dst
, NULL
, skb
);
1159 EXPORT_SYMBOL_GPL(ip6_redirect
);
1161 void ip6_sk_redirect(struct sk_buff
*skb
, struct sock
*sk
)
1163 ip6_redirect(skb
, sock_net(sk
), sk
->sk_bound_dev_if
, sk
->sk_mark
);
1165 EXPORT_SYMBOL_GPL(ip6_sk_redirect
);
1167 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
)
1169 struct net_device
*dev
= dst
->dev
;
1170 unsigned int mtu
= dst_mtu(dst
);
1171 struct net
*net
= dev_net(dev
);
1173 mtu
-= sizeof(struct ipv6hdr
) + sizeof(struct tcphdr
);
1175 if (mtu
< net
->ipv6
.sysctl
.ip6_rt_min_advmss
)
1176 mtu
= net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
1179 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1180 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1181 * IPV6_MAXPLEN is also valid and means: "any MSS,
1182 * rely only on pmtu discovery"
1184 if (mtu
> IPV6_MAXPLEN
- sizeof(struct tcphdr
))
1189 static unsigned int ip6_mtu(const struct dst_entry
*dst
)
1191 struct inet6_dev
*idev
;
1192 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
1200 idev
= __in6_dev_get(dst
->dev
);
1202 mtu
= idev
->cnf
.mtu6
;
1208 static struct dst_entry
*icmp6_dst_gc_list
;
1209 static DEFINE_SPINLOCK(icmp6_dst_lock
);
1211 struct dst_entry
*icmp6_dst_alloc(struct net_device
*dev
,
1214 struct dst_entry
*dst
;
1215 struct rt6_info
*rt
;
1216 struct inet6_dev
*idev
= in6_dev_get(dev
);
1217 struct net
*net
= dev_net(dev
);
1219 if (unlikely(!idev
))
1220 return ERR_PTR(-ENODEV
);
1222 rt
= ip6_dst_alloc(net
, dev
, 0, NULL
);
1223 if (unlikely(!rt
)) {
1225 dst
= ERR_PTR(-ENOMEM
);
1229 rt
->dst
.flags
|= DST_HOST
;
1230 rt
->dst
.output
= ip6_output
;
1231 atomic_set(&rt
->dst
.__refcnt
, 1);
1232 rt
->rt6i_dst
.addr
= fl6
->daddr
;
1233 rt
->rt6i_dst
.plen
= 128;
1234 rt
->rt6i_idev
= idev
;
1235 dst_metric_set(&rt
->dst
, RTAX_HOPLIMIT
, 0);
1237 spin_lock_bh(&icmp6_dst_lock
);
1238 rt
->dst
.next
= icmp6_dst_gc_list
;
1239 icmp6_dst_gc_list
= &rt
->dst
;
1240 spin_unlock_bh(&icmp6_dst_lock
);
1242 fib6_force_start_gc(net
);
1244 dst
= xfrm_lookup(net
, &rt
->dst
, flowi6_to_flowi(fl6
), NULL
, 0);
1250 int icmp6_dst_gc(void)
1252 struct dst_entry
*dst
, **pprev
;
1255 spin_lock_bh(&icmp6_dst_lock
);
1256 pprev
= &icmp6_dst_gc_list
;
1258 while ((dst
= *pprev
) != NULL
) {
1259 if (!atomic_read(&dst
->__refcnt
)) {
1268 spin_unlock_bh(&icmp6_dst_lock
);
1273 static void icmp6_clean_all(int (*func
)(struct rt6_info
*rt
, void *arg
),
1276 struct dst_entry
*dst
, **pprev
;
1278 spin_lock_bh(&icmp6_dst_lock
);
1279 pprev
= &icmp6_dst_gc_list
;
1280 while ((dst
= *pprev
) != NULL
) {
1281 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1282 if (func(rt
, arg
)) {
1289 spin_unlock_bh(&icmp6_dst_lock
);
1292 static int ip6_dst_gc(struct dst_ops
*ops
)
1294 unsigned long now
= jiffies
;
1295 struct net
*net
= container_of(ops
, struct net
, ipv6
.ip6_dst_ops
);
1296 int rt_min_interval
= net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
1297 int rt_max_size
= net
->ipv6
.sysctl
.ip6_rt_max_size
;
1298 int rt_elasticity
= net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
1299 int rt_gc_timeout
= net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
1300 unsigned long rt_last_gc
= net
->ipv6
.ip6_rt_last_gc
;
1303 entries
= dst_entries_get_fast(ops
);
1304 if (time_after(rt_last_gc
+ rt_min_interval
, now
) &&
1305 entries
<= rt_max_size
)
1308 net
->ipv6
.ip6_rt_gc_expire
++;
1309 fib6_run_gc(net
->ipv6
.ip6_rt_gc_expire
, net
);
1310 net
->ipv6
.ip6_rt_last_gc
= now
;
1311 entries
= dst_entries_get_slow(ops
);
1312 if (entries
< ops
->gc_thresh
)
1313 net
->ipv6
.ip6_rt_gc_expire
= rt_gc_timeout
>>1;
1315 net
->ipv6
.ip6_rt_gc_expire
-= net
->ipv6
.ip6_rt_gc_expire
>>rt_elasticity
;
1316 return entries
> rt_max_size
;
1319 int ip6_dst_hoplimit(struct dst_entry
*dst
)
1321 int hoplimit
= dst_metric_raw(dst
, RTAX_HOPLIMIT
);
1322 if (hoplimit
== 0) {
1323 struct net_device
*dev
= dst
->dev
;
1324 struct inet6_dev
*idev
;
1327 idev
= __in6_dev_get(dev
);
1329 hoplimit
= idev
->cnf
.hop_limit
;
1331 hoplimit
= dev_net(dev
)->ipv6
.devconf_all
->hop_limit
;
1336 EXPORT_SYMBOL(ip6_dst_hoplimit
);
1342 int ip6_route_add(struct fib6_config
*cfg
)
1345 struct net
*net
= cfg
->fc_nlinfo
.nl_net
;
1346 struct rt6_info
*rt
= NULL
;
1347 struct net_device
*dev
= NULL
;
1348 struct inet6_dev
*idev
= NULL
;
1349 struct fib6_table
*table
;
1352 if (cfg
->fc_dst_len
> 128 || cfg
->fc_src_len
> 128)
1354 #ifndef CONFIG_IPV6_SUBTREES
1355 if (cfg
->fc_src_len
)
1358 if (cfg
->fc_ifindex
) {
1360 dev
= dev_get_by_index(net
, cfg
->fc_ifindex
);
1363 idev
= in6_dev_get(dev
);
1368 if (cfg
->fc_metric
== 0)
1369 cfg
->fc_metric
= IP6_RT_PRIO_USER
;
1372 if (cfg
->fc_nlinfo
.nlh
&&
1373 !(cfg
->fc_nlinfo
.nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
1374 table
= fib6_get_table(net
, cfg
->fc_table
);
1376 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1377 table
= fib6_new_table(net
, cfg
->fc_table
);
1380 table
= fib6_new_table(net
, cfg
->fc_table
);
1386 rt
= ip6_dst_alloc(net
, NULL
, DST_NOCOUNT
, table
);
1393 if (cfg
->fc_flags
& RTF_EXPIRES
)
1394 rt6_set_expires(rt
, jiffies
+
1395 clock_t_to_jiffies(cfg
->fc_expires
));
1397 rt6_clean_expires(rt
);
1399 if (cfg
->fc_protocol
== RTPROT_UNSPEC
)
1400 cfg
->fc_protocol
= RTPROT_BOOT
;
1401 rt
->rt6i_protocol
= cfg
->fc_protocol
;
1403 addr_type
= ipv6_addr_type(&cfg
->fc_dst
);
1405 if (addr_type
& IPV6_ADDR_MULTICAST
)
1406 rt
->dst
.input
= ip6_mc_input
;
1407 else if (cfg
->fc_flags
& RTF_LOCAL
)
1408 rt
->dst
.input
= ip6_input
;
1410 rt
->dst
.input
= ip6_forward
;
1412 rt
->dst
.output
= ip6_output
;
1414 ipv6_addr_prefix(&rt
->rt6i_dst
.addr
, &cfg
->fc_dst
, cfg
