frag6.c revision 165118
162587Sitojun/*	$FreeBSD: head/sys/netinet6/frag6.c 165118 2006-12-12 12:17:58Z bz $	*/
295023Ssuz/*	$KAME: frag6.c,v 1.33 2002/01/07 11:34:48 kjc Exp $	*/
362587Sitojun
4139826Simp/*-
553541Sshin * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
653541Sshin * All rights reserved.
753541Sshin *
853541Sshin * Redistribution and use in source and binary forms, with or without
953541Sshin * modification, are permitted provided that the following conditions
1053541Sshin * are met:
1153541Sshin * 1. Redistributions of source code must retain the above copyright
1253541Sshin *    notice, this list of conditions and the following disclaimer.
1353541Sshin * 2. Redistributions in binary form must reproduce the above copyright
1453541Sshin *    notice, this list of conditions and the following disclaimer in the
1553541Sshin *    documentation and/or other materials provided with the distribution.
1653541Sshin * 3. Neither the name of the project nor the names of its contributors
1753541Sshin *    may be used to endorse or promote products derived from this software
1853541Sshin *    without specific prior written permission.
1953541Sshin *
2053541Sshin * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
2153541Sshin * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
2253541Sshin * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
2353541Sshin * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
2453541Sshin * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
2553541Sshin * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
2653541Sshin * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
2753541Sshin * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
2853541Sshin * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
2953541Sshin * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
3053541Sshin * SUCH DAMAGE.
3153541Sshin */
3253541Sshin
3353541Sshin#include <sys/param.h>
3453541Sshin#include <sys/systm.h>
3553541Sshin#include <sys/malloc.h>
3653541Sshin#include <sys/mbuf.h>
3753541Sshin#include <sys/domain.h>
3853541Sshin#include <sys/protosw.h>
3953541Sshin#include <sys/socket.h>
4053541Sshin#include <sys/errno.h>
4153541Sshin#include <sys/time.h>
4253541Sshin#include <sys/kernel.h>
4353541Sshin#include <sys/syslog.h>
4453541Sshin
4553541Sshin#include <net/if.h>
4653541Sshin#include <net/route.h>
4753541Sshin
4853541Sshin#include <netinet/in.h>
4953541Sshin#include <netinet/in_var.h>
5062587Sitojun#include <netinet/ip6.h>
5153541Sshin#include <netinet6/ip6_var.h>
5262587Sitojun#include <netinet/icmp6.h>
53121684Sume#include <netinet/in_systm.h>	/* for ECN definitions */
54121684Sume#include <netinet/ip.h>		/* for ECN definitions */
5553541Sshin
5653541Sshin/*
5753541Sshin * Define it to get a correct behavior on per-interface statistics.
5853541Sshin * You will need to perform an extra routing table lookup, per fragment,
5953541Sshin * to do it.  This may, or may not be, a performance hit.
6053541Sshin */
6162587Sitojun#define IN6_IFSTAT_STRICT
6253541Sshin
6362587Sitojunstatic void frag6_enq __P((struct ip6asfrag *, struct ip6asfrag *));
6462587Sitojunstatic void frag6_deq __P((struct ip6asfrag *));
6562587Sitojunstatic void frag6_insque __P((struct ip6q *, struct ip6q *));
6662587Sitojunstatic void frag6_remque __P((struct ip6q *));
6762587Sitojunstatic void frag6_freef __P((struct ip6q *));
6853541Sshin
69121346Sumestatic struct mtx ip6qlock;
70121346Sume/*
71121346Sume * These fields all protected by ip6qlock.
72121346Sume */
73121346Sumestatic u_int frag6_nfragpackets;
74121346Sumestatic u_int frag6_nfrags;
75121346Sumestatic struct	ip6q ip6q;	/* ip6 reassemble queue */
7653541Sshin
77121346Sume#define	IP6Q_LOCK_INIT()	mtx_init(&ip6qlock, "ip6qlock", NULL, MTX_DEF);
78121346Sume#define	IP6Q_LOCK()		mtx_lock(&ip6qlock)
79121346Sume#define	IP6Q_TRYLOCK()		mtx_trylock(&ip6qlock)
80121355Sume#define	IP6Q_LOCK_ASSERT()	mtx_assert(&ip6qlock, MA_OWNED)
81121346Sume#define	IP6Q_UNLOCK()		mtx_unlock(&ip6qlock)
82121345Sume
8369774Sphkstatic MALLOC_DEFINE(M_FTABLE, "fragment", "fragment reassembly header");
8462587Sitojun
8553541Sshin/*
8653541Sshin * Initialise reassembly queue and fragment identifier.
