route.c 82.2 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7
/*
 * INET		An implementation of the TCP/IP protocol suite for the LINUX
 *		operating system.  INET is implemented using the  BSD Socket
 *		interface as the means of communication with the user level.
 *
 *		ROUTE - implementation of the IP router.
 *
8
 * Authors:	Ross Biro
L
Linus Torvalds 已提交
9 10 11 12 13 14 15 16 17 18 19 20
 *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 *		Alan Cox, <gw4pts@gw4pts.ampr.org>
 *		Linus Torvalds, <Linus.Torvalds@helsinki.fi>
 *		Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
 *
 * Fixes:
 *		Alan Cox	:	Verify area fixes.
 *		Alan Cox	:	cli() protects routing changes
 *		Rui Oliveira	:	ICMP routing table updates
 *		(rco@di.uminho.pt)	Routing table insertion and update
 *		Linus Torvalds	:	Rewrote bits to be sensible
 *		Alan Cox	:	Added BSD route gw semantics
21
 *		Alan Cox	:	Super /proc >4K
L
Linus Torvalds 已提交
22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38
 *		Alan Cox	:	MTU in route table
 *		Alan Cox	: 	MSS actually. Also added the window
 *					clamper.
 *		Sam Lantinga	:	Fixed route matching in rt_del()
 *		Alan Cox	:	Routing cache support.
 *		Alan Cox	:	Removed compatibility cruft.
 *		Alan Cox	:	RTF_REJECT support.
 *		Alan Cox	:	TCP irtt support.
 *		Jonathan Naylor	:	Added Metric support.
 *	Miquel van Smoorenburg	:	BSD API fixes.
 *	Miquel van Smoorenburg	:	Metrics.
 *		Alan Cox	:	Use __u32 properly
 *		Alan Cox	:	Aligned routing errors more closely with BSD
 *					our system is still very different.
 *		Alan Cox	:	Faster /proc handling
 *	Alexey Kuznetsov	:	Massive rework to support tree based routing,
 *					routing caches and better behaviour.
39
 *
L
Linus Torvalds 已提交
40 41 42 43 44 45 46 47 48 49 50 51 52 53 54
 *		Olaf Erb	:	irtt wasn't being copied right.
 *		Bjorn Ekwall	:	Kerneld route support.
 *		Alan Cox	:	Multicast fixed (I hope)
 * 		Pavel Krauz	:	Limited broadcast fixed
 *		Mike McLagan	:	Routing by source
 *	Alexey Kuznetsov	:	End of old history. Split to fib.c and
 *					route.c and rewritten from scratch.
 *		Andi Kleen	:	Load-limit warning messages.
 *	Vitaly E. Lavrov	:	Transparent proxy revived after year coma.
 *	Vitaly E. Lavrov	:	Race condition in ip_route_input_slow.
 *	Tobias Ringstrom	:	Uninitialized res.type in ip_route_output_slow.
 *	Vladimir V. Ivanov	:	IP rule info (flowid) is really useful.
 *		Marc Boucher	:	routing by fwmark
 *	Robert Olsson		:	Added rt_cache statistics
 *	Arnaldo C. Melo		:	Convert proc stuff to seq_file
55
 *	Eric Dumazet		:	hashed spinlocks and rt_check_expire() fixes.
56 57
 * 	Ilia Sotnikov		:	Ignore TOS on PMTUD and Redirect
 * 	Ilia Sotnikov		:	Removed TOS from hash calculations
L
Linus Torvalds 已提交
58 59 60 61 62 63 64 65 66 67 68 69 70 71
 *
 *		This program is free software; you can redistribute it and/or
 *		modify it under the terms of the GNU General Public License
 *		as published by the Free Software Foundation; either version
 *		2 of the License, or (at your option) any later version.
 */

#include <linux/module.h>
#include <asm/uaccess.h>
#include <asm/system.h>
#include <linux/bitops.h>
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/mm.h>
72
#include <linux/bootmem.h>
L
Linus Torvalds 已提交
73 74 75 76 77 78 79 80 81
#include <linux/string.h>
#include <linux/socket.h>
#include <linux/sockios.h>
#include <linux/errno.h>
#include <linux/in.h>
#include <linux/inet.h>
#include <linux/netdevice.h>
#include <linux/proc_fs.h>
#include <linux/init.h>
82
#include <linux/workqueue.h>
L
Linus Torvalds 已提交
83 84 85 86 87 88 89 90 91 92
#include <linux/skbuff.h>
#include <linux/inetdevice.h>
#include <linux/igmp.h>
#include <linux/pkt_sched.h>
#include <linux/mroute.h>
#include <linux/netfilter_ipv4.h>
#include <linux/random.h>
#include <linux/jhash.h>
#include <linux/rcupdate.h>
#include <linux/times.h>
93
#include <linux/slab.h>
94
#include <net/dst.h>
95
#include <net/net_namespace.h>
L
Linus Torvalds 已提交
96 97 98 99 100 101 102 103 104 105
#include <net/protocol.h>
#include <net/ip.h>
#include <net/route.h>
#include <net/inetpeer.h>
#include <net/sock.h>
#include <net/ip_fib.h>
#include <net/arp.h>
#include <net/tcp.h>
#include <net/icmp.h>
#include <net/xfrm.h>
106
#include <net/netevent.h>
107
#include <net/rtnetlink.h>
L
Linus Torvalds 已提交
108 109 110
#ifdef CONFIG_SYSCTL
#include <linux/sysctl.h>
#endif
D
David Miller 已提交
111
#include <net/atmclip.h>
112
#include <net/secure_seq.h>
L
Linus Torvalds 已提交
113

114 115
#define RT_FL_TOS(oldflp4) \
    ((u32)(oldflp4->flowi4_tos & (IPTOS_RT_MASK | RTO_ONLINK)))
L
Linus Torvalds 已提交
116 117 118 119 120 121

#define IP_MAX_MTU	0xFFF0

#define RT_GC_TIMEOUT (300*HZ)

static int ip_rt_max_size;
122 123 124 125 126 127 128 129 130 131 132 133
static int ip_rt_gc_timeout __read_mostly	= RT_GC_TIMEOUT;
static int ip_rt_gc_interval __read_mostly	= 60 * HZ;
static int ip_rt_gc_min_interval __read_mostly	= HZ / 2;
static int ip_rt_redirect_number __read_mostly	= 9;
static int ip_rt_redirect_load __read_mostly	= HZ / 50;
static int ip_rt_redirect_silence __read_mostly	= ((HZ / 50) << (9 + 1));
static int ip_rt_error_cost __read_mostly	= HZ;
static int ip_rt_error_burst __read_mostly	= 5 * HZ;
static int ip_rt_gc_elasticity __read_mostly	= 8;
static int ip_rt_mtu_expires __read_mostly	= 10 * 60 * HZ;
static int ip_rt_min_pmtu __read_mostly		= 512 + 20 + 20;
static int ip_rt_min_advmss __read_mostly	= 256;
134
static int rt_chain_length_max __read_mostly	= 20;
L
Linus Torvalds 已提交
135 136 137 138 139 140

/*
 *	Interface to generic destination cache.
 */

static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie);
141
static unsigned int	 ipv4_default_advmss(const struct dst_entry *dst);
142
static unsigned int	 ipv4_default_mtu(const struct dst_entry *dst);
L
Linus Torvalds 已提交
143 144 145 146
static void		 ipv4_dst_destroy(struct dst_entry *dst);
static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst);
static void		 ipv4_link_failure(struct sk_buff *skb);
static void		 ip_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
147
static int rt_garbage_collect(struct dst_ops *ops);
L
Linus Torvalds 已提交
148

E
Eric Dumazet 已提交
149 150 151 152
static void ipv4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
			    int how)
{
}
L
Linus Torvalds 已提交
153

154 155
static u32 *ipv4_cow_metrics(struct dst_entry *dst, unsigned long old)
{
156 157 158 159 160
	struct rtable *rt = (struct rtable *) dst;
	struct inet_peer *peer;
	u32 *p = NULL;

	if (!rt->peer)
161
		rt_bind_peer(rt, rt->rt_dst, 1);
162

163 164
	peer = rt->peer;
	if (peer) {
165 166 167
		u32 *old_p = __DST_METRICS_PTR(old);
		unsigned long prev, new;

168 169 170
		p = peer->metrics;
		if (inet_metrics_new(peer))
			memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188

		new = (unsigned long) p;
		prev = cmpxchg(&dst->_metrics, old, new);

		if (prev != old) {
			p = __DST_METRICS_PTR(prev);
			if (prev & DST_METRICS_READ_ONLY)
				p = NULL;
		} else {
			if (rt->fi) {
				fib_info_put(rt->fi);
				rt->fi = NULL;
			}
		}
	}
	return p;
}

189 190
static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst, const void *daddr);

L
Linus Torvalds 已提交
191 192
static struct dst_ops ipv4_dst_ops = {
	.family =		AF_INET,
193
	.protocol =		cpu_to_be16(ETH_P_IP),
L
Linus Torvalds 已提交
194 195
	.gc =			rt_garbage_collect,
	.check =		ipv4_dst_check,
196
	.default_advmss =	ipv4_default_advmss,
197
	.default_mtu =		ipv4_default_mtu,
198
	.cow_metrics =		ipv4_cow_metrics,
L
Linus Torvalds 已提交
199 200 201 202 203
	.destroy =		ipv4_dst_destroy,
	.ifdown =		ipv4_dst_ifdown,
	.negative_advice =	ipv4_negative_advice,
	.link_failure =		ipv4_link_failure,
	.update_pmtu =		ip_rt_update_pmtu,
204
	.local_out =		__ip_local_out,
205
	.neigh_lookup =		ipv4_neigh_lookup,
L
Linus Torvalds 已提交
206 207 208 209
};

#define ECN_OR_COST(class)	TC_PRIO_##class

210
const __u8 ip_tos2prio[16] = {
L
Linus Torvalds 已提交
211
	TC_PRIO_BESTEFFORT,
D
Dan Siemon 已提交
212
	ECN_OR_COST(BESTEFFORT),
L
Linus Torvalds 已提交
213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244
	TC_PRIO_BESTEFFORT,
	ECN_OR_COST(BESTEFFORT),
	TC_PRIO_BULK,
	ECN_OR_COST(BULK),
	TC_PRIO_BULK,
	ECN_OR_COST(BULK),
	TC_PRIO_INTERACTIVE,
	ECN_OR_COST(INTERACTIVE),
	TC_PRIO_INTERACTIVE,
	ECN_OR_COST(INTERACTIVE),
	TC_PRIO_INTERACTIVE_BULK,
	ECN_OR_COST(INTERACTIVE_BULK),
	TC_PRIO_INTERACTIVE_BULK,
	ECN_OR_COST(INTERACTIVE_BULK)
};


/*
 * Route cache.
 */

/* The locking scheme is rather straight forward:
 *
 * 1) Read-Copy Update protects the buckets of the central route hash.
 * 2) Only writers remove entries, and they hold the lock
 *    as they look at rtable reference counts.
 * 3) Only readers acquire references to rtable entries,
 *    they do so with atomic increments and with the
 *    lock held.
 */

struct rt_hash_bucket {
245
	struct rtable __rcu	*chain;
246
};
247

248 249
#if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK) || \
	defined(CONFIG_PROVE_LOCKING)
250 251 252
/*
 * Instead of using one spinlock for each rt_hash_bucket, we use a table of spinlocks
 * The size of this table is a power of two and depends on the number of CPUS.
253
 * (on lockdep we have a quite big spinlock_t, so keep the size down there)
254
 */
255 256
#ifdef CONFIG_LOCKDEP
# define RT_HASH_LOCK_SZ	256
257
#else
258 259 260 261 262 263 264 265 266 267 268
# if NR_CPUS >= 32
#  define RT_HASH_LOCK_SZ	4096
# elif NR_CPUS >= 16
#  define RT_HASH_LOCK_SZ	2048
# elif NR_CPUS >= 8
#  define RT_HASH_LOCK_SZ	1024
# elif NR_CPUS >= 4
#  define RT_HASH_LOCK_SZ	512
# else
#  define RT_HASH_LOCK_SZ	256
# endif
269 270 271 272
#endif

static spinlock_t	*rt_hash_locks;
# define rt_hash_lock_addr(slot) &rt_hash_locks[(slot) & (RT_HASH_LOCK_SZ - 1)]
273 274 275 276 277 278 279 280 281 282 283 284 285

static __init void rt_hash_lock_init(void)
{
	int i;

	rt_hash_locks = kmalloc(sizeof(spinlock_t) * RT_HASH_LOCK_SZ,
			GFP_KERNEL);
	if (!rt_hash_locks)
		panic("IP: failed to allocate rt_hash_locks\n");

	for (i = 0; i < RT_HASH_LOCK_SZ; i++)
		spin_lock_init(&rt_hash_locks[i]);
}
286 287
#else
# define rt_hash_lock_addr(slot) NULL
288 289 290 291

static inline void rt_hash_lock_init(void)
{
}
292
#endif
L
Linus Torvalds 已提交
293

294 295 296
static struct rt_hash_bucket 	*rt_hash_table __read_mostly;
static unsigned			rt_hash_mask __read_mostly;
static unsigned int		rt_hash_log  __read_mostly;
L
Linus Torvalds 已提交
297

298
static DEFINE_PER_CPU(struct rt_cache_stat, rt_cache_stat);
299
#define RT_CACHE_STAT_INC(field) __this_cpu_inc(rt_cache_stat.field)
L
Linus Torvalds 已提交
300

301
static inline unsigned int rt_hash(__be32 daddr, __be32 saddr, int idx,
302
				   int genid)
L
Linus Torvalds 已提交
303
{
304
	return jhash_3words((__force u32)daddr, (__force u32)saddr,
305
			    idx, genid)
306
		& rt_hash_mask;
L
Linus Torvalds 已提交
307 308
}

309 310 311 312 313
static inline int rt_genid(struct net *net)
{
	return atomic_read(&net->ipv4.rt_genid);
}

L
Linus Torvalds 已提交
314 315
#ifdef CONFIG_PROC_FS
struct rt_cache_iter_state {
316
	struct seq_net_private p;
L
Linus Torvalds 已提交
317
	int bucket;
318
	int genid;
L
Linus Torvalds 已提交
319 320
};

321
static struct rtable *rt_cache_get_first(struct seq_file *seq)
L
Linus Torvalds 已提交
322
{
323
	struct rt_cache_iter_state *st = seq->private;
L
Linus Torvalds 已提交
324 325 326
	struct rtable *r = NULL;

	for (st->bucket = rt_hash_mask; st->bucket >= 0; --st->bucket) {
327
		if (!rcu_dereference_raw(rt_hash_table[st->bucket].chain))
328
			continue;
L
Linus Torvalds 已提交
329
		rcu_read_lock_bh();
330
		r = rcu_dereference_bh(rt_hash_table[st->bucket].chain);
331
		while (r) {
332
			if (dev_net(r->dst.dev) == seq_file_net(seq) &&
333
			    r->rt_genid == st->genid)
334
				return r;
335
			r = rcu_dereference_bh(r->dst.rt_next);
336
		}
L
Linus Torvalds 已提交
337 338
		rcu_read_unlock_bh();
	}
339
	return r;
L
Linus Torvalds 已提交
340 341
}

342
static struct rtable *__rt_cache_get_next(struct seq_file *seq,
343
					  struct rtable *r)
L
Linus Torvalds 已提交
344
{
345
	struct rt_cache_iter_state *st = seq->private;
346

347
	r = rcu_dereference_bh(r->dst.rt_next);
L
Linus Torvalds 已提交
348 349
	while (!r) {
		rcu_read_unlock_bh();
350 351 352
		do {
			if (--st->bucket < 0)
				return NULL;
353
		} while (!rcu_dereference_raw(rt_hash_table[st->bucket].chain));
L
Linus Torvalds 已提交
354
		rcu_read_lock_bh();
355
		r = rcu_dereference_bh(rt_hash_table[st->bucket].chain);
L
Linus Torvalds 已提交
356
	}
357
	return r;
L
Linus Torvalds 已提交
358 359
}

360
static struct rtable *rt_cache_get_next(struct seq_file *seq,
361 362
					struct rtable *r)
{
363 364
	struct rt_cache_iter_state *st = seq->private;
	while ((r = __rt_cache_get_next(seq, r)) != NULL) {
365
		if (dev_net(r->dst.dev) != seq_file_net(seq))
366
			continue;
367 368 369 370 371 372
		if (r->rt_genid == st->genid)
			break;
	}
	return r;
}

373
static struct rtable *rt_cache_get_idx(struct seq_file *seq, loff_t pos)
L
Linus Torvalds 已提交
374
{
375
	struct rtable *r = rt_cache_get_first(seq);
L
Linus Torvalds 已提交
376 377

	if (r)
378
		while (pos && (r = rt_cache_get_next(seq, r)))
L
Linus Torvalds 已提交
379 380 381 382 383 384
			--pos;
	return pos ? NULL : r;
}

static void *rt_cache_seq_start(struct seq_file *seq, loff_t *pos)
{
385 386
	struct rt_cache_iter_state *st = seq->private;
	if (*pos)
387
		return rt_cache_get_idx(seq, *pos - 1);
388
	st->genid = rt_genid(seq_file_net(seq));
389
	return SEQ_START_TOKEN;
L
Linus Torvalds 已提交
390 391 392 393
}

static void *rt_cache_seq_next(struct seq_file *seq, void *v, loff_t *pos)
{
394
	struct rtable *r;
L
Linus Torvalds 已提交
395 396

	if (v == SEQ_START_TOKEN)
397
		r = rt_cache_get_first(seq);
L
Linus Torvalds 已提交
398
	else
399
		r = rt_cache_get_next(seq, v);
L
Linus Torvalds 已提交
400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418
	++*pos;
	return r;
}

static void rt_cache_seq_stop(struct seq_file *seq, void *v)
{
	if (v && v != SEQ_START_TOKEN)
		rcu_read_unlock_bh();
}

static int rt_cache_seq_show(struct seq_file *seq, void *v)
{
	if (v == SEQ_START_TOKEN)
		seq_printf(seq, "%-127s\n",
			   "Iface\tDestination\tGateway \tFlags\t\tRefCnt\tUse\t"
			   "Metric\tSource\t\tMTU\tWindow\tIRTT\tTOS\tHHRef\t"
			   "HHUptod\tSpecDst");
	else {
		struct rtable *r = v;
419
		struct neighbour *n;
420
		int len;
L
Linus Torvalds 已提交
421

