xref: /xnu-11215.81.4/bsd/netinet/raw_ip.c (revision d4514f0bc1d3f944c22d92e68b646ac3fb40d452)
1 /*
2  * Copyright (c) 2000-2023 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 	ipoa.ipoa_sotc = sotc;
517 	ipoa.ipoa_netsvctype = netsvctype;
518 
519 	if (inp->inp_flowhash == 0) {
520 		inp_calc_flowhash(inp);
521 		ASSERT(inp->inp_flowhash != 0);
522 	}
523 
524 	/*
525 	 * If the user handed us a complete IP packet, use it.
526 	 * Otherwise, allocate an mbuf for a header and fill it in.
527 	 */
528 	if ((inp_flags & INP_HDRINCL) == 0) {
529 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
530 			m_freem(m);
531 			return EMSGSIZE;
532 		}
533 		M_PREPEND(m, sizeof(struct ip), M_WAIT, 1);
534 		if (m == NULL) {
535 			return ENOBUFS;
536 		}
537 		ip = mtod(m, struct ip *);
538 		if (tos != IPTOS_UNSPEC) {
539 			ip->ip_tos = (uint8_t)(tos & IPTOS_MASK);
540 		} else {
541 			ip->ip_tos = inp->inp_ip_tos;
542 		}
543 		if (inp->inp_flags2 & INP2_DONTFRAG) {
544 			ip->ip_off = IP_DF;
545 		} else {
546 			ip->ip_off = 0;
547 		}
548 		ip->ip_p = inp->inp_ip_p;
549 		ip->ip_len = (uint16_t)m->m_pkthdr.len;
550 		ip->ip_src = inp->inp_laddr;
551 		ip->ip_dst.s_addr = dst;
552 		ip->ip_ttl = inp->inp_ip_ttl;
553 	} else {
554 		if (m->m_pkthdr.len > IP_MAXPACKET) {
555 			m_freem(m);
556 			return EMSGSIZE;
557 		}
558 		ip = mtod(m, struct ip *);
559 		/*
560 		 * don't allow both user specified and setsockopt options,
561 		 * and don't allow packet length sizes that will crash
562 		 */
563 		if (m->m_pkthdr.len < sizeof(struct ip) ||
564 		    ((IP_VHL_HL(ip->ip_vhl) != (sizeof(*ip) >> 2)) && inp->inp_options) ||
565 		    (ip->ip_len > m->m_pkthdr.len) ||
566 		    (ip->ip_len < (IP_VHL_HL(ip->ip_vhl) << 2))) {
567 			m_freem(m);
568 			return EINVAL;
569 		}
570 		if (ip->ip_id == 0 && !(rfc6864 && IP_OFF_IS_ATOMIC(ntohs(ip->ip_off)))) {
571 			ip->ip_id = ip_randomid((uint64_t)m);
572 		}
573 		/* XXX prevent ip_output from overwriting header fields */
574 		flags |= IP_RAWOUTPUT;
575 		OSAddAtomic(1, &ipstat.ips_rawout);
576 	}
577 
578 	if (inp->inp_laddr.s_addr != INADDR_ANY) {
579 		ipoa.ipoa_flags |= IPOAF_BOUND_SRCADDR;
580 	}
581 
582 #if NECP
583 	{
584 		necp_kernel_policy_id policy_id;
585 		necp_kernel_policy_id skip_policy_id;
586 		u_int32_t route_rule_id;
587 		u_int32_t pass_flags;
588 
589 		/*
590 		 * We need a route to perform NECP route rule checks
591 		 */
592 		if ((net_qos_policy_restricted != 0 &&
593 		    ROUTE_UNUSABLE(&inp->inp_route))
594 #if CONTENT_FILTER
595 		    || cfil_faddr_use
596 #endif
597 		    ) {
598 			struct sockaddr_in to;
599 			struct sockaddr_in from;
600 			struct in_addr laddr = ip->ip_src;
601 
602 			ROUTE_RELEASE(&inp->inp_route);
603 
604 			bzero(&from, sizeof(struct sockaddr_in));
605 			from.sin_family = AF_INET;
606 			from.sin_len = sizeof(struct sockaddr_in);
607 			from.sin_addr = laddr;
608 
609 			bzero(&to, sizeof(struct sockaddr_in));
610 			to.sin_family = AF_INET;
611 			to.sin_len = sizeof(struct sockaddr_in);
612 			to.sin_addr.s_addr = ip->ip_dst.s_addr;
613 
614 			if ((error = in_pcbladdr(inp, (struct sockaddr *)&to,
615 			    &laddr, ipoa.ipoa_boundif, NULL, 1)) != 0) {
616 				printf("%s in_pcbladdr(%p) error %d\n",
