2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * IPv4 Forwarding Information Base: semantics.
8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
16 #include <linux/uaccess.h>
17 #include <linux/bitops.h>
18 #include <linux/types.h>
19 #include <linux/kernel.h>
20 #include <linux/jiffies.h>
22 #include <linux/string.h>
23 #include <linux/socket.h>
24 #include <linux/sockios.h>
25 #include <linux/errno.h>
27 #include <linux/inet.h>
28 #include <linux/inetdevice.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/proc_fs.h>
32 #include <linux/skbuff.h>
33 #include <linux/init.h>
34 #include <linux/slab.h>
35 #include <linux/netlink.h>
39 #include <net/protocol.h>
40 #include <net/route.h>
43 #include <net/ip_fib.h>
44 #include <net/netlink.h>
45 #include <net/nexthop.h>
46 #include <net/lwtunnel.h>
47 #include <net/fib_notifier.h>
49 #include "fib_lookup.h"
51 static DEFINE_SPINLOCK(fib_info_lock
);
52 static struct hlist_head
*fib_info_hash
;
53 static struct hlist_head
*fib_info_laddrhash
;
54 static unsigned int fib_info_hash_size
;
55 static unsigned int fib_info_cnt
;
57 #define DEVINDEX_HASHBITS 8
58 #define DEVINDEX_HASHSIZE (1U << DEVINDEX_HASHBITS)
59 static struct hlist_head fib_info_devhash
[DEVINDEX_HASHSIZE
];
61 #ifdef CONFIG_IP_ROUTE_MULTIPATH
63 #define for_nexthops(fi) { \
64 int nhsel; const struct fib_nh *nh; \
65 for (nhsel = 0, nh = (fi)->fib_nh; \
66 nhsel < (fi)->fib_nhs; \
69 #define change_nexthops(fi) { \
70 int nhsel; struct fib_nh *nexthop_nh; \
71 for (nhsel = 0, nexthop_nh = (struct fib_nh *)((fi)->fib_nh); \
72 nhsel < (fi)->fib_nhs; \
73 nexthop_nh++, nhsel++)
75 #else /* CONFIG_IP_ROUTE_MULTIPATH */
77 /* Hope, that gcc will optimize it to get rid of dummy loop */
79 #define for_nexthops(fi) { \
80 int nhsel; const struct fib_nh *nh = (fi)->fib_nh; \
81 for (nhsel = 0; nhsel < 1; nhsel++)
83 #define change_nexthops(fi) { \
85 struct fib_nh *nexthop_nh = (struct fib_nh *)((fi)->fib_nh); \
86 for (nhsel = 0; nhsel < 1; nhsel++)
88 #endif /* CONFIG_IP_ROUTE_MULTIPATH */
90 #define endfor_nexthops(fi) }
93 const struct fib_prop fib_props
[RTN_MAX
+ 1] = {
96 .scope
= RT_SCOPE_NOWHERE
,
100 .scope
= RT_SCOPE_UNIVERSE
,
104 .scope
= RT_SCOPE_HOST
,
108 .scope
= RT_SCOPE_LINK
,
112 .scope
= RT_SCOPE_LINK
,
116 .scope
= RT_SCOPE_UNIVERSE
,
120 .scope
= RT_SCOPE_UNIVERSE
,
122 [RTN_UNREACHABLE
] = {
123 .error
= -EHOSTUNREACH
,
124 .scope
= RT_SCOPE_UNIVERSE
,
128 .scope
= RT_SCOPE_UNIVERSE
,
132 .scope
= RT_SCOPE_UNIVERSE
,
136 .scope
= RT_SCOPE_NOWHERE
,
140 .scope
= RT_SCOPE_NOWHERE
,
144 static void rt_fibinfo_free(struct rtable __rcu
**rtp
)
146 struct rtable
*rt
= rcu_dereference_protected(*rtp
, 1);
151 /* Not even needed : RCU_INIT_POINTER(*rtp, NULL);
152 * because we waited an RCU grace period before calling
153 * free_fib_info_rcu()
156 dst_dev_put(&rt
->dst
);
157 dst_release_immediate(&rt
->dst
);
160 static void free_nh_exceptions(struct fib_nh
*nh
)
162 struct fnhe_hash_bucket
*hash
;
165 hash
= rcu_dereference_protected(nh
->nh_exceptions
, 1);
168 for (i
= 0; i
< FNHE_HASH_SIZE
; i
++) {
169 struct fib_nh_exception
*fnhe
;
171 fnhe
= rcu_dereference_protected(hash
[i
].chain
, 1);
173 struct fib_nh_exception
*next
;
175 next
= rcu_dereference_protected(fnhe
->fnhe_next
, 1);
177 rt_fibinfo_free(&fnhe
->fnhe_rth_input
);
178 rt_fibinfo_free(&fnhe
->fnhe_rth_output
);
188 static void rt_fibinfo_free_cpus(struct rtable __rcu
* __percpu
*rtp
)
195 for_each_possible_cpu(cpu
) {
198 rt
= rcu_dereference_protected(*per_cpu_ptr(rtp
, cpu
), 1);
200 dst_dev_put(&rt
->dst
);
201 dst_release_immediate(&rt
->dst
);
207 /* Release a nexthop info record */
208 static void free_fib_info_rcu(struct rcu_head
*head
)
210 struct fib_info
*fi
= container_of(head
, struct fib_info
, rcu
);
211 struct dst_metrics
*m
;
213 change_nexthops(fi
) {
214 if (nexthop_nh
->nh_dev
)
215 dev_put(nexthop_nh
->nh_dev
);
216 lwtstate_put(nexthop_nh
->nh_lwtstate
);
217 free_nh_exceptions(nexthop_nh
);
218 rt_fibinfo_free_cpus(nexthop_nh
->nh_pcpu_rth_output
);
219 rt_fibinfo_free(&nexthop_nh
->nh_rth_input
);
220 } endfor_nexthops(fi
);
223 if (m
!= &dst_default_metrics
&& refcount_dec_and_test(&m
->refcnt
))
228 void free_fib_info(struct fib_info
*fi
)
230 if (fi
->fib_dead
== 0) {
231 pr_warn("Freeing alive fib_info %p\n", fi
);
235 #ifdef CONFIG_IP_ROUTE_CLASSID
236 change_nexthops(fi
) {
237 if (nexthop_nh
->nh_tclassid
)
238 fi
->fib_net
->ipv4
.fib_num_tclassid_users
--;
239 } endfor_nexthops(fi
);
241 call_rcu(&fi
->rcu
, free_fib_info_rcu
);
243 EXPORT_SYMBOL_GPL(free_fib_info
);
245 void fib_release_info(struct fib_info
*fi
)
247 spin_lock_bh(&fib_info_lock
);
248 if (fi
&& --fi
->fib_treeref
== 0) {
249 hlist_del(&fi
->fib_hash
);
