• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 RT_GC_TIMEOUT (300*HZ)
117 
118 static int ip_rt_max_size;
119 static int ip_rt_redirect_number __read_mostly	= 9;
120 static int ip_rt_redirect_load __read_mostly	= HZ / 50;
121 static int ip_rt_redirect_silence __read_mostly	= ((HZ / 50) << (9 + 1));
122 static int ip_rt_error_cost __read_mostly	= HZ;
123 static int ip_rt_error_burst __read_mostly	= 5 * HZ;
124 static int ip_rt_mtu_expires __read_mostly	= 10 * 60 * HZ;
125 static int ip_rt_min_pmtu __read_mostly		= 512 + 20 + 20;
126 static int ip_rt_min_advmss __read_mostly	= 256;
127 
128 /*
129  *	Interface to generic destination cache.
130  */
131 
132 static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie);
133 static unsigned int	 ipv4_default_advmss(const struct dst_entry *dst);
134 static unsigned int	 ipv4_mtu(const struct dst_entry *dst);
135 static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst);
136 static void		 ipv4_link_failure(struct sk_buff *skb);
137 static void		 ip_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
138 					   struct sk_buff *skb, u32 mtu);
139 static void		 ip_do_redirect(struct dst_entry *dst, struct sock *sk,
140 					struct sk_buff *skb);
141 static void		ipv4_dst_destroy(struct dst_entry *dst);
142 
ipv4_cow_metrics(struct dst_entry * dst,unsigned long old)143 static u32 *ipv4_cow_metrics(struct dst_entry *dst, unsigned long old)
144 {
145 	WARN_ON(1);
146 	return NULL;
147 }
148 
149 static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst,
150 					   struct sk_buff *skb,
151 					   const void *daddr);
152 
153 static struct dst_ops ipv4_dst_ops = {
154 	.family =		AF_INET,
155 	.protocol =		cpu_to_be16(ETH_P_IP),
156 	.check =		ipv4_dst_check,
157 	.default_advmss =	ipv4_default_advmss,
158 	.mtu =			ipv4_mtu,
159 	.cow_metrics =		ipv4_cow_metrics,
160 	.destroy =		ipv4_dst_destroy,
161 	.negative_advice =	ipv4_negative_advice,
162 	.link_failure =		ipv4_link_failure,
163 	.update_pmtu =		ip_rt_update_pmtu,
164 	.redirect =		ip_do_redirect,
165 	.local_out =		__ip_local_out,
166 	.neigh_lookup =		ipv4_neigh_lookup,
167 };
168 
169 #define ECN_OR_COST(class)	TC_PRIO_##class
170 
171 const __u8 ip_tos2prio[16] = {
172 	TC_PRIO_BESTEFFORT,
173 	ECN_OR_COST(BESTEFFORT),
174 	TC_PRIO_BESTEFFORT,
175 	ECN_OR_COST(BESTEFFORT),
176 	TC_PRIO_BULK,
177 	ECN_OR_COST(BULK),
178 	TC_PRIO_BULK,
179 	ECN_OR_COST(BULK),
180 	TC_PRIO_INTERACTIVE,
181 	ECN_OR_COST(INTERACTIVE),
182 	TC_PRIO_INTERACTIVE,
183 	ECN_OR_COST(INTERACTIVE),
184 	TC_PRIO_INTERACTIVE_BULK,
185 	ECN_OR_COST(INTERACTIVE_BULK),
186 	TC_PRIO_INTERACTIVE_BULK,
187 	ECN_OR_COST(INTERACTIVE_BULK)
188 };
189 EXPORT_SYMBOL(ip_tos2prio);
190 
191 static DEFINE_PER_CPU(struct rt_cache_stat, rt_cache_stat);
192 #define RT_CACHE_STAT_INC(field) raw_cpu_inc(rt_cache_stat.field)
193 
194 #ifdef CONFIG_PROC_FS
rt_cache_seq_start(struct seq_file * seq,loff_t * pos)195 static void *rt_cache_seq_start(struct seq_file *seq, loff_t *pos)
196 {
197 	if (*pos)
198 		return NULL;
199 	return SEQ_START_TOKEN;
200 }
201 
rt_cache_seq_next(struct seq_file * seq,void * v,loff_t * pos)202 static void *rt_cache_seq_next(struct seq_file *seq, void *v, loff_t *pos)
203 {
204 	++*pos;
205 	return NULL;
206 }
207 
rt_cache_seq_stop(struct seq_file * seq,void * v)208 static void rt_cache_seq_stop(struct seq_file *seq, void *v)
209 {
210 }
211 
rt_cache_seq_show(struct seq_file * seq,void * v)212 static int rt_cache_seq_show(struct seq_file *seq, void *v)
213 {
214 	if (v == SEQ_START_TOKEN)
215 		seq_printf(seq, "%-127s\n",
216 			   "Iface\tDestination\tGateway \tFlags\t\tRefCnt\tUse\t"
217 			   "Metric\tSource\t\tMTU\tWindow\tIRTT\tTOS\tHHRef\t"
218 			   "HHUptod\tSpecDst");
219 	return 0;
220 }
221 
222 static const struct seq_operations rt_cache_seq_ops = {
223 	.start  = rt_cache_seq_start,
224 	.next   = rt_cache_seq_next,
225 	.stop   = rt_cache_seq_stop,
226 	.show   = rt_cache_seq_show,
227 };
228 
rt_cache_seq_open(struct inode * inode,struct file * file)229 static int rt_cache_seq_open(struct inode *inode, struct file *file)
230 {
231 	return seq_open(file, &rt_cache_seq_ops);
232 }
233 
234 static const struct file_operations rt_cache_seq_fops = {
235 	.owner	 = THIS_MODULE,
236 	.open	 = rt_cache_seq_open,
237 	.read	 = seq_read,
238 	.llseek	 = seq_lseek,
239 	.release = seq_release,
240 };
241 
242 
rt_cpu_seq_start(struct seq_file * seq,loff_t * pos)243 static void *rt_cpu_seq_start(struct seq_file *seq, loff_t *pos)
244 {
245 	int cpu;
246 
247 	if (*pos == 0)
248 		return SEQ_START_TOKEN;
249 
250 	for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
251 		if (!cpu_possible(cpu))
252 			continue;
253 		*pos = cpu+1;
254 		return &per_cpu(rt_cache_stat, cpu);
255 	}
256 	return NULL;
257 }
258 
rt_cpu_seq_next(struct seq_file * seq,void * v,loff_t * pos)259 static void *rt_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
260 {
261 	int cpu;
262 
263 	for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
264 		if (!cpu_possible(cpu))
265 			continue;
266 		*pos = cpu+1;
267 		return &per_cpu(rt_cache_stat, cpu);
268 	}
269 	return NULL;
270 
271 }
272 
rt_cpu_seq_stop(struct seq_file * seq,void * v)273 static void rt_cpu_seq_stop(struct seq_file *seq, void *v)
274 {
275 
276 }
277 
rt_cpu_seq_show(struct seq_file * seq,void * v)278 static int rt_cpu_seq_show(struct seq_file *seq, void *v)
279 {
280 	struct rt_cache_stat *st = v;
281 
282 	if (v == SEQ_START_TOKEN) {
283 		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");
284 		return 0;
285 	}
286 
287 	seq_printf(seq,"%08x  %08x %08x %08x %08x %08x %08x %08x "
288 		   " %08x %08x %08x %08x %08x %08x %08x %08x %08x \n",
289 		   dst_entries_get_slow(&ipv4_dst_ops),
290 		   0, /* st->in_hit */
291 		   st->in_slow_tot,
292 		   st->in_slow_mc,
293 		   st->in_no_route,
294 		   st->in_brd,
295 		   st->in_martian_dst,
296 		   st->in_martian_src,
297 
298 		   0, /* st->out_hit */
299 		   st->out_slow_tot,
300 		   st->out_slow_mc,
301 
302 		   0, /* st->gc_total */
303 		   0, /* st->gc_ignored */
304 		   0, /* st->gc_goal_miss */
305 		   0, /* st->gc_dst_overflow */
306 		   0, /* st->in_hlist_search */
307 		   0  /* st->out_hlist_search */
308 		);
309 	return 0;
310 }
311 
312 static const struct seq_operations rt_cpu_seq_ops = {
313 	.start  = rt_cpu_seq_start,
314 	.next   = rt_cpu_seq_next,
315 	.stop   = rt_cpu_seq_stop,
316 	.show   = rt_cpu_seq_show,
317 };
318 
319 
rt_cpu_seq_open(struct inode * inode,struct file * file)320 static int rt_cpu_seq_open(struct inode *inode, struct file *file)
321 {
322 	return seq_open(file, &rt_cpu_seq_ops);
323 }
324 
325 static const struct file_operations rt_cpu_seq_fops = {
326 	.owner	 = THIS_MODULE,
327 	.open	 = rt_cpu_seq_open,
328 	.read	 = seq_read,
329 	.llseek	 = seq_lseek,
330 	.release = seq_release,
331 };
332 
333 #ifdef CONFIG_IP_ROUTE_CLASSID
rt_acct_proc_show(struct seq_file * m,void * v)334 static int rt_acct_proc_show(struct seq_file *m, void *v)
335 {
336 	struct ip_rt_acct *dst, *src;
337 	unsigned int i, j;
338 
339 	dst = kcalloc(256, sizeof(struct ip_rt_acct), GFP_KERNEL);
340 	if (!dst)
341 		return -ENOMEM;
342 
343 	for_each_possible_cpu(i) {
344 		src = (struct ip_rt_acct *)per_cpu_ptr(ip_rt_acct, i);
345 		for (j = 0; j < 256; j++) {
346 			dst[j].o_bytes   += src[j].o_bytes;
347 			dst[j].o_packets += src[j].o_packets;
348 			dst[j].i_bytes   += src[j].i_bytes;
349 			dst[j].i_packets += src[j].i_packets;
350 		}
351 	}
352 
353 	seq_write(m, dst, 256 * sizeof(struct ip_rt_acct));
354 	kfree(dst);
355 	return 0;
356 }
357 
rt_acct_proc_open(struct inode * inode,struct file * file)358 static int rt_acct_proc_open(struct inode *inode, struct file *file)
359 {
360 	return single_open(file, rt_acct_proc_show, NULL);
361 }
362 
363 static const struct file_operations rt_acct_proc_fops = {
364 	.owner		= THIS_MODULE,
365 	.open		= rt_acct_proc_open,
366 	.read		= seq_read,
367 	.llseek		= seq_lseek,
368 	.release	= single_release,
369 };
370 #endif
371 
ip_rt_do_proc_init(struct net * net)372 static int __net_init ip_rt_do_proc_init(struct net *net)
373 {
374 	struct proc_dir_entry *pde;
375 
376 	pde = proc_create("rt_cache", S_IRUGO, net->proc_net,
377 			  &rt_cache_seq_fops);
378 	if (!pde)
379 		goto err1;
380 
381 	pde = proc_create("rt_cache", S_IRUGO,
382 			  net->proc_net_stat, &rt_cpu_seq_fops);
383 	if (!pde)
384 		goto err2;
385 
386 #ifdef CONFIG_IP_ROUTE_CLASSID
387 	pde = proc_create("rt_acct", 0, net->proc_net, &rt_acct_proc_fops);
388 	if (!pde)
389 		goto err3;
390 #endif
391 	return 0;
392 
393 #ifdef CONFIG_IP_ROUTE_CLASSID
394 err3:
395 	remove_proc_entry("rt_cache", net->proc_net_stat);
396 #endif
397 err2:
398 	remove_proc_entry("rt_cache", net->proc_net);
399 err1:
400 	return -ENOMEM;
401 }
402 
ip_rt_do_proc_exit(struct net * net)403 static void __net_exit ip_rt_do_proc_exit(struct net *net)
404 {
405 	remove_proc_entry("rt_cache", net->proc_net_stat);
406 	remove_proc_entry("rt_cache", net->proc_net);
407 #ifdef CONFIG_IP_ROUTE_CLASSID
408 	remove_proc_entry("rt_acct", net->proc_net);
409 #endif
410 }
411 
412 static struct pernet_operations ip_rt_proc_ops __net_initdata =  {
413 	.init = ip_rt_do_proc_init,
414 	.exit = ip_rt_do_proc_exit,
415 };
416 
ip_rt_proc_init(void)417 static int __init ip_rt_proc_init(void)
418 {
419 	return register_pernet_subsys(&ip_rt_proc_ops);
420 }
421 
422 #else
ip_rt_proc_init(void)423 static inline int ip_rt_proc_init(void)
424 {
425 	return 0;
426 }
427 #endif /* CONFIG_PROC_FS */
428 
rt_is_expired(const struct rtable * rth)429 static inline bool rt_is_expired(const struct rtable *rth)
430 {
431 	return rth->rt_genid != rt_genid_ipv4(dev_net(rth->dst.dev));
432 }
433 
rt_cache_flush(struct net * net)434 void rt_cache_flush(struct net *net)
435 {
436 	rt_genid_bump_ipv4(net);
437 }
438 
ipv4_neigh_lookup(const struct dst_entry * dst,struct sk_buff * skb,const void * daddr)439 static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst,
440 					   struct sk_buff *skb,
441 					   const void *daddr)
442 {
443 	struct net_device *dev = dst->dev;
444 	const __be32 *pkey = daddr;
445 	const struct rtable *rt;
446 	struct neighbour *n;
447 
448 	rt = (const struct rtable *) dst;
449 	if (rt->rt_gateway)
450 		pkey = (const __be32 *) &rt->rt_gateway;
451 	else if (skb)
452 		pkey = &ip_hdr(skb)->daddr;
453 
454 	n = __ipv4_neigh_lookup(dev, *(__force u32 *)pkey);
455 	if (n)
456 		return n;
457 	return neigh_create(&arp_tbl, pkey, dev);
458 }
459 
460 #define IP_IDENTS_SZ 2048u
461 struct ip_ident_bucket {
462 	atomic_t	id;
463 	u32		stamp32;
464 };
465 
466 static struct ip_ident_bucket *ip_idents __read_mostly;
467 
468 /* In order to protect privacy, we add a perturbation to identifiers
469  * if one generator is seldom used. This makes hard for an attacker
470  * to infer how many packets were sent between two points in time.
