1 /* $OpenBSD: ip_input.c,v 1.395 2024/06/07 18:24:16 bluhm Exp $ */
2 /* $NetBSD: ip_input.c,v 1.30 1996/03/16 23:53:58 christos Exp $ */
3
4 /*
5 * Copyright (c) 1982, 1986, 1988, 1993
6 * The Regents of the University of California. 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 University 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 REGENTS 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 REGENTS 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 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
33 */
34
35 #include "pf.h"
36 #include "carp.h"
37
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/mbuf.h>
41 #include <sys/domain.h>
42 #include <sys/mutex.h>
43 #include <sys/protosw.h>
44 #include <sys/socket.h>
45 #include <sys/socketvar.h>
46 #include <sys/sysctl.h>
47 #include <sys/pool.h>
48 #include <sys/task.h>
49
50 #include <net/if.h>
51 #include <net/if_var.h>
52 #include <net/if_dl.h>
53 #include <net/route.h>
54 #include <net/netisr.h>
55
56 #include <netinet/in.h>
57 #include <netinet/in_systm.h>
58 #include <netinet/if_ether.h>
59 #include <netinet/ip.h>
60 #include <netinet/in_pcb.h>
61 #include <netinet/in_var.h>
62 #include <netinet/ip_var.h>
63 #include <netinet/ip_icmp.h>
64 #include <net/if_types.h>
65
66 #ifdef INET6
67 #include <netinet6/ip6_var.h>
68 #endif
69
70 #if NPF > 0
71 #include <net/pfvar.h>
72 #endif
73
74 #ifdef MROUTING
75 #include <netinet/ip_mroute.h>
76 #endif
77
78 #ifdef IPSEC
79 #include <netinet/ip_ipsp.h>
80 #endif /* IPSEC */
81
82 #if NCARP > 0
83 #include <netinet/ip_carp.h>
84 #endif
85
86 /* values controllable via sysctl */
87 int ip_forwarding = 0;
88 int ipmforwarding = 0;
89 int ipmultipath = 0;
90 int ip_sendredirects = 1;
91 int ip_dosourceroute = 0;
92 int ip_defttl = IPDEFTTL;
93 int ip_mtudisc = 1;
94 int ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
95 int ip_directedbcast = 0;
96
97 /* Protects `ipq' and `ip_frags'. */
98 struct mutex ipq_mutex = MUTEX_INITIALIZER(IPL_SOFTNET);
99
100 /* IP reassembly queue */
101 LIST_HEAD(, ipq) ipq;
102
103 /* Keep track of memory used for reassembly */
104 int ip_maxqueue = 300;
105 int ip_frags = 0;
106
107 const struct sysctl_bounded_args ipctl_vars[] = {
108 #ifdef MROUTING
109 { IPCTL_MRTPROTO, &ip_mrtproto, SYSCTL_INT_READONLY },
110 #endif
111 { IPCTL_FORWARDING, &ip_forwarding, 0, 2 },
112 { IPCTL_SENDREDIRECTS, &ip_sendredirects, 0, 1 },
113 { IPCTL_DEFTTL, &ip_defttl, 0, 255 },
114 { IPCTL_DIRECTEDBCAST, &ip_directedbcast, 0, 1 },
115 { IPCTL_IPPORT_FIRSTAUTO, &ipport_firstauto, 0, 65535 },
116 { IPCTL_IPPORT_LASTAUTO, &ipport_lastauto, 0, 65535 },
117 { IPCTL_IPPORT_HIFIRSTAUTO, &ipport_hifirstauto, 0, 65535 },
118 { IPCTL_IPPORT_HILASTAUTO, &ipport_hilastauto, 0, 65535 },
119 { IPCTL_IPPORT_MAXQUEUE, &ip_maxqueue, 0, 10000 },
120 { IPCTL_MFORWARDING, &ipmforwarding, 0, 1 },
121 { IPCTL_ARPTIMEOUT, &arpt_keep, 0, INT_MAX },
122 { IPCTL_ARPDOWN, &arpt_down, 0, INT_MAX },
123 };
124
125 struct niqueue ipintrq = NIQUEUE_INITIALIZER(IPQ_MAXLEN, NETISR_IP);
126
127 struct pool ipqent_pool;
128 struct pool ipq_pool;
129
130 struct cpumem *ipcounters;
131
132 int ip_sysctl_ipstat(void *, size_t *, void *);
133
134 static struct mbuf_queue ipsend_mq;
135 static struct mbuf_queue ipsendraw_mq;
136
137 extern struct niqueue arpinq;
138
139 int ip_ours(struct mbuf **, int *, int, int);
140 int ip_dooptions(struct mbuf *, struct ifnet *, int);
141 int in_ouraddr(struct mbuf *, struct ifnet *, struct route *, int);
142
143 int ip_fragcheck(struct mbuf **, int *);
144 struct mbuf * ip_reass(struct ipqent *, struct ipq *);
145 void ip_freef(struct ipq *);
146 void ip_flush(void);
147
148 static void ip_send_dispatch(void *);
149 static void ip_sendraw_dispatch(void *);
150 static struct task ipsend_task = TASK_INITIALIZER(ip_send_dispatch, &ipsend_mq);
151 static struct task ipsendraw_task =
152 TASK_INITIALIZER(ip_sendraw_dispatch, &ipsendraw_mq);
153
154 /*
155 * Used to save the IP options in case a protocol wants to respond
156 * to an incoming packet over the same route if the packet got here
157 * using IP source routing. This allows connection establishment and
158 * maintenance when the remote end is on a network that is not known
159 * to us.
160 */
161 struct ip_srcrt {
162 int isr_nhops; /* number of hops */
163 struct in_addr isr_dst; /* final destination */
164 char isr_nop; /* one NOP to align */
165 char isr_hdr[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN & OFFSET */
166 struct in_addr isr_routes[MAX_IPOPTLEN/sizeof(struct in_addr)];
167 };
168
169 void save_rte(struct mbuf *, u_char *, struct in_addr);
170
171 /*
172 * IP initialization: fill in IP protocol switch table.
173 * All protocols not implemented in kernel go to raw IP protocol handler.
174 */
175 void
ip_init(void)176 ip_init(void)
177 {
178 const struct protosw *pr;
179 int i;
180 const u_int16_t defbaddynamicports_tcp[] = DEFBADDYNAMICPORTS_TCP;
181 const u_int16_t defbaddynamicports_udp[] = DEFBADDYNAMICPORTS_UDP;
182 const u_int16_t defrootonlyports_tcp[] = DEFROOTONLYPORTS_TCP;
183 const u_int16_t defrootonlyports_udp[] = DEFROOTONLYPORTS_UDP;
184
185 ipcounters = counters_alloc(ips_ncounters);
186
187 pool_init(&ipqent_pool, sizeof(struct ipqent), 0,
188 IPL_SOFTNET, 0, "ipqe", NULL);
189 pool_init(&ipq_pool, sizeof(struct ipq), 0,
190 IPL_SOFTNET, 0, "ipq", NULL);
191
192 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
193 if (pr == NULL)
194 panic("ip_init");
195 for (i = 0; i < IPPROTO_MAX; i++)
196 ip_protox[i] = pr - inetsw;
197 for (pr = inetdomain.dom_protosw;
198 pr < inetdomain.dom_protoswNPROTOSW; pr++)
199 if (pr->pr_domain->dom_family == PF_INET &&
200 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW &&
201 pr->pr_protocol < IPPROTO_MAX)
202 ip_protox[pr->pr_protocol] = pr - inetsw;
203 LIST_INIT(&ipq);
204
205 /* Fill in list of ports not to allocate dynamically. */
206 memset(&baddynamicports, 0, sizeof(baddynamicports));
207 for (i = 0; defbaddynamicports_tcp[i] != 0; i++)
208 DP_SET(baddynamicports.tcp, defbaddynamicports_tcp[i]);
209 for (i = 0; defbaddynamicports_udp[i] != 0; i++)
210 DP_SET(baddynamicports.udp, defbaddynamicports_udp[i]);
211
212 /* Fill in list of ports only root can bind to. */
213 memset(&rootonlyports, 0, sizeof(rootonlyports));
214 for (i = 0; defrootonlyports_tcp[i] != 0; i++)
215 DP_SET(rootonlyports.tcp, defrootonlyports_tcp[i]);
216 for (i = 0; defrootonlyports_udp[i] != 0; i++)
217 DP_SET(rootonlyports.udp, defrootonlyports_udp[i]);
218
219 mq_init(&ipsend_mq, 64, IPL_SOFTNET);
220 mq_init(&ipsendraw_mq, 64, IPL_SOFTNET);
221
222 arpinit();
223 #ifdef IPSEC
224 ipsec_init();
225 #endif
226 #ifdef MROUTING
227 rt_timer_queue_init(&ip_mrouterq, MCAST_EXPIRE_FREQUENCY,
228 &mfc_expire_route);
229 #endif
230 }
231
232 /*
233 * Enqueue packet for local delivery. Queuing is used as a boundary
234 * between the network layer (input/forward path) running with
235 * NET_LOCK_SHARED() and the transport layer needing it exclusively.
