]> git.saurik.com Git - apple/xnu.git/blob - bsd/netinet/ip_output.c
projects / apple / xnu.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
xnu-3247.10.11.tar.gz
[apple/xnu.git] / bsd / netinet / ip_output.c
1 /*
2 * Copyright (c) 2000-2014 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, 1990, 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_output.c 8.3 (Berkeley) 1/21/94
61 */
62 /*
63 * NOTICE: This file was modified by SPARTA, Inc. in 2005 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/kernel.h>
74 #include <sys/malloc.h>
75 #include <sys/mbuf.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <kern/locks.h>
80 #include <sys/sysctl.h>
81 #include <sys/mcache.h>
82 #include <sys/kdebug.h>
83
84 #include <machine/endian.h>
85 #include <pexpert/pexpert.h>
86 #include <mach/sdt.h>
87
88 #include <libkern/OSAtomic.h>
89 #include <libkern/OSByteOrder.h>
90
91 #include <net/if.h>
92 #include <net/if_dl.h>
93 #include <net/if_types.h>
94 #include <net/route.h>
95 #include <net/ntstat.h>
96 #include <net/net_osdep.h>
97 #include <net/dlil.h>
98 #include <net/net_perf.h>
99
100 #include <netinet/in.h>
101 #include <netinet/in_systm.h>
102 #include <netinet/ip.h>
103 #include <netinet/in_pcb.h>
104 #include <netinet/in_var.h>
105 #include <netinet/ip_var.h>
106 #include <netinet/kpi_ipfilter_var.h>
107
108 #if CONFIG_MACF_NET
109 #include <security/mac_framework.h>
110 #endif /* CONFIG_MACF_NET */
111
112 #define DBG_LAYER_BEG NETDBG_CODE(DBG_NETIP, 1)
113 #define DBG_LAYER_END NETDBG_CODE(DBG_NETIP, 3)
114 #define DBG_FNC_IP_OUTPUT NETDBG_CODE(DBG_NETIP, (1 << 8) | 1)
115 #define DBG_FNC_IPSEC4_OUTPUT NETDBG_CODE(DBG_NETIP, (2 << 8) | 1)
116
117 #if IPSEC
118 #include <netinet6/ipsec.h>
119 #include <netkey/key.h>
120 #if IPSEC_DEBUG
121 #include <netkey/key_debug.h>
122 #else
123 #define KEYDEBUG(lev, arg)
124 #endif
125 #endif /* IPSEC */
126
127 #if NECP
128 #include <net/necp.h>
129 #endif /* NECP */
130
131 #if IPFIREWALL
132 #include <netinet/ip_fw.h>
133 #if IPDIVERT
134 #include <netinet/ip_divert.h>
135 #endif /* IPDIVERT */
136 #endif /* IPFIREWALL */
137
138 #if DUMMYNET
139 #include <netinet/ip_dummynet.h>
140 #endif
141
142 #if PF
143 #include <net/pfvar.h>
144 #endif /* PF */
145
146 #if IPFIREWALL_FORWARD && IPFIREWALL_FORWARD_DEBUG
147 #define print_ip(a) \
148 printf("%ld.%ld.%ld.%ld", (ntohl(a.s_addr) >> 24) & 0xFF, \
149 (ntohl(a.s_addr) >> 16) & 0xFF, \
150 (ntohl(a.s_addr) >> 8) & 0xFF, \
151 (ntohl(a.s_addr)) & 0xFF);
152 #endif /* IPFIREWALL_FORWARD && IPFIREWALL_FORWARD_DEBUG */
153
154 u_short ip_id;
155
156 static int sysctl_reset_ip_output_stats SYSCTL_HANDLER_ARGS;
157 static int sysctl_ip_output_measure_bins SYSCTL_HANDLER_ARGS;
158 static int sysctl_ip_output_getperf SYSCTL_HANDLER_ARGS;
159 static void ip_out_cksum_stats(int, u_int32_t);
160 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
161 static int ip_optcopy(struct ip *, struct ip *);
162 static int ip_pcbopts(int, struct mbuf **, struct mbuf *);
163 static void imo_trace(struct ip_moptions *, int);
164 static void ip_mloopback(struct ifnet *, struct ifnet *, struct mbuf *,
165 struct sockaddr_in *, int);
166 static struct ifaddr *in_selectsrcif(struct ip *, struct route *, unsigned int);
167
168 extern struct ip_linklocal_stat ip_linklocal_stat;
169
170 /* temporary: for testing */
171 #if IPSEC
172 extern int ipsec_bypass;
173 #endif
174
175 static int ip_maxchainsent = 0;
176 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxchainsent,
177 CTLFLAG_RW | CTLFLAG_LOCKED, &ip_maxchainsent, 0,
178 "use dlil_output_list");
179 #if DEBUG
180 static int forge_ce = 0;
181 SYSCTL_INT(_net_inet_ip, OID_AUTO, forge_ce,
182 CTLFLAG_RW | CTLFLAG_LOCKED, &forge_ce, 0,
183 "Forge ECN CE");
184 #endif /* DEBUG */
185
186 static int ip_select_srcif_debug = 0;
187 SYSCTL_INT(_net_inet_ip, OID_AUTO, select_srcif_debug,
188 CTLFLAG_RW | CTLFLAG_LOCKED, &ip_select_srcif_debug, 0,
189 "log source interface selection debug info");
190
191 static int ip_output_measure = 0;
192 SYSCTL_PROC(_net_inet_ip, OID_AUTO, output_perf,
193 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
194 &ip_output_measure, 0, sysctl_reset_ip_output_stats, "I",
195 "Do time measurement");
196
197 static uint64_t ip_output_measure_bins = 0;
198 SYSCTL_PROC(_net_inet_ip, OID_AUTO, output_perf_bins,
199 CTLTYPE_QUAD | CTLFLAG_RW | CTLFLAG_LOCKED, &ip_output_measure_bins, 0,
200 sysctl_ip_output_measure_bins, "I",
201 "bins for chaining performance data histogram");
202
203 static net_perf_t net_perf;
204 SYSCTL_PROC(_net_inet_ip, OID_AUTO, output_perf_data,
205 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED,
206 0, 0, sysctl_ip_output_getperf, "S,net_perf",
207 "IP output performance data (struct net_perf, net/net_perf.h)");
208
209 #define IMO_TRACE_HIST_SIZE 32 /* size of trace history */
210
211 /* For gdb */
212 __private_extern__ unsigned int imo_trace_hist_size = IMO_TRACE_HIST_SIZE;
213
214 struct ip_moptions_dbg {
215 struct ip_moptions imo; /* ip_moptions */
216 u_int16_t imo_refhold_cnt; /* # of IMO_ADDREF */
217 u_int16_t imo_refrele_cnt; /* # of IMO_REMREF */
218 /*
219 * Alloc and free callers.
220 */
221 ctrace_t imo_alloc;
222 ctrace_t imo_free;
223 /*
224 * Circular lists of IMO_ADDREF and IMO_REMREF callers.
225 */
226 ctrace_t imo_refhold[IMO_TRACE_HIST_SIZE];
227 ctrace_t imo_refrele[IMO_TRACE_HIST_SIZE];
228 };
229
230 #if DEBUG
231 static unsigned int imo_debug = 1; /* debugging (enabled) */
232 #else
233 static unsigned int imo_debug; /* debugging (disabled) */
234 #endif /* !DEBUG */
235 static unsigned int imo_size; /* size of zone element */
236 static struct zone *imo_zone; /* zone for ip_moptions */
237
238 #define IMO_ZONE_MAX 64 /* maximum elements in zone */
239 #define IMO_ZONE_NAME "ip_moptions" /* zone name */
240
241 /*
242 * IP output. The packet in mbuf chain m contains a skeletal IP
243 * header (with len, off, ttl, proto, tos, src, dst).
244 * The mbuf chain containing the packet will be freed.
245 * The mbuf opt, if present, will not be freed.
246 */
247 int
248 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro, int flags,
249 struct ip_moptions *imo, struct ip_out_args *ipoa)
250 {
251 return (ip_output_list(m0, 0, opt, ro, flags, imo, ipoa));
252 }
253
254 /*
255 * IP output. The packet in mbuf chain m contains a skeletal IP
256 * header (with len, off, ttl, proto, tos, src, dst).
257 * The mbuf chain containing the packet will be freed.
258 * The mbuf opt, if present, will not be freed.
259 *
260 * Route ro MUST be non-NULL; if ro->ro_rt is valid, route lookup would be
261 * skipped and ro->ro_rt would be used. Otherwise the result of route
262 * lookup is stored in ro->ro_rt.
263 *
264 * In the IP forwarding case, the packet will arrive with options already
265 * inserted, so must have a NULL opt pointer.
266 */
267 int
268 ip_output_list(struct mbuf *m0, int packetchain, struct mbuf *opt,
269 struct route *ro, int flags, struct ip_moptions *imo,
270 struct ip_out_args *ipoa)
271 {
272 struct ip *ip;
273 struct ifnet *ifp = NULL; /* not refcnt'd */
274 struct mbuf *m = m0, *prevnxt = NULL, **mppn = &prevnxt;
275 int hlen = sizeof (struct ip);
276 int len = 0, error = 0;
277 struct sockaddr_in *dst = NULL;
278 struct in_ifaddr *ia = NULL, *src_ia = NULL;
279 struct in_addr pkt_dst;
280 struct ipf_pktopts *ippo = NULL;
281 ipfilter_t inject_filter_ref = NULL;
282 struct mbuf *packetlist;
283 uint32_t sw_csum, pktcnt = 0, scnt = 0, bytecnt = 0;
284 uint32_t packets_processed = 0;
285 unsigned int ifscope = IFSCOPE_NONE;
286 struct flowadv *adv = NULL;
287 struct timeval start_tv;
288 #if IPSEC
289 struct socket *so = NULL;
290 struct secpolicy *sp = NULL;
291 #endif /* IPSEC */
292 #if NECP
293 necp_kernel_policy_result necp_result = 0;
294 necp_kernel_policy_result_parameter necp_result_parameter;
295 necp_kernel_policy_id necp_matched_policy_id = 0;
296 #endif /* NECP */
297 #if IPFIREWALL
298 int ipfwoff;
299 struct sockaddr_in *next_hop_from_ipfwd_tag = NULL;
300 #endif /* IPFIREWALL */
301 #if IPFIREWALL || DUMMYNET
302 struct m_tag *tag;
303 #endif /* IPFIREWALL || DUMMYNET */
304 #if DUMMYNET
305 struct ip_out_args saved_ipoa;
306 struct sockaddr_in dst_buf;
307 #endif /* DUMMYNET */
308 struct {
309 #if IPSEC
310 struct ipsec_output_state ipsec_state;
311 #endif /* IPSEC */
312 #if NECP
313 struct route necp_route;
314 #endif /* NECP */
315 #if IPFIREWALL || DUMMYNET
316 struct ip_fw_args args;
317 #endif /* IPFIREWALL || DUMMYNET */
318 #if IPFIREWALL_FORWARD
319 struct route sro_fwd;
320 #endif /* IPFIREWALL_FORWARD */
321 #if DUMMYNET
322 struct route saved_route;
323 #endif /* DUMMYNET */
324 struct ipf_pktopts ipf_pktopts;
325 } ipobz;
326 #define ipsec_state ipobz.ipsec_state
327 #define necp_route ipobz.necp_route
328 #define args ipobz.args
329 #define sro_fwd ipobz.sro_fwd
330 #define saved_route ipobz.saved_route
331 #define ipf_pktopts ipobz.ipf_pktopts
332 union {
333 struct {
334 boolean_t select_srcif : 1; /* set once */
335 boolean_t srcbound : 1; /* set once */
336 boolean_t nocell : 1; /* set once */
337 boolean_t isbroadcast : 1;
338 boolean_t didfilter : 1;
339 boolean_t noexpensive : 1; /* set once */
340 boolean_t awdl_unrestricted : 1; /* set once */
341 #if IPFIREWALL_FORWARD
342 boolean_t fwd_rewrite_src : 1;
343 #endif /* IPFIREWALL_FORWARD */
344 };
345 uint32_t raw;
346 } ipobf = { .raw = 0 };
347
348 #define IP_CHECK_RESTRICTIONS(_ifp, _ipobf) \
349 (((_ipobf).nocell && IFNET_IS_CELLULAR(_ifp)) || \
350 ((_ipobf).noexpensive && IFNET_IS_EXPENSIVE(_ifp)) || \
351 (!(_ipobf).awdl_unrestricted && IFNET_IS_AWDL_RESTRICTED(_ifp)))
352
353 if (ip_output_measure)
354 net_perf_start_time(&net_perf, &start_tv);
355 KERNEL_DEBUG(DBG_FNC_IP_OUTPUT | DBG_FUNC_START, 0, 0, 0, 0, 0);
356
357 VERIFY(m0->m_flags & M_PKTHDR);
358 packetlist = m0;
359
360 /* zero out {ipsec_state, args, sro_fwd, saved_route, ipf_pktops} */
361 bzero(&ipobz, sizeof (ipobz));
362 ippo = &ipf_pktopts;
363
364 #if IPFIREWALL || DUMMYNET
365 if (SLIST_EMPTY(&m0->m_pkthdr.tags))
366 goto ipfw_tags_done;
367
368 /* Grab info from mtags prepended to the chain */
369 #if DUMMYNET
370 if ((tag = m_tag_locate(m0, KERNEL_MODULE_TAG_ID,
371 KERNEL_TAG_TYPE_DUMMYNET, NULL)) != NULL) {
372 struct dn_pkt_tag *dn_tag;
373
374 dn_tag = (struct dn_pkt_tag *)(tag+1);
375 args.fwa_ipfw_rule = dn_tag->dn_ipfw_rule;
376 args.fwa_pf_rule = dn_tag->dn_pf_rule;
377 opt = NULL;
378 saved_route = dn_tag->dn_ro;
379 ro = &saved_route;
380
381 imo = NULL;
382 bcopy(&dn_tag->dn_dst, &dst_buf, sizeof (dst_buf));
383 dst = &dst_buf;
384 ifp = dn_tag->dn_ifp;
385 flags = dn_tag->dn_flags;
386 if ((dn_tag->dn_flags & IP_OUTARGS)) {
387 saved_ipoa = dn_tag->dn_ipoa;
388 ipoa = &saved_ipoa;
389 }
390
391 m_tag_delete(m0, tag);
392 }
393 #endif /* DUMMYNET */
394
395 #if IPDIVERT
396 if ((tag = m_tag_locate(m0, KERNEL_MODULE_TAG_ID,
397 KERNEL_TAG_TYPE_DIVERT, NULL)) != NULL) {
398 struct divert_tag *div_tag;
399
400 div_tag = (struct divert_tag *)(tag+1);
401 args.fwa_divert_rule = div_tag->cookie;
402
403 m_tag_delete(m0, tag);
404 }
405 #endif /* IPDIVERT */
406
407 #if IPFIREWALL
408 if ((tag = m_tag_locate(m0, KERNEL_MODULE_TAG_ID,
409 KERNEL_TAG_TYPE_IPFORWARD, NULL)) != NULL) {
410 struct ip_fwd_tag *ipfwd_tag;
411
412 ipfwd_tag = (struct ip_fwd_tag *)(tag+1);
413 next_hop_from_ipfwd_tag = ipfwd_tag->next_hop;
414
415 m_tag_delete(m0, tag);
416 }
417 #endif /* IPFIREWALL */
418
419 ipfw_tags_done:
420 #endif /* IPFIREWALL || DUMMYNET */
421
422 m = m0;
423 m->m_pkthdr.pkt_flags &= ~(PKTF_LOOP|PKTF_IFAINFO);
424
425 #if IPSEC
426 if (ipsec_bypass == 0 && !(flags & IP_NOIPSEC)) {
427 /* If packet is bound to an interface, check bound policies */
428 if ((flags & IP_OUTARGS) && (ipoa != NULL) &&
429 (ipoa->ipoa_flags & IPOAF_BOUND_IF) &&
430 ipoa->ipoa_boundif != IFSCOPE_NONE) {
431 if (ipsec4_getpolicybyinterface(m, IPSEC_DIR_OUTBOUND,
432 &flags, ipoa, &sp) != 0)
433 goto bad;
434 }
435 }
436 #endif /* IPSEC */
437
438 VERIFY(ro != NULL);
439
440 if (ip_doscopedroute && (flags & IP_OUTARGS)) {
441 /*
442 * In the forwarding case, only the ifscope value is used,
443 * as source interface selection doesn't take place.
