frag6.c revision 238248
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: head/sys/netinet6/frag6.c 238248 2012-07-08 15:30:24Z bz $");
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
47#include <net/if.h>
48#include <net/route.h>
49#include <net/vnet.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#include <security/mac/mac_framework.h>
60
61/*
62 * Define it to get a correct behavior on per-interface statistics.
63 * You will need to perform an extra routing table lookup, per fragment,
64 * to do it.  This may, or may not be, a performance hit.
65 */
66#define IN6_IFSTAT_STRICT
67
68static void frag6_enq(struct ip6asfrag *, struct ip6asfrag *);
69static void frag6_deq(struct ip6asfrag *);
70static void frag6_insque(struct ip6q *, struct ip6q *);
71static void frag6_remque(struct ip6q *);
72static void frag6_freef(struct ip6q *);
73
74static struct mtx ip6qlock;
75/*
76 * These fields all protected by ip6qlock.
77 */
78static VNET_DEFINE(u_int, frag6_nfragpackets);
79static VNET_DEFINE(u_int, frag6_nfrags);
80static VNET_DEFINE(struct ip6q, ip6q);	/* ip6 reassemble queue */
81
82#define	V_frag6_nfragpackets		VNET(frag6_nfragpackets)
83#define	V_frag6_nfrags			VNET(frag6_nfrags)
84#define	V_ip6q				VNET(ip6q)
85
86#define	IP6Q_LOCK_INIT()	mtx_init(&ip6qlock, "ip6qlock", NULL, MTX_DEF);
87#define	IP6Q_LOCK()		mtx_lock(&ip6qlock)
88#define	IP6Q_TRYLOCK()		mtx_trylock(&ip6qlock)
89#define	IP6Q_LOCK_ASSERT()	mtx_assert(&ip6qlock, MA_OWNED)
90#define	IP6Q_UNLOCK()		mtx_unlock(&ip6qlock)
91
92static MALLOC_DEFINE(M_FTABLE, "fragment", "fragment reassembly header");
93
94/*
95 * Initialise reassembly queue and fragment identifier.
96 */
97static void
98frag6_change(void *tag)
99{
100
101	V_ip6_maxfragpackets = nmbclusters / 4;
102	V_ip6_maxfrags = nmbclusters / 4;
103}
104
105void
106frag6_init(void)
107{
108
109	V_ip6_maxfragpackets = nmbclusters / 4;
110	V_ip6_maxfrags = nmbclusters / 4;
111	V_ip6q.ip6q_next = V_ip6q.ip6q_prev = &V_ip6q;
112
113	if (!IS_DEFAULT_VNET(curvnet))
114		return;
115
116	EVENTHANDLER_REGISTER(nmbclusters_change,
117	    frag6_change, NULL, EVENTHANDLER_PRI_ANY);
118
119	IP6Q_LOCK_INIT();
120}
121
122/*
123 * In RFC2460, fragment and reassembly rule do not agree with each other,
124 * in terms of next header field handling in fragment header.
125 * While the sender will use the same value for all of the fragmented packets,
126 * receiver is suggested not to check the consistency.
127 *
128 * fragment rule (p20):
129 *	(2) A Fragment header containing:
130 *	The Next Header value that identifies the first header of
131 *	the Fragmentable Part of the original packet.
132 *		-> next header field is same for all fragments
133 *
134 * reassembly rule (p21):
135 *	The Next Header field of the last header of the Unfragmentable
136 *	Part is obtained from the Next Header field of the first
137 *	fragment's Fragment header.
138 *		-> should grab it from the first fragment only
139 *
140 * The following note also contradicts with fragment rule - noone is going to
141 * send different fragment with different next header field.
142 *
143 * additional note (p22):
144 *	The Next Header values in the Fragment headers of different
145 *	fragments of the same original packet may differ.  Only the value
146 *	from the Offset zero fragment packet is used for reassembly.
