1 /* $FreeBSD: src/sys/netinet6/frag6.c,v 1.2.2.6 2002/04/28 05:40:26 suz Exp $ */
2 /* $DragonFly: src/sys/netinet6/frag6.c,v 1.12 2008/01/05 14:02:40 swildner Exp $ */
3 /* $KAME: frag6.c,v 1.33 2002/01/07 11:34:48 kjc Exp $ */
6 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the project nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 #include <sys/param.h>
35 #include <sys/systm.h>
36 #include <sys/malloc.h>
38 #include <sys/domain.h>
39 #include <sys/protosw.h>
40 #include <sys/socket.h>
41 #include <sys/errno.h>
43 #include <sys/kernel.h>
44 #include <sys/syslog.h>
45 #include <sys/thread2.h>
48 #include <net/route.h>
50 #include <netinet/in.h>
51 #include <netinet/in_var.h>
52 #include <netinet/ip6.h>
53 #include <netinet6/ip6_var.h>
54 #include <netinet/icmp6.h>
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 (struct ip6asfrag *, struct ip6asfrag *);
66 static void frag6_deq (struct ip6asfrag *);
67 static void frag6_insque (struct ip6q *, struct ip6q *);
68 static void frag6_remque (struct ip6q *);
69 static void frag6_freef (struct ip6q *);
71 /* XXX we eventually need splreass6, or some real semaphore */
72 int frag6_doing_reass;
73 u_int frag6_nfragpackets;
74 struct ip6q ip6q; /* ip6 reassemble queue */
77 MALLOC_DEFINE(M_FTABLE, "fragment", "fragment reassembly header");
80 * Initialise reassembly queue and fragment identifier.
87 ip6_maxfragpackets = nmbclusters / 4;
90 * in many cases, random() here does NOT return random number
91 * as initialization during bootstrap time occur in fixed order.
94 ip6_id = krandom() ^ tv.tv_usec;
95 ip6q.ip6q_next = ip6q.ip6q_prev = &ip6q;
99 * In RFC2460, fragment and reassembly rule do not agree with each other,
100 * in terms of next header field handling in fragment header.
101 * While the sender will use the same value for all of the fragmented packets,
102 * receiver is suggested not to check the consistency.
104 * fragment rule (p20):
105 * (2) A Fragment header containing:
106 * The Next Header value that identifies the first header of
107 * the Fragmentable Part of the original packet.
108 * -> next header field is same for all fragments
110 * reassembly rule (p21):
111 * The Next Header field of the last header of the Unfragmentable
112 * Part is obtained from the Next Header field of the first
113 * fragment's Fragment header.
114 * -> should grab it from the first fragment only
116 * The following note also contradicts with fragment rule - noone is going to
117 * send different fragment with different next header field.
119 * additional note (p22):
120 * The Next Header values in the Fragment headers of different
121 * fragments of the same original packet may differ. Only the value
122 * from the Offset zero fragment packet is used for reassembly.
123 * -> should grab it from the first fragment only
125 * There is no explicit reason given in the RFC. Historical reason maybe?
