1 | /* |
2 | * Copyright (c) 2000-2021, 2023 Apple Inc. All rights reserved. |
3 | * |
4 | * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ |
5 | * |
6 | * This file contains Original Code and/or Modifications of Original Code |
7 | * as defined in and that are subject to the Apple Public Source License |
8 | * Version 2.0 (the 'License'). You may not use this file except in |
9 | * compliance with the License. The rights granted to you under the License |
10 | * may not be used to create, or enable the creation or redistribution of, |
11 | * unlawful or unlicensed copies of an Apple operating system, or to |
12 | * circumvent, violate, or enable the circumvention or violation of, any |
13 | * terms of an Apple operating system software license agreement. |
14 | * |
15 | * Please obtain a copy of the License at |
16 | * http://www.opensource.apple.com/apsl/ and read it before using this file. |
17 | * |
18 | * The Original Code and all software distributed under the License are |
19 | * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER |
20 | * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, |
21 | * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, |
22 | * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. |
23 | * Please see the License for the specific language governing rights and |
24 | * limitations under the License. |
25 | * |
26 | * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ |
27 | */ |
28 | /* |
29 | * Copyright (c) 1982, 1986, 1988, 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/net_api_stats.h> |
86 | |
87 | #include <netinet/in.h> |
88 | #include <netinet/in_systm.h> |
89 | #include <netinet/in_tclass.h> |
90 | #include <netinet/ip.h> |
91 | #include <netinet/ip6.h> |
92 | #include <netinet/in_pcb.h> |
93 | #include <netinet/in_var.h> |
94 | #include <netinet/ip_var.h> |
95 | #include <netinet6/in6_pcb.h> |
96 | #include <netinet6/ip6_var.h> |
97 | #include <netinet6/udp6_var.h> |
98 | #include <netinet/ip_icmp.h> |
99 | #include <netinet/icmp_var.h> |
100 | #include <netinet/udp.h> |
101 | #include <netinet/udp_var.h> |
102 | #include <netinet/udp_log.h> |
103 | #include <sys/kdebug.h> |
104 | |
105 | #if IPSEC |
106 | #include <netinet6/ipsec.h> |
107 | #include <netinet6/esp.h> |
108 | #include <netkey/key.h> |
109 | extern int ipsec_bypass; |
110 | extern int esp_udp_encap_port; |
111 | #endif /* IPSEC */ |
112 | |
113 | #if NECP |
114 | #include <net/necp.h> |
115 | #endif /* NECP */ |
116 | |
117 | #if FLOW_DIVERT |
118 | #include <netinet/flow_divert.h> |
119 | #endif /* FLOW_DIVERT */ |
120 | |
121 | #if CONTENT_FILTER |
122 | #include <net/content_filter.h> |
123 | #endif /* CONTENT_FILTER */ |
124 | |
125 | #if SKYWALK |
126 | #include <skywalk/core/skywalk_var.h> |
127 | #endif /* SKYWALK */ |
128 | |
129 | #include <net/sockaddr_utils.h> |
130 | |
131 | #define DBG_LAYER_IN_BEG NETDBG_CODE(DBG_NETUDP, 0) |
132 | #define DBG_LAYER_IN_END NETDBG_CODE(DBG_NETUDP, 2) |
133 | #define DBG_LAYER_OUT_BEG NETDBG_CODE(DBG_NETUDP, 1) |
134 | #define DBG_LAYER_OUT_END NETDBG_CODE(DBG_NETUDP, 3) |
135 | #define DBG_FNC_UDP_INPUT NETDBG_CODE(DBG_NETUDP, (5 << 8)) |
136 | #define DBG_FNC_UDP_OUTPUT NETDBG_CODE(DBG_NETUDP, (6 << 8) | 1) |
137 | |
138 | /* |
139 | * UDP protocol implementation. |
140 | * Per RFC 768, August, 1980. |
141 | */ |
142 | #ifndef COMPAT_42 |
143 | static int udpcksum = 1; |
144 | #else |
145 | static int udpcksum = 0; /* XXX */ |
146 | #endif |
147 | SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, |
148 | CTLFLAG_RW | CTLFLAG_LOCKED, &udpcksum, 0, "" ); |
149 | |
150 | int udp_log_in_vain = 0; |
151 | SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW | CTLFLAG_LOCKED, |
152 | &udp_log_in_vain, 0, "Log all incoming UDP packets" ); |
153 | |
154 | static int blackhole = 0; |
155 | SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW | CTLFLAG_LOCKED, |
156 | &blackhole, 0, "Do not send port unreachables for refused connects" ); |
157 | |
158 | static KALLOC_TYPE_DEFINE(inpcbzone, struct inpcb, NET_KT_DEFAULT); |
159 | |
160 | struct inpcbhead udb; /* from udp_var.h */ |
161 | #define udb6 udb /* for KAME src sync over BSD*'s */ |
162 | struct inpcbinfo udbinfo; |
163 | |
164 | #ifndef UDBHASHSIZE |
165 | #define UDBHASHSIZE 16 |
166 | #endif |
167 | |
168 | /* Garbage collection performed during most recent udp_gc() run */ |
169 | static boolean_t udp_gc_done = FALSE; |
170 | |
171 | #define log_in_vain_log(a) { log a; } |
172 | |
173 | static int udp_getstat SYSCTL_HANDLER_ARGS; |
174 | struct udpstat udpstat; /* from udp_var.h */ |
175 | SYSCTL_PROC(_net_inet_udp, UDPCTL_STATS, stats, |
176 | CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, |
177 | 0, 0, udp_getstat, "S,udpstat" , |
178 | "UDP statistics (struct udpstat, netinet/udp_var.h)" ); |
179 | |
180 | SYSCTL_INT(_net_inet_udp, OID_AUTO, pcbcount, |
181 | CTLFLAG_RD | CTLFLAG_LOCKED, &udbinfo.ipi_count, 0, |
182 | "Number of active PCBs" ); |
183 | |
184 | __private_extern__ int udp_use_randomport = 1; |
185 | SYSCTL_INT(_net_inet_udp, OID_AUTO, randomize_ports, |
186 | CTLFLAG_RW | CTLFLAG_LOCKED, &udp_use_randomport, 0, |
187 | "Randomize UDP port numbers" ); |
188 | |
189 | struct udp_in6 { |
190 | struct sockaddr_in6 uin6_sin; |
191 | u_char uin6_init_done : 1; |
192 | }; |
193 | struct udp_ip6 { |
194 | struct ip6_hdr uip6_ip6; |
195 | u_char uip6_init_done : 1; |
196 | }; |
197 | |
198 | int udp_abort(struct socket *); |
199 | int udp_attach(struct socket *, int, struct proc *); |
200 | int udp_bind(struct socket *, struct sockaddr *, struct proc *); |
201 | int udp_connect(struct socket *, struct sockaddr *, struct proc *); |
202 | int udp_connectx(struct socket *, struct sockaddr *, |
203 | struct sockaddr *, struct proc *, uint32_t, sae_associd_t, |
204 | sae_connid_t *, uint32_t, void *, uint32_t, struct uio *, user_ssize_t *); |
205 | int udp_detach(struct socket *); |
206 | int udp_disconnect(struct socket *); |
207 | int udp_disconnectx(struct socket *, sae_associd_t, sae_connid_t); |
208 | int udp_send(struct socket *, int, struct mbuf *, struct sockaddr *, |
209 | struct mbuf *, struct proc *); |
210 | static void udp_append(struct inpcb *, struct ip *, struct mbuf *, int, |
211 | struct sockaddr_in *, struct udp_in6 *, struct udp_ip6 *, struct ifnet *); |
212 | static int udp_input_checksum(struct mbuf *, struct udphdr *, int, int); |
213 | int udp_output(struct inpcb *, struct mbuf *, struct sockaddr *, |
214 | struct mbuf *, struct proc *); |
215 | static void ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip); |
216 | static void udp_gc(struct inpcbinfo *); |
217 | static int udp_defunct(struct socket *); |
218 | |
219 | struct pr_usrreqs udp_usrreqs = { |
220 | .pru_abort = udp_abort, |
221 | .pru_attach = udp_attach, |
222 | .pru_bind = udp_bind, |
223 | .pru_connect = udp_connect, |
224 | .pru_connectx = udp_connectx, |
225 | .pru_control = in_control, |
226 | .pru_detach = udp_detach, |
227 | .pru_disconnect = udp_disconnect, |
228 | .pru_disconnectx = udp_disconnectx, |
229 | .pru_peeraddr = in_getpeeraddr, |
230 | .pru_send = udp_send, |
231 | .pru_shutdown = udp_shutdown, |
232 | .pru_sockaddr = in_getsockaddr, |
233 | .pru_sosend = sosend, |
234 | .pru_soreceive = soreceive, |
235 | .pru_defunct = udp_defunct, |
236 | }; |
237 | |
238 | void |
239 | udp_init(struct protosw *pp, struct domain *dp) |
240 | { |
241 | #pragma unused(dp) |
242 | static int udp_initialized = 0; |
243 | struct inpcbinfo *pcbinfo; |
244 | |
245 | VERIFY((pp->pr_flags & (PR_INITIALIZED | PR_ATTACHED)) == PR_ATTACHED); |
246 | |
247 | if (udp_initialized) { |
248 | return; |
249 | } |
250 | udp_initialized = 1; |
251 | uint32_t pool_size = (nmbclusters << MCLSHIFT) >> MBSHIFT; |
252 | if (pool_size >= 96) { |
253 | /* Improves 10GbE UDP performance. */ |
254 | udp_recvspace = 786896; |
255 | } |
256 | |
257 | if (PE_parse_boot_argn(arg_string: "udp_log" , arg_ptr: &udp_log_enable_flags, max_arg: sizeof(udp_log_enable_flags))) { |
258 | os_log(OS_LOG_DEFAULT, "udp_init: set udp_log_enable_flags to 0x%x" , udp_log_enable_flags); |
259 | } |
260 | |
261 | LIST_INIT(&udb); |
262 | udbinfo.ipi_listhead = &udb; |
263 | udbinfo.ipi_hashbase = hashinit(UDBHASHSIZE, M_PCB, |
264 | hashmask: &udbinfo.ipi_hashmask); |
265 | udbinfo.ipi_porthashbase = hashinit(UDBHASHSIZE, M_PCB, |
266 | hashmask: &udbinfo.ipi_porthashmask); |
267 | udbinfo.ipi_zone = inpcbzone; |
268 | |
269 | pcbinfo = &udbinfo; |
270 | /* |
271 | * allocate lock group and attribute for udp pcb mutexes |
272 | */ |
273 | pcbinfo->ipi_lock_grp = lck_grp_alloc_init(grp_name: "udppcb" , |
274 | LCK_GRP_ATTR_NULL); |
275 | lck_attr_setdefault(attr: &pcbinfo->ipi_lock_attr); |
276 | lck_rw_init(lck: &pcbinfo->ipi_lock, grp: pcbinfo->ipi_lock_grp, |
277 | attr: &pcbinfo->ipi_lock_attr); |
278 | |
279 | udbinfo.ipi_gc = udp_gc; |
280 | in_pcbinfo_attach(&udbinfo); |
281 | } |
282 | |
283 | void |
284 | udp_input(struct mbuf *m, int iphlen) |
285 | { |
286 | struct ip *ip; |
287 | struct udphdr *uh; |
288 | struct inpcb *inp; |
289 | struct mbuf *opts = NULL; |
290 | int len, isbroadcast; |
291 | struct ip save_ip; |
292 | struct sockaddr *append_sa = NULL; |
293 | struct sockaddr *append_da = NULL; |
294 | struct inpcbinfo *pcbinfo = &udbinfo; |
295 | struct sockaddr_in udp_in; |
296 | struct sockaddr_in udp_dst; |
297 | struct ip_moptions *imo = NULL; |
298 | int foundmembership = 0, ret = 0; |
299 | struct udp_in6 udp_in6; |
300 | struct udp_in6 udp_dst6; |
301 | struct udp_ip6 udp_ip6; |
302 | struct ifnet *ifp = m->m_pkthdr.rcvif; |
303 | boolean_t cell = IFNET_IS_CELLULAR(ifp); |
304 | boolean_t wifi = (!cell && IFNET_IS_WIFI(ifp)); |
305 | boolean_t wired = (!wifi && IFNET_IS_WIRED(ifp)); |
306 | u_int16_t pf_tag = 0; |
307 | boolean_t is_wake_pkt = false; |
308 | boolean_t check_cfil = cfil_filter_present(); |
309 | |
310 | SOCKADDR_ZERO(&udp_in, sizeof(udp_in)); |
311 | udp_in.sin_len = sizeof(struct sockaddr_in); |
312 | udp_in.sin_family = AF_INET; |
313 | bzero(s: &udp_in6, n: sizeof(udp_in6)); |
314 | udp_in6.uin6_sin.sin6_len = sizeof(struct sockaddr_in6); |
315 | udp_in6.uin6_sin.sin6_family = AF_INET6; |
316 | |
317 | if (m->m_flags & M_PKTHDR) { |
318 | pf_tag = m_pftag(m)->pftag_tag; |
319 | if (m->m_pkthdr.pkt_flags & PKTF_WAKE_PKT) { |
320 | is_wake_pkt = true; |
321 | } |
322 | } |
323 | |
324 | udpstat.udps_ipackets++; |
325 | |
326 | KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_START, 0, 0, 0, 0, 0); |
327 | |
328 | /* Expect 32-bit aligned data pointer on strict-align platforms */ |
329 | MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m); |
330 | |
331 | m_add_crumb(m, PKT_CRUMB_UDP_INPUT); |
332 | |