8753541Sshin */
88157927Spsstatic void
89157927Spsfrag6_change(void *tag)
90157927Sps{
91157927Sps
92157927Sps	ip6_maxfragpackets = nmbclusters / 4;
93157927Sps	ip6_maxfrags = nmbclusters / 4;
94157927Sps}
95157927Sps
9653541Sshinvoid
9753541Sshinfrag6_init()
9853541Sshin{
9953541Sshin
10077969Sjesper	ip6_maxfragpackets = nmbclusters / 4;
101121345Sume	ip6_maxfrags = nmbclusters / 4;
102157927Sps	EVENTHANDLER_REGISTER(nmbclusters_change,
103157927Sps	    frag6_change, NULL, EVENTHANDLER_PRI_ANY);
10477969Sjesper
105121346Sume	IP6Q_LOCK_INIT();
106121346Sume
10753541Sshin	ip6q.ip6q_next = ip6q.ip6q_prev = &ip6q;
10853541Sshin}
10953541Sshin
11053541Sshin/*
11162587Sitojun * In RFC2460, fragment and reassembly rule do not agree with each other,
11262587Sitojun * in terms of next header field handling in fragment header.
11362587Sitojun * While the sender will use the same value for all of the fragmented packets,
11462587Sitojun * receiver is suggested not to check the consistency.
11562587Sitojun *
11662587Sitojun * fragment rule (p20):
11762587Sitojun *	(2) A Fragment header containing:
11862587Sitojun *	The Next Header value that identifies the first header of
11962587Sitojun *	the Fragmentable Part of the original packet.
12062587Sitojun *		-> next header field is same for all fragments
12162587Sitojun *
12262587Sitojun * reassembly rule (p21):
12362587Sitojun *	The Next Header field of the last header of the Unfragmentable
12462587Sitojun *	Part is obtained from the Next Header field of the first
12562587Sitojun *	fragment's Fragment header.
12662587Sitojun *		-> should grab it from the first fragment only
12762587Sitojun *
12862587Sitojun * The following note also contradicts with fragment rule - noone is going to
12962587Sitojun * send different fragment with different next header field.
13062587Sitojun *
13162587Sitojun * additional note (p22):
13262587Sitojun *	The Next Header values in the Fragment headers of different
13362587Sitojun *	fragments of the same original packet may differ.  Only the value
13462587Sitojun *	from the Offset zero fragment packet is used for reassembly.
13562587Sitojun *		-> should grab it from the first fragment only
13662587Sitojun *
13762587Sitojun * There is no explicit reason given in the RFC.  Historical reason maybe?
13862587Sitojun */
13962587Sitojun/*
14053541Sshin * Fragment input
14153541Sshin */
14253541Sshinint
14353541Sshinfrag6_input(mp, offp, proto)
14453541Sshin	struct mbuf **mp;
14553541Sshin	int *offp, proto;
14653541Sshin{
14753541Sshin	struct mbuf *m = *mp, *t;
14853541Sshin	struct ip6_hdr *ip6;
14953541Sshin	struct ip6_frag *ip6f;
15053541Sshin	struct ip6q *q6;
15162587Sitojun	struct ip6asfrag *af6, *ip6af, *af6dwn;
152121630Sume#ifdef IN6_IFSTAT_STRICT
153121630Sume	struct in6_ifaddr *ia;
154121630Sume#endif
15553541Sshin	int offset = *offp, nxt, i, next;
15653541Sshin	int first_frag = 0;
15762587Sitojun	int fragoff, frgpartlen;	/* must be larger than u_int16_t */
15853541Sshin	struct ifnet *dstifp;
159121684Sume	u_int8_t ecn, ecn0;
160165118Sbz#if 0
161165118Sbz	char ip6buf[INET6_ADDRSTRLEN];
162165118Sbz#endif
16353541Sshin
16462587Sitojun	ip6 = mtod(m, struct ip6_hdr *);
16562587Sitojun#ifndef PULLDOWN_TEST
16653541Sshin	IP6_EXTHDR_CHECK(m, offset, sizeof(struct ip6_frag), IPPROTO_DONE);
16753541Sshin	ip6f = (struct ip6_frag *)((caddr_t)ip6 + offset);
16862587Sitojun#else
16962587Sitojun	IP6_EXTHDR_GET(ip6f, struct ip6_frag *, m, offset, sizeof(*ip6f));
17062587Sitojun	if (ip6f == NULL)
171120856Sume		return (IPPROTO_DONE);
17262587Sitojun#endif
17353541Sshin
17453541Sshin	dstifp = NULL;
17553541Sshin#ifdef IN6_IFSTAT_STRICT
17653541Sshin	/* find the destination interface of the packet. */
177121630Sume	if ((ia = ip6_getdstifaddr(m)) != NULL)
178121630Sume		dstifp = ia->ia_ifp;
17953541Sshin#else
18053541Sshin	/* we are violating the spec, this is not the destination interface */
18153541Sshin	if ((m->m_flags & M_PKTHDR) != 0)
18253541Sshin		dstifp = m->m_pkthdr.rcvif;
18353541Sshin#endif
18453541Sshin
18553541Sshin	/* jumbo payload can't contain a fragment header */
18653541Sshin	if (ip6->ip6_plen == 0) {
18753541Sshin		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, offset);
18853541Sshin		in6_ifstat_inc(dstifp, ifs6_reass_fail);
18953541Sshin		return IPPROTO_DONE;
19053541Sshin	}
19153541Sshin
19253541Sshin	/*
19353541Sshin	 * check whether fragment packet's fragment length is
19453541Sshin	 * multiple of 8 octets.
