Home | History | Annotate | Line # | Download | only in netinet6
      1  1.78     ozaki /*	$NetBSD: frag6.c,v 1.78 2024/04/19 05:04:06 ozaki-r Exp $	*/
      2  1.18    itojun /*	$KAME: frag6.c,v 1.40 2002/05/27 21:40:31 itojun Exp $	*/
      3   1.3   thorpej 
      4   1.2    itojun /*
      5   1.2    itojun  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
      6   1.2    itojun  * All rights reserved.
      7  1.11    itojun  *
      8   1.2    itojun  * Redistribution and use in source and binary forms, with or without
      9   1.2    itojun  * modification, are permitted provided that the following conditions
     10   1.2    itojun  * are met:
     11   1.2    itojun  * 1. Redistributions of source code must retain the above copyright
     12   1.2    itojun  *    notice, this list of conditions and the following disclaimer.
     13   1.2    itojun  * 2. Redistributions in binary form must reproduce the above copyright
     14   1.2    itojun  *    notice, this list of conditions and the following disclaimer in the
     15   1.2    itojun  *    documentation and/or other materials provided with the distribution.
     16   1.2    itojun  * 3. Neither the name of the project nor the names of its contributors
     17   1.2    itojun  *    may be used to endorse or promote products derived from this software
     18   1.2    itojun  *    without specific prior written permission.
     19  1.11    itojun  *
     20   1.2    itojun  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
     21   1.2    itojun  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     22   1.2    itojun  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     23   1.2    itojun  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
     24   1.2    itojun  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     25   1.2    itojun  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     26   1.2    itojun  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     27   1.2    itojun  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     28   1.2    itojun  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     29   1.2    itojun  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     30   1.2    itojun  * SUCH DAMAGE.
     31   1.2    itojun  */
     32  1.16     lukem 
     33  1.16     lukem #include <sys/cdefs.h>
     34  1.78     ozaki __KERNEL_RCSID(0, "$NetBSD: frag6.c,v 1.78 2024/04/19 05:04:06 ozaki-r Exp $");
     35  1.60     ozaki 
     36  1.60     ozaki #ifdef _KERNEL_OPT
     37  1.60     ozaki #include "opt_net_mpsafe.h"
     38  1.60     ozaki #endif
     39   1.2    itojun 
     40   1.2    itojun #include <sys/param.h>
     41   1.2    itojun #include <sys/systm.h>
     42   1.2    itojun #include <sys/mbuf.h>
     43   1.2    itojun #include <sys/errno.h>
     44   1.2    itojun #include <sys/time.h>
     45  1.53     rmind #include <sys/kmem.h>
     46   1.2    itojun #include <sys/kernel.h>
     47   1.2    itojun #include <sys/syslog.h>
     48   1.2    itojun 
     49   1.2    itojun #include <net/if.h>
     50   1.2    itojun #include <net/route.h>
     51   1.2    itojun 
     52   1.2    itojun #include <netinet/in.h>
     53   1.2    itojun #include <netinet/in_var.h>
     54  1.10    itojun #include <netinet/ip6.h>
     55   1.2    itojun #include <netinet6/ip6_var.h>
     56  1.44   thorpej #include <netinet6/ip6_private.h>
     57  1.10    itojun #include <netinet/icmp6.h>
     58   1.2    itojun 
     59  1.62      maxv /*
     60  1.71      maxv  * IPv6 reassembly queue structure. Each fragment being reassembled is
     61  1.71      maxv  * attached to one of these structures.
     62  1.71      maxv  *
     63  1.71      maxv  * XXX: Would be better to use TAILQ.
     64  1.62      maxv  */
     65  1.62      maxv struct	ip6q {
     66  1.62      maxv 	u_int32_t	ip6q_head;
     67  1.62      maxv 	u_int16_t	ip6q_len;
     68  1.62      maxv 	u_int8_t	ip6q_nxt;	/* ip6f_nxt in first fragment */
     69  1.62      maxv 	u_int8_t	ip6q_hlim;
     70  1.62      maxv 	struct ip6asfrag *ip6q_down;
     71  1.62      maxv 	struct ip6asfrag *ip6q_up;
     72  1.62      maxv 	u_int32_t	ip6q_ident;
     73  1.62      maxv 	u_int8_t	ip6q_ttl;
     74  1.62      maxv 	struct in6_addr	ip6q_src, ip6q_dst;
     75  1.62      maxv 	struct ip6q	*ip6q_next;
     76  1.62      maxv 	struct ip6q	*ip6q_prev;
     77  1.62      maxv 	int		ip6q_unfrglen;	/* len of unfragmentable part */
     78  1.62      maxv 	int		ip6q_nfrag;	/* # of fragments */
     79  1.74      maxv 	int		ip6q_ipsec;	/* IPsec flags */
     80  1.62      maxv };
     81  1.62      maxv 
     82  1.62      maxv struct	ip6asfrag {
     83  1.62      maxv 	u_int32_t	ip6af_head;
     84  1.62      maxv 	u_int16_t	ip6af_len;
     85  1.62      maxv 	u_int8_t	ip6af_nxt;
     86  1.62      maxv 	u_int8_t	ip6af_hlim;
     87  1.62      maxv 	/* must not override the above members during reassembling */
     88  1.62      maxv 	struct ip6asfrag *ip6af_down;
     89  1.62      maxv 	struct ip6asfrag *ip6af_up;
     90  1.62      maxv 	struct mbuf	*ip6af_m;
     91  1.62      maxv 	int		ip6af_offset;	/* offset in ip6af_m to next header */
     92  1.62      maxv 	int		ip6af_frglen;	/* fragmentable part length */
     93  1.62      maxv 	int		ip6af_off;	/* fragment offset */
     94  1.66      maxv 	bool		ip6af_mff;	/* more fragment bit in frag off */
     95  1.62      maxv };
     96  1.62      maxv 
     97  1.39    dyoung static void frag6_enq(struct ip6asfrag *, struct ip6asfrag *);
     98  1.39    dyoung static void frag6_deq(struct ip6asfrag *);
     99  1.39    dyoung static void frag6_insque(struct ip6q *, struct ip6q *);
    100  1.39    dyoung static void frag6_remque(struct ip6q *);
    101  1.39    dyoung static void frag6_freef(struct ip6q *);
    102   1.2    itojun 
    103  1.49    dyoung static int frag6_drainwanted;
    104  1.49    dyoung 
    105  1.71      maxv static u_int frag6_nfragpackets;
    106  1.71      maxv static u_int frag6_nfrags;
    107  1.71      maxv static struct ip6q ip6q;	/* ip6 reassembly queue */
    108   1.2    itojun 
    109  1.62      maxv /* Protects ip6q */
    110  1.62      maxv static kmutex_t	frag6_lock __cacheline_aligned;
    111  1.17    itojun 
    112   1.2    itojun /*
    113   1.2    itojun  * Initialise reassembly queue and fragment identifier.
