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: releng/10.2/sys/netinet/ip_input.c 266210 2014-05-16 05:05:53Z yongari $");
34
35 #include "opt_bootp.h"
36 #include "opt_ipfw.h"
37 #include "opt_ipstealth.h"
38 #include "opt_ipsec.h"
39 #include "opt_kdtrace.h"
40 #include "opt_route.h"
41
42 #include <sys/param.h>
43 #include <sys/systm.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/rwlock.h>
53 #include <sys/sdt.h>
54 #include <sys/syslog.h>
55 #include <sys/sysctl.h>
56
57 #include <net/pfil.h>
58 #include <net/if.h>
59 #include <net/if_types.h>
60 #include <net/if_var.h>
61 #include <net/if_dl.h>
62 #include <net/route.h>
63 #include <net/netisr.h>
64 #include <net/vnet.h>
65
66 #include <netinet/in.h>
67 #include <netinet/in_kdtrace.h>
68 #include <netinet/in_systm.h>
69 #include <netinet/in_var.h>
70 #include <netinet/ip.h>
71 #include <netinet/in_pcb.h>
72 #include <netinet/ip_var.h>
73 #include <netinet/ip_fw.h>
74 #include <netinet/ip_icmp.h>
75 #include <netinet/ip_options.h>
76 #include <machine/in_cksum.h>
77 #include <netinet/ip_carp.h>
78 #ifdef IPSEC
79 #include <netinet/ip_ipsec.h>
80 #endif /* IPSEC */
81
82 #include <sys/socketvar.h>
83
84 #include <security/mac/mac_framework.h>
85
86 #ifdef CTASSERT
87 CTASSERT(sizeof(struct ip) == 20);
88 #endif
89
90 struct rwlock in_ifaddr_lock;
91 RW_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
92
93 VNET_DEFINE(int, rsvp_on);
94
95 VNET_DEFINE(int, ipforwarding);
96 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
97 &VNET_NAME(ipforwarding), 0,
98 "Enable IP forwarding between interfaces");
99
100 static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */
101 #define V_ipsendredirects VNET(ipsendredirects)
102 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
103 &VNET_NAME(ipsendredirects), 0,
104 "Enable sending IP redirects");
105
106 static VNET_DEFINE(int, ip_keepfaith);
107 #define V_ip_keepfaith VNET(ip_keepfaith)
108 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
109 &VNET_NAME(ip_keepfaith), 0,
110 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
111
112 static VNET_DEFINE(int, ip_sendsourcequench);
113 #define V_ip_sendsourcequench VNET(ip_sendsourcequench)
114 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
115 &VNET_NAME(ip_sendsourcequench), 0,
116 "Enable the transmission of source quench packets");
117
118 VNET_DEFINE(int, ip_do_randomid);
119 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
120 &VNET_NAME(ip_do_randomid), 0,
121 "Assign random ip_id values");
122
123 /*
124 * XXX - Setting ip_checkinterface mostly implements the receive side of
125 * the Strong ES model described in RFC 1122, but since the routing table
126 * and transmit implementation do not implement the Strong ES model,
127 * setting this to 1 results in an odd hybrid.
128 *
129 * XXX - ip_checkinterface currently must be disabled if you use ipnat
130 * to translate the destination address to another local interface.
131 *
132 * XXX - ip_checkinterface must be disabled if you add IP aliases
133 * to the loopback interface instead of the interface where the
134 * packets for those addresses are received.
135 */
136 static VNET_DEFINE(int, ip_checkinterface);
137 #define V_ip_checkinterface VNET(ip_checkinterface)
138 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
139 &VNET_NAME(ip_checkinterface), 0,
140 "Verify packet arrives on correct interface");
141
142 VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */
143
144 static struct netisr_handler ip_nh = {
145 .nh_name = "ip",
146 .nh_handler = ip_input,
147 .nh_proto = NETISR_IP,
148 .nh_policy = NETISR_POLICY_FLOW,
149 };
150
151 extern struct domain inetdomain;
152 extern struct protosw inetsw[];
153 u_char ip_protox[IPPROTO_MAX];
154 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
155 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
156 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
157
158 static VNET_DEFINE(uma_zone_t, ipq_zone);
159 static VNET_DEFINE(TAILQ_HEAD(ipqhead, ipq), ipq[IPREASS_NHASH]);
160 static struct mtx ipqlock;
161
162 #define V_ipq_zone VNET(ipq_zone)
163 #define V_ipq VNET(ipq)
164
165 #define IPQ_LOCK() mtx_lock(&ipqlock)
166 #define IPQ_UNLOCK() mtx_unlock(&ipqlock)
167 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
168 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
169
170 static void maxnipq_update(void);
171 static void ipq_zone_change(void *);
172 static void ip_drain_locked(void);
173
174 static VNET_DEFINE(int, maxnipq); /* Administrative limit on # reass queues. */
175 static VNET_DEFINE(int, nipq); /* Total # of reass queues */
176 #define V_maxnipq VNET(maxnipq)
177 #define V_nipq VNET(nipq)
178 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD,
179 &VNET_NAME(nipq), 0,
180 "Current number of IPv4 fragment reassembly queue entries");
181
182 static VNET_DEFINE(int, maxfragsperpacket);
183 #define V_maxfragsperpacket VNET(maxfragsperpacket)
184 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
185 &VNET_NAME(maxfragsperpacket), 0,
186 "Maximum number of IPv4 fragments allowed per packet");
187
188 #ifdef IPCTL_DEFMTU
189 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
190 &ip_mtu, 0, "Default MTU");
191 #endif
192
193 #ifdef IPSTEALTH
194 VNET_DEFINE(int, ipstealth);
195 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
196 &VNET_NAME(ipstealth), 0,
197 "IP stealth mode, no TTL decrementation on forwarding");
198 #endif
199
200 static void ip_freef(struct ipqhead *, struct ipq *);
201
202 /*
203 * IP statistics are stored in the "array" of counter(9)s.
204 */
205 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
206 VNET_PCPUSTAT_SYSINIT(ipstat);
207 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
208 "IP statistics (struct ipstat, netinet/ip_var.h)");
209
210 #ifdef VIMAGE
211 VNET_PCPUSTAT_SYSUNINIT(ipstat);
212 #endif /* VIMAGE */
213
214 /*
215 * Kernel module interface for updating ipstat. The argument is an index
216 * into ipstat treated as an array.
