xref: /freebsd/sys/netinet6/frag6.c (revision 3fe92528afe8313fecf48822dde74bad5e380f48)
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 
161 	ip6 = mtod(m, struct ip6_hdr *);
162 #ifndef PULLDOWN_TEST
163 	IP6_EXTHDR_CHECK(m, offset, sizeof(struct ip6_frag), IPPROTO_DONE);
164 	ip6f = (struct ip6_frag *)((caddr_t)ip6 + offset);
165 #else
166 	IP6_EXTHDR_GET(ip6f, struct ip6_frag *, m, offset, sizeof(*ip6f));
167 	if (ip6f == NULL)
168 		return (IPPROTO_DONE);
169 #endif
170 
171 	dstifp = NULL;
172 #ifdef IN6_IFSTAT_STRICT
173 	/* find the destination interface of the packet. */
174 	if ((ia = ip6_getdstifaddr(m)) != NULL)
175 		dstifp = ia->ia_ifp;
176 #else
177 	/* we are violating the spec, this is not the destination interface */
178 	if ((m->m_flags & M_PKTHDR) != 0)
179 		dstifp = m->m_pkthdr.rcvif;
180 #endif
181 
182 	/* jumbo payload can't contain a fragment header */
183 	if (ip6->ip6_plen == 0) {
184 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, offset);
185 		in6_ifstat_inc(dstifp, ifs6_reass_fail);
186 		return IPPROTO_DONE;
187 	}
188 
189 	/*
190 	 * check whether fragment packet's fragment length is
191 	 * multiple of 8 octets.
192 	 * sizeof(struct ip6_frag) == 8
193 	 * sizeof(struct ip6_hdr) = 40
194 	 */
195 	if ((ip6f->ip6f_offlg & IP6F_MORE_FRAG) &&
196 	    (((ntohs(ip6->ip6_plen) - offset) & 0x7) != 0)) {
197 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
198 		    offsetof(struct ip6_hdr, ip6_plen));
199 		in6_ifstat_inc(dstifp, ifs6_reass_fail);
200 		return IPPROTO_DONE;
201 	}
202 
203 	ip6stat.ip6s_fragments++;
204 	in6_ifstat_inc(dstifp, ifs6_reass_reqd);
205 
206 	/* offset now points to data portion */
207 	offset += sizeof(struct ip6_frag);
208 
209 	IP6Q_LOCK();
210 
211 	/*
212 	 * Enforce upper bound on number of fragments.
213 	 * If maxfrag is 0, never accept fragments.
214 	 * If maxfrag is -1, accept all fragments without limitation.
215 	 */
216 	if (ip6_maxfrags < 0)
217 		;
218 	else if (frag6_nfrags >= (u_int)ip6_maxfrags)
219 		goto dropfrag;
220 
221 	for (q6 = ip6q.ip6q_next; q6 != &ip6q; q6 = q6->ip6q_next)
222 		if (ip6f->ip6f_ident == q6->ip6q_ident &&
223 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &q6->ip6q_src) &&
224 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &q6->ip6q_dst))
225 			break;
226 
227 	if (q6 == &ip6q) {
228 		/*
229 		 * the first fragment to arrive, create a reassembly queue.
230 		 */
231 		first_frag = 1;
232 
233 		/*
234 		 * Enforce upper bound on number of fragmented packets
235 		 * for which we attempt reassembly;
236 		 * If maxfragpackets is 0, never accept fragments.
237 		 * If maxfragpackets is -1, accept all fragments without
238 		 * limitation.
239 		 */
240 		if (ip6_maxfragpackets < 0)
241 			;
242 		else if (frag6_nfragpackets >= (u_int)ip6_maxfragpackets)
243 			goto dropfrag;
244 		frag6_nfragpackets++;
245 		q6 = (struct ip6q *)malloc(sizeof(struct ip6q), M_FTABLE,
246 		    M_NOWAIT);
247 		if (q6 == NULL)
248 			goto dropfrag;
249 		bzero(q6, sizeof(*q6));
250 
251 		frag6_insque(q6, &ip6q);
252 
253 		/* ip6q_nxt will be filled afterwards, from 1st fragment */
254 		q6->ip6q_down	= q6->ip6q_up = (struct ip6asfrag *)q6;
255 #ifdef notyet
256 		q6->ip6q_nxtp	= (u_char *)nxtp;
257 #endif
258 		q6->ip6q_ident	= ip6f->ip6f_ident;
259 		q6->ip6q_arrive = 0; /* Is it used anywhere? */
260 		q6->ip6q_ttl 	= IPV6_FRAGTTL;
261 		q6->ip6q_src	= ip6->ip6_src;
262 		q6->ip6q_dst	= ip6->ip6_dst;
263 		q6->ip6q_unfrglen = -1;	/* The 1st fragment has not arrived. */
264 
265 		q6->ip6q_nfrag = 0;
266 	}
267 
268 	/*
269 	 * If it's the 1st fragment, record the length of the
270 	 * unfragmentable part and the next header of the fragment header.
