projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blob
? search:


f65a59878c636ee47d17540fa6dc44194e1c64a6
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / ipv4 / route.c
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 <linux/jhash.h>
93 #include <net/dst.h>
94 #include <net/net_namespace.h>
95 #include <net/protocol.h>
96 #include <net/ip.h>
97 #include <net/route.h>
98 #include <net/inetpeer.h>
99 #include <net/sock.h>
100 #include <net/ip_fib.h>
101 #include <net/arp.h>
102 #include <net/tcp.h>
103 #include <net/icmp.h>
104 #include <net/xfrm.h>
105 #include <net/netevent.h>
106 #include <net/rtnetlink.h>
107 #ifdef CONFIG_SYSCTL
108 #include <linux/sysctl.h>
109 #include <linux/kmemleak.h>
110 #endif
111 #include <net/secure_seq.h>
112
113 #define RT_FL_TOS(oldflp4) \
114 ((oldflp4)->flowi4_tos & (IPTOS_RT_MASK | RTO_ONLINK))
115
116 #define IP_MAX_MTU 0xFFF0
117
118 #define RT_GC_TIMEOUT (300*HZ)
119
120 static int ip_rt_max_size;
121 static int ip_rt_redirect_number __read_mostly = 9;
122 static int ip_rt_redirect_load __read_mostly = HZ / 50;
123 static int ip_rt_redirect_silence __read_mostly = ((HZ / 50) << (9 + 1));
124 static int ip_rt_error_cost __read_mostly = HZ;
125 static int ip_rt_error_burst __read_mostly = 5 * HZ;
126 static int ip_rt_mtu_expires __read_mostly = 10 * 60 * HZ;
127 static int ip_rt_min_pmtu __read_mostly = 512 + 20 + 20;
128 static int ip_rt_min_advmss __read_mostly = 256;
129
130 /*
131 * Interface to generic destination cache.
132 */
133
134 static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie);
135 static unsigned int ipv4_default_advmss(const struct dst_entry *dst);
136 static unsigned int ipv4_mtu(const struct dst_entry *dst);
137 static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst);
138 static void ipv4_link_failure(struct sk_buff *skb);
139 static void ip_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
140 struct sk_buff *skb, u32 mtu);
141 static void ip_do_redirect(struct dst_entry *dst, struct sock *sk,
142 struct sk_buff *skb);
143 static void ipv4_dst_destroy(struct dst_entry *dst);
144
145 static void ipv4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
146 int how)
147 {
148 }
149
150 static u32 *ipv4_cow_metrics(struct dst_entry *dst, unsigned long old)
151 {
152 WARN_ON(1);
153 return NULL;
154 }
155
156 static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst,
157 struct sk_buff *skb,
158 const void *daddr);
159
160 static struct dst_ops ipv4_dst_ops = {
161 .family = AF_INET,
162 .protocol = cpu_to_be16(ETH_P_IP),
163 .check = ipv4_dst_check,
164 .default_advmss = ipv4_default_advmss,
165 .mtu = ipv4_mtu,
166 .cow_metrics = ipv4_cow_metrics,
167 .destroy = ipv4_dst_destroy,
168 .ifdown = ipv4_dst_ifdown,
169 .negative_advice = ipv4_negative_advice,
170 .link_failure = ipv4_link_failure,
171 .update_pmtu = ip_rt_update_pmtu,
172 .redirect = ip_do_redirect,
173 .local_out = __ip_local_out,
174 .neigh_lookup = ipv4_neigh_lookup,
175 };
176
177 #define ECN_OR_COST(class) TC_PRIO_##class
178
179 const __u8 ip_tos2prio[16] = {
180 TC_PRIO_BESTEFFORT,
181 ECN_OR_COST(BESTEFFORT),
182 TC_PRIO_BESTEFFORT,
183 ECN_OR_COST(BESTEFFORT),
184 TC_PRIO_BULK,
185 ECN_OR_COST(BULK),
186 TC_PRIO_BULK,
187 ECN_OR_COST(BULK),
188 TC_PRIO_INTERACTIVE,
189 ECN_OR_COST(INTERACTIVE),
190 TC_PRIO_INTERACTIVE,
191 ECN_OR_COST(INTERACTIVE),
192 TC_PRIO_INTERACTIVE_BULK,
193 ECN_OR_COST(INTERACTIVE_BULK),
194 TC_PRIO_INTERACTIVE_BULK,
195 ECN_OR_COST(INTERACTIVE_BULK)
196 };
197 EXPORT_SYMBOL(ip_tos2prio);
198
199 static DEFINE_PER_CPU(struct rt_cache_stat, rt_cache_stat);
200 #define RT_CACHE_STAT_INC(field) __this_cpu_inc(rt_cache_stat.field)
201
202 #ifdef CONFIG_PROC_FS
203 static void *rt_cache_seq_start(struct seq_file *seq, loff_t *pos)
204 {
205 if (*pos)
206 return NULL;
207 return SEQ_START_TOKEN;
208 }
209
210 static void *rt_cache_seq_next(struct seq_file *seq, void *v, loff_t *pos)
211 {
212 ++*pos;
213 return NULL;
214 }
215
216 static void rt_cache_seq_stop(struct seq_file *seq, void *v)
217 {
218 }
219
220 static int rt_cache_seq_show(struct seq_file *seq, void *v)
221 {
222 if (v == SEQ_START_TOKEN)
223 seq_printf(seq, "%-127s\n",
224 "Iface\tDestination\tGateway \tFlags\t\tRefCnt\tUse\t"
225 "Metric\tSource\t\tMTU\tWindow\tIRTT\tTOS\tHHRef\t"
226 "HHUptod\tSpecDst");
227 return 0;
228 }
229
230 static const struct seq_operations rt_cache_seq_ops = {
231 .start = rt_cache_seq_start,
232 .next = rt_cache_seq_next,
233 .stop = rt_cache_seq_stop,
234 .show = rt_cache_seq_show,
235 };
236
237 static int rt_cache_seq_open(struct inode *inode, struct file *file)
238 {
239 return seq_open(file, &rt_cache_seq_ops);
240 }
241
242 static const struct file_operations rt_cache_seq_fops = {
243 .owner = THIS_MODULE,
244 .open = rt_cache_seq_open,
245 .read = seq_read,
246 .llseek = seq_lseek,
247 .release = seq_release,
248 };
249
250
251 static void *rt_cpu_seq_start(struct seq_file *seq, loff_t *pos)
252 {
253 int cpu;
254
255 if (*pos == 0)
256 return SEQ_START_TOKEN;
257
258 for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
259 if (!cpu_possible(cpu))
260 continue;
261 *pos = cpu+1;
262 return &per_cpu(rt_cache_stat, cpu);
263 }
264 return NULL;
265 }
266
267 static void *rt_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
268 {
269 int cpu;
270
271 for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
272 if (!cpu_possible(cpu))
273 continue;
274 *pos = cpu+1;
275 return &per_cpu(rt_cache_stat, cpu);
276 }
277 return NULL;
278
279 }
280
281 static void rt_cpu_seq_stop(struct seq_file *seq, void *v)
282 {
283
284 }
285
286 static int rt_cpu_seq_show(struct seq_file *seq, void *v)
287 {
288 struct rt_cache_stat *st = v;
289
290 if (v == SEQ_START_TOKEN) {
291 seq_printf(seq, "entries in_hit in_slow_tot in_slow_mc in_no_route in_brd in_martian_dst in_martian_src out_hit out_slow_tot out_slow_mc gc_total gc_ignored gc_goal_miss gc_dst_overflow in_hlist_search out_hlist_search\n");
292 return 0;
293 }
294
295 seq_printf(seq,"%08x %08x %08x %08x %08x %08x %08x %08x "
296 " %08x %08x %08x %08x %08x %08x %08x %08x %08x \n",
297 dst_entries_get_slow(&ipv4_dst_ops),
298 st->in_hit,
299 st->in_slow_tot,
300 st->in_slow_mc,
301 st->in_no_route,
302 st->in_brd,
303 st->in_martian_dst,
304 st->in_martian_src,
305
306 st->out_hit,
307 st->out_slow_tot,
308 st->out_slow_mc,
309
310 st->gc_total,
311 st->gc_ignored,
312 st->gc_goal_miss,
313 st->gc_dst_overflow,
314 st->in_hlist_search,
315 st->out_hlist_search
316 );
317 return 0;
318 }
319
320 static const struct seq_operations rt_cpu_seq_ops = {
321 .start = rt_cpu_seq_start,
322 .next = rt_cpu_seq_next,
323 .stop = rt_cpu_seq_stop,
324 .show = rt_cpu_seq_show,
325 };
326
327
328 static int rt_cpu_seq_open(struct inode *inode, struct file *file)
329 {
330 return seq_open(file, &rt_cpu_seq_ops);
331 }
332
333 static const struct file_operations rt_cpu_seq_fops = {
334 .owner = THIS_MODULE,
335 .open = rt_cpu_seq_open,
336 .read = seq_read,
337 .llseek = seq_lseek,
338 .release = seq_release,
339 };
340
341 #ifdef CONFIG_IP_ROUTE_CLASSID
342 static int rt_acct_proc_show(struct seq_file *m, void *v)
343 {
344 struct ip_rt_acct *dst, *src;
345 unsigned int i, j;
346
347 dst = kcalloc(256, sizeof(struct ip_rt_acct), GFP_KERNEL);
348 if (!dst)
349 return -ENOMEM;
350
351 for_each_possible_cpu(i) {
352 src = (struct ip_rt_acct *)per_cpu_ptr(ip_rt_acct, i);
353 for (j = 0; j < 256; j++) {
354 dst[j].o_bytes += src[j].o_bytes;
355 dst[j].o_packets += src[j].o_packets;
356 dst[j].i_bytes += src[j].i_bytes;
357 dst[j].i_packets += src[j].i_packets;
358 }
359 }
360
361 seq_write(m, dst, 256 * sizeof(struct ip_rt_acct));
362 kfree(dst);
363 return 0;
364 }
365
366 static int rt_acct_proc_open(struct inode *inode, struct file *file)
367 {
368 return single_open(file, rt_acct_proc_show, NULL);
369 }
370
371 static const struct file_operations rt_acct_proc_fops = {
372 .owner = THIS_MODULE,
373 .open = rt_acct_proc_open,
374 .read = seq_read,
375 .llseek = seq_lseek,
376 .release = single_release,
377 };
378 #endif
379
380 static int __net_init ip_rt_do_proc_init(struct net *net)
381 {
382 struct proc_dir_entry *pde;
383
384 pde = proc_create("rt_cache", S_IRUGO, net->proc_net,
385 &rt_cache_seq_fops);
386 if (!pde)
387 goto err1;
388
389 pde = proc_create("rt_cache", S_IRUGO,
390 net->proc_net_stat, &rt_cpu_seq_fops);
391 if (!pde)
392 goto err2;
393
394 #ifdef CONFIG_IP_ROUTE_CLASSID
395 pde = proc_create("rt_acct", 0, net->proc_net, &rt_acct_proc_fops);
396 if (!pde)
397 goto err3;
398 #endif
399 return 0;
400
401 #ifdef CONFIG_IP_ROUTE_CLASSID
402 err3:
403 remove_proc_entry("rt_cache", net->proc_net_stat);
404 #endif
405 err2:
406 remove_proc_entry("rt_cache", net->proc_net);
407 err1:
408 return -ENOMEM;
409 }
410
411 static void __net_exit ip_rt_do_proc_exit(struct net *net)
412 {
413 remove_proc_entry("rt_cache", net->proc_net_stat);
414 remove_proc_entry("rt_cache", net->proc_net);
415 #ifdef CONFIG_IP_ROUTE_CLASSID
416 remove_proc_entry("rt_acct", net->proc_net);
417 #endif
418 }
419
420 static struct pernet_operations ip_rt_proc_ops __net_initdata = {
421 .init = ip_rt_do_proc_init,
422 .exit = ip_rt_do_proc_exit,
423 };
424
425 static int __init ip_rt_proc_init(void)
426 {
427 return register_pernet_subsys(&ip_rt_proc_ops);
428 }
429
430 #else
431 static inline int ip_rt_proc_init(void)
432 {
433 return 0;
434 }
435 #endif /* CONFIG_PROC_FS */
436
437 static inline bool rt_is_expired(const struct rtable *rth)
438 {
439 return rth->rt_genid != rt_genid(dev_net(rth->dst.dev));
440 }
441
442 void rt_cache_flush(struct net *net)
443 {
444 rt_genid_bump(net);
445 }
446
447 static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst,
448 struct sk_buff *skb,
449 const void *daddr)
450 {
451 struct net_device *dev = dst->dev;
452 const __be32 *pkey = daddr;
453 const struct rtable *rt;
454 struct neighbour *n;
455
456 rt = (const struct rtable *) dst;
457 if (rt->rt_gateway)
458 pkey = (const __be32 *) &rt->rt_gateway;
459 else if (skb)
460 pkey = &ip_hdr(skb)->daddr;
461
462 n = __ipv4_neigh_lookup(dev, *(__force u32 *)pkey);
463 if (n)
464 return n;
465 return neigh_create(&arp_tbl, pkey, dev);
466 }
467
468 #define IP_IDENTS_SZ 2048u
469 struct ip_ident_bucket {
470 atomic_t id;
471 u32 stamp32;
472 };
473
474 static struct ip_ident_bucket *ip_idents __read_mostly;
475
476 /* In order to protect privacy, we add a perturbation to identifiers
477 * if one generator is seldom used. This makes hard for an attacker
478 * to infer how many packets were sent between two points in time.
