ip_fragment.c 18.0 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6
/*
 * 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.
 *
 *		The IP fragmentation functionality.
7
 *
L
Linus Torvalds 已提交
8
 * Authors:	Fred N. van Kempen <waltje@uWalt.NL.Mugnet.ORG>
9
 *		Alan Cox <alan@lxorguk.ukuu.org.uk>
L
Linus Torvalds 已提交
10 11 12 13 14 15 16 17 18 19 20 21 22
 *
 * Fixes:
 *		Alan Cox	:	Split from ip.c , see ip_input.c for history.
 *		David S. Miller :	Begin massive cleanup...
 *		Andi Kleen	:	Add sysctls.
 *		xxxx		:	Overlapfrag bug.
 *		Ultima          :       ip_expire() kernel panic.
 *		Bill Hawes	:	Frag accounting and evictor fixes.
 *		John McDonald	:	0 length frag bug.
 *		Alexey Kuznetsov:	SMP races, threading, cleanup.
 *		Patrick McHardy :	LRU queue of frag heads for evictor.
 */

H
Herbert Xu 已提交
23
#include <linux/compiler.h>
L
Linus Torvalds 已提交
24 25 26 27 28 29 30 31 32 33 34
#include <linux/module.h>
#include <linux/types.h>
#include <linux/mm.h>
#include <linux/jiffies.h>
#include <linux/skbuff.h>
#include <linux/list.h>
#include <linux/ip.h>
#include <linux/icmp.h>
#include <linux/netdevice.h>
#include <linux/jhash.h>
#include <linux/random.h>
35
#include <linux/slab.h>
36 37
#include <net/route.h>
#include <net/dst.h>
L
Linus Torvalds 已提交
38 39 40 41
#include <net/sock.h>
#include <net/ip.h>
#include <net/icmp.h>
#include <net/checksum.h>
H
Herbert Xu 已提交
42
#include <net/inetpeer.h>
43
#include <net/inet_frag.h>
L
Linus Torvalds 已提交
44 45 46 47 48 49 50 51 52 53
#include <linux/tcp.h>
#include <linux/udp.h>
#include <linux/inet.h>
#include <linux/netfilter_ipv4.h>

/* NOTE. Logic of IP defragmentation is parallel to corresponding IPv6
 * code now. If you change something here, _PLEASE_ update ipv6/reassembly.c
 * as well. Or notify me, at least. --ANK
 */

54
static int sysctl_ipfrag_max_dist __read_mostly = 64;
H
Herbert Xu 已提交
55

L
Linus Torvalds 已提交
56 57 58 59 60 61
struct ipfrag_skb_cb
{
	struct inet_skb_parm	h;
	int			offset;
};

62
#define FRAG_CB(skb)	((struct ipfrag_skb_cb *)((skb)->cb))
L
Linus Torvalds 已提交
63 64 65

/* Describe an entry in the "incomplete datagrams" queue. */
struct ipq {
66 67
	struct inet_frag_queue q;

L
Linus Torvalds 已提交
68
	u32		user;
69 70 71
	__be32		saddr;
	__be32		daddr;
	__be16		id;
L
Linus Torvalds 已提交
72
	u8		protocol;
H
Herbert Xu 已提交
73 74 75
	int             iif;
	unsigned int    rid;
	struct inet_peer *peer;
L
Linus Torvalds 已提交
76 77
};

78
static struct inet_frags ip4_frags;
L
Linus Torvalds 已提交
79

80
int ip_frag_nqueues(struct net *net)
81
{
82
	return net->ipv4.frags.nqueues;
83
}
L
Linus Torvalds 已提交
84

85
int ip_frag_mem(struct net *net)
86
{
87
	return atomic_read(&net->ipv4.frags.mem);
88
}
L
Linus Torvalds 已提交
89

90 91 92
static int ip_frag_reasm(struct ipq *qp, struct sk_buff *prev,
			 struct net_device *dev);

93 94 95 96 97
struct ip4_create_arg {
	struct iphdr *iph;
	u32 user;
};

