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>
92 #include <linux/jhash.h>
94 #include <net/net_namespace.h>
95 #include <net/protocol.h>
97 #include <net/route.h>
98 #include <net/inetpeer.h>
100 #include <net/ip_fib.h>
103 #include <net/icmp.h>
104 #include <net/xfrm.h>
105 #include <net/netevent.h>
106 #include <net/rtnetlink.h>
108 #include <linux/sysctl.h>
109 #include <linux/kmemleak.h>
111 #include <net/secure_seq.h>
113 #define RT_FL_TOS(oldflp4) \
114 ((oldflp4)->flowi4_tos & (IPTOS_RT_MASK | RTO_ONLINK))
116 #define IP_MAX_MTU 0xFFF0
118 #define RT_GC_TIMEOUT (300*HZ)
120 static int ip_rt_max_size
;
121 static int ip_rt_redirect_number __read_mostly
= 9;
122 static int ip_rt_redirect_load __read_mostly
= HZ
/ 50;
123 static int ip_rt_redirect_silence __read_mostly
= ((HZ
/ 50) << (9 + 1));
124 static int ip_rt_error_cost __read_mostly
= HZ
;
125 static int ip_rt_error_burst __read_mostly
= 5 * HZ
;
126 static int ip_rt_mtu_expires __read_mostly
= 10 * 60 * HZ
;
127 static int ip_rt_min_pmtu __read_mostly
= 512 + 20 + 20;
128 static int ip_rt_min_advmss __read_mostly
= 256;
131 * Interface to generic destination cache.
134 static struct dst_entry
*ipv4_dst_check(struct dst_entry
*dst
, u32 cookie
);
135 static unsigned int ipv4_default_advmss(const struct dst_entry
*dst
);
136 static unsigned int ipv4_mtu(const struct dst_entry
*dst
);
137 static struct dst_entry
*ipv4_negative_advice(struct dst_entry
*dst
);
138 static void ipv4_link_failure(struct sk_buff
*skb
);
139 static void ip_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
140 struct sk_buff
*skb
, u32 mtu
);
141 static void ip_do_redirect(struct dst_entry
*dst
, struct sock
*sk
,
142 struct sk_buff
*skb
);
143 static void ipv4_dst_destroy(struct dst_entry
*dst
);
145 static void ipv4_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
,
150 static u32
*ipv4_cow_metrics(struct dst_entry
*dst
, unsigned long old
)
156 static struct neighbour
*ipv4_neigh_lookup(const struct dst_entry
*dst
,
160 static struct dst_ops ipv4_dst_ops
= {
162 .protocol
= cpu_to_be16(ETH_P_IP
),
163 .check
= ipv4_dst_check
,
164 .default_advmss
= ipv4_default_advmss
,
166 .cow_metrics
= ipv4_cow_metrics
,
167 .destroy
= ipv4_dst_destroy
,
168 .ifdown
= ipv4_dst_ifdown
,
169 .negative_advice
= ipv4_negative_advice
,
170 .link_failure
= ipv4_link_failure
,
171 .update_pmtu
= ip_rt_update_pmtu
,
172 .redirect
= ip_do_redirect
,
173 .local_out
= __ip_local_out
,
174 .neigh_lookup
= ipv4_neigh_lookup
,
177 #define ECN_OR_COST(class) TC_PRIO_##class
179 const __u8 ip_tos2prio
[16] = {
181 ECN_OR_COST(BESTEFFORT
),
183 ECN_OR_COST(BESTEFFORT
),
189 ECN_OR_COST(INTERACTIVE
),
191 ECN_OR_COST(INTERACTIVE
),
192 TC_PRIO_INTERACTIVE_BULK
,
193 ECN_OR_COST(INTERACTIVE_BULK
),
194 TC_PRIO_INTERACTIVE_BULK
,
195 ECN_OR_COST(INTERACTIVE_BULK
)
197 EXPORT_SYMBOL(ip_tos2prio
);
199 static DEFINE_PER_CPU(struct rt_cache_stat
, rt_cache_stat
);
200 #define RT_CACHE_STAT_INC(field) __this_cpu_inc(rt_cache_stat.field)
202 #ifdef CONFIG_PROC_FS
203 static void *rt_cache_seq_start(struct seq_file
*seq
, loff_t
*pos
)
207 return SEQ_START_TOKEN
;
210 static void *rt_cache_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
216 static void rt_cache_seq_stop(struct seq_file
*seq
, void *v
)
220 static int rt_cache_seq_show(struct seq_file
*seq
, void *v
)
222 if (v
== SEQ_START_TOKEN
)
223 seq_printf(seq
, "%-127s\n",
224 "Iface\tDestination\tGateway \tFlags\t\tRefCnt\tUse\t"
225 "Metric\tSource\t\tMTU\tWindow\tIRTT\tTOS\tHHRef\t"
230 static const struct seq_operations rt_cache_seq_ops
= {
231 .start
= rt_cache_seq_start
,
232 .next
= rt_cache_seq_next
,
233 .stop
= rt_cache_seq_stop
,
234 .show
= rt_cache_seq_show
,
237 static int rt_cache_seq_open(struct inode
*inode
, struct file
*file
)
239 return seq_open(file
, &rt_cache_seq_ops
);
242 static const struct file_operations rt_cache_seq_fops
= {
243 .owner
= THIS_MODULE
,
244 .open
= rt_cache_seq_open
,
247 .release
= seq_release
,
251 static void *rt_cpu_seq_start(struct seq_file
*seq
, loff_t
*pos
)
256 return SEQ_START_TOKEN
;
258 for (cpu
= *pos
-1; cpu
< nr_cpu_ids
; ++cpu
) {
259 if (!cpu_possible(cpu
))
262 return &per_cpu(rt_cache_stat
, cpu
);
267 static void *rt_cpu_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
271 for (cpu
= *pos
; cpu
< nr_cpu_ids
; ++cpu
) {
272 if (!cpu_possible(cpu
))
275 return &per_cpu(rt_cache_stat
, cpu
);
281 static void rt_cpu_seq_stop(struct seq_file
*seq
, void *v
)
286 static int rt_cpu_seq_show(struct seq_file
*seq
, void *v
)
288 struct rt_cache_stat
*st
= v
;
290 if (v
== SEQ_START_TOKEN
) {
291 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");
295 seq_printf(seq
,"%08x %08x %08x %08x %08x %08x %08x %08x "
296 " %08x %08x %08x %08x %08x %08x %08x %08x %08x \n",
297 dst_entries_get_slow(&ipv4_dst_ops
),
320 static const struct seq_operations rt_cpu_seq_ops
= {
321 .start
= rt_cpu_seq_start
,
322 .next
= rt_cpu_seq_next
,
323 .stop
= rt_cpu_seq_stop
,
324 .show
= rt_cpu_seq_show
,
328 static int rt_cpu_seq_open(struct inode
*inode
, struct file
*file
)
330 return seq_open(file
, &rt_cpu_seq_ops
);
333 static const struct file_operations rt_cpu_seq_fops
= {
334 .owner
= THIS_MODULE
,
335 .open
= rt_cpu_seq_open
,
338 .release
= seq_release
,
341 #ifdef CONFIG_IP_ROUTE_CLASSID
342 static int rt_acct_proc_show(struct seq_file
*m
, void *v
)
344 struct ip_rt_acct
*dst
, *src
;
347 dst
= kcalloc(256, sizeof(struct ip_rt_acct
), GFP_KERNEL
);
351 for_each_possible_cpu(i
) {
352 src
= (struct ip_rt_acct
*)per_cpu_ptr(ip_rt_acct
, i
);
353 for (j
= 0; j
< 256; j
++) {
354 dst
[j
].o_bytes
+= src
[j
].o_bytes
;
355 dst
[j
].o_packets
+= src
[j
].o_packets
;
356 dst
[j
].i_bytes
+= src
[j
].i_bytes
;
357 dst
[j
].i_packets
+= src
[j
].i_packets
;
361 seq_write(m
, dst
, 256 * sizeof(struct ip_rt_acct
));
366 static int rt_acct_proc_open(struct inode
*inode
, struct file
*file
)
368 return single_open(file
, rt_acct_proc_show
, NULL
);
371 static const struct file_operations rt_acct_proc_fops
= {
372 .owner
= THIS_MODULE
,
373 .open
= rt_acct_proc_open
,
376 .release
= single_release
,
380 static int __net_init
ip_rt_do_proc_init(struct net
*net
)
382 struct proc_dir_entry
*pde
;
384 pde
= proc_create("rt_cache", S_IRUGO
, net
->proc_net
,
389 pde
= proc_create("rt_cache", S_IRUGO
,
390 net
->proc_net_stat
, &rt_cpu_seq_fops
);
394 #ifdef CONFIG_IP_ROUTE_CLASSID
395 pde
= proc_create("rt_acct", 0, net
->proc_net
, &rt_acct_proc_fops
);
401 #ifdef CONFIG_IP_ROUTE_CLASSID
403 remove_proc_entry("rt_cache", net
->proc_net_stat
);
406 remove_proc_entry("rt_cache", net
->proc_net
);
411 static void __net_exit
ip_rt_do_proc_exit(struct net
*net
)
413 remove_proc_entry("rt_cache", net
->proc_net_stat
);
414 remove_proc_entry("rt_cache", net
->proc_net
);
415 #ifdef CONFIG_IP_ROUTE_CLASSID
416 remove_proc_entry("rt_acct", net
->proc_net
);
420 static struct pernet_operations ip_rt_proc_ops __net_initdata
= {
421 .init
= ip_rt_do_proc_init
,
422 .exit
= ip_rt_do_proc_exit
,
425 static int __init
ip_rt_proc_init(void)
427 return register_pernet_subsys(&ip_rt_proc_ops
);
431 static inline int ip_rt_proc_init(void)
435 #endif /* CONFIG_PROC_FS */
437 static inline bool rt_is_expired(const struct rtable
*rth
)
439 return rth
->rt_genid
!= rt_genid(dev_net(rth
->dst
.dev
));
442 void rt_cache_flush(struct net
*net
)
447 static struct neighbour
*ipv4_neigh_lookup(const struct dst_entry
*dst
,
451 struct net_device
*dev
= dst
->dev
;
452 const __be32
*pkey
= daddr
;
453 const struct rtable
*rt
;
456 rt
= (const struct rtable
*) dst
;
458 pkey
= (const __be32
*) &rt
->rt_gateway
;
460 pkey
= &ip_hdr(skb
)->daddr
;
462 n
= __ipv4_neigh_lookup(dev
, *(__force u32
*)pkey
);
465 return neigh_create(&arp_tbl
, pkey
, dev
);
468 #define IP_IDENTS_SZ 2048u
469 struct ip_ident_bucket
{
474 static struct ip_ident_bucket
*ip_idents __read_mostly
;
476 /* In order to protect privacy, we add a perturbation to identifiers
477 * if one generator is seldom used. This makes hard for an attacker
478 * to infer how many packets were sent between two points in time.
