xref: /xnu-8796.121.2/bsd/netinet/raw_ip.c (revision c54f35ca767986246321eb901baf8f5ff7923f6a)
1 /*
2  * Copyright (c) 2000-2022 Apple Inc. All rights reserved.
3  *
4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5  *
6  * This file contains Original Code and/or Modifications of Original Code
7  * as defined in and that are subject to the Apple Public Source License
8  * Version 2.0 (the 'License'). You may not use this file except in
9  * compliance with the License. The rights granted to you under the License
10  * may not be used to create, or enable the creation or redistribution of,
11  * unlawful or unlicensed copies of an Apple operating system, or to
12  * circumvent, violate, or enable the circumvention or violation of, any
13  * terms of an Apple operating system software license agreement.
14  *
15  * Please obtain a copy of the License at
16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
17  *
18  * The Original Code and all software distributed under the License are
19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23  * Please see the License for the specific language governing rights and
24  * limitations under the License.
25  *
26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27  */
28 /*
29  * Copyright (c) 1982, 1986, 1988, 1993
30  *	The Regents of the University of California.  All rights reserved.
31  *
32  * Redistribution and use in source and binary forms, with or without
33  * modification, are permitted provided that the following conditions
34  * are met:
35  * 1. Redistributions of source code must retain the above copyright
36  *    notice, this list of conditions and the following disclaimer.
37  * 2. Redistributions in binary form must reproduce the above copyright
38  *    notice, this list of conditions and the following disclaimer in the
39  *    documentation and/or other materials provided with the distribution.
40  * 3. All advertising materials mentioning features or use of this software
41  *    must display the following acknowledgement:
42  *	This product includes software developed by the University of
43  *	California, Berkeley and its contributors.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)raw_ip.c	8.7 (Berkeley) 5/15/95
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 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/kernel.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/mcache.h>
75 #include <sys/proc.h>
76 #include <sys/domain.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/sysctl.h>
81 #include <libkern/OSAtomic.h>
82 #include <kern/zalloc.h>
83 
84 #include <pexpert/pexpert.h>
85 
86 #include <net/if.h>
87 #include <net/net_api_stats.h>
88 #include <net/route.h>
89 #include <net/content_filter.h>
90 
91 #define _IP_VHL
92 #include <netinet/in.h>
93 #include <netinet/in_systm.h>
94 #include <netinet/in_tclass.h>
95 #include <netinet/ip.h>
96 #include <netinet/in_pcb.h>
97 #include <netinet/in_var.h>
98 #include <netinet/ip_var.h>
99 
100 #include <netinet6/in6_pcb.h>
101 
102 
103 #if IPSEC
104 #include <netinet6/ipsec.h>
105 #endif /*IPSEC*/
106 
107 #if DUMMYNET
108 #include <netinet/ip_dummynet.h>
109 #endif /* DUMMYNET */
110 
111 int rip_detach(struct socket *);
112 int rip_abort(struct socket *);
113 int rip_disconnect(struct socket *);
114 int rip_bind(struct socket *, struct sockaddr *, struct proc *);
115 int rip_connect(struct socket *, struct sockaddr *, struct proc *);
116 int rip_shutdown(struct socket *);
117 
118 struct  inpcbhead ripcb;
119 struct  inpcbinfo ripcbinfo;
120 
121 /* control hooks for dummynet */
122 #if DUMMYNET
123 ip_dn_ctl_t *ip_dn_ctl_ptr;
124 #endif /* DUMMYNET */
125 
126 /*
127  * Nominal space allocated to a raw ip socket.
128  */
129 #define RIPSNDQ         8192
130 #define RIPRCVQ         8192
131 
132 static KALLOC_TYPE_DEFINE(ripzone, struct inpcb, NET_KT_DEFAULT);
133 
134 /*
135  * Raw interface to IP protocol.
136  */
137 
138 /*
139  * Initialize raw connection block q.
140  */
141 void
rip_init(struct protosw * pp,struct domain * dp)142 rip_init(struct protosw *pp, struct domain *dp)
143 {
144 #pragma unused(dp)
145 	static int rip_initialized = 0;
146 	struct inpcbinfo *pcbinfo;
147 
148 	VERIFY((pp->pr_flags & (PR_INITIALIZED | PR_ATTACHED)) == PR_ATTACHED);
149 
150 	if (rip_initialized) {
151 		return;
152 	}
153 	rip_initialized = 1;
154 
155 	LIST_INIT(&ripcb);
156 	ripcbinfo.ipi_listhead = &ripcb;
157 	/*
158 	 * XXX We don't use the hash list for raw IP, but it's easier
159 	 * to allocate a one entry hash list than it is to check all
160 	 * over the place for ipi_hashbase == NULL.
161 	 */
162 	ripcbinfo.ipi_hashbase = hashinit(1, M_PCB, &ripcbinfo.ipi_hashmask);
163 	ripcbinfo.ipi_porthashbase = hashinit(1, M_PCB, &ripcbinfo.ipi_porthashmask);
164 
165 	ripcbinfo.ipi_zone.zov_kt_heap = ripzone;
166 
167 	pcbinfo = &ripcbinfo;
168 	/*
169 	 * allocate lock group attribute and group for udp pcb mutexes
170 	 */
171 	pcbinfo->ipi_lock_grp = lck_grp_alloc_init("ripcb", LCK_GRP_ATTR_NULL);
172 
173 	/*
174 	 * allocate the lock attribute for udp pcb mutexes
175 	 */
176 	lck_attr_setdefault(&pcbinfo->ipi_lock_attr);
177 	lck_rw_init(&pcbinfo->ipi_lock, pcbinfo->ipi_lock_grp,
178 	    &pcbinfo->ipi_lock_attr);
179 
180 	in_pcbinfo_attach(&ripcbinfo);
181 }
182 
183 static struct   sockaddr_in ripsrc = {
184 	.sin_len = sizeof(ripsrc),
185 	.sin_family = AF_INET,
186 	.sin_port = 0,
187 	.sin_addr = { .s_addr = 0 },
188 	.sin_zero = {0, 0, 0, 0, 0, 0, 0, 0, }
189 };
190 
191 /*
192  * Setup generic address and protocol structures
193  * for raw_input routine, then pass them along with
194  * mbuf chain.
