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 __FBSDID("$FreeBSD: releng/12.0/sys/netinet/ip_input.c 337736 2018-08-14 07:54:14Z ae $");
36
37 #include "opt_bootp.h"
38 #include "opt_ipstealth.h"
39 #include "opt_ipsec.h"
40 #include "opt_route.h"
41 #include "opt_rss.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/pfil.h>
61 #include <net/if.h>
62 #include <net/if_types.h>
63 #include <net/if_var.h>
64 #include <net/if_dl.h>
65 #include <net/route.h>
66 #include <net/netisr.h>
67 #include <net/rss_config.h>
68 #include <net/vnet.h>
69
70 #include <netinet/in.h>
71 #include <netinet/in_kdtrace.h>
72 #include <netinet/in_systm.h>
73 #include <netinet/in_var.h>
74 #include <netinet/ip.h>
75 #include <netinet/in_pcb.h>
76 #include <netinet/ip_var.h>
77 #include <netinet/ip_fw.h>
78 #include <netinet/ip_icmp.h>
79 #include <netinet/ip_options.h>
80 #include <machine/in_cksum.h>
81 #include <netinet/ip_carp.h>
82 #include <netinet/in_rss.h>
83
84 #include <netipsec/ipsec_support.h>
85
86 #include <sys/socketvar.h>
87
88 #include <security/mac/mac_framework.h>
89
90 #ifdef CTASSERT
91 CTASSERT(sizeof(struct ip) == 20);
92 #endif
93
94 /* IP reassembly functions are defined in ip_reass.c. */
95 extern void ipreass_init(void);
96 extern void ipreass_drain(void);
97 extern void ipreass_slowtimo(void);
98 #ifdef VIMAGE
99 extern void ipreass_destroy(void);
100 #endif
101
102 struct rmlock in_ifaddr_lock;
103 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
104
105 VNET_DEFINE(int, rsvp_on);
106
107 VNET_DEFINE(int, ipforwarding);
108 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
109 &VNET_NAME(ipforwarding), 0,
110 "Enable IP forwarding between interfaces");
111
112 VNET_DEFINE_STATIC(int, ipsendredirects) = 1; /* XXX */
113 #define V_ipsendredirects VNET(ipsendredirects)
114 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
115 &VNET_NAME(ipsendredirects), 0,
116 "Enable sending IP redirects");
117
118 /*
119 * XXX - Setting ip_checkinterface mostly implements the receive side of
120 * the Strong ES model described in RFC 1122, but since the routing table
121 * and transmit implementation do not implement the Strong ES model,
122 * setting this to 1 results in an odd hybrid.
123 *
124 * XXX - ip_checkinterface currently must be disabled if you use ipnat
125 * to translate the destination address to another local interface.
126 *
127 * XXX - ip_checkinterface must be disabled if you add IP aliases
128 * to the loopback interface instead of the interface where the
129 * packets for those addresses are received.
130 */
131 VNET_DEFINE_STATIC(int, ip_checkinterface);
132 #define V_ip_checkinterface VNET(ip_checkinterface)
133 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
134 &VNET_NAME(ip_checkinterface), 0,
135 "Verify packet arrives on correct interface");
136
137 VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */
138
139 static struct netisr_handler ip_nh = {
140 .nh_name = "ip",
141 .nh_handler = ip_input,
142 .nh_proto = NETISR_IP,
143 #ifdef RSS
144 .nh_m2cpuid = rss_soft_m2cpuid_v4,
145 .nh_policy = NETISR_POLICY_CPU,
146 .nh_dispatch = NETISR_DISPATCH_HYBRID,
147 #else
148 .nh_policy = NETISR_POLICY_FLOW,
149 #endif
150 };
151
152 #ifdef RSS
153 /*
154 * Directly dispatched frames are currently assumed
155 * to have a flowid already calculated.
156 *
157 * It should likely have something that assert it
158 * actually has valid flow details.
159 */
160 static struct netisr_handler ip_direct_nh = {
161 .nh_name = "ip_direct",
162 .nh_handler = ip_direct_input,
163 .nh_proto = NETISR_IP_DIRECT,
164 .nh_m2cpuid = rss_soft_m2cpuid_v4,
165 .nh_policy = NETISR_POLICY_CPU,
166 .nh_dispatch = NETISR_DISPATCH_HYBRID,
167 };
168 #endif
169
170 extern struct domain inetdomain;
171 extern struct protosw inetsw[];
172 u_char ip_protox[IPPROTO_MAX];
173 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
174 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
175 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
176
177 #ifdef IPCTL_DEFMTU
178 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
179 &ip_mtu, 0, "Default MTU");
180 #endif
181
182 #ifdef IPSTEALTH
183 VNET_DEFINE(int, ipstealth);
184 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
185 &VNET_NAME(ipstealth), 0,
186 "IP stealth mode, no TTL decrementation on forwarding");
187 #endif
188
189 /*
190 * IP statistics are stored in the "array" of counter(9)s.
191 */
192 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
193 VNET_PCPUSTAT_SYSINIT(ipstat);
194 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
195 "IP statistics (struct ipstat, netinet/ip_var.h)");
196
197 #ifdef VIMAGE
198 VNET_PCPUSTAT_SYSUNINIT(ipstat);
199 #endif /* VIMAGE */
200
201 /*
202 * Kernel module interface for updating ipstat. The argument is an index
203 * into ipstat treated as an array.
