xref: /xnu-8020.101.4/bsd/netinet/raw_ip.c (revision e7776783b89a353188416a9a346c6cdb4928faad)
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 
205 	/* Expect 32-bit aligned data pointer on strict-align platforms */
206 	MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
207 
208 	ripsrc.sin_addr = ip->ip_src;
209 	lck_rw_lock_shared(&ripcbinfo.ipi_lock);
210 	LIST_FOREACH(inp, &ripcb, inp_list) {
211 		if ((inp->inp_vflag & INP_IPV4) == 0) {
212 			continue;
213 		}
214 		if (inp->inp_ip_p && (inp->inp_ip_p != ip->ip_p)) {
215 			continue;
216 		}
217 		if (inp->inp_laddr.s_addr &&
218 		    inp->inp_laddr.s_addr != ip->ip_dst.s_addr) {
219 			continue;
220 		}
221 		if (inp->inp_faddr.s_addr &&
222 		    inp->inp_faddr.s_addr != ip->ip_src.s_addr) {
223 			continue;
224 		}
225 		if (inp_restricted_recv(inp, ifp)) {
226 			continue;
227 		}
228 		if (last) {
229 			struct mbuf *n = m_copy(m, 0, (int)M_COPYALL);
230 
231 			skipit = 0;
232 
233 #if NECP
234 			if (n && !necp_socket_is_allowed_to_send_recv_v4(last, 0, 0,
235 			    &ip->ip_dst, &ip->ip_src, ifp, 0, NULL, NULL, NULL, NULL)) {
236 				m_freem(n);
237 				/* do not inject data to pcb */
238 				skipit = 1;
239 			}
240 #endif /* NECP */
241 			if (n && skipit == 0) {
242 				int error = 0;
243 				if ((last->inp_flags & INP_CONTROLOPTS) != 0 ||
244 				    SOFLOW_ENABLED(last->inp_socket) ||
245 				    (last->inp_socket->so_options & SO_TIMESTAMP) != 0 ||
246 				    (last->inp_socket->so_options & SO_TIMESTAMP_MONOTONIC) != 0 ||
247 				    (last->inp_socket->so_options & SO_TIMESTAMP_CONTINUOUS) != 0) {
248 					ret = ip_savecontrol(last, &opts, ip, n);
249 					if (ret != 0) {
250 						m_freem(n);
251 						m_freem(opts);
252 						last = inp;
253 						continue;
254 					}
255 				}
256 				if (last->inp_flags & INP_STRIPHDR
257 #if CONTENT_FILTER
258 				    /*
259 				     * If socket is subject to Content Filter, delay stripping until reinject
260 				     */
261 				    && (!CFIL_DGRAM_FILTERED(last->inp_socket))
262 #endif
263 				    ) {
264 					n->m_len -= iphlen;
265 					n->m_pkthdr.len -= iphlen;
266 					n->m_data += iphlen;
267 				}
268 				so_recv_data_stat(last->inp_socket, m, 0);
269 				if (sbappendaddr(&last->inp_socket->so_rcv,
270 				    (struct sockaddr *)&ripsrc, n,
271 				    opts, &error) != 0) {
272 					sorwakeup(last->inp_socket);
273 				} else {
274 					if (error) {
275 						/* should notify about lost packet */
276 						ipstat.ips_raw_sappend_fail++;
277 					}
278 				}
279 				opts = 0;
280 			}
281 		}
282 		last = inp;
283 	}
284 
285 	skipit = 0;
286 #if NECP
287 	if (last && !necp_socket_is_allowed_to_send_recv_v4(last, 0, 0,
288 	    &ip->ip_dst, &ip->ip_src, ifp, 0, NULL, NULL, NULL, NULL)) {
289 		m_freem(m);
290 		OSAddAtomic(1, &ipstat.ips_delivered);
291 		/* do not inject data to pcb */
292 		skipit = 1;
293 	}
294 #endif /* NECP */
295 	if (skipit == 0) {
296 		if (last) {
297 			if ((last->inp_flags & INP_CONTROLOPTS) != 0 ||
298 			    SOFLOW_ENABLED(last->inp_socket) ||
299 			    (last->inp_socket->so_options & SO_TIMESTAMP) != 0 ||
300 			    (last->inp_socket->so_options & SO_TIMESTAMP_MONOTONIC) != 0 ||
301 			    (last->inp_socket->so_options & SO_TIMESTAMP_CONTINUOUS) != 0) {
302 				ret = ip_savecontrol(last, &opts, ip, m);
303 				if (ret != 0) {
304 					m_freem(m);
305 					m_freem(opts);
306 					goto unlock;
307 				}
308 			}
309 			if (last->inp_flags & INP_STRIPHDR
310 #if CONTENT_FILTER
311 			    /*
312 			     * If socket is subject to Content Filter, delay stripping until reinject
313 			     */
314 			    && (!CFIL_DGRAM_FILTERED(last->inp_socket))
315 #endif
316 			    ) {
317 				m->m_len -= iphlen;
318 				m->m_pkthdr.len -= iphlen;
319 				m->m_data += iphlen;
320 			}
321 			so_recv_data_stat(last->inp_socket, m, 0);
322 			if (sbappendaddr(&last->inp_socket->so_rcv,
323 			    (struct sockaddr *)&ripsrc, m, opts, NULL) != 0) {
324 				sorwakeup(last->inp_socket);
325 			} else {
326 				ipstat.ips_raw_sappend_fail++;
327 			}
328 		} else {
329 			m_freem(m);
330 			OSAddAtomic(1, &ipstat.ips_noproto);
331 			OSAddAtomic(-1, &ipstat.ips_delivered);
332 		}
333 	}
334 unlock:
335 	/*
336 	 * Keep the list locked because socket filter may force the socket lock
337 	 * to be released when calling sbappendaddr() -- see rdar://7627704
338 	 */
339 	lck_rw_done(&ripcbinfo.ipi_lock);
340 }
341 
342 /*
343  * Generate IP header and pass packet to ip_output.
344  * Tack on options user may have setup with control call.
