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
;
847 EXPORT_SYMBOL(rt6_lookup
);
849 /* ip6_ins_rt is called with FREE table->tb6_lock.
850 It takes new route entry, the addition fails by any reason the
851 route is freed. In any case, if caller does not hold it, it may
855 static int __ip6_ins_rt(struct rt6_info
*rt
, struct nl_info
*info
,
856 struct nlattr
*mx
, int mx_len
)
859 struct fib6_table
*table
;
861 table
= rt
->rt6i_table
;
862 write_lock_bh(&table
->tb6_lock
);
863 err
= fib6_add(&table
->tb6_root
, rt
, info
, mx
, mx_len
);
864 write_unlock_bh(&table
->tb6_lock
);
869 int ip6_ins_rt(struct rt6_info
*rt
)
871 struct nl_info info
= {
872 .nl_net
= dev_net(rt
->dst
.dev
),
874 return __ip6_ins_rt(rt
, &info
, NULL
, 0);
877 static struct rt6_info
*rt6_alloc_cow(struct rt6_info
*ort
,
878 const struct in6_addr
*daddr
,
879 const struct in6_addr
*saddr
)
887 rt
= ip6_rt_copy(ort
, daddr
);
890 if (ort
->rt6i_dst
.plen
!= 128 &&
891 ipv6_addr_equal(&ort
->rt6i_dst
.addr
, daddr
))
892 rt
->rt6i_flags
|= RTF_ANYCAST
;
894 rt
->rt6i_flags
|= RTF_CACHE
;
896 #ifdef CONFIG_IPV6_SUBTREES
897 if (rt
->rt6i_src
.plen
&& saddr
) {
898 rt
->rt6i_src
.addr
= *saddr
;
899 rt
->rt6i_src
.plen
= 128;
907 static struct rt6_info
*rt6_alloc_clone(struct rt6_info
*ort
,
908 const struct in6_addr
*daddr
)
910 struct rt6_info
*rt
= ip6_rt_copy(ort
, daddr
);
913 rt
->rt6i_flags
|= RTF_CACHE
;
917 static struct rt6_info
*ip6_pol_route(struct net
*net
, struct fib6_table
*table
, int oif
,
918 struct flowi6
*fl6
, int flags
)
920 struct fib6_node
*fn
;
921 struct rt6_info
*rt
, *nrt
;
925 int reachable
= net
->ipv6
.devconf_all
->forwarding
? 0 : RT6_LOOKUP_F_REACHABLE
;
927 strict
|= flags
& RT6_LOOKUP_F_IFACE
;
930 read_lock_bh(&table
->tb6_lock
);
933 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
936 rt
= rt6_select(fn
, oif
, strict
| reachable
);
937 if (rt
->rt6i_nsiblings
)
938 rt
= rt6_multipath_select(rt
, fl6
, oif
, strict
| reachable
);
939 BACKTRACK(net
, &fl6
->saddr
);
940 if (rt
== net
->ipv6
.ip6_null_entry
||
941 rt
->rt6i_flags
& RTF_CACHE
)
945 read_unlock_bh(&table
->tb6_lock
);
947 if (!(rt
->rt6i_flags
& (RTF_NONEXTHOP
| RTF_GATEWAY
)))
948 nrt
= rt6_alloc_cow(rt
, &fl6
->daddr
, &fl6
->saddr
);
949 else if (!(rt
->dst
.flags
& DST_HOST
))
950 nrt
= rt6_alloc_clone(rt
, &fl6
->daddr
);
955 rt
= nrt
? : net
->ipv6
.ip6_null_entry
;
959 err
= ip6_ins_rt(nrt
);
968 * Race condition! In the gap, when table->tb6_lock was
969 * released someone could insert this route. Relookup.
980 read_unlock_bh(&table
->tb6_lock
);
982 rt
->dst
.lastuse
= jiffies
;
988 static struct rt6_info
*ip6_pol_route_input(struct net
*net
, struct fib6_table
*table
,
989 struct flowi6
*fl6
, int flags
)
991 return ip6_pol_route(net
, table
, fl6
->flowi6_iif
, fl6
, flags
);
994 static struct dst_entry
*ip6_route_input_lookup(struct net
*net
,
995 struct net_device
*dev
,
996 struct flowi6
*fl6
, int flags
)
998 if (rt6_need_strict(&fl6
->daddr
) && dev
->type
!= ARPHRD_PIMREG
)
999 flags
|= RT6_LOOKUP_F_IFACE
;
1001 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_input
);
1004 void ip6_route_input(struct sk_buff
*skb
)
1006 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
1007 struct net
*net
= dev_net(skb
->dev
);
1008 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
1009 struct flowi6 fl6
= {
1010 .flowi6_iif
= skb
->dev
->ifindex
,
1011 .daddr
= iph
->daddr
,
1012 .saddr
= iph
->saddr
,
1013 .flowlabel
= ip6_flowinfo(iph
),
1014 .flowi6_mark
= skb
->mark
,
1015 .flowi6_proto
= iph
->nexthdr
,
1018 skb_dst_set(skb
, ip6_route_input_lookup(net
, skb
->dev
, &fl6
, flags
));
1021 static struct rt6_info
*ip6_pol_route_output(struct net
*net
, struct fib6_table
*table
,
1022 struct flowi6
*fl6
, int flags
)
1024 return ip6_pol_route(net
, table
, fl6
->flowi6_oif
, fl6
, flags
);
1027 struct dst_entry
* ip6_route_output(struct net
*net
, const struct sock
*sk
,
1032 fl6
->flowi6_iif
= LOOPBACK_IFINDEX
;
1034 if ((sk
&& sk
->sk_bound_dev_if
) || rt6_need_strict(&fl6
->daddr
))
1035 flags
|= RT6_LOOKUP_F_IFACE
;
1037 if (!ipv6_addr_any(&fl6
->saddr
))
1038 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
1040 flags
|= rt6_srcprefs2flags(inet6_sk(sk
)->srcprefs
);
1042 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_output
);
1045 EXPORT_SYMBOL(ip6_route_output
);
1047 struct dst_entry
*ip6_blackhole_route(struct net
*net
, struct dst_entry
*dst_orig
)
1049 struct rt6_info
*rt
, *ort
= (struct rt6_info
*) dst_orig
;
1050 struct dst_entry
*new = NULL
;
1052 rt
= dst_alloc(&ip6_dst_blackhole_ops
, ort
->dst
.dev
, 1, DST_OBSOLETE_NONE
, 0);
1056 memset(new + 1, 0, sizeof(*rt
) - sizeof(*new));
1057 rt6_init_peer(rt
, net
->ipv6
.peers
);
1060 new->input
= dst_discard
;
1061 new->output
= dst_discard_sk
;
1063 if (dst_metrics_read_only(&ort
->dst
))
1064 new->_metrics
= ort
->dst
._metrics
;
1066 dst_copy_metrics(new, &ort
->dst
);
1067 rt
->rt6i_idev
= ort
->rt6i_idev
;
1069 in6_dev_hold(rt
->rt6i_idev
);
1071 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
1072 rt
->rt6i_flags
= ort
->rt6i_flags
;
1073 rt
->rt6i_metric
= 0;
1075 memcpy(&rt
->rt6i_dst
, &ort
->rt6i_dst
, sizeof(struct rt6key
));
1076 #ifdef CONFIG_IPV6_SUBTREES
1077 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1083 dst_release(dst_orig
);
1084 return new ? new : ERR_PTR(-ENOMEM
);
1088 * Destination cache support functions
1091 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
)
1093 struct rt6_info
*rt
;
1095 rt
= (struct rt6_info
*) dst
;
1097 /* All IPV6 dsts are created with ->obsolete set to the value
1098 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1099 * into this function always.
1101 if (rt
->rt6i_genid
!= rt_genid_ipv6(dev_net(rt
->dst
.dev
)))
1104 if (!rt
->rt6i_node
|| (rt
->rt6i_node
->fn_sernum
!= cookie
))
1107 if (rt6_check_expired(rt
))
1113 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*dst
)
1115 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1118 if (rt
->rt6i_flags
& RTF_CACHE
) {
1119 if (rt6_check_expired(rt
)) {
1131 static void ip6_link_failure(struct sk_buff
*skb
)
1133 struct rt6_info
*rt
;
1135 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, ICMPV6_ADDR_UNREACH
, 0);
1137 rt
= (struct rt6_info
*) skb_dst(skb
);
1139 if (rt
->rt6i_flags
& RTF_CACHE
) {
1143 } else if (rt
->rt6i_node
&& (rt
->rt6i_flags
& RTF_DEFAULT
)) {
1144 rt
->rt6i_node
->fn_sernum
= -1;
1149 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
1150 struct sk_buff
*skb
, u32 mtu
)
1152 struct rt6_info
*rt6
= (struct rt6_info
*)dst
;
1155 if (mtu
< dst_mtu(dst
) && rt6
->rt6i_dst
.plen
== 128) {
1156 struct net
*net
= dev_net(dst
->dev
);
1158 rt6
->rt6i_flags
|= RTF_MODIFIED
;
1159 if (mtu
< IPV6_MIN_MTU
) {
1160 u32 features
= dst_metric(dst
, RTAX_FEATURES
);
1162 features
|= RTAX_FEATURE_ALLFRAG
;
1163 dst_metric_set(dst
, RTAX_FEATURES
, features
);
1165 dst_metric_set(dst
, RTAX_MTU
, mtu
);
1166 rt6_update_expires(rt6
, net
->ipv6
.sysctl
.ip6_rt_mtu_expires
);
1170 void ip6_update_pmtu(struct sk_buff
*skb
, struct net
*net
, __be32 mtu
,
1173 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1174 struct dst_entry
*dst
;
1177 memset(&fl6
, 0, sizeof(fl6
));
1178 fl6
.flowi6_oif
= oif
;
1179 fl6
.flowi6_mark
= mark
;
1180 fl6
.daddr
= iph
->daddr
;
1181 fl6
.saddr
= iph
->saddr
;
1182 fl6
.flowlabel
= ip6_flowinfo(iph
);
1184 dst
= ip6_route_output(net
, NULL
, &fl6
);
1186 ip6_rt_update_pmtu(dst
, NULL
, skb
, ntohl(mtu
));
1189 EXPORT_SYMBOL_GPL(ip6_update_pmtu
);
1191 void ip6_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, __be32 mtu
)
1193 ip6_update_pmtu(skb
, sock_net(sk
), mtu
,
1194 sk
->sk_bound_dev_if
, sk
->sk_mark
);
1196 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu
);
1198 /* Handle redirects */
1199 struct ip6rd_flowi
{
1201 struct in6_addr gateway
;
1204 static struct rt6_info
*__ip6_route_redirect(struct net
*net
,
1205 struct fib6_table
*table
,
1209 struct ip6rd_flowi
*rdfl
= (struct ip6rd_flowi
*)fl6
;
1210 struct rt6_info
*rt
;
1211 struct fib6_node
*fn
;
1213 /* Get the "current" route for this destination and
1214 * check if the redirect has come from approriate router.