217 */
218 void
219 kmod_ipstat_inc(int statnum)
220 {
221
222 counter_u64_add(VNET(ipstat)[statnum], 1);
223 }
224
225 void
226 kmod_ipstat_dec(int statnum)
227 {
228
229 counter_u64_add(VNET(ipstat)[statnum], -1);
230 }
231
232 static int
233 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
234 {
235 int error, qlimit;
236
237 netisr_getqlimit(&ip_nh, &qlimit);
238 error = sysctl_handle_int(oidp, &qlimit, 0, req);
239 if (error || !req->newptr)
240 return (error);
241 if (qlimit < 1)
242 return (EINVAL);
243 return (netisr_setqlimit(&ip_nh, qlimit));
244 }
245 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
246 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
247 "Maximum size of the IP input queue");
248
249 static int
250 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
251 {
252 u_int64_t qdrops_long;
253 int error, qdrops;
254
255 netisr_getqdrops(&ip_nh, &qdrops_long);
256 qdrops = qdrops_long;
257 error = sysctl_handle_int(oidp, &qdrops, 0, req);
258 if (error || !req->newptr)
259 return (error);
260 if (qdrops != 0)
261 return (EINVAL);
262 netisr_clearqdrops(&ip_nh);
263 return (0);
264 }
265
266 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
267 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I",
268 "Number of packets dropped from the IP input queue");
269
270 /*
271 * IP initialization: fill in IP protocol switch table.
272 * All protocols not implemented in kernel go to raw IP protocol handler.
273 */
274 void
275 ip_init(void)
276 {
277 struct protosw *pr;
278 int i;
279
280 V_ip_id = time_second & 0xffff;
281
282 TAILQ_INIT(&V_in_ifaddrhead);
283 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
284
285 /* Initialize IP reassembly queue. */
286 for (i = 0; i < IPREASS_NHASH; i++)
287 TAILQ_INIT(&V_ipq[i]);
288 V_maxnipq = nmbclusters / 32;
289 V_maxfragsperpacket = 16;
290 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
291 NULL, UMA_ALIGN_PTR, 0);
292 maxnipq_update();
293
294 /* Initialize packet filter hooks. */
295 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
296 V_inet_pfil_hook.ph_af = AF_INET;
297 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
298 printf("%s: WARNING: unable to register pfil hook, "
299 "error %d\n", __func__, i);
300
301 /* Skip initialization of globals for non-default instances. */
302 if (!IS_DEFAULT_VNET(curvnet))
303 return;
304
305 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
306 if (pr == NULL)
307 panic("ip_init: PF_INET not found");
308
309 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
310 for (i = 0; i < IPPROTO_MAX; i++)
311 ip_protox[i] = pr - inetsw;
312 /*
313 * Cycle through IP protocols and put them into the appropriate place
314 * in ip_protox[].
315 */
316 for (pr = inetdomain.dom_protosw;
317 pr < inetdomain.dom_protoswNPROTOSW; pr++)
318 if (pr->pr_domain->dom_family == PF_INET &&
319 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
320 /* Be careful to only index valid IP protocols. */
321 if (pr->pr_protocol < IPPROTO_MAX)
322 ip_protox[pr->pr_protocol] = pr - inetsw;
323 }
324
325 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
326 NULL, EVENTHANDLER_PRI_ANY);
327
328 /* Initialize various other remaining things. */
329 IPQ_LOCK_INIT();
330 netisr_register(&ip_nh);
331 }
332
333 #ifdef VIMAGE
334 void
335 ip_destroy(void)
336 {
337 int i;
338
339 if ((i = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
340 printf("%s: WARNING: unable to unregister pfil hook, "
341 "error %d\n", __func__, i);
342
343 /* Cleanup in_ifaddr hash table; should be empty. */
344 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
345
346 IPQ_LOCK();
347 ip_drain_locked();
348 IPQ_UNLOCK();
349
350 uma_zdestroy(V_ipq_zone);
351 }
352 #endif
353
354 /*
355 * Ip input routine. Checksum and byte swap header. If fragmented
356 * try to reassemble. Process options. Pass to next level.
357 */
358 void
359 ip_input(struct mbuf *m)
360 {
361 struct ip *ip = NULL;
362 struct in_ifaddr *ia = NULL;
363 struct ifaddr *ifa;
364 struct ifnet *ifp;
365 int checkif, hlen = 0;
366 uint16_t sum, ip_len;
367 int dchg = 0; /* dest changed after fw */
368 struct in_addr odst; /* original dst address */
369
370 M_ASSERTPKTHDR(m);
371
372 if (m->m_flags & M_FASTFWD_OURS) {
373 m->m_flags &= ~M_FASTFWD_OURS;
374 /* Set up some basics that will be used later. */
375 ip = mtod(m, struct ip *);
376 hlen = ip->ip_hl << 2;
377 ip_len = ntohs(ip->ip_len);
378 goto ours;
379 }
380
381 IPSTAT_INC(ips_total);
382
383 if (m->m_pkthdr.len < sizeof(struct ip))
384 goto tooshort;
385
386 if (m->m_len < sizeof (struct ip) &&
387 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
388 IPSTAT_INC(ips_toosmall);
389 return;
390 }
391 ip = mtod(m, struct ip *);
392
393 if (ip->ip_v != IPVERSION) {
394 IPSTAT_INC(ips_badvers);
395 goto bad;
396 }
397
398 hlen = ip->ip_hl << 2;
399 if (hlen < sizeof(struct ip)) { /* minimum header length */
400 IPSTAT_INC(ips_badhlen);
401 goto bad;
402 }
403 if (hlen > m->m_len) {
404 if ((m = m_pullup(m, hlen)) == NULL) {
405 IPSTAT_INC(ips_badhlen);
406 return;
407 }
408 ip = mtod(m, struct ip *);
409 }
410
411 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
412
413 /* 127/8 must not appear on wire - RFC1122 */
414 ifp = m->m_pkthdr.rcvif;
415 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
416 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
417 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
418 IPSTAT_INC(ips_badaddr);
419 goto bad;
420 }
421 }
422
423 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
424 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
425 } else {
426 if (hlen == sizeof(struct ip)) {
427 sum = in_cksum_hdr(ip);
428 } else {
429 sum = in_cksum(m, hlen);
430 }
431 }
432 if (sum) {
433 IPSTAT_INC(ips_badsum);
434 goto bad;
435 }
436
437 #ifdef ALTQ
438 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
439 /* packet is dropped by traffic conditioner */
440 return;
441 #endif
442
443 ip_len = ntohs(ip->ip_len);
444 if (ip_len < hlen) {
445 IPSTAT_INC(ips_badlen);
446 goto bad;
447 }
448
449 /*
450 * Check that the amount of data in the buffers
451 * is as at least much as the IP header would have us expect.
452 * Trim mbufs if longer than we expect.
453 * Drop packet if shorter than we expect.
