1/*
2 * Copyright (c) 2000-2018 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28/*
29 * Copyright (c) 1982, 1986, 1988, 1993
30 * The Regents of the University of California. All rights reserved.
31 *
32 * Redistribution and use in source and binary forms, with or without
33 * modification, are permitted provided that the following conditions
34 * are met:
35 * 1. Redistributions of source code must retain the above copyright
36 * notice, this list of conditions and the following disclaimer.
37 * 2. Redistributions in binary form must reproduce the above copyright
38 * notice, this list of conditions and the following disclaimer in the
39 * documentation and/or other materials provided with the distribution.
40 * 3. All advertising materials mentioning features or use of this software
41 * must display the following acknowledgement:
42 * This product includes software developed by the University of
43 * California, Berkeley and its contributors.
44 * 4. Neither the name of the University nor the names of its contributors
45 * may be used to endorse or promote products derived from this software
46 * without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 *
60 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
61 */
62/*
63 * NOTICE: This file was modified by SPARTA, Inc. in 2007 to introduce
64 * support for mandatory and extensible security protections. This notice
65 * is included in support of clause 2.2 (b) of the Apple Public License,
66 * Version 2.0.
67 */
68
69#define _IP_VHL
70
71#include <sys/param.h>
72#include <sys/systm.h>
73#include <sys/mbuf.h>
74#include <sys/malloc.h>
75#include <sys/domain.h>
76#include <sys/protosw.h>
77#include <sys/socket.h>
78#include <sys/time.h>
79#include <sys/kernel.h>
80#include <sys/syslog.h>
81#include <sys/sysctl.h>
82#include <sys/mcache.h>
83#include <sys/socketvar.h>
84#include <sys/kdebug.h>
85#include <mach/mach_time.h>
86#include <mach/sdt.h>
87
88#include <machine/endian.h>
89#include <dev/random/randomdev.h>
90
91#include <kern/queue.h>
92#include <kern/locks.h>
93#include <libkern/OSAtomic.h>
94
95#include <pexpert/pexpert.h>
96
97#include <net/if.h>
98#include <net/if_var.h>
99#include <net/if_dl.h>
100#include <net/route.h>
101#include <net/kpi_protocol.h>
102#include <net/ntstat.h>
103#include <net/dlil.h>
104#include <net/classq/classq.h>
105#include <net/net_perf.h>
106#include <net/init.h>
107#if PF
108#include <net/pfvar.h>
109#endif /* PF */
110
111#include <netinet/in.h>
112#include <netinet/in_systm.h>
113#include <netinet/in_var.h>
114#include <netinet/in_arp.h>
115#include <netinet/ip.h>
116#include <netinet/in_pcb.h>
117#include <netinet/ip_var.h>
118#include <netinet/ip_icmp.h>
119#include <netinet/ip_fw.h>
120#include <netinet/ip_divert.h>
121#include <netinet/kpi_ipfilter_var.h>
122#include <netinet/udp.h>
123#include <netinet/udp_var.h>
124#include <netinet/bootp.h>
125#include <netinet/lro_ext.h>
126
127#if DUMMYNET
128#include <netinet/ip_dummynet.h>
129#endif /* DUMMYNET */
130
131#if CONFIG_MACF_NET
132#include <security/mac_framework.h>
133#endif /* CONFIG_MACF_NET */
134
135#if IPSEC
136#include <netinet6/ipsec.h>
137#include <netkey/key.h>
138#endif /* IPSEC */
139
140#define DBG_LAYER_BEG NETDBG_CODE(DBG_NETIP, 0)
141#define DBG_LAYER_END NETDBG_CODE(DBG_NETIP, 2)
142#define DBG_FNC_IP_INPUT NETDBG_CODE(DBG_NETIP, (2 << 8))
143
144#if IPSEC
145extern int ipsec_bypass;
146extern lck_mtx_t *sadb_mutex;
147
148lck_grp_t *sadb_stat_mutex_grp;
149lck_grp_attr_t *sadb_stat_mutex_grp_attr;
150lck_attr_t *sadb_stat_mutex_attr;
151decl_lck_mtx_data(, sadb_stat_mutex_data);
152lck_mtx_t *sadb_stat_mutex = &sadb_stat_mutex_data;
153#endif /* IPSEC */
154
155MBUFQ_HEAD(fq_head);
156
157static int frag_timeout_run; /* frag timer is scheduled to run */
158static void frag_timeout(void *);
159static void frag_sched_timeout(void);
160
161static struct ipq *ipq_alloc(int);
162static void ipq_free(struct ipq *);
163static void ipq_updateparams(void);
164static void ip_input_second_pass(struct mbuf *, struct ifnet *,
165 u_int32_t, int, int, struct ip_fw_in_args *, int);
166
167decl_lck_mtx_data(static, ipqlock);
168static lck_attr_t *ipqlock_attr;
169static lck_grp_t *ipqlock_grp;
170static lck_grp_attr_t *ipqlock_grp_attr;
171
172/* Packet reassembly stuff */
173#define IPREASS_NHASH_LOG2 6
174#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
175#define IPREASS_HMASK (IPREASS_NHASH - 1)
176#define IPREASS_HASH(x, y) \
177 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
178
179/* IP fragment reassembly queues (protected by ipqlock) */
180static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; /* ip reassembly queues */
181static int maxnipq; /* max packets in reass queues */
182static u_int32_t maxfragsperpacket; /* max frags/packet in reass queues */
183static u_int32_t nipq; /* # of packets in reass queues */
184static u_int32_t ipq_limit; /* ipq allocation limit */
185static u_int32_t ipq_count; /* current # of allocated ipq's */
186
187static int sysctl_ipforwarding SYSCTL_HANDLER_ARGS;
188static int sysctl_maxnipq SYSCTL_HANDLER_ARGS;
189static int sysctl_maxfragsperpacket SYSCTL_HANDLER_ARGS;
190
191#if (DEBUG || DEVELOPMENT)
192static int sysctl_reset_ip_input_stats SYSCTL_HANDLER_ARGS;
193static int sysctl_ip_input_measure_bins SYSCTL_HANDLER_ARGS;
194static int sysctl_ip_input_getperf SYSCTL_HANDLER_ARGS;
195#endif /* (DEBUG || DEVELOPMENT) */
196
197int ipforwarding = 0;
198SYSCTL_PROC(_net_inet_ip, IPCTL_FORWARDING, forwarding,
199 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &ipforwarding, 0,
200 sysctl_ipforwarding, "I", "Enable IP forwarding between interfaces");
201
202static int ipsendredirects = 1; /* XXX */
203SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect,
204 CTLFLAG_RW | CTLFLAG_LOCKED, &ipsendredirects, 0,
205 "Enable sending IP redirects");
206
207int ip_defttl = IPDEFTTL;
208SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW | CTLFLAG_LOCKED,
209 &ip_defttl, 0, "Maximum TTL on IP packets");
210
211static int ip_dosourceroute = 0;
212SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute,
213 CTLFLAG_RW | CTLFLAG_LOCKED, &ip_dosourceroute, 0,
214 "Enable forwarding source routed IP packets");
215
216static int ip_acceptsourceroute = 0;
217SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
218 CTLFLAG_RW | CTLFLAG_LOCKED, &ip_acceptsourceroute, 0,
219 "Enable accepting source routed IP packets");
220
221static int ip_sendsourcequench = 0;
222SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench,
223 CTLFLAG_RW | CTLFLAG_LOCKED, &ip_sendsourcequench, 0,
224 "Enable the transmission of source quench packets");
225
226SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets,
227 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &maxnipq, 0, sysctl_maxnipq,
228 "I", "Maximum number of IPv4 fragment reassembly queue entries");
229
230SYSCTL_UINT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD | CTLFLAG_LOCKED,
231 &nipq, 0, "Current number of IPv4 fragment reassembly queue entries");
232
233SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragsperpacket,
234 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &maxfragsperpacket, 0,
235 sysctl_maxfragsperpacket, "I",
236 "Maximum number of IPv4 fragments allowed per packet");
237
238static uint32_t ip_adj_clear_hwcksum = 0;
239SYSCTL_UINT(_net_inet_ip, OID_AUTO, adj_clear_hwcksum,
240 CTLFLAG_RW | CTLFLAG_LOCKED, &ip_adj_clear_hwcksum, 0,
241 "Invalidate hwcksum info when adjusting length");
242
243static uint32_t ip_adj_partial_sum = 1;
244SYSCTL_UINT(_net_inet_ip, OID_AUTO, adj_partial_sum,
245 CTLFLAG_RW | CTLFLAG_LOCKED, &ip_adj_partial_sum, 0,
246 "Perform partial sum adjustment of trailing bytes at IP layer");
247
248/*
249 * XXX - Setting ip_checkinterface mostly implements the receive side of
250 * the Strong ES model described in RFC 1122, but since the routing table
251 * and transmit implementation do not implement the Strong ES model,
252 * setting this to 1 results in an odd hybrid.
253 *
254 * XXX - ip_checkinterface currently must be disabled if you use ipnat
255 * to translate the destination address to another local interface.
256 *
257 * XXX - ip_checkinterface must be disabled if you add IP aliases
258 * to the loopback interface instead of the interface where the
259 * packets for those addresses are received.
260 */
261static int ip_checkinterface = 0;
262SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW | CTLFLAG_LOCKED,
263 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
264
265static int ip_chaining = 1;
266SYSCTL_INT(_net_inet_ip, OID_AUTO, rx_chaining, CTLFLAG_RW | CTLFLAG_LOCKED,
267 &ip_chaining, 1, "Do receive side ip address based chaining");
268
269static int ip_chainsz = 6;
270SYSCTL_INT(_net_inet_ip, OID_AUTO, rx_chainsz, CTLFLAG_RW | CTLFLAG_LOCKED,
271 &ip_chainsz, 1, "IP receive side max chaining");
272
273#if (DEBUG || DEVELOPMENT)
274static int ip_input_measure = 0;
275SYSCTL_PROC(_net_inet_ip, OID_AUTO, input_perf,
276 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
277 &ip_input_measure, 0, sysctl_reset_ip_input_stats, "I", "Do time measurement");
278
279static uint64_t ip_input_measure_bins = 0;
280SYSCTL_PROC(_net_inet_ip, OID_AUTO, input_perf_bins,
281 CTLTYPE_QUAD | CTLFLAG_RW | CTLFLAG_LOCKED, &ip_input_measure_bins, 0,
282 sysctl_ip_input_measure_bins, "I",
283 "bins for chaining performance data histogram");
284
285static net_perf_t net_perf;
286SYSCTL_PROC(_net_inet_ip, OID_AUTO, input_perf_data,
287 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED,
288 0, 0, sysctl_ip_input_getperf, "S,net_perf",
289 "IP input performance data (struct net_perf, net/net_perf.h)");
290#endif /* (DEBUG || DEVELOPMENT) */
291
292#if DIAGNOSTIC
293static int ipprintfs = 0;
294#endif
295
296struct protosw *ip_protox[IPPROTO_MAX];
297
298static lck_grp_attr_t *in_ifaddr_rwlock_grp_attr;
299static lck_grp_t *in_ifaddr_rwlock_grp;
300static lck_attr_t *in_ifaddr_rwlock_attr;
301decl_lck_rw_data(, in_ifaddr_rwlock_data);
302lck_rw_t *in_ifaddr_rwlock = &in_ifaddr_rwlock_data;
303
304/* Protected by in_ifaddr_rwlock */
305struct in_ifaddrhead in_ifaddrhead; /* first inet address */
306struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
307
308#define INADDR_NHASH 61
309static u_int32_t inaddr_nhash; /* hash table size */
310static u_int32_t inaddr_hashp; /* next largest prime */
311
312static int ip_getstat SYSCTL_HANDLER_ARGS;
313struct ipstat ipstat;
314SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats,
315 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED,
316 0, 0, ip_getstat, "S,ipstat",
317 "IP statistics (struct ipstat, netinet/ip_var.h)");
318
319#if IPCTL_DEFMTU
320SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW | CTLFLAG_LOCKED,
321 &ip_mtu, 0, "Default MTU");
322#endif /* IPCTL_DEFMTU */
323
324#if IPSTEALTH
325static int ipstealth = 0;
326SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW | CTLFLAG_LOCKED,
327 &ipstealth, 0, "");
328#endif /* IPSTEALTH */
329
330/* Firewall hooks */
331#if IPFIREWALL
332ip_fw_chk_t *ip_fw_chk_ptr;
333int fw_enable = 1;
334int fw_bypass = 1;
335int fw_one_pass = 0;
336#endif /* IPFIREWALL */
337
338#if DUMMYNET
339ip_dn_io_t *ip_dn_io_ptr;
340#endif /* DUMMYNET */
341
342SYSCTL_NODE(_net_inet_ip, OID_AUTO, linklocal,
343 CTLFLAG_RW | CTLFLAG_LOCKED, 0, "link local");
344
345struct ip_linklocal_stat ip_linklocal_stat;
346SYSCTL_STRUCT(_net_inet_ip_linklocal, OID_AUTO, stat,
347 CTLFLAG_RD | CTLFLAG_LOCKED, &ip_linklocal_stat, ip_linklocal_stat,
348 "Number of link local packets with TTL less than 255");
349
350SYSCTL_NODE(_net_inet_ip_linklocal, OID_AUTO, in,
351 CTLFLAG_RW | CTLFLAG_LOCKED, 0, "link local input");
352
353int ip_linklocal_in_allowbadttl = 1;
354SYSCTL_INT(_net_inet_ip_linklocal_in, OID_AUTO, allowbadttl,
355 CTLFLAG_RW | CTLFLAG_LOCKED, &ip_linklocal_in_allowbadttl, 0,
356 "Allow incoming link local packets with TTL less than 255");
357
358
359/*
360 * We need to save the IP options in case a protocol wants to respond
361 * to an incoming packet over the same route if the packet got here
362 * using IP source routing. This allows connection establishment and
363 * maintenance when the remote end is on a network that is not known
364 * to us.
365 */
366static int ip_nhops = 0;
367static struct ip_srcrt {
368 struct in_addr dst; /* final destination */
369 char nop; /* one NOP to align */
370 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
371 struct in_addr route[MAX_IPOPTLEN / sizeof (struct in_addr)];
372} ip_srcrt;
373
374static void in_ifaddrhashtbl_init(void);
375static void save_rte(u_char *, struct in_addr);
376static int ip_dooptions(struct mbuf *, int, struct sockaddr_in *);
377static void ip_forward(struct mbuf *, int, struct sockaddr_in *);
378static void frag_freef(struct ipqhead *, struct ipq *);
379#if IPDIVERT
380#ifdef IPDIVERT_44
381static struct mbuf *ip_reass(struct mbuf *, u_int32_t *, u_int16_t *);
382#else /* !IPDIVERT_44 */
383static struct mbuf *ip_reass(struct mbuf *, u_int16_t *, u_int16_t *);
384#endif /* !IPDIVERT_44 */
385#else /* !IPDIVERT */
386static struct mbuf *ip_reass(struct mbuf *);
387#endif /* !IPDIVERT */
388static void ip_fwd_route_copyout(struct ifnet *, struct route *);
389static void ip_fwd_route_copyin(struct ifnet *, struct route *);
390static inline u_short ip_cksum(struct mbuf *, int);
391
392int ip_use_randomid = 1;
393SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW | CTLFLAG_LOCKED,
394 &ip_use_randomid, 0, "Randomize IP packets IDs");
395
396/*
397 * On platforms which require strict alignment (currently for anything but
398 * i386 or x86_64), check if the IP header pointer is 32-bit aligned; if not,
399 * copy the contents of the mbuf chain into a new chain, and free the original
400 * one. Create some head room in the first mbuf of the new chain, in case
401 * it's needed later on.
402 */
403#if defined(__i386__) || defined(__x86_64__)
404#define IP_HDR_ALIGNMENT_FIXUP(_m, _ifp, _action) do { } while (0)
405#else /* !__i386__ && !__x86_64__ */
406#define IP_HDR_ALIGNMENT_FIXUP(_m, _ifp, _action) do { \
407 if (!IP_HDR_ALIGNED_P(mtod(_m, caddr_t))) { \
408 struct mbuf *_n; \
409 struct ifnet *__ifp = (_ifp); \
410 atomic_add_64(&(__ifp)->if_alignerrs, 1); \
411 if (((_m)->m_flags & M_PKTHDR) && \
412 (_m)->m_pkthdr.pkt_hdr != NULL) \
413 (_m)->m_pkthdr.pkt_hdr = NULL; \
414 _n = m_defrag_offset(_m, max_linkhdr, M_NOWAIT); \
415 if (_n == NULL) { \
416 atomic_add_32(&ipstat.ips_toosmall, 1); \
417 m_freem(_m); \
418 (_m) = NULL; \
419 _action; \
420 } else { \
421 VERIFY(_n != (_m)); \
422 (_m) = _n; \
423 } \
424 } \
425} while (0)
426#endif /* !__i386__ && !__x86_64__ */
427
428/*
429 * GRE input handler function, settable via ip_gre_register_input() for PPTP.
430 */
431static gre_input_func_t gre_input_func;
432
433static void
434ip_init_delayed(void)
435{
436 struct ifreq ifr;
437 int error;
438 struct sockaddr_in *sin;
439
440 bzero(&ifr, sizeof(ifr));
441 strlcpy(ifr.ifr_name, "lo0", sizeof(ifr.ifr_name));
442 sin = (struct sockaddr_in *)(void *)&ifr.ifr_addr;
443 sin->sin_len = sizeof(struct sockaddr_in);
444 sin->sin_family = AF_INET;
445 sin->sin_addr.s_addr = htonl(INADDR_LOOPBACK);
446 error = in_control(NULL, SIOCSIFADDR, (caddr_t)&ifr, lo_ifp, kernproc);
447 if (error)
448 printf("%s: failed to initialise lo0's address, error=%d\n",
449 __func__, error);
450}
451
452/*
453 * IP initialization: fill in IP protocol switch table.
454 * All protocols not implemented in kernel go to raw IP protocol handler.
455 */
456void
457ip_init(struct protosw *pp, struct domain *dp)
458{
459 static int ip_initialized = 0;
460 struct protosw *pr;
461 struct timeval tv;
462 int i;
463
464 domain_proto_mtx_lock_assert_held();
465 VERIFY((pp->pr_flags & (PR_INITIALIZED|PR_ATTACHED)) == PR_ATTACHED);
466
467 /* ipq_alloc() uses mbufs for IP fragment queue structures */
468 _CASSERT(sizeof (struct ipq) <= _MLEN);
469
470 /*
471 * Some ioctls (e.g. SIOCAIFADDR) use ifaliasreq struct, which is
472 * interchangeable with in_aliasreq; they must have the same size.
473 */
474 _CASSERT(sizeof (struct ifaliasreq) == sizeof (struct in_aliasreq));
475
476 if (ip_initialized)
477 return;
478 ip_initialized = 1;
479
480 in_ifaddr_init();
481
482 in_ifaddr_rwlock_grp_attr = lck_grp_attr_alloc_init();
483 in_ifaddr_rwlock_grp = lck_grp_alloc_init("in_ifaddr_rwlock",
484 in_ifaddr_rwlock_grp_attr);
485 in_ifaddr_rwlock_attr = lck_attr_alloc_init();
486 lck_rw_init(in_ifaddr_rwlock, in_ifaddr_rwlock_grp,
487 in_ifaddr_rwlock_attr);
488
489 TAILQ_INIT(&in_ifaddrhead);
490 in_ifaddrhashtbl_init();
491
492 ip_moptions_init();
493
494 pr = pffindproto_locked(PF_INET, IPPROTO_RAW, SOCK_RAW);
495 if (pr == NULL) {
496 panic("%s: Unable to find [PF_INET,IPPROTO_RAW,SOCK_RAW]\n",
497 __func__);
498 /* NOTREACHED */
499 }
500
501 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
502 for (i = 0; i < IPPROTO_MAX; i++)
503 ip_protox[i] = pr;
504 /*
505 * Cycle through IP protocols and put them into the appropriate place
506 * in ip_protox[], skipping protocols IPPROTO_{IP,RAW}.
