1 /* $OpenBSD: ip_input.c,v 1.403 2025/01/03 21:27:40 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 /*
87 * Locks used to protect global variables in this file:
88 * I immutable after creation
89 * a atomic operations
90 * N net lock
91 */
92
93 /* values controllable via sysctl */
94 int ip_forwarding = 0; /* [a] */
95 int ipmforwarding = 0;
96 int ipmultipath = 0;
97 int ip_sendredirects = 1; /* [a] */
98 int ip_dosourceroute = 0;
99 int ip_defttl = IPDEFTTL;
100 int ip_mtudisc = 1;
101 int ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
102 int ip_directedbcast = 0; /* [a] */
103
104 /* Protects `ipq' and `ip_frags'. */
105 struct mutex ipq_mutex = MUTEX_INITIALIZER(IPL_SOFTNET);
106
107 /* IP reassembly queue */
108 LIST_HEAD(, ipq) ipq;
109
110 /* Keep track of memory used for reassembly */
111 int ip_maxqueue = 300;
112 int ip_frags = 0;
113
114 const struct sysctl_bounded_args ipctl_vars_unlocked[] = {
115 { IPCTL_FORWARDING, &ip_forwarding, 0, 2 },
116 { IPCTL_SENDREDIRECTS, &ip_sendredirects, 0, 1 },
117 { IPCTL_DIRECTEDBCAST, &ip_directedbcast, 0, 1 },
118 };
119
120 const struct sysctl_bounded_args ipctl_vars[] = {
121 #ifdef MROUTING
122 { IPCTL_MRTPROTO, &ip_mrtproto, SYSCTL_INT_READONLY },
123 #endif
124 { IPCTL_DEFTTL, &ip_defttl, 0, 255 },
125 { IPCTL_IPPORT_FIRSTAUTO, &ipport_firstauto, 0, 65535 },
126 { IPCTL_IPPORT_LASTAUTO, &ipport_lastauto, 0, 65535 },
127 { IPCTL_IPPORT_HIFIRSTAUTO, &ipport_hifirstauto, 0, 65535 },
128 { IPCTL_IPPORT_HILASTAUTO, &ipport_hilastauto, 0, 65535 },
129 { IPCTL_IPPORT_MAXQUEUE, &ip_maxqueue, 0, 10000 },
130 { IPCTL_MFORWARDING, &ipmforwarding, 0, 1 },
131 { IPCTL_ARPTIMEOUT, &arpt_keep, 0, INT_MAX },
132 { IPCTL_ARPDOWN, &arpt_down, 0, INT_MAX },
133 };
134
135 struct niqueue ipintrq = NIQUEUE_INITIALIZER(IPQ_MAXLEN, NETISR_IP);
136
137 struct pool ipqent_pool;
138 struct pool ipq_pool;
139
140 struct cpumem *ipcounters;
141
142 int ip_sysctl_ipstat(void *, size_t *, void *);
143
144 static struct mbuf_queue ipsend_mq;
145 static struct mbuf_queue ipsendraw_mq;
146
147 extern struct niqueue arpinq;
148
149 int ip_ours(struct mbuf **, int *, int, int);
150 int ip_ours_enqueue(struct mbuf **mp, int *offp, int nxt);
151 int ip_dooptions(struct mbuf *, struct ifnet *, int);
152 int in_ouraddr(struct mbuf *, struct ifnet *, struct route *, int);
153
154 int ip_fragcheck(struct mbuf **, int *);
155 struct mbuf * ip_reass(struct ipqent *, struct ipq *);
156 void ip_freef(struct ipq *);
157 void ip_flush(void);
158
159 static void ip_send_dispatch(void *);
160 static void ip_sendraw_dispatch(void *);
161 static struct task ipsend_task = TASK_INITIALIZER(ip_send_dispatch, &ipsend_mq);
162 static struct task ipsendraw_task =
163 TASK_INITIALIZER(ip_sendraw_dispatch, &ipsendraw_mq);
164
165 /*
166 * Used to save the IP options in case a protocol wants to respond
167 * to an incoming packet over the same route if the packet got here
168 * using IP source routing. This allows connection establishment and
169 * maintenance when the remote end is on a network that is not known
170 * to us.
171 */
172 struct ip_srcrt {
173 int isr_nhops; /* number of hops */
174 struct in_addr isr_dst; /* final destination */
175 char isr_nop; /* one NOP to align */
176 char isr_hdr[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN & OFFSET */
177 struct in_addr isr_routes[MAX_IPOPTLEN/sizeof(struct in_addr)];
178 };
179
180 void save_rte(struct mbuf *, u_char *, struct in_addr);
181
182 /*
183 * IP initialization: fill in IP protocol switch table.
184 * All protocols not implemented in kernel go to raw IP protocol handler.
185 */
186 void
ip_init(void)187 ip_init(void)
188 {
189 const struct protosw *pr;
190 int i;
191 const u_int16_t defbaddynamicports_tcp[] = DEFBADDYNAMICPORTS_TCP;
192 const u_int16_t defbaddynamicports_udp[] = DEFBADDYNAMICPORTS_UDP;
193 const u_int16_t defrootonlyports_tcp[] = DEFROOTONLYPORTS_TCP;
194 const u_int16_t defrootonlyports_udp[] = DEFROOTONLYPORTS_UDP;
195
196 ipcounters = counters_alloc(ips_ncounters);
197
198 pool_init(&ipqent_pool, sizeof(struct ipqent), 0,
199 IPL_SOFTNET, 0, "ipqe", NULL);
200 pool_init(&ipq_pool, sizeof(struct ipq), 0,
201 IPL_SOFTNET, 0, "ipq", NULL);
202
203 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
204 if (pr == NULL)
205 panic("ip_init");
206 for (i = 0; i < IPPROTO_MAX; i++)
207 ip_protox[i] = pr - inetsw;
208 for (pr = inetdomain.dom_protosw;
209 pr < inetdomain.dom_protoswNPROTOSW; pr++)
210 if (pr->pr_domain->dom_family == PF_INET &&
211 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW &&
212 pr->pr_protocol < IPPROTO_MAX)
213 ip_protox[pr->pr_protocol] = pr - inetsw;
214 LIST_INIT(&ipq);
215
216 /* Fill in list of ports not to allocate dynamically. */
217 memset(&baddynamicports, 0, sizeof(baddynamicports));
218 for (i = 0; defbaddynamicports_tcp[i] != 0; i++)
219 DP_SET(baddynamicports.tcp, defbaddynamicports_tcp[i]);
220 for (i = 0; defbaddynamicports_udp[i] != 0; i++)
221 DP_SET(baddynamicports.udp, defbaddynamicports_udp[i]);
222
223 /* Fill in list of ports only root can bind to. */
224 memset(&rootonlyports, 0, sizeof(rootonlyports));
225 for (i = 0; defrootonlyports_tcp[i] != 0; i++)
226 DP_SET(rootonlyports.tcp, defrootonlyports_tcp[i]);
227 for (i = 0; defrootonlyports_udp[i] != 0; i++)
228 DP_SET(rootonlyports.udp, defrootonlyports_udp[i]);
229
230 mq_init(&ipsend_mq, 64, IPL_SOFTNET);
231 mq_init(&ipsendraw_mq, 64, IPL_SOFTNET);
232
233 arpinit();
234 #ifdef IPSEC
235 ipsec_init();
236 #endif
237 #ifdef MROUTING
238 rt_timer_queue_init(&ip_mrouterq, MCAST_EXPIRE_FREQUENCY,
239 &mfc_expire_route);
240 #endif
241 }
242
243 /*
244 * Enqueue packet for local delivery. Queuing is used as a boundary
245 * between the network layer (input/forward path) running with
246 * NET_LOCK_SHARED() and the transport layer needing it exclusively.
247 */
248 int
ip_ours(struct mbuf ** mp,int * offp,int nxt,int af)249 ip_ours(struct mbuf **mp, int *offp, int nxt, int af)
250 {
251 nxt = ip_fragcheck(mp, offp);
252 if (nxt == IPPROTO_DONE)
253 return IPPROTO_DONE;
254
255 /* We are already in a IPv4/IPv6 local deliver loop. */
256 if (af != AF_UNSPEC)
257 return nxt;
258
259 nxt = ip_deliver(mp, offp, nxt, AF_INET, 1);
260 if (nxt == IPPROTO_DONE)
261 return IPPROTO_DONE;
262
263 return ip_ours_enqueue(mp, offp, nxt);
264 }
265
266 int
ip_ours_enqueue(struct mbuf ** mp,int * offp,int nxt)267 ip_ours_enqueue(struct mbuf **mp, int *offp, int nxt)
268 {
269 /* save values for later, use after dequeue */
270 if (*offp != sizeof(struct ip)) {
271 struct m_tag *mtag;
272 struct ipoffnxt *ion;
273
274 /* mbuf tags are expensive, but only used for header options */
275 mtag = m_tag_get(PACKET_TAG_IP_OFFNXT, sizeof(*ion),
276 M_NOWAIT);
277 if (mtag == NULL) {
278 ipstat_inc(ips_idropped);
279 m_freemp(mp);
280 return IPPROTO_DONE;
281 }
282 ion = (struct ipoffnxt *)(mtag + 1);
283 ion->ion_off = *offp;
284 ion->ion_nxt = nxt;
285
286 m_tag_prepend(*mp, mtag);
287 }
288
289 niq_enqueue(&ipintrq, *mp);
290 *mp = NULL;
291 return IPPROTO_DONE;
292 }
293
294 /*
295 * Dequeue and process locally delivered packets.