236 */
237 int
ip_ours(struct mbuf ** mp,int * offp,int nxt,int af)238 ip_ours(struct mbuf **mp, int *offp, int nxt, int af)
239 {
240 nxt = ip_fragcheck(mp, offp);
241 if (nxt == IPPROTO_DONE)
242 return IPPROTO_DONE;
243
244 /* We are already in a IPv4/IPv6 local deliver loop. */
245 if (af != AF_UNSPEC)
246 return nxt;
247
248 nxt = ip_deliver(mp, offp, nxt, AF_INET, 1);
249 if (nxt == IPPROTO_DONE)
250 return IPPROTO_DONE;
251
252 /* save values for later, use after dequeue */
253 if (*offp != sizeof(struct ip)) {
254 struct m_tag *mtag;
255 struct ipoffnxt *ion;
256
257 /* mbuf tags are expensive, but only used for header options */
258 mtag = m_tag_get(PACKET_TAG_IP_OFFNXT, sizeof(*ion),
259 M_NOWAIT);
260 if (mtag == NULL) {
261 ipstat_inc(ips_idropped);
262 m_freemp(mp);
263 return IPPROTO_DONE;
264 }
265 ion = (struct ipoffnxt *)(mtag + 1);
266 ion->ion_off = *offp;
267 ion->ion_nxt = nxt;
268
269 m_tag_prepend(*mp, mtag);
270 }
271
272 niq_enqueue(&ipintrq, *mp);
273 *mp = NULL;
274 return IPPROTO_DONE;
275 }
276
277 /*
278 * Dequeue and process locally delivered packets.
279 * This is called with exclusive NET_LOCK().
280 */
281 void
ipintr(void)282 ipintr(void)
283 {
284 struct mbuf *m;
285
286 while ((m = niq_dequeue(&ipintrq)) != NULL) {
287 struct m_tag *mtag;
288 int off, nxt;
289
290 #ifdef DIAGNOSTIC
291 if ((m->m_flags & M_PKTHDR) == 0)
292 panic("ipintr no HDR");
293 #endif
294 mtag = m_tag_find(m, PACKET_TAG_IP_OFFNXT, NULL);
295 if (mtag != NULL) {
296 struct ipoffnxt *ion;
297
298 ion = (struct ipoffnxt *)(mtag + 1);
299 off = ion->ion_off;
300 nxt = ion->ion_nxt;
301
302 m_tag_delete(m, mtag);
303 } else {
304 struct ip *ip;
305
306 ip = mtod(m, struct ip *);
307 off = ip->ip_hl << 2;
308 nxt = ip->ip_p;
309 }
310
311 nxt = ip_deliver(&m, &off, nxt, AF_INET, 0);
312 KASSERT(nxt == IPPROTO_DONE);
313 }
314 }
315
316 /*
317 * IPv4 input routine.
318 *
319 * Checksum and byte swap header. Process options. Forward or deliver.
320 */
321 void
ipv4_input(struct ifnet * ifp,struct mbuf * m)322 ipv4_input(struct ifnet *ifp, struct mbuf *m)
323 {
324 int off, nxt;
325
326 off = 0;
327 nxt = ip_input_if(&m, &off, IPPROTO_IPV4, AF_UNSPEC, ifp);
328 KASSERT(nxt == IPPROTO_DONE);
329 }
330
331 struct mbuf *
ipv4_check(struct ifnet * ifp,struct mbuf * m)332 ipv4_check(struct ifnet *ifp, struct mbuf *m)
333 {
334 struct ip *ip;
335 int hlen, len;
336
337 if (m->m_len < sizeof(*ip)) {
338 m = m_pullup(m, sizeof(*ip));
339 if (m == NULL) {
340 ipstat_inc(ips_toosmall);
341 return (NULL);
342 }
343 }
344
345 ip = mtod(m, struct ip *);
346 if (ip->ip_v != IPVERSION) {
347 ipstat_inc(ips_badvers);
348 goto bad;
349 }
350
351 hlen = ip->ip_hl << 2;
352 if (hlen < sizeof(*ip)) { /* minimum header length */
353 ipstat_inc(ips_badhlen);
354 goto bad;
355 }
356 if (hlen > m->m_len) {
357 m = m_pullup(m, hlen);
358 if (m == NULL) {
359 ipstat_inc(ips_badhlen);
360 return (NULL);
361 }
362 ip = mtod(m, struct ip *);
363 }
364
365 /* 127/8 must not appear on wire - RFC1122 */
366 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
367 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
368 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
369 ipstat_inc(ips_badaddr);
370 goto bad;
371 }
372 }
373
374 if (!ISSET(m->m_pkthdr.csum_flags, M_IPV4_CSUM_IN_OK)) {
375 if (ISSET(m->m_pkthdr.csum_flags, M_IPV4_CSUM_IN_BAD)) {
376 ipstat_inc(ips_badsum);
377 goto bad;
378 }
379
380 ipstat_inc(ips_inswcsum);
381 if (in_cksum(m, hlen) != 0) {
382 ipstat_inc(ips_badsum);
383 goto bad;
384 }
385
386 SET(m->m_pkthdr.csum_flags, M_IPV4_CSUM_IN_OK);
387 }
388
389 /* Retrieve the packet length. */
390 len = ntohs(ip->ip_len);
391
392 /*
393 * Convert fields to host representation.
394 */
395 if (len < hlen) {
396 ipstat_inc(ips_badlen);
397 goto bad;
398 }
399
400 /*
401 * Check that the amount of data in the buffers
402 * is at least as much as the IP header would have us expect.
403 * Trim mbufs if longer than we expect.
404 * Drop packet if shorter than we expect.
405 */
406 if (m->m_pkthdr.len < len) {
407 ipstat_inc(ips_tooshort);
408 goto bad;
409 }
410 if (m->m_pkthdr.len > len) {
411 if (m->m_len == m->m_pkthdr.len) {
412 m->m_len = len;
413 m->m_pkthdr.len = len;
414 } else
415 m_adj(m, len - m->m_pkthdr.len);
416 }
417
418 return (m);
419 bad:
420 m_freem(m);
421 return (NULL);
422 }
423
424 int
ip_input_if(struct mbuf ** mp,int * offp,int nxt,int af,struct ifnet * ifp)425 ip_input_if(struct mbuf **mp, int *offp, int nxt, int af, struct ifnet *ifp)
426 {
427 struct route ro;
428 struct mbuf *m;
429 struct ip *ip;
430 int hlen;
431 #if NPF > 0
432 struct in_addr odst;
433 #endif
434 int flags = 0;
435
436 KASSERT(*offp == 0);
437
438 ro.ro_rt = NULL;
439 ipstat_inc(ips_total);
440 m = *mp = ipv4_check(ifp, *mp);
441 if (m == NULL)
442 goto bad;
443
444 ip = mtod(m, struct ip *);
445
446 #if NCARP > 0
447 if (carp_lsdrop(ifp, m, AF_INET, &ip->ip_src.s_addr,
448 &ip->ip_dst.s_addr, (ip->ip_p == IPPROTO_ICMP ? 0 : 1)))
449 goto bad;
450 #endif
451
452 #if NPF > 0
453 /*
454 * Packet filter
455 */
456 odst = ip->ip_dst;
457 if (pf_test(AF_INET, PF_IN, ifp, mp) != PF_PASS)
458 goto bad;
459 m = *mp;
460 if (m == NULL)
461 goto bad;
462
463 ip = mtod(m, struct ip *);
464 if (odst.s_addr != ip->ip_dst.s_addr)
465 SET(flags, IP_REDIRECT);
466 #endif
467
468 if (ip_forwarding != 0)
469 SET(flags, IP_FORWARDING);
470 if (ip_directedbcast)
471 SET(flags, IP_ALLOWBROADCAST);
472
473 hlen = ip->ip_hl << 2;
474
475 /*
476 * Process options and, if not destined for us,
477 * ship it on. ip_dooptions returns 1 when an
478 * error was detected (causing an icmp message
479 * to be sent and the original packet to be freed).
480 */
481 if (hlen > sizeof (struct ip) && ip_dooptions(m, ifp, flags)) {
482 m = *mp = NULL;
483 goto bad;
484 }
485
486 if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
487 ip->ip_dst.s_addr == INADDR_ANY) {
488 nxt = ip_ours(mp, offp, nxt, af);
489 goto out;
490 }
491
492 switch(in_ouraddr(m, ifp, &ro, flags)) {
493 case 2:
494 goto bad;
495 case 1:
496 nxt = ip_ours(mp, offp, nxt, af);
497 goto out;
498 }
499
500 if (IN_MULTICAST(ip->ip_dst.s_addr)) {
501 /*
502 * Make sure M_MCAST is set. It should theoretically
503 * already be there, but let's play safe because upper
504 * layers check for this flag.