507 */
508 VERIFY(dp == inetdomain && dp->dom_family == PF_INET);
509 TAILQ_FOREACH(pr, &dp->dom_protosw, pr_entry) {
510 VERIFY(pr->pr_domain == dp);
511 if (pr->pr_protocol != 0 && pr->pr_protocol != IPPROTO_RAW) {
512 /* Be careful to only index valid IP protocols. */
513 if (pr->pr_protocol < IPPROTO_MAX)
514 ip_protox[pr->pr_protocol] = pr;
515 }
516 }
517
518 /* IP fragment reassembly queue lock */
519 ipqlock_grp_attr = lck_grp_attr_alloc_init();
520 ipqlock_grp = lck_grp_alloc_init("ipqlock", ipqlock_grp_attr);
521 ipqlock_attr = lck_attr_alloc_init();
522 lck_mtx_init(&ipqlock, ipqlock_grp, ipqlock_attr);
523
524 lck_mtx_lock(&ipqlock);
525 /* Initialize IP reassembly queue. */
526 for (i = 0; i < IPREASS_NHASH; i++)
527 TAILQ_INIT(&ipq[i]);
528
529 maxnipq = nmbclusters / 32;
530 maxfragsperpacket = 128; /* enough for 64k in 512 byte fragments */
531 ipq_updateparams();
532 lck_mtx_unlock(&ipqlock);
533
534 getmicrotime(&tv);
535 ip_id = RandomULong() ^ tv.tv_usec;
536 ip_initid();
537
538 ipf_init();
539
540#if IPSEC
541 sadb_stat_mutex_grp_attr = lck_grp_attr_alloc_init();
542 sadb_stat_mutex_grp = lck_grp_alloc_init("sadb_stat",
543 sadb_stat_mutex_grp_attr);
544 sadb_stat_mutex_attr = lck_attr_alloc_init();
545 lck_mtx_init(sadb_stat_mutex, sadb_stat_mutex_grp,
546 sadb_stat_mutex_attr);
547
548#endif
549 arp_init();
550 net_init_add(ip_init_delayed);
551}
552
553/*
554 * Initialize IPv4 source address hash table.
555 */
556static void
557in_ifaddrhashtbl_init(void)
558{
559 int i, k, p;
560
561 if (in_ifaddrhashtbl != NULL)
562 return;
563
564 PE_parse_boot_argn("inaddr_nhash", &inaddr_nhash,
565 sizeof (inaddr_nhash));
566 if (inaddr_nhash == 0)
567 inaddr_nhash = INADDR_NHASH;
568
569 MALLOC(in_ifaddrhashtbl, struct in_ifaddrhashhead *,
570 inaddr_nhash * sizeof (*in_ifaddrhashtbl),
571 M_IFADDR, M_WAITOK | M_ZERO);
572 if (in_ifaddrhashtbl == NULL)
573 panic("in_ifaddrhashtbl_init allocation failed");
574
575 /*
576 * Generate the next largest prime greater than inaddr_nhash.
577 */
578 k = (inaddr_nhash % 2 == 0) ? inaddr_nhash + 1 : inaddr_nhash + 2;
579 for (;;) {
580 p = 1;
581 for (i = 3; i * i <= k; i += 2) {
582 if (k % i == 0)
583 p = 0;
584 }
585 if (p == 1)
586 break;
587 k += 2;
588 }
589 inaddr_hashp = k;
590}
591
592u_int32_t
593inaddr_hashval(u_int32_t key)
594{
595 /*
596 * The hash index is the computed prime times the key modulo
597 * the hash size, as documented in "Introduction to Algorithms"
598 * (Cormen, Leiserson, Rivest).
599 */
600 if (inaddr_nhash > 1)
601 return ((key * inaddr_hashp) % inaddr_nhash);
602 else
603 return (0);
604}
605
606void
607ip_proto_dispatch_in_wrapper(struct mbuf *m, int hlen, u_int8_t proto)
608{
609 ip_proto_dispatch_in(m, hlen, proto, 0);
610}
611
612__private_extern__ void
613ip_proto_dispatch_in(struct mbuf *m, int hlen, u_int8_t proto,
614 ipfilter_t inject_ipfref)
615{
616 struct ipfilter *filter;
617 int seen = (inject_ipfref == NULL);
618 int changed_header = 0;
619 struct ip *ip;
620 void (*pr_input)(struct mbuf *, int len);
621
622 if (!TAILQ_EMPTY(&ipv4_filters)) {
623 ipf_ref();
624 TAILQ_FOREACH(filter, &ipv4_filters, ipf_link) {
625 if (seen == 0) {
626 if ((struct ipfilter *)inject_ipfref == filter)
627 seen = 1;
628 } else if (filter->ipf_filter.ipf_input) {
629 errno_t result;
630
631 if (changed_header == 0) {
632 /*
633 * Perform IP header alignment fixup,
634 * if needed, before passing packet
635 * into filter(s).
636 */
637 IP_HDR_ALIGNMENT_FIXUP(m,
638 m->m_pkthdr.rcvif, ipf_unref());
639
640 /* ipf_unref() already called */
641 if (m == NULL)
642 return;
643
644 changed_header = 1;
645 ip = mtod(m, struct ip *);
646 ip->ip_len = htons(ip->ip_len + hlen);
647 ip->ip_off = htons(ip->ip_off);
648 ip->ip_sum = 0;
649 ip->ip_sum = ip_cksum_hdr_in(m, hlen);
650 }
651 result = filter->ipf_filter.ipf_input(
652 filter->ipf_filter.cookie, (mbuf_t *)&m,
653 hlen, proto);
654 if (result == EJUSTRETURN) {
655 ipf_unref();
656 return;
657 }
658 if (result != 0) {
659 ipf_unref();
660 m_freem(m);
661 return;
662 }
663 }
664 }
665 ipf_unref();
666 }
667
668 /* Perform IP header alignment fixup (post-filters), if needed */
669 IP_HDR_ALIGNMENT_FIXUP(m, m->m_pkthdr.rcvif, return);
670
671 /*
672 * If there isn't a specific lock for the protocol
673 * we're about to call, use the generic lock for AF_INET.
674 * otherwise let the protocol deal with its own locking
675 */
676 ip = mtod(m, struct ip *);
677
678 if (changed_header) {
679 ip->ip_len = ntohs(ip->ip_len) - hlen;
680 ip->ip_off = ntohs(ip->ip_off);
681 }
682
683 if ((pr_input = ip_protox[ip->ip_p]->pr_input) == NULL) {
684 m_freem(m);
685 } else if (!(ip_protox[ip->ip_p]->pr_flags & PR_PROTOLOCK)) {
686 lck_mtx_lock(inet_domain_mutex);
687 pr_input(m, hlen);
688 lck_mtx_unlock(inet_domain_mutex);
689 } else {
690 pr_input(m, hlen);
691 }
692}
693
694struct pktchain_elm {
695 struct mbuf *pkte_head;
696 struct mbuf *pkte_tail;
697 struct in_addr pkte_saddr;
698 struct in_addr pkte_daddr;
699 uint16_t pkte_npkts;
700 uint16_t pkte_proto;
701 uint32_t pkte_nbytes;
702};
703
704typedef struct pktchain_elm pktchain_elm_t;
705
706/* Store upto PKTTBL_SZ unique flows on the stack */
707#define PKTTBL_SZ 7
708
709static struct mbuf *
710ip_chain_insert(struct mbuf *packet, pktchain_elm_t *tbl)
711{
712 struct ip* ip;
713 int pkttbl_idx = 0;
714
715 ip = mtod(packet, struct ip*);
716
717 /* reusing the hash function from inaddr_hashval */
718 pkttbl_idx = inaddr_hashval(ntohs(ip->ip_src.s_addr)) % PKTTBL_SZ;
719 if (tbl[pkttbl_idx].pkte_head == NULL) {
720 tbl[pkttbl_idx].pkte_head = packet;
721 tbl[pkttbl_idx].pkte_saddr.s_addr = ip->ip_src.s_addr;
722 tbl[pkttbl_idx].pkte_daddr.s_addr = ip->ip_dst.s_addr;
723 tbl[pkttbl_idx].pkte_proto = ip->ip_p;
724 } else {
725 if ((ip->ip_dst.s_addr == tbl[pkttbl_idx].pkte_daddr.s_addr) &&
726 (ip->ip_src.s_addr == tbl[pkttbl_idx].pkte_saddr.s_addr) &&
727 (ip->ip_p == tbl[pkttbl_idx].pkte_proto)) {
728 } else {
729 return (packet);
730 }
731 }
732 if (tbl[pkttbl_idx].pkte_tail != NULL)
733 mbuf_setnextpkt(tbl[pkttbl_idx].pkte_tail, packet);
734
735 tbl[pkttbl_idx].pkte_tail = packet;
736 tbl[pkttbl_idx].pkte_npkts += 1;
737 tbl[pkttbl_idx].pkte_nbytes += packet->m_pkthdr.len;
738 return (NULL);
739}
740
741/* args is a dummy variable here for backward compatibility */
742static void
743ip_input_second_pass_loop_tbl(pktchain_elm_t *tbl, struct ip_fw_in_args *args)
744{
745 int i = 0;
746
747 for (i = 0; i < PKTTBL_SZ; i++) {
748 if (tbl[i].pkte_head != NULL) {
749 struct mbuf *m = tbl[i].pkte_head;
750 ip_input_second_pass(m, m->m_pkthdr.rcvif, 0,
751 tbl[i].pkte_npkts, tbl[i].pkte_nbytes, args, 0);
752
753 if (tbl[i].pkte_npkts > 2)
754 ipstat.ips_rxc_chainsz_gt2++;
755 if (tbl[i].pkte_npkts > 4)
756 ipstat.ips_rxc_chainsz_gt4++;
757#if (DEBUG || DEVELOPMENT)
758 if (ip_input_measure)
759 net_perf_histogram(&net_perf, tbl[i].pkte_npkts);
760#endif /* (DEBUG || DEVELOPMENT) */
761 tbl[i].pkte_head = tbl[i].pkte_tail = NULL;
762 tbl[i].pkte_npkts = 0;
763 tbl[i].pkte_nbytes = 0;
764 /* no need to initialize address and protocol in tbl */
765 }
766 }
767}
768
769static void
770ip_input_cpout_args(struct ip_fw_in_args *args, struct ip_fw_args *args1,
771 boolean_t *done_init)
772{
773 if (*done_init == FALSE) {
774 bzero(args1, sizeof(struct ip_fw_args));
775 *done_init = TRUE;
776 }
777 args1->fwa_next_hop = args->fwai_next_hop;
778 args1->fwa_ipfw_rule = args->fwai_ipfw_rule;
779 args1->fwa_pf_rule = args->fwai_pf_rule;
780 args1->fwa_divert_rule = args->fwai_divert_rule;
781}
782
783static void
784ip_input_cpin_args(struct ip_fw_args *args1, struct ip_fw_in_args *args)
785{
786 args->fwai_next_hop = args1->fwa_next_hop;
787 args->fwai_ipfw_rule = args1->fwa_ipfw_rule;
788 args->fwai_pf_rule = args1->fwa_pf_rule;
789 args->fwai_divert_rule = args1->fwa_divert_rule;
790}
791
792typedef enum {
793 IPINPUT_DOCHAIN = 0,
794 IPINPUT_DONTCHAIN,
795 IPINPUT_FREED,
796 IPINPUT_DONE
797} ipinput_chain_ret_t;
798
799static void
800ip_input_update_nstat(struct ifnet *ifp, struct in_addr src_ip,
801 u_int32_t packets, u_int32_t bytes)
802{
803 if (nstat_collect) {
804 struct rtentry *rt = ifnet_cached_rtlookup_inet(ifp,
805 src_ip);
806 if (rt != NULL) {
807 nstat_route_rx(rt, packets, bytes, 0);
808 rtfree(rt);
809 }
810 }
811}
812
813static void
814ip_input_dispatch_chain(struct mbuf *m)
815{
816 struct mbuf *tmp_mbuf = m;
817 struct mbuf *nxt_mbuf = NULL;
818 struct ip *ip = NULL;
819 unsigned int hlen;
820
821 ip = mtod(tmp_mbuf, struct ip *);
822 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
823 while(tmp_mbuf) {
824 nxt_mbuf = mbuf_nextpkt(tmp_mbuf);
825 mbuf_setnextpkt(tmp_mbuf, NULL);
826
827 if ((sw_lro) && (ip->ip_p == IPPROTO_TCP))
828 tmp_mbuf = tcp_lro(tmp_mbuf, hlen);
829 if (tmp_mbuf)
830 ip_proto_dispatch_in(tmp_mbuf, hlen, ip->ip_p, 0);
831 tmp_mbuf = nxt_mbuf;
832 if (tmp_mbuf) {
833 ip = mtod(tmp_mbuf, struct ip *);
834 /* first mbuf of chain already has adjusted ip_len */
835 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
836 ip->ip_len -= hlen;
837 }
838 }
839}
840
841static void
842ip_input_setdst_chain(struct mbuf *m, uint32_t ifindex, struct in_ifaddr *ia)
843{
844 struct mbuf *tmp_mbuf = m;
845
846 while (tmp_mbuf) {
847 ip_setdstifaddr_info(tmp_mbuf, ifindex, ia);
848 tmp_mbuf = mbuf_nextpkt(tmp_mbuf);
849 }
850}
851
852static void
853ip_input_adjust(struct mbuf *m, struct ip *ip, struct ifnet *inifp)
854{
855 boolean_t adjust = TRUE;
856
857 ASSERT(m_pktlen(m) > ip->ip_len);
858
859 /*
860 * Invalidate hardware checksum info if ip_adj_clear_hwcksum
861 * is set; useful to handle buggy drivers. Note that this
862 * should not be enabled by default, as we may get here due
863 * to link-layer padding.
864 */
865 if (ip_adj_clear_hwcksum &&
866 (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) &&
867 !(inifp->if_flags & IFF_LOOPBACK) &&
868 !(m->m_pkthdr.pkt_flags & PKTF_LOOP)) {
869 m->m_pkthdr.csum_flags &= ~CSUM_DATA_VALID;
870 m->m_pkthdr.csum_data = 0;
871 ipstat.ips_adj_hwcsum_clr++;
872 }
873
874 /*
875 * If partial checksum information is available, subtract
876 * out the partial sum of postpended extraneous bytes, and
877 * update the checksum metadata accordingly. By doing it
878 * here, the upper layer transport only needs to adjust any
879 * prepended extraneous bytes (else it will do both.)
880 */
881 if (ip_adj_partial_sum &&
882 (m->m_pkthdr.csum_flags & (CSUM_DATA_VALID|CSUM_PARTIAL)) ==
883 (CSUM_DATA_VALID|CSUM_PARTIAL)) {
884 m->m_pkthdr.csum_rx_val = m_adj_sum16(m,
885 m->m_pkthdr.csum_rx_start, m->m_pkthdr.csum_rx_start,
886 (ip->ip_len - m->m_pkthdr.csum_rx_start),
887 m->m_pkthdr.csum_rx_val);
888 } else if ((m->m_pkthdr.csum_flags &
889 (CSUM_DATA_VALID|CSUM_PARTIAL)) ==
890 (CSUM_DATA_VALID|CSUM_PARTIAL)) {
891 /*
892 * If packet has partial checksum info and we decided not
893 * to subtract the partial sum of postpended extraneous
894 * bytes here (not the default case), leave that work to
895 * be handled by the other layers. For now, only TCP, UDP
896 * layers are capable of dealing with this. For all other
897 * protocols (including fragments), trim and ditch the
898 * partial sum as those layers might not implement partial
899 * checksumming (or adjustment) at all.
900 */
901 if ((ip->ip_off & (IP_MF | IP_OFFMASK)) == 0 &&
902 (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_UDP)) {
903 adjust = FALSE;
904 } else {
905 m->m_pkthdr.csum_flags &= ~CSUM_DATA_VALID;
906 m->m_pkthdr.csum_data = 0;
907 ipstat.ips_adj_hwcsum_clr++;
908 }
909 }
910
911 if (adjust) {
912 ipstat.ips_adj++;
913 if (m->m_len == m->m_pkthdr.len) {
914 m->m_len = ip->ip_len;
915 m->m_pkthdr.len = ip->ip_len;
916 } else {
917 m_adj(m, ip->ip_len - m->m_pkthdr.len);
918 }
919 }
920}
921
922/*
923 * First pass does all essential packet validation and places on a per flow
924 * queue for doing operations that have same outcome for all packets of a flow.
925 * div_info is packet divert/tee info
926 */
927static ipinput_chain_ret_t
928ip_input_first_pass(struct mbuf *m, u_int32_t *div_info,
929 struct ip_fw_in_args *args, int *ours, struct mbuf **modm)
930{
931 struct ip *ip;
932 struct ifnet *inifp;
933 unsigned int hlen;
934 int retval = IPINPUT_DOCHAIN;
935 int len = 0;
936 struct in_addr src_ip;
937#if IPFIREWALL
938 int i;
939#endif
940#if IPFIREWALL || DUMMYNET
941 struct m_tag *copy;
942 struct m_tag *p;
943 boolean_t delete = FALSE;
944 struct ip_fw_args args1;
945 boolean_t init = FALSE;
946#endif
947 ipfilter_t inject_filter_ref = NULL;
948
949#if !IPFIREWALL
950#pragma unused (args)
951#endif
952
953#if !IPDIVERT
954#pragma unused (div_info)
955#pragma unused (ours)
956#endif
957
958#if !IPFIREWALL_FORWARD
959#pragma unused (ours)
960#endif
961
962 /* Check if the mbuf is still valid after interface filter processing */
963 MBUF_INPUT_CHECK(m, m->m_pkthdr.rcvif);
964 inifp = mbuf_pkthdr_rcvif(m);
965 VERIFY(inifp != NULL);
966
967 /* Perform IP header alignment fixup, if needed */
968 IP_HDR_ALIGNMENT_FIXUP(m, inifp, goto bad);
969
970 m->m_pkthdr.pkt_flags &= ~PKTF_FORWARDED;
971
972#if IPFIREWALL || DUMMYNET
973
974 /*
975 * Don't bother searching for tag(s) if there's none.
976 */
977 if (SLIST_EMPTY(&m->m_pkthdr.tags))
978 goto ipfw_tags_done;
979
980 /* Grab info from mtags prepended to the chain */
981 p = m_tag_first(m);
982 while (p) {
983 if (p->m_tag_id == KERNEL_MODULE_TAG_ID) {
984#if DUMMYNET
985 if (p->m_tag_type == KERNEL_TAG_TYPE_DUMMYNET) {
986 struct dn_pkt_tag *dn_tag;
987
988 dn_tag = (struct dn_pkt_tag *)(p+1);
989 args->fwai_ipfw_rule = dn_tag->dn_ipfw_rule;
990 args->fwai_pf_rule = dn_tag->dn_pf_rule;
991 delete = TRUE;
992 }
993#endif
994
995#if IPDIVERT
996 if (p->m_tag_type == KERNEL_TAG_TYPE_DIVERT) {
997 struct divert_tag *div_tag;
998
999 div_tag = (struct divert_tag *)(p+1);
1000 args->fwai_divert_rule = div_tag->cookie;
1001 delete = TRUE;
1002 }
1003#endif
1004
1005 if (p->m_tag_type == KERNEL_TAG_TYPE_IPFORWARD) {
1006 struct ip_fwd_tag *ipfwd_tag;
1007
1008 ipfwd_tag = (struct ip_fwd_tag *)(p+1);
1009 args->fwai_next_hop = ipfwd_tag->next_hop;
1010 delete = TRUE;
1011 }
1012
1013 if (delete) {
1014 copy = p;
1015 p = m_tag_next(m, p);
1016 m_tag_delete(m, copy);
1017 } else {
1018 p = m_tag_next(m, p);
1019 }
1020 } else {
1021 p = m_tag_next(m, p);
1022 }
1023 }
1024
1025#if DIAGNOSTIC
1026 if (m == NULL || !(m->m_flags & M_PKTHDR))
1027 panic("ip_input no HDR");
1028#endif
1029
1030#if DUMMYNET
1031 if (args->fwai_ipfw_rule || args->fwai_pf_rule) {
1032 /* dummynet already filtered us */
1033 ip = mtod(m, struct ip *);
1034 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1035 inject_filter_ref = ipf_get_inject_filter(m);
1036#if IPFIREWALL
1037 if (args->fwai_ipfw_rule)
1038 goto iphack;
1039#endif /* IPFIREWALL */
1040 if (args->fwai_pf_rule)
1041 goto check_with_pf;
1042 }
1043#endif /* DUMMYNET */
1044ipfw_tags_done:
1045#endif /* IPFIREWALL || DUMMYNET */
1046
1047 /*
1048 * No need to process packet twice if we've already seen it.