    114   1.2    itojun  */
    115   1.2    itojun void
    116  1.42      matt frag6_init(void)
    117   1.2    itojun {
    118   1.6    itojun 
    119   1.2    itojun 	ip6q.ip6q_next = ip6q.ip6q_prev = &ip6q;
    120  1.76     ozaki 	mutex_init(&frag6_lock, MUTEX_DEFAULT, IPL_NONE);
    121   1.2    itojun }
    122   1.2    itojun 
    123  1.77  christos static void
    124  1.77  christos frag6_dropfrag(struct ip6q *q6)
    125  1.77  christos {
    126  1.77  christos 	frag6_remque(q6);
    127  1.77  christos 	frag6_nfrags -= q6->ip6q_nfrag;
    128  1.77  christos 	kmem_intr_free(q6, sizeof(*q6));
    129  1.77  christos 	frag6_nfragpackets--;
    130  1.77  christos }
    131  1.77  christos 
    132   1.2    itojun /*
    133  1.53     rmind  * IPv6 fragment input.
    134  1.53     rmind  *
    135   1.9    itojun  * In RFC2460, fragment and reassembly rule do not agree with each other,
    136   1.9    itojun  * in terms of next header field handling in fragment header.
    137   1.9    itojun  * While the sender will use the same value for all of the fragmented packets,
    138   1.9    itojun  * receiver is suggested not to check the consistency.
    139   1.9    itojun  *
    140   1.9    itojun  * fragment rule (p20):
    141   1.9    itojun  *	(2) A Fragment header containing:
    142   1.9    itojun  *	The Next Header value that identifies the first header of
    143   1.9    itojun  *	the Fragmentable Part of the original packet.
    144   1.9    itojun  *		-> next header field is same for all fragments
    145   1.9    itojun  *
    146  1.11    itojun  * reassembly rule (p21):
    147   1.9    itojun  *	The Next Header field of the last header of the Unfragmentable
    148   1.9    itojun  *	Part is obtained from the Next Header field of the first
    149   1.9    itojun  *	fragment's Fragment header.
    150   1.9    itojun  *		-> should grab it from the first fragment only
    151   1.9    itojun  *
    152   1.9    itojun  * The following note also contradicts with fragment rule - noone is going to
    153   1.9    itojun  * send different fragment with different next header field.
    154   1.9    itojun  *
    155   1.9    itojun  * additional note (p22):
    156   1.9    itojun  *	The Next Header values in the Fragment headers of different
    157   1.9    itojun  *	fragments of the same original packet may differ.  Only the value
    158   1.9    itojun  *	from the Offset zero fragment packet is used for reassembly.
    159   1.9    itojun  *		-> should grab it from the first fragment only
    160   1.9    itojun  *
    161   1.9    itojun  * There is no explicit reason given in the RFC.  Historical reason maybe?
    162  1.70      maxv  *
    163  1.70      maxv  * XXX: It would be better to use a pool, rather than kmem.