471  */
ip_idents_reserve(u32 hash,int segs)472 u32 ip_idents_reserve(u32 hash, int segs)
473 {
474 	struct ip_ident_bucket *bucket = ip_idents + hash % IP_IDENTS_SZ;
475 	u32 old = ACCESS_ONCE(bucket->stamp32);
476 	u32 now = (u32)jiffies;
477 	u32 delta = 0;
478 
479 	if (old != now && cmpxchg(&bucket->stamp32, old, now) == old)
480 		delta = prandom_u32_max(now - old);
481 
482 	return atomic_add_return(segs + delta, &bucket->id) - segs;
483 }
484 EXPORT_SYMBOL(ip_idents_reserve);
485 
__ip_select_ident(struct iphdr * iph,int segs)486 void __ip_select_ident(struct iphdr *iph, int segs)
487 {
488 	static u32 ip_idents_hashrnd __read_mostly;
489 	u32 hash, id;
490 
491 	net_get_random_once(&ip_idents_hashrnd, sizeof(ip_idents_hashrnd));
492 
493 	hash = jhash_3words((__force u32)iph->daddr,
494 			    (__force u32)iph->saddr,
495 			    iph->protocol,
496 			    ip_idents_hashrnd);
497 	id = ip_idents_reserve(hash, segs);
498 	iph->id = htons(id);
499 }
500 EXPORT_SYMBOL(__ip_select_ident);
501 
__build_flow_key(const struct net * net,struct flowi4 * fl4,const struct sock * sk,const struct iphdr * iph,int oif,u8 tos,u8 prot,u32 mark,int flow_flags)502 static void __build_flow_key(const struct net *net, struct flowi4 *fl4,
503 			     const struct sock *sk,
504 			     const struct iphdr *iph,
505 			     int oif, u8 tos,
506 			     u8 prot, u32 mark, int flow_flags)
507 {
508 	if (sk) {
509 		const struct inet_sock *inet = inet_sk(sk);
510 
511 		oif = sk->sk_bound_dev_if;
512 		mark = sk->sk_mark;
513 		tos = RT_CONN_FLAGS(sk);
514 		prot = inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol;
515 	}
516 	flowi4_init_output(fl4, oif, mark, tos,
517 			   RT_SCOPE_UNIVERSE, prot,
518 			   flow_flags,
519 			   iph->daddr, iph->saddr, 0, 0,
520 			   sock_net_uid(net, sk));
521 }
522 
build_skb_flow_key(struct flowi4 * fl4,const struct sk_buff * skb,const struct sock * sk)523 static void build_skb_flow_key(struct flowi4 *fl4, const struct sk_buff *skb,
524 			       const struct sock *sk)
525 {
526 	const struct net *net = dev_net(skb->dev);
527 	const struct iphdr *iph = ip_hdr(skb);
528 	int oif = skb->dev->ifindex;
529 	u8 tos = RT_TOS(iph->tos);
530 	u8 prot = iph->protocol;
531 	u32 mark = skb->mark;
532 
533 	__build_flow_key(net, fl4, sk, iph, oif, tos, prot, mark, 0);
534 }
535 
build_sk_flow_key(struct flowi4 * fl4,const struct sock * sk)536 static void build_sk_flow_key(struct flowi4 *fl4, const struct sock *sk)
537 {
538 	const struct inet_sock *inet = inet_sk(sk);
539 	const struct ip_options_rcu *inet_opt;
540 	__be32 daddr = inet->inet_daddr;
541 
542 	rcu_read_lock();
543 	inet_opt = rcu_dereference(inet->inet_opt);
544 	if (inet_opt && inet_opt->opt.srr)
545 		daddr = inet_opt->opt.faddr;
546 	flowi4_init_output(fl4, sk->sk_bound_dev_if, sk->sk_mark,
547 			   RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE,
548 			   inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol,
549 			   inet_sk_flowi_flags(sk),
550 			   daddr, inet->inet_saddr, 0, 0, sk->sk_uid);
551 	rcu_read_unlock();
552 }
553 
ip_rt_build_flow_key(struct flowi4 * fl4,const struct sock * sk,const struct sk_buff * skb)554 static void ip_rt_build_flow_key(struct flowi4 *fl4, const struct sock *sk,
555 				 const struct sk_buff *skb)
556 {
557 	if (skb)
558 		build_skb_flow_key(fl4, skb, sk);
559 	else
560 		build_sk_flow_key(fl4, sk);
561 }
562 
rt_free(struct rtable * rt)563 static inline void rt_free(struct rtable *rt)
564 {
565 	call_rcu(&rt->dst.rcu_head, dst_rcu_free);
566 }
567 
568 static DEFINE_SPINLOCK(fnhe_lock);
569 
fnhe_flush_routes(struct fib_nh_exception * fnhe)570 static void fnhe_flush_routes(struct fib_nh_exception *fnhe)
571 {
572 	struct rtable *rt;
573 
574 	rt = rcu_dereference(fnhe->fnhe_rth_input);
575 	if (rt) {
576 		RCU_INIT_POINTER(fnhe->fnhe_rth_input, NULL);
577 		rt_free(rt);
578 	}
579 	rt = rcu_dereference(fnhe->fnhe_rth_output);
580 	if (rt) {
581 		RCU_INIT_POINTER(fnhe->fnhe_rth_output, NULL);
582 		rt_free(rt);
583 	}
584 }
585 
fnhe_oldest(struct fnhe_hash_bucket * hash)586 static struct fib_nh_exception *fnhe_oldest(struct fnhe_hash_bucket *hash)
587 {
588 	struct fib_nh_exception *fnhe, *oldest;
589 
590 	oldest = rcu_dereference(hash->chain);
591 	for (fnhe = rcu_dereference(oldest->fnhe_next); fnhe;
592 	     fnhe = rcu_dereference(fnhe->fnhe_next)) {
593 		if (time_before(fnhe->fnhe_stamp, oldest->fnhe_stamp))
594 			oldest = fnhe;
595 	}
596 	fnhe_flush_routes(oldest);
597 	return oldest;
598 }
599 
fnhe_hashfun(__be32 daddr)600 static inline u32 fnhe_hashfun(__be32 daddr)
601 {
602 	static u32 fnhe_hashrnd __read_mostly;
603 	u32 hval;
604 
605 	net_get_random_once(&fnhe_hashrnd, sizeof(fnhe_hashrnd));
606 	hval = jhash_1word((__force u32) daddr, fnhe_hashrnd);
607 	return hash_32(hval, FNHE_HASH_SHIFT);
608 }
609 
fill_route_from_fnhe(struct rtable * rt,struct fib_nh_exception * fnhe)610 static void fill_route_from_fnhe(struct rtable *rt, struct fib_nh_exception *fnhe)
611 {
612 	rt->rt_pmtu = fnhe->fnhe_pmtu;
613 	rt->dst.expires = fnhe->fnhe_expires;
614 
615 	if (fnhe->fnhe_gw) {
616 		rt->rt_flags |= RTCF_REDIRECTED;
617 		rt->rt_gateway = fnhe->fnhe_gw;
618 		rt->rt_uses_gateway = 1;
619 	}
620 }
621 
update_or_create_fnhe(struct fib_nh * nh,__be32 daddr,__be32 gw,u32 pmtu,unsigned long expires)622 static void update_or_create_fnhe(struct fib_nh *nh, __be32 daddr, __be32 gw,
623 				  u32 pmtu, unsigned long expires)
624 {
625 	struct fnhe_hash_bucket *hash;
626 	struct fib_nh_exception *fnhe;
627 	struct rtable *rt;
628 	u32 genid, hval;
629 	unsigned int i;
630 	int depth;
631 
632 	genid = fnhe_genid(dev_net(nh->nh_dev));
633 	hval = fnhe_hashfun(daddr);
634 
635 	spin_lock_bh(&fnhe_lock);
636 
637 	hash = rcu_dereference(nh->nh_exceptions);
638 	if (!hash) {
639 		hash = kzalloc(FNHE_HASH_SIZE * sizeof(*hash), GFP_ATOMIC);
640 		if (!hash)
641 			goto out_unlock;
642 		rcu_assign_pointer(nh->nh_exceptions, hash);
643 	}
644 
645 	hash += hval;
646 
647 	depth = 0;
648 	for (fnhe = rcu_dereference(hash->chain); fnhe;
649 	     fnhe = rcu_dereference(fnhe->fnhe_next)) {
650 		if (fnhe->fnhe_daddr == daddr)
651 			break;
652 		depth++;
653 	}
654 
655 	if (fnhe) {
656 		if (fnhe->fnhe_genid != genid)
657 			fnhe->fnhe_genid = genid;
658 		if (gw)
659 			fnhe->fnhe_gw = gw;
660 		if (pmtu)
661 			fnhe->fnhe_pmtu = pmtu;
662 		fnhe->fnhe_expires = max(1UL, expires);
663 		/* Update all cached dsts too */
664 		rt = rcu_dereference(fnhe->fnhe_rth_input);
665 		if (rt)
666 			fill_route_from_fnhe(rt, fnhe);
667 		rt = rcu_dereference(fnhe->fnhe_rth_output);
668 		if (rt)
669 			fill_route_from_fnhe(rt, fnhe);
670 	} else {
671 		if (depth > FNHE_RECLAIM_DEPTH)
672 			fnhe = fnhe_oldest(hash);
673 		else {
674 			fnhe = kzalloc(sizeof(*fnhe), GFP_ATOMIC);
675 			if (!fnhe)
676 				goto out_unlock;
677 
678 			fnhe->fnhe_next = hash->chain;
679 			rcu_assign_pointer(hash->chain, fnhe);
680 		}
681 		fnhe->fnhe_genid = genid;
682 		fnhe->fnhe_daddr = daddr;
683 		fnhe->fnhe_gw = gw;
684 		fnhe->fnhe_pmtu = pmtu;
685 		fnhe->fnhe_expires = expires;
686 
687 		/* Exception created; mark the cached routes for the nexthop
688 		 * stale, so anyone caching it rechecks if this exception
689 		 * applies to them.
690 		 */
691 		rt = rcu_dereference(nh->nh_rth_input);
692 		if (rt)
693 			rt->dst.obsolete = DST_OBSOLETE_KILL;
694 
695 		for_each_possible_cpu(i) {
696 			struct rtable __rcu **prt;
697 			prt = per_cpu_ptr(nh->nh_pcpu_rth_output, i);
698 			rt = rcu_dereference(*prt);
699 			if (rt)
700 				rt->dst.obsolete = DST_OBSOLETE_KILL;
701 		}
702 	}
703 
704 	fnhe->fnhe_stamp = jiffies;
705 
706 out_unlock:
707 	spin_unlock_bh(&fnhe_lock);
708 }
709 
__ip_do_redirect(struct rtable * rt,struct sk_buff * skb,struct flowi4 * fl4,bool kill_route)710 static void __ip_do_redirect(struct rtable *rt, struct sk_buff *skb, struct flowi4 *fl4,
711 			     bool kill_route)
712 {
713 	__be32 new_gw = icmp_hdr(skb)->un.gateway;
714 	__be32 old_gw = ip_hdr(skb)->saddr;
715 	struct net_device *dev = skb->dev;
716 	struct in_device *in_dev;
717 	struct fib_result res;
718 	struct neighbour *n;
719 	struct net *net;
720 
721 	switch (icmp_hdr(skb)->code & 7) {
722 	case ICMP_REDIR_NET:
723 	case ICMP_REDIR_NETTOS:
724 	case ICMP_REDIR_HOST:
725 	case ICMP_REDIR_HOSTTOS:
726 		break;
727 
728 	default:
729 		return;
730 	}
731 
732 	if (rt->rt_gateway != old_gw)
733 		return;
734 
735 	in_dev = __in_dev_get_rcu(dev);
736 	if (!in_dev)
737 		return;
738 
739 	net = dev_net(dev);
740 	if (new_gw == old_gw || !IN_DEV_RX_REDIRECTS(in_dev) ||
741 	    ipv4_is_multicast(new_gw) || ipv4_is_lbcast(new_gw) ||
742 	    ipv4_is_zeronet(new_gw))
743 		goto reject_redirect;
744 
745 	if (!IN_DEV_SHARED_MEDIA(in_dev)) {
746 		if (!inet_addr_onlink(in_dev, new_gw, old_gw))
747 			goto reject_redirect;
748 		if (IN_DEV_SEC_REDIRECTS(in_dev) && ip_fib_check_default(new_gw, dev))
749 			goto reject_redirect;
750 	} else {
751 		if (inet_addr_type(net, new_gw) != RTN_UNICAST)
752 			goto reject_redirect;
753 	}
754 
755 	n = ipv4_neigh_lookup(&rt->dst, NULL, &new_gw);
756 	if (!IS_ERR(n)) {
757 		if (!(n->nud_state & NUD_VALID)) {
758 			neigh_event_send(n, NULL);
759 		} else {
760 			if (fib_lookup(net, fl4, &res) == 0) {
761 				struct fib_nh *nh = &FIB_RES_NH(res);
762 
763 				update_or_create_fnhe(nh, fl4->daddr, new_gw,
764 						      0, 0);
765 			}
766 			if (kill_route)
767 				rt->dst.obsolete = DST_OBSOLETE_KILL;
768 			call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, n);
769 		}
770 		neigh_release(n);
771 	}
772 	return;
773 
774 reject_redirect:
775 #ifdef CONFIG_IP_ROUTE_VERBOSE
776 	if (IN_DEV_LOG_MARTIANS(in_dev)) {
777 		const struct iphdr *iph = (const struct iphdr *) skb->data;
778 		__be32 daddr = iph->daddr;
779 		__be32 saddr = iph->saddr;
780 
781 		net_info_ratelimited("Redirect from %pI4 on %s about %pI4 ignored\n"
782 				     "  Advised path = %pI4 -> %pI4\n",
783 				     &old_gw, dev->name, &new_gw,
784 				     &saddr, &daddr);
785 	}
786 #endif
787 	;
788 }
789 
ip_do_redirect(struct dst_entry * dst,struct sock * sk,struct sk_buff * skb)790 static void ip_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
791 {
792 	struct rtable *rt;
793 	struct flowi4 fl4;
794 	const struct iphdr *iph = (const struct iphdr *) skb->data;
795 	struct net *net = dev_net(skb->dev);
796 	int oif = skb->dev->ifindex;
797 	u8 tos = RT_TOS(iph->tos);
798 	u8 prot = iph->protocol;
799 	u32 mark = skb->mark;
800 
801 	rt = (struct rtable *) dst;
802 
803 	__build_flow_key(net, &fl4, sk, iph, oif, tos, prot, mark, 0);
804 	__ip_do_redirect(rt, skb, &fl4, true);
805 }
806 
ipv4_negative_advice(struct dst_entry * dst)807 static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst)
808 {
809 	struct rtable *rt = (struct rtable *)dst;
810 	struct dst_entry *ret = dst;
811 
812 	if (rt) {
813 		if (dst->obsolete > 0) {
814 			ip_rt_put(rt);
815 			ret = NULL;
816 		} else if ((rt->rt_flags & RTCF_REDIRECTED) ||
817 			   rt->dst.expires) {
818 			ip_rt_put(rt);
819 			ret = NULL;
820 		}
821 	}
822 	return ret;
823 }
824 
825 /*
826  * Algorithm:
827  *	1. The first ip_rt_redirect_number redirects are sent
828  *	   with exponential backoff, then we stop sending them at all,
829  *	   assuming that the host ignores our redirects.