251 hlist_del(&fi
->fib_lhash
);
252 change_nexthops(fi
) {
253 if (!nexthop_nh
->nh_dev
)
255 hlist_del(&nexthop_nh
->nh_hash
);
256 } endfor_nexthops(fi
)
260 spin_unlock_bh(&fib_info_lock
);
263 static inline int nh_comp(const struct fib_info
*fi
, const struct fib_info
*ofi
)
265 const struct fib_nh
*onh
= ofi
->fib_nh
;
268 if (nh
->nh_oif
!= onh
->nh_oif
||
269 nh
->nh_gw
!= onh
->nh_gw
||
270 nh
->nh_scope
!= onh
->nh_scope
||
271 #ifdef CONFIG_IP_ROUTE_MULTIPATH
272 nh
->nh_weight
!= onh
->nh_weight
||
274 #ifdef CONFIG_IP_ROUTE_CLASSID
275 nh
->nh_tclassid
!= onh
->nh_tclassid
||
277 lwtunnel_cmp_encap(nh
->nh_lwtstate
, onh
->nh_lwtstate
) ||
278 ((nh
->nh_flags
^ onh
->nh_flags
) & ~RTNH_COMPARE_MASK
))
281 } endfor_nexthops(fi
);
285 static inline unsigned int fib_devindex_hashfn(unsigned int val
)
287 unsigned int mask
= DEVINDEX_HASHSIZE
- 1;
290 (val
>> DEVINDEX_HASHBITS
) ^
291 (val
>> (DEVINDEX_HASHBITS
* 2))) & mask
;
294 static inline unsigned int fib_info_hashfn(const struct fib_info
*fi
)
296 unsigned int mask
= (fib_info_hash_size
- 1);
297 unsigned int val
= fi
->fib_nhs
;
299 val
^= (fi
->fib_protocol
<< 8) | fi
->fib_scope
;
300 val
^= (__force u32
)fi
->fib_prefsrc
;
301 val
^= fi
->fib_priority
;
303 val
^= fib_devindex_hashfn(nh
->nh_oif
);
304 } endfor_nexthops(fi
)
306 return (val
^ (val
>> 7) ^ (val
>> 12)) & mask
;
309 static struct fib_info
*fib_find_info(const struct fib_info
*nfi
)
311 struct hlist_head
*head
;
315 hash
= fib_info_hashfn(nfi
);
316 head
= &fib_info_hash
[hash
];
318 hlist_for_each_entry(fi
, head
, fib_hash
) {
319 if (!net_eq(fi
->fib_net
, nfi
->fib_net
))
321 if (fi
->fib_nhs
!= nfi
->fib_nhs
)
323 if (nfi
->fib_protocol
== fi
->fib_protocol
&&
324 nfi
->fib_scope
== fi
->fib_scope
&&
325 nfi
->fib_prefsrc
== fi
->fib_prefsrc
&&
326 nfi
->fib_priority
== fi
->fib_priority
&&
327 nfi
->fib_type
== fi
->fib_type
&&
328 memcmp(nfi
->fib_metrics
, fi
->fib_metrics
,
329 sizeof(u32
) * RTAX_MAX
) == 0 &&
330 !((nfi
->fib_flags
^ fi
->fib_flags
) & ~RTNH_COMPARE_MASK
) &&
331 (nfi
->fib_nhs
== 0 || nh_comp(fi
, nfi
) == 0))
338 /* Check, that the gateway is already configured.
339 * Used only by redirect accept routine.
341 int ip_fib_check_default(__be32 gw
, struct net_device
*dev
)
343 struct hlist_head
*head
;
347 spin_lock(&fib_info_lock
);
349 hash
= fib_devindex_hashfn(dev
->ifindex
);
350 head
= &fib_info_devhash
[hash
];
351 hlist_for_each_entry(nh
, head
, nh_hash
) {
352 if (nh
->nh_dev
== dev
&&
354 !(nh
->nh_flags
& RTNH_F_DEAD
)) {
355 spin_unlock(&fib_info_lock
);
360 spin_unlock(&fib_info_lock
);
365 static inline size_t fib_nlmsg_size(struct fib_info
*fi
)
367 size_t payload
= NLMSG_ALIGN(sizeof(struct rtmsg
))
368 + nla_total_size(4) /* RTA_TABLE */
369 + nla_total_size(4) /* RTA_DST */
370 + nla_total_size(4) /* RTA_PRIORITY */
371 + nla_total_size(4) /* RTA_PREFSRC */
372 + nla_total_size(TCP_CA_NAME_MAX
); /* RTAX_CC_ALGO */
374 /* space for nested metrics */
375 payload
+= nla_total_size((RTAX_MAX
* nla_total_size(4)));
378 size_t nh_encapsize
= 0;
379 /* Also handles the special case fib_nhs == 1 */
381 /* each nexthop is packed in an attribute */
382 size_t nhsize
= nla_total_size(sizeof(struct rtnexthop
));
384 /* may contain flow and gateway attribute */
385 nhsize
+= 2 * nla_total_size(4);
387 /* grab encap info */
389 if (nh
->nh_lwtstate
) {
391 nh_encapsize
+= lwtunnel_get_encap_size(
394 nh_encapsize
+= nla_total_size(2);
396 } endfor_nexthops(fi
);
398 /* all nexthops are packed in a nested attribute */
399 payload
+= nla_total_size((fi
->fib_nhs
* nhsize
) +
407 void rtmsg_fib(int event
, __be32 key
, struct fib_alias
*fa
,
408 int dst_len
, u32 tb_id
, const struct nl_info
*info
,
409 unsigned int nlm_flags
)
412 u32 seq
= info
->nlh
? info
->nlh
->nlmsg_seq
: 0;
415 skb
= nlmsg_new(fib_nlmsg_size(fa
->fa_info
), GFP_KERNEL
);
419 err
= fib_dump_info(skb
, info
->portid
, seq
, event
, tb_id
,
420 fa
->fa_type
, key
, dst_len
,
421 fa
->fa_tos
, fa
->fa_info
, nlm_flags
);
423 /* -EMSGSIZE implies BUG in fib_nlmsg_size() */
424 WARN_ON(err
== -EMSGSIZE
);
428 rtnl_notify(skb
, info
->nl_net
, info
->portid
, RTNLGRP_IPV4_ROUTE
,
429 info
->nlh
, GFP_KERNEL
);
433 rtnl_set_sk_err(info
->nl_net
, RTNLGRP_IPV4_ROUTE
, err
);
436 static int fib_detect_death(struct fib_info
*fi
, int order
,
437 struct fib_info
**last_resort
, int *last_idx
,
441 int state
= NUD_NONE
;
443 n
= neigh_lookup(&arp_tbl
, &fi
->fib_nh
[0].nh_gw
, fi
->fib_dev
);
445 state
= n
->nud_state
;
450 if (state
== NUD_REACHABLE
)
452 if ((state
& NUD_VALID
) && order
!= dflt
)
454 if ((state
& NUD_VALID
) ||
455 (*last_idx
< 0 && order
> dflt
&& state
!= NUD_INCOMPLETE
)) {
462 #ifdef CONFIG_IP_ROUTE_MULTIPATH