480 u32
ip_idents_reserve(u32 hash
, int segs
)
482 struct ip_ident_bucket
*bucket
= ip_idents
+ hash
% IP_IDENTS_SZ
;
483 u32 old
= ACCESS_ONCE(bucket
->stamp32
);
484 u32 now
= (u32
)jiffies
;
487 if (old
!= now
&& cmpxchg(&bucket
->stamp32
, old
, now
) == old
) {
488 u64 x
= prandom_u32();
491 delta
= (u32
)(x
>> 32);
494 return atomic_add_return(segs
+ delta
, &bucket
->id
) - segs
;
496 EXPORT_SYMBOL(ip_idents_reserve
);
498 void __ip_select_ident(struct iphdr
*iph
, int segs
)
500 static u32 ip_idents_hashrnd __read_mostly
;
501 static bool hashrnd_initialized
= false;
504 if (unlikely(!hashrnd_initialized
)) {
505 hashrnd_initialized
= true;
506 get_random_bytes(&ip_idents_hashrnd
, sizeof(ip_idents_hashrnd
));
509 hash
= jhash_3words((__force u32
)iph
->daddr
,
510 (__force u32
)iph
->saddr
,
513 id
= ip_idents_reserve(hash
, segs
);
516 EXPORT_SYMBOL(__ip_select_ident
);
518 static void __build_flow_key(const struct net
*net
, struct flowi4
*fl4
,
519 const struct sock
*sk
,
520 const struct iphdr
*iph
,
522 u8 prot
, u32 mark
, int flow_flags
)
525 const struct inet_sock
*inet
= inet_sk(sk
);
527 oif
= sk
->sk_bound_dev_if
;
529 tos
= RT_CONN_FLAGS(sk
);
530 prot
= inet
->hdrincl
? IPPROTO_RAW
: sk
->sk_protocol
;
532 flowi4_init_output(fl4
, oif
, mark
, tos
,
533 RT_SCOPE_UNIVERSE
, prot
,
535 iph
->daddr
, iph
->saddr
, 0, 0,
536 sock_net_uid(net
, sk
));
539 static void build_skb_flow_key(struct flowi4
*fl4
, const struct sk_buff
*skb
,
540 const struct sock
*sk
)
542 const struct iphdr
*iph
= ip_hdr(skb
);
543 int oif
= skb
->dev
->ifindex
;
544 u8 tos
= RT_TOS(iph
->tos
);
545 u8 prot
= iph
->protocol
;
546 u32 mark
= skb
->mark
;
548 __build_flow_key(sock_net(sk
), fl4
, sk
, iph
, oif
, tos
, prot
, mark
, 0);
551 static void build_sk_flow_key(struct flowi4
*fl4
, const struct sock
*sk
)
553 const struct inet_sock
*inet
= inet_sk(sk
);
554 const struct ip_options_rcu
*inet_opt
;
555 __be32 daddr
= inet
->inet_daddr
;
558 inet_opt
= rcu_dereference(inet
->inet_opt
);
559 if (inet_opt
&& inet_opt
->opt
.srr
)
560 daddr
= inet_opt
->opt
.faddr
;
561 flowi4_init_output(fl4
, sk
->sk_bound_dev_if
, sk
->sk_mark
,
562 RT_CONN_FLAGS(sk
), RT_SCOPE_UNIVERSE
,
563 inet
->hdrincl
? IPPROTO_RAW
: sk
->sk_protocol
,
564 inet_sk_flowi_flags(sk
),
565 daddr
, inet
->inet_saddr
, 0, 0, sk
->sk_uid
);
569 static void ip_rt_build_flow_key(struct flowi4
*fl4
, const struct sock
*sk
,
570 const struct sk_buff
*skb
)
573 build_skb_flow_key(fl4
, skb
, sk
);
575 build_sk_flow_key(fl4
, sk
);
578 static inline void rt_free(struct rtable
*rt
)
580 call_rcu(&rt
->dst
.rcu_head
, dst_rcu_free
);
583 static DEFINE_SPINLOCK(fnhe_lock
);
585 static struct fib_nh_exception
*fnhe_oldest(struct fnhe_hash_bucket
*hash
)
587 struct fib_nh_exception
*fnhe
, *oldest
;
590 oldest
= rcu_dereference(hash
->chain
);
591 for (fnhe
= rcu_dereference(oldest
->fnhe_next
); fnhe
;
592 fnhe
= rcu_dereference(fnhe
->fnhe_next
)) {
593 if (time_before(fnhe
->fnhe_stamp
, oldest
->fnhe_stamp
))
596 orig
= rcu_dereference(oldest
->fnhe_rth
);
598 RCU_INIT_POINTER(oldest
->fnhe_rth
, NULL
);
604 static inline u32
fnhe_hashfun(__be32 daddr
)
608 hval
= (__force u32
) daddr
;
609 hval
^= (hval
>> 11) ^ (hval
>> 22);
611 return hval
& (FNHE_HASH_SIZE
- 1);
614 static void update_or_create_fnhe(struct fib_nh
*nh
, __be32 daddr
, __be32 gw
,
615 u32 pmtu
, unsigned long expires
)
617 struct fnhe_hash_bucket
*hash
;
618 struct fib_nh_exception
*fnhe
;
620 u32 hval
= fnhe_hashfun(daddr
);
622 spin_lock_bh(&fnhe_lock
);
624 hash
= nh
->nh_exceptions
;
626 hash
= kzalloc(FNHE_HASH_SIZE
* sizeof(*hash
), GFP_ATOMIC
);
629 nh
->nh_exceptions
= hash
;
635 for (fnhe
= rcu_dereference(hash
->chain
); fnhe
;
636 fnhe
= rcu_dereference(fnhe
->fnhe_next
)) {
637 if (fnhe
->fnhe_daddr
== daddr
)
646 fnhe
->fnhe_pmtu
= pmtu
;
647 fnhe
->fnhe_expires
= expires
;
650 if (depth
> FNHE_RECLAIM_DEPTH
)
651 fnhe
= fnhe_oldest(hash
);
653 fnhe
= kzalloc(sizeof(*fnhe
), GFP_ATOMIC
);
657 fnhe
->fnhe_next
= hash
->chain
;
658 rcu_assign_pointer(hash
->chain
, fnhe
);
660 fnhe
->fnhe_daddr
= daddr
;
662 fnhe
->fnhe_pmtu
= pmtu
;
663 fnhe
->fnhe_expires
= expires
;
666 fnhe
->fnhe_stamp
= jiffies
;
669 spin_unlock_bh(&fnhe_lock
);
673 static void __ip_do_redirect(struct rtable
*rt
, struct sk_buff
*skb
, struct flowi4
*fl4
,
676 __be32 new_gw
= icmp_hdr(skb
)->un
.gateway
;
677 __be32 old_gw
= ip_hdr(skb
)->saddr
;
678 struct net_device
*dev
= skb
->dev
;
679 struct in_device
*in_dev
;
680 struct fib_result res
;
684 switch (icmp_hdr(skb
)->code
& 7) {
686 case ICMP_REDIR_NETTOS
:
687 case ICMP_REDIR_HOST
:
688 case ICMP_REDIR_HOSTTOS
:
695 if (rt
->rt_gateway
!= old_gw
)
698 in_dev
= __in_dev_get_rcu(dev
);
703 if (new_gw
== old_gw
|| !IN_DEV_RX_REDIRECTS(in_dev
) ||
704 ipv4_is_multicast(new_gw
) || ipv4_is_lbcast(new_gw
) ||
705 ipv4_is_zeronet(new_gw
))
706 goto reject_redirect
;
708 if (!IN_DEV_SHARED_MEDIA(in_dev
)) {
709 if (!inet_addr_onlink(in_dev
, new_gw
, old_gw
))
710 goto reject_redirect
;
711 if (IN_DEV_SEC_REDIRECTS(in_dev
) && ip_fib_check_default(new_gw
, dev
))
712 goto reject_redirect
;
714 if (inet_addr_type(net
, new_gw
) != RTN_UNICAST
)
715 goto reject_redirect
;
718 n
= __ipv4_neigh_lookup(rt
->dst
.dev
, new_gw
);
720 n
= neigh_create(&arp_tbl
, &new_gw
, rt
->dst
.dev
);
722 if (!(n
->nud_state
& NUD_VALID
)) {
723 neigh_event_send(n
, NULL
);
725 if (fib_lookup(net
, fl4
, &res
) == 0) {
726 struct fib_nh
*nh
= &FIB_RES_NH(res
);
728 update_or_create_fnhe(nh
, fl4
->daddr
, new_gw
,
732 rt
->dst
.obsolete
= DST_OBSOLETE_KILL
;
733 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE
, n
);
740 #ifdef CONFIG_IP_ROUTE_VERBOSE
741 if (IN_DEV_LOG_MARTIANS(in_dev
)) {
742 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
743 __be32 daddr
= iph
->daddr
;
744 __be32 saddr
= iph
->saddr
;
746 net_info_ratelimited("Redirect from %pI4 on %s about %pI4 ignored\n"
747 " Advised path = %pI4 -> %pI4\n",
748 &old_gw
, dev
->name
, &new_gw
,
755 static void ip_do_redirect(struct dst_entry
*dst
, struct sock
*sk
, struct sk_buff
*skb
)
759 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
760 int oif
= skb
->dev
->ifindex
;
761 u8 tos
= RT_TOS(iph
->tos
);
762 u8 prot
= iph
->protocol
;
763 u32 mark
= skb
->mark
;
765 rt
= (struct rtable
*) dst
;
767 __build_flow_key(sock_net(sk
), &fl4
, sk
, iph
, oif
, tos
, prot
, mark
, 0);
768 __ip_do_redirect(rt
, skb
, &fl4
, true);
771 static struct dst_entry
*ipv4_negative_advice(struct dst_entry
*dst
)
773 struct rtable
*rt
= (struct rtable
*)dst
;
774 struct dst_entry
*ret
= dst
;
777 if (dst
->obsolete
> 0) {
780 } else if ((rt
->rt_flags
& RTCF_REDIRECTED
) ||
791 * 1. The first ip_rt_redirect_number redirects are sent
792 * with exponential backoff, then we stop sending them at all,
793 * assuming that the host ignores our redirects.