830  *	2. If we did not see packets requiring redirects
831  *	   during ip_rt_redirect_silence, we assume that the host
832  *	   forgot redirected route and start to send redirects again.
833  *
834  * This algorithm is much cheaper and more intelligent than dumb load limiting
835  * in icmp.c.
836  *
837  * NOTE. Do not forget to inhibit load limiting for redirects (redundant)
838  * and "frag. need" (breaks PMTU discovery) in icmp.c.
839  */
840 
ip_rt_send_redirect(struct sk_buff * skb)841 void ip_rt_send_redirect(struct sk_buff *skb)
842 {
843 	struct rtable *rt = skb_rtable(skb);
844 	struct in_device *in_dev;
845 	struct inet_peer *peer;
846 	struct net *net;
847 	int log_martians;
848 
849 	rcu_read_lock();
850 	in_dev = __in_dev_get_rcu(rt->dst.dev);
851 	if (!in_dev || !IN_DEV_TX_REDIRECTS(in_dev)) {
852 		rcu_read_unlock();
853 		return;
854 	}
855 	log_martians = IN_DEV_LOG_MARTIANS(in_dev);
856 	rcu_read_unlock();
857 
858 	net = dev_net(rt->dst.dev);
859 	peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, 1);
860 	if (!peer) {
861 		icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST,
862 			  rt_nexthop(rt, ip_hdr(skb)->daddr));
863 		return;
864 	}
865 
866 	/* No redirected packets during ip_rt_redirect_silence;
867 	 * reset the algorithm.
868 	 */
869 	if (time_after(jiffies, peer->rate_last + ip_rt_redirect_silence))
870 		peer->rate_tokens = 0;
871 
872 	/* Too many ignored redirects; do not send anything
873 	 * set dst.rate_last to the last seen redirected packet.
874 	 */
875 	if (peer->rate_tokens >= ip_rt_redirect_number) {
876 		peer->rate_last = jiffies;
877 		goto out_put_peer;
878 	}
879 
880 	/* Check for load limit; set rate_last to the latest sent
881 	 * redirect.
882 	 */
883 	if (peer->rate_tokens == 0 ||
884 	    time_after(jiffies,
885 		       (peer->rate_last +
886 			(ip_rt_redirect_load << peer->rate_tokens)))) {
887 		__be32 gw = rt_nexthop(rt, ip_hdr(skb)->daddr);
888 
889 		icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, gw);
890 		peer->rate_last = jiffies;
891 		++peer->rate_tokens;
892 #ifdef CONFIG_IP_ROUTE_VERBOSE
893 		if (log_martians &&
894 		    peer->rate_tokens == ip_rt_redirect_number)
895 			net_warn_ratelimited("host %pI4/if%d ignores redirects for %pI4 to %pI4\n",
896 					     &ip_hdr(skb)->saddr, inet_iif(skb),
897 					     &ip_hdr(skb)->daddr, &gw);
898 #endif
899 	}
900 out_put_peer:
901 	inet_putpeer(peer);
902 }
903 
ip_error(struct sk_buff * skb)904 static int ip_error(struct sk_buff *skb)
905 {
906 	struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
907 	struct rtable *rt = skb_rtable(skb);
908 	struct inet_peer *peer;
909 	unsigned long now;
910 	struct net *net;
911 	bool send;
912 	int code;
913 
914 	/* IP on this device is disabled. */
915 	if (!in_dev)
916 		goto out;
917 
918 	net = dev_net(rt->dst.dev);
919 	if (!IN_DEV_FORWARD(in_dev)) {
920 		switch (rt->dst.error) {
921 		case EHOSTUNREACH:
922 			IP_INC_STATS_BH(net, IPSTATS_MIB_INADDRERRORS);
923 			break;
924 
925 		case ENETUNREACH:
926 			IP_INC_STATS_BH(net, IPSTATS_MIB_INNOROUTES);
927 			break;
928 		}
929 		goto out;
930 	}
931 
932 	switch (rt->dst.error) {
933 	case EINVAL:
934 	default:
935 		goto out;
936 	case EHOSTUNREACH:
937 		code = ICMP_HOST_UNREACH;
938 		break;
939 	case ENETUNREACH:
940 		code = ICMP_NET_UNREACH;
941 		IP_INC_STATS_BH(net, IPSTATS_MIB_INNOROUTES);
942 		break;
943 	case EACCES:
944 		code = ICMP_PKT_FILTERED;
945 		break;
946 	}
947 
948 	peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, 1);
949 
950 	send = true;
951 	if (peer) {
952 		now = jiffies;
953 		peer->rate_tokens += now - peer->rate_last;
954 		if (peer->rate_tokens > ip_rt_error_burst)
955 			peer->rate_tokens = ip_rt_error_burst;
956 		peer->rate_last = now;
957 		if (peer->rate_tokens >= ip_rt_error_cost)
958 			peer->rate_tokens -= ip_rt_error_cost;
959 		else
960 			send = false;
961 		inet_putpeer(peer);
962 	}
963 	if (send)
964 		icmp_send(skb, ICMP_DEST_UNREACH, code, 0);
965 
966 out:	kfree_skb(skb);
967 	return 0;
968 }
969 
__ip_rt_update_pmtu(struct rtable * rt,struct flowi4 * fl4,u32 mtu)970 static void __ip_rt_update_pmtu(struct rtable *rt, struct flowi4 *fl4, u32 mtu)
971 {
972 	struct dst_entry *dst = &rt->dst;
973 	struct fib_result res;
974 
975 	if (dst_metric_locked(dst, RTAX_MTU))
976 		return;
977 
978 	if (dst->dev->mtu < mtu)
979 		return;
980 
981 	if (mtu < ip_rt_min_pmtu)
982 		mtu = ip_rt_min_pmtu;
983 
984 	if (rt->rt_pmtu == mtu &&
985 	    time_before(jiffies, dst->expires - ip_rt_mtu_expires / 2))
986 		return;
987 
988 	rcu_read_lock();
989 	if (fib_lookup(dev_net(dst->dev), fl4, &res) == 0) {
990 		struct fib_nh *nh = &FIB_RES_NH(res);
991 
992 		update_or_create_fnhe(nh, fl4->daddr, 0, mtu,
993 				      jiffies + ip_rt_mtu_expires);
994 	}
995 	rcu_read_unlock();
996 }
997 
ip_rt_update_pmtu(struct dst_entry * dst,struct sock * sk,struct sk_buff * skb,u32 mtu)998 static void ip_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
999 			      struct sk_buff *skb, u32 mtu)
1000 {
1001 	struct rtable *rt = (struct rtable *) dst;
1002 	struct flowi4 fl4;
1003 
1004 	ip_rt_build_flow_key(&fl4, sk, skb);
1005 	__ip_rt_update_pmtu(rt, &fl4, mtu);
1006 }
1007 
ipv4_update_pmtu(struct sk_buff * skb,struct net * net,u32 mtu,int oif,u32 mark,u8 protocol,int flow_flags)1008 void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu,
1009 		      int oif, u32 mark, u8 protocol, int flow_flags)
1010 {
1011 	const struct iphdr *iph = (const struct iphdr *) skb->data;
1012 	struct flowi4 fl4;
1013 	struct rtable *rt;
1014 
1015 	if (!mark)
1016 		mark = IP4_REPLY_MARK(net, skb->mark);
1017 
1018 	__build_flow_key(net, &fl4, NULL, iph, oif,
1019 			 RT_TOS(iph->tos), protocol, mark, flow_flags);
1020 	rt = __ip_route_output_key(net, &fl4);
1021 	if (!IS_ERR(rt)) {
1022 		__ip_rt_update_pmtu(rt, &fl4, mtu);
1023 		ip_rt_put(rt);
1024 	}
1025 }
1026 EXPORT_SYMBOL_GPL(ipv4_update_pmtu);
1027 
__ipv4_sk_update_pmtu(struct sk_buff * skb,struct sock * sk,u32 mtu)1028 static void __ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu)
1029 {
1030 	const struct iphdr *iph = (const struct iphdr *) skb->data;
1031 	struct flowi4 fl4;
1032 	struct rtable *rt;
1033 
1034 	__build_flow_key(sock_net(sk), &fl4, sk, iph, 0, 0, 0, 0, 0);
1035 
1036 	if (!fl4.flowi4_mark)
1037 		fl4.flowi4_mark = IP4_REPLY_MARK(sock_net(sk), skb->mark);
1038 
1039 	rt = __ip_route_output_key(sock_net(sk), &fl4);
1040 	if (!IS_ERR(rt)) {
1041 		__ip_rt_update_pmtu(rt, &fl4, mtu);
1042 		ip_rt_put(rt);
1043 	}
1044 }
1045 
ipv4_sk_update_pmtu(struct sk_buff * skb,struct sock * sk,u32 mtu)1046 void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu)
1047 {
1048 	const struct iphdr *iph = (const struct iphdr *) skb->data;
1049 	struct flowi4 fl4;
1050 	struct rtable *rt;
1051 	struct dst_entry *odst = NULL;
1052 	bool new = false;
1053 	struct net *net = sock_net(sk);
1054 
1055 	bh_lock_sock(sk);
1056 
1057 	if (!ip_sk_accept_pmtu(sk))
1058 		goto out;
1059 
1060 	odst = sk_dst_get(sk);
1061 
1062 	if (sock_owned_by_user(sk) || !odst) {
1063 		__ipv4_sk_update_pmtu(skb, sk, mtu);
1064 		goto out;
1065 	}
1066 
1067 	__build_flow_key(net, &fl4, sk, iph, 0, 0, 0, 0, 0);
1068 
1069 	rt = (struct rtable *)odst;
1070 	if (odst->obsolete && odst->ops->check(odst, 0) == NULL) {
1071 		rt = ip_route_output_flow(sock_net(sk), &fl4, sk);
1072 		if (IS_ERR(rt))
1073 			goto out;
1074 
1075 		new = true;
1076 	}
1077 
1078 	__ip_rt_update_pmtu((struct rtable *) rt->dst.path, &fl4, mtu);
1079 
1080 	if (!dst_check(&rt->dst, 0)) {
1081 		if (new)
1082 			dst_release(&rt->dst);
1083 
1084 		rt = ip_route_output_flow(sock_net(sk), &fl4, sk);
1085 		if (IS_ERR(rt))
1086 			goto out;
1087 
1088 		new = true;
1089 	}
1090 
1091 	if (new)
1092 		sk_dst_set(sk, &rt->dst);
1093 
1094 out:
1095 	bh_unlock_sock(sk);
1096 	dst_release(odst);
1097 }
1098 EXPORT_SYMBOL_GPL(ipv4_sk_update_pmtu);
1099 
ipv4_redirect(struct sk_buff * skb,struct net * net,int oif,u32 mark,u8 protocol,int flow_flags)1100 void ipv4_redirect(struct sk_buff *skb, struct net *net,
1101 		   int oif, u32 mark, u8 protocol, int flow_flags)
1102 {
1103 	const struct iphdr *iph = (const struct iphdr *) skb->data;
1104 	struct flowi4 fl4;
1105 	struct rtable *rt;
1106 
1107 	__build_flow_key(net, &fl4, NULL, iph, oif,
1108 			 RT_TOS(iph->tos), protocol, mark, flow_flags);
1109 	rt = __ip_route_output_key(net, &fl4);
1110 	if (!IS_ERR(rt)) {
1111 		__ip_do_redirect(rt, skb, &fl4, false);
1112 		ip_rt_put(rt);
1113 	}
1114 }
1115 EXPORT_SYMBOL_GPL(ipv4_redirect);
1116 
ipv4_sk_redirect(struct sk_buff * skb,struct sock * sk)1117 void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk)
1118 {
1119 	const struct iphdr *iph = (const struct iphdr *) skb->data;
1120 	struct flowi4 fl4;
1121 	struct rtable *rt;
1122 	struct net *net = sock_net(sk);
1123 
1124 	__build_flow_key(net, &fl4, sk, iph, 0, 0, 0, 0, 0);
1125 	rt = __ip_route_output_key(net, &fl4);
1126 	if (!IS_ERR(rt)) {
1127 		__ip_do_redirect(rt, skb, &fl4, false);
1128 		ip_rt_put(rt);
1129 	}
1130 }
1131 EXPORT_SYMBOL_GPL(ipv4_sk_redirect);
1132 
ipv4_dst_check(struct dst_entry * dst,u32 cookie)1133 static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie)
1134 {
1135 	struct rtable *rt = (struct rtable *) dst;
1136 
1137 	/* All IPV4 dsts are created with ->obsolete set to the value
1138 	 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1139 	 * into this function always.
1140 	 *
1141 	 * When a PMTU/redirect information update invalidates a route,
1142 	 * this is indicated by setting obsolete to DST_OBSOLETE_KILL or
1143 	 * DST_OBSOLETE_DEAD by dst_free().