617 				    __func__, inp, error);
618 				m_freem(m);
619 				return error;
620 			}
621 
622 			inp_update_necp_policy(inp, (struct sockaddr *)&from,
623 			    (struct sockaddr *)&to, ipoa.ipoa_boundif);
624 			inp->inp_policyresult.results.qos_marking_gencount = 0;
625 		}
626 
627 		if (!necp_socket_is_allowed_to_send_recv_v4(inp, 0, 0,
628 		    &ip->ip_src, &ip->ip_dst, NULL, 0, &policy_id, &route_rule_id, &skip_policy_id, &pass_flags)) {
629 			m_freem(m);
630 			return EHOSTUNREACH;
631 		}
632 
633 		necp_mark_packet_from_socket(m, inp, policy_id, route_rule_id, skip_policy_id, pass_flags);
634 
635 		if (net_qos_policy_restricted != 0) {
636 			struct ifnet *rt_ifp = NULL;
637 
638 			if (inp->inp_route.ro_rt != NULL) {
639 				rt_ifp = inp->inp_route.ro_rt->rt_ifp;
640 			}
641 
642 			necp_socket_update_qos_marking(inp, inp->inp_route.ro_rt, route_rule_id);
643 		}
644 	}
645 #endif /* NECP */
646 	if ((so->so_flags1 & SOF1_QOSMARKING_ALLOWED)) {
647 		ipoa.ipoa_flags |= IPOAF_QOSMARKING_ALLOWED;
648 	}
649 #if IPSEC
650 	if (inp->inp_sp != NULL && ipsec_setsocket(m, so) != 0) {
651 		m_freem(m);
652 		return ENOBUFS;
653 	}
654 #endif /*IPSEC*/
655 
656 	if (ROUTE_UNUSABLE(&inp->inp_route)) {
657 		ROUTE_RELEASE(&inp->inp_route);
658 	}
659 
660 	set_packet_service_class(m, so, sotc, 0);
661 	m->m_pkthdr.pkt_flowsrc = FLOWSRC_INPCB;
662 	m->m_pkthdr.pkt_flowid = inp->inp_flowhash;
663 	m->m_pkthdr.pkt_flags |= (PKTF_FLOW_ID | PKTF_FLOW_LOCALSRC |
664 	    PKTF_FLOW_RAWSOCK);
665 	m->m_pkthdr.pkt_proto = inp->inp_ip_p;
666 	m->m_pkthdr.tx_rawip_pid = so->last_pid;
667 	m->m_pkthdr.tx_rawip_e_pid = so->e_pid;
668 	if (so->so_flags & SOF_DELEGATED) {
669 		m->m_pkthdr.tx_rawip_e_pid = so->e_pid;
670 	} else {
671 		m->m_pkthdr.tx_rawip_e_pid = 0;
672 	}
673 #if (DEBUG || DEVELOPMENT)
674 	if (so->so_flags & SOF_MARK_WAKE_PKT) {
675 		so->so_flags &= ~SOF_MARK_WAKE_PKT;
676 		m->m_pkthdr.pkt_flags |= PKTF_WAKE_PKT;
677 	}
678 #endif /* (DEBUG || DEVELOPMENT) */
679 
680 	imo = inp->inp_moptions;
681 	if (imo != NULL) {
682 		IMO_ADDREF(imo);
683 	}
684 	/*
685 	 * The domain lock is held across ip_output, so it is okay
686 	 * to pass the PCB cached route pointer directly to IP and
687 	 * the modules beneath it.
688 	 */
689 	// TODO: PASS DOWN ROUTE RULE ID
690 	error = ip_output(m, inp->inp_options, &inp->inp_route, flags,
691 	    imo, &ipoa);
692 
693 	if (imo != NULL) {
694 		IMO_REMREF(imo);
695 	}
696 
697 	if (inp->inp_route.ro_rt != NULL) {
698 		struct rtentry *rt = inp->inp_route.ro_rt;
699 		struct ifnet *outif;
700 
701 		if ((rt->rt_flags & (RTF_MULTICAST | RTF_BROADCAST)) ||
702 		    inp->inp_socket == NULL ||
703 #if CONTENT_FILTER
704 		    /* Discard temporary route for cfil case */
705 		    cfil_faddr_use ||
706 #endif
707 		    !(inp->inp_socket->so_state & SS_ISCONNECTED)) {
708 			rt = NULL;      /* unusable */
709 		}
710 		/*
711 		 * Always discard the cached route for unconnected
712 		 * socket or if it is a multicast route.
713 		 */
714 		if (rt == NULL) {
715 			ROUTE_RELEASE(&inp->inp_route);
716 		}
717 
718 		/*
719 		 * If this is a connected socket and the destination
720 		 * route is unicast, update outif with that of the
721 		 * route interface used by IP.