794 * 2. If we did not see packets requiring redirects
795 * during ip_rt_redirect_silence, we assume that the host
796 * forgot redirected route and start to send redirects again.
798 * This algorithm is much cheaper and more intelligent than dumb load limiting
801 * NOTE. Do not forget to inhibit load limiting for redirects (redundant)
802 * and "frag. need" (breaks PMTU discovery) in icmp.c.
805 void ip_rt_send_redirect(struct sk_buff
*skb
)
807 struct rtable
*rt
= skb_rtable(skb
);
808 struct in_device
*in_dev
;
809 struct inet_peer
*peer
;
814 in_dev
= __in_dev_get_rcu(rt
->dst
.dev
);
815 if (!in_dev
|| !IN_DEV_TX_REDIRECTS(in_dev
)) {
819 log_martians
= IN_DEV_LOG_MARTIANS(in_dev
);
822 net
= dev_net(rt
->dst
.dev
);
823 peer
= inet_getpeer_v4(net
->ipv4
.peers
, ip_hdr(skb
)->saddr
, 1);
825 icmp_send(skb
, ICMP_REDIRECT
, ICMP_REDIR_HOST
,
826 rt_nexthop(rt
, ip_hdr(skb
)->daddr
));
830 /* No redirected packets during ip_rt_redirect_silence;
831 * reset the algorithm.
833 if (time_after(jiffies
, peer
->rate_last
+ ip_rt_redirect_silence
))
834 peer
->rate_tokens
= 0;
836 /* Too many ignored redirects; do not send anything
837 * set dst.rate_last to the last seen redirected packet.
839 if (peer
->rate_tokens
>= ip_rt_redirect_number
) {
840 peer
->rate_last
= jiffies
;
844 /* Check for load limit; set rate_last to the latest sent
847 if (peer
->rate_tokens
== 0 ||
850 (ip_rt_redirect_load
<< peer
->rate_tokens
)))) {
851 __be32 gw
= rt_nexthop(rt
, ip_hdr(skb
)->daddr
);
853 icmp_send(skb
, ICMP_REDIRECT
, ICMP_REDIR_HOST
, gw
);
854 peer
->rate_last
= jiffies
;
856 #ifdef CONFIG_IP_ROUTE_VERBOSE
858 peer
->rate_tokens
== ip_rt_redirect_number
)
859 net_warn_ratelimited("host %pI4/if%d ignores redirects for %pI4 to %pI4\n",
860 &ip_hdr(skb
)->saddr
, inet_iif(skb
),
861 &ip_hdr(skb
)->daddr
, &gw
);
868 static int ip_error(struct sk_buff
*skb
)
870 struct in_device
*in_dev
= __in_dev_get_rcu(skb
->dev
);
871 struct rtable
*rt
= skb_rtable(skb
);
872 struct inet_peer
*peer
;
878 /* IP on this device is disabled. */
882 net
= dev_net(rt
->dst
.dev
);
883 if (!IN_DEV_FORWARD(in_dev
)) {
884 switch (rt
->dst
.error
) {
886 IP_INC_STATS_BH(net
, IPSTATS_MIB_INADDRERRORS
);
890 IP_INC_STATS_BH(net
, IPSTATS_MIB_INNOROUTES
);
896 switch (rt
->dst
.error
) {
901 code
= ICMP_HOST_UNREACH
;
904 code
= ICMP_NET_UNREACH
;
905 IP_INC_STATS_BH(net
, IPSTATS_MIB_INNOROUTES
);
908 code
= ICMP_PKT_FILTERED
;
912 peer
= inet_getpeer_v4(net
->ipv4
.peers
, ip_hdr(skb
)->saddr
, 1);
917 peer
->rate_tokens
+= now
- peer
->rate_last
;
918 if (peer
->rate_tokens
> ip_rt_error_burst
)
919 peer
->rate_tokens
= ip_rt_error_burst
;
920 peer
->rate_last
= now
;
921 if (peer
->rate_tokens
>= ip_rt_error_cost
)
922 peer
->rate_tokens
-= ip_rt_error_cost
;
928 icmp_send(skb
, ICMP_DEST_UNREACH
, code
, 0);
934 static void __ip_rt_update_pmtu(struct rtable
*rt
, struct flowi4
*fl4
, u32 mtu
)
936 struct dst_entry
*dst
= &rt
->dst
;
937 struct fib_result res
;
939 if (dst_metric_locked(dst
, RTAX_MTU
))
942 if (dst
->dev
->mtu
< mtu
)
945 if (mtu
< ip_rt_min_pmtu
)
946 mtu
= ip_rt_min_pmtu
;
949 dst
->obsolete
= DST_OBSOLETE_KILL
;
952 dst
->expires
= max(1UL, jiffies
+ ip_rt_mtu_expires
);
956 if (fib_lookup(dev_net(dst
->dev
), fl4
, &res
) == 0) {
957 struct fib_nh
*nh
= &FIB_RES_NH(res
);
959 update_or_create_fnhe(nh
, fl4
->daddr
, 0, mtu
,
960 jiffies
+ ip_rt_mtu_expires
);
965 static void ip_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
966 struct sk_buff
*skb
, u32 mtu
)
968 struct rtable
*rt
= (struct rtable
*) dst
;
971 ip_rt_build_flow_key(&fl4
, sk
, skb
);
972 __ip_rt_update_pmtu(rt
, &fl4
, mtu
);
975 void ipv4_update_pmtu(struct sk_buff
*skb
, struct net
*net
, u32 mtu
,
976 int oif
, u32 mark
, u8 protocol
, int flow_flags
)
978 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
983 mark
= IP4_REPLY_MARK(net
, skb
->mark
);
985 __build_flow_key(net
, &fl4
, NULL
, iph
, oif
,
986 RT_TOS(iph
->tos
), protocol
, mark
, flow_flags
);
987 rt
= __ip_route_output_key(net
, &fl4
);
989 __ip_rt_update_pmtu(rt
, &fl4
, mtu
);
993 EXPORT_SYMBOL_GPL(ipv4_update_pmtu
);
995 static void __ipv4_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, u32 mtu
)
997 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1001 __build_flow_key(sock_net(sk
), &fl4
, sk
, iph
, 0, 0, 0, 0, 0);
1003 if (!fl4
.flowi4_mark
)
1004 fl4
.flowi4_mark
= IP4_REPLY_MARK(sock_net(sk
), skb
->mark
);
1006 rt
= __ip_route_output_key(sock_net(sk
), &fl4
);
1008 __ip_rt_update_pmtu(rt
, &fl4
, mtu
);
1013 void ipv4_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, u32 mtu
)
1015 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1018 struct dst_entry
*odst
= NULL
;
1020 struct net
*net
= sock_net(sk
);
1023 odst
= sk_dst_get(sk
);
1025 if (sock_owned_by_user(sk
) || !odst
) {
1026 __ipv4_sk_update_pmtu(skb
, sk
, mtu
);
1030 __build_flow_key(net
, &fl4
, sk
, iph
, 0, 0, 0, 0, 0);
1032 rt
= (struct rtable
*)odst
;
1033 if (odst
->obsolete
&& odst
->ops
->check(odst
, 0) == NULL
) {
1034 rt
= ip_route_output_flow(sock_net(sk
), &fl4
, sk
);
1041 __ip_rt_update_pmtu((struct rtable
*) rt
->dst
.path
, &fl4
, mtu
);
1043 if (!dst_check(&rt
->dst
, 0)) {
1045 dst_release(&rt
->dst
);
1047 rt
= ip_route_output_flow(sock_net(sk
), &fl4
, sk
);
1055 sk_dst_set(sk
, &rt
->dst
);
1061 EXPORT_SYMBOL_GPL(ipv4_sk_update_pmtu
);
1063 void ipv4_redirect(struct sk_buff
*skb
, struct net
*net
,
1064 int oif
, u32 mark
, u8 protocol
, int flow_flags
)
1066 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1070 __build_flow_key(net
, &fl4
, NULL
, iph
, oif
,
1071 RT_TOS(iph
->tos
), protocol
, mark
, flow_flags
);
1072 rt
= __ip_route_output_key(net
, &fl4
);
1074 __ip_do_redirect(rt
, skb
, &fl4
, false);
1078 EXPORT_SYMBOL_GPL(ipv4_redirect
);
1080 void ipv4_sk_redirect(struct sk_buff
*skb
, struct sock
*sk
)
1082 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1085 struct net
*net
= sock_net(sk
);
1087 __build_flow_key(net
, &fl4
, sk
, iph
, 0, 0, 0, 0, 0);
1088 rt
= __ip_route_output_key(net
, &fl4
);
1090 __ip_do_redirect(rt
, skb
, &fl4
, false);
1094 EXPORT_SYMBOL_GPL(ipv4_sk_redirect
);
1096 static struct dst_entry
*ipv4_dst_check(struct dst_entry
*dst
, u32 cookie
)
1098 struct rtable
*rt
= (struct rtable
*) dst
;
1100 /* All IPV4 dsts are created with ->obsolete set to the value
1101 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1102 * into this function always.
1104 * When a PMTU/redirect information update invalidates a
1105 * route, this is indicated by setting obsolete to
1106 * DST_OBSOLETE_KILL.
1108 if (dst
->obsolete
== DST_OBSOLETE_KILL
|| rt_is_expired(rt
))
1113 static void ipv4_link_failure(struct sk_buff
*skb
)
1117 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_HOST_UNREACH
, 0);
1119 rt
= skb_rtable(skb
);
1121 dst_set_expires(&rt
->dst
, 0);
1124 static int ip_rt_bug(struct sk_buff
*skb
)
1126 pr_debug("%s: %pI4 -> %pI4, %s\n",
1127 __func__
, &ip_hdr(skb
)->saddr
, &ip_hdr(skb
)->daddr
,
1128 skb
->dev
? skb
->dev
->name
: "?");
1135 We do not cache source address of outgoing interface,
1136 because it is used only by IP RR, TS and SRR options,
1137 so that it out of fast path.
