1 /*-
2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
30 */
31
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34
35 #include "opt_bootp.h"
36 #include "opt_ipstealth.h"
37 #include "opt_ipsec.h"
38 #include "opt_route.h"
39 #include "opt_rss.h"
40
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/hhook.h>
44 #include <sys/mbuf.h>
45 #include <sys/malloc.h>
46 #include <sys/domain.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/time.h>
50 #include <sys/kernel.h>
51 #include <sys/lock.h>
52 #include <sys/rmlock.h>
53 #include <sys/rwlock.h>
54 #include <sys/sdt.h>
55 #include <sys/syslog.h>
56 #include <sys/sysctl.h>
57
58 #include <net/pfil.h>
59 #include <net/if.h>
60 #include <net/if_types.h>
61 #include <net/if_var.h>
62 #include <net/if_dl.h>
63 #include <net/route.h>
64 #include <net/netisr.h>
65 #include <net/rss_config.h>
66 #include <net/vnet.h>
67
68 #include <netinet/in.h>
69 #include <netinet/in_kdtrace.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/in_var.h>
72 #include <netinet/ip.h>
73 #include <netinet/in_pcb.h>
74 #include <netinet/ip_var.h>
75 #include <netinet/ip_fw.h>
76 #include <netinet/ip_icmp.h>
77 #include <netinet/ip_options.h>
78 #include <machine/in_cksum.h>
79 #include <netinet/ip_carp.h>
80 #include <netinet/in_rss.h>
81
82 #include <netipsec/ipsec_support.h>
83
84 #include <sys/socketvar.h>
85
86 #include <security/mac/mac_framework.h>
87
88 #ifdef CTASSERT
89 CTASSERT(sizeof(struct ip) == 20);
90 #endif
91
92 /* IP reassembly functions are defined in ip_reass.c. */
93 extern void ipreass_init(void);
94 extern void ipreass_drain(void);
95 extern void ipreass_slowtimo(void);
96 #ifdef VIMAGE
97 extern void ipreass_destroy(void);
98 #endif
99
100 struct rmlock in_ifaddr_lock;
101 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
102
103 VNET_DEFINE(int, rsvp_on);
104
105 VNET_DEFINE(int, ipforwarding);
106 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
107 &VNET_NAME(ipforwarding), 0,
108 "Enable IP forwarding between interfaces");
109
110 static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */
111 #define V_ipsendredirects VNET(ipsendredirects)
112 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
113 &VNET_NAME(ipsendredirects), 0,
114 "Enable sending IP redirects");
115
116 /*
117 * XXX - Setting ip_checkinterface mostly implements the receive side of
118 * the Strong ES model described in RFC 1122, but since the routing table
119 * and transmit implementation do not implement the Strong ES model,
120 * setting this to 1 results in an odd hybrid.
121 *
122 * XXX - ip_checkinterface currently must be disabled if you use ipnat
123 * to translate the destination address to another local interface.
124 *
125 * XXX - ip_checkinterface must be disabled if you add IP aliases
126 * to the loopback interface instead of the interface where the
127 * packets for those addresses are received.
128 */
129 static VNET_DEFINE(int, ip_checkinterface);
130 #define V_ip_checkinterface VNET(ip_checkinterface)
131 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
132 &VNET_NAME(ip_checkinterface), 0,
133 "Verify packet arrives on correct interface");
134
135 VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */
136
137 static struct netisr_handler ip_nh = {
138 .nh_name = "ip",
139 .nh_handler = ip_input,
140 .nh_proto = NETISR_IP,
141 #ifdef RSS
142 .nh_m2cpuid = rss_soft_m2cpuid_v4,
143 .nh_policy = NETISR_POLICY_CPU,
144 .nh_dispatch = NETISR_DISPATCH_HYBRID,
145 #else
146 .nh_policy = NETISR_POLICY_FLOW,
147 #endif
148 };
149
150 #ifdef RSS
151 /*
152 * Directly dispatched frames are currently assumed
153 * to have a flowid already calculated.
154 *
155 * It should likely have something that assert it
156 * actually has valid flow details.
157 */
158 static struct netisr_handler ip_direct_nh = {
159 .nh_name = "ip_direct",
160 .nh_handler = ip_direct_input,
161 .nh_proto = NETISR_IP_DIRECT,
162 .nh_m2cpuid = rss_soft_m2cpuid_v4,
163 .nh_policy = NETISR_POLICY_CPU,
164 .nh_dispatch = NETISR_DISPATCH_HYBRID,
165 };
166 #endif
167
168 extern struct domain inetdomain;
169 extern struct protosw inetsw[];
170 u_char ip_protox[IPPROTO_MAX];
171 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
172 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
173 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
174
175 #ifdef IPCTL_DEFMTU
176 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
177 &ip_mtu, 0, "Default MTU");
178 #endif
179
180 #ifdef IPSTEALTH
181 VNET_DEFINE(int, ipstealth);
182 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
183 &VNET_NAME(ipstealth), 0,
184 "IP stealth mode, no TTL decrementation on forwarding");
185 #endif
186
187 /*
188 * IP statistics are stored in the "array" of counter(9)s.