444 */
445 if ((ipobf.select_srcif = (!(flags & IP_FORWARDING) &&
446 (ipoa->ipoa_flags & IPOAF_SELECT_SRCIF)))) {
447 ipf_pktopts.ippo_flags |= IPPOF_SELECT_SRCIF;
448 }
449
450 if ((ipoa->ipoa_flags & IPOAF_BOUND_IF) &&
451 ipoa->ipoa_boundif != IFSCOPE_NONE) {
452 ifscope = ipoa->ipoa_boundif;
453 ipf_pktopts.ippo_flags |=
454 (IPPOF_BOUND_IF | (ifscope << IPPOF_SHIFT_IFSCOPE));
455 }
456
457 /* double negation needed for bool bit field */
458 ipobf.srcbound = !!(ipoa->ipoa_flags & IPOAF_BOUND_SRCADDR);
459 if (ipobf.srcbound)
460 ipf_pktopts.ippo_flags |= IPPOF_BOUND_SRCADDR;
461 } else {
462 ipobf.select_srcif = FALSE;
463 ipobf.srcbound = FALSE;
464 ifscope = IFSCOPE_NONE;
465 if (flags & IP_OUTARGS) {
466 ipoa->ipoa_boundif = IFSCOPE_NONE;
467 ipoa->ipoa_flags &= ~(IPOAF_SELECT_SRCIF |
468 IPOAF_BOUND_IF | IPOAF_BOUND_SRCADDR);
469 }
470 }
471
472 if (flags & IP_OUTARGS) {
473 if (ipoa->ipoa_flags & IPOAF_NO_CELLULAR) {
474 ipobf.nocell = TRUE;
475 ipf_pktopts.ippo_flags |= IPPOF_NO_IFT_CELLULAR;
476 }
477 if (ipoa->ipoa_flags & IPOAF_NO_EXPENSIVE) {
478 ipobf.noexpensive = TRUE;
479 ipf_pktopts.ippo_flags |= IPPOF_NO_IFF_EXPENSIVE;
480 }
481 if (ipoa->ipoa_flags & IPOAF_AWDL_UNRESTRICTED)
482 ipobf.awdl_unrestricted = TRUE;
483 adv = &ipoa->ipoa_flowadv;
484 adv->code = FADV_SUCCESS;
485 ipoa->ipoa_retflags = 0;
486 }
487
488 #if IPSEC
489 if (ipsec_bypass == 0 && !(flags & IP_NOIPSEC)) {
490 so = ipsec_getsocket(m);
491 if (so != NULL) {
492 (void) ipsec_setsocket(m, NULL);
493 }
494 }
495 #endif /* IPSEC */
496
497 #if DUMMYNET
498 if (args.fwa_ipfw_rule != NULL || args.fwa_pf_rule != NULL) {
499 /* dummynet already saw us */
500 ip = mtod(m, struct ip *);
501 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
502 pkt_dst = ip->ip_dst;
503 if (ro->ro_rt != NULL) {
504 RT_LOCK_SPIN(ro->ro_rt);
505 ia = (struct in_ifaddr *)ro->ro_rt->rt_ifa;
506 if (ia) {
507 /* Become a regular mutex */
508 RT_CONVERT_LOCK(ro->ro_rt);
509 IFA_ADDREF(&ia->ia_ifa);
510 }
511 RT_UNLOCK(ro->ro_rt);
512 }
513
514 #if IPFIREWALL
515 if (args.fwa_ipfw_rule != NULL)
516 goto skip_ipsec;
517 #endif /* IPFIREWALL */
518 if (args.fwa_pf_rule != NULL)
519 goto sendit;
520 }
521 #endif /* DUMMYNET */
522
523 loopit:
524 packets_processed++;
525 ipobf.isbroadcast = FALSE;
526 ipobf.didfilter = FALSE;
527 #if IPFIREWALL_FORWARD
528 ipobf.fwd_rewrite_src = FALSE;
529 #endif /* IPFIREWALL_FORWARD */
530
531 VERIFY(m->m_flags & M_PKTHDR);
532 /*
533 * No need to proccess packet twice if we've already seen it.
534 */
535 if (!SLIST_EMPTY(&m->m_pkthdr.tags))
536 inject_filter_ref = ipf_get_inject_filter(m);
537 else
538 inject_filter_ref = NULL;
539
540 if (opt) {
541 m = ip_insertoptions(m, opt, &len);
542 hlen = len;
543 /* Update the chain */
544 if (m != m0) {
545 if (m0 == packetlist)
546 packetlist = m;
547 m0 = m;
548 }
549 }
550 ip = mtod(m, struct ip *);
551
552 #if IPFIREWALL
553 /*
554 * rdar://8542331
555 *
556 * When dealing with a packet chain, we need to reset "next_hop"
557 * because "dst" may have been changed to the gateway address below
558 * for the previous packet of the chain. This could cause the route
559 * to be inavertandly changed to the route to the gateway address
560 * (instead of the route to the destination).
561 */
562 args.fwa_next_hop = next_hop_from_ipfwd_tag;
563 pkt_dst = args.fwa_next_hop ? args.fwa_next_hop->sin_addr : ip->ip_dst;
564 #else /* !IPFIREWALL */
565 pkt_dst = ip->ip_dst;
566 #endif /* !IPFIREWALL */
567
568 /*
569 * We must not send if the packet is destined to network zero.
570 * RFC1122 3.2.1.3 (a) and (b).
571 */
572 if (IN_ZERONET(ntohl(pkt_dst.s_addr))) {
573 error = EHOSTUNREACH;
574 goto bad;
575 }
576
577 /*
578 * Fill in IP header.
579 */
580 if (!(flags & (IP_FORWARDING|IP_RAWOUTPUT))) {
581 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2);
582 ip->ip_off &= IP_DF;
583 ip->ip_id = ip_randomid();
584 OSAddAtomic(1, &ipstat.ips_localout);
585 } else {
586 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
587 }
588
589 #if DEBUG
590 /* For debugging, we let the stack forge congestion */
591 if (forge_ce != 0 &&
592 ((ip->ip_tos & IPTOS_ECN_MASK) == IPTOS_ECN_ECT1 ||
593 (ip->ip_tos & IPTOS_ECN_MASK) == IPTOS_ECN_ECT0)) {
594 ip->ip_tos = (ip->ip_tos & ~IPTOS_ECN_MASK) | IPTOS_ECN_CE;
595 forge_ce--;
596 }
597 #endif /* DEBUG */
598
599 KERNEL_DEBUG(DBG_LAYER_BEG, ip->ip_dst.s_addr, ip->ip_src.s_addr,
600 ip->ip_p, ip->ip_off, ip->ip_len);
601
602 dst = SIN(&ro->ro_dst);
603
604 /*
605 * If there is a cached route,
606 * check that it is to the same destination
607 * and is still up. If not, free it and try again.
608 * The address family should also be checked in case of sharing the
609 * cache with IPv6.
610 */
611
612 if (ro->ro_rt != NULL) {
613 if (ROUTE_UNUSABLE(ro) && ip->ip_src.s_addr != INADDR_ANY &&
614 !(flags & (IP_ROUTETOIF | IP_FORWARDING))) {
615 src_ia = ifa_foraddr(ip->ip_src.s_addr);
616 if (src_ia == NULL) {
617 error = EADDRNOTAVAIL;
618 goto bad;
619 }
620 IFA_REMREF(&src_ia->ia_ifa);
621 src_ia = NULL;
622 }
623 /*
624 * Test rt_flags without holding rt_lock for performance
625 * reasons; if the route is down it will hopefully be
626 * caught by the layer below (since it uses this route
627 * as a hint) or during the next transmit.
628 */
629 if (ROUTE_UNUSABLE(ro) || dst->sin_family != AF_INET ||
630 dst->sin_addr.s_addr != pkt_dst.s_addr)
631 ROUTE_RELEASE(ro);
632
633 /*
634 * If we're doing source interface selection, we may not
635 * want to use this route; only synch up the generation
636 * count otherwise.
637 */
638 if (!ipobf.select_srcif && ro->ro_rt != NULL &&
639 RT_GENID_OUTOFSYNC(ro->ro_rt))
640 RT_GENID_SYNC(ro->ro_rt);
641 }
642 if (ro->ro_rt == NULL) {
643 bzero(dst, sizeof (*dst));
644 dst->sin_family = AF_INET;
645 dst->sin_len = sizeof (*dst);
646 dst->sin_addr = pkt_dst;
647 }
648 /*
649 * If routing to interface only,
650 * short circuit routing lookup.
651 */
652 if (flags & IP_ROUTETOIF) {
653 if (ia != NULL)
654 IFA_REMREF(&ia->ia_ifa);
655 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
656 ia = ifatoia(ifa_ifwithnet(sintosa(dst)));
657 if (ia == NULL) {
658 OSAddAtomic(1, &ipstat.ips_noroute);
659 error = ENETUNREACH;
660 goto bad;
661 }
662 }
663 ifp = ia->ia_ifp;
664 ip->ip_ttl = 1;
665 ipobf.isbroadcast = in_broadcast(dst->sin_addr, ifp);
666 /*
667 * For consistency with other cases below. Loopback
668 * multicast case is handled separately by ip_mloopback().
669 */
670 if ((ifp->if_flags & IFF_LOOPBACK) &&
671 !IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
672 m->m_pkthdr.rcvif = ifp;
673 ip_setsrcifaddr_info(m, ifp->if_index, NULL);
674 ip_setdstifaddr_info(m, ifp->if_index, NULL);
675 }
676 } else if (IN_MULTICAST(ntohl(pkt_dst.s_addr)) &&
677 imo != NULL && (ifp = imo->imo_multicast_ifp) != NULL) {
678 /*
679 * Bypass the normal routing lookup for multicast
680 * packets if the interface is specified.
681 */
682 ipobf.isbroadcast = FALSE;
683 if (ia != NULL)
684 IFA_REMREF(&ia->ia_ifa);
685
686 /* Macro takes reference on ia */
687 IFP_TO_IA(ifp, ia);
688 } else {
689 struct ifaddr *ia0 = NULL;
690 boolean_t cloneok = FALSE;
691 /*
692 * Perform source interface selection; the source IP address
693 * must belong to one of the addresses of the interface used
694 * by the route. For performance reasons, do this only if
695 * there is no route, or if the routing table has changed,
696 * or if we haven't done source interface selection on this
697 * route (for this PCB instance) before.
698 */
699 if (ipobf.select_srcif &&
700 ip->ip_src.s_addr != INADDR_ANY && (ROUTE_UNUSABLE(ro) ||
701 !(ro->ro_flags & ROF_SRCIF_SELECTED))) {
702 /* Find the source interface */
703 ia0 = in_selectsrcif(ip, ro, ifscope);
704
705 /*
706 * If the source address belongs to a restricted
707 * interface and the caller forbids our using
708 * interfaces of such type, pretend that there is no
709 * route.
710 */
711 if (ia0 != NULL &&
712 IP_CHECK_RESTRICTIONS(ia0->ifa_ifp, ipobf)) {
713 IFA_REMREF(ia0);
714 ia0 = NULL;
715 error = EHOSTUNREACH;
716 if (flags & IP_OUTARGS)
717 ipoa->ipoa_retflags |= IPOARF_IFDENIED;
718 goto bad;
719 }
720
721 /*
722 * If the source address is spoofed (in the case of
723 * IP_RAWOUTPUT on an unbounded socket), or if this
724 * is destined for local/loopback, just let it go out
725 * using the interface of the route. Otherwise,
726 * there's no interface having such an address,
727 * so bail out.
728 */
729 if (ia0 == NULL && (!(flags & IP_RAWOUTPUT) ||
730 ipobf.srcbound) && ifscope != lo_ifp->if_index) {
731 error = EADDRNOTAVAIL;
732 goto bad;
733 }
734
735 /*
736 * If the caller didn't explicitly specify the scope,
737 * pick it up from the source interface. If the cached
738 * route was wrong and was blown away as part of source
739 * interface selection, don't mask out RTF_PRCLONING
740 * since that route may have been allocated by the ULP,
741 * unless the IP header was created by the caller or
742 * the destination is IPv4 LLA. The check for the
743 * latter is needed because IPv4 LLAs are never scoped
744 * in the current implementation, and we don't want to
745 * replace the resolved IPv4 LLA route with one whose
746 * gateway points to that of the default gateway on
747 * the primary interface of the system.
748 */
749 if (ia0 != NULL) {
750 if (ifscope == IFSCOPE_NONE)
751 ifscope = ia0->ifa_ifp->if_index;
752 cloneok = (!(flags & IP_RAWOUTPUT) &&
753 !(IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))));
754 }
755 }
756
757 /*
758 * If this is the case, we probably don't want to allocate
759 * a protocol-cloned route since we didn't get one from the
760 * ULP. This lets TCP do its thing, while not burdening
761 * forwarding or ICMP with the overhead of cloning a route.
762 * Of course, we still want to do any cloning requested by
763 * the link layer, as this is probably required in all cases
764 * for correct operation (as it is for ARP).
765 */
766 if (ro->ro_rt == NULL) {
767 unsigned long ign = RTF_PRCLONING;
768 /*
769 * We make an exception here: if the destination
770 * address is INADDR_BROADCAST, allocate a protocol-
771 * cloned host route so that we end up with a route
772 * marked with the RTF_BROADCAST flag. Otherwise,
773 * we would end up referring to the default route,
774 * instead of creating a cloned host route entry.
775 * That would introduce inconsistencies between ULPs
776 * that allocate a route and those that don't. The
777 * RTF_BROADCAST route is important since we'd want
778 * to send out undirected IP broadcast packets using
779 * link-level broadcast address. Another exception
780 * is for ULP-created routes that got blown away by
781 * source interface selection (see above).
782 *
783 * These exceptions will no longer be necessary when
784 * the RTF_PRCLONING scheme is no longer present.
785 */
786 if (cloneok || dst->sin_addr.s_addr == INADDR_BROADCAST)
787 ign &= ~RTF_PRCLONING;
788
789 /*
790 * Loosen the route lookup criteria if the ifscope
791 * corresponds to the loopback interface; this is
792 * needed to support Application Layer Gateways
793 * listening on loopback, in conjunction with packet
794 * filter redirection rules. The final source IP
795 * address will be rewritten by the packet filter
796 * prior to the RFC1122 loopback check below.
797 */
798 if (ifscope == lo_ifp->if_index)
799 rtalloc_ign(ro, ign);
800 else
801 rtalloc_scoped_ign(ro, ign, ifscope);
802
803 /*
804 * If the route points to a cellular/expensive interface
805 * and the caller forbids our using interfaces of such type,
806 * pretend that there is no route.
807 */
808 if (ro->ro_rt != NULL) {
809 RT_LOCK_SPIN(ro->ro_rt);
810 if (IP_CHECK_RESTRICTIONS(ro->ro_rt->rt_ifp,
811 ipobf)) {
812 RT_UNLOCK(ro->ro_rt);
813 ROUTE_RELEASE(ro);
814 if (flags & IP_OUTARGS) {
815 ipoa->ipoa_retflags |=
816 IPOARF_IFDENIED;
817 }
818 } else {
819 RT_UNLOCK(ro->ro_rt);
820 }
821 }
822 }
823
824 if (ro->ro_rt == NULL) {
825 OSAddAtomic(1, &ipstat.ips_noroute);
826 error = EHOSTUNREACH;
827 if (ia0 != NULL) {
828 IFA_REMREF(ia0);
829 ia0 = NULL;
830 }
831 goto bad;
832 }
833
834 if (ia != NULL)
835 IFA_REMREF(&ia->ia_ifa);
836 RT_LOCK_SPIN(ro->ro_rt);
837 ia = ifatoia(ro->ro_rt->rt_ifa);
838 if (ia != NULL) {
839 /* Become a regular mutex */
840 RT_CONVERT_LOCK(ro->ro_rt);
841 IFA_ADDREF(&ia->ia_ifa);
842 }
843 /*
844 * Note: ia_ifp may not be the same as rt_ifp; the latter
845 * is what we use for determining outbound i/f, mtu, etc.
846 */
847 ifp = ro->ro_rt->rt_ifp;
848 ro->ro_rt->rt_use++;
849 if (ro->ro_rt->rt_flags & RTF_GATEWAY) {
850 dst = SIN(ro->ro_rt->rt_gateway);
851 }
852 if (ro->ro_rt->rt_flags & RTF_HOST) {
853 /* double negation needed for bool bit field */
854 ipobf.isbroadcast =
855 !!(ro->ro_rt->rt_flags & RTF_BROADCAST);
856 } else {
857 /* Become a regular mutex */
858 RT_CONVERT_LOCK(ro->ro_rt);
859 ipobf.isbroadcast = in_broadcast(dst->sin_addr, ifp);
860 }
861 /*
862 * For consistency with IPv6, as well as to ensure that
863 * IP_RECVIF is set correctly for packets that are sent
864 * to one of the local addresses. ia (rt_ifa) would have
865 * been fixed up by rt_setif for local routes. This
866 * would make it appear as if the packet arrives on the
867 * interface which owns the local address. Loopback
868 * multicast case is handled separately by ip_mloopback().
