xref: /freebsd/sys/netinet6/frag6.c (revision f0a75d274af375d15b97b830966b99a02b7db911)
1 /*	$FreeBSD$	*/
2 /*	$KAME: frag6.c,v 1.33 2002/01/07 11:34:48 kjc Exp $	*/
3 
4 /*-
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #include <sys/param.h>
34 #include <sys/systm.h>
35 #include <sys/malloc.h>
36 #include <sys/mbuf.h>
37 #include <sys/domain.h>
38 #include <sys/protosw.h>
39 #include <sys/socket.h>
40 #include <sys/errno.h>
41 #include <sys/time.h>
42 #include <sys/kernel.h>
43 #include <sys/syslog.h>
44 
45 #include <net/if.h>
46 #include <net/route.h>
47 
48 #include <netinet/in.h>
49 #include <netinet/in_var.h>
50 #include <netinet/ip6.h>
51 #include <netinet6/ip6_var.h>
52 #include <netinet/icmp6.h>
53 #include <netinet/in_systm.h>	/* for ECN definitions */
54 #include <netinet/ip.h>		/* for ECN definitions */
55 
56 /*
57  * Define it to get a correct behavior on per-interface statistics.
58  * You will need to perform an extra routing table lookup, per fragment,
59  * to do it.  This may, or may not be, a performance hit.
60  */
61 #define IN6_IFSTAT_STRICT
62 
63 static void frag6_enq __P((struct ip6asfrag *, struct ip6asfrag *));
64 static void frag6_deq __P((struct ip6asfrag *));
65 static void frag6_insque __P((struct ip6q *, struct ip6q *));
66 static void frag6_remque __P((struct ip6q *));
67 static void frag6_freef __P((struct ip6q *));
68 
69 static struct mtx ip6qlock;
70 /*
71  * These fields all protected by ip6qlock.
72  */
73 static u_int frag6_nfragpackets;
74 static u_int frag6_nfrags;
75 static struct	ip6q ip6q;	/* ip6 reassemble queue */
76 
77 #define	IP6Q_LOCK_INIT()	mtx_init(&ip6qlock, "ip6qlock", NULL, MTX_DEF);
78 #define	IP6Q_LOCK()		mtx_lock(&ip6qlock)
79 #define	IP6Q_TRYLOCK()		mtx_trylock(&ip6qlock)
80 #define	IP6Q_LOCK_ASSERT()	mtx_assert(&ip6qlock, MA_OWNED)
81 #define	IP6Q_UNLOCK()		mtx_unlock(&ip6qlock)
82 
83 static MALLOC_DEFINE(M_FTABLE, "fragment", "fragment reassembly header");
84 
85 /*
86  * Initialise reassembly queue and fragment identifier.
87  */
88 static void
89 frag6_change(void *tag)
90 {
91 
92 	ip6_maxfragpackets = nmbclusters / 4;
93 	ip6_maxfrags = nmbclusters / 4;
94 }
95 
96 void
97 frag6_init()
98 {
99 
100 	ip6_maxfragpackets = nmbclusters / 4;
101 	ip6_maxfrags = nmbclusters / 4;
102 	EVENTHANDLER_REGISTER(nmbclusters_change,
103 	    frag6_change, NULL, EVENTHANDLER_PRI_ANY);
104 
105 	IP6Q_LOCK_INIT();
106 
107 	ip6q.ip6q_next = ip6q.ip6q_prev = &ip6q;
108 }
109 
110 /*
111  * In RFC2460, fragment and reassembly rule do not agree with each other,
112  * in terms of next header field handling in fragment header.
113  * While the sender will use the same value for all of the fragmented packets,
114  * receiver is suggested not to check the consistency.
115  *
116  * fragment rule (p20):
117  *	(2) A Fragment header containing:
118  *	The Next Header value that identifies the first header of
119  *	the Fragmentable Part of the original packet.
120  *		-> next header field is same for all fragments
121  *
122  * reassembly rule (p21):
123  *	The Next Header field of the last header of the Unfragmentable
124  *	Part is obtained from the Next Header field of the first
125  *	fragment's Fragment header.