    164   1.9    itojun  */
    165  1.53     rmind int
    166  1.53     rmind frag6_input(struct mbuf **mp, int *offp, int proto)
    167   1.2    itojun {
    168  1.40    dyoung 	struct rtentry *rt;
    169   1.2    itojun 	struct mbuf *m = *mp, *t;
    170   1.2    itojun 	struct ip6_hdr *ip6;
    171   1.2    itojun 	struct ip6_frag *ip6f;
    172   1.2    itojun 	struct ip6q *q6;
    173   1.9    itojun 	struct ip6asfrag *af6, *ip6af, *af6dwn;
    174   1.2    itojun 	int offset = *offp, nxt, i, next;
    175  1.74      maxv 	int ipsecflags = m->m_flags & (M_DECRYPTED|M_AUTHIPHDR);
    176   1.2    itojun 	int first_frag = 0;
    177   1.9    itojun 	int fragoff, frgpartlen;	/* must be larger than u_int16_t */
    178   1.7    itojun 	struct ifnet *dstifp;
    179  1.37    dyoung 	static struct route ro;
    180  1.37    dyoung 	union {
    181  1.37    dyoung 		struct sockaddr		dst;
    182  1.37    dyoung 		struct sockaddr_in6	dst6;
    183  1.37    dyoung 	} u;
    184   1.2    itojun 
    185   1.7    itojun 	ip6 = mtod(m, struct ip6_hdr *);
    186   1.7    itojun 	IP6_EXTHDR_GET(ip6f, struct ip6_frag *, m, offset, sizeof(*ip6f));
    187   1.7    itojun 	if (ip6f == NULL)
    188  1.53     rmind 		return IPPROTO_DONE;
    189   1.2    itojun 
    190   1.7    itojun 	dstifp = NULL;
    191   1.7    itojun 	/* find the destination interface of the packet. */
    192  1.37    dyoung 	sockaddr_in6_init(&u.dst6, &ip6->ip6_dst, 0, 0, 0);
    193  1.75     ozaki 	if ((rt = rtcache_lookup(&ro, &u.dst)) != NULL)
    194  1.40    dyoung 		dstifp = ((struct in6_ifaddr *)rt->rt_ifa)->ia_ifp;
    195   1.2    itojun 
    196   1.2    itojun 	/* jumbo payload can't contain a fragment header */
    197   1.2    itojun 	if (ip6->ip6_plen == 0) {
    198   1.2    itojun 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, offset);
    199   1.7    itojun 		in6_ifstat_inc(dstifp, ifs6_reass_fail);
    200  1.57     ozaki 		goto done;
    201   1.2    itojun 	}
    202   1.2    itojun 
    203   1.2    itojun 	/*
    204  1.64      maxv 	 * Check whether fragment packet's fragment length is non-zero and
    205  1.11    itojun 	 * multiple of 8 octets.
    206   1.2    itojun 	 * sizeof(struct ip6_frag) == 8
    207   1.2    itojun 	 * sizeof(struct ip6_hdr) = 40
    208   1.2    itojun 	 */
    209  1.78     ozaki 	frgpartlen = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen) - offset
    210  1.78     ozaki 	    - sizeof(struct ip6_frag);
    211  1.78     ozaki 	if ((frgpartlen == 0) ||
    212  1.78     ozaki 	    ((ip6f->ip6f_offlg & IP6F_MORE_FRAG) && (frgpartlen & 0x7) != 0)) {
    213  1.18    itojun 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
    214  1.18    itojun 		    offsetof(struct ip6_hdr, ip6_plen));
    215   1.7    itojun 		in6_ifstat_inc(dstifp, ifs6_reass_fail);
    216  1.57     ozaki 		goto done;
    217   1.2    itojun 	}
    218   1.2    itojun 
    219  1.44   thorpej 	IP6_STATINC(IP6_STAT_FRAGMENTS);
    220   1.7    itojun 	in6_ifstat_inc(dstifp, ifs6_reass_reqd);
    221  1.20    itojun 
    222   1.9    itojun 	/* offset now points to data portion */
    223   1.2    itojun 	offset += sizeof(struct ip6_frag);
    224   1.2    itojun 
    225  1.54  christos 	/*
    226  1.62      maxv 	 * RFC6946: A host that receives an IPv6 packet which includes
    227  1.62      maxv 	 * a Fragment Header with the "Fragment Offset" equal to 0 and
    228  1.55  christos 	 * the "M" bit equal to 0 MUST process such packet in isolation
    229  1.62      maxv 	 * from any other packets/fragments.
    230  1.71      maxv 	 *
    231  1.71      maxv 	 * XXX: Would be better to remove this fragment header entirely,
    232  1.71      maxv 	 * for us not to get confused later when looking back at the
    233  1.71      maxv 	 * previous headers in the chain.
    234  1.54  christos 	 */
    235  1.54  christos 	fragoff = ntohs(ip6f->ip6f_offlg & IP6F_OFF_MASK);
    236  1.54  christos 	if (fragoff == 0 && !(ip6f->ip6f_offlg & IP6F_MORE_FRAG)) {
    237  1.54  christos 		IP6_STATINC(IP6_STAT_REASSEMBLED);
    238  1.54  christos 		in6_ifstat_inc(dstifp, ifs6_reass_ok);
    239  1.54  christos 		*offp = offset;
    240  1.58     ozaki 		rtcache_unref(rt, &ro);
    241  1.62      maxv 		return ip6f->ip6f_nxt;
    242  1.54  christos 	}
    243  1.54  christos 
    244  1.50    zoltan 	mutex_enter(&frag6_lock);
    245  1.12    itojun 
    246  1.18    itojun 	/*
    247  1.18    itojun 	 * Enforce upper bound on number of fragments.
    248  1.18    itojun 	 * If maxfrag is 0, never accept fragments.
    249  1.18    itojun 	 * If maxfrag is -1, accept all fragments without limitation.
    250  1.18    itojun 	 */
    251  1.18    itojun 	if (ip6_maxfrags < 0)
    252  1.18    itojun 		;
    253  1.18    itojun 	else if (frag6_nfrags >= (u_int)ip6_maxfrags)
    254  1.18    itojun 		goto dropfrag;
    255  1.18    itojun 
    256   1.2    itojun 	for (q6 = ip6q.ip6q_next; q6 != &ip6q; q6 = q6->ip6q_next)
    257   1.2    itojun 		if (ip6f->ip6f_ident == q6->ip6q_ident &&
    258   1.2    itojun 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &q6->ip6q_src) &&
    259   1.2    itojun 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &q6->ip6q_dst))
    260   1.2    itojun 			break;
    261   1.2    itojun 
    262  1.74      maxv 	if (q6 != &ip6q) {
    263  1.74      maxv 		/* All fragments must have the same IPsec flags. */
    264  1.74      maxv 		if (q6->ip6q_ipsec != ipsecflags) {
    265  1.74      maxv 			goto dropfrag;
    266  1.74      maxv 		}
    267  1.74      maxv 	}
    268  1.74      maxv 
    269   1.2    itojun 	if (q6 == &ip6q) {
    270   1.2    itojun 		/*
    271   1.2    itojun 		 * the first fragment to arrive, create a reassembly queue.