19553541Sshin	 * sizeof(struct ip6_frag) == 8
19653541Sshin	 * sizeof(struct ip6_hdr) = 40
19753541Sshin	 */
19853541Sshin	if ((ip6f->ip6f_offlg & IP6F_MORE_FRAG) &&
19953541Sshin	    (((ntohs(ip6->ip6_plen) - offset) & 0x7) != 0)) {
200120891Sume		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
201120891Sume		    offsetof(struct ip6_hdr, ip6_plen));
20253541Sshin		in6_ifstat_inc(dstifp, ifs6_reass_fail);
20353541Sshin		return IPPROTO_DONE;
20453541Sshin	}
20553541Sshin
20653541Sshin	ip6stat.ip6s_fragments++;
20753541Sshin	in6_ifstat_inc(dstifp, ifs6_reass_reqd);
208120891Sume
20962587Sitojun	/* offset now points to data portion */
21053541Sshin	offset += sizeof(struct ip6_frag);
21153541Sshin
212121345Sume	IP6Q_LOCK();
21378064Sume
214121345Sume	/*
215121345Sume	 * Enforce upper bound on number of fragments.
216121345Sume	 * If maxfrag is 0, never accept fragments.
217121345Sume	 * If maxfrag is -1, accept all fragments without limitation.
218121345Sume	 */
219121345Sume	if (ip6_maxfrags < 0)
220121345Sume		;
221121345Sume	else if (frag6_nfrags >= (u_int)ip6_maxfrags)
222121345Sume		goto dropfrag;
223121345Sume
22453541Sshin	for (q6 = ip6q.ip6q_next; q6 != &ip6q; q6 = q6->ip6q_next)
22553541Sshin		if (ip6f->ip6f_ident == q6->ip6q_ident &&
22653541Sshin		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &q6->ip6q_src) &&
22753541Sshin		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &q6->ip6q_dst))
22853541Sshin			break;
22953541Sshin
23053541Sshin	if (q6 == &ip6q) {
23153541Sshin		/*
23253541Sshin		 * the first fragment to arrive, create a reassembly queue.
23353541Sshin		 */
23453541Sshin		first_frag = 1;
23553541Sshin
23653541Sshin		/*
23753541Sshin		 * Enforce upper bound on number of fragmented packets
23853541Sshin		 * for which we attempt reassembly;
239121345Sume		 * If maxfragpackets is 0, never accept fragments.
240121345Sume		 * If maxfragpackets is -1, accept all fragments without
241121345Sume		 * limitation.
24253541Sshin		 */
24378064Sume		if (ip6_maxfragpackets < 0)
24478064Sume			;
24578064Sume		else if (frag6_nfragpackets >= (u_int)ip6_maxfragpackets)
24678064Sume			goto dropfrag;
24778064Sume		frag6_nfragpackets++;
24853541Sshin		q6 = (struct ip6q *)malloc(sizeof(struct ip6q), M_FTABLE,
249121607Sume		    M_NOWAIT);
25053541Sshin		if (q6 == NULL)
25153541Sshin			goto dropfrag;
25262587Sitojun		bzero(q6, sizeof(*q6));
25353541Sshin
25453541Sshin		frag6_insque(q6, &ip6q);
25553541Sshin
25662587Sitojun		/* ip6q_nxt will be filled afterwards, from 1st fragment */
25753541Sshin		q6->ip6q_down	= q6->ip6q_up = (struct ip6asfrag *)q6;
25862587Sitojun#ifdef notyet
25962587Sitojun		q6->ip6q_nxtp	= (u_char *)nxtp;
26062587Sitojun#endif
26153541Sshin		q6->ip6q_ident	= ip6f->ip6f_ident;
26253541Sshin		q6->ip6q_arrive = 0; /* Is it used anywhere? */
26353541Sshin		q6->ip6q_ttl 	= IPV6_FRAGTTL;
26453541Sshin		q6->ip6q_src	= ip6->ip6_src;
26553541Sshin		q6->ip6q_dst	= ip6->ip6_dst;
26653541Sshin		q6->ip6q_unfrglen = -1;	/* The 1st fragment has not arrived. */
267121345Sume
268121345Sume		q6->ip6q_nfrag = 0;
26953541Sshin	}
27053541Sshin
27153541Sshin	/*
27253541Sshin	 * If it's the 1st fragment, record the length of the
27353541Sshin	 * unfragmentable part and the next header of the fragment header.