479 */
480 u32 ip_idents_reserve(u32 hash, int segs)
481 {
482 struct ip_ident_bucket *bucket = ip_idents + hash % IP_IDENTS_SZ;
483 u32 old = ACCESS_ONCE(bucket->stamp32);
484 u32 now = (u32)jiffies;
485 u32 delta = 0;
486
487 if (old != now && cmpxchg(&bucket->stamp32, old, now) == old) {
488 u64 x = prandom_u32();
489
490 x *= (now - old);
491 delta = (u32)(x >> 32);
492 }
493
494 return atomic_add_return(segs + delta, &bucket->id) - segs;
495 }
496 EXPORT_SYMBOL(ip_idents_reserve);
497
498 void __ip_select_ident(struct iphdr *iph, int segs)
499 {
500 static u32 ip_idents_hashrnd __read_mostly;
501 static bool hashrnd_initialized = false;
502 u32 hash, id;
503
504 if (unlikely(!hashrnd_initialized)) {
505 hashrnd_initialized = true;
506 get_random_bytes(&ip_idents_hashrnd, sizeof(ip_idents_hashrnd));
507 }
508
509 hash = jhash_3words((__force u32)iph->daddr,
510 (__force u32)iph->saddr,
511 iph->protocol,
512 ip_idents_hashrnd);
513 id = ip_idents_reserve(hash, segs);
514 iph->id = htons(id);
515 }
516 EXPORT_SYMBOL(__ip_select_ident);
517
518 static void __build_flow_key(struct flowi4 *fl4, struct sock *sk,
519 const struct iphdr *iph,
520 int oif, u8 tos,
521 u8 prot, u32 mark, int flow_flags)
522 {
523 if (sk) {
524 const struct inet_sock *inet = inet_sk(sk);
525
526 oif = sk->sk_bound_dev_if;
527 mark = sk->sk_mark;
528 tos = RT_CONN_FLAGS(sk);
529 prot = inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol;
530 }
531 flowi4_init_output(fl4, oif, mark, tos,
532 RT_SCOPE_UNIVERSE, prot,
533 flow_flags,
534 iph->daddr, iph->saddr, 0, 0);
535 }
536
537 static void build_skb_flow_key(struct flowi4 *fl4, const struct sk_buff *skb,
538 const struct sock *sk)
539 {
540 const struct iphdr *iph = ip_hdr(skb);
541 int oif = skb->dev->ifindex;
542 u8 tos = RT_TOS(iph->tos);
543 u8 prot = iph->protocol;
544 u32 mark = skb->mark;
545
546 __build_flow_key(fl4, sk, iph, oif, tos, prot, mark, 0);
547 }
548
549 static void build_sk_flow_key(struct flowi4 *fl4, const struct sock *sk)
550 {
551 const struct inet_sock *inet = inet_sk(sk);
552 const struct ip_options_rcu *inet_opt;
553 __be32 daddr = inet->inet_daddr;
554
555 rcu_read_lock();
556 inet_opt = rcu_dereference(inet->inet_opt);
557 if (inet_opt && inet_opt->opt.srr)
558 daddr = inet_opt->opt.faddr;
559 flowi4_init_output(fl4, sk->sk_bound_dev_if, sk->sk_mark,
560 RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE,
561 inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol,
562 inet_sk_flowi_flags(sk),
563 daddr, inet->inet_saddr, 0, 0);
564 rcu_read_unlock();
565 }
566
567 static void ip_rt_build_flow_key(struct flowi4 *fl4, const struct sock *sk,
568 const struct sk_buff *skb)
569 {
570 if (skb)
571 build_skb_flow_key(fl4, skb, sk);
572 else
573 build_sk_flow_key(fl4, sk);
574 }
575
576 static inline void rt_free(struct rtable *rt)
577 {
578 call_rcu(&rt->dst.rcu_head, dst_rcu_free);
579 }
580
581 static DEFINE_SPINLOCK(fnhe_lock);
582
583 static struct fib_nh_exception *fnhe_oldest(struct fnhe_hash_bucket *hash)
584 {
585 struct fib_nh_exception *fnhe, *oldest;
586 struct rtable *orig;
587
588 oldest = rcu_dereference(hash->chain);
589 for (fnhe = rcu_dereference(oldest->fnhe_next); fnhe;
590 fnhe = rcu_dereference(fnhe->fnhe_next)) {
591 if (time_before(fnhe->fnhe_stamp, oldest->fnhe_stamp))
592 oldest = fnhe;
593 }
594 orig = rcu_dereference(oldest->fnhe_rth);
595 if (orig) {
596 RCU_INIT_POINTER(oldest->fnhe_rth, NULL);
597 rt_free(orig);
598 }
599 return oldest;
600 }
601
602 static inline u32 fnhe_hashfun(__be32 daddr)
603 {
604 u32 hval;
605
606 hval = (__force u32) daddr;
607 hval ^= (hval >> 11) ^ (hval >> 22);
608
609 return hval & (FNHE_HASH_SIZE - 1);
610 }
611
612 static 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 int depth;
618 u32 hval = fnhe_hashfun(daddr);
619
620 spin_lock_bh(&fnhe_lock);
621
622 hash = nh->nh_exceptions;
623 if (!hash) {
624 hash = kzalloc(FNHE_HASH_SIZE * sizeof(*hash), GFP_ATOMIC);
625 if (!hash)
626 goto out_unlock;
627 nh->nh_exceptions = hash;
628 }
629
630 hash += hval;
631
632 depth = 0;
633 for (fnhe = rcu_dereference(hash->chain); fnhe;
634 fnhe = rcu_dereference(fnhe->fnhe_next)) {
635 if (fnhe->fnhe_daddr == daddr)
636 break;
637 depth++;
638 }
639
640 if (fnhe) {
641 if (gw)
642 fnhe->fnhe_gw = gw;
643 if (pmtu) {
644 fnhe->fnhe_pmtu = pmtu;
645 fnhe->fnhe_expires = expires;
646 }
647 } else {
648 if (depth > FNHE_RECLAIM_DEPTH)
649 fnhe = fnhe_oldest(hash);
650 else {
651 fnhe = kzalloc(sizeof(*fnhe), GFP_ATOMIC);
652 if (!fnhe)
653 goto out_unlock;
654
655 fnhe->fnhe_next = hash->chain;
656 rcu_assign_pointer(hash->chain, fnhe);
657 }
658 fnhe->fnhe_daddr = daddr;
659 fnhe->fnhe_gw = gw;
660 fnhe->fnhe_pmtu = pmtu;
661 fnhe->fnhe_expires = expires;
662 }
663
664 fnhe->fnhe_stamp = jiffies;
665
666 out_unlock:
667 spin_unlock_bh(&fnhe_lock);
668 return;
669 }
670
671 static void __ip_do_redirect(struct rtable *rt, struct sk_buff *skb, struct flowi4 *fl4,
672 bool kill_route)
673 {
674 __be32 new_gw = icmp_hdr(skb)->un.gateway;
675 __be32 old_gw = ip_hdr(skb)->saddr;
676 struct net_device *dev = skb->dev;
677 struct in_device *in_dev;
678 struct fib_result res;
679 struct neighbour *n;
680 struct net *net;
681
682 switch (icmp_hdr(skb)->code & 7) {
683 case ICMP_REDIR_NET:
684 case ICMP_REDIR_NETTOS:
685 case ICMP_REDIR_HOST:
686 case ICMP_REDIR_HOSTTOS:
687 break;
688
689 default:
690 return;
691 }
692
693 if (rt->rt_gateway != old_gw)
694 return;
695
696 in_dev = __in_dev_get_rcu(dev);
697 if (!in_dev)
698 return;
699
700 net = dev_net(dev);
701 if (new_gw == old_gw || !IN_DEV_RX_REDIRECTS(in_dev) ||
702 ipv4_is_multicast(new_gw) || ipv4_is_lbcast(new_gw) ||
703 ipv4_is_zeronet(new_gw))
704 goto reject_redirect;
705
706 if (!IN_DEV_SHARED_MEDIA(in_dev)) {
707 if (!inet_addr_onlink(in_dev, new_gw, old_gw))
708 goto reject_redirect;
709 if (IN_DEV_SEC_REDIRECTS(in_dev) && ip_fib_check_default(new_gw, dev))
710 goto reject_redirect;
711 } else {
712 if (inet_addr_type(net, new_gw) != RTN_UNICAST)
713 goto reject_redirect;
714 }
715
716 n = __ipv4_neigh_lookup(rt->dst.dev, new_gw);
717 if (!n)
718 n = neigh_create(&arp_tbl, &new_gw, rt->dst.dev);
719 if (!IS_ERR(n)) {
720 if (!(n->nud_state & NUD_VALID)) {
721 neigh_event_send(n, NULL);
722 } else {
723 if (fib_lookup(net, fl4, &res) == 0) {
724 struct fib_nh *nh = &FIB_RES_NH(res);
725
726 update_or_create_fnhe(nh, fl4->daddr, new_gw,
727 0, 0);
728 }
729 if (kill_route)
730 rt->dst.obsolete = DST_OBSOLETE_KILL;
731 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, n);
732 }
733 neigh_release(n);
734 }
735 return;
736
737 reject_redirect:
738 #ifdef CONFIG_IP_ROUTE_VERBOSE
739 if (IN_DEV_LOG_MARTIANS(in_dev)) {
740 const struct iphdr *iph = (const struct iphdr *) skb->data;
741 __be32 daddr = iph->daddr;
742 __be32 saddr = iph->saddr;
743
744 net_info_ratelimited("Redirect from %pI4 on %s about %pI4 ignored\n"
745 " Advised path = %pI4 -> %pI4\n",
746 &old_gw, dev->name, &new_gw,
747 &saddr, &daddr);
748 }
749 #endif
750 ;
751 }
752
753 static void ip_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
754 {
755 struct rtable *rt;
756 struct flowi4 fl4;
757 const struct iphdr *iph = (const struct iphdr *) skb->data;
758 int oif = skb->dev->ifindex;
759 u8 tos = RT_TOS(iph->tos);
760 u8 prot = iph->protocol;
761 u32 mark = skb->mark;
762
763 rt = (struct rtable *) dst;
764
765 __build_flow_key(&fl4, sk, iph, oif, tos, prot, mark, 0);
766 __ip_do_redirect(rt, skb, &fl4, true);
767 }
768
769 static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst)
770 {
771 struct rtable *rt = (struct rtable *)dst;
772 struct dst_entry *ret = dst;
773
774 if (rt) {
775 if (dst->obsolete > 0) {
776 ip_rt_put(rt);
777 ret = NULL;
778 } else if ((rt->rt_flags & RTCF_REDIRECTED) ||
779 rt->dst.expires) {
780 ip_rt_put(rt);
781 ret = NULL;
782 }
783 }
784 return ret;
785 }
786
787 /*
788 * Algorithm:
789 * 1. The first ip_rt_redirect_number redirects are sent
790 * with exponential backoff, then we stop sending them at all,
791 * assuming that the host ignores our redirects.
