2 /* $KAME: frag6.c,v 1.33 2002/01/07 11:34:48 kjc Exp $ */
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 #include <sys/param.h>
34 #include <sys/systm.h>
35 #include <sys/malloc.h>
37 #include <sys/domain.h>
38 #include <sys/protosw.h>
39 #include <sys/socket.h>
40 #include <sys/errno.h>
42 #include <sys/kernel.h>
43 #include <sys/syslog.h>
46 #include <net/route.h>
48 #include <netinet/in.h>
49 #include <netinet/in_var.h>
50 #include <netinet/ip6.h>
51 #include <netinet6/ip6_var.h>
52 #include <netinet/icmp6.h>
53 #include <netinet/in_systm.h> /* for ECN definitions */
54 #include <netinet/ip.h> /* for ECN definitions */
56 #include <net/net_osdep.h>
59 * Define it to get a correct behavior on per-interface statistics.
60 * You will need to perform an extra routing table lookup, per fragment,
61 * to do it. This may, or may not be, a performance hit.
63 #define IN6_IFSTAT_STRICT
65 static void frag6_enq __P((struct ip6asfrag *, struct ip6asfrag *));
66 static void frag6_deq __P((struct ip6asfrag *));
67 static void frag6_insque __P((struct ip6q *, struct ip6q *));
68 static void frag6_remque __P((struct ip6q *));
69 static void frag6_freef __P((struct ip6q *));
71 static struct mtx ip6qlock;
73 * These fields all protected by ip6qlock.
75 static u_int frag6_nfragpackets;
76 static u_int frag6_nfrags;
77 static struct ip6q ip6q; /* ip6 reassemble queue */
79 #define IP6Q_LOCK_INIT() mtx_init(&ip6qlock, "ip6qlock", NULL, MTX_DEF);
80 #define IP6Q_LOCK() mtx_lock(&ip6qlock)
81 #define IP6Q_TRYLOCK() mtx_trylock(&ip6qlock)
82 #define IP6Q_LOCK_ASSERT() mtx_assert(&ip6qlock, MA_OWNED)
83 #define IP6Q_UNLOCK() mtx_unlock(&ip6qlock)
85 static MALLOC_DEFINE(M_FTABLE, "fragment", "fragment reassembly header");
88 * Initialise reassembly queue and fragment identifier.
94 ip6_maxfragpackets = nmbclusters / 4;
95 ip6_maxfrags = nmbclusters / 4;
99 ip6q.ip6q_next = ip6q.ip6q_prev = &ip6q;
103 * In RFC2460, fragment and reassembly rule do not agree with each other,
104 * in terms of next header field handling in fragment header.
105 * While the sender will use the same value for all of the fragmented packets,
106 * receiver is suggested not to check the consistency.
108 * fragment rule (p20):
109 * (2) A Fragment header containing:
110 * The Next Header value that identifies the first header of
111 * the Fragmentable Part of the original packet.
112 * -> next header field is same for all fragments
114 * reassembly rule (p21):
115 * The Next Header field of the last header of the Unfragmentable
116 * Part is obtained from the Next Header field of the first
117 * fragment's Fragment header.
118 * -> should grab it from the first fragment only
120 * The following note also contradicts with fragment rule - noone is going to
121 * send different fragment with different next header field.
123 * additional note (p22):
124 * The Next Header values in the Fragment headers of different
125 * fragments of the same original packet may differ. Only the value
126 * from the Offset zero fragment packet is used for reassembly.
127 * -> should grab it from the first fragment only
129 * There is no explicit reason given in the RFC. Historical reason maybe?