296 * This is called with exclusive NET_LOCK().
297 */
298 void
ipintr(void)299 ipintr(void)
300 {
301 struct mbuf *m;
302
303 while ((m = niq_dequeue(&ipintrq)) != NULL) {
304 struct m_tag *mtag;
305 int off, nxt;
306
307 #ifdef DIAGNOSTIC
308 if ((m->m_flags & M_PKTHDR) == 0)
309 panic("ipintr no HDR");
310 #endif
311 mtag = m_tag_find(m, PACKET_TAG_IP_OFFNXT, NULL);
312 if (mtag != NULL) {
313 struct ipoffnxt *ion;
314
315 ion = (struct ipoffnxt *)(mtag + 1);
316 off = ion->ion_off;
317 nxt = ion->ion_nxt;
318
319 m_tag_delete(m, mtag);
320 } else {
321 struct ip *ip;
322
323 ip = mtod(m, struct ip *);
324 off = ip->ip_hl << 2;
325 nxt = ip->ip_p;
326 }
327
328 nxt = ip_deliver(&m, &off, nxt, AF_INET, 0);
329 KASSERT(nxt == IPPROTO_DONE);
330 }
331 }
332
333 /*
334 * IPv4 input routine.
335 *
336 * Checksum and byte swap header. Process options. Forward or deliver.
337 */
338 void
ipv4_input(struct ifnet * ifp,struct mbuf * m)339 ipv4_input(struct ifnet *ifp, struct mbuf *m)
340 {
341 int off, nxt;
342
343 off = 0;
344 nxt = ip_input_if(&m, &off, IPPROTO_IPV4, AF_UNSPEC, ifp);
345 KASSERT(nxt == IPPROTO_DONE);
346 }
347
348 struct mbuf *
ipv4_check(struct ifnet * ifp,struct mbuf * m)349 ipv4_check(struct ifnet *ifp, struct mbuf *m)
350 {
351 struct ip *ip;
352 int hlen, len;
353
354 if (m->m_len < sizeof(*ip)) {
355 m = m_pullup(m, sizeof(*ip));
356 if (m == NULL) {
357 ipstat_inc(ips_toosmall);
358 return (NULL);
359 }
360 }
361
362 ip = mtod(m, struct ip *);
363 if (ip->ip_v != IPVERSION) {
364 ipstat_inc(ips_badvers);
365 goto bad;
366 }
367
368 hlen = ip->ip_hl << 2;
369 if (hlen < sizeof(*ip)) { /* minimum header length */
370 ipstat_inc(ips_badhlen);
371 goto bad;
372 }
373 if (hlen > m->m_len) {
374 m = m_pullup(m, hlen);
375 if (m == NULL) {
376 ipstat_inc(ips_badhlen);
377 return (NULL);
378 }
379 ip = mtod(m, struct ip *);
380 }
381
382 /* 127/8 must not appear on wire - RFC1122 */
383 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
384 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
385 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
386 ipstat_inc(ips_badaddr);
387 goto bad;
388 }
389 }
390
391 if (!ISSET(m->m_pkthdr.csum_flags, M_IPV4_CSUM_IN_OK)) {
392 if (ISSET(m->m_pkthdr.csum_flags, M_IPV4_CSUM_IN_BAD)) {
393 ipstat_inc(ips_badsum);
394 goto bad;
395 }
396
397 ipstat_inc(ips_inswcsum);
398 if (in_cksum(m, hlen) != 0) {
399 ipstat_inc(ips_badsum);
400 goto bad;
401 }
402
403 SET(m->m_pkthdr.csum_flags, M_IPV4_CSUM_IN_OK);
404 }
405
406 /* Retrieve the packet length. */
407 len = ntohs(ip->ip_len);
408
409 /*
410 * Convert fields to host representation.
411 */
412 if (len < hlen) {
413 ipstat_inc(ips_badlen);
414 goto bad;
415 }
416
417 /*
418 * Check that the amount of data in the buffers
419 * is at least as much as the IP header would have us expect.
420 * Trim mbufs if longer than we expect.
421 * Drop packet if shorter than we expect.
422 */
423 if (m->m_pkthdr.len < len) {
424 ipstat_inc(ips_tooshort);
425 goto bad;
426 }
427 if (m->m_pkthdr.len > len) {
428 if (m->m_len == m->m_pkthdr.len) {
429 m->m_len = len;
430 m->m_pkthdr.len = len;
431 } else
432 m_adj(m, len - m->m_pkthdr.len);
433 }
434
435 return (m);
436 bad:
437 m_freem(m);
438 return (NULL);
439 }
440
441 int
ip_input_if(struct mbuf ** mp,int * offp,int nxt,int af,struct ifnet * ifp)442 ip_input_if(struct mbuf **mp, int *offp, int nxt, int af, struct ifnet *ifp)
443 {
444 struct route ro;
445 struct mbuf *m;
446 struct ip *ip;
447 int hlen;
448 #if NPF > 0
449 struct in_addr odst;
450 #endif
451 int flags = 0;
452
453 KASSERT(*offp == 0);
454
455 ro.ro_rt = NULL;
456 ipstat_inc(ips_total);
457 m = *mp = ipv4_check(ifp, *mp);
458 if (m == NULL)
459 goto bad;
460
461 ip = mtod(m, struct ip *);
462
463 #if NCARP > 0
464 if (carp_lsdrop(ifp, m, AF_INET, &ip->ip_src.s_addr,
465 &ip->ip_dst.s_addr, (ip->ip_p == IPPROTO_ICMP ? 0 : 1)))
466 goto bad;
467 #endif
468
469 #if NPF > 0
470 /*
471 * Packet filter
472 */
473 odst = ip->ip_dst;
474 if (pf_test(AF_INET, PF_IN, ifp, mp) != PF_PASS)
475 goto bad;
476 m = *mp;
477 if (m == NULL)
478 goto bad;
479
480 ip = mtod(m, struct ip *);
481 if (odst.s_addr != ip->ip_dst.s_addr)
482 SET(flags, IP_REDIRECT);
483 #endif
484
485 switch (atomic_load_int(&ip_forwarding)) {
486 case 2:
487 SET(flags, IP_FORWARDING_IPSEC);
488 /* FALLTHROUGH */
489 case 1:
490 SET(flags, IP_FORWARDING);
491 break;
492 }
493 if (atomic_load_int(&ip_directedbcast))
494 SET(flags, IP_ALLOWBROADCAST);
495
496 hlen = ip->ip_hl << 2;
497
498 /*
499 * Process options and, if not destined for us,
500 * ship it on. ip_dooptions returns 1 when an
501 * error was detected (causing an icmp message
502 * to be sent and the original packet to be freed).
503 */
504 if (hlen > sizeof (struct ip) && ip_dooptions(m, ifp, flags)) {
505 m = *mp = NULL;
506 goto bad;
507 }
508
509 if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
510 ip->ip_dst.s_addr == INADDR_ANY) {
511 nxt = ip_ours(mp, offp, nxt, af);
512 goto out;
513 }
514
515 switch(in_ouraddr(m, ifp, &ro, flags)) {
516 case 2:
517 goto bad;
518 case 1:
519 nxt = ip_ours(mp, offp, nxt, af);
520 goto out;
521 }
522
523 if (IN_MULTICAST(ip->ip_dst.s_addr)) {
524 /*
525 * Make sure M_MCAST is set. It should theoretically
526 * already be there, but let's play safe because upper
527 * layers check for this flag.
528 */
529 m->m_flags |= M_MCAST;
530
531 #ifdef MROUTING
532 if (ipmforwarding && ip_mrouter[ifp->if_rdomain]) {
533 int error;
534
535 if (m->m_flags & M_EXT) {
536 if ((m = *mp = m_pullup(m, hlen)) == NULL) {
537 ipstat_inc(ips_toosmall);
538 goto bad;
539 }
540 ip = mtod(m, struct ip *);
541 }
542 /*
543 * If we are acting as a multicast router, all
544 * incoming multicast packets are passed to the
545 * kernel-level multicast forwarding function.