464 static int fib_count_nexthops(struct rtnexthop
*rtnh
, int remaining
,
465 struct netlink_ext_ack
*extack
)
469 while (rtnh_ok(rtnh
, remaining
)) {
471 rtnh
= rtnh_next(rtnh
, &remaining
);
474 /* leftover implies invalid nexthop configuration, discard it */
476 NL_SET_ERR_MSG(extack
,
477 "Invalid nexthop configuration - extra data after nexthops");
484 static int fib_get_nhs(struct fib_info
*fi
, struct rtnexthop
*rtnh
,
485 int remaining
, struct fib_config
*cfg
,
486 struct netlink_ext_ack
*extack
)
490 change_nexthops(fi
) {
493 if (!rtnh_ok(rtnh
, remaining
)) {
494 NL_SET_ERR_MSG(extack
,
495 "Invalid nexthop configuration - extra data after nexthop");
499 if (rtnh
->rtnh_flags
& (RTNH_F_DEAD
| RTNH_F_LINKDOWN
)) {
500 NL_SET_ERR_MSG(extack
,
501 "Invalid flags for nexthop - can not contain DEAD or LINKDOWN");
505 nexthop_nh
->nh_flags
=
506 (cfg
->fc_flags
& ~0xFF) | rtnh
->rtnh_flags
;
507 nexthop_nh
->nh_oif
= rtnh
->rtnh_ifindex
;
508 nexthop_nh
->nh_weight
= rtnh
->rtnh_hops
+ 1;
510 attrlen
= rtnh_attrlen(rtnh
);
512 struct nlattr
*nla
, *attrs
= rtnh_attrs(rtnh
);
514 nla
= nla_find(attrs
, attrlen
, RTA_GATEWAY
);
515 nexthop_nh
->nh_gw
= nla
? nla_get_in_addr(nla
) : 0;
516 #ifdef CONFIG_IP_ROUTE_CLASSID
517 nla
= nla_find(attrs
, attrlen
, RTA_FLOW
);
518 nexthop_nh
->nh_tclassid
= nla
? nla_get_u32(nla
) : 0;
519 if (nexthop_nh
->nh_tclassid
)
520 fi
->fib_net
->ipv4
.fib_num_tclassid_users
++;
522 nla
= nla_find(attrs
, attrlen
, RTA_ENCAP
);
524 struct lwtunnel_state
*lwtstate
;
525 struct nlattr
*nla_entype
;
527 nla_entype
= nla_find(attrs
, attrlen
,
530 NL_SET_BAD_ATTR(extack
, nla
);
531 NL_SET_ERR_MSG(extack
,
532 "Encap type is missing");
536 ret
= lwtunnel_build_state(nla_get_u16(
542 nexthop_nh
->nh_lwtstate
=
543 lwtstate_get(lwtstate
);
547 rtnh
= rtnh_next(rtnh
, &remaining
);
548 } endfor_nexthops(fi
);
559 static void fib_rebalance(struct fib_info
*fi
)
563 struct in_device
*in_dev
;
570 if (nh
->nh_flags
& RTNH_F_DEAD
)
573 in_dev
= __in_dev_get_rtnl(nh
->nh_dev
);
576 IN_DEV_IGNORE_ROUTES_WITH_LINKDOWN(in_dev
) &&
577 nh
->nh_flags
& RTNH_F_LINKDOWN
)
580 total
+= nh
->nh_weight
;
581 } endfor_nexthops(fi
);
584 change_nexthops(fi
) {
587 in_dev
= __in_dev_get_rtnl(nexthop_nh
->nh_dev
);
589 if (nexthop_nh
->nh_flags
& RTNH_F_DEAD
) {
592 IN_DEV_IGNORE_ROUTES_WITH_LINKDOWN(in_dev
) &&
593 nexthop_nh
->nh_flags
& RTNH_F_LINKDOWN
) {
596 w
+= nexthop_nh
->nh_weight
;
597 upper_bound
= DIV_ROUND_CLOSEST_ULL((u64
)w
<< 31,
601 atomic_set(&nexthop_nh
->nh_upper_bound
, upper_bound
);
602 } endfor_nexthops(fi
);
604 #else /* CONFIG_IP_ROUTE_MULTIPATH */
606 #define fib_rebalance(fi) do { } while (0)
608 #endif /* CONFIG_IP_ROUTE_MULTIPATH */
610 static int fib_encap_match(u16 encap_type
,
611 struct nlattr
*encap
,
612 const struct fib_nh
*nh
,
613 const struct fib_config
*cfg
,
614 struct netlink_ext_ack
*extack
)
616 struct lwtunnel_state
*lwtstate
;
619 if (encap_type
== LWTUNNEL_ENCAP_NONE
)
622 ret
= lwtunnel_build_state(encap_type
, encap
, AF_INET
,
623 cfg
, &lwtstate
, extack
);
625 result
= lwtunnel_cmp_encap(lwtstate
, nh
->nh_lwtstate
);
626 lwtstate_free(lwtstate
);
632 int fib_nh_match(struct fib_config
*cfg
, struct fib_info
*fi
,
633 struct netlink_ext_ack
*extack
)
635 #ifdef CONFIG_IP_ROUTE_MULTIPATH
636 struct rtnexthop
*rtnh
;
640 if (cfg
->fc_priority
&& cfg
->fc_priority
!= fi
->fib_priority
)
643 if (cfg
->fc_oif
|| cfg
->fc_gw
) {
645 if (fib_encap_match(cfg
->fc_encap_type
, cfg
->fc_encap
,
646 fi
->fib_nh
, cfg
, extack
))
649 if ((!cfg
->fc_oif
|| cfg
->fc_oif
== fi
->fib_nh
->nh_oif
) &&
650 (!cfg
->fc_gw
|| cfg
->fc_gw
== fi
->fib_nh
->nh_gw
))
655 #ifdef CONFIG_IP_ROUTE_MULTIPATH
660 remaining
= cfg
->fc_mp_len
;
665 if (!rtnh_ok(rtnh
, remaining
))
668 if (rtnh
->rtnh_ifindex
&& rtnh
->rtnh_ifindex
!= nh
->nh_oif
)
671 attrlen
= rtnh_attrlen(rtnh
);
673 struct nlattr
*nla
, *attrs
= rtnh_attrs(rtnh
);
675 nla
= nla_find(attrs
, attrlen
, RTA_GATEWAY
);
676 if (nla
&& nla_get_in_addr(nla
) != nh
->nh_gw
)
678 #ifdef CONFIG_IP_ROUTE_CLASSID
679 nla
= nla_find(attrs
, attrlen
, RTA_FLOW
);
680 if (nla
&& nla_get_u32(nla
) != nh
->nh_tclassid
)
685 rtnh
= rtnh_next(rtnh
, &remaining
);
686 } endfor_nexthops(fi
);
691 bool fib_metrics_match(struct fib_config
*cfg
, struct fib_info
*fi
)
699 nla_for_each_attr(nla
, cfg
->fc_mx
, cfg
->fc_mx_len
, remaining
) {
700 int type
= nla_type(nla
);
708 if (type
== RTAX_CC_ALGO
) {
709 char tmp
[TCP_CA_NAME_MAX
];
712 nla_strlcpy(tmp
, nla
, sizeof(tmp
));
713 val
= tcp_ca_get_key_by_name(fi
->fib_net
, tmp
, &ecn_ca
);
715 val
= nla_get_u32(nla
);
718 if (fi
->fib_metrics
->metrics
[type
- 1] != val
)
730 * Semantics of nexthop is very messy by historical reasons.