135 frag6_input(mp, offp, proto)
139 struct mbuf *m = *mp, *t;
141 struct ip6_frag *ip6f;
143 struct ip6asfrag *af6, *ip6af, *af6dwn;
144 #ifdef IN6_IFSTAT_STRICT
145 struct in6_ifaddr *ia;
147 int offset = *offp, nxt, i, next;
149 int fragoff, frgpartlen; /* must be larger than u_int16_t */
150 struct ifnet *dstifp;
153 ip6 = mtod(m, struct ip6_hdr *);
154 #ifndef PULLDOWN_TEST
155 IP6_EXTHDR_CHECK(m, offset, sizeof(struct ip6_frag), IPPROTO_DONE);
156 ip6f = (struct ip6_frag *)((caddr_t)ip6 + offset);
158 IP6_EXTHDR_GET(ip6f, struct ip6_frag *, m, offset, sizeof(*ip6f));
160 return (IPPROTO_DONE);
164 #ifdef IN6_IFSTAT_STRICT
165 /* find the destination interface of the packet. */
166 if ((ia = ip6_getdstifaddr(m)) != NULL)
169 /* we are violating the spec, this is not the destination interface */
170 if ((m->m_flags & M_PKTHDR) != 0)
171 dstifp = m->m_pkthdr.rcvif;
174 /* jumbo payload can't contain a fragment header */
175 if (ip6->ip6_plen == 0) {
176 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, offset);
177 in6_ifstat_inc(dstifp, ifs6_reass_fail);
182 * check whether fragment packet's fragment length is
183 * multiple of 8 octets.
184 * sizeof(struct ip6_frag) == 8
185 * sizeof(struct ip6_hdr) = 40
187 if ((ip6f->ip6f_offlg & IP6F_MORE_FRAG) &&
188 (((ntohs(ip6->ip6_plen) - offset) & 0x7) != 0)) {
189 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
190 offsetof(struct ip6_hdr, ip6_plen));
191 in6_ifstat_inc(dstifp, ifs6_reass_fail);
195 ip6stat.ip6s_fragments++;
196 in6_ifstat_inc(dstifp, ifs6_reass_reqd);
198 /* offset now points to data portion */
199 offset += sizeof(struct ip6_frag);
204 * Enforce upper bound on number of fragments.
205 * If maxfrag is 0, never accept fragments.
206 * If maxfrag is -1, accept all fragments without limitation.
208 if (ip6_maxfrags < 0)
210 else if (frag6_nfrags >= (u_int)ip6_maxfrags)
213 for (q6 = ip6q.ip6q_next; q6 != &ip6q; q6 = q6->ip6q_next)
214 if (ip6f->ip6f_ident == q6->ip6q_ident &&
215 IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &q6->ip6q_src) &&
216 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &q6->ip6q_dst))
221 * the first fragment to arrive, create a reassembly queue.
226 * Enforce upper bound on number of fragmented packets
227 * for which we attempt reassembly;
228 * If maxfragpackets is 0, never accept fragments.
229 * If maxfragpackets is -1, accept all fragments without
232 if (ip6_maxfragpackets < 0)
234 else if (frag6_nfragpackets >= (u_int)ip6_maxfragpackets)
236 frag6_nfragpackets++;
237 q6 = (struct ip6q *)malloc(sizeof(struct ip6q), M_FTABLE,
241 bzero(q6, sizeof(*q6));
243 frag6_insque(q6, &ip6q);
245 /* ip6q_nxt will be filled afterwards, from 1st fragment */
246 q6->ip6q_down = q6->ip6q_up = (struct ip6asfrag *)q6;
248 q6->ip6q_nxtp = (u_char *)nxtp;
250 q6->ip6q_ident = ip6f->ip6f_ident;
251 q6->ip6q_arrive = 0; /* Is it used anywhere? */
252 q6->ip6q_ttl = IPV6_FRAGTTL;
253 q6->ip6q_src = ip6->ip6_src;
254 q6->ip6q_dst = ip6->ip6_dst;
255 q6->ip6q_unfrglen = -1; /* The 1st fragment has not arrived. */
261 * If it's the 1st fragment, record the length of the
262 * unfragmentable part and the next header of the fragment header.