195  */
196 void
rip_input(struct mbuf * m,int iphlen)197 rip_input(struct mbuf *m, int iphlen)
198 {
199 	struct ip *ip = mtod(m, struct ip *);
200 	struct inpcb *inp;
201 	struct inpcb *last = 0;
202 	struct mbuf *opts = 0;
203 	int skipit = 0, ret = 0;
204 	struct ifnet *ifp = m->m_pkthdr.rcvif;
205 	boolean_t is_wake_pkt = false;
206 
207 	/* Expect 32-bit aligned data pointer on strict-align platforms */
208 	MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
209 
210 	if ((m->m_flags & M_PKTHDR) && (m->m_pkthdr.pkt_flags & PKTF_WAKE_PKT)) {
211 		is_wake_pkt = true;
212 	}
213 
214 	ripsrc.sin_addr = ip->ip_src;
215 	lck_rw_lock_shared(&ripcbinfo.ipi_lock);
216 	LIST_FOREACH(inp, &ripcb, inp_list) {
217 		if ((inp->inp_vflag & INP_IPV4) == 0) {
218 			continue;
219 		}
220 		if (inp->inp_ip_p && (inp->inp_ip_p != ip->ip_p)) {
221 			continue;
222 		}
223 		if (inp->inp_laddr.s_addr &&
224 		    inp->inp_laddr.s_addr != ip->ip_dst.s_addr) {
225 			continue;
226 		}
227 		if (inp->inp_faddr.s_addr &&
228 		    inp->inp_faddr.s_addr != ip->ip_src.s_addr) {
229 			continue;
230 		}
231 		if (inp_restricted_recv(inp, ifp)) {
232 			continue;
233 		}
234 		if (last) {
235 			struct mbuf *n = m_copy(m, 0, (int)M_COPYALL);
236 
237 			skipit = 0;
238 
239 #if NECP
240 			if (n && !necp_socket_is_allowed_to_send_recv_v4(last, 0, 0,
241 			    &ip->ip_dst, &ip->ip_src, ifp, 0, NULL, NULL, NULL, NULL)) {
242 				m_freem(n);
243 				/* do not inject data to pcb */
244 				skipit = 1;
245 			}
246 #endif /* NECP */
247 			if (n && skipit == 0) {
248 				int error = 0;
249 				if ((last->inp_flags & INP_CONTROLOPTS) != 0 ||
250 				    SOFLOW_ENABLED(last->inp_socket) ||
251 				    SO_RECV_CONTROL_OPTS(last->inp_socket)) {
252 					ret = ip_savecontrol(last, &opts, ip, n);
253 					if (ret != 0) {
254 						m_freem(n);
255 						m_freem(opts);
256 						last = inp;
257 						continue;
258 					}
259 				}
260 				if (last->inp_flags & INP_STRIPHDR
261 #if CONTENT_FILTER
262 				    /*
263 				     * If socket is subject to Content Filter, delay stripping until reinject
264 				     */
265 				    && (!CFIL_DGRAM_FILTERED(last->inp_socket))
266 #endif
267 				    ) {
268 					n->m_len -= iphlen;
269 					n->m_pkthdr.len -= iphlen;
270 					n->m_data += iphlen;
271 				}
272 				so_recv_data_stat(last->inp_socket, m, 0);
273 				if (sbappendaddr(&last->inp_socket->so_rcv,
274 				    (struct sockaddr *)&ripsrc, n,
275 				    opts, &error) != 0) {
276 					sorwakeup(last->inp_socket);
277 				} else {
278 					if (error) {
279 						/* should notify about lost packet */
280 						ipstat.ips_raw_sappend_fail++;
281 					}
282 				}
283 				if (is_wake_pkt) {
284 					soevent(last->in6p_socket,
285 					    SO_FILT_HINT_LOCKED | SO_FILT_HINT_WAKE_PKT);
286 				}
287 				opts = 0;
288 			}
289 		}
290 		last = inp;
291 	}
292 
293 	skipit = 0;
294 #if NECP
295 	if (last && !necp_socket_is_allowed_to_send_recv_v4(last, 0, 0,
296 	    &ip->ip_dst, &ip->ip_src, ifp, 0, NULL, NULL, NULL, NULL)) {
297 		m_freem(m);
298 		OSAddAtomic(1, &ipstat.ips_delivered);
299 		/* do not inject data to pcb */
300 		skipit = 1;
301 	}
302 #endif /* NECP */
303 	if (skipit == 0) {
304 		if (last) {
305 			if ((last->inp_flags & INP_CONTROLOPTS) != 0 ||
306 			    SOFLOW_ENABLED(last->inp_socket) ||
307 			    SO_RECV_CONTROL_OPTS(last->inp_socket)) {
308 				ret = ip_savecontrol(last, &opts, ip, m);
309 				if (ret != 0) {
310 					m_freem(m);
311 					m_freem(opts);
312 					goto unlock;
313 				}
314 			}
315 			if (last->inp_flags & INP_STRIPHDR
316 #if CONTENT_FILTER
317 			    /*
318 			     * If socket is subject to Content Filter, delay stripping until reinject
319 			     */
320 			    && (!CFIL_DGRAM_FILTERED(last->inp_socket))
321 #endif
322 			    ) {
323 				m->m_len -= iphlen;
324 				m->m_pkthdr.len -= iphlen;
325 				m->m_data += iphlen;
326 			}
327 			so_recv_data_stat(last->inp_socket, m, 0);
328 			if (sbappendaddr(&last->inp_socket->so_rcv,
329 			    (struct sockaddr *)&ripsrc, m, opts, NULL) != 0) {
330 				sorwakeup(last->inp_socket);
331 			} else {
332 				ipstat.ips_raw_sappend_fail++;
333 			}
334 			if (is_wake_pkt) {
335 				soevent(last->in6p_socket,
336 				    SO_FILT_HINT_LOCKED | SO_FILT_HINT_WAKE_PKT);
337 			}
338 		} else {
339 			m_freem(m);
340 			OSAddAtomic(1, &ipstat.ips_noproto);
341 			OSAddAtomic(-1, &ipstat.ips_delivered);
342 		}
343 	}
344 unlock:
345 	/*
346 	 * Keep the list locked because socket filter may force the socket lock
347 	 * to be released when calling sbappendaddr() -- see rdar://7627704
348 	 */
349 	lck_rw_done(&ripcbinfo.ipi_lock);
350 }
351 
352 /*
353  * Generate IP header and pass packet to ip_output.
354  * Tack on options user may have setup with control call.
355  */
356 int
rip_output(struct mbuf * m,struct socket * so,u_int32_t dst,struct mbuf * control)357 rip_output(
358 	struct mbuf *m,
359 	struct socket *so,
360 	u_int32_t dst,
361 	struct mbuf *control)
362 {
363 	struct ip *ip;
364 	struct inpcb *inp = sotoinpcb(so);
365 	int flags = (so->so_options & SO_DONTROUTE) | IP_ALLOWBROADCAST;
366 	int inp_flags = inp ? inp->inp_flags : 0;
367 	struct ip_out_args ipoa;
368 	struct ip_moptions *imo;
369 	int tos = IPTOS_UNSPEC;
370 	int error = 0;
371 #if CONTENT_FILTER
372 	struct m_tag *cfil_tag = NULL;
373 	bool cfil_faddr_use = false;
374 	uint32_t cfil_so_state_change_cnt = 0;
375 	uint32_t cfil_so_options = 0;
376 	int cfil_inp_flags = 0;
377 	struct sockaddr *cfil_faddr = NULL;
378 	struct sockaddr_in *cfil_sin;
379 	u_int32_t cfil_dst = 0;
380 #endif
381 
382 #if CONTENT_FILTER
383 	/*
384 	 * If socket is subject to Content Filter and no addr is passed in,
385 	 * retrieve CFIL saved state from mbuf and use it if necessary.
