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
32 #include "opt_bootp.h"
33 #include "opt_inet.h"
34 #include "opt_ipstealth.h"
35 #include "opt_ipsec.h"
36 #include "opt_route.h"
37 #include "opt_rss.h"
38 #include "opt_sctp.h"
39
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/hhook.h>
43 #include <sys/mbuf.h>
44 #include <sys/malloc.h>
45 #include <sys/domain.h>
46 #include <sys/protosw.h>
47 #include <sys/socket.h>
48 #include <sys/time.h>
49 #include <sys/kernel.h>
50 #include <sys/lock.h>
51 #include <sys/rmlock.h>
52 #include <sys/rwlock.h>
53 #include <sys/sdt.h>
54 #include <sys/syslog.h>
55 #include <sys/sysctl.h>
56
57 #include <net/if.h>
58 #include <net/if_types.h>
59 #include <net/if_var.h>
60 #include <net/if_dl.h>
61 #include <net/if_private.h>
62 #include <net/pfil.h>
63 #include <net/route.h>
64 #include <net/route/nhop.h>
65 #include <net/netisr.h>
66 #include <net/rss_config.h>
67 #include <net/vnet.h>
68
69 #include <netinet/in.h>
70 #include <netinet/in_kdtrace.h>
71 #include <netinet/in_systm.h>
72 #include <netinet/in_var.h>
73 #include <netinet/ip.h>
74 #include <netinet/in_fib.h>
75 #include <netinet/in_pcb.h>
76 #include <netinet/ip_var.h>
77 #include <netinet/ip_encap.h>
78 #include <netinet/ip_fw.h>
79 #include <netinet/ip_icmp.h>
80 #include <netinet/igmp_var.h>
81 #include <netinet/ip_options.h>
82 #include <machine/in_cksum.h>
83 #include <netinet/ip_carp.h>
84 #include <netinet/in_rss.h>
85 #ifdef SCTP
86 #include <netinet/sctp_var.h>
87 #endif
88
89 #include <netipsec/ipsec_support.h>
90
91 #include <sys/socketvar.h>
92
93 #include <security/mac/mac_framework.h>
94
95 #ifdef CTASSERT
96 CTASSERT(sizeof(struct ip) == 20);
97 #endif
98
99 /* IP reassembly functions are defined in ip_reass.c. */
100 extern void ipreass_init(void);
101 extern void ipreass_vnet_init(void);
102 #ifdef VIMAGE
103 extern void ipreass_destroy(void);
104 #endif
105
106 VNET_DEFINE(int, rsvp_on);
107
108 VNET_DEFINE(int, ipforwarding);
109 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
110 &VNET_NAME(ipforwarding), 0,
111 "Enable IP forwarding between interfaces");
112
113 /*
114 * Respond with an ICMP host redirect when we forward a packet out of
115 * the same interface on which it was received. See RFC 792.
116 */
117 VNET_DEFINE(int, ipsendredirects) = 1;
118 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
119 &VNET_NAME(ipsendredirects), 0,
120 "Enable sending IP redirects");
121
122 VNET_DEFINE_STATIC(bool, ip_strong_es) = false;
123 #define V_ip_strong_es VNET(ip_strong_es)
124 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, rfc1122_strong_es,
125 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_strong_es), false,
126 "Packet's IP destination address must match address on arrival interface");
127
128 VNET_DEFINE_STATIC(bool, ip_sav) = true;
129 #define V_ip_sav VNET(ip_sav)
130 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, source_address_validation,
131 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_sav), true,
132 "Drop incoming packets with source address that is a local address");
133
134 /* Packet filter hooks */
135 VNET_DEFINE(pfil_head_t, inet_pfil_head);
136 VNET_DEFINE(pfil_head_t, inet_local_pfil_head);
137
138 static struct netisr_handler ip_nh = {
139 .nh_name = "ip",
140 .nh_handler = ip_input,
141 .nh_proto = NETISR_IP,
142 #ifdef RSS
143 .nh_m2cpuid = rss_soft_m2cpuid_v4,
144 .nh_policy = NETISR_POLICY_CPU,
145 .nh_dispatch = NETISR_DISPATCH_HYBRID,
146 #else
147 .nh_policy = NETISR_POLICY_FLOW,
148 #endif
149 };
150
151 #ifdef RSS
152 /*
153 * Directly dispatched frames are currently assumed
154 * to have a flowid already calculated.
155 *
156 * It should likely have something that assert it
157 * actually has valid flow details.
158 */
159 static struct netisr_handler ip_direct_nh = {
160 .nh_name = "ip_direct",
161 .nh_handler = ip_direct_input,
162 .nh_proto = NETISR_IP_DIRECT,
163 .nh_m2cpuid = rss_soft_m2cpuid_v4,
164 .nh_policy = NETISR_POLICY_CPU,
165 .nh_dispatch = NETISR_DISPATCH_HYBRID,
166 };
167 #endif
168
169 ipproto_input_t *ip_protox[IPPROTO_MAX] = {
170 [0 ... IPPROTO_MAX - 1] = rip_input };
171 ipproto_ctlinput_t *ip_ctlprotox[IPPROTO_MAX] = {
172 [0 ... IPPROTO_MAX - 1] = rip_ctlinput };
173
174 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
175 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
176 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
177
178 /* Make sure it is safe to use hashinit(9) on CK_LIST. */
179 CTASSERT(sizeof(struct in_ifaddrhashhead) == sizeof(LIST_HEAD(, in_addr)));
180
181 #ifdef IPCTL_DEFMTU
182 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
183 &ip_mtu, 0, "Default MTU");
184 #endif
185
186 #ifdef IPSTEALTH
187 VNET_DEFINE(int, ipstealth);
188 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
189 &VNET_NAME(ipstealth), 0,
190 "IP stealth mode, no TTL decrementation on forwarding");
191 #endif
192
193 /*
194 * IP statistics are stored in the "array" of counter(9)s.