345  */
346 int
rip_output(struct mbuf * m,struct socket * so,u_int32_t dst,struct mbuf * control)347 rip_output(
348 	struct mbuf *m,
349 	struct socket *so,
350 	u_int32_t dst,
351 	struct mbuf *control)
352 {
353 	struct ip *ip;
354 	struct inpcb *inp = sotoinpcb(so);
355 	int flags = (so->so_options & SO_DONTROUTE) | IP_ALLOWBROADCAST;
356 	int inp_flags = inp ? inp->inp_flags : 0;
357 	struct ip_out_args ipoa;
358 	struct ip_moptions *imo;
359 	int tos = IPTOS_UNSPEC;
360 	int error = 0;
361 #if CONTENT_FILTER
362 	struct m_tag *cfil_tag = NULL;
363 	bool cfil_faddr_use = false;
364 	uint32_t cfil_so_state_change_cnt = 0;
365 	uint32_t cfil_so_options = 0;
366 	int cfil_inp_flags = 0;
367 	struct sockaddr *cfil_faddr = NULL;
368 	struct sockaddr_in *cfil_sin;
369 #endif
370 
371 #if CONTENT_FILTER
372 	/*
373 	 * If socket is subject to Content Filter and no addr is passed in,
374 	 * retrieve CFIL saved state from mbuf and use it if necessary.
375 	 */
376 	if (CFIL_DGRAM_FILTERED(so) && dst == INADDR_ANY) {
377 		cfil_tag = cfil_dgram_get_socket_state(m, &cfil_so_state_change_cnt, &cfil_so_options, &cfil_faddr, &cfil_inp_flags);
378 		if (cfil_tag) {
379 			cfil_sin = SIN(cfil_faddr);
380 			flags = (cfil_so_options & SO_DONTROUTE) | IP_ALLOWBROADCAST;
381 			inp_flags = cfil_inp_flags;
382 			if (inp && inp->inp_faddr.s_addr == INADDR_ANY) {
383 				/*
384 				 * Socket is unconnected, simply use the saved faddr as 'addr' to go through
385 				 * the connect/disconnect logic.
386 				 */
387 				dst = cfil_sin->sin_addr.s_addr;
388 			} else if ((so->so_state_change_cnt != cfil_so_state_change_cnt) &&
389 			    (inp->inp_fport != cfil_sin->sin_port ||
390 			    inp->inp_faddr.s_addr != cfil_sin->sin_addr.s_addr)) {
391 				/*
392 				 * Socket is connected but socket state and dest addr/port changed.
393 				 * We need to use the saved faddr and socket options.
394 				 */
395 				cfil_faddr_use = true;
396 			}
397 			m_tag_free(cfil_tag);
398 		}
399 	}
400 #endif
401 
402 	if (so->so_state & SS_ISCONNECTED) {
403 		if (dst != INADDR_ANY) {
404 			if (m != NULL) {
405 				m_freem(m);
406 			}
407 			if (control != NULL) {
408 				m_freem(control);
409 			}
410 			return EISCONN;
411 		}
412 		dst = cfil_faddr_use ? cfil_sin->sin_addr.s_addr : inp->inp_faddr.s_addr;
413 	} else {
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 ENOTCONN;
422 		}
423 	}
424 
425 	bzero(&ipoa, sizeof(ipoa));
426 	ipoa.ipoa_boundif = IFSCOPE_NONE;
427 	ipoa.ipoa_flags = IPOAF_SELECT_SRCIF;
428 
429 	int sotc = SO_TC_UNSPEC;
430 	int netsvctype = _NET_SERVICE_TYPE_UNSPEC;
431 
432 
433 	if (control != NULL) {
434 		tos = so_tos_from_control(control);
435 		sotc = so_tc_from_control(control, &netsvctype);
436 
437 		m_freem(control);
438 		control = NULL;
439 	}
440 	if (sotc == SO_TC_UNSPEC) {
441 		sotc = so->so_traffic_class;
442 		netsvctype = so->so_netsvctype;
443 	}
444 
445 	if (inp == NULL
446 #if NECP
447 	    || (necp_socket_should_use_flow_divert(inp))
448 #endif /* NECP */
449 	    ) {
450 		if (m != NULL) {
451 			m_freem(m);
452 		}
453 		VERIFY(control == NULL);
454 		return inp == NULL ? EINVAL : EPROTOTYPE;
455 	}
456 
457 	flags |= IP_OUTARGS;
458 	/* If socket was bound to an ifindex, tell ip_output about it */
459 	if (inp->inp_flags & INP_BOUND_IF) {
460 		ipoa.ipoa_boundif = inp->inp_boundifp->if_index;
461 		ipoa.ipoa_flags |= IPOAF_BOUND_IF;
462 	}
463 	if (INP_NO_CELLULAR(inp)) {
464 		ipoa.ipoa_flags |=  IPOAF_NO_CELLULAR;
465 	}
466 	if (INP_NO_EXPENSIVE(inp)) {
467 		ipoa.ipoa_flags |=  IPOAF_NO_EXPENSIVE;
468 	}
469 	if (INP_NO_CONSTRAINED(inp)) {
470 		ipoa.ipoa_flags |=  IPOAF_NO_CONSTRAINED;
471 	}
472 	if (INP_AWDL_UNRESTRICTED(inp)) {
473 		ipoa.ipoa_flags |=  IPOAF_AWDL_UNRESTRICTED;
474 	}
475 	ipoa.ipoa_sotc = sotc;
476 	ipoa.ipoa_netsvctype = netsvctype;
477 
478 	if (inp->inp_flowhash == 0) {
479 		inp->inp_flowhash = inp_calc_flowhash(inp);
480 	}
481 
482 	/*
483 	 * If the user handed us a complete IP packet, use it.
484 	 * Otherwise, allocate an mbuf for a header and fill it in.