792 * 2. If we did not see packets requiring redirects
793 * during ip_rt_redirect_silence, we assume that the host
794 * forgot redirected route and start to send redirects again.
795 *
796 * This algorithm is much cheaper and more intelligent than dumb load limiting
797 * in icmp.c.
798 *
799 * NOTE. Do not forget to inhibit load limiting for redirects (redundant)
800 * and "frag. need" (breaks PMTU discovery) in icmp.c.
801 */
802
803 void ip_rt_send_redirect(struct sk_buff *skb)
804 {
805 struct rtable *rt = skb_rtable(skb);
806 struct in_device *in_dev;
807 struct inet_peer *peer;
808 struct net *net;
809 int log_martians;
810
811 rcu_read_lock();
812 in_dev = __in_dev_get_rcu(rt->dst.dev);
813 if (!in_dev || !IN_DEV_TX_REDIRECTS(in_dev)) {
814 rcu_read_unlock();
815 return;
816 }
817 log_martians = IN_DEV_LOG_MARTIANS(in_dev);
818 rcu_read_unlock();
819
820 net = dev_net(rt->dst.dev);
821 peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, 1);
822 if (!peer) {
823 icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST,
824 rt_nexthop(rt, ip_hdr(skb)->daddr));
825 return;
826 }
827
828 /* No redirected packets during ip_rt_redirect_silence;
829 * reset the algorithm.
830 */
831 if (time_after(jiffies, peer->rate_last + ip_rt_redirect_silence))
832 peer->rate_tokens = 0;
833
834 /* Too many ignored redirects; do not send anything
835 * set dst.rate_last to the last seen redirected packet.
836 */
837 if (peer->rate_tokens >= ip_rt_redirect_number) {
838 peer->rate_last = jiffies;
839 goto out_put_peer;
840 }
841
842 /* Check for load limit; set rate_last to the latest sent
843 * redirect.
844 */
845 if (peer->rate_tokens == 0 ||
846 time_after(jiffies,
847 (peer->rate_last +
848 (ip_rt_redirect_load << peer->rate_tokens)))) {
849 __be32 gw = rt_nexthop(rt, ip_hdr(skb)->daddr);
850
851 icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, gw);
852 peer->rate_last = jiffies;
853 ++peer->rate_tokens;
854 #ifdef CONFIG_IP_ROUTE_VERBOSE
855 if (log_martians &&
856 peer->rate_tokens == ip_rt_redirect_number)
857 net_warn_ratelimited("host %pI4/if%d ignores redirects for %pI4 to %pI4\n",
858 &ip_hdr(skb)->saddr, inet_iif(skb),
859 &ip_hdr(skb)->daddr, &gw);
860 #endif
861 }
862 out_put_peer:
863 inet_putpeer(peer);
864 }
865
866 static int ip_error(struct sk_buff *skb)
867 {
868 struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
869 struct rtable *rt = skb_rtable(skb);
870 struct inet_peer *peer;
871 unsigned long now;
872 struct net *net;
873 bool send;
874 int code;
875
876 /* IP on this device is disabled. */
877 if (!in_dev)
878 goto out;
879
880 net = dev_net(rt->dst.dev);
881 if (!IN_DEV_FORWARD(in_dev)) {
882 switch (rt->dst.error) {
883 case EHOSTUNREACH:
884 IP_INC_STATS_BH(net, IPSTATS_MIB_INADDRERRORS);
885 break;
886
887 case ENETUNREACH:
888 IP_INC_STATS_BH(net, IPSTATS_MIB_INNOROUTES);
889 break;
890 }
891 goto out;
892 }
893
894 switch (rt->dst.error) {
895 case EINVAL:
896 default:
897 goto out;
898 case EHOSTUNREACH:
899 code = ICMP_HOST_UNREACH;
900 break;
901 case ENETUNREACH:
902 code = ICMP_NET_UNREACH;
903 IP_INC_STATS_BH(net, IPSTATS_MIB_INNOROUTES);
904 break;
905 case EACCES:
906 code = ICMP_PKT_FILTERED;
907 break;
908 }
909
910 peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, 1);
911
912 send = true;
913 if (peer) {
914 now = jiffies;
915 peer->rate_tokens += now - peer->rate_last;
916 if (peer->rate_tokens > ip_rt_error_burst)
917 peer->rate_tokens = ip_rt_error_burst;
918 peer->rate_last = now;
919 if (peer->rate_tokens >= ip_rt_error_cost)
920 peer->rate_tokens -= ip_rt_error_cost;
921 else
922 send = false;
923 inet_putpeer(peer);
924 }
925 if (send)
926 icmp_send(skb, ICMP_DEST_UNREACH, code, 0);
927
928 out: kfree_skb(skb);
929 return 0;
930 }
931
932 static void __ip_rt_update_pmtu(struct rtable *rt, struct flowi4 *fl4, u32 mtu)
933 {
934 struct dst_entry *dst = &rt->dst;
935 struct fib_result res;
936
937 if (dst_metric_locked(dst, RTAX_MTU))
938 return;
939
940 if (dst->dev->mtu < mtu)
941 return;
942
943 if (mtu < ip_rt_min_pmtu)
944 mtu = ip_rt_min_pmtu;
945
946 if (!rt->rt_pmtu) {
947 dst->obsolete = DST_OBSOLETE_KILL;
948 } else {
949 rt->rt_pmtu = mtu;
950 dst->expires = max(1UL, jiffies + ip_rt_mtu_expires);
951 }
952
953 rcu_read_lock();
954 if (fib_lookup(dev_net(dst->dev), fl4, &res) == 0) {
955 struct fib_nh *nh = &FIB_RES_NH(res);
956
957 update_or_create_fnhe(nh, fl4->daddr, 0, mtu,
958 jiffies + ip_rt_mtu_expires);
959 }
960 rcu_read_unlock();
961 }
962
963 static void ip_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
964 struct sk_buff *skb, u32 mtu)
965 {
966 struct rtable *rt = (struct rtable *) dst;
967 struct flowi4 fl4;
968
969 ip_rt_build_flow_key(&fl4, sk, skb);
970 __ip_rt_update_pmtu(rt, &fl4, mtu);
971 }
972
973 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu,
974 int oif, u32 mark, u8 protocol, int flow_flags)
975 {
976 const struct iphdr *iph = (const struct iphdr *) skb->data;
977 struct flowi4 fl4;
978 struct rtable *rt;
979
980 if (!mark)
981 mark = IP4_REPLY_MARK(net, skb->mark);
982
983 __build_flow_key(&fl4, NULL, iph, oif,
984 RT_TOS(iph->tos), protocol, mark, flow_flags);
985 rt = __ip_route_output_key(net, &fl4);
986 if (!IS_ERR(rt)) {
987 __ip_rt_update_pmtu(rt, &fl4, mtu);
988 ip_rt_put(rt);
989 }
990 }
991 EXPORT_SYMBOL_GPL(ipv4_update_pmtu);
992
993 static void __ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu)
994 {
995 const struct iphdr *iph = (const struct iphdr *) skb->data;
996 struct flowi4 fl4;
997 struct rtable *rt;
998
999 __build_flow_key(&fl4, sk, iph, 0, 0, 0, 0, 0);
1000
1001 if (!fl4.flowi4_mark)
1002 fl4.flowi4_mark = IP4_REPLY_MARK(sock_net(sk), skb->mark);
1003
1004 rt = __ip_route_output_key(sock_net(sk), &fl4);
1005 if (!IS_ERR(rt)) {
1006 __ip_rt_update_pmtu(rt, &fl4, mtu);
1007 ip_rt_put(rt);
1008 }
1009 }
1010
1011 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu)
1012 {
1013 const struct iphdr *iph = (const struct iphdr *) skb->data;
1014 struct flowi4 fl4;
1015 struct rtable *rt;
1016 struct dst_entry *odst = NULL;
1017 bool new = false;
1018
1019 bh_lock_sock(sk);
1020 odst = sk_dst_get(sk);
1021
1022 if (sock_owned_by_user(sk) || !odst) {
1023 __ipv4_sk_update_pmtu(skb, sk, mtu);
1024 goto out;
1025 }
1026
1027 __build_flow_key(&fl4, sk, iph, 0, 0, 0, 0, 0);
1028
1029 rt = (struct rtable *)odst;
1030 if (odst->obsolete && odst->ops->check(odst, 0) == NULL) {
1031 rt = ip_route_output_flow(sock_net(sk), &fl4, sk);
1032 if (IS_ERR(rt))
1033 goto out;
1034
1035 new = true;
1036 }
1037
1038 __ip_rt_update_pmtu((struct rtable *) rt->dst.path, &fl4, mtu);
1039
1040 if (!dst_check(&rt->dst, 0)) {
1041 if (new)
1042 dst_release(&rt->dst);
1043
1044 rt = ip_route_output_flow(sock_net(sk), &fl4, sk);
1045 if (IS_ERR(rt))
1046 goto out;
1047
1048 new = true;
1049 }
1050
1051 if (new)
1052 sk_dst_set(sk, &rt->dst);
1053
1054 out:
1055 bh_unlock_sock(sk);
1056 dst_release(odst);
1057 }
1058 EXPORT_SYMBOL_GPL(ipv4_sk_update_pmtu);
1059
1060 void ipv4_redirect(struct sk_buff *skb, struct net *net,
1061 int oif, u32 mark, u8 protocol, int flow_flags)
1062 {
1063 const struct iphdr *iph = (const struct iphdr *) skb->data;
1064 struct flowi4 fl4;
1065 struct rtable *rt;
1066
1067 __build_flow_key(&fl4, NULL, iph, oif,
1068 RT_TOS(iph->tos), protocol, mark, flow_flags);
1069 rt = __ip_route_output_key(net, &fl4);
1070 if (!IS_ERR(rt)) {
1071 __ip_do_redirect(rt, skb, &fl4, false);
1072 ip_rt_put(rt);
1073 }
1074 }
1075 EXPORT_SYMBOL_GPL(ipv4_redirect);
1076
1077 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk)
1078 {
1079 const struct iphdr *iph = (const struct iphdr *) skb->data;
1080 struct flowi4 fl4;
1081 struct rtable *rt;
1082
1083 __build_flow_key(&fl4, sk, iph, 0, 0, 0, 0, 0);
1084 rt = __ip_route_output_key(sock_net(sk), &fl4);
1085 if (!IS_ERR(rt)) {
1086 __ip_do_redirect(rt, skb, &fl4, false);
1087 ip_rt_put(rt);
1088 }
1089 }
1090 EXPORT_SYMBOL_GPL(ipv4_sk_redirect);
1091
1092 static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie)
1093 {
1094 struct rtable *rt = (struct rtable *) dst;
1095
1096 /* All IPV4 dsts are created with ->obsolete set to the value
1097 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1098 * into this function always.
1099 *
1100 * When a PMTU/redirect information update invalidates a
1101 * route, this is indicated by setting obsolete to
1102 * DST_OBSOLETE_KILL.
1103 */
1104 if (dst->obsolete == DST_OBSOLETE_KILL || rt_is_expired(rt))
1105 return NULL;
1106 return dst;
1107 }
1108
1109 static void ipv4_link_failure(struct sk_buff *skb)
1110 {
1111 struct rtable *rt;
1112
1113 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0);
1114
1115 rt = skb_rtable(skb);
1116 if (rt)
1117 dst_set_expires(&rt->dst, 0);
1118 }
1119
1120 static int ip_rt_bug(struct sk_buff *skb)
1121 {
1122 pr_debug("%s: %pI4 -> %pI4, %s\n",
1123 __func__, &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr,
1124 skb->dev ? skb->dev->name : "?");
1125 kfree_skb(skb);
1126 WARN_ON(1);
1127 return 0;
1128 }
1129
1130 /*
1131 We do not cache source address of outgoing interface,
1132 because it is used only by IP RR, TS and SRR options,
1133 so that it out of fast path.
1134
1135 BTW remember: "addr" is allowed to be not aligned
1136 in IP options!
