1 /* $FreeBSD: src/sys/netinet6/frag6.c,v 1.2.2.6 2002/04/28 05:40:26 suz Exp $ */
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>
44 #include <sys/thread2.h>
47 #include <net/route.h>
49 #include <netinet/in.h>
50 #include <netinet/in_var.h>
51 #include <netinet/ip6.h>
52 #include <netinet6/ip6_var.h>
53 #include <netinet/icmp6.h>
55 #include <net/net_osdep.h>
58 * Define it to get a correct behavior on per-interface statistics.
59 * You will need to perform an extra routing table lookup, per fragment,
60 * to do it. This may, or may not be, a performance hit.
62 #define IN6_IFSTAT_STRICT
64 static void frag6_enq (struct ip6asfrag *, struct ip6asfrag *);
65 static void frag6_deq (struct ip6asfrag *);
66 static void frag6_insque (struct ip6q *, struct ip6q *);
67 static void frag6_remque (struct ip6q *);
68 static void frag6_freef (struct ip6q *);
70 /* XXX we eventually need splreass6, or some real semaphore */
71 int frag6_doing_reass;
72 u_int frag6_nfragpackets;
73 struct ip6q ip6q; /* ip6 reassemble queue */
76 MALLOC_DEFINE(M_FTABLE, "fragment", "fragment reassembly header");
79 * Initialise reassembly queue and fragment identifier.
86 ip6_maxfragpackets = nmbclusters / 4;
89 * in many cases, random() here does NOT return random number
90 * as initialization during bootstrap time occur in fixed order.
93 ip6_id = krandom() ^ tv.tv_usec;
94 ip6q.ip6q_next = ip6q.ip6q_prev = &ip6q;
98 * In RFC2460, fragment and reassembly rule do not agree with each other,
99 * in terms of next header field handling in fragment header.
100 * While the sender will use the same value for all of the fragmented packets,
101 * receiver is suggested not to check the consistency.
103 * fragment rule (p20):
104 * (2) A Fragment header containing:
105 * The Next Header value that identifies the first header of
106 * the Fragmentable Part of the original packet.
107 * -> next header field is same for all fragments
109 * reassembly rule (p21):
110 * The Next Header field of the last header of the Unfragmentable
111 * Part is obtained from the Next Header field of the first
112 * fragment's Fragment header.
113 * -> should grab it from the first fragment only
115 * The following note also contradicts with fragment rule - noone is going to
116 * send different fragment with different next header field.
118 * additional note (p22):
119 * The Next Header values in the Fragment headers of different
120 * fragments of the same original packet may differ. Only the value
121 * from the Offset zero fragment packet is used for reassembly.
122 * -> should grab it from the first fragment only
124 * There is no explicit reason given in the RFC. Historical reason maybe?
130 frag6_input(struct mbuf **mp, int *offp, int proto)
132 struct mbuf *m = *mp, *t;
134 struct ip6_frag *ip6f;
136 struct ip6asfrag *af6, *ip6af, *af6dwn;
137 int offset = *offp, nxt, i, next;
139 int fragoff, frgpartlen; /* must be larger than u_int16_t */
140 struct ifnet *dstifp;
141 #ifdef IN6_IFSTAT_STRICT
142 static struct route_in6 ro;
143 struct sockaddr_in6 *dst;
146 ip6 = mtod(m, struct ip6_hdr *);
147 #ifndef PULLDOWN_TEST
148 IP6_EXTHDR_CHECK(m, offset, sizeof(struct ip6_frag), IPPROTO_DONE);
149 ip6f = (struct ip6_frag *)((caddr_t)ip6 + offset);
151 IP6_EXTHDR_GET(ip6f, struct ip6_frag *, m, offset, sizeof(*ip6f));
157 #ifdef IN6_IFSTAT_STRICT
158 /* find the destination interface of the packet. */
159 dst = (struct sockaddr_in6 *)&ro.ro_dst;
161 (!(ro.ro_rt->rt_flags & RTF_UP) ||
162 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
166 if (ro.ro_rt == NULL) {
167 bzero(dst, sizeof(*dst));
168 dst->sin6_family = AF_INET6;
169 dst->sin6_len = sizeof(struct sockaddr_in6);
170 dst->sin6_addr = ip6->ip6_dst;
172 rtalloc((struct route *)&ro);
173 if (ro.ro_rt != NULL && ro.ro_rt->rt_ifa != NULL)
174 dstifp = ((struct in6_ifaddr *)ro.ro_rt->rt_ifa)->ia_ifp;
176 /* we are violating the spec, this is not the destination interface */
177 if (m->m_flags & M_PKTHDR)
178 dstifp = m->m_pkthdr.rcvif;
181 /* jumbo payload can't contain a fragment header */
182 if (ip6->ip6_plen == 0) {
183 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, offset);
184 in6_ifstat_inc(dstifp, ifs6_reass_fail);
189 * check whether fragment packet's fragment length is
190 * multiple of 8 octets.