386 	 */
387 	if (CFIL_DGRAM_FILTERED(so) && dst == INADDR_ANY) {
388 		cfil_tag = cfil_dgram_get_socket_state(m, &cfil_so_state_change_cnt, &cfil_so_options, &cfil_faddr, &cfil_inp_flags);
389 		if (cfil_tag) {
390 			cfil_sin = SIN(cfil_faddr);
391 			flags = (cfil_so_options & SO_DONTROUTE) | IP_ALLOWBROADCAST;
392 			inp_flags = cfil_inp_flags;
393 			if (inp && inp->inp_faddr.s_addr == INADDR_ANY) {
394 				/*
395 				 * Socket is unconnected, simply use the saved faddr as 'addr' to go through
396 				 * the connect/disconnect logic.
397 				 */
398 				dst = cfil_sin->sin_addr.s_addr;
399 			} else if ((so->so_state_change_cnt != cfil_so_state_change_cnt) &&
400 			    (inp->inp_fport != cfil_sin->sin_port ||
401 			    inp->inp_faddr.s_addr != cfil_sin->sin_addr.s_addr)) {
402 				/*
403 				 * Socket is connected but socket state and dest addr/port changed.
404 				 * We need to use the saved faddr and socket options.
405 				 */
406 				cfil_faddr_use = true;
407 				cfil_dst = cfil_sin->sin_addr.s_addr;
408 			}
409 			m_tag_free(cfil_tag);
410 		}
411 	}
412 #endif
413 
414 	if (so->so_state & SS_ISCONNECTED) {
415 		if (dst != INADDR_ANY) {
416 			if (m != NULL) {
417 				m_freem(m);
418 			}
419 			if (control != NULL) {
420 				m_freem(control);
421 			}
422 			return EISCONN;
423 		}
424 		dst = cfil_faddr_use ? cfil_dst : inp->inp_faddr.s_addr;
425 	} else {
426 		if (dst == INADDR_ANY) {
427 			if (m != NULL) {
428 				m_freem(m);
429 			}
430 			if (control != NULL) {
431 				m_freem(control);
432 			}
433 			return ENOTCONN;
434 		}
435 	}
436 
437 	bzero(&ipoa, sizeof(ipoa));
438 	ipoa.ipoa_boundif = IFSCOPE_NONE;
439 	ipoa.ipoa_flags = IPOAF_SELECT_SRCIF;
440 
441 	int sotc = SO_TC_UNSPEC;
442 	int netsvctype = _NET_SERVICE_TYPE_UNSPEC;
443 
444 
445 	if (control != NULL) {
446 		tos = so_tos_from_control(control);
447 		sotc = so_tc_from_control(control, &netsvctype);
448 
449 		m_freem(control);
450 		control = NULL;
451 	}
452 	if (sotc == SO_TC_UNSPEC) {
453 		sotc = so->so_traffic_class;
454 		netsvctype = so->so_netsvctype;
455 	}
456 
457 	if (inp == NULL
458 #if NECP
459 	    || (necp_socket_should_use_flow_divert(inp))
460 #endif /* NECP */
461 	    ) {
462 		if (m != NULL) {
463 			m_freem(m);
464 		}
465 		VERIFY(control == NULL);
466 		return inp == NULL ? EINVAL : EPROTOTYPE;
467 	}
468 
469 	flags |= IP_OUTARGS;
470 	/* If socket was bound to an ifindex, tell ip_output about it */
471 	if (inp->inp_flags & INP_BOUND_IF) {
472 		ipoa.ipoa_boundif = inp->inp_boundifp->if_index;
473 		ipoa.ipoa_flags |= IPOAF_BOUND_IF;
474 	}
475 	if (INP_NO_CELLULAR(inp)) {
476 		ipoa.ipoa_flags |=  IPOAF_NO_CELLULAR;
477 	}
478 	if (INP_NO_EXPENSIVE(inp)) {
479 		ipoa.ipoa_flags |=  IPOAF_NO_EXPENSIVE;
480 	}
481 	if (INP_NO_CONSTRAINED(inp)) {
482 		ipoa.ipoa_flags |=  IPOAF_NO_CONSTRAINED;
483 	}
484 	if (INP_AWDL_UNRESTRICTED(inp)) {
485 		ipoa.ipoa_flags |=  IPOAF_AWDL_UNRESTRICTED;
486 	}
487 	ipoa.ipoa_sotc = sotc;
488 	ipoa.ipoa_netsvctype = netsvctype;
489 
490 	if (inp->inp_flowhash == 0) {
491 		inp_calc_flowhash(inp);
492 		ASSERT(inp->inp_flowhash != 0);
493 	}
494 
495 	/*
496 	 * If the user handed us a complete IP packet, use it.
497 	 * Otherwise, allocate an mbuf for a header and fill it in.
