xref: /xnu-8792.61.2/bsd/netinet/raw_ip.c (revision 42e220869062b56f8d7d0726fd4c88954f87902c)
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 /*
133  * Raw interface to IP protocol.
134  */
135 
136 /*
137  * Initialize raw connection block q.
138  */
139 void
rip_init(struct protosw * pp,struct domain * dp)140 rip_init(struct protosw *pp, struct domain *dp)
141 {
142 #pragma unused(dp)
143 	static int rip_initialized = 0;
144 	struct inpcbinfo *pcbinfo;
145 
146 	VERIFY((pp->pr_flags & (PR_INITIALIZED | PR_ATTACHED)) == PR_ATTACHED);
147 
148 	if (rip_initialized) {
149 		return;
150 	}
151 	rip_initialized = 1;
152 
153 	LIST_INIT(&ripcb);
154 	ripcbinfo.ipi_listhead = &ripcb;
155 	/*
156 	 * XXX We don't use the hash list for raw IP, but it's easier
157 	 * to allocate a one entry hash list than it is to check all
158 	 * over the place for ipi_hashbase == NULL.
159 	 */
160 	ripcbinfo.ipi_hashbase = hashinit(1, M_PCB, &ripcbinfo.ipi_hashmask);
161 	ripcbinfo.ipi_porthashbase = hashinit(1, M_PCB, &ripcbinfo.ipi_porthashmask);
162 
163 	ripcbinfo.ipi_zone = zone_create("ripzone", sizeof(struct inpcb),
164 	    ZC_NONE);
165 
166 	pcbinfo = &ripcbinfo;
167 	/*
168 	 * allocate lock group attribute and group for udp pcb mutexes
169 	 */
170 	pcbinfo->ipi_lock_grp = lck_grp_alloc_init("ripcb", LCK_GRP_ATTR_NULL);
171 
172 	/*
173 	 * allocate the lock attribute for udp pcb mutexes
174 	 */
175 	lck_attr_setdefault(&pcbinfo->ipi_lock_attr);
176 	lck_rw_init(&pcbinfo->ipi_lock, pcbinfo->ipi_lock_grp,
177 	    &pcbinfo->ipi_lock_attr);
178 
179 	in_pcbinfo_attach(&ripcbinfo);
180 }
181 
182 static struct   sockaddr_in ripsrc = {
183 	.sin_len = sizeof(ripsrc),
184 	.sin_family = AF_INET,
185 	.sin_port = 0,
186 	.sin_addr = { .s_addr = 0 },
187 	.sin_zero = {0, 0, 0, 0, 0, 0, 0, 0, }
188 };
189 
190 /*
191  * Setup generic address and protocol structures
192  * for raw_input routine, then pass them along with
193  * mbuf chain.
194  */
195 void
rip_input(struct mbuf * m,int iphlen)196 rip_input(struct mbuf *m, int iphlen)
197 {
198 	struct ip *ip = mtod(m, struct ip *);
199 	struct inpcb *inp;
200 	struct inpcb *last = 0;
201 	struct mbuf *opts = 0;
202 	int skipit = 0, ret = 0;
203 	struct ifnet *ifp = m->m_pkthdr.rcvif;
204 	boolean_t is_wake_pkt = false;
205 
206 	/* Expect 32-bit aligned data pointer on strict-align platforms */
207 	MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
208 
209 	if ((m->m_flags & M_PKTHDR) && (m->m_pkthdr.pkt_flags & PKTF_WAKE_PKT)) {
210 		is_wake_pkt = true;
211 	}
212 
213 	ripsrc.sin_addr = ip->ip_src;
214 	lck_rw_lock_shared(&ripcbinfo.ipi_lock);
215 	LIST_FOREACH(inp, &ripcb, inp_list) {
216 		if ((inp->inp_vflag & INP_IPV4) == 0) {
217 			continue;
218 		}
219 		if (inp->inp_ip_p && (inp->inp_ip_p != ip->ip_p)) {
220 			continue;
221 		}
222 		if (inp->inp_laddr.s_addr &&
223 		    inp->inp_laddr.s_addr != ip->ip_dst.s_addr) {
224 			continue;
225 		}
226 		if (inp->inp_faddr.s_addr &&
227 		    inp->inp_faddr.s_addr != ip->ip_src.s_addr) {
228 			continue;
229 		}
230 		if (inp_restricted_recv(inp, ifp)) {
231 			continue;
232 		}
233 		if (last) {
234 			struct mbuf *n = m_copy(m, 0, (int)M_COPYALL);
235 
236 			skipit = 0;
237 
238 #if NECP
239 			if (n && !necp_socket_is_allowed_to_send_recv_v4(last, 0, 0,
240 			    &ip->ip_dst, &ip->ip_src, ifp, 0, NULL, NULL, NULL, NULL)) {
241 				m_freem(n);
242 				/* do not inject data to pcb */
243 				skipit = 1;
244 			}
245 #endif /* NECP */
246 			if (n && skipit == 0) {
247 				int error = 0;
248 				if ((last->inp_flags & INP_CONTROLOPTS) != 0 ||
249 				    SOFLOW_ENABLED(last->inp_socket) ||
250 				    SO_RECV_CONTROL_OPTS(last->inp_socket)) {
251 					ret = ip_savecontrol(last, &opts, ip, n);
252 					if (ret != 0) {
253 						m_freem(n);
254 						m_freem(opts);
255 						last = inp;
256 						continue;
257 					}
258 				}
259 				if (last->inp_flags & INP_STRIPHDR
260 #if CONTENT_FILTER
261 				    /*
262 				     * If socket is subject to Content Filter, delay stripping until reinject
263 				     */
264 				    && (!CFIL_DGRAM_FILTERED(last->inp_socket))
265 #endif
266 				    ) {
267 					n->m_len -= iphlen;
268 					n->m_pkthdr.len -= iphlen;
269 					n->m_data += iphlen;
270 				}
271 				so_recv_data_stat(last->inp_socket, m, 0);