191 * sizeof(struct ip6_frag) == 8
192 * sizeof(struct ip6_hdr) = 40
194 if ((ip6f->ip6f_offlg & IP6F_MORE_FRAG) &&
195 (((ntohs(ip6->ip6_plen) - offset) & 0x7) != 0)) {
196 icmp6_error(m, ICMP6_PARAM_PROB,
197 ICMP6_PARAMPROB_HEADER,
198 offsetof(struct ip6_hdr, ip6_plen));
199 in6_ifstat_inc(dstifp, ifs6_reass_fail);
203 ip6stat.ip6s_fragments++;
204 in6_ifstat_inc(dstifp, ifs6_reass_reqd);
206 /* offset now points to data portion */
207 offset += sizeof(struct ip6_frag);
209 frag6_doing_reass = 1;
211 for (q6 = ip6q.ip6q_next; q6 != &ip6q; q6 = q6->ip6q_next)
212 if (ip6f->ip6f_ident == q6->ip6q_ident &&
213 IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &q6->ip6q_src) &&
214 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &q6->ip6q_dst))
219 * the first fragment to arrive, create a reassembly queue.
224 * Enforce upper bound on number of fragmented packets
225 * for which we attempt reassembly;
226 * If maxfrag is 0, never accept fragments.
227 * If maxfrag is -1, accept all fragments without limitation.
229 if (ip6_maxfragpackets < 0)
231 else if (frag6_nfragpackets >= (u_int)ip6_maxfragpackets)
233 frag6_nfragpackets++;
234 q6 = (struct ip6q *)kmalloc(sizeof(struct ip6q), M_FTABLE,
239 frag6_insque(q6, &ip6q);
241 /* ip6q_nxt will be filled afterwards, from 1st fragment */
242 q6->ip6q_down = q6->ip6q_up = (struct ip6asfrag *)q6;
244 q6->ip6q_nxtp = (u_char *)nxtp;
246 q6->ip6q_ident = ip6f->ip6f_ident;
247 q6->ip6q_arrive = 0; /* Is it used anywhere? */
248 q6->ip6q_ttl = IPV6_FRAGTTL;
249 q6->ip6q_src = ip6->ip6_src;
250 q6->ip6q_dst = ip6->ip6_dst;
251 q6->ip6q_unfrglen = -1; /* The 1st fragment has not arrived. */
255 * If it's the 1st fragment, record the length of the
256 * unfragmentable part and the next header of the fragment header.
258 fragoff = ntohs(ip6f->ip6f_offlg & IP6F_OFF_MASK);
260 q6->ip6q_unfrglen = offset - sizeof(struct ip6_hdr)
261 - sizeof(struct ip6_frag);
262 q6->ip6q_nxt = ip6f->ip6f_nxt;
266 * Check that the reassembled packet would not exceed 65535 bytes
268 * If it would exceed, discard the fragment and return an ICMP error.
270 frgpartlen = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen) - offset;
271 if (q6->ip6q_unfrglen >= 0) {
272 /* The 1st fragment has already arrived. */
273 if (q6->ip6q_unfrglen + fragoff + frgpartlen > IPV6_MAXPACKET) {
274 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
275 offset - sizeof(struct ip6_frag) +
276 offsetof(struct ip6_frag, ip6f_offlg));
277 frag6_doing_reass = 0;
278 return (IPPROTO_DONE);
281 else if (fragoff + frgpartlen > IPV6_MAXPACKET) {
282 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
283 offset - sizeof(struct ip6_frag) +
284 offsetof(struct ip6_frag, ip6f_offlg));
285 frag6_doing_reass = 0;
286 return (IPPROTO_DONE);
289 * If it's the first fragment, do the above check for each
290 * fragment already stored in the reassembly queue.
