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>
69 RT6_NUD_FAIL_HARD
= -3,
70 RT6_NUD_FAIL_PROBE
= -2,
71 RT6_NUD_FAIL_DO_RR
= -1,
75 static struct rt6_info
*ip6_rt_copy(struct rt6_info
*ort
,
76 const struct in6_addr
*dest
);
77 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
);
78 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
);
79 static unsigned int ip6_mtu(const struct dst_entry
*dst
);
80 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*);
81 static void ip6_dst_destroy(struct dst_entry
*);
82 static void ip6_dst_ifdown(struct dst_entry
*,
83 struct net_device
*dev
, int how
);
84 static int ip6_dst_gc(struct dst_ops
*ops
);
86 static int ip6_pkt_discard(struct sk_buff
*skb
);
87 static int ip6_pkt_discard_out(struct sock
*sk
, struct sk_buff
*skb
);
88 static int ip6_pkt_prohibit(struct sk_buff
*skb
);
89 static int ip6_pkt_prohibit_out(struct sock
*sk
, struct sk_buff
*skb
);
90 static void ip6_link_failure(struct sk_buff
*skb
);
91 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
92 struct sk_buff
*skb
, u32 mtu
);
93 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
,
95 static int rt6_score_route(struct rt6_info
*rt
, int oif
, int strict
);
97 #ifdef CONFIG_IPV6_ROUTE_INFO
98 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
99 const struct in6_addr
*prefix
, int prefixlen
,
100 const struct in6_addr
*gwaddr
, int ifindex
,
102 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
103 const struct in6_addr
*prefix
, int prefixlen
,
104 const struct in6_addr
*gwaddr
, int ifindex
);
107 static void rt6_bind_peer(struct rt6_info
*rt
, int create
)
109 struct inet_peer_base
*base
;
110 struct inet_peer
*peer
;
112 base
= inetpeer_base_ptr(rt
->_rt6i_peer
);
116 peer
= inet_getpeer_v6(base
, &rt
->rt6i_dst
.addr
, create
);
118 if (!rt6_set_peer(rt
, peer
))
123 static struct inet_peer
*__rt6_get_peer(struct rt6_info
*rt
, int create
)
125 if (rt6_has_peer(rt
))
126 return rt6_peer_ptr(rt
);
128 rt6_bind_peer(rt
, create
);
129 return (rt6_has_peer(rt
) ? rt6_peer_ptr(rt
) : NULL
);
132 static struct inet_peer
*rt6_get_peer_create(struct rt6_info
*rt
)
134 return __rt6_get_peer(rt
, 1);
137 static u32
*ipv6_cow_metrics(struct dst_entry
*dst
, unsigned long old
)
139 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
140 struct inet_peer
*peer
;
143 if (!(rt
->dst
.flags
& DST_HOST
))
146 peer
= rt6_get_peer_create(rt
);
148 u32
*old_p
= __DST_METRICS_PTR(old
);
149 unsigned long prev
, new;
152 if (inet_metrics_new(peer
) ||
153 (old
& DST_METRICS_FORCE_OVERWRITE
))
154 memcpy(p
, old_p
, sizeof(u32
) * RTAX_MAX
);
156 new = (unsigned long) p
;
157 prev
= cmpxchg(&dst
->_metrics
, old
, new);
160 p
= __DST_METRICS_PTR(prev
);
161 if (prev
& DST_METRICS_READ_ONLY
)
168 static inline const void *choose_neigh_daddr(struct rt6_info
*rt
,
172 struct in6_addr
*p
= &rt
->rt6i_gateway
;
174 if (!ipv6_addr_any(p
))
175 return (const void *) p
;
177 return &ipv6_hdr(skb
)->daddr
;
181 static struct neighbour
*ip6_neigh_lookup(const struct dst_entry
*dst
,
185 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
188 daddr
= choose_neigh_daddr(rt
, skb
, daddr
);
189 n
= __ipv6_neigh_lookup(dst
->dev
, daddr
);
192 return neigh_create(&nd_tbl
, daddr
, dst
->dev
);
195 static struct dst_ops ip6_dst_ops_template
= {
197 .protocol
= cpu_to_be16(ETH_P_IPV6
),
200 .check
= ip6_dst_check
,
201 .default_advmss
= ip6_default_advmss
,
203 .cow_metrics
= ipv6_cow_metrics
,
204 .destroy
= ip6_dst_destroy
,
205 .ifdown
= ip6_dst_ifdown
,
206 .negative_advice
= ip6_negative_advice
,
207 .link_failure
= ip6_link_failure
,
208 .update_pmtu
= ip6_rt_update_pmtu
,
209 .redirect
= rt6_do_redirect
,
210 .local_out
= __ip6_local_out
,
211 .neigh_lookup
= ip6_neigh_lookup
,
214 static unsigned int ip6_blackhole_mtu(const struct dst_entry
*dst
)
216 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
218 return mtu
? : dst
->dev
->mtu
;
221 static void ip6_rt_blackhole_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
222 struct sk_buff
*skb
, u32 mtu
)
226 static void ip6_rt_blackhole_redirect(struct dst_entry
*dst
, struct sock
*sk
,
231 static u32
*ip6_rt_blackhole_cow_metrics(struct dst_entry
*dst
,
237 static struct dst_ops ip6_dst_blackhole_ops
= {
239 .protocol
= cpu_to_be16(ETH_P_IPV6
),
240 .destroy
= ip6_dst_destroy
,
241 .check
= ip6_dst_check
,
242 .mtu
= ip6_blackhole_mtu
,
243 .default_advmss
= ip6_default_advmss
,
244 .update_pmtu
= ip6_rt_blackhole_update_pmtu
,
245 .redirect
= ip6_rt_blackhole_redirect
,
246 .cow_metrics
= ip6_rt_blackhole_cow_metrics
,
247 .neigh_lookup
= ip6_neigh_lookup
,
250 static const u32 ip6_template_metrics
[RTAX_MAX
] = {
251 [RTAX_HOPLIMIT
- 1] = 0,
254 static const struct rt6_info ip6_null_entry_template
= {
256 .__refcnt
= ATOMIC_INIT(1),
258 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
259 .error
= -ENETUNREACH
,
260 .input
= ip6_pkt_discard
,
261 .output
= ip6_pkt_discard_out
,
263 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
264 .rt6i_protocol
= RTPROT_KERNEL
,
265 .rt6i_metric
= ~(u32
) 0,
266 .rt6i_ref
= ATOMIC_INIT(1),
269 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
271 static const struct rt6_info ip6_prohibit_entry_template
= {
273 .__refcnt
= ATOMIC_INIT(1),
275 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
277 .input
= ip6_pkt_prohibit
,
278 .output
= ip6_pkt_prohibit_out
,
280 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
281 .rt6i_protocol
= RTPROT_KERNEL
,
282 .rt6i_metric
= ~(u32
) 0,
283 .rt6i_ref
= ATOMIC_INIT(1),
286 static const struct rt6_info ip6_blk_hole_entry_template
= {
288 .__refcnt
= ATOMIC_INIT(1),
290 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
292 .input
= dst_discard
,
293 .output
= dst_discard_sk
,
295 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
296 .rt6i_protocol
= RTPROT_KERNEL
,
297 .rt6i_metric
= ~(u32
) 0,
298 .rt6i_ref
= ATOMIC_INIT(1),
303 /* allocate dst with ip6_dst_ops */
304 static inline struct rt6_info
*ip6_dst_alloc(struct net
*net
,
305 struct net_device
*dev
,
307 struct fib6_table
*table
)
309 struct rt6_info
*rt
= dst_alloc(&net
->ipv6
.ip6_dst_ops
, dev
,
310 0, DST_OBSOLETE_FORCE_CHK
, flags
);
313 struct dst_entry
*dst
= &rt
->dst
;
315 memset(dst
+ 1, 0, sizeof(*rt
) - sizeof(*dst
));
316 rt6_init_peer(rt
, table
? &table
->tb6_peers
: net
->ipv6
.peers
);
317 rt
->rt6i_genid
= rt_genid_ipv6(net
);
318 INIT_LIST_HEAD(&rt
->rt6i_siblings
);
323 static void ip6_dst_destroy(struct dst_entry
*dst
)
325 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
326 struct inet6_dev
*idev
= rt
->rt6i_idev
;
327 struct dst_entry
*from
= dst
->from
;
329 if (!(rt
->dst
.flags
& DST_HOST
))
330 dst_destroy_metrics_generic(dst
);
333 rt
->rt6i_idev
= NULL
;
340 if (rt6_has_peer(rt
)) {
341 struct inet_peer
*peer
= rt6_peer_ptr(rt
);
346 static void ip6_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
,
349 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
350 struct inet6_dev
*idev
= rt
->rt6i_idev
;
351 struct net_device
*loopback_dev
=
352 dev_net(dev
)->loopback_dev
;
354 if (dev
!= loopback_dev
) {
355 if (idev
&& idev
->dev
== dev
) {
356 struct inet6_dev
*loopback_idev
=
357 in6_dev_get(loopback_dev
);
359 rt
->rt6i_idev
= loopback_idev
;
366 static bool rt6_check_expired(const struct rt6_info
*rt
)
368 if (rt
->rt6i_flags
& RTF_EXPIRES
) {
369 if (time_after(jiffies
, rt
->dst
.expires
))
371 } else if (rt
->dst
.from
) {
372 return rt6_check_expired((struct rt6_info
*) rt
->dst
.from
);
377 /* Multipath route selection:
378 * Hash based function using packet header and flowlabel.
379 * Adapted from fib_info_hashfn()
381 static int rt6_info_hash_nhsfn(unsigned int candidate_count
,
382 const struct flowi6
*fl6
)
384 unsigned int val
= fl6
->flowi6_proto
;
386 val
^= ipv6_addr_hash(&fl6
->daddr
);
387 val
^= ipv6_addr_hash(&fl6
->saddr
);
389 /* Work only if this not encapsulated */
390 switch (fl6
->flowi6_proto
) {
394 val
^= (__force u16
)fl6
->fl6_sport
;
395 val
^= (__force u16
)fl6
->fl6_dport
;
399 val
^= (__force u16
)fl6
->fl6_icmp_type
;
400 val
^= (__force u16
)fl6
->fl6_icmp_code
;
403 /* RFC6438 recommands to use flowlabel */
404 val
^= (__force u32
)fl6
->flowlabel
;
406 /* Perhaps, we need to tune, this function? */
407 val
= val
^ (val
>> 7) ^ (val
>> 12);
408 return val
% candidate_count
;
411 static struct rt6_info
*rt6_multipath_select(struct rt6_info
*match
,
412 struct flowi6
*fl6
, int oif
,
415 struct rt6_info
*sibling
, *next_sibling
;
418 route_choosen
= rt6_info_hash_nhsfn(match
->rt6i_nsiblings
+ 1, fl6
);
419 /* Don't change the route, if route_choosen == 0
420 * (siblings does not include ourself)
423 list_for_each_entry_safe(sibling
, next_sibling
,
424 &match
->rt6i_siblings
, rt6i_siblings
) {
426 if (route_choosen
== 0) {
427 if (rt6_score_route(sibling
, oif
, strict
) < 0)
437 * Route lookup. Any table->tb6_lock is implied.
440 static inline struct rt6_info
*rt6_device_match(struct net
*net
,
442 const struct in6_addr
*saddr
,
446 struct rt6_info
*local
= NULL
;
447 struct rt6_info
*sprt
;
449 if (!oif
&& ipv6_addr_any(saddr
))
452 for (sprt
= rt
; sprt
; sprt
= sprt
->dst
.rt6_next
) {
453 struct net_device
*dev
= sprt
->dst
.dev
;
456 if (dev
->ifindex
== oif
)
458 if (dev
->flags
& IFF_LOOPBACK
) {
459 if (!sprt
->rt6i_idev
||
460 sprt
->rt6i_idev
->dev
->ifindex
!= oif
) {
461 if (flags
& RT6_LOOKUP_F_IFACE
&& oif
)
463 if (local
&& (!oif
||
464 local
->rt6i_idev
->dev
->ifindex
== oif
))
470 if (ipv6_chk_addr(net
, saddr
, dev
,
471 flags
& RT6_LOOKUP_F_IFACE
))
480 if (flags
& RT6_LOOKUP_F_IFACE
)
481 return net
->ipv6
.ip6_null_entry
;
487 #ifdef CONFIG_IPV6_ROUTER_PREF
488 struct __rt6_probe_work
{
489 struct work_struct work
;
490 struct in6_addr target
;
491 struct net_device
*dev
;
494 static void rt6_probe_deferred(struct work_struct
*w
)
496 struct in6_addr mcaddr
;
497 struct __rt6_probe_work
*work
=
498 container_of(w
, struct __rt6_probe_work
, work
);
500 addrconf_addr_solict_mult(&work
->target
, &mcaddr
);
501 ndisc_send_ns(work
->dev
, NULL
, &work
->target
, &mcaddr
, NULL
);
506 static void rt6_probe(struct rt6_info
*rt
)
508 struct neighbour
*neigh
;
510 * Okay, this does not seem to be appropriate
511 * for now, however, we need to check if it
512 * is really so; aka Router Reachability Probing.
