reassembly.c 19.2 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 16 17
 *
 *	$Id: reassembly.c,v 1.26 2001/03/07 22:00:57 davem Exp $
 *
 *	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.
 */

18 19
/*
 *	Fixes:
L
Linus Torvalds 已提交
20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44
 *	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.
 */
#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>
45
#include <linux/skbuff.h>
L
Linus Torvalds 已提交
46 47 48 49 50

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

#include <net/ipv6.h>
51
#include <net/ip6_route.h>
L
Linus Torvalds 已提交
52 53 54 55 56
#include <net/protocol.h>
#include <net/transp_v6.h>
#include <net/rawv6.h>
#include <net/ndisc.h>
#include <net/addrconf.h>
57
#include <net/inet_frag.h>
L
Linus Torvalds 已提交
58

59 60
int sysctl_ip6frag_high_thresh __read_mostly = 256*1024;
int sysctl_ip6frag_low_thresh __read_mostly = 192*1024;
L
Linus Torvalds 已提交
61

62
int sysctl_ip6frag_time __read_mostly = IPV6_FRAG_TIMEOUT;
L
Linus Torvalds 已提交
63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78

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

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


/*
 *	Equivalent of ipv4 struct ipq
 */

struct frag_queue
{
79
	struct inet_frag_queue	q;
L
Linus Torvalds 已提交
80

A
Al Viro 已提交
81
	__be32			id;		/* fragment id		*/
L
Linus Torvalds 已提交
82 83 84 85 86 87 88 89
	struct in6_addr		saddr;
	struct in6_addr		daddr;

	int			iif;
	unsigned int		csum;
	__u16			nhoffset;
};

90
static struct inet_frags ip6_frags;
L
Linus Torvalds 已提交
91

92 93 94 95
int ip6_frag_nqueues(void)
{
	return ip6_frags.nqueues;
}
L
Linus Torvalds 已提交
96

97 98 99 100
int ip6_frag_mem(void)
{
	return atomic_read(&ip6_frags.mem);
}
L
Linus Torvalds 已提交
101

102 103 104
static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
			  struct net_device *dev);

L
Linus Torvalds 已提交
105 106
static __inline__ void __fq_unlink(struct frag_queue *fq)
{
107 108
	hlist_del(&fq->q.list);
	list_del(&fq->q.lru_list);
109
	ip6_frags.nqueues--;
L
Linus Torvalds 已提交
110 111 112 113
}

static __inline__ void fq_unlink(struct frag_queue *fq)
{
114
	write_lock(&ip6_frags.lock);
L
Linus Torvalds 已提交
115
	__fq_unlink(fq);
116
	write_unlock(&ip6_frags.lock);
L
Linus Torvalds 已提交
117 118
}

119 120 121 122
/*
 * 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.
 */
A
Al Viro 已提交
123
static unsigned int ip6qhashfn(__be32 id, struct in6_addr *saddr,
L
Linus Torvalds 已提交
124 125 126 127
			       struct in6_addr *daddr)
{
	u32 a, b, c;

A
Al Viro 已提交
128 129 130
	a = (__force u32)saddr->s6_addr32[0];
	b = (__force u32)saddr->s6_addr32[1];
	c = (__force u32)saddr->s6_addr32[2];
L
Linus Torvalds 已提交
131 132 133

	a += JHASH_GOLDEN_RATIO;
	b += JHASH_GOLDEN_RATIO;
134
	c += ip6_frags.rnd;
L
Linus Torvalds 已提交
135 136
	__jhash_mix(a, b, c);

A
Al Viro 已提交
137 138 139
	a += (__force u32)saddr->s6_addr32[3];
	b += (__force u32)daddr->s6_addr32[0];
	c += (__force u32)daddr->s6_addr32[1];
L
Linus Torvalds 已提交
140 141
	__jhash_mix(a, b, c);

A
Al Viro 已提交
142 143 144
	a += (__force u32)daddr->s6_addr32[2];
	b += (__force u32)daddr->s6_addr32[3];
	c += (__force u32)id;
L
Linus Torvalds 已提交
145 146
	__jhash_mix(a, b, c);

147
	return c & (INETFRAGS_HASHSZ - 1);
L
Linus Torvalds 已提交
148 149
}

150
int sysctl_ip6frag_secret_interval __read_mostly = 10 * 60 * HZ;
L
Linus Torvalds 已提交
151 152 153 154 155 156

static void ip6_frag_secret_rebuild(unsigned long dummy)
{
	unsigned long now = jiffies;
	int i;

