1/*
2 * Copyright (c) 2000-2011 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28
29/*	$FreeBSD: src/sys/netinet6/frag6.c,v 1.2.2.5 2001/07/03 11:01:50 ume Exp $	*/
30/*	$KAME: frag6.c,v 1.31 2001/05/17 13:45:34 jinmei Exp $	*/
31
32/*
33 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
34 * All rights reserved.
35 *
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
38 * are met:
39 * 1. Redistributions of source code must retain the above copyright
40 *    notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 *    notice, this list of conditions and the following disclaimer in the
43 *    documentation and/or other materials provided with the distribution.
44 * 3. Neither the name of the project nor the names of its contributors
45 *    may be used to endorse or promote products derived from this software
46 *    without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 */
60
61#include <sys/param.h>
62#include <sys/systm.h>
63#include <sys/malloc.h>
64#include <sys/mcache.h>
65#include <sys/mbuf.h>
66#include <sys/domain.h>
67#include <sys/protosw.h>
68#include <sys/socket.h>
69#include <sys/errno.h>
70#include <sys/time.h>
71#include <sys/kernel.h>
72#include <sys/syslog.h>
73#include <kern/queue.h>
74#include <kern/locks.h>
75
76#include <net/if.h>
77#include <net/route.h>
78
79#include <netinet/in.h>
80#include <netinet/in_var.h>
81#include <netinet/ip.h>
82#include <netinet/ip6.h>
83#include <netinet6/ip6_var.h>
84#include <netinet/icmp6.h>
85
86#include <net/net_osdep.h>
87
88/*
89 * Define it to get a correct behavior on per-interface statistics.
90 * You will need to perform an extra routing table lookup, per fragment,
91 * to do it.  This may, or may not be, a performance hit.
92 */
93#define IN6_IFSTAT_STRICT
94
95static void frag6_enq(struct ip6asfrag *, struct ip6asfrag *);
96static void frag6_deq(struct ip6asfrag *);
97static void frag6_insque(struct ip6q *, struct ip6q *);
98static void frag6_remque(struct ip6q *);
99static void frag6_freef(struct ip6q *);
100
101/* XXX we eventually need splreass6, or some real semaphore */
102int frag6_doing_reass;
103u_int frag6_nfragpackets;
104static u_int frag6_nfrags;
105struct	ip6q ip6q;	/* ip6 reassemble queue */
106
107
108extern lck_mtx_t *inet6_domain_mutex;
109/*
110 * Initialise reassembly queue and fragment identifier.
111 */
112void
113frag6_init()
114{
115	struct timeval tv;
116
117	ip6_maxfragpackets = nmbclusters / 32;
118	ip6_maxfrags = nmbclusters / 4;
119
120	/*
121	 * in many cases, random() here does NOT return random number
122	 * as initialization during bootstrap time occur in fixed order.
123	 */
124	microtime(&tv);
125	ip6_id = random() ^ tv.tv_usec;
126	ip6q.ip6q_next = ip6q.ip6q_prev = &ip6q;
127}
128
129/*
130 * In RFC2460, fragment and reassembly rule do not agree with each other,
131 * in terms of next header field handling in fragment header.
132 * While the sender will use the same value for all of the fragmented packets,
133 * receiver is suggested not to check the consistency.
134 *
135 * fragment rule (p20):
136 *	(2) A Fragment header containing:
137 *	The Next Header value that identifies the first header of
138 *	the Fragmentable Part of the original packet.
139 *		-> next header field is same for all fragments
140 *
141 * reassembly rule (p21):
142 *	The Next Header field of the last header of the Unfragmentable
143 *	Part is obtained from the Next Header field of the first
144 *	fragment's Fragment header.
145 *		-> should grab it from the first fragment only
146 *
147 * The following note also contradicts with fragment rule - noone is going to
148 * send different fragment with different next header field.