454 */
455 if (m->m_pkthdr.len < ip_len) {
456 tooshort:
457 IPSTAT_INC(ips_tooshort);
458 goto bad;
459 }
460 if (m->m_pkthdr.len > ip_len) {
461 if (m->m_len == m->m_pkthdr.len) {
462 m->m_len = ip_len;
463 m->m_pkthdr.len = ip_len;
464 } else
465 m_adj(m, ip_len - m->m_pkthdr.len);
466 }
467 #ifdef IPSEC
468 /*
469 * Bypass packet filtering for packets previously handled by IPsec.
470 */
471 if (ip_ipsec_filtertunnel(m))
472 goto passin;
473 #endif /* IPSEC */
474
475 /*
476 * Run through list of hooks for input packets.
477 *
478 * NB: Beware of the destination address changing (e.g.
479 * by NAT rewriting). When this happens, tell
480 * ip_forward to do the right thing.
481 */
482
483 /* Jump over all PFIL processing if hooks are not active. */
484 if (!PFIL_HOOKED(&V_inet_pfil_hook))
485 goto passin;
486
487 odst = ip->ip_dst;
488 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
489 return;
490 if (m == NULL) /* consumed by filter */
491 return;
492
493 ip = mtod(m, struct ip *);
494 dchg = (odst.s_addr != ip->ip_dst.s_addr);
495 ifp = m->m_pkthdr.rcvif;
496
497 if (m->m_flags & M_FASTFWD_OURS) {
498 m->m_flags &= ~M_FASTFWD_OURS;
499 goto ours;
500 }
501 if (m->m_flags & M_IP_NEXTHOP) {
502 dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL);
503 if (dchg != 0) {
504 /*
505 * Directly ship the packet on. This allows
506 * forwarding packets originally destined to us
507 * to some other directly connected host.
508 */
509 ip_forward(m, 1);
510 return;
511 }
512 }
513 passin:
514
515 /*
516 * Process options and, if not destined for us,
517 * ship it on. ip_dooptions returns 1 when an
518 * error was detected (causing an icmp message
519 * to be sent and the original packet to be freed).
520 */
521 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
522 return;
523
524 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
525 * matter if it is destined to another node, or whether it is
526 * a multicast one, RSVP wants it! and prevents it from being forwarded
527 * anywhere else. Also checks if the rsvp daemon is running before
528 * grabbing the packet.
529 */
530 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
531 goto ours;
532
533 /*
534 * Check our list of addresses, to see if the packet is for us.
535 * If we don't have any addresses, assume any unicast packet
536 * we receive might be for us (and let the upper layers deal
537 * with it).
538 */
539 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
540 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
541 goto ours;
542
543 /*
544 * Enable a consistency check between the destination address
545 * and the arrival interface for a unicast packet (the RFC 1122
546 * strong ES model) if IP forwarding is disabled and the packet
547 * is not locally generated and the packet is not subject to
548 * 'ipfw fwd'.
549 *
550 * XXX - Checking also should be disabled if the destination
551 * address is ipnat'ed to a different interface.
552 *
553 * XXX - Checking is incompatible with IP aliases added
554 * to the loopback interface instead of the interface where
555 * the packets are received.
556 *
557 * XXX - This is the case for carp vhost IPs as well so we
558 * insert a workaround. If the packet got here, we already
559 * checked with carp_iamatch() and carp_forus().
560 */
561 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
562 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
563 ifp->if_carp == NULL && (dchg == 0);
564
565 /*
566 * Check for exact addresses in the hash bucket.
567 */
568 /* IN_IFADDR_RLOCK(); */
569 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
570 /*
571 * If the address matches, verify that the packet
572 * arrived via the correct interface if checking is
573 * enabled.
574 */
575 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
576 (!checkif || ia->ia_ifp == ifp)) {
577 ifa_ref(&ia->ia_ifa);
578 /* IN_IFADDR_RUNLOCK(); */
579 goto ours;
580 }
581 }
582 /* IN_IFADDR_RUNLOCK(); */
583
584 /*
585 * Check for broadcast addresses.
586 *
587 * Only accept broadcast packets that arrive via the matching
588 * interface. Reception of forwarded directed broadcasts would
589 * be handled via ip_forward() and ether_output() with the loopback
590 * into the stack for SIMPLEX interfaces handled by ether_output().
591 */
592 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
593 IF_ADDR_RLOCK(ifp);
594 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
595 if (ifa->ifa_addr->sa_family != AF_INET)
596 continue;
597 ia = ifatoia(ifa);
598 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
599 ip->ip_dst.s_addr) {
600 ifa_ref(ifa);
601 IF_ADDR_RUNLOCK(ifp);
602 goto ours;
603 }
604 #ifdef BOOTP_COMPAT
605 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
606 ifa_ref(ifa);
607 IF_ADDR_RUNLOCK(ifp);
608 goto ours;
609 }
610 #endif
611 }
612 IF_ADDR_RUNLOCK(ifp);
613 ia = NULL;
614 }
615 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
616 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
617 IPSTAT_INC(ips_cantforward);
618 m_freem(m);
619 return;
620 }
621 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
622 if (V_ip_mrouter) {
623 /*
624 * If we are acting as a multicast router, all
625 * incoming multicast packets are passed to the
626 * kernel-level multicast forwarding function.
627 * The packet is returned (relatively) intact; if
628 * ip_mforward() returns a non-zero value, the packet
629 * must be discarded, else it may be accepted below.
630 */
631 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
632 IPSTAT_INC(ips_cantforward);
633 m_freem(m);
634 return;
635 }
636
637 /*
638 * The process-level routing daemon needs to receive
639 * all multicast IGMP packets, whether or not this
640 * host belongs to their destination groups.
641 */
642 if (ip->ip_p == IPPROTO_IGMP)
643 goto ours;
644 IPSTAT_INC(ips_forward);
645 }
646 /*
647 * Assume the packet is for us, to avoid prematurely taking
648 * a lock on the in_multi hash. Protocols must perform
649 * their own filtering and update statistics accordingly.
650 */
651 goto ours;
652 }
653 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
654 goto ours;
655 if (ip->ip_dst.s_addr == INADDR_ANY)
656 goto ours;
657
658 /*
659 * FAITH(Firewall Aided Internet Translator)
660 */
661 if (ifp && ifp->if_type == IFT_FAITH) {
662 if (V_ip_keepfaith) {
663 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
664 goto ours;
665 }
666 m_freem(m);
667 return;
668 }
669
670 /*
671 * Not for us; forward if possible and desirable.
672 */
673 if (V_ipforwarding == 0) {
674 IPSTAT_INC(ips_cantforward);
675 m_freem(m);
676 } else {
677 #ifdef IPSEC
678 if (ip_ipsec_fwd(m))
679 goto bad;
680 #endif /* IPSEC */
681 ip_forward(m, dchg);
682 }
683 return;
684
685 ours:
686 #ifdef IPSTEALTH
687 /*
688 * IPSTEALTH: Process non-routing options only
689 * if the packet is destined for us.