157 158 159
	write_lock(&ip6_frags.lock);
	get_random_bytes(&ip6_frags.rnd, sizeof(u32));
	for (i = 0; i < INETFRAGS_HASHSZ; i++) {
L
Linus Torvalds 已提交
160
		struct frag_queue *q;
161
		struct hlist_node *p, *n;
L
Linus Torvalds 已提交
162

163
		hlist_for_each_entry_safe(q, p, n, &ip6_frags.hash[i], q.list) {
L
Linus Torvalds 已提交
164 165 166 167 168
			unsigned int hval = ip6qhashfn(q->id,
						       &q->saddr,
						       &q->daddr);

			if (hval != i) {
169
				hlist_del(&q->q.list);
L
Linus Torvalds 已提交
170 171

				/* Relink to new hash chain. */
172
				hlist_add_head(&q->q.list,
173
					       &ip6_frags.hash[hval]);
L
Linus Torvalds 已提交
174

175
			}
L
Linus Torvalds 已提交
176 177
		}
	}
178
	write_unlock(&ip6_frags.lock);
L
Linus Torvalds 已提交
179

180
	mod_timer(&ip6_frags.secret_timer, now + sysctl_ip6frag_secret_interval);
L
Linus Torvalds 已提交
181 182 183 184 185 186 187
}

/* Memory Tracking Functions. */
static inline void frag_kfree_skb(struct sk_buff *skb, int *work)
{
	if (work)
		*work -= skb->truesize;
188
	atomic_sub(skb->truesize, &ip6_frags.mem);
L
Linus Torvalds 已提交
189 190 191 192 193 194 195
	kfree_skb(skb);
}

static inline void frag_free_queue(struct frag_queue *fq, int *work)
{
	if (work)
		*work -= sizeof(struct frag_queue);
196
	atomic_sub(sizeof(struct frag_queue), &ip6_frags.mem);
L
Linus Torvalds 已提交
197 198 199 200 201
	kfree(fq);
}

static inline struct frag_queue *frag_alloc_queue(void)
{
202
	struct frag_queue *fq = kzalloc(sizeof(struct frag_queue), GFP_ATOMIC);
L
Linus Torvalds 已提交
203 204 205

	if(!fq)
		return NULL;
206
	atomic_add(sizeof(struct frag_queue), &ip6_frags.mem);
L
Linus Torvalds 已提交
207 208 209 210 211 212 213 214 215 216
	return fq;
}

/* Destruction primitives. */

/* Complete destruction of fq. */
static void ip6_frag_destroy(struct frag_queue *fq, int *work)
{
	struct sk_buff *fp;

217 218
	BUG_TRAP(fq->q.last_in&COMPLETE);
	BUG_TRAP(del_timer(&fq->q.timer) == 0);
L
Linus Torvalds 已提交
219 220

	/* Release all fragment data. */
221
	fp = fq->q.fragments;
L
Linus Torvalds 已提交
222 223 224 225 226 227 228 229 230 231 232 233
	while (fp) {
		struct sk_buff *xp = fp->next;

		frag_kfree_skb(fp, work);
		fp = xp;
	}

	frag_free_queue(fq, work);
}

static __inline__ void fq_put(struct frag_queue *fq, int *work)
{
234
	if (atomic_dec_and_test(&fq->q.refcnt))
L
Linus Torvalds 已提交
235 236 237 238 239 240 241 242
		ip6_frag_destroy(fq, work);
}

/* Kill fq entry. It is not destroyed immediately,
 * because caller (and someone more) holds reference count.
 */
static __inline__ void fq_kill(struct frag_queue *fq)
{
243 244
	if (del_timer(&fq->q.timer))
		atomic_dec(&fq->q.refcnt);
L
Linus Torvalds 已提交
245

246
	if (!(fq->q.last_in & COMPLETE)) {
L
Linus Torvalds 已提交
247
		fq_unlink(fq);
248 249
		atomic_dec(&fq->q.refcnt);
		fq->q.last_in |= COMPLETE;
L
Linus Torvalds 已提交
250 251 252
	}
}

253
static void ip6_evictor(struct inet6_dev *idev)
L
Linus Torvalds 已提交
254 255 256 257 258
{
	struct frag_queue *fq;
	struct list_head *tmp;
	int work;