272 				if (sbappendaddr(&last->inp_socket->so_rcv,
273 				    (struct sockaddr *)&ripsrc, n,
274 				    opts, &error) != 0) {
275 					sorwakeup(last->inp_socket);
276 				} else {
277 					if (error) {
278 						/* should notify about lost packet */
279 						ipstat.ips_raw_sappend_fail++;
280 					}
281 				}
282 				if (is_wake_pkt) {
283 					soevent(last->in6p_socket,
284 					    SO_FILT_HINT_LOCKED | SO_FILT_HINT_WAKE_PKT);
285 				}
286 				opts = 0;
287 			}
288 		}
289 		last = inp;
290 	}
291 
292 	skipit = 0;
293 #if NECP
294 	if (last && !necp_socket_is_allowed_to_send_recv_v4(last, 0, 0,
295 	    &ip->ip_dst, &ip->ip_src, ifp, 0, NULL, NULL, NULL, NULL)) {
296 		m_freem(m);
297 		OSAddAtomic(1, &ipstat.ips_delivered);
298 		/* do not inject data to pcb */
299 		skipit = 1;
300 	}
301 #endif /* NECP */
302 	if (skipit == 0) {
303 		if (last) {
304 			if ((last->inp_flags & INP_CONTROLOPTS) != 0 ||
305 			    SOFLOW_ENABLED(last->inp_socket) ||
306 			    SO_RECV_CONTROL_OPTS(last->inp_socket)) {
307 				ret = ip_savecontrol(last, &opts, ip, m);
308 				if (ret != 0) {
309 					m_freem(m);
310 					m_freem(opts);
311 					goto unlock;
312 				}
313 			}
314 			if (last->inp_flags & INP_STRIPHDR
315 #if CONTENT_FILTER
316 			    /*
317 			     * If socket is subject to Content Filter, delay stripping until reinject
318 			     */
319 			    && (!CFIL_DGRAM_FILTERED(last->inp_socket))
320 #endif
321 			    ) {
322 				m->m_len -= iphlen;
323 				m->m_pkthdr.len -= iphlen;
324 				m->m_data += iphlen;
325 			}
326 			so_recv_data_stat(last->inp_socket, m, 0);
327 			if (sbappendaddr(&last->inp_socket->so_rcv,
328 			    (struct sockaddr *)&ripsrc, m, opts, NULL) != 0) {
329 				sorwakeup(last->inp_socket);
330 			} else {
331 				ipstat.ips_raw_sappend_fail++;
332 			}
333 			if (is_wake_pkt) {
334 				soevent(last->in6p_socket,
335 				    SO_FILT_HINT_LOCKED | SO_FILT_HINT_WAKE_PKT);
336 			}
337 		} else {
338 			m_freem(m);
339 			OSAddAtomic(1, &ipstat.ips_noproto);
340 			OSAddAtomic(-1, &ipstat.ips_delivered);
341 		}
342 	}
343 unlock:
344 	/*
345 	 * Keep the list locked because socket filter may force the socket lock
346 	 * to be released when calling sbappendaddr() -- see rdar://7627704
347 	 */
348 	lck_rw_done(&ripcbinfo.ipi_lock);
349 }
350 
351 /*
352  * Generate IP header and pass packet to ip_output.
353  * Tack on options user may have setup with control call.
354  */
355 int
rip_output(struct mbuf * m,struct socket * so,u_int32_t dst,struct mbuf * control)356 rip_output(
357 	struct mbuf *m,
358 	struct socket *so,
359 	u_int32_t dst,
360 	struct mbuf *control)
361 {
362 	struct ip *ip;
363 	struct inpcb *inp = sotoinpcb(so);
364 	int flags = (so->so_options & SO_DONTROUTE) | IP_ALLOWBROADCAST;
365 	int inp_flags = inp ? inp->inp_flags : 0;
366 	struct ip_out_args ipoa;
367 	struct ip_moptions *imo;
368 	int tos = IPTOS_UNSPEC;
369 	int error = 0;
370 #if CONTENT_FILTER
371 	struct m_tag *cfil_tag = NULL;
372 	bool cfil_faddr_use = false;
373 	uint32_t cfil_so_state_change_cnt = 0;
374 	uint32_t cfil_so_options = 0;
375 	int cfil_inp_flags = 0;
376 	struct sockaddr *cfil_faddr = NULL;
377 	struct sockaddr_in *cfil_sin;
378 	u_int32_t cfil_dst = 0;
379 #endif
380 
381 #if CONTENT_FILTER
382 	/*
383 	 * If socket is subject to Content Filter and no addr is passed in,
384 	 * retrieve CFIL saved state from mbuf and use it if necessary.
385 	 */
386 	if (CFIL_DGRAM_FILTERED(so) && dst == INADDR_ANY) {
387 		cfil_tag = cfil_dgram_get_socket_state(m, &cfil_so_state_change_cnt, &cfil_so_options, &cfil_faddr, &cfil_inp_flags);
388 		if (cfil_tag) {
389 			cfil_sin = SIN(cfil_faddr);
390 			flags = (cfil_so_options & SO_DONTROUTE) | IP_ALLOWBROADCAST;
391 			inp_flags = cfil_inp_flags;
392 			if (inp && inp->inp_faddr.s_addr == INADDR_ANY) {
393 				/*
394 				 * Socket is unconnected, simply use the saved faddr as 'addr' to go through
395 				 * the connect/disconnect logic.
396 				 */
397 				dst = cfil_sin->sin_addr.s_addr;
398 			} else if ((so->so_state_change_cnt != cfil_so_state_change_cnt) &&
399 			    (inp->inp_fport != cfil_sin->sin_port ||
400 			    inp->inp_faddr.s_addr != cfil_sin->sin_addr.s_addr)) {
401 				/*
402 				 * Socket is connected but socket state and dest addr/port changed.
403 				 * We need to use the saved faddr and socket options.