869 */
870 if (ia != NULL && (ifp->if_flags & IFF_LOOPBACK) &&
871 !IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
872 uint32_t srcidx;
873
874 m->m_pkthdr.rcvif = ia->ia_ifa.ifa_ifp;
875
876 if (ia0 != NULL)
877 srcidx = ia0->ifa_ifp->if_index;
878 else if ((ro->ro_flags & ROF_SRCIF_SELECTED) &&
879 ro->ro_srcia != NULL)
880 srcidx = ro->ro_srcia->ifa_ifp->if_index;
881 else
882 srcidx = 0;
883
884 ip_setsrcifaddr_info(m, srcidx, NULL);
885 ip_setdstifaddr_info(m, 0, ia);
886 }
887 RT_UNLOCK(ro->ro_rt);
888 if (ia0 != NULL) {
889 IFA_REMREF(ia0);
890 ia0 = NULL;
891 }
892 }
893
894 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
895 struct ifnet *srcifp = NULL;
896 struct in_multi *inm;
897 u_int32_t vif;
898 u_int8_t ttl = IP_DEFAULT_MULTICAST_TTL;
899 u_int8_t loop = IP_DEFAULT_MULTICAST_LOOP;
900
901 m->m_flags |= M_MCAST;
902 /*
903 * IP destination address is multicast. Make sure "dst"
904 * still points to the address in "ro". (It may have been
905 * changed to point to a gateway address, above.)
906 */
907 dst = SIN(&ro->ro_dst);
908 /*
909 * See if the caller provided any multicast options
910 */
911 if (imo != NULL) {
912 IMO_LOCK(imo);
913 vif = imo->imo_multicast_vif;
914 ttl = imo->imo_multicast_ttl;
915 loop = imo->imo_multicast_loop;
916 if (!(flags & IP_RAWOUTPUT))
917 ip->ip_ttl = ttl;
918 if (imo->imo_multicast_ifp != NULL)
919 ifp = imo->imo_multicast_ifp;
920 IMO_UNLOCK(imo);
921 } else if (!(flags & IP_RAWOUTPUT)) {
922 vif = -1;
923 ip->ip_ttl = ttl;
924 }
925 /*
926 * Confirm that the outgoing interface supports multicast.
927 */
928 if (imo == NULL || vif == -1) {
929 if (!(ifp->if_flags & IFF_MULTICAST)) {
930 OSAddAtomic(1, &ipstat.ips_noroute);
931 error = ENETUNREACH;
932 goto bad;
933 }
934 }
935 /*
936 * If source address not specified yet, use address
937 * of outgoing interface.
938 */
939 if (ip->ip_src.s_addr == INADDR_ANY) {
940 struct in_ifaddr *ia1;
941 lck_rw_lock_shared(in_ifaddr_rwlock);
942 TAILQ_FOREACH(ia1, &in_ifaddrhead, ia_link) {
943 IFA_LOCK_SPIN(&ia1->ia_ifa);
944 if (ia1->ia_ifp == ifp) {
945 ip->ip_src = IA_SIN(ia1)->sin_addr;
946 srcifp = ifp;
947 IFA_UNLOCK(&ia1->ia_ifa);
948 break;
949 }
950 IFA_UNLOCK(&ia1->ia_ifa);
951 }
952 lck_rw_done(in_ifaddr_rwlock);
953 if (ip->ip_src.s_addr == INADDR_ANY) {
954 error = ENETUNREACH;
955 goto bad;
956 }
957 }
958
959 in_multihead_lock_shared();
960 IN_LOOKUP_MULTI(&pkt_dst, ifp, inm);
961 in_multihead_lock_done();
962 if (inm != NULL && (imo == NULL || loop)) {
963 /*
964 * If we belong to the destination multicast group
965 * on the outgoing interface, and the caller did not
966 * forbid loopback, loop back a copy.
967 */
968 if (!TAILQ_EMPTY(&ipv4_filters)) {
969 struct ipfilter *filter;
970 int seen = (inject_filter_ref == NULL);
971
972 if (imo != NULL) {
973 ipf_pktopts.ippo_flags |=
974 IPPOF_MCAST_OPTS;
975 ipf_pktopts.ippo_mcast_ifnet = ifp;
976 ipf_pktopts.ippo_mcast_ttl = ttl;
977 ipf_pktopts.ippo_mcast_loop = loop;
978 }
979
980 ipf_ref();
981
982 /*
983 * 4135317 - always pass network byte
984 * order to filter
985 */
986 #if BYTE_ORDER != BIG_ENDIAN
987 HTONS(ip->ip_len);
988 HTONS(ip->ip_off);
989 #endif
990 TAILQ_FOREACH(filter, &ipv4_filters, ipf_link) {
991 if (seen == 0) {
992 if ((struct ipfilter *)
993 inject_filter_ref == filter)
994 seen = 1;
995 } else if (filter->ipf_filter.
996 ipf_output != NULL) {
997 errno_t result;
998 result = filter->ipf_filter.
999 ipf_output(filter->
1000 ipf_filter.cookie,
1001 (mbuf_t *)&m, ippo);
1002 if (result == EJUSTRETURN) {
1003 ipf_unref();
1004 INM_REMREF(inm);
1005 goto done;
1006 }
1007 if (result != 0) {
1008 ipf_unref();
1009 INM_REMREF(inm);
1010 goto bad;
1011 }
1012 }
1013 }
1014
1015 /* set back to host byte order */
1016 ip = mtod(m, struct ip *);
1017 #if BYTE_ORDER != BIG_ENDIAN
1018 NTOHS(ip->ip_len);
1019 NTOHS(ip->ip_off);
1020 #endif
1021 ipf_unref();
1022 ipobf.didfilter = TRUE;
1023 }
1024 ip_mloopback(srcifp, ifp, m, dst, hlen);
1025 }
1026 if (inm != NULL)
1027 INM_REMREF(inm);
1028 /*
1029 * Multicasts with a time-to-live of zero may be looped-
1030 * back, above, but must not be transmitted on a network.
1031 * Also, multicasts addressed to the loopback interface
1032 * are not sent -- the above call to ip_mloopback() will
1033 * loop back a copy if this host actually belongs to the
1034 * destination group on the loopback interface.
1035 */
1036 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
1037 m_freem(m);
1038 goto done;
1039 }
1040
1041 goto sendit;
1042 }
1043 /*
1044 * If source address not specified yet, use address
1045 * of outgoing interface.
1046 */
1047 if (ip->ip_src.s_addr == INADDR_ANY) {
1048 IFA_LOCK_SPIN(&ia->ia_ifa);
1049 ip->ip_src = IA_SIN(ia)->sin_addr;
1050 IFA_UNLOCK(&ia->ia_ifa);
1051 #if IPFIREWALL_FORWARD
1052 /*
1053 * Keep note that we did this - if the firewall changes
1054 * the next-hop, our interface may change, changing the
1055 * default source IP. It's a shame so much effort happens
1056 * twice. Oh well.
1057 */
1058 ipobf.fwd_rewrite_src = TRUE;
1059 #endif /* IPFIREWALL_FORWARD */
1060 }
1061
1062 /*
1063 * Look for broadcast address and
1064 * and verify user is allowed to send
1065 * such a packet.
1066 */
1067 if (ipobf.isbroadcast) {
1068 if (!(ifp->if_flags & IFF_BROADCAST)) {
1069 error = EADDRNOTAVAIL;
1070 goto bad;
1071 }
1072 if (!(flags & IP_ALLOWBROADCAST)) {
1073 error = EACCES;
1074 goto bad;
1075 }
1076 /* don't allow broadcast messages to be fragmented */
1077 if ((u_short)ip->ip_len > ifp->if_mtu) {
1078 error = EMSGSIZE;
1079 goto bad;
1080 }
1081 m->m_flags |= M_BCAST;
1082 } else {
1083 m->m_flags &= ~M_BCAST;
1084 }
1085
1086 sendit:
1087 #if PF
1088 /* Invoke outbound packet filter */
1089 if (PF_IS_ENABLED) {
1090 int rc;
1091
1092 m0 = m; /* Save for later */
1093 #if DUMMYNET
1094 args.fwa_m = m;
1095 args.fwa_next_hop = dst;
1096 args.fwa_oif = ifp;
1097 args.fwa_ro = ro;
1098 args.fwa_dst = dst;
1099 args.fwa_oflags = flags;
1100 if (flags & IP_OUTARGS)
1101 args.fwa_ipoa = ipoa;
1102 rc = pf_af_hook(ifp, mppn, &m, AF_INET, FALSE, &args);
1103 #else /* DUMMYNET */
1104 rc = pf_af_hook(ifp, mppn, &m, AF_INET, FALSE, NULL);
1105 #endif /* DUMMYNET */
1106 if (rc != 0 || m == NULL) {
1107 /* Move to the next packet */
1108 m = *mppn;
1109
1110 /* Skip ahead if first packet in list got dropped */
1111 if (packetlist == m0)
1112 packetlist = m;
1113
1114 if (m != NULL) {
1115 m0 = m;
1116 /* Next packet in the chain */
1117 goto loopit;
1118 } else if (packetlist != NULL) {
1119 /* No more packet; send down the chain */
1120 goto sendchain;
1121 }
1122 /* Nothing left; we're done */
1123 goto done;
1124 }
1125 m0 = m;
1126 ip = mtod(m, struct ip *);
1127 pkt_dst = ip->ip_dst;
1128 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1129 }
1130 #endif /* PF */
1131 /*
1132 * Force IP TTL to 255 following draft-ietf-zeroconf-ipv4-linklocal.txt
1133 */
1134 if (IN_LINKLOCAL(ntohl(ip->ip_src.s_addr)) ||
1135 IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
1136 ip_linklocal_stat.iplls_out_total++;
1137 if (ip->ip_ttl != MAXTTL) {
1138 ip_linklocal_stat.iplls_out_badttl++;
1139 ip->ip_ttl = MAXTTL;
1140 }
1141 }
1142
1143 if (!ipobf.didfilter && !TAILQ_EMPTY(&ipv4_filters)) {
1144 struct ipfilter *filter;
1145 int seen = (inject_filter_ref == NULL);
1146 ipf_pktopts.ippo_flags &= ~IPPOF_MCAST_OPTS;
1147
1148 /*
1149 * Check that a TSO frame isn't passed to a filter.
1150 * This could happen if a filter is inserted while
1151 * TCP is sending the TSO packet.
1152 */
1153 if (m->m_pkthdr.csum_flags & CSUM_TSO_IPV4) {
1154 error = EMSGSIZE;
1155 goto bad;
1156 }
1157
1158 ipf_ref();
1159
1160 /* 4135317 - always pass network byte order to filter */
1161 #if BYTE_ORDER != BIG_ENDIAN
1162 HTONS(ip->ip_len);
1163 HTONS(ip->ip_off);
1164 #endif
1165 TAILQ_FOREACH(filter, &ipv4_filters, ipf_link) {
1166 if (seen == 0) {
1167 if ((struct ipfilter *)inject_filter_ref ==
1168 filter)
1169 seen = 1;
1170 } else if (filter->ipf_filter.ipf_output) {
1171 errno_t result;
1172 result = filter->ipf_filter.
1173 ipf_output(filter->ipf_filter.cookie,
1174 (mbuf_t *)&m, ippo);
1175 if (result == EJUSTRETURN) {
1176 ipf_unref();
1177 goto done;
1178 }
1179 if (result != 0) {
1180 ipf_unref();
1181 goto bad;
1182 }
1183 }
1184 }
1185 /* set back to host byte order */
1186 ip = mtod(m, struct ip *);
1187 #if BYTE_ORDER != BIG_ENDIAN
1188 NTOHS(ip->ip_len);
1189 NTOHS(ip->ip_off);
1190 #endif
1191 ipf_unref();
1192 }
1193
1194 #if NECP
1195 /* Process Network Extension Policy. Will Pass, Drop, or Rebind packet. */
1196 necp_matched_policy_id = necp_ip_output_find_policy_match (m,
1197 flags, (flags & IP_OUTARGS) ? ipoa : NULL, &necp_result, &necp_result_parameter);
1198 if (necp_matched_policy_id) {
1199 necp_mark_packet_from_ip(m, necp_matched_policy_id);
1200 switch (necp_result) {
1201 case NECP_KERNEL_POLICY_RESULT_PASS:
1202 /* Check if the interface is allowed */
1203 if (!necp_packet_is_allowed_over_interface(m, ifp)) {
1204 error = EHOSTUNREACH;
1205 goto bad;
1206 }
1207 goto skip_ipsec;
1208 case NECP_KERNEL_POLICY_RESULT_DROP:
1209 case NECP_KERNEL_POLICY_RESULT_SOCKET_DIVERT:
1210 /* Flow divert packets should be blocked at the IP layer */
1211 error = EHOSTUNREACH;
1212 goto bad;
1213 case NECP_KERNEL_POLICY_RESULT_IP_TUNNEL: {
1214 /* Verify that the packet is being routed to the tunnel */
1215 struct ifnet *policy_ifp = necp_get_ifnet_from_result_parameter(&necp_result_parameter);
1216 if (policy_ifp == ifp) {
1217 /* Check if the interface is allowed */
1218 if (!necp_packet_is_allowed_over_interface(m, ifp)) {
1219 error = EHOSTUNREACH;
1220 goto bad;
1221 }
1222 goto skip_ipsec;
1223 } else {
1224 if (necp_packet_can_rebind_to_ifnet(m, policy_ifp, &necp_route, AF_INET)) {
1225 /* Check if the interface is allowed */
1226 if (!necp_packet_is_allowed_over_interface(m, policy_ifp)) {
1227 error = EHOSTUNREACH;
1228 goto bad;
1229 }
1230
1231 /* Set ifp to the tunnel interface, since it is compatible with the packet */
1232 ifp = policy_ifp;
1233 ro = &necp_route;
1234 goto skip_ipsec;
1235 } else {
1236 error = ENETUNREACH;
1237 goto bad;
1238 }
1239 }
1240 break;
1241 }
1242 default:
1243 break;
1244 }
1245 }
1246 /* Catch-all to check if the interface is allowed */
1247 if (!necp_packet_is_allowed_over_interface(m, ifp)) {
1248 error = EHOSTUNREACH;
1249 goto bad;
1250 }
1251 #endif /* NECP */
1252
1253 #if IPSEC
1254 if (ipsec_bypass != 0 || (flags & IP_NOIPSEC))
1255 goto skip_ipsec;
1256
1257 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_START, 0, 0, 0, 0, 0);
1258
1259 if (sp == NULL) {
1260 /* get SP for this packet */
1261 if (so != NULL) {
1262 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND,
1263 so, &error);
1264 } else {
1265 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND,
1266 flags, &error);
1267 }
1268 if (sp == NULL) {
1269 IPSEC_STAT_INCREMENT(ipsecstat.out_inval);
1270 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END,
1271 0, 0, 0, 0, 0);
1272 goto bad;
1273 }
1274 }
1275
1276 error = 0;
1277
1278 /* check policy */
1279 switch (sp->policy) {
1280 case IPSEC_POLICY_DISCARD:
1281 case IPSEC_POLICY_GENERATE:
1282 /*
1283 * This packet is just discarded.
1284 */
1285 IPSEC_STAT_INCREMENT(ipsecstat.out_polvio);
1286 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END,
1287 1, 0, 0, 0, 0);
1288 goto bad;
1289
1290 case IPSEC_POLICY_BYPASS:
1291 case IPSEC_POLICY_NONE:
1292 /* no need to do IPsec. */
1293 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END,
1294 2, 0, 0, 0, 0);
1295 goto skip_ipsec;
1296
1297 case IPSEC_POLICY_IPSEC:
1298 if (sp->req == NULL) {
1299 /* acquire a policy */
1300 error = key_spdacquire(sp);
1301 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END,
1302 3, 0, 0, 0, 0);
1303 goto bad;
1304 }
1305 if (sp->ipsec_if) {
1306 /* Verify the redirect to ipsec interface */
1307 if (sp->ipsec_if == ifp) {
1308 goto skip_ipsec;
1309 }
1310 goto bad;
1311 }
1312 break;
1313
1314 case IPSEC_POLICY_ENTRUST:
1315 default:
1316 printf("ip_output: Invalid policy found. %d\n", sp->policy);
1317 }
1318 {
1319 ipsec_state.m = m;
1320 if (flags & IP_ROUTETOIF) {
1321 bzero(&ipsec_state.ro, sizeof (ipsec_state.ro));
1322 } else {
1323 route_copyout(&ipsec_state.ro, ro, sizeof (ipsec_state.ro));
1324 }
1325 ipsec_state.dst = SA(dst);
1326
1327 ip->ip_sum = 0;
1328
1329 /*
1330 * XXX
1331 * delayed checksums are not currently compatible with IPsec
1332 */
1333 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
1334 in_delayed_cksum(m);
1335
1336 #if BYTE_ORDER != BIG_ENDIAN
1337 HTONS(ip->ip_len);
1338 HTONS(ip->ip_off);
1339 #endif
1340
1341 DTRACE_IP6(send, struct mbuf *, m, struct inpcb *, NULL,
1342 struct ip *, ip, struct ifnet *, ifp,
1343 struct ip *, ip, struct ip6_hdr *, NULL);
1344
1345 error = ipsec4_output(&ipsec_state, sp, flags);
1346
1347 m0 = m = ipsec_state.m;
1348
1349 #if DUMMYNET
1350 /*
1351 * If we're about to use the route in ipsec_state
1352 * and this came from dummynet, cleaup now.