514 * Router Reachability Probe MUST be rate-limited
515 * to no more than one per minute.
517 if (!rt
|| !(rt
->rt6i_flags
& RTF_GATEWAY
))
520 neigh
= __ipv6_neigh_lookup_noref(rt
->dst
.dev
, &rt
->rt6i_gateway
);
522 write_lock(&neigh
->lock
);
523 if (neigh
->nud_state
& NUD_VALID
)
528 time_after(jiffies
, neigh
->updated
+ rt
->rt6i_idev
->cnf
.rtr_probe_interval
)) {
529 struct __rt6_probe_work
*work
;
531 work
= kmalloc(sizeof(*work
), GFP_ATOMIC
);
534 __neigh_set_probe_once(neigh
);
537 write_unlock(&neigh
->lock
);
540 INIT_WORK(&work
->work
, rt6_probe_deferred
);
541 work
->target
= rt
->rt6i_gateway
;
542 dev_hold(rt
->dst
.dev
);
543 work
->dev
= rt
->dst
.dev
;
544 schedule_work(&work
->work
);
548 write_unlock(&neigh
->lock
);
550 rcu_read_unlock_bh();
553 static inline void rt6_probe(struct rt6_info
*rt
)
559 * Default Router Selection (RFC 2461 6.3.6)
561 static inline int rt6_check_dev(struct rt6_info
*rt
, int oif
)
563 struct net_device
*dev
= rt
->dst
.dev
;
564 if (!oif
|| dev
->ifindex
== oif
)
566 if ((dev
->flags
& IFF_LOOPBACK
) &&
567 rt
->rt6i_idev
&& rt
->rt6i_idev
->dev
->ifindex
== oif
)
572 static inline enum rt6_nud_state
rt6_check_neigh(struct rt6_info
*rt
)
574 struct neighbour
*neigh
;
575 enum rt6_nud_state ret
= RT6_NUD_FAIL_HARD
;
577 if (rt
->rt6i_flags
& RTF_NONEXTHOP
||
578 !(rt
->rt6i_flags
& RTF_GATEWAY
))
579 return RT6_NUD_SUCCEED
;
582 neigh
= __ipv6_neigh_lookup_noref(rt
->dst
.dev
, &rt
->rt6i_gateway
);
584 read_lock(&neigh
->lock
);
585 if (neigh
->nud_state
& NUD_VALID
)
586 ret
= RT6_NUD_SUCCEED
;
587 #ifdef CONFIG_IPV6_ROUTER_PREF
588 else if (!(neigh
->nud_state
& NUD_FAILED
))
589 ret
= RT6_NUD_SUCCEED
;
591 ret
= RT6_NUD_FAIL_PROBE
;
593 read_unlock(&neigh
->lock
);
595 ret
= IS_ENABLED(CONFIG_IPV6_ROUTER_PREF
) ?
596 RT6_NUD_SUCCEED
: RT6_NUD_FAIL_DO_RR
;
598 rcu_read_unlock_bh();
603 static int rt6_score_route(struct rt6_info
*rt
, int oif
,
608 m
= rt6_check_dev(rt
, oif
);
609 if (!m
&& (strict
& RT6_LOOKUP_F_IFACE
))
610 return RT6_NUD_FAIL_HARD
;
611 #ifdef CONFIG_IPV6_ROUTER_PREF
612 m
|= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt
->rt6i_flags
)) << 2;
614 if (strict
& RT6_LOOKUP_F_REACHABLE
) {
615 int n
= rt6_check_neigh(rt
);
622 static struct rt6_info
*find_match(struct rt6_info
*rt
, int oif
, int strict
,
623 int *mpri
, struct rt6_info
*match
,
627 bool match_do_rr
= false;
629 if (rt6_check_expired(rt
))
632 m
= rt6_score_route(rt
, oif
, strict
);
633 if (m
== RT6_NUD_FAIL_DO_RR
) {
635 m
= 0; /* lowest valid score */
636 } else if (m
== RT6_NUD_FAIL_HARD
) {
640 if (strict
& RT6_LOOKUP_F_REACHABLE
)
643 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
645 *do_rr
= match_do_rr
;
653 static struct rt6_info
*find_rr_leaf(struct fib6_node
*fn
,
654 struct rt6_info
*rr_head
,
655 u32 metric
, int oif
, int strict
,
658 struct rt6_info
*rt
, *match
;
662 for (rt
= rr_head
; rt
&& rt
->rt6i_metric
== metric
;
663 rt
= rt
->dst
.rt6_next
)
664 match
= find_match(rt
, oif
, strict
, &mpri
, match
, do_rr
);
665 for (rt
= fn
->leaf
; rt
&& rt
!= rr_head
&& rt
->rt6i_metric
== metric
;
666 rt
= rt
->dst
.rt6_next
)
667 match
= find_match(rt
, oif
, strict
, &mpri
, match
, do_rr
);
672 static struct rt6_info
*rt6_select(struct fib6_node
*fn
, int oif
, int strict
)
674 struct rt6_info
*match
, *rt0
;
680 fn
->rr_ptr
= rt0
= fn
->leaf
;
682 match
= find_rr_leaf(fn
, rt0
, rt0
->rt6i_metric
, oif
, strict
,
686 struct rt6_info
*next
= rt0
->dst
.rt6_next
;
688 /* no entries matched; do round-robin */
689 if (!next
|| next
->rt6i_metric
!= rt0
->rt6i_metric
)
696 net
= dev_net(rt0
->dst
.dev
);
697 return match
? match
: net
->ipv6
.ip6_null_entry
;
700 #ifdef CONFIG_IPV6_ROUTE_INFO
701 int rt6_route_rcv(struct net_device
*dev
, u8
*opt
, int len
,
702 const struct in6_addr
*gwaddr
)
704 struct net
*net
= dev_net(dev
);
705 struct route_info
*rinfo
= (struct route_info
*) opt
;
706 struct in6_addr prefix_buf
, *prefix
;
708 unsigned long lifetime
;
711 if (len
< sizeof(struct route_info
)) {
715 /* Sanity check for prefix_len and length */
716 if (rinfo
->length
> 3) {
718 } else if (rinfo
->prefix_len
> 128) {
720 } else if (rinfo
->prefix_len
> 64) {
721 if (rinfo
->length
< 2) {
724 } else if (rinfo
->prefix_len
> 0) {
725 if (rinfo
->length
< 1) {
730 pref
= rinfo
->route_pref
;
731 if (pref
== ICMPV6_ROUTER_PREF_INVALID
)
734 lifetime
= addrconf_timeout_fixup(ntohl(rinfo
->lifetime
), HZ
);
736 if (rinfo
->length
== 3)
737 prefix
= (struct in6_addr
*)rinfo
->prefix
;
739 /* this function is safe */
740 ipv6_addr_prefix(&prefix_buf
,
741 (struct in6_addr
*)rinfo
->prefix
,
743 prefix
= &prefix_buf
;
746 if (rinfo
->prefix_len
== 0)
747 rt
= rt6_get_dflt_router(gwaddr
, dev
);
749 rt
= rt6_get_route_info(net
, prefix
, rinfo
->prefix_len
,
750 gwaddr
, dev
->ifindex
);
752 if (rt
&& !lifetime
) {
758 rt
= rt6_add_route_info(net
, prefix
, rinfo
->prefix_len
, gwaddr
, dev
->ifindex
,
761 rt
->rt6i_flags
= RTF_ROUTEINFO
|
762 (rt
->rt6i_flags
& ~RTF_PREF_MASK
) | RTF_PREF(pref
);
765 if (!addrconf_finite_timeout(lifetime
))
766 rt6_clean_expires(rt
);
768 rt6_set_expires(rt
, jiffies
+ HZ
* lifetime
);
776 #define BACKTRACK(__net, saddr) \
778 if (rt == __net->ipv6.ip6_null_entry) { \
779 struct fib6_node *pn; \
781 if (fn->fn_flags & RTN_TL_ROOT) \
784 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
785 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
788 if (fn->fn_flags & RTN_RTINFO) \
794 static struct rt6_info
*ip6_pol_route_lookup(struct net
*net
,
795 struct fib6_table
*table
,
796 struct flowi6
*fl6
, int flags
)
798 struct fib6_node
*fn
;
801 read_lock_bh(&table
->tb6_lock
);
802 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
805 rt
= rt6_device_match(net
, rt
, &fl6
->saddr
, fl6
->flowi6_oif
, flags
);
806 if (rt
->rt6i_nsiblings
&& fl6
->flowi6_oif
== 0)
807 rt
= rt6_multipath_select(rt
, fl6
, fl6
->flowi6_oif
, flags
);
808 BACKTRACK(net
, &fl6
->saddr
);
810 dst_use(&rt
->dst
, jiffies
);
811 read_unlock_bh(&table
->tb6_lock
);
816 struct dst_entry
*ip6_route_lookup(struct net
*net
, struct flowi6
*fl6
,
819 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_lookup
);
821 EXPORT_SYMBOL_GPL(ip6_route_lookup
);
823 struct rt6_info
*rt6_lookup(struct net
*net
, const struct in6_addr
*daddr
,
824 const struct in6_addr
*saddr
, int oif
, int strict
)
826 struct flowi6 fl6
= {
830 struct dst_entry
*dst
;
831 int flags
= strict
? RT6_LOOKUP_F_IFACE
: 0;
834 memcpy(&fl6
.saddr
, saddr
, sizeof(*saddr
));
835 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
838 dst
= fib6_rule_lookup(net
, &fl6
, flags
, ip6_pol_route_lookup
);
840 return (struct rt6_info
*) dst
;
846 EXPORT_SYMBOL(rt6_lookup
);
848 /* ip6_ins_rt is called with FREE table->tb6_lock.
849 It takes new route entry, the addition fails by any reason the
850 route is freed. In any case, if caller does not hold it, it may
854 static int __ip6_ins_rt(struct rt6_info
*rt
, struct nl_info
*info
,
855 struct nlattr
*mx
, int mx_len
)
858 struct fib6_table
*table
;
860 table
= rt
->rt6i_table
;
861 write_lock_bh(&table
->tb6_lock
);
862 err
= fib6_add(&table
->tb6_root
, rt
, info
, mx
, mx_len
);
863 write_unlock_bh(&table
->tb6_lock
);
868 int ip6_ins_rt(struct rt6_info
*rt
)
870 struct nl_info info
= {
871 .nl_net
= dev_net(rt
->dst
.dev
),
873 return __ip6_ins_rt(rt
, &info
, NULL
, 0);
876 static struct rt6_info
*rt6_alloc_cow(struct rt6_info
*ort
,
877 const struct in6_addr
*daddr
,
878 const struct in6_addr
*saddr
)
886 rt
= ip6_rt_copy(ort
, daddr
);
889 if (ort
->rt6i_dst
.plen
!= 128 &&
890 ipv6_addr_equal(&ort
->rt6i_dst
.addr
, daddr
))
891 rt
->rt6i_flags
|= RTF_ANYCAST
;
893 rt
->rt6i_flags
|= RTF_CACHE
;
895 #ifdef CONFIG_IPV6_SUBTREES
896 if (rt
->rt6i_src
.plen
&& saddr
) {
897 rt
->rt6i_src
.addr
= *saddr
;
898 rt
->rt6i_src
.plen
= 128;
906 static struct rt6_info
*rt6_alloc_clone(struct rt6_info
*ort
,
907 const struct in6_addr
*daddr
)
909 struct rt6_info
*rt
= ip6_rt_copy(ort
, daddr
);
912 rt
->rt6i_flags
|= RTF_CACHE
;
916 static struct rt6_info
*ip6_pol_route(struct net
*net
, struct fib6_table
*table
, int oif
,
917 struct flowi6
*fl6
, int flags
)
919 struct fib6_node
*fn
;
920 struct rt6_info
*rt
, *nrt
;
924 int reachable
= net
->ipv6
.devconf_all
->forwarding
? 0 : RT6_LOOKUP_F_REACHABLE
;
926 strict
|= flags
& RT6_LOOKUP_F_IFACE
;
929 read_lock_bh(&table
->tb6_lock
);
932 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
935 rt
= rt6_select(fn
, oif
, strict
| reachable
);
936 if (rt
->rt6i_nsiblings
)
937 rt
= rt6_multipath_select(rt
, fl6
, oif
, strict
| reachable
);
938 BACKTRACK(net
, &fl6
->saddr
);
939 if (rt
== net
->ipv6
.ip6_null_entry
||
940 rt
->rt6i_flags
& RTF_CACHE
)
944 read_unlock_bh(&table
->tb6_lock
);
946 if (!(rt
->rt6i_flags
& (RTF_NONEXTHOP
| RTF_GATEWAY
)))
947 nrt
= rt6_alloc_cow(rt
, &fl6
->daddr
, &fl6
->saddr
);
948 else if (!(rt
->dst
.flags
& DST_HOST
))
949 nrt
= rt6_alloc_clone(rt
, &fl6
->daddr
);
954 rt
= nrt
? : net
->ipv6
.ip6_null_entry
;
958 err
= ip6_ins_rt(nrt
);
967 * Race condition! In the gap, when table->tb6_lock was
968 * released someone could insert this route. Relookup.
