1 /* $FreeBSD: src/sys/netinet6/frag6.c,v 1.2.2.5 2001/07/03 11:01:50 ume Exp $ */
2 /* $KAME: frag6.c,v 1.31 2001/05/17 13:45:34 jinmei 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 <kern/queue.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
__P((struct ip6asfrag
*, struct ip6asfrag
*));
65 static void frag6_deq
__P((struct ip6asfrag
*));
66 static void frag6_insque
__P((struct ip6q
*, struct ip6q
*));
67 static void frag6_remque
__P((struct ip6q
*));
68 static void frag6_freef
__P((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");
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
= random() ^ 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(mp
, offp
)
135 struct mbuf
*m
= *mp
, *t
;
137 struct ip6_frag
*ip6f
;
139 struct ip6asfrag
*af6
, *ip6af
, *af6dwn
;
140 int offset
= *offp
, nxt
, i
, next
;
142 int fragoff
, frgpartlen
; /* must be larger than u_int16_t */
143 struct ifnet
*dstifp
;
144 #ifdef IN6_IFSTAT_STRICT
145 static struct route_in6 ro
;
146 struct sockaddr_in6
*dst
;
149 ip6
= mtod(m
, struct ip6_hdr
*);
150 #ifndef PULLDOWN_TEST
151 IP6_EXTHDR_CHECK(m
, offset
, sizeof(struct ip6_frag
), IPPROTO_DONE
);
152 ip6f
= (struct ip6_frag
*)((caddr_t
)ip6
+ offset
);
154 IP6_EXTHDR_GET(ip6f
, struct ip6_frag
*, m
, offset
, sizeof(*ip6f
));
160 #ifdef IN6_IFSTAT_STRICT
161 /* find the destination interface of the packet. */
162 dst
= (struct sockaddr_in6
*)&ro
.ro_dst
;
164 && ((ro
.ro_rt
->rt_flags
& RTF_UP
) == 0
165 || !IN6_ARE_ADDR_EQUAL(&dst
->sin6_addr
, &ip6
->ip6_dst
))) {
167 ro
.ro_rt
= (struct rtentry
*)0;
169 if (ro
.ro_rt
== NULL
) {
170 bzero(dst
, sizeof(*dst
));
171 dst
->sin6_family
= AF_INET6
;
172 dst
->sin6_len
= sizeof(struct sockaddr_in6
);
173 dst
->sin6_addr
= ip6
->ip6_dst
;
175 rtalloc((struct route
*)&ro
);
176 if (ro
.ro_rt
!= NULL
&& ro
.ro_rt
->rt_ifa
!= NULL
)
177 dstifp
= ((struct in6_ifaddr
*)ro
.ro_rt
->rt_ifa
)->ia_ifp
;
179 /* we are violating the spec, this is not the destination interface */
180 if ((m
->m_flags
& M_PKTHDR
) != 0)
181 dstifp
= m
->m_pkthdr
.rcvif
;
184 /* jumbo payload can't contain a fragment header */
185 if (ip6
->ip6_plen
== 0) {
186 icmp6_error(m
, ICMP6_PARAM_PROB
, ICMP6_PARAMPROB_HEADER
, offset
);
187 in6_ifstat_inc(dstifp
, ifs6_reass_fail
);
192 * check whether fragment packet's fragment length is
193 * multiple of 8 octets.
194 * sizeof(struct ip6_frag) == 8
195 * sizeof(struct ip6_hdr) = 40
197 if ((ip6f
->ip6f_offlg
& IP6F_MORE_FRAG
) &&
198 (((ntohs(ip6
->ip6_plen
) - offset
) & 0x7) != 0)) {
199 icmp6_error(m
, ICMP6_PARAM_PROB
,
200 ICMP6_PARAMPROB_HEADER
,
201 offsetof(struct ip6_hdr
, ip6_plen
));
202 in6_ifstat_inc(dstifp
, ifs6_reass_fail
);
206 ip6stat
.ip6s_fragments
++;
207 in6_ifstat_inc(dstifp
, ifs6_reass_reqd
);
209 /* offset now points to data portion */
210 offset
+= sizeof(struct ip6_frag
);
212 frag6_doing_reass
= 1;
214 for (q6
= ip6q
.ip6q_next
; q6
!= &ip6q
; q6
= q6
->ip6q_next
)
215 if (ip6f
->ip6f_ident
== q6
->ip6q_ident
&&
216 IN6_ARE_ADDR_EQUAL(&ip6
->ip6_src
, &q6
->ip6q_src
) &&
217 IN6_ARE_ADDR_EQUAL(&ip6
->ip6_dst
, &q6
->ip6q_dst
))
222 * the first fragment to arrive, create a reassembly queue.