126  *		-> should grab it from the first fragment only
127  *
128  * The following note also contradicts with fragment rule - noone is going to
129  * send different fragment with different next header field.
130  *
131  * additional note (p22):
132  *	The Next Header values in the Fragment headers of different
133  *	fragments of the same original packet may differ.  Only the value
134  *	from the Offset zero fragment packet is used for reassembly.
135  *		-> should grab it from the first fragment only
136  *
137  * There is no explicit reason given in the RFC.  Historical reason maybe?
138  */
139 /*
140  * Fragment input
141  */
142 int
143 frag6_input(mp, offp, proto)
144 	struct mbuf **mp;
145 	int *offp, proto;
146 {
147 	struct mbuf *m = *mp, *t;
148 	struct ip6_hdr *ip6;
149 	struct ip6_frag *ip6f;
150 	struct ip6q *q6;
151 	struct ip6asfrag *af6, *ip6af, *af6dwn;
152 #ifdef IN6_IFSTAT_STRICT
153 	struct in6_ifaddr *ia;
154 #endif
155 	int offset = *offp, nxt, i, next;
156 	int first_frag = 0;
157 	int fragoff, frgpartlen;	/* must be larger than u_int16_t */
158 	struct ifnet *dstifp;
159 	u_int8_t ecn, ecn0;
160 #if 0
161 	char ip6buf[INET6_ADDRSTRLEN];
162 #endif
163 
164 	ip6 = mtod(m, struct ip6_hdr *);
165 #ifndef PULLDOWN_TEST
166 	IP6_EXTHDR_CHECK(m, offset, sizeof(struct ip6_frag), IPPROTO_DONE);
167 	ip6f = (struct ip6_frag *)((caddr_t)ip6 + offset);
168 #else
169 	IP6_EXTHDR_GET(ip6f, struct ip6_frag *, m, offset, sizeof(*ip6f));
170 	if (ip6f == NULL)
171 		return (IPPROTO_DONE);
172 #endif
173 
174 	dstifp = NULL;
175 #ifdef IN6_IFSTAT_STRICT
176 	/* find the destination interface of the packet. */
177 	if ((ia = ip6_getdstifaddr(m)) != NULL)
178 		dstifp = ia->ia_ifp;
179 #else
180 	/* we are violating the spec, this is not the destination interface */
181 	if ((m->m_flags & M_PKTHDR) != 0)
182 		dstifp = m->m_pkthdr.rcvif;
183 #endif
184 
185 	/* jumbo payload can't contain a fragment header */
186 	if (ip6->ip6_plen == 0) {
187 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, offset);
188 		in6_ifstat_inc(dstifp, ifs6_reass_fail);
189 		return IPPROTO_DONE;
190 	}
191 
192 	/*
193 	 * check whether fragment packet's fragment length is
194 	 * multiple of 8 octets.
195 	 * sizeof(struct ip6_frag) == 8
196 	 * sizeof(struct ip6_hdr) = 40
197 	 */
198 	if ((ip6f->ip6f_offlg & IP6F_MORE_FRAG) &&
199 	    (((ntohs(ip6->ip6_plen) - offset) & 0x7) != 0)) {
200 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
201 		    offsetof(struct ip6_hdr, ip6_plen));
202 		in6_ifstat_inc(dstifp, ifs6_reass_fail);
203 		return IPPROTO_DONE;
204 	}
205 
206 	ip6stat.ip6s_fragments++;
207 	in6_ifstat_inc(dstifp, ifs6_reass_reqd);
208 
209 	/* offset now points to data portion */
210 	offset += sizeof(struct ip6_frag);
211 
212 	IP6Q_LOCK();
213 
214 	/*
215 	 * Enforce upper bound on number of fragments.
216 	 * If maxfrag is 0, never accept fragments.
217 	 * If maxfrag is -1, accept all fragments without limitation.