189 */
190 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
191 VNET_PCPUSTAT_SYSINIT(ipstat);
192 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
193 "IP statistics (struct ipstat, netinet/ip_var.h)");
194
195 #ifdef VIMAGE
196 VNET_PCPUSTAT_SYSUNINIT(ipstat);
197 #endif /* VIMAGE */
198
199 /*
200 * Kernel module interface for updating ipstat. The argument is an index
201 * into ipstat treated as an array.
202 */
203 void
204 kmod_ipstat_inc(int statnum)
205 {
206
207 counter_u64_add(VNET(ipstat)[statnum], 1);
208 }
209
210 void
211 kmod_ipstat_dec(int statnum)
212 {
213
214 counter_u64_add(VNET(ipstat)[statnum], -1);
215 }
216
217 static int
218 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
219 {
220 int error, qlimit;
221
222 netisr_getqlimit(&ip_nh, &qlimit);
223 error = sysctl_handle_int(oidp, &qlimit, 0, req);
224 if (error || !req->newptr)
225 return (error);
226 if (qlimit < 1)
227 return (EINVAL);
228 return (netisr_setqlimit(&ip_nh, qlimit));
229 }
230 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
231 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
232 "Maximum size of the IP input queue");
233
234 static int
235 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
236 {
237 u_int64_t qdrops_long;
238 int error, qdrops;
239
240 netisr_getqdrops(&ip_nh, &qdrops_long);
241 qdrops = qdrops_long;
242 error = sysctl_handle_int(oidp, &qdrops, 0, req);
243 if (error || !req->newptr)
244 return (error);
245 if (qdrops != 0)
246 return (EINVAL);
247 netisr_clearqdrops(&ip_nh);
248 return (0);
249 }
250
251 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
252 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I",
253 "Number of packets dropped from the IP input queue");
254
255 #ifdef RSS
256 static int
257 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
258 {
259 int error, qlimit;
260
261 netisr_getqlimit(&ip_direct_nh, &qlimit);
262 error = sysctl_handle_int(oidp, &qlimit, 0, req);
263 if (error || !req->newptr)
264 return (error);
265 if (qlimit < 1)
266 return (EINVAL);
267 return (netisr_setqlimit(&ip_direct_nh, qlimit));
268 }
269 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
270 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
271 "I", "Maximum size of the IP direct input queue");
272
273 static int
274 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
275 {
276 u_int64_t qdrops_long;
277 int error, qdrops;
278
279 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
280 qdrops = qdrops_long;
281 error = sysctl_handle_int(oidp, &qdrops, 0, req);
282 if (error || !req->newptr)
283 return (error);
284 if (qdrops != 0)
285 return (EINVAL);
286 netisr_clearqdrops(&ip_direct_nh);
287 return (0);
288 }
289
290 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
291 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I",
292 "Number of packets dropped from the IP direct input queue");
293 #endif /* RSS */
294
295 /*
296 * IP initialization: fill in IP protocol switch table.
297 * All protocols not implemented in kernel go to raw IP protocol handler.
298 */
299 void
300 ip_init(void)
301 {
302 struct protosw *pr;
303 int i;
304
305 TAILQ_INIT(&V_in_ifaddrhead);
306 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
307
308 /* Initialize IP reassembly queue. */
309 ipreass_init();
310
311 /* Initialize packet filter hooks. */
312 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
313 V_inet_pfil_hook.ph_af = AF_INET;
314 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
315 printf("%s: WARNING: unable to register pfil hook, "
316 "error %d\n", __func__, i);
317
318 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
319 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
320 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
321 printf("%s: WARNING: unable to register input helper hook\n",
322 __func__);
323 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
324 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
325 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
326 printf("%s: WARNING: unable to register output helper hook\n",
327 __func__);
328
329 /* Skip initialization of globals for non-default instances. */
330 #ifdef VIMAGE
331 if (!IS_DEFAULT_VNET(curvnet)) {
332 netisr_register_vnet(&ip_nh);
333 #ifdef RSS
334 netisr_register_vnet(&ip_direct_nh);
335 #endif
336 return;
337 }
338 #endif
339
340 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
341 if (pr == NULL)
342 panic("ip_init: PF_INET not found");
343
344 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
345 for (i = 0; i < IPPROTO_MAX; i++)
346 ip_protox[i] = pr - inetsw;
347 /*
348 * Cycle through IP protocols and put them into the appropriate place
349 * in ip_protox[].
