]> git.saurik.com Git - apple/xnu.git/blob - bsd/netinet/udp_usrreq.c
0c4945e79b07f3e149b2767993493c677a7d39a9
[apple/xnu.git] / bsd / netinet / udp_usrreq.c
1 /*
2 * Copyright (c) 2000-2017 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
80 #include <net/if.h>
81 #include <net/if_types.h>
82 #include <net/route.h>
83 #include <net/dlil.h>
84 #include <net/net_api_stats.h>
85
86 #include <netinet/in.h>
87 #include <netinet/in_systm.h>
88 #include <netinet/in_tclass.h>
89 #include <netinet/ip.h>
90 #if INET6
91 #include <netinet/ip6.h>
92 #endif /* INET6 */
93 #include <netinet/in_pcb.h>
94 #include <netinet/in_var.h>
95 #include <netinet/ip_var.h>
96 #if INET6
97 #include <netinet6/in6_pcb.h>
98 #include <netinet6/ip6_var.h>
99 #include <netinet6/udp6_var.h>
100 #endif /* INET6 */
101 #include <netinet/ip_icmp.h>
102 #include <netinet/icmp_var.h>
103 #include <netinet/udp.h>
104 #include <netinet/udp_var.h>
105 #include <sys/kdebug.h>
106
107 #if IPSEC
108 #include <netinet6/ipsec.h>
109 #include <netinet6/esp.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 #define DBG_LAYER_IN_BEG NETDBG_CODE(DBG_NETUDP, 0)
123 #define DBG_LAYER_IN_END NETDBG_CODE(DBG_NETUDP, 2)
124 #define DBG_LAYER_OUT_BEG NETDBG_CODE(DBG_NETUDP, 1)
125 #define DBG_LAYER_OUT_END NETDBG_CODE(DBG_NETUDP, 3)
126 #define DBG_FNC_UDP_INPUT NETDBG_CODE(DBG_NETUDP, (5 << 8))
127 #define DBG_FNC_UDP_OUTPUT NETDBG_CODE(DBG_NETUDP, (6 << 8) | 1)
128
129 /*
130 * UDP protocol implementation.
131 * Per RFC 768, August, 1980.
132 */
133 #ifndef COMPAT_42
134 static int udpcksum = 1;
135 #else
136 static int udpcksum = 0; /* XXX */
137 #endif
138 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum,
139 CTLFLAG_RW | CTLFLAG_LOCKED, &udpcksum, 0, "");
140
141 int udp_log_in_vain = 0;
142 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW | CTLFLAG_LOCKED,
143 &udp_log_in_vain, 0, "Log all incoming UDP packets");
144
145 static int blackhole = 0;
146 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW | CTLFLAG_LOCKED,
147 &blackhole, 0, "Do not send port unreachables for refused connects");
148
149 struct inpcbhead udb; /* from udp_var.h */
150 #define udb6 udb /* for KAME src sync over BSD*'s */
151 struct inpcbinfo udbinfo;
152
153 #ifndef UDBHASHSIZE
154 #define UDBHASHSIZE 16
155 #endif
156
157 /* Garbage collection performed during most recent udp_gc() run */
158 static boolean_t udp_gc_done = FALSE;
159
160 #if IPFIREWALL
161 extern int fw_verbose;
162 extern void ipfwsyslog(int level, const char *format, ...);
163 extern void ipfw_stealth_stats_incr_udp(void);
164
165 /* Apple logging, log to ipfw.log */
166 #define log_in_vain_log(a) { \
167 if ((udp_log_in_vain == 3) && (fw_verbose == 2)) { \
168 ipfwsyslog a; \
169 } else if ((udp_log_in_vain == 4) && (fw_verbose == 2)) { \
170 ipfw_stealth_stats_incr_udp(); \
171 } else { \
172 log a; \
173 } \
174 }
175 #else /* !IPFIREWALL */
176 #define log_in_vain_log(a) { log a; }
177 #endif /* !IPFIREWALL */
178
179 static int udp_getstat SYSCTL_HANDLER_ARGS;
180 struct udpstat udpstat; /* from udp_var.h */
181 SYSCTL_PROC(_net_inet_udp, UDPCTL_STATS, stats,
182 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED,
183 0, 0, udp_getstat, "S,udpstat",
184 "UDP statistics (struct udpstat, netinet/udp_var.h)");
185
186 SYSCTL_INT(_net_inet_udp, OID_AUTO, pcbcount,
187 CTLFLAG_RD | CTLFLAG_LOCKED, &udbinfo.ipi_count, 0,
188 "Number of active PCBs");
189
190 __private_extern__ int udp_use_randomport = 1;
191 SYSCTL_INT(_net_inet_udp, OID_AUTO, randomize_ports,
192 CTLFLAG_RW | CTLFLAG_LOCKED, &udp_use_randomport, 0,
193 "Randomize UDP port numbers");
194
195 #if INET6
196 struct udp_in6 {
197 struct sockaddr_in6 uin6_sin;
198 u_char uin6_init_done : 1;
199 };
200 struct udp_ip6 {
201 struct ip6_hdr uip6_ip6;
202 u_char uip6_init_done : 1;
203 };
204
205 int udp_abort(struct socket *);
206 int udp_attach(struct socket *, int, struct proc *);
207 int udp_bind(struct socket *, struct sockaddr *, struct proc *);
208 int udp_connect(struct socket *, struct sockaddr *, struct proc *);
209 int udp_connectx(struct socket *, struct sockaddr *,
210 struct sockaddr *, struct proc *, uint32_t, sae_associd_t,
211 sae_connid_t *, uint32_t, void *, uint32_t, struct uio *, user_ssize_t *);
212 int udp_detach(struct socket *);
213 int udp_disconnect(struct socket *);
214 int udp_disconnectx(struct socket *, sae_associd_t, sae_connid_t);
215 int udp_send(struct socket *, int, struct mbuf *, struct sockaddr *,
216 struct mbuf *, struct proc *);
217 static void udp_append(struct inpcb *, struct ip *, struct mbuf *, int,
218 struct sockaddr_in *, struct udp_in6 *, struct udp_ip6 *, struct ifnet *);
219 #else /* !INET6 */
220 static void udp_append(struct inpcb *, struct ip *, struct mbuf *, int,
221 struct sockaddr_in *, struct ifnet *);
222 #endif /* !INET6 */
223 static int udp_input_checksum(struct mbuf *, struct udphdr *, int, int);
224 int udp_output(struct inpcb *, struct mbuf *, struct sockaddr *,
225 struct mbuf *, struct proc *);
226 static void ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip);
227 static void udp_gc(struct inpcbinfo *);
228
229 struct pr_usrreqs udp_usrreqs = {
230 .pru_abort = udp_abort,
231 .pru_attach = udp_attach,
232 .pru_bind = udp_bind,
233 .pru_connect = udp_connect,
234 .pru_connectx = udp_connectx,
235 .pru_control = in_control,
236 .pru_detach = udp_detach,
237 .pru_disconnect = udp_disconnect,
238 .pru_disconnectx = udp_disconnectx,
239 .pru_peeraddr = in_getpeeraddr,
240 .pru_send = udp_send,
241 .pru_shutdown = udp_shutdown,
242 .pru_sockaddr = in_getsockaddr,
243 .pru_sosend = sosend,
244 .pru_soreceive = soreceive,
245 .pru_soreceive_list = soreceive_list,
246 };
247
248 void
249 udp_init(struct protosw *pp, struct domain *dp)
250 {
251 #pragma unused(dp)
252 static int udp_initialized = 0;
253 vm_size_t str_size;
254 struct inpcbinfo *pcbinfo;
255
256 VERIFY((pp->pr_flags & (PR_INITIALIZED|PR_ATTACHED)) == PR_ATTACHED);
257
258 if (udp_initialized)
259 return;
260 udp_initialized = 1;
261
262 LIST_INIT(&udb);
263 udbinfo.ipi_listhead = &udb;
264 udbinfo.ipi_hashbase = hashinit(UDBHASHSIZE, M_PCB,
265 &udbinfo.ipi_hashmask);
266 udbinfo.ipi_porthashbase = hashinit(UDBHASHSIZE, M_PCB,
267 &udbinfo.ipi_porthashmask);
268 str_size = (vm_size_t) sizeof (struct inpcb);
269 udbinfo.ipi_zone = zinit(str_size, 80000*str_size, 8192, "udpcb");
270
271 pcbinfo = &udbinfo;
272 /*
273 * allocate lock group attribute and group for udp pcb mutexes
274 */
275 pcbinfo->ipi_lock_grp_attr = lck_grp_attr_alloc_init();
276 pcbinfo->ipi_lock_grp = lck_grp_alloc_init("udppcb",
277 pcbinfo->ipi_lock_grp_attr);
278 pcbinfo->ipi_lock_attr = lck_attr_alloc_init();
279 if ((pcbinfo->ipi_lock = lck_rw_alloc_init(pcbinfo->ipi_lock_grp,
280 pcbinfo->ipi_lock_attr)) == NULL) {
281 panic("%s: unable to allocate PCB lock\n", __func__);
282 /* NOTREACHED */
283 }
284
285 udbinfo.ipi_gc = udp_gc;
286 in_pcbinfo_attach(&udbinfo);
287 }
288
289 void
290 udp_input(struct mbuf *m, int iphlen)
291 {
292 struct ip *ip;
293 struct udphdr *uh;
294 struct inpcb *inp;
295 struct mbuf *opts = NULL;
296 int len, isbroadcast;
297 struct ip save_ip;
298 struct sockaddr *append_sa;
299 struct inpcbinfo *pcbinfo = &udbinfo;
300 struct sockaddr_in udp_in;
301 struct ip_moptions *imo = NULL;
302 int foundmembership = 0, ret = 0;
303 #if INET6
304 struct udp_in6 udp_in6;
305 struct udp_ip6 udp_ip6;
306 #endif /* INET6 */
307 struct ifnet *ifp = m->m_pkthdr.rcvif;
308 boolean_t cell = IFNET_IS_CELLULAR(ifp);
309 boolean_t wifi = (!cell && IFNET_IS_WIFI(ifp));
310 boolean_t wired = (!wifi && IFNET_IS_WIRED(ifp));
311
312 bzero(&udp_in, sizeof (udp_in));
313 udp_in.sin_len = sizeof (struct sockaddr_in);
314 udp_in.sin_family = AF_INET;
315 #if INET6
316 bzero(&udp_in6, sizeof (udp_in6));
317 udp_in6.uin6_sin.sin6_len = sizeof (struct sockaddr_in6);
318 udp_in6.uin6_sin.sin6_family = AF_INET6;
319 #endif /* INET6 */
320
321 udpstat.udps_ipackets++;
322
323 KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_START, 0, 0, 0, 0, 0);
324
325 /* Expect 32-bit aligned data pointer on strict-align platforms */
326 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
327
328 /*
329 * Strip IP options, if any; should skip this,
330 * make available to user, and use on returned packets,
331 * but we don't yet have a way to check the checksum
332 * with options still present.
333 */
334 if (iphlen > sizeof (struct ip)) {
335 ip_stripoptions(m);
336 iphlen = sizeof (struct ip);
337 }
338
339 /*
340 * Get IP and UDP header together in first mbuf.
341 */
342 ip = mtod(m, struct ip *);
343 if (m->m_len < iphlen + sizeof (struct udphdr)) {
344 m = m_pullup(m, iphlen + sizeof (struct udphdr));
345 if (m == NULL) {
346 udpstat.udps_hdrops++;
347 KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END,
348 0, 0, 0, 0, 0);
349 return;
350 }
351 ip = mtod(m, struct ip *);
352 }
353 uh = (struct udphdr *)(void *)((caddr_t)ip + iphlen);
354
355 /* destination port of 0 is illegal, based on RFC768. */
356 if (uh->uh_dport == 0) {
357 IF_UDP_STATINC(ifp, port0);
358 goto bad;
359 }
360
361 KERNEL_DEBUG(DBG_LAYER_IN_BEG, uh->uh_dport, uh->uh_sport,
362 ip->ip_src.s_addr, ip->ip_dst.s_addr, uh->uh_ulen);
363
364 /*
365 * Make mbuf data length reflect UDP length.
366 * If not enough data to reflect UDP length, drop.
367 */
368 len = ntohs((u_short)uh->uh_ulen);
369 if (ip->ip_len != len) {
370 if (len > ip->ip_len || len < sizeof (struct udphdr)) {
371 udpstat.udps_badlen++;
372 IF_UDP_STATINC(ifp, badlength);
373 goto bad;
374 }
375 m_adj(m, len - ip->ip_len);
376 /* ip->ip_len = len; */
377 }
378 /*
379 * Save a copy of the IP header in case we want restore it
380 * for sending an ICMP error message in response.
381 */
382 save_ip = *ip;
383
384 /*
385 * Checksum extended UDP header and data.
386 */
387 if (udp_input_checksum(m, uh, iphlen, len))
388 goto bad;
389
390 isbroadcast = in_broadcast(ip->ip_dst, ifp);
391
392 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || isbroadcast) {
393 int reuse_sock = 0, mcast_delivered = 0;
394
395 lck_rw_lock_shared(pcbinfo->ipi_lock);
396 /*
397 * Deliver a multicast or broadcast datagram to *all* sockets
398 * for which the local and remote addresses and ports match
399 * those of the incoming datagram. This allows more than
400 * one process to receive multi/broadcasts on the same port.