422
		n = dst_get_neighbour(&r->dst);
423 424
		seq_printf(seq, "%s\t%08X\t%08X\t%8X\t%d\t%u\t%d\t"
			      "%08X\t%d\t%u\t%u\t%02X\t%d\t%1d\t%08X%n",
425
			r->dst.dev ? r->dst.dev->name : "*",
426 427
			(__force u32)r->rt_dst,
			(__force u32)r->rt_gateway,
428 429
			r->rt_flags, atomic_read(&r->dst.__refcnt),
			r->dst.__use, 0, (__force u32)r->rt_src,
430
			dst_metric_advmss(&r->dst) + 40,
431 432 433
			dst_metric(&r->dst, RTAX_WINDOW),
			(int)((dst_metric(&r->dst, RTAX_RTT) >> 3) +
			      dst_metric(&r->dst, RTAX_RTTVAR)),
434
			r->rt_key_tos,
435
			-1,
436
			(n && (n->nud_state & NUD_CONNECTED)) ? 1 : 0,
437 438 439
			r->rt_spec_dst, &len);

		seq_printf(seq, "%*s\n", 127 - len, "");
440 441
	}
	return 0;
L
Linus Torvalds 已提交
442 443
}

444
static const struct seq_operations rt_cache_seq_ops = {
L
Linus Torvalds 已提交
445 446 447 448 449 450 451 452
	.start  = rt_cache_seq_start,
	.next   = rt_cache_seq_next,
	.stop   = rt_cache_seq_stop,
	.show   = rt_cache_seq_show,
};

static int rt_cache_seq_open(struct inode *inode, struct file *file)
{
453
	return seq_open_net(inode, file, &rt_cache_seq_ops,
454
			sizeof(struct rt_cache_iter_state));
L
Linus Torvalds 已提交
455 456
}

457
static const struct file_operations rt_cache_seq_fops = {
L
Linus Torvalds 已提交
458 459 460 461
	.owner	 = THIS_MODULE,
	.open	 = rt_cache_seq_open,
	.read	 = seq_read,
	.llseek	 = seq_lseek,
462
	.release = seq_release_net,
L
Linus Torvalds 已提交
463 464 465 466 467 468 469 470 471 472
};


static void *rt_cpu_seq_start(struct seq_file *seq, loff_t *pos)
{
	int cpu;

	if (*pos == 0)
		return SEQ_START_TOKEN;

473
	for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
L
Linus Torvalds 已提交
474 475 476
		if (!cpu_possible(cpu))
			continue;
		*pos = cpu+1;
477
		return &per_cpu(rt_cache_stat, cpu);
L
Linus Torvalds 已提交
478 479 480 481 482 483 484 485
	}
	return NULL;
}

static void *rt_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
{
	int cpu;

486
	for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
L
Linus Torvalds 已提交
487 488 489
		if (!cpu_possible(cpu))
			continue;
		*pos = cpu+1;
490
		return &per_cpu(rt_cache_stat, cpu);
L
Linus Torvalds 已提交
491 492
	}
	return NULL;
493

L
Linus Torvalds 已提交
494 495 496 497 498 499 500 501 502 503 504 505
}

static void rt_cpu_seq_stop(struct seq_file *seq, void *v)
{

}

static int rt_cpu_seq_show(struct seq_file *seq, void *v)
{
	struct rt_cache_stat *st = v;

	if (v == SEQ_START_TOKEN) {
506
		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");
L
Linus Torvalds 已提交
507 508
		return 0;
	}
509

L
Linus Torvalds 已提交
510 511
	seq_printf(seq,"%08x  %08x %08x %08x %08x %08x %08x %08x "
		   " %08x %08x %08x %08x %08x %08x %08x %08x %08x \n",
512
		   dst_entries_get_slow(&ipv4_dst_ops),
L
Linus Torvalds 已提交
513 514 515 516 517 518 519 520 521 522
		   st->in_hit,
		   st->in_slow_tot,
		   st->in_slow_mc,
		   st->in_no_route,
		   st->in_brd,
		   st->in_martian_dst,
		   st->in_martian_src,

		   st->out_hit,
		   st->out_slow_tot,
523
		   st->out_slow_mc,
L
Linus Torvalds 已提交
524 525 526 527 528 529 530 531 532 533 534

		   st->gc_total,
		   st->gc_ignored,
		   st->gc_goal_miss,
		   st->gc_dst_overflow,
		   st->in_hlist_search,
		   st->out_hlist_search
		);
	return 0;
}

535
static const struct seq_operations rt_cpu_seq_ops = {
L
Linus Torvalds 已提交
536 537 538 539 540 541 542 543 544 545 546 547
	.start  = rt_cpu_seq_start,
	.next   = rt_cpu_seq_next,
	.stop   = rt_cpu_seq_stop,
	.show   = rt_cpu_seq_show,
};


static int rt_cpu_seq_open(struct inode *inode, struct file *file)
{
	return seq_open(file, &rt_cpu_seq_ops);
}

548
static const struct file_operations rt_cpu_seq_fops = {
L
Linus Torvalds 已提交
549 550 551 552 553 554 555
	.owner	 = THIS_MODULE,
	.open	 = rt_cpu_seq_open,
	.read	 = seq_read,
	.llseek	 = seq_lseek,
	.release = seq_release,
};

556
#ifdef CONFIG_IP_ROUTE_CLASSID
557
static int rt_acct_proc_show(struct seq_file *m, void *v)
558
{
559 560 561 562 563 564 565 566 567 568 569 570 571 572 573
	struct ip_rt_acct *dst, *src;
	unsigned int i, j;

	dst = kcalloc(256, sizeof(struct ip_rt_acct), GFP_KERNEL);
	if (!dst)
		return -ENOMEM;

	for_each_possible_cpu(i) {
		src = (struct ip_rt_acct *)per_cpu_ptr(ip_rt_acct, i);
		for (j = 0; j < 256; j++) {
			dst[j].o_bytes   += src[j].o_bytes;
			dst[j].o_packets += src[j].o_packets;
			dst[j].i_bytes   += src[j].i_bytes;
			dst[j].i_packets += src[j].i_packets;
		}
574 575
	}

576 577 578 579
	seq_write(m, dst, 256 * sizeof(struct ip_rt_acct));
	kfree(dst);
	return 0;
}
580

581 582 583
static int rt_acct_proc_open(struct inode *inode, struct file *file)
{
	return single_open(file, rt_acct_proc_show, NULL);
584
}
585 586 587 588 589 590 591 592

static const struct file_operations rt_acct_proc_fops = {
	.owner		= THIS_MODULE,
	.open		= rt_acct_proc_open,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
};
593
#endif
594

595
static int __net_init ip_rt_do_proc_init(struct net *net)
596 597 598 599 600 601 602 603
{
	struct proc_dir_entry *pde;

	pde = proc_net_fops_create(net, "rt_cache", S_IRUGO,
			&rt_cache_seq_fops);
	if (!pde)
		goto err1;

604 605
	pde = proc_create("rt_cache", S_IRUGO,
			  net->proc_net_stat, &rt_cpu_seq_fops);
606 607 608
	if (!pde)
		goto err2;

609
#ifdef CONFIG_IP_ROUTE_CLASSID
610
	pde = proc_create("rt_acct", 0, net->proc_net, &rt_acct_proc_fops);
611 612 613 614 615
	if (!pde)
		goto err3;
#endif
	return 0;

616
#ifdef CONFIG_IP_ROUTE_CLASSID
617 618 619 620 621 622 623 624
err3:
	remove_proc_entry("rt_cache", net->proc_net_stat);
#endif
err2:
	remove_proc_entry("rt_cache", net->proc_net);
err1:
	return -ENOMEM;
}
625 626 627 628 629

static void __net_exit ip_rt_do_proc_exit(struct net *net)
{
	remove_proc_entry("rt_cache", net->proc_net_stat);
	remove_proc_entry("rt_cache", net->proc_net);
630
#ifdef CONFIG_IP_ROUTE_CLASSID
631
	remove_proc_entry("rt_acct", net->proc_net);
632
#endif
633 634 635 636 637 638 639 640 641 642 643 644
}

static struct pernet_operations ip_rt_proc_ops __net_initdata =  {
	.init = ip_rt_do_proc_init,
	.exit = ip_rt_do_proc_exit,
};

static int __init ip_rt_proc_init(void)
{
	return register_pernet_subsys(&ip_rt_proc_ops);
}

645
#else
646
static inline int ip_rt_proc_init(void)
647 648 649
{
	return 0;
}
L
Linus Torvalds 已提交
650
#endif /* CONFIG_PROC_FS */
651

S
Stephen Hemminger 已提交
652
static inline void rt_free(struct rtable *rt)
L
Linus Torvalds 已提交
653
{
654
	call_rcu_bh(&rt->dst.rcu_head, dst_rcu_free);
L
Linus Torvalds 已提交
655 656
}

S
Stephen Hemminger 已提交
657
static inline void rt_drop(struct rtable *rt)
L
Linus Torvalds 已提交
658 659
{
	ip_rt_put(rt);
660
	call_rcu_bh(&rt->dst.rcu_head, dst_rcu_free);
L
Linus Torvalds 已提交
661 662
}

S
Stephen Hemminger 已提交
663
static inline int rt_fast_clean(struct rtable *rth)
L
Linus Torvalds 已提交
664 665 666 667
{
	/* Kill broadcast/multicast entries very aggresively, if they
	   collide in hash table with more useful entries */
	return (rth->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) &&
668
		rt_is_input_route(rth) && rth->dst.rt_next;
L
Linus Torvalds 已提交
669 670
}

S
Stephen Hemminger 已提交
671
static inline int rt_valuable(struct rtable *rth)
L
Linus Torvalds 已提交
672 673
{
	return (rth->rt_flags & (RTCF_REDIRECTED | RTCF_NOTIFY)) ||
674
		(rth->peer && rth->peer->pmtu_expires);
L
Linus Torvalds 已提交
675 676 677 678 679 680 681
}

static int rt_may_expire(struct rtable *rth, unsigned long tmo1, unsigned long tmo2)
{
	unsigned long age;
	int ret = 0;

682
	if (atomic_read(&rth->dst.__refcnt))
L
Linus Torvalds 已提交
683 684
		goto out;

685
	age = jiffies - rth->dst.lastuse;
L
Linus Torvalds 已提交
686 687 688 689 690 691 692 693 694 695 696 697 698 699
	if ((age <= tmo1 && !rt_fast_clean(rth)) ||
	    (age <= tmo2 && rt_valuable(rth)))
		goto out;
	ret = 1;
out:	return ret;
}

/* Bits of score are:
 * 31: very valuable
 * 30: not quite useless
 * 29..0: usage counter
 */
static inline u32 rt_score(struct rtable *rt)
{
700
	u32 score = jiffies - rt->dst.lastuse;
L
Linus Torvalds 已提交
701 702 703 704 705 706

	score = ~score & ~(3<<30);

	if (rt_valuable(rt))
		score |= (1<<31);

707
	if (rt_is_output_route(rt) ||
L
Linus Torvalds 已提交
708 709 710 711 712 713
	    !(rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST|RTCF_LOCAL)))
		score |= (1<<30);

	return score;
}

714 715 716 717 718 719
static inline bool rt_caching(const struct net *net)
{
	return net->ipv4.current_rt_cache_rebuild_count <=
		net->ipv4.sysctl_rt_cache_rebuild_count;
}

720 721
static inline bool compare_hash_inputs(const struct rtable *rt1,
				       const struct rtable *rt2)
722
{
723 724 725
	return ((((__force u32)rt1->rt_key_dst ^ (__force u32)rt2->rt_key_dst) |
		((__force u32)rt1->rt_key_src ^ (__force u32)rt2->rt_key_src) |
		(rt1->rt_iif ^ rt2->rt_iif)) == 0);
726 727
}

728
static inline int compare_keys(struct rtable *rt1, struct rtable *rt2)
L
Linus Torvalds 已提交
729
{
730 731 732
	return (((__force u32)rt1->rt_key_dst ^ (__force u32)rt2->rt_key_dst) |
		((__force u32)rt1->rt_key_src ^ (__force u32)rt2->rt_key_src) |
		(rt1->rt_mark ^ rt2->rt_mark) |
733
		(rt1->rt_key_tos ^ rt2->rt_key_tos) |
734
		(rt1->rt_route_iif ^ rt2->rt_route_iif) |
735 736
		(rt1->rt_oif ^ rt2->rt_oif) |
		(rt1->rt_iif ^ rt2->rt_iif)) == 0;
L
Linus Torvalds 已提交
737 738
}

739 740
static inline int compare_netns(struct rtable *rt1, struct rtable *rt2)
{
741
	return net_eq(dev_net(rt1->dst.dev), dev_net(rt2->dst.dev));
742 743
}

744 745
static inline int rt_is_expired(struct rtable *rth)
{
746
	return rth->rt_genid != rt_genid(dev_net(rth->dst.dev));
747 748
}

749 750 751 752 753
/*
 * Perform a full scan of hash table and free all entries.
 * Can be called by a softirq or a process.
 * In the later case, we want to be reschedule if necessary
 */
754
static void rt_do_flush(struct net *net, int process_context)
755 756 757 758 759
{
	unsigned int i;
	struct rtable *rth, *next;

	for (i = 0; i <= rt_hash_mask; i++) {
760 761 762
		struct rtable __rcu **pprev;
		struct rtable *list;

763 764
		if (process_context && need_resched())
			cond_resched();
765
		rth = rcu_dereference_raw(rt_hash_table[i].chain);
766 767 768 769
		if (!rth)
			continue;

		spin_lock_bh(rt_hash_lock_addr(i));
770

771 772 773
		list = NULL;
		pprev = &rt_hash_table[i].chain;
		rth = rcu_dereference_protected(*pprev,
774
			lockdep_is_held(rt_hash_lock_addr(i)));
775

776 777
		while (rth) {
			next = rcu_dereference_protected(rth->dst.rt_next,
778
				lockdep_is_held(rt_hash_lock_addr(i)));
779 780 781 782 783 784

			if (!net ||
			    net_eq(dev_net(rth->dst.dev), net)) {
				rcu_assign_pointer(*pprev, next);
				rcu_assign_pointer(rth->dst.rt_next, list);
				list = rth;
785
			} else {
786
				pprev = &rth->dst.rt_next;
787
			}
788
			rth = next;
789
		}
790

791 792
		spin_unlock_bh(rt_hash_lock_addr(i));

793 794 795
		for (; list; list = next) {
			next = rcu_dereference_protected(list->dst.rt_next, 1);
			rt_free(list);
796 797 798 799
		}
	}
}

800 801 802 803 804 805 806 807 808 809 810
/*
 * While freeing expired entries, we compute average chain length
 * and standard deviation, using fixed-point arithmetic.
 * This to have an estimation of rt_chain_length_max
 *  rt_chain_length_max = max(elasticity, AVG + 4*SD)
 * We use 3 bits for frational part, and 29 (or 61) for magnitude.
 */

#define FRACT_BITS 3
#define ONE (1UL << FRACT_BITS)

E
Eric Dumazet 已提交
811 812 813 814 815 816 817 818 819 820 821 822
/*
 * Given a hash chain and an item in this hash chain,
 * find if a previous entry has the same hash_inputs
 * (but differs on tos, mark or oif)
 * Returns 0 if an alias is found.
 * Returns ONE if rth has no alias before itself.
 */
static int has_noalias(const struct rtable *head, const struct rtable *rth)
{
	const struct rtable *aux = head;

	while (aux != rth) {
823
		if (compare_hash_inputs(aux, rth))
E
Eric Dumazet 已提交
824
			return 0;
825
		aux = rcu_dereference_protected(aux->dst.rt_next, 1);
E
Eric Dumazet 已提交
826 827 828 829
	}
	return ONE;
}

830
/*
L
Lucas De Marchi 已提交
831
 * Perturbation of rt_genid by a small quantity [1..256]
832 833 834
 * Using 8 bits of shuffling ensure we can call rt_cache_invalidate()
 * many times (2^24) without giving recent rt_genid.
 * Jenkins hash is strong enough that litle changes of rt_genid are OK.
L
Linus Torvalds 已提交
835
 */
836
static void rt_cache_invalidate(struct net *net)
L
Linus Torvalds 已提交
837
{
838
	unsigned char shuffle;
L
Linus Torvalds 已提交
839

840
	get_random_bytes(&shuffle, sizeof(shuffle));
841
	atomic_add(shuffle + 1U, &net->ipv4.rt_genid);
L
Linus Torvalds 已提交
842 843
}

844 845 846 847
/*
 * delay < 0  : invalidate cache (fast : entries will be deleted later)
 * delay >= 0 : invalidate & flush cache (can be long)
 */
848
void rt_cache_flush(struct net *net, int delay)
L
Linus Torvalds 已提交
849
{
850
	rt_cache_invalidate(net);
851
	if (delay >= 0)
852
		rt_do_flush(net, !in_softirq());
L
Linus Torvalds 已提交
853 854
}

855
/* Flush previous cache invalidated entries from the cache */
856
void rt_cache_flush_batch(struct net *net)
857
{
858
	rt_do_flush(net, !in_softirq());
859 860
}

861 862
static void rt_emergency_hash_rebuild(struct net *net)
{
863
	if (net_ratelimit())
864
		printk(KERN_WARNING "Route hash chain too long!\n");
865
	rt_cache_invalidate(net);
866 867
}

L
Linus Torvalds 已提交
868 869 870 871 872 873 874 875 876 877 878 879 880
/*
   Short description of GC goals.

   We want to build algorithm, which will keep routing cache
   at some equilibrium point, when number of aged off entries
   is kept approximately equal to newly generated ones.