350 */
351 for (pr = inetdomain.dom_protosw;
352 pr < inetdomain.dom_protoswNPROTOSW; pr++)
353 if (pr->pr_domain->dom_family == PF_INET &&
354 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
355 /* Be careful to only index valid IP protocols. */
356 if (pr->pr_protocol < IPPROTO_MAX)
357 ip_protox[pr->pr_protocol] = pr - inetsw;
358 }
359
360 netisr_register(&ip_nh);
361 #ifdef RSS
362 netisr_register(&ip_direct_nh);
363 #endif
364 }
365
366 #ifdef VIMAGE
367 static void
368 ip_destroy(void *unused __unused)
369 {
370 struct ifnet *ifp;
371 int error;
372
373 #ifdef RSS
374 netisr_unregister_vnet(&ip_direct_nh);
375 #endif
376 netisr_unregister_vnet(&ip_nh);
377
378 if ((error = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
379 printf("%s: WARNING: unable to unregister pfil hook, "
380 "error %d\n", __func__, error);
381
382 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
383 if (error != 0) {
384 printf("%s: WARNING: unable to deregister input helper hook "
385 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
386 "error %d returned\n", __func__, error);
387 }
388 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
389 if (error != 0) {
390 printf("%s: WARNING: unable to deregister output helper hook "
391 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
392 "error %d returned\n", __func__, error);
393 }
394
395 /* Remove the IPv4 addresses from all interfaces. */
396 in_ifscrub_all();
397
398 /* Make sure the IPv4 routes are gone as well. */
399 IFNET_RLOCK();
400 TAILQ_FOREACH(ifp, &V_ifnet, if_link)
401 rt_flushifroutes_af(ifp, AF_INET);
402 IFNET_RUNLOCK();
403
404 /* Destroy IP reassembly queue. */
405 ipreass_destroy();
406
407 /* Cleanup in_ifaddr hash table; should be empty. */
408 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
409 }
410
411 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
412 #endif
413
414 #ifdef RSS
415 /*
416 * IP direct input routine.
417 *
418 * This is called when reinjecting completed fragments where
419 * all of the previous checking and book-keeping has been done.
420 */
421 void
422 ip_direct_input(struct mbuf *m)
423 {
424 struct ip *ip;
425 int hlen;
426
427 ip = mtod(m, struct ip *);
428 hlen = ip->ip_hl << 2;
429
430 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
431 if (IPSEC_ENABLED(ipv4)) {
432 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
433 return;
434 }
435 #endif /* IPSEC */
436 IPSTAT_INC(ips_delivered);
437 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
438 return;
439 }
440 #endif
441
442 /*
443 * Ip input routine. Checksum and byte swap header. If fragmented
444 * try to reassemble. Process options. Pass to next level.
445 */
446 void
447 ip_input(struct mbuf *m)
448 {
449 struct rm_priotracker in_ifa_tracker;
450 struct ip *ip = NULL;
451 struct in_ifaddr *ia = NULL;
452 struct ifaddr *ifa;
453 struct ifnet *ifp;
454 int checkif, hlen = 0;
455 uint16_t sum, ip_len;
456 int dchg = 0; /* dest changed after fw */
457 struct in_addr odst; /* original dst address */
458
459 M_ASSERTPKTHDR(m);
460
461 if (m->m_flags & M_FASTFWD_OURS) {
462 m->m_flags &= ~M_FASTFWD_OURS;
463 /* Set up some basics that will be used later. */
464 ip = mtod(m, struct ip *);
465 hlen = ip->ip_hl << 2;
466 ip_len = ntohs(ip->ip_len);
467 goto ours;
468 }
469
470 IPSTAT_INC(ips_total);
471
472 if (m->m_pkthdr.len < sizeof(struct ip))
473 goto tooshort;
474
475 if (m->m_len < sizeof (struct ip) &&
476 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
477 IPSTAT_INC(ips_toosmall);
478 return;
479 }
480 ip = mtod(m, struct ip *);
481
482 if (ip->ip_v != IPVERSION) {
483 IPSTAT_INC(ips_badvers);
484 goto bad;
485 }
486
487 hlen = ip->ip_hl << 2;
488 if (hlen < sizeof(struct ip)) { /* minimum header length */
489 IPSTAT_INC(ips_badhlen);
490 goto bad;
491 }
492 if (hlen > m->m_len) {
493 if ((m = m_pullup(m, hlen)) == NULL) {
494 IPSTAT_INC(ips_badhlen);
495 return;
496 }
497 ip = mtod(m, struct ip *);
498 }
499
500 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
501
502 /* 127/8 must not appear on wire - RFC1122 */
503 ifp = m->m_pkthdr.rcvif;
504 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
505 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
506 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
507 IPSTAT_INC(ips_badaddr);
508 goto bad;
509 }
510 }
511
512 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
513 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
514 } else {
515 if (hlen == sizeof(struct ip)) {
516 sum = in_cksum_hdr(ip);
517 } else {
518 sum = in_cksum(m, hlen);
519 }
520 }
521 if (sum) {
522 IPSTAT_INC(ips_badsum);
523 goto bad;
524 }
525
526 #ifdef ALTQ
527 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
528 /* packet is dropped by traffic conditioner */
529 return;
530 #endif
531
532 ip_len = ntohs(ip->ip_len);
533 if (ip_len < hlen) {
534 IPSTAT_INC(ips_badlen);
535 goto bad;
536 }
537
538 /*
539 * Check that the amount of data in the buffers
540 * is as at least much as the IP header would have us expect.
541 * Trim mbufs if longer than we expect.
542 * Drop packet if shorter than we expect.
