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 * ROUTE - implementation of the IP router.
9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 * Alan Cox, <gw4pts@gw4pts.ampr.org>
11 * Linus Torvalds, <Linus.Torvalds@helsinki.fi>
12 * Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
15 * Alan Cox : Verify area fixes.
16 * Alan Cox : cli() protects routing changes
17 * Rui Oliveira : ICMP routing table updates
18 * (rco@di.uminho.pt) Routing table insertion and update
19 * Linus Torvalds : Rewrote bits to be sensible
20 * Alan Cox : Added BSD route gw semantics
21 * Alan Cox : Super /proc >4K
22 * Alan Cox : MTU in route table
23 * Alan Cox : MSS actually. Also added the window
25 * Sam Lantinga : Fixed route matching in rt_del()
26 * Alan Cox : Routing cache support.
27 * Alan Cox : Removed compatibility cruft.
28 * Alan Cox : RTF_REJECT support.
29 * Alan Cox : TCP irtt support.
30 * Jonathan Naylor : Added Metric support.
31 * Miquel van Smoorenburg : BSD API fixes.
32 * Miquel van Smoorenburg : Metrics.
33 * Alan Cox : Use __u32 properly
34 * Alan Cox : Aligned routing errors more closely with BSD
35 * our system is still very different.
36 * Alan Cox : Faster /proc handling
37 * Alexey Kuznetsov : Massive rework to support tree based routing,
38 * routing caches and better behaviour.
40 * Olaf Erb : irtt wasn't being copied right.
41 * Bjorn Ekwall : Kerneld route support.
42 * Alan Cox : Multicast fixed (I hope)
43 * Pavel Krauz : Limited broadcast fixed
44 * Mike McLagan : Routing by source
45 * Alexey Kuznetsov : End of old history. Split to fib.c and
46 * route.c and rewritten from scratch.
47 * Andi Kleen : Load-limit warning messages.
48 * Vitaly E. Lavrov : Transparent proxy revived after year coma.
49 * Vitaly E. Lavrov : Race condition in ip_route_input_slow.
50 * Tobias Ringstrom : Uninitialized res.type in ip_route_output_slow.
51 * Vladimir V. Ivanov : IP rule info (flowid) is really useful.
52 * Marc Boucher : routing by fwmark
53 * Robert Olsson : Added rt_cache statistics
54 * Arnaldo C. Melo : Convert proc stuff to seq_file
55 * Eric Dumazet : hashed spinlocks and rt_check_expire() fixes.
56 * Ilia Sotnikov : Ignore TOS on PMTUD and Redirect
57 * Ilia Sotnikov : Removed TOS from hash calculations
59 * This program is free software; you can redistribute it and/or
60 * modify it under the terms of the GNU General Public License
61 * as published by the Free Software Foundation; either version
62 * 2 of the License, or (at your option) any later version.
65 #define pr_fmt(fmt) "IPv4: " fmt
67 #include <linux/module.h>
68 #include <asm/uaccess.h>
69 #include <linux/bitops.h>
70 #include <linux/types.h>
71 #include <linux/kernel.h>
73 #include <linux/string.h>
74 #include <linux/socket.h>
75 #include <linux/sockios.h>
76 #include <linux/errno.h>
78 #include <linux/inet.h>
79 #include <linux/netdevice.h>
80 #include <linux/proc_fs.h>
81 #include <linux/init.h>
82 #include <linux/skbuff.h>
83 #include <linux/inetdevice.h>
84 #include <linux/igmp.h>
85 #include <linux/pkt_sched.h>
86 #include <linux/mroute.h>
87 #include <linux/netfilter_ipv4.h>
88 #include <linux/random.h>
89 #include <linux/rcupdate.h>
90 #include <linux/times.h>
91 #include <linux/slab.h>
93 #include <net/net_namespace.h>
94 #include <net/protocol.h>
96 #include <net/route.h>
97 #include <net/inetpeer.h>
99 #include <net/ip_fib.h>
102 #include <net/icmp.h>
103 #include <net/xfrm.h>
104 #include <net/netevent.h>
105 #include <net/rtnetlink.h>
107 #include <linux/sysctl.h>
108 #include <linux/kmemleak.h>
110 #include <net/secure_seq.h>
112 #define RT_FL_TOS(oldflp4) \
113 ((oldflp4)->flowi4_tos & (IPTOS_RT_MASK | RTO_ONLINK))
115 #define RT_GC_TIMEOUT (300*HZ)
117 static int ip_rt_max_size
;
118 static int ip_rt_redirect_number __read_mostly
= 9;
119 static int ip_rt_redirect_load __read_mostly
= HZ
/ 50;
120 static int ip_rt_redirect_silence __read_mostly
= ((HZ
/ 50) << (9 + 1));
121 static int ip_rt_error_cost __read_mostly
= HZ
;
122 static int ip_rt_error_burst __read_mostly
= 5 * HZ
;
123 static int ip_rt_mtu_expires __read_mostly
= 10 * 60 * HZ
;
124 static int ip_rt_min_pmtu __read_mostly
= 512 + 20 + 20;
125 static int ip_rt_min_advmss __read_mostly
= 256;
128 * Interface to generic destination cache.
131 static struct dst_entry
*ipv4_dst_check(struct dst_entry
*dst
, u32 cookie
);
132 static unsigned int ipv4_default_advmss(const struct dst_entry
*dst
);
133 static unsigned int ipv4_mtu(const struct dst_entry
*dst
);
134 static struct dst_entry
*ipv4_negative_advice(struct dst_entry
*dst
);
135 static void ipv4_link_failure(struct sk_buff
*skb
);
136 static void ip_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
137 struct sk_buff
*skb
, u32 mtu
);
138 static void ip_do_redirect(struct dst_entry
*dst
, struct sock
*sk
,
139 struct sk_buff
*skb
);
140 static void ipv4_dst_destroy(struct dst_entry
*dst
);
142 static void ipv4_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
,
147 static u32
*ipv4_cow_metrics(struct dst_entry
*dst
, unsigned long old
)
153 static struct neighbour
*ipv4_neigh_lookup(const struct dst_entry
*dst
,
157 static struct dst_ops ipv4_dst_ops
= {
159 .protocol
= cpu_to_be16(ETH_P_IP
),
160 .check
= ipv4_dst_check
,
161 .default_advmss
= ipv4_default_advmss
,
163 .cow_metrics
= ipv4_cow_metrics
,
164 .destroy
= ipv4_dst_destroy
,
165 .ifdown
= ipv4_dst_ifdown
,
166 .negative_advice
= ipv4_negative_advice
,
167 .link_failure
= ipv4_link_failure
,
168 .update_pmtu
= ip_rt_update_pmtu
,
169 .redirect
= ip_do_redirect
,
170 .local_out
= __ip_local_out
,
171 .neigh_lookup
= ipv4_neigh_lookup
,
174 #define ECN_OR_COST(class) TC_PRIO_##class
176 const __u8 ip_tos2prio
[16] = {
178 ECN_OR_COST(BESTEFFORT
),
180 ECN_OR_COST(BESTEFFORT
),
186 ECN_OR_COST(INTERACTIVE
),
188 ECN_OR_COST(INTERACTIVE
),
189 TC_PRIO_INTERACTIVE_BULK
,
190 ECN_OR_COST(INTERACTIVE_BULK
),
191 TC_PRIO_INTERACTIVE_BULK
,
192 ECN_OR_COST(INTERACTIVE_BULK
)
194 EXPORT_SYMBOL(ip_tos2prio
);
196 static DEFINE_PER_CPU(struct rt_cache_stat
, rt_cache_stat
);
197 #define RT_CACHE_STAT_INC(field) __this_cpu_inc(rt_cache_stat.field)
199 #ifdef CONFIG_PROC_FS
200 static void *rt_cache_seq_start(struct seq_file
*seq
, loff_t
*pos
)
204 return SEQ_START_TOKEN
;
207 static void *rt_cache_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
213 static void rt_cache_seq_stop(struct seq_file
*seq
, void *v
)
217 static int rt_cache_seq_show(struct seq_file
*seq
, void *v
)
219 if (v
== SEQ_START_TOKEN
)
220 seq_printf(seq
, "%-127s\n",
221 "Iface\tDestination\tGateway \tFlags\t\tRefCnt\tUse\t"
222 "Metric\tSource\t\tMTU\tWindow\tIRTT\tTOS\tHHRef\t"
227 static const struct seq_operations rt_cache_seq_ops
= {
228 .start
= rt_cache_seq_start
,
229 .next
= rt_cache_seq_next
,
230 .stop
= rt_cache_seq_stop
,
231 .show
= rt_cache_seq_show
,
234 static int rt_cache_seq_open(struct inode
*inode
, struct file
*file
)
236 return seq_open(file
, &rt_cache_seq_ops
);
239 static const struct file_operations rt_cache_seq_fops
= {
240 .owner
= THIS_MODULE
,
241 .open
= rt_cache_seq_open
,
244 .release
= seq_release
,
248 static void *rt_cpu_seq_start(struct seq_file
*seq
, loff_t
*pos
)
253 return SEQ_START_TOKEN
;
255 for (cpu
= *pos
-1; cpu
< nr_cpu_ids
; ++cpu
) {
256 if (!cpu_possible(cpu
))
259 return &per_cpu(rt_cache_stat
, cpu
);
264 static void *rt_cpu_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
268 for (cpu
= *pos
; cpu
< nr_cpu_ids
; ++cpu
) {
269 if (!cpu_possible(cpu
))
272 return &per_cpu(rt_cache_stat
, cpu
);
278 static void rt_cpu_seq_stop(struct seq_file
*seq
, void *v
)
283 static int rt_cpu_seq_show(struct seq_file
*seq
, void *v
)
285 struct rt_cache_stat
*st
= v
;
287 if (v
== SEQ_START_TOKEN
) {
288 seq_printf(seq
, "entries in_hit in_slow_tot in_slow_mc in_no_route in_brd in_martian_dst in_martian_src out_hit out_slow_tot out_slow_mc gc_total gc_ignored gc_goal_miss gc_dst_overflow in_hlist_search out_hlist_search\n");
292 seq_printf(seq
,"%08x %08x %08x %08x %08x %08x %08x %08x "
293 " %08x %08x %08x %08x %08x %08x %08x %08x %08x \n",
294 dst_entries_get_slow(&ipv4_dst_ops
),
307 0, /* st->gc_total */
308 0, /* st->gc_ignored */
309 0, /* st->gc_goal_miss */
310 0, /* st->gc_dst_overflow */
311 0, /* st->in_hlist_search */
312 0 /* st->out_hlist_search */
317 static const struct seq_operations rt_cpu_seq_ops
= {
318 .start
= rt_cpu_seq_start
,
319 .next
= rt_cpu_seq_next
,
320 .stop
= rt_cpu_seq_stop
,
321 .show
= rt_cpu_seq_show