98
static unsigned int ipqhashfn(__be16 id, __be32 saddr, __be32 daddr, u8 prot)
L
Linus Torvalds 已提交
99
{
100 101
	return jhash_3words((__force u32)id << 16 | prot,
			    (__force u32)saddr, (__force u32)daddr,
102
			    ip4_frags.rnd) & (INETFRAGS_HASHSZ - 1);
L
Linus Torvalds 已提交
103 104
}

105
static unsigned int ip4_hashfn(struct inet_frag_queue *q)
L
Linus Torvalds 已提交
106
{
107
	struct ipq *ipq;
L
Linus Torvalds 已提交
108

109 110
	ipq = container_of(q, struct ipq, q);
	return ipqhashfn(ipq->id, ipq->saddr, ipq->daddr, ipq->protocol);
L
Linus Torvalds 已提交
111 112
}

113 114 115 116 117 118 119 120 121 122 123 124 125
static int ip4_frag_match(struct inet_frag_queue *q, void *a)
{
	struct ipq *qp;
	struct ip4_create_arg *arg = a;

	qp = container_of(q, struct ipq, q);
	return (qp->id == arg->iph->id &&
			qp->saddr == arg->iph->saddr &&
			qp->daddr == arg->iph->daddr &&
			qp->protocol == arg->iph->protocol &&
			qp->user == arg->user);
}

L
Linus Torvalds 已提交
126
/* Memory Tracking Functions. */
127 128
static __inline__ void frag_kfree_skb(struct netns_frags *nf,
		struct sk_buff *skb, int *work)
L
Linus Torvalds 已提交
129 130 131
{
	if (work)
		*work -= skb->truesize;
132
	atomic_sub(skb->truesize, &nf->mem);
L
Linus Torvalds 已提交
133 134 135
	kfree_skb(skb);
}

136 137 138 139 140 141 142 143 144 145 146 147 148 149
static void ip4_frag_init(struct inet_frag_queue *q, void *a)
{
	struct ipq *qp = container_of(q, struct ipq, q);
	struct ip4_create_arg *arg = a;

	qp->protocol = arg->iph->protocol;
	qp->id = arg->iph->id;
	qp->saddr = arg->iph->saddr;
	qp->daddr = arg->iph->daddr;
	qp->user = arg->user;
	qp->peer = sysctl_ipfrag_max_dist ?
		inet_getpeer(arg->iph->saddr, 1) : NULL;
}

150
static __inline__ void ip4_frag_free(struct inet_frag_queue *q)
L
Linus Torvalds 已提交
151
{
152 153 154 155 156
	struct ipq *qp;

	qp = container_of(q, struct ipq, q);
	if (qp->peer)
		inet_putpeer(qp->peer);
L
Linus Torvalds 已提交
157 158 159 160 161
}


/* Destruction primitives. */

162
static __inline__ void ipq_put(struct ipq *ipq)
L
Linus Torvalds 已提交
163
{
P
Pavel Emelyanov 已提交
164
	inet_frag_put(&ipq->q, &ip4_frags);
L
Linus Torvalds 已提交
165 166 167 168 169 170 171
}

/* Kill ipq entry. It is not destroyed immediately,
 * because caller (and someone more) holds reference count.
 */
static void ipq_kill(struct ipq *ipq)
{
172
	inet_frag_kill(&ipq->q, &ip4_frags);
L
Linus Torvalds 已提交
173 174
}

175
/* Memory limiting on fragments.  Evictor trashes the oldest
L
Linus Torvalds 已提交
176 177
 * fragment queue until we are back under the threshold.
 */
178
static void ip_evictor(struct net *net)
L
Linus Torvalds 已提交
179
{
180 181
	int evicted;

182
	evicted = inet_frag_evictor(&net->ipv4.frags, &ip4_frags);
183
	if (evicted)
184
		IP_ADD_STATS_BH(net, IPSTATS_MIB_REASMFAILS, evicted);
L
Linus Torvalds 已提交
185 186 187 188 189 190 191
}