546 * The packet is returned (relatively) intact; if
547 * ip_mforward() returns a non-zero value, the packet
548 * must be discarded, else it may be accepted below.
549 *
550 * (The IP ident field is put in the same byte order
551 * as expected when ip_mforward() is called from
552 * ip_output().)
553 */
554 KERNEL_LOCK();
555 error = ip_mforward(m, ifp, flags);
556 KERNEL_UNLOCK();
557 if (error) {
558 ipstat_inc(ips_cantforward);
559 goto bad;
560 }
561
562 /*
563 * The process-level routing daemon needs to receive
564 * all multicast IGMP packets, whether or not this
565 * host belongs to their destination groups.
566 */
567 if (ip->ip_p == IPPROTO_IGMP) {
568 nxt = ip_ours(mp, offp, nxt, af);
569 goto out;
570 }
571 ipstat_inc(ips_forward);
572 }
573 #endif
574 /*
575 * See if we belong to the destination multicast group on the
576 * arrival interface.
577 */
578 if (!in_hasmulti(&ip->ip_dst, ifp)) {
579 ipstat_inc(ips_notmember);
580 if (!IN_LOCAL_GROUP(ip->ip_dst.s_addr))
581 ipstat_inc(ips_cantforward);
582 goto bad;
583 }
584 nxt = ip_ours(mp, offp, nxt, af);
585 goto out;
586 }
587
588 #if NCARP > 0
589 if (ip->ip_p == IPPROTO_ICMP &&
590 carp_lsdrop(ifp, m, AF_INET, &ip->ip_src.s_addr,
591 &ip->ip_dst.s_addr, 1))
592 goto bad;
593 #endif
594 /*
595 * Not for us; forward if possible and desirable.
596 */
597 if (!ISSET(flags, IP_FORWARDING)) {
598 ipstat_inc(ips_cantforward);
599 goto bad;
600 }
601 #ifdef IPSEC
602 if (ipsec_in_use) {
603 int rv;
604
605 rv = ipsec_forward_check(m, hlen, AF_INET);
606 if (rv != 0) {
607 ipstat_inc(ips_cantforward);
608 goto bad;
609 }
610 /*
611 * Fall through, forward packet. Outbound IPsec policy
612 * checking will occur in ip_output().
613 */
614 }
615 #endif /* IPSEC */
616
617 ip_forward(m, ifp, &ro, flags);
618 *mp = NULL;
619 rtfree(ro.ro_rt);
620 return IPPROTO_DONE;
621 bad:
622 nxt = IPPROTO_DONE;
623 m_freemp(mp);
624 out:
625 rtfree(ro.ro_rt);
626 return nxt;
627 }
628
629 int
ip_fragcheck(struct mbuf ** mp,int * offp)630 ip_fragcheck(struct mbuf **mp, int *offp)
631 {
632 struct ip *ip;
633 struct ipq *fp;
634 struct ipqent *ipqe;
635 int hlen;
636 uint16_t mff;
637
638 ip = mtod(*mp, struct ip *);
639 hlen = ip->ip_hl << 2;
640
641 /*
642 * If offset or more fragments are set, must reassemble.
643 * Otherwise, nothing need be done.
644 * (We could look in the reassembly queue to see
645 * if the packet was previously fragmented,
646 * but it's not worth the time; just let them time out.)
647 */
648 if (ISSET(ip->ip_off, htons(IP_OFFMASK | IP_MF))) {
649 if ((*mp)->m_flags & M_EXT) { /* XXX */
650 if ((*mp = m_pullup(*mp, hlen)) == NULL) {
651 ipstat_inc(ips_toosmall);
652 return IPPROTO_DONE;
653 }
654 ip = mtod(*mp, struct ip *);
655 }
656
657 /*
658 * Adjust ip_len to not reflect header,
659 * set ipqe_mff if more fragments are expected,
660 * convert offset of this to bytes.
661 */
662 ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
663 mff = ISSET(ip->ip_off, htons(IP_MF));
664 if (mff) {
665 /*
666 * Make sure that fragments have a data length
667 * that's a non-zero multiple of 8 bytes.
668 */
669 if (ntohs(ip->ip_len) == 0 ||
670 (ntohs(ip->ip_len) & 0x7) != 0) {
671 ipstat_inc(ips_badfrags);
672 m_freemp(mp);
673 return IPPROTO_DONE;
674 }
675 }
676 ip->ip_off = htons(ntohs(ip->ip_off) << 3);
677
678 mtx_enter(&ipq_mutex);
679
680 /*
681 * Look for queue of fragments
682 * of this datagram.
683 */
684 LIST_FOREACH(fp, &ipq, ipq_q) {
685 if (ip->ip_id == fp->ipq_id &&
686 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
687 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
688 ip->ip_p == fp->ipq_p)
689 break;
690 }
691
692 /*
693 * If datagram marked as having more fragments
694 * or if this is not the first fragment,
695 * attempt reassembly; if it succeeds, proceed.
696 */
697 if (mff || ip->ip_off) {
698 ipstat_inc(ips_fragments);
699 if (ip_frags + 1 > ip_maxqueue) {
700 ip_flush();
701 ipstat_inc(ips_rcvmemdrop);
702 goto bad;
703 }
704
705 ipqe = pool_get(&ipqent_pool, PR_NOWAIT);
706 if (ipqe == NULL) {
707 ipstat_inc(ips_rcvmemdrop);
708 goto bad;
709 }
710 ip_frags++;
711 ipqe->ipqe_mff = mff;
712 ipqe->ipqe_m = *mp;
713 ipqe->ipqe_ip = ip;
714 *mp = ip_reass(ipqe, fp);
715 if (*mp == NULL)
716 goto bad;
717 ipstat_inc(ips_reassembled);
718 ip = mtod(*mp, struct ip *);
719 hlen = ip->ip_hl << 2;
720 ip->ip_len = htons(ntohs(ip->ip_len) + hlen);
721 } else {
722 if (fp != NULL)
723 ip_freef(fp);
724 }
725
726 mtx_leave(&ipq_mutex);
727 }
728
729 *offp = hlen;
730 return ip->ip_p;
731
732 bad:
733 mtx_leave(&ipq_mutex);
734 m_freemp(mp);
735 return IPPROTO_DONE;
736 }
737
738 #ifndef INET6
739 #define IPSTAT_INC(name) ipstat_inc(ips_##name)
740 #else
741 #define IPSTAT_INC(name) (af == AF_INET ? \
742 ipstat_inc(ips_##name) : ip6stat_inc(ip6s_##name))
743 #endif
744
745 int
ip_deliver(struct mbuf ** mp,int * offp,int nxt,int af,int shared)746 ip_deliver(struct mbuf **mp, int *offp, int nxt, int af, int shared)
747 {
748 #ifdef INET6
749 int nest = 0;
750 #endif
751
752 /*
753 * Tell launch routine the next header
754 */
755 IPSTAT_INC(delivered);
756
757 while (nxt != IPPROTO_DONE) {
758 const struct protosw *psw;
759 int naf;
760
761 switch (af) {
762 case AF_INET:
763 psw = &inetsw[ip_protox[nxt]];
764 break;
765 #ifdef INET6
766 case AF_INET6:
767 psw = &inet6sw[ip6_protox[nxt]];
768 break;
769 #endif
770 }
771 if (shared && !ISSET(psw->pr_flags, PR_MPINPUT)) {
772 /* delivery not finished, decrement counter, queue */
773 switch (af) {
774 case AF_INET:
775 counters_dec(ipcounters, ips_delivered);
776 return ip_ours_enqueue(mp, offp, nxt);
777 #ifdef INET6
778 case AF_INET6:
779 counters_dec(ip6counters, ip6s_delivered);
780 return ip6_ours_enqueue(mp, offp, nxt);
781 #endif
782 }
783 break;
784 }
785
786 #ifdef INET6
787 if (af == AF_INET6 &&
788 ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
789 ip6stat_inc(ip6s_toomanyhdr);
790 goto bad;
791 }
792 #endif
793
794 /*
795 * protection against faulty packet - there should be
796 * more sanity checks in header chain processing.