195 */
196 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
197 VNET_PCPUSTAT_SYSINIT(ipstat);
198 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
199 "IP statistics (struct ipstat, netinet/ip_var.h)");
200
201 #ifdef VIMAGE
202 VNET_PCPUSTAT_SYSUNINIT(ipstat);
203 #endif /* VIMAGE */
204
205 /*
206 * Kernel module interface for updating ipstat. The argument is an index
207 * into ipstat treated as an array.
208 */
209 void
kmod_ipstat_inc(int statnum)210 kmod_ipstat_inc(int statnum)
211 {
212
213 counter_u64_add(VNET(ipstat)[statnum], 1);
214 }
215
216 void
kmod_ipstat_dec(int statnum)217 kmod_ipstat_dec(int statnum)
218 {
219
220 counter_u64_add(VNET(ipstat)[statnum], -1);
221 }
222
223 static int
sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)224 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
225 {
226 int error, qlimit;
227
228 netisr_getqlimit(&ip_nh, &qlimit);
229 error = sysctl_handle_int(oidp, &qlimit, 0, req);
230 if (error || !req->newptr)
231 return (error);
232 if (qlimit < 1)
233 return (EINVAL);
234 return (netisr_setqlimit(&ip_nh, qlimit));
235 }
236 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
237 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
238 sysctl_netinet_intr_queue_maxlen, "I",
239 "Maximum size of the IP input queue");
240
241 static int
sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)242 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
243 {
244 u_int64_t qdrops_long;
245 int error, qdrops;
246
247 netisr_getqdrops(&ip_nh, &qdrops_long);
248 qdrops = qdrops_long;
249 error = sysctl_handle_int(oidp, &qdrops, 0, req);
250 if (error || !req->newptr)
251 return (error);
252 if (qdrops != 0)
253 return (EINVAL);
254 netisr_clearqdrops(&ip_nh);
255 return (0);
256 }
257
258 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
259 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
260 0, 0, sysctl_netinet_intr_queue_drops, "I",
261 "Number of packets dropped from the IP input queue");
262
263 #ifdef RSS
264 static int
sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)265 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
266 {
267 int error, qlimit;
268
269 netisr_getqlimit(&ip_direct_nh, &qlimit);
270 error = sysctl_handle_int(oidp, &qlimit, 0, req);
271 if (error || !req->newptr)
272 return (error);
273 if (qlimit < 1)
274 return (EINVAL);
275 return (netisr_setqlimit(&ip_direct_nh, qlimit));
276 }
277 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
278 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
279 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
280 "I", "Maximum size of the IP direct input queue");
281
282 static int
sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)283 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
284 {
285 u_int64_t qdrops_long;
286 int error, qdrops;
287
288 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
289 qdrops = qdrops_long;
290 error = sysctl_handle_int(oidp, &qdrops, 0, req);
291 if (error || !req->newptr)
292 return (error);
293 if (qdrops != 0)
294 return (EINVAL);
295 netisr_clearqdrops(&ip_direct_nh);
296 return (0);
297 }
298
299 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
300 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
301 sysctl_netinet_intr_direct_queue_drops, "I",
302 "Number of packets dropped from the IP direct input queue");
303 #endif /* RSS */
304
305 /*
306 * IP initialization: fill in IP protocol switch table.
307 * All protocols not implemented in kernel go to raw IP protocol handler.
308 */
309 static void
ip_vnet_init(void * arg __unused)310 ip_vnet_init(void *arg __unused)
311 {
312 struct pfil_head_args args;
313
314 CK_STAILQ_INIT(&V_in_ifaddrhead);
315 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
316
317 /* Initialize IP reassembly queue. */
318 ipreass_vnet_init();
319
320 /* Initialize packet filter hooks. */
321 args.pa_version = PFIL_VERSION;
322 args.pa_flags = PFIL_IN | PFIL_OUT;
323 args.pa_type = PFIL_TYPE_IP4;
324 args.pa_headname = PFIL_INET_NAME;
325 V_inet_pfil_head = pfil_head_register(&args);
326
327 args.pa_flags = PFIL_OUT;
328 args.pa_headname = PFIL_INET_LOCAL_NAME;
329 V_inet_local_pfil_head = pfil_head_register(&args);
330
331 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
332 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
333 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
334 printf("%s: WARNING: unable to register input helper hook\n",
335 __func__);
336 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
337 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
338 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
339 printf("%s: WARNING: unable to register output helper hook\n",
340 __func__);
341
342 #ifdef VIMAGE
343 netisr_register_vnet(&ip_nh);
344 #ifdef RSS
345 netisr_register_vnet(&ip_direct_nh);
346 #endif
347 #endif
348 }
349 VNET_SYSINIT(ip_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
350 ip_vnet_init, NULL);
351
352 static void
ip_init(const void * unused __unused)353 ip_init(const void *unused __unused)
354 {
355 struct ifnet *ifp;
356
357 ipreass_init();
358
359 /*
360 * Register statically compiled protocols, that are unlikely to
361 * ever become dynamic.