    272   1.2    itojun 		 */
    273   1.2    itojun 		first_frag = 1;
    274   1.2    itojun 
    275   1.2    itojun 		/*
    276   1.2    itojun 		 * Enforce upper bound on number of fragmented packets
    277  1.11    itojun 		 * for which we attempt reassembly;
    278  1.18    itojun 		 * If maxfragpackets is 0, never accept fragments.
    279  1.18    itojun 		 * If maxfragpackets is -1, accept all fragments without
    280  1.18    itojun 		 * limitation.
    281   1.2    itojun 		 */
    282  1.13    itojun 		if (ip6_maxfragpackets < 0)
    283  1.13    itojun 			;
    284  1.13    itojun 		else if (frag6_nfragpackets >= (u_int)ip6_maxfragpackets)
    285  1.13    itojun 			goto dropfrag;
    286  1.13    itojun 		frag6_nfragpackets++;
    287  1.53     rmind 
    288  1.53     rmind 		q6 = kmem_intr_zalloc(sizeof(struct ip6q), KM_NOSLEEP);
    289  1.53     rmind 		if (q6 == NULL) {
    290   1.2    itojun 			goto dropfrag;
    291  1.53     rmind 		}
    292   1.2    itojun 		frag6_insque(q6, &ip6q);
    293   1.2    itojun 
    294   1.9    itojun 		/* ip6q_nxt will be filled afterwards, from 1st fragment */
    295   1.2    itojun 		q6->ip6q_down	= q6->ip6q_up = (struct ip6asfrag *)q6;
    296   1.2    itojun 		q6->ip6q_ident	= ip6f->ip6f_ident;
    297   1.2    itojun 		q6->ip6q_ttl 	= IPV6_FRAGTTL;
    298   1.2    itojun 		q6->ip6q_src	= ip6->ip6_src;
    299   1.2    itojun 		q6->ip6q_dst	= ip6->ip6_dst;
    300   1.2    itojun 		q6->ip6q_unfrglen = -1;	/* The 1st fragment has not arrived. */
    301  1.18    itojun 		q6->ip6q_nfrag = 0;
    302  1.74      maxv 		q6->ip6q_ipsec = ipsecflags;
    303   1.2    itojun 	}
    304   1.2    itojun 
    305   1.2    itojun 	/*
    306   1.2    itojun 	 * If it's the 1st fragment, record the length of the
    307   1.2    itojun 	 * unfragmentable part and the next header of the fragment header.
    308   1.2    itojun 	 */
    309   1.2    itojun 	if (fragoff == 0) {
    310  1.18    itojun 		q6->ip6q_unfrglen = offset - sizeof(struct ip6_hdr) -
    311  1.18    itojun 		    sizeof(struct ip6_frag);
    312   1.2    itojun 		q6->ip6q_nxt = ip6f->ip6f_nxt;
    313   1.2    itojun 	}
    314   1.2    itojun 
    315   1.2    itojun 	/*
    316   1.2    itojun 	 * Check that the reassembled packet would not exceed 65535 bytes
    317  1.62      maxv 	 * in size. If it would exceed, discard the fragment and return an
    318  1.62      maxv 	 * ICMP error.
    319   1.2    itojun 	 */
    320   1.2    itojun 	if (q6->ip6q_unfrglen >= 0) {
    321   1.2    itojun 		/* The 1st fragment has already arrived. */
    322   1.2    itojun 		if (q6->ip6q_unfrglen + fragoff + frgpartlen > IPV6_MAXPACKET) {
    323  1.50    zoltan 			mutex_exit(&frag6_lock);
    324   1.2    itojun 			icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
    325  1.18    itojun 			    offset - sizeof(struct ip6_frag) +
    326  1.18    itojun 			    offsetof(struct ip6_frag, ip6f_offlg));
    327  1.57     ozaki 			goto done;
    328   1.2    itojun 		}
    329  1.18    itojun 	} else if (fragoff + frgpartlen > IPV6_MAXPACKET) {
    330  1.50    zoltan 		mutex_exit(&frag6_lock);
    331   1.2    itojun 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
    332  1.62      maxv 		    offset - sizeof(struct ip6_frag) +
    333  1.62      maxv 		    offsetof(struct ip6_frag, ip6f_offlg));
    334  1.57     ozaki 		goto done;
    335   1.2    itojun 	}
    336  1.62      maxv 
    337   1.2    itojun 	/*
    338   1.2    itojun 	 * If it's the first fragment, do the above check for each
    339   1.2    itojun 	 * fragment already stored in the reassembly queue.
