]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame_incremental - net/ipv4/route.c
projects / mirror_ubuntu-artful-kernel.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
of: of_mdio: export symbol of_mdiobus_link_phydev
[mirror_ubuntu-artful-kernel.git] / net / ipv4 / route.c
... / ...
CommitLineData
1/*
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.
5 *
6 * ROUTE - implementation of the IP router.
7 *
8 * Authors: Ross Biro
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>
13 *
14 * Fixes:
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
24 * clamper.
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.
39 *
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
58 *
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.
63 */
64
65#define pr_fmt(fmt) "IPv4: " fmt
66
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>
72#include <linux/mm.h>
73#include <linux/string.h>
74#include <linux/socket.h>
75#include <linux/sockios.h>
76#include <linux/errno.h>
77#include <linux/in.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 <net/dst.h>
93#include <net/net_namespace.h>
94#include <net/protocol.h>
95#include <net/ip.h>
96#include <net/route.h>
97#include <net/inetpeer.h>
98#include <net/sock.h>
99#include <net/ip_fib.h>
100#include <net/arp.h>
101#include <net/tcp.h>
102#include <net/icmp.h>
103#include <net/xfrm.h>
104#include <net/netevent.h>
105#include <net/rtnetlink.h>
106#ifdef CONFIG_SYSCTL
107#include <linux/sysctl.h>
108#include <linux/kmemleak.h>
109#endif
110#include <net/secure_seq.h>
111
112#define RT_FL_TOS(oldflp4) \
113 ((oldflp4)->flowi4_tos & (IPTOS_RT_MASK | RTO_ONLINK))
114
115#define RT_GC_TIMEOUT (300*HZ)
116
117static int ip_rt_max_size;
118static int ip_rt_redirect_number __read_mostly = 9;
119static int ip_rt_redirect_load __read_mostly = HZ / 50;
120static int ip_rt_redirect_silence __read_mostly = ((HZ / 50) << (9 + 1));
121static int ip_rt_error_cost __read_mostly = HZ;
122static int ip_rt_error_burst __read_mostly = 5 * HZ;
123static int ip_rt_mtu_expires __read_mostly = 10 * 60 * HZ;
124static int ip_rt_min_pmtu __read_mostly = 512 + 20 + 20;
125static int ip_rt_min_advmss __read_mostly = 256;
126
127/*
128 * Interface to generic destination cache.
129 */
130
131static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie);
132static unsigned int ipv4_default_advmss(const struct dst_entry *dst);
133static unsigned int ipv4_mtu(const struct dst_entry *dst);
134static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst);
135static void ipv4_link_failure(struct sk_buff *skb);
136static void ip_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
137 struct sk_buff *skb, u32 mtu);
138static void ip_do_redirect(struct dst_entry *dst, struct sock *sk,
139 struct sk_buff *skb);
140static void ipv4_dst_destroy(struct dst_entry *dst);
141
142static u32 *ipv4_cow_metrics(struct dst_entry *dst, unsigned long old)
143{
144 WARN_ON(1);
145 return NULL;
146}
147
148static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst,
149 struct sk_buff *skb,
150 const void *daddr);
151
152static struct dst_ops ipv4_dst_ops = {
153 .family = AF_INET,
154 .protocol = cpu_to_be16(ETH_P_IP),
155 .check = ipv4_dst_check,
156 .default_advmss = ipv4_default_advmss,
157 .mtu = ipv4_mtu,
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,
166};
167
168#define ECN_OR_COST(class) TC_PRIO_##class
169
170const __u8 ip_tos2prio[16] = {
171 TC_PRIO_BESTEFFORT,
172 ECN_OR_COST(BESTEFFORT),
173 TC_PRIO_BESTEFFORT,
174 ECN_OR_COST(BESTEFFORT),
175 TC_PRIO_BULK,
176 ECN_OR_COST(BULK),
177 TC_PRIO_BULK,
178 ECN_OR_COST(BULK),
179 TC_PRIO_INTERACTIVE,
180 ECN_OR_COST(INTERACTIVE),
181 TC_PRIO_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)
187};
188EXPORT_SYMBOL(ip_tos2prio);
189
190static DEFINE_PER_CPU(struct rt_cache_stat, rt_cache_stat);
191#define RT_CACHE_STAT_INC(field) raw_cpu_inc(rt_cache_stat.field)
192
193#ifdef CONFIG_PROC_FS
194static void *rt_cache_seq_start(struct seq_file *seq, loff_t *pos)
195{
196 if (*pos)
197 return NULL;
198 return SEQ_START_TOKEN;
199}
200
201static void *rt_cache_seq_next(struct seq_file *seq, void *v, loff_t *pos)
202{
203 ++*pos;
204 return NULL;
205}
206
207static void rt_cache_seq_stop(struct seq_file *seq, void *v)
208{
209}
210
211static int rt_cache_seq_show(struct seq_file *seq, void *v)
212{
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"
217 "HHUptod\tSpecDst");
218 return 0;
219}
220
221static 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,
226};
227
228static int rt_cache_seq_open(struct inode *inode, struct file *file)
229{
230 return seq_open(file, &rt_cache_seq_ops);
231}
232
233static const struct file_operations rt_cache_seq_fops = {
234 .owner = THIS_MODULE,
235 .open = rt_cache_seq_open,
236 .read = seq_read,
237 .llseek = seq_lseek,
238 .release = seq_release,
239};
240
241
242static void *rt_cpu_seq_start(struct seq_file *seq, loff_t *pos)
243{
244 int cpu;
245
246 if (*pos == 0)
247 return SEQ_START_TOKEN;
248
249 for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
250 if (!cpu_possible(cpu))
251 continue;
252 *pos = cpu+1;
253 return &per_cpu(rt_cache_stat, cpu);
254 }
255 return NULL;
256}
257
258static void *rt_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
259{
260 int cpu;
261
262 for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
263 if (!cpu_possible(cpu))
264 continue;
265 *pos = cpu+1;
266 return &per_cpu(rt_cache_stat, cpu);
267 }
268 return NULL;
269
270}
271
272static void rt_cpu_seq_stop(struct seq_file *seq, void *v)
273{
274
275}
276
277static int rt_cpu_seq_show(struct seq_file *seq, void *v)
278{
279 struct rt_cache_stat *st = v;
280
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");
283 return 0;
284 }
285
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),
289 0, /* st->in_hit */
290 st->in_slow_tot,
291 st->in_slow_mc,
292 st->in_no_route,
293 st->in_brd,
294 st->in_martian_dst,
295 st->in_martian_src,
296
297 0, /* st->out_hit */
298 st->out_slow_tot,
299 st->out_slow_mc,
300
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 */
307 );
308 return 0;
309}
310
311static 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,
316};
317
318
319static int rt_cpu_seq_open(struct inode *inode, struct file *file)
320{
321 return seq_open(file, &rt_cpu_seq_ops);
322}
323
324static const struct file_operations rt_cpu_seq_fops = {
325 .owner = THIS_MODULE,
326 .open = rt_cpu_seq_open,
327 .read = seq_read,
328 .llseek = seq_lseek,
329 .release = seq_release,
330};
331
332#ifdef CONFIG_IP_ROUTE_CLASSID
333static int rt_acct_proc_show(struct seq_file *m, void *v)
334{
335 struct ip_rt_acct *dst, *src;
336 unsigned int i, j;
337
338 dst = kcalloc(256, sizeof(struct ip_rt_acct), GFP_KERNEL);
339 if (!dst)
340 return -ENOMEM;
341
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;
349 }
350 }
351
352 seq_write(m, dst, 256 * sizeof(struct ip_rt_acct));
353 kfree(dst);
354 return 0;
355}
356
357static int rt_acct_proc_open(struct inode *inode, struct file *file)
358{
359 return single_open(file, rt_acct_proc_show, NULL);
360}
361
362static const struct file_operations rt_acct_proc_fops = {
363 .owner = THIS_MODULE,
364 .open = rt_acct_proc_open,
365 .read = seq_read,
366 .llseek = seq_lseek,
367 .release = single_release,
368};
369#endif
370
371static int __net_init ip_rt_do_proc_init(struct net *net)
372{
373 struct proc_dir_entry *pde;
374
375 pde = proc_create("rt_cache", S_IRUGO, net->proc_net,
376 &rt_cache_seq_fops);
377 if (!pde)
378 goto err1;
379
380 pde = proc_create("rt_cache", S_IRUGO,
381 net->proc_net_stat, &rt_cpu_seq_fops);
382 if (!pde)
383 goto err2;
384
385#ifdef CONFIG_IP_ROUTE_CLASSID
386 pde = proc_create("rt_acct", 0, net->proc_net, &rt_acct_proc_fops);
387 if (!pde)
388 goto err3;
389#endif
390 return 0;
391
392#ifdef CONFIG_IP_ROUTE_CLASSID
393err3:
394 remove_proc_entry("rt_cache", net->proc_net_stat);
395#endif
396err2:
397 remove_proc_entry("rt_cache", net->proc_net);
398err1:
399 return -ENOMEM;
400}
401
402static void __net_exit ip_rt_do_proc_exit(struct net *net)
403{
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);
408#endif
409}
410
411static struct pernet_operations ip_rt_proc_ops __net_initdata = {
412 .init = ip_rt_do_proc_init,
413 .exit = ip_rt_do_proc_exit,
414};
415
416static int __init ip_rt_proc_init(void)
417{
418 return register_pernet_subsys(&ip_rt_proc_ops);
419}
420
421#else
422static inline int ip_rt_proc_init(void)
423{
424 return 0;
425}
426#endif /* CONFIG_PROC_FS */
427
428static inline bool rt_is_expired(const struct rtable *rth)
429{
430 return rth->rt_genid != rt_genid_ipv4(dev_net(rth->dst.dev));
431}
432
433void rt_cache_flush(struct net *net)
434{
435 rt_genid_bump_ipv4(net);
436}
437
438static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst,
439 struct sk_buff *skb,
440 const void *daddr)
441{
442 struct net_device *dev = dst->dev;
443 const __be32 *pkey = daddr;
444 const struct rtable *rt;
445 struct neighbour *n;
446
447 rt = (const struct rtable *) dst;
448 if (rt->rt_gateway)
449 pkey = (const __be32 *) &rt->rt_gateway;
450 else if (skb)
451 pkey = &ip_hdr(skb)->daddr;
452
453 n = __ipv4_neigh_lookup(dev, *(__force u32 *)pkey);
454 if (n)
455 return n;
456 return neigh_create(&arp_tbl, pkey, dev);
457}
458
459/*
460 * Peer allocation may fail only in serious out-of-memory conditions. However
461 * we still can generate some output.
462 * Random ID selection looks a bit dangerous because we have no chances to
463 * select ID being unique in a reasonable period of time.
464 * But broken packet identifier may be better than no packet at all.