362 */
363 IPPROTO_REGISTER(IPPROTO_ICMP, icmp_input, NULL);
364 IPPROTO_REGISTER(IPPROTO_IGMP, igmp_input, NULL);
365 IPPROTO_REGISTER(IPPROTO_RSVP, rsvp_input, NULL);
366 IPPROTO_REGISTER(IPPROTO_IPV4, encap4_input, NULL);
367 IPPROTO_REGISTER(IPPROTO_MOBILE, encap4_input, NULL);
368 IPPROTO_REGISTER(IPPROTO_ETHERIP, encap4_input, NULL);
369 IPPROTO_REGISTER(IPPROTO_GRE, encap4_input, NULL);
370 IPPROTO_REGISTER(IPPROTO_IPV6, encap4_input, NULL);
371 IPPROTO_REGISTER(IPPROTO_PIM, encap4_input, NULL);
372 #ifdef SCTP /* XXX: has a loadable & static version */
373 IPPROTO_REGISTER(IPPROTO_SCTP, sctp_input, sctp_ctlinput);
374 #endif
375
376 netisr_register(&ip_nh);
377 #ifdef RSS
378 netisr_register(&ip_direct_nh);
379 #endif
380 /*
381 * XXXGL: we use SYSINIT() here, but go over V_ifnet. It was the same
382 * way before dom_ifattach removal. This worked because when any
383 * non-default vnet is created, there are no interfaces inside.
384 * Eventually this needs to be fixed.
385 */
386 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
387 in_ifattach(NULL, ifp);
388 }
389 SYSINIT(ip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_init, NULL);
390
391 #ifdef VIMAGE
392 static void
ip_destroy(void * unused __unused)393 ip_destroy(void *unused __unused)
394 {
395 int error;
396
397 #ifdef RSS
398 netisr_unregister_vnet(&ip_direct_nh);
399 #endif
400 netisr_unregister_vnet(&ip_nh);
401
402 pfil_head_unregister(V_inet_pfil_head);
403 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
404 if (error != 0) {
405 printf("%s: WARNING: unable to deregister input helper hook "
406 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
407 "error %d returned\n", __func__, error);
408 }
409 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
410 if (error != 0) {
411 printf("%s: WARNING: unable to deregister output helper hook "
412 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
413 "error %d returned\n", __func__, error);
414 }
415
416 /* Remove the IPv4 addresses from all interfaces. */
417 in_ifscrub_all();
418
419 /* Make sure the IPv4 routes are gone as well. */
420 rib_flush_routes_family(AF_INET);
421
422 /* Destroy IP reassembly queue. */
423 ipreass_destroy();
424
425 /* Cleanup in_ifaddr hash table; should be empty. */
426 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
427 }
428
429 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
430 #endif
431
432 #ifdef RSS
433 /*
434 * IP direct input routine.
435 *
436 * This is called when reinjecting completed fragments where
437 * all of the previous checking and book-keeping has been done.
438 */
439 void
ip_direct_input(struct mbuf * m)440 ip_direct_input(struct mbuf *m)
441 {
442 struct ip *ip;
443 int hlen;
444
445 ip = mtod(m, struct ip *);
446 hlen = ip->ip_hl << 2;
447
448 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
449 if (IPSEC_ENABLED(ipv4)) {
450 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
451 return;
452 }
453 #endif /* IPSEC */
454 IPSTAT_INC(ips_delivered);
455 ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
456 }
457 #endif
458
459 /*
460 * Ip input routine. Checksum and byte swap header. If fragmented
461 * try to reassemble. Process options. Pass to next level.