485 	 */
486 	if ((inp_flags & INP_HDRINCL) == 0) {
487 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
488 			m_freem(m);
489 			return EMSGSIZE;
490 		}
491 		M_PREPEND(m, sizeof(struct ip), M_WAIT, 1);
492 		if (m == NULL) {
493 			return ENOBUFS;
494 		}
495 		ip = mtod(m, struct ip *);
496 		if (tos != IPTOS_UNSPEC) {
497 			ip->ip_tos = (uint8_t)(tos & IPTOS_MASK);
498 		} else {
499 			ip->ip_tos = inp->inp_ip_tos;
500 		}
501 		if (inp->inp_flags2 & INP2_DONTFRAG) {
502 			ip->ip_off = IP_DF;
503 		} else {
504 			ip->ip_off = 0;
505 		}
506 		ip->ip_p = inp->inp_ip_p;
507 		ip->ip_len = (uint16_t)m->m_pkthdr.len;
508 		ip->ip_src = inp->inp_laddr;
509 		ip->ip_dst.s_addr = dst;
510 		ip->ip_ttl = inp->inp_ip_ttl;
511 	} else {
512 		if (m->m_pkthdr.len > IP_MAXPACKET) {
513 			m_freem(m);
514 			return EMSGSIZE;
515 		}
516 		ip = mtod(m, struct ip *);
517 		/*
518 		 * don't allow both user specified and setsockopt options,
519 		 * and don't allow packet length sizes that will crash
520 		 */
521 		if (m->m_pkthdr.len < sizeof(struct ip) ||
522 		    ((IP_VHL_HL(ip->ip_vhl) != (sizeof(*ip) >> 2)) && inp->inp_options) ||
523 		    (ip->ip_len > m->m_pkthdr.len) ||
524 		    (ip->ip_len < (IP_VHL_HL(ip->ip_vhl) << 2))) {
525 			m_freem(m);
526 			return EINVAL;
527 		}
528 		if (ip->ip_id == 0 && !(rfc6864 && IP_OFF_IS_ATOMIC(ntohs(ip->ip_off)))) {
529 			ip->ip_id = ip_randomid((uint64_t)m);
530 		}
531 		/* XXX prevent ip_output from overwriting header fields */
532 		flags |= IP_RAWOUTPUT;
533 		OSAddAtomic(1, &ipstat.ips_rawout);
534 	}
535 
536 	if (inp->inp_laddr.s_addr != INADDR_ANY) {
537 		ipoa.ipoa_flags |= IPOAF_BOUND_SRCADDR;
538 	}
539 
540 #if NECP
541 	{
542 		necp_kernel_policy_id policy_id;
543 		necp_kernel_policy_id skip_policy_id;
544 		u_int32_t route_rule_id;
545 		u_int32_t pass_flags;
546 
547 		/*
548 		 * We need a route to perform NECP route rule checks
549 		 */
550 		if ((net_qos_policy_restricted != 0 &&
551 		    ROUTE_UNUSABLE(&inp->inp_route))
552 #if CONTENT_FILTER
553 		    || cfil_faddr_use
554 #endif
555 		    ) {
556 			struct sockaddr_in to;
557 			struct sockaddr_in from;
558 			struct in_addr laddr = ip->ip_src;
559 
560 			ROUTE_RELEASE(&inp->inp_route);
561 
562 			bzero(&from, sizeof(struct sockaddr_in));
563 			from.sin_family = AF_INET;
564 			from.sin_len = sizeof(struct sockaddr_in);
565 			from.sin_addr = laddr;
566 
567 			bzero(&to, sizeof(struct sockaddr_in));
568 			to.sin_family = AF_INET;
569 			to.sin_len = sizeof(struct sockaddr_in);
570 			to.sin_addr.s_addr = ip->ip_dst.s_addr;
571 
572 			if ((error = in_pcbladdr(inp, (struct sockaddr *)&to,
573 			    &laddr, ipoa.ipoa_boundif, NULL, 1)) != 0) {
574 				printf("%s in_pcbladdr(%p) error %d\n",
575 				    __func__, inp, error);
576 				m_freem(m);
577 				return error;
578 			}
579 
580 			inp_update_necp_policy(inp, (struct sockaddr *)&from,
581 			    (struct sockaddr *)&to, ipoa.ipoa_boundif);
582 			inp->inp_policyresult.results.qos_marking_gencount = 0;
583 		}
584 
585 		if (!necp_socket_is_allowed_to_send_recv_v4(inp, 0, 0,
586 		    &ip->ip_src, &ip->ip_dst, NULL, 0, &policy_id, &route_rule_id, &skip_policy_id, &pass_flags)) {
587 			m_freem(m);
588 			return EHOSTUNREACH;
589 		}
590 
591 		necp_mark_packet_from_socket(m, inp, policy_id, route_rule_id, skip_policy_id, pass_flags);
592 
593 		if (net_qos_policy_restricted != 0) {
594 			struct ifnet *rt_ifp = NULL;
595 
596 			if (inp->inp_route.ro_rt != NULL) {
597 				rt_ifp = inp->inp_route.ro_rt->rt_ifp;
598 			}
599 
600 			necp_socket_update_qos_marking(inp, inp->inp_route.ro_rt, route_rule_id);
601 		}
602 	}
603 #endif /* NECP */
604 	if ((so->so_flags1 & SOF1_QOSMARKING_ALLOWED)) {
605 		ipoa.ipoa_flags |= IPOAF_QOSMARKING_ALLOWED;
606 	}
607 #if IPSEC
608 	if (inp->inp_sp != NULL && ipsec_setsocket(m, so) != 0) {
609 		m_freem(m);
610 		return ENOBUFS;
611 	}
612 #endif /*IPSEC*/
613 
614 	if (ROUTE_UNUSABLE(&inp->inp_route)) {
615 		ROUTE_RELEASE(&inp->inp_route);
616 	}
617 
618 	set_packet_service_class(m, so, sotc, 0);
619 	m->m_pkthdr.pkt_flowsrc = FLOWSRC_INPCB;
620 	m->m_pkthdr.pkt_flowid = inp->inp_flowhash;
621 	m->m_pkthdr.pkt_flags |= (PKTF_FLOW_ID | PKTF_FLOW_LOCALSRC |
622 	    PKTF_FLOW_RAWSOCK);
623 	m->m_pkthdr.pkt_proto = inp->inp_ip_p;
624 	m->m_pkthdr.tx_rawip_pid = so->last_pid;
625 	m->m_pkthdr.tx_rawip_e_pid = so->e_pid;
626 	if (so->so_flags & SOF_DELEGATED) {
627 		m->m_pkthdr.tx_rawip_e_pid = so->e_pid;
628 	} else {
629 		m->m_pkthdr.tx_rawip_e_pid = 0;
630 	}
631 #if (DEBUG || DEVELOPMENT)
632 	if (so->so_flags & SOF_MARK_WAKE_PKT) {
633 		so->so_flags &= ~SOF_MARK_WAKE_PKT;
634 		m->m_pkthdr.pkt_flags |= PKTF_WAKE_PKT;
635 	}
636 #endif /* (DEBUG || DEVELOPMENT) */
637 
638 	imo = inp->inp_moptions;
639 	if (imo != NULL) {
640 		IMO_ADDREF(imo);
641 	}
642 	/*
643 	 * The domain lock is held across ip_output, so it is okay
644 	 * to pass the PCB cached route pointer directly to IP and
645 	 * the modules beneath it.