1353 */
1354 if (ro == &saved_route &&
1355 (!(flags & IP_ROUTETOIF) || ipsec_state.tunneled))
1356 ROUTE_RELEASE(ro);
1357 #endif /* DUMMYNET */
1358
1359 if (flags & IP_ROUTETOIF) {
1360 /*
1361 * if we have tunnel mode SA, we may need to ignore
1362 * IP_ROUTETOIF.
1363 */
1364 if (ipsec_state.tunneled) {
1365 flags &= ~IP_ROUTETOIF;
1366 ro = &ipsec_state.ro;
1367 }
1368 } else {
1369 ro = &ipsec_state.ro;
1370 }
1371 dst = SIN(ipsec_state.dst);
1372 if (error) {
1373 /* mbuf is already reclaimed in ipsec4_output. */
1374 m0 = NULL;
1375 switch (error) {
1376 case EHOSTUNREACH:
1377 case ENETUNREACH:
1378 case EMSGSIZE:
1379 case ENOBUFS:
1380 case ENOMEM:
1381 break;
1382 default:
1383 printf("ip4_output (ipsec): error code %d\n", error);
1384 /* FALLTHRU */
1385 case ENOENT:
1386 /* don't show these error codes to the user */
1387 error = 0;
1388 break;
1389 }
1390 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END,
1391 4, 0, 0, 0, 0);
1392 goto bad;
1393 }
1394 }
1395
1396 /* be sure to update variables that are affected by ipsec4_output() */
1397 ip = mtod(m, struct ip *);
1398
1399 #ifdef _IP_VHL
1400 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1401 #else /* !_IP_VHL */
1402 hlen = ip->ip_hl << 2;
1403 #endif /* !_IP_VHL */
1404 /* Check that there wasn't a route change and src is still valid */
1405 if (ROUTE_UNUSABLE(ro)) {
1406 ROUTE_RELEASE(ro);
1407 VERIFY(src_ia == NULL);
1408 if (ip->ip_src.s_addr != INADDR_ANY &&
1409 !(flags & (IP_ROUTETOIF | IP_FORWARDING)) &&
1410 (src_ia = ifa_foraddr(ip->ip_src.s_addr)) == NULL) {
1411 error = EADDRNOTAVAIL;
1412 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END,
1413 5, 0, 0, 0, 0);
1414 goto bad;
1415 }
1416 if (src_ia != NULL) {
1417 IFA_REMREF(&src_ia->ia_ifa);
1418 src_ia = NULL;
1419 }
1420 }
1421
1422 if (ro->ro_rt == NULL) {
1423 if (!(flags & IP_ROUTETOIF)) {
1424 printf("%s: can't update route after "
1425 "IPsec processing\n", __func__);
1426 error = EHOSTUNREACH; /* XXX */
1427 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END,
1428 6, 0, 0, 0, 0);
1429 goto bad;
1430 }
1431 } else {
1432 if (ia != NULL)
1433 IFA_REMREF(&ia->ia_ifa);
1434 RT_LOCK_SPIN(ro->ro_rt);
1435 ia = ifatoia(ro->ro_rt->rt_ifa);
1436 if (ia != NULL) {
1437 /* Become a regular mutex */
1438 RT_CONVERT_LOCK(ro->ro_rt);
1439 IFA_ADDREF(&ia->ia_ifa);
1440 }
1441 ifp = ro->ro_rt->rt_ifp;
1442 RT_UNLOCK(ro->ro_rt);
1443 }
1444
1445 /* make it flipped, again. */
1446 #if BYTE_ORDER != BIG_ENDIAN
1447 NTOHS(ip->ip_len);
1448 NTOHS(ip->ip_off);
1449 #endif
1450 KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END,
1451 7, 0xff, 0xff, 0xff, 0xff);
1452
1453 /* Pass to filters again */
1454 if (!TAILQ_EMPTY(&ipv4_filters)) {
1455 struct ipfilter *filter;
1456
1457 ipf_pktopts.ippo_flags &= ~IPPOF_MCAST_OPTS;
1458
1459 /*
1460 * Check that a TSO frame isn't passed to a filter.
1461 * This could happen if a filter is inserted while
1462 * TCP is sending the TSO packet.
1463 */
1464 if (m->m_pkthdr.csum_flags & CSUM_TSO_IPV4) {
1465 error = EMSGSIZE;
1466 goto bad;
1467 }
1468
1469 ipf_ref();
1470
1471 /* 4135317 - always pass network byte order to filter */
1472 #if BYTE_ORDER != BIG_ENDIAN
1473 HTONS(ip->ip_len);
1474 HTONS(ip->ip_off);
1475 #endif
1476 TAILQ_FOREACH(filter, &ipv4_filters, ipf_link) {
1477 if (filter->ipf_filter.ipf_output) {
1478 errno_t result;
1479 result = filter->ipf_filter.
1480 ipf_output(filter->ipf_filter.cookie,
1481 (mbuf_t *)&m, ippo);
1482 if (result == EJUSTRETURN) {
1483 ipf_unref();
1484 goto done;
1485 }
1486 if (result != 0) {
1487 ipf_unref();
1488 goto bad;
1489 }
1490 }
1491 }
1492 /* set back to host byte order */
1493 ip = mtod(m, struct ip *);
1494 #if BYTE_ORDER != BIG_ENDIAN
1495 NTOHS(ip->ip_len);
1496 NTOHS(ip->ip_off);
1497 #endif
1498 ipf_unref();
1499 }
1500 skip_ipsec:
1501 #endif /* IPSEC */
1502
1503 #if IPFIREWALL
1504 /*
1505 * Check with the firewall...
1506 * but not if we are already being fwd'd from a firewall.
1507 */
1508 if (fw_enable && IPFW_LOADED && !args.fwa_next_hop) {
1509 struct sockaddr_in *old = dst;
1510
1511 args.fwa_m = m;
1512 args.fwa_next_hop = dst;
1513 args.fwa_oif = ifp;
1514 ipfwoff = ip_fw_chk_ptr(&args);
1515 m = args.fwa_m;
1516 dst = args.fwa_next_hop;
1517
1518 /*
1519 * On return we must do the following:
1520 * IP_FW_PORT_DENY_FLAG -> drop the pkt (XXX new)
1521 * 1<=off<= 0xffff -> DIVERT
1522 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe
1523 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet
1524 * dst != old -> IPFIREWALL_FORWARD
1525 * off==0, dst==old -> accept
1526 * If some of the above modules is not compiled in, then
1527 * we should't have to check the corresponding condition
1528 * (because the ipfw control socket should not accept
1529 * unsupported rules), but better play safe and drop
1530 * packets in case of doubt.
1531 */
1532 m0 = m;
1533 if ((ipfwoff & IP_FW_PORT_DENY_FLAG) || m == NULL) {
1534 if (m)
1535 m_freem(m);
1536 error = EACCES;
1537 goto done;
1538 }
1539 ip = mtod(m, struct ip *);
1540
1541 if (ipfwoff == 0 && dst == old) { /* common case */
1542 goto pass;
1543 }
1544 #if DUMMYNET
1545 if (DUMMYNET_LOADED && (ipfwoff & IP_FW_PORT_DYNT_FLAG) != 0) {
1546 /*
1547 * pass the pkt to dummynet. Need to include
1548 * pipe number, m, ifp, ro, dst because these are
1549 * not recomputed in the next pass.
1550 * All other parameters have been already used and
1551 * so they are not needed anymore.
1552 * XXX note: if the ifp or ro entry are deleted
1553 * while a pkt is in dummynet, we are in trouble!
1554 */
1555 args.fwa_ro = ro;
1556 args.fwa_dst = dst;
1557 args.fwa_oflags = flags;
1558 if (flags & IP_OUTARGS)
1559 args.fwa_ipoa = ipoa;
1560
1561 error = ip_dn_io_ptr(m, ipfwoff & 0xffff, DN_TO_IP_OUT,
1562 &args, DN_CLIENT_IPFW);
1563 goto done;
1564 }
1565 #endif /* DUMMYNET */
1566 #if IPDIVERT
1567 if (ipfwoff != 0 && (ipfwoff & IP_FW_PORT_DYNT_FLAG) == 0) {
1568 struct mbuf *clone = NULL;
1569
1570 /* Clone packet if we're doing a 'tee' */
1571 if ((ipfwoff & IP_FW_PORT_TEE_FLAG) != 0)
1572 clone = m_dup(m, M_DONTWAIT);
1573 /*
1574 * XXX
1575 * delayed checksums are not currently compatible
1576 * with divert sockets.
1577 */
1578 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
1579 in_delayed_cksum(m);
1580
1581 /* Restore packet header fields to original values */
1582
1583 #if BYTE_ORDER != BIG_ENDIAN
1584 HTONS(ip->ip_len);
1585 HTONS(ip->ip_off);
1586 #endif
1587
1588 /* Deliver packet to divert input routine */
1589 divert_packet(m, 0, ipfwoff & 0xffff,
1590 args.fwa_divert_rule);
1591
1592 /* If 'tee', continue with original packet */
1593 if (clone != NULL) {
1594 m0 = m = clone;
1595 ip = mtod(m, struct ip *);
1596 goto pass;
1597 }
1598 goto done;
1599 }
1600 #endif /* IPDIVERT */
1601 #if IPFIREWALL_FORWARD
1602 /*
1603 * Here we check dst to make sure it's directly reachable on
1604 * the interface we previously thought it was.
1605 * If it isn't (which may be likely in some situations) we have
1606 * to re-route it (ie, find a route for the next-hop and the
1607 * associated interface) and set them here. This is nested
1608 * forwarding which in most cases is undesirable, except where
1609 * such control is nigh impossible. So we do it here.
1610 * And I'm babbling.
1611 */
1612 if (ipfwoff == 0 && old != dst) {
1613 struct in_ifaddr *ia_fw;
1614 struct route *ro_fwd = &sro_fwd;
1615
1616 #if IPFIREWALL_FORWARD_DEBUG
1617 printf("IPFIREWALL_FORWARD: New dst ip: ");
1618 print_ip(dst->sin_addr);
1619 printf("\n");
1620 #endif /* IPFIREWALL_FORWARD_DEBUG */
1621 /*
1622 * We need to figure out if we have been forwarded
1623 * to a local socket. If so then we should somehow
1624 * "loop back" to ip_input, and get directed to the
1625 * PCB as if we had received this packet. This is
1626 * because it may be dificult to identify the packets
1627 * you want to forward until they are being output
1628 * and have selected an interface. (e.g. locally
1629 * initiated packets) If we used the loopback inteface,
1630 * we would not be able to control what happens
1631 * as the packet runs through ip_input() as
1632 * it is done through a ISR.
1633 */
1634 lck_rw_lock_shared(in_ifaddr_rwlock);
1635 TAILQ_FOREACH(ia_fw, &in_ifaddrhead, ia_link) {
1636 /*
1637 * If the addr to forward to is one
1638 * of ours, we pretend to
1639 * be the destination for this packet.
1640 */
1641 IFA_LOCK_SPIN(&ia_fw->ia_ifa);
1642 if (IA_SIN(ia_fw)->sin_addr.s_addr ==
1643 dst->sin_addr.s_addr) {
1644 IFA_UNLOCK(&ia_fw->ia_ifa);
1645 break;
1646 }
1647 IFA_UNLOCK(&ia_fw->ia_ifa);
1648 }
1649 lck_rw_done(in_ifaddr_rwlock);
1650 if (ia_fw) {
1651 /* tell ip_input "dont filter" */
1652 struct m_tag *fwd_tag;
1653 struct ip_fwd_tag *ipfwd_tag;
1654
1655 fwd_tag = m_tag_create(KERNEL_MODULE_TAG_ID,
1656 KERNEL_TAG_TYPE_IPFORWARD,
1657 sizeof (*ipfwd_tag), M_NOWAIT, m);
1658 if (fwd_tag == NULL) {
1659 error = ENOBUFS;
1660 goto bad;
1661 }
1662
1663 ipfwd_tag = (struct ip_fwd_tag *)(fwd_tag+1);
1664 ipfwd_tag->next_hop = args.fwa_next_hop;
1665
1666 m_tag_prepend(m, fwd_tag);
1667
1668 if (m->m_pkthdr.rcvif == NULL)
1669 m->m_pkthdr.rcvif = lo_ifp;
1670
1671 #if BYTE_ORDER != BIG_ENDIAN
1672 HTONS(ip->ip_len);
1673 HTONS(ip->ip_off);
1674 #endif
1675 mbuf_outbound_finalize(m, PF_INET, 0);
1676
1677 /*
1678 * we need to call dlil_output to run filters
1679 * and resync to avoid recursion loops.
1680 */
1681 if (lo_ifp) {
1682 dlil_output(lo_ifp, PF_INET, m, NULL,
1683 SA(dst), 0, adv);
1684 } else {
1685 printf("%s: no loopback ifp for "
1686 "forwarding!!!\n", __func__);
1687 }
1688 goto done;
1689 }
1690 /*
1691 * Some of the logic for this was nicked from above.
1692 *
1693 * This rewrites the cached route in a local PCB.
1694 * Is this what we want to do?
1695 */
1696 ROUTE_RELEASE(ro_fwd);
1697 bcopy(dst, &ro_fwd->ro_dst, sizeof (*dst));
1698
1699 rtalloc_ign(ro_fwd, RTF_PRCLONING);
1700
1701 if (ro_fwd->ro_rt == NULL) {
1702 OSAddAtomic(1, &ipstat.ips_noroute);
1703 error = EHOSTUNREACH;
1704 goto bad;
1705 }
1706
1707 RT_LOCK_SPIN(ro_fwd->ro_rt);
1708 ia_fw = ifatoia(ro_fwd->ro_rt->rt_ifa);
1709 if (ia_fw != NULL) {
1710 /* Become a regular mutex */
1711 RT_CONVERT_LOCK(ro_fwd->ro_rt);
1712 IFA_ADDREF(&ia_fw->ia_ifa);
1713 }
1714 ifp = ro_fwd->ro_rt->rt_ifp;
1715 ro_fwd->ro_rt->rt_use++;
1716 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY)
1717 dst = SIN(ro_fwd->ro_rt->rt_gateway);
1718 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) {
1719 /* double negation needed for bool bit field */
1720 ipobf.isbroadcast =
1721 !!(ro_fwd->ro_rt->rt_flags & RTF_BROADCAST);
1722 } else {
1723 /* Become a regular mutex */
1724 RT_CONVERT_LOCK(ro_fwd->ro_rt);
1725 ipobf.isbroadcast =
1726 in_broadcast(dst->sin_addr, ifp);
1727 }
1728 RT_UNLOCK(ro_fwd->ro_rt);
1729 ROUTE_RELEASE(ro);
1730 ro->ro_rt = ro_fwd->ro_rt;
1731 ro_fwd->ro_rt = NULL;
1732 dst = SIN(&ro_fwd->ro_dst);
1733
1734 /*
1735 * If we added a default src ip earlier,
1736 * which would have been gotten from the-then
1737 * interface, do it again, from the new one.
1738 */
1739 if (ia_fw != NULL) {
1740 if (ipobf.fwd_rewrite_src) {
1741 IFA_LOCK_SPIN(&ia_fw->ia_ifa);
1742 ip->ip_src = IA_SIN(ia_fw)->sin_addr;
1743 IFA_UNLOCK(&ia_fw->ia_ifa);
1744 }
1745 IFA_REMREF(&ia_fw->ia_ifa);
1746 }
1747 goto pass;
1748 }
1749 #endif /* IPFIREWALL_FORWARD */
1750 /*
1751 * if we get here, none of the above matches, and
1752 * we have to drop the pkt
1753 */
1754 m_freem(m);
1755 error = EACCES; /* not sure this is the right error msg */
1756 goto done;
1757 }
1758
1759 pass:
1760 #endif /* IPFIREWALL */
1761
1762 /* 127/8 must not appear on wire - RFC1122 */
1763 if (!(ifp->if_flags & IFF_LOOPBACK) &&
1764 ((ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
1765 (ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)) {
1766 OSAddAtomic(1, &ipstat.ips_badaddr);
1767 error = EADDRNOTAVAIL;
1768 goto bad;
1769 }
1770
1771 ip_output_checksum(ifp, m, (IP_VHL_HL(ip->ip_vhl) << 2),
1772 ip->ip_len, &sw_csum);
1773
1774 /*
1775 * If small enough for interface, or the interface will take
1776 * care of the fragmentation for us, can just send directly.