149 *
150 * additional note (p22):
151 *	The Next Header values in the Fragment headers of different
152 *	fragments of the same original packet may differ.  Only the value
153 *	from the Offset zero fragment packet is used for reassembly.
154 *		-> should grab it from the first fragment only
155 *
156 * There is no explicit reason given in the RFC.  Historical reason maybe?
157 */
158/*
159 * Fragment input
160 * NOTE: this function is called with the inet6_domain_mutex held from ip6_input.
161 * 	 inet6_domain_mutex is protecting he frag6 queue manipulation.
162 */
163int
164frag6_input(struct mbuf **mp, int *offp, int proto)
165{
166#pragma unused(proto)
167	struct mbuf *m = *mp, *t;
168	struct ip6_hdr *ip6;
169	struct ip6_frag *ip6f;
170	struct ip6q *q6;
171	struct ip6asfrag *af6, *ip6af, *af6dwn;
172	int offset = *offp, nxt, i, next;
173	int first_frag = 0;
174	int fragoff, frgpartlen;	/* must be larger than u_int16_t */
175	struct ifnet *dstifp;
176	struct ifaddr *ifa = NULL;
177	u_int8_t ecn, ecn0;
178
179#ifdef IN6_IFSTAT_STRICT
180	struct route_in6 ro;
181	struct sockaddr_in6 *dst;
182#endif
183
184	/* Expect 32-bit aligned data pointer on strict-align platforms */
185	MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
186
187	ip6 = mtod(m, struct ip6_hdr *);
188#ifndef PULLDOWN_TEST
189	IP6_EXTHDR_CHECK(m, offset, sizeof(struct ip6_frag), return IPPROTO_DONE);
190	ip6f = (struct ip6_frag *)((caddr_t)ip6 + offset);
191#else
192	IP6_EXTHDR_GET(ip6f, struct ip6_frag *, m, offset, sizeof(*ip6f));
193	if (ip6f == NULL)
194		return IPPROTO_DONE;
195#endif
196
197	dstifp = NULL;
198#ifdef IN6_IFSTAT_STRICT
199	/* find the destination interface of the packet. */
200	bzero(&ro, sizeof (ro));
201	dst = (struct sockaddr_in6 *)&ro.ro_dst;
202	dst->sin6_family = AF_INET6;
203	dst->sin6_len = sizeof (struct sockaddr_in6);
204	dst->sin6_addr = ip6->ip6_dst;
205
206	rtalloc((struct route *)&ro);
207	if (ro.ro_rt != NULL) {
208		RT_LOCK(ro.ro_rt);
209		if ((ifa = ro.ro_rt->rt_ifa) != NULL) {
210			IFA_ADDREF(ifa);
211			dstifp = ((struct in6_ifaddr *)ro.ro_rt->rt_ifa)->ia_ifp;
212		}
213		RT_UNLOCK(ro.ro_rt);
214		rtfree(ro.ro_rt);
215		ro.ro_rt = NULL;
216	}
217#else
218	/* we are violating the spec, this is not the destination interface */
219	if ((m->m_flags & M_PKTHDR) != 0)
220		dstifp = m->m_pkthdr.rcvif;
221#endif
222
223	/* jumbo payload can't contain a fragment header */
224	if (ip6->ip6_plen == 0) {
225		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, offset);
226		in6_ifstat_inc(dstifp, ifs6_reass_fail);
227		if (ifa != NULL)
228			IFA_REMREF(ifa);
229		return IPPROTO_DONE;
230	}
231
232	/*
233	 * check whether fragment packet's fragment length is
234	 * multiple of 8 octets.