404 				 */
405 				cfil_faddr_use = true;
406 				cfil_dst = cfil_sin->sin_addr.s_addr;
407 			}
408 			m_tag_free(cfil_tag);
409 		}
410 	}
411 #endif
412 
413 	if (so->so_state & SS_ISCONNECTED) {
414 		if (dst != INADDR_ANY) {
415 			if (m != NULL) {
416 				m_freem(m);
417 			}
418 			if (control != NULL) {
419 				m_freem(control);
420 			}
421 			return EISCONN;
422 		}
423 		dst = cfil_faddr_use ? cfil_dst : inp->inp_faddr.s_addr;
424 	} else {
425 		if (dst == INADDR_ANY) {
426 			if (m != NULL) {
427 				m_freem(m);
428 			}
429 			if (control != NULL) {
430 				m_freem(control);
431 			}
432 			return ENOTCONN;
433 		}
434 	}
435 
436 	bzero(&ipoa, sizeof(ipoa));
437 	ipoa.ipoa_boundif = IFSCOPE_NONE;
438 	ipoa.ipoa_flags = IPOAF_SELECT_SRCIF;
439 
440 	int sotc = SO_TC_UNSPEC;
441 	int netsvctype = _NET_SERVICE_TYPE_UNSPEC;
442 
443 
444 	if (control != NULL) {
445 		tos = so_tos_from_control(control);
446 		sotc = so_tc_from_control(control, &netsvctype);
447 
448 		m_freem(control);
449 		control = NULL;
450 	}
451 	if (sotc == SO_TC_UNSPEC) {
452 		sotc = so->so_traffic_class;
453 		netsvctype = so->so_netsvctype;
454 	}
455 
456 	if (inp == NULL
457 #if NECP
458 	    || (necp_socket_should_use_flow_divert(inp))
459 #endif /* NECP */
460 	    ) {
461 		if (m != NULL) {
462 			m_freem(m);
463 		}
464 		VERIFY(control == NULL);
465 		return inp == NULL ? EINVAL : EPROTOTYPE;
466 	}
467 
468 	flags |= IP_OUTARGS;
469 	/* If socket was bound to an ifindex, tell ip_output about it */
470 	if (inp->inp_flags & INP_BOUND_IF) {
471 		ipoa.ipoa_boundif = inp->inp_boundifp->if_index;
472 		ipoa.ipoa_flags |= IPOAF_BOUND_IF;
473 	}
474 	if (INP_NO_CELLULAR(inp)) {
475 		ipoa.ipoa_flags |=  IPOAF_NO_CELLULAR;
476 	}
477 	if (INP_NO_EXPENSIVE(inp)) {
478 		ipoa.ipoa_flags |=  IPOAF_NO_EXPENSIVE;
479 	}
480 	if (INP_NO_CONSTRAINED(inp)) {
481 		ipoa.ipoa_flags |=  IPOAF_NO_CONSTRAINED;
482 	}
483 	if (INP_AWDL_UNRESTRICTED(inp)) {
484 		ipoa.ipoa_flags |=  IPOAF_AWDL_UNRESTRICTED;
485 	}
486 	ipoa.ipoa_sotc = sotc;
487 	ipoa.ipoa_netsvctype = netsvctype;
488 
489 	if (inp->inp_flowhash == 0) {
490 		inp_calc_flowhash(inp);
491 		ASSERT(inp->inp_flowhash != 0);
492 	}
493 
494 	/*
495 	 * If the user handed us a complete IP packet, use it.
496 	 * Otherwise, allocate an mbuf for a header and fill it in.
497 	 */
498 	if ((inp_flags & INP_HDRINCL) == 0) {
499 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
500 			m_freem(m);
501 			return EMSGSIZE;
502 		}
503 		M_PREPEND(m, sizeof(struct ip), M_WAIT, 1);
504 		if (m == NULL) {
505 			return ENOBUFS;
506 		}
507 		ip = mtod(m, struct ip *);
508 		if (tos != IPTOS_UNSPEC) {
509 			ip->ip_tos = (uint8_t)(tos & IPTOS_MASK);
510 		} else {
511 			ip->ip_tos = inp->inp_ip_tos;
512 		}
513 		if (inp->inp_flags2 & INP2_DONTFRAG) {
514 			ip->ip_off = IP_DF;
515 		} else {
516 			ip->ip_off = 0;
517 		}
518 		ip->ip_p = inp->inp_ip_p;
519 		ip->ip_len = (uint16_t)m->m_pkthdr.len;
520 		ip->ip_src = inp->inp_laddr;
521 		ip->ip_dst.s_addr = dst;
522 		ip->ip_ttl = inp->inp_ip_ttl;
523 	} else {
524 		if (m->m_pkthdr.len > IP_MAXPACKET) {
525 			m_freem(m);
526 			return EMSGSIZE;
527 		}
528 		ip = mtod(m, struct ip *);
529 		/*
530 		 * don't allow both user specified and setsockopt options,
531 		 * and don't allow packet length sizes that will crash
532 		 */
533 		if (m->m_pkthdr.len < sizeof(struct ip) ||
534 		    ((IP_VHL_HL(ip->ip_vhl) != (sizeof(*ip) >> 2)) && inp->inp_options) ||
535 		    (ip->ip_len > m->m_pkthdr.len) ||
536 		    (ip->ip_len < (IP_VHL_HL(ip->ip_vhl) << 2))) {
537 			m_freem(m);
538 			return EINVAL;
539 		}
540 		if (ip->ip_id == 0 && !(rfc6864 && IP_OFF_IS_ATOMIC(ntohs(ip->ip_off)))) {
541 			ip->ip_id = ip_randomid((uint64_t)m);
542 		}
543 		/* XXX prevent ip_output from overwriting header fields */
544 		flags |= IP_RAWOUTPUT;
545 		OSAddAtomic(1, &ipstat.ips_rawout);
546 	}
547 
548 	if (inp->inp_laddr.s_addr != INADDR_ANY) {
549 		ipoa.ipoa_flags |= IPOAF_BOUND_SRCADDR;
550 	}
551 
552 #if NECP
553 	{
554 		necp_kernel_policy_id policy_id;
555 		necp_kernel_policy_id skip_policy_id;
556 		u_int32_t route_rule_id;
557 		u_int32_t pass_flags;
558 
559 		/*
560 		 * We need a route to perform NECP route rule checks
561 		 */
562 		if ((net_qos_policy_restricted != 0 &&
563 		    ROUTE_UNUSABLE(&inp->inp_route))
564 #if CONTENT_FILTER
565 		    || cfil_faddr_use
566 #endif
567 		    ) {
568 			struct sockaddr_in to;
569 			struct sockaddr_in from;