147 *		-> should grab it from the first fragment only
148 *
149 * There is no explicit reason given in the RFC.  Historical reason maybe?
150 */
151/*
152 * Fragment input
153 */
154int
155frag6_input(struct mbuf **mp, int *offp, int proto)
156{
157	struct mbuf *m = *mp, *t;
158	struct ip6_hdr *ip6;
159	struct ip6_frag *ip6f;
160	struct ip6q *q6;
161	struct ip6asfrag *af6, *ip6af, *af6dwn;
162#ifdef IN6_IFSTAT_STRICT
163	struct in6_ifaddr *ia;
164#endif
165	int offset = *offp, nxt, i, next;
166	int first_frag = 0;
167	int fragoff, frgpartlen;	/* must be larger than u_int16_t */
168	struct ifnet *dstifp;
169	u_int8_t ecn, ecn0;
170#if 0
171	char ip6buf[INET6_ADDRSTRLEN];
172#endif
173
174	ip6 = mtod(m, struct ip6_hdr *);
175#ifndef PULLDOWN_TEST
176	IP6_EXTHDR_CHECK(m, offset, sizeof(struct ip6_frag), IPPROTO_DONE);
177	ip6f = (struct ip6_frag *)((caddr_t)ip6 + offset);
178#else
179	IP6_EXTHDR_GET(ip6f, struct ip6_frag *, m, offset, sizeof(*ip6f));
180	if (ip6f == NULL)
181		return (IPPROTO_DONE);
182#endif
183
184	dstifp = NULL;
185#ifdef IN6_IFSTAT_STRICT
186	/* find the destination interface of the packet. */
187	if ((ia = ip6_getdstifaddr(m)) != NULL) {
188		dstifp = ia->ia_ifp;
189		ifa_free(&ia->ia_ifa);
190	}
191#else
192	/* we are violating the spec, this is not the destination interface */
193	if ((m->m_flags & M_PKTHDR) != 0)
194		dstifp = m->m_pkthdr.rcvif;
195#endif
196
197	/* jumbo payload can't contain a fragment header */
198	if (ip6->ip6_plen == 0) {
199		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, offset);
200		in6_ifstat_inc(dstifp, ifs6_reass_fail);
201		return IPPROTO_DONE;
202	}
203
204	/*
205	 * check whether fragment packet's fragment length is
206	 * multiple of 8 octets.
207	 * sizeof(struct ip6_frag) == 8
208	 * sizeof(struct ip6_hdr) = 40
209	 */
210	if ((ip6f->ip6f_offlg & IP6F_MORE_FRAG) &&
211	    (((ntohs(ip6->ip6_plen) - offset) & 0x7) != 0)) {
212		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
213		    offsetof(struct ip6_hdr, ip6_plen));
214		in6_ifstat_inc(dstifp, ifs6_reass_fail);
215		return IPPROTO_DONE;
216	}
217
218	V_ip6stat.ip6s_fragments++;
219	in6_ifstat_inc(dstifp, ifs6_reass_reqd);
220
221	/* offset now points to data portion */
222	offset += sizeof(struct ip6_frag);
223
224	/*
225	 * XXX-BZ RFC XXXX (draft-gont-6man-ipv6-atomic-fragments)
226	 * Handle "atomic" fragments (offset and m bit set to 0) upfront,
227	 * unrelated to any reassembly.  Just skip the fragment header.
228	 */
229	if ((ip6f->ip6f_offlg & ~IP6F_RESERVED_MASK) == 0) {
230		/* XXX-BZ we want dedicated counters for this. */
231		V_ip6stat.ip6s_reassembled++;
232		in6_ifstat_inc(dstifp, ifs6_reass_ok);
233		*offp = offset;
234		return (ip6f->ip6f_nxt);
235	}
236
237	IP6Q_LOCK();
238
239	/*
240	 * Enforce upper bound on number of fragments.
241	 * If maxfrag is 0, never accept fragments.