979 read_unlock_bh(&table
->tb6_lock
);
981 rt
->dst
.lastuse
= jiffies
;
987 static struct rt6_info
*ip6_pol_route_input(struct net
*net
, struct fib6_table
*table
,
988 struct flowi6
*fl6
, int flags
)
990 return ip6_pol_route(net
, table
, fl6
->flowi6_iif
, fl6
, flags
);
993 static struct dst_entry
*ip6_route_input_lookup(struct net
*net
,
994 struct net_device
*dev
,
995 struct flowi6
*fl6
, int flags
)
997 if (rt6_need_strict(&fl6
->daddr
) && dev
->type
!= ARPHRD_PIMREG
)
998 flags
|= RT6_LOOKUP_F_IFACE
;
1000 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_input
);
1003 void ip6_route_input(struct sk_buff
*skb
)
1005 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
1006 struct net
*net
= dev_net(skb
->dev
);
1007 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
1008 struct flowi6 fl6
= {
1009 .flowi6_iif
= skb
->dev
->ifindex
,
1010 .daddr
= iph
->daddr
,
1011 .saddr
= iph
->saddr
,
1012 .flowlabel
= ip6_flowinfo(iph
),
1013 .flowi6_mark
= skb
->mark
,
1014 .flowi6_proto
= iph
->nexthdr
,
1017 skb_dst_set(skb
, ip6_route_input_lookup(net
, skb
->dev
, &fl6
, flags
));
1020 static struct rt6_info
*ip6_pol_route_output(struct net
*net
, struct fib6_table
*table
,
1021 struct flowi6
*fl6
, int flags
)
1023 return ip6_pol_route(net
, table
, fl6
->flowi6_oif
, fl6
, flags
);
1026 struct dst_entry
*ip6_route_output(struct net
*net
, const struct sock
*sk
,
1031 fl6
->flowi6_iif
= LOOPBACK_IFINDEX
;
1033 if ((sk
&& sk
->sk_bound_dev_if
) || rt6_need_strict(&fl6
->daddr
))
1034 flags
|= RT6_LOOKUP_F_IFACE
;
1036 if (!ipv6_addr_any(&fl6
->saddr
))
1037 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
1039 flags
|= rt6_srcprefs2flags(inet6_sk(sk
)->srcprefs
);
1041 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_output
);
1043 EXPORT_SYMBOL(ip6_route_output
);
1045 struct dst_entry
*ip6_blackhole_route(struct net
*net
, struct dst_entry
*dst_orig
)
1047 struct rt6_info
*rt
, *ort
= (struct rt6_info
*) dst_orig
;
1048 struct dst_entry
*new = NULL
;
1050 rt
= dst_alloc(&ip6_dst_blackhole_ops
, ort
->dst
.dev
, 1, DST_OBSOLETE_NONE
, 0);
1054 memset(new + 1, 0, sizeof(*rt
) - sizeof(*new));
1055 rt6_init_peer(rt
, net
->ipv6
.peers
);
1058 new->input
= dst_discard
;
1059 new->output
= dst_discard_sk
;
1061 if (dst_metrics_read_only(&ort
->dst
))
1062 new->_metrics
= ort
->dst
._metrics
;
1064 dst_copy_metrics(new, &ort
->dst
);
1065 rt
->rt6i_idev
= ort
->rt6i_idev
;
1067 in6_dev_hold(rt
->rt6i_idev
);
1069 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
1070 rt
->rt6i_flags
= ort
->rt6i_flags
;
1071 rt
->rt6i_metric
= 0;
1073 memcpy(&rt
->rt6i_dst
, &ort
->rt6i_dst
, sizeof(struct rt6key
));
1074 #ifdef CONFIG_IPV6_SUBTREES
1075 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1081 dst_release(dst_orig
);
1082 return new ? new : ERR_PTR(-ENOMEM
);
1086 * Destination cache support functions
1089 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
)
1091 struct rt6_info
*rt
;
1093 rt
= (struct rt6_info
*) dst
;
1095 /* All IPV6 dsts are created with ->obsolete set to the value
1096 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1097 * into this function always.
1099 if (rt
->rt6i_genid
!= rt_genid_ipv6(dev_net(rt
->dst
.dev
)))
1102 if (!rt
->rt6i_node
|| (rt
->rt6i_node
->fn_sernum
!= cookie
))
1105 if (rt6_check_expired(rt
))
1111 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*dst
)
1113 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1116 if (rt
->rt6i_flags
& RTF_CACHE
) {
1117 if (rt6_check_expired(rt
)) {
1129 static void ip6_link_failure(struct sk_buff
*skb
)
1131 struct rt6_info
*rt
;
1133 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, ICMPV6_ADDR_UNREACH
, 0);
1135 rt
= (struct rt6_info
*) skb_dst(skb
);
1137 if (rt
->rt6i_flags
& RTF_CACHE
) {
1141 } else if (rt
->rt6i_node
&& (rt
->rt6i_flags
& RTF_DEFAULT
)) {
1142 rt
->rt6i_node
->fn_sernum
= -1;
1147 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
1148 struct sk_buff
*skb
, u32 mtu
)
1150 struct rt6_info
*rt6
= (struct rt6_info
*)dst
;
1153 if (mtu
< dst_mtu(dst
) && rt6
->rt6i_dst
.plen
== 128) {
1154 struct net
*net
= dev_net(dst
->dev
);
1156 rt6
->rt6i_flags
|= RTF_MODIFIED
;
1157 if (mtu
< IPV6_MIN_MTU
) {
1158 u32 features
= dst_metric(dst
, RTAX_FEATURES
);
1160 features
|= RTAX_FEATURE_ALLFRAG
;
1161 dst_metric_set(dst
, RTAX_FEATURES
, features
);
1163 dst_metric_set(dst
, RTAX_MTU
, mtu
);
1164 rt6_update_expires(rt6
, net
->ipv6
.sysctl
.ip6_rt_mtu_expires
);
1168 void ip6_update_pmtu(struct sk_buff
*skb
, struct net
*net
, __be32 mtu
,
1171 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1172 struct dst_entry
*dst
;
1175 memset(&fl6
, 0, sizeof(fl6
));
1176 fl6
.flowi6_oif
= oif
;
1177 fl6
.flowi6_mark
= mark
? mark
: IP6_REPLY_MARK(net
, skb
->mark
);
1178 fl6
.daddr
= iph
->daddr
;
1179 fl6
.saddr
= iph
->saddr
;
1180 fl6
.flowlabel
= ip6_flowinfo(iph
);
1182 dst
= ip6_route_output(net
, NULL
, &fl6
);
1184 ip6_rt_update_pmtu(dst
, NULL
, skb
, ntohl(mtu
));
1187 EXPORT_SYMBOL_GPL(ip6_update_pmtu
);
1189 void ip6_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, __be32 mtu
)
1191 ip6_update_pmtu(skb
, sock_net(sk
), mtu
,
1192 sk
->sk_bound_dev_if
, sk
->sk_mark
);
1194 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu
);
1196 /* Handle redirects */
1197 struct ip6rd_flowi
{
1199 struct in6_addr gateway
;
1202 static struct rt6_info
*__ip6_route_redirect(struct net
*net
,
1203 struct fib6_table
*table
,
1207 struct ip6rd_flowi
*rdfl
= (struct ip6rd_flowi
*)fl6
;
1208 struct rt6_info
*rt
;
1209 struct fib6_node
*fn
;
1211 /* Get the "current" route for this destination and
1212 * check if the redirect has come from approriate router.
1214 * RFC 4861 specifies that redirects should only be
1215 * accepted if they come from the nexthop to the target.