543 */
544 if (m->m_pkthdr.len < ip_len) {
545 tooshort:
546 IPSTAT_INC(ips_tooshort);
547 goto bad;
548 }
549 if (m->m_pkthdr.len > ip_len) {
550 if (m->m_len == m->m_pkthdr.len) {
551 m->m_len = ip_len;
552 m->m_pkthdr.len = ip_len;
553 } else
554 m_adj(m, ip_len - m->m_pkthdr.len);
555 }
556
557 /*
558 * Try to forward the packet, but if we fail continue.
559 * ip_tryforward() does not generate redirects, so fall
560 * through to normal processing if redirects are required.
561 * ip_tryforward() does inbound and outbound packet firewall
562 * processing. If firewall has decided that destination becomes
563 * our local address, it sets M_FASTFWD_OURS flag. In this
564 * case skip another inbound firewall processing and update
565 * ip pointer.
566 */
567 if (V_ipforwarding != 0 && V_ipsendredirects == 0
568 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
569 && (!IPSEC_ENABLED(ipv4) ||
570 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
571 #endif
572 ) {
573 if ((m = ip_tryforward(m)) == NULL)
574 return;
575 if (m->m_flags & M_FASTFWD_OURS) {
576 m->m_flags &= ~M_FASTFWD_OURS;
577 ip = mtod(m, struct ip *);
578 goto ours;
579 }
580 }
581
582 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
583 /*
584 * Bypass packet filtering for packets previously handled by IPsec.
585 */
586 if (IPSEC_ENABLED(ipv4) &&
587 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
588 goto passin;
589 #endif
590
591 /*
592 * Run through list of hooks for input packets.
593 *
594 * NB: Beware of the destination address changing (e.g.
595 * by NAT rewriting). When this happens, tell
596 * ip_forward to do the right thing.
597 */
598
599 /* Jump over all PFIL processing if hooks are not active. */
600 if (!PFIL_HOOKED(&V_inet_pfil_hook))
601 goto passin;
602
603 odst = ip->ip_dst;
604 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, 0, NULL) != 0)
605 return;
606 if (m == NULL) /* consumed by filter */
607 return;
608
609 ip = mtod(m, struct ip *);
610 dchg = (odst.s_addr != ip->ip_dst.s_addr);
611 ifp = m->m_pkthdr.rcvif;
612
613 if (m->m_flags & M_FASTFWD_OURS) {
614 m->m_flags &= ~M_FASTFWD_OURS;
615 goto ours;
616 }
617 if (m->m_flags & M_IP_NEXTHOP) {
618 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
619 /*
620 * Directly ship the packet on. This allows
621 * forwarding packets originally destined to us
622 * to some other directly connected host.
623 */
624 ip_forward(m, 1);
625 return;
626 }
627 }
628 passin:
629
630 /*
631 * Process options and, if not destined for us,
632 * ship it on. ip_dooptions returns 1 when an
633 * error was detected (causing an icmp message
634 * to be sent and the original packet to be freed).
635 */
636 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
637 return;
638
639 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
640 * matter if it is destined to another node, or whether it is
641 * a multicast one, RSVP wants it! and prevents it from being forwarded
642 * anywhere else. Also checks if the rsvp daemon is running before
643 * grabbing the packet.
644 */
645 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
646 goto ours;
647
648 /*
649 * Check our list of addresses, to see if the packet is for us.
650 * If we don't have any addresses, assume any unicast packet
651 * we receive might be for us (and let the upper layers deal
652 * with it).
653 */
654 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
655 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
656 goto ours;
657
658 /*
659 * Enable a consistency check between the destination address
660 * and the arrival interface for a unicast packet (the RFC 1122
661 * strong ES model) if IP forwarding is disabled and the packet
662 * is not locally generated and the packet is not subject to
663 * 'ipfw fwd'.
664 *
665 * XXX - Checking also should be disabled if the destination
666 * address is ipnat'ed to a different interface.
667 *
668 * XXX - Checking is incompatible with IP aliases added
669 * to the loopback interface instead of the interface where
670 * the packets are received.
671 *
672 * XXX - This is the case for carp vhost IPs as well so we
673 * insert a workaround. If the packet got here, we already
674 * checked with carp_iamatch() and carp_forus().
675 */
676 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
677 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
678 ifp->if_carp == NULL && (dchg == 0);
679
680 /*
681 * Check for exact addresses in the hash bucket.
682 */
683 IN_IFADDR_RLOCK(&in_ifa_tracker);
684 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
685 /*
686 * If the address matches, verify that the packet
687 * arrived via the correct interface if checking is
688 * enabled.
689 */
690 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
691 (!checkif || ia->ia_ifp == ifp)) {
692 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
693 counter_u64_add(ia->ia_ifa.ifa_ibytes,
694 m->m_pkthdr.len);
695 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
696 goto ours;
697 }
698 }
699 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
700
701 /*
702 * Check for broadcast addresses.
703 *
704 * Only accept broadcast packets that arrive via the matching
705 * interface. Reception of forwarded directed broadcasts would
706 * be handled via ip_forward() and ether_output() with the loopback
707 * into the stack for SIMPLEX interfaces handled by ether_output().