462 */
463 void
ip_input(struct mbuf * m)464 ip_input(struct mbuf *m)
465 {
466 struct ip *ip = NULL;
467 struct in_ifaddr *ia = NULL;
468 struct ifaddr *ifa;
469 struct ifnet *ifp;
470 int hlen = 0;
471 uint16_t sum, ip_len;
472 int dchg = 0; /* dest changed after fw */
473 struct in_addr odst; /* original dst address */
474 bool strong_es;
475
476 M_ASSERTPKTHDR(m);
477 NET_EPOCH_ASSERT();
478
479 if (m->m_flags & M_FASTFWD_OURS) {
480 m->m_flags &= ~M_FASTFWD_OURS;
481 /* Set up some basics that will be used later. */
482 ip = mtod(m, struct ip *);
483 hlen = ip->ip_hl << 2;
484 ip_len = ntohs(ip->ip_len);
485 goto ours;
486 }
487
488 IPSTAT_INC(ips_total);
489
490 if (__predict_false(m->m_pkthdr.len < sizeof(struct ip)))
491 goto tooshort;
492
493 if (m->m_len < sizeof(struct ip)) {
494 m = m_pullup(m, sizeof(struct ip));
495 if (__predict_false(m == NULL)) {
496 IPSTAT_INC(ips_toosmall);
497 return;
498 }
499 }
500 ip = mtod(m, struct ip *);
501
502 if (__predict_false(ip->ip_v != IPVERSION)) {
503 IPSTAT_INC(ips_badvers);
504 goto bad;
505 }
506
507 hlen = ip->ip_hl << 2;
508 if (__predict_false(hlen < sizeof(struct ip))) { /* minimum header length */
509 IPSTAT_INC(ips_badhlen);
510 goto bad;
511 }
512 if (hlen > m->m_len) {
513 m = m_pullup(m, hlen);
514 if (__predict_false(m == NULL)) {
515 IPSTAT_INC(ips_badhlen);
516 return;
517 }
518 ip = mtod(m, struct ip *);
519 }
520
521 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
522
523 /* IN_LOOPBACK must not appear on the wire - RFC1122 */
524 ifp = m->m_pkthdr.rcvif;
525 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
526 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
527 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
528 IPSTAT_INC(ips_badaddr);
529 goto bad;
530 }
531 }
532
533 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
534 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
535 } else {
536 if (hlen == sizeof(struct ip)) {
537 sum = in_cksum_hdr(ip);
538 } else {
539 sum = in_cksum(m, hlen);
540 }
541 }
542 if (__predict_false(sum)) {
543 IPSTAT_INC(ips_badsum);
544 goto bad;
545 }
546
547 ip_len = ntohs(ip->ip_len);
548 if (__predict_false(ip_len < hlen)) {
549 IPSTAT_INC(ips_badlen);
550 goto bad;
551 }
552
553 /*
554 * Check that the amount of data in the buffers
555 * is as at least much as the IP header would have us expect.
556 * Trim mbufs if longer than we expect.
557 * Drop packet if shorter than we expect.
558 */
559 if (__predict_false(m->m_pkthdr.len < ip_len)) {
560 tooshort:
561 IPSTAT_INC(ips_tooshort);
562 goto bad;
563 }
564 if (m->m_pkthdr.len > ip_len) {
565 if (m->m_len == m->m_pkthdr.len) {
566 m->m_len = ip_len;
567 m->m_pkthdr.len = ip_len;
568 } else
569 m_adj(m, ip_len - m->m_pkthdr.len);
570 }
571
572 /*
573 * Try to forward the packet, but if we fail continue.
574 * ip_tryforward() may generate redirects these days.
575 * XXX the logic below falling through to normal processing
576 * if redirects are required should be revisited as well.
577 * ip_tryforward() does inbound and outbound packet firewall
578 * processing. If firewall has decided that destination becomes
579 * our local address, it sets M_FASTFWD_OURS flag. In this
580 * case skip another inbound firewall processing and update
581 * ip pointer.
582 */
583 if (V_ipforwarding != 0
584 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
585 && (!IPSEC_ENABLED(ipv4) ||
586 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
587 #endif
588 ) {
589 /*
590 * ip_dooptions() was run so we can ignore the source route (or
591 * any IP options case) case for redirects in ip_tryforward().
592 */
593 if ((m = ip_tryforward(m)) == NULL)
594 return;
595 if (m->m_flags & M_FASTFWD_OURS) {
596 m->m_flags &= ~M_FASTFWD_OURS;
597 ip = mtod(m, struct ip *);
598 goto ours;
599 }
600 }
601
602 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
603 /*
604 * Bypass packet filtering for packets previously handled by IPsec.
605 */
606 if (IPSEC_ENABLED(ipv4) &&
607 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
608 goto passin;
609 #endif
610
611 /*
612 * Run through list of hooks for input packets.
613 *
614 * NB: Beware of the destination address changing (e.g.
615 * by NAT rewriting). When this happens, tell
616 * ip_forward to do the right thing.
617 */
618
619 /* Jump over all PFIL processing if hooks are not active. */
620 if (!PFIL_HOOKED_IN(V_inet_pfil_head))
621 goto passin;
622
623 odst = ip->ip_dst;
624 if (pfil_mbuf_in(V_inet_pfil_head, &m, ifp, NULL) !=
625 PFIL_PASS)
626 return;
627
628 ip = mtod(m, struct ip *);
629 dchg = (odst.s_addr != ip->ip_dst.s_addr);
630
631 if (m->m_flags & M_FASTFWD_OURS) {
632 m->m_flags &= ~M_FASTFWD_OURS;
633 goto ours;
634 }
635 if (m->m_flags & M_IP_NEXTHOP) {
636 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
637 /*
638 * Directly ship the packet on. This allows
639 * forwarding packets originally destined to us
640 * to some other directly connected host.
641 */
642 ip_forward(m, 1);
643 return;
644 }
645 }
646 passin:
647 /*
648 * The unspecified address can appear only as a src address - RFC1122.
649 *
650 * The check is deferred to here to give firewalls a chance to block
651 * (and log) such packets. ip_tryforward() will not process such
652 * packets.