27453541Sshin	 */
27553541Sshin	fragoff = ntohs(ip6f->ip6f_offlg & IP6F_OFF_MASK);
27653541Sshin	if (fragoff == 0) {
277120891Sume		q6->ip6q_unfrglen = offset - sizeof(struct ip6_hdr) -
278120891Sume		    sizeof(struct ip6_frag);
27953541Sshin		q6->ip6q_nxt = ip6f->ip6f_nxt;
28053541Sshin	}
28153541Sshin
28253541Sshin	/*
28353541Sshin	 * Check that the reassembled packet would not exceed 65535 bytes
28453541Sshin	 * in size.
28553541Sshin	 * If it would exceed, discard the fragment and return an ICMP error.
28653541Sshin	 */
28762587Sitojun	frgpartlen = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen) - offset;
28853541Sshin	if (q6->ip6q_unfrglen >= 0) {
28953541Sshin		/* The 1st fragment has already arrived. */
29053541Sshin		if (q6->ip6q_unfrglen + fragoff + frgpartlen > IPV6_MAXPACKET) {
29153541Sshin			icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
292120891Sume			    offset - sizeof(struct ip6_frag) +
293120891Sume			    offsetof(struct ip6_frag, ip6f_offlg));
294121345Sume			IP6Q_UNLOCK();
295120856Sume			return (IPPROTO_DONE);
29653541Sshin		}
297120891Sume	} else if (fragoff + frgpartlen > IPV6_MAXPACKET) {
29853541Sshin		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
299120891Sume		    offset - sizeof(struct ip6_frag) +
300120891Sume		    offsetof(struct ip6_frag, ip6f_offlg));
301121345Sume		IP6Q_UNLOCK();
302120856Sume		return (IPPROTO_DONE);
30353541Sshin	}
30453541Sshin	/*
30553541Sshin	 * If it's the first fragment, do the above check for each
30653541Sshin	 * fragment already stored in the reassembly queue.
30753541Sshin	 */
30853541Sshin	if (fragoff == 0) {
30953541Sshin		for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
31053541Sshin		     af6 = af6dwn) {
31153541Sshin			af6dwn = af6->ip6af_down;
31253541Sshin
31353541Sshin			if (q6->ip6q_unfrglen + af6->ip6af_off + af6->ip6af_frglen >
31453541Sshin			    IPV6_MAXPACKET) {
31553541Sshin				struct mbuf *merr = IP6_REASS_MBUF(af6);
31653541Sshin				struct ip6_hdr *ip6err;
31753541Sshin				int erroff = af6->ip6af_offset;
31853541Sshin
31953541Sshin				/* dequeue the fragment. */
32053541Sshin				frag6_deq(af6);
32162587Sitojun				free(af6, M_FTABLE);
32253541Sshin
32353541Sshin				/* adjust pointer. */
32453541Sshin				ip6err = mtod(merr, struct ip6_hdr *);
32553541Sshin
32653541Sshin				/*
32753541Sshin				 * Restore source and destination addresses
32853541Sshin				 * in the erroneous IPv6 header.
32953541Sshin				 */
33053541Sshin				ip6err->ip6_src = q6->ip6q_src;
33153541Sshin				ip6err->ip6_dst = q6->ip6q_dst;
33253541Sshin
33353541Sshin				icmp6_error(merr, ICMP6_PARAM_PROB,
334120891Sume				    ICMP6_PARAMPROB_HEADER,
335120891Sume				    erroff - sizeof(struct ip6_frag) +
336120891Sume				    offsetof(struct ip6_frag, ip6f_offlg));
33753541Sshin			}
33853541Sshin		}
33953541Sshin	}
34053541Sshin
34162587Sitojun	ip6af = (struct ip6asfrag *)malloc(sizeof(struct ip6asfrag), M_FTABLE,
342121607Sume	    M_NOWAIT);
34362587Sitojun	if (ip6af == NULL)
34462587Sitojun		goto dropfrag;
34562587Sitojun	bzero(ip6af, sizeof(*ip6af));
34662587Sitojun	ip6af->ip6af_head = ip6->ip6_flow;
34762587Sitojun	ip6af->ip6af_len = ip6->ip6_plen;
34862587Sitojun	ip6af->ip6af_nxt = ip6->ip6_nxt;
34962587Sitojun	ip6af->ip6af_hlim = ip6->ip6_hlim;
35053541Sshin	ip6af->ip6af_mff = ip6f->ip6f_offlg & IP6F_MORE_FRAG;
35153541Sshin	ip6af->ip6af_off = fragoff;
35253541Sshin	ip6af->ip6af_frglen = frgpartlen;
35353541Sshin	ip6af->ip6af_offset = offset;
35453541Sshin	IP6_REASS_MBUF(ip6af) = m;
35553541Sshin
35653541Sshin	if (first_frag) {
35753541Sshin		af6 = (struct ip6asfrag *)q6;
35853541Sshin		goto insert;
35953541Sshin	}
36053541Sshin
36153541Sshin	/*
362121684Sume	 * Handle ECN by comparing this segment with the first one;
363121684Sume	 * if CE is set, do not lose CE.