242	 * If maxfrag is -1, accept all fragments without limitation.
243	 */
244	if (V_ip6_maxfrags < 0)
245		;
246	else if (V_frag6_nfrags >= (u_int)V_ip6_maxfrags)
247		goto dropfrag;
248
249	for (q6 = V_ip6q.ip6q_next; q6 != &V_ip6q; q6 = q6->ip6q_next)
250		if (ip6f->ip6f_ident == q6->ip6q_ident &&
251		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &q6->ip6q_src) &&
252		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &q6->ip6q_dst)
253#ifdef MAC
254		    && mac_ip6q_match(m, q6)
255#endif
256		    )
257			break;
258
259	if (q6 == &V_ip6q) {
260		/*
261		 * the first fragment to arrive, create a reassembly queue.
262		 */
263		first_frag = 1;
264
265		/*
266		 * Enforce upper bound on number of fragmented packets
267		 * for which we attempt reassembly;
268		 * If maxfragpackets is 0, never accept fragments.
269		 * If maxfragpackets is -1, accept all fragments without
270		 * limitation.
271		 */
272		if (V_ip6_maxfragpackets < 0)
273			;
274		else if (V_frag6_nfragpackets >= (u_int)V_ip6_maxfragpackets)
275			goto dropfrag;
276		V_frag6_nfragpackets++;
277		q6 = (struct ip6q *)malloc(sizeof(struct ip6q), M_FTABLE,
278		    M_NOWAIT);
279		if (q6 == NULL)
280			goto dropfrag;
281		bzero(q6, sizeof(*q6));
282#ifdef MAC
283		if (mac_ip6q_init(q6, M_NOWAIT) != 0) {
284			free(q6, M_FTABLE);
285			goto dropfrag;
286		}
287		mac_ip6q_create(m, q6);
288#endif
289		frag6_insque(q6, &V_ip6q);
290
291		/* ip6q_nxt will be filled afterwards, from 1st fragment */
292		q6->ip6q_down	= q6->ip6q_up = (struct ip6asfrag *)q6;
293#ifdef notyet
294		q6->ip6q_nxtp	= (u_char *)nxtp;
295#endif
296		q6->ip6q_ident	= ip6f->ip6f_ident;
297		q6->ip6q_ttl	= IPV6_FRAGTTL;
298		q6->ip6q_src	= ip6->ip6_src;
299		q6->ip6q_dst	= ip6->ip6_dst;
300		q6->ip6q_ecn	=
301		    (ntohl(ip6->ip6_flow) >> 20) & IPTOS_ECN_MASK;
302		q6->ip6q_unfrglen = -1;	/* The 1st fragment has not arrived. */
303
304		q6->ip6q_nfrag = 0;
305	}
306
307	/*
308	 * If it's the 1st fragment, record the length of the
309	 * unfragmentable part and the next header of the fragment header.
310	 */
311	fragoff = ntohs(ip6f->ip6f_offlg & IP6F_OFF_MASK);
312	if (fragoff == 0) {
313		q6->ip6q_unfrglen = offset - sizeof(struct ip6_hdr) -
314		    sizeof(struct ip6_frag);
315		q6->ip6q_nxt = ip6f->ip6f_nxt;
316	}
317
318	/*
319	 * Check that the reassembled packet would not exceed 65535 bytes
320	 * in size.
321	 * If it would exceed, discard the fragment and return an ICMP error.
322	 */
323	frgpartlen = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen) - offset;
324	if (q6->ip6q_unfrglen >= 0) {
325		/* The 1st fragment has already arrived. */
326		if (q6->ip6q_unfrglen + fragoff + frgpartlen > IPV6_MAXPACKET) {
327			icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
328			    offset - sizeof(struct ip6_frag) +
329			    offsetof(struct ip6_frag, ip6f_offlg));
330			IP6Q_UNLOCK();
331			return (IPPROTO_DONE);
332		}
333	} else if (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		IP6Q_UNLOCK();
338		return (IPPROTO_DONE);
339	}
340	/*
341	 * If it's the first fragment, do the above check for each
342	 * fragment already stored in the reassembly queue.