1216 * RFC 4861 specifies that redirects should only be
1217 * accepted if they come from the nexthop to the target.
1218 * Due to the way the routes are chosen, this notion
1219 * is a bit fuzzy and one might need to check all possible
1223 read_lock_bh(&table
->tb6_lock
);
1224 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
1226 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1227 if (rt6_check_expired(rt
))
1231 if (!(rt
->rt6i_flags
& RTF_GATEWAY
))
1233 if (fl6
->flowi6_oif
!= rt
->dst
.dev
->ifindex
)
1235 if (!ipv6_addr_equal(&rdfl
->gateway
, &rt
->rt6i_gateway
))
1241 rt
= net
->ipv6
.ip6_null_entry
;
1242 else if (rt
->dst
.error
) {
1243 rt
= net
->ipv6
.ip6_null_entry
;
1246 BACKTRACK(net
, &fl6
->saddr
);
1250 read_unlock_bh(&table
->tb6_lock
);
1255 static struct dst_entry
*ip6_route_redirect(struct net
*net
,
1256 const struct flowi6
*fl6
,
1257 const struct in6_addr
*gateway
)
1259 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
1260 struct ip6rd_flowi rdfl
;
1263 rdfl
.gateway
= *gateway
;
1265 return fib6_rule_lookup(net
, &rdfl
.fl6
,
1266 flags
, __ip6_route_redirect
);
1269 void ip6_redirect(struct sk_buff
*skb
, struct net
*net
, int oif
, u32 mark
)
1271 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1272 struct dst_entry
*dst
;
1275 memset(&fl6
, 0, sizeof(fl6
));
1276 fl6
.flowi6_iif
= LOOPBACK_IFINDEX
;
1277 fl6
.flowi6_oif
= oif
;
1278 fl6
.flowi6_mark
= mark
;
1279 fl6
.daddr
= iph
->daddr
;
1280 fl6
.saddr
= iph
->saddr
;
1281 fl6
.flowlabel
= ip6_flowinfo(iph
);
1283 dst
= ip6_route_redirect(net
, &fl6
, &ipv6_hdr(skb
)->saddr
);
1284 rt6_do_redirect(dst
, NULL
, skb
);
1287 EXPORT_SYMBOL_GPL(ip6_redirect
);
1289 void ip6_redirect_no_header(struct sk_buff
*skb
, struct net
*net
, int oif
,
1292 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
1293 const struct rd_msg
*msg
= (struct rd_msg
*)icmp6_hdr(skb
);
1294 struct dst_entry
*dst
;
1297 memset(&fl6
, 0, sizeof(fl6
));
1298 fl6
.flowi6_iif
= LOOPBACK_IFINDEX
;
1299 fl6
.flowi6_oif
= oif
;
1300 fl6
.flowi6_mark
= mark
;
1301 fl6
.daddr
= msg
->dest
;
1302 fl6
.saddr
= iph
->daddr
;
1304 dst
= ip6_route_redirect(net
, &fl6
, &iph
->saddr
);
1305 rt6_do_redirect(dst
, NULL
, skb
);
1309 void ip6_sk_redirect(struct sk_buff
*skb
, struct sock
*sk
)
1311 ip6_redirect(skb
, sock_net(sk
), sk
->sk_bound_dev_if
, sk
->sk_mark
);
1313 EXPORT_SYMBOL_GPL(ip6_sk_redirect
);
1315 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
)
1317 struct net_device
*dev
= dst
->dev
;
1318 unsigned int mtu
= dst_mtu(dst
);
1319 struct net
*net
= dev_net(dev
);
1321 mtu
-= sizeof(struct ipv6hdr
) + sizeof(struct tcphdr
);
1323 if (mtu
< net
->ipv6
.sysctl
.ip6_rt_min_advmss
)
1324 mtu
= net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
1327 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1328 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1329 * IPV6_MAXPLEN is also valid and means: "any MSS,
1330 * rely only on pmtu discovery"
1332 if (mtu
> IPV6_MAXPLEN
- sizeof(struct tcphdr
))
1337 static unsigned int ip6_mtu(const struct dst_entry
*dst
)
1339 struct inet6_dev
*idev
;
1340 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
1348 idev
= __in6_dev_get(dst
->dev
);
1350 mtu
= idev
->cnf
.mtu6
;
1354 return min_t(unsigned int, mtu
, IP6_MAX_MTU
);
1357 static struct dst_entry
*icmp6_dst_gc_list
;
1358 static DEFINE_SPINLOCK(icmp6_dst_lock
);
1360 struct dst_entry
*icmp6_dst_alloc(struct net_device
*dev
,
1363 struct dst_entry
*dst
;
1364 struct rt6_info
*rt
;
1365 struct inet6_dev
*idev
= in6_dev_get(dev
);
1366 struct net
*net
= dev_net(dev
);
1368 if (unlikely(!idev
))
1369 return ERR_PTR(-ENODEV
);
1371 rt
= ip6_dst_alloc(net
, dev
, 0, NULL
);
1372 if (unlikely(!rt
)) {
1374 dst
= ERR_PTR(-ENOMEM
);
1378 rt
->dst
.flags
|= DST_HOST
;
1379 rt
->dst
.output
= ip6_output
;
1380 atomic_set(&rt
->dst
.__refcnt
, 1);
1381 rt
->rt6i_gateway
= fl6
->daddr
;
1382 rt
->rt6i_dst
.addr
= fl6
->daddr
;
1383 rt
->rt6i_dst
.plen
= 128;
1384 rt
->rt6i_idev
= idev
;
1385 dst_metric_set(&rt
->dst
, RTAX_HOPLIMIT
, 0);
1387 spin_lock_bh(&icmp6_dst_lock
);
1388 rt
->dst
.next
= icmp6_dst_gc_list
;
1389 icmp6_dst_gc_list
= &rt
->dst
;
1390 spin_unlock_bh(&icmp6_dst_lock
);
1392 fib6_force_start_gc(net
);
1394 dst
= xfrm_lookup(net
, &rt
->dst
, flowi6_to_flowi(fl6
), NULL
, 0);
1400 int icmp6_dst_gc(void)
1402 struct dst_entry
*dst
, **pprev
;
1405 spin_lock_bh(&icmp6_dst_lock
);
1406 pprev
= &icmp6_dst_gc_list
;
1408 while ((dst
= *pprev
) != NULL
) {
1409 if (!atomic_read(&dst
->__refcnt
)) {
1418 spin_unlock_bh(&icmp6_dst_lock
);
1423 static void icmp6_clean_all(int (*func
)(struct rt6_info
*rt
, void *arg
),
1426 struct dst_entry
*dst
, **pprev
;
1428 spin_lock_bh(&icmp6_dst_lock
);
1429 pprev
= &icmp6_dst_gc_list
;
1430 while ((dst
= *pprev
) != NULL
) {
1431 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1432 if (func(rt
, arg
)) {
1439 spin_unlock_bh(&icmp6_dst_lock
);
1442 static int ip6_dst_gc(struct dst_ops
*ops
)
1444 struct net
*net
= container_of(ops
, struct net
, ipv6
.ip6_dst_ops
);
1445 int rt_min_interval
= net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
1446 int rt_max_size
= net
->ipv6
.sysctl
.ip6_rt_max_size
;
1447 int rt_elasticity
= net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
1448 int rt_gc_timeout
= net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
1449 unsigned long rt_last_gc
= net
->ipv6
.ip6_rt_last_gc
;
1452 entries
= dst_entries_get_fast(ops
);
1453 if (time_after(rt_last_gc
+ rt_min_interval
, jiffies
) &&
1454 entries
<= rt_max_size
)
1457 net
->ipv6
.ip6_rt_gc_expire
++;
1458 fib6_run_gc(net
->ipv6
.ip6_rt_gc_expire
, net
, entries
> rt_max_size
);
1459 entries
= dst_entries_get_slow(ops
);
1460 if (entries
< ops
->gc_thresh
)
1461 net
->ipv6
.ip6_rt_gc_expire
= rt_gc_timeout
>>1;
1463 net
->ipv6
.ip6_rt_gc_expire
-= net
->ipv6
.ip6_rt_gc_expire
>>rt_elasticity
;
1464 return entries
> rt_max_size
;
1471 int ip6_route_add(struct fib6_config
*cfg
)
1474 struct net
*net
= cfg
->fc_nlinfo
.nl_net
;
1475 struct rt6_info
*rt
= NULL
;
1476 struct net_device
*dev
= NULL
;
1477 struct inet6_dev
*idev
= NULL
;
1478 struct fib6_table
*table
;
1481 if (cfg
->fc_dst_len
> 128 || cfg
->fc_src_len
> 128)
1483 #ifndef CONFIG_IPV6_SUBTREES
1484 if (cfg
->fc_src_len
)
1487 if (cfg