722 		 */
723 		if (rt != NULL &&
724 		    (outif = rt->rt_ifp) != inp->inp_last_outifp) {
725 			inp->inp_last_outifp = outif;
726 		}
727 	} else {
728 		ROUTE_RELEASE(&inp->inp_route);
729 	}
730 
731 	/*
732 	 * If output interface was cellular/expensive/constrained, and this socket is
733 	 * denied access to it, generate an event.
734 	 */
735 	if (error != 0 && (ipoa.ipoa_flags & IPOAF_R_IFDENIED) &&
736 	    (INP_NO_CELLULAR(inp) || INP_NO_EXPENSIVE(inp) || INP_NO_CONSTRAINED(inp))) {
737 		soevent(so, (SO_FILT_HINT_LOCKED | SO_FILT_HINT_IFDENIED));
738 	}
739 
740 	return error;
741 }
742 
743 
744 /*
745  * Raw IP socket option processing.
746  */
747 int
rip_ctloutput(struct socket * so,struct sockopt * sopt)748 rip_ctloutput(struct socket *so, struct sockopt *sopt)
749 {
750 	struct  inpcb *inp = sotoinpcb(so);
751 	int     error, optval;
752 
753 	/* Allow <SOL_SOCKET,SO_BINDTODEVICE> at this level */
754 	if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_BINDTODEVICE) {
755 		return ip_ctloutput(so, sopt);
756 	}
757 
758 	if (sopt->sopt_level != IPPROTO_IP) {
759 		return EINVAL;
760 	}
761 
762 	error = 0;
763 
764 	switch (sopt->sopt_dir) {
765 	case SOPT_GET:
766 		switch (sopt->sopt_name) {
767 		case IP_HDRINCL:
768 			optval = inp->inp_flags & INP_HDRINCL;
769 			error = sooptcopyout(sopt, &optval, sizeof optval);
770 			break;
771 
772 		case IP_STRIPHDR:
773 			optval = inp->inp_flags & INP_STRIPHDR;
774 			error = sooptcopyout(sopt, &optval, sizeof optval);
775 			break;
776 
777 
778 #if DUMMYNET
779 		case IP_DUMMYNET_GET:
780 			if (!DUMMYNET_LOADED) {
781 				ip_dn_init();
782 			}
783 			if (DUMMYNET_LOADED) {
784 				error = ip_dn_ctl_ptr(sopt);
785 			} else {
786 				error = ENOPROTOOPT;
787 			}
788 			break;
789 #endif /* DUMMYNET */
790 
791 		default:
792 			error = ip_ctloutput(so, sopt);
793 			break;
794 		}
795 		break;
796 
797 	case SOPT_SET:
798 		switch (sopt->sopt_name) {
799 		case IP_HDRINCL:
800 			error = sooptcopyin(sopt, &optval, sizeof optval,
801 			    sizeof optval);
802 			if (error) {
803 				break;
804 			}
805 			if (optval) {
806 				inp->inp_flags |= INP_HDRINCL;
807 			} else {
808 				inp->inp_flags &= ~INP_HDRINCL;
809 			}
810 			break;
811 
812 		case IP_STRIPHDR:
813 			error = sooptcopyin(sopt, &optval, sizeof optval,
814 			    sizeof optval);
815 			if (error) {
816 				break;
817 			}
818 			if (optval) {
819 				inp->inp_flags |= INP_STRIPHDR;
820 			} else {
821 				inp->inp_flags &= ~INP_STRIPHDR;
822 			}
823 			break;
824 
825 
826 #if DUMMYNET
827 		case IP_DUMMYNET_CONFIGURE:
828 		case IP_DUMMYNET_DEL:
829 		case IP_DUMMYNET_FLUSH:
830 			if (!DUMMYNET_LOADED) {
831 				ip_dn_init();
832 			}
833 			if (DUMMYNET_LOADED) {
834 				error = ip_dn_ctl_ptr(sopt);
835 			} else {
836 				error = ENOPROTOOPT;
837 			}
838 			break;
839 #endif /* DUMMYNET */
840 
841 		case SO_FLUSH:
842 			if ((error = sooptcopyin(sopt, &optval, sizeof(optval),
843 			    sizeof(optval))) != 0) {
844 				break;
845 			}
846 
847 			error = inp_flush(inp, optval);
848 			break;
849 
850 		default:
851 			error = ip_ctloutput(so, sopt);
852 			break;
853 		}
854 		break;
855 	}
856 
857 	return error;
858 }
859 
860 /*
861  * This function exists solely to receive the PRC_IFDOWN messages which
862  * are sent by if_down().  It looks for an ifaddr whose ifa_addr is sa,
863  * and calls in_ifadown() to remove all routes corresponding to that address.
