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
, 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,
535 sk
? sock_i_uid(sk
) : 0);
538 static void build_skb_flow_key(struct flowi4
*fl4
, const struct sk_buff
*skb
,
541 const struct iphdr
*iph
= ip_hdr(skb
);
542 int oif
= skb
->dev
->ifindex
;
543 u8 tos
= RT_TOS(iph
->tos
);
544 u8 prot
= iph
->protocol
;
545 u32 mark
= skb
->mark
;
547 __build_flow_key(fl4
, sk
, iph
, oif
, tos
, prot
, mark
, 0);
550 static void build_sk_flow_key(struct flowi4
*fl4
, struct sock
*sk
)
552 const struct inet_sock
*inet
= inet_sk(sk
);
553 const struct ip_options_rcu
*inet_opt
;
554 __be32 daddr
= inet
->inet_daddr
;
557 inet_opt
= rcu_dereference(inet
->inet_opt
);
558 if (inet_opt
&& inet_opt
->opt
.srr
)
559 daddr
= inet_opt
->opt
.faddr
;
560 flowi4_init_output(fl4
, sk
->sk_bound_dev_if
, sk
->sk_mark
,
561 RT_CONN_FLAGS(sk
), RT_SCOPE_UNIVERSE
,
562 inet
->hdrincl
? IPPROTO_RAW
: sk
->sk_protocol
,
563 inet_sk_flowi_flags(sk
),
564 daddr
, inet
->inet_saddr
, 0, 0,
569 static void ip_rt_build_flow_key(struct flowi4
*fl4
, 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(&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(&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(&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
;
1022 odst
= sk_dst_get(sk
);
1024 if (sock_owned_by_user(sk
) || !odst
) {
1025 __ipv4_sk_update_pmtu(skb
, sk
, mtu
);
1029 __build_flow_key(&fl4
, sk
, iph
, 0, 0, 0, 0, 0);
1031 rt
= (struct rtable
*)odst
;
1032 if (odst
->obsolete
&& odst
->ops
->check(odst
, 0) == NULL
) {
1033 rt
= ip_route_output_flow(sock_net(sk
), &fl4
, sk
);
1040 __ip_rt_update_pmtu((struct rtable
*) rt
->dst
.path
, &fl4
, mtu
);
1042 if (!dst_check(&rt
->dst
, 0)) {
1044 dst_release(&rt
->dst
);
1046 rt
= ip_route_output_flow(sock_net(sk
), &fl4
, sk
);
1054 sk_dst_set(sk
, &rt
->dst
);
1060 EXPORT_SYMBOL_GPL(ipv4_sk_update_pmtu
);
1062 void ipv4_redirect(struct sk_buff
*skb
, struct net
*net
,
1063 int oif
, u32 mark
, u8 protocol
, int flow_flags
)
1065 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1069 __build_flow_key(&fl4
, NULL
, iph
, oif
,
1070 RT_TOS(iph
->tos
), protocol
, mark
, flow_flags
);
1071 rt
= __ip_route_output_key(net
, &fl4
);
1073 __ip_do_redirect(rt
, skb
, &fl4
, false);
1077 EXPORT_SYMBOL_GPL(ipv4_redirect
);
1079 void ipv4_sk_redirect(struct sk_buff
*skb
, struct sock
*sk
)
1081 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1085 __build_flow_key(&fl4
, sk
, iph
, 0, 0, 0, 0, 0);
1086 rt
= __ip_route_output_key(sock_net(sk
), &fl4
);
1088 __ip_do_redirect(rt
, skb
, &fl4
, false);
1092 EXPORT_SYMBOL_GPL(ipv4_sk_redirect
);
1094 static struct dst_entry
*ipv4_dst_check(struct dst_entry
*dst
, u32 cookie
)
1096 struct rtable
*rt
= (struct rtable
*) dst
;
1098 /* All IPV4 dsts are created with ->obsolete set to the value
1099 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1100 * into this function always.
1102 * When a PMTU/redirect information update invalidates a
1103 * route, this is indicated by setting obsolete to
1104 * DST_OBSOLETE_KILL.
1106 if (dst
->obsolete
== DST_OBSOLETE_KILL
|| rt_is_expired(rt
))
1111 static void ipv4_link_failure(struct sk_buff
*skb
)
1115 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_HOST_UNREACH
, 0);
1117 rt
= skb_rtable(skb
);
1119 dst_set_expires(&rt
->dst
, 0);
1122 static int ip_rt_bug(struct sk_buff
*skb
)
1124 pr_debug("%s: %pI4 -> %pI4, %s\n",
1125 __func__
, &ip_hdr(skb
)->saddr
, &ip_hdr(skb
)->daddr
,
1126 skb
->dev
? skb
->dev
->name
: "?");
1133 We do not cache source address of outgoing interface,
1134 because it is used only by IP RR, TS and SRR options,
1135 so that it out of fast path.
1137 BTW remember: "addr" is allowed to be not aligned
1141 void ip_rt_get_source(u8
*addr
, struct sk_buff
*skb
, struct rtable
*rt
)
1145 if (rt_is_output_route(rt
))
1146 src
= ip_hdr(skb
)->saddr
;
1148 struct fib_result res
;
1154 memset(&fl4
, 0, sizeof(fl4
));
1155 fl4
.daddr
= iph
->daddr
;
1156 fl4
.saddr
= iph
->saddr
;
1157 fl4
.flowi4_tos
= RT_TOS(iph
->tos
);
1158 fl4
.flowi4_oif
= rt
->dst
.dev
->ifindex
;
1159 fl4
.flowi4_iif
= skb
->dev
->ifindex
;
1160 fl4
.flowi4_mark
= skb
->mark
;
1163 if (fib_lookup(dev_net(rt
->dst
.dev
), &fl4
, &res
) == 0)
1164 src
= FIB_RES_PREFSRC(dev_net(rt
->dst
.dev
), res
);
1166 src
= inet_select_addr(rt
->dst
.dev
,
1167 rt_nexthop(rt
, iph
->daddr
),
1171 memcpy(addr
, &src
, 4);