,
325 static int rt_cpu_seq_open(struct inode
*inode
, struct file
*file
)
327 return seq_open(file
, &rt_cpu_seq_ops
);
330 static const struct file_operations rt_cpu_seq_fops
= {
331 .owner
= THIS_MODULE
,
332 .open
= rt_cpu_seq_open
,
335 .release
= seq_release
,
338 #ifdef CONFIG_IP_ROUTE_CLASSID
339 static int rt_acct_proc_show(struct seq_file
*m
, void *v
)
341 struct ip_rt_acct
*dst
, *src
;
344 dst
= kcalloc(256, sizeof(struct ip_rt_acct
), GFP_KERNEL
);
348 for_each_possible_cpu(i
) {
349 src
= (struct ip_rt_acct
*)per_cpu_ptr(ip_rt_acct
, i
);
350 for (j
= 0; j
< 256; j
++) {
351 dst
[j
].o_bytes
+= src
[j
].o_bytes
;
352 dst
[j
].o_packets
+= src
[j
].o_packets
;
353 dst
[j
].i_bytes
+= src
[j
].i_bytes
;
354 dst
[j
].i_packets
+= src
[j
].i_packets
;
358 seq_write(m
, dst
, 256 * sizeof(struct ip_rt_acct
));
363 static int rt_acct_proc_open(struct inode
*inode
, struct file
*file
)
365 return single_open(file
, rt_acct_proc_show
, NULL
);
368 static const struct file_operations rt_acct_proc_fops
= {
369 .owner
= THIS_MODULE
,
370 .open
= rt_acct_proc_open
,
373 .release
= single_release
,
377 static int __net_init
ip_rt_do_proc_init(struct net
*net
)
379 struct proc_dir_entry
*pde
;
381 pde
= proc_create("rt_cache", S_IRUGO
, net
->proc_net
,
386 pde
= proc_create("rt_cache", S_IRUGO
,
387 net
->proc_net_stat
, &rt_cpu_seq_fops
);
391 #ifdef CONFIG_IP_ROUTE_CLASSID
392 pde
= proc_create("rt_acct", 0, net
->proc_net
, &rt_acct_proc_fops
);
398 #ifdef CONFIG_IP_ROUTE_CLASSID
400 remove_proc_entry("rt_cache", net
->proc_net_stat
);
403 remove_proc_entry("rt_cache", net
->proc_net
);
408 static void __net_exit
ip_rt_do_proc_exit(struct net
*net
)
410 remove_proc_entry("rt_cache", net
->proc_net_stat
);
411 remove_proc_entry("rt_cache", net
->proc_net
);
412 #ifdef CONFIG_IP_ROUTE_CLASSID
413 remove_proc_entry("rt_acct", net
->proc_net
);
417 static struct pernet_operations ip_rt_proc_ops __net_initdata
= {
418 .init
= ip_rt_do_proc_init
,
419 .exit
= ip_rt_do_proc_exit
,
422 static int __init
ip_rt_proc_init(void)
424 return register_pernet_subsys(&ip_rt_proc_ops
);
428 static inline int ip_rt_proc_init(void)
432 #endif /* CONFIG_PROC_FS */
434 static inline bool rt_is_expired(const struct rtable
*rth
)
436 return rth
->rt_genid
!= rt_genid_ipv4(dev_net(rth
->dst
.dev
));
439 void rt_cache_flush(struct net
*net
)
441 rt_genid_bump_ipv4(net
);
444 static struct neighbour
*ipv4_neigh_lookup(const struct dst_entry
*dst
,
448 struct net_device
*dev
= dst
->dev
;
449 const __be32
*pkey
= daddr
;
450 const struct rtable
*rt
;
453 rt
= (const struct rtable
*) dst
;
455 pkey
= (const __be32
*) &rt
->rt_gateway
;
457 pkey
= &ip_hdr(skb
)->daddr
;
459 n
= __ipv4_neigh_lookup(dev
, *(__force u32
*)pkey
);
462 return neigh_create(&arp_tbl
, pkey
, dev
);
466 * Peer allocation may fail only in serious out-of-memory conditions. However
467 * we still can generate some output.
468 * Random ID selection looks a bit dangerous because we have no chances to
469 * select ID being unique in a reasonable period of time.
470 * But broken packet identifier may be better than no packet at all.
472 static void ip_select_fb_ident(struct iphdr
*iph
)
474 static DEFINE_SPINLOCK(ip_fb_id_lock
);
475 static u32 ip_fallback_id
;
478 spin_lock_bh(&ip_fb_id_lock
);
479 salt
= secure_ip_id((__force __be32
)ip_fallback_id
^ iph
->daddr
);
480 iph
->id
= htons(salt
& 0xFFFF);
481 ip_fallback_id
= salt
;
482 spin_unlock_bh(&ip_fb_id_lock
);
485 void __ip_select_ident(struct iphdr
*iph
, struct dst_entry
*dst
, int more
)
487 struct net
*net
= dev_net(dst
->dev
);
488 struct inet_peer
*peer
;
490 peer
= inet_getpeer_v4(net
->ipv4
.peers
, iph
->daddr
, 1);
492 iph
->id
= htons(inet_getid(peer
, more
));
497 ip_select_fb_ident(iph
);
499 EXPORT_SYMBOL(__ip_select_ident
);
501 static void __build_flow_key(struct flowi4
*fl4
, const struct sock
*sk
,
502 const struct iphdr
*iph
,
504 u8 prot
, u32 mark
, int flow_flags
)
507 const struct inet_sock
*inet
= inet_sk(sk
);
509 oif
= sk
->sk_bound_dev_if
;
511 tos
= RT_CONN_FLAGS(sk
);
512 prot
= inet
->hdrincl
? IPPROTO_RAW
: sk
->sk_protocol
;
514 flowi4_init_output(fl4
, oif
, mark
, tos
,
515 RT_SCOPE_UNIVERSE
, prot
,
517 iph
->daddr
, iph
->saddr
, 0, 0);
520 static void build_skb_flow_key(struct flowi4
*fl4
, const struct sk_buff
*skb
,
521 const struct sock
*sk
)
523 const struct iphdr
*iph
= ip_hdr(skb
);
524 int oif
= skb
->dev
->ifindex
;
525 u8 tos
= RT_TOS(iph
->tos
);
526 u8 prot
= iph
->protocol
;
527 u32 mark
= skb
->mark
;
529 __build_flow_key(fl4
, sk
, iph
, oif
, tos
, prot
, mark
, 0);
532 static void build_sk_flow_key(struct flowi4
*fl4
, const struct sock
*sk
)
534 const struct inet_sock
*inet
= inet_sk(sk
);
535 const struct ip_options_rcu
*inet_opt
;
536 __be32 daddr
= inet
->inet_daddr
;
539 inet_opt
= rcu_dereference(inet
->inet_opt
);
540 if (inet_opt
&& inet_opt
->opt
.srr
)
541 daddr
= inet_opt
->opt
.faddr
;
542 flowi4_init_output(fl4
, sk
->sk_bound_dev_if
, sk
->sk_mark
,
543 RT_CONN_FLAGS(sk
), RT_SCOPE_UNIVERSE
,
544 inet
->hdrincl
? IPPROTO_RAW
: sk
->sk_protocol
,
545 inet_sk_flowi_flags(sk
),
546 daddr
, inet
->inet_saddr
, 0, 0);
550 static void ip_rt_build_flow_key(struct flowi4
*fl4
, const struct sock
*sk
,
551 const struct sk_buff
*skb
)
554 build_skb_flow_key(fl4
, skb
, sk
);
556 build_sk_flow_key(fl4
, sk
);
559 static inline void rt_free(struct rtable
*rt
)
561 call_rcu(&rt
->dst
.rcu_head
, dst_rcu_free
);
564 static DEFINE_SPINLOCK(fnhe_lock
);
566 static void fnhe_flush_routes(struct fib_nh_exception
*fnhe
)
570 rt
= rcu_dereference(fnhe
->fnhe_rth_input
);
572 RCU_INIT_POINTER(fnhe
->fnhe_rth_input
, NULL
);
575 rt
= rcu_dereference(fnhe
->fnhe_rth_output
);
577 RCU_INIT_POINTER(fnhe
->fnhe_rth_output
, NULL
);
582 static struct fib_nh_exception
*fnhe_oldest(struct fnhe_hash_bucket
*hash
)
584 struct fib_nh_exception
*fnhe
, *oldest
;
586 oldest
= rcu_dereference(hash
->chain
);
587 for (fnhe
= rcu_dereference(oldest
->fnhe_next
); fnhe
;
588 fnhe
= rcu_dereference(fnhe
->fnhe_next
)) {
589 if (time_before(fnhe
->fnhe_stamp
, oldest
->fnhe_stamp
))
592 fnhe_flush_routes(oldest
);
596 static inline u32
fnhe_hashfun(__be32 daddr
)
600 hval
= (__force u32
) daddr
;
601 hval
^= (hval
>> 11) ^ (hval
>> 22);
603 return hval
& (FNHE_HASH_SIZE
- 1);
606 static void fill_route_from_fnhe(struct rtable
*rt
, struct fib_nh_exception
*fnhe
)
608 rt
->rt_pmtu
= fnhe
->fnhe_pmtu
;
609 rt
->dst
.expires
= fnhe
->fnhe_expires
;
612 rt
->rt_flags
|= RTCF_REDIRECTED
;
613 rt
->rt_gateway
= fnhe
->fnhe_gw
;
614 rt
->rt_uses_gateway
= 1;
618 static void update_or_create_fnhe(struct fib_nh
*nh
, __be32 daddr
, __be32 gw
,
619 u32 pmtu
, unsigned long expires
)
621 struct fnhe_hash_bucket
*hash
;
622 struct fib_nh_exception
*fnhe
;
626 u32 hval
= fnhe_hashfun(daddr
);
628 spin_lock_bh(&fnhe_lock
);
630 hash
= nh
->nh_exceptions
;
632 hash
= kzalloc(FNHE_HASH_SIZE
* sizeof(*hash
), GFP_ATOMIC
);
635 nh
->nh_exceptions
= hash
;
641 for (fnhe
= rcu_dereference(hash
->chain
); fnhe
;
642 fnhe
= rcu_dereference(fnhe
->fnhe_next
)) {
643 if (fnhe
->fnhe_daddr
== daddr
)
652 fnhe
->fnhe_pmtu
= pmtu
;
653 fnhe
->fnhe_expires
= max(1UL, expires
);
655 /* Update all cached dsts too */
656 rt
= rcu_dereference(fnhe
->fnhe_rth_input
);
658 fill_route_from_fnhe(rt
, fnhe
);
659 rt
= rcu_dereference(fnhe
->fnhe_rth_output
);
661 fill_route_from_fnhe(rt
, fnhe
);
663 if (depth
> FNHE_RECLAIM_DEPTH
)
664 fnhe
= fnhe_oldest(hash
);
666 fnhe
= kzalloc(sizeof(*fnhe
), GFP_ATOMIC
);
670 fnhe
->fnhe_next
= hash
->chain
;
671 rcu_assign_pointer(hash
->chain
, fnhe
);
673 fnhe
->fnhe_genid
= fnhe_genid(dev_net(nh
->nh_dev
));
674 fnhe
->fnhe_daddr
= daddr
;
676 fnhe
->fnhe_pmtu
= pmtu
;
677 fnhe
->fnhe_expires
= expires
;
679 /* Exception created; mark the cached routes for the nexthop
680 * stale, so anyone caching it rechecks if this exception
683 rt
= rcu_dereference(nh
->nh_rth_input
);
685 rt
->dst
.obsolete
= DST_OBSOLETE_KILL
;
687 for_each_possible_cpu(i
) {
688 struct rtable __rcu
**prt
;
689 prt
= per_cpu_ptr(nh
->nh_pcpu_rth_output
, i
);
690 rt
= rcu_dereference(*prt
);
692 rt
->dst
.obsolete
= DST_OBSOLETE_KILL
;
696 fnhe
->fnhe_stamp
= jiffies
;
699 spin_unlock_bh(&fnhe_lock
);
703 static void __ip_do_redirect(struct rtable
*rt
, struct sk_buff
*skb
, struct flowi4
*fl4
,
706 __be32 new_gw
= icmp_hdr(skb
)->un
.gateway
;
707 __be32 old_gw
= ip_hdr(skb
)->saddr
;
708 struct net_device
*dev