465 */
466static void ip_select_fb_ident(struct iphdr *iph)
467{
468 static DEFINE_SPINLOCK(ip_fb_id_lock);
469 static u32 ip_fallback_id;
470 u32 salt;
471
472 spin_lock_bh(&ip_fb_id_lock);
473 salt = secure_ip_id((__force __be32)ip_fallback_id ^ iph->daddr);
474 iph->id = htons(salt & 0xFFFF);
475 ip_fallback_id = salt;
476 spin_unlock_bh(&ip_fb_id_lock);
477}
478
479void __ip_select_ident(struct iphdr *iph, struct dst_entry *dst, int more)
480{
481 struct net *net = dev_net(dst->dev);
482 struct inet_peer *peer;
483
484 peer = inet_getpeer_v4(net->ipv4.peers, iph->daddr, 1);
485 if (peer) {
486 iph->id = htons(inet_getid(peer, more));
487 inet_putpeer(peer);
488 return;
489 }
490
491 ip_select_fb_ident(iph);
492}
493EXPORT_SYMBOL(__ip_select_ident);
494
495static void __build_flow_key(struct flowi4 *fl4, const struct sock *sk,
496 const struct iphdr *iph,
497 int oif, u8 tos,
498 u8 prot, u32 mark, int flow_flags)
499{
500 if (sk) {
501 const struct inet_sock *inet = inet_sk(sk);
502
503 oif = sk->sk_bound_dev_if;
504 mark = sk->sk_mark;
505 tos = RT_CONN_FLAGS(sk);
506 prot = inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol;
507 }
508 flowi4_init_output(fl4, oif, mark, tos,
509 RT_SCOPE_UNIVERSE, prot,
510 flow_flags,
511 iph->daddr, iph->saddr, 0, 0);
512}
513
514static void build_skb_flow_key(struct flowi4 *fl4, const struct sk_buff *skb,
515 const struct sock *sk)
516{
517 const struct iphdr *iph = ip_hdr(skb);
518 int oif = skb->dev->ifindex;
519 u8 tos = RT_TOS(iph->tos);
520 u8 prot = iph->protocol;
521 u32 mark = skb->mark;
522
523 __build_flow_key(fl4, sk, iph, oif, tos, prot, mark, 0);
524}
525
526static void build_sk_flow_key(struct flowi4 *fl4, const struct sock *sk)
527{
528 const struct inet_sock *inet = inet_sk(sk);
529 const struct ip_options_rcu *inet_opt;
530 __be32 daddr = inet->inet_daddr;
531
532 rcu_read_lock();
533 inet_opt = rcu_dereference(inet->inet_opt);
534 if (inet_opt && inet_opt->opt.srr)
535 daddr = inet_opt->opt.faddr;
536 flowi4_init_output(fl4, sk->sk_bound_dev_if, sk->sk_mark,
537 RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE,
538 inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol,
539 inet_sk_flowi_flags(sk),
540 daddr, inet->inet_saddr, 0, 0);
541 rcu_read_unlock();
542}
543
544static void ip_rt_build_flow_key(struct flowi4 *fl4, const struct sock *sk,
545 const struct sk_buff *skb)
546{
547 if (skb)
548 build_skb_flow_key(fl4, skb, sk);
549 else
550 build_sk_flow_key(fl4, sk);
551}
552
553static inline void rt_free(struct rtable *rt)
554{
555 call_rcu(&rt->dst.rcu_head, dst_rcu_free);
556}
557
558static DEFINE_SPINLOCK(fnhe_lock);
559
560static void fnhe_flush_routes(struct fib_nh_exception *fnhe)
561{
562 struct rtable *rt;
563
564 rt = rcu_dereference(fnhe->fnhe_rth_input);
565 if (rt) {
566 RCU_INIT_POINTER(fnhe->fnhe_rth_input, NULL);
567 rt_free(rt);
568 }
569 rt = rcu_dereference(fnhe->fnhe_rth_output);
570 if (rt) {
571 RCU_INIT_POINTER(fnhe->fnhe_rth_output, NULL);
572 rt_free(rt);
573 }
574}
575
576static struct fib_nh_exception *fnhe_oldest(struct fnhe_hash_bucket *hash)
577{
578 struct fib_nh_exception *fnhe, *oldest;
579
580 oldest = rcu_dereference(hash->chain);
581 for (fnhe = rcu_dereference(oldest->fnhe_next); fnhe;
582 fnhe = rcu_dereference(fnhe->fnhe_next)) {
583 if (time_before(fnhe->fnhe_stamp, oldest->fnhe_stamp))
584 oldest = fnhe;
585 }
586 fnhe_flush_routes(oldest);
587 return oldest;
588}
589
590static inline u32 fnhe_hashfun(__be32 daddr)
591{
592 u32 hval;
593
594 hval = (__force u32) daddr;
595 hval ^= (hval >> 11) ^ (hval >> 22);
596
597 return hval & (FNHE_HASH_SIZE - 1);
598}
599
600static void fill_route_from_fnhe(struct rtable *rt, struct fib_nh_exception *fnhe)
601{
602 rt->rt_pmtu = fnhe->fnhe_pmtu;
603 rt->dst.expires = fnhe->fnhe_expires;
604
605 if (fnhe->fnhe_gw) {
606 rt->rt_flags |= RTCF_REDIRECTED;
607 rt->rt_gateway = fnhe->fnhe_gw;
608 rt->rt_uses_gateway = 1;
609 }
610}
611
612static void update_or_create_fnhe(struct fib_nh *nh, __be32 daddr, __be32 gw,
613 u32 pmtu, unsigned long expires)
614{
615 struct fnhe_hash_bucket *hash;
616 struct fib_nh_exception *fnhe;
617 struct rtable *rt;
618 unsigned int i;
619 int depth;
620 u32 hval = fnhe_hashfun(daddr);
621
622 spin_lock_bh(&fnhe_lock);
623
624 hash = nh->nh_exceptions;
625 if (!hash) {
626 hash = kzalloc(FNHE_HASH_SIZE * sizeof(*hash), GFP_ATOMIC);
627 if (!hash)
628 goto out_unlock;
629 nh->nh_exceptions = hash;
630 }
631
632 hash += hval;
633
634 depth = 0;
635 for (fnhe = rcu_dereference(hash->chain); fnhe;
636 fnhe = rcu_dereference(fnhe->fnhe_next)) {
637 if (fnhe->fnhe_daddr == daddr)
638 break;
639 depth++;
640 }
641
642 if (fnhe) {
643 if (gw)
644 fnhe->fnhe_gw = gw;
645 if (pmtu) {
646 fnhe->fnhe_pmtu = pmtu;
647 fnhe->fnhe_expires = max(1UL, expires);
648 }
649 /* Update all cached dsts too */
650 rt = rcu_dereference(fnhe->fnhe_rth_input);
651 if (rt)
652 fill_route_from_fnhe(rt, fnhe);
653 rt = rcu_dereference(fnhe->fnhe_rth_output);
654 if (rt)
655 fill_route_from_fnhe(rt, fnhe);
656 } else {
657 if (depth > FNHE_RECLAIM_DEPTH)
658 fnhe = fnhe_oldest(hash);
659 else {
660 fnhe = kzalloc(sizeof(*fnhe), GFP_ATOMIC);
661 if (!fnhe)
662 goto out_unlock;
663
664 fnhe->fnhe_next = hash->chain;
665 rcu_assign_pointer(hash->chain, fnhe);
666 }
667 fnhe->fnhe_genid = fnhe_genid(dev_net(nh->nh_dev));
668 fnhe->fnhe_daddr = daddr;
669 fnhe->fnhe_gw = gw;
670 fnhe->fnhe_pmtu = pmtu;
671 fnhe->fnhe_expires = expires;
672
673 /* Exception created; mark the cached routes for the nexthop
674 * stale, so anyone caching it rechecks if this exception
675 * applies to them.
676 */
677 rt = rcu_dereference(nh->nh_rth_input);
678 if (rt)
679 rt->dst.obsolete = DST_OBSOLETE_KILL;
680
681 for_each_possible_cpu(i) {
682 struct rtable __rcu **prt;
683 prt = per_cpu_ptr(nh->nh_pcpu_rth_output, i);
684 rt = rcu_dereference(*prt);
685 if (rt)
686 rt->dst.obsolete = DST_OBSOLETE_KILL;
687 }
688 }
689
690 fnhe->fnhe_stamp = jiffies;
691
692out_unlock:
693 spin_unlock_bh(&fnhe_lock);
694}
695
696static void __ip_do_redirect(struct rtable *rt, struct sk_buff *skb, struct flowi4 *fl4,
697 bool kill_route)
698{
699 __be32 new_gw = icmp_hdr(skb)->un.gateway;
700 __be32 old_gw = ip_hdr(skb)->saddr;
701 struct net_device *dev = skb->dev;
702 struct in_device *in_dev;
703 struct fib_result res;
704 struct neighbour *n;
705 struct net *net;
706
707 switch (icmp_hdr(skb)->code & 7) {
708 case ICMP_REDIR_NET:
709 case ICMP_REDIR_NETTOS:
710 case ICMP_REDIR_HOST:
711 case ICMP_REDIR_HOSTTOS:
712 break;
713
714 default:
715 return;
716 }
717
718 if (rt->rt_gateway != old_gw)
719 return;
720
721 in_dev = __in_dev_get_rcu(dev);
722 if (!in_dev)
723 return;
724
725 net = dev_net(dev);
726 if (new_gw == old_gw || !IN_DEV_RX_REDIRECTS(in_dev) ||
727 ipv4_is_multicast(new_gw) || ipv4_is_lbcast(new_gw) ||
728 ipv4_is_zeronet(new_gw))
729 goto reject_redirect;
730
731 if (!IN_DEV_SHARED_MEDIA(in_dev)) {
732 if (!inet_addr_onlink(in_dev, new_gw, old_gw))
733 goto reject_redirect;
734 if (IN_DEV_SEC_REDIRECTS(in_dev) && ip_fib_check_default(new_gw, dev))
735 goto reject_redirect;
736 } else {
737 if (inet_addr_type(net, new_gw) != RTN_UNICAST)
738 goto reject_redirect;
739 }
740
741 n = ipv4_neigh_lookup(&rt->dst, NULL, &new_gw);
742 if (n) {
743 if (!(n->nud_state & NUD_VALID)) {
744 neigh_event_send(n, NULL);
745 } else {
746 if (fib_lookup(net, fl4, &res) == 0) {
747 struct fib_nh *nh = &FIB_RES_NH(res);
748
749 update_or_create_fnhe(nh, fl4->daddr, new_gw,
750 0, 0);
751 }
752 if (kill_route)
753 rt->dst.obsolete = DST_OBSOLETE_KILL;
754 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, n);
755 }
756 neigh_release(n);
757 }
758 return;
759
760reject_redirect:
761#ifdef CONFIG_IP_ROUTE_VERBOSE
762 if (IN_DEV_LOG_MARTIANS(in_dev)) {
763 const struct iphdr *iph = (const struct iphdr *) skb->data;
764 __be32 daddr = iph->daddr;
765 __be32 saddr = iph->saddr;
766
767 net_info_ratelimited("Redirect from %pI4 on %s about %pI4 ignored\n"
768 " Advised path = %pI4 -> %pI4\n",
769 &old_gw, dev->name, &new_gw,
770 &saddr, &daddr);
771 }
772#endif
773 ;
774}
775
776static void ip_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
777{
778 struct rtable *rt;
779 struct flowi4 fl4;
780 const struct iphdr *iph = (const struct iphdr *) skb->data;
781 int oif = skb->dev->ifindex;
782 u8 tos = RT_TOS(iph->tos);
783 u8 prot = iph->protocol;
784 u32 mark = skb->mark;
785
786 rt = (struct rtable *) dst;
787
788 __build_flow_key(&fl4, sk, iph, oif, tos, prot, mark, 0);
789 __ip_do_redirect(rt, skb, &fl4, true);
790}
791
792static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst)
793{
794 struct rtable *rt = (struct rtable *)dst;
795 struct dst_entry *ret = dst;
796
797 if (rt) {
798 if (dst->obsolete > 0) {
799 ip_rt_put(rt);
800 ret = NULL;
801 } else if ((rt->rt_flags & RTCF_REDIRECTED) ||
802 rt->dst.expires) {
803 ip_rt_put(rt);
804 ret = NULL;
805 }
806 }
807 return ret;
808}
809
810/*
811 * Algorithm:
812 * 1. The first ip_rt_redirect_number redirects are sent
813 * with exponential backoff, then we stop sending them at all,
814 * assuming that the host ignores our redirects.
815 * 2. If we did not see packets requiring redirects
816 * during ip_rt_redirect_silence, we assume that the host
817 * forgot redirected route and start to send redirects again.
818 *
819 * This algorithm is much cheaper and more intelligent than dumb load limiting
820 * in icmp.c.
821 *
822 * NOTE. Do not forget to inhibit load limiting for redirects (redundant)
823 * and "frag. need" (breaks PMTU discovery) in icmp.c.
