This source file includes following definitions.
- ipv4_cow_metrics
- rt_cache_seq_start
- rt_cache_seq_next
- rt_cache_seq_stop
- rt_cache_seq_show
- rt_cache_seq_open
- rt_cpu_seq_start
- rt_cpu_seq_next
- rt_cpu_seq_stop
- rt_cpu_seq_show
- rt_cpu_seq_open
- rt_acct_proc_show
- ip_rt_do_proc_init
- ip_rt_do_proc_exit
- ip_rt_proc_init
- ip_rt_proc_init
- rt_is_expired
- rt_cache_flush
- ipv4_neigh_lookup
- ipv4_confirm_neigh
- ip_idents_reserve
- __ip_select_ident
- __build_flow_key
- build_skb_flow_key
- build_sk_flow_key
- ip_rt_build_flow_key
- fnhe_flush_routes
- fnhe_oldest
- fnhe_hashfun
- fill_route_from_fnhe
- update_or_create_fnhe
- __ip_do_redirect
- ip_do_redirect
- ipv4_negative_advice
- ip_rt_send_redirect
- ip_error
- __ip_rt_update_pmtu
- ip_rt_update_pmtu
- ipv4_update_pmtu
- __ipv4_sk_update_pmtu
- ipv4_sk_update_pmtu
- ipv4_redirect
- ipv4_sk_redirect
- ipv4_dst_check
- ipv4_send_dest_unreach
- ipv4_link_failure
- ip_rt_bug
- ip_rt_get_source
- set_class_tag
- ipv4_default_advmss
- ipv4_mtu
- ip_del_fnhe
- find_exception
- ip_mtu_from_fib_result
- rt_bind_exception
- rt_cache_route
- rt_add_uncached_list
- rt_del_uncached_list
- ipv4_dst_destroy
- rt_flush_dev
- rt_cache_valid
- rt_set_nexthop
- rt_dst_alloc
- rt_dst_clone
- ip_mc_validate_source
- ip_route_input_mc
- ip_handle_martian_source
- __mkroute_input
- ip_multipath_l3_keys
- fib_multipath_hash
- ip_mkroute_input
- ip_route_input_slow
- ip_route_input_noref
- ip_route_input_rcu
- __mkroute_output
- ip_route_output_key_hash
- ip_route_output_key_hash_rcu
- ipv4_blackhole_dst_check
- ipv4_blackhole_mtu
- ipv4_rt_blackhole_update_pmtu
- ipv4_rt_blackhole_redirect
- ipv4_rt_blackhole_cow_metrics
- ipv4_blackhole_route
- ip_route_output_flow
- rt_fill_info
- fnhe_dump_bucket
- fib_dump_info_fnhe
- inet_rtm_getroute_build_skb
- inet_rtm_valid_getroute_req
- inet_rtm_getroute
- ip_rt_multicast_event
- ipv4_sysctl_rtcache_flush
- sysctl_route_net_init
- sysctl_route_net_exit
- rt_genid_init
- ipv4_inetpeer_init
- ipv4_inetpeer_exit
- ip_rt_init
- ip_static_sysctl_init
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61 #define pr_fmt(fmt) "IPv4: " fmt
62
63 #include <linux/module.h>
64 #include <linux/uaccess.h>
65 #include <linux/bitops.h>
66 #include <linux/types.h>
67 #include <linux/kernel.h>
68 #include <linux/mm.h>
69 #include <linux/string.h>
70 #include <linux/socket.h>
71 #include <linux/sockios.h>
72 #include <linux/errno.h>
73 #include <linux/in.h>
74 #include <linux/inet.h>
75 #include <linux/netdevice.h>
76 #include <linux/proc_fs.h>
77 #include <linux/init.h>
78 #include <linux/skbuff.h>
79 #include <linux/inetdevice.h>
80 #include <linux/igmp.h>
81 #include <linux/pkt_sched.h>
82 #include <linux/mroute.h>
83 #include <linux/netfilter_ipv4.h>
84 #include <linux/random.h>
85 #include <linux/rcupdate.h>
86 #include <linux/times.h>
87 #include <linux/slab.h>
88 #include <linux/jhash.h>
89 #include <net/dst.h>
90 #include <net/dst_metadata.h>
91 #include <net/net_namespace.h>
92 #include <net/protocol.h>
93 #include <net/ip.h>
94 #include <net/route.h>
95 #include <net/inetpeer.h>
96 #include <net/sock.h>
97 #include <net/ip_fib.h>
98 #include <net/nexthop.h>
99 #include <net/arp.h>
100 #include <net/tcp.h>
101 #include <net/icmp.h>
102 #include <net/xfrm.h>
103 #include <net/lwtunnel.h>
104 #include <net/netevent.h>
105 #include <net/rtnetlink.h>
106 #ifdef CONFIG_SYSCTL
107 #include <linux/sysctl.h>
108 #endif
109 #include <net/secure_seq.h>
110 #include <net/ip_tunnels.h>
111 #include <net/l3mdev.h>
112
113 #include "fib_lookup.h"
114
115 #define RT_FL_TOS(oldflp4) \
116 ((oldflp4)->flowi4_tos & (IPTOS_RT_MASK | RTO_ONLINK))
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 u32 ip_rt_min_pmtu __read_mostly = 512 + 20 + 20;
128 static int ip_rt_min_advmss __read_mostly = 256;
129
130 static int ip_rt_gc_timeout __read_mostly = RT_GC_TIMEOUT;
131
132
133
134
135
136 static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie);
137 static unsigned int ipv4_default_advmss(const struct dst_entry *dst);
138 static unsigned int ipv4_mtu(const struct dst_entry *dst);
139 static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst);
140 static void ipv4_link_failure(struct sk_buff *skb);
141 static void ip_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
142 struct sk_buff *skb, u32 mtu,
143 bool confirm_neigh);
144 static void ip_do_redirect(struct dst_entry *dst, struct sock *sk,
145 struct sk_buff *skb);
146 static void ipv4_dst_destroy(struct dst_entry *dst);
147
148 static u32 *ipv4_cow_metrics(struct dst_entry *dst, unsigned long old)
149 {
150 WARN_ON(1);
151 return NULL;
152 }
153
154 static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst,
155 struct sk_buff *skb,
156 const void *daddr);
157 static void ipv4_confirm_neigh(const struct dst_entry *dst, const void *daddr);
158
159 static struct dst_ops ipv4_dst_ops = {
160 .family = AF_INET,
161 .check = ipv4_dst_check,
162 .default_advmss = ipv4_default_advmss,
163 .mtu = ipv4_mtu,
164 .cow_metrics = ipv4_cow_metrics,
165 .destroy = ipv4_dst_destroy,
166 .negative_advice = ipv4_negative_advice,
167 .link_failure = ipv4_link_failure,
168 .update_pmtu = ip_rt_update_pmtu,
169 .redirect = ip_do_redirect,
170 .local_out = __ip_local_out,
171 .neigh_lookup = ipv4_neigh_lookup,
172 .confirm_neigh = ipv4_confirm_neigh,
173 };
174
175 #define ECN_OR_COST(class) TC_PRIO_##class
176
177 const __u8 ip_tos2prio[16] = {
178 TC_PRIO_BESTEFFORT,
179 ECN_OR_COST(BESTEFFORT),
180 TC_PRIO_BESTEFFORT,
181 ECN_OR_COST(BESTEFFORT),
182 TC_PRIO_BULK,
183 ECN_OR_COST(BULK),
184 TC_PRIO_BULK,
185 ECN_OR_COST(BULK),
186 TC_PRIO_INTERACTIVE,
187 ECN_OR_COST(INTERACTIVE),
188 TC_PRIO_INTERACTIVE,
189 ECN_OR_COST(INTERACTIVE),
190 TC_PRIO_INTERACTIVE_BULK,
191 ECN_OR_COST(INTERACTIVE_BULK),
192 TC_PRIO_INTERACTIVE_BULK,
193 ECN_OR_COST(INTERACTIVE_BULK)
194 };
195 EXPORT_SYMBOL(ip_tos2prio);
196
197 static DEFINE_PER_CPU(struct rt_cache_stat, rt_cache_stat);
198 #define RT_CACHE_STAT_INC(field) raw_cpu_inc(rt_cache_stat.field)
199
200 #ifdef CONFIG_PROC_FS
201 static void *rt_cache_seq_start(struct seq_file *seq, loff_t *pos)
202 {
203 if (*pos)
204 return NULL;
205 return SEQ_START_TOKEN;
206 }
207
208 static void *rt_cache_seq_next(struct seq_file *seq, void *v, loff_t *pos)
209 {
210 ++*pos;
211 return NULL;
212 }
213
214 static void rt_cache_seq_stop(struct seq_file *seq, void *v)
215 {
216 }
217
218 static int rt_cache_seq_show(struct seq_file *seq, void *v)
219 {
220 if (v == SEQ_START_TOKEN)
221 seq_printf(seq, "%-127s\n",
222 "Iface\tDestination\tGateway \tFlags\t\tRefCnt\tUse\t"
223 "Metric\tSource\t\tMTU\tWindow\tIRTT\tTOS\tHHRef\t"
224 "HHUptod\tSpecDst");
225 return 0;
226 }
227
228 static const struct seq_operations rt_cache_seq_ops = {
229 .start = rt_cache_seq_start,
230 .next = rt_cache_seq_next,
231 .stop = rt_cache_seq_stop,
232 .show = rt_cache_seq_show,
233 };
234
235 static int rt_cache_seq_open(struct inode *inode, struct file *file)
236 {
237 return seq_open(file, &rt_cache_seq_ops);
238 }
239
240 static const struct file_operations rt_cache_seq_fops = {
241 .open = rt_cache_seq_open,
242 .read = seq_read,
243 .llseek = seq_lseek,
244 .release = seq_release,
245 };
246
247
248 static void *rt_cpu_seq_start(struct seq_file *seq, loff_t *pos)
249 {
250 int cpu;
251
252 if (*pos == 0)
253 return SEQ_START_TOKEN;
254
255 for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
256 if (!cpu_possible(cpu))
257 continue;
258 *pos = cpu+1;
259 return &per_cpu(rt_cache_stat, cpu);
260 }
261 return NULL;
262 }
263
264 static void *rt_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
265 {
266 int cpu;
267
268 for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
269 if (!cpu_possible(cpu))
270 continue;
271 *pos = cpu+1;
272 return &per_cpu(rt_cache_stat, cpu);
273 }
274 return NULL;
275
276 }
277
278 static void rt_cpu_seq_stop(struct seq_file *seq, void *v)
279 {
280
281 }
282
283 static int rt_cpu_seq_show(struct seq_file *seq, void *v)
284 {
285 struct rt_cache_stat *st = v;
286
287 if (v == SEQ_START_TOKEN) {
288 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");
289 return 0;
290 }
291
292 seq_printf(seq,"%08x %08x %08x %08x %08x %08x %08x %08x "
293 " %08x %08x %08x %08x %08x %08x %08x %08x %08x \n",
294 dst_entries_get_slow(&ipv4_dst_ops),
295 0,
296 st->in_slow_tot,
297 st->in_slow_mc,
298 st->in_no_route,
299 st->in_brd,
300 st->in_martian_dst,
301 st->in_martian_src,
302
303 0,
304 st->out_slow_tot,
305 st->out_slow_mc,
306
307 0,
308 0,
309 0,
310 0,
311 0,
312 0
313 );
314 return 0;
315 }
316
317 static const struct seq_operations rt_cpu_seq_ops = {
318 .start = rt_cpu_seq_start,
319 .next = rt_cpu_seq_next,
320 .stop = rt_cpu_seq_stop,
321 .show = rt_cpu_seq_show,
322 };
323
324
325 static int rt_cpu_seq_open(struct inode *inode, struct file *file)
326 {
327 return seq_open(file, &rt_cpu_seq_ops);
328 }
329
330 static const struct file_operations rt_cpu_seq_fops = {
331 .open = rt_cpu_seq_open,
332 .read = seq_read,
333 .llseek = seq_lseek,
334 .release = seq_release,
335 };
336
337 #ifdef CONFIG_IP_ROUTE_CLASSID
338 static int rt_acct_proc_show(struct seq_file *m, void *v)
339 {
340 struct ip_rt_acct *dst, *src;
341 unsigned int i, j;
342
343 dst = kcalloc(256, sizeof(struct ip_rt_acct), GFP_KERNEL);
344 if (!dst)
345 return -ENOMEM;
346
347 for_each_possible_cpu(i) {
348 src = (struct ip_rt_acct *)per_cpu_ptr(ip_rt_acct, i);
349 for (j = 0; j < 256; j++) {
350 dst[j].o_bytes += src[j].o_bytes;
351 dst[j].o_packets += src[j].o_packets;
352 dst[j].i_bytes += src[j].i_bytes;
353 dst[j].i_packets += src[j].i_packets;
354 }
355 }
356
357 seq_write(m, dst, 256 * sizeof(struct ip_rt_acct));
358 kfree(dst);
359 return 0;
360 }
361 #endif
362
363 static int __net_init ip_rt_do_proc_init(struct net *net)
364 {
365 struct proc_dir_entry *pde;
366
367 pde = proc_create("rt_cache", 0444, net->proc_net,
368 &rt_cache_seq_fops);
369 if (!pde)
370 goto err1;
371
372 pde = proc_create("rt_cache", 0444,
373 net->proc_net_stat, &rt_cpu_seq_fops);
374 if (!pde)
375 goto err2;
376
377 #ifdef CONFIG_IP_ROUTE_CLASSID
378 pde = proc_create_single("rt_acct", 0, net->proc_net,
379 rt_acct_proc_show);
380 if (!pde)
381 goto err3;
382 #endif
383 return 0;
384
385 #ifdef CONFIG_IP_ROUTE_CLASSID
386 err3:
387 remove_proc_entry("rt_cache", net->proc_net_stat);
388 #endif
389 err2:
390 remove_proc_entry("rt_cache", net->proc_net);
391 err1:
392 return -ENOMEM;
393 }
394
395 static void __net_exit ip_rt_do_proc_exit(struct net *net)
396 {
397 remove_proc_entry("rt_cache", net->proc_net_stat);
398 remove_proc_entry("rt_cache", net->proc_net);
399 #ifdef CONFIG_IP_ROUTE_CLASSID
400 remove_proc_entry("rt_acct", net->proc_net);
401 #endif
402 }
403
404 static struct pernet_operations ip_rt_proc_ops __net_initdata = {
405 .init = ip_rt_do_proc_init,
406 .exit = ip_rt_do_proc_exit,
407 };
408
409 static int __init ip_rt_proc_init(void)
410 {
411 return register_pernet_subsys(&ip_rt_proc_ops);
412 }
413
414 #else
415 static inline int ip_rt_proc_init(void)
416 {
417 return 0;
418 }
419 #endif
420
421 static inline bool rt_is_expired(const struct rtable *rth)
