xref: /freebsd/sys/netinet6/frag6.c (revision ae2cbf4c649fecd3302a3bea16672345582d2562)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: frag6.c,v 1.33 2002/01/07 11:34:48 kjc Exp $
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/malloc.h>
38 #include <sys/mbuf.h>
39 #include <sys/domain.h>
40 #include <sys/protosw.h>
41 #include <sys/socket.h>
42 #include <sys/errno.h>
43 #include <sys/time.h>
44 #include <sys/kernel.h>
45 #include <sys/syslog.h>
46 #include <sys/vimage.h>
47 
48 #include <net/if.h>
49 #include <net/route.h>
50 
51 #include <netinet/in.h>
52 #include <netinet/in_var.h>
53 #include <netinet/ip6.h>
54 #include <netinet6/ip6_var.h>
55 #include <netinet/icmp6.h>
56 #include <netinet/in_systm.h>	/* for ECN definitions */
57 #include <netinet/ip.h>		/* for ECN definitions */
58 
59 /*
60  * Define it to get a correct behavior on per-interface statistics.
61  * You will need to perform an extra routing table lookup, per fragment,
62  * to do it.  This may, or may not be, a performance hit.
63  */
64 #define IN6_IFSTAT_STRICT
65 
66 static void frag6_enq(struct ip6asfrag *, struct ip6asfrag *);
67 static void frag6_deq(struct ip6asfrag *);
68 static void frag6_insque(struct ip6q *, struct ip6q *);
69 static void frag6_remque(struct ip6q *);
70 static void frag6_freef(struct ip6q *);
71 
72 static struct mtx ip6qlock;
73 /*
74  * These fields all protected by ip6qlock.
75  */
76 static u_int frag6_nfragpackets;
77 static u_int frag6_nfrags;
78 static struct	ip6q ip6q;	/* ip6 reassemble queue */
79 
80 #define	IP6Q_LOCK_INIT()	mtx_init(&ip6qlock, "ip6qlock", NULL, MTX_DEF);
81 #define	IP6Q_LOCK()		mtx_lock(&ip6qlock)
82 #define	IP6Q_TRYLOCK()		mtx_trylock(&ip6qlock)
83 #define	IP6Q_LOCK_ASSERT()	mtx_assert(&ip6qlock, MA_OWNED)
84 #define	IP6Q_UNLOCK()		mtx_unlock(&ip6qlock)
85 
86 static MALLOC_DEFINE(M_FTABLE, "fragment", "fragment reassembly header");
87 
88 /*
89  * Initialise reassembly queue and fragment identifier.
90  */
91 static void
92 frag6_change(void *tag)
93 {
94 
95 	V_ip6_maxfragpackets = nmbclusters / 4;
96 	V_ip6_maxfrags = nmbclusters / 4;
97 }
98 
99 void
100 frag6_init(void)
101 {
102 
103 	V_ip6_maxfragpackets = nmbclusters / 4;
104 	V_ip6_maxfrags = nmbclusters / 4;
105 	EVENTHANDLER_REGISTER(nmbclusters_change,
106 	    frag6_change, NULL, EVENTHANDLER_PRI_ANY);
107 
108 	IP6Q_LOCK_INIT();
109 
110 	V_ip6q.ip6q_next = V_ip6q.ip6q_prev = &V_ip6q;
111 }
112 
113 /*
114  * In RFC2460, fragment and reassembly rule do not agree with each other,
115  * in terms of next header field handling in fragment header.
116  * While the sender will use the same value for all of the fragmented packets,
117  * receiver is suggested not to check the consistency.
118  *
119  * fragment rule (p20):
120  *	(2) A Fragment header containing:
121  *	The Next Header value that identifies the first header of
122  *	the Fragmentable Part of the original packet.
123  *		-> next header field is same for all fragments
124  *
125  * reassembly rule (p21):
126  *	The Next Header field of the last header of the Unfragmentable
127  *	Part is obtained from the Next Header field of the first
128  *	fragment's Fragment header.
129  *		-> should grab it from the first fragment only
130  *
131  * The following note also contradicts with fragment rule - noone is going to
132  * send different fragment with different next header field.
133  *
134  * additional note (p22):
135  *	The Next Header values in the Fragment headers of different
136  *	fragments of the same original packet may differ.  Only the value
137  *	from the Offset zero fragment packet is used for reassembly.