->fc_dst_len
);
1415 rt
->rt6i_dst
.plen
= cfg
->fc_dst_len
;
1416 if (rt
->rt6i_dst
.plen
== 128)
1417 rt
->dst
.flags
|= DST_HOST
;
1419 if (!(rt
->dst
.flags
& DST_HOST
) && cfg
->fc_mx
) {
1420 u32
*metrics
= kzalloc(sizeof(u32
) * RTAX_MAX
, GFP_KERNEL
);
1425 dst_init_metrics(&rt
->dst
, metrics
, 0);
1427 #ifdef CONFIG_IPV6_SUBTREES
1428 ipv6_addr_prefix(&rt
->rt6i_src
.addr
, &cfg
->fc_src
, cfg
->fc_src_len
);
1429 rt
->rt6i_src
.plen
= cfg
->fc_src_len
;
1432 rt
->rt6i_metric
= cfg
->fc_metric
;
1434 /* We cannot add true routes via loopback here,
1435 they would result in kernel looping; promote them to reject routes
1437 if ((cfg
->fc_flags
& RTF_REJECT
) ||
1438 (dev
&& (dev
->flags
& IFF_LOOPBACK
) &&
1439 !(addr_type
& IPV6_ADDR_LOOPBACK
) &&
1440 !(cfg
->fc_flags
& RTF_LOCAL
))) {
1441 /* hold loopback dev/idev if we haven't done so. */
1442 if (dev
!= net
->loopback_dev
) {
1447 dev
= net
->loopback_dev
;
1449 idev
= in6_dev_get(dev
);
1455 rt
->dst
.output
= ip6_pkt_discard_out
;
1456 rt
->dst
.input
= ip6_pkt_discard
;
1457 rt
->rt6i_flags
= RTF_REJECT
|RTF_NONEXTHOP
;
1458 switch (cfg
->fc_type
) {
1460 rt
->dst
.error
= -EINVAL
;
1463 rt
->dst
.error
= -EACCES
;
1466 rt
->dst
.error
= -EAGAIN
;
1469 rt
->dst
.error
= -ENETUNREACH
;
1475 if (cfg
->fc_flags
& RTF_GATEWAY
) {
1476 const struct in6_addr
*gw_addr
;
1479 gw_addr
= &cfg
->fc_gateway
;
1480 rt
->rt6i_gateway
= *gw_addr
;
1481 gwa_type
= ipv6_addr_type(gw_addr
);
1483 if (gwa_type
!= (IPV6_ADDR_LINKLOCAL
|IPV6_ADDR_UNICAST
)) {
1484 struct rt6_info
*grt
;
1486 /* IPv6 strictly inhibits using not link-local
1487 addresses as nexthop address.
1488 Otherwise, router will not able to send redirects.
1489 It is very good, but in some (rare!) circumstances
1490 (SIT, PtP, NBMA NOARP links) it is handy to allow
1491 some exceptions. --ANK
1494 if (!(gwa_type
& IPV6_ADDR_UNICAST
))
1497 grt
= rt6_lookup(net
, gw_addr
, NULL
, cfg
->fc_ifindex
, 1);
1499 err
= -EHOSTUNREACH
;
1503 if (dev
!= grt
->dst
.dev
) {
1509 idev
= grt
->rt6i_idev
;
1511 in6_dev_hold(grt
->rt6i_idev
);
1513 if (!(grt
->rt6i_flags
& RTF_GATEWAY
))
1521 if (!dev
|| (dev
->flags
& IFF_LOOPBACK
))
1529 if (!ipv6_addr_any(&cfg
->fc_prefsrc
)) {
1530 if (!ipv6_chk_addr(net
, &cfg
->fc_prefsrc
, dev
, 0)) {
1534 rt
->rt6i_prefsrc
.addr
= cfg
->fc_prefsrc
;
1535 rt
->rt6i_prefsrc
.plen
= 128;
1537 rt
->rt6i_prefsrc
.plen
= 0;
1539 rt
->rt6i_flags
= cfg
->fc_flags
;
1546 nla_for_each_attr(nla
, cfg
->fc_mx
, cfg
->fc_mx_len
, remaining
) {
1547 int type
= nla_type(nla
);
1550 if (type
> RTAX_MAX
) {
1555 dst_metric_set(&rt
->dst
, type
, nla_get_u32(nla
));
1561 rt
->rt6i_idev
= idev
;
1562 rt
->rt6i_table
= table
;
1564 cfg
->fc_nlinfo
.nl_net
= dev_net(dev
);
1566 return __ip6_ins_rt(rt
, &cfg
->fc_nlinfo
);
1578 static int __ip6_del_rt(struct rt6_info
*rt
, struct nl_info
*info
)
1581 struct fib6_table
*table
;
1582 struct net
*net
= dev_net(rt
->dst
.dev
);
1584 if (rt
== net
->ipv6
.ip6_null_entry
) {
1589 table
= rt
->rt6i_table
;
1590 write_lock_bh(&table
->tb6_lock
);
1591 err
= fib6_del(rt
, info
);
1592 write_unlock_bh(&table
->tb6_lock
);
1599 int ip6_del_rt(struct rt6_info
*rt
)
1601 struct nl_info info
= {
1602 .nl_net
= dev_net(rt
->dst
.dev
),
1604 return __ip6_del_rt(rt
, &info
);
1607 static int ip6_route_del(struct fib6_config
*cfg
)
1609 struct fib6_table
*table
;
1610 struct fib6_node
*fn
;
1611 struct rt6_info
*rt
;
1614 table
= fib6_get_table(cfg
->fc_nlinfo
.nl_net
, cfg
->fc_table
);
1618 read_lock_bh(&table
->tb6_lock
);
1620 fn
= fib6_locate(&table
->tb6_root
,
1621 &cfg
->fc_dst
, cfg
->fc_dst_len
,
1622 &cfg
->fc_src
, cfg
->fc_src_len
);
1625 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1626 if (cfg
->fc_ifindex
&&
1628 rt
->dst
.dev
->ifindex
!= cfg
->fc_ifindex
))
1630 if (cfg
->fc_flags
& RTF_GATEWAY
&&
1631 !ipv6_addr_equal(&cfg
->fc_gateway
, &rt
->rt6i_gateway
))
1633 if (cfg
->fc_metric
&& cfg
->fc_metric
!= rt
->rt6i_metric
)
1636 read_unlock_bh(&table
->tb6_lock
);
1638 return __ip6_del_rt(rt
, &cfg
->fc_nlinfo
);
1641 read_unlock_bh(&table
->tb6_lock
);
1646 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
, struct sk_buff
*skb
)
1648 struct net
*net
= dev_net(skb
->dev
);
1649 struct netevent_redirect netevent
;
1650 struct rt6_info
*rt
, *nrt
= NULL
;
1651 struct ndisc_options ndopts
;
1652 struct inet6_dev
*in6_dev
;
1653 struct neighbour
*neigh
;
1655 int optlen
, on_link
;
1658 optlen
= skb_tail_pointer(skb
) - skb_transport_header(skb
);
1659 optlen
-= sizeof(*msg
);
1662 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1666 msg
= (struct rd_msg
*)icmp6_hdr(skb
);
1668 if (ipv6_addr_is_multicast(&msg
->dest
)) {
1669 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1674 if (ipv6_addr_equal(&msg
->dest
, &msg
->target
)) {
1676 } else if (ipv6_addr_type(&msg
->target
) !=
1677 (IPV6_ADDR_UNICAST
|IPV6_ADDR_LINKLOCAL
)) {
1678 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1682 in6_dev
= __in6_dev_get(skb
->dev
);
1685 if (in6_dev
->cnf
.forwarding
|| !in6_dev
->cnf
.accept_redirects
)
1689 * The IP source address of the Redirect MUST be the same as the current
1690 * first-hop router for the specified ICMP Destination Address.
1693 if (!ndisc_parse_options(msg
->opt
, optlen
, &ndopts
)) {
1694 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1699 if (ndopts
.nd_opts_tgt_lladdr
) {
1700 lladdr
= ndisc_opt_addr_data(ndopts
.nd_opts_tgt_lladdr
,
1703 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1708 rt
= (struct rt6_info
*) dst
;
1709 if (rt
== net
->ipv6
.ip6_null_entry
) {
1710 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1714 /* Redirect received -> path was valid.