505 */
506 m->m_flags |= M_MCAST;
507
508 #ifdef MROUTING
509 if (ipmforwarding && ip_mrouter[ifp->if_rdomain]) {
510 int error;
511
512 if (m->m_flags & M_EXT) {
513 if ((m = *mp = m_pullup(m, hlen)) == NULL) {
514 ipstat_inc(ips_toosmall);
515 goto bad;
516 }
517 ip = mtod(m, struct ip *);
518 }
519 /*
520 * If we are acting as a multicast router, all
521 * incoming multicast packets are passed to the
522 * kernel-level multicast forwarding function.
523 * The packet is returned (relatively) intact; if
524 * ip_mforward() returns a non-zero value, the packet
525 * must be discarded, else it may be accepted below.
526 *
527 * (The IP ident field is put in the same byte order
528 * as expected when ip_mforward() is called from
529 * ip_output().)
530 */
531 KERNEL_LOCK();
532 error = ip_mforward(m, ifp);
533 KERNEL_UNLOCK();
534 if (error) {
535 ipstat_inc(ips_cantforward);
536 goto bad;
537 }
538
539 /*
540 * The process-level routing daemon needs to receive
541 * all multicast IGMP packets, whether or not this
542 * host belongs to their destination groups.
543 */
544 if (ip->ip_p == IPPROTO_IGMP) {
545 nxt = ip_ours(mp, offp, nxt, af);
546 goto out;
547 }
548 ipstat_inc(ips_forward);
549 }
550 #endif
551 /*
552 * See if we belong to the destination multicast group on the
553 * arrival interface.
554 */
555 if (!in_hasmulti(&ip->ip_dst, ifp)) {
556 ipstat_inc(ips_notmember);
557 if (!IN_LOCAL_GROUP(ip->ip_dst.s_addr))
558 ipstat_inc(ips_cantforward);
559 goto bad;
560 }
561 nxt = ip_ours(mp, offp, nxt, af);
562 goto out;
563 }
564
565 #if NCARP > 0
566 if (ip->ip_p == IPPROTO_ICMP &&
567 carp_lsdrop(ifp, m, AF_INET, &ip->ip_src.s_addr,
568 &ip->ip_dst.s_addr, 1))
569 goto bad;
570 #endif
571 /*
572 * Not for us; forward if possible and desirable.
573 */
574 if (!ISSET(flags, IP_FORWARDING)) {
575 ipstat_inc(ips_cantforward);
576 goto bad;
577 }
578 #ifdef IPSEC
579 if (ipsec_in_use) {
580 int rv;
581
582 rv = ipsec_forward_check(m, hlen, AF_INET);
583 if (rv != 0) {
584 ipstat_inc(ips_cantforward);
585 goto bad;
586 }
587 /*
588 * Fall through, forward packet. Outbound IPsec policy
589 * checking will occur in ip_output().
590 */
591 }
592 #endif /* IPSEC */
593
594 ip_forward(m, ifp, &ro, flags);
595 *mp = NULL;
596 rtfree(ro.ro_rt);
597 return IPPROTO_DONE;
598 bad:
599 nxt = IPPROTO_DONE;
600 m_freemp(mp);
601 out:
602 rtfree(ro.ro_rt);
603 return nxt;
604 }
605
606 int
ip_fragcheck(struct mbuf ** mp,int * offp)607 ip_fragcheck(struct mbuf **mp, int *offp)
608 {
609 struct ip *ip;
610 struct ipq *fp;
611 struct ipqent *ipqe;
612 int hlen;
613 uint16_t mff;
614
615 ip = mtod(*mp, struct ip *);
616 hlen = ip->ip_hl << 2;
617
618 /*
619 * If offset or more fragments are set, must reassemble.
620 * Otherwise, nothing need be done.
621 * (We could look in the reassembly queue to see
622 * if the packet was previously fragmented,
623 * but it's not worth the time; just let them time out.)
624 */
625 if (ISSET(ip->ip_off, htons(IP_OFFMASK | IP_MF))) {
626 if ((*mp)->m_flags & M_EXT) { /* XXX */
627 if ((*mp = m_pullup(*mp, hlen)) == NULL) {
628 ipstat_inc(ips_toosmall);
629 return IPPROTO_DONE;
630 }
631 ip = mtod(*mp, struct ip *);
632 }
633
634 /*
635 * Adjust ip_len to not reflect header,
636 * set ipqe_mff if more fragments are expected,
637 * convert offset of this to bytes.
638 */
639 ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
640 mff = ISSET(ip->ip_off, htons(IP_MF));
641 if (mff) {
642 /*
643 * Make sure that fragments have a data length
644 * that's a non-zero multiple of 8 bytes.
645 */
646 if (ntohs(ip->ip_len) == 0 ||
647 (ntohs(ip->ip_len) & 0x7) != 0) {
648 ipstat_inc(ips_badfrags);
649 m_freemp(mp);
650 return IPPROTO_DONE;
651 }
652 }
653 ip->ip_off = htons(ntohs(ip->ip_off) << 3);
654
655 mtx_enter(&ipq_mutex);
656
657 /*
658 * Look for queue of fragments
659 * of this datagram.
660 */
661 LIST_FOREACH(fp, &ipq, ipq_q) {
662 if (ip->ip_id == fp->ipq_id &&
663 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
664 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
665 ip->ip_p == fp->ipq_p)
666 break;
667 }
668
669 /*
670 * If datagram marked as having more fragments
671 * or if this is not the first fragment,
672 * attempt reassembly; if it succeeds, proceed.
673 */
674 if (mff || ip->ip_off) {
675 ipstat_inc(ips_fragments);
676 if (ip_frags + 1 > ip_maxqueue) {
677 ip_flush();
678 ipstat_inc(ips_rcvmemdrop);
679 goto bad;
680 }
681
682 ipqe = pool_get(&ipqent_pool, PR_NOWAIT);
683 if (ipqe == NULL) {
684 ipstat_inc(ips_rcvmemdrop);
685 goto bad;
686 }
687 ip_frags++;
688 ipqe->ipqe_mff = mff;
689 ipqe->ipqe_m = *mp;
690 ipqe->ipqe_ip = ip;
691 *mp = ip_reass(ipqe, fp);
692 if (*mp == NULL)
693 goto bad;
694 ipstat_inc(ips_reassembled);
695 ip = mtod(*mp, struct ip *);
696 hlen = ip->ip_hl << 2;
697 ip->ip_len = htons(ntohs(ip->ip_len) + hlen);
698 } else {
699 if (fp != NULL)
700 ip_freef(fp);
701 }
702
703 mtx_leave(&ipq_mutex);
704 }
705
706 *offp = hlen;
707 return ip->ip_p;
708
709 bad:
710 mtx_leave(&ipq_mutex);
711 m_freemp(mp);
712 return IPPROTO_DONE;
713 }
714
715 #ifndef INET6
716 #define IPSTAT_INC(name) ipstat_inc(ips_##name)
717 #else
718 #define IPSTAT_INC(name) (af == AF_INET ? \
719 ipstat_inc(ips_##name) : ip6stat_inc(ip6s_##name))
720 #endif
721
722 int
ip_deliver(struct mbuf ** mp,int * offp,int nxt,int af,int shared)723 ip_deliver(struct mbuf **mp, int *offp, int nxt, int af, int shared)
724 {
725 #ifdef INET6
726 int nest = 0;
727 #endif
728
729 /*
730 * Tell launch routine the next header
731 */
732 IPSTAT_INC(delivered);
733
734 while (nxt != IPPROTO_DONE) {
735 const struct protosw *psw;
736 int naf;
737
738 switch (af) {
739 case AF_INET:
740 psw = &inetsw[ip_protox[nxt]];
741 break;
742 #ifdef INET6
743 case AF_INET6:
744 psw = &inet6sw[ip6_protox[nxt]];
745 break;
746 #endif
747 }
748 if (shared && !ISSET(psw->pr_flags, PR_MPINPUT)) {
749 /* delivery not finished, decrement counter, queue */
750 switch (af) {
751 case AF_INET:
752 counters_dec(ipcounters, ips_delivered);
753 break;
754 #ifdef INET6
755 case AF_INET6:
756 counters_dec(ip6counters, ip6s_delivered);
757 break;
758 #endif
759 }
760 break;
761 }
762
763 #ifdef INET6
764 if (af == AF_INET6 &&
765 ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
766 ip6stat_inc(ip6s_toomanyhdr);
767 goto bad;
768 }
769 #endif
770
771 /*
772 * protection against faulty packet - there should be
773 * more sanity checks in header chain processing.