271 	 */
272 	fragoff = ntohs(ip6f->ip6f_offlg & IP6F_OFF_MASK);
273 	if (fragoff == 0) {
274 		q6->ip6q_unfrglen = offset - sizeof(struct ip6_hdr) -
275 		    sizeof(struct ip6_frag);
276 		q6->ip6q_nxt = ip6f->ip6f_nxt;
277 	}
278 
279 	/*
280 	 * Check that the reassembled packet would not exceed 65535 bytes
281 	 * in size.
282 	 * If it would exceed, discard the fragment and return an ICMP error.
283 	 */
284 	frgpartlen = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen) - offset;
285 	if (q6->ip6q_unfrglen >= 0) {
286 		/* The 1st fragment has already arrived. */
287 		if (q6->ip6q_unfrglen + fragoff + frgpartlen > IPV6_MAXPACKET) {
288 			icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
289 			    offset - sizeof(struct ip6_frag) +
290 			    offsetof(struct ip6_frag, ip6f_offlg));
291 			IP6Q_UNLOCK();
292 			return (IPPROTO_DONE);
293 		}
294 	} else if (fragoff + frgpartlen > IPV6_MAXPACKET) {
295 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
296 		    offset - sizeof(struct ip6_frag) +
297 		    offsetof(struct ip6_frag, ip6f_offlg));
298 		IP6Q_UNLOCK();
299 		return (IPPROTO_DONE);
300 	}
301 	/*
302 	 * If it's the first fragment, do the above check for each
303 	 * fragment already stored in the reassembly queue.
304 	 */
305 	if (fragoff == 0) {
306 		for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
307 		     af6 = af6dwn) {
308 			af6dwn = af6->ip6af_down;
309 
310 			if (q6->ip6q_unfrglen + af6->ip6af_off + af6->ip6af_frglen >
311 			    IPV6_MAXPACKET) {
312 				struct mbuf *merr = IP6_REASS_MBUF(af6);
313 				struct ip6_hdr *ip6err;
314 				int erroff = af6->ip6af_offset;
315 
316 				/* dequeue the fragment. */
317 				frag6_deq(af6);
318 				free(af6, M_FTABLE);
319 
320 				/* adjust pointer. */
321 				ip6err = mtod(merr, struct ip6_hdr *);
322 
323 				/*
324 				 * Restore source and destination addresses
325 				 * in the erroneous IPv6 header.
326 				 */
327 				ip6err->ip6_src = q6->ip6q_src;
328 				ip6err->ip6_dst = q6->ip6q_dst;
329 
330 				icmp6_error(merr, ICMP6_PARAM_PROB,
331 				    ICMP6_PARAMPROB_HEADER,
332 				    erroff - sizeof(struct ip6_frag) +
333 				    offsetof(struct ip6_frag, ip6f_offlg));
334 			}
335 		}
336 	}
337 
338 	ip6af = (struct ip6asfrag *)malloc(sizeof(struct ip6asfrag), M_FTABLE,
339 	    M_NOWAIT);
340 	if (ip6af == NULL)
341 		goto dropfrag;
342 	bzero(ip6af, sizeof(*ip6af));
343 	ip6af->ip6af_head = ip6->ip6_flow;
344 	ip6af->ip6af_len = ip6->ip6_plen;
345 	ip6af->ip6af_nxt = ip6->ip6_nxt;
346 	ip6af->ip6af_hlim = ip6->ip6_hlim;
347 	ip6af->ip6af_mff = ip6f->ip6f_offlg & IP6F_MORE_FRAG;
348 	ip6af->ip6af_off = fragoff;
349 	ip6af->ip6af_frglen = frgpartlen;
350 	ip6af->ip6af_offset = offset;
351 	IP6_REASS_MBUF(ip6af) = m;
352 
353 	if (first_frag) {
354 		af6 = (struct ip6asfrag *)q6;
355 		goto insert;
356 	}
357 
358 	/*
359 	 * Handle ECN by comparing this segment with the first one;
360 	 * if CE is set, do not lose CE.
361 	 * drop if CE and not-ECT are mixed for the same packet.