1049 */
1050 if (!SLIST_EMPTY(&m->m_pkthdr.tags))
1051 inject_filter_ref = ipf_get_inject_filter(m);
1052 if (inject_filter_ref != NULL) {
1053 ip = mtod(m, struct ip *);
1054 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1055
1056 DTRACE_IP6(receive, struct mbuf *, m, struct inpcb *, NULL,
1057 struct ip *, ip, struct ifnet *, inifp,
1058 struct ip *, ip, struct ip6_hdr *, NULL);
1059
1060 ip->ip_len = ntohs(ip->ip_len) - hlen;
1061 ip->ip_off = ntohs(ip->ip_off);
1062 ip_proto_dispatch_in(m, hlen, ip->ip_p, inject_filter_ref);
1063 return (IPINPUT_DONE);
1064 }
1065
1066 if (m->m_pkthdr.len < sizeof (struct ip)) {
1067 OSAddAtomic(1, &ipstat.ips_total);
1068 OSAddAtomic(1, &ipstat.ips_tooshort);
1069 m_freem(m);
1070 return (IPINPUT_FREED);
1071 }
1072
1073 if (m->m_len < sizeof (struct ip) &&
1074 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
1075 OSAddAtomic(1, &ipstat.ips_total);
1076 OSAddAtomic(1, &ipstat.ips_toosmall);
1077 return (IPINPUT_FREED);
1078 }
1079
1080 ip = mtod(m, struct ip *);
1081 *modm = m;
1082
1083 KERNEL_DEBUG(DBG_LAYER_BEG, ip->ip_dst.s_addr, ip->ip_src.s_addr,
1084 ip->ip_p, ip->ip_off, ip->ip_len);
1085
1086 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
1087 OSAddAtomic(1, &ipstat.ips_total);
1088 OSAddAtomic(1, &ipstat.ips_badvers);
1089 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1090 m_freem(m);
1091 return (IPINPUT_FREED);
1092 }
1093
1094 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1095 if (hlen < sizeof (struct ip)) {
1096 OSAddAtomic(1, &ipstat.ips_total);
1097 OSAddAtomic(1, &ipstat.ips_badhlen);
1098 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1099 m_freem(m);
1100 return (IPINPUT_FREED);
1101 }
1102
1103 if (hlen > m->m_len) {
1104 if ((m = m_pullup(m, hlen)) == NULL) {
1105 OSAddAtomic(1, &ipstat.ips_total);
1106 OSAddAtomic(1, &ipstat.ips_badhlen);
1107 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1108 return (IPINPUT_FREED);
1109 }
1110 ip = mtod(m, struct ip *);
1111 *modm = m;
1112 }
1113
1114 /* 127/8 must not appear on wire - RFC1122 */
1115 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
1116 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
1117 /*
1118 * Allow for the following exceptions:
1119 *
1120 * 1. If the packet was sent to loopback (i.e. rcvif
1121 * would have been set earlier at output time.)
1122 *
1123 * 2. If the packet was sent out on loopback from a local
1124 * source address which belongs to a non-loopback
1125 * interface (i.e. rcvif may not necessarily be a
1126 * loopback interface, hence the test for PKTF_LOOP.)
1127 * Unlike IPv6, there is no interface scope ID, and
1128 * therefore we don't care so much about PKTF_IFINFO.
1129 */
1130 if (!(inifp->if_flags & IFF_LOOPBACK) &&
1131 !(m->m_pkthdr.pkt_flags & PKTF_LOOP)) {
1132 OSAddAtomic(1, &ipstat.ips_total);
1133 OSAddAtomic(1, &ipstat.ips_badaddr);
1134 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1135 m_freem(m);
1136 return (IPINPUT_FREED);
1137 }
1138 }
1139
1140 /* IPv4 Link-Local Addresses as defined in RFC3927 */
1141 if ((IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
1142 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr)))) {
1143 ip_linklocal_stat.iplls_in_total++;
1144 if (ip->ip_ttl != MAXTTL) {
1145 OSAddAtomic(1, &ip_linklocal_stat.iplls_in_badttl);
1146 /* Silently drop link local traffic with bad TTL */
1147 if (!ip_linklocal_in_allowbadttl) {
1148 OSAddAtomic(1, &ipstat.ips_total);
1149 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1150 m_freem(m);
1151 return (IPINPUT_FREED);
1152 }
1153 }
1154 }
1155
1156 if (ip_cksum(m, hlen)) {
1157 OSAddAtomic(1, &ipstat.ips_total);
1158 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1159 m_freem(m);
1160 return (IPINPUT_FREED);
1161 }
1162
1163 DTRACE_IP6(receive, struct mbuf *, m, struct inpcb *, NULL,
1164 struct ip *, ip, struct ifnet *, inifp,
1165 struct ip *, ip, struct ip6_hdr *, NULL);
1166
1167 /*
1168 * Convert fields to host representation.
1169 */
1170#if BYTE_ORDER != BIG_ENDIAN
1171 NTOHS(ip->ip_len);
1172#endif
1173
1174 if (ip->ip_len < hlen) {
1175 OSAddAtomic(1, &ipstat.ips_total);
1176 OSAddAtomic(1, &ipstat.ips_badlen);
1177 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1178 m_freem(m);
1179 return (IPINPUT_FREED);
1180 }
1181
1182#if BYTE_ORDER != BIG_ENDIAN
1183 NTOHS(ip->ip_off);
1184#endif
1185
1186 /*
1187 * Check that the amount of data in the buffers
1188 * is as at least much as the IP header would have us expect.
1189 * Trim mbufs if longer than we expect.
1190 * Drop packet if shorter than we expect.
1191 */
1192 if (m->m_pkthdr.len < ip->ip_len) {
1193 OSAddAtomic(1, &ipstat.ips_total);
1194 OSAddAtomic(1, &ipstat.ips_tooshort);
1195 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1196 m_freem(m);
1197 return (IPINPUT_FREED);
1198 }
1199
1200 if (m->m_pkthdr.len > ip->ip_len) {
1201 ip_input_adjust(m, ip, inifp);
1202 }
1203
1204 /* for consistency */
1205 m->m_pkthdr.pkt_proto = ip->ip_p;
1206
1207 /* for netstat route statistics */
1208 src_ip = ip->ip_src;
1209 len = m->m_pkthdr.len;
1210
1211#if DUMMYNET
1212check_with_pf:
1213#endif
1214#if PF
1215 /* Invoke inbound packet filter */
1216 if (PF_IS_ENABLED) {
1217 int error;
1218 ip_input_cpout_args(args, &args1, &init);
1219 ip = mtod(m, struct ip *);
1220 src_ip = ip->ip_src;
1221
1222#if DUMMYNET
1223 error = pf_af_hook(inifp, NULL, &m, AF_INET, TRUE, &args1);
1224#else
1225 error = pf_af_hook(inifp, NULL, &m, AF_INET, TRUE, NULL);
1226#endif /* DUMMYNET */
1227 if (error != 0 || m == NULL) {
1228 if (m != NULL) {
1229 panic("%s: unexpected packet %p\n",
1230 __func__, m);
1231 /* NOTREACHED */
1232 }
1233 /* Already freed by callee */
1234 ip_input_update_nstat(inifp, src_ip, 1, len);
1235 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1236 OSAddAtomic(1, &ipstat.ips_total);
1237 return (IPINPUT_FREED);
1238 }
1239 ip = mtod(m, struct ip *);
1240 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1241 *modm = m;
1242 ip_input_cpin_args(&args1, args);
1243 }
1244#endif /* PF */
1245
1246#if IPSEC
1247 if (ipsec_bypass == 0 && ipsec_gethist(m, NULL)) {
1248 retval = IPINPUT_DONTCHAIN; /* XXX scope for chaining here? */
1249 goto pass;
1250 }
1251#endif
1252
1253#if IPFIREWALL
1254#if DUMMYNET
1255iphack:
1256#endif /* DUMMYNET */
1257 /*
1258 * Check if we want to allow this packet to be processed.
1259 * Consider it to be bad if not.
1260 */
1261 if (fw_enable && IPFW_LOADED) {
1262#if IPFIREWALL_FORWARD
1263 /*
1264 * If we've been forwarded from the output side, then
1265 * skip the firewall a second time
1266 */
1267 if (args->fwai_next_hop) {
1268 *ours = 1;
1269 return (IPINPUT_DONTCHAIN);
1270 }
1271#endif /* IPFIREWALL_FORWARD */
1272 ip_input_cpout_args(args, &args1, &init);
1273 args1.fwa_m = m;
1274
1275 i = ip_fw_chk_ptr(&args1);
1276 m = args1.fwa_m;
1277
1278 if ((i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
1279 if (m)
1280 m_freem(m);
1281 ip_input_update_nstat(inifp, src_ip, 1, len);
1282 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1283 OSAddAtomic(1, &ipstat.ips_total);
1284 return (IPINPUT_FREED);
1285 }
1286 ip = mtod(m, struct ip *); /* just in case m changed */
1287 *modm = m;
1288 ip_input_cpin_args(&args1, args);
1289
1290 if (i == 0 && args->fwai_next_hop == NULL) { /* common case */
1291 goto pass;
1292 }
1293#if DUMMYNET
1294 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) {
1295 /* Send packet to the appropriate pipe */
1296 ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args1,
1297 DN_CLIENT_IPFW);
1298 ip_input_update_nstat(inifp, src_ip, 1, len);
1299 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1300 OSAddAtomic(1, &ipstat.ips_total);
1301 return (IPINPUT_FREED);
1302 }
1303#endif /* DUMMYNET */
1304#if IPDIVERT
1305 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
1306 /* Divert or tee packet */
1307 *div_info = i;
1308 *ours = 1;
1309 return (IPINPUT_DONTCHAIN);
1310 }
1311#endif
1312#if IPFIREWALL_FORWARD
1313 if (i == 0 && args->fwai_next_hop != NULL) {
1314 retval = IPINPUT_DONTCHAIN;
1315 goto pass;
1316 }
1317#endif
1318 /*
1319 * if we get here, the packet must be dropped
1320 */
1321 ip_input_update_nstat(inifp, src_ip, 1, len);
1322 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1323 m_freem(m);
1324 OSAddAtomic(1, &ipstat.ips_total);
1325 return (IPINPUT_FREED);
1326 }
1327#endif /* IPFIREWALL */
1328#if IPSEC | IPFIREWALL
1329pass:
1330#endif
1331 /*
1332 * Process options and, if not destined for us,
1333 * ship it on. ip_dooptions returns 1 when an
1334 * error was detected (causing an icmp message
1335 * to be sent and the original packet to be freed).
1336 */
1337 ip_nhops = 0; /* for source routed packets */
1338#if IPFIREWALL
1339 if (hlen > sizeof (struct ip) &&
1340 ip_dooptions(m, 0, args->fwai_next_hop)) {
1341#else /* !IPFIREWALL */
1342 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, NULL)) {
1343#endif /* !IPFIREWALL */
1344 ip_input_update_nstat(inifp, src_ip, 1, len);
1345 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1346 OSAddAtomic(1, &ipstat.ips_total);
1347 return (IPINPUT_FREED);
1348 }
1349
1350 /*
1351 * Don't chain fragmented packets as the process of determining
1352 * if it is our fragment or someone else's plus the complexity of
1353 * divert and fw args makes it harder to do chaining.
1354 */
1355 if (ip->ip_off & ~(IP_DF | IP_RF))
1356 return (IPINPUT_DONTCHAIN);
1357
1358 /* Allow DHCP/BootP responses through */
1359 if ((inifp->if_eflags & IFEF_AUTOCONFIGURING) &&
1360 hlen == sizeof (struct ip) && ip->ip_p == IPPROTO_UDP) {
1361 struct udpiphdr *ui;
1362
1363 if (m->m_len < sizeof (struct udpiphdr) &&
1364 (m = m_pullup(m, sizeof (struct udpiphdr))) == NULL) {
1365 OSAddAtomic(1, &udpstat.udps_hdrops);
1366 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1367 OSAddAtomic(1, &ipstat.ips_total);
1368 return (IPINPUT_FREED);
1369 }
1370 *modm = m;
1371 ui = mtod(m, struct udpiphdr *);
1372 if (ntohs(ui->ui_dport) == IPPORT_BOOTPC) {
1373 ip_setdstifaddr_info(m, inifp->if_index, NULL);
1374 return (IPINPUT_DONTCHAIN);
1375 }
1376 }
1377
1378 /* Avoid chaining raw sockets as ipsec checks occur later for them */
1379 if (ip_protox[ip->ip_p]->pr_flags & PR_LASTHDR)
1380 return (IPINPUT_DONTCHAIN);
1381
1382 return (retval);
1383#if !defined(__i386__) && !defined(__x86_64__)
1384bad:
1385 m_freem(m);
1386 return (IPINPUT_FREED);
1387#endif
1388}
1389
1390static void
1391ip_input_second_pass(struct mbuf *m, struct ifnet *inifp, u_int32_t div_info,
1392 int npkts_in_chain, int bytes_in_chain, struct ip_fw_in_args *args, int ours)
1393{
1394 unsigned int checkif;
1395 struct mbuf *tmp_mbuf = NULL;
1396 struct in_ifaddr *ia = NULL;
1397 struct in_addr pkt_dst;
1398 unsigned int hlen;
1399
1400#if !IPFIREWALL
1401#pragma unused (args)
1402#endif
1403
1404#if !IPDIVERT
1405#pragma unused (div_info)
1406#endif
1407
1408 struct ip *ip = mtod(m, struct ip *);
1409 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1410
1411 OSAddAtomic(npkts_in_chain, &ipstat.ips_total);
1412
1413 /*
1414 * Naively assume we can attribute inbound data to the route we would
1415 * use to send to this destination. Asymmetric routing breaks this
1416 * assumption, but it still allows us to account for traffic from
1417 * a remote node in the routing table.
1418 * this has a very significant performance impact so we bypass
1419 * if nstat_collect is disabled. We may also bypass if the
1420 * protocol is tcp in the future because tcp will have a route that
1421 * we can use to attribute the data to. That does mean we would not
1422 * account for forwarded tcp traffic.
1423 */
1424 ip_input_update_nstat(inifp, ip->ip_src, npkts_in_chain,
1425 bytes_in_chain);
1426
1427 if (ours)
1428 goto ours;
1429
1430 /*
1431 * Check our list of addresses, to see if the packet is for us.
1432 * If we don't have any addresses, assume any unicast packet
1433 * we receive might be for us (and let the upper layers deal
1434 * with it).
1435 */
1436 tmp_mbuf = m;
1437 if (TAILQ_EMPTY(&in_ifaddrhead)) {
1438 while (tmp_mbuf) {
1439 if (!(tmp_mbuf->m_flags & (M_MCAST|M_BCAST))) {
1440 ip_setdstifaddr_info(tmp_mbuf, inifp->if_index,
1441 NULL);
1442 }
1443 tmp_mbuf = mbuf_nextpkt(tmp_mbuf);
1444 }
1445 goto ours;
1446 }
1447 /*
1448 * Cache the destination address of the packet; this may be
1449 * changed by use of 'ipfw fwd'.
1450 */
1451#if IPFIREWALL
1452 pkt_dst = args->fwai_next_hop == NULL ?
1453 ip->ip_dst : args->fwai_next_hop->sin_addr;
1454#else /* !IPFIREWALL */
1455 pkt_dst = ip->ip_dst;
1456#endif /* !IPFIREWALL */
1457
1458 /*
1459 * Enable a consistency check between the destination address
1460 * and the arrival interface for a unicast packet (the RFC 1122
1461 * strong ES model) if IP forwarding is disabled and the packet
1462 * is not locally generated and the packet is not subject to
1463 * 'ipfw fwd'.
1464 *
1465 * XXX - Checking also should be disabled if the destination
1466 * address is ipnat'ed to a different interface.
1467 *
1468 * XXX - Checking is incompatible with IP aliases added
1469 * to the loopback interface instead of the interface where
1470 * the packets are received.
1471 */
1472 checkif = ip_checkinterface && (ipforwarding == 0) &&
1473 !(inifp->if_flags & IFF_LOOPBACK) &&
1474 !(m->m_pkthdr.pkt_flags & PKTF_LOOP)
1475#if IPFIREWALL
1476 && (args->fwai_next_hop == NULL);
1477#else /* !IPFIREWALL */
1478 ;
1479#endif /* !IPFIREWALL */
1480
1481 /*
1482 * Check for exact addresses in the hash bucket.
1483 */
1484 lck_rw_lock_shared(in_ifaddr_rwlock);
1485 TAILQ_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
1486 /*
1487 * If the address matches, verify that the packet
1488 * arrived via the correct interface if checking is
1489 * enabled.
1490 */
1491 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
1492 (!checkif || ia->ia_ifp == inifp)) {
1493 ip_input_setdst_chain(m, 0, ia);
1494 lck_rw_done(in_ifaddr_rwlock);
1495 goto ours;
1496 }
1497 }
1498 lck_rw_done(in_ifaddr_rwlock);
1499
1500 /*
1501 * Check for broadcast addresses.
1502 *
1503 * Only accept broadcast packets that arrive via the matching
1504 * interface. Reception of forwarded directed broadcasts would be
1505 * handled via ip_forward() and ether_frameout() with the loopback
1506 * into the stack for SIMPLEX interfaces handled by ether_frameout().
1507 */
1508 if (inifp->if_flags & IFF_BROADCAST) {
1509 struct ifaddr *ifa;
1510
1511 ifnet_lock_shared(inifp);
1512 TAILQ_FOREACH(ifa, &inifp->if_addrhead, ifa_link) {
1513 if (ifa->ifa_addr->sa_family != AF_INET) {
1514 continue;
1515 }
1516 ia = ifatoia(ifa);
1517 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
1518 pkt_dst.s_addr || ia->ia_netbroadcast.s_addr ==
1519 pkt_dst.s_addr) {
1520 ip_input_setdst_chain(m, 0, ia);
1521 ifnet_lock_done(inifp);
1522 goto ours;
1523 }
1524 }
1525 ifnet_lock_done(inifp);
1526 }
1527
1528 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
1529 struct in_multi *inm;
1530 /*
1531 * See if we belong to the destination multicast group on the
1532 * arrival interface.
1533 */
1534 in_multihead_lock_shared();
1535 IN_LOOKUP_MULTI(&ip->ip_dst, inifp, inm);
1536 in_multihead_lock_done();
1537 if (inm == NULL) {
1538 OSAddAtomic(npkts_in_chain, &ipstat.ips_notmember);
1539 m_freem_list(m);
1540 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1541 return;
1542 }
1543 ip_input_setdst_chain(m, inifp->if_index, NULL);
1544 INM_REMREF(inm);
1545 goto ours;
1546 }
1547
1548 if (ip->ip_dst.s_addr == (u_int32_t)INADDR_BROADCAST ||
1549 ip->ip_dst.s_addr == INADDR_ANY) {
1550 ip_input_setdst_chain(m, inifp->if_index, NULL);
1551 goto ours;
1552 }
1553
1554 if (ip->ip_p == IPPROTO_UDP) {
1555 struct udpiphdr *ui;
1556 ui = mtod(m, struct udpiphdr *);
1557 if (ntohs(ui->ui_dport) == IPPORT_BOOTPC) {
1558 goto ours;
1559 }
1560 }
1561
1562 tmp_mbuf = m;
1563 struct mbuf *nxt_mbuf = NULL;
1564 while (tmp_mbuf) {
1565 nxt_mbuf = mbuf_nextpkt(tmp_mbuf);
1566 /*
1567 * Not for us; forward if possible and desirable.
1568 */
1569 mbuf_setnextpkt(tmp_mbuf, NULL);
1570 if (ipforwarding == 0) {
1571 OSAddAtomic(1, &ipstat.ips_cantforward);
1572 m_freem(tmp_mbuf);
1573 } else {
1574#if IPFIREWALL
1575 ip_forward(tmp_mbuf, 0, args->fwai_next_hop);
1576#else
1577 ip_forward(tmp_mbuf, 0, NULL);
1578#endif
1579 }
1580 tmp_mbuf = nxt_mbuf;
1581 }
1582 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1583 return;
1584ours:
1585 /*
1586 * If offset or IP_MF are set, must reassemble.
1587 */
1588 if (ip->ip_off & ~(IP_DF | IP_RF)) {
1589 VERIFY(npkts_in_chain == 1);
1590 /*
1591 * ip_reass() will return a different mbuf, and update
1592 * the divert info in div_info and args->fwai_divert_rule.
1593 */
1594#if IPDIVERT
1595 m = ip_reass(m, (u_int16_t *)&div_info, &args->fwai_divert_rule);
1596#else
1597 m = ip_reass(m);
1598#endif
1599 if (m == NULL)
1600 return;
1601 ip = mtod(m, struct ip *);
1602 /* Get the header length of the reassembled packet */
1603 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1604#if IPDIVERT
1605 /* Restore original checksum before diverting packet */
1606 if (div_info != 0) {
1607 VERIFY(npkts_in_chain == 1);
1608#if BYTE_ORDER != BIG_ENDIAN
1609 HTONS(ip->ip_len);
1610 HTONS(ip->ip_off);
1611#endif
1612 ip->ip_sum = 0;
1613 ip->ip_sum = ip_cksum_hdr_in(m, hlen);
1614#if BYTE_ORDER != BIG_ENDIAN
1615 NTOHS(ip->ip_off);
1616 NTOHS(ip->ip_len);
1617#endif
1618 }
1619#endif
1620 }
1621
1622 /*
1623 * Further protocols expect the packet length to be w/o the
1624 * IP header.
1625 */
1626 ip->ip_len -= hlen;
1627
1628#if IPDIVERT
1629 /*
1630 * Divert or tee packet to the divert protocol if required.
1631 *
1632 * If div_info is zero then cookie should be too, so we shouldn't
1633 * need to clear them here. Assume divert_packet() does so also.
1634 */
1635 if (div_info != 0) {
1636 struct mbuf *clone = NULL;
1637 VERIFY(npkts_in_chain == 1);
1638
1639 /* Clone packet if we're doing a 'tee' */
1640 if (div_info & IP_FW_PORT_TEE_FLAG)
1641 clone = m_dup(m, M_DONTWAIT);
1642
1643 /* Restore packet header fields to original values */
1644 ip->ip_len += hlen;
1645
1646#if BYTE_ORDER != BIG_ENDIAN
1647 HTONS(ip->ip_len);
1648 HTONS(ip->ip_off);
1649#endif
1650 /* Deliver packet to divert input routine */
1651 OSAddAtomic(1, &ipstat.ips_delivered);
1652 divert_packet(m, 1, div_info & 0xffff, args->fwai_divert_rule);
1653
1654 /* If 'tee', continue with original packet */
1655 if (clone == NULL) {
1656 return;
1657 }
1658 m = clone;
1659 ip = mtod(m, struct ip *);
1660 }
1661#endif
1662
1663#if IPSEC
1664 /*
1665 * enforce IPsec policy checking if we are seeing last header.