1715 * Look, redirects are sent only in response to data packets,
1716 * so that this nexthop apparently is reachable. --ANK
1718 dst_confirm(&rt
->dst
);
1720 neigh
= __neigh_lookup(&nd_tbl
, &msg
->target
, skb
->dev
, 1);
1725 * We have finally decided to accept it.
1728 neigh_update(neigh
, lladdr
, NUD_STALE
,
1729 NEIGH_UPDATE_F_WEAK_OVERRIDE
|
1730 NEIGH_UPDATE_F_OVERRIDE
|
1731 (on_link
? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER
|
1732 NEIGH_UPDATE_F_ISROUTER
))
1735 nrt
= ip6_rt_copy(rt
, &msg
->dest
);
1739 nrt
->rt6i_flags
= RTF_GATEWAY
|RTF_UP
|RTF_DYNAMIC
|RTF_CACHE
;
1741 nrt
->rt6i_flags
&= ~RTF_GATEWAY
;
1743 nrt
->rt6i_gateway
= *(struct in6_addr
*)neigh
->primary_key
;
1745 if (ip6_ins_rt(nrt
))
1748 netevent
.old
= &rt
->dst
;
1749 netevent
.new = &nrt
->dst
;
1750 netevent
.daddr
= &msg
->dest
;
1751 netevent
.neigh
= neigh
;
1752 call_netevent_notifiers(NETEVENT_REDIRECT
, &netevent
);
1754 if (rt
->rt6i_flags
& RTF_CACHE
) {
1755 rt
= (struct rt6_info
*) dst_clone(&rt
->dst
);
1760 neigh_release(neigh
);
1764 * Misc support functions
1767 static struct rt6_info
*ip6_rt_copy(struct rt6_info
*ort
,
1768 const struct in6_addr
*dest
)
1770 struct net
*net
= dev_net(ort
->dst
.dev
);
1771 struct rt6_info
*rt
= ip6_dst_alloc(net
, ort
->dst
.dev
, 0,
1775 rt
->dst
.input
= ort
->dst
.input
;
1776 rt
->dst
.output
= ort
->dst
.output
;
1777 rt
->dst
.flags
|= DST_HOST
;
1779 rt
->rt6i_dst
.addr
= *dest
;
1780 rt
->rt6i_dst
.plen
= 128;
1781 dst_copy_metrics(&rt
->dst
, &ort
->dst
);
1782 rt
->dst
.error
= ort
->dst
.error
;
1783 rt
->rt6i_idev
= ort
->rt6i_idev
;
1785 in6_dev_hold(rt
->rt6i_idev
);
1786 rt
->dst
.lastuse
= jiffies
;
1788 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
1789 rt
->rt6i_flags
= ort
->rt6i_flags
;
1790 if ((ort
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
)) ==
1791 (RTF_DEFAULT
| RTF_ADDRCONF
))
1792 rt6_set_from(rt
, ort
);
1793 rt
->rt6i_metric
= 0;
1795 #ifdef CONFIG_IPV6_SUBTREES
1796 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1798 memcpy(&rt
->rt6i_prefsrc
, &ort
->rt6i_prefsrc
, sizeof(struct rt6key
));
1799 rt
->rt6i_table
= ort
->rt6i_table
;
1804 #ifdef CONFIG_IPV6_ROUTE_INFO
1805 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
1806 const struct in6_addr
*prefix
, int prefixlen
,
1807 const struct in6_addr
*gwaddr
, int ifindex
)
1809 struct fib6_node
*fn
;
1810 struct rt6_info
*rt
= NULL
;
1811 struct fib6_table
*table
;
1813 table
= fib6_get_table(net
, RT6_TABLE_INFO
);
1817 read_lock_bh(&table
->tb6_lock
);
1818 fn
= fib6_locate(&table
->tb6_root
, prefix
,prefixlen
, NULL
, 0);
1822 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1823 if (rt
->dst
.dev
->ifindex
!= ifindex
)
1825 if ((rt
->rt6i_flags
& (RTF_ROUTEINFO
|RTF_GATEWAY
)) != (RTF_ROUTEINFO
|RTF_GATEWAY
))
1827 if (!ipv6_addr_equal(&rt
->rt6i_gateway
, gwaddr
))
1833 read_unlock_bh(&table
->tb6_lock
);
1837 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
1838 const struct in6_addr
*prefix
, int prefixlen
,
1839 const struct in6_addr
*gwaddr
, int ifindex
,
1842 struct fib6_config cfg
= {
1843 .fc_table
= RT6_TABLE_INFO
,
1844 .fc_metric
= IP6_RT_PRIO_USER
,
1845 .fc_ifindex
= ifindex
,
1846 .fc_dst_len
= prefixlen
,
1847 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_ROUTEINFO
|
1848 RTF_UP
| RTF_PREF(pref
),
1849 .fc_nlinfo
.portid
= 0,
1850 .fc_nlinfo
.nlh
= NULL
,
1851 .fc_nlinfo
.nl_net
= net
,
1854 cfg
.fc_dst
= *prefix
;
1855 cfg
.fc_gateway
= *gwaddr
;
1857 /* We should treat it as a default route if prefix length is 0. */
1859 cfg
.fc_flags
|= RTF_DEFAULT
;
1861 ip6_route_add(&cfg
);
1863 return rt6_get_route_info(net
, prefix
, prefixlen
, gwaddr
, ifindex
);
1867 struct rt6_info
*rt6_get_dflt_router(const struct in6_addr
*addr
, struct net_device
*dev
)
1869 struct rt6_info
*rt
;
1870 struct fib6_table
*table
;
1872 table
= fib6_get_table(dev_net(dev
), RT6_TABLE_DFLT
);
1876 read_lock_bh(&table
->tb6_lock
);
1877 for (rt
= table
->tb6_root
.leaf
; rt
; rt
=rt
->dst
.rt6_next
) {
1878 if (dev
== rt
->dst
.dev
&&
1879 ((rt
->rt6i_flags
& (RTF_ADDRCONF
| RTF_DEFAULT
)) == (RTF_ADDRCONF
| RTF_DEFAULT
)) &&
1880 ipv6_addr_equal(&rt
->rt6i_gateway
, addr
))
1885 read_unlock_bh(&table
->tb6_lock
);
1889 struct rt6_info
*rt6_add_dflt_router(const struct in6_addr
*gwaddr
,
1890 struct net_device
*dev
,
1893 struct fib6_config cfg
= {
1894 .fc_table
= RT6_TABLE_DFLT
,
1895 .fc_metric
= IP6_RT_PRIO_USER
,
1896 .fc_ifindex
= dev
->ifindex
,
1897 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_DEFAULT
|
1898 RTF_UP
| RTF_EXPIRES
| RTF_PREF(pref
),
1899 .fc_nlinfo
.portid
= 0,
1900 .fc_nlinfo
.nlh
= NULL
,
1901 .fc_nlinfo
.nl_net
= dev_net(dev
),
1904 cfg
.fc_gateway
= *gwaddr
;
1906 ip6_route_add(&cfg
);
1908 return rt6_get_dflt_router(gwaddr
, dev
);
1911 void rt6_purge_dflt_routers(struct net
*net
)
1913 struct rt6_info
*rt
;
1914 struct fib6_table
*table
;
1916 /* NOTE: Keep consistent with rt6_get_dflt_router */
1917 table
= fib6_get_table(net
, RT6_TABLE_DFLT
);
1922 read_lock_bh(&table
->tb6_lock
);
1923 for (rt
= table
->tb6_root
.leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1924 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
) &&
1925 (!rt
->rt6i_idev
|| rt
->rt6i_idev
->cnf
.accept_ra
!= 2)) {
1927 read_unlock_bh(&table
->tb6_lock
);