570 			struct in_addr laddr = ip->ip_src;
571 
572 			ROUTE_RELEASE(&inp->inp_route);
573 
574 			bzero(&from, sizeof(struct sockaddr_in));
575 			from.sin_family = AF_INET;
576 			from.sin_len = sizeof(struct sockaddr_in);
577 			from.sin_addr = laddr;
578 
579 			bzero(&to, sizeof(struct sockaddr_in));
580 			to.sin_family = AF_INET;
581 			to.sin_len = sizeof(struct sockaddr_in);
582 			to.sin_addr.s_addr = ip->ip_dst.s_addr;
583 
584 			if ((error = in_pcbladdr(inp, (struct sockaddr *)&to,
585 			    &laddr, ipoa.ipoa_boundif, NULL, 1)) != 0) {
586 				printf("%s in_pcbladdr(%p) error %d\n",
587 				    __func__, inp, error);
588 				m_freem(m);
589 				return error;
590 			}
591 
592 			inp_update_necp_policy(inp, (struct sockaddr *)&from,
593 			    (struct sockaddr *)&to, ipoa.ipoa_boundif);
594 			inp->inp_policyresult.results.qos_marking_gencount = 0;
595 		}
596 
597 		if (!necp_socket_is_allowed_to_send_recv_v4(inp, 0, 0,
598 		    &ip->ip_src, &ip->ip_dst, NULL, 0, &policy_id, &route_rule_id, &skip_policy_id, &pass_flags)) {
599 			m_freem(m);
600 			return EHOSTUNREACH;
601 		}
602 
603 		necp_mark_packet_from_socket(m, inp, policy_id, route_rule_id, skip_policy_id, pass_flags);
604 
605 		if (net_qos_policy_restricted != 0) {
606 			struct ifnet *rt_ifp = NULL;
607 
608 			if (inp->inp_route.ro_rt != NULL) {
609 				rt_ifp = inp->inp_route.ro_rt->rt_ifp;
610 			}
611 
612 			necp_socket_update_qos_marking(inp, inp->inp_route.ro_rt, route_rule_id);
613 		}
614 	}
615 #endif /* NECP */
616 	if ((so->so_flags1 & SOF1_QOSMARKING_ALLOWED)) {
617 		ipoa.ipoa_flags |= IPOAF_QOSMARKING_ALLOWED;
618 	}
619 #if IPSEC
620 	if (inp->inp_sp != NULL && ipsec_setsocket(m, so) != 0) {
621 		m_freem(m);
622 		return ENOBUFS;
623 	}
624 #endif /*IPSEC*/
625 
626 	if (ROUTE_UNUSABLE(&inp->inp_route)) {
627 		ROUTE_RELEASE(&inp->inp_route);
628 	}
629 
630 	set_packet_service_class(m, so, sotc, 0);
631 	m->m_pkthdr.pkt_flowsrc = FLOWSRC_INPCB;
632 	m->m_pkthdr.pkt_flowid = inp->inp_flowhash;
633 	m->m_pkthdr.pkt_flags |= (PKTF_FLOW_ID | PKTF_FLOW_LOCALSRC |
634 	    PKTF_FLOW_RAWSOCK);
635 	m->m_pkthdr.pkt_proto = inp->inp_ip_p;
636 	m->m_pkthdr.tx_rawip_pid = so->last_pid;
637 	m->m_pkthdr.tx_rawip_e_pid = so->e_pid;
638 	if (so->so_flags & SOF_DELEGATED) {
639 		m->m_pkthdr.tx_rawip_e_pid = so->e_pid;
640 	} else {
641 		m->m_pkthdr.tx_rawip_e_pid = 0;
642 	}
643 #if (DEBUG || DEVELOPMENT)
644 	if (so->so_flags & SOF_MARK_WAKE_PKT) {
645 		so->so_flags &= ~SOF_MARK_WAKE_PKT;
646 		m->m_pkthdr.pkt_flags |= PKTF_WAKE_PKT;
647 	}
648 #endif /* (DEBUG || DEVELOPMENT) */
649 
650 	imo = inp->inp_moptions;
651 	if (imo != NULL) {
652 		IMO_ADDREF(imo);
653 	}
654 	/*
655 	 * The domain lock is held across ip_output, so it is okay
656 	 * to pass the PCB cached route pointer directly to IP and
657 	 * the modules beneath it.
658 	 */
659 	// TODO: PASS DOWN ROUTE RULE ID
660 	error = ip_output(m, inp->inp_options, &inp->inp_route, flags,
661 	    imo, &ipoa);
662 
663 	if (imo != NULL) {
664 		IMO_REMREF(imo);
665 	}
666 
667 	if (inp->inp_route.ro_rt != NULL) {
668 		struct rtentry *rt = inp->inp_route.ro_rt;
669 		struct ifnet *outif;
670 
671 		if ((rt->rt_flags & (RTF_MULTICAST | RTF_BROADCAST)) ||
672 		    inp->inp_socket == NULL ||
673 #if CONTENT_FILTER
674 		    /* Discard temporary route for cfil case */
675 		    cfil_faddr_use ||
676 #endif
677 		    !(inp->inp_socket->so_state & SS_ISCONNECTED)) {
678 			rt = NULL;      /* unusable */
679 		}
680 		/*
681 		 * Always discard the cached route for unconnected
682 		 * socket or if it is a multicast route.
683 		 */
684 		if (rt == NULL) {
685 			ROUTE_RELEASE(&inp->inp_route);
686 		}
687 
688 		/*
689 		 * If this is a connected socket and the destination
690 		 * route is unicast, update outif with that of the
691 		 * route interface used by IP.
692 		 */
693 		if (rt != NULL &&
694 		    (outif = rt->rt_ifp) != inp->inp_last_outifp) {
695 			inp->inp_last_outifp = outif;
696 		}
697 	} else {
698 		ROUTE_RELEASE(&inp->inp_route);
699 	}
700 
701 	/*
702 	 * If output interface was cellular/expensive/constrained, and this socket is
703 	 * denied access to it, generate an event.
704 	 */
705 	if (error != 0 && (ipoa.ipoa_flags & IPOAF_R_IFDENIED) &&
706 	    (INP_NO_CELLULAR(inp) || INP_NO_EXPENSIVE(inp) || INP_NO_CONSTRAINED(inp))) {
707 		soevent(so, (SO_FILT_HINT_LOCKED | SO_FILT_HINT_IFDENIED));
708 	}
709 
710 	return error;
711 }
712 
713 
714 /*
715  * Raw IP socket option processing.