401 * (This really ought to be done for unicast datagrams as
402 * well, but that would cause problems with existing
403 * applications that open both address-specific sockets and
404 * a wildcard socket listening to the same port -- they would
405 * end up receiving duplicates of every unicast datagram.
406 * Those applications open the multiple sockets to overcome an
407 * inadequacy of the UDP socket interface, but for backwards
408 * compatibility we avoid the problem here rather than
409 * fixing the interface. Maybe 4.5BSD will remedy this?)
410 */
411
412 /*
413 * Construct sockaddr format source address.
414 */
415 udp_in.sin_port = uh->uh_sport;
416 udp_in.sin_addr = ip->ip_src;
417 /*
418 * Locate pcb(s) for datagram.
419 * (Algorithm copied from raw_intr().)
420 */
421 #if INET6
422 udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
423 #endif /* INET6 */
424 LIST_FOREACH(inp, &udb, inp_list) {
425 #if IPSEC
426 int skipit;
427 #endif /* IPSEC */
428
429 if (inp->inp_socket == NULL)
430 continue;
431 if (inp != sotoinpcb(inp->inp_socket)) {
432 panic("%s: bad so back ptr inp=%p\n",
433 __func__, inp);
434 /* NOTREACHED */
435 }
436 #if INET6
437 if ((inp->inp_vflag & INP_IPV4) == 0)
438 continue;
439 #endif /* INET6 */
440 if (inp_restricted_recv(inp, ifp))
441 continue;
442
443 if ((inp->inp_moptions == NULL) &&
444 (ntohl(ip->ip_dst.s_addr) !=
445 INADDR_ALLHOSTS_GROUP) && (isbroadcast == 0))
446 continue;
447
448 if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) ==
449 WNT_STOPUSING)
450 continue;
451
452 udp_lock(inp->inp_socket, 1, 0);
453
454 if (in_pcb_checkstate(inp, WNT_RELEASE, 1) ==
455 WNT_STOPUSING) {
456 udp_unlock(inp->inp_socket, 1, 0);
457 continue;
458 }
459
460 if (inp->inp_lport != uh->uh_dport) {
461 udp_unlock(inp->inp_socket, 1, 0);
462 continue;
463 }
464 if (inp->inp_laddr.s_addr != INADDR_ANY) {
465 if (inp->inp_laddr.s_addr !=
466 ip->ip_dst.s_addr) {
467 udp_unlock(inp->inp_socket, 1, 0);
468 continue;
469 }
470 }
471 if (inp->inp_faddr.s_addr != INADDR_ANY) {
472 if (inp->inp_faddr.s_addr !=
473 ip->ip_src.s_addr ||
474 inp->inp_fport != uh->uh_sport) {
475 udp_unlock(inp->inp_socket, 1, 0);
476 continue;
477 }
478 }
479
480 if (isbroadcast == 0 && (ntohl(ip->ip_dst.s_addr) !=
481 INADDR_ALLHOSTS_GROUP)) {
482 struct sockaddr_in group;
483 int blocked;
484
485 if ((imo = inp->inp_moptions) == NULL) {
486 udp_unlock(inp->inp_socket, 1, 0);
487 continue;
488 }
489 IMO_LOCK(imo);
490
491 bzero(&group, sizeof (struct sockaddr_in));
492 group.sin_len = sizeof (struct sockaddr_in);
493 group.sin_family = AF_INET;
494 group.sin_addr = ip->ip_dst;
495
496 blocked = imo_multi_filter(imo, ifp,
497 &group, &udp_in);
498 if (blocked == MCAST_PASS)
499 foundmembership = 1;
500
501 IMO_UNLOCK(imo);
502 if (!foundmembership) {
503 udp_unlock(inp->inp_socket, 1, 0);
504 if (blocked == MCAST_NOTSMEMBER ||
505 blocked == MCAST_MUTED)
506 udpstat.udps_filtermcast++;
507 continue;
508 }
509 foundmembership = 0;
510 }
511
512 reuse_sock = (inp->inp_socket->so_options &
513 (SO_REUSEPORT|SO_REUSEADDR));
514
515 #if NECP
516 skipit = 0;
517 if (!necp_socket_is_allowed_to_send_recv_v4(inp,
518 uh->uh_dport, uh->uh_sport, &ip->ip_dst,
519 &ip->ip_src, ifp, NULL, NULL)) {
520 /* do not inject data to pcb */
521 skipit = 1;
522 }
523 if (skipit == 0)
524 #endif /* NECP */
525 {
526 struct mbuf *n = NULL;
527
528 if (reuse_sock)
529 n = m_copy(m, 0, M_COPYALL);
530 #if INET6
531 udp_append(inp, ip, m,
532 iphlen + sizeof (struct udphdr),
533 &udp_in, &udp_in6, &udp_ip6, ifp);
534 #else /* !INET6 */
535 udp_append(inp, ip, m,
536 iphlen + sizeof (struct udphdr),
537 &udp_in, ifp);
538 #endif /* !INET6 */
539 mcast_delivered++;
540
541 m = n;
542 }
543 udp_unlock(inp->inp_socket, 1, 0);
544
545 /*
546 * Don't look for additional matches if this one does
547 * not have either the SO_REUSEPORT or SO_REUSEADDR
548 * socket options set. This heuristic avoids searching
549 * through all pcbs in the common case of a non-shared
550 * port. It assumes that an application will never
551 * clear these options after setting them.
552 */
553 if (reuse_sock == 0 || m == NULL)
554 break;
555
556 /*
557 * Expect 32-bit aligned data pointer on strict-align
558 * platforms.
559 */
560 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
561 /*
562 * Recompute IP and UDP header pointers for new mbuf
563 */
564 ip = mtod(m, struct ip *);
565 uh = (struct udphdr *)(void *)((caddr_t)ip + iphlen);
566 }
567 lck_rw_done(pcbinfo->ipi_lock);
568
569 if (mcast_delivered == 0) {
570 /*
571 * No matching pcb found; discard datagram.
572 * (No need to send an ICMP Port Unreachable
573 * for a broadcast or multicast datgram.)
574 */
575 udpstat.udps_noportbcast++;
576 IF_UDP_STATINC(ifp, port_unreach);
577 goto bad;
578 }
579
580 /* free the extra copy of mbuf or skipped by IPSec */
581 if (m != NULL)
582 m_freem(m);
583 KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0);
584 return;
585 }
586
587 #if IPSEC
588 /*
589 * UDP to port 4500 with a payload where the first four bytes are
590 * not zero is a UDP encapsulated IPSec packet. Packets where
591 * the payload is one byte and that byte is 0xFF are NAT keepalive
592 * packets. Decapsulate the ESP packet and carry on with IPSec input
593 * or discard the NAT keep-alive.
594 */
595 if (ipsec_bypass == 0 && (esp_udp_encap_port & 0xFFFF) != 0 &&
596 uh->uh_dport == ntohs((u_short)esp_udp_encap_port)) {
597 int payload_len = len - sizeof (struct udphdr) > 4 ? 4 :
598 len - sizeof (struct udphdr);
599
600 if (m->m_len < iphlen + sizeof (struct udphdr) + payload_len) {
601 if ((m = m_pullup(m, iphlen + sizeof (struct udphdr) +
602 payload_len)) == NULL) {
603 udpstat.udps_hdrops++;
604 KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END,
605 0, 0, 0, 0, 0);
606 return;
607 }
608 /*
609 * Expect 32-bit aligned data pointer on strict-align
610 * platforms.
611 */
612 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
613
614 ip = mtod(m, struct ip *);
615 uh = (struct udphdr *)(void *)((caddr_t)ip + iphlen);
616 }
617 /* Check for NAT keepalive packet */
618 if (payload_len == 1 && *(u_int8_t *)
619 ((caddr_t)uh + sizeof (struct udphdr)) == 0xFF) {
620 m_freem(m);
621 KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END,
622 0, 0, 0, 0, 0);
623 return;
624 } else if (payload_len == 4 && *(u_int32_t *)(void *)
625 ((caddr_t)uh + sizeof (struct udphdr)) != 0) {
626 /* UDP encapsulated IPSec packet to pass through NAT */
627 KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END,
628 0, 0, 0, 0, 0);
629 /* preserve the udp header */
630 esp4_input(m, iphlen + sizeof (struct udphdr));
631 return;
632 }
633 }
634 #endif /* IPSEC */
635
636 /*
637 * Locate pcb for datagram.
638 */
639 inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
640 ip->ip_dst, uh->uh_dport, 1, ifp);
641 if (inp == NULL) {
642 IF_UDP_STATINC(ifp, port_unreach);
643
644 if (udp_log_in_vain) {
645 char buf[MAX_IPv4_STR_LEN];
646 char buf2[MAX_IPv4_STR_LEN];
647
648 /* check src and dst address */
649 if (udp_log_in_vain < 3) {
650 log(LOG_INFO, "Connection attempt to "
651 "UDP %s:%d from %s:%d\n", inet_ntop(AF_INET,
652 &ip->ip_dst, buf, sizeof (buf)),
653 ntohs(uh->uh_dport), inet_ntop(AF_INET,
654 &ip->ip_src, buf2, sizeof (buf2)),
655 ntohs(uh->uh_sport));
656 } else if (!(m->m_flags & (M_BCAST | M_MCAST)) &&
657 ip->ip_dst.s_addr != ip->ip_src.s_addr) {
658 log_in_vain_log((LOG_INFO,
659 "Stealth Mode connection attempt to "
660 "UDP %s:%d from %s:%d\n", inet_ntop(AF_INET,
661 &ip->ip_dst, buf, sizeof (buf)),
662 ntohs(uh->uh_dport), inet_ntop(AF_INET,
663 &ip->ip_src, buf2, sizeof (buf2)),
664 ntohs(uh->uh_sport)))
665 }
666 }
667 udpstat.udps_noport++;
668 if (m->m_flags & (M_BCAST | M_MCAST)) {
669 udpstat.udps_noportbcast++;
670 goto bad;
671 }
672 #if ICMP_BANDLIM
673 if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
674 goto bad;
675 #endif /* ICMP_BANDLIM */
676 if (blackhole)
677 if (ifp && ifp->if_type != IFT_LOOP)
678 goto bad;
679 *ip = save_ip;
680 ip->ip_len += iphlen;
681 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
682 KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0);
683 return;
684 }
685 udp_lock(inp->inp_socket, 1, 0);
686
687 if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
688 udp_unlock(inp->inp_socket, 1, 0);
689 IF_UDP_STATINC(ifp, cleanup);
690 goto bad;
691 }
692 #if NECP
693 if (!necp_socket_is_allowed_to_send_recv_v4(inp, uh->uh_dport,
694 uh->uh_sport, &ip->ip_dst, &ip->ip_src, ifp, NULL, NULL)) {
695 udp_unlock(inp->inp_socket, 1, 0);
696 IF_UDP_STATINC(ifp, badipsec);
697 goto bad;
698 }
699 #endif /* NECP */
700
701 /*
702 * Construct sockaddr format source address.
703 * Stuff source address and datagram in user buffer.
704 */
705 udp_in.sin_port = uh->uh_sport;
706 udp_in.sin_addr = ip->ip_src;
707 if ((inp->inp_flags & INP_CONTROLOPTS) != 0 ||
708 (inp->inp_socket->so_options & SO_TIMESTAMP) != 0 ||
709 (inp->inp_socket->so_options & SO_TIMESTAMP_MONOTONIC) != 0) {
710 #if INET6
711 if (inp->inp_vflag & INP_IPV6) {
712 int savedflags;
713
714 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
715 savedflags = inp->inp_flags;
716 inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
717 ret = ip6_savecontrol(inp, m, &opts);
718 inp->inp_flags = savedflags;
719 } else
720 #endif /* INET6 */
721 {
722 ret = ip_savecontrol(inp, &opts, ip, m);
723 }
724 if (ret != 0) {
725 udp_unlock(inp->inp_socket, 1, 0);
726 goto bad;
727 }
728 }
729 m_adj(m, iphlen + sizeof (struct udphdr));
730
731 KERNEL_DEBUG(DBG_LAYER_IN_END, uh->uh_dport, uh->uh_sport,
732 save_ip.ip_src.s_addr, save_ip.ip_dst.s_addr, uh->uh_ulen);
733
734 #if INET6
735 if (inp->inp_vflag & INP_IPV6) {
736 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
737 append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
738 } else
739 #endif /* INET6 */
740 {
741 append_sa = (struct sockaddr *)&udp_in;
742 }
743 if (nstat_collect) {
744 INP_ADD_STAT(inp, cell, wifi, wired, rxpackets, 1);
745 INP_ADD_STAT(inp, cell, wifi, wired, rxbytes, m->m_pkthdr.len);
746 inp_set_activity_bitmap(inp);
747 }
748 so_recv_data_stat(inp->inp_socket, m, 0);
749 if (sbappendaddr(&inp->inp_socket->so_rcv, append_sa,
750 m, opts, NULL) == 0) {
751 udpstat.udps_fullsock++;
752 } else {
753 sorwakeup(inp->inp_socket);
754 }
755 udp_unlock(inp->inp_socket, 1, 0);
756 KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0);
757 return;
758 bad:
759 m_freem(m);
760 if (opts)
761 m_freem(opts);
762 KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0);
763 }
764
765 #if INET6
766 static void
767 ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip)
768 {
769 bzero(ip6, sizeof (*ip6));
770
771 ip6->ip6_vfc = IPV6_VERSION;
772 ip6->ip6_plen = ip->ip_len;
773 ip6->ip6_nxt = ip->ip_p;
774 ip6->ip6_hlim = ip->ip_ttl;
775 if (ip->ip_src.s_addr) {
776 ip6->ip6_src.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
777 ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
778 }
779 if (ip->ip_dst.s_addr) {
780 ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
781 ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
782 }
783 }
784 #endif /* INET6 */
785
786 /*
787 * subroutine of udp_input(), mainly for source code readability.