1666 * note that we do not visit this with protocols with pcb layer
1667 * code - like udp/tcp/raw ip.
1668 */
1669 if (ipsec_bypass == 0 && (ip_protox[ip->ip_p]->pr_flags & PR_LASTHDR)) {
1670 VERIFY(npkts_in_chain == 1);
1671 if (ipsec4_in_reject(m, NULL)) {
1672 IPSEC_STAT_INCREMENT(ipsecstat.in_polvio);
1673 goto bad;
1674 }
1675 }
1676#endif /* IPSEC */
1677
1678 /*
1679 * Switch out to protocol's input routine.
1680 */
1681 OSAddAtomic(npkts_in_chain, &ipstat.ips_delivered);
1682
1683#if IPFIREWALL
1684 if (args->fwai_next_hop && ip->ip_p == IPPROTO_TCP) {
1685 /* TCP needs IPFORWARD info if available */
1686 struct m_tag *fwd_tag;
1687 struct ip_fwd_tag *ipfwd_tag;
1688
1689 VERIFY(npkts_in_chain == 1);
1690 fwd_tag = m_tag_create(KERNEL_MODULE_TAG_ID,
1691 KERNEL_TAG_TYPE_IPFORWARD, sizeof (*ipfwd_tag),
1692 M_NOWAIT, m);
1693 if (fwd_tag == NULL)
1694 goto bad;
1695
1696 ipfwd_tag = (struct ip_fwd_tag *)(fwd_tag+1);
1697 ipfwd_tag->next_hop = args->fwai_next_hop;
1698
1699 m_tag_prepend(m, fwd_tag);
1700
1701 KERNEL_DEBUG(DBG_LAYER_END, ip->ip_dst.s_addr,
1702 ip->ip_src.s_addr, ip->ip_p, ip->ip_off, ip->ip_len);
1703
1704 /* TCP deals with its own locking */
1705 ip_proto_dispatch_in(m, hlen, ip->ip_p, 0);
1706 } else {
1707 KERNEL_DEBUG(DBG_LAYER_END, ip->ip_dst.s_addr,
1708 ip->ip_src.s_addr, ip->ip_p, ip->ip_off, ip->ip_len);
1709
1710 ip_input_dispatch_chain(m);
1711
1712 }
1713#else /* !IPFIREWALL */
1714 ip_input_dispatch_chain(m);
1715
1716#endif /* !IPFIREWALL */
1717 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1718 return;
1719bad:
1720 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
1721 m_freem(m);
1722}
1723
1724void
1725ip_input_process_list(struct mbuf *packet_list)
1726{
1727 pktchain_elm_t pktchain_tbl[PKTTBL_SZ];
1728
1729 struct mbuf *packet = NULL;
1730 struct mbuf *modm = NULL; /* modified mbuf */
1731 int retval = 0;
1732 u_int32_t div_info = 0;
1733 int ours = 0;
1734#if (DEBUG || DEVELOPMENT)
1735 struct timeval start_tv;
1736#endif /* (DEBUG || DEVELOPMENT) */
1737 int num_pkts = 0;
1738 int chain = 0;
1739 struct ip_fw_in_args args;
1740
1741 if (ip_chaining == 0) {
1742 struct mbuf *m = packet_list;
1743#if (DEBUG || DEVELOPMENT)
1744 if (ip_input_measure)
1745 net_perf_start_time(&net_perf, &start_tv);
1746#endif /* (DEBUG || DEVELOPMENT) */
1747
1748 while (m) {
1749 packet_list = mbuf_nextpkt(m);
1750 mbuf_setnextpkt(m, NULL);
1751 ip_input(m);
1752 m = packet_list;
1753 num_pkts++;
1754 }
1755#if (DEBUG || DEVELOPMENT)
1756 if (ip_input_measure)
1757 net_perf_measure_time(&net_perf, &start_tv, num_pkts);
1758#endif /* (DEBUG || DEVELOPMENT) */
1759 return;
1760 }
1761#if (DEBUG || DEVELOPMENT)
1762 if (ip_input_measure)
1763 net_perf_start_time(&net_perf, &start_tv);
1764#endif /* (DEBUG || DEVELOPMENT) */
1765
1766 bzero(&pktchain_tbl, sizeof(pktchain_tbl));
1767restart_list_process:
1768 chain = 0;
1769 for (packet = packet_list; packet; packet = packet_list) {
1770 packet_list = mbuf_nextpkt(packet);
1771 mbuf_setnextpkt(packet, NULL);
1772
1773 num_pkts++;
1774 modm = NULL;
1775 div_info = 0;
1776 bzero(&args, sizeof (args));
1777
1778 retval = ip_input_first_pass(packet, &div_info, &args,
1779 &ours, &modm);
1780
1781 if (retval == IPINPUT_DOCHAIN) {
1782 if (modm)
1783 packet = modm;
1784 packet = ip_chain_insert(packet, &pktchain_tbl[0]);
1785 if (packet == NULL) {
1786 ipstat.ips_rxc_chained++;
1787 chain++;
1788 if (chain > ip_chainsz)
1789 break;
1790 } else {
1791 ipstat.ips_rxc_collisions++;
1792 break;
1793 }
1794 } else if (retval == IPINPUT_DONTCHAIN) {
1795 /* in order to preserve order, exit from chaining */
1796 if (modm)
1797 packet = modm;
1798 ipstat.ips_rxc_notchain++;
1799 break;
1800 } else {
1801 /* packet was freed or delivered, do nothing. */
1802 }
1803 }
1804
1805 /* do second pass here for pktchain_tbl */
1806 if (chain)
1807 ip_input_second_pass_loop_tbl(&pktchain_tbl[0], &args);
1808
1809 if (packet) {
1810 /*
1811 * equivalent update in chaining case if performed in
1812 * ip_input_second_pass_loop_tbl().
1813 */
1814#if (DEBUG || DEVELOPMENT)
1815 if (ip_input_measure)
1816 net_perf_histogram(&net_perf, 1);
1817#endif /* (DEBUG || DEVELOPMENT) */
1818 ip_input_second_pass(packet, packet->m_pkthdr.rcvif, div_info,
1819 1, packet->m_pkthdr.len, &args, ours);
1820 }
1821
1822 if (packet_list)
1823 goto restart_list_process;
1824
1825#if (DEBUG || DEVELOPMENT)
1826 if (ip_input_measure)
1827 net_perf_measure_time(&net_perf, &start_tv, num_pkts);
1828#endif /* (DEBUG || DEVELOPMENT) */
1829}
1830/*
1831 * Ip input routine. Checksum and byte swap header. If fragmented
1832 * try to reassemble. Process options. Pass to next level.
1833 */
1834void
1835ip_input(struct mbuf *m)
1836{
1837 struct ip *ip;
1838 struct in_ifaddr *ia = NULL;
1839 unsigned int hlen, checkif;
1840 u_short sum = 0;
1841 struct in_addr pkt_dst;
1842#if IPFIREWALL
1843 int i;
1844 u_int32_t div_info = 0; /* packet divert/tee info */
1845#endif
1846#if IPFIREWALL || DUMMYNET
1847 struct ip_fw_args args;
1848 struct m_tag *tag;
1849#endif
1850 ipfilter_t inject_filter_ref = NULL;
1851 struct ifnet *inifp;
1852
1853 /* Check if the mbuf is still valid after interface filter processing */
1854 MBUF_INPUT_CHECK(m, m->m_pkthdr.rcvif);
1855 inifp = m->m_pkthdr.rcvif;
1856 VERIFY(inifp != NULL);
1857
1858 ipstat.ips_rxc_notlist++;
1859
1860 /* Perform IP header alignment fixup, if needed */
1861 IP_HDR_ALIGNMENT_FIXUP(m, inifp, goto bad);
1862
1863 m->m_pkthdr.pkt_flags &= ~PKTF_FORWARDED;
1864
1865#if IPFIREWALL || DUMMYNET
1866 bzero(&args, sizeof (struct ip_fw_args));
1867
1868 /*
1869 * Don't bother searching for tag(s) if there's none.
1870 */
1871 if (SLIST_EMPTY(&m->m_pkthdr.tags))
1872 goto ipfw_tags_done;
1873
1874 /* Grab info from mtags prepended to the chain */
1875#if DUMMYNET
1876 if ((tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
1877 KERNEL_TAG_TYPE_DUMMYNET, NULL)) != NULL) {
1878 struct dn_pkt_tag *dn_tag;
1879
1880 dn_tag = (struct dn_pkt_tag *)(tag+1);
1881 args.fwa_ipfw_rule = dn_tag->dn_ipfw_rule;
1882 args.fwa_pf_rule = dn_tag->dn_pf_rule;
1883
1884 m_tag_delete(m, tag);
1885 }
1886#endif /* DUMMYNET */
1887
1888#if IPDIVERT
1889 if ((tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
1890 KERNEL_TAG_TYPE_DIVERT, NULL)) != NULL) {
1891 struct divert_tag *div_tag;
1892
1893 div_tag = (struct divert_tag *)(tag+1);
1894 args.fwa_divert_rule = div_tag->cookie;
1895
1896 m_tag_delete(m, tag);
1897 }
1898#endif
1899
1900 if ((tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
1901 KERNEL_TAG_TYPE_IPFORWARD, NULL)) != NULL) {
1902 struct ip_fwd_tag *ipfwd_tag;
1903
1904 ipfwd_tag = (struct ip_fwd_tag *)(tag+1);
1905 args.fwa_next_hop = ipfwd_tag->next_hop;
1906
1907 m_tag_delete(m, tag);
1908 }
1909
1910#if DIAGNOSTIC
1911 if (m == NULL || !(m->m_flags & M_PKTHDR))
1912 panic("ip_input no HDR");
1913#endif
1914
1915#if DUMMYNET
1916 if (args.fwa_ipfw_rule || args.fwa_pf_rule) {
1917 /* dummynet already filtered us */
1918 ip = mtod(m, struct ip *);
1919 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1920 inject_filter_ref = ipf_get_inject_filter(m);
1921#if IPFIREWALL
1922 if (args.fwa_ipfw_rule)
1923 goto iphack;
1924#endif /* IPFIREWALL */
1925 if (args.fwa_pf_rule)
1926 goto check_with_pf;
1927 }
1928#endif /* DUMMYNET */
1929ipfw_tags_done:
1930#endif /* IPFIREWALL || DUMMYNET */
1931
1932 /*
1933 * No need to process packet twice if we've already seen it.
1934 */
1935 if (!SLIST_EMPTY(&m->m_pkthdr.tags))
1936 inject_filter_ref = ipf_get_inject_filter(m);
1937 if (inject_filter_ref != NULL) {
1938 ip = mtod(m, struct ip *);
1939 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1940
1941 DTRACE_IP6(receive, struct mbuf *, m, struct inpcb *, NULL,
1942 struct ip *, ip, struct ifnet *, inifp,
1943 struct ip *, ip, struct ip6_hdr *, NULL);
1944
1945 ip->ip_len = ntohs(ip->ip_len) - hlen;
1946 ip->ip_off = ntohs(ip->ip_off);
1947 ip_proto_dispatch_in(m, hlen, ip->ip_p, inject_filter_ref);
1948 return;
1949 }
1950
1951 OSAddAtomic(1, &ipstat.ips_total);
1952 if (m->m_pkthdr.len < sizeof (struct ip))
1953 goto tooshort;
1954
1955 if (m->m_len < sizeof (struct ip) &&
1956 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
1957 OSAddAtomic(1, &ipstat.ips_toosmall);
1958 return;
1959 }
1960 ip = mtod(m, struct ip *);
1961
1962 KERNEL_DEBUG(DBG_LAYER_BEG, ip->ip_dst.s_addr, ip->ip_src.s_addr,
1963 ip->ip_p, ip->ip_off, ip->ip_len);
1964
1965 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
1966 OSAddAtomic(1, &ipstat.ips_badvers);
1967 goto bad;
1968 }
1969
1970 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1971 if (hlen < sizeof (struct ip)) { /* minimum header length */
1972 OSAddAtomic(1, &ipstat.ips_badhlen);
1973 goto bad;
1974 }
1975 if (hlen > m->m_len) {
1976 if ((m = m_pullup(m, hlen)) == NULL) {
1977 OSAddAtomic(1, &ipstat.ips_badhlen);
1978 return;
1979 }
1980 ip = mtod(m, struct ip *);
1981 }
1982
1983 /* 127/8 must not appear on wire - RFC1122 */
1984 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
1985 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
1986 /*
1987 * Allow for the following exceptions:
1988 *
1989 * 1. If the packet was sent to loopback (i.e. rcvif
1990 * would have been set earlier at output time.)
1991 *
1992 * 2. If the packet was sent out on loopback from a local
1993 * source address which belongs to a non-loopback
1994 * interface (i.e. rcvif may not necessarily be a
1995 * loopback interface, hence the test for PKTF_LOOP.)
1996 * Unlike IPv6, there is no interface scope ID, and
1997 * therefore we don't care so much about PKTF_IFINFO.
1998 */
1999 if (!(inifp->if_flags & IFF_LOOPBACK) &&
2000 !(m->m_pkthdr.pkt_flags & PKTF_LOOP)) {
2001 OSAddAtomic(1, &ipstat.ips_badaddr);
2002 goto bad;
2003 }
2004 }
2005
2006 /* IPv4 Link-Local Addresses as defined in RFC3927 */
2007 if ((IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
2008 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr)))) {
2009 ip_linklocal_stat.iplls_in_total++;
2010 if (ip->ip_ttl != MAXTTL) {
2011 OSAddAtomic(1, &ip_linklocal_stat.iplls_in_badttl);
2012 /* Silently drop link local traffic with bad TTL */
2013 if (!ip_linklocal_in_allowbadttl)
2014 goto bad;
2015 }
2016 }
2017
2018 sum = ip_cksum(m, hlen);
2019 if (sum) {
2020 goto bad;
2021 }
2022
2023 DTRACE_IP6(receive, struct mbuf *, m, struct inpcb *, NULL,
2024 struct ip *, ip, struct ifnet *, inifp,
2025 struct ip *, ip, struct ip6_hdr *, NULL);
2026
2027 /*
2028 * Naively assume we can attribute inbound data to the route we would
2029 * use to send to this destination. Asymmetric routing breaks this
2030 * assumption, but it still allows us to account for traffic from
2031 * a remote node in the routing table.
2032 * this has a very significant performance impact so we bypass
2033 * if nstat_collect is disabled. We may also bypass if the
2034 * protocol is tcp in the future because tcp will have a route that
2035 * we can use to attribute the data to. That does mean we would not
2036 * account for forwarded tcp traffic.
2037 */
2038 if (nstat_collect) {
2039 struct rtentry *rt =
2040 ifnet_cached_rtlookup_inet(inifp, ip->ip_src);
2041 if (rt != NULL) {
2042 nstat_route_rx(rt, 1, m->m_pkthdr.len, 0);
2043 rtfree(rt);
2044 }
2045 }
2046
2047 /*
2048 * Convert fields to host representation.
2049 */
2050#if BYTE_ORDER != BIG_ENDIAN
2051 NTOHS(ip->ip_len);
2052#endif
2053
2054 if (ip->ip_len < hlen) {
2055 OSAddAtomic(1, &ipstat.ips_badlen);
2056 goto bad;
2057 }
2058
2059#if BYTE_ORDER != BIG_ENDIAN
2060 NTOHS(ip->ip_off);
2061#endif
2062 /*
2063 * Check that the amount of data in the buffers
2064 * is as at least much as the IP header would have us expect.
2065 * Trim mbufs if longer than we expect.
2066 * Drop packet if shorter than we expect.
2067 */
2068 if (m->m_pkthdr.len < ip->ip_len) {
2069tooshort:
2070 OSAddAtomic(1, &ipstat.ips_tooshort);
2071 goto bad;
2072 }
2073 if (m->m_pkthdr.len > ip->ip_len) {
2074 ip_input_adjust(m, ip, inifp);
2075 }
2076
2077 /* for consistency */
2078 m->m_pkthdr.pkt_proto = ip->ip_p;
2079
2080#if DUMMYNET
2081check_with_pf:
2082#endif
2083#if PF
2084 /* Invoke inbound packet filter */
2085 if (PF_IS_ENABLED) {
2086 int error;
2087#if DUMMYNET
2088 error = pf_af_hook(inifp, NULL, &m, AF_INET, TRUE, &args);
2089#else
2090 error = pf_af_hook(inifp, NULL, &m, AF_INET, TRUE, NULL);
2091#endif /* DUMMYNET */
2092 if (error != 0 || m == NULL) {
2093 if (m != NULL) {
2094 panic("%s: unexpected packet %p\n",
2095 __func__, m);
2096 /* NOTREACHED */
2097 }
2098 /* Already freed by callee */
2099 return;
2100 }
2101 ip = mtod(m, struct ip *);
2102 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
2103 }
2104#endif /* PF */
2105
2106#if IPSEC
2107 if (ipsec_bypass == 0 && ipsec_gethist(m, NULL))
2108 goto pass;
2109#endif
2110
2111#if IPFIREWALL
2112#if DUMMYNET
2113iphack:
2114#endif /* DUMMYNET */
2115 /*
2116 * Check if we want to allow this packet to be processed.
2117 * Consider it to be bad if not.
2118 */
2119 if (fw_enable && IPFW_LOADED) {
2120#if IPFIREWALL_FORWARD
2121 /*
2122 * If we've been forwarded from the output side, then
2123 * skip the firewall a second time
2124 */
2125 if (args.fwa_next_hop)
2126 goto ours;
2127#endif /* IPFIREWALL_FORWARD */
2128
2129 args.fwa_m = m;
2130
2131 i = ip_fw_chk_ptr(&args);
2132 m = args.fwa_m;
2133
2134 if ((i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
2135 if (m)
2136 m_freem(m);
2137 return;
2138 }
2139 ip = mtod(m, struct ip *); /* just in case m changed */
2140
2141 if (i == 0 && args.fwa_next_hop == NULL) { /* common case */
2142 goto pass;
2143 }
2144#if DUMMYNET
2145 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) {
2146 /* Send packet to the appropriate pipe */
2147 ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args,
2148 DN_CLIENT_IPFW);
2149 return;
2150 }
2151#endif /* DUMMYNET */
2152#if IPDIVERT
2153 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
2154 /* Divert or tee packet */
2155 div_info = i;
2156 goto ours;
2157 }
2158#endif
2159#if IPFIREWALL_FORWARD
2160 if (i == 0 && args.fwa_next_hop != NULL) {
2161 goto pass;
2162 }
2163#endif
2164 /*
2165 * if we get here, the packet must be dropped
2166 */
2167 m_freem(m);
2168 return;
2169 }
2170#endif /* IPFIREWALL */
2171#if IPSEC | IPFIREWALL
2172pass:
2173#endif
2174 /*
2175 * Process options and, if not destined for us,
2176 * ship it on. ip_dooptions returns 1 when an
2177 * error was detected (causing an icmp message
2178 * to be sent and the original packet to be freed).
2179 */
2180 ip_nhops = 0; /* for source routed packets */
2181#if IPFIREWALL
2182 if (hlen > sizeof (struct ip) &&
2183 ip_dooptions(m, 0, args.fwa_next_hop)) {
2184#else /* !IPFIREWALL */
2185 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, NULL)) {
2186#endif /* !IPFIREWALL */
2187 return;
2188 }
2189
2190 /*
2191 * Check our list of addresses, to see if the packet is for us.
2192 * If we don't have any addresses, assume any unicast packet
2193 * we receive might be for us (and let the upper layers deal
2194 * with it).
2195 */
2196 if (TAILQ_EMPTY(&in_ifaddrhead) && !(m->m_flags & (M_MCAST|M_BCAST))) {
2197 ip_setdstifaddr_info(m, inifp->if_index, NULL);
2198 goto ours;
2199 }
2200
2201 /*
2202 * Cache the destination address of the packet; this may be
2203 * changed by use of 'ipfw fwd'.
2204 */
2205#if IPFIREWALL
2206 pkt_dst = args.fwa_next_hop == NULL ?
2207 ip->ip_dst : args.fwa_next_hop->sin_addr;
2208#else /* !IPFIREWALL */
2209 pkt_dst = ip->ip_dst;
2210#endif /* !IPFIREWALL */
2211
2212 /*
2213 * Enable a consistency check between the destination address
2214 * and the arrival interface for a unicast packet (the RFC 1122
2215 * strong ES model) if IP forwarding is disabled and the packet
2216 * is not locally generated and the packet is not subject to
2217 * 'ipfw fwd'.
2218 *
2219 * XXX - Checking also should be disabled if the destination
2220 * address is ipnat'ed to a different interface.
