xref: /netbsd/sys/netinet6/frag6.c (revision bf9ec67e)
1 /*	$NetBSD: frag6.c,v 1.19 2002/05/28 10:11:50 itojun Exp $	*/
2 /*	$KAME: frag6.c,v 1.40 2002/05/27 21:40:31 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
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.
19  *
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
30  * SUCH DAMAGE.
31  */
32 
33 #include <sys/cdefs.h>
34 __KERNEL_RCSID(0, "$NetBSD: frag6.c,v 1.19 2002/05/28 10:11:50 itojun Exp $");
35 
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/malloc.h>
39 #include <sys/mbuf.h>
40 #include <sys/domain.h>
41 #include <sys/protosw.h>
42 #include <sys/socket.h>
43 #include <sys/errno.h>
44 #include <sys/time.h>
45 #include <sys/kernel.h>
46 #include <sys/syslog.h>
47 
48 #include <net/if.h>
49 #include <net/route.h>
50 
51 #include <netinet/in.h>
52 #include <netinet/in_var.h>
53 #include <netinet/ip6.h>
54 #include <netinet6/in6_pcb.h>
55 #include <netinet6/ip6_var.h>
56 #include <netinet/icmp6.h>
57 
58 #include <net/net_osdep.h>
59 
60 /*
61  * Define it to get a correct behavior on per-interface statistics.
62  * You will need to perform an extra routing table lookup, per fragment,
63  * to do it.  This may, or may not be, a performance hit.
64  */
65 #define IN6_IFSTAT_STRICT
66 
67 static void frag6_enq __P((struct ip6asfrag *, struct ip6asfrag *));
68 static void frag6_deq __P((struct ip6asfrag *));
69 static void frag6_insque __P((struct ip6q *, struct ip6q *));
70 static void frag6_remque __P((struct ip6q *));
71 static void frag6_freef __P((struct ip6q *));
72 
73 static int ip6q_locked;
74 u_int frag6_nfragpackets;
75 u_int frag6_nfrags;
76 struct	ip6q ip6q;	/* ip6 reassemble queue */
77 
78 static __inline int ip6q_lock_try __P((void));
79 static __inline void ip6q_unlock __P((void));
80 
81 static __inline int
82 ip6q_lock_try()
83 {
84 	int s;
85 
86 	/*
87 	 * Use splvm() -- we're bloking things that would cause
88 	 * mbuf allocation.
89 	 */
90 	s = splvm();
91 	if (ip6q_locked) {
92 		splx(s);
93 		return (0);
94 	}
95 	ip6q_locked = 1;
96 	splx(s);
97 	return (1);
98 }
99 
100 static __inline void
101 ip6q_unlock()
102 {
103 	int s;
104 
105 	s = splvm();
106 	ip6q_locked = 0;
107 	splx(s);
108 }
109 
110 #ifdef DIAGNOSTIC
111 #define	IP6Q_LOCK()							\
112 do {									\
113 	if (ip6q_lock_try() == 0) {					\
114 		printf("%s:%d: ip6q already locked\n", __FILE__, __LINE__); \
115 		panic("ip6q_lock");					\
116 	}								\
117 } while (0)
118 #define	IP6Q_LOCK_CHECK()						\
119 do {									\
120 	if (ip6q_locked == 0) {						\
121 		printf("%s:%d: ip6q lock not held\n", __FILE__, __LINE__); \
122 		panic("ip6q lock check");				\
123 	}								\
124 } while (0)
125 #else
126 #define	IP6Q_LOCK()		(void) ip6q_lock_try()
127 #define	IP6Q_LOCK_CHECK()	/* nothing */
128 #endif
129 
130 #define	IP6Q_UNLOCK()		ip6q_unlock()
131 
132 #ifndef offsetof		/* XXX */
133 #define	offsetof(type, member)	((size_t)(&((type *)0)->member))
134 #endif
135 
136 /*
137  * Initialise reassembly queue and fragment identifier.
138  */
139 void
140 frag6_init()
141 {
142 
143 	ip6_id = arc4random();
144 	ip6q.ip6q_next = ip6q.ip6q_prev = &ip6q;
145 }
146 
147 /*
148  * In RFC2460, fragment and reassembly rule do not agree with each other,
149  * in terms of next header field handling in fragment header.