498 	 */
499 	if ((inp_flags & INP_HDRINCL) == 0) {
500 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
501 			m_freem(m);
502 			return EMSGSIZE;
503 		}
504 		M_PREPEND(m, sizeof(struct ip), M_WAIT, 1);
505 		if (m == NULL) {
506 			return ENOBUFS;
507 		}
508 		ip = mtod(m, struct ip *);
509 		if (tos != IPTOS_UNSPEC) {
510 			ip->ip_tos = (uint8_t)(tos & IPTOS_MASK);
511 		} else {
512 			ip->ip_tos = inp->inp_ip_tos;
513 		}
514 		if (inp->inp_flags2 & INP2_DONTFRAG) {
515 			ip->ip_off = IP_DF;
516 		} else {
517 			ip->ip_off = 0;
518 		}
519 		ip->ip_p = inp->inp_ip_p;
520 		ip->ip_len = (uint16_t)m->m_pkthdr.len;
521 		ip->ip_src = inp->inp_laddr;
522 		ip->ip_dst.s_addr = dst;
523 		ip->ip_ttl = inp->inp_ip_ttl;
524 	} else {
525 		if (m->m_pkthdr.len > IP_MAXPACKET) {
526 			m_freem(m);
527 			return EMSGSIZE;
528 		}
529 		ip = mtod(m, struct ip *);
530 		/*
531 		 * don't allow both user specified and setsockopt options,
532 		 * and don't allow packet length sizes that will crash
533 		 */
534 		if (m->m_pkthdr.len < sizeof(struct ip) ||
535 		    ((IP_VHL_HL(ip->ip_vhl) != (sizeof(*ip) >> 2)) && inp->inp_options) ||
536 		    (ip->ip_len > m->m_pkthdr.len) ||
537 		    (ip->ip_len < (IP_VHL_HL(ip->ip_vhl) << 2))) {
538 			m_freem(m);
539 			return EINVAL;
540 		}
541 		if (ip->ip_id == 0 && !(rfc6864 && IP_OFF_IS_ATOMIC(ntohs(ip->ip_off)))) {
542 			ip->ip_id = ip_randomid((uint64_t)m);
543 		}
544 		/* XXX prevent ip_output from overwriting header fields */
545 		flags |= IP_RAWOUTPUT;
546 		OSAddAtomic(1, &ipstat.ips_rawout);
547 	}
548 
549 	if (inp->inp_laddr.s_addr != INADDR_ANY) {
550 		ipoa.ipoa_flags |= IPOAF_BOUND_SRCADDR;
551 	}
552 
553 #if NECP
554 	{
555 		necp_kernel_policy_id policy_id;
556 		necp_kernel_policy_id skip_policy_id;
557 		u_int32_t route_rule_id;
558 		u_int32_t pass_flags;
559 
560 		/*
561 		 * We need a route to perform NECP route rule checks
562 		 */
563 		if ((net_qos_policy_restricted != 0 &&
564 		    ROUTE_UNUSABLE(&inp->inp_route))
565 #if CONTENT_FILTER
566 		    || cfil_faddr_use
567 #endif
568 		    ) {
569 			struct sockaddr_in to;
570 			struct sockaddr_in from;
571 			struct in_addr laddr = ip->ip_src;
572 
573 			ROUTE_RELEASE(&inp->inp_route);
574 
575 			bzero(&from, sizeof(struct sockaddr_in));
576 			from.sin_family = AF_INET;
577 			from.sin_len = sizeof(struct sockaddr_in);
578 			from.sin_addr = laddr;
579 
580 			bzero(&to, sizeof(struct sockaddr_in));
581 			to.sin_family = AF_INET;
582 			to.sin_len = sizeof(struct sockaddr_in);
583 			to.sin_addr.s_addr = ip->ip_dst.s_addr;
584 
585 			if ((error = in_pcbladdr(inp, (struct sockaddr *)&to,
586 			    &laddr, ipoa.ipoa_boundif, NULL, 1)) != 0) {
587 				printf("%s in_pcbladdr(%p) error %d\n",
588 				    __func__, inp, error);
589 				m_freem(m);
590 				return error;
591 			}
592 
593 			inp_update_necp_policy(inp, (struct sockaddr *)&from,
594 			    (struct sockaddr *)&to, ipoa.ipoa_boundif);
595 			inp->inp_policyresult.results.qos_marking_gencount = 0;
596 		}
597 
598 		if (!necp_socket_is_allowed_to_send_recv_v4(inp, 0, 0,
599 		    &ip->ip_src, &ip->ip_dst, NULL, 0, &policy_id, &route_rule_id, &skip_policy_id, &pass_flags)) {
600 			m_freem(m);
601 			return EHOSTUNREACH;
602 		}
603 
604 		necp_mark_packet_from_socket(m, inp, policy_id, route_rule_id, skip_policy_id, pass_flags);
605 
606 		if (net_qos_policy_restricted != 0) {
607 			struct ifnet *rt_ifp = NULL;
608 
609 			if (inp->inp_route.ro_rt != NULL) {
610 				rt_ifp = inp->inp_route.ro_rt->rt_ifp;
611 			}
612 
613 			necp_socket_update_qos_marking(inp, inp->inp_route.ro_rt, route_rule_id);
614 		}
615 	}
616 #endif /* NECP */
617 	if ((so->so_flags1 & SOF1_QOSMARKING_ALLOWED)) {
618 		ipoa.ipoa_flags |= IPOAF_QOSMARKING_ALLOWED;
619 	}
620 #if IPSEC
621 	if (inp->inp_sp != NULL && ipsec_setsocket(m, so) != 0) {
622 		m_freem(m);
623 		return ENOBUFS;
624 	}
625 #endif /*IPSEC*/
626 
627 	if (ROUTE_UNUSABLE(&inp->inp_route)) {
628 		ROUTE_RELEASE(&inp->inp_route);
629 	}
630 
631 	set_packet_service_class(m, so, sotc, 0);
632 	m->m_pkthdr.pkt_flowsrc = FLOWSRC_INPCB;
633 	m->m_pkthdr.pkt_flowid = inp->inp_flowhash;
634 	m->m_pkthdr.pkt_flags |= (PKTF_FLOW_ID | PKTF_FLOW_LOCALSRC |
635 	    PKTF_FLOW_RAWSOCK);
636 	m->m_pkthdr.pkt_proto = inp->inp_ip_p;
637 	m->m_pkthdr.tx_rawip_pid = so->last_pid;
638 	m->m_pkthdr.tx_rawip_e_pid = so->e_pid;
639 	if (so->so_flags & SOF_DELEGATED) {
640 		m->m_pkthdr.tx_rawip_e_pid = so->e_pid;
641 	} else {
642 		m->m_pkthdr.tx_rawip_e_pid = 0;
643 	}
644 #if (DEBUG || DEVELOPMENT)
645 	if (so->so_flags & SOF_MARK_WAKE_PKT) {
646 		so->so_flags &= ~SOF_MARK_WAKE_PKT;
647 		m->m_pkthdr.pkt_flags |= PKTF_WAKE_PKT;
648 	}
649 #endif /* (DEBUG || DEVELOPMENT) */
650 
651 	imo = inp->inp_moptions;
652 	if (imo != NULL) {
653 		IMO_ADDREF(imo);
654 	}
655 	/*
656 	 * The domain lock is held across ip_output, so it is okay
657 	 * to pass the PCB cached route pointer directly to IP and
658 	 * the modules beneath it.
659 	 */
660 	// TODO: PASS DOWN ROUTE RULE ID
661 	error = ip_output(m, inp->inp_options, &inp->inp_route, flags,
662 	    imo, &ipoa);
663 
664 	if (imo != NULL) {
665 		IMO_REMREF(imo);
666 	}
667 
668 	if (inp->inp_route.ro_rt != NULL) {
669 		struct rtentry *rt = inp->inp_route.ro_rt;
670 		struct ifnet *outif;
671 
672 		if ((rt->rt_flags & (RTF_MULTICAST | RTF_BROADCAST)) ||
673 		    inp->inp_socket == NULL ||
674 #if CONTENT_FILTER
675 		    /* Discard temporary route for cfil case */
676 		    cfil_faddr_use ||
677 #endif
678 		    !(inp->inp_socket->so_state & SS_ISCONNECTED)) {
679 			rt = NULL;      /* unusable */
680 		}
681 		/*
682 		 * Always discard the cached route for unconnected
683 		 * socket or if it is a multicast route.
684 		 */
685 		if (rt == NULL) {
686 			ROUTE_RELEASE(&inp->inp_route);
687 		}
688 
689 		/*
690 		 * If this is a connected socket and the destination
691 		 * route is unicast, update outif with that of the
692 		 * route interface used by IP.