= skb
->dev
;
709 struct in_device
*in_dev
;
710 struct fib_result res
;
714 switch (icmp_hdr(skb
)->code
& 7) {
716 case ICMP_REDIR_NETTOS
:
717 case ICMP_REDIR_HOST
:
718 case ICMP_REDIR_HOSTTOS
:
725 if (rt
->rt_gateway
!= old_gw
)
728 in_dev
= __in_dev_get_rcu(dev
);
733 if (new_gw
== old_gw
|| !IN_DEV_RX_REDIRECTS(in_dev
) ||
734 ipv4_is_multicast(new_gw
) || ipv4_is_lbcast(new_gw
) ||
735 ipv4_is_zeronet(new_gw
))
736 goto reject_redirect
;
738 if (!IN_DEV_SHARED_MEDIA(in_dev
)) {
739 if (!inet_addr_onlink(in_dev
, new_gw
, old_gw
))
740 goto reject_redirect
;
741 if (IN_DEV_SEC_REDIRECTS(in_dev
) && ip_fib_check_default(new_gw
, dev
))
742 goto reject_redirect
;
744 if (inet_addr_type(net
, new_gw
) != RTN_UNICAST
)
745 goto reject_redirect
;
748 n
= ipv4_neigh_lookup(&rt
->dst
, NULL
, &new_gw
);
750 if (!(n
->nud_state
& NUD_VALID
)) {
751 neigh_event_send(n
, NULL
);
753 if (fib_lookup(net
, fl4
, &res
) == 0) {
754 struct fib_nh
*nh
= &FIB_RES_NH(res
);
756 update_or_create_fnhe(nh
, fl4
->daddr
, new_gw
,
760 rt
->dst
.obsolete
= DST_OBSOLETE_KILL
;
761 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE
, n
);
768 #ifdef CONFIG_IP_ROUTE_VERBOSE
769 if (IN_DEV_LOG_MARTIANS(in_dev
)) {
770 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
771 __be32 daddr
= iph
->daddr
;
772 __be32 saddr
= iph
->saddr
;
774 net_info_ratelimited("Redirect from %pI4 on %s about %pI4 ignored\n"
775 " Advised path = %pI4 -> %pI4\n",
776 &old_gw
, dev
->name
, &new_gw
,
783 static void ip_do_redirect(struct dst_entry
*dst
, struct sock
*sk
, struct sk_buff
*skb
)
787 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
788 int oif
= skb
->dev
->ifindex
;
789 u8 tos
= RT_TOS(iph
->tos
);
790 u8 prot
= iph
->protocol
;
791 u32 mark
= skb
->mark
;
793 rt
= (struct rtable
*) dst
;
795 __build_flow_key(&fl4
, sk
, iph
, oif
, tos
, prot
, mark
, 0);
796 __ip_do_redirect(rt
, skb
, &fl4
, true);
799 static struct dst_entry
*ipv4_negative_advice(struct dst_entry
*dst
)
801 struct rtable
*rt
= (struct rtable
*)dst
;
802 struct dst_entry
*ret
= dst
;
805 if (dst
->obsolete
> 0) {
808 } else if ((rt
->rt_flags
& RTCF_REDIRECTED
) ||
819 * 1. The first ip_rt_redirect_number redirects are sent
820 * with exponential backoff, then we stop sending them at all,
821 * assuming that the host ignores our redirects.
822 * 2. If we did not see packets requiring redirects
823 * during ip_rt_redirect_silence, we assume that the host
824 * forgot redirected route and start to send redirects again.
826 * This algorithm is much cheaper and more intelligent than dumb load limiting
829 * NOTE. Do not forget to inhibit load limiting for redirects (redundant)
830 * and "frag. need" (breaks PMTU discovery) in icmp.c.
833 void ip_rt_send_redirect(struct sk_buff
*skb
)
835 struct rtable
*rt
= skb_rtable(skb
);
836 struct in_device
*in_dev
;
837 struct inet_peer
*peer
;
842 in_dev
= __in_dev_get_rcu(rt
->dst
.dev
);
843 if (!in_dev
|| !IN_DEV_TX_REDIRECTS(in_dev
)) {
847 log_martians
= IN_DEV_LOG_MARTIANS(in_dev
);
850 net
= dev_net(rt
->dst
.dev
);
851 peer
= inet_getpeer_v4(net
->ipv4
.peers
, ip_hdr(skb
)->saddr
, 1);
853 icmp_send(skb
, ICMP_REDIRECT
, ICMP_REDIR_HOST
,
854 rt_nexthop(rt
, ip_hdr(skb
)->daddr
));
858 /* No redirected packets during ip_rt_redirect_silence;
859 * reset the algorithm.
861 if (time_after(jiffies
, peer
->rate_last
+ ip_rt_redirect_silence
))
862 peer
->rate_tokens
= 0;
864 /* Too many ignored redirects; do not send anything
865 * set dst.rate_last to the last seen redirected packet.
867 if (peer
->rate_tokens
>= ip_rt_redirect_number
) {
868 peer
->rate_last
= jiffies
;
872 /* Check for load limit; set rate_last to the latest sent
875 if (peer
->rate_tokens
== 0 ||
878 (ip_rt_redirect_load
<< peer
->rate_tokens
)))) {
879 __be32 gw
= rt_nexthop(rt
, ip_hdr(skb
)->daddr
);
881 icmp_send(skb
, ICMP_REDIRECT
, ICMP_REDIR_HOST
, gw
);
882 peer
->rate_last
= jiffies
;
884 #ifdef CONFIG_IP_ROUTE_VERBOSE
886 peer
->rate_tokens
== ip_rt_redirect_number
)
887 net_warn_ratelimited("host %pI4/if%d ignores redirects for %pI4 to %pI4\n",
888 &ip_hdr(skb
)->saddr
, inet_iif(skb
),
889 &ip_hdr(skb
)->daddr
, &gw
);
896 static int ip_error(struct sk_buff
*skb
)
898 struct in_device
*in_dev
= __in_dev_get_rcu(skb
->dev
);
899 struct rtable
*rt
= skb_rtable(skb
);
900 struct inet_peer
*peer
;
906 net
= dev_net(rt
->dst
.dev
);
907 if (!IN_DEV_FORWARD(in_dev
)) {
908 switch (rt
->dst
.error
) {
910 IP_INC_STATS_BH(net
, IPSTATS_MIB_INADDRERRORS
);
914 IP_INC_STATS_BH(net
, IPSTATS_MIB_INNOROUTES
);
920 switch (rt
->dst
.error
) {
925 code
= ICMP_HOST_UNREACH
;
928 code
= ICMP_NET_UNREACH
;
929 IP_INC_STATS_BH(net
, IPSTATS_MIB_INNOROUTES
);
932 code
= ICMP_PKT_FILTERED
;
936 peer
= inet_getpeer_v4(net
->ipv4
.peers
, ip_hdr(skb
)->saddr
, 1);
941 peer
->rate_tokens
+= now
- peer
->rate_last
;
942 if (peer
->rate_tokens
> ip_rt_error_burst
)
943 peer
->rate_tokens
= ip_rt_error_burst
;
944 peer
->rate_last
= now
;
945 if (peer
->rate_tokens
>= ip_rt_error_cost
)
946 peer
->rate_tokens
-= ip_rt_error_cost
;
952 icmp_send(skb
, ICMP_DEST_UNREACH
, code
, 0);
958 static void __ip_rt_update_pmtu(struct rtable
*rt
, struct flowi4
*fl4
, u32 mtu
)
960 struct dst_entry
*dst
= &rt
->dst
;
961 struct fib_result res
;
963 if (dst_metric_locked(dst
, RTAX_MTU
))
966 if (dst
->dev
->mtu
< mtu
)
969 if (mtu
< ip_rt_min_pmtu
)
970 mtu
= ip_rt_min_pmtu
;
972 if (rt
->rt_pmtu
== mtu
&&
973 time_before(jiffies
, dst
->expires
- ip_rt_mtu_expires
/ 2))
977 if (fib_lookup(dev_net(dst
->dev
), fl4
, &res
) == 0) {
978 struct fib_nh
*nh
= &FIB_RES_NH(res
);
980 update_or_create_fnhe(nh
, fl4
->daddr
, 0, mtu
,
981 jiffies
+ ip_rt_mtu_expires
);
986 static void ip_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
987 struct sk_buff
*skb
, u32 mtu
)
989 struct rtable
*rt
= (struct rtable
*) dst
;
992 ip_rt_build_flow_key(&fl4
, sk
, skb
);
993 __ip_rt_update_pmtu(rt
, &fl4
, mtu
);
996 void ipv4_update_pmtu(struct sk_buff
*skb
, struct net
*net
, u32 mtu
,
997 int oif
, u32 mark
, u8 protocol
, int flow_flags
)
999 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1003 __build_flow_key(&fl4
, NULL
, iph
, oif
,
1004 RT_TOS(iph
->tos
), protocol
, mark
, flow_flags
);
1005 rt
= __ip_route_output_key(net
, &fl4
);
1007 __ip_rt_update_pmtu(rt
, &fl4
, mtu
);
1011 EXPORT_SYMBOL_GPL(ipv4_update_pmtu
);
1013 static void __ipv4_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, u32 mtu
)
1015 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1019 __build_flow_key(&fl4
, sk
, iph
, 0, 0, 0, 0, 0);
1020 rt
= __ip_route_output_key(sock_net(sk
), &fl4
);
1022 __ip_rt_update_pmtu(rt
, &fl4
, mtu
);
1027 void ipv4_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, u32 mtu
)
1029 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1032 struct dst_entry
*dst
;
1037 if (!ip_sk_accept_pmtu(sk
))
1040 rt
= (struct rtable
*) __sk_dst_get(sk
);
1042 if (sock_owned_by_user(sk
) || !rt
) {
1043 __ipv4_sk_update_pmtu(skb
, sk
, mtu
);
1047 __build_flow_key(&fl4
, sk
, iph
, 0, 0, 0, 0, 0);
1049 if (!__sk_dst_check(sk
, 0)) {
1050 rt
= ip_route_output_flow(sock_net(sk
), &fl4
, sk
);
1057 __ip_rt_update_pmtu((struct rtable
*) rt
->dst
.path
, &fl4
, mtu
);
1059 dst
= dst_check(&rt
->dst
, 0);
1062 dst_release(&rt
->dst
);
1064 rt
= ip_route_output_flow(sock_net(sk
), &fl4
, sk
);
1072 __sk_dst_set(sk
, &rt
->dst
);
1077 EXPORT_SYMBOL_GPL(ipv4_sk_update_pmtu
);
1079 void ipv4_redirect(struct sk_buff
*skb
, struct net
*net
,
1080 int oif
, u32 mark
, u8 protocol
, int flow_flags
)
1082 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1086 __build_flow_key(&fl4
, NULL
, iph
, oif
,
1087 RT_TOS(iph
->tos
), protocol
, mark
, flow_flags
);
1088 rt
= __ip_route_output_key(net
, &fl4
);
1090 __ip_do_redirect(rt
, skb
, &fl4
, false);
1094 EXPORT_SYMBOL_GPL(ipv4_redirect
);
1096 void ipv4_sk_redirect(struct sk_buff
*skb
, struct sock
*sk
)
1098 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1102 __build_flow_key(&fl4
, sk
, iph
, 0, 0, 0, 0, 0);
1103 rt
= __ip_route_output_key(sock_net(sk
), &fl4
);
1105 __ip_do_redirect(rt
, skb
, &fl4
, false);
1109 EXPORT_SYMBOL_GPL(ipv4_sk_redirect
);
1111 static struct dst_entry
*ipv4_dst_check(struct dst_entry
*dst
, u32 cookie
)
1113 struct rtable
*rt
= (struct rtable
*) dst
;
1115 /* All IPV4 dsts are created with ->obsolete set to the value
1116 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1117 * into this function always.