150  * While the sender will use the same value for all of the fragmented packets,
151  * receiver is suggested not to check the consistency.
152  *
153  * fragment rule (p20):
154  *	(2) A Fragment header containing:
155  *	The Next Header value that identifies the first header of
156  *	the Fragmentable Part of the original packet.
157  *		-> next header field is same for all fragments
158  *
159  * reassembly rule (p21):
160  *	The Next Header field of the last header of the Unfragmentable
161  *	Part is obtained from the Next Header field of the first
162  *	fragment's Fragment header.
163  *		-> should grab it from the first fragment only
164  *
165  * The following note also contradicts with fragment rule - noone is going to
166  * send different fragment with different next header field.
167  *
168  * additional note (p22):
169  *	The Next Header values in the Fragment headers of different
170  *	fragments of the same original packet may differ.  Only the value
171  *	from the Offset zero fragment packet is used for reassembly.
172  *		-> should grab it from the first fragment only
173  *
174  * There is no explicit reason given in the RFC.  Historical reason maybe?
175  */
176 /*
177  * Fragment input
178  */
179 int
180 frag6_input(mp, offp, proto)
181 	struct mbuf **mp;
182 	int *offp, proto;
183 {
184 	struct mbuf *m = *mp, *t;
185 	struct ip6_hdr *ip6;
186 	struct ip6_frag *ip6f;
187 	struct ip6q *q6;
188 	struct ip6asfrag *af6, *ip6af, *af6dwn;
189 	int offset = *offp, nxt, i, next;
190 	int first_frag = 0;
191 	int fragoff, frgpartlen;	/* must be larger than u_int16_t */
192 	struct ifnet *dstifp;
193 #ifdef IN6_IFSTAT_STRICT
194 	static struct route_in6 ro;
195 	struct sockaddr_in6 *dst;
196 #endif
197 
198 	ip6 = mtod(m, struct ip6_hdr *);
199 #ifndef PULLDOWN_TEST
200 	IP6_EXTHDR_CHECK(m, offset, sizeof(struct ip6_frag), IPPROTO_DONE);
201 	ip6f = (struct ip6_frag *)((caddr_t)ip6 + offset);
202 #else
203 	IP6_EXTHDR_GET(ip6f, struct ip6_frag *, m, offset, sizeof(*ip6f));
204 	if (ip6f == NULL)
205 		return IPPROTO_DONE;
206 #endif
207 
208 	dstifp = NULL;
209 #ifdef IN6_IFSTAT_STRICT
210 	/* find the destination interface of the packet. */
211 	dst = (struct sockaddr_in6 *)&ro.ro_dst;
212 	if (ro.ro_rt
213 	 && ((ro.ro_rt->rt_flags & RTF_UP) == 0
214 	  || !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
215 		RTFREE(ro.ro_rt);
216 		ro.ro_rt = (struct rtentry *)0;
217 	}
218 	if (ro.ro_rt == NULL) {
219 		bzero(dst, sizeof(*dst));
220 		dst->sin6_family = AF_INET6;
221 		dst->sin6_len = sizeof(struct sockaddr_in6);
222 		dst->sin6_addr = ip6->ip6_dst;
223 	}
224 	rtalloc((struct route *)&ro);
225 	if (ro.ro_rt != NULL && ro.ro_rt->rt_ifa != NULL)
226 		dstifp = ((struct in6_ifaddr *)ro.ro_rt->rt_ifa)->ia_ifp;
227 #else
228 	/* we are violating the spec, this is not the destination interface */
229 	if ((m->m_flags & M_PKTHDR) != 0)
230 		dstifp = m->m_pkthdr.rcvif;
231 #endif
232 
233 	/* jumbo payload can't contain a fragment header */
234 	if (ip6->ip6_plen == 0) {
235 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, offset);
236 		in6_ifstat_inc(dstifp, ifs6_reass_fail);
237 		return IPPROTO_DONE;
238 	}
239 
240 	/*
241 	 * check whether fragment packet's fragment length is
242 	 * multiple of 8 octets.
