xref: /xnu-11417.140.69/bsd/netinet/raw_ip.c (revision 43a90889846e00bfb5cf1d255cdc0a701a1e05a4)
1 /*
2  * Copyright (c) 2000-2024 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 #include <net/sockaddr_utils.h>
91 
92 #define _IP_VHL
93 #include <netinet/in.h>
94 #include <netinet/in_systm.h>
95 #include <netinet/in_tclass.h>
96 #include <netinet/ip.h>
97 #include <netinet/in_pcb.h>
98 #include <netinet/in_var.h>
99 #include <netinet/ip_var.h>
100 
101 #include <netinet6/in6_pcb.h>
102 
103 
104 #if IPSEC
105 #include <netinet6/ipsec.h>
106 #endif /*IPSEC*/
107 
108 #if DUMMYNET
109 #include <netinet/ip_dummynet.h>
110 #endif /* DUMMYNET */
111 
112 int rip_detach(struct socket *);
113 int rip_abort(struct socket *);
114 int rip_disconnect(struct socket *);
115 int rip_bind(struct socket *, struct sockaddr *, struct proc *);
116 int rip_connect(struct socket *, struct sockaddr *, struct proc *);
117 int rip_shutdown(struct socket *);
118 
119 struct  inpcbhead ripcb;
120 struct  inpcbinfo ripcbinfo;
121 
122 /* control hooks for dummynet */
123 #if DUMMYNET
124 ip_dn_ctl_t *ip_dn_ctl_ptr;
125 #endif /* DUMMYNET */
126 
127 /*
128  * Nominal space allocated to a raw ip socket.
129  */
130 #define RIPSNDQ         8192
131 #define RIPRCVQ         8192
132 
133 static KALLOC_TYPE_DEFINE(ripzone, struct inpcb, NET_KT_DEFAULT);
134 
135 /*
136  * Raw interface to IP protocol.
137  */
138 
139 /*
140  * Initialize raw connection block q.
141  */
142 void
rip_init(struct protosw * pp,struct domain * dp)143 rip_init(struct protosw *pp, struct domain *dp)
144 {
145 #pragma unused(dp)
146 	static int rip_initialized = 0;
147 	struct inpcbinfo *pcbinfo;
148 
149 	VERIFY((pp->pr_flags & (PR_INITIALIZED | PR_ATTACHED)) == PR_ATTACHED);
150 
151 	if (rip_initialized) {
152 		return;
153 	}
154 	rip_initialized = 1;
155 
156 	LIST_INIT(&ripcb);
157 	ripcbinfo.ipi_listhead = &ripcb;
158 	/*
159 	 * XXX We don't use the hash list for raw IP, but it's easier
160 	 * to allocate a one entry hash list than it is to check all
161 	 * over the place for ipi_hashbase == NULL.
162 	 */
163 	hashinit_counted_by(1, ripcbinfo.ipi_hashbase,
164 	    ripcbinfo.ipi_hashbase_count);
165 	ripcbinfo.ipi_hashmask = ripcbinfo.ipi_hashbase_count - 1;
166 	hashinit_counted_by(1, ripcbinfo.ipi_porthashbase,
167 	    ripcbinfo.ipi_porthashbase_count);
168 	ripcbinfo.ipi_porthashmask = ripcbinfo.ipi_porthashbase_count - 1;
169 	ripcbinfo.ipi_zone = ripzone;
170 
171 	pcbinfo = &ripcbinfo;
172 	/*
173 	 * allocate lock group attribute and group for udp pcb mutexes
174 	 */
175 	pcbinfo->ipi_lock_grp = lck_grp_alloc_init("ripcb", LCK_GRP_ATTR_NULL);
176 
177 	/*
178 	 * allocate the lock attribute for udp pcb mutexes
179 	 */
180 	lck_attr_setdefault(&pcbinfo->ipi_lock_attr);
181 	lck_rw_init(&pcbinfo->ipi_lock, pcbinfo->ipi_lock_grp,
182 	    &pcbinfo->ipi_lock_attr);
183 
184 	in_pcbinfo_attach(&ripcbinfo);
185 }
186 
187 static uint32_t
rip_inp_input(struct inpcb * inp,struct mbuf * m,int iphlen)188 rip_inp_input(struct inpcb *inp, struct mbuf *m, int iphlen)
189 {
190 	struct ip *ip = mtod(m, struct ip *);
191 	struct ifnet *ifp = m->m_pkthdr.rcvif;
192 	struct sockaddr_in ripsrc = {
193 		.sin_len = sizeof(ripsrc),
194 		.sin_family = AF_INET,
195 		.sin_port = 0,
196 		.sin_addr = { .s_addr = 0 },
197 		.sin_zero = {0, 0, 0, 0, 0, 0, 0, 0, }
198 	};
199 	mbuf_ref_t opts = NULL;
200 	boolean_t is_wake_pkt = false;
201 	uint32_t num_delivered = 0;
202 
203 #if NECP
204 	if (!necp_socket_is_allowed_to_send_recv_v4(inp, 0, 0,
205 	    &ip->ip_dst, &ip->ip_src, ifp, 0, NULL, NULL, NULL, NULL)) {
206 		/* do not inject data to pcb */
207 		m_freem(m);
208 		goto done;
209 	}
210 #endif /* NECP */
211 
212 	ripsrc.sin_addr = ip->ip_src;
213 
214 	if ((m->m_flags & M_PKTHDR) && (m->m_pkthdr.pkt_flags & PKTF_WAKE_PKT)) {
215 		is_wake_pkt = true;
216 	}
217 
218 	if ((inp->inp_flags & INP_CONTROLOPTS) != 0 ||
219 	    SOFLOW_ENABLED(inp->inp_socket) ||
220 	    SO_RECV_CONTROL_OPTS(inp->inp_socket)) {
221 		if (ip_savecontrol(inp, &opts, ip, m) != 0) {
222 			m_freem(m);
223 			m_freem(opts);
224 			goto done;
225 		}
226 	}
227 	if (inp->inp_flags & INP_STRIPHDR
228 #if CONTENT_FILTER
229 	    /*
230 	     * If socket is subject to Content Filter, delay stripping until reinject
231 	     */
232 	    && (!CFIL_DGRAM_FILTERED(inp->inp_socket))
233 #endif
234 	    ) {
235 		m->m_len -= iphlen;
236 		m->m_pkthdr.len -= iphlen;
237 		m->m_data += iphlen;
238 	}
239 	so_recv_data_stat(inp->inp_socket, m, 0);
240 	if (sbappendaddr(&inp->inp_socket->so_rcv,
241 	    (struct sockaddr *)&ripsrc, m, opts, NULL) != 0) {
242 		num_delivered = 1;
243 		sorwakeup(inp->inp_socket);
244 		if (is_wake_pkt) {
245 			soevent(inp->in6p_socket,
246 			    SO_FILT_HINT_LOCKED | SO_FILT_HINT_WAKE_PKT);
247 		}
248 	} else {
249 		ipstat.ips_raw_sappend_fail++;
250 	}
251 done:
252 	return num_delivered;
253 }
254 
255 /*
256  * The first pass is for IPv4 socket and the second pass for IPv6
257  */
258 static bool
rip_input_inner(struct mbuf * m,int iphlen,bool is_ipv4_pass,uint32_t * total_delivered)259 rip_input_inner(struct mbuf *m, int iphlen, bool is_ipv4_pass, uint32_t *total_delivered)