1174 #ifdef CONFIG_IP_ROUTE_CLASSID
1175 static void set_class_tag(struct rtable
*rt
, u32 tag
)
1177 if (!(rt
->dst
.tclassid
& 0xFFFF))
1178 rt
->dst
.tclassid
|= tag
& 0xFFFF;
1179 if (!(rt
->dst
.tclassid
& 0xFFFF0000))
1180 rt
->dst
.tclassid
|= tag
& 0xFFFF0000;
1184 static unsigned int ipv4_default_advmss(const struct dst_entry
*dst
)
1186 unsigned int advmss
= dst_metric_raw(dst
, RTAX_ADVMSS
);
1189 advmss
= max_t(unsigned int, dst
->dev
->mtu
- 40,
1191 if (advmss
> 65535 - 40)
1192 advmss
= 65535 - 40;
1197 static unsigned int ipv4_mtu(const struct dst_entry
*dst
)
1199 const struct rtable
*rt
= (const struct rtable
*) dst
;
1200 unsigned int mtu
= rt
->rt_pmtu
;
1202 if (!mtu
|| time_after_eq(jiffies
, rt
->dst
.expires
))
1203 mtu
= dst_metric_raw(dst
, RTAX_MTU
);
1208 mtu
= dst
->dev
->mtu
;
1210 if (unlikely(dst_metric_locked(dst
, RTAX_MTU
))) {
1211 if (rt
->rt_uses_gateway
&& mtu
> 576)
1215 if (mtu
> IP_MAX_MTU
)
1221 static struct fib_nh_exception
*find_exception(struct fib_nh
*nh
, __be32 daddr
)
1223 struct fnhe_hash_bucket
*hash
= nh
->nh_exceptions
;
1224 struct fib_nh_exception
*fnhe
;
1230 hval
= fnhe_hashfun(daddr
);
1232 for (fnhe
= rcu_dereference(hash
[hval
].chain
); fnhe
;
1233 fnhe
= rcu_dereference(fnhe
->fnhe_next
)) {
1234 if (fnhe
->fnhe_daddr
== daddr
)
1240 static bool rt_bind_exception(struct rtable
*rt
, struct fib_nh_exception
*fnhe
,
1245 spin_lock_bh(&fnhe_lock
);
1247 if (daddr
== fnhe
->fnhe_daddr
) {
1248 struct rtable
*orig
= rcu_dereference(fnhe
->fnhe_rth
);
1249 if (orig
&& rt_is_expired(orig
)) {
1251 fnhe
->fnhe_pmtu
= 0;
1252 fnhe
->fnhe_expires
= 0;
1254 if (fnhe
->fnhe_pmtu
) {
1255 unsigned long expires
= fnhe
->fnhe_expires
;
1256 unsigned long diff
= expires
- jiffies
;
1258 if (time_before(jiffies
, expires
)) {
1259 rt
->rt_pmtu
= fnhe
->fnhe_pmtu
;
1260 dst_set_expires(&rt
->dst
, diff
);
1263 if (fnhe
->fnhe_gw
) {
1264 rt
->rt_flags
|= RTCF_REDIRECTED
;
1265 rt
->rt_gateway
= fnhe
->fnhe_gw
;
1266 rt
->rt_uses_gateway
= 1;
1267 } else if (!rt
->rt_gateway
)
1268 rt
->rt_gateway
= daddr
;
1270 rcu_assign_pointer(fnhe
->fnhe_rth
, rt
);
1274 fnhe
->fnhe_stamp
= jiffies
;
1277 spin_unlock_bh(&fnhe_lock
);
1282 static bool rt_cache_route(struct fib_nh
*nh
, struct rtable
*rt
)
1284 struct rtable
*orig
, *prev
, **p
;
1287 if (rt_is_input_route(rt
)) {
1288 p
= (struct rtable
**)&nh
->nh_rth_input
;
1290 p
= (struct rtable
**)__this_cpu_ptr(nh
->nh_pcpu_rth_output
);
1294 prev
= cmpxchg(p
, orig
, rt
);
1304 static DEFINE_SPINLOCK(rt_uncached_lock
);
1305 static LIST_HEAD(rt_uncached_list
);
1307 static void rt_add_uncached_list(struct rtable
*rt
)
1309 spin_lock_bh(&rt_uncached_lock
);
1310 list_add_tail(&rt
->rt_uncached
, &rt_uncached_list
);
1311 spin_unlock_bh(&rt_uncached_lock
);
1314 static void ipv4_dst_destroy(struct dst_entry
*dst
)
1316 struct rtable
*rt
= (struct rtable
*) dst
;
1318 if (!list_empty(&rt
->rt_uncached
)) {
1319 spin_lock_bh(&rt_uncached_lock
);
1320 list_del(&rt
->rt_uncached
);
1321 spin_unlock_bh(&rt_uncached_lock
);
1325 void rt_flush_dev(struct net_device
*dev
)
1327 if (!list_empty(&rt_uncached_list
)) {
1328 struct net
*net
= dev_net(dev
);
1331 spin_lock_bh(&rt_uncached_lock
);
1332 list_for_each_entry(rt
, &rt_uncached_list
, rt_uncached
) {
1333 if (rt
->dst
.dev
!= dev
)
1335 rt
->dst
.dev
= net
->loopback_dev
;
1336 dev_hold(rt
->dst
.dev
);
1339 spin_unlock_bh(&rt_uncached_lock
);
1343 static bool rt_cache_valid(const struct rtable
*rt
)
1346 rt
->dst
.obsolete
== DST_OBSOLETE_FORCE_CHK
&&
1350 static void rt_set_nexthop(struct rtable
*rt
, __be32 daddr
,
1351 const struct fib_result
*res
,
1352 struct fib_nh_exception
*fnhe
,
1353 struct fib_info
*fi
, u16 type
, u32 itag
)
1355 bool cached
= false;
1358 struct fib_nh
*nh
= &FIB_RES_NH(*res
);
1360 if (nh
->nh_gw
&& nh
->nh_scope
== RT_SCOPE_LINK
) {
1361 rt
->rt_gateway
= nh
->nh_gw
;
1362 rt
->rt_uses_gateway
= 1;
1364 dst_init_metrics(&rt
->dst
, fi
->fib_metrics
, true);
1365 #ifdef CONFIG_IP_ROUTE_CLASSID
1366 rt
->dst
.tclassid
= nh
->nh_tclassid
;
1369 cached
= rt_bind_exception(rt
, fnhe
, daddr
);
1370 else if (!(rt
->dst
.flags
& DST_NOCACHE
))
1371 cached
= rt_cache_route(nh
, rt
);
1372 if (unlikely(!cached
)) {
1373 /* Routes we intend to cache in nexthop exception or
1374 * FIB nexthop have the DST_NOCACHE bit clear.
1375 * However, if we are unsuccessful at storing this
1376 * route into the cache we really need to set it.