2221 *
2222 * XXX - Checking is incompatible with IP aliases added
2223 * to the loopback interface instead of the interface where
2224 * the packets are received.
2225 */
2226 checkif = ip_checkinterface && (ipforwarding == 0) &&
2227 !(inifp->if_flags & IFF_LOOPBACK) &&
2228 !(m->m_pkthdr.pkt_flags & PKTF_LOOP)
2229#if IPFIREWALL
2230 && (args.fwa_next_hop == NULL);
2231#else /* !IPFIREWALL */
2232 ;
2233#endif /* !IPFIREWALL */
2234
2235 /*
2236 * Check for exact addresses in the hash bucket.
2237 */
2238 lck_rw_lock_shared(in_ifaddr_rwlock);
2239 TAILQ_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
2240 /*
2241 * If the address matches, verify that the packet
2242 * arrived via the correct interface if checking is
2243 * enabled.
2244 */
2245 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
2246 (!checkif || ia->ia_ifp == inifp)) {
2247 ip_setdstifaddr_info(m, 0, ia);
2248 lck_rw_done(in_ifaddr_rwlock);
2249 goto ours;
2250 }
2251 }
2252 lck_rw_done(in_ifaddr_rwlock);
2253
2254 /*
2255 * Check for broadcast addresses.
2256 *
2257 * Only accept broadcast packets that arrive via the matching
2258 * interface. Reception of forwarded directed broadcasts would be
2259 * handled via ip_forward() and ether_frameout() with the loopback
2260 * into the stack for SIMPLEX interfaces handled by ether_frameout().
2261 */
2262 if (inifp->if_flags & IFF_BROADCAST) {
2263 struct ifaddr *ifa;
2264
2265 ifnet_lock_shared(inifp);
2266 TAILQ_FOREACH(ifa, &inifp->if_addrhead, ifa_link) {
2267 if (ifa->ifa_addr->sa_family != AF_INET) {
2268 continue;
2269 }
2270 ia = ifatoia(ifa);
2271 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
2272 pkt_dst.s_addr || ia->ia_netbroadcast.s_addr ==
2273 pkt_dst.s_addr) {
2274 ip_setdstifaddr_info(m, 0, ia);
2275 ifnet_lock_done(inifp);
2276 goto ours;
2277 }
2278 }
2279 ifnet_lock_done(inifp);
2280 }
2281
2282 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
2283 struct in_multi *inm;
2284 /*
2285 * See if we belong to the destination multicast group on the
2286 * arrival interface.
2287 */
2288 in_multihead_lock_shared();
2289 IN_LOOKUP_MULTI(&ip->ip_dst, inifp, inm);
2290 in_multihead_lock_done();
2291 if (inm == NULL) {
2292 OSAddAtomic(1, &ipstat.ips_notmember);
2293 m_freem(m);
2294 return;
2295 }
2296 ip_setdstifaddr_info(m, inifp->if_index, NULL);
2297 INM_REMREF(inm);
2298 goto ours;
2299 }
2300 if (ip->ip_dst.s_addr == (u_int32_t)INADDR_BROADCAST ||
2301 ip->ip_dst.s_addr == INADDR_ANY) {
2302 ip_setdstifaddr_info(m, inifp->if_index, NULL);
2303 goto ours;
2304 }
2305
2306 /* Allow DHCP/BootP responses through */
2307 if ((inifp->if_eflags & IFEF_AUTOCONFIGURING) &&
2308 hlen == sizeof (struct ip) && ip->ip_p == IPPROTO_UDP) {
2309 struct udpiphdr *ui;
2310
2311 if (m->m_len < sizeof (struct udpiphdr) &&
2312 (m = m_pullup(m, sizeof (struct udpiphdr))) == NULL) {
2313 OSAddAtomic(1, &udpstat.udps_hdrops);
2314 return;
2315 }
2316 ui = mtod(m, struct udpiphdr *);
2317 if (ntohs(ui->ui_dport) == IPPORT_BOOTPC) {
2318 ip_setdstifaddr_info(m, inifp->if_index, NULL);
2319 goto ours;
2320 }
2321 ip = mtod(m, struct ip *); /* in case it changed */
2322 }
2323
2324 /*
2325 * Not for us; forward if possible and desirable.
2326 */
2327 if (ipforwarding == 0) {
2328 OSAddAtomic(1, &ipstat.ips_cantforward);
2329 m_freem(m);
2330 } else {
2331#if IPFIREWALL
2332 ip_forward(m, 0, args.fwa_next_hop);
2333#else
2334 ip_forward(m, 0, NULL);
2335#endif
2336 }
2337 return;
2338
2339ours:
2340 /*
2341 * If offset or IP_MF are set, must reassemble.
2342 */
2343 if (ip->ip_off & ~(IP_DF | IP_RF)) {
2344 /*
2345 * ip_reass() will return a different mbuf, and update
2346 * the divert info in div_info and args.fwa_divert_rule.
2347 */
2348#if IPDIVERT
2349 m = ip_reass(m, (u_int16_t *)&div_info, &args.fwa_divert_rule);
2350#else
2351 m = ip_reass(m);
2352#endif
2353 if (m == NULL)
2354 return;
2355 ip = mtod(m, struct ip *);
2356 /* Get the header length of the reassembled packet */
2357 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
2358#if IPDIVERT
2359 /* Restore original checksum before diverting packet */
2360 if (div_info != 0) {
2361#if BYTE_ORDER != BIG_ENDIAN
2362 HTONS(ip->ip_len);
2363 HTONS(ip->ip_off);
2364#endif
2365 ip->ip_sum = 0;
2366 ip->ip_sum = ip_cksum_hdr_in(m, hlen);
2367#if BYTE_ORDER != BIG_ENDIAN
2368 NTOHS(ip->ip_off);
2369 NTOHS(ip->ip_len);
2370#endif
2371 }
2372#endif
2373 }
2374
2375 /*
2376 * Further protocols expect the packet length to be w/o the
2377 * IP header.
2378 */
2379 ip->ip_len -= hlen;
2380
2381#if IPDIVERT
2382 /*
2383 * Divert or tee packet to the divert protocol if required.
2384 *
2385 * If div_info is zero then cookie should be too, so we shouldn't
2386 * need to clear them here. Assume divert_packet() does so also.
2387 */
2388 if (div_info != 0) {
2389 struct mbuf *clone = NULL;
2390
2391 /* Clone packet if we're doing a 'tee' */
2392 if (div_info & IP_FW_PORT_TEE_FLAG)
2393 clone = m_dup(m, M_DONTWAIT);
2394
2395 /* Restore packet header fields to original values */
2396 ip->ip_len += hlen;
2397
2398#if BYTE_ORDER != BIG_ENDIAN
2399 HTONS(ip->ip_len);
2400 HTONS(ip->ip_off);
2401#endif
2402 /* Deliver packet to divert input routine */
2403 OSAddAtomic(1, &ipstat.ips_delivered);
2404 divert_packet(m, 1, div_info & 0xffff, args.fwa_divert_rule);
2405
2406 /* If 'tee', continue with original packet */
2407 if (clone == NULL) {
2408 return;
2409 }
2410 m = clone;
2411 ip = mtod(m, struct ip *);
2412 }
2413#endif
2414
2415#if IPSEC
2416 /*
2417 * enforce IPsec policy checking if we are seeing last header.
2418 * note that we do not visit this with protocols with pcb layer
2419 * code - like udp/tcp/raw ip.
2420 */
2421 if (ipsec_bypass == 0 && (ip_protox[ip->ip_p]->pr_flags & PR_LASTHDR)) {
2422 if (ipsec4_in_reject(m, NULL)) {
2423 IPSEC_STAT_INCREMENT(ipsecstat.in_polvio);
2424 goto bad;
2425 }
2426 }
2427#endif /* IPSEC */
2428
2429 /*
2430 * Switch out to protocol's input routine.
2431 */
2432 OSAddAtomic(1, &ipstat.ips_delivered);
2433
2434#if IPFIREWALL
2435 if (args.fwa_next_hop && ip->ip_p == IPPROTO_TCP) {
2436 /* TCP needs IPFORWARD info if available */
2437 struct m_tag *fwd_tag;
2438 struct ip_fwd_tag *ipfwd_tag;
2439
2440 fwd_tag = m_tag_create(KERNEL_MODULE_TAG_ID,
2441 KERNEL_TAG_TYPE_IPFORWARD, sizeof (*ipfwd_tag),
2442 M_NOWAIT, m);
2443 if (fwd_tag == NULL)
2444 goto bad;
2445
2446 ipfwd_tag = (struct ip_fwd_tag *)(fwd_tag+1);
2447 ipfwd_tag->next_hop = args.fwa_next_hop;
2448
2449 m_tag_prepend(m, fwd_tag);
2450
2451 KERNEL_DEBUG(DBG_LAYER_END, ip->ip_dst.s_addr,
2452 ip->ip_src.s_addr, ip->ip_p, ip->ip_off, ip->ip_len);
2453
2454 /* TCP deals with its own locking */
2455 ip_proto_dispatch_in(m, hlen, ip->ip_p, 0);
2456 } else {
2457 KERNEL_DEBUG(DBG_LAYER_END, ip->ip_dst.s_addr,
2458 ip->ip_src.s_addr, ip->ip_p, ip->ip_off, ip->ip_len);
2459
2460 if ((sw_lro) && (ip->ip_p == IPPROTO_TCP)) {
2461 m = tcp_lro(m, hlen);
2462 if (m == NULL)
2463 return;
2464 }
2465
2466 ip_proto_dispatch_in(m, hlen, ip->ip_p, 0);
2467 }
2468#else /* !IPFIREWALL */
2469 if ((sw_lro) && (ip->ip_p == IPPROTO_TCP)) {
2470 m = tcp_lro(m, hlen);
2471 if (m == NULL)
2472 return;
2473 }
2474 ip_proto_dispatch_in(m, hlen, ip->ip_p, 0);
2475#endif /* !IPFIREWALL */
2476 return;
2477
2478bad:
2479 KERNEL_DEBUG(DBG_LAYER_END, 0, 0, 0, 0, 0);
2480 m_freem(m);
2481}
2482
2483static void
2484ipq_updateparams(void)
2485{
2486 LCK_MTX_ASSERT(&ipqlock, LCK_MTX_ASSERT_OWNED);
2487 /*
2488 * -1 for unlimited allocation.
2489 */
2490 if (maxnipq < 0)
2491 ipq_limit = 0;
2492 /*
2493 * Positive number for specific bound.
2494 */
2495 if (maxnipq > 0)
2496 ipq_limit = maxnipq;
2497 /*
2498 * Zero specifies no further fragment queue allocation -- set the
2499 * bound very low, but rely on implementation elsewhere to actually
2500 * prevent allocation and reclaim current queues.
2501 */
2502 if (maxnipq == 0)
2503 ipq_limit = 1;
2504 /*
2505 * Arm the purge timer if not already and if there's work to do
2506 */
2507 frag_sched_timeout();
2508}
2509
2510static int
2511sysctl_maxnipq SYSCTL_HANDLER_ARGS
2512{
2513#pragma unused(arg1, arg2)
2514 int error, i;
2515
2516 lck_mtx_lock(&ipqlock);
2517 i = maxnipq;
2518 error = sysctl_handle_int(oidp, &i, 0, req);
2519 if (error || req->newptr == USER_ADDR_NULL)
2520 goto done;
2521 /* impose bounds */
2522 if (i < -1 || i > (nmbclusters / 4)) {
2523 error = EINVAL;
2524 goto done;
2525 }
2526 maxnipq = i;
2527 ipq_updateparams();
2528done:
2529 lck_mtx_unlock(&ipqlock);
2530 return (error);
2531}
2532
2533static int
2534sysctl_maxfragsperpacket SYSCTL_HANDLER_ARGS
2535{
2536#pragma unused(arg1, arg2)
2537 int error, i;
2538
2539 lck_mtx_lock(&ipqlock);
2540 i = maxfragsperpacket;
2541 error = sysctl_handle_int(oidp, &i, 0, req);
2542 if (error || req->newptr == USER_ADDR_NULL)
2543 goto done;
2544 maxfragsperpacket = i;
2545 ipq_updateparams(); /* see if we need to arm timer */
2546done:
2547 lck_mtx_unlock(&ipqlock);
2548 return (error);
2549}
2550
2551/*
2552 * Take incoming datagram fragment and try to reassemble it into
2553 * whole datagram. If a chain for reassembly of this datagram already
2554 * exists, then it is given as fp; otherwise have to make a chain.
2555 *
2556 * When IPDIVERT enabled, keep additional state with each packet that
2557 * tells us if we need to divert or tee the packet we're building.
2558 *
2559 * The IP header is *NOT* adjusted out of iplen (but in host byte order).
2560 */
2561static struct mbuf *
2562#if IPDIVERT
2563ip_reass(struct mbuf *m,
2564#ifdef IPDIVERT_44
2565 u_int32_t *divinfo,
2566#else /* IPDIVERT_44 */
2567 u_int16_t *divinfo,
2568#endif /* IPDIVERT_44 */
2569 u_int16_t *divcookie)
2570#else /* IPDIVERT */
2571ip_reass(struct mbuf *m)
2572#endif /* IPDIVERT */
2573{
2574 struct ip *ip;
2575 struct mbuf *p, *q, *nq, *t;
2576 struct ipq *fp = NULL;
2577 struct ipqhead *head;
2578 int i, hlen, next;
2579 u_int8_t ecn, ecn0;
2580 uint32_t csum, csum_flags;
2581 uint16_t hash;
2582 struct fq_head dfq;
2583
2584 MBUFQ_INIT(&dfq); /* for deferred frees */
2585
2586 /* If maxnipq or maxfragsperpacket is 0, never accept fragments. */
2587 if (maxnipq == 0 || maxfragsperpacket == 0) {
2588 ipstat.ips_fragments++;
2589 ipstat.ips_fragdropped++;
2590 m_freem(m);
2591 if (nipq > 0) {
2592 lck_mtx_lock(&ipqlock);
2593 frag_sched_timeout(); /* purge stale fragments */
2594 lck_mtx_unlock(&ipqlock);
2595 }
2596 return (NULL);
2597 }
2598
2599 ip = mtod(m, struct ip *);
2600 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
2601
2602 lck_mtx_lock(&ipqlock);
2603
2604 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
2605 head = &ipq[hash];
2606
2607 /*
2608 * Look for queue of fragments
2609 * of this datagram.
2610 */
2611 TAILQ_FOREACH(fp, head, ipq_list) {
2612 if (ip->ip_id == fp->ipq_id &&
2613 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
2614 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
2615#if CONFIG_MACF_NET
2616 mac_ipq_label_compare(m, fp) &&
2617#endif
2618 ip->ip_p == fp->ipq_p)
2619 goto found;
2620 }
2621
2622 fp = NULL;
2623
2624 /*
2625 * Attempt to trim the number of allocated fragment queues if it
2626 * exceeds the administrative limit.
2627 */
2628 if ((nipq > (unsigned)maxnipq) && (maxnipq > 0)) {
2629 /*
2630 * drop something from the tail of the current queue
2631 * before proceeding further
2632 */
2633 struct ipq *fq = TAILQ_LAST(head, ipqhead);
2634 if (fq == NULL) { /* gak */
2635 for (i = 0; i < IPREASS_NHASH; i++) {
2636 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
2637 if (r) {
2638 ipstat.ips_fragtimeout += r->ipq_nfrags;
2639 frag_freef(&ipq[i], r);
2640 break;
2641 }
2642 }
2643 } else {
2644 ipstat.ips_fragtimeout += fq->ipq_nfrags;
2645 frag_freef(head, fq);
2646 }
2647 }
2648
2649found:
2650 /*
2651 * Leverage partial checksum offload for IP fragments. Narrow down
2652 * the scope to cover only UDP without IP options, as that is the
2653 * most common case.
2654 *
2655 * Perform 1's complement adjustment of octets that got included/
2656 * excluded in the hardware-calculated checksum value. Ignore cases
2657 * where the value includes the entire IPv4 header span, as the sum
2658 * for those octets would already be 0 by the time we get here; IP
2659 * has already performed its header checksum validation. Also take
2660 * care of any trailing bytes and subtract out their partial sum.
2661 */
2662 if (ip->ip_p == IPPROTO_UDP && hlen == sizeof (struct ip) &&
2663 (m->m_pkthdr.csum_flags &
2664 (CSUM_DATA_VALID | CSUM_PARTIAL | CSUM_PSEUDO_HDR)) ==
2665 (CSUM_DATA_VALID | CSUM_PARTIAL)) {
2666 uint32_t start = m->m_pkthdr.csum_rx_start;
2667 int32_t trailer = (m_pktlen(m) - ip->ip_len);
2668 uint32_t swbytes = (uint32_t)trailer;
2669
2670 csum = m->m_pkthdr.csum_rx_val;
2671
2672 ASSERT(trailer >= 0);
2673 if ((start != 0 && start != hlen) || trailer != 0) {
2674#if BYTE_ORDER != BIG_ENDIAN
2675 if (start < hlen) {
2676 HTONS(ip->ip_len);
2677 HTONS(ip->ip_off);
2678 }
2679#endif /* BYTE_ORDER != BIG_ENDIAN */
2680 /* callee folds in sum */
2681 csum = m_adj_sum16(m, start, hlen,
2682 (ip->ip_len - hlen), csum);
2683 if (hlen > start)
2684 swbytes += (hlen - start);
2685 else
2686 swbytes += (start - hlen);
2687#if BYTE_ORDER != BIG_ENDIAN
2688 if (start < hlen) {
2689 NTOHS(ip->ip_off);
2690 NTOHS(ip->ip_len);
2691 }
2692#endif /* BYTE_ORDER != BIG_ENDIAN */
2693 }
2694 csum_flags = m->m_pkthdr.csum_flags;
2695
2696 if (swbytes != 0)
2697 udp_in_cksum_stats(swbytes);
2698 if (trailer != 0)
2699 m_adj(m, -trailer);
2700 } else {
2701 csum = 0;
2702 csum_flags = 0;
2703 }
2704
2705 /* Invalidate checksum */
2706 m->m_pkthdr.csum_flags &= ~CSUM_DATA_VALID;
2707
2708 ipstat.ips_fragments++;
2709
2710 /*
2711 * Adjust ip_len to not reflect header,
2712 * convert offset of this to bytes.
2713 */
2714 ip->ip_len -= hlen;
2715 if (ip->ip_off & IP_MF) {
2716 /*
2717 * Make sure that fragments have a data length
2718 * that's a non-zero multiple of 8 bytes.
2719 */
2720 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
2721 OSAddAtomic(1, &ipstat.ips_toosmall);
2722 /*
2723 * Reassembly queue may have been found if previous
2724 * fragments were valid; given that this one is bad,
2725 * we need to drop it. Make sure to set fp to NULL
2726 * if not already, since we don't want to decrement
2727 * ipq_nfrags as it doesn't include this packet.
2728 */
2729 fp = NULL;
2730 goto dropfrag;
2731 }
2732 m->m_flags |= M_FRAG;
2733 } else {
2734 /* Clear the flag in case packet comes from loopback */
2735 m->m_flags &= ~M_FRAG;
2736 }
2737 ip->ip_off <<= 3;
2738
2739 m->m_pkthdr.pkt_hdr = ip;
2740
2741 /* Previous ip_reass() started here. */
2742 /*
2743 * Presence of header sizes in mbufs
2744 * would confuse code below.
2745 */
2746 m->m_data += hlen;
2747 m->m_len -= hlen;
2748
2749 /*
2750 * If first fragment to arrive, create a reassembly queue.
2751 */
2752 if (fp == NULL) {
2753 fp = ipq_alloc(M_DONTWAIT);
2754 if (fp == NULL)
2755 goto dropfrag;
2756#if CONFIG_MACF_NET
2757 if (mac_ipq_label_init(fp, M_NOWAIT) != 0) {
2758 ipq_free(fp);
2759 fp = NULL;
2760 goto dropfrag;
2761 }
2762 mac_ipq_label_associate(m, fp);
2763#endif
2764 TAILQ_INSERT_HEAD(head, fp, ipq_list);
2765 nipq++;
2766 fp->ipq_nfrags = 1;
2767 fp->ipq_ttl = IPFRAGTTL;
2768 fp->ipq_p = ip->ip_p;
2769 fp->ipq_id = ip->ip_id;
2770 fp->ipq_src = ip->ip_src;
2771 fp->ipq_dst = ip->ip_dst;
2772 fp->ipq_frags = m;
2773 m->m_nextpkt = NULL;
2774 /*
2775 * If the first fragment has valid checksum offload
2776 * info, the rest of fragments are eligible as well.
2777 */
2778 if (csum_flags != 0) {
2779 fp->ipq_csum = csum;
2780 fp->ipq_csum_flags = csum_flags;
2781 }
2782#if IPDIVERT
2783 /*
2784 * Transfer firewall instructions to the fragment structure.
2785 * Only trust info in the fragment at offset 0.