/*
 * Oops, a fragment queue timed out.  Kill it and send an ICMP reply.
 */
static void ip_expire(unsigned long arg)
{
192
	struct ipq *qp;
193
	struct net *net;
194 195

	qp = container_of((struct inet_frag_queue *) arg, struct ipq, q);
196
	net = container_of(qp->q.net, struct net, ipv4.frags);
L
Linus Torvalds 已提交
197

198
	spin_lock(&qp->q.lock);
L
Linus Torvalds 已提交
199

200
	if (qp->q.last_in & INET_FRAG_COMPLETE)
L
Linus Torvalds 已提交
201 202 203 204
		goto out;

	ipq_kill(qp);

205 206
	IP_INC_STATS_BH(net, IPSTATS_MIB_REASMTIMEOUT);
	IP_INC_STATS_BH(net, IPSTATS_MIB_REASMFAILS);
L
Linus Torvalds 已提交
207

208
	if ((qp->q.last_in & INET_FRAG_FIRST_IN) && qp->q.fragments != NULL) {
209
		struct sk_buff *head = qp->q.fragments;
210

E
Eric Dumazet 已提交
211 212
		rcu_read_lock();
		head->dev = dev_get_by_index_rcu(net, qp->iif);
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
		if (!head->dev)
			goto out_rcu_unlock;

		/*
		 * Only search router table for the head fragment,
		 * when defraging timeout at PRE_ROUTING HOOK.
		 */
		if (qp->user == IP_DEFRAG_CONNTRACK_IN && !skb_dst(head)) {
			const struct iphdr *iph = ip_hdr(head);
			int err = ip_route_input(head, iph->daddr, iph->saddr,
						 iph->tos, head->dev);
			if (unlikely(err))
				goto out_rcu_unlock;

			/*
			 * Only an end host needs to send an ICMP
			 * "Fragment Reassembly Timeout" message, per RFC792.
			 */
			if (skb_rtable(head)->rt_type != RTN_LOCAL)
				goto out_rcu_unlock;

		}

		/* Send an ICMP "Fragment Reassembly Timeout" message. */
		icmp_send(head, ICMP_TIME_EXCEEDED, ICMP_EXC_FRAGTIME, 0);
out_rcu_unlock:
239 240
		rcu_read_unlock();
	}
L
Linus Torvalds 已提交
241
out:
242
	spin_unlock(&qp->q.lock);
243
	ipq_put(qp);
L
Linus Torvalds 已提交
244 245
}

246 247 248
/* Find the correct entry in the "incomplete datagrams" queue for
 * this IP datagram, and create new one, if nothing is found.
 */
249
static inline struct ipq *ip_find(struct net *net, struct iphdr *iph, u32 user)
L
Linus Torvalds 已提交
250
{
251 252
	struct inet_frag_queue *q;
	struct ip4_create_arg arg;
253
	unsigned int hash;
L
Linus Torvalds 已提交
254

255 256
	arg.iph = iph;
	arg.user = user;
257 258

	read_lock(&ip4_frags.lock);
259
	hash = ipqhashfn(iph->id, iph->saddr, iph->daddr, iph->protocol);
L
Linus Torvalds 已提交
260

261
	q = inet_frag_find(&net->ipv4.frags, &ip4_frags, &arg, hash);
262 263
	if (q == NULL)
		goto out_nomem;
L
Linus Torvalds 已提交
264

265
	return container_of(q, struct ipq, q);
L
Linus Torvalds 已提交
266 267

out_nomem:
268
	LIMIT_NETDEBUG(KERN_ERR "ip_frag_create: no memory left !\n");
L
Linus Torvalds 已提交
269 270 271
	return NULL;
}

H
Herbert Xu 已提交
272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287
/* Is the fragment too far ahead to be part of ipq? */
static inline int ip_frag_too_far(struct ipq *qp)
{
	struct inet_peer *peer = qp->peer;
	unsigned int max = sysctl_ipfrag_max_dist;
	unsigned int start, end;

	int rc;

	if (!peer || !max)
		return 0;

	start = qp->rid;
	end = atomic_inc_return(&peer->rid);
	qp->rid = end;