864  * It also receives the PRC_IFUP messages from if_up() and reinstalls the
865  * interface routes.
866  */
867 void
rip_ctlinput(int cmd,struct sockaddr * sa,__unused void * vip,__unused struct ifnet * ifp)868 rip_ctlinput(
869 	int cmd,
870 	struct sockaddr *sa,
871 	__unused void *vip,
872 	__unused struct ifnet *ifp)
873 {
874 	struct in_ifaddr *ia = NULL;
875 	struct ifnet *iaifp = NULL;
876 	int err = 0;
877 	int flags, done = 0;
878 
879 	switch (cmd) {
880 	case PRC_IFDOWN:
881 		lck_rw_lock_shared(&in_ifaddr_rwlock);
882 		for (ia = in_ifaddrhead.tqh_first; ia;
883 		    ia = ia->ia_link.tqe_next) {
884 			IFA_LOCK(&ia->ia_ifa);
885 			if (ia->ia_ifa.ifa_addr == sa &&
886 			    (ia->ia_flags & IFA_ROUTE)) {
887 				done = 1;
888 				ifa_addref(&ia->ia_ifa);
889 				IFA_UNLOCK(&ia->ia_ifa);
890 				lck_rw_done(&in_ifaddr_rwlock);
891 				lck_mtx_lock(rnh_lock);
892 				/*
893 				 * in_ifscrub kills the interface route.
894 				 */
895 				in_ifscrub(ia->ia_ifp, ia, 1);
896 				/*
897 				 * in_ifadown gets rid of all the rest of
898 				 * the routes.  This is not quite the right
899 				 * thing to do, but at least if we are running
900 				 * a routing process they will come back.
901 				 */
902 				in_ifadown(&ia->ia_ifa, 1);
903 				lck_mtx_unlock(rnh_lock);
904 				ifa_remref(&ia->ia_ifa);
905 				break;
906 			}
907 			IFA_UNLOCK(&ia->ia_ifa);
908 		}
909 		if (!done) {
910 			lck_rw_done(&in_ifaddr_rwlock);
911 		}
912 		break;
913 
914 	case PRC_IFUP:
915 		lck_rw_lock_shared(&in_ifaddr_rwlock);
916 		for (ia = in_ifaddrhead.tqh_first; ia;
917 		    ia = ia->ia_link.tqe_next) {
918 			IFA_LOCK(&ia->ia_ifa);
919 			if (ia->ia_ifa.ifa_addr == sa) {
920 				/* keep it locked */
921 				break;
922 			}
923 			IFA_UNLOCK(&ia->ia_ifa);
924 		}
925 		if (ia == NULL || (ia->ia_flags & IFA_ROUTE) ||
926 		    (ia->ia_ifa.ifa_debug & IFD_NOTREADY)) {
927 			if (ia != NULL) {
928 				IFA_UNLOCK(&ia->ia_ifa);
929 			}
930 			lck_rw_done(&in_ifaddr_rwlock);
931 			return;
932 		}
933 		ifa_addref(&ia->ia_ifa);
934 		IFA_UNLOCK(&ia->ia_ifa);
935 		lck_rw_done(&in_ifaddr_rwlock);
936 
937 		flags = RTF_UP;
938 		iaifp = ia->ia_ifa.ifa_ifp;
939 
940 		if ((iaifp->if_flags & IFF_LOOPBACK)
941 		    || (iaifp->if_flags & IFF_POINTOPOINT)) {
942 			flags |= RTF_HOST;
943 		}
944 
945 		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
946 		if (err == 0) {
947 			IFA_LOCK_SPIN(&ia->ia_ifa);
948 			ia->ia_flags |= IFA_ROUTE;
949 			IFA_UNLOCK(&ia->ia_ifa);
950 		}
951 		ifa_remref(&ia->ia_ifa);
952 		break;
953 	}
954 }
955 
956 u_int32_t       rip_sendspace = RIPSNDQ;
957 u_int32_t       rip_recvspace = RIPRCVQ;
958 
959 SYSCTL_INT(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW | CTLFLAG_LOCKED,
960     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
961 SYSCTL_INT(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW | CTLFLAG_LOCKED,
962     &rip_recvspace, 0, "Maximum incoming raw IP datagram size");
