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 RT_GC_TIMEOUT (300*HZ)
118 static int ip_rt_max_size
;
119 static int ip_rt_redirect_number __read_mostly
= 9;
120 static int ip_rt_redirect_load __read_mostly
= HZ
/ 50;
121 static int ip_rt_redirect_silence __read_mostly
= ((HZ
/ 50) << (9 + 1));
122 static int ip_rt_error_cost __read_mostly
= HZ
;
123 static int ip_rt_error_burst __read_mostly
= 5 * HZ
;
124 static int ip_rt_mtu_expires __read_mostly
= 10 * 60 * HZ
;
125 static int ip_rt_min_pmtu __read_mostly
= 512 + 20 + 20;
126 static int ip_rt_min_advmss __read_mostly
= 256;
129 * Interface to generic destination cache.
132 static struct dst_entry
*ipv4_dst_check(struct dst_entry
*dst
, u32 cookie
);
133 static unsigned int ipv4_default_advmss(const struct dst_entry
*dst
);
134 static unsigned int ipv4_mtu(const struct dst_entry
*dst
);
135 static struct dst_entry
*ipv4_negative_advice(struct dst_entry
*dst
);
136 static void ipv4_link_failure(struct sk_buff
*skb
);
137 static void ip_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
138 struct sk_buff
*skb
, u32 mtu
);
139 static void ip_do_redirect(struct dst_entry
*dst
, struct sock
*sk
,
140 struct sk_buff
*skb
);
141 static void ipv4_dst_destroy(struct dst_entry
*dst
);
143 static u32
*ipv4_cow_metrics(struct dst_entry
*dst
, unsigned long old
)
149 static struct neighbour
*ipv4_neigh_lookup(const struct dst_entry
*dst
,
153 static struct dst_ops ipv4_dst_ops
= {
155 .check
= ipv4_dst_check
,
156 .default_advmss
= ipv4_default_advmss
,
158 .cow_metrics
= ipv4_cow_metrics
,
159 .destroy
= ipv4_dst_destroy
,
160 .negative_advice
= ipv4_negative_advice
,
161 .link_failure
= ipv4_link_failure
,
162 .update_pmtu
= ip_rt_update_pmtu
,
163 .redirect
= ip_do_redirect
,
164 .local_out
= __ip_local_out
,
165 .neigh_lookup
= ipv4_neigh_lookup
,
168 #define ECN_OR_COST(class) TC_PRIO_##class
170 const __u8 ip_tos2prio
[16] = {
172 ECN_OR_COST(BESTEFFORT
),
174 ECN_OR_COST(BESTEFFORT
),
180 ECN_OR_COST(INTERACTIVE
),
182 ECN_OR_COST(INTERACTIVE
),
183 TC_PRIO_INTERACTIVE_BULK
,
184 ECN_OR_COST(INTERACTIVE_BULK
),
185 TC_PRIO_INTERACTIVE_BULK
,
186 ECN_OR_COST(INTERACTIVE_BULK
)
188 EXPORT_SYMBOL(ip_tos2prio
);
190 static DEFINE_PER_CPU(struct rt_cache_stat
, rt_cache_stat
);
191 #define RT_CACHE_STAT_INC(field) raw_cpu_inc(rt_cache_stat.field)
193 #ifdef CONFIG_PROC_FS
194 static void *rt_cache_seq_start(struct seq_file
*seq
, loff_t
*pos
)
198 return SEQ_START_TOKEN
;
201 static void *rt_cache_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
207 static void rt_cache_seq_stop(struct seq_file
*seq
, void *v
)
211 static int rt_cache_seq_show(struct seq_file
*seq
, void *v
)
213 if (v
== SEQ_START_TOKEN
)
214 seq_printf(seq
, "%-127s\n",
215 "Iface\tDestination\tGateway \tFlags\t\tRefCnt\tUse\t"
216 "Metric\tSource\t\tMTU\tWindow\tIRTT\tTOS\tHHRef\t"
221 static const struct seq_operations rt_cache_seq_ops
= {
222 .start
= rt_cache_seq_start
,
223 .next
= rt_cache_seq_next
,
224 .stop
= rt_cache_seq_stop
,
225 .show
= rt_cache_seq_show
,
228 static int rt_cache_seq_open(struct inode
*inode
, struct file
*file
)
230 return seq_open(file
, &rt_cache_seq_ops
);
233 static const struct file_operations rt_cache_seq_fops
= {
234 .owner
= THIS_MODULE
,
235 .open
= rt_cache_seq_open
,
238 .release
= seq_release
,
242 static void *rt_cpu_seq_start(struct seq_file
*seq
, loff_t
*pos
)
247 return SEQ_START_TOKEN
;
249 for (cpu
= *pos
-1; cpu
< nr_cpu_ids
; ++cpu
) {
250 if (!cpu_possible(cpu
))
253 return &per_cpu(rt_cache_stat
, cpu
);
258 static void *rt_cpu_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
262 for (cpu
= *pos
; cpu
< nr_cpu_ids
; ++cpu
) {
263 if (!cpu_possible(cpu
))
266 return &per_cpu(rt_cache_stat
, cpu
);
272 static void rt_cpu_seq_stop(struct seq_file
*seq
, void *v
)
277 static int rt_cpu_seq_show(struct seq_file
*seq
, void *v
)
279 struct rt_cache_stat
*st
= v
;
281 if (v
== SEQ_START_TOKEN
) {
282 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");
286 seq_printf(seq
,"%08x %08x %08x %08x %08x %08x %08x %08x "
287 " %08x %08x %08x %08x %08x %08x %08x %08x %08x \n",
288 dst_entries_get_slow(&ipv4_dst_ops
),
301 0, /* st->gc_total */
302 0, /* st->gc_ignored */
303 0, /* st->gc_goal_miss */
304 0, /* st->gc_dst_overflow */
305 0, /* st->in_hlist_search */
306 0 /* st->out_hlist_search */
311 static const struct seq_operations rt_cpu_seq_ops
= {
312 .start
= rt_cpu_seq_start
,
313 .next
= rt_cpu_seq_next
,
314 .stop
= rt_cpu_seq_stop
,
315 .show
= rt_cpu_seq_show
,
319 static int rt_cpu_seq_open(struct inode
*inode
, struct file
*file
)
321 return seq_open(file
, &rt_cpu_seq_ops
);
324 static const struct file_operations rt_cpu_seq_fops
= {
325 .owner
= THIS_MODULE
,
326 .open
= rt_cpu_seq_open
,
329 .release
= seq_release
,
332 #ifdef CONFIG_IP_ROUTE_CLASSID
333 static int rt_acct_proc_show(struct seq_file
*m
, void *v
)
335 struct ip_rt_acct
*dst
, *src
;
338 dst
= kcalloc(256, sizeof(struct ip_rt_acct
), GFP_KERNEL
);
342 for_each_possible_cpu(i
) {
343 src
= (struct ip_rt_acct
*)per_cpu_ptr(ip_rt_acct
, i
);
344 for (j
= 0; j
< 256; j
++) {
345 dst
[j
].o_bytes
+= src
[j
].o_bytes
;
346 dst
[j
].o_packets
+= src
[j
].o_packets
;
347 dst
[j
].i_bytes
+= src
[j
].i_bytes
;
348 dst
[j
].i_packets
+= src
[j
].i_packets
;
352 seq_write(m
, dst
, 256 * sizeof(struct ip_rt_acct
));
357 static int rt_acct_proc_open(struct inode
*inode
, struct file
*file
)
359 return single_open(file
, rt_acct_proc_show
, NULL
);
362 static const struct file_operations rt_acct_proc_fops
= {
363 .owner
= THIS_MODULE
,
364 .open
= rt_acct_proc_open
,
367 .release
= single_release
,
371 static int __net_init
ip_rt_do_proc_init(struct net
*net
)
373 struct proc_dir_entry
*pde
;
375 pde
= proc_create("rt_cache", S_IRUGO
, net
->proc_net
,
380 pde
= proc_create("rt_cache", S_IRUGO
,
381 net
->proc_net_stat
, &rt_cpu_seq_fops
);
385 #ifdef CONFIG_IP_ROUTE_CLASSID
386 pde
= proc_create("rt_acct", 0, net
->proc_net
, &rt_acct_proc_fops
);
392 #ifdef CONFIG_IP_ROUTE_CLASSID
394 remove_proc_entry("rt_cache", net
->proc_net_stat
);
397 remove_proc_entry("rt_cache", net
->proc_net
);
402 static void __net_exit
ip_rt_do_proc_exit(struct net
*net
)
404 remove_proc_entry("rt_cache", net
->proc_net_stat
);
405 remove_proc_entry("rt_cache", net
->proc_net
);
406 #ifdef CONFIG_IP_ROUTE_CLASSID
407 remove_proc_entry("rt_acct", net
->proc_net
);
411 static struct pernet_operations ip_rt_proc_ops __net_initdata
= {
412 .init
= ip_rt_do_proc_init
,
413 .exit
= ip_rt_do_proc_exit
,
416 static int __init
ip_rt_proc_init(void)
418 return register_pernet_subsys(&ip_rt_proc_ops
);
422 static inline int ip_rt_proc_init(void)
426 #endif /* CONFIG_PROC_FS */
428 static inline bool rt_is_expired(const struct rtable
*rth
)
430 return rth
->rt_genid
!= rt_genid_ipv4(dev_net(rth
->dst
.dev
));
433 void rt_cache_flush(struct net
*net
)
435 rt_genid_bump_ipv4(net
);
438 static struct neighbour
*ipv4_neigh_lookup(const struct dst_entry
*dst
,
442 struct net_device
*dev
= dst
->dev
;
443 const __be32
*pkey
= daddr
;
444 const struct rtable
*rt
;
447 rt
= (const struct rtable
*) dst
;
449 pkey
= (const __be32
*) &rt
->rt_gateway
;
451 pkey
= &ip_hdr(skb
)->daddr
;
453 n
= __ipv4_neigh_lookup(dev
, *(__force u32
*)pkey
);
456 return neigh_create(&arp_tbl
, pkey
, dev
);
459 #define IP_IDENTS_SZ 2048u
461 static atomic_t
*ip_idents __read_mostly
;
462 static u32
*ip_tstamps __read_mostly
;
464 /* In order to protect privacy, we add a perturbation to identifiers
465 * if one generator is seldom used. This makes hard for an attacker
466 * to infer how many packets were sent between two points in time.