422 {
423 return rth->rt_genid != rt_genid_ipv4(dev_net(rth->dst.dev));
424 }
425
426 void rt_cache_flush(struct net *net)
427 {
428 rt_genid_bump_ipv4(net);
429 }
430
431 static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst,
432 struct sk_buff *skb,
433 const void *daddr)
434 {
435 const struct rtable *rt = container_of(dst, struct rtable, dst);
436 struct net_device *dev = dst->dev;
437 struct neighbour *n;
438
439 rcu_read_lock_bh();
440
441 if (likely(rt->rt_gw_family == AF_INET)) {
442 n = ip_neigh_gw4(dev, rt->rt_gw4);
443 } else if (rt->rt_gw_family == AF_INET6) {
444 n = ip_neigh_gw6(dev, &rt->rt_gw6);
445 } else {
446 __be32 pkey;
447
448 pkey = skb ? ip_hdr(skb)->daddr : *((__be32 *) daddr);
449 n = ip_neigh_gw4(dev, pkey);
450 }
451
452 if (!IS_ERR(n) && !refcount_inc_not_zero(&n->refcnt))
453 n = NULL;
454
455 rcu_read_unlock_bh();
456
457 return n;
458 }
459
460 static void ipv4_confirm_neigh(const struct dst_entry *dst, const void *daddr)
461 {
462 const struct rtable *rt = container_of(dst, struct rtable, dst);
463 struct net_device *dev = dst->dev;
464 const __be32 *pkey = daddr;
465
466 if (rt->rt_gw_family == AF_INET) {
467 pkey = (const __be32 *)&rt->rt_gw4;
468 } else if (rt->rt_gw_family == AF_INET6) {
469 return __ipv6_confirm_neigh_stub(dev, &rt->rt_gw6);
470 } else if (!daddr ||
471 (rt->rt_flags &
472 (RTCF_MULTICAST | RTCF_BROADCAST | RTCF_LOCAL))) {
473 return;
474 }
475 __ipv4_confirm_neigh(dev, *(__force u32 *)pkey);
476 }
477
478 #define IP_IDENTS_SZ 2048u
479
480 static atomic_t *ip_idents __read_mostly;
481 static u32 *ip_tstamps __read_mostly;
482
483
484
485
486
487 u32 ip_idents_reserve(u32 hash, int segs)
488 {
489 u32 *p_tstamp = ip_tstamps + hash % IP_IDENTS_SZ;
490 atomic_t *p_id = ip_idents + hash % IP_IDENTS_SZ;
491 u32 old = READ_ONCE(*p_tstamp);
492 u32 now = (u32)jiffies;
493 u32 delta = 0;
494
495 if (old != now && cmpxchg(p_tstamp, old, now) == old)
496 delta = prandom_u32_max(now - old);
497
498
499
500
501
502 return atomic_add_return(segs + delta, p_id) - segs;
503 }
504 EXPORT_SYMBOL(ip_idents_reserve);
505
506 void __ip_select_ident(struct net *net, struct iphdr *iph, int segs)
507 {
508 u32 hash, id;
509
510
511 if (unlikely(siphash_key_is_zero(&net->ipv4.ip_id_key)))
512 get_random_bytes(&net->ipv4.ip_id_key,
513 sizeof(net->ipv4.ip_id_key));
514
515 hash = siphash_3u32((__force u32)iph->daddr,
516 (__force u32)iph->saddr,
517 iph->protocol,
518 &net->ipv4.ip_id_key);
519 id = ip_idents_reserve(hash, segs);
520 iph->id = htons(id);
521 }
522 EXPORT_SYMBOL(__ip_select_ident);
523
524 static void __build_flow_key(const struct net *net, struct flowi4 *fl4,
525 const struct sock *sk,
526 const struct iphdr *iph,
527 int oif, u8 tos,
528 u8 prot, u32 mark, int flow_flags)
529 {
530 if (sk) {
531 const struct inet_sock *inet = inet_sk(sk);
532
533 oif = sk->sk_bound_dev_if;
534 mark = sk->sk_mark;
535 tos = RT_CONN_FLAGS(sk);
536 prot = inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol;
537 }
538 flowi4_init_output(fl4, oif, mark, tos,
539 RT_SCOPE_UNIVERSE, prot,
540 flow_flags,
541 iph->daddr, iph->saddr, 0, 0,
542 sock_net_uid(net, sk));
543 }
544
545 static void build_skb_flow_key(struct flowi4 *fl4, const struct sk_buff *skb,
546 const struct sock *sk)
547 {
548 const struct net *net = dev_net(skb->dev);
549 const struct iphdr *iph = ip_hdr(skb);
550 int oif = skb->dev->ifindex;
551 u8 tos = RT_TOS(iph->tos);
552 u8 prot = iph->protocol;
553 u32 mark = skb->mark;
554
555 __build_flow_key(net, fl4, sk, iph, oif, tos, prot, mark, 0);
556 }
557
558 static void build_sk_flow_key(struct flowi4 *fl4, const struct sock *sk)
559 {
560 const struct inet_sock *inet = inet_sk(sk);
561 const struct ip_options_rcu *inet_opt;
562 __be32 daddr = inet->inet_daddr;
563
564 rcu_read_lock();
565 inet_opt = rcu_dereference(inet->inet_opt);
566 if (inet_opt && inet_opt->opt.srr)
567 daddr = inet_opt->opt.faddr;
568 flowi4_init_output(fl4, sk->sk_bound_dev_if, sk->sk_mark,
569 RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE,
570 inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol,
571 inet_sk_flowi_flags(sk),
572 daddr, inet->inet_saddr, 0, 0, sk->sk_uid);
573 rcu_read_unlock();
574 }
575
576 static void ip_rt_build_flow_key(struct flowi4 *fl4, const struct sock *sk,
577 const struct sk_buff *skb)
578 {
579 if (skb)
580 build_skb_flow_key(fl4, skb, sk);
581 else
582 build_sk_flow_key(fl4, sk);
583 }
584
585 static DEFINE_SPINLOCK(fnhe_lock);
586
587 static void fnhe_flush_routes(struct fib_nh_exception *fnhe)
588 {
589 struct rtable *rt;
590
591 rt = rcu_dereference(fnhe->fnhe_rth_input);
592 if (rt) {
593 RCU_INIT_POINTER(fnhe->fnhe_rth_input, NULL);
594 dst_dev_put(&rt->dst);
595 dst_release(&rt->dst);
596 }
597 rt = rcu_dereference(fnhe->fnhe_rth_output);
598 if (rt) {
599 RCU_INIT_POINTER(fnhe->fnhe_rth_output, NULL);
600 dst_dev_put(&rt->dst);
601 dst_release(&rt->dst);
602 }
603 }
604
605 static struct fib_nh_exception *fnhe_oldest(struct fnhe_hash_bucket *hash)
606 {
607 struct fib_nh_exception *fnhe, *oldest;
608
609 oldest = rcu_dereference(hash->chain);
610 for (fnhe = rcu_dereference(oldest->fnhe_next); fnhe;
611 fnhe = rcu_dereference(fnhe->fnhe_next)) {
612 if (time_before(fnhe->fnhe_stamp, oldest->fnhe_stamp))
613 oldest = fnhe;
614 }
615 fnhe_flush_routes(oldest);
616 return oldest;
617 }
618
619 static inline u32 fnhe_hashfun(__be32 daddr)
620 {
621 static u32 fnhe_hashrnd __read_mostly;
622 u32 hval;
623
624 net_get_random_once(&fnhe_hashrnd, sizeof(fnhe_hashrnd));
625 hval = jhash_1word((__force u32) daddr, fnhe_hashrnd);
626 return hash_32(hval, FNHE_HASH_SHIFT);
627 }
628
629 static void fill_route_from_fnhe(struct rtable *rt, struct fib_nh_exception *fnhe)
630 {
631 rt->rt_pmtu = fnhe->fnhe_pmtu;
632 rt->rt_mtu_locked = fnhe->fnhe_mtu_locked;
633 rt->dst.expires = fnhe->fnhe_expires;
634
635 if (fnhe->fnhe_gw) {
636 rt->rt_flags |= RTCF_REDIRECTED;
637 rt->rt_uses_gateway = 1;
638 rt->rt_gw_family = AF_INET;
639 rt->rt_gw4 = fnhe->fnhe_gw;
640 }
641 }
642
643 static void update_or_create_fnhe(struct fib_nh_common *nhc, __be32 daddr,
644 __be32 gw, u32 pmtu, bool lock,
645 unsigned long expires)
646 {
647 struct fnhe_hash_bucket *hash;
648 struct fib_nh_exception *fnhe;
649 struct rtable *rt;
650 u32 genid, hval;
651 unsigned int i;
652 int depth;
653
654 genid = fnhe_genid(dev_net(nhc->nhc_dev));
655 hval = fnhe_hashfun(daddr);
656
657 spin_lock_bh(&fnhe_lock);
658
659 hash = rcu_dereference(nhc->nhc_exceptions);
660 if (!hash) {
661 hash = kcalloc(FNHE_HASH_SIZE, sizeof(*hash), GFP_ATOMIC);
662 if (!hash)
663 goto out_unlock;
664 rcu_assign_pointer(nhc->nhc_exceptions, hash);
665 }
666
667 hash += hval;
668
669 depth = 0;
670 for (fnhe = rcu_dereference(hash->chain); fnhe;
671 fnhe = rcu_dereference(fnhe->fnhe_next)) {
672 if (fnhe->fnhe_daddr == daddr)
673 break;
674 depth++;
675 }
676
677 if (fnhe) {
678 if (fnhe->fnhe_genid != genid)
679 fnhe->fnhe_genid = genid;
680 if (gw)
681 fnhe->fnhe_gw = gw;
682 if (pmtu) {
683 fnhe->fnhe_pmtu = pmtu;
684 fnhe->fnhe_mtu_locked = lock;
685 }
686 fnhe->fnhe_expires = max(1UL, expires);
687
688 rt = rcu_dereference(fnhe->fnhe_rth_input);
689 if (rt)
690 fill_route_from_fnhe(rt, fnhe);
691 rt = rcu_dereference(fnhe->fnhe_rth_output);
692 if (rt)
693 fill_route_from_fnhe(rt, fnhe);
694 } else {
695 if (depth > FNHE_RECLAIM_DEPTH)
696 fnhe = fnhe_oldest(hash);
697 else {
698 fnhe = kzalloc(sizeof(*fnhe), GFP_ATOMIC);
699 if (!fnhe)
700 goto out_unlock;
701
702 fnhe->fnhe_next = hash->chain;
703 rcu_assign_pointer(hash->chain, fnhe);
704 }
705 fnhe->fnhe_genid = genid;
706 fnhe->fnhe_daddr = daddr;
707 fnhe->fnhe_gw = gw;
708 fnhe->fnhe_pmtu = pmtu;
709 fnhe->fnhe_mtu_locked = lock;
710 fnhe->fnhe_expires = max(1UL, expires);
711
712
713
714
715
716 rt = rcu_dereference(nhc->nhc_rth_input);
717 if (rt)
718 rt->dst.obsolete = DST_OBSOLETE_KILL;
719
720 for_each_possible_cpu(i) {
721 struct rtable __rcu **prt;
722 prt = per_cpu_ptr(nhc->nhc_pcpu_rth_output, i);
723 rt = rcu_dereference(*prt);
724 if (rt)
725 rt->dst.obsolete = DST_OBSOLETE_KILL;
726 }
727 }
728
729 fnhe->fnhe_stamp = jiffies;
730
731 out_unlock:
732 spin_unlock_bh(&fnhe_lock);
733 }
734
735 static void __ip_do_redirect(struct rtable *rt, struct sk_buff *skb, struct flowi4 *fl4,
736 bool kill_route)
737 {
738 __be32 new_gw = icmp_hdr(skb)->un.gateway;
739 __be32 old_gw = ip_hdr(skb)->saddr;
740 struct net_device *dev = skb->dev;
741 struct in_device *in_dev;
742 struct fib_result res;
743 struct neighbour *n;
744 struct net *net;
745
746 switch (icmp_hdr(skb)->code & 7) {
747 case ICMP_REDIR_NET:
748 case ICMP_REDIR_NETTOS:
749 case ICMP_REDIR_HOST:
750 case ICMP_REDIR_HOSTTOS:
751 break;
752
753 default:
754 return;
755 }
756
757 if (rt->rt_gw_family != AF_INET || rt->rt_gw4 != old_gw)
758 return;
759
760 in_dev = __in_dev_get_rcu(dev);
761 if (!in_dev)
762 return;
763
764 net = dev_net(dev);
765 if (new_gw == old_gw || !IN_DEV_RX_REDIRECTS(in_dev) ||
766 ipv4_is_multicast(new_gw) || ipv4_is_lbcast(new_gw) ||
767 ipv4_is_zeronet(new_gw))
768 goto reject_redirect;
769
770 if (!IN_DEV_SHARED_MEDIA(in_dev)) {
771 if (!inet_addr_onlink(in_dev, new_gw, old_gw))
772 goto reject_redirect;
773 if (IN_DEV_SEC_REDIRECTS(in_dev) && ip_fib_check_default(new_gw, dev))
774 goto reject_redirect;
775 } else {
776 if (inet_addr_type(net, new_gw) != RTN_UNICAST)
777 goto reject_redirect;
778 }
779
780 n = __ipv4_neigh_lookup(rt->dst.dev, new_gw);
781 if (!n)
782 n = neigh_create(&arp_tbl, &new_gw, rt->dst.dev);
783 if (!IS_ERR(n)) {
784 if (!(n->nud_state & NUD_VALID)) {
785 neigh_event_send(n, NULL);
786 } else {
787 if (fib_lookup(net, fl4, &res, 0) == 0) {
788 struct fib_nh_common *nhc = FIB_RES_NHC(res);
789
790 update_or_create_fnhe(nhc, fl4->daddr, new_gw,
791 0, false,
792 jiffies + ip_rt_gc_timeout);
793 }
794 if (kill_route)
795 rt->dst.obsolete = DST_OBSOLETE_KILL;
796 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, n);
797 }
798 neigh_release(n);
799 }
800 return;
801
802 reject_redirect:
803 #ifdef CONFIG_IP_ROUTE_VERBOSE
804 if (IN_DEV_LOG_MARTIANS(in_dev)) {
805 const struct iphdr *iph = (const struct iphdr *) skb->data;
806 __be32 daddr = iph->daddr;
807 __be32 saddr = iph->saddr;
808
809 net_info_ratelimited("Redirect from %pI4 on %s about %pI4 ignored\n"
810 " Advised path = %pI4 -> %pI4\n",
811 &old_gw, dev->name, &new_gw,
812 &saddr, &daddr);
813 }
814 #endif
815 ;
816 }
817
818 static void ip_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
819 {
820 struct rtable *rt;
821 struct flowi4 fl4;
822 const struct iphdr *iph = (const struct iphdr *) skb->data;
823 struct net *net = dev_net(skb->dev);
824 int oif = skb->dev->ifindex;
825 u8 tos = RT_TOS(iph->tos);
826 u8 prot = iph->protocol;
827 u32 mark = skb->mark;
828
829 rt = (struct rtable *) dst;
830
831 __build_flow_key(net, &fl4, sk, iph, oif, tos, prot, mark, 0);
832 __ip_do_redirect(rt, skb, &fl4, true);
833 }
834
835 static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst)
836 {
837 struct rtable *rt = (struct rtable *)dst;
838 struct dst_entry *ret = dst;
839
840 if (rt) {
841 if (dst->obsolete > 0) {