788 */
789 static void
790 #if INET6
791 udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off,
792 struct sockaddr_in *pudp_in, struct udp_in6 *pudp_in6,
793 struct udp_ip6 *pudp_ip6, struct ifnet *ifp)
794 #else /* !INET6 */
795 udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off,
796 struct sockaddr_in *pudp_in, struct ifnet *ifp)
797 #endif /* !INET6 */
798 {
799 struct sockaddr *append_sa;
800 struct mbuf *opts = 0;
801 boolean_t cell = IFNET_IS_CELLULAR(ifp);
802 boolean_t wifi = (!cell && IFNET_IS_WIFI(ifp));
803 boolean_t wired = (!wifi && IFNET_IS_WIRED(ifp));
804 int ret = 0;
805
806 #if CONFIG_MACF_NET
807 if (mac_inpcb_check_deliver(last, n, AF_INET, SOCK_DGRAM) != 0) {
808 m_freem(n);
809 return;
810 }
811 #endif /* CONFIG_MACF_NET */
812 if ((last->inp_flags & INP_CONTROLOPTS) != 0 ||
813 (last->inp_socket->so_options & SO_TIMESTAMP) != 0 ||
814 (last->inp_socket->so_options & SO_TIMESTAMP_MONOTONIC) != 0) {
815 #if INET6
816 if (last->inp_vflag & INP_IPV6) {
817 int savedflags;
818
819 if (pudp_ip6->uip6_init_done == 0) {
820 ip_2_ip6_hdr(&pudp_ip6->uip6_ip6, ip);
821 pudp_ip6->uip6_init_done = 1;
822 }
823 savedflags = last->inp_flags;
824 last->inp_flags &= ~INP_UNMAPPABLEOPTS;
825 ret = ip6_savecontrol(last, n, &opts);
826 if (ret != 0) {
827 last->inp_flags = savedflags;
828 goto error;
829 }
830 last->inp_flags = savedflags;
831 } else
832 #endif /* INET6 */
833 {
834 ret = ip_savecontrol(last, &opts, ip, n);
835 if (ret != 0) {
836 goto error;
837 }
838 }
839 }
840 #if INET6
841 if (last->inp_vflag & INP_IPV6) {
842 if (pudp_in6->uin6_init_done == 0) {
843 in6_sin_2_v4mapsin6(pudp_in, &pudp_in6->uin6_sin);
844 pudp_in6->uin6_init_done = 1;
845 }
846 append_sa = (struct sockaddr *)&pudp_in6->uin6_sin;
847 } else
848 #endif /* INET6 */
849 append_sa = (struct sockaddr *)pudp_in;
850 if (nstat_collect) {
851 INP_ADD_STAT(last, cell, wifi, wired, rxpackets, 1);
852 INP_ADD_STAT(last, cell, wifi, wired, rxbytes,
853 n->m_pkthdr.len);
854 inp_set_activity_bitmap(last);
855 }
856 so_recv_data_stat(last->inp_socket, n, 0);
857 m_adj(n, off);
858 if (sbappendaddr(&last->inp_socket->so_rcv, append_sa,
859 n, opts, NULL) == 0) {
860 udpstat.udps_fullsock++;
861 } else {
862 sorwakeup(last->inp_socket);
863 }
864 return;
865 error:
866 m_freem(n);
867 m_freem(opts);
868 }
869
870 /*
871 * Notify a udp user of an asynchronous error;
872 * just wake up so that he can collect error status.
873 */
874 void
875 udp_notify(struct inpcb *inp, int errno)
876 {
877 inp->inp_socket->so_error = errno;
878 sorwakeup(inp->inp_socket);
879 sowwakeup(inp->inp_socket);
880 }
881
882 void
883 udp_ctlinput(int cmd, struct sockaddr *sa, void *vip, __unused struct ifnet * ifp)
884 {
885 struct ip *ip = vip;
886 void (*notify)(struct inpcb *, int) = udp_notify;
887 struct in_addr faddr;
888 struct inpcb *inp = NULL;
889
890 faddr = ((struct sockaddr_in *)(void *)sa)->sin_addr;
891 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
892 return;
893
894 if (PRC_IS_REDIRECT(cmd)) {
895 ip = 0;
896 notify = in_rtchange;
897 } else if (cmd == PRC_HOSTDEAD) {
898 ip = 0;
899 } else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) {
900 return;
901 }
902 if (ip) {
903 struct udphdr uh;
904
905 bcopy(((caddr_t)ip + (ip->ip_hl << 2)), &uh, sizeof (uh));
906 inp = in_pcblookup_hash(&udbinfo, faddr, uh.uh_dport,
907 ip->ip_src, uh.uh_sport, 0, NULL);
908 if (inp != NULL && inp->inp_socket != NULL) {
909 udp_lock(inp->inp_socket, 1, 0);
910 if (in_pcb_checkstate(inp, WNT_RELEASE, 1) ==
911 WNT_STOPUSING) {
912 udp_unlock(inp->inp_socket, 1, 0);
913 return;
914 }
915 (*notify)(inp, inetctlerrmap[cmd]);
916 udp_unlock(inp->inp_socket, 1, 0);
917 }
918 } else {
919 in_pcbnotifyall(&udbinfo, faddr, inetctlerrmap[cmd], notify);
920 }
921 }
922
923 int
924 udp_ctloutput(struct socket *so, struct sockopt *sopt)
925 {
926 int error = 0, optval = 0;
927 struct inpcb *inp;
928
929 /* Allow <SOL_SOCKET,SO_FLUSH> at this level */
930 if (sopt->sopt_level != IPPROTO_UDP &&
931 !(sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_FLUSH))
932 return (ip_ctloutput(so, sopt));
933
934 inp = sotoinpcb(so);
935
936 switch (sopt->sopt_dir) {
937 case SOPT_SET:
938 switch (sopt->sopt_name) {
939 case UDP_NOCKSUM:
940 /* This option is settable only for UDP over IPv4 */
941 if (!(inp->inp_vflag & INP_IPV4)) {
942 error = EINVAL;
943 break;
944 }
945
946 if ((error = sooptcopyin(sopt, &optval, sizeof (optval),
947 sizeof (optval))) != 0)
948 break;
949
950 if (optval != 0)
951 inp->inp_flags |= INP_UDP_NOCKSUM;
952 else
953 inp->inp_flags &= ~INP_UDP_NOCKSUM;
954 break;
955 case UDP_KEEPALIVE_OFFLOAD:
956 {
957 struct udp_keepalive_offload ka;
958 /*
959 * If the socket is not connected, the stack will
960 * not know the destination address to put in the
961 * keepalive datagram. Return an error now instead
962 * of failing later.
963 */
964 if (!(so->so_state & SS_ISCONNECTED)) {
965 error = EINVAL;
966 break;
967 }
968 if (sopt->sopt_valsize != sizeof(ka)) {
969 error = EINVAL;
970 break;
971 }
972 if ((error = sooptcopyin(sopt, &ka, sizeof(ka),
973 sizeof(ka))) != 0)
974 break;
975
976 /* application should specify the type */
977 if (ka.ka_type == 0)
978 return (EINVAL);
979
980 if (ka.ka_interval == 0) {
981 /*
982 * if interval is 0, disable the offload
983 * mechanism
984 */
985 if (inp->inp_keepalive_data != NULL)
986 FREE(inp->inp_keepalive_data,
987 M_TEMP);
988 inp->inp_keepalive_data = NULL;
989 inp->inp_keepalive_datalen = 0;
990 inp->inp_keepalive_interval = 0;
991 inp->inp_keepalive_type = 0;
992 inp->inp_flags2 &= ~INP2_KEEPALIVE_OFFLOAD;
993 } else {
994 if (inp->inp_keepalive_data != NULL) {
995 FREE(inp->inp_keepalive_data,
996 M_TEMP);
997 inp->inp_keepalive_data = NULL;
998 }
999
1000 inp->inp_keepalive_datalen = min(
1001 ka.ka_data_len,
1002 UDP_KEEPALIVE_OFFLOAD_DATA_SIZE);
1003 if (inp->inp_keepalive_datalen > 0) {
1004 MALLOC(inp->inp_keepalive_data,
1005 u_int8_t *,
1006 inp->inp_keepalive_datalen,
1007 M_TEMP, M_WAITOK);
1008 if (inp->inp_keepalive_data == NULL) {
1009 inp->inp_keepalive_datalen = 0;
1010 error = ENOMEM;
1011 break;
1012 }
1013 bcopy(ka.ka_data,
1014 inp->inp_keepalive_data,
1015 inp->inp_keepalive_datalen);
1016 } else {
1017 inp->inp_keepalive_datalen = 0;
1018 }
1019 inp->inp_keepalive_interval =
1020 min(UDP_KEEPALIVE_INTERVAL_MAX_SECONDS,
1021 ka.ka_interval);
1022 inp->inp_keepalive_type = ka.ka_type;
1023 inp->inp_flags2 |= INP2_KEEPALIVE_OFFLOAD;
1024 }
1025 break;
1026 }
1027 case SO_FLUSH:
1028 if ((error = sooptcopyin(sopt, &optval, sizeof (optval),
1029 sizeof (optval))) != 0)
1030 break;
1031
1032 error = inp_flush(inp, optval);
1033 break;
1034
1035 default:
1036 error = ENOPROTOOPT;
1037 break;
1038 }
1039 break;
1040
1041 case SOPT_GET:
1042 switch (sopt->sopt_name) {
1043 case UDP_NOCKSUM:
1044 optval = inp->inp_flags & INP_UDP_NOCKSUM;
1045 break;
1046
1047 default:
1048 error = ENOPROTOOPT;
1049 break;
1050 }
1051 if (error == 0)
1052 error = sooptcopyout(sopt, &optval, sizeof (optval));
1053 break;
1054 }
1055 return (error);
1056 }
1057
1058 static int
1059 udp_pcblist SYSCTL_HANDLER_ARGS
1060 {
1061 #pragma unused(oidp, arg1, arg2)
1062 int error, i, n;
1063 struct inpcb *inp, **inp_list;
1064 inp_gen_t gencnt;
1065 struct xinpgen xig;
1066
1067 /*
1068 * The process of preparing the TCB list is too time-consuming and
1069 * resource-intensive to repeat twice on every request.
1070 */
1071 lck_rw_lock_exclusive(udbinfo.ipi_lock);
1072 if (req->oldptr == USER_ADDR_NULL) {
1073 n = udbinfo.ipi_count;
1074 req->oldidx = 2 * (sizeof (xig))
1075 + (n + n/8) * sizeof (struct xinpcb);
1076 lck_rw_done(udbinfo.ipi_lock);
1077 return (0);
1078 }
1079
1080 if (req->newptr != USER_ADDR_NULL) {
1081 lck_rw_done(udbinfo.ipi_lock);
1082 return (EPERM);
1083 }
1084
1085 /*
1086 * OK, now we're committed to doing something.