1144 	 */
1145 	if (dst->obsolete != DST_OBSOLETE_FORCE_CHK || rt_is_expired(rt))
1146 		return NULL;
1147 	return dst;
1148 }
1149 
ipv4_link_failure(struct sk_buff * skb)1150 static void ipv4_link_failure(struct sk_buff *skb)
1151 {
1152 	struct rtable *rt;
1153 
1154 	icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0);
1155 
1156 	rt = skb_rtable(skb);
1157 	if (rt)
1158 		dst_set_expires(&rt->dst, 0);
1159 }
1160 
ip_rt_bug(struct sock * sk,struct sk_buff * skb)1161 static int ip_rt_bug(struct sock *sk, struct sk_buff *skb)
1162 {
1163 	pr_debug("%s: %pI4 -> %pI4, %s\n",
1164 		 __func__, &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr,
1165 		 skb->dev ? skb->dev->name : "?");
1166 	kfree_skb(skb);
1167 	WARN_ON(1);
1168 	return 0;
1169 }
1170 
1171 /*
1172    We do not cache source address of outgoing interface,
1173    because it is used only by IP RR, TS and SRR options,
1174    so that it out of fast path.
1175 
1176    BTW remember: "addr" is allowed to be not aligned
1177    in IP options!
1178  */
1179 
ip_rt_get_source(u8 * addr,struct sk_buff * skb,struct rtable * rt)1180 void ip_rt_get_source(u8 *addr, struct sk_buff *skb, struct rtable *rt)
1181 {
1182 	__be32 src;
1183 
1184 	if (rt_is_output_route(rt))
1185 		src = ip_hdr(skb)->saddr;
1186 	else {
1187 		struct fib_result res;
1188 		struct flowi4 fl4;
1189 		struct iphdr *iph;
1190 
1191 		iph = ip_hdr(skb);
1192 
1193 		memset(&fl4, 0, sizeof(fl4));
1194 		fl4.daddr = iph->daddr;
1195 		fl4.saddr = iph->saddr;
1196 		fl4.flowi4_tos = RT_TOS(iph->tos);
1197 		fl4.flowi4_oif = rt->dst.dev->ifindex;
1198 		fl4.flowi4_iif = skb->dev->ifindex;
1199 		fl4.flowi4_mark = skb->mark;
1200 
1201 		rcu_read_lock();
1202 		if (fib_lookup(dev_net(rt->dst.dev), &fl4, &res) == 0)
1203 			src = FIB_RES_PREFSRC(dev_net(rt->dst.dev), res);
1204 		else
1205 			src = inet_select_addr(rt->dst.dev,
1206 					       rt_nexthop(rt, iph->daddr),
1207 					       RT_SCOPE_UNIVERSE);
1208 		rcu_read_unlock();
1209 	}
1210 	memcpy(addr, &src, 4);
1211 }
1212 
1213 #ifdef CONFIG_IP_ROUTE_CLASSID
set_class_tag(struct rtable * rt,u32 tag)1214 static void set_class_tag(struct rtable *rt, u32 tag)
1215 {
1216 	if (!(rt->dst.tclassid & 0xFFFF))
1217 		rt->dst.tclassid |= tag & 0xFFFF;
1218 	if (!(rt->dst.tclassid & 0xFFFF0000))
1219 		rt->dst.tclassid |= tag & 0xFFFF0000;
1220 }
1221 #endif
1222 
ipv4_default_advmss(const struct dst_entry * dst)1223 static unsigned int ipv4_default_advmss(const struct dst_entry *dst)
1224 {
1225 	unsigned int advmss = dst_metric_raw(dst, RTAX_ADVMSS);
1226 
1227 	if (advmss == 0) {
1228 		advmss = max_t(unsigned int, dst->dev->mtu - 40,
1229 			       ip_rt_min_advmss);
1230 		if (advmss > 65535 - 40)
1231 			advmss = 65535 - 40;
1232 	}
1233 	return advmss;
1234 }
1235 
ipv4_mtu(const struct dst_entry * dst)1236 static unsigned int ipv4_mtu(const struct dst_entry *dst)
1237 {
1238 	const struct rtable *rt = (const struct rtable *) dst;
1239 	unsigned int mtu = rt->rt_pmtu;
1240 
1241 	if (!mtu || time_after_eq(jiffies, rt->dst.expires))
1242 		mtu = dst_metric_raw(dst, RTAX_MTU);
1243 
1244 	if (mtu)
1245 		return mtu;
1246 
1247 	mtu = dst->dev->mtu;
1248 
1249 	if (unlikely(dst_metric_locked(dst, RTAX_MTU))) {
1250 		if (rt->rt_uses_gateway && mtu > 576)
1251 			mtu = 576;
1252 	}
1253 
1254 	return min_t(unsigned int, mtu, IP_MAX_MTU);
1255 }
1256 
find_exception(struct fib_nh * nh,__be32 daddr)1257 static struct fib_nh_exception *find_exception(struct fib_nh *nh, __be32 daddr)
1258 {
1259 	struct fnhe_hash_bucket *hash = rcu_dereference(nh->nh_exceptions);
1260 	struct fib_nh_exception *fnhe;
1261 	u32 hval;
1262 
1263 	if (!hash)
1264 		return NULL;
1265 
1266 	hval = fnhe_hashfun(daddr);
1267 
1268 	for (fnhe = rcu_dereference(hash[hval].chain); fnhe;
1269 	     fnhe = rcu_dereference(fnhe->fnhe_next)) {
1270 		if (fnhe->fnhe_daddr == daddr)
1271 			return fnhe;
1272 	}
1273 	return NULL;
1274 }
1275 
rt_bind_exception(struct rtable * rt,struct fib_nh_exception * fnhe,__be32 daddr)1276 static bool rt_bind_exception(struct rtable *rt, struct fib_nh_exception *fnhe,
1277 			      __be32 daddr)
1278 {
1279 	bool ret = false;
1280 
1281 	spin_lock_bh(&fnhe_lock);
1282 
1283 	if (daddr == fnhe->fnhe_daddr) {
1284 		struct rtable __rcu **porig;
1285 		struct rtable *orig;
1286 		int genid = fnhe_genid(dev_net(rt->dst.dev));
1287 
1288 		if (rt_is_input_route(rt))
1289 			porig = &fnhe->fnhe_rth_input;
1290 		else
1291 			porig = &fnhe->fnhe_rth_output;
1292 		orig = rcu_dereference(*porig);
1293 
1294 		if (fnhe->fnhe_genid != genid) {
1295 			fnhe->fnhe_genid = genid;
1296 			fnhe->fnhe_gw = 0;
1297 			fnhe->fnhe_pmtu = 0;
1298 			fnhe->fnhe_expires = 0;
1299 			fnhe_flush_routes(fnhe);
1300 			orig = NULL;
1301 		}
1302 		fill_route_from_fnhe(rt, fnhe);
1303 		if (!rt->rt_gateway)
1304 			rt->rt_gateway = daddr;
1305 
1306 		if (!(rt->dst.flags & DST_NOCACHE)) {
1307 			rcu_assign_pointer(*porig, rt);
1308 			if (orig)
1309 				rt_free(orig);
1310 			ret = true;
1311 		}
1312 
1313 		fnhe->fnhe_stamp = jiffies;
1314 	}
1315 	spin_unlock_bh(&fnhe_lock);
1316 
1317 	return ret;
1318 }
1319 
rt_cache_route(struct fib_nh * nh,struct rtable * rt)1320 static bool rt_cache_route(struct fib_nh *nh, struct rtable *rt)
1321 {
1322 	struct rtable *orig, *prev, **p;
1323 	bool ret = true;
1324 
1325 	if (rt_is_input_route(rt)) {
1326 		p = (struct rtable **)&nh->nh_rth_input;
1327 	} else {
1328 		p = (struct rtable **)raw_cpu_ptr(nh->nh_pcpu_rth_output);
1329 	}
1330 	orig = *p;
1331 
1332 	prev = cmpxchg(p, orig, rt);
1333 	if (prev == orig) {
1334 		if (orig)
1335 			rt_free(orig);
1336 	} else
1337 		ret = false;
1338 
1339 	return ret;
1340 }
1341 
1342 static DEFINE_SPINLOCK(rt_uncached_lock);
1343 static LIST_HEAD(rt_uncached_list);
1344 
rt_add_uncached_list(struct rtable * rt)1345 static void rt_add_uncached_list(struct rtable *rt)
1346 {
1347 	spin_lock_bh(&rt_uncached_lock);
1348 	list_add_tail(&rt->rt_uncached, &rt_uncached_list);
1349 	spin_unlock_bh(&rt_uncached_lock);
1350 }
1351 
ipv4_dst_destroy(struct dst_entry * dst)1352 static void ipv4_dst_destroy(struct dst_entry *dst)
1353 {
1354 	struct rtable *rt = (struct rtable *) dst;
1355 
1356 	if (!list_empty(&rt->rt_uncached)) {
1357 		spin_lock_bh(&rt_uncached_lock);
1358 		list_del(&rt->rt_uncached);
1359 		spin_unlock_bh(&rt_uncached_lock);
1360 	}
1361 }
1362 
rt_flush_dev(struct net_device * dev)1363 void rt_flush_dev(struct net_device *dev)
1364 {
1365 	if (!list_empty(&rt_uncached_list)) {
1366 		struct net *net = dev_net(dev);
1367 		struct rtable *rt;
1368 
1369 		spin_lock_bh(&rt_uncached_lock);
1370 		list_for_each_entry(rt, &rt_uncached_list, rt_uncached) {
1371 			if (rt->dst.dev != dev)
1372 				continue;
1373 			rt->dst.dev = net->loopback_dev;
1374 			dev_hold(rt->dst.dev);
1375 			dev_put(dev);
1376 		}
1377 		spin_unlock_bh(&rt_uncached_lock);
1378 	}
1379 }
1380 
rt_cache_valid(const struct rtable * rt)1381 static bool rt_cache_valid(const struct rtable *rt)
1382 {
1383 	return	rt &&
1384 		rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1385 		!rt_is_expired(rt);
1386 }
1387 
rt_set_nexthop(struct rtable * rt,__be32 daddr,const struct fib_result * res,struct fib_nh_exception * fnhe,struct fib_info * fi,u16 type,u32 itag)1388 static void rt_set_nexthop(struct rtable *rt, __be32 daddr,
1389 			   const struct fib_result *res,
1390 			   struct fib_nh_exception *fnhe,
1391 			   struct fib_info *fi, u16 type, u32 itag)
1392 {
1393 	bool cached = false;
1394 
1395 	if (fi) {
1396 		struct fib_nh *nh = &FIB_RES_NH(*res);
1397 
1398 		if (nh->nh_gw && nh->nh_scope == RT_SCOPE_LINK) {
1399 			rt->rt_gateway = nh->nh_gw;
1400 			rt->rt_uses_gateway = 1;
1401 		}
1402 		dst_init_metrics(&rt->dst, fi->fib_metrics, true);
1403 #ifdef CONFIG_IP_ROUTE_CLASSID
1404 		rt->dst.tclassid = nh->nh_tclassid;
1405 #endif
1406 		if (unlikely(fnhe))
1407 			cached = rt_bind_exception(rt, fnhe, daddr);
1408 		else if (!(rt->dst.flags & DST_NOCACHE))
1409 			cached = rt_cache_route(nh, rt);
1410 		if (unlikely(!cached)) {
1411 			/* Routes we intend to cache in nexthop exception or
1412 			 * FIB nexthop have the DST_NOCACHE bit clear.
1413 			 * However, if we are unsuccessful at storing this
1414 			 * route into the cache we really need to set it.
1415 			 */
1416 			rt->dst.flags |= DST_NOCACHE;
1417 			if (!rt->rt_gateway)
1418 				rt->rt_gateway = daddr;
1419 			rt_add_uncached_list(rt);
1420 		}
1421 	} else
1422 		rt_add_uncached_list(rt);
1423 
1424 #ifdef CONFIG_IP_ROUTE_CLASSID
1425 #ifdef CONFIG_IP_MULTIPLE_TABLES
1426 	set_class_tag(rt, res->tclassid);
1427 #endif
1428 	set_class_tag(rt, itag);
1429 #endif
1430 }
1431 
rt_dst_alloc(struct net_device * dev,bool nopolicy,bool noxfrm,bool will_cache)1432 static struct rtable *rt_dst_alloc(struct net_device *dev,
1433 				   bool nopolicy, bool noxfrm, bool will_cache)
1434 {
1435 	return dst_alloc(&ipv4_dst_ops, dev, 1, DST_OBSOLETE_FORCE_CHK,
1436 			 (will_cache ? 0 : (DST_HOST | DST_NOCACHE)) |
1437 			 (nopolicy ? DST_NOPOLICY : 0) |
1438 			 (noxfrm ? DST_NOXFRM : 0));
1439 }
1440 
1441 /* called in rcu_read_lock() section */
ip_route_input_mc(struct sk_buff * skb,__be32 daddr,__be32 saddr,u8 tos,struct net_device * dev,int our)1442 static int ip_route_input_mc(struct sk_buff *skb, __be32 daddr, __be32 saddr,
1443 				u8 tos, struct net_device *dev, int our)
1444 {
1445 	struct rtable *rth;
1446 	struct in_device *in_dev = __in_dev_get_rcu(dev);
1447 	u32 itag = 0;
1448 	int err;
1449 
1450 	/* Primary sanity checks. */
1451 
1452 	if (in_dev == NULL)
1453 		return -EINVAL;
1454 
1455 	if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) ||
1456 	    skb->protocol != htons(ETH_P_IP))
1457 		goto e_inval;
1458 
1459 	if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev)))
1460 		if (ipv4_is_loopback(saddr))
1461 			goto e_inval;
1462 
1463 	if (ipv4_is_zeronet(saddr)) {
1464 		if (!ipv4_is_local_multicast(daddr))
1465 			goto e_inval;
1466 	} else {
1467 		err = fib_validate_source(skb, saddr, 0, tos, 0, dev,
1468 					  in_dev, &itag);
1469 		if (err < 0)
1470 			goto e_err;
1471 	}
1472 	rth = rt_dst_alloc(dev_net(dev)->loopback_dev,
1473 			   IN_DEV_CONF_GET(in_dev, NOPOLICY), false, false);
1474 	if (!rth)
1475 		goto e_nobufs;
1476 
1477 #ifdef CONFIG_IP_ROUTE_CLASSID
1478 	rth->dst.tclassid = itag;
1479 #endif
1480 	rth->dst.output = ip_rt_bug;
1481 
1482 	rth->rt_genid	= rt_genid_ipv4(dev_net(dev));
1483 	rth->rt_flags	= RTCF_MULTICAST;
1484 	rth->rt_type	= RTN_MULTICAST;
1485 	rth->rt_is_input= 1;
1486 	rth->rt_iif	= 0;
1487 	rth->rt_pmtu	= 0;
1488 	rth->rt_gateway	= 0;
1489 	rth->rt_uses_gateway = 0;
1490 	INIT_LIST_HEAD(&rth->rt_uncached);
1491 	if (our) {
1492 		rth->dst.input= ip_local_deliver;
1493 		rth->rt_flags |= RTCF_LOCAL;
1494 	}
1495 
1496 #ifdef CONFIG_IP_MROUTE
1497 	if (!ipv4_is_local_multicast(daddr) && IN_DEV_MFORWARD(in_dev))
1498 		rth->dst.input = ip_mr_input;
1499 #endif
1500 	RT_CACHE_STAT_INC(in_slow_mc);
1501 
1502 	skb_dst_set(skb, &rth->dst);
1503 	return 0;
1504 
1505 e_nobufs:
1506 	return -ENOBUFS;
1507 e_inval:
1508 	return -EINVAL;
1509 e_err:
1510 	return err;
1511 }
1512 
1513 
ip_handle_martian_source(struct net_device * dev,struct in_device * in_dev,struct sk_buff * skb,__be32 daddr,__be32 saddr)1514 static void ip_handle_martian_source(struct net_device *dev,
1515 				     struct in_device *in_dev,
1516 				     struct sk_buff *skb,
1517 				     __be32 daddr,
1518 				     __be32 saddr)
1519 {
1520 	RT_CACHE_STAT_INC(in_martian_src);
1521 #ifdef CONFIG_IP_ROUTE_VERBOSE
1522 	if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) {
1523 		/*
1524 		 *	RFC1812 recommendation, if source is martian,
1525 		 *	the only hint is MAC header.