1139 BTW remember: "addr" is allowed to be not aligned
1143 void ip_rt_get_source(u8
*addr
, struct sk_buff
*skb
, struct rtable
*rt
)
1147 if (rt_is_output_route(rt
))
1148 src
= ip_hdr(skb
)->saddr
;
1150 struct fib_result res
;
1156 memset(&fl4
, 0, sizeof(fl4
));
1157 fl4
.daddr
= iph
->daddr
;
1158 fl4
.saddr
= iph
->saddr
;
1159 fl4
.flowi4_tos
= RT_TOS(iph
->tos
);
1160 fl4
.flowi4_oif
= rt
->dst
.dev
->ifindex
;
1161 fl4
.flowi4_iif
= skb
->dev
->ifindex
;
1162 fl4
.flowi4_mark
= skb
->mark
;
1165 if (fib_lookup(dev_net(rt
->dst
.dev
), &fl4
, &res
) == 0)
1166 src
= FIB_RES_PREFSRC(dev_net(rt
->dst
.dev
), res
);
1168 src
= inet_select_addr(rt
->dst
.dev
,
1169 rt_nexthop(rt
, iph
->daddr
),
1173 memcpy(addr
, &src
, 4);
1176 #ifdef CONFIG_IP_ROUTE_CLASSID
1177 static void set_class_tag(struct rtable
*rt
, u32 tag
)
1179 if (!(rt
->dst
.tclassid
& 0xFFFF))
1180 rt
->dst
.tclassid
|= tag
& 0xFFFF;
1181 if (!(rt
->dst
.tclassid
& 0xFFFF0000))
1182 rt
->dst
.tclassid
|= tag
& 0xFFFF0000;
1186 static unsigned int ipv4_default_advmss(const struct dst_entry
*dst
)
1188 unsigned int advmss
= dst_metric_raw(dst
, RTAX_ADVMSS
);
1191 advmss
= max_t(unsigned int, dst
->dev
->mtu
- 40,
1193 if (advmss
> 65535 - 40)
1194 advmss
= 65535 - 40;
1199 static unsigned int ipv4_mtu(const struct dst_entry
*dst
)
1201 const struct rtable
*rt
= (const struct rtable
*) dst
;
1202 unsigned int mtu
= rt
->rt_pmtu
;
1204 if (!mtu
|| time_after_eq(jiffies
, rt
->dst
.expires
))
1205 mtu
= dst_metric_raw(dst
, RTAX_MTU
);
1210 mtu
= dst
->dev
->mtu
;
1212 if (unlikely(dst_metric_locked(dst
, RTAX_MTU
))) {
1213 if (rt
->rt_uses_gateway
&& mtu
> 576)
1217 if (mtu
> IP_MAX_MTU
)
1223 static struct fib_nh_exception
*find_exception(struct fib_nh
*nh
, __be32 daddr
)
1225 struct fnhe_hash_bucket
*hash
= nh
->nh_exceptions
;
1226 struct fib_nh_exception
*fnhe
;
1232 hval
= fnhe_hashfun(daddr
);
1234 for (fnhe
= rcu_dereference(hash
[hval
].chain
); fnhe
;
1235 fnhe
= rcu_dereference(fnhe
->fnhe_next
)) {
1236 if (fnhe
->fnhe_daddr
== daddr
)
1242 static bool rt_bind_exception(struct rtable
*rt
, struct fib_nh_exception
*fnhe
,
1247 spin_lock_bh(&fnhe_lock
);
1249 if (daddr
== fnhe
->fnhe_daddr
) {
1250 struct rtable
*orig
= rcu_dereference(fnhe
->fnhe_rth
);
1251 if (orig
&& rt_is_expired(orig
)) {
1253 fnhe
->fnhe_pmtu
= 0;
1254 fnhe
->fnhe_expires
= 0;
1256 if (fnhe
->fnhe_pmtu
) {
1257 unsigned long expires
= fnhe
->fnhe_expires
;
1258 unsigned long diff
= expires
- jiffies
;
1260 if (time_before(jiffies
, expires
)) {
1261 rt
->rt_pmtu
= fnhe
->fnhe_pmtu
;
1262 dst_set_expires(&rt
->dst
, diff
);
1265 if (fnhe
->fnhe_gw
) {
1266 rt
->rt_flags
|= RTCF_REDIRECTED
;
1267 rt
->rt_gateway
= fnhe
->fnhe_gw
;
1268 rt
->rt_uses_gateway
= 1;
1269 } else if (!rt
->rt_gateway
)
1270 rt
->rt_gateway
= daddr
;
1272 rcu_assign_pointer(fnhe
->fnhe_rth
, rt
);
1276 fnhe
->fnhe_stamp
= jiffies
;
1279 spin_unlock_bh(&fnhe_lock
);
1284 static bool rt_cache_route(struct fib_nh
*nh
, struct rtable
*rt
)
1286 struct rtable
*orig
, *prev
, **p
;
1289 if (rt_is_input_route(rt
)) {
1290 p
= (struct rtable
**)&nh
->nh_rth_input
;
1292 p
= (struct rtable
**)__this_cpu_ptr(nh
->nh_pcpu_rth_output
);
1296 prev
= cmpxchg(p
, orig
, rt
);
1306 static DEFINE_SPINLOCK(rt_uncached_lock
);
1307 static LIST_HEAD(rt_uncached_list
);
1309 static void rt_add_uncached_list(struct rtable
*rt
)
1311 spin_lock_bh(&rt_uncached_lock
);
1312 list_add_tail(&rt
->rt_uncached
, &rt_uncached_list
);
1313 spin_unlock_bh(&rt_uncached_lock
);
1316 static void ipv4_dst_destroy(struct dst_entry
*dst
)
1318 struct rtable
*rt
= (struct rtable
*) dst
;
1320 if (!list_empty(&rt
->rt_uncached
)) {
1321 spin_lock_bh(&rt_uncached_lock
);
1322 list_del(&rt
->rt_uncached
);
1323 spin_unlock_bh(&rt_uncached_lock
);
1327 void rt_flush_dev(struct net_device
*dev
)
1329 if (!list_empty(&rt_uncached_list
)) {
1330 struct net
*net
= dev_net(dev
);
1333 spin_lock_bh(&rt_uncached_lock
);
1334 list_for_each_entry(rt
, &rt_uncached_list
, rt_uncached
) {
1335 if (rt
->dst
.dev
!= dev
)
1337 rt
->dst
.dev
= net
->loopback_dev
;
1338 dev_hold(rt
->dst
.dev
);
1341 spin_unlock_bh(&rt_uncached_lock
);
1345 static bool rt_cache_valid(const struct rtable
*rt
)
1348 rt
->dst
.obsolete
== DST_OBSOLETE_FORCE_CHK
&&
1352 static void rt_set_nexthop(struct rtable
*rt
, __be32 daddr
,
1353 const struct fib_result
*res
,
1354 struct fib_nh_exception
*fnhe
,
1355 struct fib_info
*fi
, u16 type
, u32 itag
)
1357 bool cached
= false;
1360 struct fib_nh
*nh
= &FIB_RES_NH(*res
);
1362 if (nh
->nh_gw
&& nh
->nh_scope
== RT_SCOPE_LINK
) {
1363 rt
->rt_gateway
= nh
->nh_gw
;
1364 rt
->rt_uses_gateway
= 1;
1366 dst_init_metrics(&rt
->dst
, fi
->fib_metrics
, true);
1367 #ifdef CONFIG_IP_ROUTE_CLASSID
1368 rt
->dst
.tclassid
= nh
->nh_tclassid
;
1371 cached
= rt_bind_exception(rt
, fnhe
, daddr
);
1372 else if (!(rt
->dst
.flags
& DST_NOCACHE
))
1373 cached
= rt_cache_route(nh
, rt
);
1374 if (unlikely(!cached
)) {
1375 /* Routes we intend to cache in nexthop exception or
1376 * FIB nexthop have the DST_NOCACHE bit clear.
1377 * However, if we are unsuccessful at storing this
1378 * route into the cache we really need to set it.
1380 rt
->dst
.flags
|= DST_NOCACHE
;
1381 if (!rt
->rt_gateway
)
1382 rt
->rt_gateway
= daddr
;
1383 rt_add_uncached_list(rt
);
1386 rt_add_uncached_list(rt
);
1388 #ifdef CONFIG_IP_ROUTE_CLASSID
1389 #ifdef CONFIG_IP_MULTIPLE_TABLES
1390 set_class_tag(rt
, res
->tclassid
);
1392 set_class_tag(rt
, itag
);
1396 static struct rtable
*rt_dst_alloc(struct net_device
*dev
,
1397 bool nopolicy
, bool noxfrm
, bool will_cache
)
1399 return dst_alloc(&ipv4_dst_ops
, dev
, 1, DST_OBSOLETE_FORCE_CHK
,
1400 (will_cache
? 0 : (DST_HOST
| DST_NOCACHE
)) |
1401 (nopolicy
? DST_NOPOLICY
: 0) |
1402 (noxfrm
? DST_NOXFRM
: 0));
1405 /* called in rcu_read_lock() section */
1406 static int ip_route_input_mc(struct sk_buff
*skb
, __be32 daddr
, __be32 saddr
,
1407 u8 tos
, struct net_device
*dev
, int our
)
1410 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
1414 /* Primary sanity checks. */
1419 if (ipv4_is_multicast(saddr
) || ipv4_is_lbcast(saddr
) ||
1420 skb
->protocol
!= htons(ETH_P_IP
))
1423 if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev
)))
1424 if (ipv4_is_loopback(saddr
))
1427 if (ipv4_is_zeronet(saddr
)) {
1428 if (!ipv4_is_local_multicast(daddr
))
1431 err
= fib_validate_source(skb
, saddr
, 0, tos
, 0, dev
,
1436 rth
= rt_dst_alloc(dev_net(dev
)->loopback_dev
,
1437 IN_DEV_CONF_GET(in_dev
, NOPOLICY
), false, false);
1441 #ifdef CONFIG_IP_ROUTE_CLASSID
1442 rth
->dst
.tclassid
= itag
;
1444 rth
->dst
.output
= ip_rt_bug
;
1446 rth
->rt_genid
= rt_genid(dev_net(dev
));
1447 rth
->rt_flags
= RTCF_MULTICAST
;
1448 rth
->rt_type
= RTN_MULTICAST
;
1449 rth
->rt_is_input
= 1;
1452 rth
->rt_gateway
= 0;
1453 rth
->rt_uses_gateway
= 0;
1454 INIT_LIST_HEAD(&rth
->rt_uncached
);
1456 rth
->dst
.input
= ip_local_deliver
;
1457 rth
->rt_flags
|= RTCF_LOCAL
;
1460 #ifdef CONFIG_IP_MROUTE
1461 if (!ipv4_is_local_multicast(daddr
) && IN_DEV_MFORWARD(in_dev
))
1462 rth
->dst
.input
= ip_mr_input
;
1464 RT_CACHE_STAT_INC(in_slow_mc
);
1466 skb_dst_set(skb
, &rth
->dst
);
1478 static void ip_handle_martian_source(struct net_device
*dev
,
1479 struct in_device
*in_dev
,
1480 struct sk_buff
*skb
,
1484 RT_CACHE_STAT_INC(in_martian_src
);
1485 #ifdef CONFIG_IP_ROUTE_VERBOSE
1486 if (IN_DEV_LOG_MARTIANS(in_dev
) && net_ratelimit()) {
1488 * RFC1812 recommendation, if source is martian,
1489 * the only hint is MAC header.