690 */
691 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) {
692 if (ia != NULL)
693 ifa_free(&ia->ia_ifa);
694 return;
695 }
696 #endif /* IPSTEALTH */
697
698 /* Count the packet in the ip address stats */
699 if (ia != NULL) {
700 ia->ia_ifa.if_ipackets++;
701 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
702 ifa_free(&ia->ia_ifa);
703 }
704
705 /*
706 * Attempt reassembly; if it succeeds, proceed.
707 * ip_reass() will return a different mbuf.
708 */
709 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
710 /* XXXGL: shouldn't we save & set m_flags? */
711 m = ip_reass(m);
712 if (m == NULL)
713 return;
714 ip = mtod(m, struct ip *);
715 /* Get the header length of the reassembled packet */
716 hlen = ip->ip_hl << 2;
717 }
718
719 #ifdef IPSEC
720 /*
721 * enforce IPsec policy checking if we are seeing last header.
722 * note that we do not visit this with protocols with pcb layer
723 * code - like udp/tcp/raw ip.
724 */
725 if (ip_ipsec_input(m))
726 goto bad;
727 #endif /* IPSEC */
728
729 /*
730 * Switch out to protocol's input routine.
731 */
732 IPSTAT_INC(ips_delivered);
733
734 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
735 return;
736 bad:
737 m_freem(m);
738 }
739
740 /*
741 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
742 * max has slightly different semantics than the sysctl, for historical
743 * reasons.
744 */
745 static void
746 maxnipq_update(void)
747 {
748
749 /*
750 * -1 for unlimited allocation.
751 */
752 if (V_maxnipq < 0)
753 uma_zone_set_max(V_ipq_zone, 0);
754 /*
755 * Positive number for specific bound.
756 */
757 if (V_maxnipq > 0)
758 uma_zone_set_max(V_ipq_zone, V_maxnipq);
759 /*
760 * Zero specifies no further fragment queue allocation -- set the
761 * bound very low, but rely on implementation elsewhere to actually
762 * prevent allocation and reclaim current queues.
763 */
764 if (V_maxnipq == 0)
765 uma_zone_set_max(V_ipq_zone, 1);
766 }
767
768 static void
769 ipq_zone_change(void *tag)
770 {
771
772 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) {
773 V_maxnipq = nmbclusters / 32;
774 maxnipq_update();
775 }
776 }
777
778 static int
779 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
780 {
781 int error, i;
782
783 i = V_maxnipq;
784 error = sysctl_handle_int(oidp, &i, 0, req);
785 if (error || !req->newptr)
786 return (error);
787
788 /*
789 * XXXRW: Might be a good idea to sanity check the argument and place
790 * an extreme upper bound.
791 */
792 if (i < -1)
793 return (EINVAL);
794 V_maxnipq = i;
795 maxnipq_update();
796 return (0);
797 }
798
799 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
800 NULL, 0, sysctl_maxnipq, "I",
801 "Maximum number of IPv4 fragment reassembly queue entries");
802
803 #define M_IP_FRAG M_PROTO9
804
805 /*
806 * Take incoming datagram fragment and try to reassemble it into
807 * whole datagram. If the argument is the first fragment or one
808 * in between the function will return NULL and store the mbuf
809 * in the fragment chain. If the argument is the last fragment
810 * the packet will be reassembled and the pointer to the new
811 * mbuf returned for further processing. Only m_tags attached
812 * to the first packet/fragment are preserved.
813 * The IP header is *NOT* adjusted out of iplen.
814 */
815 struct mbuf *
816 ip_reass(struct mbuf *m)
817 {
818 struct ip *ip;
819 struct mbuf *p, *q, *nq, *t;
820 struct ipq *fp = NULL;
821 struct ipqhead *head;
822 int i, hlen, next;
823 u_int8_t ecn, ecn0;
824 u_short hash;
825
826 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
827 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) {
828 IPSTAT_INC(ips_fragments);
829 IPSTAT_INC(ips_fragdropped);
830 m_freem(m);
831 return (NULL);
832 }
833
834 ip = mtod(m, struct ip *);
835 hlen = ip->ip_hl << 2;
836
837 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
838 head = &V_ipq[hash];
839 IPQ_LOCK();
840
841 /*
842 * Look for queue of fragments
843 * of this datagram.
844 */
845 TAILQ_FOREACH(fp, head, ipq_list)
846 if (ip->ip_id == fp->ipq_id &&
847 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
848 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
849 #ifdef MAC
850 mac_ipq_match(m, fp) &&
851 #endif
852 ip->ip_p == fp->ipq_p)
853 goto found;
854
855 fp = NULL;
856
857 /*
858 * Attempt to trim the number of allocated fragment queues if it
859 * exceeds the administrative limit.
860 */
861 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) {
862 /*
863 * drop something from the tail of the current queue
864 * before proceeding further
865 */
866 struct ipq *q = TAILQ_LAST(head, ipqhead);
867 if (q == NULL) { /* gak */
868 for (i = 0; i < IPREASS_NHASH; i++) {
869 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead);
870 if (r) {
871 IPSTAT_ADD(ips_fragtimeout,
872 r->ipq_nfrags);
873 ip_freef(&V_ipq[i], r);
874 break;
875 }
876 }
877 } else {
878 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags);
879 ip_freef(head, q);
880 }
881 }
882
883 found:
884 /*
885 * Adjust ip_len to not reflect header,
886 * convert offset of this to bytes.
887 */
888 ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
889 if (ip->ip_off & htons(IP_MF)) {
890 /*
891 * Make sure that fragments have a data length
892 * that's a non-zero multiple of 8 bytes.
893 */
894 if (ip->ip_len == htons(0) || (ntohs(ip->ip_len) & 0x7) != 0) {
895 IPSTAT_INC(ips_toosmall); /* XXX */
896 goto dropfrag;
897 }
898 m->m_flags |= M_IP_FRAG;
899 } else
900 m->m_flags &= ~M_IP_FRAG;
901 ip->ip_off = htons(ntohs(ip->ip_off) << 3);
902
903 /*
904 * Attempt reassembly; if it succeeds, proceed.
905 * ip_reass() will return a different mbuf.
906 */
907 IPSTAT_INC(ips_fragments);
908 m->m_pkthdr.PH_loc.ptr = ip;
909
910 /* Previous ip_reass() started here. */
911 /*
912 * Presence of header sizes in mbufs
913 * would confuse code below.
914 */
915 m->m_data += hlen;
916 m->m_len -= hlen;
917
918 /*
919 * If first fragment to arrive, create a reassembly queue.