468 u32
ip_idents_reserve(u32 hash
, int segs
)
470 u32
*p_tstamp
= ip_tstamps
+ hash
% IP_IDENTS_SZ
;
471 atomic_t
*p_id
= ip_idents
+ hash
% IP_IDENTS_SZ
;
472 u32 old
= ACCESS_ONCE(*p_tstamp
);
473 u32 now
= (u32
)jiffies
;
476 if (old
!= now
&& cmpxchg(p_tstamp
, old
, now
) == old
)
477 delta
= prandom_u32_max(now
- old
);
479 return atomic_add_return(segs
+ delta
, p_id
) - segs
;
481 EXPORT_SYMBOL(ip_idents_reserve
);
483 void __ip_select_ident(struct net
*net
, struct iphdr
*iph
, int segs
)
485 static u32 ip_idents_hashrnd __read_mostly
;
488 net_get_random_once(&ip_idents_hashrnd
, sizeof(ip_idents_hashrnd
));
490 hash
= jhash_3words((__force u32
)iph
->daddr
,
491 (__force u32
)iph
->saddr
,
492 iph
->protocol
^ net_hash_mix(net
),
494 id
= ip_idents_reserve(hash
, segs
);
497 EXPORT_SYMBOL(__ip_select_ident
);
499 static void __build_flow_key(struct flowi4
*fl4
, const struct sock
*sk
,
500 const struct iphdr
*iph
,
502 u8 prot
, u32 mark
, int flow_flags
)
505 const struct inet_sock
*inet
= inet_sk(sk
);
507 oif
= sk
->sk_bound_dev_if
;
509 tos
= RT_CONN_FLAGS(sk
);
510 prot
= inet
->hdrincl
? IPPROTO_RAW
: sk
->sk_protocol
;
512 flowi4_init_output(fl4
, oif
, mark
, tos
,
513 RT_SCOPE_UNIVERSE
, prot
,
515 iph
->daddr
, iph
->saddr
, 0, 0);
518 static void build_skb_flow_key(struct flowi4
*fl4
, const struct sk_buff
*skb
,
519 const struct sock
*sk
)
521 const struct iphdr
*iph
= ip_hdr(skb
);
522 int oif
= skb
->dev
->ifindex
;
523 u8 tos
= RT_TOS(iph
->tos
);
524 u8 prot
= iph
->protocol
;
525 u32 mark
= skb
->mark
;
527 __build_flow_key(fl4
, sk
, iph
, oif
, tos
, prot
, mark
, 0);
530 static void build_sk_flow_key(struct flowi4
*fl4
, const struct sock
*sk
)
532 const struct inet_sock
*inet
= inet_sk(sk
);
533 const struct ip_options_rcu
*inet_opt
;
534 __be32 daddr
= inet
->inet_daddr
;
537 inet_opt
= rcu_dereference(inet
->inet_opt
);
538 if (inet_opt
&& inet_opt
->opt
.srr
)
539 daddr
= inet_opt
->opt
.faddr
;
540 flowi4_init_output(fl4
, sk
->sk_bound_dev_if
, sk
->sk_mark
,
541 RT_CONN_FLAGS(sk
), RT_SCOPE_UNIVERSE
,
542 inet
->hdrincl
? IPPROTO_RAW
: sk
->sk_protocol
,
543 inet_sk_flowi_flags(sk
),
544 daddr
, inet
->inet_saddr
, 0, 0);
548 static void ip_rt_build_flow_key(struct flowi4
*fl4
, const struct sock
*sk
,
549 const struct sk_buff
*skb
)
552 build_skb_flow_key(fl4
, skb
, sk
);
554 build_sk_flow_key(fl4
, sk
);
557 static inline void rt_free(struct rtable
*rt
)
559 call_rcu(&rt
->dst
.rcu_head
, dst_rcu_free
);
562 static DEFINE_SPINLOCK(fnhe_lock
);
564 static void fnhe_flush_routes(struct fib_nh_exception
*fnhe
)
568 rt
= rcu_dereference(fnhe
->fnhe_rth_input
);
570 RCU_INIT_POINTER(fnhe
->fnhe_rth_input
, NULL
);
573 rt
= rcu_dereference(fnhe
->fnhe_rth_output
);
575 RCU_INIT_POINTER(fnhe
->fnhe_rth_output
, NULL
);
580 static struct fib_nh_exception
*fnhe_oldest(struct fnhe_hash_bucket
*hash
)
582 struct fib_nh_exception
*fnhe
, *oldest
;
584 oldest
= rcu_dereference(hash
->chain
);
585 for (fnhe
= rcu_dereference(oldest
->fnhe_next
); fnhe
;
586 fnhe
= rcu_dereference(fnhe
->fnhe_next
)) {
587 if (time_before(fnhe
->fnhe_stamp
, oldest
->fnhe_stamp
))
590 fnhe_flush_routes(oldest
);
594 static inline u32
fnhe_hashfun(__be32 daddr
)
596 static u32 fnhe_hashrnd __read_mostly
;
599 net_get_random_once(&fnhe_hashrnd
, sizeof(fnhe_hashrnd
));
600 hval
= jhash_1word((__force u32
) daddr
, fnhe_hashrnd
);
601 return hash_32(hval
, FNHE_HASH_SHIFT
);
604 static void fill_route_from_fnhe(struct rtable
*rt
, struct fib_nh_exception
*fnhe
)
606 rt
->rt_pmtu
= fnhe
->fnhe_pmtu
;
607 rt
->dst
.expires
= fnhe
->fnhe_expires
;
610 rt
->rt_flags
|= RTCF_REDIRECTED
;
611 rt
->rt_gateway
= fnhe
->fnhe_gw
;
612 rt
->rt_uses_gateway
= 1;
616 static void update_or_create_fnhe(struct fib_nh
*nh
, __be32 daddr
, __be32 gw
,
617 u32 pmtu
, unsigned long expires
)
619 struct fnhe_hash_bucket
*hash
;
620 struct fib_nh_exception
*fnhe
;
624 u32 hval
= fnhe_hashfun(daddr
);
626 spin_lock_bh(&fnhe_lock
);
628 hash
= rcu_dereference(nh
->nh_exceptions
);
630 hash
= kzalloc(FNHE_HASH_SIZE
* sizeof(*hash
), GFP_ATOMIC
);
633 rcu_assign_pointer(nh
->nh_exceptions
, hash
);
639 for (fnhe
= rcu_dereference(hash
->chain
); fnhe
;
640 fnhe
= rcu_dereference(fnhe
->fnhe_next
)) {
641 if (fnhe
->fnhe_daddr
== daddr
)
650 fnhe
->fnhe_pmtu
= pmtu
;
651 fnhe
->fnhe_expires
= max(1UL, expires
);
653 /* Update all cached dsts too */
654 rt
= rcu_dereference(fnhe
->fnhe_rth_input
);
656 fill_route_from_fnhe(rt
, fnhe
);
657 rt
= rcu_dereference(fnhe
->fnhe_rth_output
);
659 fill_route_from_fnhe(rt
, fnhe
);
661 if (depth
> FNHE_RECLAIM_DEPTH
)
662 fnhe
= fnhe_oldest(hash
);
664 fnhe
= kzalloc(sizeof(*fnhe
), GFP_ATOMIC
);
668 fnhe
->fnhe_next
= hash
->chain
;
669 rcu_assign_pointer(hash
->chain
, fnhe
);
671 fnhe
->fnhe_genid
= fnhe_genid(dev_net(nh
->nh_dev
));
672 fnhe
->fnhe_daddr
= daddr
;
674 fnhe
->fnhe_pmtu
= pmtu
;
675 fnhe
->fnhe_expires
= expires
;
677 /* Exception created; mark the cached routes for the nexthop
678 * stale, so anyone caching it rechecks if this exception
681 rt
= rcu_dereference(nh
->nh_rth_input
);
683 rt
->dst
.obsolete
= DST_OBSOLETE_KILL
;
685 for_each_possible_cpu(i
) {
686 struct rtable __rcu
**prt
;
687 prt
= per_cpu_ptr(nh
->nh_pcpu_rth_output
, i
);
688 rt
= rcu_dereference(*prt
);
690 rt
->dst
.obsolete
= DST_OBSOLETE_KILL
;
694 fnhe
->fnhe_stamp
= jiffies
;
697 spin_unlock_bh(&fnhe_lock
);
700 static void __ip_do_redirect(struct rtable
*rt
, struct sk_buff
*skb
, struct flowi4
*fl4
,
703 __be32 new_gw
= icmp_hdr(skb
)->un
.gateway
;
704 __be32 old_gw
= ip_hdr(skb
)->saddr
;
705 struct net_device
*dev
= skb
->dev
;
706 struct in_device
*in_dev
;
707 struct fib_result res
;
711 switch (icmp_hdr(skb
)->code
& 7) {
713 case ICMP_REDIR_NETTOS
:
714 case ICMP_REDIR_HOST
:
715 case ICMP_REDIR_HOSTTOS
:
722 if (rt
->rt_gateway
!= old_gw
)
725 in_dev
= __in_dev_get_rcu(dev
);
730 if (new_gw
== old_gw
|| !IN_DEV_RX_REDIRECTS(in_dev
) ||
731 ipv4_is_multicast(new_gw
) || ipv4_is_lbcast(new_gw
) ||
732 ipv4_is_zeronet(new_gw
))
733 goto reject_redirect
;
735 if (!IN_DEV_SHARED_MEDIA(in_dev
)) {
736 if (!inet_addr_onlink(in_dev
, new_gw
, old_gw
))
737 goto reject_redirect
;
738 if (IN_DEV_SEC_REDIRECTS(in_dev
) && ip_fib_check_default(new_gw
, dev
))
739 goto reject_redirect
;
741 if (inet_addr_type(net
, new_gw
) != RTN_UNICAST
)
742 goto reject_redirect
;
745 n
= ipv4_neigh_lookup(&rt
->dst
, NULL
, &new_gw
);
747 if (!(n
->nud_state
& NUD_VALID
)) {
748 neigh_event_send(n
, NULL
);
750 if (fib_lookup(net
, fl4
, &res
, 0) == 0) {
751 struct fib_nh
*nh
= &FIB_RES_NH(res
);
753 update_or_create_fnhe(nh
, fl4
->daddr
, new_gw
,
757 rt
->dst
.obsolete
= DST_OBSOLETE_KILL
;
758 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE
, n
);
765 #ifdef CONFIG_IP_ROUTE_VERBOSE
766 if (IN_DEV_LOG_MARTIANS(in_dev
)) {
767 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
768 __be32 daddr
= iph
->daddr
;
769 __be32 saddr
= iph
->saddr
;
771 net_info_ratelimited("Redirect from %pI4 on %s about %pI4 ignored\n"
772 " Advised path = %pI4 -> %pI4\n",
773 &old_gw
, dev
->name
, &new_gw
,
780 static void ip_do_redirect(struct dst_entry
*dst
, struct sock
*sk
, struct sk_buff
*skb
)
784 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
785 int oif
= skb
->dev
->ifindex
;
786 u8 tos
= RT_TOS(iph
->tos
);
787 u8 prot
= iph
->protocol
;
788 u32 mark
= skb
->mark
;
790 rt
= (struct rtable
*) dst
;
792 __build_flow_key(&fl4
, sk
, iph
, oif
, tos
, prot
, mark
, 0);
793 __ip_do_redirect(rt
, skb
, &fl4
, true);
796 static struct dst_entry
*ipv4_negative_advice(struct dst_entry
*dst
)
798 struct rtable
*rt
= (struct rtable
*)dst
;
799 struct dst_entry
*ret
= dst
;
802 if (dst
->obsolete
> 0) {
805 } else if ((rt
->rt_flags
& RTCF_REDIRECTED
) ||
816 * 1. The first ip_rt_redirect_number redirects are sent
817 * with exponential backoff, then we stop sending them at all,
818 * assuming that the host ignores our redirects.