1777 */
1778 if ((u_short)ip->ip_len <= ifp->if_mtu || TSO_IPV4_OK(ifp, m) ||
1779 (!(ip->ip_off & IP_DF) && (ifp->if_hwassist & CSUM_FRAGMENT))) {
1780 #if BYTE_ORDER != BIG_ENDIAN
1781 HTONS(ip->ip_len);
1782 HTONS(ip->ip_off);
1783 #endif
1784
1785 ip->ip_sum = 0;
1786 if (sw_csum & CSUM_DELAY_IP) {
1787 ip->ip_sum = ip_cksum_hdr_out(m, hlen);
1788 sw_csum &= ~CSUM_DELAY_IP;
1789 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_IP;
1790 }
1791
1792 #if IPSEC
1793 /* clean ipsec history once it goes out of the node */
1794 if (ipsec_bypass == 0 && !(flags & IP_NOIPSEC))
1795 ipsec_delaux(m);
1796 #endif /* IPSEC */
1797 if ((m->m_pkthdr.csum_flags & CSUM_TSO_IPV4) &&
1798 (m->m_pkthdr.tso_segsz > 0))
1799 scnt += m->m_pkthdr.len / m->m_pkthdr.tso_segsz;
1800 else
1801 scnt++;
1802
1803 if (packetchain == 0) {
1804 if (ro->ro_rt != NULL && nstat_collect)
1805 nstat_route_tx(ro->ro_rt, scnt,
1806 m->m_pkthdr.len, 0);
1807
1808 error = dlil_output(ifp, PF_INET, m, ro->ro_rt,
1809 SA(dst), 0, adv);
1810 if (dlil_verbose && error) {
1811 printf("dlil_output error on interface %s: %d\n",
1812 ifp->if_xname, error);
1813 }
1814 scnt = 0;
1815 goto done;
1816 } else {
1817 /*
1818 * packet chaining allows us to reuse the
1819 * route for all packets
1820 */
1821 bytecnt += m->m_pkthdr.len;
1822 mppn = &m->m_nextpkt;
1823 m = m->m_nextpkt;
1824 if (m == NULL) {
1825 #if PF
1826 sendchain:
1827 #endif /* PF */
1828 if (pktcnt > ip_maxchainsent)
1829 ip_maxchainsent = pktcnt;
1830 if (ro->ro_rt != NULL && nstat_collect)
1831 nstat_route_tx(ro->ro_rt, scnt,
1832 bytecnt, 0);
1833
1834 error = dlil_output(ifp, PF_INET, packetlist,
1835 ro->ro_rt, SA(dst), 0, adv);
1836 if (dlil_verbose && error) {
1837 printf("dlil_output error on interface %s: %d\n",
1838 ifp->if_xname, error);
1839 }
1840 pktcnt = 0;
1841 scnt = 0;
1842 bytecnt = 0;
1843 goto done;
1844
1845 }
1846 m0 = m;
1847 pktcnt++;
1848 goto loopit;
1849 }
1850 }
1851 /*
1852 * Too large for interface; fragment if possible.
1853 * Must be able to put at least 8 bytes per fragment.
1854 * Balk when DF bit is set or the interface didn't support TSO.
1855 */
1856 if ((ip->ip_off & IP_DF) || pktcnt > 0 ||
1857 (m->m_pkthdr.csum_flags & CSUM_TSO_IPV4)) {
1858 error = EMSGSIZE;
1859 /*
1860 * This case can happen if the user changed the MTU
1861 * of an interface after enabling IP on it. Because
1862 * most netifs don't keep track of routes pointing to
1863 * them, there is no way for one to update all its
1864 * routes when the MTU is changed.
1865 */
1866 if (ro->ro_rt) {
1867 RT_LOCK_SPIN(ro->ro_rt);
1868 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) &&
1869 !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) &&
1870 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
1871 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
1872 }
1873 RT_UNLOCK(ro->ro_rt);
1874 }
1875 if (pktcnt > 0) {
1876 m0 = packetlist;
1877 }
1878 OSAddAtomic(1, &ipstat.ips_cantfrag);
1879 goto bad;
1880 }
1881
1882 error = ip_fragment(m, ifp, ifp->if_mtu, sw_csum);
1883 if (error != 0) {
1884 m0 = m = NULL;
1885 goto bad;
1886 }
1887
1888 KERNEL_DEBUG(DBG_LAYER_END, ip->ip_dst.s_addr,
1889 ip->ip_src.s_addr, ip->ip_p, ip->ip_off, ip->ip_len);
1890
1891 for (m = m0; m; m = m0) {
1892 m0 = m->m_nextpkt;
1893 m->m_nextpkt = 0;
1894 #if IPSEC
1895 /* clean ipsec history once it goes out of the node */
1896 if (ipsec_bypass == 0 && !(flags & IP_NOIPSEC))
1897 ipsec_delaux(m);
1898 #endif /* IPSEC */
1899 if (error == 0) {
1900 if ((packetchain != 0) && (pktcnt > 0)) {
1901 panic("%s: mix of packet in packetlist is "
1902 "wrong=%p", __func__, packetlist);
1903 /* NOTREACHED */
1904 }
1905 if (ro->ro_rt != NULL && nstat_collect) {
1906 nstat_route_tx(ro->ro_rt, 1,
1907 m->m_pkthdr.len, 0);
1908 }
1909 error = dlil_output(ifp, PF_INET, m, ro->ro_rt,
1910 SA(dst), 0, adv);
1911 if (dlil_verbose && error) {
1912 printf("dlil_output error on interface %s: %d\n",
1913 ifp->if_xname, error);
1914 }
1915 } else {
1916 m_freem(m);
1917 }
1918 }
1919
1920 if (error == 0)
1921 OSAddAtomic(1, &ipstat.ips_fragmented);
1922
1923 done:
1924 if (ia != NULL) {
1925 IFA_REMREF(&ia->ia_ifa);
1926 ia = NULL;
1927 }
1928 #if IPSEC
1929 ROUTE_RELEASE(&ipsec_state.ro);
1930 if (sp != NULL) {
1931 KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
1932 printf("DP ip_output call free SP:%x\n", sp));
1933 key_freesp(sp, KEY_SADB_UNLOCKED);
1934 }
1935 #endif /* IPSEC */
1936 #if NECP
1937 ROUTE_RELEASE(&necp_route);
1938 #endif /* NECP */
1939 #if DUMMYNET
1940 ROUTE_RELEASE(&saved_route);
1941 #endif /* DUMMYNET */
1942 #if IPFIREWALL_FORWARD
1943 ROUTE_RELEASE(&sro_fwd);
1944 #endif /* IPFIREWALL_FORWARD */
1945
1946 KERNEL_DEBUG(DBG_FNC_IP_OUTPUT | DBG_FUNC_END, error, 0, 0, 0, 0);
1947 if (ip_output_measure) {
1948 net_perf_measure_time(&net_perf, &start_tv, packets_processed);
1949 net_perf_histogram(&net_perf, packets_processed);
1950 }
1951 return (error);
1952 bad:
1953 if (pktcnt > 0)
1954 m0 = packetlist;
1955 m_freem_list(m0);
1956 goto done;
1957
1958 #undef ipsec_state
1959 #undef args
1960 #undef sro_fwd
1961 #undef saved_route
1962 #undef ipf_pktopts
1963 #undef IP_CHECK_RESTRICTIONS
1964 }
1965
1966 int
1967 ip_fragment(struct mbuf *m, struct ifnet *ifp, unsigned long mtu, int sw_csum)
1968 {
1969 struct ip *ip, *mhip;
1970 int len, hlen, mhlen, firstlen, off, error = 0;
1971 struct mbuf **mnext = &m->m_nextpkt, *m0;
1972 int nfrags = 1;
1973
1974 ip = mtod(m, struct ip *);
1975 #ifdef _IP_VHL
1976 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1977 #else /* !_IP_VHL */
1978 hlen = ip->ip_hl << 2;
1979 #endif /* !_IP_VHL */
1980
1981 firstlen = len = (mtu - hlen) &~ 7;
1982 if (len < 8) {
1983 m_freem(m);
1984 return (EMSGSIZE);
1985 }
1986
1987 /*
1988 * if the interface will not calculate checksums on
1989 * fragmented packets, then do it here.
1990 */
1991 if ((m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) &&
1992 !(ifp->if_hwassist & CSUM_IP_FRAGS))
1993 in_delayed_cksum(m);
1994
1995 /*
1996 * Loop through length of segment after first fragment,
1997 * make new header and copy data of each part and link onto chain.
1998 */
1999 m0 = m;
2000 mhlen = sizeof (struct ip);
2001 for (off = hlen + len; off < (u_short)ip->ip_len; off += len) {
2002 MGETHDR(m, M_DONTWAIT, MT_HEADER); /* MAC-OK */
2003 if (m == NULL) {
2004 error = ENOBUFS;
2005 OSAddAtomic(1, &ipstat.ips_odropped);
2006 goto sendorfree;
2007 }
2008 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
2009 m->m_data += max_linkhdr;
2010 mhip = mtod(m, struct ip *);
2011 *mhip = *ip;
2012 if (hlen > sizeof (struct ip)) {
2013 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
2014 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2);
2015 }
2016 m->m_len = mhlen;
2017 mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
2018 if (ip->ip_off & IP_MF)
2019 mhip->ip_off |= IP_MF;
2020 if (off + len >= (u_short)ip->ip_len)
2021 len = (u_short)ip->ip_len - off;
2022 else
2023 mhip->ip_off |= IP_MF;
2024 mhip->ip_len = htons((u_short)(len + mhlen));
2025 m->m_next = m_copy(m0, off, len);
2026 if (m->m_next == NULL) {
2027 (void) m_free(m);
2028 error = ENOBUFS; /* ??? */
2029 OSAddAtomic(1, &ipstat.ips_odropped);
2030 goto sendorfree;
2031 }
2032 m->m_pkthdr.len = mhlen + len;
2033 m->m_pkthdr.rcvif = NULL;
2034 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
2035
2036 M_COPY_CLASSIFIER(m, m0);
2037 M_COPY_PFTAG(m, m0);
2038
2039 #if CONFIG_MACF_NET
2040 mac_netinet_fragment(m0, m);
2041 #endif /* CONFIG_MACF_NET */
2042
2043 #if BYTE_ORDER != BIG_ENDIAN
2044 HTONS(mhip->ip_off);
2045 #endif
2046
2047 mhip->ip_sum = 0;
2048 if (sw_csum & CSUM_DELAY_IP) {
2049 mhip->ip_sum = ip_cksum_hdr_out(m, mhlen);
2050 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_IP;
2051 }
2052 *mnext = m;
2053 mnext = &m->m_nextpkt;
2054 nfrags++;
2055 }
2056 OSAddAtomic(nfrags, &ipstat.ips_ofragments);
2057
2058 /* set first/last markers for fragment chain */
2059 m->m_flags |= M_LASTFRAG;
2060 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
2061 m0->m_pkthdr.csum_data = nfrags;
2062
2063 /*
2064 * Update first fragment by trimming what's been copied out
2065 * and updating header, then send each fragment (in order).
2066 */
2067 m = m0;
2068 m_adj(m, hlen + firstlen - (u_short)ip->ip_len);
2069 m->m_pkthdr.len = hlen + firstlen;
2070 ip->ip_len = htons((u_short)m->m_pkthdr.len);
2071 ip->ip_off |= IP_MF;
2072
2073 #if BYTE_ORDER != BIG_ENDIAN
2074 HTONS(ip->ip_off);
2075 #endif
2076
2077 ip->ip_sum = 0;
2078 if (sw_csum & CSUM_DELAY_IP) {
2079 ip->ip_sum = ip_cksum_hdr_out(m, hlen);
2080 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_IP;
2081 }
2082 sendorfree:
2083 if (error)
2084 m_freem_list(m0);
2085
2086 return (error);
2087 }
2088
2089 static void
2090 ip_out_cksum_stats(int proto, u_int32_t len)
2091 {
2092 switch (proto) {
2093 case IPPROTO_TCP:
2094 tcp_out_cksum_stats(len);
2095 break;
2096 case IPPROTO_UDP:
2097 udp_out_cksum_stats(len);
2098 break;
2099 default:
2100 /* keep only TCP or UDP stats for now */
2101 break;
2102 }
2103 }
2104
2105 /*
2106 * Process a delayed payload checksum calculation (outbound path.)
2107 *
2108 * hoff is the number of bytes beyond the mbuf data pointer which
2109 * points to the IP header.
2110 *
2111 * Returns a bitmask representing all the work done in software.
2112 */
2113 uint32_t
2114 in_finalize_cksum(struct mbuf *m, uint32_t hoff, uint32_t csum_flags)
2115 {
2116 unsigned char buf[15 << 2] __attribute__((aligned(8)));
2117 struct ip *ip;
2118 uint32_t offset, _hlen, mlen, hlen, len, sw_csum;
2119 uint16_t csum, ip_len;
2120
2121 _CASSERT(sizeof (csum) == sizeof (uint16_t));
2122 VERIFY(m->m_flags & M_PKTHDR);
2123
2124 sw_csum = (csum_flags & m->m_pkthdr.csum_flags);
2125
2126 if ((sw_csum &= (CSUM_DELAY_IP | CSUM_DELAY_DATA)) == 0)
2127 goto done;
2128
2129 mlen = m->m_pkthdr.len; /* total mbuf len */
2130
2131 /* sanity check (need at least simple IP header) */
2132 if (mlen < (hoff + sizeof (*ip))) {
2133 panic("%s: mbuf %p pkt len (%u) < hoff+ip_hdr "
2134 "(%u+%u)\n", __func__, m, mlen, hoff,
2135 (uint32_t)sizeof (*ip));
2136 /* NOTREACHED */
2137 }
2138
2139 /*
2140 * In case the IP header is not contiguous, or not 32-bit aligned,
2141 * or if we're computing the IP header checksum, copy it to a local
2142 * buffer. Copy only the simple IP header here (IP options case
2143 * is handled below.)
2144 */
2145 if ((sw_csum & CSUM_DELAY_IP) || (hoff + sizeof (*ip)) > m->m_len ||
2146 !IP_HDR_ALIGNED_P(mtod(m, caddr_t) + hoff)) {
2147 m_copydata(m, hoff, sizeof (*ip), (caddr_t)buf);
2148 ip = (struct ip *)(void *)buf;
2149 _hlen = sizeof (*ip);
2150 } else {
2151 ip = (struct ip *)(void *)(m->m_data + hoff);
2152 _hlen = 0;
2153 }
2154
2155 hlen = IP_VHL_HL(ip->ip_vhl) << 2; /* IP header len */
2156
2157 /* sanity check */
2158 if (mlen < (hoff + hlen)) {
2159 panic("%s: mbuf %p pkt too short (%d) for IP header (%u), "
2160 "hoff %u", __func__, m, mlen, hlen, hoff);
2161 /* NOTREACHED */
2162 }
2163
2164 /*
2165 * We could be in the context of an IP or interface filter; in the
2166 * former case, ip_len would be in host (correct) order while for
2167 * the latter it would be in network order. Because of this, we
2168 * attempt to interpret the length field by comparing it against
2169 * the actual packet length. If the comparison fails, byte swap
2170 * the length and check again. If it still fails, use the actual
2171 * packet length. This also covers the trailing bytes case.
2172 */
2173 ip_len = ip->ip_len;
2174 if (ip_len != (mlen - hoff)) {
2175 ip_len = OSSwapInt16(ip_len);
2176 if (ip_len != (mlen - hoff)) {
2177 printf("%s: mbuf 0x%llx proto %d IP len %d (%x) "
2178 "[swapped %d (%x)] doesn't match actual packet "
2179 "length; %d is used instead\n", __func__,
2180 (uint64_t)VM_KERNEL_ADDRPERM(m), ip->ip_p,
2181 ip->ip_len, ip->ip_len, ip_len, ip_len,
2182 (mlen - hoff));
2183 ip_len = mlen - hoff;
2184 }
2185 }
2186
2187 len = ip_len - hlen; /* csum span */
2188
2189 if (sw_csum & CSUM_DELAY_DATA) {
2190 uint16_t ulpoff;
2191
2192 /*
2193 * offset is added to the lower 16-bit value of csum_data,
2194 * which is expected to contain the ULP offset; therefore
2195 * CSUM_PARTIAL offset adjustment must be undone.
2196 */
2197 if ((m->m_pkthdr.csum_flags & (CSUM_PARTIAL|CSUM_DATA_VALID)) ==
2198 (CSUM_PARTIAL|CSUM_DATA_VALID)) {
2199 /*
2200 * Get back the original ULP offset (this will
2201 * undo the CSUM_PARTIAL logic in ip_output.)