->fc_ifindex
) {
1489 dev
= dev_get_by_index(net
, cfg
->fc_ifindex
);
1492 idev
= in6_dev_get(dev
);
1497 if (cfg
->fc_metric
== 0)
1498 cfg
->fc_metric
= IP6_RT_PRIO_USER
;
1501 if (cfg
->fc_nlinfo
.nlh
&&
1502 !(cfg
->fc_nlinfo
.nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
1503 table
= fib6_get_table(net
, cfg
->fc_table
);
1505 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1506 table
= fib6_new_table(net
, cfg
->fc_table
);
1509 table
= fib6_new_table(net
, cfg
->fc_table
);
1515 rt
= ip6_dst_alloc(net
, NULL
, (cfg
->fc_flags
& RTF_ADDRCONF
) ? 0 : DST_NOCOUNT
, table
);
1522 if (cfg
->fc_flags
& RTF_EXPIRES
)
1523 rt6_set_expires(rt
, jiffies
+
1524 clock_t_to_jiffies(cfg
->fc_expires
));
1526 rt6_clean_expires(rt
);
1528 if (cfg
->fc_protocol
== RTPROT_UNSPEC
)
1529 cfg
->fc_protocol
= RTPROT_BOOT
;
1530 rt
->rt6i_protocol
= cfg
->fc_protocol
;
1532 addr_type
= ipv6_addr_type(&cfg
->fc_dst
);
1534 if (addr_type
& IPV6_ADDR_MULTICAST
)
1535 rt
->dst
.input
= ip6_mc_input
;
1536 else if (cfg
->fc_flags
& RTF_LOCAL
)
1537 rt
->dst
.input
= ip6_input
;
1539 rt
->dst
.input
= ip6_forward
;
1541 rt
->dst
.output
= ip6_output
;
1543 ipv6_addr_prefix(&rt
->rt6i_dst
.addr
, &cfg
->fc_dst
, cfg
->fc_dst_len
);
1544 rt
->rt6i_dst
.plen
= cfg
->fc_dst_len
;
1545 if (rt
->rt6i_dst
.plen
== 128) {
1546 rt
->dst
.flags
|= DST_HOST
;
1547 dst_metrics_set_force_overwrite(&rt
->dst
);
1550 #ifdef CONFIG_IPV6_SUBTREES
1551 ipv6_addr_prefix(&rt
->rt6i_src
.addr
, &cfg
->fc_src
, cfg
->fc_src_len
);
1552 rt
->rt6i_src
.plen
= cfg
->fc_src_len
;
1555 rt
->rt6i_metric
= cfg
->fc_metric
;
1557 /* We cannot add true routes via loopback here,
1558 they would result in kernel looping; promote them to reject routes
1560 if ((cfg
->fc_flags
& RTF_REJECT
) ||
1561 (dev
&& (dev
->flags
& IFF_LOOPBACK
) &&
1562 !(addr_type
& IPV6_ADDR_LOOPBACK
) &&
1563 !(cfg
->fc_flags
& RTF_LOCAL
))) {
1564 /* hold loopback dev/idev if we haven't done so. */
1565 if (dev
!= net
->loopback_dev
) {
1570 dev
= net
->loopback_dev
;
1572 idev
= in6_dev_get(dev
);
1578 rt
->rt6i_flags
= RTF_REJECT
|RTF_NONEXTHOP
;
1579 switch (cfg
->fc_type
) {
1581 rt
->dst
.error
= -EINVAL
;
1582 rt
->dst
.output
= dst_discard_sk
;
1583 rt
->dst
.input
= dst_discard
;
1586 rt
->dst
.error
= -EACCES
;
1587 rt
->dst
.output
= ip6_pkt_prohibit_out
;
1588 rt
->dst
.input
= ip6_pkt_prohibit
;
1592 rt
->dst
.error
= (cfg
->fc_type
== RTN_THROW
) ? -EAGAIN
1594 rt
->dst
.output
= ip6_pkt_discard_out
;
1595 rt
->dst
.input
= ip6_pkt_discard
;
1601 if (cfg
->fc_flags
& RTF_GATEWAY
) {
1602 const struct in6_addr
*gw_addr
;
1605 gw_addr
= &cfg
->fc_gateway
;
1606 rt
->rt6i_gateway
= *gw_addr
;
1607 gwa_type
= ipv6_addr_type(gw_addr
);
1609 if (gwa_type
!= (IPV6_ADDR_LINKLOCAL
|IPV6_ADDR_UNICAST
)) {
1610 struct rt6_info
*grt
;
1612 /* IPv6 strictly inhibits using not link-local
1613 addresses as nexthop address.
1614 Otherwise, router will not able to send redirects.
1615 It is very good, but in some (rare!) circumstances
1616 (SIT, PtP, NBMA NOARP links) it is handy to allow
1617 some exceptions. --ANK
1620 if (!(gwa_type
& IPV6_ADDR_UNICAST
))
1623 grt
= rt6_lookup(net
, gw_addr
, NULL
, cfg
->fc_ifindex
, 1);
1625 err
= -EHOSTUNREACH
;
1629 if (dev
!= grt
->dst
.dev
) {
1635 idev
= grt
->rt6i_idev
;
1637 in6_dev_hold(grt
->rt6i_idev
);
1639 if (!(grt
->rt6i_flags
& RTF_GATEWAY
))
1647 if (!dev
|| (dev
->flags
& IFF_LOOPBACK
))
1655 if (!ipv6_addr_any(&cfg
->fc_prefsrc
)) {
1656 if (!ipv6_chk_addr(net
, &cfg
->fc_prefsrc
, dev
, 0)) {
1660 rt
->rt6i_prefsrc
.addr
= cfg
->fc_prefsrc
;
1661 rt
->rt6i_prefsrc
.plen
= 128;
1663 rt
->rt6i_prefsrc
.plen
= 0;
1665 rt
->rt6i_flags
= cfg
->fc_flags
;
1669 rt
->rt6i_idev
= idev
;
1670 rt
->rt6i_table
= table
;
1672 cfg
->fc_nlinfo
.nl_net
= dev_net(dev
);
1674 return __ip6_ins_rt(rt
, &cfg
->fc_nlinfo
, cfg
->fc_mx
, cfg
->fc_mx_len
);
1686 static int __ip6_del_rt(struct rt6_info
*rt
, struct nl_info
*info
)
1689 struct fib6_table
*table
;
1690 struct net
*net
= dev_net(rt
->dst
.dev
);
1692 if (rt
== net
->ipv6
.ip6_null_entry
) {
1697 table
= rt
->rt6i_table
;
1698 write_lock_bh(&table
->tb6_lock
);
1699 err
= fib6_del(rt
, info
);
1700 write_unlock_bh(&table
->tb6_lock
);
1707 int ip6_del_rt(struct rt6_info
*rt
)
1709 struct nl_info info
= {
1710 .nl_net
= dev_net(rt
->dst
.dev
),
1712 return __ip6_del_rt(rt
, &info
);
1715 static int ip6_route_del(struct fib6_config
*cfg
)
1717 struct fib6_table
*table
;
1718 struct fib6_node
*fn
;
1719 struct rt6_info
*rt
;
1722 table
= fib6_get_table(cfg
->fc_nlinfo
.nl_net
, cfg
->fc_table
);
1726 read_lock_bh(&table
->tb6_lock
);
1728 fn
= fib6_locate(&table
->tb6_root
,
1729 &cfg
->fc_dst
, cfg
->fc_dst_len
,
1730 &cfg
->fc_src
, cfg
->fc_src_len
);
1733 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1734 if (cfg
->fc_ifindex
&&
1736 rt
->dst
.dev
->ifindex
!= cfg
->fc_ifindex
))
1738 if (cfg
->fc_flags
& RTF_GATEWAY
&&
1739 !ipv6_addr_equal(&cfg
->fc_gateway
, &rt
->rt6i_gateway
))
1741 if (cfg
->fc_metric
&& cfg
->fc_metric
!= rt
->rt6i_metric
)
1744 read_unlock_bh(&table
->tb6_lock
);
1746 return __ip6_del_rt(rt
, &cfg
->fc_nlinfo
);
1749 read_unlock_bh(&table
->tb6_lock
);
1754 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
, struct sk_buff
*skb
)
1756 struct net
*net
= dev_net(skb
->dev
);
1757 struct netevent_redirect netevent
;
1758 struct rt6_info
*rt
, *nrt
= NULL
;
1759 struct ndisc_options ndopts
;
1760 struct inet6_dev
*in6_dev
;
1761 struct neighbour
*neigh
;
1763 int optlen
, on_link
;
1766 optlen
= skb_tail_pointer(skb
) - skb_transport_header(skb
);
1767 optlen
-= sizeof(*msg
);
1770 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1774 msg
= (struct rd_msg
*)icmp6_hdr(skb
);
1776 if (ipv6_addr_is_multicast(&msg
->dest
)) {
1777 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1782 if (ipv6_addr_equal(&msg
->dest
, &msg
->target
)) {
1784 } else if (ipv6_addr_type(&msg
->target
) !=
1785 (IPV6_ADDR_UNICAST
|IPV6_ADDR_LINKLOCAL
)) {
1786 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1790 in6_dev
= __in6_dev_get(skb
->dev
);
1793 if (in6_dev
->cnf
.forwarding
|| !in6_dev
->cnf
.accept_redirects
)
1797 * The IP source address of the Redirect MUST be the same as the current
1798 * first-hop router for the specified ICMP Destination Address.