774 */
775 if ((*mp)->m_pkthdr.len < *offp) {
776 IPSTAT_INC(tooshort);
777 goto bad;
778 }
779
780 #ifdef IPSEC
781 if (ipsec_in_use) {
782 if (ipsec_local_check(*mp, *offp, nxt, af) != 0) {
783 IPSTAT_INC(cantforward);
784 goto bad;
785 }
786 }
787 /* Otherwise, just fall through and deliver the packet */
788 #endif
789
790 switch (nxt) {
791 case IPPROTO_IPV4:
792 naf = AF_INET;
793 ipstat_inc(ips_delivered);
794 break;
795 #ifdef INET6
796 case IPPROTO_IPV6:
797 naf = AF_INET6;
798 ip6stat_inc(ip6s_delivered);
799 break;
800 #endif
801 default:
802 naf = af;
803 break;
804 }
805 nxt = (*psw->pr_input)(mp, offp, nxt, af);
806 af = naf;
807 }
808 return nxt;
809 bad:
810 m_freemp(mp);
811 return IPPROTO_DONE;
812 }
813 #undef IPSTAT_INC
814
815 int
in_ouraddr(struct mbuf * m,struct ifnet * ifp,struct route * ro,int flags)816 in_ouraddr(struct mbuf *m, struct ifnet *ifp, struct route *ro, int flags)
817 {
818 struct rtentry *rt;
819 struct ip *ip;
820 int match = 0;
821
822 #if NPF > 0
823 switch (pf_ouraddr(m)) {
824 case 0:
825 return (0);
826 case 1:
827 return (1);
828 default:
829 /* pf does not know it */
830 break;
831 }
832 #endif
833
834 ip = mtod(m, struct ip *);
835
836 rt = route_mpath(ro, &ip->ip_dst, &ip->ip_src, m->m_pkthdr.ph_rtableid);
837 if (rt != NULL) {
838 if (ISSET(rt->rt_flags, RTF_LOCAL))
839 match = 1;
840
841 /*
842 * If directedbcast is enabled we only consider it local
843 * if it is received on the interface with that address.
844 */
845 if (ISSET(rt->rt_flags, RTF_BROADCAST) &&
846 (!ISSET(flags, IP_ALLOWBROADCAST) ||
847 rt->rt_ifidx == ifp->if_index)) {
848 match = 1;
849
850 /* Make sure M_BCAST is set */
851 m->m_flags |= M_BCAST;
852 }
853 }
854
855 if (!match) {
856 struct ifaddr *ifa;
857
858 /*
859 * No local address or broadcast address found, so check for
860 * ancient classful broadcast addresses.
861 * It must have been broadcast on the link layer, and for an
862 * address on the interface it was received on.
863 */
864 if (!ISSET(m->m_flags, M_BCAST) ||
865 !IN_CLASSFULBROADCAST(ip->ip_dst.s_addr, ip->ip_dst.s_addr))
866 return (0);
867
868 if (ifp->if_rdomain != rtable_l2(m->m_pkthdr.ph_rtableid))
869 return (0);
870 /*
871 * The check in the loop assumes you only rx a packet on an UP
872 * interface, and that M_BCAST will only be set on a BROADCAST
873 * interface.
874 */
875 NET_ASSERT_LOCKED();
876 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
877 if (ifa->ifa_addr->sa_family != AF_INET)
878 continue;
879
880 if (IN_CLASSFULBROADCAST(ip->ip_dst.s_addr,
881 ifatoia(ifa)->ia_addr.sin_addr.s_addr)) {
882 match = 1;
883 break;
884 }
885 }
886 } else if (!ISSET(flags, IP_FORWARDING) &&
887 rt->rt_ifidx != ifp->if_index &&
888 !((ifp->if_flags & IFF_LOOPBACK) || (ifp->if_type == IFT_ENC) ||
889 (m->m_pkthdr.pf.flags & PF_TAG_TRANSLATE_LOCALHOST))) {
890 /* received on wrong interface. */
891 #if NCARP > 0
892 struct ifnet *out_if;
893
894 /*
895 * Virtual IPs on carp interfaces need to be checked also
896 * against the parent interface and other carp interfaces
897 * sharing the same parent.
898 */
899 out_if = if_get(rt->rt_ifidx);
900 if (!(out_if && carp_strict_addr_chk(out_if, ifp))) {
901 ipstat_inc(ips_wrongif);
902 match = 2;
903 }
904 if_put(out_if);
905 #else
906 ipstat_inc(ips_wrongif);
907 match = 2;
908 #endif
909 }
910
911 return (match);
912 }
913
914 /*
915 * Take incoming datagram fragment and try to
916 * reassemble it into whole datagram. If a chain for
917 * reassembly of this datagram already exists, then it
918 * is given as fp; otherwise have to make a chain.
919 */
920 struct mbuf *
ip_reass(struct ipqent * ipqe,struct ipq * fp)921 ip_reass(struct ipqent *ipqe, struct ipq *fp)
922 {
923 struct mbuf *m = ipqe->ipqe_m;
924 struct ipqent *nq, *p, *q;
925 struct ip *ip;
926 struct mbuf *t;
927 int hlen = ipqe->ipqe_ip->ip_hl << 2;
928 int i, next;
929 u_int8_t ecn, ecn0;
930
931 MUTEX_ASSERT_LOCKED(&ipq_mutex);
932
933 /*
934 * Presence of header sizes in mbufs
935 * would confuse code below.
936 */
937 m->m_data += hlen;
938 m->m_len -= hlen;
939
940 /*
941 * If first fragment to arrive, create a reassembly queue.
942 */
943 if (fp == NULL) {
944 fp = pool_get(&ipq_pool, PR_NOWAIT);
945 if (fp == NULL)
946 goto dropfrag;
947 LIST_INSERT_HEAD(&ipq, fp, ipq_q);
948 fp->ipq_ttl = IPFRAGTTL;
949 fp->ipq_p = ipqe->ipqe_ip->ip_p;
950 fp->ipq_id = ipqe->ipqe_ip->ip_id;
951 LIST_INIT(&fp->ipq_fragq);
952 fp->ipq_src = ipqe->ipqe_ip->ip_src;
953 fp->ipq_dst = ipqe->ipqe_ip->ip_dst;
954 p = NULL;
955 goto insert;
956 }
957
958 /*
959 * Handle ECN by comparing this segment with the first one;
960 * if CE is set, do not lose CE.
961 * drop if CE and not-ECT are mixed for the same packet.
962 */
963 ecn = ipqe->ipqe_ip->ip_tos & IPTOS_ECN_MASK;
964 ecn0 = LIST_FIRST(&fp->ipq_fragq)->ipqe_ip->ip_tos & IPTOS_ECN_MASK;
965 if (ecn == IPTOS_ECN_CE) {
966 if (ecn0 == IPTOS_ECN_NOTECT)
967 goto dropfrag;
968 if (ecn0 != IPTOS_ECN_CE)
969 LIST_FIRST(&fp->ipq_fragq)->ipqe_ip->ip_tos |=
970 IPTOS_ECN_CE;
971 }
972 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
973 goto dropfrag;
974
975 /*
976 * Find a segment which begins after this one does.
977 */
978 for (p = NULL, q = LIST_FIRST(&fp->ipq_fragq); q != NULL;
979 p = q, q = LIST_NEXT(q, ipqe_q))
980 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off))
981 break;
982
983 /*
984 * If there is a preceding segment, it may provide some of
985 * our data already. If so, drop the data from the incoming
986 * segment. If it provides all of our data, drop us.
987 */
988 if (p != NULL) {
989 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) -
990 ntohs(ipqe->ipqe_ip->ip_off);
991 if (i > 0) {
992 if (i >= ntohs(ipqe->ipqe_ip->ip_len))
993 goto dropfrag;
994 m_adj(ipqe->ipqe_m, i);
995 ipqe->ipqe_ip->ip_off =
996 htons(ntohs(ipqe->ipqe_ip->ip_off) + i);
997 ipqe->ipqe_ip->ip_len =
998 htons(ntohs(ipqe->ipqe_ip->ip_len) - i);
999 }
1000 }
1001
1002 /*
1003 * While we overlap succeeding segments trim them or,
1004 * if they are completely covered, dequeue them.