2202 */
2203 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_tx_stuff -
2204 m->m_pkthdr.csum_tx_start);
2205 }
2206
2207 ulpoff = (m->m_pkthdr.csum_data & 0xffff); /* ULP csum offset */
2208 offset = hoff + hlen; /* ULP header */
2209
2210 if (mlen < (ulpoff + sizeof (csum))) {
2211 panic("%s: mbuf %p pkt len (%u) proto %d invalid ULP "
2212 "cksum offset (%u) cksum flags 0x%x\n", __func__,
2213 m, mlen, ip->ip_p, ulpoff, m->m_pkthdr.csum_flags);
2214 /* NOTREACHED */
2215 }
2216
2217 csum = inet_cksum(m, 0, offset, len);
2218
2219 /* Update stats */
2220 ip_out_cksum_stats(ip->ip_p, len);
2221
2222 /* RFC1122 4.1.3.4 */
2223 if (csum == 0 && (m->m_pkthdr.csum_flags & CSUM_UDP))
2224 csum = 0xffff;
2225
2226 /* Insert the checksum in the ULP csum field */
2227 offset += ulpoff;
2228 if (offset + sizeof (csum) > m->m_len) {
2229 m_copyback(m, offset, sizeof (csum), &csum);
2230 } else if (IP_HDR_ALIGNED_P(mtod(m, char *) + hoff)) {
2231 *(uint16_t *)(void *)(mtod(m, char *) + offset) = csum;
2232 } else {
2233 bcopy(&csum, (mtod(m, char *) + offset), sizeof (csum));
2234 }
2235 m->m_pkthdr.csum_flags &=
2236 ~(CSUM_DELAY_DATA | CSUM_DATA_VALID | CSUM_PARTIAL);
2237 }
2238
2239 if (sw_csum & CSUM_DELAY_IP) {
2240 /* IP header must be in the local buffer */
2241 VERIFY(_hlen == sizeof (*ip));
2242 if (_hlen != hlen) {
2243 VERIFY(hlen <= sizeof (buf));
2244 m_copydata(m, hoff, hlen, (caddr_t)buf);
2245 ip = (struct ip *)(void *)buf;
2246 _hlen = hlen;
2247 }
2248
2249 /*
2250 * Compute the IP header checksum as if the IP length
2251 * is the length which we believe is "correct"; see
2252 * how ip_len gets calculated above. Note that this
2253 * is done on the local copy and not on the real one.
2254 */
2255 ip->ip_len = htons(ip_len);
2256 ip->ip_sum = 0;
2257 csum = in_cksum_hdr_opt(ip);
2258
2259 /* Update stats */
2260 ipstat.ips_snd_swcsum++;
2261 ipstat.ips_snd_swcsum_bytes += hlen;
2262
2263 /*
2264 * Insert only the checksum in the existing IP header
2265 * csum field; all other fields are left unchanged.
2266 */
2267 offset = hoff + offsetof(struct ip, ip_sum);
2268 if (offset + sizeof (csum) > m->m_len) {
2269 m_copyback(m, offset, sizeof (csum), &csum);
2270 } else if (IP_HDR_ALIGNED_P(mtod(m, char *) + hoff)) {
2271 *(uint16_t *)(void *)(mtod(m, char *) + offset) = csum;
2272 } else {
2273 bcopy(&csum, (mtod(m, char *) + offset), sizeof (csum));
2274 }
2275 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_IP;
2276 }
2277
2278 done:
2279 return (sw_csum);
2280 }
2281
2282 /*
2283 * Insert IP options into preformed packet.
2284 * Adjust IP destination as required for IP source routing,
2285 * as indicated by a non-zero in_addr at the start of the options.
2286 *
2287 * XXX This routine assumes that the packet has no options in place.
2288 */
2289 static struct mbuf *
2290 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
2291 {
2292 struct ipoption *p = mtod(opt, struct ipoption *);
2293 struct mbuf *n;
2294 struct ip *ip = mtod(m, struct ip *);
2295 unsigned optlen;
2296
2297 optlen = opt->m_len - sizeof (p->ipopt_dst);
2298 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET)
2299 return (m); /* XXX should fail */
2300 if (p->ipopt_dst.s_addr)
2301 ip->ip_dst = p->ipopt_dst;
2302 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
2303 MGETHDR(n, M_DONTWAIT, MT_HEADER); /* MAC-OK */
2304 if (n == NULL)
2305 return (m);
2306 n->m_pkthdr.rcvif = 0;
2307 #if CONFIG_MACF_NET
2308 mac_mbuf_label_copy(m, n);
2309 #endif /* CONFIG_MACF_NET */
2310 n->m_pkthdr.len = m->m_pkthdr.len + optlen;
2311 m->m_len -= sizeof (struct ip);
2312 m->m_data += sizeof (struct ip);
2313 n->m_next = m;
2314 m = n;
2315 m->m_len = optlen + sizeof (struct ip);
2316 m->m_data += max_linkhdr;
2317 (void) memcpy(mtod(m, void *), ip, sizeof (struct ip));
2318 } else {
2319 m->m_data -= optlen;
2320 m->m_len += optlen;
2321 m->m_pkthdr.len += optlen;
2322 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof (struct ip));
2323 }
2324 ip = mtod(m, struct ip *);
2325 bcopy(p->ipopt_list, ip + 1, optlen);
2326 *phlen = sizeof (struct ip) + optlen;
2327 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2);
2328 ip->ip_len += optlen;
2329 return (m);
2330 }
2331
2332 /*
2333 * Copy options from ip to jp,
2334 * omitting those not copied during fragmentation.
2335 */
2336 static int
2337 ip_optcopy(struct ip *ip, struct ip *jp)
2338 {
2339 u_char *cp, *dp;
2340 int opt, optlen, cnt;
2341
2342 cp = (u_char *)(ip + 1);
2343 dp = (u_char *)(jp + 1);
2344 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
2345 for (; cnt > 0; cnt -= optlen, cp += optlen) {
2346 opt = cp[0];
2347 if (opt == IPOPT_EOL)
2348 break;
2349 if (opt == IPOPT_NOP) {
2350 /* Preserve for IP mcast tunnel's LSRR alignment. */
2351 *dp++ = IPOPT_NOP;
2352 optlen = 1;
2353 continue;
2354 }
2355 #if DIAGNOSTIC
2356 if (cnt < IPOPT_OLEN + sizeof (*cp)) {
2357 panic("malformed IPv4 option passed to ip_optcopy");
2358 /* NOTREACHED */
2359 }
2360 #endif
2361 optlen = cp[IPOPT_OLEN];
2362 #if DIAGNOSTIC
2363 if (optlen < IPOPT_OLEN + sizeof (*cp) || optlen > cnt) {
2364 panic("malformed IPv4 option passed to ip_optcopy");
2365 /* NOTREACHED */
2366 }
2367 #endif
2368 /* bogus lengths should have been caught by ip_dooptions */
2369 if (optlen > cnt)
2370 optlen = cnt;
2371 if (IPOPT_COPIED(opt)) {
2372 bcopy(cp, dp, optlen);
2373 dp += optlen;
2374 }
2375 }
2376 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
2377 *dp++ = IPOPT_EOL;
2378 return (optlen);
2379 }
2380
2381 /*
2382 * IP socket option processing.
2383 */
2384 int
2385 ip_ctloutput(struct socket *so, struct sockopt *sopt)
2386 {
2387 struct inpcb *inp = sotoinpcb(so);
2388 int error, optval;
2389
2390 error = optval = 0;
2391 if (sopt->sopt_level != IPPROTO_IP)
2392 return (EINVAL);
2393
2394 switch (sopt->sopt_dir) {
2395 case SOPT_SET:
2396 switch (sopt->sopt_name) {
2397 #ifdef notyet
2398 case IP_RETOPTS:
2399 #endif
2400 case IP_OPTIONS: {
2401 struct mbuf *m;
2402
2403 if (sopt->sopt_valsize > MLEN) {
2404 error = EMSGSIZE;
2405 break;
2406 }
2407 MGET(m, sopt->sopt_p != kernproc ? M_WAIT : M_DONTWAIT,
2408 MT_HEADER);
2409 if (m == NULL) {
2410 error = ENOBUFS;
2411 break;
2412 }
2413 m->m_len = sopt->sopt_valsize;
2414 error = sooptcopyin(sopt, mtod(m, char *),
2415 m->m_len, m->m_len);
2416 if (error)
2417 break;
2418
2419 return (ip_pcbopts(sopt->sopt_name,
2420 &inp->inp_options, m));
2421 }
2422
2423 case IP_TOS:
2424 case IP_TTL:
2425 case IP_RECVOPTS:
2426 case IP_RECVRETOPTS:
2427 case IP_RECVDSTADDR:
2428 case IP_RECVIF:
2429 case IP_RECVTTL:
2430 case IP_RECVPKTINFO:
2431 error = sooptcopyin(sopt, &optval, sizeof (optval),
2432 sizeof (optval));
2433 if (error)
2434 break;
2435
2436 switch (sopt->sopt_name) {
2437 case IP_TOS:
2438 inp->inp_ip_tos = optval;
2439 break;
2440
2441 case IP_TTL:
2442 inp->inp_ip_ttl = optval;
2443 break;
2444 #define OPTSET(bit) \
2445 if (optval) \
2446 inp->inp_flags |= bit; \
2447 else \
2448 inp->inp_flags &= ~bit;
2449
2450 case IP_RECVOPTS:
2451 OPTSET(INP_RECVOPTS);
2452 break;
2453
2454 case IP_RECVRETOPTS:
2455 OPTSET(INP_RECVRETOPTS);
2456 break;
2457
2458 case IP_RECVDSTADDR:
2459 OPTSET(INP_RECVDSTADDR);
2460 break;
2461
2462 case IP_RECVIF:
2463 OPTSET(INP_RECVIF);
2464 break;
2465
2466 case IP_RECVTTL:
2467 OPTSET(INP_RECVTTL);
2468 break;
2469
2470 case IP_RECVPKTINFO:
2471 OPTSET(INP_PKTINFO);
2472 break;
2473 }
2474 break;
2475 #undef OPTSET
2476
2477 #if CONFIG_FORCE_OUT_IFP
2478 /*
2479 * Apple private interface, similar to IP_BOUND_IF, except
2480 * that the parameter is a NULL-terminated string containing
2481 * the name of the network interface; an emptry string means
2482 * unbind. Applications are encouraged to use IP_BOUND_IF
2483 * instead, as that is the current "official" API.
2484 */
2485 case IP_FORCE_OUT_IFP: {
2486 char ifname[IFNAMSIZ];
2487 unsigned int ifscope;
2488
2489 /* This option is settable only for IPv4 */
2490 if (!(inp->inp_vflag & INP_IPV4)) {
2491 error = EINVAL;
2492 break;
2493 }
2494
2495 /* Verify interface name parameter is sane */
2496 if (sopt->sopt_valsize > sizeof (ifname)) {
2497 error = EINVAL;
2498 break;
2499 }
2500
2501 /* Copy the interface name */
2502 if (sopt->sopt_valsize != 0) {
2503 error = sooptcopyin(sopt, ifname,
2504 sizeof (ifname), sopt->sopt_valsize);
2505 if (error)
2506 break;
2507 }
2508
2509 if (sopt->sopt_valsize == 0 || ifname[0] == '\0') {
2510 /* Unbind this socket from any interface */
2511 ifscope = IFSCOPE_NONE;
2512 } else {
2513 ifnet_t ifp;
2514
2515 /* Verify name is NULL terminated */
2516 if (ifname[sopt->sopt_valsize - 1] != '\0') {
2517 error = EINVAL;
2518 break;
2519 }
2520
2521 /* Bail out if given bogus interface name */
2522 if (ifnet_find_by_name(ifname, &ifp) != 0) {
2523 error = ENXIO;
2524 break;
2525 }
2526
2527 /* Bind this socket to this interface */
2528 ifscope = ifp->if_index;
2529
2530 /*
2531 * Won't actually free; since we don't release
2532 * this later, we should do it now.
2533 */
2534 ifnet_release(ifp);
2535 }
2536 error = inp_bindif(inp, ifscope, NULL);
2537 }
2538 break;
2539 #endif /* CONFIG_FORCE_OUT_IFP */
2540 /*
2541 * Multicast socket options are processed by the in_mcast
2542 * module.
2543 */
2544 case IP_MULTICAST_IF:
2545 case IP_MULTICAST_IFINDEX:
2546 case IP_MULTICAST_VIF:
2547 case IP_MULTICAST_TTL:
2548 case IP_MULTICAST_LOOP:
2549 case IP_ADD_MEMBERSHIP:
2550 case IP_DROP_MEMBERSHIP:
2551 case IP_ADD_SOURCE_MEMBERSHIP:
2552 case IP_DROP_SOURCE_MEMBERSHIP:
2553 case IP_BLOCK_SOURCE:
2554 case IP_UNBLOCK_SOURCE:
2555 case IP_MSFILTER:
2556 case MCAST_JOIN_GROUP:
2557 case MCAST_LEAVE_GROUP:
2558 case MCAST_JOIN_SOURCE_GROUP:
2559 case MCAST_LEAVE_SOURCE_GROUP:
2560 case MCAST_BLOCK_SOURCE:
2561 case MCAST_UNBLOCK_SOURCE:
2562 error = inp_setmoptions(inp, sopt);
2563 break;
2564
2565 case IP_PORTRANGE:
2566 error = sooptcopyin(sopt, &optval, sizeof (optval),
2567 sizeof (optval));
2568 if (error)
2569 break;
2570
2571 switch (optval) {
2572 case IP_PORTRANGE_DEFAULT:
2573 inp->inp_flags &= ~(INP_LOWPORT);
2574 inp->inp_flags &= ~(INP_HIGHPORT);
2575 break;
2576
2577 case IP_PORTRANGE_HIGH:
2578 inp->inp_flags &= ~(INP_LOWPORT);
2579 inp->inp_flags |= INP_HIGHPORT;
2580 break;
2581
2582 case IP_PORTRANGE_LOW:
2583 inp->inp_flags &= ~(INP_HIGHPORT);
2584 inp->inp_flags |= INP_LOWPORT;
2585 break;
2586
2587 default:
2588 error = EINVAL;
2589 break;
2590 }
2591 break;
2592
2593 #if IPSEC
2594 case IP_IPSEC_POLICY: {
2595 caddr_t req = NULL;
2596 size_t len = 0;
2597 int priv;
2598 struct mbuf *m;
2599 int optname;
2600
2601 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
2602 break;
2603 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
2604 break;
2605 priv = (proc_suser(sopt->sopt_p) == 0);
2606 if (m) {
2607 req = mtod(m, caddr_t);
2608 len = m->m_len;
2609 }
2610 optname = sopt->sopt_name;
2611 error = ipsec4_set_policy(inp, optname, req, len, priv);
2612 m_freem(m);
2613 break;
2614 }
2615 #endif /* IPSEC */
2616
2617 #if TRAFFIC_MGT
2618 case IP_TRAFFIC_MGT_BACKGROUND: {
2619 unsigned background = 0;
2620
2621 error = sooptcopyin(sopt, &background,
2622 sizeof (background), sizeof (background));
2623 if (error)
2624 break;
2625
2626 if (background) {
2627 socket_set_traffic_mgt_flags_locked(so,
2628 TRAFFIC_MGT_SO_BACKGROUND);
2629 } else {
2630 socket_clear_traffic_mgt_flags_locked(so,
2631 TRAFFIC_MGT_SO_BACKGROUND);
2632 }
2633
2634 break;
2635 }
2636 #endif /* TRAFFIC_MGT */
2637
2638 /*
2639 * On a multihomed system, scoped routing can be used to
2640 * restrict the source interface used for sending packets.
2641 * The socket option IP_BOUND_IF binds a particular AF_INET
2642 * socket to an interface such that data sent on the socket
2643 * is restricted to that interface. This is unlike the
2644 * SO_DONTROUTE option where the routing table is bypassed;
2645 * therefore it allows for a greater flexibility and control
2646 * over the system behavior, and does not place any restriction
2647 * on the destination address type (e.g. unicast, multicast,
2648 * or broadcast if applicable) or whether or not the host is
2649 * directly reachable. Note that in the multicast transmit
2650 * case, IP_MULTICAST_{IF,IFINDEX} takes precedence over
2651 * IP_BOUND_IF, since the former practically bypasses the
2652 * routing table; in this case, IP_BOUND_IF sets the default
2653 * interface used for sending multicast packets in the absence
2654 * of an explicit multicast transmit interface.