204 */
205 void
206 kmod_ipstat_inc(int statnum)
207 {
208
209 counter_u64_add(VNET(ipstat)[statnum], 1);
210 }
211
212 void
213 kmod_ipstat_dec(int statnum)
214 {
215
216 counter_u64_add(VNET(ipstat)[statnum], -1);
217 }
218
219 static int
220 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
221 {
222 int error, qlimit;
223
224 netisr_getqlimit(&ip_nh, &qlimit);
225 error = sysctl_handle_int(oidp, &qlimit, 0, req);
226 if (error || !req->newptr)
227 return (error);
228 if (qlimit < 1)
229 return (EINVAL);
230 return (netisr_setqlimit(&ip_nh, qlimit));
231 }
232 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
233 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
234 "Maximum size of the IP input queue");
235
236 static int
237 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
238 {
239 u_int64_t qdrops_long;
240 int error, qdrops;
241
242 netisr_getqdrops(&ip_nh, &qdrops_long);
243 qdrops = qdrops_long;
244 error = sysctl_handle_int(oidp, &qdrops, 0, req);
245 if (error || !req->newptr)
246 return (error);
247 if (qdrops != 0)
248 return (EINVAL);
249 netisr_clearqdrops(&ip_nh);
250 return (0);
251 }
252
253 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
254 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I",
255 "Number of packets dropped from the IP input queue");
256
257 #ifdef RSS
258 static int
259 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
260 {
261 int error, qlimit;
262
263 netisr_getqlimit(&ip_direct_nh, &qlimit);
264 error = sysctl_handle_int(oidp, &qlimit, 0, req);
265 if (error || !req->newptr)
266 return (error);
267 if (qlimit < 1)
268 return (EINVAL);
269 return (netisr_setqlimit(&ip_direct_nh, qlimit));
270 }
271 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
272 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
273 "I", "Maximum size of the IP direct input queue");
274
275 static int
276 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
277 {
278 u_int64_t qdrops_long;
279 int error, qdrops;
280
281 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
282 qdrops = qdrops_long;
283 error = sysctl_handle_int(oidp, &qdrops, 0, req);
284 if (error || !req->newptr)
285 return (error);
286 if (qdrops != 0)
287 return (EINVAL);
288 netisr_clearqdrops(&ip_direct_nh);
289 return (0);
290 }
291
292 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
293 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I",
294 "Number of packets dropped from the IP direct input queue");
295 #endif /* RSS */
296
297 /*
298 * IP initialization: fill in IP protocol switch table.
299 * All protocols not implemented in kernel go to raw IP protocol handler.
300 */
301 void
302 ip_init(void)
303 {
304 struct protosw *pr;
305 int i;
306
307 CK_STAILQ_INIT(&V_in_ifaddrhead);
308 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
309
310 /* Initialize IP reassembly queue. */
311 ipreass_init();
312
313 /* Initialize packet filter hooks. */
314 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
315 V_inet_pfil_hook.ph_af = AF_INET;
316 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
317 printf("%s: WARNING: unable to register pfil hook, "
318 "error %d\n", __func__, i);
319
320 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
321 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
322 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
323 printf("%s: WARNING: unable to register input helper hook\n",
324 __func__);
325 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
326 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
327 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
328 printf("%s: WARNING: unable to register output helper hook\n",
329 __func__);
330
331 /* Skip initialization of globals for non-default instances. */
332 #ifdef VIMAGE
333 if (!IS_DEFAULT_VNET(curvnet)) {
334 netisr_register_vnet(&ip_nh);
335 #ifdef RSS
336 netisr_register_vnet(&ip_direct_nh);
337 #endif
338 return;
339 }
340 #endif
341
342 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
343 if (pr == NULL)
344 panic("ip_init: PF_INET not found");
345
346 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
347 for (i = 0; i < IPPROTO_MAX; i++)
348 ip_protox[i] = pr - inetsw;
349 /*
350 * Cycle through IP protocols and put them into the appropriate place
351 * in ip_protox[].
352 */
353 for (pr = inetdomain.dom_protosw;
354 pr < inetdomain.dom_protoswNPROTOSW; pr++)
355 if (pr->pr_domain->dom_family == PF_INET &&
356 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
357 /* Be careful to only index valid IP protocols. */
358 if (pr->pr_protocol < IPPROTO_MAX)
359 ip_protox[pr->pr_protocol] = pr - inetsw;
360 }
361
362 netisr_register(&ip_nh);
363 #ifdef RSS
364 netisr_register(&ip_direct_nh);
365 #endif
366 }
367
368 #ifdef VIMAGE
369 static void
370 ip_destroy(void *unused __unused)
371 {
372 struct ifnet *ifp;
373 int error;
374
375 #ifdef RSS
376 netisr_unregister_vnet(&ip_direct_nh);
377 #endif
378 netisr_unregister_vnet(&ip_nh);
379
380 if ((error = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
381 printf("%s: WARNING: unable to unregister pfil hook, "
382 "error %d\n", __func__, error);
383
384 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
385 if (error != 0) {
386 printf("%s: WARNING: unable to deregister input helper hook "
387 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
388 "error %d returned\n", __func__, error);
389 }
390 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
391 if (error != 0) {
392 printf("%s: WARNING: unable to deregister output helper hook "
393 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
394 "error %d returned\n", __func__, error);
395 }
396
397 /* Remove the IPv4 addresses from all interfaces. */
398 in_ifscrub_all();
399
400 /* Make sure the IPv4 routes are gone as well. */
401 IFNET_RLOCK();
402 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
403 rt_flushifroutes_af(ifp, AF_INET);
404 IFNET_RUNLOCK();
405
406 /* Destroy IP reassembly queue. */
407 ipreass_destroy();
408
409 /* Cleanup in_ifaddr hash table; should be empty. */
410 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
411 }
412
413 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
414 #endif
415
416 #ifdef RSS
417 /*
418 * IP direct input routine.