1491 pr_warn("martian source %pI4 from %pI4, on dev %s\n",
1492 &daddr
, &saddr
, dev
->name
);
1493 if (dev
->hard_header_len
&& skb_mac_header_was_set(skb
)) {
1494 print_hex_dump(KERN_WARNING
, "ll header: ",
1495 DUMP_PREFIX_OFFSET
, 16, 1,
1496 skb_mac_header(skb
),
1497 dev
->hard_header_len
, true);
1503 /* called in rcu_read_lock() section */
1504 static int __mkroute_input(struct sk_buff
*skb
,
1505 const struct fib_result
*res
,
1506 struct in_device
*in_dev
,
1507 __be32 daddr
, __be32 saddr
, u32 tos
)
1511 struct in_device
*out_dev
;
1512 unsigned int flags
= 0;
1516 /* get a working reference to the output device */
1517 out_dev
= __in_dev_get_rcu(FIB_RES_DEV(*res
));
1518 if (out_dev
== NULL
) {
1519 net_crit_ratelimited("Bug in ip_route_input_slow(). Please report.\n");
1523 err
= fib_validate_source(skb
, saddr
, daddr
, tos
, FIB_RES_OIF(*res
),
1524 in_dev
->dev
, in_dev
, &itag
);
1526 ip_handle_martian_source(in_dev
->dev
, in_dev
, skb
, daddr
,
1532 do_cache
= res
->fi
&& !itag
;
1533 if (out_dev
== in_dev
&& err
&& IN_DEV_TX_REDIRECTS(out_dev
) &&
1534 skb
->protocol
== htons(ETH_P_IP
) &&
1535 (IN_DEV_SHARED_MEDIA(out_dev
) ||
1536 inet_addr_onlink(out_dev
, saddr
, FIB_RES_GW(*res
))))
1537 IPCB(skb
)->flags
|= IPSKB_DOREDIRECT
;
1539 if (skb
->protocol
!= htons(ETH_P_IP
)) {
1540 /* Not IP (i.e. ARP). Do not create route, if it is
1541 * invalid for proxy arp. DNAT routes are always valid.
1543 * Proxy arp feature have been extended to allow, ARP
1544 * replies back to the same interface, to support
1545 * Private VLAN switch technologies. See arp.c.
1547 if (out_dev
== in_dev
&&
1548 IN_DEV_PROXY_ARP_PVLAN(in_dev
) == 0) {
1555 rth
= rcu_dereference(FIB_RES_NH(*res
).nh_rth_input
);
1556 if (rt_cache_valid(rth
)) {
1557 skb_dst_set_noref(skb
, &rth
->dst
);
1562 rth
= rt_dst_alloc(out_dev
->dev
,
1563 IN_DEV_CONF_GET(in_dev
, NOPOLICY
),
1564 IN_DEV_CONF_GET(out_dev
, NOXFRM
), do_cache
);
1570 rth
->rt_genid
= rt_genid(dev_net(rth
->dst
.dev
));
1571 rth
->rt_flags
= flags
;
1572 rth
->rt_type
= res
->type
;
1573 rth
->rt_is_input
= 1;
1576 rth
->rt_gateway
= 0;
1577 rth
->rt_uses_gateway
= 0;
1578 INIT_LIST_HEAD(&rth
->rt_uncached
);
1579 RT_CACHE_STAT_INC(in_slow_tot
);
1581 rth
->dst
.input
= ip_forward
;
1582 rth
->dst
.output
= ip_output
;
1584 rt_set_nexthop(rth
, daddr
, res
, NULL
, res
->fi
, res
->type
, itag
);
1585 skb_dst_set(skb
, &rth
->dst
);
1592 static int ip_mkroute_input(struct sk_buff
*skb
,
1593 struct fib_result
*res
,
1594 const struct flowi4
*fl4
,
1595 struct in_device
*in_dev
,
1596 __be32 daddr
, __be32 saddr
, u32 tos
)
1598 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1599 if (res
->fi
&& res
->fi
->fib_nhs
> 1)
1600 fib_select_multipath(res
);
1603 /* create a routing cache entry */
1604 return __mkroute_input(skb
, res
, in_dev
, daddr
, saddr
, tos
);
1608 * NOTE. We drop all the packets that has local source
1609 * addresses, because every properly looped back packet
1610 * must have correct destination already attached by output routine.
1612 * Such approach solves two big problems:
1613 * 1. Not simplex devices are handled properly.
1614 * 2. IP spoofing attempts are filtered with 100% of guarantee.
1615 * called with rcu_read_lock()
1618 static int ip_route_input_slow(struct sk_buff
*skb
, __be32 daddr
, __be32 saddr
,
1619 u8 tos
, struct net_device
*dev
)
1621 struct fib_result res
;
1622 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
1624 unsigned int flags
= 0;
1628 struct net
*net
= dev_net(dev
);
1631 /* IP on this device is disabled. */
1636 /* Check for the most weird martians, which can be not detected
1640 if (ipv4_is_multicast(saddr
) || ipv4_is_lbcast(saddr
))
1641 goto martian_source
;
1644 if (ipv4_is_lbcast(daddr
) || (saddr
== 0 && daddr
== 0))
1647 /* Accept zero addresses only to limited broadcast;
1648 * I even do not know to fix it or not. Waiting for complains :-)
1650 if (ipv4_is_zeronet(saddr
))
1651 goto martian_source
;
1653 if (ipv4_is_zeronet(daddr
))
1654 goto martian_destination
;
1656 /* Following code try to avoid calling IN_DEV_NET_ROUTE_LOCALNET(),
1657 * and call it once if daddr or/and saddr are loopback addresses
1659 if (ipv4_is_loopback(daddr
)) {
1660 if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev
, net
))
1661 goto martian_destination
;
1662 } else if (ipv4_is_loopback(saddr
)) {
1663 if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev
, net
))
1664 goto martian_source
;
1668 * Now we are ready to route packet.
1671 fl4
.flowi4_iif
= dev
->ifindex
;
1672 fl4
.flowi4_mark
= skb
->mark
;
1673 fl4
.flowi4_tos
= tos
;
1674 fl4
.flowi4_scope
= RT_SCOPE_UNIVERSE
;
1677 err
= fib_lookup(net
, &fl4
, &res
);
1681 if (res
.type
== RTN_BROADCAST
)
1684 if (res
.type
== RTN_LOCAL
) {
1685 err
= fib_validate_source(skb
, saddr
, daddr
, tos
,
1687 dev
, in_dev
, &itag
);
1689 goto martian_source_keep_err
;
1693 if (!IN_DEV_FORWARD(in_dev
))
1695 if (res
.type
!= RTN_UNICAST
)
1696 goto martian_destination
;
1698 err
= ip_mkroute_input(skb
, &res
, &fl4
, in_dev
, daddr
, saddr
, tos
);
1702 if (skb
->protocol
!= htons(ETH_P_IP
))
1705 if (!ipv4_is_zeronet(saddr
)) {
1706 err
= fib_validate_source(skb
, saddr
, 0, tos
, 0, dev
,
1709 goto martian_source_keep_err
;
1711 flags
|= RTCF_BROADCAST
;
1712 res
.type
= RTN_BROADCAST
;
1713 RT_CACHE_STAT_INC(in_brd
);
1719 rth
= rcu_dereference(FIB_RES_NH(res
).nh_rth_input
);
1720 if (rt_cache_valid(rth
)) {
1721 skb_dst_set_noref(skb
, &rth
->dst
);
1729 rth
= rt_dst_alloc(net
->loopback_dev
,
1730 IN_DEV_CONF_GET(in_dev
, NOPOLICY
), false, do_cache
);
1734 rth
->dst
.input
= ip_local_deliver
;
1735 rth
->dst
.output
= ip_rt_bug
;
1736 #ifdef CONFIG_IP_ROUTE_CLASSID
1737 rth
->dst
.tclassid
= itag
;
1740 rth
->rt_genid
= rt_genid(net
);
1741 rth
->rt_flags
= flags
|RTCF_LOCAL
;
1742 rth
->rt_type
= res
.type
;
1743 rth
->rt_is_input
= 1;
1746 rth
->rt_gateway
= 0;
1747 rth
->rt_uses_gateway
= 0;
1748 INIT_LIST_HEAD(&rth
->rt_uncached
);
1749 RT_CACHE_STAT_INC(in_slow_tot
);
1750 if (res
.type
== RTN_UNREACHABLE
) {
1751 rth
->dst
.input
= ip_error
;
1752 rth
->dst
.error
= -err
;
1753 rth
->rt_flags
&= ~RTCF_LOCAL
;
1756 if (unlikely(!rt_cache_route(&FIB_RES_NH(res
), rth
))) {
1757 rth
->dst
.flags
|= DST_NOCACHE
;
1758 rt_add_uncached_list(rth
);
1761 skb_dst_set(skb
, &rth
->dst
);
1766 RT_CACHE_STAT_INC(in_no_route
);
1767 res
.type
= RTN_UNREACHABLE
;
1773 * Do not cache martian addresses: they should be logged (RFC1812)
1775 martian_destination
:
1776 RT_CACHE_STAT_INC(in_martian_dst
);
1777 #ifdef CONFIG_IP_ROUTE_VERBOSE
1778 if (IN_DEV_LOG_MARTIANS(in_dev
))
1779 net_warn_ratelimited("martian destination %pI4 from %pI4, dev %s\n",
1780 &daddr
, &saddr
, dev
->name
);
1793 martian_source_keep_err
:
1794 ip_handle_martian_source(dev
, in_dev
, skb
, daddr
, saddr
);
1798 int ip_route_input_noref(struct sk_buff
*skb
, __be32 daddr
, __be32 saddr
,
1799 u8 tos
, struct net_device
*dev
)
1803 tos
&= IPTOS_RT_MASK
;
1806 /* Multicast recognition logic is moved from route cache to here.