708 */
709 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
710 IF_ADDR_RLOCK(ifp);
711 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
712 if (ifa->ifa_addr->sa_family != AF_INET)
713 continue;
714 ia = ifatoia(ifa);
715 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
716 ip->ip_dst.s_addr) {
717 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
718 counter_u64_add(ia->ia_ifa.ifa_ibytes,
719 m->m_pkthdr.len);
720 IF_ADDR_RUNLOCK(ifp);
721 goto ours;
722 }
723 #ifdef BOOTP_COMPAT
724 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
725 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
726 counter_u64_add(ia->ia_ifa.ifa_ibytes,
727 m->m_pkthdr.len);
728 IF_ADDR_RUNLOCK(ifp);
729 goto ours;
730 }
731 #endif
732 }
733 IF_ADDR_RUNLOCK(ifp);
734 ia = NULL;
735 }
736 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
737 /*
738 * RFC 3927 2.7: Do not forward multicast packets from
739 * IN_LINKLOCAL.
740 */
741 if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
742 /*
743 * If we are acting as a multicast router, all
744 * incoming multicast packets are passed to the
745 * kernel-level multicast forwarding function.
746 * The packet is returned (relatively) intact; if
747 * ip_mforward() returns a non-zero value, the packet
748 * must be discarded, else it may be accepted below.
749 */
750 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
751 IPSTAT_INC(ips_cantforward);
752 m_freem(m);
753 return;
754 }
755
756 /*
757 * The process-level routing daemon needs to receive
758 * all multicast IGMP packets, whether or not this
759 * host belongs to their destination groups.
760 */
761 if (ip->ip_p == IPPROTO_IGMP)
762 goto ours;
763 IPSTAT_INC(ips_forward);
764 }
765 /*
766 * Assume the packet is for us, to avoid prematurely taking
767 * a lock on the in_multi hash. Protocols must perform
768 * their own filtering and update statistics accordingly.
769 */
770 goto ours;
771 }
772 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
773 goto ours;
774 if (ip->ip_dst.s_addr == INADDR_ANY)
775 goto ours;
776 /* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
777 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
778 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
779 IPSTAT_INC(ips_cantforward);
780 m_freem(m);
781 return;
782 }
783
784 /*
785 * Not for us; forward if possible and desirable.
786 */
787 if (V_ipforwarding == 0) {
788 IPSTAT_INC(ips_cantforward);
789 m_freem(m);
790 } else {
791 ip_forward(m, dchg);
792 }
793 return;
794
795 ours:
796 #ifdef IPSTEALTH
797 /*
798 * IPSTEALTH: Process non-routing options only
799 * if the packet is destined for us.
800 */
801 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
802 return;
803 #endif /* IPSTEALTH */
804
805 /*
806 * Attempt reassembly; if it succeeds, proceed.
807 * ip_reass() will return a different mbuf.
808 */
809 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
810 /* XXXGL: shouldn't we save & set m_flags? */
811 m = ip_reass(m);
812 if (m == NULL)
813 return;
814 ip = mtod(m, struct ip *);
815 /* Get the header length of the reassembled packet */
816 hlen = ip->ip_hl << 2;
817 }
818
819 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
820 if (IPSEC_ENABLED(ipv4)) {
821 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
822 return;
823 }
824 #endif /* IPSEC */
825
826 /*
827 * Switch out to protocol's input routine.
828 */
829 IPSTAT_INC(ips_delivered);
830
831 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
832 return;
833 bad:
834 m_freem(m);
835 }
836
837 /*
838 * IP timer processing;
839 * if a timer expires on a reassembly
840 * queue, discard it.
841 */
842 void
843 ip_slowtimo(void)
844 {
845 VNET_ITERATOR_DECL(vnet_iter);
846
847 VNET_LIST_RLOCK_NOSLEEP();
848 VNET_FOREACH(vnet_iter) {
849 CURVNET_SET(vnet_iter);
850 ipreass_slowtimo();
851 CURVNET_RESTORE();
852 }
853 VNET_LIST_RUNLOCK_NOSLEEP();
854 }
855
856 void
857 ip_drain(void)
858 {
859 VNET_ITERATOR_DECL(vnet_iter);
860
861 VNET_LIST_RLOCK_NOSLEEP();
862 VNET_FOREACH(vnet_iter) {
863 CURVNET_SET(vnet_iter);
864 ipreass_drain();
865 CURVNET_RESTORE();
866 }
867 VNET_LIST_RUNLOCK_NOSLEEP();
868 }
869
870 /*
871 * The protocol to be inserted into ip_protox[] must be already registered
872 * in inetsw[], either statically or through pf_proto_register().
873 */
874 int
875 ipproto_register(short ipproto)
876 {
877 struct protosw *pr;
878
879 /* Sanity checks. */
880 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
881 return (EPROTONOSUPPORT);
882
883 /*
884 * The protocol slot must not be occupied by another protocol
885 * already. An index pointing to IPPROTO_RAW is unused.
