xref: /xnu-11417.140.69/bsd/netinet/udp_usrreq.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, 1990, 1993, 1995
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  *	@(#)udp_usrreq.c	8.6 (Berkeley) 5/23/95
61  */
62 
63 #include <sys/param.h>
64 #include <sys/systm.h>
65 #include <sys/kernel.h>
66 #include <sys/malloc.h>
67 #include <sys/mbuf.h>
68 #include <sys/domain.h>
69 #include <sys/protosw.h>
70 #include <sys/socket.h>
71 #include <sys/socketvar.h>
72 #include <sys/sysctl.h>
73 #include <sys/syslog.h>
74 #include <sys/mcache.h>
75 #include <net/ntstat.h>
76 
77 #include <kern/zalloc.h>
78 #include <mach/boolean.h>
79 #include <pexpert/pexpert.h>
80 
81 #include <net/if.h>
82 #include <net/if_types.h>
83 #include <net/route.h>
84 #include <net/dlil.h>
85 #include <net/droptap.h>
86 #include <net/net_api_stats.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_systm.h>
90 #include <netinet/in_tclass.h>
91 #include <netinet/ip.h>
92 #include <netinet/ip6.h>
93 #include <netinet/in_pcb.h>
94 #include <netinet/in_var.h>
95 #include <netinet/ip_var.h>
96 #include <netinet6/in6_pcb.h>
97 #include <netinet6/ip6_var.h>
98 #include <netinet6/udp6_var.h>
99 #include <netinet/ip_icmp.h>
100 #include <netinet/icmp_var.h>
101 #include <netinet/udp.h>
102 #include <netinet/udp_var.h>
103 #include <netinet/udp_log.h>
104 #include <sys/kdebug.h>
105 
106 #if IPSEC
107 #include <netinet6/ipsec.h>
108 #include <netinet6/esp.h>
109 #include <netkey/key.h>
110 extern int ipsec_bypass;
111 extern int esp_udp_encap_port;
112 #endif /* IPSEC */
113 
114 #if NECP
115 #include <net/necp.h>
116 #endif /* NECP */
117 
118 #if FLOW_DIVERT
119 #include <netinet/flow_divert.h>
120 #endif /* FLOW_DIVERT */
121 
122 #if CONTENT_FILTER
123 #include <net/content_filter.h>
124 #endif /* CONTENT_FILTER */
125 
126 #if SKYWALK
127 #include <skywalk/core/skywalk_var.h>
128 #endif /* SKYWALK */
129 
130 #include <net/sockaddr_utils.h>
131 
132 #define DBG_LAYER_IN_BEG        NETDBG_CODE(DBG_NETUDP, 0)
133 #define DBG_LAYER_IN_END        NETDBG_CODE(DBG_NETUDP, 2)
134 #define DBG_LAYER_OUT_BEG       NETDBG_CODE(DBG_NETUDP, 1)
135 #define DBG_LAYER_OUT_END       NETDBG_CODE(DBG_NETUDP, 3)
136 #define DBG_FNC_UDP_INPUT       NETDBG_CODE(DBG_NETUDP, (5 << 8))
137 #define DBG_FNC_UDP_OUTPUT      NETDBG_CODE(DBG_NETUDP, (6 << 8) | 1)
138 
139 /*
140  * UDP protocol implementation.
141  * Per RFC 768, August, 1980.
142  */
143 #ifndef COMPAT_42
144 static int udpcksum = 1;
145 #else
146 static int udpcksum = 0;                /* XXX */
147 #endif
148 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum,
149     CTLFLAG_RW | CTLFLAG_LOCKED, &udpcksum, 0, "");
150 
151 int udp_log_in_vain = 0;
152 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW | CTLFLAG_LOCKED,
153     &udp_log_in_vain, 0, "Log all incoming UDP packets");
154 
155 static int blackhole = 0;
156 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW | CTLFLAG_LOCKED,
157     &blackhole, 0, "Do not send port unreachables for refused connects");
158 
159 static KALLOC_TYPE_DEFINE(inpcbzone, struct inpcb, NET_KT_DEFAULT);
160 
161 struct inpcbhead udb;           /* from udp_var.h */
162 #define udb6    udb  /* for KAME src sync over BSD*'s */
163 struct inpcbinfo udbinfo;
164 
165 #ifndef UDBHASHSIZE
166 #define UDBHASHSIZE 16
167 #endif
168 
169 /* Garbage collection performed during most recent udp_gc() run */
170 static boolean_t udp_gc_done = FALSE;
171 
172 #define log_in_vain_log(a) { log a; }
173 
174 static int udp_getstat SYSCTL_HANDLER_ARGS;
175 struct  udpstat udpstat;        /* from udp_var.h */
176 SYSCTL_PROC(_net_inet_udp, UDPCTL_STATS, stats,
177     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED,
178     0, 0, udp_getstat, "S,udpstat",
179     "UDP statistics (struct udpstat, netinet/udp_var.h)");
180 
181 SYSCTL_UINT(_net_inet_udp, OID_AUTO, pcbcount,
182     CTLFLAG_RD | CTLFLAG_LOCKED, &udbinfo.ipi_count, 0,
183     "Number of active PCBs");
184 
185 __private_extern__ int udp_use_randomport = 1;
186 SYSCTL_INT(_net_inet_udp, OID_AUTO, randomize_ports,
187     CTLFLAG_RW | CTLFLAG_LOCKED, &udp_use_randomport, 0,
188     "Randomize UDP port numbers");
189 
190 struct udp_in6 {
191 	struct sockaddr_in6     uin6_sin;
192 	u_char                  uin6_init_done : 1;
193 };
194 struct udp_ip6 {
195 	struct ip6_hdr          uip6_ip6;
196 	u_char                  uip6_init_done : 1;
197 };
198 
199 int udp_abort(struct socket *);
200 int udp_attach(struct socket *, int, struct proc *);
201 int udp_bind(struct socket *, struct sockaddr *, struct proc *);
202 int udp_connect(struct socket *, struct sockaddr *, struct proc *);
203 int udp_connectx(struct socket *, struct sockaddr *,
204     struct sockaddr *, struct proc *, uint32_t, sae_associd_t,
205     sae_connid_t *, uint32_t, void *, uint32_t, struct uio *, user_ssize_t *);
206 int udp_detach(struct socket *);
207 int udp_disconnect(struct socket *);
208 int udp_disconnectx(struct socket *, sae_associd_t, sae_connid_t);
209 int udp_send(struct socket *, int, struct mbuf *, struct sockaddr *,
210     struct mbuf *, struct proc *);
211 static void udp_append(struct inpcb *, struct ip *, struct mbuf *, int,
212     struct sockaddr_in *, struct udp_in6 *, struct udp_ip6 *, struct ifnet *);
213 static int udp_input_checksum(struct mbuf *, struct udphdr *, int, int);
214 int udp_output(struct inpcb *, struct mbuf *, struct sockaddr *,
215     struct mbuf *, struct proc *);
216 static void ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip);
217 static void udp_gc(struct inpcbinfo *);
218 static int udp_defunct(struct socket *);
219 
220 struct pr_usrreqs udp_usrreqs = {
221 	.pru_abort =            udp_abort,
222 	.pru_attach =           udp_attach,
223 	.pru_bind =             udp_bind,
224 	.pru_connect =          udp_connect,
225 	.pru_connectx =         udp_connectx,
226 	.pru_control =          in_control,
227 	.pru_detach =           udp_detach,
228 	.pru_disconnect =       udp_disconnect,
229 	.pru_disconnectx =      udp_disconnectx,
230 	.pru_peeraddr =         in_getpeeraddr,
231 	.pru_send =             udp_send,
232 	.pru_shutdown =         udp_shutdown,
233 	.pru_sockaddr =         in_getsockaddr,
234 	.pru_sosend =           sosend,
235 	.pru_soreceive =        soreceive,
236 	.pru_defunct =          udp_defunct,
237 };
238 
239 void
udp_init(struct protosw * pp,struct domain * dp)240 udp_init(struct protosw *pp, struct domain *dp)
241 {
242 #pragma unused(dp)
243 	static int udp_initialized = 0;
244 	struct inpcbinfo        *pcbinfo;
245 
246 	VERIFY((pp->pr_flags & (PR_INITIALIZED | PR_ATTACHED)) == PR_ATTACHED);
247 
248 	if (udp_initialized) {
249 		return;
250 	}
251 	udp_initialized = 1;
252 	uint32_t pool_size = (nmbclusters << MCLSHIFT) >> MBSHIFT;
253 	if (pool_size >= 96) {
254 		/* Improves 10GbE UDP performance. */
255 		udp_recvspace = 786896;
256 	}
257 
258 	if (PE_parse_boot_argn("udp_log", &udp_log_enable_flags, sizeof(udp_log_enable_flags))) {
259 		os_log(OS_LOG_DEFAULT, "udp_init: set udp_log_enable_flags to 0x%x", udp_log_enable_flags);
260 	}
261 
262 	LIST_INIT(&udb);
263 	udbinfo.ipi_listhead = &udb;
264 	hashinit_counted_by(UDBHASHSIZE, udbinfo.ipi_hashbase,
265 	    udbinfo.ipi_hashbase_count);
266 	udbinfo.ipi_hashmask = udbinfo.ipi_hashbase_count - 1;
267 	hashinit_counted_by(UDBHASHSIZE, udbinfo.ipi_porthashbase,
268 	    udbinfo.ipi_porthashbase_count);
269 	udbinfo.ipi_porthashmask = udbinfo.ipi_porthashbase_count - 1;
270 	udbinfo.ipi_zone = inpcbzone;
271 
272 	pcbinfo = &udbinfo;
273 	/*
274 	 * allocate lock group and attribute for udp pcb mutexes
275 	 */
276 	pcbinfo->ipi_lock_grp = lck_grp_alloc_init("udppcb",
277 	    LCK_GRP_ATTR_NULL);
278 	lck_attr_setdefault(&pcbinfo->ipi_lock_attr);
279 	lck_rw_init(&pcbinfo->ipi_lock, pcbinfo->ipi_lock_grp,
280 	    &pcbinfo->ipi_lock_attr);
281 
282 	udbinfo.ipi_gc = udp_gc;
283 	in_pcbinfo_attach(&udbinfo);
284 }
285 
286 void
udp_input(struct mbuf * m,int iphlen)287 udp_input(struct mbuf *m, int iphlen)
288 {
289 	struct ip *ip;
290 	struct udphdr *uh;
291 	struct inpcb *inp;
292 	mbuf_ref_t opts = NULL;
293 	int len, isbroadcast;
294 	struct ip save_ip;
295 	struct sockaddr *append_sa = NULL;
296 	struct sockaddr *append_da = NULL;
297 	struct inpcbinfo *pcbinfo = &udbinfo;
298 	struct sockaddr_in udp_in;
299 	struct sockaddr_in udp_dst;
300 	struct ip_moptions *imo = NULL;
301 	int foundmembership = 0, ret = 0;
302 	struct udp_in6 udp_in6;
303 	struct udp_in6 udp_dst6;
304 	struct udp_ip6 udp_ip6;
305 	struct ifnet *ifp = m->m_pkthdr.rcvif;
306 	u_int16_t pf_tag = 0;
307 	boolean_t is_wake_pkt = false;
308 	boolean_t check_cfil = cfil_filter_present();
309 	drop_reason_t drop_reason = DROP_REASON_UNSPECIFIED;
310 
311 	SOCKADDR_ZERO(&udp_in, sizeof(udp_in));
312 	udp_in.sin_len = sizeof(struct sockaddr_in);
313 	udp_in.sin_family = AF_INET;
314 	bzero(&udp_in6, sizeof(udp_in6));
315 	udp_in6.uin6_sin.sin6_len = sizeof(struct sockaddr_in6);
316 	udp_in6.uin6_sin.sin6_family = AF_INET6;
317 
318 	if (m->m_flags & M_PKTHDR) {
319 		pf_tag = m_pftag(m)->pftag_tag;
320 		if (m->m_pkthdr.pkt_flags & PKTF_WAKE_PKT) {
321 			is_wake_pkt = true;
322 		}
323 	}
324 
325 	udpstat.udps_ipackets++;
326 
327 	KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_START, 0, 0, 0, 0, 0);
328 
329 	/* Expect 32-bit aligned data pointer on strict-align platforms */
330 	MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
331 
332 	m_add_crumb(m, PKT_CRUMB_UDP_INPUT);
333 
334 	/*
335 	 * Strip IP options, if any; should skip this,
336 	 * make available to user, and use on returned packets,
337 	 * but we don't yet have a way to check the checksum
338 	 * with options still present.
339 	 */
340 	if (iphlen > sizeof(struct ip)) {
341 		ip_stripoptions(m);
342 		iphlen = sizeof(struct ip);
343 	}
344 
345 	/*
346 	 * Get IP and UDP header together in first mbuf.
347 	 */
348 	ip = mtod(m, struct ip *);
349 	if (m->m_len < iphlen + sizeof(struct udphdr)) {
350 		m = m_pullup(m, iphlen + sizeof(struct udphdr));
351 		if (m == NULL) {
352 			udpstat.udps_hdrops++;
353 			KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END,
354 			    0, 0, 0, 0, 0);
355 			return;
356 		}
357 		ip = mtod(m, struct ip *);
358 	}
359 	uh = (struct udphdr *)(void *)((caddr_t)ip + iphlen);
360 
361 	/* destination port of 0 is illegal, based on RFC768. */
362 	if (uh->uh_dport == 0) {
363 		drop_reason = DROP_REASON_UDP_DST_PORT_ZERO;
364 		IF_UDP_STATINC(ifp, port0);
365 		goto bad;
366 	}
367 
368 	KERNEL_DEBUG(DBG_LAYER_IN_BEG, uh->uh_dport, uh->uh_sport,
369 	    ip->ip_src.s_addr, ip->ip_dst.s_addr, uh->uh_ulen);
370 
371 	/*
372 	 * Make mbuf data length reflect UDP length.
373 	 * If not enough data to reflect UDP length, drop.
374 	 */
375 	len = ntohs((u_short)uh->uh_ulen);
376 	if (ip->ip_len != len) {
377 		if (len > ip->ip_len || len < sizeof(struct udphdr)) {
378 			udpstat.udps_badlen++;
379 			IF_UDP_STATINC(ifp, badlength);
380 			drop_reason = DROP_REASON_UDP_BAD_LENGTH;
381 			goto bad;
382 		}
383 		m_adj(m, len - ip->ip_len);
384 		/* ip->ip_len = len; */
385 	}
386 	/*
387 	 * Save a copy of the IP header in case we want restore it
388 	 * for sending an ICMP error message in response.
389 	 */
390 	save_ip = *ip;
391 
392 	/*
393 	 * Checksum extended UDP header and data.
394 	 */
395 	if (udp_input_checksum(m, uh, iphlen, len)) {
396 		drop_reason = DROP_REASON_UDP_BAD_CHECKSUM;
397 		goto bad;
398 	}
399 
400 	isbroadcast = in_broadcast(ip->ip_dst, ifp);
401 
402 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || isbroadcast) {
403 		int reuse_sock = 0, mcast_delivered = 0;
404 
405 		lck_rw_lock_shared(&pcbinfo->ipi_lock);
406 		/*
407 		 * Deliver a multicast or broadcast datagram to *all* sockets
408 		 * for which the local and remote addresses and ports match
409 		 * those of the incoming datagram.  This allows more than
410 		 * one process to receive multi/broadcasts on the same port.
411 		 * (This really ought to be done for unicast datagrams as
412 		 * well, but that would cause problems with existing
413 		 * applications that open both address-specific sockets and
414 		 * a wildcard socket listening to the same port -- they would
415 		 * end up receiving duplicates of every unicast datagram.
416 		 * Those applications open the multiple sockets to overcome an
417 		 * inadequacy of the UDP socket interface, but for backwards
418 		 * compatibility we avoid the problem here rather than
419 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
420 		 */
421 
422 		/*
423 		 * Construct sockaddr format source address.
424 		 */
425 		udp_in.sin_port = uh->uh_sport;
426 		udp_in.sin_addr = ip->ip_src;
427 		/*
428 		 * Locate pcb(s) for datagram.
429 		 * (Algorithm copied from raw_intr().)