260 {
261 	struct inpcb *inp;
262 	struct inpcb *last = NULL;
263 	struct ip *ip = mtod(m, struct ip *);
264 	struct ifnet *ifp = m->m_pkthdr.rcvif;
265 	bool need_ipv6_pass = false;
266 	uint32_t num_delivered = 0;
267 
268 	lck_rw_lock_shared(&ripcbinfo.ipi_lock);
269 	LIST_FOREACH(inp, &ripcb, inp_list) {
270 		if (is_ipv4_pass) {
271 			if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) != INP_IPV4) {
272 				/* Tell if we need to an IPv6 pass */
273 				need_ipv6_pass = true;
274 				continue;
275 			}
276 		} else {
277 			if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) != (INP_IPV4 | INP_IPV6)) {
278 				continue;
279 			}
280 		}
281 		if (inp->inp_ip_p && (inp->inp_ip_p != ip->ip_p)) {
282 			continue;
283 		}
284 		if (inp->inp_laddr.s_addr &&
285 		    inp->inp_laddr.s_addr != ip->ip_dst.s_addr) {
286 			continue;
287 		}
288 		if (inp->inp_faddr.s_addr &&
289 		    inp->inp_faddr.s_addr != ip->ip_src.s_addr) {
290 			continue;
291 		}
292 		if (inp_restricted_recv(inp, ifp)) {
293 			continue;
294 		}
295 		if (last != NULL) {
296 			struct mbuf *n = m_copym_mode(m, 0, (int)M_COPYALL, M_DONTWAIT, NULL, NULL, M_COPYM_MUST_COPY_HDR);
297 
298 			if (n == NULL) {
299 				continue;
300 			}
301 			num_delivered += rip_inp_input(last, n, iphlen);
302 		}
303 		last = inp;
304 	}
305 
306 	/*
307 	 * Consume the orignal mbuf 'm' if:
308 	 * - it is the first pass and there is no IPv6 raw socket
309 	 * - it is the second pass for IPv6
310 	 */
311 	if (need_ipv6_pass == false || is_ipv4_pass == false) {
312 		if (last != NULL) {
313 			num_delivered += rip_inp_input(last, m, iphlen);
314 		} else {
315 			m_freem(m);
316 		}
317 	} else {
318 		if (last != NULL) {
319 			struct mbuf *n = m_copym_mode(m, 0, (int)M_COPYALL, M_DONTWAIT, NULL, NULL, M_COPYM_MUST_COPY_HDR);
320 
321 			if (n != NULL) {
322 				num_delivered += rip_inp_input(last, n, iphlen);
323 			}
324 		}
325 	}
326 	/*
327 	 * Keep the list locked because socket filter may force the socket lock
328 	 * to be released when calling sbappendaddr() -- see rdar://7627704
329 	 */
330 	lck_rw_done(&ripcbinfo.ipi_lock);
331 
332 	*total_delivered += num_delivered;
333 
334 	return need_ipv6_pass;
335 }
336 
337 
338 /*
339  * Setup generic address and protocol structures
340  * for raw_input routine, then pass them along with
341  * mbuf chain.
342  */
343 void
rip_input(struct mbuf * m,int iphlen)344 rip_input(struct mbuf *m, int iphlen)
345 {
346 	uint32_t num_delivered = 0;
347 	bool need_v6_pass = false;
348 
349 	/* Expect 32-bit aligned data pointer on strict-align platforms */
350 	MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
351 
352 	/*
353 	 * First pass for raw IPv4 sockets that are protected by the inet_domain_mutex lock
354 	 */
355 	need_v6_pass = rip_input_inner(m, iphlen, true, &num_delivered);
356 
357 	/*
358 	 * For the IPv6 pass we need to switch to the inet6_domain_mutex lock
359 	 * to protect the raw IPv6 sockets
360 	 */
361 	if (need_v6_pass) {
362 		lck_mtx_unlock(inet_domain_mutex);
363 
364 		lck_mtx_lock(inet6_domain_mutex);
365 		rip_input_inner(m, iphlen, false, &num_delivered);
366 		lck_mtx_unlock(inet6_domain_mutex);
367 
368 		lck_mtx_lock(inet_domain_mutex);
369 	}
370 
371 	if (num_delivered > 0) {
372 		OSAddAtomic(1, &ipstat.ips_delivered);
373 	} else {
374 		OSAddAtomic(1, &ipstat.ips_noproto);
375 	}
376 }
377 
378 /*
379  * Generate IP header and pass packet to ip_output.
380  * Tack on options user may have setup with control call.
381  */
382 int
rip_output(struct mbuf * m,struct socket * so,u_int32_t dst,struct mbuf * control)383 rip_output(
384 	struct mbuf *m,
385 	struct socket *so,
386 	u_int32_t dst,
387 	struct mbuf *control)
388 {
389 	struct ip *ip;
390 	struct inpcb *inp = sotoinpcb(so);
391 	int flags = (so->so_options & SO_DONTROUTE) | IP_ALLOWBROADCAST;
392 	int inp_flags = inp ? inp->inp_flags : 0;
393 	struct ip_out_args ipoa;
394 	struct ip_moptions *imo;
395 	int tos = IPTOS_UNSPEC;
396 	int error = 0;
397 #if CONTENT_FILTER
398 	struct m_tag *cfil_tag = NULL;
399 	bool cfil_faddr_use = false;
400 	uint32_t cfil_so_state_change_cnt = 0;
401 	uint32_t cfil_so_options = 0;
402 	int cfil_inp_flags = 0;
403 	struct sockaddr *__single cfil_faddr = NULL;
404 	struct sockaddr_in *__single cfil_sin;
405 	u_int32_t cfil_dst = 0;
406 #endif
407 
408 #if CONTENT_FILTER
409 	/*
410 	 * If socket is subject to Content Filter and no addr is passed in,
411 	 * retrieve CFIL saved state from mbuf and use it if necessary.
412 	 */
413 	if (CFIL_DGRAM_FILTERED(so) && dst == INADDR_ANY) {
414 		cfil_tag = cfil_dgram_get_socket_state(m, &cfil_so_state_change_cnt, &cfil_so_options, &cfil_faddr, &cfil_inp_flags);
415 		if (cfil_tag) {
416 			cfil_sin = SIN(cfil_faddr);
417 			flags = (cfil_so_options & SO_DONTROUTE) | IP_ALLOWBROADCAST;
418 			inp_flags = cfil_inp_flags;
419 			if (inp && inp->inp_faddr.s_addr == INADDR_ANY) {
420 				/*
421 				 * Socket is unconnected, simply use the saved faddr as 'addr' to go through
422 				 * the connect/disconnect logic.