731 * We have to take into account, that:
732 * a) gateway can be actually local interface address,
733 * so that gatewayed route is direct.
734 * b) gateway must be on-link address, possibly
735 * described not by an ifaddr, but also by a direct route.
736 * c) If both gateway and interface are specified, they should not
738 * d) If we use tunnel routes, gateway could be not on-link.
740 * Attempt to reconcile all of these (alas, self-contradictory) conditions
741 * results in pretty ugly and hairy code with obscure logic.
743 * I chose to generalized it instead, so that the size
744 * of code does not increase practically, but it becomes
746 * Every prefix is assigned a "scope" value: "host" is local address,
747 * "link" is direct route,
748 * [ ... "site" ... "interior" ... ]
749 * and "universe" is true gateway route with global meaning.
751 * Every prefix refers to a set of "nexthop"s (gw, oif),
752 * where gw must have narrower scope. This recursion stops
753 * when gw has LOCAL scope or if "nexthop" is declared ONLINK,
754 * which means that gw is forced to be on link.
756 * Code is still hairy, but now it is apparently logically
757 * consistent and very flexible. F.e. as by-product it allows
758 * to co-exists in peace independent exterior and interior
761 * Normally it looks as following.
763 * {universe prefix} -> (gw, oif) [scope link]
765 * |-> {link prefix} -> (gw, oif) [scope local]
767 * |-> {local prefix} (terminal node)
769 static int fib_check_nh(struct fib_config
*cfg
, struct fib_nh
*nh
,
770 struct netlink_ext_ack
*extack
)
774 struct net_device
*dev
;
776 net
= cfg
->fc_nlinfo
.nl_net
;
778 struct fib_result res
;
780 if (nh
->nh_flags
& RTNH_F_ONLINK
) {
781 unsigned int addr_type
;
783 if (cfg
->fc_scope
>= RT_SCOPE_LINK
) {
784 NL_SET_ERR_MSG(extack
,
785 "Nexthop has invalid scope");
788 dev
= __dev_get_by_index(net
, nh
->nh_oif
);
791 if (!(dev
->flags
& IFF_UP
)) {
792 NL_SET_ERR_MSG(extack
,
793 "Nexthop device is not up");
796 addr_type
= inet_addr_type_dev_table(net
, dev
, nh
->nh_gw
);
797 if (addr_type
!= RTN_UNICAST
) {
798 NL_SET_ERR_MSG(extack
,
799 "Nexthop has invalid gateway");
802 if (!netif_carrier_ok(dev
))
803 nh
->nh_flags
|= RTNH_F_LINKDOWN
;
806 nh
->nh_scope
= RT_SCOPE_LINK
;
811 struct fib_table
*tbl
= NULL
;
812 struct flowi4 fl4
= {
814 .flowi4_scope
= cfg
->fc_scope
+ 1,
815 .flowi4_oif
= nh
->nh_oif
,
816 .flowi4_iif
= LOOPBACK_IFINDEX
,
819 /* It is not necessary, but requires a bit of thinking */
820 if (fl4
.flowi4_scope
< RT_SCOPE_LINK
)
821 fl4
.flowi4_scope
= RT_SCOPE_LINK
;
824 tbl
= fib_get_table(net
, cfg
->fc_table
);
827 err
= fib_table_lookup(tbl
, &fl4
, &res
,
828 FIB_LOOKUP_IGNORE_LINKSTATE
|
831 /* on error or if no table given do full lookup. This
832 * is needed for example when nexthops are in the local
833 * table rather than the given table
836 err
= fib_lookup(net
, &fl4
, &res
,
837 FIB_LOOKUP_IGNORE_LINKSTATE
);
841 NL_SET_ERR_MSG(extack
,
842 "Nexthop has invalid gateway");
848 if (res
.type
!= RTN_UNICAST
&& res
.type
!= RTN_LOCAL
) {
849 NL_SET_ERR_MSG(extack
, "Nexthop has invalid gateway");
852 nh
->nh_scope
= res
.scope
;
853 nh
->nh_oif
= FIB_RES_OIF(res
);
854 nh
->nh_dev
= dev
= FIB_RES_DEV(res
);
856 NL_SET_ERR_MSG(extack
,
857 "No egress device for nexthop gateway");
861 if (!netif_carrier_ok(dev
))
862 nh
->nh_flags
|= RTNH_F_LINKDOWN
;
863 err
= (dev
->flags
& IFF_UP
) ? 0 : -ENETDOWN
;
865 struct in_device
*in_dev
;
867 if (nh
->nh_flags
& (RTNH_F_PERVASIVE
| RTNH_F_ONLINK
)) {
868 NL_SET_ERR_MSG(extack
,
869 "Invalid flags for nexthop - PERVASIVE and ONLINK can not be set");
874 in_dev
= inetdev_by_index(net
, nh
->nh_oif
);
878 if (!(in_dev
->dev
->flags
& IFF_UP
)) {
879 NL_SET_ERR_MSG(extack
, "Device for nexthop is not up");
882 nh
->nh_dev
= in_dev
->dev
;
883 dev_hold(nh
->nh_dev
);
884 nh
->nh_scope
= RT_SCOPE_HOST
;
885 if (!netif_carrier_ok(nh
->nh_dev
))
886 nh
->nh_flags
|= RTNH_F_LINKDOWN
;
894 static inline unsigned int fib_laddr_hashfn(__be32 val
)
896 unsigned int mask
= (fib_info_hash_size
- 1);
898 return ((__force u32
)val
^
899 ((__force u32
)val
>> 7) ^
900 ((__force u32
)val
>> 14)) & mask
;
903 static struct hlist_head
*fib_info_hash_alloc(int bytes
)
905 if (bytes
<= PAGE_SIZE
)
906 return kzalloc(bytes
, GFP_KERNEL
);
908 return (struct hlist_head
*)
909 __get_free_pages(GFP_KERNEL
| __GFP_ZERO
,
913 static void fib_info_hash_free(struct hlist_head
*hash
, int bytes
)
918 if (bytes
<= PAGE_SIZE
)
921 free_pages((unsigned long) hash
, get_order(bytes
));
924 static void fib_info_hash_move(struct hlist_head