259
	work = atomic_read(&ip6_frags.mem) - sysctl_ip6frag_low_thresh;
L
Linus Torvalds 已提交
260 261 262 263
	if (work <= 0)
		return;

	while(work > 0) {
264 265 266
		read_lock(&ip6_frags.lock);
		if (list_empty(&ip6_frags.lru_list)) {
			read_unlock(&ip6_frags.lock);
L
Linus Torvalds 已提交
267 268
			return;
		}
269
		tmp = ip6_frags.lru_list.next;
270 271
		fq = list_entry(tmp, struct frag_queue, q.lru_list);
		atomic_inc(&fq->q.refcnt);
272
		read_unlock(&ip6_frags.lock);
L
Linus Torvalds 已提交
273

274 275
		spin_lock(&fq->q.lock);
		if (!(fq->q.last_in&COMPLETE))
L
Linus Torvalds 已提交
276
			fq_kill(fq);
277
		spin_unlock(&fq->q.lock);
L
Linus Torvalds 已提交
278 279

		fq_put(fq, &work);
280
		IP6_INC_STATS_BH(idev, IPSTATS_MIB_REASMFAILS);
L
Linus Torvalds 已提交
281 282 283 284 285 286
	}
}

static void ip6_frag_expire(unsigned long data)
{
	struct frag_queue *fq = (struct frag_queue *) data;
287
	struct net_device *dev = NULL;
L
Linus Torvalds 已提交
288

289
	spin_lock(&fq->q.lock);
L
Linus Torvalds 已提交
290

291
	if (fq->q.last_in & COMPLETE)
L
Linus Torvalds 已提交
292 293 294 295
		goto out;

	fq_kill(fq);

296
	dev = dev_get_by_index(&init_net, fq->iif);
297 298 299 300 301 302 303
	if (!dev)
		goto out;

	rcu_read_lock();
	IP6_INC_STATS_BH(__in6_dev_get(dev), IPSTATS_MIB_REASMTIMEOUT);
	IP6_INC_STATS_BH(__in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
	rcu_read_unlock();
L
Linus Torvalds 已提交
304

305
	/* Don't send error if the first segment did not arrive. */
306
	if (!(fq->q.last_in&FIRST_IN) || !fq->q.fragments)
307 308 309 310 311 312 313
		goto out;

	/*
	   But use as source device on which LAST ARRIVED
	   segment was received. And do not use fq->dev
	   pointer directly, device might already disappeared.
	 */
314 315
	fq->q.fragments->dev = dev;
	icmpv6_send(fq->q.fragments, ICMPV6_TIME_EXCEED, ICMPV6_EXC_FRAGTIME, 0, dev);
L
Linus Torvalds 已提交
316
out:
317 318
	if (dev)
		dev_put(dev);
319
	spin_unlock(&fq->q.lock);
L
Linus Torvalds 已提交
320 321 322 323 324 325
	fq_put(fq, NULL);
}

/* Creation primitives. */


326
static struct frag_queue *ip6_frag_intern(struct frag_queue *fq_in)
L
Linus Torvalds 已提交
327 328
{
	struct frag_queue *fq;
329
	unsigned int hash;
330 331 332
#ifdef CONFIG_SMP
	struct hlist_node *n;
#endif
L
Linus Torvalds 已提交
333

334
	write_lock(&ip6_frags.lock);
335
	hash = ip6qhashfn(fq_in->id, &fq_in->saddr, &fq_in->daddr);
L
Linus Torvalds 已提交
336
#ifdef CONFIG_SMP
337
	hlist_for_each_entry(fq, n, &ip6_frags.hash[hash], q.list) {
338
		if (fq->id == fq_in->id &&
L
Linus Torvalds 已提交
339 340
		    ipv6_addr_equal(&fq_in->saddr, &fq->saddr) &&
		    ipv6_addr_equal(&fq_in->daddr, &fq->daddr)) {
341
			atomic_inc(&fq->q.refcnt);
342
			write_unlock(&ip6_frags.lock);
343
			fq_in->q.last_in |= COMPLETE;
L
Linus Torvalds 已提交
344 345 346 347 348 349 350
			fq_put(fq_in, NULL);
			return fq;
		}
	}
#endif
	fq = fq_in;