716  */
717 int
rip_ctloutput(struct socket * so,struct sockopt * sopt)718 rip_ctloutput(struct socket *so, struct sockopt *sopt)
719 {
720 	struct  inpcb *inp = sotoinpcb(so);
721 	int     error, optval;
722 
723 	/* Allow <SOL_SOCKET,SO_FLUSH> at this level */
724 	if (sopt->sopt_level != IPPROTO_IP &&
725 	    !(sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_FLUSH)) {
726 		return EINVAL;
727 	}
728 
729 	error = 0;
730 
731 	switch (sopt->sopt_dir) {
732 	case SOPT_GET:
733 		switch (sopt->sopt_name) {
734 		case IP_HDRINCL:
735 			optval = inp->inp_flags & INP_HDRINCL;
736 			error = sooptcopyout(sopt, &optval, sizeof optval);
737 			break;
738 
739 		case IP_STRIPHDR:
740 			optval = inp->inp_flags & INP_STRIPHDR;
741 			error = sooptcopyout(sopt, &optval, sizeof optval);
742 			break;
743 
744 
745 #if DUMMYNET
746 		case IP_DUMMYNET_GET:
747 			if (!DUMMYNET_LOADED) {
748 				ip_dn_init();
749 			}
750 			if (DUMMYNET_LOADED) {
751 				error = ip_dn_ctl_ptr(sopt);
752 			} else {
753 				error = ENOPROTOOPT;
754 			}
755 			break;
756 #endif /* DUMMYNET */
757 
758 		default:
759 			error = ip_ctloutput(so, sopt);
760 			break;
761 		}
762 		break;
763 
764 	case SOPT_SET:
765 		switch (sopt->sopt_name) {
766 		case IP_HDRINCL:
767 			error = sooptcopyin(sopt, &optval, sizeof optval,
768 			    sizeof optval);
769 			if (error) {
770 				break;
771 			}
772 			if (optval) {
773 				inp->inp_flags |= INP_HDRINCL;
774 			} else {
775 				inp->inp_flags &= ~INP_HDRINCL;
776 			}
777 			break;
778 
779 		case IP_STRIPHDR:
780 			error = sooptcopyin(sopt, &optval, sizeof optval,
781 			    sizeof optval);
782 			if (error) {
783 				break;
784 			}
785 			if (optval) {
786 				inp->inp_flags |= INP_STRIPHDR;
787 			} else {
788 				inp->inp_flags &= ~INP_STRIPHDR;
789 			}
790 			break;
791 
792 
793 #if DUMMYNET
794 		case IP_DUMMYNET_CONFIGURE:
795 		case IP_DUMMYNET_DEL:
796 		case IP_DUMMYNET_FLUSH:
797 			if (!DUMMYNET_LOADED) {
798 				ip_dn_init();
799 			}
800 			if (DUMMYNET_LOADED) {
801 				error = ip_dn_ctl_ptr(sopt);
802 			} else {
803 				error = ENOPROTOOPT;
804 			}
805 			break;
806 #endif /* DUMMYNET */
807 
808 		case SO_FLUSH:
809 			if ((error = sooptcopyin(sopt, &optval, sizeof(optval),
810 			    sizeof(optval))) != 0) {
811 				break;
812 			}
813 
814 			error = inp_flush(inp, optval);
815 			break;
816 
817 		default:
818 			error = ip_ctloutput(so, sopt);
819 			break;
820 		}
821 		break;
822 	}
823 
824 	return error;
825 }
826 
827 /*
828  * This function exists solely to receive the PRC_IFDOWN messages which
829  * are sent by if_down().  It looks for an ifaddr whose ifa_addr is sa,
830  * and calls in_ifadown() to remove all routes corresponding to that address.
831  * It also receives the PRC_IFUP messages from if_up() and reinstalls the
832  * interface routes.
833  */
834 void
rip_ctlinput(int cmd,struct sockaddr * sa,__unused void * vip,__unused struct ifnet * ifp)835 rip_ctlinput(
836 	int cmd,
837 	struct sockaddr *sa,
838 	__unused void *vip,
839 	__unused struct ifnet *ifp)
840 {
841 	struct in_ifaddr *ia = NULL;
842 	struct ifnet *iaifp = NULL;
843 	int err = 0;
844 	int flags, done = 0;
845 
846 	switch (cmd) {
847 	case PRC_IFDOWN:
848 		lck_rw_lock_shared(&in_ifaddr_rwlock);
849 		for (ia = in_ifaddrhead.tqh_first; ia;
850 		    ia = ia->ia_link.tqe_next) {
851 			IFA_LOCK(&ia->ia_ifa);
852 			if (ia->ia_ifa.ifa_addr == sa &&
853 			    (ia->ia_flags & IFA_ROUTE)) {
854 				done = 1;
855 				IFA_ADDREF_LOCKED(&ia->ia_ifa);
856 				IFA_UNLOCK(&ia->ia_ifa);
857 				lck_rw_done(&in_ifaddr_rwlock);
858 				lck_mtx_lock(rnh_lock);
859 				/*
860 				 * in_ifscrub kills the interface route.
861 				 */
862 				in_ifscrub(ia->ia_ifp, ia, 1);
863 				/*
864 				 * in_ifadown gets rid of all the rest of
865 				 * the routes.  This is not quite the right
866 				 * thing to do, but at least if we are running
867 				 * a routing process they will come back.
868 				 */
869 				in_ifadown(&ia->ia_ifa, 1);
870 				lck_mtx_unlock(rnh_lock);
871 				IFA_REMREF(&ia->ia_ifa);
872 				break;
873 			}
874 			IFA_UNLOCK(&ia->ia_ifa);
875 		}
876 		if (!done) {
877 			lck_rw_done(&in_ifaddr_rwlock);
878 		}
879 		break;
880 
881 	case PRC_IFUP:
882 		lck_rw_lock_shared(&in_ifaddr_rwlock);
883 		for (ia = in_ifaddrhead.tqh_first; ia;
884 		    ia = ia->ia_link.tqe_next) {
885 			IFA_LOCK(&ia->ia_ifa);
886 			if (ia->ia_ifa.ifa_addr == sa) {
887 				/* keep it locked */
888 				break;
889 			}
890 			IFA_UNLOCK(&ia->ia_ifa);
891 		}
892 		if (ia == NULL || (ia->ia_flags & IFA_ROUTE) ||
893 		    (ia->ia_ifa.ifa_debug & IFD_NOTREADY)) {
894 			if (ia != NULL) {
895 				IFA_UNLOCK(&ia->ia_ifa);
896 			}
897 			lck_rw_done(&in_ifaddr_rwlock);
898 			return;
899 		}
900 		IFA_ADDREF_LOCKED(&ia->ia_ifa);
901 		IFA_UNLOCK(&ia->ia_ifa);
902 		lck_rw_done(&in_ifaddr_rwlock);
903 
904 		flags = RTF_UP;
905 		iaifp = ia->ia_ifa.ifa_ifp;
906 
907 		if ((iaifp->if_flags & IFF_LOOPBACK)
908 		    || (iaifp->if_flags & IFF_POINTOPOINT)) {