653 */
654 if (__predict_false(ntohl(ip->ip_dst.s_addr) == INADDR_ANY)) {
655 IPSTAT_INC(ips_badaddr);
656 goto bad;
657 }
658
659 /*
660 * Process options and, if not destined for us,
661 * ship it on. ip_dooptions returns 1 when an
662 * error was detected (causing an icmp message
663 * to be sent and the original packet to be freed).
664 */
665 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
666 return;
667
668 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
669 * matter if it is destined to another node, or whether it is
670 * a multicast one, RSVP wants it! and prevents it from being forwarded
671 * anywhere else. Also checks if the rsvp daemon is running before
672 * grabbing the packet.
673 */
674 if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
675 goto ours;
676
677 /*
678 * Check our list of addresses, to see if the packet is for us.
679 * If we don't have any addresses, assume any unicast packet
680 * we receive might be for us (and let the upper layers deal
681 * with it).
682 */
683 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
684 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
685 goto ours;
686
687 /*
688 * Enable a consistency check between the destination address
689 * and the arrival interface for a unicast packet (the RFC 1122
690 * strong ES model) with a list of additional predicates:
691 * - if IP forwarding is disabled
692 * - the packet is not locally generated
693 * - the packet is not subject to 'ipfw fwd'
694 * - Interface is not running CARP. If the packet got here, we already
695 * checked it with carp_iamatch() and carp_forus().
696 */
697 strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
698 ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
699 ifp->if_carp == NULL && (dchg == 0);
700
701 /*
702 * Check for exact addresses in the hash bucket.
703 */
704 CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
705 if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
706 continue;
707
708 /*
709 * net.inet.ip.rfc1122_strong_es: the address matches, verify
710 * that the packet arrived via the correct interface.
711 */
712 if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
713 IPSTAT_INC(ips_badaddr);
714 goto bad;
715 }
716
717 /*
718 * net.inet.ip.source_address_validation: drop incoming
719 * packets that pretend to be ours.
720 */
721 if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
722 __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
723 IPSTAT_INC(ips_badaddr);
724 goto bad;
725 }
726
727 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
728 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
729 goto ours;
730 }
731
732 /*
733 * Check for broadcast addresses.
734 *
735 * Only accept broadcast packets that arrive via the matching
736 * interface. Reception of forwarded directed broadcasts would
737 * be handled via ip_forward() and ether_output() with the loopback
738 * into the stack for SIMPLEX interfaces handled by ether_output().
739 */
740 if (ifp->if_flags & IFF_BROADCAST) {
741 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
742 if (ifa->ifa_addr->sa_family != AF_INET)
743 continue;
744 ia = ifatoia(ifa);
745 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
746 ip->ip_dst.s_addr) {
747 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
748 counter_u64_add(ia->ia_ifa.ifa_ibytes,
749 m->m_pkthdr.len);
750 goto ours;
751 }
752 #ifdef BOOTP_COMPAT
753 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
754 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
755 counter_u64_add(ia->ia_ifa.ifa_ibytes,
756 m->m_pkthdr.len);
757 goto ours;
758 }
759 #endif
760 }
761 ia = NULL;
762 }
763 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
764 /*
765 * RFC 3927 2.7: Do not forward multicast packets from
766 * IN_LINKLOCAL.
767 */
768 if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
769 /*
770 * If we are acting as a multicast router, all
771 * incoming multicast packets are passed to the
772 * kernel-level multicast forwarding function.
773 * The packet is returned (relatively) intact; if
774 * ip_mforward() returns a non-zero value, the packet
775 * must be discarded, else it may be accepted below.
776 */
777 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
778 IPSTAT_INC(ips_cantforward);
779 m_freem(m);
780 return;
781 }
782
783 /*
784 * The process-level routing daemon needs to receive
785 * all multicast IGMP packets, whether or not this
786 * host belongs to their destination groups.
787 */
788 if (ip->ip_p == IPPROTO_IGMP) {
789 goto ours;
790 }
791 IPSTAT_INC(ips_forward);
792 }
793 /*
794 * Assume the packet is for us, to avoid prematurely taking
795 * a lock on the in_multi hash. Protocols must perform
796 * their own filtering and update statistics accordingly.
797 */
798 goto ours;
799 }
800 if (in_broadcast(ip->ip_dst))
801 goto ours;
802 /* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
803 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
804 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
805 IPSTAT_INC(ips_cantforward);
806 m_freem(m);
807 return;
808 }
809
810 /*
811 * Not for us; forward if possible and desirable.
812 */
813 if (V_ipforwarding == 0) {
814 IPSTAT_INC(ips_cantforward);
815 m_freem(m);
816 } else {
817 ip_forward(m, dchg);
818 }
819 return;
820
821 ours:
822 #ifdef IPSTEALTH
823 /*
824 * IPSTEALTH: Process non-routing options only
825 * if the packet is destined for us.
826 */
827 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
828 return;
829 #endif /* IPSTEALTH */
830
831 /*
832 * We are going to ship the packet to the local protocol stack. Call the
833 * filter again for this 'output' action, allowing redirect-like rules
834 * to adjust the source address.
