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: stable/10/sys/netinet/ip_input.c 340961 2018-11-26 12:47:12Z eugen $");
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
kmod_ipstat_inc(int statnum)219 kmod_ipstat_inc(int statnum)
220 {
221
222 counter_u64_add(VNET(ipstat)[statnum], 1);
223 }
224
225 void
kmod_ipstat_dec(int statnum)226 kmod_ipstat_dec(int statnum)
227 {
228
229 counter_u64_add(VNET(ipstat)[statnum], -1);
230 }
231
232 static int
sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)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
sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)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
ip_init(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
ip_destroy(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
ip_input(struct mbuf * m)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 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
503 /*
504 * Directly ship the packet on. This allows
505 * forwarding packets originally destined to us
506 * to some other directly connected host.
507 */
508 ip_forward(m, 1);
509 return;
510 }
511 }
512 passin:
513
514 /*
515 * Process options and, if not destined for us,
516 * ship it on. ip_dooptions returns 1 when an
517 * error was detected (causing an icmp message
518 * to be sent and the original packet to be freed).
519 */
520 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
521 return;
522
523 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
524 * matter if it is destined to another node, or whether it is
525 * a multicast one, RSVP wants it! and prevents it from being forwarded
526 * anywhere else. Also checks if the rsvp daemon is running before
527 * grabbing the packet.
528 */
529 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
530 goto ours;
531
532 /*
533 * Check our list of addresses, to see if the packet is for us.
534 * If we don't have any addresses, assume any unicast packet
535 * we receive might be for us (and let the upper layers deal
536 * with it).
537 */
538 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
539 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
540 goto ours;
541
542 /*
543 * Enable a consistency check between the destination address
544 * and the arrival interface for a unicast packet (the RFC 1122
545 * strong ES model) if IP forwarding is disabled and the packet
546 * is not locally generated and the packet is not subject to
547 * 'ipfw fwd'.
548 *
549 * XXX - Checking also should be disabled if the destination
550 * address is ipnat'ed to a different interface.
551 *
552 * XXX - Checking is incompatible with IP aliases added
553 * to the loopback interface instead of the interface where
554 * the packets are received.
555 *
556 * XXX - This is the case for carp vhost IPs as well so we
557 * insert a workaround. If the packet got here, we already
558 * checked with carp_iamatch() and carp_forus().
559 */
560 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
561 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
562 ifp->if_carp == NULL && (dchg == 0);
563
564 /*
565 * Check for exact addresses in the hash bucket.
566 */
567 IN_IFADDR_RLOCK();
568 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
569 /*
570 * If the address matches, verify that the packet
571 * arrived via the correct interface if checking is
572 * enabled.
573 */
574 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
575 (!checkif || ia->ia_ifp == ifp)) {
576 ifa_ref(&ia->ia_ifa);
577 IN_IFADDR_RUNLOCK();
578 goto ours;
579 }
580 }
581 IN_IFADDR_RUNLOCK();
582
583 /*
584 * Check for broadcast addresses.
585 *
586 * Only accept broadcast packets that arrive via the matching
587 * interface. Reception of forwarded directed broadcasts would
588 * be handled via ip_forward() and ether_output() with the loopback
589 * into the stack for SIMPLEX interfaces handled by ether_output().
590 */
591 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
592 IF_ADDR_RLOCK(ifp);
593 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
594 if (ifa->ifa_addr->sa_family != AF_INET)
595 continue;
596 ia = ifatoia(ifa);
597 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
598 ip->ip_dst.s_addr) {
599 ifa_ref(ifa);
600 IF_ADDR_RUNLOCK(ifp);
601 goto ours;
602 }
603 #ifdef BOOTP_COMPAT
604 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
605 ifa_ref(ifa);
606 IF_ADDR_RUNLOCK(ifp);
607 goto ours;
608 }
609 #endif
610 }
611 IF_ADDR_RUNLOCK(ifp);
612 ia = NULL;
613 }
614 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
615 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
616 IPSTAT_INC(ips_cantforward);
617 m_freem(m);
618 return;
619 }
620 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
621 if (V_ip_mrouter) {
622 /*
623 * If we are acting as a multicast router, all
624 * incoming multicast packets are passed to the
625 * kernel-level multicast forwarding function.
626 * The packet is returned (relatively) intact; if
627 * ip_mforward() returns a non-zero value, the packet
628 * must be discarded, else it may be accepted below.
629 */
630 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
631 IPSTAT_INC(ips_cantforward);
632 m_freem(m);
633 return;
634 }
635
636 /*
637 * The process-level routing daemon needs to receive
638 * all multicast IGMP packets, whether or not this
639 * host belongs to their destination groups.
640 */
641 if (ip->ip_p == IPPROTO_IGMP)
642 goto ours;
643 IPSTAT_INC(ips_forward);
644 }
645 /*
646 * Assume the packet is for us, to avoid prematurely taking
647 * a lock on the in_multi hash. Protocols must perform
648 * their own filtering and update statistics accordingly.