1932 read_unlock_bh(&table
->tb6_lock
);
1935 static void rtmsg_to_fib6_config(struct net
*net
,
1936 struct in6_rtmsg
*rtmsg
,
1937 struct fib6_config
*cfg
)
1939 memset(cfg
, 0, sizeof(*cfg
));
1941 cfg
->fc_table
= RT6_TABLE_MAIN
;
1942 cfg
->fc_ifindex
= rtmsg
->rtmsg_ifindex
;
1943 cfg
->fc_metric
= rtmsg
->rtmsg_metric
;
1944 cfg
->fc_expires
= rtmsg
->rtmsg_info
;
1945 cfg
->fc_dst_len
= rtmsg
->rtmsg_dst_len
;
1946 cfg
->fc_src_len
= rtmsg
->rtmsg_src_len
;
1947 cfg
->fc_flags
= rtmsg
->rtmsg_flags
;
1949 cfg
->fc_nlinfo
.nl_net
= net
;
1951 cfg
->fc_dst
= rtmsg
->rtmsg_dst
;
1952 cfg
->fc_src
= rtmsg
->rtmsg_src
;
1953 cfg
->fc_gateway
= rtmsg
->rtmsg_gateway
;
1956 int ipv6_route_ioctl(struct net
*net
, unsigned int cmd
, void __user
*arg
)
1958 struct fib6_config cfg
;
1959 struct in6_rtmsg rtmsg
;
1963 case SIOCADDRT
: /* Add a route */
1964 case SIOCDELRT
: /* Delete a route */
1965 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1967 err
= copy_from_user(&rtmsg
, arg
,
1968 sizeof(struct in6_rtmsg
));
1972 rtmsg_to_fib6_config(net
, &rtmsg
, &cfg
);
1977 err
= ip6_route_add(&cfg
);
1980 err
= ip6_route_del(&cfg
);
1994 * Drop the packet on the floor
1997 static int ip6_pkt_drop(struct sk_buff
*skb
, u8 code
, int ipstats_mib_noroutes
)
2000 struct dst_entry
*dst
= skb_dst(skb
);
2001 switch (ipstats_mib_noroutes
) {
2002 case IPSTATS_MIB_INNOROUTES
:
2003 type
= ipv6_addr_type(&ipv6_hdr(skb
)->daddr
);
2004 if (type
== IPV6_ADDR_ANY
) {
2005 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2006 IPSTATS_MIB_INADDRERRORS
);
2010 case IPSTATS_MIB_OUTNOROUTES
:
2011 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2012 ipstats_mib_noroutes
);
2015 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, code
, 0);
2020 static int ip6_pkt_discard(struct sk_buff
*skb
)
2022 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_INNOROUTES
);
2025 static int ip6_pkt_discard_out(struct sk_buff
*skb
)
2027 skb
->dev
= skb_dst(skb
)->dev
;
2028 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_OUTNOROUTES
);
2031 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2033 static int ip6_pkt_prohibit(struct sk_buff
*skb
)
2035 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_INNOROUTES
);
2038 static int ip6_pkt_prohibit_out(struct sk_buff
*skb
)
2040 skb
->dev
= skb_dst(skb
)->dev
;
2041 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_OUTNOROUTES
);
2047 * Allocate a dst for local (unicast / anycast) address.
2050 struct rt6_info
*addrconf_dst_alloc(struct inet6_dev
*idev
,
2051 const struct in6_addr
*addr
,
2054 struct net
*net
= dev_net(idev
->dev
);
2055 struct rt6_info
*rt
= ip6_dst_alloc(net
, net
->loopback_dev
, 0, NULL
);
2058 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2059 return ERR_PTR(-ENOMEM
);
2064 rt
->dst
.flags
|= DST_HOST
;
2065 rt
->dst
.input
= ip6_input
;
2066 rt
->dst
.output
= ip6_output
;
2067 rt
->rt6i_idev
= idev
;
2069 rt
->rt6i_flags
= RTF_UP
| RTF_NONEXTHOP
;
2071 rt
->rt6i_flags
|= RTF_ANYCAST
;
2073 rt
->rt6i_flags
|= RTF_LOCAL
;
2075 rt
->rt6i_dst
.addr
= *addr
;
2076 rt
->rt6i_dst
.plen
= 128;
2077 rt
->rt6i_table
= fib6_get_table(net
, RT6_TABLE_LOCAL
);
2079 atomic_set(&rt
->dst
.__refcnt
, 1);
2084 int ip6_route_get_saddr(struct net
*net
,
2085 struct rt6_info
*rt
,
2086 const struct in6_addr
*daddr
,
2088 struct in6_addr
*saddr
)
2090 struct inet6_dev
*idev
= ip6_dst_idev((struct dst_entry
*)rt
);
2092 if (rt
->rt6i_prefsrc
.plen
)
2093 *saddr
= rt
->rt6i_prefsrc
.addr
;
2095 err
= ipv6_dev_get_saddr(net
, idev
? idev
->dev
: NULL
,
2096 daddr
, prefs
, saddr
);
2100 /* remove deleted ip from prefsrc entries */
2101 struct arg_dev_net_ip
{
2102 struct net_device
*dev
;
2104 struct in6_addr
*addr
;
2107 static int fib6_remove_prefsrc(struct rt6_info
*rt
, void *arg
)
2109 struct net_device
*dev
= ((struct arg_dev_net_ip
*)arg
)->dev
;
2110 struct net
*net
= ((struct arg_dev_net_ip
*)arg
)->net
;
2111 struct in6_addr
*addr
= ((struct arg_dev_net_ip
*)arg
)->addr
;
2113 if (((void *)rt
->dst
.dev
== dev
|| !dev
) &&
2114 rt
!= net
->ipv6
.ip6_null_entry
&&
2115 ipv6_addr_equal(addr
, &rt
->rt6i_prefsrc
.addr
)) {
2116 /* remove prefsrc entry */
2117 rt
->rt6i_prefsrc
.plen
= 0;
2122 void rt6_remove_prefsrc(struct inet6_ifaddr
*ifp
)
2124 struct net
*net
= dev_net(ifp
->idev
->dev
);
2125 struct arg_dev_net_ip adni
= {
2126 .dev
= ifp
->idev
->dev
,
2130 fib6_clean_all(net
, fib6_remove_prefsrc
, 0, &adni
);
2133 struct arg_dev_net
{
2134 struct net_device
*dev
;
2138 static int fib6_ifdown(struct rt6_info
*rt
, void *arg
)
2140 const struct arg_dev_net
*adn
= arg
;
2141 const struct net_device
*dev
= adn
->dev
;
2143 if ((rt
->dst
.dev
== dev
|| !dev
) &&
2144 rt
!= adn
->net
->ipv6
.ip6_null_entry
)
2150 void rt6_ifdown(struct net
*net
, struct net_device
*dev
)
2152 struct arg_dev_net adn
= {
2157 fib6_clean_all(net
, fib6_ifdown
, 0, &adn
);
2158 icmp6_clean_all(fib6_ifdown
, &adn
);
2161 struct rt6_mtu_change_arg
{
2162 struct net_device
*dev
;
2166 static int rt6_mtu_change_route(struct rt6_info
*rt
, void *p_arg
)
2168 struct rt6_mtu_change_arg
*arg
= (struct rt6_mtu_change_arg
*) p_arg
;
2169 struct inet6_dev
*idev
;
2171 /* In IPv6 pmtu discovery is not optional,
2172 so that RTAX_MTU lock cannot disable it.
2173 We still use this lock to block changes
2174 caused by addrconf/ndisc.