   Current expiration strength is variable "expire".
   We try to adjust it dynamically, so that if networking
   is idle expires is large enough to keep enough of warm entries,
   and when load increases it reduces to limit cache size.
 */

881
static int rt_garbage_collect(struct dst_ops *ops)
L
Linus Torvalds 已提交
882 883 884 885 886
{
	static unsigned long expire = RT_GC_TIMEOUT;
	static unsigned long last_gc;
	static int rover;
	static int equilibrium;
887 888
	struct rtable *rth;
	struct rtable __rcu **rthp;
L
Linus Torvalds 已提交
889 890
	unsigned long now = jiffies;
	int goal;
891
	int entries = dst_entries_get_fast(&ipv4_dst_ops);
L
Linus Torvalds 已提交
892 893 894 895 896 897 898 899 900

	/*
	 * Garbage collection is pretty expensive,
	 * do not make it too frequently.
	 */

	RT_CACHE_STAT_INC(gc_total);

	if (now - last_gc < ip_rt_gc_min_interval &&
901
	    entries < ip_rt_max_size) {
L
Linus Torvalds 已提交
902 903 904 905
		RT_CACHE_STAT_INC(gc_ignored);
		goto out;
	}

906
	entries = dst_entries_get_slow(&ipv4_dst_ops);
L
Linus Torvalds 已提交
907
	/* Calculate number of entries, which we want to expire now. */
908
	goal = entries - (ip_rt_gc_elasticity << rt_hash_log);
L
Linus Torvalds 已提交
909 910 911
	if (goal <= 0) {
		if (equilibrium < ipv4_dst_ops.gc_thresh)
			equilibrium = ipv4_dst_ops.gc_thresh;
912
		goal = entries - equilibrium;
L
Linus Torvalds 已提交
913
		if (goal > 0) {
914
			equilibrium += min_t(unsigned int, goal >> 1, rt_hash_mask + 1);
915
			goal = entries - equilibrium;
L
Linus Torvalds 已提交
916 917 918 919 920
		}
	} else {
		/* We are in dangerous area. Try to reduce cache really
		 * aggressively.
		 */
921
		goal = max_t(unsigned int, goal >> 1, rt_hash_mask + 1);
922
		equilibrium = entries - goal;
L
Linus Torvalds 已提交
923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940
	}

	if (now - last_gc >= ip_rt_gc_min_interval)
		last_gc = now;

	if (goal <= 0) {
		equilibrium += goal;
		goto work_done;
	}

	do {
		int i, k;

		for (i = rt_hash_mask, k = rover; i >= 0; i--) {
			unsigned long tmo = expire;

			k = (k + 1) & rt_hash_mask;
			rthp = &rt_hash_table[k].chain;
941
			spin_lock_bh(rt_hash_lock_addr(k));
942 943
			while ((rth = rcu_dereference_protected(*rthp,
					lockdep_is_held(rt_hash_lock_addr(k)))) != NULL) {
944
				if (!rt_is_expired(rth) &&
945
					!rt_may_expire(rth, tmo, expire)) {
L
Linus Torvalds 已提交
946
					tmo >>= 1;
947
					rthp = &rth->dst.rt_next;
L
Linus Torvalds 已提交
948 949
					continue;
				}
950
				*rthp = rth->dst.rt_next;
L
Linus Torvalds 已提交
951 952 953
				rt_free(rth);
				goal--;
			}
954
			spin_unlock_bh(rt_hash_lock_addr(k));
L
Linus Torvalds 已提交
955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978
			if (goal <= 0)
				break;
		}
		rover = k;

		if (goal <= 0)
			goto work_done;

		/* Goal is not achieved. We stop process if:

		   - if expire reduced to zero. Otherwise, expire is halfed.
		   - if table is not full.
		   - if we are called from interrupt.
		   - jiffies check is just fallback/debug loop breaker.
		     We will not spin here for long time in any case.
		 */

		RT_CACHE_STAT_INC(gc_goal_miss);

		if (expire == 0)
			break;

		expire >>= 1;

979
		if (dst_entries_get_fast(&ipv4_dst_ops) < ip_rt_max_size)
L
Linus Torvalds 已提交
980 981 982
			goto out;
	} while (!in_softirq() && time_before_eq(jiffies, now));

983 984 985
	if (dst_entries_get_fast(&ipv4_dst_ops) < ip_rt_max_size)
		goto out;
	if (dst_entries_get_slow(&ipv4_dst_ops) < ip_rt_max_size)
L
Linus Torvalds 已提交
986 987 988 989 990 991 992 993 994
		goto out;
	if (net_ratelimit())
		printk(KERN_WARNING "dst cache overflow\n");
	RT_CACHE_STAT_INC(gc_dst_overflow);
	return 1;

work_done:
	expire += ip_rt_gc_min_interval;
	if (expire > ip_rt_gc_timeout ||
995 996
	    dst_entries_get_fast(&ipv4_dst_ops) < ipv4_dst_ops.gc_thresh ||
	    dst_entries_get_slow(&ipv4_dst_ops) < ipv4_dst_ops.gc_thresh)
L
Linus Torvalds 已提交
997 998 999 1000
		expire = ip_rt_gc_timeout;
out:	return 0;
}

E
Eric Dumazet 已提交
1001 1002 1003 1004 1005 1006 1007 1008 1009 1010
/*
 * Returns number of entries in a hash chain that have different hash_inputs
 */
static int slow_chain_length(const struct rtable *head)
{
	int length = 0;
	const struct rtable *rth = head;

	while (rth) {
		length += has_noalias(head, rth);
1011
		rth = rcu_dereference_protected(rth->dst.rt_next, 1);
E
Eric Dumazet 已提交
1012 1013 1014 1015
	}
	return length >> FRACT_BITS;
}

1016
static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst, const void *daddr)
D
David Miller 已提交
1017 1018
{
	struct neigh_table *tbl = &arp_tbl;
1019 1020 1021
	static const __be32 inaddr_any = 0;
	struct net_device *dev = dst->dev;
	const __be32 *pkey = daddr;
D
David Miller 已提交
1022 1023 1024 1025 1026 1027 1028
	struct neighbour *n;

#if defined(CONFIG_ATM_CLIP) || defined(CONFIG_ATM_CLIP_MODULE)
	if (dev->type == ARPHRD_ATM)
		tbl = clip_tbl_hook;
#endif
	if (dev->flags & (IFF_LOOPBACK | IFF_POINTOPOINT))
1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039
		pkey = &inaddr_any;

	n = __ipv4_neigh_lookup(tbl, dev, *(__force u32 *)pkey);
	if (n)
		return n;
	return neigh_create(tbl, pkey, dev);
}

static int rt_bind_neighbour(struct rtable *rt)
{
	struct neighbour *n = ipv4_neigh_lookup(&rt->dst, &rt->rt_gateway);
D
David Miller 已提交
1040 1041
	if (IS_ERR(n))
		return PTR_ERR(n);
1042
	dst_set_neighbour(&rt->dst, n);
D
David Miller 已提交
1043 1044 1045 1046

	return 0;
}

1047 1048
static struct rtable *rt_intern_hash(unsigned hash, struct rtable *rt,
				     struct sk_buff *skb, int ifindex)
L
Linus Torvalds 已提交
1049
{
1050 1051
	struct rtable	*rth, *cand;
	struct rtable __rcu **rthp, **candp;
L
Linus Torvalds 已提交
1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063
	unsigned long	now;
	u32 		min_score;
	int		chain_length;
	int attempts = !in_softirq();

restart:
	chain_length = 0;
	min_score = ~(u32)0;
	cand = NULL;
	candp = NULL;
	now = jiffies;

1064
	if (!rt_caching(dev_net(rt->dst.dev))) {
1065 1066 1067 1068 1069 1070 1071 1072
		/*
		 * If we're not caching, just tell the caller we
		 * were successful and don't touch the route.  The
		 * caller hold the sole reference to the cache entry, and
		 * it will be released when the caller is done with it.
		 * If we drop it here, the callers have no way to resolve routes
		 * when we're not caching.  Instead, just point *rp at rt, so
		 * the caller gets a single use out of the route
1073 1074 1075
		 * Note that we do rt_free on this new route entry, so that
		 * once its refcount hits zero, we are still able to reap it
		 * (Thanks Alexey)
E
Eric Dumazet 已提交
1076 1077 1078
		 * Note: To avoid expensive rcu stuff for this uncached dst,
		 * we set DST_NOCACHE so that dst_release() can free dst without
		 * waiting a grace period.
1079
		 */
1080

E
Eric Dumazet 已提交
1081
		rt->dst.flags |= DST_NOCACHE;
1082
		if (rt->rt_type == RTN_UNICAST || rt_is_output_route(rt)) {
D
David Miller 已提交
1083
			int err = rt_bind_neighbour(rt);
1084 1085 1086 1087
			if (err) {
				if (net_ratelimit())
					printk(KERN_WARNING
					    "Neighbour table failure & not caching routes.\n");
E
Eric Dumazet 已提交
1088
				ip_rt_put(rt);
1089
				return ERR_PTR(err);
1090 1091 1092 1093
			}
		}

		goto skip_hashing;
1094 1095
	}

L
Linus Torvalds 已提交
1096 1097
	rthp = &rt_hash_table[hash].chain;

1098
	spin_lock_bh(rt_hash_lock_addr(hash));
1099 1100
	while ((rth = rcu_dereference_protected(*rthp,
			lockdep_is_held(rt_hash_lock_addr(hash)))) != NULL) {
1101
		if (rt_is_expired(rth)) {
1102
			*rthp = rth->dst.rt_next;
1103 1104 1105
			rt_free(rth);
			continue;
		}
1106
		if (compare_keys(rth, rt) && compare_netns(rth, rt)) {
L
Linus Torvalds 已提交
1107
			/* Put it first */
1108
			*rthp = rth->dst.rt_next;
L
Linus Torvalds 已提交
1109 1110 1111 1112 1113
			/*
			 * Since lookup is lockfree, the deletion
			 * must be visible to another weakly ordered CPU before
			 * the insertion at the start of the hash chain.
			 */
1114
			rcu_assign_pointer(rth->dst.rt_next,
L
Linus Torvalds 已提交
1115 1116 1117 1118 1119 1120 1121
					   rt_hash_table[hash].chain);
			/*
			 * Since lookup is lockfree, the update writes
			 * must be ordered for consistency on SMP.
			 */
			rcu_assign_pointer(rt_hash_table[hash].chain, rth);

1122
			dst_use(&rth->dst, now);
1123
			spin_unlock_bh(rt_hash_lock_addr(hash));
L
Linus Torvalds 已提交
1124 1125

			rt_drop(rt);
1126
			if (skb)
1127
				skb_dst_set(skb, &rth->dst);
1128
			return rth;
L
Linus Torvalds 已提交
1129 1130
		}

1131
		if (!atomic_read(&rth->dst.__refcnt)) {
L
Linus Torvalds 已提交
1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142
			u32 score = rt_score(rth);

			if (score <= min_score) {
				cand = rth;
				candp = rthp;
				min_score = score;
			}
		}

		chain_length++;

1143
		rthp = &rth->dst.rt_next;
L
Linus Torvalds 已提交
1144 1145 1146 1147 1148 1149 1150 1151 1152 1153
	}

	if (cand) {
		/* ip_rt_gc_elasticity used to be average length of chain
		 * length, when exceeded gc becomes really aggressive.
		 *
		 * The second limit is less certain. At the moment it allows
		 * only 2 entries per bucket. We will see.
		 */
		if (chain_length > ip_rt_gc_elasticity) {
1154
			*candp = cand->dst.rt_next;
L
Linus Torvalds 已提交
1155 1156
			rt_free(cand);
		}
1157
	} else {
E
Eric Dumazet 已提交
1158 1159
		if (chain_length > rt_chain_length_max &&
		    slow_chain_length(rt_hash_table[hash].chain) > rt_chain_length_max) {
1160
			struct net *net = dev_net(rt->dst.dev);
1161
			int num = ++net->ipv4.current_rt_cache_rebuild_count;
1162
			if (!rt_caching(net)) {
1163
				printk(KERN_WARNING "%s: %d rebuilds is over limit, route caching disabled\n",
1164
					rt->dst.dev->name, num);
1165
			}
1166
			rt_emergency_hash_rebuild(net);
1167 1168
			spin_unlock_bh(rt_hash_lock_addr(hash));

1169
			hash = rt_hash(rt->rt_key_dst, rt->rt_key_src,
1170 1171
					ifindex, rt_genid(net));
			goto restart;
1172
		}
L
Linus Torvalds 已提交
1173 1174 1175 1176 1177
	}

	/* Try to bind route to arp only if it is output
	   route or unicast forwarding path.
	 */
1178
	if (rt->rt_type == RTN_UNICAST || rt_is_output_route(rt)) {
D
David Miller 已提交
1179
		int err = rt_bind_neighbour(rt);
L
Linus Torvalds 已提交
1180
		if (err) {
1181
			spin_unlock_bh(rt_hash_lock_addr(hash));
L
Linus Torvalds 已提交
1182 1183 1184

			if (err != -ENOBUFS) {
				rt_drop(rt);
1185
				return ERR_PTR(err);
L
Linus Torvalds 已提交
1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196
			}

			/* Neighbour tables are full and nothing
			   can be released. Try to shrink route cache,
			   it is most likely it holds some neighbour records.
			 */
			if (attempts-- > 0) {
				int saved_elasticity = ip_rt_gc_elasticity;
				int saved_int = ip_rt_gc_min_interval;
				ip_rt_gc_elasticity	= 1;
				ip_rt_gc_min_interval	= 0;
1197
				rt_garbage_collect(&ipv4_dst_ops);
L
Linus Torvalds 已提交
1198 1199 1200 1201 1202 1203
				ip_rt_gc_min_interval	= saved_int;
				ip_rt_gc_elasticity	= saved_elasticity;
				goto restart;
			}

			if (net_ratelimit())
1204
				printk(KERN_WARNING "ipv4: Neighbour table overflow.\n");
L
Linus Torvalds 已提交
1205
			rt_drop(rt);
1206
			return ERR_PTR(-ENOBUFS);
L
Linus Torvalds 已提交
1207 1208 1209
		}
	}

1210
	rt->dst.rt_next = rt_hash_table[hash].chain;
1211

1212 1213
	/*
	 * Since lookup is lockfree, we must make sure
L
Lucas De Marchi 已提交
1214
	 * previous writes to rt are committed to memory
1215 1216
	 * before making rt visible to other CPUS.
	 */
1217
	rcu_assign_pointer(rt_hash_table[hash].chain, rt);
1218

1219
	spin_unlock_bh(rt_hash_lock_addr(hash));
1220

1221
skip_hashing:
1222
	if (skb)
1223
		skb_dst_set(skb, &rt->dst);
1224
	return rt;
L
Linus Torvalds 已提交
1225 1226
}

1227 1228 1229 1230 1231 1232 1233
static atomic_t __rt_peer_genid = ATOMIC_INIT(0);

static u32 rt_peer_genid(void)
{
	return atomic_read(&__rt_peer_genid);
}

1234
void rt_bind_peer(struct rtable *rt, __be32 daddr, int create)
L
Linus Torvalds 已提交
1235 1236 1237
{
	struct inet_peer *peer;

1238
	peer = inet_getpeer_v4(daddr, create);
L
Linus Torvalds 已提交
1239

1240
	if (peer && cmpxchg(&rt->peer, NULL, peer) != NULL)
L
Linus Torvalds 已提交
1241
		inet_putpeer(peer);
1242 1243
	else
		rt->rt_peer_genid = rt_peer_genid();
L
Linus Torvalds 已提交
1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259
}

/*
 * Peer allocation may fail only in serious out-of-memory conditions.  However
 * we still can generate some output.
 * Random ID selection looks a bit dangerous because we have no chances to
 * select ID being unique in a reasonable period of time.
 * But broken packet identifier may be better than no packet at all.
 */
static void ip_select_fb_ident(struct iphdr *iph)
{
	static DEFINE_SPINLOCK(ip_fb_id_lock);
	static u32 ip_fallback_id;
	u32 salt;

	spin_lock_bh(&ip_fb_id_lock);
1260
	salt = secure_ip_id((__force __be32)ip_fallback_id ^ iph->daddr);
L
Linus Torvalds 已提交
1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271
	iph->id = htons(salt & 0xFFFF);
	ip_fallback_id = salt;
	spin_unlock_bh(&ip_fb_id_lock);
}

void __ip_select_ident(struct iphdr *iph, struct dst_entry *dst, int more)
{
	struct rtable *rt = (struct rtable *) dst;

	if (rt) {
		if (rt->peer == NULL)
1272
			rt_bind_peer(rt, rt->rt_dst, 1);
L
Linus Torvalds 已提交
1273 1274 1275 1276 1277 1278 1279 1280 1281

		/* If peer is attached to destination, it is never detached,
		   so that we need not to grab a lock to dereference it.
		 */
		if (rt->peer) {
			iph->id = htons(inet_getid(rt->peer, more));
			return;
		}
	} else
1282
		printk(KERN_DEBUG "rt_bind_peer(0) @%p\n",
1283
		       __builtin_return_address(0));
L
Linus Torvalds 已提交
1284 1285 1286

	ip_select_fb_ident(iph);
}
E
Eric Dumazet 已提交
1287
EXPORT_SYMBOL(__ip_select_ident);
L
Linus Torvalds 已提交
1288 1289 1290

static void rt_del(unsigned hash, struct rtable *rt)
{
1291 1292
	struct rtable __rcu **rthp;
	struct rtable *aux;
L
Linus Torvalds 已提交
1293

1294
	rthp = &rt_hash_table[hash].chain;
1295
	spin_lock_bh(rt_hash_lock_addr(hash));
L
Linus Torvalds 已提交
1296
	ip_rt_put(rt);
1297 1298
	while ((aux = rcu_dereference_protected(*rthp,
			lockdep_is_held(rt_hash_lock_addr(hash)))) != NULL) {
1299
		if (aux == rt || rt_is_expired(aux)) {
1300
			*rthp = aux->dst.rt_next;
1301 1302
			rt_free(aux);
			continue;
L
Linus Torvalds 已提交
1303
		}
1304
		rthp = &aux->dst.rt_next;
1305
	}
1306
	spin_unlock_bh(rt_hash_lock_addr(hash));
L
Linus Torvalds 已提交
1307 1308
}

1309
/* called in rcu_read_lock() section */
A
Al Viro 已提交
1310 1311
void ip_rt_redirect(__be32 old_gw, __be32 daddr, __be32 new_gw,
		    __be32 saddr, struct net_device *dev)
L
Linus Torvalds 已提交
1312
{
1313
	struct in_device *in_dev = __in_dev_get_rcu(dev);
1314
	struct inet_peer *peer;
1315
	struct net *net;
L
Linus Torvalds 已提交
1316 1317 1318 1319

	if (!in_dev)
		return;

1320
	net = dev_net(dev);
1321 1322 1323
	if (new_gw == old_gw || !IN_DEV_RX_REDIRECTS(in_dev) ||
	    ipv4_is_multicast(new_gw) || ipv4_is_lbcast(new_gw) ||
	    ipv4_is_zeronet(new_gw))
L
Linus Torvalds 已提交
1324 1325 1326 1327 1328 1329 1330 1331
		goto reject_redirect;

	if (!IN_DEV_SHARED_MEDIA(in_dev)) {
		if (!inet_addr_onlink(in_dev, new_gw, old_gw))
			goto reject_redirect;
		if (IN_DEV_SEC_REDIRECTS(in_dev) && ip_fib_check_default(new_gw, dev))
			goto reject_redirect;
	} else {
1332
		if (inet_addr_type(net, new_gw) != RTN_UNICAST)
L
Linus Torvalds 已提交
1333 1334 1335
			goto reject_redirect;
	}