364121684Sume	 * drop if CE and not-ECT are mixed for the same packet.
365121684Sume	 */
366121684Sume	ecn = (ntohl(ip6->ip6_flow) >> 20) & IPTOS_ECN_MASK;
367121684Sume	ecn0 = (ntohl(q6->ip6q_down->ip6af_head) >> 20) & IPTOS_ECN_MASK;
368121684Sume	if (ecn == IPTOS_ECN_CE) {
369121684Sume		if (ecn0 == IPTOS_ECN_NOTECT) {
370121684Sume			free(ip6af, M_FTABLE);
371121684Sume			goto dropfrag;
372121684Sume		}
373121684Sume		if (ecn0 != IPTOS_ECN_CE)
374121684Sume			q6->ip6q_down->ip6af_head |= htonl(IPTOS_ECN_CE << 20);
375121684Sume	}
376121684Sume	if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) {
377121684Sume		free(ip6af, M_FTABLE);
378121684Sume		goto dropfrag;
379121684Sume	}
380121684Sume
381121684Sume	/*
38253541Sshin	 * Find a segment which begins after this one does.
38353541Sshin	 */
38453541Sshin	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
38553541Sshin	     af6 = af6->ip6af_down)
38653541Sshin		if (af6->ip6af_off > ip6af->ip6af_off)
38753541Sshin			break;
38853541Sshin
38962587Sitojun#if 0
39053541Sshin	/*
39162587Sitojun	 * If there is a preceding segment, it may provide some of
39262587Sitojun	 * our data already.  If so, drop the data from the incoming
39362587Sitojun	 * segment.  If it provides all of our data, drop us.
39462587Sitojun	 */
39562587Sitojun	if (af6->ip6af_up != (struct ip6asfrag *)q6) {
39662587Sitojun		i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
39762587Sitojun			- ip6af->ip6af_off;
39862587Sitojun		if (i > 0) {
39962587Sitojun			if (i >= ip6af->ip6af_frglen)
40062587Sitojun				goto dropfrag;
40162587Sitojun			m_adj(IP6_REASS_MBUF(ip6af), i);
40262587Sitojun			ip6af->ip6af_off += i;
40362587Sitojun			ip6af->ip6af_frglen -= i;
40462587Sitojun		}
40562587Sitojun	}
40662587Sitojun
40762587Sitojun	/*
40862587Sitojun	 * While we overlap succeeding segments trim them or,
40962587Sitojun	 * if they are completely covered, dequeue them.
41062587Sitojun	 */
41162587Sitojun	while (af6 != (struct ip6asfrag *)q6 &&
41262587Sitojun	       ip6af->ip6af_off + ip6af->ip6af_frglen > af6->ip6af_off) {
41362587Sitojun		i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
41462587Sitojun		if (i < af6->ip6af_frglen) {
41562587Sitojun			af6->ip6af_frglen -= i;
41662587Sitojun			af6->ip6af_off += i;
41762587Sitojun			m_adj(IP6_REASS_MBUF(af6), i);
41862587Sitojun			break;
41962587Sitojun		}
42062587Sitojun		af6 = af6->ip6af_down;
42162587Sitojun		m_freem(IP6_REASS_MBUF(af6->ip6af_up));
42262587Sitojun		frag6_deq(af6->ip6af_up);
42362587Sitojun	}
42462587Sitojun#else
42562587Sitojun	/*
42653541Sshin	 * If the incoming framgent overlaps some existing fragments in
42753541Sshin	 * the reassembly queue, drop it, since it is dangerous to override
42853541Sshin	 * existing fragments from a security point of view.
429121345Sume	 * We don't know which fragment is the bad guy - here we trust
430121345Sume	 * fragment that came in earlier, with no real reason.
43153541Sshin	 */
43253541Sshin	if (af6->ip6af_up != (struct ip6asfrag *)q6) {
43353541Sshin		i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
43453541Sshin			- ip6af->ip6af_off;
43553541Sshin		if (i > 0) {
43676899Ssumikawa#if 0				/* suppress the noisy log */
43753541Sshin			log(LOG_ERR, "%d bytes of a fragment from %s "
43853541Sshin			    "overlaps the previous fragment\n",
439165118Sbz			    i, ip6_sprintf(ip6buf, &q6->ip6q_src));
44076899Ssumikawa#endif
44176899Ssumikawa			free(ip6af, M_FTABLE);
44253541Sshin			goto dropfrag;
44353541Sshin		}
44453541Sshin	}
44553541Sshin	if (af6 != (struct ip6asfrag *)q6) {
44653541Sshin		i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
44753541Sshin		if (i > 0) {
44876899Ssumikawa#if 0				/* suppress the noisy log */
44953541Sshin			log(LOG_ERR, "%d bytes of a fragment from %s "
45053541Sshin			    "overlaps the succeeding fragment",
451165118Sbz			    i, ip6_sprintf(ip6buf, &q6->ip6q_src));
45276899Ssumikawa#endif
45376899Ssumikawa			free(ip6af, M_FTABLE);
45453541Sshin			goto dropfrag;
45553541Sshin		}
45653541Sshin	}
45762587Sitojun#endif
45853541Sshin
45953541Sshininsert:
46053541Sshin
46153541Sshin	/*
46253541Sshin	 * Stick new segment in its place;
46353541Sshin	 * check for complete reassembly.