824 */
825
826void ip_rt_send_redirect(struct sk_buff *skb)
827{
828 struct rtable *rt = skb_rtable(skb);
829 struct in_device *in_dev;
830 struct inet_peer *peer;
831 struct net *net;
832 int log_martians;
833
834 rcu_read_lock();
835 in_dev = __in_dev_get_rcu(rt->dst.dev);
836 if (!in_dev || !IN_DEV_TX_REDIRECTS(in_dev)) {
837 rcu_read_unlock();
838 return;
839 }
840 log_martians = IN_DEV_LOG_MARTIANS(in_dev);
841 rcu_read_unlock();
842
843 net = dev_net(rt->dst.dev);
844 peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, 1);
845 if (!peer) {
846 icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST,
847 rt_nexthop(rt, ip_hdr(skb)->daddr));
848 return;
849 }
850
851 /* No redirected packets during ip_rt_redirect_silence;
852 * reset the algorithm.
853 */
854 if (time_after(jiffies, peer->rate_last + ip_rt_redirect_silence))
855 peer->rate_tokens = 0;
856
857 /* Too many ignored redirects; do not send anything
858 * set dst.rate_last to the last seen redirected packet.
859 */
860 if (peer->rate_tokens >= ip_rt_redirect_number) {
861 peer->rate_last = jiffies;
862 goto out_put_peer;
863 }
864
865 /* Check for load limit; set rate_last to the latest sent
866 * redirect.
867 */
868 if (peer->rate_tokens == 0 ||
869 time_after(jiffies,
870 (peer->rate_last +
871 (ip_rt_redirect_load << peer->rate_tokens)))) {
872 __be32 gw = rt_nexthop(rt, ip_hdr(skb)->daddr);
873
874 icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, gw);
875 peer->rate_last = jiffies;
876 ++peer->rate_tokens;
877#ifdef CONFIG_IP_ROUTE_VERBOSE
878 if (log_martians &&
879 peer->rate_tokens == ip_rt_redirect_number)
880 net_warn_ratelimited("host %pI4/if%d ignores redirects for %pI4 to %pI4\n",
881 &ip_hdr(skb)->saddr, inet_iif(skb),
882 &ip_hdr(skb)->daddr, &gw);
883#endif
884 }
885out_put_peer:
886 inet_putpeer(peer);
887}
888
889static int ip_error(struct sk_buff *skb)
890{
891 struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
892 struct rtable *rt = skb_rtable(skb);
893 struct inet_peer *peer;
894 unsigned long now;
895 struct net *net;
896 bool send;
897 int code;
898
899 net = dev_net(rt->dst.dev);
900 if (!IN_DEV_FORWARD(in_dev)) {
901 switch (rt->dst.error) {
902 case EHOSTUNREACH:
903 IP_INC_STATS_BH(net, IPSTATS_MIB_INADDRERRORS);
904 break;
905
906 case ENETUNREACH:
907 IP_INC_STATS_BH(net, IPSTATS_MIB_INNOROUTES);
908 break;
909 }
910 goto out;
911 }
912
913 switch (rt->dst.error) {
914 case EINVAL:
915 default:
916 goto out;
917 case EHOSTUNREACH:
918 code = ICMP_HOST_UNREACH;
919 break;
920 case ENETUNREACH:
921 code = ICMP_NET_UNREACH;
922 IP_INC_STATS_BH(net, IPSTATS_MIB_INNOROUTES);
923 break;
924 case EACCES:
925 code = ICMP_PKT_FILTERED;
926 break;
927 }
928
929 peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, 1);
930
931 send = true;
932 if (peer) {
933 now = jiffies;
934 peer->rate_tokens += now - peer->rate_last;
935 if (peer->rate_tokens > ip_rt_error_burst)
936 peer->rate_tokens = ip_rt_error_burst;
937 peer->rate_last = now;
938 if (peer->rate_tokens >= ip_rt_error_cost)
939 peer->rate_tokens -= ip_rt_error_cost;
940 else
941 send = false;
942 inet_putpeer(peer);
943 }
944 if (send)
945 icmp_send(skb, ICMP_DEST_UNREACH, code, 0);
946
947out: kfree_skb(skb);
948 return 0;
949}
950
951static void __ip_rt_update_pmtu(struct rtable *rt, struct flowi4 *fl4, u32 mtu)
952{
953 struct dst_entry *dst = &rt->dst;
954 struct fib_result res;
955
956 if (dst_metric_locked(dst, RTAX_MTU))
957 return;
958
959 if (dst->dev->mtu < mtu)
960 return;
961
962 if (mtu < ip_rt_min_pmtu)
963 mtu = ip_rt_min_pmtu;
964
965 if (rt->rt_pmtu == mtu &&
966 time_before(jiffies, dst->expires - ip_rt_mtu_expires / 2))
967 return;
968
969 rcu_read_lock();
970 if (fib_lookup(dev_net(dst->dev), fl4, &res) == 0) {
971 struct fib_nh *nh = &FIB_RES_NH(res);
972
973 update_or_create_fnhe(nh, fl4->daddr, 0, mtu,
974 jiffies + ip_rt_mtu_expires);
975 }
976 rcu_read_unlock();
977}
978
979static void ip_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
980 struct sk_buff *skb, u32 mtu)
981{
982 struct rtable *rt = (struct rtable *) dst;
983 struct flowi4 fl4;
984
985 ip_rt_build_flow_key(&fl4, sk, skb);
986 __ip_rt_update_pmtu(rt, &fl4, mtu);
987}
988
989void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu,
990 int oif, u32 mark, u8 protocol, int flow_flags)
991{
992 const struct iphdr *iph = (const struct iphdr *) skb->data;
993 struct flowi4 fl4;
994 struct rtable *rt;
995
996 if (!mark)
997 mark = IP4_REPLY_MARK(net, skb->mark);
998
999 __build_flow_key(&fl4, NULL, iph, oif,
1000 RT_TOS(iph->tos), protocol, mark, flow_flags);
1001 rt = __ip_route_output_key(net, &fl4);
1002 if (!IS_ERR(rt)) {
1003 __ip_rt_update_pmtu(rt, &fl4, mtu);
1004 ip_rt_put(rt);
1005 }
1006}
1007EXPORT_SYMBOL_GPL(ipv4_update_pmtu);
1008
1009static void __ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu)
1010{
1011 const struct iphdr *iph = (const struct iphdr *) skb->data;
1012 struct flowi4 fl4;
1013 struct rtable *rt;
1014
1015 __build_flow_key(&fl4, sk, iph, 0, 0, 0, 0, 0);
1016
1017 if (!fl4.flowi4_mark)
1018 fl4.flowi4_mark = IP4_REPLY_MARK(sock_net(sk), skb->mark);
1019
1020 rt = __ip_route_output_key(sock_net(sk), &fl4);
1021 if (!IS_ERR(rt)) {
1022 __ip_rt_update_pmtu(rt, &fl4, mtu);
1023 ip_rt_put(rt);
1024 }
1025}
1026
1027void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu)
1028{
1029 const struct iphdr *iph = (const struct iphdr *) skb->data;
1030 struct flowi4 fl4;
1031 struct rtable *rt;
1032 struct dst_entry *dst;
1033 bool new = false;
1034
1035 bh_lock_sock(sk);
1036
1037 if (!ip_sk_accept_pmtu(sk))
1038 goto out;
1039
1040 rt = (struct rtable *) __sk_dst_get(sk);
1041
1042 if (sock_owned_by_user(sk) || !rt) {
1043 __ipv4_sk_update_pmtu(skb, sk, mtu);
1044 goto out;
1045 }
1046
1047 __build_flow_key(&fl4, sk, iph, 0, 0, 0, 0, 0);
1048
1049 if (!__sk_dst_check(sk, 0)) {
1050 rt = ip_route_output_flow(sock_net(sk), &fl4, sk);
1051 if (IS_ERR(rt))
1052 goto out;
1053
1054 new = true;
1055 }
1056
1057 __ip_rt_update_pmtu((struct rtable *) rt->dst.path, &fl4, mtu);
1058
1059 dst = dst_check(&rt->dst, 0);
1060 if (!dst) {
1061 if (new)
1062 dst_release(&rt->dst);
1063
1064 rt = ip_route_output_flow(sock_net(sk), &fl4, sk);
1065 if (IS_ERR(rt))
1066 goto out;
1067
1068 new = true;
1069 }
1070
1071 if (new)
1072 __sk_dst_set(sk, &rt->dst);
1073
1074out:
1075 bh_unlock_sock(sk);
1076}
1077EXPORT_SYMBOL_GPL(ipv4_sk_update_pmtu);
1078
1079void ipv4_redirect(struct sk_buff *skb, struct net *net,
1080 int oif, u32 mark, u8 protocol, int flow_flags)
1081{
1082 const struct iphdr *iph = (const struct iphdr *) skb->data;
1083 struct flowi4 fl4;
1084 struct rtable *rt;
1085
1086 __build_flow_key(&fl4, NULL, iph, oif,
1087 RT_TOS(iph->tos), protocol, mark, flow_flags);
1088 rt = __ip_route_output_key(net, &fl4);
1089 if (!IS_ERR(rt)) {
1090 __ip_do_redirect(rt, skb, &fl4, false);
1091 ip_rt_put(rt);
1092 }
1093}
1094EXPORT_SYMBOL_GPL(ipv4_redirect);
1095
1096void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk)
1097{
1098 const struct iphdr *iph = (const struct iphdr *) skb->data;
1099 struct flowi4 fl4;
1100 struct rtable *rt;
1101
1102 __build_flow_key(&fl4, sk, iph, 0, 0, 0, 0, 0);
1103 rt = __ip_route_output_key(sock_net(sk), &fl4);
1104 if (!IS_ERR(rt)) {
1105 __ip_do_redirect(rt, skb, &fl4, false);
1106 ip_rt_put(rt);
1107 }
1108}
1109EXPORT_SYMBOL_GPL(ipv4_sk_redirect);
1110
1111static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie)
1112{
1113 struct rtable *rt = (struct rtable *) dst;
1114
1115 /* All IPV4 dsts are created with ->obsolete set to the value
1116 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1117 * into this function always.
1118 *
1119 * When a PMTU/redirect information update invalidates a route,
1120 * this is indicated by setting obsolete to DST_OBSOLETE_KILL or
1121 * DST_OBSOLETE_DEAD by dst_free().
1122 */
1123 if (dst->obsolete != DST_OBSOLETE_FORCE_CHK || rt_is_expired(rt))
1124 return NULL;
1125 return dst;
1126}
1127
1128static void ipv4_link_failure(struct sk_buff *skb)
1129{
1130 struct rtable *rt;
1131
1132 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0);
1133
1134 rt = skb_rtable(skb);
1135 if (rt)
1136 dst_set_expires(&rt->dst, 0);
1137}
1138
1139static int ip_rt_bug(struct sock *sk, struct sk_buff *skb)
1140{
1141 pr_debug("%s: %pI4 -> %pI4, %s\n",
1142 __func__, &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr,
1143 skb->dev ? skb->dev->name : "?");
1144 kfree_skb(skb);
1145 WARN_ON(1);
1146 return 0;
1147}
1148
1149/*
1150 We do not cache source address of outgoing interface,
1151 because it is used only by IP RR, TS and SRR options,
1152 so that it out of fast path.
1153
1154 BTW remember: "addr" is allowed to be not aligned
1155 in IP options!