419 *
420 * This is called when reinjecting completed fragments where
421 * all of the previous checking and book-keeping has been done.
422 */
423 void
424 ip_direct_input(struct mbuf *m)
425 {
426 struct ip *ip;
427 int hlen;
428
429 ip = mtod(m, struct ip *);
430 hlen = ip->ip_hl << 2;
431
432 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
433 if (IPSEC_ENABLED(ipv4)) {
434 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
435 return;
436 }
437 #endif /* IPSEC */
438 IPSTAT_INC(ips_delivered);
439 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
440 return;
441 }
442 #endif
443
444 /*
445 * Ip input routine. Checksum and byte swap header. If fragmented
446 * try to reassemble. Process options. Pass to next level.
447 */
448 void
449 ip_input(struct mbuf *m)
450 {
451 struct ip *ip = NULL;
452 struct in_ifaddr *ia = NULL;
453 struct ifaddr *ifa;
454 struct ifnet *ifp;
455 int checkif, hlen = 0;
456 uint16_t sum, ip_len;
457 int dchg = 0; /* dest changed after fw */
458 struct in_addr odst; /* original dst address */
459
460 M_ASSERTPKTHDR(m);
461
462 if (m->m_flags & M_FASTFWD_OURS) {
463 m->m_flags &= ~M_FASTFWD_OURS;
464 /* Set up some basics that will be used later. */
465 ip = mtod(m, struct ip *);
466 hlen = ip->ip_hl << 2;
467 ip_len = ntohs(ip->ip_len);
468 goto ours;
469 }
470
471 IPSTAT_INC(ips_total);
472
473 if (m->m_pkthdr.len < sizeof(struct ip))
474 goto tooshort;
475
476 if (m->m_len < sizeof (struct ip) &&
477 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
478 IPSTAT_INC(ips_toosmall);
479 return;
480 }
481 ip = mtod(m, struct ip *);
482
483 if (ip->ip_v != IPVERSION) {
484 IPSTAT_INC(ips_badvers);
485 goto bad;
486 }
487
488 hlen = ip->ip_hl << 2;
489 if (hlen < sizeof(struct ip)) { /* minimum header length */
490 IPSTAT_INC(ips_badhlen);
491 goto bad;
492 }
493 if (hlen > m->m_len) {
494 if ((m = m_pullup(m, hlen)) == NULL) {
495 IPSTAT_INC(ips_badhlen);
496 return;
497 }
498 ip = mtod(m, struct ip *);
499 }
500
501 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
502
503 /* 127/8 must not appear on wire - RFC1122 */
504 ifp = m->m_pkthdr.rcvif;
505 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
506 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
507 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
508 IPSTAT_INC(ips_badaddr);
509 goto bad;
510 }
511 }
512
513 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
514 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
515 } else {
516 if (hlen == sizeof(struct ip)) {
517 sum = in_cksum_hdr(ip);
518 } else {
519 sum = in_cksum(m, hlen);
520 }
521 }
522 if (sum) {
523 IPSTAT_INC(ips_badsum);
524 goto bad;
525 }
526
527 #ifdef ALTQ
528 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
529 /* packet is dropped by traffic conditioner */
530 return;
531 #endif
532
533 ip_len = ntohs(ip->ip_len);
534 if (ip_len < hlen) {
535 IPSTAT_INC(ips_badlen);
536 goto bad;
537 }
538
539 /*
540 * Check that the amount of data in the buffers
541 * is as at least much as the IP header would have us expect.
542 * Trim mbufs if longer than we expect.
543 * Drop packet if shorter than we expect.
544 */
545 if (m->m_pkthdr.len < ip_len) {
546 tooshort:
547 IPSTAT_INC(ips_tooshort);
548 goto bad;
549 }
550 if (m->m_pkthdr.len > ip_len) {
551 if (m->m_len == m->m_pkthdr.len) {
552 m->m_len = ip_len;
553 m->m_pkthdr.len = ip_len;
554 } else
555 m_adj(m, ip_len - m->m_pkthdr.len);
556 }
557
558 /*
559 * Try to forward the packet, but if we fail continue.
560 * ip_tryforward() does not generate redirects, so fall
561 * through to normal processing if redirects are required.
562 * ip_tryforward() does inbound and outbound packet firewall
563 * processing. If firewall has decided that destination becomes
564 * our local address, it sets M_FASTFWD_OURS flag. In this
565 * case skip another inbound firewall processing and update
566 * ip pointer.