362 	 */
363 	ecn = (ntohl(ip6->ip6_flow) >> 20) & IPTOS_ECN_MASK;
364 	ecn0 = (ntohl(q6->ip6q_down->ip6af_head) >> 20) & IPTOS_ECN_MASK;
365 	if (ecn == IPTOS_ECN_CE) {
366 		if (ecn0 == IPTOS_ECN_NOTECT) {
367 			free(ip6af, M_FTABLE);
368 			goto dropfrag;
369 		}
370 		if (ecn0 != IPTOS_ECN_CE)
371 			q6->ip6q_down->ip6af_head |= htonl(IPTOS_ECN_CE << 20);
372 	}
373 	if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) {
374 		free(ip6af, M_FTABLE);
375 		goto dropfrag;
376 	}
377 
378 	/*
379 	 * Find a segment which begins after this one does.
380 	 */
381 	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
382 	     af6 = af6->ip6af_down)
383 		if (af6->ip6af_off > ip6af->ip6af_off)
384 			break;
385 
386 #if 0
387 	/*
388 	 * If there is a preceding segment, it may provide some of
389 	 * our data already.  If so, drop the data from the incoming
390 	 * segment.  If it provides all of our data, drop us.
391 	 */
392 	if (af6->ip6af_up != (struct ip6asfrag *)q6) {
393 		i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
394 			- ip6af->ip6af_off;
395 		if (i > 0) {
396 			if (i >= ip6af->ip6af_frglen)
397 				goto dropfrag;
398 			m_adj(IP6_REASS_MBUF(ip6af), i);
399 			ip6af->ip6af_off += i;
400 			ip6af->ip6af_frglen -= i;
401 		}
402 	}
403 
404 	/*
405 	 * While we overlap succeeding segments trim them or,
406 	 * if they are completely covered, dequeue them.
407 	 */
408 	while (af6 != (struct ip6asfrag *)q6 &&
409 	       ip6af->ip6af_off + ip6af->ip6af_frglen > af6->ip6af_off) {
410 		i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
411 		if (i < af6->ip6af_frglen) {
412 			af6->ip6af_frglen -= i;
413 			af6->ip6af_off += i;
414 			m_adj(IP6_REASS_MBUF(af6), i);
415 			break;
416 		}
417 		af6 = af6->ip6af_down;
418 		m_freem(IP6_REASS_MBUF(af6->ip6af_up));
419 		frag6_deq(af6->ip6af_up);
420 	}
421 #else
422 	/*
423 	 * If the incoming framgent overlaps some existing fragments in
424 	 * the reassembly queue, drop it, since it is dangerous to override
425 	 * existing fragments from a security point of view.
426 	 * We don't know which fragment is the bad guy - here we trust
427 	 * fragment that came in earlier, with no real reason.
428 	 */
429 	if (af6->ip6af_up != (struct ip6asfrag *)q6) {
430 		i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
431 			- ip6af->ip6af_off;
432 		if (i > 0) {
433 #if 0				/* suppress the noisy log */
434 			log(LOG_ERR, "%d bytes of a fragment from %s "
435 			    "overlaps the previous fragment\n",
436 			    i, ip6_sprintf(&q6->ip6q_src));
437 #endif
438 			free(ip6af, M_FTABLE);
439 			goto dropfrag;
440 		}
441 	}
442 	if (af6 != (struct ip6asfrag *)q6) {
443 		i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
444 		if (i > 0) {
445 #if 0				/* suppress the noisy log */
446 			log(LOG_ERR, "%d bytes of a fragment from %s "
447 			    "overlaps the succeeding fragment",
448 			    i, ip6_sprintf(&q6->ip6q_src));
449 #endif
450 			free(ip6af, M_FTABLE);
451 			goto dropfrag;
452 		}
453 	}
454 #endif
455 
456 insert:
457 
458 	/*
459 	 * Stick new segment in its place;
460 	 * check for complete reassembly.
461 	 * Move to front of packet queue, as we are
462 	 * the most recently active fragmented packet.
463 	 */
464 	frag6_enq(ip6af, af6->ip6af_up);
465 	frag6_nfrags++;
466 	q6->ip6q_nfrag++;
467 #if 0 /* xxx */
468 	if (q6 != ip6q.ip6q_next) {
469 		frag6_remque(q6);
470 		frag6_insque(q6, &ip6q);
471 	}
472 #endif
473 	next = 0;
474 	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
475 	     af6 = af6->ip6af_down) {
476 		if (af6->ip6af_off != next) {
477 			IP6Q_UNLOCK();
478 			return IPPROTO_DONE;
479 		}
480 		next += af6->ip6af_frglen;
481 	}
482 	if (af6->ip6af_up->ip6af_mff) {
483 		IP6Q_UNLOCK();
484 		return IPPROTO_DONE;
485 	}
486 
487 	/*
488 	 * Reassembly is complete; concatenate fragments.