131 frag6_input(struct mbuf **mp, int *offp, int proto)
133 struct mbuf *m = *mp, *t;
135 struct ip6_frag *ip6f;
137 struct ip6asfrag *af6, *ip6af, *af6dwn;
138 int offset = *offp, nxt, i, next;
140 int fragoff, frgpartlen; /* must be larger than u_int16_t */
141 struct ifnet *dstifp;
142 #ifdef IN6_IFSTAT_STRICT
143 static struct route_in6 ro;
144 struct sockaddr_in6 *dst;
147 ip6 = mtod(m, struct ip6_hdr *);
148 #ifndef PULLDOWN_TEST
149 IP6_EXTHDR_CHECK(m, offset, sizeof(struct ip6_frag), IPPROTO_DONE);
150 ip6f = (struct ip6_frag *)((caddr_t)ip6 + offset);
152 IP6_EXTHDR_GET(ip6f, struct ip6_frag *, m, offset, sizeof(*ip6f));
158 #ifdef IN6_IFSTAT_STRICT
159 /* find the destination interface of the packet. */
160 dst = (struct sockaddr_in6 *)&ro.ro_dst;
162 (!(ro.ro_rt->rt_flags & RTF_UP) ||
163 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
167 if (ro.ro_rt == NULL) {
168 bzero(dst, sizeof(*dst));
169 dst->sin6_family = AF_INET6;
170 dst->sin6_len = sizeof(struct sockaddr_in6);
171 dst->sin6_addr = ip6->ip6_dst;
173 rtalloc((struct route *)&ro);
174 if (ro.ro_rt != NULL && ro.ro_rt->rt_ifa != NULL)
175 dstifp = ((struct in6_ifaddr *)ro.ro_rt->rt_ifa)->ia_ifp;
177 /* we are violating the spec, this is not the destination interface */
178 if (m->m_flags & M_PKTHDR)
179 dstifp = m->m_pkthdr.rcvif;
182 /* jumbo payload can't contain a fragment header */
183 if (ip6->ip6_plen == 0) {
184 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, offset);
185 in6_ifstat_inc(dstifp, ifs6_reass_fail);
190 * check whether fragment packet's fragment length is
191 * multiple of 8 octets.
192 * sizeof(struct ip6_frag) == 8
193 * sizeof(struct ip6_hdr) = 40
195 if ((ip6f->ip6f_offlg & IP6F_MORE_FRAG) &&
196 (((ntohs(ip6->ip6_plen) - offset) & 0x7) != 0)) {
197 icmp6_error(m, ICMP6_PARAM_PROB,
198 ICMP6_PARAMPROB_HEADER,
199 offsetof(struct ip6_hdr, ip6_plen));
200 in6_ifstat_inc(dstifp, ifs6_reass_fail);
204 ip6stat.ip6s_fragments++;
205 in6_ifstat_inc(dstifp, ifs6_reass_reqd);
207 /* offset now points to data portion */
208 offset += sizeof(struct ip6_frag);
210 frag6_doing_reass = 1;
212 for (q6 = ip6q.ip6q_next; q6 != &ip6q; q6 = q6->ip6q_next)
213 if (ip6f->ip6f_ident == q6->ip6q_ident &&
214 IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &q6->ip6q_src) &&
215 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &q6->ip6q_dst))
220 * the first fragment to arrive, create a reassembly queue.
225 * Enforce upper bound on number of fragmented packets
226 * for which we attempt reassembly;
227 * If maxfrag is 0, never accept fragments.
228 * If maxfrag is -1, accept all fragments without limitation.
230 if (ip6_maxfragpackets < 0)
232 else if (frag6_nfragpackets >= (u_int)ip6_maxfragpackets)
234 frag6_nfragpackets++;
235 q6 = (struct ip6q *)kmalloc(sizeof(struct ip6q), M_FTABLE,
240 frag6_insque(q6, &ip6q);
242 /* ip6q_nxt will be filled afterwards, from 1st fragment */
243 q6->ip6q_down = q6->ip6q_up = (struct ip6asfrag *)q6;
245 q6->ip6q_nxtp = (u_char *)nxtp;
247 q6->ip6q_ident = ip6f->ip6f_ident;
248 q6->ip6q_arrive = 0; /* Is it used anywhere? */
249 q6->ip6q_ttl = IPV6_FRAGTTL;
250 q6->ip6q_src = ip6->ip6_src;
251 q6->ip6q_dst = ip6->ip6_dst;
252 q6->ip6q_unfrglen = -1; /* The 1st fragment has not arrived. */
256 * If it's the 1st fragment, record the length of the
257 * unfragmentable part and the next header of the fragment header.