909 			flags |= RTF_HOST;
910 		}
911 
912 		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
913 		if (err == 0) {
914 			IFA_LOCK_SPIN(&ia->ia_ifa);
915 			ia->ia_flags |= IFA_ROUTE;
916 			IFA_UNLOCK(&ia->ia_ifa);
917 		}
918 		IFA_REMREF(&ia->ia_ifa);
919 		break;
920 	}
921 }
922 
923 u_int32_t       rip_sendspace = RIPSNDQ;
924 u_int32_t       rip_recvspace = RIPRCVQ;
925 
926 SYSCTL_INT(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW | CTLFLAG_LOCKED,
927     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
928 SYSCTL_INT(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW | CTLFLAG_LOCKED,
929     &rip_recvspace, 0, "Maximum incoming raw IP datagram size");
930 SYSCTL_UINT(_net_inet_raw, OID_AUTO, pcbcount, CTLFLAG_RD | CTLFLAG_LOCKED,
931     &ripcbinfo.ipi_count, 0, "Number of active PCBs");
932 
933 static int
rip_attach(struct socket * so,int proto,struct proc * p)934 rip_attach(struct socket *so, int proto, struct proc *p)
935 {
936 	struct inpcb *inp;
937 	int error;
938 
939 	inp = sotoinpcb(so);
940 	if (inp) {
941 		panic("rip_attach");
942 	}
943 	if ((so->so_state & SS_PRIV) == 0) {
944 		return EPERM;
945 	}
946 	if (proto > UINT8_MAX) {
947 		return EINVAL;
948 	}
949 
950 	error = soreserve(so, rip_sendspace, rip_recvspace);
951 	if (error) {
952 		return error;
953 	}
954 	error = in_pcballoc(so, &ripcbinfo, p);
955 	if (error) {
956 		return error;
957 	}
958 	inp = (struct inpcb *)so->so_pcb;
959 	inp->inp_vflag |= INP_IPV4;
960 	VERIFY(proto <= UINT8_MAX);
961 	inp->inp_ip_p = (u_char)proto;
962 	inp->inp_ip_ttl = (u_char)ip_defttl;
963 	return 0;
964 }
965 
966 __private_extern__ int
rip_detach(struct socket * so)967 rip_detach(struct socket *so)
968 {
969 	struct inpcb *inp;
970 
971 	inp = sotoinpcb(so);
972 	if (inp == 0) {
973 		panic("rip_detach");
974 	}
975 	in_pcbdetach(inp);
976 	return 0;
977 }
978 
979 __private_extern__ int
rip_abort(struct socket * so)980 rip_abort(struct socket *so)
981 {
982 	soisdisconnected(so);
983 	return rip_detach(so);
984 }
985 
986 __private_extern__ int
rip_disconnect(struct socket * so)987 rip_disconnect(struct socket *so)
988 {
989 	if ((so->so_state & SS_ISCONNECTED) == 0) {
990 		return ENOTCONN;
991 	}
992 	return rip_abort(so);
993 }
994 
995 __private_extern__ int
rip_bind(struct socket * so,struct sockaddr * nam,struct proc * p)996 rip_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
997 {
998 #pragma unused(p)
999 	struct inpcb *inp = sotoinpcb(so);
1000 	struct sockaddr_in sin;
1001 	struct ifaddr *ifa = NULL;
1002 	struct ifnet *outif = NULL;
1003 
1004 	if (inp == NULL
1005 #if NECP
1006 	    || (necp_socket_should_use_flow_divert(inp))
1007 #endif /* NECP */
1008 	    ) {
1009 		return inp == NULL ? EINVAL : EPROTOTYPE;
1010 	}
1011 
1012 	if (nam->sa_len != sizeof(struct sockaddr_in)) {
1013 		return EINVAL;
1014 	}
1015 
1016 	/* Sanitized local copy for interface address searches */
1017 	bzero(&sin, sizeof(sin));
1018 	sin.sin_family = AF_INET;
1019 	sin.sin_len = sizeof(struct sockaddr_in);
1020 	sin.sin_addr.s_addr = SIN(nam)->sin_addr.s_addr;
1021 
1022 	if (TAILQ_EMPTY(&ifnet_head) ||
1023 	    (sin.sin_family != AF_INET && sin.sin_family != AF_IMPLINK) ||
1024 	    (sin.sin_addr.s_addr && (ifa = ifa_ifwithaddr(SA(&sin))) == 0)) {
1025 		return EADDRNOTAVAIL;
1026 	} else if (ifa) {
1027 		/*
1028 		 * Opportunistically determine the outbound
1029 		 * interface that may be used; this may not
1030 		 * hold true if we end up using a route
1031 		 * going over a different interface, e.g.
1032 		 * when sending to a local address.  This
1033 		 * will get updated again after sending.
1034 		 */
1035 		IFA_LOCK(ifa);
1036 		outif = ifa->ifa_ifp;
1037 		IFA_UNLOCK(ifa);
1038 		IFA_REMREF(ifa);
1039 	}
1040 	inp->inp_laddr = sin.sin_addr;
1041 	inp->inp_last_outifp = outif;
1042 
1043 	return 0;
1044 }
1045 
1046 __private_extern__ int
rip_connect(struct socket * so,struct sockaddr * nam,__unused struct proc * p)1047 rip_connect(struct socket *so, struct sockaddr *nam, __unused  struct proc *p)
1048 {
1049 	struct inpcb *inp = sotoinpcb(so);
1050 	struct sockaddr_in *addr = (struct sockaddr_in *)(void *)nam;
1051 
1052 	if (inp == NULL
1053 #if NECP
1054 	    || (necp_socket_should_use_flow_divert(inp))
1055 #endif /* NECP */
1056 	    ) {
1057 		return inp == NULL ? EINVAL : EPROTOTYPE;
1058 	}
1059 	if (nam->sa_len != sizeof(*addr)) {
1060 		return EINVAL;
1061 	}
1062 	if (TAILQ_EMPTY(&ifnet_head)) {
1063 		return EADDRNOTAVAIL;
1064 	}
1065 	if ((addr->sin_family != AF_INET) &&
1066 	    (addr->sin_family != AF_IMPLINK)) {
1067 		return EAFNOSUPPORT;
1068 	}
1069 
1070 	if (!(so->so_flags1 & SOF1_CONNECT_COUNTED)) {
1071 		so->so_flags1 |= SOF1_CONNECT_COUNTED;
1072 		INC_ATOMIC_INT64_LIM(net_api_stats.nas_socket_inet_dgram_connected);
1073 	}
1074 
1075 	inp->inp_faddr = addr->sin_addr;
1076 	soisconnected(so);
1077 
1078 	return 0;
1079 }
1080 
1081 __private_extern__ int
rip_shutdown(struct socket * so)1082 rip_shutdown(struct socket *so)