333 | /* |
334 | * Strip IP options, if any; should skip this, |
335 | * make available to user, and use on returned packets, |
336 | * but we don't yet have a way to check the checksum |
337 | * with options still present. |
338 | */ |
339 | if (iphlen > sizeof(struct ip)) { |
340 | ip_stripoptions(m); |
341 | iphlen = sizeof(struct ip); |
342 | } |
343 | |
344 | /* |
345 | * Get IP and UDP header together in first mbuf. |
346 | */ |
347 | ip = mtod(m, struct ip *); |
348 | if (m->m_len < iphlen + sizeof(struct udphdr)) { |
349 | m = m_pullup(m, iphlen + sizeof(struct udphdr)); |
350 | if (m == NULL) { |
351 | udpstat.udps_hdrops++; |
352 | KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, |
353 | 0, 0, 0, 0, 0); |
354 | return; |
355 | } |
356 | ip = mtod(m, struct ip *); |
357 | } |
358 | uh = (struct udphdr *)(void *)((caddr_t)ip + iphlen); |
359 | |
360 | /* destination port of 0 is illegal, based on RFC768. */ |
361 | if (uh->uh_dport == 0) { |
362 | IF_UDP_STATINC(ifp, port0); |
363 | goto bad; |
364 | } |
365 | |
366 | KERNEL_DEBUG(DBG_LAYER_IN_BEG, uh->uh_dport, uh->uh_sport, |
367 | ip->ip_src.s_addr, ip->ip_dst.s_addr, uh->uh_ulen); |
368 | |
369 | /* |
370 | * Make mbuf data length reflect UDP length. |
371 | * If not enough data to reflect UDP length, drop. |
372 | */ |
373 | len = ntohs((u_short)uh->uh_ulen); |
374 | if (ip->ip_len != len) { |
375 | if (len > ip->ip_len || len < sizeof(struct udphdr)) { |
376 | udpstat.udps_badlen++; |
377 | IF_UDP_STATINC(ifp, badlength); |
378 | goto bad; |
379 | } |
380 | m_adj(m, len - ip->ip_len); |
381 | /* ip->ip_len = len; */ |
382 | } |
383 | /* |
384 | * Save a copy of the IP header in case we want restore it |
385 | * for sending an ICMP error message in response. |
386 | */ |
387 | save_ip = *ip; |
388 | |
389 | /* |
390 | * Checksum extended UDP header and data. |
391 | */ |
392 | if (udp_input_checksum(m, uh, iphlen, len)) { |
393 | goto bad; |
394 | } |
395 | |
396 | isbroadcast = in_broadcast(ip->ip_dst, ifp); |
397 | |
398 | if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || isbroadcast) { |
399 | int reuse_sock = 0, mcast_delivered = 0; |
400 | |
401 | lck_rw_lock_shared(lck: &pcbinfo->ipi_lock); |
402 | /* |
403 | * Deliver a multicast or broadcast datagram to *all* sockets |
404 | * for which the local and remote addresses and ports match |
405 | * those of the incoming datagram. This allows more than |
406 | * one process to receive multi/broadcasts on the same port. |
407 | * (This really ought to be done for unicast datagrams as |
408 | * well, but that would cause problems with existing |
409 | * applications that open both address-specific sockets and |
410 | * a wildcard socket listening to the same port -- they would |
411 | * end up receiving duplicates of every unicast datagram. |
412 | * Those applications open the multiple sockets to overcome an |
413 | * inadequacy of the UDP socket interface, but for backwards |
414 | * compatibility we avoid the problem here rather than |
415 | * fixing the interface. Maybe 4.5BSD will remedy this?) |
416 | */ |
417 | |
418 | /* |
419 | * Construct sockaddr format source address. |
420 | */ |
421 | udp_in.sin_port = uh->uh_sport; |
422 | udp_in.sin_addr = ip->ip_src; |
423 | /* |
424 | * Locate pcb(s) for datagram. |
425 | * (Algorithm copied from raw_intr().) |
426 | */ |
427 | udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0; |
428 | LIST_FOREACH(inp, &udb, inp_list) { |
429 | #if IPSEC |
430 | int skipit; |
431 | #endif /* IPSEC */ |
432 | |
433 | if (inp->inp_socket == NULL) { |
434 | continue; |
435 | } |
436 | if (inp != sotoinpcb(inp->inp_socket)) { |
437 | panic("%s: bad so back ptr inp=%p" , |
438 | __func__, inp); |
439 | /* NOTREACHED */ |
440 | } |
441 | if ((inp->inp_vflag & INP_IPV4) == 0) { |
442 | continue; |
443 | } |
444 | if (inp_restricted_recv(inp, ifp)) { |
445 | continue; |
446 | } |
447 | |
448 | if ((inp->inp_moptions == NULL) && |
449 | (ntohl(ip->ip_dst.s_addr) != |
450 | INADDR_ALLHOSTS_GROUP) && (isbroadcast == 0)) { |
451 | continue; |
452 | } |
453 | /* |
454 | * Skip unbound sockets before taking the lock on the socket as |
455 | * the test with the destination port in the header will fail |
456 | */ |
457 | if (inp->inp_lport == 0) { |
458 | continue; |
459 | } |
460 | |
461 | if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) == |
462 | WNT_STOPUSING) { |
463 | continue; |
464 | } |
465 | |
466 | udp_lock(inp->inp_socket, 1, 0); |
467 | |
468 | if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == |
469 | WNT_STOPUSING) { |
470 | udp_unlock(inp->inp_socket, 1, 0); |
471 | continue; |
472 | } |
473 | |
474 | if (inp->inp_lport != uh->uh_dport) { |
475 | udp_unlock(inp->inp_socket, 1, 0); |
476 | continue; |
477 | } |
478 | if (inp->inp_laddr.s_addr != INADDR_ANY) { |
479 | if (inp->inp_laddr.s_addr != |
480 | ip->ip_dst.s_addr) { |
481 | udp_unlock(inp->inp_socket, 1, 0); |
482 | continue; |
483 | } |
484 | } |
485 | if (inp->inp_faddr.s_addr != INADDR_ANY) { |
486 | if (inp->inp_faddr.s_addr != |
487 | ip->ip_src.s_addr || |
488 | inp->inp_fport != uh->uh_sport) { |
489 | udp_unlock(inp->inp_socket, 1, 0); |
490 | continue; |
491 | } |
492 | } |
493 | |
494 | if (isbroadcast == 0 && (ntohl(ip->ip_dst.s_addr) != |
495 | INADDR_ALLHOSTS_GROUP)) { |
496 | struct sockaddr_in group; |
497 | int blocked; |
498 | |
499 | if ((imo = inp->inp_moptions) == NULL) { |
500 | udp_unlock(inp->inp_socket, 1, 0); |
501 | continue; |
502 | } |
503 | IMO_LOCK(imo); |
504 | |
505 | SOCKADDR_ZERO(&group, sizeof(struct sockaddr_in)); |
506 | group.sin_len = sizeof(struct sockaddr_in); |
507 | group.sin_family = AF_INET; |
508 | group.sin_addr = ip->ip_dst; |
509 | |
510 | blocked = imo_multi_filter(imo, ifp, |
511 | &group, &udp_in); |
512 | if (blocked == MCAST_PASS) { |
513 | foundmembership = 1; |
514 | } |
515 | |
516 | IMO_UNLOCK(imo); |
517 | if (!foundmembership) { |
518 | udp_unlock(inp->inp_socket, 1, 0); |
519 | if (blocked == MCAST_NOTSMEMBER || |
520 | blocked == MCAST_MUTED) { |
521 | udpstat.udps_filtermcast++; |
522 | } |
523 | continue; |
524 | } |
525 | foundmembership = 0; |
526 | } |
527 | |
528 | reuse_sock = (inp->inp_socket->so_options & |
529 | (SO_REUSEPORT | SO_REUSEADDR)); |
530 | |
531 | #if NECP |
532 | skipit = 0; |
533 | if (!necp_socket_is_allowed_to_send_recv_v4(inp, |
534 | local_port: uh->uh_dport, remote_port: uh->uh_sport, local_addr: &ip->ip_dst, |
535 | remote_addr: &ip->ip_src, input_interface: ifp, pf_tag, NULL, NULL, NULL, NULL)) { |
536 | /* do not inject data to pcb */ |
537 | skipit = 1; |
538 | } |
539 | if (skipit == 0) |
540 | #endif /* NECP */ |
541 | { |
542 | struct mbuf *n = NULL; |
543 | |
544 | if (reuse_sock) { |
545 | n = m_copy(m, 0, M_COPYALL); |
546 | } |
547 | udp_append(inp, ip, m, |
548 | iphlen + sizeof(struct udphdr), |
549 | &udp_in, &udp_in6, &udp_ip6, ifp); |
550 | mcast_delivered++; |
551 | |
552 | m = n; |
553 | } |
554 | if (is_wake_pkt) { |
555 | soevent(so: inp->inp_socket, SO_FILT_HINT_LOCKED | SO_FILT_HINT_WAKE_PKT); |
556 | } |
557 | |
558 | udp_unlock(inp->inp_socket, 1, 0); |
559 | |
560 | |
561 | /* |
562 | * Don't look for additional matches if this one does |
563 | * not have either the SO_REUSEPORT or SO_REUSEADDR |
564 | * socket options set. This heuristic avoids searching |
565 | * through all pcbs in the common case of a non-shared |
566 | * port. It assumes that an application will never |
567 | * clear these options after setting them. |
568 | */ |
569 | if (reuse_sock == 0 || m == NULL) { |
570 | break; |
571 | } |
572 | |
573 | /* |
574 | * Expect 32-bit aligned data pointer on strict-align |
575 | * platforms. |
576 | */ |
577 | MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m); |
578 | /* |
579 | * Recompute IP and UDP header pointers for new mbuf |
580 | */ |
581 | ip = mtod(m, struct ip *); |
582 | uh = (struct udphdr *)(void *)((caddr_t)ip + iphlen); |
583 | } |
584 | lck_rw_done(lck: &pcbinfo->ipi_lock); |
585 | |
586 | if (mcast_delivered == 0) { |
587 | /* |
588 | * No matching pcb found; discard datagram. |
589 | * (No need to send an ICMP Port Unreachable |
590 | * for a broadcast or multicast datgram.) |
591 | */ |
592 | udpstat.udps_noportbcast++; |
593 | IF_UDP_STATINC(ifp, port_unreach); |
594 | goto bad; |
595 | } |
596 | |
597 | /* free the extra copy of mbuf or skipped by IPsec */ |
598 | if (m != NULL) { |
599 | m_freem(m); |
600 | } |
601 | KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0); |
602 | return; |
603 | } |
604 | |
605 | #if IPSEC |
606 | /* |
607 | * UDP to port 4500 with a payload where the first four bytes are |
608 | * not zero is a UDP encapsulated IPsec packet. Packets where |
609 | * the payload is one byte and that byte is 0xFF are NAT keepalive |
610 | * packets. Decapsulate the ESP packet and carry on with IPsec input |
611 | * or discard the NAT keep-alive. |
612 | */ |
613 | if (ipsec_bypass == 0 && (esp_udp_encap_port & 0xFFFF) != 0 && |
614 | (uh->uh_dport == ntohs((u_short)esp_udp_encap_port) || |
615 | uh->uh_sport == ntohs((u_short)esp_udp_encap_port))) { |
616 | /* |
617 | * Check if ESP or keepalive: |
618 | * 1. If the destination port of the incoming packet is 4500. |
619 | * 2. If the source port of the incoming packet is 4500, |
620 | * then check the SADB to match IP address and port. |
621 | */ |
622 | bool check_esp = true; |
623 | if (uh->uh_dport != ntohs((u_short)esp_udp_encap_port)) { |
624 | check_esp = key_checksa_present(AF_INET, src: (caddr_t)&ip->ip_dst, |
625 | dst: (caddr_t)&ip->ip_src, src_port: uh->uh_dport, |
626 | dst_port: uh->uh_sport, IFSCOPE_NONE, IFSCOPE_NONE); |
627 | } |
628 | |
629 | if (check_esp) { |
630 | int payload_len = len - sizeof(struct udphdr) > 4 ? 4 : |
631 | len - sizeof(struct udphdr); |
632 | |
633 | if (m->m_len < iphlen + sizeof(struct udphdr) + payload_len) { |
634 | if ((m = m_pullup(m, iphlen + sizeof(struct udphdr) + |
635 | payload_len)) == NULL) { |
636 | udpstat.udps_hdrops++; |
637 | KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, |
638 | 0, 0, 0, 0, 0); |
639 | return; |
640 | } |
641 | /* |
642 | * Expect 32-bit aligned data pointer on strict-align |
643 | * platforms. |
644 | */ |
645 | MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m); |
646 | |
647 | ip = mtod(m, struct ip *); |
648 | uh = (struct udphdr *)(void *)((caddr_t)ip + iphlen); |
649 | } |
650 | /* Check for NAT keepalive packet */ |
651 | if (payload_len == 1 && *(u_int8_t *) |
652 | ((caddr_t)uh + sizeof(struct udphdr)) == 0xFF) { |
653 | m_freem(m); |
654 | KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, |
655 | 0, 0, 0, 0, 0); |
656 | return; |
657 | } else if (payload_len == 4 && *(u_int32_t *)(void *) |