1378 rt
->dst
.flags
|= DST_NOCACHE
;
1379 if (!rt
->rt_gateway
)
1380 rt
->rt_gateway
= daddr
;
1381 rt_add_uncached_list(rt
);
1384 rt_add_uncached_list(rt
);
1386 #ifdef CONFIG_IP_ROUTE_CLASSID
1387 #ifdef CONFIG_IP_MULTIPLE_TABLES
1388 set_class_tag(rt
, res
->tclassid
);
1390 set_class_tag(rt
, itag
);
1394 static struct rtable
*rt_dst_alloc(struct net_device
*dev
,
1395 bool nopolicy
, bool noxfrm
, bool will_cache
)
1397 return dst_alloc(&ipv4_dst_ops
, dev
, 1, DST_OBSOLETE_FORCE_CHK
,
1398 (will_cache
? 0 : (DST_HOST
| DST_NOCACHE
)) |
1399 (nopolicy
? DST_NOPOLICY
: 0) |
1400 (noxfrm
? DST_NOXFRM
: 0));
1403 /* called in rcu_read_lock() section */
1404 static int ip_route_input_mc(struct sk_buff
*skb
, __be32 daddr
, __be32 saddr
,
1405 u8 tos
, struct net_device
*dev
, int our
)
1408 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
1412 /* Primary sanity checks. */
1417 if (ipv4_is_multicast(saddr
) || ipv4_is_lbcast(saddr
) ||
1418 skb
->protocol
!= htons(ETH_P_IP
))
1421 if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev
)))
1422 if (ipv4_is_loopback(saddr
))
1425 if (ipv4_is_zeronet(saddr
)) {
1426 if (!ipv4_is_local_multicast(daddr
))
1429 err
= fib_validate_source(skb
, saddr
, 0, tos
, 0, dev
,
1434 rth
= rt_dst_alloc(dev_net(dev
)->loopback_dev
,
1435 IN_DEV_CONF_GET(in_dev
, NOPOLICY
), false, false);
1439 #ifdef CONFIG_IP_ROUTE_CLASSID
1440 rth
->dst
.tclassid
= itag
;
1442 rth
->dst
.output
= ip_rt_bug
;
1444 rth
->rt_genid
= rt_genid(dev_net(dev
));
1445 rth
->rt_flags
= RTCF_MULTICAST
;
1446 rth
->rt_type
= RTN_MULTICAST
;
1447 rth
->rt_is_input
= 1;
1450 rth
->rt_gateway
= 0;
1451 rth
->rt_uses_gateway
= 0;
1452 INIT_LIST_HEAD(&rth
->rt_uncached
);
1454 rth
->dst
.input
= ip_local_deliver
;
1455 rth
->rt_flags
|= RTCF_LOCAL
;
1458 #ifdef CONFIG_IP_MROUTE
1459 if (!ipv4_is_local_multicast(daddr
) && IN_DEV_MFORWARD(in_dev
))
1460 rth
->dst
.input
= ip_mr_input
;
1462 RT_CACHE_STAT_INC(in_slow_mc
);
1464 skb_dst_set(skb
, &rth
->dst
);
1476 static void ip_handle_martian_source(struct net_device
*dev
,
1477 struct in_device
*in_dev
,
1478 struct sk_buff
*skb
,
1482 RT_CACHE_STAT_INC(in_martian_src
);
1483 #ifdef CONFIG_IP_ROUTE_VERBOSE
1484 if (IN_DEV_LOG_MARTIANS(in_dev
) && net_ratelimit()) {
1486 * RFC1812 recommendation, if source is martian,
1487 * the only hint is MAC header.
1489 pr_warn("martian source %pI4 from %pI4, on dev %s\n",
1490 &daddr
, &saddr
, dev
->name
);
1491 if (dev
->hard_header_len
&& skb_mac_header_was_set(skb
)) {
1492 print_hex_dump(KERN_WARNING
, "ll header: ",
1493 DUMP_PREFIX_OFFSET
, 16, 1,
1494 skb_mac_header(skb
),
1495 dev
->hard_header_len
, true);
1501 /* called in rcu_read_lock() section */
1502 static int __mkroute_input(struct sk_buff
*skb
,
1503 const struct fib_result
*res
,
1504 struct in_device
*in_dev
,
1505 __be32 daddr
, __be32 saddr
, u32 tos
)
1509 struct in_device
*out_dev
;
1510 unsigned int flags
= 0;
1514 /* get a working reference to the output device */
1515 out_dev
= __in_dev_get_rcu(FIB_RES_DEV(*res
));
1516 if (out_dev
== NULL
) {
1517 net_crit_ratelimited("Bug in ip_route_input_slow(). Please report.\n");
1521 err
= fib_validate_source(skb
, saddr
, daddr
, tos
, FIB_RES_OIF(*res
),
1522 in_dev
->dev
, in_dev
, &itag
);
1524 ip_handle_martian_source(in_dev
->dev
, in_dev
, skb
, daddr
,
1530 do_cache
= res
->fi
&& !itag
;
1531 if (out_dev
== in_dev
&& err
&& IN_DEV_TX_REDIRECTS(out_dev
) &&
1532 skb
->protocol
== htons(ETH_P_IP
) &&
1533 (IN_DEV_SHARED_MEDIA(out_dev
) ||
1534 inet_addr_onlink(out_dev
, saddr
, FIB_RES_GW(*res
))))
1535 IPCB(skb
)->flags
|= IPSKB_DOREDIRECT
;
1537 if (skb
->protocol
!= htons(ETH_P_IP
)) {
1538 /* Not IP (i.e. ARP). Do not create route, if it is
1539 * invalid for proxy arp. DNAT routes are always valid.
1541 * Proxy arp feature have been extended to allow, ARP
1542 * replies back to the same interface, to support
1543 * Private VLAN switch technologies. See arp.c.
1545 if (out_dev
== in_dev
&&
1546 IN_DEV_PROXY_ARP_PVLAN(in_dev
) == 0) {
1553 rth
= rcu_dereference(FIB_RES_NH(*res
).nh_rth_input
);
1554 if (rt_cache_valid(rth
)) {
1555 skb_dst_set_noref(skb
, &rth
->dst
);
1560 rth
= rt_dst_alloc(out_dev
->dev
,
1561 IN_DEV_CONF_GET(in_dev
, NOPOLICY
),
1562 IN_DEV_CONF_GET(out_dev
, NOXFRM
), do_cache
);
1568 rth
->rt_genid
= rt_genid(dev_net(rth
->dst
.dev
));
1569 rth
->rt_flags
= flags
;
1570 rth
->rt_type
= res
->type
;
1571 rth
->rt_is_input
= 1;
1574 rth
->rt_gateway
= 0;
1575 rth
->rt_uses_gateway
= 0;
1576 INIT_LIST_HEAD(&rth
->rt_uncached
);
1577 RT_CACHE_STAT_INC(in_slow_tot
);
1579 rth
->dst
.input
= ip_forward
;
1580 rth
->dst
.output
= ip_output
;
1582 rt_set_nexthop(rth
, daddr
, res
, NULL
, res
->fi
, res
->type
, itag
);
1583 skb_dst_set(skb
, &rth
->dst
);
1590 static int ip_mkroute_input(struct sk_buff
*skb
,
1591 struct fib_result
*res
,
1592 const struct flowi4
*fl4
,
1593 struct in_device
*in_dev
,
1594 __be32 daddr
, __be32 saddr
, u32 tos
)
1596 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1597 if (res
->fi
&& res
->fi
->fib_nhs
> 1)
1598 fib_select_multipath(res
);
1601 /* create a routing cache entry */
1602 return __mkroute_input(skb
, res
, in_dev
, daddr
, saddr
, tos
);
1606 * NOTE. We drop all the packets that has local source
1607 * addresses, because every properly looped back packet
1608 * must have correct destination already attached by output routine.
1610 * Such approach solves two big problems:
1611 * 1. Not simplex devices are handled properly.
1612 * 2. IP spoofing attempts are filtered with 100% of guarantee.
1613 * called with rcu_read_lock()
1616 static int ip_route_input_slow(struct sk_buff
*skb
, __be32 daddr
, __be32 saddr
,
1617 u8 tos
, struct net_device
*dev
)
1619 struct fib_result res
;
1620 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
1622 unsigned int flags
= 0;
1626 struct net
*net
= dev_net(dev
);
1629 /* IP on this device is disabled. */
1634 /* Check for the most weird martians, which can be not detected
1638 if (ipv4_is_multicast(saddr
) || ipv4_is_lbcast(saddr
))
1639 goto martian_source
;
1642 if (ipv4_is_lbcast(daddr
) || (saddr
== 0 && daddr
== 0))
1645 /* Accept zero addresses only to limited broadcast;
1646 * I even do not know to fix it or not. Waiting for complains :-)
1648 if (ipv4_is_zeronet(saddr
))
1649 goto martian_source
;
1651 if (ipv4_is_zeronet(daddr
))
1652 goto martian_destination
;
1654 /* Following code try to avoid calling IN_DEV_NET_ROUTE_LOCALNET(),
1655 * and call it once if daddr or/and saddr are loopback addresses
1657 if (ipv4_is_loopback(daddr
)) {
1658 if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev
, net
))
1659 goto martian_destination
;
1660 } else if (ipv4_is_loopback(saddr
)) {
1661 if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev
, net
))
1662 goto martian_source
;
1666 * Now we are ready to route packet.