2786 */
2787 if (ip->ip_off == 0) {
2788#ifdef IPDIVERT_44
2789 fp->ipq_div_info = *divinfo;
2790#else
2791 fp->ipq_divert = *divinfo;
2792#endif
2793 fp->ipq_div_cookie = *divcookie;
2794 }
2795 *divinfo = 0;
2796 *divcookie = 0;
2797#endif /* IPDIVERT */
2798 m = NULL; /* nothing to return */
2799 goto done;
2800 } else {
2801 fp->ipq_nfrags++;
2802#if CONFIG_MACF_NET
2803 mac_ipq_label_update(m, fp);
2804#endif
2805 }
2806
2807#define GETIP(m) ((struct ip *)((m)->m_pkthdr.pkt_hdr))
2808
2809 /*
2810 * Handle ECN by comparing this segment with the first one;
2811 * if CE is set, do not lose CE.
2812 * drop if CE and not-ECT are mixed for the same packet.
2813 */
2814 ecn = ip->ip_tos & IPTOS_ECN_MASK;
2815 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
2816 if (ecn == IPTOS_ECN_CE) {
2817 if (ecn0 == IPTOS_ECN_NOTECT)
2818 goto dropfrag;
2819 if (ecn0 != IPTOS_ECN_CE)
2820 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
2821 }
2822 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
2823 goto dropfrag;
2824
2825 /*
2826 * Find a segment which begins after this one does.
2827 */
2828 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
2829 if (GETIP(q)->ip_off > ip->ip_off)
2830 break;
2831
2832 /*
2833 * If there is a preceding segment, it may provide some of
2834 * our data already. If so, drop the data from the incoming
2835 * segment. If it provides all of our data, drop us, otherwise
2836 * stick new segment in the proper place.
2837 *
2838 * If some of the data is dropped from the preceding
2839 * segment, then it's checksum is invalidated.
2840 */
2841 if (p) {
2842 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
2843 if (i > 0) {
2844 if (i >= ip->ip_len)
2845 goto dropfrag;
2846 m_adj(m, i);
2847 fp->ipq_csum_flags = 0;
2848 ip->ip_off += i;
2849 ip->ip_len -= i;
2850 }
2851 m->m_nextpkt = p->m_nextpkt;
2852 p->m_nextpkt = m;
2853 } else {
2854 m->m_nextpkt = fp->ipq_frags;
2855 fp->ipq_frags = m;
2856 }
2857
2858 /*
2859 * While we overlap succeeding segments trim them or,
2860 * if they are completely covered, dequeue them.
2861 */
2862 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
2863 q = nq) {
2864 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
2865 if (i < GETIP(q)->ip_len) {
2866 GETIP(q)->ip_len -= i;
2867 GETIP(q)->ip_off += i;
2868 m_adj(q, i);
2869 fp->ipq_csum_flags = 0;
2870 break;
2871 }
2872 nq = q->m_nextpkt;
2873 m->m_nextpkt = nq;
2874 ipstat.ips_fragdropped++;
2875 fp->ipq_nfrags--;
2876 /* defer freeing until after lock is dropped */
2877 MBUFQ_ENQUEUE(&dfq, q);
2878 }
2879
2880 /*
2881 * If this fragment contains similar checksum offload info
2882 * as that of the existing ones, accumulate checksum. Otherwise,
2883 * invalidate checksum offload info for the entire datagram.
2884 */
2885 if (csum_flags != 0 && csum_flags == fp->ipq_csum_flags)
2886 fp->ipq_csum += csum;
2887 else if (fp->ipq_csum_flags != 0)
2888 fp->ipq_csum_flags = 0;
2889
2890#if IPDIVERT
2891 /*
2892 * Transfer firewall instructions to the fragment structure.
2893 * Only trust info in the fragment at offset 0.
2894 */
2895 if (ip->ip_off == 0) {
2896#ifdef IPDIVERT_44
2897 fp->ipq_div_info = *divinfo;
2898#else
2899 fp->ipq_divert = *divinfo;
2900#endif
2901 fp->ipq_div_cookie = *divcookie;
2902 }
2903 *divinfo = 0;
2904 *divcookie = 0;
2905#endif /* IPDIVERT */
2906
2907 /*
2908 * Check for complete reassembly and perform frag per packet
2909 * limiting.
2910 *
2911 * Frag limiting is performed here so that the nth frag has
2912 * a chance to complete the packet before we drop the packet.
2913 * As a result, n+1 frags are actually allowed per packet, but
2914 * only n will ever be stored. (n = maxfragsperpacket.)
2915 *
2916 */
2917 next = 0;
2918 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
2919 if (GETIP(q)->ip_off != next) {
2920 if (fp->ipq_nfrags > maxfragsperpacket) {
2921 ipstat.ips_fragdropped += fp->ipq_nfrags;
2922 frag_freef(head, fp);
2923 }
2924 m = NULL; /* nothing to return */
2925 goto done;
2926 }
2927 next += GETIP(q)->ip_len;
2928 }
2929 /* Make sure the last packet didn't have the IP_MF flag */
2930 if (p->m_flags & M_FRAG) {
2931 if (fp->ipq_nfrags > maxfragsperpacket) {
2932 ipstat.ips_fragdropped += fp->ipq_nfrags;
2933 frag_freef(head, fp);
2934 }
2935 m = NULL; /* nothing to return */
2936 goto done;
2937 }
2938
2939 /*
2940 * Reassembly is complete. Make sure the packet is a sane size.
2941 */
2942 q = fp->ipq_frags;
2943 ip = GETIP(q);
2944 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
2945 ipstat.ips_toolong++;
2946 ipstat.ips_fragdropped += fp->ipq_nfrags;
2947 frag_freef(head, fp);
2948 m = NULL; /* nothing to return */
2949 goto done;
2950 }
2951
2952 /*
2953 * Concatenate fragments.
2954 */
2955 m = q;
2956 t = m->m_next;
2957 m->m_next = NULL;
2958 m_cat(m, t);
2959 nq = q->m_nextpkt;
2960 q->m_nextpkt = NULL;
2961 for (q = nq; q != NULL; q = nq) {
2962 nq = q->m_nextpkt;
2963 q->m_nextpkt = NULL;
2964 m_cat(m, q);
2965 }
2966
2967 /*
2968 * Store partial hardware checksum info from the fragment queue;
2969 * the receive start offset is set to 20 bytes (see code at the
2970 * top of this routine.)
2971 */
2972 if (fp->ipq_csum_flags != 0) {
2973 csum = fp->ipq_csum;
2974
2975 ADDCARRY(csum);
2976
2977 m->m_pkthdr.csum_rx_val = csum;
2978 m->m_pkthdr.csum_rx_start = sizeof (struct ip);
2979 m->m_pkthdr.csum_flags = fp->ipq_csum_flags;
2980 } else if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) ||
2981 (m->m_pkthdr.pkt_flags & PKTF_LOOP)) {
2982 /* loopback checksums are always OK */
2983 m->m_pkthdr.csum_data = 0xffff;
2984 m->m_pkthdr.csum_flags &= ~CSUM_PARTIAL;
2985 m->m_pkthdr.csum_flags =
2986 CSUM_DATA_VALID | CSUM_PSEUDO_HDR |
2987 CSUM_IP_CHECKED | CSUM_IP_VALID;
2988 }
2989
2990#if IPDIVERT
2991 /*
2992 * Extract firewall instructions from the fragment structure.
2993 */
2994#ifdef IPDIVERT_44
2995 *divinfo = fp->ipq_div_info;
2996#else
2997 *divinfo = fp->ipq_divert;
2998#endif
2999 *divcookie = fp->ipq_div_cookie;
3000#endif /* IPDIVERT */
3001
3002#if CONFIG_MACF_NET
3003 mac_mbuf_label_associate_ipq(fp, m);
3004 mac_ipq_label_destroy(fp);
3005#endif
3006 /*
3007 * Create header for new ip packet by modifying header of first
3008 * packet; dequeue and discard fragment reassembly header.
3009 * Make header visible.
3010 */
3011 ip->ip_len = (IP_VHL_HL(ip->ip_vhl) << 2) + next;
3012 ip->ip_src = fp->ipq_src;
3013 ip->ip_dst = fp->ipq_dst;
3014
3015 fp->ipq_frags = NULL; /* return to caller as 'm' */
3016 frag_freef(head, fp);
3017 fp = NULL;
3018
3019 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
3020 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
3021 /* some debugging cruft by sklower, below, will go away soon */
3022 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
3023 m_fixhdr(m);
3024 ipstat.ips_reassembled++;
3025
3026 /* arm the purge timer if not already and if there's work to do */
3027 frag_sched_timeout();
3028 lck_mtx_unlock(&ipqlock);
3029 /* perform deferred free (if needed) now that lock is dropped */
3030 if (!MBUFQ_EMPTY(&dfq))
3031 MBUFQ_DRAIN(&dfq);
3032 VERIFY(MBUFQ_EMPTY(&dfq));
3033 return (m);
3034
3035done:
3036 VERIFY(m == NULL);
3037 /* arm the purge timer if not already and if there's work to do */
3038 frag_sched_timeout();
3039 lck_mtx_unlock(&ipqlock);
3040 /* perform deferred free (if needed) */
3041 if (!MBUFQ_EMPTY(&dfq))
3042 MBUFQ_DRAIN(&dfq);
3043 VERIFY(MBUFQ_EMPTY(&dfq));
3044 return (NULL);
3045
3046dropfrag:
3047#if IPDIVERT
3048 *divinfo = 0;
3049 *divcookie = 0;
3050#endif /* IPDIVERT */
3051 ipstat.ips_fragdropped++;
3052 if (fp != NULL)
3053 fp->ipq_nfrags--;
3054 /* arm the purge timer if not already and if there's work to do */
3055 frag_sched_timeout();
3056 lck_mtx_unlock(&ipqlock);
3057 m_freem(m);
3058 /* perform deferred free (if needed) */
3059 if (!MBUFQ_EMPTY(&dfq))
3060 MBUFQ_DRAIN(&dfq);
3061 VERIFY(MBUFQ_EMPTY(&dfq));
3062 return (NULL);
3063#undef GETIP
3064}
3065
3066/*
3067 * Free a fragment reassembly header and all
3068 * associated datagrams.
3069 */
3070static void
3071frag_freef(struct ipqhead *fhp, struct ipq *fp)
3072{
3073 LCK_MTX_ASSERT(&ipqlock, LCK_MTX_ASSERT_OWNED);
3074
3075 fp->ipq_nfrags = 0;
3076 if (fp->ipq_frags != NULL) {
3077 m_freem_list(fp->ipq_frags);
3078 fp->ipq_frags = NULL;
3079 }
3080 TAILQ_REMOVE(fhp, fp, ipq_list);
3081 nipq--;
3082 ipq_free(fp);
3083}
3084
3085/*
3086 * IP reassembly timer processing
3087 */
3088static void
3089frag_timeout(void *arg)
3090{
3091#pragma unused(arg)
3092 struct ipq *fp;
3093 int i;
3094
3095 /*
3096 * Update coarse-grained networking timestamp (in sec.); the idea
3097 * is to piggy-back on the timeout callout to update the counter
3098 * returnable via net_uptime().
3099 */
3100 net_update_uptime();
3101
3102 lck_mtx_lock(&ipqlock);
3103 for (i = 0; i < IPREASS_NHASH; i++) {
3104 for (fp = TAILQ_FIRST(&ipq[i]); fp; ) {
3105 struct ipq *fpp;
3106
3107 fpp = fp;
3108 fp = TAILQ_NEXT(fp, ipq_list);
3109 if (--fpp->ipq_ttl == 0) {
3110 ipstat.ips_fragtimeout += fpp->ipq_nfrags;
3111 frag_freef(&ipq[i], fpp);
3112 }
3113 }
3114 }
3115 /*
3116 * If we are over the maximum number of fragments
3117 * (due to the limit being lowered), drain off
3118 * enough to get down to the new limit.
3119 */
3120 if (maxnipq >= 0 && nipq > (unsigned)maxnipq) {
3121 for (i = 0; i < IPREASS_NHASH; i++) {
3122 while (nipq > (unsigned)maxnipq &&
3123 !TAILQ_EMPTY(&ipq[i])) {
3124 ipstat.ips_fragdropped +=
3125 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
3126 frag_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
3127 }
3128 }
3129 }
3130 /* re-arm the purge timer if there's work to do */
3131 frag_timeout_run = 0;
3132 frag_sched_timeout();
3133 lck_mtx_unlock(&ipqlock);
3134}
3135
3136static void
3137frag_sched_timeout(void)
3138{
3139 LCK_MTX_ASSERT(&ipqlock, LCK_MTX_ASSERT_OWNED);
3140
3141 if (!frag_timeout_run && nipq > 0) {
3142 frag_timeout_run = 1;
3143 timeout(frag_timeout, NULL, hz);
3144 }
3145}
3146
3147/*
3148 * Drain off all datagram fragments.
3149 */
3150static void
3151frag_drain(void)
3152{
3153 int i;
3154
3155 lck_mtx_lock(&ipqlock);
3156 for (i = 0; i < IPREASS_NHASH; i++) {
3157 while (!TAILQ_EMPTY(&ipq[i])) {
3158 ipstat.ips_fragdropped +=
3159 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
3160 frag_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
3161 }
3162 }
3163 lck_mtx_unlock(&ipqlock);
3164}
3165
3166static struct ipq *
3167ipq_alloc(int how)
3168{
3169 struct mbuf *t;
3170 struct ipq *fp;
3171
3172 /*
3173 * See comments in ipq_updateparams(). Keep the count separate
3174 * from nipq since the latter represents the elements already
3175 * in the reassembly queues.
3176 */
3177 if (ipq_limit > 0 && ipq_count > ipq_limit)
3178 return (NULL);
3179
3180 t = m_get(how, MT_FTABLE);
3181 if (t != NULL) {
3182 atomic_add_32(&ipq_count, 1);
3183 fp = mtod(t, struct ipq *);
3184 bzero(fp, sizeof (*fp));
3185 } else {
3186 fp = NULL;
3187 }
3188 return (fp);
3189}
3190
3191static void
3192ipq_free(struct ipq *fp)
3193{
3194 (void) m_free(dtom(fp));
3195 atomic_add_32(&ipq_count, -1);
3196}
3197
3198/*
3199 * Drain callback
3200 */
3201void
3202ip_drain(void)
3203{
3204 frag_drain(); /* fragments */
3205 in_rtqdrain(); /* protocol cloned routes */
3206 in_arpdrain(NULL); /* cloned routes: ARP */
3207}
3208
3209/*
3210 * Do option processing on a datagram,
3211 * possibly discarding it if bad options are encountered,
3212 * or forwarding it if source-routed.
3213 * The pass argument is used when operating in the IPSTEALTH
3214 * mode to tell what options to process:
3215 * [LS]SRR (pass 0) or the others (pass 1).
3216 * The reason for as many as two passes is that when doing IPSTEALTH,
3217 * non-routing options should be processed only if the packet is for us.
3218 * Returns 1 if packet has been forwarded/freed,
3219 * 0 if the packet should be processed further.
3220 */
3221static int
3222ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
3223{
3224#pragma unused(pass)
3225 struct ip *ip = mtod(m, struct ip *);
3226 u_char *cp;
3227 struct ip_timestamp *ipt;
3228 struct in_ifaddr *ia;
3229 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
3230 struct in_addr *sin, dst;
3231 u_int32_t ntime;
3232 struct sockaddr_in ipaddr = {
3233 sizeof (ipaddr), AF_INET, 0, { 0 }, { 0, } };
3234
3235 /* Expect 32-bit aligned data pointer on strict-align platforms */
3236 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
3237
3238 dst = ip->ip_dst;
3239 cp = (u_char *)(ip + 1);
3240 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
3241 for (; cnt > 0; cnt -= optlen, cp += optlen) {
3242 opt = cp[IPOPT_OPTVAL];
3243 if (opt == IPOPT_EOL)
3244 break;
3245 if (opt == IPOPT_NOP)
3246 optlen = 1;
3247 else {
3248 if (cnt < IPOPT_OLEN + sizeof (*cp)) {
3249 code = &cp[IPOPT_OLEN] - (u_char *)ip;
3250 goto bad;
3251 }
3252 optlen = cp[IPOPT_OLEN];
3253 if (optlen < IPOPT_OLEN + sizeof (*cp) ||
3254 optlen > cnt) {
3255 code = &cp[IPOPT_OLEN] - (u_char *)ip;
3256 goto bad;
3257 }
3258 }
3259 switch (opt) {
3260
3261 default:
3262 break;
3263
3264 /*
3265 * Source routing with record.
3266 * Find interface with current destination address.
3267 * If none on this machine then drop if strictly routed,
3268 * or do nothing if loosely routed.
3269 * Record interface address and bring up next address
3270 * component. If strictly routed make sure next
3271 * address is on directly accessible net.
3272 */
3273 case IPOPT_LSRR:
3274 case IPOPT_SSRR:
3275 if (optlen < IPOPT_OFFSET + sizeof (*cp)) {
3276 code = &cp[IPOPT_OLEN] - (u_char *)ip;
3277 goto bad;
3278 }
3279 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
3280 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
3281 goto bad;
3282 }
3283 ipaddr.sin_addr = ip->ip_dst;
3284 ia = (struct in_ifaddr *)ifa_ifwithaddr(SA(&ipaddr));
3285 if (ia == NULL) {
3286 if (opt == IPOPT_SSRR) {
3287 type = ICMP_UNREACH;
3288 code = ICMP_UNREACH_SRCFAIL;
3289 goto bad;
3290 }
3291 if (!ip_dosourceroute)
3292 goto nosourcerouting;
3293 /*
3294 * Loose routing, and not at next destination
3295 * yet; nothing to do except forward.
3296 */
3297 break;
3298 } else {
3299 IFA_REMREF(&ia->ia_ifa);
3300 ia = NULL;
3301 }
3302 off--; /* 0 origin */
3303 if (off > optlen - (int)sizeof (struct in_addr)) {
3304 /*
3305 * End of source route. Should be for us.
3306 */
3307 if (!ip_acceptsourceroute)
3308 goto nosourcerouting;
3309 save_rte(cp, ip->ip_src);
3310 break;
3311 }
3312
3313 if (!ip_dosourceroute) {
3314 if (ipforwarding) {
3315 char buf[MAX_IPv4_STR_LEN];
3316 char buf2[MAX_IPv4_STR_LEN];
3317 /*
3318 * Acting as a router, so generate ICMP
3319 */
3320nosourcerouting:
3321 log(LOG_WARNING,
3322 "attempted source route from %s "
3323 "to %s\n",
3324 inet_ntop(AF_INET, &ip->ip_src,
3325 buf, sizeof (buf)),
3326 inet_ntop(AF_INET, &ip->ip_dst,
3327 buf2, sizeof (buf2)));
3328 type = ICMP_UNREACH;
3329 code = ICMP_UNREACH_SRCFAIL;
3330 goto bad;
3331 } else {
3332 /*
3333 * Not acting as a router,
3334 * so silently drop.
3335 */
3336 OSAddAtomic(1, &ipstat.ips_cantforward);
3337 m_freem(m);
3338 return (1);
3339 }
3340 }
3341
3342 /*
3343 * locate outgoing interface
3344 */
3345 (void) memcpy(&ipaddr.sin_addr, cp + off,
3346 sizeof (ipaddr.sin_addr));
3347
3348 if (opt == IPOPT_SSRR) {
3349#define INA struct in_ifaddr *
3350 if ((ia = (INA)ifa_ifwithdstaddr(
3351 SA(&ipaddr))) == NULL) {
3352 ia = (INA)ifa_ifwithnet(SA(&ipaddr));
3353 }
3354 } else {
3355 ia = ip_rtaddr(ipaddr.sin_addr);
3356 }
3357 if (ia == NULL) {
3358 type = ICMP_UNREACH;
3359 code = ICMP_UNREACH_SRCFAIL;
3360 goto bad;
3361 }
3362 ip->ip_dst = ipaddr.sin_addr;
3363 IFA_LOCK(&ia->ia_ifa);
3364 (void) memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
3365 sizeof (struct in_addr));
3366 IFA_UNLOCK(&ia->ia_ifa);
3367 IFA_REMREF(&ia->ia_ifa);
3368 ia = NULL;
3369 cp[IPOPT_OFFSET] += sizeof (struct in_addr);
3370 /*
3371 * Let ip_intr's mcast routing check handle mcast pkts
3372 */
3373 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
3374 break;
3375
3376 case IPOPT_RR:
3377 if (optlen < IPOPT_OFFSET + sizeof (*cp)) {
3378 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
3379 goto bad;
3380 }
3381 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
3382 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
3383 goto bad;
3384 }
3385 /*
3386 * If no space remains, ignore.
3387 */
3388 off--; /* 0 origin */
3389 if (off > optlen - (int)sizeof (struct in_addr))
3390 break;
3391 (void) memcpy(&ipaddr.sin_addr, &ip->ip_dst,
3392 sizeof (ipaddr.sin_addr));
3393 /*
3394 * locate outgoing interface; if we're the destination,
3395 * use the incoming interface (should be same).