1087 */
1088 gencnt = udbinfo.ipi_gencnt;
1089 n = udbinfo.ipi_count;
1090
1091 bzero(&xig, sizeof (xig));
1092 xig.xig_len = sizeof (xig);
1093 xig.xig_count = n;
1094 xig.xig_gen = gencnt;
1095 xig.xig_sogen = so_gencnt;
1096 error = SYSCTL_OUT(req, &xig, sizeof (xig));
1097 if (error) {
1098 lck_rw_done(udbinfo.ipi_lock);
1099 return (error);
1100 }
1101 /*
1102 * We are done if there is no pcb
1103 */
1104 if (n == 0) {
1105 lck_rw_done(udbinfo.ipi_lock);
1106 return (0);
1107 }
1108
1109 inp_list = _MALLOC(n * sizeof (*inp_list), M_TEMP, M_WAITOK);
1110 if (inp_list == 0) {
1111 lck_rw_done(udbinfo.ipi_lock);
1112 return (ENOMEM);
1113 }
1114
1115 for (inp = LIST_FIRST(udbinfo.ipi_listhead), i = 0; inp && i < n;
1116 inp = LIST_NEXT(inp, inp_list)) {
1117 if (inp->inp_gencnt <= gencnt &&
1118 inp->inp_state != INPCB_STATE_DEAD)
1119 inp_list[i++] = inp;
1120 }
1121 n = i;
1122
1123 error = 0;
1124 for (i = 0; i < n; i++) {
1125 struct xinpcb xi;
1126
1127 inp = inp_list[i];
1128
1129 if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) == WNT_STOPUSING)
1130 continue;
1131 udp_lock(inp->inp_socket, 1, 0);
1132 if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
1133 udp_unlock(inp->inp_socket, 1, 0);
1134 continue;
1135 }
1136 if (inp->inp_gencnt > gencnt) {
1137 udp_unlock(inp->inp_socket, 1, 0);
1138 continue;
1139 }
1140
1141 bzero(&xi, sizeof (xi));
1142 xi.xi_len = sizeof (xi);
1143 /* XXX should avoid extra copy */
1144 inpcb_to_compat(inp, &xi.xi_inp);
1145 if (inp->inp_socket)
1146 sotoxsocket(inp->inp_socket, &xi.xi_socket);
1147
1148 udp_unlock(inp->inp_socket, 1, 0);
1149
1150 error = SYSCTL_OUT(req, &xi, sizeof (xi));
1151 }
1152 if (!error) {
1153 /*
1154 * Give the user an updated idea of our state.
1155 * If the generation differs from what we told
1156 * her before, she knows that something happened
1157 * while we were processing this request, and it
1158 * might be necessary to retry.
1159 */
1160 bzero(&xig, sizeof (xig));
1161 xig.xig_len = sizeof (xig);
1162 xig.xig_gen = udbinfo.ipi_gencnt;
1163 xig.xig_sogen = so_gencnt;
1164 xig.xig_count = udbinfo.ipi_count;
1165 error = SYSCTL_OUT(req, &xig, sizeof (xig));
1166 }
1167 FREE(inp_list, M_TEMP);
1168 lck_rw_done(udbinfo.ipi_lock);
1169 return (error);
1170 }
1171
1172 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist,
1173 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0, udp_pcblist,
1174 "S,xinpcb", "List of active UDP sockets");
1175
1176 #if !CONFIG_EMBEDDED
1177
1178 static int
1179 udp_pcblist64 SYSCTL_HANDLER_ARGS
1180 {
1181 #pragma unused(oidp, arg1, arg2)
1182 int error, i, n;
1183 struct inpcb *inp, **inp_list;
1184 inp_gen_t gencnt;
1185 struct xinpgen xig;
1186
1187 /*
1188 * The process of preparing the TCB list is too time-consuming and
1189 * resource-intensive to repeat twice on every request.
1190 */
1191 lck_rw_lock_shared(udbinfo.ipi_lock);
1192 if (req->oldptr == USER_ADDR_NULL) {
1193 n = udbinfo.ipi_count;
1194 req->oldidx =
1195 2 * (sizeof (xig)) + (n + n/8) * sizeof (struct xinpcb64);
1196 lck_rw_done(udbinfo.ipi_lock);
1197 return (0);
1198 }
1199
1200 if (req->newptr != USER_ADDR_NULL) {
1201 lck_rw_done(udbinfo.ipi_lock);
1202 return (EPERM);
1203 }
1204
1205 /*
1206 * OK, now we're committed to doing something.
1207 */
1208 gencnt = udbinfo.ipi_gencnt;
1209 n = udbinfo.ipi_count;
1210
1211 bzero(&xig, sizeof (xig));
1212 xig.xig_len = sizeof (xig);
1213 xig.xig_count = n;
1214 xig.xig_gen = gencnt;
1215 xig.xig_sogen = so_gencnt;
1216 error = SYSCTL_OUT(req, &xig, sizeof (xig));
1217 if (error) {
1218 lck_rw_done(udbinfo.ipi_lock);
1219 return (error);
1220 }
1221 /*
1222 * We are done if there is no pcb
1223 */
1224 if (n == 0) {
1225 lck_rw_done(udbinfo.ipi_lock);
1226 return (0);
1227 }
1228
1229 inp_list = _MALLOC(n * sizeof (*inp_list), M_TEMP, M_WAITOK);
1230 if (inp_list == 0) {
1231 lck_rw_done(udbinfo.ipi_lock);
1232 return (ENOMEM);
1233 }
1234
1235 for (inp = LIST_FIRST(udbinfo.ipi_listhead), i = 0; inp && i < n;
1236 inp = LIST_NEXT(inp, inp_list)) {
1237 if (inp->inp_gencnt <= gencnt &&
1238 inp->inp_state != INPCB_STATE_DEAD)
1239 inp_list[i++] = inp;
1240 }
1241 n = i;
1242
1243 error = 0;
1244 for (i = 0; i < n; i++) {
1245 struct xinpcb64 xi;
1246
1247 inp = inp_list[i];
1248
1249 if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) == WNT_STOPUSING)
1250 continue;
1251 udp_lock(inp->inp_socket, 1, 0);
1252 if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
1253 udp_unlock(inp->inp_socket, 1, 0);
1254 continue;
1255 }
1256 if (inp->inp_gencnt > gencnt) {
1257 udp_unlock(inp->inp_socket, 1, 0);
1258 continue;
1259 }
1260
1261 bzero(&xi, sizeof (xi));
1262 xi.xi_len = sizeof (xi);
1263 inpcb_to_xinpcb64(inp, &xi);
1264 if (inp->inp_socket)
1265 sotoxsocket64(inp->inp_socket, &xi.xi_socket);
1266
1267 udp_unlock(inp->inp_socket, 1, 0);
1268
1269 error = SYSCTL_OUT(req, &xi, sizeof (xi));
1270 }
1271 if (!error) {
1272 /*
1273 * Give the user an updated idea of our state.
1274 * If the generation differs from what we told
1275 * her before, she knows that something happened
1276 * while we were processing this request, and it
1277 * might be necessary to retry.
1278 */
1279 bzero(&xig, sizeof (xig));
1280 xig.xig_len = sizeof (xig);
1281 xig.xig_gen = udbinfo.ipi_gencnt;
1282 xig.xig_sogen = so_gencnt;
1283 xig.xig_count = udbinfo.ipi_count;
1284 error = SYSCTL_OUT(req, &xig, sizeof (xig));
1285 }
1286 FREE(inp_list, M_TEMP);
1287 lck_rw_done(udbinfo.ipi_lock);
1288 return (error);
1289 }
1290
1291 SYSCTL_PROC(_net_inet_udp, OID_AUTO, pcblist64,
1292 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0, udp_pcblist64,
1293 "S,xinpcb64", "List of active UDP sockets");
1294
1295 #endif /* !CONFIG_EMBEDDED */
1296
1297 static int
1298 udp_pcblist_n SYSCTL_HANDLER_ARGS
1299 {
1300 #pragma unused(oidp, arg1, arg2)
1301 return (get_pcblist_n(IPPROTO_UDP, req, &udbinfo));
1302 }
1303
1304 SYSCTL_PROC(_net_inet_udp, OID_AUTO, pcblist_n,
1305 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0, udp_pcblist_n,
1306 "S,xinpcb_n", "List of active UDP sockets");
1307
1308 __private_extern__ void
1309 udp_get_ports_used(uint32_t ifindex, int protocol, uint32_t flags,
1310 bitstr_t *bitfield)
1311 {
1312 inpcb_get_ports_used(ifindex, protocol, flags, bitfield,
1313 &udbinfo);
1314 }
1315
1316 __private_extern__ uint32_t
1317 udp_count_opportunistic(unsigned int ifindex, u_int32_t flags)
1318 {
1319 return (inpcb_count_opportunistic(ifindex, &udbinfo, flags));
1320 }
1321
1322 __private_extern__ uint32_t
1323 udp_find_anypcb_byaddr(struct ifaddr *ifa)
1324 {
1325 return (inpcb_find_anypcb_byaddr(ifa, &udbinfo));
1326 }
1327
1328 static int
1329 udp_check_pktinfo(struct mbuf *control, struct ifnet **outif,
1330 struct in_addr *laddr)
1331 {
1332 struct cmsghdr *cm = 0;
1333 struct in_pktinfo *pktinfo;
1334 struct ifnet *ifp;
1335
1336 if (outif != NULL)
1337 *outif = NULL;
1338
1339 /*
1340 * XXX: Currently, we assume all the optional information is stored
1341 * in a single mbuf.
1342 */
1343 if (control->m_next)
1344 return (EINVAL);
1345
1346 if (control->m_len < CMSG_LEN(0))
1347 return (EINVAL);
1348
1349 for (cm = M_FIRST_CMSGHDR(control); cm;
1350 cm = M_NXT_CMSGHDR(control, cm)) {
1351 if (cm->cmsg_len < sizeof (struct cmsghdr) ||
1352 cm->cmsg_len > control->m_len)
1353 return (EINVAL);
1354
1355 if (cm->cmsg_level != IPPROTO_IP || cm->cmsg_type != IP_PKTINFO)
1356 continue;
1357
1358 if (cm->cmsg_len != CMSG_LEN(sizeof (struct in_pktinfo)))
1359 return (EINVAL);
1360
1361 pktinfo = (struct in_pktinfo *)(void *)CMSG_DATA(cm);
1362
1363 /* Check for a valid ifindex in pktinfo */
1364 ifnet_head_lock_shared();
1365
1366 if (pktinfo->ipi_ifindex > if_index) {
1367 ifnet_head_done();
1368 return (ENXIO);
1369 }
1370
1371 /*
1372 * If ipi_ifindex is specified it takes precedence
1373 * over ipi_spec_dst.
1374 */
1375 if (pktinfo->ipi_ifindex) {
1376 ifp = ifindex2ifnet[pktinfo->ipi_ifindex];
1377 if (ifp == NULL) {
1378 ifnet_head_done();
1379 return (ENXIO);
1380 }
1381 if (outif != NULL) {
1382 ifnet_reference(ifp);
1383 *outif = ifp;
1384 }
1385 ifnet_head_done();
1386 laddr->s_addr = INADDR_ANY;
1387 break;
1388 }
1389
1390 ifnet_head_done();
1391
1392 /*
1393 * Use the provided ipi_spec_dst address for temp
1394 * source address.
1395 */
1396 *laddr = pktinfo->ipi_spec_dst;
1397 break;
1398 }
1399 return (0);
1400 }
1401
1402 int
1403 udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr,
1404 struct mbuf *control, struct proc *p)
1405 {
1406 struct udpiphdr *ui;
1407 int len = m->m_pkthdr.len;
1408 struct sockaddr_in *sin;
1409 struct in_addr origladdr, laddr, faddr, pi_laddr;
1410 u_short lport, fport;
1411 int error = 0, udp_dodisconnect = 0, pktinfo = 0;
1412 struct socket *so = inp->inp_socket;
1413 int soopts = 0;
1414 struct mbuf *inpopts;
1415 struct ip_moptions *mopts;
1416 struct route ro;
1417 struct ip_out_args ipoa =
1418 { IFSCOPE_NONE, { 0 }, IPOAF_SELECT_SRCIF, 0, 0, 0 };
1419 struct ifnet *outif = NULL;
1420 struct flowadv *adv = &ipoa.ipoa_flowadv;
1421 int sotc = SO_TC_UNSPEC;
1422 int netsvctype = _NET_SERVICE_TYPE_UNSPEC;
1423 struct ifnet *origoutifp = NULL;
1424 int flowadv = 0;
1425
1426 /* Enable flow advisory only when connected */
1427 flowadv = (so->so_state & SS_ISCONNECTED) ? 1 : 0;
1428 pi_laddr.s_addr = INADDR_ANY;
1429
1430 KERNEL_DEBUG(DBG_FNC_UDP_OUTPUT | DBG_FUNC_START, 0, 0, 0, 0, 0);
1431
1432 socket_lock_assert_owned(so);
1433 if (control != NULL) {
1434 sotc = so_tc_from_control(control, &netsvctype);
1435 VERIFY(outif == NULL);
1436 error = udp_check_pktinfo(control, &outif, &pi_laddr);
1437 m_freem(control);
1438 control = NULL;
1439 if (error)
1440 goto release;
1441 pktinfo++;
1442 if (outif != NULL)
1443 ipoa.ipoa_boundif = outif->if_index;
1444 }
1445 if (sotc == SO_TC_UNSPEC) {
1446 sotc = so->so_traffic_class;
1447 netsvctype = so->so_netsvctype;
1448 }
1449
1450 KERNEL_DEBUG(DBG_LAYER_OUT_BEG, inp->inp_fport, inp->inp_lport,
1451 inp->inp_laddr.s_addr, inp->inp_faddr.s_addr,
1452 (htons((u_short)len + sizeof (struct udphdr))));
1453
1454 if (len + sizeof (struct udpiphdr) > IP_MAXPACKET) {
1455 error = EMSGSIZE;
1456 goto release;
1457 }
1458
1459 if (flowadv && INP_WAIT_FOR_IF_FEEDBACK(inp)) {
1460 /*
1461 * The socket is flow-controlled, drop the packets
1462 * until the inp is not flow controlled
1463 */
1464 error = ENOBUFS;
1465 goto release;
1466 }
1467 /*
1468 * If socket was bound to an ifindex, tell ip_output about it.