567 */
568 if (V_ipforwarding != 0 && V_ipsendredirects == 0
569 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
570 && (!IPSEC_ENABLED(ipv4) ||
571 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
572 #endif
573 ) {
574 if ((m = ip_tryforward(m)) == NULL)
575 return;
576 if (m->m_flags & M_FASTFWD_OURS) {
577 m->m_flags &= ~M_FASTFWD_OURS;
578 ip = mtod(m, struct ip *);
579 goto ours;
580 }
581 }
582
583 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
584 /*
585 * Bypass packet filtering for packets previously handled by IPsec.
586 */
587 if (IPSEC_ENABLED(ipv4) &&
588 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
589 goto passin;
590 #endif
591
592 /*
593 * Run through list of hooks for input packets.
594 *
595 * NB: Beware of the destination address changing (e.g.
596 * by NAT rewriting). When this happens, tell
597 * ip_forward to do the right thing.
598 */
599
600 /* Jump over all PFIL processing if hooks are not active. */
601 if (!PFIL_HOOKED(&V_inet_pfil_hook))
602 goto passin;
603
604 odst = ip->ip_dst;
605 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, 0, NULL) != 0)
606 return;
607 if (m == NULL) /* consumed by filter */
608 return;
609
610 ip = mtod(m, struct ip *);
611 dchg = (odst.s_addr != ip->ip_dst.s_addr);
612 ifp = m->m_pkthdr.rcvif;
613
614 if (m->m_flags & M_FASTFWD_OURS) {
615 m->m_flags &= ~M_FASTFWD_OURS;
616 goto ours;
617 }
618 if (m->m_flags & M_IP_NEXTHOP) {
619 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
620 /*
621 * Directly ship the packet on. This allows
622 * forwarding packets originally destined to us
623 * to some other directly connected host.
624 */
625 ip_forward(m, 1);
626 return;
627 }
628 }
629 passin:
630
631 /*
632 * Process options and, if not destined for us,
633 * ship it on. ip_dooptions returns 1 when an
634 * error was detected (causing an icmp message
635 * to be sent and the original packet to be freed).
636 */
637 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
638 return;
639
640 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
641 * matter if it is destined to another node, or whether it is
642 * a multicast one, RSVP wants it! and prevents it from being forwarded
643 * anywhere else. Also checks if the rsvp daemon is running before
644 * grabbing the packet.
645 */
646 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
647 goto ours;
648
649 /*
650 * Check our list of addresses, to see if the packet is for us.
651 * If we don't have any addresses, assume any unicast packet
652 * we receive might be for us (and let the upper layers deal
653 * with it).
654 */
655 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
656 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
657 goto ours;
658
659 /*
660 * Enable a consistency check between the destination address
661 * and the arrival interface for a unicast packet (the RFC 1122
662 * strong ES model) if IP forwarding is disabled and the packet
663 * is not locally generated and the packet is not subject to
664 * 'ipfw fwd'.
665 *
666 * XXX - Checking also should be disabled if the destination
667 * address is ipnat'ed to a different interface.
668 *
669 * XXX - Checking is incompatible with IP aliases added
670 * to the loopback interface instead of the interface where
671 * the packets are received.
672 *
673 * XXX - This is the case for carp vhost IPs as well so we
674 * insert a workaround. If the packet got here, we already
675 * checked with carp_iamatch() and carp_forus().
676 */
677 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
678 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
679 ifp->if_carp == NULL && (dchg == 0);
680
681 /*
682 * Check for exact addresses in the hash bucket.
683 */
684 /* IN_IFADDR_RLOCK(); */
685 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
686 /*
687 * If the address matches, verify that the packet
688 * arrived via the correct interface if checking is
689 * enabled.
690 */
691 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
692 (!checkif || ia->ia_ifp == ifp)) {
693 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
694 counter_u64_add(ia->ia_ifa.ifa_ibytes,
695 m->m_pkthdr.len);
696 /* IN_IFADDR_RUNLOCK(); */
697 goto ours;
698 }
699 }
700 /* IN_IFADDR_RUNLOCK(); */
701
702 /*
703 * Check for broadcast addresses.
704 *
705 * Only accept broadcast packets that arrive via the matching
706 * interface. Reception of forwarded directed broadcasts would
707 * be handled via ip_forward() and ether_output() with the loopback
708 * into the stack for SIMPLEX interfaces handled by ether_output().
709 */
710 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
711 IF_ADDR_RLOCK(ifp);
712 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
713 if (ifa->ifa_addr->sa_family != AF_INET)
714 continue;
715 ia = ifatoia(ifa);
716 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
717 ip->ip_dst.s_addr) {
718 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
719 counter_u64_add(ia->ia_ifa.ifa_ibytes,
720 m->m_pkthdr.len);
721 IF_ADDR_RUNLOCK(ifp);
722 goto ours;
723 }
724 #ifdef BOOTP_COMPAT
725 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
726 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
727 counter_u64_add(ia->ia_ifa.ifa_ibytes,
728 m->m_pkthdr.len);
729 IF_ADDR_RUNLOCK(ifp);
730 goto ours;
731 }
732 #endif
733 }
734 IF_ADDR_RUNLOCK(ifp);
735 ia = NULL;
736 }
737 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
738 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
739 IPSTAT_INC(ips_cantforward);
740 m_freem(m);
741 return;
742 }
743 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
744 if (V_ip_mrouter) {
745 /*
746 * If we are acting as a multicast router, all
747 * incoming multicast packets are passed to the
748 * kernel-level multicast forwarding function.