423 				 */
424 				dst = cfil_sin->sin_addr.s_addr;
425 			} else if ((so->so_state_change_cnt != cfil_so_state_change_cnt) &&
426 			    (inp->inp_fport != cfil_sin->sin_port ||
427 			    inp->inp_faddr.s_addr != cfil_sin->sin_addr.s_addr)) {
428 				/*
429 				 * Socket is connected but socket state and dest addr/port changed.
430 				 * We need to use the saved faddr and socket options.
431 				 */
432 				cfil_faddr_use = true;
433 				cfil_dst = cfil_sin->sin_addr.s_addr;
434 			}
435 			m_tag_free(cfil_tag);
436 		}
437 	}
438 #endif
439 
440 	if (so->so_state & SS_ISCONNECTED) {
441 		if (dst != INADDR_ANY) {
442 			if (m != NULL) {
443 				m_freem(m);
444 			}
445 			if (control != NULL) {
446 				m_freem(control);
447 			}
448 			return EISCONN;
449 		}
450 		dst = cfil_faddr_use ? cfil_dst : inp->inp_faddr.s_addr;
451 	} else {
452 		if (dst == INADDR_ANY) {
453 			if (m != NULL) {
454 				m_freem(m);
455 			}
456 			if (control != NULL) {
457 				m_freem(control);
458 			}
459 			return ENOTCONN;
460 		}
461 	}
462 
463 	bzero(&ipoa, sizeof(ipoa));
464 	ipoa.ipoa_boundif = IFSCOPE_NONE;
465 	ipoa.ipoa_flags = IPOAF_SELECT_SRCIF;
466 
467 	int sotc = SO_TC_UNSPEC;
468 	int netsvctype = _NET_SERVICE_TYPE_UNSPEC;
469 
470 
471 	if (control != NULL) {
472 		tos = so_tos_from_control(control);
473 		sotc = so_tc_from_control(control, &netsvctype);
474 
475 		m_freem(control);
476 		control = NULL;
477 	}
478 	if (sotc == SO_TC_UNSPEC) {
479 		sotc = so->so_traffic_class;
480 		netsvctype = so->so_netsvctype;
481 	}
482 
483 	if (inp == NULL
484 #if NECP
485 	    || (necp_socket_should_use_flow_divert(inp))
486 #endif /* NECP */
487 	    ) {
488 		if (m != NULL) {
489 			m_freem(m);
490 		}
491 		VERIFY(control == NULL);
492 		return inp == NULL ? EINVAL : EPROTOTYPE;
493 	}
494 
495 	flags |= IP_OUTARGS;
496 	/* If socket was bound to an ifindex, tell ip_output about it */
497 	if (inp->inp_flags & INP_BOUND_IF) {
498 		ipoa.ipoa_boundif = inp->inp_boundifp->if_index;
499 		ipoa.ipoa_flags |= IPOAF_BOUND_IF;
500 	}
501 	if (INP_NO_CELLULAR(inp)) {
502 		ipoa.ipoa_flags |=  IPOAF_NO_CELLULAR;
503 	}
504 	if (INP_NO_EXPENSIVE(inp)) {
505 		ipoa.ipoa_flags |=  IPOAF_NO_EXPENSIVE;
506 	}
507 	if (INP_NO_CONSTRAINED(inp)) {
508 		ipoa.ipoa_flags |=  IPOAF_NO_CONSTRAINED;
509 	}
510 	if (INP_AWDL_UNRESTRICTED(inp)) {
511 		ipoa.ipoa_flags |=  IPOAF_AWDL_UNRESTRICTED;
512 	}
513 	if (INP_MANAGEMENT_ALLOWED(inp)) {
514 		ipoa.ipoa_flags |=  IPOAF_MANAGEMENT_ALLOWED;
515 	}
516 	if (INP_ULTRA_CONSTRAINED_ALLOWED(inp)) {
517 		ipoa.ipoa_flags |=  IPOAF_ULTRA_CONSTRAINED_ALLOWED;
518 	}
519 	ipoa.ipoa_sotc = sotc;
520 	ipoa.ipoa_netsvctype = netsvctype;
521 
522 	if (inp->inp_flowhash == 0) {
523 		inp_calc_flowhash(inp);
524 		ASSERT(inp->inp_flowhash != 0);
525 	}
526 
527 	/*
528 	 * If the user handed us a complete IP packet, use it.
529 	 * Otherwise, allocate an mbuf for a header and fill it in.
530 	 */
531 	if ((inp_flags & INP_HDRINCL) == 0) {
532 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
533 			m_freem(m);
534 			return EMSGSIZE;
535 		}
536 		M_PREPEND(m, sizeof(struct ip), M_WAIT, 1);
537 		if (m == NULL) {
538 			return ENOBUFS;
539 		}
540 		ip = mtod(m, struct ip *);
541 		if (tos != IPTOS_UNSPEC) {
542 			ip->ip_tos = (uint8_t)(tos & IPTOS_MASK);
543 		} else {
544 			ip->ip_tos = inp->inp_ip_tos;
545 		}
546 		if (inp->inp_flags2 & INP2_DONTFRAG) {
547 			ip->ip_off = IP_DF;
548 		} else {
549 			ip->ip_off = 0;
550 		}
551 		ip->ip_p = inp->inp_ip_p;
552 		ip->ip_len = (uint16_t)m->m_pkthdr.len;
553 		ip->ip_src = inp->inp_laddr;
554 		ip->ip_dst.s_addr = dst;
555 		ip->ip_ttl = inp->inp_ip_ttl;
556 	} else {
557 		if (m->m_pkthdr.len > IP_MAXPACKET) {
558 			m_freem(m);
559 			return EMSGSIZE;
560 		}
561 		ip = mtod(m, struct ip *);
562 		/*
563 		 * don't allow both user specified and setsockopt options,
564 		 * and don't allow packet length sizes that will crash
565 		 */
566 		if (m->m_pkthdr.len < sizeof(struct ip) ||
567 		    ((IP_VHL_HL(ip->ip_vhl) != (sizeof(*ip) >> 2)) && inp->inp_options) ||
568 		    (ip->ip_len > m->m_pkthdr.len) ||
569 		    (ip->ip_len < (IP_VHL_HL(ip->ip_vhl) << 2))) {
570 			m_freem(m);
571 			return EINVAL;
572 		}
573 		if (ip->ip_id == 0 && !(rfc6864 && IP_OFF_IS_ATOMIC(ntohs(ip->ip_off)))) {
574 			ip->ip_id = ip_randomid((uint64_t)m);
575 		}
576 		/* XXX prevent ip_output from overwriting header fields */
577 		flags |= IP_RAWOUTPUT;
578 		OSAddAtomic(1, &ipstat.ips_rawout);
579 	}
580 
581 	if (inp->inp_laddr.s_addr != INADDR_ANY) {
582 		ipoa.ipoa_flags |= IPOAF_BOUND_SRCADDR;
583 	}
584 
585 #if NECP
586 	{