138  *		-> should grab it from the first fragment only
139  *
140  * There is no explicit reason given in the RFC.  Historical reason maybe?
141  */
142 /*
143  * Fragment input
144  */
145 int
146 frag6_input(struct mbuf **mp, int *offp, int proto)
147 {
148 	struct mbuf *m = *mp, *t;
149 	struct ip6_hdr *ip6;
150 	struct ip6_frag *ip6f;
151 	struct ip6q *q6;
152 	struct ip6asfrag *af6, *ip6af, *af6dwn;
153 #ifdef IN6_IFSTAT_STRICT
154 	struct in6_ifaddr *ia;
155 #endif
156 	int offset = *offp, nxt, i, next;
157 	int first_frag = 0;
158 	int fragoff, frgpartlen;	/* must be larger than u_int16_t */
159 	struct ifnet *dstifp;
160 	u_int8_t ecn, ecn0;
161 #if 0
162 	char ip6buf[INET6_ADDRSTRLEN];
163 #endif
164 
165 	ip6 = mtod(m, struct ip6_hdr *);
166 #ifndef PULLDOWN_TEST
167 	IP6_EXTHDR_CHECK(m, offset, sizeof(struct ip6_frag), IPPROTO_DONE);
168 	ip6f = (struct ip6_frag *)((caddr_t)ip6 + offset);
169 #else
170 	IP6_EXTHDR_GET(ip6f, struct ip6_frag *, m, offset, sizeof(*ip6f));
171 	if (ip6f == NULL)
172 		return (IPPROTO_DONE);
173 #endif
174 
175 	dstifp = NULL;
176 #ifdef IN6_IFSTAT_STRICT
177 	/* find the destination interface of the packet. */
178 	if ((ia = ip6_getdstifaddr(m)) != NULL)
179 		dstifp = ia->ia_ifp;
180 #else
181 	/* we are violating the spec, this is not the destination interface */
182 	if ((m->m_flags & M_PKTHDR) != 0)
183 		dstifp = m->m_pkthdr.rcvif;
184 #endif
185 
186 	/* jumbo payload can't contain a fragment header */
187 	if (ip6->ip6_plen == 0) {
188 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, offset);
189 		in6_ifstat_inc(dstifp, ifs6_reass_fail);
190 		return IPPROTO_DONE;
191 	}
192 
193 	/*
194 	 * check whether fragment packet's fragment length is
195 	 * multiple of 8 octets.
196 	 * sizeof(struct ip6_frag) == 8
197 	 * sizeof(struct ip6_hdr) = 40
198 	 */
199 	if ((ip6f->ip6f_offlg & IP6F_MORE_FRAG) &&
200 	    (((ntohs(ip6->ip6_plen) - offset) & 0x7) != 0)) {
201 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
202 		    offsetof(struct ip6_hdr, ip6_plen));
203 		in6_ifstat_inc(dstifp, ifs6_reass_fail);
204 		return IPPROTO_DONE;
205 	}
206 
207 	V_ip6stat.ip6s_fragments++;
208 	in6_ifstat_inc(dstifp, ifs6_reass_reqd);
209 
210 	/* offset now points to data portion */
211 	offset += sizeof(struct ip6_frag);
212 
213 	IP6Q_LOCK();
214 
215 	/*
216 	 * Enforce upper bound on number of fragments.
217 	 * If maxfrag is 0, never accept fragments.
218 	 * If maxfrag is -1, accept all fragments without limitation.
219 	 */
220 	if (V_ip6_maxfrags < 0)
221 		;
222 	else if (V_frag6_nfrags >= (u_int)V_ip6_maxfrags)
223 		goto dropfrag;
224 
225 	for (q6 = V_ip6q.ip6q_next; q6 != &V_ip6q; q6 = q6->ip6q_next)
226 		if (ip6f->ip6f_ident == q6->ip6q_ident &&
227 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &q6->ip6q_src) &&
228 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &q6->ip6q_dst))
229 			break;
230 
231 	if (q6 == &V_ip6q) {
232 		/*
233 		 * the first fragment to arrive, create a reassembly queue.
234 		 */
235 		first_frag = 1;
236 
237 		/*
238 		 * Enforce upper bound on number of fragmented packets
239 		 * for which we attempt reassembly;
240 		 * If maxfragpackets is 0, never accept fragments.