1083 {
1084 	socantsendmore(so);
1085 	return 0;
1086 }
1087 
1088 __private_extern__ int
rip_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct proc * p)1089 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
1090     struct mbuf *control, struct proc *p)
1091 {
1092 #pragma unused(flags, p)
1093 	struct inpcb *inp = sotoinpcb(so);
1094 	u_int32_t dst = INADDR_ANY;
1095 	int error = 0;
1096 
1097 	if (inp == NULL
1098 #if NECP
1099 	    || (necp_socket_should_use_flow_divert(inp) && (error = EPROTOTYPE))
1100 #endif /* NECP */
1101 	    ) {
1102 		if (inp == NULL) {
1103 			error = EINVAL;
1104 		} else {
1105 			error = EPROTOTYPE;
1106 		}
1107 		goto bad;
1108 	}
1109 
1110 	if (nam != NULL) {
1111 		dst = ((struct sockaddr_in *)(void *)nam)->sin_addr.s_addr;
1112 	}
1113 	return rip_output(m, so, dst, control);
1114 
1115 bad:
1116 	VERIFY(error != 0);
1117 
1118 	if (m != NULL) {
1119 		m_freem(m);
1120 	}
1121 	if (control != NULL) {
1122 		m_freem(control);
1123 	}
1124 
1125 	return error;
1126 }
1127 
1128 /* note: rip_unlock is called from different protos  instead of the generic socket_unlock,
1129  * it will handle the socket dealloc on last reference
1130  * */
1131 int
rip_unlock(struct socket * so,int refcount,void * debug)1132 rip_unlock(struct socket *so, int refcount, void *debug)
1133 {
1134 	void *lr_saved;
1135 	struct inpcb *inp = sotoinpcb(so);
1136 
1137 	if (debug == NULL) {
1138 		lr_saved = __builtin_return_address(0);
1139 	} else {
1140 		lr_saved = debug;
1141 	}
1142 
1143 	if (refcount) {
1144 		if (so->so_usecount <= 0) {
1145 			panic("rip_unlock: bad refoucnt so=%p val=%x lrh= %s",
1146 			    so, so->so_usecount, solockhistory_nr(so));
1147 			/* NOTREACHED */
1148 		}
1149 		so->so_usecount--;
1150 		if (so->so_usecount == 0 && (inp->inp_wantcnt == WNT_STOPUSING)) {
1151 			/* cleanup after last reference */
1152 			lck_mtx_unlock(so->so_proto->pr_domain->dom_mtx);
1153 			lck_rw_lock_exclusive(&ripcbinfo.ipi_lock);
1154 			if (inp->inp_state != INPCB_STATE_DEAD) {
1155 				if (SOCK_CHECK_DOM(so, PF_INET6)) {
1156 					in6_pcbdetach(inp);
1157 				} else {
1158 					in_pcbdetach(inp);
1159 				}
1160 			}
1161 			in_pcbdispose(inp);
1162 			lck_rw_done(&ripcbinfo.ipi_lock);
1163 			return 0;
1164 		}
1165 	}
1166 	so->unlock_lr[so->next_unlock_lr] = lr_saved;
1167 	so->next_unlock_lr = (so->next_unlock_lr + 1) % SO_LCKDBG_MAX;
1168 	lck_mtx_unlock(so->so_proto->pr_domain->dom_mtx);
1169 	return 0;
1170 }
1171 
1172 static int
1173 rip_pcblist SYSCTL_HANDLER_ARGS
1174 {
1175 #pragma unused(oidp, arg1, arg2)
1176 	int error, i, n, sz;
1177 	struct inpcb *inp, **inp_list;
1178 	inp_gen_t gencnt;
1179 	struct xinpgen xig;
1180 
1181 	/*
1182 	 * The process of preparing the TCB list is too time-consuming and
1183 	 * resource-intensive to repeat twice on every request.
1184 	 */
1185 	lck_rw_lock_exclusive(&ripcbinfo.ipi_lock);
1186 	if (req->oldptr == USER_ADDR_NULL) {
1187 		n = ripcbinfo.ipi_count;
1188 		req->oldidx = 2 * (sizeof xig)
1189 		    + (n + n / 8) * sizeof(struct xinpcb);
1190 		lck_rw_done(&ripcbinfo.ipi_lock);
1191 		return 0;
1192 	}
1193 
1194 	if (req->newptr != USER_ADDR_NULL) {
1195 		lck_rw_done(&ripcbinfo.ipi_lock);
1196 		return EPERM;
1197 	}
1198 
1199 	/*
1200 	 * OK, now we're committed to doing something.
1201 	 */
1202 	gencnt = ripcbinfo.ipi_gencnt;
1203 	sz = n = ripcbinfo.ipi_count;
1204 
1205 	bzero(&xig, sizeof(xig));
1206 	xig.xig_len = sizeof xig;
1207 	xig.xig_count = n;
1208 	xig.xig_gen = gencnt;
1209 	xig.xig_sogen = so_gencnt;
1210 	error = SYSCTL_OUT(req, &xig, sizeof xig);
1211 	if (error) {
1212 		lck_rw_done(&ripcbinfo.ipi_lock);
1213 		return error;
1214 	}
1215 	/*
1216 	 * We are done if there is no pcb
1217 	 */
1218 	if (n == 0) {
1219 		lck_rw_done(&ripcbinfo.ipi_lock);
1220 		return 0;
1221 	}
1222 
1223 	inp_list = kalloc_type(struct inpcb *, n, Z_WAITOK);
1224 	if (inp_list == NULL) {
1225 		lck_rw_done(&ripcbinfo.ipi_lock);
1226 		return ENOMEM;
1227 	}
1228 
1229 	for (inp = ripcbinfo.ipi_listhead->lh_first, i = 0; inp && i < n;
1230 	    inp = inp->inp_list.le_next) {
1231 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1232 			inp_list[i++] = inp;
1233 		}
1234 	}
1235 	n = i;
1236 
1237 	error = 0;
1238 	for (i = 0; i < n; i++) {
1239 		inp = inp_list[i];
1240 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1241 			struct xinpcb xi;
1242 
1243 			bzero(&xi, sizeof(xi));
1244 			xi.xi_len = sizeof xi;
1245 			/* XXX should avoid extra copy */
1246 			inpcb_to_compat(inp, &xi.xi_inp);
1247 			if (inp->inp_socket) {
1248 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
1249 			}
1250 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1251 		}
1252 	}
1253 	if (!error) {
1254 		/*
1255 		 * Give the user an updated idea of our state.
1256 		 * If the generation differs from what we told
1257 		 * her before, she knows that something happened
1258 		 * while we were processing this request, and it
1259 		 * might be necessary to retry.