1526 		 */
1527 		pr_warn("martian source %pI4 from %pI4, on dev %s\n",
1528 			&daddr, &saddr, dev->name);
1529 		if (dev->hard_header_len && skb_mac_header_was_set(skb)) {
1530 			print_hex_dump(KERN_WARNING, "ll header: ",
1531 				       DUMP_PREFIX_OFFSET, 16, 1,
1532 				       skb_mac_header(skb),
1533 				       dev->hard_header_len, true);
1534 		}
1535 	}
1536 #endif
1537 }
1538 
1539 /* called in rcu_read_lock() section */
__mkroute_input(struct sk_buff * skb,const struct fib_result * res,struct in_device * in_dev,__be32 daddr,__be32 saddr,u32 tos)1540 static int __mkroute_input(struct sk_buff *skb,
1541 			   const struct fib_result *res,
1542 			   struct in_device *in_dev,
1543 			   __be32 daddr, __be32 saddr, u32 tos)
1544 {
1545 	struct fib_nh_exception *fnhe;
1546 	struct rtable *rth;
1547 	int err;
1548 	struct in_device *out_dev;
1549 	unsigned int flags = 0;
1550 	bool do_cache;
1551 	u32 itag = 0;
1552 
1553 	/* get a working reference to the output device */
1554 	out_dev = __in_dev_get_rcu(FIB_RES_DEV(*res));
1555 	if (out_dev == NULL) {
1556 		net_crit_ratelimited("Bug in ip_route_input_slow(). Please report.\n");
1557 		return -EINVAL;
1558 	}
1559 
1560 	err = fib_validate_source(skb, saddr, daddr, tos, FIB_RES_OIF(*res),
1561 				  in_dev->dev, in_dev, &itag);
1562 	if (err < 0) {
1563 		ip_handle_martian_source(in_dev->dev, in_dev, skb, daddr,
1564 					 saddr);
1565 
1566 		goto cleanup;
1567 	}
1568 
1569 	do_cache = res->fi && !itag;
1570 	if (out_dev == in_dev && err && IN_DEV_TX_REDIRECTS(out_dev) &&
1571 	    skb->protocol == htons(ETH_P_IP) &&
1572 	    (IN_DEV_SHARED_MEDIA(out_dev) ||
1573 	     inet_addr_onlink(out_dev, saddr, FIB_RES_GW(*res))))
1574 		IPCB(skb)->flags |= IPSKB_DOREDIRECT;
1575 
1576 	if (skb->protocol != htons(ETH_P_IP)) {
1577 		/* Not IP (i.e. ARP). Do not create route, if it is
1578 		 * invalid for proxy arp. DNAT routes are always valid.
1579 		 *
1580 		 * Proxy arp feature have been extended to allow, ARP
1581 		 * replies back to the same interface, to support
1582 		 * Private VLAN switch technologies. See arp.c.
1583 		 */
1584 		if (out_dev == in_dev &&
1585 		    IN_DEV_PROXY_ARP_PVLAN(in_dev) == 0) {
1586 			err = -EINVAL;
1587 			goto cleanup;
1588 		}
1589 	}
1590 
1591 	fnhe = find_exception(&FIB_RES_NH(*res), daddr);
1592 	if (do_cache) {
1593 		if (fnhe != NULL)
1594 			rth = rcu_dereference(fnhe->fnhe_rth_input);
1595 		else
1596 			rth = rcu_dereference(FIB_RES_NH(*res).nh_rth_input);
1597 
1598 		if (rt_cache_valid(rth)) {
1599 			skb_dst_set_noref(skb, &rth->dst);
1600 			goto out;
1601 		}
1602 	}
1603 
1604 	rth = rt_dst_alloc(out_dev->dev,
1605 			   IN_DEV_CONF_GET(in_dev, NOPOLICY),
1606 			   IN_DEV_CONF_GET(out_dev, NOXFRM), do_cache);
1607 	if (!rth) {
1608 		err = -ENOBUFS;
1609 		goto cleanup;
1610 	}
1611 
1612 	rth->rt_genid = rt_genid_ipv4(dev_net(rth->dst.dev));
1613 	rth->rt_flags = flags;
1614 	rth->rt_type = res->type;
1615 	rth->rt_is_input = 1;
1616 	rth->rt_iif 	= 0;
1617 	rth->rt_pmtu	= 0;
1618 	rth->rt_gateway	= 0;
1619 	rth->rt_uses_gateway = 0;
1620 	INIT_LIST_HEAD(&rth->rt_uncached);
1621 	RT_CACHE_STAT_INC(in_slow_tot);
1622 
1623 	rth->dst.input = ip_forward;
1624 	rth->dst.output = ip_output;
1625 
1626 	rt_set_nexthop(rth, daddr, res, fnhe, res->fi, res->type, itag);
1627 	skb_dst_set(skb, &rth->dst);
1628 out:
1629 	err = 0;
1630  cleanup:
1631 	return err;
1632 }
1633 
ip_mkroute_input(struct sk_buff * skb,struct fib_result * res,const struct flowi4 * fl4,struct in_device * in_dev,__be32 daddr,__be32 saddr,u32 tos)1634 static int ip_mkroute_input(struct sk_buff *skb,
1635 			    struct fib_result *res,
1636 			    const struct flowi4 *fl4,
1637 			    struct in_device *in_dev,
1638 			    __be32 daddr, __be32 saddr, u32 tos)
1639 {
1640 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1641 	if (res->fi && res->fi->fib_nhs > 1)
1642 		fib_select_multipath(res);
1643 #endif
1644 
1645 	/* create a routing cache entry */
1646 	return __mkroute_input(skb, res, in_dev, daddr, saddr, tos);
1647 }
1648 
1649 /*
1650  *	NOTE. We drop all the packets that has local source
1651  *	addresses, because every properly looped back packet
1652  *	must have correct destination already attached by output routine.
1653  *
1654  *	Such approach solves two big problems:
1655  *	1. Not simplex devices are handled properly.
1656  *	2. IP spoofing attempts are filtered with 100% of guarantee.
1657  *	called with rcu_read_lock()
1658  */
1659 
ip_route_input_slow(struct sk_buff * skb,__be32 daddr,__be32 saddr,u8 tos,struct net_device * dev)1660 static int ip_route_input_slow(struct sk_buff *skb, __be32 daddr, __be32 saddr,
1661 			       u8 tos, struct net_device *dev)
1662 {
1663 	struct fib_result res;
1664 	struct in_device *in_dev = __in_dev_get_rcu(dev);
1665 	struct flowi4	fl4;
1666 	unsigned int	flags = 0;
1667 	u32		itag = 0;
1668 	struct rtable	*rth;
1669 	int		err = -EINVAL;
1670 	struct net    *net = dev_net(dev);
1671 	bool do_cache;
1672 
1673 	/* IP on this device is disabled. */
1674 
1675 	if (!in_dev)
1676 		goto out;
1677 
1678 	/* Check for the most weird martians, which can be not detected
1679 	   by fib_lookup.
1680 	 */
1681 
1682 	if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr))
1683 		goto martian_source;
1684 
1685 	res.fi = NULL;
1686 	if (ipv4_is_lbcast(daddr) || (saddr == 0 && daddr == 0))
1687 		goto brd_input;
1688 
1689 	/* Accept zero addresses only to limited broadcast;
1690 	 * I even do not know to fix it or not. Waiting for complains :-)
1691 	 */
1692 	if (ipv4_is_zeronet(saddr))
1693 		goto martian_source;
1694 
1695 	if (ipv4_is_zeronet(daddr))
1696 		goto martian_destination;
1697 
1698 	/* Following code try to avoid calling IN_DEV_NET_ROUTE_LOCALNET(),
1699 	 * and call it once if daddr or/and saddr are loopback addresses
1700 	 */
1701 	if (ipv4_is_loopback(daddr)) {
1702 		if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev, net))
1703 			goto martian_destination;
1704 	} else if (ipv4_is_loopback(saddr)) {
1705 		if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev, net))
1706 			goto martian_source;
1707 	}
1708 
1709 	/*
1710 	 *	Now we are ready to route packet.
1711 	 */
1712 	fl4.flowi4_oif = 0;
1713 	fl4.flowi4_iif = dev->ifindex;
1714 	fl4.flowi4_mark = skb->mark;
1715 	fl4.flowi4_tos = tos;
1716 	fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
1717 	fl4.daddr = daddr;
1718 	fl4.saddr = saddr;
1719 	err = fib_lookup(net, &fl4, &res);
1720 	if (err != 0) {
1721 		if (!IN_DEV_FORWARD(in_dev))
1722 			err = -EHOSTUNREACH;
1723 		goto no_route;
1724 	}
1725 
1726 	if (res.type == RTN_BROADCAST)
1727 		goto brd_input;
1728 
1729 	if (res.type == RTN_LOCAL) {
1730 		err = fib_validate_source(skb, saddr, daddr, tos,
1731 					  0, dev, in_dev, &itag);
1732 		if (err < 0)
1733 			goto martian_source_keep_err;
1734 		goto local_input;
1735 	}
1736 
1737 	if (!IN_DEV_FORWARD(in_dev)) {
1738 		err = -EHOSTUNREACH;
1739 		goto no_route;
1740 	}
1741 	if (res.type != RTN_UNICAST)
1742 		goto martian_destination;
1743 
1744 	err = ip_mkroute_input(skb, &res, &fl4, in_dev, daddr, saddr, tos);
1745 out:	return err;
1746 
1747 brd_input:
1748 	if (skb->protocol != htons(ETH_P_IP))
1749 		goto e_inval;
1750 
1751 	if (!ipv4_is_zeronet(saddr)) {
1752 		err = fib_validate_source(skb, saddr, 0, tos, 0, dev,
1753 					  in_dev, &itag);
1754 		if (err < 0)
1755 			goto martian_source_keep_err;
1756 	}
1757 	flags |= RTCF_BROADCAST;
1758 	res.type = RTN_BROADCAST;
1759 	RT_CACHE_STAT_INC(in_brd);
1760 
1761 local_input:
1762 	do_cache = false;
1763 	if (res.fi) {
1764 		if (!itag) {
1765 			rth = rcu_dereference(FIB_RES_NH(res).nh_rth_input);
1766 			if (rt_cache_valid(rth)) {
1767 				skb_dst_set_noref(skb, &rth->dst);
1768 				err = 0;
1769 				goto out;
1770 			}
1771 			do_cache = true;
1772 		}
1773 	}
1774 
1775 	rth = rt_dst_alloc(net->loopback_dev,
1776 			   IN_DEV_CONF_GET(in_dev, NOPOLICY), false, do_cache);
1777 	if (!rth)
1778 		goto e_nobufs;
1779 
1780 	rth->dst.input= ip_local_deliver;
1781 	rth->dst.output= ip_rt_bug;
1782 #ifdef CONFIG_IP_ROUTE_CLASSID
1783 	rth->dst.tclassid = itag;
1784 #endif
1785 
1786 	rth->rt_genid = rt_genid_ipv4(net);
1787 	rth->rt_flags 	= flags|RTCF_LOCAL;
1788 	rth->rt_type	= res.type;
1789 	rth->rt_is_input = 1;
1790 	rth->rt_iif	= 0;
1791 	rth->rt_pmtu	= 0;
1792 	rth->rt_gateway	= 0;
1793 	rth->rt_uses_gateway = 0;
1794 	INIT_LIST_HEAD(&rth->rt_uncached);
1795 	RT_CACHE_STAT_INC(in_slow_tot);
1796 	if (res.type == RTN_UNREACHABLE) {
1797 		rth->dst.input= ip_error;
1798 		rth->dst.error= -err;
1799 		rth->rt_flags 	&= ~RTCF_LOCAL;
1800 	}
1801 	if (do_cache) {
1802 		if (unlikely(!rt_cache_route(&FIB_RES_NH(res), rth))) {
1803 			rth->dst.flags |= DST_NOCACHE;
1804 			rt_add_uncached_list(rth);
1805 		}
1806 	}
1807 	skb_dst_set(skb, &rth->dst);
1808 	err = 0;
1809 	goto out;
1810 
1811 no_route:
1812 	RT_CACHE_STAT_INC(in_no_route);
1813 	res.type = RTN_UNREACHABLE;
1814 	res.fi = NULL;
1815 	goto local_input;
1816 
1817 	/*
1818 	 *	Do not cache martian addresses: they should be logged (RFC1812)
1819 	 */
1820 martian_destination:
1821 	RT_CACHE_STAT_INC(in_martian_dst);
1822 #ifdef CONFIG_IP_ROUTE_VERBOSE
1823 	if (IN_DEV_LOG_MARTIANS(in_dev))
1824 		net_warn_ratelimited("martian destination %pI4 from %pI4, dev %s\n",
1825 				     &daddr, &saddr, dev->name);
1826 #endif
1827 
1828 e_inval:
1829 	err = -EINVAL;
1830 	goto out;
1831 
1832 e_nobufs:
1833 	err = -ENOBUFS;
1834 	goto out;
1835 
1836 martian_source:
1837 	err = -EINVAL;
1838 martian_source_keep_err:
1839 	ip_handle_martian_source(dev, in_dev, skb, daddr, saddr);
1840 	goto out;
1841 }
1842 
ip_route_input_noref(struct sk_buff * skb,__be32 daddr,__be32 saddr,u8 tos,struct net_device * dev)1843 int ip_route_input_noref(struct sk_buff *skb, __be32 daddr, __be32 saddr,
1844 			 u8 tos, struct net_device *dev)
1845 {
1846 	int res;
1847 
1848 	tos &= IPTOS_RT_MASK;
1849 	rcu_read_lock();
1850 
1851 	/* Multicast recognition logic is moved from route cache to here.