1336 1337 1338
	peer = inet_getpeer_v4(daddr, 1);
	if (peer) {
		peer->redirect_learned.a4 = new_gw;
1339

1340
		inet_putpeer(peer);
L
Linus Torvalds 已提交
1341

1342
		atomic_inc(&__rt_peer_genid);
L
Linus Torvalds 已提交
1343 1344 1345 1346 1347 1348
	}
	return;

reject_redirect:
#ifdef CONFIG_IP_ROUTE_VERBOSE
	if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit())
1349 1350 1351 1352
		printk(KERN_INFO "Redirect from %pI4 on %s about %pI4 ignored.\n"
			"  Advised path = %pI4 -> %pI4\n",
		       &old_gw, dev->name, &new_gw,
		       &saddr, &daddr);
L
Linus Torvalds 已提交
1353
#endif
1354
	;
L
Linus Torvalds 已提交
1355 1356
}

E
Eric Dumazet 已提交
1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373
static bool peer_pmtu_expired(struct inet_peer *peer)
{
	unsigned long orig = ACCESS_ONCE(peer->pmtu_expires);

	return orig &&
	       time_after_eq(jiffies, orig) &&
	       cmpxchg(&peer->pmtu_expires, orig, 0) == orig;
}

static bool peer_pmtu_cleaned(struct inet_peer *peer)
{
	unsigned long orig = ACCESS_ONCE(peer->pmtu_expires);

	return orig &&
	       cmpxchg(&peer->pmtu_expires, orig, 0) == orig;
}

L
Linus Torvalds 已提交
1374 1375
static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst)
{
1376
	struct rtable *rt = (struct rtable *)dst;
L
Linus Torvalds 已提交
1377 1378 1379
	struct dst_entry *ret = dst;

	if (rt) {
T
Timo Teräs 已提交
1380
		if (dst->obsolete > 0) {
L
Linus Torvalds 已提交
1381 1382
			ip_rt_put(rt);
			ret = NULL;
1383
		} else if (rt->rt_flags & RTCF_REDIRECTED) {
1384 1385
			unsigned hash = rt_hash(rt->rt_key_dst, rt->rt_key_src,
						rt->rt_oif,
1386
						rt_genid(dev_net(dst->dev)));
L
Linus Torvalds 已提交
1387 1388
			rt_del(hash, rt);
			ret = NULL;
E
Eric Dumazet 已提交
1389 1390
		} else if (rt->peer && peer_pmtu_expired(rt->peer)) {
			dst_metric_set(dst, RTAX_MTU, rt->peer->pmtu_orig);
L
Linus Torvalds 已提交
1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413
		}
	}
	return ret;
}

/*
 * Algorithm:
 *	1. The first ip_rt_redirect_number redirects are sent
 *	   with exponential backoff, then we stop sending them at all,
 *	   assuming that the host ignores our redirects.
 *	2. If we did not see packets requiring redirects
 *	   during ip_rt_redirect_silence, we assume that the host
 *	   forgot redirected route and start to send redirects again.
 *
 * This algorithm is much cheaper and more intelligent than dumb load limiting
 * in icmp.c.
 *
 * NOTE. Do not forget to inhibit load limiting for redirects (redundant)
 * and "frag. need" (breaks PMTU discovery) in icmp.c.
 */

void ip_rt_send_redirect(struct sk_buff *skb)
{
E
Eric Dumazet 已提交
1414
	struct rtable *rt = skb_rtable(skb);
1415
	struct in_device *in_dev;
1416
	struct inet_peer *peer;
1417
	int log_martians;
L
Linus Torvalds 已提交
1418

1419
	rcu_read_lock();
1420
	in_dev = __in_dev_get_rcu(rt->dst.dev);
1421 1422
	if (!in_dev || !IN_DEV_TX_REDIRECTS(in_dev)) {
		rcu_read_unlock();
L
Linus Torvalds 已提交
1423
		return;
1424 1425 1426
	}
	log_martians = IN_DEV_LOG_MARTIANS(in_dev);
	rcu_read_unlock();
L
Linus Torvalds 已提交
1427

1428
	if (!rt->peer)
1429
		rt_bind_peer(rt, rt->rt_dst, 1);
1430 1431 1432 1433 1434 1435
	peer = rt->peer;
	if (!peer) {
		icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, rt->rt_gateway);
		return;
	}

L
Linus Torvalds 已提交
1436 1437 1438
	/* No redirected packets during ip_rt_redirect_silence;
	 * reset the algorithm.
	 */
1439 1440
	if (time_after(jiffies, peer->rate_last + ip_rt_redirect_silence))
		peer->rate_tokens = 0;
L
Linus Torvalds 已提交
1441 1442

	/* Too many ignored redirects; do not send anything
1443
	 * set dst.rate_last to the last seen redirected packet.
L
Linus Torvalds 已提交
1444
	 */
1445 1446
	if (peer->rate_tokens >= ip_rt_redirect_number) {
		peer->rate_last = jiffies;
1447
		return;
L
Linus Torvalds 已提交
1448 1449 1450 1451 1452
	}

	/* Check for load limit; set rate_last to the latest sent
	 * redirect.
	 */
1453
	if (peer->rate_tokens == 0 ||
1454
	    time_after(jiffies,
1455 1456
		       (peer->rate_last +
			(ip_rt_redirect_load << peer->rate_tokens)))) {
L
Linus Torvalds 已提交
1457
		icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, rt->rt_gateway);
1458 1459
		peer->rate_last = jiffies;
		++peer->rate_tokens;
L
Linus Torvalds 已提交
1460
#ifdef CONFIG_IP_ROUTE_VERBOSE
1461
		if (log_martians &&
1462
		    peer->rate_tokens == ip_rt_redirect_number &&
L
Linus Torvalds 已提交
1463
		    net_ratelimit())
1464
			printk(KERN_WARNING "host %pI4/if%d ignores redirects for %pI4 to %pI4.\n",
1465
			       &ip_hdr(skb)->saddr, rt->rt_iif,
1466
				&rt->rt_dst, &rt->rt_gateway);
L
Linus Torvalds 已提交
1467 1468 1469 1470 1471 1472
#endif
	}
}

static int ip_error(struct sk_buff *skb)
{
E
Eric Dumazet 已提交
1473
	struct rtable *rt = skb_rtable(skb);
1474
	struct inet_peer *peer;
L
Linus Torvalds 已提交
1475
	unsigned long now;
1476
	bool send;
L
Linus Torvalds 已提交
1477 1478
	int code;

1479
	switch (rt->dst.error) {
J
Joe Perches 已提交
1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493
	case EINVAL:
	default:
		goto out;
	case EHOSTUNREACH:
		code = ICMP_HOST_UNREACH;
		break;
	case ENETUNREACH:
		code = ICMP_NET_UNREACH;
		IP_INC_STATS_BH(dev_net(rt->dst.dev),
				IPSTATS_MIB_INNOROUTES);
		break;
	case EACCES:
		code = ICMP_PKT_FILTERED;
		break;
L
Linus Torvalds 已提交
1494 1495
	}

1496
	if (!rt->peer)
1497
		rt_bind_peer(rt, rt->rt_dst, 1);
1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510
	peer = rt->peer;

	send = true;
	if (peer) {
		now = jiffies;
		peer->rate_tokens += now - peer->rate_last;
		if (peer->rate_tokens > ip_rt_error_burst)
			peer->rate_tokens = ip_rt_error_burst;
		peer->rate_last = now;
		if (peer->rate_tokens >= ip_rt_error_cost)
			peer->rate_tokens -= ip_rt_error_cost;
		else
			send = false;
L
Linus Torvalds 已提交
1511
	}
1512 1513
	if (send)
		icmp_send(skb, ICMP_DEST_UNREACH, code, 0);
L
Linus Torvalds 已提交
1514 1515 1516

out:	kfree_skb(skb);
	return 0;
1517
}
L
Linus Torvalds 已提交
1518 1519 1520 1521 1522 1523

/*
 *	The last two values are not from the RFC but
 *	are needed for AMPRnet AX.25 paths.
 */

1524
static const unsigned short mtu_plateau[] =
L
Linus Torvalds 已提交
1525 1526
{32000, 17914, 8166, 4352, 2002, 1492, 576, 296, 216, 128 };

S
Stephen Hemminger 已提交
1527
static inline unsigned short guess_mtu(unsigned short old_mtu)
L
Linus Torvalds 已提交
1528 1529
{
	int i;
1530

L
Linus Torvalds 已提交
1531 1532 1533 1534 1535 1536
	for (i = 0; i < ARRAY_SIZE(mtu_plateau); i++)
		if (old_mtu > mtu_plateau[i])
			return mtu_plateau[i];
	return 68;
}

1537
unsigned short ip_rt_frag_needed(struct net *net, const struct iphdr *iph,
1538 1539
				 unsigned short new_mtu,
				 struct net_device *dev)
L
Linus Torvalds 已提交
1540 1541 1542
{
	unsigned short old_mtu = ntohs(iph->tot_len);
	unsigned short est_mtu = 0;
1543
	struct inet_peer *peer;
L
Linus Torvalds 已提交
1544

1545 1546 1547
	peer = inet_getpeer_v4(iph->daddr, 1);
	if (peer) {
		unsigned short mtu = new_mtu;
L
Linus Torvalds 已提交
1548

1549 1550 1551 1552 1553 1554 1555 1556 1557 1558
		if (new_mtu < 68 || new_mtu >= old_mtu) {
			/* BSD 4.2 derived systems incorrectly adjust
			 * tot_len by the IP header length, and report
			 * a zero MTU in the ICMP message.
			 */
			if (mtu == 0 &&
			    old_mtu >= 68 + (iph->ihl << 2))
				old_mtu -= iph->ihl << 2;
			mtu = guess_mtu(old_mtu);
		}
1559

1560 1561 1562
		if (mtu < ip_rt_min_pmtu)
			mtu = ip_rt_min_pmtu;
		if (!peer->pmtu_expires || mtu < peer->pmtu_learned) {
H
Hiroaki SHIMODA 已提交
1563 1564 1565 1566 1567 1568
			unsigned long pmtu_expires;

			pmtu_expires = jiffies + ip_rt_mtu_expires;
			if (!pmtu_expires)
				pmtu_expires = 1UL;

1569 1570
			est_mtu = mtu;
			peer->pmtu_learned = mtu;
H
Hiroaki SHIMODA 已提交
1571
			peer->pmtu_expires = pmtu_expires;
1572
		}
L
Linus Torvalds 已提交
1573

1574
		inet_putpeer(peer);
L
Linus Torvalds 已提交
1575

1576
		atomic_inc(&__rt_peer_genid);
L
Linus Torvalds 已提交
1577 1578 1579 1580
	}
	return est_mtu ? : new_mtu;
}

1581 1582
static void check_peer_pmtu(struct dst_entry *dst, struct inet_peer *peer)
{
E
Eric Dumazet 已提交
1583
	unsigned long expires = ACCESS_ONCE(peer->pmtu_expires);
1584

E
Eric Dumazet 已提交
1585 1586
	if (!expires)
		return;
H
Hiroaki SHIMODA 已提交
1587
	if (time_before(jiffies, expires)) {
1588 1589 1590 1591 1592 1593 1594 1595 1596 1597
		u32 orig_dst_mtu = dst_mtu(dst);
		if (peer->pmtu_learned < orig_dst_mtu) {
			if (!peer->pmtu_orig)
				peer->pmtu_orig = dst_metric_raw(dst, RTAX_MTU);
			dst_metric_set(dst, RTAX_MTU, peer->pmtu_learned);
		}
	} else if (cmpxchg(&peer->pmtu_expires, expires, 0) == expires)
		dst_metric_set(dst, RTAX_MTU, peer->pmtu_orig);
}

L
Linus Torvalds 已提交
1598 1599
static void ip_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
{
1600 1601 1602 1603 1604 1605
	struct rtable *rt = (struct rtable *) dst;
	struct inet_peer *peer;

	dst_confirm(dst);

	if (!rt->peer)
1606
		rt_bind_peer(rt, rt->rt_dst, 1);
1607 1608
	peer = rt->peer;
	if (peer) {
E
Eric Dumazet 已提交
1609 1610
		unsigned long pmtu_expires = ACCESS_ONCE(peer->pmtu_expires);

1611
		if (mtu < ip_rt_min_pmtu)
L
Linus Torvalds 已提交
1612
			mtu = ip_rt_min_pmtu;
E
Eric Dumazet 已提交
1613
		if (!pmtu_expires || mtu < peer->pmtu_learned) {
H
Hiroaki SHIMODA 已提交
1614 1615 1616 1617 1618

			pmtu_expires = jiffies + ip_rt_mtu_expires;
			if (!pmtu_expires)
				pmtu_expires = 1UL;

1619
			peer->pmtu_learned = mtu;
H
Hiroaki SHIMODA 已提交
1620
			peer->pmtu_expires = pmtu_expires;
1621 1622 1623

			atomic_inc(&__rt_peer_genid);
			rt->rt_peer_genid = rt_peer_genid();
L
Linus Torvalds 已提交
1624
		}
H
Hiroaki SHIMODA 已提交
1625
		check_peer_pmtu(dst, peer);
L
Linus Torvalds 已提交
1626 1627 1628
	}
}

1629 1630 1631 1632
static int check_peer_redir(struct dst_entry *dst, struct inet_peer *peer)
{
	struct rtable *rt = (struct rtable *) dst;
	__be32 orig_gw = rt->rt_gateway;
1633
	struct neighbour *n, *old_n;
1634 1635 1636 1637

	dst_confirm(&rt->dst);

	rt->rt_gateway = peer->redirect_learned.a4;
1638 1639 1640 1641 1642 1643 1644

	n = ipv4_neigh_lookup(&rt->dst, &rt->rt_gateway);
	if (IS_ERR(n))
		return PTR_ERR(n);
	old_n = xchg(&rt->dst._neighbour, n);
	if (old_n)
		neigh_release(old_n);
1645 1646 1647
	if (!n || !(n->nud_state & NUD_VALID)) {
		if (n)
			neigh_event_send(n, NULL);
1648 1649 1650 1651
		rt->rt_gateway = orig_gw;
		return -EAGAIN;
	} else {
		rt->rt_flags |= RTCF_REDIRECTED;
1652
		call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, n);
1653 1654 1655 1656
	}
	return 0;
}

L
Linus Torvalds 已提交
1657 1658
static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie)
{
1659 1660 1661
	struct rtable *rt = (struct rtable *) dst;

	if (rt_is_expired(rt))
T
Timo Teräs 已提交
1662
		return NULL;
1663
	if (rt->rt_peer_genid != rt_peer_genid()) {
1664 1665
		struct inet_peer *peer;

1666
		if (!rt->peer)
1667
			rt_bind_peer(rt, rt->rt_dst, 0);
1668

1669
		peer = rt->peer;
E
Eric Dumazet 已提交
1670
		if (peer) {
1671 1672
			check_peer_pmtu(dst, peer);

E
Eric Dumazet 已提交
1673 1674 1675 1676 1677
			if (peer->redirect_learned.a4 &&
			    peer->redirect_learned.a4 != rt->rt_gateway) {
				if (check_peer_redir(dst, peer))
					return NULL;
			}
1678 1679
		}

1680 1681
		rt->rt_peer_genid = rt_peer_genid();
	}
T
Timo Teräs 已提交
1682
	return dst;
L
Linus Torvalds 已提交
1683 1684 1685 1686 1687 1688 1689
}

static void ipv4_dst_destroy(struct dst_entry *dst)
{
	struct rtable *rt = (struct rtable *) dst;
	struct inet_peer *peer = rt->peer;

1690 1691 1692 1693
	if (rt->fi) {
		fib_info_put(rt->fi);
		rt->fi = NULL;
	}
L
Linus Torvalds 已提交
1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706
	if (peer) {
		rt->peer = NULL;
		inet_putpeer(peer);
	}
}


static void ipv4_link_failure(struct sk_buff *skb)
{
	struct rtable *rt;

	icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0);

E
Eric Dumazet 已提交
1707
	rt = skb_rtable(skb);
E
Eric Dumazet 已提交
1708 1709
	if (rt && rt->peer && peer_pmtu_cleaned(rt->peer))
		dst_metric_set(&rt->dst, RTAX_MTU, rt->peer->pmtu_orig);
L
Linus Torvalds 已提交
1710 1711 1712 1713
}

static int ip_rt_bug(struct sk_buff *skb)
{
1714 1715
	printk(KERN_DEBUG "ip_rt_bug: %pI4 -> %pI4, %s\n",
		&ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr,
L
Linus Torvalds 已提交
1716 1717
		skb->dev ? skb->dev->name : "?");
	kfree_skb(skb);
1718
	WARN_ON(1);
L
Linus Torvalds 已提交
1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730
	return 0;
}

/*
   We do not cache source address of outgoing interface,
   because it is used only by IP RR, TS and SRR options,
   so that it out of fast path.