1156 */
1157
1158void ip_rt_get_source(u8 *addr, struct sk_buff *skb, struct rtable *rt)
1159{
1160 __be32 src;
1161
1162 if (rt_is_output_route(rt))
1163 src = ip_hdr(skb)->saddr;
1164 else {
1165 struct fib_result res;
1166 struct flowi4 fl4;
1167 struct iphdr *iph;
1168
1169 iph = ip_hdr(skb);
1170
1171 memset(&fl4, 0, sizeof(fl4));
1172 fl4.daddr = iph->daddr;
1173 fl4.saddr = iph->saddr;
1174 fl4.flowi4_tos = RT_TOS(iph->tos);
1175 fl4.flowi4_oif = rt->dst.dev->ifindex;
1176 fl4.flowi4_iif = skb->dev->ifindex;
1177 fl4.flowi4_mark = skb->mark;
1178
1179 rcu_read_lock();
1180 if (fib_lookup(dev_net(rt->dst.dev), &fl4, &res) == 0)
1181 src = FIB_RES_PREFSRC(dev_net(rt->dst.dev), res);
1182 else
1183 src = inet_select_addr(rt->dst.dev,
1184 rt_nexthop(rt, iph->daddr),
1185 RT_SCOPE_UNIVERSE);
1186 rcu_read_unlock();
1187 }
1188 memcpy(addr, &src, 4);
1189}
1190
1191#ifdef CONFIG_IP_ROUTE_CLASSID
1192static void set_class_tag(struct rtable *rt, u32 tag)
1193{
1194 if (!(rt->dst.tclassid & 0xFFFF))
1195 rt->dst.tclassid |= tag & 0xFFFF;
1196 if (!(rt->dst.tclassid & 0xFFFF0000))
1197 rt->dst.tclassid |= tag & 0xFFFF0000;
1198}
1199#endif
1200
1201static unsigned int ipv4_default_advmss(const struct dst_entry *dst)
1202{
1203 unsigned int advmss = dst_metric_raw(dst, RTAX_ADVMSS);
1204
1205 if (advmss == 0) {
1206 advmss = max_t(unsigned int, dst->dev->mtu - 40,
1207 ip_rt_min_advmss);
1208 if (advmss > 65535 - 40)
1209 advmss = 65535 - 40;
1210 }
1211 return advmss;
1212}
1213
1214static unsigned int ipv4_mtu(const struct dst_entry *dst)
1215{
1216 const struct rtable *rt = (const struct rtable *) dst;
1217 unsigned int mtu = rt->rt_pmtu;
1218
1219 if (!mtu || time_after_eq(jiffies, rt->dst.expires))
1220 mtu = dst_metric_raw(dst, RTAX_MTU);
1221
1222 if (mtu)
1223 return mtu;
1224
1225 mtu = dst->dev->mtu;
1226
1227 if (unlikely(dst_metric_locked(dst, RTAX_MTU))) {
1228 if (rt->rt_uses_gateway && mtu > 576)
1229 mtu = 576;
1230 }
1231
1232 return min_t(unsigned int, mtu, IP_MAX_MTU);
1233}
1234
1235static struct fib_nh_exception *find_exception(struct fib_nh *nh, __be32 daddr)
1236{
1237 struct fnhe_hash_bucket *hash = nh->nh_exceptions;
1238 struct fib_nh_exception *fnhe;
1239 u32 hval;
1240
1241 if (!hash)
1242 return NULL;
1243
1244 hval = fnhe_hashfun(daddr);
1245
1246 for (fnhe = rcu_dereference(hash[hval].chain); fnhe;
1247 fnhe = rcu_dereference(fnhe->fnhe_next)) {
1248 if (fnhe->fnhe_daddr == daddr)
1249 return fnhe;
1250 }
1251 return NULL;
1252}
1253
1254static bool rt_bind_exception(struct rtable *rt, struct fib_nh_exception *fnhe,
1255 __be32 daddr)
1256{
1257 bool ret = false;
1258
1259 spin_lock_bh(&fnhe_lock);
1260
1261 if (daddr == fnhe->fnhe_daddr) {
1262 struct rtable __rcu **porig;
1263 struct rtable *orig;
1264 int genid = fnhe_genid(dev_net(rt->dst.dev));
1265
1266 if (rt_is_input_route(rt))
1267 porig = &fnhe->fnhe_rth_input;
1268 else
1269 porig = &fnhe->fnhe_rth_output;
1270 orig = rcu_dereference(*porig);
1271
1272 if (fnhe->fnhe_genid != genid) {
1273 fnhe->fnhe_genid = genid;
1274 fnhe->fnhe_gw = 0;
1275 fnhe->fnhe_pmtu = 0;
1276 fnhe->fnhe_expires = 0;
1277 fnhe_flush_routes(fnhe);
1278 orig = NULL;
1279 }
1280 fill_route_from_fnhe(rt, fnhe);
1281 if (!rt->rt_gateway)
1282 rt->rt_gateway = daddr;
1283
1284 if (!(rt->dst.flags & DST_NOCACHE)) {
1285 rcu_assign_pointer(*porig, rt);
1286 if (orig)
1287 rt_free(orig);
1288 ret = true;
1289 }
1290
1291 fnhe->fnhe_stamp = jiffies;
1292 }
1293 spin_unlock_bh(&fnhe_lock);
1294
1295 return ret;
1296}
1297
1298static bool rt_cache_route(struct fib_nh *nh, struct rtable *rt)
1299{
1300 struct rtable *orig, *prev, **p;
1301 bool ret = true;
1302
1303 if (rt_is_input_route(rt)) {
1304 p = (struct rtable **)&nh->nh_rth_input;
1305 } else {
1306 p = (struct rtable **)__this_cpu_ptr(nh->nh_pcpu_rth_output);
1307 }
1308 orig = *p;
1309
1310 prev = cmpxchg(p, orig, rt);
1311 if (prev == orig) {
1312 if (orig)
1313 rt_free(orig);
1314 } else
1315 ret = false;
1316
1317 return ret;
1318}
1319
1320static DEFINE_SPINLOCK(rt_uncached_lock);
1321static LIST_HEAD(rt_uncached_list);
1322
1323static void rt_add_uncached_list(struct rtable *rt)
1324{
1325 spin_lock_bh(&rt_uncached_lock);
1326 list_add_tail(&rt->rt_uncached, &rt_uncached_list);
1327 spin_unlock_bh(&rt_uncached_lock);
1328}
1329
1330static void ipv4_dst_destroy(struct dst_entry *dst)
1331{
1332 struct rtable *rt = (struct rtable *) dst;
1333
1334 if (!list_empty(&rt->rt_uncached)) {
1335 spin_lock_bh(&rt_uncached_lock);
1336 list_del(&rt->rt_uncached);
1337 spin_unlock_bh(&rt_uncached_lock);
1338 }
1339}
1340
1341void rt_flush_dev(struct net_device *dev)
1342{
1343 if (!list_empty(&rt_uncached_list)) {
1344 struct net *net = dev_net(dev);
1345 struct rtable *rt;
1346
1347 spin_lock_bh(&rt_uncached_lock);
1348 list_for_each_entry(rt, &rt_uncached_list, rt_uncached) {
1349 if (rt->dst.dev != dev)
1350 continue;
1351 rt->dst.dev = net->loopback_dev;
1352 dev_hold(rt->dst.dev);
1353 dev_put(dev);
1354 }
1355 spin_unlock_bh(&rt_uncached_lock);
1356 }
1357}
1358
1359static bool rt_cache_valid(const struct rtable *rt)
1360{
1361 return rt &&
1362 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1363 !rt_is_expired(rt);
1364}
1365
1366static void rt_set_nexthop(struct rtable *rt, __be32 daddr,
1367 const struct fib_result *res,
1368 struct fib_nh_exception *fnhe,
1369 struct fib_info *fi, u16 type, u32 itag)
1370{
1371 bool cached = false;
1372
1373 if (fi) {
1374 struct fib_nh *nh = &FIB_RES_NH(*res);
1375
1376 if (nh->nh_gw && nh->nh_scope == RT_SCOPE_LINK) {
1377 rt->rt_gateway = nh->nh_gw;
1378 rt->rt_uses_gateway = 1;
1379 }
1380 dst_init_metrics(&rt->dst, fi->fib_metrics, true);
1381#ifdef CONFIG_IP_ROUTE_CLASSID
1382 rt->dst.tclassid = nh->nh_tclassid;
1383#endif
1384 if (unlikely(fnhe))
1385 cached = rt_bind_exception(rt, fnhe, daddr);
1386 else if (!(rt->dst.flags & DST_NOCACHE))
1387 cached = rt_cache_route(nh, rt);
1388 if (unlikely(!cached)) {
1389 /* Routes we intend to cache in nexthop exception or
1390 * FIB nexthop have the DST_NOCACHE bit clear.
1391 * However, if we are unsuccessful at storing this
1392 * route into the cache we really need to set it.
1393 */
1394 rt->dst.flags |= DST_NOCACHE;
1395 if (!rt->rt_gateway)
1396 rt->rt_gateway = daddr;
1397 rt_add_uncached_list(rt);
1398 }
1399 } else
1400 rt_add_uncached_list(rt);
1401
1402#ifdef CONFIG_IP_ROUTE_CLASSID
1403#ifdef CONFIG_IP_MULTIPLE_TABLES
1404 set_class_tag(rt, res->tclassid);
1405#endif
1406 set_class_tag(rt, itag);
1407#endif
1408}
1409
1410static struct rtable *rt_dst_alloc(struct net_device *dev,
1411 bool nopolicy, bool noxfrm, bool will_cache)
1412{
1413 return dst_alloc(&ipv4_dst_ops, dev, 1, DST_OBSOLETE_FORCE_CHK,
1414 (will_cache ? 0 : (DST_HOST | DST_NOCACHE)) |
1415 (nopolicy ? DST_NOPOLICY : 0) |
1416 (noxfrm ? DST_NOXFRM : 0));
1417}
1418
1419/* called in rcu_read_lock() section */
1420static int ip_route_input_mc(struct sk_buff *skb, __be32 daddr, __be32 saddr,
1421 u8 tos, struct net_device *dev, int our)
1422{
1423 struct rtable *rth;
1424 struct in_device *in_dev = __in_dev_get_rcu(dev);
1425 u32 itag = 0;
1426 int err;
1427
1428 /* Primary sanity checks. */
1429
1430 if (in_dev == NULL)
1431 return -EINVAL;
1432
1433 if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) ||
1434 skb->protocol != htons(ETH_P_IP))
1435 goto e_inval;
1436
1437 if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev)))
1438 if (ipv4_is_loopback(saddr))
1439 goto e_inval;
1440
1441 if (ipv4_is_zeronet(saddr)) {
1442 if (!ipv4_is_local_multicast(daddr))
1443 goto e_inval;
1444 } else {
1445 err = fib_validate_source(skb, saddr, 0, tos, 0, dev,
1446 in_dev, &itag);
1447 if (err < 0)
1448 goto e_err;
1449 }
1450 rth = rt_dst_alloc(dev_net(dev)->loopback_dev,
1451 IN_DEV_CONF_GET(in_dev, NOPOLICY), false, false);
1452 if (!rth)
1453 goto e_nobufs;
1454
1455#ifdef CONFIG_IP_ROUTE_CLASSID
1456 rth->dst.tclassid = itag;
1457#endif
1458 rth->dst.output = ip_rt_bug;
1459
1460 rth->rt_genid = rt_genid_ipv4(dev_net(dev));
1461 rth->rt_flags = RTCF_MULTICAST;
1462 rth->rt_type = RTN_MULTICAST;
1463 rth->rt_is_input= 1;
1464 rth->rt_iif = 0;
1465 rth->rt_pmtu = 0;
1466 rth->rt_gateway = 0;
1467 rth->rt_uses_gateway = 0;
1468 INIT_LIST_HEAD(&rth->rt_uncached);
1469 if (our) {
1470 rth->dst.input= ip_local_deliver;
1471 rth->rt_flags |= RTCF_LOCAL;
1472 }
1473
1474#ifdef CONFIG_IP_MROUTE
1475 if (!ipv4_is_local_multicast(daddr) && IN_DEV_MFORWARD(in_dev))
1476 rth->dst.input = ip_mr_input;
1477#endif
1478 RT_CACHE_STAT_INC(in_slow_mc);
1479
1480 skb_dst_set(skb, &rth->dst);
1481 return 0;
1482
1483e_nobufs:
1484 return -ENOBUFS;
1485e_inval:
1486 return -EINVAL;
1487e_err:
1488 return err;
1489}
1490
1491
1492static void ip_handle_martian_source(struct net_device *dev,
1493 struct in_device *in_dev,
1494 struct sk_buff *skb,
1495 __be32 daddr,
1496 __be32 saddr)
1497{
1498 RT_CACHE_STAT_INC(in_martian_src);
1499#ifdef CONFIG_IP_ROUTE_VERBOSE
1500 if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) {
1501 /*
1502 * RFC1812 recommendation, if source is martian,
1503 * the only hint is MAC header.