*new_info_hash
,
925 struct hlist_head
*new_laddrhash
,
926 unsigned int new_size
)
928 struct hlist_head
*old_info_hash
, *old_laddrhash
;
929 unsigned int old_size
= fib_info_hash_size
;
930 unsigned int i
, bytes
;
932 spin_lock_bh(&fib_info_lock
);
933 old_info_hash
= fib_info_hash
;
934 old_laddrhash
= fib_info_laddrhash
;
935 fib_info_hash_size
= new_size
;
937 for (i
= 0; i
< old_size
; i
++) {
938 struct hlist_head
*head
= &fib_info_hash
[i
];
939 struct hlist_node
*n
;
942 hlist_for_each_entry_safe(fi
, n
, head
, fib_hash
) {
943 struct hlist_head
*dest
;
944 unsigned int new_hash
;
946 new_hash
= fib_info_hashfn(fi
);
947 dest
= &new_info_hash
[new_hash
];
948 hlist_add_head(&fi
->fib_hash
, dest
);
951 fib_info_hash
= new_info_hash
;
953 for (i
= 0; i
< old_size
; i
++) {
954 struct hlist_head
*lhead
= &fib_info_laddrhash
[i
];
955 struct hlist_node
*n
;
958 hlist_for_each_entry_safe(fi
, n
, lhead
, fib_lhash
) {
959 struct hlist_head
*ldest
;
960 unsigned int new_hash
;
962 new_hash
= fib_laddr_hashfn(fi
->fib_prefsrc
);
963 ldest
= &new_laddrhash
[new_hash
];
964 hlist_add_head(&fi
->fib_lhash
, ldest
);
967 fib_info_laddrhash
= new_laddrhash
;
969 spin_unlock_bh(&fib_info_lock
);
971 bytes
= old_size
* sizeof(struct hlist_head
*);
972 fib_info_hash_free(old_info_hash
, bytes
);
973 fib_info_hash_free(old_laddrhash
, bytes
);
976 __be32
fib_info_update_nh_saddr(struct net
*net
, struct fib_nh
*nh
)
978 nh
->nh_saddr
= inet_select_addr(nh
->nh_dev
,
980 nh
->nh_parent
->fib_scope
);
981 nh
->nh_saddr_genid
= atomic_read(&net
->ipv4
.dev_addr_genid
);
986 static bool fib_valid_prefsrc(struct fib_config
*cfg
, __be32 fib_prefsrc
)
988 if (cfg
->fc_type
!= RTN_LOCAL
|| !cfg
->fc_dst
||
989 fib_prefsrc
!= cfg
->fc_dst
) {
990 u32 tb_id
= cfg
->fc_table
;
993 if (tb_id
== RT_TABLE_MAIN
)
994 tb_id
= RT_TABLE_LOCAL
;
996 rc
= inet_addr_type_table(cfg
->fc_nlinfo
.nl_net
,
999 if (rc
!= RTN_LOCAL
&& tb_id
!= RT_TABLE_LOCAL
) {
1000 rc
= inet_addr_type_table(cfg
->fc_nlinfo
.nl_net
,
1001 fib_prefsrc
, RT_TABLE_LOCAL
);
1004 if (rc
!= RTN_LOCAL
)
1011 fib_convert_metrics(struct fib_info
*fi
, const struct fib_config
*cfg
)
1013 bool ecn_ca
= false;
1020 nla_for_each_attr(nla
, cfg
->fc_mx
, cfg
->fc_mx_len
, remaining
) {
1021 int type
= nla_type(nla
);
1026 if (type
> RTAX_MAX
)
1029 if (type
== RTAX_CC_ALGO
) {
1030 char tmp
[TCP_CA_NAME_MAX
];
1032 nla_strlcpy(tmp
, nla
, sizeof(tmp
));
1033 val
= tcp_ca_get_key_by_name(fi
->fib_net
, tmp
, &ecn_ca
);
1034 if (val
== TCP_CA_UNSPEC
)
1037 val
= nla_get_u32(nla
);
1039 if (type
== RTAX_ADVMSS
&& val
> 65535 - 40)
1041 if (type
== RTAX_MTU
&& val
> 65535 - 15)
1043 if (type
== RTAX_HOPLIMIT
&& val
> 255)
1045 if (type
== RTAX_FEATURES
&& (val
& ~RTAX_FEATURE_MASK
))
1047 fi
->fib_metrics
->metrics
[type
- 1] = val
;
1051 fi
->fib_metrics
->metrics
[RTAX_FEATURES
- 1] |= DST_FEATURE_ECN_CA
;
1056 struct fib_info
*fib_create_info(struct fib_config
*cfg
,
1057 struct netlink_ext_ack
*extack
)
1060 struct fib_info
*fi
= NULL
;
1061 struct fib_info
*ofi
;
1063 struct net
*net
= cfg
->fc_nlinfo
.nl_net
;
1065 if (cfg
->fc_type
> RTN_MAX
)
1068 /* Fast check to catch the most weird cases */
1069 if (fib_props
[cfg
->fc_type
].scope
> cfg
->fc_scope
) {
1070 NL_SET_ERR_MSG(extack
, "Invalid scope");
1074 if (cfg
->fc_flags
& (RTNH_F_DEAD
| RTNH_F_LINKDOWN
)) {
1075 NL_SET_ERR_MSG(extack
,
1076 "Invalid rtm_flags - can not contain DEAD or LINKDOWN");
1080 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1082 nhs
= fib_count_nexthops(cfg
->fc_mp
, cfg
->fc_mp_len
, extack
);
1089 if (fib_info_cnt
>= fib_info_hash_size
) {
1090 unsigned int new_size
= fib_info_hash_size
<< 1;
1091 struct hlist_head
*new_info_hash
;
1092 struct hlist_head
*new_laddrhash
;
1097 bytes
= new_size
* sizeof(struct hlist_head
*);
1098 new_info_hash
= fib_info_hash_alloc(bytes
);
1099 new_laddrhash
= fib_info_hash_alloc(bytes
);
1100 if (!new_info_hash
|| !new_laddrhash
) {
1101 fib_info_hash_free(new_info_hash
, bytes
);
1102 fib_info_hash_free(new_laddrhash
, bytes
);
1104 fib_info_hash_move(new_info_hash
, new_laddrhash
, new_size
);
1106 if (!fib_info_hash_size
)
1110 fi
= kzalloc(sizeof(*fi
)+nhs
*sizeof(struct fib_nh
), GFP_KERNEL
);
1114 fi
->fib_metrics
= kzalloc(sizeof(*fi
->fib_metrics
), GFP_KERNEL
);
1115 if (unlikely(!fi
->fib_metrics
)) {
1117 return ERR_PTR(err
);
1119 refcount_set(&fi
->fib_metrics
->refcnt
, 1);
1121 fi
->fib_metrics
= (struct dst_metrics
*)&dst_default_metrics
;
1125 fi
->fib_protocol
= cfg
->fc_protocol
;
1126 fi
->fib_scope
= cfg
->fc_scope
;
1127 fi
->fib_flags
= cfg