288
	rc = qp->q.fragments && (end - start) > max;
H
Herbert Xu 已提交
289 290

	if (rc) {
291 292 293 294
		struct net *net;

		net = container_of(qp->q.net, struct net, ipv4.frags);
		IP_INC_STATS_BH(net, IPSTATS_MIB_REASMFAILS);
H
Herbert Xu 已提交
295 296 297 298 299 300 301 302 303
	}

	return rc;
}

static int ip_frag_reinit(struct ipq *qp)
{
	struct sk_buff *fp;

304
	if (!mod_timer(&qp->q.timer, jiffies + qp->q.net->timeout)) {
305
		atomic_inc(&qp->q.refcnt);
H
Herbert Xu 已提交
306 307 308
		return -ETIMEDOUT;
	}

309
	fp = qp->q.fragments;
H
Herbert Xu 已提交
310 311
	do {
		struct sk_buff *xp = fp->next;
312
		frag_kfree_skb(qp->q.net, fp, NULL);
H
Herbert Xu 已提交
313 314 315
		fp = xp;
	} while (fp);

316 317 318 319
	qp->q.last_in = 0;
	qp->q.len = 0;
	qp->q.meat = 0;
	qp->q.fragments = NULL;
H
Herbert Xu 已提交
320 321 322 323 324
	qp->iif = 0;

	return 0;
}

L
Linus Torvalds 已提交
325
/* Add new segment to existing queue. */
326
static int ip_frag_queue(struct ipq *qp, struct sk_buff *skb)
L
Linus Torvalds 已提交
327 328
{
	struct sk_buff *prev, *next;
329
	struct net_device *dev;
L
Linus Torvalds 已提交
330 331
	int flags, offset;
	int ihl, end;
332
	int err = -ENOENT;
L
Linus Torvalds 已提交
333

334
	if (qp->q.last_in & INET_FRAG_COMPLETE)
L
Linus Torvalds 已提交
335 336
		goto err;

H
Herbert Xu 已提交
337
	if (!(IPCB(skb)->flags & IPSKB_FRAG_COMPLETE) &&
338 339
	    unlikely(ip_frag_too_far(qp)) &&
	    unlikely(err = ip_frag_reinit(qp))) {
H
Herbert Xu 已提交
340 341 342 343
		ipq_kill(qp);
		goto err;
	}

344
	offset = ntohs(ip_hdr(skb)->frag_off);
L
Linus Torvalds 已提交
345 346 347
	flags = offset & ~IP_OFFSET;
	offset &= IP_OFFSET;
	offset <<= 3;		/* offset is in 8-byte chunks */
348
	ihl = ip_hdrlen(skb);
L
Linus Torvalds 已提交
349 350

	/* Determine the position of this fragment. */
351
	end = offset + skb->len - ihl;
352
	err = -EINVAL;
L
Linus Torvalds 已提交
353 354 355 356 357 358

	/* Is this the final fragment? */
	if ((flags & IP_MF) == 0) {
		/* If we already have some bits beyond end
		 * or have different end, the segment is corrrupted.
		 */
359
		if (end < qp->q.len ||
360
		    ((qp->q.last_in & INET_FRAG_LAST_IN) && end != qp->q.len))
L
Linus Torvalds 已提交
361
			goto err;
362
		qp->q.last_in |= INET_FRAG_LAST_IN;
363
		qp->q.len = end;
L
Linus Torvalds 已提交
364 365 366 367 368 369
	} else {
		if (end&7) {
			end &= ~7;
			if (skb->ip_summed != CHECKSUM_UNNECESSARY)
				skb->ip_summed = CHECKSUM_NONE;
		}
370
		if (end > qp->q.len) {
L
Linus Torvalds 已提交
371
			/* Some bits beyond end -> corruption. */
372
			if (qp->q.last_in & INET_FRAG_LAST_IN)
L
Linus Torvalds 已提交
373
				goto err;
374
			qp->q.len = end;
L
Linus Torvalds 已提交
375 376 377 378 379
		}
	}
	if (end == offset)
		goto err;