430 		 */
431 		udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
432 		LIST_FOREACH(inp, &udb, inp_list) {
433 #if IPSEC
434 			int skipit;
435 #endif /* IPSEC */
436 
437 			if (inp->inp_socket == NULL) {
438 				continue;
439 			}
440 			if (inp != sotoinpcb(inp->inp_socket)) {
441 				panic("%s: bad so back ptr inp=%p",
442 				    __func__, inp);
443 				/* NOTREACHED */
444 			}
445 			if ((inp->inp_vflag & INP_IPV4) == 0) {
446 				continue;
447 			}
448 			if (inp_restricted_recv(inp, ifp)) {
449 				continue;
450 			}
451 
452 			if ((inp->inp_moptions == NULL) &&
453 			    (ntohl(ip->ip_dst.s_addr) !=
454 			    INADDR_ALLHOSTS_GROUP) && (isbroadcast == 0)) {
455 				continue;
456 			}
457 			/*
458 			 * Skip unbound sockets before taking the lock on the socket as
459 			 * the test with the destination port in the header will fail
460 			 */
461 			if (inp->inp_lport == 0) {
462 				continue;
463 			}
464 
465 			if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) ==
466 			    WNT_STOPUSING) {
467 				continue;
468 			}
469 
470 			udp_lock(inp->inp_socket, 1, 0);
471 
472 			if (in_pcb_checkstate(inp, WNT_RELEASE, 1) ==
473 			    WNT_STOPUSING) {
474 				udp_unlock(inp->inp_socket, 1, 0);
475 				continue;
476 			}
477 
478 			if (inp->inp_lport != uh->uh_dport) {
479 				udp_unlock(inp->inp_socket, 1, 0);
480 				continue;
481 			}
482 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
483 				if (inp->inp_laddr.s_addr !=
484 				    ip->ip_dst.s_addr) {
485 					udp_unlock(inp->inp_socket, 1, 0);
486 					continue;
487 				}
488 			}
489 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
490 				if (inp->inp_faddr.s_addr !=
491 				    ip->ip_src.s_addr ||
492 				    inp->inp_fport != uh->uh_sport) {
493 					udp_unlock(inp->inp_socket, 1, 0);
494 					continue;
495 				}
496 			}
497 
498 			if (isbroadcast == 0 && (ntohl(ip->ip_dst.s_addr) !=
499 			    INADDR_ALLHOSTS_GROUP)) {
500 				struct sockaddr_in group;
501 				int blocked;
502 
503 				if ((imo = inp->inp_moptions) == NULL) {
504 					udp_unlock(inp->inp_socket, 1, 0);
505 					continue;
506 				}
507 				IMO_LOCK(imo);
508 
509 				SOCKADDR_ZERO(&group, sizeof(struct sockaddr_in));
510 				group.sin_len = sizeof(struct sockaddr_in);
511 				group.sin_family = AF_INET;
512 				group.sin_addr = ip->ip_dst;
513 
514 				blocked = imo_multi_filter(imo, ifp,
515 				    &group, &udp_in);
516 				if (blocked == MCAST_PASS) {
517 					foundmembership = 1;
518 				}
519 
520 				IMO_UNLOCK(imo);
521 				if (!foundmembership) {
522 					udp_unlock(inp->inp_socket, 1, 0);
523 					if (blocked == MCAST_NOTSMEMBER ||
524 					    blocked == MCAST_MUTED) {
525 						udpstat.udps_filtermcast++;
526 					}
527 					continue;
528 				}
529 				foundmembership = 0;
530 			}
531 
532 			reuse_sock = (inp->inp_socket->so_options &
533 			    (SO_REUSEPORT | SO_REUSEADDR));
534 
535 #if NECP
536 			skipit = 0;
537 			if (!necp_socket_is_allowed_to_send_recv_v4(inp,
538 			    uh->uh_dport, uh->uh_sport, &ip->ip_dst,
539 			    &ip->ip_src, ifp, pf_tag, NULL, NULL, NULL, NULL)) {
540 				/* do not inject data to pcb */
541 				skipit = 1;
542 				UDP_LOG_DROP_NECP(ip, uh, inp, false);
543 			}
544 			if (skipit == 0)
545 #endif /* NECP */
546 			{
547 				mbuf_ref_t n = NULL;
548 
549 				if (reuse_sock) {
550 					n = m_copy(m, 0, M_COPYALL);
551 				}
552 				udp_append(inp, ip, m,
553 				    iphlen + sizeof(struct udphdr),
554 				    &udp_in, &udp_in6, &udp_ip6, ifp);
555 				mcast_delivered++;
556 
557 				m = n;
558 			}
559 			if (is_wake_pkt) {
560 				soevent(inp->inp_socket, SO_FILT_HINT_LOCKED | SO_FILT_HINT_WAKE_PKT);
561 			}
562 
563 			udp_unlock(inp->inp_socket, 1, 0);
564 
565 
566 			/*
567 			 * Don't look for additional matches if this one does
568 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
569 			 * socket options set.  This heuristic avoids searching
570 			 * through all pcbs in the common case of a non-shared
571 			 * port.  It assumes that an application will never
572 			 * clear these options after setting them.
573 			 */
574 			if (reuse_sock == 0 || m == NULL) {
575 				break;
576 			}
577 
578 			/*
579 			 * Expect 32-bit aligned data pointer on strict-align
580 			 * platforms.
581 			 */
582 			MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
583 			/*
584 			 * Recompute IP and UDP header pointers for new mbuf
585 			 */
586 			ip = mtod(m, struct ip *);
587 			uh = (struct udphdr *)(void *)((caddr_t)ip + iphlen);
588 		}
589 		lck_rw_done(&pcbinfo->ipi_lock);
590 
591 		if (mcast_delivered == 0) {
592 			/*
593 			 * No matching pcb found; discard datagram.
594 			 * (No need to send an ICMP Port Unreachable
595 			 * for a broadcast or multicast datgram.)
596 			 */
597 			udpstat.udps_noportbcast++;
598 			IF_UDP_STATINC(ifp, port_unreach);
599 			drop_reason = DROP_REASON_UDP_PORT_UNREACHEABLE;
600 			goto bad;
601 		}
602 
603 		/* free the extra copy of mbuf or skipped by IPsec */
604 		if (m != NULL) {
605 			m_freem(m);
606 		}
607 		KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0);
608 		return;
609 	}
610 
611 #if IPSEC
612 	/*
613 	 * UDP to port 4500 with a payload where the first four bytes are
614 	 * not zero is a UDP encapsulated IPsec packet. Packets where
615 	 * the payload is one byte and that byte is 0xFF are NAT keepalive
616 	 * packets. Decapsulate the ESP packet and carry on with IPsec input
617 	 * or discard the NAT keep-alive.
618 	 */
619 	if (ipsec_bypass == 0 && (esp_udp_encap_port & 0xFFFF) != 0 &&
620 	    (uh->uh_dport == ntohs((u_short)esp_udp_encap_port) ||
621 	    uh->uh_sport == ntohs((u_short)esp_udp_encap_port))) {
622 		/*
623 		 * Check if ESP or keepalive:
624 		 *      1. If the destination port of the incoming packet is 4500.
625 		 *      2. If the source port of the incoming packet is 4500,
626 		 *         then check the SADB to match IP address and port.
627 		 */
628 		bool check_esp = true;
629 		if (uh->uh_dport != ntohs((u_short)esp_udp_encap_port)) {
630 			union sockaddr_in_4_6 src = {};
631 			union sockaddr_in_4_6 dst = {};
632 
633 			ipsec_fill_ip_sockaddr_4_6(&src, ip->ip_src, uh->uh_sport);
634 			ipsec_fill_ip_sockaddr_4_6(&dst, ip->ip_dst, uh->uh_dport);
635 
636 			check_esp = key_checksa_present(&dst, &src);
637 		}
638 
639 		if (check_esp) {
640 			int payload_len = len - sizeof(struct udphdr) > 4 ? 4 :
641 			    len - sizeof(struct udphdr);
642 
643 			if (m->m_len < iphlen + sizeof(struct udphdr) + payload_len) {
644 				if ((m = m_pullup(m, iphlen + sizeof(struct udphdr) +
645 				    payload_len)) == NULL) {
646 					udpstat.udps_hdrops++;
647 					KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END,
648 					    0, 0, 0, 0, 0);
649 					return;
650 				}
651 				/*
652 				 * Expect 32-bit aligned data pointer on strict-align
653 				 * platforms.
654 				 */
655 				MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
656 
657 				ip = mtod(m, struct ip *);
658 				uh = (struct udphdr *)(void *)((caddr_t)ip + iphlen);
659 			}
660 			/* Check for NAT keepalive packet */
661 			if (payload_len == 1 && *(u_int8_t *)
662 			    ((caddr_t)uh + sizeof(struct udphdr)) == 0xFF) {
663 				m_freem(m);
664 				KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END,
665 				    0, 0, 0, 0, 0);
666 				return;
667 			} else if (payload_len == 4 && *(u_int32_t *)(void *)
668 			    ((caddr_t)uh + sizeof(struct udphdr)) != 0) {
669 				/* UDP encapsulated IPsec packet to pass through NAT */
670 				KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END,
671 				    0, 0, 0, 0, 0);
672 				/* preserve the udp header */
673 				esp4_input(m, iphlen + sizeof(struct udphdr));
674 				return;
675 			}
676 		}
677 	}
678 #endif /* IPSEC */
679 
680 	/*
681 	 * Locate pcb for datagram.
682 	 */
683 	inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
684 	    ip->ip_dst, uh->uh_dport, 1, ifp);
685 	if (inp == NULL) {
686 		IF_UDP_STATINC(ifp, port_unreach);
687 
688 		if (udp_log_in_vain) {
689 			char buf[MAX_IPv4_STR_LEN];
690 			char buf2[MAX_IPv4_STR_LEN];
691 
692 			/* check src and dst address */
693 			if (udp_log_in_vain < 3) {
694 				log(LOG_INFO, "Connection attempt to "
695 				    "UDP %s:%d from %s:%d\n", inet_ntop(AF_INET,
696 				    &ip->ip_dst, buf, sizeof(buf)),
697 				    ntohs(uh->uh_dport), inet_ntop(AF_INET,
698 				    &ip->ip_src, buf2, sizeof(buf2)),
699 				    ntohs(uh->uh_sport));
700 			} else if (!(m->m_flags & (M_BCAST | M_MCAST)) &&
701 			    ip->ip_dst.s_addr != ip->ip_src.s_addr) {
702 				log_in_vain_log((LOG_INFO,
703 				    "Stealth Mode connection attempt to "
704 				    "UDP %s:%d from %s:%d\n", inet_ntop(AF_INET,
705 				    &ip->ip_dst, buf, sizeof(buf)),
706 				    ntohs(uh->uh_dport), inet_ntop(AF_INET,
707 				    &ip->ip_src, buf2, sizeof(buf2)),
708 				    ntohs(uh->uh_sport)))
709 			}
710 		}
711 		udpstat.udps_noport++;
712 		if (m->m_flags & (M_BCAST | M_MCAST)) {
713 			udpstat.udps_noportbcast++;
714 			drop_reason = DROP_REASON_UDP_PORT_UNREACHEABLE;
715 			goto bad;
716 		}
717 		if (blackhole) {
718 			if (ifp && ifp->if_type != IFT_LOOP) {
719 				drop_reason = DROP_REASON_UDP_PORT_UNREACHEABLE;
720 				goto bad;
721 			}
722 		}
723 		if (if_link_heuristics_enabled(ifp)) {
724 			drop_reason = DROP_REASON_UDP_PORT_UNREACHEABLE;
725 			IF_UDP_STATINC(ifp, linkheur_stealthdrop);
726 			goto bad;
727 		}
728 		*ip = save_ip;
729 		ip->ip_len += iphlen;
730 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
731 		KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0);
732 		return;
733 	}
734 	udp_lock(inp->inp_socket, 1, 0);
735 
736 	if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
737 		udp_unlock(inp->inp_socket, 1, 0);
738 		IF_UDP_STATINC(ifp, cleanup);
739 		drop_reason = DROP_REASON_UDP_SOCKET_CLOSING;
740 		goto bad;
741 	}
742 #if NECP
743 	if (!necp_socket_is_allowed_to_send_recv_v4(inp, uh->uh_dport,
744 	    uh->uh_sport, &ip->ip_dst, &ip->ip_src, ifp, pf_tag, NULL, NULL, NULL, NULL)) {
745 		udp_unlock(inp->inp_socket, 1, 0);
746 		IF_UDP_STATINC(ifp, badipsec);
747 		drop_reason = DROP_REASON_UDP_NECP;
748 		goto bad;
749 	}
750 #endif /* NECP */
751 
752 	/*
753 	 * Construct sockaddr format source address.
754 	 * Stuff source address and datagram in user buffer.
755 	 */
756 	udp_in.sin_port = uh->uh_sport;
757 	udp_in.sin_addr = ip->ip_src;
758 	if ((inp->inp_flags & INP_CONTROLOPTS) != 0 ||
759 	    SOFLOW_ENABLED(inp->inp_socket) ||
760 	    SO_RECV_CONTROL_OPTS(inp->inp_socket)) {
761 		if (inp->inp_vflag & INP_IPV6 || inp->inp_vflag & INP_V4MAPPEDV6) {
762 			int savedflags;
763 
764 			ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
765 			savedflags = inp->inp_flags;
766 			inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
767 			ret = ip6_savecontrol(inp, m, &opts);
768 			inp->inp_flags = savedflags;
769 		} else {
770 			ret = ip_savecontrol(inp, &opts, ip, m);
771 		}
772 		if (ret != 0) {
773 			udp_unlock(inp->inp_socket, 1, 0);
774 			drop_reason = DROP_REASON_UDP_CANNOT_SAVE_CONTROL;
775 			goto bad;
776 		}
777 	}
778 	m_adj(m, iphlen + sizeof(struct udphdr));
779 
780 	KERNEL_DEBUG(DBG_LAYER_IN_END, uh->uh_dport, uh->uh_sport,
781 	    save_ip.ip_src.s_addr, save_ip.ip_dst.s_addr, uh->uh_ulen);
782 
783 	if (inp->inp_vflag & INP_IPV6) {
784 		in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
785 		append_sa = SA(&udp_in6.uin6_sin);
786 	} else {
787 		append_sa = SA(&udp_in);
788 	}
789 	if (nstat_collect) {
790 		stats_functional_type ifnet_count_type = IFNET_COUNT_TYPE(ifp);
791 		INP_ADD_STAT(inp, ifnet_count_type, rxpackets, 1);
792 		INP_ADD_STAT(inp, ifnet_count_type, rxbytes, m->m_pkthdr.len);
793 		inp_set_activity_bitmap(inp);
794 	}
795 #if CONTENT_FILTER && NECP
796 	if (check_cfil && inp != NULL && inp->inp_policyresult.results.filter_control_unit == 0) {
797 		if (inp->inp_vflag & INP_IPV6) {
798 			bzero(&udp_dst6, sizeof(udp_dst6));
799 			udp_dst6.uin6_sin.sin6_len = sizeof(struct sockaddr_in6);
800 			udp_dst6.uin6_sin.sin6_family = AF_INET6;
801 			in6_sin_2_v4mapsin6(&udp_dst, &udp_dst6.uin6_sin);
802 			append_da = SA(&udp_dst6.uin6_sin);
803 		} else {
804 			SOCKADDR_ZERO(&udp_dst, sizeof(udp_dst));
805 			udp_dst.sin_len = sizeof(struct sockaddr_in);
806 			udp_dst.sin_family = AF_INET;
807 			udp_dst.sin_port = uh->uh_dport;
808 			udp_dst.sin_addr = ip->ip_dst;
809 			append_da = SA(&udp_dst);
810 		}
811 		// Override the dst input here so NECP can pick up the policy
812 		// and CFIL can find an existing control socket.
813 		necp_socket_find_policy_match(inp, append_da, append_sa, 0);
814 	}
815 #endif /* CONTENT_FILTER and NECP */
816 	so_recv_data_stat(inp->inp_socket, m, 0);
817 	if (sbappendaddr(&inp->inp_socket->so_rcv, append_sa,
818 	    m, opts, NULL) == 0) {
819 		udpstat.udps_fullsock++;
820 	} else {
821 		sorwakeup(inp->inp_socket);
822 	}
823 	if (is_wake_pkt) {
824 		soevent(inp->inp_socket, SO_FILT_HINT_LOCKED | SO_FILT_HINT_WAKE_PKT);
825 	}
826 	udp_unlock(inp->inp_socket, 1, 0);
827 	KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0);
828 	return;
829 bad:
830 	m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, drop_reason, NULL, 0);
831 	if (opts) {
832 		m_freem(opts);
833 	}
834 	KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0);
835 }
836 
837 static void
ip_2_ip6_hdr(struct ip6_hdr * ip6,struct ip * ip)838 ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip)
839 {
840 	bzero(ip6, sizeof(*ip6));
841 
842 	ip6->ip6_vfc = IPV6_VERSION;
843 	ip6->ip6_plen = ip->ip_len;
844 	ip6->ip6_nxt = ip->ip_p;
845 	ip6->ip6_hlim = ip->ip_ttl;
846 	if (ip->ip_src.s_addr) {
847 		ip6->ip6_src.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
848 		ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
849 	}
850 	if (ip->ip_dst.s_addr) {
851 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
852 		ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
853 	}
854 }
855 
856 /*
857  * subroutine of udp_input(), mainly for source code readability.