1669 fl4
.flowi4_iif
= dev
->ifindex
;
1670 fl4
.flowi4_mark
= skb
->mark
;
1671 fl4
.flowi4_tos
= tos
;
1672 fl4
.flowi4_scope
= RT_SCOPE_UNIVERSE
;
1675 err
= fib_lookup(net
, &fl4
, &res
);
1679 if (res
.type
== RTN_BROADCAST
)
1682 if (res
.type
== RTN_LOCAL
) {
1683 err
= fib_validate_source(skb
, saddr
, daddr
, tos
,
1685 dev
, in_dev
, &itag
);
1687 goto martian_source_keep_err
;
1691 if (!IN_DEV_FORWARD(in_dev
))
1693 if (res
.type
!= RTN_UNICAST
)
1694 goto martian_destination
;
1696 err
= ip_mkroute_input(skb
, &res
, &fl4
, in_dev
, daddr
, saddr
, tos
);
1700 if (skb
->protocol
!= htons(ETH_P_IP
))
1703 if (!ipv4_is_zeronet(saddr
)) {
1704 err
= fib_validate_source(skb
, saddr
, 0, tos
, 0, dev
,
1707 goto martian_source_keep_err
;
1709 flags
|= RTCF_BROADCAST
;
1710 res
.type
= RTN_BROADCAST
;
1711 RT_CACHE_STAT_INC(in_brd
);
1717 rth
= rcu_dereference(FIB_RES_NH(res
).nh_rth_input
);
1718 if (rt_cache_valid(rth
)) {
1719 skb_dst_set_noref(skb
, &rth
->dst
);
1727 rth
= rt_dst_alloc(net
->loopback_dev
,
1728 IN_DEV_CONF_GET(in_dev
, NOPOLICY
), false, do_cache
);
1732 rth
->dst
.input
= ip_local_deliver
;
1733 rth
->dst
.output
= ip_rt_bug
;
1734 #ifdef CONFIG_IP_ROUTE_CLASSID
1735 rth
->dst
.tclassid
= itag
;
1738 rth
->rt_genid
= rt_genid(net
);
1739 rth
->rt_flags
= flags
|RTCF_LOCAL
;
1740 rth
->rt_type
= res
.type
;
1741 rth
->rt_is_input
= 1;
1744 rth
->rt_gateway
= 0;
1745 rth
->rt_uses_gateway
= 0;
1746 INIT_LIST_HEAD(&rth
->rt_uncached
);
1747 RT_CACHE_STAT_INC(in_slow_tot
);
1748 if (res
.type
== RTN_UNREACHABLE
) {
1749 rth
->dst
.input
= ip_error
;
1750 rth
->dst
.error
= -err
;
1751 rth
->rt_flags
&= ~RTCF_LOCAL
;
1754 if (unlikely(!rt_cache_route(&FIB_RES_NH(res
), rth
))) {
1755 rth
->dst
.flags
|= DST_NOCACHE
;
1756 rt_add_uncached_list(rth
);
1759 skb_dst_set(skb
, &rth
->dst
);
1764 RT_CACHE_STAT_INC(in_no_route
);
1765 res
.type
= RTN_UNREACHABLE
;
1771 * Do not cache martian addresses: they should be logged (RFC1812)
1773 martian_destination
:
1774 RT_CACHE_STAT_INC(in_martian_dst
);
1775 #ifdef CONFIG_IP_ROUTE_VERBOSE
1776 if (IN_DEV_LOG_MARTIANS(in_dev
))
1777 net_warn_ratelimited("martian destination %pI4 from %pI4, dev %s\n",
1778 &daddr
, &saddr
, dev
->name
);
1791 martian_source_keep_err
:
1792 ip_handle_martian_source(dev
, in_dev
, skb
, daddr
, saddr
);
1796 int ip_route_input_noref(struct sk_buff
*skb
, __be32 daddr
, __be32 saddr
,
1797 u8 tos
, struct net_device
*dev
)
1803 /* Multicast recognition logic is moved from route cache to here.
1804 The problem was that too many Ethernet cards have broken/missing
1805 hardware multicast filters :-( As result the host on multicasting
1806 network acquires a lot of useless route cache entries, sort of
1807 SDR messages from all the world. Now we try to get rid of them.
1808 Really, provided software IP multicast filter is organized
1809 reasonably (at least, hashed), it does not result in a slowdown
1810 comparing with route cache reject entries.
1811 Note, that multicast routers are not affected, because
1812 route cache entry is created eventually.
1814 if (ipv4_is_multicast(daddr
)) {
1815 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
1818 int our
= ip_check_mc_rcu(in_dev
, daddr
, saddr
,
1819 ip_hdr(skb
)->protocol
);
1821 #ifdef CONFIG_IP_MROUTE
1823 (!ipv4_is_local_multicast(daddr
) &&
1824 IN_DEV_MFORWARD(in_dev
))
1827 int res
= ip_route_input_mc(skb
, daddr
, saddr
,
1836 res
= ip_route_input_slow(skb
, daddr
, saddr
, tos
, dev
);
1840 EXPORT_SYMBOL(ip_route_input_noref
);
1842 /* called with rcu_read_lock() */
1843 static struct rtable
*__mkroute_output(const struct fib_result
*res
,
1844 const struct flowi4
*fl4
, int orig_oif
,
1845 struct net_device
*dev_out
,
1848 struct fib_info
*fi
= res
->fi
;
1849 struct fib_nh_exception
*fnhe
;
1850 struct in_device
*in_dev
;
1851 u16 type
= res
->type
;
1855 in_dev
= __in_dev_get_rcu(dev_out
);
1857 return ERR_PTR(-EINVAL
);
1859 if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev
)))
1860 if (ipv4_is_loopback(fl4
->saddr
) && !(dev_out
->flags
& IFF_LOOPBACK
))
1861 return ERR_PTR(-EINVAL
);
1863 if (ipv4_is_lbcast(fl4
->daddr
))
1864 type
= RTN_BROADCAST
;
1865 else if (ipv4_is_multicast(fl4
->daddr
))
1866 type
= RTN_MULTICAST
;
1867 else if (ipv4_is_zeronet(fl4
->daddr
))
1868 return ERR_PTR(-EINVAL
);
1870 if (dev_out
->flags
& IFF_LOOPBACK
)
1871 flags
|= RTCF_LOCAL
;
1874 if (type
== RTN_BROADCAST
) {
1875 flags
|= RTCF_BROADCAST
| RTCF_LOCAL
;
1877 } else if (type
== RTN_MULTICAST
) {
1878 flags
|= RTCF_MULTICAST
| RTCF_LOCAL
;
1879 if (!ip_check_mc_rcu(in_dev
, fl4
->daddr
, fl4
->saddr
,
1881 flags
&= ~RTCF_LOCAL
;
1884 /* If multicast route do not exist use
1885 * default one, but do not gateway in this case.