259 fragoff = ntohs(ip6f->ip6f_offlg & IP6F_OFF_MASK);
261 q6->ip6q_unfrglen = offset - sizeof(struct ip6_hdr)
262 - sizeof(struct ip6_frag);
263 q6->ip6q_nxt = ip6f->ip6f_nxt;
267 * Check that the reassembled packet would not exceed 65535 bytes
269 * If it would exceed, discard the fragment and return an ICMP error.
271 frgpartlen = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen) - offset;
272 if (q6->ip6q_unfrglen >= 0) {
273 /* The 1st fragment has already arrived. */
274 if (q6->ip6q_unfrglen + fragoff + frgpartlen > IPV6_MAXPACKET) {
275 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
276 offset - sizeof(struct ip6_frag) +
277 offsetof(struct ip6_frag, ip6f_offlg));
278 frag6_doing_reass = 0;
279 return (IPPROTO_DONE);
282 else if (fragoff + frgpartlen > IPV6_MAXPACKET) {
283 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
284 offset - sizeof(struct ip6_frag) +
285 offsetof(struct ip6_frag, ip6f_offlg));
286 frag6_doing_reass = 0;
287 return (IPPROTO_DONE);
290 * If it's the first fragment, do the above check for each
291 * fragment already stored in the reassembly queue.
294 for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
296 af6dwn = af6->ip6af_down;
298 if (q6->ip6q_unfrglen + af6->ip6af_off + af6->ip6af_frglen >
300 struct mbuf *merr = IP6_REASS_MBUF(af6);
301 struct ip6_hdr *ip6err;
302 int erroff = af6->ip6af_offset;
304 /* dequeue the fragment. */
306 kfree(af6, M_FTABLE);
308 /* adjust pointer. */
309 ip6err = mtod(merr, struct ip6_hdr *);
312 * Restore source and destination addresses
313 * in the erroneous IPv6 header.
315 ip6err->ip6_src = q6->ip6q_src;
316 ip6err->ip6_dst = q6->ip6q_dst;
318 icmp6_error(merr, ICMP6_PARAM_PROB,
319 ICMP6_PARAMPROB_HEADER,
320 erroff - sizeof(struct ip6_frag) +
321 offsetof(struct ip6_frag, ip6f_offlg));
326 ip6af = (struct ip6asfrag *)kmalloc(sizeof(struct ip6asfrag), M_FTABLE,
330 ip6af->ip6af_head = ip6->ip6_flow;
331 ip6af->ip6af_len = ip6->ip6_plen;
332 ip6af->ip6af_nxt = ip6->ip6_nxt;
333 ip6af->ip6af_hlim = ip6->ip6_hlim;
334 ip6af->ip6af_mff = ip6f->ip6f_offlg & IP6F_MORE_FRAG;
335 ip6af->ip6af_off = fragoff;
336 ip6af->ip6af_frglen = frgpartlen;
337 ip6af->ip6af_offset = offset;
338 IP6_REASS_MBUF(ip6af) = m;
341 af6 = (struct ip6asfrag *)q6;
346 * Find a segment which begins after this one does.
348 for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
349 af6 = af6->ip6af_down)
350 if (af6->ip6af_off > ip6af->ip6af_off)
355 * If there is a preceding segment, it may provide some of
356 * our data already. If so, drop the data from the incoming
357 * segment. If it provides all of our data, drop us.
359 if (af6->ip6af_up != (struct ip6asfrag *)q6) {
360 i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
363 if (i >= ip6af->ip6af_frglen)
365 m_adj(IP6_REASS_MBUF(ip6af), i);
366 ip6af->ip6af_off += i;
367 ip6af->ip6af_frglen -= i;
372 * While we overlap succeeding segments trim them or,
373 * if they are completely covered, dequeue them.