   BTW remember: "addr" is allowed to be not aligned
   in IP options!
 */

1731
void ip_rt_get_source(u8 *addr, struct sk_buff *skb, struct rtable *rt)
L
Linus Torvalds 已提交
1732
{
A
Al Viro 已提交
1733
	__be32 src;
L
Linus Torvalds 已提交
1734

1735
	if (rt_is_output_route(rt))
1736
		src = ip_hdr(skb)->saddr;
E
Eric Dumazet 已提交
1737
	else {
1738 1739 1740 1741 1742 1743 1744 1745 1746
		struct fib_result res;
		struct flowi4 fl4;
		struct iphdr *iph;

		iph = ip_hdr(skb);

		memset(&fl4, 0, sizeof(fl4));
		fl4.daddr = iph->daddr;
		fl4.saddr = iph->saddr;
1747
		fl4.flowi4_tos = RT_TOS(iph->tos);
1748 1749 1750
		fl4.flowi4_oif = rt->dst.dev->ifindex;
		fl4.flowi4_iif = skb->dev->ifindex;
		fl4.flowi4_mark = skb->mark;
1751

E
Eric Dumazet 已提交
1752
		rcu_read_lock();
1753
		if (fib_lookup(dev_net(rt->dst.dev), &fl4, &res) == 0)
1754
			src = FIB_RES_PREFSRC(dev_net(rt->dst.dev), res);
E
Eric Dumazet 已提交
1755 1756
		else
			src = inet_select_addr(rt->dst.dev, rt->rt_gateway,
L
Linus Torvalds 已提交
1757
					RT_SCOPE_UNIVERSE);
E
Eric Dumazet 已提交
1758 1759
		rcu_read_unlock();
	}
L
Linus Torvalds 已提交
1760 1761 1762
	memcpy(addr, &src, 4);
}

1763
#ifdef CONFIG_IP_ROUTE_CLASSID
L
Linus Torvalds 已提交
1764 1765
static void set_class_tag(struct rtable *rt, u32 tag)
{
1766 1767 1768 1769
	if (!(rt->dst.tclassid & 0xFFFF))
		rt->dst.tclassid |= tag & 0xFFFF;
	if (!(rt->dst.tclassid & 0xFFFF0000))
		rt->dst.tclassid |= tag & 0xFFFF0000;
L
Linus Torvalds 已提交
1770 1771 1772
}
#endif

1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785
static unsigned int ipv4_default_advmss(const struct dst_entry *dst)
{
	unsigned int advmss = dst_metric_raw(dst, RTAX_ADVMSS);

	if (advmss == 0) {
		advmss = max_t(unsigned int, dst->dev->mtu - 40,
			       ip_rt_min_advmss);
		if (advmss > 65535 - 40)
			advmss = 65535 - 40;
	}
	return advmss;
}

1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802
static unsigned int ipv4_default_mtu(const struct dst_entry *dst)
{
	unsigned int mtu = dst->dev->mtu;

	if (unlikely(dst_metric_locked(dst, RTAX_MTU))) {
		const struct rtable *rt = (const struct rtable *) dst;

		if (rt->rt_gateway != rt->rt_dst && mtu > 576)
			mtu = 576;
	}

	if (mtu > IP_MAX_MTU)
		mtu = IP_MAX_MTU;

	return mtu;
}

1803
static void rt_init_metrics(struct rtable *rt, const struct flowi4 *fl4,
1804
			    struct fib_info *fi)
D
David S. Miller 已提交
1805
{
1806 1807
	struct inet_peer *peer;
	int create = 0;
D
David S. Miller 已提交
1808

1809 1810 1811
	/* If a peer entry exists for this destination, we must hook
	 * it up in order to get at cached metrics.
	 */
1812
	if (fl4 && (fl4->flowi4_flags & FLOWI_FLAG_PRECOW_METRICS))
1813 1814
		create = 1;

1815
	rt->peer = peer = inet_getpeer_v4(rt->rt_dst, create);
1816
	if (peer) {
1817
		rt->rt_peer_genid = rt_peer_genid();
D
David S. Miller 已提交
1818 1819 1820 1821
		if (inet_metrics_new(peer))
			memcpy(peer->metrics, fi->fib_metrics,
			       sizeof(u32) * RTAX_MAX);
		dst_init_metrics(&rt->dst, peer->metrics, false);
1822

E
Eric Dumazet 已提交
1823
		check_peer_pmtu(&rt->dst, peer);
1824 1825 1826 1827 1828
		if (peer->redirect_learned.a4 &&
		    peer->redirect_learned.a4 != rt->rt_gateway) {
			rt->rt_gateway = peer->redirect_learned.a4;
			rt->rt_flags |= RTCF_REDIRECTED;
		}
1829 1830 1831 1832 1833 1834
	} else {
		if (fi->fib_metrics != (u32 *) dst_default_metrics) {
			rt->fi = fi;
			atomic_inc(&fi->fib_clntref);
		}
		dst_init_metrics(&rt->dst, fi->fib_metrics, true);
D
David S. Miller 已提交
1835 1836 1837
	}
}

1838
static void rt_set_nexthop(struct rtable *rt, const struct flowi4 *fl4,
1839
			   const struct fib_result *res,
1840
			   struct fib_info *fi, u16 type, u32 itag)
L
Linus Torvalds 已提交
1841
{
1842
	struct dst_entry *dst = &rt->dst;
L
Linus Torvalds 已提交
1843 1844 1845 1846 1847

	if (fi) {
		if (FIB_RES_GW(*res) &&
		    FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK)
			rt->rt_gateway = FIB_RES_GW(*res);
1848
		rt_init_metrics(rt, fl4, fi);
1849
#ifdef CONFIG_IP_ROUTE_CLASSID
1850
		dst->tclassid = FIB_RES_NH(*res).nh_tclassid;
L
Linus Torvalds 已提交
1851
#endif
1852
	}
1853 1854 1855

	if (dst_mtu(dst) > IP_MAX_MTU)
		dst_metric_set(dst, RTAX_MTU, IP_MAX_MTU);
1856
	if (dst_metric_raw(dst, RTAX_ADVMSS) > 65535 - 40)
1857
		dst_metric_set(dst, RTAX_ADVMSS, 65535 - 40);
L
Linus Torvalds 已提交
1858

1859
#ifdef CONFIG_IP_ROUTE_CLASSID
L
Linus Torvalds 已提交
1860 1861 1862 1863 1864 1865 1866
#ifdef CONFIG_IP_MULTIPLE_TABLES
	set_class_tag(rt, fib_rules_tclass(res));
#endif
	set_class_tag(rt, itag);
#endif
}

1867 1868
static struct rtable *rt_dst_alloc(struct net_device *dev,
				   bool nopolicy, bool noxfrm)
1869
{
1870 1871 1872 1873
	return dst_alloc(&ipv4_dst_ops, dev, 1, -1,
			 DST_HOST |
			 (nopolicy ? DST_NOPOLICY : 0) |
			 (noxfrm ? DST_NOXFRM : 0));
1874 1875
}

1876
/* called in rcu_read_lock() section */
A
Al Viro 已提交
1877
static int ip_route_input_mc(struct sk_buff *skb, __be32 daddr, __be32 saddr,
L
Linus Torvalds 已提交
1878 1879
				u8 tos, struct net_device *dev, int our)
{
1880
	unsigned int hash;
L
Linus Torvalds 已提交
1881
	struct rtable *rth;
A
Al Viro 已提交
1882
	__be32 spec_dst;
1883
	struct in_device *in_dev = __in_dev_get_rcu(dev);
L
Linus Torvalds 已提交
1884
	u32 itag = 0;
1885
	int err;
L
Linus Torvalds 已提交
1886 1887 1888 1889 1890 1891

	/* Primary sanity checks. */

	if (in_dev == NULL)
		return -EINVAL;

1892
	if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) ||
1893
	    ipv4_is_loopback(saddr) || skb->protocol != htons(ETH_P_IP))
L
Linus Torvalds 已提交
1894 1895
		goto e_inval;

1896 1897
	if (ipv4_is_zeronet(saddr)) {
		if (!ipv4_is_local_multicast(daddr))
L
Linus Torvalds 已提交
1898 1899
			goto e_inval;
		spec_dst = inet_select_addr(dev, 0, RT_SCOPE_LINK);
1900
	} else {
1901 1902
		err = fib_validate_source(skb, saddr, 0, tos, 0, dev, &spec_dst,
					  &itag);
1903 1904 1905
		if (err < 0)
			goto e_err;
	}
1906 1907
	rth = rt_dst_alloc(init_net.loopback_dev,
			   IN_DEV_CONF_GET(in_dev, NOPOLICY), false);
L
Linus Torvalds 已提交
1908 1909 1910
	if (!rth)
		goto e_nobufs;

1911 1912 1913
#ifdef CONFIG_IP_ROUTE_CLASSID
	rth->dst.tclassid = itag;
#endif
1914
	rth->dst.output = ip_rt_bug;
L
Linus Torvalds 已提交
1915

1916 1917
	rth->rt_key_dst	= daddr;
	rth->rt_key_src	= saddr;
1918 1919 1920
	rth->rt_genid	= rt_genid(dev_net(dev));
	rth->rt_flags	= RTCF_MULTICAST;
	rth->rt_type	= RTN_MULTICAST;
1921
	rth->rt_key_tos	= tos;
1922
	rth->rt_dst	= daddr;
L
Linus Torvalds 已提交
1923
	rth->rt_src	= saddr;
1924
	rth->rt_route_iif = dev->ifindex;
1925 1926
	rth->rt_iif	= dev->ifindex;
	rth->rt_oif	= 0;
1927
	rth->rt_mark    = skb->mark;
L
Linus Torvalds 已提交
1928 1929
	rth->rt_gateway	= daddr;
	rth->rt_spec_dst= spec_dst;
1930 1931 1932
	rth->rt_peer_genid = 0;
	rth->peer = NULL;
	rth->fi = NULL;
L
Linus Torvalds 已提交
1933
	if (our) {
1934
		rth->dst.input= ip_local_deliver;
L
Linus Torvalds 已提交
1935 1936 1937 1938
		rth->rt_flags |= RTCF_LOCAL;
	}

#ifdef CONFIG_IP_MROUTE
1939
	if (!ipv4_is_local_multicast(daddr) && IN_DEV_MFORWARD(in_dev))
1940
		rth->dst.input = ip_mr_input;
L
Linus Torvalds 已提交
1941 1942 1943
#endif
	RT_CACHE_STAT_INC(in_slow_mc);

1944
	hash = rt_hash(daddr, saddr, dev->ifindex, rt_genid(dev_net(dev)));
1945
	rth = rt_intern_hash(hash, rth, skb, dev->ifindex);
E
Eric Dumazet 已提交
1946
	return IS_ERR(rth) ? PTR_ERR(rth) : 0;
L
Linus Torvalds 已提交
1947 1948 1949 1950

e_nobufs:
	return -ENOBUFS;
e_inval:
1951
	return -EINVAL;
1952 1953
e_err:
	return err;
L
Linus Torvalds 已提交
1954 1955 1956 1957 1958 1959
}


static void ip_handle_martian_source(struct net_device *dev,
				     struct in_device *in_dev,
				     struct sk_buff *skb,
A
Al Viro 已提交
1960 1961
				     __be32 daddr,
				     __be32 saddr)
L
Linus Torvalds 已提交
1962 1963 1964 1965 1966 1967 1968 1969
{
	RT_CACHE_STAT_INC(in_martian_src);
#ifdef CONFIG_IP_ROUTE_VERBOSE
	if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) {
		/*
		 *	RFC1812 recommendation, if source is martian,
		 *	the only hint is MAC header.
		 */
1970 1971
		printk(KERN_WARNING "martian source %pI4 from %pI4, on dev %s\n",
			&daddr, &saddr, dev->name);
1972
		if (dev->hard_header_len && skb_mac_header_was_set(skb)) {
L
Linus Torvalds 已提交
1973
			int i;
1974
			const unsigned char *p = skb_mac_header(skb);
L
Linus Torvalds 已提交
1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
			printk(KERN_WARNING "ll header: ");
			for (i = 0; i < dev->hard_header_len; i++, p++) {
				printk("%02x", *p);
				if (i < (dev->hard_header_len - 1))
					printk(":");
			}
			printk("\n");
		}
	}
#endif
}

1987
/* called in rcu_read_lock() section */
S
Stephen Hemminger 已提交
1988
static int __mkroute_input(struct sk_buff *skb,
1989
			   const struct fib_result *res,
S
Stephen Hemminger 已提交
1990 1991 1992
			   struct in_device *in_dev,
			   __be32 daddr, __be32 saddr, u32 tos,
			   struct rtable **result)
L
Linus Torvalds 已提交
1993 1994 1995 1996
{
	struct rtable *rth;
	int err;
	struct in_device *out_dev;
1997
	unsigned int flags = 0;
1998 1999
	__be32 spec_dst;
	u32 itag;
L
Linus Torvalds 已提交
2000 2001

	/* get a working reference to the output device */
2002
	out_dev = __in_dev_get_rcu(FIB_RES_DEV(*res));
L
Linus Torvalds 已提交
2003 2004 2005 2006 2007 2008 2009 2010
	if (out_dev == NULL) {
		if (net_ratelimit())
			printk(KERN_CRIT "Bug in ip_route_input" \
			       "_slow(). Please, report\n");
		return -EINVAL;
	}


2011 2012
	err = fib_validate_source(skb, saddr, daddr, tos, FIB_RES_OIF(*res),
				  in_dev->dev, &spec_dst, &itag);
L
Linus Torvalds 已提交
2013
	if (err < 0) {
2014
		ip_handle_martian_source(in_dev->dev, in_dev, skb, daddr,
L
Linus Torvalds 已提交
2015
					 saddr);
2016

L
Linus Torvalds 已提交
2017 2018 2019 2020 2021 2022
		goto cleanup;
	}

	if (err)
		flags |= RTCF_DIRECTSRC;

2023
	if (out_dev == in_dev && err &&
L
Linus Torvalds 已提交
2024 2025 2026 2027 2028 2029 2030
	    (IN_DEV_SHARED_MEDIA(out_dev) ||
	     inet_addr_onlink(out_dev, saddr, FIB_RES_GW(*res))))
		flags |= RTCF_DOREDIRECT;

	if (skb->protocol != htons(ETH_P_IP)) {
		/* Not IP (i.e. ARP). Do not create route, if it is
		 * invalid for proxy arp. DNAT routes are always valid.
2031 2032 2033 2034
		 *
		 * Proxy arp feature have been extended to allow, ARP
		 * replies back to the same interface, to support
		 * Private VLAN switch technologies. See arp.c.
L
Linus Torvalds 已提交
2035
		 */
2036 2037
		if (out_dev == in_dev &&
		    IN_DEV_PROXY_ARP_PVLAN(in_dev) == 0) {
L
Linus Torvalds 已提交
2038 2039 2040 2041 2042
			err = -EINVAL;
			goto cleanup;
		}
	}

2043 2044
	rth = rt_dst_alloc(out_dev->dev,
			   IN_DEV_CONF_GET(in_dev, NOPOLICY),
2045
			   IN_DEV_CONF_GET(out_dev, NOXFRM));
L
Linus Torvalds 已提交
2046 2047 2048 2049 2050
	if (!rth) {
		err = -ENOBUFS;
		goto cleanup;
	}

2051 2052
	rth->rt_key_dst	= daddr;
	rth->rt_key_src	= saddr;
2053 2054 2055
	rth->rt_genid = rt_genid(dev_net(rth->dst.dev));
	rth->rt_flags = flags;
	rth->rt_type = res->type;
2056
	rth->rt_key_tos	= tos;
2057
	rth->rt_dst	= daddr;
L
Linus Torvalds 已提交
2058
	rth->rt_src	= saddr;
2059
	rth->rt_route_iif = in_dev->dev->ifindex;
2060 2061
	rth->rt_iif 	= in_dev->dev->ifindex;
	rth->rt_oif 	= 0;
2062 2063
	rth->rt_mark    = skb->mark;
	rth->rt_gateway	= daddr;
L
Linus Torvalds 已提交
2064
	rth->rt_spec_dst= spec_dst;
2065 2066 2067
	rth->rt_peer_genid = 0;
	rth->peer = NULL;
	rth->fi = NULL;
L
Linus Torvalds 已提交
2068

2069 2070
	rth->dst.input = ip_forward;
	rth->dst.output = ip_output;
L
Linus Torvalds 已提交
2071

2072
	rt_set_nexthop(rth, NULL, res, res->fi, res->type, itag);
L
Linus Torvalds 已提交
2073 2074 2075 2076 2077

	*result = rth;
	err = 0;
 cleanup:
	return err;
2078
}
L
Linus Torvalds 已提交
2079

S
Stephen Hemminger 已提交
2080 2081
static int ip_mkroute_input(struct sk_buff *skb,
			    struct fib_result *res,
2082
			    const struct flowi4 *fl4,
S
Stephen Hemminger 已提交
2083 2084
			    struct in_device *in_dev,
			    __be32 daddr, __be32 saddr, u32 tos)
L
Linus Torvalds 已提交
2085
{
C
Chuck Short 已提交
2086
	struct rtable* rth = NULL;
L
Linus Torvalds 已提交
2087 2088 2089 2090
	int err;
	unsigned hash;

#ifdef CONFIG_IP_ROUTE_MULTIPATH
2091
	if (res->fi && res->fi->fib_nhs > 1)
2092
		fib_select_multipath(res);
L
Linus Torvalds 已提交
2093 2094 2095 2096 2097 2098 2099 2100
#endif

	/* create a routing cache entry */
	err = __mkroute_input(skb, res, in_dev, daddr, saddr, tos, &rth);
	if (err)
		return err;

	/* put it into the cache */
2101
	hash = rt_hash(daddr, saddr, fl4->flowi4_iif,
2102
		       rt_genid(dev_net(rth->dst.dev)));
2103
	rth = rt_intern_hash(hash, rth, skb, fl4->flowi4_iif);
2104 2105 2106
	if (IS_ERR(rth))
		return PTR_ERR(rth);
	return 0;
L
Linus Torvalds 已提交
2107 2108 2109 2110 2111 2112 2113 2114 2115 2116
}

/*
 *	NOTE. We drop all the packets that has local source
 *	addresses, because every properly looped back packet
 *	must have correct destination already attached by output routine.
 *
 *	Such approach solves two big problems:
 *	1. Not simplex devices are handled properly.
 *	2. IP spoofing attempts are filtered with 100% of guarantee.
E
Eric Dumazet 已提交
2117
 *	called with rcu_read_lock()
L
Linus Torvalds 已提交
2118 2119
 */

A
Al Viro 已提交
2120
static int ip_route_input_slow(struct sk_buff *skb, __be32 daddr, __be32 saddr,
L
Linus Torvalds 已提交
2121 2122 2123
			       u8 tos, struct net_device *dev)
{
	struct fib_result res;
2124
	struct in_device *in_dev = __in_dev_get_rcu(dev);
2125
	struct flowi4	fl4;
L
Linus Torvalds 已提交
2126 2127 2128 2129
	unsigned	flags = 0;
	u32		itag = 0;
	struct rtable * rth;
	unsigned	hash;
A
Al Viro 已提交
2130
	__be32		spec_dst;
L
Linus Torvalds 已提交
2131
	int		err = -EINVAL;
2132
	struct net    * net = dev_net(dev);
L
Linus Torvalds 已提交
2133 2134 2135 2136 2137 2138 2139 2140 2141 2142

	/* IP on this device is disabled. */

	if (!in_dev)
		goto out;

	/* Check for the most weird martians, which can be not detected
	   by fib_lookup.
	 */

