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(struct flowi4
*fl4
, const struct sock
*sk
,
519 const struct iphdr
*iph
,
521 u8 prot
, u32 mark
, int flow_flags
)
524 const struct inet_sock
*inet
= inet_sk(sk
);
526 oif
= sk
->sk_bound_dev_if
;
528 tos
= RT_CONN_FLAGS(sk
);
529 prot
= inet
->hdrincl
? IPPROTO_RAW
: sk
->sk_protocol
;
531 flowi4_init_output(fl4
, oif
, mark
, tos
,
532 RT_SCOPE_UNIVERSE
, prot
,
534 iph
->daddr
, iph
->saddr
, 0, 0);
537 static void build_skb_flow_key(struct flowi4
*fl4
, const struct sk_buff
*skb
,
538 const struct sock
*sk
)
540 const struct iphdr
*iph
= ip_hdr(skb
);
541 int oif
= skb
->dev
->ifindex
;
542 u8 tos
= RT_TOS(iph
->tos
);
543 u8 prot
= iph
->protocol
;
544 u32 mark
= skb
->mark
;
546 __build_flow_key(fl4
, sk
, iph
, oif
, tos
, prot
, mark
, 0);
549 static void build_sk_flow_key(struct flowi4
*fl4
, const struct sock
*sk
)
551 const struct inet_sock
*inet
= inet_sk(sk
);
552 const struct ip_options_rcu
*inet_opt
;
553 __be32 daddr
= inet
->inet_daddr
;
556 inet_opt
= rcu_dereference(inet
->inet_opt
);
557 if (inet_opt
&& inet_opt
->opt
.srr
)
558 daddr
= inet_opt
->opt
.faddr
;
559 flowi4_init_output(fl4
, sk
->sk_bound_dev_if
, sk
->sk_mark
,
560 RT_CONN_FLAGS(sk
), RT_SCOPE_UNIVERSE
,
561 inet
->hdrincl
? IPPROTO_RAW
: sk
->sk_protocol
,
562 inet_sk_flowi_flags(sk
),
563 daddr
, inet
->inet_saddr
, 0, 0);
567 static void ip_rt_build_flow_key(struct flowi4
*fl4
, const struct sock
*sk
,
568 const struct sk_buff
*skb
)
571 build_skb_flow_key(fl4
, skb
, sk
);
573 build_sk_flow_key(fl4
, sk
);
576 static inline void rt_free(struct rtable
*rt
)
578 call_rcu(&rt
->dst
.rcu_head
, dst_rcu_free
);
581 static DEFINE_SPINLOCK(fnhe_lock
);
583 static struct fib_nh_exception
*fnhe_oldest(struct fnhe_hash_bucket
*hash
)
585 struct fib_nh_exception
*fnhe
, *oldest
;
588 oldest
= rcu_dereference(hash
->chain
);
589 for (fnhe
= rcu_dereference(oldest
->fnhe_next
); fnhe
;
590 fnhe
= rcu_dereference(fnhe
->fnhe_next
)) {
591 if (time_before(fnhe
->fnhe_stamp
, oldest
->fnhe_stamp
))
594 orig
= rcu_dereference(oldest
->fnhe_rth
);
596 RCU_INIT_POINTER(oldest
->fnhe_rth
, NULL
);
602 static inline u32
fnhe_hashfun(__be32 daddr
)
606 hval
= (__force u32
) daddr
;
607 hval
^= (hval
>> 11) ^ (hval
>> 22);
609 return hval
& (FNHE_HASH_SIZE
- 1);
612 static void update_or_create_fnhe(struct fib_nh
*nh
, __be32 daddr
, __be32 gw
,
613 u32 pmtu
, unsigned long expires
)
615 struct fnhe_hash_bucket
*hash
;
616 struct fib_nh_exception
*fnhe
;
618 u32 hval
= fnhe_hashfun(daddr
);
620 spin_lock_bh(&fnhe_lock
);
622 hash
= nh
->nh_exceptions
;
624 hash
= kzalloc(FNHE_HASH_SIZE
* sizeof(*hash
), GFP_ATOMIC
);
627 nh
->nh_exceptions
= hash
;
633 for (fnhe
= rcu_dereference(hash
->chain
); fnhe
;
634 fnhe
= rcu_dereference(fnhe
->fnhe_next
)) {
635 if (fnhe
->fnhe_daddr
== daddr
)
644 fnhe
->fnhe_pmtu
= pmtu
;
645 fnhe
->fnhe_expires
= expires
;
648 if (depth
> FNHE_RECLAIM_DEPTH
)
649 fnhe
= fnhe_oldest(hash
);
651 fnhe
= kzalloc(sizeof(*fnhe
), GFP_ATOMIC
);
655 fnhe
->fnhe_next
= hash
->chain
;
656 rcu_assign_pointer(hash
->chain
, fnhe
);
658 fnhe
->fnhe_daddr
= daddr
;
660 fnhe
->fnhe_pmtu
= pmtu
;
661 fnhe
->fnhe_expires
= expires
;
664 fnhe
->fnhe_stamp
= jiffies
;
667 spin_unlock_bh(&fnhe_lock
);
671 static void __ip_do_redirect(struct rtable
*rt
, struct sk_buff
*skb
, struct flowi4
*fl4
,
674 __be32 new_gw
= icmp_hdr(skb
)->un
.gateway
;
675 __be32 old_gw
= ip_hdr(skb
)->saddr
;
676 struct net_device
*dev
= skb
->dev
;
677 struct in_device
*in_dev
;
678 struct fib_result res
;
682 switch (icmp_hdr(skb
)->code
& 7) {
684 case ICMP_REDIR_NETTOS
:
685 case ICMP_REDIR_HOST
:
686 case ICMP_REDIR_HOSTTOS
:
693 if (rt
->rt_gateway
!= old_gw
)
696 in_dev
= __in_dev_get_rcu(dev
);
701 if (new_gw
== old_gw
|| !IN_DEV_RX_REDIRECTS(in_dev
) ||
702 ipv4_is_multicast(new_gw
) || ipv4_is_lbcast(new_gw
) ||
703 ipv4_is_zeronet(new_gw
))
704 goto reject_redirect
;
706 if (!IN_DEV_SHARED_MEDIA(in_dev
)) {
707 if (!inet_addr_onlink(in_dev
, new_gw
, old_gw
))
708 goto reject_redirect
;
709 if (IN_DEV_SEC_REDIRECTS(in_dev
) && ip_fib_check_default(new_gw
, dev
))
710 goto reject_redirect
;
712 if (inet_addr_type(net
, new_gw
) != RTN_UNICAST
)
713 goto reject_redirect
;
716 n
= __ipv4_neigh_lookup(rt
->dst
.dev
, new_gw
);
718 n
= neigh_create(&arp_tbl
, &new_gw
, rt
->dst
.dev
);
720 if (!(n
->nud_state
& NUD_VALID
)) {
721 neigh_event_send(n
, NULL
);
723 if (fib_lookup(net
, fl4
, &res
) == 0) {
724 struct fib_nh
*nh
= &FIB_RES_NH(res
);
726 update_or_create_fnhe(nh
, fl4
->daddr
, new_gw
,
730 rt
->dst
.obsolete
= DST_OBSOLETE_KILL
;
731 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE
, n
);
738 #ifdef CONFIG_IP_ROUTE_VERBOSE
739 if (IN_DEV_LOG_MARTIANS(in_dev
)) {
740 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
741 __be32 daddr
= iph
->daddr
;
742 __be32 saddr
= iph
->saddr
;
744 net_info_ratelimited("Redirect from %pI4 on %s about %pI4 ignored\n"
745 " Advised path = %pI4 -> %pI4\n",
746 &old_gw
, dev
->name
, &new_gw
,
753 static void ip_do_redirect(struct dst_entry
*dst
, struct sock
*sk
, struct sk_buff
*skb
)
757 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
758 int oif
= skb
->dev
->ifindex
;
759 u8 tos
= RT_TOS(iph
->tos
);
760 u8 prot
= iph
->protocol
;
761 u32 mark
= skb
->mark
;
763 rt
= (struct rtable
*) dst
;
765 __build_flow_key(&fl4
, sk
, iph
, oif
, tos
, prot
, mark
, 0);
766 __ip_do_redirect(rt
, skb
, &fl4
, true);
769 static struct dst_entry
*ipv4_negative_advice(struct dst_entry
*dst
)
771 struct rtable
*rt
= (struct rtable
*)dst
;
772 struct dst_entry
*ret
= dst
;
775 if (dst
->obsolete
> 0) {
778 } else if ((rt
->rt_flags
& RTCF_REDIRECTED
) ||
789 * 1. The first ip_rt_redirect_number redirects are sent
790 * with exponential backoff, then we stop sending them at all,
791 * assuming that the host ignores our redirects.