    340   1.2    itojun 	 */
    341   1.2    itojun 	if (fragoff == 0) {
    342   1.2    itojun 		for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
    343   1.2    itojun 		     af6 = af6dwn) {
    344   1.2    itojun 			af6dwn = af6->ip6af_down;
    345   1.2    itojun 
    346   1.2    itojun 			if (q6->ip6q_unfrglen + af6->ip6af_off + af6->ip6af_frglen >
    347   1.2    itojun 			    IPV6_MAXPACKET) {
    348  1.62      maxv 				struct mbuf *merr = af6->ip6af_m;
    349   1.2    itojun 				struct ip6_hdr *ip6err;
    350   1.2    itojun 				int erroff = af6->ip6af_offset;
    351   1.2    itojun 
    352   1.2    itojun 				/* dequeue the fragment. */
    353   1.2    itojun 				frag6_deq(af6);
    354  1.53     rmind 				kmem_intr_free(af6, sizeof(struct ip6asfrag));
    355   1.2    itojun 
    356   1.2    itojun 				/* adjust pointer. */
    357   1.2    itojun 				ip6err = mtod(merr, struct ip6_hdr *);
    358   1.2    itojun 
    359   1.2    itojun 				/*
    360   1.2    itojun 				 * Restore source and destination addresses
    361   1.2    itojun 				 * in the erroneous IPv6 header.
    362   1.2    itojun 				 */
    363   1.2    itojun 				ip6err->ip6_src = q6->ip6q_src;
    364   1.2    itojun 				ip6err->ip6_dst = q6->ip6q_dst;
    365   1.2    itojun 
    366   1.2    itojun 				icmp6_error(merr, ICMP6_PARAM_PROB,
    367  1.18    itojun 				    ICMP6_PARAMPROB_HEADER,
    368  1.18    itojun 				    erroff - sizeof(struct ip6_frag) +
    369  1.18    itojun 				    offsetof(struct ip6_frag, ip6f_offlg));
    370   1.2    itojun 			}
    371   1.2    itojun 		}
    372   1.2    itojun 	}
    373   1.2    itojun 
    374  1.53     rmind 	ip6af = kmem_intr_zalloc(sizeof(struct ip6asfrag), KM_NOSLEEP);
    375  1.53     rmind 	if (ip6af == NULL) {
    376   1.9    itojun 		goto dropfrag;
    377  1.53     rmind 	}
    378   1.9    itojun 	ip6af->ip6af_head = ip6->ip6_flow;
    379   1.9    itojun 	ip6af->ip6af_len = ip6->ip6_plen;
    380   1.9    itojun 	ip6af->ip6af_nxt = ip6->ip6_nxt;
    381   1.9    itojun 	ip6af->ip6af_hlim = ip6->ip6_hlim;
    382  1.66      maxv 	ip6af->ip6af_mff = (ip6f->ip6f_offlg & IP6F_MORE_FRAG) != 0;
    383   1.2    itojun 	ip6af->ip6af_off = fragoff;
    384   1.2    itojun 	ip6af->ip6af_frglen = frgpartlen;
    385   1.2    itojun 	ip6af->ip6af_offset = offset;
    386  1.62      maxv 	ip6af->ip6af_m = m;
    387   1.2    itojun 
    388   1.2    itojun 	if (first_frag) {
    389   1.2    itojun 		af6 = (struct ip6asfrag *)q6;
    390   1.2    itojun 		goto insert;
    391   1.2    itojun 	}
    392   1.2    itojun 
    393   1.2    itojun 	/*
    394   1.2    itojun 	 * Find a segment which begins after this one does.
    395   1.2    itojun 	 */
    396   1.2    itojun 	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
    397   1.2    itojun 	     af6 = af6->ip6af_down)
    398   1.2    itojun 		if (af6->ip6af_off > ip6af->ip6af_off)
    399   1.2    itojun 			break;
    400   1.2    itojun 
    401   1.2    itojun 	/*
    402  1.53     rmind 	 * If the incoming fragment overlaps some existing fragments in
    403  1.53     rmind 	 * the reassembly queue - drop it as per RFC 5722.
    404   1.2    itojun 	 */
    405   1.2    itojun 	if (af6->ip6af_up != (struct ip6asfrag *)q6) {
    406   1.2    itojun 		i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
    407   1.2    itojun 			- ip6af->ip6af_off;
    408   1.2    itojun 		if (i > 0) {
    409  1.53     rmind 			kmem_intr_free(ip6af, sizeof(struct ip6asfrag));
    410   1.2    itojun 			goto dropfrag;
    411   1.2    itojun 		}
    412   1.2    itojun 	}
    413   1.2    itojun 	if (af6 != (struct ip6asfrag *)q6) {
    414   1.2    itojun 		i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
    415   1.2    itojun 		if (i > 0) {
    416  1.53     rmind 			kmem_intr_free(ip6af, sizeof(struct ip6asfrag));
    417   1.2    itojun 			goto dropfrag;
    418   1.2    itojun 		}
    419   1.2    itojun 	}
    420   1.2    itojun 
    421   1.2    itojun insert:
    422   1.2    itojun 	/*
    423  1.63      maxv 	 * Stick new segment in its place.
    424   1.2    itojun 	 */
    425   1.2    itojun 	frag6_enq(ip6af, af6->ip6af_up);
    426  1.18    itojun 	frag6_nfrags++;
    427  1.18    itojun 	q6->ip6q_nfrag++;
    428  1.62      maxv 
    429  1.62      maxv 	/*
    430  1.62      maxv 	 * Check for complete reassembly.
    431  1.62      maxv 	 */
    432   1.2    itojun 	next = 0;
    433   1.2    itojun 	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
    434   1.2    itojun 	     af6 = af6->ip6af_down) {
    435   1.2    itojun 		if (af6->ip6af_off != next) {
    436  1.50    zoltan 			mutex_exit(&frag6_lock);
    437  1.57     ozaki 			goto done;
    438   1.2    itojun 		}
    439   1.2    itojun 		next += af6->ip6af_frglen;
    440   1.2    itojun 	}
    441  1.66      maxv 	if (af6->ip6af_up->ip6af_mff) {
    442  1.50    zoltan 		mutex_exit(&frag6_lock);
    443  1.57     ozaki 		goto done;
    444   1.2    itojun 	}
    445   1.2    itojun 
    446   1.2    itojun 	/*
    447   1.2    itojun 	 * Reassembly is complete; concatenate fragments.