963 SYSCTL_UINT(_net_inet_raw, OID_AUTO, pcbcount, CTLFLAG_RD | CTLFLAG_LOCKED,
964     &ripcbinfo.ipi_count, 0, "Number of active PCBs");
965 
966 static int
rip_attach(struct socket * so,int proto,struct proc * p)967 rip_attach(struct socket *so, int proto, struct proc *p)
968 {
969 	struct inpcb *inp;
970 	int error;
971 
972 	inp = sotoinpcb(so);
973 	if (inp) {
974 		panic("rip_attach");
975 	}
976 	if ((so->so_state & SS_PRIV) == 0) {
977 		return EPERM;
978 	}
979 	if (proto > UINT8_MAX) {
980 		return EINVAL;
981 	}
982 
983 	error = soreserve(so, rip_sendspace, rip_recvspace);
984 	if (error) {
985 		return error;
986 	}
987 	error = in_pcballoc(so, &ripcbinfo, p);
988 	if (error) {
989 		return error;
990 	}
991 	inp = (struct inpcb *)so->so_pcb;
992 	inp->inp_vflag |= INP_IPV4;
993 	VERIFY(proto <= UINT8_MAX);
994 	inp->inp_ip_p = (u_char)proto;
995 	inp->inp_ip_ttl = (u_char)ip_defttl;
996 	return 0;
997 }
998 
999 __private_extern__ int
rip_detach(struct socket * so)1000 rip_detach(struct socket *so)
1001 {
1002 	struct inpcb *inp;
1003 
1004 	inp = sotoinpcb(so);
1005 	if (inp == 0) {
1006 		panic("rip_detach");
1007 	}
1008 	in_pcbdetach(inp);
1009 	return 0;
1010 }
1011 
1012 __private_extern__ int
rip_abort(struct socket * so)1013 rip_abort(struct socket *so)
1014 {
1015 	soisdisconnected(so);
1016 	return rip_detach(so);
1017 }
1018 
1019 __private_extern__ int
rip_disconnect(struct socket * so)1020 rip_disconnect(struct socket *so)
1021 {
1022 	if ((so->so_state & SS_ISCONNECTED) == 0) {
1023 		return ENOTCONN;
1024 	}
1025 	return rip_abort(so);
1026 }
1027 
1028 __private_extern__ int
rip_bind(struct socket * so,struct sockaddr * nam,struct proc * p)1029 rip_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
1030 {
1031 #pragma unused(p)
1032 	struct inpcb *inp = sotoinpcb(so);
1033 	struct sockaddr_in sin;
1034 	struct ifaddr *ifa = NULL;
1035 	struct ifnet *outif = NULL;
1036 
1037 	if (inp == NULL
1038 #if NECP
1039 	    || (necp_socket_should_use_flow_divert(inp))
1040 #endif /* NECP */
1041 	    ) {
1042 		return inp == NULL ? EINVAL : EPROTOTYPE;
1043 	}
1044 
1045 	if (nam->sa_len != sizeof(struct sockaddr_in)) {
1046 		return EINVAL;
1047 	}
1048 
1049 	/* Sanitized local copy for interface address searches */
1050 	bzero(&sin, sizeof(sin));
1051 	sin.sin_family = AF_INET;
1052 	sin.sin_len = sizeof(struct sockaddr_in);
1053 	sin.sin_addr.s_addr = SIN(nam)->sin_addr.s_addr;
1054 
1055 	if (TAILQ_EMPTY(&ifnet_head) ||
1056 	    (sin.sin_family != AF_INET && sin.sin_family != AF_IMPLINK) ||
1057 	    (sin.sin_addr.s_addr && (ifa = ifa_ifwithaddr(SA(&sin))) == 0)) {
1058 		return EADDRNOTAVAIL;
1059 	} else if (ifa) {
1060 		/*
1061 		 * Opportunistically determine the outbound
1062 		 * interface that may be used; this may not
1063 		 * hold true if we end up using a route
1064 		 * going over a different interface, e.g.
1065 		 * when sending to a local address.  This
1066 		 * will get updated again after sending.