243 	 * sizeof(struct ip6_frag) == 8
244 	 * sizeof(struct ip6_hdr) = 40
245 	 */
246 	if ((ip6f->ip6f_offlg & IP6F_MORE_FRAG) &&
247 	    (((ntohs(ip6->ip6_plen) - offset) & 0x7) != 0)) {
248 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
249 		    offsetof(struct ip6_hdr, ip6_plen));
250 		in6_ifstat_inc(dstifp, ifs6_reass_fail);
251 		return IPPROTO_DONE;
252 	}
253 
254 	ip6stat.ip6s_fragments++;
255 	in6_ifstat_inc(dstifp, ifs6_reass_reqd);
256 
257 	/* offset now points to data portion */
258 	offset += sizeof(struct ip6_frag);
259 
260 	IP6Q_LOCK();
261 
262 	/*
263 	 * Enforce upper bound on number of fragments.
264 	 * If maxfrag is 0, never accept fragments.
265 	 * If maxfrag is -1, accept all fragments without limitation.
266 	 */
267 	if (ip6_maxfrags < 0)
268 		;
269 	else if (frag6_nfrags >= (u_int)ip6_maxfrags)
270 		goto dropfrag;
271 
272 	for (q6 = ip6q.ip6q_next; q6 != &ip6q; q6 = q6->ip6q_next)
273 		if (ip6f->ip6f_ident == q6->ip6q_ident &&
274 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &q6->ip6q_src) &&
275 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &q6->ip6q_dst))
276 			break;
277 
278 	if (q6 == &ip6q) {
279 		/*
280 		 * the first fragment to arrive, create a reassembly queue.
281 		 */
282 		first_frag = 1;
283 
284 		/*
285 		 * Enforce upper bound on number of fragmented packets
286 		 * for which we attempt reassembly;
287 		 * If maxfragpackets is 0, never accept fragments.
288 		 * If maxfragpackets is -1, accept all fragments without
289 		 * limitation.
290 		 */
291 		if (ip6_maxfragpackets < 0)
292 			;
293 		else if (frag6_nfragpackets >= (u_int)ip6_maxfragpackets)
294 			goto dropfrag;
295 		frag6_nfragpackets++;
296 		q6 = (struct ip6q *)malloc(sizeof(struct ip6q), M_FTABLE,
297 		    M_DONTWAIT);
298 		if (q6 == NULL)
299 			goto dropfrag;
300 		bzero(q6, sizeof(*q6));
301 
302 		frag6_insque(q6, &ip6q);
303 
304 		/* ip6q_nxt will be filled afterwards, from 1st fragment */
305 		q6->ip6q_down	= q6->ip6q_up = (struct ip6asfrag *)q6;
306 #ifdef notyet
307 		q6->ip6q_nxtp	= (u_char *)nxtp;
308 #endif
309 		q6->ip6q_ident	= ip6f->ip6f_ident;
310 		q6->ip6q_arrive = 0; /* Is it used anywhere? */
311 		q6->ip6q_ttl 	= IPV6_FRAGTTL;
312 		q6->ip6q_src	= ip6->ip6_src;
313 		q6->ip6q_dst	= ip6->ip6_dst;
314 		q6->ip6q_unfrglen = -1;	/* The 1st fragment has not arrived. */
315 
316 		q6->ip6q_nfrag = 0;
317 	}
318 
319 	/*
320 	 * If it's the 1st fragment, record the length of the
321 	 * unfragmentable part and the next header of the fragment header.
322 	 */
323 	fragoff = ntohs(ip6f->ip6f_offlg & IP6F_OFF_MASK);
324 	if (fragoff == 0) {
325 		q6->ip6q_unfrglen = offset - sizeof(struct ip6_hdr) -
326 		    sizeof(struct ip6_frag);
327 		q6->ip6q_nxt = ip6f->ip6f_nxt;
328 	}
329 
330 	/*
331 	 * Check that the reassembled packet would not exceed 65535 bytes
332 	 * in size.
333 	 * If it would exceed, discard the fragment and return an ICMP error.