792 * 2. If we did not see packets requiring redirects
793 * during ip_rt_redirect_silence, we assume that the host
794 * forgot redirected route and start to send redirects again.
796 * This algorithm is much cheaper and more intelligent than dumb load limiting
799 * NOTE. Do not forget to inhibit load limiting for redirects (redundant)
800 * and "frag. need" (breaks PMTU discovery) in icmp.c.
803 void ip_rt_send_redirect(struct sk_buff
*skb
)
805 struct rtable
*rt
= skb_rtable(skb
);
806 struct in_device
*in_dev
;
807 struct inet_peer
*peer
;
812 in_dev
= __in_dev_get_rcu(rt
->dst
.dev
);
813 if (!in_dev
|| !IN_DEV_TX_REDIRECTS(in_dev
)) {
817 log_martians
= IN_DEV_LOG_MARTIANS(in_dev
);
820 net
= dev_net(rt
->dst
.dev
);
821 peer
= inet_getpeer_v4(net
->ipv4
.peers
, ip_hdr(skb
)->saddr
, 1);
823 icmp_send(skb
, ICMP_REDIRECT
, ICMP_REDIR_HOST
,
824 rt_nexthop(rt
, ip_hdr(skb
)->daddr
));
828 /* No redirected packets during ip_rt_redirect_silence;
829 * reset the algorithm.
831 if (time_after(jiffies
, peer
->rate_last
+ ip_rt_redirect_silence
))
832 peer
->rate_tokens
= 0;
834 /* Too many ignored redirects; do not send anything
835 * set dst.rate_last to the last seen redirected packet.
837 if (peer
->rate_tokens
>= ip_rt_redirect_number
) {
838 peer
->rate_last
= jiffies
;
842 /* Check for load limit; set rate_last to the latest sent
845 if (peer
->rate_tokens
== 0 ||
848 (ip_rt_redirect_load
<< peer
->rate_tokens
)))) {
849 __be32 gw
= rt_nexthop(rt
, ip_hdr(skb
)->daddr
);
851 icmp_send(skb
, ICMP_REDIRECT
, ICMP_REDIR_HOST
, gw
);
852 peer
->rate_last
= jiffies
;
854 #ifdef CONFIG_IP_ROUTE_VERBOSE
856 peer
->rate_tokens
== ip_rt_redirect_number
)
857 net_warn_ratelimited("host %pI4/if%d ignores redirects for %pI4 to %pI4\n",
858 &ip_hdr(skb
)->saddr
, inet_iif(skb
),
859 &ip_hdr(skb
)->daddr
, &gw
);
866 static int ip_error(struct sk_buff
*skb
)
868 struct in_device
*in_dev
= __in_dev_get_rcu(skb
->dev
);
869 struct rtable
*rt
= skb_rtable(skb
);
870 struct inet_peer
*peer
;
876 /* IP on this device is disabled. */
880 net
= dev_net(rt
->dst
.dev
);
881 if (!IN_DEV_FORWARD(in_dev
)) {
882 switch (rt
->dst
.error
) {
884 IP_INC_STATS_BH(net
, IPSTATS_MIB_INADDRERRORS
);
888 IP_INC_STATS_BH(net
, IPSTATS_MIB_INNOROUTES
);
894 switch (rt
->dst
.error
) {
899 code
= ICMP_HOST_UNREACH
;
902 code
= ICMP_NET_UNREACH
;
903 IP_INC_STATS_BH(net
, IPSTATS_MIB_INNOROUTES
);
906 code
= ICMP_PKT_FILTERED
;
910 peer
= inet_getpeer_v4(net
->ipv4
.peers
, ip_hdr(skb
)->saddr
, 1);
915 peer
->rate_tokens
+= now
- peer
->rate_last
;
916 if (peer
->rate_tokens
> ip_rt_error_burst
)
917 peer
->rate_tokens
= ip_rt_error_burst
;
918 peer
->rate_last
= now
;
919 if (peer
->rate_tokens
>= ip_rt_error_cost
)
920 peer
->rate_tokens
-= ip_rt_error_cost
;
926 icmp_send(skb
, ICMP_DEST_UNREACH
, code
, 0);
932 static void __ip_rt_update_pmtu(struct rtable
*rt
, struct flowi4
*fl4
, u32 mtu
)
934 struct dst_entry
*dst
= &rt
->dst
;
935 struct fib_result res
;
937 if (dst_metric_locked(dst
, RTAX_MTU
))
940 if (dst
->dev
->mtu
< mtu
)
943 if (mtu
< ip_rt_min_pmtu
)
944 mtu
= ip_rt_min_pmtu
;
947 dst
->obsolete
= DST_OBSOLETE_KILL
;
950 dst
->expires
= max(1UL, jiffies
+ ip_rt_mtu_expires
);
954 if (fib_lookup(dev_net(dst
->dev
), fl4
, &res
) == 0) {
955 struct fib_nh
*nh
= &FIB_RES_NH(res
);
957 update_or_create_fnhe(nh
, fl4
->daddr
, 0, mtu
,
958 jiffies
+ ip_rt_mtu_expires
);
963 static void ip_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
964 struct sk_buff
*skb
, u32 mtu
)
966 struct rtable
*rt
= (struct rtable
*) dst
;
969 ip_rt_build_flow_key(&fl4
, sk
, skb
);
970 __ip_rt_update_pmtu(rt
, &fl4
, mtu
);
973 void ipv4_update_pmtu(struct sk_buff
*skb
, struct net
*net
, u32 mtu
,
974 int oif
, u32 mark
, u8 protocol
, int flow_flags
)
976 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
980 __build_flow_key(&fl4
, NULL
, iph
, oif
,
981 RT_TOS(iph
->tos
), protocol
, mark
, flow_flags
);
982 rt
= __ip_route_output_key(net
, &fl4
);
984 __ip_rt_update_pmtu(rt
, &fl4
, mtu
);
988 EXPORT_SYMBOL_GPL(ipv4_update_pmtu
);
990 static void __ipv4_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, u32 mtu
)
992 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
996 __build_flow_key(&fl4
, sk
, iph
, 0, 0, 0, 0, 0);
997 rt
= __ip_route_output_key(sock_net(sk
), &fl4
);
999 __ip_rt_update_pmtu(rt
, &fl4
, mtu
);
1004 void ipv4_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, u32 mtu
)
1006 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1009 struct dst_entry
*odst
= NULL
;
1013 odst
= sk_dst_get(sk
);
1015 if (sock_owned_by_user(sk
) || !odst
) {
1016 __ipv4_sk_update_pmtu(skb
, sk
, mtu
);
1020 __build_flow_key(&fl4
, sk
, iph
, 0, 0, 0, 0, 0);
1022 rt
= (struct rtable
*)odst
;
1023 if (odst
->obsolete
&& odst
->ops
->check(odst
, 0) == NULL
) {
1024 rt
= ip_route_output_flow(sock_net(sk
), &fl4
, sk
);
1031 __ip_rt_update_pmtu((struct rtable
*) rt
->dst
.path
, &fl4
, mtu
);
1033 if (!dst_check(&rt
->dst
, 0)) {
1035 dst_release(&rt
->dst
);
1037 rt
= ip_route_output_flow(sock_net(sk
), &fl4
, sk
);
1045 sk_dst_set(sk
, &rt
->dst
);
1051 EXPORT_SYMBOL_GPL(ipv4_sk_update_pmtu
);
1053 void ipv4_redirect(struct sk_buff
*skb
, struct net
*net
,
1054 int oif
, u32 mark
, u8 protocol
, int flow_flags
)
1056 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1060 __build_flow_key(&fl4
, NULL
, iph
, oif
,
1061 RT_TOS(iph
->tos
), protocol
, mark
, flow_flags
);
1062 rt
= __ip_route_output_key(net
, &fl4
);
1064 __ip_do_redirect(rt
, skb
, &fl4
, false);
1068 EXPORT_SYMBOL_GPL(ipv4_redirect
);
1070 void ipv4_sk_redirect(struct sk_buff
*skb
, struct sock
*sk
)
1072 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1076 __build_flow_key(&fl4
, sk
, iph
, 0, 0, 0, 0, 0);
1077 rt
= __ip_route_output_key(sock_net(sk
), &fl4
);
1079 __ip_do_redirect(rt
, skb
, &fl4
, false);
1083 EXPORT_SYMBOL_GPL(ipv4_sk_redirect
);
1085 static struct dst_entry
*ipv4_dst_check(struct dst_entry
*dst
, u32 cookie
)
1087 struct rtable
*rt
= (struct rtable
*) dst
;
1089 /* All IPV4 dsts are created with ->obsolete set to the value
1090 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1091 * into this function always.
1093 * When a PMTU/redirect information update invalidates a
1094 * route, this is indicated by setting obsolete to
1095 * DST_OBSOLETE_KILL.