343	 */
344	if (fragoff == 0) {
345		for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
346		     af6 = af6dwn) {
347			af6dwn = af6->ip6af_down;
348
349			if (q6->ip6q_unfrglen + af6->ip6af_off + af6->ip6af_frglen >
350			    IPV6_MAXPACKET) {
351				struct mbuf *merr = IP6_REASS_MBUF(af6);
352				struct ip6_hdr *ip6err;
353				int erroff = af6->ip6af_offset;
354
355				/* dequeue the fragment. */
356				frag6_deq(af6);
357				free(af6, M_FTABLE);
358
359				/* adjust pointer. */
360				ip6err = mtod(merr, struct ip6_hdr *);
361
362				/*
363				 * Restore source and destination addresses
364				 * in the erroneous IPv6 header.
365				 */
366				ip6err->ip6_src = q6->ip6q_src;
367				ip6err->ip6_dst = q6->ip6q_dst;
368
369				icmp6_error(merr, ICMP6_PARAM_PROB,
370				    ICMP6_PARAMPROB_HEADER,
371				    erroff - sizeof(struct ip6_frag) +
372				    offsetof(struct ip6_frag, ip6f_offlg));
373			}
374		}
375	}
376
377	ip6af = (struct ip6asfrag *)malloc(sizeof(struct ip6asfrag), M_FTABLE,
378	    M_NOWAIT);
379	if (ip6af == NULL)
380		goto dropfrag;
381	bzero(ip6af, sizeof(*ip6af));
382	ip6af->ip6af_mff = ip6f->ip6f_offlg & IP6F_MORE_FRAG;
383	ip6af->ip6af_off = fragoff;
384	ip6af->ip6af_frglen = frgpartlen;
385	ip6af->ip6af_offset = offset;
386	IP6_REASS_MBUF(ip6af) = m;
387
388	if (first_frag) {
389		af6 = (struct ip6asfrag *)q6;
390		goto insert;
391	}
392
393	/*
394	 * Handle ECN by comparing this segment with the first one;
395	 * if CE is set, do not lose CE.
396	 * drop if CE and not-ECT are mixed for the same packet.
397	 */
398	ecn = (ntohl(ip6->ip6_flow) >> 20) & IPTOS_ECN_MASK;
399	ecn0 = q6->ip6q_ecn;
400	if (ecn == IPTOS_ECN_CE) {
401		if (ecn0 == IPTOS_ECN_NOTECT) {
402			free(ip6af, M_FTABLE);
403			goto dropfrag;
404		}
405		if (ecn0 != IPTOS_ECN_CE)
406			q6->ip6q_ecn = IPTOS_ECN_CE;
407	}
408	if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) {
409		free(ip6af, M_FTABLE);
410		goto dropfrag;
411	}
412
413	/*
414	 * Find a segment which begins after this one does.
415	 */
416	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
417	     af6 = af6->ip6af_down)
418		if (af6->ip6af_off > ip6af->ip6af_off)
419			break;
420
421#if 0
422	/*
423	 * If there is a preceding segment, it may provide some of
424	 * our data already.  If so, drop the data from the incoming
425	 * segment.  If it provides all of our data, drop us.
426	 */
427	if (af6->ip6af_up != (struct ip6asfrag *)q6) {
428		i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
429			- ip6af->ip6af_off;
430		if (i > 0) {
431			if (i >= ip6af->ip6af_frglen)
432				goto dropfrag;
433			m_adj(IP6_REASS_MBUF(ip6af), i);
434			ip6af->ip6af_off += i;
435			ip6af->ip6af_frglen -= i;
436		}
437	}
438
439	/*
440	 * While we overlap succeeding segments trim them or,
441	 * if they are completely covered, dequeue them.