351 352
	if (!mod_timer(&fq->q.timer, jiffies + sysctl_ip6frag_time))
		atomic_inc(&fq->q.refcnt);
L
Linus Torvalds 已提交
353

354
	atomic_inc(&fq->q.refcnt);
355
	hlist_add_head(&fq->q.list, &ip6_frags.hash[hash]);
356
	INIT_LIST_HEAD(&fq->q.lru_list);
357 358 359
	list_add_tail(&fq->q.lru_list, &ip6_frags.lru_list);
	ip6_frags.nqueues++;
	write_unlock(&ip6_frags.lock);
L
Linus Torvalds 已提交
360 361 362 363 364
	return fq;
}


static struct frag_queue *
A
Al Viro 已提交
365
ip6_frag_create(__be32 id, struct in6_addr *src, struct in6_addr *dst,
366
		struct inet6_dev *idev)
L
Linus Torvalds 已提交
367 368 369 370 371 372 373 374 375 376
{
	struct frag_queue *fq;

	if ((fq = frag_alloc_queue()) == NULL)
		goto oom;

	fq->id = id;
	ipv6_addr_copy(&fq->saddr, src);
	ipv6_addr_copy(&fq->daddr, dst);

377 378 379 380 381
	init_timer(&fq->q.timer);
	fq->q.timer.function = ip6_frag_expire;
	fq->q.timer.data = (long) fq;
	spin_lock_init(&fq->q.lock);
	atomic_set(&fq->q.refcnt, 1);
L
Linus Torvalds 已提交
382

383
	return ip6_frag_intern(fq);
L
Linus Torvalds 已提交
384 385

oom:
386
	IP6_INC_STATS_BH(idev, IPSTATS_MIB_REASMFAILS);
L
Linus Torvalds 已提交
387 388 389 390
	return NULL;
}

static __inline__ struct frag_queue *
A
Al Viro 已提交
391
fq_find(__be32 id, struct in6_addr *src, struct in6_addr *dst,
392
	struct inet6_dev *idev)
L
Linus Torvalds 已提交
393 394
{
	struct frag_queue *fq;
395
	struct hlist_node *n;
396
	unsigned int hash;
L
Linus Torvalds 已提交
397

398
	read_lock(&ip6_frags.lock);
399
	hash = ip6qhashfn(id, src, dst);
400
	hlist_for_each_entry(fq, n, &ip6_frags.hash[hash], q.list) {
401
		if (fq->id == id &&
L
Linus Torvalds 已提交
402 403
		    ipv6_addr_equal(src, &fq->saddr) &&
		    ipv6_addr_equal(dst, &fq->daddr)) {
404
			atomic_inc(&fq->q.refcnt);
405
			read_unlock(&ip6_frags.lock);
L
Linus Torvalds 已提交
406 407 408
			return fq;
		}
	}
409
	read_unlock(&ip6_frags.lock);
L
Linus Torvalds 已提交
410

411
	return ip6_frag_create(id, src, dst, idev);
L
Linus Torvalds 已提交
412 413 414
}


415
static int ip6_frag_queue(struct frag_queue *fq, struct sk_buff *skb,
L
Linus Torvalds 已提交
416 417 418
			   struct frag_hdr *fhdr, int nhoff)
{
	struct sk_buff *prev, *next;
419
	struct net_device *dev;
L
Linus Torvalds 已提交
420 421
	int offset, end;

422
	if (fq->q.last_in & COMPLETE)
L
Linus Torvalds 已提交
423 424 425
		goto err;

	offset = ntohs(fhdr->frag_off) & ~0x7;
426 427
	end = offset + (ntohs(ipv6_hdr(skb)->payload_len) -
			((u8 *)(fhdr + 1) - (u8 *)(ipv6_hdr(skb) + 1)));
L
Linus Torvalds 已提交
428 429

	if ((unsigned int)end > IPV6_MAXPLEN) {
430 431
		IP6_INC_STATS_BH(ip6_dst_idev(skb->dst),
				 IPSTATS_MIB_INHDRERRORS);
432 433 434
		icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
				  ((u8 *)&fhdr->frag_off -
				   skb_network_header(skb)));
435
		return -1;
L
Linus Torvalds 已提交
436 437
	}

