1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 *
31 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
32 */
33
34 #include <sys/cdefs.h>
35 #include "opt_bootp.h"
36 #include "opt_inet.h"
37 #include "opt_ipstealth.h"
38 #include "opt_ipsec.h"
39 #include "opt_route.h"
40 #include "opt_rss.h"
41 #include "opt_sctp.h"
42
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/hhook.h>
46 #include <sys/mbuf.h>
47 #include <sys/malloc.h>
48 #include <sys/domain.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
51 #include <sys/time.h>
52 #include <sys/kernel.h>
53 #include <sys/lock.h>
54 #include <sys/rmlock.h>
55 #include <sys/rwlock.h>
56 #include <sys/sdt.h>
57 #include <sys/syslog.h>
58 #include <sys/sysctl.h>
59
60 #include <net/if.h>
61 #include <net/if_types.h>
62 #include <net/if_var.h>
63 #include <net/if_dl.h>
64 #include <net/if_private.h>
65 #include <net/pfil.h>
66 #include <net/route.h>
67 #include <net/route/nhop.h>
68 #include <net/netisr.h>
69 #include <net/rss_config.h>
70 #include <net/vnet.h>
71
72 #include <netinet/in.h>
73 #include <netinet/in_kdtrace.h>
74 #include <netinet/in_systm.h>
75 #include <netinet/in_var.h>
76 #include <netinet/ip.h>
77 #include <netinet/in_fib.h>
78 #include <netinet/in_pcb.h>
79 #include <netinet/ip_var.h>
80 #include <netinet/ip_encap.h>
81 #include <netinet/ip_fw.h>
82 #include <netinet/ip_icmp.h>
83 #include <netinet/igmp_var.h>
84 #include <netinet/ip_options.h>
85 #include <machine/in_cksum.h>
86 #include <netinet/ip_carp.h>
87 #include <netinet/in_rss.h>
88 #ifdef SCTP
89 #include <netinet/sctp_var.h>
90 #endif
91
92 #include <netipsec/ipsec_support.h>
93
94 #include <sys/socketvar.h>
95
96 #include <security/mac/mac_framework.h>
97
98 #ifdef CTASSERT
99 CTASSERT(sizeof(struct ip) == 20);
100 #endif
101
102 /* IP reassembly functions are defined in ip_reass.c. */
103 extern void ipreass_init(void);
104 extern void ipreass_vnet_init(void);
105 #ifdef VIMAGE
106 extern void ipreass_destroy(void);
107 #endif
108
109 VNET_DEFINE(int, rsvp_on);
110
111 VNET_DEFINE(int, ipforwarding);
112 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
113 &VNET_NAME(ipforwarding), 0,
114 "Enable IP forwarding between interfaces");
115
116 /*
117 * Respond with an ICMP host redirect when we forward a packet out of
118 * the same interface on which it was received. See RFC 792.
119 */
120 VNET_DEFINE(int, ipsendredirects) = 1;
121 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
122 &VNET_NAME(ipsendredirects), 0,
123 "Enable sending IP redirects");
124
125 VNET_DEFINE_STATIC(bool, ip_strong_es) = false;
126 #define V_ip_strong_es VNET(ip_strong_es)
127 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, rfc1122_strong_es,
128 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_strong_es), false,
129 "Packet's IP destination address must match address on arrival interface");
130
131 VNET_DEFINE_STATIC(bool, ip_sav) = true;
132 #define V_ip_sav VNET(ip_sav)
133 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, source_address_validation,
134 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_sav), true,
135 "Drop incoming packets with source address that is a local address");
136
137 /* Packet filter hooks */
138 VNET_DEFINE(pfil_head_t, inet_pfil_head);
139 VNET_DEFINE(pfil_head_t, inet_local_pfil_head);
140
141 static struct netisr_handler ip_nh = {
142 .nh_name = "ip",
143 .nh_handler = ip_input,
144 .nh_proto = NETISR_IP,
145 #ifdef RSS
146 .nh_m2cpuid = rss_soft_m2cpuid_v4,
147 .nh_policy = NETISR_POLICY_CPU,
148 .nh_dispatch = NETISR_DISPATCH_HYBRID,
149 #else
150 .nh_policy = NETISR_POLICY_FLOW,
151 #endif
152 };
153
154 #ifdef RSS
155 /*
156 * Directly dispatched frames are currently assumed
157 * to have a flowid already calculated.
158 *
159 * It should likely have something that assert it
160 * actually has valid flow details.
161 */
162 static struct netisr_handler ip_direct_nh = {
163 .nh_name = "ip_direct",
164 .nh_handler = ip_direct_input,
165 .nh_proto = NETISR_IP_DIRECT,
166 .nh_m2cpuid = rss_soft_m2cpuid_v4,
167 .nh_policy = NETISR_POLICY_CPU,
168 .nh_dispatch = NETISR_DISPATCH_HYBRID,
169 };
170 #endif
171
172 ipproto_input_t *ip_protox[IPPROTO_MAX] = {
173 [0 ... IPPROTO_MAX - 1] = rip_input };
174 ipproto_ctlinput_t *ip_ctlprotox[IPPROTO_MAX] = {
175 [0 ... IPPROTO_MAX - 1] = rip_ctlinput };
176
177 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
178 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
179 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
180
181 /* Make sure it is safe to use hashinit(9) on CK_LIST. */
182 CTASSERT(sizeof(struct in_ifaddrhashhead) == sizeof(LIST_HEAD(, in_addr)));
183
184 #ifdef IPCTL_DEFMTU
185 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
186 &ip_mtu, 0, "Default MTU");
187 #endif
188
189 #ifdef IPSTEALTH
190 VNET_DEFINE(int, ipstealth);
191 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
192 &VNET_NAME(ipstealth), 0,
193 "IP stealth mode, no TTL decrementation on forwarding");
194 #endif
195
196 /*
197 * IP statistics are stored in the "array" of counter(9)s.