649 */
650 goto ours;
651 }
652 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
653 goto ours;
654 if (ip->ip_dst.s_addr == INADDR_ANY)
655 goto ours;
656
657 /*
658 * FAITH(Firewall Aided Internet Translator)
659 */
660 if (ifp && ifp->if_type == IFT_FAITH) {
661 if (V_ip_keepfaith) {
662 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
663 goto ours;
664 }
665 m_freem(m);
666 return;
667 }
668
669 /*
670 * Not for us; forward if possible and desirable.
671 */
672 if (V_ipforwarding == 0) {
673 IPSTAT_INC(ips_cantforward);
674 m_freem(m);
675 } else {
676 #ifdef IPSEC
677 if (ip_ipsec_fwd(m))
678 goto bad;
679 #endif /* IPSEC */
680 ip_forward(m, dchg);
681 }
682 return;
683
684 ours:
685 #ifdef IPSTEALTH
686 /*
687 * IPSTEALTH: Process non-routing options only
688 * if the packet is destined for us.
689 */
690 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) {
691 if (ia != NULL)
692 ifa_free(&ia->ia_ifa);
693 return;
694 }
695 #endif /* IPSTEALTH */
696
697 /* Count the packet in the ip address stats */
698 if (ia != NULL) {
699 ia->ia_ifa.if_ipackets++;
700 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
701 ifa_free(&ia->ia_ifa);
702 }
703
704 /*
705 * Attempt reassembly; if it succeeds, proceed.
706 * ip_reass() will return a different mbuf.
707 */
708 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
709 /* XXXGL: shouldn't we save & set m_flags? */
710 m = ip_reass(m);
711 if (m == NULL)
712 return;
713 ip = mtod(m, struct ip *);
714 /* Get the header length of the reassembled packet */
715 hlen = ip->ip_hl << 2;
716 }
717
718 #ifdef IPSEC
719 /*
720 * enforce IPsec policy checking if we are seeing last header.
721 * note that we do not visit this with protocols with pcb layer
722 * code - like udp/tcp/raw ip.
723 */
724 if (ip_ipsec_input(m))
725 goto bad;
726 #endif /* IPSEC */
727
728 /*
729 * Switch out to protocol's input routine.
730 */
731 IPSTAT_INC(ips_delivered);
732
733 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
734 return;
735 bad:
736 m_freem(m);
737 }
738
739 /*
740 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
741 * max has slightly different semantics than the sysctl, for historical
742 * reasons.
743 */
744 static void
maxnipq_update(void)745 maxnipq_update(void)
746 {
747
748 /*
749 * -1 for unlimited allocation.
750 */
751 if (V_maxnipq < 0)
752 uma_zone_set_max(V_ipq_zone, 0);
753 /*
754 * Positive number for specific bound.
755 */
756 if (V_maxnipq > 0)
757 uma_zone_set_max(V_ipq_zone, V_maxnipq);
758 /*
759 * Zero specifies no further fragment queue allocation -- set the
760 * bound very low, but rely on implementation elsewhere to actually
761 * prevent allocation and reclaim current queues.
762 */
763 if (V_maxnipq == 0)
764 uma_zone_set_max(V_ipq_zone, 1);
765 }
766
767 static void
ipq_zone_change(void * tag)768 ipq_zone_change(void *tag)
769 {
770
771 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) {
772 V_maxnipq = nmbclusters / 32;
773 maxnipq_update();
774 }
775 }
776
777 static int
sysctl_maxnipq(SYSCTL_HANDLER_ARGS)778 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
779 {
780 int error, i;
781
782 i = V_maxnipq;
783 error = sysctl_handle_int(oidp, &i, 0, req);
784 if (error || !req->newptr)
785 return (error);
786
787 /*
788 * XXXRW: Might be a good idea to sanity check the argument and place
789 * an extreme upper bound.
790 */
791 if (i < -1)
792 return (EINVAL);
793 V_maxnipq = i;
794 maxnipq_update();
795 return (0);
796 }
797
798 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
799 NULL, 0, sysctl_maxnipq, "I",
800 "Maximum number of IPv4 fragment reassembly queue entries");
801
802 #define M_IP_FRAG M_PROTO9
803
804 /*
805 * Take incoming datagram fragment and try to reassemble it into
806 * whole datagram. If the argument is the first fragment or one
807 * in between the function will return NULL and store the mbuf
808 * in the fragment chain. If the argument is the last fragment
809 * the packet will be reassembled and the pointer to the new
810 * mbuf returned for further processing. Only m_tags attached
811 * to the first packet/fragment are preserved.
812 * The IP header is *NOT* adjusted out of iplen.