842 ip_rt_put(rt);
843 ret = NULL;
844 } else if ((rt->rt_flags & RTCF_REDIRECTED) ||
845 rt->dst.expires) {
846 ip_rt_put(rt);
847 ret = NULL;
848 }
849 }
850 return ret;
851 }
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869 void ip_rt_send_redirect(struct sk_buff *skb)
870 {
871 struct rtable *rt = skb_rtable(skb);
872 struct in_device *in_dev;
873 struct inet_peer *peer;
874 struct net *net;
875 int log_martians;
876 int vif;
877
878 rcu_read_lock();
879 in_dev = __in_dev_get_rcu(rt->dst.dev);
880 if (!in_dev || !IN_DEV_TX_REDIRECTS(in_dev)) {
881 rcu_read_unlock();
882 return;
883 }
884 log_martians = IN_DEV_LOG_MARTIANS(in_dev);
885 vif = l3mdev_master_ifindex_rcu(rt->dst.dev);
886 rcu_read_unlock();
887
888 net = dev_net(rt->dst.dev);
889 peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, vif, 1);
890 if (!peer) {
891 icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST,
892 rt_nexthop(rt, ip_hdr(skb)->daddr));
893 return;
894 }
895
896
897
898
899 if (time_after(jiffies, peer->rate_last + ip_rt_redirect_silence)) {
900 peer->rate_tokens = 0;
901 peer->n_redirects = 0;
902 }
903
904
905
906
907 if (peer->n_redirects >= ip_rt_redirect_number) {
908 peer->rate_last = jiffies;
909 goto out_put_peer;
910 }
911
912
913
914
915 if (peer->n_redirects == 0 ||
916 time_after(jiffies,
917 (peer->rate_last +
918 (ip_rt_redirect_load << peer->n_redirects)))) {
919 __be32 gw = rt_nexthop(rt, ip_hdr(skb)->daddr);
920
921 icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, gw);
922 peer->rate_last = jiffies;
923 ++peer->n_redirects;
924 #ifdef CONFIG_IP_ROUTE_VERBOSE
925 if (log_martians &&
926 peer->n_redirects == ip_rt_redirect_number)
927 net_warn_ratelimited("host %pI4/if%d ignores redirects for %pI4 to %pI4\n",
928 &ip_hdr(skb)->saddr, inet_iif(skb),
929 &ip_hdr(skb)->daddr, &gw);
930 #endif
931 }
932 out_put_peer:
933 inet_putpeer(peer);
934 }
935
936 static int ip_error(struct sk_buff *skb)
937 {
938 struct rtable *rt = skb_rtable(skb);
939 struct net_device *dev = skb->dev;
940 struct in_device *in_dev;
941 struct inet_peer *peer;
942 unsigned long now;
943 struct net *net;
944 bool send;
945 int code;
946
947 if (netif_is_l3_master(skb->dev)) {
948 dev = __dev_get_by_index(dev_net(skb->dev), IPCB(skb)->iif);
949 if (!dev)
950 goto out;
951 }
952
953 in_dev = __in_dev_get_rcu(dev);
954
955
956 if (!in_dev)
957 goto out;
958
959 net = dev_net(rt->dst.dev);
960 if (!IN_DEV_FORWARD(in_dev)) {
961 switch (rt->dst.error) {
962 case EHOSTUNREACH:
963 __IP_INC_STATS(net, IPSTATS_MIB_INADDRERRORS);
964 break;
965
966 case ENETUNREACH:
967 __IP_INC_STATS(net, IPSTATS_MIB_INNOROUTES);
968 break;
969 }
970 goto out;
971 }
972
973 switch (rt->dst.error) {
974 case EINVAL:
975 default:
976 goto out;
977 case EHOSTUNREACH:
978 code = ICMP_HOST_UNREACH;
979 break;
980 case ENETUNREACH:
981 code = ICMP_NET_UNREACH;
982 __IP_INC_STATS(net, IPSTATS_MIB_INNOROUTES);
983 break;
984 case EACCES:
985 code = ICMP_PKT_FILTERED;
986 break;
987 }
988
989 peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr,
990 l3mdev_master_ifindex(skb->dev), 1);
991
992 send = true;
993 if (peer) {
994 now = jiffies;
995 peer->rate_tokens += now - peer->rate_last;
996 if (peer->rate_tokens > ip_rt_error_burst)
997 peer->rate_tokens = ip_rt_error_burst;
998 peer->rate_last = now;
999 if (peer->rate_tokens >= ip_rt_error_cost)
1000 peer->rate_tokens -= ip_rt_error_cost;
1001 else
1002 send = false;
1003 inet_putpeer(peer);
1004 }
1005 if (send)
1006 icmp_send(skb, ICMP_DEST_UNREACH, code, 0);
1007
1008 out: kfree_skb(skb);
1009 return 0;
1010 }
1011
1012 static void __ip_rt_update_pmtu(struct rtable *rt, struct flowi4 *fl4, u32 mtu)
1013 {
1014 struct dst_entry *dst = &rt->dst;
1015 u32 old_mtu = ipv4_mtu(dst);
1016 struct fib_result res;
1017 bool lock = false;
1018
1019 if (ip_mtu_locked(dst))
1020 return;
1021
1022 if (old_mtu < mtu)
1023 return;
1024
1025 if (mtu < ip_rt_min_pmtu) {
1026 lock = true;
1027 mtu = min(old_mtu, ip_rt_min_pmtu);
1028 }
1029
1030 if (rt->rt_pmtu == mtu && !lock &&
1031 time_before(jiffies, dst->expires - ip_rt_mtu_expires / 2))
1032 return;
1033
1034 rcu_read_lock();
1035 if (fib_lookup(dev_net(dst->dev), fl4, &res, 0) == 0) {
1036 struct fib_nh_common *nhc = FIB_RES_NHC(res);
1037
1038 update_or_create_fnhe(nhc, fl4->daddr, 0, mtu, lock,
1039 jiffies + ip_rt_mtu_expires);
1040 }
1041 rcu_read_unlock();
1042 }
1043
1044 static void ip_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1045 struct sk_buff *skb, u32 mtu,
1046 bool confirm_neigh)
1047 {
1048 struct rtable *rt = (struct rtable *) dst;
1049 struct flowi4 fl4;
1050
1051 ip_rt_build_flow_key(&fl4, sk, skb);
1052 __ip_rt_update_pmtu(rt, &fl4, mtu);
1053 }
1054
1055 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu,
1056 int oif, u8 protocol)
1057 {
1058 const struct iphdr *iph = (const struct iphdr *) skb->data;
1059 struct flowi4 fl4;
1060 struct rtable *rt;
1061 u32 mark = IP4_REPLY_MARK(net, skb->mark);
1062
1063 __build_flow_key(net, &fl4, NULL, iph, oif,
1064 RT_TOS(iph->tos), protocol, mark, 0);
1065 rt = __ip_route_output_key(net, &fl4);
1066 if (!IS_ERR(rt)) {
1067 __ip_rt_update_pmtu(rt, &fl4, mtu);
1068 ip_rt_put(rt);
1069 }
1070 }
1071 EXPORT_SYMBOL_GPL(ipv4_update_pmtu);
1072
1073 static void __ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu)
1074 {
1075 const struct iphdr *iph = (const struct iphdr *) skb->data;
1076 struct flowi4 fl4;
1077 struct rtable *rt;
1078
1079 __build_flow_key(sock_net(sk), &fl4, sk, iph, 0, 0, 0, 0, 0);
1080
1081 if (!fl4.flowi4_mark)
1082 fl4.flowi4_mark = IP4_REPLY_MARK(sock_net(sk), skb->mark);
1083
1084 rt = __ip_route_output_key(sock_net(sk), &fl4);
1085 if (!IS_ERR(rt)) {
1086 __ip_rt_update_pmtu(rt, &fl4, mtu);
1087 ip_rt_put(rt);
1088 }
1089 }
1090
1091 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu)
1092 {
1093 const struct iphdr *iph = (const struct iphdr *) skb->data;
1094 struct flowi4 fl4;
1095 struct rtable *rt;
1096 struct dst_entry *odst = NULL;
1097 bool new = false;
1098 struct net *net = sock_net(sk);
1099
1100 bh_lock_sock(sk);
1101
1102 if (!ip_sk_accept_pmtu(sk))
1103 goto out;
1104
1105 odst = sk_dst_get(sk);
1106
1107 if (sock_owned_by_user(sk) || !odst) {
1108 __ipv4_sk_update_pmtu(skb, sk, mtu);
1109 goto out;
1110 }
1111
1112 __build_flow_key(net, &fl4, sk, iph, 0, 0, 0, 0, 0);
1113
1114 rt = (struct rtable *)odst;
1115 if (odst->obsolete && !odst->ops->check(odst, 0)) {
1116 rt = ip_route_output_flow(sock_net(sk), &fl4, sk);
1117 if (IS_ERR(rt))
1118 goto out;
1119
1120 new = true;
1121 }
1122
1123 __ip_rt_update_pmtu((struct rtable *) xfrm_dst_path(&rt->dst), &fl4, mtu);
1124
1125 if (!dst_check(&rt->dst, 0)) {
1126 if (new)
1127 dst_release(&rt->dst);
1128
1129 rt = ip_route_output_flow(sock_net(sk), &fl4, sk);
1130 if (IS_ERR(rt))
1131 goto out;
1132
1133 new = true;
1134 }
1135
1136 if (new)
1137 sk_dst_set(sk, &rt->dst);
1138
1139 out:
1140 bh_unlock_sock(sk);
1141 dst_release(odst);
1142 }
1143 EXPORT_SYMBOL_GPL(ipv4_sk_update_pmtu);
1144
1145 void ipv4_redirect(struct sk_buff *skb, struct net *net,
1146 int oif, u8 protocol)
1147 {
1148 const struct iphdr *iph = (const struct iphdr *) skb->data;
1149 struct flowi4 fl4;
1150 struct rtable *rt;
1151
1152 __build_flow_key(net, &fl4, NULL, iph, oif,
1153 RT_TOS(iph->tos), protocol, 0, 0);
1154 rt = __ip_route_output_key(net, &fl4);
1155 if (!IS_ERR(rt)) {
1156 __ip_do_redirect(rt, skb, &fl4, false);
1157 ip_rt_put(rt);
1158 }
1159 }
1160 EXPORT_SYMBOL_GPL(ipv4_redirect);
1161
1162 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk)
1163 {
1164 const struct iphdr *iph = (const struct iphdr *) skb->data;
1165 struct flowi4 fl4;
1166 struct rtable *rt;
1167 struct net *net = sock_net(sk);
1168
1169 __build_flow_key(net, &fl4, sk, iph, 0, 0, 0, 0, 0);
1170 rt = __ip_route_output_key(net, &fl4);
1171 if (!IS_ERR(rt)) {
1172 __ip_do_redirect(rt, skb, &fl4, false);
1173 ip_rt_put(rt);
1174 }
1175 }
1176 EXPORT_SYMBOL_GPL(ipv4_sk_redirect);
1177
1178 static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie)
1179 {
1180 struct rtable *rt = (struct rtable *) dst;
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190 if (dst->obsolete != DST_OBSOLETE_FORCE_CHK || rt_is_expired(rt))
1191 return NULL;
1192 return dst;
1193 }
1194
1195 static void ipv4_send_dest_unreach(struct sk_buff *skb)
1196 {
1197 struct ip_options opt;
1198 int res;
1199
1200
1201
1202
1203 if (!pskb_network_may_pull(skb, sizeof(struct iphdr)) ||
1204 ip_hdr(skb)->version != 4 || ip_hdr(skb)->ihl < 5)
1205 return;
1206
1207 memset(&opt, 0, sizeof(opt));
1208 if (ip_hdr(skb)->ihl > 5) {
1209 if (!pskb_network_may_pull(skb, ip_hdr(skb)->ihl * 4))
1210 return;
1211 opt.optlen = ip_hdr(skb)->ihl * 4 - sizeof(struct iphdr);
1212
1213 rcu_read_lock();
1214 res = __ip_options_compile(dev_net(skb->dev), &opt, skb, NULL);
1215 rcu_read_unlock();
1216
1217 if (res)
1218 return;
1219 }
1220 __icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0, &opt);
1221 }
1222
1223 static void ipv4_link_failure(struct sk_buff *skb)
1224 {
1225 struct rtable *rt;
1226
1227 ipv4_send_dest_unreach(skb);
1228
1229 rt = skb_rtable(skb);
1230 if (rt)
1231 dst_set_expires(&rt->dst, 0);
1232 }
1233
1234 static int ip_rt_bug(struct net *net, struct sock *sk, struct sk_buff *skb)
1235 {
1236 pr_debug("%s: %pI4 -> %pI4, %s\n",
1237 __func__, &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr,
1238 skb->dev ? skb->dev->name : "?");
1239 kfree_skb(skb);
1240 WARN_ON(1);
1241 return 0;
1242 }
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253 void ip_rt_get_source(u8 *addr, struct sk_buff *skb, struct rtable *rt)
1254 {
1255 __be32 src;
1256
1257 if (rt_is_output_route(rt))
1258 src = ip_hdr(skb)->saddr;
1259 else {
1260 struct fib_result res;
1261 struct iphdr *iph = ip_hdr(skb);
1262 struct flowi4 fl4 = {
1263 .daddr = iph->daddr,
1264 .saddr = iph->saddr,
1265 .flowi4_tos = RT_TOS(iph->tos),
1266 .flowi4_oif = rt->dst.dev->ifindex,
1267 .flowi4_iif = skb->dev->ifindex,
1268 .flowi4_mark = skb->mark,
1269 };
1270
1271 rcu_read_lock();
1272 if (fib_lookup(dev_net(rt->dst.dev), &fl4, &res, 0) == 0)
1273 src = fib_result_prefsrc(dev_net(rt->dst.dev), &res);
1274 else
1275 src = inet_select_addr(rt->dst.dev,
1276 rt_nexthop(rt, iph->daddr),
1277 RT_SCOPE_UNIVERSE);
1278 rcu_read_unlock();
1279 }
1280 memcpy(addr, &src, 4);
1281 }
1282
1283 #ifdef CONFIG_IP_ROUTE_CLASSID
1284 static void set_class_tag(struct rtable *rt, u32 tag)
1285 {
1286 if (!(rt->dst.tclassid & 0xFFFF))
1287 rt->dst.tclassid |= tag & 0xFFFF;
1288 if (!(rt->dst.tclassid & 0xFFFF0000))
1289 rt->dst.tclassid |= tag & 0xFFFF0000;
1290 }
1291 #endif
1292
1293 static unsigned int ipv4_default_advmss(const struct dst_entry *dst)
1294 {
1295 unsigned int header_size = sizeof(struct tcphdr) + sizeof(struct iphdr);
1296 unsigned int advmss = max_t(unsigned int, ipv4_mtu(dst) - header_size,
1297 ip_rt_min_advmss);
1298
1299 return min(advmss, IPV4_MAX_PMTU - header_size);
1300 }
1301
1302 static unsigned int ipv4_mtu(const struct dst_entry *dst)
1303 {
1304 const struct rtable *rt = (const struct rtable *) dst;
1305 unsigned int mtu = rt->rt_pmtu;
1306