1097 if (dst
->obsolete
== DST_OBSOLETE_KILL
|| rt_is_expired(rt
))
1102 static void ipv4_link_failure(struct sk_buff
*skb
)
1106 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_HOST_UNREACH
, 0);
1108 rt
= skb_rtable(skb
);
1110 dst_set_expires(&rt
->dst
, 0);
1113 static int ip_rt_bug(struct sk_buff
*skb
)
1115 pr_debug("%s: %pI4 -> %pI4, %s\n",
1116 __func__
, &ip_hdr(skb
)->saddr
, &ip_hdr(skb
)->daddr
,
1117 skb
->dev
? skb
->dev
->name
: "?");
1124 We do not cache source address of outgoing interface,
1125 because it is used only by IP RR, TS and SRR options,
1126 so that it out of fast path.
1128 BTW remember: "addr" is allowed to be not aligned
1132 void ip_rt_get_source(u8
*addr
, struct sk_buff
*skb
, struct rtable
*rt
)
1136 if (rt_is_output_route(rt
))
1137 src
= ip_hdr(skb
)->saddr
;
1139 struct fib_result res
;
1145 memset(&fl4
, 0, sizeof(fl4
));
1146 fl4
.daddr
= iph
->daddr
;
1147 fl4
.saddr
= iph
->saddr
;
1148 fl4
.flowi4_tos
= RT_TOS(iph
->tos
);
1149 fl4
.flowi4_oif
= rt
->dst
.dev
->ifindex
;
1150 fl4
.flowi4_iif
= skb
->dev
->ifindex
;
1151 fl4
.flowi4_mark
= skb
->mark
;
1154 if (fib_lookup(dev_net(rt
->dst
.dev
), &fl4
, &res
) == 0)
1155 src
= FIB_RES_PREFSRC(dev_net(rt
->dst
.dev
), res
);
1157 src
= inet_select_addr(rt
->dst
.dev
,
1158 rt_nexthop(rt
, iph
->daddr
),
1162 memcpy(addr
, &src
, 4);
1165 #ifdef CONFIG_IP_ROUTE_CLASSID
1166 static void set_class_tag(struct rtable
*rt
, u32 tag
)
1168 if (!(rt
->dst
.tclassid
& 0xFFFF))
1169 rt
->dst
.tclassid
|= tag
& 0xFFFF;
1170 if (!(rt
->dst
.tclassid
& 0xFFFF0000))
1171 rt
->dst
.tclassid
|= tag
& 0xFFFF0000;
1175 static unsigned int ipv4_default_advmss(const struct dst_entry
*dst
)
1177 unsigned int advmss
= dst_metric_raw(dst
, RTAX_ADVMSS
);
1180 advmss
= max_t(unsigned int, dst
->dev
->mtu
- 40,
1182 if (advmss
> 65535 - 40)
1183 advmss
= 65535 - 40;
1188 static unsigned int ipv4_mtu(const struct dst_entry
*dst
)
1190 const struct rtable
*rt
= (const struct rtable
*) dst
;
1191 unsigned int mtu
= rt
->rt_pmtu
;
1193 if (!mtu
|| time_after_eq(jiffies
, rt
->dst
.expires
))
1194 mtu
= dst_metric_raw(dst
, RTAX_MTU
);
1199 mtu
= dst
->dev
->mtu
;
1201 if (unlikely(dst_metric_locked(dst
, RTAX_MTU
))) {
1202 if (rt
->rt_uses_gateway
&& mtu
> 576)
1206 if (mtu
> IP_MAX_MTU
)
1212 static struct fib_nh_exception
*find_exception(struct fib_nh
*nh
, __be32 daddr
)
1214 struct fnhe_hash_bucket
*hash
= nh
->nh_exceptions
;
1215 struct fib_nh_exception
*fnhe
;
1221 hval
= fnhe_hashfun(daddr
);
1223 for (fnhe
= rcu_dereference(hash
[hval
].chain
); fnhe
;
1224 fnhe
= rcu_dereference(fnhe
->fnhe_next
)) {
1225 if (fnhe
->fnhe_daddr
== daddr
)
1231 static bool rt_bind_exception(struct rtable
*rt
, struct fib_nh_exception
*fnhe
,
1236 spin_lock_bh(&fnhe_lock
);
1238 if (daddr
== fnhe
->fnhe_daddr
) {
1239 struct rtable
*orig
= rcu_dereference(fnhe
->fnhe_rth
);
1240 if (orig
&& rt_is_expired(orig
)) {
1242 fnhe
->fnhe_pmtu
= 0;
1243 fnhe
->fnhe_expires
= 0;
1245 if (fnhe
->fnhe_pmtu
) {
1246 unsigned long expires
= fnhe
->fnhe_expires
;
1247 unsigned long diff
= expires
- jiffies
;
1249 if (time_before(jiffies
, expires
)) {
1250 rt
->rt_pmtu
= fnhe
->fnhe_pmtu
;
1251 dst_set_expires(&rt
->dst
, diff
);
1254 if (fnhe
->fnhe_gw
) {
1255 rt
->rt_flags
|= RTCF_REDIRECTED
;
1256 rt
->rt_gateway
= fnhe
->fnhe_gw
;
1257 rt
->rt_uses_gateway
= 1;
1258 } else if (!rt
->rt_gateway
)
1259 rt
->rt_gateway
= daddr
;
1261 rcu_assign_pointer(fnhe
->fnhe_rth
, rt
);
1265 fnhe
->fnhe_stamp
= jiffies
;
1268 spin_unlock_bh(&fnhe_lock
);
1273 static bool rt_cache_route(struct fib_nh
*nh
, struct rtable
*rt
)
1275 struct rtable
*orig
, *prev
, **p
;
1278 if (rt_is_input_route(rt
)) {
1279 p
= (struct rtable
**)&nh
->nh_rth_input
;
1281 p
= (struct rtable
**)__this_cpu_ptr(nh
->nh_pcpu_rth_output
);
1285 prev
= cmpxchg(p
, orig
, rt
);
1295 static DEFINE_SPINLOCK(rt_uncached_lock
);
1296 static LIST_HEAD(rt_uncached_list
);
1298 static void rt_add_uncached_list(struct rtable
*rt
)
1300 spin_lock_bh(&rt_uncached_lock
);
1301 list_add_tail(&rt
->rt_uncached
, &rt_uncached_list
);
1302 spin_unlock_bh(&rt_uncached_lock
);
1305 static void ipv4_dst_destroy(struct dst_entry
*dst
)
1307 struct rtable
*rt
= (struct rtable
*) dst
;
1309 if (!list_empty(&rt
->rt_uncached
)) {
1310 spin_lock_bh(&rt_uncached_lock
);
1311 list_del(&rt
->rt_uncached
);
1312 spin_unlock_bh(&rt_uncached_lock
);
1316 void rt_flush_dev(struct net_device
*dev
)
1318 if (!list_empty(&rt_uncached_list
)) {
1319 struct net
*net
= dev_net(dev
);
1322 spin_lock_bh(&rt_uncached_lock
);
1323 list_for_each_entry(rt
, &rt_uncached_list
, rt_uncached
) {
1324 if (rt
->dst
.dev
!= dev
)
1326 rt
->dst
.dev
= net
->loopback_dev
;
1327 dev_hold(rt
->dst
.dev
);
1330 spin_unlock_bh(&rt_uncached_lock
);
1334 static bool rt_cache_valid(const struct rtable
*rt
)
1337 rt
->dst
.obsolete
== DST_OBSOLETE_FORCE_CHK
&&
1341 static void rt_set_nexthop(struct rtable
*rt
, __be32 daddr
,
1342 const struct fib_result
*res
,
1343 struct fib_nh_exception
*fnhe
,
1344 struct fib_info
*fi
, u16 type
, u32 itag
)
1346 bool cached
= false;
1349 struct fib_nh
*nh
= &FIB_RES_NH(*res
);
1351 if (nh
->nh_gw
&& nh
->nh_scope
== RT_SCOPE_LINK
) {
1352 rt
->rt_gateway
= nh
->nh_gw
;
1353 rt
->rt_uses_gateway
= 1;
1355 dst_init_metrics(&rt
->dst
, fi
->fib_metrics
, true);
1356 #ifdef CONFIG_IP_ROUTE_CLASSID
1357 rt
->dst
.tclassid
= nh
->nh_tclassid
;
1360 cached
= rt_bind_exception(rt
, fnhe
, daddr
);
1361 else if (!(rt
->dst
.flags
& DST_NOCACHE
))
1362 cached
= rt_cache_route(nh
, rt
);
1363 if (unlikely(!cached
)) {
1364 /* Routes we intend to cache in nexthop exception or
1365 * FIB nexthop have the DST_NOCACHE bit clear.
1366 * However, if we are unsuccessful at storing this
1367 * route into the cache we really need to set it.