920 */
921 if (fp == NULL) {
922 fp = uma_zalloc(V_ipq_zone, M_NOWAIT);
923 if (fp == NULL)
924 goto dropfrag;
925 #ifdef MAC
926 if (mac_ipq_init(fp, M_NOWAIT) != 0) {
927 uma_zfree(V_ipq_zone, fp);
928 fp = NULL;
929 goto dropfrag;
930 }
931 mac_ipq_create(m, fp);
932 #endif
933 TAILQ_INSERT_HEAD(head, fp, ipq_list);
934 V_nipq++;
935 fp->ipq_nfrags = 1;
936 fp->ipq_ttl = IPFRAGTTL;
937 fp->ipq_p = ip->ip_p;
938 fp->ipq_id = ip->ip_id;
939 fp->ipq_src = ip->ip_src;
940 fp->ipq_dst = ip->ip_dst;
941 fp->ipq_frags = m;
942 m->m_nextpkt = NULL;
943 goto done;
944 } else {
945 fp->ipq_nfrags++;
946 #ifdef MAC
947 mac_ipq_update(m, fp);
948 #endif
949 }
950
951 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.PH_loc.ptr))
952
953 /*
954 * Handle ECN by comparing this segment with the first one;
955 * if CE is set, do not lose CE.
956 * drop if CE and not-ECT are mixed for the same packet.
957 */
958 ecn = ip->ip_tos & IPTOS_ECN_MASK;
959 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
960 if (ecn == IPTOS_ECN_CE) {
961 if (ecn0 == IPTOS_ECN_NOTECT)
962 goto dropfrag;
963 if (ecn0 != IPTOS_ECN_CE)
964 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
965 }
966 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
967 goto dropfrag;
968
969 /*
970 * Find a segment which begins after this one does.
971 */
972 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
973 if (ntohs(GETIP(q)->ip_off) > ntohs(ip->ip_off))
974 break;
975
976 /*
977 * If there is a preceding segment, it may provide some of
978 * our data already. If so, drop the data from the incoming
979 * segment. If it provides all of our data, drop us, otherwise
980 * stick new segment in the proper place.
981 *
982 * If some of the data is dropped from the preceding
983 * segment, then it's checksum is invalidated.
984 */
985 if (p) {
986 i = ntohs(GETIP(p)->ip_off) + ntohs(GETIP(p)->ip_len) -
987 ntohs(ip->ip_off);
988 if (i > 0) {
989 if (i >= ntohs(ip->ip_len))
990 goto dropfrag;
991 m_adj(m, i);
992 m->m_pkthdr.csum_flags = 0;
993 ip->ip_off = htons(ntohs(ip->ip_off) + i);
994 ip->ip_len = htons(ntohs(ip->ip_len) - i);
995 }
996 m->m_nextpkt = p->m_nextpkt;
997 p->m_nextpkt = m;
998 } else {
999 m->m_nextpkt = fp->ipq_frags;
1000 fp->ipq_frags = m;
1001 }
1002
1003 /*
1004 * While we overlap succeeding segments trim them or,
1005 * if they are completely covered, dequeue them.
1006 */
1007 for (; q != NULL && ntohs(ip->ip_off) + ntohs(ip->ip_len) >
1008 ntohs(GETIP(q)->ip_off); q = nq) {
1009 i = (ntohs(ip->ip_off) + ntohs(ip->ip_len)) -
1010 ntohs(GETIP(q)->ip_off);
1011 if (i < ntohs(GETIP(q)->ip_len)) {
1012 GETIP(q)->ip_len = htons(ntohs(GETIP(q)->ip_len) - i);
1013 GETIP(q)->ip_off = htons(ntohs(GETIP(q)->ip_off) + i);
1014 m_adj(q, i);
1015 q->m_pkthdr.csum_flags = 0;
1016 break;
1017 }
1018 nq = q->m_nextpkt;
1019 m->m_nextpkt = nq;
1020 IPSTAT_INC(ips_fragdropped);
1021 fp->ipq_nfrags--;
1022 m_freem(q);
1023 }
1024
1025 /*
1026 * Check for complete reassembly and perform frag per packet
1027 * limiting.
1028 *
1029 * Frag limiting is performed here so that the nth frag has
1030 * a chance to complete the packet before we drop the packet.
1031 * As a result, n+1 frags are actually allowed per packet, but
1032 * only n will ever be stored. (n = maxfragsperpacket.)
1033 *
1034 */
1035 next = 0;
1036 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1037 if (ntohs(GETIP(q)->ip_off) != next) {
1038 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1039 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1040 ip_freef(head, fp);
1041 }
1042 goto done;
1043 }
1044 next += ntohs(GETIP(q)->ip_len);
1045 }
1046 /* Make sure the last packet didn't have the IP_MF flag */
1047 if (p->m_flags & M_IP_FRAG) {
1048 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1049 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1050 ip_freef(head, fp);
1051 }
1052 goto done;
1053 }
1054
1055 /*
1056 * Reassembly is complete. Make sure the packet is a sane size.
1057 */
1058 q = fp->ipq_frags;
1059 ip = GETIP(q);
1060 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1061 IPSTAT_INC(ips_toolong);
1062 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1063 ip_freef(head, fp);
1064 goto done;
1065 }
1066
1067 /*
1068 * Concatenate fragments.
1069 */
1070 m = q;
1071 t = m->m_next;
1072 m->m_next = NULL;
1073 m_cat(m, t);
1074 nq = q->m_nextpkt;
1075 q->m_nextpkt = NULL;
1076 for (q = nq; q != NULL; q = nq) {
1077 nq = q->m_nextpkt;
1078 q->m_nextpkt = NULL;
1079 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1080 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1081 m_cat(m, q);
1082 }
1083 /*
1084 * In order to do checksumming faster we do 'end-around carry' here
1085 * (and not in for{} loop), though it implies we are not going to
1086 * reassemble more than 64k fragments.
1087 */
1088 while (m->m_pkthdr.csum_data & 0xffff0000)
1089 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1090 (m->m_pkthdr.csum_data >> 16);
1091 #ifdef MAC
1092 mac_ipq_reassemble(fp, m);
1093 mac_ipq_destroy(fp);
1094 #endif
1095
1096 /*
1097 * Create header for new ip packet by modifying header of first
1098 * packet; dequeue and discard fragment reassembly header.
1099 * Make header visible.