334 	 */
335 	frgpartlen = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen) - offset;
336 	if (q6->ip6q_unfrglen >= 0) {
337 		/* The 1st fragment has already arrived. */
338 		if (q6->ip6q_unfrglen + fragoff + frgpartlen > IPV6_MAXPACKET) {
339 			icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
340 			    offset - sizeof(struct ip6_frag) +
341 			    offsetof(struct ip6_frag, ip6f_offlg));
342 			IP6Q_UNLOCK();
343 			return(IPPROTO_DONE);
344 		}
345 	} else if (fragoff + frgpartlen > IPV6_MAXPACKET) {
346 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
347 			    offset - sizeof(struct ip6_frag) +
348 				offsetof(struct ip6_frag, ip6f_offlg));
349 		IP6Q_UNLOCK();
350 		return(IPPROTO_DONE);
351 	}
352 	/*
353 	 * If it's the first fragment, do the above check for each
354 	 * fragment already stored in the reassembly queue.
355 	 */
356 	if (fragoff == 0) {
357 		for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
358 		     af6 = af6dwn) {
359 			af6dwn = af6->ip6af_down;
360 
361 			if (q6->ip6q_unfrglen + af6->ip6af_off + af6->ip6af_frglen >
362 			    IPV6_MAXPACKET) {
363 				struct mbuf *merr = IP6_REASS_MBUF(af6);
364 				struct ip6_hdr *ip6err;
365 				int erroff = af6->ip6af_offset;
366 
367 				/* dequeue the fragment. */
368 				frag6_deq(af6);
369 				free(af6, M_FTABLE);
370 
371 				/* adjust pointer. */
372 				ip6err = mtod(merr, struct ip6_hdr *);
373 
374 				/*
375 				 * Restore source and destination addresses
376 				 * in the erroneous IPv6 header.
377 				 */
378 				ip6err->ip6_src = q6->ip6q_src;
379 				ip6err->ip6_dst = q6->ip6q_dst;
380 
381 				icmp6_error(merr, ICMP6_PARAM_PROB,
382 				    ICMP6_PARAMPROB_HEADER,
383 				    erroff - sizeof(struct ip6_frag) +
384 				    offsetof(struct ip6_frag, ip6f_offlg));
385 			}
386 		}
387 	}
388 
389 	ip6af = (struct ip6asfrag *)malloc(sizeof(struct ip6asfrag), M_FTABLE,
390 	    M_DONTWAIT);
391 	if (ip6af == NULL)
392 		goto dropfrag;
393 	bzero(ip6af, sizeof(*ip6af));
394 	ip6af->ip6af_head = ip6->ip6_flow;
395 	ip6af->ip6af_len = ip6->ip6_plen;
396 	ip6af->ip6af_nxt = ip6->ip6_nxt;
397 	ip6af->ip6af_hlim = ip6->ip6_hlim;
398 	ip6af->ip6af_mff = ip6f->ip6f_offlg & IP6F_MORE_FRAG;
399 	ip6af->ip6af_off = fragoff;
400 	ip6af->ip6af_frglen = frgpartlen;
401 	ip6af->ip6af_offset = offset;
402 	IP6_REASS_MBUF(ip6af) = m;
403 
404 	if (first_frag) {
405 		af6 = (struct ip6asfrag *)q6;
406 		goto insert;
407 	}
408 
409 	/*
410 	 * Find a segment which begins after this one does.
411 	 */
412 	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
413 	     af6 = af6->ip6af_down)
414 		if (af6->ip6af_off > ip6af->ip6af_off)
415 			break;
416 
417 #if 0
418 	/*
419 	 * If there is a preceding segment, it may provide some of
420 	 * our data already.  If so, drop the data from the incoming
421 	 * segment.  If it provides all of our data, drop us.
422 	 */
423 	if (af6->ip6af_up != (struct ip6asfrag *)q6) {
424 		i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
425 			- ip6af->ip6af_off;
426 		if (i > 0) {
427 			if (i >= ip6af->ip6af_frglen)
428 				goto dropfrag;
429 			m_adj(IP6_REASS_MBUF(ip6af), i);
430 			ip6af->ip6af_off += i;
431 			ip6af->ip6af_frglen -= i;
432 		}
433 	}
434 
435 	/*
436 	 * While we overlap succeeding segments trim them or,
437 	 * if they are completely covered, dequeue them.