218 	 */
219 	if (ip6_maxfrags < 0)
220 		;
221 	else if (frag6_nfrags >= (u_int)ip6_maxfrags)
222 		goto dropfrag;
223 
224 	for (q6 = ip6q.ip6q_next; q6 != &ip6q; q6 = q6->ip6q_next)
225 		if (ip6f->ip6f_ident == q6->ip6q_ident &&
226 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &q6->ip6q_src) &&
227 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &q6->ip6q_dst))
228 			break;
229 
230 	if (q6 == &ip6q) {
231 		/*
232 		 * the first fragment to arrive, create a reassembly queue.
233 		 */
234 		first_frag = 1;
235 
236 		/*
237 		 * Enforce upper bound on number of fragmented packets
238 		 * for which we attempt reassembly;
239 		 * If maxfragpackets is 0, never accept fragments.
240 		 * If maxfragpackets is -1, accept all fragments without
241 		 * limitation.
242 		 */
243 		if (ip6_maxfragpackets < 0)
244 			;
245 		else if (frag6_nfragpackets >= (u_int)ip6_maxfragpackets)
246 			goto dropfrag;
247 		frag6_nfragpackets++;
248 		q6 = (struct ip6q *)malloc(sizeof(struct ip6q), M_FTABLE,
249 		    M_NOWAIT);
250 		if (q6 == NULL)
251 			goto dropfrag;
252 		bzero(q6, sizeof(*q6));
253 
254 		frag6_insque(q6, &ip6q);
255 
256 		/* ip6q_nxt will be filled afterwards, from 1st fragment */
257 		q6->ip6q_down	= q6->ip6q_up = (struct ip6asfrag *)q6;
258 #ifdef notyet
259 		q6->ip6q_nxtp	= (u_char *)nxtp;
260 #endif
261 		q6->ip6q_ident	= ip6f->ip6f_ident;
262 		q6->ip6q_arrive = 0; /* Is it used anywhere? */
263 		q6->ip6q_ttl 	= IPV6_FRAGTTL;
264 		q6->ip6q_src	= ip6->ip6_src;
265 		q6->ip6q_dst	= ip6->ip6_dst;
266 		q6->ip6q_unfrglen = -1;	/* The 1st fragment has not arrived. */
267 
268 		q6->ip6q_nfrag = 0;
269 	}
270 
271 	/*
272 	 * If it's the 1st fragment, record the length of the
273 	 * unfragmentable part and the next header of the fragment header.
274 	 */
275 	fragoff = ntohs(ip6f->ip6f_offlg & IP6F_OFF_MASK);
276 	if (fragoff == 0) {
277 		q6->ip6q_unfrglen = offset - sizeof(struct ip6_hdr) -
278 		    sizeof(struct ip6_frag);
279 		q6->ip6q_nxt = ip6f->ip6f_nxt;
280 	}
281 
282 	/*
283 	 * Check that the reassembled packet would not exceed 65535 bytes
284 	 * in size.
285 	 * If it would exceed, discard the fragment and return an ICMP error.
286 	 */
287 	frgpartlen = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen) - offset;
288 	if (q6->ip6q_unfrglen >= 0) {
289 		/* The 1st fragment has already arrived. */
290 		if (q6->ip6q_unfrglen + fragoff + frgpartlen > IPV6_MAXPACKET) {
291 			icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
292 			    offset - sizeof(struct ip6_frag) +
293 			    offsetof(struct ip6_frag, ip6f_offlg));
294 			IP6Q_UNLOCK();
295 			return (IPPROTO_DONE);
296 		}
297 	} else if (fragoff + frgpartlen > IPV6_MAXPACKET) {
298 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
299 		    offset - sizeof(struct ip6_frag) +
300 		    offsetof(struct ip6_frag, ip6f_offlg));
301 		IP6Q_UNLOCK();
302 		return (IPPROTO_DONE);
303 	}
304 	/*
305 	 * If it's the first fragment, do the above check for each
306 	 * fragment already stored in the reassembly queue.