->fc_flags
;
1128 fi
->fib_priority
= cfg
->fc_priority
;
1129 fi
->fib_prefsrc
= cfg
->fc_prefsrc
;
1130 fi
->fib_type
= cfg
->fc_type
;
1131 fi
->fib_tb_id
= cfg
->fc_table
;
1134 change_nexthops(fi
) {
1135 nexthop_nh
->nh_parent
= fi
;
1136 nexthop_nh
->nh_pcpu_rth_output
= alloc_percpu(struct rtable __rcu
*);
1137 if (!nexthop_nh
->nh_pcpu_rth_output
)
1139 } endfor_nexthops(fi
)
1141 err
= fib_convert_metrics(fi
, cfg
);
1146 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1147 err
= fib_get_nhs(fi
, cfg
->fc_mp
, cfg
->fc_mp_len
, cfg
, extack
);
1150 if (cfg
->fc_oif
&& fi
->fib_nh
->nh_oif
!= cfg
->fc_oif
) {
1151 NL_SET_ERR_MSG(extack
,
1152 "Nexthop device index does not match RTA_OIF");
1155 if (cfg
->fc_gw
&& fi
->fib_nh
->nh_gw
!= cfg
->fc_gw
) {
1156 NL_SET_ERR_MSG(extack
,
1157 "Nexthop gateway does not match RTA_GATEWAY");
1160 #ifdef CONFIG_IP_ROUTE_CLASSID
1161 if (cfg
->fc_flow
&& fi
->fib_nh
->nh_tclassid
!= cfg
->fc_flow
) {
1162 NL_SET_ERR_MSG(extack
,
1163 "Nexthop class id does not match RTA_FLOW");
1168 NL_SET_ERR_MSG(extack
,
1169 "Multipath support not enabled in kernel");
1173 struct fib_nh
*nh
= fi
->fib_nh
;
1175 if (cfg
->fc_encap
) {
1176 struct lwtunnel_state
*lwtstate
;
1178 if (cfg
->fc_encap_type
== LWTUNNEL_ENCAP_NONE
) {
1179 NL_SET_ERR_MSG(extack
,
1180 "LWT encap type not specified");
1183 err
= lwtunnel_build_state(cfg
->fc_encap_type
,
1184 cfg
->fc_encap
, AF_INET
, cfg
,
1189 nh
->nh_lwtstate
= lwtstate_get(lwtstate
);
1191 nh
->nh_oif
= cfg
->fc_oif
;
1192 nh
->nh_gw
= cfg
->fc_gw
;
1193 nh
->nh_flags
= cfg
->fc_flags
;
1194 #ifdef CONFIG_IP_ROUTE_CLASSID
1195 nh
->nh_tclassid
= cfg
->fc_flow
;
1196 if (nh
->nh_tclassid
)
1197 fi
->fib_net
->ipv4
.fib_num_tclassid_users
++;
1199 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1204 if (fib_props
[cfg
->fc_type
].error
) {
1205 if (cfg
->fc_gw
|| cfg
->fc_oif
|| cfg
->fc_mp
) {
1206 NL_SET_ERR_MSG(extack
,
1207 "Gateway, device and multipath can not be specified for this route type");
1212 switch (cfg
->fc_type
) {
1220 NL_SET_ERR_MSG(extack
, "Invalid route type");
1225 if (cfg
->fc_scope
> RT_SCOPE_HOST
) {
1226 NL_SET_ERR_MSG(extack
, "Invalid scope");
1230 if (cfg
->fc_scope
== RT_SCOPE_HOST
) {
1231 struct fib_nh
*nh
= fi
->fib_nh
;
1233 /* Local address is added. */
1235 NL_SET_ERR_MSG(extack
,
1236 "Route with host scope can not have multiple nexthops");
1240 NL_SET_ERR_MSG(extack
,
1241 "Route with host scope can not have a gateway");
1244 nh
->nh_scope
= RT_SCOPE_NOWHERE
;
1245 nh
->nh_dev
= dev_get_by_index(net
, fi
->fib_nh
->nh_oif
);
1252 change_nexthops(fi
) {
1253 err
= fib_check_nh(cfg
, nexthop_nh
, extack
);
1256 if (nexthop_nh
->nh_flags
& RTNH_F_LINKDOWN
)
1258 } endfor_nexthops(fi
)
1259 if (linkdown
== fi
->fib_nhs
)
1260 fi
->fib_flags
|= RTNH_F_LINKDOWN
;
1263 if (fi
->fib_prefsrc
&& !fib_valid_prefsrc(cfg
, fi
->fib_prefsrc
)) {
1264 NL_SET_ERR_MSG(extack
, "Invalid prefsrc address");
1268 change_nexthops(fi
) {
1269 fib_info_update_nh_saddr(net
, nexthop_nh
);
1270 } endfor_nexthops(fi
)
1275 ofi
= fib_find_info(fi
);
1284 refcount_set(&fi
->fib_clntref
, 1);
1285 spin_lock_bh(&fib_info_lock
);
1286 hlist_add_head(&fi
->fib_hash
,
1287 &fib_info_hash
[fib_info_hashfn(fi
)]);
1288 if (fi
->fib_prefsrc
) {
1289 struct hlist_head
*head
;
1291 head
= &fib_info_laddrhash
[fib_laddr_hashfn(fi
->fib_prefsrc
)];
1292 hlist_add_head(&fi
->fib_lhash
, head
);
1294 change_nexthops(fi
) {
1295 struct hlist_head
*head
;
1298 if (!nexthop_nh
->nh_dev
)
1300 hash
= fib_devindex_hashfn(nexthop_nh
->nh_dev
->ifindex
);
1301 head
= &fib_info_devhash
[hash
];
1302 hlist_add_head(&nexthop_nh
->nh_hash
, head
);
1303 } endfor_nexthops(fi
)
1304 spin_unlock_bh(&fib_info_lock
);
1316 return ERR_PTR(err
);
1319 int fib_dump_info(struct sk_buff
*skb
, u32 portid
, u32 seq
, int event
,
1320 u32 tb_id
, u8 type
, __be32 dst
, int dst_len
, u8 tos
,
1321 struct fib_info
*fi
, unsigned int flags
)
1323 struct nlmsghdr
*nlh
;
1326 nlh
= nlmsg_put(skb
, portid
, seq
, event
, sizeof(*rtm
), flags
);
1330 rtm
= nlmsg_data(nlh
);
1331 rtm
->rtm_family
= AF_INET
;
1332 rtm
->rtm_dst_len
= dst_len
;
1333 rtm
->rtm_src_len
= 0;
1336 rtm
->rtm_table
= tb_id
;
1338 rtm
->rtm_table
= RT_TABLE_COMPAT
;
1339 if (nla_put_u32(skb
, RTA_TABLE
, tb_id
))
1340 goto nla_put_failure
;
1341 rtm
->rtm_type
= type
;
1342 rtm
->rtm_flags
= fi
->fib_flags
;
1343 rtm
->rtm_scope
= fi
->fib_scope
;
1344 rtm
->rtm_protocol
= fi
->fib_protocol
;
1346 if (rtm
->rtm_dst_len
&&
1347 nla_put_in_addr(skb
, RTA_DST
, dst
))
1348 goto nla_put_failure
;
1349 if (fi