1888 if (fi
&& res
->prefixlen
< 4)
1890 } else if ((type
== RTN_LOCAL
) && (orig_oif
!= 0) &&
1891 (orig_oif
!= dev_out
->ifindex
)) {
1892 /* For local routes that require a particular output interface
1893 * we do not want to cache the result. Caching the result
1894 * causes incorrect behaviour when there are multiple source
1895 * addresses on the interface, the end result being that if the
1896 * intended recipient is waiting on that interface for the
1897 * packet he won't receive it because it will be delivered on
1898 * the loopback interface and the IP_PKTINFO ipi_ifindex will
1899 * be set to the loopback interface as well.
1905 do_cache
&= fi
!= NULL
;
1907 struct rtable __rcu
**prth
;
1908 struct fib_nh
*nh
= &FIB_RES_NH(*res
);
1910 fnhe
= find_exception(nh
, fl4
->daddr
);
1912 prth
= &fnhe
->fnhe_rth
;
1914 if (unlikely(fl4
->flowi4_flags
&
1915 FLOWI_FLAG_KNOWN_NH
&&
1917 nh
->nh_scope
== RT_SCOPE_LINK
))) {
1921 prth
= __this_cpu_ptr(nh
->nh_pcpu_rth_output
);
1923 rth
= rcu_dereference(*prth
);
1924 if (rt_cache_valid(rth
)) {
1925 dst_hold(&rth
->dst
);
1931 rth
= rt_dst_alloc(dev_out
,
1932 IN_DEV_CONF_GET(in_dev
, NOPOLICY
),
1933 IN_DEV_CONF_GET(in_dev
, NOXFRM
),
1936 return ERR_PTR(-ENOBUFS
);
1938 rth
->dst
.output
= ip_output
;
1940 rth
->rt_genid
= rt_genid(dev_net(dev_out
));
1941 rth
->rt_flags
= flags
;
1942 rth
->rt_type
= type
;
1943 rth
->rt_is_input
= 0;
1944 rth
->rt_iif
= orig_oif
? : 0;
1946 rth
->rt_gateway
= 0;
1947 rth
->rt_uses_gateway
= 0;
1948 INIT_LIST_HEAD(&rth
->rt_uncached
);
1950 RT_CACHE_STAT_INC(out_slow_tot
);
1952 if (flags
& RTCF_LOCAL
)
1953 rth
->dst
.input
= ip_local_deliver
;
1954 if (flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
)) {
1955 if (flags
& RTCF_LOCAL
&&
1956 !(dev_out
->flags
& IFF_LOOPBACK
)) {
1957 rth
->dst
.output
= ip_mc_output
;
1958 RT_CACHE_STAT_INC(out_slow_mc
);
1960 #ifdef CONFIG_IP_MROUTE
1961 if (type
== RTN_MULTICAST
) {
1962 if (IN_DEV_MFORWARD(in_dev
) &&
1963 !ipv4_is_local_multicast(fl4
->daddr
)) {
1964 rth
->dst
.input
= ip_mr_input
;
1965 rth
->dst
.output
= ip_mc_output
;
1971 rt_set_nexthop(rth
, fl4
->daddr
, res
, fnhe
, fi
, type
, 0);
1977 * Major route resolver routine.
1980 struct rtable
*__ip_route_output_key(struct net
*net
, struct flowi4
*fl4
)
1982 struct net_device
*dev_out
= NULL
;
1983 __u8 tos
= RT_FL_TOS(fl4
);
1984 unsigned int flags
= 0;
1985 struct fib_result res
;
1993 orig_oif
= fl4
->flowi4_oif
;
1995 fl4
->flowi4_iif
= LOOPBACK_IFINDEX
;
1996 fl4
->flowi4_tos
= tos
& IPTOS_RT_MASK
;
1997 fl4
->flowi4_scope
= ((tos
& RTO_ONLINK
) ?
1998 RT_SCOPE_LINK
: RT_SCOPE_UNIVERSE
);
2002 rth
= ERR_PTR(-EINVAL
);
2003 if (ipv4_is_multicast(fl4
->saddr
) ||
2004 ipv4_is_lbcast(fl4
->saddr
) ||
2005 ipv4_is_zeronet(fl4
->saddr
))
2008 /* I removed check for oif == dev_out->oif here.
2009 It was wrong for two reasons:
2010 1. ip_dev_find(net, saddr) can return wrong iface, if saddr
2011 is assigned to multiple interfaces.
2012 2. Moreover, we are allowed to send packets with saddr
2013 of another iface. --ANK
2016 if (fl4
->flowi4_oif
== 0 &&
2017 (ipv4_is_multicast(fl4
->daddr
) ||
2018 ipv4_is_lbcast(fl4
->daddr
))) {
2019 /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2020 dev_out
= __ip_dev_find(net
, fl4
->saddr
, false);
2021 if (dev_out
== NULL
)
2024 /* Special hack: user can direct multicasts
2025 and limited broadcast via necessary interface
2026 without fiddling with IP_MULTICAST_IF or IP_PKTINFO.
2027 This hack is not just for fun, it allows
2028 vic,vat and friends to work.
2029 They bind socket to loopback, set ttl to zero
2030 and expect that it will work.
2031 From the viewpoint of routing cache they are broken,
2032 because we are not allowed to build multicast path
2033 with loopback source addr (look, routing cache
2034 cannot know, that ttl is zero, so that packet
2035 will not leave this host and route is valid).
2036 Luckily, this hack is good workaround.
2039 fl4
->flowi4_oif
= dev_out
->ifindex
;
2043 if (!(fl4
->flowi4_flags
& FLOWI_FLAG_ANYSRC
)) {
2044 /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2045 if (!__ip_dev_find(net
, fl4
->saddr
, false))
2051 if (fl4
->flowi4_oif
) {
2052 dev_out
= dev_get_by_index_rcu(net
, fl4
->flowi4_oif
);
2053 rth
= ERR_PTR(-ENODEV
);
2054 if (dev_out
== NULL
)
2057 /* RACE: Check return value of inet_select_addr instead. */
2058 if (!(dev_out
->flags
& IFF_UP
) || !__in_dev_get_rcu(dev_out
)) {
2059 rth
= ERR_PTR(-ENETUNREACH
);
2062 if (ipv4_is_local_multicast(fl4
->daddr
) ||
2063 ipv4_is_lbcast(fl4
->daddr
)) {
2065 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2070 if (ipv4_is_multicast(fl4
->daddr
))
2071 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2073 else if (!fl4
->daddr
)
2074 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2080 fl4
->daddr
= fl4
->saddr
;
2082 fl4
->daddr
= fl4
->saddr
= htonl(INADDR_LOOPBACK
);
2083 dev_out
= net
->loopback_dev
;
2084 fl4
->flowi4_oif
= LOOPBACK_IFINDEX
;
2085 res
.type
= RTN_LOCAL
;
2086 flags
|= RTCF_LOCAL
;
2090 if (fib_lookup(net
, fl4
, &res
)) {
2093 if (fl4
->flowi4_oif
) {
2094 /* Apparently, routing tables are wrong. Assume,
2095 that the destination is on link.