307 	 */
308 	if (fragoff == 0) {
309 		for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
310 		     af6 = af6dwn) {
311 			af6dwn = af6->ip6af_down;
312 
313 			if (q6->ip6q_unfrglen + af6->ip6af_off + af6->ip6af_frglen >
314 			    IPV6_MAXPACKET) {
315 				struct mbuf *merr = IP6_REASS_MBUF(af6);
316 				struct ip6_hdr *ip6err;
317 				int erroff = af6->ip6af_offset;
318 
319 				/* dequeue the fragment. */
320 				frag6_deq(af6);
321 				free(af6, M_FTABLE);
322 
323 				/* adjust pointer. */
324 				ip6err = mtod(merr, struct ip6_hdr *);
325 
326 				/*
327 				 * Restore source and destination addresses
328 				 * in the erroneous IPv6 header.
329 				 */
330 				ip6err->ip6_src = q6->ip6q_src;
331 				ip6err->ip6_dst = q6->ip6q_dst;
332 
333 				icmp6_error(merr, ICMP6_PARAM_PROB,
334 				    ICMP6_PARAMPROB_HEADER,
335 				    erroff - sizeof(struct ip6_frag) +
336 				    offsetof(struct ip6_frag, ip6f_offlg));
337 			}
338 		}
339 	}
340 
341 	ip6af = (struct ip6asfrag *)malloc(sizeof(struct ip6asfrag), M_FTABLE,
342 	    M_NOWAIT);
343 	if (ip6af == NULL)
344 		goto dropfrag;
345 	bzero(ip6af, sizeof(*ip6af));
346 	ip6af->ip6af_head = ip6->ip6_flow;
347 	ip6af->ip6af_len = ip6->ip6_plen;
348 	ip6af->ip6af_nxt = ip6->ip6_nxt;
349 	ip6af->ip6af_hlim = ip6->ip6_hlim;
350 	ip6af->ip6af_mff = ip6f->ip6f_offlg & IP6F_MORE_FRAG;
351 	ip6af->ip6af_off = fragoff;
352 	ip6af->ip6af_frglen = frgpartlen;
353 	ip6af->ip6af_offset = offset;
354 	IP6_REASS_MBUF(ip6af) = m;
355 
356 	if (first_frag) {
357 		af6 = (struct ip6asfrag *)q6;
358 		goto insert;
359 	}
360 
361 	/*
362 	 * Handle ECN by comparing this segment with the first one;
363 	 * if CE is set, do not lose CE.
364 	 * drop if CE and not-ECT are mixed for the same packet.
365 	 */
366 	ecn = (ntohl(ip6->ip6_flow) >> 20) & IPTOS_ECN_MASK;
367 	ecn0 = (ntohl(q6->ip6q_down->ip6af_head) >> 20) & IPTOS_ECN_MASK;
368 	if (ecn == IPTOS_ECN_CE) {
369 		if (ecn0 == IPTOS_ECN_NOTECT) {
370 			free(ip6af, M_FTABLE);
371 			goto dropfrag;
372 		}
373 		if (ecn0 != IPTOS_ECN_CE)
374 			q6->ip6q_down->ip6af_head |= htonl(IPTOS_ECN_CE << 20);
375 	}
376 	if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) {
377 		free(ip6af, M_FTABLE);
378 		goto dropfrag;
379 	}
380 
381 	/*
382 	 * Find a segment which begins after this one does.
383 	 */
384 	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
385 	     af6 = af6->ip6af_down)
386 		if (af6->ip6af_off > ip6af->ip6af_off)
387 			break;
388 
389 #if 0
390 	/*
391 	 * If there is a preceding segment, it may provide some of
392 	 * our data already.  If so, drop the data from the incoming
393 	 * segment.  If it provides all of our data, drop us.
394 	 */
395 	if (af6->ip6af_up != (struct ip6asfrag *)q6) {
396 		i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
397 			- ip6af->ip6af_off;
398 		if (i > 0) {
399 			if (i >= ip6af->ip6af_frglen)
400 				goto dropfrag;
401 			m_adj(IP6_REASS_MBUF(ip6af), i);
402 			ip6af->ip6af_off += i;
403 			ip6af->ip6af_frglen -= i;
404 		}
405 	}
406 
407 	/*
408 	 * While we overlap succeeding segments trim them or,
409 	 * if they are completely covered, dequeue them.