1307 if (!mtu || time_after_eq(jiffies, rt->dst.expires))
1308 mtu = dst_metric_raw(dst, RTAX_MTU);
1309
1310 if (mtu)
1311 return mtu;
1312
1313 mtu = READ_ONCE(dst->dev->mtu);
1314
1315 if (unlikely(ip_mtu_locked(dst))) {
1316 if (rt->rt_uses_gateway && mtu > 576)
1317 mtu = 576;
1318 }
1319
1320 mtu = min_t(unsigned int, mtu, IP_MAX_MTU);
1321
1322 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
1323 }
1324
1325 static void ip_del_fnhe(struct fib_nh_common *nhc, __be32 daddr)
1326 {
1327 struct fnhe_hash_bucket *hash;
1328 struct fib_nh_exception *fnhe, __rcu **fnhe_p;
1329 u32 hval = fnhe_hashfun(daddr);
1330
1331 spin_lock_bh(&fnhe_lock);
1332
1333 hash = rcu_dereference_protected(nhc->nhc_exceptions,
1334 lockdep_is_held(&fnhe_lock));
1335 hash += hval;
1336
1337 fnhe_p = &hash->chain;
1338 fnhe = rcu_dereference_protected(*fnhe_p, lockdep_is_held(&fnhe_lock));
1339 while (fnhe) {
1340 if (fnhe->fnhe_daddr == daddr) {
1341 rcu_assign_pointer(*fnhe_p, rcu_dereference_protected(
1342 fnhe->fnhe_next, lockdep_is_held(&fnhe_lock)));
1343
1344
1345
1346 fnhe->fnhe_daddr = 0;
1347 fnhe_flush_routes(fnhe);
1348 kfree_rcu(fnhe, rcu);
1349 break;
1350 }
1351 fnhe_p = &fnhe->fnhe_next;
1352 fnhe = rcu_dereference_protected(fnhe->fnhe_next,
1353 lockdep_is_held(&fnhe_lock));
1354 }
1355
1356 spin_unlock_bh(&fnhe_lock);
1357 }
1358
1359 static struct fib_nh_exception *find_exception(struct fib_nh_common *nhc,
1360 __be32 daddr)
1361 {
1362 struct fnhe_hash_bucket *hash = rcu_dereference(nhc->nhc_exceptions);
1363 struct fib_nh_exception *fnhe;
1364 u32 hval;
1365
1366 if (!hash)
1367 return NULL;
1368
1369 hval = fnhe_hashfun(daddr);
1370
1371 for (fnhe = rcu_dereference(hash[hval].chain); fnhe;
1372 fnhe = rcu_dereference(fnhe->fnhe_next)) {
1373 if (fnhe->fnhe_daddr == daddr) {
1374 if (fnhe->fnhe_expires &&
1375 time_after(jiffies, fnhe->fnhe_expires)) {
1376 ip_del_fnhe(nhc, daddr);
1377 break;
1378 }
1379 return fnhe;
1380 }
1381 }
1382 return NULL;
1383 }
1384
1385
1386
1387
1388
1389
1390
1391 u32 ip_mtu_from_fib_result(struct fib_result *res, __be32 daddr)
1392 {
1393 struct fib_nh_common *nhc = res->nhc;
1394 struct net_device *dev = nhc->nhc_dev;
1395 struct fib_info *fi = res->fi;
1396 u32 mtu = 0;
1397
1398 if (dev_net(dev)->ipv4.sysctl_ip_fwd_use_pmtu ||
1399 fi->fib_metrics->metrics[RTAX_LOCK - 1] & (1 << RTAX_MTU))
1400 mtu = fi->fib_mtu;
1401
1402 if (likely(!mtu)) {
1403 struct fib_nh_exception *fnhe;
1404
1405 fnhe = find_exception(nhc, daddr);
1406 if (fnhe && !time_after_eq(jiffies, fnhe->fnhe_expires))
1407 mtu = fnhe->fnhe_pmtu;
1408 }
1409
1410 if (likely(!mtu))
1411 mtu = min(READ_ONCE(dev->mtu), IP_MAX_MTU);
1412
1413 return mtu - lwtunnel_headroom(nhc->nhc_lwtstate, mtu);
1414 }
1415
1416 static bool rt_bind_exception(struct rtable *rt, struct fib_nh_exception *fnhe,
1417 __be32 daddr, const bool do_cache)
1418 {
1419 bool ret = false;
1420
1421 spin_lock_bh(&fnhe_lock);
1422
1423 if (daddr == fnhe->fnhe_daddr) {
1424 struct rtable __rcu **porig;
1425 struct rtable *orig;
1426 int genid = fnhe_genid(dev_net(rt->dst.dev));
1427
1428 if (rt_is_input_route(rt))
1429 porig = &fnhe->fnhe_rth_input;
1430 else
1431 porig = &fnhe->fnhe_rth_output;
1432 orig = rcu_dereference(*porig);
1433
1434 if (fnhe->fnhe_genid != genid) {
1435 fnhe->fnhe_genid = genid;
1436 fnhe->fnhe_gw = 0;
1437 fnhe->fnhe_pmtu = 0;
1438 fnhe->fnhe_expires = 0;
1439 fnhe->fnhe_mtu_locked = false;
1440 fnhe_flush_routes(fnhe);
1441 orig = NULL;
1442 }
1443 fill_route_from_fnhe(rt, fnhe);
1444 if (!rt->rt_gw4) {
1445 rt->rt_gw4 = daddr;
1446 rt->rt_gw_family = AF_INET;
1447 }
1448
1449 if (do_cache) {
1450 dst_hold(&rt->dst);
1451 rcu_assign_pointer(*porig, rt);
1452 if (orig) {
1453 dst_dev_put(&orig->dst);
1454 dst_release(&orig->dst);
1455 }
1456 ret = true;
1457 }
1458
1459 fnhe->fnhe_stamp = jiffies;
1460 }
1461 spin_unlock_bh(&fnhe_lock);
1462
1463 return ret;
1464 }
1465
1466 static bool rt_cache_route(struct fib_nh_common *nhc, struct rtable *rt)
1467 {
1468 struct rtable *orig, *prev, **p;
1469 bool ret = true;
1470
1471 if (rt_is_input_route(rt)) {
1472 p = (struct rtable **)&nhc->nhc_rth_input;
1473 } else {
1474 p = (struct rtable **)raw_cpu_ptr(nhc->nhc_pcpu_rth_output);
1475 }
1476 orig = *p;
1477
1478
1479
1480
1481 dst_hold(&rt->dst);
1482 prev = cmpxchg(p, orig, rt);
1483 if (prev == orig) {
1484 if (orig) {
1485 rt_add_uncached_list(orig);
1486 dst_release(&orig->dst);
1487 }
1488 } else {
1489 dst_release(&rt->dst);
1490 ret = false;
1491 }
1492
1493 return ret;
1494 }
1495
1496 struct uncached_list {
1497 spinlock_t lock;
1498 struct list_head head;
1499 };
1500
1501 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt_uncached_list);
1502
1503 void rt_add_uncached_list(struct rtable *rt)
1504 {
1505 struct uncached_list *ul = raw_cpu_ptr(&rt_uncached_list);
1506
1507 rt->rt_uncached_list = ul;
1508
1509 spin_lock_bh(&ul->lock);
1510 list_add_tail(&rt->rt_uncached, &ul->head);
1511 spin_unlock_bh(&ul->lock);
1512 }
1513
1514 void rt_del_uncached_list(struct rtable *rt)
1515 {
1516 if (!list_empty(&rt->rt_uncached)) {
1517 struct uncached_list *ul = rt->rt_uncached_list;
1518
1519 spin_lock_bh(&ul->lock);
1520 list_del(&rt->rt_uncached);
1521 spin_unlock_bh(&ul->lock);
1522 }
1523 }
1524
1525 static void ipv4_dst_destroy(struct dst_entry *dst)
1526 {
1527 struct rtable *rt = (struct rtable *)dst;
1528
1529 ip_dst_metrics_put(dst);
1530 rt_del_uncached_list(rt);
1531 }
1532
1533 void rt_flush_dev(struct net_device *dev)
1534 {
1535 struct rtable *rt;
1536 int cpu;
1537
1538 for_each_possible_cpu(cpu) {
1539 struct uncached_list *ul = &per_cpu(rt_uncached_list, cpu);
1540
1541 spin_lock_bh(&ul->lock);
1542 list_for_each_entry(rt, &ul->head, rt_uncached) {
1543 if (rt->dst.dev != dev)
1544 continue;
1545 rt->dst.dev = blackhole_netdev;
1546 dev_hold(rt->dst.dev);
1547 dev_put(dev);
1548 }
1549 spin_unlock_bh(&ul->lock);
1550 }
1551 }
1552
1553 static bool rt_cache_valid(const struct rtable *rt)
1554 {
1555 return rt &&
1556 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1557 !rt_is_expired(rt);
1558 }
1559
1560 static void rt_set_nexthop(struct rtable *rt, __be32 daddr,
1561 const struct fib_result *res,
1562 struct fib_nh_exception *fnhe,
1563 struct fib_info *fi, u16 type, u32 itag,
1564 const bool do_cache)
1565 {
1566 bool cached = false;
1567
1568 if (fi) {
1569 struct fib_nh_common *nhc = FIB_RES_NHC(*res);
1570
1571 if (nhc->nhc_gw_family && nhc->nhc_scope == RT_SCOPE_LINK) {
1572 rt->rt_uses_gateway = 1;
1573 rt->rt_gw_family = nhc->nhc_gw_family;
1574
1575 if (likely(nhc->nhc_gw_family == AF_INET))
1576 rt->rt_gw4 = nhc->nhc_gw.ipv4;
1577 else
1578 rt->rt_gw6 = nhc->nhc_gw.ipv6;
1579 }
1580
1581 ip_dst_init_metrics(&rt->dst, fi->fib_metrics);
1582
1583 #ifdef CONFIG_IP_ROUTE_CLASSID
1584 if (nhc->nhc_family == AF_INET) {
1585 struct fib_nh *nh;
1586
1587 nh = container_of(nhc, struct fib_nh, nh_common);
1588 rt->dst.tclassid = nh->nh_tclassid;
1589 }
1590 #endif
1591 rt->dst.lwtstate = lwtstate_get(nhc->nhc_lwtstate);
1592 if (unlikely(fnhe))
1593 cached = rt_bind_exception(rt, fnhe, daddr, do_cache);
1594 else if (do_cache)
1595 cached = rt_cache_route(nhc, rt);
1596 if (unlikely(!cached)) {
1597
1598
1599
1600
1601
1602 if (!rt->rt_gw4) {
1603 rt->rt_gw_family = AF_INET;
1604 rt->rt_gw4 = daddr;
1605 }
1606 rt_add_uncached_list(rt);
1607 }
1608 } else
1609 rt_add_uncached_list(rt);
1610
1611 #ifdef CONFIG_IP_ROUTE_CLASSID
1612 #ifdef CONFIG_IP_MULTIPLE_TABLES
1613 set_class_tag(rt, res->tclassid);
1614 #endif
1615 set_class_tag(rt, itag);
1616 #endif
1617 }
1618
1619 struct rtable *rt_dst_alloc(struct net_device *dev,
1620 unsigned int flags, u16 type,
1621 bool nopolicy, bool noxfrm, bool will_cache)
1622 {
1623 struct rtable *rt;
1624
1625 rt = dst_alloc(&ipv4_dst_ops, dev, 1, DST_OBSOLETE_FORCE_CHK,
1626 (will_cache ? 0 : DST_HOST) |
1627 (nopolicy ? DST_NOPOLICY : 0) |
1628 (noxfrm ? DST_NOXFRM : 0));
1629
1630 if (rt) {
1631 rt->rt_genid = rt_genid_ipv4(dev_net(dev));
1632 rt->rt_flags = flags;
1633 rt->rt_type = type;
1634 rt->rt_is_input = 0;
1635 rt->rt_iif = 0;
1636 rt->rt_pmtu = 0;
1637 rt->rt_mtu_locked = 0;
1638 rt->rt_uses_gateway = 0;
1639 rt->rt_gw_family = 0;
1640 rt->rt_gw4 = 0;
1641 INIT_LIST_HEAD(&rt->rt_uncached);
1642
1643 rt->dst.output = ip_output;
1644 if (flags & RTCF_LOCAL)
1645 rt->dst.input = ip_local_deliver;
1646 }
1647
1648 return rt;
1649 }
1650 EXPORT_SYMBOL(rt_dst_alloc);
1651
1652 struct rtable *rt_dst_clone(struct net_device *dev, struct rtable *rt)
1653 {
1654 struct rtable *new_rt;
1655
1656 new_rt = dst_alloc(&ipv4_dst_ops, dev, 1, DST_OBSOLETE_FORCE_CHK,
1657 rt->dst.flags);
1658
1659 if (new_rt) {
1660 new_rt->rt_genid = rt_genid_ipv4(dev_net(dev));
1661 new_rt->rt_flags = rt->rt_flags;
1662 new_rt->rt_type = rt->rt_type;
1663 new_rt->rt_is_input = rt->rt_is_input;
1664 new_rt->rt_iif = rt->rt_iif;
1665 new_rt->rt_pmtu = rt->rt_pmtu;
1666 new_rt->rt_mtu_locked = rt->rt_mtu_locked;
1667 new_rt->rt_gw_family = rt->rt_gw_family;
1668 if (rt->rt_gw_family == AF_INET)
1669 new_rt->rt_gw4 = rt->rt_gw4;
1670 else if (rt->rt_gw_family == AF_INET6)
1671 new_rt->rt_gw6 = rt->rt_gw6;
1672 INIT_LIST_HEAD(&new_rt->rt_uncached);
1673
1674 new_rt->dst.flags |= DST_HOST;
1675 new_rt->dst.input = rt->dst.input;
1676 new_rt->dst.output = rt->dst.output;
1677 new_rt->dst.error = rt->dst.error;
1678 new_rt->dst.lastuse = jiffies;
1679 new_rt->dst.lwtstate = lwtstate_get(rt->dst.lwtstate);
1680 }
1681 return new_rt;
1682 }
1683 EXPORT_SYMBOL(rt_dst_clone);
1684
1685
1686 int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr,
1687 u8 tos, struct net_device *dev,
1688 struct in_device *in_dev, u32 *itag)
1689 {
1690 int err;
1691
1692
1693 if (!in_dev)
1694 return -EINVAL;
1695
1696 if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) ||
1697 skb->protocol != htons(ETH_P_IP))
1698 return -EINVAL;
1699
1700 if (ipv4_is_loopback(saddr) && !IN_DEV_ROUTE_LOCALNET(in_dev))
1701 return -EINVAL;
1702
1703 if (ipv4_is_zeronet(saddr)) {
1704 if (!ipv4_is_local_multicast(daddr) &&
1705 ip_hdr(skb)->protocol != IPPROTO_IGMP)
1706 return -EINVAL;
1707 } else {
1708 err = fib_validate_source(skb, saddr, 0, tos, 0, dev,
1709 in_dev, itag);
1710 if (err < 0)
1711 return err;
1712 }
1713 return 0;
1714 }
1715
1716
1717 static int ip_route_input_mc(struct sk_buff *skb, __be32 daddr, __be32 saddr,
1718 u8 tos, struct net_device *dev, int our)
1719 {
1720 struct in_device *in_dev = __in_dev_get_rcu(dev);
1721 unsigned int flags = RTCF_MULTICAST;
1722 struct rtable *rth;
1723 u32 itag = 0;
1724 int err;
1725
1726 err = ip_mc_validate_source(skb, daddr, saddr, tos, dev, in_dev, &itag);
1727 if (err)
1728 return err;
1729
1730 if (our)
1731 flags |= RTCF_LOCAL;
1732
1733 rth = rt_dst_alloc(dev_net(dev)->loopback_dev, flags, RTN_MULTICAST,
1734 IN_DEV_CONF_GET(in_dev, NOPOLICY), false, false);
1735 if (!rth)
1736 return -ENOBUFS;
1737
1738 #ifdef CONFIG_IP_ROUTE_CLASSID
1739 rth->dst.tclassid = itag;
1740 #endif
1741 rth->dst.output = ip_rt_bug;
1742 rth->rt_is_input= 1;
1743
1744 #ifdef CONFIG_IP_MROUTE
1745 if (!ipv4_is_local_multicast(daddr) && IN_DEV_MFORWARD(in_dev))