380
	err = -ENOMEM;
L
Linus Torvalds 已提交
381 382
	if (pskb_pull(skb, ihl) == NULL)
		goto err;
383 384 385

	err = pskb_trim_rcsum(skb, end - offset);
	if (err)
L
Linus Torvalds 已提交
386 387 388 389 390 391 392
		goto err;

	/* Find out which fragments are in front and at the back of us
	 * in the chain of fragments so far.  We must know where to put
	 * this fragment, right?
	 */
	prev = NULL;
393
	for (next = qp->q.fragments; next != NULL; next = next->next) {
L
Linus Torvalds 已提交
394 395 396 397 398 399 400 401 402 403 404 405 406 407
		if (FRAG_CB(next)->offset >= offset)
			break;	/* bingo! */
		prev = next;
	}

	/* We found where to put this one.  Check for overlap with
	 * preceding fragment, and, if needed, align things so that
	 * any overlaps are eliminated.
	 */
	if (prev) {
		int i = (FRAG_CB(prev)->offset + prev->len) - offset;

		if (i > 0) {
			offset += i;
408
			err = -EINVAL;
L
Linus Torvalds 已提交
409 410
			if (end <= offset)
				goto err;
411
			err = -ENOMEM;
L
Linus Torvalds 已提交
412 413 414 415 416 417 418
			if (!pskb_pull(skb, i))
				goto err;
			if (skb->ip_summed != CHECKSUM_UNNECESSARY)
				skb->ip_summed = CHECKSUM_NONE;
		}
	}

419 420
	err = -ENOMEM;

L
Linus Torvalds 已提交
421 422 423 424 425 426 427 428 429 430
	while (next && FRAG_CB(next)->offset < end) {
		int i = end - FRAG_CB(next)->offset; /* overlap is 'i' bytes */

		if (i < next->len) {
			/* Eat head of the next overlapped fragment
			 * and leave the loop. The next ones cannot overlap.
			 */
			if (!pskb_pull(next, i))
				goto err;
			FRAG_CB(next)->offset += i;
431
			qp->q.meat -= i;
L
Linus Torvalds 已提交
432 433 434 435 436 437
			if (next->ip_summed != CHECKSUM_UNNECESSARY)
				next->ip_summed = CHECKSUM_NONE;
			break;
		} else {
			struct sk_buff *free_it = next;

438
			/* Old fragment is completely overridden with
L
Linus Torvalds 已提交
439 440 441 442 443 444 445
			 * new one drop it.
			 */
			next = next->next;

			if (prev)
				prev->next = next;
			else
446
				qp->q.fragments = next;
L
Linus Torvalds 已提交
447

448
			qp->q.meat -= free_it->len;
449
			frag_kfree_skb(qp->q.net, free_it, NULL);
L
Linus Torvalds 已提交
450 451 452 453 454 455 456 457 458 459
		}
	}

	FRAG_CB(skb)->offset = offset;

	/* Insert this fragment in the chain of fragments. */
	skb->next = next;
	if (prev)
		prev->next = skb;
	else
460
		qp->q.fragments = skb;
L
Linus Torvalds 已提交
461

462 463 464 465 466
	dev = skb->dev;
	if (dev) {
		qp->iif = dev->ifindex;
		skb->dev = NULL;
	}
467 468
	qp->q.stamp = skb->tstamp;
	qp->q.meat += skb->len;
469
	atomic_add(skb->truesize, &qp->q.net->mem);
L
Linus Torvalds 已提交
470
	if (offset == 0)
471
		qp->q.last_in |= INET_FRAG_FIRST_IN;
L
Linus Torvalds 已提交
472

473 474
	if (qp->q.last_in == (INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN) &&
	    qp->q.meat == qp->q.len)
475 476
		return ip_frag_reasm(qp, prev, dev);

477
	write_lock(&ip4_frags.lock);
478
	list_move_tail(&qp->q.lru_list, &qp->q.net->lru_list);
479
	write_unlock(&ip4_frags.lock);
480
	return -EINPROGRESS;
L
Linus Torvalds 已提交
481 482 483

err:
	kfree_skb(skb);
484
	return err;
L
Linus Torvalds 已提交
485 486 487 488 489
}