1807 The problem was that too many Ethernet cards have broken/missing
1808 hardware multicast filters :-( As result the host on multicasting
1809 network acquires a lot of useless route cache entries, sort of
1810 SDR messages from all the world. Now we try to get rid of them.
1811 Really, provided software IP multicast filter is organized
1812 reasonably (at least, hashed), it does not result in a slowdown
1813 comparing with route cache reject entries.
1814 Note, that multicast routers are not affected, because
1815 route cache entry is created eventually.
1817 if (ipv4_is_multicast(daddr
)) {
1818 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
1821 int our
= ip_check_mc_rcu(in_dev
, daddr
, saddr
,
1822 ip_hdr(skb
)->protocol
);
1824 #ifdef CONFIG_IP_MROUTE
1826 (!ipv4_is_local_multicast(daddr
) &&
1827 IN_DEV_MFORWARD(in_dev
))
1830 int res
= ip_route_input_mc(skb
, daddr
, saddr
,
1839 res
= ip_route_input_slow(skb
, daddr
, saddr
, tos
, dev
);
1843 EXPORT_SYMBOL(ip_route_input_noref
);
1845 /* called with rcu_read_lock() */
1846 static struct rtable
*__mkroute_output(const struct fib_result
*res
,
1847 const struct flowi4
*fl4
, int orig_oif
,
1848 struct net_device
*dev_out
,
1851 struct fib_info
*fi
= res
->fi
;
1852 struct fib_nh_exception
*fnhe
;
1853 struct in_device
*in_dev
;
1854 u16 type
= res
->type
;
1858 in_dev
= __in_dev_get_rcu(dev_out
);
1860 return ERR_PTR(-EINVAL
);
1862 if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev
)))
1863 if (ipv4_is_loopback(fl4
->saddr
) && !(dev_out
->flags
& IFF_LOOPBACK
))
1864 return ERR_PTR(-EINVAL
);
1866 if (ipv4_is_lbcast(fl4
->daddr
))
1867 type
= RTN_BROADCAST
;
1868 else if (ipv4_is_multicast(fl4
->daddr
))
1869 type
= RTN_MULTICAST
;
1870 else if (ipv4_is_zeronet(fl4
->daddr
))
1871 return ERR_PTR(-EINVAL
);
1873 if (dev_out
->flags
& IFF_LOOPBACK
)
1874 flags
|= RTCF_LOCAL
;
1877 if (type
== RTN_BROADCAST
) {
1878 flags
|= RTCF_BROADCAST
| RTCF_LOCAL
;
1880 } else if (type
== RTN_MULTICAST
) {
1881 flags
|= RTCF_MULTICAST
| RTCF_LOCAL
;
1882 if (!ip_check_mc_rcu(in_dev
, fl4
->daddr
, fl4
->saddr
,
1884 flags
&= ~RTCF_LOCAL
;
1887 /* If multicast route do not exist use
1888 * default one, but do not gateway in this case.
1891 if (fi
&& res
->prefixlen
< 4)
1893 } else if ((type
== RTN_LOCAL
) && (orig_oif
!= 0) &&
1894 (orig_oif
!= dev_out
->ifindex
)) {
1895 /* For local routes that require a particular output interface
1896 * we do not want to cache the result. Caching the result
1897 * causes incorrect behaviour when there are multiple source
1898 * addresses on the interface, the end result being that if the
1899 * intended recipient is waiting on that interface for the
1900 * packet he won't receive it because it will be delivered on
1901 * the loopback interface and the IP_PKTINFO ipi_ifindex will
1902 * be set to the loopback interface as well.
1908 do_cache
&= fi
!= NULL
;
1910 struct rtable __rcu
**prth
;
1911 struct fib_nh
*nh
= &FIB_RES_NH(*res
);
1913 fnhe
= find_exception(nh
, fl4
->daddr
);
1915 prth
= &fnhe
->fnhe_rth
;
1917 if (unlikely(fl4
->flowi4_flags
&
1918 FLOWI_FLAG_KNOWN_NH
&&
1920 nh
->nh_scope
== RT_SCOPE_LINK
))) {
1924 prth
= __this_cpu_ptr(nh
->nh_pcpu_rth_output
);
1926 rth
= rcu_dereference(*prth
);
1927 if (rt_cache_valid(rth
)) {
1928 dst_hold(&rth
->dst
);
1934 rth
= rt_dst_alloc(dev_out
,
1935 IN_DEV_CONF_GET(in_dev
, NOPOLICY
),
1936 IN_DEV_CONF_GET(in_dev
, NOXFRM
),
1939 return ERR_PTR(-ENOBUFS
);
1941 rth
->dst
.output
= ip_output
;
1943 rth
->rt_genid
= rt_genid(dev_net(dev_out
));
1944 rth
->rt_flags
= flags
;
1945 rth
->rt_type
= type
;
1946 rth
->rt_is_input
= 0;
1947 rth
->rt_iif
= orig_oif
? : 0;
1949 rth
->rt_gateway
= 0;
1950 rth
->rt_uses_gateway
= 0;
1951 INIT_LIST_HEAD(&rth
->rt_uncached
);
1953 RT_CACHE_STAT_INC(out_slow_tot
);
1955 if (flags
& RTCF_LOCAL
)
1956 rth
->dst
.input
= ip_local_deliver
;
1957 if (flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
)) {
1958 if (flags
& RTCF_LOCAL
&&
1959 !(dev_out
->flags
& IFF_LOOPBACK
)) {
1960 rth
->dst
.output
= ip_mc_output
;
1961 RT_CACHE_STAT_INC(out_slow_mc
);
1963 #ifdef CONFIG_IP_MROUTE
1964 if (type
== RTN_MULTICAST
) {
1965 if (IN_DEV_MFORWARD(in_dev
) &&
1966 !ipv4_is_local_multicast(fl4
->daddr
)) {
1967 rth
->dst
.input
= ip_mr_input
;
1968 rth
->dst
.output
= ip_mc_output
;
1974 rt_set_nexthop(rth
, fl4
->daddr
, res
, fnhe
, fi
, type
, 0);
1980 * Major route resolver routine.
1983 struct rtable
*__ip_route_output_key(struct net
*net
, struct flowi4
*fl4
)
1985 struct net_device
*dev_out
= NULL
;
1986 __u8 tos
= RT_FL_TOS(fl4
);
1987 unsigned int flags
= 0;
1988 struct fib_result res
;
1996 orig_oif
= fl4
->flowi4_oif
;
1998 fl4
->flowi4_iif
= LOOPBACK_IFINDEX
;
1999 fl4
->flowi4_tos
= tos
& IPTOS_RT_MASK
;
2000 fl4
->flowi4_scope
= ((tos
& RTO_ONLINK
) ?
2001 RT_SCOPE_LINK
: RT_SCOPE_UNIVERSE
);
2005 rth
= ERR_PTR(-EINVAL
);
2006 if (ipv4_is_multicast(fl4
->saddr
) ||
2007 ipv4_is_lbcast(fl4
->saddr
) ||
2008 ipv4_is_zeronet(fl4
->saddr
))
2011 /* I removed check for oif == dev_out->oif here.
2012 It was wrong for two reasons:
2013 1. ip_dev_find(net, saddr) can return wrong iface, if saddr
2014 is assigned to multiple interfaces.
2015 2. Moreover, we are allowed to send packets with saddr
2016 of another iface. --ANK
2019 if (fl4
->flowi4_oif
== 0 &&
2020 (ipv4_is_multicast(fl4
->daddr
) ||
2021 ipv4_is_lbcast(fl4
->daddr
))) {
2022 /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2023 dev_out
= __ip_dev_find(net
, fl4
->saddr
, false);
2024 if (dev_out
== NULL
)
2027 /* Special hack: user can direct multicasts
2028 and limited broadcast via necessary interface
2029 without fiddling with IP_MULTICAST_IF or IP_PKTINFO.
2030 This hack is not just for fun, it allows
2031 vic,vat and friends to work.
2032 They bind socket to loopback, set ttl to zero
2033 and expect that it will work.
2034 From the viewpoint of routing cache they are broken,
2035 because we are not allowed to build multicast path
2036 with loopback source addr (look, routing cache
2037 cannot know, that ttl is zero, so that packet
2038 will not leave this host and route is valid).
2039 Luckily, this hack is good workaround.
2042 fl4
->flowi4_oif
= dev_out
->ifindex
;
2046 if (!(fl4
->flowi4_flags
& FLOWI_FLAG_ANYSRC
)) {
2047 /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2048 if (!__ip_dev_find(net
, fl4
->saddr
, false))
2054 if (fl4
->flowi4_oif
) {
2055 dev_out
= dev_get_by_index_rcu(net
, fl4
->flowi4_oif
);
2056 rth
= ERR_PTR(-ENODEV
);
2057 if (dev_out
== NULL
)
2060 /* RACE: Check return value of inet_select_addr instead. */
2061 if (!(dev_out
->flags
& IFF_UP
) || !__in_dev_get_rcu(dev_out
)) {
2062 rth
= ERR_PTR(-ENETUNREACH
);
2065 if (ipv4_is_local_multicast(fl4
->daddr
) ||
2066 ipv4_is_lbcast(fl4
->daddr
)) {
2068 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2073 if (ipv4_is_multicast(fl4
->daddr
))
2074 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2076 else if (!fl4
->daddr
)
2077 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2083 fl4
->daddr
= fl4
->saddr
;
2085 fl4
->daddr
= fl4
->saddr
= htonl(INADDR_LOOPBACK
);
2086 dev_out
= net
->loopback_dev
;
2087 fl4
->flowi4_oif
= LOOPBACK_IFINDEX
;
2088 res
.type
= RTN_LOCAL
;
2089 flags
|= RTCF_LOCAL
;
2093 if (fib_lookup(net
, fl4
, &res
)) {
2096 if (fl4
->flowi4_oif
) {
2097 /* Apparently, routing tables are wrong. Assume,
2098 that the destination is on link.