1469 * If the ancillary IP_PKTINFO option contains an interface index,
1470 * it takes precedence over the one specified by IP_BOUND_IF.
1471 */
1472 if (ipoa.ipoa_boundif == IFSCOPE_NONE &&
1473 (inp->inp_flags & INP_BOUND_IF)) {
1474 VERIFY(inp->inp_boundifp != NULL);
1475 ifnet_reference(inp->inp_boundifp); /* for this routine */
1476 if (outif != NULL)
1477 ifnet_release(outif);
1478 outif = inp->inp_boundifp;
1479 ipoa.ipoa_boundif = outif->if_index;
1480 }
1481 if (INP_NO_CELLULAR(inp))
1482 ipoa.ipoa_flags |= IPOAF_NO_CELLULAR;
1483 if (INP_NO_EXPENSIVE(inp))
1484 ipoa.ipoa_flags |= IPOAF_NO_EXPENSIVE;
1485 if (INP_AWDL_UNRESTRICTED(inp))
1486 ipoa.ipoa_flags |= IPOAF_AWDL_UNRESTRICTED;
1487 ipoa.ipoa_sotc = sotc;
1488 ipoa.ipoa_netsvctype = netsvctype;
1489 soopts |= IP_OUTARGS;
1490
1491 /*
1492 * If there was a routing change, discard cached route and check
1493 * that we have a valid source address. Reacquire a new source
1494 * address if INADDR_ANY was specified.
1495 */
1496 if (ROUTE_UNUSABLE(&inp->inp_route)) {
1497 struct in_ifaddr *ia = NULL;
1498
1499 ROUTE_RELEASE(&inp->inp_route);
1500
1501 /* src address is gone? */
1502 if (inp->inp_laddr.s_addr != INADDR_ANY &&
1503 (ia = ifa_foraddr(inp->inp_laddr.s_addr)) == NULL) {
1504 if (!(inp->inp_flags & INP_INADDR_ANY) ||
1505 (so->so_state & SS_ISCONNECTED)) {
1506 /*
1507 * Rdar://5448998
1508 * If the source address is gone, return an
1509 * error if:
1510 * - the source was specified
1511 * - the socket was already connected
1512 */
1513 soevent(so, (SO_FILT_HINT_LOCKED |
1514 SO_FILT_HINT_NOSRCADDR));
1515 error = EADDRNOTAVAIL;
1516 goto release;
1517 } else {
1518 /* new src will be set later */
1519 inp->inp_laddr.s_addr = INADDR_ANY;
1520 inp->inp_last_outifp = NULL;
1521 }
1522 }
1523 if (ia != NULL)
1524 IFA_REMREF(&ia->ia_ifa);
1525 }
1526
1527 /*
1528 * IP_PKTINFO option check. If a temporary scope or src address
1529 * is provided, use it for this packet only and make sure we forget
1530 * it after sending this datagram.
1531 */
1532 if (pi_laddr.s_addr != INADDR_ANY ||
1533 (ipoa.ipoa_boundif != IFSCOPE_NONE && pktinfo)) {
1534 /* temp src address for this datagram only */
1535 laddr = pi_laddr;
1536 origladdr.s_addr = INADDR_ANY;
1537 /* we don't want to keep the laddr or route */
1538 udp_dodisconnect = 1;
1539 /* remember we don't care about src addr */
1540 inp->inp_flags |= INP_INADDR_ANY;
1541 } else {
1542 origladdr = laddr = inp->inp_laddr;
1543 }
1544
1545 origoutifp = inp->inp_last_outifp;
1546 faddr = inp->inp_faddr;
1547 lport = inp->inp_lport;
1548 fport = inp->inp_fport;
1549
1550 if (addr) {
1551 sin = (struct sockaddr_in *)(void *)addr;
1552 if (faddr.s_addr != INADDR_ANY) {
1553 error = EISCONN;
1554 goto release;
1555 }
1556 if (lport == 0) {
1557 /*
1558 * In case we don't have a local port set, go through
1559 * the full connect. We don't have a local port yet
1560 * (i.e., we can't be looked up), so it's not an issue
1561 * if the input runs at the same time we do this.
1562 */
1563 /* if we have a source address specified, use that */
1564 if (pi_laddr.s_addr != INADDR_ANY)
1565 inp->inp_laddr = pi_laddr;
1566 /*
1567 * If a scope is specified, use it. Scope from
1568 * IP_PKTINFO takes precendence over the the scope
1569 * set via INP_BOUND_IF.
1570 */
1571 error = in_pcbconnect(inp, addr, p, ipoa.ipoa_boundif,
1572 &outif);
1573 if (error)
1574 goto release;
1575
1576 laddr = inp->inp_laddr;
1577 lport = inp->inp_lport;
1578 faddr = inp->inp_faddr;
1579 fport = inp->inp_fport;
1580 udp_dodisconnect = 1;
1581
1582 /* synch up in case in_pcbladdr() overrides */
1583 if (outif != NULL && ipoa.ipoa_boundif != IFSCOPE_NONE)
1584 ipoa.ipoa_boundif = outif->if_index;
1585 } else {
1586 /*
1587 * Fast path case
1588 *
1589 * We have a full address and a local port; use those
1590 * info to build the packet without changing the pcb
1591 * and interfering with the input path. See 3851370.
1592 *
1593 * Scope from IP_PKTINFO takes precendence over the
1594 * the scope set via INP_BOUND_IF.
1595 */
1596 if (laddr.s_addr == INADDR_ANY) {
1597 if ((error = in_pcbladdr(inp, addr, &laddr,
1598 ipoa.ipoa_boundif, &outif, 0)) != 0)
1599 goto release;
1600 /*
1601 * from pcbconnect: remember we don't
1602 * care about src addr.
1603 */
1604 inp->inp_flags |= INP_INADDR_ANY;
1605
1606 /* synch up in case in_pcbladdr() overrides */
1607 if (outif != NULL &&
1608 ipoa.ipoa_boundif != IFSCOPE_NONE)
1609 ipoa.ipoa_boundif = outif->if_index;
1610 }
1611
1612 faddr = sin->sin_addr;
1613 fport = sin->sin_port;
1614 }
1615 } else {
1616 if (faddr.s_addr == INADDR_ANY) {
1617 error = ENOTCONN;
1618 goto release;
1619 }
1620 }
1621
1622 #if CONFIG_MACF_NET
1623 mac_mbuf_label_associate_inpcb(inp, m);
1624 #endif /* CONFIG_MACF_NET */
1625
1626 if (inp->inp_flowhash == 0)
1627 inp->inp_flowhash = inp_calc_flowhash(inp);
1628
1629 if (fport == htons(53) && !(so->so_flags1 & SOF1_DNS_COUNTED)) {
1630 so->so_flags1 |= SOF1_DNS_COUNTED;
1631 INC_ATOMIC_INT64_LIM(net_api_stats.nas_socket_inet_dgram_dns);
1632 }
1633
1634 /*
1635 * Calculate data length and get a mbuf
1636 * for UDP and IP headers.
1637 */
1638 M_PREPEND(m, sizeof (struct udpiphdr), M_DONTWAIT, 1);
1639 if (m == 0) {
1640 error = ENOBUFS;
1641 goto abort;
1642 }
1643
1644 /*
1645 * Fill in mbuf with extended UDP header
1646 * and addresses and length put into network format.
1647 */
1648 ui = mtod(m, struct udpiphdr *);
1649 bzero(ui->ui_x1, sizeof (ui->ui_x1)); /* XXX still needed? */
1650 ui->ui_pr = IPPROTO_UDP;
1651 ui->ui_src = laddr;
1652 ui->ui_dst = faddr;
1653 ui->ui_sport = lport;
1654 ui->ui_dport = fport;
1655 ui->ui_ulen = htons((u_short)len + sizeof (struct udphdr));
1656
1657 /*
1658 * Set up checksum and output datagram.
1659 */
1660 if (udpcksum && !(inp->inp_flags & INP_UDP_NOCKSUM)) {
1661 ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
1662 htons((u_short)len + sizeof (struct udphdr) + IPPROTO_UDP));
1663 m->m_pkthdr.csum_flags = (CSUM_UDP|CSUM_ZERO_INVERT);
1664 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1665 } else {
1666 ui->ui_sum = 0;
1667 }
1668 ((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len;
1669 ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */
1670 ((struct ip *)ui)->ip_tos = inp->inp_ip_tos; /* XXX */
1671 udpstat.udps_opackets++;
1672
1673 KERNEL_DEBUG(DBG_LAYER_OUT_END, ui->ui_dport, ui->ui_sport,
1674 ui->ui_src.s_addr, ui->ui_dst.s_addr, ui->ui_ulen);
1675
1676 #if NECP
1677 {
1678 necp_kernel_policy_id policy_id;
1679 u_int32_t route_rule_id;
1680
1681 /*
1682 * We need a route to perform NECP route rule checks
1683 */
1684 if (net_qos_policy_restricted != 0 &&
1685 ROUTE_UNUSABLE(&inp->inp_route)) {
1686 struct sockaddr_in to;
1687 struct sockaddr_in from;
1688
1689 ROUTE_RELEASE(&inp->inp_route);
1690
1691 bzero(&from, sizeof(struct sockaddr_in));
1692 from.sin_family = AF_INET;
1693 from.sin_len = sizeof(struct sockaddr_in);
1694 from.sin_addr = laddr;
1695
1696 bzero(&to, sizeof(struct sockaddr_in));
1697 to.sin_family = AF_INET;
1698 to.sin_len = sizeof(struct sockaddr_in);
1699 to.sin_addr = faddr;
1700
1701 inp->inp_route.ro_dst.sa_family = AF_INET;
1702 inp->inp_route.ro_dst.sa_len = sizeof(struct sockaddr_in);
1703 ((struct sockaddr_in *)(void *)&inp->inp_route.ro_dst)->sin_addr =
1704 faddr;
1705
1706 rtalloc_scoped(&inp->inp_route, ipoa.ipoa_boundif);
1707
1708 inp_update_necp_policy(inp, (struct sockaddr *)&from,
1709 (struct sockaddr *)&to, ipoa.ipoa_boundif);
1710 inp->inp_policyresult.results.qos_marking_gencount = 0;
1711 }
1712
1713 if (!necp_socket_is_allowed_to_send_recv_v4(inp, lport, fport,
1714 &laddr, &faddr, NULL, &policy_id, &route_rule_id)) {
1715 error = EHOSTUNREACH;
1716 goto abort;
1717 }
1718
1719 necp_mark_packet_from_socket(m, inp, policy_id, route_rule_id);
1720
1721 if (net_qos_policy_restricted != 0) {
1722 necp_socket_update_qos_marking(inp,
1723 inp->inp_route.ro_rt, NULL, route_rule_id);
1724 }
1725 }
1726 #endif /* NECP */
1727 if ((so->so_flags1 & SOF1_QOSMARKING_ALLOWED))
1728 ipoa.ipoa_flags |= IPOAF_QOSMARKING_ALLOWED;
1729
1730 #if IPSEC
1731 if (inp->inp_sp != NULL && ipsec_setsocket(m, inp->inp_socket) != 0) {
1732 error = ENOBUFS;
1733 goto abort;
1734 }
1735 #endif /* IPSEC */
1736
1737 inpopts = inp->inp_options;
1738 soopts |= (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST));
1739 mopts = inp->inp_moptions;
1740 if (mopts != NULL) {
1741 IMO_LOCK(mopts);
1742 IMO_ADDREF_LOCKED(mopts);
1743 if (IN_MULTICAST(ntohl(ui->ui_dst.s_addr)) &&
1744 mopts->imo_multicast_ifp != NULL) {
1745 /* no reference needed */
1746 inp->inp_last_outifp = mopts->imo_multicast_ifp;
1747
1748 }
1749 IMO_UNLOCK(mopts);
1750 }
1751
1752 /* Copy the cached route and take an extra reference */
1753 inp_route_copyout(inp, &ro);
1754
1755 set_packet_service_class(m, so, sotc, 0);
1756 m->m_pkthdr.pkt_flowsrc = FLOWSRC_INPCB;
1757 m->m_pkthdr.pkt_flowid = inp->inp_flowhash;
1758 m->m_pkthdr.pkt_proto = IPPROTO_UDP;
1759 m->m_pkthdr.pkt_flags |= (PKTF_FLOW_ID | PKTF_FLOW_LOCALSRC);
1760 if (flowadv)
1761 m->m_pkthdr.pkt_flags |= PKTF_FLOW_ADV;
1762
1763 if (ipoa.ipoa_boundif != IFSCOPE_NONE)
1764 ipoa.ipoa_flags |= IPOAF_BOUND_IF;
1765
1766 if (laddr.s_addr != INADDR_ANY)
1767 ipoa.ipoa_flags |= IPOAF_BOUND_SRCADDR;
1768
1769 inp->inp_sndinprog_cnt++;
1770
1771 socket_unlock(so, 0);
1772 error = ip_output(m, inpopts, &ro, soopts, mopts, &ipoa);
1773 m = NULL;
1774 socket_lock(so, 0);
1775 if (mopts != NULL)
1776 IMO_REMREF(mopts);
1777
1778 if (error == 0 && nstat_collect) {
1779 boolean_t cell, wifi, wired;
1780
1781 if (ro.ro_rt != NULL) {
1782 cell = IFNET_IS_CELLULAR(ro.ro_rt->rt_ifp);
1783 wifi = (!cell && IFNET_IS_WIFI(ro.ro_rt->rt_ifp));
1784 wired = (!wifi && IFNET_IS_WIRED(ro.ro_rt->rt_ifp));
1785 } else {
1786 cell = wifi = wired = FALSE;
1787 }
1788 INP_ADD_STAT(inp, cell, wifi, wired, txpackets, 1);
1789 INP_ADD_STAT(inp, cell, wifi, wired, txbytes, len);
1790 inp_set_activity_bitmap(inp);
1791 }
1792
1793 if (flowadv && (adv->code == FADV_FLOW_CONTROLLED ||
1794 adv->code == FADV_SUSPENDED)) {
1795 /*
1796 * return a hint to the application that
1797 * the packet has been dropped
1798 */
1799 error = ENOBUFS;
1800 inp_set_fc_state(inp, adv->code);
1801 }
1802
1803 VERIFY(inp->inp_sndinprog_cnt > 0);
1804 if ( --inp->inp_sndinprog_cnt == 0)
1805 inp->inp_flags &= ~(INP_FC_FEEDBACK);
1806
1807 /* Synchronize PCB cached route */
1808 inp_route_copyin(inp, &ro);
1809
1810 abort:
1811 if (udp_dodisconnect) {
1812 /* Always discard the cached route for unconnected socket */
1813 ROUTE_RELEASE(&inp->inp_route);
1814 in_pcbdisconnect(inp);
1815 inp->inp_laddr = origladdr; /* XXX rehash? */
1816 /* no reference needed */
1817 inp->inp_last_outifp = origoutifp;
1818
1819 } else if (inp->inp_route.ro_rt != NULL) {
1820 struct rtentry *rt = inp->inp_route.ro_rt;
1821 struct ifnet *outifp;
1822
1823 if (rt->rt_flags & (RTF_MULTICAST|RTF_BROADCAST))
1824 rt = NULL; /* unusable */
1825 /*
1826 * Always discard if it is a multicast or broadcast route.