1369 rt
->dst
.flags
|= DST_NOCACHE
;
1370 if (!rt
->rt_gateway
)
1371 rt
->rt_gateway
= daddr
;
1372 rt_add_uncached_list(rt
);
1375 rt_add_uncached_list(rt
);
1377 #ifdef CONFIG_IP_ROUTE_CLASSID
1378 #ifdef CONFIG_IP_MULTIPLE_TABLES
1379 set_class_tag(rt
, res
->tclassid
);
1381 set_class_tag(rt
, itag
);
1385 static struct rtable
*rt_dst_alloc(struct net_device
*dev
,
1386 bool nopolicy
, bool noxfrm
, bool will_cache
)
1388 return dst_alloc(&ipv4_dst_ops
, dev
, 1, DST_OBSOLETE_FORCE_CHK
,
1389 (will_cache
? 0 : (DST_HOST
| DST_NOCACHE
)) |
1390 (nopolicy
? DST_NOPOLICY
: 0) |
1391 (noxfrm
? DST_NOXFRM
: 0));
1394 /* called in rcu_read_lock() section */
1395 static int ip_route_input_mc(struct sk_buff
*skb
, __be32 daddr
, __be32 saddr
,
1396 u8 tos
, struct net_device
*dev
, int our
)
1399 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
1403 /* Primary sanity checks. */
1408 if (ipv4_is_multicast(saddr
) || ipv4_is_lbcast(saddr
) ||
1409 skb
->protocol
!= htons(ETH_P_IP
))
1412 if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev
)))
1413 if (ipv4_is_loopback(saddr
))
1416 if (ipv4_is_zeronet(saddr
)) {
1417 if (!ipv4_is_local_multicast(daddr
))
1420 err
= fib_validate_source(skb
, saddr
, 0, tos
, 0, dev
,
1425 rth
= rt_dst_alloc(dev_net(dev
)->loopback_dev
,
1426 IN_DEV_CONF_GET(in_dev
, NOPOLICY
), false, false);
1430 #ifdef CONFIG_IP_ROUTE_CLASSID
1431 rth
->dst
.tclassid
= itag
;
1433 rth
->dst
.output
= ip_rt_bug
;
1435 rth
->rt_genid
= rt_genid(dev_net(dev
));
1436 rth
->rt_flags
= RTCF_MULTICAST
;
1437 rth
->rt_type
= RTN_MULTICAST
;
1438 rth
->rt_is_input
= 1;
1441 rth
->rt_gateway
= 0;
1442 rth
->rt_uses_gateway
= 0;
1443 INIT_LIST_HEAD(&rth
->rt_uncached
);
1445 rth
->dst
.input
= ip_local_deliver
;
1446 rth
->rt_flags
|= RTCF_LOCAL
;
1449 #ifdef CONFIG_IP_MROUTE
1450 if (!ipv4_is_local_multicast(daddr
) && IN_DEV_MFORWARD(in_dev
))
1451 rth
->dst
.input
= ip_mr_input
;
1453 RT_CACHE_STAT_INC(in_slow_mc
);
1455 skb_dst_set(skb
, &rth
->dst
);
1467 static void ip_handle_martian_source(struct net_device
*dev
,
1468 struct in_device
*in_dev
,
1469 struct sk_buff
*skb
,
1473 RT_CACHE_STAT_INC(in_martian_src
);
1474 #ifdef CONFIG_IP_ROUTE_VERBOSE
1475 if (IN_DEV_LOG_MARTIANS(in_dev
) && net_ratelimit()) {
1477 * RFC1812 recommendation, if source is martian,
1478 * the only hint is MAC header.
1480 pr_warn("martian source %pI4 from %pI4, on dev %s\n",
1481 &daddr
, &saddr
, dev
->name
);
1482 if (dev
->hard_header_len
&& skb_mac_header_was_set(skb
)) {
1483 print_hex_dump(KERN_WARNING
, "ll header: ",
1484 DUMP_PREFIX_OFFSET
, 16, 1,
1485 skb_mac_header(skb
),
1486 dev
->hard_header_len
, true);
1492 /* called in rcu_read_lock() section */
1493 static int __mkroute_input(struct sk_buff
*skb
,
1494 const struct fib_result
*res
,
1495 struct in_device
*in_dev
,
1496 __be32 daddr
, __be32 saddr
, u32 tos
)
1500 struct in_device
*out_dev
;
1501 unsigned int flags
= 0;
1505 /* get a working reference to the output device */
1506 out_dev
= __in_dev_get_rcu(FIB_RES_DEV(*res
));
1507 if (out_dev
== NULL
) {
1508 net_crit_ratelimited("Bug in ip_route_input_slow(). Please report.\n");
1512 err
= fib_validate_source(skb
, saddr
, daddr
, tos
, FIB_RES_OIF(*res
),
1513 in_dev
->dev
, in_dev
, &itag
);
1515 ip_handle_martian_source(in_dev
->dev
, in_dev
, skb
, daddr
,
1521 do_cache
= res
->fi
&& !itag
;
1522 if (out_dev
== in_dev
&& err
&& IN_DEV_TX_REDIRECTS(out_dev
) &&
1523 skb
->protocol
== htons(ETH_P_IP
) &&
1524 (IN_DEV_SHARED_MEDIA(out_dev
) ||
1525 inet_addr_onlink(out_dev
, saddr
, FIB_RES_GW(*res
))))
1526 IPCB(skb
)->flags
|= IPSKB_DOREDIRECT
;
1528 if (skb
->protocol
!= htons(ETH_P_IP
)) {
1529 /* Not IP (i.e. ARP). Do not create route, if it is
1530 * invalid for proxy arp. DNAT routes are always valid.
1532 * Proxy arp feature have been extended to allow, ARP
1533 * replies back to the same interface, to support
1534 * Private VLAN switch technologies. See arp.c.
1536 if (out_dev
== in_dev
&&
1537 IN_DEV_PROXY_ARP_PVLAN(in_dev
) == 0) {
1544 rth
= rcu_dereference(FIB_RES_NH(*res
).nh_rth_input
);
1545 if (rt_cache_valid(rth
)) {
1546 skb_dst_set_noref(skb
, &rth
->dst
);
1551 rth
= rt_dst_alloc(out_dev
->dev
,
1552 IN_DEV_CONF_GET(in_dev
, NOPOLICY
),
1553 IN_DEV_CONF_GET(out_dev
, NOXFRM
), do_cache
);
1559 rth
->rt_genid
= rt_genid(dev_net(rth
->dst
.dev
));
1560 rth
->rt_flags
= flags
;
1561 rth
->rt_type
= res
->type
;
1562 rth
->rt_is_input
= 1;
1565 rth
->rt_gateway
= 0;
1566 rth
->rt_uses_gateway
= 0;
1567 INIT_LIST_HEAD(&rth
->rt_uncached
);
1568 RT_CACHE_STAT_INC(in_slow_tot
);
1570 rth
->dst
.input
= ip_forward
;
1571 rth
->dst
.output
= ip_output
;
1573 rt_set_nexthop(rth
, daddr
, res
, NULL
, res
->fi
, res
->type
, itag
);
1574 skb_dst_set(skb
, &rth
->dst
);
1581 static int ip_mkroute_input(struct sk_buff
*skb
,
1582 struct fib_result
*res
,
1583 const struct flowi4
*fl4
,
1584 struct in_device
*in_dev
,
1585 __be32 daddr
, __be32 saddr
, u32 tos
)
1587 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1588 if (res
->fi
&& res
->fi
->fib_nhs
> 1)
1589 fib_select_multipath(res
);
1592 /* create a routing cache entry */
1593 return __mkroute_input(skb
, res
, in_dev
, daddr
, saddr
, tos
);
1597 * NOTE. We drop all the packets that has local source
1598 * addresses, because every properly looped back packet
1599 * must have correct destination already attached by output routine.
1601 * Such approach solves two big problems:
1602 * 1. Not simplex devices are handled properly.
1603 * 2. IP spoofing attempts are filtered with 100% of guarantee.
1604 * called with rcu_read_lock()
1607 static int ip_route_input_slow(struct sk_buff
*skb
, __be32 daddr
, __be32 saddr
,
1608 u8 tos
, struct net_device
*dev
)
1610 struct fib_result res
;
1611 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
1613 unsigned int flags
= 0;
1617 struct net
*net
= dev_net(dev
);
1620 /* IP on this device is disabled. */
1625 /* Check for the most weird martians, which can be not detected
1629 if (ipv4_is_multicast(saddr
) || ipv4_is_lbcast(saddr
))
1630 goto martian_source
;
1633 if (ipv4_is_lbcast(daddr
) || (saddr
== 0 && daddr
== 0))
1636 /* Accept zero addresses only to limited broadcast;
1637 * I even do not know to fix it or not. Waiting for complains :-)
1639 if (ipv4_is_zeronet(saddr
))
1640 goto martian_source
;
1642 if (ipv4_is_zeronet(daddr
))
1643 goto martian_destination
;
1645 /* Following code try to avoid calling IN_DEV_NET_ROUTE_LOCALNET(),
1646 * and call it once if daddr or/and saddr are loopback addresses
1648 if (ipv4_is_loopback(daddr
)) {
1649 if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev
, net
))
1650 goto martian_destination
;
1651 } else if (ipv4_is_loopback(saddr
)) {
1652 if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev
, net
))
1653 goto martian_source
;
1657 * Now we are ready to route packet.