438 	 */
439 	while (af6 != (struct ip6asfrag *)q6 &&
440 	       ip6af->ip6af_off + ip6af->ip6af_frglen > af6->ip6af_off) {
441 		i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
442 		if (i < af6->ip6af_frglen) {
443 			af6->ip6af_frglen -= i;
444 			af6->ip6af_off += i;
445 			m_adj(IP6_REASS_MBUF(af6), i);
446 			break;
447 		}
448 		af6 = af6->ip6af_down;
449 		m_freem(IP6_REASS_MBUF(af6->ip6af_up));
450 		frag6_deq(af6->ip6af_up);
451 	}
452 #else
453 	/*
454 	 * If the incoming framgent overlaps some existing fragments in
455 	 * the reassembly queue, drop it, since it is dangerous to override
456 	 * existing fragments from a security point of view.
457 	 * We don't know which fragment is the bad guy - here we trust
458 	 * fragment that came in earlier, with no real reason.
459 	 */
460 	if (af6->ip6af_up != (struct ip6asfrag *)q6) {
461 		i = af6->ip6af_up->ip6af_off + af6->ip6af_up->ip6af_frglen
462 			- ip6af->ip6af_off;
463 		if (i > 0) {
464 #if 0				/* suppress the noisy log */
465 			log(LOG_ERR, "%d bytes of a fragment from %s "
466 			    "overlaps the previous fragment\n",
467 			    i, ip6_sprintf(&q6->ip6q_src));
468 #endif
469 			free(ip6af, M_FTABLE);
470 			goto dropfrag;
471 		}
472 	}
473 	if (af6 != (struct ip6asfrag *)q6) {
474 		i = (ip6af->ip6af_off + ip6af->ip6af_frglen) - af6->ip6af_off;
475 		if (i > 0) {
476 #if 0				/* suppress the noisy log */
477 			log(LOG_ERR, "%d bytes of a fragment from %s "
478 			    "overlaps the succeeding fragment",
479 			    i, ip6_sprintf(&q6->ip6q_src));
480 #endif
481 			free(ip6af, M_FTABLE);
482 			goto dropfrag;
483 		}
484 	}
485 #endif
486 
487 insert:
488 
489 	/*
490 	 * Stick new segment in its place;
491 	 * check for complete reassembly.
492 	 * Move to front of packet queue, as we are
493 	 * the most recently active fragmented packet.
494 	 */
495 	frag6_enq(ip6af, af6->ip6af_up);
496 	frag6_nfrags++;
497 	q6->ip6q_nfrag++;
498 #if 0 /* xxx */
499 	if (q6 != ip6q.ip6q_next) {
500 		frag6_remque(q6);
501 		frag6_insque(q6, &ip6q);
502 	}
503 #endif
504 	next = 0;
505 	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
506 	     af6 = af6->ip6af_down) {
507 		if (af6->ip6af_off != next) {
508 			IP6Q_UNLOCK();
509 			return IPPROTO_DONE;
510 		}
511 		next += af6->ip6af_frglen;
512 	}
513 	if (af6->ip6af_up->ip6af_mff) {
514 		IP6Q_UNLOCK();
515 		return IPPROTO_DONE;
516 	}
517 
518 	/*
519 	 * Reassembly is complete; concatenate fragments.
520 	 */
521 	ip6af = q6->ip6q_down;
522 	t = m = IP6_REASS_MBUF(ip6af);
523 	af6 = ip6af->ip6af_down;
524 	frag6_deq(ip6af);
525 	while (af6 != (struct ip6asfrag *)q6) {
526 		af6dwn = af6->ip6af_down;
527 		frag6_deq(af6);
528 		while (t->m_next)
529 			t = t->m_next;
530 		t->m_next = IP6_REASS_MBUF(af6);
531 		m_adj(t->m_next, af6->ip6af_offset);
532 		free(af6, M_FTABLE);
533 		af6 = af6dwn;
534 	}
535 
536 	/* adjust offset to point where the original next header starts */
537 	offset = ip6af->ip6af_offset - sizeof(struct ip6_frag);
538 	free(ip6af, M_FTABLE);
539 	ip6 = mtod(m, struct ip6_hdr *);
540 	ip6->ip6_plen = htons((u_short)next + offset - sizeof(struct ip6_hdr));
541 	ip6->ip6_src = q6->ip6q_src;
542 	ip6->ip6_dst = q6->ip6q_dst;
543 	nxt = q6->ip6q_nxt;
544 #ifdef notyet
545 	*q6->ip6q_nxtp = (u_char)(nxt & 0xff);
546 #endif
547 
548 	/*
549 	 * Delete frag6 header with as a few cost as possible.