/* Build a new IP datagram from all its fragments. */

490 491
static int ip_frag_reasm(struct ipq *qp, struct sk_buff *prev,
			 struct net_device *dev)
L
Linus Torvalds 已提交
492
{
493
	struct net *net = container_of(qp->q.net, struct net, ipv4.frags);
L
Linus Torvalds 已提交
494
	struct iphdr *iph;
495
	struct sk_buff *fp, *head = qp->q.fragments;
L
Linus Torvalds 已提交
496 497
	int len;
	int ihlen;
498
	int err;
L
Linus Torvalds 已提交
499 500 501

	ipq_kill(qp);

502 503 504 505 506 507 508 509 510 511
	/* Make the one we just received the head. */
	if (prev) {
		head = prev->next;
		fp = skb_clone(head, GFP_ATOMIC);
		if (!fp)
			goto out_nomem;

		fp->next = head->next;
		prev->next = fp;

512 513
		skb_morph(head, qp->q.fragments);
		head->next = qp->q.fragments->next;
514

515 516
		kfree_skb(qp->q.fragments);
		qp->q.fragments = head;
517 518
	}

519 520
	WARN_ON(head == NULL);
	WARN_ON(FRAG_CB(head)->offset != 0);
L
Linus Torvalds 已提交
521 522

	/* Allocate a new buffer for the datagram. */
523
	ihlen = ip_hdrlen(head);
524
	len = ihlen + qp->q.len;
L
Linus Torvalds 已提交
525

526
	err = -E2BIG;
S
Stephen Hemminger 已提交
527
	if (len > 65535)
L
Linus Torvalds 已提交
528 529 530 531 532 533 534 535 536
		goto out_oversize;

	/* Head of list must not be cloned. */
	if (skb_cloned(head) && pskb_expand_head(head, 0, 0, GFP_ATOMIC))
		goto out_nomem;

	/* If the first fragment is fragmented itself, we split
	 * it to two chunks: the first with data and paged part
	 * and the second, holding only fragments. */
537
	if (skb_has_frags(head)) {
L
Linus Torvalds 已提交
538 539 540 541 542 543 544 545
		struct sk_buff *clone;
		int i, plen = 0;

		if ((clone = alloc_skb(0, GFP_ATOMIC)) == NULL)
			goto out_nomem;
		clone->next = head->next;
		head->next = clone;
		skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list;
546
		skb_frag_list_init(head);
L
Linus Torvalds 已提交
547 548 549 550 551 552 553
		for (i=0; i<skb_shinfo(head)->nr_frags; i++)
			plen += skb_shinfo(head)->frags[i].size;
		clone->len = clone->data_len = head->data_len - plen;
		head->data_len -= clone->len;
		head->len -= clone->len;
		clone->csum = 0;
		clone->ip_summed = head->ip_summed;
554
		atomic_add(clone->truesize, &qp->q.net->mem);
L
Linus Torvalds 已提交
555 556 557
	}

	skb_shinfo(head)->frag_list = head->next;
558
	skb_push(head, head->data - skb_network_header(head));
559
	atomic_sub(head->truesize, &qp->q.net->mem);
L
Linus Torvalds 已提交
560 561 562 563 564 565

	for (fp=head->next; fp; fp = fp->next) {
		head->data_len += fp->len;
		head->len += fp->len;
		if (head->ip_summed != fp->ip_summed)
			head->ip_summed = CHECKSUM_NONE;
566
		else if (head->ip_summed == CHECKSUM_COMPLETE)
L
Linus Torvalds 已提交
567 568
			head->csum = csum_add(head->csum, fp->csum);
		head->truesize += fp->truesize;
569
		atomic_sub(fp->truesize, &qp->q.net->mem);
L
Linus Torvalds 已提交
570 571 572 573
	}

	head->next = NULL;
	head->dev = dev;
574
	head->tstamp = qp->q.stamp;
L
Linus Torvalds 已提交
575