858  */
859 static void
udp_append(struct inpcb * last,struct ip * ip,struct mbuf * n,int off,struct sockaddr_in * pudp_in,struct udp_in6 * pudp_in6,struct udp_ip6 * pudp_ip6,struct ifnet * ifp)860 udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off,
861     struct sockaddr_in *pudp_in, struct udp_in6 *pudp_in6,
862     struct udp_ip6 *pudp_ip6, struct ifnet *ifp)
863 {
864 	struct sockaddr *append_sa;
865 	mbuf_ref_t opts = NULL;
866 	int ret = 0;
867 
868 	if ((last->inp_flags & INP_CONTROLOPTS) != 0 ||
869 	    SOFLOW_ENABLED(last->inp_socket) ||
870 	    SO_RECV_CONTROL_OPTS(last->inp_socket)) {
871 		if (last->inp_vflag & INP_IPV6 || last->inp_vflag & INP_V4MAPPEDV6) {
872 			int savedflags;
873 
874 			if (pudp_ip6->uip6_init_done == 0) {
875 				ip_2_ip6_hdr(&pudp_ip6->uip6_ip6, ip);
876 				pudp_ip6->uip6_init_done = 1;
877 			}
878 			savedflags = last->inp_flags;
879 			last->inp_flags &= ~INP_UNMAPPABLEOPTS;
880 			ret = ip6_savecontrol(last, n, &opts);
881 			if (ret != 0) {
882 				last->inp_flags = savedflags;
883 				UDP_LOG(last, "ip6_savecontrol error %d", ret);
884 				goto error;
885 			}
886 			last->inp_flags = savedflags;
887 		} else {
888 			ret = ip_savecontrol(last, &opts, ip, n);
889 			if (ret != 0) {
890 				UDP_LOG(last, "ip_savecontrol error %d", ret);
891 				goto error;
892 			}
893 		}
894 	}
895 	if (last->inp_vflag & INP_IPV6) {
896 		if (pudp_in6->uin6_init_done == 0) {
897 			in6_sin_2_v4mapsin6(pudp_in, &pudp_in6->uin6_sin);
898 			pudp_in6->uin6_init_done = 1;
899 		}
900 		append_sa = SA(&pudp_in6->uin6_sin);
901 	} else {
902 		append_sa = SA(pudp_in);
903 	}
904 	if (nstat_collect) {
905 		stats_functional_type ifnet_count_type = IFNET_COUNT_TYPE(ifp);
906 		INP_ADD_STAT(last, ifnet_count_type, rxpackets, 1);
907 		INP_ADD_STAT(last, ifnet_count_type, rxbytes,
908 		    n->m_pkthdr.len);
909 		inp_set_activity_bitmap(last);
910 	}
911 	so_recv_data_stat(last->inp_socket, n, 0);
912 	m_adj(n, off);
913 	if (sbappendaddr(&last->inp_socket->so_rcv, append_sa,
914 	    n, opts, NULL) == 0) {
915 		udpstat.udps_fullsock++;
916 		UDP_LOG(last, "sbappendaddr full receive socket buffer");
917 	} else {
918 		sorwakeup(last->inp_socket);
919 	}
920 	return;
921 error:
922 	m_freem(n);
923 	m_freem(opts);
924 }
925 
926 /*
927  * Notify a udp user of an asynchronous error;
928  * just wake up so that he can collect error status.
929  */
930 void
udp_notify(struct inpcb * inp,int errno)931 udp_notify(struct inpcb *inp, int errno)
932 {
933 	inp->inp_socket->so_error = (u_short)errno;
934 	sorwakeup(inp->inp_socket);
935 	sowwakeup(inp->inp_socket);
936 }
937 
938 void
udp_ctlinput(int cmd,struct sockaddr * sa,void * vip,__unused struct ifnet * ifp)939 udp_ctlinput(int cmd, struct sockaddr *sa, void *vip, __unused struct ifnet * ifp)
940 {
941 	struct ipctlparam *__single ctl_param = vip;
942 	struct ip *ip = NULL;
943 	mbuf_ref_t m = NULL;
944 	void (*notify)(struct inpcb *, int) = udp_notify;
945 	struct in_addr faddr;
946 	struct inpcb *inp = NULL;
947 	struct icmp *icp = NULL;
948 	size_t off;
949 
950 	if (ctl_param != NULL) {
951 		ip = ctl_param->ipc_icmp_ip;
952 		icp = ctl_param->ipc_icmp;
953 		m = ctl_param->ipc_m;
954 		off = ctl_param->ipc_off;
955 	} else {
956 		ip = NULL;
957 		icp = NULL;
958 		m = NULL;
959 		off = 0;
960 	}
961 
962 	faddr = SIN(sa)->sin_addr;
963 	if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) {
964 		return;
965 	}
966 
967 	if (PRC_IS_REDIRECT(cmd)) {
968 		ip = 0;
969 		notify = in_rtchange;
970 	} else if (cmd == PRC_HOSTDEAD) {
971 		ip = 0;
972 	} else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) {
973 		return;
974 	}
975 	if (ip) {
976 		struct udphdr uh;
977 
978 		/* Check if we can safely get the ports from the UDP header */
979 		if (m == NULL ||
980 		    (m->m_len < off + sizeof(uh))) {
981 			/* Insufficient length */
982 			return;
983 		}
984 
985 		bcopy(m_mtod_current(m) + off, &uh, sizeof(uh));
986 		inp = in_pcblookup_hash(&udbinfo, faddr, uh.uh_dport,
987 		    ip->ip_src, uh.uh_sport, 0, NULL);
988 
989 		if (inp != NULL && inp->inp_socket != NULL) {
990 			udp_lock(inp->inp_socket, 1, 0);
991 			if (in_pcb_checkstate(inp, WNT_RELEASE, 1) ==
992 			    WNT_STOPUSING) {
993 				udp_unlock(inp->inp_socket, 1, 0);
994 				return;
995 			}
996 			if (cmd == PRC_MSGSIZE && !uuid_is_null(inp->necp_client_uuid)) {
997 				uuid_t null_uuid;
998 				uuid_clear(null_uuid);
999 				necp_update_flow_protoctl_event(null_uuid, inp->necp_client_uuid,
1000 				    PRC_MSGSIZE, ntohs(icp->icmp_nextmtu), 0);
1001 				/*
1002 				 * Avoid calling udp_notify() to set so_error
1003 				 * when using Network.framework since the notification
1004 				 * of PRC_MSGSIZE has been delivered through NECP.
1005 				 */
1006 			} else {
1007 				(*notify)(inp, inetctlerrmap[cmd]);
1008 			}
1009 			udp_unlock(inp->inp_socket, 1, 0);
1010 		}
1011 #if SKYWALK
1012 		else {
1013 			union sockaddr_in_4_6 sock_laddr;
1014 			struct protoctl_ev_val prctl_ev_val;
1015 			bzero(&prctl_ev_val, sizeof(prctl_ev_val));
1016 			bzero(&sock_laddr, sizeof(sock_laddr));
1017 
1018 			if (cmd == PRC_MSGSIZE) {
1019 				prctl_ev_val.val = ntohs(icp->icmp_nextmtu);
1020 			}
1021 
1022 			sock_laddr.sin.sin_family = AF_INET;
1023 			sock_laddr.sin.sin_len = sizeof(sock_laddr.sin);
1024 			sock_laddr.sin.sin_addr = ip->ip_src;
1025 
1026 			protoctl_event_enqueue_nwk_wq_entry(ifp,
1027 			    SA(&sock_laddr), sa,
1028 			    uh.uh_sport, uh.uh_dport, IPPROTO_UDP,
1029 			    cmd, &prctl_ev_val);
1030 		}
1031 #endif /* SKYWALK */
1032 	} else {
1033 		in_pcbnotifyall(&udbinfo, faddr, inetctlerrmap[cmd], notify);
1034 	}
1035 }
1036 
1037 int
udp_ctloutput(struct socket * so,struct sockopt * sopt)1038 udp_ctloutput(struct socket *so, struct sockopt *sopt)
1039 {
1040 	int     error = 0, optval = 0;
1041 	struct  inpcb *inp;
1042 
1043 	/* Allow <SOL_SOCKET,SO_FLUSH/SO_BINDTODEVICE> at this level */
1044 	if (sopt->sopt_level == SOL_SOCKET) {
1045 		if (sopt->sopt_name == SO_BINDTODEVICE) {
1046 			if (SOCK_CHECK_DOM(so, PF_INET6)) {
1047 				error = ip6_ctloutput(so, sopt);
1048 			} else {
1049 				error = ip_ctloutput(so, sopt);
1050 			}
1051 			return error;
1052 		} else if (sopt->sopt_name != SO_FLUSH) {
1053 			return EINVAL;
1054 		}
1055 	}
1056 	if (sopt->sopt_level != IPPROTO_UDP) {
1057 		if (SOCK_CHECK_DOM(so, PF_INET6)) {
1058 			error = ip6_ctloutput(so, sopt);
1059 		} else {
1060 			error = ip_ctloutput(so, sopt);
1061 		}
1062 		return error;
1063 	}
1064 
1065 	inp = sotoinpcb(so);
1066 
1067 	switch (sopt->sopt_dir) {
1068 	case SOPT_SET:
1069 		switch (sopt->sopt_name) {
1070 		case UDP_NOCKSUM:
1071 			/* This option is settable only for UDP over IPv4 */
1072 			if (!(inp->inp_vflag & INP_IPV4)) {
1073 				error = EINVAL;
1074 				break;
1075 			}
1076 
1077 			if ((error = sooptcopyin(sopt, &optval, sizeof(optval),
1078 			    sizeof(optval))) != 0) {
1079 				break;
1080 			}
1081 
1082 			if (optval != 0) {
1083 				inp->inp_flags |= INP_UDP_NOCKSUM;
1084 			} else {
1085 				inp->inp_flags &= ~INP_UDP_NOCKSUM;
1086 			}
1087 			break;
1088 		case UDP_KEEPALIVE_OFFLOAD:
1089 		{
1090 			struct udp_keepalive_offload ka;
1091 			/*
1092 			 * If the socket is not connected, the stack will
1093 			 * not know the destination address to put in the
1094 			 * keepalive datagram. Return an error now instead
1095 			 * of failing later.
1096 			 */
1097 			if (!(so->so_state & SS_ISCONNECTED)) {
1098 				error = EINVAL;
1099 				break;
1100 			}
1101 			if (sopt->sopt_valsize != sizeof(ka)) {
1102 				error = EINVAL;
1103 				break;
1104 			}
1105 			if ((error = sooptcopyin(sopt, &ka, sizeof(ka),
1106 			    sizeof(ka))) != 0) {
1107 				break;
1108 			}
1109 
1110 			/* application should specify the type */
1111 			if (ka.ka_type == 0) {
1112 				return EINVAL;
1113 			}
1114 
1115 			if (ka.ka_interval == 0) {
1116 				/*
1117 				 * if interval is 0, disable the offload
1118 				 * mechanism
1119 				 */
1120 				if (inp->inp_keepalive_data != NULL) {
1121 					kfree_data_sized_by(inp->inp_keepalive_data,
1122 					    inp->inp_keepalive_datalen);
1123 				}
1124 				inp->inp_keepalive_data = NULL;
1125 				inp->inp_keepalive_datalen = 0;
1126 				inp->inp_keepalive_interval = 0;
1127 				inp->inp_keepalive_type = 0;
1128 				inp->inp_flags2 &= ~INP2_KEEPALIVE_OFFLOAD;
1129 			} else {
1130 				if (inp->inp_keepalive_data != NULL) {
1131 					kfree_data_sized_by(inp->inp_keepalive_data,
1132 					    inp->inp_keepalive_datalen);
1133 				}
1134 
1135 				uint8_t datalen = (uint8_t)min(
1136 					ka.ka_data_len,
1137 					UDP_KEEPALIVE_OFFLOAD_DATA_SIZE);
1138 				if (datalen > 0) {
1139 					uint8_t *data = kalloc_data(datalen, Z_WAITOK);
1140 					if (data == NULL) {
1141 						inp->inp_keepalive_data = NULL;
1142 						inp->inp_keepalive_datalen = 0;
1143 						error = ENOMEM;
1144 						break;
1145 					} else {
1146 						inp->inp_keepalive_data = data;
1147 						inp->inp_keepalive_datalen = datalen;
1148 					}
1149 					bcopy(ka.ka_data,
1150 					    inp->inp_keepalive_data,
1151 					    inp->inp_keepalive_datalen);
1152 				} else {
1153 					inp->inp_keepalive_datalen = 0;
1154 					inp->inp_keepalive_data = NULL;
1155 				}
1156 				inp->inp_keepalive_interval = (uint8_t)
1157 				    min(UDP_KEEPALIVE_INTERVAL_MAX_SECONDS,
1158 				    ka.ka_interval);
1159 				inp->inp_keepalive_type = ka.ka_type;
1160 				inp->inp_flags2 |= INP2_KEEPALIVE_OFFLOAD;
1161 			}
1162 			break;
1163 		}
1164 		case SO_FLUSH:
1165 			if ((error = sooptcopyin(sopt, &optval, sizeof(optval),
1166 			    sizeof(optval))) != 0) {
1167 				break;
1168 			}
1169 
1170 			error = inp_flush(inp, optval);
1171 			break;
1172 
1173 		default:
1174 			error = ENOPROTOOPT;
1175 			break;
1176 		}
1177 		break;
1178 
1179 	case SOPT_GET:
1180 		switch (sopt->sopt_name) {
1181 		case UDP_NOCKSUM:
1182 			optval = inp->inp_flags & INP_UDP_NOCKSUM;
1183 			break;
1184 
1185 		default:
1186 			error = ENOPROTOOPT;
1187 			break;
1188 		}
1189 		if (error == 0) {
1190 			error = sooptcopyout(sopt, &optval, sizeof(optval));
1191 		}
1192 		break;
1193 	}
1194 	return error;
1195 }
1196 
1197 static int
1198 udp_pcblist SYSCTL_HANDLER_ARGS
1199 {
1200 #pragma unused(oidp, arg1, arg2)
1201 	int error, i, n, sz;
1202 	struct inpcb *inp, **inp_list;
1203 	inp_gen_t gencnt;
1204 	struct xinpgen xig;
1205 
1206 	/*
1207 	 * The process of preparing the TCB list is too time-consuming and
1208 	 * resource-intensive to repeat twice on every request.
1209 	 */
1210 	lck_rw_lock_exclusive(&udbinfo.ipi_lock);
1211 	if (req->oldptr == USER_ADDR_NULL) {
1212 		n = udbinfo.ipi_count;
1213 		req->oldidx = 2 * (sizeof(xig))
1214 		    + (n + n / 8) * sizeof(struct xinpcb);
1215 		lck_rw_done(&udbinfo.ipi_lock);
1216 		return 0;
1217 	}
1218 
1219 	if (req->newptr != USER_ADDR_NULL) {
1220 		lck_rw_done(&udbinfo.ipi_lock);
1221 		return EPERM;
1222 	}
1223 
1224 	/*
1225 	 * OK, now we're committed to doing something.