819 * 2. If we did not see packets requiring redirects
820 * during ip_rt_redirect_silence, we assume that the host
821 * forgot redirected route and start to send redirects again.
823 * This algorithm is much cheaper and more intelligent than dumb load limiting
826 * NOTE. Do not forget to inhibit load limiting for redirects (redundant)
827 * and "frag. need" (breaks PMTU discovery) in icmp.c.
830 void ip_rt_send_redirect(struct sk_buff
*skb
)
832 struct rtable
*rt
= skb_rtable(skb
);
833 struct in_device
*in_dev
;
834 struct inet_peer
*peer
;
839 in_dev
= __in_dev_get_rcu(rt
->dst
.dev
);
840 if (!in_dev
|| !IN_DEV_TX_REDIRECTS(in_dev
)) {
844 log_martians
= IN_DEV_LOG_MARTIANS(in_dev
);
847 net
= dev_net(rt
->dst
.dev
);
848 peer
= inet_getpeer_v4(net
->ipv4
.peers
, ip_hdr(skb
)->saddr
, 1);
850 icmp_send(skb
, ICMP_REDIRECT
, ICMP_REDIR_HOST
,
851 rt_nexthop(rt
, ip_hdr(skb
)->daddr
));
855 /* No redirected packets during ip_rt_redirect_silence;
856 * reset the algorithm.
858 if (time_after(jiffies
, peer
->rate_last
+ ip_rt_redirect_silence
))
859 peer
->rate_tokens
= 0;
861 /* Too many ignored redirects; do not send anything
862 * set dst.rate_last to the last seen redirected packet.
864 if (peer
->rate_tokens
>= ip_rt_redirect_number
) {
865 peer
->rate_last
= jiffies
;
869 /* Check for load limit; set rate_last to the latest sent
872 if (peer
->rate_tokens
== 0 ||
875 (ip_rt_redirect_load
<< peer
->rate_tokens
)))) {
876 __be32 gw
= rt_nexthop(rt
, ip_hdr(skb
)->daddr
);
878 icmp_send(skb
, ICMP_REDIRECT
, ICMP_REDIR_HOST
, gw
);
879 peer
->rate_last
= jiffies
;
881 #ifdef CONFIG_IP_ROUTE_VERBOSE
883 peer
->rate_tokens
== ip_rt_redirect_number
)
884 net_warn_ratelimited("host %pI4/if%d ignores redirects for %pI4 to %pI4\n",
885 &ip_hdr(skb
)->saddr
, inet_iif(skb
),
886 &ip_hdr(skb
)->daddr
, &gw
);
893 static int ip_error(struct sk_buff
*skb
)
895 struct in_device
*in_dev
= __in_dev_get_rcu(skb
->dev
);
896 struct rtable
*rt
= skb_rtable(skb
);
897 struct inet_peer
*peer
;
903 /* IP on this device is disabled. */
907 net
= dev_net(rt
->dst
.dev
);
908 if (!IN_DEV_FORWARD(in_dev
)) {
909 switch (rt
->dst
.error
) {
911 IP_INC_STATS_BH(net
, IPSTATS_MIB_INADDRERRORS
);
915 IP_INC_STATS_BH(net
, IPSTATS_MIB_INNOROUTES
);
921 switch (rt
->dst
.error
) {
926 code
= ICMP_HOST_UNREACH
;
929 code
= ICMP_NET_UNREACH
;
930 IP_INC_STATS_BH(net
, IPSTATS_MIB_INNOROUTES
);
933 code
= ICMP_PKT_FILTERED
;
937 peer
= inet_getpeer_v4(net
->ipv4
.peers
, ip_hdr(skb
)->saddr
, 1);
942 peer
->rate_tokens
+= now
- peer
->rate_last
;
943 if (peer
->rate_tokens
> ip_rt_error_burst
)
944 peer
->rate_tokens
= ip_rt_error_burst
;
945 peer
->rate_last
= now
;
946 if (peer
->rate_tokens
>= ip_rt_error_cost
)
947 peer
->rate_tokens
-= ip_rt_error_cost
;
953 icmp_send(skb
, ICMP_DEST_UNREACH
, code
, 0);
959 static void __ip_rt_update_pmtu(struct rtable
*rt
, struct flowi4
*fl4
, u32 mtu
)
961 struct dst_entry
*dst
= &rt
->dst
;
962 struct fib_result res
;
964 if (dst_metric_locked(dst
, RTAX_MTU
))
967 if (ipv4_mtu(dst
) < mtu
)
970 if (mtu
< ip_rt_min_pmtu
)
971 mtu
= ip_rt_min_pmtu
;
973 if (rt
->rt_pmtu
== mtu
&&
974 time_before(jiffies
, dst
->expires
- ip_rt_mtu_expires
/ 2))
978 if (fib_lookup(dev_net(dst
->dev
), fl4
, &res
, 0) == 0) {
979 struct fib_nh
*nh
= &FIB_RES_NH(res
);
981 update_or_create_fnhe(nh
, fl4
->daddr
, 0, mtu
,
982 jiffies
+ ip_rt_mtu_expires
);
987 static void ip_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
988 struct sk_buff
*skb
, u32 mtu
)
990 struct rtable
*rt
= (struct rtable
*) dst
;
993 ip_rt_build_flow_key(&fl4
, sk
, skb
);
994 __ip_rt_update_pmtu(rt
, &fl4
, mtu
);
997 void ipv4_update_pmtu(struct sk_buff
*skb
, struct net
*net
, u32 mtu
,
998 int oif
, u32 mark
, u8 protocol
, int flow_flags
)
1000 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1005 mark
= IP4_REPLY_MARK(net
, skb
->mark
);
1007 __build_flow_key(&fl4
, NULL
, iph
, oif
,
1008 RT_TOS(iph
->tos
), protocol
, mark
, flow_flags
);
1009 rt
= __ip_route_output_key(net
, &fl4
);
1011 __ip_rt_update_pmtu(rt
, &fl4
, mtu
);
1015 EXPORT_SYMBOL_GPL(ipv4_update_pmtu
);
1017 static void __ipv4_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, u32 mtu
)
1019 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1023 __build_flow_key(&fl4
, sk
, iph
, 0, 0, 0, 0, 0);
1025 if (!fl4
.flowi4_mark
)
1026 fl4
.flowi4_mark
= IP4_REPLY_MARK(sock_net(sk
), skb
->mark
);
1028 rt
= __ip_route_output_key(sock_net(sk
), &fl4
);
1030 __ip_rt_update_pmtu(rt
, &fl4
, mtu
);
1035 void ipv4_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, u32 mtu
)
1037 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1040 struct dst_entry
*odst
= NULL
;
1045 if (!ip_sk_accept_pmtu(sk
))
1048 odst
= sk_dst_get(sk
);
1050 if (sock_owned_by_user(sk
) || !odst
) {
1051 __ipv4_sk_update_pmtu(skb
, sk
, mtu
);
1055 __build_flow_key(&fl4
, sk
, iph
, 0, 0, 0, 0, 0);
1057 rt
= (struct rtable
*)odst
;
1058 if (odst
->obsolete
&& !odst
->ops
->check(odst
, 0)) {
1059 rt
= ip_route_output_flow(sock_net(sk
), &fl4
, sk
);
1066 __ip_rt_update_pmtu((struct rtable
*) rt
->dst
.path
, &fl4
, mtu
);
1068 if (!dst_check(&rt
->dst
, 0)) {
1070 dst_release(&rt
->dst
);
1072 rt
= ip_route_output_flow(sock_net(sk
), &fl4
, sk
);
1080 sk_dst_set(sk
, &rt
->dst
);
1086 EXPORT_SYMBOL_GPL(ipv4_sk_update_pmtu
);
1088 void ipv4_redirect(struct sk_buff
*skb
, struct net
*net
,
1089 int oif
, u32 mark
, u8 protocol
, int flow_flags
)
1091 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1095 __build_flow_key(&fl4
, NULL
, iph
, oif
,
1096 RT_TOS(iph
->tos
), protocol
, mark
, flow_flags
);
1097 rt
= __ip_route_output_key(net
, &fl4
);
1099 __ip_do_redirect(rt
, skb
, &fl4
, false);
1103 EXPORT_SYMBOL_GPL(ipv4_redirect
);
1105 void ipv4_sk_redirect(struct sk_buff
*skb
, struct sock
*sk
)
1107 const struct iphdr
*iph
= (const struct iphdr
*) skb
->data
;
1111 __build_flow_key(&fl4
, sk
, iph
, 0, 0, 0, 0, 0);
1112 rt
= __ip_route_output_key(sock_net(sk
), &fl4
);
1114 __ip_do_redirect(rt
, skb
, &fl4
, false);
1118 EXPORT_SYMBOL_GPL(ipv4_sk_redirect
);
1120 static struct dst_entry
*ipv4_dst_check(struct dst_entry
*dst
, u32 cookie
)
1122 struct rtable
*rt
= (struct rtable
*) dst
;
1124 /* All IPV4 dsts are created with ->obsolete set to the value
1125 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1126 * into this function always.
1128 * When a PMTU/redirect information update invalidates a route,
1129 * this is indicated by setting obsolete to DST_OBSOLETE_KILL or
1130 * DST_OBSOLETE_DEAD by dst_free().
1132 if (dst
->obsolete
!= DST_OBSOLETE_FORCE_CHK
|| rt_is_expired(rt
))
1137 static void ipv4_link_failure(struct sk_buff
*skb
)
1141 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_HOST_UNREACH
, 0);
1143 rt
= skb_rtable(skb
);
1145 dst_set_expires(&rt
->dst
, 0);
1148 static int ip_rt_bug(struct sock
*sk
, struct sk_buff
*skb
)
1150 pr_debug("%s: %pI4 -> %pI4, %s\n",
1151 __func__
, &ip_hdr(skb
)->saddr
, &ip_hdr(skb
)->daddr
,
1152 skb
->dev
? skb
->dev
->name
: "?");
1159 We do not cache source address of outgoing interface,
1160 because it is used only by IP RR, TS and SRR options,
1161 so that it out of fast path.