    448   1.2    itojun 	 */
    449   1.2    itojun 	ip6af = q6->ip6q_down;
    450  1.62      maxv 	t = m = ip6af->ip6af_m;
    451   1.2    itojun 	af6 = ip6af->ip6af_down;
    452   1.9    itojun 	frag6_deq(ip6af);
    453   1.2    itojun 	while (af6 != (struct ip6asfrag *)q6) {
    454   1.9    itojun 		af6dwn = af6->ip6af_down;
    455   1.9    itojun 		frag6_deq(af6);
    456   1.2    itojun 		while (t->m_next)
    457   1.2    itojun 			t = t->m_next;
    458  1.62      maxv 		t->m_next = af6->ip6af_m;
    459   1.9    itojun 		m_adj(t->m_next, af6->ip6af_offset);
    460  1.73      maxv 		m_remove_pkthdr(t->m_next);
    461  1.53     rmind 		kmem_intr_free(af6, sizeof(struct ip6asfrag));
    462   1.9    itojun 		af6 = af6dwn;
    463   1.2    itojun 	}
    464   1.2    itojun 
    465   1.2    itojun 	/* adjust offset to point where the original next header starts */
    466   1.2    itojun 	offset = ip6af->ip6af_offset - sizeof(struct ip6_frag);
    467  1.53     rmind 	kmem_intr_free(ip6af, sizeof(struct ip6asfrag));
    468  1.77  christos 	next += offset - sizeof(struct ip6_hdr);
    469  1.77  christos 	if ((u_int)next > IPV6_MAXPACKET) {
    470  1.77  christos 		frag6_dropfrag(q6);
    471  1.77  christos 		goto dropfrag;
    472  1.77  christos 	}
    473   1.9    itojun 	ip6 = mtod(m, struct ip6_hdr *);
    474  1.77  christos 	ip6->ip6_plen = htons(next);
    475   1.2    itojun 	ip6->ip6_src = q6->ip6q_src;
    476   1.2    itojun 	ip6->ip6_dst = q6->ip6q_dst;
    477   1.2    itojun 	nxt = q6->ip6q_nxt;
    478   1.2    itojun 
    479   1.2    itojun 	/*
    480  1.62      maxv 	 * Delete frag6 header.
    481   1.2    itojun 	 */
    482  1.48   mlelstv 	if (m->m_len >= offset + sizeof(struct ip6_frag)) {
    483  1.36  christos 		memmove((char *)ip6 + sizeof(struct ip6_frag), ip6, offset);
    484   1.9    itojun 		m->m_data += sizeof(struct ip6_frag);
    485   1.9    itojun 		m->m_len -= sizeof(struct ip6_frag);
    486   1.9    itojun 	} else {
    487   1.9    itojun 		/* this comes with no copy if the boundary is on cluster */
    488   1.9    itojun 		if ((t = m_split(m, offset, M_DONTWAIT)) == NULL) {
    489  1.77  christos 			frag6_dropfrag(q6);
    490   1.9    itojun 			goto dropfrag;
    491   1.9    itojun 		}
    492   1.9    itojun 		m_adj(t, sizeof(struct ip6_frag));
    493   1.9    itojun 		m_cat(m, t);
    494   1.2    itojun 	}
    495   1.2    itojun 
    496  1.77  christos 	frag6_dropfrag(q6);
    497   1.2    itojun 
    498  1.67      maxv 	{
    499  1.67      maxv 		KASSERT(m->m_flags & M_PKTHDR);
    500   1.2    itojun 		int plen = 0;
    501  1.65      maxv 		for (t = m; t; t = t->m_next) {
    502   1.2    itojun 			plen += t->m_len;
    503  1.65      maxv 		}
    504   1.2    itojun 		m->m_pkthdr.len = plen;
    505  1.68      maxv 		/* XXX XXX: clear csum_flags? */
    506   1.2    itojun 	}
    507  1.20    itojun 
    508  1.65      maxv 	/*
    509  1.65      maxv 	 * Restore NXT to the original.
    510  1.65      maxv 	 */
    511  1.65      maxv 	{
    512  1.65      maxv 		const int prvnxt = ip6_get_prevhdr(m, offset);
    513  1.65      maxv 		uint8_t *prvnxtp;
    514  1.65      maxv 
    515  1.65      maxv 		IP6_EXTHDR_GET(prvnxtp, uint8_t *, m, prvnxt,
    516  1.65      maxv 		    sizeof(*prvnxtp));
    517  1.65      maxv 		if (prvnxtp == NULL) {
    518  1.65      maxv 			goto dropfrag;
    519  1.65      maxv 		}
    520  1.65      maxv 		*prvnxtp = nxt;
    521  1.65      maxv 	}
    522  1.65      maxv 
    523  1.44   thorpej 	IP6_STATINC(IP6_STAT_REASSEMBLED);
    524   1.7    itojun 	in6_ifstat_inc(dstifp, ifs6_reass_ok);
    525  1.58     ozaki 	rtcache_unref(rt, &ro);
    526  1.69      maxv 	mutex_exit(&frag6_lock);
    527   1.2    itojun 
    528   1.2    itojun 	/*
    529  1.69      maxv 	 * Tell launch routine the next header.