2177 idev
= __in6_dev_get(arg
->dev
);
2181 /* For administrative MTU increase, there is no way to discover
2182 IPv6 PMTU increase, so PMTU increase should be updated here.
2183 Since RFC 1981 doesn't include administrative MTU increase
2184 update PMTU increase is a MUST. (i.e. jumbo frame)
2187 If new MTU is less than route PMTU, this new MTU will be the
2188 lowest MTU in the path, update the route PMTU to reflect PMTU
2189 decreases; if new MTU is greater than route PMTU, and the
2190 old MTU is the lowest MTU in the path, update the route PMTU
2191 to reflect the increase. In this case if the other nodes' MTU
2192 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2195 if (rt
->dst
.dev
== arg
->dev
&&
2196 !dst_metric_locked(&rt
->dst
, RTAX_MTU
) &&
2197 (dst_mtu(&rt
->dst
) >= arg
->mtu
||
2198 (dst_mtu(&rt
->dst
) < arg
->mtu
&&
2199 dst_mtu(&rt
->dst
) == idev
->cnf
.mtu6
))) {
2200 dst_metric_set(&rt
->dst
, RTAX_MTU
, arg
->mtu
);
2205 void rt6_mtu_change(struct net_device
*dev
, unsigned int mtu
)
2207 struct rt6_mtu_change_arg arg
= {
2212 fib6_clean_all(dev_net(dev
), rt6_mtu_change_route
, 0, &arg
);
2215 static const struct nla_policy rtm_ipv6_policy
[RTA_MAX
+1] = {
2216 [RTA_GATEWAY
] = { .len
= sizeof(struct in6_addr
) },
2217 [RTA_OIF
] = { .type
= NLA_U32
},
2218 [RTA_IIF
] = { .type
= NLA_U32
},
2219 [RTA_PRIORITY
] = { .type
= NLA_U32
},
2220 [RTA_METRICS
] = { .type
= NLA_NESTED
},
2221 [RTA_MULTIPATH
] = { .len
= sizeof(struct rtnexthop
) },
2224 static int rtm_to_fib6_config(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
2225 struct fib6_config
*cfg
)
2228 struct nlattr
*tb
[RTA_MAX
+1];
2231 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2236 rtm
= nlmsg_data(nlh
);
2237 memset(cfg
, 0, sizeof(*cfg
));
2239 cfg
->fc_table
= rtm
->rtm_table
;
2240 cfg
->fc_dst_len
= rtm
->rtm_dst_len
;
2241 cfg
->fc_src_len
= rtm
->rtm_src_len
;
2242 cfg
->fc_flags
= RTF_UP
;
2243 cfg
->fc_protocol
= rtm
->rtm_protocol
;
2244 cfg
->fc_type
= rtm
->rtm_type
;
2246 if (rtm
->rtm_type
== RTN_UNREACHABLE
||
2247 rtm
->rtm_type
== RTN_BLACKHOLE
||
2248 rtm
->rtm_type
== RTN_PROHIBIT
||
2249 rtm
->rtm_type
== RTN_THROW
)
2250 cfg
->fc_flags
|= RTF_REJECT
;
2252 if (rtm
->rtm_type
== RTN_LOCAL
)
2253 cfg
->fc_flags
|= RTF_LOCAL
;
2255 cfg
->fc_nlinfo
.portid
= NETLINK_CB(skb
).portid
;
2256 cfg
->fc_nlinfo
.nlh
= nlh
;
2257 cfg
->fc_nlinfo
.nl_net
= sock_net(skb
->sk
);
2259 if (tb
[RTA_GATEWAY
]) {
2260 nla_memcpy(&cfg
->fc_gateway
, tb
[RTA_GATEWAY
], 16);
2261 cfg
->fc_flags
|= RTF_GATEWAY
;
2265 int plen
= (rtm
->rtm_dst_len
+ 7) >> 3;
2267 if (nla_len(tb
[RTA_DST
]) < plen
)
2270 nla_memcpy(&cfg
->fc_dst
, tb
[RTA_DST
], plen
);
2274 int plen
= (rtm
->rtm_src_len
+ 7) >> 3;
2276 if (nla_len(tb
[RTA_SRC
]) < plen
)
2279 nla_memcpy(&cfg
->fc_src
, tb
[RTA_SRC
], plen
);
2282 if (tb
[RTA_PREFSRC
])
2283 nla_memcpy(&cfg
->fc_prefsrc
, tb
[RTA_PREFSRC
], 16);
2286 cfg
->fc_ifindex
= nla_get_u32(tb
[RTA_OIF
]);
2288 if (tb
[RTA_PRIORITY
])
2289 cfg
->fc_metric
= nla_get_u32(tb
[RTA_PRIORITY
]);
2291 if (tb
[RTA_METRICS
]) {
2292 cfg
->fc_mx
= nla_data(tb
[RTA_METRICS
]);
2293 cfg
->fc_mx_len
= nla_len(tb
[RTA_METRICS
]);
2297 cfg
->fc_table
= nla_get_u32(tb
[RTA_TABLE
]);
2299 if (tb
[RTA_MULTIPATH
]) {
2300 cfg
->fc_mp
= nla_data(tb
[RTA_MULTIPATH
]);
2301 cfg
->fc_mp_len
= nla_len(tb
[RTA_MULTIPATH
]);
2309 static int ip6_route_multipath(struct fib6_config
*cfg
, int add
)
2311 struct fib6_config r_cfg
;
2312 struct rtnexthop
*rtnh
;
2315 int err
= 0, last_err
= 0;
2318 rtnh
= (struct rtnexthop
*)cfg
->fc_mp
;
2319 remaining
= cfg
->fc_mp_len
;
2321 /* Parse a Multipath Entry */
2322 while (rtnh_ok(rtnh
, remaining
)) {
2323 memcpy(&r_cfg
, cfg
, sizeof(*cfg
));
2324 if (rtnh
->rtnh_ifindex
)
2325 r_cfg
.fc_ifindex
= rtnh
->rtnh_ifindex
;
2327 attrlen
= rtnh_attrlen(rtnh
);
2329 struct nlattr
*nla
, *attrs
= rtnh_attrs(rtnh
);
2331 nla
= nla_find(attrs
, attrlen
, RTA_GATEWAY
);
2333 nla_memcpy(&r_cfg
.fc_gateway
, nla
, 16);
2334 r_cfg
.fc_flags
|= RTF_GATEWAY
;
2337 err
= add
? ip6_route_add(&r_cfg
) : ip6_route_del(&r_cfg
);
2340 /* If we are trying to remove a route, do not stop the
2341 * loop when ip6_route_del() fails (because next hop is
2342 * already gone), we should try to remove all next hops.
2345 /* If add fails, we should try to delete all
2346 * next hops that have been already added.
2352 /* Because each route is added like a single route we remove
2353 * this flag after the first nexthop (if there is a collision,
2354 * we have already fail to add the first nexthop:
2355 * fib6_add_rt2node() has reject it).