1660 fl4
.flowi4_iif
= dev
->ifindex
;
1661 fl4
.flowi4_mark
= skb
->mark
;
1662 fl4
.flowi4_tos
= tos
;
1663 fl4
.flowi4_scope
= RT_SCOPE_UNIVERSE
;
1666 err
= fib_lookup(net
, &fl4
, &res
);
1670 if (res
.type
== RTN_BROADCAST
)
1673 if (res
.type
== RTN_LOCAL
) {
1674 err
= fib_validate_source(skb
, saddr
, daddr
, tos
,
1676 dev
, in_dev
, &itag
);
1678 goto martian_source_keep_err
;
1682 if (!IN_DEV_FORWARD(in_dev
))
1684 if (res
.type
!= RTN_UNICAST
)
1685 goto martian_destination
;
1687 err
= ip_mkroute_input(skb
, &res
, &fl4
, in_dev
, daddr
, saddr
, tos
);
1691 if (skb
->protocol
!= htons(ETH_P_IP
))
1694 if (!ipv4_is_zeronet(saddr
)) {
1695 err
= fib_validate_source(skb
, saddr
, 0, tos
, 0, dev
,
1698 goto martian_source_keep_err
;
1700 flags
|= RTCF_BROADCAST
;
1701 res
.type
= RTN_BROADCAST
;
1702 RT_CACHE_STAT_INC(in_brd
);
1708 rth
= rcu_dereference(FIB_RES_NH(res
).nh_rth_input
);
1709 if (rt_cache_valid(rth
)) {
1710 skb_dst_set_noref(skb
, &rth
->dst
);
1718 rth
= rt_dst_alloc(net
->loopback_dev
,
1719 IN_DEV_CONF_GET(in_dev
, NOPOLICY
), false, do_cache
);
1723 rth
->dst
.input
= ip_local_deliver
;
1724 rth
->dst
.output
= ip_rt_bug
;
1725 #ifdef CONFIG_IP_ROUTE_CLASSID
1726 rth
->dst
.tclassid
= itag
;
1729 rth
->rt_genid
= rt_genid(net
);
1730 rth
->rt_flags
= flags
|RTCF_LOCAL
;
1731 rth
->rt_type
= res
.type
;
1732 rth
->rt_is_input
= 1;
1735 rth
->rt_gateway
= 0;
1736 rth
->rt_uses_gateway
= 0;
1737 INIT_LIST_HEAD(&rth
->rt_uncached
);
1738 RT_CACHE_STAT_INC(in_slow_tot
);
1739 if (res
.type
== RTN_UNREACHABLE
) {
1740 rth
->dst
.input
= ip_error
;
1741 rth
->dst
.error
= -err
;
1742 rth
->rt_flags
&= ~RTCF_LOCAL
;
1745 if (unlikely(!rt_cache_route(&FIB_RES_NH(res
), rth
))) {
1746 rth
->dst
.flags
|= DST_NOCACHE
;
1747 rt_add_uncached_list(rth
);
1750 skb_dst_set(skb
, &rth
->dst
);
1755 RT_CACHE_STAT_INC(in_no_route
);
1756 res
.type
= RTN_UNREACHABLE
;
1762 * Do not cache martian addresses: they should be logged (RFC1812)
1764 martian_destination
:
1765 RT_CACHE_STAT_INC(in_martian_dst
);
1766 #ifdef CONFIG_IP_ROUTE_VERBOSE
1767 if (IN_DEV_LOG_MARTIANS(in_dev
))
1768 net_warn_ratelimited("martian destination %pI4 from %pI4, dev %s\n",
1769 &daddr
, &saddr
, dev
->name
);
1782 martian_source_keep_err
:
1783 ip_handle_martian_source(dev
, in_dev
, skb
, daddr
, saddr
);
1787 int ip_route_input_noref(struct sk_buff
*skb
, __be32 daddr
, __be32 saddr
,
1788 u8 tos
, struct net_device
*dev
)
1792 tos
&= IPTOS_RT_MASK
;
1795 /* Multicast recognition logic is moved from route cache to here.
1796 The problem was that too many Ethernet cards have broken/missing
1797 hardware multicast filters :-( As result the host on multicasting
1798 network acquires a lot of useless route cache entries, sort of
1799 SDR messages from all the world. Now we try to get rid of them.
1800 Really, provided software IP multicast filter is organized
1801 reasonably (at least, hashed), it does not result in a slowdown
1802 comparing with route cache reject entries.
1803 Note, that multicast routers are not affected, because
1804 route cache entry is created eventually.
1806 if (ipv4_is_multicast(daddr
)) {
1807 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
1810 int our
= ip_check_mc_rcu(in_dev
, daddr
, saddr
,
1811 ip_hdr(skb
)->protocol
);
1813 #ifdef CONFIG_IP_MROUTE
1815 (!ipv4_is_local_multicast(daddr
) &&
1816 IN_DEV_MFORWARD(in_dev
))
1819 int res
= ip_route_input_mc(skb
, daddr
, saddr
,
1828 res
= ip_route_input_slow(skb
, daddr
, saddr
, tos
, dev
);
1832 EXPORT_SYMBOL(ip_route_input_noref
);
1834 /* called with rcu_read_lock() */
1835 static struct rtable
*__mkroute_output(const struct fib_result
*res
,
1836 const struct flowi4
*fl4
, int orig_oif
,
1837 struct net_device
*dev_out
,
1840 struct fib_info
*fi
= res
->fi
;
1841 struct fib_nh_exception
*fnhe
;
1842 struct in_device
*in_dev
;
1843 u16 type
= res
->type
;
1847 in_dev
= __in_dev_get_rcu(dev_out
);
1849 return ERR_PTR(-EINVAL
);
1851 if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev
)))
1852 if (ipv4_is_loopback(fl4
->saddr
) && !(dev_out
->flags
& IFF_LOOPBACK
))
1853 return ERR_PTR(-EINVAL
);
1855 if (ipv4_is_lbcast(fl4
->daddr
))
1856 type
= RTN_BROADCAST
;
1857 else if (ipv4_is_multicast(fl4
->daddr
))
1858 type
= RTN_MULTICAST
;
1859 else if (ipv4_is_zeronet(fl4
->daddr
))
1860 return ERR_PTR(-EINVAL
);
1862 if (dev_out
->flags
& IFF_LOOPBACK
)
1863 flags
|= RTCF_LOCAL
;
1866 if (type
== RTN_BROADCAST
) {
1867 flags
|= RTCF_BROADCAST
| RTCF_LOCAL
;
1869 } else if (type
== RTN_MULTICAST
) {
1870 flags
|= RTCF_MULTICAST
| RTCF_LOCAL
;
1871 if (!ip_check_mc_rcu(in_dev
, fl4
->daddr
, fl4
->saddr
,
1873 flags
&= ~RTCF_LOCAL
;
1876 /* If multicast route do not exist use
1877 * default one, but do not gateway in this case.
1880 if (fi
&& res
->prefixlen
< 4)
1882 } else if ((type
== RTN_LOCAL
) && (orig_oif
!= 0) &&
1883 (orig_oif
!= dev_out
->ifindex
)) {
1884 /* For local routes that require a particular output interface
1885 * we do not want to cache the result. Caching the result
1886 * causes incorrect behaviour when there are multiple source
1887 * addresses on the interface, the end result being that if the
1888 * intended recipient is waiting on that interface for the
1889 * packet he won't receive it because it will be delivered on
1890 * the loopback interface and the IP_PKTINFO ipi_ifindex will
1891 * be set to the loopback interface as well.