198 */
199 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
200 VNET_PCPUSTAT_SYSINIT(ipstat);
201 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
202 "IP statistics (struct ipstat, netinet/ip_var.h)");
203
204 #ifdef VIMAGE
205 VNET_PCPUSTAT_SYSUNINIT(ipstat);
206 #endif /* VIMAGE */
207
208 /*
209 * Kernel module interface for updating ipstat. The argument is an index
210 * into ipstat treated as an array.
211 */
212 void
kmod_ipstat_inc(int statnum)213 kmod_ipstat_inc(int statnum)
214 {
215
216 counter_u64_add(VNET(ipstat)[statnum], 1);
217 }
218
219 void
kmod_ipstat_dec(int statnum)220 kmod_ipstat_dec(int statnum)
221 {
222
223 counter_u64_add(VNET(ipstat)[statnum], -1);
224 }
225
226 static int
sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)227 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
228 {
229 int error, qlimit;
230
231 netisr_getqlimit(&ip_nh, &qlimit);
232 error = sysctl_handle_int(oidp, &qlimit, 0, req);
233 if (error || !req->newptr)
234 return (error);
235 if (qlimit < 1)
236 return (EINVAL);
237 return (netisr_setqlimit(&ip_nh, qlimit));
238 }
239 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
240 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
241 sysctl_netinet_intr_queue_maxlen, "I",
242 "Maximum size of the IP input queue");
243
244 static int
sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)245 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
246 {
247 u_int64_t qdrops_long;
248 int error, qdrops;
249
250 netisr_getqdrops(&ip_nh, &qdrops_long);
251 qdrops = qdrops_long;
252 error = sysctl_handle_int(oidp, &qdrops, 0, req);
253 if (error || !req->newptr)
254 return (error);
255 if (qdrops != 0)
256 return (EINVAL);
257 netisr_clearqdrops(&ip_nh);
258 return (0);
259 }
260
261 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
262 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
263 0, 0, sysctl_netinet_intr_queue_drops, "I",
264 "Number of packets dropped from the IP input queue");
265
266 #ifdef RSS
267 static int
sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)268 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
269 {
270 int error, qlimit;
271
272 netisr_getqlimit(&ip_direct_nh, &qlimit);
273 error = sysctl_handle_int(oidp, &qlimit, 0, req);
274 if (error || !req->newptr)
275 return (error);
276 if (qlimit < 1)
277 return (EINVAL);
278 return (netisr_setqlimit(&ip_direct_nh, qlimit));
279 }
280 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
281 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
282 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
283 "I", "Maximum size of the IP direct input queue");
284
285 static int
sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)286 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
287 {
288 u_int64_t qdrops_long;
289 int error, qdrops;
290
291 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
292 qdrops = qdrops_long;
293 error = sysctl_handle_int(oidp, &qdrops, 0, req);
294 if (error || !req->newptr)
295 return (error);
296 if (qdrops != 0)
297 return (EINVAL);
298 netisr_clearqdrops(&ip_direct_nh);
299 return (0);
300 }
301
302 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
303 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
304 sysctl_netinet_intr_direct_queue_drops, "I",
305 "Number of packets dropped from the IP direct input queue");
306 #endif /* RSS */
307
308 /*
309 * IP initialization: fill in IP protocol switch table.
310 * All protocols not implemented in kernel go to raw IP protocol handler.
311 */
312 static void
ip_vnet_init(void * arg __unused)313 ip_vnet_init(void *arg __unused)
314 {
315 struct pfil_head_args args;
316
317 CK_STAILQ_INIT(&V_in_ifaddrhead);
318 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
319
320 /* Initialize IP reassembly queue. */
321 ipreass_vnet_init();
322
323 /* Initialize packet filter hooks. */
324 args.pa_version = PFIL_VERSION;
325 args.pa_flags = PFIL_IN | PFIL_OUT;
326 args.pa_type = PFIL_TYPE_IP4;
327 args.pa_headname = PFIL_INET_NAME;
328 V_inet_pfil_head = pfil_head_register(&args);
329
330 args.pa_flags = PFIL_OUT;
331 args.pa_headname = PFIL_INET_LOCAL_NAME;
332 V_inet_local_pfil_head = pfil_head_register(&args);
333
334 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
335 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
336 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
337 printf("%s: WARNING: unable to register input helper hook\n",
338 __func__);
339 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
340 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
341 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
342 printf("%s: WARNING: unable to register output helper hook\n",
343 __func__);
344
345 #ifdef VIMAGE
346 netisr_register_vnet(&ip_nh);
347 #ifdef RSS
348 netisr_register_vnet(&ip_direct_nh);
349 #endif
350 #endif
351 }
352 VNET_SYSINIT(ip_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
353 ip_vnet_init, NULL);
354
355 static void
ip_init(const void * unused __unused)356 ip_init(const void *unused __unused)
357 {
358
359 ipreass_init();
360
361 /*
362 * Register statically compiled protocols, that are unlikely to
363 * ever become dynamic.