2357 cfg
->fc_nlinfo
.nlh
->nlmsg_flags
&= ~NLM_F_EXCL
;
2358 rtnh
= rtnh_next(rtnh
, &remaining
);
2364 static int inet6_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
* nlh
)
2366 struct fib6_config cfg
;
2369 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2374 return ip6_route_multipath(&cfg
, 0);
2376 return ip6_route_del(&cfg
);
2379 static int inet6_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
* nlh
)
2381 struct fib6_config cfg
;
2384 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2389 return ip6_route_multipath(&cfg
, 1);
2391 return ip6_route_add(&cfg
);
2394 static inline size_t rt6_nlmsg_size(void)
2396 return NLMSG_ALIGN(sizeof(struct rtmsg
))
2397 + nla_total_size(16) /* RTA_SRC */
2398 + nla_total_size(16) /* RTA_DST */
2399 + nla_total_size(16) /* RTA_GATEWAY */
2400 + nla_total_size(16) /* RTA_PREFSRC */
2401 + nla_total_size(4) /* RTA_TABLE */
2402 + nla_total_size(4) /* RTA_IIF */
2403 + nla_total_size(4) /* RTA_OIF */
2404 + nla_total_size(4) /* RTA_PRIORITY */
2405 + RTAX_MAX
* nla_total_size(4) /* RTA_METRICS */
2406 + nla_total_size(sizeof(struct rta_cacheinfo
));
2409 static int rt6_fill_node(struct net
*net
,
2410 struct sk_buff
*skb
, struct rt6_info
*rt
,
2411 struct in6_addr
*dst
, struct in6_addr
*src
,
2412 int iif
, int type
, u32 portid
, u32 seq
,
2413 int prefix
, int nowait
, unsigned int flags
)
2416 struct nlmsghdr
*nlh
;
2420 if (prefix
) { /* user wants prefix routes only */
2421 if (!(rt
->rt6i_flags
& RTF_PREFIX_RT
)) {
2422 /* success since this is not a prefix route */
2427 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*rtm
), flags
);
2431 rtm
= nlmsg_data(nlh
);
2432 rtm
->rtm_family
= AF_INET6
;
2433 rtm
->rtm_dst_len
= rt
->rt6i_dst
.plen
;
2434 rtm
->rtm_src_len
= rt
->rt6i_src
.plen
;
2437 table
= rt
->rt6i_table
->tb6_id
;
2439 table
= RT6_TABLE_UNSPEC
;
2440 rtm
->rtm_table
= table
;
2441 if (nla_put_u32(skb
, RTA_TABLE
, table
))
2442 goto nla_put_failure
;
2443 if (rt
->rt6i_flags
& RTF_REJECT
) {
2444 switch (rt
->dst
.error
) {
2446 rtm
->rtm_type
= RTN_BLACKHOLE
;
2449 rtm
->rtm_type
= RTN_PROHIBIT
;
2452 rtm
->rtm_type
= RTN_THROW
;
2455 rtm
->rtm_type
= RTN_UNREACHABLE
;
2459 else if (rt
->rt6i_flags
& RTF_LOCAL
)
2460 rtm
->rtm_type
= RTN_LOCAL
;
2461 else if (rt
->dst
.dev
&& (rt
->dst
.dev
->flags
& IFF_LOOPBACK
))
2462 rtm
->rtm_type
= RTN_LOCAL
;
2464 rtm
->rtm_type
= RTN_UNICAST
;
2466 rtm
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2467 rtm
->rtm_protocol
= rt
->rt6i_protocol
;
2468 if (rt
->rt6i_flags
& RTF_DYNAMIC
)
2469 rtm
->rtm_protocol
= RTPROT_REDIRECT
;
2470 else if (rt
->rt6i_flags
& RTF_ADDRCONF
) {
2471 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ROUTEINFO
))
2472 rtm
->rtm_protocol
= RTPROT_RA
;
2474 rtm
->rtm_protocol
= RTPROT_KERNEL
;
2477 if (rt
->rt6i_flags
& RTF_CACHE
)
2478 rtm
->rtm_flags
|= RTM_F_CLONED
;
2481 if (nla_put(skb
, RTA_DST
, 16, dst
))
2482 goto nla_put_failure
;
2483 rtm
->rtm_dst_len
= 128;
2484 } else if (rtm
->rtm_dst_len
)
2485 if (nla_put(skb
, RTA_DST
, 16, &rt
->rt6i_dst
.addr
))
2486 goto nla_put_failure
;
2487 #ifdef CONFIG_IPV6_SUBTREES
2489 if (nla_put(skb
, RTA_SRC
, 16, src
))
2490 goto nla_put_failure
;
2491 rtm
->rtm_src_len
= 128;
2492 } else if (rtm
->rtm_src_len
&&
2493 nla_put(skb
, RTA_SRC
, 16, &rt
->rt6i_src
.addr
))
2494 goto nla_put_failure
;
2497 #ifdef CONFIG_IPV6_MROUTE
2498 if (ipv6_addr_is_multicast(&rt
->rt6i_dst
.addr
)) {
2499 int err
= ip6mr_get_route(net
, skb
, rtm
, nowait
);
2504 goto nla_put_failure
;
2506 if (err
== -EMSGSIZE
)
2507 goto nla_put_failure
;
2512 if (nla_put_u32(skb
, RTA_IIF
, iif
))
2513 goto nla_put_failure
;
2515 struct in6_addr saddr_buf
;
2516 if (ip6_route_get_saddr(net
, rt
, dst
, 0, &saddr_buf
) == 0 &&
2517 nla_put(skb
, RTA_PREFSRC
, 16, &saddr_buf
))
2518 goto nla_put_failure
;
2521 if (rt
->rt6i_prefsrc
.plen
) {
2522 struct in6_addr saddr_buf
;
2523 saddr_buf
= rt
->rt6i_prefsrc
.addr
;
2524 if (nla_put(skb
, RTA_PREFSRC
, 16, &saddr_buf
))
2525 goto nla_put_failure
;
2528 if (rtnetlink_put_metrics(skb
, dst_metrics_ptr(&rt
->dst
)) < 0)
2529 goto nla_put_failure
;
2531 if (rt
->rt6i_flags
& RTF_GATEWAY
) {
2532 if (nla_put(skb
, RTA_GATEWAY
, 16, &rt
->rt6i_gateway
) < 0)
2533 goto nla_put_failure
;
2537 nla_put_u32(skb
, RTA_OIF
, rt
->dst
.dev
->ifindex
))
2538 goto nla_put_failure
;
2539 if (nla_put_u32(skb
, RTA_PRIORITY
, rt
->rt6i_metric
))
2540 goto nla_put_failure
;
2542 expires
= (rt
->rt6i_flags
& RTF_EXPIRES
) ? rt
->dst
.expires
- jiffies
: 0;
2544 if (rtnl_put_cacheinfo(skb
, &rt
->dst
, 0, expires
, rt
->dst
.error
) < 0)
2545 goto nla_put_failure
;
2547 return nlmsg_end(skb
, nlh
);
2550 nlmsg_cancel(skb
, nlh
);
2554 int rt6_dump_route(struct rt6_info
*rt
, void *p_arg
)
2556 struct rt6_rtnl_dump_arg
*arg
= (struct rt6_rtnl_dump_arg
*) p_arg
;
2559 if (nlmsg_len(arg
->cb
->nlh
) >= sizeof(struct rtmsg
)) {
2560 struct rtmsg
*rtm
= nlmsg_data(arg
->cb
->nlh
);
2561 prefix
= (rtm
->rtm_flags
& RTM_F_PREFIX
) != 0;
2565 return rt6_fill_node(arg
->net
,
2566 arg
->skb
, rt
, NULL
, NULL
, 0, RTM_NEWROUTE
,
2567 NETLINK_CB(arg
->cb
->skb
).portid
, arg
->cb
->nlh
->nlmsg_seq
,
2568 prefix
, 0, NLM_F_MULTI
);
2571 static int inet6_rtm_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
* nlh
)
2573 struct net
*net
= sock_net(in_skb
->sk
);
2574 struct nlattr
*tb
[RTA_MAX
+1];
2575 struct rt6_info
*rt
;
2576 struct sk_buff
*skb
;
2579 int err
, iif
= 0, oif
= 0;
2581 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2586 memset(&fl6
, 0, sizeof(fl6
));
2589 if (nla_len(tb
[RTA_SRC
]) < sizeof(struct in6_addr
))
2592 fl6
.saddr
= *(struct in6_addr
*)nla_data(tb
[RTA_SRC
]);
2596 if (nla_len(tb
[RTA_DST
]) < sizeof(struct in6_addr
))
2599 fl6
.daddr
= *(struct in6_addr
*)nla_data(tb
[RTA_DST
]);
2603 iif
= nla_get_u32(tb
[RTA_IIF
]);
2606 oif
= nla_get_u32(tb
[RTA_OIF
]);
2609 struct net_device
*dev
;
2612 dev
= __dev_get_by_index(net
, iif
);
2618 fl6
.flowi6_iif
= iif
;
2620 if (!ipv6_addr_any(&fl6
.saddr
))
2621 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
2623 rt
= (struct rt6_info
*)ip6_route_input_lookup(net
, dev
, &fl6
,
2626 fl6
.flowi6_oif
= oif
;
2628 rt
= (struct rt6_info
*)ip6_route_output(net
, NULL
, &fl6
);
2631 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
2638 /* Reserve room for dummy headers, this skb can pass
2639 through good chunk of routing engine.