813 */
814 struct mbuf *
ip_reass(struct mbuf * m)815 ip_reass(struct mbuf *m)
816 {
817 struct ip *ip;
818 struct mbuf *p, *q, *nq, *t;
819 struct ipq *fp = NULL;
820 struct ipqhead *head;
821 int i, hlen, next;
822 u_int8_t ecn, ecn0;
823 u_short hash;
824
825 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
826 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) {
827 IPSTAT_INC(ips_fragments);
828 IPSTAT_INC(ips_fragdropped);
829 m_freem(m);
830 return (NULL);
831 }
832
833 ip = mtod(m, struct ip *);
834 hlen = ip->ip_hl << 2;
835
836 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
837 head = &V_ipq[hash];
838 IPQ_LOCK();
839
840 /*
841 * Look for queue of fragments
842 * of this datagram.
843 */
844 TAILQ_FOREACH(fp, head, ipq_list)
845 if (ip->ip_id == fp->ipq_id &&
846 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
847 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
848 #ifdef MAC
849 mac_ipq_match(m, fp) &&
850 #endif
851 ip->ip_p == fp->ipq_p)
852 goto found;
853
854 fp = NULL;
855
856 /*
857 * Attempt to trim the number of allocated fragment queues if it
858 * exceeds the administrative limit.
859 */
860 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) {
861 /*
862 * drop something from the tail of the current queue
863 * before proceeding further
864 */
865 struct ipq *q = TAILQ_LAST(head, ipqhead);
866 if (q == NULL) { /* gak */
867 for (i = 0; i < IPREASS_NHASH; i++) {
868 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead);
869 if (r) {
870 IPSTAT_ADD(ips_fragtimeout,
871 r->ipq_nfrags);
872 ip_freef(&V_ipq[i], r);
873 break;
874 }
875 }
876 } else {
877 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags);
878 ip_freef(head, q);
879 }
880 }
881
882 found:
883 /*
884 * Adjust ip_len to not reflect header,
885 * convert offset of this to bytes.
886 */
887 ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
888 if (ip->ip_off & htons(IP_MF)) {
889 /*
890 * Make sure that fragments have a data length
891 * that's a non-zero multiple of 8 bytes.
892 */
893 if (ip->ip_len == htons(0) || (ntohs(ip->ip_len) & 0x7) != 0) {
894 IPSTAT_INC(ips_toosmall); /* XXX */
895 goto dropfrag;
896 }
897 m->m_flags |= M_IP_FRAG;
898 } else
899 m->m_flags &= ~M_IP_FRAG;
900 ip->ip_off = htons(ntohs(ip->ip_off) << 3);
901
902 /*
903 * Attempt reassembly; if it succeeds, proceed.
904 * ip_reass() will return a different mbuf.
905 */
906 IPSTAT_INC(ips_fragments);
907 m->m_pkthdr.PH_loc.ptr = ip;
908
909 /* Previous ip_reass() started here. */
910 /*
911 * Presence of header sizes in mbufs
912 * would confuse code below.
913 */
914 m->m_data += hlen;
915 m->m_len -= hlen;
916
917 /*
918 * If first fragment to arrive, create a reassembly queue.
919 */
920 if (fp == NULL) {
921 fp = uma_zalloc(V_ipq_zone, M_NOWAIT);
922 if (fp == NULL)
923 goto dropfrag;
924 #ifdef MAC
925 if (mac_ipq_init(fp, M_NOWAIT) != 0) {
926 uma_zfree(V_ipq_zone, fp);
927 fp = NULL;
928 goto dropfrag;
929 }
930 mac_ipq_create(m, fp);
931 #endif
932 TAILQ_INSERT_HEAD(head, fp, ipq_list);
933 V_nipq++;
934 fp->ipq_nfrags = 1;
935 fp->ipq_ttl = IPFRAGTTL;
936 fp->ipq_p = ip->ip_p;
937 fp->ipq_id = ip->ip_id;
938 fp->ipq_src = ip->ip_src;
939 fp->ipq_dst = ip->ip_dst;
940 fp->ipq_frags = m;
941 m->m_nextpkt = NULL;
942 goto done;
943 } else {
944 fp->ipq_nfrags++;
945 #ifdef MAC
946 mac_ipq_update(m, fp);
947 #endif
948 }
949
950 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.PH_loc.ptr))
951
952 /*
953 * Handle ECN by comparing this segment with the first one;
954 * if CE is set, do not lose CE.
955 * drop if CE and not-ECT are mixed for the same packet.
956 */
957 ecn = ip->ip_tos & IPTOS_ECN_MASK;
958 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
959 if (ecn == IPTOS_ECN_CE) {
960 if (ecn0 == IPTOS_ECN_NOTECT)
961 goto dropfrag;
962 if (ecn0 != IPTOS_ECN_CE)
963 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
964 }
965 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
966 goto dropfrag;
967
968 /*
969 * Find a segment which begins after this one does.
970 */
971 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
972 if (ntohs(GETIP(q)->ip_off) > ntohs(ip->ip_off))
973 break;
974
975 /*
976 * If there is a preceding segment, it may provide some of
977 * our data already. If so, drop the data from the incoming
978 * segment. If it provides all of our data, drop us, otherwise
979 * stick new segment in the proper place.
980 *
981 * If some of the data is dropped from the preceding
982 * segment, then it's checksum is invalidated.