1260 		 */
1261 		bzero(&xig, sizeof(xig));
1262 		xig.xig_len = sizeof xig;
1263 		xig.xig_gen = ripcbinfo.ipi_gencnt;
1264 		xig.xig_sogen = so_gencnt;
1265 		xig.xig_count = ripcbinfo.ipi_count;
1266 		error = SYSCTL_OUT(req, &xig, sizeof xig);
1267 	}
1268 
1269 	lck_rw_done(&ripcbinfo.ipi_lock);
1270 	kfree_type(struct inpcb *, sz, inp_list);
1271 	return error;
1272 }
1273 
1274 SYSCTL_PROC(_net_inet_raw, OID_AUTO /*XXX*/, pcblist,
1275     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
1276     rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
1277 
1278 #if XNU_TARGET_OS_OSX
1279 
1280 static int
1281 rip_pcblist64 SYSCTL_HANDLER_ARGS
1282 {
1283 #pragma unused(oidp, arg1, arg2)
1284 	int error, i, n, sz;
1285 	struct inpcb *inp, **inp_list;
1286 	inp_gen_t gencnt;
1287 	struct xinpgen xig;
1288 
1289 	/*
1290 	 * The process of preparing the TCB list is too time-consuming and
1291 	 * resource-intensive to repeat twice on every request.
1292 	 */
1293 	lck_rw_lock_exclusive(&ripcbinfo.ipi_lock);
1294 	if (req->oldptr == USER_ADDR_NULL) {
1295 		n = ripcbinfo.ipi_count;
1296 		req->oldidx = 2 * (sizeof xig)
1297 		    + (n + n / 8) * sizeof(struct xinpcb64);
1298 		lck_rw_done(&ripcbinfo.ipi_lock);
1299 		return 0;
1300 	}
1301 
1302 	if (req->newptr != USER_ADDR_NULL) {
1303 		lck_rw_done(&ripcbinfo.ipi_lock);
1304 		return EPERM;
1305 	}
1306 
1307 	/*
1308 	 * OK, now we're committed to doing something.
1309 	 */
1310 	gencnt = ripcbinfo.ipi_gencnt;
1311 	sz = n = ripcbinfo.ipi_count;
1312 
1313 	bzero(&xig, sizeof(xig));
1314 	xig.xig_len = sizeof xig;
1315 	xig.xig_count = n;
1316 	xig.xig_gen = gencnt;
1317 	xig.xig_sogen = so_gencnt;
1318 	error = SYSCTL_OUT(req, &xig, sizeof xig);
1319 	if (error) {
1320 		lck_rw_done(&ripcbinfo.ipi_lock);
1321 		return error;
1322 	}
1323 	/*
1324 	 * We are done if there is no pcb
1325 	 */
1326 	if (n == 0) {
1327 		lck_rw_done(&ripcbinfo.ipi_lock);
1328 		return 0;
1329 	}
1330 
1331 	inp_list = kalloc_type(struct inpcb *, n, Z_WAITOK);
1332 	if (inp_list == NULL) {
1333 		lck_rw_done(&ripcbinfo.ipi_lock);
1334 		return ENOMEM;
1335 	}
1336 
1337 	for (inp = ripcbinfo.ipi_listhead->lh_first, i = 0; inp && i < n;
1338 	    inp = inp->inp_list.le_next) {
1339 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1340 			inp_list[i++] = inp;
1341 		}
1342 	}
1343 	n = i;
1344 
1345 	error = 0;
1346 	for (i = 0; i < n; i++) {
1347 		inp = inp_list[i];
1348 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1349 			struct xinpcb64 xi;
1350 
1351 			bzero(&xi, sizeof(xi));
1352 			xi.xi_len = sizeof xi;
1353 			inpcb_to_xinpcb64(inp, &xi);
1354 			if (inp->inp_socket) {
1355 				sotoxsocket64(inp->inp_socket, &xi.xi_socket);
1356 			}
1357 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1358 		}
1359 	}
1360 	if (!error) {
1361 		/*
1362 		 * Give the user an updated idea of our state.
1363 		 * If the generation differs from what we told
1364 		 * her before, she knows that something happened
1365 		 * while we were processing this request, and it
1366 		 * might be necessary to retry.
1367 		 */
1368 		bzero(&xig, sizeof(xig));
1369 		xig.xig_len = sizeof xig;
1370 		xig.xig_gen = ripcbinfo.ipi_gencnt;
1371 		xig.xig_sogen = so_gencnt;
1372 		xig.xig_count = ripcbinfo.ipi_count;
1373 		error = SYSCTL_OUT(req, &xig, sizeof xig);
1374 	}
1375 
1376 	lck_rw_done(&ripcbinfo.ipi_lock);
1377 	kfree_type(struct inpcb *, sz, inp_list);
1378 	return error;
1379 }
1380 
1381 SYSCTL_PROC(_net_inet_raw, OID_AUTO, pcblist64,
1382     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
1383     rip_pcblist64, "S,xinpcb64", "List of active raw IP sockets");
1384 
1385 #endif /* XNU_TARGET_OS_OSX */
1386 
1387 
1388 static int
1389 rip_pcblist_n SYSCTL_HANDLER_ARGS
1390 {
1391 #pragma unused(oidp, arg1, arg2)
1392 	int error = 0;
1393 
1394 	error = get_pcblist_n(IPPROTO_IP, req, &ripcbinfo);
1395 
1396 	return error;
1397 }
1398 
1399 SYSCTL_PROC(_net_inet_raw, OID_AUTO, pcblist_n,
1400     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
1401     rip_pcblist_n, "S,xinpcb_n", "List of active raw IP sockets");
1402 
1403 struct pr_usrreqs rip_usrreqs = {
1404 	.pru_abort =            rip_abort,
1405 	.pru_attach =           rip_attach,
1406 	.pru_bind =             rip_bind,
1407 	.pru_connect =          rip_connect,
1408 	.pru_control =          in_control,
1409 	.pru_detach =           rip_detach,
1410 	.pru_disconnect =       rip_disconnect,
1411 	.pru_peeraddr =         in_getpeeraddr,
1412 	.pru_send =             rip_send,
1413 	.pru_shutdown =         rip_shutdown,
1414 	.pru_sockaddr =         in_getsockaddr,
1415 	.pru_sosend =           sosend,
1416 	.pru_soreceive =        soreceive,
1417 };
1418 /* DSEP Review Done pl-20051213-v02 @3253 */
1419