1119 * When a PMTU/redirect information update invalidates a route,
1120 * this is indicated by setting obsolete to DST_OBSOLETE_KILL or
1121 * DST_OBSOLETE_DEAD by dst_free().
1123 if (dst
->obsolete
!= DST_OBSOLETE_FORCE_CHK
|| rt_is_expired(rt
))
1128 static void ipv4_link_failure(struct sk_buff
*skb
)
1132 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_HOST_UNREACH
, 0);
1134 rt
= skb_rtable(skb
);
1136 dst_set_expires(&rt
->dst
, 0);
1139 static int ip_rt_bug(struct sk_buff
*skb
)
1141 pr_debug("%s: %pI4 -> %pI4, %s\n",
1142 __func__
, &ip_hdr(skb
)->saddr
, &ip_hdr(skb
)->daddr
,
1143 skb
->dev
? skb
->dev
->name
: "?");
1150 We do not cache source address of outgoing interface,
1151 because it is used only by IP RR, TS and SRR options,
1152 so that it out of fast path.
1154 BTW remember: "addr" is allowed to be not aligned
1158 void ip_rt_get_source(u8
*addr
, struct sk_buff
*skb
, struct rtable
*rt
)
1162 if (rt_is_output_route(rt
))
1163 src
= ip_hdr(skb
)->saddr
;
1165 struct fib_result res
;
1171 memset(&fl4
, 0, sizeof(fl4
));
1172 fl4
.daddr
= iph
->daddr
;
1173 fl4
.saddr
= iph
->saddr
;
1174 fl4
.flowi4_tos
= RT_TOS(iph
->tos
);
1175 fl4
.flowi4_oif
= rt
->dst
.dev
->ifindex
;
1176 fl4
.flowi4_iif
= skb
->dev
->ifindex
;
1177 fl4
.flowi4_mark
= skb
->mark
;
1180 if (fib_lookup(dev_net(rt
->dst
.dev
), &fl4
, &res
) == 0)
1181 src
= FIB_RES_PREFSRC(dev_net(rt
->dst
.dev
), res
);
1183 src
= inet_select_addr(rt
->dst
.dev
,
1184 rt_nexthop(rt
, iph
->daddr
),
1188 memcpy(addr
, &src
, 4);
1191 #ifdef CONFIG_IP_ROUTE_CLASSID
1192 static void set_class_tag(struct rtable
*rt
, u32 tag
)
1194 if (!(rt
->dst
.tclassid
& 0xFFFF))
1195 rt
->dst
.tclassid
|= tag
& 0xFFFF;
1196 if (!(rt
->dst
.tclassid
& 0xFFFF0000))
1197 rt
->dst
.tclassid
|= tag
& 0xFFFF0000;
1201 static unsigned int ipv4_default_advmss(const struct dst_entry
*dst
)
1203 unsigned int advmss
= dst_metric_raw(dst
, RTAX_ADVMSS
);
1206 advmss
= max_t(unsigned int, dst
->dev
->mtu
- 40,
1208 if (advmss
> 65535 - 40)
1209 advmss
= 65535 - 40;
1214 static unsigned int ipv4_mtu(const struct dst_entry
*dst
)
1216 const struct rtable
*rt
= (const struct rtable
*) dst
;
1217 unsigned int mtu
= rt
->rt_pmtu
;
1219 if (!mtu
|| time_after_eq(jiffies
, rt
->dst
.expires
))
1220 mtu
= dst_metric_raw(dst
, RTAX_MTU
);
1225 mtu
= dst
->dev
->mtu
;
1227 if (unlikely(dst_metric_locked(dst
, RTAX_MTU
))) {
1228 if (rt
->rt_uses_gateway
&& mtu
> 576)
1232 return min_t(unsigned int, mtu
, IP_MAX_MTU
);
1235 static struct fib_nh_exception
*find_exception(struct fib_nh
*nh
, __be32 daddr
)
1237 struct fnhe_hash_bucket
*hash
= nh
->nh_exceptions
;
1238 struct fib_nh_exception
*fnhe
;
1244 hval
= fnhe_hashfun(daddr
);
1246 for (fnhe
= rcu_dereference(hash
[hval
].chain
); fnhe
;
1247 fnhe
= rcu_dereference(fnhe
->fnhe_next
)) {
1248 if (fnhe
->fnhe_daddr
== daddr
)
1254 static bool rt_bind_exception(struct rtable
*rt
, struct fib_nh_exception
*fnhe
,
1259 spin_lock_bh(&fnhe_lock
);
1261 if (daddr
== fnhe
->fnhe_daddr
) {
1262 struct rtable __rcu
**porig
;
1263 struct rtable
*orig
;
1264 int genid
= fnhe_genid(dev_net(rt
->dst
.dev
));
1266 if (rt_is_input_route(rt
))
1267 porig
= &fnhe
->fnhe_rth_input
;
1269 porig
= &fnhe
->fnhe_rth_output
;
1270 orig
= rcu_dereference(*porig
);
1272 if (fnhe
->fnhe_genid
!= genid
) {
1273 fnhe
->fnhe_genid
= genid
;
1275 fnhe
->fnhe_pmtu
= 0;
1276 fnhe
->fnhe_expires
= 0;
1277 fnhe_flush_routes(fnhe
);
1280 fill_route_from_fnhe(rt
, fnhe
);
1281 if (!rt
->rt_gateway
)
1282 rt
->rt_gateway
= daddr
;
1284 if (!(rt
->dst
.flags
& DST_NOCACHE
)) {
1285 rcu_assign_pointer(*porig
, rt
);
1291 fnhe
->fnhe_stamp
= jiffies
;
1293 spin_unlock_bh(&fnhe_lock
);
1298 static bool rt_cache_route(struct fib_nh
*nh
, struct rtable
*rt
)
1300 struct rtable
*orig
, *prev
, **p
;
1303 if (rt_is_input_route(rt
)) {
1304 p
= (struct rtable
**)&nh
->nh_rth_input
;
1306 p
= (struct rtable
**)__this_cpu_ptr(nh
->nh_pcpu_rth_output
);
1310 prev
= cmpxchg(p
, orig
, rt
);
1320 static DEFINE_SPINLOCK(rt_uncached_lock
);
1321 static LIST_HEAD(rt_uncached_list
);
1323 static void rt_add_uncached_list(struct rtable
*rt
)
1325 spin_lock_bh(&rt_uncached_lock
);
1326 list_add_tail(&rt
->rt_uncached
, &rt_uncached_list
);
1327 spin_unlock_bh(&rt_uncached_lock
);
1330 static void ipv4_dst_destroy(struct dst_entry
*dst
)
1332 struct rtable
*rt
= (struct rtable
*) dst
;
1334 if (!list_empty(&rt
->rt_uncached
)) {
1335 spin_lock_bh(&rt_uncached_lock
);
1336 list_del(&rt
->rt_uncached
);
1337 spin_unlock_bh(&rt_uncached_lock
);
1341 void rt_flush_dev(struct net_device
*dev
)
1343 if (!list_empty(&rt_uncached_list
)) {
1344 struct net
*net
= dev_net(dev
);
1347 spin_lock_bh(&rt_uncached_lock
);
1348 list_for_each_entry(rt
, &rt_uncached_list
, rt_uncached
) {
1349 if (rt
->dst
.dev
!= dev
)
1351 rt
->dst
.dev
= net
->loopback_dev
;
1352 dev_hold(rt
->dst
.dev
);
1355 spin_unlock_bh(&rt_uncached_lock
);
1359 static bool rt_cache_valid(const struct rtable
*rt
)
1362 rt
->dst
.obsolete
== DST_OBSOLETE_FORCE_CHK
&&
1366 static void rt_set_nexthop(struct rtable
*rt
, __be32 daddr
,
1367 const struct fib_result
*res
,
1368 struct fib_nh_exception
*fnhe
,
1369 struct fib_info
*fi
, u16 type
, u32 itag
)
1371 bool cached
= false;
1374 struct fib_nh
*nh
= &FIB_RES_NH(*res
);
1376 if (nh
->nh_gw
&& nh
->nh_scope
== RT_SCOPE_LINK
) {
1377 rt
->rt_gateway
= nh
->nh_gw
;
1378 rt
->rt_uses_gateway
= 1;
1380 dst_init_metrics(&rt
->dst
, fi
->fib_metrics
, true);
1381 #ifdef CONFIG_IP_ROUTE_CLASSID
1382 rt
->dst
.tclassid
= nh
->nh_tclassid
;
1385 cached
= rt_bind_exception(rt
, fnhe
, daddr
);
1386 else if (!(rt
->dst
.flags
& DST_NOCACHE
))
1387 cached
= rt_cache_route(nh
, rt
);
1388 if (unlikely(!cached
)) {
1389 /* Routes we intend to cache in nexthop exception or
1390 * FIB nexthop have the DST_NOCACHE bit clear.