1216 * Due to the way the routes are chosen, this notion
1217 * is a bit fuzzy and one might need to check all possible
1221 read_lock_bh(&table
->tb6_lock
);
1222 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
1224 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1225 if (rt6_check_expired(rt
))
1229 if (!(rt
->rt6i_flags
& RTF_GATEWAY
))
1231 if (fl6
->flowi6_oif
!= rt
->dst
.dev
->ifindex
)
1233 if (!ipv6_addr_equal(&rdfl
->gateway
, &rt
->rt6i_gateway
))
1239 rt
= net
->ipv6
.ip6_null_entry
;
1240 else if (rt
->dst
.error
) {
1241 rt
= net
->ipv6
.ip6_null_entry
;
1244 BACKTRACK(net
, &fl6
->saddr
);
1248 read_unlock_bh(&table
->tb6_lock
);
1253 static struct dst_entry
*ip6_route_redirect(struct net
*net
,
1254 const struct flowi6
*fl6
,
1255 const struct in6_addr
*gateway
)
1257 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
1258 struct ip6rd_flowi rdfl
;
1261 rdfl
.gateway
= *gateway
;
1263 return fib6_rule_lookup(net
, &rdfl
.fl6
,
1264 flags
, __ip6_route_redirect
);
1267 void ip6_redirect(struct sk_buff
*skb
, struct net
*net
, int oif
, u32 mark
)
1269 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1270 struct dst_entry
*dst
;
1273 memset(&fl6
, 0, sizeof(fl6
));
1274 fl6
.flowi6_iif
= LOOPBACK_IFINDEX
;
1275 fl6
.flowi6_oif
= oif
;
1276 fl6
.flowi6_mark
= mark
;
1277 fl6
.daddr
= iph
->daddr
;
1278 fl6
.saddr
= iph
->saddr
;
1279 fl6
.flowlabel
= ip6_flowinfo(iph
);
1281 dst
= ip6_route_redirect(net
, &fl6
, &ipv6_hdr(skb
)->saddr
);
1282 rt6_do_redirect(dst
, NULL
, skb
);
1285 EXPORT_SYMBOL_GPL(ip6_redirect
);
1287 void ip6_redirect_no_header(struct sk_buff
*skb
, struct net
*net
, int oif
,
1290 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
1291 const struct rd_msg
*msg
= (struct rd_msg
*)icmp6_hdr(skb
);
1292 struct dst_entry
*dst
;
1295 memset(&fl6
, 0, sizeof(fl6
));
1296 fl6
.flowi6_iif
= LOOPBACK_IFINDEX
;
1297 fl6
.flowi6_oif
= oif
;
1298 fl6
.flowi6_mark
= mark
;
1299 fl6
.daddr
= msg
->dest
;
1300 fl6
.saddr
= iph
->daddr
;
1302 dst
= ip6_route_redirect(net
, &fl6
, &iph
->saddr
);
1303 rt6_do_redirect(dst
, NULL
, skb
);
1307 void ip6_sk_redirect(struct sk_buff
*skb
, struct sock
*sk
)
1309 ip6_redirect(skb
, sock_net(sk
), sk
->sk_bound_dev_if
, sk
->sk_mark
);
1311 EXPORT_SYMBOL_GPL(ip6_sk_redirect
);
1313 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
)
1315 struct net_device
*dev
= dst
->dev
;
1316 unsigned int mtu
= dst_mtu(dst
);
1317 struct net
*net
= dev_net(dev
);
1319 mtu
-= sizeof(struct ipv6hdr
) + sizeof(struct tcphdr
);
1321 if (mtu
< net
->ipv6
.sysctl
.ip6_rt_min_advmss
)
1322 mtu
= net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
1325 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1326 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1327 * IPV6_MAXPLEN is also valid and means: "any MSS,
1328 * rely only on pmtu discovery"
1330 if (mtu
> IPV6_MAXPLEN
- sizeof(struct tcphdr
))
1335 static unsigned int ip6_mtu(const struct dst_entry
*dst
)
1337 struct inet6_dev
*idev
;
1338 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
1346 idev
= __in6_dev_get(dst
->dev
);
1348 mtu
= idev
->cnf
.mtu6
;
1352 return min_t(unsigned int, mtu
, IP6_MAX_MTU
);
1355 static struct dst_entry
*icmp6_dst_gc_list
;
1356 static DEFINE_SPINLOCK(icmp6_dst_lock
);
1358 struct dst_entry
*icmp6_dst_alloc(struct net_device
*dev
,
1361 struct dst_entry
*dst
;
1362 struct rt6_info
*rt
;
1363 struct inet6_dev
*idev
= in6_dev_get(dev
);
1364 struct net
*net
= dev_net(dev
);
1366 if (unlikely(!idev
))
1367 return ERR_PTR(-ENODEV
);
1369 rt
= ip6_dst_alloc(net
, dev
, 0, NULL
);
1370 if (unlikely(!rt
)) {
1372 dst
= ERR_PTR(-ENOMEM
);
1376 rt
->dst
.flags
|= DST_HOST
;
1377 rt
->dst
.output
= ip6_output
;
1378 atomic_set(&rt
->dst
.__refcnt
, 1);
1379 rt
->rt6i_gateway
= fl6
->daddr
;
1380 rt
->rt6i_dst
.addr
= fl6
->daddr
;
1381 rt
->rt6i_dst
.plen
= 128;
1382 rt
->rt6i_idev
= idev
;
1383 dst_metric_set(&rt
->dst
, RTAX_HOPLIMIT
, 0);
1385 spin_lock_bh(&icmp6_dst_lock
);
1386 rt
->dst
.next
= icmp6_dst_gc_list
;
1387 icmp6_dst_gc_list
= &rt
->dst
;
1388 spin_unlock_bh(&icmp6_dst_lock
);
1390 fib6_force_start_gc(net
);
1392 dst
= xfrm_lookup(net
, &rt
->dst
, flowi6_to_flowi(fl6
), NULL
, 0);
1398 int icmp6_dst_gc(void)
1400 struct dst_entry
*dst
, **pprev
;
1403 spin_lock_bh(&icmp6_dst_lock
);
1404 pprev
= &icmp6_dst_gc_list
;
1406 while ((dst
= *pprev
) != NULL
) {
1407 if (!atomic_read(&dst
->__refcnt
)) {
1416 spin_unlock_bh(&icmp6_dst_lock
);
1421 static void icmp6_clean_all(int (*func
)(struct rt6_info
*rt
, void *arg
),
1424 struct dst_entry
*dst
, **pprev
;
1426 spin_lock_bh(&icmp6_dst_lock
);
1427 pprev
= &icmp6_dst_gc_list
;
1428 while ((dst
= *pprev
) != NULL
) {
1429 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1430 if (func(rt
, arg
)) {
1437 spin_unlock_bh(&icmp6_dst_lock
);
1440 static int ip6_dst_gc(struct dst_ops
*ops
)
1442 struct net
*net
= container_of(ops
, struct net
, ipv6
.ip6_dst_ops
);
1443 int rt_min_interval
= net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
1444 int rt_max_size
= net
->ipv6
.sysctl
.ip6_rt_max_size
;
1445 int rt_elasticity
= net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
1446 int rt_gc_timeout
= net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
1447 unsigned long rt_last_gc
= net
->ipv6
.ip6_rt_last_gc
;
1450 entries
= dst_entries_get_fast(ops
);
1451 if (time_after(rt_last_gc
+ rt_min_interval
, jiffies
) &&
1452 entries
<= rt_max_size
)
1455 net
->ipv6
.ip6_rt_gc_expire
++;
1456 fib6_run_gc(net
->ipv6
.ip6_rt_gc_expire
, net
, true);
1457 entries
= dst_entries_get_slow(ops
);
1458 if (entries
< ops
->gc_thresh
)
1459 net
->ipv6
.ip6_rt_gc_expire
= rt_gc_timeout
>>1;
1461 net
->ipv6
.ip6_rt_gc_expire
-= net
->ipv6
.ip6_rt_gc_expire
>>rt_elasticity
;
1462 return entries
> rt_max_size
;
1469 int ip6_route_add(struct fib6_config
*cfg
)
1472 struct net
*net
= cfg
->fc_nlinfo
.nl_net
;
1473 struct rt6_info
*rt
= NULL
;
1474 struct net_device
*dev
= NULL
;
1475 struct inet6_dev
*idev
= NULL
;
1476 struct fib6_table
*table
;
1479 if (cfg
->fc_dst_len
> 128 || cfg
->fc_src_len
> 128)
1481 #ifndef CONFIG_IPV6_SUBTREES
1482 if (cfg
->fc_src_len
)
1485 if (cfg
->fc_ifindex
) {
1487 dev
= dev_get_by_index(net
, cfg
->fc_ifindex
);
1490 idev
= in6_dev_get(dev
);
1495 if (cfg
->fc_metric
== 0)
1496 cfg
->fc_metric
= IP6_RT_PRIO_USER
;
1499 if (cfg
->fc_nlinfo
.nlh
&&
1500 !(cfg
->fc_nlinfo
.nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
1501 table
= fib6_get_table(net
, cfg
->fc_table
);
1503 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1504 table
= fib6_new_table(net
, cfg
->fc_table
);
1507 table
= fib6_new_table(net
, cfg
->fc_table
);
1513 rt
= ip6_dst_alloc(net
, NULL
, (cfg
->fc_flags
& RTF_ADDRCONF
) ? 0 : DST_NOCOUNT
, table
);
1520 if (cfg
->fc_flags
& RTF_EXPIRES
)
1521 rt6_set_expires(rt
, jiffies
+
1522 clock_t_to_jiffies(cfg
->fc_expires
));
1524 rt6_clean_expires(rt
);
1526 if (cfg
->fc_protocol
== RTPROT_UNSPEC
)
1527 cfg
->fc_protocol
= RTPROT_BOOT
;
1528 rt
->rt6i_protocol
= cfg
->fc_protocol
;
1530 addr_type
= ipv6_addr_type(&cfg
->fc_dst
);
1532 if (addr_type
& IPV6_ADDR_MULTICAST
)
1533 rt
->dst
.input
= ip6_mc_input
;
1534 else if (cfg
->fc_flags
& RTF_LOCAL
)
1535 rt
->dst
.input
= ip6_input
;
1537 rt
->dst
.input
= ip6_forward
;
1539 rt
->dst
.output
= ip6_output
;
1541 ipv6_addr_prefix(&rt
->rt6i_dst
.addr
, &cfg
->fc_dst
, cfg
->fc_dst_len
);
1542 rt
->rt6i_dst
.plen
= cfg
->fc_dst_len
;
1543 if (rt
->rt6i_dst
.plen
== 128) {
1544 rt
->dst
.flags
|= DST_HOST
;
1545 dst_metrics_set_force_overwrite(&rt
->dst
);
1548 #ifdef CONFIG_IPV6_SUBTREES
1549 ipv6_addr_prefix(&rt
->rt6i_src
.addr
, &cfg
->fc_src
, cfg
->fc_src_len
);
1550 rt
->rt6i_src
.plen
= cfg
->fc_src_len
;
1553 rt
->rt6i_metric
= cfg
->fc_metric
;
1555 /* We cannot add true routes via loopback here,
1556 they would result in kernel looping; promote them to reject routes
1558 if ((cfg
->fc_flags
& RTF_REJECT
) ||
1559 (dev
&& (dev
->flags
& IFF_LOOPBACK
) &&
1560 !(addr_type
& IPV6_ADDR_LOOPBACK
) &&
1561 !(cfg
->fc_flags
& RTF_LOCAL
))) {
1562 /* hold loopback dev/idev if we haven't done so. */
1563 if (dev
!= net
->loopback_dev
) {
1568 dev
= net
->loopback_dev
;
1570 idev
= in6_dev_get(dev
);
1576 rt
->rt6i_flags
= RTF_REJECT
|RTF_NONEXTHOP
;
1577 switch (cfg
->fc_type
) {
1579 rt
->dst
.error
= -EINVAL
;
1580 rt
->dst
.output
= dst_discard_sk
;
1581 rt
->dst
.input
= dst_discard
;
1584 rt
->dst
.error
= -EACCES
;
1585 rt
->dst
.output
= ip6_pkt_prohibit_out
;
1586 rt
->dst
.input
= ip6_pkt_prohibit
;
1590 rt
->dst
.error
= (cfg
->fc_type
== RTN_THROW
) ? -EAGAIN
1592 rt
->dst
.output
= ip6_pkt_discard_out
;
1593 rt
->dst
.input
= ip6_pkt_discard
;
1599 if (cfg
->fc_flags
& RTF_GATEWAY
) {
1600 const struct in6_addr
*gw_addr
;
1603 gw_addr
= &cfg
->fc_gateway
;
1604 rt
->rt6i_gateway
= *gw_addr
;
1605 gwa_type
= ipv6_addr_type(gw_addr
);
1607 if (gwa_type
!= (IPV6_ADDR_LINKLOCAL
|IPV6_ADDR_UNICAST
)) {
1608 struct rt6_info
*grt
;
1610 /* IPv6 strictly inhibits using not link-local
1611 addresses as nexthop address.
1612 Otherwise, router will not able to send redirects.
1613 It is very good, but in some (rare!) circumstances
1614 (SIT, PtP, NBMA NOARP links) it is handy to allow
1615 some exceptions. --ANK
1618 if (!(gwa_type
& IPV6_ADDR_UNICAST
))
1621 grt
= rt6_lookup(net
, gw_addr
, NULL
, cfg
->fc_ifindex
, 1);
1623 err
= -EHOSTUNREACH
;
1627 if (dev
!= grt
->dst
.dev
) {
1633 idev
= grt
->rt6i_idev
;
1635 in6_dev_hold(grt
->rt6i_idev
);
1637 if (!(grt
->rt6i_flags
& RTF_GATEWAY
))
1645 if (!dev
|| (dev
->flags
& IFF_LOOPBACK
))
1653 if (!ipv6_addr_any(&cfg
->fc_prefsrc
)) {
1654 if (!ipv6_chk_addr(net
, &cfg
->fc_prefsrc
, dev
, 0)) {
1658 rt
->rt6i_prefsrc
.addr
= cfg
->fc_prefsrc
;
1659 rt
->rt6i_prefsrc
.plen
= 128;
1661 rt
->rt6i_prefsrc
.plen
= 0;
1663 rt
->rt6i_flags
= cfg
->fc_flags
;
1667 rt
->rt6i_idev
= idev
;
1668 rt
->rt6i_table
= table
;
1670 cfg
->fc_nlinfo
.nl_net
= dev_net(dev
);
1672 return __ip6_ins_rt(rt
, &cfg
->fc_nlinfo
, cfg
->fc_mx
, cfg
->fc_mx_len
);
1684 static int __ip6_del_rt(struct rt6_info
*rt
, struct nl_info
*info
)
1687 struct fib6_table
*table
;
1688 struct net
*net
= dev_net(rt
->dst
.dev
);
1690 if (rt
== net
->ipv6
.ip6_null_entry
) {
1695 table
= rt
->rt6i_table
;
1696 write_lock_bh(&table
->tb6_lock
);
1697 err
= fib6_del(rt
, info
);
1698 write_unlock_bh(&table
->tb6_lock
);
1705 int ip6_del_rt(struct rt6_info
*rt
)
1707 struct nl_info info
= {
1708 .nl_net
= dev_net(rt
->dst
.dev
),
1710 return __ip6_del_rt(rt
, &info
);
1713 static int ip6_route_del(struct fib6_config
*cfg
)
1715 struct fib6_table
*table
;
1716 struct fib6_node
*fn
;
1717 struct rt6_info
*rt
;
1720 table
= fib6_get_table(cfg
->fc_nlinfo
.nl_net
, cfg
->fc_table
);
1724 read_lock_bh(&table
->tb6_lock
);
1726 fn
= fib6_locate(&table
->tb6_root
,
1727 &cfg
->fc_dst
, cfg
->fc_dst_len
,
1728 &cfg
->fc_src
, cfg
->fc_src_len
);
1731 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1732 if (cfg
->fc_ifindex
&&
1734 rt
->dst
.dev
->ifindex
!= cfg
->fc_ifindex
))
1736 if (cfg
->fc_flags
& RTF_GATEWAY
&&
1737 !ipv6_addr_equal(&cfg
->fc_gateway
, &rt
->rt6i_gateway
))
1739 if (cfg
->fc_metric
&& cfg
->fc_metric
!= rt
->rt6i_metric
)
1742 read_unlock_bh(&table
->tb6_lock
);
1744 return __ip6_del_rt(rt
, &cfg
->fc_nlinfo
);
1747 read_unlock_bh(&table
->tb6_lock
);
1752 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
, struct sk_buff
*skb
)
1754 struct net
*net
= dev_net(skb
->dev
);
1755 struct netevent_redirect netevent
;
1756 struct rt6_info
*rt
, *nrt
= NULL
;
1757 struct ndisc_options ndopts
;
1758 struct inet6_dev
*in6_dev
;
1759 struct neighbour
*neigh
;
1761 int optlen
, on_link
;
1764 optlen
= skb_tail_pointer(skb
) - skb_transport_header(skb
);
1765 optlen
-= sizeof(*msg
);
1768 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1772 msg
= (struct rd_msg
*)icmp6_hdr(skb
);
1774 if (ipv6_addr_is_multicast(&msg
->dest
)) {
1775 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1780 if (ipv6_addr_equal(&msg
->dest
, &msg
->target
)) {
1782 } else if (ipv6_addr_type(&msg
->target
) !=
1783 (IPV6_ADDR_UNICAST
|IPV6_ADDR_LINKLOCAL
)) {
1784 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1788 in6_dev
= __in6_dev_get(skb
->dev
);
1791 if (in6_dev
->cnf
.forwarding
|| !in6_dev
->cnf
.accept_redirects
)
1795 * The IP source address of the Redirect MUST be the same as the current
1796 * first-hop router for the specified ICMP Destination Address.
