reassembly.c 18.8 KB
Newer Older
L
Linus Torvalds 已提交
1 2
/*
 *	IPv6 fragment reassembly
3
 *	Linux INET6 implementation
L
Linus Torvalds 已提交
4 5
 *
 *	Authors:
6
 *	Pedro Roque		<roque@di.fc.ul.pt>
L
Linus Torvalds 已提交
7 8 9 10 11 12 13 14 15
 *
 *	Based on: net/ipv4/ip_fragment.c
 *
 *	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.
 */

16 17
/*
 *	Fixes:
L
Linus Torvalds 已提交
18 19 20 21 22 23 24 25 26 27 28
 *	Andi Kleen	Make it work with multiple hosts.
 *			More RFC compliance.
 *
 *      Horst von Brand Add missing #include <linux/string.h>
 *	Alexey Kuznetsov	SMP races, threading, cleanup.
 *	Patrick McHardy		LRU queue of frag heads for evictor.
 *	Mitsuru KANDA @USAGI	Register inet6_protocol{}.
 *	David Stevens and
 *	YOSHIFUJI,H. @USAGI	Always remove fragment header to
 *				calculate ICV correctly.
 */
29 30 31

#define pr_fmt(fmt) "IPv6: " fmt

L
Linus Torvalds 已提交
32 33 34 35 36 37 38 39 40 41 42 43 44 45
#include <linux/errno.h>
#include <linux/types.h>
#include <linux/string.h>
#include <linux/socket.h>
#include <linux/sockios.h>
#include <linux/jiffies.h>
#include <linux/net.h>
#include <linux/list.h>
#include <linux/netdevice.h>
#include <linux/in6.h>
#include <linux/ipv6.h>
#include <linux/icmpv6.h>
#include <linux/random.h>
#include <linux/jhash.h>
46
#include <linux/skbuff.h>
47
#include <linux/slab.h>
48
#include <linux/export.h>
L
Linus Torvalds 已提交
49 50 51 52 53

#include <net/sock.h>
#include <net/snmp.h>

#include <net/ipv6.h>
54
#include <net/ip6_route.h>
L
Linus Torvalds 已提交
55 56 57 58 59
#include <net/protocol.h>
#include <net/transp_v6.h>
#include <net/rawv6.h>
#include <net/ndisc.h>
#include <net/addrconf.h>
60
#include <net/inet_frag.h>
61
#include <net/inet_ecn.h>
L
Linus Torvalds 已提交
62 63 64 65 66 67 68 69 70

struct ip6frag_skb_cb
{
	struct inet6_skb_parm	h;
	int			offset;
};

#define FRAG6_CB(skb)	((struct ip6frag_skb_cb*)((skb)->cb))

71 72 73 74
static inline u8 ip6_frag_ecn(const struct ipv6hdr *ipv6h)
{
	return 1 << (ipv6_get_dsfield(ipv6h) & INET_ECN_MASK);
}
L
Linus Torvalds 已提交
75

76
static struct inet_frags ip6_frags;
L
Linus Torvalds 已提交
77

78 79 80
static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
			  struct net_device *dev);

81 82 83 84
/*
 * callers should be careful not to use the hash value outside the ipfrag_lock
 * as doing so could race with ipfrag_hash_rnd being recalculated.
 */
85 86
unsigned int inet6_hash_frag(__be32 id, const struct in6_addr *saddr,
			     const struct in6_addr *daddr, u32 rnd)
L
Linus Torvalds 已提交
87
{
88 89
	u32 c;

E
Eric Dumazet 已提交
90 91
	c = jhash_3words(ipv6_addr_hash(saddr), ipv6_addr_hash(daddr),
			 (__force u32)id, rnd);
L
Linus Torvalds 已提交
92

93
	return c & (INETFRAGS_HASHSZ - 1);
L
Linus Torvalds 已提交
94
}
95
EXPORT_SYMBOL_GPL(inet6_hash_frag);
L
Linus Torvalds 已提交
96