489 	 */
490 	ip6af = q6->ip6q_down;
491 	t = m = IP6_REASS_MBUF(ip6af);
492 	af6 = ip6af->ip6af_down;
493 	frag6_deq(ip6af);
494 	while (af6 != (struct ip6asfrag *)q6) {
495 		af6dwn = af6->ip6af_down;
496 		frag6_deq(af6);
497 		while (t->m_next)
498 			t = t->m_next;
499 		t->m_next = IP6_REASS_MBUF(af6);
500 		m_adj(t->m_next, af6->ip6af_offset);
501 		free(af6, M_FTABLE);
502 		af6 = af6dwn;
503 	}
504 
505 	/* adjust offset to point where the original next header starts */
506 	offset = ip6af->ip6af_offset - sizeof(struct ip6_frag);
507 	free(ip6af, M_FTABLE);
508 	ip6 = mtod(m, struct ip6_hdr *);
509 	ip6->ip6_plen = htons((u_short)next + offset - sizeof(struct ip6_hdr));
510 	ip6->ip6_src = q6->ip6q_src;
511 	ip6->ip6_dst = q6->ip6q_dst;
512 	nxt = q6->ip6q_nxt;
513 #ifdef notyet
514 	*q6->ip6q_nxtp = (u_char)(nxt & 0xff);
515 #endif
516 
517 	/*
518 	 * Delete frag6 header with as a few cost as possible.
519 	 */
520 	if (offset < m->m_len) {
521 		ovbcopy((caddr_t)ip6, (caddr_t)ip6 + sizeof(struct ip6_frag),
522 			offset);
523 		m->m_data += sizeof(struct ip6_frag);
524 		m->m_len -= sizeof(struct ip6_frag);
525 	} else {
526 		/* this comes with no copy if the boundary is on cluster */
527 		if ((t = m_split(m, offset, M_DONTWAIT)) == NULL) {
528 			frag6_remque(q6);
529 			frag6_nfrags -= q6->ip6q_nfrag;
530 			free(q6, M_FTABLE);
531 			frag6_nfragpackets--;
532 			goto dropfrag;
533 		}
534 		m_adj(t, sizeof(struct ip6_frag));
535 		m_cat(m, t);
536 	}
537 
538 	/*
539 	 * Store NXT to the original.
540 	 */
541 	{
542 		char *prvnxtp = ip6_get_prevhdr(m, offset); /* XXX */
543 		*prvnxtp = nxt;
544 	}
545 
546 	frag6_remque(q6);
547 	frag6_nfrags -= q6->ip6q_nfrag;
548 	free(q6, M_FTABLE);
549 	frag6_nfragpackets--;
550 
551 	if (m->m_flags & M_PKTHDR) { /* Isn't it always true? */
552 		int plen = 0;
553 		for (t = m; t; t = t->m_next)
554 			plen += t->m_len;
555 		m->m_pkthdr.len = plen;
556 	}
557 
558 	ip6stat.ip6s_reassembled++;
559 	in6_ifstat_inc(dstifp, ifs6_reass_ok);
560 
561 	/*
562 	 * Tell launch routine the next header
563 	 */
564 
565 	*mp = m;
566 	*offp = offset;
567 
568 	IP6Q_UNLOCK();
569 	return nxt;
570 
571  dropfrag:
572 	IP6Q_UNLOCK();
573 	in6_ifstat_inc(dstifp, ifs6_reass_fail);
574 	ip6stat.ip6s_fragdropped++;
575 	m_freem(m);
576 	return IPPROTO_DONE;
577 }
578 
579 /*
580  * Free a fragment reassembly header and all
581  * associated datagrams.
582  */
583 void
584 frag6_freef(q6)
585 	struct ip6q *q6;
586 {
587 	struct ip6asfrag *af6, *down6;
588 
589 	IP6Q_LOCK_ASSERT();
590 
591 	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
592 	     af6 = down6) {
593 		struct mbuf *m = IP6_REASS_MBUF(af6);
594 
595 		down6 = af6->ip6af_down;
596 		frag6_deq(af6);
597 
598 		/*
599 		 * Return ICMP time exceeded error for the 1st fragment.
600 		 * Just free other fragments.