1137 */
1138
1139 void ip_rt_get_source(u8 *addr, struct sk_buff *skb, struct rtable *rt)
1140 {
1141 __be32 src;
1142
1143 if (rt_is_output_route(rt))
1144 src = ip_hdr(skb)->saddr;
1145 else {
1146 struct fib_result res;
1147 struct flowi4 fl4;
1148 struct iphdr *iph;
1149
1150 iph = ip_hdr(skb);
1151
1152 memset(&fl4, 0, sizeof(fl4));
1153 fl4.daddr = iph->daddr;
1154 fl4.saddr = iph->saddr;
1155 fl4.flowi4_tos = RT_TOS(iph->tos);
1156 fl4.flowi4_oif = rt->dst.dev->ifindex;
1157 fl4.flowi4_iif = skb->dev->ifindex;
1158 fl4.flowi4_mark = skb->mark;
1159
1160 rcu_read_lock();
1161 if (fib_lookup(dev_net(rt->dst.dev), &fl4, &res) == 0)
1162 src = FIB_RES_PREFSRC(dev_net(rt->dst.dev), res);
1163 else
1164 src = inet_select_addr(rt->dst.dev,
1165 rt_nexthop(rt, iph->daddr),
1166 RT_SCOPE_UNIVERSE);
1167 rcu_read_unlock();
1168 }
1169 memcpy(addr, &src, 4);
1170 }
1171
1172 #ifdef CONFIG_IP_ROUTE_CLASSID
1173 static void set_class_tag(struct rtable *rt, u32 tag)
1174 {
1175 if (!(rt->dst.tclassid & 0xFFFF))
1176 rt->dst.tclassid |= tag & 0xFFFF;
1177 if (!(rt->dst.tclassid & 0xFFFF0000))
1178 rt->dst.tclassid |= tag & 0xFFFF0000;
1179 }
1180 #endif
1181
1182 static unsigned int ipv4_default_advmss(const struct dst_entry *dst)
1183 {
1184 unsigned int advmss = dst_metric_raw(dst, RTAX_ADVMSS);
1185
1186 if (advmss == 0) {
1187 advmss = max_t(unsigned int, dst->dev->mtu - 40,
1188 ip_rt_min_advmss);
1189 if (advmss > 65535 - 40)
1190 advmss = 65535 - 40;
1191 }
1192 return advmss;
1193 }
1194
1195 static unsigned int ipv4_mtu(const struct dst_entry *dst)
1196 {
1197 const struct rtable *rt = (const struct rtable *) dst;
1198 unsigned int mtu = rt->rt_pmtu;
1199
1200 if (!mtu || time_after_eq(jiffies, rt->dst.expires))
1201 mtu = dst_metric_raw(dst, RTAX_MTU);
1202
1203 if (mtu)
1204 return mtu;
1205
1206 mtu = dst->dev->mtu;
1207
1208 if (unlikely(dst_metric_locked(dst, RTAX_MTU))) {
1209 if (rt->rt_uses_gateway && mtu > 576)
1210 mtu = 576;
1211 }
1212
1213 if (mtu > IP_MAX_MTU)
1214 mtu = IP_MAX_MTU;
1215
1216 return mtu;
1217 }
1218
1219 static struct fib_nh_exception *find_exception(struct fib_nh *nh, __be32 daddr)
1220 {
1221 struct fnhe_hash_bucket *hash = nh->nh_exceptions;
1222 struct fib_nh_exception *fnhe;
1223 u32 hval;
1224
1225 if (!hash)
1226 return NULL;
1227
1228 hval = fnhe_hashfun(daddr);
1229
1230 for (fnhe = rcu_dereference(hash[hval].chain); fnhe;
1231 fnhe = rcu_dereference(fnhe->fnhe_next)) {
1232 if (fnhe->fnhe_daddr == daddr)
1233 return fnhe;
1234 }
1235 return NULL;
1236 }
1237
1238 static bool rt_bind_exception(struct rtable *rt, struct fib_nh_exception *fnhe,
1239 __be32 daddr)
1240 {
1241 bool ret = false;
1242
1243 spin_lock_bh(&fnhe_lock);
1244
1245 if (daddr == fnhe->fnhe_daddr) {
1246 struct rtable *orig = rcu_dereference(fnhe->fnhe_rth);
1247 if (orig && rt_is_expired(orig)) {
1248 fnhe->fnhe_gw = 0;
1249 fnhe->fnhe_pmtu = 0;
1250 fnhe->fnhe_expires = 0;
1251 }
1252 if (fnhe->fnhe_pmtu) {
1253 unsigned long expires = fnhe->fnhe_expires;
1254 unsigned long diff = expires - jiffies;
1255
1256 if (time_before(jiffies, expires)) {
1257 rt->rt_pmtu = fnhe->fnhe_pmtu;
1258 dst_set_expires(&rt->dst, diff);
1259 }
1260 }
1261 if (fnhe->fnhe_gw) {
1262 rt->rt_flags |= RTCF_REDIRECTED;
1263 rt->rt_gateway = fnhe->fnhe_gw;
1264 rt->rt_uses_gateway = 1;
1265 } else if (!rt->rt_gateway)
1266 rt->rt_gateway = daddr;
1267
1268 rcu_assign_pointer(fnhe->fnhe_rth, rt);
1269 if (orig)
1270 rt_free(orig);
1271
1272 fnhe->fnhe_stamp = jiffies;
1273 ret = true;
1274 }
1275 spin_unlock_bh(&fnhe_lock);
1276
1277 return ret;
1278 }
1279
1280 static bool rt_cache_route(struct fib_nh *nh, struct rtable *rt)
1281 {
1282 struct rtable *orig, *prev, **p;
1283 bool ret = true;
1284
1285 if (rt_is_input_route(rt)) {
1286 p = (struct rtable **)&nh->nh_rth_input;
1287 } else {
1288 p = (struct rtable **)__this_cpu_ptr(nh->nh_pcpu_rth_output);
1289 }
1290 orig = *p;
1291
1292 prev = cmpxchg(p, orig, rt);
1293 if (prev == orig) {
1294 if (orig)
1295 rt_free(orig);
1296 } else
1297 ret = false;
1298
1299 return ret;
1300 }
1301
1302 static DEFINE_SPINLOCK(rt_uncached_lock);
1303 static LIST_HEAD(rt_uncached_list);
1304
1305 static void rt_add_uncached_list(struct rtable *rt)
1306 {
1307 spin_lock_bh(&rt_uncached_lock);
1308 list_add_tail(&rt->rt_uncached, &rt_uncached_list);
1309 spin_unlock_bh(&rt_uncached_lock);
1310 }
1311
1312 static void ipv4_dst_destroy(struct dst_entry *dst)
1313 {
1314 struct rtable *rt = (struct rtable *) dst;
1315
1316 if (!list_empty(&rt->rt_uncached)) {
1317 spin_lock_bh(&rt_uncached_lock);
1318 list_del(&rt->rt_uncached);
1319 spin_unlock_bh(&rt_uncached_lock);
1320 }
1321 }
1322
1323 void rt_flush_dev(struct net_device *dev)
1324 {
1325 if (!list_empty(&rt_uncached_list)) {
1326 struct net *net = dev_net(dev);
1327 struct rtable *rt;
1328
1329 spin_lock_bh(&rt_uncached_lock);
1330 list_for_each_entry(rt, &rt_uncached_list, rt_uncached) {
1331 if (rt->dst.dev != dev)
1332 continue;
1333 rt->dst.dev = net->loopback_dev;
1334 dev_hold(rt->dst.dev);
1335 dev_put(dev);
1336 }
1337 spin_unlock_bh(&rt_uncached_lock);
1338 }
1339 }
1340
1341 static bool rt_cache_valid(const struct rtable *rt)
1342 {
1343 return rt &&
1344 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1345 !rt_is_expired(rt);
1346 }
1347
1348 static void rt_set_nexthop(struct rtable *rt, __be32 daddr,
1349 const struct fib_result *res,
1350 struct fib_nh_exception *fnhe,
1351 struct fib_info *fi, u16 type, u32 itag)
1352 {
1353 bool cached = false;
1354
1355 if (fi) {
1356 struct fib_nh *nh = &FIB_RES_NH(*res);
1357
1358 if (nh->nh_gw && nh->nh_scope == RT_SCOPE_LINK) {
1359 rt->rt_gateway = nh->nh_gw;
1360 rt->rt_uses_gateway = 1;
1361 }
1362 dst_init_metrics(&rt->dst, fi->fib_metrics, true);
1363 #ifdef CONFIG_IP_ROUTE_CLASSID
1364 rt->dst.tclassid = nh->nh_tclassid;
1365 #endif
1366 if (unlikely(fnhe))
1367 cached = rt_bind_exception(rt, fnhe, daddr);
1368 else if (!(rt->dst.flags & DST_NOCACHE))
1369 cached = rt_cache_route(nh, rt);
1370 if (unlikely(!cached)) {
1371 /* Routes we intend to cache in nexthop exception or
1372 * FIB nexthop have the DST_NOCACHE bit clear.
1373 * However, if we are unsuccessful at storing this
1374 * route into the cache we really need to set it.
1375 */
1376 rt->dst.flags |= DST_NOCACHE;
1377 if (!rt->rt_gateway)
1378 rt->rt_gateway = daddr;
1379 rt_add_uncached_list(rt);
1380 }
1381 } else
1382 rt_add_uncached_list(rt);
1383
1384 #ifdef CONFIG_IP_ROUTE_CLASSID
1385 #ifdef CONFIG_IP_MULTIPLE_TABLES
1386 set_class_tag(rt, res->tclassid);
1387 #endif
1388 set_class_tag(rt, itag);
1389 #endif
1390 }
1391
1392 static struct rtable *rt_dst_alloc(struct net_device *dev,
1393 bool nopolicy, bool noxfrm, bool will_cache)
1394 {
1395 return dst_alloc(&ipv4_dst_ops, dev, 1, DST_OBSOLETE_FORCE_CHK,
1396 (will_cache ? 0 : (DST_HOST | DST_NOCACHE)) |
1397 (nopolicy ? DST_NOPOLICY : 0) |
1398 (noxfrm ? DST_NOXFRM : 0));
1399 }
1400
1401 /* called in rcu_read_lock() section */
1402 static int ip_route_input_mc(struct sk_buff *skb, __be32 daddr, __be32 saddr,
1403 u8 tos, struct net_device *dev, int our)
1404 {
1405 struct rtable *rth;
1406 struct in_device *in_dev = __in_dev_get_rcu(dev);
1407 u32 itag = 0;
1408 int err;
1409
1410 /* Primary sanity checks. */
1411
1412 if (in_dev == NULL)
1413 return -EINVAL;
1414
1415 if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) ||
1416 skb->protocol != htons(ETH_P_IP))
1417 goto e_inval;
1418
1419 if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev)))
1420 if (ipv4_is_loopback(saddr))
1421 goto e_inval;
1422
1423 if (ipv4_is_zeronet(saddr)) {
1424 if (!ipv4_is_local_multicast(daddr))
1425 goto e_inval;
1426 } else {
1427 err = fib_validate_source(skb, saddr, 0, tos, 0, dev,
1428 in_dev, &itag);
1429 if (err < 0)
1430 goto e_err;
1431 }
1432 rth = rt_dst_alloc(dev_net(dev)->loopback_dev,
1433 IN_DEV_CONF_GET(in_dev, NOPOLICY), false, false);
1434 if (!rth)
1435 goto e_nobufs;
1436
1437 #ifdef CONFIG_IP_ROUTE_CLASSID
1438 rth->dst.tclassid = itag;
1439 #endif
1440 rth->dst.output = ip_rt_bug;
1441
1442 rth->rt_genid = rt_genid(dev_net(dev));
1443 rth->rt_flags = RTCF_MULTICAST;
1444 rth->rt_type = RTN_MULTICAST;
1445 rth->rt_is_input= 1;
1446 rth->rt_iif = 0;
1447 rth->rt_pmtu = 0;
1448 rth->rt_gateway = 0;
1449 rth->rt_uses_gateway = 0;
1450 INIT_LIST_HEAD(&rth->rt_uncached);
1451 if (our) {
1452 rth->dst.input= ip_local_deliver;
1453 rth->rt_flags |= RTCF_LOCAL;
1454 }
1455
1456 #ifdef CONFIG_IP_MROUTE
1457 if (!ipv4_is_local_multicast(daddr) && IN_DEV_MFORWARD(in_dev))
1458 rth->dst.input = ip_mr_input;
1459 #endif
1460 RT_CACHE_STAT_INC(in_slow_mc);
1461
1462 skb_dst_set(skb, &rth->dst);
1463 return 0;
1464
1465 e_nobufs:
1466 return -ENOBUFS;
1467 e_inval:
1468 return -EINVAL;
1469 e_err:
1470 return err;
1471 }
1472
1473
1474 static void ip_handle_martian_source(struct net_device *dev,
1475 struct in_device *in_dev,
1476 struct sk_buff *skb,
1477 __be32 daddr,
1478 __be32 saddr)
1479 {
1480 RT_CACHE_STAT_INC(in_martian_src);
1481 #ifdef CONFIG_IP_ROUTE_VERBOSE
1482 if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) {
1483 /*
1484 * RFC1812 recommendation, if source is martian,
1485 * the only hint is MAC header.