1801 if (!ndisc_parse_options(msg
->opt
, optlen
, &ndopts
)) {
1802 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1807 if (ndopts
.nd_opts_tgt_lladdr
) {
1808 lladdr
= ndisc_opt_addr_data(ndopts
.nd_opts_tgt_lladdr
,
1811 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1816 rt
= (struct rt6_info
*) dst
;
1817 if (rt
== net
->ipv6
.ip6_null_entry
) {
1818 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1822 /* Redirect received -> path was valid.
1823 * Look, redirects are sent only in response to data packets,
1824 * so that this nexthop apparently is reachable. --ANK
1826 dst_confirm(&rt
->dst
);
1828 neigh
= __neigh_lookup(&nd_tbl
, &msg
->target
, skb
->dev
, 1);
1833 * We have finally decided to accept it.
1836 neigh_update(neigh
, lladdr
, NUD_STALE
,
1837 NEIGH_UPDATE_F_WEAK_OVERRIDE
|
1838 NEIGH_UPDATE_F_OVERRIDE
|
1839 (on_link
? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER
|
1840 NEIGH_UPDATE_F_ISROUTER
))
1843 nrt
= ip6_rt_copy(rt
, &msg
->dest
);
1847 nrt
->rt6i_flags
= RTF_GATEWAY
|RTF_UP
|RTF_DYNAMIC
|RTF_CACHE
;
1849 nrt
->rt6i_flags
&= ~RTF_GATEWAY
;
1851 nrt
->rt6i_gateway
= *(struct in6_addr
*)neigh
->primary_key
;
1853 if (ip6_ins_rt(nrt
))
1856 netevent
.old
= &rt
->dst
;
1857 netevent
.new = &nrt
->dst
;
1858 netevent
.daddr
= &msg
->dest
;
1859 netevent
.neigh
= neigh
;
1860 call_netevent_notifiers(NETEVENT_REDIRECT
, &netevent
);
1862 if (rt
->rt6i_flags
& RTF_CACHE
) {
1863 rt
= (struct rt6_info
*) dst_clone(&rt
->dst
);
1868 neigh_release(neigh
);
1872 * Misc support functions
1875 static struct rt6_info
*ip6_rt_copy(struct rt6_info
*ort
,
1876 const struct in6_addr
*dest
)
1878 struct net
*net
= dev_net(ort
->dst
.dev
);
1879 struct rt6_info
*rt
= ip6_dst_alloc(net
, ort
->dst
.dev
, 0,
1883 rt
->dst
.input
= ort
->dst
.input
;
1884 rt
->dst
.output
= ort
->dst
.output
;
1885 rt
->dst
.flags
|= DST_HOST
;
1887 rt
->rt6i_dst
.addr
= *dest
;
1888 rt
->rt6i_dst
.plen
= 128;
1889 dst_copy_metrics(&rt
->dst
, &ort
->dst
);
1890 rt
->dst
.error
= ort
->dst
.error
;
1891 rt
->rt6i_idev
= ort
->rt6i_idev
;
1893 in6_dev_hold(rt
->rt6i_idev
);
1894 rt
->dst
.lastuse
= jiffies
;
1896 if (ort
->rt6i_flags
& RTF_GATEWAY
)
1897 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
1899 rt
->rt6i_gateway
= *dest
;
1900 rt
->rt6i_flags
= ort
->rt6i_flags
;
1901 rt6_set_from(rt
, ort
);
1902 rt
->rt6i_metric
= 0;
1904 #ifdef CONFIG_IPV6_SUBTREES
1905 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1907 memcpy(&rt
->rt6i_prefsrc
, &ort
->rt6i_prefsrc
, sizeof(struct rt6key
));
1908 rt
->rt6i_table
= ort
->rt6i_table
;
1913 #ifdef CONFIG_IPV6_ROUTE_INFO
1914 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
1915 const struct in6_addr
*prefix
, int prefixlen
,
1916 const struct in6_addr
*gwaddr
, int ifindex
)
1918 struct fib6_node
*fn
;
1919 struct rt6_info
*rt
= NULL
;
1920 struct fib6_table
*table
;
1922 table
= fib6_get_table(net
, RT6_TABLE_INFO
);
1926 read_lock_bh(&table
->tb6_lock
);
1927 fn
= fib6_locate(&table
->tb6_root
, prefix
,prefixlen
, NULL
, 0);
1931 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1932 if (rt
->dst
.dev
->ifindex
!= ifindex
)
1934 if ((rt
->rt6i_flags
& (RTF_ROUTEINFO
|RTF_GATEWAY
)) != (RTF_ROUTEINFO
|RTF_GATEWAY
))
1936 if (!ipv6_addr_equal(&rt
->rt6i_gateway
, gwaddr
))
1942 read_unlock_bh(&table
->tb6_lock
);
1946 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
1947 const struct in6_addr
*prefix
, int prefixlen
,
1948 const struct in6_addr
*gwaddr
, int ifindex
,
1951 struct fib6_config cfg
= {
1952 .fc_table
= RT6_TABLE_INFO
,
1953 .fc_metric
= IP6_RT_PRIO_USER
,
1954 .fc_ifindex
= ifindex
,
1955 .fc_dst_len
= prefixlen
,
1956 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_ROUTEINFO
|
1957 RTF_UP
| RTF_PREF(pref
),
1958 .fc_nlinfo
.portid
= 0,
1959 .fc_nlinfo
.nlh
= NULL
,
1960 .fc_nlinfo
.nl_net
= net
,
1963 cfg
.fc_dst
= *prefix
;
1964 cfg
.fc_gateway
= *gwaddr
;
1966 /* We should treat it as a default route if prefix length is 0. */
1968 cfg
.fc_flags
|= RTF_DEFAULT
;
1970 ip6_route_add(&cfg
);
1972 return rt6_get_route_info(net
, prefix
, prefixlen
, gwaddr
, ifindex
);
1976 struct rt6_info
*rt6_get_dflt_router(const struct in6_addr
*addr
, struct net_device
*dev
)
1978 struct rt6_info
*rt
;
1979 struct fib6_table
*table
;
1981 table
= fib6_get_table(dev_net(dev
), RT6_TABLE_DFLT
);
1985 read_lock_bh(&table
->tb6_lock
);
1986 for (rt
= table
->tb6_root
.leaf
; rt
; rt
=rt
->dst
.rt6_next
) {
1987 if (dev
== rt
->dst
.dev
&&
1988 ((rt
->rt6i_flags
& (RTF_ADDRCONF
| RTF_DEFAULT
)) == (RTF_ADDRCONF
| RTF_DEFAULT
)) &&
1989 ipv6_addr_equal(&rt
->rt6i_gateway
, addr
))
1994 read_unlock_bh(&table
->tb6_lock
);
1998 struct rt6_info
*rt6_add_dflt_router(const struct in6_addr
*gwaddr
,
1999 struct net_device
*dev
,
2002 struct fib6_config cfg
= {
2003 .fc_table
= RT6_TABLE_DFLT
,
2004 .fc_metric
= IP6_RT_PRIO_USER
,
2005 .fc_ifindex
= dev
->ifindex
,
2006 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_DEFAULT
|
2007 RTF_UP
| RTF_EXPIRES
| RTF_PREF(pref
),
2008 .fc_nlinfo
.portid
= 0,
2009 .fc_nlinfo
.nlh
= NULL
,
2010 .fc_nlinfo
.nl_net
= dev_net(dev
),
2013 cfg
.fc_gateway
= *gwaddr
;
2015 ip6_route_add(&cfg
);
2017 return rt6_get_dflt_router(gwaddr
, dev
);
2020 void rt6_purge_dflt_routers(struct net
*net
)
2022 struct rt6_info
*rt
;
2023 struct fib6_table
*table
;
2025 /* NOTE: Keep consistent with rt6_get_dflt_router */
2026 table
= fib6_get_table(net
, RT6_TABLE_DFLT
);
2031 read_lock_bh(&table
->tb6_lock
);
2032 for (rt
= table
->tb6_root
.leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
2033 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
) &&
2034 (!rt
->rt6i_idev
|| rt
->rt6i_idev
->cnf
.accept_ra
!= 2)) {
2036 read_unlock_bh(&table
->tb6_lock
);
2041 read_unlock_bh(&table
->tb6_lock
);
2044 static void rtmsg_to_fib6_config(struct net
*net
,
2045 struct in6_rtmsg
*rtmsg
,
2046 struct fib6_config
*cfg
)
2048 memset(cfg
, 0, sizeof(*cfg
));
2050 cfg
->fc_table
= RT6_TABLE_MAIN
;
2051 cfg
->fc_ifindex
= rtmsg
->rtmsg_ifindex
;
2052 cfg
->fc_metric
= rtmsg
->rtmsg_metric
;
2053 cfg
->fc_expires
= rtmsg
->rtmsg_info
;
2054 cfg
->fc_dst_len
= rtmsg
->rtmsg_dst_len
;
2055 cfg
->fc_src_len
= rtmsg
->rtmsg_src_len
;
2056 cfg
->fc_flags
= rtmsg
->rtmsg_flags
;
2058 cfg
->fc_nlinfo
.nl_net
= net
;
2060 cfg
->fc_dst
= rtmsg
->rtmsg_dst
;
2061 cfg
->fc_src
= rtmsg
->rtmsg_src
;
2062 cfg
->fc_gateway
= rtmsg
->rtmsg_gateway
;
2065 int ipv6_route_ioctl(struct net
*net
, unsigned int cmd
, void __user
*arg
)
2067 struct fib6_config cfg
;
2068 struct in6_rtmsg rtmsg
;
2072 case SIOCADDRT
: /* Add a route */
2073 case SIOCDELRT
: /* Delete a route */
2074 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
2076 err
= copy_from_user(&rtmsg
, arg
,
2077 sizeof(struct in6_rtmsg
));
2081 rtmsg_to_fib6_config(net
, &rtmsg
, &cfg
);
2086 err
= ip6_route_add(&cfg
);
2089 err
= ip6_route_del(&cfg
);
2103 * Drop the packet on the floor
2106 static int ip6_pkt_drop(struct sk_buff
*skb
, u8 code
, int ipstats_mib_noroutes
)
2109 struct dst_entry
*dst
= skb_dst(skb
);
2110 switch (ipstats_mib_noroutes
) {
2111 case IPSTATS_MIB_INNOROUTES
:
2112 type
= ipv6_addr_type(&ipv6_hdr(skb
)->daddr
);
2113 if (type
== IPV6_ADDR_ANY
) {
2114 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2115 IPSTATS_MIB_INADDRERRORS
);
2119 case IPSTATS_MIB_OUTNOROUTES
:
2120 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2121 ipstats_mib_noroutes
);
2124 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, code
, 0);
2129 static int ip6_pkt_discard(struct sk_buff
*skb
)
2131 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_INNOROUTES
);
2134 static int ip6_pkt_discard_out(struct sock
*sk
, struct sk_buff
*skb
)
2136 skb
->dev
= skb_dst(skb
)->dev
;
2137 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_OUTNOROUTES
);
2140 static int ip6_pkt_prohibit(struct sk_buff
*skb
)
2142 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_INNOROUTES
);
2145 static int ip6_pkt_prohibit_out(struct sock
*sk
, struct sk_buff
*skb
)
2147 skb
->dev
= skb_dst(skb
)->dev
;
2148 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_OUTNOROUTES
);
2152 * Allocate a dst for local (unicast / anycast) address.
