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 1.11 2006/10/24 06:18:42 hsu 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
))) {
165 ro
.ro_rt
= (struct rtentry
*)NULL
;
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
,
239 bzero(q6
, sizeof(*q6
));
241 frag6_insque(q6
, &ip6q
);
243 /* ip6q_nxt will be filled afterwards, from 1st fragment */
244 q6
->ip6q_down
= q6
->ip6q_up
= (struct ip6asfrag
*)q6
;
246 q6
->ip6q_nxtp
= (u_char
*)nxtp
;
248 q6
->ip6q_ident
= ip6f
->ip6f_ident
;
249 q6
->ip6q_arrive
= 0; /* Is it used anywhere? */
250 q6
->ip6q_ttl
= IPV6_FRAGTTL
;
251 q6
->ip6q_src
= ip6
->ip6_src
;
252 q6
->ip6q_dst
= ip6
->ip6_dst
;
253 q6
->ip6q_unfrglen
= -1; /* The 1st fragment has not arrived. */
257 * If it's the 1st fragment, record the length of the
258 * unfragmentable part and the next header of the fragment header.
260 fragoff
= ntohs(ip6f
->ip6f_offlg
& IP6F_OFF_MASK
);
262 q6
->ip6q_unfrglen
= offset
- sizeof(struct ip6_hdr
)
263 - sizeof(struct ip6_frag
);
264 q6
->ip6q_nxt
= ip6f
->ip6f_nxt
;
268 * Check that the reassembled packet would not exceed 65535 bytes
270 * If it would exceed, discard the fragment and return an ICMP error.
272 frgpartlen
= sizeof(struct ip6_hdr
) + ntohs(ip6
->ip6_plen
) - offset
;
273 if (q6
->ip6q_unfrglen
>= 0) {
274 /* The 1st fragment has already arrived. */
275 if (q6
->ip6q_unfrglen
+ fragoff
+ frgpartlen
> IPV6_MAXPACKET
) {
276 icmp6_error(m
, ICMP6_PARAM_PROB
, ICMP6_PARAMPROB_HEADER
,
277 offset
- sizeof(struct ip6_frag
) +
278 offsetof(struct ip6_frag
, ip6f_offlg
));
279 frag6_doing_reass
= 0;
280 return (IPPROTO_DONE
);
283 else if (fragoff
+ frgpartlen
> IPV6_MAXPACKET
) {
284 icmp6_error(m
, ICMP6_PARAM_PROB
, ICMP6_PARAMPROB_HEADER
,
285 offset
- sizeof(struct ip6_frag
) +
286 offsetof(struct ip6_frag
, ip6f_offlg
));
287 frag6_doing_reass
= 0;
288 return (IPPROTO_DONE
);
291 * If it's the first fragment, do the above check for each
292 * fragment already stored in the reassembly queue.
295 for (af6
= q6
->ip6q_down
; af6
!= (struct ip6asfrag
*)q6
;
297 af6dwn
= af6
->ip6af_down
;
299 if (q6
->ip6q_unfrglen
+ af6
->ip6af_off
+ af6
->ip6af_frglen
>
301 struct mbuf
*merr
= IP6_REASS_MBUF(af6
);
302 struct ip6_hdr
*ip6err
;
303 int erroff
= af6
->ip6af_offset
;
305 /* dequeue the fragment. */
307 kfree(af6
, M_FTABLE
);
309 /* adjust pointer. */
310 ip6err
= mtod(merr
, struct ip6_hdr
*);
313 * Restore source and destination addresses
314 * in the erroneous IPv6 header.
316 ip6err
->ip6_src
= q6
->ip6q_src
;
317 ip6err
->ip6_dst
= q6
->ip6q_dst
;
319 icmp6_error(merr
, ICMP6_PARAM_PROB
,
320 ICMP6_PARAMPROB_HEADER
,
321 erroff
- sizeof(struct ip6_frag
) +
322 offsetof(struct ip6_frag
, ip6f_offlg
));
327 ip6af
= (struct ip6asfrag
*)kmalloc(sizeof(struct ip6asfrag
), M_FTABLE
,
331 bzero(ip6af
, sizeof(*ip6af
));
332 ip6af
->ip6af_head
= ip6
->ip6_flow
;
333 ip6af
->ip6af_len
= ip6
->ip6_plen
;
334 ip6af
->ip6af_nxt
= ip6
->ip6_nxt
;
335 ip6af
->ip6af_hlim
= ip6
->ip6_hlim
;
336 ip6af
->ip6af_mff
= ip6f
->ip6f_offlg
& IP6F_MORE_FRAG
;
337 ip6af
->ip6af_off
= fragoff
;
338 ip6af
->ip6af_frglen
= frgpartlen
;
339 ip6af
->ip6af_offset
= offset
;
340 IP6_REASS_MBUF(ip6af
) = m
;
343 af6
= (struct ip6asfrag
*)q6
;
348 * Find a segment which begins after this one does.