1005 */
1006 for (; q != NULL &&
1007 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) >
1008 ntohs(q->ipqe_ip->ip_off); q = nq) {
1009 i = (ntohs(ipqe->ipqe_ip->ip_off) +
1010 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off);
1011 if (i < ntohs(q->ipqe_ip->ip_len)) {
1012 q->ipqe_ip->ip_len =
1013 htons(ntohs(q->ipqe_ip->ip_len) - i);
1014 q->ipqe_ip->ip_off =
1015 htons(ntohs(q->ipqe_ip->ip_off) + i);
1016 m_adj(q->ipqe_m, i);
1017 break;
1018 }
1019 nq = LIST_NEXT(q, ipqe_q);
1020 m_freem(q->ipqe_m);
1021 LIST_REMOVE(q, ipqe_q);
1022 pool_put(&ipqent_pool, q);
1023 ip_frags--;
1024 }
1025
1026 insert:
1027 /*
1028 * Stick new segment in its place;
1029 * check for complete reassembly.
1030 */
1031 if (p == NULL) {
1032 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q);
1033 } else {
1034 LIST_INSERT_AFTER(p, ipqe, ipqe_q);
1035 }
1036 next = 0;
1037 for (p = NULL, q = LIST_FIRST(&fp->ipq_fragq); q != NULL;
1038 p = q, q = LIST_NEXT(q, ipqe_q)) {
1039 if (ntohs(q->ipqe_ip->ip_off) != next)
1040 return (0);
1041 next += ntohs(q->ipqe_ip->ip_len);
1042 }
1043 if (p->ipqe_mff)
1044 return (0);
1045
1046 /*
1047 * Reassembly is complete. Check for a bogus message size and
1048 * concatenate fragments.
1049 */
1050 q = LIST_FIRST(&fp->ipq_fragq);
1051 ip = q->ipqe_ip;
1052 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) {
1053 ipstat_inc(ips_toolong);
1054 ip_freef(fp);
1055 return (0);
1056 }
1057 m = q->ipqe_m;
1058 t = m->m_next;
1059 m->m_next = 0;
1060 m_cat(m, t);
1061 nq = LIST_NEXT(q, ipqe_q);
1062 pool_put(&ipqent_pool, q);
1063 ip_frags--;
1064 for (q = nq; q != NULL; q = nq) {
1065 t = q->ipqe_m;
1066 nq = LIST_NEXT(q, ipqe_q);
1067 pool_put(&ipqent_pool, q);
1068 ip_frags--;
1069 m_removehdr(t);
1070 m_cat(m, t);
1071 }
1072
1073 /*
1074 * Create header for new ip packet by
1075 * modifying header of first packet;
1076 * dequeue and discard fragment reassembly header.
1077 * Make header visible.
1078 */
1079 ip->ip_len = htons(next);
1080 ip->ip_src = fp->ipq_src;
1081 ip->ip_dst = fp->ipq_dst;
1082 LIST_REMOVE(fp, ipq_q);
1083 pool_put(&ipq_pool, fp);
1084 m->m_len += (ip->ip_hl << 2);
1085 m->m_data -= (ip->ip_hl << 2);
1086 m_calchdrlen(m);
1087 return (m);
1088
1089 dropfrag:
1090 ipstat_inc(ips_fragdropped);
1091 m_freem(m);
1092 pool_put(&ipqent_pool, ipqe);
1093 ip_frags--;
1094 return (NULL);
1095 }
1096
1097 /*
1098 * Free a fragment reassembly header and all
1099 * associated datagrams.
1100 */
1101 void
ip_freef(struct ipq * fp)1102 ip_freef(struct ipq *fp)
1103 {
1104 struct ipqent *q;
1105
1106 MUTEX_ASSERT_LOCKED(&ipq_mutex);
1107
1108 while ((q = LIST_FIRST(&fp->ipq_fragq)) != NULL) {
1109 LIST_REMOVE(q, ipqe_q);
1110 m_freem(q->ipqe_m);
1111 pool_put(&ipqent_pool, q);
1112 ip_frags--;
1113 }
1114 LIST_REMOVE(fp, ipq_q);
1115 pool_put(&ipq_pool, fp);
1116 }
1117
1118 /*
1119 * IP timer processing;
1120 * if a timer expires on a reassembly queue, discard it.
1121 */
1122 void
ip_slowtimo(void)1123 ip_slowtimo(void)
1124 {
1125 struct ipq *fp, *nfp;
1126
1127 mtx_enter(&ipq_mutex);
1128 LIST_FOREACH_SAFE(fp, &ipq, ipq_q, nfp) {
1129 if (--fp->ipq_ttl == 0) {
1130 ipstat_inc(ips_fragtimeout);
1131 ip_freef(fp);
1132 }
1133 }
1134 mtx_leave(&ipq_mutex);
1135 }
1136
1137 /*
1138 * Flush a bunch of datagram fragments, till we are down to 75%.
1139 */
1140 void
ip_flush(void)1141 ip_flush(void)
1142 {
1143 int max = 50;
1144
1145 MUTEX_ASSERT_LOCKED(&ipq_mutex);
1146
1147 while (!LIST_EMPTY(&ipq) && ip_frags > ip_maxqueue * 3 / 4 && --max) {
1148 ipstat_inc(ips_fragdropped);
1149 ip_freef(LIST_FIRST(&ipq));
1150 }
1151 }
1152
1153 /*
1154 * Do option processing on a datagram,
1155 * possibly discarding it if bad options are encountered,
1156 * or forwarding it if source-routed.
1157 * Returns 1 if packet has been forwarded/freed,
1158 * 0 if the packet should be processed further.
1159 */
1160 int
ip_dooptions(struct mbuf * m,struct ifnet * ifp,int flags)1161 ip_dooptions(struct mbuf *m, struct ifnet *ifp, int flags)
1162 {
1163 struct ip *ip = mtod(m, struct ip *);
1164 unsigned int rtableid = m->m_pkthdr.ph_rtableid;
1165 struct rtentry *rt;
1166 struct sockaddr_in ipaddr;
1167 u_char *cp;
1168 struct ip_timestamp ipt;
1169 struct in_ifaddr *ia;
1170 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1171 struct in_addr sin, dst;
1172 u_int32_t ntime;
1173
1174 dst = ip->ip_dst;
1175 cp = (u_char *)(ip + 1);
1176 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1177
1178 KERNEL_LOCK();
1179 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1180 opt = cp[IPOPT_OPTVAL];
1181 if (opt == IPOPT_EOL)
1182 break;
1183 if (opt == IPOPT_NOP)
1184 optlen = 1;
1185 else {
1186 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1187 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1188 goto bad;
1189 }
1190 optlen = cp[IPOPT_OLEN];
1191 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1192 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1193 goto bad;
1194 }
1195 }
1196
1197 switch (opt) {
1198
1199 default:
1200 break;
1201
1202 /*
1203 * Source routing with record.
1204 * Find interface with current destination address.
1205 * If none on this machine then drop if strictly routed,
1206 * or do nothing if loosely routed.
1207 * Record interface address and bring up next address
1208 * component. If strictly routed make sure next
1209 * address is on directly accessible net.
1210 */
1211 case IPOPT_LSRR:
1212 case IPOPT_SSRR:
1213 if (!ip_dosourceroute) {
1214 type = ICMP_UNREACH;
1215 code = ICMP_UNREACH_SRCFAIL;
1216 goto bad;
1217 }
1218 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1219 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1220 goto bad;
1221 }
1222 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1223 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1224 goto bad;
1225 }
1226 memset(&ipaddr, 0, sizeof(ipaddr));
1227 ipaddr.sin_family = AF_INET;
1228 ipaddr.sin_len = sizeof(ipaddr);
1229 ipaddr.sin_addr = ip->ip_dst;
1230 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr),
1231 m->m_pkthdr.ph_rtableid));
1232 if (ia == NULL) {
1233 if (opt == IPOPT_SSRR) {
1234 type = ICMP_UNREACH;
1235 code = ICMP_UNREACH_SRCFAIL;
1236 goto bad;
1237 }
1238 /*
1239 * Loose routing, and not at next destination
1240 * yet; nothing to do except forward.
1241 */
1242 break;
1243 }
1244 off--; /* 0 origin */
1245 if ((off + sizeof(struct in_addr)) > optlen) {
1246 /*
1247 * End of source route. Should be for us.
1248 */
1249 save_rte(m, cp, ip->ip_src);
1250 break;
1251 }
1252
1253 /*
1254 * locate outgoing interface
1255 */
1256 memset(&ipaddr, 0, sizeof(ipaddr));
1257 ipaddr.sin_family = AF_INET;
1258 ipaddr.sin_len = sizeof(ipaddr);
1259 memcpy(&ipaddr.sin_addr, cp + off,
1260 sizeof(ipaddr.sin_addr));
1261 /* keep packet in the virtual instance */
1262 rt = rtalloc(sintosa(&ipaddr), RT_RESOLVE, rtableid);
1263 if (!rtisvalid(rt) || ((opt == IPOPT_SSRR) &&
1264 ISSET(rt->rt_flags, RTF_GATEWAY))) {
1265 type = ICMP_UNREACH;
1266 code = ICMP_UNREACH_SRCFAIL;
1267 rtfree(rt);
1268 goto bad;
1269 }
1270 ia = ifatoia(rt->rt_ifa);
1271 memcpy(cp + off, &ia->ia_addr.sin_addr,
1272 sizeof(struct in_addr));
1273 rtfree(rt);
1274 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1275 ip->ip_dst = ipaddr.sin_addr;
1276 /*
1277 * Let ip_intr's mcast routing check handle mcast pkts
1278 */
1279 forward = !IN_MULTICAST(ip->ip_dst.s_addr);
1280 break;
1281
1282 case IPOPT_RR:
1283 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1284 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1285 goto bad;
1286 }
1287 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1288 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1289 goto bad;
1290 }
1291
1292 /*
1293 * If no space remains, ignore.