1852 	   The problem was that too many Ethernet cards have broken/missing
1853 	   hardware multicast filters :-( As result the host on multicasting
1854 	   network acquires a lot of useless route cache entries, sort of
1855 	   SDR messages from all the world. Now we try to get rid of them.
1856 	   Really, provided software IP multicast filter is organized
1857 	   reasonably (at least, hashed), it does not result in a slowdown
1858 	   comparing with route cache reject entries.
1859 	   Note, that multicast routers are not affected, because
1860 	   route cache entry is created eventually.
1861 	 */
1862 	if (ipv4_is_multicast(daddr)) {
1863 		struct in_device *in_dev = __in_dev_get_rcu(dev);
1864 
1865 		if (in_dev) {
1866 			int our = ip_check_mc_rcu(in_dev, daddr, saddr,
1867 						  ip_hdr(skb)->protocol);
1868 			if (our
1869 #ifdef CONFIG_IP_MROUTE
1870 				||
1871 			    (!ipv4_is_local_multicast(daddr) &&
1872 			     IN_DEV_MFORWARD(in_dev))
1873 #endif
1874 			   ) {
1875 				int res = ip_route_input_mc(skb, daddr, saddr,
1876 							    tos, dev, our);
1877 				rcu_read_unlock();
1878 				return res;
1879 			}
1880 		}
1881 		rcu_read_unlock();
1882 		return -EINVAL;
1883 	}
1884 	res = ip_route_input_slow(skb, daddr, saddr, tos, dev);
1885 	rcu_read_unlock();
1886 	return res;
1887 }
1888 EXPORT_SYMBOL(ip_route_input_noref);
1889 
1890 /* called with rcu_read_lock() */
__mkroute_output(const struct fib_result * res,const struct flowi4 * fl4,int orig_oif,struct net_device * dev_out,unsigned int flags)1891 static struct rtable *__mkroute_output(const struct fib_result *res,
1892 				       const struct flowi4 *fl4, int orig_oif,
1893 				       struct net_device *dev_out,
1894 				       unsigned int flags)
1895 {
1896 	struct fib_info *fi = res->fi;
1897 	struct fib_nh_exception *fnhe;
1898 	struct in_device *in_dev;
1899 	u16 type = res->type;
1900 	struct rtable *rth;
1901 	bool do_cache;
1902 
1903 	in_dev = __in_dev_get_rcu(dev_out);
1904 	if (!in_dev)
1905 		return ERR_PTR(-EINVAL);
1906 
1907 	if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev)))
1908 		if (ipv4_is_loopback(fl4->saddr) && !(dev_out->flags & IFF_LOOPBACK))
1909 			return ERR_PTR(-EINVAL);
1910 
1911 	if (ipv4_is_lbcast(fl4->daddr))
1912 		type = RTN_BROADCAST;
1913 	else if (ipv4_is_multicast(fl4->daddr))
1914 		type = RTN_MULTICAST;
1915 	else if (ipv4_is_zeronet(fl4->daddr))
1916 		return ERR_PTR(-EINVAL);
1917 
1918 	if (dev_out->flags & IFF_LOOPBACK)
1919 		flags |= RTCF_LOCAL;
1920 
1921 	do_cache = true;
1922 	if (type == RTN_BROADCAST) {
1923 		flags |= RTCF_BROADCAST | RTCF_LOCAL;
1924 		fi = NULL;
1925 	} else if (type == RTN_MULTICAST) {
1926 		flags |= RTCF_MULTICAST | RTCF_LOCAL;
1927 		if (!ip_check_mc_rcu(in_dev, fl4->daddr, fl4->saddr,
1928 				     fl4->flowi4_proto))
1929 			flags &= ~RTCF_LOCAL;
1930 		else
1931 			do_cache = false;
1932 		/* If multicast route do not exist use
1933 		 * default one, but do not gateway in this case.
1934 		 * Yes, it is hack.
1935 		 */
1936 		if (fi && res->prefixlen < 4)
1937 			fi = NULL;
1938 	} else if ((type == RTN_LOCAL) && (orig_oif != 0) &&
1939 		   (orig_oif != dev_out->ifindex)) {
1940 		/* For local routes that require a particular output interface
1941 		 * we do not want to cache the result.  Caching the result
1942 		 * causes incorrect behaviour when there are multiple source
1943 		 * addresses on the interface, the end result being that if the
1944 		 * intended recipient is waiting on that interface for the
1945 		 * packet he won't receive it because it will be delivered on
1946 		 * the loopback interface and the IP_PKTINFO ipi_ifindex will
1947 		 * be set to the loopback interface as well.
1948 		 */
1949 		fi = NULL;
1950 	}
1951 
1952 	fnhe = NULL;
1953 	do_cache &= fi != NULL;
1954 	if (do_cache) {
1955 		struct rtable __rcu **prth;
1956 		struct fib_nh *nh = &FIB_RES_NH(*res);
1957 
1958 		fnhe = find_exception(nh, fl4->daddr);
1959 		if (fnhe)
1960 			prth = &fnhe->fnhe_rth_output;
1961 		else {
1962 			if (unlikely(fl4->flowi4_flags &
1963 				     FLOWI_FLAG_KNOWN_NH &&
1964 				     !(nh->nh_gw &&
1965 				       nh->nh_scope == RT_SCOPE_LINK))) {
1966 				do_cache = false;
1967 				goto add;
1968 			}
1969 			prth = raw_cpu_ptr(nh->nh_pcpu_rth_output);
1970 		}
1971 		rth = rcu_dereference(*prth);
1972 		if (rt_cache_valid(rth)) {
1973 			dst_hold(&rth->dst);
1974 			return rth;
1975 		}
1976 	}
1977 
1978 add:
1979 	rth = rt_dst_alloc(dev_out,
1980 			   IN_DEV_CONF_GET(in_dev, NOPOLICY),
1981 			   IN_DEV_CONF_GET(in_dev, NOXFRM),
1982 			   do_cache);
1983 	if (!rth)
1984 		return ERR_PTR(-ENOBUFS);
1985 
1986 	rth->dst.output = ip_output;
1987 
1988 	rth->rt_genid = rt_genid_ipv4(dev_net(dev_out));
1989 	rth->rt_flags	= flags;
1990 	rth->rt_type	= type;
1991 	rth->rt_is_input = 0;
1992 	rth->rt_iif	= orig_oif ? : 0;
1993 	rth->rt_pmtu	= 0;
1994 	rth->rt_gateway = 0;
1995 	rth->rt_uses_gateway = 0;
1996 	INIT_LIST_HEAD(&rth->rt_uncached);
1997 
1998 	RT_CACHE_STAT_INC(out_slow_tot);
1999 
2000 	if (flags & RTCF_LOCAL)
2001 		rth->dst.input = ip_local_deliver;
2002 	if (flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
2003 		if (flags & RTCF_LOCAL &&
2004 		    !(dev_out->flags & IFF_LOOPBACK)) {
2005 			rth->dst.output = ip_mc_output;
2006 			RT_CACHE_STAT_INC(out_slow_mc);
2007 		}
2008 #ifdef CONFIG_IP_MROUTE
2009 		if (type == RTN_MULTICAST) {
2010 			if (IN_DEV_MFORWARD(in_dev) &&
2011 			    !ipv4_is_local_multicast(fl4->daddr)) {
2012 				rth->dst.input = ip_mr_input;
2013 				rth->dst.output = ip_mc_output;
2014 			}
2015 		}
2016 #endif
2017 	}
2018 
2019 	rt_set_nexthop(rth, fl4->daddr, res, fnhe, fi, type, 0);
2020 
2021 	return rth;
2022 }
2023 
2024 /*
2025  * Major route resolver routine.
2026  */
2027 
__ip_route_output_key(struct net * net,struct flowi4 * fl4)2028 struct rtable *__ip_route_output_key(struct net *net, struct flowi4 *fl4)
2029 {
2030 	struct net_device *dev_out = NULL;
2031 	__u8 tos = RT_FL_TOS(fl4);
2032 	unsigned int flags = 0;
2033 	struct fib_result res;
2034 	struct rtable *rth;
2035 	int orig_oif;
2036 
2037 	res.tclassid	= 0;
2038 	res.fi		= NULL;
2039 	res.table	= NULL;
2040 
2041 	orig_oif = fl4->flowi4_oif;
2042 
2043 	fl4->flowi4_iif = LOOPBACK_IFINDEX;
2044 	fl4->flowi4_tos = tos & IPTOS_RT_MASK;
2045 	fl4->flowi4_scope = ((tos & RTO_ONLINK) ?
2046 			 RT_SCOPE_LINK : RT_SCOPE_UNIVERSE);
2047 
2048 	rcu_read_lock();
2049 	if (fl4->saddr) {
2050 		rth = ERR_PTR(-EINVAL);
2051 		if (ipv4_is_multicast(fl4->saddr) ||
2052 		    ipv4_is_lbcast(fl4->saddr) ||
2053 		    ipv4_is_zeronet(fl4->saddr))
2054 			goto out;
2055 
2056 		/* I removed check for oif == dev_out->oif here.
2057 		   It was wrong for two reasons:
2058 		   1. ip_dev_find(net, saddr) can return wrong iface, if saddr
2059 		      is assigned to multiple interfaces.
2060 		   2. Moreover, we are allowed to send packets with saddr
2061 		      of another iface. --ANK
2062 		 */
2063 
2064 		if (fl4->flowi4_oif == 0 &&
2065 		    (ipv4_is_multicast(fl4->daddr) ||
2066 		     ipv4_is_lbcast(fl4->daddr))) {
2067 			/* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2068 			dev_out = __ip_dev_find(net, fl4->saddr, false);
2069 			if (dev_out == NULL)
2070 				goto out;
2071 
2072 			/* Special hack: user can direct multicasts
2073 			   and limited broadcast via necessary interface
2074 			   without fiddling with IP_MULTICAST_IF or IP_PKTINFO.
2075 			   This hack is not just for fun, it allows
2076 			   vic,vat and friends to work.
2077 			   They bind socket to loopback, set ttl to zero
2078 			   and expect that it will work.
2079 			   From the viewpoint of routing cache they are broken,
2080 			   because we are not allowed to build multicast path
2081 			   with loopback source addr (look, routing cache
2082 			   cannot know, that ttl is zero, so that packet
2083 			   will not leave this host and route is valid).
2084 			   Luckily, this hack is good workaround.
2085 			 */
2086 
2087 			fl4->flowi4_oif = dev_out->ifindex;
2088 			goto make_route;
2089 		}
2090 
2091 		if (!(fl4->flowi4_flags & FLOWI_FLAG_ANYSRC)) {
2092 			/* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2093 			if (!__ip_dev_find(net, fl4->saddr, false))
2094 				goto out;
2095 		}
2096 	}
2097 
2098 
2099 	if (fl4->flowi4_oif) {
2100 		dev_out = dev_get_by_index_rcu(net, fl4->flowi4_oif);
2101 		rth = ERR_PTR(-ENODEV);
2102 		if (dev_out == NULL)
2103 			goto out;
2104 
2105 		/* RACE: Check return value of inet_select_addr instead. */
2106 		if (!(dev_out->flags & IFF_UP) || !__in_dev_get_rcu(dev_out)) {
2107 			rth = ERR_PTR(-ENETUNREACH);
2108 			goto out;
2109 		}
2110 		if (ipv4_is_local_multicast(fl4->daddr) ||
2111 		    ipv4_is_lbcast(fl4->daddr)) {
2112 			if (!fl4->saddr)
2113 				fl4->saddr = inet_select_addr(dev_out, 0,
2114 							      RT_SCOPE_LINK);
2115 			goto make_route;
2116 		}
2117 		if (!fl4->saddr) {
2118 			if (ipv4_is_multicast(fl4->daddr))
2119 				fl4->saddr = inet_select_addr(dev_out, 0,
2120 							      fl4->flowi4_scope);
2121 			else if (!fl4->daddr)
2122 				fl4->saddr = inet_select_addr(dev_out, 0,
2123 							      RT_SCOPE_HOST);
2124 		}
2125 	}
2126 
2127 	if (!fl4->daddr) {
2128 		fl4->daddr = fl4->saddr;
2129 		if (!fl4->daddr)
2130 			fl4->daddr = fl4->saddr = htonl(INADDR_LOOPBACK);
2131 		dev_out = net->loopback_dev;
2132 		fl4->flowi4_oif = LOOPBACK_IFINDEX;
2133 		res.type = RTN_LOCAL;
2134 		flags |= RTCF_LOCAL;
2135 		goto make_route;
2136 	}
2137 
2138 	if (fib_lookup(net, fl4, &res)) {
2139 		res.fi = NULL;
2140 		res.table = NULL;
2141 		if (fl4->flowi4_oif) {
2142 			/* Apparently, routing tables are wrong. Assume,
2143 			   that the destination is on link.