235	 * sizeof(struct ip6_frag) == 8
236	 * sizeof(struct ip6_hdr) = 40
237	 */
238	if ((ip6f->ip6f_offlg & IP6F_MORE_FRAG) &&
239	    (((ntohs(ip6->ip6_plen) - offset) & 0x7) != 0)) {
240		icmp6_error(m, ICMP6_PARAM_PROB,
241			    ICMP6_PARAMPROB_HEADER,
242			    offsetof(struct ip6_hdr, ip6_plen));
243		in6_ifstat_inc(dstifp, ifs6_reass_fail);
244		if (ifa != NULL)
245			IFA_REMREF(ifa);
246		return IPPROTO_DONE;
247	}
248
249	ip6stat.ip6s_fragments++;
250	in6_ifstat_inc(dstifp, ifs6_reass_reqd);
251
252	/* offset now points to data portion */
253	offset += sizeof(struct ip6_frag);
254
255	frag6_doing_reass = 1;
256
257	/*
258	 * Enforce upper bound on number of fragments.
259	 * If maxfrag is 0, never accept fragments.
260	 * If maxfrag is -1, accept all fragments without limitation.
261	 */
262	if (ip6_maxfrags < 0)
263		;
264	else if (frag6_nfrags >= (u_int)ip6_maxfrags)
265		goto dropfrag;
266
267	for (q6 = ip6q.ip6q_next; q6 != &ip6q; q6 = q6->ip6q_next)
268		if (ip6f->ip6f_ident == q6->ip6q_ident &&
269		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &q6->ip6q_src) &&
270		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &q6->ip6q_dst))
271			break;
272
273	if (q6 == &ip6q) {
274		/*
275		 * the first fragment to arrive, create a reassembly queue.
276		 */
277		first_frag = 1;
278
279		/*
280		 * Enforce upper bound on number of fragmented packets
281		 * for which we attempt reassembly;
282		 * If maxfrag is 0, never accept fragments.
283		 * If maxfrag is -1, accept all fragments without limitation.
284		 */
285		if (ip6_maxfragpackets < 0)
286			;
287		else if (frag6_nfragpackets >= (u_int)ip6_maxfragpackets)
288			goto dropfrag;
289		frag6_nfragpackets++;
290		q6 = (struct ip6q *)_MALLOC(sizeof(struct ip6q), M_FTABLE,
291			M_DONTWAIT);
292		if (q6 == NULL)
293			goto dropfrag;
294		bzero(q6, sizeof(*q6));
295
296		frag6_insque(q6, &ip6q);
297
298		/* ip6q_nxt will be filled afterwards, from 1st fragment */
299		q6->ip6q_down	= q6->ip6q_up = (struct ip6asfrag *)q6;
300#ifdef notyet
301		q6->ip6q_nxtp	= (u_char *)nxtp;
302#endif
303		q6->ip6q_ident	= ip6f->ip6f_ident;
304		q6->ip6q_ttl 	= IPV6_FRAGTTL;
305		q6->ip6q_src	= ip6->ip6_src;
306		q6->ip6q_dst	= ip6->ip6_dst;
307		q6->ip6q_ecn	=
308		    (ntohl(ip6->ip6_flow) >> 20) & IPTOS_ECN_MASK;
309		q6->ip6q_unfrglen = -1;	/* The 1st fragment has not arrived. */
310
311		q6->ip6q_nfrag	= 0;
312	}
313
314	/*
315	 * If it's the 1st fragment, record the length of the
316	 * unfragmentable part and the next header of the fragment header.
317	 */
318	fragoff = ntohs(ip6f->ip6f_offlg & IP6F_OFF_MASK);
319	if (fragoff == 0) {
320		q6->ip6q_unfrglen = offset - sizeof(struct ip6_hdr)
321			- sizeof(struct ip6_frag);
322		q6->ip6q_nxt = ip6f->ip6f_nxt;
323	}
324
325	/*
326	 * Check that the reassembled packet would not exceed 65535 bytes
327	 * in size.
328	 * If it would exceed, discard the fragment and return an ICMP error.