1799 if (!ndisc_parse_options(msg
->opt
, optlen
, &ndopts
)) {
1800 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1805 if (ndopts
.nd_opts_tgt_lladdr
) {
1806 lladdr
= ndisc_opt_addr_data(ndopts
.nd_opts_tgt_lladdr
,
1809 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1814 rt
= (struct rt6_info
*) dst
;
1815 if (rt
== net
->ipv6
.ip6_null_entry
) {
1816 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1820 /* Redirect received -> path was valid.
1821 * Look, redirects are sent only in response to data packets,
1822 * so that this nexthop apparently is reachable. --ANK
1824 dst_confirm(&rt
->dst
);
1826 neigh
= __neigh_lookup(&nd_tbl
, &msg
->target
, skb
->dev
, 1);
1831 * We have finally decided to accept it.
1834 neigh_update(neigh
, lladdr
, NUD_STALE
,
1835 NEIGH_UPDATE_F_WEAK_OVERRIDE
|
1836 NEIGH_UPDATE_F_OVERRIDE
|
1837 (on_link
? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER
|
1838 NEIGH_UPDATE_F_ISROUTER
))
1841 nrt
= ip6_rt_copy(rt
, &msg
->dest
);
1845 nrt
->rt6i_flags
= RTF_GATEWAY
|RTF_UP
|RTF_DYNAMIC
|RTF_CACHE
;
1847 nrt
->rt6i_flags
&= ~RTF_GATEWAY
;
1849 nrt
->rt6i_gateway
= *(struct in6_addr
*)neigh
->primary_key
;
1851 if (ip6_ins_rt(nrt
))
1854 netevent
.old
= &rt
->dst
;
1855 netevent
.new = &nrt
->dst
;
1856 netevent
.daddr
= &msg
->dest
;
1857 netevent
.neigh
= neigh
;
1858 call_netevent_notifiers(NETEVENT_REDIRECT
, &netevent
);
1860 if (rt
->rt6i_flags
& RTF_CACHE
) {
1861 rt
= (struct rt6_info
*) dst_clone(&rt
->dst
);
1866 neigh_release(neigh
);
1870 * Misc support functions
1873 static struct rt6_info
*ip6_rt_copy(struct rt6_info
*ort
,
1874 const struct in6_addr
*dest
)
1876 struct net
*net
= dev_net(ort
->dst
.dev
);
1877 struct rt6_info
*rt
= ip6_dst_alloc(net
, ort
->dst
.dev
, 0,
1881 rt
->dst
.input
= ort
->dst
.input
;
1882 rt
->dst
.output
= ort
->dst
.output
;
1883 rt
->dst
.flags
|= DST_HOST
;
1885 rt
->rt6i_dst
.addr
= *dest
;
1886 rt
->rt6i_dst
.plen
= 128;
1887 dst_copy_metrics(&rt
->dst
, &ort
->dst
);
1888 rt
->dst
.error
= ort
->dst
.error
;
1889 rt
->rt6i_idev
= ort
->rt6i_idev
;
1891 in6_dev_hold(rt
->rt6i_idev
);
1892 rt
->dst
.lastuse
= jiffies
;
1894 if (ort
->rt6i_flags
& RTF_GATEWAY
)
1895 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
1897 rt
->rt6i_gateway
= *dest
;
1898 rt
->rt6i_flags
= ort
->rt6i_flags
;
1899 rt6_set_from(rt
, ort
);
1900 rt
->rt6i_metric
= 0;
1902 #ifdef CONFIG_IPV6_SUBTREES
1903 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1905 memcpy(&rt
->rt6i_prefsrc
, &ort
->rt6i_prefsrc
, sizeof(struct rt6key
));
1906 rt
->rt6i_table
= ort
->rt6i_table
;
1911 #ifdef CONFIG_IPV6_ROUTE_INFO
1912 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
1913 const struct in6_addr
*prefix
, int prefixlen
,
1914 const struct in6_addr
*gwaddr
, int ifindex
)
1916 struct fib6_node
*fn
;
1917 struct rt6_info
*rt
= NULL
;
1918 struct fib6_table
*table
;
1920 table
= fib6_get_table(net
, RT6_TABLE_INFO
);
1924 read_lock_bh(&table
->tb6_lock
);
1925 fn
= fib6_locate(&table
->tb6_root
, prefix
, prefixlen
, NULL
, 0);
1929 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1930 if (rt
->dst
.dev
->ifindex
!= ifindex
)
1932 if ((rt
->rt6i_flags
& (RTF_ROUTEINFO
|RTF_GATEWAY
)) != (RTF_ROUTEINFO
|RTF_GATEWAY
))
1934 if (!ipv6_addr_equal(&rt
->rt6i_gateway
, gwaddr
))
1940 read_unlock_bh(&table
->tb6_lock
);
1944 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
1945 const struct in6_addr
*prefix
, int prefixlen
,
1946 const struct in6_addr
*gwaddr
, int ifindex
,
1949 struct fib6_config cfg
= {
1950 .fc_table
= RT6_TABLE_INFO
,
1951 .fc_metric
= IP6_RT_PRIO_USER
,
1952 .fc_ifindex
= ifindex
,
1953 .fc_dst_len
= prefixlen
,
1954 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_ROUTEINFO
|
1955 RTF_UP
| RTF_PREF(pref
),
1956 .fc_nlinfo
.portid
= 0,
1957 .fc_nlinfo
.nlh
= NULL
,
1958 .fc_nlinfo
.nl_net
= net
,
1961 cfg
.fc_dst
= *prefix
;
1962 cfg
.fc_gateway
= *gwaddr
;
1964 /* We should treat it as a default route if prefix length is 0. */
1966 cfg
.fc_flags
|= RTF_DEFAULT
;
1968 ip6_route_add(&cfg
);
1970 return rt6_get_route_info(net
, prefix
, prefixlen
, gwaddr
, ifindex
);
1974 struct rt6_info
*rt6_get_dflt_router(const struct in6_addr
*addr
, struct net_device
*dev
)
1976 struct rt6_info
*rt
;
1977 struct fib6_table
*table
;
1979 table
= fib6_get_table(dev_net(dev
), RT6_TABLE_DFLT
);
1983 read_lock_bh(&table
->tb6_lock
);
1984 for (rt
= table
->tb6_root
.leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1985 if (dev
== rt
->dst
.dev
&&
1986 ((rt
->rt6i_flags
& (RTF_ADDRCONF
| RTF_DEFAULT
)) == (RTF_ADDRCONF
| RTF_DEFAULT
)) &&
1987 ipv6_addr_equal(&rt
->rt6i_gateway
, addr
))
1992 read_unlock_bh(&table
->tb6_lock
);
1996 struct rt6_info
*rt6_add_dflt_router(const struct in6_addr
*gwaddr
,
1997 struct net_device
*dev
,
2000 struct fib6_config cfg
= {
2001 .fc_table
= RT6_TABLE_DFLT
,
2002 .fc_metric
= IP6_RT_PRIO_USER
,
2003 .fc_ifindex
= dev
->ifindex
,
2004 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_DEFAULT
|
2005 RTF_UP
| RTF_EXPIRES
| RTF_PREF(pref
),
2006 .fc_nlinfo
.portid
= 0,
2007 .fc_nlinfo
.nlh
= NULL
,
2008 .fc_nlinfo
.nl_net
= dev_net(dev
),
2011 cfg
.fc_gateway
= *gwaddr
;
2013 ip6_route_add(&cfg
);
2015 return rt6_get_dflt_router(gwaddr
, dev
);
2018 void rt6_purge_dflt_routers(struct net
*net
)
2020 struct rt6_info
*rt
;
2021 struct fib6_table
*table
;
2023 /* NOTE: Keep consistent with rt6_get_dflt_router */
2024 table
= fib6_get_table(net
, RT6_TABLE_DFLT
);
2029 read_lock_bh(&table
->tb6_lock
);
2030 for (rt
= table
->tb6_root
.leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
2031 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
) &&
2032 (!rt
->rt6i_idev
|| rt
->rt6i_idev
->cnf
.accept_ra
!= 2)) {
2034 read_unlock_bh(&table
->tb6_lock
);
2039 read_unlock_bh(&table
->tb6_lock
);
2042 static void rtmsg_to_fib6_config(struct net
*net
,
2043 struct in6_rtmsg
*rtmsg
,
2044 struct fib6_config
*cfg
)
2046 memset(cfg
, 0, sizeof(*cfg
));
2048 cfg
->fc_table
= RT6_TABLE_MAIN
;
2049 cfg
->fc_ifindex
= rtmsg
->rtmsg_ifindex
;
2050 cfg
->fc_metric
= rtmsg
->rtmsg_metric
;
2051 cfg
->fc_expires
= rtmsg
->rtmsg_info
;
2052 cfg
->fc_dst_len
= rtmsg
->rtmsg_dst_len
;
2053 cfg
->fc_src_len
= rtmsg
->rtmsg_src_len
;
2054 cfg
->fc_flags
= rtmsg
->rtmsg_flags
;
2056 cfg
->fc_nlinfo
.nl_net
= net
;
2058 cfg
->fc_dst
= rtmsg
->rtmsg_dst
;
2059 cfg
->fc_src
= rtmsg
->rtmsg_src
;
2060 cfg
->fc_gateway
= rtmsg
->rtmsg_gateway
;
2063 int ipv6_route_ioctl(struct net
*net
, unsigned int cmd
, void __user
*arg
)
2065 struct fib6_config cfg
;
2066 struct in6_rtmsg rtmsg
;
2070 case SIOCADDRT
: /* Add a route */
2071 case SIOCDELRT
: /* Delete a route */
2072 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
2074 err
= copy_from_user(&rtmsg
, arg
,
2075 sizeof(struct in6_rtmsg
));
2079 rtmsg_to_fib6_config(net
, &rtmsg
, &cfg
);
2084 err
= ip6_route_add(&cfg
);
2087 err
= ip6_route_del(&cfg
);
2101 * Drop the packet on the floor
2104 static int ip6_pkt_drop(struct sk_buff
*skb
, u8 code
, int ipstats_mib_noroutes
)
2107 struct dst_entry
*dst
= skb_dst(skb
);
2108 switch (ipstats_mib_noroutes
) {
2109 case IPSTATS_MIB_INNOROUTES
:
2110 type
= ipv6_addr_type(&ipv6_hdr(skb
)->daddr
);
2111 if (type
== IPV6_ADDR_ANY
) {
2112 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2113 IPSTATS_MIB_INADDRERRORS
);
2117 case IPSTATS_MIB_OUTNOROUTES
:
2118 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2119 ipstats_mib_noroutes
);
2122 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, code
, 0);
2127 static int ip6_pkt_discard(struct sk_buff
*skb
)
2129 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_INNOROUTES
);
2132 static int ip6_pkt_discard_out(struct sock
*sk
, struct sk_buff
*skb
)
2134 skb
->dev
= skb_dst(skb
)->dev
;
2135 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_OUTNOROUTES
);
2138 static int ip6_pkt_prohibit(struct sk_buff
*skb
)
2140 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_INNOROUTES
);
2143 static int ip6_pkt_prohibit_out(struct sock
*sk
, struct sk_buff
*skb
)
2145 skb
->dev
= skb_dst(skb
)->dev
;
2146 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_OUTNOROUTES
);
2150 * Allocate a dst for local (unicast / anycast) address.