601 		 */
602 		if (af6->ip6af_off == 0) {
603 			struct ip6_hdr *ip6;
604 
605 			/* adjust pointer */
606 			ip6 = mtod(m, struct ip6_hdr *);
607 
608 			/* restore source and destination addresses */
609 			ip6->ip6_src = q6->ip6q_src;
610 			ip6->ip6_dst = q6->ip6q_dst;
611 
612 			icmp6_error(m, ICMP6_TIME_EXCEEDED,
613 				    ICMP6_TIME_EXCEED_REASSEMBLY, 0);
614 		} else
615 			m_freem(m);
616 		free(af6, M_FTABLE);
617 	}
618 	frag6_remque(q6);
619 	frag6_nfrags -= q6->ip6q_nfrag;
620 	free(q6, M_FTABLE);
621 	frag6_nfragpackets--;
622 }
623 
624 /*
625  * Put an ip fragment on a reassembly chain.
626  * Like insque, but pointers in middle of structure.
627  */
628 void
629 frag6_enq(af6, up6)
630 	struct ip6asfrag *af6, *up6;
631 {
632 
633 	IP6Q_LOCK_ASSERT();
634 
635 	af6->ip6af_up = up6;
636 	af6->ip6af_down = up6->ip6af_down;
637 	up6->ip6af_down->ip6af_up = af6;
638 	up6->ip6af_down = af6;
639 }
640 
641 /*
642  * To frag6_enq as remque is to insque.
643  */
644 void
645 frag6_deq(af6)
646 	struct ip6asfrag *af6;
647 {
648 
649 	IP6Q_LOCK_ASSERT();
650 
651 	af6->ip6af_up->ip6af_down = af6->ip6af_down;
652 	af6->ip6af_down->ip6af_up = af6->ip6af_up;
653 }
654 
655 void
656 frag6_insque(new, old)
657 	struct ip6q *new, *old;
658 {
659 
660 	IP6Q_LOCK_ASSERT();
661 
662 	new->ip6q_prev = old;
663 	new->ip6q_next = old->ip6q_next;
664 	old->ip6q_next->ip6q_prev= new;
665 	old->ip6q_next = new;
666 }
667 
668 void
669 frag6_remque(p6)
670 	struct ip6q *p6;
671 {
672 
673 	IP6Q_LOCK_ASSERT();
674 
675 	p6->ip6q_prev->ip6q_next = p6->ip6q_next;
676 	p6->ip6q_next->ip6q_prev = p6->ip6q_prev;
677 }
678 
679 /*
680  * IPv6 reassembling timer processing;
681  * if a timer expires on a reassembly
682  * queue, discard it.
683  */
684 void
685 frag6_slowtimo()
686 {
687 	struct ip6q *q6;
688 
689 #if 0
690 	GIANT_REQUIRED;	/* XXX bz: ip6_forward_rt */
691 #endif
692 
693 	IP6Q_LOCK();
694 	q6 = ip6q.ip6q_next;
695 	if (q6)
696 		while (q6 != &ip6q) {
697 			--q6->ip6q_ttl;
698 			q6 = q6->ip6q_next;
699 			if (q6->ip6q_prev->ip6q_ttl == 0) {
700 				ip6stat.ip6s_fragtimeout++;
701 				/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
702 				frag6_freef(q6->ip6q_prev);
703 			}
704 		}
705 	/*
706 	 * If we are over the maximum number of fragments
707 	 * (due to the limit being lowered), drain off
708 	 * enough to get down to the new limit.
709 	 */
710 	while (frag6_nfragpackets > (u_int)ip6_maxfragpackets &&
711 	    ip6q.ip6q_prev) {
712 		ip6stat.ip6s_fragoverflow++;
713 		/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
714 		frag6_freef(ip6q.ip6q_prev);
715 	}
716 	IP6Q_UNLOCK();
717 
718 #if 0
719 	/*
720 	 * Routing changes might produce a better route than we last used;
721 	 * make sure we notice eventually, even if forwarding only for one
722 	 * destination and the cache is never replaced.
723 	 */
724 	if (ip6_forward_rt.ro_rt) {
725 		RTFREE(ip6_forward_rt.ro_rt);
726 		ip6_forward_rt.ro_rt = 0;
727 	}
728 	if (ipsrcchk_rt.ro_rt) {
729 		RTFREE(ipsrcchk_rt.ro_rt);
730 		ipsrcchk_rt.ro_rt = 0;
731 	}
732 #endif
733 }
734 
735 /*
736  * Drain off all datagram fragments.
737  */
738 void
739 frag6_drain()
740 {
741 
742 	if (IP6Q_TRYLOCK() == 0)
743 		return;
744 	while (ip6q.ip6q_next != &ip6q) {
745 		ip6stat.ip6s_fragdropped++;
746 		/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
747 		frag6_freef(ip6q.ip6q_next);
748 	}
749 	IP6Q_UNLOCK();
750 }
751