2155 struct rt6_info
*addrconf_dst_alloc(struct inet6_dev
*idev
,
2156 const struct in6_addr
*addr
,
2159 struct net
*net
= dev_net(idev
->dev
);
2160 struct rt6_info
*rt
= ip6_dst_alloc(net
, net
->loopback_dev
,
2163 return ERR_PTR(-ENOMEM
);
2167 rt
->dst
.flags
|= DST_HOST
;
2168 rt
->dst
.input
= ip6_input
;
2169 rt
->dst
.output
= ip6_output
;
2170 rt
->rt6i_idev
= idev
;
2172 rt
->rt6i_flags
= RTF_UP
| RTF_NONEXTHOP
;
2174 rt
->rt6i_flags
|= RTF_ANYCAST
;
2176 rt
->rt6i_flags
|= RTF_LOCAL
;
2178 rt
->rt6i_gateway
= *addr
;
2179 rt
->rt6i_dst
.addr
= *addr
;
2180 rt
->rt6i_dst
.plen
= 128;
2181 rt
->rt6i_table
= fib6_get_table(net
, RT6_TABLE_LOCAL
);
2183 atomic_set(&rt
->dst
.__refcnt
, 1);
2188 int ip6_route_get_saddr(struct net
*net
,
2189 struct rt6_info
*rt
,
2190 const struct in6_addr
*daddr
,
2192 struct in6_addr
*saddr
)
2194 struct inet6_dev
*idev
= ip6_dst_idev((struct dst_entry
*)rt
);
2196 if (rt
->rt6i_prefsrc
.plen
)
2197 *saddr
= rt
->rt6i_prefsrc
.addr
;
2199 err
= ipv6_dev_get_saddr(net
, idev
? idev
->dev
: NULL
,
2200 daddr
, prefs
, saddr
);
2204 /* remove deleted ip from prefsrc entries */
2205 struct arg_dev_net_ip
{
2206 struct net_device
*dev
;
2208 struct in6_addr
*addr
;
2211 static int fib6_remove_prefsrc(struct rt6_info
*rt
, void *arg
)
2213 struct net_device
*dev
= ((struct arg_dev_net_ip
*)arg
)->dev
;
2214 struct net
*net
= ((struct arg_dev_net_ip
*)arg
)->net
;
2215 struct in6_addr
*addr
= ((struct arg_dev_net_ip
*)arg
)->addr
;
2217 if (((void *)rt
->dst
.dev
== dev
|| !dev
) &&
2218 rt
!= net
->ipv6
.ip6_null_entry
&&
2219 ipv6_addr_equal(addr
, &rt
->rt6i_prefsrc
.addr
)) {
2220 /* remove prefsrc entry */
2221 rt
->rt6i_prefsrc
.plen
= 0;
2226 void rt6_remove_prefsrc(struct inet6_ifaddr
*ifp
)
2228 struct net
*net
= dev_net(ifp
->idev
->dev
);
2229 struct arg_dev_net_ip adni
= {
2230 .dev
= ifp
->idev
->dev
,
2234 fib6_clean_all(net
, fib6_remove_prefsrc
, &adni
);
2237 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2238 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2240 /* Remove routers and update dst entries when gateway turn into host. */
2241 static int fib6_clean_tohost(struct rt6_info
*rt
, void *arg
)
2243 struct in6_addr
*gateway
= (struct in6_addr
*)arg
;
2245 if ((((rt
->rt6i_flags
& RTF_RA_ROUTER
) == RTF_RA_ROUTER
) ||
2246 ((rt
->rt6i_flags
& RTF_CACHE_GATEWAY
) == RTF_CACHE_GATEWAY
)) &&
2247 ipv6_addr_equal(gateway
, &rt
->rt6i_gateway
)) {
2253 void rt6_clean_tohost(struct net
*net
, struct in6_addr
*gateway
)
2255 fib6_clean_all(net
, fib6_clean_tohost
, gateway
);
2258 struct arg_dev_net
{
2259 struct net_device
*dev
;
2263 static int fib6_ifdown(struct rt6_info
*rt
, void *arg
)
2265 const struct arg_dev_net
*adn
= arg
;
2266 const struct net_device
*dev
= adn
->dev
;
2268 if ((rt
->dst
.dev
== dev
|| !dev
) &&
2269 rt
!= adn
->net
->ipv6
.ip6_null_entry
)
2275 void rt6_ifdown(struct net
*net
, struct net_device
*dev
)
2277 struct arg_dev_net adn
= {
2282 fib6_clean_all(net
, fib6_ifdown
, &adn
);
2283 icmp6_clean_all(fib6_ifdown
, &adn
);
2286 struct rt6_mtu_change_arg
{
2287 struct net_device
*dev
;
2291 static int rt6_mtu_change_route(struct rt6_info
*rt
, void *p_arg
)
2293 struct rt6_mtu_change_arg
*arg
= (struct rt6_mtu_change_arg
*) p_arg
;
2294 struct inet6_dev
*idev
;
2296 /* In IPv6 pmtu discovery is not optional,
2297 so that RTAX_MTU lock cannot disable it.
2298 We still use this lock to block changes
2299 caused by addrconf/ndisc.
2302 idev
= __in6_dev_get(arg
->dev
);