1827 */
1828 if (rt == NULL)
1829 ROUTE_RELEASE(&inp->inp_route);
1830
1831 /*
1832 * If the destination route is unicast, update outifp with
1833 * that of the route interface used by IP.
1834 */
1835 if (rt != NULL &&
1836 (outifp = rt->rt_ifp) != inp->inp_last_outifp) {
1837 inp->inp_last_outifp = outifp; /* no reference needed */
1838
1839 so->so_pktheadroom = P2ROUNDUP(
1840 sizeof(struct udphdr) +
1841 sizeof(struct ip) +
1842 ifnet_hdrlen(outifp) +
1843 ifnet_mbuf_packetpreamblelen(outifp),
1844 sizeof(u_int32_t));
1845 }
1846 } else {
1847 ROUTE_RELEASE(&inp->inp_route);
1848 }
1849
1850 /*
1851 * If output interface was cellular/expensive, and this socket is
1852 * denied access to it, generate an event.
1853 */
1854 if (error != 0 && (ipoa.ipoa_retflags & IPOARF_IFDENIED) &&
1855 (INP_NO_CELLULAR(inp) || INP_NO_EXPENSIVE(inp)))
1856 soevent(so, (SO_FILT_HINT_LOCKED|SO_FILT_HINT_IFDENIED));
1857
1858 release:
1859 KERNEL_DEBUG(DBG_FNC_UDP_OUTPUT | DBG_FUNC_END, error, 0, 0, 0, 0);
1860
1861 if (m != NULL)
1862 m_freem(m);
1863
1864 if (outif != NULL)
1865 ifnet_release(outif);
1866
1867 return (error);
1868 }
1869
1870 u_int32_t udp_sendspace = 9216; /* really max datagram size */
1871 /* 187 1K datagrams (approx 192 KB) */
1872 u_int32_t udp_recvspace = 187 * (1024 +
1873 #if INET6
1874 sizeof (struct sockaddr_in6)
1875 #else /* !INET6 */
1876 sizeof (struct sockaddr_in)
1877 #endif /* !INET6 */
1878 );
1879
1880 /* Check that the values of udp send and recv space do not exceed sb_max */
1881 static int
1882 sysctl_udp_sospace(struct sysctl_oid *oidp, void *arg1, int arg2,
1883 struct sysctl_req *req)
1884 {
1885 #pragma unused(arg1, arg2)
1886 u_int32_t new_value = 0, *space_p = NULL;
1887 int changed = 0, error = 0;
1888 u_quad_t sb_effective_max = (sb_max/(MSIZE+MCLBYTES)) * MCLBYTES;
1889
1890 switch (oidp->oid_number) {
1891 case UDPCTL_RECVSPACE:
1892 space_p = &udp_recvspace;
1893 break;
1894 case UDPCTL_MAXDGRAM:
1895 space_p = &udp_sendspace;
1896 break;
1897 default:
1898 return (EINVAL);
1899 }
1900 error = sysctl_io_number(req, *space_p, sizeof (u_int32_t),
1901 &new_value, &changed);
1902 if (changed) {
1903 if (new_value > 0 && new_value <= sb_effective_max)
1904 *space_p = new_value;
1905 else
1906 error = ERANGE;
1907 }
1908 return (error);
1909 }
1910
1911 SYSCTL_PROC(_net_inet_udp, UDPCTL_RECVSPACE, recvspace,
1912 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &udp_recvspace, 0,
1913 &sysctl_udp_sospace, "IU", "Maximum incoming UDP datagram size");
1914
1915 SYSCTL_PROC(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram,
1916 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &udp_sendspace, 0,
1917 &sysctl_udp_sospace, "IU", "Maximum outgoing UDP datagram size");
1918
1919 int
1920 udp_abort(struct socket *so)
1921 {
1922 struct inpcb *inp;
1923
1924 inp = sotoinpcb(so);
1925 if (inp == NULL) {
1926 panic("%s: so=%p null inp\n", __func__, so);
1927 /* NOTREACHED */
1928 }
1929 soisdisconnected(so);
1930 in_pcbdetach(inp);
1931 return (0);
1932 }
1933
1934 int
1935 udp_attach(struct socket *so, int proto, struct proc *p)
1936 {
1937 #pragma unused(proto)
1938 struct inpcb *inp;
1939 int error;
1940
1941 inp = sotoinpcb(so);
1942 if (inp != NULL) {
1943 panic("%s so=%p inp=%p\n", __func__, so, inp);
1944 /* NOTREACHED */
1945 }
1946 error = in_pcballoc(so, &udbinfo, p);
1947 if (error != 0)
1948 return (error);
1949 error = soreserve(so, udp_sendspace, udp_recvspace);
1950 if (error != 0)
1951 return (error);
1952 inp = (struct inpcb *)so->so_pcb;
1953 inp->inp_vflag |= INP_IPV4;
1954 inp->inp_ip_ttl = ip_defttl;
1955 if (nstat_collect)
1956 nstat_udp_new_pcb(inp);
1957 return (0);
1958 }
1959
1960 int
1961 udp_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
1962 {
1963 struct inpcb *inp;
1964 int error;
1965
1966 if (nam->sa_family != 0 && nam->sa_family != AF_INET &&
1967 nam->sa_family != AF_INET6)
1968 return (EAFNOSUPPORT);
1969
1970 inp = sotoinpcb(so);
1971 if (inp == NULL)
1972 return (EINVAL);
1973 error = in_pcbbind(inp, nam, p);
1974
1975 #if NECP
1976 /* Update NECP client with bind result if not in middle of connect */
1977 if (error == 0 &&
1978 (inp->inp_flags2 & INP2_CONNECT_IN_PROGRESS) &&
1979 !uuid_is_null(inp->necp_client_uuid)) {
1980 socket_unlock(so, 0);
1981 necp_client_assign_from_socket(so->last_pid, inp->necp_client_uuid, inp);
1982 socket_lock(so, 0);
1983 }
1984 #endif /* NECP */
1985
1986 return (error);
1987 }
1988
1989 int
1990 udp_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
1991 {
1992 struct inpcb *inp;
1993 int error;
1994
1995 inp = sotoinpcb(so);
1996 if (inp == NULL)
1997 return (EINVAL);
1998 if (inp->inp_faddr.s_addr != INADDR_ANY)
1999 return (EISCONN);
2000
2001 if (!(so->so_flags1 & SOF1_CONNECT_COUNTED)) {
2002 so->so_flags1 |= SOF1_CONNECT_COUNTED;
2003 INC_ATOMIC_INT64_LIM(net_api_stats.nas_socket_inet_dgram_connected);
2004 }
2005
2006 #if NECP
2007 #if FLOW_DIVERT
2008 if (necp_socket_should_use_flow_divert(inp)) {
2009 uint32_t fd_ctl_unit =
2010 necp_socket_get_flow_divert_control_unit(inp);
2011 if (fd_ctl_unit > 0) {
2012 error = flow_divert_pcb_init(so, fd_ctl_unit);
2013 if (error == 0) {
2014 error = flow_divert_connect_out(so, nam, p);
2015 }
2016 } else {
2017 error = ENETDOWN;
2018 }
2019 return (error);
2020 }
2021 #endif /* FLOW_DIVERT */
2022 #endif /* NECP */
2023
2024 error = in_pcbconnect(inp, nam, p, IFSCOPE_NONE, NULL);
2025 if (error == 0) {
2026 #if NECP
2027 /* Update NECP client with connected five-tuple */
2028 if (!uuid_is_null(inp->necp_client_uuid)) {
2029 socket_unlock(so, 0);
2030 necp_client_assign_from_socket(so->last_pid, inp->necp_client_uuid, inp);
2031 socket_lock(so, 0);
2032 }
2033 #endif /* NECP */
2034
2035 soisconnected(so);
2036 if (inp->inp_flowhash == 0)
2037 inp->inp_flowhash = inp_calc_flowhash(inp);
2038 }
2039 return (error);
2040 }
2041
2042 int
2043 udp_connectx_common(struct socket *so, int af, struct sockaddr *src, struct sockaddr *dst,
2044 struct proc *p, uint32_t ifscope, sae_associd_t aid, sae_connid_t *pcid,
2045 uint32_t flags, void *arg, uint32_t arglen,
2046 struct uio *uio, user_ssize_t *bytes_written)
2047 {
2048 #pragma unused(aid, flags, arg, arglen)
2049 struct inpcb *inp = sotoinpcb(so);
2050 int error = 0;
2051 user_ssize_t datalen = 0;
2052
2053 if (inp == NULL)
2054 return (EINVAL);
2055
2056 VERIFY(dst != NULL);
2057
2058 ASSERT(!(inp->inp_flags2 & INP2_CONNECT_IN_PROGRESS));
2059 inp->inp_flags2 |= INP2_CONNECT_IN_PROGRESS;
2060
2061 #if NECP
2062 inp_update_necp_policy(inp, src, dst, ifscope);
2063 #endif /* NECP */
2064
2065 /* bind socket to the specified interface, if requested */
2066 if (ifscope != IFSCOPE_NONE &&
2067 (error = inp_bindif(inp, ifscope, NULL)) != 0) {
2068 goto done;
2069 }
2070
2071 /* if source address and/or port is specified, bind to it */
2072 if (src != NULL) {
2073 error = sobindlock(so, src, 0); /* already locked */
2074 if (error != 0) {
2075 goto done;
2076 }
2077 }
2078
2079 switch (af) {
2080 case AF_INET:
2081 error = udp_connect(so, dst, p);
2082 break;
2083 #if INET6
2084 case AF_INET6:
2085 error = udp6_connect(so, dst, p);
2086 break;
2087 #endif /* INET6 */
2088 default:
2089 VERIFY(0);
2090 /* NOTREACHED */
2091 }
2092
2093 if (error != 0) {
2094 goto done;
2095 }
2096
2097 /*
2098 * If there is data, copy it. DATA_IDEMPOTENT is ignored.
2099 * CONNECT_RESUME_ON_READ_WRITE is ignored.
2100 */
2101 if (uio != NULL) {
2102 socket_unlock(so, 0);
2103
2104 VERIFY(bytes_written != NULL);
2105
2106 datalen = uio_resid(uio);
2107 error = so->so_proto->pr_usrreqs->pru_sosend(so, NULL,
2108 (uio_t)uio, NULL, NULL, 0);
2109 socket_lock(so, 0);
2110
2111 /* If error returned is EMSGSIZE, for example, disconnect */
2112 if (error == 0 || error == EWOULDBLOCK)
2113 *bytes_written = datalen - uio_resid(uio);
2114 else
2115 (void) so->so_proto->pr_usrreqs->pru_disconnectx(so,
2116 SAE_ASSOCID_ANY, SAE_CONNID_ANY);
2117 /*
2118 * mask the EWOULDBLOCK error so that the caller
2119 * knows that atleast the connect was successful.