2143
	if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) ||
2144
	    ipv4_is_loopback(saddr))
L
Linus Torvalds 已提交
2145 2146
		goto martian_source;

2147
	if (ipv4_is_lbcast(daddr) || (saddr == 0 && daddr == 0))
L
Linus Torvalds 已提交
2148 2149 2150 2151 2152
		goto brd_input;

	/* Accept zero addresses only to limited broadcast;
	 * I even do not know to fix it or not. Waiting for complains :-)
	 */
2153
	if (ipv4_is_zeronet(saddr))
L
Linus Torvalds 已提交
2154 2155
		goto martian_source;

2156
	if (ipv4_is_zeronet(daddr) || ipv4_is_loopback(daddr))
L
Linus Torvalds 已提交
2157 2158 2159 2160 2161
		goto martian_destination;

	/*
	 *	Now we are ready to route packet.
	 */
2162 2163 2164 2165 2166 2167 2168 2169
	fl4.flowi4_oif = 0;
	fl4.flowi4_iif = dev->ifindex;
	fl4.flowi4_mark = skb->mark;
	fl4.flowi4_tos = tos;
	fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
	fl4.daddr = daddr;
	fl4.saddr = saddr;
	err = fib_lookup(net, &fl4, &res);
E
Eric Dumazet 已提交
2170
	if (err != 0) {
L
Linus Torvalds 已提交
2171
		if (!IN_DEV_FORWARD(in_dev))
2172
			goto e_hostunreach;
L
Linus Torvalds 已提交
2173 2174 2175 2176 2177 2178 2179 2180 2181
		goto no_route;
	}

	RT_CACHE_STAT_INC(in_slow_tot);

	if (res.type == RTN_BROADCAST)
		goto brd_input;

	if (res.type == RTN_LOCAL) {
2182
		err = fib_validate_source(skb, saddr, daddr, tos,
E
Eric Dumazet 已提交
2183
					  net->loopback_dev->ifindex,
2184
					  dev, &spec_dst, &itag);
2185 2186 2187
		if (err < 0)
			goto martian_source_keep_err;
		if (err)
L
Linus Torvalds 已提交
2188 2189 2190 2191 2192 2193
			flags |= RTCF_DIRECTSRC;
		spec_dst = daddr;
		goto local_input;
	}

	if (!IN_DEV_FORWARD(in_dev))
2194
		goto e_hostunreach;
L
Linus Torvalds 已提交
2195 2196 2197
	if (res.type != RTN_UNICAST)
		goto martian_destination;

2198
	err = ip_mkroute_input(skb, &res, &fl4, in_dev, daddr, saddr, tos);
L
Linus Torvalds 已提交
2199 2200 2201 2202 2203 2204
out:	return err;

brd_input:
	if (skb->protocol != htons(ETH_P_IP))
		goto e_inval;

2205
	if (ipv4_is_zeronet(saddr))
L
Linus Torvalds 已提交
2206 2207
		spec_dst = inet_select_addr(dev, 0, RT_SCOPE_LINK);
	else {
2208 2209
		err = fib_validate_source(skb, saddr, 0, tos, 0, dev, &spec_dst,
					  &itag);
L
Linus Torvalds 已提交
2210
		if (err < 0)
2211
			goto martian_source_keep_err;
L
Linus Torvalds 已提交
2212 2213 2214 2215 2216 2217 2218 2219
		if (err)
			flags |= RTCF_DIRECTSRC;
	}
	flags |= RTCF_BROADCAST;
	res.type = RTN_BROADCAST;
	RT_CACHE_STAT_INC(in_brd);

local_input:
2220 2221
	rth = rt_dst_alloc(net->loopback_dev,
			   IN_DEV_CONF_GET(in_dev, NOPOLICY), false);
L
Linus Torvalds 已提交
2222 2223 2224
	if (!rth)
		goto e_nobufs;

2225
	rth->dst.input= ip_local_deliver;
2226
	rth->dst.output= ip_rt_bug;
2227 2228 2229
#ifdef CONFIG_IP_ROUTE_CLASSID
	rth->dst.tclassid = itag;
#endif
L
Linus Torvalds 已提交
2230

2231 2232
	rth->rt_key_dst	= daddr;
	rth->rt_key_src	= saddr;
2233 2234 2235
	rth->rt_genid = rt_genid(net);
	rth->rt_flags 	= flags|RTCF_LOCAL;
	rth->rt_type	= res.type;
2236
	rth->rt_key_tos	= tos;
2237
	rth->rt_dst	= daddr;
L
Linus Torvalds 已提交
2238
	rth->rt_src	= saddr;
2239
#ifdef CONFIG_IP_ROUTE_CLASSID
2240
	rth->dst.tclassid = itag;
L
Linus Torvalds 已提交
2241
#endif
2242
	rth->rt_route_iif = dev->ifindex;
2243
	rth->rt_iif	= dev->ifindex;
2244 2245
	rth->rt_oif	= 0;
	rth->rt_mark    = skb->mark;
L
Linus Torvalds 已提交
2246 2247
	rth->rt_gateway	= daddr;
	rth->rt_spec_dst= spec_dst;
2248 2249 2250
	rth->rt_peer_genid = 0;
	rth->peer = NULL;
	rth->fi = NULL;
L
Linus Torvalds 已提交
2251
	if (res.type == RTN_UNREACHABLE) {
2252 2253
		rth->dst.input= ip_error;
		rth->dst.error= -err;
L
Linus Torvalds 已提交
2254 2255
		rth->rt_flags 	&= ~RTCF_LOCAL;
	}
2256 2257
	hash = rt_hash(daddr, saddr, fl4.flowi4_iif, rt_genid(net));
	rth = rt_intern_hash(hash, rth, skb, fl4.flowi4_iif);
2258 2259 2260
	err = 0;
	if (IS_ERR(rth))
		err = PTR_ERR(rth);
E
Eric Dumazet 已提交
2261
	goto out;
L
Linus Torvalds 已提交
2262 2263 2264 2265 2266

no_route:
	RT_CACHE_STAT_INC(in_no_route);
	spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE);
	res.type = RTN_UNREACHABLE;
2267 2268
	if (err == -ESRCH)
		err = -ENETUNREACH;
L
Linus Torvalds 已提交
2269 2270 2271 2272 2273 2274 2275 2276 2277
	goto local_input;

	/*
	 *	Do not cache martian addresses: they should be logged (RFC1812)
	 */
martian_destination:
	RT_CACHE_STAT_INC(in_martian_dst);
#ifdef CONFIG_IP_ROUTE_VERBOSE
	if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit())
2278 2279
		printk(KERN_WARNING "martian destination %pI4 from %pI4, dev %s\n",
			&daddr, &saddr, dev->name);
L
Linus Torvalds 已提交
2280
#endif
2281 2282

e_hostunreach:
2283
	err = -EHOSTUNREACH;
E
Eric Dumazet 已提交
2284
	goto out;
2285

L
Linus Torvalds 已提交
2286 2287
e_inval:
	err = -EINVAL;
E
Eric Dumazet 已提交
2288
	goto out;
L
Linus Torvalds 已提交
2289 2290 2291

e_nobufs:
	err = -ENOBUFS;
E
Eric Dumazet 已提交
2292
	goto out;
L
Linus Torvalds 已提交
2293 2294

martian_source:
2295 2296
	err = -EINVAL;
martian_source_keep_err:
L
Linus Torvalds 已提交
2297
	ip_handle_martian_source(dev, in_dev, skb, daddr, saddr);
E
Eric Dumazet 已提交
2298
	goto out;
L
Linus Torvalds 已提交
2299 2300
}

2301 2302
int ip_route_input_common(struct sk_buff *skb, __be32 daddr, __be32 saddr,
			   u8 tos, struct net_device *dev, bool noref)
L
Linus Torvalds 已提交
2303 2304 2305 2306
{
	struct rtable * rth;
	unsigned	hash;
	int iif = dev->ifindex;
2307
	struct net *net;
2308
	int res;
L
Linus Torvalds 已提交
2309

2310
	net = dev_net(dev);
2311

2312 2313
	rcu_read_lock();

2314 2315 2316
	if (!rt_caching(net))
		goto skip_cache;

L
Linus Torvalds 已提交
2317
	tos &= IPTOS_RT_MASK;
2318
	hash = rt_hash(daddr, saddr, iif, rt_genid(net));
L
Linus Torvalds 已提交
2319 2320

	for (rth = rcu_dereference(rt_hash_table[hash].chain); rth;
2321
	     rth = rcu_dereference(rth->dst.rt_next)) {
2322 2323 2324
		if ((((__force u32)rth->rt_key_dst ^ (__force u32)daddr) |
		     ((__force u32)rth->rt_key_src ^ (__force u32)saddr) |
		     (rth->rt_iif ^ iif) |
2325
		     (rth->rt_key_tos ^ tos)) == 0 &&
2326
		    rt_is_input_route(rth) &&
2327
		    rth->rt_mark == skb->mark &&
2328
		    net_eq(dev_net(rth->dst.dev), net) &&
2329
		    !rt_is_expired(rth)) {
2330
			if (noref) {
2331 2332
				dst_use_noref(&rth->dst, jiffies);
				skb_dst_set_noref(skb, &rth->dst);
2333
			} else {
2334 2335
				dst_use(&rth->dst, jiffies);
				skb_dst_set(skb, &rth->dst);
2336
			}
L
Linus Torvalds 已提交
2337 2338 2339 2340 2341 2342 2343
			RT_CACHE_STAT_INC(in_hit);
			rcu_read_unlock();
			return 0;
		}
		RT_CACHE_STAT_INC(in_hlist_search);
	}

2344
skip_cache:
L
Linus Torvalds 已提交
2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355
	/* Multicast recognition logic is moved from route cache to here.
	   The problem was that too many Ethernet cards have broken/missing
	   hardware multicast filters :-( As result the host on multicasting
	   network acquires a lot of useless route cache entries, sort of
	   SDR messages from all the world. Now we try to get rid of them.
	   Really, provided software IP multicast filter is organized
	   reasonably (at least, hashed), it does not result in a slowdown
	   comparing with route cache reject entries.
	   Note, that multicast routers are not affected, because
	   route cache entry is created eventually.
	 */
2356
	if (ipv4_is_multicast(daddr)) {
2357
		struct in_device *in_dev = __in_dev_get_rcu(dev);
L
Linus Torvalds 已提交
2358

2359
		if (in_dev) {
2360 2361
			int our = ip_check_mc_rcu(in_dev, daddr, saddr,
						  ip_hdr(skb)->protocol);
L
Linus Torvalds 已提交
2362 2363
			if (our
#ifdef CONFIG_IP_MROUTE
2364 2365 2366
				||
			    (!ipv4_is_local_multicast(daddr) &&
			     IN_DEV_MFORWARD(in_dev))
L
Linus Torvalds 已提交
2367
#endif
2368
			   ) {
2369 2370
				int res = ip_route_input_mc(skb, daddr, saddr,
							    tos, dev, our);
L
Linus Torvalds 已提交
2371
				rcu_read_unlock();
2372
				return res;
L
Linus Torvalds 已提交
2373 2374 2375 2376 2377
			}
		}
		rcu_read_unlock();
		return -EINVAL;
	}
2378 2379 2380
	res = ip_route_input_slow(skb, daddr, saddr, tos, dev);
	rcu_read_unlock();
	return res;
L
Linus Torvalds 已提交
2381
}
2382
EXPORT_SYMBOL(ip_route_input_common);
L
Linus Torvalds 已提交
2383

E
Eric Dumazet 已提交
2384
/* called with rcu_read_lock() */
2385
static struct rtable *__mkroute_output(const struct fib_result *res,
2386
				       const struct flowi4 *fl4,
2387 2388
				       __be32 orig_daddr, __be32 orig_saddr,
				       int orig_oif, struct net_device *dev_out,
2389
				       unsigned int flags)
L
Linus Torvalds 已提交
2390
{
2391
	struct fib_info *fi = res->fi;
2392
	u32 tos = RT_FL_TOS(fl4);
2393
	struct in_device *in_dev;
2394
	u16 type = res->type;
2395
	struct rtable *rth;
L
Linus Torvalds 已提交
2396

2397
	if (ipv4_is_loopback(fl4->saddr) && !(dev_out->flags & IFF_LOOPBACK))
2398
		return ERR_PTR(-EINVAL);
L
Linus Torvalds 已提交
2399

2400
	if (ipv4_is_lbcast(fl4->daddr))
2401
		type = RTN_BROADCAST;
2402
	else if (ipv4_is_multicast(fl4->daddr))
2403
		type = RTN_MULTICAST;
2404
	else if (ipv4_is_zeronet(fl4->daddr))
2405
		return ERR_PTR(-EINVAL);
L
Linus Torvalds 已提交
2406 2407 2408 2409

	if (dev_out->flags & IFF_LOOPBACK)
		flags |= RTCF_LOCAL;

E
Eric Dumazet 已提交
2410
	in_dev = __in_dev_get_rcu(dev_out);
E
Eric Dumazet 已提交
2411
	if (!in_dev)
2412
		return ERR_PTR(-EINVAL);
E
Eric Dumazet 已提交
2413

2414
	if (type == RTN_BROADCAST) {
L
Linus Torvalds 已提交
2415
		flags |= RTCF_BROADCAST | RTCF_LOCAL;
2416 2417
		fi = NULL;
	} else if (type == RTN_MULTICAST) {
E
Eric Dumazet 已提交
2418
		flags |= RTCF_MULTICAST | RTCF_LOCAL;
2419 2420
		if (!ip_check_mc_rcu(in_dev, fl4->daddr, fl4->saddr,
				     fl4->flowi4_proto))
L
Linus Torvalds 已提交
2421 2422
			flags &= ~RTCF_LOCAL;
		/* If multicast route do not exist use
E
Eric Dumazet 已提交
2423 2424
		 * default one, but do not gateway in this case.
		 * Yes, it is hack.
L
Linus Torvalds 已提交
2425
		 */
2426 2427
		if (fi && res->prefixlen < 4)
			fi = NULL;
L
Linus Torvalds 已提交
2428 2429
	}

2430 2431
	rth = rt_dst_alloc(dev_out,
			   IN_DEV_CONF_GET(in_dev, NOPOLICY),
2432
			   IN_DEV_CONF_GET(in_dev, NOXFRM));
2433
	if (!rth)
2434
		return ERR_PTR(-ENOBUFS);
2435

2436 2437
	rth->dst.output = ip_output;

2438 2439
	rth->rt_key_dst	= orig_daddr;
	rth->rt_key_src	= orig_saddr;
2440 2441 2442
	rth->rt_genid = rt_genid(dev_net(dev_out));
	rth->rt_flags	= flags;
	rth->rt_type	= type;
2443
	rth->rt_key_tos	= tos;
2444 2445
	rth->rt_dst	= fl4->daddr;
	rth->rt_src	= fl4->saddr;
2446
	rth->rt_route_iif = 0;
2447 2448 2449
	rth->rt_iif	= orig_oif ? : dev_out->ifindex;
	rth->rt_oif	= orig_oif;
	rth->rt_mark    = fl4->flowi4_mark;
2450 2451
	rth->rt_gateway = fl4->daddr;
	rth->rt_spec_dst= fl4->saddr;
2452 2453 2454
	rth->rt_peer_genid = 0;
	rth->peer = NULL;
	rth->fi = NULL;
L
Linus Torvalds 已提交
2455 2456 2457 2458

	RT_CACHE_STAT_INC(out_slow_tot);

	if (flags & RTCF_LOCAL) {
2459
		rth->dst.input = ip_local_deliver;
2460
		rth->rt_spec_dst = fl4->daddr;
L
Linus Torvalds 已提交
2461 2462
	}
	if (flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
2463
		rth->rt_spec_dst = fl4->saddr;
2464
		if (flags & RTCF_LOCAL &&
L
Linus Torvalds 已提交
2465
		    !(dev_out->flags & IFF_LOOPBACK)) {
2466
			rth->dst.output = ip_mc_output;
L
Linus Torvalds 已提交
2467 2468 2469
			RT_CACHE_STAT_INC(out_slow_mc);
		}
#ifdef CONFIG_IP_MROUTE
2470
		if (type == RTN_MULTICAST) {
L
Linus Torvalds 已提交
2471
			if (IN_DEV_MFORWARD(in_dev) &&
2472
			    !ipv4_is_local_multicast(fl4->daddr)) {
2473 2474
				rth->dst.input = ip_mr_input;
				rth->dst.output = ip_mc_output;
L
Linus Torvalds 已提交
2475 2476 2477 2478 2479
			}
		}
#endif
	}

2480
	rt_set_nexthop(rth, fl4, res, fi, type, 0);
L
Linus Torvalds 已提交
2481

2482
	return rth;
L
Linus Torvalds 已提交
2483 2484 2485 2486
}

/*
 * Major route resolver routine.
2487
 * called with rcu_read_lock();
L
Linus Torvalds 已提交
2488 2489
 */

2490
static struct rtable *ip_route_output_slow(struct net *net, struct flowi4 *fl4)
L
Linus Torvalds 已提交
2491 2492
{
	struct net_device *dev_out = NULL;
2493 2494 2495
	u32 tos	= RT_FL_TOS(fl4);
	unsigned int flags = 0;
	struct fib_result res;
2496
	struct rtable *rth;
2497 2498 2499
	__be32 orig_daddr;
	__be32 orig_saddr;
	int orig_oif;
L
Linus Torvalds 已提交
2500 2501 2502 2503 2504 2505

	res.fi		= NULL;
#ifdef CONFIG_IP_MULTIPLE_TABLES
	res.r		= NULL;
#endif

2506 2507 2508 2509 2510 2511 2512 2513
	orig_daddr = fl4->daddr;
	orig_saddr = fl4->saddr;
	orig_oif = fl4->flowi4_oif;

	fl4->flowi4_iif = net->loopback_dev->ifindex;
	fl4->flowi4_tos = tos & IPTOS_RT_MASK;
	fl4->flowi4_scope = ((tos & RTO_ONLINK) ?
			 RT_SCOPE_LINK : RT_SCOPE_UNIVERSE);
2514

2515
	rcu_read_lock();
2516
	if (fl4->saddr) {
2517
		rth = ERR_PTR(-EINVAL);
2518 2519 2520
		if (ipv4_is_multicast(fl4->saddr) ||
		    ipv4_is_lbcast(fl4->saddr) ||
		    ipv4_is_zeronet(fl4->saddr))
L
Linus Torvalds 已提交
2521 2522 2523 2524
			goto out;