835 */
836 if (PFIL_HOOKED_OUT(V_inet_local_pfil_head)) {
837 if (pfil_mbuf_out(V_inet_local_pfil_head, &m, V_loif, NULL) !=
838 PFIL_PASS)
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 if (ro.ro_nh->nh_flags & (NHF_BLACKHOLE | NHF_BROADCAST)) {
958 IPSTAT_INC(ips_cantforward);
959 m_freem(m);
960 NH_FREE(ro.ro_nh);
961 return;
962 }
963 if (ro.ro_nh->nh_flags & NHF_REJECT) {
964 IPSTAT_INC(ips_cantforward);
965 NH_FREE(ro.ro_nh);
966 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
967 return;
968 }
969 ia = ifatoia(ro.ro_nh->nh_ifa);
970 } else
971 ia = NULL;
972 /*
973 * Save the IP header and at most 8 bytes of the payload,
974 * in case we need to generate an ICMP message to the src.
975 *
976 * XXX this can be optimized a lot by saving the data in a local
977 * buffer on the stack (72 bytes at most), and only allocating the
978 * mbuf if really necessary. The vast majority of the packets
979 * are forwarded without having to send an ICMP back (either
980 * because unnecessary, or because rate limited), so we are
981 * really we are wasting a lot of work here.
982 *
983 * We don't use m_copym() because it might return a reference
984 * to a shared cluster. Both this function and ip_output()
985 * assume exclusive access to the IP header in `m', so any
986 * data in a cluster may change before we reach icmp_error().
987 */
988 mcopy = m_gethdr(M_NOWAIT, m->m_type);
989 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
990 /*
991 * It's probably ok if the pkthdr dup fails (because
992 * the deep copy of the tag chain failed), but for now
993 * be conservative and just discard the copy since
994 * code below may some day want the tags.
995 */
996 m_free(mcopy);
997 mcopy = NULL;
998 }
999 if (mcopy != NULL) {
1000 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1001 mcopy->m_pkthdr.len = mcopy->m_len;
1002 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1003 }
1004 #ifdef IPSTEALTH
1005 if (V_ipstealth == 0)
1006 #endif
1007 ip->ip_ttl -= IPTTLDEC;
1008 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1009 if (IPSEC_ENABLED(ipv4)) {
1010 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1011 /* mbuf consumed by IPsec */
1012 RO_NHFREE(&ro);
1013 m_freem(mcopy);
1014 if (error != EINPROGRESS)
1015 IPSTAT_INC(ips_cantforward);
1016 return;
1017 }
1018 /* No IPsec processing required */
1019 }
1020 #endif /* IPSEC */
1021 /*
1022 * If forwarding packet using same interface that it came in on,
1023 * perhaps should send a redirect to sender to shortcut a hop.
1024 * Only send redirect if source is sending directly to us,
1025 * and if packet was not source routed (or has any options).
1026 * Also, don't send redirect if forwarding using a default route
1027 * or a route modified by a redirect.
1028 */
1029 dest.s_addr = 0;
1030 if (!srcrt && V_ipsendredirects &&
1031 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1032 struct nhop_object *nh;
1033
1034 nh = ro.ro_nh;
1035
1036 if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1037 struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1038 u_long src = ntohl(ip->ip_src.s_addr);
1039
1040 if (nh_ia != NULL &&
1041 (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1042 /* Router requirements says to only send host redirects */
1043 type = ICMP_REDIRECT;
1044 code = ICMP_REDIRECT_HOST;
1045 if (nh->nh_flags & NHF_GATEWAY) {
1046 if (nh->gw_sa.sa_family == AF_INET)
1047 dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1048 else /* Do not redirect in case gw is AF_INET6 */
1049 type = 0;
1050 } else
1051 dest.s_addr = ip->ip_dst.s_addr;
1052 }
1053 }
1054 }
1055
1056 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1057
1058 if (error == EMSGSIZE && ro.ro_nh)
1059 mtu = ro.ro_nh->nh_mtu;
1060 RO_NHFREE(&ro);
1061
1062 if (error)
1063 IPSTAT_INC(ips_cantforward);
1064 else {
1065 IPSTAT_INC(ips_forward);
1066 if (type)
1067 IPSTAT_INC(ips_redirectsent);
1068 else {
1069 if (mcopy)
1070 m_freem(mcopy);
1071 return;
1072 }
1073 }
1074 if (mcopy == NULL)
1075 return;
1076
1077 switch (error) {
1078 case 0: /* forwarded, but need redirect */
1079 /* type, code set above */
1080 break;
1081
1082 case ENETUNREACH:
1083 case EHOSTUNREACH:
1084 case ENETDOWN:
1085 case EHOSTDOWN:
1086 default:
1087 type = ICMP_UNREACH;
1088 code = ICMP_UNREACH_HOST;
1089 break;
1090
1091 case EMSGSIZE:
1092 type = ICMP_UNREACH;
1093 code = ICMP_UNREACH_NEEDFRAG;
1094 /*
1095 * If the MTU was set before make sure we are below the
1096 * interface MTU.
1097 * If the MTU wasn't set before use the interface mtu or
1098 * fall back to the next smaller mtu step compared to the
1099 * current packet size.