2098 Because we are allowed to send to iface
2099 even if it has NO routes and NO assigned
2100 addresses. When oif is specified, routing
2101 tables are looked up with only one purpose:
2102 to catch if destination is gatewayed, rather than
2103 direct. Moreover, if MSG_DONTROUTE is set,
2104 we send packet, ignoring both routing tables
2105 and ifaddr state. --ANK
2108 We could make it even if oif is unknown,
2109 likely IPv6, but we do not.
2112 if (fl4
->saddr
== 0)
2113 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2115 res
.type
= RTN_UNICAST
;
2118 rth
= ERR_PTR(-ENETUNREACH
);
2122 if (res
.type
== RTN_LOCAL
) {
2124 if (res
.fi
->fib_prefsrc
)
2125 fl4
->saddr
= res
.fi
->fib_prefsrc
;
2127 fl4
->saddr
= fl4
->daddr
;
2129 dev_out
= net
->loopback_dev
;
2130 fl4
->flowi4_oif
= dev_out
->ifindex
;
2131 flags
|= RTCF_LOCAL
;
2135 #ifdef CONFIG_IP_ROUTE_MULTIPATH
2136 if (res
.fi
->fib_nhs
> 1 && fl4
->flowi4_oif
== 0)
2137 fib_select_multipath(&res
);
2140 if (!res
.prefixlen
&&
2141 res
.table
->tb_num_default
> 1 &&
2142 res
.type
== RTN_UNICAST
&& !fl4
->flowi4_oif
)
2143 fib_select_default(&res
);
2146 fl4
->saddr
= FIB_RES_PREFSRC(net
, res
);
2148 dev_out
= FIB_RES_DEV(res
);
2149 fl4
->flowi4_oif
= dev_out
->ifindex
;
2153 rth
= __mkroute_output(&res
, fl4
, orig_oif
, dev_out
, flags
);
2159 EXPORT_SYMBOL_GPL(__ip_route_output_key
);
2161 static struct dst_entry
*ipv4_blackhole_dst_check(struct dst_entry
*dst
, u32 cookie
)
2166 static unsigned int ipv4_blackhole_mtu(const struct dst_entry
*dst
)
2168 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
2170 return mtu
? : dst
->dev
->mtu
;
2173 static void ipv4_rt_blackhole_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
2174 struct sk_buff
*skb
, u32 mtu
)
2178 static void ipv4_rt_blackhole_redirect(struct dst_entry
*dst
, struct sock
*sk
,
2179 struct sk_buff
*skb
)
2183 static u32
*ipv4_rt_blackhole_cow_metrics(struct dst_entry
*dst
,
2189 static struct dst_ops ipv4_dst_blackhole_ops
= {
2191 .protocol
= cpu_to_be16(ETH_P_IP
),
2192 .check
= ipv4_blackhole_dst_check
,
2193 .mtu
= ipv4_blackhole_mtu
,
2194 .default_advmss
= ipv4_default_advmss
,
2195 .update_pmtu
= ipv4_rt_blackhole_update_pmtu
,
2196 .redirect
= ipv4_rt_blackhole_redirect
,
2197 .cow_metrics
= ipv4_rt_blackhole_cow_metrics
,
2198 .neigh_lookup
= ipv4_neigh_lookup
,
2201 struct dst_entry
*ipv4_blackhole_route(struct net
*net
, struct dst_entry
*dst_orig
)
2203 struct rtable
*ort
= (struct rtable
*) dst_orig
;
2206 rt
= dst_alloc(&ipv4_dst_blackhole_ops
, NULL
, 1, DST_OBSOLETE_NONE
, 0);
2208 struct dst_entry
*new = &rt
->dst
;
2211 new->input
= dst_discard
;
2212 new->output
= dst_discard
;
2214 new->dev
= ort
->dst
.dev
;
2218 rt
->rt_is_input
= ort
->rt_is_input
;
2219 rt
->rt_iif
= ort
->rt_iif
;
2220 rt
->rt_pmtu
= ort
->rt_pmtu
;
2222 rt
->rt_genid
= rt_genid(net
);
2223 rt
->rt_flags
= ort
->rt_flags
;
2224 rt
->rt_type
= ort
->rt_type
;
2225 rt
->rt_gateway
= ort
->rt_gateway
;
2226 rt
->rt_uses_gateway
= ort
->rt_uses_gateway
;
2228 INIT_LIST_HEAD(&rt
->rt_uncached
);
2233 dst_release(dst_orig
);
2235 return rt
? &rt
->dst
: ERR_PTR(-ENOMEM
);
2238 struct rtable
*ip_route_output_flow(struct net
*net
, struct flowi4
*flp4
,
2241 struct rtable
*rt
= __ip_route_output_key(net
, flp4
);
2246 if (flp4
->flowi4_proto
)
2247 rt
= (struct rtable
*) xfrm_lookup(net
, &rt
->dst
,
2248 flowi4_to_flowi(flp4
),
2253 EXPORT_SYMBOL_GPL(ip_route_output_flow
);
2255 static int rt_fill_info(struct net
*net
, __be32 dst
, __be32 src
,
2256 struct flowi4
*fl4
, struct sk_buff
*skb
, u32 portid
,
2257 u32 seq
, int event
, int nowait
, unsigned int flags
)
2259 struct rtable
*rt
= skb_rtable(skb
);
2261 struct nlmsghdr
*nlh
;
2262 unsigned long expires
= 0;
2264 u32 metrics
[RTAX_MAX
];
2266 nlh
= nlmsg_put(skb
, portid
, seq
, event
, sizeof(*r
), flags
);
2270 r
= nlmsg_data(nlh
);
2271 r
->rtm_family
= AF_INET
;
2272 r
->rtm_dst_len
= 32;
2274 r
->rtm_tos
= fl4
->flowi4_tos
;
2275 r
->rtm_table
= RT_TABLE_MAIN
;
2276 if (nla_put_u32(skb
, RTA_TABLE
, RT_TABLE_MAIN
))
2277 goto nla_put_failure
;
2278 r
->rtm_type
= rt
->rt_type
;
2279 r
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2280 r
->rtm_protocol
= RTPROT_UNSPEC
;
2281 r
->rtm_flags
= (rt
->rt_flags
& ~0xFFFF) | RTM_F_CLONED
;
2282 if (rt
->rt_flags
& RTCF_NOTIFY
)
2283 r
->rtm_flags
|= RTM_F_NOTIFY
;
2284 if (IPCB(skb
)->flags
& IPSKB_DOREDIRECT
)
2285 r
->rtm_flags
|= RTCF_DOREDIRECT
;
2287 if (nla_put_be32(skb
, RTA_DST
, dst
))
2288 goto nla_put_failure
;
2290 r
->rtm_src_len
= 32;
2291 if (nla_put_be32(skb
, RTA_SRC
, src
))
2292 goto nla_put_failure
;
2295 nla_put_u32(skb
, RTA_OIF
, rt
->dst
.dev
->ifindex
))
2296 goto nla_put_failure
;
2297 #ifdef CONFIG_IP_ROUTE_CLASSID
2298 if (rt
->dst
.tclassid
&&
2299 nla_put_u32(skb
, RTA_FLOW
, rt
->dst
.tclassid
))
2300 goto nla_put_failure
;
2302 if (!rt_is_input_route(rt