2101 Because we are allowed to send to iface
2102 even if it has NO routes and NO assigned
2103 addresses. When oif is specified, routing
2104 tables are looked up with only one purpose:
2105 to catch if destination is gatewayed, rather than
2106 direct. Moreover, if MSG_DONTROUTE is set,
2107 we send packet, ignoring both routing tables
2108 and ifaddr state. --ANK
2111 We could make it even if oif is unknown,
2112 likely IPv6, but we do not.
2115 if (fl4
->saddr
== 0)
2116 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2118 res
.type
= RTN_UNICAST
;
2121 rth
= ERR_PTR(-ENETUNREACH
);
2125 if (res
.type
== RTN_LOCAL
) {
2127 if (res
.fi
->fib_prefsrc
)
2128 fl4
->saddr
= res
.fi
->fib_prefsrc
;
2130 fl4
->saddr
= fl4
->daddr
;
2132 dev_out
= net
->loopback_dev
;
2133 fl4
->flowi4_oif
= dev_out
->ifindex
;
2134 flags
|= RTCF_LOCAL
;
2138 #ifdef CONFIG_IP_ROUTE_MULTIPATH
2139 if (res
.fi
->fib_nhs
> 1 && fl4
->flowi4_oif
== 0)
2140 fib_select_multipath(&res
);
2143 if (!res
.prefixlen
&&
2144 res
.table
->tb_num_default
> 1 &&
2145 res
.type
== RTN_UNICAST
&& !fl4
->flowi4_oif
)
2146 fib_select_default(&res
);
2149 fl4
->saddr
= FIB_RES_PREFSRC(net
, res
);
2151 dev_out
= FIB_RES_DEV(res
);
2152 fl4
->flowi4_oif
= dev_out
->ifindex
;
2156 rth
= __mkroute_output(&res
, fl4
, orig_oif
, dev_out
, flags
);
2162 EXPORT_SYMBOL_GPL(__ip_route_output_key
);
2164 static struct dst_entry
*ipv4_blackhole_dst_check(struct dst_entry
*dst
, u32 cookie
)
2169 static unsigned int ipv4_blackhole_mtu(const struct dst_entry
*dst
)
2171 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
2173 return mtu
? : dst
->dev
->mtu
;
2176 static void ipv4_rt_blackhole_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
2177 struct sk_buff
*skb
, u32 mtu
)
2181 static void ipv4_rt_blackhole_redirect(struct dst_entry
*dst
, struct sock
*sk
,
2182 struct sk_buff
*skb
)
2186 static u32
*ipv4_rt_blackhole_cow_metrics(struct dst_entry
*dst
,
2192 static struct dst_ops ipv4_dst_blackhole_ops
= {
2194 .protocol
= cpu_to_be16(ETH_P_IP
),
2195 .check
= ipv4_blackhole_dst_check
,
2196 .mtu
= ipv4_blackhole_mtu
,
2197 .default_advmss
= ipv4_default_advmss
,
2198 .update_pmtu
= ipv4_rt_blackhole_update_pmtu
,
2199 .redirect
= ipv4_rt_blackhole_redirect
,
2200 .cow_metrics
= ipv4_rt_blackhole_cow_metrics
,
2201 .neigh_lookup
= ipv4_neigh_lookup
,
2204 struct dst_entry
*ipv4_blackhole_route(struct net
*net
, struct dst_entry
*dst_orig
)
2206 struct rtable
*ort
= (struct rtable
*) dst_orig
;
2209 rt
= dst_alloc(&ipv4_dst_blackhole_ops
, NULL
, 1, DST_OBSOLETE_NONE
, 0);
2211 struct dst_entry
*new = &rt
->dst
;
2214 new->input
= dst_discard
;
2215 new->output
= dst_discard
;
2217 new->dev
= ort
->dst
.dev
;
2221 rt
->rt_is_input
= ort
->rt_is_input
;
2222 rt
->rt_iif
= ort
->rt_iif
;
2223 rt
->rt_pmtu
= ort
->rt_pmtu
;
2225 rt
->rt_genid
= rt_genid(net
);
2226 rt
->rt_flags
= ort
->rt_flags
;
2227 rt
->rt_type
= ort
->rt_type
;
2228 rt
->rt_gateway
= ort
->rt_gateway
;
2229 rt
->rt_uses_gateway
= ort
->rt_uses_gateway
;
2231 INIT_LIST_HEAD(&rt
->rt_uncached
);
2236 dst_release(dst_orig
);
2238 return rt
? &rt
->dst
: ERR_PTR(-ENOMEM
);
2241 struct rtable
*ip_route_output_flow(struct net
*net
, struct flowi4
*flp4
,
2244 struct rtable
*rt
= __ip_route_output_key(net
, flp4
);
2249 if (flp4
->flowi4_proto
)
2250 rt
= (struct rtable
*) xfrm_lookup(net
, &rt
->dst
,
2251 flowi4_to_flowi(flp4
),
2256 EXPORT_SYMBOL_GPL(ip_route_output_flow
);
2258 static int rt_fill_info(struct net
*net
, __be32 dst
, __be32 src
,
2259 struct flowi4
*fl4
, struct sk_buff
*skb
, u32 portid
,
2260 u32 seq
, int event
, int nowait
, unsigned int flags
)
2262 struct rtable
*rt
= skb_rtable(skb
);
2264 struct nlmsghdr
*nlh
;
2265 unsigned long expires
= 0;
2267 u32 metrics
[RTAX_MAX
];
2269 nlh
= nlmsg_put(skb
, portid
, seq
, event
, sizeof(*r
), flags
);
2273 r
= nlmsg_data(nlh
);
2274 r
->rtm_family
= AF_INET
;
2275 r
->rtm_dst_len
= 32;
2277 r
->rtm_tos
= fl4
->flowi4_tos
;
2278 r
->rtm_table
= RT_TABLE_MAIN
;
2279 if (nla_put_u32(skb
, RTA_TABLE
, RT_TABLE_MAIN
))
2280 goto nla_put_failure
;
2281 r
->rtm_type
= rt
->rt_type
;
2282 r
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2283 r
->rtm_protocol
= RTPROT_UNSPEC
;
2284 r
->rtm_flags
= (rt
->rt_flags
& ~0xFFFF) | RTM_F_CLONED
;
2285 if (rt
->rt_flags
& RTCF_NOTIFY
)
2286 r
->rtm_flags
|= RTM_F_NOTIFY
;
2287 if (IPCB(skb
)->flags
& IPSKB_DOREDIRECT
)
2288 r
->rtm_flags
|= RTCF_DOREDIRECT
;
2290 if (nla_put_be32(skb
, RTA_DST
, dst
))
2291 goto nla_put_failure
;
2293 r
->rtm_src_len
= 32;
2294 if (nla_put_be32(skb
, RTA_SRC
, src
))
2295 goto nla_put_failure
;
2298 nla_put_u32(skb
, RTA_OIF
, rt
->dst
.dev
->ifindex
))
2299 goto nla_put_failure
;
2300 #ifdef CONFIG_IP_ROUTE_CLASSID
2301 if (rt
->dst
.tclassid
&&
2302 nla_put_u32(skb
, RTA_FLOW
, rt
->dst
.tclassid
))
2303 goto nla_put_failure
;
2305 if (!rt_is_input_route(rt
) &&
2306 fl4
->saddr
!= src
) {
2307 if (nla_put_be32(skb
, RTA_PREFSRC
, fl4
->saddr
))
2308 goto nla_put_failure
;
2310 if (rt
->rt_uses_gateway
&&
2311 nla_put_be32(skb
, RTA_GATEWAY
, rt
->rt_gateway
))
2312 goto nla_put_failure
;
2314 expires
= rt
->dst
.expires
;
2316 unsigned long now
= jiffies
;
2318 if (time_before(now
, expires
))
2324 memcpy(metrics
, dst_metrics_ptr(&rt
->dst
), sizeof(metrics
));
2325 if (rt
->rt_pmtu
&& expires
)
2326 metrics
[RTAX_MTU
- 1] = rt
->rt_pmtu
;
2327 if (rtnetlink_put_metrics(skb
, metrics
) < 0)
2328 goto nla_put_failure
;
2330 if (fl4
->flowi4_mark
&&
2331 nla_put_u32(skb
, RTA_MARK
, fl4
->flowi4_mark
))
2332 goto nla_put_failure
;
2334 if (!uid_eq(fl4
->flowi4_uid
, INVALID_UID
) &&
2335 nla_put_u32(skb
, RTA_UID
,
2336 from_kuid_munged(current_user_ns(), fl4
->flowi4_uid
)))
2337 goto nla_put_failure
;
2339 error
= rt
->dst
.error
;
2341 if (rt_is_input_route(rt
)) {
2342 #ifdef CONFIG_IP_MROUTE
2343 if (ipv4_is_multicast(dst
) && !ipv4_is_local_multicast(dst
) &&
2344 IPV4_DEVCONF_ALL(net
, MC_FORWARDING
)) {
2345 int err
= ipmr_get_route(net
, skb
,
2346 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
, skb
->dev
->ifindex
))
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
;
2391 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv4_policy
);
2395 rtm
= nlmsg_data(nlh
);
2397 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
2403 /* Reserve room for dummy headers, this skb can pass
2404 through good chunk of routing engine.