241 		 * If maxfragpackets is -1, accept all fragments without
242 		 * limitation.
243 		 */
244 		if (V_ip6_maxfragpackets < 0)
245 			;
246 		else if (V_frag6_nfragpackets >= (u_int)V_ip6_maxfragpackets)
247 			goto dropfrag;
248 		V_frag6_nfragpackets++;
249 		q6 = (struct ip6q *)malloc(sizeof(struct ip6q), M_FTABLE,
250 		    M_NOWAIT);
251 		if (q6 == NULL)
252 			goto dropfrag;
253 		bzero(q6, sizeof(*q6));
254 
255 		frag6_insque(q6, &V_ip6q);
256 
257 		/* ip6q_nxt will be filled afterwards, from 1st fragment */
258 		q6->ip6q_down	= q6->ip6q_up = (struct ip6asfrag *)q6;
259 #ifdef notyet
260 		q6->ip6q_nxtp	= (u_char *)nxtp;
261 #endif
262 		q6->ip6q_ident	= ip6f->ip6f_ident;
263 		q6->ip6q_ttl	= IPV6_FRAGTTL;
264 		q6->ip6q_src	= ip6->ip6_src;
265 		q6->ip6q_dst	= ip6->ip6_dst;
266 		q6->ip6q_ecn	=
267 		    (ntohl(ip6->ip6_flow) >> 20) & IPTOS_ECN_MASK;
268 		q6->ip6q_unfrglen = -1;	/* The 1st fragment has not arrived. */
269 
270 		q6->ip6q_nfrag = 0;
271 	}
272 
273 	/*
274 	 * If it's the 1st fragment, record the length of the
275 	 * unfragmentable part and the next header of the fragment header.
276 	 */
277 	fragoff = ntohs(ip6f->ip6f_offlg & IP6F_OFF_MASK);
278 	if (fragoff == 0) {
279 		q6->ip6q_unfrglen = offset - sizeof(struct ip6_hdr) -
280 		    sizeof(struct ip6_frag);
281 		q6->ip6q_nxt = ip6f->ip6f_nxt;
282 	}
283 
284 	/*
285 	 * Check that the reassembled packet would not exceed 65535 bytes
286 	 * in size.
287 	 * If it would exceed, discard the fragment and return an ICMP error.
288 	 */
289 	frgpartlen = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen) - offset;
290 	if (q6->ip6q_unfrglen >= 0) {
291 		/* The 1st fragment has already arrived. */
292 		if (q6->ip6q_unfrglen + fragoff + frgpartlen > IPV6_MAXPACKET) {
293 			icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
294 			    offset - sizeof(struct ip6_frag) +
295 			    offsetof(struct ip6_frag, ip6f_offlg));
296 			IP6Q_UNLOCK();
297 			return (IPPROTO_DONE);
298 		}
299 	} else if (fragoff + frgpartlen > IPV6_MAXPACKET) {
300 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
301 		    offset - sizeof(struct ip6_frag) +
302 		    offsetof(struct ip6_frag, ip6f_offlg));
303 		IP6Q_UNLOCK();
304 		return (IPPROTO_DONE);
305 	}
306 	/*
307 	 * If it's the first fragment, do the above check for each
308 	 * fragment already stored in the reassembly queue.
309 	 */
310 	if (fragoff == 0) {
311 		for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
312 		     af6 = af6dwn) {
313 			af6dwn = af6->ip6af_down;
314 
315 			if (q6->ip6q_unfrglen + af6->ip6af_off + af6->ip6af_frglen >
316 			    IPV6_MAXPACKET) {
317 				struct mbuf *merr = IP6_REASS_MBUF(af6);
318 				struct ip6_hdr *ip6err;
319 				int erroff = af6->ip6af_offset;
320 
321 				/* dequeue the fragment. */
322 				frag6_deq(af6);
323 				free(af6, M_FTABLE);
324 
325 				/* adjust pointer. */
326 				ip6err = mtod(merr, struct ip6_hdr *);
327 
328 				/*
329 				 * Restore source and destination addresses
330 				 * in the erroneous IPv6 header.