329	 */
330	frgpartlen = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen) - offset;
331	if (q6->ip6q_unfrglen >= 0) {
332		/* The 1st fragment has already arrived. */
333		if (q6->ip6q_unfrglen + fragoff + frgpartlen > IPV6_MAXPACKET) {
334			icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
335				    offset - sizeof(struct ip6_frag) +
336					offsetof(struct ip6_frag, ip6f_offlg));
337			frag6_doing_reass = 0;
338			if (ifa != NULL)
339				IFA_REMREF(ifa);
340			return(IPPROTO_DONE);
341		}
342	}
343	else if (fragoff + frgpartlen > IPV6_MAXPACKET) {
344		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
345			    offset - sizeof(struct ip6_frag) +
346				offsetof(struct ip6_frag, ip6f_offlg));
347		frag6_doing_reass = 0;
348		if (ifa != NULL)
349			IFA_REMREF(ifa);
350		return(IPPROTO_DONE);
351	}
352	/*
353	 * If it's the first fragment, do the above check for each
354	 * fragment already stored in the reassembly queue.
355	 */
356	if (fragoff == 0) {
357		for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
358		     af6 = af6dwn) {
359			af6dwn = af6->ip6af_down;
360
361			if (q6->ip6q_unfrglen + af6->ip6af_off + af6->ip6af_frglen >
362			    IPV6_MAXPACKET) {
363				struct mbuf *merr = IP6_REASS_MBUF(af6);
364				struct ip6_hdr *ip6err;
365				int erroff = af6->ip6af_offset;
366
367				/* dequeue the fragment. */
368				frag6_deq(af6);
369				FREE(af6, M_FTABLE);
370
371				/* adjust pointer. */
372				ip6err = mtod(merr, struct ip6_hdr *);
373
374				/*
375				 * Restore source and destination addresses
376				 * in the erroneous IPv6 header.
377				 */
378				ip6err->ip6_src = q6->ip6q_src;
379				ip6err->ip6_dst = q6->ip6q_dst;
380
381				icmp6_error(merr, ICMP6_PARAM_PROB,
382					    ICMP6_PARAMPROB_HEADER,
383					    erroff - sizeof(struct ip6_frag) +
384						offsetof(struct ip6_frag, ip6f_offlg));
385			}
386		}
387	}
388
389	ip6af = (struct ip6asfrag *)_MALLOC(sizeof(struct ip6asfrag), M_FTABLE,
390	    M_DONTWAIT);
391	if (ip6af == NULL)
392		goto dropfrag;
393	bzero(ip6af, sizeof(*ip6af));
394	ip6af->ip6af_mff = ip6f->ip6f_offlg & IP6F_MORE_FRAG;
395	ip6af->ip6af_off = fragoff;
396	ip6af->ip6af_frglen = frgpartlen;
397	ip6af->ip6af_offset = offset;
398	IP6_REASS_MBUF(ip6af) = m;
399
400	if (first_frag) {
401		af6 = (struct ip6asfrag *)q6;
402		goto insert;
403	}
404
405	/*
406	 * Handle ECN by comparing this segment with the first one;
407	 * if CE is set, do not lose CE.
408	 * drop if CE and not-ECT are mixed for the same packet.
409	 */
410	ecn = (ntohl(ip6->ip6_flow) >> 20) & IPTOS_ECN_MASK;
411	ecn0 = q6->ip6q_ecn;
412	if (ecn == IPTOS_ECN_CE) {
413		if (ecn0 == IPTOS_ECN_NOTECT) {
414			FREE(ip6af, M_FTABLE);
415			goto dropfrag;
416		}
417		if (ecn0 != IPTOS_ECN_CE)
418			q6->ip6q_ecn = IPTOS_ECN_CE;
419	}
420	if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) {
421		FREE(ip6af, M_FTABLE);
422		goto dropfrag;
423	}
424
425	/*
426	 * Find a segment which begins after this one does.
427	 */
428	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
429	     af6 = af6->ip6af_down)
430		if (af6->ip6af_off > ip6af->ip6af_off)
431			break;
432
433#if 0
434	/*
435	 * If there is a preceding segment, it may provide some of
436	 * our data already.  If so, drop the data from the incoming
437	 * segment.  If it provides all of our data, drop us.