646 	 */
647 	// TODO: PASS DOWN ROUTE RULE ID
648 	error = ip_output(m, inp->inp_options, &inp->inp_route, flags,
649 	    imo, &ipoa);
650 
651 	if (imo != NULL) {
652 		IMO_REMREF(imo);
653 	}
654 
655 	if (inp->inp_route.ro_rt != NULL) {
656 		struct rtentry *rt = inp->inp_route.ro_rt;
657 		struct ifnet *outif;
658 
659 		if ((rt->rt_flags & (RTF_MULTICAST | RTF_BROADCAST)) ||
660 		    inp->inp_socket == NULL ||
661 #if CONTENT_FILTER
662 		    /* Discard temporary route for cfil case */
663 		    cfil_faddr_use ||
664 #endif
665 		    !(inp->inp_socket->so_state & SS_ISCONNECTED)) {
666 			rt = NULL;      /* unusable */
667 		}
668 		/*
669 		 * Always discard the cached route for unconnected
670 		 * socket or if it is a multicast route.
671 		 */
672 		if (rt == NULL) {
673 			ROUTE_RELEASE(&inp->inp_route);
674 		}
675 
676 		/*
677 		 * If this is a connected socket and the destination
678 		 * route is unicast, update outif with that of the
679 		 * route interface used by IP.
680 		 */
681 		if (rt != NULL &&
682 		    (outif = rt->rt_ifp) != inp->inp_last_outifp) {
683 			inp->inp_last_outifp = outif;
684 		}
685 	} else {
686 		ROUTE_RELEASE(&inp->inp_route);
687 	}
688 
689 	/*
690 	 * If output interface was cellular/expensive/constrained, and this socket is
691 	 * denied access to it, generate an event.
692 	 */
693 	if (error != 0 && (ipoa.ipoa_flags & IPOAF_R_IFDENIED) &&
694 	    (INP_NO_CELLULAR(inp) || INP_NO_EXPENSIVE(inp) || INP_NO_CONSTRAINED(inp))) {
695 		soevent(so, (SO_FILT_HINT_LOCKED | SO_FILT_HINT_IFDENIED));
696 	}
697 
698 	return error;
699 }
700 
701 
702 /*
703  * Raw IP socket option processing.
704  */
705 int
rip_ctloutput(struct socket * so,struct sockopt * sopt)706 rip_ctloutput(struct socket *so, struct sockopt *sopt)
707 {
708 	struct  inpcb *inp = sotoinpcb(so);
709 	int     error, optval;
710 
711 	/* Allow <SOL_SOCKET,SO_FLUSH> at this level */
712 	if (sopt->sopt_level != IPPROTO_IP &&
713 	    !(sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_FLUSH)) {
714 		return EINVAL;
715 	}
716 
717 	error = 0;
718 
719 	switch (sopt->sopt_dir) {
720 	case SOPT_GET:
721 		switch (sopt->sopt_name) {
722 		case IP_HDRINCL:
723 			optval = inp->inp_flags & INP_HDRINCL;
724 			error = sooptcopyout(sopt, &optval, sizeof optval);
725 			break;
726 
727 		case IP_STRIPHDR:
728 			optval = inp->inp_flags & INP_STRIPHDR;
729 			error = sooptcopyout(sopt, &optval, sizeof optval);
730 			break;
731 
732 
733 #if DUMMYNET
734 		case IP_DUMMYNET_GET:
735 			if (!DUMMYNET_LOADED) {
736 				ip_dn_init();
737 			}
738 			if (DUMMYNET_LOADED) {
739 				error = ip_dn_ctl_ptr(sopt);
740 			} else {
741 				error = ENOPROTOOPT;
742 			}
743 			break;
744 #endif /* DUMMYNET */
745 
746 		default:
747 			error = ip_ctloutput(so, sopt);
748 			break;
749 		}
750 		break;
751 
752 	case SOPT_SET:
753 		switch (sopt->sopt_name) {
754 		case IP_HDRINCL:
755 			error = sooptcopyin(sopt, &optval, sizeof optval,
756 			    sizeof optval);
757 			if (error) {
758 				break;
759 			}
760 			if (optval) {
761 				inp->inp_flags |= INP_HDRINCL;
762 			} else {
763 				inp->inp_flags &= ~INP_HDRINCL;
764 			}
765 			break;
766 
767 		case IP_STRIPHDR:
768 			error = sooptcopyin(sopt, &optval, sizeof optval,
769 			    sizeof optval);
770 			if (error) {
771 				break;
772 			}
773 			if (optval) {
774 				inp->inp_flags |= INP_STRIPHDR;
775 			} else {
776 				inp->inp_flags &= ~INP_STRIPHDR;
777 			}
778 			break;
779 
780 
781 #if DUMMYNET
782 		case IP_DUMMYNET_CONFIGURE:
783 		case IP_DUMMYNET_DEL:
784 		case IP_DUMMYNET_FLUSH:
785 			if (!DUMMYNET_LOADED) {
786 				ip_dn_init();
787 			}
788 			if (DUMMYNET_LOADED) {
789 				error = ip_dn_ctl_ptr(sopt);
790 			} else {
791 				error = ENOPROTOOPT;
792 			}
793 			break;
794 #endif /* DUMMYNET */
795 
796 		case SO_FLUSH:
797 			if ((error = sooptcopyin(sopt, &optval, sizeof(optval),
798 			    sizeof(optval))) != 0) {
799 				break;
800 			}
801 
802 			error = inp_flush(inp, optval);
803 			break;
804 
805 		default:
806 			error = ip_ctloutput(so, sopt);
807 			break;
808 		}
809 		break;
810 	}
811 
812 	return error;
813 }
814 
815 /*
816  * This function exists solely to receive the PRC_IFDOWN messages which
817  * are sent by if_down().  It looks for an ifaddr whose ifa_addr is sa,
818  * and calls in_ifadown() to remove all routes corresponding to that address.