227 * Enforce upper bound on number of fragmented packets
228 * for which we attempt reassembly;
229 * If maxfrag is 0, never accept fragments.
230 * If maxfrag is -1, accept all fragments without limitation.
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. */
259 * If it's the 1st fragment, record the length of the
260 * unfragmentable part and the next header of the fragment header.
262 fragoff
= ntohs(ip6f
->ip6f_offlg
& IP6F_OFF_MASK
);
264 q6
->ip6q_unfrglen
= offset
- sizeof(struct ip6_hdr
)
265 - sizeof(struct ip6_frag
);
266 q6
->ip6q_nxt
= ip6f
->ip6f_nxt
;
270 * Check that the reassembled packet would not exceed 65535 bytes
272 * If it would exceed, discard the fragment and return an ICMP error.
274 frgpartlen
= sizeof(struct ip6_hdr
) + ntohs(ip6
->ip6_plen
) - offset
;
275 if (q6
->ip6q_unfrglen
>= 0) {
276 /* The 1st fragment has already arrived. */
277 if (q6
->ip6q_unfrglen
+ fragoff
+ frgpartlen
> IPV6_MAXPACKET
) {
278 icmp6_error(m
, ICMP6_PARAM_PROB
, ICMP6_PARAMPROB_HEADER
,
279 offset
- sizeof(struct ip6_frag
) +
280 offsetof(struct ip6_frag
, ip6f_offlg
));
281 frag6_doing_reass
= 0;
282 return(IPPROTO_DONE
);
285 else if (fragoff
+ frgpartlen
> IPV6_MAXPACKET
) {
286 icmp6_error(m
, ICMP6_PARAM_PROB
, ICMP6_PARAMPROB_HEADER
,
287 offset
- sizeof(struct ip6_frag
) +
288 offsetof(struct ip6_frag
, ip6f_offlg
));
289 frag6_doing_reass
= 0;
290 return(IPPROTO_DONE
);
293 * If it's the first fragment, do the above check for each
294 * fragment already stored in the reassembly queue.
297 for (af6
= q6
->ip6q_down
; af6
!= (struct ip6asfrag
*)q6
;
299 af6dwn
= af6
->ip6af_down
;
301 if (q6
->ip6q_unfrglen
+ af6
->ip6af_off
+ af6
->ip6af_frglen
>
303 struct mbuf
*merr
= IP6_REASS_MBUF(af6
);
304 struct ip6_hdr
*ip6err
;
305 int erroff
= af6
->ip6af_offset
;
307 /* dequeue the fragment. */
311 /* adjust pointer. */
312 ip6err
= mtod(merr
, struct ip6_hdr
*);
315 * Restore source and destination addresses
316 * in the erroneous IPv6 header.
318 ip6err
->ip6_src
= q6
->ip6q_src
;
319 ip6err
->ip6_dst
= q6
->ip6q_dst
;
321 icmp6_error(merr
, ICMP6_PARAM_PROB
,
322 ICMP6_PARAMPROB_HEADER
,
323 erroff
- sizeof(struct ip6_frag
) +
324 offsetof(struct ip6_frag
, ip6f_offlg
));
329 ip6af
= (struct ip6asfrag
*)_MALLOC(sizeof(struct ip6asfrag
), M_FTABLE
,
333 bzero(ip6af
, sizeof(*ip6af
));
334 ip6af
->ip6af_head
= ip6
->ip6_flow
;
335 ip6af
->ip6af_len
= ip6
->ip6_plen
;
336 ip6af
->ip6af_nxt
= ip6
->ip6_nxt
;
337 ip6af
->ip6af_hlim
= ip6
->ip6_hlim
;
338 ip6af
->ip6af_mff
= ip6f
->ip6f_offlg
& IP6F_MORE_FRAG
;
339 ip6af
->ip6af_off
= fragoff
;
340 ip6af
->ip6af_frglen
= frgpartlen
;
341 ip6af
->ip6af_offset
= offset
;
342 IP6_REASS_MBUF(ip6af
) = m
;
345 af6
= (struct ip6asfrag
*)q6
;
350 * Find a segment which begins after this one does.
352 for (af6
= q6
->ip6q_down
; af6
!= (struct ip6asfrag
*)q6
;
353 af6
= af6
->ip6af_down
)
354 if (af6
->ip6af_off
> ip6af
->ip6af_off
)
359 * If there is a preceding segment, it may provide some of
360 * our data already. If so, drop the data from the incoming
361 * segment. If it provides all of our data, drop us.