2655 */
2656 case IP_BOUND_IF:
2657 /* This option is settable only for IPv4 */
2658 if (!(inp->inp_vflag & INP_IPV4)) {
2659 error = EINVAL;
2660 break;
2661 }
2662
2663 error = sooptcopyin(sopt, &optval, sizeof (optval),
2664 sizeof (optval));
2665
2666 if (error)
2667 break;
2668
2669 error = inp_bindif(inp, optval, NULL);
2670 break;
2671
2672 case IP_NO_IFT_CELLULAR:
2673 /* This option is settable only for IPv4 */
2674 if (!(inp->inp_vflag & INP_IPV4)) {
2675 error = EINVAL;
2676 break;
2677 }
2678
2679 error = sooptcopyin(sopt, &optval, sizeof (optval),
2680 sizeof (optval));
2681
2682 if (error)
2683 break;
2684
2685 /* once set, it cannot be unset */
2686 if (!optval && INP_NO_CELLULAR(inp)) {
2687 error = EINVAL;
2688 break;
2689 }
2690
2691 error = so_set_restrictions(so,
2692 SO_RESTRICT_DENY_CELLULAR);
2693 break;
2694
2695 case IP_OUT_IF:
2696 /* This option is not settable */
2697 error = EINVAL;
2698 break;
2699
2700 default:
2701 error = ENOPROTOOPT;
2702 break;
2703 }
2704 break;
2705
2706 case SOPT_GET:
2707 switch (sopt->sopt_name) {
2708 case IP_OPTIONS:
2709 case IP_RETOPTS:
2710 if (inp->inp_options) {
2711 error = sooptcopyout(sopt,
2712 mtod(inp->inp_options, char *),
2713 inp->inp_options->m_len);
2714 } else {
2715 sopt->sopt_valsize = 0;
2716 }
2717 break;
2718
2719 case IP_TOS:
2720 case IP_TTL:
2721 case IP_RECVOPTS:
2722 case IP_RECVRETOPTS:
2723 case IP_RECVDSTADDR:
2724 case IP_RECVIF:
2725 case IP_RECVTTL:
2726 case IP_PORTRANGE:
2727 case IP_RECVPKTINFO:
2728 switch (sopt->sopt_name) {
2729
2730 case IP_TOS:
2731 optval = inp->inp_ip_tos;
2732 break;
2733
2734 case IP_TTL:
2735 optval = inp->inp_ip_ttl;
2736 break;
2737
2738 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
2739
2740 case IP_RECVOPTS:
2741 optval = OPTBIT(INP_RECVOPTS);
2742 break;
2743
2744 case IP_RECVRETOPTS:
2745 optval = OPTBIT(INP_RECVRETOPTS);
2746 break;
2747
2748 case IP_RECVDSTADDR:
2749 optval = OPTBIT(INP_RECVDSTADDR);
2750 break;
2751
2752 case IP_RECVIF:
2753 optval = OPTBIT(INP_RECVIF);
2754 break;
2755
2756 case IP_RECVTTL:
2757 optval = OPTBIT(INP_RECVTTL);
2758 break;
2759
2760 case IP_PORTRANGE:
2761 if (inp->inp_flags & INP_HIGHPORT)
2762 optval = IP_PORTRANGE_HIGH;
2763 else if (inp->inp_flags & INP_LOWPORT)
2764 optval = IP_PORTRANGE_LOW;
2765 else
2766 optval = 0;
2767 break;
2768
2769 case IP_RECVPKTINFO:
2770 optval = OPTBIT(INP_PKTINFO);
2771 break;
2772 }
2773 error = sooptcopyout(sopt, &optval, sizeof (optval));
2774 break;
2775
2776 case IP_MULTICAST_IF:
2777 case IP_MULTICAST_IFINDEX:
2778 case IP_MULTICAST_VIF:
2779 case IP_MULTICAST_TTL:
2780 case IP_MULTICAST_LOOP:
2781 case IP_MSFILTER:
2782 error = inp_getmoptions(inp, sopt);
2783 break;
2784
2785 #if IPSEC
2786 case IP_IPSEC_POLICY: {
2787 error = 0; /* This option is no longer supported */
2788 break;
2789 }
2790 #endif /* IPSEC */
2791
2792 #if TRAFFIC_MGT
2793 case IP_TRAFFIC_MGT_BACKGROUND: {
2794 unsigned background = (so->so_traffic_mgt_flags &
2795 TRAFFIC_MGT_SO_BACKGROUND) ? 1 : 0;
2796 return (sooptcopyout(sopt, &background,
2797 sizeof (background)));
2798 break;
2799 }
2800 #endif /* TRAFFIC_MGT */
2801
2802 case IP_BOUND_IF:
2803 if (inp->inp_flags & INP_BOUND_IF)
2804 optval = inp->inp_boundifp->if_index;
2805 error = sooptcopyout(sopt, &optval, sizeof (optval));
2806 break;
2807
2808 case IP_NO_IFT_CELLULAR:
2809 optval = INP_NO_CELLULAR(inp) ? 1 : 0;
2810 error = sooptcopyout(sopt, &optval, sizeof (optval));
2811 break;
2812
2813 case IP_OUT_IF:
2814 optval = (inp->inp_last_outifp != NULL) ?
2815 inp->inp_last_outifp->if_index : 0;
2816 error = sooptcopyout(sopt, &optval, sizeof (optval));
2817 break;
2818
2819 default:
2820 error = ENOPROTOOPT;
2821 break;
2822 }
2823 break;
2824 }
2825 return (error);
2826 }
2827
2828 /*
2829 * Set up IP options in pcb for insertion in output packets.
2830 * Store in mbuf with pointer in pcbopt, adding pseudo-option
2831 * with destination address if source routed.
2832 */
2833 static int
2834 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m)
2835 {
2836 #pragma unused(optname)
2837 int cnt, optlen;
2838 u_char *cp;
2839 u_char opt;
2840
2841 /* turn off any old options */
2842 if (*pcbopt)
2843 (void) m_free(*pcbopt);
2844 *pcbopt = 0;
2845 if (m == (struct mbuf *)0 || m->m_len == 0) {
2846 /*
2847 * Only turning off any previous options.
2848 */
2849 if (m)
2850 (void) m_free(m);
2851 return (0);
2852 }
2853
2854 if (m->m_len % sizeof (int32_t))
2855 goto bad;
2856
2857 /*
2858 * IP first-hop destination address will be stored before
2859 * actual options; move other options back
2860 * and clear it when none present.
2861 */
2862 if (m->m_data + m->m_len + sizeof (struct in_addr) >= &m->m_dat[MLEN])
2863 goto bad;
2864 cnt = m->m_len;
2865 m->m_len += sizeof (struct in_addr);
2866 cp = mtod(m, u_char *) + sizeof (struct in_addr);
2867 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
2868 bzero(mtod(m, caddr_t), sizeof (struct in_addr));
2869
2870 for (; cnt > 0; cnt -= optlen, cp += optlen) {
2871 opt = cp[IPOPT_OPTVAL];
2872 if (opt == IPOPT_EOL)
2873 break;
2874 if (opt == IPOPT_NOP)
2875 optlen = 1;
2876 else {
2877 if (cnt < IPOPT_OLEN + sizeof (*cp))
2878 goto bad;
2879 optlen = cp[IPOPT_OLEN];
2880 if (optlen < IPOPT_OLEN + sizeof (*cp) || optlen > cnt)
2881 goto bad;
2882 }
2883 switch (opt) {
2884
2885 default:
2886 break;
2887
2888 case IPOPT_LSRR:
2889 case IPOPT_SSRR:
2890 /*
2891 * user process specifies route as:
2892 * ->A->B->C->D
2893 * D must be our final destination (but we can't
2894 * check that since we may not have connected yet).
2895 * A is first hop destination, which doesn't appear in
2896 * actual IP option, but is stored before the options.
2897 */
2898 if (optlen < IPOPT_MINOFF - 1 + sizeof (struct in_addr))
2899 goto bad;
2900 m->m_len -= sizeof (struct in_addr);
2901 cnt -= sizeof (struct in_addr);
2902 optlen -= sizeof (struct in_addr);
2903 cp[IPOPT_OLEN] = optlen;
2904 /*
2905 * Move first hop before start of options.
2906 */
2907 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
2908 sizeof (struct in_addr));
2909 /*
2910 * Then copy rest of options back
2911 * to close up the deleted entry.
2912 */
2913 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
2914 sizeof (struct in_addr)),
2915 (caddr_t)&cp[IPOPT_OFFSET+1],
2916 (unsigned)cnt + sizeof (struct in_addr));
2917 break;
2918 }
2919 }
2920 if (m->m_len > MAX_IPOPTLEN + sizeof (struct in_addr))
2921 goto bad;
2922 *pcbopt = m;
2923 return (0);
2924
2925 bad:
2926 (void) m_free(m);
2927 return (EINVAL);
2928 }
2929
2930 void
2931 ip_moptions_init(void)
2932 {
2933 PE_parse_boot_argn("ifa_debug", &imo_debug, sizeof (imo_debug));
2934
2935 imo_size = (imo_debug == 0) ? sizeof (struct ip_moptions) :
2936 sizeof (struct ip_moptions_dbg);
2937
2938 imo_zone = zinit(imo_size, IMO_ZONE_MAX * imo_size, 0,
2939 IMO_ZONE_NAME);
2940 if (imo_zone == NULL) {
2941 panic("%s: failed allocating %s", __func__, IMO_ZONE_NAME);
2942 /* NOTREACHED */
2943 }
2944 zone_change(imo_zone, Z_EXPAND, TRUE);
2945 }
2946
2947 void
2948 imo_addref(struct ip_moptions *imo, int locked)
2949 {
2950 if (!locked)
2951 IMO_LOCK(imo);
2952 else
2953 IMO_LOCK_ASSERT_HELD(imo);
2954
2955 if (++imo->imo_refcnt == 0) {
2956 panic("%s: imo %p wraparound refcnt\n", __func__, imo);
2957 /* NOTREACHED */
2958 } else if (imo->imo_trace != NULL) {
2959 (*imo->imo_trace)(imo, TRUE);
2960 }
2961
2962 if (!locked)
2963 IMO_UNLOCK(imo);
2964 }
2965
2966 void
2967 imo_remref(struct ip_moptions *imo)
2968 {
2969 int i;
2970
2971 IMO_LOCK(imo);
2972 if (imo->imo_refcnt == 0) {
2973 panic("%s: imo %p negative refcnt", __func__, imo);
2974 /* NOTREACHED */
2975 } else if (imo->imo_trace != NULL) {
2976 (*imo->imo_trace)(imo, FALSE);
2977 }
2978
2979 --imo->imo_refcnt;
2980 if (imo->imo_refcnt > 0) {
2981 IMO_UNLOCK(imo);
2982 return;
2983 }
2984
2985 for (i = 0; i < imo->imo_num_memberships; ++i) {
2986 struct in_mfilter *imf;
2987
2988 imf = imo->imo_mfilters ? &imo->imo_mfilters[i] : NULL;
2989 if (imf != NULL)
2990 imf_leave(imf);
2991
2992 (void) in_leavegroup(imo->imo_membership[i], imf);
2993
2994 if (imf != NULL)
2995 imf_purge(imf);
2996
2997 INM_REMREF(imo->imo_membership[i]);
2998 imo->imo_membership[i] = NULL;
2999 }
3000 imo->imo_num_memberships = 0;
3001 if (imo->imo_mfilters != NULL) {
3002 FREE(imo->imo_mfilters, M_INMFILTER);
3003 imo->imo_mfilters = NULL;
3004 }
3005 if (imo->imo_membership != NULL) {
3006 FREE(imo->imo_membership, M_IPMOPTS);
3007 imo->imo_membership = NULL;
3008 }
3009 IMO_UNLOCK(imo);
3010
3011 lck_mtx_destroy(&imo->imo_lock, ifa_mtx_grp);
3012
3013 if (!(imo->imo_debug & IFD_ALLOC)) {
3014 panic("%s: imo %p cannot be freed", __func__, imo);
3015 /* NOTREACHED */
3016 }
3017 zfree(imo_zone, imo);
3018 }
3019
3020 static void
3021 imo_trace(struct ip_moptions *imo, int refhold)
3022 {
3023 struct ip_moptions_dbg *imo_dbg = (struct ip_moptions_dbg *)imo;
3024 ctrace_t *tr;
3025 u_int32_t idx;
3026 u_int16_t *cnt;
3027
3028 if (!(imo->imo_debug & IFD_DEBUG)) {
3029 panic("%s: imo %p has no debug structure", __func__, imo);
3030 /* NOTREACHED */
3031 }
3032 if (refhold) {
3033 cnt = &imo_dbg->imo_refhold_cnt;
3034 tr = imo_dbg->imo_refhold;
3035 } else {
3036 cnt = &imo_dbg->imo_refrele_cnt;
3037 tr = imo_dbg->imo_refrele;
3038 }
3039
3040 idx = atomic_add_16_ov(cnt, 1) % IMO_TRACE_HIST_SIZE;
3041 ctrace_record(&tr[idx]);
3042 }
3043
3044 struct ip_moptions *
3045 ip_allocmoptions(int how)
3046 {
3047 struct ip_moptions *imo;
3048
3049 imo = (how == M_WAITOK) ? zalloc(imo_zone) : zalloc_noblock(imo_zone);
3050 if (imo != NULL) {
3051 bzero(imo, imo_size);
3052 lck_mtx_init(&imo->imo_lock, ifa_mtx_grp, ifa_mtx_attr);
3053 imo->imo_debug |= IFD_ALLOC;
3054 if (imo_debug != 0) {
3055 imo->imo_debug |= IFD_DEBUG;
3056 imo->imo_trace = imo_trace;
3057 }
3058 IMO_ADDREF(imo);
3059 }
3060
3061 return (imo);
3062 }
3063
3064 /*
3065 * Routine called from ip_output() to loop back a copy of an IP multicast
3066 * packet to the input queue of a specified interface. Note that this
3067 * calls the output routine of the loopback "driver", but with an interface
3068 * pointer that might NOT be a loopback interface -- evil, but easier than
3069 * replicating that code here.
3070 */
3071 static void
3072 ip_mloopback(struct ifnet *srcifp, struct ifnet *origifp, struct mbuf *m,
3073 struct sockaddr_in *dst, int hlen)
3074 {
3075 struct mbuf *copym;
3076 struct ip *ip;
3077
3078 if (lo_ifp == NULL)
3079 return;
3080
3081 /*
3082 * Copy the packet header as it's needed for the checksum
3083 * Make sure to deep-copy IP header portion in case the data
3084 * is in an mbuf cluster, so that we can safely override the IP
3085 * header portion later.
3086 */
3087 copym = m_copym_mode(m, 0, M_COPYALL, M_DONTWAIT, M_COPYM_COPY_HDR);
3088 if (copym != NULL && ((copym->m_flags & M_EXT) || copym->m_len < hlen))
3089 copym = m_pullup(copym, hlen);
3090
3091 if (copym == NULL)
3092 return;
3093
3094 /*
3095 * We don't bother to fragment if the IP length is greater
3096 * than the interface's MTU. Can this possibly matter?
3097 */
3098 ip = mtod(copym, struct ip *);
3099 #if BYTE_ORDER != BIG_ENDIAN
3100 HTONS(ip->ip_len);
3101 HTONS(ip->ip_off);
3102 #endif
3103 ip->ip_sum = 0;
3104 ip->ip_sum = ip_cksum_hdr_out(copym, hlen);
3105
3106 /*
3107 * Mark checksum as valid unless receive checksum offload is
3108 * disabled; if so, compute checksum in software. If the
3109 * interface itself is lo0, this will be overridden by if_loop.
3110 */
3111 if (hwcksum_rx) {
3112 copym->m_pkthdr.csum_flags &= ~CSUM_PARTIAL;
3113 copym->m_pkthdr.csum_flags |=
3114 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
3115 copym->m_pkthdr.csum_data = 0xffff;
3116 } else if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3117 #if BYTE_ORDER != BIG_ENDIAN
3118 NTOHS(ip->ip_len);
3119 #endif
3120 in_delayed_cksum(copym);
3121 #if BYTE_ORDER != BIG_ENDIAN
3122 HTONS(ip->ip_len);
3123 #endif
3124 }
3125
3126 /*
3127 * Stuff the 'real' ifp into the pkthdr, to be used in matching
3128 * in ip_input(); we need the loopback ifp/dl_tag passed as args
3129 * to make the loopback driver compliant with the data link
3130 * requirements.
3131 */
3132 copym->m_pkthdr.rcvif = origifp;
3133
3134 /*
3135 * Also record the source interface (which owns the source address).
3136 * This is basically a stripped down version of ifa_foraddr().
3137 */
3138 if (srcifp == NULL) {
3139 struct in_ifaddr *ia;
3140
3141 lck_rw_lock_shared(in_ifaddr_rwlock);
3142 TAILQ_FOREACH(ia, INADDR_HASH(ip->ip_src.s_addr), ia_hash) {
3143 IFA_LOCK_SPIN(&ia->ia_ifa);
3144 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_src.s_addr) {
3145 srcifp = ia->ia_ifp;
3146 IFA_UNLOCK(&ia->ia_ifa);
3147 break;
3148 }
3149 IFA_UNLOCK(&ia->ia_ifa);
3150 }
3151 lck_rw_done(in_ifaddr_rwlock);
3152 }
3153 if (srcifp != NULL)
3154 ip_setsrcifaddr_info(copym, srcifp->if_index, NULL);
3155 ip_setdstifaddr_info(copym, origifp->if_index, NULL);
3156
3157 dlil_output(lo_ifp, PF_INET, copym, NULL, SA(dst), 0, NULL);
3158 }
3159
3160 /*
3161 * Given a source IP address (and route, if available), determine the best
3162 * interface to send the packet from. Checking for (and updating) the
3163 * ROF_SRCIF_SELECTED flag in the pcb-supplied route placeholder is done
3164 * without any locks based on the assumption that ip_output() is single-
3165 * threaded per-pcb, i.e. for any given pcb there can only be one thread
3166 * performing output at the IP layer.