983 */
984 if (p) {
985 i = ntohs(GETIP(p)->ip_off) + ntohs(GETIP(p)->ip_len) -
986 ntohs(ip->ip_off);
987 if (i > 0) {
988 if (i >= ntohs(ip->ip_len))
989 goto dropfrag;
990 m_adj(m, i);
991 m->m_pkthdr.csum_flags = 0;
992 ip->ip_off = htons(ntohs(ip->ip_off) + i);
993 ip->ip_len = htons(ntohs(ip->ip_len) - i);
994 }
995 m->m_nextpkt = p->m_nextpkt;
996 p->m_nextpkt = m;
997 } else {
998 m->m_nextpkt = fp->ipq_frags;
999 fp->ipq_frags = m;
1000 }
1001
1002 /*
1003 * While we overlap succeeding segments trim them or,
1004 * if they are completely covered, dequeue them.
1005 */
1006 for (; q != NULL && ntohs(ip->ip_off) + ntohs(ip->ip_len) >
1007 ntohs(GETIP(q)->ip_off); q = nq) {
1008 i = (ntohs(ip->ip_off) + ntohs(ip->ip_len)) -
1009 ntohs(GETIP(q)->ip_off);
1010 if (i < ntohs(GETIP(q)->ip_len)) {
1011 GETIP(q)->ip_len = htons(ntohs(GETIP(q)->ip_len) - i);
1012 GETIP(q)->ip_off = htons(ntohs(GETIP(q)->ip_off) + i);
1013 m_adj(q, i);
1014 q->m_pkthdr.csum_flags = 0;
1015 break;
1016 }
1017 nq = q->m_nextpkt;
1018 m->m_nextpkt = nq;
1019 IPSTAT_INC(ips_fragdropped);
1020 fp->ipq_nfrags--;
1021 m_freem(q);
1022 }
1023
1024 /*
1025 * Check for complete reassembly and perform frag per packet
1026 * limiting.
1027 *
1028 * Frag limiting is performed here so that the nth frag has
1029 * a chance to complete the packet before we drop the packet.
1030 * As a result, n+1 frags are actually allowed per packet, but
1031 * only n will ever be stored. (n = maxfragsperpacket.)
1032 *
1033 */
1034 next = 0;
1035 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1036 if (ntohs(GETIP(q)->ip_off) != next) {
1037 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1038 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1039 ip_freef(head, fp);
1040 }
1041 goto done;
1042 }
1043 next += ntohs(GETIP(q)->ip_len);
1044 }
1045 /* Make sure the last packet didn't have the IP_MF flag */
1046 if (p->m_flags & M_IP_FRAG) {
1047 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1048 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1049 ip_freef(head, fp);
1050 }
1051 goto done;
1052 }
1053
1054 /*
1055 * Reassembly is complete. Make sure the packet is a sane size.
1056 */
1057 q = fp->ipq_frags;
1058 ip = GETIP(q);
1059 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1060 IPSTAT_INC(ips_toolong);
1061 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1062 ip_freef(head, fp);
1063 goto done;
1064 }
1065
1066 /*
1067 * Concatenate fragments.
1068 */
1069 m = q;
1070 t = m->m_next;
1071 m->m_next = NULL;
1072 m_cat(m, t);
1073 nq = q->m_nextpkt;
1074 q->m_nextpkt = NULL;
1075 for (q = nq; q != NULL; q = nq) {
1076 nq = q->m_nextpkt;
1077 q->m_nextpkt = NULL;
1078 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1079 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1080 m_cat(m, q);
1081 }
1082 /*
1083 * In order to do checksumming faster we do 'end-around carry' here
1084 * (and not in for{} loop), though it implies we are not going to
1085 * reassemble more than 64k fragments.
1086 */
1087 while (m->m_pkthdr.csum_data & 0xffff0000)
1088 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1089 (m->m_pkthdr.csum_data >> 16);
1090 #ifdef MAC
1091 mac_ipq_reassemble(fp, m);
1092 mac_ipq_destroy(fp);
1093 #endif
1094
1095 /*
1096 * Create header for new ip packet by modifying header of first
1097 * packet; dequeue and discard fragment reassembly header.
1098 * Make header visible.
1099 */
1100 ip->ip_len = htons((ip->ip_hl << 2) + next);
1101 ip->ip_src = fp->ipq_src;
1102 ip->ip_dst = fp->ipq_dst;
1103 TAILQ_REMOVE(head, fp, ipq_list);
1104 V_nipq--;
1105 uma_zfree(V_ipq_zone, fp);
1106 m->m_len += (ip->ip_hl << 2);
1107 m->m_data -= (ip->ip_hl << 2);
1108 /* some debugging cruft by sklower, below, will go away soon */
1109 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1110 m_fixhdr(m);
1111 IPSTAT_INC(ips_reassembled);
1112 IPQ_UNLOCK();
1113 return (m);
1114
1115 dropfrag:
1116 IPSTAT_INC(ips_fragdropped);
1117 if (fp != NULL)
1118 fp->ipq_nfrags--;
1119 m_freem(m);
1120 done:
1121 IPQ_UNLOCK();
1122 return (NULL);
1123
1124 #undef GETIP
1125 }
1126
1127 /*
1128 * Free a fragment reassembly header and all
1129 * associated datagrams.