693 		 */
694 		if (rt != NULL &&
695 		    (outif = rt->rt_ifp) != inp->inp_last_outifp) {
696 			inp->inp_last_outifp = outif;
697 		}
698 	} else {
699 		ROUTE_RELEASE(&inp->inp_route);
700 	}
701 
702 	/*
703 	 * If output interface was cellular/expensive/constrained, and this socket is
704 	 * denied access to it, generate an event.
705 	 */
706 	if (error != 0 && (ipoa.ipoa_flags & IPOAF_R_IFDENIED) &&
707 	    (INP_NO_CELLULAR(inp) || INP_NO_EXPENSIVE(inp) || INP_NO_CONSTRAINED(inp))) {
708 		soevent(so, (SO_FILT_HINT_LOCKED | SO_FILT_HINT_IFDENIED));
709 	}
710 
711 	return error;
712 }
713 
714 
715 /*
716  * Raw IP socket option processing.
717  */
718 int
rip_ctloutput(struct socket * so,struct sockopt * sopt)719 rip_ctloutput(struct socket *so, struct sockopt *sopt)
720 {
721 	struct  inpcb *inp = sotoinpcb(so);
722 	int     error, optval;
723 
724 	/* Allow <SOL_SOCKET,SO_FLUSH> at this level */
725 	if (sopt->sopt_level != IPPROTO_IP &&
726 	    !(sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_FLUSH)) {
727 		return EINVAL;
728 	}
729 
730 	error = 0;
731 
732 	switch (sopt->sopt_dir) {
733 	case SOPT_GET:
734 		switch (sopt->sopt_name) {
735 		case IP_HDRINCL:
736 			optval = inp->inp_flags & INP_HDRINCL;
737 			error = sooptcopyout(sopt, &optval, sizeof optval);
738 			break;
739 
740 		case IP_STRIPHDR:
741 			optval = inp->inp_flags & INP_STRIPHDR;
742 			error = sooptcopyout(sopt, &optval, sizeof optval);
743 			break;
744 
745 
746 #if DUMMYNET
747 		case IP_DUMMYNET_GET:
748 			if (!DUMMYNET_LOADED) {
749 				ip_dn_init();
750 			}
751 			if (DUMMYNET_LOADED) {
752 				error = ip_dn_ctl_ptr(sopt);
753 			} else {
754 				error = ENOPROTOOPT;
755 			}
756 			break;
757 #endif /* DUMMYNET */
758 
759 		default:
760 			error = ip_ctloutput(so, sopt);
761 			break;
762 		}
763 		break;
764 
765 	case SOPT_SET:
766 		switch (sopt->sopt_name) {
767 		case IP_HDRINCL:
768 			error = sooptcopyin(sopt, &optval, sizeof optval,
769 			    sizeof optval);
770 			if (error) {
771 				break;
772 			}
773 			if (optval) {
774 				inp->inp_flags |= INP_HDRINCL;
775 			} else {
776 				inp->inp_flags &= ~INP_HDRINCL;
777 			}
778 			break;
779 
780 		case IP_STRIPHDR:
781 			error = sooptcopyin(sopt, &optval, sizeof optval,
782 			    sizeof optval);
783 			if (error) {
784 				break;
785 			}
786 			if (optval) {
787 				inp->inp_flags |= INP_STRIPHDR;
788 			} else {
789 				inp->inp_flags &= ~INP_STRIPHDR;
790 			}
791 			break;
792 
793 
794 #if DUMMYNET
795 		case IP_DUMMYNET_CONFIGURE:
796 		case IP_DUMMYNET_DEL:
797 		case IP_DUMMYNET_FLUSH:
798 			if (!DUMMYNET_LOADED) {
799 				ip_dn_init();
800 			}
801 			if (DUMMYNET_LOADED) {
802 				error = ip_dn_ctl_ptr(sopt);
803 			} else {
804 				error = ENOPROTOOPT;
805 			}
806 			break;
807 #endif /* DUMMYNET */
808 
809 		case SO_FLUSH:
810 			if ((error = sooptcopyin(sopt, &optval, sizeof(optval),
811 			    sizeof(optval))) != 0) {
812 				break;
813 			}
814 
815 			error = inp_flush(inp, optval);
816 			break;
817 
818 		default:
819 			error = ip_ctloutput(so, sopt);
820 			break;
821 		}
822 		break;
823 	}
824 
825 	return error;
826 }
827 
828 /*
829  * This function exists solely to receive the PRC_IFDOWN messages which
830  * are sent by if_down().  It looks for an ifaddr whose ifa_addr is sa,
831  * and calls in_ifadown() to remove all routes corresponding to that address.
832  * It also receives the PRC_IFUP messages from if_up() and reinstalls the
833  * interface routes.
834  */
835 void
rip_ctlinput(int cmd,struct sockaddr * sa,__unused void * vip,__unused struct ifnet * ifp)836 rip_ctlinput(
837 	int cmd,
838 	struct sockaddr *sa,
839 	__unused void *vip,
840 	__unused struct ifnet *ifp)
841 {
842 	struct in_ifaddr *ia = NULL;
843 	struct ifnet *iaifp = NULL;
844 	int err = 0;
845 	int flags, done = 0;
846 
847 	switch (cmd) {
848 	case PRC_IFDOWN:
849 		lck_rw_lock_shared(&in_ifaddr_rwlock);
850 		for (ia = in_ifaddrhead.tqh_first; ia;
851 		    ia = ia->ia_link.tqe_next) {
852 			IFA_LOCK(&ia->ia_ifa);
853 			if (ia->ia_ifa.ifa_addr == sa &&
854 			    (ia->ia_flags & IFA_ROUTE)) {
855 				done = 1;
856 				IFA_ADDREF_LOCKED(&ia->ia_ifa);
857 				IFA_UNLOCK(&ia->ia_ifa);
858 				lck_rw_done(&in_ifaddr_rwlock);
859 				lck_mtx_lock(rnh_lock);
860 				/*
861 				 * in_ifscrub kills the interface route.
862 				 */
863 				in_ifscrub(ia->ia_ifp, ia, 1);
864 				/*
865 				 * in_ifadown gets rid of all the rest of
866 				 * the routes.  This is not quite the right
867 				 * thing to do, but at least if we are running
868 				 * a routing process they will come back.