1100 */
1101 ip->ip_len = htons((ip->ip_hl << 2) + next);
1102 ip->ip_src = fp->ipq_src;
1103 ip->ip_dst = fp->ipq_dst;
1104 TAILQ_REMOVE(head, fp, ipq_list);
1105 V_nipq--;
1106 uma_zfree(V_ipq_zone, fp);
1107 m->m_len += (ip->ip_hl << 2);
1108 m->m_data -= (ip->ip_hl << 2);
1109 /* some debugging cruft by sklower, below, will go away soon */
1110 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1111 m_fixhdr(m);
1112 IPSTAT_INC(ips_reassembled);
1113 IPQ_UNLOCK();
1114 return (m);
1115
1116 dropfrag:
1117 IPSTAT_INC(ips_fragdropped);
1118 if (fp != NULL)
1119 fp->ipq_nfrags--;
1120 m_freem(m);
1121 done:
1122 IPQ_UNLOCK();
1123 return (NULL);
1124
1125 #undef GETIP
1126 }
1127
1128 /*
1129 * Free a fragment reassembly header and all
1130 * associated datagrams.
1131 */
1132 static void
1133 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1134 {
1135 struct mbuf *q;
1136
1137 IPQ_LOCK_ASSERT();
1138
1139 while (fp->ipq_frags) {
1140 q = fp->ipq_frags;
1141 fp->ipq_frags = q->m_nextpkt;
1142 m_freem(q);
1143 }
1144 TAILQ_REMOVE(fhp, fp, ipq_list);
1145 uma_zfree(V_ipq_zone, fp);
1146 V_nipq--;
1147 }
1148
1149 /*
1150 * IP timer processing;
1151 * if a timer expires on a reassembly
1152 * queue, discard it.
1153 */
1154 void
1155 ip_slowtimo(void)
1156 {
1157 VNET_ITERATOR_DECL(vnet_iter);
1158 struct ipq *fp;
1159 int i;
1160
1161 VNET_LIST_RLOCK_NOSLEEP();
1162 IPQ_LOCK();
1163 VNET_FOREACH(vnet_iter) {
1164 CURVNET_SET(vnet_iter);
1165 for (i = 0; i < IPREASS_NHASH; i++) {
1166 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) {
1167 struct ipq *fpp;
1168
1169 fpp = fp;
1170 fp = TAILQ_NEXT(fp, ipq_list);
1171 if(--fpp->ipq_ttl == 0) {
1172 IPSTAT_ADD(ips_fragtimeout,
1173 fpp->ipq_nfrags);
1174 ip_freef(&V_ipq[i], fpp);
1175 }
1176 }
1177 }
1178 /*
1179 * If we are over the maximum number of fragments
1180 * (due to the limit being lowered), drain off
1181 * enough to get down to the new limit.
1182 */
1183 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) {
1184 for (i = 0; i < IPREASS_NHASH; i++) {
1185 while (V_nipq > V_maxnipq &&
1186 !TAILQ_EMPTY(&V_ipq[i])) {
1187 IPSTAT_ADD(ips_fragdropped,
1188 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1189 ip_freef(&V_ipq[i],
1190 TAILQ_FIRST(&V_ipq[i]));
1191 }
1192 }
1193 }
1194 CURVNET_RESTORE();
1195 }
1196 IPQ_UNLOCK();
1197 VNET_LIST_RUNLOCK_NOSLEEP();
1198 }
1199
1200 /*
1201 * Drain off all datagram fragments.
1202 */
1203 static void
1204 ip_drain_locked(void)
1205 {
1206 int i;
1207
1208 IPQ_LOCK_ASSERT();
1209
1210 for (i = 0; i < IPREASS_NHASH; i++) {
1211 while(!TAILQ_EMPTY(&V_ipq[i])) {
1212 IPSTAT_ADD(ips_fragdropped,
1213 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1214 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i]));
1215 }
1216 }
1217 }
1218
1219 void
1220 ip_drain(void)
1221 {
1222 VNET_ITERATOR_DECL(vnet_iter);
1223
1224 VNET_LIST_RLOCK_NOSLEEP();
1225 IPQ_LOCK();
1226 VNET_FOREACH(vnet_iter) {
1227 CURVNET_SET(vnet_iter);
1228 ip_drain_locked();
1229 CURVNET_RESTORE();
1230 }
1231 IPQ_UNLOCK();
1232 VNET_LIST_RUNLOCK_NOSLEEP();
1233 in_rtqdrain();
1234 }
1235
1236 /*
1237 * The protocol to be inserted into ip_protox[] must be already registered
1238 * in inetsw[], either statically or through pf_proto_register().
1239 */
1240 int
1241 ipproto_register(short ipproto)
1242 {
1243 struct protosw *pr;
1244
1245 /* Sanity checks. */
1246 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
1247 return (EPROTONOSUPPORT);
1248
1249 /*
1250 * The protocol slot must not be occupied by another protocol
1251 * already. An index pointing to IPPROTO_RAW is unused.
1252 */
1253 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1254 if (pr == NULL)
1255 return (EPFNOSUPPORT);
1256 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1257 return (EEXIST);
1258
1259 /* Find the protocol position in inetsw[] and set the index. */
1260 for (pr = inetdomain.dom_protosw;
1261 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1262 if (pr->pr_domain->dom_family == PF_INET &&
1263 pr->pr_protocol && pr->pr_protocol == ipproto) {
1264 ip_protox[pr->pr_protocol] = pr - inetsw;
1265 return (0);
1266 }
1267 }
1268 return (EPROTONOSUPPORT);
1269 }
1270
1271 int
1272 ipproto_unregister(short ipproto)
1273 {
1274 struct protosw *pr;
1275
1276 /* Sanity checks. */
1277 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
1278 return (EPROTONOSUPPORT);
1279
1280 /* Check if the protocol was indeed registered. */
1281 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1282 if (pr == NULL)
1283 return (EPFNOSUPPORT);
1284 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1285 return (ENOENT);
1286
1287 /* Reset the protocol slot to IPPROTO_RAW. */
1288 ip_protox[ipproto] = pr - inetsw;
1289 return (0);
1290 }
1291
1292 /*
1293 * Given address of next destination (final or next hop), return (referenced)
1294 * internet address info of interface to be used to get there.
1295 */
1296 struct in_ifaddr *
1297 ip_rtaddr(struct in_addr dst, u_int fibnum)
1298 {
1299 struct route sro;
1300 struct sockaddr_in *sin;
1301 struct in_ifaddr *ia;
1302
1303 bzero(&sro, sizeof(sro));
1304 sin = (struct sockaddr_in *)&sro.ro_dst;
1305 sin->sin_family = AF_INET;
1306 sin->sin_len = sizeof(*sin);
1307 sin->sin_addr = dst;
1308 in_rtalloc_ign(&sro, 0, fibnum);
1309
1310 if (sro.ro_rt == NULL)
1311 return (NULL);
1312
1313 ia = ifatoia(sro.ro_rt->rt_ifa);
1314 ifa_ref(&ia->ia_ifa);
1315 RTFREE(sro.ro_rt);
1316 return (ia);
1317 }
1318
1319 u_char inetctlerrmap[PRC_NCMDS] = {
1320 0, 0, 0, 0,
1321 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1322 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1323 EMSGSIZE, EHOSTUNREACH, 0, 0,
1324 0, 0, EHOSTUNREACH, 0,
1325 ENOPROTOOPT, ECONNREFUSED
1326 };
1327
1328 /*
1329 * Forward a packet. If some error occurs return the sender
1330 * an icmp packet. Note we can't always generate a meaningful
1331 * icmp message because icmp doesn't have a large enough repertoire
1332 * of codes and types.