1391 * However, if we are unsuccessful at storing this
1392 * route into the cache we really need to set it.
1394 rt
->dst
.flags
|= DST_NOCACHE
;
1395 if (!rt
->rt_gateway
)
1396 rt
->rt_gateway
= daddr
;
1397 rt_add_uncached_list(rt
);
1400 rt_add_uncached_list(rt
);
1402 #ifdef CONFIG_IP_ROUTE_CLASSID
1403 #ifdef CONFIG_IP_MULTIPLE_TABLES
1404 set_class_tag(rt
, res
->tclassid
);
1406 set_class_tag(rt
, itag
);
1410 static struct rtable
*rt_dst_alloc(struct net_device
*dev
,
1411 bool nopolicy
, bool noxfrm
, bool will_cache
)
1413 return dst_alloc(&ipv4_dst_ops
, dev
, 1, DST_OBSOLETE_FORCE_CHK
,
1414 (will_cache
? 0 : (DST_HOST
| DST_NOCACHE
)) |
1415 (nopolicy
? DST_NOPOLICY
: 0) |
1416 (noxfrm
? DST_NOXFRM
: 0));
1419 /* called in rcu_read_lock() section */
1420 static int ip_route_input_mc(struct sk_buff
*skb
, __be32 daddr
, __be32 saddr
,
1421 u8 tos
, struct net_device
*dev
, int our
)
1424 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
1428 /* Primary sanity checks. */
1433 if (ipv4_is_multicast(saddr
) || ipv4_is_lbcast(saddr
) ||
1434 skb
->protocol
!= htons(ETH_P_IP
))
1437 if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev
)))
1438 if (ipv4_is_loopback(saddr
))
1441 if (ipv4_is_zeronet(saddr
)) {
1442 if (!ipv4_is_local_multicast(daddr
))
1445 err
= fib_validate_source(skb
, saddr
, 0, tos
, 0, dev
,
1450 rth
= rt_dst_alloc(dev_net(dev
)->loopback_dev
,
1451 IN_DEV_CONF_GET(in_dev
, NOPOLICY
), false, false);
1455 #ifdef CONFIG_IP_ROUTE_CLASSID
1456 rth
->dst
.tclassid
= itag
;
1458 rth
->dst
.output
= ip_rt_bug
;
1460 rth
->rt_genid
= rt_genid_ipv4(dev_net(dev
));
1461 rth
->rt_flags
= RTCF_MULTICAST
;
1462 rth
->rt_type
= RTN_MULTICAST
;
1463 rth
->rt_is_input
= 1;
1466 rth
->rt_gateway
= 0;
1467 rth
->rt_uses_gateway
= 0;
1468 INIT_LIST_HEAD(&rth
->rt_uncached
);
1470 rth
->dst
.input
= ip_local_deliver
;
1471 rth
->rt_flags
|= RTCF_LOCAL
;
1474 #ifdef CONFIG_IP_MROUTE
1475 if (!ipv4_is_local_multicast(daddr
) && IN_DEV_MFORWARD(in_dev
))
1476 rth
->dst
.input
= ip_mr_input
;
1478 RT_CACHE_STAT_INC(in_slow_mc
);
1480 skb_dst_set(skb
, &rth
->dst
);
1492 static void ip_handle_martian_source(struct net_device
*dev
,
1493 struct in_device
*in_dev
,
1494 struct sk_buff
*skb
,
1498 RT_CACHE_STAT_INC(in_martian_src
);
1499 #ifdef CONFIG_IP_ROUTE_VERBOSE
1500 if (IN_DEV_LOG_MARTIANS(in_dev
) && net_ratelimit()) {
1502 * RFC1812 recommendation, if source is martian,
1503 * the only hint is MAC header.
1505 pr_warn("martian source %pI4 from %pI4, on dev %s\n",
1506 &daddr
, &saddr
, dev
->name
);
1507 if (dev
->hard_header_len
&& skb_mac_header_was_set(skb
)) {
1508 print_hex_dump(KERN_WARNING
, "ll header: ",
1509 DUMP_PREFIX_OFFSET
, 16, 1,
1510 skb_mac_header(skb
),
1511 dev
->hard_header_len
, true);
1517 /* called in rcu_read_lock() section */
1518 static int __mkroute_input(struct sk_buff
*skb
,
1519 const struct fib_result
*res
,
1520 struct in_device
*in_dev
,
1521 __be32 daddr
, __be32 saddr
, u32 tos
)
1523 struct fib_nh_exception
*fnhe
;
1526 struct in_device
*out_dev
;
1527 unsigned int flags
= 0;
1531 /* get a working reference to the output device */
1532 out_dev
= __in_dev_get_rcu(FIB_RES_DEV(*res
));
1533 if (out_dev
== NULL
) {
1534 net_crit_ratelimited("Bug in ip_route_input_slow(). Please report.\n");
1538 err
= fib_validate_source(skb
, saddr
, daddr
, tos
, FIB_RES_OIF(*res
),
1539 in_dev
->dev
, in_dev
, &itag
);
1541 ip_handle_martian_source(in_dev
->dev
, in_dev
, skb
, daddr
,
1547 do_cache
= res
->fi
&& !itag
;
1548 if (out_dev
== in_dev
&& err
&& IN_DEV_TX_REDIRECTS(out_dev
) &&
1549 (IN_DEV_SHARED_MEDIA(out_dev
) ||
1550 inet_addr_onlink(out_dev
, saddr
, FIB_RES_GW(*res
)))) {
1551 flags
|= RTCF_DOREDIRECT
;
1555 if (skb
->protocol
!= htons(ETH_P_IP
)) {
1556 /* Not IP (i.e. ARP). Do not create route, if it is
1557 * invalid for proxy arp. DNAT routes are always valid.
1559 * Proxy arp feature have been extended to allow, ARP
1560 * replies back to the same interface, to support
1561 * Private VLAN switch technologies. See arp.c.
1563 if (out_dev
== in_dev
&&
1564 IN_DEV_PROXY_ARP_PVLAN(in_dev
) == 0) {
1570 fnhe
= find_exception(&FIB_RES_NH(*res
), daddr
);
1573 rth
= rcu_dereference(fnhe
->fnhe_rth_input
);
1575 rth
= rcu_dereference(FIB_RES_NH(*res
).nh_rth_input
);
1577 if (rt_cache_valid(rth
)) {
1578 skb_dst_set_noref(skb
, &rth
->dst
);
1583 rth
= rt_dst_alloc(out_dev
->dev
,
1584 IN_DEV_CONF_GET(in_dev
, NOPOLICY
),
1585 IN_DEV_CONF_GET(out_dev
, NOXFRM
), do_cache
);
1591 rth
->rt_genid
= rt_genid_ipv4(dev_net(rth
->dst
.dev
));
1592 rth
->rt_flags
= flags
;
1593 rth
->rt_type
= res
->type
;
1594 rth
->rt_is_input
= 1;
1597 rth
->rt_gateway
= 0;
1598 rth
->rt_uses_gateway
= 0;
1599 INIT_LIST_HEAD(&rth
->rt_uncached
);
1601 rth
->dst
.input
= ip_forward
;
1602 rth
->dst
.output
= ip_output
;
1604 rt_set_nexthop(rth
, daddr
, res
, fnhe
, res
->fi
, res
->type
, itag
);
1605 skb_dst_set(skb
, &rth
->dst
);
1612 static int ip_mkroute_input(struct sk_buff
*skb
,
1613 struct fib_result
*res
,
1614 const struct flowi4
*fl4
,
1615 struct in_device
*in_dev
,
1616 __be32 daddr
, __be32 saddr
, u32 tos
)
1618 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1619 if (res
->fi
&& res
->fi
->fib_nhs
> 1)
1620 fib_select_multipath(res
);
1623 /* create a routing cache entry */
1624 return __mkroute_input(skb
, res
, in_dev
, daddr
, saddr
, tos
);
1628 * NOTE. We drop all the packets that has local source
1629 * addresses, because every properly looped back packet
1630 * must have correct destination already attached by output routine.
1632 * Such approach solves two big problems:
1633 * 1. Not simplex devices are handled properly.
1634 * 2. IP spoofing attempts are filtered with 100% of guarantee.
1635 * called with rcu_read_lock()
1638 static int ip_route_input_slow(struct sk_buff
*skb
, __be32 daddr
, __be32 saddr
,
1639 u8 tos
, struct net_device
*dev
)
1641 struct fib_result res
;
1642 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
1644 unsigned int flags
= 0;
1648 struct net
*net
= dev_net(dev
);
1651 /* IP on this device is disabled. */
1656 /* Check for the most weird martians, which can be not detected
1660 if (ipv4_is_multicast(saddr
) || ipv4_is_lbcast(saddr
))
1661 goto martian_source
;
1664 if (ipv4_is_lbcast(daddr
) || (saddr
== 0 && daddr
== 0))
1667 /* Accept zero addresses only to limited broadcast;
1668 * I even do not know to fix it or not. Waiting for complains :-)
1670 if (ipv4_is_zeronet(saddr
))
1671 goto martian_source
;
1673 if (ipv4_is_zeronet(daddr
))
1674 goto martian_destination
;
1676 /* Following code try to avoid calling IN_DEV_NET_ROUTE_LOCALNET(),
1677 * and call it once if daddr or/and saddr are loopback addresses
1679 if (ipv4_is_loopback(daddr
)) {
1680 if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev
, net
))
1681 goto martian_destination
;
1682 } else if (ipv4_is_loopback(saddr
)) {
1683 if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev
, net
))
1684 goto martian_source
;
1688 * Now we are ready to route packet.