350 for (af6
= q6
->ip6q_down
; af6
!= (struct ip6asfrag
*)q6
;
351 af6
= af6
->ip6af_down
)
352 if (af6
->ip6af_off
> ip6af
->ip6af_off
)
357 * If there is a preceding segment, it may provide some of
358 * our data already. If so, drop the data from the incoming
359 * segment. If it provides all of our data, drop us.
361 if (af6
->ip6af_up
!= (struct ip6asfrag
*)q6
) {
362 i
= af6
->ip6af_up
->ip6af_off
+ af6
->ip6af_up
->ip6af_frglen
365 if (i
>= ip6af
->ip6af_frglen
)
367 m_adj(IP6_REASS_MBUF(ip6af
), i
);
368 ip6af
->ip6af_off
+= i
;
369 ip6af
->ip6af_frglen
-= i
;
374 * While we overlap succeeding segments trim them or,
375 * if they are completely covered, dequeue them.
377 while (af6
!= (struct ip6asfrag
*)q6
&&
378 ip6af
->ip6af_off
+ ip6af
->ip6af_frglen
> af6
->ip6af_off
) {
379 i
= (ip6af
->ip6af_off
+ ip6af
->ip6af_frglen
) - af6
->ip6af_off
;
380 if (i
< af6
->ip6af_frglen
) {
381 af6
->ip6af_frglen
-= i
;
383 m_adj(IP6_REASS_MBUF(af6
), i
);
386 af6
= af6
->ip6af_down
;
387 m_freem(IP6_REASS_MBUF(af6
->ip6af_up
));
388 frag6_deq(af6
->ip6af_up
);
392 * If the incoming framgent overlaps some existing fragments in
393 * the reassembly queue, drop it, since it is dangerous to override
394 * existing fragments from a security point of view.
396 if (af6
->ip6af_up
!= (struct ip6asfrag
*)q6
) {
397 i
= af6
->ip6af_up
->ip6af_off
+ af6
->ip6af_up
->ip6af_frglen
400 #if 0 /* suppress the noisy log */
401 log(LOG_ERR
, "%d bytes of a fragment from %s "
402 "overlaps the previous fragment\n",
403 i
, ip6_sprintf(&q6
->ip6q_src
));
405 kfree(ip6af
, M_FTABLE
);
409 if (af6
!= (struct ip6asfrag
*)q6
) {
410 i
= (ip6af
->ip6af_off
+ ip6af
->ip6af_frglen
) - af6
->ip6af_off
;
412 #if 0 /* suppress the noisy log */
413 log(LOG_ERR
, "%d bytes of a fragment from %s "
414 "overlaps the succeeding fragment",
415 i
, ip6_sprintf(&q6
->ip6q_src
));
417 kfree(ip6af
, M_FTABLE
);
426 * Stick new segment in its place;
427 * check for complete reassembly.
428 * Move to front of packet queue, as we are
429 * the most recently active fragmented packet.
431 frag6_enq(ip6af
, af6
->ip6af_up
);
433 if (q6
!= ip6q
.ip6q_next
) {
435 frag6_insque(q6
, &ip6q
);
439 for (af6
= q6
->ip6q_down
; af6
!= (struct ip6asfrag
*)q6
;
440 af6
= af6
->ip6af_down
) {
441 if (af6
->ip6af_off
!= next
) {
442 frag6_doing_reass
= 0;
445 next
+= af6
->ip6af_frglen
;
447 if (af6
->ip6af_up
->ip6af_mff
) {
448 frag6_doing_reass
= 0;
453 * Reassembly is complete; concatenate fragments.
455 ip6af
= q6
->ip6q_down
;
456 t
= m
= IP6_REASS_MBUF(ip6af
);
457 af6
= ip6af
->ip6af_down
;
459 while (af6
!= (struct ip6asfrag
*)q6
) {
460 af6dwn
= af6
->ip6af_down
;
464 t
->m_next
= IP6_REASS_MBUF(af6
);
465 m_adj(t
->m_next
, af6
->ip6af_offset
);
466 kfree(af6
, M_FTABLE
);
470 /* adjust offset to point where the original next header starts */
471 offset
= ip6af
->ip6af_offset
- sizeof(struct ip6_frag
);
472 kfree(ip6af
, M_FTABLE
);
473 ip6
= mtod(m
, struct ip6_hdr
*);
474 ip6
->ip6_plen
= htons((u_short
)next
+ offset
- sizeof(struct ip6_hdr
));
475 ip6
->ip6_src
= q6
->ip6q_src
;
476 ip6
->ip6_dst
= q6
->ip6q_dst
;
479 *q6
->ip6q_nxtp
= (u_char
)(nxt
& 0xff);
483 * Delete frag6 header with as a few cost as possible.