658 | ((caddr_t)uh + sizeof(struct udphdr)) != 0) { |
659 | /* UDP encapsulated IPsec packet to pass through NAT */ |
660 | KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, |
661 | 0, 0, 0, 0, 0); |
662 | /* preserve the udp header */ |
663 | esp4_input(m, off: iphlen + sizeof(struct udphdr)); |
664 | return; |
665 | } |
666 | } |
667 | } |
668 | #endif /* IPSEC */ |
669 | |
670 | /* |
671 | * Locate pcb for datagram. |
672 | */ |
673 | inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport, |
674 | ip->ip_dst, uh->uh_dport, 1, ifp); |
675 | if (inp == NULL) { |
676 | IF_UDP_STATINC(ifp, port_unreach); |
677 | |
678 | if (udp_log_in_vain) { |
679 | char buf[MAX_IPv4_STR_LEN]; |
680 | char buf2[MAX_IPv4_STR_LEN]; |
681 | |
682 | /* check src and dst address */ |
683 | if (udp_log_in_vain < 3) { |
684 | log(LOG_INFO, "Connection attempt to " |
685 | "UDP %s:%d from %s:%d\n" , inet_ntop(AF_INET, |
686 | &ip->ip_dst, buf, sizeof(buf)), |
687 | ntohs(uh->uh_dport), inet_ntop(AF_INET, |
688 | &ip->ip_src, buf2, sizeof(buf2)), |
689 | ntohs(uh->uh_sport)); |
690 | } else if (!(m->m_flags & (M_BCAST | M_MCAST)) && |
691 | ip->ip_dst.s_addr != ip->ip_src.s_addr) { |
692 | log_in_vain_log((LOG_INFO, |
693 | "Stealth Mode connection attempt to " |
694 | "UDP %s:%d from %s:%d\n" , inet_ntop(AF_INET, |
695 | &ip->ip_dst, buf, sizeof(buf)), |
696 | ntohs(uh->uh_dport), inet_ntop(AF_INET, |
697 | &ip->ip_src, buf2, sizeof(buf2)), |
698 | ntohs(uh->uh_sport))) |
699 | } |
700 | } |
701 | udpstat.udps_noport++; |
702 | if (m->m_flags & (M_BCAST | M_MCAST)) { |
703 | udpstat.udps_noportbcast++; |
704 | goto bad; |
705 | } |
706 | if (blackhole) { |
707 | if (ifp && ifp->if_type != IFT_LOOP) { |
708 | goto bad; |
709 | } |
710 | } |
711 | *ip = save_ip; |
712 | ip->ip_len += iphlen; |
713 | icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0); |
714 | KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0); |
715 | return; |
716 | } |
717 | udp_lock(inp->inp_socket, 1, 0); |
718 | |
719 | if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) { |
720 | udp_unlock(inp->inp_socket, 1, 0); |
721 | IF_UDP_STATINC(ifp, cleanup); |
722 | goto bad; |
723 | } |
724 | #if NECP |
725 | if (!necp_socket_is_allowed_to_send_recv_v4(inp, local_port: uh->uh_dport, |
726 | remote_port: uh->uh_sport, local_addr: &ip->ip_dst, remote_addr: &ip->ip_src, input_interface: ifp, pf_tag, NULL, NULL, NULL, NULL)) { |
727 | udp_unlock(inp->inp_socket, 1, 0); |
728 | IF_UDP_STATINC(ifp, badipsec); |
729 | goto bad; |
730 | } |
731 | #endif /* NECP */ |
732 | |
733 | /* |
734 | * Construct sockaddr format source address. |
735 | * Stuff source address and datagram in user buffer. |
736 | */ |
737 | udp_in.sin_port = uh->uh_sport; |
738 | udp_in.sin_addr = ip->ip_src; |
739 | if ((inp->inp_flags & INP_CONTROLOPTS) != 0 || |
740 | SOFLOW_ENABLED(inp->inp_socket) || |
741 | SO_RECV_CONTROL_OPTS(inp->inp_socket)) { |
742 | if (inp->inp_vflag & INP_IPV6 || inp->inp_vflag & INP_V4MAPPEDV6) { |
743 | int savedflags; |
744 | |
745 | ip_2_ip6_hdr(ip6: &udp_ip6.uip6_ip6, ip); |
746 | savedflags = inp->inp_flags; |
747 | inp->inp_flags &= ~INP_UNMAPPABLEOPTS; |
748 | ret = ip6_savecontrol(inp, m, &opts); |
749 | inp->inp_flags = savedflags; |
750 | } else { |
751 | ret = ip_savecontrol(inp, &opts, ip, m); |
752 | } |
753 | if (ret != 0) { |
754 | udp_unlock(inp->inp_socket, 1, 0); |
755 | goto bad; |
756 | } |
757 | } |
758 | m_adj(m, iphlen + sizeof(struct udphdr)); |
759 | |
760 | KERNEL_DEBUG(DBG_LAYER_IN_END, uh->uh_dport, uh->uh_sport, |
761 | save_ip.ip_src.s_addr, save_ip.ip_dst.s_addr, uh->uh_ulen); |
762 | |
763 | if (inp->inp_vflag & INP_IPV6) { |
764 | in6_sin_2_v4mapsin6(sin: &udp_in, sin6: &udp_in6.uin6_sin); |
765 | append_sa = SA(&udp_in6.uin6_sin); |
766 | } else { |
767 | append_sa = SA(&udp_in); |
768 | } |
769 | if (nstat_collect) { |
770 | INP_ADD_STAT(inp, cell, wifi, wired, rxpackets, 1); |
771 | INP_ADD_STAT(inp, cell, wifi, wired, rxbytes, m->m_pkthdr.len); |
772 | inp_set_activity_bitmap(inp); |
773 | } |
774 | #if CONTENT_FILTER && NECP |
775 | if (check_cfil && inp != NULL && inp->inp_policyresult.results.filter_control_unit == 0) { |
776 | if (inp->inp_vflag & INP_IPV6) { |
777 | bzero(s: &udp_dst6, n: sizeof(udp_dst6)); |
778 | udp_dst6.uin6_sin.sin6_len = sizeof(struct sockaddr_in6); |
779 | udp_dst6.uin6_sin.sin6_family = AF_INET6; |
780 | in6_sin_2_v4mapsin6(sin: &udp_dst, sin6: &udp_dst6.uin6_sin); |
781 | append_da = SA(&udp_dst6.uin6_sin); |
782 | } else { |
783 | SOCKADDR_ZERO(&udp_dst, sizeof(udp_dst)); |
784 | udp_dst.sin_len = sizeof(struct sockaddr_in); |
785 | udp_dst.sin_family = AF_INET; |
786 | udp_dst.sin_port = uh->uh_dport; |
787 | udp_dst.sin_addr = ip->ip_dst; |
788 | append_da = SA(&udp_dst); |
789 | } |
790 | // Override the dst input here so NECP can pick up the policy |
791 | // and CFIL can find an existing control socket. |
792 | necp_socket_find_policy_match(inp, override_local_addr: append_da, override_remote_addr: append_sa, override_bound_interface: 0); |
793 | } |
794 | #endif /* CONTENT_FILTER and NECP */ |
795 | so_recv_data_stat(inp->inp_socket, m, 0); |
796 | if (sbappendaddr(sb: &inp->inp_socket->so_rcv, asa: append_sa, |
797 | m0: m, control: opts, NULL) == 0) { |
798 | udpstat.udps_fullsock++; |
799 | } else { |
800 | sorwakeup(so: inp->inp_socket); |
801 | } |
802 | if (is_wake_pkt) { |
803 | soevent(so: inp->inp_socket, SO_FILT_HINT_LOCKED | SO_FILT_HINT_WAKE_PKT); |
804 | } |
805 | udp_unlock(inp->inp_socket, 1, 0); |
806 | KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0); |
807 | return; |
808 | bad: |
809 | m_freem(m); |
810 | if (opts) { |
811 | m_freem(opts); |
812 | } |
813 | KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0); |
814 | } |
815 | |
816 | static void |
817 | ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip) |
818 | { |
819 | bzero(s: ip6, n: sizeof(*ip6)); |
820 | |
821 | ip6->ip6_vfc = IPV6_VERSION; |
822 | ip6->ip6_plen = ip->ip_len; |
823 | ip6->ip6_nxt = ip->ip_p; |
824 | ip6->ip6_hlim = ip->ip_ttl; |
825 | if (ip->ip_src.s_addr) { |
826 | ip6->ip6_src.s6_addr32[2] = IPV6_ADDR_INT32_SMP; |
827 | ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr; |
828 | } |
829 | if (ip->ip_dst.s_addr) { |
830 | ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_SMP; |
831 | ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr; |
832 | } |
833 | } |
834 | |
835 | /* |
836 | * subroutine of udp_input(), mainly for source code readability. |
837 | */ |
838 | static void |
839 | udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off, |
840 | struct sockaddr_in *pudp_in, struct udp_in6 *pudp_in6, |
841 | struct udp_ip6 *pudp_ip6, struct ifnet *ifp) |
842 | { |
843 | struct sockaddr *append_sa; |
844 | struct mbuf *opts = 0; |
845 | boolean_t cell = IFNET_IS_CELLULAR(ifp); |
846 | boolean_t wifi = (!cell && IFNET_IS_WIFI(ifp)); |
847 | boolean_t wired = (!wifi && IFNET_IS_WIRED(ifp)); |
848 | int ret = 0; |
849 | |
850 | if ((last->inp_flags & INP_CONTROLOPTS) != 0 || |
851 | SOFLOW_ENABLED(last->inp_socket) || |
852 | SO_RECV_CONTROL_OPTS(last->inp_socket)) { |
853 | if (last->inp_vflag & INP_IPV6 || last->inp_vflag & INP_V4MAPPEDV6) { |
854 | int savedflags; |
855 | |
856 | if (pudp_ip6->uip6_init_done == 0) { |
857 | ip_2_ip6_hdr(ip6: &pudp_ip6->uip6_ip6, ip); |
858 | pudp_ip6->uip6_init_done = 1; |
859 | } |
860 | savedflags = last->inp_flags; |
861 | last->inp_flags &= ~INP_UNMAPPABLEOPTS; |
862 | ret = ip6_savecontrol(last, n, &opts); |
863 | if (ret != 0) { |
864 | last->inp_flags = savedflags; |
865 | goto error; |
866 | } |
867 | last->inp_flags = savedflags; |
868 | } else { |
869 | ret = ip_savecontrol(last, &opts, ip, n); |
870 | if (ret != 0) { |
871 | goto error; |
872 | } |
873 | } |
874 | } |
875 | if (last->inp_vflag & INP_IPV6) { |
876 | if (pudp_in6->uin6_init_done == 0) { |
877 | in6_sin_2_v4mapsin6(sin: pudp_in, sin6: &pudp_in6->uin6_sin); |
878 | pudp_in6->uin6_init_done = 1; |
879 | } |
880 | append_sa = SA(&pudp_in6->uin6_sin); |
881 | } else { |
882 | append_sa = SA(pudp_in); |
883 | } |
884 | if (nstat_collect) { |
885 | INP_ADD_STAT(last, cell, wifi, wired, rxpackets, 1); |
886 | INP_ADD_STAT(last, cell, wifi, wired, rxbytes, |
887 | n->m_pkthdr.len); |
888 | inp_set_activity_bitmap(inp: last); |
889 | } |
890 | so_recv_data_stat(last->inp_socket, n, 0); |
891 | m_adj(n, off); |
892 | if (sbappendaddr(sb: &last->inp_socket->so_rcv, asa: append_sa, |
893 | m0: n, control: opts, NULL) == 0) { |
894 | udpstat.udps_fullsock++; |
895 | } else { |
896 | sorwakeup(so: last->inp_socket); |
897 | } |
898 | return; |
899 | error: |
900 | m_freem(n); |
901 | m_freem(opts); |
902 | } |
903 | |
904 | /* |
905 | * Notify a udp user of an asynchronous error; |
906 | * just wake up so that he can collect error status. |
907 | */ |
908 | void |
909 | udp_notify(struct inpcb *inp, int errno) |
910 | { |
911 | inp->inp_socket->so_error = (u_short)errno; |
912 | sorwakeup(so: inp->inp_socket); |
913 | sowwakeup(so: inp->inp_socket); |
914 | } |
915 | |
916 | void |
917 | udp_ctlinput(int cmd, struct sockaddr *sa, void *vip, __unused struct ifnet * ifp) |
918 | { |
919 | struct ipctlparam *ctl_param = vip; |
920 | struct ip *ip = NULL; |
921 | struct mbuf *m = NULL; |
922 | void (*notify)(struct inpcb *, int) = udp_notify; |
923 | struct in_addr faddr; |
924 | struct inpcb *inp = NULL; |
925 | struct icmp *icp = NULL; |
926 | size_t off; |
927 | |
928 | if (ctl_param != NULL) { |
929 | ip = ctl_param->ipc_icmp_ip; |
930 | icp = ctl_param->ipc_icmp; |
931 | m = ctl_param->ipc_m; |
932 | off = ctl_param->ipc_off; |
933 | } else { |
934 | ip = NULL; |
935 | icp = NULL; |
936 | m = NULL; |
937 | off = 0; |
938 | } |
939 | |
940 | faddr = SIN(sa)->sin_addr; |
941 | if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) { |
942 | return; |
943 | } |
944 | |
945 | if (PRC_IS_REDIRECT(cmd)) { |
946 | ip = 0; |
947 | notify = in_rtchange; |
948 | } else if (cmd == PRC_HOSTDEAD) { |
949 | ip = 0; |
950 | } else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) { |
951 | return; |
952 | } |
953 | if (ip) { |
954 | struct udphdr uh; |
955 | |
956 | /* Check if we can safely get the ports from the UDP header */ |
957 | if (m == NULL || |
958 | (m->m_len < off + sizeof(uh))) { |
959 | /* Insufficient length */ |
960 | return; |
961 | } |
962 | |
963 | bcopy(src: m_mtod_current(m) + off, dst: &uh, n: sizeof(uh)); |
964 | inp = in_pcblookup_hash(&udbinfo, faddr, uh.uh_dport, |
965 | ip->ip_src, uh.uh_sport, 0, NULL); |
966 | |
967 | if (inp != NULL && inp->inp_socket != NULL) { |
968 | udp_lock(inp->inp_socket, 1, 0); |
969 | if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == |
970 | WNT_STOPUSING) { |
971 | udp_unlock(inp->inp_socket, 1, 0); |
972 | return; |
973 | } |
974 | if (cmd == PRC_MSGSIZE && !uuid_is_null(uu: inp->necp_client_uuid)) { |
975 | uuid_t null_uuid; |
976 | uuid_clear(uu: null_uuid); |
977 | necp_update_flow_protoctl_event(netagent_uuid: null_uuid, client_id: inp->necp_client_uuid, |
978 | PRC_MSGSIZE, ntohs(icp->icmp_nextmtu), protoctl_event_tcp_seq_num: 0); |
979 | /* |
980 | * Avoid calling udp_notify() to set so_error |
981 | * when using Network.framework since the notification |
982 | * of PRC_MSGSIZE has been delivered through NECP. |
983 | */ |
984 | } else { |
985 | (*notify)(inp, inetctlerrmap[cmd]); |
986 | } |
987 | udp_unlock(inp->inp_socket, 1, 0); |
988 | } |
989 | #if SKYWALK |
990 | else { |
991 | union sockaddr_in_4_6 sock_laddr; |
992 | struct protoctl_ev_val prctl_ev_val; |
993 | bzero(s: &prctl_ev_val, n: sizeof(prctl_ev_val)); |
994 | bzero(s: &sock_laddr, n: sizeof(sock_laddr)); |
995 | |
996 | if (cmd == PRC_MSGSIZE) { |
997 | prctl_ev_val.val = ntohs(icp->icmp_nextmtu); |
998 | } |
999 | |
1000 | sock_laddr.sin.sin_family = AF_INET; |
1001 | sock_laddr.sin.sin_len = sizeof(sock_laddr.sin); |
1002 | sock_laddr.sin.sin_addr = ip->ip_src; |
1003 | |
1004 | protoctl_event_enqueue_nwk_wq_entry(ifp, |
1005 | SA(&sock_laddr), p_raddr: sa, |
1006 | lport: uh.uh_sport, rport: uh.uh_dport, IPPROTO_UDP, |
1007 | protoctl_event_code: cmd, p_protoctl_ev_val: &prctl_ev_val); |
1008 | } |
1009 | #endif /* SKYWALK */ |
1010 | } else { |
1011 | in_pcbnotifyall(&udbinfo, faddr, inetctlerrmap[cmd], notify); |
1012 | } |
1013 | } |
1014 | |
1015 | int |
1016 | udp_ctloutput(struct socket *so, struct sockopt *sopt) |
1017 | { |
1018 | int error = 0, optval = 0; |
1019 | struct inpcb *inp; |
1020 | |
1021 | /* Allow <SOL_SOCKET,SO_FLUSH> at this level */ |
1022 | if (sopt->sopt_level != IPPROTO_UDP && |
1023 | !(sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_FLUSH)) { |
1024 | if (SOCK_CHECK_DOM(so, PF_INET6)) { |
1025 | error = ip6_ctloutput(so, sopt); |
1026 | } else { |
1027 | error = ip_ctloutput(so, sopt); |
1028 | } |
1029 | return error; |
1030 | } |
1031 | |
1032 | inp = sotoinpcb(so); |
1033 | |
1034 | switch (sopt->sopt_dir) { |
1035 | case SOPT_SET: |
1036 | switch (sopt->sopt_name) { |
1037 | case UDP_NOCKSUM: |
1038 | /* This option is settable only for UDP over IPv4 */ |
1039 | if (!(inp->inp_vflag & INP_IPV4)) { |
1040 | error = EINVAL; |
1041 | break; |
1042 | } |
1043 | |
1044 | if ((error = sooptcopyin(sopt, &optval, len: sizeof(optval), |
1045 | minlen: sizeof(optval))) != 0) { |
1046 | break; |
1047 | } |
1048 | |
1049 | if (optval != 0) { |
1050 | inp->inp_flags |= INP_UDP_NOCKSUM; |
1051 | } else { |
1052 | inp->inp_flags &= ~INP_UDP_NOCKSUM; |
1053 | } |
1054 | break; |
1055 | case UDP_KEEPALIVE_OFFLOAD: |
1056 | { |
1057 | struct udp_keepalive_offload ka; |
1058 | /* |
1059 | * If the socket is not connected, the stack will |
1060 | * not know the destination address to put in the |
1061 | * keepalive datagram. Return an error now instead |
1062 | * of failing later. |
1063 | */ |
1064 | if (!(so->so_state & SS_ISCONNECTED)) { |
1065 | error = EINVAL; |
1066 | break; |
1067 | } |
1068 | if (sopt->sopt_valsize != sizeof(ka)) { |
1069 | error = EINVAL; |
1070 | break; |
1071 | } |
1072 | if ((error = sooptcopyin(sopt, &ka, len: sizeof(ka), |
1073 | minlen: sizeof(ka))) != 0) { |
1074 | break; |
1075 | } |
1076 | |
1077 | /* application should specify the type */ |
1078 | if (ka.ka_type == 0) { |
1079 | return EINVAL; |
1080 | } |
1081 | |
1082 | if (ka.ka_interval == 0) { |
1083 | /* |
1084 | * if interval is 0, disable the offload |
1085 | * mechanism |
1086 | */ |
1087 | if (inp->inp_keepalive_data != NULL) { |
1088 | kfree_data(inp->inp_keepalive_data, |
1089 | inp->inp_keepalive_datalen); |
1090 | } |
1091 | inp->inp_keepalive_data = NULL; |
1092 | inp->inp_keepalive_datalen = 0; |
1093 | inp->inp_keepalive_interval = 0; |
1094 | inp->inp_keepalive_type = 0; |
1095 | inp->inp_flags2 &= ~INP2_KEEPALIVE_OFFLOAD; |
1096 | } else { |
1097 | if (inp->inp_keepalive_data != NULL) { |
1098 | kfree_data(inp->inp_keepalive_data, |
1099 | inp->inp_keepalive_datalen); |
1100 | inp->inp_keepalive_data = NULL; |
1101 | } |
1102 | |
1103 | inp->inp_keepalive_datalen = (uint8_t)min( |
1104 | a: ka.ka_data_len, |
1105 | UDP_KEEPALIVE_OFFLOAD_DATA_SIZE); |
1106 | if (inp->inp_keepalive_datalen > 0) { |
1107 | inp->inp_keepalive_data = (u_int8_t *)kalloc_data( |
1108 | inp->inp_keepalive_datalen, Z_WAITOK); |
1109 | if (inp->inp_keepalive_data == NULL) { |
1110 | inp->inp_keepalive_datalen = 0; |
1111 | error = ENOMEM; |
1112 | break; |
1113 | } |
1114 | bcopy(src: ka.ka_data, |
1115 | dst: inp->inp_keepalive_data, |
1116 | n: inp->inp_keepalive_datalen); |
1117 | } else { |
1118 | inp->inp_keepalive_datalen = 0; |
1119 | } |
1120 | inp->inp_keepalive_interval = (uint8_t) |
1121 | min(UDP_KEEPALIVE_INTERVAL_MAX_SECONDS, |
1122 | b: ka.ka_interval); |
1123 | inp->inp_keepalive_type = ka.ka_type; |
1124 | inp->inp_flags2 |= INP2_KEEPALIVE_OFFLOAD; |
1125 | } |
1126 | break; |
1127 | } |
1128 | case SO_FLUSH: |
1129 | if ((error = sooptcopyin(sopt, &optval, len: sizeof(optval), |
1130 | minlen: sizeof(optval))) != 0) { |
1131 | break; |
1132 | } |
1133 | |
1134 | error = inp_flush(inp, optval); |
1135 | break; |
1136 | |
1137 | default: |
1138 | error = ENOPROTOOPT; |
1139 | break; |
1140 | } |
1141 | break; |
1142 | |
1143 | case SOPT_GET: |
1144 | switch (sopt->sopt_name) { |
1145 | case UDP_NOCKSUM: |
1146 | optval = inp->inp_flags & INP_UDP_NOCKSUM; |
1147 | break; |
1148 | |
1149 | default: |
1150 | error = ENOPROTOOPT; |
1151 | break; |
1152 | } |
1153 | if (error == 0) { |
1154 | error = sooptcopyout(sopt, data: &optval, len: sizeof(optval)); |
1155 | } |
1156 | break; |
1157 | } |
1158 | return error; |
1159 | } |
1160 | |
1161 | static int |
1162 | udp_pcblist SYSCTL_HANDLER_ARGS |
1163 | { |
1164 | #pragma unused(oidp, arg1, arg2) |
1165 | int error, i, n, sz; |
1166 | struct inpcb *inp, **inp_list; |
1167 | inp_gen_t gencnt; |
1168 | struct xinpgen xig; |
1169 | |
1170 | /* |
1171 | * The process of preparing the TCB list is too time-consuming and |
1172 | * resource-intensive to repeat twice on every request. |
1173 | */ |
1174 | lck_rw_lock_exclusive(lck: &udbinfo.ipi_lock); |
1175 | if (req->oldptr == USER_ADDR_NULL) { |
1176 | n = udbinfo.ipi_count; |
1177 | req->oldidx = 2 * (sizeof(xig)) |
1178 | + (n + n / 8) * sizeof(struct xinpcb); |
1179 | lck_rw_done(lck: &udbinfo.ipi_lock); |
1180 | return 0; |
1181 | } |
1182 | |
1183 | if (req->newptr != USER_ADDR_NULL) { |
1184 | lck_rw_done(lck: &udbinfo.ipi_lock); |
1185 | return EPERM; |
1186 | } |
1187 | |
1188 | /* |
1189 | * OK, now we're committed to doing something. |
1190 | */ |
1191 | gencnt = udbinfo.ipi_gencnt; |
1192 | sz = n = udbinfo.ipi_count; |
1193 | |
1194 | bzero(s: &xig, n: sizeof(xig)); |
1195 | xig.xig_len = sizeof(xig); |
1196 | xig.xig_count = n; |
1197 | xig.xig_gen = gencnt; |
1198 | xig.xig_sogen = so_gencnt; |
1199 | error = SYSCTL_OUT(req, &xig, sizeof(xig)); |
1200 | if (error) { |
1201 | lck_rw_done(lck: &udbinfo.ipi_lock); |
1202 | return error; |
1203 | } |
1204 | /* |
1205 | * We are done if there is no pcb |
1206 | */ |
1207 | if (n == 0) { |
1208 | lck_rw_done(lck: &udbinfo.ipi_lock); |
1209 | return 0; |
1210 | } |
1211 | |
1212 | inp_list = kalloc_type(struct inpcb *, n, Z_WAITOK); |
1213 | if (inp_list == NULL) { |
1214 | lck_rw_done(lck: &udbinfo.ipi_lock); |
1215 | return ENOMEM; |
1216 | } |
1217 | |
1218 | for (inp = LIST_FIRST(udbinfo.ipi_listhead), i = 0; inp && i < n; |
1219 | inp = LIST_NEXT(inp, inp_list)) { |
1220 | if (inp->inp_gencnt <= gencnt && |
1221 | inp->inp_state != INPCB_STATE_DEAD) { |
1222 | inp_list[i++] = inp; |
1223 | } |
1224 | } |
1225 | n = i; |
1226 | |
1227 | error = 0; |
1228 | for (i = 0; i < n; i++) { |
1229 | struct xinpcb xi; |
1230 | |
1231 | inp = inp_list[i]; |
1232 | |
1233 | if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) == WNT_STOPUSING) { |
1234 | continue; |
1235 | } |
1236 | udp_lock(inp->inp_socket, 1, 0); |
1237 | if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) { |
1238 | udp_unlock(inp->inp_socket, 1, 0); |
1239 | continue; |
1240 | } |
1241 | if (inp->inp_gencnt > gencnt) { |
1242 | udp_unlock(inp->inp_socket, 1, 0); |
1243 | continue; |
1244 | } |
1245 | |
1246 | bzero(s: &xi, n: sizeof(xi)); |
1247 | xi.xi_len = sizeof(xi); |
1248 | /* XXX should avoid extra copy */ |
1249 | inpcb_to_compat(inp, &xi.xi_inp); |
1250 | if (inp->inp_socket) { |
1251 | sotoxsocket(so: inp->inp_socket, xso: &xi.xi_socket); |
1252 | } |
1253 | |
1254 | udp_unlock(inp->inp_socket, 1, 0); |
1255 | |
1256 | error = SYSCTL_OUT(req, &xi, sizeof(xi)); |
1257 | } |
1258 | if (!error) { |
1259 | /* |
1260 | * Give the user an updated idea of our state. |
1261 | * If the generation differs from what we told |
1262 | * her before, she knows that something happened |
1263 | * while we were processing this request, and it |
1264 | * might be necessary to retry. |
1265 | */ |
1266 | bzero(s: &xig, n: sizeof(xig)); |
1267 | xig.xig_len = sizeof(xig); |
1268 | xig.xig_gen = udbinfo.ipi_gencnt; |
1269 | xig.xig_sogen = so_gencnt; |
1270 | xig.xig_count = udbinfo.ipi_count; |
1271 | error = SYSCTL_OUT(req, &xig, sizeof(xig)); |
1272 | } |
1273 | |
1274 | lck_rw_done(lck: &udbinfo.ipi_lock); |
1275 | kfree_type(struct inpcb *, sz, inp_list); |
1276 | return error; |
1277 | } |
1278 | |
1279 | SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, |
1280 | CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0, udp_pcblist, |
1281 | "S,xinpcb" , "List of active UDP sockets" ); |
1282 | |
1283 | #if XNU_TARGET_OS_OSX |
1284 | |
1285 | static int |
1286 | udp_pcblist64 SYSCTL_HANDLER_ARGS |
1287 | { |
1288 | #pragma unused(oidp, arg1, arg2) |
1289 | int error, i, n, sz; |
1290 | struct inpcb *inp, **inp_list; |
1291 | inp_gen_t gencnt; |
1292 | struct xinpgen xig; |
1293 | |
1294 | /* |
1295 | * The process of preparing the TCB list is too time-consuming and |
1296 | * resource-intensive to repeat twice on every request. |
1297 | */ |
1298 | lck_rw_lock_shared(lck: &udbinfo.ipi_lock); |
1299 | if (req->oldptr == USER_ADDR_NULL) { |
1300 | n = udbinfo.ipi_count; |
1301 | req->oldidx = |
1302 | 2 * (sizeof(xig)) + (n + n / 8) * sizeof(struct xinpcb64); |
1303 | lck_rw_done(lck: &udbinfo.ipi_lock); |
1304 | return 0; |
1305 | } |
1306 | |
1307 | if (req->newptr != USER_ADDR_NULL) { |
1308 | lck_rw_done(lck: &udbinfo.ipi_lock); |
1309 | return EPERM; |
1310 | } |
1311 | |
1312 | /* |
1313 | * OK, now we're committed to doing something. |
1314 | */ |
1315 | gencnt = udbinfo.ipi_gencnt; |