2120 */
2121 if (error == EWOULDBLOCK)
2122 error = 0;
2123 }
2124
2125 if (error == 0 && pcid != NULL)
2126 *pcid = 1; /* there is only 1 connection for UDP */
2127
2128 done:
2129 inp->inp_flags2 &= ~INP2_CONNECT_IN_PROGRESS;
2130 return (error);
2131 }
2132
2133 int
2134 udp_connectx(struct socket *so, struct sockaddr *src,
2135 struct sockaddr *dst, struct proc *p, uint32_t ifscope,
2136 sae_associd_t aid, sae_connid_t *pcid, uint32_t flags, void *arg,
2137 uint32_t arglen, struct uio *uio, user_ssize_t *bytes_written)
2138 {
2139 return (udp_connectx_common(so, AF_INET, src, dst,
2140 p, ifscope, aid, pcid, flags, arg, arglen, uio, bytes_written));
2141 }
2142
2143 int
2144 udp_detach(struct socket *so)
2145 {
2146 struct inpcb *inp;
2147
2148 inp = sotoinpcb(so);
2149 if (inp == NULL) {
2150 panic("%s: so=%p null inp\n", __func__, so);
2151 /* NOTREACHED */
2152 }
2153
2154 /*
2155 * If this is a socket that does not want to wakeup the device
2156 * for it's traffic, the application might be waiting for
2157 * close to complete before going to sleep. Send a notification
2158 * for this kind of sockets
2159 */
2160 if (so->so_options & SO_NOWAKEFROMSLEEP)
2161 socket_post_kev_msg_closed(so);
2162
2163 in_pcbdetach(inp);
2164 inp->inp_state = INPCB_STATE_DEAD;
2165 return (0);
2166 }
2167
2168 int
2169 udp_disconnect(struct socket *so)
2170 {
2171 struct inpcb *inp;
2172
2173 inp = sotoinpcb(so);
2174 if (inp == NULL
2175 #if NECP
2176 || (necp_socket_should_use_flow_divert(inp))
2177 #endif /* NECP */
2178 )
2179 return (inp == NULL ? EINVAL : EPROTOTYPE);
2180 if (inp->inp_faddr.s_addr == INADDR_ANY)
2181 return (ENOTCONN);
2182
2183 in_pcbdisconnect(inp);
2184
2185 /* reset flow controlled state, just in case */
2186 inp_reset_fc_state(inp);
2187
2188 inp->inp_laddr.s_addr = INADDR_ANY;
2189 so->so_state &= ~SS_ISCONNECTED; /* XXX */
2190 inp->inp_last_outifp = NULL;
2191
2192 return (0);
2193 }
2194
2195 int
2196 udp_disconnectx(struct socket *so, sae_associd_t aid, sae_connid_t cid)
2197 {
2198 #pragma unused(cid)
2199 if (aid != SAE_ASSOCID_ANY && aid != SAE_ASSOCID_ALL)
2200 return (EINVAL);
2201
2202 return (udp_disconnect(so));
2203 }
2204
2205 int
2206 udp_send(struct socket *so, int flags, struct mbuf *m,
2207 struct sockaddr *addr, struct mbuf *control, struct proc *p)
2208 {
2209 #ifndef FLOW_DIVERT
2210 #pragma unused(flags)
2211 #endif /* !(FLOW_DIVERT) */
2212 struct inpcb *inp;
2213
2214 inp = sotoinpcb(so);
2215 if (inp == NULL) {
2216 if (m != NULL)
2217 m_freem(m);
2218 if (control != NULL)
2219 m_freem(control);
2220 return (EINVAL);
2221 }
2222
2223 #if NECP
2224 #if FLOW_DIVERT
2225 if (necp_socket_should_use_flow_divert(inp)) {
2226 /* Implicit connect */
2227 return (flow_divert_implicit_data_out(so, flags, m, addr,
2228 control, p));
2229 }
2230 #endif /* FLOW_DIVERT */
2231 #endif /* NECP */
2232
2233 return (udp_output(inp, m, addr, control, p));
2234 }
2235
2236 int
2237 udp_shutdown(struct socket *so)
2238 {
2239 struct inpcb *inp;
2240
2241 inp = sotoinpcb(so);
2242 if (inp == NULL)
2243 return (EINVAL);
2244 socantsendmore(so);
2245 return (0);
2246 }
2247
2248 int
2249 udp_lock(struct socket *so, int refcount, void *debug)
2250 {
2251 void *lr_saved;
2252
2253 if (debug == NULL)
2254 lr_saved = __builtin_return_address(0);
2255 else
2256 lr_saved = debug;
2257
2258 if (so->so_pcb != NULL) {
2259 LCK_MTX_ASSERT(&((struct inpcb *)so->so_pcb)->inpcb_mtx,
2260 LCK_MTX_ASSERT_NOTOWNED);
2261 lck_mtx_lock(&((struct inpcb *)so->so_pcb)->inpcb_mtx);
2262 } else {
2263 panic("%s: so=%p NO PCB! lr=%p lrh= %s\n", __func__,
2264 so, lr_saved, solockhistory_nr(so));
2265 /* NOTREACHED */
2266 }
2267 if (refcount)
2268 so->so_usecount++;
2269
2270 so->lock_lr[so->next_lock_lr] = lr_saved;
2271 so->next_lock_lr = (so->next_lock_lr+1) % SO_LCKDBG_MAX;
2272 return (0);
2273 }
2274
2275 int
2276 udp_unlock(struct socket *so, int refcount, void *debug)
2277 {
2278 void *lr_saved;
2279
2280 if (debug == NULL)
2281 lr_saved = __builtin_return_address(0);
2282 else
2283 lr_saved = debug;
2284
2285 if (refcount) {
2286 VERIFY(so->so_usecount > 0);
2287 so->so_usecount--;
2288 }
2289 if (so->so_pcb == NULL) {
2290 panic("%s: so=%p NO PCB! lr=%p lrh= %s\n", __func__,
2291 so, lr_saved, solockhistory_nr(so));
2292 /* NOTREACHED */
2293 } else {
2294 LCK_MTX_ASSERT(&((struct inpcb *)so->so_pcb)->inpcb_mtx,
2295 LCK_MTX_ASSERT_OWNED);
2296 so->unlock_lr[so->next_unlock_lr] = lr_saved;
2297 so->next_unlock_lr = (so->next_unlock_lr+1) % SO_LCKDBG_MAX;
2298 lck_mtx_unlock(&((struct inpcb *)so->so_pcb)->inpcb_mtx);
2299 }
2300 return (0);
2301 }
2302
2303 lck_mtx_t *
2304 udp_getlock(struct socket *so, int flags)
2305 {
2306 #pragma unused(flags)
2307 struct inpcb *inp = sotoinpcb(so);
2308
2309 if (so->so_pcb == NULL) {
2310 panic("%s: so=%p NULL so_pcb lrh= %s\n", __func__,
2311 so, solockhistory_nr(so));
2312 /* NOTREACHED */
2313 }
2314 return (&inp->inpcb_mtx);
2315 }
2316
2317 /*
2318 * UDP garbage collector callback (inpcb_timer_func_t).
2319 *
2320 * Returns > 0 to keep timer active.
2321 */
2322 static void
2323 udp_gc(struct inpcbinfo *ipi)
2324 {
2325 struct inpcb *inp, *inpnxt;
2326 struct socket *so;
2327
2328 if (lck_rw_try_lock_exclusive(ipi->ipi_lock) == FALSE) {
2329 if (udp_gc_done == TRUE) {
2330 udp_gc_done = FALSE;
2331 /* couldn't get the lock, must lock next time */
2332 atomic_add_32(&ipi->ipi_gc_req.intimer_fast, 1);
2333 return;
2334 }
2335 lck_rw_lock_exclusive(ipi->ipi_lock);
2336 }
2337
2338 udp_gc_done = TRUE;
2339
2340 for (inp = udb.lh_first; inp != NULL; inp = inpnxt) {
2341 inpnxt = inp->inp_list.le_next;
2342
2343 /*
2344 * Skip unless it's STOPUSING; garbage collector will
2345 * be triggered by in_pcb_checkstate() upon setting
2346 * wantcnt to that value. If the PCB is already dead,
2347 * keep gc active to anticipate wantcnt changing.
2348 */
2349 if (inp->inp_wantcnt != WNT_STOPUSING)
2350 continue;
2351
2352 /*
2353 * Skip if busy, no hurry for cleanup. Keep gc active
2354 * and try the lock again during next round.
2355 */
2356 if (!socket_try_lock(inp->inp_socket)) {
2357 atomic_add_32(&ipi->ipi_gc_req.intimer_fast, 1);
2358 continue;
2359 }
2360
2361 /*
2362 * Keep gc active unless usecount is 0.
2363 */
2364 so = inp->inp_socket;
2365 if (so->so_usecount == 0) {
2366 if (inp->inp_state != INPCB_STATE_DEAD) {
2367 #if INET6
2368 if (SOCK_CHECK_DOM(so, PF_INET6))
2369 in6_pcbdetach(inp);
2370 else
2371 #endif /* INET6 */
2372 in_pcbdetach(inp);
2373 }
2374 in_pcbdispose(inp);
2375 } else {
2376 socket_unlock(so, 0);
2377 atomic_add_32(&ipi->ipi_gc_req.intimer_fast, 1);
2378 }
2379 }
2380 lck_rw_done(ipi->ipi_lock);
2381 }
2382
2383 static int
2384 udp_getstat SYSCTL_HANDLER_ARGS
2385 {
2386 #pragma unused(oidp, arg1, arg2)
2387 if (req->oldptr == USER_ADDR_NULL)
2388 req->oldlen = (size_t)sizeof (struct udpstat);
2389
2390 return (SYSCTL_OUT(req, &udpstat, MIN(sizeof (udpstat), req->oldlen)));
2391 }
2392
2393 void
2394 udp_in_cksum_stats(u_int32_t len)
2395 {
2396 udpstat.udps_rcv_swcsum++;
2397 udpstat.udps_rcv_swcsum_bytes += len;
2398 }
2399
2400 void
2401 udp_out_cksum_stats(u_int32_t len)
2402 {
2403 udpstat.udps_snd_swcsum++;
2404 udpstat.udps_snd_swcsum_bytes += len;
2405 }
2406
2407 #if INET6
2408 void
2409 udp_in6_cksum_stats(u_int32_t len)
2410 {
2411 udpstat.udps_rcv6_swcsum++;
2412 udpstat.udps_rcv6_swcsum_bytes += len;
2413 }
2414
2415 void
2416 udp_out6_cksum_stats(u_int32_t len)
2417 {
2418 udpstat.udps_snd6_swcsum++;
2419 udpstat.udps_snd6_swcsum_bytes += len;
2420 }
2421 #endif /* INET6 */
2422
2423 /*
2424 * Checksum extended UDP header and data.
2425 */
2426 static int
2427 udp_input_checksum(struct mbuf *m, struct udphdr *uh, int off, int ulen)
2428 {
2429 struct ifnet *ifp = m->m_pkthdr.rcvif;
2430 struct ip *ip = mtod(m, struct ip *);
2431 struct ipovly *ipov = (struct ipovly *)ip;
2432
2433 if (uh->uh_sum == 0) {
2434 udpstat.udps_nosum++;
2435 return (0);
2436 }
2437
2438 /* ip_stripoptions() must have been called before we get here */
2439 ASSERT((ip->ip_hl << 2) == sizeof (*ip));
2440
2441 if ((hwcksum_rx || (ifp->if_flags & IFF_LOOPBACK) ||
2442 (m->m_pkthdr.pkt_flags & PKTF_LOOP)) &&
2443 (m->m_pkthdr.csum_flags & CSUM_DATA_VALID)) {
2444 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
2445 uh->uh_sum = m->m_pkthdr.csum_rx_val;
2446 } else {
2447 uint32_t sum = m->m_pkthdr.csum_rx_val;
2448 uint32_t start = m->m_pkthdr.csum_rx_start;
2449 int32_t trailer = (m_pktlen(m) - (off + ulen));
2450
2451 /*
2452 * Perform 1's complement adjustment of octets
2453 * that got included/excluded in the hardware-
2454 * calculated checksum value. Ignore cases
2455 * where the value already includes the entire
2456 * IP header span, as the sum for those octets
2457 * would already be 0 by the time we get here;
2458 * IP has already performed its header checksum
2459 * checks. If we do need to adjust, restore
2460 * the original fields in the IP header when
2461 * computing the adjustment value. Also take
2462 * care of any trailing bytes and subtract out
2463 * their partial sum.