1100 */
1101 if (mtu != 0) {
1102 if (ia != NULL)
1103 mtu = min(mtu, ia->ia_ifp->if_mtu);
1104 } else {
1105 if (ia != NULL)
1106 mtu = ia->ia_ifp->if_mtu;
1107 else
1108 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1109 }
1110 IPSTAT_INC(ips_cantfrag);
1111 break;
1112
1113 case ENOBUFS:
1114 case EACCES: /* ipfw denied packet */
1115 m_freem(mcopy);
1116 return;
1117 }
1118 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1119 }
1120
1121 #define CHECK_SO_CT(sp, ct) \
1122 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1123
1124 void
ip_savecontrol(struct inpcb * inp,struct mbuf ** mp,struct ip * ip,struct mbuf * m)1125 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1126 struct mbuf *m)
1127 {
1128 bool stamped;
1129
1130 stamped = false;
1131 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1132 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1133 struct bintime boottimebin, bt;
1134 struct timespec ts1;
1135
1136 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1137 M_TSTMP)) {
1138 mbuf_tstmp2timespec(m, &ts1);
1139 timespec2bintime(&ts1, &bt);
1140 getboottimebin(&boottimebin);
1141 bintime_add(&bt, &boottimebin);
1142 } else {
1143 bintime(&bt);
1144 }
1145 *mp = sbcreatecontrol(&bt, sizeof(bt), SCM_BINTIME,
1146 SOL_SOCKET, M_NOWAIT);
1147 if (*mp != NULL) {
1148 mp = &(*mp)->m_next;
1149 stamped = true;
1150 }
1151 }
1152 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1153 struct bintime boottimebin, bt1;
1154 struct timespec ts1;
1155 struct timeval tv;
1156
1157 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1158 M_TSTMP)) {
1159 mbuf_tstmp2timespec(m, &ts1);
1160 timespec2bintime(&ts1, &bt1);
1161 getboottimebin(&boottimebin);
1162 bintime_add(&bt1, &boottimebin);
1163 bintime2timeval(&bt1, &tv);
1164 } else {
1165 microtime(&tv);
1166 }
1167 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), SCM_TIMESTAMP,
1168 SOL_SOCKET, M_NOWAIT);
1169 if (*mp != NULL) {
1170 mp = &(*mp)->m_next;
1171 stamped = true;
1172 }
1173 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1174 struct bintime boottimebin;
1175 struct timespec ts, ts1;
1176
1177 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1178 M_TSTMP)) {
1179 mbuf_tstmp2timespec(m, &ts);
1180 getboottimebin(&boottimebin);
1181 bintime2timespec(&boottimebin, &ts1);
1182 timespecadd(&ts, &ts1, &ts);
1183 } else {
1184 nanotime(&ts);
1185 }
1186 *mp = sbcreatecontrol(&ts, sizeof(ts), SCM_REALTIME,
1187 SOL_SOCKET, M_NOWAIT);
1188 if (*mp != NULL) {
1189 mp = &(*mp)->m_next;
1190 stamped = true;
1191 }
1192 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1193 struct timespec ts;
1194
1195 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1196 M_TSTMP))
1197 mbuf_tstmp2timespec(m, &ts);
1198 else
1199 nanouptime(&ts);
1200 *mp = sbcreatecontrol(&ts, sizeof(ts), SCM_MONOTONIC,
1201 SOL_SOCKET, M_NOWAIT);
1202 if (*mp != NULL) {
1203 mp = &(*mp)->m_next;
1204 stamped = true;
1205 }
1206 }
1207 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1208 M_TSTMP)) {
1209 struct sock_timestamp_info sti;
1210
1211 bzero(&sti, sizeof(sti));
1212 sti.st_info_flags = ST_INFO_HW;
1213 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1214 sti.st_info_flags |= ST_INFO_HW_HPREC;
1215 *mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1216 SOL_SOCKET, M_NOWAIT);
1217 if (*mp != NULL)
1218 mp = &(*mp)->m_next;
1219 }
1220 if (inp->inp_flags & INP_RECVDSTADDR) {
1221 *mp = sbcreatecontrol(&ip->ip_dst, sizeof(struct in_addr),
1222 IP_RECVDSTADDR, IPPROTO_IP, M_NOWAIT);
1223 if (*mp)
1224 mp = &(*mp)->m_next;
1225 }
1226 if (inp->inp_flags & INP_RECVTTL) {
1227 *mp = sbcreatecontrol(&ip->ip_ttl, sizeof(u_char), IP_RECVTTL,
1228 IPPROTO_IP, M_NOWAIT);
1229 if (*mp)
1230 mp = &(*mp)->m_next;
1231 }
1232 #ifdef notyet
1233 /* XXX
1234 * Moving these out of udp_input() made them even more broken
1235 * than they already were.
1236 */
1237 /* options were tossed already */
1238 if (inp->inp_flags & INP_RECVOPTS) {
1239 *mp = sbcreatecontrol(opts_deleted_above,
1240 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP, M_NOWAIT);
1241 if (*mp)
1242 mp = &(*mp)->m_next;
1243 }
1244 /* ip_srcroute doesn't do what we want here, need to fix */
1245 if (inp->inp_flags & INP_RECVRETOPTS) {
1246 *mp = sbcreatecontrol(ip_srcroute(m), sizeof(struct in_addr),
1247 IP_RECVRETOPTS, IPPROTO_IP, M_NOWAIT);
1248 if (*mp)
1249 mp = &(*mp)->m_next;
1250 }
1251 #endif
1252 if (inp->inp_flags & INP_RECVIF) {
1253 struct ifnet *ifp;
1254 struct sdlbuf {
1255 struct sockaddr_dl sdl;
1256 u_char pad[32];
1257 } sdlbuf;
1258 struct sockaddr_dl *sdp;
1259 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1260
1261 if ((ifp = m->m_pkthdr.rcvif)) {
1262 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1263 /*
1264 * Change our mind and don't try copy.