886 */
887 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
888 if (pr == NULL)
889 return (EPFNOSUPPORT);
890 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
891 return (EEXIST);
892
893 /* Find the protocol position in inetsw[] and set the index. */
894 for (pr = inetdomain.dom_protosw;
895 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
896 if (pr->pr_domain->dom_family == PF_INET &&
897 pr->pr_protocol && pr->pr_protocol == ipproto) {
898 ip_protox[pr->pr_protocol] = pr - inetsw;
899 return (0);
900 }
901 }
902 return (EPROTONOSUPPORT);
903 }
904
905 int
906 ipproto_unregister(short ipproto)
907 {
908 struct protosw *pr;
909
910 /* Sanity checks. */
911 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
912 return (EPROTONOSUPPORT);
913
914 /* Check if the protocol was indeed registered. */
915 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
916 if (pr == NULL)
917 return (EPFNOSUPPORT);
918 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
919 return (ENOENT);
920
921 /* Reset the protocol slot to IPPROTO_RAW. */
922 ip_protox[ipproto] = pr - inetsw;
923 return (0);
924 }
925
926 u_char inetctlerrmap[PRC_NCMDS] = {
927 0, 0, 0, 0,
928 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
929 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
930 EMSGSIZE, EHOSTUNREACH, 0, 0,
931 0, 0, EHOSTUNREACH, 0,
932 ENOPROTOOPT, ECONNREFUSED
933 };
934
935 /*
936 * Forward a packet. If some error occurs return the sender
937 * an icmp packet. Note we can't always generate a meaningful
938 * icmp message because icmp doesn't have a large enough repertoire
939 * of codes and types.
940 *
941 * If not forwarding, just drop the packet. This could be confusing
942 * if ipforwarding was zero but some routing protocol was advancing
943 * us as a gateway to somewhere. However, we must let the routing
944 * protocol deal with that.
945 *
946 * The srcrt parameter indicates whether the packet is being forwarded
947 * via a source route.
948 */
949 void
950 ip_forward(struct mbuf *m, int srcrt)
951 {
952 struct ip *ip = mtod(m, struct ip *);
953 struct in_ifaddr *ia;
954 struct mbuf *mcopy;
955 struct sockaddr_in *sin;
956 struct in_addr dest;
957 struct route ro;
958 int error, type = 0, code = 0, mtu = 0;
959
960 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
961 IPSTAT_INC(ips_cantforward);
962 m_freem(m);
963 return;
964 }
965 if (
966 #ifdef IPSTEALTH
967 V_ipstealth == 0 &&
968 #endif
969 ip->ip_ttl <= IPTTLDEC) {
970 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
971 return;
972 }
973
974 bzero(&ro, sizeof(ro));
975 sin = (struct sockaddr_in *)&ro.ro_dst;
976 sin->sin_family = AF_INET;
977 sin->sin_len = sizeof(*sin);
978 sin->sin_addr = ip->ip_dst;
979 #ifdef RADIX_MPATH
980 rtalloc_mpath_fib(&ro,
981 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
982 M_GETFIB(m));
983 #else
984 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
985 #endif
986 if (ro.ro_rt != NULL) {
987 ia = ifatoia(ro.ro_rt->rt_ifa);
988 ifa_ref(&ia->ia_ifa);
989 } else
990 ia = NULL;
991 /*
992 * Save the IP header and at most 8 bytes of the payload,
993 * in case we need to generate an ICMP message to the src.
994 *
995 * XXX this can be optimized a lot by saving the data in a local
996 * buffer on the stack (72 bytes at most), and only allocating the
997 * mbuf if really necessary. The vast majority of the packets
998 * are forwarded without having to send an ICMP back (either
999 * because unnecessary, or because rate limited), so we are
1000 * really we are wasting a lot of work here.
1001 *
1002 * We don't use m_copy() because it might return a reference
1003 * to a shared cluster. Both this function and ip_output()
1004 * assume exclusive access to the IP header in `m', so any
1005 * data in a cluster may change before we reach icmp_error().
1006 */
1007 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1008 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1009 /*
1010 * It's probably ok if the pkthdr dup fails (because
1011 * the deep copy of the tag chain failed), but for now
1012 * be conservative and just discard the copy since
1013 * code below may some day want the tags.
1014 */
1015 m_free(mcopy);
1016 mcopy = NULL;
1017 }
1018 if (mcopy != NULL) {
1019 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1020 mcopy->m_pkthdr.len = mcopy->m_len;
1021 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1022 }
1023 #ifdef IPSTEALTH
1024 if (V_ipstealth == 0)
1025 #endif
1026 ip->ip_ttl -= IPTTLDEC;
1027 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1028 if (IPSEC_ENABLED(ipv4)) {
1029 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1030 /* mbuf consumed by IPsec */
1031 if (ia != NULL)
1032 ifa_free(&ia->ia_ifa);
1033 m_freem(mcopy);
1034 if (error != EINPROGRESS)
1035 IPSTAT_INC(ips_cantforward);
1036 return;
1037 }
1038 /* No IPsec processing required */
1039 }
1040 #endif /* IPSEC */
1041 /*
1042 * If forwarding packet using same interface that it came in on,
1043 * perhaps should send a redirect to sender to shortcut a hop.
1044 * Only send redirect if source is sending directly to us,
1045 * and if packet was not source routed (or has any options).