		/* I removed check for oif == dev_out->oif here.
		   It was wrong for two reasons:
2525 2526
		   1. ip_dev_find(net, saddr) can return wrong iface, if saddr
		      is assigned to multiple interfaces.
L
Linus Torvalds 已提交
2527 2528 2529 2530
		   2. Moreover, we are allowed to send packets with saddr
		      of another iface. --ANK
		 */

2531 2532 2533
		if (fl4->flowi4_oif == 0 &&
		    (ipv4_is_multicast(fl4->daddr) ||
		     ipv4_is_lbcast(fl4->daddr))) {
2534
			/* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2535
			dev_out = __ip_dev_find(net, fl4->saddr, false);
2536 2537 2538
			if (dev_out == NULL)
				goto out;

L
Linus Torvalds 已提交
2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553
			/* Special hack: user can direct multicasts
			   and limited broadcast via necessary interface
			   without fiddling with IP_MULTICAST_IF or IP_PKTINFO.
			   This hack is not just for fun, it allows
			   vic,vat and friends to work.
			   They bind socket to loopback, set ttl to zero
			   and expect that it will work.
			   From the viewpoint of routing cache they are broken,
			   because we are not allowed to build multicast path
			   with loopback source addr (look, routing cache
			   cannot know, that ttl is zero, so that packet
			   will not leave this host and route is valid).
			   Luckily, this hack is good workaround.
			 */

2554
			fl4->flowi4_oif = dev_out->ifindex;
L
Linus Torvalds 已提交
2555 2556
			goto make_route;
		}
2557

2558
		if (!(fl4->flowi4_flags & FLOWI_FLAG_ANYSRC)) {
2559
			/* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */
2560
			if (!__ip_dev_find(net, fl4->saddr, false))
2561 2562
				goto out;
		}
L
Linus Torvalds 已提交
2563 2564 2565
	}


2566 2567
	if (fl4->flowi4_oif) {
		dev_out = dev_get_by_index_rcu(net, fl4->flowi4_oif);
2568
		rth = ERR_PTR(-ENODEV);
L
Linus Torvalds 已提交
2569 2570
		if (dev_out == NULL)
			goto out;
2571 2572

		/* RACE: Check return value of inet_select_addr instead. */
2573
		if (!(dev_out->flags & IFF_UP) || !__in_dev_get_rcu(dev_out)) {
2574
			rth = ERR_PTR(-ENETUNREACH);
2575 2576
			goto out;
		}
2577 2578 2579 2580 2581
		if (ipv4_is_local_multicast(fl4->daddr) ||
		    ipv4_is_lbcast(fl4->daddr)) {
			if (!fl4->saddr)
				fl4->saddr = inet_select_addr(dev_out, 0,
							      RT_SCOPE_LINK);
L
Linus Torvalds 已提交
2582 2583
			goto make_route;
		}
2584 2585 2586 2587 2588 2589 2590
		if (fl4->saddr) {
			if (ipv4_is_multicast(fl4->daddr))
				fl4->saddr = inet_select_addr(dev_out, 0,
							      fl4->flowi4_scope);
			else if (!fl4->daddr)
				fl4->saddr = inet_select_addr(dev_out, 0,
							      RT_SCOPE_HOST);
L
Linus Torvalds 已提交
2591 2592 2593
		}
	}

2594 2595 2596 2597
	if (!fl4->daddr) {
		fl4->daddr = fl4->saddr;
		if (!fl4->daddr)
			fl4->daddr = fl4->saddr = htonl(INADDR_LOOPBACK);
2598
		dev_out = net->loopback_dev;
2599
		fl4->flowi4_oif = net->loopback_dev->ifindex;
L
Linus Torvalds 已提交
2600 2601 2602 2603 2604
		res.type = RTN_LOCAL;
		flags |= RTCF_LOCAL;
		goto make_route;
	}

2605
	if (fib_lookup(net, fl4, &res)) {
L
Linus Torvalds 已提交
2606
		res.fi = NULL;
2607
		if (fl4->flowi4_oif) {
L
Linus Torvalds 已提交
2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625
			/* Apparently, routing tables are wrong. Assume,
			   that the destination is on link.

			   WHY? DW.
			   Because we are allowed to send to iface
			   even if it has NO routes and NO assigned
			   addresses. When oif is specified, routing
			   tables are looked up with only one purpose:
			   to catch if destination is gatewayed, rather than
			   direct. Moreover, if MSG_DONTROUTE is set,
			   we send packet, ignoring both routing tables
			   and ifaddr state. --ANK


			   We could make it even if oif is unknown,
			   likely IPv6, but we do not.
			 */

2626 2627 2628
			if (fl4->saddr == 0)
				fl4->saddr = inet_select_addr(dev_out, 0,
							      RT_SCOPE_LINK);
L
Linus Torvalds 已提交
2629 2630 2631
			res.type = RTN_UNICAST;
			goto make_route;
		}
2632
		rth = ERR_PTR(-ENETUNREACH);
L
Linus Torvalds 已提交
2633 2634 2635 2636
		goto out;
	}

	if (res.type == RTN_LOCAL) {
2637
		if (!fl4->saddr) {
2638
			if (res.fi->fib_prefsrc)
2639
				fl4->saddr = res.fi->fib_prefsrc;
2640
			else
2641
				fl4->saddr = fl4->daddr;
2642
		}
2643
		dev_out = net->loopback_dev;
2644
		fl4->flowi4_oif = dev_out->ifindex;
L
Linus Torvalds 已提交
2645 2646 2647 2648 2649 2650
		res.fi = NULL;
		flags |= RTCF_LOCAL;
		goto make_route;
	}

#ifdef CONFIG_IP_ROUTE_MULTIPATH
2651
	if (res.fi->fib_nhs > 1 && fl4->flowi4_oif == 0)
2652
		fib_select_multipath(&res);
L
Linus Torvalds 已提交
2653 2654
	else
#endif
2655 2656
	if (!res.prefixlen &&
	    res.table->tb_num_default > 1 &&
2657
	    res.type == RTN_UNICAST && !fl4->flowi4_oif)
2658
		fib_select_default(&res);
L
Linus Torvalds 已提交
2659

2660 2661
	if (!fl4->saddr)
		fl4->saddr = FIB_RES_PREFSRC(net, res);
L
Linus Torvalds 已提交
2662 2663

	dev_out = FIB_RES_DEV(res);
2664
	fl4->flowi4_oif = dev_out->ifindex;
L
Linus Torvalds 已提交
2665 2666 2667


make_route:
2668 2669
	rth = __mkroute_output(&res, fl4, orig_daddr, orig_saddr, orig_oif,
			       dev_out, flags);
2670
	if (!IS_ERR(rth)) {
2671 2672
		unsigned int hash;

2673
		hash = rt_hash(orig_daddr, orig_saddr, orig_oif,
2674
			       rt_genid(dev_net(dev_out)));
2675
		rth = rt_intern_hash(hash, rth, NULL, orig_oif);
2676
	}
L
Linus Torvalds 已提交
2677

2678 2679
out:
	rcu_read_unlock();
2680
	return rth;
L
Linus Torvalds 已提交
2681 2682
}

2683
struct rtable *__ip_route_output_key(struct net *net, struct flowi4 *flp4)
L
Linus Torvalds 已提交
2684 2685
{
	struct rtable *rth;
2686
	unsigned int hash;
L
Linus Torvalds 已提交
2687

2688 2689 2690
	if (!rt_caching(net))
		goto slow_output;

2691
	hash = rt_hash(flp4->daddr, flp4->saddr, flp4->flowi4_oif, rt_genid(net));
L
Linus Torvalds 已提交
2692 2693

	rcu_read_lock_bh();
2694
	for (rth = rcu_dereference_bh(rt_hash_table[hash].chain); rth;
2695
		rth = rcu_dereference_bh(rth->dst.rt_next)) {
2696 2697
		if (rth->rt_key_dst == flp4->daddr &&
		    rth->rt_key_src == flp4->saddr &&
2698
		    rt_is_output_route(rth) &&
2699 2700
		    rth->rt_oif == flp4->flowi4_oif &&
		    rth->rt_mark == flp4->flowi4_mark &&
2701
		    !((rth->rt_key_tos ^ flp4->flowi4_tos) &
2702
			    (IPTOS_RT_MASK | RTO_ONLINK)) &&
2703
		    net_eq(dev_net(rth->dst.dev), net) &&
2704
		    !rt_is_expired(rth)) {
2705
			dst_use(&rth->dst, jiffies);
L
Linus Torvalds 已提交
2706 2707
			RT_CACHE_STAT_INC(out_hit);
			rcu_read_unlock_bh();
2708 2709 2710 2711
			if (!flp4->saddr)
				flp4->saddr = rth->rt_src;
			if (!flp4->daddr)
				flp4->daddr = rth->rt_dst;
2712
			return rth;
L
Linus Torvalds 已提交
2713 2714 2715 2716 2717
		}
		RT_CACHE_STAT_INC(out_hlist_search);
	}
	rcu_read_unlock_bh();

2718
slow_output:
2719
	return ip_route_output_slow(net, flp4);
L
Linus Torvalds 已提交
2720
}
2721 2722
EXPORT_SYMBOL_GPL(__ip_route_output_key);

2723 2724 2725 2726 2727
static struct dst_entry *ipv4_blackhole_dst_check(struct dst_entry *dst, u32 cookie)
{
	return NULL;
}

2728 2729 2730 2731 2732
static unsigned int ipv4_blackhole_default_mtu(const struct dst_entry *dst)
{
	return 0;
}

2733 2734 2735 2736
static void ipv4_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
{
}

2737 2738 2739 2740 2741 2742
static u32 *ipv4_rt_blackhole_cow_metrics(struct dst_entry *dst,
					  unsigned long old)
{
	return NULL;
}

2743 2744
static struct dst_ops ipv4_dst_blackhole_ops = {
	.family			=	AF_INET,
2745
	.protocol		=	cpu_to_be16(ETH_P_IP),
2746
	.destroy		=	ipv4_dst_destroy,
2747
	.check			=	ipv4_blackhole_dst_check,
2748
	.default_mtu		=	ipv4_blackhole_default_mtu,
2749
	.default_advmss		=	ipv4_default_advmss,
2750
	.update_pmtu		=	ipv4_rt_blackhole_update_pmtu,
2751
	.cow_metrics		=	ipv4_rt_blackhole_cow_metrics,
2752
	.neigh_lookup		=	ipv4_neigh_lookup,
2753 2754
};

2755
struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2756
{
2757
	struct rtable *rt = dst_alloc(&ipv4_dst_blackhole_ops, NULL, 1, 0, 0);
2758
	struct rtable *ort = (struct rtable *) dst_orig;
2759 2760

	if (rt) {
2761
		struct dst_entry *new = &rt->dst;
2762 2763

		new->__use = 1;
2764 2765
		new->input = dst_discard;
		new->output = dst_discard;
2766
		dst_copy_metrics(new, &ort->dst);
2767

2768
		new->dev = ort->dst.dev;
2769 2770 2771
		if (new->dev)
			dev_hold(new->dev);

2772 2773
		rt->rt_key_dst = ort->rt_key_dst;
		rt->rt_key_src = ort->rt_key_src;
2774
		rt->rt_key_tos = ort->rt_key_tos;
2775
		rt->rt_route_iif = ort->rt_route_iif;
2776 2777 2778
		rt->rt_iif = ort->rt_iif;
		rt->rt_oif = ort->rt_oif;
		rt->rt_mark = ort->rt_mark;
2779

2780
		rt->rt_genid = rt_genid(net);
2781 2782 2783 2784 2785 2786 2787 2788 2789
		rt->rt_flags = ort->rt_flags;
		rt->rt_type = ort->rt_type;
		rt->rt_dst = ort->rt_dst;
		rt->rt_src = ort->rt_src;
		rt->rt_gateway = ort->rt_gateway;
		rt->rt_spec_dst = ort->rt_spec_dst;
		rt->peer = ort->peer;
		if (rt->peer)
			atomic_inc(&rt->peer->refcnt);
2790 2791 2792
		rt->fi = ort->fi;
		if (rt->fi)
			atomic_inc(&rt->fi->fib_clntref);
2793 2794 2795 2796

		dst_free(new);
	}

2797 2798 2799
	dst_release(dst_orig);

	return rt ? &rt->dst : ERR_PTR(-ENOMEM);
2800 2801
}

2802
struct rtable *ip_route_output_flow(struct net *net, struct flowi4 *flp4,
2803
				    struct sock *sk)
L
Linus Torvalds 已提交
2804
{
2805
	struct rtable *rt = __ip_route_output_key(net, flp4);
L
Linus Torvalds 已提交
2806

2807 2808
	if (IS_ERR(rt))
		return rt;
L
Linus Torvalds 已提交
2809

2810
	if (flp4->flowi4_proto)
2811 2812 2813
		rt = (struct rtable *) xfrm_lookup(net, &rt->dst,
						   flowi4_to_flowi(flp4),
						   sk, 0);
L
Linus Torvalds 已提交
2814

2815
	return rt;
L
Linus Torvalds 已提交
2816
}
2817 2818
EXPORT_SYMBOL_GPL(ip_route_output_flow);

2819 2820
static int rt_fill_info(struct net *net,
			struct sk_buff *skb, u32 pid, u32 seq, int event,
2821
			int nowait, unsigned int flags)
L
Linus Torvalds 已提交
2822
{
E
Eric Dumazet 已提交
2823
	struct rtable *rt = skb_rtable(skb);
L
Linus Torvalds 已提交
2824
	struct rtmsg *r;
2825
	struct nlmsghdr *nlh;
E
Eric Dumazet 已提交
2826 2827
	long expires = 0;
	const struct inet_peer *peer = rt->peer;
2828
	u32 id = 0, ts = 0, tsage = 0, error;
2829 2830 2831

	nlh = nlmsg_put(skb, pid, seq, event, sizeof(*r), flags);
	if (nlh == NULL)
2832
		return -EMSGSIZE;
2833 2834

	r = nlmsg_data(nlh);
L
Linus Torvalds 已提交
2835 2836 2837
	r->rtm_family	 = AF_INET;
	r->rtm_dst_len	= 32;
	r->rtm_src_len	= 0;
2838
	r->rtm_tos	= rt->rt_key_tos;
L
Linus Torvalds 已提交
2839
	r->rtm_table	= RT_TABLE_MAIN;
2840
	NLA_PUT_U32(skb, RTA_TABLE, RT_TABLE_MAIN);
L
Linus Torvalds 已提交
2841 2842 2843 2844 2845 2846
	r->rtm_type	= rt->rt_type;
	r->rtm_scope	= RT_SCOPE_UNIVERSE;
	r->rtm_protocol = RTPROT_UNSPEC;
	r->rtm_flags	= (rt->rt_flags & ~0xFFFF) | RTM_F_CLONED;
	if (rt->rt_flags & RTCF_NOTIFY)
		r->rtm_flags |= RTM_F_NOTIFY;
2847

2848
	NLA_PUT_BE32(skb, RTA_DST, rt->rt_dst);
2849

2850
	if (rt->rt_key_src) {
L
Linus Torvalds 已提交
2851
		r->rtm_src_len = 32;
2852
		NLA_PUT_BE32(skb, RTA_SRC, rt->rt_key_src);
L
Linus Torvalds 已提交
2853
	}
2854 2855
	if (rt->dst.dev)
		NLA_PUT_U32(skb, RTA_OIF, rt->dst.dev->ifindex);
2856
#ifdef CONFIG_IP_ROUTE_CLASSID
2857 2858
	if (rt->dst.tclassid)
		NLA_PUT_U32(skb, RTA_FLOW, rt->dst.tclassid);
L
Linus Torvalds 已提交
2859
#endif
2860
	if (rt_is_input_route(rt))
2861
		NLA_PUT_BE32(skb, RTA_PREFSRC, rt->rt_spec_dst);
2862
	else if (rt->rt_src != rt->rt_key_src)
2863
		NLA_PUT_BE32(skb, RTA_PREFSRC, rt->rt_src);
2864

L
Linus Torvalds 已提交
2865
	if (rt->rt_dst != rt->rt_gateway)
2866
		NLA_PUT_BE32(skb, RTA_GATEWAY, rt->rt_gateway);
2867

2868
	if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2869 2870
		goto nla_put_failure;

2871 2872
	if (rt->rt_mark)
		NLA_PUT_BE32(skb, RTA_MARK, rt->rt_mark);
E
Eric Dumazet 已提交
2873

2874
	error = rt->dst.error;
E
Eric Dumazet 已提交
2875
	if (peer) {
2876
		inet_peer_refcheck(rt->peer);
E
Eric Dumazet 已提交
2877 2878 2879 2880
		id = atomic_read(&peer->ip_id_count) & 0xffff;
		if (peer->tcp_ts_stamp) {
			ts = peer->tcp_ts;
			tsage = get_seconds() - peer->tcp_ts_stamp;
L
Linus Torvalds 已提交
2881
		}
E
Eric Dumazet 已提交
2882 2883 2884
		expires = ACCESS_ONCE(peer->pmtu_expires);
		if (expires)
			expires -= jiffies;
L
Linus Torvalds 已提交
2885
	}
2886

2887
	if (rt_is_input_route(rt)) {
L
Linus Torvalds 已提交
2888
#ifdef CONFIG_IP_MROUTE
2889
		__be32 dst = rt->rt_dst;
L
Linus Torvalds 已提交
2890

2891
		if (ipv4_is_multicast(dst) && !ipv4_is_local_multicast(dst) &&
2892
		    IPV4_DEVCONF_ALL(net, MC_FORWARDING)) {
2893 2894 2895
			int err = ipmr_get_route(net, skb,
						 rt->rt_src, rt->rt_dst,
						 r, nowait);
L
Linus Torvalds 已提交
2896 2897 2898 2899
			if (err <= 0) {
				if (!nowait) {
					if (err == 0)
						return 0;
2900
					goto nla_put_failure;
L
Linus Torvalds 已提交
2901 2902
				} else {
					if (err == -EMSGSIZE)
2903
						goto nla_put_failure;
2904
					error = err;
L
Linus Torvalds 已提交
2905 2906 2907 2908
				}
			}
		} else
#endif
2909
			NLA_PUT_U32(skb, RTA_IIF, rt->rt_iif);
L
Linus Torvalds 已提交
2910 2911
	}