1316 | sz = n = udbinfo.ipi_count; |
1317 | |
1318 | bzero(s: &xig, n: sizeof(xig)); |
1319 | xig.xig_len = sizeof(xig); |
1320 | xig.xig_count = n; |
1321 | xig.xig_gen = gencnt; |
1322 | xig.xig_sogen = so_gencnt; |
1323 | error = SYSCTL_OUT(req, &xig, sizeof(xig)); |
1324 | if (error) { |
1325 | lck_rw_done(lck: &udbinfo.ipi_lock); |
1326 | return error; |
1327 | } |
1328 | /* |
1329 | * We are done if there is no pcb |
1330 | */ |
1331 | if (n == 0) { |
1332 | lck_rw_done(lck: &udbinfo.ipi_lock); |
1333 | return 0; |
1334 | } |
1335 | |
1336 | inp_list = kalloc_type(struct inpcb *, n, Z_WAITOK); |
1337 | if (inp_list == NULL) { |
1338 | lck_rw_done(lck: &udbinfo.ipi_lock); |
1339 | return ENOMEM; |
1340 | } |
1341 | |
1342 | for (inp = LIST_FIRST(udbinfo.ipi_listhead), i = 0; inp && i < n; |
1343 | inp = LIST_NEXT(inp, inp_list)) { |
1344 | if (inp->inp_gencnt <= gencnt && |
1345 | inp->inp_state != INPCB_STATE_DEAD) { |
1346 | inp_list[i++] = inp; |
1347 | } |
1348 | } |
1349 | n = i; |
1350 | |
1351 | error = 0; |
1352 | for (i = 0; i < n; i++) { |
1353 | struct xinpcb64 xi; |
1354 | |
1355 | inp = inp_list[i]; |
1356 | |
1357 | if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) == WNT_STOPUSING) { |
1358 | continue; |
1359 | } |
1360 | udp_lock(inp->inp_socket, 1, 0); |
1361 | if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) { |
1362 | udp_unlock(inp->inp_socket, 1, 0); |
1363 | continue; |
1364 | } |
1365 | if (inp->inp_gencnt > gencnt) { |
1366 | udp_unlock(inp->inp_socket, 1, 0); |
1367 | continue; |
1368 | } |
1369 | |
1370 | bzero(s: &xi, n: sizeof(xi)); |
1371 | xi.xi_len = sizeof(xi); |
1372 | inpcb_to_xinpcb64(inp, &xi); |
1373 | if (inp->inp_socket) { |
1374 | sotoxsocket64(so: inp->inp_socket, xso: &xi.xi_socket); |
1375 | } |
1376 | |
1377 | udp_unlock(inp->inp_socket, 1, 0); |
1378 | |
1379 | error = SYSCTL_OUT(req, &xi, sizeof(xi)); |
1380 | } |
1381 | if (!error) { |
1382 | /* |
1383 | * Give the user an updated idea of our state. |
1384 | * If the generation differs from what we told |
1385 | * her before, she knows that something happened |
1386 | * while we were processing this request, and it |
1387 | * might be necessary to retry. |
1388 | */ |
1389 | bzero(s: &xig, n: sizeof(xig)); |
1390 | xig.xig_len = sizeof(xig); |
1391 | xig.xig_gen = udbinfo.ipi_gencnt; |
1392 | xig.xig_sogen = so_gencnt; |
1393 | xig.xig_count = udbinfo.ipi_count; |
1394 | error = SYSCTL_OUT(req, &xig, sizeof(xig)); |
1395 | } |
1396 | |
1397 | lck_rw_done(lck: &udbinfo.ipi_lock); |
1398 | kfree_type(struct inpcb *, sz, inp_list); |
1399 | return error; |
1400 | } |
1401 | |
1402 | SYSCTL_PROC(_net_inet_udp, OID_AUTO, pcblist64, |
1403 | CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0, udp_pcblist64, |
1404 | "S,xinpcb64" , "List of active UDP sockets" ); |
1405 | |
1406 | #endif /* XNU_TARGET_OS_OSX */ |
1407 | |
1408 | static int |
1409 | udp_pcblist_n SYSCTL_HANDLER_ARGS |
1410 | { |
1411 | #pragma unused(oidp, arg1, arg2) |
1412 | return get_pcblist_n(IPPROTO_UDP, req, &udbinfo); |
1413 | } |
1414 | |
1415 | SYSCTL_PROC(_net_inet_udp, OID_AUTO, pcblist_n, |
1416 | CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0, udp_pcblist_n, |
1417 | "S,xinpcb_n" , "List of active UDP sockets" ); |
1418 | |
1419 | __private_extern__ void |
1420 | udp_get_ports_used(ifnet_t ifp, int protocol, uint32_t flags, |
1421 | bitstr_t *bitfield) |
1422 | { |
1423 | inpcb_get_ports_used(ifp, protocol, flags, bitfield, |
1424 | &udbinfo); |
1425 | } |
1426 | |
1427 | __private_extern__ uint32_t |
1428 | udp_count_opportunistic(unsigned int ifindex, u_int32_t flags) |
1429 | { |
1430 | return inpcb_count_opportunistic(ifindex, &udbinfo, flags); |
1431 | } |
1432 | |
1433 | __private_extern__ uint32_t |
1434 | udp_find_anypcb_byaddr(struct ifaddr *ifa) |
1435 | { |
1436 | #if SKYWALK |
1437 | if (netns_is_enabled()) { |
1438 | return netns_find_anyres_byaddr(ifa, IPPROTO_UDP); |
1439 | } else |
1440 | #endif /* SKYWALK */ |
1441 | return inpcb_find_anypcb_byaddr(ifa, &udbinfo); |
1442 | } |
1443 | |
1444 | static int |
1445 | udp_check_pktinfo(struct mbuf *control, struct ifnet **outif, |
1446 | struct in_addr *laddr) |
1447 | { |
1448 | struct cmsghdr *cm = 0; |
1449 | struct in_pktinfo *pktinfo; |
1450 | struct ifnet *ifp; |
1451 | |
1452 | if (outif != NULL) { |
1453 | *outif = NULL; |
1454 | } |
1455 | |
1456 | /* |
1457 | * XXX: Currently, we assume all the optional information is stored |
1458 | * in a single mbuf. |
1459 | */ |
1460 | if (control->m_next) { |
1461 | return EINVAL; |
1462 | } |
1463 | |
1464 | if (control->m_len < CMSG_LEN(0)) { |
1465 | return EINVAL; |
1466 | } |
1467 | |
1468 | for (cm = M_FIRST_CMSGHDR(control); |
1469 | is_cmsg_valid(control, cmsg: cm); |
1470 | cm = M_NXT_CMSGHDR(control, cm)) { |
1471 | if (cm->cmsg_level != IPPROTO_IP || |
1472 | cm->cmsg_type != IP_PKTINFO) { |
1473 | continue; |
1474 | } |
1475 | |
1476 | if (cm->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo))) { |
1477 | return EINVAL; |
1478 | } |
1479 | |
1480 | pktinfo = (struct in_pktinfo *)(void *)CMSG_DATA(cm); |
1481 | |
1482 | /* Check for a valid ifindex in pktinfo */ |
1483 | ifnet_head_lock_shared(); |
1484 | |
1485 | if (pktinfo->ipi_ifindex > if_index) { |
1486 | ifnet_head_done(); |
1487 | return ENXIO; |
1488 | } |
1489 | |
1490 | /* |
1491 | * If ipi_ifindex is specified it takes precedence |
1492 | * over ipi_spec_dst. |
1493 | */ |
1494 | if (pktinfo->ipi_ifindex) { |
1495 | ifp = ifindex2ifnet[pktinfo->ipi_ifindex]; |
1496 | if (ifp == NULL) { |
1497 | ifnet_head_done(); |
1498 | return ENXIO; |
1499 | } |
1500 | if (outif != NULL) { |
1501 | ifnet_reference(interface: ifp); |
1502 | *outif = ifp; |
1503 | } |
1504 | ifnet_head_done(); |
1505 | laddr->s_addr = INADDR_ANY; |
1506 | break; |
1507 | } |
1508 | |
1509 | ifnet_head_done(); |
1510 | |
1511 | /* |
1512 | * Use the provided ipi_spec_dst address for temp |
1513 | * source address. |
1514 | */ |
1515 | *laddr = pktinfo->ipi_spec_dst; |
1516 | break; |
1517 | } |
1518 | return 0; |
1519 | } |
1520 | |
1521 | int |
1522 | udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr, |
1523 | struct mbuf *control, struct proc *p) |
1524 | { |
1525 | struct udpiphdr *ui; |
1526 | int len = m->m_pkthdr.len; |
1527 | struct sockaddr_in *sin; |
1528 | struct in_addr origladdr, laddr, faddr, pi_laddr; |
1529 | u_short lport, fport; |
1530 | int error = 0, udp_dodisconnect = 0, pktinfo = 0; |
1531 | struct socket *so = inp->inp_socket; |
1532 | int soopts = 0; |
1533 | struct mbuf *inpopts; |
1534 | struct ip_moptions *mopts; |
1535 | struct route ro; |
1536 | struct ip_out_args ipoa; |
1537 | bool sndinprog_cnt_used = false; |
1538 | #if CONTENT_FILTER |
1539 | struct m_tag *cfil_tag = NULL; |
1540 | bool cfil_faddr_use = false; |
1541 | uint32_t cfil_so_state_change_cnt = 0; |
1542 | uint32_t cfil_so_options = 0; |
1543 | struct sockaddr *cfil_faddr = NULL; |
1544 | #endif |
1545 | bool check_qos_marking_again = (so->so_flags1 & SOF1_QOSMARKING_POLICY_OVERRIDE) ? FALSE : TRUE; |
1546 | |
1547 | bzero(s: &ipoa, n: sizeof(ipoa)); |
1548 | ipoa.ipoa_boundif = IFSCOPE_NONE; |
1549 | ipoa.ipoa_flags = IPOAF_SELECT_SRCIF; |
1550 | |
1551 | struct ifnet *outif = NULL; |
1552 | struct flowadv *adv = &ipoa.ipoa_flowadv; |
1553 | int sotc = SO_TC_UNSPEC; |
1554 | int netsvctype = _NET_SERVICE_TYPE_UNSPEC; |
1555 | struct ifnet *origoutifp = NULL; |
1556 | int flowadv = 0; |
1557 | int tos = IPTOS_UNSPEC; |
1558 | |
1559 | /* Enable flow advisory only when connected */ |
1560 | flowadv = (so->so_state & SS_ISCONNECTED) ? 1 : 0; |
1561 | pi_laddr.s_addr = INADDR_ANY; |
1562 | |
1563 | KERNEL_DEBUG(DBG_FNC_UDP_OUTPUT | DBG_FUNC_START, 0, 0, 0, 0, 0); |
1564 | |
1565 | socket_lock_assert_owned(so); |
1566 | |
1567 | #if CONTENT_FILTER |
1568 | /* |
1569 | * If socket is subject to UDP Content Filter and no addr is passed in, |
1570 | * retrieve CFIL saved state from mbuf and use it if necessary. |
1571 | */ |
1572 | if (CFIL_DGRAM_FILTERED(so) && !addr) { |
1573 | cfil_tag = cfil_dgram_get_socket_state(m, state_change_cnt: &cfil_so_state_change_cnt, options: &cfil_so_options, faddr: &cfil_faddr, NULL); |
1574 | if (cfil_tag) { |
1575 | sin = SIN(cfil_faddr); |
1576 | if (inp && inp->inp_faddr.s_addr == INADDR_ANY) { |
1577 | /* |
1578 | * Socket is unconnected, simply use the saved faddr as 'addr' to go through |
1579 | * the connect/disconnect logic. |
1580 | */ |
1581 | addr = SA(cfil_faddr); |
1582 | } else if ((so->so_state_change_cnt != cfil_so_state_change_cnt) && |
1583 | (inp->inp_fport != sin->sin_port || |
1584 | inp->inp_faddr.s_addr != sin->sin_addr.s_addr)) { |
1585 | /* |
1586 | * Socket is connected but socket state and dest addr/port changed. |
1587 | * We need to use the saved faddr info. |
1588 | */ |
1589 | cfil_faddr_use = true; |
1590 | } |
1591 | } |
1592 | } |
1593 | #endif |
1594 | |
1595 | if (control != NULL) { |
1596 | tos = so_tos_from_control(control); |
1597 | sotc = so_tc_from_control(control, &netsvctype); |
1598 | VERIFY(outif == NULL); |
1599 | error = udp_check_pktinfo(control, outif: &outif, laddr: &pi_laddr); |
1600 | m_freem(control); |
1601 | control = NULL; |
1602 | if (error) { |
1603 | goto release; |
1604 | } |
1605 | if (outif != NULL) { |
1606 | pktinfo++; |
1607 | ipoa.ipoa_boundif = outif->if_index; |
1608 | } |
1609 | } |
1610 | if (sotc == SO_TC_UNSPEC) { |
1611 | sotc = so->so_traffic_class; |
1612 | netsvctype = so->so_netsvctype; |
1613 | } |
1614 | |
1615 | KERNEL_DEBUG(DBG_LAYER_OUT_BEG, inp->inp_fport, inp->inp_lport, |
1616 | inp->inp_laddr.s_addr, inp->inp_faddr.s_addr, |
1617 | (htons((u_short)len + sizeof(struct udphdr)))); |
1618 | |
1619 | if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) { |
1620 | error = EMSGSIZE; |
1621 | goto release; |
1622 | } |
1623 | |
1624 | if (flowadv && INP_WAIT_FOR_IF_FEEDBACK(inp)) { |
1625 | /* |
1626 | * The socket is flow-controlled, drop the packets |
1627 | * until the inp is not flow controlled |
1628 | */ |
1629 | error = ENOBUFS; |
1630 | goto release; |
1631 | } |
1632 | /* |
1633 | * If socket was bound to an ifindex, tell ip_output about it. |
1634 | * If the ancillary IP_PKTINFO option contains an interface index, |
1635 | * it takes precedence over the one specified by IP_BOUND_IF. |
1636 | */ |
1637 | if (ipoa.ipoa_boundif == IFSCOPE_NONE && |
1638 | (inp->inp_flags & INP_BOUND_IF)) { |
1639 | VERIFY(inp->inp_boundifp != NULL); |
1640 | ifnet_reference(interface: inp->inp_boundifp); /* for this routine */ |
1641 | if (outif != NULL) { |
1642 | ifnet_release(interface: outif); |
1643 | } |
1644 | outif = inp->inp_boundifp; |
1645 | ipoa.ipoa_boundif = outif->if_index; |