2153 struct rt6_info
*addrconf_dst_alloc(struct inet6_dev
*idev
,
2154 const struct in6_addr
*addr
,
2157 struct net
*net
= dev_net(idev
->dev
);
2158 struct rt6_info
*rt
= ip6_dst_alloc(net
, net
->loopback_dev
,
2161 return ERR_PTR(-ENOMEM
);
2165 rt
->dst
.flags
|= DST_HOST
;
2166 rt
->dst
.input
= ip6_input
;
2167 rt
->dst
.output
= ip6_output
;
2168 rt
->rt6i_idev
= idev
;
2170 rt
->rt6i_flags
= RTF_UP
| RTF_NONEXTHOP
;
2172 rt
->rt6i_flags
|= RTF_ANYCAST
;
2174 rt
->rt6i_flags
|= RTF_LOCAL
;
2176 rt
->rt6i_gateway
= *addr
;
2177 rt
->rt6i_dst
.addr
= *addr
;
2178 rt
->rt6i_dst
.plen
= 128;
2179 rt
->rt6i_table
= fib6_get_table(net
, RT6_TABLE_LOCAL
);
2181 atomic_set(&rt
->dst
.__refcnt
, 1);
2186 int ip6_route_get_saddr(struct net
*net
,
2187 struct rt6_info
*rt
,
2188 const struct in6_addr
*daddr
,
2190 struct in6_addr
*saddr
)
2192 struct inet6_dev
*idev
= ip6_dst_idev((struct dst_entry
*)rt
);
2194 if (rt
->rt6i_prefsrc
.plen
)
2195 *saddr
= rt
->rt6i_prefsrc
.addr
;
2197 err
= ipv6_dev_get_saddr(net
, idev
? idev
->dev
: NULL
,
2198 daddr
, prefs
, saddr
);
2202 /* remove deleted ip from prefsrc entries */
2203 struct arg_dev_net_ip
{
2204 struct net_device
*dev
;
2206 struct in6_addr
*addr
;
2209 static int fib6_remove_prefsrc(struct rt6_info
*rt
, void *arg
)
2211 struct net_device
*dev
= ((struct arg_dev_net_ip
*)arg
)->dev
;
2212 struct net
*net
= ((struct arg_dev_net_ip
*)arg
)->net
;
2213 struct in6_addr
*addr
= ((struct arg_dev_net_ip
*)arg
)->addr
;
2215 if (((void *)rt
->dst
.dev
== dev
|| !dev
) &&
2216 rt
!= net
->ipv6
.ip6_null_entry
&&
2217 ipv6_addr_equal(addr
, &rt
->rt6i_prefsrc
.addr
)) {
2218 /* remove prefsrc entry */
2219 rt
->rt6i_prefsrc
.plen
= 0;
2224 void rt6_remove_prefsrc(struct inet6_ifaddr
*ifp
)
2226 struct net
*net
= dev_net(ifp
->idev
->dev
);
2227 struct arg_dev_net_ip adni
= {
2228 .dev
= ifp
->idev
->dev
,
2232 fib6_clean_all(net
, fib6_remove_prefsrc
, &adni
);
2235 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2236 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2238 /* Remove routers and update dst entries when gateway turn into host. */
2239 static int fib6_clean_tohost(struct rt6_info
*rt
, void *arg
)
2241 struct in6_addr
*gateway
= (struct in6_addr
*)arg
;
2243 if ((((rt
->rt6i_flags
& RTF_RA_ROUTER
) == RTF_RA_ROUTER
) ||
2244 ((rt
->rt6i_flags
& RTF_CACHE_GATEWAY
) == RTF_CACHE_GATEWAY
)) &&
2245 ipv6_addr_equal(gateway
, &rt
->rt6i_gateway
)) {
2251 void rt6_clean_tohost(struct net
*net
, struct in6_addr
*gateway
)
2253 fib6_clean_all(net
, fib6_clean_tohost
, gateway
);
2256 struct arg_dev_net
{
2257 struct net_device
*dev
;
2261 static int fib6_ifdown(struct rt6_info
*rt
, void *arg
)
2263 const struct arg_dev_net
*adn
= arg
;
2264 const struct net_device
*dev
= adn
->dev
;
2266 if ((rt
->dst
.dev
== dev
|| !dev
) &&
2267 rt
!= adn
->net
->ipv6
.ip6_null_entry
)
2273 void rt6_ifdown(struct net
*net
, struct net_device
*dev
)
2275 struct arg_dev_net adn
= {
2280 fib6_clean_all(net
, fib6_ifdown
, &adn
);
2281 icmp6_clean_all(fib6_ifdown
, &adn
);
2284 struct rt6_mtu_change_arg
{
2285 struct net_device
*dev
;
2289 static int rt6_mtu_change_route(struct rt6_info
*rt
, void *p_arg
)
2291 struct rt6_mtu_change_arg
*arg
= (struct rt6_mtu_change_arg
*) p_arg
;
2292 struct inet6_dev
*idev
;
2294 /* In IPv6 pmtu discovery is not optional,
2295 so that RTAX_MTU lock cannot disable it.
2296 We still use this lock to block changes
2297 caused by addrconf/ndisc.
2300 idev
= __in6_dev_get(arg
->dev
);
2304 /* For administrative MTU increase, there is no way to discover
2305 IPv6 PMTU increase, so PMTU increase should be updated here.
2306 Since RFC 1981 doesn't include administrative MTU increase
2307 update PMTU increase is a MUST. (i.e. jumbo frame)
2310 If new MTU is less than route PMTU, this new MTU will be the
2311 lowest MTU in the path, update the route PMTU to reflect PMTU
2312 decreases; if new MTU is greater than route PMTU, and the
2313 old MTU is the lowest MTU in the path, update the route PMTU
2314 to reflect the increase. In this case if the other nodes' MTU
2315 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2318 if (rt
->dst
.dev
== arg
->dev
&&
2319 !dst_metric_locked(&rt
->dst
, RTAX_MTU
) &&
2320 (dst_mtu(&rt
->dst
) >= arg
->mtu
||
2321 (dst_mtu(&rt
->dst
) < arg
->mtu
&&
2322 dst_mtu(&rt
->dst
) == idev
->cnf
.mtu6
))) {
2323 dst_metric_set(&rt
->dst
, RTAX_MTU
, arg
->mtu
);
2328 void rt6_mtu_change(struct net_device
*dev
, unsigned int mtu
)
2330 struct rt6_mtu_change_arg arg
= {
2335 fib6_clean_all(dev_net(dev
), rt6_mtu_change_route
, &arg
);
2338 static const struct nla_policy rtm_ipv6_policy
[RTA_MAX
+1] = {
2339 [RTA_GATEWAY
] = { .len
= sizeof(struct in6_addr
) },
2340 [RTA_OIF
] = { .type
= NLA_U32
},
2341 [RTA_IIF
] = { .type
= NLA_U32
},
2342 [RTA_PRIORITY
] = { .type
= NLA_U32
},
2343 [RTA_METRICS
] = { .type
= NLA_NESTED
},
2344 [RTA_MULTIPATH
] = { .len
= sizeof(struct rtnexthop
) },
2347 static int rtm_to_fib6_config(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
2348 struct fib6_config
*cfg
)
2351 struct nlattr
*tb
[RTA_MAX
+1];
2354 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2359 rtm
= nlmsg_data(nlh
);
2360 memset(cfg
, 0, sizeof(*cfg
));
2362 cfg
->fc_table
= rtm
->rtm_table
;
2363 cfg
->fc_dst_len
= rtm
->rtm_dst_len
;
2364 cfg
->fc_src_len
= rtm
->rtm_src_len
;
2365 cfg
->fc_flags
= RTF_UP
;
2366 cfg
->fc_protocol
= rtm
->rtm_protocol
;
2367 cfg
->fc_type
= rtm
->rtm_type
;
2369 if (rtm
->rtm_type
== RTN_UNREACHABLE
||
2370 rtm
->rtm_type
== RTN_BLACKHOLE
||
2371 rtm
->rtm_type
== RTN_PROHIBIT
||
2372 rtm
->rtm_type
== RTN_THROW
)
2373 cfg
->fc_flags
|= RTF_REJECT
;
2375 if (rtm
->rtm_type
== RTN_LOCAL
)
2376 cfg
->fc_flags
|= RTF_LOCAL
;
2378 cfg
->fc_nlinfo
.portid
= NETLINK_CB(skb
).portid
;
2379 cfg
->fc_nlinfo
.nlh
= nlh
;
2380 cfg
->fc_nlinfo
.nl_net
= sock_net(skb
->sk
);
2382 if (tb
[RTA_GATEWAY
]) {
2383 nla_memcpy(&cfg
->fc_gateway
, tb
[RTA_GATEWAY
], 16);
2384 cfg
->fc_flags
|= RTF_GATEWAY
;
2388 int plen
= (rtm
->rtm_dst_len
+ 7) >> 3;
2390 if (nla_len(tb
[RTA_DST
]) < plen
)
2393 nla_memcpy(&cfg
->fc_dst
, tb
[RTA_DST
], plen
);
2397 int plen
= (rtm
->rtm_src_len
+ 7) >> 3;
2399 if (nla_len(tb
[RTA_SRC
]) < plen
)
2402 nla_memcpy(&cfg
->fc_src
, tb
[RTA_SRC
], plen
);
2405 if (tb
[RTA_PREFSRC
])
2406 nla_memcpy(&cfg
->fc_prefsrc
, tb
[RTA_PREFSRC
], 16);
2409 cfg
->fc_ifindex
= nla_get_u32(tb
[RTA_OIF
]);
2411 if (tb
[RTA_PRIORITY
])
2412 cfg
->fc_metric
= nla_get_u32(tb
[RTA_PRIORITY
]);
2414 if (tb
[RTA_METRICS
]) {
2415 cfg
->fc_mx
= nla_data(tb
[RTA_METRICS
]);
2416 cfg
->fc_mx_len
= nla_len(tb
[RTA_METRICS
]);
2420 cfg
->fc_table
= nla_get_u32(tb
[RTA_TABLE
]);
2422 if (tb
[RTA_MULTIPATH
]) {
2423 cfg
->fc_mp
= nla_data(tb
[RTA_MULTIPATH
]);
2424 cfg
->fc_mp_len
= nla_len(tb
[RTA_MULTIPATH
]);
2432 static int ip6_route_multipath(struct fib6_config
*cfg
, int add
)
2434 struct fib6_config r_cfg
;
2435 struct rtnexthop
*rtnh
;
2438 int err
= 0, last_err
= 0;
2441 rtnh
= (struct rtnexthop
*)cfg
->fc_mp
;
2442 remaining
= cfg
->fc_mp_len
;
2444 /* Parse a Multipath Entry */
2445 while (rtnh_ok(rtnh
, remaining
)) {
2446 memcpy(&r_cfg
, cfg
, sizeof(*cfg
));
2447 if (rtnh
->rtnh_ifindex
)
2448 r_cfg
.fc_ifindex
= rtnh
->rtnh_ifindex
;
2450 attrlen
= rtnh_attrlen(rtnh
);
2452 struct nlattr
*nla
, *attrs
= rtnh_attrs(rtnh
);
2454 nla
= nla_find(attrs
, attrlen
, RTA_GATEWAY
);
2456 nla_memcpy(&r_cfg
.fc_gateway
, nla
, 16);
2457 r_cfg
.fc_flags
|= RTF_GATEWAY
;
2460 err
= add
? ip6_route_add(&r_cfg
) : ip6_route_del(&r_cfg
);
2463 /* If we are trying to remove a route, do not stop the
2464 * loop when ip6_route_del() fails (because next hop is
2465 * already gone), we should try to remove all next hops.
2468 /* If add fails, we should try to delete all
2469 * next hops that have been already added.
2475 /* Because each route is added like a single route we remove
2476 * this flag after the first nexthop (if there is a collision,
2477 * we have already fail to add the first nexthop:
2478 * fib6_add_rt2node() has reject it).