2144 
2145 			   WHY? DW.
2146 			   Because we are allowed to send to iface
2147 			   even if it has NO routes and NO assigned
2148 			   addresses. When oif is specified, routing
2149 			   tables are looked up with only one purpose:
2150 			   to catch if destination is gatewayed, rather than
2151 			   direct. Moreover, if MSG_DONTROUTE is set,
2152 			   we send packet, ignoring both routing tables
2153 			   and ifaddr state. --ANK
2154 
2155 
2156 			   We could make it even if oif is unknown,
2157 			   likely IPv6, but we do not.
2158 			 */
2159 
2160 			if (fl4->saddr == 0)
2161 				fl4->saddr = inet_select_addr(dev_out, 0,
2162 							      RT_SCOPE_LINK);
2163 			res.type = RTN_UNICAST;
2164 			goto make_route;
2165 		}
2166 		rth = ERR_PTR(-ENETUNREACH);
2167 		goto out;
2168 	}
2169 
2170 	if (res.type == RTN_LOCAL) {
2171 		if (!fl4->saddr) {
2172 			if (res.fi->fib_prefsrc)
2173 				fl4->saddr = res.fi->fib_prefsrc;
2174 			else
2175 				fl4->saddr = fl4->daddr;
2176 		}
2177 		dev_out = net->loopback_dev;
2178 		fl4->flowi4_oif = dev_out->ifindex;
2179 		flags |= RTCF_LOCAL;
2180 		goto make_route;
2181 	}
2182 
2183 #ifdef CONFIG_IP_ROUTE_MULTIPATH
2184 	if (res.fi->fib_nhs > 1 && fl4->flowi4_oif == 0)
2185 		fib_select_multipath(&res);
2186 	else
2187 #endif
2188 	if (!res.prefixlen &&
2189 	    res.table->tb_num_default > 1 &&
2190 	    res.type == RTN_UNICAST && !fl4->flowi4_oif)
2191 		fib_select_default(&res);
2192 
2193 	if (!fl4->saddr)
2194 		fl4->saddr = FIB_RES_PREFSRC(net, res);
2195 
2196 	dev_out = FIB_RES_DEV(res);
2197 	fl4->flowi4_oif = dev_out->ifindex;
2198 
2199 
2200 make_route:
2201 	rth = __mkroute_output(&res, fl4, orig_oif, dev_out, flags);
2202 
2203 out:
2204 	rcu_read_unlock();
2205 	return rth;
2206 }
2207 EXPORT_SYMBOL_GPL(__ip_route_output_key);
2208 
ipv4_blackhole_dst_check(struct dst_entry * dst,u32 cookie)2209 static struct dst_entry *ipv4_blackhole_dst_check(struct dst_entry *dst, u32 cookie)
2210 {
2211 	return NULL;
2212 }
2213 
ipv4_blackhole_mtu(const struct dst_entry * dst)2214 static unsigned int ipv4_blackhole_mtu(const struct dst_entry *dst)
2215 {
2216 	unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2217 
2218 	return mtu ? : dst->dev->mtu;
2219 }
2220 
ipv4_rt_blackhole_update_pmtu(struct dst_entry * dst,struct sock * sk,struct sk_buff * skb,u32 mtu)2221 static void ipv4_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
2222 					  struct sk_buff *skb, u32 mtu)
2223 {
2224 }
2225 
ipv4_rt_blackhole_redirect(struct dst_entry * dst,struct sock * sk,struct sk_buff * skb)2226 static void ipv4_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
2227 				       struct sk_buff *skb)
2228 {
2229 }
2230 
ipv4_rt_blackhole_cow_metrics(struct dst_entry * dst,unsigned long old)2231 static u32 *ipv4_rt_blackhole_cow_metrics(struct dst_entry *dst,
2232 					  unsigned long old)
2233 {
2234 	return NULL;
2235 }
2236 
2237 static struct dst_ops ipv4_dst_blackhole_ops = {
2238 	.family			=	AF_INET,
2239 	.protocol		=	cpu_to_be16(ETH_P_IP),
2240 	.check			=	ipv4_blackhole_dst_check,
2241 	.mtu			=	ipv4_blackhole_mtu,
2242 	.default_advmss		=	ipv4_default_advmss,
2243 	.update_pmtu		=	ipv4_rt_blackhole_update_pmtu,
2244 	.redirect		=	ipv4_rt_blackhole_redirect,
2245 	.cow_metrics		=	ipv4_rt_blackhole_cow_metrics,
2246 	.neigh_lookup		=	ipv4_neigh_lookup,
2247 };
2248 
ipv4_blackhole_route(struct net * net,struct dst_entry * dst_orig)2249 struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2250 {
2251 	struct rtable *ort = (struct rtable *) dst_orig;
2252 	struct rtable *rt;
2253 
2254 	rt = dst_alloc(&ipv4_dst_blackhole_ops, NULL, 1, DST_OBSOLETE_NONE, 0);
2255 	if (rt) {
2256 		struct dst_entry *new = &rt->dst;
2257 
2258 		new->__use = 1;
2259 		new->input = dst_discard;
2260 		new->output = dst_discard_sk;
2261 
2262 		new->dev = ort->dst.dev;
2263 		if (new->dev)
2264 			dev_hold(new->dev);
2265 
2266 		rt->rt_is_input = ort->rt_is_input;
2267 		rt->rt_iif = ort->rt_iif;
2268 		rt->rt_pmtu = ort->rt_pmtu;
2269 
2270 		rt->rt_genid = rt_genid_ipv4(net);
2271 		rt->rt_flags = ort->rt_flags;
2272 		rt->rt_type = ort->rt_type;
2273 		rt->rt_gateway = ort->rt_gateway;
2274 		rt->rt_uses_gateway = ort->rt_uses_gateway;
2275 
2276 		INIT_LIST_HEAD(&rt->rt_uncached);
2277 
2278 		dst_free(new);
2279 	}
2280 
2281 	dst_release(dst_orig);
2282 
2283 	return rt ? &rt->dst : ERR_PTR(-ENOMEM);
2284 }
2285 
ip_route_output_flow(struct net * net,struct flowi4 * flp4,struct sock * sk)2286 struct rtable *ip_route_output_flow(struct net *net, struct flowi4 *flp4,
2287 				    struct sock *sk)
2288 {
2289 	struct rtable *rt = __ip_route_output_key(net, flp4);
2290 
2291 	if (IS_ERR(rt))
2292 		return rt;
2293 
2294 	if (flp4->flowi4_proto)
2295 		rt = (struct rtable *)xfrm_lookup_route(net, &rt->dst,
2296 							flowi4_to_flowi(flp4),
2297 							sk, 0);
2298 
2299 	return rt;
2300 }
2301 EXPORT_SYMBOL_GPL(ip_route_output_flow);
2302 
rt_fill_info(struct net * net,__be32 dst,__be32 src,struct flowi4 * fl4,struct sk_buff * skb,u32 portid,u32 seq,int event,int nowait,unsigned int flags)2303 static int rt_fill_info(struct net *net,  __be32 dst, __be32 src,
2304 			struct flowi4 *fl4, struct sk_buff *skb, u32 portid,
2305 			u32 seq, int event, int nowait, unsigned int flags)
2306 {
2307 	struct rtable *rt = skb_rtable(skb);
2308 	struct rtmsg *r;
2309 	struct nlmsghdr *nlh;
2310 	unsigned long expires = 0;
2311 	u32 error;
2312 	u32 metrics[RTAX_MAX];
2313 
2314 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*r), flags);
2315 	if (nlh == NULL)
2316 		return -EMSGSIZE;
2317 
2318 	r = nlmsg_data(nlh);
2319 	r->rtm_family	 = AF_INET;
2320 	r->rtm_dst_len	= 32;
2321 	r->rtm_src_len	= 0;
2322 	r->rtm_tos	= fl4->flowi4_tos;
2323 	r->rtm_table	= RT_TABLE_MAIN;
2324 	if (nla_put_u32(skb, RTA_TABLE, RT_TABLE_MAIN))
2325 		goto nla_put_failure;
2326 	r->rtm_type	= rt->rt_type;
2327 	r->rtm_scope	= RT_SCOPE_UNIVERSE;
2328 	r->rtm_protocol = RTPROT_UNSPEC;
2329 	r->rtm_flags	= (rt->rt_flags & ~0xFFFF) | RTM_F_CLONED;
2330 	if (rt->rt_flags & RTCF_NOTIFY)
2331 		r->rtm_flags |= RTM_F_NOTIFY;
2332 	if (IPCB(skb)->flags & IPSKB_DOREDIRECT)
2333 		r->rtm_flags |= RTCF_DOREDIRECT;
2334 
2335 	if (nla_put_be32(skb, RTA_DST, dst))
2336 		goto nla_put_failure;
2337 	if (src) {
2338 		r->rtm_src_len = 32;
2339 		if (nla_put_be32(skb, RTA_SRC, src))
2340 			goto nla_put_failure;
2341 	}
2342 	if (rt->dst.dev &&
2343 	    nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2344 		goto nla_put_failure;
2345 #ifdef CONFIG_IP_ROUTE_CLASSID
2346 	if (rt->dst.tclassid &&
2347 	    nla_put_u32(skb, RTA_FLOW, rt->dst.tclassid))
2348 		goto nla_put_failure;
2349 #endif
2350 	if (!rt_is_input_route(rt) &&
2351 	    fl4->saddr != src) {
2352 		if (nla_put_be32(skb, RTA_PREFSRC, fl4->saddr))
2353 			goto nla_put_failure;
2354 	}
2355 	if (rt->rt_uses_gateway &&
2356 	    nla_put_be32(skb, RTA_GATEWAY, rt->rt_gateway))
2357 		goto nla_put_failure;
2358 
2359 	expires = rt->dst.expires;
2360 	if (expires) {
2361 		unsigned long now = jiffies;
2362 
2363 		if (time_before(now, expires))
2364 			expires -= now;
2365 		else
2366 			expires = 0;
2367 	}
2368 
2369 	memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
2370 	if (rt->rt_pmtu && expires)
2371 		metrics[RTAX_MTU - 1] = rt->rt_pmtu;
2372 	if (rtnetlink_put_metrics(skb, metrics) < 0)
2373 		goto nla_put_failure;
2374 
2375 	if (fl4->flowi4_mark &&
2376 	    nla_put_u32(skb, RTA_MARK, fl4->flowi4_mark))
2377 		goto nla_put_failure;
2378 
2379 	if (!uid_eq(fl4->flowi4_uid, INVALID_UID) &&
2380 	    nla_put_u32(skb, RTA_UID,
2381 			from_kuid_munged(current_user_ns(), fl4->flowi4_uid)))
2382 		goto nla_put_failure;
2383 
2384 	error = rt->dst.error;
2385 
2386 	if (rt_is_input_route(rt)) {
2387 #ifdef CONFIG_IP_MROUTE
2388 		if (ipv4_is_multicast(dst) && !ipv4_is_local_multicast(dst) &&
2389 		    IPV4_DEVCONF_ALL(net, MC_FORWARDING)) {
2390 			int err = ipmr_get_route(net, skb,
2391 						 fl4->saddr, fl4->daddr,
2392 						 r, nowait, portid);
2393 
2394 			if (err <= 0) {
2395 				if (!nowait) {
2396 					if (err == 0)
2397 						return 0;
2398 					goto nla_put_failure;
2399 				} else {
2400 					if (err == -EMSGSIZE)
2401 						goto nla_put_failure;
2402 					error = err;
2403 				}
2404 			}
2405 		} else
2406 #endif
2407 			if (nla_put_u32(skb, RTA_IIF, skb->dev->ifindex))
2408 				goto nla_put_failure;
2409 	}
2410 
2411 	if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, error) < 0)
2412 		goto nla_put_failure;
2413 
2414 	return nlmsg_end(skb, nlh);
2415 
2416 nla_put_failure:
2417 	nlmsg_cancel(skb, nlh);
2418 	return -EMSGSIZE;
2419 }
2420 
inet_rtm_getroute(struct sk_buff * in_skb,struct nlmsghdr * nlh)2421 static int inet_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
2422 {
2423 	struct net *net = sock_net(in_skb->sk);
2424 	struct rtmsg *rtm;
2425 	struct nlattr *tb[RTA_MAX+1];
2426 	struct rtable *rt = NULL;
2427 	struct flowi4 fl4;
2428 	__be32 dst = 0;
2429 	__be32 src = 0;
2430 	u32 iif;
2431 	int err;
2432 	int mark;
2433 	struct sk_buff *skb;
2434 	kuid_t uid;
2435 
2436 	err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv4_policy);
2437 	if (err < 0)
2438 		goto errout;
2439 
2440 	rtm = nlmsg_data(nlh);
2441 
2442 	skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2443 	if (skb == NULL) {
2444 		err = -ENOBUFS;
2445 		goto errout;
2446 	}
2447 
2448 	/* Reserve room for dummy headers, this skb can pass
2449 	   through good chunk of routing engine.