797 */
798 if ((*mp)->m_pkthdr.len < *offp) {
799 IPSTAT_INC(tooshort);
800 goto bad;
801 }
802
803 #ifdef IPSEC
804 if (ipsec_in_use) {
805 if (ipsec_local_check(*mp, *offp, nxt, af) != 0) {
806 IPSTAT_INC(cantforward);
807 goto bad;
808 }
809 }
810 /* Otherwise, just fall through and deliver the packet */
811 #endif
812
813 switch (nxt) {
814 case IPPROTO_IPV4:
815 naf = AF_INET;
816 ipstat_inc(ips_delivered);
817 break;
818 #ifdef INET6
819 case IPPROTO_IPV6:
820 naf = AF_INET6;
821 ip6stat_inc(ip6s_delivered);
822 break;
823 #endif
824 default:
825 naf = af;
826 break;
827 }
828 nxt = (*psw->pr_input)(mp, offp, nxt, af);
829 af = naf;
830 }
831 return nxt;
832 bad:
833 m_freemp(mp);
834 return IPPROTO_DONE;
835 }
836 #undef IPSTAT_INC
837
838 int
in_ouraddr(struct mbuf * m,struct ifnet * ifp,struct route * ro,int flags)839 in_ouraddr(struct mbuf *m, struct ifnet *ifp, struct route *ro, int flags)
840 {
841 struct rtentry *rt;
842 struct ip *ip;
843 int match = 0;
844
845 #if NPF > 0
846 switch (pf_ouraddr(m)) {
847 case 0:
848 return (0);
849 case 1:
850 return (1);
851 default:
852 /* pf does not know it */
853 break;
854 }
855 #endif
856
857 ip = mtod(m, struct ip *);
858
859 rt = route_mpath(ro, &ip->ip_dst, &ip->ip_src, m->m_pkthdr.ph_rtableid);
860 if (rt != NULL) {
861 if (ISSET(rt->rt_flags, RTF_LOCAL))
862 match = 1;
863
864 /*
865 * If directedbcast is enabled we only consider it local
866 * if it is received on the interface with that address.
867 */
868 if (ISSET(rt->rt_flags, RTF_BROADCAST) &&
869 (!ISSET(flags, IP_ALLOWBROADCAST) ||
870 rt->rt_ifidx == ifp->if_index)) {
871 match = 1;
872
873 /* Make sure M_BCAST is set */
874 m->m_flags |= M_BCAST;
875 }
876 }
877
878 if (!match) {
879 struct ifaddr *ifa;
880
881 /*
882 * No local address or broadcast address found, so check for
883 * ancient classful broadcast addresses.
884 * It must have been broadcast on the link layer, and for an
885 * address on the interface it was received on.
886 */
887 if (!ISSET(m->m_flags, M_BCAST) ||
888 !IN_CLASSFULBROADCAST(ip->ip_dst.s_addr, ip->ip_dst.s_addr))
889 return (0);
890
891 if (ifp->if_rdomain != rtable_l2(m->m_pkthdr.ph_rtableid))
892 return (0);
893 /*
894 * The check in the loop assumes you only rx a packet on an UP
895 * interface, and that M_BCAST will only be set on a BROADCAST
896 * interface.
897 */
898 NET_ASSERT_LOCKED();
899 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
900 if (ifa->ifa_addr->sa_family != AF_INET)
901 continue;
902
903 if (IN_CLASSFULBROADCAST(ip->ip_dst.s_addr,
904 ifatoia(ifa)->ia_addr.sin_addr.s_addr)) {
905 match = 1;
906 break;
907 }
908 }
909 } else if (!ISSET(flags, IP_FORWARDING) &&
910 rt->rt_ifidx != ifp->if_index &&
911 !((ifp->if_flags & IFF_LOOPBACK) || (ifp->if_type == IFT_ENC) ||
912 (m->m_pkthdr.pf.flags & PF_TAG_TRANSLATE_LOCALHOST))) {
913 /* received on wrong interface. */
914 #if NCARP > 0
915 struct ifnet *out_if;
916
917 /*
918 * Virtual IPs on carp interfaces need to be checked also
919 * against the parent interface and other carp interfaces
920 * sharing the same parent.
921 */
922 out_if = if_get(rt->rt_ifidx);
923 if (!(out_if && carp_strict_addr_chk(out_if, ifp))) {
924 ipstat_inc(ips_wrongif);
925 match = 2;
926 }
927 if_put(out_if);
928 #else
929 ipstat_inc(ips_wrongif);
930 match = 2;
931 #endif
932 }
933
934 return (match);
935 }
936
937 /*
938 * Take incoming datagram fragment and try to
939 * reassemble it into whole datagram. If a chain for
940 * reassembly of this datagram already exists, then it
941 * is given as fp; otherwise have to make a chain.
942 */
943 struct mbuf *
ip_reass(struct ipqent * ipqe,struct ipq * fp)944 ip_reass(struct ipqent *ipqe, struct ipq *fp)
945 {
946 struct mbuf *m = ipqe->ipqe_m;
947 struct ipqent *nq, *p, *q;
948 struct ip *ip;
949 struct mbuf *t;
950 int hlen = ipqe->ipqe_ip->ip_hl << 2;
951 int i, next;
952 u_int8_t ecn, ecn0;
953
954 MUTEX_ASSERT_LOCKED(&ipq_mutex);
955
956 /*
957 * Presence of header sizes in mbufs
958 * would confuse code below.
959 */
960 m->m_data += hlen;
961 m->m_len -= hlen;
962
963 /*
964 * If first fragment to arrive, create a reassembly queue.
965 */
966 if (fp == NULL) {
967 fp = pool_get(&ipq_pool, PR_NOWAIT);
968 if (fp == NULL)
969 goto dropfrag;
970 LIST_INSERT_HEAD(&ipq, fp, ipq_q);
971 fp->ipq_ttl = IPFRAGTTL;
972 fp->ipq_p = ipqe->ipqe_ip->ip_p;
973 fp->ipq_id = ipqe->ipqe_ip->ip_id;
974 LIST_INIT(&fp->ipq_fragq);
975 fp->ipq_src = ipqe->ipqe_ip->ip_src;
976 fp->ipq_dst = ipqe->ipqe_ip->ip_dst;
977 p = NULL;
978 goto insert;
979 }
980
981 /*
982 * Handle ECN by comparing this segment with the first one;
983 * if CE is set, do not lose CE.
984 * drop if CE and not-ECT are mixed for the same packet.
985 */
986 ecn = ipqe->ipqe_ip->ip_tos & IPTOS_ECN_MASK;
987 ecn0 = LIST_FIRST(&fp->ipq_fragq)->ipqe_ip->ip_tos & IPTOS_ECN_MASK;
988 if (ecn == IPTOS_ECN_CE) {
989 if (ecn0 == IPTOS_ECN_NOTECT)
990 goto dropfrag;
991 if (ecn0 != IPTOS_ECN_CE)
992 LIST_FIRST(&fp->ipq_fragq)->ipqe_ip->ip_tos |=
993 IPTOS_ECN_CE;
994 }
995 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
996 goto dropfrag;
997
998 /*
999 * Find a segment which begins after this one does.
1000 */
1001 for (p = NULL, q = LIST_FIRST(&fp->ipq_fragq); q != NULL;
1002 p = q, q = LIST_NEXT(q, ipqe_q))
1003 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off))
1004 break;
1005
1006 /*
1007 * If there is a preceding segment, it may provide some of
1008 * our data already. If so, drop the data from the incoming
1009 * segment. If it provides all of our data, drop us.
1010 */
1011 if (p != NULL) {
1012 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) -
1013 ntohs(ipqe->ipqe_ip->ip_off);
1014 if (i > 0) {
1015 if (i >= ntohs(ipqe->ipqe_ip->ip_len))
1016 goto dropfrag;
1017 m_adj(ipqe->ipqe_m, i);
1018 ipqe->ipqe_ip->ip_off =
1019 htons(ntohs(ipqe->ipqe_ip->ip_off) + i);
1020 ipqe->ipqe_ip->ip_len =
1021 htons(ntohs(ipqe->ipqe_ip->ip_len) - i);
1022 }
1023 }
1024
1025 /*
1026 * While we overlap succeeding segments trim them or,
1027 * if they are completely covered, dequeue them.
1028 */
1029 for (; q != NULL &&
1030 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) >
1031 ntohs(q->ipqe_ip->ip_off); q = nq) {
1032 i = (ntohs(ipqe->ipqe_ip->ip_off) +
1033 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off);
1034 if (i < ntohs(q->ipqe_ip->ip_len)) {
1035 q->ipqe_ip->ip_len =
1036 htons(ntohs(q->ipqe_ip->ip_len) - i);
1037 q->ipqe_ip->ip_off =
1038 htons(ntohs(q->ipqe_ip->ip_off) + i);
1039 m_adj(q->ipqe_m, i);
1040 break;
1041 }
1042 nq = LIST_NEXT(q, ipqe_q);
1043 m_freem(q->ipqe_m);
1044 LIST_REMOVE(q, ipqe_q);
1045 pool_put(&ipqent_pool, q);
1046 ip_frags--;
1047 }
1048
1049 insert:
1050 /*
1051 * Stick new segment in its place;
1052 * check for complete reassembly.