364 */
365 IPPROTO_REGISTER(IPPROTO_ICMP, icmp_input, NULL);
366 IPPROTO_REGISTER(IPPROTO_IGMP, igmp_input, NULL);
367 IPPROTO_REGISTER(IPPROTO_RSVP, rsvp_input, NULL);
368 IPPROTO_REGISTER(IPPROTO_IPV4, encap4_input, NULL);
369 IPPROTO_REGISTER(IPPROTO_MOBILE, encap4_input, NULL);
370 IPPROTO_REGISTER(IPPROTO_ETHERIP, encap4_input, NULL);
371 IPPROTO_REGISTER(IPPROTO_GRE, encap4_input, NULL);
372 IPPROTO_REGISTER(IPPROTO_IPV6, encap4_input, NULL);
373 IPPROTO_REGISTER(IPPROTO_PIM, encap4_input, NULL);
374 #ifdef SCTP /* XXX: has a loadable & static version */
375 IPPROTO_REGISTER(IPPROTO_SCTP, sctp_input, sctp_ctlinput);
376 #endif
377
378 netisr_register(&ip_nh);
379 #ifdef RSS
380 netisr_register(&ip_direct_nh);
381 #endif
382 }
383 SYSINIT(ip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_init, NULL);
384
385 #ifdef VIMAGE
386 static void
ip_destroy(void * unused __unused)387 ip_destroy(void *unused __unused)
388 {
389 int error;
390
391 #ifdef RSS
392 netisr_unregister_vnet(&ip_direct_nh);
393 #endif
394 netisr_unregister_vnet(&ip_nh);
395
396 pfil_head_unregister(V_inet_pfil_head);
397 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
398 if (error != 0) {
399 printf("%s: WARNING: unable to deregister input helper hook "
400 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
401 "error %d returned\n", __func__, error);
402 }
403 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
404 if (error != 0) {
405 printf("%s: WARNING: unable to deregister output helper hook "
406 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
407 "error %d returned\n", __func__, error);
408 }
409
410 /* Remove the IPv4 addresses from all interfaces. */
411 in_ifscrub_all();
412
413 /* Make sure the IPv4 routes are gone as well. */
414 rib_flush_routes_family(AF_INET);
415
416 /* Destroy IP reassembly queue. */
417 ipreass_destroy();
418
419 /* Cleanup in_ifaddr hash table; should be empty. */
420 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
421 }
422
423 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
424 #endif
425
426 #ifdef RSS
427 /*
428 * IP direct input routine.
429 *
430 * This is called when reinjecting completed fragments where
431 * all of the previous checking and book-keeping has been done.
432 */
433 void
ip_direct_input(struct mbuf * m)434 ip_direct_input(struct mbuf *m)
435 {
436 struct ip *ip;
437 int hlen;
438
439 ip = mtod(m, struct ip *);
440 hlen = ip->ip_hl << 2;
441
442 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
443 if (IPSEC_ENABLED(ipv4)) {
444 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
445 return;
446 }
447 #endif /* IPSEC */
448 IPSTAT_INC(ips_delivered);
449 ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
450 }
451 #endif
452
453 /*
454 * Ip input routine. Checksum and byte swap header. If fragmented
455 * try to reassemble. Process options. Pass to next level.
456 */
457 void
ip_input(struct mbuf * m)458 ip_input(struct mbuf *m)
459 {
460 struct ip *ip = NULL;
461 struct in_ifaddr *ia = NULL;
462 struct ifaddr *ifa;
463 struct ifnet *ifp;
464 int hlen = 0;
465 uint16_t sum, ip_len;
466 int dchg = 0; /* dest changed after fw */
467 struct in_addr odst; /* original dst address */
468 bool strong_es;
469
470 M_ASSERTPKTHDR(m);
471 NET_EPOCH_ASSERT();
472
473 if (m->m_flags & M_FASTFWD_OURS) {
474 m->m_flags &= ~M_FASTFWD_OURS;
475 /* Set up some basics that will be used later. */
476 ip = mtod(m, struct ip *);
477 hlen = ip->ip_hl << 2;
478 ip_len = ntohs(ip->ip_len);
479 goto ours;
480 }
481
482 IPSTAT_INC(ips_total);
483
484 if (__predict_false(m->m_pkthdr.len < sizeof(struct ip)))
485 goto tooshort;
486
487 if (m->m_len < sizeof(struct ip)) {
488 m = m_pullup(m, sizeof(struct ip));
489 if (__predict_false(m == NULL)) {
490 IPSTAT_INC(ips_toosmall);
491 return;
492 }
493 }
494 ip = mtod(m, struct ip *);
495
496 if (__predict_false(ip->ip_v != IPVERSION)) {
497 IPSTAT_INC(ips_badvers);
498 goto bad;
499 }
500
501 hlen = ip->ip_hl << 2;
502 if (__predict_false(hlen < sizeof(struct ip))) { /* minimum header length */
503 IPSTAT_INC(ips_badhlen);
504 goto bad;
505 }
506 if (hlen > m->m_len) {
507 m = m_pullup(m, hlen);
508 if (__predict_false(m == NULL)) {
509 IPSTAT_INC(ips_badhlen);
510 return;
511 }
512 ip = mtod(m, struct ip *);
513 }
514
515 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
516
517 /* IN_LOOPBACK must not appear on the wire - RFC1122 */
518 ifp = m->m_pkthdr.rcvif;
519 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
520 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
521 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
522 IPSTAT_INC(ips_badaddr);
523 goto bad;
524 }
525 }
526
527 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
528 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
529 } else {
530 if (hlen == sizeof(struct ip)) {
531 sum = in_cksum_hdr(ip);
532 } else {
533 sum = in_cksum(m, hlen);
534 }
535 }
536 if (__predict_false(sum)) {
537 IPSTAT_INC(ips_badsum);
538 goto bad;
539 }
540
541 ip_len = ntohs(ip->ip_len);
542 if (__predict_false(ip_len < hlen)) {
543 IPSTAT_INC(ips_badlen);
544 goto bad;
545 }
546
547 /*
548 * Check that the amount of data in the buffers
549 * is as at least much as the IP header would have us expect.