363 if (af6
->ip6af_up
!= (struct ip6asfrag
*)q6
) {
364 i
= af6
->ip6af_up
->ip6af_off
+ af6
->ip6af_up
->ip6af_frglen
367 if (i
>= ip6af
->ip6af_frglen
)
369 m_adj(IP6_REASS_MBUF(ip6af
), i
);
370 ip6af
->ip6af_off
+= i
;
371 ip6af
->ip6af_frglen
-= i
;
376 * While we overlap succeeding segments trim them or,
377 * if they are completely covered, dequeue them.
379 while (af6
!= (struct ip6asfrag
*)q6
&&
380 ip6af
->ip6af_off
+ ip6af
->ip6af_frglen
> af6
->ip6af_off
) {
381 i
= (ip6af
->ip6af_off
+ ip6af
->ip6af_frglen
) - af6
->ip6af_off
;
382 if (i
< af6
->ip6af_frglen
) {
383 af6
->ip6af_frglen
-= i
;
385 m_adj(IP6_REASS_MBUF(af6
), i
);
388 af6
= af6
->ip6af_down
;
389 m_freem(IP6_REASS_MBUF(af6
->ip6af_up
));
390 frag6_deq(af6
->ip6af_up
);
394 * If the incoming framgent overlaps some existing fragments in
395 * the reassembly queue, drop it, since it is dangerous to override
396 * existing fragments from a security point of view.
398 if (af6
->ip6af_up
!= (struct ip6asfrag
*)q6
) {
399 i
= af6
->ip6af_up
->ip6af_off
+ af6
->ip6af_up
->ip6af_frglen
402 #if 0 /* suppress the noisy log */
403 log(LOG_ERR
, "%d bytes of a fragment from %s "
404 "overlaps the previous fragment\n",
405 i
, ip6_sprintf(&q6
->ip6q_src
));
407 FREE(ip6af
, M_FTABLE
);
411 if (af6
!= (struct ip6asfrag
*)q6
) {
412 i
= (ip6af
->ip6af_off
+ ip6af
->ip6af_frglen
) - af6
->ip6af_off
;
414 #if 0 /* suppress the noisy log */
415 log(LOG_ERR
, "%d bytes of a fragment from %s "
416 "overlaps the succeeding fragment",
417 i
, ip6_sprintf(&q6
->ip6q_src
));
419 FREE(ip6af
, M_FTABLE
);
428 * Stick new segment in its place;
429 * check for complete reassembly.
430 * Move to front of packet queue, as we are
431 * the most recently active fragmented packet.
433 frag6_enq(ip6af
, af6
->ip6af_up
);
435 if (q6
!= ip6q
.ip6q_next
) {
437 frag6_insque(q6
, &ip6q
);
441 for (af6
= q6
->ip6q_down
; af6
!= (struct ip6asfrag
*)q6
;
442 af6
= af6
->ip6af_down
) {
443 if (af6
->ip6af_off
!= next
) {
444 frag6_doing_reass
= 0;
447 next
+= af6
->ip6af_frglen
;
449 if (af6
->ip6af_up
->ip6af_mff
) {
450 frag6_doing_reass
= 0;
455 * Reassembly is complete; concatenate fragments.
457 ip6af
= q6
->ip6q_down
;
458 t
= m
= IP6_REASS_MBUF(ip6af
);
459 af6
= ip6af
->ip6af_down
;
461 while (af6
!= (struct ip6asfrag
*)q6
) {
462 af6dwn
= af6
->ip6af_down
;
466 t
->m_next
= IP6_REASS_MBUF(af6
);
467 m_adj(t
->m_next
, af6
->ip6af_offset
);
472 /* adjust offset to point where the original next header starts */
473 offset
= ip6af
->ip6af_offset
- sizeof(struct ip6_frag
);
474 FREE(ip6af
, M_FTABLE
);
475 ip6
= mtod(m
, struct ip6_hdr
*);
476 ip6
->ip6_plen
= htons((u_short
)next
+ offset
- sizeof(struct ip6_hdr
));
477 ip6
->ip6_src
= q6
->ip6q_src
;
478 ip6
->ip6_dst
= q6
->ip6q_dst
;
481 *q6
->ip6q_nxtp
= (u_char
)(nxt
& 0xff);
485 * Delete frag6 header with as a few cost as possible.
487 if (offset
< m
->m_len
) {
488 ovbcopy((caddr_t
)ip6
, (caddr_t
)ip6
+ sizeof(struct ip6_frag
),
490 m
->m_data
+= sizeof(struct ip6_frag
);
491 m
->m_len
-= sizeof(struct ip6_frag
);
493 /* this comes with no copy if the boundary is on cluster */
494 if ((t
= m_split(m
, offset
, M_DONTWAIT
)) == NULL
) {
497 frag6_nfragpackets
--;
500 m_adj(t
, sizeof(struct ip6_frag
));
505 * Store NXT to the original.