1746 rth->dst.input = ip_mr_input;
1747 #endif
1748 RT_CACHE_STAT_INC(in_slow_mc);
1749
1750 skb_dst_set(skb, &rth->dst);
1751 return 0;
1752 }
1753
1754
1755 static void ip_handle_martian_source(struct net_device *dev,
1756 struct in_device *in_dev,
1757 struct sk_buff *skb,
1758 __be32 daddr,
1759 __be32 saddr)
1760 {
1761 RT_CACHE_STAT_INC(in_martian_src);
1762 #ifdef CONFIG_IP_ROUTE_VERBOSE
1763 if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) {
1764
1765
1766
1767
1768 pr_warn("martian source %pI4 from %pI4, on dev %s\n",
1769 &daddr, &saddr, dev->name);
1770 if (dev->hard_header_len && skb_mac_header_was_set(skb)) {
1771 print_hex_dump(KERN_WARNING, "ll header: ",
1772 DUMP_PREFIX_OFFSET, 16, 1,
1773 skb_mac_header(skb),
1774 dev->hard_header_len, false);
1775 }
1776 }
1777 #endif
1778 }
1779
1780
1781 static int __mkroute_input(struct sk_buff *skb,
1782 const struct fib_result *res,
1783 struct in_device *in_dev,
1784 __be32 daddr, __be32 saddr, u32 tos)
1785 {
1786 struct fib_nh_common *nhc = FIB_RES_NHC(*res);
1787 struct net_device *dev = nhc->nhc_dev;
1788 struct fib_nh_exception *fnhe;
1789 struct rtable *rth;
1790 int err;
1791 struct in_device *out_dev;
1792 bool do_cache;
1793 u32 itag = 0;
1794
1795
1796 out_dev = __in_dev_get_rcu(dev);
1797 if (!out_dev) {
1798 net_crit_ratelimited("Bug in ip_route_input_slow(). Please report.\n");
1799 return -EINVAL;
1800 }
1801
1802 err = fib_validate_source(skb, saddr, daddr, tos, FIB_RES_OIF(*res),
1803 in_dev->dev, in_dev, &itag);
1804 if (err < 0) {
1805 ip_handle_martian_source(in_dev->dev, in_dev, skb, daddr,
1806 saddr);
1807
1808 goto cleanup;
1809 }
1810
1811 do_cache = res->fi && !itag;
1812 if (out_dev == in_dev && err && IN_DEV_TX_REDIRECTS(out_dev) &&
1813 skb->protocol == htons(ETH_P_IP)) {
1814 __be32 gw;
1815
1816 gw = nhc->nhc_gw_family == AF_INET ? nhc->nhc_gw.ipv4 : 0;
1817 if (IN_DEV_SHARED_MEDIA(out_dev) ||
1818 inet_addr_onlink(out_dev, saddr, gw))
1819 IPCB(skb)->flags |= IPSKB_DOREDIRECT;
1820 }
1821
1822 if (skb->protocol != htons(ETH_P_IP)) {
1823
1824
1825
1826
1827
1828
1829
1830 if (out_dev == in_dev &&
1831 IN_DEV_PROXY_ARP_PVLAN(in_dev) == 0) {
1832 err = -EINVAL;
1833 goto cleanup;
1834 }
1835 }
1836
1837 fnhe = find_exception(nhc, daddr);
1838 if (do_cache) {
1839 if (fnhe)
1840 rth = rcu_dereference(fnhe->fnhe_rth_input);
1841 else
1842 rth = rcu_dereference(nhc->nhc_rth_input);
1843 if (rt_cache_valid(rth)) {
1844 skb_dst_set_noref(skb, &rth->dst);
1845 goto out;
1846 }
1847 }
1848
1849 rth = rt_dst_alloc(out_dev->dev, 0, res->type,
1850 IN_DEV_CONF_GET(in_dev, NOPOLICY),
1851 IN_DEV_CONF_GET(out_dev, NOXFRM), do_cache);
1852 if (!rth) {
1853 err = -ENOBUFS;
1854 goto cleanup;
1855 }
1856
1857 rth->rt_is_input = 1;
1858 RT_CACHE_STAT_INC(in_slow_tot);
1859
1860 rth->dst.input = ip_forward;
1861
1862 rt_set_nexthop(rth, daddr, res, fnhe, res->fi, res->type, itag,
1863 do_cache);
1864 lwtunnel_set_redirect(&rth->dst);
1865 skb_dst_set(skb, &rth->dst);
1866 out:
1867 err = 0;
1868 cleanup:
1869 return err;
1870 }
1871
1872 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1873
1874
1875
1876 static void ip_multipath_l3_keys(const struct sk_buff *skb,
1877 struct flow_keys *hash_keys)
1878 {
1879 const struct iphdr *outer_iph = ip_hdr(skb);
1880 const struct iphdr *key_iph = outer_iph;
1881 const struct iphdr *inner_iph;
1882 const struct icmphdr *icmph;
1883 struct iphdr _inner_iph;
1884 struct icmphdr _icmph;
1885
1886 if (likely(outer_iph->protocol != IPPROTO_ICMP))
1887 goto out;
1888
1889 if (unlikely((outer_iph->frag_off & htons(IP_OFFSET)) != 0))
1890 goto out;
1891
1892 icmph = skb_header_pointer(skb, outer_iph->ihl * 4, sizeof(_icmph),
1893 &_icmph);
1894 if (!icmph)
1895 goto out;
1896
1897 if (icmph->type != ICMP_DEST_UNREACH &&
1898 icmph->type != ICMP_REDIRECT &&
1899 icmph->type != ICMP_TIME_EXCEEDED &&
1900 icmph->type != ICMP_PARAMETERPROB)
1901 goto out;
1902
1903 inner_iph = skb_header_pointer(skb,
1904 outer_iph->ihl * 4 + sizeof(_icmph),
1905 sizeof(_inner_iph), &_inner_iph);
1906 if (!inner_iph)
1907 goto out;
1908
1909 key_iph = inner_iph;
1910 out:
1911 hash_keys->addrs.v4addrs.src = key_iph->saddr;
1912 hash_keys->addrs.v4addrs.dst = key_iph->daddr;
1913 }
1914
1915
1916 int fib_multipath_hash(const struct net *net, const struct flowi4 *fl4,
1917 const struct sk_buff *skb, struct flow_keys *flkeys)
1918 {
1919 u32 multipath_hash = fl4 ? fl4->flowi4_multipath_hash : 0;
1920 struct flow_keys hash_keys;
1921 u32 mhash;
1922
1923 switch (net->ipv4.sysctl_fib_multipath_hash_policy) {
1924 case 0:
1925 memset(&hash_keys, 0, sizeof(hash_keys));
1926 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1927 if (skb) {
1928 ip_multipath_l3_keys(skb, &hash_keys);
1929 } else {
1930 hash_keys.addrs.v4addrs.src = fl4->saddr;
1931 hash_keys.addrs.v4addrs.dst = fl4->daddr;
1932 }
1933 break;
1934 case 1:
1935
1936 if (skb) {
1937 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
1938 struct flow_keys keys;
1939
1940
1941 if (skb->l4_hash)
1942 return skb_get_hash_raw(skb) >> 1;
1943
1944 memset(&hash_keys, 0, sizeof(hash_keys));
1945
1946 if (!flkeys) {
1947 skb_flow_dissect_flow_keys(skb, &keys, flag);
1948 flkeys = &keys;
1949 }
1950
1951 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1952 hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
1953 hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
1954 hash_keys.ports.src = flkeys->ports.src;
1955 hash_keys.ports.dst = flkeys->ports.dst;
1956 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
1957 } else {
1958 memset(&hash_keys, 0, sizeof(hash_keys));
1959 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1960 hash_keys.addrs.v4addrs.src = fl4->saddr;
1961 hash_keys.addrs.v4addrs.dst = fl4->daddr;
1962 hash_keys.ports.src = fl4->fl4_sport;
1963 hash_keys.ports.dst = fl4->fl4_dport;
1964 hash_keys.basic.ip_proto = fl4->flowi4_proto;
1965 }
1966 break;
1967 case 2:
1968 memset(&hash_keys, 0, sizeof(hash_keys));
1969
1970 if (skb) {
1971 struct flow_keys keys;
1972
1973 skb_flow_dissect_flow_keys(skb, &keys, 0);
1974
1975 if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
1976 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1977 hash_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src;
1978 hash_keys.addrs.v4addrs.dst = keys.addrs.v4addrs.dst;
1979 } else if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
1980 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1981 hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
1982 hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
1983 hash_keys.tags.flow_label = keys.tags.flow_label;
1984 hash_keys.basic.ip_proto = keys.basic.ip_proto;
1985 } else {
1986
1987 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1988 ip_multipath_l3_keys(skb, &hash_keys);
1989 }
1990 } else {
1991
1992 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1993 hash_keys.addrs.v4addrs.src = fl4->saddr;
1994 hash_keys.addrs.v4addrs.dst = fl4->daddr;
1995 }
1996 break;
1997 }
1998 mhash = flow_hash_from_keys(&hash_keys);
1999
2000 if (multipath_hash)
2001 mhash = jhash_2words(mhash, multipath_hash, 0);
2002
2003 return mhash >> 1;
2004 }
2005 #endif
2006
2007 static int ip_mkroute_input(struct sk_buff *skb,
2008 struct fib_result *res,
2009 struct in_device *in_dev,
2010 __be32 daddr, __be32 saddr, u32 tos,
2011 struct flow_keys *hkeys)
2012 {
2013 #ifdef CONFIG_IP_ROUTE_MULTIPATH
2014 if (res->fi && fib_info_num_path(res->fi) > 1) {
2015 int h = fib_multipath_hash(res->fi->fib_net, NULL, skb, hkeys);
2016
2017 fib_select_multipath(res, h);
2018 }
2019 #endif
2020
2021
2022 return __mkroute_input(skb, res, in_dev, daddr, saddr, tos);
2023 }
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036 static int ip_route_input_slow(struct sk_buff *skb, __be32 daddr, __be32 saddr,
2037 u8 tos, struct net_device *dev,
2038 struct fib_result *res)
2039 {
2040 struct in_device *in_dev = __in_dev_get_rcu(dev);
2041 struct flow_keys *flkeys = NULL, _flkeys;
2042 struct net *net = dev_net(dev);
2043 struct ip_tunnel_info *tun_info;
2044 int err = -EINVAL;
2045 unsigned int flags = 0;
2046 u32 itag = 0;
2047 struct rtable *rth;
2048 struct flowi4 fl4;
2049 bool do_cache = true;
2050
2051
2052
2053 if (!in_dev)
2054 goto out;
2055
2056
2057
2058
2059
2060 tun_info = skb_tunnel_info(skb);
2061 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2062 fl4.flowi4_tun_key.tun_id = tun_info->key.tun_id;
2063 else
2064 fl4.flowi4_tun_key.tun_id = 0;
2065 skb_dst_drop(skb);
2066
2067 if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr))
2068 goto martian_source;
2069
2070 res->fi = NULL;
2071 res->table = NULL;
2072 if (ipv4_is_lbcast(daddr) || (saddr == 0 && daddr == 0))
2073 goto brd_input;
2074
2075
2076
2077
2078 if (ipv4_is_zeronet(saddr))
2079 goto martian_source;
2080
2081 if (ipv4_is_zeronet(daddr))
2082 goto martian_destination;
2083
2084
2085
2086
2087 if (ipv4_is_loopback(daddr)) {
2088 if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev, net))
2089 goto martian_destination;
2090 } else if (ipv4_is_loopback(saddr)) {
2091 if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev, net))
2092 goto martian_source;
2093 }
2094
2095
2096
2097
2098 fl4.flowi4_oif = 0;
2099 fl4.flowi4_iif = dev->ifindex;
2100 fl4.flowi4_mark = skb->mark;
2101 fl4.flowi4_tos = tos;
2102 fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
2103 fl4.flowi4_flags = 0;
2104 fl4.daddr = daddr;
2105 fl4.saddr = saddr;
2106 fl4.flowi4_uid = sock_net_uid(net, NULL);
2107
2108 if (fib4_rules_early_flow_dissect(net, skb, &fl4, &_flkeys)) {
2109 flkeys = &_flkeys;
2110 } else {
2111 fl4.flowi4_proto = 0;
2112 fl4.fl4_sport = 0;
2113 fl4.fl4_dport = 0;
2114 }
2115
2116 err = fib_lookup(net, &fl4, res, 0);
2117 if (err != 0) {
2118 if (!IN_DEV_FORWARD(in_dev))
2119 err = -EHOSTUNREACH;
2120 goto no_route;
2121 }
2122
2123 if (res->type == RTN_BROADCAST) {
2124 if (IN_DEV_BFORWARD(in_dev))
2125 goto make_route;
2126
2127 if (IPV4_DEVCONF_ALL(net, BC_FORWARDING))
2128 do_cache = false;
2129 goto brd_input;
2130 }
2131
2132 if (res->type == RTN_LOCAL) {
2133 err = fib_validate_source(skb, saddr, daddr, tos,
2134 0, dev, in_dev, &itag);
2135 if (err < 0)
2136 goto martian_source;
2137 goto local_input;
2138 }
2139
2140 if (!IN_DEV_FORWARD(in_dev)) {
2141 err = -EHOSTUNREACH;
2142 goto no_route;
2143 }
2144 if (res->type != RTN_UNICAST)
2145 goto martian_destination;
2146
2147 make_route:
2148 err = ip_mkroute_input(skb, res, in_dev, daddr, saddr, tos, flkeys);
2149 out: return err;
2150
2151 brd_input:
2152 if (skb->protocol != htons(ETH_P_IP))
2153 goto e_inval;
2154
2155 if (!ipv4_is_zeronet(saddr)) {
2156 err = fib_validate_source(skb, saddr, 0, tos, 0, dev,
2157 in_dev, &itag);
2158 if (err < 0)
2159 goto martian_source;
2160 }
2161 flags |= RTCF_BROADCAST;
2162 res->type = RTN_BROADCAST;
2163 RT_CACHE_STAT_INC(in_brd);
2164
2165 local_input:
2166 do_cache &= res->fi && !itag;
2167 if (do_cache) {
2168 struct fib_nh_common *nhc = FIB_RES_NHC(*res);
2169
2170 rth = rcu_dereference(nhc->nhc_rth_input);
2171 if (rt_cache_valid(rth)) {
2172 skb_dst_set_noref(skb, &rth->dst);