438 439
	if (skb->ip_summed == CHECKSUM_COMPLETE) {
		const unsigned char *nh = skb_network_header(skb);
440
		skb->csum = csum_sub(skb->csum,
441 442 443
				     csum_partial(nh, (u8 *)(fhdr + 1) - nh,
						  0));
	}
L
Linus Torvalds 已提交
444 445 446 447 448 449

	/* 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.
		 */
450 451
		if (end < fq->q.len ||
		    ((fq->q.last_in & LAST_IN) && end != fq->q.len))
L
Linus Torvalds 已提交
452
			goto err;
453 454
		fq->q.last_in |= LAST_IN;
		fq->q.len = end;
L
Linus Torvalds 已提交
455 456 457 458 459 460 461 462
	} 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
			 */
463 464
			IP6_INC_STATS_BH(ip6_dst_idev(skb->dst),
					 IPSTATS_MIB_INHDRERRORS);
465
			icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
L
Linus Torvalds 已提交
466
					  offsetof(struct ipv6hdr, payload_len));
467
			return -1;
L
Linus Torvalds 已提交
468
		}
469
		if (end > fq->q.len) {
L
Linus Torvalds 已提交
470
			/* Some bits beyond end -> corruption. */
471
			if (fq->q.last_in & LAST_IN)
L
Linus Torvalds 已提交
472
				goto err;
473
			fq->q.len = end;
L
Linus Torvalds 已提交
474 475 476 477 478 479 480 481 482
		}
	}

	if (end == offset)
		goto err;

	/* Point into the IP datagram 'data' part. */
	if (!pskb_pull(skb, (u8 *) (fhdr + 1) - skb->data))
		goto err;
483

484 485
	if (pskb_trim_rcsum(skb, end - offset))
		goto err;
L
Linus Torvalds 已提交
486 487 488 489 490 491

	/* 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;
492
	for(next = fq->q.fragments; next != NULL; next = next->next) {
L
Linus Torvalds 已提交
493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528
		if (FRAG6_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 = (FRAG6_CB(prev)->offset + prev->len) - offset;

		if (i > 0) {
			offset += i;
			if (end <= offset)
				goto err;
			if (!pskb_pull(skb, i))
				goto err;
			if (skb->ip_summed != CHECKSUM_UNNECESSARY)
				skb->ip_summed = CHECKSUM_NONE;
		}
	}

	/* Look for overlap with succeeding segments.
	 * If we can merge fragments, do it.
	 */
	while (next && FRAG6_CB(next)->offset < end) {
		int i = end - FRAG6_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;
			FRAG6_CB(next)->offset += i;	/* next fragment */
529
			fq->q.meat -= i;
L
Linus Torvalds 已提交
530 531 532 533 534 535 536 537 538 539 540 541 542 543
			if (next->ip_summed != CHECKSUM_UNNECESSARY)
				next->ip_summed = CHECKSUM_NONE;
			break;
		} else {
			struct sk_buff *free_it = next;

			/* Old fragment is completely overridden with
			 * new one drop it.
			 */
			next = next->next;

			if (prev)
				prev->next = next;
			else
544
				fq->q.fragments = next;
L
Linus Torvalds 已提交
545

546
			fq->q.meat -= free_it->len;
L
Linus Torvalds 已提交
547 548 549 550 551 552 553 554 555 556 557
			frag_kfree_skb(free_it, NULL);
		}
	}

	FRAG6_CB(skb)->offset = offset;

	/* Insert this fragment in the chain of fragments. */
	skb->next = next;
	if (prev)
		prev->next = skb;
	else
558
		fq->q.fragments = skb;
L
Linus Torvalds 已提交
559

560 561 562 563 564
	dev = skb->dev;
	if (dev) {
		fq->iif = dev->ifindex;
		skb->dev = NULL;
	}
565 566
	fq->q.stamp = skb->tstamp;
	fq->q.meat += skb->len;
567
	atomic_add(skb->truesize, &ip6_frags.mem);
L
Linus Torvalds 已提交
568 569 570 571 572 573

	/* The first fragment.
	 * nhoffset is obtained from the first fragment, of course.
	 */
	if (offset == 0) {
		fq->nhoffset = nhoff;
574
		fq->q.last_in |= FIRST_IN;
L
Linus Torvalds 已提交
575
	}
576

577
	if (fq->q.last_in == (FIRST_IN | LAST_IN) && fq->q.meat == fq->q.len)
578 579
		return ip6_frag_reasm(fq, prev, dev);