97
static unsigned int ip6_hashfn(struct inet_frag_queue *q)
L
Linus Torvalds 已提交
98
{
99
	struct frag_queue *fq;
L
Linus Torvalds 已提交
100

101
	fq = container_of(q, struct frag_queue, q);
102
	return inet6_hash_frag(fq->id, &fq->saddr, &fq->daddr, ip6_frags.rnd);
L
Linus Torvalds 已提交
103 104
}

105
bool ip6_frag_match(struct inet_frag_queue *q, void *a)
106 107 108 109 110
{
	struct frag_queue *fq;
	struct ip6_create_arg *arg = a;

	fq = container_of(q, struct frag_queue, q);
111 112 113 114
	return	fq->id == arg->id &&
		fq->user == arg->user &&
		ipv6_addr_equal(&fq->saddr, arg->src) &&
		ipv6_addr_equal(&fq->daddr, arg->dst);
115 116 117
}
EXPORT_SYMBOL(ip6_frag_match);

118
void ip6_frag_init(struct inet_frag_queue *q, void *a)
L
Linus Torvalds 已提交
119
{
120 121 122 123
	struct frag_queue *fq = container_of(q, struct frag_queue, q);
	struct ip6_create_arg *arg = a;

	fq->id = arg->id;
124
	fq->user = arg->user;
A
Alexey Dobriyan 已提交
125 126
	fq->saddr = *arg->src;
	fq->daddr = *arg->dst;
127
	fq->ecn = arg->ecn;
L
Linus Torvalds 已提交
128
}
129
EXPORT_SYMBOL(ip6_frag_init);
L
Linus Torvalds 已提交
130

131 132
void ip6_expire_frag_queue(struct net *net, struct frag_queue *fq,
			   struct inet_frags *frags)
L
Linus Torvalds 已提交
133
{
134
	struct net_device *dev = NULL;
135

136
	spin_lock(&fq->q.lock);
L
Linus Torvalds 已提交
137

138
	if (fq->q.last_in & INET_FRAG_COMPLETE)
L
Linus Torvalds 已提交
139 140
		goto out;

141
	inet_frag_kill(&fq->q, frags);
L
Linus Torvalds 已提交
142

E
Eric Dumazet 已提交
143 144
	rcu_read_lock();
	dev = dev_get_by_index_rcu(net, fq->iif);
145
	if (!dev)
E
Eric Dumazet 已提交
146
		goto out_rcu_unlock;
147

148 149
	IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMTIMEOUT);
	IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
L
Linus Torvalds 已提交
150

151
	/* Don't send error if the first segment did not arrive. */
152
	if (!(fq->q.last_in & INET_FRAG_FIRST_IN) || !fq->q.fragments)
E
Eric Dumazet 已提交
153
		goto out_rcu_unlock;
154 155 156 157 158 159

	/*
	   But use as source device on which LAST ARRIVED
	   segment was received. And do not use fq->dev
	   pointer directly, device might already disappeared.
	 */
160
	fq->q.fragments->dev = dev;
161
	icmpv6_send(fq->q.fragments, ICMPV6_TIME_EXCEED, ICMPV6_EXC_FRAGTIME, 0);
E
Eric Dumazet 已提交
162 163
out_rcu_unlock:
	rcu_read_unlock();
L
Linus Torvalds 已提交
164
out:
165
	spin_unlock(&fq->q.lock);
166 167 168 169 170 171 172 173 174 175 176 177 178
	inet_frag_put(&fq->q, frags);
}
EXPORT_SYMBOL(ip6_expire_frag_queue);

static void ip6_frag_expire(unsigned long data)
{
	struct frag_queue *fq;
	struct net *net;

	fq = container_of((struct inet_frag_queue *)data, struct frag_queue, q);
	net = container_of(fq->q.net, struct net, ipv6.frags);

	ip6_expire_frag_queue(net, fq, &ip6_frags);
L
Linus Torvalds 已提交
179 180
}

181
static __inline__ struct frag_queue *
182 183
fq_find(struct net *net, __be32 id, const struct in6_addr *src,
	const struct in6_addr *dst, u8 ecn)
L
Linus Torvalds 已提交
184
{
185 186
	struct inet_frag_queue *q;
	struct ip6_create_arg arg;
187
	unsigned int hash;
L
Linus Torvalds 已提交
188