2306 /* For administrative MTU increase, there is no way to discover
2307 IPv6 PMTU increase, so PMTU increase should be updated here.
2308 Since RFC 1981 doesn't include administrative MTU increase
2309 update PMTU increase is a MUST. (i.e. jumbo frame)
2312 If new MTU is less than route PMTU, this new MTU will be the
2313 lowest MTU in the path, update the route PMTU to reflect PMTU
2314 decreases; if new MTU is greater than route PMTU, and the
2315 old MTU is the lowest MTU in the path, update the route PMTU
2316 to reflect the increase. In this case if the other nodes' MTU
2317 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2320 if (rt
->dst
.dev
== arg
->dev
&&
2321 !dst_metric_locked(&rt
->dst
, RTAX_MTU
) &&
2322 (dst_mtu(&rt
->dst
) >= arg
->mtu
||
2323 (dst_mtu(&rt
->dst
) < arg
->mtu
&&
2324 dst_mtu(&rt
->dst
) == idev
->cnf
.mtu6
))) {
2325 dst_metric_set(&rt
->dst
, RTAX_MTU
, arg
->mtu
);
2330 void rt6_mtu_change(struct net_device
*dev
, unsigned int mtu
)
2332 struct rt6_mtu_change_arg arg
= {
2337 fib6_clean_all(dev_net(dev
), rt6_mtu_change_route
, &arg
);
2340 static const struct nla_policy rtm_ipv6_policy
[RTA_MAX
+1] = {
2341 [RTA_GATEWAY
] = { .len
= sizeof(struct in6_addr
) },
2342 [RTA_OIF
] = { .type
= NLA_U32
},
2343 [RTA_IIF
] = { .type
= NLA_U32
},
2344 [RTA_PRIORITY
] = { .type
= NLA_U32
},
2345 [RTA_METRICS
] = { .type
= NLA_NESTED
},
2346 [RTA_MULTIPATH
] = { .len
= sizeof(struct rtnexthop
) },
2349 static int rtm_to_fib6_config(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
2350 struct fib6_config
*cfg
)
2353 struct nlattr
*tb
[RTA_MAX
+1];
2356 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2361 rtm
= nlmsg_data(nlh
);
2362 memset(cfg
, 0, sizeof(*cfg
));
2364 cfg
->fc_table
= rtm
->rtm_table
;
2365 cfg
->fc_dst_len
= rtm
->rtm_dst_len
;
2366 cfg
->fc_src_len
= rtm
->rtm_src_len
;
2367 cfg
->fc_flags
= RTF_UP
;
2368 cfg
->fc_protocol
= rtm
->rtm_protocol
;
2369 cfg
->fc_type
= rtm
->rtm_type
;
2371 if (rtm
->rtm_type
== RTN_UNREACHABLE
||
2372 rtm
->rtm_type
== RTN_BLACKHOLE
||
2373 rtm
->rtm_type
== RTN_PROHIBIT
||
2374 rtm
->rtm_type
== RTN_THROW
)
2375 cfg
->fc_flags
|= RTF_REJECT
;
2377 if (rtm
->rtm_type
== RTN_LOCAL
)
2378 cfg
->fc_flags
|= RTF_LOCAL
;
2380 cfg
->fc_nlinfo
.portid
= NETLINK_CB(skb
).portid
;
2381 cfg
->fc_nlinfo
.nlh
= nlh
;
2382 cfg
->fc_nlinfo
.nl_net
= sock_net(skb
->sk
);
2384 if (tb
[RTA_GATEWAY
]) {
2385 nla_memcpy(&cfg
->fc_gateway
, tb
[RTA_GATEWAY
], 16);
2386 cfg
->fc_flags
|= RTF_GATEWAY
;
2390 int plen
= (rtm
->rtm_dst_len
+ 7) >> 3;
2392 if (nla_len(tb
[RTA_DST
]) < plen
)
2395 nla_memcpy(&cfg
->fc_dst
, tb
[RTA_DST
], plen
);
2399 int plen
= (rtm
->rtm_src_len
+ 7) >> 3;
2401 if (nla_len(tb
[RTA_SRC
]) < plen
)
2404 nla_memcpy(&cfg
->fc_src
, tb
[RTA_SRC
], plen
);
2407 if (tb
[RTA_PREFSRC
])
2408 nla_memcpy(&cfg
->fc_prefsrc
, tb
[RTA_PREFSRC
], 16);
2411 cfg
->fc_ifindex
= nla_get_u32(tb
[RTA_OIF
]);
2413 if (tb
[RTA_PRIORITY
])
2414 cfg
->fc_metric
= nla_get_u32(tb
[RTA_PRIORITY
]);
2416 if (tb
[RTA_METRICS
]) {
2417 cfg
->fc_mx
= nla_data(tb
[RTA_METRICS
]);
2418 cfg
->fc_mx_len
= nla_len(tb
[RTA_METRICS
]);
2422 cfg
->fc_table
= nla_get_u32(tb
[RTA_TABLE
]);
2424 if (tb
[RTA_MULTIPATH
]) {
2425 cfg
->fc_mp
= nla_data(tb
[RTA_MULTIPATH
]);
2426 cfg
->fc_mp_len
= nla_len(tb
[RTA_MULTIPATH
]);
2434 static int ip6_route_multipath(struct fib6_config
*cfg
, int add
)
2436 struct fib6_config r_cfg
;
2437 struct rtnexthop
*rtnh
;
2440 int err
= 0, last_err
= 0;
2443 rtnh
= (struct rtnexthop
*)cfg
->fc_mp
;
2444 remaining
= cfg
->fc_mp_len
;
2446 /* Parse a Multipath Entry */
2447 while (rtnh_ok(rtnh
, remaining
)) {
2448 memcpy(&r_cfg
, cfg
, sizeof(*cfg
));
2449 if (rtnh
->rtnh_ifindex
)
2450 r_cfg
.fc_ifindex
= rtnh
->rtnh_ifindex
;
2452 attrlen
= rtnh_attrlen(rtnh
);
2454 struct nlattr
*nla
, *attrs
= rtnh_attrs(rtnh
);
2456 nla
= nla_find(attrs
, attrlen
, RTA_GATEWAY
);
2458 nla_memcpy(&r_cfg
.fc_gateway
, nla
, 16);
2459 r_cfg
.fc_flags
|= RTF_GATEWAY
;
2462 err
= add
? ip6_route_add(&r_cfg
) : ip6_route_del(&r_cfg
);
2465 /* If we are trying to remove a route, do not stop the
2466 * loop when ip6_route_del() fails (because next hop is
2467 * already gone), we should try to remove all next hops.
2470 /* If add fails, we should try to delete all
2471 * next hops that have been already added.
2477 /* Because each route is added like a single route we remove
2478 * this flag after the first nexthop (if there is a collision,
2479 * we have already fail to add the first nexthop:
2480 * fib6_add_rt2node() has reject it).
2482 cfg
->fc_nlinfo
.nlh
->nlmsg_flags
&= ~NLM_F_EXCL
;
2483 rtnh
= rtnh_next(rtnh
, &remaining
);
2489 static int inet6_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
* nlh
)
2491 struct fib6_config cfg
;
2494 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2499 return ip6_route_multipath(&cfg
, 0);
2501 return ip6_route_del(&cfg
);
2504 static int inet6_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
* nlh
)
2506 struct fib6_config cfg
;
2509 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2514 return ip6_route_multipath(&cfg
, 1);
2516 return ip6_route_add(&cfg
);
2519 static inline size_t rt6_nlmsg_size(void)
2521 return NLMSG_ALIGN(sizeof(struct rtmsg
))
2522 + nla_total_size(16) /* RTA_SRC */
2523 + nla_total_size(16) /* RTA_DST */
2524 + nla_total_size(16) /* RTA_GATEWAY */
2525 + nla_total_size(16) /* RTA_PREFSRC */
2526 + nla_total_size(4) /* RTA_TABLE */
2527 + nla_total_size(4) /* RTA_IIF */
2528 + nla_total_size(4) /* RTA_OIF */
2529 + nla_total_size(4) /* RTA_PRIORITY */
2530 + RTAX_MAX
* nla_total_size(4) /* RTA_METRICS */
2531 + nla_total_size(sizeof(struct rta_cacheinfo
));
2534 static int rt6_fill_node(struct net
*net
,
2535 struct sk_buff
*skb
, struct rt6_info
*rt
,
2536 struct in6_addr
*dst
, struct in6_addr
*src
,
2537 int iif
, int type
, u32 portid
, u32 seq
,
2538 int prefix
, int nowait
, unsigned int flags
)
2541 struct nlmsghdr
*nlh
;
2545 if (prefix
) { /* user wants prefix routes only */
2546 if (!(rt
->rt6i_flags
& RTF_PREFIX_RT
)) {
2547 /* success since this is not a prefix route */
2552 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*rtm
), flags
);
2556 rtm
= nlmsg_data(nlh
);
2557 rtm
->rtm_family
= AF_INET6
;
2558 rtm
->rtm_dst_len
= rt
->rt6i_dst
.plen
;
2559 rtm
->rtm_src_len
= rt
->rt6i_src
.plen
;
2562 table
= rt
->rt6i_table
->tb6_id
;
2564 table
= RT6_TABLE_UNSPEC
;
2565 rtm
->rtm_table
= table
;
2566 if (nla_put_u32(skb
, RTA_TABLE
, table
))
2567 goto nla_put_failure
;
2568 if (rt
->rt6i_flags
& RTF_REJECT
) {
2569 switch (rt
->dst
.error
) {
2571 rtm
->rtm_type
= RTN_BLACKHOLE
;
2574 rtm
->rtm_type
= RTN_PROHIBIT
;
2577 rtm
->rtm_type
= RTN_THROW
;
2580 rtm
->rtm_type
= RTN_UNREACHABLE
;
2584 else if (rt
->rt6i_flags
& RTF_LOCAL
)
2585 rtm
->rtm_type
= RTN_LOCAL
;
2586 else if (rt
->dst
.dev
&& (rt
->dst
.dev
->flags
& IFF_LOOPBACK
))
2587 rtm
->rtm_type
= RTN_LOCAL
;
2589 rtm
->rtm_type
= RTN_UNICAST
;
2591 rtm
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2592 rtm
->rtm_protocol
= rt
->rt6i_protocol
;
2593 if (rt
->rt6i_flags
& RTF_DYNAMIC
)
2594 rtm
->rtm_protocol
= RTPROT_REDIRECT
;
2595 else if (rt
->rt6i_flags
& RTF_ADDRCONF
) {
2596 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ROUTEINFO
))
2597 rtm
->rtm_protocol
= RTPROT_RA
;
2599 rtm
->rtm_protocol
= RTPROT_KERNEL
;
2602 if (rt
->rt6i_flags
& RTF_CACHE
)
2603 rtm
->rtm_flags
|= RTM_F_CLONED
;
2606 if (nla_put(skb
, RTA_DST
, 16, dst
))
2607 goto nla_put_failure
;
2608 rtm
->rtm_dst_len
= 128;
2609 } else if (rtm
->rtm_dst_len
)
2610 if (nla_put(skb
, RTA_DST
, 16, &rt
->rt6i_dst
.addr
))
2611 goto nla_put_failure
;
2612 #ifdef CONFIG_IPV6_SUBTREES
2614 if (nla_put(skb
, RTA_SRC
, 16, src
))
2615 goto nla_put_failure
;
2616 rtm
->rtm_src_len
= 128;
2617 } else if (rtm
->rtm_src_len
&&
2618 nla_put(skb