438	 */
439	if (af6->ip6af_up != (struct ip6asfrag *)q6) {
440		i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
441			- ip6af->ip6af_off;
442		if (i > 0) {
443			if (i >= ip6af->ip6af_frglen)
444				goto dropfrag;
445			m_adj(IP6_REASS_MBUF(ip6af), i);
446			ip6af->ip6af_off += i;
447			ip6af->ip6af_frglen -= i;
448		}
449	}
450
451	/*
452	 * While we overlap succeeding segments trim them or,
453	 * if they are completely covered, dequeue them.
454	 */
455	while (af6 != (struct ip6asfrag *)q6 &&
456	       ip6af->ip6af_off + ip6af->ip6af_frglen > af6->ip6af_off) {
457		i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
458		if (i < af6->ip6af_frglen) {
459			af6->ip6af_frglen -= i;
460			af6->ip6af_off += i;
461			m_adj(IP6_REASS_MBUF(af6), i);
462			break;
463		}
464		af6 = af6->ip6af_down;
465		m_freem(IP6_REASS_MBUF(af6->ip6af_up));
466		frag6_deq(af6->ip6af_up);
467	}
468#else
469	/*
470	 * If the incoming framgent overlaps some existing fragments in
471	 * the reassembly queue, drop it, since it is dangerous to override
472	 * existing fragments from a security point of view.
473	 * We don't know which fragment is the bad guy - here we trust
474	 * fragment that came in earlier, with no real reason.
475	 *
476	 * Note: due to changes after disabling this part, mbuf passed to
477	 * m_adj() below now does not meet the requirement.
478	 */
479	if (af6->ip6af_up != (struct ip6asfrag *)q6) {
480		i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
481			- ip6af->ip6af_off;
482		if (i > 0) {
483#if 0				/* suppress the noisy log */
484			log(LOG_ERR, "%d bytes of a fragment from %s "
485			    "overlaps the previous fragment\n",
486			    i, ip6_sprintf(&q6->ip6q_src));
487#endif
488			FREE(ip6af, M_FTABLE);
489			goto dropfrag;
490		}
491	}
492	if (af6 != (struct ip6asfrag *)q6) {
493		i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
494		if (i > 0) {
495#if 0				/* suppress the noisy log */
496			log(LOG_ERR, "%d bytes of a fragment from %s "
497			    "overlaps the succeeding fragment",
498			    i, ip6_sprintf(&q6->ip6q_src));
499#endif
500			FREE(ip6af, M_FTABLE);
501			goto dropfrag;
502		}
503	}
504#endif
505
506insert:
507
508	/*
509	 * Stick new segment in its place;
510	 * check for complete reassembly.
511	 * Move to front of packet queue, as we are
512	 * the most recently active fragmented packet.
513	 */
514	frag6_enq(ip6af, af6->ip6af_up);
515	frag6_nfrags++;
516	q6->ip6q_nfrag++;
517#if 0 /* xxx */
518	if (q6 != ip6q.ip6q_next) {
519		frag6_remque(q6);
520		frag6_insque(q6, &ip6q);
521	}
522#endif
523	next = 0;
524	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
525	     af6 = af6->ip6af_down) {
526		if (af6->ip6af_off != next) {
527			frag6_doing_reass = 0;
528			if (ifa != NULL)
529				IFA_REMREF(ifa);
530			return IPPROTO_DONE;
531		}
532		next += af6->ip6af_frglen;
533	}
534	if (af6->ip6af_up->ip6af_mff) {
535		frag6_doing_reass = 0;
536		if (ifa != NULL)
537			IFA_REMREF(ifa);
538		return IPPROTO_DONE;
539	}
540
541	/*
542	 * Reassembly is complete; concatenate fragments.