1067 		 */
1068 		IFA_LOCK(ifa);
1069 		outif = ifa->ifa_ifp;
1070 		IFA_UNLOCK(ifa);
1071 		ifa_remref(ifa);
1072 	}
1073 	inp->inp_laddr = sin.sin_addr;
1074 	inp->inp_last_outifp = outif;
1075 
1076 	return 0;
1077 }
1078 
1079 __private_extern__ int
rip_connect(struct socket * so,struct sockaddr * nam,__unused struct proc * p)1080 rip_connect(struct socket *so, struct sockaddr *nam, __unused  struct proc *p)
1081 {
1082 	struct inpcb *inp = sotoinpcb(so);
1083 	struct sockaddr_in *addr = (struct sockaddr_in *)(void *)nam;
1084 
1085 	if (inp == NULL
1086 #if NECP
1087 	    || (necp_socket_should_use_flow_divert(inp))
1088 #endif /* NECP */
1089 	    ) {
1090 		return inp == NULL ? EINVAL : EPROTOTYPE;
1091 	}
1092 	if (nam->sa_len != sizeof(*addr)) {
1093 		return EINVAL;
1094 	}
1095 	if (TAILQ_EMPTY(&ifnet_head)) {
1096 		return EADDRNOTAVAIL;
1097 	}
1098 	if ((addr->sin_family != AF_INET) &&
1099 	    (addr->sin_family != AF_IMPLINK)) {
1100 		return EAFNOSUPPORT;
1101 	}
1102 
1103 	if (!(so->so_flags1 & SOF1_CONNECT_COUNTED)) {
1104 		so->so_flags1 |= SOF1_CONNECT_COUNTED;
1105 		INC_ATOMIC_INT64_LIM(net_api_stats.nas_socket_inet_dgram_connected);
1106 	}
1107 
1108 	inp->inp_faddr = addr->sin_addr;
1109 	soisconnected(so);
1110 
1111 	return 0;
1112 }
1113 
1114 __private_extern__ int
rip_shutdown(struct socket * so)1115 rip_shutdown(struct socket *so)
1116 {
1117 	socantsendmore(so);
1118 	return 0;
1119 }
1120 
1121 __private_extern__ int
rip_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct proc * p)1122 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
1123     struct mbuf *control, struct proc *p)
1124 {
1125 #pragma unused(flags, p)
1126 	struct inpcb *inp = sotoinpcb(so);
1127 	u_int32_t dst = INADDR_ANY;
1128 	int error = 0;
1129 
1130 	if (inp == NULL
1131 #if NECP
1132 	    || (necp_socket_should_use_flow_divert(inp) && (error = EPROTOTYPE))
1133 #endif /* NECP */
1134 	    ) {
1135 		if (inp == NULL) {
1136 			error = EINVAL;
1137 		} else {
1138 			error = EPROTOTYPE;
1139 		}
1140 		goto bad;
1141 	}
1142 	so_update_tx_data_stats(so, 1, m->m_pkthdr.len);
1143 
1144 	if (nam != NULL) {
1145 		dst = ((struct sockaddr_in *)(void *)nam)->sin_addr.s_addr;
1146 	}
1147 	return rip_output(m, so, dst, control);
1148 
1149 bad:
1150 	VERIFY(error != 0);
1151 
1152 	if (m != NULL) {
1153 		m_freem(m);
1154 	}
1155 	if (control != NULL) {
1156 		m_freem(control);
1157 	}
1158 
1159 	return error;
1160 }
1161 
1162 /* note: rip_unlock is called from different protos  instead of the generic socket_unlock,
1163  * it will handle the socket dealloc on last reference
1164  * */
1165 int
rip_unlock(struct socket * so,int refcount,void * debug)1166 rip_unlock(struct socket *so, int refcount, void *debug)
1167 {
1168 	void *__single lr_saved;
1169 	struct inpcb *inp = sotoinpcb(so);
1170 
1171 	if (debug == NULL) {
1172 		lr_saved = __unsafe_forge_single(void *, __builtin_return_address(0));
1173 	} else {
1174 		lr_saved = debug;
1175 	}
1176 
1177 	if (refcount) {
1178 		if (so->so_usecount <= 0) {
1179 			panic("rip_unlock: bad refoucnt so=%p val=%x lrh= %s",
1180 			    so, so->so_usecount, solockhistory_nr(so));
1181 			/* NOTREACHED */
1182 		}
1183 		so->so_usecount--;
1184 		if (so->so_usecount == 0 && (inp->inp_wantcnt == WNT_STOPUSING)) {
1185 			/* cleanup after last reference */
1186 			lck_mtx_unlock(so->so_proto->pr_domain->dom_mtx);
1187 			lck_rw_lock_exclusive(&ripcbinfo.ipi_lock);
1188 			if (inp->inp_state != INPCB_STATE_DEAD) {
1189 				if (SOCK_CHECK_DOM(so, PF_INET6)) {
1190 					in6_pcbdetach(inp);
1191 				} else {
1192 					in_pcbdetach(inp);
1193 				}
1194 			}
1195 			in_pcbdispose(inp);
1196 			lck_rw_done(&ripcbinfo.ipi_lock);
1197 			return 0;
1198 		}
1199 	}
1200 	so->unlock_lr[so->next_unlock_lr] = lr_saved;
1201 	so->next_unlock_lr = (so->next_unlock_lr + 1) % SO_LCKDBG_MAX;
1202 	lck_mtx_unlock(so->so_proto->pr_domain->dom_mtx);
1203 	return 0;
1204 }
1205 
1206 static int
1207 rip_pcblist SYSCTL_HANDLER_ARGS
1208 {
1209 #pragma unused(oidp, arg1, arg2)
1210 	int error, i, n, sz;
1211 	struct inpcb *inp, **inp_list;
1212 	inp_gen_t gencnt;
1213 	struct xinpgen xig;
1214 
1215 	/*
1216 	 * The process of preparing the TCB list is too time-consuming and
1217 	 * resource-intensive to repeat twice on every request.