442	 */
443	while (af6 != (struct ip6asfrag *)q6 &&
444	       ip6af->ip6af_off + ip6af->ip6af_frglen > af6->ip6af_off) {
445		i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
446		if (i < af6->ip6af_frglen) {
447			af6->ip6af_frglen -= i;
448			af6->ip6af_off += i;
449			m_adj(IP6_REASS_MBUF(af6), i);
450			break;
451		}
452		af6 = af6->ip6af_down;
453		m_freem(IP6_REASS_MBUF(af6->ip6af_up));
454		frag6_deq(af6->ip6af_up);
455	}
456#else
457	/*
458	 * If the incoming framgent overlaps some existing fragments in
459	 * the reassembly queue, drop it, since it is dangerous to override
460	 * existing fragments from a security point of view.
461	 * We don't know which fragment is the bad guy - here we trust
462	 * fragment that came in earlier, with no real reason.
463	 *
464	 * Note: due to changes after disabling this part, mbuf passed to
465	 * m_adj() below now does not meet the requirement.
466	 */
467	if (af6->ip6af_up != (struct ip6asfrag *)q6) {
468		i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
469			- ip6af->ip6af_off;
470		if (i > 0) {
471#if 0				/* suppress the noisy log */
472			log(LOG_ERR, "%d bytes of a fragment from %s "
473			    "overlaps the previous fragment\n",
474			    i, ip6_sprintf(ip6buf, &q6->ip6q_src));
475#endif
476			free(ip6af, M_FTABLE);
477			goto dropfrag;
478		}
479	}
480	if (af6 != (struct ip6asfrag *)q6) {
481		i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->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 succeeding fragment",
486			    i, ip6_sprintf(ip6buf, &q6->ip6q_src));
487#endif
488			free(ip6af, M_FTABLE);
489			goto dropfrag;
490		}
491	}
492#endif
493
494insert:
495#ifdef MAC
496	if (!first_frag)
497		mac_ip6q_update(m, q6);
498#endif
499
500	/*
501	 * Stick new segment in its place;
502	 * check for complete reassembly.
503	 * Move to front of packet queue, as we are
504	 * the most recently active fragmented packet.
505	 */
506	frag6_enq(ip6af, af6->ip6af_up);
507	V_frag6_nfrags++;
508	q6->ip6q_nfrag++;
509#if 0 /* xxx */
510	if (q6 != V_ip6q.ip6q_next) {
511		frag6_remque(q6);
512		frag6_insque(q6, &V_ip6q);
513	}
514#endif
515	next = 0;
516	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
517	     af6 = af6->ip6af_down) {
518		if (af6->ip6af_off != next) {
519			IP6Q_UNLOCK();
520			return IPPROTO_DONE;
521		}
522		next += af6->ip6af_frglen;
523	}
524	if (af6->ip6af_up->ip6af_mff) {
525		IP6Q_UNLOCK();
526		return IPPROTO_DONE;
527	}
528
529	/*
530	 * Reassembly is complete; concatenate fragments.