1046 * Also, don't send redirect if forwarding using a default route
1047 * or a route modified by a redirect.
1048 */
1049 dest.s_addr = 0;
1050 if (!srcrt && V_ipsendredirects &&
1051 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1052 struct rtentry *rt;
1053
1054 rt = ro.ro_rt;
1055
1056 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1057 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1058 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1059 u_long src = ntohl(ip->ip_src.s_addr);
1060
1061 if (RTA(rt) &&
1062 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1063 if (rt->rt_flags & RTF_GATEWAY)
1064 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1065 else
1066 dest.s_addr = ip->ip_dst.s_addr;
1067 /* Router requirements says to only send host redirects */
1068 type = ICMP_REDIRECT;
1069 code = ICMP_REDIRECT_HOST;
1070 }
1071 }
1072 }
1073
1074 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1075
1076 if (error == EMSGSIZE && ro.ro_rt)
1077 mtu = ro.ro_rt->rt_mtu;
1078 RO_RTFREE(&ro);
1079
1080 if (error)
1081 IPSTAT_INC(ips_cantforward);
1082 else {
1083 IPSTAT_INC(ips_forward);
1084 if (type)
1085 IPSTAT_INC(ips_redirectsent);
1086 else {
1087 if (mcopy)
1088 m_freem(mcopy);
1089 if (ia != NULL)
1090 ifa_free(&ia->ia_ifa);
1091 return;
1092 }
1093 }
1094 if (mcopy == NULL) {
1095 if (ia != NULL)
1096 ifa_free(&ia->ia_ifa);
1097 return;
1098 }
1099
1100 switch (error) {
1101
1102 case 0: /* forwarded, but need redirect */
1103 /* type, code set above */
1104 break;
1105
1106 case ENETUNREACH:
1107 case EHOSTUNREACH:
1108 case ENETDOWN:
1109 case EHOSTDOWN:
1110 default:
1111 type = ICMP_UNREACH;
1112 code = ICMP_UNREACH_HOST;
1113 break;
1114
1115 case EMSGSIZE:
1116 type = ICMP_UNREACH;
1117 code = ICMP_UNREACH_NEEDFRAG;
1118 /*
1119 * If the MTU was set before make sure we are below the
1120 * interface MTU.
1121 * If the MTU wasn't set before use the interface mtu or
1122 * fall back to the next smaller mtu step compared to the
1123 * current packet size.
1124 */
1125 if (mtu != 0) {
1126 if (ia != NULL)
1127 mtu = min(mtu, ia->ia_ifp->if_mtu);
1128 } else {
1129 if (ia != NULL)
1130 mtu = ia->ia_ifp->if_mtu;
1131 else
1132 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1133 }
1134 IPSTAT_INC(ips_cantfrag);
1135 break;
1136
1137 case ENOBUFS:
1138 case EACCES: /* ipfw denied packet */
1139 m_freem(mcopy);
1140 if (ia != NULL)
1141 ifa_free(&ia->ia_ifa);
1142 return;
1143 }
1144 if (ia != NULL)
1145 ifa_free(&ia->ia_ifa);
1146 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1147 }
1148
1149 #define CHECK_SO_CT(sp, ct) \
1150 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1151
1152 void
1153 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1154 struct mbuf *m)
1155 {
1156
1157 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1158 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1159 struct bintime bt;
1160
1161 bintime(&bt);
1162 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1163 SCM_BINTIME, SOL_SOCKET);
1164 if (*mp)
1165 mp = &(*mp)->m_next;
1166 }
1167 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1168 struct timeval tv;
1169
1170 microtime(&tv);
1171 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1172 SCM_TIMESTAMP, SOL_SOCKET);
1173 if (*mp)
1174 mp = &(*mp)->m_next;
1175 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1176 struct timespec ts;
1177
1178 nanotime(&ts);
1179 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1180 SCM_REALTIME, SOL_SOCKET);
1181 if (*mp)
1182 mp = &(*mp)->m_next;
1183 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1184 struct timespec ts;
1185
1186 nanouptime(&ts);
1187 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1188 SCM_MONOTONIC, SOL_SOCKET);
1189 if (*mp)
1190 mp = &(*mp)->m_next;
1191 }
1192 if (inp->inp_flags & INP_RECVDSTADDR) {
1193 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1194 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1195 if (*mp)
1196 mp = &(*mp)->m_next;
1197 }
1198 if (inp->inp_flags & INP_RECVTTL) {
1199 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1200 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1201 if (*mp)
1202 mp = &(*mp)->m_next;
1203 }
1204 #ifdef notyet
1205 /* XXX
1206 * Moving these out of udp_input() made them even more broken
1207 * than they already were.