    530   1.2    itojun 	 */
    531   1.2    itojun 	*mp = m;
    532   1.2    itojun 	*offp = offset;
    533   1.2    itojun 	return nxt;
    534   1.2    itojun 
    535   1.2    itojun  dropfrag:
    536  1.50    zoltan 	mutex_exit(&frag6_lock);
    537   1.7    itojun 	in6_ifstat_inc(dstifp, ifs6_reass_fail);
    538  1.44   thorpej 	IP6_STATINC(IP6_STAT_FRAGDROPPED);
    539   1.2    itojun 	m_freem(m);
    540  1.57     ozaki  done:
    541  1.58     ozaki 	rtcache_unref(rt, &ro);
    542   1.2    itojun 	return IPPROTO_DONE;
    543   1.2    itojun }
    544   1.2    itojun 
    545  1.50    zoltan int
    546  1.50    zoltan ip6_reass_packet(struct mbuf **mp, int offset)
    547  1.50    zoltan {
    548  1.50    zoltan 
    549  1.53     rmind 	if (frag6_input(mp, &offset, IPPROTO_IPV6) == IPPROTO_DONE) {
    550  1.50    zoltan 		*mp = NULL;
    551  1.53     rmind 		return EINVAL;
    552  1.50    zoltan 	}
    553  1.53     rmind 	return 0;
    554  1.50    zoltan }
    555  1.50    zoltan 
    556   1.2    itojun /*
    557   1.2    itojun  * Free a fragment reassembly header and all
    558   1.2    itojun  * associated datagrams.
    559   1.2    itojun  */
    560  1.62      maxv static void
    561  1.38  christos frag6_freef(struct ip6q *q6)
    562   1.2    itojun {
    563   1.2    itojun 	struct ip6asfrag *af6, *down6;
    564   1.2    itojun 
    565  1.50    zoltan 	KASSERT(mutex_owned(&frag6_lock));
    566  1.17    itojun 
    567   1.2    itojun 	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
    568   1.2    itojun 	     af6 = down6) {
    569  1.62      maxv 		struct mbuf *m = af6->ip6af_m;
    570   1.2    itojun 
    571   1.2    itojun 		down6 = af6->ip6af_down;
    572   1.2    itojun 		frag6_deq(af6);
    573   1.2    itojun 
    574   1.2    itojun 		/*
    575   1.2    itojun 		 * Return ICMP time exceeded error for the 1st fragment.
    576   1.2    itojun 		 * Just free other fragments.
    577   1.2    itojun 		 */
    578   1.2    itojun 		if (af6->ip6af_off == 0) {
    579   1.2    itojun 			struct ip6_hdr *ip6;
    580   1.2    itojun 
    581   1.2    itojun 			/* adjust pointer */
    582   1.2    itojun 			ip6 = mtod(m, struct ip6_hdr *);
    583   1.2    itojun 
    584  1.63      maxv 			/* restore source and destination addresses */
    585   1.2    itojun 			ip6->ip6_src = q6->ip6q_src;
    586   1.2    itojun 			ip6->ip6_dst = q6->ip6q_dst;
    587   1.2    itojun 
    588   1.2    itojun 			icmp6_error(m, ICMP6_TIME_EXCEEDED,
    589   1.2    itojun 				    ICMP6_TIME_EXCEED_REASSEMBLY, 0);
    590  1.53     rmind 		} else {
    591   1.2    itojun 			m_freem(m);
    592  1.53     rmind 		}
    593  1.53     rmind 		kmem_intr_free(af6, sizeof(struct ip6asfrag));
    594   1.2    itojun 	}
    595  1.62      maxv 
    596  1.77  christos 	frag6_dropfrag(q6);
    597   1.2    itojun }
    598   1.2    itojun 
    599   1.2    itojun /*
    600   1.2    itojun  * Put an ip fragment on a reassembly chain.
    601   1.2    itojun  * Like insque, but pointers in middle of structure.
    602   1.2    itojun  */
    603   1.2    itojun void
    604  1.38  christos frag6_enq(struct ip6asfrag *af6, struct ip6asfrag *up6)
    605   1.2    itojun {
    606  1.17    itojun 
    607  1.50    zoltan 	KASSERT(mutex_owned(&frag6_lock));
    608  1.17    itojun 
    609   1.2    itojun 	af6->ip6af_up = up6;
    610   1.2    itojun 	af6->ip6af_down = up6->ip6af_down;
    611   1.2    itojun 	up6->ip6af_down->ip6af_up = af6;
    612   1.2    itojun 	up6->ip6af_down = af6;
    613   1.2    itojun }
    614   1.2    itojun 
    615   1.2    itojun /*
    616   1.2    itojun  * To frag6_enq as remque is to insque.
    617   1.2    itojun  */
    618   1.2    itojun void
    619  1.38  christos frag6_deq(struct ip6asfrag *af6)
    620   1.2    itojun {
    621  1.17    itojun 
    622  1.50    zoltan 	KASSERT(mutex_owned(&frag6_lock));
    623  1.17    itojun 
    624   1.2    itojun 	af6->ip6af_up->ip6af_down = af6->ip6af_down;
    625   1.2    itojun 	af6->ip6af_down->ip6af_up = af6->ip6af_up;
    626   1.2    itojun }
    627   1.2    itojun 
    628  1.62      maxv /*
    629  1.62      maxv  * Insert newq after oldq.