1897 do_cache
&= fi
!= NULL
;
1899 struct rtable __rcu
**prth
;
1900 struct fib_nh
*nh
= &FIB_RES_NH(*res
);
1902 fnhe
= find_exception(nh
, fl4
->daddr
);
1904 prth
= &fnhe
->fnhe_rth
;
1906 if (unlikely(fl4
->flowi4_flags
&
1907 FLOWI_FLAG_KNOWN_NH
&&
1909 nh
->nh_scope
== RT_SCOPE_LINK
))) {
1913 prth
= __this_cpu_ptr(nh
->nh_pcpu_rth_output
);
1915 rth
= rcu_dereference(*prth
);
1916 if (rt_cache_valid(rth
)) {
1917 dst_hold(&rth
->dst
);
1923 rth
= rt_dst_alloc(dev_out
,
1924 IN_DEV_CONF_GET(in_dev
, NOPOLICY
),
1925 IN_DEV_CONF_GET(in_dev
, NOXFRM
),
1928 return ERR_PTR(-ENOBUFS
);
1930 rth
->dst
.output
= ip_output
;
1932 rth
->rt_genid
= rt_genid(dev_net(dev_out
));
1933 rth
->rt_flags
= flags
;
1934 rth
->rt_type
= type
;
1935 rth
->rt_is_input
= 0;
1936 rth
->rt_iif
= orig_oif
? : 0;
1938 rth
->rt_gateway
= 0;
1939 rth
->rt_uses_gateway
= 0;
1940 INIT_LIST_HEAD(&rth
->rt_uncached
);
1942 RT_CACHE_STAT_INC(out_slow_tot
);
1944 if (flags
& RTCF_LOCAL
)
1945 rth
->dst
.input
= ip_local_deliver
;
1946 if (flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
)) {
1947 if (flags
& RTCF_LOCAL
&&
1948 !(dev_out
->flags
& IFF_LOOPBACK
)) {
1949 rth
->dst
.output
= ip_mc_output
;
1950 RT_CACHE_STAT_INC(out_slow_mc
);
1952 #ifdef CONFIG_IP_MROUTE
1953 if (type
== RTN_MULTICAST
) {
1954 if (IN_DEV_MFORWARD(in_dev
) &&
1955 !ipv4_is_local_multicast(fl4
->daddr
)) {
1956 rth
->dst
.input
= ip_mr_input
;
1957 rth
->dst
.output
= ip_mc_output
;
1963 rt_set_nexthop(rth
, fl4
->daddr
, res
, fnhe
, fi
, type
, 0);
1969 * Major route resolver routine.
1972 struct rtable
*__ip_route_output_key(struct net
*net
, struct flowi4
*fl4
)
1974 struct net_device
*dev_out
= NULL
;
1975 __u8 tos
= RT_FL_TOS(fl4
);
1976 unsigned int flags
= 0;
1977 struct fib_result res
;
1985 orig_oif
= fl4
->flowi4_oif
;
1987 fl4
->flowi4_iif
= LOOPBACK_IFINDEX
;
1988 fl4
->flowi4_tos
= tos
& IPTOS_RT_MASK
;
1989 fl4
->flowi4_scope
= ((tos
& RTO_ONLINK
) ?
1990 RT_SCOPE_LINK
: RT_SCOPE_UNIVERSE
);
1994 rth
= ERR_PTR(-EINVAL
);
1995 if (ipv4_is_multicast(fl4
->saddr
) ||
1996 ipv4_is_lbcast(fl4
->saddr
) ||
1997 ipv4_is_zeronet(fl4
->saddr
))
2000 /* I removed check for oif == dev_out->oif here.
2001 It was wrong for two reasons:
2002 1. ip_dev_find(net, saddr) can return wrong iface, if saddr
2003 is assigned to multiple interfaces.
2004 2. Moreover, we are allowed to send packets with saddr
2005 of another iface. --ANK
2008 if (fl4
->flowi4_oif
== 0 &&
2009 (ipv4_is_multicast(fl4
->daddr
) ||
2010 ipv4_is_lbcast(fl4
->daddr
))) {
2011 /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2012 dev_out
= __ip_dev_find(net
, fl4
->saddr
, false);
2013 if (dev_out
== NULL
)
2016 /* Special hack: user can direct multicasts
2017 and limited broadcast via necessary interface
2018 without fiddling with IP_MULTICAST_IF or IP_PKTINFO.
2019 This hack is not just for fun, it allows
2020 vic,vat and friends to work.
2021 They bind socket to loopback, set ttl to zero
2022 and expect that it will work.
2023 From the viewpoint of routing cache they are broken,
2024 because we are not allowed to build multicast path
2025 with loopback source addr (look, routing cache
2026 cannot know, that ttl is zero, so that packet
2027 will not leave this host and route is valid).
2028 Luckily, this hack is good workaround.
2031 fl4
->flowi4_oif
= dev_out
->ifindex
;
2035 if (!(fl4
->flowi4_flags
& FLOWI_FLAG_ANYSRC
)) {
2036 /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2037 if (!__ip_dev_find(net
, fl4
->saddr
, false))
2043 if (fl4
->flowi4_oif
) {
2044 dev_out
= dev_get_by_index_rcu(net
, fl4
->flowi4_oif
);
2045 rth
= ERR_PTR(-ENODEV
);
2046 if (dev_out
== NULL
)
2049 /* RACE: Check return value of inet_select_addr instead. */
2050 if (!(dev_out
->flags
& IFF_UP
) || !__in_dev_get_rcu(dev_out
)) {
2051 rth
= ERR_PTR(-ENETUNREACH
);
2054 if (ipv4_is_local_multicast(fl4
->daddr
) ||
2055 ipv4_is_lbcast(fl4
->daddr
)) {
2057 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2062 if (ipv4_is_multicast(fl4
->daddr
))
2063 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2065 else if (!fl4
->daddr
)
2066 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2072 fl4
->daddr
= fl4
->saddr
;
2074 fl4
->daddr
= fl4
->saddr
= htonl(INADDR_LOOPBACK
);
2075 dev_out
= net
->loopback_dev
;
2076 fl4
->flowi4_oif
= LOOPBACK_IFINDEX
;
2077 res
.type
= RTN_LOCAL
;
2078 flags
|= RTCF_LOCAL
;
2082 if (fib_lookup(net
, fl4
, &res
)) {
2085 if (fl4
->flowi4_oif
) {
2086 /* Apparently, routing tables are wrong. Assume,
2087 that the destination is on link.
2090 Because we are allowed to send to iface
2091 even if it has NO routes and NO assigned
2092 addresses. When oif is specified, routing
2093 tables are looked up with only one purpose:
2094 to catch if destination is gatewayed, rather than
2095 direct. Moreover, if MSG_DONTROUTE is set,
2096 we send packet, ignoring both routing tables
2097 and ifaddr state. --ANK
2100 We could make it even if oif is unknown,
2101 likely IPv6, but we do not.
2104 if (fl4
->saddr
== 0)
2105 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2107 res
.type
= RTN_UNICAST
;
2110 rth
= ERR_PTR(-ENETUNREACH
);
2114 if (res
.type
== RTN_LOCAL
) {
2116 if (res
.fi
->fib_prefsrc
)
2117 fl4
->saddr
= res
.fi
->fib_prefsrc
;
2119 fl4
->saddr
= fl4
->daddr
;
2121 dev_out
= net
->loopback_dev
;
2122 fl4
->flowi4_oif
= dev_out
->ifindex
;
2123 flags
|= RTCF_LOCAL
;
2127 #ifdef CONFIG_IP_ROUTE_MULTIPATH
2128 if (res
.fi
->fib_nhs
> 1 && fl4
->flowi4_oif
== 0)
2129 fib_select_multipath(&res
);
2132 if (!res
.prefixlen
&&
2133 res
.table
->tb_num_default
> 1 &&
2134 res
.type