819  * It also receives the PRC_IFUP messages from if_up() and reinstalls the
820  * interface routes.
821  */
822 void
rip_ctlinput(int cmd,struct sockaddr * sa,__unused void * vip,__unused struct ifnet * ifp)823 rip_ctlinput(
824 	int cmd,
825 	struct sockaddr *sa,
826 	__unused void *vip,
827 	__unused struct ifnet *ifp)
828 {
829 	struct in_ifaddr *ia = NULL;
830 	struct ifnet *iaifp = NULL;
831 	int err = 0;
832 	int flags, done = 0;
833 
834 	switch (cmd) {
835 	case PRC_IFDOWN:
836 		lck_rw_lock_shared(&in_ifaddr_rwlock);
837 		for (ia = in_ifaddrhead.tqh_first; ia;
838 		    ia = ia->ia_link.tqe_next) {
839 			IFA_LOCK(&ia->ia_ifa);
840 			if (ia->ia_ifa.ifa_addr == sa &&
841 			    (ia->ia_flags & IFA_ROUTE)) {
842 				done = 1;
843 				IFA_ADDREF_LOCKED(&ia->ia_ifa);
844 				IFA_UNLOCK(&ia->ia_ifa);
845 				lck_rw_done(&in_ifaddr_rwlock);
846 				lck_mtx_lock(rnh_lock);
847 				/*
848 				 * in_ifscrub kills the interface route.
849 				 */
850 				in_ifscrub(ia->ia_ifp, ia, 1);
851 				/*
852 				 * in_ifadown gets rid of all the rest of
853 				 * the routes.  This is not quite the right
854 				 * thing to do, but at least if we are running
855 				 * a routing process they will come back.
856 				 */
857 				in_ifadown(&ia->ia_ifa, 1);
858 				lck_mtx_unlock(rnh_lock);
859 				IFA_REMREF(&ia->ia_ifa);
860 				break;
861 			}
862 			IFA_UNLOCK(&ia->ia_ifa);
863 		}
864 		if (!done) {
865 			lck_rw_done(&in_ifaddr_rwlock);
866 		}
867 		break;
868 
869 	case PRC_IFUP:
870 		lck_rw_lock_shared(&in_ifaddr_rwlock);
871 		for (ia = in_ifaddrhead.tqh_first; ia;
872 		    ia = ia->ia_link.tqe_next) {
873 			IFA_LOCK(&ia->ia_ifa);
874 			if (ia->ia_ifa.ifa_addr == sa) {
875 				/* keep it locked */
876 				break;
877 			}
878 			IFA_UNLOCK(&ia->ia_ifa);
879 		}
880 		if (ia == NULL || (ia->ia_flags & IFA_ROUTE) ||
881 		    (ia->ia_ifa.ifa_debug & IFD_NOTREADY)) {
882 			if (ia != NULL) {
883 				IFA_UNLOCK(&ia->ia_ifa);
884 			}
885 			lck_rw_done(&in_ifaddr_rwlock);
886 			return;
887 		}
888 		IFA_ADDREF_LOCKED(&ia->ia_ifa);
889 		IFA_UNLOCK(&ia->ia_ifa);
890 		lck_rw_done(&in_ifaddr_rwlock);
891 
892 		flags = RTF_UP;
893 		iaifp = ia->ia_ifa.ifa_ifp;
894 
895 		if ((iaifp->if_flags & IFF_LOOPBACK)
896 		    || (iaifp->if_flags & IFF_POINTOPOINT)) {
897 			flags |= RTF_HOST;
898 		}
899 
900 		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
901 		if (err == 0) {
902 			IFA_LOCK_SPIN(&ia->ia_ifa);
903 			ia->ia_flags |= IFA_ROUTE;
904 			IFA_UNLOCK(&ia->ia_ifa);
905 		}
906 		IFA_REMREF(&ia->ia_ifa);
907 		break;
908 	}
909 }
910 
911 u_int32_t       rip_sendspace = RIPSNDQ;
912 u_int32_t       rip_recvspace = RIPRCVQ;
913 
914 SYSCTL_INT(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW | CTLFLAG_LOCKED,
915     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
916 SYSCTL_INT(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW | CTLFLAG_LOCKED,
917     &rip_recvspace, 0, "Maximum incoming raw IP datagram size");
918 SYSCTL_UINT(_net_inet_raw, OID_AUTO, pcbcount, CTLFLAG_RD | CTLFLAG_LOCKED,
919     &ripcbinfo.ipi_count, 0, "Number of active PCBs");
920 
921 static int
rip_attach(struct socket * so,int proto,struct proc * p)922 rip_attach(struct socket *so, int proto, struct proc *p)
923 {
924 	struct inpcb *inp;
925 	int error;
926 
927 	inp = sotoinpcb(so);
928 	if (inp) {
929 		panic("rip_attach");
930 	}
931 	if ((so->so_state & SS_PRIV) == 0) {
932 		return EPERM;
933 	}
934 	if (proto > UINT8_MAX) {
935 		return EINVAL;
936 	}
937 
938 	error = soreserve(so, rip_sendspace, rip_recvspace);
939 	if (error) {
940 		return error;
941 	}
942 	error = in_pcballoc(so, &ripcbinfo, p);
943 	if (error) {
944 		return error;
945 	}
946 	inp = (struct inpcb *)so->so_pcb;
947 	inp->inp_vflag |= INP_IPV4;
948 	VERIFY(proto <= UINT8_MAX);
949 	inp->inp_ip_p = (u_char)proto;
950 	inp->inp_ip_ttl = (u_char)ip_defttl;
951 	return 0;
952 }
953 
954 __private_extern__ int
rip_detach(struct socket * so)955 rip_detach(struct socket *so)
956 {
957 	struct inpcb *inp;
958 
959 	inp = sotoinpcb(so);
960 	if (inp == 0) {
961 		panic("rip_detach");