1486 */
1487 pr_warn("martian source %pI4 from %pI4, on dev %s\n",
1488 &daddr, &saddr, dev->name);
1489 if (dev->hard_header_len && skb_mac_header_was_set(skb)) {
1490 print_hex_dump(KERN_WARNING, "ll header: ",
1491 DUMP_PREFIX_OFFSET, 16, 1,
1492 skb_mac_header(skb),
1493 dev->hard_header_len, true);
1494 }
1495 }
1496 #endif
1497 }
1498
1499 /* called in rcu_read_lock() section */
1500 static int __mkroute_input(struct sk_buff *skb,
1501 const struct fib_result *res,
1502 struct in_device *in_dev,
1503 __be32 daddr, __be32 saddr, u32 tos)
1504 {
1505 struct rtable *rth;
1506 int err;
1507 struct in_device *out_dev;
1508 unsigned int flags = 0;
1509 bool do_cache;
1510 u32 itag = 0;
1511
1512 /* get a working reference to the output device */
1513 out_dev = __in_dev_get_rcu(FIB_RES_DEV(*res));
1514 if (out_dev == NULL) {
1515 net_crit_ratelimited("Bug in ip_route_input_slow(). Please report.\n");
1516 return -EINVAL;
1517 }
1518
1519 err = fib_validate_source(skb, saddr, daddr, tos, FIB_RES_OIF(*res),
1520 in_dev->dev, in_dev, &itag);
1521 if (err < 0) {
1522 ip_handle_martian_source(in_dev->dev, in_dev, skb, daddr,
1523 saddr);
1524
1525 goto cleanup;
1526 }
1527
1528 do_cache = res->fi && !itag;
1529 if (out_dev == in_dev && err && IN_DEV_TX_REDIRECTS(out_dev) &&
1530 skb->protocol == htons(ETH_P_IP) &&
1531 (IN_DEV_SHARED_MEDIA(out_dev) ||
1532 inet_addr_onlink(out_dev, saddr, FIB_RES_GW(*res))))
1533 IPCB(skb)->flags |= IPSKB_DOREDIRECT;
1534
1535 if (skb->protocol != htons(ETH_P_IP)) {
1536 /* Not IP (i.e. ARP). Do not create route, if it is
1537 * invalid for proxy arp. DNAT routes are always valid.
1538 *
1539 * Proxy arp feature have been extended to allow, ARP
1540 * replies back to the same interface, to support
1541 * Private VLAN switch technologies. See arp.c.
1542 */
1543 if (out_dev == in_dev &&
1544 IN_DEV_PROXY_ARP_PVLAN(in_dev) == 0) {
1545 err = -EINVAL;
1546 goto cleanup;
1547 }
1548 }
1549
1550 if (do_cache) {
1551 rth = rcu_dereference(FIB_RES_NH(*res).nh_rth_input);
1552 if (rt_cache_valid(rth)) {
1553 skb_dst_set_noref(skb, &rth->dst);
1554 goto out;
1555 }
1556 }
1557
1558 rth = rt_dst_alloc(out_dev->dev,
1559 IN_DEV_CONF_GET(in_dev, NOPOLICY),
1560 IN_DEV_CONF_GET(out_dev, NOXFRM), do_cache);
1561 if (!rth) {
1562 err = -ENOBUFS;
1563 goto cleanup;
1564 }
1565
1566 rth->rt_genid = rt_genid(dev_net(rth->dst.dev));
1567 rth->rt_flags = flags;
1568 rth->rt_type = res->type;
1569 rth->rt_is_input = 1;
1570 rth->rt_iif = 0;
1571 rth->rt_pmtu = 0;
1572 rth->rt_gateway = 0;
1573 rth->rt_uses_gateway = 0;
1574 INIT_LIST_HEAD(&rth->rt_uncached);
1575 RT_CACHE_STAT_INC(in_slow_tot);
1576
1577 rth->dst.input = ip_forward;
1578 rth->dst.output = ip_output;
1579
1580 rt_set_nexthop(rth, daddr, res, NULL, res->fi, res->type, itag);
1581 skb_dst_set(skb, &rth->dst);
1582 out:
1583 err = 0;
1584 cleanup:
1585 return err;
1586 }
1587
1588 static int ip_mkroute_input(struct sk_buff *skb,
1589 struct fib_result *res,
1590 const struct flowi4 *fl4,
1591 struct in_device *in_dev,
1592 __be32 daddr, __be32 saddr, u32 tos)
1593 {
1594 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1595 if (res->fi && res->fi->fib_nhs > 1)
1596 fib_select_multipath(res);
1597 #endif
1598
1599 /* create a routing cache entry */
1600 return __mkroute_input(skb, res, in_dev, daddr, saddr, tos);
1601 }
1602
1603 /*
1604 * NOTE. We drop all the packets that has local source
1605 * addresses, because every properly looped back packet
1606 * must have correct destination already attached by output routine.
1607 *
1608 * Such approach solves two big problems:
1609 * 1. Not simplex devices are handled properly.
1610 * 2. IP spoofing attempts are filtered with 100% of guarantee.
1611 * called with rcu_read_lock()
1612 */
1613
1614 static int ip_route_input_slow(struct sk_buff *skb, __be32 daddr, __be32 saddr,
1615 u8 tos, struct net_device *dev)
1616 {
1617 struct fib_result res;
1618 struct in_device *in_dev = __in_dev_get_rcu(dev);
1619 struct flowi4 fl4;
1620 unsigned int flags = 0;
1621 u32 itag = 0;
1622 struct rtable *rth;
1623 int err = -EINVAL;
1624 struct net *net = dev_net(dev);
1625 bool do_cache;
1626
1627 /* IP on this device is disabled. */
1628
1629 if (!in_dev)
1630 goto out;
1631
1632 /* Check for the most weird martians, which can be not detected
1633 by fib_lookup.
1634 */
1635
1636 if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr))
1637 goto martian_source;
1638
1639 res.fi = NULL;
1640 if (ipv4_is_lbcast(daddr) || (saddr == 0 && daddr == 0))
1641 goto brd_input;
1642
1643 /* Accept zero addresses only to limited broadcast;
1644 * I even do not know to fix it or not. Waiting for complains :-)
1645 */
1646 if (ipv4_is_zeronet(saddr))
1647 goto martian_source;
1648
1649 if (ipv4_is_zeronet(daddr))
1650 goto martian_destination;
1651
1652 /* Following code try to avoid calling IN_DEV_NET_ROUTE_LOCALNET(),
1653 * and call it once if daddr or/and saddr are loopback addresses
1654 */
1655 if (ipv4_is_loopback(daddr)) {
1656 if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev, net))
1657 goto martian_destination;
1658 } else if (ipv4_is_loopback(saddr)) {
1659 if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev, net))
1660 goto martian_source;
1661 }
1662
1663 /*
1664 * Now we are ready to route packet.
1665 */
1666 fl4.flowi4_oif = 0;
1667 fl4.flowi4_iif = dev->ifindex;
1668 fl4.flowi4_mark = skb->mark;
1669 fl4.flowi4_tos = tos;
1670 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
1671 fl4.daddr = daddr;
1672 fl4.saddr = saddr;
1673 err = fib_lookup(net, &fl4, &res);
1674 if (err != 0)
1675 goto no_route;
1676
1677 if (res.type == RTN_BROADCAST)
1678 goto brd_input;
1679
1680 if (res.type == RTN_LOCAL) {
1681 err = fib_validate_source(skb, saddr, daddr, tos,
1682 LOOPBACK_IFINDEX,
1683 dev, in_dev, &itag);
1684 if (err < 0)
1685 goto martian_source_keep_err;
1686 goto local_input;
1687 }
1688
1689 if (!IN_DEV_FORWARD(in_dev))
1690 goto no_route;
1691 if (res.type != RTN_UNICAST)
1692 goto martian_destination;
1693
1694 err = ip_mkroute_input(skb, &res, &fl4, in_dev, daddr, saddr, tos);
1695 out: return err;
1696
1697 brd_input:
1698 if (skb->protocol != htons(ETH_P_IP))
1699 goto e_inval;
1700
1701 if (!ipv4_is_zeronet(saddr)) {
1702 err = fib_validate_source(skb, saddr, 0, tos, 0, dev,
1703 in_dev, &itag);
1704 if (err < 0)
1705 goto martian_source_keep_err;
1706 }
1707 flags |= RTCF_BROADCAST;
1708 res.type = RTN_BROADCAST;
1709 RT_CACHE_STAT_INC(in_brd);
1710
1711 local_input:
1712 do_cache = false;
1713 if (res.fi) {
1714 if (!itag) {
1715 rth = rcu_dereference(FIB_RES_NH(res).nh_rth_input);
1716 if (rt_cache_valid(rth)) {
1717 skb_dst_set_noref(skb, &rth->dst);
1718 err = 0;
1719 goto out;
1720 }
1721 do_cache = true;
1722 }
1723 }
1724
1725 rth = rt_dst_alloc(net->loopback_dev,
1726 IN_DEV_CONF_GET(in_dev, NOPOLICY), false, do_cache);
1727 if (!rth)
1728 goto e_nobufs;
1729
1730 rth->dst.input= ip_local_deliver;
1731 rth->dst.output= ip_rt_bug;
1732 #ifdef CONFIG_IP_ROUTE_CLASSID
1733 rth->dst.tclassid = itag;
1734 #endif
1735
1736 rth->rt_genid = rt_genid(net);
1737 rth->rt_flags = flags|RTCF_LOCAL;
1738 rth->rt_type = res.type;
1739 rth->rt_is_input = 1;
1740 rth->rt_iif = 0;
1741 rth->rt_pmtu = 0;
1742 rth->rt_gateway = 0;
1743 rth->rt_uses_gateway = 0;
1744 INIT_LIST_HEAD(&rth->rt_uncached);
1745 RT_CACHE_STAT_INC(in_slow_tot);
1746 if (res.type == RTN_UNREACHABLE) {
1747 rth->dst.input= ip_error;
1748 rth->dst.error= -err;
1749 rth->rt_flags &= ~RTCF_LOCAL;
1750 }
1751 if (do_cache) {
1752 if (unlikely(!rt_cache_route(&FIB_RES_NH(res), rth))) {
1753 rth->dst.flags |= DST_NOCACHE;
1754 rt_add_uncached_list(rth);
1755 }
1756 }
1757 skb_dst_set(skb, &rth->dst);
1758 err = 0;
1759 goto out;
1760
1761 no_route:
1762 RT_CACHE_STAT_INC(in_no_route);
1763 res.type = RTN_UNREACHABLE;
1764 if (err == -ESRCH)
1765 err = -ENETUNREACH;
1766 goto local_input;
1767
1768 /*
1769 * Do not cache martian addresses: they should be logged (RFC1812)
1770 */
1771 martian_destination:
1772 RT_CACHE_STAT_INC(in_martian_dst);
1773 #ifdef CONFIG_IP_ROUTE_VERBOSE
1774 if (IN_DEV_LOG_MARTIANS(in_dev))
1775 net_warn_ratelimited("martian destination %pI4 from %pI4, dev %s\n",
1776 &daddr, &saddr, dev->name);
1777 #endif
1778
1779 e_inval:
1780 err = -EINVAL;
1781 goto out;
1782
1783 e_nobufs:
1784 err = -ENOBUFS;
1785 goto out;
1786
1787 martian_source:
1788 err = -EINVAL;
1789 martian_source_keep_err:
1790 ip_handle_martian_source(dev, in_dev, skb, daddr, saddr);
1791 goto out;
1792 }
1793
1794 int ip_route_input_noref(struct sk_buff *skb, __be32 daddr, __be32 saddr,
1795 u8 tos, struct net_device *dev)
1796 {
1797 int res;
1798
1799 tos &= IPTOS_RT_MASK;
1800 rcu_read_lock();
1801
1802 /* Multicast recognition logic is moved from route cache to here.