46453541Sshin	 * Move to front of packet queue, as we are
46553541Sshin	 * the most recently active fragmented packet.
46653541Sshin	 */
46753541Sshin	frag6_enq(ip6af, af6->ip6af_up);
468121345Sume	frag6_nfrags++;
469121345Sume	q6->ip6q_nfrag++;
47062587Sitojun#if 0 /* xxx */
47162587Sitojun	if (q6 != ip6q.ip6q_next) {
47262587Sitojun		frag6_remque(q6);
47362587Sitojun		frag6_insque(q6, &ip6q);
47462587Sitojun	}
47562587Sitojun#endif
47653541Sshin	next = 0;
47753541Sshin	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
47853541Sshin	     af6 = af6->ip6af_down) {
47953541Sshin		if (af6->ip6af_off != next) {
480121345Sume			IP6Q_UNLOCK();
48153541Sshin			return IPPROTO_DONE;
48253541Sshin		}
48353541Sshin		next += af6->ip6af_frglen;
48453541Sshin	}
48553541Sshin	if (af6->ip6af_up->ip6af_mff) {
486121345Sume		IP6Q_UNLOCK();
48753541Sshin		return IPPROTO_DONE;
48853541Sshin	}
48953541Sshin
49053541Sshin	/*
49153541Sshin	 * Reassembly is complete; concatenate fragments.
49253541Sshin	 */
49353541Sshin	ip6af = q6->ip6q_down;
49453541Sshin	t = m = IP6_REASS_MBUF(ip6af);
49553541Sshin	af6 = ip6af->ip6af_down;
49662587Sitojun	frag6_deq(ip6af);
49753541Sshin	while (af6 != (struct ip6asfrag *)q6) {
49862587Sitojun		af6dwn = af6->ip6af_down;
49962587Sitojun		frag6_deq(af6);
50053541Sshin		while (t->m_next)
50153541Sshin			t = t->m_next;
50253541Sshin		t->m_next = IP6_REASS_MBUF(af6);
50362587Sitojun		m_adj(t->m_next, af6->ip6af_offset);
50462587Sitojun		free(af6, M_FTABLE);
50562587Sitojun		af6 = af6dwn;
50653541Sshin	}
50753541Sshin
50853541Sshin	/* adjust offset to point where the original next header starts */
50953541Sshin	offset = ip6af->ip6af_offset - sizeof(struct ip6_frag);
51062587Sitojun	free(ip6af, M_FTABLE);
51162587Sitojun	ip6 = mtod(m, struct ip6_hdr *);
51253541Sshin	ip6->ip6_plen = htons((u_short)next + offset - sizeof(struct ip6_hdr));
51353541Sshin	ip6->ip6_src = q6->ip6q_src;
51453541Sshin	ip6->ip6_dst = q6->ip6q_dst;
51553541Sshin	nxt = q6->ip6q_nxt;
51662587Sitojun#ifdef notyet
51762587Sitojun	*q6->ip6q_nxtp = (u_char)(nxt & 0xff);
51862587Sitojun#endif
51953541Sshin
52053541Sshin	/*
52153541Sshin	 * Delete frag6 header with as a few cost as possible.
52253541Sshin	 */
52362587Sitojun	if (offset < m->m_len) {
52453541Sshin		ovbcopy((caddr_t)ip6, (caddr_t)ip6 + sizeof(struct ip6_frag),
52553541Sshin			offset);
52662587Sitojun		m->m_data += sizeof(struct ip6_frag);
52762587Sitojun		m->m_len -= sizeof(struct ip6_frag);
52862587Sitojun	} else {
52962587Sitojun		/* this comes with no copy if the boundary is on cluster */
530111119Simp		if ((t = m_split(m, offset, M_DONTWAIT)) == NULL) {
53162587Sitojun			frag6_remque(q6);
532121345Sume			frag6_nfrags -= q6->ip6q_nfrag;
53362587Sitojun			free(q6, M_FTABLE);
53462587Sitojun			frag6_nfragpackets--;
53562587Sitojun			goto dropfrag;
53662587Sitojun		}
53762587Sitojun		m_adj(t, sizeof(struct ip6_frag));
53862587Sitojun		m_cat(m, t);
53953541Sshin	}
54053541Sshin
54153541Sshin	/*
54253541Sshin	 * Store NXT to the original.