== RTN_UNICAST
&& !fl4
->flowi4_oif
)
2135 fib_select_default(&res
);
2138 fl4
->saddr
= FIB_RES_PREFSRC(net
, res
);
2140 dev_out
= FIB_RES_DEV(res
);
2141 fl4
->flowi4_oif
= dev_out
->ifindex
;
2145 rth
= __mkroute_output(&res
, fl4
, orig_oif
, dev_out
, flags
);
2151 EXPORT_SYMBOL_GPL(__ip_route_output_key
);
2153 static struct dst_entry
*ipv4_blackhole_dst_check(struct dst_entry
*dst
, u32 cookie
)
2158 static unsigned int ipv4_blackhole_mtu(const struct dst_entry
*dst
)
2160 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
2162 return mtu
? : dst
->dev
->mtu
;
2165 static void ipv4_rt_blackhole_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
2166 struct sk_buff
*skb
, u32 mtu
)
2170 static void ipv4_rt_blackhole_redirect(struct dst_entry
*dst
, struct sock
*sk
,
2171 struct sk_buff
*skb
)
2175 static u32
*ipv4_rt_blackhole_cow_metrics(struct dst_entry
*dst
,
2181 static struct dst_ops ipv4_dst_blackhole_ops
= {
2183 .protocol
= cpu_to_be16(ETH_P_IP
),
2184 .check
= ipv4_blackhole_dst_check
,
2185 .mtu
= ipv4_blackhole_mtu
,
2186 .default_advmss
= ipv4_default_advmss
,
2187 .update_pmtu
= ipv4_rt_blackhole_update_pmtu
,
2188 .redirect
= ipv4_rt_blackhole_redirect
,
2189 .cow_metrics
= ipv4_rt_blackhole_cow_metrics
,
2190 .neigh_lookup
= ipv4_neigh_lookup
,
2193 struct dst_entry
*ipv4_blackhole_route(struct net
*net
, struct dst_entry
*dst_orig
)
2195 struct rtable
*ort
= (struct rtable
*) dst_orig
;
2198 rt
= dst_alloc(&ipv4_dst_blackhole_ops
, NULL
, 1, DST_OBSOLETE_NONE
, 0);
2200 struct dst_entry
*new = &rt
->dst
;
2203 new->input
= dst_discard
;
2204 new->output
= dst_discard
;
2206 new->dev
= ort
->dst
.dev
;
2210 rt
->rt_is_input
= ort
->rt_is_input
;
2211 rt
->rt_iif
= ort
->rt_iif
;
2212 rt
->rt_pmtu
= ort
->rt_pmtu
;
2214 rt
->rt_genid
= rt_genid(net
);
2215 rt
->rt_flags
= ort
->rt_flags
;
2216 rt
->rt_type
= ort
->rt_type
;
2217 rt
->rt_gateway
= ort
->rt_gateway
;
2218 rt
->rt_uses_gateway
= ort
->rt_uses_gateway
;
2220 INIT_LIST_HEAD(&rt
->rt_uncached
);
2225 dst_release(dst_orig
);
2227 return rt
? &rt
->dst
: ERR_PTR(-ENOMEM
);
2230 struct rtable
*ip_route_output_flow(struct net
*net
, struct flowi4
*flp4
,
2233 struct rtable
*rt
= __ip_route_output_key(net
, flp4
);
2238 if (flp4
->flowi4_proto
)
2239 rt
= (struct rtable
*) xfrm_lookup(net
, &rt
->dst
,
2240 flowi4_to_flowi(flp4
),
2245 EXPORT_SYMBOL_GPL(ip_route_output_flow
);
2247 static int rt_fill_info(struct net
*net
, __be32 dst
, __be32 src
,
2248 struct flowi4
*fl4
, struct sk_buff
*skb
, u32 portid
,
2249 u32 seq
, int event
, int nowait
, unsigned int flags
)
2251 struct rtable
*rt
= skb_rtable(skb
);
2253 struct nlmsghdr
*nlh
;
2254 unsigned long expires
= 0;
2256 u32 metrics
[RTAX_MAX
];
2258 nlh
= nlmsg_put(skb
, portid
, seq
, event
, sizeof(*r
), flags
);
2262 r
= nlmsg_data(nlh
);
2263 r
->rtm_family
= AF_INET
;
2264 r
->rtm_dst_len
= 32;
2266 r
->rtm_tos
= fl4
->flowi4_tos
;
2267 r
->rtm_table
= RT_TABLE_MAIN
;
2268 if (nla_put_u32(skb
, RTA_TABLE
, RT_TABLE_MAIN
))
2269 goto nla_put_failure
;
2270 r
->rtm_type
= rt
->rt_type
;
2271 r
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2272 r
->rtm_protocol
= RTPROT_UNSPEC
;
2273 r
->rtm_flags
= (rt
->rt_flags
& ~0xFFFF) | RTM_F_CLONED
;
2274 if (rt
->rt_flags
& RTCF_NOTIFY
)
2275 r
->rtm_flags
|= RTM_F_NOTIFY
;
2276 if (IPCB(skb
)->flags
& IPSKB_DOREDIRECT
)
2277 r
->rtm_flags
|= RTCF_DOREDIRECT
;
2279 if (nla_put_be32(skb
, RTA_DST
, dst
))
2280 goto nla_put_failure
;
2282 r
->rtm_src_len
= 32;
2283 if (nla_put_be32(skb
, RTA_SRC
, src
))
2284 goto nla_put_failure
;
2287 nla_put_u32(skb
, RTA_OIF
, rt
->dst
.dev
->ifindex
))
2288 goto nla_put_failure
;
2289 #ifdef CONFIG_IP_ROUTE_CLASSID
2290 if (rt
->dst
.tclassid
&&
2291 nla_put_u32(skb
, RTA_FLOW
, rt
->dst
.tclassid
))
2292 goto nla_put_failure
;
2294 if (!rt_is_input_route(rt
) &&
2295 fl4
->saddr
!= src
) {
2296 if (nla_put_be32(skb
, RTA_PREFSRC
, fl4
->saddr
))
2297 goto nla_put_failure
;
2299 if (rt
->rt_uses_gateway
&&
2300 nla_put_be32(skb
, RTA_GATEWAY
, rt
->rt_gateway
))
2301 goto nla_put_failure
;
2303 expires
= rt
->dst
.expires
;
2305 unsigned long now
= jiffies
;
2307 if (time_before(now
, expires
))
2313 memcpy(metrics
, dst_metrics_ptr(&rt
->dst
), sizeof(metrics
));
2314 if (rt
->rt_pmtu
&& expires
)
2315 metrics
[RTAX_MTU
- 1] = rt
->rt_pmtu
;
2316 if (rtnetlink_put_metrics(skb
, metrics
) < 0)
2317 goto nla_put_failure
;
2319 if (fl4
->flowi4_mark
&&
2320 nla_put_u32(skb
, RTA_MARK
, fl4
->flowi4_mark
))
2321 goto nla_put_failure
;
2323 error
= rt
->dst
.error
;
2325 if (rt_is_input_route(rt
)) {
2326 #ifdef CONFIG_IP_MROUTE
2327 if (ipv4_is_multicast(dst
) && !ipv4_is_local_multicast(dst
) &&
2328 IPV4_DEVCONF_ALL(net
, MC_FORWARDING
)) {
2329 int err
= ipmr_get_route(net
, skb
,
2330 fl4
->saddr
, fl4
->daddr
,
2337 goto nla_put_failure
;
2339 if (err
== -EMSGSIZE
)
2340 goto nla_put_failure
;
2346 if (nla_put_u32(skb
, RTA_IIF
, skb
->dev
->ifindex
))
2347 goto nla_put_failure
;
2350 if (rtnl_put_cacheinfo(skb
, &rt
->dst
, 0, expires
, error
) < 0)
2351 goto nla_put_failure
;
2353 return nlmsg_end(skb
, nlh
);
2356 nlmsg_cancel(skb
, nlh
);
2360 static int inet_rtm_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
*nlh
)
2362 struct net
*net
= sock_net(in_skb
->sk
);
2364 struct nlattr
*tb
[RTA_MAX
+1];
2365 struct rtable
*rt
= NULL
;
2372 struct sk_buff
*skb
;
2374 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv4_policy
);
2378 rtm
= nlmsg_data(nlh
);
2380 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
2386 /* Reserve room for dummy headers, this skb can pass
2387 through good chunk of routing engine.