1130 */
1131 static void
ip_freef(struct ipqhead * fhp,struct ipq * fp)1132 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1133 {
1134 struct mbuf *q;
1135
1136 IPQ_LOCK_ASSERT();
1137
1138 while (fp->ipq_frags) {
1139 q = fp->ipq_frags;
1140 fp->ipq_frags = q->m_nextpkt;
1141 m_freem(q);
1142 }
1143 TAILQ_REMOVE(fhp, fp, ipq_list);
1144 uma_zfree(V_ipq_zone, fp);
1145 V_nipq--;
1146 }
1147
1148 /*
1149 * IP timer processing;
1150 * if a timer expires on a reassembly
1151 * queue, discard it.
1152 */
1153 void
ip_slowtimo(void)1154 ip_slowtimo(void)
1155 {
1156 VNET_ITERATOR_DECL(vnet_iter);
1157 struct ipq *fp;
1158 int i;
1159
1160 VNET_LIST_RLOCK_NOSLEEP();
1161 IPQ_LOCK();
1162 VNET_FOREACH(vnet_iter) {
1163 CURVNET_SET(vnet_iter);
1164 for (i = 0; i < IPREASS_NHASH; i++) {
1165 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) {
1166 struct ipq *fpp;
1167
1168 fpp = fp;
1169 fp = TAILQ_NEXT(fp, ipq_list);
1170 if(--fpp->ipq_ttl == 0) {
1171 IPSTAT_ADD(ips_fragtimeout,
1172 fpp->ipq_nfrags);
1173 ip_freef(&V_ipq[i], fpp);
1174 }
1175 }
1176 }
1177 /*
1178 * If we are over the maximum number of fragments
1179 * (due to the limit being lowered), drain off
1180 * enough to get down to the new limit.
1181 */
1182 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) {
1183 for (i = 0; i < IPREASS_NHASH; i++) {
1184 while (V_nipq > V_maxnipq &&
1185 !TAILQ_EMPTY(&V_ipq[i])) {
1186 IPSTAT_ADD(ips_fragdropped,
1187 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1188 ip_freef(&V_ipq[i],
1189 TAILQ_FIRST(&V_ipq[i]));
1190 }
1191 }
1192 }
1193 CURVNET_RESTORE();
1194 }
1195 IPQ_UNLOCK();
1196 VNET_LIST_RUNLOCK_NOSLEEP();
1197 }
1198
1199 /*
1200 * Drain off all datagram fragments.
1201 */
1202 static void
ip_drain_locked(void)1203 ip_drain_locked(void)
1204 {
1205 int i;
1206
1207 IPQ_LOCK_ASSERT();
1208
1209 for (i = 0; i < IPREASS_NHASH; i++) {
1210 while(!TAILQ_EMPTY(&V_ipq[i])) {
1211 IPSTAT_ADD(ips_fragdropped,
1212 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1213 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i]));
1214 }
1215 }
1216 }
1217
1218 void
ip_drain(void)1219 ip_drain(void)
1220 {
1221 VNET_ITERATOR_DECL(vnet_iter);
1222
1223 VNET_LIST_RLOCK_NOSLEEP();
1224 IPQ_LOCK();
1225 VNET_FOREACH(vnet_iter) {
1226 CURVNET_SET(vnet_iter);
1227 ip_drain_locked();
1228 CURVNET_RESTORE();
1229 }
1230 IPQ_UNLOCK();
1231 VNET_LIST_RUNLOCK_NOSLEEP();
1232 in_rtqdrain();
1233 }
1234
1235 /*
1236 * The protocol to be inserted into ip_protox[] must be already registered
1237 * in inetsw[], either statically or through pf_proto_register().
1238 */
1239 int
ipproto_register(short ipproto)1240 ipproto_register(short ipproto)
1241 {
1242 struct protosw *pr;
1243
1244 /* Sanity checks. */
1245 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
1246 return (EPROTONOSUPPORT);
1247
1248 /*
1249 * The protocol slot must not be occupied by another protocol
1250 * already. An index pointing to IPPROTO_RAW is unused.
1251 */
1252 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1253 if (pr == NULL)
1254 return (EPFNOSUPPORT);
1255 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1256 return (EEXIST);
1257
1258 /* Find the protocol position in inetsw[] and set the index. */
1259 for (pr = inetdomain.dom_protosw;
1260 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1261 if (pr->pr_domain->dom_family == PF_INET &&
1262 pr->pr_protocol && pr->pr_protocol == ipproto) {
1263 ip_protox[pr->pr_protocol] = pr - inetsw;
1264 return (0);
1265 }
1266 }
1267 return (EPROTONOSUPPORT);
1268 }
1269
1270 int
ipproto_unregister(short ipproto)1271 ipproto_unregister(short ipproto)
1272 {
1273 struct protosw *pr;
1274
1275 /* Sanity checks. */
1276 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
1277 return (EPROTONOSUPPORT);
1278
1279 /* Check if the protocol was indeed registered. */
1280 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1281 if (pr == NULL)
1282 return (EPFNOSUPPORT);
1283 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1284 return (ENOENT);
1285
1286 /* Reset the protocol slot to IPPROTO_RAW. */
1287 ip_protox[ipproto] = pr - inetsw;
1288 return (0);
1289 }
1290
1291 /*
1292 * Given address of next destination (final or next hop), return (referenced)
1293 * internet address info of interface to be used to get there.