1691 fl4
.flowi4_iif
= dev
->ifindex
;
1692 fl4
.flowi4_mark
= skb
->mark
;
1693 fl4
.flowi4_tos
= tos
;
1694 fl4
.flowi4_scope
= RT_SCOPE_UNIVERSE
;
1697 err
= fib_lookup(net
, &fl4
, &res
);
1701 RT_CACHE_STAT_INC(in_slow_tot
);
1703 if (res
.type
== RTN_BROADCAST
)
1706 if (res
.type
== RTN_LOCAL
) {
1707 err
= fib_validate_source(skb
, saddr
, daddr
, tos
,
1709 dev
, in_dev
, &itag
);
1711 goto martian_source_keep_err
;
1715 if (!IN_DEV_FORWARD(in_dev
))
1717 if (res
.type
!= RTN_UNICAST
)
1718 goto martian_destination
;
1720 err
= ip_mkroute_input(skb
, &res
, &fl4
, in_dev
, daddr
, saddr
, tos
);
1724 if (skb
->protocol
!= htons(ETH_P_IP
))
1727 if (!ipv4_is_zeronet(saddr
)) {
1728 err
= fib_validate_source(skb
, saddr
, 0, tos
, 0, dev
,
1731 goto martian_source_keep_err
;
1733 flags
|= RTCF_BROADCAST
;
1734 res
.type
= RTN_BROADCAST
;
1735 RT_CACHE_STAT_INC(in_brd
);
1741 rth
= rcu_dereference(FIB_RES_NH(res
).nh_rth_input
);
1742 if (rt_cache_valid(rth
)) {
1743 skb_dst_set_noref(skb
, &rth
->dst
);
1751 rth
= rt_dst_alloc(net
->loopback_dev
,
1752 IN_DEV_CONF_GET(in_dev
, NOPOLICY
), false, do_cache
);
1756 rth
->dst
.input
= ip_local_deliver
;
1757 rth
->dst
.output
= ip_rt_bug
;
1758 #ifdef CONFIG_IP_ROUTE_CLASSID
1759 rth
->dst
.tclassid
= itag
;
1762 rth
->rt_genid
= rt_genid_ipv4(net
);
1763 rth
->rt_flags
= flags
|RTCF_LOCAL
;
1764 rth
->rt_type
= res
.type
;
1765 rth
->rt_is_input
= 1;
1768 rth
->rt_gateway
= 0;
1769 rth
->rt_uses_gateway
= 0;
1770 INIT_LIST_HEAD(&rth
->rt_uncached
);
1771 if (res
.type
== RTN_UNREACHABLE
) {
1772 rth
->dst
.input
= ip_error
;
1773 rth
->dst
.error
= -err
;
1774 rth
->rt_flags
&= ~RTCF_LOCAL
;
1777 if (unlikely(!rt_cache_route(&FIB_RES_NH(res
), rth
))) {
1778 rth
->dst
.flags
|= DST_NOCACHE
;
1779 rt_add_uncached_list(rth
);
1782 skb_dst_set(skb
, &rth
->dst
);
1787 RT_CACHE_STAT_INC(in_no_route
);
1788 res
.type
= RTN_UNREACHABLE
;
1794 * Do not cache martian addresses: they should be logged (RFC1812)
1796 martian_destination
:
1797 RT_CACHE_STAT_INC(in_martian_dst
);
1798 #ifdef CONFIG_IP_ROUTE_VERBOSE
1799 if (IN_DEV_LOG_MARTIANS(in_dev
))
1800 net_warn_ratelimited("martian destination %pI4 from %pI4, dev %s\n",
1801 &daddr
, &saddr
, dev
->name
);
1814 martian_source_keep_err
:
1815 ip_handle_martian_source(dev
, in_dev
, skb
, daddr
, saddr
);
1819 int ip_route_input_noref(struct sk_buff
*skb
, __be32 daddr
, __be32 saddr
,
1820 u8 tos
, struct net_device
*dev
)
1826 /* Multicast recognition logic is moved from route cache to here.
1827 The problem was that too many Ethernet cards have broken/missing
1828 hardware multicast filters :-( As result the host on multicasting
1829 network acquires a lot of useless route cache entries, sort of
1830 SDR messages from all the world. Now we try to get rid of them.
1831 Really, provided software IP multicast filter is organized
1832 reasonably (at least, hashed), it does not result in a slowdown
1833 comparing with route cache reject entries.
1834 Note, that multicast routers are not affected, because
1835 route cache entry is created eventually.
1837 if (ipv4_is_multicast(daddr
)) {
1838 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
1841 int our
= ip_check_mc_rcu(in_dev
, daddr
, saddr
,
1842 ip_hdr(skb
)->protocol
);
1844 #ifdef CONFIG_IP_MROUTE
1846 (!ipv4_is_local_multicast(daddr
) &&
1847 IN_DEV_MFORWARD(in_dev
))
1850 int res
= ip_route_input_mc(skb
, daddr
, saddr
,
1859 res
= ip_route_input_slow(skb
, daddr
, saddr
, tos
, dev
);
1863 EXPORT_SYMBOL(ip_route_input_noref
);
1865 /* called with rcu_read_lock() */
1866 static struct rtable
*__mkroute_output(const struct fib_result
*res
,
1867 const struct flowi4
*fl4
, int orig_oif
,
1868 struct net_device
*dev_out
,
1871 struct fib_info
*fi
= res
->fi
;
1872 struct fib_nh_exception
*fnhe
;
1873 struct in_device
*in_dev
;
1874 u16 type
= res
->type
;
1878 in_dev
= __in_dev_get_rcu(dev_out
);
1880 return ERR_PTR(-EINVAL
);
1882 if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev
)))
1883 if (ipv4_is_loopback(fl4
->saddr
) && !(dev_out
->flags
& IFF_LOOPBACK
))
1884 return ERR_PTR(-EINVAL
);
1886 if (ipv4_is_lbcast(fl4
->daddr
))
1887 type
= RTN_BROADCAST
;
1888 else if (ipv4_is_multicast(fl4
->daddr
))
1889 type
= RTN_MULTICAST
;
1890 else if (ipv4_is_zeronet(fl4
->daddr
))
1891 return ERR_PTR(-EINVAL
);
1893 if (dev_out
->flags
& IFF_LOOPBACK
)
1894 flags
|= RTCF_LOCAL
;
1897 if (type
== RTN_BROADCAST
) {
1898 flags
|= RTCF_BROADCAST
| RTCF_LOCAL
;
1900 } else if (type
== RTN_MULTICAST
) {
1901 flags
|= RTCF_MULTICAST
| RTCF_LOCAL
;
1902 if (!ip_check_mc_rcu(in_dev
, fl4
->daddr
, fl4
->saddr
,
1904 flags
&= ~RTCF_LOCAL
;
1907 /* If multicast route do not exist use
1908 * default one, but do not gateway in this case.
1911 if (fi
&& res
->prefixlen
< 4)
1916 do_cache
&= fi
!= NULL
;
1918 struct rtable __rcu
**prth
;
1919 struct fib_nh
*nh
= &FIB_RES_NH(*res
);
1921 fnhe
= find_exception(nh
, fl4
->daddr
);
1923 prth
= &fnhe
->fnhe_rth_output
;
1925 if (unlikely(fl4
->flowi4_flags
&
1926 FLOWI_FLAG_KNOWN_NH
&&
1928 nh
->nh_scope
== RT_SCOPE_LINK
))) {
1932 prth
= __this_cpu_ptr(nh
->nh_pcpu_rth_output
);
1934 rth
= rcu_dereference(*prth
);
1935 if (rt_cache_valid(rth
)) {
1936 dst_hold(&rth
->dst
);
1942 rth
= rt_dst_alloc(dev_out
,
1943 IN_DEV_CONF_GET(in_dev
, NOPOLICY
),
1944 IN_DEV_CONF_GET(in_dev
, NOXFRM
),
1947 return ERR_PTR(-ENOBUFS
);
1949 rth
->dst
.output
= ip_output
;
1951 rth
->rt_genid
= rt_genid_ipv4(dev_net(dev_out
));
1952 rth
->rt_flags
= flags
;
1953 rth
->rt_type
= type
;
1954 rth
->rt_is_input
= 0;
1955 rth
->rt_iif
= orig_oif
? : 0;
1957 rth
->rt_gateway
= 0;
1958 rth
->rt_uses_gateway
= 0;
1959 INIT_LIST_HEAD(&rth
->rt_uncached
);
1961 RT_CACHE_STAT_INC(out_slow_tot
);
1963 if (flags
& RTCF_LOCAL
)
1964 rth
->dst
.input
= ip_local_deliver
;
1965 if (flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
)) {
1966 if (flags
& RTCF_LOCAL
&&
1967 !(dev_out
->flags
& IFF_LOOPBACK
)) {
1968 rth
->dst
.output
= ip_mc_output
;
1969 RT_CACHE_STAT_INC(out_slow_mc
);
1971 #ifdef CONFIG_IP_MROUTE
1972 if (type
== RTN_MULTICAST
) {
1973 if (IN_DEV_MFORWARD(in_dev
) &&
1974 !ipv4_is_local_multicast(fl4
->daddr
)) {
1975 rth
->dst
.input
= ip_mr_input
;
1976 rth
->dst
.output
= ip_mc_output
;
1982 rt_set_nexthop(rth
, fl4
->daddr
, res
, fnhe
, fi
, type
, 0);
1988 * Major route resolver routine.
1991 struct rtable
*__ip_route_output_key(struct net
*net
, struct flowi4
*fl4
)
1993 struct net_device
*dev_out
= NULL
;
1994 __u8 tos
= RT_FL_TOS(fl4
);
1995 unsigned int flags
= 0;
1996 struct fib_result res
;
2004 orig_oif
= fl4
->flowi4_oif
;
2006 fl4
->flowi4_iif
= LOOPBACK_IFINDEX
;
2007 fl4
->flowi4_tos
= tos
& IPTOS_RT_MASK
;
2008 fl4
->flowi4_scope
= ((tos
& RTO_ONLINK
) ?
2009 RT_SCOPE_LINK
: RT_SCOPE_UNIVERSE
);
2013 rth
= ERR_PTR(-EINVAL
);
2014 if (ipv4_is_multicast(fl4
->saddr
) ||
2015 ipv4_is_lbcast(fl4
->saddr
) ||
2016 ipv4_is_zeronet(fl4
->saddr
))
2019 /* I removed check for oif == dev_out->oif here.
2020 It was wrong for two reasons:
2021 1. ip_dev_find(net, saddr) can return wrong iface, if saddr
2022 is assigned to multiple interfaces.
2023 2. Moreover, we are allowed to send packets with saddr
2024 of another iface. --ANK
2027 if (fl4
->flowi4_oif
== 0 &&
2028 (ipv4_is_multicast(fl4
->daddr
) ||
2029 ipv4_is_lbcast(fl4
->daddr
))) {
2030 /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2031 dev_out
= __ip_dev_find(net
, fl4
->saddr
, false);
2032 if (dev_out
== NULL
)
2035 /* Special hack: user can direct multicasts
2036 and limited broadcast via necessary interface
2037 without fiddling with IP_MULTICAST_IF or IP_PKTINFO.
2038 This hack is not just for fun, it allows
2039 vic,vat and friends to work.
2040 They bind socket to loopback, set ttl to zero
2041 and expect that it will work.
2042 From the viewpoint of routing cache they are broken,
2043 because we are not allowed to build multicast path
2044 with loopback source addr (look, routing cache
2045 cannot know, that ttl is zero, so that packet
2046 will not leave this host and route is valid).