2641 skb_reset_mac_header(skb
);
2642 skb_reserve(skb
, MAX_HEADER
+ sizeof(struct ipv6hdr
));
2644 skb_dst_set(skb
, &rt
->dst
);
2646 err
= rt6_fill_node(net
, skb
, rt
, &fl6
.daddr
, &fl6
.saddr
, iif
,
2647 RTM_NEWROUTE
, NETLINK_CB(in_skb
).portid
,
2648 nlh
->nlmsg_seq
, 0, 0, 0);
2654 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).portid
);
2659 void inet6_rt_notify(int event
, struct rt6_info
*rt
, struct nl_info
*info
)
2661 struct sk_buff
*skb
;
2662 struct net
*net
= info
->nl_net
;
2667 seq
= info
->nlh
? info
->nlh
->nlmsg_seq
: 0;
2669 skb
= nlmsg_new(rt6_nlmsg_size(), gfp_any());
2673 err
= rt6_fill_node(net
, skb
, rt
, NULL
, NULL
, 0,
2674 event
, info
->portid
, seq
, 0, 0, 0);
2676 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2677 WARN_ON(err
== -EMSGSIZE
);
2681 rtnl_notify(skb
, net
, info
->portid
, RTNLGRP_IPV6_ROUTE
,
2682 info
->nlh
, gfp_any());
2686 rtnl_set_sk_err(net
, RTNLGRP_IPV6_ROUTE
, err
);
2689 static int ip6_route_dev_notify(struct notifier_block
*this,
2690 unsigned long event
, void *ptr
)
2692 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
2693 struct net
*net
= dev_net(dev
);
2695 if (event
== NETDEV_REGISTER
&& (dev
->flags
& IFF_LOOPBACK
)) {
2696 net
->ipv6
.ip6_null_entry
->dst
.dev
= dev
;
2697 net
->ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(dev
);
2698 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2699 net
->ipv6
.ip6_prohibit_entry
->dst
.dev
= dev
;
2700 net
->ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(dev
);
2701 net
->ipv6
.ip6_blk_hole_entry
->dst
.dev
= dev
;
2702 net
->ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(dev
);
2713 #ifdef CONFIG_PROC_FS
2724 static int rt6_info_route(struct rt6_info
*rt
, void *p_arg
)
2726 struct seq_file
*m
= p_arg
;
2728 seq_printf(m
, "%pi6 %02x ", &rt
->rt6i_dst
.addr
, rt
->rt6i_dst
.plen
);
2730 #ifdef CONFIG_IPV6_SUBTREES
2731 seq_printf(m
, "%pi6 %02x ", &rt
->rt6i_src
.addr
, rt
->rt6i_src
.plen
);
2733 seq_puts(m
, "00000000000000000000000000000000 00 ");
2735 if (rt
->rt6i_flags
& RTF_GATEWAY
) {
2736 seq_printf(m
, "%pi6", &rt
->rt6i_gateway
);
2738 seq_puts(m
, "00000000000000000000000000000000");
2740 seq_printf(m
, " %08x %08x %08x %08x %8s\n",
2741 rt
->rt6i_metric
, atomic_read(&rt
->dst
.__refcnt
),
2742 rt
->dst
.__use
, rt
->rt6i_flags
,
2743 rt
->dst
.dev
? rt
->dst
.dev
->name
: "");
2747 static int ipv6_route_show(struct seq_file
*m
, void *v
)
2749 struct net
*net
= (struct net
*)m
->private;
2750 fib6_clean_all_ro(net
, rt6_info_route
, 0, m
);
2754 static int ipv6_route_open(struct inode
*inode
, struct file
*file
)
2756 return single_open_net(inode
, file
, ipv6_route_show
);
2759 static const struct file_operations ipv6_route_proc_fops
= {
2760 .owner
= THIS_MODULE
,
2761 .open
= ipv6_route_open
,
2763 .llseek
= seq_lseek
,
2764 .release
= single_release_net
,
2767 static int rt6_stats_seq_show(struct seq_file
*seq
, void *v
)
2769 struct net
*net
= (struct net
*)seq
->private;
2770 seq_printf(seq
, "%04x %04x %04x %04x %04x %04x %04x\n",
2771 net
->ipv6
.rt6_stats
->fib_nodes
,
2772 net
->ipv6
.rt6_stats
->fib_route_nodes
,
2773 net
->ipv6
.rt6_stats
->fib_rt_alloc
,
2774 net
->ipv6
.rt6_stats
->fib_rt_entries
,
2775 net
->ipv6
.rt6_stats
->fib_rt_cache
,
2776 dst_entries_get_slow(&net
->ipv6
.ip6_dst_ops
),
2777 net
->ipv6
.rt6_stats
->fib_discarded_routes
);
2782 static int rt6_stats_seq_open(struct inode
*inode
, struct file
*file
)
2784 return single_open_net(inode
, file
, rt6_stats_seq_show
);
2787 static const struct file_operations rt6_stats_seq_fops
= {
2788 .owner
= THIS_MODULE
,
2789 .open
= rt6_stats_seq_open
,
2791 .llseek
= seq_lseek
,
2792 .release
= single_release_net
,
2794 #endif /* CONFIG_PROC_FS */
2796 #ifdef CONFIG_SYSCTL
2799 int ipv6_sysctl_rtcache_flush(struct ctl_table
*ctl
, int write
,
2800 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
2807 net
= (struct net
*)ctl
->extra1
;
2808 delay
= net
->ipv6
.sysctl
.flush_delay
;
2809 proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
2810 fib6_run_gc(delay
<= 0 ? ~0UL : (unsigned long)delay
, net
);
2814 struct ctl_table ipv6_route_table_template
[] = {
2816 .procname
= "flush",
2817 .data
= &init_net
.ipv6
.sysctl
.flush_delay
,
2818 .maxlen
= sizeof(int),
2820 .proc_handler
= ipv6_sysctl_rtcache_flush
2823 .procname
= "gc_thresh",
2824 .data
= &ip6_dst_ops_template
.gc_thresh
,
2825 .maxlen
= sizeof(int),
2827 .proc_handler
= proc_dointvec
,
2830 .procname
= "max_size",
2831 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_max_size
,
2832 .maxlen
= sizeof(int),
2834 .proc_handler
= proc_dointvec
,
2837 .procname
= "gc_min_interval",
2838 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
2839 .maxlen
= sizeof(int),
2841 .proc_handler
= proc_dointvec_jiffies
,
2844 .procname
= "gc_timeout",
2845 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_timeout
,
2846 .maxlen
= sizeof(int),
2848 .proc_handler
= proc_dointvec_jiffies
,
2851 .procname
= "gc_interval",
2852 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_interval
,
2853 .maxlen
= sizeof(int),
2855 .proc_handler
= proc_dointvec_jiffies
,
2858 .procname
= "gc_elasticity",
2859 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_elasticity
,
2860 .maxlen
= sizeof(int),
2862 .proc_handler
= proc_dointvec
,
2865 .procname
= "mtu_expires",
2866 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_mtu_expires
,
2867 .maxlen
= sizeof(int),
2869 .proc_handler
= proc_dointvec_jiffies
,
2872 .procname
= "min_adv_mss",
2873 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_min_advmss
,
2874 .maxlen
= sizeof(int),
2876 .proc_handler
= proc_dointvec
,
2879 .procname
= "gc_min_interval_ms",
2880 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
2881 .maxlen
= sizeof(int),
2883 .proc_handler
= proc_dointvec_ms_jiffies
,
2888 struct ctl_table
* __net_init
ipv6_route_sysctl_init(struct net
*net
)
2890 struct ctl_table
*table
;
2892 table
= kmemdup(ipv6_route_table_template
,
2893 sizeof(ipv6_route_table_template
),
2897 table
[0].data
= &net
->ipv6
.sysctl
.flush_delay
;
2898 table
[0].extra1
= net
;
2899 table
[1].data
= &net
->ipv6
.ip6_dst_ops
.gc_thresh
;
2900 table
[2].data
= &net
->ipv6
.sysctl
.ip6_rt_max_size
;
2901 table
[3].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
2902 table
[4].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
2903 table
[5].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
2904 table
[6].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
2905 table
[7].data
= &net
->ipv6
.sysctl
.ip6_rt_mtu_expires
;
2906 table
[8].data
= &net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
2907 table
[9].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
2909 /* Don't export sysctls to unprivileged users */