1504 */
1505 pr_warn("martian source %pI4 from %pI4, on dev %s\n",
1506 &daddr, &saddr, dev->name);
1507 if (dev->hard_header_len && skb_mac_header_was_set(skb)) {
1508 print_hex_dump(KERN_WARNING, "ll header: ",
1509 DUMP_PREFIX_OFFSET, 16, 1,
1510 skb_mac_header(skb),
1511 dev->hard_header_len, true);
1512 }
1513 }
1514#endif
1515}
1516
1517/* called in rcu_read_lock() section */
1518static int __mkroute_input(struct sk_buff *skb,
1519 const struct fib_result *res,
1520 struct in_device *in_dev,
1521 __be32 daddr, __be32 saddr, u32 tos)
1522{
1523 struct fib_nh_exception *fnhe;
1524 struct rtable *rth;
1525 int err;
1526 struct in_device *out_dev;
1527 unsigned int flags = 0;
1528 bool do_cache;
1529 u32 itag = 0;
1530
1531 /* get a working reference to the output device */
1532 out_dev = __in_dev_get_rcu(FIB_RES_DEV(*res));
1533 if (out_dev == NULL) {
1534 net_crit_ratelimited("Bug in ip_route_input_slow(). Please report.\n");
1535 return -EINVAL;
1536 }
1537
1538 err = fib_validate_source(skb, saddr, daddr, tos, FIB_RES_OIF(*res),
1539 in_dev->dev, in_dev, &itag);
1540 if (err < 0) {
1541 ip_handle_martian_source(in_dev->dev, in_dev, skb, daddr,
1542 saddr);
1543
1544 goto cleanup;
1545 }
1546
1547 do_cache = res->fi && !itag;
1548 if (out_dev == in_dev && err && IN_DEV_TX_REDIRECTS(out_dev) &&
1549 (IN_DEV_SHARED_MEDIA(out_dev) ||
1550 inet_addr_onlink(out_dev, saddr, FIB_RES_GW(*res)))) {
1551 flags |= RTCF_DOREDIRECT;
1552 do_cache = false;
1553 }
1554
1555 if (skb->protocol != htons(ETH_P_IP)) {
1556 /* Not IP (i.e. ARP). Do not create route, if it is
1557 * invalid for proxy arp. DNAT routes are always valid.
1558 *
1559 * Proxy arp feature have been extended to allow, ARP
1560 * replies back to the same interface, to support
1561 * Private VLAN switch technologies. See arp.c.
1562 */
1563 if (out_dev == in_dev &&
1564 IN_DEV_PROXY_ARP_PVLAN(in_dev) == 0) {
1565 err = -EINVAL;
1566 goto cleanup;
1567 }
1568 }
1569
1570 fnhe = find_exception(&FIB_RES_NH(*res), daddr);
1571 if (do_cache) {
1572 if (fnhe != NULL)
1573 rth = rcu_dereference(fnhe->fnhe_rth_input);
1574 else
1575 rth = rcu_dereference(FIB_RES_NH(*res).nh_rth_input);
1576
1577 if (rt_cache_valid(rth)) {
1578 skb_dst_set_noref(skb, &rth->dst);
1579 goto out;
1580 }
1581 }
1582
1583 rth = rt_dst_alloc(out_dev->dev,
1584 IN_DEV_CONF_GET(in_dev, NOPOLICY),
1585 IN_DEV_CONF_GET(out_dev, NOXFRM), do_cache);
1586 if (!rth) {
1587 err = -ENOBUFS;
1588 goto cleanup;
1589 }
1590
1591 rth->rt_genid = rt_genid_ipv4(dev_net(rth->dst.dev));
1592 rth->rt_flags = flags;
1593 rth->rt_type = res->type;
1594 rth->rt_is_input = 1;
1595 rth->rt_iif = 0;
1596 rth->rt_pmtu = 0;
1597 rth->rt_gateway = 0;
1598 rth->rt_uses_gateway = 0;
1599 INIT_LIST_HEAD(&rth->rt_uncached);
1600 RT_CACHE_STAT_INC(in_slow_tot);
1601
1602 rth->dst.input = ip_forward;
1603 rth->dst.output = ip_output;
1604
1605 rt_set_nexthop(rth, daddr, res, fnhe, res->fi, res->type, itag);
1606 skb_dst_set(skb, &rth->dst);
1607out:
1608 err = 0;
1609 cleanup:
1610 return err;
1611}
1612
1613static int ip_mkroute_input(struct sk_buff *skb,
1614 struct fib_result *res,
1615 const struct flowi4 *fl4,
1616 struct in_device *in_dev,
1617 __be32 daddr, __be32 saddr, u32 tos)
1618{
1619#ifdef CONFIG_IP_ROUTE_MULTIPATH
1620 if (res->fi && res->fi->fib_nhs > 1)
1621 fib_select_multipath(res);
1622#endif
1623
1624 /* create a routing cache entry */
1625 return __mkroute_input(skb, res, in_dev, daddr, saddr, tos);
1626}
1627
1628/*
1629 * NOTE. We drop all the packets that has local source
1630 * addresses, because every properly looped back packet
1631 * must have correct destination already attached by output routine.
1632 *
1633 * Such approach solves two big problems:
1634 * 1. Not simplex devices are handled properly.
1635 * 2. IP spoofing attempts are filtered with 100% of guarantee.
1636 * called with rcu_read_lock()
1637 */
1638
1639static int ip_route_input_slow(struct sk_buff *skb, __be32 daddr, __be32 saddr,
1640 u8 tos, struct net_device *dev)
1641{
1642 struct fib_result res;
1643 struct in_device *in_dev = __in_dev_get_rcu(dev);
1644 struct flowi4 fl4;
1645 unsigned int flags = 0;
1646 u32 itag = 0;
1647 struct rtable *rth;
1648 int err = -EINVAL;
1649 struct net *net = dev_net(dev);
1650 bool do_cache;
1651
1652 /* IP on this device is disabled. */
1653
1654 if (!in_dev)
1655 goto out;
1656
1657 /* Check for the most weird martians, which can be not detected
1658 by fib_lookup.
1659 */
1660
1661 if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr))
1662 goto martian_source;
1663
1664 res.fi = NULL;
1665 if (ipv4_is_lbcast(daddr) || (saddr == 0 && daddr == 0))
1666 goto brd_input;
1667
1668 /* Accept zero addresses only to limited broadcast;
1669 * I even do not know to fix it or not. Waiting for complains :-)
1670 */
1671 if (ipv4_is_zeronet(saddr))
1672 goto martian_source;
1673
1674 if (ipv4_is_zeronet(daddr))
1675 goto martian_destination;
1676
1677 /* Following code try to avoid calling IN_DEV_NET_ROUTE_LOCALNET(),
1678 * and call it once if daddr or/and saddr are loopback addresses
1679 */
1680 if (ipv4_is_loopback(daddr)) {
1681 if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev, net))
1682 goto martian_destination;
1683 } else if (ipv4_is_loopback(saddr)) {
1684 if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev, net))
1685 goto martian_source;
1686 }
1687
1688 /*
1689 * Now we are ready to route packet.
1690 */
1691 fl4.flowi4_oif = 0;
1692 fl4.flowi4_iif = dev->ifindex;
1693 fl4.flowi4_mark = skb->mark;
1694 fl4.flowi4_tos = tos;
1695 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
1696 fl4.daddr = daddr;
1697 fl4.saddr = saddr;
1698 err = fib_lookup(net, &fl4, &res);
1699 if (err != 0) {
1700 if (!IN_DEV_FORWARD(in_dev))
1701 err = -EHOSTUNREACH;
1702 goto no_route;
1703 }
1704
1705 if (res.type == RTN_BROADCAST)
1706 goto brd_input;
1707
1708 if (res.type == RTN_LOCAL) {
1709 err = fib_validate_source(skb, saddr, daddr, tos,
1710 0, dev, in_dev, &itag);
1711 if (err < 0)
1712 goto martian_source_keep_err;
1713 goto local_input;
1714 }
1715
1716 if (!IN_DEV_FORWARD(in_dev)) {
1717 err = -EHOSTUNREACH;
1718 goto no_route;
1719 }
1720 if (res.type != RTN_UNICAST)
1721 goto martian_destination;
1722
1723 err = ip_mkroute_input(skb, &res, &fl4, in_dev, daddr, saddr, tos);
1724out: return err;
1725
1726brd_input:
1727 if (skb->protocol != htons(ETH_P_IP))
1728 goto e_inval;
1729
1730 if (!ipv4_is_zeronet(saddr)) {
1731 err = fib_validate_source(skb, saddr, 0, tos, 0, dev,
1732 in_dev, &itag);
1733 if (err < 0)
1734 goto martian_source_keep_err;
1735 }
1736 flags |= RTCF_BROADCAST;
1737 res.type = RTN_BROADCAST;
1738 RT_CACHE_STAT_INC(in_brd);
1739
1740local_input:
1741 do_cache = false;
1742 if (res.fi) {
1743 if (!itag) {
1744 rth = rcu_dereference(FIB_RES_NH(res).nh_rth_input);
1745 if (rt_cache_valid(rth)) {
1746 skb_dst_set_noref(skb, &rth->dst);
1747 err = 0;
1748 goto out;
1749 }
1750 do_cache = true;
1751 }
1752 }
1753
1754 rth = rt_dst_alloc(net->loopback_dev,
1755 IN_DEV_CONF_GET(in_dev, NOPOLICY), false, do_cache);
1756 if (!rth)
1757 goto e_nobufs;
1758
1759 rth->dst.input= ip_local_deliver;
1760 rth->dst.output= ip_rt_bug;
1761#ifdef CONFIG_IP_ROUTE_CLASSID
1762 rth->dst.tclassid = itag;
1763#endif
1764
1765 rth->rt_genid = rt_genid_ipv4(net);
1766 rth->rt_flags = flags|RTCF_LOCAL;
1767 rth->rt_type = res.type;
1768 rth->rt_is_input = 1;
1769 rth->rt_iif = 0;
1770 rth->rt_pmtu = 0;
1771 rth->rt_gateway = 0;
1772 rth->rt_uses_gateway = 0;
1773 INIT_LIST_HEAD(&rth->rt_uncached);
1774 RT_CACHE_STAT_INC(in_slow_tot);
1775 if (res.type == RTN_UNREACHABLE) {
1776 rth->dst.input= ip_error;
1777 rth->dst.error= -err;
1778 rth->rt_flags &= ~RTCF_LOCAL;
1779 }
1780 if (do_cache) {
1781 if (unlikely(!rt_cache_route(&FIB_RES_NH(res), rth))) {
1782 rth->dst.flags |= DST_NOCACHE;
1783 rt_add_uncached_list(rth);
1784 }
1785 }
1786 skb_dst_set(skb, &rth->dst);
1787 err = 0;
1788 goto out;
1789
1790no_route:
1791 RT_CACHE_STAT_INC(in_no_route);
1792 res.type = RTN_UNREACHABLE;
1793 if (err == -ESRCH)
1794 err = -ENETUNREACH;
1795 goto local_input;
1796
1797 /*
1798 * Do not cache martian addresses: they should be logged (RFC1812)
1799 */
1800martian_destination:
1801 RT_CACHE_STAT_INC(in_martian_dst);
1802#ifdef CONFIG_IP_ROUTE_VERBOSE
1803 if (IN_DEV_LOG_MARTIANS(in_dev))
1804 net_warn_ratelimited("martian destination %pI4 from %pI4, dev %s\n",
1805 &daddr, &saddr, dev->name);
1806#endif
1807
1808e_inval:
1809 err = -EINVAL;
1810 goto out;
1811
1812e_nobufs:
1813 err = -ENOBUFS;
1814 goto out;
1815
1816martian_source:
1817 err = -EINVAL;
1818martian_source_keep_err:
1819 ip_handle_martian_source(dev, in_dev, skb, daddr, saddr);
1820 goto out;
1821}
1822
1823int ip_route_input_noref(struct sk_buff *skb, __be32 daddr, __be32 saddr,
1824 u8 tos, struct net_device *dev)
1825{
1826 int res;
1827
1828 rcu_read_lock();
1829
1830 /* Multicast recognition logic is moved from route cache to here.