54353541Sshin	 */
54453541Sshin	{
54553541Sshin		char *prvnxtp = ip6_get_prevhdr(m, offset); /* XXX */
54653541Sshin		*prvnxtp = nxt;
54753541Sshin	}
54853541Sshin
54953541Sshin	frag6_remque(q6);
550121345Sume	frag6_nfrags -= q6->ip6q_nfrag;
55153541Sshin	free(q6, M_FTABLE);
55253541Sshin	frag6_nfragpackets--;
55353541Sshin
55453541Sshin	if (m->m_flags & M_PKTHDR) { /* Isn't it always true? */
55553541Sshin		int plen = 0;
55653541Sshin		for (t = m; t; t = t->m_next)
55753541Sshin			plen += t->m_len;
55853541Sshin		m->m_pkthdr.len = plen;
55953541Sshin	}
560120891Sume
56153541Sshin	ip6stat.ip6s_reassembled++;
56253541Sshin	in6_ifstat_inc(dstifp, ifs6_reass_ok);
56353541Sshin
56453541Sshin	/*
56553541Sshin	 * Tell launch routine the next header
56653541Sshin	 */
56753541Sshin
56853541Sshin	*mp = m;
56953541Sshin	*offp = offset;
57053541Sshin
571121345Sume	IP6Q_UNLOCK();
57253541Sshin	return nxt;
57353541Sshin
57453541Sshin dropfrag:
575121346Sume	IP6Q_UNLOCK();
57653541Sshin	in6_ifstat_inc(dstifp, ifs6_reass_fail);
57753541Sshin	ip6stat.ip6s_fragdropped++;
57853541Sshin	m_freem(m);
57953541Sshin	return IPPROTO_DONE;
58053541Sshin}
58153541Sshin
58253541Sshin/*
58353541Sshin * Free a fragment reassembly header and all
58453541Sshin * associated datagrams.
58553541Sshin */
58653541Sshinvoid
58753541Sshinfrag6_freef(q6)
58853541Sshin	struct ip6q *q6;
58953541Sshin{
59053541Sshin	struct ip6asfrag *af6, *down6;
59153541Sshin
592121355Sume	IP6Q_LOCK_ASSERT();
593121345Sume
59453541Sshin	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
59553541Sshin	     af6 = down6) {
59653541Sshin		struct mbuf *m = IP6_REASS_MBUF(af6);
59753541Sshin
59853541Sshin		down6 = af6->ip6af_down;
59953541Sshin		frag6_deq(af6);
60053541Sshin
60153541Sshin		/*
60253541Sshin		 * Return ICMP time exceeded error for the 1st fragment.
60353541Sshin		 * Just free other fragments.
60453541Sshin		 */
60553541Sshin		if (af6->ip6af_off == 0) {
60653541Sshin			struct ip6_hdr *ip6;
60753541Sshin
60853541Sshin			/* adjust pointer */
60953541Sshin			ip6 = mtod(m, struct ip6_hdr *);
61053541Sshin
611120891Sume			/* restore source and destination addresses */
61253541Sshin			ip6->ip6_src = q6->ip6q_src;
61353541Sshin			ip6->ip6_dst = q6->ip6q_dst;
61453541Sshin
61553541Sshin			icmp6_error(m, ICMP6_TIME_EXCEEDED,
61653541Sshin				    ICMP6_TIME_EXCEED_REASSEMBLY, 0);
61762587Sitojun		} else
61853541Sshin			m_freem(m);
61962587Sitojun		free(af6, M_FTABLE);
62053541Sshin	}
62153541Sshin	frag6_remque(q6);
622121345Sume	frag6_nfrags -= q6->ip6q_nfrag;
62353541Sshin	free(q6, M_FTABLE);
62453541Sshin	frag6_nfragpackets--;
62553541Sshin}
62653541Sshin
62753541Sshin/*
62853541Sshin * Put an ip fragment on a reassembly chain.
62953541Sshin * Like insque, but pointers in middle of structure.
63053541Sshin */
63153541Sshinvoid
63253541Sshinfrag6_enq(af6, up6)
63353541Sshin	struct ip6asfrag *af6, *up6;
63453541Sshin{
635121345Sume
636121355Sume	IP6Q_LOCK_ASSERT();
637121345Sume
63853541Sshin	af6->ip6af_up = up6;
63953541Sshin	af6->ip6af_down = up6->ip6af_down;
64053541Sshin	up6->ip6af_down->ip6af_up = af6;
64153541Sshin	up6->ip6af_down = af6;
64253541Sshin}
64353541Sshin
64453541Sshin/*
64553541Sshin * To frag6_enq as remque is to insque.