189
	arg.id = id;
190
	arg.user = IP6_DEFRAG_LOCAL_DELIVER;
191 192
	arg.src = src;
	arg.dst = dst;
193
	arg.ecn = ecn;
194 195

	read_lock(&ip6_frags.lock);
196
	hash = inet6_hash_frag(id, src, dst, ip6_frags.rnd);
L
Linus Torvalds 已提交
197

198
	q = inet_frag_find(&net->ipv6.frags, &ip6_frags, &arg, hash);
199 200
	if (IS_ERR_OR_NULL(q)) {
		inet_frag_maybe_warn_overflow(q, pr_fmt());
201
		return NULL;
202
	}
203
	return container_of(q, struct frag_queue, q);
L
Linus Torvalds 已提交
204 205
}

206
static int ip6_frag_queue(struct frag_queue *fq, struct sk_buff *skb,
L
Linus Torvalds 已提交
207 208 209
			   struct frag_hdr *fhdr, int nhoff)
{
	struct sk_buff *prev, *next;
210
	struct net_device *dev;
L
Linus Torvalds 已提交
211
	int offset, end;
E
Eric Dumazet 已提交
212
	struct net *net = dev_net(skb_dst(skb)->dev);
213
	u8 ecn;
L
Linus Torvalds 已提交
214

215
	if (fq->q.last_in & INET_FRAG_COMPLETE)
L
Linus Torvalds 已提交
216 217 218
		goto err;

	offset = ntohs(fhdr->frag_off) & ~0x7;
219 220
	end = offset + (ntohs(ipv6_hdr(skb)->payload_len) -
			((u8 *)(fhdr + 1) - (u8 *)(ipv6_hdr(skb) + 1)));
L
Linus Torvalds 已提交
221 222

	if ((unsigned int)end > IPV6_MAXPLEN) {
E
Eric Dumazet 已提交
223
		IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
224
				 IPSTATS_MIB_INHDRERRORS);
225 226 227
		icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
				  ((u8 *)&fhdr->frag_off -
				   skb_network_header(skb)));
228
		return -1;
L
Linus Torvalds 已提交
229 230
	}

231 232
	ecn = ip6_frag_ecn(ipv6_hdr(skb));

233 234
	if (skb->ip_summed == CHECKSUM_COMPLETE) {
		const unsigned char *nh = skb_network_header(skb);
235
		skb->csum = csum_sub(skb->csum,
236 237 238
				     csum_partial(nh, (u8 *)(fhdr + 1) - nh,
						  0));
	}
L
Linus Torvalds 已提交
239 240 241 242 243 244

	/* Is this the final fragment? */
	if (!(fhdr->frag_off & htons(IP6_MF))) {
		/* If we already have some bits beyond end
		 * or have different end, the segment is corrupted.
		 */
245
		if (end < fq->q.len ||
246
		    ((fq->q.last_in & INET_FRAG_LAST_IN) && end != fq->q.len))
L
Linus Torvalds 已提交
247
			goto err;
248
		fq->q.last_in |= INET_FRAG_LAST_IN;
249
		fq->q.len = end;
L
Linus Torvalds 已提交
250 251 252 253 254 255 256 257
	} else {
		/* Check if the fragment is rounded to 8 bytes.
		 * Required by the RFC.
		 */
		if (end & 0x7) {
			/* RFC2460 says always send parameter problem in
			 * this case. -DaveM
			 */
E
Eric Dumazet 已提交
258
			IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
259
					 IPSTATS_MIB_INHDRERRORS);
260
			icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
L
Linus Torvalds 已提交
261
					  offsetof(struct ipv6hdr, payload_len));
262
			return -1;
L
Linus Torvalds 已提交
263
		}
264
		if (end > fq->q.len) {
L
Linus Torvalds 已提交
265
			/* Some bits beyond end -> corruption. */
266
			if (fq->q.last_in & INET_FRAG_LAST_IN)
L
Linus Torvalds 已提交
267
				goto err;
268
			fq->q.len = end;
L
Linus Torvalds 已提交
269 270 271 272 273 274 275 276 277
		}
	}

	if (end == offset)
		goto err;