576
	iph = ip_hdr(head);
L
Linus Torvalds 已提交
577 578
	iph->frag_off = 0;
	iph->tot_len = htons(len);
579
	IP_INC_STATS_BH(net, IPSTATS_MIB_REASMOKS);
580
	qp->q.fragments = NULL;
581
	return 0;
L
Linus Torvalds 已提交
582 583

out_nomem:
584
	LIMIT_NETDEBUG(KERN_ERR "IP: queue_glue: no memory for gluing "
585
			      "queue %p\n", qp);
586
	err = -ENOMEM;
L
Linus Torvalds 已提交
587 588 589
	goto out_fail;
out_oversize:
	if (net_ratelimit())
590 591
		printk(KERN_INFO "Oversized IP packet from %pI4.\n",
			&qp->saddr);
L
Linus Torvalds 已提交
592
out_fail:
593
	IP_INC_STATS_BH(net, IPSTATS_MIB_REASMFAILS);
594
	return err;
L
Linus Torvalds 已提交
595 596 597
}

/* Process an incoming IP datagram fragment. */
598
int ip_defrag(struct sk_buff *skb, u32 user)
L
Linus Torvalds 已提交
599 600
{
	struct ipq *qp;
601
	struct net *net;
602

E
Eric Dumazet 已提交
603
	net = skb->dev ? dev_net(skb->dev) : dev_net(skb_dst(skb)->dev);
604
	IP_INC_STATS_BH(net, IPSTATS_MIB_REASMREQDS);
L
Linus Torvalds 已提交
605 606

	/* Start by cleaning up the memory. */
607
	if (atomic_read(&net->ipv4.frags.mem) > net->ipv4.frags.high_thresh)
608
		ip_evictor(net);
L
Linus Torvalds 已提交
609 610

	/* Lookup (or create) queue header */
611
	if ((qp = ip_find(net, ip_hdr(skb), user)) != NULL) {
612
		int ret;
L
Linus Torvalds 已提交
613

614
		spin_lock(&qp->q.lock);
L
Linus Torvalds 已提交
615

616
		ret = ip_frag_queue(qp, skb);
L
Linus Torvalds 已提交
617

618
		spin_unlock(&qp->q.lock);
619
		ipq_put(qp);
620
		return ret;
L
Linus Torvalds 已提交
621 622
	}

623
	IP_INC_STATS_BH(net, IPSTATS_MIB_REASMFAILS);
L
Linus Torvalds 已提交
624
	kfree_skb(skb);
625
	return -ENOMEM;
L
Linus Torvalds 已提交
626 627
}

628 629 630
#ifdef CONFIG_SYSCTL
static int zero;

631
static struct ctl_table ip4_frags_ns_ctl_table[] = {
632 633
	{
		.procname	= "ipfrag_high_thresh",
634
		.data		= &init_net.ipv4.frags.high_thresh,
635 636
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
637
		.proc_handler	= proc_dointvec
638 639 640
	},
	{
		.procname	= "ipfrag_low_thresh",
641
		.data		= &init_net.ipv4.frags.low_thresh,
642 643
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
644
		.proc_handler	= proc_dointvec
645 646 647
	},
	{
		.procname	= "ipfrag_time",
648
		.data		= &init_net.ipv4.frags.timeout,
649 650
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
651
		.proc_handler	= proc_dointvec_jiffies,
652
	},
653 654 655 656
	{ }
};

static struct ctl_table ip4_frags_ctl_table[] = {
657 658
	{
		.procname	= "ipfrag_secret_interval",
659
		.data		= &ip4_frags.secret_interval,
660 661
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
662
		.proc_handler	= proc_dointvec_jiffies,
663 664 665 666 667 668
	},
	{
		.procname	= "ipfrag_max_dist",
		.data		= &sysctl_ipfrag_max_dist,
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
669
		.proc_handler	= proc_dointvec_minmax,
670 671 672 673 674
		.extra1		= &zero
	},
	{ }
};

675
static int __net_init ip4_frags_ns_ctl_register(struct net *net)
676
{
677
	struct ctl_table *table;
678 679
	struct ctl_table_header *hdr;