375 while (af6 != (struct ip6asfrag *)q6 &&
376 ip6af->ip6af_off + ip6af->ip6af_frglen > af6->ip6af_off) {
377 i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
378 if (i < af6->ip6af_frglen) {
379 af6->ip6af_frglen -= i;
381 m_adj(IP6_REASS_MBUF(af6), i);
384 af6 = af6->ip6af_down;
385 m_freem(IP6_REASS_MBUF(af6->ip6af_up));
386 frag6_deq(af6->ip6af_up);
390 * If the incoming framgent overlaps some existing fragments in
391 * the reassembly queue, drop it, since it is dangerous to override
392 * existing fragments from a security point of view.
394 if (af6->ip6af_up != (struct ip6asfrag *)q6) {
395 i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
398 #if 0 /* suppress the noisy log */
399 log(LOG_ERR, "%d bytes of a fragment from %s "
400 "overlaps the previous fragment\n",
401 i, ip6_sprintf(&q6->ip6q_src));
403 kfree(ip6af, M_FTABLE);
407 if (af6 != (struct ip6asfrag *)q6) {
408 i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
410 #if 0 /* suppress the noisy log */
411 log(LOG_ERR, "%d bytes of a fragment from %s "
412 "overlaps the succeeding fragment",
413 i, ip6_sprintf(&q6->ip6q_src));
415 kfree(ip6af, M_FTABLE);
424 * Stick new segment in its place;
425 * check for complete reassembly.
426 * Move to front of packet queue, as we are
427 * the most recently active fragmented packet.
429 frag6_enq(ip6af, af6->ip6af_up);
431 if (q6 != ip6q.ip6q_next) {
433 frag6_insque(q6, &ip6q);
437 for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
438 af6 = af6->ip6af_down) {
439 if (af6->ip6af_off != next) {
440 frag6_doing_reass = 0;
443 next += af6->ip6af_frglen;
445 if (af6->ip6af_up->ip6af_mff) {
446 frag6_doing_reass = 0;
451 * Reassembly is complete; concatenate fragments.
453 ip6af = q6->ip6q_down;
454 t = m = IP6_REASS_MBUF(ip6af);
455 af6 = ip6af->ip6af_down;
457 while (af6 != (struct ip6asfrag *)q6) {
458 af6dwn = af6->ip6af_down;
462 t->m_next = IP6_REASS_MBUF(af6);
463 m_adj(t->m_next, af6->ip6af_offset);
464 kfree(af6, M_FTABLE);
468 /* adjust offset to point where the original next header starts */
469 offset = ip6af->ip6af_offset - sizeof(struct ip6_frag);
470 kfree(ip6af, M_FTABLE);
471 ip6 = mtod(m, struct ip6_hdr *);
472 ip6->ip6_plen = htons((u_short)next + offset - sizeof(struct ip6_hdr));
473 ip6->ip6_src = q6->ip6q_src;
474 ip6->ip6_dst = q6->ip6q_dst;
477 *q6->ip6q_nxtp = (u_char)(nxt & 0xff);
481 * Delete frag6 header with as a few cost as possible.
483 if (offset < m->m_len) {
484 ovbcopy((caddr_t)ip6, (caddr_t)ip6 + sizeof(struct ip6_frag),
486 m->m_data += sizeof(struct ip6_frag);
487 m->m_len -= sizeof(struct ip6_frag);
489 /* this comes with no copy if the boundary is on cluster */
490 if ((t = m_split(m, offset, MB_DONTWAIT)) == NULL) {
493 frag6_nfragpackets--;
496 m_adj(t, sizeof(struct ip6_frag));
501 * Store NXT to the original.
504 char *prvnxtp = ip6_get_prevhdr(m, offset); /* XXX */
510 frag6_nfragpackets--;
512 if (m->m_flags & M_PKTHDR) { /* Isn't it always true? */
514 for (t = m; t; t = t->m_next)
516 m->m_pkthdr.len = plen;
519 ip6stat.ip6s_reassembled++;
520 in6_ifstat_inc(dstifp, ifs6_reass_ok);
523 * Reassembly complete, return the next protocol.