	/* Point into the IP datagram 'data' part. */
	if (!pskb_pull(skb, (u8 *) (fhdr + 1) - skb->data))
		goto err;
278

279 280
	if (pskb_trim_rcsum(skb, end - offset))
		goto err;
L
Linus Torvalds 已提交
281 282 283 284 285

	/* 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?
	 */
286 287 288 289 290
	prev = fq->q.fragments_tail;
	if (!prev || FRAG6_CB(prev)->offset < offset) {
		next = NULL;
		goto found;
	}
L
Linus Torvalds 已提交
291
	prev = NULL;
292
	for(next = fq->q.fragments; next != NULL; next = next->next) {
L
Linus Torvalds 已提交
293 294 295 296 297
		if (FRAG6_CB(next)->offset >= offset)
			break;	/* bingo! */
		prev = next;
	}

298
found:
E
Eric Dumazet 已提交
299 300
	/* RFC5722, Section 4, amended by Errata ID : 3089
	 *                          When reassembling an IPv6 datagram, if
301 302
	 *   one or more its constituent fragments is determined to be an
	 *   overlapping fragment, the entire datagram (and any constituent
E
Eric Dumazet 已提交
303
	 *   fragments) MUST be silently discarded.
L
Linus Torvalds 已提交
304 305
	 */

306 307
	/* Check for overlap with preceding fragment. */
	if (prev &&
S
Shan Wei 已提交
308
	    (FRAG6_CB(prev)->offset + prev->len) > offset)
309
		goto discard_fq;
L
Linus Torvalds 已提交
310

311 312 313
	/* Look for overlap with succeeding segment. */
	if (next && FRAG6_CB(next)->offset < end)
		goto discard_fq;
L
Linus Torvalds 已提交
314 315 316 317 318

	FRAG6_CB(skb)->offset = offset;

	/* Insert this fragment in the chain of fragments. */
	skb->next = next;
319 320
	if (!next)
		fq->q.fragments_tail = skb;
L
Linus Torvalds 已提交
321 322 323
	if (prev)
		prev->next = skb;
	else
324
		fq->q.fragments = skb;
L
Linus Torvalds 已提交
325

326 327 328 329 330
	dev = skb->dev;
	if (dev) {
		fq->iif = dev->ifindex;
		skb->dev = NULL;
	}
331 332
	fq->q.stamp = skb->tstamp;
	fq->q.meat += skb->len;
333
	fq->ecn |= ecn;
334
	add_frag_mem_limit(&fq->q, skb->truesize);
L
Linus Torvalds 已提交
335 336 337 338 339 340

	/* The first fragment.
	 * nhoffset is obtained from the first fragment, of course.
	 */
	if (offset == 0) {
		fq->nhoffset = nhoff;
341
		fq->q.last_in |= INET_FRAG_FIRST_IN;
L
Linus Torvalds 已提交
342
	}
343

344
	if (fq->q.last_in == (INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN) &&
345 346 347 348 349 350 351 352 353
	    fq->q.meat == fq->q.len) {
		int res;
		unsigned long orefdst = skb->_skb_refdst;

		skb->_skb_refdst = 0UL;
		res = ip6_frag_reasm(fq, prev, dev);
		skb->_skb_refdst = orefdst;
		return res;
	}
354

355
	skb_dst_drop(skb);
356
	inet_frag_lru_move(&fq->q);
357
	return -1;
L
Linus Torvalds 已提交
358

359
discard_fq:
360
	inet_frag_kill(&fq->q, &ip6_frags);
L
Linus Torvalds 已提交
361
err:
E
Eric Dumazet 已提交
362
	IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
363
		      IPSTATS_MIB_REASMFAILS);
L
Linus Torvalds 已提交
364
	kfree_skb(skb);
365
	return -1;
L
Linus Torvalds 已提交
366 367 368 369 370 371 372 373 374 375 376
}