869 				 */
870 				in_ifadown(&ia->ia_ifa, 1);
871 				lck_mtx_unlock(rnh_lock);
872 				IFA_REMREF(&ia->ia_ifa);
873 				break;
874 			}
875 			IFA_UNLOCK(&ia->ia_ifa);
876 		}
877 		if (!done) {
878 			lck_rw_done(&in_ifaddr_rwlock);
879 		}
880 		break;
881 
882 	case PRC_IFUP:
883 		lck_rw_lock_shared(&in_ifaddr_rwlock);
884 		for (ia = in_ifaddrhead.tqh_first; ia;
885 		    ia = ia->ia_link.tqe_next) {
886 			IFA_LOCK(&ia->ia_ifa);
887 			if (ia->ia_ifa.ifa_addr == sa) {
888 				/* keep it locked */
889 				break;
890 			}
891 			IFA_UNLOCK(&ia->ia_ifa);
892 		}
893 		if (ia == NULL || (ia->ia_flags & IFA_ROUTE) ||
894 		    (ia->ia_ifa.ifa_debug & IFD_NOTREADY)) {
895 			if (ia != NULL) {
896 				IFA_UNLOCK(&ia->ia_ifa);
897 			}
898 			lck_rw_done(&in_ifaddr_rwlock);
899 			return;
900 		}
901 		IFA_ADDREF_LOCKED(&ia->ia_ifa);
902 		IFA_UNLOCK(&ia->ia_ifa);
903 		lck_rw_done(&in_ifaddr_rwlock);
904 
905 		flags = RTF_UP;
906 		iaifp = ia->ia_ifa.ifa_ifp;
907 
908 		if ((iaifp->if_flags & IFF_LOOPBACK)
909 		    || (iaifp->if_flags & IFF_POINTOPOINT)) {
910 			flags |= RTF_HOST;
911 		}
912 
913 		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
914 		if (err == 0) {
915 			IFA_LOCK_SPIN(&ia->ia_ifa);
916 			ia->ia_flags |= IFA_ROUTE;
917 			IFA_UNLOCK(&ia->ia_ifa);
918 		}
919 		IFA_REMREF(&ia->ia_ifa);
920 		break;
921 	}
922 }
923 
924 u_int32_t       rip_sendspace = RIPSNDQ;
925 u_int32_t       rip_recvspace = RIPRCVQ;
926 
927 SYSCTL_INT(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW | CTLFLAG_LOCKED,
928     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
929 SYSCTL_INT(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW | CTLFLAG_LOCKED,
930     &rip_recvspace, 0, "Maximum incoming raw IP datagram size");
931 SYSCTL_UINT(_net_inet_raw, OID_AUTO, pcbcount, CTLFLAG_RD | CTLFLAG_LOCKED,
932     &ripcbinfo.ipi_count, 0, "Number of active PCBs");
933 
934 static int
rip_attach(struct socket * so,int proto,struct proc * p)935 rip_attach(struct socket *so, int proto, struct proc *p)
936 {
937 	struct inpcb *inp;
938 	int error;
939 
940 	inp = sotoinpcb(so);
941 	if (inp) {
942 		panic("rip_attach");
943 	}
944 	if ((so->so_state & SS_PRIV) == 0) {
945 		return EPERM;
946 	}
947 	if (proto > UINT8_MAX) {
948 		return EINVAL;
949 	}
950 
951 	error = soreserve(so, rip_sendspace, rip_recvspace);
952 	if (error) {
953 		return error;
954 	}
955 	error = in_pcballoc(so, &ripcbinfo, p);
956 	if (error) {
957 		return error;
958 	}
959 	inp = (struct inpcb *)so->so_pcb;
960 	inp->inp_vflag |= INP_IPV4;
961 	VERIFY(proto <= UINT8_MAX);
962 	inp->inp_ip_p = (u_char)proto;
963 	inp->inp_ip_ttl = (u_char)ip_defttl;
964 	return 0;
965 }
966 
967 __private_extern__ int
rip_detach(struct socket * so)968 rip_detach(struct socket *so)
969 {
970 	struct inpcb *inp;
971 
972 	inp = sotoinpcb(so);
973 	if (inp == 0) {
974 		panic("rip_detach");
975 	}
976 	in_pcbdetach(inp);
977 	return 0;
978 }
979 
980 __private_extern__ int
rip_abort(struct socket * so)981 rip_abort(struct socket *so)
982 {
983 	soisdisconnected(so);
984 	return rip_detach(so);
985 }
986 
987 __private_extern__ int
rip_disconnect(struct socket * so)988 rip_disconnect(struct socket *so)
989 {
990 	if ((so->so_state & SS_ISCONNECTED) == 0) {
991 		return ENOTCONN;
992 	}
993 	return rip_abort(so);
994 }
995 
996 __private_extern__ int
rip_bind(struct socket * so,struct sockaddr * nam,struct proc * p)997 rip_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
998 {
999 #pragma unused(p)
1000 	struct inpcb *inp = sotoinpcb(so);
1001 	struct sockaddr_in sin;
1002 	struct ifaddr *ifa = NULL;
1003 	struct ifnet *outif = NULL;
1004 
1005 	if (inp == NULL
1006 #if NECP
1007 	    || (necp_socket_should_use_flow_divert(inp))
1008 #endif /* NECP */
1009 	    ) {
1010 		return inp == NULL ? EINVAL : EPROTOTYPE;
1011 	}
1012 
1013 	if (nam->sa_len != sizeof(struct sockaddr_in)) {
1014 		return EINVAL;
1015 	}
1016 
1017 	/* Sanitized local copy for interface address searches */
1018 	bzero(&sin, sizeof(sin));
1019 	sin.sin_family = AF_INET;
1020 	sin.sin_len = sizeof(struct sockaddr_in);
1021 	sin.sin_addr.s_addr = SIN(nam)->sin_addr.s_addr;
1022 
1023 	if (TAILQ_EMPTY(&ifnet_head) ||
1024 	    (sin.sin_family != AF_INET && sin.sin_family != AF_IMPLINK) ||
1025 	    (sin.sin_addr.s_addr && (ifa = ifa_ifwithaddr(SA(&sin))) == 0)) {
1026 		return EADDRNOTAVAIL;
1027 	} else if (ifa) {
1028 		/*
1029 		 * Opportunistically determine the outbound
1030 		 * interface that may be used; this may not
1031 		 * hold true if we end up using a route
1032 		 * going over a different interface, e.g.
1033 		 * when sending to a local address.  This
1034 		 * will get updated again after sending.