680
	table = ip4_frags_ns_ctl_table;
O
Octavian Purdila 已提交
681
	if (!net_eq(net, &init_net)) {
682
		table = kmemdup(table, sizeof(ip4_frags_ns_ctl_table), GFP_KERNEL);
683 684 685
		if (table == NULL)
			goto err_alloc;

686 687
		table[0].data = &net->ipv4.frags.high_thresh;
		table[1].data = &net->ipv4.frags.low_thresh;
688
		table[2].data = &net->ipv4.frags.timeout;
689 690 691 692 693 694 695 696 697 698
	}

	hdr = register_net_sysctl_table(net, net_ipv4_ctl_path, table);
	if (hdr == NULL)
		goto err_reg;

	net->ipv4.frags_hdr = hdr;
	return 0;

err_reg:
O
Octavian Purdila 已提交
699
	if (!net_eq(net, &init_net))
700 701 702 703 704
		kfree(table);
err_alloc:
	return -ENOMEM;
}

705
static void __net_exit ip4_frags_ns_ctl_unregister(struct net *net)
706 707 708 709 710 711
{
	struct ctl_table *table;

	table = net->ipv4.frags_hdr->ctl_table_arg;
	unregister_net_sysctl_table(net->ipv4.frags_hdr);
	kfree(table);
712
}
713 714 715 716 717

static void ip4_frags_ctl_register(void)
{
	register_net_sysctl_rotable(net_ipv4_ctl_path, ip4_frags_ctl_table);
}
718
#else
719
static inline int ip4_frags_ns_ctl_register(struct net *net)
720 721 722
{
	return 0;
}
723

724
static inline void ip4_frags_ns_ctl_unregister(struct net *net)
725 726
{
}
727 728 729 730

static inline void ip4_frags_ctl_register(void)
{
}
731 732
#endif

733
static int __net_init ipv4_frags_init_net(struct net *net)
734
{
735 736 737 738 739 740 741 742
	/*
	 * Fragment cache limits. We will commit 256K at one time. Should we
	 * cross that limit we will prune down to 192K. This should cope with
	 * even the most extreme cases without allowing an attacker to
	 * measurably harm machine performance.
	 */
	net->ipv4.frags.high_thresh = 256 * 1024;
	net->ipv4.frags.low_thresh = 192 * 1024;
743 744 745 746 747 748 749
	/*
	 * Important NOTE! Fragment queue must be destroyed before MSL expires.
	 * RFC791 is wrong proposing to prolongate timer each fragment arrival
	 * by TTL.
	 */
	net->ipv4.frags.timeout = IP_FRAG_TIME;

750 751
	inet_frags_init_net(&net->ipv4.frags);

752
	return ip4_frags_ns_ctl_register(net);
753 754
}

755
static void __net_exit ipv4_frags_exit_net(struct net *net)
756
{
757
	ip4_frags_ns_ctl_unregister(net);
758 759 760 761 762 763 764 765
	inet_frags_exit_net(&net->ipv4.frags, &ip4_frags);
}

static struct pernet_operations ip4_frags_ops = {
	.init = ipv4_frags_init_net,
	.exit = ipv4_frags_exit_net,
};

766
void __init ipfrag_init(void)
L
Linus Torvalds 已提交
767
{
768
	ip4_frags_ctl_register();
769
	register_pernet_subsys(&ip4_frags_ops);
770
	ip4_frags.hashfn = ip4_hashfn;
771
	ip4_frags.constructor = ip4_frag_init;
772 773 774
	ip4_frags.destructor = ip4_frag_free;
	ip4_frags.skb_free = NULL;
	ip4_frags.qsize = sizeof(struct ipq);
775
	ip4_frags.match = ip4_frag_match;
776
	ip4_frags.frag_expire = ip_expire;
777
	ip4_frags.secret_interval = 10 * 60 * HZ;
778
	inet_frags_init(&ip4_frags);
L
Linus Torvalds 已提交
779 780 781
}

EXPORT_SYMBOL(ip_defrag);