1053 */
1054 if (p == NULL) {
1055 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q);
1056 } else {
1057 LIST_INSERT_AFTER(p, ipqe, ipqe_q);
1058 }
1059 next = 0;
1060 for (p = NULL, q = LIST_FIRST(&fp->ipq_fragq); q != NULL;
1061 p = q, q = LIST_NEXT(q, ipqe_q)) {
1062 if (ntohs(q->ipqe_ip->ip_off) != next)
1063 return (0);
1064 next += ntohs(q->ipqe_ip->ip_len);
1065 }
1066 if (p->ipqe_mff)
1067 return (0);
1068
1069 /*
1070 * Reassembly is complete. Check for a bogus message size and
1071 * concatenate fragments.
1072 */
1073 q = LIST_FIRST(&fp->ipq_fragq);
1074 ip = q->ipqe_ip;
1075 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) {
1076 ipstat_inc(ips_toolong);
1077 ip_freef(fp);
1078 return (0);
1079 }
1080 m = q->ipqe_m;
1081 t = m->m_next;
1082 m->m_next = 0;
1083 m_cat(m, t);
1084 nq = LIST_NEXT(q, ipqe_q);
1085 pool_put(&ipqent_pool, q);
1086 ip_frags--;
1087 for (q = nq; q != NULL; q = nq) {
1088 t = q->ipqe_m;
1089 nq = LIST_NEXT(q, ipqe_q);
1090 pool_put(&ipqent_pool, q);
1091 ip_frags--;
1092 m_removehdr(t);
1093 m_cat(m, t);
1094 }
1095
1096 /*
1097 * Create header for new ip packet by
1098 * modifying header of first packet;
1099 * dequeue and discard fragment reassembly header.
1100 * Make header visible.
1101 */
1102 ip->ip_len = htons(next);
1103 ip->ip_src = fp->ipq_src;
1104 ip->ip_dst = fp->ipq_dst;
1105 LIST_REMOVE(fp, ipq_q);
1106 pool_put(&ipq_pool, fp);
1107 m->m_len += (ip->ip_hl << 2);
1108 m->m_data -= (ip->ip_hl << 2);
1109 m_calchdrlen(m);
1110 return (m);
1111
1112 dropfrag:
1113 ipstat_inc(ips_fragdropped);
1114 m_freem(m);
1115 pool_put(&ipqent_pool, ipqe);
1116 ip_frags--;
1117 return (NULL);
1118 }
1119
1120 /*
1121 * Free a fragment reassembly header and all
1122 * associated datagrams.
1123 */
1124 void
ip_freef(struct ipq * fp)1125 ip_freef(struct ipq *fp)
1126 {
1127 struct ipqent *q;
1128
1129 MUTEX_ASSERT_LOCKED(&ipq_mutex);
1130
1131 while ((q = LIST_FIRST(&fp->ipq_fragq)) != NULL) {
1132 LIST_REMOVE(q, ipqe_q);
1133 m_freem(q->ipqe_m);
1134 pool_put(&ipqent_pool, q);
1135 ip_frags--;
1136 }
1137 LIST_REMOVE(fp, ipq_q);
1138 pool_put(&ipq_pool, fp);
1139 }
1140
1141 /*
1142 * IP timer processing;
1143 * if a timer expires on a reassembly queue, discard it.
1144 */
1145 void
ip_slowtimo(void)1146 ip_slowtimo(void)
1147 {
1148 struct ipq *fp, *nfp;
1149
1150 mtx_enter(&ipq_mutex);
1151 LIST_FOREACH_SAFE(fp, &ipq, ipq_q, nfp) {
1152 if (--fp->ipq_ttl == 0) {
1153 ipstat_inc(ips_fragtimeout);
1154 ip_freef(fp);
1155 }
1156 }
1157 mtx_leave(&ipq_mutex);
1158 }
1159
1160 /*
1161 * Flush a bunch of datagram fragments, till we are down to 75%.
1162 */
1163 void
ip_flush(void)1164 ip_flush(void)
1165 {
1166 int max = 50;
1167
1168 MUTEX_ASSERT_LOCKED(&ipq_mutex);
1169
1170 while (!LIST_EMPTY(&ipq) && ip_frags > ip_maxqueue * 3 / 4 && --max) {
1171 ipstat_inc(ips_fragdropped);
1172 ip_freef(LIST_FIRST(&ipq));
1173 }
1174 }
1175
1176 /*
1177 * Do option processing on a datagram,
1178 * possibly discarding it if bad options are encountered,
1179 * or forwarding it if source-routed.
1180 * Returns 1 if packet has been forwarded/freed,
1181 * 0 if the packet should be processed further.
1182 */
1183 int
ip_dooptions(struct mbuf * m,struct ifnet * ifp,int flags)1184 ip_dooptions(struct mbuf *m, struct ifnet *ifp, int flags)
1185 {
1186 struct ip *ip = mtod(m, struct ip *);
1187 unsigned int rtableid = m->m_pkthdr.ph_rtableid;
1188 struct rtentry *rt;
1189 struct sockaddr_in ipaddr;
1190 u_char *cp;
1191 struct ip_timestamp ipt;
1192 struct in_ifaddr *ia;
1193 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1194 struct in_addr sin, dst;
1195 u_int32_t ntime;
1196
1197 dst = ip->ip_dst;
1198 cp = (u_char *)(ip + 1);
1199 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1200
1201 KERNEL_LOCK();
1202 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1203 opt = cp[IPOPT_OPTVAL];
1204 if (opt == IPOPT_EOL)
1205 break;
1206 if (opt == IPOPT_NOP)
1207 optlen = 1;
1208 else {
1209 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1210 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1211 goto bad;
1212 }
1213 optlen = cp[IPOPT_OLEN];
1214 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1215 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1216 goto bad;
1217 }
1218 }
1219
1220 switch (opt) {
1221
1222 default:
1223 break;
1224
1225 /*
1226 * Source routing with record.
1227 * Find interface with current destination address.
1228 * If none on this machine then drop if strictly routed,
1229 * or do nothing if loosely routed.
1230 * Record interface address and bring up next address
1231 * component. If strictly routed make sure next
1232 * address is on directly accessible net.
1233 */
1234 case IPOPT_LSRR:
1235 case IPOPT_SSRR:
1236 if (!ip_dosourceroute) {
1237 type = ICMP_UNREACH;
1238 code = ICMP_UNREACH_SRCFAIL;
1239 goto bad;
1240 }
1241 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1242 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1243 goto bad;
1244 }
1245 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1246 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1247 goto bad;
1248 }
1249 memset(&ipaddr, 0, sizeof(ipaddr));
1250 ipaddr.sin_family = AF_INET;
1251 ipaddr.sin_len = sizeof(ipaddr);
1252 ipaddr.sin_addr = ip->ip_dst;
1253 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr),
1254 m->m_pkthdr.ph_rtableid));
1255 if (ia == NULL) {
1256 if (opt == IPOPT_SSRR) {
1257 type = ICMP_UNREACH;
1258 code = ICMP_UNREACH_SRCFAIL;
1259 goto bad;
1260 }
1261 /*
1262 * Loose routing, and not at next destination
1263 * yet; nothing to do except forward.
1264 */
1265 break;
1266 }
1267 off--; /* 0 origin */
1268 if ((off + sizeof(struct in_addr)) > optlen) {
1269 /*
1270 * End of source route. Should be for us.
1271 */
1272 save_rte(m, cp, ip->ip_src);
1273 break;
1274 }
1275
1276 /*
1277 * locate outgoing interface
1278 */
1279 memset(&ipaddr, 0, sizeof(ipaddr));
1280 ipaddr.sin_family = AF_INET;
1281 ipaddr.sin_len = sizeof(ipaddr);
1282 memcpy(&ipaddr.sin_addr, cp + off,
1283 sizeof(ipaddr.sin_addr));
1284 /* keep packet in the virtual instance */
1285 rt = rtalloc(sintosa(&ipaddr), RT_RESOLVE, rtableid);
1286 if (!rtisvalid(rt) || ((opt == IPOPT_SSRR) &&
1287 ISSET(rt->rt_flags, RTF_GATEWAY))) {
1288 type = ICMP_UNREACH;
1289 code = ICMP_UNREACH_SRCFAIL;
1290 rtfree(rt);
1291 goto bad;
1292 }
1293 ia = ifatoia(rt->rt_ifa);
1294 memcpy(cp + off, &ia->ia_addr.sin_addr,
1295 sizeof(struct in_addr));
1296 rtfree(rt);
1297 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1298 ip->ip_dst = ipaddr.sin_addr;
1299 /*
1300 * Let ip_intr's mcast routing check handle mcast pkts
1301 */
1302 forward = !IN_MULTICAST(ip->ip_dst.s_addr);
1303 break;
1304
1305 case IPOPT_RR:
1306 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1307 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1308 goto bad;
1309 }
1310 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1311 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1312 goto bad;
1313 }
1314
1315 /*
1316 * If no space remains, ignore.