/*
 *	Check if this packet is complete.
 *	Returns NULL on failure by any reason, and pointer
 *	to current nexthdr field in reassembled frame.
 *
 *	It is called with locked fq, and caller must check that
 *	queue is eligible for reassembly i.e. it is not COMPLETE,
 *	the last and the first frames arrived and all the bits are here.
 */
377
static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
L
Linus Torvalds 已提交
378 379
			  struct net_device *dev)
{
380
	struct net *net = container_of(fq->q.net, struct net, ipv6.frags);
381
	struct sk_buff *fp, *head = fq->q.fragments;
L
Linus Torvalds 已提交
382 383
	int    payload_len;
	unsigned int nhoff;
384
	int sum_truesize;
385
	u8 ecn;
L
Linus Torvalds 已提交
386

387
	inet_frag_kill(&fq->q, &ip6_frags);
L
Linus Torvalds 已提交
388

389 390 391 392
	ecn = ip_frag_ecn_table[fq->ecn];
	if (unlikely(ecn == 0xff))
		goto out_fail;

393 394 395 396 397 398 399 400 401
	/* Make the one we just received the head. */
	if (prev) {
		head = prev->next;
		fp = skb_clone(head, GFP_ATOMIC);

		if (!fp)
			goto out_oom;

		fp->next = head->next;
402 403
		if (!fp->next)
			fq->q.fragments_tail = fp;
404 405
		prev->next = fp;

406 407
		skb_morph(head, fq->q.fragments);
		head->next = fq->q.fragments->next;
408

409
		consume_skb(fq->q.fragments);
410
		fq->q.fragments = head;
411 412
	}

413 414
	WARN_ON(head == NULL);
	WARN_ON(FRAG6_CB(head)->offset != 0);
L
Linus Torvalds 已提交
415 416

	/* Unfragmented part is taken from the first segment. */
417
	payload_len = ((head->data - skb_network_header(head)) -
418
		       sizeof(struct ipv6hdr) + fq->q.len -
419
		       sizeof(struct frag_hdr));
L
Linus Torvalds 已提交
420 421 422 423
	if (payload_len > IPV6_MAXPLEN)
		goto out_oversize;

	/* Head of list must not be cloned. */
424
	if (skb_unclone(head, GFP_ATOMIC))
L
Linus Torvalds 已提交
425 426 427 428 429
		goto out_oom;

	/* 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. */
430
	if (skb_has_frag_list(head)) {
L
Linus Torvalds 已提交
431 432 433 434 435 436 437 438
		struct sk_buff *clone;
		int i, plen = 0;

		if ((clone = alloc_skb(0, GFP_ATOMIC)) == NULL)
			goto out_oom;
		clone->next = head->next;
		head->next = clone;
		skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list;
439
		skb_frag_list_init(head);
E
Eric Dumazet 已提交
440 441
		for (i = 0; i < skb_shinfo(head)->nr_frags; i++)
			plen += skb_frag_size(&skb_shinfo(head)->frags[i]);
L
Linus Torvalds 已提交
442 443 444 445 446
		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;
447
		add_frag_mem_limit(&fq->q, clone->truesize);
L
Linus Torvalds 已提交
448 449 450 451 452
	}

	/* We have to remove fragment header from datagram and to relocate
	 * header in order to calculate ICV correctly. */
	nhoff = fq->nhoffset;
453
	skb_network_header(head)[nhoff] = skb_transport_header(head)[0];
454
	memmove(head->head + sizeof(struct frag_hdr), head->head,
L
Linus Torvalds 已提交
455
		(head->data - head->head) - sizeof(struct frag_hdr));
456 457
	head->mac_header += sizeof(struct frag_hdr);
	head->network_header += sizeof(struct frag_hdr);
L
Linus Torvalds 已提交
458

459
	skb_reset_transport_header(head);
460
	skb_push(head, head->data - skb_network_header(head));
L
Linus Torvalds 已提交
461