2173 err = 0;
2174 goto out;
2175 }
2176 }
2177
2178 rth = rt_dst_alloc(l3mdev_master_dev_rcu(dev) ? : net->loopback_dev,
2179 flags | RTCF_LOCAL, res->type,
2180 IN_DEV_CONF_GET(in_dev, NOPOLICY), false, do_cache);
2181 if (!rth)
2182 goto e_nobufs;
2183
2184 rth->dst.output= ip_rt_bug;
2185 #ifdef CONFIG_IP_ROUTE_CLASSID
2186 rth->dst.tclassid = itag;
2187 #endif
2188 rth->rt_is_input = 1;
2189
2190 RT_CACHE_STAT_INC(in_slow_tot);
2191 if (res->type == RTN_UNREACHABLE) {
2192 rth->dst.input= ip_error;
2193 rth->dst.error= -err;
2194 rth->rt_flags &= ~RTCF_LOCAL;
2195 }
2196
2197 if (do_cache) {
2198 struct fib_nh_common *nhc = FIB_RES_NHC(*res);
2199
2200 rth->dst.lwtstate = lwtstate_get(nhc->nhc_lwtstate);
2201 if (lwtunnel_input_redirect(rth->dst.lwtstate)) {
2202 WARN_ON(rth->dst.input == lwtunnel_input);
2203 rth->dst.lwtstate->orig_input = rth->dst.input;
2204 rth->dst.input = lwtunnel_input;
2205 }
2206
2207 if (unlikely(!rt_cache_route(nhc, rth)))
2208 rt_add_uncached_list(rth);
2209 }
2210 skb_dst_set(skb, &rth->dst);
2211 err = 0;
2212 goto out;
2213
2214 no_route:
2215 RT_CACHE_STAT_INC(in_no_route);
2216 res->type = RTN_UNREACHABLE;
2217 res->fi = NULL;
2218 res->table = NULL;
2219 goto local_input;
2220
2221
2222
2223
2224 martian_destination:
2225 RT_CACHE_STAT_INC(in_martian_dst);
2226 #ifdef CONFIG_IP_ROUTE_VERBOSE
2227 if (IN_DEV_LOG_MARTIANS(in_dev))
2228 net_warn_ratelimited("martian destination %pI4 from %pI4, dev %s\n",
2229 &daddr, &saddr, dev->name);
2230 #endif
2231
2232 e_inval:
2233 err = -EINVAL;
2234 goto out;
2235
2236 e_nobufs:
2237 err = -ENOBUFS;
2238 goto out;
2239
2240 martian_source:
2241 ip_handle_martian_source(dev, in_dev, skb, daddr, saddr);
2242 goto out;
2243 }
2244
2245 int ip_route_input_noref(struct sk_buff *skb, __be32 daddr, __be32 saddr,
2246 u8 tos, struct net_device *dev)
2247 {
2248 struct fib_result res;
2249 int err;
2250
2251 tos &= IPTOS_RT_MASK;
2252 rcu_read_lock();
2253 err = ip_route_input_rcu(skb, daddr, saddr, tos, dev, &res);
2254 rcu_read_unlock();
2255
2256 return err;
2257 }
2258 EXPORT_SYMBOL(ip_route_input_noref);
2259
2260
2261 int ip_route_input_rcu(struct sk_buff *skb, __be32 daddr, __be32 saddr,
2262 u8 tos, struct net_device *dev, struct fib_result *res)
2263 {
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275 if (ipv4_is_multicast(daddr)) {
2276 struct in_device *in_dev = __in_dev_get_rcu(dev);
2277 int our = 0;
2278 int err = -EINVAL;
2279
2280 if (!in_dev)
2281 return err;
2282 our = ip_check_mc_rcu(in_dev, daddr, saddr,
2283 ip_hdr(skb)->protocol);
2284
2285
2286 if (!our && netif_is_l3_slave(dev)) {
2287 struct in_device *l3_in_dev;
2288
2289 l3_in_dev = __in_dev_get_rcu(skb->dev);
2290 if (l3_in_dev)
2291 our = ip_check_mc_rcu(l3_in_dev, daddr, saddr,
2292 ip_hdr(skb)->protocol);
2293 }
2294
2295 if (our
2296 #ifdef CONFIG_IP_MROUTE
2297 ||
2298 (!ipv4_is_local_multicast(daddr) &&
2299 IN_DEV_MFORWARD(in_dev))
2300 #endif
2301 ) {
2302 err = ip_route_input_mc(skb, daddr, saddr,
2303 tos, dev, our);
2304 }
2305 return err;
2306 }
2307
2308 return ip_route_input_slow(skb, daddr, saddr, tos, dev, res);
2309 }
2310
2311
2312 static struct rtable *__mkroute_output(const struct fib_result *res,
2313 const struct flowi4 *fl4, int orig_oif,
2314 struct net_device *dev_out,
2315 unsigned int flags)
2316 {
2317 struct fib_info *fi = res->fi;
2318 struct fib_nh_exception *fnhe;
2319 struct in_device *in_dev;
2320 u16 type = res->type;
2321 struct rtable *rth;
2322 bool do_cache;
2323
2324 in_dev = __in_dev_get_rcu(dev_out);
2325 if (!in_dev)
2326 return ERR_PTR(-EINVAL);
2327
2328 if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev)))
2329 if (ipv4_is_loopback(fl4->saddr) &&
2330 !(dev_out->flags & IFF_LOOPBACK) &&
2331 !netif_is_l3_master(dev_out))
2332 return ERR_PTR(-EINVAL);
2333
2334 if (ipv4_is_lbcast(fl4->daddr))
2335 type = RTN_BROADCAST;
2336 else if (ipv4_is_multicast(fl4->daddr))
2337 type = RTN_MULTICAST;
2338 else if (ipv4_is_zeronet(fl4->daddr))
2339 return ERR_PTR(-EINVAL);
2340
2341 if (dev_out->flags & IFF_LOOPBACK)
2342 flags |= RTCF_LOCAL;
2343
2344 do_cache = true;
2345 if (type == RTN_BROADCAST) {
2346 flags |= RTCF_BROADCAST | RTCF_LOCAL;
2347 fi = NULL;
2348 } else if (type == RTN_MULTICAST) {
2349 flags |= RTCF_MULTICAST | RTCF_LOCAL;
2350 if (!ip_check_mc_rcu(in_dev, fl4->daddr, fl4->saddr,
2351 fl4->flowi4_proto))
2352 flags &= ~RTCF_LOCAL;
2353 else
2354 do_cache = false;
2355
2356
2357
2358
2359 if (fi && res->prefixlen < 4)
2360 fi = NULL;
2361 } else if ((type == RTN_LOCAL) && (orig_oif != 0) &&
2362 (orig_oif != dev_out->ifindex)) {
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372 do_cache = false;
2373 }
2374
2375 fnhe = NULL;
2376 do_cache &= fi != NULL;
2377 if (fi) {
2378 struct fib_nh_common *nhc = FIB_RES_NHC(*res);
2379 struct rtable __rcu **prth;
2380
2381 fnhe = find_exception(nhc, fl4->daddr);
2382 if (!do_cache)
2383 goto add;
2384 if (fnhe) {
2385 prth = &fnhe->fnhe_rth_output;
2386 } else {
2387 if (unlikely(fl4->flowi4_flags &
2388 FLOWI_FLAG_KNOWN_NH &&
2389 !(nhc->nhc_gw_family &&
2390 nhc->nhc_scope == RT_SCOPE_LINK))) {
2391 do_cache = false;
2392 goto add;
2393 }
2394 prth = raw_cpu_ptr(nhc->nhc_pcpu_rth_output);
2395 }
2396 rth = rcu_dereference(*prth);
2397 if (rt_cache_valid(rth) && dst_hold_safe(&rth->dst))
2398 return rth;
2399 }
2400
2401 add:
2402 rth = rt_dst_alloc(dev_out, flags, type,
2403 IN_DEV_CONF_GET(in_dev, NOPOLICY),
2404 IN_DEV_CONF_GET(in_dev, NOXFRM),
2405 do_cache);
2406 if (!rth)
2407 return ERR_PTR(-ENOBUFS);
2408
2409 rth->rt_iif = orig_oif;
2410
2411 RT_CACHE_STAT_INC(out_slow_tot);
2412
2413 if (flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
2414 if (flags & RTCF_LOCAL &&
2415 !(dev_out->flags & IFF_LOOPBACK)) {
2416 rth->dst.output = ip_mc_output;
2417 RT_CACHE_STAT_INC(out_slow_mc);
2418 }
2419 #ifdef CONFIG_IP_MROUTE
2420 if (type == RTN_MULTICAST) {
2421 if (IN_DEV_MFORWARD(in_dev) &&
2422 !ipv4_is_local_multicast(fl4->daddr)) {
2423 rth->dst.input = ip_mr_input;
2424 rth->dst.output = ip_mc_output;
2425 }
2426 }
2427 #endif
2428 }
2429
2430 rt_set_nexthop(rth, fl4->daddr, res, fnhe, fi, type, 0, do_cache);
2431 lwtunnel_set_redirect(&rth->dst);
2432
2433 return rth;
2434 }
2435
2436
2437
2438
2439
2440 struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *fl4,
2441 const struct sk_buff *skb)
2442 {
2443 __u8 tos = RT_FL_TOS(fl4);
2444 struct fib_result res = {
2445 .type = RTN_UNSPEC,
2446 .fi = NULL,
2447 .table = NULL,
2448 .tclassid = 0,
2449 };
2450 struct rtable *rth;
2451
2452 fl4->flowi4_iif = LOOPBACK_IFINDEX;
2453 fl4->flowi4_tos = tos & IPTOS_RT_MASK;
2454 fl4->flowi4_scope = ((tos & RTO_ONLINK) ?
2455 RT_SCOPE_LINK : RT_SCOPE_UNIVERSE);
2456
2457 rcu_read_lock();
2458 rth = ip_route_output_key_hash_rcu(net, fl4, &res, skb);
2459 rcu_read_unlock();
2460
2461 return rth;
2462 }
2463 EXPORT_SYMBOL_GPL(ip_route_output_key_hash);
2464
2465 struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *fl4,
2466 struct fib_result *res,
2467 const struct sk_buff *skb)
2468 {
2469 struct net_device *dev_out = NULL;
2470 int orig_oif = fl4->flowi4_oif;
2471 unsigned int flags = 0;
2472 struct rtable *rth;
2473 int err;
2474
2475 if (fl4->saddr) {
2476 if (ipv4_is_multicast(fl4->saddr) ||
2477 ipv4_is_lbcast(fl4->saddr) ||
2478 ipv4_is_zeronet(fl4->saddr)) {
2479 rth = ERR_PTR(-EINVAL);
2480 goto out;
2481 }
2482
2483 rth = ERR_PTR(-ENETUNREACH);
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493 if (fl4->flowi4_oif == 0 &&
2494 (ipv4_is_multicast(fl4->daddr) ||
2495 ipv4_is_lbcast(fl4->daddr))) {
2496
2497 dev_out = __ip_dev_find(net, fl4->saddr, false);
2498 if (!dev_out)
2499 goto out;
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516 fl4->flowi4_oif = dev_out->ifindex;
2517 goto make_route;
2518 }
2519
2520 if (!(fl4->flowi4_flags & FLOWI_FLAG_ANYSRC)) {
2521
2522 if (!__ip_dev_find(net, fl4->saddr, false))
2523 goto out;
2524 }
2525 }
2526
2527
2528 if (fl4->flowi4_oif) {
2529 dev_out = dev_get_by_index_rcu(net, fl4->flowi4_oif);
2530 rth = ERR_PTR(-ENODEV);
2531 if (!dev_out)
2532 goto out;
2533
2534
2535 if (!(dev_out->flags & IFF_UP) || !__in_dev_get_rcu(dev_out)) {
2536 rth = ERR_PTR(-ENETUNREACH);
2537 goto out;
2538 }
2539 if (ipv4_is_local_multicast(fl4->daddr) ||
2540 ipv4_is_lbcast(fl4->daddr) ||
2541 fl4->flowi4_proto == IPPROTO_IGMP) {
2542 if (!fl4->saddr)
2543 fl4->saddr = inet_select_addr(dev_out, 0,
2544 RT_SCOPE_LINK);
2545 goto make_route;
2546 }
2547 if (!fl4->saddr) {
2548 if (ipv4_is_multicast(fl4->daddr))
2549 fl4->saddr = inet_select_addr(dev_out, 0,
2550 fl4->flowi4_scope);
2551 else if (!fl4->daddr)
2552 fl4->saddr = inet_select_addr(dev_out, 0,
2553 RT_SCOPE_HOST);
2554 }
2555 }
2556
2557 if (!fl4->daddr) {
2558 fl4->daddr = fl4->saddr;
2559 if (!fl4->daddr)
2560 fl4->daddr = fl4->saddr = htonl(INADDR_LOOPBACK);
2561 dev_out = net->loopback_dev;
2562 fl4->flowi4_oif = LOOPBACK_IFINDEX;
2563 res->type = RTN_LOCAL;
2564 flags |= RTCF_LOCAL;
2565 goto make_route;
2566 }
2567
2568 err = fib_lookup(net, fl4, res, 0);
2569 if (err) {
2570 res->fi = NULL;
2571 res->table = NULL;
2572 if (fl4->flowi4_oif &&
2573 (ipv4_is_multicast(fl4->daddr) ||
2574 !netif_index_is_l3_master(net, fl4->flowi4_oif))) {
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593 if (fl4->saddr == 0)
2594 fl4->saddr = inet_select_addr(dev_out, 0,
2595 RT_SCOPE_LINK);
2596 res->type = RTN_UNICAST;
2597 goto make_route;
2598 }
2599 rth = ERR_PTR(err);
2600 goto out;
2601 }
2602
2603 if (res->type == RTN_LOCAL) {
2604 if (!fl4->saddr) {
2605 if (res->fi->fib_prefsrc)
2606 fl4->saddr = res->fi->fib_prefsrc;
2607 else
2608 fl4->saddr = fl4->daddr;
2609 }
2610
2611
2612 dev_out = l3mdev_master_dev_rcu(FIB_RES_DEV(*res)) ? :
2613 net->loopback_dev;
2614
2615
2616
2617
2618 orig_oif = FIB_RES_OIF(*res);
2619
2620 fl4->flowi4_oif = dev_out->ifindex;
2621 flags |= RTCF_LOCAL;
2622 goto make_route;
2623 }
2624
2625 fib_select_path(net, res, fl4, skb);
2626
2627 dev_out = FIB_RES_DEV(*res);
2628 fl4->flowi4_oif = dev_out->ifindex;
2629
2630
2631 make_route:
2632 rth = __mkroute_output(res, fl4, orig_oif, dev_out, flags);
2633
2634 out:
2635 return rth;
2636 }
2637
2638 static struct dst_entry *ipv4_blackhole_dst_check(struct dst_entry *dst, u32 cookie)
2639 {
2640 return NULL;
2641 }
2642
2643 static unsigned int ipv4_blackhole_mtu(const struct dst_entry *dst)
2644 {
2645 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2646
2647 return mtu ? : dst->dev->mtu;
2648 }
2649
2650 static void ipv4_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
2651 struct sk_buff *skb, u32 mtu,
2652 bool confirm_neigh)
2653 {
2654 }
2655
2656 static void ipv4_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
2657 struct sk_buff *skb)
2658 {
2659 }
2660
2661 static u32 *ipv4_rt_blackhole_cow_metrics(struct dst_entry *dst,
2662 unsigned long old)
2663 {
2664 return NULL;
2665 }
2666
2667 static struct dst_ops ipv4_dst_blackhole_ops = {
2668 .family = AF_INET,
2669 .check = ipv4_blackhole_dst_check,
2670 .mtu = ipv4_blackhole_mtu,
2671 .default_advmss = ipv4_default_advmss,
2672 .update_pmtu = ipv4_rt_blackhole_update_pmtu,