293 for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
295 af6dwn = af6->ip6af_down;
297 if (q6->ip6q_unfrglen + af6->ip6af_off + af6->ip6af_frglen >
299 struct mbuf *merr = IP6_REASS_MBUF(af6);
300 struct ip6_hdr *ip6err;
301 int erroff = af6->ip6af_offset;
303 /* dequeue the fragment. */
305 kfree(af6, M_FTABLE);
307 /* adjust pointer. */
308 ip6err = mtod(merr, struct ip6_hdr *);
311 * Restore source and destination addresses
312 * in the erroneous IPv6 header.
314 ip6err->ip6_src = q6->ip6q_src;
315 ip6err->ip6_dst = q6->ip6q_dst;
317 icmp6_error(merr, ICMP6_PARAM_PROB,
318 ICMP6_PARAMPROB_HEADER,
319 erroff - sizeof(struct ip6_frag) +
320 offsetof(struct ip6_frag, ip6f_offlg));
325 ip6af = (struct ip6asfrag *)kmalloc(sizeof(struct ip6asfrag), M_FTABLE,
329 ip6af->ip6af_head = ip6->ip6_flow;
330 ip6af->ip6af_len = ip6->ip6_plen;
331 ip6af->ip6af_nxt = ip6->ip6_nxt;
332 ip6af->ip6af_hlim = ip6->ip6_hlim;
333 ip6af->ip6af_mff = ip6f->ip6f_offlg & IP6F_MORE_FRAG;
334 ip6af->ip6af_off = fragoff;
335 ip6af->ip6af_frglen = frgpartlen;
336 ip6af->ip6af_offset = offset;
337 IP6_REASS_MBUF(ip6af) = m;
340 af6 = (struct ip6asfrag *)q6;
345 * Find a segment which begins after this one does.
347 for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
348 af6 = af6->ip6af_down)
349 if (af6->ip6af_off > ip6af->ip6af_off)
353 * RFC 5722: Drop overlapping fragments
355 if (af6->ip6af_up != (struct ip6asfrag *)q6) {
356 i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
359 kfree(ip6af, M_FTABLE);
363 if (af6 != (struct ip6asfrag *)q6) {
364 i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
366 kfree(ip6af, M_FTABLE);
374 * Stick new segment in its place;
375 * check for complete reassembly.
376 * Move to front of packet queue, as we are
377 * the most recently active fragmented packet.
379 frag6_enq(ip6af, af6->ip6af_up);
381 if (q6 != ip6q.ip6q_next) {
383 frag6_insque(q6, &ip6q);
387 for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
388 af6 = af6->ip6af_down) {
389 if (af6->ip6af_off != next) {
390 frag6_doing_reass = 0;
393 next += af6->ip6af_frglen;
395 if (af6->ip6af_up->ip6af_mff) {
396 frag6_doing_reass = 0;
401 * Reassembly is complete; concatenate fragments.
403 ip6af = q6->ip6q_down;
404 t = m = IP6_REASS_MBUF(ip6af);
405 af6 = ip6af->ip6af_down;
407 while (af6 != (struct ip6asfrag *)q6) {
408 af6dwn = af6->ip6af_down;
412 t->m_next = IP6_REASS_MBUF(af6);
413 m_adj(t->m_next, af6->ip6af_offset);
414 kfree(af6, M_FTABLE);
418 /* adjust offset to point where the original next header starts */
419 offset = ip6af->ip6af_offset - sizeof(struct ip6_frag);
420 kfree(ip6af, M_FTABLE);
421 ip6 = mtod(m, struct ip6_hdr *);
422 ip6->ip6_plen = htons((u_short)next + offset - sizeof(struct ip6_hdr));
423 ip6->ip6_src = q6->ip6q_src;
424 ip6->ip6_dst = q6->ip6q_dst;
427 *q6->ip6q_nxtp = (u_char)(nxt & 0xff);
431 * Delete frag6 header with as a few cost as possible.
433 if (offset < m->m_len) {
434 ovbcopy((caddr_t)ip6, (caddr_t)ip6 + sizeof(struct ip6_frag),
436 m->m_data += sizeof(struct ip6_frag);
437 m->m_len -= sizeof(struct ip6_frag);
439 /* this comes with no copy if the boundary is on cluster */
440 if ((t = m_split(m, offset, M_NOWAIT)) == NULL) {
443 frag6_nfragpackets--;
446 m_adj(t, sizeof(struct ip6_frag));
451 * Store NXT to the original.