2389 skb_reset_mac_header(skb
);
2390 skb_reset_network_header(skb
);
2392 /* Bugfix: need to give ip_route_input enough of an IP header to not gag. */
2393 ip_hdr(skb
)->protocol
= IPPROTO_ICMP
;
2394 skb_reserve(skb
, MAX_HEADER
+ sizeof(struct iphdr
));
2396 src
= tb
[RTA_SRC
] ? nla_get_be32(tb
[RTA_SRC
]) : 0;
2397 dst
= tb
[RTA_DST
] ? nla_get_be32(tb
[RTA_DST
]) : 0;
2398 iif
= tb
[RTA_IIF
] ? nla_get_u32(tb
[RTA_IIF
]) : 0;
2399 mark
= tb
[RTA_MARK
] ? nla_get_u32(tb
[RTA_MARK
]) : 0;
2401 memset(&fl4
, 0, sizeof(fl4
));
2404 fl4
.flowi4_tos
= rtm
->rtm_tos
;
2405 fl4
.flowi4_oif
= tb
[RTA_OIF
] ? nla_get_u32(tb
[RTA_OIF
]) : 0;
2406 fl4
.flowi4_mark
= mark
;
2409 struct net_device
*dev
;
2411 dev
= __dev_get_by_index(net
, iif
);
2417 skb
->protocol
= htons(ETH_P_IP
);
2421 err
= ip_route_input(skb
, dst
, src
, rtm
->rtm_tos
, dev
);
2424 rt
= skb_rtable(skb
);
2425 if (err
== 0 && rt
->dst
.error
)
2426 err
= -rt
->dst
.error
;
2428 rt
= ip_route_output_key(net
, &fl4
);
2438 skb_dst_set(skb
, &rt
->dst
);
2439 if (rtm
->rtm_flags
& RTM_F_NOTIFY
)
2440 rt
->rt_flags
|= RTCF_NOTIFY
;
2442 err
= rt_fill_info(net
, dst
, src
, &fl4
, skb
,
2443 NETLINK_CB(in_skb
).portid
, nlh
->nlmsg_seq
,
2444 RTM_NEWROUTE
, 0, 0);
2448 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).portid
);
2457 int ip_rt_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2462 void ip_rt_multicast_event(struct in_device
*in_dev
)
2464 rt_cache_flush(dev_net(in_dev
->dev
));
2467 #ifdef CONFIG_SYSCTL
2468 static int ip_rt_gc_timeout __read_mostly
= RT_GC_TIMEOUT
;
2469 static int ip_rt_gc_interval __read_mostly
= 60 * HZ
;
2470 static int ip_rt_gc_min_interval __read_mostly
= HZ
/ 2;
2471 static int ip_rt_gc_elasticity __read_mostly
= 8;
2473 static int ipv4_sysctl_rtcache_flush(ctl_table
*__ctl
, int write
,
2474 void __user
*buffer
,
2475 size_t *lenp
, loff_t
*ppos
)
2478 rt_cache_flush((struct net
*)__ctl
->extra1
);
2485 static ctl_table ipv4_route_table
[] = {
2487 .procname
= "gc_thresh",
2488 .data
= &ipv4_dst_ops
.gc_thresh
,
2489 .maxlen
= sizeof(int),
2491 .proc_handler
= proc_dointvec
,
2494 .procname
= "max_size",
2495 .data
= &ip_rt_max_size
,
2496 .maxlen
= sizeof(int),
2498 .proc_handler
= proc_dointvec
,
2501 /* Deprecated. Use gc_min_interval_ms */
2503 .procname
= "gc_min_interval",
2504 .data
= &ip_rt_gc_min_interval
,
2505 .maxlen
= sizeof(int),
2507 .proc_handler
= proc_dointvec_jiffies
,
2510 .procname
= "gc_min_interval_ms",
2511 .data
= &ip_rt_gc_min_interval
,
2512 .maxlen
= sizeof(int),
2514 .proc_handler
= proc_dointvec_ms_jiffies
,
2517 .procname
= "gc_timeout",
2518 .data
= &ip_rt_gc_timeout
,
2519 .maxlen
= sizeof(int),
2521 .proc_handler
= proc_dointvec_jiffies
,
2524 .procname
= "gc_interval",
2525 .data
= &ip_rt_gc_interval
,
2526 .maxlen
= sizeof(int),
2528 .proc_handler
= proc_dointvec_jiffies
,
2531 .procname
= "redirect_load",
2532 .data
= &ip_rt_redirect_load
,
2533 .maxlen
= sizeof(int),
2535 .proc_handler
= proc_dointvec
,
2538 .procname
= "redirect_number",
2539 .data
= &ip_rt_redirect_number
,
2540 .maxlen
= sizeof(int),
2542 .proc_handler
= proc_dointvec
,
2545 .procname
= "redirect_silence",
2546 .data
= &ip_rt_redirect_silence
,
2547 .maxlen
= sizeof(int),
2549 .proc_handler
= proc_dointvec
,
2552 .procname
= "error_cost",
2553 .data
= &ip_rt_error_cost
,
2554 .maxlen
= sizeof(int),
2556 .proc_handler
= proc_dointvec
,
2559 .procname
= "error_burst",
2560 .data
= &ip_rt_error_burst
,
2561 .maxlen
= sizeof(int),
2563 .proc_handler
= proc_dointvec
,
2566 .procname
= "gc_elasticity",
2567 .data
= &ip_rt_gc_elasticity
,
2568 .maxlen
= sizeof(int),
2570 .proc_handler
= proc_dointvec
,
2573 .procname
= "mtu_expires",
2574 .data
= &ip_rt_mtu_expires
,
2575 .maxlen
= sizeof(int),
2577 .proc_handler
= proc_dointvec_jiffies
,
2580 .procname
= "min_pmtu",
2581 .data
= &ip_rt_min_pmtu
,
2582 .maxlen
= sizeof(int),
2584 .proc_handler
= proc_dointvec
,
2587 .procname
= "min_adv_mss",
2588 .data
= &ip_rt_min_advmss
,
2589 .maxlen
= sizeof(int),
2591 .proc_handler
= proc_dointvec
,
2596 static struct ctl_table ipv4_route_flush_table
[] = {
2598 .procname
= "flush",
2599 .maxlen
= sizeof(int),
2601 .proc_handler
= ipv4_sysctl_rtcache_flush
,
2606 static __net_init
int sysctl_route_net_init(struct net
*net
)
2608 struct ctl_table
*tbl
;
2610 tbl
= ipv4_route_flush_table
;
2611 if (!net_eq(net
, &init_net
)) {
2612 tbl
= kmemdup(tbl
, sizeof(ipv4_route_flush_table
), GFP_KERNEL
);
2616 /* Don't export sysctls to unprivileged users */
2617 if (net
->user_ns
!= &init_user_ns
)
2618 tbl
[0].procname
= NULL
;
2620 tbl
[0].extra1
= net
;
2622 net
->ipv4
.route_hdr
= register_net_sysctl(net
, "net/ipv4/route", tbl
);
2623 if (net
->ipv4
.route_hdr
== NULL
)
2628 if (tbl
!= ipv4_route_flush_table
)
2634 static __net_exit
void sysctl_route_net_exit(struct net
*net
)
2636 struct ctl_table
*tbl
;
2638 tbl
= net
->ipv4
.route_hdr
->ctl_table_arg
;
2639 unregister_net_sysctl_table(net
->ipv4
.route_hdr
);
2640 BUG_ON(tbl
== ipv4_route_flush_table
);
2644 static __net_initdata
struct pernet_operations sysctl_route_ops
= {
2645 .init
= sysctl_route_net_init
,
2646 .exit
= sysctl_route_net_exit
,
2650 static __net_init
int rt_genid_init(struct net
*net
)
2652 atomic_set(&net
->rt_genid
, 0);
2653 get_random_bytes(&net
->ipv4
.dev_addr_genid
,
2654 sizeof(net
->ipv4
.dev_addr_genid
));
2658 static __net_initdata
struct pernet_operations rt_genid_ops
= {
2659 .init
= rt_genid_init
,
2662 static int __net_init
ipv4_inetpeer_init(struct net
*net
)
2664 struct inet_peer_base
*bp
= kmalloc(sizeof(*bp
), GFP_KERNEL
);
2668 inet_peer_base_init(bp
);
2669 net
->ipv4
.peers
= bp
;
2673 static void __net_exit
ipv4_inetpeer_exit(struct net
*net
)
2675 struct inet_peer_base
*bp
= net
->ipv4
.peers
;
2677 net
->ipv4
.peers
= NULL
;
2678 inetpeer_invalidate_tree(bp
);
2682 static __net_initdata
struct pernet_operations ipv4_inetpeer_ops
= {
2683 .init
= ipv4_inetpeer_init
,
2684 .exit
= ipv4_inetpeer_exit
,
2687 #ifdef CONFIG_IP_ROUTE_CLASSID
2688 struct ip_rt_acct __percpu
*ip_rt_acct __read_mostly
;
2689 #endif /* CONFIG_IP_ROUTE_CLASSID */
2691 int __init
ip_rt_init(void)
2695 ip_idents
= kmalloc(IP_IDENTS_SZ
* sizeof(*ip_idents
), GFP_KERNEL
);
2697 panic("IP: failed to allocate ip_idents\n");
2699 prandom_bytes(ip_idents
, IP_IDENTS_SZ
* sizeof(*ip_idents
));
2701 #ifdef CONFIG_IP_ROUTE_CLASSID
2702 ip_rt_acct
= __alloc_percpu(256 * sizeof(struct ip_rt_acct
), __alignof__(struct ip_rt_acct
));
2704 panic("IP: failed to allocate ip_rt_acct\n");
2707 ipv4_dst_ops
.kmem_cachep
=
2708 kmem_cache_create("ip_dst_cache", sizeof(struct rtable
), 0,
2709 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
);
2711 ipv4_dst_blackhole_ops
.kmem_cachep
= ipv4_dst_ops
.kmem_cachep
;
2713 if (dst_entries_init(&ipv4_dst_ops
) < 0)
2714 panic("IP: failed to allocate ipv4_dst_ops counter\n");
2716 if (dst_entries_init(&ipv4_dst_blackhole_ops
) < 0)
2717 panic("IP: failed to allocate ipv4_dst_blackhole_ops counter\n");
2719 ipv4_dst_ops
.gc_thresh
= ~0;
2720 ip_rt_max_size
= INT_MAX
;
2725 if (ip_rt_proc_init())
2726 pr_err("Unable to create route proc files\n");
2731 rtnl_register(PF_INET
, RTM_GETROUTE
, inet_rtm_getroute
, NULL
, NULL
);
2733 #ifdef CONFIG_SYSCTL
2734 register_pernet_subsys(&sysctl_route_ops
);
2736 register_pernet_subsys(&rt_genid_ops
);
2737 register_pernet_subsys(&ipv4_inetpeer_ops
);
2741 #ifdef CONFIG_SYSCTL
2743 * We really need to sanitize the damn ipv4 init order, then all
2744 * this nonsense will go away.
2746 void __init
ip_static_sysctl_init(void)
2748 register_net_sysctl(&init_net
, "net/ipv4/route", ipv4_route_table
);