2673 .redirect = ipv4_rt_blackhole_redirect,
2674 .cow_metrics = ipv4_rt_blackhole_cow_metrics,
2675 .neigh_lookup = ipv4_neigh_lookup,
2676 };
2677
2678 struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2679 {
2680 struct rtable *ort = (struct rtable *) dst_orig;
2681 struct rtable *rt;
2682
2683 rt = dst_alloc(&ipv4_dst_blackhole_ops, NULL, 1, DST_OBSOLETE_DEAD, 0);
2684 if (rt) {
2685 struct dst_entry *new = &rt->dst;
2686
2687 new->__use = 1;
2688 new->input = dst_discard;
2689 new->output = dst_discard_out;
2690
2691 new->dev = net->loopback_dev;
2692 if (new->dev)
2693 dev_hold(new->dev);
2694
2695 rt->rt_is_input = ort->rt_is_input;
2696 rt->rt_iif = ort->rt_iif;
2697 rt->rt_pmtu = ort->rt_pmtu;
2698 rt->rt_mtu_locked = ort->rt_mtu_locked;
2699
2700 rt->rt_genid = rt_genid_ipv4(net);
2701 rt->rt_flags = ort->rt_flags;
2702 rt->rt_type = ort->rt_type;
2703 rt->rt_uses_gateway = ort->rt_uses_gateway;
2704 rt->rt_gw_family = ort->rt_gw_family;
2705 if (rt->rt_gw_family == AF_INET)
2706 rt->rt_gw4 = ort->rt_gw4;
2707 else if (rt->rt_gw_family == AF_INET6)
2708 rt->rt_gw6 = ort->rt_gw6;
2709
2710 INIT_LIST_HEAD(&rt->rt_uncached);
2711 }
2712
2713 dst_release(dst_orig);
2714
2715 return rt ? &rt->dst : ERR_PTR(-ENOMEM);
2716 }
2717
2718 struct rtable *ip_route_output_flow(struct net *net, struct flowi4 *flp4,
2719 const struct sock *sk)
2720 {
2721 struct rtable *rt = __ip_route_output_key(net, flp4);
2722
2723 if (IS_ERR(rt))
2724 return rt;
2725
2726 if (flp4->flowi4_proto)
2727 rt = (struct rtable *)xfrm_lookup_route(net, &rt->dst,
2728 flowi4_to_flowi(flp4),
2729 sk, 0);
2730
2731 return rt;
2732 }
2733 EXPORT_SYMBOL_GPL(ip_route_output_flow);
2734
2735
2736 static int rt_fill_info(struct net *net, __be32 dst, __be32 src,
2737 struct rtable *rt, u32 table_id, struct flowi4 *fl4,
2738 struct sk_buff *skb, u32 portid, u32 seq,
2739 unsigned int flags)
2740 {
2741 struct rtmsg *r;
2742 struct nlmsghdr *nlh;
2743 unsigned long expires = 0;
2744 u32 error;
2745 u32 metrics[RTAX_MAX];
2746
2747 nlh = nlmsg_put(skb, portid, seq, RTM_NEWROUTE, sizeof(*r), flags);
2748 if (!nlh)
2749 return -EMSGSIZE;
2750
2751 r = nlmsg_data(nlh);
2752 r->rtm_family = AF_INET;
2753 r->rtm_dst_len = 32;
2754 r->rtm_src_len = 0;
2755 r->rtm_tos = fl4 ? fl4->flowi4_tos : 0;
2756 r->rtm_table = table_id < 256 ? table_id : RT_TABLE_COMPAT;
2757 if (nla_put_u32(skb, RTA_TABLE, table_id))
2758 goto nla_put_failure;
2759 r->rtm_type = rt->rt_type;
2760 r->rtm_scope = RT_SCOPE_UNIVERSE;
2761 r->rtm_protocol = RTPROT_UNSPEC;
2762 r->rtm_flags = (rt->rt_flags & ~0xFFFF) | RTM_F_CLONED;
2763 if (rt->rt_flags & RTCF_NOTIFY)
2764 r->rtm_flags |= RTM_F_NOTIFY;
2765 if (IPCB(skb)->flags & IPSKB_DOREDIRECT)
2766 r->rtm_flags |= RTCF_DOREDIRECT;
2767
2768 if (nla_put_in_addr(skb, RTA_DST, dst))
2769 goto nla_put_failure;
2770 if (src) {
2771 r->rtm_src_len = 32;
2772 if (nla_put_in_addr(skb, RTA_SRC, src))
2773 goto nla_put_failure;
2774 }
2775 if (rt->dst.dev &&
2776 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2777 goto nla_put_failure;
2778 #ifdef CONFIG_IP_ROUTE_CLASSID
2779 if (rt->dst.tclassid &&
2780 nla_put_u32(skb, RTA_FLOW, rt->dst.tclassid))
2781 goto nla_put_failure;
2782 #endif
2783 if (fl4 && !rt_is_input_route(rt) &&
2784 fl4->saddr != src) {
2785 if (nla_put_in_addr(skb, RTA_PREFSRC, fl4->saddr))
2786 goto nla_put_failure;
2787 }
2788 if (rt->rt_uses_gateway) {
2789 if (rt->rt_gw_family == AF_INET &&
2790 nla_put_in_addr(skb, RTA_GATEWAY, rt->rt_gw4)) {
2791 goto nla_put_failure;
2792 } else if (rt->rt_gw_family == AF_INET6) {
2793 int alen = sizeof(struct in6_addr);
2794 struct nlattr *nla;
2795 struct rtvia *via;
2796
2797 nla = nla_reserve(skb, RTA_VIA, alen + 2);
2798 if (!nla)
2799 goto nla_put_failure;
2800
2801 via = nla_data(nla);
2802 via->rtvia_family = AF_INET6;
2803 memcpy(via->rtvia_addr, &rt->rt_gw6, alen);
2804 }
2805 }
2806
2807 expires = rt->dst.expires;
2808 if (expires) {
2809 unsigned long now = jiffies;
2810
2811 if (time_before(now, expires))
2812 expires -= now;
2813 else
2814 expires = 0;
2815 }
2816
2817 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
2818 if (rt->rt_pmtu && expires)
2819 metrics[RTAX_MTU - 1] = rt->rt_pmtu;
2820 if (rt->rt_mtu_locked && expires)
2821 metrics[RTAX_LOCK - 1] |= BIT(RTAX_MTU);
2822 if (rtnetlink_put_metrics(skb, metrics) < 0)
2823 goto nla_put_failure;
2824
2825 if (fl4) {
2826 if (fl4->flowi4_mark &&
2827 nla_put_u32(skb, RTA_MARK, fl4->flowi4_mark))
2828 goto nla_put_failure;
2829
2830 if (!uid_eq(fl4->flowi4_uid, INVALID_UID) &&
2831 nla_put_u32(skb, RTA_UID,
2832 from_kuid_munged(current_user_ns(),
2833 fl4->flowi4_uid)))
2834 goto nla_put_failure;
2835
2836 if (rt_is_input_route(rt)) {
2837 #ifdef CONFIG_IP_MROUTE
2838 if (ipv4_is_multicast(dst) &&
2839 !ipv4_is_local_multicast(dst) &&
2840 IPV4_DEVCONF_ALL(net, MC_FORWARDING)) {
2841 int err = ipmr_get_route(net, skb,
2842 fl4->saddr, fl4->daddr,
2843 r, portid);
2844
2845 if (err <= 0) {
2846 if (err == 0)
2847 return 0;
2848 goto nla_put_failure;
2849 }
2850 } else
2851 #endif
2852 if (nla_put_u32(skb, RTA_IIF, fl4->flowi4_iif))
2853 goto nla_put_failure;
2854 }
2855 }
2856
2857 error = rt->dst.error;
2858
2859 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, error) < 0)
2860 goto nla_put_failure;
2861
2862 nlmsg_end(skb, nlh);
2863 return 0;
2864
2865 nla_put_failure:
2866 nlmsg_cancel(skb, nlh);
2867 return -EMSGSIZE;
2868 }
2869
2870 static int fnhe_dump_bucket(struct net *net, struct sk_buff *skb,
2871 struct netlink_callback *cb, u32 table_id,
2872 struct fnhe_hash_bucket *bucket, int genid,
2873 int *fa_index, int fa_start, unsigned int flags)
2874 {
2875 int i;
2876
2877 for (i = 0; i < FNHE_HASH_SIZE; i++) {
2878 struct fib_nh_exception *fnhe;
2879
2880 for (fnhe = rcu_dereference(bucket[i].chain); fnhe;
2881 fnhe = rcu_dereference(fnhe->fnhe_next)) {
2882 struct rtable *rt;
2883 int err;
2884
2885 if (*fa_index < fa_start)
2886 goto next;
2887
2888 if (fnhe->fnhe_genid != genid)
2889 goto next;
2890
2891 if (fnhe->fnhe_expires &&
2892 time_after(jiffies, fnhe->fnhe_expires))
2893 goto next;
2894
2895 rt = rcu_dereference(fnhe->fnhe_rth_input);
2896 if (!rt)
2897 rt = rcu_dereference(fnhe->fnhe_rth_output);
2898 if (!rt)
2899 goto next;
2900
2901 err = rt_fill_info(net, fnhe->fnhe_daddr, 0, rt,
2902 table_id, NULL, skb,
2903 NETLINK_CB(cb->skb).portid,
2904 cb->nlh->nlmsg_seq, flags);
2905 if (err)
2906 return err;
2907 next:
2908 (*fa_index)++;
2909 }
2910 }
2911
2912 return 0;
2913 }
2914
2915 int fib_dump_info_fnhe(struct sk_buff *skb, struct netlink_callback *cb,
2916 u32 table_id, struct fib_info *fi,
2917 int *fa_index, int fa_start, unsigned int flags)
2918 {
2919 struct net *net = sock_net(cb->skb->sk);
2920 int nhsel, genid = fnhe_genid(net);
2921
2922 for (nhsel = 0; nhsel < fib_info_num_path(fi); nhsel++) {
2923 struct fib_nh_common *nhc = fib_info_nhc(fi, nhsel);
2924 struct fnhe_hash_bucket *bucket;
2925 int err;
2926
2927 if (nhc->nhc_flags & RTNH_F_DEAD)
2928 continue;
2929
2930 rcu_read_lock();
2931 bucket = rcu_dereference(nhc->nhc_exceptions);
2932 err = 0;
2933 if (bucket)
2934 err = fnhe_dump_bucket(net, skb, cb, table_id, bucket,
2935 genid, fa_index, fa_start,
2936 flags);
2937 rcu_read_unlock();
2938 if (err)
2939 return err;
2940 }
2941
2942 return 0;
2943 }
2944
2945 static struct sk_buff *inet_rtm_getroute_build_skb(__be32 src, __be32 dst,
2946 u8 ip_proto, __be16 sport,
2947 __be16 dport)
2948 {
2949 struct sk_buff *skb;
2950 struct iphdr *iph;
2951
2952 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2953 if (!skb)
2954 return NULL;
2955
2956
2957
2958
2959 skb_reset_mac_header(skb);
2960 skb_reset_network_header(skb);
2961 skb->protocol = htons(ETH_P_IP);
2962 iph = skb_put(skb, sizeof(struct iphdr));
2963 iph->protocol = ip_proto;
2964 iph->saddr = src;
2965 iph->daddr = dst;
2966 iph->version = 0x4;
2967 iph->frag_off = 0;
2968 iph->ihl = 0x5;
2969 skb_set_transport_header(skb, skb->len);
2970
2971 switch (iph->protocol) {
2972 case IPPROTO_UDP: {
2973 struct udphdr *udph;
2974
2975 udph = skb_put_zero(skb, sizeof(struct udphdr));
2976 udph->source = sport;
2977 udph->dest = dport;
2978 udph->len = sizeof(struct udphdr);
2979 udph->check = 0;
2980 break;
2981 }
2982 case IPPROTO_TCP: {
2983 struct tcphdr *tcph;
2984
2985 tcph = skb_put_zero(skb, sizeof(struct tcphdr));
2986 tcph->source = sport;
2987 tcph->dest = dport;
2988 tcph->doff = sizeof(struct tcphdr) / 4;
2989 tcph->rst = 1;
2990 tcph->check = ~tcp_v4_check(sizeof(struct tcphdr),
2991 src, dst, 0);
2992 break;
2993 }
2994 case IPPROTO_ICMP: {
2995 struct icmphdr *icmph;
2996
2997 icmph = skb_put_zero(skb, sizeof(struct icmphdr));
2998 icmph->type = ICMP_ECHO;
2999 icmph->code = 0;
3000 }
3001 }
3002
3003 return skb;
3004 }
3005
3006 static int inet_rtm_valid_getroute_req(struct sk_buff *skb,
3007 const struct nlmsghdr *nlh,
3008 struct nlattr **tb,
3009 struct netlink_ext_ack *extack)
3010 {
3011 struct rtmsg *rtm;
3012 int i, err;
3013
3014 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
3015 NL_SET_ERR_MSG(extack,
3016 "ipv4: Invalid header for route get request");
3017 return -EINVAL;
3018 }
3019
3020 if (!netlink_strict_get_check(skb))
3021 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
3022 rtm_ipv4_policy, extack);
3023
3024 rtm = nlmsg_data(nlh);
3025 if ((rtm->rtm_src_len && rtm->rtm_src_len != 32) ||
3026 (rtm->rtm_dst_len && rtm->rtm_dst_len != 32) ||
3027 rtm->rtm_table || rtm->rtm_protocol ||
3028 rtm->rtm_scope || rtm->rtm_type) {
3029 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for route get request");
3030 return -EINVAL;
3031 }
3032
3033 if (rtm->rtm_flags & ~(RTM_F_NOTIFY |
3034 RTM_F_LOOKUP_TABLE |
3035 RTM_F_FIB_MATCH)) {
3036 NL_SET_ERR_MSG(extack, "ipv4: Unsupported rtm_flags for route get request");
3037 return -EINVAL;
3038 }
3039
3040 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
3041 rtm_ipv4_policy, extack);
3042 if (err)
3043 return err;
3044
3045 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
3046 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
3047 NL_SET_ERR_MSG(extack, "ipv4: rtm_src_len and rtm_dst_len must be 32 for IPv4");
3048 return -EINVAL;
3049 }
3050
3051 for (i = 0; i <= RTA_MAX; i++) {
3052 if (!tb[i])
3053 continue;
3054
3055 switch (i) {
3056 case RTA_IIF:
3057 case RTA_OIF:
3058 case RTA_SRC:
3059 case RTA_DST:
3060 case RTA_IP_PROTO:
3061 case RTA_SPORT:
3062 case RTA_DPORT:
3063 case RTA_MARK:
3064 case RTA_UID:
3065 break;
3066 default:
3067 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in route get request");
3068 return -EINVAL;
3069 }
3070 }
3071
3072 return 0;
3073 }
3074
3075 static int inet_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3076 struct netlink_ext_ack *extack)
3077 {
3078 struct net *net = sock_net(in_skb->sk);
3079 struct nlattr *tb[RTA_MAX+1];
3080 u32 table_id = RT_TABLE_MAIN;
3081 __be16 sport = 0, dport = 0;
3082 struct fib_result res = {};
3083 u8 ip_proto = IPPROTO_UDP;
3084 struct rtable *rt = NULL;
3085 struct sk_buff *skb;
3086 struct rtmsg *rtm;
3087 struct flowi4 fl4 = {};
3088 __be32 dst = 0;
3089 __be32 src = 0;