1035 		 */
1036 		IFA_LOCK(ifa);
1037 		outif = ifa->ifa_ifp;
1038 		IFA_UNLOCK(ifa);
1039 		IFA_REMREF(ifa);
1040 	}
1041 	inp->inp_laddr = sin.sin_addr;
1042 	inp->inp_last_outifp = outif;
1043 
1044 	return 0;
1045 }
1046 
1047 __private_extern__ int
rip_connect(struct socket * so,struct sockaddr * nam,__unused struct proc * p)1048 rip_connect(struct socket *so, struct sockaddr *nam, __unused  struct proc *p)
1049 {
1050 	struct inpcb *inp = sotoinpcb(so);
1051 	struct sockaddr_in *addr = (struct sockaddr_in *)(void *)nam;
1052 
1053 	if (inp == NULL
1054 #if NECP
1055 	    || (necp_socket_should_use_flow_divert(inp))
1056 #endif /* NECP */
1057 	    ) {
1058 		return inp == NULL ? EINVAL : EPROTOTYPE;
1059 	}
1060 	if (nam->sa_len != sizeof(*addr)) {
1061 		return EINVAL;
1062 	}
1063 	if (TAILQ_EMPTY(&ifnet_head)) {
1064 		return EADDRNOTAVAIL;
1065 	}
1066 	if ((addr->sin_family != AF_INET) &&
1067 	    (addr->sin_family != AF_IMPLINK)) {
1068 		return EAFNOSUPPORT;
1069 	}
1070 
1071 	if (!(so->so_flags1 & SOF1_CONNECT_COUNTED)) {
1072 		so->so_flags1 |= SOF1_CONNECT_COUNTED;
1073 		INC_ATOMIC_INT64_LIM(net_api_stats.nas_socket_inet_dgram_connected);
1074 	}
1075 
1076 	inp->inp_faddr = addr->sin_addr;
1077 	soisconnected(so);
1078 
1079 	return 0;
1080 }
1081 
1082 __private_extern__ int
rip_shutdown(struct socket * so)1083 rip_shutdown(struct socket *so)
1084 {
1085 	socantsendmore(so);
1086 	return 0;
1087 }
1088 
1089 __private_extern__ int
rip_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct proc * p)1090 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
1091     struct mbuf *control, struct proc *p)
1092 {
1093 #pragma unused(flags, p)
1094 	struct inpcb *inp = sotoinpcb(so);
1095 	u_int32_t dst = INADDR_ANY;
1096 	int error = 0;
1097 
1098 	if (inp == NULL
1099 #if NECP
1100 	    || (necp_socket_should_use_flow_divert(inp) && (error = EPROTOTYPE))
1101 #endif /* NECP */
1102 	    ) {
1103 		if (inp == NULL) {
1104 			error = EINVAL;
1105 		} else {
1106 			error = EPROTOTYPE;
1107 		}
1108 		goto bad;
1109 	}
1110 
1111 	if (nam != NULL) {
1112 		dst = ((struct sockaddr_in *)(void *)nam)->sin_addr.s_addr;
1113 	}
1114 	return rip_output(m, so, dst, control);
1115 
1116 bad:
1117 	VERIFY(error != 0);
1118 
1119 	if (m != NULL) {
1120 		m_freem(m);
1121 	}
1122 	if (control != NULL) {
1123 		m_freem(control);
1124 	}
1125 
1126 	return error;
1127 }
1128 
1129 /* note: rip_unlock is called from different protos  instead of the generic socket_unlock,
1130  * it will handle the socket dealloc on last reference
1131  * */
1132 int
rip_unlock(struct socket * so,int refcount,void * debug)1133 rip_unlock(struct socket *so, int refcount, void *debug)
1134 {
1135 	void *lr_saved;
1136 	struct inpcb *inp = sotoinpcb(so);
1137 
1138 	if (debug == NULL) {
1139 		lr_saved = __builtin_return_address(0);
1140 	} else {
1141 		lr_saved = debug;
1142 	}
1143 
1144 	if (refcount) {
1145 		if (so->so_usecount <= 0) {
1146 			panic("rip_unlock: bad refoucnt so=%p val=%x lrh= %s",
1147 			    so, so->so_usecount, solockhistory_nr(so));
1148 			/* NOTREACHED */
1149 		}
1150 		so->so_usecount--;
1151 		if (so->so_usecount == 0 && (inp->inp_wantcnt == WNT_STOPUSING)) {
1152 			/* cleanup after last reference */
1153 			lck_mtx_unlock(so->so_proto->pr_domain->dom_mtx);
1154 			lck_rw_lock_exclusive(&ripcbinfo.ipi_lock);
1155 			if (inp->inp_state != INPCB_STATE_DEAD) {
1156 				if (SOCK_CHECK_DOM(so, PF_INET6)) {
1157 					in6_pcbdetach(inp);
1158 				} else {
1159 					in_pcbdetach(inp);
1160 				}
1161 			}
1162 			in_pcbdispose(inp);
1163 			lck_rw_done(&ripcbinfo.ipi_lock);
1164 			return 0;
1165 		}
1166 	}
1167 	so->unlock_lr[so->next_unlock_lr] = lr_saved;
1168 	so->next_unlock_lr = (so->next_unlock_lr + 1) % SO_LCKDBG_MAX;
1169 	lck_mtx_unlock(so->so_proto->pr_domain->dom_mtx);
1170 	return 0;
1171 }
1172 
1173 static int
1174 rip_pcblist SYSCTL_HANDLER_ARGS
1175 {
1176 #pragma unused(oidp, arg1, arg2)
1177 	int error, i, n, sz;
1178 	struct inpcb *inp, **inp_list;
1179 	inp_gen_t gencnt;
1180 	struct xinpgen xig;
1181 
1182 	/*
1183 	 * The process of preparing the TCB list is too time-consuming and
1184 	 * resource-intensive to repeat twice on every request.
1185 	 */
1186 	lck_rw_lock_exclusive(&ripcbinfo.ipi_lock);
1187 	if (req->oldptr == USER_ADDR_NULL) {
1188 		n = ripcbinfo.ipi_count;
1189 		req->oldidx = 2 * (sizeof xig)
1190 		    + (n + n / 8) * sizeof(struct xinpcb);
1191 		lck_rw_done(&ripcbinfo.ipi_lock);
1192 		return 0;
1193 	}
1194 
1195 	if (req->newptr != USER_ADDR_NULL) {
1196 		lck_rw_done(&ripcbinfo.ipi_lock);
1197 		return EPERM;
1198 	}
1199 
1200 	/*
1201 	 * OK, now we're committed to doing something.
1202 	 */
1203 	gencnt = ripcbinfo.ipi_gencnt;
1204 	sz = n = ripcbinfo.ipi_count;
1205 
1206 	bzero(&xig, sizeof(xig));
1207 	xig.xig_len = sizeof xig;
1208 	xig.xig_count = n;
1209 	xig.xig_gen = gencnt;
1210 	xig.xig_sogen = so_gencnt;
1211 	error = SYSCTL_OUT(req, &xig, sizeof xig);
1212 	if (error) {
1213 		lck_rw_done(&ripcbinfo.ipi_lock);
1214 		return error;
1215 	}
1216 	/*
1217 	 * We are done if there is no pcb
1218 	 */
1219 	if (n == 0) {
1220 		lck_rw_done(&ripcbinfo.ipi_lock);
1221 		return 0;
1222 	}
1223 
1224 	inp_list = kalloc_type(struct inpcb *, n, Z_WAITOK);
1225 	if (inp_list == NULL) {
1226 		lck_rw_done(&ripcbinfo.ipi_lock);
1227 		return ENOMEM;
1228 	}
1229 
1230 	for (inp = ripcbinfo.ipi_listhead->lh_first, i = 0; inp && i < n;
1231 	    inp = inp->inp_list.le_next) {
1232 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1233 			inp_list[i++] = inp;
1234 		}
1235 	}
1236 	n = i;
1237 
1238 	error = 0;
1239 	for (i = 0; i < n; i++) {
1240 		inp = inp_list[i];
1241 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1242 			struct xinpcb xi;
1243 
1244 			bzero(&xi, sizeof(xi));
1245 			xi.xi_len = sizeof xi;
1246 			/* XXX should avoid extra copy */
1247 			inpcb_to_compat(inp, &xi.xi_inp);
1248 			if (inp->inp_socket) {
1249 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
1250 			}
1251 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1252 		}
1253 	}
1254 	if (!error) {
1255 		/*
1256 		 * Give the user an updated idea of our state.