1163 BTW remember: "addr" is allowed to be not aligned
1167 void ip_rt_get_source(u8
*addr
, struct sk_buff
*skb
, struct rtable
*rt
)
1171 if (rt_is_output_route(rt
))
1172 src
= ip_hdr(skb
)->saddr
;
1174 struct fib_result res
;
1180 memset(&fl4
, 0, sizeof(fl4
));
1181 fl4
.daddr
= iph
->daddr
;
1182 fl4
.saddr
= iph
->saddr
;
1183 fl4
.flowi4_tos
= RT_TOS(iph
->tos
);
1184 fl4
.flowi4_oif
= rt
->dst
.dev
->ifindex
;
1185 fl4
.flowi4_iif
= skb
->dev
->ifindex
;
1186 fl4
.flowi4_mark
= skb
->mark
;
1189 if (fib_lookup(dev_net(rt
->dst
.dev
), &fl4
, &res
, 0) == 0)
1190 src
= FIB_RES_PREFSRC(dev_net(rt
->dst
.dev
), res
);
1192 src
= inet_select_addr(rt
->dst
.dev
,
1193 rt_nexthop(rt
, iph
->daddr
),
1197 memcpy(addr
, &src
, 4);
1200 #ifdef CONFIG_IP_ROUTE_CLASSID
1201 static void set_class_tag(struct rtable
*rt
, u32 tag
)
1203 if (!(rt
->dst
.tclassid
& 0xFFFF))
1204 rt
->dst
.tclassid
|= tag
& 0xFFFF;
1205 if (!(rt
->dst
.tclassid
& 0xFFFF0000))
1206 rt
->dst
.tclassid
|= tag
& 0xFFFF0000;
1210 static unsigned int ipv4_default_advmss(const struct dst_entry
*dst
)
1212 unsigned int advmss
= dst_metric_raw(dst
, RTAX_ADVMSS
);
1215 advmss
= max_t(unsigned int, dst
->dev
->mtu
- 40,
1217 if (advmss
> 65535 - 40)
1218 advmss
= 65535 - 40;
1223 static unsigned int ipv4_mtu(const struct dst_entry
*dst
)
1225 const struct rtable
*rt
= (const struct rtable
*) dst
;
1226 unsigned int mtu
= rt
->rt_pmtu
;
1228 if (!mtu
|| time_after_eq(jiffies
, rt
->dst
.expires
))
1229 mtu
= dst_metric_raw(dst
, RTAX_MTU
);
1234 mtu
= dst
->dev
->mtu
;
1236 if (unlikely(dst_metric_locked(dst
, RTAX_MTU
))) {
1237 if (rt
->rt_uses_gateway
&& mtu
> 576)
1241 return min_t(unsigned int, mtu
, IP_MAX_MTU
);
1244 static struct fib_nh_exception
*find_exception(struct fib_nh
*nh
, __be32 daddr
)
1246 struct fnhe_hash_bucket
*hash
= rcu_dereference(nh
->nh_exceptions
);
1247 struct fib_nh_exception
*fnhe
;
1253 hval
= fnhe_hashfun(daddr
);
1255 for (fnhe
= rcu_dereference(hash
[hval
].chain
); fnhe
;
1256 fnhe
= rcu_dereference(fnhe
->fnhe_next
)) {
1257 if (fnhe
->fnhe_daddr
== daddr
)
1263 static bool rt_bind_exception(struct rtable
*rt
, struct fib_nh_exception
*fnhe
,
1268 spin_lock_bh(&fnhe_lock
);
1270 if (daddr
== fnhe
->fnhe_daddr
) {
1271 struct rtable __rcu
**porig
;
1272 struct rtable
*orig
;
1273 int genid
= fnhe_genid(dev_net(rt
->dst
.dev
));
1275 if (rt_is_input_route(rt
))
1276 porig
= &fnhe
->fnhe_rth_input
;
1278 porig
= &fnhe
->fnhe_rth_output
;
1279 orig
= rcu_dereference(*porig
);
1281 if (fnhe
->fnhe_genid
!= genid
) {
1282 fnhe
->fnhe_genid
= genid
;
1284 fnhe
->fnhe_pmtu
= 0;
1285 fnhe
->fnhe_expires
= 0;
1286 fnhe_flush_routes(fnhe
);
1289 fill_route_from_fnhe(rt
, fnhe
);
1290 if (!rt
->rt_gateway
)
1291 rt
->rt_gateway
= daddr
;
1293 if (!(rt
->dst
.flags
& DST_NOCACHE
)) {
1294 rcu_assign_pointer(*porig
, rt
);
1300 fnhe
->fnhe_stamp
= jiffies
;
1302 spin_unlock_bh(&fnhe_lock
);
1307 static bool rt_cache_route(struct fib_nh
*nh
, struct rtable
*rt
)
1309 struct rtable
*orig
, *prev
, **p
;
1312 if (rt_is_input_route(rt
)) {
1313 p
= (struct rtable
**)&nh
->nh_rth_input
;
1315 p
= (struct rtable
**)raw_cpu_ptr(nh
->nh_pcpu_rth_output
);
1319 prev
= cmpxchg(p
, orig
, rt
);
1329 struct uncached_list
{
1331 struct list_head head
;
1334 static DEFINE_PER_CPU_ALIGNED(struct uncached_list
, rt_uncached_list
);
1336 static void rt_add_uncached_list(struct rtable
*rt
)
1338 struct uncached_list
*ul
= raw_cpu_ptr(&rt_uncached_list
);
1340 rt
->rt_uncached_list
= ul
;
1342 spin_lock_bh(&ul
->lock
);
1343 list_add_tail(&rt
->rt_uncached
, &ul
->head
);
1344 spin_unlock_bh(&ul
->lock
);
1347 static void ipv4_dst_destroy(struct dst_entry
*dst
)
1349 struct rtable
*rt
= (struct rtable
*) dst
;
1351 if (!list_empty(&rt
->rt_uncached
)) {
1352 struct uncached_list
*ul
= rt
->rt_uncached_list
;
1354 spin_lock_bh(&ul
->lock
);
1355 list_del(&rt
->rt_uncached
);
1356 spin_unlock_bh(&ul
->lock
);
1360 void rt_flush_dev(struct net_device
*dev
)
1362 struct net
*net
= dev_net(dev
);
1366 for_each_possible_cpu(cpu
) {
1367 struct uncached_list
*ul
= &per_cpu(rt_uncached_list
, cpu
);
1369 spin_lock_bh(&ul
->lock
);
1370 list_for_each_entry(rt
, &ul
->head
, rt_uncached
) {
1371 if (rt
->dst
.dev
!= dev
)
1373 rt
->dst
.dev
= net
->loopback_dev
;
1374 dev_hold(rt
->dst
.dev
);
1377 spin_unlock_bh(&ul
->lock
);
1381 static bool rt_cache_valid(const struct rtable
*rt
)
1384 rt
->dst
.obsolete
== DST_OBSOLETE_FORCE_CHK
&&
1388 static void rt_set_nexthop(struct rtable
*rt
, __be32 daddr
,
1389 const struct fib_result
*res
,
1390 struct fib_nh_exception
*fnhe
,
1391 struct fib_info
*fi
, u16 type
, u32 itag
)
1393 bool cached
= false;
1396 struct fib_nh
*nh
= &FIB_RES_NH(*res
);
1398 if (nh
->nh_gw
&& nh
->nh_scope
== RT_SCOPE_LINK
) {
1399 rt
->rt_gateway
= nh
->nh_gw
;
1400 rt
->rt_uses_gateway
= 1;
1402 dst_init_metrics(&rt
->dst
, fi
->fib_metrics
, true);
1403 #ifdef CONFIG_IP_ROUTE_CLASSID
1404 rt
->dst
.tclassid
= nh
->nh_tclassid
;
1407 cached
= rt_bind_exception(rt
, fnhe
, daddr
);
1408 else if (!(rt
->dst
.flags
& DST_NOCACHE
))
1409 cached
= rt_cache_route(nh
, rt
);
1410 if (unlikely(!cached
)) {
1411 /* Routes we intend to cache in nexthop exception or
1412 * FIB nexthop have the DST_NOCACHE bit clear.
1413 * However, if we are unsuccessful at storing this
1414 * route into the cache we really need to set it.
1416 rt
->dst
.flags
|= DST_NOCACHE
;
1417 if (!rt
->rt_gateway
)
1418 rt
->rt_gateway
= daddr
;
1419 rt_add_uncached_list(rt
);
1422 rt_add_uncached_list(rt
);
1424 #ifdef CONFIG_IP_ROUTE_CLASSID
1425 #ifdef CONFIG_IP_MULTIPLE_TABLES
1426 set_class_tag(rt
, res
->tclassid
);
1428 set_class_tag(rt
, itag
);
1432 static struct rtable
*rt_dst_alloc(struct net_device
*dev
,
1433 bool nopolicy
, bool noxfrm
, bool will_cache
)
1435 return dst_alloc(&ipv4_dst_ops
, dev
, 1, DST_OBSOLETE_FORCE_CHK
,
1436 (will_cache
? 0 : (DST_HOST
| DST_NOCACHE
)) |
1437 (nopolicy
? DST_NOPOLICY
: 0) |
1438 (noxfrm
? DST_NOXFRM
: 0));
1441 /* called in rcu_read_lock() section */
1442 static int ip_route_input_mc(struct sk_buff
*skb
, __be32 daddr
, __be32 saddr
,
1443 u8 tos
, struct net_device
*dev
, int our
)
1446 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
1450 /* Primary sanity checks. */
1455 if (ipv4_is_multicast(saddr
) || ipv4_is_lbcast(saddr
) ||
1456 skb
->protocol
!= htons(ETH_P_IP
))
1459 if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev
)))
1460 if (ipv4_is_loopback(saddr
))
1463 if (ipv4_is_zeronet(saddr
)) {
1464 if (!ipv4_is_local_multicast(daddr
))
1467 err
= fib_validate_source(skb
, saddr
, 0, tos
, 0, dev
,
1472 rth
= rt_dst_alloc(dev_net(dev
)->loopback_dev
,
1473 IN_DEV_CONF_GET(in_dev
, NOPOLICY
), false, false);
1477 #ifdef CONFIG_IP_ROUTE_CLASSID
1478 rth
->dst
.tclassid
= itag
;
1480 rth
->dst
.output
= ip_rt_bug
;
1482 rth
->rt_genid
= rt_genid_ipv4(dev_net(dev
));
1483 rth
->rt_flags
= RTCF_MULTICAST
;
1484 rth
->rt_type
= RTN_MULTICAST
;
1485 rth
->rt_is_input
= 1;
1488 rth
->rt_gateway
= 0;
1489 rth
->rt_uses_gateway
= 0;
1490 INIT_LIST_HEAD(&rth
->rt_uncached
);
1492 rth
->dst
.input
= ip_local_deliver
;
1493 rth
->rt_flags
|= RTCF_LOCAL
;
1496 #ifdef CONFIG_IP_MROUTE
1497 if (!ipv4_is_local_multicast(daddr
) && IN_DEV_MFORWARD(in_dev
))
1498 rth
->dst
.input
= ip_mr_input
;
1500 RT_CACHE_STAT_INC(in_slow_mc
);
1502 skb_dst_set(skb
, &rth
->dst
);
1514 static void ip_handle_martian_source(struct net_device
*dev
,
1515 struct in_device
*in_dev
,
1516 struct sk_buff
*skb
,
1520 RT_CACHE_STAT_INC(in_martian_src
);
1521 #ifdef CONFIG_IP_ROUTE_VERBOSE
1522 if (IN_DEV_LOG_MARTIANS(in_dev
) && net_ratelimit()) {
1524 * RFC1812 recommendation, if source is martian,
1525 * the only hint is MAC header.