962 	}
963 	in_pcbdetach(inp);
964 	return 0;
965 }
966 
967 __private_extern__ int
rip_abort(struct socket * so)968 rip_abort(struct socket *so)
969 {
970 	soisdisconnected(so);
971 	return rip_detach(so);
972 }
973 
974 __private_extern__ int
rip_disconnect(struct socket * so)975 rip_disconnect(struct socket *so)
976 {
977 	if ((so->so_state & SS_ISCONNECTED) == 0) {
978 		return ENOTCONN;
979 	}
980 	return rip_abort(so);
981 }
982 
983 __private_extern__ int
rip_bind(struct socket * so,struct sockaddr * nam,struct proc * p)984 rip_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
985 {
986 #pragma unused(p)
987 	struct inpcb *inp = sotoinpcb(so);
988 	struct sockaddr_in sin;
989 	struct ifaddr *ifa = NULL;
990 	struct ifnet *outif = NULL;
991 
992 	if (inp == NULL
993 #if NECP
994 	    || (necp_socket_should_use_flow_divert(inp))
995 #endif /* NECP */
996 	    ) {
997 		return inp == NULL ? EINVAL : EPROTOTYPE;
998 	}
999 
1000 	if (nam->sa_len != sizeof(struct sockaddr_in)) {
1001 		return EINVAL;
1002 	}
1003 
1004 	/* Sanitized local copy for interface address searches */
1005 	bzero(&sin, sizeof(sin));
1006 	sin.sin_family = AF_INET;
1007 	sin.sin_len = sizeof(struct sockaddr_in);
1008 	sin.sin_addr.s_addr = SIN(nam)->sin_addr.s_addr;
1009 
1010 	if (TAILQ_EMPTY(&ifnet_head) ||
1011 	    (sin.sin_family != AF_INET && sin.sin_family != AF_IMPLINK) ||
1012 	    (sin.sin_addr.s_addr && (ifa = ifa_ifwithaddr(SA(&sin))) == 0)) {
1013 		return EADDRNOTAVAIL;
1014 	} else if (ifa) {
1015 		/*
1016 		 * Opportunistically determine the outbound
1017 		 * interface that may be used; this may not
1018 		 * hold true if we end up using a route
1019 		 * going over a different interface, e.g.
1020 		 * when sending to a local address.  This
1021 		 * will get updated again after sending.
1022 		 */
1023 		IFA_LOCK(ifa);
1024 		outif = ifa->ifa_ifp;
1025 		IFA_UNLOCK(ifa);
1026 		IFA_REMREF(ifa);
1027 	}
1028 	inp->inp_laddr = sin.sin_addr;
1029 	inp->inp_last_outifp = outif;
1030 
1031 	return 0;
1032 }
1033 
1034 __private_extern__ int
rip_connect(struct socket * so,struct sockaddr * nam,__unused struct proc * p)1035 rip_connect(struct socket *so, struct sockaddr *nam, __unused  struct proc *p)
1036 {
1037 	struct inpcb *inp = sotoinpcb(so);
1038 	struct sockaddr_in *addr = (struct sockaddr_in *)(void *)nam;
1039 
1040 	if (inp == NULL
1041 #if NECP
1042 	    || (necp_socket_should_use_flow_divert(inp))
1043 #endif /* NECP */
1044 	    ) {
1045 		return inp == NULL ? EINVAL : EPROTOTYPE;
1046 	}
1047 	if (nam->sa_len != sizeof(*addr)) {
1048 		return EINVAL;
1049 	}
1050 	if (TAILQ_EMPTY(&ifnet_head)) {
1051 		return EADDRNOTAVAIL;
1052 	}
1053 	if ((addr->sin_family != AF_INET) &&
1054 	    (addr->sin_family != AF_IMPLINK)) {
1055 		return EAFNOSUPPORT;
1056 	}
1057 
1058 	if (!(so->so_flags1 & SOF1_CONNECT_COUNTED)) {
1059 		so->so_flags1 |= SOF1_CONNECT_COUNTED;
1060 		INC_ATOMIC_INT64_LIM(net_api_stats.nas_socket_inet_dgram_connected);
1061 	}
1062 
1063 	inp->inp_faddr = addr->sin_addr;
1064 	soisconnected(so);
1065 
1066 	return 0;
1067 }
1068 
1069 __private_extern__ int
rip_shutdown(struct socket * so)1070 rip_shutdown(struct socket *so)
1071 {
1072 	socantsendmore(so);
1073 	return 0;
1074 }
1075 
1076 __private_extern__ int
rip_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct proc * p)1077 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
1078     struct mbuf *control, struct proc *p)
1079 {
1080 #pragma unused(flags, p)
1081 	struct inpcb *inp = sotoinpcb(so);
1082 	u_int32_t dst = INADDR_ANY;
1083 	int error = 0;
1084 
1085 	if (inp == NULL
1086 #if NECP
1087 	    || (necp_socket_should_use_flow_divert(inp) && (error = EPROTOTYPE))
1088 #endif /* NECP */
1089 	    ) {
1090 		if (inp == NULL) {
1091 			error = EINVAL;
1092 		} else {
1093 			error = EPROTOTYPE;
1094 		}
1095 		goto bad;
1096 	}
1097 
1098 	if (nam != NULL) {
1099 		dst = ((struct sockaddr_in *)(void *)nam)->sin_addr.s_addr;
1100 	}
1101 	return rip_output(m, so, dst, control);
1102 
1103 bad:
1104 	VERIFY(error != 0);
1105 
1106 	if (m != NULL) {
1107 		m_freem(m);