1317 */
1318 off--; /* 0 origin */
1319 if ((off + sizeof(struct in_addr)) > optlen)
1320 break;
1321 memset(&ipaddr, 0, sizeof(ipaddr));
1322 ipaddr.sin_family = AF_INET;
1323 ipaddr.sin_len = sizeof(ipaddr);
1324 ipaddr.sin_addr = ip->ip_dst;
1325 /*
1326 * locate outgoing interface; if we're the destination,
1327 * use the incoming interface (should be same).
1328 * Again keep the packet inside the virtual instance.
1329 */
1330 rt = rtalloc(sintosa(&ipaddr), RT_RESOLVE, rtableid);
1331 if (!rtisvalid(rt)) {
1332 type = ICMP_UNREACH;
1333 code = ICMP_UNREACH_HOST;
1334 rtfree(rt);
1335 goto bad;
1336 }
1337 ia = ifatoia(rt->rt_ifa);
1338 memcpy(cp + off, &ia->ia_addr.sin_addr,
1339 sizeof(struct in_addr));
1340 rtfree(rt);
1341 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1342 break;
1343
1344 case IPOPT_TS:
1345 code = cp - (u_char *)ip;
1346 if (optlen < sizeof(struct ip_timestamp))
1347 goto bad;
1348 memcpy(&ipt, cp, sizeof(struct ip_timestamp));
1349 if (ipt.ipt_ptr < 5 || ipt.ipt_len < 5)
1350 goto bad;
1351 if (ipt.ipt_ptr - 1 + sizeof(u_int32_t) > ipt.ipt_len) {
1352 if (++ipt.ipt_oflw == 0)
1353 goto bad;
1354 break;
1355 }
1356 memcpy(&sin, cp + ipt.ipt_ptr - 1, sizeof sin);
1357 switch (ipt.ipt_flg) {
1358
1359 case IPOPT_TS_TSONLY:
1360 break;
1361
1362 case IPOPT_TS_TSANDADDR:
1363 if (ipt.ipt_ptr - 1 + sizeof(u_int32_t) +
1364 sizeof(struct in_addr) > ipt.ipt_len)
1365 goto bad;
1366 memset(&ipaddr, 0, sizeof(ipaddr));
1367 ipaddr.sin_family = AF_INET;
1368 ipaddr.sin_len = sizeof(ipaddr);
1369 ipaddr.sin_addr = dst;
1370 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr),
1371 ifp));
1372 if (ia == NULL)
1373 continue;
1374 memcpy(&sin, &ia->ia_addr.sin_addr,
1375 sizeof(struct in_addr));
1376 ipt.ipt_ptr += sizeof(struct in_addr);
1377 break;
1378
1379 case IPOPT_TS_PRESPEC:
1380 if (ipt.ipt_ptr - 1 + sizeof(u_int32_t) +
1381 sizeof(struct in_addr) > ipt.ipt_len)
1382 goto bad;
1383 memset(&ipaddr, 0, sizeof(ipaddr));
1384 ipaddr.sin_family = AF_INET;
1385 ipaddr.sin_len = sizeof(ipaddr);
1386 ipaddr.sin_addr = sin;
1387 if (ifa_ifwithaddr(sintosa(&ipaddr),
1388 m->m_pkthdr.ph_rtableid) == NULL)
1389 continue;
1390 ipt.ipt_ptr += sizeof(struct in_addr);
1391 break;
1392
1393 default:
1394 /* XXX can't take &ipt->ipt_flg */
1395 code = (u_char *)&ipt.ipt_ptr -
1396 (u_char *)ip + 1;
1397 goto bad;
1398 }
1399 ntime = iptime();
1400 memcpy(cp + ipt.ipt_ptr - 1, &ntime, sizeof(u_int32_t));
1401 ipt.ipt_ptr += sizeof(u_int32_t);
1402 }
1403 }
1404 KERNEL_UNLOCK();
1405 if (forward && ISSET(flags, IP_FORWARDING)) {
1406 ip_forward(m, ifp, NULL, flags | IP_REDIRECT);
1407 return (1);
1408 }
1409 return (0);
1410 bad:
1411 KERNEL_UNLOCK();
1412 icmp_error(m, type, code, 0, 0);
1413 ipstat_inc(ips_badoptions);
1414 return (1);
1415 }
1416
1417 /*
1418 * Save incoming source route for use in replies,
1419 * to be picked up later by ip_srcroute if the receiver is interested.
1420 */
1421 void
save_rte(struct mbuf * m,u_char * option,struct in_addr dst)1422 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
1423 {
1424 struct ip_srcrt *isr;
1425 struct m_tag *mtag;
1426 unsigned olen;
1427
1428 olen = option[IPOPT_OLEN];
1429 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes))
1430 return;
1431
1432 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT);
1433 if (mtag == NULL) {
1434 ipstat_inc(ips_idropped);
1435 return;
1436 }
1437 isr = (struct ip_srcrt *)(mtag + 1);
1438
1439 memcpy(isr->isr_hdr, option, olen);
1440 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1441 isr->isr_dst = dst;
1442 m_tag_prepend(m, mtag);
1443 }
1444
1445 /*
1446 * Retrieve incoming source route for use in replies,
1447 * in the same form used by setsockopt.
1448 * The first hop is placed before the options, will be removed later.
1449 */
1450 struct mbuf *
ip_srcroute(struct mbuf * m0)1451 ip_srcroute(struct mbuf *m0)
1452 {
1453 struct in_addr *p, *q;
1454 struct mbuf *m;
1455 struct ip_srcrt *isr;
1456 struct m_tag *mtag;
1457
1458 if (!ip_dosourceroute)
1459 return (NULL);
1460
1461 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE, NULL);
1462 if (mtag == NULL)
1463 return (NULL);
1464 isr = (struct ip_srcrt *)(mtag + 1);
1465
1466 if (isr->isr_nhops == 0)
1467 return (NULL);
1468 m = m_get(M_DONTWAIT, MT_SOOPTS);
1469 if (m == NULL) {
1470 ipstat_inc(ips_idropped);
1471 return (NULL);
1472 }
1473
1474 #define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr))
1475
1476 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */
1477 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ;
1478
1479 /*
1480 * First save first hop for return route
1481 */
1482 p = &(isr->isr_routes[isr->isr_nhops - 1]);
1483 *(mtod(m, struct in_addr *)) = *p--;
1484
1485 /*
1486 * Copy option fields and padding (nop) to mbuf.
1487 */
1488 isr->isr_nop = IPOPT_NOP;
1489 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF;
1490 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &isr->isr_nop,
1491 OPTSIZ);
1492 q = (struct in_addr *)(mtod(m, caddr_t) +
1493 sizeof(struct in_addr) + OPTSIZ);
1494 #undef OPTSIZ
1495 /*
1496 * Record return path as an IP source route,
1497 * reversing the path (pointers are now aligned).
1498 */
1499 while (p >= isr->isr_routes) {
1500 *q++ = *p--;
1501 }
1502 /*
1503 * Last hop goes to final destination.
1504 */
1505 *q = isr->isr_dst;
1506 m_tag_delete(m0, (struct m_tag *)isr);
1507 return (m);
1508 }
1509
1510 /*
1511 * Strip out IP options, at higher level protocol in the kernel.
1512 */
1513 void
ip_stripoptions(struct mbuf * m)1514 ip_stripoptions(struct mbuf *m)
1515 {
1516 int i;
1517 struct ip *ip = mtod(m, struct ip *);
1518 caddr_t opts;
1519 int olen;
1520
1521 olen = (ip->ip_hl<<2) - sizeof (struct ip);
1522 opts = (caddr_t)(ip + 1);
1523 i = m->m_len - (sizeof (struct ip) + olen);
1524 memmove(opts, opts + olen, i);
1525 m->m_len -= olen;
1526 if (m->m_flags & M_PKTHDR)
1527 m->m_pkthdr.len -= olen;
1528 ip->ip_hl = sizeof(struct ip) >> 2;
1529 ip->ip_len = htons(ntohs(ip->ip_len) - olen);
1530 }
1531
1532 const u_char inetctlerrmap[PRC_NCMDS] = {
1533 0, 0, 0, 0,
1534 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1535 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1536 EMSGSIZE, EHOSTUNREACH, 0, 0,
1537 0, 0, 0, 0,
1538 ENOPROTOOPT
1539 };
1540
1541 /*
1542 * Forward a packet. If some error occurs return the sender
1543 * an icmp packet. Note we can't always generate a meaningful
1544 * icmp message because icmp doesn't have a large enough repertoire
1545 * of codes and types.
1546 *
1547 * If not forwarding, just drop the packet. This could be confusing
1548 * if ip_forwarding was zero but some routing protocol was advancing
1549 * us as a gateway to somewhere. However, we must let the routing
1550 * protocol deal with that.
1551 *
1552 * The srcrt parameter indicates whether the packet is being forwarded
1553 * via a source route.