) &&
2303 fl4
->saddr
!= src
) {
2304 if (nla_put_be32(skb
, RTA_PREFSRC
, fl4
->saddr
))
2305 goto nla_put_failure
;
2307 if (rt
->rt_uses_gateway
&&
2308 nla_put_be32(skb
, RTA_GATEWAY
, rt
->rt_gateway
))
2309 goto nla_put_failure
;
2311 expires
= rt
->dst
.expires
;
2313 unsigned long now
= jiffies
;
2315 if (time_before(now
, expires
))
2321 memcpy(metrics
, dst_metrics_ptr(&rt
->dst
), sizeof(metrics
));
2322 if (rt
->rt_pmtu
&& expires
)
2323 metrics
[RTAX_MTU
- 1] = rt
->rt_pmtu
;
2324 if (rtnetlink_put_metrics(skb
, metrics
) < 0)
2325 goto nla_put_failure
;
2327 if (fl4
->flowi4_mark
&&
2328 nla_put_u32(skb
, RTA_MARK
, fl4
->flowi4_mark
))
2329 goto nla_put_failure
;
2331 if (!uid_eq(fl4
->flowi4_uid
, INVALID_UID
) &&
2332 nla_put_u32(skb
, RTA_UID
,
2333 from_kuid_munged(current_user_ns(), fl4
->flowi4_uid
)))
2334 goto nla_put_failure
;
2336 error
= rt
->dst
.error
;
2338 if (rt_is_input_route(rt
)) {
2339 #ifdef CONFIG_IP_MROUTE
2340 if (ipv4_is_multicast(dst
) && !ipv4_is_local_multicast(dst
) &&
2341 IPV4_DEVCONF_ALL(net
, MC_FORWARDING
)) {
2342 int err
= ipmr_get_route(net
, skb
,
2343 fl4
->saddr
, fl4
->daddr
,
2350 goto nla_put_failure
;
2352 if (err
== -EMSGSIZE
)
2353 goto nla_put_failure
;
2359 if (nla_put_u32(skb
, RTA_IIF
, skb
->dev
->ifindex
))
2360 goto nla_put_failure
;
2363 if (rtnl_put_cacheinfo(skb
, &rt
->dst
, 0, expires
, error
) < 0)
2364 goto nla_put_failure
;
2366 return nlmsg_end(skb
, nlh
);
2369 nlmsg_cancel(skb
, nlh
);
2373 static int inet_rtm_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
*nlh
)
2375 struct net
*net
= sock_net(in_skb
->sk
);
2377 struct nlattr
*tb
[RTA_MAX
+1];
2378 struct rtable
*rt
= NULL
;
2385 struct sk_buff
*skb
;
2388 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv4_policy
);
2392 rtm
= nlmsg_data(nlh
);
2394 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
2400 /* Reserve room for dummy headers, this skb can pass
2401 through good chunk of routing engine.
2403 skb_reset_mac_header(skb
);
2404 skb_reset_network_header(skb
);
2406 /* Bugfix: need to give ip_route_input enough of an IP header to not gag. */
2407 ip_hdr(skb
)->protocol
= IPPROTO_ICMP
;
2408 skb_reserve(skb
, MAX_HEADER
+ sizeof(struct iphdr
));
2410 src
= tb
[RTA_SRC
] ? nla_get_be32(tb
[RTA_SRC
]) : 0;
2411 dst
= tb
[RTA_DST
] ? nla_get_be32(tb
[RTA_DST
]) : 0;
2412 iif
= tb
[RTA_IIF
] ? nla_get_u32(tb
[RTA_IIF
]) : 0;
2413 mark
= tb
[RTA_MARK
] ? nla_get_u32(tb
[RTA_MARK
]) : 0;
2415 uid
= make_kuid(current_user_ns(), nla_get_u32(tb
[RTA_UID
]));
2417 uid
= (iif
? INVALID_UID
: current_uid());
2419 memset(&fl4
, 0, sizeof(fl4
));
2422 fl4
.flowi4_tos
= rtm
->rtm_tos
;
2423 fl4
.flowi4_oif
= tb
[RTA_OIF
] ? nla_get_u32(tb
[RTA_OIF
]) : 0;
2424 fl4
.flowi4_mark
= mark
;
2425 fl4
.flowi4_uid
= uid
;
2428 struct net_device
*dev
;
2430 dev
= __dev_get_by_index(net
, iif
);
2436 skb
->protocol
= htons(ETH_P_IP
);
2440 err
= ip_route_input(skb
, dst
, src
, rtm
->rtm_tos
, dev
);
2443 rt
= skb_rtable(skb
);
2444 if (err
== 0 && rt
->dst
.error
)
2445 err
= -rt
->dst
.error
;
2447 rt
= ip_route_output_key(net
, &fl4
);
2457 skb_dst_set(skb
, &rt
->dst
);
2458 if (rtm
->rtm_flags
& RTM_F_NOTIFY
)
2459 rt
->rt_flags
|= RTCF_NOTIFY
;
2461 err
= rt_fill_info(net
, dst
, src
, &fl4
, skb
,
2462 NETLINK_CB(in_skb
).portid
, nlh
->nlmsg_seq
,
2463 RTM_NEWROUTE
, 0, 0);
2467 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).portid
);
2476 int ip_rt_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2481 void ip_rt_multicast_event(struct in_device
*in_dev
)
2483 rt_cache_flush(dev_net(in_dev
->dev
));
2486 #ifdef CONFIG_SYSCTL
2487 static int ip_rt_gc_timeout __read_mostly
= RT_GC_TIMEOUT
;
2488 static int ip_rt_gc_interval __read_mostly
= 60 * HZ
;
2489 static int ip_rt_gc_min_interval __read_mostly
= HZ
/ 2;
2490 static int ip_rt_gc_elasticity __read_mostly
= 8;
2492 static int ipv4_sysctl_rtcache_flush(ctl_table
*__ctl
, int write
,
2493 void __user
*buffer
,
2494 size_t *lenp
, loff_t
*ppos
)
2497 rt_cache_flush((struct net
*)__ctl
->extra1
);
2504 static ctl_table ipv4_route_table
[] = {
2506 .procname
= "gc_thresh",
2507 .data
= &ipv4_dst_ops
.gc_thresh
,
2508 .maxlen
= sizeof(int),
2510 .proc_handler
= proc_dointvec
,
2513 .procname
= "max_size",
2514 .data
= &ip_rt_max_size
,
2515 .maxlen
= sizeof(int),
2517 .proc_handler
= proc_dointvec
,
2520 /* Deprecated. Use gc_min_interval_ms */
2522 .procname
= "gc_min_interval",
2523 .data
= &ip_rt_gc_min_interval
,
2524 .maxlen
= sizeof(int),
2526 .proc_handler
= proc_dointvec_jiffies
,
2529 .procname
= "gc_min_interval_ms",
2530 .data
= &ip_rt_gc_min_interval
,
2531 .maxlen
= sizeof(int),
2533 .proc_handler
= proc_dointvec_ms_jiffies
,
2536 .procname
= "gc_timeout",
2537 .data
= &ip_rt_gc_timeout
,
2538 .maxlen
= sizeof(int),
2540 .proc_handler
= proc_dointvec_jiffies
,
2543 .procname
= "gc_interval",