1333 *
1334 * If not forwarding, just drop the packet. This could be confusing
1335 * if ipforwarding was zero but some routing protocol was advancing
1336 * us as a gateway to somewhere. However, we must let the routing
1337 * protocol deal with that.
1338 *
1339 * The srcrt parameter indicates whether the packet is being forwarded
1340 * via a source route.
1341 */
1342 void
1343 ip_forward(struct mbuf *m, int srcrt)
1344 {
1345 struct ip *ip = mtod(m, struct ip *);
1346 struct in_ifaddr *ia;
1347 struct mbuf *mcopy;
1348 struct in_addr dest;
1349 struct route ro;
1350 int error, type = 0, code = 0, mtu = 0;
1351
1352 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1353 IPSTAT_INC(ips_cantforward);
1354 m_freem(m);
1355 return;
1356 }
1357 #ifdef IPSTEALTH
1358 if (!V_ipstealth) {
1359 #endif
1360 if (ip->ip_ttl <= IPTTLDEC) {
1361 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1362 0, 0);
1363 return;
1364 }
1365 #ifdef IPSTEALTH
1366 }
1367 #endif
1368
1369 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
1370 #ifndef IPSEC
1371 /*
1372 * 'ia' may be NULL if there is no route for this destination.
1373 * In case of IPsec, Don't discard it just yet, but pass it to
1374 * ip_output in case of outgoing IPsec policy.
1375 */
1376 if (!srcrt && ia == NULL) {
1377 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1378 return;
1379 }
1380 #endif
1381
1382 /*
1383 * Save the IP header and at most 8 bytes of the payload,
1384 * in case we need to generate an ICMP message to the src.
1385 *
1386 * XXX this can be optimized a lot by saving the data in a local
1387 * buffer on the stack (72 bytes at most), and only allocating the
1388 * mbuf if really necessary. The vast majority of the packets
1389 * are forwarded without having to send an ICMP back (either
1390 * because unnecessary, or because rate limited), so we are
1391 * really we are wasting a lot of work here.
1392 *
1393 * We don't use m_copy() because it might return a reference
1394 * to a shared cluster. Both this function and ip_output()
1395 * assume exclusive access to the IP header in `m', so any
1396 * data in a cluster may change before we reach icmp_error().
1397 */
1398 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1399 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1400 /*
1401 * It's probably ok if the pkthdr dup fails (because
1402 * the deep copy of the tag chain failed), but for now
1403 * be conservative and just discard the copy since
1404 * code below may some day want the tags.
1405 */
1406 m_free(mcopy);
1407 mcopy = NULL;
1408 }
1409 if (mcopy != NULL) {
1410 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1411 mcopy->m_pkthdr.len = mcopy->m_len;
1412 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1413 }
1414
1415 #ifdef IPSTEALTH
1416 if (!V_ipstealth) {
1417 #endif
1418 ip->ip_ttl -= IPTTLDEC;
1419 #ifdef IPSTEALTH
1420 }
1421 #endif
1422
1423 /*
1424 * If forwarding packet using same interface that it came in on,
1425 * perhaps should send a redirect to sender to shortcut a hop.
1426 * Only send redirect if source is sending directly to us,
1427 * and if packet was not source routed (or has any options).
1428 * Also, don't send redirect if forwarding using a default route
1429 * or a route modified by a redirect.
1430 */
1431 dest.s_addr = 0;
1432 if (!srcrt && V_ipsendredirects &&
1433 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1434 struct sockaddr_in *sin;
1435 struct rtentry *rt;
1436
1437 bzero(&ro, sizeof(ro));
1438 sin = (struct sockaddr_in *)&ro.ro_dst;
1439 sin->sin_family = AF_INET;
1440 sin->sin_len = sizeof(*sin);
1441 sin->sin_addr = ip->ip_dst;
1442 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
1443
1444 rt = ro.ro_rt;
1445
1446 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1447 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1448 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1449 u_long src = ntohl(ip->ip_src.s_addr);
1450
1451 if (RTA(rt) &&
1452 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1453 if (rt->rt_flags & RTF_GATEWAY)
1454 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1455 else
1456 dest.s_addr = ip->ip_dst.s_addr;
1457 /* Router requirements says to only send host redirects */
1458 type = ICMP_REDIRECT;
1459 code = ICMP_REDIRECT_HOST;
1460 }
1461 }
1462 if (rt)
1463 RTFREE(rt);
1464 }
1465
1466 /*
1467 * Try to cache the route MTU from ip_output so we can consider it for
1468 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191.
1469 */
1470 bzero(&ro, sizeof(ro));
1471
1472 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1473
1474 if (error == EMSGSIZE && ro.ro_rt)
1475 mtu = ro.ro_rt->rt_mtu;
1476 RO_RTFREE(&ro);
1477
1478 if (error)
1479 IPSTAT_INC(ips_cantforward);
1480 else {
1481 IPSTAT_INC(ips_forward);
1482 if (type)
1483 IPSTAT_INC(ips_redirectsent);
1484 else {
1485 if (mcopy)
1486 m_freem(mcopy);
1487 if (ia != NULL)
1488 ifa_free(&ia->ia_ifa);
1489 return;
1490 }
1491 }
1492 if (mcopy == NULL) {
1493 if (ia != NULL)
1494 ifa_free(&ia->ia_ifa);
1495 return;
1496 }
1497
1498 switch (error) {
1499
1500 case 0: /* forwarded, but need redirect */
1501 /* type, code set above */
1502 break;
1503
1504 case ENETUNREACH:
1505 case EHOSTUNREACH:
1506 case ENETDOWN:
1507 case EHOSTDOWN:
1508 default:
1509 type = ICMP_UNREACH;
1510 code = ICMP_UNREACH_HOST;
1511 break;
1512
1513 case EMSGSIZE:
1514 type = ICMP_UNREACH;
1515 code = ICMP_UNREACH_NEEDFRAG;
1516
1517 #ifdef IPSEC
1518 /*
1519 * If IPsec is configured for this path,
1520 * override any possibly mtu value set by ip_output.
1521 */
1522 mtu = ip_ipsec_mtu(mcopy, mtu);
1523 #endif /* IPSEC */
1524 /*
1525 * If the MTU was set before make sure we are below the
1526 * interface MTU.