1831 The problem was that too many Ethernet cards have broken/missing
1832 hardware multicast filters :-( As result the host on multicasting
1833 network acquires a lot of useless route cache entries, sort of
1834 SDR messages from all the world. Now we try to get rid of them.
1835 Really, provided software IP multicast filter is organized
1836 reasonably (at least, hashed), it does not result in a slowdown
1837 comparing with route cache reject entries.
1838 Note, that multicast routers are not affected, because
1839 route cache entry is created eventually.
1840 */
1841 if (ipv4_is_multicast(daddr)) {
1842 struct in_device *in_dev = __in_dev_get_rcu(dev);
1843
1844 if (in_dev) {
1845 int our = ip_check_mc_rcu(in_dev, daddr, saddr,
1846 ip_hdr(skb)->protocol);
1847 if (our
1848#ifdef CONFIG_IP_MROUTE
1849 ||
1850 (!ipv4_is_local_multicast(daddr) &&
1851 IN_DEV_MFORWARD(in_dev))
1852#endif
1853 ) {
1854 int res = ip_route_input_mc(skb, daddr, saddr,
1855 tos, dev, our);
1856 rcu_read_unlock();
1857 return res;
1858 }
1859 }
1860 rcu_read_unlock();
1861 return -EINVAL;
1862 }
1863 res = ip_route_input_slow(skb, daddr, saddr, tos, dev);
1864 rcu_read_unlock();
1865 return res;
1866}
1867EXPORT_SYMBOL(ip_route_input_noref);
1868
1869/* called with rcu_read_lock() */
1870static struct rtable *__mkroute_output(const struct fib_result *res,
1871 const struct flowi4 *fl4, int orig_oif,
1872 struct net_device *dev_out,
1873 unsigned int flags)
1874{
1875 struct fib_info *fi = res->fi;
1876 struct fib_nh_exception *fnhe;
1877 struct in_device *in_dev;
1878 u16 type = res->type;
1879 struct rtable *rth;
1880 bool do_cache;
1881
1882 in_dev = __in_dev_get_rcu(dev_out);
1883 if (!in_dev)
1884 return ERR_PTR(-EINVAL);
1885
1886 if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev)))
1887 if (ipv4_is_loopback(fl4->saddr) && !(dev_out->flags & IFF_LOOPBACK))
1888 return ERR_PTR(-EINVAL);
1889
1890 if (ipv4_is_lbcast(fl4->daddr))
1891 type = RTN_BROADCAST;
1892 else if (ipv4_is_multicast(fl4->daddr))
1893 type = RTN_MULTICAST;
1894 else if (ipv4_is_zeronet(fl4->daddr))
1895 return ERR_PTR(-EINVAL);
1896
1897 if (dev_out->flags & IFF_LOOPBACK)
1898 flags |= RTCF_LOCAL;
1899
1900 do_cache = true;
1901 if (type == RTN_BROADCAST) {
1902 flags |= RTCF_BROADCAST | RTCF_LOCAL;
1903 fi = NULL;
1904 } else if (type == RTN_MULTICAST) {
1905 flags |= RTCF_MULTICAST | RTCF_LOCAL;
1906 if (!ip_check_mc_rcu(in_dev, fl4->daddr, fl4->saddr,
1907 fl4->flowi4_proto))
1908 flags &= ~RTCF_LOCAL;
1909 else
1910 do_cache = false;
1911 /* If multicast route do not exist use
1912 * default one, but do not gateway in this case.
1913 * Yes, it is hack.
1914 */
1915 if (fi && res->prefixlen < 4)
1916 fi = NULL;
1917 }
1918
1919 fnhe = NULL;
1920 do_cache &= fi != NULL;
1921 if (do_cache) {
1922 struct rtable __rcu **prth;
1923 struct fib_nh *nh = &FIB_RES_NH(*res);
1924
1925 fnhe = find_exception(nh, fl4->daddr);
1926 if (fnhe)
1927 prth = &fnhe->fnhe_rth_output;
1928 else {
1929 if (unlikely(fl4->flowi4_flags &
1930 FLOWI_FLAG_KNOWN_NH &&
1931 !(nh->nh_gw &&
1932 nh->nh_scope == RT_SCOPE_LINK))) {
1933 do_cache = false;
1934 goto add;
1935 }
1936 prth = __this_cpu_ptr(nh->nh_pcpu_rth_output);
1937 }
1938 rth = rcu_dereference(*prth);
1939 if (rt_cache_valid(rth)) {
1940 dst_hold(&rth->dst);
1941 return rth;
1942 }
1943 }
1944
1945add:
1946 rth = rt_dst_alloc(dev_out,
1947 IN_DEV_CONF_GET(in_dev, NOPOLICY),
1948 IN_DEV_CONF_GET(in_dev, NOXFRM),
1949 do_cache);
1950 if (!rth)
1951 return ERR_PTR(-ENOBUFS);
1952
1953 rth->dst.output = ip_output;
1954
1955 rth->rt_genid = rt_genid_ipv4(dev_net(dev_out));
1956 rth->rt_flags = flags;
1957 rth->rt_type = type;
1958 rth->rt_is_input = 0;
1959 rth->rt_iif = orig_oif ? : 0;
1960 rth->rt_pmtu = 0;
1961 rth->rt_gateway = 0;
1962 rth->rt_uses_gateway = 0;
1963 INIT_LIST_HEAD(&rth->rt_uncached);
1964
1965 RT_CACHE_STAT_INC(out_slow_tot);
1966
1967 if (flags & RTCF_LOCAL)
1968 rth->dst.input = ip_local_deliver;
1969 if (flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
1970 if (flags & RTCF_LOCAL &&
1971 !(dev_out->flags & IFF_LOOPBACK)) {
1972 rth->dst.output = ip_mc_output;
1973 RT_CACHE_STAT_INC(out_slow_mc);
1974 }
1975#ifdef CONFIG_IP_MROUTE
1976 if (type == RTN_MULTICAST) {
1977 if (IN_DEV_MFORWARD(in_dev) &&
1978 !ipv4_is_local_multicast(fl4->daddr)) {
1979 rth->dst.input = ip_mr_input;
1980 rth->dst.output = ip_mc_output;
1981 }
1982 }
1983#endif
1984 }
1985
1986 rt_set_nexthop(rth, fl4->daddr, res, fnhe, fi, type, 0);
1987
1988 return rth;
1989}
1990
1991/*
1992 * Major route resolver routine.
1993 */
1994
1995struct rtable *__ip_route_output_key(struct net *net, struct flowi4 *fl4)
1996{
1997 struct net_device *dev_out = NULL;
1998 __u8 tos = RT_FL_TOS(fl4);
1999 unsigned int flags = 0;
2000 struct fib_result res;
2001 struct rtable *rth;
2002 int orig_oif;
2003
2004 res.tclassid = 0;
2005 res.fi = NULL;
2006 res.table = NULL;
2007
2008 orig_oif = fl4->flowi4_oif;
2009
2010 fl4->flowi4_iif = LOOPBACK_IFINDEX;
2011 fl4->flowi4_tos = tos & IPTOS_RT_MASK;
2012 fl4->flowi4_scope = ((tos & RTO_ONLINK) ?
2013 RT_SCOPE_LINK : RT_SCOPE_UNIVERSE);
2014
2015 rcu_read_lock();
2016 if (fl4->saddr) {
2017 rth = ERR_PTR(-EINVAL);
2018 if (ipv4_is_multicast(fl4->saddr) ||
2019 ipv4_is_lbcast(fl4->saddr) ||
2020 ipv4_is_zeronet(fl4->saddr))
2021 goto out;
2022
2023 /* I removed check for oif == dev_out->oif here.
2024 It was wrong for two reasons:
2025 1. ip_dev_find(net, saddr) can return wrong iface, if saddr
2026 is assigned to multiple interfaces.
2027 2. Moreover, we are allowed to send packets with saddr
2028 of another iface. --ANK
2029 */
2030
2031 if (fl4->flowi4_oif == 0 &&
2032 (ipv4_is_multicast(fl4->daddr) ||
2033 ipv4_is_lbcast(fl4->daddr))) {
2034 /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2035 dev_out = __ip_dev_find(net, fl4->saddr, false);
2036 if (dev_out == NULL)
2037 goto out;
2038
2039 /* Special hack: user can direct multicasts
2040 and limited broadcast via necessary interface
2041 without fiddling with IP_MULTICAST_IF or IP_PKTINFO.
2042 This hack is not just for fun, it allows
2043 vic,vat and friends to work.
2044 They bind socket to loopback, set ttl to zero
2045 and expect that it will work.
2046 From the viewpoint of routing cache they are broken,
2047 because we are not allowed to build multicast path
2048 with loopback source addr (look, routing cache
2049 cannot know, that ttl is zero, so that packet
2050 will not leave this host and route is valid).
2051 Luckily, this hack is good workaround.
2052 */
2053
2054 fl4->flowi4_oif = dev_out->ifindex;
2055 goto make_route;
2056 }
2057
2058 if (!(fl4->flowi4_flags & FLOWI_FLAG_ANYSRC)) {
2059 /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2060 if (!__ip_dev_find(net, fl4->saddr, false))
2061 goto out;
2062 }
2063 }
2064
2065
2066 if (fl4->flowi4_oif) {
2067 dev_out = dev_get_by_index_rcu(net, fl4->flowi4_oif);
2068 rth = ERR_PTR(-ENODEV);
2069 if (dev_out == NULL)
2070 goto out;
2071
2072 /* RACE: Check return value of inet_select_addr instead. */
2073 if (!(dev_out->flags & IFF_UP) || !__in_dev_get_rcu(dev_out)) {
2074 rth = ERR_PTR(-ENETUNREACH);
2075 goto out;
2076 }
2077 if (ipv4_is_local_multicast(fl4->daddr) ||
2078 ipv4_is_lbcast(fl4->daddr)) {
2079 if (!fl4->saddr)
2080 fl4->saddr = inet_select_addr(dev_out, 0,
2081 RT_SCOPE_LINK);
2082 goto make_route;
2083 }
2084 if (!fl4->saddr) {
2085 if (ipv4_is_multicast(fl4->daddr))
2086 fl4->saddr = inet_select_addr(dev_out, 0,
2087 fl4->flowi4_scope);
2088 else if (!fl4->daddr)
2089 fl4->saddr = inet_select_addr(dev_out, 0,
2090 RT_SCOPE_HOST);
2091 }
2092 }
2093
2094 if (!fl4->daddr) {
2095 fl4->daddr = fl4->saddr;
2096 if (!fl4->daddr)
2097 fl4->daddr = fl4->saddr = htonl(INADDR_LOOPBACK);
2098 dev_out = net->loopback_dev;
2099 fl4->flowi4_oif = LOOPBACK_IFINDEX;
2100 res.type = RTN_LOCAL;
2101 flags |= RTCF_LOCAL;
2102 goto make_route;
2103 }
2104
2105 if (fib_lookup(net, fl4, &res)) {
2106 res.fi = NULL;
2107 res.table = NULL;
2108 if (fl4->flowi4_oif) {
2109 /* Apparently, routing tables are wrong. Assume,
2110 that the destination is on link.