1108 	}
1109 	if (control != NULL) {
1110 		m_freem(control);
1111 	}
1112 
1113 	return error;
1114 }
1115 
1116 /* note: rip_unlock is called from different protos  instead of the generic socket_unlock,
1117  * it will handle the socket dealloc on last reference
1118  * */
1119 int
rip_unlock(struct socket * so,int refcount,void * debug)1120 rip_unlock(struct socket *so, int refcount, void *debug)
1121 {
1122 	void *lr_saved;
1123 	struct inpcb *inp = sotoinpcb(so);
1124 
1125 	if (debug == NULL) {
1126 		lr_saved = __builtin_return_address(0);
1127 	} else {
1128 		lr_saved = debug;
1129 	}
1130 
1131 	if (refcount) {
1132 		if (so->so_usecount <= 0) {
1133 			panic("rip_unlock: bad refoucnt so=%p val=%x lrh= %s",
1134 			    so, so->so_usecount, solockhistory_nr(so));
1135 			/* NOTREACHED */
1136 		}
1137 		so->so_usecount--;
1138 		if (so->so_usecount == 0 && (inp->inp_wantcnt == WNT_STOPUSING)) {
1139 			/* cleanup after last reference */
1140 			lck_mtx_unlock(so->so_proto->pr_domain->dom_mtx);
1141 			lck_rw_lock_exclusive(&ripcbinfo.ipi_lock);
1142 			if (inp->inp_state != INPCB_STATE_DEAD) {
1143 				if (SOCK_CHECK_DOM(so, PF_INET6)) {
1144 					in6_pcbdetach(inp);
1145 				} else {
1146 					in_pcbdetach(inp);
1147 				}
1148 			}
1149 			in_pcbdispose(inp);
1150 			lck_rw_done(&ripcbinfo.ipi_lock);
1151 			return 0;
1152 		}
1153 	}
1154 	so->unlock_lr[so->next_unlock_lr] = lr_saved;
1155 	so->next_unlock_lr = (so->next_unlock_lr + 1) % SO_LCKDBG_MAX;
1156 	lck_mtx_unlock(so->so_proto->pr_domain->dom_mtx);
1157 	return 0;
1158 }
1159 
1160 static int
1161 rip_pcblist SYSCTL_HANDLER_ARGS
1162 {
1163 #pragma unused(oidp, arg1, arg2)
1164 	int error, i, n, sz;
1165 	struct inpcb *inp, **inp_list;
1166 	inp_gen_t gencnt;
1167 	struct xinpgen xig;
1168 
1169 	/*
1170 	 * The process of preparing the TCB list is too time-consuming and
1171 	 * resource-intensive to repeat twice on every request.
1172 	 */
1173 	lck_rw_lock_exclusive(&ripcbinfo.ipi_lock);
1174 	if (req->oldptr == USER_ADDR_NULL) {
1175 		n = ripcbinfo.ipi_count;
1176 		req->oldidx = 2 * (sizeof xig)
1177 		    + (n + n / 8) * sizeof(struct xinpcb);
1178 		lck_rw_done(&ripcbinfo.ipi_lock);
1179 		return 0;
1180 	}
1181 
1182 	if (req->newptr != USER_ADDR_NULL) {
1183 		lck_rw_done(&ripcbinfo.ipi_lock);
1184 		return EPERM;
1185 	}
1186 
1187 	/*
1188 	 * OK, now we're committed to doing something.
1189 	 */
1190 	gencnt = ripcbinfo.ipi_gencnt;
1191 	sz = n = ripcbinfo.ipi_count;
1192 
1193 	bzero(&xig, sizeof(xig));
1194 	xig.xig_len = sizeof xig;
1195 	xig.xig_count = n;
1196 	xig.xig_gen = gencnt;
1197 	xig.xig_sogen = so_gencnt;
1198 	error = SYSCTL_OUT(req, &xig, sizeof xig);
1199 	if (error) {
1200 		lck_rw_done(&ripcbinfo.ipi_lock);
1201 		return error;
1202 	}
1203 	/*
1204 	 * We are done if there is no pcb
1205 	 */
1206 	if (n == 0) {
1207 		lck_rw_done(&ripcbinfo.ipi_lock);
1208 		return 0;
1209 	}
1210 
1211 	inp_list = kalloc_type(struct inpcb *, n, Z_WAITOK);
1212 	if (inp_list == NULL) {
1213 		lck_rw_done(&ripcbinfo.ipi_lock);
1214 		return ENOMEM;
1215 	}
1216 
1217 	for (inp = ripcbinfo.ipi_listhead->lh_first, i = 0; inp && i < n;
1218 	    inp = inp->inp_list.le_next) {
1219 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1220 			inp_list[i++] = inp;
1221 		}
1222 	}
1223 	n = i;
1224 
1225 	error = 0;
1226 	for (i = 0; i < n; i++) {
1227 		inp = inp_list[i];
1228 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1229 			struct xinpcb xi;
1230 
1231 			bzero(&xi, sizeof(xi));
1232 			xi.xi_len = sizeof xi;
1233 			/* XXX should avoid extra copy */
1234 			inpcb_to_compat(inp, &xi.xi_inp);
1235 			if (inp->inp_socket) {
1236 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
1237 			}
1238 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1239 		}
1240 	}
1241 	if (!error) {
1242 		/*
1243 		 * Give the user an updated idea of our state.
1244 		 * If the generation differs from what we told
1245 		 * her before, she knows that something happened
1246 		 * while we were processing this request, and it
1247 		 * might be necessary to retry.