1646 | } |
1647 | if (INP_NO_CELLULAR(inp)) { |
1648 | ipoa.ipoa_flags |= IPOAF_NO_CELLULAR; |
1649 | } |
1650 | if (INP_NO_EXPENSIVE(inp)) { |
1651 | ipoa.ipoa_flags |= IPOAF_NO_EXPENSIVE; |
1652 | } |
1653 | if (INP_NO_CONSTRAINED(inp)) { |
1654 | ipoa.ipoa_flags |= IPOAF_NO_CONSTRAINED; |
1655 | } |
1656 | if (INP_AWDL_UNRESTRICTED(inp)) { |
1657 | ipoa.ipoa_flags |= IPOAF_AWDL_UNRESTRICTED; |
1658 | } |
1659 | if (INP_MANAGEMENT_ALLOWED(inp)) { |
1660 | ipoa.ipoa_flags |= IPOAF_MANAGEMENT_ALLOWED; |
1661 | } |
1662 | ipoa.ipoa_sotc = sotc; |
1663 | ipoa.ipoa_netsvctype = netsvctype; |
1664 | soopts |= IP_OUTARGS; |
1665 | |
1666 | /* |
1667 | * If there was a routing change, discard cached route and check |
1668 | * that we have a valid source address. Reacquire a new source |
1669 | * address if INADDR_ANY was specified. |
1670 | * |
1671 | * If we are using cfil saved state, go through this cache cleanup |
1672 | * so that we can get a new route. |
1673 | */ |
1674 | if (ROUTE_UNUSABLE(&inp->inp_route) |
1675 | #if CONTENT_FILTER |
1676 | || cfil_faddr_use |
1677 | #endif |
1678 | ) { |
1679 | struct in_ifaddr *ia = NULL; |
1680 | |
1681 | ROUTE_RELEASE(&inp->inp_route); |
1682 | |
1683 | /* src address is gone? */ |
1684 | if (inp->inp_laddr.s_addr != INADDR_ANY && |
1685 | (ia = ifa_foraddr(inp->inp_laddr.s_addr)) == NULL) { |
1686 | if (!(inp->inp_flags & INP_INADDR_ANY) || |
1687 | (so->so_state & SS_ISCONNECTED)) { |
1688 | /* |
1689 | * Rdar://5448998 |
1690 | * If the source address is gone, return an |
1691 | * error if: |
1692 | * - the source was specified |
1693 | * - the socket was already connected |
1694 | */ |
1695 | soevent(so, hint: (SO_FILT_HINT_LOCKED | |
1696 | SO_FILT_HINT_NOSRCADDR)); |
1697 | error = EADDRNOTAVAIL; |
1698 | goto release; |
1699 | } else { |
1700 | /* new src will be set later */ |
1701 | inp->inp_laddr.s_addr = INADDR_ANY; |
1702 | inp->inp_last_outifp = NULL; |
1703 | #if SKYWALK |
1704 | if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) { |
1705 | netns_set_ifnet(token: &inp->inp_netns_token, NULL); |
1706 | } |
1707 | #endif /* SKYWALK */ |
1708 | } |
1709 | } |
1710 | if (ia != NULL) { |
1711 | ifa_remref(ifa: &ia->ia_ifa); |
1712 | } |
1713 | } |
1714 | |
1715 | /* |
1716 | * IP_PKTINFO option check. If a temporary scope or src address |
1717 | * is provided, use it for this packet only and make sure we forget |
1718 | * it after sending this datagram. |
1719 | */ |
1720 | if (pi_laddr.s_addr != INADDR_ANY || |
1721 | (ipoa.ipoa_boundif != IFSCOPE_NONE && pktinfo)) { |
1722 | /* temp src address for this datagram only */ |
1723 | laddr = pi_laddr; |
1724 | origladdr.s_addr = INADDR_ANY; |
1725 | /* we don't want to keep the laddr or route */ |
1726 | udp_dodisconnect = 1; |
1727 | /* remember we don't care about src addr */ |
1728 | inp->inp_flags |= INP_INADDR_ANY; |
1729 | } else { |
1730 | origladdr = laddr = inp->inp_laddr; |
1731 | } |
1732 | |
1733 | origoutifp = inp->inp_last_outifp; |
1734 | faddr = inp->inp_faddr; |
1735 | lport = inp->inp_lport; |
1736 | fport = inp->inp_fport; |
1737 | |
1738 | #if CONTENT_FILTER |
1739 | if (cfil_faddr_use) { |
1740 | faddr = SIN(cfil_faddr)->sin_addr; |
1741 | fport = SIN(cfil_faddr)->sin_port; |
1742 | } |
1743 | #endif |
1744 | inp->inp_sndinprog_cnt++; |
1745 | sndinprog_cnt_used = true; |
1746 | |
1747 | if (addr) { |
1748 | sin = SIN(addr); |
1749 | if (faddr.s_addr != INADDR_ANY) { |
1750 | error = EISCONN; |
1751 | goto release; |
1752 | } |
1753 | if (lport == 0) { |
1754 | /* |
1755 | * In case we don't have a local port set, go through |
1756 | * the full connect. We don't have a local port yet |
1757 | * (i.e., we can't be looked up), so it's not an issue |
1758 | * if the input runs at the same time we do this. |
1759 | */ |
1760 | /* if we have a source address specified, use that */ |
1761 | if (pi_laddr.s_addr != INADDR_ANY) { |
1762 | inp->inp_laddr = pi_laddr; |
1763 | } |
1764 | /* |
1765 | * If a scope is specified, use it. Scope from |
1766 | * IP_PKTINFO takes precendence over the the scope |
1767 | * set via INP_BOUND_IF. |
1768 | */ |
1769 | error = in_pcbconnect(inp, addr, p, ipoa.ipoa_boundif, |
1770 | &outif); |
1771 | if (error) { |
1772 | goto release; |
1773 | } |
1774 | |
1775 | laddr = inp->inp_laddr; |
1776 | lport = inp->inp_lport; |
1777 | faddr = inp->inp_faddr; |
1778 | fport = inp->inp_fport; |
1779 | udp_dodisconnect = 1; |
1780 | |
1781 | /* synch up in case in_pcbladdr() overrides */ |
1782 | if (outif != NULL && ipoa.ipoa_boundif != IFSCOPE_NONE) { |
1783 | ipoa.ipoa_boundif = outif->if_index; |
1784 | } |
1785 | } else { |
1786 | /* |
1787 | * Fast path case |
1788 | * |
1789 | * We have a full address and a local port; use those |
1790 | * info to build the packet without changing the pcb |
1791 | * and interfering with the input path. See 3851370. |
1792 | * |
1793 | * Scope from IP_PKTINFO takes precendence over the |
1794 | * the scope set via INP_BOUND_IF. |
1795 | */ |
1796 | if (laddr.s_addr == INADDR_ANY) { |
1797 | if ((error = in_pcbladdr(inp, addr, &laddr, |
1798 | ipoa.ipoa_boundif, &outif, 0)) != 0) { |
1799 | goto release; |
1800 | } |
1801 | /* |
1802 | * from pcbconnect: remember we don't |
1803 | * care about src addr. |
1804 | */ |
1805 | inp->inp_flags |= INP_INADDR_ANY; |
1806 | |
1807 | /* synch up in case in_pcbladdr() overrides */ |
1808 | if (outif != NULL && |
1809 | ipoa.ipoa_boundif != IFSCOPE_NONE) { |
1810 | ipoa.ipoa_boundif = outif->if_index; |
1811 | } |
1812 | } |
1813 | |
1814 | faddr = sin->sin_addr; |
1815 | fport = sin->sin_port; |
1816 | } |
1817 | } else { |
1818 | if (faddr.s_addr == INADDR_ANY) { |
1819 | error = ENOTCONN; |
1820 | goto release; |
1821 | } |
1822 | } |
1823 | |
1824 | if (inp->inp_flowhash == 0) { |
1825 | inp_calc_flowhash(inp); |
1826 | ASSERT(inp->inp_flowhash != 0); |
1827 | } |
1828 | |
1829 | if (fport == htons(53) && !(so->so_flags1 & SOF1_DNS_COUNTED)) { |
1830 | so->so_flags1 |= SOF1_DNS_COUNTED; |
1831 | INC_ATOMIC_INT64_LIM(net_api_stats.nas_socket_inet_dgram_dns); |
1832 | } |
1833 | |
1834 | /* |
1835 | * Calculate data length and get a mbuf |
1836 | * for UDP and IP headers. |
1837 | */ |
1838 | M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT, 1); |
1839 | if (m == 0) { |
1840 | error = ENOBUFS; |
1841 | goto abort; |
1842 | } |
1843 | |
1844 | /* |
1845 | * Fill in mbuf with extended UDP header |
1846 | * and addresses and length put into network format. |
1847 | */ |
1848 | ui = mtod(m, struct udpiphdr *); |
1849 | bzero(s: ui->ui_x1, n: sizeof(ui->ui_x1)); /* XXX still needed? */ |
1850 | ui->ui_pr = IPPROTO_UDP; |
1851 | ui->ui_src = laddr; |
1852 | ui->ui_dst = faddr; |
1853 | ui->ui_sport = lport; |
1854 | ui->ui_dport = fport; |
1855 | ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr)); |
1856 | |
1857 | /* |
1858 | * Set the Don't Fragment bit in the IP header. |
1859 | */ |
1860 | if (inp->inp_flags2 & INP2_DONTFRAG) { |
1861 | struct ip *ip; |
1862 | |
1863 | ip = (struct ip *)&ui->ui_i; |
1864 | ip->ip_off |= IP_DF; |
1865 | } |
1866 | |
1867 | /* |
1868 | * Set up checksum to pseudo header checksum and output datagram. |
1869 | * |
1870 | * Treat flows to be CLAT46'd as IPv6 flow and compute checksum |
1871 | * no matter what, as IPv6 mandates checksum for UDP. |
1872 | * |
1873 | * Here we only compute the one's complement sum of the pseudo header. |
1874 | * The payload computation and final complement is delayed to much later |
1875 | * in IP processing to decide if remaining computation needs to be done |
1876 | * through offload. |
1877 | * |
1878 | * That is communicated by setting CSUM_UDP in csum_flags. |
1879 | * The offset of checksum from the start of ULP header is communicated |
1880 | * through csum_data. |
1881 | * |
1882 | * Note since this already contains the pseudo checksum header, any |
1883 | * later operation at IP layer that modify the values used here must |
1884 | * update the checksum as well (for example NAT etc). |
1885 | */ |
1886 | if ((inp->inp_flags2 & INP2_CLAT46_FLOW) || |
1887 | (udpcksum && !(inp->inp_flags & INP_UDP_NOCKSUM))) { |
1888 | ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr, |
1889 | htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP)); |
1890 | m->m_pkthdr.csum_flags = (CSUM_UDP | CSUM_ZERO_INVERT); |
1891 | m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); |
1892 | } else { |
1893 | ui->ui_sum = 0; |
1894 | } |
1895 | ((struct ip *)ui)->ip_len = (uint16_t)(sizeof(struct udpiphdr) + len); |
1896 | ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */ |
1897 | if (tos != IPTOS_UNSPEC) { |
1898 | ((struct ip *)ui)->ip_tos = (uint8_t)(tos & IPTOS_MASK); |
1899 | } else { |
1900 | ((struct ip *)ui)->ip_tos = inp->inp_ip_tos; /* XXX */ |
1901 | } |
1902 | udpstat.udps_opackets++; |
1903 | |
1904 | KERNEL_DEBUG(DBG_LAYER_OUT_END, ui->ui_dport, ui->ui_sport, |
1905 | ui->ui_src.s_addr, ui->ui_dst.s_addr, ui->ui_ulen); |
1906 | |
1907 | #if NECP |
1908 | { |
1909 | necp_kernel_policy_id policy_id; |
1910 | necp_kernel_policy_id skip_policy_id; |
1911 | u_int32_t route_rule_id; |
1912 | u_int32_t pass_flags; |
1913 | |
1914 | /* |
1915 | * We need a route to perform NECP route rule checks |
1916 | */ |
1917 | if (net_qos_policy_restricted != 0 && |
1918 | ROUTE_UNUSABLE(&inp->inp_route)) { |
1919 | struct sockaddr_in to; |
1920 | struct sockaddr_in from; |
1921 | |
1922 | ROUTE_RELEASE(&inp->inp_route); |
1923 | |
1924 | SOCKADDR_ZERO(&from, sizeof(struct sockaddr_in)); |
1925 | from.sin_family = AF_INET; |
1926 | from.sin_len = sizeof(struct sockaddr_in); |
1927 | from.sin_addr = laddr; |
1928 | |
1929 | SOCKADDR_ZERO(&to, sizeof(struct sockaddr_in)); |
1930 | to.sin_family = AF_INET; |
1931 | to.sin_len = sizeof(struct sockaddr_in); |
1932 | to.sin_addr = faddr; |
1933 | |
1934 | inp->inp_route.ro_dst.sa_family = AF_INET; |
1935 | inp->inp_route.ro_dst.sa_len = sizeof(struct sockaddr_in); |
1936 | SIN(&inp->inp_route.ro_dst)->sin_addr = faddr; |
1937 | |
1938 | rtalloc_scoped(&inp->inp_route, ipoa.ipoa_boundif); |
1939 | |
1940 | inp_update_necp_policy(inp, SA(&from), |
1941 | SA(&to), ipoa.ipoa_boundif); |
1942 | inp->inp_policyresult.results.qos_marking_gencount = 0; |
1943 | } |
1944 | |
1945 | if (!necp_socket_is_allowed_to_send_recv_v4(inp, local_port: lport, remote_port: fport, |
1946 | local_addr: &laddr, remote_addr: &faddr, NULL, pf_tag: 0, return_policy_id: &policy_id, return_route_rule_id: &route_rule_id, return_skip_policy_id: &skip_policy_id, return_pass_flags: &pass_flags)) { |
1947 | error = EHOSTUNREACH; |
1948 | goto abort; |
1949 | } |
1950 | |
1951 | necp_mark_packet_from_socket(packet: m, inp, policy_id, route_rule_id, skip_policy_id, pass_flags); |