, RTA_SRC
, 16, &rt
->rt6i_src
.addr
))
2619 goto nla_put_failure
;
2622 #ifdef CONFIG_IPV6_MROUTE
2623 if (ipv6_addr_is_multicast(&rt
->rt6i_dst
.addr
)) {
2624 int err
= ip6mr_get_route(net
, skb
, rtm
, nowait
);
2629 goto nla_put_failure
;
2631 if (err
== -EMSGSIZE
)
2632 goto nla_put_failure
;
2637 if (nla_put_u32(skb
, RTA_IIF
, iif
))
2638 goto nla_put_failure
;
2640 struct in6_addr saddr_buf
;
2641 if (ip6_route_get_saddr(net
, rt
, dst
, 0, &saddr_buf
) == 0 &&
2642 nla_put(skb
, RTA_PREFSRC
, 16, &saddr_buf
))
2643 goto nla_put_failure
;
2646 if (rt
->rt6i_prefsrc
.plen
) {
2647 struct in6_addr saddr_buf
;
2648 saddr_buf
= rt
->rt6i_prefsrc
.addr
;
2649 if (nla_put(skb
, RTA_PREFSRC
, 16, &saddr_buf
))
2650 goto nla_put_failure
;
2653 if (rtnetlink_put_metrics(skb
, dst_metrics_ptr(&rt
->dst
)) < 0)
2654 goto nla_put_failure
;
2656 if (rt
->rt6i_flags
& RTF_GATEWAY
) {
2657 if (nla_put(skb
, RTA_GATEWAY
, 16, &rt
->rt6i_gateway
) < 0)
2658 goto nla_put_failure
;
2662 nla_put_u32(skb
, RTA_OIF
, rt
->dst
.dev
->ifindex
))
2663 goto nla_put_failure
;
2664 if (nla_put_u32(skb
, RTA_PRIORITY
, rt
->rt6i_metric
))
2665 goto nla_put_failure
;
2667 expires
= (rt
->rt6i_flags
& RTF_EXPIRES
) ? rt
->dst
.expires
- jiffies
: 0;
2669 if (rtnl_put_cacheinfo(skb
, &rt
->dst
, 0, expires
, rt
->dst
.error
) < 0)
2670 goto nla_put_failure
;
2672 return nlmsg_end(skb
, nlh
);
2675 nlmsg_cancel(skb
, nlh
);
2679 int rt6_dump_route(struct rt6_info
*rt
, void *p_arg
)
2681 struct rt6_rtnl_dump_arg
*arg
= (struct rt6_rtnl_dump_arg
*) p_arg
;
2684 if (nlmsg_len(arg
->cb
->nlh
) >= sizeof(struct rtmsg
)) {
2685 struct rtmsg
*rtm
= nlmsg_data(arg
->cb
->nlh
);
2686 prefix
= (rtm
->rtm_flags
& RTM_F_PREFIX
) != 0;
2690 return rt6_fill_node(arg
->net
,
2691 arg
->skb
, rt
, NULL
, NULL
, 0, RTM_NEWROUTE
,
2692 NETLINK_CB(arg
->cb
->skb
).portid
, arg
->cb
->nlh
->nlmsg_seq
,
2693 prefix
, 0, NLM_F_MULTI
);
2696 static int inet6_rtm_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
* nlh
)
2698 struct net
*net
= sock_net(in_skb
->sk
);
2699 struct nlattr
*tb
[RTA_MAX
+1];
2700 struct rt6_info
*rt
;
2701 struct sk_buff
*skb
;
2704 int err
, iif
= 0, oif
= 0;
2706 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2711 memset(&fl6
, 0, sizeof(fl6
));
2714 if (nla_len(tb
[RTA_SRC
]) < sizeof(struct in6_addr
))
2717 fl6
.saddr
= *(struct in6_addr
*)nla_data(tb
[RTA_SRC
]);
2721 if (nla_len(tb
[RTA_DST
]) < sizeof(struct in6_addr
))
2724 fl6
.daddr
= *(struct in6_addr
*)nla_data(tb
[RTA_DST
]);
2728 iif
= nla_get_u32(tb
[RTA_IIF
]);
2731 oif
= nla_get_u32(tb
[RTA_OIF
]);
2734 fl6
.flowi6_mark
= nla_get_u32(tb
[RTA_MARK
]);
2737 struct net_device
*dev
;
2740 dev
= __dev_get_by_index(net
, iif
);
2746 fl6
.flowi6_iif
= iif
;
2748 if (!ipv6_addr_any(&fl6
.saddr
))
2749 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
2751 rt
= (struct rt6_info
*)ip6_route_input_lookup(net
, dev
, &fl6
,
2754 fl6
.flowi6_oif
= oif
;
2756 rt
= (struct rt6_info
*)ip6_route_output(net
, NULL
, &fl6
);
2759 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
2766 /* Reserve room for dummy headers, this skb can pass
2767 through good chunk of routing engine.
2769 skb_reset_mac_header(skb
);
2770 skb_reserve(skb
, MAX_HEADER
+ sizeof(struct ipv6hdr
));
2772 skb_dst_set(skb
, &rt
->dst
);
2774 err
= rt6_fill_node(net
, skb
, rt
, &fl6
.daddr
, &fl6
.saddr
, iif
,
2775 RTM_NEWROUTE
, NETLINK_CB(in_skb
).portid
,
2776 nlh
->nlmsg_seq
, 0, 0, 0);
2782 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).portid
);
2787 void inet6_rt_notify(int event
, struct rt6_info
*rt
, struct nl_info
*info
)
2789 struct sk_buff
*skb
;
2790 struct net
*net
= info
->nl_net
;
2795 seq
= info
->nlh
? info
->nlh
->nlmsg_seq
: 0;
2797 skb
= nlmsg_new(rt6_nlmsg_size(), gfp_any());
2801 err
= rt6_fill_node(net
, skb
, rt
, NULL
, NULL
, 0,
2802 event
, info
->portid
, seq
, 0, 0, 0);
2804 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2805 WARN_ON(err
== -EMSGSIZE
);
2809 rtnl_notify(skb
, net
, info
->portid
, RTNLGRP_IPV6_ROUTE
,
2810 info
->nlh
, gfp_any());
2814 rtnl_set_sk_err(net
, RTNLGRP_IPV6_ROUTE
, err
);
2817 static int ip6_route_dev_notify(struct notifier_block
*this,
2818 unsigned long event
, void *ptr
)
2820 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
2821 struct net
*net
= dev_net(dev
);
2823 if (event
== NETDEV_REGISTER
&& (dev
->flags
& IFF_LOOPBACK
)) {
2824 net
->ipv6
.ip6_null_entry
->dst
.dev
= dev
;
2825 net
->ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(dev
);
2826 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2827 net
->ipv6
.ip6_prohibit_entry
->dst
.dev
= dev
;
2828 net
->ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(dev
);
2829 net
->ipv6
.ip6_blk_hole_entry
->dst
.dev
= dev
;
2830 net
->ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(dev
);
2841 #ifdef CONFIG_PROC_FS
2843 static const struct file_operations ipv6_route_proc_fops
= {
2844 .owner
= THIS_MODULE
,
2845 .open
= ipv6_route_open
,
2847 .llseek
= seq_lseek
,
2848 .release
= seq_release_net
,
2851 static int rt6_stats_seq_show(struct seq_file
*seq
, void *v
)
2853 struct net
*net
= (struct net
*)seq
->private;
2854 seq_printf(seq
, "%04x %04x %04x %04x %04x %04x %04x\n",
2855 net
->ipv6
.rt6_stats
->fib_nodes
,
2856 net
->ipv6
.rt6_stats
->fib_route_nodes
,
2857 net
->ipv6
.rt6_stats
->fib_rt_alloc
,
2858 net
->ipv6
.rt6_stats
->fib_rt_entries
,
2859 net
->ipv6
.rt6_stats
->fib_rt_cache
,
2860 dst_entries_get_slow(&net
->ipv6
.ip6_dst_ops
),
2861 net
->ipv6
.rt6_stats
->fib_discarded_routes
);
2866 static int rt6_stats_seq_open(struct inode
*inode
, struct file
*file
)
2868 return single_open_net(inode
, file
, rt6_stats_seq_show
);
2871 static const struct file_operations rt6_stats_seq_fops
= {
2872 .owner
= THIS_MODULE
,
2873 .open
= rt6_stats_seq_open
,
2875 .llseek
= seq_lseek
,
2876 .release
= single_release_net
,
2878 #endif /* CONFIG_PROC_FS */
2880 #ifdef CONFIG_SYSCTL
2883 int ipv6_sysctl_rtcache_flush(struct ctl_table
*ctl
, int write
,
2884 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
2891 net
= (struct net
*)ctl
->extra1
;
2892 delay
= net
->ipv6
.sysctl
.flush_delay
;
2893 proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
2894 fib6_run_gc(delay
<= 0 ? 0 : (unsigned long)delay
, net
, delay
> 0);
2898 struct ctl_table ipv6_route_table_template
[] = {
2900 .procname
= "flush",
2901 .data
= &init_net
.ipv6
.sysctl
.flush_delay
,
2902 .maxlen
= sizeof(int),
2904 .proc_handler
= ipv6_sysctl_rtcache_flush
2907 .procname
= "gc_thresh",
2908 .data
= &ip6_dst_ops_template
.gc_thresh
,
2909 .maxlen
= sizeof(int),
2911 .proc_handler
= proc_dointvec
,
2914 .procname
= "max_size",
2915 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_max_size
,
2916 .maxlen
= sizeof(int),
2918 .proc_handler
= proc_dointvec
,
2921 .procname
= "gc_min_interval",
2922 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
2923 .maxlen
= sizeof(int),
2925 .proc_handler
= proc_dointvec_jiffies
,
2928 .procname
= "gc_timeout",
2929 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_timeout
,
2930 .maxlen
= sizeof(int),
2932 .proc_handler
= proc_dointvec_jiffies
,
2935 .procname
= "gc_interval",
2936 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_interval
,
2937 .maxlen
= sizeof(int),
2939 .proc_handler
= proc_dointvec_jiffies
,
2942 .procname
= "gc_elasticity",
2943 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_elasticity
,
2944 .maxlen
= sizeof(int),
2946 .proc_handler
= proc_dointvec
,
2949 .procname
= "mtu_expires",
2950 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_mtu_expires
,
2951 .maxlen
= sizeof(int),
2953 .proc_handler
= proc_dointvec_jiffies
,
2956 .procname
= "min_adv_mss",
2957 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_min_advmss
,
2958 .maxlen
= sizeof(int),
2960 .proc_handler
= proc_dointvec
,
2963 .procname
= "gc_min_interval_ms",