1554 */
1555 void
ip_forward(struct mbuf * m,struct ifnet * ifp,struct route * ro,int flags)1556 ip_forward(struct mbuf *m, struct ifnet *ifp, struct route *ro, int flags)
1557 {
1558 struct ip *ip = mtod(m, struct ip *);
1559 struct route iproute;
1560 struct rtentry *rt;
1561 u_int rtableid = m->m_pkthdr.ph_rtableid;
1562 u_int8_t loopcnt = m->m_pkthdr.ph_loopcnt;
1563 u_int icmp_len;
1564 char icmp_buf[68];
1565 CTASSERT(sizeof(icmp_buf) <= MHLEN);
1566 u_short mflags, pfflags;
1567 struct mbuf *mcopy;
1568 int error = 0, type = 0, code = 0, destmtu = 0;
1569 u_int32_t dest;
1570
1571 dest = 0;
1572 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1573 ipstat_inc(ips_cantforward);
1574 m_freem(m);
1575 goto done;
1576 }
1577 if (ip->ip_ttl <= IPTTLDEC) {
1578 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1579 goto done;
1580 }
1581
1582 if (ro == NULL) {
1583 ro = &iproute;
1584 ro->ro_rt = NULL;
1585 }
1586 rt = route_mpath(ro, &ip->ip_dst, &ip->ip_src, rtableid);
1587 if (rt == NULL) {
1588 ipstat_inc(ips_noroute);
1589 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1590 goto done;
1591 }
1592
1593 /*
1594 * Save at most 68 bytes of the packet in case we need to generate
1595 * an ICMP message to the src. The data is saved on the stack.
1596 * A new mbuf is only allocated when ICMP is actually created.
1597 */
1598 icmp_len = min(sizeof(icmp_buf), ntohs(ip->ip_len));
1599 mflags = m->m_flags;
1600 pfflags = m->m_pkthdr.pf.flags;
1601 m_copydata(m, 0, icmp_len, icmp_buf);
1602
1603 ip->ip_ttl -= IPTTLDEC;
1604
1605 /*
1606 * If forwarding packet using same interface that it came in on,
1607 * perhaps should send a redirect to sender to shortcut a hop.
1608 * Only send redirect if source is sending directly to us,
1609 * and if packet was not source routed (or has any options).
1610 * Also, don't send redirect if forwarding using a default route
1611 * or a route modified by a redirect.
1612 * Don't send redirect if we advertise destination's arp address
1613 * as ours (proxy arp).
1614 */
1615 if (rt->rt_ifidx == ifp->if_index &&
1616 !ISSET(rt->rt_flags, RTF_DYNAMIC|RTF_MODIFIED) &&
1617 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
1618 !ISSET(flags, IP_REDIRECT) &&
1619 atomic_load_int(&ip_sendredirects) &&
1620 !arpproxy(satosin(rt_key(rt))->sin_addr, rtableid)) {
1621 if ((ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_netmask) ==
1622 ifatoia(rt->rt_ifa)->ia_net) {
1623 if (rt->rt_flags & RTF_GATEWAY)
1624 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1625 else
1626 dest = ip->ip_dst.s_addr;
1627 /* Router requirements says to only send host redirects */
1628 type = ICMP_REDIRECT;
1629 code = ICMP_REDIRECT_HOST;
1630 }
1631 }
1632
1633 error = ip_output(m, NULL, ro, flags | IP_FORWARDING, NULL, NULL, 0);
1634 rt = ro->ro_rt;
1635 if (error)
1636 ipstat_inc(ips_cantforward);
1637 else {
1638 ipstat_inc(ips_forward);
1639 if (type)
1640 ipstat_inc(ips_redirectsent);
1641 else
1642 goto done;
1643 }
1644 switch (error) {
1645 case 0: /* forwarded, but need redirect */
1646 /* type, code set above */
1647 break;
1648
1649 case EMSGSIZE:
1650 type = ICMP_UNREACH;
1651 code = ICMP_UNREACH_NEEDFRAG;
1652 if (rt != NULL) {
1653 u_int rtmtu;
1654
1655 rtmtu = atomic_load_int(&rt->rt_mtu);
1656 if (rtmtu != 0) {
1657 destmtu = rtmtu;
1658 } else {
1659 struct ifnet *destifp;
1660
1661 destifp = if_get(rt->rt_ifidx);
1662 if (destifp != NULL)
1663 destmtu = destifp->if_mtu;
1664 if_put(destifp);
1665 }
1666 }
1667 ipstat_inc(ips_cantfrag);
1668 if (destmtu == 0)
1669 goto done;
1670 break;
1671
1672 case EACCES:
1673 /*
1674 * pf(4) blocked the packet. There is no need to send an ICMP
1675 * packet back since pf(4) takes care of it.
1676 */
1677 goto done;
1678
1679 case ENOBUFS:
1680 /*
1681 * a router should not generate ICMP_SOURCEQUENCH as
1682 * required in RFC1812 Requirements for IP Version 4 Routers.
1683 * source quench could be a big problem under DoS attacks,
1684 * or the underlying interface is rate-limited.
1685 */
1686 goto done;
1687
1688 case ENETUNREACH: /* shouldn't happen, checked above */
1689 case EHOSTUNREACH:
1690 case ENETDOWN:
1691 case EHOSTDOWN:
1692 default:
1693 type = ICMP_UNREACH;
1694 code = ICMP_UNREACH_HOST;
1695 break;
1696 }
1697
1698 mcopy = m_gethdr(M_DONTWAIT, MT_DATA);
1699 if (mcopy == NULL)
1700 goto done;
1701 mcopy->m_len = mcopy->m_pkthdr.len = icmp_len;
1702 mcopy->m_flags |= (mflags & M_COPYFLAGS);
1703 mcopy->m_pkthdr.ph_rtableid = rtableid;
1704 mcopy->m_pkthdr.ph_ifidx = ifp->if_index;
1705 mcopy->m_pkthdr.ph_loopcnt = loopcnt;
1706 mcopy->m_pkthdr.pf.flags |= (pfflags & PF_TAG_GENERATED);
1707 memcpy(mcopy->m_data, icmp_buf, icmp_len);
1708 icmp_error(mcopy, type, code, dest, destmtu);
1709
1710 done:
1711 if (ro == &iproute)
1712 rtfree(ro->ro_rt);
1713 }
1714
1715 int
ip_sysctl(int * name,u_int namelen,void * oldp,size_t * oldlenp,void * newp,size_t newlen)1716 ip_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
1717 size_t newlen)
1718 {
1719 #ifdef MROUTING
1720 extern struct mrtstat mrtstat;
1721 #endif
1722 int oldval, error;
1723
1724 /* Almost all sysctl names at this level are terminal. */
1725 if (namelen != 1 && name[0] != IPCTL_IFQUEUE &&
1726 name[0] != IPCTL_ARPQUEUE)
1727 return (ENOTDIR);
1728
1729 switch (name[0]) {
1730 case IPCTL_SOURCEROUTE:
1731 NET_LOCK();
1732 error = sysctl_securelevel_int(oldp, oldlenp, newp, newlen,
1733 &ip_dosourceroute);
1734 NET_UNLOCK();
1735 return (error);
1736 case IPCTL_MTUDISC:
1737 NET_LOCK();
1738 error = sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtudisc);
1739 if (ip_mtudisc == 0)
1740 rt_timer_queue_flush(&ip_mtudisc_timeout_q);
1741 NET_UNLOCK();
1742 return error;
1743 case IPCTL_MTUDISCTIMEOUT:
1744 NET_LOCK();
1745 error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
1746 &ip_mtudisc_timeout, 0, INT_MAX);
1747 rt_timer_queue_change(&ip_mtudisc_timeout_q,
1748 ip_mtudisc_timeout);
1749 NET_UNLOCK();
1750 return (error);
1751 #ifdef IPSEC
1752 case IPCTL_ENCDEBUG:
1753 case IPCTL_IPSEC_STATS:
1754 case IPCTL_IPSEC_EXPIRE_ACQUIRE:
1755 case IPCTL_IPSEC_EMBRYONIC_SA_TIMEOUT:
1756 case IPCTL_IPSEC_REQUIRE_PFS:
1757 case IPCTL_IPSEC_SOFT_ALLOCATIONS:
1758 case IPCTL_IPSEC_ALLOCATIONS:
1759 case IPCTL_IPSEC_SOFT_BYTES:
1760 case IPCTL_IPSEC_BYTES:
1761 case IPCTL_IPSEC_TIMEOUT:
1762 case IPCTL_IPSEC_SOFT_TIMEOUT:
1763 case IPCTL_IPSEC_SOFT_FIRSTUSE:
1764 case IPCTL_IPSEC_FIRSTUSE:
1765 case IPCTL_IPSEC_ENC_ALGORITHM:
1766 case IPCTL_IPSEC_AUTH_ALGORITHM:
1767 case IPCTL_IPSEC_IPCOMP_ALGORITHM:
1768 return (ipsec_sysctl(name, namelen, oldp, oldlenp, newp,
1769 newlen));
1770 #endif
1771 case IPCTL_IFQUEUE:
1772 return (sysctl_niq(name + 1, namelen - 1,
1773 oldp, oldlenp, newp, newlen, &ipintrq));
1774 case IPCTL_ARPQUEUE:
1775 return (sysctl_niq(name + 1, namelen - 1,
1776 oldp, oldlenp, newp, newlen, &arpinq));
1777 case IPCTL_ARPQUEUED:
1778 return (sysctl_rdint(oldp, oldlenp, newp,
1779 atomic_load_int(&la_hold_total)));
1780 case IPCTL_STATS:
1781 return (ip_sysctl_ipstat(oldp, oldlenp, newp));
1782 #ifdef MROUTING
1783 case IPCTL_MRTSTATS:
1784 return (sysctl_rdstruct(oldp, oldlenp, newp,
1785 &mrtstat, sizeof(mrtstat)));
1786 case IPCTL_MRTMFC:
1787 if (newp)
1788 return (EPERM);
1789 NET_LOCK();
1790 error = mrt_sysctl_mfc(oldp, oldlenp);
1791 NET_UNLOCK();
1792 return (error);
1793 case IPCTL_MRTVIF:
1794 if (newp)
1795 return (EPERM);
1796 NET_LOCK();
1797 error = mrt_sysctl_vif(oldp, oldlenp);
1798 NET_UNLOCK();
1799 return (error);
1800 #else
1801 case IPCTL_MRTPROTO:
1802 case IPCTL_MRTSTATS:
1803 case IPCTL_MRTMFC:
1804 case IPCTL_MRTVIF:
1805 return (EOPNOTSUPP);
1806 #endif
1807 case IPCTL_MULTIPATH:
1808 NET_LOCK();
1809 oldval = ipmultipath;
1810 error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
1811 &ipmultipath, 0, 1);
1812 if (oldval != ipmultipath)
1813 atomic_inc_long(&rtgeneration);
1814 NET_UNLOCK();
1815 return (error);
1816 case IPCTL_FORWARDING:
1817 case IPCTL_SENDREDIRECTS:
1818 case IPCTL_DIRECTEDBCAST:
1819 return (sysctl_bounded_arr(
1820 ipctl_vars_unlocked, nitems(ipctl_vars_unlocked),
1821 name, namelen, oldp, oldlenp, newp, newlen));
1822 default:
1823 NET_LOCK();
1824 error = sysctl_bounded_arr(ipctl_vars, nitems(ipctl_vars),
1825 name, namelen, oldp, oldlenp, newp, newlen);
1826 NET_UNLOCK();
1827 return (error);
1828 }
1829 /* NOTREACHED */
1830 }
1831
1832 int
ip_sysctl_ipstat(void * oldp,size_t * oldlenp,void * newp)1833 ip_sysctl_ipstat(void *oldp, size_t *oldlenp, void *newp)
1834 {
1835 uint64_t counters[ips_ncounters];
1836 struct ipstat ipstat;
1837 u_long *words = (u_long *)&ipstat;
1838 int i;
1839
1840 CTASSERT(sizeof(ipstat) == (nitems(counters) * sizeof(u_long)));
1841 memset(&ipstat, 0, sizeof ipstat);
1842 counters_read(ipcounters, counters, nitems(counters), NULL);
1843
1844 for (i = 0; i < nitems(counters); i++)
1845 words[i] = (u_long)counters[i];
1846
1847 return (sysctl_rdstruct(oldp, oldlenp, newp, &ipstat, sizeof(ipstat)));
1848 }
1849
1850 void
ip_savecontrol(struct inpcb * inp,struct mbuf ** mp,struct ip * ip,struct mbuf * m)1851 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1852 struct mbuf *m)
1853 {
1854 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1855 struct timeval tv;
1856
1857 m_microtime(m, &tv);
1858 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1859 SCM_TIMESTAMP, SOL_SOCKET);
1860 if (*mp)
1861 mp = &(*mp)->m_next;
1862 }
1863
1864 if (inp->inp_flags & INP_RECVDSTADDR) {
1865 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1866 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1867 if (*mp)
1868 mp = &(*mp)->m_next;
1869 }
1870 #ifdef notyet
1871 /* this code is broken and will probably never be fixed. */
1872 /* options were tossed already */
1873 if (inp->inp_flags & INP_RECVOPTS) {
1874 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1875 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1876 if (*mp)
1877 mp = &(*mp)->m_next;
1878 }
1879 /* ip_srcroute doesn't do what we want here, need to fix */
1880 if (inp->inp_flags & INP_RECVRETOPTS) {
1881 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1882 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1883 if (*mp)
1884 mp = &(*mp)->m_next;
1885 }
1886 #endif
1887 if (inp->inp_flags & INP_RECVIF) {
1888 struct sockaddr_dl sdl;
1889 struct ifnet *ifp;
1890
1891 ifp = if_get(m->m_pkthdr.ph_ifidx);
1892 if (ifp == NULL || ifp->if_sadl == NULL) {
1893 memset(&sdl, 0, sizeof(sdl));
1894 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]);
1895 sdl.sdl_family = AF_LINK;
1896 sdl.sdl_index = ifp != NULL ? ifp->if_index : 0;
1897 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0;
1898 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len,
1899 IP_RECVIF, IPPROTO_IP);
1900 } else {
1901 *mp = sbcreatecontrol((caddr_t) ifp->if_sadl,
1902 ifp->if_sadl->sdl_len, IP_RECVIF, IPPROTO_IP);
1903 }
1904 if (*mp)
1905 mp = &(*mp)->m_next;
1906 if_put(ifp);
1907 }
1908 if (inp->inp_flags & INP_RECVTTL) {
1909 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1910 sizeof(u_int8_t), IP_RECVTTL, IPPROTO_IP);
1911 if (*mp)
1912 mp = &(*mp)->m_next;
1913 }
1914 if (inp->inp_flags & INP_RECVRTABLE) {
1915 u_int rtableid = inp->inp_rtableid;
1916
1917 #if NPF > 0
1918 if (m && m->m_pkthdr.pf.flags & PF_TAG_DIVERTED) {
1919 struct pf_divert *divert;
1920
1921 divert = pf_find_divert(m);
1922 KASSERT(divert != NULL);
1923 rtableid = divert->rdomain;
1924 }
1925 #endif
1926
1927 *mp = sbcreatecontrol((caddr_t) &rtableid,
1928 sizeof(u_int), IP_RECVRTABLE, IPPROTO_IP);
1929 if (*mp)
1930 mp = &(*mp)->m_next;
1931 }
1932 }
1933
1934 void
ip_send_do_dispatch(void * xmq,int flags)1935 ip_send_do_dispatch(void *xmq, int flags)
1936 {
1937 struct mbuf_queue *mq = xmq;
1938 struct mbuf *m;
1939 struct mbuf_list ml;
1940 struct m_tag *mtag;
1941
1942 mq_delist(mq, &ml);
1943 if (ml_empty(&ml))
1944 return;
1945
1946 NET_LOCK_SHARED();
1947 while ((m = ml_dequeue(&ml)) != NULL) {
1948 u_int32_t ipsecflowinfo = 0;
1949
1950 if ((mtag = m_tag_find(m, PACKET_TAG_IPSEC_FLOWINFO, NULL))
1951 != NULL) {
1952 ipsecflowinfo = *(u_int32_t *)(mtag + 1);
1953 m_tag_delete(m, mtag);
1954 }
1955 ip_output(m, NULL, NULL, flags, NULL, NULL, ipsecflowinfo);
1956 }
1957 NET_UNLOCK_SHARED();
1958 }
1959
1960 void
ip_sendraw_dispatch(void * xmq)1961 ip_sendraw_dispatch(void *xmq)
1962 {
1963 ip_send_do_dispatch(xmq, IP_RAWOUTPUT);
1964 }
1965
1966 void
ip_send_dispatch(void * xmq)1967 ip_send_dispatch(void *xmq)
1968 {
1969 ip_send_do_dispatch(xmq, 0);
1970 }
1971
1972 void
ip_send(struct mbuf * m)1973 ip_send(struct mbuf *m)
1974 {
1975 mq_enqueue(&ipsend_mq, m);
1976 task_add(net_tq(0), &ipsend_task);
1977 }
1978
1979 void
ip_send_raw(struct mbuf * m)1980 ip_send_raw(struct mbuf *m)
1981 {
1982 mq_enqueue(&ipsendraw_mq, m);
1983 task_add(net_tq(0), &ipsendraw_task);
1984 }
1985