550 * Trim mbufs if longer than we expect.
551 * Drop packet if shorter than we expect.
552 */
553 if (__predict_false(m->m_pkthdr.len < ip_len)) {
554 tooshort:
555 IPSTAT_INC(ips_tooshort);
556 goto bad;
557 }
558 if (m->m_pkthdr.len > ip_len) {
559 if (m->m_len == m->m_pkthdr.len) {
560 m->m_len = ip_len;
561 m->m_pkthdr.len = ip_len;
562 } else
563 m_adj(m, ip_len - m->m_pkthdr.len);
564 }
565
566 /*
567 * Try to forward the packet, but if we fail continue.
568 * ip_tryforward() may generate redirects these days.
569 * XXX the logic below falling through to normal processing
570 * if redirects are required should be revisited as well.
571 * ip_tryforward() does inbound and outbound packet firewall
572 * processing. If firewall has decided that destination becomes
573 * our local address, it sets M_FASTFWD_OURS flag. In this
574 * case skip another inbound firewall processing and update
575 * ip pointer.
576 */
577 if (V_ipforwarding != 0
578 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
579 && (!IPSEC_ENABLED(ipv4) ||
580 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
581 #endif
582 ) {
583 /*
584 * ip_dooptions() was run so we can ignore the source route (or
585 * any IP options case) case for redirects in ip_tryforward().
586 */
587 if ((m = ip_tryforward(m)) == NULL)
588 return;
589 if (m->m_flags & M_FASTFWD_OURS) {
590 m->m_flags &= ~M_FASTFWD_OURS;
591 ip = mtod(m, struct ip *);
592 goto ours;
593 }
594 }
595
596 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
597 /*
598 * Bypass packet filtering for packets previously handled by IPsec.
599 */
600 if (IPSEC_ENABLED(ipv4) &&
601 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
602 goto passin;
603 #endif
604
605 /*
606 * Run through list of hooks for input packets.
607 *
608 * NB: Beware of the destination address changing (e.g.
609 * by NAT rewriting). When this happens, tell
610 * ip_forward to do the right thing.
611 */
612
613 /* Jump over all PFIL processing if hooks are not active. */
614 if (!PFIL_HOOKED_IN(V_inet_pfil_head))
615 goto passin;
616
617 odst = ip->ip_dst;
618 if (pfil_mbuf_in(V_inet_pfil_head, &m, ifp, NULL) !=
619 PFIL_PASS)
620 return;
621 if (m == NULL) /* consumed by filter */
622 return;
623
624 ip = mtod(m, struct ip *);
625 dchg = (odst.s_addr != ip->ip_dst.s_addr);
626
627 if (m->m_flags & M_FASTFWD_OURS) {
628 m->m_flags &= ~M_FASTFWD_OURS;
629 goto ours;
630 }
631 if (m->m_flags & M_IP_NEXTHOP) {
632 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
633 /*
634 * Directly ship the packet on. This allows
635 * forwarding packets originally destined to us
636 * to some other directly connected host.
637 */
638 ip_forward(m, 1);
639 return;
640 }
641 }
642 passin:
643 /*
644 * The unspecified address can appear only as a src address - RFC1122.
645 *
646 * The check is deferred to here to give firewalls a chance to block
647 * (and log) such packets. ip_tryforward() will not process such
648 * packets.
649 */
650 if (__predict_false(ntohl(ip->ip_dst.s_addr) == INADDR_ANY)) {
651 IPSTAT_INC(ips_badaddr);
652 goto bad;
653 }
654
655 /*
656 * Process options and, if not destined for us,
657 * ship it on. ip_dooptions returns 1 when an
658 * error was detected (causing an icmp message
659 * to be sent and the original packet to be freed).
660 */
661 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
662 return;
663
664 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
665 * matter if it is destined to another node, or whether it is
666 * a multicast one, RSVP wants it! and prevents it from being forwarded
667 * anywhere else. Also checks if the rsvp daemon is running before
668 * grabbing the packet.
669 */
670 if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
671 goto ours;
672
673 /*
674 * Check our list of addresses, to see if the packet is for us.
675 * If we don't have any addresses, assume any unicast packet
676 * we receive might be for us (and let the upper layers deal
677 * with it).
678 */
679 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
680 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
681 goto ours;
682
683 /*
684 * Enable a consistency check between the destination address
685 * and the arrival interface for a unicast packet (the RFC 1122
686 * strong ES model) with a list of additional predicates:
687 * - if IP forwarding is disabled
688 * - the packet is not locally generated
689 * - the packet is not subject to 'ipfw fwd'
690 * - Interface is not running CARP. If the packet got here, we already
691 * checked it with carp_iamatch() and carp_forus().
692 */
693 strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
694 ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
695 ifp->if_carp == NULL && (dchg == 0);
696
697 /*
698 * Check for exact addresses in the hash bucket.
699 */
700 CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
701 if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
702 continue;
703
704 /*
705 * net.inet.ip.rfc1122_strong_es: the address matches, verify
706 * that the packet arrived via the correct interface.
707 */
708 if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
709 IPSTAT_INC(ips_badaddr);
710 goto bad;
711 }
712
713 /*
714 * net.inet.ip.source_address_validation: drop incoming
715 * packets that pretend to be ours.