1803 The problem was that too many Ethernet cards have broken/missing
1804 hardware multicast filters :-( As result the host on multicasting
1805 network acquires a lot of useless route cache entries, sort of
1806 SDR messages from all the world. Now we try to get rid of them.
1807 Really, provided software IP multicast filter is organized
1808 reasonably (at least, hashed), it does not result in a slowdown
1809 comparing with route cache reject entries.
1810 Note, that multicast routers are not affected, because
1811 route cache entry is created eventually.
1812 */
1813 if (ipv4_is_multicast(daddr)) {
1814 struct in_device *in_dev = __in_dev_get_rcu(dev);
1815
1816 if (in_dev) {
1817 int our = ip_check_mc_rcu(in_dev, daddr, saddr,
1818 ip_hdr(skb)->protocol);
1819 if (our
1820 #ifdef CONFIG_IP_MROUTE
1821 ||
1822 (!ipv4_is_local_multicast(daddr) &&
1823 IN_DEV_MFORWARD(in_dev))
1824 #endif
1825 ) {
1826 int res = ip_route_input_mc(skb, daddr, saddr,
1827 tos, dev, our);
1828 rcu_read_unlock();
1829 return res;
1830 }
1831 }
1832 rcu_read_unlock();
1833 return -EINVAL;
1834 }
1835 res = ip_route_input_slow(skb, daddr, saddr, tos, dev);
1836 rcu_read_unlock();
1837 return res;
1838 }
1839 EXPORT_SYMBOL(ip_route_input_noref);
1840
1841 /* called with rcu_read_lock() */
1842 static struct rtable *__mkroute_output(const struct fib_result *res,
1843 const struct flowi4 *fl4, int orig_oif,
1844 struct net_device *dev_out,
1845 unsigned int flags)
1846 {
1847 struct fib_info *fi = res->fi;
1848 struct fib_nh_exception *fnhe;
1849 struct in_device *in_dev;
1850 u16 type = res->type;
1851 struct rtable *rth;
1852 bool do_cache;
1853
1854 in_dev = __in_dev_get_rcu(dev_out);
1855 if (!in_dev)
1856 return ERR_PTR(-EINVAL);
1857
1858 if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev)))
1859 if (ipv4_is_loopback(fl4->saddr) && !(dev_out->flags & IFF_LOOPBACK))
1860 return ERR_PTR(-EINVAL);
1861
1862 if (ipv4_is_lbcast(fl4->daddr))
1863 type = RTN_BROADCAST;
1864 else if (ipv4_is_multicast(fl4->daddr))
1865 type = RTN_MULTICAST;
1866 else if (ipv4_is_zeronet(fl4->daddr))
1867 return ERR_PTR(-EINVAL);
1868
1869 if (dev_out->flags & IFF_LOOPBACK)
1870 flags |= RTCF_LOCAL;
1871
1872 do_cache = true;
1873 if (type == RTN_BROADCAST) {
1874 flags |= RTCF_BROADCAST | RTCF_LOCAL;
1875 fi = NULL;
1876 } else if (type == RTN_MULTICAST) {
1877 flags |= RTCF_MULTICAST | RTCF_LOCAL;
1878 if (!ip_check_mc_rcu(in_dev, fl4->daddr, fl4->saddr,
1879 fl4->flowi4_proto))
1880 flags &= ~RTCF_LOCAL;
1881 else
1882 do_cache = false;
1883 /* If multicast route do not exist use
1884 * default one, but do not gateway in this case.
1885 * Yes, it is hack.
1886 */
1887 if (fi && res->prefixlen < 4)
1888 fi = NULL;
1889 } else if ((type == RTN_LOCAL) && (orig_oif != 0) &&
1890 (orig_oif != dev_out->ifindex)) {
1891 /* For local routes that require a particular output interface
1892 * we do not want to cache the result. Caching the result
1893 * causes incorrect behaviour when there are multiple source
1894 * addresses on the interface, the end result being that if the
1895 * intended recipient is waiting on that interface for the
1896 * packet he won't receive it because it will be delivered on
1897 * the loopback interface and the IP_PKTINFO ipi_ifindex will
1898 * be set to the loopback interface as well.
1899 */
1900 fi = NULL;
1901 }
1902
1903 fnhe = NULL;
1904 do_cache &= fi != NULL;
1905 if (do_cache) {
1906 struct rtable __rcu **prth;
1907 struct fib_nh *nh = &FIB_RES_NH(*res);
1908
1909 fnhe = find_exception(nh, fl4->daddr);
1910 if (fnhe)
1911 prth = &fnhe->fnhe_rth;
1912 else {
1913 if (unlikely(fl4->flowi4_flags &
1914 FLOWI_FLAG_KNOWN_NH &&
1915 !(nh->nh_gw &&
1916 nh->nh_scope == RT_SCOPE_LINK))) {
1917 do_cache = false;
1918 goto add;
1919 }
1920 prth = __this_cpu_ptr(nh->nh_pcpu_rth_output);
1921 }
1922 rth = rcu_dereference(*prth);
1923 if (rt_cache_valid(rth)) {
1924 dst_hold(&rth->dst);
1925 return rth;
1926 }
1927 }
1928
1929 add:
1930 rth = rt_dst_alloc(dev_out,
1931 IN_DEV_CONF_GET(in_dev, NOPOLICY),
1932 IN_DEV_CONF_GET(in_dev, NOXFRM),
1933 do_cache);
1934 if (!rth)
1935 return ERR_PTR(-ENOBUFS);
1936
1937 rth->dst.output = ip_output;
1938
1939 rth->rt_genid = rt_genid(dev_net(dev_out));
1940 rth->rt_flags = flags;
1941 rth->rt_type = type;
1942 rth->rt_is_input = 0;
1943 rth->rt_iif = orig_oif ? : 0;
1944 rth->rt_pmtu = 0;
1945 rth->rt_gateway = 0;
1946 rth->rt_uses_gateway = 0;
1947 INIT_LIST_HEAD(&rth->rt_uncached);
1948
1949 RT_CACHE_STAT_INC(out_slow_tot);
1950
1951 if (flags & RTCF_LOCAL)
1952 rth->dst.input = ip_local_deliver;
1953 if (flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
1954 if (flags & RTCF_LOCAL &&
1955 !(dev_out->flags & IFF_LOOPBACK)) {
1956 rth->dst.output = ip_mc_output;
1957 RT_CACHE_STAT_INC(out_slow_mc);
1958 }
1959 #ifdef CONFIG_IP_MROUTE
1960 if (type == RTN_MULTICAST) {
1961 if (IN_DEV_MFORWARD(in_dev) &&
1962 !ipv4_is_local_multicast(fl4->daddr)) {
1963 rth->dst.input = ip_mr_input;
1964 rth->dst.output = ip_mc_output;
1965 }
1966 }
1967 #endif
1968 }
1969
1970 rt_set_nexthop(rth, fl4->daddr, res, fnhe, fi, type, 0);
1971
1972 return rth;
1973 }
1974
1975 /*
1976 * Major route resolver routine.
1977 */
1978
1979 struct rtable *__ip_route_output_key(struct net *net, struct flowi4 *fl4)
1980 {
1981 struct net_device *dev_out = NULL;
1982 __u8 tos = RT_FL_TOS(fl4);
1983 unsigned int flags = 0;
1984 struct fib_result res;
1985 struct rtable *rth;
1986 int orig_oif;
1987
1988 res.tclassid = 0;
1989 res.fi = NULL;
1990 res.table = NULL;
1991
1992 orig_oif = fl4->flowi4_oif;
1993
1994 fl4->flowi4_iif = LOOPBACK_IFINDEX;
1995 fl4->flowi4_tos = tos & IPTOS_RT_MASK;
1996 fl4->flowi4_scope = ((tos & RTO_ONLINK) ?
1997 RT_SCOPE_LINK : RT_SCOPE_UNIVERSE);
1998
1999 rcu_read_lock();
2000 if (fl4->saddr) {
2001 rth = ERR_PTR(-EINVAL);
2002 if (ipv4_is_multicast(fl4->saddr) ||
2003 ipv4_is_lbcast(fl4->saddr) ||
2004 ipv4_is_zeronet(fl4->saddr))
2005 goto out;
2006
2007 /* I removed check for oif == dev_out->oif here.
2008 It was wrong for two reasons:
2009 1. ip_dev_find(net, saddr) can return wrong iface, if saddr
2010 is assigned to multiple interfaces.
2011 2. Moreover, we are allowed to send packets with saddr
2012 of another iface. --ANK
2013 */
2014
2015 if (fl4->flowi4_oif == 0 &&
2016 (ipv4_is_multicast(fl4->daddr) ||
2017 ipv4_is_lbcast(fl4->daddr))) {
2018 /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2019 dev_out = __ip_dev_find(net, fl4->saddr, false);
2020 if (dev_out == NULL)
2021 goto out;
2022
2023 /* Special hack: user can direct multicasts
2024 and limited broadcast via necessary interface
2025 without fiddling with IP_MULTICAST_IF or IP_PKTINFO.
2026 This hack is not just for fun, it allows
2027 vic,vat and friends to work.
2028 They bind socket to loopback, set ttl to zero
2029 and expect that it will work.
2030 From the viewpoint of routing cache they are broken,
2031 because we are not allowed to build multicast path
2032 with loopback source addr (look, routing cache
2033 cannot know, that ttl is zero, so that packet
2034 will not leave this host and route is valid).
2035 Luckily, this hack is good workaround.
2036 */
2037
2038 fl4->flowi4_oif = dev_out->ifindex;
2039 goto make_route;
2040 }
2041
2042 if (!(fl4->flowi4_flags & FLOWI_FLAG_ANYSRC)) {
2043 /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2044 if (!__ip_dev_find(net, fl4->saddr, false))
2045 goto out;
2046 }
2047 }
2048
2049
2050 if (fl4->flowi4_oif) {
2051 dev_out = dev_get_by_index_rcu(net, fl4->flowi4_oif);
2052 rth = ERR_PTR(-ENODEV);
2053 if (dev_out == NULL)
2054 goto out;
2055
2056 /* RACE: Check return value of inet_select_addr instead. */
2057 if (!(dev_out->flags & IFF_UP) || !__in_dev_get_rcu(dev_out)) {
2058 rth = ERR_PTR(-ENETUNREACH);
2059 goto out;
2060 }
2061 if (ipv4_is_local_multicast(fl4->daddr) ||
2062 ipv4_is_lbcast(fl4->daddr)) {
2063 if (!fl4->saddr)
2064 fl4->saddr = inet_select_addr(dev_out, 0,
2065 RT_SCOPE_LINK);
2066 goto make_route;
2067 }
2068 if (!fl4->saddr) {
2069 if (ipv4_is_multicast(fl4->daddr))
2070 fl4->saddr = inet_select_addr(dev_out, 0,
2071 fl4->flowi4_scope);
2072 else if (!fl4->daddr)
2073 fl4->saddr = inet_select_addr(dev_out, 0,
2074 RT_SCOPE_HOST);
2075 }
2076 }
2077
2078 if (!fl4->daddr) {
2079 fl4->daddr = fl4->saddr;
2080 if (!fl4->daddr)
2081 fl4->daddr = fl4->saddr = htonl(INADDR_LOOPBACK);
2082 dev_out = net->loopback_dev;
2083 fl4->flowi4_oif = LOOPBACK_IFINDEX;
2084 res.type = RTN_LOCAL;
2085 flags |= RTCF_LOCAL;
2086 goto make_route;
2087 }
2088
2089 if (fib_lookup(net, fl4, &res)) {
2090 res.fi = NULL;
2091 res.table = NULL;
2092 if (fl4->flowi4_oif) {
2093 /* Apparently, routing tables are wrong. Assume,
2094 that the destination is on link.