1226 	 */
1227 	gencnt = udbinfo.ipi_gencnt;
1228 	sz = n = udbinfo.ipi_count;
1229 
1230 	bzero(&xig, sizeof(xig));
1231 	xig.xig_len = sizeof(xig);
1232 	xig.xig_count = n;
1233 	xig.xig_gen = gencnt;
1234 	xig.xig_sogen = so_gencnt;
1235 	error = SYSCTL_OUT(req, &xig, sizeof(xig));
1236 	if (error) {
1237 		lck_rw_done(&udbinfo.ipi_lock);
1238 		return error;
1239 	}
1240 	/*
1241 	 * We are done if there is no pcb
1242 	 */
1243 	if (n == 0) {
1244 		lck_rw_done(&udbinfo.ipi_lock);
1245 		return 0;
1246 	}
1247 
1248 	inp_list = kalloc_type(struct inpcb *, n, Z_WAITOK);
1249 	if (inp_list == NULL) {
1250 		lck_rw_done(&udbinfo.ipi_lock);
1251 		return ENOMEM;
1252 	}
1253 
1254 	for (inp = LIST_FIRST(udbinfo.ipi_listhead), i = 0; inp && i < n;
1255 	    inp = LIST_NEXT(inp, inp_list)) {
1256 		if (inp->inp_gencnt <= gencnt &&
1257 		    inp->inp_state != INPCB_STATE_DEAD) {
1258 			inp_list[i++] = inp;
1259 		}
1260 	}
1261 	n = i;
1262 
1263 	error = 0;
1264 	for (i = 0; i < n; i++) {
1265 		struct xinpcb xi;
1266 
1267 		inp = inp_list[i];
1268 
1269 		if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) == WNT_STOPUSING) {
1270 			continue;
1271 		}
1272 		udp_lock(inp->inp_socket, 1, 0);
1273 		if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
1274 			udp_unlock(inp->inp_socket, 1, 0);
1275 			continue;
1276 		}
1277 		if (inp->inp_gencnt > gencnt) {
1278 			udp_unlock(inp->inp_socket, 1, 0);
1279 			continue;
1280 		}
1281 
1282 		bzero(&xi, sizeof(xi));
1283 		xi.xi_len = sizeof(xi);
1284 		/* XXX should avoid extra copy */
1285 		inpcb_to_compat(inp, &xi.xi_inp);
1286 		if (inp->inp_socket) {
1287 			sotoxsocket(inp->inp_socket, &xi.xi_socket);
1288 		}
1289 
1290 		udp_unlock(inp->inp_socket, 1, 0);
1291 
1292 		error = SYSCTL_OUT(req, &xi, sizeof(xi));
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 = udbinfo.ipi_gencnt;
1305 		xig.xig_sogen = so_gencnt;
1306 		xig.xig_count = udbinfo.ipi_count;
1307 		error = SYSCTL_OUT(req, &xig, sizeof(xig));
1308 	}
1309 
1310 	lck_rw_done(&udbinfo.ipi_lock);
1311 	kfree_type(struct inpcb *, sz, inp_list);
1312 	return error;
1313 }
1314 
1315 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist,
1316     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0, udp_pcblist,
1317     "S,xinpcb", "List of active UDP sockets");
1318 
1319 #if XNU_TARGET_OS_OSX
1320 
1321 static int
1322 udp_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_shared(&udbinfo.ipi_lock);
1335 	if (req->oldptr == USER_ADDR_NULL) {
1336 		n = udbinfo.ipi_count;
1337 		req->oldidx =
1338 		    2 * (sizeof(xig)) + (n + n / 8) * sizeof(struct xinpcb64);
1339 		lck_rw_done(&udbinfo.ipi_lock);
1340 		return 0;
1341 	}
1342 
1343 	if (req->newptr != USER_ADDR_NULL) {
1344 		lck_rw_done(&udbinfo.ipi_lock);
1345 		return EPERM;
1346 	}
1347 
1348 	/*
1349 	 * OK, now we're committed to doing something.
1350 	 */
1351 	gencnt = udbinfo.ipi_gencnt;
1352 	sz = n = udbinfo.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(&udbinfo.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(&udbinfo.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(&udbinfo.ipi_lock);
1375 		return ENOMEM;
1376 	}
1377 
1378 	for (inp = LIST_FIRST(udbinfo.ipi_listhead), i = 0; inp && i < n;
1379 	    inp = LIST_NEXT(inp, inp_list)) {
1380 		if (inp->inp_gencnt <= gencnt &&
1381 		    inp->inp_state != INPCB_STATE_DEAD) {
1382 			inp_list[i++] = inp;
1383 		}
1384 	}
1385 	n = i;
1386 
1387 	error = 0;
1388 	for (i = 0; i < n; i++) {
1389 		struct xinpcb64 xi;
1390 
1391 		inp = inp_list[i];
1392 
1393 		if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) == WNT_STOPUSING) {
1394 			continue;
1395 		}
1396 		udp_lock(inp->inp_socket, 1, 0);
1397 		if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
1398 			udp_unlock(inp->inp_socket, 1, 0);
1399 			continue;
1400 		}
1401 		if (inp->inp_gencnt > gencnt) {
1402 			udp_unlock(inp->inp_socket, 1, 0);
1403 			continue;
1404 		}
1405 
1406 		bzero(&xi, sizeof(xi));
1407 		xi.xi_len = sizeof(xi);
1408 		inpcb_to_xinpcb64(inp, &xi);
1409 		if (inp->inp_socket) {
1410 			sotoxsocket64(inp->inp_socket, &xi.xi_socket);
1411 		}
1412 
1413 		udp_unlock(inp->inp_socket, 1, 0);
1414 
1415 		error = SYSCTL_OUT(req, &xi, sizeof(xi));
1416 	}
1417 	if (!error) {
1418 		/*
1419 		 * Give the user an updated idea of our state.
1420 		 * If the generation differs from what we told
1421 		 * her before, she knows that something happened
1422 		 * while we were processing this request, and it
1423 		 * might be necessary to retry.
1424 		 */
1425 		bzero(&xig, sizeof(xig));
1426 		xig.xig_len = sizeof(xig);
1427 		xig.xig_gen = udbinfo.ipi_gencnt;
1428 		xig.xig_sogen = so_gencnt;
1429 		xig.xig_count = udbinfo.ipi_count;
1430 		error = SYSCTL_OUT(req, &xig, sizeof(xig));
1431 	}
1432 
1433 	lck_rw_done(&udbinfo.ipi_lock);
1434 	kfree_type(struct inpcb *, sz, inp_list);
1435 	return error;
1436 }
1437 
1438 SYSCTL_PROC(_net_inet_udp, OID_AUTO, pcblist64,
1439     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0, udp_pcblist64,
1440     "S,xinpcb64", "List of active UDP sockets");
1441 
1442 #endif /* XNU_TARGET_OS_OSX */
1443 
1444 static int
1445 udp_pcblist_n SYSCTL_HANDLER_ARGS
1446 {
1447 #pragma unused(oidp, arg1, arg2)
1448 	return get_pcblist_n(IPPROTO_UDP, req, &udbinfo);
1449 }
1450 
1451 SYSCTL_PROC(_net_inet_udp, OID_AUTO, pcblist_n,
1452     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0, udp_pcblist_n,
1453     "S,xinpcb_n", "List of active UDP sockets");
1454 
1455 __private_extern__ void
udp_get_ports_used(ifnet_t ifp,int protocol,uint32_t flags,bitstr_t * __counted_by (bitstr_size (IP_PORTRANGE_SIZE))bitfield)1456 udp_get_ports_used(ifnet_t ifp, int protocol, uint32_t flags,
1457     bitstr_t *__counted_by(bitstr_size(IP_PORTRANGE_SIZE)) bitfield)
1458 {
1459 	inpcb_get_ports_used(ifp, protocol, flags, bitfield,
1460 	    &udbinfo);
1461 }
1462 
1463 __private_extern__ uint32_t
udp_count_opportunistic(unsigned int ifindex,u_int32_t flags)1464 udp_count_opportunistic(unsigned int ifindex, u_int32_t flags)
1465 {
1466 	return inpcb_count_opportunistic(ifindex, &udbinfo, flags);
1467 }
1468 
1469 __private_extern__ uint32_t
udp_find_anypcb_byaddr(struct ifaddr * ifa)1470 udp_find_anypcb_byaddr(struct ifaddr *ifa)
1471 {
1472 #if SKYWALK
1473 	if (netns_is_enabled()) {
1474 		return netns_find_anyres_byaddr(ifa, IPPROTO_UDP);
1475 	} else
1476 #endif /* SKYWALK */
1477 	return inpcb_find_anypcb_byaddr(ifa, &udbinfo);
1478 }
1479 
1480 static int
udp_check_pktinfo(struct mbuf * control,struct ifnet ** outif,struct in_addr * laddr)1481 udp_check_pktinfo(struct mbuf *control, struct ifnet **outif,
1482     struct in_addr *laddr)
1483 {
1484 	struct cmsghdr *cm = 0;
1485 	struct in_pktinfo *pktinfo;
1486 	struct ifnet *ifp;
1487 
1488 	if (outif != NULL) {
1489 		*outif = NULL;
1490 	}
1491 
1492 	/*
1493 	 * XXX: Currently, we assume all the optional information is stored
1494 	 * in a single mbuf.
1495 	 */
1496 	if (control->m_next) {
1497 		return EINVAL;
1498 	}
1499 
1500 	if (control->m_len < CMSG_LEN(0)) {
1501 		return EINVAL;
1502 	}
1503 
1504 	for (cm = M_FIRST_CMSGHDR(control);
1505 	    is_cmsg_valid(control, cm);
1506 	    cm = M_NXT_CMSGHDR(control, cm)) {
1507 		if (cm->cmsg_level != IPPROTO_IP ||
1508 		    cm->cmsg_type != IP_PKTINFO) {
1509 			continue;
1510 		}
1511 
1512 		if (cm->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo))) {
1513 			return EINVAL;
1514 		}
1515 
1516 		pktinfo =  (struct in_pktinfo *)(void *)CMSG_DATA(cm);
1517 
1518 		/* Check for a valid ifindex in pktinfo */
1519 		ifnet_head_lock_shared();
1520 
1521 		if (pktinfo->ipi_ifindex > if_index) {
1522 			ifnet_head_done();
1523 			return ENXIO;
1524 		}
1525 
1526 		/*
1527 		 * If ipi_ifindex is specified it takes precedence
1528 		 * over ipi_spec_dst.
1529 		 */
1530 		if (pktinfo->ipi_ifindex) {
1531 			ifp = ifindex2ifnet[pktinfo->ipi_ifindex];
1532 			if (ifp == NULL) {
1533 				ifnet_head_done();
1534 				return ENXIO;
1535 			}
1536 			if (outif != NULL) {
1537 				ifnet_reference(ifp);
1538 				*outif = ifp;
1539 			}
1540 			ifnet_head_done();
1541 			laddr->s_addr = INADDR_ANY;
1542 			break;
1543 		}
1544 
1545 		ifnet_head_done();
1546 
1547 		/*
1548 		 * Use the provided ipi_spec_dst address for temp
1549 		 * source address.
1550 		 */
1551 		*laddr = pktinfo->ipi_spec_dst;
1552 		break;
1553 	}
1554 	return 0;
1555 }
1556 
1557 int
udp_output(struct inpcb * inp,struct mbuf * m,struct sockaddr * addr,struct mbuf * control,struct proc * p)1558 udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr,
1559     struct mbuf *control, struct proc *p)
1560 {
1561 	struct udpiphdr *ui;
1562 	int len = m->m_pkthdr.len;
1563 	struct sockaddr_in *sin;
1564 	struct in_addr origladdr, laddr, faddr, pi_laddr;
1565 	u_short lport, fport;
1566 	int error = 0, udp_dodisconnect = 0, pktinfo = 0;
1567 	struct socket *so = inp->inp_socket;
1568 	int soopts = 0;
1569 	struct mbuf *inpopts;
1570 	struct ip_moptions *__single mopts;
1571 	struct route ro;
1572 	struct ip_out_args ipoa;
1573 	bool sndinprog_cnt_used = false;
1574 #if CONTENT_FILTER
1575 	struct m_tag *__single cfil_tag = NULL;
1576 	bool cfil_faddr_use = false;
1577 	uint32_t cfil_so_state_change_cnt = 0;
1578 	uint32_t cfil_so_options = 0;
1579 	struct sockaddr *__single cfil_faddr = NULL;
1580 #endif
1581 	bool check_qos_marking_again = (so->so_flags1 & SOF1_QOSMARKING_POLICY_OVERRIDE) ? FALSE : TRUE;
1582 
1583 	bzero(&ipoa, sizeof(ipoa));
1584 	ipoa.ipoa_boundif = IFSCOPE_NONE;
1585 	ipoa.ipoa_flags = IPOAF_SELECT_SRCIF;
1586 
1587 	ifnet_ref_t outif = NULL;
1588 	struct flowadv *adv = &ipoa.ipoa_flowadv;
1589 	int sotc = SO_TC_UNSPEC;
1590 	int netsvctype = _NET_SERVICE_TYPE_UNSPEC;
1591 	struct ifnet *origoutifp = NULL;
1592 	int flowadv = 0;
1593 	int tos = IPTOS_UNSPEC;
1594 
1595 	/* Enable flow advisory only when connected */
1596 	flowadv = (so->so_state & SS_ISCONNECTED) ? 1 : 0;
1597 	pi_laddr.s_addr = INADDR_ANY;
1598 
1599 	KERNEL_DEBUG(DBG_FNC_UDP_OUTPUT | DBG_FUNC_START, 0, 0, 0, 0, 0);
1600 
1601 	socket_lock_assert_owned(so);
1602 
1603 #if CONTENT_FILTER
1604 	/*
1605 	 * If socket is subject to UDP Content Filter and no addr is passed in,
1606 	 * retrieve CFIL saved state from mbuf and use it if necessary.
1607 	 */
1608 	if (CFIL_DGRAM_FILTERED(so) && !addr) {
1609 		cfil_tag = cfil_dgram_get_socket_state(m, &cfil_so_state_change_cnt, &cfil_so_options, &cfil_faddr, NULL);
1610 		if (cfil_tag) {
1611 			sin = SIN(cfil_faddr);
1612 			if (inp && inp->inp_faddr.s_addr == INADDR_ANY) {
1613 				/*
1614 				 * Socket is unconnected, simply use the saved faddr as 'addr' to go through
1615 				 * the connect/disconnect logic.
1616 				 */
1617 				addr = SA(cfil_faddr);
1618 			} else if ((so->so_state_change_cnt != cfil_so_state_change_cnt) &&
1619 			    (inp->inp_fport != sin->sin_port ||
1620 			    inp->inp_faddr.s_addr != sin->sin_addr.s_addr)) {
1621 				/*
1622 				 * Socket is connected but socket state and dest addr/port changed.
1623 				 * We need to use the saved faddr info.
1624 				 */
1625 				cfil_faddr_use = true;
1626 			}
1627 		}
1628 	}
1629 #endif
1630 
1631 	if (control != NULL) {
1632 		tos = so_tos_from_control(control);
1633 		sotc = so_tc_from_control(control, &netsvctype);
1634 		VERIFY(outif == NULL);
1635 		error = udp_check_pktinfo(control, &outif, &pi_laddr);
1636 		m_freem(control);
1637 		control = NULL;
1638 		if (error) {
1639 			UDP_LOG(inp, "udp_check_pktinfo error %d", error);
1640 			goto release;
1641 		}
1642 		if (outif != NULL) {
1643 			pktinfo++;
1644 			ipoa.ipoa_boundif = outif->if_index;
1645 		}
1646 	}
1647 	if (sotc == SO_TC_UNSPEC) {
1648 		sotc = so->so_traffic_class;
1649 		netsvctype = so->so_netsvctype;
1650 	}
1651 
1652 	KERNEL_DEBUG(DBG_LAYER_OUT_BEG, inp->inp_fport, inp->inp_lport,
1653 	    inp->inp_laddr.s_addr, inp->inp_faddr.s_addr,
1654 	    (htons((u_short)len + sizeof(struct udphdr))));
1655 
1656 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
1657 		error = EMSGSIZE;
1658 		UDP_LOG(inp, "len %d too big error EMSGSIZE", len);
1659 		goto release;
1660 	}
1661 
1662 	if (flowadv && INP_WAIT_FOR_IF_FEEDBACK(inp)) {
1663 		/*
1664 		 * The socket is flow-controlled, drop the packets
1665 		 * until the inp is not flow controlled
1666 		 */
1667 		error = ENOBUFS;
1668 		UDP_LOG(inp, "flow controlled error ENOBUFS");
1669 		goto release;
1670 	}
1671 	/*
1672 	 * If socket was bound to an ifindex, tell ip_output about it.
1673 	 * If the ancillary IP_PKTINFO option contains an interface index,
1674 	 * it takes precedence over the one specified by IP_BOUND_IF.