543	 */
544	ip6af = q6->ip6q_down;
545	t = m = IP6_REASS_MBUF(ip6af);
546	af6 = ip6af->ip6af_down;
547	frag6_deq(ip6af);
548	while (af6 != (struct ip6asfrag *)q6) {
549		af6dwn = af6->ip6af_down;
550		frag6_deq(af6);
551		while (t->m_next)
552			t = t->m_next;
553		t->m_next = IP6_REASS_MBUF(af6);
554		m_adj(t->m_next, af6->ip6af_offset);
555		FREE(af6, M_FTABLE);
556		af6 = af6dwn;
557	}
558
559	/* adjust offset to point where the original next header starts */
560	offset = ip6af->ip6af_offset - sizeof(struct ip6_frag);
561	FREE(ip6af, M_FTABLE);
562	ip6 = mtod(m, struct ip6_hdr *);
563	ip6->ip6_plen = htons((u_short)next + offset - sizeof(struct ip6_hdr));
564	ip6->ip6_src = q6->ip6q_src;
565	ip6->ip6_dst = q6->ip6q_dst;
566	if (q6->ip6q_ecn == IPTOS_ECN_CE)
567		ip6->ip6_flow |= htonl(IPTOS_ECN_CE << 20);
568
569	nxt = q6->ip6q_nxt;
570#if notyet
571	*q6->ip6q_nxtp = (u_char)(nxt & 0xff);
572#endif
573
574	/* Delete frag6 header */
575	if (m->m_len >= offset + sizeof(struct ip6_frag)) {
576		/* This is the only possible case with !PULLDOWN_TEST */
577		ovbcopy((caddr_t)ip6, (caddr_t)ip6 + sizeof(struct ip6_frag),
578			offset);
579		m->m_data += sizeof(struct ip6_frag);
580		m->m_len -= sizeof(struct ip6_frag);
581	} else {
582		/* this comes with no copy if the boundary is on cluster */
583		if ((t = m_split(m, offset, M_DONTWAIT)) == NULL) {
584			frag6_remque(q6);
585			frag6_nfrags -= q6->ip6q_nfrag;
586			FREE(q6, M_FTABLE);
587			frag6_nfragpackets--;
588			goto dropfrag;
589		}
590		m_adj(t, sizeof(struct ip6_frag));
591		m_cat(m, t);
592	}
593
594	/*
595	 * Store NXT to the original.
596	 */
597	{
598		char *prvnxtp = ip6_get_prevhdr(m, offset); /* XXX */
599		*prvnxtp = nxt;
600	}
601
602	frag6_remque(q6);
603	frag6_nfrags -= q6->ip6q_nfrag;
604	FREE(q6, M_FTABLE);
605	frag6_nfragpackets--;
606
607	if (m->m_flags & M_PKTHDR) { /* Isn't it always true? */
608		int plen = 0;
609		for (t = m; t; t = t->m_next)
610			plen += t->m_len;
611		m->m_pkthdr.len = plen;
612	}
613
614	ip6stat.ip6s_reassembled++;
615	in6_ifstat_inc(dstifp, ifs6_reass_ok);
616
617	/*
618	 * Tell launch routine the next header
619	 */
620
621	*mp = m;
622	*offp = offset;
623
624	frag6_doing_reass = 0;
625	if (ifa != NULL)
626		IFA_REMREF(ifa);
627	return nxt;
628
629 dropfrag:
630	in6_ifstat_inc(dstifp, ifs6_reass_fail);
631	ip6stat.ip6s_fragdropped++;
632	m_freem(m);
633	frag6_doing_reass = 0;
634	if (ifa != NULL)
635		IFA_REMREF(ifa);
636	return IPPROTO_DONE;
637}
638
639/*
640 * Free a fragment reassembly header and all
641 * associated datagrams.
642 */
643void
644frag6_freef(q6)
645	struct ip6q *q6;
646{
647	struct ip6asfrag *af6, *down6;
648
649	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
650	     af6 = down6) {
651		struct mbuf *m = IP6_REASS_MBUF(af6);
652
653		down6 = af6->ip6af_down;
654		frag6_deq(af6);
655
656		/*
657		 * Return ICMP time exceeded error for the 1st fragment.