2464 */
2465 ASSERT(trailer >= 0);
2466 if ((m->m_pkthdr.csum_flags & CSUM_PARTIAL) &&
2467 ((start != 0 && start != off) || trailer != 0)) {
2468 uint32_t swbytes = (uint32_t)trailer;
2469
2470 if (start < off) {
2471 ip->ip_len += sizeof (*ip);
2472 #if BYTE_ORDER != BIG_ENDIAN
2473 HTONS(ip->ip_len);
2474 HTONS(ip->ip_off);
2475 #endif /* BYTE_ORDER != BIG_ENDIAN */
2476 }
2477 /* callee folds in sum */
2478 sum = m_adj_sum16(m, start, off, ulen, sum);
2479 if (off > start)
2480 swbytes += (off - start);
2481 else
2482 swbytes += (start - off);
2483
2484 if (start < off) {
2485 #if BYTE_ORDER != BIG_ENDIAN
2486 NTOHS(ip->ip_off);
2487 NTOHS(ip->ip_len);
2488 #endif /* BYTE_ORDER != BIG_ENDIAN */
2489 ip->ip_len -= sizeof (*ip);
2490 }
2491
2492 if (swbytes != 0)
2493 udp_in_cksum_stats(swbytes);
2494 if (trailer != 0)
2495 m_adj(m, -trailer);
2496 }
2497
2498 /* callee folds in sum */
2499 uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
2500 ip->ip_dst.s_addr, sum + htonl(ulen + IPPROTO_UDP));
2501 }
2502 uh->uh_sum ^= 0xffff;
2503 } else {
2504 uint16_t ip_sum;
2505 char b[9];
2506
2507 bcopy(ipov->ih_x1, b, sizeof (ipov->ih_x1));
2508 bzero(ipov->ih_x1, sizeof (ipov->ih_x1));
2509 ip_sum = ipov->ih_len;
2510 ipov->ih_len = uh->uh_ulen;
2511 uh->uh_sum = in_cksum(m, ulen + sizeof (struct ip));
2512 bcopy(b, ipov->ih_x1, sizeof (ipov->ih_x1));
2513 ipov->ih_len = ip_sum;
2514
2515 udp_in_cksum_stats(ulen);
2516 }
2517
2518 if (uh->uh_sum != 0) {
2519 udpstat.udps_badsum++;
2520 IF_UDP_STATINC(ifp, badchksum);
2521 return (-1);
2522 }
2523
2524 return (0);
2525 }
2526
2527 void
2528 udp_fill_keepalive_offload_frames(ifnet_t ifp,
2529 struct ifnet_keepalive_offload_frame *frames_array,
2530 u_int32_t frames_array_count, size_t frame_data_offset,
2531 u_int32_t *used_frames_count)
2532 {
2533 struct inpcb *inp;
2534 inp_gen_t gencnt;
2535 u_int32_t frame_index = *used_frames_count;
2536
2537 if (ifp == NULL || frames_array == NULL ||
2538 frames_array_count == 0 ||
2539 frame_index >= frames_array_count ||
2540 frame_data_offset >= IFNET_KEEPALIVE_OFFLOAD_FRAME_DATA_SIZE)
2541 return;
2542
2543 lck_rw_lock_shared(udbinfo.ipi_lock);
2544 gencnt = udbinfo.ipi_gencnt;
2545 LIST_FOREACH(inp, udbinfo.ipi_listhead, inp_list) {
2546 struct socket *so;
2547 u_int8_t *data;
2548 struct ifnet_keepalive_offload_frame *frame;
2549 struct mbuf *m = NULL;
2550
2551 if (frame_index >= frames_array_count)
2552 break;
2553
2554 if (inp->inp_gencnt > gencnt ||
2555 inp->inp_state == INPCB_STATE_DEAD)
2556 continue;
2557
2558 if ((so = inp->inp_socket) == NULL ||
2559 (so->so_state & SS_DEFUNCT))
2560 continue;
2561 /*
2562 * check for keepalive offload flag without socket
2563 * lock to avoid a deadlock
2564 */
2565 if (!(inp->inp_flags2 & INP2_KEEPALIVE_OFFLOAD)) {
2566 continue;
2567 }
2568
2569 udp_lock(so, 1, 0);
2570 if (!(inp->inp_vflag & (INP_IPV4 | INP_IPV6))) {
2571 udp_unlock(so, 1, 0);
2572 continue;
2573 }
2574 if ((inp->inp_vflag & INP_IPV4) &&
2575 (inp->inp_laddr.s_addr == INADDR_ANY ||
2576 inp->inp_faddr.s_addr == INADDR_ANY)) {
2577 udp_unlock(so, 1, 0);
2578 continue;
2579 }
2580 if ((inp->inp_vflag & INP_IPV6) &&
2581 (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ||
2582 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))) {
2583 udp_unlock(so, 1, 0);
2584 continue;
2585 }
2586 if (inp->inp_lport == 0 || inp->inp_fport == 0) {
2587 udp_unlock(so, 1, 0);
2588 continue;
2589 }
2590 if (inp->inp_last_outifp == NULL ||
2591 inp->inp_last_outifp->if_index != ifp->if_index) {
2592 udp_unlock(so, 1, 0);
2593 continue;
2594 }
2595 if ((inp->inp_vflag & INP_IPV4)) {
2596 if ((frame_data_offset + sizeof(struct udpiphdr) +
2597 inp->inp_keepalive_datalen) >
2598 IFNET_KEEPALIVE_OFFLOAD_FRAME_DATA_SIZE) {
2599 udp_unlock(so, 1, 0);
2600 continue;
2601 }
2602 if ((sizeof(struct udpiphdr) +
2603 inp->inp_keepalive_datalen) > _MHLEN) {
2604 udp_unlock(so, 1, 0);
2605 continue;
2606 }
2607 } else {
2608 if ((frame_data_offset + sizeof(struct ip6_hdr) +
2609 sizeof(struct udphdr) +
2610 inp->inp_keepalive_datalen) >
2611 IFNET_KEEPALIVE_OFFLOAD_FRAME_DATA_SIZE) {
2612 udp_unlock(so, 1, 0);
2613 continue;
2614 }
2615 if ((sizeof(struct ip6_hdr) + sizeof(struct udphdr) +
2616 inp->inp_keepalive_datalen) > _MHLEN) {
2617 udp_unlock(so, 1, 0);
2618 continue;
2619 }
2620 }
2621 MGETHDR(m, M_WAIT, MT_HEADER);
2622 if (m == NULL) {
2623 udp_unlock(so, 1, 0);
2624 continue;
2625 }
2626 /*
2627 * This inp has all the information that is needed to
2628 * generate an offload frame.
2629 */
2630 if (inp->inp_vflag & INP_IPV4) {
2631 struct ip *ip;
2632 struct udphdr *udp;
2633
2634 frame = &frames_array[frame_index];
2635 frame->length = frame_data_offset +
2636 sizeof(struct udpiphdr) +
2637 inp->inp_keepalive_datalen;
2638 frame->ether_type =
2639 IFNET_KEEPALIVE_OFFLOAD_FRAME_ETHERTYPE_IPV4;
2640 frame->interval = inp->inp_keepalive_interval;
2641 switch (inp->inp_keepalive_type) {
2642 case UDP_KEEPALIVE_OFFLOAD_TYPE_AIRPLAY:
2643 frame->type =
2644 IFNET_KEEPALIVE_OFFLOAD_FRAME_AIRPLAY;
2645 break;
2646 default:
2647 break;
2648 }
2649 data = mtod(m, u_int8_t *);
2650 bzero(data, sizeof(struct udpiphdr));
2651 ip = (__typeof__(ip))(void *)data;
2652 udp = (__typeof__(udp))(void *) (data +
2653 sizeof(struct ip));
2654 m->m_len = sizeof(struct udpiphdr);
2655 data = data + sizeof(struct udpiphdr);
2656 if (inp->inp_keepalive_datalen > 0 &&
2657 inp->inp_keepalive_data != NULL) {
2658 bcopy(inp->inp_keepalive_data, data,
2659 inp->inp_keepalive_datalen);
2660 m->m_len += inp->inp_keepalive_datalen;
2661 }
2662 m->m_pkthdr.len = m->m_len;
2663
2664 ip->ip_v = IPVERSION;
2665 ip->ip_hl = (sizeof(struct ip) >> 2);
2666 ip->ip_p = IPPROTO_UDP;
2667 ip->ip_len = htons(sizeof(struct udpiphdr) +
2668 (u_short)inp->inp_keepalive_datalen);
2669 ip->ip_ttl = inp->inp_ip_ttl;
2670 ip->ip_tos |= (inp->inp_ip_tos & ~IPTOS_ECN_MASK);
2671 ip->ip_src = inp->inp_laddr;
2672 ip->ip_dst = inp->inp_faddr;
2673 ip->ip_sum = in_cksum_hdr_opt(ip);
2674
2675 udp->uh_sport = inp->inp_lport;
2676 udp->uh_dport = inp->inp_fport;
2677 udp->uh_ulen = htons(sizeof(struct udphdr) +
2678 (u_short)inp->inp_keepalive_datalen);
2679
2680 if (!(inp->inp_flags & INP_UDP_NOCKSUM)) {
2681 udp->uh_sum = in_pseudo(ip->ip_src.s_addr,
2682 ip->ip_dst.s_addr,
2683 htons(sizeof(struct udphdr) +
2684 (u_short)inp->inp_keepalive_datalen +
2685 IPPROTO_UDP));
2686 m->m_pkthdr.csum_flags =
2687 (CSUM_UDP|CSUM_ZERO_INVERT);
2688 m->m_pkthdr.csum_data = offsetof(struct udphdr,
2689 uh_sum);
2690 }
2691 m->m_pkthdr.pkt_proto = IPPROTO_UDP;
2692 in_delayed_cksum(m);
2693 bcopy(m->m_data, frame->data + frame_data_offset,
2694 m->m_len);
2695 } else {
2696 struct ip6_hdr *ip6;
2697 struct udphdr *udp6;
2698
2699 VERIFY(inp->inp_vflag & INP_IPV6);
2700 frame = &frames_array[frame_index];
2701 frame->length = frame_data_offset +
2702 sizeof(struct ip6_hdr) +
2703 sizeof(struct udphdr) +
2704 inp->inp_keepalive_datalen;
2705 frame->ether_type =
2706 IFNET_KEEPALIVE_OFFLOAD_FRAME_ETHERTYPE_IPV6;
2707 frame->interval = inp->inp_keepalive_interval;
2708 switch (inp->inp_keepalive_type) {
2709 case UDP_KEEPALIVE_OFFLOAD_TYPE_AIRPLAY:
2710 frame->type =
2711 IFNET_KEEPALIVE_OFFLOAD_FRAME_AIRPLAY;
2712 break;
2713 default:
2714 break;
2715 }
2716 data = mtod(m, u_int8_t *);
2717 bzero(data, sizeof(struct ip6_hdr) + sizeof(struct udphdr));
2718 ip6 = (__typeof__(ip6))(void *)data;
2719 udp6 = (__typeof__(udp6))(void *)(data +
2720 sizeof(struct ip6_hdr));
2721 m->m_len = sizeof(struct ip6_hdr) +
2722 sizeof(struct udphdr);
2723 data = data + (sizeof(struct ip6_hdr) +
2724 sizeof(struct udphdr));
2725 if (inp->inp_keepalive_datalen > 0 &&
2726 inp->inp_keepalive_data != NULL) {
2727 bcopy(inp->inp_keepalive_data, data,
2728 inp->inp_keepalive_datalen);
2729 m->m_len += inp->inp_keepalive_datalen;
2730 }
2731 m->m_pkthdr.len = m->m_len;
2732 ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK;
2733 ip6->ip6_flow = ip6->ip6_flow & ~IPV6_FLOW_ECN_MASK;
2734 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2735 ip6->ip6_vfc |= IPV6_VERSION;
2736 ip6->ip6_nxt = IPPROTO_UDP;
2737 ip6->ip6_hlim = ip6_defhlim;
2738 ip6->ip6_plen = htons(sizeof(struct udphdr) +
2739 (u_short)inp->inp_keepalive_datalen);
2740 ip6->ip6_src = inp->in6p_laddr;
2741 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
2742 ip6->ip6_src.s6_addr16[1] = 0;
2743
2744 ip6->ip6_dst = inp->in6p_faddr;
2745 if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
2746 ip6->ip6_dst.s6_addr16[1] = 0;
2747
2748 udp6->uh_sport = inp->in6p_lport;
2749 udp6->uh_dport = inp->in6p_fport;
2750 udp6->uh_ulen = htons(sizeof(struct udphdr) +
2751 (u_short)inp->inp_keepalive_datalen);
2752 if (!(inp->inp_flags & INP_UDP_NOCKSUM)) {
2753 udp6->uh_sum = in6_pseudo(&ip6->ip6_src,
2754 &ip6->ip6_dst,
2755 htonl(sizeof(struct udphdr) +
2756 (u_short)inp->inp_keepalive_datalen +
2757 IPPROTO_UDP));
2758 m->m_pkthdr.csum_flags =
2759 (CSUM_UDPIPV6|CSUM_ZERO_INVERT);
2760 m->m_pkthdr.csum_data = offsetof(struct udphdr,
2761 uh_sum);
2762 }
2763 m->m_pkthdr.pkt_proto = IPPROTO_UDP;
2764 in6_delayed_cksum(m);
2765 bcopy(m->m_data, frame->data + frame_data_offset,
2766 m->m_len);
2767 }
2768 if (m != NULL) {
2769 m_freem(m);
2770 m = NULL;
2771 }
2772 frame_index++;
2773 udp_unlock(so, 1, 0);
2774 }
2775 lck_rw_done(udbinfo.ipi_lock);
2776 *used_frames_count = frame_index;
2777 }