1294 */
1295 struct in_ifaddr *
ip_rtaddr(struct in_addr dst,u_int fibnum)1296 ip_rtaddr(struct in_addr dst, u_int fibnum)
1297 {
1298 struct route sro;
1299 struct sockaddr_in *sin;
1300 struct in_ifaddr *ia;
1301
1302 bzero(&sro, sizeof(sro));
1303 sin = (struct sockaddr_in *)&sro.ro_dst;
1304 sin->sin_family = AF_INET;
1305 sin->sin_len = sizeof(*sin);
1306 sin->sin_addr = dst;
1307 in_rtalloc_ign(&sro, 0, fibnum);
1308
1309 if (sro.ro_rt == NULL)
1310 return (NULL);
1311
1312 ia = ifatoia(sro.ro_rt->rt_ifa);
1313 ifa_ref(&ia->ia_ifa);
1314 RTFREE(sro.ro_rt);
1315 return (ia);
1316 }
1317
1318 u_char inetctlerrmap[PRC_NCMDS] = {
1319 0, 0, 0, 0,
1320 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1321 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1322 EMSGSIZE, EHOSTUNREACH, 0, 0,
1323 0, 0, EHOSTUNREACH, 0,
1324 ENOPROTOOPT, ECONNREFUSED
1325 };
1326
1327 /*
1328 * Forward a packet. If some error occurs return the sender
1329 * an icmp packet. Note we can't always generate a meaningful
1330 * icmp message because icmp doesn't have a large enough repertoire
1331 * of codes and types.
1332 *
1333 * If not forwarding, just drop the packet. This could be confusing
1334 * if ipforwarding was zero but some routing protocol was advancing
1335 * us as a gateway to somewhere. However, we must let the routing
1336 * protocol deal with that.
1337 *
1338 * The srcrt parameter indicates whether the packet is being forwarded
1339 * via a source route.
1340 */
1341 void
ip_forward(struct mbuf * m,int srcrt)1342 ip_forward(struct mbuf *m, int srcrt)
1343 {
1344 struct ip *ip = mtod(m, struct ip *);
1345 struct in_ifaddr *ia;
1346 struct mbuf *mcopy;
1347 struct sockaddr_in *sin;
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 bzero(&ro, sizeof(ro));
1370 sin = (struct sockaddr_in *)&ro.ro_dst;
1371 sin->sin_family = AF_INET;
1372 sin->sin_len = sizeof(*sin);
1373 sin->sin_addr = ip->ip_dst;
1374 #ifdef RADIX_MPATH
1375 rtalloc_mpath_fib(&ro,
1376 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
1377 M_GETFIB(m));
1378 #else
1379 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
1380 #endif
1381 if (ro.ro_rt != NULL) {
1382 ia = ifatoia(ro.ro_rt->rt_ifa);
1383 ifa_ref(&ia->ia_ifa);
1384 } else
1385 ia = NULL;
1386 #ifndef IPSEC
1387 /*
1388 * 'ia' may be NULL if there is no route for this destination.
1389 * In case of IPsec, Don't discard it just yet, but pass it to
1390 * ip_output in case of outgoing IPsec policy.
1391 */
1392 if (!srcrt && ia == NULL) {
1393 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1394 RO_RTFREE(&ro);
1395 return;
1396 }
1397 #endif
1398
1399 /*
1400 * Save the IP header and at most 8 bytes of the payload,
1401 * in case we need to generate an ICMP message to the src.
1402 *
1403 * XXX this can be optimized a lot by saving the data in a local
1404 * buffer on the stack (72 bytes at most), and only allocating the
1405 * mbuf if really necessary. The vast majority of the packets
1406 * are forwarded without having to send an ICMP back (either
1407 * because unnecessary, or because rate limited), so we are
1408 * really we are wasting a lot of work here.
1409 *
1410 * We don't use m_copy() because it might return a reference
1411 * to a shared cluster. Both this function and ip_output()
1412 * assume exclusive access to the IP header in `m', so any
1413 * data in a cluster may change before we reach icmp_error().
1414 */
1415 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1416 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1417 /*
1418 * It's probably ok if the pkthdr dup fails (because
1419 * the deep copy of the tag chain failed), but for now
1420 * be conservative and just discard the copy since
1421 * code below may some day want the tags.