587 		necp_kernel_policy_id policy_id;
588 		necp_kernel_policy_id skip_policy_id;
589 		u_int32_t route_rule_id;
590 		u_int32_t pass_flags;
591 
592 		/*
593 		 * We need a route to perform NECP route rule checks
594 		 */
595 		if ((net_qos_policy_restricted != 0 &&
596 		    ROUTE_UNUSABLE(&inp->inp_route))
597 #if CONTENT_FILTER
598 		    || cfil_faddr_use
599 #endif
600 		    ) {
601 			struct sockaddr_in to;
602 			struct sockaddr_in from;
603 			struct in_addr laddr = ip->ip_src;
604 
605 			ROUTE_RELEASE(&inp->inp_route);
606 
607 			bzero(&from, sizeof(struct sockaddr_in));
608 			from.sin_family = AF_INET;
609 			from.sin_len = sizeof(struct sockaddr_in);
610 			from.sin_addr = laddr;
611 
612 			bzero(&to, sizeof(struct sockaddr_in));
613 			to.sin_family = AF_INET;
614 			to.sin_len = sizeof(struct sockaddr_in);
615 			to.sin_addr.s_addr = ip->ip_dst.s_addr;
616 
617 			if ((error = in_pcbladdr(inp, (struct sockaddr *)&to,
618 			    &laddr, ipoa.ipoa_boundif, NULL, 1)) != 0) {
619 				printf("%s in_pcbladdr(%p) error %d\n",
620 				    __func__, inp, error);
621 				m_freem(m);
622 				return error;
623 			}
624 
625 			inp_update_necp_policy(inp, (struct sockaddr *)&from,
626 			    (struct sockaddr *)&to, ipoa.ipoa_boundif);
627 			inp->inp_policyresult.results.qos_marking_gencount = 0;
628 		}
629 
630 		if (!necp_socket_is_allowed_to_send_recv_v4(inp, 0, 0,
631 		    &ip->ip_src, &ip->ip_dst, NULL, 0, &policy_id, &route_rule_id, &skip_policy_id, &pass_flags)) {
632 			m_freem(m);
633 			return EHOSTUNREACH;
634 		}
635 
636 		necp_mark_packet_from_socket(m, inp, policy_id, route_rule_id, skip_policy_id, pass_flags);
637 
638 		if (net_qos_policy_restricted != 0) {
639 			struct ifnet *rt_ifp = NULL;
640 
641 			if (inp->inp_route.ro_rt != NULL) {
642 				rt_ifp = inp->inp_route.ro_rt->rt_ifp;
643 			}
644 
645 			necp_socket_update_qos_marking(inp, inp->inp_route.ro_rt, route_rule_id);
646 		}
647 	}
648 #endif /* NECP */
649 	if ((so->so_flags1 & SOF1_QOSMARKING_ALLOWED)) {
650 		ipoa.ipoa_flags |= IPOAF_QOSMARKING_ALLOWED;
651 	}
652 #if IPSEC
653 	if (inp->inp_sp != NULL && ipsec_setsocket(m, so) != 0) {
654 		m_freem(m);
655 		return ENOBUFS;
656 	}
657 #endif /*IPSEC*/
658 
659 	if (ROUTE_UNUSABLE(&inp->inp_route)) {
660 		ROUTE_RELEASE(&inp->inp_route);
661 	}
662 
663 	set_packet_service_class(m, so, sotc, 0);
664 	m->m_pkthdr.pkt_flowsrc = FLOWSRC_INPCB;
665 	m->m_pkthdr.pkt_flowid = inp->inp_flowhash;
666 	m->m_pkthdr.pkt_flags |= (PKTF_FLOW_ID | PKTF_FLOW_LOCALSRC |
667 	    PKTF_FLOW_RAWSOCK);
668 	m->m_pkthdr.pkt_proto = inp->inp_ip_p;
669 	m->m_pkthdr.tx_rawip_pid = so->last_pid;
670 	m->m_pkthdr.tx_rawip_e_pid = so->e_pid;
671 	if (so->so_flags & SOF_DELEGATED) {
672 		m->m_pkthdr.tx_rawip_e_pid = so->e_pid;
673 	} else {
674 		m->m_pkthdr.tx_rawip_e_pid = 0;
675 	}
676 #if (DEBUG || DEVELOPMENT)
677 	if (so->so_flags & SOF_MARK_WAKE_PKT) {
678 		so->so_flags &= ~SOF_MARK_WAKE_PKT;
679 		m->m_pkthdr.pkt_flags |= PKTF_WAKE_PKT;
680 	}
681 #endif /* (DEBUG || DEVELOPMENT) */
682 
683 	imo = inp->inp_moptions;
684 	if (imo != NULL) {
685 		IMO_ADDREF(imo);
686 	}
687 	/*
688 	 * The domain lock is held across ip_output, so it is okay
689 	 * to pass the PCB cached route pointer directly to IP and
690 	 * the modules beneath it.
691 	 */
692 	// TODO: PASS DOWN ROUTE RULE ID
693 	error = ip_output(m, inp->inp_options, &inp->inp_route, flags,
694 	    imo, &ipoa);
695 
696 	if (imo != NULL) {
697 		IMO_REMREF(imo);
698 	}
699 
700 	if (inp->inp_route.ro_rt != NULL) {
701 		struct rtentry *rt = inp->inp_route.ro_rt;
702 		struct ifnet *outif;
703 
704 		if ((rt->rt_flags & (RTF_MULTICAST | RTF_BROADCAST)) ||
705 		    inp->inp_socket == NULL ||
706 #if CONTENT_FILTER
707 		    /* Discard temporary route for cfil case */
708 		    cfil_faddr_use ||
709 #endif
710 		    !(inp->inp_socket->so_state & SS_ISCONNECTED)) {
711 			rt = NULL;      /* unusable */
712 		}
713 		/*
714 		 * Always discard the cached route for unconnected
715 		 * socket or if it is a multicast route.
716 		 */
717 		if (rt == NULL) {
718 			ROUTE_RELEASE(&inp->inp_route);
719 		}
720 
721 		/*
722 		 * If this is a connected socket and the destination
723 		 * route is unicast, update outif with that of the
724 		 * route interface used by IP.
725 		 */
726 		if (rt != NULL &&
727 		    (outif = rt->rt_ifp) != inp->inp_last_outifp) {
728 			inp->inp_last_outifp = outif;
729 		}
730 	} else {
731 		ROUTE_RELEASE(&inp->inp_route);
732 	}
733 
734 	/*
735 	 * If output interface was cellular/expensive/constrained, and this socket is
736 	 * denied access to it, generate an event.