716 */
717 if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
718 __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
719 IPSTAT_INC(ips_badaddr);
720 goto bad;
721 }
722
723 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
724 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
725 goto ours;
726 }
727
728 /*
729 * Check for broadcast addresses.
730 *
731 * Only accept broadcast packets that arrive via the matching
732 * interface. Reception of forwarded directed broadcasts would
733 * be handled via ip_forward() and ether_output() with the loopback
734 * into the stack for SIMPLEX interfaces handled by ether_output().
735 */
736 if (ifp->if_flags & IFF_BROADCAST) {
737 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
738 if (ifa->ifa_addr->sa_family != AF_INET)
739 continue;
740 ia = ifatoia(ifa);
741 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
742 ip->ip_dst.s_addr) {
743 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
744 counter_u64_add(ia->ia_ifa.ifa_ibytes,
745 m->m_pkthdr.len);
746 goto ours;
747 }
748 #ifdef BOOTP_COMPAT
749 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
750 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
751 counter_u64_add(ia->ia_ifa.ifa_ibytes,
752 m->m_pkthdr.len);
753 goto ours;
754 }
755 #endif
756 }
757 ia = NULL;
758 }
759 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
760 /*
761 * RFC 3927 2.7: Do not forward multicast packets from
762 * IN_LINKLOCAL.
763 */
764 if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
765 /*
766 * If we are acting as a multicast router, all
767 * incoming multicast packets are passed to the
768 * kernel-level multicast forwarding function.
769 * The packet is returned (relatively) intact; if
770 * ip_mforward() returns a non-zero value, the packet
771 * must be discarded, else it may be accepted below.
772 */
773 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
774 IPSTAT_INC(ips_cantforward);
775 m_freem(m);
776 return;
777 }
778
779 /*
780 * The process-level routing daemon needs to receive
781 * all multicast IGMP packets, whether or not this
782 * host belongs to their destination groups.
783 */
784 if (ip->ip_p == IPPROTO_IGMP) {
785 goto ours;
786 }
787 IPSTAT_INC(ips_forward);
788 }
789 /*
790 * Assume the packet is for us, to avoid prematurely taking
791 * a lock on the in_multi hash. Protocols must perform
792 * their own filtering and update statistics accordingly.
793 */
794 goto ours;
795 }
796 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
797 goto ours;
798 if (ip->ip_dst.s_addr == INADDR_ANY)
799 goto ours;
800 /* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
801 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
802 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
803 IPSTAT_INC(ips_cantforward);
804 m_freem(m);
805 return;
806 }
807
808 /*
809 * Not for us; forward if possible and desirable.
810 */
811 if (V_ipforwarding == 0) {
812 IPSTAT_INC(ips_cantforward);
813 m_freem(m);
814 } else {
815 ip_forward(m, dchg);
816 }
817 return;
818
819 ours:
820 #ifdef IPSTEALTH
821 /*
822 * IPSTEALTH: Process non-routing options only
823 * if the packet is destined for us.
824 */
825 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
826 return;
827 #endif /* IPSTEALTH */
828
829 /*
830 * We are going to ship the packet to the local protocol stack. Call the
831 * filter again for this 'output' action, allowing redirect-like rules
832 * to adjust the source address.
833 */
834 if (PFIL_HOOKED_OUT(V_inet_local_pfil_head)) {
835 if (pfil_mbuf_out(V_inet_local_pfil_head, &m, V_loif, NULL) !=
836 PFIL_PASS)
837 return;
838 if (m == NULL) /* consumed by filter */
839 return;
840 ip = mtod(m, struct ip *);
841 }
842
843 /*
844 * Attempt reassembly; if it succeeds, proceed.
845 * ip_reass() will return a different mbuf.
846 */
847 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
848 /* XXXGL: shouldn't we save & set m_flags? */
849 m = ip_reass(m);
850 if (m == NULL)
851 return;
852 ip = mtod(m, struct ip *);
853 /* Get the header length of the reassembled packet */
854 hlen = ip->ip_hl << 2;
855 }
856
857 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
858 if (IPSEC_ENABLED(ipv4)) {
859 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
860 return;
861 }
862 #endif /* IPSEC */
863
864 /*
865 * Switch out to protocol's input routine.
866 */
867 IPSTAT_INC(ips_delivered);
868
869 ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
870 return;
871 bad:
872 m_freem(m);
873 }
874
875 int
ipproto_register(uint8_t proto,ipproto_input_t input,ipproto_ctlinput_t ctl)876 ipproto_register(uint8_t proto, ipproto_input_t input, ipproto_ctlinput_t ctl)
877 {
878
879 MPASS(proto > 0);
880
881 /*
882 * The protocol slot must not be occupied by another protocol
883 * already. An index pointing to rip_input() is unused.
884 */
885 if (ip_protox[proto] == rip_input) {
886 ip_protox[proto] = input;
887 ip_ctlprotox[proto] = ctl;
888 return (0);
889 } else
890 return (EEXIST);
891 }
892
893 int
ipproto_unregister(uint8_t proto)894 ipproto_unregister(uint8_t proto)
895 {
896
897 MPASS(proto > 0);
898
899 if (ip_protox[proto] != rip_input) {
900 ip_protox[proto] = rip_input;
901 ip_ctlprotox[proto] = rip_ctlinput;
902 return (0);
903 } else
904 return (ENOENT);
905 }
906
907 /*
908 * Forward a packet. If some error occurs return the sender
909 * an icmp packet. Note we can't always generate a meaningful
910 * icmp message because icmp doesn't have a large enough repertoire
911 * of codes and types.