2095
2096 WHY? DW.
2097 Because we are allowed to send to iface
2098 even if it has NO routes and NO assigned
2099 addresses. When oif is specified, routing
2100 tables are looked up with only one purpose:
2101 to catch if destination is gatewayed, rather than
2102 direct. Moreover, if MSG_DONTROUTE is set,
2103 we send packet, ignoring both routing tables
2104 and ifaddr state. --ANK
2105
2106
2107 We could make it even if oif is unknown,
2108 likely IPv6, but we do not.
2109 */
2110
2111 if (fl4->saddr == 0)
2112 fl4->saddr = inet_select_addr(dev_out, 0,
2113 RT_SCOPE_LINK);
2114 res.type = RTN_UNICAST;
2115 goto make_route;
2116 }
2117 rth = ERR_PTR(-ENETUNREACH);
2118 goto out;
2119 }
2120
2121 if (res.type == RTN_LOCAL) {
2122 if (!fl4->saddr) {
2123 if (res.fi->fib_prefsrc)
2124 fl4->saddr = res.fi->fib_prefsrc;
2125 else
2126 fl4->saddr = fl4->daddr;
2127 }
2128 dev_out = net->loopback_dev;
2129 fl4->flowi4_oif = dev_out->ifindex;
2130 flags |= RTCF_LOCAL;
2131 goto make_route;
2132 }
2133
2134 #ifdef CONFIG_IP_ROUTE_MULTIPATH
2135 if (res.fi->fib_nhs > 1 && fl4->flowi4_oif == 0)
2136 fib_select_multipath(&res);
2137 else
2138 #endif
2139 if (!res.prefixlen &&
2140 res.table->tb_num_default > 1 &&
2141 res.type == RTN_UNICAST && !fl4->flowi4_oif)
2142 fib_select_default(&res);
2143
2144 if (!fl4->saddr)
2145 fl4->saddr = FIB_RES_PREFSRC(net, res);
2146
2147 dev_out = FIB_RES_DEV(res);
2148 fl4->flowi4_oif = dev_out->ifindex;
2149
2150
2151 make_route:
2152 rth = __mkroute_output(&res, fl4, orig_oif, dev_out, flags);
2153
2154 out:
2155 rcu_read_unlock();
2156 return rth;
2157 }
2158 EXPORT_SYMBOL_GPL(__ip_route_output_key);
2159
2160 static struct dst_entry *ipv4_blackhole_dst_check(struct dst_entry *dst, u32 cookie)
2161 {
2162 return NULL;
2163 }
2164
2165 static unsigned int ipv4_blackhole_mtu(const struct dst_entry *dst)
2166 {
2167 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2168
2169 return mtu ? : dst->dev->mtu;
2170 }
2171
2172 static void ipv4_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
2173 struct sk_buff *skb, u32 mtu)
2174 {
2175 }
2176
2177 static void ipv4_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
2178 struct sk_buff *skb)
2179 {
2180 }
2181
2182 static u32 *ipv4_rt_blackhole_cow_metrics(struct dst_entry *dst,
2183 unsigned long old)
2184 {
2185 return NULL;
2186 }
2187
2188 static struct dst_ops ipv4_dst_blackhole_ops = {
2189 .family = AF_INET,
2190 .protocol = cpu_to_be16(ETH_P_IP),
2191 .check = ipv4_blackhole_dst_check,
2192 .mtu = ipv4_blackhole_mtu,
2193 .default_advmss = ipv4_default_advmss,
2194 .update_pmtu = ipv4_rt_blackhole_update_pmtu,
2195 .redirect = ipv4_rt_blackhole_redirect,
2196 .cow_metrics = ipv4_rt_blackhole_cow_metrics,
2197 .neigh_lookup = ipv4_neigh_lookup,
2198 };
2199
2200 struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2201 {
2202 struct rtable *ort = (struct rtable *) dst_orig;
2203 struct rtable *rt;
2204
2205 rt = dst_alloc(&ipv4_dst_blackhole_ops, NULL, 1, DST_OBSOLETE_NONE, 0);
2206 if (rt) {
2207 struct dst_entry *new = &rt->dst;
2208
2209 new->__use = 1;
2210 new->input = dst_discard;
2211 new->output = dst_discard;
2212
2213 new->dev = ort->dst.dev;
2214 if (new->dev)
2215 dev_hold(new->dev);
2216
2217 rt->rt_is_input = ort->rt_is_input;
2218 rt->rt_iif = ort->rt_iif;
2219 rt->rt_pmtu = ort->rt_pmtu;
2220
2221 rt->rt_genid = rt_genid(net);
2222 rt->rt_flags = ort->rt_flags;
2223 rt->rt_type = ort->rt_type;
2224 rt->rt_gateway = ort->rt_gateway;
2225 rt->rt_uses_gateway = ort->rt_uses_gateway;
2226
2227 INIT_LIST_HEAD(&rt->rt_uncached);
2228
2229 dst_free(new);
2230 }
2231
2232 dst_release(dst_orig);
2233
2234 return rt ? &rt->dst : ERR_PTR(-ENOMEM);
2235 }
2236
2237 struct rtable *ip_route_output_flow(struct net *net, struct flowi4 *flp4,
2238 struct sock *sk)
2239 {
2240 struct rtable *rt = __ip_route_output_key(net, flp4);
2241
2242 if (IS_ERR(rt))
2243 return rt;
2244
2245 if (flp4->flowi4_proto)
2246 rt = (struct rtable *) xfrm_lookup(net, &rt->dst,
2247 flowi4_to_flowi(flp4),
2248 sk, 0);
2249
2250 return rt;
2251 }
2252 EXPORT_SYMBOL_GPL(ip_route_output_flow);
2253
2254 static int rt_fill_info(struct net *net, __be32 dst, __be32 src,
2255 struct flowi4 *fl4, struct sk_buff *skb, u32 portid,
2256 u32 seq, int event, int nowait, unsigned int flags)
2257 {
2258 struct rtable *rt = skb_rtable(skb);
2259 struct rtmsg *r;
2260 struct nlmsghdr *nlh;
2261 unsigned long expires = 0;
2262 u32 error;
2263 u32 metrics[RTAX_MAX];
2264
2265 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*r), flags);
2266 if (nlh == NULL)
2267 return -EMSGSIZE;
2268
2269 r = nlmsg_data(nlh);
2270 r->rtm_family = AF_INET;
2271 r->rtm_dst_len = 32;
2272 r->rtm_src_len = 0;
2273 r->rtm_tos = fl4->flowi4_tos;
2274 r->rtm_table = RT_TABLE_MAIN;
2275 if (nla_put_u32(skb, RTA_TABLE, RT_TABLE_MAIN))
2276 goto nla_put_failure;
2277 r->rtm_type = rt->rt_type;
2278 r->rtm_scope = RT_SCOPE_UNIVERSE;
2279 r->rtm_protocol = RTPROT_UNSPEC;
2280 r->rtm_flags = (rt->rt_flags & ~0xFFFF) | RTM_F_CLONED;
2281 if (rt->rt_flags & RTCF_NOTIFY)
2282 r->rtm_flags |= RTM_F_NOTIFY;
2283 if (IPCB(skb)->flags & IPSKB_DOREDIRECT)
2284 r->rtm_flags |= RTCF_DOREDIRECT;
2285
2286 if (nla_put_be32(skb, RTA_DST, dst))
2287 goto nla_put_failure;
2288 if (src) {
2289 r->rtm_src_len = 32;
2290 if (nla_put_be32(skb, RTA_SRC, src))
2291 goto nla_put_failure;
2292 }
2293 if (rt->dst.dev &&
2294 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2295 goto nla_put_failure;
2296 #ifdef CONFIG_IP_ROUTE_CLASSID
2297 if (rt->dst.tclassid &&
2298 nla_put_u32(skb, RTA_FLOW, rt->dst.tclassid))
2299 goto nla_put_failure;
2300 #endif
2301 if (!rt_is_input_route(rt) &&
2302 fl4->saddr != src) {
2303 if (nla_put_be32(skb, RTA_PREFSRC, fl4->saddr))
2304 goto nla_put_failure;
2305 }
2306 if (rt->rt_uses_gateway &&
2307 nla_put_be32(skb, RTA_GATEWAY, rt->rt_gateway))
2308 goto nla_put_failure;
2309
2310 expires = rt->dst.expires;
2311 if (expires) {
2312 unsigned long now = jiffies;
2313
2314 if (time_before(now, expires))
2315 expires -= now;
2316 else
2317 expires = 0;
2318 }
2319
2320 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
2321 if (rt->rt_pmtu && expires)
2322 metrics[RTAX_MTU - 1] = rt->rt_pmtu;
2323 if (rtnetlink_put_metrics(skb, metrics) < 0)
2324 goto nla_put_failure;
2325
2326 if (fl4->flowi4_mark &&
2327 nla_put_u32(skb, RTA_MARK, fl4->flowi4_mark))
2328 goto nla_put_failure;
2329
2330 error = rt->dst.error;
2331
2332 if (rt_is_input_route(rt)) {
2333 #ifdef CONFIG_IP_MROUTE
2334 if (ipv4_is_multicast(dst) && !ipv4_is_local_multicast(dst) &&
2335 IPV4_DEVCONF_ALL(net, MC_FORWARDING)) {
2336 int err = ipmr_get_route(net, skb,
2337 fl4->saddr, fl4->daddr,
2338 r, nowait, portid);
2339
2340 if (err <= 0) {
2341 if (!nowait) {
2342 if (err == 0)
2343 return 0;
2344 goto nla_put_failure;
2345 } else {
2346 if (err == -EMSGSIZE)
2347 goto nla_put_failure;
2348 error = err;
2349 }
2350 }
2351 } else
2352 #endif
2353 if (nla_put_u32(skb, RTA_IIF, skb->dev->ifindex))
2354 goto nla_put_failure;
2355 }
2356
2357 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, error) < 0)
2358 goto nla_put_failure;
2359
2360 return nlmsg_end(skb, nlh);
2361
2362 nla_put_failure:
2363 nlmsg_cancel(skb, nlh);
2364 return -EMSGSIZE;
2365 }
2366
2367 static int inet_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
2368 {
2369 struct net *net = sock_net(in_skb->sk);
2370 struct rtmsg *rtm;
2371 struct nlattr *tb[RTA_MAX+1];
2372 struct rtable *rt = NULL;
2373 struct flowi4 fl4;
2374 __be32 dst = 0;
2375 __be32 src = 0;
2376 u32 iif;
2377 int err;
2378 int mark;
2379 struct sk_buff *skb;
2380
2381 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv4_policy);
2382 if (err < 0)
2383 goto errout;
2384
2385 rtm = nlmsg_data(nlh);
2386
2387 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2388 if (skb == NULL) {
2389 err = -ENOBUFS;
2390 goto errout;
2391 }
2392
2393 /* Reserve room for dummy headers, this skb can pass
2394 through good chunk of routing engine.