1218 	 */
1219 	lck_rw_lock_exclusive(&ripcbinfo.ipi_lock);
1220 	if (req->oldptr == USER_ADDR_NULL) {
1221 		n = ripcbinfo.ipi_count;
1222 		req->oldidx = 2 * (sizeof xig)
1223 		    + (n + n / 8) * sizeof(struct xinpcb);
1224 		lck_rw_done(&ripcbinfo.ipi_lock);
1225 		return 0;
1226 	}
1227 
1228 	if (req->newptr != USER_ADDR_NULL) {
1229 		lck_rw_done(&ripcbinfo.ipi_lock);
1230 		return EPERM;
1231 	}
1232 
1233 	/*
1234 	 * OK, now we're committed to doing something.
1235 	 */
1236 	gencnt = ripcbinfo.ipi_gencnt;
1237 	sz = n = ripcbinfo.ipi_count;
1238 
1239 	bzero(&xig, sizeof(xig));
1240 	xig.xig_len = sizeof xig;
1241 	xig.xig_count = n;
1242 	xig.xig_gen = gencnt;
1243 	xig.xig_sogen = so_gencnt;
1244 	error = SYSCTL_OUT(req, &xig, sizeof xig);
1245 	if (error) {
1246 		lck_rw_done(&ripcbinfo.ipi_lock);
1247 		return error;
1248 	}
1249 	/*
1250 	 * We are done if there is no pcb
1251 	 */
1252 	if (n == 0) {
1253 		lck_rw_done(&ripcbinfo.ipi_lock);
1254 		return 0;
1255 	}
1256 
1257 	inp_list = kalloc_type(struct inpcb *, n, Z_WAITOK);
1258 	if (inp_list == NULL) {
1259 		lck_rw_done(&ripcbinfo.ipi_lock);
1260 		return ENOMEM;
1261 	}
1262 
1263 	for (inp = ripcbinfo.ipi_listhead->lh_first, i = 0; inp && i < n;
1264 	    inp = inp->inp_list.le_next) {
1265 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1266 			inp_list[i++] = inp;
1267 		}
1268 	}
1269 	n = i;
1270 
1271 	error = 0;
1272 	for (i = 0; i < n; i++) {
1273 		inp = inp_list[i];
1274 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1275 			struct xinpcb xi;
1276 
1277 			bzero(&xi, sizeof(xi));
1278 			xi.xi_len = sizeof xi;
1279 			/* XXX should avoid extra copy */
1280 			inpcb_to_compat(inp, &xi.xi_inp);
1281 			if (inp->inp_socket) {
1282 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
1283 			}
1284 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1285 		}
1286 	}
1287 	if (!error) {
1288 		/*
1289 		 * Give the user an updated idea of our state.
1290 		 * If the generation differs from what we told
1291 		 * her before, she knows that something happened
1292 		 * while we were processing this request, and it
1293 		 * might be necessary to retry.
1294 		 */
1295 		bzero(&xig, sizeof(xig));
1296 		xig.xig_len = sizeof xig;
1297 		xig.xig_gen = ripcbinfo.ipi_gencnt;
1298 		xig.xig_sogen = so_gencnt;
1299 		xig.xig_count = ripcbinfo.ipi_count;
1300 		error = SYSCTL_OUT(req, &xig, sizeof xig);
1301 	}
1302 
1303 	lck_rw_done(&ripcbinfo.ipi_lock);
1304 	kfree_type(struct inpcb *, sz, inp_list);
1305 	return error;
1306 }
1307 
1308 SYSCTL_PROC(_net_inet_raw, OID_AUTO /*XXX*/, pcblist,
1309     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
1310     rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
1311 
1312 #if XNU_TARGET_OS_OSX
1313 
1314 static int
1315 rip_pcblist64 SYSCTL_HANDLER_ARGS
1316 {
1317 #pragma unused(oidp, arg1, arg2)
1318 	int error, i, n, sz;
1319 	struct inpcb *inp, **inp_list;
1320 	inp_gen_t gencnt;
1321 	struct xinpgen xig;
1322 
1323 	/*
1324 	 * The process of preparing the TCB list is too time-consuming and
1325 	 * resource-intensive to repeat twice on every request.
1326 	 */
1327 	lck_rw_lock_exclusive(&ripcbinfo.ipi_lock);
1328 	if (req->oldptr == USER_ADDR_NULL) {
1329 		n = ripcbinfo.ipi_count;
1330 		req->oldidx = 2 * (sizeof xig)
1331 		    + (n + n / 8) * sizeof(struct xinpcb64);
1332 		lck_rw_done(&ripcbinfo.ipi_lock);
1333 		return 0;
1334 	}
1335 
1336 	if (req->newptr != USER_ADDR_NULL) {
1337 		lck_rw_done(&ripcbinfo.ipi_lock);
1338 		return EPERM;
1339 	}
1340 
1341 	/*
1342 	 * OK, now we're committed to doing something.