912 *
913 * If not forwarding, just drop the packet. This could be confusing
914 * if ipforwarding was zero but some routing protocol was advancing
915 * us as a gateway to somewhere. However, we must let the routing
916 * protocol deal with that.
917 *
918 * The srcrt parameter indicates whether the packet is being forwarded
919 * via a source route.
920 */
921 void
ip_forward(struct mbuf * m,int srcrt)922 ip_forward(struct mbuf *m, int srcrt)
923 {
924 struct ip *ip = mtod(m, struct ip *);
925 struct in_ifaddr *ia;
926 struct mbuf *mcopy;
927 struct sockaddr_in *sin;
928 struct in_addr dest;
929 struct route ro;
930 uint32_t flowid;
931 int error, type = 0, code = 0, mtu = 0;
932
933 NET_EPOCH_ASSERT();
934
935 if (m->m_flags & (M_BCAST|M_MCAST) || !in_canforward(ip->ip_dst)) {
936 IPSTAT_INC(ips_cantforward);
937 m_freem(m);
938 return;
939 }
940 if (
941 #ifdef IPSTEALTH
942 V_ipstealth == 0 &&
943 #endif
944 ip->ip_ttl <= IPTTLDEC) {
945 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
946 return;
947 }
948
949 bzero(&ro, sizeof(ro));
950 sin = (struct sockaddr_in *)&ro.ro_dst;
951 sin->sin_family = AF_INET;
952 sin->sin_len = sizeof(*sin);
953 sin->sin_addr = ip->ip_dst;
954 flowid = m->m_pkthdr.flowid;
955 ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
956 if (ro.ro_nh != NULL) {
957 ia = ifatoia(ro.ro_nh->nh_ifa);
958 } else
959 ia = NULL;
960 /*
961 * Save the IP header and at most 8 bytes of the payload,
962 * in case we need to generate an ICMP message to the src.
963 *
964 * XXX this can be optimized a lot by saving the data in a local
965 * buffer on the stack (72 bytes at most), and only allocating the
966 * mbuf if really necessary. The vast majority of the packets
967 * are forwarded without having to send an ICMP back (either
968 * because unnecessary, or because rate limited), so we are
969 * really we are wasting a lot of work here.
970 *
971 * We don't use m_copym() because it might return a reference
972 * to a shared cluster. Both this function and ip_output()
973 * assume exclusive access to the IP header in `m', so any
974 * data in a cluster may change before we reach icmp_error().
975 */
976 mcopy = m_gethdr(M_NOWAIT, m->m_type);
977 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
978 /*
979 * It's probably ok if the pkthdr dup fails (because
980 * the deep copy of the tag chain failed), but for now
981 * be conservative and just discard the copy since
982 * code below may some day want the tags.
983 */
984 m_free(mcopy);
985 mcopy = NULL;
986 }
987 if (mcopy != NULL) {
988 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
989 mcopy->m_pkthdr.len = mcopy->m_len;
990 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
991 }
992 #ifdef IPSTEALTH
993 if (V_ipstealth == 0)
994 #endif
995 ip->ip_ttl -= IPTTLDEC;
996 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
997 if (IPSEC_ENABLED(ipv4)) {
998 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
999 /* mbuf consumed by IPsec */
1000 RO_NHFREE(&ro);
1001 m_freem(mcopy);
1002 if (error != EINPROGRESS)
1003 IPSTAT_INC(ips_cantforward);
1004 return;
1005 }
1006 /* No IPsec processing required */
1007 }
1008 #endif /* IPSEC */
1009 /*
1010 * If forwarding packet using same interface that it came in on,
1011 * perhaps should send a redirect to sender to shortcut a hop.
1012 * Only send redirect if source is sending directly to us,
1013 * and if packet was not source routed (or has any options).
1014 * Also, don't send redirect if forwarding using a default route
1015 * or a route modified by a redirect.
1016 */
1017 dest.s_addr = 0;
1018 if (!srcrt && V_ipsendredirects &&
1019 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1020 struct nhop_object *nh;
1021
1022 nh = ro.ro_nh;
1023
1024 if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1025 struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1026 u_long src = ntohl(ip->ip_src.s_addr);
1027
1028 if (nh_ia != NULL &&
1029 (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1030 /* Router requirements says to only send host redirects */
1031 type = ICMP_REDIRECT;
1032 code = ICMP_REDIRECT_HOST;
1033 if (nh->nh_flags & NHF_GATEWAY) {
1034 if (nh->gw_sa.sa_family == AF_INET)
1035 dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1036 else /* Do not redirect in case gw is AF_INET6 */
1037 type = 0;
1038 } else
1039 dest.s_addr = ip->ip_dst.s_addr;
1040 }
1041 }
1042 }
1043
1044 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1045
1046 if (error == EMSGSIZE && ro.ro_nh)
1047 mtu = ro.ro_nh->nh_mtu;
1048 RO_NHFREE(&ro);
1049
1050 if (error)
1051 IPSTAT_INC(ips_cantforward);
1052 else {
1053 IPSTAT_INC(ips_forward);
1054 if (type)
1055 IPSTAT_INC(ips_redirectsent);
1056 else {
1057 if (mcopy)
1058 m_freem(mcopy);
1059 return;
1060 }
1061 }
1062 if (mcopy == NULL)
1063 return;
1064
1065 switch (error) {
1066 case 0: /* forwarded, but need redirect */
1067 /* type, code set above */
1068 break;
1069
1070 case ENETUNREACH:
1071 case EHOSTUNREACH:
1072 case ENETDOWN:
1073 case EHOSTDOWN:
1074 default:
1075 type = ICMP_UNREACH;
1076 code = ICMP_UNREACH_HOST;
1077 break;
1078
1079 case EMSGSIZE:
1080 type = ICMP_UNREACH;
1081 code = ICMP_UNREACH_NEEDFRAG;
1082 /*
1083 * If the MTU was set before make sure we are below the
1084 * interface MTU.