2480 cfg
->fc_nlinfo
.nlh
->nlmsg_flags
&= ~NLM_F_EXCL
;
2481 rtnh
= rtnh_next(rtnh
, &remaining
);
2487 static int inet6_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
2489 struct fib6_config cfg
;
2492 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2497 return ip6_route_multipath(&cfg
, 0);
2499 return ip6_route_del(&cfg
);
2502 static int inet6_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
2504 struct fib6_config cfg
;
2507 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2512 return ip6_route_multipath(&cfg
, 1);
2514 return ip6_route_add(&cfg
);
2517 static inline size_t rt6_nlmsg_size(void)
2519 return NLMSG_ALIGN(sizeof(struct rtmsg
))
2520 + nla_total_size(16) /* RTA_SRC */
2521 + nla_total_size(16) /* RTA_DST */
2522 + nla_total_size(16) /* RTA_GATEWAY */
2523 + nla_total_size(16) /* RTA_PREFSRC */
2524 + nla_total_size(4) /* RTA_TABLE */
2525 + nla_total_size(4) /* RTA_IIF */
2526 + nla_total_size(4) /* RTA_OIF */
2527 + nla_total_size(4) /* RTA_PRIORITY */
2528 + RTAX_MAX
* nla_total_size(4) /* RTA_METRICS */
2529 + nla_total_size(sizeof(struct rta_cacheinfo
));
2532 static int rt6_fill_node(struct net
*net
,
2533 struct sk_buff
*skb
, struct rt6_info
*rt
,
2534 struct in6_addr
*dst
, struct in6_addr
*src
,
2535 int iif
, int type
, u32 portid
, u32 seq
,
2536 int prefix
, int nowait
, unsigned int flags
)
2539 struct nlmsghdr
*nlh
;
2543 if (prefix
) { /* user wants prefix routes only */
2544 if (!(rt
->rt6i_flags
& RTF_PREFIX_RT
)) {
2545 /* success since this is not a prefix route */
2550 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*rtm
), flags
);
2554 rtm
= nlmsg_data(nlh
);
2555 rtm
->rtm_family
= AF_INET6
;
2556 rtm
->rtm_dst_len
= rt
->rt6i_dst
.plen
;
2557 rtm
->rtm_src_len
= rt
->rt6i_src
.plen
;
2560 table
= rt
->rt6i_table
->tb6_id
;
2562 table
= RT6_TABLE_UNSPEC
;
2563 rtm
->rtm_table
= table
;
2564 if (nla_put_u32(skb
, RTA_TABLE
, table
))
2565 goto nla_put_failure
;
2566 if (rt
->rt6i_flags
& RTF_REJECT
) {
2567 switch (rt
->dst
.error
) {
2569 rtm
->rtm_type
= RTN_BLACKHOLE
;
2572 rtm
->rtm_type
= RTN_PROHIBIT
;
2575 rtm
->rtm_type
= RTN_THROW
;
2578 rtm
->rtm_type
= RTN_UNREACHABLE
;
2582 else if (rt
->rt6i_flags
& RTF_LOCAL
)
2583 rtm
->rtm_type
= RTN_LOCAL
;
2584 else if (rt
->dst
.dev
&& (rt
->dst
.dev
->flags
& IFF_LOOPBACK
))
2585 rtm
->rtm_type
= RTN_LOCAL
;
2587 rtm
->rtm_type
= RTN_UNICAST
;
2589 rtm
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2590 rtm
->rtm_protocol
= rt
->rt6i_protocol
;
2591 if (rt
->rt6i_flags
& RTF_DYNAMIC
)
2592 rtm
->rtm_protocol
= RTPROT_REDIRECT
;
2593 else if (rt
->rt6i_flags
& RTF_ADDRCONF
) {
2594 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ROUTEINFO
))
2595 rtm
->rtm_protocol
= RTPROT_RA
;
2597 rtm
->rtm_protocol
= RTPROT_KERNEL
;
2600 if (rt
->rt6i_flags
& RTF_CACHE
)
2601 rtm
->rtm_flags
|= RTM_F_CLONED
;
2604 if (nla_put(skb
, RTA_DST
, 16, dst
))
2605 goto nla_put_failure
;
2606 rtm
->rtm_dst_len
= 128;
2607 } else if (rtm
->rtm_dst_len
)
2608 if (nla_put(skb
, RTA_DST
, 16, &rt
->rt6i_dst
.addr
))
2609 goto nla_put_failure
;
2610 #ifdef CONFIG_IPV6_SUBTREES
2612 if (nla_put(skb
, RTA_SRC
, 16, src
))
2613 goto nla_put_failure
;
2614 rtm
->rtm_src_len
= 128;
2615 } else if (rtm
->rtm_src_len
&&
2616 nla_put(skb
, RTA_SRC
, 16, &rt
->rt6i_src
.addr
))
2617 goto nla_put_failure
;
2620 #ifdef CONFIG_IPV6_MROUTE
2621 if (ipv6_addr_is_multicast(&rt
->rt6i_dst
.addr
)) {
2622 int err
= ip6mr_get_route(net
, skb
, rtm
, nowait
);
2627 goto nla_put_failure
;
2629 if (err
== -EMSGSIZE
)
2630 goto nla_put_failure
;
2635 if (nla_put_u32(skb
, RTA_IIF
, iif
))
2636 goto nla_put_failure
;
2638 struct in6_addr saddr_buf
;
2639 if (ip6_route_get_saddr(net
, rt
, dst
, 0, &saddr_buf
) == 0 &&
2640 nla_put(skb
, RTA_PREFSRC
, 16, &saddr_buf
))
2641 goto nla_put_failure
;
2644 if (rt
->rt6i_prefsrc
.plen
) {
2645 struct in6_addr saddr_buf
;
2646 saddr_buf
= rt
->rt6i_prefsrc
.addr
;
2647 if (nla_put(skb
, RTA_PREFSRC
, 16, &saddr_buf
))
2648 goto nla_put_failure
;
2651 if (rtnetlink_put_metrics(skb
, dst_metrics_ptr(&rt
->dst
)) < 0)
2652 goto nla_put_failure
;
2654 if (rt
->rt6i_flags
& RTF_GATEWAY
) {
2655 if (nla_put(skb
, RTA_GATEWAY
, 16, &rt
->rt6i_gateway
) < 0)
2656 goto nla_put_failure
;
2660 nla_put_u32(skb
, RTA_OIF
, rt
->dst
.dev
->ifindex
))
2661 goto nla_put_failure
;
2662 if (nla_put_u32(skb
, RTA_PRIORITY
, rt
->rt6i_metric
))
2663 goto nla_put_failure
;
2665 expires
= (rt
->rt6i_flags
& RTF_EXPIRES
) ? rt
->dst
.expires
- jiffies
: 0;
2667 if (rtnl_put_cacheinfo(skb
, &rt
->dst
, 0, expires
, rt
->dst
.error
) < 0)
2668 goto nla_put_failure
;
2670 return nlmsg_end(skb
, nlh
);
2673 nlmsg_cancel(skb
, nlh
);
2677 int rt6_dump_route(struct rt6_info
*rt
, void *p_arg
)
2679 struct rt6_rtnl_dump_arg
*arg
= (struct rt6_rtnl_dump_arg
*) p_arg
;
2682 if (nlmsg_len(arg
->cb
->nlh
) >= sizeof(struct rtmsg
)) {
2683 struct rtmsg
*rtm
= nlmsg_data(arg
->cb
->nlh
);
2684 prefix
= (rtm
->rtm_flags
& RTM_F_PREFIX
) != 0;
2688 return rt6_fill_node(arg
->net
,
2689 arg
->skb
, rt
, NULL
, NULL
, 0, RTM_NEWROUTE
,
2690 NETLINK_CB(arg
->cb
->skb
).portid
, arg
->cb
->nlh
->nlmsg_seq
,
2691 prefix
, 0, NLM_F_MULTI
);
2694 static int inet6_rtm_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
*nlh
)
2696 struct net
*net
= sock_net(in_skb
->sk
);
2697 struct nlattr
*tb
[RTA_MAX
+1];
2698 struct rt6_info
*rt
;
2699 struct sk_buff
*skb
;
2702 int err
, iif
= 0, oif
= 0;
2704 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2709 memset(&fl6
, 0, sizeof(fl6
));
2712 if (nla_len(tb
[RTA_SRC
]) < sizeof(struct in6_addr
))
2715 fl6
.saddr
= *(struct in6_addr
*)nla_data(tb
[RTA_SRC
]);
2719 if (nla_len(tb
[RTA_DST
]) < sizeof(struct in6_addr
))
2722 fl6
.daddr
= *(struct in6_addr
*)nla_data(tb
[RTA_DST
]);
2726 iif
= nla_get_u32(tb
[RTA_IIF
]);
2729 oif
= nla_get_u32(tb
[RTA_OIF
]);
2732 fl6
.flowi6_mark
= nla_get_u32(tb
[RTA_MARK
]);
2735 struct net_device
*dev
;
2738 dev
= __dev_get_by_index(net
, iif
);
2744 fl6
.flowi6_iif
= iif
;
2746 if (!ipv6_addr_any(&fl6
.saddr
))
2747 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
2749 rt
= (struct rt6_info
*)ip6_route_input_lookup(net
, dev
, &fl6
,
2752 fl6
.flowi6_oif
= oif
;
2754 rt
= (struct rt6_info
*)ip6_route_output(net
, NULL
, &fl6
);
2757 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
2764 /* Reserve room for dummy headers, this skb can pass
2765 through good chunk of routing engine.