1248 		 */
1249 		bzero(&xig, sizeof(xig));
1250 		xig.xig_len = sizeof xig;
1251 		xig.xig_gen = ripcbinfo.ipi_gencnt;
1252 		xig.xig_sogen = so_gencnt;
1253 		xig.xig_count = ripcbinfo.ipi_count;
1254 		error = SYSCTL_OUT(req, &xig, sizeof xig);
1255 	}
1256 
1257 	lck_rw_done(&ripcbinfo.ipi_lock);
1258 	kfree_type(struct inpcb *, sz, inp_list);
1259 	return error;
1260 }
1261 
1262 SYSCTL_PROC(_net_inet_raw, OID_AUTO /*XXX*/, pcblist,
1263     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
1264     rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
1265 
1266 #if XNU_TARGET_OS_OSX
1267 
1268 static int
1269 rip_pcblist64 SYSCTL_HANDLER_ARGS
1270 {
1271 #pragma unused(oidp, arg1, arg2)
1272 	int error, i, n, sz;
1273 	struct inpcb *inp, **inp_list;
1274 	inp_gen_t gencnt;
1275 	struct xinpgen xig;
1276 
1277 	/*
1278 	 * The process of preparing the TCB list is too time-consuming and
1279 	 * resource-intensive to repeat twice on every request.
1280 	 */
1281 	lck_rw_lock_exclusive(&ripcbinfo.ipi_lock);
1282 	if (req->oldptr == USER_ADDR_NULL) {
1283 		n = ripcbinfo.ipi_count;
1284 		req->oldidx = 2 * (sizeof xig)
1285 		    + (n + n / 8) * sizeof(struct xinpcb64);
1286 		lck_rw_done(&ripcbinfo.ipi_lock);
1287 		return 0;
1288 	}
1289 
1290 	if (req->newptr != USER_ADDR_NULL) {
1291 		lck_rw_done(&ripcbinfo.ipi_lock);
1292 		return EPERM;
1293 	}
1294 
1295 	/*
1296 	 * OK, now we're committed to doing something.
1297 	 */
1298 	gencnt = ripcbinfo.ipi_gencnt;
1299 	sz = n = ripcbinfo.ipi_count;
1300 
1301 	bzero(&xig, sizeof(xig));
1302 	xig.xig_len = sizeof xig;
1303 	xig.xig_count = n;
1304 	xig.xig_gen = gencnt;
1305 	xig.xig_sogen = so_gencnt;
1306 	error = SYSCTL_OUT(req, &xig, sizeof xig);
1307 	if (error) {
1308 		lck_rw_done(&ripcbinfo.ipi_lock);
1309 		return error;
1310 	}
1311 	/*
1312 	 * We are done if there is no pcb
1313 	 */
1314 	if (n == 0) {
1315 		lck_rw_done(&ripcbinfo.ipi_lock);
1316 		return 0;
1317 	}
1318 
1319 	inp_list = kalloc_type(struct inpcb *, n, Z_WAITOK);
1320 	if (inp_list == NULL) {
1321 		lck_rw_done(&ripcbinfo.ipi_lock);
1322 		return ENOMEM;
1323 	}
1324 
1325 	for (inp = ripcbinfo.ipi_listhead->lh_first, i = 0; inp && i < n;
1326 	    inp = inp->inp_list.le_next) {
1327 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1328 			inp_list[i++] = inp;
1329 		}
1330 	}
1331 	n = i;
1332 
1333 	error = 0;
1334 	for (i = 0; i < n; i++) {
1335 		inp = inp_list[i];
1336 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1337 			struct xinpcb64 xi;
1338 
1339 			bzero(&xi, sizeof(xi));
1340 			xi.xi_len = sizeof xi;
1341 			inpcb_to_xinpcb64(inp, &xi);
1342 			if (inp->inp_socket) {
1343 				sotoxsocket64(inp->inp_socket, &xi.xi_socket);
1344 			}
1345 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1346 		}
1347 	}
1348 	if (!error) {
1349 		/*
1350 		 * Give the user an updated idea of our state.
1351 		 * If the generation differs from what we told
1352 		 * her before, she knows that something happened
1353 		 * while we were processing this request, and it
1354 		 * might be necessary to retry.
1355 		 */
1356 		bzero(&xig, sizeof(xig));
1357 		xig.xig_len = sizeof xig;
1358 		xig.xig_gen = ripcbinfo.ipi_gencnt;
1359 		xig.xig_sogen = so_gencnt;
1360 		xig.xig_count = ripcbinfo.ipi_count;
1361 		error = SYSCTL_OUT(req, &xig, sizeof xig);
1362 	}
1363 
1364 	lck_rw_done(&ripcbinfo.ipi_lock);
1365 	kfree_type(struct inpcb *, sz, inp_list);
1366 	return error;
1367 }
1368 
1369 SYSCTL_PROC(_net_inet_raw, OID_AUTO, pcblist64,
1370     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
1371     rip_pcblist64, "S,xinpcb64", "List of active raw IP sockets");
1372 
1373 #endif /* XNU_TARGET_OS_OSX */
1374 
1375 
1376 static int
1377 rip_pcblist_n SYSCTL_HANDLER_ARGS
1378 {
1379 #pragma unused(oidp, arg1, arg2)
1380 	int error = 0;
1381 
1382 	error = get_pcblist_n(IPPROTO_IP, req, &ripcbinfo);
1383 
1384 	return error;
1385 }
1386 
1387 SYSCTL_PROC(_net_inet_raw, OID_AUTO, pcblist_n,
1388     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
1389     rip_pcblist_n, "S,xinpcb_n", "List of active raw IP sockets");
1390 
1391 struct pr_usrreqs rip_usrreqs = {
1392 	.pru_abort =            rip_abort,
1393 	.pru_attach =           rip_attach,
1394 	.pru_bind =             rip_bind,
1395 	.pru_connect =          rip_connect,
1396 	.pru_control =          in_control,
1397 	.pru_detach =           rip_detach,
1398 	.pru_disconnect =       rip_disconnect,
1399 	.pru_peeraddr =         in_getpeeraddr,
1400 	.pru_send =             rip_send,
1401 	.pru_shutdown =         rip_shutdown,
1402 	.pru_sockaddr =         in_getsockaddr,
1403 	.pru_sosend =           sosend,
1404 	.pru_soreceive =        soreceive,
1405 };
1406 /* DSEP Review Done pl-20051213-v02 @3253 */
1407