1085 * If the MTU wasn't set before use the interface mtu or
1086 * fall back to the next smaller mtu step compared to the
1087 * current packet size.
1088 */
1089 if (mtu != 0) {
1090 if (ia != NULL)
1091 mtu = min(mtu, ia->ia_ifp->if_mtu);
1092 } else {
1093 if (ia != NULL)
1094 mtu = ia->ia_ifp->if_mtu;
1095 else
1096 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1097 }
1098 IPSTAT_INC(ips_cantfrag);
1099 break;
1100
1101 case ENOBUFS:
1102 case EACCES: /* ipfw denied packet */
1103 m_freem(mcopy);
1104 return;
1105 }
1106 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1107 }
1108
1109 #define CHECK_SO_CT(sp, ct) \
1110 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1111
1112 void
ip_savecontrol(struct inpcb * inp,struct mbuf ** mp,struct ip * ip,struct mbuf * m)1113 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1114 struct mbuf *m)
1115 {
1116 bool stamped;
1117
1118 stamped = false;
1119 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1120 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1121 struct bintime boottimebin, bt;
1122 struct timespec ts1;
1123
1124 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1125 M_TSTMP)) {
1126 mbuf_tstmp2timespec(m, &ts1);
1127 timespec2bintime(&ts1, &bt);
1128 getboottimebin(&boottimebin);
1129 bintime_add(&bt, &boottimebin);
1130 } else {
1131 bintime(&bt);
1132 }
1133 *mp = sbcreatecontrol(&bt, sizeof(bt), SCM_BINTIME,
1134 SOL_SOCKET, M_NOWAIT);
1135 if (*mp != NULL) {
1136 mp = &(*mp)->m_next;
1137 stamped = true;
1138 }
1139 }
1140 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1141 struct bintime boottimebin, bt1;
1142 struct timespec ts1;
1143 struct timeval tv;
1144
1145 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1146 M_TSTMP)) {
1147 mbuf_tstmp2timespec(m, &ts1);
1148 timespec2bintime(&ts1, &bt1);
1149 getboottimebin(&boottimebin);
1150 bintime_add(&bt1, &boottimebin);
1151 bintime2timeval(&bt1, &tv);
1152 } else {
1153 microtime(&tv);
1154 }
1155 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), SCM_TIMESTAMP,
1156 SOL_SOCKET, M_NOWAIT);
1157 if (*mp != NULL) {
1158 mp = &(*mp)->m_next;
1159 stamped = true;
1160 }
1161 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1162 struct bintime boottimebin;
1163 struct timespec ts, ts1;
1164
1165 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1166 M_TSTMP)) {
1167 mbuf_tstmp2timespec(m, &ts);
1168 getboottimebin(&boottimebin);
1169 bintime2timespec(&boottimebin, &ts1);
1170 timespecadd(&ts, &ts1, &ts);
1171 } else {
1172 nanotime(&ts);
1173 }
1174 *mp = sbcreatecontrol(&ts, sizeof(ts), SCM_REALTIME,
1175 SOL_SOCKET, M_NOWAIT);
1176 if (*mp != NULL) {
1177 mp = &(*mp)->m_next;
1178 stamped = true;
1179 }
1180 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1181 struct timespec ts;
1182
1183 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1184 M_TSTMP))
1185 mbuf_tstmp2timespec(m, &ts);
1186 else
1187 nanouptime(&ts);
1188 *mp = sbcreatecontrol(&ts, sizeof(ts), SCM_MONOTONIC,
1189 SOL_SOCKET, M_NOWAIT);
1190 if (*mp != NULL) {
1191 mp = &(*mp)->m_next;
1192 stamped = true;
1193 }
1194 }
1195 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1196 M_TSTMP)) {
1197 struct sock_timestamp_info sti;
1198
1199 bzero(&sti, sizeof(sti));
1200 sti.st_info_flags = ST_INFO_HW;
1201 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1202 sti.st_info_flags |= ST_INFO_HW_HPREC;
1203 *mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1204 SOL_SOCKET, M_NOWAIT);
1205 if (*mp != NULL)
1206 mp = &(*mp)->m_next;
1207 }
1208 if (inp->inp_flags & INP_RECVDSTADDR) {
1209 *mp = sbcreatecontrol(&ip->ip_dst, sizeof(struct in_addr),
1210 IP_RECVDSTADDR, IPPROTO_IP, M_NOWAIT);
1211 if (*mp)
1212 mp = &(*mp)->m_next;
1213 }
1214 if (inp->inp_flags & INP_RECVTTL) {
1215 *mp = sbcreatecontrol(&ip->ip_ttl, sizeof(u_char), IP_RECVTTL,
1216 IPPROTO_IP, M_NOWAIT);
1217 if (*mp)
1218 mp = &(*mp)->m_next;
1219 }
1220 #ifdef notyet
1221 /* XXX
1222 * Moving these out of udp_input() made them even more broken
1223 * than they already were.