264 fragoff = ntohs(ip6f->ip6f_offlg & IP6F_OFF_MASK);
266 q6->ip6q_unfrglen = offset - sizeof(struct ip6_hdr) -
267 sizeof(struct ip6_frag);
268 q6->ip6q_nxt = ip6f->ip6f_nxt;
272 * Check that the reassembled packet would not exceed 65535 bytes
274 * If it would exceed, discard the fragment and return an ICMP error.
276 frgpartlen = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen) - offset;
277 if (q6->ip6q_unfrglen >= 0) {
278 /* The 1st fragment has already arrived. */
279 if (q6->ip6q_unfrglen + fragoff + frgpartlen > IPV6_MAXPACKET) {
280 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
281 offset - sizeof(struct ip6_frag) +
282 offsetof(struct ip6_frag, ip6f_offlg));
284 return (IPPROTO_DONE);
286 } else if (fragoff + frgpartlen > IPV6_MAXPACKET) {
287 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
288 offset - sizeof(struct ip6_frag) +
289 offsetof(struct ip6_frag, ip6f_offlg));
291 return (IPPROTO_DONE);
294 * If it's the first fragment, do the above check for each
295 * fragment already stored in the reassembly queue.
298 for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
300 af6dwn = af6->ip6af_down;
302 if (q6->ip6q_unfrglen + af6->ip6af_off + af6->ip6af_frglen >
304 struct mbuf *merr = IP6_REASS_MBUF(af6);
305 struct ip6_hdr *ip6err;
306 int erroff = af6->ip6af_offset;
308 /* dequeue the fragment. */
312 /* adjust pointer. */
313 ip6err = mtod(merr, struct ip6_hdr *);
316 * Restore source and destination addresses
317 * in the erroneous IPv6 header.
319 ip6err->ip6_src = q6->ip6q_src;
320 ip6err->ip6_dst = q6->ip6q_dst;
322 icmp6_error(merr, ICMP6_PARAM_PROB,
323 ICMP6_PARAMPROB_HEADER,
324 erroff - sizeof(struct ip6_frag) +
325 offsetof(struct ip6_frag, ip6f_offlg));
330 ip6af = (struct ip6asfrag *)malloc(sizeof(struct ip6asfrag), M_FTABLE,
334 bzero(ip6af, sizeof(*ip6af));
335 ip6af->ip6af_head = ip6->ip6_flow;
336 ip6af->ip6af_len = ip6->ip6_plen;
337 ip6af->ip6af_nxt = ip6->ip6_nxt;
338 ip6af->ip6af_hlim = ip6->ip6_hlim;
339 ip6af->ip6af_mff = ip6f->ip6f_offlg & IP6F_MORE_FRAG;
340 ip6af->ip6af_off = fragoff;
341 ip6af->ip6af_frglen = frgpartlen;
342 ip6af->ip6af_offset = offset;
343 IP6_REASS_MBUF(ip6af) = m;
346 af6 = (struct ip6asfrag *)q6;
351 * Handle ECN by comparing this segment with the first one;
352 * if CE is set, do not lose CE.
353 * drop if CE and not-ECT are mixed for the same packet.
355 ecn = (ntohl(ip6->ip6_flow) >> 20) & IPTOS_ECN_MASK;
356 ecn0 = (ntohl(q6->ip6q_down->ip6af_head) >> 20) & IPTOS_ECN_MASK;
357 if (ecn == IPTOS_ECN_CE) {
358 if (ecn0 == IPTOS_ECN_NOTECT) {
359 free(ip6af, M_FTABLE);
362 if (ecn0 != IPTOS_ECN_CE)
363 q6->ip6q_down->ip6af_head |= htonl(IPTOS_ECN_CE << 20);
365 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) {
366 free(ip6af, M_FTABLE);
371 * Find a segment which begins after this one does.
373 for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
374 af6 = af6->ip6af_down)
375 if (af6->ip6af_off > ip6af->ip6af_off)
380 * If there is a preceding segment, it may provide some of
381 * our data already. If so, drop the data from the incoming
382 * segment. If it provides all of our data, drop us.