508 char *prvnxtp
= ip6_get_prevhdr(m
, offset
); /* XXX */
514 frag6_nfragpackets
--;
516 if (m
->m_flags
& M_PKTHDR
) { /* Isn't it always true? */
518 for (t
= m
; t
; t
= t
->m_next
)
520 m
->m_pkthdr
.len
= plen
;
523 ip6stat
.ip6s_reassembled
++;
524 in6_ifstat_inc(dstifp
, ifs6_reass_ok
);
527 * Tell launch routine the next header
533 frag6_doing_reass
= 0;
537 in6_ifstat_inc(dstifp
, ifs6_reass_fail
);
538 ip6stat
.ip6s_fragdropped
++;
540 frag6_doing_reass
= 0;
545 * Free a fragment reassembly header and all
546 * associated datagrams.
552 struct ip6asfrag
*af6
, *down6
;
554 for (af6
= q6
->ip6q_down
; af6
!= (struct ip6asfrag
*)q6
;
556 struct mbuf
*m
= IP6_REASS_MBUF(af6
);
558 down6
= af6
->ip6af_down
;
562 * Return ICMP time exceeded error for the 1st fragment.
563 * Just free other fragments.
565 if (af6
->ip6af_off
== 0) {
569 ip6
= mtod(m
, struct ip6_hdr
*);
571 /* restoure source and destination addresses */
572 ip6
->ip6_src
= q6
->ip6q_src
;
573 ip6
->ip6_dst
= q6
->ip6q_dst
;
575 icmp6_error(m
, ICMP6_TIME_EXCEEDED
,
576 ICMP6_TIME_EXCEED_REASSEMBLY
, 0);
584 frag6_nfragpackets
--;
588 * Put an ip fragment on a reassembly chain.
589 * Like insque, but pointers in middle of structure.
593 struct ip6asfrag
*af6
, *up6
;
596 af6
->ip6af_down
= up6
->ip6af_down
;
597 up6
->ip6af_down
->ip6af_up
= af6
;
598 up6
->ip6af_down
= af6
;
602 * To frag6_enq as remque is to insque.
606 struct ip6asfrag
*af6
;
608 af6
->ip6af_up
->ip6af_down
= af6
->ip6af_down
;
609 af6
->ip6af_down
->ip6af_up
= af6
->ip6af_up
;
613 frag6_insque(new, old
)
614 struct ip6q
*new, *old
;
616 new->ip6q_prev
= old
;
617 new->ip6q_next
= old
->ip6q_next
;
618 old
->ip6q_next
->ip6q_prev
= new;
619 old
->ip6q_next
= new;
626 p6
->ip6q_prev
->ip6q_next
= p6
->ip6q_next
;
627 p6
->ip6q_next
->ip6q_prev
= p6
->ip6q_prev
;
631 * IPv6 reassembling timer processing;
632 * if a timer expires on a reassembly
641 frag6_doing_reass
= 1;
644 while (q6
!= &ip6q
) {
647 if (q6
->ip6q_prev
->ip6q_ttl
== 0) {
648 ip6stat
.ip6s_fragtimeout
++;
649 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
650 frag6_freef(q6
->ip6q_prev
);
654 * If we are over the maximum number of fragments
655 * (due to the limit being lowered), drain off
656 * enough to get down to the new limit.
658 while (frag6_nfragpackets
> (u_int
)ip6_maxfragpackets
&&
660 ip6stat
.ip6s_fragoverflow
++;
661 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
662 frag6_freef(ip6q
.ip6q_prev
);
664 frag6_doing_reass
= 0;
668 * Routing changes might produce a better route than we last used;
669 * make sure we notice eventually, even if forwarding only for one
670 * destination and the cache is never replaced.
672 if (ip6_forward_rt
.ro_rt
) {
673 rtfree(ip6_forward_rt
.ro_rt
);
674 ip6_forward_rt
.ro_rt
= 0;
676 if (ipsrcchk_rt
.ro_rt
) {
677 rtfree(ipsrcchk_rt
.ro_rt
);
678 ipsrcchk_rt
.ro_rt
= 0;
686 * Drain off all datagram fragments.
691 if (frag6_doing_reass
)
693 while (ip6q
.ip6q_next
!= &ip6q
) {
694 ip6stat
.ip6s_fragdropped
++;
695 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
696 frag6_freef(ip6q
.ip6q_next
);