749 * The packet is returned (relatively) intact; if
750 * ip_mforward() returns a non-zero value, the packet
751 * must be discarded, else it may be accepted below.
752 */
753 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
754 IPSTAT_INC(ips_cantforward);
755 m_freem(m);
756 return;
757 }
758
759 /*
760 * The process-level routing daemon needs to receive
761 * all multicast IGMP packets, whether or not this
762 * host belongs to their destination groups.
763 */
764 if (ip->ip_p == IPPROTO_IGMP)
765 goto ours;
766 IPSTAT_INC(ips_forward);
767 }
768 /*
769 * Assume the packet is for us, to avoid prematurely taking
770 * a lock on the in_multi hash. Protocols must perform
771 * their own filtering and update statistics accordingly.
772 */
773 goto ours;
774 }
775 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
776 goto ours;
777 if (ip->ip_dst.s_addr == INADDR_ANY)
778 goto ours;
779
780 /*
781 * Not for us; forward if possible and desirable.
782 */
783 if (V_ipforwarding == 0) {
784 IPSTAT_INC(ips_cantforward);
785 m_freem(m);
786 } else {
787 ip_forward(m, dchg);
788 }
789 return;
790
791 ours:
792 #ifdef IPSTEALTH
793 /*
794 * IPSTEALTH: Process non-routing options only
795 * if the packet is destined for us.
796 */
797 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
798 return;
799 #endif /* IPSTEALTH */
800
801 /*
802 * Attempt reassembly; if it succeeds, proceed.
803 * ip_reass() will return a different mbuf.
804 */
805 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
806 /* XXXGL: shouldn't we save & set m_flags? */
807 m = ip_reass(m);
808 if (m == NULL)
809 return;
810 ip = mtod(m, struct ip *);
811 /* Get the header length of the reassembled packet */
812 hlen = ip->ip_hl << 2;
813 }
814
815 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
816 if (IPSEC_ENABLED(ipv4)) {
817 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
818 return;
819 }
820 #endif /* IPSEC */
821
822 /*
823 * Switch out to protocol's input routine.
824 */
825 IPSTAT_INC(ips_delivered);
826
827 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
828 return;
829 bad:
830 m_freem(m);
831 }
832
833 /*
834 * IP timer processing;
835 * if a timer expires on a reassembly
836 * queue, discard it.
837 */
838 void
839 ip_slowtimo(void)
840 {
841 VNET_ITERATOR_DECL(vnet_iter);
842
843 VNET_LIST_RLOCK_NOSLEEP();
844 VNET_FOREACH(vnet_iter) {
845 CURVNET_SET(vnet_iter);
846 ipreass_slowtimo();
847 CURVNET_RESTORE();
848 }
849 VNET_LIST_RUNLOCK_NOSLEEP();
850 }
851
852 void
853 ip_drain(void)
854 {
855 VNET_ITERATOR_DECL(vnet_iter);
856
857 VNET_LIST_RLOCK_NOSLEEP();
858 VNET_FOREACH(vnet_iter) {
859 CURVNET_SET(vnet_iter);
860 ipreass_drain();
861 CURVNET_RESTORE();
862 }
863 VNET_LIST_RUNLOCK_NOSLEEP();
864 }
865
866 /*
867 * The protocol to be inserted into ip_protox[] must be already registered
868 * in inetsw[], either statically or through pf_proto_register().
869 */
870 int
871 ipproto_register(short ipproto)
872 {
873 struct protosw *pr;
874
875 /* Sanity checks. */
876 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
877 return (EPROTONOSUPPORT);
878
879 /*
880 * The protocol slot must not be occupied by another protocol
881 * already. An index pointing to IPPROTO_RAW is unused.
882 */
883 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
884 if (pr == NULL)
885 return (EPFNOSUPPORT);
886 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
887 return (EEXIST);
888
889 /* Find the protocol position in inetsw[] and set the index. */
890 for (pr = inetdomain.dom_protosw;
891 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
892 if (pr->pr_domain->dom_family == PF_INET &&
893 pr->pr_protocol && pr->pr_protocol == ipproto) {
894 ip_protox[pr->pr_protocol] = pr - inetsw;
895 return (0);
896 }
897 }
898 return (EPROTONOSUPPORT);
899 }
900
901 int
902 ipproto_unregister(short ipproto)
903 {
904 struct protosw *pr;
905
906 /* Sanity checks. */
907 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
908 return (EPROTONOSUPPORT);
909
910 /* Check if the protocol was indeed registered. */
911 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
912 if (pr == NULL)
913 return (EPFNOSUPPORT);
914 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
915 return (ENOENT);
916
917 /* Reset the protocol slot to IPPROTO_RAW. */
918 ip_protox[ipproto] = pr - inetsw;
919 return (0);
920 }
921
922 u_char inetctlerrmap[PRC_NCMDS] = {
923 0, 0, 0, 0,
924 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
925 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
926 EMSGSIZE, EHOSTUNREACH, 0, 0,
927 0, 0, EHOSTUNREACH, 0,
928 ENOPROTOOPT, ECONNREFUSED
929 };
930
931 /*
932 * Forward a packet. If some error occurs return the sender
933 * an icmp packet. Note we can't always generate a meaningful
934 * icmp message because icmp doesn't have a large enough repertoire
935 * of codes and types.