2910 if (net
->user_ns
!= &init_user_ns
)
2911 table
[0].procname
= NULL
;
2918 static int __net_init
ip6_route_net_init(struct net
*net
)
2922 memcpy(&net
->ipv6
.ip6_dst_ops
, &ip6_dst_ops_template
,
2923 sizeof(net
->ipv6
.ip6_dst_ops
));
2925 if (dst_entries_init(&net
->ipv6
.ip6_dst_ops
) < 0)
2926 goto out_ip6_dst_ops
;
2928 net
->ipv6
.ip6_null_entry
= kmemdup(&ip6_null_entry_template
,
2929 sizeof(*net
->ipv6
.ip6_null_entry
),
2931 if (!net
->ipv6
.ip6_null_entry
)
2932 goto out_ip6_dst_entries
;
2933 net
->ipv6
.ip6_null_entry
->dst
.path
=
2934 (struct dst_entry
*)net
->ipv6
.ip6_null_entry
;
2935 net
->ipv6
.ip6_null_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
2936 dst_init_metrics(&net
->ipv6
.ip6_null_entry
->dst
,
2937 ip6_template_metrics
, true);
2939 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2940 net
->ipv6
.ip6_prohibit_entry
= kmemdup(&ip6_prohibit_entry_template
,
2941 sizeof(*net
->ipv6
.ip6_prohibit_entry
),
2943 if (!net
->ipv6
.ip6_prohibit_entry
)
2944 goto out_ip6_null_entry
;
2945 net
->ipv6
.ip6_prohibit_entry
->dst
.path
=
2946 (struct dst_entry
*)net
->ipv6
.ip6_prohibit_entry
;
2947 net
->ipv6
.ip6_prohibit_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
2948 dst_init_metrics(&net
->ipv6
.ip6_prohibit_entry
->dst
,
2949 ip6_template_metrics
, true);
2951 net
->ipv6
.ip6_blk_hole_entry
= kmemdup(&ip6_blk_hole_entry_template
,
2952 sizeof(*net
->ipv6
.ip6_blk_hole_entry
),
2954 if (!net
->ipv6
.ip6_blk_hole_entry
)
2955 goto out_ip6_prohibit_entry
;
2956 net
->ipv6
.ip6_blk_hole_entry
->dst
.path
=
2957 (struct dst_entry
*)net
->ipv6
.ip6_blk_hole_entry
;
2958 net
->ipv6
.ip6_blk_hole_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
2959 dst_init_metrics(&net
->ipv6
.ip6_blk_hole_entry
->dst
,
2960 ip6_template_metrics
, true);
2963 net
->ipv6
.sysctl
.flush_delay
= 0;
2964 net
->ipv6
.sysctl
.ip6_rt_max_size
= 4096;
2965 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
= HZ
/ 2;
2966 net
->ipv6
.sysctl
.ip6_rt_gc_timeout
= 60*HZ
;
2967 net
->ipv6
.sysctl
.ip6_rt_gc_interval
= 30*HZ
;
2968 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
= 9;
2969 net
->ipv6
.sysctl
.ip6_rt_mtu_expires
= 10*60*HZ
;
2970 net
->ipv6
.sysctl
.ip6_rt_min_advmss
= IPV6_MIN_MTU
- 20 - 40;
2972 net
->ipv6
.ip6_rt_gc_expire
= 30*HZ
;
2978 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2979 out_ip6_prohibit_entry
:
2980 kfree(net
->ipv6
.ip6_prohibit_entry
);
2982 kfree(net
->ipv6
.ip6_null_entry
);
2984 out_ip6_dst_entries
:
2985 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
2990 static void __net_exit
ip6_route_net_exit(struct net
*net
)
2992 kfree(net
->ipv6
.ip6_null_entry
);
2993 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2994 kfree(net
->ipv6
.ip6_prohibit_entry
);
2995 kfree(net
->ipv6
.ip6_blk_hole_entry
);
2997 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
3000 static int __net_init
ip6_route_net_init_late(struct net
*net
)
3002 #ifdef CONFIG_PROC_FS
3003 proc_create("ipv6_route", 0, net
->proc_net
, &ipv6_route_proc_fops
);
3004 proc_create("rt6_stats", S_IRUGO
, net
->proc_net
, &rt6_stats_seq_fops
);
3009 static void __net_exit
ip6_route_net_exit_late(struct net
*net
)
3011 #ifdef CONFIG_PROC_FS
3012 remove_proc_entry("ipv6_route", net
->proc_net
);
3013 remove_proc_entry("rt6_stats", net
->proc_net
);
3017 static struct pernet_operations ip6_route_net_ops
= {
3018 .init
= ip6_route_net_init
,
3019 .exit
= ip6_route_net_exit
,
3022 static int __net_init
ipv6_inetpeer_init(struct net
*net
)
3024 struct inet_peer_base
*bp
= kmalloc(sizeof(*bp
), GFP_KERNEL
);
3028 inet_peer_base_init(bp
);
3029 net
->ipv6
.peers
= bp
;
3033 static void __net_exit
ipv6_inetpeer_exit(struct net
*net
)
3035 struct inet_peer_base
*bp
= net
->ipv6
.peers
;
3037 net
->ipv6
.peers
= NULL
;
3038 inetpeer_invalidate_tree(bp
);
3042 static struct pernet_operations ipv6_inetpeer_ops
= {
3043 .init
= ipv6_inetpeer_init
,
3044 .exit
= ipv6_inetpeer_exit
,
3047 static struct pernet_operations ip6_route_net_late_ops
= {
3048 .init
= ip6_route_net_init_late
,
3049 .exit
= ip6_route_net_exit_late
,
3052 static struct notifier_block ip6_route_dev_notifier
= {
3053 .notifier_call
= ip6_route_dev_notify
,
3057 int __init
ip6_route_init(void)
3062 ip6_dst_ops_template
.kmem_cachep
=
3063 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info
), 0,
3064 SLAB_HWCACHE_ALIGN
, NULL
);
3065 if (!ip6_dst_ops_template
.kmem_cachep
)
3068 ret
= dst_entries_init(&ip6_dst_blackhole_ops
);
3070 goto out_kmem_cache
;
3072 ret
= register_pernet_subsys(&ipv6_inetpeer_ops
);
3074 goto out_dst_entries
;
3076 ret
= register_pernet_subsys(&ip6_route_net_ops
);
3078 goto out_register_inetpeer
;
3080 ip6_dst_blackhole_ops
.kmem_cachep
= ip6_dst_ops_template
.kmem_cachep
;
3082 /* Registering of the loopback is done before this portion of code,
3083 * the loopback reference in rt6_info will not be taken, do it
3084 * manually for init_net */
3085 init_net
.ipv6
.ip6_null_entry
->dst
.dev
= init_net
.loopback_dev
;
3086 init_net
.ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3087 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3088 init_net
.ipv6
.ip6_prohibit_entry
->dst
.dev
= init_net
.loopback_dev
;
3089 init_net
.ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3090 init_net
.ipv6
.ip6_blk_hole_entry
->dst
.dev
= init_net
.loopback_dev
;
3091 init_net
.ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3095 goto out_register_subsys
;
3101 ret
= fib6_rules_init();
3105 ret
= register_pernet_subsys(&ip6_route_net_late_ops
);
3107 goto fib6_rules_init
;
3110 if (__rtnl_register(PF_INET6
, RTM_NEWROUTE
, inet6_rtm_newroute
, NULL
, NULL
) ||
3111 __rtnl_register(PF_INET6
, RTM_DELROUTE
, inet6_rtm_delroute
, NULL
, NULL
) ||
3112 __rtnl_register(PF_INET6
, RTM_GETROUTE
, inet6_rtm_getroute
, NULL
, NULL
))
3113 goto out_register_late_subsys
;
3115 ret
= register_netdevice_notifier(&ip6_route_dev_notifier
);
3117 goto out_register_late_subsys
;
3122 out_register_late_subsys
:
3123 unregister_pernet_subsys(&ip6_route_net_late_ops
);
3125 fib6_rules_cleanup();
3130 out_register_subsys
:
3131 unregister_pernet_subsys(&ip6_route_net_ops
);
3132 out_register_inetpeer
:
3133 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
3135 dst_entries_destroy(&ip6_dst_blackhole_ops
);
3137 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);
3141 void ip6_route_cleanup(void)
3143 unregister_netdevice_notifier(&ip6_route_dev_notifier
);
3144 unregister_pernet_subsys(&ip6_route_net_late_ops
);
3145 fib6_rules_cleanup();
3148 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
3149 unregister_pernet_subsys(&ip6_route_net_ops
);
3150 dst_entries_destroy(&ip6_dst_blackhole_ops
);
3151 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);