454 char *prvnxtp = ip6_get_prevhdr(m, offset); /* XXX */
460 frag6_nfragpackets--;
462 if (m->m_flags & M_PKTHDR) { /* Isn't it always true? */
464 for (t = m; t; t = t->m_next)
466 m->m_pkthdr.len = plen;
469 ip6stat.ip6s_reassembled++;
470 in6_ifstat_inc(dstifp, ifs6_reass_ok);
473 * Reassembly complete, return the next protocol.
474 * Be sure to clear M_HASH to force the packet
475 * to be re-characterized.
477 m->m_flags &= ~M_HASH;
482 frag6_doing_reass = 0;
486 in6_ifstat_inc(dstifp, ifs6_reass_fail);
487 ip6stat.ip6s_fragdropped++;
489 frag6_doing_reass = 0;
494 * Free a fragment reassembly header and all
495 * associated datagrams.
498 frag6_freef(struct ip6q *q6)
500 struct ip6asfrag *af6, *down6;
502 for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
504 struct mbuf *m = IP6_REASS_MBUF(af6);
506 down6 = af6->ip6af_down;
510 * Return ICMP time exceeded error for the 1st fragment.
511 * Just free other fragments.
513 if (af6->ip6af_off == 0) {
517 ip6 = mtod(m, struct ip6_hdr *);
519 /* restoure source and destination addresses */
520 ip6->ip6_src = q6->ip6q_src;
521 ip6->ip6_dst = q6->ip6q_dst;
523 icmp6_error(m, ICMP6_TIME_EXCEEDED,
524 ICMP6_TIME_EXCEED_REASSEMBLY, 0);
527 kfree(af6, M_FTABLE);
531 frag6_nfragpackets--;
535 * Put an ip fragment on a reassembly chain.
536 * Like insque, but pointers in middle of structure.
539 frag6_enq(struct ip6asfrag *af6, struct ip6asfrag *up6)
542 af6->ip6af_down = up6->ip6af_down;
543 up6->ip6af_down->ip6af_up = af6;
544 up6->ip6af_down = af6;
548 * To frag6_enq as remque is to insque.
551 frag6_deq(struct ip6asfrag *af6)
553 af6->ip6af_up->ip6af_down = af6->ip6af_down;
554 af6->ip6af_down->ip6af_up = af6->ip6af_up;
558 frag6_insque(struct ip6q *new, struct ip6q *old)
560 new->ip6q_prev = old;
561 new->ip6q_next = old->ip6q_next;
562 old->ip6q_next->ip6q_prev= new;
563 old->ip6q_next = new;
567 frag6_remque(struct ip6q *p6)
569 p6->ip6q_prev->ip6q_next = p6->ip6q_next;
570 p6->ip6q_next->ip6q_prev = p6->ip6q_prev;
574 * IPv6 reassembling timer processing;
575 * if a timer expires on a reassembly
584 frag6_doing_reass = 1;
587 while (q6 != &ip6q) {
590 if (q6->ip6q_prev->ip6q_ttl == 0) {
591 ip6stat.ip6s_fragtimeout++;
592 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
593 frag6_freef(q6->ip6q_prev);
597 * If we are over the maximum number of fragments
598 * (due to the limit being lowered), drain off
599 * enough to get down to the new limit.
601 while (frag6_nfragpackets > (u_int)ip6_maxfragpackets &&
603 ip6stat.ip6s_fragoverflow++;
604 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
605 frag6_freef(ip6q.ip6q_prev);
607 frag6_doing_reass = 0;
611 * Routing changes might produce a better route than we last used;
612 * make sure we notice eventually, even if forwarding only for one
613 * destination and the cache is never replaced.
615 if (ip6_forward_rt.ro_rt) {
616 RTFREE(ip6_forward_rt.ro_rt);
617 ip6_forward_rt.ro_rt = NULL;
619 if (ipsrcchk_rt.ro_rt) {
620 RTFREE(ipsrcchk_rt.ro_rt);
621 ipsrcchk_rt.ro_rt = NULL;
629 * Drain off all datagram fragments.
634 if (frag6_doing_reass)
636 while (ip6q.ip6q_next != &ip6q) {
637 ip6stat.ip6s_fragdropped++;
638 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
639 frag6_freef(ip6q.ip6q_next);