1675 	 */
1676 	if (ipoa.ipoa_boundif == IFSCOPE_NONE &&
1677 	    (inp->inp_flags & INP_BOUND_IF)) {
1678 		VERIFY(inp->inp_boundifp != NULL);
1679 		ifnet_reference(inp->inp_boundifp);     /* for this routine */
1680 		if (outif != NULL) {
1681 			ifnet_release(outif);
1682 		}
1683 		outif = inp->inp_boundifp;
1684 		ipoa.ipoa_boundif = outif->if_index;
1685 	}
1686 	if (INP_NO_CELLULAR(inp)) {
1687 		ipoa.ipoa_flags |=  IPOAF_NO_CELLULAR;
1688 	}
1689 	if (INP_NO_EXPENSIVE(inp)) {
1690 		ipoa.ipoa_flags |=  IPOAF_NO_EXPENSIVE;
1691 	}
1692 	if (INP_NO_CONSTRAINED(inp)) {
1693 		ipoa.ipoa_flags |=  IPOAF_NO_CONSTRAINED;
1694 	}
1695 	if (INP_AWDL_UNRESTRICTED(inp)) {
1696 		ipoa.ipoa_flags |=  IPOAF_AWDL_UNRESTRICTED;
1697 	}
1698 	if (INP_MANAGEMENT_ALLOWED(inp)) {
1699 		ipoa.ipoa_flags |= IPOAF_MANAGEMENT_ALLOWED;
1700 	}
1701 	if (INP_ULTRA_CONSTRAINED_ALLOWED(inp)) {
1702 		ipoa.ipoa_flags |= IPOAF_ULTRA_CONSTRAINED_ALLOWED;
1703 	}
1704 	ipoa.ipoa_sotc = sotc;
1705 	ipoa.ipoa_netsvctype = netsvctype;
1706 	soopts |= IP_OUTARGS;
1707 
1708 	/*
1709 	 * If there was a routing change, discard cached route and check
1710 	 * that we have a valid source address.  Reacquire a new source
1711 	 * address if INADDR_ANY was specified.
1712 	 *
1713 	 * If we are using cfil saved state, go through this cache cleanup
1714 	 * so that we can get a new route.
1715 	 */
1716 	if (ROUTE_UNUSABLE(&inp->inp_route)
1717 #if CONTENT_FILTER
1718 	    || cfil_faddr_use
1719 #endif
1720 	    ) {
1721 		struct in_ifaddr *ia = NULL;
1722 
1723 		ROUTE_RELEASE(&inp->inp_route);
1724 
1725 		/* src address is gone? */
1726 		if (inp->inp_laddr.s_addr != INADDR_ANY &&
1727 		    (ia = ifa_foraddr(inp->inp_laddr.s_addr)) == NULL) {
1728 			if (!(inp->inp_flags & INP_INADDR_ANY) ||
1729 			    (so->so_state & SS_ISCONNECTED)) {
1730 				/*
1731 				 * Rdar://5448998
1732 				 * If the source address is gone, return an
1733 				 * error if:
1734 				 * - the source was specified
1735 				 * - the socket was already connected
1736 				 */
1737 				soevent(so, (SO_FILT_HINT_LOCKED |
1738 				    SO_FILT_HINT_NOSRCADDR));
1739 				error = EADDRNOTAVAIL;
1740 				UDP_LOG(inp, "source address not available error EADDRNOTAVAIL");
1741 				goto release;
1742 			} else {
1743 				/* new src will be set later */
1744 				inp->inp_laddr.s_addr = INADDR_ANY;
1745 				inp->inp_last_outifp = NULL;
1746 #if SKYWALK
1747 				if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
1748 					netns_set_ifnet(&inp->inp_netns_token, NULL);
1749 				}
1750 #endif /* SKYWALK */
1751 			}
1752 		}
1753 		if (ia != NULL) {
1754 			ifa_remref(&ia->ia_ifa);
1755 		}
1756 	}
1757 
1758 	/*
1759 	 * IP_PKTINFO option check.  If a temporary scope or src address
1760 	 * is provided, use it for this packet only and make sure we forget
1761 	 * it after sending this datagram.
1762 	 */
1763 	if (pi_laddr.s_addr != INADDR_ANY ||
1764 	    (ipoa.ipoa_boundif != IFSCOPE_NONE && pktinfo)) {
1765 		/* temp src address for this datagram only */
1766 		laddr = pi_laddr;
1767 		origladdr.s_addr = INADDR_ANY;
1768 		/* we don't want to keep the laddr or route */
1769 		udp_dodisconnect = 1;
1770 		/* remember we don't care about src addr */
1771 		inp->inp_flags |= INP_INADDR_ANY;
1772 	} else {
1773 		origladdr = laddr = inp->inp_laddr;
1774 	}
1775 
1776 	origoutifp = inp->inp_last_outifp;
1777 	faddr = inp->inp_faddr;
1778 	lport = inp->inp_lport;
1779 	fport = inp->inp_fport;
1780 
1781 #if CONTENT_FILTER
1782 	if (cfil_faddr_use) {
1783 		faddr = SIN(cfil_faddr)->sin_addr;
1784 		fport = SIN(cfil_faddr)->sin_port;
1785 	}
1786 #endif
1787 	inp->inp_sndinprog_cnt++;
1788 	sndinprog_cnt_used = true;
1789 
1790 	if (addr) {
1791 		sin = SIN(addr);
1792 		if (faddr.s_addr != INADDR_ANY) {
1793 			error = EISCONN;
1794 			UDP_LOG(inp, "socket already connected error EISCONN");
1795 			goto release;
1796 		}
1797 		if (lport == 0) {
1798 			inp_enter_bind_in_progress(so);
1799 
1800 			/*
1801 			 * In case we don't have a local port set, go through
1802 			 * the full connect.  We don't have a local port yet
1803 			 * (i.e., we can't be looked up), so it's not an issue
1804 			 * if the input runs at the same time we do this.
1805 			 */
1806 			/* if we have a source address specified, use that */
1807 			if (pi_laddr.s_addr != INADDR_ANY) {
1808 				inp->inp_laddr = pi_laddr;
1809 			}
1810 			/*
1811 			 * If a scope is specified, use it.  Scope from
1812 			 * IP_PKTINFO takes precendence over the the scope
1813 			 * set via INP_BOUND_IF.
1814 			 */
1815 			error = in_pcbconnect(inp, addr, p, ipoa.ipoa_boundif,
1816 			    &outif);
1817 
1818 			inp_exit_bind_in_progress(so);
1819 
1820 			if (error) {
1821 				UDP_LOG(inp, "in_pcbconnect error %d", error);
1822 				goto release;
1823 			}
1824 
1825 			laddr = inp->inp_laddr;
1826 			lport = inp->inp_lport;
1827 			faddr = inp->inp_faddr;
1828 			fport = inp->inp_fport;
1829 			udp_dodisconnect = 1;
1830 
1831 			/* synch up in case in_pcbladdr() overrides */
1832 			if (outif != NULL && ipoa.ipoa_boundif != IFSCOPE_NONE) {
1833 				ipoa.ipoa_boundif = outif->if_index;
1834 			}
1835 		} else {
1836 			/*
1837 			 * Fast path case
1838 			 *
1839 			 * We have a full address and a local port; use those
1840 			 * info to build the packet without changing the pcb
1841 			 * and interfering with the input path. See 3851370.
1842 			 *
1843 			 * Scope from IP_PKTINFO takes precendence over the
1844 			 * the scope set via INP_BOUND_IF.
1845 			 */
1846 			if (laddr.s_addr == INADDR_ANY) {
1847 				if ((error = in_pcbladdr(inp, addr, &laddr,
1848 				    ipoa.ipoa_boundif, &outif, 0)) != 0) {
1849 					UDP_LOG(inp, "in_pcbladdr error %d", error);
1850 					goto release;
1851 				}
1852 				/*
1853 				 * from pcbconnect: remember we don't
1854 				 * care about src addr.
1855 				 */
1856 				inp->inp_flags |= INP_INADDR_ANY;
1857 
1858 				/* synch up in case in_pcbladdr() overrides */
1859 				if (outif != NULL &&
1860 				    ipoa.ipoa_boundif != IFSCOPE_NONE) {
1861 					ipoa.ipoa_boundif = outif->if_index;
1862 				}
1863 			}
1864 
1865 			faddr = sin->sin_addr;
1866 			fport = sin->sin_port;
1867 		}
1868 	} else {
1869 		if (faddr.s_addr == INADDR_ANY) {
1870 			error = ENOTCONN;
1871 			UDP_LOG(inp, "not connected error ENOTCONN");
1872 			goto release;
1873 		}
1874 	}
1875 
1876 	if (inp->inp_flowhash == 0) {
1877 		inp_calc_flowhash(inp);
1878 		ASSERT(inp->inp_flowhash != 0);
1879 	}
1880 
1881 	if (fport == htons(53) && !(so->so_flags1 & SOF1_DNS_COUNTED)) {
1882 		so->so_flags1 |= SOF1_DNS_COUNTED;
1883 		INC_ATOMIC_INT64_LIM(net_api_stats.nas_socket_inet_dgram_dns);
1884 	}
1885 
1886 	/*
1887 	 * Calculate data length and get a mbuf
1888 	 * for UDP and IP headers.
1889 	 */
1890 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT, 1);
1891 	if (m == 0) {
1892 		error = ENOBUFS;
1893 		UDP_LOG(inp, "M_PREPEND error ENOBUFS");
1894 		goto abort;
1895 	}
1896 
1897 	/*
1898 	 * Fill in mbuf with extended UDP header
1899 	 * and addresses and length put into network format.
1900 	 */
1901 	ui = mtod(m, struct udpiphdr *);
1902 	bzero(ui->ui_x1, sizeof(ui->ui_x1));    /* XXX still needed? */
1903 	ui->ui_pr = IPPROTO_UDP;
1904 	ui->ui_src = laddr;
1905 	ui->ui_dst = faddr;
1906 	ui->ui_sport = lport;
1907 	ui->ui_dport = fport;
1908 	ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
1909 
1910 	/*
1911 	 * Set the Don't Fragment bit in the IP header.
1912 	 */
1913 	if (inp->inp_flags2 & INP2_DONTFRAG) {
1914 		struct ip *ip;
1915 
1916 		ip = (struct ip *)&ui->ui_i;
1917 		ip->ip_off |= IP_DF;
1918 	}
1919 
1920 	/*
1921 	 * Set up checksum to pseudo header checksum and output datagram.
1922 	 *
1923 	 * Treat flows to be CLAT46'd as IPv6 flow and compute checksum
1924 	 * no matter what, as IPv6 mandates checksum for UDP.
1925 	 *
1926 	 * Here we only compute the one's complement sum of the pseudo header.
1927 	 * The payload computation and final complement is delayed to much later
1928 	 * in IP processing to decide if remaining computation needs to be done
1929 	 * through offload.
1930 	 *
1931 	 * That is communicated by setting CSUM_UDP in csum_flags.
1932 	 * The offset of checksum from the start of ULP header is communicated
1933 	 * through csum_data.
1934 	 *
1935 	 * Note since this already contains the pseudo checksum header, any
1936 	 * later operation at IP layer that modify the values used here must
1937 	 * update the checksum as well (for example NAT etc).
1938 	 */
1939 	if ((inp->inp_flags2 & INP2_CLAT46_FLOW) ||
1940 	    (udpcksum && !(inp->inp_flags & INP_UDP_NOCKSUM))) {
1941 		ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
1942 		    htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
1943 		m->m_pkthdr.csum_flags = (CSUM_UDP | CSUM_ZERO_INVERT);
1944 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1945 	} else {
1946 		ui->ui_sum = 0;
1947 	}
1948 	((struct ip *)ui)->ip_len = (uint16_t)(sizeof(struct udpiphdr) + len);
1949 	((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl;    /* XXX */
1950 	if (tos != IPTOS_UNSPEC) {
1951 		((struct ip *)ui)->ip_tos = (uint8_t)(tos & IPTOS_MASK);
1952 	} else {
1953 		((struct ip *)ui)->ip_tos = inp->inp_ip_tos;    /* XXX */
1954 	}
1955 	udpstat.udps_opackets++;
1956 
1957 	KERNEL_DEBUG(DBG_LAYER_OUT_END, ui->ui_dport, ui->ui_sport,
1958 	    ui->ui_src.s_addr, ui->ui_dst.s_addr, ui->ui_ulen);
1959 
1960 #if NECP
1961 	{
1962 		necp_kernel_policy_id policy_id;
1963 		necp_kernel_policy_id skip_policy_id;
1964 		u_int32_t route_rule_id;
1965 		u_int32_t pass_flags;
1966 
1967 		/*
1968 		 * We need a route to perform NECP route rule checks
1969 		 */
1970 		if (net_qos_policy_restricted != 0 &&
1971 		    ROUTE_UNUSABLE(&inp->inp_route)) {
1972 			struct sockaddr_in to;
1973 			struct sockaddr_in from;
1974 
1975 			ROUTE_RELEASE(&inp->inp_route);
1976 
1977 			SOCKADDR_ZERO(&from, sizeof(struct sockaddr_in));
1978 			from.sin_family = AF_INET;
1979 			from.sin_len = sizeof(struct sockaddr_in);
1980 			from.sin_addr = laddr;
1981 
1982 			SOCKADDR_ZERO(&to, sizeof(struct sockaddr_in));
1983 			to.sin_family = AF_INET;
1984 			to.sin_len = sizeof(struct sockaddr_in);
1985 			to.sin_addr = faddr;
1986 
1987 			inp->inp_route.ro_dst.sa_family = AF_INET;
1988 			inp->inp_route.ro_dst.sa_len = sizeof(struct sockaddr_in);
1989 			SIN(&inp->inp_route.ro_dst)->sin_addr = faddr;
1990 
1991 			rtalloc_scoped(&inp->inp_route, ipoa.ipoa_boundif);
1992 
1993 			inp_update_necp_policy(inp, SA(&from),
1994 			    SA(&to), ipoa.ipoa_boundif);
1995 			inp->inp_policyresult.results.qos_marking_gencount = 0;
1996 		}
1997 
1998 		if (!necp_socket_is_allowed_to_send_recv_v4(inp, lport, fport,
1999 		    &laddr, &faddr, NULL, 0, &policy_id, &route_rule_id, &skip_policy_id, &pass_flags)) {
2000 			error = EHOSTUNREACH;
2001 			UDP_LOG_DROP_NECP((struct ip *)&ui->ui_i, &ui->ui_u, inp, true);
2002 			m_drop(m, DROPTAP_FLAG_DIR_OUT | DROPTAP_FLAG_L2_MISSING, DROP_REASON_UDP_NECP, NULL, 0);
2003 			m = NULL;
2004 			goto abort;
2005 		}
2006 
2007 		necp_mark_packet_from_socket(m, inp, policy_id, route_rule_id, skip_policy_id, pass_flags);
2008 
2009 		if (net_qos_policy_restricted != 0) {
2010 			necp_socket_update_qos_marking(inp, inp->inp_route.ro_rt, route_rule_id);
2011 		}
2012 	}
2013 #endif /* NECP */
2014 	if ((so->so_flags1 & SOF1_QOSMARKING_ALLOWED)) {
2015 		ipoa.ipoa_flags |= IPOAF_QOSMARKING_ALLOWED;
2016 	}
2017 	if (check_qos_marking_again) {
2018 		ipoa.ipoa_flags |= IPOAF_REDO_QOSMARKING_POLICY;
2019 	}
2020 	ipoa.qos_marking_gencount = inp->inp_policyresult.results.qos_marking_gencount;
2021 
2022 #if IPSEC
2023 	if (inp->inp_sp != NULL && ipsec_setsocket(m, inp->inp_socket) != 0) {
2024 		error = ENOBUFS;
2025 		UDP_LOG_DROP_PCB((struct ip *)&ui->ui_i, &ui->ui_u, inp, true, "ipsec_setsocket error ENOBUFS");
2026 		m_drop(m, DROPTAP_FLAG_DIR_OUT | DROPTAP_FLAG_L2_MISSING, DROP_REASON_UDP_IPSEC, NULL, 0);
2027 		m = NULL;
2028 		goto abort;
2029 	}
2030 #endif /* IPSEC */
2031 
2032 	inpopts = inp->inp_options;
2033 #if CONTENT_FILTER
2034 	if (cfil_tag && (inp->inp_socket->so_options != cfil_so_options)) {
2035 		soopts |= (cfil_so_options & (SO_DONTROUTE | SO_BROADCAST));
2036 	} else
2037 #endif
2038 	soopts |= (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST));
2039 
2040 	mopts = inp->inp_moptions;
2041 	if (mopts != NULL) {
2042 		IMO_LOCK(mopts);
2043 		IMO_ADDREF_LOCKED(mopts);
2044 		if (IN_MULTICAST(ntohl(ui->ui_dst.s_addr)) &&
2045 		    mopts->imo_multicast_ifp != NULL) {
2046 			/* no reference needed */
2047 			inp->inp_last_outifp = mopts->imo_multicast_ifp;
2048 #if SKYWALK
2049 			if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
2050 				netns_set_ifnet(&inp->inp_netns_token,
2051 				    inp->inp_last_outifp);
2052 			}