3167 *
3168 * This routine is analogous to in6_selectroute() for IPv6.
3169 */
3170 static struct ifaddr *
3171 in_selectsrcif(struct ip *ip, struct route *ro, unsigned int ifscope)
3172 {
3173 struct ifaddr *ifa = NULL;
3174 struct in_addr src = ip->ip_src;
3175 struct in_addr dst = ip->ip_dst;
3176 struct ifnet *rt_ifp;
3177 char s_src[MAX_IPv4_STR_LEN], s_dst[MAX_IPv4_STR_LEN];
3178
3179 VERIFY(src.s_addr != INADDR_ANY);
3180
3181 if (ip_select_srcif_debug) {
3182 (void) inet_ntop(AF_INET, &src.s_addr, s_src, sizeof (s_src));
3183 (void) inet_ntop(AF_INET, &dst.s_addr, s_dst, sizeof (s_dst));
3184 }
3185
3186 if (ro->ro_rt != NULL)
3187 RT_LOCK(ro->ro_rt);
3188
3189 rt_ifp = (ro->ro_rt != NULL) ? ro->ro_rt->rt_ifp : NULL;
3190
3191 /*
3192 * Given the source IP address, find a suitable source interface
3193 * to use for transmission; if the caller has specified a scope,
3194 * optimize the search by looking at the addresses only for that
3195 * interface. This is still suboptimal, however, as we need to
3196 * traverse the per-interface list.
3197 */
3198 if (ifscope != IFSCOPE_NONE || ro->ro_rt != NULL) {
3199 unsigned int scope = ifscope;
3200
3201 /*
3202 * If no scope is specified and the route is stale (pointing
3203 * to a defunct interface) use the current primary interface;
3204 * this happens when switching between interfaces configured
3205 * with the same IP address. Otherwise pick up the scope
3206 * information from the route; the ULP may have looked up a
3207 * correct route and we just need to verify it here and mark
3208 * it with the ROF_SRCIF_SELECTED flag below.
3209 */
3210 if (scope == IFSCOPE_NONE) {
3211 scope = rt_ifp->if_index;
3212 if (scope != get_primary_ifscope(AF_INET) &&
3213 ROUTE_UNUSABLE(ro))
3214 scope = get_primary_ifscope(AF_INET);
3215 }
3216
3217 ifa = (struct ifaddr *)ifa_foraddr_scoped(src.s_addr, scope);
3218
3219 if (ifa == NULL && ip->ip_p != IPPROTO_UDP &&
3220 ip->ip_p != IPPROTO_TCP && ipforwarding) {
3221 /*
3222 * If forwarding is enabled, and if the packet isn't
3223 * TCP or UDP, check if the source address belongs
3224 * to one of our own interfaces; if so, demote the
3225 * interface scope and do a route lookup right below.
3226 */
3227 ifa = (struct ifaddr *)ifa_foraddr(src.s_addr);
3228 if (ifa != NULL) {
3229 IFA_REMREF(ifa);
3230 ifa = NULL;
3231 ifscope = IFSCOPE_NONE;
3232 }
3233 }
3234
3235 if (ip_select_srcif_debug && ifa != NULL) {
3236 if (ro->ro_rt != NULL) {
3237 printf("%s->%s ifscope %d->%d ifa_if %s "
3238 "ro_if %s\n", s_src, s_dst, ifscope,
3239 scope, if_name(ifa->ifa_ifp),
3240 if_name(rt_ifp));
3241 } else {
3242 printf("%s->%s ifscope %d->%d ifa_if %s\n",
3243 s_src, s_dst, ifscope, scope,
3244 if_name(ifa->ifa_ifp));
3245 }
3246 }
3247 }
3248
3249 /*
3250 * Slow path; search for an interface having the corresponding source
3251 * IP address if the scope was not specified by the caller, and:
3252 *
3253 * 1) There currently isn't any route, or,
3254 * 2) The interface used by the route does not own that source
3255 * IP address; in this case, the route will get blown away
3256 * and we'll do a more specific scoped search using the newly
3257 * found interface.
3258 */
3259 if (ifa == NULL && ifscope == IFSCOPE_NONE) {
3260 ifa = (struct ifaddr *)ifa_foraddr(src.s_addr);
3261
3262 /*
3263 * If we have the IP address, but not the route, we don't
3264 * really know whether or not it belongs to the correct
3265 * interface (it could be shared across multiple interfaces.)
3266 * The only way to find out is to do a route lookup.
3267 */
3268 if (ifa != NULL && ro->ro_rt == NULL) {
3269 struct rtentry *rt;
3270 struct sockaddr_in sin;
3271 struct ifaddr *oifa = NULL;
3272
3273 bzero(&sin, sizeof (sin));
3274 sin.sin_family = AF_INET;
3275 sin.sin_len = sizeof (sin);
3276 sin.sin_addr = dst;
3277
3278 lck_mtx_lock(rnh_lock);
3279 if ((rt = rt_lookup(TRUE, SA(&sin), NULL,
3280 rt_tables[AF_INET], IFSCOPE_NONE)) != NULL) {
3281 RT_LOCK(rt);
3282 /*
3283 * If the route uses a different interface,
3284 * use that one instead. The IP address of
3285 * the ifaddr that we pick up here is not
3286 * relevant.
3287 */
3288 if (ifa->ifa_ifp != rt->rt_ifp) {
3289 oifa = ifa;
3290 ifa = rt->rt_ifa;
3291 IFA_ADDREF(ifa);
3292 RT_UNLOCK(rt);
3293 } else {
3294 RT_UNLOCK(rt);
3295 }
3296 rtfree_locked(rt);
3297 }
3298 lck_mtx_unlock(rnh_lock);
3299
3300 if (oifa != NULL) {
3301 struct ifaddr *iifa;
3302
3303 /*
3304 * See if the interface pointed to by the
3305 * route is configured with the source IP
3306 * address of the packet.
3307 */
3308 iifa = (struct ifaddr *)ifa_foraddr_scoped(
3309 src.s_addr, ifa->ifa_ifp->if_index);
3310
3311 if (iifa != NULL) {
3312 /*
3313 * Found it; drop the original one
3314 * as well as the route interface
3315 * address, and use this instead.
3316 */
3317 IFA_REMREF(oifa);
3318 IFA_REMREF(ifa);
3319 ifa = iifa;
3320 } else if (!ipforwarding ||
3321 (rt->rt_flags & RTF_GATEWAY)) {
3322 /*
3323 * This interface doesn't have that
3324 * source IP address; drop the route
3325 * interface address and just use the
3326 * original one, and let the caller
3327 * do a scoped route lookup.
3328 */
3329 IFA_REMREF(ifa);
3330 ifa = oifa;
3331 } else {
3332 /*
3333 * Forwarding is enabled and the source
3334 * address belongs to one of our own
3335 * interfaces which isn't the outgoing
3336 * interface, and we have a route, and
3337 * the destination is on a network that
3338 * is directly attached (onlink); drop
3339 * the original one and use the route
3340 * interface address instead.
3341 */
3342 IFA_REMREF(oifa);
3343 }
3344 }
3345 } else if (ifa != NULL && ro->ro_rt != NULL &&
3346 !(ro->ro_rt->rt_flags & RTF_GATEWAY) &&
3347 ifa->ifa_ifp != ro->ro_rt->rt_ifp && ipforwarding) {
3348 /*
3349 * Forwarding is enabled and the source address belongs
3350 * to one of our own interfaces which isn't the same
3351 * as the interface used by the known route; drop the
3352 * original one and use the route interface address.
3353 */
3354 IFA_REMREF(ifa);
3355 ifa = ro->ro_rt->rt_ifa;
3356 IFA_ADDREF(ifa);
3357 }
3358
3359 if (ip_select_srcif_debug && ifa != NULL) {
3360 printf("%s->%s ifscope %d ifa_if %s\n",
3361 s_src, s_dst, ifscope, if_name(ifa->ifa_ifp));
3362 }
3363 }
3364
3365 if (ro->ro_rt != NULL)
3366 RT_LOCK_ASSERT_HELD(ro->ro_rt);
3367 /*
3368 * If there is a non-loopback route with the wrong interface, or if
3369 * there is no interface configured with such an address, blow it
3370 * away. Except for local/loopback, we look for one with a matching
3371 * interface scope/index.
3372 */
3373 if (ro->ro_rt != NULL &&
3374 (ifa == NULL || (ifa->ifa_ifp != rt_ifp && rt_ifp != lo_ifp) ||
3375 !(ro->ro_rt->rt_flags & RTF_UP))) {
3376 if (ip_select_srcif_debug) {
3377 if (ifa != NULL) {
3378 printf("%s->%s ifscope %d ro_if %s != "
3379 "ifa_if %s (cached route cleared)\n",
3380 s_src, s_dst, ifscope, if_name(rt_ifp),
3381 if_name(ifa->ifa_ifp));
3382 } else {
3383 printf("%s->%s ifscope %d ro_if %s "
3384 "(no ifa_if found)\n",
3385 s_src, s_dst, ifscope, if_name(rt_ifp));
3386 }
3387 }
3388
3389 RT_UNLOCK(ro->ro_rt);
3390 ROUTE_RELEASE(ro);
3391
3392 /*
3393 * If the destination is IPv4 LLA and the route's interface
3394 * doesn't match the source interface, then the source IP
3395 * address is wrong; it most likely belongs to the primary
3396 * interface associated with the IPv4 LL subnet. Drop the
3397 * packet rather than letting it go out and return an error
3398 * to the ULP. This actually applies not only to IPv4 LL
3399 * but other shared subnets; for now we explicitly test only
3400 * for the former case and save the latter for future.
3401 */
3402 if (IN_LINKLOCAL(ntohl(dst.s_addr)) &&
3403 !IN_LINKLOCAL(ntohl(src.s_addr)) && ifa != NULL) {
3404 IFA_REMREF(ifa);
3405 ifa = NULL;
3406 }
3407 }
3408
3409 if (ip_select_srcif_debug && ifa == NULL) {
3410 printf("%s->%s ifscope %d (neither ro_if/ifa_if found)\n",
3411 s_src, s_dst, ifscope);
3412 }
3413
3414 /*
3415 * If there is a route, mark it accordingly. If there isn't one,
3416 * we'll get here again during the next transmit (possibly with a
3417 * route) and the flag will get set at that point. For IPv4 LLA
3418 * destination, mark it only if the route has been fully resolved;
3419 * otherwise we want to come back here again when the route points
3420 * to the interface over which the ARP reply arrives on.
3421 */
3422 if (ro->ro_rt != NULL && (!IN_LINKLOCAL(ntohl(dst.s_addr)) ||
3423 (ro->ro_rt->rt_gateway->sa_family == AF_LINK &&
3424 SDL(ro->ro_rt->rt_gateway)->sdl_alen != 0))) {
3425 if (ifa != NULL)
3426 IFA_ADDREF(ifa); /* for route */
3427 if (ro->ro_srcia != NULL)
3428 IFA_REMREF(ro->ro_srcia);
3429 ro->ro_srcia = ifa;
3430 ro->ro_flags |= ROF_SRCIF_SELECTED;
3431 RT_GENID_SYNC(ro->ro_rt);
3432 }
3433
3434 if (ro->ro_rt != NULL)
3435 RT_UNLOCK(ro->ro_rt);
3436
3437 return (ifa);
3438 }
3439
3440 void
3441 ip_output_checksum(struct ifnet *ifp, struct mbuf *m, int hlen, int ip_len,
3442 uint32_t *sw_csum)
3443 {
3444 int tso = TSO_IPV4_OK(ifp, m);
3445 uint32_t hwcap = ifp->if_hwassist;
3446
3447 m->m_pkthdr.csum_flags |= CSUM_IP;
3448
3449 if (!hwcksum_tx) {
3450 /* do all in software; hardware checksum offload is disabled */
3451 *sw_csum = (CSUM_DELAY_DATA | CSUM_DELAY_IP) &
3452 m->m_pkthdr.csum_flags;
3453 } else {
3454 /* do in software what the hardware cannot */
3455 *sw_csum = m->m_pkthdr.csum_flags &
3456 ~IF_HWASSIST_CSUM_FLAGS(hwcap);
3457 }
3458
3459 if (hlen != sizeof (struct ip)) {
3460 *sw_csum |= ((CSUM_DELAY_DATA | CSUM_DELAY_IP) &
3461 m->m_pkthdr.csum_flags);
3462 } else if (!(*sw_csum & CSUM_DELAY_DATA) && (hwcap & CSUM_PARTIAL)) {
3463 /*
3464 * Partial checksum offload, if non-IP fragment, and TCP only
3465 * (no UDP support, as the hardware may not be able to convert
3466 * +0 to -0 (0xffff) per RFC1122 4.1.3.4.)
3467 */
3468 if (hwcksum_tx && !tso &&
3469 (m->m_pkthdr.csum_flags & CSUM_TCP) &&
3470 ip_len <= ifp->if_mtu) {
3471 uint16_t start = sizeof (struct ip);
3472 uint16_t ulpoff = m->m_pkthdr.csum_data & 0xffff;
3473 m->m_pkthdr.csum_flags |=
3474 (CSUM_DATA_VALID | CSUM_PARTIAL);
3475 m->m_pkthdr.csum_tx_stuff = (ulpoff + start);
3476 m->m_pkthdr.csum_tx_start = start;
3477 /* do IP hdr chksum in software */
3478 *sw_csum = CSUM_DELAY_IP;
3479 } else {
3480 *sw_csum |= (CSUM_DELAY_DATA & m->m_pkthdr.csum_flags);
3481 }
3482 }
3483
3484 if (*sw_csum & CSUM_DELAY_DATA) {
3485 in_delayed_cksum(m);
3486 *sw_csum &= ~CSUM_DELAY_DATA;
3487 }
3488
3489 if (hwcksum_tx) {
3490 /*
3491 * Drop off bits that aren't supported by hardware;
3492 * also make sure to preserve non-checksum related bits.
3493 */
3494 m->m_pkthdr.csum_flags =
3495 ((m->m_pkthdr.csum_flags &
3496 (IF_HWASSIST_CSUM_FLAGS(hwcap) | CSUM_DATA_VALID)) |
3497 (m->m_pkthdr.csum_flags & ~IF_HWASSIST_CSUM_MASK));
3498 } else {
3499 /* drop all bits; hardware checksum offload is disabled */
3500 m->m_pkthdr.csum_flags = 0;
3501 }
3502 }
3503
3504 /*
3505 * GRE protocol output for PPP/PPTP
3506 */
3507 int
3508 ip_gre_output(struct mbuf *m)
3509 {
3510 struct route ro;
3511 int error;
3512
3513 bzero(&ro, sizeof (ro));
3514
3515 error = ip_output(m, NULL, &ro, 0, NULL, NULL);
3516
3517 ROUTE_RELEASE(&ro);
3518
3519 return (error);
3520 }
3521
3522 static int
3523 sysctl_reset_ip_output_stats SYSCTL_HANDLER_ARGS
3524 {
3525 #pragma unused(arg1, arg2)
3526 int error, i;
3527
3528 i = ip_output_measure;
3529 error = sysctl_handle_int(oidp, &i, 0, req);
3530 if (error || req->newptr == USER_ADDR_NULL)
3531 goto done;
3532 /* impose bounds */
3533 if (i < 0 || i > 1) {
3534 error = EINVAL;
3535 goto done;
3536 }
3537 if (ip_output_measure != i && i == 1) {
3538 net_perf_initialize(&net_perf, ip_output_measure_bins);
3539 }
3540 ip_output_measure = i;
3541 done:
3542 return (error);
3543 }
3544
3545 static int
3546 sysctl_ip_output_measure_bins SYSCTL_HANDLER_ARGS
3547 {
3548 #pragma unused(arg1, arg2)
3549 int error;
3550 uint64_t i;
3551
3552 i = ip_output_measure_bins;
3553 error = sysctl_handle_quad(oidp, &i, 0, req);
3554 if (error || req->newptr == USER_ADDR_NULL)
3555 goto done;
3556 /* validate data */
3557 if (!net_perf_validate_bins(i)) {
3558 error = EINVAL;
3559 goto done;
3560 }
3561 ip_output_measure_bins = i;
3562 done:
3563 return (error);
3564 }
3565
3566 static int
3567 sysctl_ip_output_getperf SYSCTL_HANDLER_ARGS
3568 {
3569 #pragma unused(oidp, arg1, arg2)
3570 if (req->oldptr == USER_ADDR_NULL)
3571 req->oldlen = (size_t)sizeof (struct ipstat);
3572
3573 return (SYSCTL_OUT(req, &net_perf, MIN(sizeof (net_perf), req->oldlen)));
3574 }
3575
mirror of XNU