1527 pr_warn("martian source %pI4 from %pI4, on dev %s\n",
1528 &daddr
, &saddr
, dev
->name
);
1529 if (dev
->hard_header_len
&& skb_mac_header_was_set(skb
)) {
1530 print_hex_dump(KERN_WARNING
, "ll header: ",
1531 DUMP_PREFIX_OFFSET
, 16, 1,
1532 skb_mac_header(skb
),
1533 dev
->hard_header_len
, true);
1539 /* called in rcu_read_lock() section */
1540 static int __mkroute_input(struct sk_buff
*skb
,
1541 const struct fib_result
*res
,
1542 struct in_device
*in_dev
,
1543 __be32 daddr
, __be32 saddr
, u32 tos
)
1545 struct fib_nh_exception
*fnhe
;
1548 struct in_device
*out_dev
;
1552 /* get a working reference to the output device */
1553 out_dev
= __in_dev_get_rcu(FIB_RES_DEV(*res
));
1555 net_crit_ratelimited("Bug in ip_route_input_slow(). Please report.\n");
1559 err
= fib_validate_source(skb
, saddr
, daddr
, tos
, FIB_RES_OIF(*res
),
1560 in_dev
->dev
, in_dev
, &itag
);
1562 ip_handle_martian_source(in_dev
->dev
, in_dev
, skb
, daddr
,
1568 do_cache
= res
->fi
&& !itag
;
1569 if (out_dev
== in_dev
&& err
&& IN_DEV_TX_REDIRECTS(out_dev
) &&
1570 skb
->protocol
== htons(ETH_P_IP
) &&
1571 (IN_DEV_SHARED_MEDIA(out_dev
) ||
1572 inet_addr_onlink(out_dev
, saddr
, FIB_RES_GW(*res
))))
1573 IPCB(skb
)->flags
|= IPSKB_DOREDIRECT
;
1575 if (skb
->protocol
!= htons(ETH_P_IP
)) {
1576 /* Not IP (i.e. ARP). Do not create route, if it is
1577 * invalid for proxy arp. DNAT routes are always valid.
1579 * Proxy arp feature have been extended to allow, ARP
1580 * replies back to the same interface, to support
1581 * Private VLAN switch technologies. See arp.c.
1583 if (out_dev
== in_dev
&&
1584 IN_DEV_PROXY_ARP_PVLAN(in_dev
) == 0) {
1590 fnhe
= find_exception(&FIB_RES_NH(*res
), daddr
);
1593 rth
= rcu_dereference(fnhe
->fnhe_rth_input
);
1595 rth
= rcu_dereference(FIB_RES_NH(*res
).nh_rth_input
);
1597 if (rt_cache_valid(rth
)) {
1598 skb_dst_set_noref(skb
, &rth
->dst
);
1603 rth
= rt_dst_alloc(out_dev
->dev
,
1604 IN_DEV_CONF_GET(in_dev
, NOPOLICY
),
1605 IN_DEV_CONF_GET(out_dev
, NOXFRM
), do_cache
);
1611 rth
->rt_genid
= rt_genid_ipv4(dev_net(rth
->dst
.dev
));
1613 rth
->rt_type
= res
->type
;
1614 rth
->rt_is_input
= 1;
1617 rth
->rt_gateway
= 0;
1618 rth
->rt_uses_gateway
= 0;
1619 INIT_LIST_HEAD(&rth
->rt_uncached
);
1620 RT_CACHE_STAT_INC(in_slow_tot
);
1622 rth
->dst
.input
= ip_forward
;
1623 rth
->dst
.output
= ip_output
;
1625 rt_set_nexthop(rth
, daddr
, res
, fnhe
, res
->fi
, res
->type
, itag
);
1626 skb_dst_set(skb
, &rth
->dst
);
1633 static int ip_mkroute_input(struct sk_buff
*skb
,
1634 struct fib_result
*res
,
1635 const struct flowi4
*fl4
,
1636 struct in_device
*in_dev
,
1637 __be32 daddr
, __be32 saddr
, u32 tos
)
1639 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1640 if (res
->fi
&& res
->fi
->fib_nhs
> 1)
1641 fib_select_multipath(res
);
1644 /* create a routing cache entry */
1645 return __mkroute_input(skb
, res
, in_dev
, daddr
, saddr
, tos
);
1649 * NOTE. We drop all the packets that has local source
1650 * addresses, because every properly looped back packet
1651 * must have correct destination already attached by output routine.
1653 * Such approach solves two big problems:
1654 * 1. Not simplex devices are handled properly.
1655 * 2. IP spoofing attempts are filtered with 100% of guarantee.
1656 * called with rcu_read_lock()
1659 static int ip_route_input_slow(struct sk_buff
*skb
, __be32 daddr
, __be32 saddr
,
1660 u8 tos
, struct net_device
*dev
)
1662 struct fib_result res
;
1663 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
1665 unsigned int flags
= 0;
1669 struct net
*net
= dev_net(dev
);
1672 /* IP on this device is disabled. */
1677 /* Check for the most weird martians, which can be not detected
1681 if (ipv4_is_multicast(saddr
) || ipv4_is_lbcast(saddr
))
1682 goto martian_source
;
1685 if (ipv4_is_lbcast(daddr
) || (saddr
== 0 && daddr
== 0))
1688 /* Accept zero addresses only to limited broadcast;
1689 * I even do not know to fix it or not. Waiting for complains :-)
1691 if (ipv4_is_zeronet(saddr
))
1692 goto martian_source
;
1694 if (ipv4_is_zeronet(daddr
))
1695 goto martian_destination
;
1697 /* Following code try to avoid calling IN_DEV_NET_ROUTE_LOCALNET(),
1698 * and call it once if daddr or/and saddr are loopback addresses
1700 if (ipv4_is_loopback(daddr
)) {
1701 if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev
, net
))
1702 goto martian_destination
;
1703 } else if (ipv4_is_loopback(saddr
)) {
1704 if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev
, net
))
1705 goto martian_source
;
1709 * Now we are ready to route packet.
1712 fl4
.flowi4_iif
= dev
->ifindex
;
1713 fl4
.flowi4_mark
= skb
->mark
;
1714 fl4
.flowi4_tos
= tos
;
1715 fl4
.flowi4_scope
= RT_SCOPE_UNIVERSE
;
1718 err
= fib_lookup(net
, &fl4
, &res
, 0);
1720 if (!IN_DEV_FORWARD(in_dev
))
1721 err
= -EHOSTUNREACH
;
1725 if (res
.type
== RTN_BROADCAST
)
1728 if (res
.type
== RTN_LOCAL
) {
1729 err
= fib_validate_source(skb
, saddr
, daddr
, tos
,
1730 0, dev
, in_dev
, &itag
);
1732 goto martian_source_keep_err
;
1736 if (!IN_DEV_FORWARD(in_dev
)) {
1737 err
= -EHOSTUNREACH
;
1740 if (res
.type
!= RTN_UNICAST
)
1741 goto martian_destination
;
1743 err
= ip_mkroute_input(skb
, &res
, &fl4
, in_dev
, daddr
, saddr
, tos
);
1747 if (skb
->protocol
!= htons(ETH_P_IP
))
1750 if (!ipv4_is_zeronet(saddr
)) {
1751 err
= fib_validate_source(skb
, saddr
, 0, tos
, 0, dev
,
1754 goto martian_source_keep_err
;
1756 flags
|= RTCF_BROADCAST
;
1757 res
.type
= RTN_BROADCAST
;
1758 RT_CACHE_STAT_INC(in_brd
);
1764 rth
= rcu_dereference(FIB_RES_NH(res
).nh_rth_input
);
1765 if (rt_cache_valid(rth
)) {
1766 skb_dst_set_noref(skb
, &rth
->dst
);
1774 rth
= rt_dst_alloc(net
->loopback_dev
,
1775 IN_DEV_CONF_GET(in_dev
, NOPOLICY
), false, do_cache
);
1779 rth
->dst
.input
= ip_local_deliver
;
1780 rth
->dst
.output
= ip_rt_bug
;
1781 #ifdef CONFIG_IP_ROUTE_CLASSID
1782 rth
->dst
.tclassid
= itag
;
1785 rth
->rt_genid
= rt_genid_ipv4(net
);
1786 rth
->rt_flags
= flags
|RTCF_LOCAL
;
1787 rth
->rt_type
= res
.type
;
1788 rth
->rt_is_input
= 1;
1791 rth
->rt_gateway
= 0;
1792 rth
->rt_uses_gateway
= 0;
1793 INIT_LIST_HEAD(&rth
->rt_uncached
);
1794 RT_CACHE_STAT_INC(in_slow_tot
);
1795 if (res
.type
== RTN_UNREACHABLE
) {
1796 rth
->dst
.input
= ip_error
;
1797 rth
->dst
.error
= -err
;
1798 rth
->rt_flags
&= ~RTCF_LOCAL
;
1801 if (unlikely(!rt_cache_route(&FIB_RES_NH(res
), rth
))) {
1802 rth
->dst
.flags
|= DST_NOCACHE
;
1803 rt_add_uncached_list(rth
);
1806 skb_dst_set(skb
, &rth
->dst
);
1811 RT_CACHE_STAT_INC(in_no_route
);
1812 res
.type
= RTN_UNREACHABLE
;
1817 * Do not cache martian addresses: they should be logged (RFC1812)
1819 martian_destination
:
1820 RT_CACHE_STAT_INC(in_martian_dst
);
1821 #ifdef CONFIG_IP_ROUTE_VERBOSE
1822 if (IN_DEV_LOG_MARTIANS(in_dev
))
1823 net_warn_ratelimited("martian destination %pI4 from %pI4, dev %s\n",
1824 &daddr
, &saddr
, dev
->name
);
1837 martian_source_keep_err
:
1838 ip_handle_martian_source(dev
, in_dev
, skb
, daddr
, saddr
);
1842 int ip_route_input_noref(struct sk_buff
*skb
, __be32 daddr
, __be32 saddr
,
1843 u8 tos
, struct net_device
*dev
)
1849 /* Multicast recognition logic is moved from route cache to here.