331 				 */
332 				ip6err->ip6_src = q6->ip6q_src;
333 				ip6err->ip6_dst = q6->ip6q_dst;
334 
335 				icmp6_error(merr, ICMP6_PARAM_PROB,
336 				    ICMP6_PARAMPROB_HEADER,
337 				    erroff - sizeof(struct ip6_frag) +
338 				    offsetof(struct ip6_frag, ip6f_offlg));
339 			}
340 		}
341 	}
342 
343 	ip6af = (struct ip6asfrag *)malloc(sizeof(struct ip6asfrag), M_FTABLE,
344 	    M_NOWAIT);
345 	if (ip6af == NULL)
346 		goto dropfrag;
347 	bzero(ip6af, sizeof(*ip6af));
348 	ip6af->ip6af_mff = ip6f->ip6f_offlg & IP6F_MORE_FRAG;
349 	ip6af->ip6af_off = fragoff;
350 	ip6af->ip6af_frglen = frgpartlen;
351 	ip6af->ip6af_offset = offset;
352 	IP6_REASS_MBUF(ip6af) = m;
353 
354 	if (first_frag) {
355 		af6 = (struct ip6asfrag *)q6;
356 		goto insert;
357 	}
358 
359 	/*
360 	 * Handle ECN by comparing this segment with the first one;
361 	 * if CE is set, do not lose CE.
362 	 * drop if CE and not-ECT are mixed for the same packet.
363 	 */
364 	ecn = (ntohl(ip6->ip6_flow) >> 20) & IPTOS_ECN_MASK;
365 	ecn0 = q6->ip6q_ecn;
366 	if (ecn == IPTOS_ECN_CE) {
367 		if (ecn0 == IPTOS_ECN_NOTECT) {
368 			free(ip6af, M_FTABLE);
369 			goto dropfrag;
370 		}
371 		if (ecn0 != IPTOS_ECN_CE)
372 			q6->ip6q_ecn = IPTOS_ECN_CE;
373 	}
374 	if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) {
375 		free(ip6af, M_FTABLE);
376 		goto dropfrag;
377 	}
378 
379 	/*
380 	 * Find a segment which begins after this one does.
381 	 */
382 	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
383 	     af6 = af6->ip6af_down)
384 		if (af6->ip6af_off > ip6af->ip6af_off)
385 			break;
386 
387 #if 0
388 	/*
389 	 * If there is a preceding segment, it may provide some of
390 	 * our data already.  If so, drop the data from the incoming
391 	 * segment.  If it provides all of our data, drop us.
392 	 */
393 	if (af6->ip6af_up != (struct ip6asfrag *)q6) {
394 		i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
395 			- ip6af->ip6af_off;
396 		if (i > 0) {
397 			if (i >= ip6af->ip6af_frglen)
398 				goto dropfrag;
399 			m_adj(IP6_REASS_MBUF(ip6af), i);
400 			ip6af->ip6af_off += i;
401 			ip6af->ip6af_frglen -= i;
402 		}
403 	}
404 
405 	/*
406 	 * While we overlap succeeding segments trim them or,
407 	 * if they are completely covered, dequeue them.
408 	 */
409 	while (af6 != (struct ip6asfrag *)q6 &&
410 	       ip6af->ip6af_off + ip6af->ip6af_frglen > af6->ip6af_off) {
411 		i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
412 		if (i < af6->ip6af_frglen) {
413 			af6->ip6af_frglen -= i;
414 			af6->ip6af_off += i;
415 			m_adj(IP6_REASS_MBUF(af6), i);
416 			break;
417 		}
418 		af6 = af6->ip6af_down;
419 		m_freem(IP6_REASS_MBUF(af6->ip6af_up));
420 		frag6_deq(af6->ip6af_up);
421 	}
422 #else
423 	/*
424 	 * If the incoming framgent overlaps some existing fragments in
425 	 * the reassembly queue, drop it, since it is dangerous to override
426 	 * existing fragments from a security point of view.
427 	 * We don't know which fragment is the bad guy - here we trust
428 	 * fragment that came in earlier, with no real reason.
429 	 *
430 	 * Note: due to changes after disabling this part, mbuf passed to
431 	 * m_adj() below now does not meet the requirement.