410 	 */
411 	while (af6 != (struct ip6asfrag *)q6 &&
412 	       ip6af->ip6af_off + ip6af->ip6af_frglen > af6->ip6af_off) {
413 		i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
414 		if (i < af6->ip6af_frglen) {
415 			af6->ip6af_frglen -= i;
416 			af6->ip6af_off += i;
417 			m_adj(IP6_REASS_MBUF(af6), i);
418 			break;
419 		}
420 		af6 = af6->ip6af_down;
421 		m_freem(IP6_REASS_MBUF(af6->ip6af_up));
422 		frag6_deq(af6->ip6af_up);
423 	}
424 #else
425 	/*
426 	 * If the incoming framgent overlaps some existing fragments in
427 	 * the reassembly queue, drop it, since it is dangerous to override
428 	 * existing fragments from a security point of view.
429 	 * We don't know which fragment is the bad guy - here we trust
430 	 * fragment that came in earlier, with no real reason.
431 	 */
432 	if (af6->ip6af_up != (struct ip6asfrag *)q6) {
433 		i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
434 			- ip6af->ip6af_off;
435 		if (i > 0) {
436 #if 0				/* suppress the noisy log */
437 			log(LOG_ERR, "%d bytes of a fragment from %s "
438 			    "overlaps the previous fragment\n",
439 			    i, ip6_sprintf(ip6buf, &q6->ip6q_src));
440 #endif
441 			free(ip6af, M_FTABLE);
442 			goto dropfrag;
443 		}
444 	}
445 	if (af6 != (struct ip6asfrag *)q6) {
446 		i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
447 		if (i > 0) {
448 #if 0				/* suppress the noisy log */
449 			log(LOG_ERR, "%d bytes of a fragment from %s "
450 			    "overlaps the succeeding fragment",
451 			    i, ip6_sprintf(ip6buf, &q6->ip6q_src));
452 #endif
453 			free(ip6af, M_FTABLE);
454 			goto dropfrag;
455 		}
456 	}
457 #endif
458 
459 insert:
460 
461 	/*
462 	 * Stick new segment in its place;
463 	 * check for complete reassembly.
464 	 * Move to front of packet queue, as we are
465 	 * the most recently active fragmented packet.
466 	 */
467 	frag6_enq(ip6af, af6->ip6af_up);
468 	frag6_nfrags++;
469 	q6->ip6q_nfrag++;
470 #if 0 /* xxx */
471 	if (q6 != ip6q.ip6q_next) {
472 		frag6_remque(q6);
473 		frag6_insque(q6, &ip6q);
474 	}
475 #endif
476 	next = 0;
477 	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
478 	     af6 = af6->ip6af_down) {
479 		if (af6->ip6af_off != next) {
480 			IP6Q_UNLOCK();
481 			return IPPROTO_DONE;
482 		}
483 		next += af6->ip6af_frglen;
484 	}
485 	if (af6->ip6af_up->ip6af_mff) {
486 		IP6Q_UNLOCK();
487 		return IPPROTO_DONE;
488 	}
489 
490 	/*
491 	 * Reassembly is complete; concatenate fragments.
492 	 */
493 	ip6af = q6->ip6q_down;
494 	t = m = IP6_REASS_MBUF(ip6af);
495 	af6 = ip6af->ip6af_down;
496 	frag6_deq(ip6af);
497 	while (af6 != (struct ip6asfrag *)q6) {
498 		af6dwn = af6->ip6af_down;
499 		frag6_deq(af6);
500 		while (t->m_next)
501 			t = t->m_next;
502 		t->m_next = IP6_REASS_MBUF(af6);
503 		m_adj(t->m_next, af6->ip6af_offset);
504 		free(af6, M_FTABLE);
505 		af6 = af6dwn;
506 	}
507 
508 	/* adjust offset to point where the original next header starts */
509 	offset = ip6af->ip6af_offset - sizeof(struct ip6_frag);
510 	free(ip6af, M_FTABLE);
511 	ip6 = mtod(m, struct ip6_hdr *);
512 	ip6->ip6_plen = htons((u_short)next + offset - sizeof(struct ip6_hdr));
513 	ip6->ip6_src = q6->ip6q_src;
514 	ip6->ip6_dst = q6->ip6q_dst;
515 	nxt = q6->ip6q_nxt;
516 #ifdef notyet
517 	*q6->ip6q_nxtp = (u_char)(nxt & 0xff);
518 #endif
519 
520 	/*
521 	 * Delete frag6 header with as a few cost as possible.