2406 skb_reset_mac_header(skb
);
2407 skb_reset_network_header(skb
);
2409 /* Bugfix: need to give ip_route_input enough of an IP header to not gag. */
2410 ip_hdr(skb
)->protocol
= IPPROTO_ICMP
;
2411 skb_reserve(skb
, MAX_HEADER
+ sizeof(struct iphdr
));
2413 src
= tb
[RTA_SRC
] ? nla_get_be32(tb
[RTA_SRC
]) : 0;
2414 dst
= tb
[RTA_DST
] ? nla_get_be32(tb
[RTA_DST
]) : 0;
2415 iif
= tb
[RTA_IIF
] ? nla_get_u32(tb
[RTA_IIF
]) : 0;
2416 mark
= tb
[RTA_MARK
] ? nla_get_u32(tb
[RTA_MARK
]) : 0;
2418 uid
= make_kuid(current_user_ns(), nla_get_u32(tb
[RTA_UID
]));
2420 uid
= (iif
? INVALID_UID
: current_uid());
2422 memset(&fl4
, 0, sizeof(fl4
));
2425 fl4
.flowi4_tos
= rtm
->rtm_tos
;
2426 fl4
.flowi4_oif
= tb
[RTA_OIF
] ? nla_get_u32(tb
[RTA_OIF
]) : 0;
2427 fl4
.flowi4_mark
= mark
;
2428 fl4
.flowi4_uid
= uid
;
2431 struct net_device
*dev
;
2433 dev
= __dev_get_by_index(net
, iif
);
2439 skb
->protocol
= htons(ETH_P_IP
);
2443 err
= ip_route_input(skb
, dst
, src
, rtm
->rtm_tos
, dev
);
2446 rt
= skb_rtable(skb
);
2447 if (err
== 0 && rt
->dst
.error
)
2448 err
= -rt
->dst
.error
;
2450 rt
= ip_route_output_key(net
, &fl4
);
2460 skb_dst_set(skb
, &rt
->dst
);
2461 if (rtm
->rtm_flags
& RTM_F_NOTIFY
)
2462 rt
->rt_flags
|= RTCF_NOTIFY
;
2464 err
= rt_fill_info(net
, dst
, src
, &fl4
, skb
,
2465 NETLINK_CB(in_skb
).portid
, nlh
->nlmsg_seq
,
2466 RTM_NEWROUTE
, 0, 0);
2470 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).portid
);
2479 int ip_rt_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2484 void ip_rt_multicast_event(struct in_device
*in_dev
)
2486 rt_cache_flush(dev_net(in_dev
->dev
));
2489 #ifdef CONFIG_SYSCTL
2490 static int ip_rt_gc_timeout __read_mostly
= RT_GC_TIMEOUT
;
2491 static int ip_rt_gc_interval __read_mostly
= 60 * HZ
;
2492 static int ip_rt_gc_min_interval __read_mostly
= HZ
/ 2;
2493 static int ip_rt_gc_elasticity __read_mostly
= 8;
2495 static int ipv4_sysctl_rtcache_flush(ctl_table
*__ctl
, int write
,
2496 void __user
*buffer
,
2497 size_t *lenp
, loff_t
*ppos
)
2500 rt_cache_flush((struct net
*)__ctl
->extra1
);
2507 static ctl_table ipv4_route_table
[] = {
2509 .procname
= "gc_thresh",
2510 .data
= &ipv4_dst_ops
.gc_thresh
,
2511 .maxlen
= sizeof(int),
2513 .proc_handler
= proc_dointvec
,
2516 .procname
= "max_size",
2517 .data
= &ip_rt_max_size
,
2518 .maxlen
= sizeof(int),
2520 .proc_handler
= proc_dointvec
,
2523 /* Deprecated. Use gc_min_interval_ms */
2525 .procname
= "gc_min_interval",
2526 .data
= &ip_rt_gc_min_interval
,
2527 .maxlen
= sizeof(int),
2529 .proc_handler
= proc_dointvec_jiffies
,
2532 .procname
= "gc_min_interval_ms",
2533 .data
= &ip_rt_gc_min_interval
,
2534 .maxlen
= sizeof(int),
2536 .proc_handler
= proc_dointvec_ms_jiffies
,
2539 .procname
= "gc_timeout",
2540 .data
= &ip_rt_gc_timeout
,
2541 .maxlen
= sizeof(int),
2543 .proc_handler
= proc_dointvec_jiffies
,
2546 .procname
= "gc_interval",
2547 .data
= &ip_rt_gc_interval
,
2548 .maxlen
= sizeof(int),
2550 .proc_handler
= proc_dointvec_jiffies
,
2553 .procname
= "redirect_load",
2554 .data
= &ip_rt_redirect_load
,
2555 .maxlen
= sizeof(int),
2557 .proc_handler
= proc_dointvec
,
2560 .procname
= "redirect_number",
2561 .data
= &ip_rt_redirect_number
,
2562 .maxlen
= sizeof(int),
2564 .proc_handler
= proc_dointvec
,
2567 .procname
= "redirect_silence",
2568 .data
= &ip_rt_redirect_silence
,
2569 .maxlen
= sizeof(int),
2571 .proc_handler
= proc_dointvec
,
2574 .procname
= "error_cost",
2575 .data
= &ip_rt_error_cost
,
2576 .maxlen
= sizeof(int),
2578 .proc_handler
= proc_dointvec
,
2581 .procname
= "error_burst",
2582 .data
= &ip_rt_error_burst
,
2583 .maxlen
= sizeof(int),
2585 .proc_handler
= proc_dointvec
,
2588 .procname
= "gc_elasticity",
2589 .data
= &ip_rt_gc_elasticity
,
2590 .maxlen
= sizeof(int),
2592 .proc_handler
= proc_dointvec
,
2595 .procname
= "mtu_expires",
2596 .data
= &ip_rt_mtu_expires
,
2597 .maxlen
= sizeof(int),
2599 .proc_handler
= proc_dointvec_jiffies
,
2602 .procname
= "min_pmtu",
2603 .data
= &ip_rt_min_pmtu
,
2604 .maxlen
= sizeof(int),
2606 .proc_handler
= proc_dointvec
,
2609 .procname
= "min_adv_mss",
2610 .data
= &ip_rt_min_advmss
,
2611 .maxlen
= sizeof(int),
2613 .proc_handler
= proc_dointvec
,
2618 static struct ctl_table ipv4_route_flush_table
[] = {
2620 .procname
= "flush",
2621 .maxlen
= sizeof(int),
2623 .proc_handler
= ipv4_sysctl_rtcache_flush
,
2628 static __net_init
int sysctl_route_net_init(struct net
*net
)
2630 struct ctl_table
*tbl
;
2632 tbl
= ipv4_route_flush_table
;
2633 if (!net_eq(net
, &init_net
)) {
2634 tbl
= kmemdup(tbl
, sizeof(ipv4_route_flush_table
), GFP_KERNEL
);
2638 /* Don't export sysctls to unprivileged users */
2639 if (net
->user_ns
!= &init_user_ns
)
2640 tbl
[0].procname
= NULL
;
2642 tbl
[0].extra1
= net
;
2644 net
->ipv4
.route_hdr
= register_net_sysctl(net
, "net/ipv4/route", tbl
);
2645 if (net
->ipv4
.route_hdr
== NULL
)
2650 if (tbl
!= ipv4_route_flush_table
)
2656 static __net_exit
void sysctl_route_net_exit(struct net
*net
)
2658 struct ctl_table
*tbl
;
2660 tbl
= net
->ipv4
.route_hdr
->ctl_table_arg
;
2661 unregister_net_sysctl_table(net
->ipv4
.route_hdr
);
2662 BUG_ON(tbl
== ipv4_route_flush_table
);
2666 static __net_initdata
struct pernet_operations sysctl_route_ops
= {
2667 .init
= sysctl_route_net_init
,
2668 .exit
= sysctl_route_net_exit
,
2672 static __net_init
int rt_genid_init(struct net
*net
)
2674 atomic_set(&net
->rt_genid
, 0);
2675 get_random_bytes(&net
->ipv4
.dev_addr_genid
,
2676 sizeof(net
->ipv4
.dev_addr_genid
));
2680 static __net_initdata
struct pernet_operations rt_genid_ops
= {
2681 .init
= rt_genid_init
,
2684 static int __net_init
ipv4_inetpeer_init(struct net
*net
)
2686 struct inet_peer_base
*bp
= kmalloc(sizeof(*bp
), GFP_KERNEL
);
2690 inet_peer_base_init(bp
);
2691 net
->ipv4
.peers
= bp
;
2695 static void __net_exit
ipv4_inetpeer_exit(struct net
*net
)
2697 struct inet_peer_base
*bp
= net
->ipv4
.peers
;
2699 net
->ipv4
.peers
= NULL
;
2700 inetpeer_invalidate_tree(bp
);
2704 static __net_initdata
struct pernet_operations ipv4_inetpeer_ops
= {
2705 .init
= ipv4_inetpeer_init
,
2706 .exit
= ipv4_inetpeer_exit
,
2709 #ifdef CONFIG_IP_ROUTE_CLASSID
2710 struct ip_rt_acct __percpu
*ip_rt_acct __read_mostly
;
2711 #endif /* CONFIG_IP_ROUTE_CLASSID */
2713 int __init
ip_rt_init(void)
2717 ip_idents
= kmalloc(IP_IDENTS_SZ
* sizeof(*ip_idents
), GFP_KERNEL
);
2719 panic("IP: failed to allocate ip_idents\n");
2721 prandom_bytes(ip_idents
, IP_IDENTS_SZ
* sizeof(*ip_idents
));
2723 #ifdef CONFIG_IP_ROUTE_CLASSID
2724 ip_rt_acct
= __alloc_percpu(256 * sizeof(struct ip_rt_acct
), __alignof__(struct ip_rt_acct
));
2726 panic("IP: failed to allocate ip_rt_acct\n");
2729 ipv4_dst_ops
.kmem_cachep
=
2730 kmem_cache_create("ip_dst_cache", sizeof(struct rtable
), 0,
2731 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
);
2733 ipv4_dst_blackhole_ops
.kmem_cachep
= ipv4_dst_ops
.kmem_cachep
;
2735 if (dst_entries_init(&ipv4_dst_ops
) < 0)
2736 panic("IP: failed to allocate ipv4_dst_ops counter\n");
2738 if (dst_entries_init(&ipv4_dst_blackhole_ops
) < 0)
2739 panic("IP: failed to allocate ipv4_dst_blackhole_ops counter\n");
2741 ipv4_dst_ops
.gc_thresh
= ~0;
2742 ip_rt_max_size
= INT_MAX
;
2747 if (ip_rt_proc_init())
2748 pr_err("Unable to create route proc files\n");
2753 rtnl_register(PF_INET
, RTM_GETROUTE
, inet_rtm_getroute
, NULL
, NULL
);
2755 #ifdef CONFIG_SYSCTL
2756 register_pernet_subsys(&sysctl_route_ops
);
2758 register_pernet_subsys(&rt_genid_ops
);
2759 register_pernet_subsys(&ipv4_inetpeer_ops
);
2763 #ifdef CONFIG_SYSCTL
2765 * We really need to sanitize the damn ipv4 init order, then all
2766 * this nonsense will go away.
2768 void __init
ip_static_sysctl_init(void)
2770 register_net_sysctl(&init_net
, "net/ipv4/route", ipv4_route_table
);