->fib_priority
&&
1350 nla_put_u32(skb
, RTA_PRIORITY
, fi
->fib_priority
))
1351 goto nla_put_failure
;
1352 if (rtnetlink_put_metrics(skb
, fi
->fib_metrics
->metrics
) < 0)
1353 goto nla_put_failure
;
1355 if (fi
->fib_prefsrc
&&
1356 nla_put_in_addr(skb
, RTA_PREFSRC
, fi
->fib_prefsrc
))
1357 goto nla_put_failure
;
1358 if (fi
->fib_nhs
== 1) {
1359 if (fi
->fib_nh
->nh_gw
&&
1360 nla_put_in_addr(skb
, RTA_GATEWAY
, fi
->fib_nh
->nh_gw
))
1361 goto nla_put_failure
;
1362 if (fi
->fib_nh
->nh_oif
&&
1363 nla_put_u32(skb
, RTA_OIF
, fi
->fib_nh
->nh_oif
))
1364 goto nla_put_failure
;
1365 if (fi
->fib_nh
->nh_flags
& RTNH_F_LINKDOWN
) {
1366 struct in_device
*in_dev
;
1369 in_dev
= __in_dev_get_rcu(fi
->fib_nh
->nh_dev
);
1371 IN_DEV_IGNORE_ROUTES_WITH_LINKDOWN(in_dev
))
1372 rtm
->rtm_flags
|= RTNH_F_DEAD
;
1375 if (fi
->fib_nh
->nh_flags
& RTNH_F_OFFLOAD
)
1376 rtm
->rtm_flags
|= RTNH_F_OFFLOAD
;
1377 #ifdef CONFIG_IP_ROUTE_CLASSID
1378 if (fi
->fib_nh
[0].nh_tclassid
&&
1379 nla_put_u32(skb
, RTA_FLOW
, fi
->fib_nh
[0].nh_tclassid
))
1380 goto nla_put_failure
;
1382 if (fi
->fib_nh
->nh_lwtstate
&&
1383 lwtunnel_fill_encap(skb
, fi
->fib_nh
->nh_lwtstate
) < 0)
1384 goto nla_put_failure
;
1386 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1387 if (fi
->fib_nhs
> 1) {
1388 struct rtnexthop
*rtnh
;
1391 mp
= nla_nest_start(skb
, RTA_MULTIPATH
);
1393 goto nla_put_failure
;
1396 rtnh
= nla_reserve_nohdr(skb
, sizeof(*rtnh
));
1398 goto nla_put_failure
;
1400 rtnh
->rtnh_flags
= nh
->nh_flags
& 0xFF;
1401 if (nh
->nh_flags
& RTNH_F_LINKDOWN
) {
1402 struct in_device
*in_dev
;
1405 in_dev
= __in_dev_get_rcu(nh
->nh_dev
);
1407 IN_DEV_IGNORE_ROUTES_WITH_LINKDOWN(in_dev
))
1408 rtnh
->rtnh_flags
|= RTNH_F_DEAD
;
1411 rtnh
->rtnh_hops
= nh
->nh_weight
- 1;
1412 rtnh
->rtnh_ifindex
= nh
->nh_oif
;
1415 nla_put_in_addr(skb
, RTA_GATEWAY
, nh
->nh_gw
))
1416 goto nla_put_failure
;
1417 #ifdef CONFIG_IP_ROUTE_CLASSID
1418 if (nh
->nh_tclassid
&&
1419 nla_put_u32(skb
, RTA_FLOW
, nh
->nh_tclassid
))
1420 goto nla_put_failure
;
1422 if (nh
->nh_lwtstate
&&
1423 lwtunnel_fill_encap(skb
, nh
->nh_lwtstate
) < 0)
1424 goto nla_put_failure
;
1426 /* length of rtnetlink header + attributes */
1427 rtnh
->rtnh_len
= nlmsg_get_pos(skb
) - (void *) rtnh
;
1428 } endfor_nexthops(fi
);
1430 nla_nest_end(skb
, mp
);
1433 nlmsg_end(skb
, nlh
);
1437 nlmsg_cancel(skb
, nlh
);
1443 * - local address disappeared -> we must delete all the entries
1445 * - device went down -> we must shutdown all nexthops going via it.
1447 int fib_sync_down_addr(struct net_device
*dev
, __be32 local
)
1450 unsigned int hash
= fib_laddr_hashfn(local
);
1451 struct hlist_head
*head
= &fib_info_laddrhash
[hash
];
1452 struct net
*net
= dev_net(dev
);
1453 int tb_id
= l3mdev_fib_table(dev
);
1454 struct fib_info
*fi
;
1456 if (!fib_info_laddrhash
|| local
== 0)
1459 hlist_for_each_entry(fi
, head
, fib_lhash
) {
1460 if (!net_eq(fi
->fib_net
, net
) ||
1461 fi
->fib_tb_id
!= tb_id
)
1463 if (fi
->fib_prefsrc
== local
) {
1464 fi
->fib_flags
|= RTNH_F_DEAD
;
1471 static int call_fib_nh_notifiers(struct fib_nh
*fib_nh
,
1472 enum fib_event_type event_type
)
1474 struct in_device
*in_dev
= __in_dev_get_rtnl(fib_nh
->nh_dev
);
1475 struct fib_nh_notifier_info info
= {
1479 switch (event_type
) {
1480 case FIB_EVENT_NH_ADD
:
1481 if (fib_nh
->nh_flags
& RTNH_F_DEAD
)
1483 if (IN_DEV_IGNORE_ROUTES_WITH_LINKDOWN(in_dev
) &&
1484 fib_nh
->nh_flags
& RTNH_F_LINKDOWN
)
1486 return call_fib4_notifiers(dev_net(fib_nh
->nh_dev
), event_type
,
1488 case FIB_EVENT_NH_DEL
:
1489 if ((in_dev
&& IN_DEV_IGNORE_ROUTES_WITH_LINKDOWN(in_dev
) &&
1490 fib_nh
->nh_flags
& RTNH_F_LINKDOWN
) ||
1491 (fib_nh
->nh_flags
& RTNH_F_DEAD
))
1492 return call_fib4_notifiers(dev_net(fib_nh
->nh_dev
),
1493 event_type
, &info
.info
);
1501 /* Event force Flags Description
1502 * NETDEV_CHANGE 0 LINKDOWN Carrier OFF, not for scope host
1503 * NETDEV_DOWN 0 LINKDOWN|DEAD Link down, not for scope host
1504 * NETDEV_DOWN 1 LINKDOWN|DEAD Last address removed
1505 * NETDEV_UNREGISTER 1 LINKDOWN|DEAD Device removed
1507 int fib_sync_down_dev(struct net_device
*dev
, unsigned long event
, bool force
)
1510 int scope
= RT_SCOPE_NOWHERE
;
1511 struct fib_info
*prev_fi
= NULL
;
1512 unsigned int hash
= fib_devindex_hashfn(dev
->ifindex
);
1513 struct hlist_head
*head
= &fib_info_devhash
[hash
];
1519 hlist_for_each_entry(nh
, head
, nh_hash
) {
1520 struct fib_info
*fi
= nh
->nh_parent
;
1523 BUG_ON(!fi
->fib_nhs
);
1524 if (nh
->nh_dev
!= dev
|| fi
== prev_fi
)
1528 change_nexthops(fi
) {
1529 if (nexthop_nh