462 463 464 465 466 467 468
	sum_truesize = head->truesize;
	for (fp = head->next; fp;) {
		bool headstolen;
		int delta;
		struct sk_buff *next = fp->next;

		sum_truesize += fp->truesize;
L
Linus Torvalds 已提交
469 470
		if (head->ip_summed != fp->ip_summed)
			head->ip_summed = CHECKSUM_NONE;
471
		else if (head->ip_summed == CHECKSUM_COMPLETE)
L
Linus Torvalds 已提交
472
			head->csum = csum_add(head->csum, fp->csum);
473 474 475 476 477 478 479 480 481 482 483

		if (skb_try_coalesce(head, fp, &headstolen, &delta)) {
			kfree_skb_partial(fp, headstolen);
		} else {
			if (!skb_shinfo(head)->frag_list)
				skb_shinfo(head)->frag_list = fp;
			head->data_len += fp->len;
			head->len += fp->len;
			head->truesize += fp->truesize;
		}
		fp = next;
L
Linus Torvalds 已提交
484
	}
485
	sub_frag_mem_limit(&fq->q, sum_truesize);
L
Linus Torvalds 已提交
486 487 488

	head->next = NULL;
	head->dev = dev;
489
	head->tstamp = fq->q.stamp;
490
	ipv6_hdr(head)->payload_len = htons(payload_len);
491
	ipv6_change_dsfield(ipv6_hdr(head), 0xff, ecn);
492
	IP6CB(head)->nhoff = nhoff;
L
Linus Torvalds 已提交
493 494

	/* Yes, and fold redundant checksum back. 8) */
495
	if (head->ip_summed == CHECKSUM_COMPLETE)
496
		head->csum = csum_partial(skb_network_header(head),
497
					  skb_network_header_len(head),
498
					  head->csum);
L
Linus Torvalds 已提交
499

500
	rcu_read_lock();
501
	IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMOKS);
502
	rcu_read_unlock();
503
	fq->q.fragments = NULL;
504
	fq->q.fragments_tail = NULL;
L
Linus Torvalds 已提交
505 506 507
	return 1;

out_oversize:
508
	net_dbg_ratelimited("ip6_frag_reasm: payload len = %d\n", payload_len);
L
Linus Torvalds 已提交
509 510
	goto out_fail;
out_oom:
511
	net_dbg_ratelimited("ip6_frag_reasm: no memory for reassembly\n");
L
Linus Torvalds 已提交
512
out_fail:
513
	rcu_read_lock();
514
	IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
515
	rcu_read_unlock();
L
Linus Torvalds 已提交
516 517 518
	return -1;
}

519
static int ipv6_frag_rcv(struct sk_buff *skb)
L
Linus Torvalds 已提交
520 521 522
{
	struct frag_hdr *fhdr;
	struct frag_queue *fq;
523
	const struct ipv6hdr *hdr = ipv6_hdr(skb);
E
Eric Dumazet 已提交
524
	struct net *net = dev_net(skb_dst(skb)->dev);
525
	int evicted;
L
Linus Torvalds 已提交
526

E
Eric Dumazet 已提交
527
	IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMREQDS);
L
Linus Torvalds 已提交
528 529

	/* Jumbo payload inhibits frag. header */
530 531 532
	if (hdr->payload_len==0)
		goto fail_hdr;

533
	if (!pskb_may_pull(skb, (skb_transport_offset(skb) +
534 535
				 sizeof(struct frag_hdr))))
		goto fail_hdr;
L
Linus Torvalds 已提交
536

537
	hdr = ipv6_hdr(skb);
538
	fhdr = (struct frag_hdr *)skb_transport_header(skb);
L
Linus Torvalds 已提交
539 540 541

	if (!(fhdr->frag_off & htons(0xFFF9))) {
		/* It is not a fragmented frame */
542
		skb->transport_header += sizeof(struct frag_hdr);
543
		IP6_INC_STATS_BH(net,
E
Eric Dumazet 已提交
544
				 ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMOKS);
L
Linus Torvalds 已提交
545