936 *
937 * If not forwarding, just drop the packet. This could be confusing
938 * if ipforwarding was zero but some routing protocol was advancing
939 * us as a gateway to somewhere. However, we must let the routing
940 * protocol deal with that.
941 *
942 * The srcrt parameter indicates whether the packet is being forwarded
943 * via a source route.
944 */
945 void
946 ip_forward(struct mbuf *m, int srcrt)
947 {
948 struct ip *ip = mtod(m, struct ip *);
949 struct in_ifaddr *ia;
950 struct mbuf *mcopy;
951 struct sockaddr_in *sin;
952 struct in_addr dest;
953 struct route ro;
954 int error, type = 0, code = 0, mtu = 0;
955
956 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
957 IPSTAT_INC(ips_cantforward);
958 m_freem(m);
959 return;
960 }
961 if (
962 #ifdef IPSTEALTH
963 V_ipstealth == 0 &&
964 #endif
965 ip->ip_ttl <= IPTTLDEC) {
966 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
967 return;
968 }
969
970 bzero(&ro, sizeof(ro));
971 sin = (struct sockaddr_in *)&ro.ro_dst;
972 sin->sin_family = AF_INET;
973 sin->sin_len = sizeof(*sin);
974 sin->sin_addr = ip->ip_dst;
975 #ifdef RADIX_MPATH
976 rtalloc_mpath_fib(&ro,
977 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
978 M_GETFIB(m));
979 #else
980 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
981 #endif
982 NET_EPOCH_ENTER();
983 if (ro.ro_rt != NULL) {
984 ia = ifatoia(ro.ro_rt->rt_ifa);
985 } else
986 ia = NULL;
987 /*
988 * Save the IP header and at most 8 bytes of the payload,
989 * in case we need to generate an ICMP message to the src.
990 *
991 * XXX this can be optimized a lot by saving the data in a local
992 * buffer on the stack (72 bytes at most), and only allocating the
993 * mbuf if really necessary. The vast majority of the packets
994 * are forwarded without having to send an ICMP back (either
995 * because unnecessary, or because rate limited), so we are
996 * really we are wasting a lot of work here.
997 *
998 * We don't use m_copym() because it might return a reference
999 * to a shared cluster. Both this function and ip_output()
1000 * assume exclusive access to the IP header in `m', so any
1001 * data in a cluster may change before we reach icmp_error().
1002 */
1003 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1004 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1005 /*
1006 * It's probably ok if the pkthdr dup fails (because
1007 * the deep copy of the tag chain failed), but for now
1008 * be conservative and just discard the copy since
1009 * code below may some day want the tags.
1010 */
1011 m_free(mcopy);
1012 mcopy = NULL;
1013 }
1014 if (mcopy != NULL) {
1015 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1016 mcopy->m_pkthdr.len = mcopy->m_len;
1017 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1018 }
1019 #ifdef IPSTEALTH
1020 if (V_ipstealth == 0)
1021 #endif
1022 ip->ip_ttl -= IPTTLDEC;
1023 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1024 if (IPSEC_ENABLED(ipv4)) {
1025 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1026 /* mbuf consumed by IPsec */
1027 m_freem(mcopy);
1028 if (error != EINPROGRESS)
1029 IPSTAT_INC(ips_cantforward);
1030 goto out;
1031 }
1032 /* No IPsec processing required */
1033 }
1034 #endif /* IPSEC */
1035 /*
1036 * If forwarding packet using same interface that it came in on,
1037 * perhaps should send a redirect to sender to shortcut a hop.
1038 * Only send redirect if source is sending directly to us,
1039 * and if packet was not source routed (or has any options).
1040 * Also, don't send redirect if forwarding using a default route
1041 * or a route modified by a redirect.
1042 */
1043 dest.s_addr = 0;
1044 if (!srcrt && V_ipsendredirects &&
1045 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1046 struct rtentry *rt;
1047
1048 rt = ro.ro_rt;
1049
1050 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1051 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1052 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1053 u_long src = ntohl(ip->ip_src.s_addr);
1054
1055 if (RTA(rt) &&
1056 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1057 if (rt->rt_flags & RTF_GATEWAY)
1058 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1059 else
1060 dest.s_addr = ip->ip_dst.s_addr;
1061 /* Router requirements says to only send host redirects */
1062 type = ICMP_REDIRECT;
1063 code = ICMP_REDIRECT_HOST;
1064 }
1065 }
1066 }
1067
1068 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1069
1070 if (error == EMSGSIZE && ro.ro_rt)
1071 mtu = ro.ro_rt->rt_mtu;
1072 RO_RTFREE(&ro);
1073
1074 if (error)
1075 IPSTAT_INC(ips_cantforward);
1076 else {
1077 IPSTAT_INC(ips_forward);
1078 if (type)
1079 IPSTAT_INC(ips_redirectsent);
1080 else {
1081 if (mcopy)
1082 m_freem(mcopy);
1083 goto out;
1084 }
1085 }
1086 if (mcopy == NULL)
1087 goto out;
1088
1089
1090 switch (error) {
1091
1092 case 0: /* forwarded, but need redirect */
1093 /* type, code set above */
1094 break;
1095
1096 case ENETUNREACH:
1097 case EHOSTUNREACH:
1098 case ENETDOWN:
1099 case EHOSTDOWN:
1100 default:
1101 type = ICMP_UNREACH;
1102 code = ICMP_UNREACH_HOST;
1103 break;
1104
1105 case EMSGSIZE:
1106 type = ICMP_UNREACH;
1107 code = ICMP_UNREACH_NEEDFRAG;
1108 /*
1109 * If the MTU was set before make sure we are below the
1110 * interface MTU.