1343 	 */
1344 	gencnt = ripcbinfo.ipi_gencnt;
1345 	sz = n = ripcbinfo.ipi_count;
1346 
1347 	bzero(&xig, sizeof(xig));
1348 	xig.xig_len = sizeof xig;
1349 	xig.xig_count = n;
1350 	xig.xig_gen = gencnt;
1351 	xig.xig_sogen = so_gencnt;
1352 	error = SYSCTL_OUT(req, &xig, sizeof xig);
1353 	if (error) {
1354 		lck_rw_done(&ripcbinfo.ipi_lock);
1355 		return error;
1356 	}
1357 	/*
1358 	 * We are done if there is no pcb
1359 	 */
1360 	if (n == 0) {
1361 		lck_rw_done(&ripcbinfo.ipi_lock);
1362 		return 0;
1363 	}
1364 
1365 	inp_list = kalloc_type(struct inpcb *, n, Z_WAITOK);
1366 	if (inp_list == NULL) {
1367 		lck_rw_done(&ripcbinfo.ipi_lock);
1368 		return ENOMEM;
1369 	}
1370 
1371 	for (inp = ripcbinfo.ipi_listhead->lh_first, i = 0; inp && i < n;
1372 	    inp = inp->inp_list.le_next) {
1373 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1374 			inp_list[i++] = inp;
1375 		}
1376 	}
1377 	n = i;
1378 
1379 	error = 0;
1380 	for (i = 0; i < n; i++) {
1381 		inp = inp_list[i];
1382 		if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
1383 			struct xinpcb64 xi;
1384 
1385 			bzero(&xi, sizeof(xi));
1386 			xi.xi_len = sizeof xi;
1387 			inpcb_to_xinpcb64(inp, &xi);
1388 			if (inp->inp_socket) {
1389 				sotoxsocket64(inp->inp_socket, &xi.xi_socket);
1390 			}
1391 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1392 		}
1393 	}
1394 	if (!error) {
1395 		/*
1396 		 * Give the user an updated idea of our state.
1397 		 * If the generation differs from what we told
1398 		 * her before, she knows that something happened
1399 		 * while we were processing this request, and it
1400 		 * might be necessary to retry.
1401 		 */
1402 		bzero(&xig, sizeof(xig));
1403 		xig.xig_len = sizeof xig;
1404 		xig.xig_gen = ripcbinfo.ipi_gencnt;
1405 		xig.xig_sogen = so_gencnt;
1406 		xig.xig_count = ripcbinfo.ipi_count;
1407 		error = SYSCTL_OUT(req, &xig, sizeof xig);
1408 	}
1409 
1410 	lck_rw_done(&ripcbinfo.ipi_lock);
1411 	kfree_type(struct inpcb *, sz, inp_list);
1412 	return error;
1413 }
1414 
1415 SYSCTL_PROC(_net_inet_raw, OID_AUTO, pcblist64,
1416     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
1417     rip_pcblist64, "S,xinpcb64", "List of active raw IP sockets");
1418 
1419 #endif /* XNU_TARGET_OS_OSX */
1420 
1421 
1422 static int
1423 rip_pcblist_n SYSCTL_HANDLER_ARGS
1424 {
1425 #pragma unused(oidp, arg1, arg2)
1426 	int error = 0;
1427 
1428 	error = get_pcblist_n(IPPROTO_IP, req, &ripcbinfo);
1429 
1430 	return error;
1431 }
1432 
1433 SYSCTL_PROC(_net_inet_raw, OID_AUTO, pcblist_n,
1434     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
1435     rip_pcblist_n, "S,xinpcb_n", "List of active raw IP sockets");
1436 
1437 struct pr_usrreqs rip_usrreqs = {
1438 	.pru_abort =            rip_abort,
1439 	.pru_attach =           rip_attach,
1440 	.pru_bind =             rip_bind,
1441 	.pru_connect =          rip_connect,
1442 	.pru_control =          in_control,
1443 	.pru_detach =           rip_detach,
1444 	.pru_disconnect =       rip_disconnect,
1445 	.pru_peeraddr =         in_getpeeraddr,
1446 	.pru_send =             rip_send,
1447 	.pru_shutdown =         rip_shutdown,
1448 	.pru_sockaddr =         in_getsockaddr,
1449 	.pru_sosend =           sosend,
1450 	.pru_soreceive =        soreceive,
1451 };
1452 /* DSEP Review Done pl-20051213-v02 @3253 */
1453