1208 */
1209 /* options were tossed already */
1210 if (inp->inp_flags & INP_RECVOPTS) {
1211 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1212 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1213 if (*mp)
1214 mp = &(*mp)->m_next;
1215 }
1216 /* ip_srcroute doesn't do what we want here, need to fix */
1217 if (inp->inp_flags & INP_RECVRETOPTS) {
1218 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1219 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1220 if (*mp)
1221 mp = &(*mp)->m_next;
1222 }
1223 #endif
1224 if (inp->inp_flags & INP_RECVIF) {
1225 struct ifnet *ifp;
1226 struct sdlbuf {
1227 struct sockaddr_dl sdl;
1228 u_char pad[32];
1229 } sdlbuf;
1230 struct sockaddr_dl *sdp;
1231 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1232
1233 if ((ifp = m->m_pkthdr.rcvif) &&
1234 ifp->if_index && ifp->if_index <= V_if_index) {
1235 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1236 /*
1237 * Change our mind and don't try copy.
1238 */
1239 if (sdp->sdl_family != AF_LINK ||
1240 sdp->sdl_len > sizeof(sdlbuf)) {
1241 goto makedummy;
1242 }
1243 bcopy(sdp, sdl2, sdp->sdl_len);
1244 } else {
1245 makedummy:
1246 sdl2->sdl_len =
1247 offsetof(struct sockaddr_dl, sdl_data[0]);
1248 sdl2->sdl_family = AF_LINK;
1249 sdl2->sdl_index = 0;
1250 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1251 }
1252 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1253 IP_RECVIF, IPPROTO_IP);
1254 if (*mp)
1255 mp = &(*mp)->m_next;
1256 }
1257 if (inp->inp_flags & INP_RECVTOS) {
1258 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1259 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1260 if (*mp)
1261 mp = &(*mp)->m_next;
1262 }
1263
1264 if (inp->inp_flags2 & INP_RECVFLOWID) {
1265 uint32_t flowid, flow_type;
1266
1267 flowid = m->m_pkthdr.flowid;
1268 flow_type = M_HASHTYPE_GET(m);
1269
1270 /*
1271 * XXX should handle the failure of one or the
1272 * other - don't populate both?
1273 */
1274 *mp = sbcreatecontrol((caddr_t) &flowid,
1275 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1276 if (*mp)
1277 mp = &(*mp)->m_next;
1278 *mp = sbcreatecontrol((caddr_t) &flow_type,
1279 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1280 if (*mp)
1281 mp = &(*mp)->m_next;
1282 }
1283
1284 #ifdef RSS
1285 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1286 uint32_t flowid, flow_type;
1287 uint32_t rss_bucketid;
1288
1289 flowid = m->m_pkthdr.flowid;
1290 flow_type = M_HASHTYPE_GET(m);
1291
1292 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1293 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1294 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1295 if (*mp)
1296 mp = &(*mp)->m_next;
1297 }
1298 }
1299 #endif
1300 }
1301
1302 /*
1303 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1304 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1305 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1306 * compiled.
1307 */
1308 static VNET_DEFINE(int, ip_rsvp_on);
1309 VNET_DEFINE(struct socket *, ip_rsvpd);
1310
1311 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1312
1313 int
1314 ip_rsvp_init(struct socket *so)
1315 {
1316
1317 if (so->so_type != SOCK_RAW ||
1318 so->so_proto->pr_protocol != IPPROTO_RSVP)
1319 return EOPNOTSUPP;
1320
1321 if (V_ip_rsvpd != NULL)
1322 return EADDRINUSE;
1323
1324 V_ip_rsvpd = so;
1325 /*
1326 * This may seem silly, but we need to be sure we don't over-increment
1327 * the RSVP counter, in case something slips up.
1328 */
1329 if (!V_ip_rsvp_on) {
1330 V_ip_rsvp_on = 1;
1331 V_rsvp_on++;
1332 }
1333
1334 return 0;
1335 }
1336
1337 int
1338 ip_rsvp_done(void)
1339 {
1340
1341 V_ip_rsvpd = NULL;
1342 /*
1343 * This may seem silly, but we need to be sure we don't over-decrement
1344 * the RSVP counter, in case something slips up.
1345 */
1346 if (V_ip_rsvp_on) {
1347 V_ip_rsvp_on = 0;
1348 V_rsvp_on--;
1349 }
1350 return 0;
1351 }
1352
1353 int
1354 rsvp_input(struct mbuf **mp, int *offp, int proto)
1355 {
1356 struct mbuf *m;
1357
1358 m = *mp;
1359 *mp = NULL;
1360
1361 if (rsvp_input_p) { /* call the real one if loaded */
1362 *mp = m;
1363 rsvp_input_p(mp, offp, proto);
1364 return (IPPROTO_DONE);
1365 }
1366
1367 /* Can still get packets with rsvp_on = 0 if there is a local member
1368 * of the group to which the RSVP packet is addressed. But in this
1369 * case we want to throw the packet away.
1370 */
1371
1372 if (!V_rsvp_on) {
1373 m_freem(m);
1374 return (IPPROTO_DONE);
1375 }
1376
1377 if (V_ip_rsvpd != NULL) {
1378 *mp = m;
1379 rip_input(mp, offp, proto);
1380 return (IPPROTO_DONE);
1381 }
1382 /* Drop the packet */
1383 m_freem(m);
1384 return (IPPROTO_DONE);
1385 }
Cache object: 85b7c8d510e0688d53a23be2730aaa20
|