522 	 */
523 	if (offset < m->m_len) {
524 		ovbcopy((caddr_t)ip6, (caddr_t)ip6 + sizeof(struct ip6_frag),
525 			offset);
526 		m->m_data += sizeof(struct ip6_frag);
527 		m->m_len -= sizeof(struct ip6_frag);
528 	} else {
529 		/* this comes with no copy if the boundary is on cluster */
530 		if ((t = m_split(m, offset, M_DONTWAIT)) == NULL) {
531 			frag6_remque(q6);
532 			frag6_nfrags -= q6->ip6q_nfrag;
533 			free(q6, M_FTABLE);
534 			frag6_nfragpackets--;
535 			goto dropfrag;
536 		}
537 		m_adj(t, sizeof(struct ip6_frag));
538 		m_cat(m, t);
539 	}
540 
541 	/*
542 	 * Store NXT to the original.
543 	 */
544 	{
545 		char *prvnxtp = ip6_get_prevhdr(m, offset); /* XXX */
546 		*prvnxtp = nxt;
547 	}
548 
549 	frag6_remque(q6);
550 	frag6_nfrags -= q6->ip6q_nfrag;
551 	free(q6, M_FTABLE);
552 	frag6_nfragpackets--;
553 
554 	if (m->m_flags & M_PKTHDR) { /* Isn't it always true? */
555 		int plen = 0;
556 		for (t = m; t; t = t->m_next)
557 			plen += t->m_len;
558 		m->m_pkthdr.len = plen;
559 	}
560 
561 	ip6stat.ip6s_reassembled++;
562 	in6_ifstat_inc(dstifp, ifs6_reass_ok);
563 
564 	/*
565 	 * Tell launch routine the next header
566 	 */
567 
568 	*mp = m;
569 	*offp = offset;
570 
571 	IP6Q_UNLOCK();
572 	return nxt;
573 
574  dropfrag:
575 	IP6Q_UNLOCK();
576 	in6_ifstat_inc(dstifp, ifs6_reass_fail);
577 	ip6stat.ip6s_fragdropped++;
578 	m_freem(m);
579 	return IPPROTO_DONE;
580 }
581 
582 /*
583  * Free a fragment reassembly header and all
584  * associated datagrams.
585  */
586 void
587 frag6_freef(q6)
588 	struct ip6q *q6;
589 {
590 	struct ip6asfrag *af6, *down6;
591 
592 	IP6Q_LOCK_ASSERT();
593 
594 	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
595 	     af6 = down6) {
596 		struct mbuf *m = IP6_REASS_MBUF(af6);
597 
598 		down6 = af6->ip6af_down;
599 		frag6_deq(af6);
600 
601 		/*
602 		 * Return ICMP time exceeded error for the 1st fragment.
603 		 * Just free other fragments.
604 		 */
605 		if (af6->ip6af_off == 0) {
606 			struct ip6_hdr *ip6;
607 
608 			/* adjust pointer */
609 			ip6 = mtod(m, struct ip6_hdr *);
610 
611 			/* restore source and destination addresses */
612 			ip6->ip6_src = q6->ip6q_src;
613 			ip6->ip6_dst = q6->ip6q_dst;
614 
615 			icmp6_error(m, ICMP6_TIME_EXCEEDED,
616 				    ICMP6_TIME_EXCEED_REASSEMBLY, 0);
617 		} else
618 			m_freem(m);
619 		free(af6, M_FTABLE);
620 	}
621 	frag6_remque(q6);
622 	frag6_nfrags -= q6->ip6q_nfrag;
623 	free(q6, M_FTABLE);
624 	frag6_nfragpackets--;
625 }
626 
627 /*
628  * Put an ip fragment on a reassembly chain.