550 	 */
551 	if (offset < m->m_len) {
552 		ovbcopy((caddr_t)ip6, (caddr_t)ip6 + sizeof(struct ip6_frag),
553 			offset);
554 		m->m_data += sizeof(struct ip6_frag);
555 		m->m_len -= sizeof(struct ip6_frag);
556 	} else {
557 		/* this comes with no copy if the boundary is on cluster */
558 		if ((t = m_split(m, offset, M_DONTWAIT)) == NULL) {
559 			frag6_remque(q6);
560 			frag6_nfrags -= q6->ip6q_nfrag;
561 			free(q6, M_FTABLE);
562 			frag6_nfragpackets--;
563 			goto dropfrag;
564 		}
565 		m_adj(t, sizeof(struct ip6_frag));
566 		m_cat(m, t);
567 	}
568 
569 	/*
570 	 * Store NXT to the original.
571 	 */
572 	{
573 		char *prvnxtp = ip6_get_prevhdr(m, offset); /* XXX */
574 		*prvnxtp = nxt;
575 	}
576 
577 	frag6_remque(q6);
578 	frag6_nfrags -= q6->ip6q_nfrag;
579 	free(q6, M_FTABLE);
580 	frag6_nfragpackets--;
581 
582 	if (m->m_flags & M_PKTHDR) { /* Isn't it always true? */
583 		int plen = 0;
584 		for (t = m; t; t = t->m_next)
585 			plen += t->m_len;
586 		m->m_pkthdr.len = plen;
587 	}
588 
589 	ip6stat.ip6s_reassembled++;
590 	in6_ifstat_inc(dstifp, ifs6_reass_ok);
591 
592 	/*
593 	 * Tell launch routine the next header
594 	 */
595 
596 	*mp = m;
597 	*offp = offset;
598 
599 	IP6Q_UNLOCK();
600 	return nxt;
601 
602  dropfrag:
603 	in6_ifstat_inc(dstifp, ifs6_reass_fail);
604 	ip6stat.ip6s_fragdropped++;
605 	m_freem(m);
606 	IP6Q_UNLOCK();
607 	return IPPROTO_DONE;
608 }
609 
610 /*
611  * Free a fragment reassembly header and all
612  * associated datagrams.
613  */
614 void
615 frag6_freef(q6)
616 	struct ip6q *q6;
617 {
618 	struct ip6asfrag *af6, *down6;
619 
620 	IP6Q_LOCK_CHECK();
621 
622 	for (af6 = q6->ip6q_down; af6 != (struct ip6asfrag *)q6;
623 	     af6 = down6) {
624 		struct mbuf *m = IP6_REASS_MBUF(af6);
625 
626 		down6 = af6->ip6af_down;
627 		frag6_deq(af6);
628 
629 		/*
630 		 * Return ICMP time exceeded error for the 1st fragment.
631 		 * Just free other fragments.
632 		 */
633 		if (af6->ip6af_off == 0) {
634 			struct ip6_hdr *ip6;
635 
636 			/* adjust pointer */
637 			ip6 = mtod(m, struct ip6_hdr *);
638 
639 			/* restoure source and destination addresses */
640 			ip6->ip6_src = q6->ip6q_src;
641 			ip6->ip6_dst = q6->ip6q_dst;
642 
643 			icmp6_error(m, ICMP6_TIME_EXCEEDED,
644 				    ICMP6_TIME_EXCEED_REASSEMBLY, 0);
645 		} else
646 			m_freem(m);
647 		free(af6, M_FTABLE);
648 	}
649 	frag6_remque(q6);
650 	frag6_nfrags -= q6->ip6q_nfrag;
651 	free(q6, M_FTABLE);
652 	frag6_nfragpackets--;
653 }
654 
655 /*
656  * Put an ip fragment on a reassembly chain.