1294 */
1295 off--; /* 0 origin */
1296 if ((off + sizeof(struct in_addr)) > optlen)
1297 break;
1298 memset(&ipaddr, 0, sizeof(ipaddr));
1299 ipaddr.sin_family = AF_INET;
1300 ipaddr.sin_len = sizeof(ipaddr);
1301 ipaddr.sin_addr = ip->ip_dst;
1302 /*
1303 * locate outgoing interface; if we're the destination,
1304 * use the incoming interface (should be same).
1305 * Again keep the packet inside the virtual instance.
1306 */
1307 rt = rtalloc(sintosa(&ipaddr), RT_RESOLVE, rtableid);
1308 if (!rtisvalid(rt)) {
1309 type = ICMP_UNREACH;
1310 code = ICMP_UNREACH_HOST;
1311 rtfree(rt);
1312 goto bad;
1313 }
1314 ia = ifatoia(rt->rt_ifa);
1315 memcpy(cp + off, &ia->ia_addr.sin_addr,
1316 sizeof(struct in_addr));
1317 rtfree(rt);
1318 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1319 break;
1320
1321 case IPOPT_TS:
1322 code = cp - (u_char *)ip;
1323 if (optlen < sizeof(struct ip_timestamp))
1324 goto bad;
1325 memcpy(&ipt, cp, sizeof(struct ip_timestamp));
1326 if (ipt.ipt_ptr < 5 || ipt.ipt_len < 5)
1327 goto bad;
1328 if (ipt.ipt_ptr - 1 + sizeof(u_int32_t) > ipt.ipt_len) {
1329 if (++ipt.ipt_oflw == 0)
1330 goto bad;
1331 break;
1332 }
1333 memcpy(&sin, cp + ipt.ipt_ptr - 1, sizeof sin);
1334 switch (ipt.ipt_flg) {
1335
1336 case IPOPT_TS_TSONLY:
1337 break;
1338
1339 case IPOPT_TS_TSANDADDR:
1340 if (ipt.ipt_ptr - 1 + sizeof(u_int32_t) +
1341 sizeof(struct in_addr) > ipt.ipt_len)
1342 goto bad;
1343 memset(&ipaddr, 0, sizeof(ipaddr));
1344 ipaddr.sin_family = AF_INET;
1345 ipaddr.sin_len = sizeof(ipaddr);
1346 ipaddr.sin_addr = dst;
1347 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr),
1348 ifp));
1349 if (ia == NULL)
1350 continue;
1351 memcpy(&sin, &ia->ia_addr.sin_addr,
1352 sizeof(struct in_addr));
1353 ipt.ipt_ptr += sizeof(struct in_addr);
1354 break;
1355
1356 case IPOPT_TS_PRESPEC:
1357 if (ipt.ipt_ptr - 1 + sizeof(u_int32_t) +
1358 sizeof(struct in_addr) > ipt.ipt_len)
1359 goto bad;
1360 memset(&ipaddr, 0, sizeof(ipaddr));
1361 ipaddr.sin_family = AF_INET;
1362 ipaddr.sin_len = sizeof(ipaddr);
1363 ipaddr.sin_addr = sin;
1364 if (ifa_ifwithaddr(sintosa(&ipaddr),
1365 m->m_pkthdr.ph_rtableid) == NULL)
1366 continue;
1367 ipt.ipt_ptr += sizeof(struct in_addr);
1368 break;
1369
1370 default:
1371 /* XXX can't take &ipt->ipt_flg */
1372 code = (u_char *)&ipt.ipt_ptr -
1373 (u_char *)ip + 1;
1374 goto bad;
1375 }
1376 ntime = iptime();
1377 memcpy(cp + ipt.ipt_ptr - 1, &ntime, sizeof(u_int32_t));
1378 ipt.ipt_ptr += sizeof(u_int32_t);
1379 }
1380 }
1381 KERNEL_UNLOCK();
1382 if (forward && ISSET(flags, IP_FORWARDING)) {
1383 ip_forward(m, ifp, NULL, flags | IP_REDIRECT);
1384 return (1);
1385 }
1386 return (0);
1387 bad:
1388 KERNEL_UNLOCK();
1389 icmp_error(m, type, code, 0, 0);
1390 ipstat_inc(ips_badoptions);
1391 return (1);
1392 }
1393
1394 /*
1395 * Save incoming source route for use in replies,
1396 * to be picked up later by ip_srcroute if the receiver is interested.
1397 */
1398 void
save_rte(struct mbuf * m,u_char * option,struct in_addr dst)1399 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
1400 {
1401 struct ip_srcrt *isr;
1402 struct m_tag *mtag;
1403 unsigned olen;
1404
1405 olen = option[IPOPT_OLEN];
1406 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes))
1407 return;
1408
1409 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT);
1410 if (mtag == NULL) {
1411 ipstat_inc(ips_idropped);
1412 return;
1413 }
1414 isr = (struct ip_srcrt *)(mtag + 1);
1415
1416 memcpy(isr->isr_hdr, option, olen);
1417 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1418 isr->isr_dst = dst;
1419 m_tag_prepend(m, mtag);
1420 }
1421
1422 /*
1423 * Retrieve incoming source route for use in replies,
1424 * in the same form used by setsockopt.
1425 * The first hop is placed before the options, will be removed later.
1426 */
1427 struct mbuf *
ip_srcroute(struct mbuf * m0)1428 ip_srcroute(struct mbuf *m0)
1429 {
1430 struct in_addr *p, *q;
1431 struct mbuf *m;
1432 struct ip_srcrt *isr;
1433 struct m_tag *mtag;
1434
1435 if (!ip_dosourceroute)
1436 return (NULL);
1437
1438 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE, NULL);
1439 if (mtag == NULL)
1440 return (NULL);
1441 isr = (struct ip_srcrt *)(mtag + 1);
1442
1443 if (isr->isr_nhops == 0)
1444 return (NULL);
1445 m = m_get(M_DONTWAIT, MT_SOOPTS);
1446 if (m == NULL) {
1447 ipstat_inc(ips_idropped);
1448 return (NULL);
1449 }
1450
1451 #define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr))
1452
1453 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */
1454 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ;
1455
1456 /*
1457 * First save first hop for return route
1458 */
1459 p = &(isr->isr_routes[isr->isr_nhops - 1]);
1460 *(mtod(m, struct in_addr *)) = *p--;
1461
1462 /*
1463 * Copy option fields and padding (nop) to mbuf.
1464 */
1465 isr->isr_nop = IPOPT_NOP;
1466 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF;
1467 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &isr->isr_nop,
1468 OPTSIZ);
1469 q = (struct in_addr *)(mtod(m, caddr_t) +
1470 sizeof(struct in_addr) + OPTSIZ);
1471 #undef OPTSIZ
1472 /*
1473 * Record return path as an IP source route,
1474 * reversing the path (pointers are now aligned).
1475 */
1476 while (p >= isr->isr_routes) {
1477 *q++ = *p--;
1478 }
1479 /*
1480 * Last hop goes to final destination.
1481 */
1482 *q = isr->isr_dst;
1483 m_tag_delete(m0, (struct m_tag *)isr);
1484 return (m);
1485 }
1486
1487 /*
1488 * Strip out IP options, at higher level protocol in the kernel.
1489 */
1490 void
ip_stripoptions(struct mbuf * m)1491 ip_stripoptions(struct mbuf *m)
1492 {
1493 int i;
1494 struct ip *ip = mtod(m, struct ip *);
1495 caddr_t opts;
1496 int olen;
1497
1498 olen = (ip->ip_hl<<2) - sizeof (struct ip);
1499 opts = (caddr_t)(ip + 1);
1500 i = m->m_len - (sizeof (struct ip) + olen);
1501 memmove(opts, opts + olen, i);
1502 m->m_len -= olen;
1503 if (m->m_flags & M_PKTHDR)
1504 m->m_pkthdr.len -= olen;
1505 ip->ip_hl = sizeof(struct ip) >> 2;
1506 ip->ip_len = htons(ntohs(ip->ip_len) - olen);
1507 }
1508
1509 const u_char inetctlerrmap[PRC_NCMDS] = {
1510 0, 0, 0, 0,
1511 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1512 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1513 EMSGSIZE, EHOSTUNREACH, 0, 0,
1514 0, 0, 0, 0,
1515 ENOPROTOOPT
1516 };
1517
1518 /*
1519 * Forward a packet. If some error occurs return the sender
1520 * an icmp packet. Note we can't always generate a meaningful
1521 * icmp message because icmp doesn't have a large enough repertoire
1522 * of codes and types.