3396 */
3397 if ((ia = (INA)ifa_ifwithaddr(SA(&ipaddr))) == NULL) {
3398 if ((ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) {
3399 type = ICMP_UNREACH;
3400 code = ICMP_UNREACH_HOST;
3401 goto bad;
3402 }
3403 }
3404 IFA_LOCK(&ia->ia_ifa);
3405 (void) memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
3406 sizeof (struct in_addr));
3407 IFA_UNLOCK(&ia->ia_ifa);
3408 IFA_REMREF(&ia->ia_ifa);
3409 ia = NULL;
3410 cp[IPOPT_OFFSET] += sizeof (struct in_addr);
3411 break;
3412
3413 case IPOPT_TS:
3414 code = cp - (u_char *)ip;
3415 ipt = (struct ip_timestamp *)(void *)cp;
3416 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
3417 code = (u_char *)&ipt->ipt_len - (u_char *)ip;
3418 goto bad;
3419 }
3420 if (ipt->ipt_ptr < 5) {
3421 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
3422 goto bad;
3423 }
3424 if (ipt->ipt_ptr >
3425 ipt->ipt_len - (int)sizeof (int32_t)) {
3426 if (++ipt->ipt_oflw == 0) {
3427 code = (u_char *)&ipt->ipt_ptr -
3428 (u_char *)ip;
3429 goto bad;
3430 }
3431 break;
3432 }
3433 sin = (struct in_addr *)(void *)(cp + ipt->ipt_ptr - 1);
3434 switch (ipt->ipt_flg) {
3435
3436 case IPOPT_TS_TSONLY:
3437 break;
3438
3439 case IPOPT_TS_TSANDADDR:
3440 if (ipt->ipt_ptr - 1 + sizeof (n_time) +
3441 sizeof (struct in_addr) > ipt->ipt_len) {
3442 code = (u_char *)&ipt->ipt_ptr -
3443 (u_char *)ip;
3444 goto bad;
3445 }
3446 ipaddr.sin_addr = dst;
3447 ia = (INA)ifaof_ifpforaddr(SA(&ipaddr),
3448 m->m_pkthdr.rcvif);
3449 if (ia == NULL)
3450 continue;
3451 IFA_LOCK(&ia->ia_ifa);
3452 (void) memcpy(sin, &IA_SIN(ia)->sin_addr,
3453 sizeof (struct in_addr));
3454 IFA_UNLOCK(&ia->ia_ifa);
3455 ipt->ipt_ptr += sizeof (struct in_addr);
3456 IFA_REMREF(&ia->ia_ifa);
3457 ia = NULL;
3458 break;
3459
3460 case IPOPT_TS_PRESPEC:
3461 if (ipt->ipt_ptr - 1 + sizeof (n_time) +
3462 sizeof (struct in_addr) > ipt->ipt_len) {
3463 code = (u_char *)&ipt->ipt_ptr -
3464 (u_char *)ip;
3465 goto bad;
3466 }
3467 (void) memcpy(&ipaddr.sin_addr, sin,
3468 sizeof (struct in_addr));
3469 if ((ia = (struct in_ifaddr *)ifa_ifwithaddr(
3470 SA(&ipaddr))) == NULL)
3471 continue;
3472 IFA_REMREF(&ia->ia_ifa);
3473 ia = NULL;
3474 ipt->ipt_ptr += sizeof (struct in_addr);
3475 break;
3476
3477 default:
3478 /* XXX can't take &ipt->ipt_flg */
3479 code = (u_char *)&ipt->ipt_ptr -
3480 (u_char *)ip + 1;
3481 goto bad;
3482 }
3483 ntime = iptime();
3484 (void) memcpy(cp + ipt->ipt_ptr - 1, &ntime,
3485 sizeof (n_time));
3486 ipt->ipt_ptr += sizeof (n_time);
3487 }
3488 }
3489 if (forward && ipforwarding) {
3490 ip_forward(m, 1, next_hop);
3491 return (1);
3492 }
3493 return (0);
3494bad:
3495 icmp_error(m, type, code, 0, 0);
3496 OSAddAtomic(1, &ipstat.ips_badoptions);
3497 return (1);
3498}
3499
3500/*
3501 * Check for the presence of the IP Router Alert option [RFC2113]
3502 * in the header of an IPv4 datagram.
3503 *
3504 * This call is not intended for use from the forwarding path; it is here
3505 * so that protocol domains may check for the presence of the option.
3506 * Given how FreeBSD's IPv4 stack is currently structured, the Router Alert
3507 * option does not have much relevance to the implementation, though this
3508 * may change in future.
3509 * Router alert options SHOULD be passed if running in IPSTEALTH mode and
3510 * we are not the endpoint.
3511 * Length checks on individual options should already have been peformed
3512 * by ip_dooptions() therefore they are folded under DIAGNOSTIC here.
3513 *
3514 * Return zero if not present or options are invalid, non-zero if present.
3515 */
3516int
3517ip_checkrouteralert(struct mbuf *m)
3518{
3519 struct ip *ip = mtod(m, struct ip *);
3520 u_char *cp;
3521 int opt, optlen, cnt, found_ra;
3522
3523 found_ra = 0;
3524 cp = (u_char *)(ip + 1);
3525 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
3526 for (; cnt > 0; cnt -= optlen, cp += optlen) {
3527 opt = cp[IPOPT_OPTVAL];
3528 if (opt == IPOPT_EOL)
3529 break;
3530 if (opt == IPOPT_NOP)
3531 optlen = 1;
3532 else {
3533#ifdef DIAGNOSTIC
3534 if (cnt < IPOPT_OLEN + sizeof (*cp))
3535 break;
3536#endif
3537 optlen = cp[IPOPT_OLEN];
3538#ifdef DIAGNOSTIC
3539 if (optlen < IPOPT_OLEN + sizeof (*cp) || optlen > cnt)
3540 break;
3541#endif
3542 }
3543 switch (opt) {
3544 case IPOPT_RA:
3545#ifdef DIAGNOSTIC
3546 if (optlen != IPOPT_OFFSET + sizeof (uint16_t) ||
3547 (*((uint16_t *)(void *)&cp[IPOPT_OFFSET]) != 0))
3548 break;
3549 else
3550#endif
3551 found_ra = 1;
3552 break;
3553 default:
3554 break;
3555 }
3556 }
3557
3558 return (found_ra);
3559}
3560
3561/*
3562 * Given address of next destination (final or next hop),
3563 * return internet address info of interface to be used to get there.
3564 */
3565struct in_ifaddr *
3566ip_rtaddr(struct in_addr dst)
3567{
3568 struct sockaddr_in *sin;
3569 struct ifaddr *rt_ifa;
3570 struct route ro;
3571
3572 bzero(&ro, sizeof (ro));
3573 sin = SIN(&ro.ro_dst);
3574 sin->sin_family = AF_INET;
3575 sin->sin_len = sizeof (*sin);
3576 sin->sin_addr = dst;
3577
3578 rtalloc_ign(&ro, RTF_PRCLONING);
3579 if (ro.ro_rt == NULL) {
3580 ROUTE_RELEASE(&ro);
3581 return (NULL);
3582 }
3583
3584 RT_LOCK(ro.ro_rt);
3585 if ((rt_ifa = ro.ro_rt->rt_ifa) != NULL)
3586 IFA_ADDREF(rt_ifa);
3587 RT_UNLOCK(ro.ro_rt);
3588 ROUTE_RELEASE(&ro);
3589
3590 return ((struct in_ifaddr *)rt_ifa);
3591}
3592
3593/*
3594 * Save incoming source route for use in replies,
3595 * to be picked up later by ip_srcroute if the receiver is interested.
3596 */
3597void
3598save_rte(u_char *option, struct in_addr dst)
3599{
3600 unsigned olen;
3601
3602 olen = option[IPOPT_OLEN];
3603#if DIAGNOSTIC
3604 if (ipprintfs)
3605 printf("save_rte: olen %d\n", olen);
3606#endif
3607 if (olen > sizeof (ip_srcrt) - (1 + sizeof (dst)))
3608 return;
3609 bcopy(option, ip_srcrt.srcopt, olen);
3610 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof (struct in_addr);
3611 ip_srcrt.dst = dst;
3612}
3613
3614/*
3615 * Retrieve incoming source route for use in replies,
3616 * in the same form used by setsockopt.
3617 * The first hop is placed before the options, will be removed later.
3618 */
3619struct mbuf *
3620ip_srcroute(void)
3621{
3622 struct in_addr *p, *q;
3623 struct mbuf *m;
3624
3625 if (ip_nhops == 0)
3626 return (NULL);
3627
3628 m = m_get(M_DONTWAIT, MT_HEADER);
3629 if (m == NULL)
3630 return (NULL);
3631
3632#define OPTSIZ (sizeof (ip_srcrt.nop) + sizeof (ip_srcrt.srcopt))
3633
3634 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
3635 m->m_len = ip_nhops * sizeof (struct in_addr) +
3636 sizeof (struct in_addr) + OPTSIZ;
3637#if DIAGNOSTIC
3638 if (ipprintfs)
3639 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
3640#endif
3641
3642 /*
3643 * First save first hop for return route
3644 */
3645 p = &ip_srcrt.route[ip_nhops - 1];
3646 *(mtod(m, struct in_addr *)) = *p--;
3647#if DIAGNOSTIC
3648 if (ipprintfs)
3649 printf(" hops %lx",
3650 (u_int32_t)ntohl(mtod(m, struct in_addr *)->s_addr));
3651#endif
3652
3653 /*
3654 * Copy option fields and padding (nop) to mbuf.
3655 */
3656 ip_srcrt.nop = IPOPT_NOP;
3657 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
3658 (void) memcpy(mtod(m, caddr_t) + sizeof (struct in_addr),
3659 &ip_srcrt.nop, OPTSIZ);
3660 q = (struct in_addr *)(void *)(mtod(m, caddr_t) +
3661 sizeof (struct in_addr) + OPTSIZ);
3662#undef OPTSIZ
3663 /*
3664 * Record return path as an IP source route,
3665 * reversing the path (pointers are now aligned).
3666 */
3667 while (p >= ip_srcrt.route) {
3668#if DIAGNOSTIC
3669 if (ipprintfs)
3670 printf(" %lx", (u_int32_t)ntohl(q->s_addr));
3671#endif
3672 *q++ = *p--;
3673 }
3674 /*
3675 * Last hop goes to final destination.
3676 */
3677 *q = ip_srcrt.dst;
3678#if DIAGNOSTIC
3679 if (ipprintfs)
3680 printf(" %lx\n", (u_int32_t)ntohl(q->s_addr));
3681#endif
3682 return (m);
3683}
3684
3685/*
3686 * Strip out IP options, at higher level protocol in the kernel.
3687 */
3688void
3689ip_stripoptions(struct mbuf *m)
3690{
3691 int i;
3692 struct ip *ip = mtod(m, struct ip *);
3693 caddr_t opts;
3694 int olen;
3695
3696 /* Expect 32-bit aligned data pointer on strict-align platforms */
3697 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
3698
3699 /* use bcopy() since it supports overlapping range */
3700 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
3701 opts = (caddr_t)(ip + 1);
3702 i = m->m_len - (sizeof (struct ip) + olen);
3703 bcopy(opts + olen, opts, (unsigned)i);
3704 m->m_len -= olen;
3705 if (m->m_flags & M_PKTHDR)
3706 m->m_pkthdr.len -= olen;
3707 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof (struct ip) >> 2);
3708
3709 /*
3710 * We expect ip_{off,len} to be in host order by now, and
3711 * that the original IP header length has been subtracted
3712 * out from ip_len. Temporarily adjust ip_len for checksum
3713 * recalculation, and restore it afterwards.
3714 */
3715 ip->ip_len += sizeof (struct ip);
3716
3717 /* recompute checksum now that IP header is smaller */
3718#if BYTE_ORDER != BIG_ENDIAN
3719 HTONS(ip->ip_len);
3720 HTONS(ip->ip_off);
3721#endif /* BYTE_ORDER != BIG_ENDIAN */
3722 ip->ip_sum = in_cksum_hdr(ip);
3723#if BYTE_ORDER != BIG_ENDIAN
3724 NTOHS(ip->ip_off);
3725 NTOHS(ip->ip_len);
3726#endif /* BYTE_ORDER != BIG_ENDIAN */
3727
3728 ip->ip_len -= sizeof (struct ip);
3729}
3730
3731u_char inetctlerrmap[PRC_NCMDS] = {
3732 0, 0, 0, 0,
3733 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
3734 ENETUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
3735 EMSGSIZE, EHOSTUNREACH, 0, 0,
3736 0, 0, 0, 0,
3737 ENOPROTOOPT, ECONNREFUSED
3738};
3739
3740static int
3741sysctl_ipforwarding SYSCTL_HANDLER_ARGS
3742{
3743#pragma unused(arg1, arg2)
3744 int i, was_ipforwarding = ipforwarding;
3745
3746 i = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
3747 if (i != 0 || req->newptr == USER_ADDR_NULL)
3748 return (i);
3749
3750 if (was_ipforwarding && !ipforwarding) {
3751 /* clean up IPv4 forwarding cached routes */
3752 ifnet_head_lock_shared();
3753 for (i = 0; i <= if_index; i++) {
3754 struct ifnet *ifp = ifindex2ifnet[i];
3755 if (ifp != NULL) {
3756 lck_mtx_lock(&ifp->if_cached_route_lock);
3757 ROUTE_RELEASE(&ifp->if_fwd_route);
3758 bzero(&ifp->if_fwd_route,
3759 sizeof (ifp->if_fwd_route));
3760 lck_mtx_unlock(&ifp->if_cached_route_lock);
3761 }
3762 }
3763 ifnet_head_done();
3764 }
3765
3766 return (0);
3767}
3768
3769/*
3770 * Similar to inp_route_{copyout,copyin} routines except that these copy
3771 * out the cached IPv4 forwarding route from struct ifnet instead of the
3772 * inpcb. See comments for those routines for explanations.
3773 */
3774static void
3775ip_fwd_route_copyout(struct ifnet *ifp, struct route *dst)
3776{
3777 struct route *src = &ifp->if_fwd_route;
3778
3779 lck_mtx_lock_spin(&ifp->if_cached_route_lock);
3780 lck_mtx_convert_spin(&ifp->if_cached_route_lock);
3781
3782 /* Minor sanity check */
3783 if (src->ro_rt != NULL && rt_key(src->ro_rt)->sa_family != AF_INET)
3784 panic("%s: wrong or corrupted route: %p", __func__, src);
3785
3786 route_copyout(dst, src, sizeof (*dst));
3787
3788 lck_mtx_unlock(&ifp->if_cached_route_lock);
3789}
3790
3791static void
3792ip_fwd_route_copyin(struct ifnet *ifp, struct route *src)
3793{
3794 struct route *dst = &ifp->if_fwd_route;
3795
3796 lck_mtx_lock_spin(&ifp->if_cached_route_lock);
3797 lck_mtx_convert_spin(&ifp->if_cached_route_lock);
3798
3799 /* Minor sanity check */
3800 if (src->ro_rt != NULL && rt_key(src->ro_rt)->sa_family != AF_INET)
3801 panic("%s: wrong or corrupted route: %p", __func__, src);
3802
3803 if (ifp->if_fwd_cacheok)
3804 route_copyin(src, dst, sizeof (*src));
3805
3806 lck_mtx_unlock(&ifp->if_cached_route_lock);
3807}
3808
3809/*
3810 * Forward a packet. If some error occurs return the sender
3811 * an icmp packet. Note we can't always generate a meaningful
3812 * icmp message because icmp doesn't have a large enough repertoire
3813 * of codes and types.
3814 *
3815 * If not forwarding, just drop the packet. This could be confusing
3816 * if ipforwarding was zero but some routing protocol was advancing
3817 * us as a gateway to somewhere. However, we must let the routing
3818 * protocol deal with that.
3819 *
3820 * The srcrt parameter indicates whether the packet is being forwarded
3821 * via a source route.
3822 */
3823static void
3824ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop)
3825{
3826#if !IPFIREWALL
3827#pragma unused(next_hop)
3828#endif
3829 struct ip *ip = mtod(m, struct ip *);
3830 struct sockaddr_in *sin;
3831 struct rtentry *rt;
3832 struct route fwd_rt;
3833 int error, type = 0, code = 0;
3834 struct mbuf *mcopy;
3835 n_long dest;
3836 struct in_addr pkt_dst;
3837 u_int32_t nextmtu = 0, len;
3838 struct ip_out_args ipoa;
3839 struct ifnet *rcvifp = m->m_pkthdr.rcvif;
3840
3841 bzero(&ipoa, sizeof(ipoa));
3842 ipoa.ipoa_boundif = IFSCOPE_NONE;
3843 ipoa.ipoa_sotc = SO_TC_UNSPEC;
3844 ipoa.ipoa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
3845
3846#if IPSEC
3847 struct secpolicy *sp = NULL;
3848 int ipsecerror;
3849#endif /* IPSEC */
3850#if PF
3851 struct pf_mtag *pf_mtag;
3852#endif /* PF */
3853
3854 dest = 0;
3855#if IPFIREWALL
3856 /*
3857 * Cache the destination address of the packet; this may be
3858 * changed by use of 'ipfw fwd'.
3859 */
3860 pkt_dst = ((next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst);
3861#else /* !IPFIREWALL */
3862 pkt_dst = ip->ip_dst;
3863#endif /* !IPFIREWALL */
3864
3865#if DIAGNOSTIC
3866 if (ipprintfs)
3867 printf("forward: src %lx dst %lx ttl %x\n",
3868 (u_int32_t)ip->ip_src.s_addr, (u_int32_t)pkt_dst.s_addr,
3869 ip->ip_ttl);
3870#endif
3871
3872 if (m->m_flags & (M_BCAST|M_MCAST) || !in_canforward(pkt_dst)) {
3873 OSAddAtomic(1, &ipstat.ips_cantforward);
3874 m_freem(m);
3875 return;
3876 }
3877#if IPSTEALTH
3878 if (!ipstealth) {
3879#endif /* IPSTEALTH */
3880 if (ip->ip_ttl <= IPTTLDEC) {
3881 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
3882 dest, 0);
3883 return;
3884 }
3885#if IPSTEALTH
3886 }
3887#endif /* IPSTEALTH */
3888
3889#if PF
3890 pf_mtag = pf_find_mtag(m);
3891 if (pf_mtag != NULL && pf_mtag->pftag_rtableid != IFSCOPE_NONE) {
3892 ipoa.ipoa_boundif = pf_mtag->pftag_rtableid;
3893 ipoa.ipoa_flags |= IPOAF_BOUND_IF;
3894 }
3895#endif /* PF */
3896
3897 ip_fwd_route_copyout(rcvifp, &fwd_rt);
3898
3899 sin = SIN(&fwd_rt.ro_dst);
3900 if (ROUTE_UNUSABLE(&fwd_rt) || pkt_dst.s_addr != sin->sin_addr.s_addr) {
3901 ROUTE_RELEASE(&fwd_rt);
3902
3903 sin->sin_family = AF_INET;
3904 sin->sin_len = sizeof (*sin);
3905 sin->sin_addr = pkt_dst;
3906
3907 rtalloc_scoped_ign(&fwd_rt, RTF_PRCLONING, ipoa.ipoa_boundif);
3908 if (fwd_rt.ro_rt == NULL) {
3909 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
3910 goto done;
3911 }
3912 }
3913 rt = fwd_rt.ro_rt;
3914
3915 /*
3916 * Save the IP header and at most 8 bytes of the payload,
3917 * in case we need to generate an ICMP message to the src.
3918 *
3919 * We don't use m_copy() because it might return a reference
3920 * to a shared cluster. Both this function and ip_output()
3921 * assume exclusive access to the IP header in `m', so any
3922 * data in a cluster may change before we reach icmp_error().
3923 */
3924 MGET(mcopy, M_DONTWAIT, m->m_type);
3925 if (mcopy != NULL) {
3926 M_COPY_PKTHDR(mcopy, m);
3927 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
3928 (int)ip->ip_len);
3929 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
3930 }
3931
3932#if IPSTEALTH
3933 if (!ipstealth) {
3934#endif /* IPSTEALTH */
3935 ip->ip_ttl -= IPTTLDEC;
3936#if IPSTEALTH
3937 }
3938#endif /* IPSTEALTH */
3939
3940 /*
3941 * If forwarding packet using same interface that it came in on,
3942 * perhaps should send a redirect to sender to shortcut a hop.
3943 * Only send redirect if source is sending directly to us,
3944 * and if packet was not source routed (or has any options).
3945 * Also, don't send redirect if forwarding using a default route
3946 * or a route modified by a redirect.
3947 */
3948 RT_LOCK_SPIN(rt);
3949 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
3950 !(rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) &&
3951 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
3952 ipsendredirects && !srcrt && rt->rt_ifa != NULL) {
3953 struct in_ifaddr *ia = (struct in_ifaddr *)rt->rt_ifa;
3954 u_int32_t src = ntohl(ip->ip_src.s_addr);
3955
3956 /* Become a regular mutex */
3957 RT_CONVERT_LOCK(rt);
3958 IFA_LOCK_SPIN(&ia->ia_ifa);
3959 if ((src & ia->ia_subnetmask) == ia->ia_subnet) {
3960 if (rt->rt_flags & RTF_GATEWAY)
3961 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
3962 else
3963 dest = pkt_dst.s_addr;
3964 /*
3965 * Router requirements says to only send
3966 * host redirects.