1422 */
1423 m_free(mcopy);
1424 mcopy = NULL;
1425 }
1426 if (mcopy != NULL) {
1427 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1428 mcopy->m_pkthdr.len = mcopy->m_len;
1429 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1430 }
1431
1432 #ifdef IPSTEALTH
1433 if (!V_ipstealth) {
1434 #endif
1435 ip->ip_ttl -= IPTTLDEC;
1436 #ifdef IPSTEALTH
1437 }
1438 #endif
1439
1440 /*
1441 * If forwarding packet using same interface that it came in on,
1442 * perhaps should send a redirect to sender to shortcut a hop.
1443 * Only send redirect if source is sending directly to us,
1444 * and if packet was not source routed (or has any options).
1445 * Also, don't send redirect if forwarding using a default route
1446 * or a route modified by a redirect.
1447 */
1448 dest.s_addr = 0;
1449 if (!srcrt && V_ipsendredirects &&
1450 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1451 struct rtentry *rt;
1452
1453 rt = ro.ro_rt;
1454
1455 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1456 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1457 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1458 u_long src = ntohl(ip->ip_src.s_addr);
1459
1460 if (RTA(rt) &&
1461 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1462 if (rt->rt_flags & RTF_GATEWAY)
1463 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1464 else
1465 dest.s_addr = ip->ip_dst.s_addr;
1466 /* Router requirements says to only send host redirects */
1467 type = ICMP_REDIRECT;
1468 code = ICMP_REDIRECT_HOST;
1469 }
1470 }
1471 }
1472
1473 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1474
1475 if (error == EMSGSIZE && ro.ro_rt)
1476 mtu = ro.ro_rt->rt_mtu;
1477 RO_RTFREE(&ro);
1478
1479 if (error)
1480 IPSTAT_INC(ips_cantforward);
1481 else {
1482 IPSTAT_INC(ips_forward);
1483 if (type)
1484 IPSTAT_INC(ips_redirectsent);
1485 else {
1486 if (mcopy)
1487 m_freem(mcopy);
1488 if (ia != NULL)
1489 ifa_free(&ia->ia_ifa);
1490 return;
1491 }
1492 }
1493 if (mcopy == NULL) {
1494 if (ia != NULL)
1495 ifa_free(&ia->ia_ifa);
1496 return;
1497 }
1498
1499 switch (error) {
1500
1501 case 0: /* forwarded, but need redirect */
1502 /* type, code set above */
1503 break;
1504
1505 case ENETUNREACH:
1506 case EHOSTUNREACH:
1507 case ENETDOWN:
1508 case EHOSTDOWN:
1509 default:
1510 type = ICMP_UNREACH;
1511 code = ICMP_UNREACH_HOST;
1512 break;
1513
1514 case EMSGSIZE:
1515 type = ICMP_UNREACH;
1516 code = ICMP_UNREACH_NEEDFRAG;
1517
1518 #ifdef IPSEC
1519 /*
1520 * If IPsec is configured for this path,
1521 * override any possibly mtu value set by ip_output.
1522 */
1523 mtu = ip_ipsec_mtu(mcopy, mtu);
1524 #endif /* IPSEC */
1525 /*
1526 * If the MTU was set before make sure we are below the
1527 * interface MTU.
1528 * If the MTU wasn't set before use the interface mtu or
1529 * fall back to the next smaller mtu step compared to the
1530 * current packet size.
1531 */
1532 if (mtu != 0) {
1533 if (ia != NULL)
1534 mtu = min(mtu, ia->ia_ifp->if_mtu);
1535 } else {
1536 if (ia != NULL)
1537 mtu = ia->ia_ifp->if_mtu;
1538 else
1539 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1540 }
1541 IPSTAT_INC(ips_cantfrag);
1542 break;
1543
1544 case ENOBUFS:
1545 /*
1546 * A router should not generate ICMP_SOURCEQUENCH as
1547 * required in RFC1812 Requirements for IP Version 4 Routers.
1548 * Source quench could be a big problem under DoS attacks,
1549 * or if the underlying interface is rate-limited.
1550 * Those who need source quench packets may re-enable them
1551 * via the net.inet.ip.sendsourcequench sysctl.
1552 */
1553 if (V_ip_sendsourcequench == 0) {
1554 m_freem(mcopy);
1555 if (ia != NULL)
1556 ifa_free(&ia->ia_ifa);
1557 return;
1558 } else {
1559 type = ICMP_SOURCEQUENCH;
1560 code = 0;
1561 }
1562 break;
1563
1564 case EACCES: /* ipfw denied packet */
1565 m_freem(mcopy);
1566 if (ia != NULL)
1567 ifa_free(&ia->ia_ifa);
1568 return;
1569 }
1570 if (ia != NULL)
1571 ifa_free(&ia->ia_ifa);
1572 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1573 }
1574
1575 void
ip_savecontrol(struct inpcb * inp,struct mbuf ** mp,struct ip * ip,struct mbuf * m)1576 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1577 struct mbuf *m)
1578 {
1579
1580 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1581 struct bintime bt;
1582
1583 bintime(&bt);
1584 if (inp->inp_socket->so_options & SO_BINTIME) {
1585 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1586 SCM_BINTIME, SOL_SOCKET);
1587 if (*mp)
1588 mp = &(*mp)->m_next;
1589 }
1590 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1591 struct timeval tv;
1592
1593 bintime2timeval(&bt, &tv);
1594 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1595 SCM_TIMESTAMP, SOL_SOCKET);
1596 if (*mp)
1597 mp = &(*mp)->m_next;
1598 }
1599 }
1600 if (inp->inp_flags & INP_RECVDSTADDR) {
1601 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1602 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1603 if (*mp)
1604 mp = &(*mp)->m_next;
1605 }
1606 if (inp->inp_flags & INP_RECVTTL) {
1607 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1608 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1609 if (*mp)
1610 mp = &(*mp)->m_next;
1611 }
1612 #ifdef notyet
1613 /* XXX
1614 * Moving these out of udp_input() made them even more broken
1615 * than they already were.