    630  1.62      maxv  */
    631  1.11    itojun void
    632  1.56      matt frag6_insque(struct ip6q *newq, struct ip6q *oldq)
    633   1.2    itojun {
    634  1.17    itojun 
    635  1.50    zoltan 	KASSERT(mutex_owned(&frag6_lock));
    636  1.17    itojun 
    637  1.56      matt 	newq->ip6q_prev = oldq;
    638  1.56      matt 	newq->ip6q_next = oldq->ip6q_next;
    639  1.62      maxv 	oldq->ip6q_next->ip6q_prev = newq;
    640  1.56      matt 	oldq->ip6q_next = newq;
    641   1.2    itojun }
    642   1.2    itojun 
    643  1.62      maxv /*
    644  1.62      maxv  * Unlink p6.
    645  1.62      maxv  */
    646   1.2    itojun void
    647  1.38  christos frag6_remque(struct ip6q *p6)
    648   1.2    itojun {
    649  1.17    itojun 
    650  1.50    zoltan 	KASSERT(mutex_owned(&frag6_lock));
    651  1.17    itojun 
    652   1.2    itojun 	p6->ip6q_prev->ip6q_next = p6->ip6q_next;
    653   1.2    itojun 	p6->ip6q_next->ip6q_prev = p6->ip6q_prev;
    654   1.2    itojun }
    655   1.2    itojun 
    656  1.49    dyoung void
    657  1.49    dyoung frag6_fasttimo(void)
    658  1.49    dyoung {
    659  1.60     ozaki 
    660  1.61     ozaki 	SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
    661  1.51  jakllsch 
    662  1.49    dyoung 	if (frag6_drainwanted) {
    663  1.49    dyoung 		frag6_drain();
    664  1.49    dyoung 		frag6_drainwanted = 0;
    665  1.49    dyoung 	}
    666  1.51  jakllsch 
    667  1.61     ozaki 	SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
    668  1.49    dyoung }
    669  1.49    dyoung 
    670   1.2    itojun /*
    671  1.11    itojun  * IPv6 reassembling timer processing;
    672   1.2    itojun  * if a timer expires on a reassembly
    673   1.2    itojun  * queue, discard it.
    674   1.2    itojun  */
    675   1.2    itojun void
    676  1.42      matt frag6_slowtimo(void)
    677   1.2    itojun {
    678   1.2    itojun 	struct ip6q *q6;
    679  1.45        ad 
    680  1.61     ozaki 	SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
    681  1.51  jakllsch 
    682  1.50    zoltan 	mutex_enter(&frag6_lock);
    683   1.2    itojun 	q6 = ip6q.ip6q_next;
    684  1.71      maxv 	if (q6) {
    685   1.2    itojun 		while (q6 != &ip6q) {
    686   1.2    itojun 			--q6->ip6q_ttl;
    687   1.2    itojun 			q6 = q6->ip6q_next;
    688   1.2    itojun 			if (q6->ip6q_prev->ip6q_ttl == 0) {
    689  1.44   thorpej 				IP6_STATINC(IP6_STAT_FRAGTIMEOUT);
    690   1.7    itojun 				/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
    691   1.2    itojun 				frag6_freef(q6->ip6q_prev);
    692   1.2    itojun 			}
    693   1.2    itojun 		}
    694  1.71      maxv 	}
    695  1.71      maxv 
    696   1.2    itojun 	/*
    697   1.2    itojun 	 * If we are over the maximum number of fragments
    698   1.2    itojun 	 * (due to the limit being lowered), drain off
    699   1.2    itojun 	 * enough to get down to the new limit.
    700   1.2    itojun 	 */
    701  1.13    itojun 	while (frag6_nfragpackets > (u_int)ip6_maxfragpackets &&
    702  1.13    itojun 	    ip6q.ip6q_prev) {
    703  1.44   thorpej 		IP6_STATINC(IP6_STAT_FRAGOVERFLOW);
    704   1.7    itojun 		/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
    705   1.2    itojun 		frag6_freef(ip6q.ip6q_prev);
    706   1.2    itojun 	}
    707  1.50    zoltan 	mutex_exit(&frag6_lock);
    708   1.2    itojun 
    709  1.61     ozaki 	SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
    710  1.51  jakllsch 
    711   1.2    itojun #if 0
    712   1.2    itojun 	/*
    713   1.2    itojun 	 * Routing changes might produce a better route than we last used;
    714   1.2    itojun 	 * make sure we notice eventually, even if forwarding only for one
    715   1.2    itojun 	 * destination and the cache is never replaced.
    716   1.2    itojun 	 */
    717  1.37    dyoung 	rtcache_free(&ip6_forward_rt);
    718  1.37    dyoung 	rtcache_free(&ipsrcchk_rt);
    719   1.2    itojun #endif
    720   1.2    itojun }
    721   1.2    itojun 
    722  1.49    dyoung void
    723  1.49    dyoung frag6_drainstub(void)
    724  1.49    dyoung {
    725  1.49    dyoung 	frag6_drainwanted = 1;
    726  1.49    dyoung }
    727  1.49    dyoung 
    728   1.2    itojun /*
    729   1.2    itojun  * Drain off all datagram fragments.
    730   1.2    itojun  */
    731   1.2    itojun void
    732  1.42      matt frag6_drain(void)
    733   1.2    itojun {
    734  1.17    itojun 
    735  1.50    zoltan 	if (mutex_tryenter(&frag6_lock)) {
    736  1.45        ad 		while (ip6q.ip6q_next != &ip6q) {
    737  1.45        ad 			IP6_STATINC(IP6_STAT_FRAGDROPPED);
    738  1.45        ad 			/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
    739  1.45        ad 			frag6_freef(ip6q.ip6q_next);
    740  1.45        ad 		}
    741  1.50    zoltan 		mutex_exit(&frag6_lock);
    742   1.2    itojun 	}
    743   1.2    itojun }
    744