3090 kuid_t uid;
3091 u32 iif;
3092 int err;
3093 int mark;
3094
3095 err = inet_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
3096 if (err < 0)
3097 return err;
3098
3099 rtm = nlmsg_data(nlh);
3100 src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0;
3101 dst = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0;
3102 iif = tb[RTA_IIF] ? nla_get_u32(tb[RTA_IIF]) : 0;
3103 mark = tb[RTA_MARK] ? nla_get_u32(tb[RTA_MARK]) : 0;
3104 if (tb[RTA_UID])
3105 uid = make_kuid(current_user_ns(), nla_get_u32(tb[RTA_UID]));
3106 else
3107 uid = (iif ? INVALID_UID : current_uid());
3108
3109 if (tb[RTA_IP_PROTO]) {
3110 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
3111 &ip_proto, AF_INET, extack);
3112 if (err)
3113 return err;
3114 }
3115
3116 if (tb[RTA_SPORT])
3117 sport = nla_get_be16(tb[RTA_SPORT]);
3118
3119 if (tb[RTA_DPORT])
3120 dport = nla_get_be16(tb[RTA_DPORT]);
3121
3122 skb = inet_rtm_getroute_build_skb(src, dst, ip_proto, sport, dport);
3123 if (!skb)
3124 return -ENOBUFS;
3125
3126 fl4.daddr = dst;
3127 fl4.saddr = src;
3128 fl4.flowi4_tos = rtm->rtm_tos;
3129 fl4.flowi4_oif = tb[RTA_OIF] ? nla_get_u32(tb[RTA_OIF]) : 0;
3130 fl4.flowi4_mark = mark;
3131 fl4.flowi4_uid = uid;
3132 if (sport)
3133 fl4.fl4_sport = sport;
3134 if (dport)
3135 fl4.fl4_dport = dport;
3136 fl4.flowi4_proto = ip_proto;
3137
3138 rcu_read_lock();
3139
3140 if (iif) {
3141 struct net_device *dev;
3142
3143 dev = dev_get_by_index_rcu(net, iif);
3144 if (!dev) {
3145 err = -ENODEV;
3146 goto errout_rcu;
3147 }
3148
3149 fl4.flowi4_iif = iif;
3150 skb->dev = dev;
3151 skb->mark = mark;
3152 err = ip_route_input_rcu(skb, dst, src, rtm->rtm_tos,
3153 dev, &res);
3154
3155 rt = skb_rtable(skb);
3156 if (err == 0 && rt->dst.error)
3157 err = -rt->dst.error;
3158 } else {
3159 fl4.flowi4_iif = LOOPBACK_IFINDEX;
3160 skb->dev = net->loopback_dev;
3161 rt = ip_route_output_key_hash_rcu(net, &fl4, &res, skb);
3162 err = 0;
3163 if (IS_ERR(rt))
3164 err = PTR_ERR(rt);
3165 else
3166 skb_dst_set(skb, &rt->dst);
3167 }
3168
3169 if (err)
3170 goto errout_rcu;
3171
3172 if (rtm->rtm_flags & RTM_F_NOTIFY)
3173 rt->rt_flags |= RTCF_NOTIFY;
3174
3175 if (rtm->rtm_flags & RTM_F_LOOKUP_TABLE)
3176 table_id = res.table ? res.table->tb_id : 0;
3177
3178
3179 skb_trim(skb, 0);
3180 skb_reset_network_header(skb);
3181 skb_reset_transport_header(skb);
3182 skb_reset_mac_header(skb);
3183
3184 if (rtm->rtm_flags & RTM_F_FIB_MATCH) {
3185 if (!res.fi) {
3186 err = fib_props[res.type].error;
3187 if (!err)
3188 err = -EHOSTUNREACH;
3189 goto errout_rcu;
3190 }
3191 err = fib_dump_info(skb, NETLINK_CB(in_skb).portid,
3192 nlh->nlmsg_seq, RTM_NEWROUTE, table_id,
3193 rt->rt_type, res.prefix, res.prefixlen,
3194 fl4.flowi4_tos, res.fi, 0);
3195 } else {
3196 err = rt_fill_info(net, dst, src, rt, table_id, &fl4, skb,
3197 NETLINK_CB(in_skb).portid,
3198 nlh->nlmsg_seq, 0);
3199 }
3200 if (err < 0)
3201 goto errout_rcu;
3202
3203 rcu_read_unlock();
3204
3205 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3206
3207 errout_free:
3208 return err;
3209 errout_rcu:
3210 rcu_read_unlock();
3211 kfree_skb(skb);
3212 goto errout_free;
3213 }
3214
3215 void ip_rt_multicast_event(struct in_device *in_dev)
3216 {
3217 rt_cache_flush(dev_net(in_dev->dev));
3218 }
3219
3220 #ifdef CONFIG_SYSCTL
3221 static int ip_rt_gc_interval __read_mostly = 60 * HZ;
3222 static int ip_rt_gc_min_interval __read_mostly = HZ / 2;
3223 static int ip_rt_gc_elasticity __read_mostly = 8;
3224 static int ip_min_valid_pmtu __read_mostly = IPV4_MIN_MTU;
3225
3226 static int ipv4_sysctl_rtcache_flush(struct ctl_table *__ctl, int write,
3227 void __user *buffer,
3228 size_t *lenp, loff_t *ppos)
3229 {
3230 struct net *net = (struct net *)__ctl->extra1;
3231
3232 if (write) {
3233 rt_cache_flush(net);
3234 fnhe_genid_bump(net);
3235 return 0;
3236 }
3237
3238 return -EINVAL;
3239 }
3240
3241 static struct ctl_table ipv4_route_table[] = {
3242 {
3243 .procname = "gc_thresh",
3244 .data = &ipv4_dst_ops.gc_thresh,
3245 .maxlen = sizeof(int),
3246 .mode = 0644,
3247 .proc_handler = proc_dointvec,
3248 },
3249 {
3250 .procname = "max_size",
3251 .data = &ip_rt_max_size,
3252 .maxlen = sizeof(int),
3253 .mode = 0644,
3254 .proc_handler = proc_dointvec,
3255 },
3256 {
3257
3258
3259 .procname = "gc_min_interval",
3260 .data = &ip_rt_gc_min_interval,
3261 .maxlen = sizeof(int),
3262 .mode = 0644,
3263 .proc_handler = proc_dointvec_jiffies,
3264 },
3265 {
3266 .procname = "gc_min_interval_ms",
3267 .data = &ip_rt_gc_min_interval,
3268 .maxlen = sizeof(int),
3269 .mode = 0644,
3270 .proc_handler = proc_dointvec_ms_jiffies,
3271 },
3272 {
3273 .procname = "gc_timeout",
3274 .data = &ip_rt_gc_timeout,
3275 .maxlen = sizeof(int),
3276 .mode = 0644,
3277 .proc_handler = proc_dointvec_jiffies,
3278 },
3279 {
3280 .procname = "gc_interval",
3281 .data = &ip_rt_gc_interval,
3282 .maxlen = sizeof(int),
3283 .mode = 0644,
3284 .proc_handler = proc_dointvec_jiffies,
3285 },
3286 {
3287 .procname = "redirect_load",
3288 .data = &ip_rt_redirect_load,
3289 .maxlen = sizeof(int),
3290 .mode = 0644,
3291 .proc_handler = proc_dointvec,
3292 },
3293 {
3294 .procname = "redirect_number",
3295 .data = &ip_rt_redirect_number,
3296 .maxlen = sizeof(int),
3297 .mode = 0644,
3298 .proc_handler = proc_dointvec,
3299 },
3300 {
3301 .procname = "redirect_silence",
3302 .data = &ip_rt_redirect_silence,
3303 .maxlen = sizeof(int),
3304 .mode = 0644,
3305 .proc_handler = proc_dointvec,
3306 },
3307 {
3308 .procname = "error_cost",
3309 .data = &ip_rt_error_cost,
3310 .maxlen = sizeof(int),
3311 .mode = 0644,
3312 .proc_handler = proc_dointvec,
3313 },
3314 {
3315 .procname = "error_burst",
3316 .data = &ip_rt_error_burst,
3317 .maxlen = sizeof(int),
3318 .mode = 0644,
3319 .proc_handler = proc_dointvec,
3320 },
3321 {
3322 .procname = "gc_elasticity",
3323 .data = &ip_rt_gc_elasticity,
3324 .maxlen = sizeof(int),
3325 .mode = 0644,
3326 .proc_handler = proc_dointvec,
3327 },
3328 {
3329 .procname = "mtu_expires",
3330 .data = &ip_rt_mtu_expires,
3331 .maxlen = sizeof(int),
3332 .mode = 0644,
3333 .proc_handler = proc_dointvec_jiffies,
3334 },
3335 {
3336 .procname = "min_pmtu",
3337 .data = &ip_rt_min_pmtu,
3338 .maxlen = sizeof(int),
3339 .mode = 0644,
3340 .proc_handler = proc_dointvec_minmax,
3341 .extra1 = &ip_min_valid_pmtu,
3342 },
3343 {
3344 .procname = "min_adv_mss",
3345 .data = &ip_rt_min_advmss,
3346 .maxlen = sizeof(int),
3347 .mode = 0644,
3348 .proc_handler = proc_dointvec,
3349 },
3350 { }
3351 };
3352
3353 static const char ipv4_route_flush_procname[] = "flush";
3354
3355 static struct ctl_table ipv4_route_flush_table[] = {
3356 {
3357 .procname = ipv4_route_flush_procname,
3358 .maxlen = sizeof(int),
3359 .mode = 0200,
3360 .proc_handler = ipv4_sysctl_rtcache_flush,
3361 },
3362 { },
3363 };
3364
3365 static __net_init int sysctl_route_net_init(struct net *net)
3366 {
3367 struct ctl_table *tbl;
3368
3369 tbl = ipv4_route_flush_table;
3370 if (!net_eq(net, &init_net)) {
3371 tbl = kmemdup(tbl, sizeof(ipv4_route_flush_table), GFP_KERNEL);
3372 if (!tbl)
3373 goto err_dup;
3374
3375
3376 if (net->user_ns != &init_user_ns) {
3377 if (tbl[0].procname != ipv4_route_flush_procname)
3378 tbl[0].procname = NULL;
3379 }
3380 }
3381 tbl[0].extra1 = net;
3382
3383 net->ipv4.route_hdr = register_net_sysctl(net, "net/ipv4/route", tbl);
3384 if (!net->ipv4.route_hdr)
3385 goto err_reg;
3386 return 0;
3387
3388 err_reg:
3389 if (tbl != ipv4_route_flush_table)
3390 kfree(tbl);
3391 err_dup:
3392 return -ENOMEM;
3393 }
3394
3395 static __net_exit void sysctl_route_net_exit(struct net *net)
3396 {
3397 struct ctl_table *tbl;
3398
3399 tbl = net->ipv4.route_hdr->ctl_table_arg;
3400 unregister_net_sysctl_table(net->ipv4.route_hdr);
3401 BUG_ON(tbl == ipv4_route_flush_table);
3402 kfree(tbl);
3403 }
3404
3405 static __net_initdata struct pernet_operations sysctl_route_ops = {
3406 .init = sysctl_route_net_init,
3407 .exit = sysctl_route_net_exit,
3408 };
3409 #endif
3410
3411 static __net_init int rt_genid_init(struct net *net)
3412 {
3413 atomic_set(&net->ipv4.rt_genid, 0);
3414 atomic_set(&net->fnhe_genid, 0);
3415 atomic_set(&net->ipv4.dev_addr_genid, get_random_int());
3416 return 0;
3417 }
3418
3419 static __net_initdata struct pernet_operations rt_genid_ops = {
3420 .init = rt_genid_init,
3421 };
3422
3423 static int __net_init ipv4_inetpeer_init(struct net *net)
3424 {
3425 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3426
3427 if (!bp)
3428 return -ENOMEM;
3429 inet_peer_base_init(bp);
3430 net->ipv4.peers = bp;
3431 return 0;
3432 }
3433
3434 static void __net_exit ipv4_inetpeer_exit(struct net *net)
3435 {
3436 struct inet_peer_base *bp = net->ipv4.peers;
3437
3438 net->ipv4.peers = NULL;
3439 inetpeer_invalidate_tree(bp);
3440 kfree(bp);
3441 }
3442
3443 static __net_initdata struct pernet_operations ipv4_inetpeer_ops = {
3444 .init = ipv4_inetpeer_init,
3445 .exit = ipv4_inetpeer_exit,
3446 };
3447
3448 #ifdef CONFIG_IP_ROUTE_CLASSID
3449 struct ip_rt_acct __percpu *ip_rt_acct __read_mostly;
3450 #endif
3451
3452 int __init ip_rt_init(void)
3453 {
3454 int cpu;
3455
3456 ip_idents = kmalloc_array(IP_IDENTS_SZ, sizeof(*ip_idents),
3457 GFP_KERNEL);
3458 if (!ip_idents)
3459 panic("IP: failed to allocate ip_idents\n");
3460
3461 prandom_bytes(ip_idents, IP_IDENTS_SZ * sizeof(*ip_idents));
3462
3463 ip_tstamps = kcalloc(IP_IDENTS_SZ, sizeof(*ip_tstamps), GFP_KERNEL);
3464 if (!ip_tstamps)
3465 panic("IP: failed to allocate ip_tstamps\n");
3466
3467 for_each_possible_cpu(cpu) {
3468 struct uncached_list *ul = &per_cpu(rt_uncached_list, cpu);
3469
3470 INIT_LIST_HEAD(&ul->head);
3471 spin_lock_init(&ul->lock);
3472 }
3473 #ifdef CONFIG_IP_ROUTE_CLASSID
3474 ip_rt_acct = __alloc_percpu(256 * sizeof(struct ip_rt_acct), __alignof__(struct ip_rt_acct));
3475 if (!ip_rt_acct)
3476 panic("IP: failed to allocate ip_rt_acct\n");
3477 #endif
3478
3479 ipv4_dst_ops.kmem_cachep =
3480 kmem_cache_create("ip_dst_cache", sizeof(struct rtable), 0,
3481 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
3482
3483 ipv4_dst_blackhole_ops.kmem_cachep = ipv4_dst_ops.kmem_cachep;
3484
3485 if (dst_entries_init(&ipv4_dst_ops) < 0)
3486 panic("IP: failed to allocate ipv4_dst_ops counter\n");
3487
3488 if (dst_entries_init(&ipv4_dst_blackhole_ops) < 0)
3489 panic("IP: failed to allocate ipv4_dst_blackhole_ops counter\n");
3490
3491 ipv4_dst_ops.gc_thresh = ~0;
3492 ip_rt_max_size = INT_MAX;
3493
3494 devinet_init();
3495 ip_fib_init();
3496
3497 if (ip_rt_proc_init())
3498 pr_err("Unable to create route proc files\n");
3499 #ifdef CONFIG_XFRM
3500 xfrm_init();
3501 xfrm4_init();
3502 #endif
3503 rtnl_register(PF_INET, RTM_GETROUTE, inet_rtm_getroute, NULL,
3504 RTNL_FLAG_DOIT_UNLOCKED);
3505
3506 #ifdef CONFIG_SYSCTL
3507 register_pernet_subsys(&sysctl_route_ops);
3508 #endif
3509 register_pernet_subsys(&rt_genid_ops);
3510 register_pernet_subsys(&ipv4_inetpeer_ops);
3511 return 0;
3512 }
3513
3514 #ifdef CONFIG_SYSCTL
3515
3516
3517
3518
3519 void __init ip_static_sysctl_init(void)
3520 {
3521 register_net_sysctl(&init_net, "net/ipv4/route", ipv4_route_table);
3522 }
3523 #endif