629  * Like insque, but pointers in middle of structure.
630  */
631 void
632 frag6_enq(af6, up6)
633 	struct ip6asfrag *af6, *up6;
634 {
635 
636 	IP6Q_LOCK_ASSERT();
637 
638 	af6->ip6af_up = up6;
639 	af6->ip6af_down = up6->ip6af_down;
640 	up6->ip6af_down->ip6af_up = af6;
641 	up6->ip6af_down = af6;
642 }
643 
644 /*
645  * To frag6_enq as remque is to insque.
646  */
647 void
648 frag6_deq(af6)
649 	struct ip6asfrag *af6;
650 {
651 
652 	IP6Q_LOCK_ASSERT();
653 
654 	af6->ip6af_up->ip6af_down = af6->ip6af_down;
655 	af6->ip6af_down->ip6af_up = af6->ip6af_up;
656 }
657 
658 void
659 frag6_insque(new, old)
660 	struct ip6q *new, *old;
661 {
662 
663 	IP6Q_LOCK_ASSERT();
664 
665 	new->ip6q_prev = old;
666 	new->ip6q_next = old->ip6q_next;
667 	old->ip6q_next->ip6q_prev= new;
668 	old->ip6q_next = new;
669 }
670 
671 void
672 frag6_remque(p6)
673 	struct ip6q *p6;
674 {
675 
676 	IP6Q_LOCK_ASSERT();
677 
678 	p6->ip6q_prev->ip6q_next = p6->ip6q_next;
679 	p6->ip6q_next->ip6q_prev = p6->ip6q_prev;
680 }
681 
682 /*
683  * IPv6 reassembling timer processing;
684  * if a timer expires on a reassembly
685  * queue, discard it.
686  */
687 void
688 frag6_slowtimo()
689 {
690 	struct ip6q *q6;
691 
692 #if 0
693 	GIANT_REQUIRED;	/* XXX bz: ip6_forward_rt */
694 #endif
695 
696 	IP6Q_LOCK();
697 	q6 = ip6q.ip6q_next;
698 	if (q6)
699 		while (q6 != &ip6q) {
700 			--q6->ip6q_ttl;
701 			q6 = q6->ip6q_next;
702 			if (q6->ip6q_prev->ip6q_ttl == 0) {
703 				ip6stat.ip6s_fragtimeout++;
704 				/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
705 				frag6_freef(q6->ip6q_prev);
706 			}
707 		}
708 	/*
709 	 * If we are over the maximum number of fragments
710 	 * (due to the limit being lowered), drain off
711 	 * enough to get down to the new limit.
712 	 */
713 	while (frag6_nfragpackets > (u_int)ip6_maxfragpackets &&
714 	    ip6q.ip6q_prev) {
715 		ip6stat.ip6s_fragoverflow++;
716 		/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
717 		frag6_freef(ip6q.ip6q_prev);
718 	}
719 	IP6Q_UNLOCK();
720 
721 #if 0
722 	/*
723 	 * Routing changes might produce a better route than we last used;
724 	 * make sure we notice eventually, even if forwarding only for one
725 	 * destination and the cache is never replaced.
726 	 */
727 	if (ip6_forward_rt.ro_rt) {
728 		RTFREE(ip6_forward_rt.ro_rt);
729 		ip6_forward_rt.ro_rt = 0;
730 	}
731 	if (ipsrcchk_rt.ro_rt) {
732 		RTFREE(ipsrcchk_rt.ro_rt);
733 		ipsrcchk_rt.ro_rt = 0;
734 	}
735 #endif
736 }
737 
738 /*
739  * Drain off all datagram fragments.
740  */
741 void
742 frag6_drain()
743 {
744 
745 	if (IP6Q_TRYLOCK() == 0)
746 		return;
747 	while (ip6q.ip6q_next != &ip6q) {
748 		ip6stat.ip6s_fragdropped++;
749 		/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
750 		frag6_freef(ip6q.ip6q_next);
751 	}
752 	IP6Q_UNLOCK();
753 }
754