1257 		 * If the generation differs from what we told
1258 		 * her before, she knows that something happened
1259 		 * while we were processing this request, and it
1260 		 * might be necessary to retry.
1261 		 */
1262 		bzero(&xig, sizeof(xig));
1263 		xig.xig_len = sizeof xig;
1264 		xig.xig_gen = ripcbinfo.ipi_gencnt;
1265 		xig.xig_sogen = so_gencnt;
1266 		xig.xig_count = ripcbinfo.ipi_count;
1267 		error = SYSCTL_OUT(req, &xig, sizeof xig);
1268 	}
1269 
1270 	lck_rw_done(&ripcbinfo.ipi_lock);
1271 	kfree_type(struct inpcb *, sz, inp_list);
1272 	return error;
1273 }
1274 
1275 SYSCTL_PROC(_net_inet_raw, OID_AUTO /*XXX*/, pcblist,
1276     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
1277     rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
1278 
1279 #if XNU_TARGET_OS_OSX
1280 
1281 static int
1282 rip_pcblist64 SYSCTL_HANDLER_ARGS
1283 {
1284 #pragma unused(oidp, arg1, arg2)
1285 	int error, i, n, sz;
1286 	struct inpcb *inp, **inp_list;
1287 	inp_gen_t gencnt;
1288 	struct xinpgen xig;
1289 
1290 	/*
1291 	 * The process of preparing the TCB list is too time-consuming and
1292 	 * resource-intensive to repeat twice on every request.
1293 	 */
1294 	lck_rw_lock_exclusive(&ripcbinfo.ipi_lock);
1295 	if (req->oldptr == USER_ADDR_NULL) {
1296 		n = ripcbinfo.ipi_count;
1297 		req->oldidx = 2 * (sizeof xig)
1298 		    + (n + n / 8) * sizeof(struct xinpcb64);
1299 		lck_rw_done(&ripcbinfo.ipi_lock);
1300 		return 0;
1301 	}
1302 
1303 	if (req->newptr != USER_ADDR_NULL) {
1304 		lck_rw_done(&ripcbinfo.ipi_lock);
1305 		return EPERM;
1306 	}
1307 
1308 	/*
1309 	 * OK, now we're committed to doing something.
1310 	 */
1311 	gencnt = ripcbinfo.ipi_gencnt;
1312 	sz = n = ripcbinfo.ipi_count;
1313 
1314 	bzero(&xig, sizeof(xig));
1315 	xig.xig_len = sizeof xig;
1316 	xig.xig_count = n;
1317 	xig.xig_gen = gencnt;
1318 	xig.xig_sogen = so_gencnt;
1319 	error = SYSCTL_OUT(req, &xig, sizeof xig);
1320 	if (error) {
1321 		lck_rw_done(&ripcbinfo.ipi_lock);
1322 		return error;
1323 	}
1324 	/*
1325 	 * We are done if there is no pcb
1326 	 */
1327 	if (n == 0) {
1328 		lck_rw_done(&ripcbinfo.ipi_lock);
1329 		return 0;
1330 	}
1331 
1332 	inp_list = kalloc_type(struct inpcb *, n, Z_WAITOK);
1333 	if (inp_list == NULL) {
1334 		lck_rw_done(&ripcbinfo.ipi_lock);
1335 		return ENOMEM;
1336 	}
1337 
1338 	for (inp = ripcbinfo.ipi_listhead->lh_first, i = 0; inp && i < n;
1339 	    inp = inp->inp_list.le_next) {
1340 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1341 			inp_list[i++] = inp;
1342 		}
1343 	}
1344 	n = i;
1345 
1346 	error = 0;
1347 	for (i = 0; i < n; i++) {
1348 		inp = inp_list[i];
1349 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1350 			struct xinpcb64 xi;
1351 
1352 			bzero(&xi, sizeof(xi));
1353 			xi.xi_len = sizeof xi;
1354 			inpcb_to_xinpcb64(inp, &xi);
1355 			if (inp->inp_socket) {
1356 				sotoxsocket64(inp->inp_socket, &xi.xi_socket);
1357 			}
1358 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1359 		}
1360 	}
1361 	if (!error) {
1362 		/*
1363 		 * Give the user an updated idea of our state.
1364 		 * If the generation differs from what we told
1365 		 * her before, she knows that something happened
1366 		 * while we were processing this request, and it
1367 		 * might be necessary to retry.
1368 		 */
1369 		bzero(&xig, sizeof(xig));
1370 		xig.xig_len = sizeof xig;
1371 		xig.xig_gen = ripcbinfo.ipi_gencnt;
1372 		xig.xig_sogen = so_gencnt;
1373 		xig.xig_count = ripcbinfo.ipi_count;
1374 		error = SYSCTL_OUT(req, &xig, sizeof xig);
1375 	}
1376 
1377 	lck_rw_done(&ripcbinfo.ipi_lock);
1378 	kfree_type(struct inpcb *, sz, inp_list);
1379 	return error;
1380 }
1381 
1382 SYSCTL_PROC(_net_inet_raw, OID_AUTO, pcblist64,
1383     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
1384     rip_pcblist64, "S,xinpcb64", "List of active raw IP sockets");
1385 
1386 #endif /* XNU_TARGET_OS_OSX */
1387 
1388 
1389 static int
1390 rip_pcblist_n SYSCTL_HANDLER_ARGS
1391 {
1392 #pragma unused(oidp, arg1, arg2)
1393 	int error = 0;
1394 
1395 	error = get_pcblist_n(IPPROTO_IP, req, &ripcbinfo);
1396 
1397 	return error;
1398 }
1399 
1400 SYSCTL_PROC(_net_inet_raw, OID_AUTO, pcblist_n,
1401     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
1402     rip_pcblist_n, "S,xinpcb_n", "List of active raw IP sockets");
1403 
1404 struct pr_usrreqs rip_usrreqs = {
1405 	.pru_abort =            rip_abort,
1406 	.pru_attach =           rip_attach,
1407 	.pru_bind =             rip_bind,
1408 	.pru_connect =          rip_connect,
1409 	.pru_control =          in_control,
1410 	.pru_detach =           rip_detach,
1411 	.pru_disconnect =       rip_disconnect,
1412 	.pru_peeraddr =         in_getpeeraddr,
1413 	.pru_send =             rip_send,
1414 	.pru_shutdown =         rip_shutdown,
1415 	.pru_sockaddr =         in_getsockaddr,
1416 	.pru_sosend =           sosend,
1417 	.pru_soreceive =        soreceive,
1418 };
1419 /* DSEP Review Done pl-20051213-v02 @3253 */
1420