531	 */
532	ip6af = q6->ip6q_down;
533	t = m = IP6_REASS_MBUF(ip6af);
534	af6 = ip6af->ip6af_down;
535	frag6_deq(ip6af);
536	while (af6 != (struct ip6asfrag *)q6) {
537		af6dwn = af6->ip6af_down;
538		frag6_deq(af6);
539		while (t->m_next)
540			t = t->m_next;
541		t->m_next = IP6_REASS_MBUF(af6);
542		m_adj(t->m_next, af6->ip6af_offset);
543		free(af6, M_FTABLE);
544		af6 = af6dwn;
545	}
546
547	/* adjust offset to point where the original next header starts */
548	offset = ip6af->ip6af_offset - sizeof(struct ip6_frag);
549	free(ip6af, M_FTABLE);
550	ip6 = mtod(m, struct ip6_hdr *);
551	ip6->ip6_plen = htons((u_short)next + offset - sizeof(struct ip6_hdr));
552	if (q6->ip6q_ecn == IPTOS_ECN_CE)
553		ip6->ip6_flow |= htonl(IPTOS_ECN_CE << 20);
554	nxt = q6->ip6q_nxt;
555#ifdef notyet
556	*q6->ip6q_nxtp = (u_char)(nxt & 0xff);
557#endif
558
559	/* Delete frag6 header */
560	if (m->m_len >= offset + sizeof(struct ip6_frag)) {
561		/* This is the only possible case with !PULLDOWN_TEST */
562		ovbcopy((caddr_t)ip6, (caddr_t)ip6 + sizeof(struct ip6_frag),
563		    offset);
564		m->m_data += sizeof(struct ip6_frag);
565		m->m_len -= sizeof(struct ip6_frag);
566	} else {
567		/* this comes with no copy if the boundary is on cluster */
568		if ((t = m_split(m, offset, M_DONTWAIT)) == NULL) {
569			frag6_remque(q6);
570			V_frag6_nfrags -= q6->ip6q_nfrag;
571#ifdef MAC
572			mac_ip6q_destroy(q6);
573#endif
574			free(q6, M_FTABLE);
575			V_frag6_nfragpackets--;
576			goto dropfrag;
577		}
578		m_adj(t, sizeof(struct ip6_frag));
579		m_cat(m, t);
580	}
581
582	/*
583	 * Store NXT to the original.
584	 */
585	{
586		char *prvnxtp = ip6_get_prevhdr(m, offset); /* XXX */
587		*prvnxtp = nxt;
588	}
589
590	frag6_remque(q6);
591	V_frag6_nfrags -= q6->ip6q_nfrag;
592#ifdef MAC
593	mac_ip6q_reassemble(q6, m);
594	mac_ip6q_destroy(q6);
595#endif
596	free(q6, M_FTABLE);
597	V_frag6_nfragpackets--;
598
599	if (m->m_flags & M_PKTHDR) { /* Isn't it always true? */
600		int plen = 0;
601		for (t = m; t; t = t->m_next)
602			plen += t->m_len;
603		m->m_pkthdr.len = plen;
604	}
605
606	V_ip6stat.ip6s_reassembled++;
607	in6_ifstat_inc(dstifp, ifs6_reass_ok);
608
609	/*
610	 * Tell launch routine the next header
611	 */
612
613	*mp = m;
614	*offp = offset;
615
616	IP6Q_UNLOCK();
617	return nxt;
618
619 dropfrag:
620	IP6Q_UNLOCK();
621	in6_ifstat_inc(dstifp, ifs6_reass_fail);
622	V_ip6stat.ip6s_fragdropped++;
623	m_freem(m);
624	return IPPROTO_DONE;
625}
626
627/*
628 * Free a fragment reassembly header and all
629 * associated datagrams.
630 */
631void
632frag6_freef(struct ip6q *q6)
633{
634	struct ip6asfrag *af6, *down6;
635
636	IP6Q_LOCK_ASSERT();
637
638	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
639	     af6 = down6) {
640		struct mbuf *m = IP6_REASS_MBUF(af6);
641
642		down6 = af6->ip6af_down;
643		frag6_deq(af6);
644
645		/*
646		 * Return ICMP time exceeded error for the 1st fragment.
647		 * Just free other fragments.
648		 */
649		if (af6->ip6af_off == 0) {
650			struct ip6_hdr *ip6;
651
652			/* adjust pointer */
653			ip6 = mtod(m, struct ip6_hdr *);
654
655			/* restore source and destination addresses */
656			ip6->ip6_src = q6->ip6q_src;
657			ip6->ip6_dst = q6->ip6q_dst;
658
659			icmp6_error(m, ICMP6_TIME_EXCEEDED,
660				    ICMP6_TIME_EXCEED_REASSEMBLY, 0);
661		} else
662			m_freem(m);
663		free(af6, M_FTABLE);
664	}
665	frag6_remque(q6);
666	V_frag6_nfrags -= q6->ip6q_nfrag;
667#ifdef MAC
668	mac_ip6q_destroy(q6);
669#endif
670	free(q6, M_FTABLE);
671	V_frag6_nfragpackets--;
672}
673
674/*
675 * Put an ip fragment on a reassembly chain.