64653541Sshin */
64753541Sshinvoid
64853541Sshinfrag6_deq(af6)
64953541Sshin	struct ip6asfrag *af6;
65053541Sshin{
651121345Sume
652121355Sume	IP6Q_LOCK_ASSERT();
653121345Sume
65453541Sshin	af6->ip6af_up->ip6af_down = af6->ip6af_down;
65553541Sshin	af6->ip6af_down->ip6af_up = af6->ip6af_up;
65653541Sshin}
65753541Sshin
65853541Sshinvoid
65953541Sshinfrag6_insque(new, old)
66053541Sshin	struct ip6q *new, *old;
66153541Sshin{
662121345Sume
663121355Sume	IP6Q_LOCK_ASSERT();
664121345Sume
66553541Sshin	new->ip6q_prev = old;
66653541Sshin	new->ip6q_next = old->ip6q_next;
66753541Sshin	old->ip6q_next->ip6q_prev= new;
66853541Sshin	old->ip6q_next = new;
66953541Sshin}
67053541Sshin
67153541Sshinvoid
67253541Sshinfrag6_remque(p6)
67353541Sshin	struct ip6q *p6;
67453541Sshin{
675121345Sume
676121355Sume	IP6Q_LOCK_ASSERT();
677121345Sume
67853541Sshin	p6->ip6q_prev->ip6q_next = p6->ip6q_next;
67953541Sshin	p6->ip6q_next->ip6q_prev = p6->ip6q_prev;
68053541Sshin}
68153541Sshin
68253541Sshin/*
68378064Sume * IPv6 reassembling timer processing;
68453541Sshin * if a timer expires on a reassembly
68553541Sshin * queue, discard it.
68653541Sshin */
68753541Sshinvoid
68853541Sshinfrag6_slowtimo()
68953541Sshin{
69053541Sshin	struct ip6q *q6;
69153541Sshin
692158295Sbz#if 0
693158295Sbz	GIANT_REQUIRED;	/* XXX bz: ip6_forward_rt */
694158295Sbz#endif
695158295Sbz
696121345Sume	IP6Q_LOCK();
69753541Sshin	q6 = ip6q.ip6q_next;
69853541Sshin	if (q6)
69953541Sshin		while (q6 != &ip6q) {
70053541Sshin			--q6->ip6q_ttl;
70153541Sshin			q6 = q6->ip6q_next;
70253541Sshin			if (q6->ip6q_prev->ip6q_ttl == 0) {
70353541Sshin				ip6stat.ip6s_fragtimeout++;
70453541Sshin				/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
70553541Sshin				frag6_freef(q6->ip6q_prev);
70653541Sshin			}
70753541Sshin		}
70853541Sshin	/*
70953541Sshin	 * If we are over the maximum number of fragments
71053541Sshin	 * (due to the limit being lowered), drain off
71153541Sshin	 * enough to get down to the new limit.
71253541Sshin	 */
71378064Sume	while (frag6_nfragpackets > (u_int)ip6_maxfragpackets &&
71478064Sume	    ip6q.ip6q_prev) {
71553541Sshin		ip6stat.ip6s_fragoverflow++;
71653541Sshin		/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
71753541Sshin		frag6_freef(ip6q.ip6q_prev);
71853541Sshin	}
719121345Sume	IP6Q_UNLOCK();
72062587Sitojun
72162587Sitojun#if 0
72262587Sitojun	/*
72362587Sitojun	 * Routing changes might produce a better route than we last used;
72462587Sitojun	 * make sure we notice eventually, even if forwarding only for one
72562587Sitojun	 * destination and the cache is never replaced.
72662587Sitojun	 */
72762587Sitojun	if (ip6_forward_rt.ro_rt) {
72862587Sitojun		RTFREE(ip6_forward_rt.ro_rt);
72962587Sitojun		ip6_forward_rt.ro_rt = 0;
73062587Sitojun	}
73162587Sitojun	if (ipsrcchk_rt.ro_rt) {
73262587Sitojun		RTFREE(ipsrcchk_rt.ro_rt);
73362587Sitojun		ipsrcchk_rt.ro_rt = 0;
73462587Sitojun	}
73562587Sitojun#endif
73653541Sshin}
73753541Sshin
73853541Sshin/*
73953541Sshin * Drain off all datagram fragments.
74053541Sshin */
74153541Sshinvoid
74253541Sshinfrag6_drain()
74353541Sshin{
744121345Sume
745121346Sume	if (IP6Q_TRYLOCK() == 0)
74653541Sshin		return;
74753541Sshin	while (ip6q.ip6q_next != &ip6q) {
74853541Sshin		ip6stat.ip6s_fragdropped++;
74953541Sshin		/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
75053541Sshin		frag6_freef(ip6q.ip6q_next);
75153541Sshin	}
752121345Sume	IP6Q_UNLOCK();
75353541Sshin}
754