->nh_flags
& RTNH_F_DEAD
)
1531 else if (nexthop_nh
->nh_dev
== dev
&&
1532 nexthop_nh
->nh_scope
!= scope
) {
1535 case NETDEV_UNREGISTER
:
1536 nexthop_nh
->nh_flags
|= RTNH_F_DEAD
;
1539 nexthop_nh
->nh_flags
|= RTNH_F_LINKDOWN
;
1542 call_fib_nh_notifiers(nexthop_nh
,
1546 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1547 if (event
== NETDEV_UNREGISTER
&&
1548 nexthop_nh
->nh_dev
== dev
) {
1553 } endfor_nexthops(fi
)
1554 if (dead
== fi
->fib_nhs
) {
1557 case NETDEV_UNREGISTER
:
1558 fi
->fib_flags
|= RTNH_F_DEAD
;
1561 fi
->fib_flags
|= RTNH_F_LINKDOWN
;
1573 /* Must be invoked inside of an RCU protected region. */
1574 static void fib_select_default(const struct flowi4
*flp
, struct fib_result
*res
)
1576 struct fib_info
*fi
= NULL
, *last_resort
= NULL
;
1577 struct hlist_head
*fa_head
= res
->fa_head
;
1578 struct fib_table
*tb
= res
->table
;
1579 u8 slen
= 32 - res
->prefixlen
;
1580 int order
= -1, last_idx
= -1;
1581 struct fib_alias
*fa
, *fa1
= NULL
;
1582 u32 last_prio
= res
->fi
->fib_priority
;
1585 hlist_for_each_entry_rcu(fa
, fa_head
, fa_list
) {
1586 struct fib_info
*next_fi
= fa
->fa_info
;
1588 if (fa
->fa_slen
!= slen
)
1590 if (fa
->fa_tos
&& fa
->fa_tos
!= flp
->flowi4_tos
)
1592 if (fa
->tb_id
!= tb
->tb_id
)
1594 if (next_fi
->fib_priority
> last_prio
&&
1595 fa
->fa_tos
== last_tos
) {
1600 if (next_fi
->fib_flags
& RTNH_F_DEAD
)
1602 last_tos
= fa
->fa_tos
;
1603 last_prio
= next_fi
->fib_priority
;
1605 if (next_fi
->fib_scope
!= res
->scope
||
1606 fa
->fa_type
!= RTN_UNICAST
)
1608 if (!next_fi
->fib_nh
[0].nh_gw
||
1609 next_fi
->fib_nh
[0].nh_scope
!= RT_SCOPE_LINK
)
1612 fib_alias_accessed(fa
);
1615 if (next_fi
!= res
->fi
)
1618 } else if (!fib_detect_death(fi
, order
, &last_resort
,
1619 &last_idx
, fa1
->fa_default
)) {
1620 fib_result_assign(res
, fi
);
1621 fa1
->fa_default
= order
;
1628 if (order
<= 0 || !fi
) {
1630 fa1
->fa_default
= -1;
1634 if (!fib_detect_death(fi
, order
, &last_resort
, &last_idx
,
1636 fib_result_assign(res
, fi
);
1637 fa1
->fa_default
= order
;
1642 fib_result_assign(res
, last_resort
);
1643 fa1
->fa_default
= last_idx
;
1649 * Dead device goes up. We wake up dead nexthops.
1650 * It takes sense only on multipath routes.
1652 int fib_sync_up(struct net_device
*dev
, unsigned int nh_flags
)
1654 struct fib_info
*prev_fi
;
1656 struct hlist_head
*head
;
1660 if (!(dev
->flags
& IFF_UP
))
1663 if (nh_flags
& RTNH_F_DEAD
) {
1664 unsigned int flags
= dev_get_flags(dev
);
1666 if (flags
& (IFF_RUNNING
| IFF_LOWER_UP
))
1667 nh_flags
|= RTNH_F_LINKDOWN
;
1671 hash
= fib_devindex_hashfn(dev
->ifindex
);
1672 head
= &fib_info_devhash
[hash
];
1675 hlist_for_each_entry(nh
, head
, nh_hash
) {
1676 struct fib_info
*fi
= nh
->nh_parent
;
1679 BUG_ON(!fi
->fib_nhs
);
1680 if (nh
->nh_dev
!= dev
|| fi
== prev_fi
)
1685 change_nexthops(fi
) {
1686 if (!(nexthop_nh
->nh_flags
& nh_flags
)) {
1690 if (!nexthop_nh
->nh_dev
||
1691 !(nexthop_nh
->nh_dev
->flags
& IFF_UP
))
1693 if (nexthop_nh
->nh_dev
!= dev
||
1694 !__in_dev_get_rtnl(dev
))
1697 nexthop_nh
->nh_flags
&= ~nh_flags
;
1698 call_fib_nh_notifiers(nexthop_nh
, FIB_EVENT_NH_ADD
);
1699 } endfor_nexthops(fi
)
1702 fi
->fib_flags
&= ~nh_flags
;
1712 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1713 static bool fib_good_nh(const struct fib_nh
*nh
)
1715 int state
= NUD_REACHABLE
;
1717 if (nh
->nh_scope
== RT_SCOPE_LINK
) {
1718 struct neighbour
*n
;
1722 n
= __ipv4_neigh_lookup_noref(nh
->nh_dev
,
1723 (__force u32
)nh
->nh_gw
);
1725 state
= n
->nud_state
;
1727 rcu_read_unlock_bh();
1730 return !!(state
& NUD_VALID
);
1733 void fib_select_multipath(struct fib_result
*res
, int hash
)
1735 struct fib_info
*fi
= res
->fi
;
1736 struct net
*net
= fi
->fib_net
;
1740 if (hash
> atomic_read(&nh
->nh_upper_bound
))
1743 if (!net
->ipv4
.sysctl_fib_multipath_use_neigh
||
1745 res
->nh_sel
= nhsel
;
1749 res
->nh_sel
= nhsel
;
1752 } endfor_nexthops(fi
);
1756 void fib_select_path(struct net
*net
, struct fib_result
*res
,
1757 struct flowi4
*fl4
, const struct sk_buff
*skb
)
1761 oif_check
= (fl4
->flowi4_oif
== 0 ||
1762 fl4
->flowi4_flags
& FLOWI_FLAG_SKIP_NH_OIF
);
1764 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1765 if (res
->fi
->fib_nhs
> 1 && oif_check
) {
1766 int h
= fib_multipath_hash(res
->fi
, fl4
, skb
);
1768 fib_select_multipath(res
, h
);
1772 if (!res
->prefixlen
&&
1773 res
->table
->tb_num_default
> 1 &&
1774 res
->type
== RTN_UNICAST
&& oif_check
)
1775 fib_select_default(fl4
, res
);
1778 fl4
->saddr
= FIB_RES_PREFSRC(net
, *res
);
1780 EXPORT_SYMBOL_GPL(fib_select_path
);