1616 */
1617 /* options were tossed already */
1618 if (inp->inp_flags & INP_RECVOPTS) {
1619 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1620 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1621 if (*mp)
1622 mp = &(*mp)->m_next;
1623 }
1624 /* ip_srcroute doesn't do what we want here, need to fix */
1625 if (inp->inp_flags & INP_RECVRETOPTS) {
1626 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1627 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1628 if (*mp)
1629 mp = &(*mp)->m_next;
1630 }
1631 #endif
1632 if (inp->inp_flags & INP_RECVIF) {
1633 struct ifnet *ifp;
1634 struct sdlbuf {
1635 struct sockaddr_dl sdl;
1636 u_char pad[32];
1637 } sdlbuf;
1638 struct sockaddr_dl *sdp;
1639 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1640
1641 if ((ifp = m->m_pkthdr.rcvif) &&
1642 ifp->if_index && ifp->if_index <= V_if_index) {
1643 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1644 /*
1645 * Change our mind and don't try copy.
1646 */
1647 if (sdp->sdl_family != AF_LINK ||
1648 sdp->sdl_len > sizeof(sdlbuf)) {
1649 goto makedummy;
1650 }
1651 bcopy(sdp, sdl2, sdp->sdl_len);
1652 } else {
1653 makedummy:
1654 sdl2->sdl_len =
1655 offsetof(struct sockaddr_dl, sdl_data[0]);
1656 sdl2->sdl_family = AF_LINK;
1657 sdl2->sdl_index = 0;
1658 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1659 }
1660 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1661 IP_RECVIF, IPPROTO_IP);
1662 if (*mp)
1663 mp = &(*mp)->m_next;
1664 }
1665 if (inp->inp_flags & INP_RECVTOS) {
1666 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1667 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1668 if (*mp)
1669 mp = &(*mp)->m_next;
1670 }
1671 }
1672
1673 /*
1674 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1675 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1676 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1677 * compiled.
1678 */
1679 static VNET_DEFINE(int, ip_rsvp_on);
1680 VNET_DEFINE(struct socket *, ip_rsvpd);
1681
1682 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1683
1684 int
ip_rsvp_init(struct socket * so)1685 ip_rsvp_init(struct socket *so)
1686 {
1687
1688 if (so->so_type != SOCK_RAW ||
1689 so->so_proto->pr_protocol != IPPROTO_RSVP)
1690 return EOPNOTSUPP;
1691
1692 if (V_ip_rsvpd != NULL)
1693 return EADDRINUSE;
1694
1695 V_ip_rsvpd = so;
1696 /*
1697 * This may seem silly, but we need to be sure we don't over-increment
1698 * the RSVP counter, in case something slips up.
1699 */
1700 if (!V_ip_rsvp_on) {
1701 V_ip_rsvp_on = 1;
1702 V_rsvp_on++;
1703 }
1704
1705 return 0;
1706 }
1707
1708 int
ip_rsvp_done(void)1709 ip_rsvp_done(void)
1710 {
1711
1712 V_ip_rsvpd = NULL;
1713 /*
1714 * This may seem silly, but we need to be sure we don't over-decrement
1715 * the RSVP counter, in case something slips up.
1716 */
1717 if (V_ip_rsvp_on) {
1718 V_ip_rsvp_on = 0;
1719 V_rsvp_on--;
1720 }
1721 return 0;
1722 }
1723
1724 void
rsvp_input(struct mbuf * m,int off)1725 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
1726 {
1727
1728 if (rsvp_input_p) { /* call the real one if loaded */
1729 rsvp_input_p(m, off);
1730 return;
1731 }
1732
1733 /* Can still get packets with rsvp_on = 0 if there is a local member
1734 * of the group to which the RSVP packet is addressed. But in this
1735 * case we want to throw the packet away.
1736 */
1737
1738 if (!V_rsvp_on) {
1739 m_freem(m);
1740 return;
1741 }
1742
1743 if (V_ip_rsvpd != NULL) {
1744 rip_input(m, off);
1745 return;
1746 }
1747 /* Drop the packet */
1748 m_freem(m);
1749 }
1750