2450 	 */
2451 	skb_reset_mac_header(skb);
2452 	skb_reset_network_header(skb);
2453 
2454 	/* Bugfix: need to give ip_route_input enough of an IP header to not gag. */
2455 	ip_hdr(skb)->protocol = IPPROTO_ICMP;
2456 	skb_reserve(skb, MAX_HEADER + sizeof(struct iphdr));
2457 
2458 	src = tb[RTA_SRC] ? nla_get_be32(tb[RTA_SRC]) : 0;
2459 	dst = tb[RTA_DST] ? nla_get_be32(tb[RTA_DST]) : 0;
2460 	iif = tb[RTA_IIF] ? nla_get_u32(tb[RTA_IIF]) : 0;
2461 	mark = tb[RTA_MARK] ? nla_get_u32(tb[RTA_MARK]) : 0;
2462 	if (tb[RTA_UID])
2463 		uid = make_kuid(current_user_ns(), nla_get_u32(tb[RTA_UID]));
2464 	else
2465 		uid = (iif ? INVALID_UID : current_uid());
2466 
2467 	memset(&fl4, 0, sizeof(fl4));
2468 	fl4.daddr = dst;
2469 	fl4.saddr = src;
2470 	fl4.flowi4_tos = rtm->rtm_tos;
2471 	fl4.flowi4_oif = tb[RTA_OIF] ? nla_get_u32(tb[RTA_OIF]) : 0;
2472 	fl4.flowi4_mark = mark;
2473 	fl4.flowi4_uid = uid;
2474 
2475 	if (iif) {
2476 		struct net_device *dev;
2477 
2478 		dev = __dev_get_by_index(net, iif);
2479 		if (dev == NULL) {
2480 			err = -ENODEV;
2481 			goto errout_free;
2482 		}
2483 
2484 		skb->protocol	= htons(ETH_P_IP);
2485 		skb->dev	= dev;
2486 		skb->mark	= mark;
2487 		local_bh_disable();
2488 		err = ip_route_input(skb, dst, src, rtm->rtm_tos, dev);
2489 		local_bh_enable();
2490 
2491 		rt = skb_rtable(skb);
2492 		if (err == 0 && rt->dst.error)
2493 			err = -rt->dst.error;
2494 	} else {
2495 		rt = ip_route_output_key(net, &fl4);
2496 
2497 		err = 0;
2498 		if (IS_ERR(rt))
2499 			err = PTR_ERR(rt);
2500 	}
2501 
2502 	if (err)
2503 		goto errout_free;
2504 
2505 	skb_dst_set(skb, &rt->dst);
2506 	if (rtm->rtm_flags & RTM_F_NOTIFY)
2507 		rt->rt_flags |= RTCF_NOTIFY;
2508 
2509 	err = rt_fill_info(net, dst, src, &fl4, skb,
2510 			   NETLINK_CB(in_skb).portid, nlh->nlmsg_seq,
2511 			   RTM_NEWROUTE, 0, 0);
2512 	if (err <= 0)
2513 		goto errout_free;
2514 
2515 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2516 errout:
2517 	return err;
2518 
2519 errout_free:
2520 	kfree_skb(skb);
2521 	goto errout;
2522 }
2523 
ip_rt_multicast_event(struct in_device * in_dev)2524 void ip_rt_multicast_event(struct in_device *in_dev)
2525 {
2526 	rt_cache_flush(dev_net(in_dev->dev));
2527 }
2528 
2529 #ifdef CONFIG_SYSCTL
2530 static int ip_rt_gc_timeout __read_mostly	= RT_GC_TIMEOUT;
2531 static int ip_rt_gc_interval __read_mostly  = 60 * HZ;
2532 static int ip_rt_gc_min_interval __read_mostly	= HZ / 2;
2533 static int ip_rt_gc_elasticity __read_mostly	= 8;
2534 
ipv4_sysctl_rtcache_flush(struct ctl_table * __ctl,int write,void __user * buffer,size_t * lenp,loff_t * ppos)2535 static int ipv4_sysctl_rtcache_flush(struct ctl_table *__ctl, int write,
2536 					void __user *buffer,
2537 					size_t *lenp, loff_t *ppos)
2538 {
2539 	struct net *net = (struct net *)__ctl->extra1;
2540 
2541 	if (write) {
2542 		rt_cache_flush(net);
2543 		fnhe_genid_bump(net);
2544 		return 0;
2545 	}
2546 
2547 	return -EINVAL;
2548 }
2549 
2550 static struct ctl_table ipv4_route_table[] = {
2551 	{
2552 		.procname	= "gc_thresh",
2553 		.data		= &ipv4_dst_ops.gc_thresh,
2554 		.maxlen		= sizeof(int),
2555 		.mode		= 0644,
2556 		.proc_handler	= proc_dointvec,
2557 	},
2558 	{
2559 		.procname	= "max_size",
2560 		.data		= &ip_rt_max_size,
2561 		.maxlen		= sizeof(int),
2562 		.mode		= 0644,
2563 		.proc_handler	= proc_dointvec,
2564 	},
2565 	{
2566 		/*  Deprecated. Use gc_min_interval_ms */
2567 
2568 		.procname	= "gc_min_interval",
2569 		.data		= &ip_rt_gc_min_interval,
2570 		.maxlen		= sizeof(int),
2571 		.mode		= 0644,
2572 		.proc_handler	= proc_dointvec_jiffies,
2573 	},
2574 	{
2575 		.procname	= "gc_min_interval_ms",
2576 		.data		= &ip_rt_gc_min_interval,
2577 		.maxlen		= sizeof(int),
2578 		.mode		= 0644,
2579 		.proc_handler	= proc_dointvec_ms_jiffies,
2580 	},
2581 	{
2582 		.procname	= "gc_timeout",
2583 		.data		= &ip_rt_gc_timeout,
2584 		.maxlen		= sizeof(int),
2585 		.mode		= 0644,
2586 		.proc_handler	= proc_dointvec_jiffies,
2587 	},
2588 	{
2589 		.procname	= "gc_interval",
2590 		.data		= &ip_rt_gc_interval,
2591 		.maxlen		= sizeof(int),
2592 		.mode		= 0644,
2593 		.proc_handler	= proc_dointvec_jiffies,
2594 	},
2595 	{
2596 		.procname	= "redirect_load",
2597 		.data		= &ip_rt_redirect_load,
2598 		.maxlen		= sizeof(int),
2599 		.mode		= 0644,
2600 		.proc_handler	= proc_dointvec,
2601 	},
2602 	{
2603 		.procname	= "redirect_number",
2604 		.data		= &ip_rt_redirect_number,
2605 		.maxlen		= sizeof(int),
2606 		.mode		= 0644,
2607 		.proc_handler	= proc_dointvec,
2608 	},
2609 	{
2610 		.procname	= "redirect_silence",
2611 		.data		= &ip_rt_redirect_silence,
2612 		.maxlen		= sizeof(int),
2613 		.mode		= 0644,
2614 		.proc_handler	= proc_dointvec,
2615 	},
2616 	{
2617 		.procname	= "error_cost",
2618 		.data		= &ip_rt_error_cost,
2619 		.maxlen		= sizeof(int),
2620 		.mode		= 0644,
2621 		.proc_handler	= proc_dointvec,
2622 	},
2623 	{
2624 		.procname	= "error_burst",
2625 		.data		= &ip_rt_error_burst,
2626 		.maxlen		= sizeof(int),
2627 		.mode		= 0644,
2628 		.proc_handler	= proc_dointvec,
2629 	},
2630 	{
2631 		.procname	= "gc_elasticity",
2632 		.data		= &ip_rt_gc_elasticity,
2633 		.maxlen		= sizeof(int),
2634 		.mode		= 0644,
2635 		.proc_handler	= proc_dointvec,
2636 	},
2637 	{
2638 		.procname	= "mtu_expires",
2639 		.data		= &ip_rt_mtu_expires,
2640 		.maxlen		= sizeof(int),
2641 		.mode		= 0644,
2642 		.proc_handler	= proc_dointvec_jiffies,
2643 	},
2644 	{
2645 		.procname	= "min_pmtu",
2646 		.data		= &ip_rt_min_pmtu,
2647 		.maxlen		= sizeof(int),
2648 		.mode		= 0644,
2649 		.proc_handler	= proc_dointvec,
2650 	},
2651 	{
2652 		.procname	= "min_adv_mss",
2653 		.data		= &ip_rt_min_advmss,
2654 		.maxlen		= sizeof(int),
2655 		.mode		= 0644,
2656 		.proc_handler	= proc_dointvec,
2657 	},
2658 	{ }
2659 };
2660 
2661 static struct ctl_table ipv4_route_flush_table[] = {
2662 	{
2663 		.procname	= "flush",
2664 		.maxlen		= sizeof(int),
2665 		.mode		= 0200,
2666 		.proc_handler	= ipv4_sysctl_rtcache_flush,
2667 	},
2668 	{ },
2669 };
2670 
sysctl_route_net_init(struct net * net)2671 static __net_init int sysctl_route_net_init(struct net *net)
2672 {
2673 	struct ctl_table *tbl;
2674 
2675 	tbl = ipv4_route_flush_table;
2676 	if (!net_eq(net, &init_net)) {
2677 		tbl = kmemdup(tbl, sizeof(ipv4_route_flush_table), GFP_KERNEL);
2678 		if (tbl == NULL)
2679 			goto err_dup;
2680 
2681 		/* Don't export sysctls to unprivileged users */
2682 		if (net->user_ns != &init_user_ns)
2683 			tbl[0].procname = NULL;
2684 	}
2685 	tbl[0].extra1 = net;
2686 
2687 	net->ipv4.route_hdr = register_net_sysctl(net, "net/ipv4/route", tbl);
2688 	if (net->ipv4.route_hdr == NULL)
2689 		goto err_reg;
2690 	return 0;
2691 
2692 err_reg:
2693 	if (tbl != ipv4_route_flush_table)
2694 		kfree(tbl);
2695 err_dup:
2696 	return -ENOMEM;
2697 }
2698 
sysctl_route_net_exit(struct net * net)2699 static __net_exit void sysctl_route_net_exit(struct net *net)
2700 {
2701 	struct ctl_table *tbl;
2702 
2703 	tbl = net->ipv4.route_hdr->ctl_table_arg;
2704 	unregister_net_sysctl_table(net->ipv4.route_hdr);
2705 	BUG_ON(tbl == ipv4_route_flush_table);
2706 	kfree(tbl);
2707 }
2708 
2709 static __net_initdata struct pernet_operations sysctl_route_ops = {
2710 	.init = sysctl_route_net_init,
2711 	.exit = sysctl_route_net_exit,
2712 };
2713 #endif
2714 
rt_genid_init(struct net * net)2715 static __net_init int rt_genid_init(struct net *net)
2716 {
2717 	atomic_set(&net->ipv4.rt_genid, 0);
2718 	atomic_set(&net->fnhe_genid, 0);
2719 	get_random_bytes(&net->ipv4.dev_addr_genid,
2720 			 sizeof(net->ipv4.dev_addr_genid));
2721 	return 0;
2722 }
2723 
2724 static __net_initdata struct pernet_operations rt_genid_ops = {
2725 	.init = rt_genid_init,
2726 };
2727 
ipv4_inetpeer_init(struct net * net)2728 static int __net_init ipv4_inetpeer_init(struct net *net)
2729 {
2730 	struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
2731 
2732 	if (!bp)
2733 		return -ENOMEM;
2734 	inet_peer_base_init(bp);
2735 	net->ipv4.peers = bp;
2736 	return 0;
2737 }
2738 
ipv4_inetpeer_exit(struct net * net)2739 static void __net_exit ipv4_inetpeer_exit(struct net *net)
2740 {
2741 	struct inet_peer_base *bp = net->ipv4.peers;
2742 
2743 	net->ipv4.peers = NULL;
2744 	inetpeer_invalidate_tree(bp);
2745 	kfree(bp);
2746 }
2747 
2748 static __net_initdata struct pernet_operations ipv4_inetpeer_ops = {
2749 	.init	=	ipv4_inetpeer_init,
2750 	.exit	=	ipv4_inetpeer_exit,
2751 };
2752 
2753 #ifdef CONFIG_IP_ROUTE_CLASSID
2754 struct ip_rt_acct __percpu *ip_rt_acct __read_mostly;
2755 #endif /* CONFIG_IP_ROUTE_CLASSID */
2756 
ip_rt_init(void)2757 int __init ip_rt_init(void)
2758 {
2759 	int rc = 0;
2760 
2761 	ip_idents = kmalloc(IP_IDENTS_SZ * sizeof(*ip_idents), GFP_KERNEL);
2762 	if (!ip_idents)
2763 		panic("IP: failed to allocate ip_idents\n");
2764 
2765 	prandom_bytes(ip_idents, IP_IDENTS_SZ * sizeof(*ip_idents));
2766 
2767 #ifdef CONFIG_IP_ROUTE_CLASSID
2768 	ip_rt_acct = __alloc_percpu(256 * sizeof(struct ip_rt_acct), __alignof__(struct ip_rt_acct));
2769 	if (!ip_rt_acct)
2770 		panic("IP: failed to allocate ip_rt_acct\n");
2771 #endif
2772 
2773 	ipv4_dst_ops.kmem_cachep =
2774 		kmem_cache_create("ip_dst_cache", sizeof(struct rtable), 0,
2775 				  SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
2776 
2777 	ipv4_dst_blackhole_ops.kmem_cachep = ipv4_dst_ops.kmem_cachep;
2778 
2779 	if (dst_entries_init(&ipv4_dst_ops) < 0)
2780 		panic("IP: failed to allocate ipv4_dst_ops counter\n");
2781 
2782 	if (dst_entries_init(&ipv4_dst_blackhole_ops) < 0)
2783 		panic("IP: failed to allocate ipv4_dst_blackhole_ops counter\n");
2784 
2785 	ipv4_dst_ops.gc_thresh = ~0;
2786 	ip_rt_max_size = INT_MAX;
2787 
2788 	devinet_init();
2789 	ip_fib_init();
2790 
2791 	if (ip_rt_proc_init())
2792 		pr_err("Unable to create route proc files\n");
2793 #ifdef CONFIG_XFRM
2794 	xfrm_init();
2795 	xfrm4_init();
2796 #endif
2797 	rtnl_register(PF_INET, RTM_GETROUTE, inet_rtm_getroute, NULL, NULL);
2798 
2799 #ifdef CONFIG_SYSCTL
2800 	register_pernet_subsys(&sysctl_route_ops);
2801 #endif
2802 	register_pernet_subsys(&rt_genid_ops);
2803 	register_pernet_subsys(&ipv4_inetpeer_ops);
2804 	return rc;
2805 }
2806 
2807 #ifdef CONFIG_SYSCTL
2808 /*
2809  * We really need to sanitize the damn ipv4 init order, then all
2810  * this nonsense will go away.
2811  */
ip_static_sysctl_init(void)2812 void __init ip_static_sysctl_init(void)
2813 {
2814 	register_net_sysctl(&init_net, "net/ipv4/route", ipv4_route_table);
2815 }
2816 #endif
2817