2111
2112 WHY? DW.
2113 Because we are allowed to send to iface
2114 even if it has NO routes and NO assigned
2115 addresses. When oif is specified, routing
2116 tables are looked up with only one purpose:
2117 to catch if destination is gatewayed, rather than
2118 direct. Moreover, if MSG_DONTROUTE is set,
2119 we send packet, ignoring both routing tables
2120 and ifaddr state. --ANK
2121
2122
2123 We could make it even if oif is unknown,
2124 likely IPv6, but we do not.
2125 */
2126
2127 if (fl4->saddr == 0)
2128 fl4->saddr = inet_select_addr(dev_out, 0,
2129 RT_SCOPE_LINK);
2130 res.type = RTN_UNICAST;
2131 goto make_route;
2132 }
2133 rth = ERR_PTR(-ENETUNREACH);
2134 goto out;
2135 }
2136
2137 if (res.type == RTN_LOCAL) {
2138 if (!fl4->saddr) {
2139 if (res.fi->fib_prefsrc)
2140 fl4->saddr = res.fi->fib_prefsrc;
2141 else
2142 fl4->saddr = fl4->daddr;
2143 }
2144 dev_out = net->loopback_dev;
2145 fl4->flowi4_oif = dev_out->ifindex;
2146 flags |= RTCF_LOCAL;
2147 goto make_route;
2148 }
2149
2150#ifdef CONFIG_IP_ROUTE_MULTIPATH
2151 if (res.fi->fib_nhs > 1 && fl4->flowi4_oif == 0)
2152 fib_select_multipath(&res);
2153 else
2154#endif
2155 if (!res.prefixlen &&
2156 res.table->tb_num_default > 1 &&
2157 res.type == RTN_UNICAST && !fl4->flowi4_oif)
2158 fib_select_default(&res);
2159
2160 if (!fl4->saddr)
2161 fl4->saddr = FIB_RES_PREFSRC(net, res);
2162
2163 dev_out = FIB_RES_DEV(res);
2164 fl4->flowi4_oif = dev_out->ifindex;
2165
2166
2167make_route:
2168 rth = __mkroute_output(&res, fl4, orig_oif, dev_out, flags);
2169
2170out:
2171 rcu_read_unlock();
2172 return rth;
2173}
2174EXPORT_SYMBOL_GPL(__ip_route_output_key);
2175
2176static struct dst_entry *ipv4_blackhole_dst_check(struct dst_entry *dst, u32 cookie)
2177{
2178 return NULL;
2179}
2180
2181static unsigned int ipv4_blackhole_mtu(const struct dst_entry *dst)
2182{
2183 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2184
2185 return mtu ? : dst->dev->mtu;
2186}
2187
2188static void ipv4_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
2189 struct sk_buff *skb, u32 mtu)
2190{
2191}
2192
2193static void ipv4_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
2194 struct sk_buff *skb)
2195{
2196}
2197
2198static u32 *ipv4_rt_blackhole_cow_metrics(struct dst_entry *dst,
2199 unsigned long old)
2200{
2201 return NULL;
2202}
2203
2204static struct dst_ops ipv4_dst_blackhole_ops = {
2205 .family = AF_INET,
2206 .protocol = cpu_to_be16(ETH_P_IP),
2207 .check = ipv4_blackhole_dst_check,
2208 .mtu = ipv4_blackhole_mtu,
2209 .default_advmss = ipv4_default_advmss,
2210 .update_pmtu = ipv4_rt_blackhole_update_pmtu,
2211 .redirect = ipv4_rt_blackhole_redirect,
2212 .cow_metrics = ipv4_rt_blackhole_cow_metrics,
2213 .neigh_lookup = ipv4_neigh_lookup,
2214};
2215
2216struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2217{
2218 struct rtable *ort = (struct rtable *) dst_orig;
2219 struct rtable *rt;
2220
2221 rt = dst_alloc(&ipv4_dst_blackhole_ops, NULL, 1, DST_OBSOLETE_NONE, 0);
2222 if (rt) {
2223 struct dst_entry *new = &rt->dst;
2224
2225 new->__use = 1;
2226 new->input = dst_discard;
2227 new->output = dst_discard_sk;
2228
2229 new->dev = ort->dst.dev;
2230 if (new->dev)
2231 dev_hold(new->dev);
2232
2233 rt->rt_is_input = ort->rt_is_input;
2234 rt->rt_iif = ort->rt_iif;
2235 rt->rt_pmtu = ort->rt_pmtu;
2236
2237 rt->rt_genid = rt_genid_ipv4(net);
2238 rt->rt_flags = ort->rt_flags;
2239 rt->rt_type = ort->rt_type;
2240 rt->rt_gateway = ort->rt_gateway;
2241 rt->rt_uses_gateway = ort->rt_uses_gateway;
2242
2243 INIT_LIST_HEAD(&rt->rt_uncached);
2244
2245 dst_free(new);
2246 }
2247
2248 dst_release(dst_orig);
2249
2250 return rt ? &rt->dst : ERR_PTR(-ENOMEM);
2251}
2252
2253struct rtable *ip_route_output_flow(struct net *net, struct flowi4 *flp4,
2254 struct sock *sk)
2255{
2256 struct rtable *rt = __ip_route_output_key(net, flp4);
2257
2258 if (IS_ERR(rt))
2259 return rt;
2260
2261 if (flp4->flowi4_proto)
2262 rt = (struct rtable *) xfrm_lookup(net, &rt->dst,
2263 flowi4_to_flowi(flp4),
2264 sk, 0);
2265
2266 return rt;
2267}
2268EXPORT_SYMBOL_GPL(ip_route_output_flow);
2269
2270static int rt_fill_info(struct net *net, __be32 dst, __be32 src,
2271 struct flowi4 *fl4, struct sk_buff *skb, u32 portid,
2272 u32 seq, int event, int nowait, unsigned int flags)
2273{
2274 struct rtable *rt = skb_rtable(skb);
2275 struct rtmsg *r;
2276 struct nlmsghdr *nlh;
2277 unsigned long expires = 0;
2278 u32 error;
2279 u32 metrics[RTAX_MAX];
2280
2281 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*r), flags);
2282 if (nlh == NULL)
2283 return -EMSGSIZE;
2284
2285 r = nlmsg_data(nlh);
2286 r->rtm_family = AF_INET;
2287 r->rtm_dst_len = 32;
2288 r->rtm_src_len = 0;
2289 r->rtm_tos = fl4->flowi4_tos;
2290 r->rtm_table = RT_TABLE_MAIN;
2291 if (nla_put_u32(skb, RTA_TABLE, RT_TABLE_MAIN))
2292 goto nla_put_failure;
2293 r->rtm_type = rt->rt_type;
2294 r->rtm_scope = RT_SCOPE_UNIVERSE;
2295 r->rtm_protocol = RTPROT_UNSPEC;
2296 r->rtm_flags = (rt->rt_flags & ~0xFFFF) | RTM_F_CLONED;
2297 if (rt->rt_flags & RTCF_NOTIFY)
2298 r->rtm_flags |= RTM_F_NOTIFY;
2299
2300 if (nla_put_be32(skb, RTA_DST, dst))
2301 goto nla_put_failure;
2302 if (src) {
2303 r->rtm_src_len = 32;
2304 if (nla_put_be32(skb, RTA_SRC, src))
2305 goto nla_put_failure;
2306 }
2307 if (rt->dst.dev &&
2308 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2309 goto nla_put_failure;
2310#ifdef CONFIG_IP_ROUTE_CLASSID
2311 if (rt->dst.tclassid &&
2312 nla_put_u32(skb, RTA_FLOW, rt->dst.tclassid))
2313 goto nla_put_failure;
2314#endif
2315 if (!rt_is_input_route(rt) &&
2316 fl4->saddr != src) {
2317 if (nla_put_be32(skb, RTA_PREFSRC, fl4->saddr))
2318 goto nla_put_failure;
2319 }
2320 if (rt->rt_uses_gateway &&
2321 nla_put_be32(skb, RTA_GATEWAY, rt->rt_gateway))
2322 goto nla_put_failure;
2323
2324 expires = rt->dst.expires;
2325 if (expires) {
2326 unsigned long now = jiffies;
2327
2328 if (time_before(now, expires))
2329 expires -= now;
2330 else
2331 expires = 0;
2332 }
2333
2334 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
2335 if (rt->rt_pmtu && expires)
2336 metrics[RTAX_MTU - 1] = rt->rt_pmtu;
2337 if (rtnetlink_put_metrics(skb, metrics) < 0)
2338 goto nla_put_failure;
2339
2340 if (fl4->flowi4_mark &&
2341 nla_put_u32(skb, RTA_MARK, fl4->flowi4_mark))
2342 goto nla_put_failure;
2343
2344 error = rt->dst.error;
2345
2346 if (rt_is_input_route(rt)) {
2347#ifdef CONFIG_IP_MROUTE
2348 if (ipv4_is_multicast(dst) && !ipv4_is_local_multicast(dst) &&
2349 IPV4_DEVCONF_ALL(net, MC_FORWARDING)) {
2350 int err = ipmr_get_route(net, skb,
2351 fl4->saddr, fl4->daddr,
2352 r, nowait);
2353 if (err <= 0) {
2354 if (!nowait) {
2355 if (err == 0)
2356 return 0;
2357 goto nla_put_failure;
2358 } else {
2359 if (err == -EMSGSIZE)
2360 goto nla_put_failure;
2361 error = err;
2362 }
2363 }
2364 } else
2365#endif
2366 if (nla_put_u32(skb, RTA_IIF, skb->dev->ifindex))
2367 goto nla_put_failure;
2368 }
2369
2370 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, error) < 0)
2371 goto nla_put_failure;
2372
2373 return nlmsg_end(skb, nlh);
2374
2375nla_put_failure:
2376 nlmsg_cancel(skb, nlh);
2377 return -EMSGSIZE;
2378}
2379
2380static int inet_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
2381{
2382 struct net *net = sock_net(in_skb->sk);
2383 struct rtmsg *rtm;
2384 struct nlattr *tb[RTA_MAX+1];
2385 struct rtable *rt = NULL;
2386 struct flowi4 fl4;
2387 __be32 dst = 0;
2388 __be32 src = 0;
2389 u32 iif;
2390 int err;
2391 int mark;
2392 struct sk_buff *skb;
2393
2394 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv4_policy);
2395 if (err < 0)
2396 goto errout;
2397
2398 rtm = nlmsg_data(nlh);
2399
2400 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2401 if (skb == NULL) {
2402 err = -ENOBUFS;
2403 goto errout;
2404 }
2405
2406 /* Reserve room for dummy headers, this skb can pass
2407 through good chunk of routing engine.