1952 | |
1953 | if (net_qos_policy_restricted != 0) { |
1954 | necp_socket_update_qos_marking(inp, route: inp->inp_route.ro_rt, route_rule_id); |
1955 | } |
1956 | } |
1957 | #endif /* NECP */ |
1958 | if ((so->so_flags1 & SOF1_QOSMARKING_ALLOWED)) { |
1959 | ipoa.ipoa_flags |= IPOAF_QOSMARKING_ALLOWED; |
1960 | } |
1961 | if (check_qos_marking_again) { |
1962 | ipoa.ipoa_flags |= IPOAF_REDO_QOSMARKING_POLICY; |
1963 | } |
1964 | ipoa.qos_marking_gencount = inp->inp_policyresult.results.qos_marking_gencount; |
1965 | |
1966 | #if IPSEC |
1967 | if (inp->inp_sp != NULL && ipsec_setsocket(m, inp->inp_socket) != 0) { |
1968 | error = ENOBUFS; |
1969 | goto abort; |
1970 | } |
1971 | #endif /* IPSEC */ |
1972 | |
1973 | inpopts = inp->inp_options; |
1974 | #if CONTENT_FILTER |
1975 | if (cfil_tag && (inp->inp_socket->so_options != cfil_so_options)) { |
1976 | soopts |= (cfil_so_options & (SO_DONTROUTE | SO_BROADCAST)); |
1977 | } else |
1978 | #endif |
1979 | soopts |= (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)); |
1980 | |
1981 | mopts = inp->inp_moptions; |
1982 | if (mopts != NULL) { |
1983 | IMO_LOCK(mopts); |
1984 | IMO_ADDREF_LOCKED(mopts); |
1985 | if (IN_MULTICAST(ntohl(ui->ui_dst.s_addr)) && |
1986 | mopts->imo_multicast_ifp != NULL) { |
1987 | /* no reference needed */ |
1988 | inp->inp_last_outifp = mopts->imo_multicast_ifp; |
1989 | #if SKYWALK |
1990 | if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) { |
1991 | netns_set_ifnet(token: &inp->inp_netns_token, |
1992 | ifp: inp->inp_last_outifp); |
1993 | } |
1994 | #endif /* SKYWALK */ |
1995 | } |
1996 | IMO_UNLOCK(mopts); |
1997 | } |
1998 | |
1999 | /* Copy the cached route and take an extra reference */ |
2000 | inp_route_copyout(inp, &ro); |
2001 | |
2002 | set_packet_service_class(m, so, sotc, 0); |
2003 | m->m_pkthdr.pkt_flowsrc = FLOWSRC_INPCB; |
2004 | m->m_pkthdr.pkt_flowid = inp->inp_flowhash; |
2005 | m->m_pkthdr.pkt_proto = IPPROTO_UDP; |
2006 | m->m_pkthdr.pkt_flags |= (PKTF_FLOW_ID | PKTF_FLOW_LOCALSRC); |
2007 | if (flowadv) { |
2008 | m->m_pkthdr.pkt_flags |= PKTF_FLOW_ADV; |
2009 | } |
2010 | m->m_pkthdr.tx_udp_pid = so->last_pid; |
2011 | if (so->so_flags & SOF_DELEGATED) { |
2012 | m->m_pkthdr.tx_udp_e_pid = so->e_pid; |
2013 | } else { |
2014 | m->m_pkthdr.tx_udp_e_pid = 0; |
2015 | } |
2016 | #if (DEBUG || DEVELOPMENT) |
2017 | if (so->so_flags & SOF_MARK_WAKE_PKT) { |
2018 | so->so_flags &= ~SOF_MARK_WAKE_PKT; |
2019 | m->m_pkthdr.pkt_flags |= PKTF_WAKE_PKT; |
2020 | } |
2021 | #endif /* (DEBUG || DEVELOPMENT) */ |
2022 | |
2023 | m_add_crumb(m, PKT_CRUMB_UDP_OUTPUT); |
2024 | |
2025 | if (ipoa.ipoa_boundif != IFSCOPE_NONE) { |
2026 | ipoa.ipoa_flags |= IPOAF_BOUND_IF; |
2027 | } |
2028 | |
2029 | if (laddr.s_addr != INADDR_ANY) { |
2030 | ipoa.ipoa_flags |= IPOAF_BOUND_SRCADDR; |
2031 | } |
2032 | |
2033 | socket_unlock(so, refcount: 0); |
2034 | error = ip_output(m, inpopts, &ro, soopts, mopts, &ipoa); |
2035 | m = NULL; |
2036 | socket_lock(so, refcount: 0); |
2037 | if (mopts != NULL) { |
2038 | IMO_REMREF(mopts); |
2039 | } |
2040 | |
2041 | if (check_qos_marking_again) { |
2042 | inp->inp_policyresult.results.qos_marking_gencount = ipoa.qos_marking_gencount; |
2043 | |
2044 | if (ipoa.ipoa_flags & IPOAF_QOSMARKING_ALLOWED) { |
2045 | inp->inp_socket->so_flags1 |= SOF1_QOSMARKING_ALLOWED; |
2046 | } else { |
2047 | inp->inp_socket->so_flags1 &= ~SOF1_QOSMARKING_ALLOWED; |
2048 | } |
2049 | } |
2050 | |
2051 | if (error == 0 && nstat_collect) { |
2052 | boolean_t cell, wifi, wired; |
2053 | |
2054 | if (ro.ro_rt != NULL) { |
2055 | cell = IFNET_IS_CELLULAR(ro.ro_rt->rt_ifp); |
2056 | wifi = (!cell && IFNET_IS_WIFI(ro.ro_rt->rt_ifp)); |
2057 | wired = (!wifi && IFNET_IS_WIRED(ro.ro_rt->rt_ifp)); |
2058 | } else { |
2059 | cell = wifi = wired = FALSE; |
2060 | } |
2061 | INP_ADD_STAT(inp, cell, wifi, wired, txpackets, 1); |
2062 | INP_ADD_STAT(inp, cell, wifi, wired, txbytes, len); |
2063 | inp_set_activity_bitmap(inp); |
2064 | } |
2065 | |
2066 | if (flowadv && (adv->code == FADV_FLOW_CONTROLLED || |
2067 | adv->code == FADV_SUSPENDED)) { |
2068 | /* |
2069 | * return a hint to the application that |
2070 | * the packet has been dropped |
2071 | */ |
2072 | error = ENOBUFS; |
2073 | inp_set_fc_state(inp, advcode: adv->code); |
2074 | } |
2075 | |
2076 | /* Synchronize PCB cached route */ |
2077 | inp_route_copyin(inp, &ro); |
2078 | |
2079 | if (inp->inp_route.ro_rt != NULL) { |
2080 | if (IS_LOCALNET_ROUTE(inp->inp_route.ro_rt)) { |
2081 | inp->inp_flags2 |= INP2_LAST_ROUTE_LOCAL; |
2082 | } else { |
2083 | inp->inp_flags2 &= ~INP2_LAST_ROUTE_LOCAL; |
2084 | } |
2085 | } |
2086 | |
2087 | abort: |
2088 | if (udp_dodisconnect) { |
2089 | /* Always discard the cached route for unconnected socket */ |
2090 | ROUTE_RELEASE(&inp->inp_route); |
2091 | in_pcbdisconnect(inp); |
2092 | inp->inp_laddr = origladdr; /* XXX rehash? */ |
2093 | /* no reference needed */ |
2094 | inp->inp_last_outifp = origoutifp; |
2095 | #if SKYWALK |
2096 | if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) { |
2097 | netns_set_ifnet(token: &inp->inp_netns_token, |
2098 | ifp: inp->inp_last_outifp); |
2099 | } |
2100 | #endif /* SKYWALK */ |
2101 | } else if (inp->inp_route.ro_rt != NULL) { |
2102 | struct rtentry *rt = inp->inp_route.ro_rt; |
2103 | struct ifnet *outifp; |
2104 | |
2105 | if (rt->rt_flags & (RTF_MULTICAST | RTF_BROADCAST)) { |
2106 | rt = NULL; /* unusable */ |
2107 | } |
2108 | #if CONTENT_FILTER |
2109 | /* |
2110 | * Discard temporary route for cfil case |
2111 | */ |
2112 | if (cfil_faddr_use) { |
2113 | rt = NULL; /* unusable */ |
2114 | } |
2115 | #endif |
2116 | |
2117 | /* |
2118 | * Always discard if it is a multicast or broadcast route. |
2119 | */ |
2120 | if (rt == NULL) { |
2121 | ROUTE_RELEASE(&inp->inp_route); |
2122 | } |
2123 | |
2124 | /* |
2125 | * If the destination route is unicast, update outifp with |
2126 | * that of the route interface used by IP. |
2127 | */ |
2128 | if (rt != NULL && |
2129 | (outifp = rt->rt_ifp) != inp->inp_last_outifp) { |
2130 | inp->inp_last_outifp = outifp; /* no reference needed */ |
2131 | #if SKYWALK |
2132 | if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) { |
2133 | netns_set_ifnet(token: &inp->inp_netns_token, |
2134 | ifp: inp->inp_last_outifp); |
2135 | } |
2136 | #endif /* SKYWALK */ |
2137 | |
2138 | so->so_pktheadroom = (uint16_t)P2ROUNDUP( |
2139 | sizeof(struct udphdr) + |
2140 | sizeof(struct ip) + |
2141 | ifnet_hdrlen(outifp) + |
2142 | ifnet_mbuf_packetpreamblelen(outifp), |
2143 | sizeof(u_int32_t)); |
2144 | } |
2145 | } else { |
2146 | ROUTE_RELEASE(&inp->inp_route); |
2147 | } |
2148 | |
2149 | /* |
2150 | * If output interface was cellular/expensive, and this socket is |
2151 | * denied access to it, generate an event. |
2152 | */ |
2153 | if (error != 0 && (ipoa.ipoa_flags & IPOAF_R_IFDENIED) && |
2154 | (INP_NO_CELLULAR(inp) || INP_NO_EXPENSIVE(inp) || INP_NO_CONSTRAINED(inp))) { |
2155 | soevent(so, hint: (SO_FILT_HINT_LOCKED | SO_FILT_HINT_IFDENIED)); |
2156 | } |
2157 | |
2158 | release: |
2159 | KERNEL_DEBUG(DBG_FNC_UDP_OUTPUT | DBG_FUNC_END, error, 0, 0, 0, 0); |
2160 | |
2161 | if (m != NULL) { |
2162 | m_freem(m); |
2163 | } |
2164 | |
2165 | if (outif != NULL) { |
2166 | ifnet_release(interface: outif); |
2167 | } |
2168 | |
2169 | #if CONTENT_FILTER |
2170 | if (cfil_tag) { |
2171 | m_tag_free(cfil_tag); |
2172 | } |
2173 | #endif |
2174 | if (sndinprog_cnt_used) { |
2175 | VERIFY(inp->inp_sndinprog_cnt > 0); |
2176 | if (--inp->inp_sndinprog_cnt == 0) { |
2177 | inp->inp_flags &= ~(INP_FC_FEEDBACK); |
2178 | if (inp->inp_sndingprog_waiters > 0) { |
2179 | wakeup(chan: &inp->inp_sndinprog_cnt); |
2180 | } |
2181 | } |
2182 | sndinprog_cnt_used = false; |
2183 | } |
2184 | |
2185 | return error; |
2186 | } |
2187 | |
2188 | u_int32_t udp_sendspace = 9216; /* really max datagram size */ |
2189 | /* 187 1K datagrams (approx 192 KB) */ |
2190 | u_int32_t udp_recvspace = 187 * (1024 + sizeof(struct sockaddr_in6)); |
2191 | |
2192 | /* Check that the values of udp send and recv space do not exceed sb_max */ |
2193 | static int |
2194 | sysctl_udp_sospace(struct sysctl_oid *oidp, void *arg1, int arg2, |
2195 | struct sysctl_req *req) |
2196 | { |
2197 | #pragma unused(arg1, arg2) |
2198 | u_int32_t new_value = 0, *space_p = NULL; |
2199 | int changed = 0, error = 0; |
2200 | |
2201 | switch (oidp->oid_number) { |
2202 | case UDPCTL_RECVSPACE: |
2203 | space_p = &udp_recvspace; |
2204 | break; |
2205 | case UDPCTL_MAXDGRAM: |
2206 | space_p = &udp_sendspace; |
2207 | break; |
2208 | default: |
2209 | return EINVAL; |
2210 | } |
2211 | error = sysctl_io_number(req, bigValue: *space_p, valueSize: sizeof(u_int32_t), |
2212 | pValue: &new_value, changed: &changed); |
2213 | if (changed) { |
2214 | if (new_value > 0 && new_value <= sb_max) { |
2215 | *space_p = new_value; |
2216 | } else { |
2217 | error = ERANGE; |
2218 | } |
2219 | } |
2220 | return error; |
2221 | } |
2222 | |
2223 | SYSCTL_PROC(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, |
2224 | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &udp_recvspace, 0, |
2225 | &sysctl_udp_sospace, "IU" , "Maximum incoming UDP datagram size" ); |
2226 | |
2227 | SYSCTL_PROC(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, |
2228 | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &udp_sendspace, 0, |
2229 | &sysctl_udp_sospace, "IU" , "Maximum outgoing UDP datagram size" ); |
2230 | |
2231 | int |
2232 | udp_abort(struct socket *so) |
2233 | { |
2234 | struct inpcb *inp; |
2235 | |
2236 | inp = sotoinpcb(so); |
2237 | if (inp == NULL) { |
2238 | panic("%s: so=%p null inp" , __func__, so); |
2239 | /* NOTREACHED */ |
2240 | } |
2241 | soisdisconnected(so); |
2242 | in_pcbdetach(inp); |
2243 | return 0; |
2244 | } |
2245 | |
2246 | int |
2247 | udp_attach(struct socket *so, int proto, struct proc *p) |
2248 | { |
2249 | #pragma unused(proto) |
2250 | struct inpcb *inp; |
2251 | int error; |
2252 | |
2253 | error = soreserve(so, sndcc: udp_sendspace, rcvcc: udp_recvspace); |
2254 | if (error != 0) { |
2255 | return error; |
2256 | } |
2257 | inp = sotoinpcb(so); |
2258 | if (inp != NULL) { |
2259 | panic("%s so=%p inp=%p" , __func__, so, inp); |
2260 | /* NOTREACHED */ |
2261 | } |
2262 | error = in_pcballoc(so, &udbinfo, p); |
2263 | if (error != 0) { |
2264 | return error; |
2265 | } |
2266 | inp = (struct inpcb *)so->so_pcb; |
2267 | inp->inp_vflag |= INP_IPV4; |
2268 | inp->inp_ip_ttl = (uint8_t)ip_defttl; |
2269 | if (nstat_collect) { |
2270 | nstat_udp_new_pcb(inp); |
2271 | } |
2272 | return 0; |
2273 | } |
2274 | |
2275 | int |
2276 | udp_bind(struct socket *so, struct sockaddr *nam, struct proc *p) |