524 * Be sure to clear M_HASH to force the packet
525 * to be re-characterized.
527 m->m_flags &= ~M_HASH;
532 frag6_doing_reass = 0;
536 in6_ifstat_inc(dstifp, ifs6_reass_fail);
537 ip6stat.ip6s_fragdropped++;
539 frag6_doing_reass = 0;
544 * Free a fragment reassembly header and all
545 * associated datagrams.
548 frag6_freef(struct ip6q *q6)
550 struct ip6asfrag *af6, *down6;
552 for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
554 struct mbuf *m = IP6_REASS_MBUF(af6);
556 down6 = af6->ip6af_down;
560 * Return ICMP time exceeded error for the 1st fragment.
561 * Just free other fragments.
563 if (af6->ip6af_off == 0) {
567 ip6 = mtod(m, struct ip6_hdr *);
569 /* restoure source and destination addresses */
570 ip6->ip6_src = q6->ip6q_src;
571 ip6->ip6_dst = q6->ip6q_dst;
573 icmp6_error(m, ICMP6_TIME_EXCEEDED,
574 ICMP6_TIME_EXCEED_REASSEMBLY, 0);
577 kfree(af6, M_FTABLE);
581 frag6_nfragpackets--;
585 * Put an ip fragment on a reassembly chain.
586 * Like insque, but pointers in middle of structure.
589 frag6_enq(struct ip6asfrag *af6, struct ip6asfrag *up6)
592 af6->ip6af_down = up6->ip6af_down;
593 up6->ip6af_down->ip6af_up = af6;
594 up6->ip6af_down = af6;
598 * To frag6_enq as remque is to insque.
601 frag6_deq(struct ip6asfrag *af6)
603 af6->ip6af_up->ip6af_down = af6->ip6af_down;
604 af6->ip6af_down->ip6af_up = af6->ip6af_up;
608 frag6_insque(struct ip6q *new, struct ip6q *old)
610 new->ip6q_prev = old;
611 new->ip6q_next = old->ip6q_next;
612 old->ip6q_next->ip6q_prev= new;
613 old->ip6q_next = new;
617 frag6_remque(struct ip6q *p6)
619 p6->ip6q_prev->ip6q_next = p6->ip6q_next;
620 p6->ip6q_next->ip6q_prev = p6->ip6q_prev;
624 * IPv6 reassembling timer processing;
625 * if a timer expires on a reassembly
634 frag6_doing_reass = 1;
637 while (q6 != &ip6q) {
640 if (q6->ip6q_prev->ip6q_ttl == 0) {
641 ip6stat.ip6s_fragtimeout++;
642 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
643 frag6_freef(q6->ip6q_prev);
647 * If we are over the maximum number of fragments
648 * (due to the limit being lowered), drain off
649 * enough to get down to the new limit.
651 while (frag6_nfragpackets > (u_int)ip6_maxfragpackets &&
653 ip6stat.ip6s_fragoverflow++;
654 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
655 frag6_freef(ip6q.ip6q_prev);
657 frag6_doing_reass = 0;
661 * Routing changes might produce a better route than we last used;
662 * make sure we notice eventually, even if forwarding only for one
663 * destination and the cache is never replaced.
665 if (ip6_forward_rt.ro_rt) {
666 RTFREE(ip6_forward_rt.ro_rt);
667 ip6_forward_rt.ro_rt = NULL;
669 if (ipsrcchk_rt.ro_rt) {
670 RTFREE(ipsrcchk_rt.ro_rt);
671 ipsrcchk_rt.ro_rt = NULL;
679 * Drain off all datagram fragments.
684 if (frag6_doing_reass)
686 while (ip6q.ip6q_next != &ip6q) {
687 ip6stat.ip6s_fragdropped++;
688 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
689 frag6_freef(ip6q.ip6q_next);