1850 The problem was that too many Ethernet cards have broken/missing
1851 hardware multicast filters :-( As result the host on multicasting
1852 network acquires a lot of useless route cache entries, sort of
1853 SDR messages from all the world. Now we try to get rid of them.
1854 Really, provided software IP multicast filter is organized
1855 reasonably (at least, hashed), it does not result in a slowdown
1856 comparing with route cache reject entries.
1857 Note, that multicast routers are not affected, because
1858 route cache entry is created eventually.
1860 if (ipv4_is_multicast(daddr
)) {
1861 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
1864 int our
= ip_check_mc_rcu(in_dev
, daddr
, saddr
,
1865 ip_hdr(skb
)->protocol
);
1867 #ifdef CONFIG_IP_MROUTE
1869 (!ipv4_is_local_multicast(daddr
) &&
1870 IN_DEV_MFORWARD(in_dev
))
1873 int res
= ip_route_input_mc(skb
, daddr
, saddr
,
1882 res
= ip_route_input_slow(skb
, daddr
, saddr
, tos
, dev
);
1886 EXPORT_SYMBOL(ip_route_input_noref
);
1888 /* called with rcu_read_lock() */
1889 static struct rtable
*__mkroute_output(const struct fib_result
*res
,
1890 const struct flowi4
*fl4
, int orig_oif
,
1891 struct net_device
*dev_out
,
1894 struct fib_info
*fi
= res
->fi
;
1895 struct fib_nh_exception
*fnhe
;
1896 struct in_device
*in_dev
;
1897 u16 type
= res
->type
;
1901 in_dev
= __in_dev_get_rcu(dev_out
);
1903 return ERR_PTR(-EINVAL
);
1905 if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev
)))
1906 if (ipv4_is_loopback(fl4
->saddr
) && !(dev_out
->flags
& IFF_LOOPBACK
))
1907 return ERR_PTR(-EINVAL
);
1909 if (ipv4_is_lbcast(fl4
->daddr
))
1910 type
= RTN_BROADCAST
;
1911 else if (ipv4_is_multicast(fl4
->daddr
))
1912 type
= RTN_MULTICAST
;
1913 else if (ipv4_is_zeronet(fl4
->daddr
))
1914 return ERR_PTR(-EINVAL
);
1916 if (dev_out
->flags
& IFF_LOOPBACK
)
1917 flags
|= RTCF_LOCAL
;
1920 if (type
== RTN_BROADCAST
) {
1921 flags
|= RTCF_BROADCAST
| RTCF_LOCAL
;
1923 } else if (type
== RTN_MULTICAST
) {
1924 flags
|= RTCF_MULTICAST
| RTCF_LOCAL
;
1925 if (!ip_check_mc_rcu(in_dev
, fl4
->daddr
, fl4
->saddr
,
1927 flags
&= ~RTCF_LOCAL
;
1930 /* If multicast route do not exist use
1931 * default one, but do not gateway in this case.
1934 if (fi
&& res
->prefixlen
< 4)
1939 do_cache
&= fi
!= NULL
;
1941 struct rtable __rcu
**prth
;
1942 struct fib_nh
*nh
= &FIB_RES_NH(*res
);
1944 fnhe
= find_exception(nh
, fl4
->daddr
);
1946 prth
= &fnhe
->fnhe_rth_output
;
1948 if (unlikely(fl4
->flowi4_flags
&
1949 FLOWI_FLAG_KNOWN_NH
&&
1951 nh
->nh_scope
== RT_SCOPE_LINK
))) {
1955 prth
= raw_cpu_ptr(nh
->nh_pcpu_rth_output
);
1957 rth
= rcu_dereference(*prth
);
1958 if (rt_cache_valid(rth
)) {
1959 dst_hold(&rth
->dst
);
1965 rth
= rt_dst_alloc(dev_out
,
1966 IN_DEV_CONF_GET(in_dev
, NOPOLICY
),
1967 IN_DEV_CONF_GET(in_dev
, NOXFRM
),
1970 return ERR_PTR(-ENOBUFS
);
1972 rth
->dst
.output
= ip_output
;
1974 rth
->rt_genid
= rt_genid_ipv4(dev_net(dev_out
));
1975 rth
->rt_flags
= flags
;
1976 rth
->rt_type
= type
;
1977 rth
->rt_is_input
= 0;
1978 rth
->rt_iif
= orig_oif
? : 0;
1980 rth
->rt_gateway
= 0;
1981 rth
->rt_uses_gateway
= 0;
1982 INIT_LIST_HEAD(&rth
->rt_uncached
);
1984 RT_CACHE_STAT_INC(out_slow_tot
);
1986 if (flags
& RTCF_LOCAL
)
1987 rth
->dst
.input
= ip_local_deliver
;
1988 if (flags
& (RTCF_BROADCAST
| RTCF_MULTICAST
)) {
1989 if (flags
& RTCF_LOCAL
&&
1990 !(dev_out
->flags
& IFF_LOOPBACK
)) {
1991 rth
->dst
.output
= ip_mc_output
;
1992 RT_CACHE_STAT_INC(out_slow_mc
);
1994 #ifdef CONFIG_IP_MROUTE
1995 if (type
== RTN_MULTICAST
) {
1996 if (IN_DEV_MFORWARD(in_dev
) &&
1997 !ipv4_is_local_multicast(fl4
->daddr
)) {
1998 rth
->dst
.input
= ip_mr_input
;
1999 rth
->dst
.output
= ip_mc_output
;
2005 rt_set_nexthop(rth
, fl4
->daddr
, res
, fnhe
, fi
, type
, 0);
2011 * Major route resolver routine.
2014 struct rtable
*__ip_route_output_key(struct net
*net
, struct flowi4
*fl4
)
2016 struct net_device
*dev_out
= NULL
;
2017 __u8 tos
= RT_FL_TOS(fl4
);
2018 unsigned int flags
= 0;
2019 struct fib_result res
;
2027 orig_oif
= fl4
->flowi4_oif
;
2029 fl4
->flowi4_iif
= LOOPBACK_IFINDEX
;
2030 fl4
->flowi4_tos
= tos
& IPTOS_RT_MASK
;
2031 fl4
->flowi4_scope
= ((tos
& RTO_ONLINK
) ?
2032 RT_SCOPE_LINK
: RT_SCOPE_UNIVERSE
);
2036 rth
= ERR_PTR(-EINVAL
);
2037 if (ipv4_is_multicast(fl4
->saddr
) ||
2038 ipv4_is_lbcast(fl4
->saddr
) ||
2039 ipv4_is_zeronet(fl4
->saddr
))
2042 /* I removed check for oif == dev_out->oif here.
2043 It was wrong for two reasons:
2044 1. ip_dev_find(net, saddr) can return wrong iface, if saddr
2045 is assigned to multiple interfaces.
2046 2. Moreover, we are allowed to send packets with saddr
2047 of another iface. --ANK
2050 if (fl4
->flowi4_oif
== 0 &&
2051 (ipv4_is_multicast(fl4
->daddr
) ||
2052 ipv4_is_lbcast(fl4
->daddr
))) {
2053 /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2054 dev_out
= __ip_dev_find(net
, fl4
->saddr
, false);
2058 /* Special hack: user can direct multicasts
2059 and limited broadcast via necessary interface
2060 without fiddling with IP_MULTICAST_IF or IP_PKTINFO.
2061 This hack is not just for fun, it allows
2062 vic,vat and friends to work.
2063 They bind socket to loopback, set ttl to zero
2064 and expect that it will work.
2065 From the viewpoint of routing cache they are broken,
2066 because we are not allowed to build multicast path
2067 with loopback source addr (look, routing cache
2068 cannot know, that ttl is zero, so that packet
2069 will not leave this host and route is valid).
2070 Luckily, this hack is good workaround.
2073 fl4
->flowi4_oif
= dev_out
->ifindex
;
2077 if (!(fl4
->flowi4_flags
& FLOWI_FLAG_ANYSRC
)) {
2078 /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2079 if (!__ip_dev_find(net
, fl4
->saddr
, false))
2085 if (fl4
->flowi4_oif
) {
2086 dev_out
= dev_get_by_index_rcu(net
, fl4
->flowi4_oif
);
2087 rth
= ERR_PTR(-ENODEV
);
2091 /* RACE: Check return value of inet_select_addr instead. */
2092 if (!(dev_out
->flags
& IFF_UP
) || !__in_dev_get_rcu(dev_out
)) {
2093 rth
= ERR_PTR(-ENETUNREACH
);
2096 if (ipv4_is_local_multicast(fl4
->daddr
) ||
2097 ipv4_is_lbcast(fl4
->daddr
) ||
2098 fl4
->flowi4_proto
== IPPROTO_IGMP
) {
2100 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2105 if (ipv4_is_multicast(fl4
->daddr
))
2106 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2108 else if (!fl4
->daddr
)
2109 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2115 fl4
->daddr
= fl4
->saddr
;
2117 fl4
->daddr
= fl4
->saddr
= htonl(INADDR_LOOPBACK
);
2118 dev_out
= net
->loopback_dev
;
2119 fl4
->flowi4_oif
= LOOPBACK_IFINDEX
;
2120 res
.type
= RTN_LOCAL
;
2121 flags
|= RTCF_LOCAL
;
2125 if (fib_lookup(net
, fl4
, &res
, 0)) {
2128 if (fl4
->flowi4_oif
) {
2129 /* Apparently, routing tables are wrong. Assume,
2130 that the destination is on link.
2133 Because we are allowed to send to iface
2134 even if it has NO routes and NO assigned
2135 addresses. When oif is specified, routing
2136 tables are looked up with only one purpose:
2137 to catch if destination is gatewayed, rather than
2138 direct. Moreover, if MSG_DONTROUTE is set,
2139 we send packet, ignoring both routing tables
2140 and ifaddr state. --ANK
2143 We could make it even if oif is unknown,
2144 likely IPv6, but we do not.