2395 */
2396 skb_reset_mac_header(skb);
2397 skb_reset_network_header(skb);
2398
2399 /* Bugfix: need to give ip_route_input enough of an IP header to not gag. */
2400 ip_hdr(skb)->protocol = IPPROTO_ICMP;
2401 skb_reserve(skb, MAX_HEADER + sizeof(struct iphdr));
2402
2403 src = tb[RTA_SRC] ? nla_get_be32(tb[RTA_SRC]) : 0;
2404 dst = tb[RTA_DST] ? nla_get_be32(tb[RTA_DST]) : 0;
2405 iif = tb[RTA_IIF] ? nla_get_u32(tb[RTA_IIF]) : 0;
2406 mark = tb[RTA_MARK] ? nla_get_u32(tb[RTA_MARK]) : 0;
2407
2408 memset(&fl4, 0, sizeof(fl4));
2409 fl4.daddr = dst;
2410 fl4.saddr = src;
2411 fl4.flowi4_tos = rtm->rtm_tos;
2412 fl4.flowi4_oif = tb[RTA_OIF] ? nla_get_u32(tb[RTA_OIF]) : 0;
2413 fl4.flowi4_mark = mark;
2414
2415 if (iif) {
2416 struct net_device *dev;
2417
2418 dev = __dev_get_by_index(net, iif);
2419 if (dev == NULL) {
2420 err = -ENODEV;
2421 goto errout_free;
2422 }
2423
2424 skb->protocol = htons(ETH_P_IP);
2425 skb->dev = dev;
2426 skb->mark = mark;
2427 local_bh_disable();
2428 err = ip_route_input(skb, dst, src, rtm->rtm_tos, dev);
2429 local_bh_enable();
2430
2431 rt = skb_rtable(skb);
2432 if (err == 0 && rt->dst.error)
2433 err = -rt->dst.error;
2434 } else {
2435 rt = ip_route_output_key(net, &fl4);
2436
2437 err = 0;
2438 if (IS_ERR(rt))
2439 err = PTR_ERR(rt);
2440 }
2441
2442 if (err)
2443 goto errout_free;
2444
2445 skb_dst_set(skb, &rt->dst);
2446 if (rtm->rtm_flags & RTM_F_NOTIFY)
2447 rt->rt_flags |= RTCF_NOTIFY;
2448
2449 err = rt_fill_info(net, dst, src, &fl4, skb,
2450 NETLINK_CB(in_skb).portid, nlh->nlmsg_seq,
2451 RTM_NEWROUTE, 0, 0);
2452 if (err <= 0)
2453 goto errout_free;
2454
2455 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2456 errout:
2457 return err;
2458
2459 errout_free:
2460 kfree_skb(skb);
2461 goto errout;
2462 }
2463
2464 int ip_rt_dump(struct sk_buff *skb, struct netlink_callback *cb)
2465 {
2466 return skb->len;
2467 }
2468
2469 void ip_rt_multicast_event(struct in_device *in_dev)
2470 {
2471 rt_cache_flush(dev_net(in_dev->dev));
2472 }
2473
2474 #ifdef CONFIG_SYSCTL
2475 static int ip_rt_gc_timeout __read_mostly = RT_GC_TIMEOUT;
2476 static int ip_rt_gc_interval __read_mostly = 60 * HZ;
2477 static int ip_rt_gc_min_interval __read_mostly = HZ / 2;
2478 static int ip_rt_gc_elasticity __read_mostly = 8;
2479
2480 static int ipv4_sysctl_rtcache_flush(ctl_table *__ctl, int write,
2481 void __user *buffer,
2482 size_t *lenp, loff_t *ppos)
2483 {
2484 if (write) {
2485 rt_cache_flush((struct net *)__ctl->extra1);
2486 return 0;
2487 }
2488
2489 return -EINVAL;
2490 }
2491
2492 static ctl_table ipv4_route_table[] = {
2493 {
2494 .procname = "gc_thresh",
2495 .data = &ipv4_dst_ops.gc_thresh,
2496 .maxlen = sizeof(int),
2497 .mode = 0644,
2498 .proc_handler = proc_dointvec,
2499 },
2500 {
2501 .procname = "max_size",
2502 .data = &ip_rt_max_size,
2503 .maxlen = sizeof(int),
2504 .mode = 0644,
2505 .proc_handler = proc_dointvec,
2506 },
2507 {
2508 /* Deprecated. Use gc_min_interval_ms */
2509
2510 .procname = "gc_min_interval",
2511 .data = &ip_rt_gc_min_interval,
2512 .maxlen = sizeof(int),
2513 .mode = 0644,
2514 .proc_handler = proc_dointvec_jiffies,
2515 },
2516 {
2517 .procname = "gc_min_interval_ms",
2518 .data = &ip_rt_gc_min_interval,
2519 .maxlen = sizeof(int),
2520 .mode = 0644,
2521 .proc_handler = proc_dointvec_ms_jiffies,
2522 },
2523 {
2524 .procname = "gc_timeout",
2525 .data = &ip_rt_gc_timeout,
2526 .maxlen = sizeof(int),
2527 .mode = 0644,
2528 .proc_handler = proc_dointvec_jiffies,
2529 },
2530 {
2531 .procname = "gc_interval",
2532 .data = &ip_rt_gc_interval,
2533 .maxlen = sizeof(int),
2534 .mode = 0644,
2535 .proc_handler = proc_dointvec_jiffies,
2536 },
2537 {
2538 .procname = "redirect_load",
2539 .data = &ip_rt_redirect_load,
2540 .maxlen = sizeof(int),
2541 .mode = 0644,
2542 .proc_handler = proc_dointvec,
2543 },
2544 {
2545 .procname = "redirect_number",
2546 .data = &ip_rt_redirect_number,
2547 .maxlen = sizeof(int),
2548 .mode = 0644,
2549 .proc_handler = proc_dointvec,
2550 },
2551 {
2552 .procname = "redirect_silence",
2553 .data = &ip_rt_redirect_silence,
2554 .maxlen = sizeof(int),
2555 .mode = 0644,
2556 .proc_handler = proc_dointvec,
2557 },
2558 {
2559 .procname = "error_cost",
2560 .data = &ip_rt_error_cost,
2561 .maxlen = sizeof(int),
2562 .mode = 0644,
2563 .proc_handler = proc_dointvec,
2564 },
2565 {
2566 .procname = "error_burst",
2567 .data = &ip_rt_error_burst,
2568 .maxlen = sizeof(int),
2569 .mode = 0644,
2570 .proc_handler = proc_dointvec,
2571 },
2572 {
2573 .procname = "gc_elasticity",
2574 .data = &ip_rt_gc_elasticity,
2575 .maxlen = sizeof(int),
2576 .mode = 0644,
2577 .proc_handler = proc_dointvec,
2578 },
2579 {
2580 .procname = "mtu_expires",
2581 .data = &ip_rt_mtu_expires,
2582 .maxlen = sizeof(int),
2583 .mode = 0644,
2584 .proc_handler = proc_dointvec_jiffies,
2585 },
2586 {
2587 .procname = "min_pmtu",
2588 .data = &ip_rt_min_pmtu,
2589 .maxlen = sizeof(int),
2590 .mode = 0644,
2591 .proc_handler = proc_dointvec,
2592 },
2593 {
2594 .procname = "min_adv_mss",
2595 .data = &ip_rt_min_advmss,
2596 .maxlen = sizeof(int),
2597 .mode = 0644,
2598 .proc_handler = proc_dointvec,
2599 },
2600 { }
2601 };
2602
2603 static struct ctl_table ipv4_route_flush_table[] = {
2604 {
2605 .procname = "flush",
2606 .maxlen = sizeof(int),
2607 .mode = 0200,
2608 .proc_handler = ipv4_sysctl_rtcache_flush,
2609 },
2610 { },
2611 };
2612
2613 static __net_init int sysctl_route_net_init(struct net *net)
2614 {
2615 struct ctl_table *tbl;
2616
2617 tbl = ipv4_route_flush_table;
2618 if (!net_eq(net, &init_net)) {
2619 tbl = kmemdup(tbl, sizeof(ipv4_route_flush_table), GFP_KERNEL);
2620 if (tbl == NULL)
2621 goto err_dup;
2622
2623 /* Don't export sysctls to unprivileged users */
2624 if (net->user_ns != &init_user_ns)
2625 tbl[0].procname = NULL;
2626 }
2627 tbl[0].extra1 = net;
2628
2629 net->ipv4.route_hdr = register_net_sysctl(net, "net/ipv4/route", tbl);
2630 if (net->ipv4.route_hdr == NULL)
2631 goto err_reg;
2632 return 0;
2633
2634 err_reg:
2635 if (tbl != ipv4_route_flush_table)
2636 kfree(tbl);
2637 err_dup:
2638 return -ENOMEM;
2639 }
2640
2641 static __net_exit void sysctl_route_net_exit(struct net *net)
2642 {
2643 struct ctl_table *tbl;
2644
2645 tbl = net->ipv4.route_hdr->ctl_table_arg;
2646 unregister_net_sysctl_table(net->ipv4.route_hdr);
2647 BUG_ON(tbl == ipv4_route_flush_table);
2648 kfree(tbl);
2649 }
2650
2651 static __net_initdata struct pernet_operations sysctl_route_ops = {
2652 .init = sysctl_route_net_init,
2653 .exit = sysctl_route_net_exit,
2654 };
2655 #endif
2656
2657 static __net_init int rt_genid_init(struct net *net)
2658 {
2659 atomic_set(&net->rt_genid, 0);
2660 get_random_bytes(&net->ipv4.dev_addr_genid,
2661 sizeof(net->ipv4.dev_addr_genid));
2662 return 0;
2663 }
2664
2665 static __net_initdata struct pernet_operations rt_genid_ops = {
2666 .init = rt_genid_init,
2667 };
2668
2669 static int __net_init ipv4_inetpeer_init(struct net *net)
2670 {
2671 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
2672
2673 if (!bp)
2674 return -ENOMEM;
2675 inet_peer_base_init(bp);
2676 net->ipv4.peers = bp;
2677 return 0;
2678 }
2679
2680 static void __net_exit ipv4_inetpeer_exit(struct net *net)
2681 {
2682 struct inet_peer_base *bp = net->ipv4.peers;
2683
2684 net->ipv4.peers = NULL;
2685 inetpeer_invalidate_tree(bp);
2686 kfree(bp);
2687 }
2688
2689 static __net_initdata struct pernet_operations ipv4_inetpeer_ops = {
2690 .init = ipv4_inetpeer_init,
2691 .exit = ipv4_inetpeer_exit,
2692 };
2693
2694 #ifdef CONFIG_IP_ROUTE_CLASSID
2695 struct ip_rt_acct __percpu *ip_rt_acct __read_mostly;
2696 #endif /* CONFIG_IP_ROUTE_CLASSID */
2697
2698 int __init ip_rt_init(void)
2699 {
2700 int rc = 0;
2701
2702 ip_idents = kmalloc(IP_IDENTS_SZ * sizeof(*ip_idents), GFP_KERNEL);
2703 if (!ip_idents)
2704 panic("IP: failed to allocate ip_idents\n");
2705
2706 prandom_bytes(ip_idents, IP_IDENTS_SZ * sizeof(*ip_idents));
2707
2708 #ifdef CONFIG_IP_ROUTE_CLASSID
2709 ip_rt_acct = __alloc_percpu(256 * sizeof(struct ip_rt_acct), __alignof__(struct ip_rt_acct));
2710 if (!ip_rt_acct)
2711 panic("IP: failed to allocate ip_rt_acct\n");
2712 #endif
2713
2714 ipv4_dst_ops.kmem_cachep =
2715 kmem_cache_create("ip_dst_cache", sizeof(struct rtable), 0,
2716 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
2717
2718 ipv4_dst_blackhole_ops.kmem_cachep = ipv4_dst_ops.kmem_cachep;
2719
2720 if (dst_entries_init(&ipv4_dst_ops) < 0)
2721 panic("IP: failed to allocate ipv4_dst_ops counter\n");
2722
2723 if (dst_entries_init(&ipv4_dst_blackhole_ops) < 0)
2724 panic("IP: failed to allocate ipv4_dst_blackhole_ops counter\n");
2725
2726 ipv4_dst_ops.gc_thresh = ~0;
2727 ip_rt_max_size = INT_MAX;
2728
2729 devinet_init();
2730 ip_fib_init();
2731
2732 if (ip_rt_proc_init())
2733 pr_err("Unable to create route proc files\n");
2734 #ifdef CONFIG_XFRM
2735 xfrm_init();
2736 xfrm4_init();
2737 #endif
2738 rtnl_register(PF_INET, RTM_GETROUTE, inet_rtm_getroute, NULL, NULL);
2739
2740 #ifdef CONFIG_SYSCTL
2741 register_pernet_subsys(&sysctl_route_ops);
2742 #endif
2743 register_pernet_subsys(&rt_genid_ops);
2744 register_pernet_subsys(&ipv4_inetpeer_ops);
2745 return rc;
2746 }
2747
2748 #ifdef CONFIG_SYSCTL
2749 /*
2750 * We really need to sanitize the damn ipv4 init order, then all
2751 * this nonsense will go away.
2752 */
2753 void __init ip_static_sysctl_init(void)
2754 {
2755 register_net_sysctl(&init_net, "net/ipv4/route", ipv4_route_table);
2756 }
2757 #endif
kernel source for telekom puls (alcatel/mediatek)