2053 #endif /* SKYWALK */
2054 		}
2055 		IMO_UNLOCK(mopts);
2056 	}
2057 
2058 	/* Copy the cached route and take an extra reference */
2059 	inp_route_copyout(inp, &ro);
2060 
2061 	set_packet_service_class(m, so, sotc, 0);
2062 	m->m_pkthdr.pkt_flowsrc = FLOWSRC_INPCB;
2063 	m->m_pkthdr.pkt_flowid = inp->inp_flowhash;
2064 	m->m_pkthdr.pkt_proto = IPPROTO_UDP;
2065 	m->m_pkthdr.pkt_flags |= (PKTF_FLOW_ID | PKTF_FLOW_LOCALSRC);
2066 	if (flowadv) {
2067 		m->m_pkthdr.pkt_flags |= PKTF_FLOW_ADV;
2068 	}
2069 	m->m_pkthdr.tx_udp_pid = so->last_pid;
2070 	if (so->so_flags & SOF_DELEGATED) {
2071 		m->m_pkthdr.tx_udp_e_pid = so->e_pid;
2072 	} else {
2073 		m->m_pkthdr.tx_udp_e_pid = 0;
2074 	}
2075 #if (DEBUG || DEVELOPMENT)
2076 	if (so->so_flags & SOF_MARK_WAKE_PKT) {
2077 		so->so_flags &= ~SOF_MARK_WAKE_PKT;
2078 		m->m_pkthdr.pkt_flags |= PKTF_WAKE_PKT;
2079 	}
2080 #endif /* (DEBUG || DEVELOPMENT) */
2081 
2082 	m_add_crumb(m, PKT_CRUMB_UDP_OUTPUT);
2083 
2084 	if (ipoa.ipoa_boundif != IFSCOPE_NONE) {
2085 		ipoa.ipoa_flags |= IPOAF_BOUND_IF;
2086 	}
2087 
2088 	if (laddr.s_addr != INADDR_ANY) {
2089 		ipoa.ipoa_flags |= IPOAF_BOUND_SRCADDR;
2090 	}
2091 
2092 	socket_unlock(so, 0);
2093 	error = ip_output(m, inpopts, &ro, soopts, mopts, &ipoa);
2094 	m = NULL;
2095 	socket_lock(so, 0);
2096 	if (mopts != NULL) {
2097 		IMO_REMREF(mopts);
2098 	}
2099 
2100 	if (check_qos_marking_again) {
2101 		inp->inp_policyresult.results.qos_marking_gencount = ipoa.qos_marking_gencount;
2102 
2103 		if (ipoa.ipoa_flags & IPOAF_QOSMARKING_ALLOWED) {
2104 			inp->inp_socket->so_flags1 |= SOF1_QOSMARKING_ALLOWED;
2105 		} else {
2106 			inp->inp_socket->so_flags1 &= ~SOF1_QOSMARKING_ALLOWED;
2107 		}
2108 	}
2109 
2110 	if (error == 0 && nstat_collect) {
2111 		stats_functional_type ifnet_count_type = stats_functional_type_none;
2112 
2113 		if (ro.ro_rt != NULL) {
2114 			ifnet_count_type = IFNET_COUNT_TYPE(ro.ro_rt->rt_ifp);
2115 		}
2116 		INP_ADD_STAT(inp, ifnet_count_type, txpackets, 1);
2117 		INP_ADD_STAT(inp, ifnet_count_type, txbytes, len);
2118 		inp_set_activity_bitmap(inp);
2119 	}
2120 
2121 	if (flowadv && (adv->code == FADV_FLOW_CONTROLLED ||
2122 	    adv->code == FADV_SUSPENDED)) {
2123 		/*
2124 		 * return a hint to the application that
2125 		 * the packet has been dropped
2126 		 */
2127 		error = ENOBUFS;
2128 		inp_set_fc_state(inp, adv->code);
2129 	}
2130 
2131 	/* Synchronize PCB cached route */
2132 	inp_route_copyin(inp, &ro);
2133 
2134 	if (inp->inp_route.ro_rt != NULL) {
2135 		if (IS_LOCALNET_ROUTE(inp->inp_route.ro_rt)) {
2136 			inp->inp_flags2 |= INP2_LAST_ROUTE_LOCAL;
2137 		} else {
2138 			inp->inp_flags2 &= ~INP2_LAST_ROUTE_LOCAL;
2139 		}
2140 	}
2141 
2142 abort:
2143 	if (udp_dodisconnect) {
2144 		/* Always discard the cached route for unconnected socket */
2145 		ROUTE_RELEASE(&inp->inp_route);
2146 		in_pcbdisconnect(inp);
2147 		inp->inp_laddr = origladdr;     /* XXX rehash? */
2148 		/* no reference needed */
2149 		inp->inp_last_outifp = origoutifp;
2150 #if SKYWALK
2151 		if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
2152 			netns_set_ifnet(&inp->inp_netns_token,
2153 			    inp->inp_last_outifp);
2154 		}
2155 #endif /* SKYWALK */
2156 	} else if (inp->inp_route.ro_rt != NULL) {
2157 		struct rtentry *rt = inp->inp_route.ro_rt;
2158 		struct ifnet *outifp;
2159 
2160 		if (rt->rt_flags & (RTF_MULTICAST | RTF_BROADCAST)) {
2161 			rt = NULL;      /* unusable */
2162 		}
2163 #if CONTENT_FILTER
2164 		/*
2165 		 * Discard temporary route for cfil case
2166 		 */
2167 		if (cfil_faddr_use) {
2168 			rt = NULL;      /* unusable */
2169 		}
2170 #endif
2171 
2172 		/*
2173 		 * Always discard if it is a multicast or broadcast route.
2174 		 */
2175 		if (rt == NULL) {
2176 			ROUTE_RELEASE(&inp->inp_route);
2177 		}
2178 
2179 		/*
2180 		 * If the destination route is unicast, update outifp with
2181 		 * that of the route interface used by IP.
2182 		 */
2183 		if (rt != NULL &&
2184 		    (outifp = rt->rt_ifp) != inp->inp_last_outifp) {
2185 			inp->inp_last_outifp = outifp; /* no reference needed */
2186 #if SKYWALK
2187 			if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
2188 				netns_set_ifnet(&inp->inp_netns_token,
2189 				    inp->inp_last_outifp);
2190 			}
2191 #endif /* SKYWALK */
2192 
2193 			so->so_pktheadroom = (uint16_t)P2ROUNDUP(
2194 				sizeof(struct udphdr) +
2195 				sizeof(struct ip) +
2196 				ifnet_hdrlen(outifp) +
2197 				ifnet_mbuf_packetpreamblelen(outifp),
2198 				sizeof(u_int32_t));
2199 		}
2200 	} else {
2201 		ROUTE_RELEASE(&inp->inp_route);
2202 	}
2203 
2204 	/*
2205 	 * If output interface was cellular/expensive, and this socket is
2206 	 * denied access to it, generate an event.
2207 	 */
2208 	if (error != 0 && (ipoa.ipoa_flags & IPOAF_R_IFDENIED) &&
2209 	    (INP_NO_CELLULAR(inp) || INP_NO_EXPENSIVE(inp) || INP_NO_CONSTRAINED(inp))) {
2210 		soevent(so, (SO_FILT_HINT_LOCKED | SO_FILT_HINT_IFDENIED));
2211 	}
2212 
2213 release:
2214 	KERNEL_DEBUG(DBG_FNC_UDP_OUTPUT | DBG_FUNC_END, error, 0, 0, 0, 0);
2215 
2216 	if (m != NULL) {
2217 		m_freem(m);
2218 	}
2219 
2220 	if (outif != NULL) {
2221 		ifnet_release(outif);
2222 	}
2223 
2224 #if CONTENT_FILTER
2225 	if (cfil_tag) {
2226 		m_tag_free(cfil_tag);
2227 	}
2228 #endif
2229 	if (sndinprog_cnt_used) {
2230 		VERIFY(inp->inp_sndinprog_cnt > 0);
2231 		if (--inp->inp_sndinprog_cnt == 0) {
2232 			inp->inp_flags &= ~(INP_FC_FEEDBACK);
2233 			if (inp->inp_sndingprog_waiters > 0) {
2234 				wakeup(&inp->inp_sndinprog_cnt);
2235 			}
2236 		}
2237 		sndinprog_cnt_used = false;
2238 	}
2239 
2240 	return error;
2241 }
2242 
2243 u_int32_t       udp_sendspace = 9216;           /* really max datagram size */
2244 /* 187 1K datagrams (approx 192 KB) */
2245 u_int32_t       udp_recvspace = 187 * (1024 + sizeof(struct sockaddr_in6));
2246 
2247 /* Check that the values of udp send and recv space do not exceed sb_max */
2248 static int
sysctl_udp_sospace(struct sysctl_oid * oidp,void * arg1,int arg2,struct sysctl_req * req)2249 sysctl_udp_sospace(struct sysctl_oid *oidp, void *arg1, int arg2,
2250     struct sysctl_req *req)
2251 {
2252 #pragma unused(arg1, arg2)
2253 	u_int32_t new_value = 0, *space_p = NULL;
2254 	int changed = 0, error = 0;
2255 
2256 	switch (oidp->oid_number) {
2257 	case UDPCTL_RECVSPACE:
2258 		space_p = &udp_recvspace;
2259 		break;
2260 	case UDPCTL_MAXDGRAM:
2261 		space_p = &udp_sendspace;
2262 		break;
2263 	default:
2264 		return EINVAL;
2265 	}
2266 	error = sysctl_io_number(req, *space_p, sizeof(u_int32_t),
2267 	    &new_value, &changed);
2268 	if (changed) {
2269 		if (new_value > 0 && new_value <= sb_max) {
2270 			*space_p = new_value;
2271 		} else {
2272 			error = ERANGE;
2273 		}
2274 	}
2275 	return error;
2276 }
2277 
2278 SYSCTL_PROC(_net_inet_udp, UDPCTL_RECVSPACE, recvspace,
2279     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &udp_recvspace, 0,
2280     &sysctl_udp_sospace, "IU", "Maximum incoming UDP datagram size");
2281 
2282 SYSCTL_PROC(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram,
2283     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &udp_sendspace, 0,
2284     &sysctl_udp_sospace, "IU", "Maximum outgoing UDP datagram size");
2285 
2286 int
udp_abort(struct socket * so)2287 udp_abort(struct socket *so)
2288 {
2289 	struct inpcb *inp;
2290 
2291 	inp = sotoinpcb(so);
2292 	if (inp == NULL) {
2293 		panic("%s: so=%p null inp", __func__, so);
2294 		/* NOTREACHED */
2295 	}
2296 	soisdisconnected(so);
2297 	in_pcbdetach(inp);
2298 	return 0;
2299 }
2300 
2301 int
udp_attach(struct socket * so,int proto,struct proc * p)2302 udp_attach(struct socket *so, int proto, struct proc *p)
2303 {
2304 #pragma unused(proto)
2305 	struct inpcb *inp;
2306 	int error;
2307 
2308 	error = soreserve(so, udp_sendspace, udp_recvspace);
2309 	if (error != 0) {
2310 		return error;
2311 	}
2312 	inp = sotoinpcb(so);
2313 	if (inp != NULL) {
2314 		panic("%s so=%p inp=%p", __func__, so, inp);
2315 		/* NOTREACHED */
2316 	}
2317 	error = in_pcballoc(so, &udbinfo, p);
2318 	if (error != 0) {
2319 		return error;
2320 	}
2321 	inp = (struct inpcb *)so->so_pcb;
2322 	inp->inp_vflag |= INP_IPV4;
2323 	inp->inp_ip_ttl = (uint8_t)ip_defttl;
2324 	if (nstat_collect) {
2325 		nstat_udp_new_pcb(inp);
2326 	}
2327 	return 0;
2328 }
2329 
2330 int
udp_bind(struct socket * so,struct sockaddr * nam,struct proc * p)2331 udp_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
2332 {
2333 	struct inpcb *inp;
2334 	int error;
2335 
2336 	if (nam->sa_family != 0 && nam->sa_family != AF_INET &&
2337 	    nam->sa_family != AF_INET6) {
2338 		return EAFNOSUPPORT;
2339 	}
2340 
2341 	inp = sotoinpcb(so);
2342 	if (inp == NULL) {
2343 		return EINVAL;
2344 	}
2345 
2346 	inp_enter_bind_in_progress(so);
2347 
2348 	error = in_pcbbind(inp, nam, NULL, p);
2349 
2350 #if NECP
2351 	/* Update NECP client with bind result if not in middle of connect */
2352 	if (error == 0 &&
2353 	    (inp->inp_flags2 & INP2_CONNECT_IN_PROGRESS) &&
2354 	    !uuid_is_null(inp->necp_client_uuid)) {
2355 		socket_unlock(so, 0);
2356 		necp_client_assign_from_socket(so->last_pid, inp->necp_client_uuid, inp);
2357 		socket_lock(so, 0);
2358 	}
2359 #endif /* NECP */
2360 
2361 	UDP_LOG_BIND(inp, error);
2362 
2363 	inp_exit_bind_in_progress(so);
2364 
2365 	return error;
2366 }
2367 
2368 int
udp_connect(struct socket * so,struct sockaddr * nam,struct proc * p)2369 udp_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
2370 {
2371 	struct inpcb *inp;
2372 	int error;
2373 
2374 	inp = sotoinpcb(so);
2375 	if (inp == NULL) {
2376 		return EINVAL;
2377 	}
2378 	inp_enter_bind_in_progress(so);
2379 
2380 	if (inp->inp_faddr.s_addr != INADDR_ANY) {
2381 		error = EISCONN;
2382 		goto done;
2383 	}
2384 
2385 	if (!(so->so_flags1 & SOF1_CONNECT_COUNTED)) {
2386 		so->so_flags1 |= SOF1_CONNECT_COUNTED;
2387 		INC_ATOMIC_INT64_LIM(net_api_stats.nas_socket_inet_dgram_connected);
2388 	}
2389 
2390 #if NECP
2391 #if FLOW_DIVERT
2392 	if (necp_socket_should_use_flow_divert(inp)) {
2393 		error = flow_divert_pcb_init(so);
2394 		if (error == 0) {
2395 			error = flow_divert_connect_out(so, nam, p);
2396 		}
2397 		goto done;
2398 	} else {
2399 		so->so_flags1 |= SOF1_FLOW_DIVERT_SKIP;
2400 	}
2401 #endif /* FLOW_DIVERT */
2402 #endif /* NECP */
2403 
2404 	error = in_pcbconnect(inp, nam, p, IFSCOPE_NONE, NULL);
2405 	if (error == 0) {
2406 #if NECP
2407 		/* Update NECP client with connected five-tuple */
2408 		if (!uuid_is_null(inp->necp_client_uuid)) {
2409 			socket_unlock(so, 0);
2410 			necp_client_assign_from_socket(so->last_pid, inp->necp_client_uuid, inp);
2411 			socket_lock(so, 0);
2412 		}
2413 #endif /* NECP */
2414 
2415 		soisconnected(so);
2416 		if (inp->inp_flowhash == 0) {
2417 			inp_calc_flowhash(inp);
2418 			ASSERT(inp->inp_flowhash != 0);
2419 		}
2420 		inp->inp_connect_timestamp = mach_continuous_time();
2421 	}
2422 done:
2423 	UDP_LOG_CONNECT(inp, error);
2424 
2425 	inp_exit_bind_in_progress(so);
2426 
2427 	return error;
2428 }
2429 
2430 int
udp_connectx_common(struct socket * so,int af,struct sockaddr * src,struct sockaddr * dst,struct proc * p,uint32_t ifscope,sae_associd_t aid,sae_connid_t * pcid,uint32_t flags,void * arg,uint32_t arglen,struct uio * uio,user_ssize_t * bytes_written)2431 udp_connectx_common(struct socket *so, int af, struct sockaddr *src, struct sockaddr *dst,
2432     struct proc *p, uint32_t ifscope, sae_associd_t aid, sae_connid_t *pcid,
2433     uint32_t flags, void *arg, uint32_t arglen,
2434     struct uio *uio, user_ssize_t *bytes_written)
2435 {
2436 #pragma unused(aid, flags, arg, arglen)
2437 	struct inpcb *inp = sotoinpcb(so);
2438 	int error = 0;
2439 	user_ssize_t datalen = 0;
2440 
2441 	if (inp == NULL) {
2442 		return EINVAL;
2443 	}
2444 
2445 	VERIFY(dst != NULL);
2446 
2447 	ASSERT(!(inp->inp_flags2 & INP2_CONNECT_IN_PROGRESS));
2448 	inp->inp_flags2 |= INP2_CONNECT_IN_PROGRESS;
2449 
2450 #if NECP
2451 	inp_update_necp_policy(inp, src, dst, ifscope);
2452 #endif /* NECP */
2453 
2454 	/* bind socket to the specified interface, if requested */
2455 	if (ifscope != IFSCOPE_NONE &&
2456 	    (error = inp_bindif(inp, ifscope, NULL)) != 0) {
2457 		goto done;
2458 	}
2459 
2460 	/* if source address and/or port is specified, bind to it */
2461 	if (src != NULL) {
2462 		error = sobindlock(so, src, 0); /* already locked */
2463 		if (error != 0) {
2464 			goto done;
2465 		}
2466 	}
2467 
2468 	switch (af) {
2469 	case AF_INET:
2470 		error = udp_connect(so, dst, p);
2471 		break;
2472 	case AF_INET6:
2473 		error = udp6_connect(so, dst, p);
2474 		break;
2475 	default:
2476 		VERIFY(0);
2477 		/* NOTREACHED */
2478 	}
2479 
2480 	if (error != 0) {
2481 		goto done;
2482 	}
2483 
2484 	/*
2485 	 * If there is data, copy it. DATA_IDEMPOTENT is ignored.