1523 *
1524 * If not forwarding, just drop the packet. This could be confusing
1525 * if ip_forwarding was zero but some routing protocol was advancing
1526 * us as a gateway to somewhere. However, we must let the routing
1527 * protocol deal with that.
1528 *
1529 * The srcrt parameter indicates whether the packet is being forwarded
1530 * via a source route.
1531 */
1532 void
ip_forward(struct mbuf * m,struct ifnet * ifp,struct route * ro,int flags)1533 ip_forward(struct mbuf *m, struct ifnet *ifp, struct route *ro, int flags)
1534 {
1535 struct mbuf mfake, *mcopy;
1536 struct ip *ip = mtod(m, struct ip *);
1537 struct route iproute;
1538 struct rtentry *rt;
1539 int error = 0, type = 0, code = 0, destmtu = 0, fake = 0, len;
1540 u_int32_t dest;
1541
1542 dest = 0;
1543 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1544 ipstat_inc(ips_cantforward);
1545 m_freem(m);
1546 goto done;
1547 }
1548 if (ip->ip_ttl <= IPTTLDEC) {
1549 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1550 goto done;
1551 }
1552
1553 if (ro == NULL) {
1554 ro = &iproute;
1555 ro->ro_rt = NULL;
1556 }
1557 rt = route_mpath(ro, &ip->ip_dst, &ip->ip_src, m->m_pkthdr.ph_rtableid);
1558 if (rt == NULL) {
1559 ipstat_inc(ips_noroute);
1560 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1561 goto done;
1562 }
1563
1564 /*
1565 * Save at most 68 bytes of the packet in case
1566 * we need to generate an ICMP message to the src.
1567 * The data is saved in the mbuf on the stack that
1568 * acts as a temporary storage not intended to be
1569 * passed down the IP stack or to the mfree.
1570 */
1571 memset(&mfake.m_hdr, 0, sizeof(mfake.m_hdr));
1572 mfake.m_type = m->m_type;
1573 if (m_dup_pkthdr(&mfake, m, M_DONTWAIT) == 0) {
1574 mfake.m_data = mfake.m_pktdat;
1575 len = min(ntohs(ip->ip_len), 68);
1576 m_copydata(m, 0, len, mfake.m_pktdat);
1577 mfake.m_pkthdr.len = mfake.m_len = len;
1578 #if NPF > 0
1579 pf_pkt_addr_changed(&mfake);
1580 #endif /* NPF > 0 */
1581 fake = 1;
1582 }
1583
1584 ip->ip_ttl -= IPTTLDEC;
1585
1586 /*
1587 * If forwarding packet using same interface that it came in on,
1588 * perhaps should send a redirect to sender to shortcut a hop.
1589 * Only send redirect if source is sending directly to us,
1590 * and if packet was not source routed (or has any options).
1591 * Also, don't send redirect if forwarding using a default route
1592 * or a route modified by a redirect.
1593 * Don't send redirect if we advertise destination's arp address
1594 * as ours (proxy arp).
1595 */
1596 if ((rt->rt_ifidx == ifp->if_index) &&
1597 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1598 satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1599 ip_sendredirects && !ISSET(flags, IP_REDIRECT) &&
1600 !arpproxy(satosin(rt_key(rt))->sin_addr, m->m_pkthdr.ph_rtableid)) {
1601 if ((ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_netmask) ==
1602 ifatoia(rt->rt_ifa)->ia_net) {
1603 if (rt->rt_flags & RTF_GATEWAY)
1604 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1605 else
1606 dest = ip->ip_dst.s_addr;
1607 /* Router requirements says to only send host redirects */
1608 type = ICMP_REDIRECT;
1609 code = ICMP_REDIRECT_HOST;
1610 }
1611 }
1612
1613 error = ip_output(m, NULL, ro, flags | IP_FORWARDING, NULL, NULL, 0);
1614 rt = ro->ro_rt;
1615 if (error)
1616 ipstat_inc(ips_cantforward);
1617 else {
1618 ipstat_inc(ips_forward);
1619 if (type)
1620 ipstat_inc(ips_redirectsent);
1621 else
1622 goto done;
1623 }
1624 if (!fake)
1625 goto done;
1626
1627 switch (error) {
1628 case 0: /* forwarded, but need redirect */
1629 /* type, code set above */
1630 break;
1631
1632 case EMSGSIZE:
1633 type = ICMP_UNREACH;
1634 code = ICMP_UNREACH_NEEDFRAG;
1635 if (rt != NULL) {
1636 if (rt->rt_mtu) {
1637 destmtu = rt->rt_mtu;
1638 } else {
1639 struct ifnet *destifp;
1640
1641 destifp = if_get(rt->rt_ifidx);
1642 if (destifp != NULL)
1643 destmtu = destifp->if_mtu;
1644 if_put(destifp);
1645 }
1646 }
1647 ipstat_inc(ips_cantfrag);
1648 if (destmtu == 0)
1649 goto done;
1650 break;
1651
1652 case EACCES:
1653 /*
1654 * pf(4) blocked the packet. There is no need to send an ICMP
1655 * packet back since pf(4) takes care of it.
1656 */
1657 goto done;
1658
1659 case ENOBUFS:
1660 /*
1661 * a router should not generate ICMP_SOURCEQUENCH as
1662 * required in RFC1812 Requirements for IP Version 4 Routers.
1663 * source quench could be a big problem under DoS attacks,
1664 * or the underlying interface is rate-limited.
1665 */
1666 goto done;
1667
1668 case ENETUNREACH: /* shouldn't happen, checked above */
1669 case EHOSTUNREACH:
1670 case ENETDOWN:
1671 case EHOSTDOWN:
1672 default:
1673 type = ICMP_UNREACH;
1674 code = ICMP_UNREACH_HOST;
1675 break;
1676 }
1677 mcopy = m_copym(&mfake, 0, len, M_DONTWAIT);
1678 if (mcopy != NULL)
1679 icmp_error(mcopy, type, code, dest, destmtu);
1680
1681 done:
1682 if (ro == &iproute)
1683 rtfree(ro->ro_rt);
1684 if (fake)
1685 m_tag_delete_chain(&mfake);
1686 }
1687
1688 int
ip_sysctl(int * name,u_int namelen,void * oldp,size_t * oldlenp,void * newp,size_t newlen)1689 ip_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
1690 size_t newlen)
1691 {
1692 #ifdef MROUTING
1693 extern struct mrtstat mrtstat;
1694 #endif
1695 int oldval, error;
1696
1697 /* Almost all sysctl names at this level are terminal. */
1698 if (namelen != 1 && name[0] != IPCTL_IFQUEUE &&
1699 name[0] != IPCTL_ARPQUEUE)
1700 return (ENOTDIR);
1701
1702 switch (name[0]) {
1703 case IPCTL_SOURCEROUTE:
1704 NET_LOCK();
1705 error = sysctl_securelevel_int(oldp, oldlenp, newp, newlen,
1706 &ip_dosourceroute);
1707 NET_UNLOCK();
1708 return (error);
1709 case IPCTL_MTUDISC:
1710 NET_LOCK();
1711 error = sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtudisc);
1712 if (ip_mtudisc == 0)
1713 rt_timer_queue_flush(&ip_mtudisc_timeout_q);
1714 NET_UNLOCK();
1715 return error;
1716 case IPCTL_MTUDISCTIMEOUT:
1717 NET_LOCK();
1718 error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
1719 &ip_mtudisc_timeout, 0, INT_MAX);
1720 rt_timer_queue_change(&ip_mtudisc_timeout_q,
1721 ip_mtudisc_timeout);
1722 NET_UNLOCK();
1723 return (error);
1724 #ifdef IPSEC
1725 case IPCTL_ENCDEBUG:
1726 case IPCTL_IPSEC_STATS:
1727 case IPCTL_IPSEC_EXPIRE_ACQUIRE:
1728 case IPCTL_IPSEC_EMBRYONIC_SA_TIMEOUT:
1729 case IPCTL_IPSEC_REQUIRE_PFS:
1730 case IPCTL_IPSEC_SOFT_ALLOCATIONS:
1731 case IPCTL_IPSEC_ALLOCATIONS:
1732 case IPCTL_IPSEC_SOFT_BYTES:
1733 case IPCTL_IPSEC_BYTES:
1734 case IPCTL_IPSEC_TIMEOUT:
1735 case IPCTL_IPSEC_SOFT_TIMEOUT:
1736 case IPCTL_IPSEC_SOFT_FIRSTUSE:
1737 case IPCTL_IPSEC_FIRSTUSE:
1738 case IPCTL_IPSEC_ENC_ALGORITHM:
1739 case IPCTL_IPSEC_AUTH_ALGORITHM:
1740 case IPCTL_IPSEC_IPCOMP_ALGORITHM:
1741 return (ipsec_sysctl(name, namelen, oldp, oldlenp, newp,
1742 newlen));
1743 #endif
1744 case IPCTL_IFQUEUE:
1745 return (sysctl_niq(name + 1, namelen - 1,
1746 oldp, oldlenp, newp, newlen, &ipintrq));
1747 case IPCTL_ARPQUEUE:
1748 return (sysctl_niq(name + 1, namelen - 1,
1749 oldp, oldlenp, newp, newlen, &arpinq));
1750 case IPCTL_ARPQUEUED:
1751 return (sysctl_rdint(oldp, oldlenp, newp,
1752 atomic_load_int(&la_hold_total)));
1753 case IPCTL_STATS:
1754 return (ip_sysctl_ipstat(oldp, oldlenp, newp));
1755 #ifdef MROUTING
1756 case IPCTL_MRTSTATS:
1757 return (sysctl_rdstruct(oldp, oldlenp, newp,
1758 &mrtstat, sizeof(mrtstat)));
1759 case IPCTL_MRTMFC:
1760 if (newp)
1761 return (EPERM);
1762 NET_LOCK();
1763 error = mrt_sysctl_mfc(oldp, oldlenp);
1764 NET_UNLOCK();
1765 return (error);
1766 case IPCTL_MRTVIF:
1767 if (newp)
1768 return (EPERM);
1769 NET_LOCK();
1770 error = mrt_sysctl_vif(oldp, oldlenp);
1771 NET_UNLOCK();
1772 return (error);
1773 #else
1774 case IPCTL_MRTPROTO:
1775 case IPCTL_MRTSTATS:
1776 case IPCTL_MRTMFC:
1777 case IPCTL_MRTVIF:
1778 return (EOPNOTSUPP);
1779 #endif
1780 case IPCTL_MULTIPATH:
1781 NET_LOCK();
1782 oldval = ipmultipath;
1783 error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
1784 &ipmultipath, 0, 1);
1785 if (oldval != ipmultipath)
1786 atomic_inc_long(&rtgeneration);
1787 NET_UNLOCK();
1788 return (error);
1789 default:
1790 NET_LOCK();
1791 error = sysctl_bounded_arr(ipctl_vars, nitems(ipctl_vars),
1792 name, namelen, oldp, oldlenp, newp, newlen);
1793 NET_UNLOCK();
1794 return (error);
1795 }
1796 /* NOTREACHED */
1797 }
1798
1799 int
ip_sysctl_ipstat(void * oldp,size_t * oldlenp,void * newp)1800 ip_sysctl_ipstat(void *oldp, size_t *oldlenp, void *newp)
1801 {
1802 uint64_t counters[ips_ncounters];
1803 struct ipstat ipstat;
1804 u_long *words = (u_long *)&ipstat;
1805 int i;
1806
1807 CTASSERT(sizeof(ipstat) == (nitems(counters) * sizeof(u_long)));
1808 memset(&ipstat, 0, sizeof ipstat);
1809 counters_read(ipcounters, counters, nitems(counters), NULL);
1810
1811 for (i = 0; i < nitems(counters); i++)
1812 words[i] = (u_long)counters[i];
1813
1814 return (sysctl_rdstruct(oldp, oldlenp, newp, &ipstat, sizeof(ipstat)));
1815 }
1816
1817 void
ip_savecontrol(struct inpcb * inp,struct mbuf ** mp,struct ip * ip,struct mbuf * m)1818 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1819 struct mbuf *m)
1820 {
1821 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1822 struct timeval tv;
1823
1824 m_microtime(m, &tv);
1825 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1826 SCM_TIMESTAMP, SOL_SOCKET);
1827 if (*mp)
1828 mp = &(*mp)->m_next;
1829 }
1830
1831 if (inp->inp_flags & INP_RECVDSTADDR) {
1832 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1833 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1834 if (*mp)
1835 mp = &(*mp)->m_next;
1836 }
1837 #ifdef notyet
1838 /* this code is broken and will probably never be fixed. */
1839 /* options were tossed already */
1840 if (inp->inp_flags & INP_RECVOPTS) {
1841 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1842 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1843 if (*mp)
1844 mp = &(*mp)->m_next;
1845 }
1846 /* ip_srcroute doesn't do what we want here, need to fix */
1847 if (inp->inp_flags & INP_RECVRETOPTS) {
1848 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1849 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1850 if (*mp)
1851 mp = &(*mp)->m_next;
1852 }
1853 #endif
1854 if (inp->inp_flags & INP_RECVIF) {
1855 struct sockaddr_dl sdl;
1856 struct ifnet *ifp;
1857
1858 ifp = if_get(m->m_pkthdr.ph_ifidx);
1859 if (ifp == NULL || ifp->if_sadl == NULL) {
1860 memset(&sdl, 0, sizeof(sdl));
1861 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]);
1862 sdl.sdl_family = AF_LINK;
1863 sdl.sdl_index = ifp != NULL ? ifp->if_index : 0;
1864 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0;
1865 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len,
1866 IP_RECVIF, IPPROTO_IP);
1867 } else {
1868 *mp = sbcreatecontrol((caddr_t) ifp->if_sadl,
1869 ifp->if_sadl->sdl_len, IP_RECVIF, IPPROTO_IP);
1870 }
1871 if (*mp)
1872 mp = &(*mp)->m_next;
1873 if_put(ifp);
1874 }
1875 if (inp->inp_flags & INP_RECVTTL) {
1876 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1877 sizeof(u_int8_t), IP_RECVTTL, IPPROTO_IP);
1878 if (*mp)
1879 mp = &(*mp)->m_next;
1880 }
1881 if (inp->inp_flags & INP_RECVRTABLE) {
1882 u_int rtableid = inp->inp_rtableid;
1883
1884 #if NPF > 0
1885 if (m && m->m_pkthdr.pf.flags & PF_TAG_DIVERTED) {
1886 struct pf_divert *divert;
1887
1888 divert = pf_find_divert(m);
1889 KASSERT(divert != NULL);
1890 rtableid = divert->rdomain;
1891 }
1892 #endif
1893
1894 *mp = sbcreatecontrol((caddr_t) &rtableid,
1895 sizeof(u_int), IP_RECVRTABLE, IPPROTO_IP);
1896 if (*mp)
1897 mp = &(*mp)->m_next;
1898 }
1899 }
1900
1901 void
ip_send_do_dispatch(void * xmq,int flags)1902 ip_send_do_dispatch(void *xmq, int flags)
1903 {
1904 struct mbuf_queue *mq = xmq;
1905 struct mbuf *m;
1906 struct mbuf_list ml;
1907 struct m_tag *mtag;
1908
1909 mq_delist(mq, &ml);
1910 if (ml_empty(&ml))
1911 return;
1912
1913 NET_LOCK_SHARED();
1914 while ((m = ml_dequeue(&ml)) != NULL) {
1915 u_int32_t ipsecflowinfo = 0;
1916
1917 if ((mtag = m_tag_find(m, PACKET_TAG_IPSEC_FLOWINFO, NULL))
1918 != NULL) {
1919 ipsecflowinfo = *(u_int32_t *)(mtag + 1);
1920 m_tag_delete(m, mtag);
1921 }
1922 ip_output(m, NULL, NULL, flags, NULL, NULL, ipsecflowinfo);
1923 }
1924 NET_UNLOCK_SHARED();
1925 }
1926
1927 void
ip_sendraw_dispatch(void * xmq)1928 ip_sendraw_dispatch(void *xmq)
1929 {
1930 ip_send_do_dispatch(xmq, IP_RAWOUTPUT);
1931 }
1932
1933 void
ip_send_dispatch(void * xmq)1934 ip_send_dispatch(void *xmq)
1935 {
1936 ip_send_do_dispatch(xmq, 0);
1937 }
1938
1939 void
ip_send(struct mbuf * m)1940 ip_send(struct mbuf *m)
1941 {
1942 mq_enqueue(&ipsend_mq, m);
1943 task_add(net_tq(0), &ipsend_task);
1944 }
1945
1946 void
ip_send_raw(struct mbuf * m)1947 ip_send_raw(struct mbuf *m)
1948 {
1949 mq_enqueue(&ipsendraw_mq, m);
1950 task_add(net_tq(0), &ipsendraw_task);
1951 }
1952