384 if (af6->ip6af_up != (struct ip6asfrag *)q6) {
385 i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
388 if (i >= ip6af->ip6af_frglen)
390 m_adj(IP6_REASS_MBUF(ip6af), i);
391 ip6af->ip6af_off += i;
392 ip6af->ip6af_frglen -= i;
397 * While we overlap succeeding segments trim them or,
398 * if they are completely covered, dequeue them.
400 while (af6 != (struct ip6asfrag *)q6 &&
401 ip6af->ip6af_off + ip6af->ip6af_frglen > af6->ip6af_off) {
402 i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
403 if (i < af6->ip6af_frglen) {
404 af6->ip6af_frglen -= i;
406 m_adj(IP6_REASS_MBUF(af6), i);
409 af6 = af6->ip6af_down;
410 m_freem(IP6_REASS_MBUF(af6->ip6af_up));
411 frag6_deq(af6->ip6af_up);
415 * If the incoming framgent overlaps some existing fragments in
416 * the reassembly queue, drop it, since it is dangerous to override
417 * existing fragments from a security point of view.
418 * We don't know which fragment is the bad guy - here we trust
419 * fragment that came in earlier, with no real reason.
421 if (af6->ip6af_up != (struct ip6asfrag *)q6) {
422 i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
425 #if 0 /* suppress the noisy log */
426 log(LOG_ERR, "%d bytes of a fragment from %s "
427 "overlaps the previous fragment\n",
428 i, ip6_sprintf(&q6->ip6q_src));
430 free(ip6af, M_FTABLE);
434 if (af6 != (struct ip6asfrag *)q6) {
435 i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
437 #if 0 /* suppress the noisy log */
438 log(LOG_ERR, "%d bytes of a fragment from %s "
439 "overlaps the succeeding fragment",
440 i, ip6_sprintf(&q6->ip6q_src));
442 free(ip6af, M_FTABLE);
451 * Stick new segment in its place;
452 * check for complete reassembly.
453 * Move to front of packet queue, as we are
454 * the most recently active fragmented packet.
456 frag6_enq(ip6af, af6->ip6af_up);
460 if (q6 != ip6q.ip6q_next) {
462 frag6_insque(q6, &ip6q);
466 for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
467 af6 = af6->ip6af_down) {
468 if (af6->ip6af_off != next) {
472 next += af6->ip6af_frglen;
474 if (af6->ip6af_up->ip6af_mff) {
480 * Reassembly is complete; concatenate fragments.
482 ip6af = q6->ip6q_down;
483 t = m = IP6_REASS_MBUF(ip6af);
484 af6 = ip6af->ip6af_down;
486 while (af6 != (struct ip6asfrag *)q6) {
487 af6dwn = af6->ip6af_down;
491 t->m_next = IP6_REASS_MBUF(af6);
492 m_adj(t->m_next, af6->ip6af_offset);
497 /* adjust offset to point where the original next header starts */
498 offset = ip6af->ip6af_offset - sizeof(struct ip6_frag);
499 free(ip6af, M_FTABLE);
500 ip6 = mtod(m, struct ip6_hdr *);
501 ip6->ip6_plen = htons((u_short)next + offset - sizeof(struct ip6_hdr));
502 ip6->ip6_src = q6->ip6q_src;
503 ip6->ip6_dst = q6->ip6q_dst;
506 *q6->ip6q_nxtp = (u_char)(nxt & 0xff);
510 * Delete frag6 header with as a few cost as possible.
512 if (offset < m->m_len) {
513 ovbcopy((caddr_t)ip6, (caddr_t)ip6 + sizeof(struct ip6_frag),
515 m->m_data += sizeof(struct ip6_frag);
516 m->m_len -= sizeof(struct ip6_frag);
518 /* this comes with no copy if the boundary is on cluster */
519 if ((t = m_split(m, offset, M_DONTWAIT)) == NULL) {
521 frag6_nfrags -= q6->ip6q_nfrag;
523 frag6_nfragpackets--;
526 m_adj(t, sizeof(struct ip6_frag));
531 * Store NXT to the original.