1111 * If the MTU wasn't set before use the interface mtu or
1112 * fall back to the next smaller mtu step compared to the
1113 * current packet size.
1114 */
1115 if (mtu != 0) {
1116 if (ia != NULL)
1117 mtu = min(mtu, ia->ia_ifp->if_mtu);
1118 } else {
1119 if (ia != NULL)
1120 mtu = ia->ia_ifp->if_mtu;
1121 else
1122 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1123 }
1124 IPSTAT_INC(ips_cantfrag);
1125 break;
1126
1127 case ENOBUFS:
1128 case EACCES: /* ipfw denied packet */
1129 m_freem(mcopy);
1130 goto out;
1131 }
1132 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1133 out:
1134 NET_EPOCH_EXIT();
1135 }
1136
1137 #define CHECK_SO_CT(sp, ct) \
1138 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1139
1140 void
1141 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1142 struct mbuf *m)
1143 {
1144 bool stamped;
1145
1146 stamped = false;
1147 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1148 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1149 struct bintime boottimebin, bt;
1150 struct timespec ts1;
1151
1152 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1153 M_TSTMP)) {
1154 mbuf_tstmp2timespec(m, &ts1);
1155 timespec2bintime(&ts1, &bt);
1156 getboottimebin(&boottimebin);
1157 bintime_add(&bt, &boottimebin);
1158 } else {
1159 bintime(&bt);
1160 }
1161 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1162 SCM_BINTIME, SOL_SOCKET);
1163 if (*mp != NULL) {
1164 mp = &(*mp)->m_next;
1165 stamped = true;
1166 }
1167 }
1168 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1169 struct bintime boottimebin, bt1;
1170 struct timespec ts1;;
1171 struct timeval tv;
1172
1173 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1174 M_TSTMP)) {
1175 mbuf_tstmp2timespec(m, &ts1);
1176 timespec2bintime(&ts1, &bt1);
1177 getboottimebin(&boottimebin);
1178 bintime_add(&bt1, &boottimebin);
1179 bintime2timeval(&bt1, &tv);
1180 } else {
1181 microtime(&tv);
1182 }
1183 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1184 SCM_TIMESTAMP, SOL_SOCKET);
1185 if (*mp != NULL) {
1186 mp = &(*mp)->m_next;
1187 stamped = true;
1188 }
1189 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1190 struct bintime boottimebin;
1191 struct timespec ts, ts1;
1192
1193 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1194 M_TSTMP)) {
1195 mbuf_tstmp2timespec(m, &ts);
1196 getboottimebin(&boottimebin);
1197 bintime2timespec(&boottimebin, &ts1);
1198 timespecadd(&ts, &ts1, &ts);
1199 } else {
1200 nanotime(&ts);
1201 }
1202 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1203 SCM_REALTIME, SOL_SOCKET);
1204 if (*mp != NULL) {
1205 mp = &(*mp)->m_next;
1206 stamped = true;
1207 }
1208 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1209 struct timespec ts;
1210
1211 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1212 M_TSTMP))
1213 mbuf_tstmp2timespec(m, &ts);
1214 else
1215 nanouptime(&ts);
1216 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1217 SCM_MONOTONIC, SOL_SOCKET);
1218 if (*mp != NULL) {
1219 mp = &(*mp)->m_next;
1220 stamped = true;
1221 }
1222 }
1223 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1224 M_TSTMP)) {
1225 struct sock_timestamp_info sti;
1226
1227 bzero(&sti, sizeof(sti));
1228 sti.st_info_flags = ST_INFO_HW;
1229 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1230 sti.st_info_flags |= ST_INFO_HW_HPREC;
1231 *mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1232 SOL_SOCKET);
1233 if (*mp != NULL)
1234 mp = &(*mp)->m_next;
1235 }
1236 if (inp->inp_flags & INP_RECVDSTADDR) {
1237 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1238 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1239 if (*mp)
1240 mp = &(*mp)->m_next;
1241 }
1242 if (inp->inp_flags & INP_RECVTTL) {
1243 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1244 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1245 if (*mp)
1246 mp = &(*mp)->m_next;
1247 }
1248 #ifdef notyet
1249 /* XXX
1250 * Moving these out of udp_input() made them even more broken
1251 * than they already were.