3967 */
3968 type = ICMP_REDIRECT;
3969 code = ICMP_REDIRECT_HOST;
3970#if DIAGNOSTIC
3971 if (ipprintfs)
3972 printf("redirect (%d) to %lx\n", code,
3973 (u_int32_t)dest);
3974#endif
3975 }
3976 IFA_UNLOCK(&ia->ia_ifa);
3977 }
3978 RT_UNLOCK(rt);
3979
3980#if IPFIREWALL
3981 if (next_hop != NULL) {
3982 /* Pass IPFORWARD info if available */
3983 struct m_tag *tag;
3984 struct ip_fwd_tag *ipfwd_tag;
3985
3986 tag = m_tag_create(KERNEL_MODULE_TAG_ID,
3987 KERNEL_TAG_TYPE_IPFORWARD,
3988 sizeof (*ipfwd_tag), M_NOWAIT, m);
3989 if (tag == NULL) {
3990 error = ENOBUFS;
3991 m_freem(m);
3992 goto done;
3993 }
3994
3995 ipfwd_tag = (struct ip_fwd_tag *)(tag+1);
3996 ipfwd_tag->next_hop = next_hop;
3997
3998 m_tag_prepend(m, tag);
3999 }
4000#endif /* IPFIREWALL */
4001
4002 /* Mark this packet as being forwarded from another interface */
4003 m->m_pkthdr.pkt_flags |= PKTF_FORWARDED;
4004 len = m_pktlen(m);
4005
4006 error = ip_output(m, NULL, &fwd_rt, IP_FORWARDING | IP_OUTARGS,
4007 NULL, &ipoa);
4008
4009 /* Refresh rt since the route could have changed while in IP */
4010 rt = fwd_rt.ro_rt;
4011
4012 if (error != 0) {
4013 OSAddAtomic(1, &ipstat.ips_cantforward);
4014 } else {
4015 /*
4016 * Increment stats on the source interface; the ones
4017 * for destination interface has been taken care of
4018 * during output above by virtue of PKTF_FORWARDED.
4019 */
4020 rcvifp->if_fpackets++;
4021 rcvifp->if_fbytes += len;
4022
4023 OSAddAtomic(1, &ipstat.ips_forward);
4024 if (type != 0) {
4025 OSAddAtomic(1, &ipstat.ips_redirectsent);
4026 } else {
4027 if (mcopy != NULL) {
4028 /*
4029 * If we didn't have to go thru ipflow and
4030 * the packet was successfully consumed by
4031 * ip_output, the mcopy is rather a waste;
4032 * this could be further optimized.
4033 */
4034 m_freem(mcopy);
4035 }
4036 goto done;
4037 }
4038 }
4039 if (mcopy == NULL)
4040 goto done;
4041
4042 switch (error) {
4043 case 0: /* forwarded, but need redirect */
4044 /* type, code set above */
4045 break;
4046
4047 case ENETUNREACH: /* shouldn't happen, checked above */
4048 case EHOSTUNREACH:
4049 case ENETDOWN:
4050 case EHOSTDOWN:
4051 default:
4052 type = ICMP_UNREACH;
4053 code = ICMP_UNREACH_HOST;
4054 break;
4055
4056 case EMSGSIZE:
4057 type = ICMP_UNREACH;
4058 code = ICMP_UNREACH_NEEDFRAG;
4059
4060 if (rt == NULL) {
4061 break;
4062 } else {
4063 RT_LOCK_SPIN(rt);
4064 if (rt->rt_ifp != NULL)
4065 nextmtu = rt->rt_ifp->if_mtu;
4066 RT_UNLOCK(rt);
4067 }
4068#ifdef IPSEC
4069 if (ipsec_bypass)
4070 break;
4071
4072 /*
4073 * If the packet is routed over IPsec tunnel, tell the
4074 * originator the tunnel MTU.
4075 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
4076 * XXX quickhack!!!
4077 */
4078 sp = ipsec4_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
4079 IP_FORWARDING, &ipsecerror);
4080
4081 if (sp == NULL)
4082 break;
4083
4084 /*
4085 * find the correct route for outer IPv4
4086 * header, compute tunnel MTU.
4087 */
4088 nextmtu = 0;
4089
4090 if (sp->req != NULL &&
4091 sp->req->saidx.mode == IPSEC_MODE_TUNNEL) {
4092 struct secasindex saidx;
4093 struct secasvar *sav;
4094 struct route *ro;
4095 struct ip *ipm;
4096 int ipsechdr;
4097
4098 /* count IPsec header size */
4099 ipsechdr = ipsec_hdrsiz(sp);
4100
4101 ipm = mtod(mcopy, struct ip *);
4102 bcopy(&sp->req->saidx, &saidx, sizeof (saidx));
4103 saidx.mode = sp->req->saidx.mode;
4104 saidx.reqid = sp->req->saidx.reqid;
4105 sin = SIN(&saidx.src);
4106 if (sin->sin_len == 0) {
4107 sin->sin_len = sizeof (*sin);
4108 sin->sin_family = AF_INET;
4109 sin->sin_port = IPSEC_PORT_ANY;
4110 bcopy(&ipm->ip_src, &sin->sin_addr,
4111 sizeof (sin->sin_addr));
4112 }
4113 sin = SIN(&saidx.dst);
4114 if (sin->sin_len == 0) {
4115 sin->sin_len = sizeof (*sin);
4116 sin->sin_family = AF_INET;
4117 sin->sin_port = IPSEC_PORT_ANY;
4118 bcopy(&ipm->ip_dst, &sin->sin_addr,
4119 sizeof (sin->sin_addr));
4120 }
4121 sav = key_allocsa_policy(&saidx);
4122 if (sav != NULL) {
4123 lck_mtx_lock(sadb_mutex);
4124 if (sav->sah != NULL) {
4125 ro = (struct route *)&sav->sah->sa_route;
4126 if (ro->ro_rt != NULL) {
4127 RT_LOCK(ro->ro_rt);
4128 if (ro->ro_rt->rt_ifp != NULL) {
4129 nextmtu = ro->ro_rt->
4130 rt_ifp->if_mtu;
4131 nextmtu -= ipsechdr;
4132 }
4133 RT_UNLOCK(ro->ro_rt);
4134 }
4135 }
4136 key_freesav(sav, KEY_SADB_LOCKED);
4137 lck_mtx_unlock(sadb_mutex);
4138 }
4139 }
4140 key_freesp(sp, KEY_SADB_UNLOCKED);
4141#endif /* IPSEC */
4142 break;
4143
4144 case ENOBUFS:
4145 /*
4146 * A router should not generate ICMP_SOURCEQUENCH as
4147 * required in RFC1812 Requirements for IP Version 4 Routers.
4148 * Source quench could be a big problem under DoS attacks,
4149 * or if the underlying interface is rate-limited.
4150 * Those who need source quench packets may re-enable them
4151 * via the net.inet.ip.sendsourcequench sysctl.
4152 */
4153 if (ip_sendsourcequench == 0) {
4154 m_freem(mcopy);
4155 goto done;
4156 } else {
4157 type = ICMP_SOURCEQUENCH;
4158 code = 0;
4159 }
4160 break;
4161
4162 case EACCES: /* ipfw denied packet */
4163 m_freem(mcopy);
4164 goto done;
4165 }
4166
4167 if (type == ICMP_UNREACH && code == ICMP_UNREACH_NEEDFRAG)
4168 OSAddAtomic(1, &ipstat.ips_cantfrag);
4169
4170 icmp_error(mcopy, type, code, dest, nextmtu);
4171done:
4172 ip_fwd_route_copyin(rcvifp, &fwd_rt);
4173}
4174
4175int
4176ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
4177 struct mbuf *m)
4178{
4179 *mp = NULL;
4180 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
4181 struct timeval tv;
4182
4183 getmicrotime(&tv);
4184 mp = sbcreatecontrol_mbuf((caddr_t)&tv, sizeof (tv),
4185 SCM_TIMESTAMP, SOL_SOCKET, mp);
4186 if (*mp == NULL) {
4187 goto no_mbufs;
4188 }
4189 }
4190 if (inp->inp_socket->so_options & SO_TIMESTAMP_MONOTONIC) {
4191 uint64_t time;
4192
4193 time = mach_absolute_time();
4194 mp = sbcreatecontrol_mbuf((caddr_t)&time, sizeof (time),
4195 SCM_TIMESTAMP_MONOTONIC, SOL_SOCKET, mp);
4196 if (*mp == NULL) {
4197 goto no_mbufs;
4198 }
4199 }
4200 if (inp->inp_socket->so_options & SO_TIMESTAMP_CONTINUOUS) {
4201 uint64_t time;
4202
4203 time = mach_continuous_time();
4204 mp = sbcreatecontrol_mbuf((caddr_t)&time, sizeof (time),
4205 SCM_TIMESTAMP_CONTINUOUS, SOL_SOCKET, mp);
4206 if (*mp == NULL) {
4207 goto no_mbufs;
4208 }
4209 }
4210 if (inp->inp_flags & INP_RECVDSTADDR) {
4211 mp = sbcreatecontrol_mbuf((caddr_t)&ip->ip_dst,
4212 sizeof (struct in_addr), IP_RECVDSTADDR, IPPROTO_IP, mp);
4213 if (*mp == NULL) {
4214 goto no_mbufs;
4215 }
4216 }
4217#ifdef notyet
4218 /*
4219 * XXX
4220 * Moving these out of udp_input() made them even more broken
4221 * than they already were.
4222 */
4223 /* options were tossed already */
4224 if (inp->inp_flags & INP_RECVOPTS) {
4225 mp = sbcreatecontrol_mbuf((caddr_t)opts_deleted_above,
4226 sizeof (struct in_addr), IP_RECVOPTS, IPPROTO_IP, mp);
4227 if (*mp == NULL) {
4228 goto no_mbufs;
4229 }
4230 }
4231 /* ip_srcroute doesn't do what we want here, need to fix */
4232 if (inp->inp_flags & INP_RECVRETOPTS) {
4233 mp = sbcreatecontrol_mbuf((caddr_t)ip_srcroute(),
4234 sizeof (struct in_addr), IP_RECVRETOPTS, IPPROTO_IP, mp);
4235 if (*mp == NULL) {
4236 goto no_mbufs;
4237 }
4238 }
4239#endif /* notyet */
4240 if (inp->inp_flags & INP_RECVIF) {
4241 struct ifnet *ifp;
4242 uint8_t sdlbuf[SOCK_MAXADDRLEN + 1];
4243 struct sockaddr_dl *sdl2 = SDL(&sdlbuf);
4244
4245 /*
4246 * Make sure to accomodate the largest possible
4247 * size of SA(if_lladdr)->sa_len.
4248 */
4249 _CASSERT(sizeof (sdlbuf) == (SOCK_MAXADDRLEN + 1));
4250
4251 ifnet_head_lock_shared();
4252 if ((ifp = m->m_pkthdr.rcvif) != NULL &&
4253 ifp->if_index && (ifp->if_index <= if_index)) {
4254 struct ifaddr *ifa = ifnet_addrs[ifp->if_index - 1];
4255 struct sockaddr_dl *sdp;
4256
4257 if (!ifa || !ifa->ifa_addr)
4258 goto makedummy;
4259
4260 IFA_LOCK_SPIN(ifa);
4261 sdp = SDL(ifa->ifa_addr);
4262 /*
4263 * Change our mind and don't try copy.
4264 */
4265 if (sdp->sdl_family != AF_LINK) {
4266 IFA_UNLOCK(ifa);
4267 goto makedummy;
4268 }
4269 /* the above _CASSERT ensures sdl_len fits in sdlbuf */
4270 bcopy(sdp, sdl2, sdp->sdl_len);
4271 IFA_UNLOCK(ifa);
4272 } else {
4273makedummy:
4274 sdl2->sdl_len =
4275 offsetof(struct sockaddr_dl, sdl_data[0]);
4276 sdl2->sdl_family = AF_LINK;
4277 sdl2->sdl_index = 0;
4278 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
4279 }
4280 ifnet_head_done();
4281 mp = sbcreatecontrol_mbuf((caddr_t)sdl2, sdl2->sdl_len,
4282 IP_RECVIF, IPPROTO_IP, mp);
4283 if (*mp == NULL) {
4284 goto no_mbufs;
4285 }
4286 }
4287 if (inp->inp_flags & INP_RECVTTL) {
4288 mp = sbcreatecontrol_mbuf((caddr_t)&ip->ip_ttl,
4289 sizeof (ip->ip_ttl), IP_RECVTTL, IPPROTO_IP, mp);
4290 if (*mp == NULL) {
4291 goto no_mbufs;
4292 }
4293 }
4294 if (inp->inp_socket->so_flags & SOF_RECV_TRAFFIC_CLASS) {
4295 int tc = m_get_traffic_class(m);
4296
4297 mp = sbcreatecontrol_mbuf((caddr_t)&tc, sizeof (tc),
4298 SO_TRAFFIC_CLASS, SOL_SOCKET, mp);
4299 if (*mp == NULL) {
4300 goto no_mbufs;
4301 }
4302 }
4303 if (inp->inp_flags & INP_PKTINFO) {
4304 struct in_pktinfo pi;
4305
4306 bzero(&pi, sizeof (struct in_pktinfo));
4307 bcopy(&ip->ip_dst, &pi.ipi_addr, sizeof (struct in_addr));
4308 pi.ipi_ifindex = (m != NULL && m->m_pkthdr.rcvif != NULL) ?
4309 m->m_pkthdr.rcvif->if_index : 0;
4310
4311 mp = sbcreatecontrol_mbuf((caddr_t)&pi,
4312 sizeof (struct in_pktinfo), IP_RECVPKTINFO, IPPROTO_IP, mp);
4313 if (*mp == NULL) {
4314 goto no_mbufs;
4315 }
4316 }
4317 if (inp->inp_flags & INP_RECVTOS) {
4318 mp = sbcreatecontrol_mbuf((caddr_t)&ip->ip_tos,
4319 sizeof(u_char), IP_RECVTOS, IPPROTO_IP, mp);
4320 if (*mp == NULL) {
4321 goto no_mbufs;
4322 }
4323 }
4324 return (0);
4325
4326no_mbufs:
4327 ipstat.ips_pktdropcntrl++;
4328 return (ENOBUFS);
4329}
4330
4331static inline u_short
4332ip_cksum(struct mbuf *m, int hlen)
4333{
4334 u_short sum;
4335
4336 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
4337 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
4338 } else if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
4339 !(m->m_pkthdr.pkt_flags & PKTF_LOOP)) {
4340 /*
4341 * The packet arrived on an interface which isn't capable
4342 * of performing IP header checksum; compute it now.
4343 */
4344 sum = ip_cksum_hdr_in(m, hlen);
4345 } else {
4346 sum = 0;
4347 m->m_pkthdr.csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR |
4348 CSUM_IP_CHECKED | CSUM_IP_VALID);
4349 m->m_pkthdr.csum_data = 0xffff;
4350 }
4351
4352 if (sum != 0)
4353 OSAddAtomic(1, &ipstat.ips_badsum);
4354
4355 return (sum);
4356}
4357
4358static int
4359ip_getstat SYSCTL_HANDLER_ARGS
4360{
4361#pragma unused(oidp, arg1, arg2)
4362 if (req->oldptr == USER_ADDR_NULL)
4363 req->oldlen = (size_t)sizeof (struct ipstat);
4364
4365 return (SYSCTL_OUT(req, &ipstat, MIN(sizeof (ipstat), req->oldlen)));
4366}
4367
4368void
4369ip_setsrcifaddr_info(struct mbuf *m, uint32_t src_idx, struct in_ifaddr *ia)
4370{
4371 VERIFY(m->m_flags & M_PKTHDR);
4372
4373 /*
4374 * If the source ifaddr is specified, pick up the information
4375 * from there; otherwise just grab the passed-in ifindex as the
4376 * caller may not have the ifaddr available.
4377 */
4378 if (ia != NULL) {
4379 m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
4380 m->m_pkthdr.src_ifindex = ia->ia_ifp->if_index;
4381 } else {
4382 m->m_pkthdr.src_ifindex = src_idx;
4383 if (src_idx != 0)
4384 m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
4385 }
4386}
4387
4388void
4389ip_setdstifaddr_info(struct mbuf *m, uint32_t dst_idx, struct in_ifaddr *ia)
4390{
4391 VERIFY(m->m_flags & M_PKTHDR);
4392
4393 /*
4394 * If the destination ifaddr is specified, pick up the information
4395 * from there; otherwise just grab the passed-in ifindex as the
4396 * caller may not have the ifaddr available.
4397 */
4398 if (ia != NULL) {
4399 m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
4400 m->m_pkthdr.dst_ifindex = ia->ia_ifp->if_index;
4401 } else {
4402 m->m_pkthdr.dst_ifindex = dst_idx;
4403 if (dst_idx != 0)
4404 m->m_pkthdr.pkt_flags |= PKTF_IFAINFO;
4405 }
4406}
4407
4408int
4409ip_getsrcifaddr_info(struct mbuf *m, uint32_t *src_idx, uint32_t *iaf)
4410{
4411 VERIFY(m->m_flags & M_PKTHDR);
4412
4413 if (!(m->m_pkthdr.pkt_flags & PKTF_IFAINFO))
4414 return (-1);
4415
4416 if (src_idx != NULL)
4417 *src_idx = m->m_pkthdr.src_ifindex;
4418
4419 if (iaf != NULL)
4420 *iaf = 0;
4421
4422 return (0);
4423}
4424
4425int
4426ip_getdstifaddr_info(struct mbuf *m, uint32_t *dst_idx, uint32_t *iaf)
4427{
4428 VERIFY(m->m_flags & M_PKTHDR);
4429
4430 if (!(m->m_pkthdr.pkt_flags & PKTF_IFAINFO))
4431 return (-1);
4432
4433 if (dst_idx != NULL)
4434 *dst_idx = m->m_pkthdr.dst_ifindex;
4435
4436 if (iaf != NULL)
4437 *iaf = 0;
4438
4439 return (0);
4440}
4441
4442/*
4443 * Protocol input handler for IPPROTO_GRE.
4444 */
4445void
4446gre_input(struct mbuf *m, int off)
4447{
4448 gre_input_func_t fn = gre_input_func;
4449
4450 /*
4451 * If there is a registered GRE input handler, pass mbuf to it.
4452 */
4453 if (fn != NULL) {
4454 lck_mtx_unlock(inet_domain_mutex);
4455 m = fn(m, off, (mtod(m, struct ip *))->ip_p);
4456 lck_mtx_lock(inet_domain_mutex);
4457 }
4458
4459 /*
4460 * If no matching tunnel that is up is found, we inject
4461 * the mbuf to raw ip socket to see if anyone picks it up.
4462 */
4463 if (m != NULL)
4464 rip_input(m, off);
4465}
4466
4467/*
4468 * Private KPI for PPP/PPTP.
4469 */
4470int
4471ip_gre_register_input(gre_input_func_t fn)
4472{
4473 lck_mtx_lock(inet_domain_mutex);
4474 gre_input_func = fn;
4475 lck_mtx_unlock(inet_domain_mutex);
4476
4477 return (0);
4478}
4479
4480#if (DEBUG || DEVELOPMENT)
4481static int
4482sysctl_reset_ip_input_stats SYSCTL_HANDLER_ARGS
4483{
4484#pragma unused(arg1, arg2)
4485 int error, i;
4486
4487 i = ip_input_measure;
4488 error = sysctl_handle_int(oidp, &i, 0, req);
4489 if (error || req->newptr == USER_ADDR_NULL)
4490 goto done;
4491 /* impose bounds */
4492 if (i < 0 || i > 1) {
4493 error = EINVAL;
4494 goto done;
4495 }
4496 if (ip_input_measure != i && i == 1) {
4497 net_perf_initialize(&net_perf, ip_input_measure_bins);
4498 }
4499 ip_input_measure = i;
4500done:
4501 return (error);
4502}
4503
4504static int
4505sysctl_ip_input_measure_bins SYSCTL_HANDLER_ARGS
4506{
4507#pragma unused(arg1, arg2)
4508 int error;
4509 uint64_t i;
4510
4511 i = ip_input_measure_bins;
4512 error = sysctl_handle_quad(oidp, &i, 0, req);
4513 if (error || req->newptr == USER_ADDR_NULL)
4514 goto done;
4515 /* validate data */
4516 if (!net_perf_validate_bins(i)) {
4517 error = EINVAL;
4518 goto done;
4519 }
4520 ip_input_measure_bins = i;
4521done:
4522 return (error);
4523}
4524
4525static int
4526sysctl_ip_input_getperf SYSCTL_HANDLER_ARGS
4527{
4528#pragma unused(oidp, arg1, arg2)
4529 if (req->oldptr == USER_ADDR_NULL)
4530 req->oldlen = (size_t)sizeof (struct ipstat);
4531
4532 return (SYSCTL_OUT(req, &net_perf, MIN(sizeof (net_perf), req->oldlen)));
4533}
4534#endif /* (DEBUG || DEVELOPMENT) */
4535