737 	 */
738 	if (error != 0 && (ipoa.ipoa_flags & IPOAF_R_IFDENIED) &&
739 	    (INP_NO_CELLULAR(inp) || INP_NO_EXPENSIVE(inp) || INP_NO_CONSTRAINED(inp))) {
740 		soevent(so, (SO_FILT_HINT_LOCKED | SO_FILT_HINT_IFDENIED));
741 	}
742 
743 	return error;
744 }
745 
746 
747 /*
748  * Raw IP socket option processing.
749  */
750 int
rip_ctloutput(struct socket * so,struct sockopt * sopt)751 rip_ctloutput(struct socket *so, struct sockopt *sopt)
752 {
753 	struct  inpcb *inp = sotoinpcb(so);
754 	int     error, optval;
755 
756 	/* Allow <SOL_SOCKET,SO_BINDTODEVICE> at this level */
757 	if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_BINDTODEVICE) {
758 		return ip_ctloutput(so, sopt);
759 	}
760 
761 	if (sopt->sopt_level != IPPROTO_IP) {
762 		return EINVAL;
763 	}
764 
765 	error = 0;
766 
767 	switch (sopt->sopt_dir) {
768 	case SOPT_GET:
769 		switch (sopt->sopt_name) {
770 		case IP_HDRINCL:
771 			optval = inp->inp_flags & INP_HDRINCL;
772 			error = sooptcopyout(sopt, &optval, sizeof optval);
773 			break;
774 
775 		case IP_STRIPHDR:
776 			optval = inp->inp_flags & INP_STRIPHDR;
777 			error = sooptcopyout(sopt, &optval, sizeof optval);
778 			break;
779 
780 
781 #if DUMMYNET
782 		case IP_DUMMYNET_GET:
783 			if (!DUMMYNET_LOADED) {
784 				ip_dn_init();
785 			}
786 			if (DUMMYNET_LOADED) {
787 				error = ip_dn_ctl_ptr(sopt);
788 			} else {
789 				error = ENOPROTOOPT;
790 			}
791 			break;
792 #endif /* DUMMYNET */
793 
794 		default:
795 			error = ip_ctloutput(so, sopt);
796 			break;
797 		}
798 		break;
799 
800 	case SOPT_SET:
801 		switch (sopt->sopt_name) {
802 		case IP_HDRINCL:
803 			error = sooptcopyin(sopt, &optval, sizeof optval,
804 			    sizeof optval);
805 			if (error) {
806 				break;
807 			}
808 			if (optval) {
809 				inp->inp_flags |= INP_HDRINCL;
810 			} else {
811 				inp->inp_flags &= ~INP_HDRINCL;
812 			}
813 			break;
814 
815 		case IP_STRIPHDR:
816 			error = sooptcopyin(sopt, &optval, sizeof optval,
817 			    sizeof optval);
818 			if (error) {
819 				break;
820 			}
821 			if (optval) {
822 				inp->inp_flags |= INP_STRIPHDR;
823 			} else {
824 				inp->inp_flags &= ~INP_STRIPHDR;
825 			}
826 			break;
827 
828 
829 #if DUMMYNET
830 		case IP_DUMMYNET_CONFIGURE:
831 		case IP_DUMMYNET_DEL:
832 		case IP_DUMMYNET_FLUSH:
833 			if (!DUMMYNET_LOADED) {
834 				ip_dn_init();
835 			}
836 			if (DUMMYNET_LOADED) {
837 				error = ip_dn_ctl_ptr(sopt);
838 			} else {
839 				error = ENOPROTOOPT;
840 			}
841 			break;
842 #endif /* DUMMYNET */
843 
844 		case SO_FLUSH:
845 			if ((error = sooptcopyin(sopt, &optval, sizeof(optval),
846 			    sizeof(optval))) != 0) {
847 				break;
848 			}
849 
850 			error = inp_flush(inp, optval);
851 			break;
852 
853 		default:
854 			error = ip_ctloutput(so, sopt);
855 			break;
856 		}
857 		break;
858 	}
859 
860 	return error;
861 }
862 
863 /*
864  * This function exists solely to receive the PRC_IFDOWN messages which
865  * are sent by if_down().  It looks for an ifaddr whose ifa_addr is sa,
866  * and calls in_ifadown() to remove all routes corresponding to that address.
867  * It also receives the PRC_IFUP messages from if_up() and reinstalls the
868  * interface routes.
869  */
870 void
rip_ctlinput(int cmd,struct sockaddr * sa,__unused void * vip,__unused struct ifnet * ifp)871 rip_ctlinput(
872 	int cmd,
873 	struct sockaddr *sa,
874 	__unused void *vip,
875 	__unused struct ifnet *ifp)
876 {
877 	struct in_ifaddr *ia = NULL;
878 	struct ifnet *iaifp = NULL;
879 	int err = 0;
880 	int flags, done = 0;
881 
882 	switch (cmd) {
883 	case PRC_IFDOWN:
884 		lck_rw_lock_shared(&in_ifaddr_rwlock);
885 		for (ia = in_ifaddrhead.tqh_first; ia;
886 		    ia = ia->ia_link.tqe_next) {
887 			IFA_LOCK(&ia->ia_ifa);
888 			if (ia->ia_ifa.ifa_addr == sa &&
889 			    (ia->ia_flags & IFA_ROUTE)) {
890 				done = 1;
891 				ifa_addref(&ia->ia_ifa);
892 				IFA_UNLOCK(&ia->ia_ifa);
893 				lck_rw_done(&in_ifaddr_rwlock);
894 				lck_mtx_lock(rnh_lock);
895 				/*
896 				 * in_ifscrub kills the interface route.
897 				 */
898 				in_ifscrub(ia->ia_ifp, ia, 1);
899 				/*
900 				 * in_ifadown gets rid of all the rest of
901 				 * the routes.  This is not quite the right
902 				 * thing to do, but at least if we are running
903 				 * a routing process they will come back.