2047 Luckily, this hack is good workaround.
2050 fl4
->flowi4_oif
= dev_out
->ifindex
;
2054 if (!(fl4
->flowi4_flags
& FLOWI_FLAG_ANYSRC
)) {
2055 /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2056 if (!__ip_dev_find(net
, fl4
->saddr
, false))
2062 if (fl4
->flowi4_oif
) {
2063 dev_out
= dev_get_by_index_rcu(net
, fl4
->flowi4_oif
);
2064 rth
= ERR_PTR(-ENODEV
);
2065 if (dev_out
== NULL
)
2068 /* RACE: Check return value of inet_select_addr instead. */
2069 if (!(dev_out
->flags
& IFF_UP
) || !__in_dev_get_rcu(dev_out
)) {
2070 rth
= ERR_PTR(-ENETUNREACH
);
2073 if (ipv4_is_local_multicast(fl4
->daddr
) ||
2074 ipv4_is_lbcast(fl4
->daddr
)) {
2076 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2081 if (ipv4_is_multicast(fl4
->daddr
))
2082 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2084 else if (!fl4
->daddr
)
2085 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2091 fl4
->daddr
= fl4
->saddr
;
2093 fl4
->daddr
= fl4
->saddr
= htonl(INADDR_LOOPBACK
);
2094 dev_out
= net
->loopback_dev
;
2095 fl4
->flowi4_oif
= LOOPBACK_IFINDEX
;
2096 res
.type
= RTN_LOCAL
;
2097 flags
|= RTCF_LOCAL
;
2101 if (fib_lookup(net
, fl4
, &res
)) {
2104 if (fl4
->flowi4_oif
) {
2105 /* Apparently, routing tables are wrong. Assume,
2106 that the destination is on link.
2109 Because we are allowed to send to iface
2110 even if it has NO routes and NO assigned
2111 addresses. When oif is specified, routing
2112 tables are looked up with only one purpose:
2113 to catch if destination is gatewayed, rather than
2114 direct. Moreover, if MSG_DONTROUTE is set,
2115 we send packet, ignoring both routing tables
2116 and ifaddr state. --ANK
2119 We could make it even if oif is unknown,
2120 likely IPv6, but we do not.
2123 if (fl4
->saddr
== 0)
2124 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2126 res
.type
= RTN_UNICAST
;
2129 rth
= ERR_PTR(-ENETUNREACH
);
2133 if (res
.type
== RTN_LOCAL
) {
2135 if (res
.fi
->fib_prefsrc
)
2136 fl4
->saddr
= res
.fi
->fib_prefsrc
;
2138 fl4
->saddr
= fl4
->daddr
;
2140 dev_out
= net
->loopback_dev
;
2141 fl4
->flowi4_oif
= dev_out
->ifindex
;
2142 flags
|= RTCF_LOCAL
;
2146 #ifdef CONFIG_IP_ROUTE_MULTIPATH
2147 if (res
.fi
->fib_nhs
> 1 && fl4
->flowi4_oif
== 0)
2148 fib_select_multipath(&res
);
2151 if (!res
.prefixlen
&&
2152 res
.table
->tb_num_default
> 1 &&
2153 res
.type
== RTN_UNICAST
&& !fl4
->flowi4_oif
)
2154 fib_select_default(&res
);
2157 fl4
->saddr
= FIB_RES_PREFSRC(net
, res
);
2159 dev_out
= FIB_RES_DEV(res
);
2160 fl4
->flowi4_oif
= dev_out
->ifindex
;
2164 rth
= __mkroute_output(&res
, fl4
, orig_oif
, dev_out
, flags
);
2170 EXPORT_SYMBOL_GPL(__ip_route_output_key
);
2172 static struct dst_entry
*ipv4_blackhole_dst_check(struct dst_entry
*dst
, u32 cookie
)
2177 static unsigned int ipv4_blackhole_mtu(const struct dst_entry
*dst
)
2179 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
2181 return mtu
? : dst
->dev
->mtu
;
2184 static void ipv4_rt_blackhole_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
2185 struct sk_buff
*skb
, u32 mtu
)
2189 static void ipv4_rt_blackhole_redirect(struct dst_entry
*dst
, struct sock
*sk
,
2190 struct sk_buff
*skb
)
2194 static u32
*ipv4_rt_blackhole_cow_metrics(struct dst_entry
*dst
,
2200 static struct dst_ops ipv4_dst_blackhole_ops
= {
2202 .protocol
= cpu_to_be16(ETH_P_IP
),
2203 .check
= ipv4_blackhole_dst_check
,
2204 .mtu
= ipv4_blackhole_mtu
,
2205 .default_advmss
= ipv4_default_advmss
,
2206 .update_pmtu
= ipv4_rt_blackhole_update_pmtu
,
2207 .redirect
= ipv4_rt_blackhole_redirect
,
2208 .cow_metrics
= ipv4_rt_blackhole_cow_metrics
,
2209 .neigh_lookup
= ipv4_neigh_lookup
,
2212 struct dst_entry
*ipv4_blackhole_route(struct net
*net
, struct dst_entry
*dst_orig
)
2214 struct rtable
*ort
= (struct rtable
*) dst_orig
;
2217 rt
= dst_alloc(&ipv4_dst_blackhole_ops
, NULL
, 1, DST_OBSOLETE_NONE
, 0);
2219 struct dst_entry
*new = &rt
->dst
;
2222 new->input
= dst_discard
;
2223 new->output
= dst_discard
;
2225 new->dev
= ort
->dst
.dev
;
2229 rt
->rt_is_input
= ort
->rt_is_input
;
2230 rt
->rt_iif
= ort
->rt_iif
;
2231 rt
->rt_pmtu
= ort
->rt_pmtu
;
2233 rt
->rt_genid
= rt_genid_ipv4(net
);
2234 rt
->rt_flags
= ort
->rt_flags
;
2235 rt
->rt_type
= ort
->rt_type
;
2236 rt
->rt_gateway
= ort
->rt_gateway
;
2237 rt
->rt_uses_gateway
= ort
->rt_uses_gateway
;
2239 INIT_LIST_HEAD(&rt
->rt_uncached
);
2244 dst_release(dst_orig
);
2246 return rt
? &rt
->dst
: ERR_PTR(-ENOMEM
);
2249 struct rtable
*ip_route_output_flow(struct net
*net
, struct flowi4
*flp4
,
2252 struct rtable
*rt
= __ip_route_output_key(net
, flp4
);
2257 if (flp4
->flowi4_proto
)
2258 rt
= (struct rtable
*) xfrm_lookup(net
, &rt
->dst
,
2259 flowi4_to_flowi(flp4
),
2264 EXPORT_SYMBOL_GPL(ip_route_output_flow
);
2266 static int rt_fill_info(struct net
*net
, __be32 dst
, __be32 src
,
2267 struct flowi4
*fl4
, struct sk_buff
*skb
, u32 portid
,
2268 u32 seq
, int event
, int nowait
, unsigned int flags
)
2270 struct rtable
*rt
= skb_rtable(skb
);
2272 struct nlmsghdr
*nlh
;
2273 unsigned long expires
= 0;
2275 u32 metrics
[RTAX_MAX
];
2277 nlh
= nlmsg_put(skb
, portid
, seq
, event
, sizeof(*r
), flags
);
2281 r
= nlmsg_data(nlh
);
2282 r
->rtm_family
= AF_INET
;
2283 r
->rtm_dst_len
= 32;
2285 r
->rtm_tos
= fl4
->flowi4_tos
;
2286 r
->rtm_table
= RT_TABLE_MAIN
;
2287 if (nla_put_u32(skb
, RTA_TABLE
, RT_TABLE_MAIN
))
2288 goto nla_put_failure
;
2289 r
->rtm_type
= rt
->rt_type
;
2290 r
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2291 r
->rtm_protocol
= RTPROT_UNSPEC
;
2292 r
->rtm_flags
= (rt
->rt_flags
& ~0xFFFF) | RTM_F_CLONED
;
2293 if (rt
->rt_flags
& RTCF_NOTIFY
)
2294 r
->rtm_flags
|= RTM_F_NOTIFY
;
2296 if (nla_put_be32(skb
, RTA_DST
, dst
))
2297 goto nla_put_failure
;
2299 r
->rtm_src_len
= 32;
2300 if (nla_put_be32(skb
, RTA_SRC
, src
))
2301 goto nla_put_failure
;
2304 nla_put_u32(skb
, RTA_OIF
, rt
->dst
.dev
->ifindex
))
2305 goto nla_put_failure
;
2306 #ifdef CONFIG_IP_ROUTE_CLASSID
2307 if (rt
->dst
.tclassid
&&
2308 nla_put_u32(skb
, RTA_FLOW
, rt
->dst
.tclassid
))
2309 goto nla_put_failure
;
2311 if (!rt_is_input_route(rt
) &&
2312 fl4
->saddr
!= src
) {
2313 if (nla_put_be32(skb
, RTA_PREFSRC
, fl4
->saddr
))
2314 goto nla_put_failure
;
2316 if (rt
->rt_uses_gateway
&&
2317 nla_put_be32(skb
, RTA_GATEWAY
, rt
->rt_gateway
))
2318 goto nla_put_failure
;
2320 expires
= rt
->dst
.expires
;
2322 unsigned long now
= jiffies
;
2324 if (time_before(now
, expires
))
2330 memcpy(metrics
, dst_metrics_ptr(&rt
->dst
), sizeof(metrics
));
2331 if (rt
->rt_pmtu
&& expires
)
2332 metrics
[RTAX_MTU
- 1] = rt
->rt_pmtu
;
2333 if (rtnetlink_put_metrics(skb
, metrics
) < 0)
2334 goto nla_put_failure
;
2336 if (fl4
->flowi4_mark
&&
2337 nla_put_u32(skb
, RTA_MARK
, fl4
->flowi4_mark
))
2338 goto nla_put_failure
;
2340 error
= rt
->dst
.error
;
2342 if (rt_is_input_route(rt
)) {
2343 #ifdef CONFIG_IP_MROUTE
2344 if (ipv4_is_multicast(dst
) && !ipv4_is_local_multicast(dst
) &&
2345 IPV4_DEVCONF_ALL(net
, MC_FORWARDING
)) {
2346 int err
= ipmr_get_route(net
, skb
,
2347 fl4
->saddr
, fl4
->daddr
,
2353 goto nla_put_failure
;
2355 if (err
== -EMSGSIZE
)
2356 goto nla_put_failure
;
2362 if (nla_put_u32(skb
, RTA_IIF
, rt
->rt_iif
))
2363 goto nla_put_failure
;
2366 if (rtnl_put_cacheinfo(skb
, &rt
->dst
, 0, expires
, error
) < 0)
2367 goto nla_put_failure
;
2369 return nlmsg_end(skb
, nlh
);
2372 nlmsg_cancel(skb
, nlh
);
2376 static int inet_rtm_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
*nlh
)
2378 struct net
*net
= sock_net(in_skb
->sk
);
2380 struct nlattr
*tb
[RTA_MAX
+1];
2381 struct rtable
*rt
= NULL
;
2388 struct sk_buff
*skb
;
2390 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv4_policy
);
2394 rtm
= nlmsg_data(nlh
);
2396 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
2402 /* Reserve room for dummy headers, this skb can pass
2403 through good chunk of routing engine.