1252 */
1253 /* options were tossed already */
1254 if (inp->inp_flags & INP_RECVOPTS) {
1255 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1256 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1257 if (*mp)
1258 mp = &(*mp)->m_next;
1259 }
1260 /* ip_srcroute doesn't do what we want here, need to fix */
1261 if (inp->inp_flags & INP_RECVRETOPTS) {
1262 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1263 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1264 if (*mp)
1265 mp = &(*mp)->m_next;
1266 }
1267 #endif
1268 if (inp->inp_flags & INP_RECVIF) {
1269 struct ifnet *ifp;
1270 struct sdlbuf {
1271 struct sockaddr_dl sdl;
1272 u_char pad[32];
1273 } sdlbuf;
1274 struct sockaddr_dl *sdp;
1275 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1276
1277 if ((ifp = m->m_pkthdr.rcvif) &&
1278 ifp->if_index && ifp->if_index <= V_if_index) {
1279 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1280 /*
1281 * Change our mind and don't try copy.
1282 */
1283 if (sdp->sdl_family != AF_LINK ||
1284 sdp->sdl_len > sizeof(sdlbuf)) {
1285 goto makedummy;
1286 }
1287 bcopy(sdp, sdl2, sdp->sdl_len);
1288 } else {
1289 makedummy:
1290 sdl2->sdl_len =
1291 offsetof(struct sockaddr_dl, sdl_data[0]);
1292 sdl2->sdl_family = AF_LINK;
1293 sdl2->sdl_index = 0;
1294 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1295 }
1296 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1297 IP_RECVIF, IPPROTO_IP);
1298 if (*mp)
1299 mp = &(*mp)->m_next;
1300 }
1301 if (inp->inp_flags & INP_RECVTOS) {
1302 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1303 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1304 if (*mp)
1305 mp = &(*mp)->m_next;
1306 }
1307
1308 if (inp->inp_flags2 & INP_RECVFLOWID) {
1309 uint32_t flowid, flow_type;
1310
1311 flowid = m->m_pkthdr.flowid;
1312 flow_type = M_HASHTYPE_GET(m);
1313
1314 /*
1315 * XXX should handle the failure of one or the
1316 * other - don't populate both?
1317 */
1318 *mp = sbcreatecontrol((caddr_t) &flowid,
1319 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1320 if (*mp)
1321 mp = &(*mp)->m_next;
1322 *mp = sbcreatecontrol((caddr_t) &flow_type,
1323 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1324 if (*mp)
1325 mp = &(*mp)->m_next;
1326 }
1327
1328 #ifdef RSS
1329 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1330 uint32_t flowid, flow_type;
1331 uint32_t rss_bucketid;
1332
1333 flowid = m->m_pkthdr.flowid;
1334 flow_type = M_HASHTYPE_GET(m);
1335
1336 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1337 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1338 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1339 if (*mp)
1340 mp = &(*mp)->m_next;
1341 }
1342 }
1343 #endif
1344 }
1345
1346 /*
1347 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1348 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1349 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1350 * compiled.
1351 */
1352 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1353 VNET_DEFINE(struct socket *, ip_rsvpd);
1354
1355 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1356
1357 int
1358 ip_rsvp_init(struct socket *so)
1359 {
1360
1361 if (so->so_type != SOCK_RAW ||
1362 so->so_proto->pr_protocol != IPPROTO_RSVP)
1363 return EOPNOTSUPP;
1364
1365 if (V_ip_rsvpd != NULL)
1366 return EADDRINUSE;
1367
1368 V_ip_rsvpd = so;
1369 /*
1370 * This may seem silly, but we need to be sure we don't over-increment
1371 * the RSVP counter, in case something slips up.
1372 */
1373 if (!V_ip_rsvp_on) {
1374 V_ip_rsvp_on = 1;
1375 V_rsvp_on++;
1376 }
1377
1378 return 0;
1379 }
1380
1381 int
1382 ip_rsvp_done(void)
1383 {
1384
1385 V_ip_rsvpd = NULL;
1386 /*
1387 * This may seem silly, but we need to be sure we don't over-decrement
1388 * the RSVP counter, in case something slips up.
1389 */
1390 if (V_ip_rsvp_on) {
1391 V_ip_rsvp_on = 0;
1392 V_rsvp_on--;
1393 }
1394 return 0;
1395 }
1396
1397 int
1398 rsvp_input(struct mbuf **mp, int *offp, int proto)
1399 {
1400 struct mbuf *m;
1401
1402 m = *mp;
1403 *mp = NULL;
1404
1405 if (rsvp_input_p) { /* call the real one if loaded */
1406 *mp = m;
1407 rsvp_input_p(mp, offp, proto);
1408 return (IPPROTO_DONE);
1409 }
1410
1411 /* Can still get packets with rsvp_on = 0 if there is a local member
1412 * of the group to which the RSVP packet is addressed. But in this
1413 * case we want to throw the packet away.
1414 */
1415
1416 if (!V_rsvp_on) {
1417 m_freem(m);
1418 return (IPPROTO_DONE);
1419 }
1420
1421 if (V_ip_rsvpd != NULL) {
1422 *mp = m;
1423 rip_input(mp, offp, proto);
1424 return (IPPROTO_DONE);
1425 }
1426 /* Drop the packet */
1427 m_freem(m);
1428 return (IPPROTO_DONE);
1429 }
Cache object: a6459de590c9b17d264d2270b0593905
|