904 				 */
905 				in_ifadown(&ia->ia_ifa, 1);
906 				lck_mtx_unlock(rnh_lock);
907 				ifa_remref(&ia->ia_ifa);
908 				break;
909 			}
910 			IFA_UNLOCK(&ia->ia_ifa);
911 		}
912 		if (!done) {
913 			lck_rw_done(&in_ifaddr_rwlock);
914 		}
915 		break;
916 
917 	case PRC_IFUP:
918 		lck_rw_lock_shared(&in_ifaddr_rwlock);
919 		for (ia = in_ifaddrhead.tqh_first; ia;
920 		    ia = ia->ia_link.tqe_next) {
921 			IFA_LOCK(&ia->ia_ifa);
922 			if (ia->ia_ifa.ifa_addr == sa) {
923 				/* keep it locked */
924 				break;
925 			}
926 			IFA_UNLOCK(&ia->ia_ifa);
927 		}
928 		if (ia == NULL || (ia->ia_flags & IFA_ROUTE) ||
929 		    (ia->ia_ifa.ifa_debug & IFD_NOTREADY)) {
930 			if (ia != NULL) {
931 				IFA_UNLOCK(&ia->ia_ifa);
932 			}
933 			lck_rw_done(&in_ifaddr_rwlock);
934 			return;
935 		}
936 		ifa_addref(&ia->ia_ifa);
937 		IFA_UNLOCK(&ia->ia_ifa);
938 		lck_rw_done(&in_ifaddr_rwlock);
939 
940 		flags = RTF_UP;
941 		iaifp = ia->ia_ifa.ifa_ifp;
942 
943 		if ((iaifp->if_flags & IFF_LOOPBACK)
944 		    || (iaifp->if_flags & IFF_POINTOPOINT)) {
945 			flags |= RTF_HOST;
946 		}
947 
948 		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
949 		if (err == 0) {
950 			IFA_LOCK_SPIN(&ia->ia_ifa);
951 			ia->ia_flags |= IFA_ROUTE;
952 			IFA_UNLOCK(&ia->ia_ifa);
953 		}
954 		ifa_remref(&ia->ia_ifa);
955 		break;
956 	}
957 }
958 
959 u_int32_t       rip_sendspace = RIPSNDQ;
960 u_int32_t       rip_recvspace = RIPRCVQ;
961 
962 SYSCTL_INT(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW | CTLFLAG_LOCKED,
963     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
964 SYSCTL_INT(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW | CTLFLAG_LOCKED,
965     &rip_recvspace, 0, "Maximum incoming raw IP datagram size");
966 SYSCTL_UINT(_net_inet_raw, OID_AUTO, pcbcount, CTLFLAG_RD | CTLFLAG_LOCKED,
967     &ripcbinfo.ipi_count, 0, "Number of active PCBs");
968 
969 static int
rip_attach(struct socket * so,int proto,struct proc * p)970 rip_attach(struct socket *so, int proto, struct proc *p)
971 {
972 	struct inpcb *inp;
973 	int error;
974 
975 	inp = sotoinpcb(so);
976 	if (inp) {
977 		panic("rip_attach");
978 	}
979 	if ((so->so_state & SS_PRIV) == 0) {
980 		return EPERM;
981 	}
982 	if (proto > UINT8_MAX) {
983 		return EINVAL;
984 	}
985 
986 	error = soreserve(so, rip_sendspace, rip_recvspace);
987 	if (error) {
988 		return error;
989 	}
990 	error = in_pcballoc(so, &ripcbinfo, p);
991 	if (error) {
992 		return error;
993 	}
994 	inp = (struct inpcb *)so->so_pcb;
995 	inp->inp_vflag |= INP_IPV4;
996 	VERIFY(proto <= UINT8_MAX);
997 	inp->inp_ip_p = (u_char)proto;
998 	inp->inp_ip_ttl = (u_char)ip_defttl;
999 	return 0;
1000 }
1001 
1002 __private_extern__ int
rip_detach(struct socket * so)1003 rip_detach(struct socket *so)
1004 {
1005 	struct inpcb *inp;
1006 
1007 	inp = sotoinpcb(so);
1008 	if (inp == 0) {
1009 		panic("rip_detach");
1010 	}
1011 	in_pcbdetach(inp);
1012 	return 0;
1013 }
1014 
1015 __private_extern__ int
rip_abort(struct socket * so)1016 rip_abort(struct socket *so)
1017 {
1018 	soisdisconnected(so);
1019 	return rip_detach(so);
1020 }
1021 
1022 __private_extern__ int
rip_disconnect(struct socket * so)1023 rip_disconnect(struct socket *so)
1024 {
1025 	if ((so->so_state & SS_ISCONNECTED) == 0) {
1026 		return ENOTCONN;
1027 	}
1028 	return rip_abort(so);
1029 }
1030 
1031 __private_extern__ int
rip_bind(struct socket * so,struct sockaddr * nam,struct proc * p)1032 rip_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
1033 {
1034 #pragma unused(p)
1035 	struct inpcb *inp = sotoinpcb(so);
1036 	struct sockaddr_in sin;
1037 	struct ifaddr *ifa = NULL;
1038 	struct ifnet *outif = NULL;
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 
1048 	if (nam->sa_len != sizeof(struct sockaddr_in)) {
1049 		return EINVAL;
1050 	}
1051 
1052 	/* Sanitized local copy for interface address searches */
1053 	bzero(&sin, sizeof(sin));
1054 	sin.sin_family = AF_INET;
1055 	sin.sin_len = sizeof(struct sockaddr_in);
1056 	sin.sin_addr.s_addr = SIN(nam)->sin_addr.s_addr;
1057 
1058 	if (TAILQ_EMPTY(&ifnet_head) ||
1059 	    (sin.sin_family != AF_INET && sin.sin_family != AF_IMPLINK) ||
1060 	    (sin.sin_addr.s_addr && (ifa = ifa_ifwithaddr(SA(&sin))) == 0)) {
1061 		return EADDRNOTAVAIL;
1062 	} else if (ifa) {
1063 		/*
1064 		 * Opportunistically determine the outbound
1065 		 * interface that may be used; this may not
1066 		 * hold true if we end up using a route
1067 		 * going over a different interface, e.g.
1068 		 * when sending to a local address.  This
1069 		 * will get updated again after sending.