2147 if (fl4
->saddr
== 0)
2148 fl4
->saddr
= inet_select_addr(dev_out
, 0,
2150 res
.type
= RTN_UNICAST
;
2153 rth
= ERR_PTR(-ENETUNREACH
);
2157 if (res
.type
== RTN_LOCAL
) {
2159 if (res
.fi
->fib_prefsrc
)
2160 fl4
->saddr
= res
.fi
->fib_prefsrc
;
2162 fl4
->saddr
= fl4
->daddr
;
2164 dev_out
= net
->loopback_dev
;
2165 fl4
->flowi4_oif
= dev_out
->ifindex
;
2166 flags
|= RTCF_LOCAL
;
2170 #ifdef CONFIG_IP_ROUTE_MULTIPATH
2171 if (res
.fi
->fib_nhs
> 1 && fl4
->flowi4_oif
== 0)
2172 fib_select_multipath(&res
);
2175 if (!res
.prefixlen
&&
2176 res
.table
->tb_num_default
> 1 &&
2177 res
.type
== RTN_UNICAST
&& !fl4
->flowi4_oif
)
2178 fib_select_default(&res
);
2181 fl4
->saddr
= FIB_RES_PREFSRC(net
, res
);
2183 dev_out
= FIB_RES_DEV(res
);
2184 fl4
->flowi4_oif
= dev_out
->ifindex
;
2188 rth
= __mkroute_output(&res
, fl4
, orig_oif
, dev_out
, flags
);
2194 EXPORT_SYMBOL_GPL(__ip_route_output_key
);
2196 static struct dst_entry
*ipv4_blackhole_dst_check(struct dst_entry
*dst
, u32 cookie
)
2201 static unsigned int ipv4_blackhole_mtu(const struct dst_entry
*dst
)
2203 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
2205 return mtu
? : dst
->dev
->mtu
;
2208 static void ipv4_rt_blackhole_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
2209 struct sk_buff
*skb
, u32 mtu
)
2213 static void ipv4_rt_blackhole_redirect(struct dst_entry
*dst
, struct sock
*sk
,
2214 struct sk_buff
*skb
)
2218 static u32
*ipv4_rt_blackhole_cow_metrics(struct dst_entry
*dst
,
2224 static struct dst_ops ipv4_dst_blackhole_ops
= {
2226 .check
= ipv4_blackhole_dst_check
,
2227 .mtu
= ipv4_blackhole_mtu
,
2228 .default_advmss
= ipv4_default_advmss
,
2229 .update_pmtu
= ipv4_rt_blackhole_update_pmtu
,
2230 .redirect
= ipv4_rt_blackhole_redirect
,
2231 .cow_metrics
= ipv4_rt_blackhole_cow_metrics
,
2232 .neigh_lookup
= ipv4_neigh_lookup
,
2235 struct dst_entry
*ipv4_blackhole_route(struct net
*net
, struct dst_entry
*dst_orig
)
2237 struct rtable
*ort
= (struct rtable
*) dst_orig
;
2240 rt
= dst_alloc(&ipv4_dst_blackhole_ops
, NULL
, 1, DST_OBSOLETE_NONE
, 0);
2242 struct dst_entry
*new = &rt
->dst
;
2245 new->input
= dst_discard
;
2246 new->output
= dst_discard_sk
;
2248 new->dev
= ort
->dst
.dev
;
2252 rt
->rt_is_input
= ort
->rt_is_input
;
2253 rt
->rt_iif
= ort
->rt_iif
;
2254 rt
->rt_pmtu
= ort
->rt_pmtu
;
2256 rt
->rt_genid
= rt_genid_ipv4(net
);
2257 rt
->rt_flags
= ort
->rt_flags
;
2258 rt
->rt_type
= ort
->rt_type
;
2259 rt
->rt_gateway
= ort
->rt_gateway
;
2260 rt
->rt_uses_gateway
= ort
->rt_uses_gateway
;
2262 INIT_LIST_HEAD(&rt
->rt_uncached
);
2267 dst_release(dst_orig
);
2269 return rt
? &rt
->dst
: ERR_PTR(-ENOMEM
);
2272 struct rtable
*ip_route_output_flow(struct net
*net
, struct flowi4
*flp4
,
2275 struct rtable
*rt
= __ip_route_output_key(net
, flp4
);
2280 if (flp4
->flowi4_proto
)
2281 rt
= (struct rtable
*)xfrm_lookup_route(net
, &rt
->dst
,
2282 flowi4_to_flowi(flp4
),
2287 EXPORT_SYMBOL_GPL(ip_route_output_flow
);
2289 static int rt_fill_info(struct net
*net
, __be32 dst
, __be32 src
,
2290 struct flowi4
*fl4
, struct sk_buff
*skb
, u32 portid
,
2291 u32 seq
, int event
, int nowait
, unsigned int flags
)
2293 struct rtable
*rt
= skb_rtable(skb
);
2295 struct nlmsghdr
*nlh
;
2296 unsigned long expires
= 0;
2298 u32 metrics
[RTAX_MAX
];
2300 nlh
= nlmsg_put(skb
, portid
, seq
, event
, sizeof(*r
), flags
);
2304 r
= nlmsg_data(nlh
);
2305 r
->rtm_family
= AF_INET
;
2306 r
->rtm_dst_len
= 32;
2308 r
->rtm_tos
= fl4
->flowi4_tos
;
2309 r
->rtm_table
= RT_TABLE_MAIN
;
2310 if (nla_put_u32(skb
, RTA_TABLE
, RT_TABLE_MAIN
))
2311 goto nla_put_failure
;
2312 r
->rtm_type
= rt
->rt_type
;
2313 r
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2314 r
->rtm_protocol
= RTPROT_UNSPEC
;
2315 r
->rtm_flags
= (rt
->rt_flags
& ~0xFFFF) | RTM_F_CLONED
;
2316 if (rt
->rt_flags
& RTCF_NOTIFY
)
2317 r
->rtm_flags
|= RTM_F_NOTIFY
;
2318 if (IPCB(skb
)->flags
& IPSKB_DOREDIRECT
)
2319 r
->rtm_flags
|= RTCF_DOREDIRECT
;
2321 if (nla_put_in_addr(skb
, RTA_DST
, dst
))
2322 goto nla_put_failure
;
2324 r
->rtm_src_len
= 32;
2325 if (nla_put_in_addr(skb
, RTA_SRC
, src
))
2326 goto nla_put_failure
;
2329 nla_put_u32(skb
, RTA_OIF
, rt
->dst
.dev
->ifindex
))
2330 goto nla_put_failure
;
2331 #ifdef CONFIG_IP_ROUTE_CLASSID
2332 if (rt
->dst
.tclassid
&&
2333 nla_put_u32(skb
, RTA_FLOW
, rt
->dst
.tclassid
))
2334 goto nla_put_failure
;
2336 if (!rt_is_input_route(rt
) &&
2337 fl4
->saddr
!= src
) {
2338 if (nla_put_in_addr(skb
, RTA_PREFSRC
, fl4
->saddr
))
2339 goto nla_put_failure
;
2341 if (rt
->rt_uses_gateway
&&
2342 nla_put_in_addr(skb
, RTA_GATEWAY
, rt
->rt_gateway
))
2343 goto nla_put_failure
;
2345 expires
= rt
->dst
.expires
;
2347 unsigned long now
= jiffies
;
2349 if (time_before(now
, expires
))
2355 memcpy(metrics
, dst_metrics_ptr(&rt
->dst
), sizeof(metrics
));
2356 if (rt
->rt_pmtu
&& expires
)
2357 metrics
[RTAX_MTU
- 1] = rt
->rt_pmtu
;
2358 if (rtnetlink_put_metrics(skb
, metrics
) < 0)
2359 goto nla_put_failure
;
2361 if (fl4
->flowi4_mark
&&
2362 nla_put_u32(skb
, RTA_MARK
, fl4
->flowi4_mark
))
2363 goto nla_put_failure
;
2365 error
= rt
->dst
.error
;
2367 if (rt_is_input_route(rt
)) {
2368 #ifdef CONFIG_IP_MROUTE
2369 if (ipv4_is_multicast(dst
) && !ipv4_is_local_multicast(dst
) &&
2370 IPV4_DEVCONF_ALL(net
, MC_FORWARDING
)) {
2371 int err
= ipmr_get_route(net
, skb
,
2372 fl4
->saddr
, fl4
->daddr
,
2378 goto nla_put_failure
;
2380 if (err
== -EMSGSIZE
)
2381 goto nla_put_failure
;
2387 if (nla_put_u32(skb
, RTA_IIF
, skb
->dev
->ifindex
))
2388 goto nla_put_failure
;
2391 if (rtnl_put_cacheinfo(skb
, &rt
->dst
, 0, expires
, error
) < 0)
2392 goto nla_put_failure
;
2394 nlmsg_end(skb
, nlh
);
2398 nlmsg_cancel(skb
, nlh
);
2402 static int inet_rtm_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
*nlh
)
2404 struct net
*net
= sock_net(in_skb
->sk
);
2406 struct nlattr
*tb
[RTA_MAX
+1];
2407 struct rtable
*rt
= NULL
;
2414 struct sk_buff
*skb
;
2416 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv4_policy
);
2420 rtm
= nlmsg_data(nlh
);
2422 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
2428 /* Reserve room for dummy headers, this skb can pass
2429 through good chunk of routing engine.