2767 skb_reset_mac_header(skb
);
2768 skb_reserve(skb
, MAX_HEADER
+ sizeof(struct ipv6hdr
));
2770 skb_dst_set(skb
, &rt
->dst
);
2772 err
= rt6_fill_node(net
, skb
, rt
, &fl6
.daddr
, &fl6
.saddr
, iif
,
2773 RTM_NEWROUTE
, NETLINK_CB(in_skb
).portid
,
2774 nlh
->nlmsg_seq
, 0, 0, 0);
2780 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).portid
);
2785 void inet6_rt_notify(int event
, struct rt6_info
*rt
, struct nl_info
*info
)
2787 struct sk_buff
*skb
;
2788 struct net
*net
= info
->nl_net
;
2793 seq
= info
->nlh
? info
->nlh
->nlmsg_seq
: 0;
2795 skb
= nlmsg_new(rt6_nlmsg_size(), gfp_any());
2799 err
= rt6_fill_node(net
, skb
, rt
, NULL
, NULL
, 0,
2800 event
, info
->portid
, seq
, 0, 0, 0);
2802 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2803 WARN_ON(err
== -EMSGSIZE
);
2807 rtnl_notify(skb
, net
, info
->portid
, RTNLGRP_IPV6_ROUTE
,
2808 info
->nlh
, gfp_any());
2812 rtnl_set_sk_err(net
, RTNLGRP_IPV6_ROUTE
, err
);
2815 static int ip6_route_dev_notify(struct notifier_block
*this,
2816 unsigned long event
, void *ptr
)
2818 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
2819 struct net
*net
= dev_net(dev
);
2821 if (event
== NETDEV_REGISTER
&& (dev
->flags
& IFF_LOOPBACK
)) {
2822 net
->ipv6
.ip6_null_entry
->dst
.dev
= dev
;
2823 net
->ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(dev
);
2824 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2825 net
->ipv6
.ip6_prohibit_entry
->dst
.dev
= dev
;
2826 net
->ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(dev
);
2827 net
->ipv6
.ip6_blk_hole_entry
->dst
.dev
= dev
;
2828 net
->ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(dev
);
2839 #ifdef CONFIG_PROC_FS
2841 static const struct file_operations ipv6_route_proc_fops
= {
2842 .owner
= THIS_MODULE
,
2843 .open
= ipv6_route_open
,
2845 .llseek
= seq_lseek
,
2846 .release
= seq_release_net
,
2849 static int rt6_stats_seq_show(struct seq_file
*seq
, void *v
)
2851 struct net
*net
= (struct net
*)seq
->private;
2852 seq_printf(seq
, "%04x %04x %04x %04x %04x %04x %04x\n",
2853 net
->ipv6
.rt6_stats
->fib_nodes
,
2854 net
->ipv6
.rt6_stats
->fib_route_nodes
,
2855 net
->ipv6
.rt6_stats
->fib_rt_alloc
,
2856 net
->ipv6
.rt6_stats
->fib_rt_entries
,
2857 net
->ipv6
.rt6_stats
->fib_rt_cache
,
2858 dst_entries_get_slow(&net
->ipv6
.ip6_dst_ops
),
2859 net
->ipv6
.rt6_stats
->fib_discarded_routes
);
2864 static int rt6_stats_seq_open(struct inode
*inode
, struct file
*file
)
2866 return single_open_net(inode
, file
, rt6_stats_seq_show
);
2869 static const struct file_operations rt6_stats_seq_fops
= {
2870 .owner
= THIS_MODULE
,
2871 .open
= rt6_stats_seq_open
,
2873 .llseek
= seq_lseek
,
2874 .release
= single_release_net
,
2876 #endif /* CONFIG_PROC_FS */
2878 #ifdef CONFIG_SYSCTL
2881 int ipv6_sysctl_rtcache_flush(struct ctl_table
*ctl
, int write
,
2882 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
2889 net
= (struct net
*)ctl
->extra1
;
2890 delay
= net
->ipv6
.sysctl
.flush_delay
;
2891 proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
2892 fib6_run_gc(delay
<= 0 ? 0 : (unsigned long)delay
, net
, delay
> 0);
2896 struct ctl_table ipv6_route_table_template
[] = {
2898 .procname
= "flush",
2899 .data
= &init_net
.ipv6
.sysctl
.flush_delay
,
2900 .maxlen
= sizeof(int),
2902 .proc_handler
= ipv6_sysctl_rtcache_flush
2905 .procname
= "gc_thresh",
2906 .data
= &ip6_dst_ops_template
.gc_thresh
,
2907 .maxlen
= sizeof(int),
2909 .proc_handler
= proc_dointvec
,
2912 .procname
= "max_size",
2913 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_max_size
,
2914 .maxlen
= sizeof(int),
2916 .proc_handler
= proc_dointvec
,
2919 .procname
= "gc_min_interval",
2920 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
2921 .maxlen
= sizeof(int),
2923 .proc_handler
= proc_dointvec_jiffies
,
2926 .procname
= "gc_timeout",
2927 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_timeout
,
2928 .maxlen
= sizeof(int),
2930 .proc_handler
= proc_dointvec_jiffies
,
2933 .procname
= "gc_interval",
2934 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_interval
,
2935 .maxlen
= sizeof(int),
2937 .proc_handler
= proc_dointvec_jiffies
,
2940 .procname
= "gc_elasticity",
2941 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_elasticity
,
2942 .maxlen
= sizeof(int),
2944 .proc_handler
= proc_dointvec
,
2947 .procname
= "mtu_expires",
2948 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_mtu_expires
,
2949 .maxlen
= sizeof(int),
2951 .proc_handler
= proc_dointvec_jiffies
,
2954 .procname
= "min_adv_mss",
2955 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_min_advmss
,
2956 .maxlen
= sizeof(int),
2958 .proc_handler
= proc_dointvec
,
2961 .procname
= "gc_min_interval_ms",
2962 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
2963 .maxlen
= sizeof(int),
2965 .proc_handler
= proc_dointvec_ms_jiffies
,
2970 struct ctl_table
* __net_init
ipv6_route_sysctl_init(struct net
*net
)
2972 struct ctl_table
*table
;
2974 table
= kmemdup(ipv6_route_table_template
,
2975 sizeof(ipv6_route_table_template
),
2979 table
[0].data
= &net
->ipv6
.sysctl
.flush_delay
;
2980 table
[0].extra1
= net
;
2981 table
[1].data
= &net
->ipv6
.ip6_dst_ops
.gc_thresh
;
2982 table
[2].data
= &net
->ipv6
.sysctl
.ip6_rt_max_size
;
2983 table
[3].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
2984 table
[4].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
2985 table
[5].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
2986 table
[6].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
2987 table
[7].data
= &net
->ipv6
.sysctl
.ip6_rt_mtu_expires
;
2988 table
[8].data
= &net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
2989 table
[9].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
2991 /* Don't export sysctls to unprivileged users */
2992 if (net
->user_ns
!= &init_user_ns
)
2993 table
[0].procname
= NULL
;
3000 static int __net_init
ip6_route_net_init(struct net
*net
)
3004 memcpy(&net
->ipv6
.ip6_dst_ops
, &ip6_dst_ops_template
,
3005 sizeof(net
->ipv6
.ip6_dst_ops
));
3007 if (dst_entries_init(&net
->ipv6
.ip6_dst_ops
) < 0)
3008 goto out_ip6_dst_ops
;
3010 net
->ipv6
.ip6_null_entry
= kmemdup(&ip6_null_entry_template
,
3011 sizeof(*net
->ipv6
.ip6_null_entry
),
3013 if (!net
->ipv6
.ip6_null_entry
)
3014 goto out_ip6_dst_entries
;
3015 net
->ipv6
.ip6_null_entry
->dst
.path
=
3016 (struct dst_entry
*)net
->ipv6
.ip6_null_entry
;
3017 net
->ipv6
.ip6_null_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3018 dst_init_metrics(&net
->ipv6
.ip6_null_entry
->dst
,
3019 ip6_template_metrics
, true);
3021 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3022 net
->ipv6
.ip6_prohibit_entry
= kmemdup(&ip6_prohibit_entry_template
,
3023 sizeof(*net
->ipv6
.ip6_prohibit_entry
),
3025 if (!net
->ipv6
.ip6_prohibit_entry
)
3026 goto out_ip6_null_entry
;
3027 net
->ipv6
.ip6_prohibit_entry
->dst
.path
=
3028 (struct dst_entry
*)net
->ipv6
.ip6_prohibit_entry
;
3029 net
->ipv6
.ip6_prohibit_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3030 dst_init_metrics(&net
->ipv6
.ip6_prohibit_entry
->dst
,
3031 ip6_template_metrics
, true);
3033 net
->ipv6
.ip6_blk_hole_entry
= kmemdup(&ip6_blk_hole_entry_template
,
3034 sizeof(*net
->ipv6
.ip6_blk_hole_entry
),
3036 if (!net
->ipv6
.ip6_blk_hole_entry
)
3037 goto out_ip6_prohibit_entry
;
3038 net
->ipv6
.ip6_blk_hole_entry
->dst
.path
=
3039 (struct dst_entry
*)net
->ipv6
.ip6_blk_hole_entry
;
3040 net
->ipv6
.ip6_blk_hole_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3041 dst_init_metrics(&net
->ipv6
.ip6_blk_hole_entry
->dst
,
3042 ip6_template_metrics
, true);
3045 net
->ipv6
.sysctl
.flush_delay
= 0;
3046 net
->ipv6
.sysctl
.ip6_rt_max_size
= 4096;
3047 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
= HZ
/ 2;
3048 net
->ipv6
.sysctl
.ip6_rt_gc_timeout
= 60*HZ
;
3049 net
->ipv6
.sysctl
.ip6_rt_gc_interval
= 30*HZ
;
3050 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
= 9;
3051 net
->ipv6
.sysctl
.ip6_rt_mtu_expires
= 10*60*HZ
;
3052 net
->ipv6
.sysctl
.ip6_rt_min_advmss
= IPV6_MIN_MTU
- 20 - 40;
3054 net
->ipv6
.ip6_rt_gc_expire
= 30*HZ
;
3060 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3061 out_ip6_prohibit_entry
:
3062 kfree(net
->ipv6
.ip6_prohibit_entry
);
3064 kfree(net
->ipv6
.ip6_null_entry
);
3066 out_ip6_dst_entries
:
3067 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
3072 static void __net_exit
ip6_route_net_exit(struct net
*net
)
3074 kfree(net
->ipv6
.ip6_null_entry
);
3075 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3076 kfree(net
->ipv6
.ip6_prohibit_entry
);
3077 kfree(net
->ipv6
.ip6_blk_hole_entry
);
3079 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
3082 static int __net_init
ip6_route_net_init_late(struct net
*net
)
3084 #ifdef CONFIG_PROC_FS
3085 proc_create("ipv6_route", 0, net
->proc_net
, &ipv6_route_proc_fops
);
3086 proc_create("rt6_stats", S_IRUGO
, net
->proc_net
, &rt6_stats_seq_fops
);
3091 static void __net_exit
ip6_route_net_exit_late(struct net
*net
)
3093 #ifdef CONFIG_PROC_FS
3094 remove_proc_entry("ipv6_route", net
->proc_net
);
3095 remove_proc_entry("rt6_stats", net
->proc_net
);
3099 static struct pernet_operations ip6_route_net_ops
= {
3100 .init
= ip6_route_net_init
,
3101 .exit
= ip6_route_net_exit
,
3104 static int __net_init
ipv6_inetpeer_init(struct net
*net
)
3106 struct inet_peer_base
*bp
= kmalloc(sizeof(*bp
), GFP_KERNEL
);
3110 inet_peer_base_init(bp
);
3111 net
->ipv6
.peers
= bp
;
3115 static void __net_exit
ipv6_inetpeer_exit(struct net
*net
)
3117 struct inet_peer_base
*bp
= net
->ipv6
.peers
;
3119 net
->ipv6
.peers
= NULL
;
3120 inetpeer_invalidate_tree(bp
);
3124 static struct pernet_operations ipv6_inetpeer_ops
= {
3125 .init
= ipv6_inetpeer_init
,
3126 .exit
= ipv6_inetpeer_exit
,
3129 static struct pernet_operations ip6_route_net_late_ops
= {
3130 .init
= ip6_route_net_init_late
,
3131 .exit
= ip6_route_net_exit_late
,
3134 static struct notifier_block ip6_route_dev_notifier
= {
3135 .notifier_call
= ip6_route_dev_notify
,
3139 int __init
ip6_route_init(void)
3144 ip6_dst_ops_template
.kmem_cachep
=
3145 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info
), 0,
3146 SLAB_HWCACHE_ALIGN
, NULL
);
3147 if (!ip6_dst_ops_template
.kmem_cachep
)
3150 ret
= dst_entries_init(&ip6_dst_blackhole_ops
);
3152 goto out_kmem_cache
;
3154 ret
= register_pernet_subsys(&ipv6_inetpeer_ops
);
3156 goto out_dst_entries
;
3158 ret
= register_pernet_subsys(&ip6_route_net_ops
);
3160 goto out_register_inetpeer
;
3162 ip6_dst_blackhole_ops
.kmem_cachep
= ip6_dst_ops_template
.kmem_cachep
;
3164 /* Registering of the loopback is done before this portion of code,
3165 * the loopback reference in rt6_info will not be taken, do it
3166 * manually for init_net */
3167 init_net
.ipv6
.ip6_null_entry
->dst
.dev
= init_net
.loopback_dev
;
3168 init_net
.ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3169 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3170 init_net
.ipv6
.ip6_prohibit_entry
->dst
.dev
= init_net
.loopback_dev
;
3171 init_net
.ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3172 init_net
.ipv6
.ip6_blk_hole_entry
->dst
.dev
= init_net
.loopback_dev
;
3173 init_net
.ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3177 goto out_register_subsys
;
3183 ret
= fib6_rules_init();
3187 ret
= register_pernet_subsys(&ip6_route_net_late_ops
);
3189 goto fib6_rules_init
;
3192 if (__rtnl_register(PF_INET6
, RTM_NEWROUTE
, inet6_rtm_newroute
, NULL
, NULL
) ||
3193 __rtnl_register(PF_INET6
, RTM_DELROUTE
, inet6_rtm_delroute
, NULL
, NULL
) ||
3194 __rtnl_register(PF_INET6
, RTM_GETROUTE
, inet6_rtm_getroute
, NULL
, NULL
))
3195 goto out_register_late_subsys
;
3197 ret
= register_netdevice_notifier(&ip6_route_dev_notifier
);
3199 goto out_register_late_subsys
;
3204 out_register_late_subsys
:
3205 unregister_pernet_subsys(&ip6_route_net_late_ops
);
3207 fib6_rules_cleanup();
3212 out_register_subsys
:
3213 unregister_pernet_subsys(&ip6_route_net_ops
);
3214 out_register_inetpeer
:
3215 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
3217 dst_entries_destroy(&ip6_dst_blackhole_ops
);
3219 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);
3223 void ip6_route_cleanup(void)
3225 unregister_netdevice_notifier(&ip6_route_dev_notifier
);
3226 unregister_pernet_subsys(&ip6_route_net_late_ops
);
3227 fib6_rules_cleanup();
3230 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
3231 unregister_pernet_subsys(&ip6_route_net_ops
);
3232 dst_entries_destroy(&ip6_dst_blackhole_ops
);
3233 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);