1070 		 */
1071 		IFA_LOCK(ifa);
1072 		outif = ifa->ifa_ifp;
1073 		IFA_UNLOCK(ifa);
1074 		ifa_remref(ifa);
1075 	}
1076 	inp->inp_laddr = sin.sin_addr;
1077 	inp->inp_last_outifp = outif;
1078 
1079 	return 0;
1080 }
1081 
1082 __private_extern__ int
rip_connect(struct socket * so,struct sockaddr * nam,__unused struct proc * p)1083 rip_connect(struct socket *so, struct sockaddr *nam, __unused  struct proc *p)
1084 {
1085 	struct inpcb *inp = sotoinpcb(so);
1086 	struct sockaddr_in *addr = (struct sockaddr_in *)(void *)nam;
1087 
1088 	if (inp == NULL
1089 #if NECP
1090 	    || (necp_socket_should_use_flow_divert(inp))
1091 #endif /* NECP */
1092 	    ) {
1093 		return inp == NULL ? EINVAL : EPROTOTYPE;
1094 	}
1095 	if (nam->sa_len != sizeof(*addr)) {
1096 		return EINVAL;
1097 	}
1098 	if (TAILQ_EMPTY(&ifnet_head)) {
1099 		return EADDRNOTAVAIL;
1100 	}
1101 	if ((addr->sin_family != AF_INET) &&
1102 	    (addr->sin_family != AF_IMPLINK)) {
1103 		return EAFNOSUPPORT;
1104 	}
1105 
1106 	if (!(so->so_flags1 & SOF1_CONNECT_COUNTED)) {
1107 		so->so_flags1 |= SOF1_CONNECT_COUNTED;
1108 		INC_ATOMIC_INT64_LIM(net_api_stats.nas_socket_inet_dgram_connected);
1109 	}
1110 
1111 	inp->inp_faddr = addr->sin_addr;
1112 	soisconnected(so);
1113 
1114 	return 0;
1115 }
1116 
1117 __private_extern__ int
rip_shutdown(struct socket * so)1118 rip_shutdown(struct socket *so)
1119 {
1120 	socantsendmore(so);
1121 	return 0;
1122 }
1123 
1124 __private_extern__ int
rip_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct proc * p)1125 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
1126     struct mbuf *control, struct proc *p)
1127 {
1128 #pragma unused(flags, p)
1129 	struct inpcb *inp = sotoinpcb(so);
1130 	u_int32_t dst = INADDR_ANY;
1131 	int error = 0;
1132 
1133 	if (inp == NULL
1134 #if NECP
1135 	    || (necp_socket_should_use_flow_divert(inp) && (error = EPROTOTYPE))
1136 #endif /* NECP */
1137 	    ) {
1138 		if (inp == NULL) {
1139 			error = EINVAL;
1140 		} else {
1141 			error = EPROTOTYPE;
1142 		}
1143 		goto bad;
1144 	}
1145 
1146 	in_pcb_check_management_entitled(inp);
1147 	in_pcb_check_ultra_constrained_entitled(inp);
1148 
1149 	so_update_tx_data_stats(so, 1, m->m_pkthdr.len);
1150 
1151 	if (nam != NULL) {
1152 		dst = ((struct sockaddr_in *)(void *)nam)->sin_addr.s_addr;
1153 	}
1154 	return rip_output(m, so, dst, control);
1155 
1156 bad:
1157 	VERIFY(error != 0);
1158 
1159 	if (m != NULL) {
1160 		m_freem(m);
1161 	}
1162 	if (control != NULL) {
1163 		m_freem(control);
1164 	}
1165 
1166 	return error;
1167 }
1168 
1169 /* note: rip_unlock is called from different protos  instead of the generic socket_unlock,
1170  * it will handle the socket dealloc on last reference
1171  * */
1172 int
rip_unlock(struct socket * so,int refcount,void * debug)1173 rip_unlock(struct socket *so, int refcount, void *debug)
1174 {
1175 	void *__single lr_saved;
1176 	struct inpcb *inp = sotoinpcb(so);
1177 
1178 	if (debug == NULL) {
1179 		lr_saved = __unsafe_forge_single(void *, __builtin_return_address(0));
1180 	} else {
1181 		lr_saved = debug;
1182 	}
1183 
1184 	if (refcount) {
1185 		if (so->so_usecount <= 0) {
1186 			panic("rip_unlock: bad refoucnt so=%p val=%x lrh= %s",
1187 			    so, so->so_usecount, solockhistory_nr(so));
1188 			/* NOTREACHED */
1189 		}
1190 		so->so_usecount--;
1191 		if (so->so_usecount == 0 && (inp->inp_wantcnt == WNT_STOPUSING)) {
1192 			/* cleanup after last reference */
1193 			lck_mtx_unlock(so->so_proto->pr_domain->dom_mtx);
1194 			lck_rw_lock_exclusive(&ripcbinfo.ipi_lock);
1195 			if (inp->inp_state != INPCB_STATE_DEAD) {
1196 				if (SOCK_CHECK_DOM(so, PF_INET6)) {
1197 					in6_pcbdetach(inp);
1198 				} else {
1199 					in_pcbdetach(inp);
1200 				}
1201 			}
1202 			in_pcbdispose(inp);
1203 			lck_rw_done(&ripcbinfo.ipi_lock);
1204 			return 0;
1205 		}
1206 	}
1207 	so->unlock_lr[so->next_unlock_lr] = lr_saved;
1208 	so->next_unlock_lr = (so->next_unlock_lr + 1) % SO_LCKDBG_MAX;
1209 	lck_mtx_unlock(so->so_proto->pr_domain->dom_mtx);
1210 	return 0;
1211 }
1212 
1213 static int
1214 rip_pcblist SYSCTL_HANDLER_ARGS
1215 {
1216 #pragma unused(oidp, arg1, arg2)
1217 	int error, i, n, sz;
1218 	struct inpcb *inp, **inp_list;
1219 	inp_gen_t gencnt;
1220 	struct xinpgen xig;
1221 
1222 	/*
1223 	 * The process of preparing the TCB list is too time-consuming and
1224 	 * resource-intensive to repeat twice on every request.
1225 	 */
1226 	lck_rw_lock_exclusive(&ripcbinfo.ipi_lock);
1227 	if (req->oldptr == USER_ADDR_NULL) {
1228 		n = ripcbinfo.ipi_count;
1229 		req->oldidx = 2 * (sizeof xig)
1230 		    + (n + n / 8) * sizeof(struct xinpcb);
1231 		lck_rw_done(&ripcbinfo.ipi_lock);
1232 		return 0;
1233 	}
1234 
1235 	if (req->newptr != USER_ADDR_NULL) {
1236 		lck_rw_done(&ripcbinfo.ipi_lock);
1237 		return EPERM;
1238 	}
1239 
1240 	/*
1241 	 * OK, now we're committed to doing something.
1242 	 */
1243 	gencnt = ripcbinfo.ipi_gencnt;
1244 	sz = n = ripcbinfo.ipi_count;
1245 
1246 	bzero(&xig, sizeof(xig));
1247 	xig.xig_len = sizeof xig;
1248 	xig.xig_count = n;
1249 	xig.xig_gen = gencnt;
1250 	xig.xig_sogen = so_gencnt;
1251 	error = SYSCTL_OUT(req, &xig, sizeof xig);
1252 	if (error) {
1253 		lck_rw_done(&ripcbinfo.ipi_lock);
1254 		return error;
1255 	}
1256 	/*
1257 	 * We are done if there is no pcb
1258 	 */
1259 	if (n == 0) {
1260 		lck_rw_done(&ripcbinfo.ipi_lock);
1261 		return 0;
1262 	}
1263 
1264 	inp_list = kalloc_type(struct inpcb *, n, Z_WAITOK);
1265 	if (inp_list == NULL) {
1266 		lck_rw_done(&ripcbinfo.ipi_lock);
1267 		return ENOMEM;
1268 	}
1269 
1270 	for (inp = ripcbinfo.ipi_listhead->lh_first, i = 0; inp && i < n;
1271 	    inp = inp->inp_list.le_next) {
1272 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1273 			inp_list[i++] = inp;
1274 		}
1275 	}
1276 	n = i;
1277 
1278 	error = 0;
1279 	for (i = 0; i < n; i++) {
1280 		inp = inp_list[i];
1281 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1282 			struct xinpcb xi;
1283 
1284 			bzero(&xi, sizeof(xi));
1285 			xi.xi_len = sizeof xi;
1286 			/* XXX should avoid extra copy */
1287 			inpcb_to_compat(inp, &xi.xi_inp);
1288 			if (inp->inp_socket) {
1289 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
1290 			}
1291 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1292 		}
1293 	}
1294 	if (!error) {
1295 		/*
1296 		 * Give the user an updated idea of our state.