546
		IP6CB(skb)->nhoff = (u8 *)fhdr - skb_network_header(skb);
L
Linus Torvalds 已提交
547 548 549
		return 1;
	}

550 551 552 553
	evicted = inet_frag_evictor(&net->ipv6.frags, &ip6_frags, false);
	if (evicted)
		IP6_ADD_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
				 IPSTATS_MIB_REASMFAILS, evicted);
L
Linus Torvalds 已提交
554

555 556
	fq = fq_find(net, fhdr->identification, &hdr->saddr, &hdr->daddr,
		     ip6_frag_ecn(hdr));
557
	if (fq != NULL) {
558
		int ret;
L
Linus Torvalds 已提交
559

560
		spin_lock(&fq->q.lock);
L
Linus Torvalds 已提交
561

562
		ret = ip6_frag_queue(fq, skb, fhdr, IP6CB(skb)->nhoff);
L
Linus Torvalds 已提交
563

564
		spin_unlock(&fq->q.lock);
565
		inet_frag_put(&fq->q, &ip6_frags);
L
Linus Torvalds 已提交
566 567 568
		return ret;
	}

E
Eric Dumazet 已提交
569
	IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMFAILS);
L
Linus Torvalds 已提交
570 571
	kfree_skb(skb);
	return -1;
572 573

fail_hdr:
E
Eric Dumazet 已提交
574
	IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INHDRERRORS);
575 576
	icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, skb_network_header_len(skb));
	return -1;
L
Linus Torvalds 已提交
577 578
}

579
static const struct inet6_protocol frag_protocol =
L
Linus Torvalds 已提交
580 581 582 583 584
{
	.handler	=	ipv6_frag_rcv,
	.flags		=	INET6_PROTO_NOPOLICY,
};

585
#ifdef CONFIG_SYSCTL
586
static struct ctl_table ip6_frags_ns_ctl_table[] = {
587 588
	{
		.procname	= "ip6frag_high_thresh",
589
		.data		= &init_net.ipv6.frags.high_thresh,
590 591
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
592
		.proc_handler	= proc_dointvec
593 594 595
	},
	{
		.procname	= "ip6frag_low_thresh",
596
		.data		= &init_net.ipv6.frags.low_thresh,
597 598
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
599
		.proc_handler	= proc_dointvec
600 601 602
	},
	{
		.procname	= "ip6frag_time",
603
		.data		= &init_net.ipv6.frags.timeout,
604 605
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
606
		.proc_handler	= proc_dointvec_jiffies,
607
	},
608 609 610 611
	{ }
};

static struct ctl_table ip6_frags_ctl_table[] = {
612 613
	{
		.procname	= "ip6frag_secret_interval",
614
		.data		= &ip6_frags.secret_interval,
615 616
		.maxlen		= sizeof(int),
		.mode		= 0644,
A
Alexey Dobriyan 已提交
617
		.proc_handler	= proc_dointvec_jiffies,
618 619 620 621
	},
	{ }
};

622
static int __net_init ip6_frags_ns_sysctl_register(struct net *net)
623
{
624
	struct ctl_table *table;
625 626
	struct ctl_table_header *hdr;

627
	table = ip6_frags_ns_ctl_table;
O
Octavian Purdila 已提交
628
	if (!net_eq(net, &init_net)) {
629
		table = kmemdup(table, sizeof(ip6_frags_ns_ctl_table), GFP_KERNEL);
630 631 632
		if (table == NULL)
			goto err_alloc;

633 634
		table[0].data = &net->ipv6.frags.high_thresh;
		table[1].data = &net->ipv6.frags.low_thresh;
635
		table[2].data = &net->ipv6.frags.timeout;
636 637 638 639

		/* Don't export sysctls to unprivileged users */
		if (net->user_ns != &init_user_ns)
			table[0].procname = NULL;
640 641
	}

642
	hdr = register_net_sysctl(net, "net/ipv6", table);
643 644 645 646 647 648 649
	if (hdr == NULL)
		goto err_reg;

	net->ipv6.sysctl.frags_hdr = hdr;
	return 0;

err_reg:
O
Octavian Purdila 已提交
650
	if (!net_eq(net, &init_net))
651 652 653 654 655
		kfree(table);
err_alloc:
	return -ENOMEM;
}