534 char *prvnxtp = ip6_get_prevhdr(m, offset); /* XXX */
539 frag6_nfrags -= q6->ip6q_nfrag;
541 frag6_nfragpackets--;
543 if (m->m_flags & M_PKTHDR) { /* Isn't it always true? */
545 for (t = m; t; t = t->m_next)
547 m->m_pkthdr.len = plen;
550 ip6stat.ip6s_reassembled++;
551 in6_ifstat_inc(dstifp, ifs6_reass_ok);
554 * Tell launch routine the next header
565 in6_ifstat_inc(dstifp, ifs6_reass_fail);
566 ip6stat.ip6s_fragdropped++;
572 * Free a fragment reassembly header and all
573 * associated datagrams.
579 struct ip6asfrag *af6, *down6;
583 for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
585 struct mbuf *m = IP6_REASS_MBUF(af6);
587 down6 = af6->ip6af_down;
591 * Return ICMP time exceeded error for the 1st fragment.
592 * Just free other fragments.
594 if (af6->ip6af_off == 0) {
598 ip6 = mtod(m, struct ip6_hdr *);
600 /* restore source and destination addresses */
601 ip6->ip6_src = q6->ip6q_src;
602 ip6->ip6_dst = q6->ip6q_dst;
604 icmp6_error(m, ICMP6_TIME_EXCEEDED,
605 ICMP6_TIME_EXCEED_REASSEMBLY, 0);
611 frag6_nfrags -= q6->ip6q_nfrag;
613 frag6_nfragpackets--;
617 * Put an ip fragment on a reassembly chain.
618 * Like insque, but pointers in middle of structure.
622 struct ip6asfrag *af6, *up6;
628 af6->ip6af_down = up6->ip6af_down;
629 up6->ip6af_down->ip6af_up = af6;
630 up6->ip6af_down = af6;
634 * To frag6_enq as remque is to insque.
638 struct ip6asfrag *af6;
643 af6->ip6af_up->ip6af_down = af6->ip6af_down;
644 af6->ip6af_down->ip6af_up = af6->ip6af_up;
648 frag6_insque(new, old)
649 struct ip6q *new, *old;
654 new->ip6q_prev = old;
655 new->ip6q_next = old->ip6q_next;
656 old->ip6q_next->ip6q_prev= new;
657 old->ip6q_next = new;
667 p6->ip6q_prev->ip6q_next = p6->ip6q_next;
668 p6->ip6q_next->ip6q_prev = p6->ip6q_prev;
672 * IPv6 reassembling timer processing;
673 * if a timer expires on a reassembly
685 while (q6 != &ip6q) {
688 if (q6->ip6q_prev->ip6q_ttl == 0) {
689 ip6stat.ip6s_fragtimeout++;
690 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
691 frag6_freef(q6->ip6q_prev);
695 * If we are over the maximum number of fragments
696 * (due to the limit being lowered), drain off
697 * enough to get down to the new limit.
699 while (frag6_nfragpackets > (u_int)ip6_maxfragpackets &&
701 ip6stat.ip6s_fragoverflow++;
702 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
703 frag6_freef(ip6q.ip6q_prev);
709 * Routing changes might produce a better route than we last used;
710 * make sure we notice eventually, even if forwarding only for one
711 * destination and the cache is never replaced.
713 if (ip6_forward_rt.ro_rt) {
714 RTFREE(ip6_forward_rt.ro_rt);
715 ip6_forward_rt.ro_rt = 0;
717 if (ipsrcchk_rt.ro_rt) {
718 RTFREE(ipsrcchk_rt.ro_rt);
719 ipsrcchk_rt.ro_rt = 0;
727 * Drain off all datagram fragments.
733 if (IP6Q_TRYLOCK() == 0)
735 while (ip6q.ip6q_next != &ip6q) {
736 ip6stat.ip6s_fragdropped++;
737 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
738 frag6_freef(ip6q.ip6q_next);