657  * Like insque, but pointers in middle of structure.
658  */
659 void
660 frag6_enq(af6, up6)
661 	struct ip6asfrag *af6, *up6;
662 {
663 
664 	IP6Q_LOCK_CHECK();
665 
666 	af6->ip6af_up = up6;
667 	af6->ip6af_down = up6->ip6af_down;
668 	up6->ip6af_down->ip6af_up = af6;
669 	up6->ip6af_down = af6;
670 }
671 
672 /*
673  * To frag6_enq as remque is to insque.
674  */
675 void
676 frag6_deq(af6)
677 	struct ip6asfrag *af6;
678 {
679 
680 	IP6Q_LOCK_CHECK();
681 
682 	af6->ip6af_up->ip6af_down = af6->ip6af_down;
683 	af6->ip6af_down->ip6af_up = af6->ip6af_up;
684 }
685 
686 void
687 frag6_insque(new, old)
688 	struct ip6q *new, *old;
689 {
690 
691 	IP6Q_LOCK_CHECK();
692 
693 	new->ip6q_prev = old;
694 	new->ip6q_next = old->ip6q_next;
695 	old->ip6q_next->ip6q_prev= new;
696 	old->ip6q_next = new;
697 }
698 
699 void
700 frag6_remque(p6)
701 	struct ip6q *p6;
702 {
703 
704 	IP6Q_LOCK_CHECK();
705 
706 	p6->ip6q_prev->ip6q_next = p6->ip6q_next;
707 	p6->ip6q_next->ip6q_prev = p6->ip6q_prev;
708 }
709 
710 /*
711  * IPv6 reassembling timer processing;
712  * if a timer expires on a reassembly
713  * queue, discard it.
714  */
715 void
716 frag6_slowtimo()
717 {
718 	struct ip6q *q6;
719 	int s = splsoftnet();
720 
721 	IP6Q_LOCK();
722 	q6 = ip6q.ip6q_next;
723 	if (q6)
724 		while (q6 != &ip6q) {
725 			--q6->ip6q_ttl;
726 			q6 = q6->ip6q_next;
727 			if (q6->ip6q_prev->ip6q_ttl == 0) {
728 				ip6stat.ip6s_fragtimeout++;
729 				/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
730 				frag6_freef(q6->ip6q_prev);
731 			}
732 		}
733 	/*
734 	 * If we are over the maximum number of fragments
735 	 * (due to the limit being lowered), drain off
736 	 * enough to get down to the new limit.
737 	 */
738 	while (frag6_nfragpackets > (u_int)ip6_maxfragpackets &&
739 	    ip6q.ip6q_prev) {
740 		ip6stat.ip6s_fragoverflow++;
741 		/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
742 		frag6_freef(ip6q.ip6q_prev);
743 	}
744 	IP6Q_UNLOCK();
745 
746 #if 0
747 	/*
748 	 * Routing changes might produce a better route than we last used;
749 	 * make sure we notice eventually, even if forwarding only for one
750 	 * destination and the cache is never replaced.
751 	 */
752 	if (ip6_forward_rt.ro_rt) {
753 		RTFREE(ip6_forward_rt.ro_rt);
754 		ip6_forward_rt.ro_rt = 0;
755 	}
756 	if (ipsrcchk_rt.ro_rt) {
757 		RTFREE(ipsrcchk_rt.ro_rt);
758 		ipsrcchk_rt.ro_rt = 0;
759 	}
760 #endif
761 
762 	splx(s);
763 }
764 
765 /*
766  * Drain off all datagram fragments.
767  */
768 void
769 frag6_drain()
770 {
771 
772 	if (ip6q_lock_try() == 0)
773 		return;
774 	while (ip6q.ip6q_next != &ip6q) {
775 		ip6stat.ip6s_fragdropped++;
776 		/* XXX in6_ifstat_inc(ifp, ifs6_reass_fail) */
777 		frag6_freef(ip6q.ip6q_next);
778 	}
779 	IP6Q_UNLOCK();
780 }
781