656
static void __net_exit ip6_frags_ns_sysctl_unregister(struct net *net)
657 658 659 660 661
{
	struct ctl_table *table;

	table = net->ipv6.sysctl.frags_hdr->ctl_table_arg;
	unregister_net_sysctl_table(net->ipv6.sysctl.frags_hdr);
662 663
	if (!net_eq(net, &init_net))
		kfree(table);
664
}
665 666 667 668 669

static struct ctl_table_header *ip6_ctl_header;

static int ip6_frags_sysctl_register(void)
{
670
	ip6_ctl_header = register_net_sysctl(&init_net, "net/ipv6",
671 672 673 674 675 676 677 678
			ip6_frags_ctl_table);
	return ip6_ctl_header == NULL ? -ENOMEM : 0;
}

static void ip6_frags_sysctl_unregister(void)
{
	unregister_net_sysctl_table(ip6_ctl_header);
}
679
#else
680
static inline int ip6_frags_ns_sysctl_register(struct net *net)
681
{
682 683
	return 0;
}
684

685
static inline void ip6_frags_ns_sysctl_unregister(struct net *net)
686 687
{
}
688 689 690 691 692 693 694 695 696

static inline int ip6_frags_sysctl_register(void)
{
	return 0;
}

static inline void ip6_frags_sysctl_unregister(void)
{
}
697
#endif
D
Daniel Lezcano 已提交
698

699
static int __net_init ipv6_frags_init_net(struct net *net)
700
{
701 702
	net->ipv6.frags.high_thresh = IPV6_FRAG_HIGH_THRESH;
	net->ipv6.frags.low_thresh = IPV6_FRAG_LOW_THRESH;
703
	net->ipv6.frags.timeout = IPV6_FRAG_TIMEOUT;
704

705 706
	inet_frags_init_net(&net->ipv6.frags);

707
	return ip6_frags_ns_sysctl_register(net);
708 709
}

710
static void __net_exit ipv6_frags_exit_net(struct net *net)
711
{
712
	ip6_frags_ns_sysctl_unregister(net);
713 714 715 716 717 718 719 720
	inet_frags_exit_net(&net->ipv6.frags, &ip6_frags);
}

static struct pernet_operations ip6_frags_ops = {
	.init = ipv6_frags_init_net,
	.exit = ipv6_frags_exit_net,
};

721
int __init ipv6_frag_init(void)
L
Linus Torvalds 已提交
722
{
723
	int ret;
L
Linus Torvalds 已提交
724

725 726 727
	ret = inet6_add_protocol(&frag_protocol, IPPROTO_FRAGMENT);
	if (ret)
		goto out;
728

729 730 731 732
	ret = ip6_frags_sysctl_register();
	if (ret)
		goto err_sysctl;

733 734 735
	ret = register_pernet_subsys(&ip6_frags_ops);
	if (ret)
		goto err_pernet;
736

737
	ip6_frags.hashfn = ip6_hashfn;
738
	ip6_frags.constructor = ip6_frag_init;
739
	ip6_frags.destructor = NULL;
740 741
	ip6_frags.skb_free = NULL;
	ip6_frags.qsize = sizeof(struct frag_queue);
742
	ip6_frags.match = ip6_frag_match;
743
	ip6_frags.frag_expire = ip6_frag_expire;
744
	ip6_frags.secret_interval = 10 * 60 * HZ;
745
	inet_frags_init(&ip6_frags);
746 747
out:
	return ret;
748 749

err_pernet:
750 751
	ip6_frags_sysctl_unregister();
err_sysctl:
752 753
	inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);
	goto out;
754 755 756 757 758
}

void ipv6_frag_exit(void)
{
	inet_frags_fini(&ip6_frags);
759
	ip6_frags_sysctl_unregister();
760
	unregister_pernet_subsys(&ip6_frags_ops);
761
	inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);
L
Linus Torvalds 已提交
762
}