432 	 */
433 	if (af6->ip6af_up != (struct ip6asfrag *)q6) {
434 		i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
435 			- ip6af->ip6af_off;
436 		if (i > 0) {
437 #if 0				/* suppress the noisy log */
438 			log(LOG_ERR, "%d bytes of a fragment from %s "
439 			    "overlaps the previous fragment\n",
440 			    i, ip6_sprintf(ip6buf, &q6->ip6q_src));
441 #endif
442 			free(ip6af, M_FTABLE);
443 			goto dropfrag;
444 		}
445 	}
446 	if (af6 != (struct ip6asfrag *)q6) {
447 		i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
448 		if (i > 0) {
449 #if 0				/* suppress the noisy log */
450 			log(LOG_ERR, "%d bytes of a fragment from %s "
451 			    "overlaps the succeeding fragment",
452 			    i, ip6_sprintf(ip6buf, &q6->ip6q_src));
453 #endif
454 			free(ip6af, M_FTABLE);
455 			goto dropfrag;
456 		}
457 	}
458 #endif
459 
460 insert:
461 
462 	/*
463 	 * Stick new segment in its place;
464 	 * check for complete reassembly.
465 	 * Move to front of packet queue, as we are
466 	 * the most recently active fragmented packet.
467 	 */
468 	frag6_enq(ip6af, af6->ip6af_up);
469 	V_frag6_nfrags++;
470 	q6->ip6q_nfrag++;
471 #if 0 /* xxx */
472 	if (q6 != V_ip6q.ip6q_next) {
473 		frag6_remque(q6);
474 		frag6_insque(q6, &V_ip6q);
475 	}
476 #endif
477 	next = 0;
478 	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
479 	     af6 = af6->ip6af_down) {
480 		if (af6->ip6af_off != next) {
481 			IP6Q_UNLOCK();
482 			return IPPROTO_DONE;
483 		}
484 		next += af6->ip6af_frglen;
485 	}
486 	if (af6->ip6af_up->ip6af_mff) {
487 		IP6Q_UNLOCK();
488 		return IPPROTO_DONE;
489 	}
490 
491 	/*
492 	 * Reassembly is complete; concatenate fragments.
493 	 */
494 	ip6af = q6->ip6q_down;
495 	t = m = IP6_REASS_MBUF(ip6af);
496 	af6 = ip6af->ip6af_down;
497 	frag6_deq(ip6af);
498 	while (af6 != (struct ip6asfrag *)q6) {
499 		af6dwn = af6->ip6af_down;
500 		frag6_deq(af6);
501 		while (t->m_next)
502 			t = t->m_next;
503 		t->m_next = IP6_REASS_MBUF(af6);
504 		m_adj(t->m_next, af6->ip6af_offset);
505 		free(af6, M_FTABLE);
506 		af6 = af6dwn;
507 	}
508 
509 	/* adjust offset to point where the original next header starts */
510 	offset = ip6af->ip6af_offset - sizeof(struct ip6_frag);
511 	free(ip6af, M_FTABLE);
512 	ip6 = mtod(m, struct ip6_hdr *);
513 	ip6->ip6_plen = htons((u_short)next + offset - sizeof(struct ip6_hdr));
514 	if (q6->ip6q_ecn == IPTOS_ECN_CE)
515 		ip6->ip6_flow |= htonl(IPTOS_ECN_CE << 20);
516 	nxt = q6->ip6q_nxt;
517 #ifdef notyet
518 	*q6->ip6q_nxtp = (u_char)(nxt & 0xff);
519 #endif
520 
521 	/* Delete frag6 header */
522 	if (m->m_len >= offset + sizeof(struct ip6_frag)) {
523 		/* This is the only possible case with !PULLDOWN_TEST */
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 			V_frag6_nfrags -= q6->ip6q_nfrag;
533 			free(q6, M_FTABLE);
534 			V_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 	V_frag6_nfrags -= q6->ip6q_nfrag;
551 	free(q6, M_FTABLE);
552 	V_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 	V_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 	V_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(struct ip6q *q6)
588 {
589 	struct ip6asfrag *af6, *down6;
590 
591 	IP6Q_LOCK_ASSERT();
592 
593 	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
594 	     af6 = down6) {
595 		struct mbuf *m = IP6_REASS_MBUF(af6);
596 
597 		down6 = af6->ip6af_down;
598 		frag6_deq(af6);
599 
600 		/*
601 		 * Return ICMP time exceeded error for the 1st fragment.
602 		 * Just free other fragments.