580 581 582
	write_lock(&ip6_frags.lock);
	list_move_tail(&fq->q.lru_list, &ip6_frags.lru_list);
	write_unlock(&ip6_frags.lock);
583
	return -1;
L
Linus Torvalds 已提交
584 585

err:
586
	IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_REASMFAILS);
L
Linus Torvalds 已提交
587
	kfree_skb(skb);
588
	return -1;
L
Linus Torvalds 已提交
589 590 591 592 593 594 595 596 597 598 599
}

/*
 *	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.
 */
600
static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
L
Linus Torvalds 已提交
601 602
			  struct net_device *dev)
{
603
	struct sk_buff *fp, *head = fq->q.fragments;
L
Linus Torvalds 已提交
604 605 606 607 608
	int    payload_len;
	unsigned int nhoff;

	fq_kill(fq);

609 610 611 612 613 614 615 616 617 618 619
	/* 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;
		prev->next = fp;

620 621
		skb_morph(head, fq->q.fragments);
		head->next = fq->q.fragments->next;
622

623 624
		kfree_skb(fq->q.fragments);
		fq->q.fragments = head;
625 626
	}

L
Linus Torvalds 已提交
627 628 629 630
	BUG_TRAP(head != NULL);
	BUG_TRAP(FRAG6_CB(head)->offset == 0);

	/* Unfragmented part is taken from the first segment. */
631
	payload_len = ((head->data - skb_network_header(head)) -
632
		       sizeof(struct ipv6hdr) + fq->q.len -
633
		       sizeof(struct frag_hdr));
L
Linus Torvalds 已提交
634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660
	if (payload_len > IPV6_MAXPLEN)
		goto out_oversize;

	/* Head of list must not be cloned. */
	if (skb_cloned(head) && pskb_expand_head(head, 0, 0, GFP_ATOMIC))
		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. */
	if (skb_shinfo(head)->frag_list) {
		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;
		skb_shinfo(head)->frag_list = NULL;
		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;
661
		atomic_add(clone->truesize, &ip6_frags.mem);
L
Linus Torvalds 已提交
662 663 664 665 666
	}

	/* We have to remove fragment header from datagram and to relocate
	 * header in order to calculate ICV correctly. */
	nhoff = fq->nhoffset;
667
	skb_network_header(head)[nhoff] = skb_transport_header(head)[0];
668
	memmove(head->head + sizeof(struct frag_hdr), head->head,
L
Linus Torvalds 已提交
669
		(head->data - head->head) - sizeof(struct frag_hdr));
670 671
	head->mac_header += sizeof(struct frag_hdr);
	head->network_header += sizeof(struct frag_hdr);
L
Linus Torvalds 已提交
672 673

	skb_shinfo(head)->frag_list = head->next;
674
	skb_reset_transport_header(head);
675
	skb_push(head, head->data - skb_network_header(head));
676
	atomic_sub(head->truesize, &ip6_frags.mem);
L
Linus Torvalds 已提交
677 678 679 680 681 682

	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;
683
		else if (head->ip_summed == CHECKSUM_COMPLETE)
L
Linus Torvalds 已提交
684 685
			head->csum = csum_add(head->csum, fp->csum);
		head->truesize += fp->truesize;
686
		atomic_sub(fp->truesize, &ip6_frags.mem);
L
Linus Torvalds 已提交
687 688 689 690
	}

	head->next = NULL;
	head->dev = dev;
691
	head->tstamp = fq->q.stamp;
692
	ipv6_hdr(head)->payload_len = htons(payload_len);
693
	IP6CB(head)->nhoff = nhoff;
L
Linus Torvalds 已提交
694 695

	/* Yes, and fold redundant checksum back. 8) */
696
	if (head->ip_summed == CHECKSUM_COMPLETE)
697
		head->csum = csum_partial(skb_network_header(head),
698
					  skb_network_header_len(head),
699
					  head->csum);
L
Linus Torvalds 已提交
700