2544 .data
= &ip_rt_gc_interval
,
2545 .maxlen
= sizeof(int),
2547 .proc_handler
= proc_dointvec_jiffies
,
2550 .procname
= "redirect_load",
2551 .data
= &ip_rt_redirect_load
,
2552 .maxlen
= sizeof(int),
2554 .proc_handler
= proc_dointvec
,
2557 .procname
= "redirect_number",
2558 .data
= &ip_rt_redirect_number
,
2559 .maxlen
= sizeof(int),
2561 .proc_handler
= proc_dointvec
,
2564 .procname
= "redirect_silence",
2565 .data
= &ip_rt_redirect_silence
,
2566 .maxlen
= sizeof(int),
2568 .proc_handler
= proc_dointvec
,
2571 .procname
= "error_cost",
2572 .data
= &ip_rt_error_cost
,
2573 .maxlen
= sizeof(int),
2575 .proc_handler
= proc_dointvec
,
2578 .procname
= "error_burst",
2579 .data
= &ip_rt_error_burst
,
2580 .maxlen
= sizeof(int),
2582 .proc_handler
= proc_dointvec
,
2585 .procname
= "gc_elasticity",
2586 .data
= &ip_rt_gc_elasticity
,
2587 .maxlen
= sizeof(int),
2589 .proc_handler
= proc_dointvec
,
2592 .procname
= "mtu_expires",
2593 .data
= &ip_rt_mtu_expires
,
2594 .maxlen
= sizeof(int),
2596 .proc_handler
= proc_dointvec_jiffies
,
2599 .procname
= "min_pmtu",
2600 .data
= &ip_rt_min_pmtu
,
2601 .maxlen
= sizeof(int),
2603 .proc_handler
= proc_dointvec
,
2606 .procname
= "min_adv_mss",
2607 .data
= &ip_rt_min_advmss
,
2608 .maxlen
= sizeof(int),
2610 .proc_handler
= proc_dointvec
,
2615 static struct ctl_table ipv4_route_flush_table
[] = {
2617 .procname
= "flush",
2618 .maxlen
= sizeof(int),
2620 .proc_handler
= ipv4_sysctl_rtcache_flush
,
2625 static __net_init
int sysctl_route_net_init(struct net
*net
)
2627 struct ctl_table
*tbl
;
2629 tbl
= ipv4_route_flush_table
;
2630 if (!net_eq(net
, &init_net
)) {
2631 tbl
= kmemdup(tbl
, sizeof(ipv4_route_flush_table
), GFP_KERNEL
);
2635 /* Don't export sysctls to unprivileged users */
2636 if (net
->user_ns
!= &init_user_ns
)
2637 tbl
[0].procname
= NULL
;
2639 tbl
[0].extra1
= net
;
2641 net
->ipv4
.route_hdr
= register_net_sysctl(net
, "net/ipv4/route", tbl
);
2642 if (net
->ipv4
.route_hdr
== NULL
)
2647 if (tbl
!= ipv4_route_flush_table
)
2653 static __net_exit
void sysctl_route_net_exit(struct net
*net
)
2655 struct ctl_table
*tbl
;
2657 tbl
= net
->ipv4
.route_hdr
->ctl_table_arg
;
2658 unregister_net_sysctl_table(net
->ipv4
.route_hdr
);
2659 BUG_ON(tbl
== ipv4_route_flush_table
);
2663 static __net_initdata
struct pernet_operations sysctl_route_ops
= {
2664 .init
= sysctl_route_net_init
,
2665 .exit
= sysctl_route_net_exit
,
2669 static __net_init
int rt_genid_init(struct net
*net
)
2671 atomic_set(&net
->rt_genid
, 0);
2672 get_random_bytes(&net
->ipv4
.dev_addr_genid
,
2673 sizeof(net
->ipv4
.dev_addr_genid
));
2677 static __net_initdata
struct pernet_operations rt_genid_ops
= {
2678 .init
= rt_genid_init
,
2681 static int __net_init
ipv4_inetpeer_init(struct net
*net
)
2683 struct inet_peer_base
*bp
= kmalloc(sizeof(*bp
), GFP_KERNEL
);
2687 inet_peer_base_init(bp
);
2688 net
->ipv4
.peers
= bp
;
2692 static void __net_exit
ipv4_inetpeer_exit(struct net
*net
)
2694 struct inet_peer_base
*bp
= net
->ipv4
.peers
;
2696 net
->ipv4
.peers
= NULL
;
2697 inetpeer_invalidate_tree(bp
);
2701 static __net_initdata
struct pernet_operations ipv4_inetpeer_ops
= {
2702 .init
= ipv4_inetpeer_init
,
2703 .exit
= ipv4_inetpeer_exit
,
2706 #ifdef CONFIG_IP_ROUTE_CLASSID
2707 struct ip_rt_acct __percpu
*ip_rt_acct __read_mostly
;
2708 #endif /* CONFIG_IP_ROUTE_CLASSID */
2710 int __init
ip_rt_init(void)
2714 ip_idents
= kmalloc(IP_IDENTS_SZ
* sizeof(*ip_idents
), GFP_KERNEL
);
2716 panic("IP: failed to allocate ip_idents\n");
2718 prandom_bytes(ip_idents
, IP_IDENTS_SZ
* sizeof(*ip_idents
));
2720 #ifdef CONFIG_IP_ROUTE_CLASSID
2721 ip_rt_acct
= __alloc_percpu(256 * sizeof(struct ip_rt_acct
), __alignof__(struct ip_rt_acct
));
2723 panic("IP: failed to allocate ip_rt_acct\n");
2726 ipv4_dst_ops
.kmem_cachep
=
2727 kmem_cache_create("ip_dst_cache", sizeof(struct rtable
), 0,
2728 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
);
2730 ipv4_dst_blackhole_ops
.kmem_cachep
= ipv4_dst_ops
.kmem_cachep
;
2732 if (dst_entries_init(&ipv4_dst_ops
) < 0)
2733 panic("IP: failed to allocate ipv4_dst_ops counter\n");
2735 if (dst_entries_init(&ipv4_dst_blackhole_ops
) < 0)
2736 panic("IP: failed to allocate ipv4_dst_blackhole_ops counter\n");
2738 ipv4_dst_ops
.gc_thresh
= ~0;
2739 ip_rt_max_size
= INT_MAX
;
2744 if (ip_rt_proc_init())
2745 pr_err("Unable to create route proc files\n");
2750 rtnl_register(PF_INET
, RTM_GETROUTE
, inet_rtm_getroute
, NULL
, NULL
);
2752 #ifdef CONFIG_SYSCTL
2753 register_pernet_subsys(&sysctl_route_ops
);
2755 register_pernet_subsys(&rt_genid_ops
);
2756 register_pernet_subsys(&ipv4_inetpeer_ops
);
2760 #ifdef CONFIG_SYSCTL
2762 * We really need to sanitize the damn ipv4 init order, then all
2763 * this nonsense will go away.
2765 void __init
ip_static_sysctl_init(void)
2767 register_net_sysctl(&init_net
, "net/ipv4/route", ipv4_route_table
);