2405 skb_reset_mac_header(skb
);
2406 skb_reset_network_header(skb
);
2408 /* Bugfix: need to give ip_route_input enough of an IP header to not gag. */
2409 ip_hdr(skb
)->protocol
= IPPROTO_ICMP
;
2410 skb_reserve(skb
, MAX_HEADER
+ sizeof(struct iphdr
));
2412 src
= tb
[RTA_SRC
] ? nla_get_be32(tb
[RTA_SRC
]) : 0;
2413 dst
= tb
[RTA_DST
] ? nla_get_be32(tb
[RTA_DST
]) : 0;
2414 iif
= tb
[RTA_IIF
] ? nla_get_u32(tb
[RTA_IIF
]) : 0;
2415 mark
= tb
[RTA_MARK
] ? nla_get_u32(tb
[RTA_MARK
]) : 0;
2417 memset(&fl4
, 0, sizeof(fl4
));
2420 fl4
.flowi4_tos
= rtm
->rtm_tos
;
2421 fl4
.flowi4_oif
= tb
[RTA_OIF
] ? nla_get_u32(tb
[RTA_OIF
]) : 0;
2422 fl4
.flowi4_mark
= mark
;
2425 struct net_device
*dev
;
2427 dev
= __dev_get_by_index(net
, iif
);
2433 skb
->protocol
= htons(ETH_P_IP
);
2437 err
= ip_route_input(skb
, dst
, src
, rtm
->rtm_tos
, dev
);
2440 rt
= skb_rtable(skb
);
2441 if (err
== 0 && rt
->dst
.error
)
2442 err
= -rt
->dst
.error
;
2444 rt
= ip_route_output_key(net
, &fl4
);
2454 skb_dst_set(skb
, &rt
->dst
);
2455 if (rtm
->rtm_flags
& RTM_F_NOTIFY
)
2456 rt
->rt_flags
|= RTCF_NOTIFY
;
2458 err
= rt_fill_info(net
, dst
, src
, &fl4
, skb
,
2459 NETLINK_CB(in_skb
).portid
, nlh
->nlmsg_seq
,
2460 RTM_NEWROUTE
, 0, 0);
2464 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).portid
);
2473 int ip_rt_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2478 void ip_rt_multicast_event(struct in_device
*in_dev
)
2480 rt_cache_flush(dev_net(in_dev
->dev
));
2483 #ifdef CONFIG_SYSCTL
2484 static int ip_rt_gc_timeout __read_mostly
= RT_GC_TIMEOUT
;
2485 static int ip_rt_gc_interval __read_mostly
= 60 * HZ
;
2486 static int ip_rt_gc_min_interval __read_mostly
= HZ
/ 2;
2487 static int ip_rt_gc_elasticity __read_mostly
= 8;
2489 static int ipv4_sysctl_rtcache_flush(struct ctl_table
*__ctl
, int write
,
2490 void __user
*buffer
,
2491 size_t *lenp
, loff_t
*ppos
)
2493 struct net
*net
= (struct net
*)__ctl
->extra1
;
2496 rt_cache_flush(net
);
2497 fnhe_genid_bump(net
);
2504 static struct ctl_table ipv4_route_table
[] = {
2506 .procname
= "gc_thresh",
2507 .data
= &ipv4_dst_ops
.gc_thresh
,
2508 .maxlen
= sizeof(int),
2510 .proc_handler
= proc_dointvec
,
2513 .procname
= "max_size",
2514 .data
= &ip_rt_max_size
,
2515 .maxlen
= sizeof(int),
2517 .proc_handler
= proc_dointvec
,
2520 /* Deprecated. Use gc_min_interval_ms */
2522 .procname
= "gc_min_interval",
2523 .data
= &ip_rt_gc_min_interval
,
2524 .maxlen
= sizeof(int),
2526 .proc_handler
= proc_dointvec_jiffies
,
2529 .procname
= "gc_min_interval_ms",
2530 .data
= &ip_rt_gc_min_interval
,
2531 .maxlen
= sizeof(int),
2533 .proc_handler
= proc_dointvec_ms_jiffies
,
2536 .procname
= "gc_timeout",
2537 .data
= &ip_rt_gc_timeout
,
2538 .maxlen
= sizeof(int),
2540 .proc_handler
= proc_dointvec_jiffies
,
2543 .procname
= "gc_interval",
2544 .data
= &ip_rt_gc_interval
,
2545 .maxlen
= sizeof(int),
2547 .proc_handler
= proc_dointvec_jiffies
,
2550 .procname
= "redirect_load",
2551 .data
= &ip_rt_redirect_load
,
2552 .maxlen
= sizeof(int),
2554 .proc_handler
= proc_dointvec
,
2557 .procname
= "redirect_number",
2558 .data
= &ip_rt_redirect_number
,
2559 .maxlen
= sizeof(int),
2561 .proc_handler
= proc_dointvec
,
2564 .procname
= "redirect_silence",
2565 .data
= &ip_rt_redirect_silence
,
2566 .maxlen
= sizeof(int),
2568 .proc_handler
= proc_dointvec
,
2571 .procname
= "error_cost",
2572 .data
= &ip_rt_error_cost
,
2573 .maxlen
= sizeof(int),
2575 .proc_handler
= proc_dointvec
,
2578 .procname
= "error_burst",
2579 .data
= &ip_rt_error_burst
,
2580 .maxlen
= sizeof(int),
2582 .proc_handler
= proc_dointvec
,
2585 .procname
= "gc_elasticity",
2586 .data
= &ip_rt_gc_elasticity
,
2587 .maxlen
= sizeof(int),
2589 .proc_handler
= proc_dointvec
,
2592 .procname
= "mtu_expires",
2593 .data
= &ip_rt_mtu_expires
,
2594 .maxlen
= sizeof(int),
2596 .proc_handler
= proc_dointvec_jiffies
,
2599 .procname
= "min_pmtu",
2600 .data
= &ip_rt_min_pmtu
,
2601 .maxlen
= sizeof(int),
2603 .proc_handler
= proc_dointvec
,
2606 .procname
= "min_adv_mss",
2607 .data
= &ip_rt_min_advmss
,
2608 .maxlen
= sizeof(int),
2610 .proc_handler
= proc_dointvec
,
2615 static struct ctl_table ipv4_route_flush_table
[] = {
2617 .procname
= "flush",
2618 .maxlen
= sizeof(int),
2620 .proc_handler
= ipv4_sysctl_rtcache_flush
,
2625 static __net_init
int sysctl_route_net_init(struct net
*net
)
2627 struct ctl_table
*tbl
;
2629 tbl
= ipv4_route_flush_table
;
2630 if (!net_eq(net
, &init_net
)) {
2631 tbl
= kmemdup(tbl
, sizeof(ipv4_route_flush_table
), GFP_KERNEL
);
2635 /* Don't export sysctls to unprivileged users */
2636 if (net
->user_ns
!= &init_user_ns
)
2637 tbl
[0].procname
= NULL
;
2639 tbl
[0].extra1
= net
;
2641 net
->ipv4
.route_hdr
= register_net_sysctl(net
, "net/ipv4/route", tbl
);
2642 if (net
->ipv4
.route_hdr
== NULL
)
2647 if (tbl
!= ipv4_route_flush_table
)
2653 static __net_exit
void sysctl_route_net_exit(struct net
*net
)
2655 struct ctl_table
*tbl
;
2657 tbl
= net
->ipv4
.route_hdr
->ctl_table_arg
;
2658 unregister_net_sysctl_table(net
->ipv4
.route_hdr
);
2659 BUG_ON(tbl
== ipv4_route_flush_table
);
2663 static __net_initdata
struct pernet_operations sysctl_route_ops
= {
2664 .init
= sysctl_route_net_init
,
2665 .exit
= sysctl_route_net_exit
,
2669 static __net_init
int rt_genid_init(struct net
*net
)
2671 atomic_set(&net
->ipv4
.rt_genid
, 0);
2672 atomic_set(&net
->fnhe_genid
, 0);
2673 get_random_bytes(&net
->ipv4
.dev_addr_genid
,
2674 sizeof(net
->ipv4
.dev_addr_genid
));
2678 static __net_initdata
struct pernet_operations rt_genid_ops
= {
2679 .init
= rt_genid_init
,
2682 static int __net_init
ipv4_inetpeer_init(struct net
*net
)
2684 struct inet_peer_base
*bp
= kmalloc(sizeof(*bp
), GFP_KERNEL
);
2688 inet_peer_base_init(bp
);
2689 net
->ipv4
.peers
= bp
;
2693 static void __net_exit
ipv4_inetpeer_exit(struct net
*net
)
2695 struct inet_peer_base
*bp
= net
->ipv4
.peers
;
2697 net
->ipv4
.peers
= NULL
;
2698 inetpeer_invalidate_tree(bp
);
2702 static __net_initdata
struct pernet_operations ipv4_inetpeer_ops
= {
2703 .init
= ipv4_inetpeer_init
,
2704 .exit
= ipv4_inetpeer_exit
,
2707 #ifdef CONFIG_IP_ROUTE_CLASSID
2708 struct ip_rt_acct __percpu
*ip_rt_acct __read_mostly
;
2709 #endif /* CONFIG_IP_ROUTE_CLASSID */
2711 int __init
ip_rt_init(void)
2715 #ifdef CONFIG_IP_ROUTE_CLASSID
2716 ip_rt_acct
= __alloc_percpu(256 * sizeof(struct ip_rt_acct
), __alignof__(struct ip_rt_acct
));
2718 panic("IP: failed to allocate ip_rt_acct\n");
2721 ipv4_dst_ops
.kmem_cachep
=
2722 kmem_cache_create("ip_dst_cache", sizeof(struct rtable
), 0,
2723 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
);
2725 ipv4_dst_blackhole_ops
.kmem_cachep
= ipv4_dst_ops
.kmem_cachep
;
2727 if (dst_entries_init(&ipv4_dst_ops
) < 0)
2728 panic("IP: failed to allocate ipv4_dst_ops counter\n");
2730 if (dst_entries_init(&ipv4_dst_blackhole_ops
) < 0)
2731 panic("IP: failed to allocate ipv4_dst_blackhole_ops counter\n");
2733 ipv4_dst_ops
.gc_thresh
= ~0;
2734 ip_rt_max_size
= INT_MAX
;
2739 if (ip_rt_proc_init())
2740 pr_err("Unable to create route proc files\n");
2745 rtnl_register(PF_INET
, RTM_GETROUTE
, inet_rtm_getroute
, NULL
, NULL
);
2747 #ifdef CONFIG_SYSCTL
2748 register_pernet_subsys(&sysctl_route_ops
);
2750 register_pernet_subsys(&rt_genid_ops
);
2751 register_pernet_subsys(&ipv4_inetpeer_ops
);
2755 #ifdef CONFIG_SYSCTL
2757 * We really need to sanitize the damn ipv4 init order, then all
2758 * this nonsense will go away.
2760 void __init
ip_static_sysctl_init(void)
2762 register_net_sysctl(&init_net
, "net/ipv4/route", ipv4_route_table
);