701 702 703
	rcu_read_lock();
	IP6_INC_STATS_BH(__in6_dev_get(dev), IPSTATS_MIB_REASMOKS);
	rcu_read_unlock();
704
	fq->q.fragments = NULL;
L
Linus Torvalds 已提交
705 706 707 708 709 710 711 712 713 714
	return 1;

out_oversize:
	if (net_ratelimit())
		printk(KERN_DEBUG "ip6_frag_reasm: payload len = %d\n", payload_len);
	goto out_fail;
out_oom:
	if (net_ratelimit())
		printk(KERN_DEBUG "ip6_frag_reasm: no memory for reassembly\n");
out_fail:
715 716 717
	rcu_read_lock();
	IP6_INC_STATS_BH(__in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
	rcu_read_unlock();
L
Linus Torvalds 已提交
718 719 720
	return -1;
}

721
static int ipv6_frag_rcv(struct sk_buff **skbp)
L
Linus Torvalds 已提交
722
{
723
	struct sk_buff *skb = *skbp;
L
Linus Torvalds 已提交
724 725
	struct frag_hdr *fhdr;
	struct frag_queue *fq;
726
	struct ipv6hdr *hdr = ipv6_hdr(skb);
L
Linus Torvalds 已提交
727

728
	IP6_INC_STATS_BH(ip6_dst_idev(skb->dst), IPSTATS_MIB_REASMREQDS);
L
Linus Torvalds 已提交
729 730 731

	/* Jumbo payload inhibits frag. header */
	if (hdr->payload_len==0) {
732
		IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_INHDRERRORS);
733 734
		icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
				  skb_network_header_len(skb));
L
Linus Torvalds 已提交
735 736
		return -1;
	}
737 738
	if (!pskb_may_pull(skb, (skb_transport_offset(skb) +
				 sizeof(struct frag_hdr)))) {
739
		IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_INHDRERRORS);
740 741
		icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
				  skb_network_header_len(skb));
L
Linus Torvalds 已提交
742 743 744
		return -1;
	}

745
	hdr = ipv6_hdr(skb);
746
	fhdr = (struct frag_hdr *)skb_transport_header(skb);
L
Linus Torvalds 已提交
747 748 749

	if (!(fhdr->frag_off & htons(0xFFF9))) {
		/* It is not a fragmented frame */
750
		skb->transport_header += sizeof(struct frag_hdr);
751
		IP6_INC_STATS_BH(ip6_dst_idev(skb->dst), IPSTATS_MIB_REASMOKS);
L
Linus Torvalds 已提交
752

753
		IP6CB(skb)->nhoff = (u8 *)fhdr - skb_network_header(skb);
L
Linus Torvalds 已提交
754 755 756
		return 1;
	}

757
	if (atomic_read(&ip6_frags.mem) > sysctl_ip6frag_high_thresh)
758
		ip6_evictor(ip6_dst_idev(skb->dst));
L
Linus Torvalds 已提交
759

760 761
	if ((fq = fq_find(fhdr->identification, &hdr->saddr, &hdr->daddr,
			  ip6_dst_idev(skb->dst))) != NULL) {
762
		int ret;
L
Linus Torvalds 已提交
763

764
		spin_lock(&fq->q.lock);
L
Linus Torvalds 已提交
765

766
		ret = ip6_frag_queue(fq, skb, fhdr, IP6CB(skb)->nhoff);
L
Linus Torvalds 已提交
767

768
		spin_unlock(&fq->q.lock);
L
Linus Torvalds 已提交
769 770 771 772
		fq_put(fq, NULL);
		return ret;
	}

773
	IP6_INC_STATS_BH(ip6_dst_idev(skb->dst), IPSTATS_MIB_REASMFAILS);
L
Linus Torvalds 已提交
774 775 776 777 778 779 780 781 782 783 784 785 786 787 788
	kfree_skb(skb);
	return -1;
}

static struct inet6_protocol frag_protocol =
{
	.handler	=	ipv6_frag_rcv,
	.flags		=	INET6_PROTO_NOPOLICY,
};

void __init ipv6_frag_init(void)
{
	if (inet6_add_protocol(&frag_protocol, IPPROTO_FRAGMENT) < 0)
		printk(KERN_ERR "ipv6_frag_init: Could not register protocol\n");

789 790 791 792
	init_timer(&ip6_frags.secret_timer);
	ip6_frags.secret_timer.function = ip6_frag_secret_rebuild;
	ip6_frags.secret_timer.expires = jiffies + sysctl_ip6frag_secret_interval;
	add_timer(&ip6_frags.secret_timer);
L
Linus Torvalds 已提交
793

794
	inet_frags_init(&ip6_frags);
L
Linus Torvalds 已提交
795
}