1265 */
1266 if (sdp->sdl_family != AF_LINK ||
1267 sdp->sdl_len > sizeof(sdlbuf)) {
1268 goto makedummy;
1269 }
1270 bcopy(sdp, sdl2, sdp->sdl_len);
1271 } else {
1272 makedummy:
1273 sdl2->sdl_len =
1274 offsetof(struct sockaddr_dl, sdl_data[0]);
1275 sdl2->sdl_family = AF_LINK;
1276 sdl2->sdl_index = 0;
1277 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1278 }
1279 *mp = sbcreatecontrol(sdl2, sdl2->sdl_len, IP_RECVIF,
1280 IPPROTO_IP, M_NOWAIT);
1281 if (*mp)
1282 mp = &(*mp)->m_next;
1283 }
1284 if (inp->inp_flags & INP_RECVTOS) {
1285 *mp = sbcreatecontrol(&ip->ip_tos, sizeof(u_char), IP_RECVTOS,
1286 IPPROTO_IP, M_NOWAIT);
1287 if (*mp)
1288 mp = &(*mp)->m_next;
1289 }
1290
1291 if (inp->inp_flags2 & INP_RECVFLOWID) {
1292 uint32_t flowid, flow_type;
1293
1294 flowid = m->m_pkthdr.flowid;
1295 flow_type = M_HASHTYPE_GET(m);
1296
1297 /*
1298 * XXX should handle the failure of one or the
1299 * other - don't populate both?
1300 */
1301 *mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IP_FLOWID,
1302 IPPROTO_IP, M_NOWAIT);
1303 if (*mp)
1304 mp = &(*mp)->m_next;
1305 *mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1306 IP_FLOWTYPE, IPPROTO_IP, M_NOWAIT);
1307 if (*mp)
1308 mp = &(*mp)->m_next;
1309 }
1310
1311 #ifdef RSS
1312 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1313 uint32_t flowid, flow_type;
1314 uint32_t rss_bucketid;
1315
1316 flowid = m->m_pkthdr.flowid;
1317 flow_type = M_HASHTYPE_GET(m);
1318
1319 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1320 *mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1321 IP_RSSBUCKETID, IPPROTO_IP, M_NOWAIT);
1322 if (*mp)
1323 mp = &(*mp)->m_next;
1324 }
1325 }
1326 #endif
1327 }
1328
1329 /*
1330 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1331 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1332 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1333 * compiled.
1334 */
1335 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1336 VNET_DEFINE(struct socket *, ip_rsvpd);
1337
1338 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1339
1340 int
ip_rsvp_init(struct socket * so)1341 ip_rsvp_init(struct socket *so)
1342 {
1343
1344 if (V_ip_rsvpd != NULL)
1345 return EADDRINUSE;
1346
1347 V_ip_rsvpd = so;
1348 /*
1349 * This may seem silly, but we need to be sure we don't over-increment
1350 * the RSVP counter, in case something slips up.
1351 */
1352 if (!V_ip_rsvp_on) {
1353 V_ip_rsvp_on = 1;
1354 V_rsvp_on++;
1355 }
1356
1357 return 0;
1358 }
1359
1360 int
ip_rsvp_done(void)1361 ip_rsvp_done(void)
1362 {
1363
1364 V_ip_rsvpd = NULL;
1365 /*
1366 * This may seem silly, but we need to be sure we don't over-decrement
1367 * the RSVP counter, in case something slips up.
1368 */
1369 if (V_ip_rsvp_on) {
1370 V_ip_rsvp_on = 0;
1371 V_rsvp_on--;
1372 }
1373 return 0;
1374 }
1375
1376 int
rsvp_input(struct mbuf ** mp,int * offp,int proto)1377 rsvp_input(struct mbuf **mp, int *offp, int proto)
1378 {
1379 struct mbuf *m;
1380
1381 m = *mp;
1382 *mp = NULL;
1383
1384 if (rsvp_input_p) { /* call the real one if loaded */
1385 *mp = m;
1386 rsvp_input_p(mp, offp, proto);
1387 return (IPPROTO_DONE);
1388 }
1389
1390 /* Can still get packets with rsvp_on = 0 if there is a local member
1391 * of the group to which the RSVP packet is addressed. But in this
1392 * case we want to throw the packet away.
1393 */
1394
1395 if (!V_rsvp_on) {
1396 m_freem(m);
1397 return (IPPROTO_DONE);
1398 }
1399
1400 if (V_ip_rsvpd != NULL) {
1401 *mp = m;
1402 rip_input(mp, offp, proto);
1403 return (IPPROTO_DONE);
1404 }
1405 /* Drop the packet */
1406 m_freem(m);
1407 return (IPPROTO_DONE);
1408 }
1409