2486 	 * CONNECT_RESUME_ON_READ_WRITE is ignored.
2487 	 */
2488 	if (uio != NULL) {
2489 		socket_unlock(so, 0);
2490 
2491 		VERIFY(bytes_written != NULL);
2492 
2493 		datalen = uio_resid(uio);
2494 		error = so->so_proto->pr_usrreqs->pru_sosend(so, NULL,
2495 		    (uio_t)uio, NULL, NULL, 0);
2496 		socket_lock(so, 0);
2497 
2498 		/* If error returned is EMSGSIZE, for example, disconnect */
2499 		if (error == 0 || error == EWOULDBLOCK) {
2500 			*bytes_written = datalen - uio_resid(uio);
2501 		} else {
2502 			(void) so->so_proto->pr_usrreqs->pru_disconnectx(so,
2503 			    SAE_ASSOCID_ANY, SAE_CONNID_ANY);
2504 		}
2505 		/*
2506 		 * mask the EWOULDBLOCK error so that the caller
2507 		 * knows that atleast the connect was successful.
2508 		 */
2509 		if (error == EWOULDBLOCK) {
2510 			error = 0;
2511 		}
2512 	}
2513 
2514 	if (error == 0 && pcid != NULL) {
2515 		*pcid = 1;      /* there is only 1 connection for UDP */
2516 	}
2517 done:
2518 	inp->inp_flags2 &= ~INP2_CONNECT_IN_PROGRESS;
2519 	return error;
2520 }
2521 
2522 int
udp_connectx(struct socket * so,struct sockaddr * src,struct sockaddr * dst,struct proc * p,uint32_t ifscope,sae_associd_t aid,sae_connid_t * pcid,uint32_t flags,void * arg,uint32_t arglen,struct uio * uio,user_ssize_t * bytes_written)2523 udp_connectx(struct socket *so, struct sockaddr *src,
2524     struct sockaddr *dst, struct proc *p, uint32_t ifscope,
2525     sae_associd_t aid, sae_connid_t *pcid, uint32_t flags, void *arg,
2526     uint32_t arglen, struct uio *uio, user_ssize_t *bytes_written)
2527 {
2528 	return udp_connectx_common(so, AF_INET, src, dst,
2529 	           p, ifscope, aid, pcid, flags, arg, arglen, uio, bytes_written);
2530 }
2531 
2532 int
udp_detach(struct socket * so)2533 udp_detach(struct socket *so)
2534 {
2535 	struct inpcb *inp;
2536 
2537 	inp = sotoinpcb(so);
2538 	if (inp == NULL) {
2539 		panic("%s: so=%p null inp", __func__, so);
2540 		/* NOTREACHED */
2541 	}
2542 
2543 	/*
2544 	 * If this is a socket that does not want to wakeup the device
2545 	 * for it's traffic, the application might be waiting for
2546 	 * close to complete before going to sleep. Send a notification
2547 	 * for this kind of sockets
2548 	 */
2549 	if (so->so_options & SO_NOWAKEFROMSLEEP) {
2550 		socket_post_kev_msg_closed(so);
2551 	}
2552 
2553 	UDP_LOG_CONNECTION_SUMMARY(inp);
2554 
2555 	in_pcbdetach(inp);
2556 	inp->inp_state = INPCB_STATE_DEAD;
2557 	return 0;
2558 }
2559 
2560 int
udp_disconnect(struct socket * so)2561 udp_disconnect(struct socket *so)
2562 {
2563 	struct inpcb *inp;
2564 
2565 	inp = sotoinpcb(so);
2566 	if (inp == NULL) {
2567 		return EINVAL;
2568 	}
2569 	if (inp->inp_faddr.s_addr == INADDR_ANY) {
2570 		return ENOTCONN;
2571 	}
2572 
2573 	UDP_LOG_CONNECTION_SUMMARY(inp);
2574 
2575 	in_pcbdisconnect(inp);
2576 
2577 	/* reset flow controlled state, just in case */
2578 	inp_reset_fc_state(inp);
2579 
2580 	inp->inp_laddr.s_addr = INADDR_ANY;
2581 	so->so_state &= ~SS_ISCONNECTED;                /* XXX */
2582 	inp->inp_last_outifp = NULL;
2583 #if SKYWALK
2584 	if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
2585 		netns_set_ifnet(&inp->inp_netns_token, NULL);
2586 	}
2587 #endif /* SKYWALK */
2588 
2589 	return 0;
2590 }
2591 
2592 int
udp_disconnectx(struct socket * so,sae_associd_t aid,sae_connid_t cid)2593 udp_disconnectx(struct socket *so, sae_associd_t aid, sae_connid_t cid)
2594 {
2595 #pragma unused(cid)
2596 	if (aid != SAE_ASSOCID_ANY && aid != SAE_ASSOCID_ALL) {
2597 		return EINVAL;
2598 	}
2599 
2600 	return udp_disconnect(so);
2601 }
2602 
2603 int
udp_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * addr,struct mbuf * control,struct proc * p)2604 udp_send(struct socket *so, int flags, struct mbuf *m,
2605     struct sockaddr *addr, struct mbuf *control, struct proc *p)
2606 {
2607 #ifndef FLOW_DIVERT
2608 #pragma unused(flags)
2609 #endif /* !(FLOW_DIVERT) */
2610 	struct inpcb *inp;
2611 	int error;
2612 
2613 	inp = sotoinpcb(so);
2614 	if (inp == NULL) {
2615 		if (m != NULL) {
2616 			m_freem(m);
2617 		}
2618 		if (control != NULL) {
2619 			m_freem(control);
2620 		}
2621 		return EINVAL;
2622 	}
2623 
2624 #if NECP
2625 #if FLOW_DIVERT
2626 	if (necp_socket_should_use_flow_divert(inp)) {
2627 		/* Implicit connect */
2628 		return flow_divert_implicit_data_out(so, flags, m, addr,
2629 		           control, p);
2630 	} else {
2631 		so->so_flags1 |= SOF1_FLOW_DIVERT_SKIP;
2632 	}
2633 #endif /* FLOW_DIVERT */
2634 #endif /* NECP */
2635 
2636 	so_update_tx_data_stats(so, 1, m->m_pkthdr.len);
2637 
2638 	in_pcb_check_management_entitled(inp);
2639 	in_pcb_check_ultra_constrained_entitled(inp);
2640 
2641 #if SKYWALK
2642 	sk_protect_t protect = sk_async_transmit_protect();
2643 #endif /* SKYWALK */
2644 	error = udp_output(inp, m, addr, control, p);
2645 #if SKYWALK
2646 	sk_async_transmit_unprotect(protect);
2647 #endif /* SKYWALK */
2648 
2649 	return error;
2650 }
2651 
2652 int
udp_shutdown(struct socket * so)2653 udp_shutdown(struct socket *so)
2654 {
2655 	struct inpcb *inp;
2656 
2657 	inp = sotoinpcb(so);
2658 	if (inp == NULL) {
2659 		return EINVAL;
2660 	}
2661 	socantsendmore(so);
2662 	return 0;
2663 }
2664 
2665 int
udp_lock(struct socket * so,int refcount,void * debug)2666 udp_lock(struct socket *so, int refcount, void *debug)
2667 {
2668 	void *__single lr_saved;
2669 
2670 	if (debug == NULL) {
2671 		lr_saved = __unsafe_forge_single(void *, __builtin_return_address(0));
2672 	} else {
2673 		lr_saved = debug;
2674 	}
2675 
2676 	if (so->so_pcb != NULL) {
2677 		LCK_MTX_ASSERT(&((struct inpcb *)so->so_pcb)->inpcb_mtx,
2678 		    LCK_MTX_ASSERT_NOTOWNED);
2679 		lck_mtx_lock(&((struct inpcb *)so->so_pcb)->inpcb_mtx);
2680 	} else {
2681 		panic("%s: so=%p NO PCB! lr=%p lrh= %s", __func__,
2682 		    so, lr_saved, solockhistory_nr(so));
2683 		/* NOTREACHED */
2684 	}
2685 	if (refcount) {
2686 		so->so_usecount++;
2687 	}
2688 
2689 	so->lock_lr[so->next_lock_lr] = lr_saved;
2690 	so->next_lock_lr = (so->next_lock_lr + 1) % SO_LCKDBG_MAX;
2691 	return 0;
2692 }
2693 
2694 int
udp_unlock(struct socket * so,int refcount,void * debug)2695 udp_unlock(struct socket *so, int refcount, void *debug)
2696 {
2697 	void *__single lr_saved;
2698 
2699 	if (debug == NULL) {
2700 		lr_saved = __unsafe_forge_single(void *, __builtin_return_address(0));
2701 	} else {
2702 		lr_saved = debug;
2703 	}
2704 
2705 	if (refcount) {
2706 		VERIFY(so->so_usecount > 0);
2707 		so->so_usecount--;
2708 	}
2709 	if (so->so_pcb == NULL) {
2710 		panic("%s: so=%p NO PCB! lr=%p lrh= %s", __func__,
2711 		    so, lr_saved, solockhistory_nr(so));
2712 		/* NOTREACHED */
2713 	} else {
2714 		LCK_MTX_ASSERT(&((struct inpcb *)so->so_pcb)->inpcb_mtx,
2715 		    LCK_MTX_ASSERT_OWNED);
2716 		so->unlock_lr[so->next_unlock_lr] = lr_saved;
2717 		so->next_unlock_lr = (so->next_unlock_lr + 1) % SO_LCKDBG_MAX;
2718 		lck_mtx_unlock(&((struct inpcb *)so->so_pcb)->inpcb_mtx);
2719 	}
2720 	return 0;
2721 }
2722 
2723 lck_mtx_t *
udp_getlock(struct socket * so,int flags)2724 udp_getlock(struct socket *so, int flags)
2725 {
2726 #pragma unused(flags)
2727 	struct inpcb *__single inp = sotoinpcb(so);
2728 
2729 	if (so->so_pcb == NULL) {
2730 		panic("%s: so=%p NULL so_pcb lrh= %s", __func__,
2731 		    so, solockhistory_nr(so));
2732 		/* NOTREACHED */
2733 	}
2734 	return &inp->inpcb_mtx;
2735 }
2736 
2737 /*
2738  * UDP garbage collector callback (inpcb_timer_func_t).
2739  *
2740  * Returns > 0 to keep timer active.
2741  */
2742 static void
udp_gc(struct inpcbinfo * ipi)2743 udp_gc(struct inpcbinfo *ipi)
2744 {
2745 	struct inpcb *inp, *inpnxt;
2746 	struct socket *so;
2747 
2748 	if (lck_rw_try_lock_exclusive(&ipi->ipi_lock) == FALSE) {
2749 		if (udp_gc_done == TRUE) {
2750 			udp_gc_done = FALSE;
2751 			/* couldn't get the lock, must lock next time */
2752 			os_atomic_inc(&ipi->ipi_gc_req.intimer_fast, relaxed);
2753 			return;
2754 		}
2755 		lck_rw_lock_exclusive(&ipi->ipi_lock);
2756 	}
2757 
2758 	udp_gc_done = TRUE;
2759 
2760 	for (inp = udb.lh_first; inp != NULL; inp = inpnxt) {
2761 		inpnxt = inp->inp_list.le_next;
2762 
2763 		/*
2764 		 * Skip unless it's STOPUSING; garbage collector will
2765 		 * be triggered by in_pcb_checkstate() upon setting
2766 		 * wantcnt to that value.  If the PCB is already dead,
2767 		 * keep gc active to anticipate wantcnt changing.
2768 		 */
2769 		if (inp->inp_wantcnt != WNT_STOPUSING) {
2770 			continue;
2771 		}
2772 
2773 		/*
2774 		 * Skip if busy, no hurry for cleanup.  Keep gc active
2775 		 * and try the lock again during next round.
2776 		 */
2777 		if (!socket_try_lock(inp->inp_socket)) {
2778 			os_atomic_inc(&ipi->ipi_gc_req.intimer_fast, relaxed);
2779 			continue;
2780 		}
2781 
2782 		/*
2783 		 * Keep gc active unless usecount is 0.
2784 		 */
2785 		so = inp->inp_socket;
2786 		if (so->so_usecount == 0) {
2787 			if (inp->inp_state != INPCB_STATE_DEAD) {
2788 				if (SOCK_CHECK_DOM(so, PF_INET6)) {
2789 					in6_pcbdetach(inp);
2790 				} else {
2791 					in_pcbdetach(inp);
2792 				}
2793 			}
2794 			in_pcbdispose(inp);
2795 		} else {
2796 			socket_unlock(so, 0);
2797 			os_atomic_inc(&ipi->ipi_gc_req.intimer_fast, relaxed);
2798 		}
2799 	}
2800 	lck_rw_done(&ipi->ipi_lock);
2801 }
2802 
2803 static int
2804 udp_getstat SYSCTL_HANDLER_ARGS
2805 {
2806 #pragma unused(oidp, arg1, arg2)
2807 	if (req->oldptr == USER_ADDR_NULL) {
2808 		req->oldlen = (size_t)sizeof(struct udpstat);
2809 	}
2810 
2811 	return SYSCTL_OUT(req, &udpstat, MIN(sizeof(udpstat), req->oldlen));
2812 }
2813 
2814 void
udp_in_cksum_stats(u_int32_t len)2815 udp_in_cksum_stats(u_int32_t len)
2816 {
2817 	udpstat.udps_rcv_swcsum++;
2818 	udpstat.udps_rcv_swcsum_bytes += len;
2819 }
2820 
2821 void
udp_out_cksum_stats(u_int32_t len)2822 udp_out_cksum_stats(u_int32_t len)
2823 {
2824 	udpstat.udps_snd_swcsum++;
2825 	udpstat.udps_snd_swcsum_bytes += len;
2826 }
2827 
2828 void
udp_in6_cksum_stats(u_int32_t len)2829 udp_in6_cksum_stats(u_int32_t len)
2830 {
2831 	udpstat.udps_rcv6_swcsum++;
2832 	udpstat.udps_rcv6_swcsum_bytes += len;
2833 }
2834 
2835 void
udp_out6_cksum_stats(u_int32_t len)2836 udp_out6_cksum_stats(u_int32_t len)
2837 {
2838 	udpstat.udps_snd6_swcsum++;
2839 	udpstat.udps_snd6_swcsum_bytes += len;
2840 }
2841 
2842 /*
2843  * Checksum extended UDP header and data.
2844  */
2845 static int
udp_input_checksum(struct mbuf * m,struct udphdr * uh,int off,int ulen)2846 udp_input_checksum(struct mbuf *m, struct udphdr *uh, int off, int ulen)
2847 {
2848 	ifnet_ref_t ifp = m->m_pkthdr.rcvif;
2849 	struct ip *__single ip = mtod(m, struct ip *);
2850 	struct ipovly *__single ipov = (struct ipovly *)ip;
2851 
2852 	if (uh->uh_sum == 0) {
2853 		udpstat.udps_nosum++;
2854 		return 0;
2855 	}
2856 
2857 	/* ip_stripoptions() must have been called before we get here */
2858 	ASSERT((ip->ip_hl << 2) == sizeof(*ip));
2859 
2860 	if ((hwcksum_rx || (ifp->if_flags & IFF_LOOPBACK) ||
2861 	    (m->m_pkthdr.pkt_flags & PKTF_LOOP)) &&
2862 	    (m->m_pkthdr.csum_flags & CSUM_DATA_VALID)) {
2863 		if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
2864 			uh->uh_sum = m->m_pkthdr.csum_rx_val;
2865 		} else {
2866 			uint32_t sum = m->m_pkthdr.csum_rx_val;
2867 			uint32_t start = m->m_pkthdr.csum_rx_start;
2868 			int32_t trailer = (m_pktlen(m) - (off + ulen));
2869 
2870 			/*
2871 			 * Perform 1's complement adjustment of octets
2872 			 * that got included/excluded in the hardware-
2873 			 * calculated checksum value.  Ignore cases
2874 			 * where the value already includes the entire
2875 			 * IP header span, as the sum for those octets
2876 			 * would already be 0 by the time we get here;
2877 			 * IP has already performed its header checksum
2878 			 * checks.  If we do need to adjust, restore
2879 			 * the original fields in the IP header when
2880 			 * computing the adjustment value.  Also take
2881 			 * care of any trailing bytes and subtract out
2882 			 * their partial sum.