1224 */
1225 /* options were tossed already */
1226 if (inp->inp_flags & INP_RECVOPTS) {
1227 *mp = sbcreatecontrol(opts_deleted_above,
1228 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP, M_NOWAIT);
1229 if (*mp)
1230 mp = &(*mp)->m_next;
1231 }
1232 /* ip_srcroute doesn't do what we want here, need to fix */
1233 if (inp->inp_flags & INP_RECVRETOPTS) {
1234 *mp = sbcreatecontrol(ip_srcroute(m), sizeof(struct in_addr),
1235 IP_RECVRETOPTS, IPPROTO_IP, M_NOWAIT);
1236 if (*mp)
1237 mp = &(*mp)->m_next;
1238 }
1239 #endif
1240 if (inp->inp_flags & INP_RECVIF) {
1241 struct ifnet *ifp;
1242 struct sdlbuf {
1243 struct sockaddr_dl sdl;
1244 u_char pad[32];
1245 } sdlbuf;
1246 struct sockaddr_dl *sdp;
1247 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1248
1249 if ((ifp = m->m_pkthdr.rcvif)) {
1250 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1251 /*
1252 * Change our mind and don't try copy.
1253 */
1254 if (sdp->sdl_family != AF_LINK ||
1255 sdp->sdl_len > sizeof(sdlbuf)) {
1256 goto makedummy;
1257 }
1258 bcopy(sdp, sdl2, sdp->sdl_len);
1259 } else {
1260 makedummy:
1261 sdl2->sdl_len =
1262 offsetof(struct sockaddr_dl, sdl_data[0]);
1263 sdl2->sdl_family = AF_LINK;
1264 sdl2->sdl_index = 0;
1265 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1266 }
1267 *mp = sbcreatecontrol(sdl2, sdl2->sdl_len, IP_RECVIF,
1268 IPPROTO_IP, M_NOWAIT);
1269 if (*mp)
1270 mp = &(*mp)->m_next;
1271 }
1272 if (inp->inp_flags & INP_RECVTOS) {
1273 *mp = sbcreatecontrol(&ip->ip_tos, sizeof(u_char), IP_RECVTOS,
1274 IPPROTO_IP, M_NOWAIT);
1275 if (*mp)
1276 mp = &(*mp)->m_next;
1277 }
1278
1279 if (inp->inp_flags2 & INP_RECVFLOWID) {
1280 uint32_t flowid, flow_type;
1281
1282 flowid = m->m_pkthdr.flowid;
1283 flow_type = M_HASHTYPE_GET(m);
1284
1285 /*
1286 * XXX should handle the failure of one or the
1287 * other - don't populate both?
1288 */
1289 *mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IP_FLOWID,
1290 IPPROTO_IP, M_NOWAIT);
1291 if (*mp)
1292 mp = &(*mp)->m_next;
1293 *mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1294 IP_FLOWTYPE, IPPROTO_IP, M_NOWAIT);
1295 if (*mp)
1296 mp = &(*mp)->m_next;
1297 }
1298
1299 #ifdef RSS
1300 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1301 uint32_t flowid, flow_type;
1302 uint32_t rss_bucketid;
1303
1304 flowid = m->m_pkthdr.flowid;
1305 flow_type = M_HASHTYPE_GET(m);
1306
1307 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1308 *mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1309 IP_RSSBUCKETID, IPPROTO_IP, M_NOWAIT);
1310 if (*mp)
1311 mp = &(*mp)->m_next;
1312 }
1313 }
1314 #endif
1315 }
1316
1317 /*
1318 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1319 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1320 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1321 * compiled.
1322 */
1323 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1324 VNET_DEFINE(struct socket *, ip_rsvpd);
1325
1326 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1327
1328 int
ip_rsvp_init(struct socket * so)1329 ip_rsvp_init(struct socket *so)
1330 {
1331
1332 if (V_ip_rsvpd != NULL)
1333 return EADDRINUSE;
1334
1335 V_ip_rsvpd = so;
1336 /*
1337 * This may seem silly, but we need to be sure we don't over-increment
1338 * the RSVP counter, in case something slips up.
1339 */
1340 if (!V_ip_rsvp_on) {
1341 V_ip_rsvp_on = 1;
1342 V_rsvp_on++;
1343 }
1344
1345 return 0;
1346 }
1347
1348 int
ip_rsvp_done(void)1349 ip_rsvp_done(void)
1350 {
1351
1352 V_ip_rsvpd = NULL;
1353 /*
1354 * This may seem silly, but we need to be sure we don't over-decrement
1355 * the RSVP counter, in case something slips up.
1356 */
1357 if (V_ip_rsvp_on) {
1358 V_ip_rsvp_on = 0;
1359 V_rsvp_on--;
1360 }
1361 return 0;
1362 }
1363
1364 int
rsvp_input(struct mbuf ** mp,int * offp,int proto)1365 rsvp_input(struct mbuf **mp, int *offp, int proto)
1366 {
1367 struct mbuf *m;
1368
1369 m = *mp;
1370 *mp = NULL;
1371
1372 if (rsvp_input_p) { /* call the real one if loaded */
1373 *mp = m;
1374 rsvp_input_p(mp, offp, proto);
1375 return (IPPROTO_DONE);
1376 }
1377
1378 /* Can still get packets with rsvp_on = 0 if there is a local member
1379 * of the group to which the RSVP packet is addressed. But in this
1380 * case we want to throw the packet away.
1381 */
1382
1383 if (!V_rsvp_on) {
1384 m_freem(m);
1385 return (IPPROTO_DONE);
1386 }
1387
1388 if (V_ip_rsvpd != NULL) {
1389 *mp = m;
1390 rip_input(mp, offp, proto);
1391 return (IPPROTO_DONE);
1392 }
1393 /* Drop the packet */
1394 m_freem(m);
1395 return (IPPROTO_DONE);
1396 }
1397