1527 * If the MTU wasn't set before use the interface mtu or
1528 * fall back to the next smaller mtu step compared to the
1529 * current packet size.
1530 */
1531 if (mtu != 0) {
1532 if (ia != NULL)
1533 mtu = min(mtu, ia->ia_ifp->if_mtu);
1534 } else {
1535 if (ia != NULL)
1536 mtu = ia->ia_ifp->if_mtu;
1537 else
1538 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1539 }
1540 IPSTAT_INC(ips_cantfrag);
1541 break;
1542
1543 case ENOBUFS:
1544 /*
1545 * A router should not generate ICMP_SOURCEQUENCH as
1546 * required in RFC1812 Requirements for IP Version 4 Routers.
1547 * Source quench could be a big problem under DoS attacks,
1548 * or if the underlying interface is rate-limited.
1549 * Those who need source quench packets may re-enable them
1550 * via the net.inet.ip.sendsourcequench sysctl.
1551 */
1552 if (V_ip_sendsourcequench == 0) {
1553 m_freem(mcopy);
1554 if (ia != NULL)
1555 ifa_free(&ia->ia_ifa);
1556 return;
1557 } else {
1558 type = ICMP_SOURCEQUENCH;
1559 code = 0;
1560 }
1561 break;
1562
1563 case EACCES: /* ipfw denied packet */
1564 m_freem(mcopy);
1565 if (ia != NULL)
1566 ifa_free(&ia->ia_ifa);
1567 return;
1568 }
1569 if (ia != NULL)
1570 ifa_free(&ia->ia_ifa);
1571 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1572 }
1573
1574 void
1575 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1576 struct mbuf *m)
1577 {
1578
1579 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1580 struct bintime bt;
1581
1582 bintime(&bt);
1583 if (inp->inp_socket->so_options & SO_BINTIME) {
1584 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1585 SCM_BINTIME, SOL_SOCKET);
1586 if (*mp)
1587 mp = &(*mp)->m_next;
1588 }
1589 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1590 struct timeval tv;
1591
1592 bintime2timeval(&bt, &tv);
1593 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1594 SCM_TIMESTAMP, SOL_SOCKET);
1595 if (*mp)
1596 mp = &(*mp)->m_next;
1597 }
1598 }
1599 if (inp->inp_flags & INP_RECVDSTADDR) {
1600 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1601 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1602 if (*mp)
1603 mp = &(*mp)->m_next;
1604 }
1605 if (inp->inp_flags & INP_RECVTTL) {
1606 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1607 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1608 if (*mp)
1609 mp = &(*mp)->m_next;
1610 }
1611 #ifdef notyet
1612 /* XXX
1613 * Moving these out of udp_input() made them even more broken
1614 * than they already were.
1615 */
1616 /* options were tossed already */
1617 if (inp->inp_flags & INP_RECVOPTS) {
1618 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1619 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1620 if (*mp)
1621 mp = &(*mp)->m_next;
1622 }
1623 /* ip_srcroute doesn't do what we want here, need to fix */
1624 if (inp->inp_flags & INP_RECVRETOPTS) {
1625 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1626 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1627 if (*mp)
1628 mp = &(*mp)->m_next;
1629 }
1630 #endif
1631 if (inp->inp_flags & INP_RECVIF) {
1632 struct ifnet *ifp;
1633 struct sdlbuf {
1634 struct sockaddr_dl sdl;
1635 u_char pad[32];
1636 } sdlbuf;
1637 struct sockaddr_dl *sdp;
1638 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1639
1640 if ((ifp = m->m_pkthdr.rcvif) &&
1641 ifp->if_index && ifp->if_index <= V_if_index) {
1642 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1643 /*
1644 * Change our mind and don't try copy.
1645 */
1646 if (sdp->sdl_family != AF_LINK ||
1647 sdp->sdl_len > sizeof(sdlbuf)) {
1648 goto makedummy;
1649 }
1650 bcopy(sdp, sdl2, sdp->sdl_len);
1651 } else {
1652 makedummy:
1653 sdl2->sdl_len =
1654 offsetof(struct sockaddr_dl, sdl_data[0]);
1655 sdl2->sdl_family = AF_LINK;
1656 sdl2->sdl_index = 0;
1657 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1658 }
1659 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1660 IP_RECVIF, IPPROTO_IP);
1661 if (*mp)
1662 mp = &(*mp)->m_next;
1663 }
1664 if (inp->inp_flags & INP_RECVTOS) {
1665 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1666 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1667 if (*mp)
1668 mp = &(*mp)->m_next;
1669 }
1670 }
1671
1672 /*
1673 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1674 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1675 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1676 * compiled.
1677 */
1678 static VNET_DEFINE(int, ip_rsvp_on);
1679 VNET_DEFINE(struct socket *, ip_rsvpd);
1680
1681 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1682
1683 int
1684 ip_rsvp_init(struct socket *so)
1685 {
1686
1687 if (so->so_type != SOCK_RAW ||
1688 so->so_proto->pr_protocol != IPPROTO_RSVP)
1689 return EOPNOTSUPP;
1690
1691 if (V_ip_rsvpd != NULL)
1692 return EADDRINUSE;
1693
1694 V_ip_rsvpd = so;
1695 /*
1696 * This may seem silly, but we need to be sure we don't over-increment
1697 * the RSVP counter, in case something slips up.
1698 */
1699 if (!V_ip_rsvp_on) {
1700 V_ip_rsvp_on = 1;
1701 V_rsvp_on++;
1702 }
1703
1704 return 0;
1705 }
1706
1707 int
1708 ip_rsvp_done(void)
1709 {
1710
1711 V_ip_rsvpd = NULL;
1712 /*
1713 * This may seem silly, but we need to be sure we don't over-decrement
1714 * the RSVP counter, in case something slips up.
1715 */
1716 if (V_ip_rsvp_on) {
1717 V_ip_rsvp_on = 0;
1718 V_rsvp_on--;
1719 }
1720 return 0;
1721 }
1722
1723 void
1724 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
1725 {
1726
1727 if (rsvp_input_p) { /* call the real one if loaded */
1728 rsvp_input_p(m, off);
1729 return;
1730 }
1731
1732 /* Can still get packets with rsvp_on = 0 if there is a local member
1733 * of the group to which the RSVP packet is addressed. But in this
1734 * case we want to throw the packet away.
1735 */
1736
1737 if (!V_rsvp_on) {
1738 m_freem(m);
1739 return;
1740 }
1741
1742 if (V_ip_rsvpd != NULL) {
1743 rip_input(m, off);
1744 return;
1745 }
1746 /* Drop the packet */
1747 m_freem(m);
1748 }
Cache object: a588ecee039f093bc7c07f9c48ac95a9
|