2883 			 */
2884 			ASSERT(trailer >= 0);
2885 			if ((m->m_pkthdr.csum_flags & CSUM_PARTIAL) &&
2886 			    ((start != 0 && start != off) || trailer != 0)) {
2887 				uint32_t swbytes = (uint32_t)trailer;
2888 
2889 				if (start < off) {
2890 					ip->ip_len += sizeof(*ip);
2891 #if BYTE_ORDER != BIG_ENDIAN
2892 					HTONS(ip->ip_len);
2893 					HTONS(ip->ip_off);
2894 #endif /* BYTE_ORDER != BIG_ENDIAN */
2895 				}
2896 				/* callee folds in sum */
2897 				sum = m_adj_sum16(m, start, off, ulen, sum);
2898 				if (off > start) {
2899 					swbytes += (off - start);
2900 				} else {
2901 					swbytes += (start - off);
2902 				}
2903 
2904 				if (start < off) {
2905 #if BYTE_ORDER != BIG_ENDIAN
2906 					NTOHS(ip->ip_off);
2907 					NTOHS(ip->ip_len);
2908 #endif /* BYTE_ORDER != BIG_ENDIAN */
2909 					ip->ip_len -= sizeof(*ip);
2910 				}
2911 
2912 				if (swbytes != 0) {
2913 					udp_in_cksum_stats(swbytes);
2914 				}
2915 				if (trailer != 0) {
2916 					m_adj(m, -trailer);
2917 				}
2918 			}
2919 
2920 			/* callee folds in sum */
2921 			uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
2922 			    ip->ip_dst.s_addr, sum + htonl(ulen + IPPROTO_UDP));
2923 		}
2924 		uh->uh_sum ^= 0xffff;
2925 	} else {
2926 		uint16_t ip_sum;
2927 		char b[9];
2928 
2929 		bcopy(ipov->ih_x1, b, sizeof(ipov->ih_x1));
2930 		bzero(ipov->ih_x1, sizeof(ipov->ih_x1));
2931 		ip_sum = ipov->ih_len;
2932 		ipov->ih_len = uh->uh_ulen;
2933 		uh->uh_sum = in_cksum(m, ulen + sizeof(struct ip));
2934 		bcopy(b, ipov->ih_x1, sizeof(ipov->ih_x1));
2935 		ipov->ih_len = ip_sum;
2936 
2937 		udp_in_cksum_stats(ulen);
2938 	}
2939 
2940 	if (uh->uh_sum != 0) {
2941 		udpstat.udps_badsum++;
2942 		IF_UDP_STATINC(ifp, badchksum);
2943 		return -1;
2944 	}
2945 
2946 	return 0;
2947 }
2948 
2949 void
udp_fill_keepalive_offload_frames(ifnet_t ifp,struct ifnet_keepalive_offload_frame * __counted_by (frames_array_count)frames_array,u_int32_t frames_array_count,size_t frame_data_offset,u_int32_t * used_frames_count)2950 udp_fill_keepalive_offload_frames(ifnet_t ifp,
2951     struct ifnet_keepalive_offload_frame *__counted_by(frames_array_count) frames_array,
2952     u_int32_t frames_array_count, size_t frame_data_offset,
2953     u_int32_t *used_frames_count)
2954 {
2955 	struct inpcb *inp;
2956 	inp_gen_t gencnt;
2957 	u_int32_t frame_index = *used_frames_count;
2958 
2959 	if (ifp == NULL || frames_array == NULL ||
2960 	    frames_array_count == 0 ||
2961 	    frame_index >= frames_array_count ||
2962 	    frame_data_offset >= IFNET_KEEPALIVE_OFFLOAD_FRAME_DATA_SIZE) {
2963 		return;
2964 	}
2965 
2966 	lck_rw_lock_shared(&udbinfo.ipi_lock);
2967 	gencnt = udbinfo.ipi_gencnt;
2968 	LIST_FOREACH(inp, udbinfo.ipi_listhead, inp_list) {
2969 		struct socket *so;
2970 		u_int8_t *data;
2971 		struct ifnet_keepalive_offload_frame *frame;
2972 		mbuf_ref_t m = NULL;
2973 
2974 		if (frame_index >= frames_array_count) {
2975 			break;
2976 		}
2977 
2978 		if (inp->inp_gencnt > gencnt ||
2979 		    inp->inp_state == INPCB_STATE_DEAD) {
2980 			continue;
2981 		}
2982 
2983 		if ((so = inp->inp_socket) == NULL ||
2984 		    (so->so_state & SS_DEFUNCT)) {
2985 			continue;
2986 		}
2987 		/*
2988 		 * check for keepalive offload flag without socket
2989 		 * lock to avoid a deadlock
2990 		 */
2991 		if (!(inp->inp_flags2 & INP2_KEEPALIVE_OFFLOAD)) {
2992 			continue;
2993 		}
2994 
2995 		udp_lock(so, 1, 0);
2996 		if (!(inp->inp_vflag & (INP_IPV4 | INP_IPV6))) {
2997 			udp_unlock(so, 1, 0);
2998 			continue;
2999 		}
3000 		if ((inp->inp_vflag & INP_IPV4) &&
3001 		    (inp->inp_laddr.s_addr == INADDR_ANY ||
3002 		    inp->inp_faddr.s_addr == INADDR_ANY)) {
3003 			udp_unlock(so, 1, 0);
3004 			continue;
3005 		}
3006 		if ((inp->inp_vflag & INP_IPV6) &&
3007 		    (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ||
3008 		    IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))) {
3009 			udp_unlock(so, 1, 0);
3010 			continue;
3011 		}
3012 		if (inp->inp_lport == 0 || inp->inp_fport == 0) {
3013 			udp_unlock(so, 1, 0);
3014 			continue;
3015 		}
3016 		if (inp->inp_last_outifp == NULL ||
3017 		    inp->inp_last_outifp->if_index != ifp->if_index) {
3018 			udp_unlock(so, 1, 0);
3019 			continue;
3020 		}
3021 		if ((inp->inp_vflag & INP_IPV4)) {
3022 			if ((frame_data_offset + sizeof(struct udpiphdr) +
3023 			    inp->inp_keepalive_datalen) >
3024 			    IFNET_KEEPALIVE_OFFLOAD_FRAME_DATA_SIZE) {
3025 				udp_unlock(so, 1, 0);
3026 				continue;
3027 			}
3028 			if ((sizeof(struct udpiphdr) +
3029 			    inp->inp_keepalive_datalen) > _MHLEN) {
3030 				udp_unlock(so, 1, 0);
3031 				continue;
3032 			}
3033 		} else {
3034 			if ((frame_data_offset + sizeof(struct ip6_hdr) +
3035 			    sizeof(struct udphdr) +
3036 			    inp->inp_keepalive_datalen) >
3037 			    IFNET_KEEPALIVE_OFFLOAD_FRAME_DATA_SIZE) {
3038 				udp_unlock(so, 1, 0);
3039 				continue;
3040 			}
3041 			if ((sizeof(struct ip6_hdr) + sizeof(struct udphdr) +
3042 			    inp->inp_keepalive_datalen) > _MHLEN) {
3043 				udp_unlock(so, 1, 0);
3044 				continue;
3045 			}
3046 		}
3047 		MGETHDR(m, M_WAIT, MT_HEADER);
3048 		if (m == NULL) {
3049 			udp_unlock(so, 1, 0);
3050 			continue;
3051 		}
3052 		/*
3053 		 * This inp has all the information that is needed to
3054 		 * generate an offload frame.
3055 		 */
3056 		if (inp->inp_vflag & INP_IPV4) {
3057 			struct ip *ip;
3058 			struct udphdr *udp;
3059 
3060 			frame = &frames_array[frame_index];
3061 			frame->length = (uint8_t)(frame_data_offset +
3062 			    sizeof(struct udpiphdr) +
3063 			    inp->inp_keepalive_datalen);
3064 			frame->ether_type =
3065 			    IFNET_KEEPALIVE_OFFLOAD_FRAME_ETHERTYPE_IPV4;
3066 			frame->interval = inp->inp_keepalive_interval;
3067 			switch (inp->inp_keepalive_type) {
3068 			case UDP_KEEPALIVE_OFFLOAD_TYPE_AIRPLAY:
3069 				frame->type =
3070 				    IFNET_KEEPALIVE_OFFLOAD_FRAME_AIRPLAY;
3071 				break;
3072 			default:
3073 				break;
3074 			}
3075 			data = mtod(m, u_int8_t *);
3076 			bzero(data, sizeof(struct udpiphdr));
3077 			ip = (__typeof__(ip))(void *)data;
3078 			udp = (__typeof__(udp))(void *) (data +
3079 			    sizeof(struct ip));
3080 			m->m_len = sizeof(struct udpiphdr);
3081 			data = data + sizeof(struct udpiphdr);
3082 			if (inp->inp_keepalive_datalen > 0 &&
3083 			    inp->inp_keepalive_data != NULL) {
3084 				bcopy(inp->inp_keepalive_data, data,
3085 				    inp->inp_keepalive_datalen);
3086 				m->m_len += inp->inp_keepalive_datalen;
3087 			}
3088 			m->m_pkthdr.len = m->m_len;
3089 
3090 			ip->ip_v = IPVERSION;
3091 			ip->ip_hl = (sizeof(struct ip) >> 2);
3092 			ip->ip_p = IPPROTO_UDP;
3093 			ip->ip_len = htons(sizeof(struct udpiphdr) +
3094 			    (u_short)inp->inp_keepalive_datalen);
3095 			ip->ip_ttl = inp->inp_ip_ttl;
3096 			ip->ip_tos |= (inp->inp_ip_tos & ~IPTOS_ECN_MASK);
3097 			ip->ip_src = inp->inp_laddr;
3098 			ip->ip_dst = inp->inp_faddr;
3099 			ip->ip_sum = in_cksum_hdr_opt(ip);
3100 
3101 			udp->uh_sport = inp->inp_lport;
3102 			udp->uh_dport = inp->inp_fport;
3103 			udp->uh_ulen = htons(sizeof(struct udphdr) +
3104 			    (u_short)inp->inp_keepalive_datalen);
3105 
3106 			if (!(inp->inp_flags & INP_UDP_NOCKSUM)) {
3107 				udp->uh_sum = in_pseudo(ip->ip_src.s_addr,
3108 				    ip->ip_dst.s_addr,
3109 				    htons(sizeof(struct udphdr) +
3110 				    (u_short)inp->inp_keepalive_datalen +
3111 				    IPPROTO_UDP));
3112 				m->m_pkthdr.csum_flags =
3113 				    (CSUM_UDP | CSUM_ZERO_INVERT);
3114 				m->m_pkthdr.csum_data = offsetof(struct udphdr,
3115 				    uh_sum);
3116 			}
3117 			m->m_pkthdr.pkt_proto = IPPROTO_UDP;
3118 			in_delayed_cksum(m);
3119 			bcopy(m_mtod_current(m), frame->data + frame_data_offset,
3120 			    m->m_len);
3121 		} else {
3122 			struct ip6_hdr *ip6;
3123 			struct udphdr *udp6;
3124 
3125 			VERIFY(inp->inp_vflag & INP_IPV6);
3126 			frame = &frames_array[frame_index];
3127 			frame->length = (uint8_t)(frame_data_offset +
3128 			    sizeof(struct ip6_hdr) +
3129 			    sizeof(struct udphdr) +
3130 			    inp->inp_keepalive_datalen);
3131 			frame->ether_type =
3132 			    IFNET_KEEPALIVE_OFFLOAD_FRAME_ETHERTYPE_IPV6;
3133 			frame->interval = inp->inp_keepalive_interval;
3134 			switch (inp->inp_keepalive_type) {
3135 			case UDP_KEEPALIVE_OFFLOAD_TYPE_AIRPLAY:
3136 				frame->type =
3137 				    IFNET_KEEPALIVE_OFFLOAD_FRAME_AIRPLAY;
3138 				break;
3139 			default:
3140 				break;
3141 			}
3142 			data = mtod(m, u_int8_t *);
3143 			bzero(data, sizeof(struct ip6_hdr) + sizeof(struct udphdr));
3144 			ip6 = (__typeof__(ip6))(void *)data;
3145 			udp6 = (__typeof__(udp6))(void *)(data +
3146 			    sizeof(struct ip6_hdr));
3147 			m->m_len = sizeof(struct ip6_hdr) +
3148 			    sizeof(struct udphdr);
3149 			data = data + (sizeof(struct ip6_hdr) +
3150 			    sizeof(struct udphdr));
3151 			if (inp->inp_keepalive_datalen > 0 &&
3152 			    inp->inp_keepalive_data != NULL) {
3153 				bcopy(inp->inp_keepalive_data, data,
3154 				    inp->inp_keepalive_datalen);
3155 				m->m_len += inp->inp_keepalive_datalen;
3156 			}
3157 			m->m_pkthdr.len = m->m_len;
3158 			ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK;
3159 			ip6->ip6_flow = ip6->ip6_flow & ~IPV6_FLOW_ECN_MASK;
3160 			ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
3161 			ip6->ip6_vfc |= IPV6_VERSION;
3162 			ip6->ip6_nxt = IPPROTO_UDP;
3163 			ip6->ip6_hlim = (uint8_t)ip6_defhlim;
3164 			ip6->ip6_plen = htons(sizeof(struct udphdr) +
3165 			    (u_short)inp->inp_keepalive_datalen);
3166 			ip6->ip6_src = inp->in6p_laddr;
3167 			if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) {
3168 				ip6->ip6_src.s6_addr16[1] = 0;
3169 			}
3170 
3171 			ip6->ip6_dst = inp->in6p_faddr;
3172 			if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
3173 				ip6->ip6_dst.s6_addr16[1] = 0;
3174 			}
3175 
3176 			udp6->uh_sport = inp->in6p_lport;
3177 			udp6->uh_dport = inp->in6p_fport;
3178 			udp6->uh_ulen = htons(sizeof(struct udphdr) +
3179 			    (u_short)inp->inp_keepalive_datalen);
3180 			if (!(inp->inp_flags & INP_UDP_NOCKSUM)) {
3181 				udp6->uh_sum = in6_pseudo(&ip6->ip6_src,
3182 				    &ip6->ip6_dst,
3183 				    htonl(sizeof(struct udphdr) +
3184 				    (u_short)inp->inp_keepalive_datalen +
3185 				    IPPROTO_UDP));
3186 				m->m_pkthdr.csum_flags =
3187 				    (CSUM_UDPIPV6 | CSUM_ZERO_INVERT);
3188 				m->m_pkthdr.csum_data = offsetof(struct udphdr,
3189 				    uh_sum);
3190 			}
3191 			m->m_pkthdr.pkt_proto = IPPROTO_UDP;
3192 			in6_delayed_cksum(m);
3193 			bcopy(m_mtod_current(m), frame->data + frame_data_offset, m->m_len);
3194 		}
3195 		if (m != NULL) {
3196 			m_freem(m);
3197 			m = NULL;
3198 		}
3199 		frame_index++;
3200 		udp_unlock(so, 1, 0);
3201 	}
3202 	lck_rw_done(&udbinfo.ipi_lock);
3203 	*used_frames_count = frame_index;
3204 }
3205 
3206 int
udp_defunct(struct socket * so)3207 udp_defunct(struct socket *so)
3208 {
3209 	struct ip_moptions *__single imo;
3210 	struct inpcb *__single inp;
3211 
3212 	inp = sotoinpcb(so);
3213 	if (inp == NULL) {
3214 		return EINVAL;
3215 	}
3216 
3217 	imo = inp->inp_moptions;
3218 	if (imo != NULL) {
3219 		struct proc *p = current_proc();
3220 
3221 		SODEFUNCTLOG("%s[%d, %s]: defuncting so 0x%llu drop multicast memberships",
3222 		    __func__, proc_pid(p), proc_best_name(p),
3223 		    so->so_gencnt);
3224 
3225 		inp->inp_moptions = NULL;
3226 
3227 		IMO_REMREF(imo);
3228 	}
3229 
3230 	return 0;
3231 }
3232