658		 * Just free other fragments.
659		 */
660		if (af6->ip6af_off == 0) {
661			struct ip6_hdr *ip6;
662
663			/* adjust pointer */
664			ip6 = mtod(m, struct ip6_hdr *);
665
666			/* restore source and destination addresses */
667			ip6->ip6_src = q6->ip6q_src;
668			ip6->ip6_dst = q6->ip6q_dst;
669			icmp6_error(m, ICMP6_TIME_EXCEEDED,
670				    ICMP6_TIME_EXCEED_REASSEMBLY, 0);
671		} else
672			m_freem(m);
673		FREE(af6, M_FTABLE);
674
675	}
676	frag6_remque(q6);
677	frag6_nfrags -= q6->ip6q_nfrag;
678	FREE(q6, M_FTABLE);
679	frag6_nfragpackets--;
680}
681
682/*
683 * Put an ip fragment on a reassembly chain.
684 * Like insque, but pointers in middle of structure.
685 */
686void
687frag6_enq(af6, up6)
688	struct ip6asfrag *af6, *up6;
689{
690	af6->ip6af_up = up6;
691	af6->ip6af_down = up6->ip6af_down;
692	up6->ip6af_down->ip6af_up = af6;
693	up6->ip6af_down = af6;
694}
695
696/*
697 * To frag6_enq as remque is to insque.
698 */
699void
700frag6_deq(af6)
701	struct ip6asfrag *af6;
702{
703	af6->ip6af_up->ip6af_down = af6->ip6af_down;
704	af6->ip6af_down->ip6af_up = af6->ip6af_up;
705}
706
707void
708frag6_insque(new, old)
709	struct ip6q *new, *old;
710{
711	new->ip6q_prev = old;
712	new->ip6q_next = old->ip6q_next;
713	old->ip6q_next->ip6q_prev= new;
714	old->ip6q_next = new;
715}
716
717void
718frag6_remque(p6)
719	struct ip6q *p6;
720{
721	p6->ip6q_prev->ip6q_next = p6->ip6q_next;
722	p6->ip6q_next->ip6q_prev = p6->ip6q_prev;
723}
724
725/*
726 * IPv6 reassembling timer processing;
727 * if a timer expires on a reassembly
728 * queue, discard it.
729 */
730void
731frag6_slowtimo()
732{
733	struct ip6q *q6;
734	lck_mtx_lock(inet6_domain_mutex);
735
736	frag6_doing_reass = 1;
737	q6 = ip6q.ip6q_next;
738	if (q6)
739		while (q6 != &ip6q) {
740			--q6->ip6q_ttl;
741			q6 = q6->ip6q_next;
742			if (q6->ip6q_prev->ip6q_ttl == 0) {
743				ip6stat.ip6s_fragtimeout++;
744				/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
745				frag6_freef(q6->ip6q_prev);
746			}
747		}
748	/*
749	 * If we are over the maximum number of fragments
750	 * (due to the limit being lowered), drain off
751	 * enough to get down to the new limit.
752	 */
753	while (frag6_nfragpackets > (u_int)ip6_maxfragpackets &&
754	    ip6q.ip6q_prev) {
755		ip6stat.ip6s_fragoverflow++;
756		/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
757		frag6_freef(ip6q.ip6q_prev);
758	}
759	frag6_doing_reass = 0;
760	lck_mtx_unlock(inet6_domain_mutex);
761}
762
763/*
764 * Drain off all datagram fragments.
765 */
766void
767frag6_drain()
768{
769	if (frag6_doing_reass)
770		return;
771	lck_mtx_lock(inet6_domain_mutex);
772	while (ip6q.ip6q_next != &ip6q) {
773		ip6stat.ip6s_fragdropped++;
774		/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
775		frag6_freef(ip6q.ip6q_next);
776	}
777	lck_mtx_unlock(inet6_domain_mutex);
778}
779