2408 */
2409 skb_reset_mac_header(skb);
2410 skb_reset_network_header(skb);
2411
2412 /* Bugfix: need to give ip_route_input enough of an IP header to not gag. */
2413 ip_hdr(skb)->protocol = IPPROTO_ICMP;
2414 skb_reserve(skb, MAX_HEADER + sizeof(struct iphdr));
2415
2416 src = tb[RTA_SRC] ? nla_get_be32(tb[RTA_SRC]) : 0;
2417 dst = tb[RTA_DST] ? nla_get_be32(tb[RTA_DST]) : 0;
2418 iif = tb[RTA_IIF] ? nla_get_u32(tb[RTA_IIF]) : 0;
2419 mark = tb[RTA_MARK] ? nla_get_u32(tb[RTA_MARK]) : 0;
2420
2421 memset(&fl4, 0, sizeof(fl4));
2422 fl4.daddr = dst;
2423 fl4.saddr = src;
2424 fl4.flowi4_tos = rtm->rtm_tos;
2425 fl4.flowi4_oif = tb[RTA_OIF] ? nla_get_u32(tb[RTA_OIF]) : 0;
2426 fl4.flowi4_mark = mark;
2427
2428 if (iif) {
2429 struct net_device *dev;
2430
2431 dev = __dev_get_by_index(net, iif);
2432 if (dev == NULL) {
2433 err = -ENODEV;
2434 goto errout_free;
2435 }
2436
2437 skb->protocol = htons(ETH_P_IP);
2438 skb->dev = dev;
2439 skb->mark = mark;
2440 local_bh_disable();
2441 err = ip_route_input(skb, dst, src, rtm->rtm_tos, dev);
2442 local_bh_enable();
2443
2444 rt = skb_rtable(skb);
2445 if (err == 0 && rt->dst.error)
2446 err = -rt->dst.error;
2447 } else {
2448 rt = ip_route_output_key(net, &fl4);
2449
2450 err = 0;
2451 if (IS_ERR(rt))
2452 err = PTR_ERR(rt);
2453 }
2454
2455 if (err)
2456 goto errout_free;
2457
2458 skb_dst_set(skb, &rt->dst);
2459 if (rtm->rtm_flags & RTM_F_NOTIFY)
2460 rt->rt_flags |= RTCF_NOTIFY;
2461
2462 err = rt_fill_info(net, dst, src, &fl4, skb,
2463 NETLINK_CB(in_skb).portid, nlh->nlmsg_seq,
2464 RTM_NEWROUTE, 0, 0);
2465 if (err <= 0)
2466 goto errout_free;
2467
2468 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2469errout:
2470 return err;
2471
2472errout_free:
2473 kfree_skb(skb);
2474 goto errout;
2475}
2476
2477void ip_rt_multicast_event(struct in_device *in_dev)
2478{
2479 rt_cache_flush(dev_net(in_dev->dev));
2480}
2481
2482#ifdef CONFIG_SYSCTL
2483static int ip_rt_gc_timeout __read_mostly = RT_GC_TIMEOUT;
2484static int ip_rt_gc_interval __read_mostly = 60 * HZ;
2485static int ip_rt_gc_min_interval __read_mostly = HZ / 2;
2486static int ip_rt_gc_elasticity __read_mostly = 8;
2487
2488static int ipv4_sysctl_rtcache_flush(struct ctl_table *__ctl, int write,
2489 void __user *buffer,
2490 size_t *lenp, loff_t *ppos)
2491{
2492 struct net *net = (struct net *)__ctl->extra1;
2493
2494 if (write) {
2495 rt_cache_flush(net);
2496 fnhe_genid_bump(net);
2497 return 0;
2498 }
2499
2500 return -EINVAL;
2501}
2502
2503static struct ctl_table ipv4_route_table[] = {
2504 {
2505 .procname = "gc_thresh",
2506 .data = &ipv4_dst_ops.gc_thresh,
2507 .maxlen = sizeof(int),
2508 .mode = 0644,
2509 .proc_handler = proc_dointvec,
2510 },
2511 {
2512 .procname = "max_size",
2513 .data = &ip_rt_max_size,
2514 .maxlen = sizeof(int),
2515 .mode = 0644,
2516 .proc_handler = proc_dointvec,
2517 },
2518 {
2519 /* Deprecated. Use gc_min_interval_ms */
2520
2521 .procname = "gc_min_interval",
2522 .data = &ip_rt_gc_min_interval,
2523 .maxlen = sizeof(int),
2524 .mode = 0644,
2525 .proc_handler = proc_dointvec_jiffies,
2526 },
2527 {
2528 .procname = "gc_min_interval_ms",
2529 .data = &ip_rt_gc_min_interval,
2530 .maxlen = sizeof(int),
2531 .mode = 0644,
2532 .proc_handler = proc_dointvec_ms_jiffies,
2533 },
2534 {
2535 .procname = "gc_timeout",
2536 .data = &ip_rt_gc_timeout,
2537 .maxlen = sizeof(int),
2538 .mode = 0644,
2539 .proc_handler = proc_dointvec_jiffies,
2540 },
2541 {
2542 .procname = "gc_interval",
2543 .data = &ip_rt_gc_interval,
2544 .maxlen = sizeof(int),
2545 .mode = 0644,
2546 .proc_handler = proc_dointvec_jiffies,
2547 },
2548 {
2549 .procname = "redirect_load",
2550 .data = &ip_rt_redirect_load,
2551 .maxlen = sizeof(int),
2552 .mode = 0644,
2553 .proc_handler = proc_dointvec,
2554 },
2555 {
2556 .procname = "redirect_number",
2557 .data = &ip_rt_redirect_number,
2558 .maxlen = sizeof(int),
2559 .mode = 0644,
2560 .proc_handler = proc_dointvec,
2561 },
2562 {
2563 .procname = "redirect_silence",
2564 .data = &ip_rt_redirect_silence,
2565 .maxlen = sizeof(int),
2566 .mode = 0644,
2567 .proc_handler = proc_dointvec,
2568 },
2569 {
2570 .procname = "error_cost",
2571 .data = &ip_rt_error_cost,
2572 .maxlen = sizeof(int),
2573 .mode = 0644,
2574 .proc_handler = proc_dointvec,
2575 },
2576 {
2577 .procname = "error_burst",
2578 .data = &ip_rt_error_burst,
2579 .maxlen = sizeof(int),
2580 .mode = 0644,
2581 .proc_handler = proc_dointvec,
2582 },
2583 {
2584 .procname = "gc_elasticity",
2585 .data = &ip_rt_gc_elasticity,
2586 .maxlen = sizeof(int),
2587 .mode = 0644,
2588 .proc_handler = proc_dointvec,
2589 },
2590 {
2591 .procname = "mtu_expires",
2592 .data = &ip_rt_mtu_expires,
2593 .maxlen = sizeof(int),
2594 .mode = 0644,
2595 .proc_handler = proc_dointvec_jiffies,
2596 },
2597 {
2598 .procname = "min_pmtu",
2599 .data = &ip_rt_min_pmtu,
2600 .maxlen = sizeof(int),
2601 .mode = 0644,
2602 .proc_handler = proc_dointvec,
2603 },
2604 {
2605 .procname = "min_adv_mss",
2606 .data = &ip_rt_min_advmss,
2607 .maxlen = sizeof(int),
2608 .mode = 0644,
2609 .proc_handler = proc_dointvec,
2610 },
2611 { }
2612};
2613
2614static struct ctl_table ipv4_route_flush_table[] = {
2615 {
2616 .procname = "flush",
2617 .maxlen = sizeof(int),
2618 .mode = 0200,
2619 .proc_handler = ipv4_sysctl_rtcache_flush,
2620 },
2621 { },
2622};
2623
2624static __net_init int sysctl_route_net_init(struct net *net)
2625{
2626 struct ctl_table *tbl;
2627
2628 tbl = ipv4_route_flush_table;
2629 if (!net_eq(net, &init_net)) {
2630 tbl = kmemdup(tbl, sizeof(ipv4_route_flush_table), GFP_KERNEL);
2631 if (tbl == NULL)
2632 goto err_dup;
2633
2634 /* Don't export sysctls to unprivileged users */
2635 if (net->user_ns != &init_user_ns)
2636 tbl[0].procname = NULL;
2637 }
2638 tbl[0].extra1 = net;
2639
2640 net->ipv4.route_hdr = register_net_sysctl(net, "net/ipv4/route", tbl);
2641 if (net->ipv4.route_hdr == NULL)
2642 goto err_reg;
2643 return 0;
2644
2645err_reg:
2646 if (tbl != ipv4_route_flush_table)
2647 kfree(tbl);
2648err_dup:
2649 return -ENOMEM;
2650}
2651
2652static __net_exit void sysctl_route_net_exit(struct net *net)
2653{
2654 struct ctl_table *tbl;
2655
2656 tbl = net->ipv4.route_hdr->ctl_table_arg;
2657 unregister_net_sysctl_table(net->ipv4.route_hdr);
2658 BUG_ON(tbl == ipv4_route_flush_table);
2659 kfree(tbl);
2660}
2661
2662static __net_initdata struct pernet_operations sysctl_route_ops = {
2663 .init = sysctl_route_net_init,
2664 .exit = sysctl_route_net_exit,
2665};
2666#endif
2667
2668static __net_init int rt_genid_init(struct net *net)
2669{
2670 atomic_set(&net->ipv4.rt_genid, 0);
2671 atomic_set(&net->fnhe_genid, 0);
2672 get_random_bytes(&net->ipv4.dev_addr_genid,
2673 sizeof(net->ipv4.dev_addr_genid));
2674 return 0;
2675}
2676
2677static __net_initdata struct pernet_operations rt_genid_ops = {
2678 .init = rt_genid_init,
2679};
2680
2681static int __net_init ipv4_inetpeer_init(struct net *net)
2682{
2683 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
2684
2685 if (!bp)
2686 return -ENOMEM;
2687 inet_peer_base_init(bp);
2688 net->ipv4.peers = bp;
2689 return 0;
2690}
2691
2692static void __net_exit ipv4_inetpeer_exit(struct net *net)
2693{
2694 struct inet_peer_base *bp = net->ipv4.peers;
2695
2696 net->ipv4.peers = NULL;
2697 inetpeer_invalidate_tree(bp);
2698 kfree(bp);
2699}
2700
2701static __net_initdata struct pernet_operations ipv4_inetpeer_ops = {
2702 .init = ipv4_inetpeer_init,
2703 .exit = ipv4_inetpeer_exit,
2704};
2705
2706#ifdef CONFIG_IP_ROUTE_CLASSID
2707struct ip_rt_acct __percpu *ip_rt_acct __read_mostly;
2708#endif /* CONFIG_IP_ROUTE_CLASSID */
2709
2710int __init ip_rt_init(void)
2711{
2712 int rc = 0;
2713
2714#ifdef CONFIG_IP_ROUTE_CLASSID
2715 ip_rt_acct = __alloc_percpu(256 * sizeof(struct ip_rt_acct), __alignof__(struct ip_rt_acct));
2716 if (!ip_rt_acct)
2717 panic("IP: failed to allocate ip_rt_acct\n");
2718#endif
2719
2720 ipv4_dst_ops.kmem_cachep =
2721 kmem_cache_create("ip_dst_cache", sizeof(struct rtable), 0,
2722 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
2723
2724 ipv4_dst_blackhole_ops.kmem_cachep = ipv4_dst_ops.kmem_cachep;
2725
2726 if (dst_entries_init(&ipv4_dst_ops) < 0)
2727 panic("IP: failed to allocate ipv4_dst_ops counter\n");
2728
2729 if (dst_entries_init(&ipv4_dst_blackhole_ops) < 0)
2730 panic("IP: failed to allocate ipv4_dst_blackhole_ops counter\n");
2731
2732 ipv4_dst_ops.gc_thresh = ~0;
2733 ip_rt_max_size = INT_MAX;
2734
2735 devinet_init();
2736 ip_fib_init();
2737
2738 if (ip_rt_proc_init())
2739 pr_err("Unable to create route proc files\n");
2740#ifdef CONFIG_XFRM
2741 xfrm_init();
2742 xfrm4_init();
2743#endif
2744 rtnl_register(PF_INET, RTM_GETROUTE, inet_rtm_getroute, NULL, NULL);
2745
2746#ifdef CONFIG_SYSCTL
2747 register_pernet_subsys(&sysctl_route_ops);
2748#endif
2749 register_pernet_subsys(&rt_genid_ops);
2750 register_pernet_subsys(&ipv4_inetpeer_ops);
2751 return rc;
2752}
2753
2754#ifdef CONFIG_SYSCTL
2755/*
2756 * We really need to sanitize the damn ipv4 init order, then all
2757 * this nonsense will go away.
2758 */
2759void __init ip_static_sysctl_init(void)
2760{
2761 register_net_sysctl(&init_net, "net/ipv4/route", ipv4_route_table);
2762}
2763#endif
Ubuntu Artful kernel mirror