1297 		 * If the generation differs from what we told
1298 		 * her before, she knows that something happened
1299 		 * while we were processing this request, and it
1300 		 * might be necessary to retry.
1301 		 */
1302 		bzero(&xig, sizeof(xig));
1303 		xig.xig_len = sizeof xig;
1304 		xig.xig_gen = ripcbinfo.ipi_gencnt;
1305 		xig.xig_sogen = so_gencnt;
1306 		xig.xig_count = ripcbinfo.ipi_count;
1307 		error = SYSCTL_OUT(req, &xig, sizeof xig);
1308 	}
1309 
1310 	lck_rw_done(&ripcbinfo.ipi_lock);
1311 	kfree_type(struct inpcb *, sz, inp_list);
1312 	return error;
1313 }
1314 
1315 SYSCTL_PROC(_net_inet_raw, OID_AUTO /*XXX*/, pcblist,
1316     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
1317     rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
1318 
1319 #if XNU_TARGET_OS_OSX
1320 
1321 static int
1322 rip_pcblist64 SYSCTL_HANDLER_ARGS
1323 {
1324 #pragma unused(oidp, arg1, arg2)
1325 	int error, i, n, sz;
1326 	struct inpcb *inp, **inp_list;
1327 	inp_gen_t gencnt;
1328 	struct xinpgen xig;
1329 
1330 	/*
1331 	 * The process of preparing the TCB list is too time-consuming and
1332 	 * resource-intensive to repeat twice on every request.
1333 	 */
1334 	lck_rw_lock_exclusive(&ripcbinfo.ipi_lock);
1335 	if (req->oldptr == USER_ADDR_NULL) {
1336 		n = ripcbinfo.ipi_count;
1337 		req->oldidx = 2 * (sizeof xig)
1338 		    + (n + n / 8) * sizeof(struct xinpcb64);
1339 		lck_rw_done(&ripcbinfo.ipi_lock);
1340 		return 0;
1341 	}
1342 
1343 	if (req->newptr != USER_ADDR_NULL) {
1344 		lck_rw_done(&ripcbinfo.ipi_lock);
1345 		return EPERM;
1346 	}
1347 
1348 	/*
1349 	 * OK, now we're committed to doing something.
1350 	 */
1351 	gencnt = ripcbinfo.ipi_gencnt;
1352 	sz = n = ripcbinfo.ipi_count;
1353 
1354 	bzero(&xig, sizeof(xig));
1355 	xig.xig_len = sizeof xig;
1356 	xig.xig_count = n;
1357 	xig.xig_gen = gencnt;
1358 	xig.xig_sogen = so_gencnt;
1359 	error = SYSCTL_OUT(req, &xig, sizeof xig);
1360 	if (error) {
1361 		lck_rw_done(&ripcbinfo.ipi_lock);
1362 		return error;
1363 	}
1364 	/*
1365 	 * We are done if there is no pcb
1366 	 */
1367 	if (n == 0) {
1368 		lck_rw_done(&ripcbinfo.ipi_lock);
1369 		return 0;
1370 	}
1371 
1372 	inp_list = kalloc_type(struct inpcb *, n, Z_WAITOK);
1373 	if (inp_list == NULL) {
1374 		lck_rw_done(&ripcbinfo.ipi_lock);
1375 		return ENOMEM;
1376 	}
1377 
1378 	for (inp = ripcbinfo.ipi_listhead->lh_first, i = 0; inp && i < n;
1379 	    inp = inp->inp_list.le_next) {
1380 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1381 			inp_list[i++] = inp;
1382 		}
1383 	}
1384 	n = i;
1385 
1386 	error = 0;
1387 	for (i = 0; i < n; i++) {
1388 		inp = inp_list[i];
1389 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1390 			struct xinpcb64 xi;
1391 
1392 			bzero(&xi, sizeof(xi));
1393 			xi.xi_len = sizeof xi;
1394 			inpcb_to_xinpcb64(inp, &xi);
1395 			if (inp->inp_socket) {
1396 				sotoxsocket64(inp->inp_socket, &xi.xi_socket);
1397 			}
1398 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1399 		}
1400 	}
1401 	if (!error) {
1402 		/*
1403 		 * Give the user an updated idea of our state.
1404 		 * If the generation differs from what we told
1405 		 * her before, she knows that something happened
1406 		 * while we were processing this request, and it
1407 		 * might be necessary to retry.
1408 		 */
1409 		bzero(&xig, sizeof(xig));
1410 		xig.xig_len = sizeof xig;
1411 		xig.xig_gen = ripcbinfo.ipi_gencnt;
1412 		xig.xig_sogen = so_gencnt;
1413 		xig.xig_count = ripcbinfo.ipi_count;
1414 		error = SYSCTL_OUT(req, &xig, sizeof xig);
1415 	}
1416 
1417 	lck_rw_done(&ripcbinfo.ipi_lock);
1418 	kfree_type(struct inpcb *, sz, inp_list);
1419 	return error;
1420 }
1421 
1422 SYSCTL_PROC(_net_inet_raw, OID_AUTO, pcblist64,
1423     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
1424     rip_pcblist64, "S,xinpcb64", "List of active raw IP sockets");
1425 
1426 #endif /* XNU_TARGET_OS_OSX */
1427 
1428 
1429 static int
1430 rip_pcblist_n SYSCTL_HANDLER_ARGS
1431 {
1432 #pragma unused(oidp, arg1, arg2)
1433 	int error = 0;
1434 
1435 	error = get_pcblist_n(IPPROTO_IP, req, &ripcbinfo);
1436 
1437 	return error;
1438 }
1439 
1440 SYSCTL_PROC(_net_inet_raw, OID_AUTO, pcblist_n,
1441     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
1442     rip_pcblist_n, "S,xinpcb_n", "List of active raw IP sockets");
1443 
1444 struct pr_usrreqs rip_usrreqs = {
1445 	.pru_abort =            rip_abort,
1446 	.pru_attach =           rip_attach,
1447 	.pru_bind =             rip_bind,
1448 	.pru_connect =          rip_connect,
1449 	.pru_control =          in_control,
1450 	.pru_detach =           rip_detach,
1451 	.pru_disconnect =       rip_disconnect,
1452 	.pru_peeraddr =         in_getpeeraddr,
1453 	.pru_send =             rip_send,
1454 	.pru_shutdown =         rip_shutdown,
1455 	.pru_sockaddr =         in_getsockaddr,
1456 	.pru_sosend =           sosend,
1457 	.pru_soreceive =        soreceive,
1458 };
1459 /* DSEP Review Done pl-20051213-v02 @3253 */
1460