2431 skb_reset_mac_header(skb
);
2432 skb_reset_network_header(skb
);
2434 /* Bugfix: need to give ip_route_input enough of an IP header to not gag. */
2435 ip_hdr(skb
)->protocol
= IPPROTO_ICMP
;
2436 skb_reserve(skb
, MAX_HEADER
+ sizeof(struct iphdr
));
2438 src
= tb
[RTA_SRC
] ? nla_get_in_addr(tb
[RTA_SRC
]) : 0;
2439 dst
= tb
[RTA_DST
] ? nla_get_in_addr(tb
[RTA_DST
]) : 0;
2440 iif
= tb
[RTA_IIF
] ? nla_get_u32(tb
[RTA_IIF
]) : 0;
2441 mark
= tb
[RTA_MARK
] ? nla_get_u32(tb
[RTA_MARK
]) : 0;
2443 memset(&fl4
, 0, sizeof(fl4
));
2446 fl4
.flowi4_tos
= rtm
->rtm_tos
;
2447 fl4
.flowi4_oif
= tb
[RTA_OIF
] ? nla_get_u32(tb
[RTA_OIF
]) : 0;
2448 fl4
.flowi4_mark
= mark
;
2451 struct net_device
*dev
;
2453 dev
= __dev_get_by_index(net
, iif
);
2459 skb
->protocol
= htons(ETH_P_IP
);
2463 err
= ip_route_input(skb
, dst
, src
, rtm
->rtm_tos
, dev
);
2466 rt
= skb_rtable(skb
);
2467 if (err
== 0 && rt
->dst
.error
)
2468 err
= -rt
->dst
.error
;
2470 rt
= ip_route_output_key(net
, &fl4
);
2480 skb_dst_set(skb
, &rt
->dst
);
2481 if (rtm
->rtm_flags
& RTM_F_NOTIFY
)
2482 rt
->rt_flags
|= RTCF_NOTIFY
;
2484 err
= rt_fill_info(net
, dst
, src
, &fl4
, skb
,
2485 NETLINK_CB(in_skb
).portid
, nlh
->nlmsg_seq
,
2486 RTM_NEWROUTE
, 0, 0);
2490 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).portid
);
2499 void ip_rt_multicast_event(struct in_device
*in_dev
)
2501 rt_cache_flush(dev_net(in_dev
->dev
));
2504 #ifdef CONFIG_SYSCTL
2505 static int ip_rt_gc_timeout __read_mostly
= RT_GC_TIMEOUT
;
2506 static int ip_rt_gc_interval __read_mostly
= 60 * HZ
;
2507 static int ip_rt_gc_min_interval __read_mostly
= HZ
/ 2;
2508 static int ip_rt_gc_elasticity __read_mostly
= 8;
2510 static int ipv4_sysctl_rtcache_flush(struct ctl_table
*__ctl
, int write
,
2511 void __user
*buffer
,
2512 size_t *lenp
, loff_t
*ppos
)
2514 struct net
*net
= (struct net
*)__ctl
->extra1
;
2517 rt_cache_flush(net
);
2518 fnhe_genid_bump(net
);
2525 static struct ctl_table ipv4_route_table
[] = {
2527 .procname
= "gc_thresh",
2528 .data
= &ipv4_dst_ops
.gc_thresh
,
2529 .maxlen
= sizeof(int),
2531 .proc_handler
= proc_dointvec
,
2534 .procname
= "max_size",
2535 .data
= &ip_rt_max_size
,
2536 .maxlen
= sizeof(int),
2538 .proc_handler
= proc_dointvec
,
2541 /* Deprecated. Use gc_min_interval_ms */
2543 .procname
= "gc_min_interval",
2544 .data
= &ip_rt_gc_min_interval
,
2545 .maxlen
= sizeof(int),
2547 .proc_handler
= proc_dointvec_jiffies
,
2550 .procname
= "gc_min_interval_ms",
2551 .data
= &ip_rt_gc_min_interval
,
2552 .maxlen
= sizeof(int),
2554 .proc_handler
= proc_dointvec_ms_jiffies
,
2557 .procname
= "gc_timeout",
2558 .data
= &ip_rt_gc_timeout
,
2559 .maxlen
= sizeof(int),
2561 .proc_handler
= proc_dointvec_jiffies
,
2564 .procname
= "gc_interval",
2565 .data
= &ip_rt_gc_interval
,
2566 .maxlen
= sizeof(int),
2568 .proc_handler
= proc_dointvec_jiffies
,
2571 .procname
= "redirect_load",
2572 .data
= &ip_rt_redirect_load
,
2573 .maxlen
= sizeof(int),
2575 .proc_handler
= proc_dointvec
,
2578 .procname
= "redirect_number",
2579 .data
= &ip_rt_redirect_number
,
2580 .maxlen
= sizeof(int),
2582 .proc_handler
= proc_dointvec
,
2585 .procname
= "redirect_silence",
2586 .data
= &ip_rt_redirect_silence
,
2587 .maxlen
= sizeof(int),
2589 .proc_handler
= proc_dointvec
,
2592 .procname
= "error_cost",
2593 .data
= &ip_rt_error_cost
,
2594 .maxlen
= sizeof(int),
2596 .proc_handler
= proc_dointvec
,
2599 .procname
= "error_burst",
2600 .data
= &ip_rt_error_burst
,
2601 .maxlen
= sizeof(int),
2603 .proc_handler
= proc_dointvec
,
2606 .procname
= "gc_elasticity",
2607 .data
= &ip_rt_gc_elasticity
,
2608 .maxlen
= sizeof(int),
2610 .proc_handler
= proc_dointvec
,
2613 .procname
= "mtu_expires",
2614 .data
= &ip_rt_mtu_expires
,
2615 .maxlen
= sizeof(int),
2617 .proc_handler
= proc_dointvec_jiffies
,
2620 .procname
= "min_pmtu",
2621 .data
= &ip_rt_min_pmtu
,
2622 .maxlen
= sizeof(int),
2624 .proc_handler
= proc_dointvec
,
2627 .procname
= "min_adv_mss",
2628 .data
= &ip_rt_min_advmss
,
2629 .maxlen
= sizeof(int),
2631 .proc_handler
= proc_dointvec
,
2636 static struct ctl_table ipv4_route_flush_table
[] = {
2638 .procname
= "flush",
2639 .maxlen
= sizeof(int),
2641 .proc_handler
= ipv4_sysctl_rtcache_flush
,
2646 static __net_init
int sysctl_route_net_init(struct net
*net
)
2648 struct ctl_table
*tbl
;
2650 tbl
= ipv4_route_flush_table
;
2651 if (!net_eq(net
, &init_net
)) {
2652 tbl
= kmemdup(tbl
, sizeof(ipv4_route_flush_table
), GFP_KERNEL
);
2656 /* Don't export sysctls to unprivileged users */
2657 if (net
->user_ns
!= &init_user_ns
)
2658 tbl
[0].procname
= NULL
;
2660 tbl
[0].extra1
= net
;
2662 net
->ipv4
.route_hdr
= register_net_sysctl(net
, "net/ipv4/route", tbl
);
2663 if (!net
->ipv4
.route_hdr
)
2668 if (tbl
!= ipv4_route_flush_table
)
2674 static __net_exit
void sysctl_route_net_exit(struct net
*net
)
2676 struct ctl_table
*tbl
;
2678 tbl
= net
->ipv4
.route_hdr
->ctl_table_arg
;
2679 unregister_net_sysctl_table(net
->ipv4
.route_hdr
);
2680 BUG_ON(tbl
== ipv4_route_flush_table
);
2684 static __net_initdata
struct pernet_operations sysctl_route_ops
= {
2685 .init
= sysctl_route_net_init
,
2686 .exit
= sysctl_route_net_exit
,
2690 static __net_init
int rt_genid_init(struct net
*net
)
2692 atomic_set(&net
->ipv4
.rt_genid
, 0);
2693 atomic_set(&net
->fnhe_genid
, 0);
2694 get_random_bytes(&net
->ipv4
.dev_addr_genid
,
2695 sizeof(net
->ipv4
.dev_addr_genid
));
2699 static __net_initdata
struct pernet_operations rt_genid_ops
= {
2700 .init
= rt_genid_init
,
2703 static int __net_init
ipv4_inetpeer_init(struct net
*net
)
2705 struct inet_peer_base
*bp
= kmalloc(sizeof(*bp
), GFP_KERNEL
);
2709 inet_peer_base_init(bp
);
2710 net
->ipv4
.peers
= bp
;
2714 static void __net_exit
ipv4_inetpeer_exit(struct net
*net
)
2716 struct inet_peer_base
*bp
= net
->ipv4
.peers
;
2718 net
->ipv4
.peers
= NULL
;
2719 inetpeer_invalidate_tree(bp
);
2723 static __net_initdata
struct pernet_operations ipv4_inetpeer_ops
= {
2724 .init
= ipv4_inetpeer_init
,
2725 .exit
= ipv4_inetpeer_exit
,
2728 #ifdef CONFIG_IP_ROUTE_CLASSID
2729 struct ip_rt_acct __percpu
*ip_rt_acct __read_mostly
;
2730 #endif /* CONFIG_IP_ROUTE_CLASSID */
2732 int __init
ip_rt_init(void)
2737 ip_idents
= kmalloc(IP_IDENTS_SZ
* sizeof(*ip_idents
), GFP_KERNEL
);
2739 panic("IP: failed to allocate ip_idents\n");
2741 prandom_bytes(ip_idents
, IP_IDENTS_SZ
* sizeof(*ip_idents
));
2743 ip_tstamps
= kcalloc(IP_IDENTS_SZ
, sizeof(*ip_tstamps
), GFP_KERNEL
);
2745 panic("IP: failed to allocate ip_tstamps\n");
2747 for_each_possible_cpu(cpu
) {
2748 struct uncached_list
*ul
= &per_cpu(rt_uncached_list
, cpu
);
2750 INIT_LIST_HEAD(&ul
->head
);
2751 spin_lock_init(&ul
->lock
);
2753 #ifdef CONFIG_IP_ROUTE_CLASSID
2754 ip_rt_acct
= __alloc_percpu(256 * sizeof(struct ip_rt_acct
), __alignof__(struct ip_rt_acct
));
2756 panic("IP: failed to allocate ip_rt_acct\n");
2759 ipv4_dst_ops
.kmem_cachep
=
2760 kmem_cache_create("ip_dst_cache", sizeof(struct rtable
), 0,
2761 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
);
2763 ipv4_dst_blackhole_ops
.kmem_cachep
= ipv4_dst_ops
.kmem_cachep
;
2765 if (dst_entries_init(&ipv4_dst_ops
) < 0)
2766 panic("IP: failed to allocate ipv4_dst_ops counter\n");
2768 if (dst_entries_init(&ipv4_dst_blackhole_ops
) < 0)
2769 panic("IP: failed to allocate ipv4_dst_blackhole_ops counter\n");
2771 ipv4_dst_ops
.gc_thresh
= ~0;
2772 ip_rt_max_size
= INT_MAX
;
2777 if (ip_rt_proc_init())
2778 pr_err("Unable to create route proc files\n");
2783 rtnl_register(PF_INET
, RTM_GETROUTE
, inet_rtm_getroute
, NULL
, NULL
);
2785 #ifdef CONFIG_SYSCTL
2786 register_pernet_subsys(&sysctl_route_ops
);
2788 register_pernet_subsys(&rt_genid_ops
);
2789 register_pernet_subsys(&ipv4_inetpeer_ops
);
2793 #ifdef CONFIG_SYSCTL
2795 * We really need to sanitize the damn ipv4 init order, then all
2796 * this nonsense will go away.
2798 void __init
ip_static_sysctl_init(void)
2800 register_net_sysctl(&init_net
, "net/ipv4/route", ipv4_route_table
);