676 * Like insque, but pointers in middle of structure.
677 */
678void
679frag6_enq(struct ip6asfrag *af6, struct ip6asfrag *up6)
680{
681
682	IP6Q_LOCK_ASSERT();
683
684	af6->ip6af_up = up6;
685	af6->ip6af_down = up6->ip6af_down;
686	up6->ip6af_down->ip6af_up = af6;
687	up6->ip6af_down = af6;
688}
689
690/*
691 * To frag6_enq as remque is to insque.
692 */
693void
694frag6_deq(struct ip6asfrag *af6)
695{
696
697	IP6Q_LOCK_ASSERT();
698
699	af6->ip6af_up->ip6af_down = af6->ip6af_down;
700	af6->ip6af_down->ip6af_up = af6->ip6af_up;
701}
702
703void
704frag6_insque(struct ip6q *new, struct ip6q *old)
705{
706
707	IP6Q_LOCK_ASSERT();
708
709	new->ip6q_prev = old;
710	new->ip6q_next = old->ip6q_next;
711	old->ip6q_next->ip6q_prev= new;
712	old->ip6q_next = new;
713}
714
715void
716frag6_remque(struct ip6q *p6)
717{
718
719	IP6Q_LOCK_ASSERT();
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(void)
732{
733	VNET_ITERATOR_DECL(vnet_iter);
734	struct ip6q *q6;
735
736	VNET_LIST_RLOCK_NOSLEEP();
737	IP6Q_LOCK();
738	VNET_FOREACH(vnet_iter) {
739		CURVNET_SET(vnet_iter);
740		q6 = V_ip6q.ip6q_next;
741		if (q6)
742			while (q6 != &V_ip6q) {
743				--q6->ip6q_ttl;
744				q6 = q6->ip6q_next;
745				if (q6->ip6q_prev->ip6q_ttl == 0) {
746					V_ip6stat.ip6s_fragtimeout++;
747					/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
748					frag6_freef(q6->ip6q_prev);
749				}
750			}
751		/*
752		 * If we are over the maximum number of fragments
753		 * (due to the limit being lowered), drain off
754		 * enough to get down to the new limit.
755		 */
756		while (V_frag6_nfragpackets > (u_int)V_ip6_maxfragpackets &&
757		    V_ip6q.ip6q_prev) {
758			V_ip6stat.ip6s_fragoverflow++;
759			/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
760			frag6_freef(V_ip6q.ip6q_prev);
761		}
762		CURVNET_RESTORE();
763	}
764	IP6Q_UNLOCK();
765	VNET_LIST_RUNLOCK_NOSLEEP();
766}
767
768/*
769 * Drain off all datagram fragments.
770 */
771void
772frag6_drain(void)
773{
774	VNET_ITERATOR_DECL(vnet_iter);
775
776	VNET_LIST_RLOCK_NOSLEEP();
777	if (IP6Q_TRYLOCK() == 0) {
778		VNET_LIST_RUNLOCK_NOSLEEP();
779		return;
780	}
781	VNET_FOREACH(vnet_iter) {
782		CURVNET_SET(vnet_iter);
783		while (V_ip6q.ip6q_next != &V_ip6q) {
784			V_ip6stat.ip6s_fragdropped++;
785			/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
786			frag6_freef(V_ip6q.ip6q_next);
787		}
788		CURVNET_RESTORE();
789	}
790	IP6Q_UNLOCK();
791	VNET_LIST_RUNLOCK_NOSLEEP();
792}
793