603 		 */
604 		if (af6->ip6af_off == 0) {
605 			struct ip6_hdr *ip6;
606 
607 			/* adjust pointer */
608 			ip6 = mtod(m, struct ip6_hdr *);
609 
610 			/* restore source and destination addresses */
611 			ip6->ip6_src = q6->ip6q_src;
612 			ip6->ip6_dst = q6->ip6q_dst;
613 
614 			icmp6_error(m, ICMP6_TIME_EXCEEDED,
615 				    ICMP6_TIME_EXCEED_REASSEMBLY, 0);
616 		} else
617 			m_freem(m);
618 		free(af6, M_FTABLE);
619 	}
620 	frag6_remque(q6);
621 	V_frag6_nfrags -= q6->ip6q_nfrag;
622 	free(q6, M_FTABLE);
623 	V_frag6_nfragpackets--;
624 }
625 
626 /*
627  * Put an ip fragment on a reassembly chain.
628  * Like insque, but pointers in middle of structure.
629  */
630 void
631 frag6_enq(struct ip6asfrag *af6, struct ip6asfrag *up6)
632 {
633 
634 	IP6Q_LOCK_ASSERT();
635 
636 	af6->ip6af_up = up6;
637 	af6->ip6af_down = up6->ip6af_down;
638 	up6->ip6af_down->ip6af_up = af6;
639 	up6->ip6af_down = af6;
640 }
641 
642 /*
643  * To frag6_enq as remque is to insque.
644  */
645 void
646 frag6_deq(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(struct ip6q *new, struct ip6q *old)
657 {
658 
659 	IP6Q_LOCK_ASSERT();
660 
661 	new->ip6q_prev = old;
662 	new->ip6q_next = old->ip6q_next;
663 	old->ip6q_next->ip6q_prev= new;
664 	old->ip6q_next = new;
665 }
666 
667 void
668 frag6_remque(struct ip6q *p6)
669 {
670 
671 	IP6Q_LOCK_ASSERT();
672 
673 	p6->ip6q_prev->ip6q_next = p6->ip6q_next;
674 	p6->ip6q_next->ip6q_prev = p6->ip6q_prev;
675 }
676 
677 /*
678  * IPv6 reassembling timer processing;
679  * if a timer expires on a reassembly
680  * queue, discard it.
681  */
682 void
683 frag6_slowtimo(void)
684 {
685 	struct ip6q *q6;
686 
687 	IP6Q_LOCK();
688 	q6 = V_ip6q.ip6q_next;
689 	if (q6)
690 		while (q6 != &V_ip6q) {
691 			--q6->ip6q_ttl;
692 			q6 = q6->ip6q_next;
693 			if (q6->ip6q_prev->ip6q_ttl == 0) {
694 				V_ip6stat.ip6s_fragtimeout++;
695 				/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
696 				frag6_freef(q6->ip6q_prev);
697 			}
698 		}
699 	/*
700 	 * If we are over the maximum number of fragments
701 	 * (due to the limit being lowered), drain off
702 	 * enough to get down to the new limit.
703 	 */
704 	while (V_frag6_nfragpackets > (u_int)V_ip6_maxfragpackets &&
705 	    V_ip6q.ip6q_prev) {
706 		V_ip6stat.ip6s_fragoverflow++;
707 		/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
708 		frag6_freef(V_ip6q.ip6q_prev);
709 	}
710 	IP6Q_UNLOCK();
711 
712 #if 0
713 	/*
714 	 * Routing changes might produce a better route than we last used;
715 	 * make sure we notice eventually, even if forwarding only for one
716 	 * destination and the cache is never replaced.
717 	 */
718 	if (V_ip6_forward_rt.ro_rt) {
719 		RTFREE(V_ip6_forward_rt.ro_rt);
720 		V_ip6_forward_rt.ro_rt = 0;
721 	}
722 	if (ipsrcchk_rt.ro_rt) {
723 		RTFREE(ipsrcchk_rt.ro_rt);
724 		ipsrcchk_rt.ro_rt = 0;
725 	}
726 #endif
727 }
728 
729 /*
730  * Drain off all datagram fragments.
731  */
732 void
733 frag6_drain(void)
734 {
735 
736 	if (IP6Q_TRYLOCK() == 0)
737 		return;
738 	while (V_ip6q.ip6q_next != &V_ip6q) {
739 		V_ip6stat.ip6s_fragdropped++;
740 		/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
741 		frag6_freef(V_ip6q.ip6q_next);
742 	}
743 	IP6Q_UNLOCK();
744 }
745