2912
	if (rtnl_put_cacheinfo(skb, &rt->dst, id, ts, tsage,
2913 2914
			       expires, error) < 0)
		goto nla_put_failure;
2915 2916

	return nlmsg_end(skb, nlh);
L
Linus Torvalds 已提交
2917

2918
nla_put_failure:
2919 2920
	nlmsg_cancel(skb, nlh);
	return -EMSGSIZE;
L
Linus Torvalds 已提交
2921 2922
}

2923
static int inet_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
L
Linus Torvalds 已提交
2924
{
2925
	struct net *net = sock_net(in_skb->sk);
2926 2927
	struct rtmsg *rtm;
	struct nlattr *tb[RTA_MAX+1];
L
Linus Torvalds 已提交
2928
	struct rtable *rt = NULL;
A
Al Viro 已提交
2929 2930 2931
	__be32 dst = 0;
	__be32 src = 0;
	u32 iif;
2932
	int err;
E
Eric Dumazet 已提交
2933
	int mark;
L
Linus Torvalds 已提交
2934 2935
	struct sk_buff *skb;

2936 2937 2938 2939 2940 2941
	err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv4_policy);
	if (err < 0)
		goto errout;

	rtm = nlmsg_data(nlh);

L
Linus Torvalds 已提交
2942
	skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2943 2944 2945 2946
	if (skb == NULL) {
		err = -ENOBUFS;
		goto errout;
	}
L
Linus Torvalds 已提交
2947 2948 2949 2950

	/* Reserve room for dummy headers, this skb can pass
	   through good chunk of routing engine.
	 */
2951
	skb_reset_mac_header(skb);
2952
	skb_reset_network_header(skb);
2953 2954

	/* Bugfix: need to give ip_route_input enough of an IP header to not gag. */
2955
	ip_hdr(skb)->protocol = IPPROTO_ICMP;
L
Linus Torvalds 已提交
2956 2957
	skb_reserve(skb, MAX_HEADER + sizeof(struct iphdr));

2958 2959
	src = tb[RTA_SRC] ? nla_get_be32(tb[RTA_SRC]) : 0;
	dst = tb[RTA_DST] ? nla_get_be32(tb[RTA_DST]) : 0;
2960
	iif = tb[RTA_IIF] ? nla_get_u32(tb[RTA_IIF]) : 0;
E
Eric Dumazet 已提交
2961
	mark = tb[RTA_MARK] ? nla_get_u32(tb[RTA_MARK]) : 0;
L
Linus Torvalds 已提交
2962 2963

	if (iif) {
2964 2965
		struct net_device *dev;

2966
		dev = __dev_get_by_index(net, iif);
2967 2968 2969 2970 2971
		if (dev == NULL) {
			err = -ENODEV;
			goto errout_free;
		}

L
Linus Torvalds 已提交
2972 2973
		skb->protocol	= htons(ETH_P_IP);
		skb->dev	= dev;
E
Eric Dumazet 已提交
2974
		skb->mark	= mark;
L
Linus Torvalds 已提交
2975 2976 2977
		local_bh_disable();
		err = ip_route_input(skb, dst, src, rtm->rtm_tos, dev);
		local_bh_enable();
2978

E
Eric Dumazet 已提交
2979
		rt = skb_rtable(skb);
2980 2981
		if (err == 0 && rt->dst.error)
			err = -rt->dst.error;
L
Linus Torvalds 已提交
2982
	} else {
2983 2984 2985 2986 2987 2988
		struct flowi4 fl4 = {
			.daddr = dst,
			.saddr = src,
			.flowi4_tos = rtm->rtm_tos,
			.flowi4_oif = tb[RTA_OIF] ? nla_get_u32(tb[RTA_OIF]) : 0,
			.flowi4_mark = mark,
2989
		};
2990
		rt = ip_route_output_key(net, &fl4);
2991 2992 2993 2994

		err = 0;
		if (IS_ERR(rt))
			err = PTR_ERR(rt);
L
Linus Torvalds 已提交
2995
	}
2996

L
Linus Torvalds 已提交
2997
	if (err)
2998
		goto errout_free;
L
Linus Torvalds 已提交
2999

3000
	skb_dst_set(skb, &rt->dst);
L
Linus Torvalds 已提交
3001 3002 3003
	if (rtm->rtm_flags & RTM_F_NOTIFY)
		rt->rt_flags |= RTCF_NOTIFY;

3004
	err = rt_fill_info(net, skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq,
3005
			   RTM_NEWROUTE, 0, 0);
3006 3007
	if (err <= 0)
		goto errout_free;
L
Linus Torvalds 已提交
3008

3009
	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3010
errout:
3011
	return err;
L
Linus Torvalds 已提交
3012

3013
errout_free:
L
Linus Torvalds 已提交
3014
	kfree_skb(skb);
3015
	goto errout;
L
Linus Torvalds 已提交
3016 3017 3018 3019 3020 3021 3022
}

int ip_rt_dump(struct sk_buff *skb,  struct netlink_callback *cb)
{
	struct rtable *rt;
	int h, s_h;
	int idx, s_idx;
3023 3024
	struct net *net;

3025
	net = sock_net(skb->sk);
L
Linus Torvalds 已提交
3026 3027

	s_h = cb->args[0];
3028 3029
	if (s_h < 0)
		s_h = 0;
L
Linus Torvalds 已提交
3030
	s_idx = idx = cb->args[1];
3031 3032 3033
	for (h = s_h; h <= rt_hash_mask; h++, s_idx = 0) {
		if (!rt_hash_table[h].chain)
			continue;
L
Linus Torvalds 已提交
3034
		rcu_read_lock_bh();
3035
		for (rt = rcu_dereference_bh(rt_hash_table[h].chain), idx = 0; rt;
3036 3037
		     rt = rcu_dereference_bh(rt->dst.rt_next), idx++) {
			if (!net_eq(dev_net(rt->dst.dev), net) || idx < s_idx)
L
Linus Torvalds 已提交
3038
				continue;
3039
			if (rt_is_expired(rt))
3040
				continue;
3041
			skb_dst_set_noref(skb, &rt->dst);
3042
			if (rt_fill_info(net, skb, NETLINK_CB(cb->skb).pid,
3043
					 cb->nlh->nlmsg_seq, RTM_NEWROUTE,
3044
					 1, NLM_F_MULTI) <= 0) {
E
Eric Dumazet 已提交
3045
				skb_dst_drop(skb);
L
Linus Torvalds 已提交
3046 3047 3048
				rcu_read_unlock_bh();
				goto done;
			}
E
Eric Dumazet 已提交
3049
			skb_dst_drop(skb);
L
Linus Torvalds 已提交
3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061
		}
		rcu_read_unlock_bh();
	}

done:
	cb->args[0] = h;
	cb->args[1] = idx;
	return skb->len;
}

void ip_rt_multicast_event(struct in_device *in_dev)
{
3062
	rt_cache_flush(dev_net(in_dev->dev), 0);
L
Linus Torvalds 已提交
3063 3064 3065
}

#ifdef CONFIG_SYSCTL
3066
static int ipv4_sysctl_rtcache_flush(ctl_table *__ctl, int write,
3067
					void __user *buffer,
L
Linus Torvalds 已提交
3068 3069 3070
					size_t *lenp, loff_t *ppos)
{
	if (write) {
3071
		int flush_delay;
3072
		ctl_table ctl;
3073
		struct net *net;
3074

3075 3076
		memcpy(&ctl, __ctl, sizeof(ctl));
		ctl.data = &flush_delay;
3077
		proc_dointvec(&ctl, write, buffer, lenp, ppos);
3078

3079
		net = (struct net *)__ctl->extra1;
3080
		rt_cache_flush(net, flush_delay);
L
Linus Torvalds 已提交
3081
		return 0;
3082
	}
L
Linus Torvalds 已提交
3083 3084 3085 3086

	return -EINVAL;
}

A
Al Viro 已提交
3087
static ctl_table ipv4_route_table[] = {
L
Linus Torvalds 已提交
3088 3089 3090 3091 3092
	{
		.procname	= "gc_thresh",
		.data		= &ipv4_dst_ops.gc_thresh,
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
3093
		.proc_handler	= proc_dointvec,
L
Linus Torvalds 已提交
3094 3095 3096 3097 3098 3099
	},
	{
		.procname	= "max_size",
		.data		= &ip_rt_max_size,
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
3100
		.proc_handler	= proc_dointvec,
L
Linus Torvalds 已提交
3101 3102 3103
	},
	{
		/*  Deprecated. Use gc_min_interval_ms */
3104

L
Linus Torvalds 已提交
3105 3106 3107 3108
		.procname	= "gc_min_interval",
		.data		= &ip_rt_gc_min_interval,
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
3109
		.proc_handler	= proc_dointvec_jiffies,
L
Linus Torvalds 已提交
3110 3111 3112 3113 3114 3115
	},
	{
		.procname	= "gc_min_interval_ms",
		.data		= &ip_rt_gc_min_interval,
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
3116
		.proc_handler	= proc_dointvec_ms_jiffies,
L
Linus Torvalds 已提交
3117 3118 3119 3120 3121 3122
	},
	{
		.procname	= "gc_timeout",
		.data		= &ip_rt_gc_timeout,
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
3123
		.proc_handler	= proc_dointvec_jiffies,
L
Linus Torvalds 已提交
3124 3125 3126 3127 3128 3129
	},
	{
		.procname	= "gc_interval",
		.data		= &ip_rt_gc_interval,
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
3130
		.proc_handler	= proc_dointvec_jiffies,
L
Linus Torvalds 已提交
3131 3132 3133 3134 3135 3136
	},
	{
		.procname	= "redirect_load",
		.data		= &ip_rt_redirect_load,
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
3137
		.proc_handler	= proc_dointvec,
L
Linus Torvalds 已提交
3138 3139 3140 3141 3142 3143
	},
	{
		.procname	= "redirect_number",
		.data		= &ip_rt_redirect_number,
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
3144
		.proc_handler	= proc_dointvec,
L
Linus Torvalds 已提交
3145 3146 3147 3148 3149 3150
	},
	{
		.procname	= "redirect_silence",
		.data		= &ip_rt_redirect_silence,
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
3151
		.proc_handler	= proc_dointvec,
L
Linus Torvalds 已提交
3152 3153 3154 3155 3156 3157
	},
	{
		.procname	= "error_cost",
		.data		= &ip_rt_error_cost,
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
3158
		.proc_handler	= proc_dointvec,
L
Linus Torvalds 已提交
3159 3160 3161 3162 3163 3164
	},
	{
		.procname	= "error_burst",
		.data		= &ip_rt_error_burst,
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
3165
		.proc_handler	= proc_dointvec,
L
Linus Torvalds 已提交
3166 3167 3168 3169 3170 3171
	},
	{
		.procname	= "gc_elasticity",
		.data		= &ip_rt_gc_elasticity,
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
3172
		.proc_handler	= proc_dointvec,
L
Linus Torvalds 已提交
3173 3174 3175 3176 3177 3178
	},
	{
		.procname	= "mtu_expires",
		.data		= &ip_rt_mtu_expires,
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
3179
		.proc_handler	= proc_dointvec_jiffies,
L
Linus Torvalds 已提交
3180 3181 3182 3183 3184 3185
	},
	{
		.procname	= "min_pmtu",
		.data		= &ip_rt_min_pmtu,
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
3186
		.proc_handler	= proc_dointvec,
L
Linus Torvalds 已提交
3187 3188 3189 3190 3191 3192
	},
	{
		.procname	= "min_adv_mss",
		.data		= &ip_rt_min_advmss,
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
3193
		.proc_handler	= proc_dointvec,
L
Linus Torvalds 已提交
3194
	},
3195
	{ }
L
Linus Torvalds 已提交
3196
};
3197

A
Al Viro 已提交
3198 3199 3200 3201
static struct ctl_table empty[1];

static struct ctl_table ipv4_skeleton[] =
{
3202
	{ .procname = "route", 
H
Hugh Dickins 已提交
3203
	  .mode = 0555, .child = ipv4_route_table},
3204
	{ .procname = "neigh", 
H
Hugh Dickins 已提交
3205
	  .mode = 0555, .child = empty},
A
Al Viro 已提交
3206 3207 3208 3209
	{ }
};

static __net_initdata struct ctl_path ipv4_path[] = {
3210 3211
	{ .procname = "net", },
	{ .procname = "ipv4", },
3212 3213 3214 3215 3216 3217 3218 3219
	{ },
};

static struct ctl_table ipv4_route_flush_table[] = {
	{
		.procname	= "flush",
		.maxlen		= sizeof(int),
		.mode		= 0200,
A
Alexey Dobriyan 已提交
3220
		.proc_handler	= ipv4_sysctl_rtcache_flush,
3221
	},
3222
	{ },
3223 3224
};

A
Al Viro 已提交
3225
static __net_initdata struct ctl_path ipv4_route_path[] = {
3226 3227 3228
	{ .procname = "net", },
	{ .procname = "ipv4", },
	{ .procname = "route", },
A
Al Viro 已提交
3229 3230 3231
	{ },
};

3232 3233 3234 3235 3236
static __net_init int sysctl_route_net_init(struct net *net)
{
	struct ctl_table *tbl;

	tbl = ipv4_route_flush_table;
O
Octavian Purdila 已提交
3237
	if (!net_eq(net, &init_net)) {
3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270
		tbl = kmemdup(tbl, sizeof(ipv4_route_flush_table), GFP_KERNEL);
		if (tbl == NULL)
			goto err_dup;
	}
	tbl[0].extra1 = net;

	net->ipv4.route_hdr =
		register_net_sysctl_table(net, ipv4_route_path, tbl);
	if (net->ipv4.route_hdr == NULL)
		goto err_reg;
	return 0;

err_reg:
	if (tbl != ipv4_route_flush_table)
		kfree(tbl);
err_dup:
	return -ENOMEM;
}

static __net_exit void sysctl_route_net_exit(struct net *net)
{
	struct ctl_table *tbl;

	tbl = net->ipv4.route_hdr->ctl_table_arg;
	unregister_net_sysctl_table(net->ipv4.route_hdr);
	BUG_ON(tbl == ipv4_route_flush_table);
	kfree(tbl);
}

static __net_initdata struct pernet_operations sysctl_route_ops = {
	.init = sysctl_route_net_init,
	.exit = sysctl_route_net_exit,
};
L
Linus Torvalds 已提交
3271 3272
#endif

3273
static __net_init int rt_genid_init(struct net *net)
3274
{
3275 3276
	get_random_bytes(&net->ipv4.rt_genid,
			 sizeof(net->ipv4.rt_genid));
3277 3278
	get_random_bytes(&net->ipv4.dev_addr_genid,
			 sizeof(net->ipv4.dev_addr_genid));
3279 3280 3281
	return 0;
}

3282 3283
static __net_initdata struct pernet_operations rt_genid_ops = {
	.init = rt_genid_init,
3284 3285 3286
};


3287
#ifdef CONFIG_IP_ROUTE_CLASSID
3288
struct ip_rt_acct __percpu *ip_rt_acct __read_mostly;
3289
#endif /* CONFIG_IP_ROUTE_CLASSID */
L
Linus Torvalds 已提交
3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302

static __initdata unsigned long rhash_entries;
static int __init set_rhash_entries(char *str)
{
	if (!str)
		return 0;
	rhash_entries = simple_strtoul(str, &str, 0);
	return 1;
}
__setup("rhash_entries=", set_rhash_entries);

int __init ip_rt_init(void)
{
3303
	int rc = 0;
L
Linus Torvalds 已提交
3304

3305
#ifdef CONFIG_IP_ROUTE_CLASSID
3306
	ip_rt_acct = __alloc_percpu(256 * sizeof(struct ip_rt_acct), __alignof__(struct ip_rt_acct));
L
Linus Torvalds 已提交
3307 3308 3309 3310
	if (!ip_rt_acct)
		panic("IP: failed to allocate ip_rt_acct\n");
#endif

A
Alexey Dobriyan 已提交
3311 3312
	ipv4_dst_ops.kmem_cachep =
		kmem_cache_create("ip_dst_cache", sizeof(struct rtable), 0,
3313
				  SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
L
Linus Torvalds 已提交
3314

3315 3316
	ipv4_dst_blackhole_ops.kmem_cachep = ipv4_dst_ops.kmem_cachep;

3317 3318 3319 3320 3321 3322
	if (dst_entries_init(&ipv4_dst_ops) < 0)
		panic("IP: failed to allocate ipv4_dst_ops counter\n");

	if (dst_entries_init(&ipv4_dst_blackhole_ops) < 0)
		panic("IP: failed to allocate ipv4_dst_blackhole_ops counter\n");

3323 3324 3325 3326
	rt_hash_table = (struct rt_hash_bucket *)
		alloc_large_system_hash("IP route cache",
					sizeof(struct rt_hash_bucket),
					rhash_entries,
3327
					(totalram_pages >= 128 * 1024) ?
3328
					15 : 17,
3329
					0,
3330 3331
					&rt_hash_log,
					&rt_hash_mask,
3332
					rhash_entries ? 0 : 512 * 1024);
3333 3334
	memset(rt_hash_table, 0, (rt_hash_mask + 1) * sizeof(struct rt_hash_bucket));
	rt_hash_lock_init();
L
Linus Torvalds 已提交
3335 3336 3337 3338 3339 3340 3341

	ipv4_dst_ops.gc_thresh = (rt_hash_mask + 1);
	ip_rt_max_size = (rt_hash_mask + 1) * 16;

	devinet_init();
	ip_fib_init();

3342
	if (ip_rt_proc_init())
3343
		printk(KERN_ERR "Unable to create route proc files\n");
L
Linus Torvalds 已提交
3344 3345
#ifdef CONFIG_XFRM
	xfrm_init();
3346
	xfrm4_init(ip_rt_max_size);
L
Linus Torvalds 已提交
3347
#endif
3348
	rtnl_register(PF_INET, RTM_GETROUTE, inet_rtm_getroute, NULL, NULL);
3349

3350 3351 3352
#ifdef CONFIG_SYSCTL
	register_pernet_subsys(&sysctl_route_ops);
#endif
3353
	register_pernet_subsys(&rt_genid_ops);
L
Linus Torvalds 已提交
3354 3355 3356
	return rc;
}

3357
#ifdef CONFIG_SYSCTL
A
Al Viro 已提交
3358 3359 3360 3361 3362 3363
/*
 * We really need to sanitize the damn ipv4 init order, then all
 * this nonsense will go away.
 */
void __init ip_static_sysctl_init(void)
{
A
Al Viro 已提交
3364
	register_sysctl_paths(ipv4_path, ipv4_skeleton);
A
Al Viro 已提交
3365
}
3366
#endif