2964 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
2965 .maxlen
= sizeof(int),
2967 .proc_handler
= proc_dointvec_ms_jiffies
,
2972 struct ctl_table
* __net_init
ipv6_route_sysctl_init(struct net
*net
)
2974 struct ctl_table
*table
;
2976 table
= kmemdup(ipv6_route_table_template
,
2977 sizeof(ipv6_route_table_template
),
2981 table
[0].data
= &net
->ipv6
.sysctl
.flush_delay
;
2982 table
[0].extra1
= net
;
2983 table
[1].data
= &net
->ipv6
.ip6_dst_ops
.gc_thresh
;
2984 table
[2].data
= &net
->ipv6
.sysctl
.ip6_rt_max_size
;
2985 table
[3].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
2986 table
[4].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
2987 table
[5].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
2988 table
[6].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
2989 table
[7].data
= &net
->ipv6
.sysctl
.ip6_rt_mtu_expires
;
2990 table
[8].data
= &net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
2991 table
[9].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
2993 /* Don't export sysctls to unprivileged users */
2994 if (net
->user_ns
!= &init_user_ns
)
2995 table
[0].procname
= NULL
;
3002 static int __net_init
ip6_route_net_init(struct net
*net
)
3006 memcpy(&net
->ipv6
.ip6_dst_ops
, &ip6_dst_ops_template
,
3007 sizeof(net
->ipv6
.ip6_dst_ops
));
3009 if (dst_entries_init(&net
->ipv6
.ip6_dst_ops
) < 0)
3010 goto out_ip6_dst_ops
;
3012 net
->ipv6
.ip6_null_entry
= kmemdup(&ip6_null_entry_template
,
3013 sizeof(*net
->ipv6
.ip6_null_entry
),
3015 if (!net
->ipv6
.ip6_null_entry
)
3016 goto out_ip6_dst_entries
;
3017 net
->ipv6
.ip6_null_entry
->dst
.path
=
3018 (struct dst_entry
*)net
->ipv6
.ip6_null_entry
;
3019 net
->ipv6
.ip6_null_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3020 dst_init_metrics(&net
->ipv6
.ip6_null_entry
->dst
,
3021 ip6_template_metrics
, true);
3023 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3024 net
->ipv6
.ip6_prohibit_entry
= kmemdup(&ip6_prohibit_entry_template
,
3025 sizeof(*net
->ipv6
.ip6_prohibit_entry
),
3027 if (!net
->ipv6
.ip6_prohibit_entry
)
3028 goto out_ip6_null_entry
;
3029 net
->ipv6
.ip6_prohibit_entry
->dst
.path
=
3030 (struct dst_entry
*)net
->ipv6
.ip6_prohibit_entry
;
3031 net
->ipv6
.ip6_prohibit_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3032 dst_init_metrics(&net
->ipv6
.ip6_prohibit_entry
->dst
,
3033 ip6_template_metrics
, true);
3035 net
->ipv6
.ip6_blk_hole_entry
= kmemdup(&ip6_blk_hole_entry_template
,
3036 sizeof(*net
->ipv6
.ip6_blk_hole_entry
),
3038 if (!net
->ipv6
.ip6_blk_hole_entry
)
3039 goto out_ip6_prohibit_entry
;
3040 net
->ipv6
.ip6_blk_hole_entry
->dst
.path
=
3041 (struct dst_entry
*)net
->ipv6
.ip6_blk_hole_entry
;
3042 net
->ipv6
.ip6_blk_hole_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3043 dst_init_metrics(&net
->ipv6
.ip6_blk_hole_entry
->dst
,
3044 ip6_template_metrics
, true);
3047 net
->ipv6
.sysctl
.flush_delay
= 0;
3048 net
->ipv6
.sysctl
.ip6_rt_max_size
= 4096;
3049 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
= HZ
/ 2;
3050 net
->ipv6
.sysctl
.ip6_rt_gc_timeout
= 60*HZ
;
3051 net
->ipv6
.sysctl
.ip6_rt_gc_interval
= 30*HZ
;
3052 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
= 9;
3053 net
->ipv6
.sysctl
.ip6_rt_mtu_expires
= 10*60*HZ
;
3054 net
->ipv6
.sysctl
.ip6_rt_min_advmss
= IPV6_MIN_MTU
- 20 - 40;
3056 net
->ipv6
.ip6_rt_gc_expire
= 30*HZ
;
3062 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3063 out_ip6_prohibit_entry
:
3064 kfree(net
->ipv6
.ip6_prohibit_entry
);
3066 kfree(net
->ipv6
.ip6_null_entry
);
3068 out_ip6_dst_entries
:
3069 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
3074 static void __net_exit
ip6_route_net_exit(struct net
*net
)
3076 kfree(net
->ipv6
.ip6_null_entry
);
3077 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3078 kfree(net
->ipv6
.ip6_prohibit_entry
);
3079 kfree(net
->ipv6
.ip6_blk_hole_entry
);
3081 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
3084 static int __net_init
ip6_route_net_init_late(struct net
*net
)
3086 #ifdef CONFIG_PROC_FS
3087 proc_create("ipv6_route", 0, net
->proc_net
, &ipv6_route_proc_fops
);
3088 proc_create("rt6_stats", S_IRUGO
, net
->proc_net
, &rt6_stats_seq_fops
);
3093 static void __net_exit
ip6_route_net_exit_late(struct net
*net
)
3095 #ifdef CONFIG_PROC_FS
3096 remove_proc_entry("ipv6_route", net
->proc_net
);
3097 remove_proc_entry("rt6_stats", net
->proc_net
);
3101 static struct pernet_operations ip6_route_net_ops
= {
3102 .init
= ip6_route_net_init
,
3103 .exit
= ip6_route_net_exit
,
3106 static int __net_init
ipv6_inetpeer_init(struct net
*net
)
3108 struct inet_peer_base
*bp
= kmalloc(sizeof(*bp
), GFP_KERNEL
);
3112 inet_peer_base_init(bp
);
3113 net
->ipv6
.peers
= bp
;
3117 static void __net_exit
ipv6_inetpeer_exit(struct net
*net
)
3119 struct inet_peer_base
*bp
= net
->ipv6
.peers
;
3121 net
->ipv6
.peers
= NULL
;
3122 inetpeer_invalidate_tree(bp
);
3126 static struct pernet_operations ipv6_inetpeer_ops
= {
3127 .init
= ipv6_inetpeer_init
,
3128 .exit
= ipv6_inetpeer_exit
,
3131 static struct pernet_operations ip6_route_net_late_ops
= {
3132 .init
= ip6_route_net_init_late
,
3133 .exit
= ip6_route_net_exit_late
,
3136 static struct notifier_block ip6_route_dev_notifier
= {
3137 .notifier_call
= ip6_route_dev_notify
,
3141 int __init
ip6_route_init(void)
3146 ip6_dst_ops_template
.kmem_cachep
=
3147 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info
), 0,
3148 SLAB_HWCACHE_ALIGN
, NULL
);
3149 if (!ip6_dst_ops_template
.kmem_cachep
)
3152 ret
= dst_entries_init(&ip6_dst_blackhole_ops
);
3154 goto out_kmem_cache
;
3156 ret
= register_pernet_subsys(&ipv6_inetpeer_ops
);
3158 goto out_dst_entries
;
3160 ret
= register_pernet_subsys(&ip6_route_net_ops
);
3162 goto out_register_inetpeer
;
3164 ip6_dst_blackhole_ops
.kmem_cachep
= ip6_dst_ops_template
.kmem_cachep
;
3166 /* Registering of the loopback is done before this portion of code,
3167 * the loopback reference in rt6_info will not be taken, do it
3168 * manually for init_net */
3169 init_net
.ipv6
.ip6_null_entry
->dst
.dev
= init_net
.loopback_dev
;
3170 init_net
.ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3171 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3172 init_net
.ipv6
.ip6_prohibit_entry
->dst
.dev
= init_net
.loopback_dev
;
3173 init_net
.ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3174 init_net
.ipv6
.ip6_blk_hole_entry
->dst
.dev
= init_net
.loopback_dev
;
3175 init_net
.ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3179 goto out_register_subsys
;
3185 ret
= fib6_rules_init();
3189 ret
= register_pernet_subsys(&ip6_route_net_late_ops
);
3191 goto fib6_rules_init
;
3194 if (__rtnl_register(PF_INET6
, RTM_NEWROUTE
, inet6_rtm_newroute
, NULL
, NULL
) ||
3195 __rtnl_register(PF_INET6
, RTM_DELROUTE
, inet6_rtm_delroute
, NULL
, NULL
) ||
3196 __rtnl_register(PF_INET6
, RTM_GETROUTE
, inet6_rtm_getroute
, NULL
, NULL
))
3197 goto out_register_late_subsys
;
3199 ret
= register_netdevice_notifier(&ip6_route_dev_notifier
);
3201 goto out_register_late_subsys
;
3206 out_register_late_subsys
:
3207 unregister_pernet_subsys(&ip6_route_net_late_ops
);
3209 fib6_rules_cleanup();
3214 out_register_subsys
:
3215 unregister_pernet_subsys(&ip6_route_net_ops
);
3216 out_register_inetpeer
:
3217 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
3219 dst_entries_destroy(&ip6_dst_blackhole_ops
);
3221 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);
3225 void ip6_route_cleanup(void)
3227 unregister_netdevice_notifier(&ip6_route_dev_notifier
);
3228 unregister_pernet_subsys(&ip6_route_net_late_ops
);
3229 fib6_rules_cleanup();
3232 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
3233 unregister_pernet_subsys(&ip6_route_net_ops
);
3234 dst_entries_destroy(&ip6_dst_blackhole_ops
);
3235 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);