485 if (offset
< m
->m_len
) {
486 ovbcopy((caddr_t
)ip6
, (caddr_t
)ip6
+ sizeof(struct ip6_frag
),
488 m
->m_data
+= sizeof(struct ip6_frag
);
489 m
->m_len
-= sizeof(struct ip6_frag
);
491 /* this comes with no copy if the boundary is on cluster */
492 if ((t
= m_split(m
, offset
, MB_DONTWAIT
)) == NULL
) {
495 frag6_nfragpackets
--;
498 m_adj(t
, sizeof(struct ip6_frag
));
503 * Store NXT to the original.
506 char *prvnxtp
= ip6_get_prevhdr(m
, offset
); /* XXX */
512 frag6_nfragpackets
--;
514 if (m
->m_flags
& M_PKTHDR
) { /* Isn't it always true? */
516 for (t
= m
; t
; t
= t
->m_next
)
518 m
->m_pkthdr
.len
= plen
;
521 ip6stat
.ip6s_reassembled
++;
522 in6_ifstat_inc(dstifp
, ifs6_reass_ok
);
525 * Tell launch routine the next header
531 frag6_doing_reass
= 0;
535 in6_ifstat_inc(dstifp
, ifs6_reass_fail
);
536 ip6stat
.ip6s_fragdropped
++;
538 frag6_doing_reass
= 0;
543 * Free a fragment reassembly header and all
544 * associated datagrams.
547 frag6_freef(struct ip6q
*q6
)
549 struct ip6asfrag
*af6
, *down6
;
551 for (af6
= q6
->ip6q_down
; af6
!= (struct ip6asfrag
*)q6
;
553 struct mbuf
*m
= IP6_REASS_MBUF(af6
);
555 down6
= af6
->ip6af_down
;
559 * Return ICMP time exceeded error for the 1st fragment.
560 * Just free other fragments.
562 if (af6
->ip6af_off
== 0) {
566 ip6
= mtod(m
, struct ip6_hdr
*);
568 /* restoure source and destination addresses */
569 ip6
->ip6_src
= q6
->ip6q_src
;
570 ip6
->ip6_dst
= q6
->ip6q_dst
;
572 icmp6_error(m
, ICMP6_TIME_EXCEEDED
,
573 ICMP6_TIME_EXCEED_REASSEMBLY
, 0);
576 kfree(af6
, M_FTABLE
);
580 frag6_nfragpackets
--;
584 * Put an ip fragment on a reassembly chain.
585 * Like insque, but pointers in middle of structure.
588 frag6_enq(struct ip6asfrag
*af6
, struct ip6asfrag
*up6
)
591 af6
->ip6af_down
= up6
->ip6af_down
;
592 up6
->ip6af_down
->ip6af_up
= af6
;
593 up6
->ip6af_down
= af6
;
597 * To frag6_enq as remque is to insque.
600 frag6_deq(struct ip6asfrag
*af6
)
602 af6
->ip6af_up
->ip6af_down
= af6
->ip6af_down
;
603 af6
->ip6af_down
->ip6af_up
= af6
->ip6af_up
;
607 frag6_insque(struct ip6q
*new, struct ip6q
*old
)
609 new->ip6q_prev
= old
;
610 new->ip6q_next
= old
->ip6q_next
;
611 old
->ip6q_next
->ip6q_prev
= new;
612 old
->ip6q_next
= new;
616 frag6_remque(struct ip6q
*p6
)
618 p6
->ip6q_prev
->ip6q_next
= p6
->ip6q_next
;
619 p6
->ip6q_next
->ip6q_prev
= p6
->ip6q_prev
;
623 * IPv6 reassembling timer processing;
624 * if a timer expires on a reassembly
633 frag6_doing_reass
= 1;
636 while (q6
!= &ip6q
) {
639 if (q6
->ip6q_prev
->ip6q_ttl
== 0) {
640 ip6stat
.ip6s_fragtimeout
++;
641 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
642 frag6_freef(q6
->ip6q_prev
);
646 * If we are over the maximum number of fragments
647 * (due to the limit being lowered), drain off
648 * enough to get down to the new limit.
650 while (frag6_nfragpackets
> (u_int
)ip6_maxfragpackets
&&
652 ip6stat
.ip6s_fragoverflow
++;
653 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
654 frag6_freef(ip6q
.ip6q_prev
);
656 frag6_doing_reass
= 0;
660 * Routing changes might produce a better route than we last used;
661 * make sure we notice eventually, even if forwarding only for one
662 * destination and the cache is never replaced.
664 if (ip6_forward_rt
.ro_rt
) {
665 RTFREE(ip6_forward_rt
.ro_rt
);
666 ip6_forward_rt
.ro_rt
= NULL
;
668 if (ipsrcchk_rt
.ro_rt
) {
669 RTFREE(ipsrcchk_rt
.ro_rt
);
670 ipsrcchk_rt
.ro_rt
= NULL
;
678 * Drain off all datagram fragments.
683 if (frag6_doing_reass
)
685 while (ip6q
.ip6q_next
!= &ip6q
) {
686 ip6stat
.ip6s_fragdropped
++;
687 /* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
688 frag6_freef(ip6q
.ip6q_next
);