2277 | { |
2278 | struct inpcb *inp; |
2279 | int error; |
2280 | |
2281 | if (nam->sa_family != 0 && nam->sa_family != AF_INET && |
2282 | nam->sa_family != AF_INET6) { |
2283 | return EAFNOSUPPORT; |
2284 | } |
2285 | |
2286 | inp = sotoinpcb(so); |
2287 | if (inp == NULL) { |
2288 | return EINVAL; |
2289 | } |
2290 | error = in_pcbbind(inp, nam, p); |
2291 | |
2292 | #if NECP |
2293 | /* Update NECP client with bind result if not in middle of connect */ |
2294 | if (error == 0 && |
2295 | (inp->inp_flags2 & INP2_CONNECT_IN_PROGRESS) && |
2296 | !uuid_is_null(uu: inp->necp_client_uuid)) { |
2297 | socket_unlock(so, refcount: 0); |
2298 | necp_client_assign_from_socket(pid: so->last_pid, client_id: inp->necp_client_uuid, inp); |
2299 | socket_lock(so, refcount: 0); |
2300 | } |
2301 | #endif /* NECP */ |
2302 | |
2303 | UDP_LOG_BIND(inp, error); |
2304 | |
2305 | return error; |
2306 | } |
2307 | |
2308 | int |
2309 | udp_connect(struct socket *so, struct sockaddr *nam, struct proc *p) |
2310 | { |
2311 | struct inpcb *inp; |
2312 | int error; |
2313 | |
2314 | inp = sotoinpcb(so); |
2315 | if (inp == NULL) { |
2316 | return EINVAL; |
2317 | } |
2318 | if (inp->inp_faddr.s_addr != INADDR_ANY) { |
2319 | return EISCONN; |
2320 | } |
2321 | |
2322 | if (!(so->so_flags1 & SOF1_CONNECT_COUNTED)) { |
2323 | so->so_flags1 |= SOF1_CONNECT_COUNTED; |
2324 | INC_ATOMIC_INT64_LIM(net_api_stats.nas_socket_inet_dgram_connected); |
2325 | } |
2326 | |
2327 | #if NECP |
2328 | #if FLOW_DIVERT |
2329 | if (necp_socket_should_use_flow_divert(inp)) { |
2330 | error = flow_divert_pcb_init(so); |
2331 | if (error == 0) { |
2332 | error = flow_divert_connect_out(so, to: nam, p); |
2333 | } |
2334 | UDP_LOG_CONNECT(inp, error); |
2335 | return error; |
2336 | } else { |
2337 | so->so_flags1 |= SOF1_FLOW_DIVERT_SKIP; |
2338 | } |
2339 | #endif /* FLOW_DIVERT */ |
2340 | #endif /* NECP */ |
2341 | |
2342 | error = in_pcbconnect(inp, nam, p, IFSCOPE_NONE, NULL); |
2343 | if (error == 0) { |
2344 | #if NECP |
2345 | /* Update NECP client with connected five-tuple */ |
2346 | if (!uuid_is_null(uu: inp->necp_client_uuid)) { |
2347 | socket_unlock(so, refcount: 0); |
2348 | necp_client_assign_from_socket(pid: so->last_pid, client_id: inp->necp_client_uuid, inp); |
2349 | socket_lock(so, refcount: 0); |
2350 | } |
2351 | #endif /* NECP */ |
2352 | |
2353 | soisconnected(so); |
2354 | if (inp->inp_flowhash == 0) { |
2355 | inp_calc_flowhash(inp); |
2356 | ASSERT(inp->inp_flowhash != 0); |
2357 | } |
2358 | inp->inp_connect_timestamp = mach_continuous_time(); |
2359 | } |
2360 | UDP_LOG_CONNECT(inp, error); |
2361 | return error; |
2362 | } |
2363 | |
2364 | int |
2365 | udp_connectx_common(struct socket *so, int af, struct sockaddr *src, struct sockaddr *dst, |
2366 | struct proc *p, uint32_t ifscope, sae_associd_t aid, sae_connid_t *pcid, |
2367 | uint32_t flags, void *arg, uint32_t arglen, |
2368 | struct uio *uio, user_ssize_t *bytes_written) |
2369 | { |
2370 | #pragma unused(aid, flags, arg, arglen) |
2371 | struct inpcb *inp = sotoinpcb(so); |
2372 | int error = 0; |
2373 | user_ssize_t datalen = 0; |
2374 | |
2375 | if (inp == NULL) { |
2376 | return EINVAL; |
2377 | } |
2378 | |
2379 | VERIFY(dst != NULL); |
2380 | |
2381 | ASSERT(!(inp->inp_flags2 & INP2_CONNECT_IN_PROGRESS)); |
2382 | inp->inp_flags2 |= INP2_CONNECT_IN_PROGRESS; |
2383 | |
2384 | #if NECP |
2385 | inp_update_necp_policy(inp, src, dst, ifscope); |
2386 | #endif /* NECP */ |
2387 | |
2388 | /* bind socket to the specified interface, if requested */ |
2389 | if (ifscope != IFSCOPE_NONE && |
2390 | (error = inp_bindif(inp, ifscope, NULL)) != 0) { |
2391 | goto done; |
2392 | } |
2393 | |
2394 | /* if source address and/or port is specified, bind to it */ |
2395 | if (src != NULL) { |
2396 | error = sobindlock(so, nam: src, dolock: 0); /* already locked */ |
2397 | if (error != 0) { |
2398 | goto done; |
2399 | } |
2400 | } |
2401 | |
2402 | switch (af) { |
2403 | case AF_INET: |
2404 | error = udp_connect(so, nam: dst, p); |
2405 | break; |
2406 | case AF_INET6: |
2407 | error = udp6_connect(so, dst, p); |
2408 | break; |
2409 | default: |
2410 | VERIFY(0); |
2411 | /* NOTREACHED */ |
2412 | } |
2413 | |
2414 | if (error != 0) { |
2415 | goto done; |
2416 | } |
2417 | |
2418 | /* |
2419 | * If there is data, copy it. DATA_IDEMPOTENT is ignored. |
2420 | * CONNECT_RESUME_ON_READ_WRITE is ignored. |
2421 | */ |
2422 | if (uio != NULL) { |
2423 | socket_unlock(so, refcount: 0); |
2424 | |
2425 | VERIFY(bytes_written != NULL); |
2426 | |
2427 | datalen = uio_resid(a_uio: uio); |
2428 | error = so->so_proto->pr_usrreqs->pru_sosend(so, NULL, |
2429 | (uio_t)uio, NULL, NULL, 0); |
2430 | socket_lock(so, refcount: 0); |
2431 | |
2432 | /* If error returned is EMSGSIZE, for example, disconnect */ |
2433 | if (error == 0 || error == EWOULDBLOCK) { |
2434 | *bytes_written = datalen - uio_resid(a_uio: uio); |
2435 | } else { |
2436 | (void) so->so_proto->pr_usrreqs->pru_disconnectx(so, |
2437 | SAE_ASSOCID_ANY, SAE_CONNID_ANY); |
2438 | } |
2439 | /* |
2440 | * mask the EWOULDBLOCK error so that the caller |
2441 | * knows that atleast the connect was successful. |
2442 | */ |
2443 | if (error == EWOULDBLOCK) { |
2444 | error = 0; |
2445 | } |
2446 | } |
2447 | |
2448 | if (error == 0 && pcid != NULL) { |
2449 | *pcid = 1; /* there is only 1 connection for UDP */ |
2450 | } |
2451 | done: |
2452 | inp->inp_flags2 &= ~INP2_CONNECT_IN_PROGRESS; |
2453 | return error; |
2454 | } |
2455 | |
2456 | int |
2457 | udp_connectx(struct socket *so, struct sockaddr *src, |
2458 | struct sockaddr *dst, struct proc *p, uint32_t ifscope, |
2459 | sae_associd_t aid, sae_connid_t *pcid, uint32_t flags, void *arg, |
2460 | uint32_t arglen, struct uio *uio, user_ssize_t *bytes_written) |
2461 | { |
2462 | return udp_connectx_common(so, AF_INET, src, dst, |
2463 | p, ifscope, aid, pcid, flags, arg, arglen, uio, bytes_written); |
2464 | } |
2465 | |
2466 | int |
2467 | udp_detach(struct socket *so) |
2468 | { |
2469 | struct inpcb *inp; |
2470 | |
2471 | inp = sotoinpcb(so); |
2472 | if (inp == NULL) { |
2473 | panic("%s: so=%p null inp" , __func__, so); |
2474 | /* NOTREACHED */ |
2475 | } |
2476 | |
2477 | /* |
2478 | * If this is a socket that does not want to wakeup the device |
2479 | * for it's traffic, the application might be waiting for |
2480 | * close to complete before going to sleep. Send a notification |
2481 | * for this kind of sockets |
2482 | */ |
2483 | if (so->so_options & SO_NOWAKEFROMSLEEP) { |
2484 | socket_post_kev_msg_closed(so); |
2485 | } |
2486 | |
2487 | UDP_LOG_CONNECTION_SUMMARY(inp); |
2488 | |
2489 | in_pcbdetach(inp); |
2490 | inp->inp_state = INPCB_STATE_DEAD; |
2491 | return 0; |
2492 | } |
2493 | |
2494 | int |
2495 | udp_disconnect(struct socket *so) |
2496 | { |
2497 | struct inpcb *inp; |
2498 | |
2499 | inp = sotoinpcb(so); |
2500 | if (inp == NULL) { |
2501 | return EINVAL; |
2502 | } |
2503 | if (inp->inp_faddr.s_addr == INADDR_ANY) { |
2504 | return ENOTCONN; |
2505 | } |
2506 | |
2507 | UDP_LOG_CONNECTION_SUMMARY(inp); |
2508 | |
2509 | in_pcbdisconnect(inp); |
2510 | |
2511 | /* reset flow controlled state, just in case */ |
2512 | inp_reset_fc_state(inp); |
2513 | |
2514 | inp->inp_laddr.s_addr = INADDR_ANY; |
2515 | so->so_state &= ~SS_ISCONNECTED; /* XXX */ |
2516 | inp->inp_last_outifp = NULL; |
2517 | #if SKYWALK |
2518 | if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) { |
2519 | netns_set_ifnet(token: &inp->inp_netns_token, NULL); |
2520 | } |
2521 | #endif /* SKYWALK */ |
2522 | |
2523 | return 0; |
2524 | } |
2525 | |
2526 | int |
2527 | udp_disconnectx(struct socket *so, sae_associd_t aid, sae_connid_t cid) |
2528 | { |
2529 | #pragma unused(cid) |
2530 | if (aid != SAE_ASSOCID_ANY && aid != SAE_ASSOCID_ALL) { |
2531 | return EINVAL; |
2532 | } |
2533 | |
2534 | return udp_disconnect(so); |
2535 | } |
2536 | |
2537 | int |
2538 | udp_send(struct socket *so, int flags, struct mbuf *m, |
2539 | struct sockaddr *addr, struct mbuf *control, struct proc *p) |
2540 | { |
2541 | #ifndef FLOW_DIVERT |
2542 | #pragma unused(flags) |
2543 | #endif /* !(FLOW_DIVERT) */ |
2544 | struct inpcb *inp; |
2545 | int error; |
2546 | |
2547 | inp = sotoinpcb(so); |
2548 | if (inp == NULL) { |
2549 | if (m != NULL) { |
2550 | m_freem(m); |
2551 | } |
2552 | if (control != NULL) { |
2553 | m_freem(control); |
2554 | } |
2555 | return EINVAL; |
2556 | } |
2557 | |
2558 | #if NECP |
2559 | #if FLOW_DIVERT |
2560 | if (necp_socket_should_use_flow_divert(inp)) { |
2561 | /* Implicit connect */ |
2562 | return flow_divert_implicit_data_out(so, flags, data: m, to: addr, |
2563 | control, p); |
2564 | } else { |
2565 | so->so_flags1 |= SOF1_FLOW_DIVERT_SKIP; |
2566 | } |
2567 | #endif /* FLOW_DIVERT */ |
2568 | #endif /* NECP */ |
2569 | |
2570 | #if SKYWALK |
2571 | sk_protect_t protect = sk_async_transmit_protect(); |
2572 | #endif /* SKYWALK */ |
2573 | error = udp_output(inp, m, addr, control, p); |
2574 | #if SKYWALK |
2575 | sk_async_transmit_unprotect(protect); |
2576 | #endif /* SKYWALK */ |
2577 | |
2578 | return error; |
2579 | } |
2580 | |
2581 | int |
2582 | udp_shutdown(struct socket *so) |
2583 | { |
2584 | struct inpcb *inp; |
2585 | |
2586 | inp = sotoinpcb(so); |
2587 | if (inp == NULL) { |
2588 | return EINVAL; |
2589 | } |
2590 | socantsendmore(so); |
2591 | return 0; |
2592 | } |
2593 | |
2594 | int |
2595 | udp_lock(struct socket *so, int refcount, void *debug) |
2596 | { |
2597 | void *lr_saved; |
2598 | |
2599 | if (debug == NULL) { |
2600 | lr_saved = __builtin_return_address(0); |
2601 | } else { |
2602 | lr_saved = debug; |
2603 | } |
2604 | |
2605 | if (so->so_pcb != NULL) { |
2606 | LCK_MTX_ASSERT(&((struct inpcb *)so->so_pcb)->inpcb_mtx, |
2607 | LCK_MTX_ASSERT_NOTOWNED); |
2608 | lck_mtx_lock(lck: &((struct inpcb *)so->so_pcb)->inpcb_mtx); |
2609 | } else { |
2610 | panic("%s: so=%p NO PCB! lr=%p lrh= %s" , __func__, |
2611 | so, lr_saved, solockhistory_nr(so)); |
2612 | /* NOTREACHED */ |
2613 | } |
2614 | if (refcount) { |
2615 | so->so_usecount++; |
2616 | } |
2617 | |
2618 | so->lock_lr[so->next_lock_lr] = lr_saved; |
2619 | so->next_lock_lr = (so->next_lock_lr + 1) % SO_LCKDBG_MAX; |
2620 | return 0; |
2621 | } |
2622 | |
2623 | int |
2624 | udp_unlock(struct socket *so, int refcount, void *debug) |
2625 | { |
2626 | void *lr_saved; |
2627 | |
2628 | if (debug == NULL) { |
2629 | lr_saved = __builtin_return_address(0); |
2630 | } else { |
2631 | lr_saved = debug; |
2632 | } |
2633 | |
2634 | if (refcount) { |
2635 | VERIFY(so->so_usecount > 0); |
2636 | so->so_usecount--; |
2637 | } |
2638 | if (so->so_pcb == NULL) { |
2639 | panic("%s: so=%p NO PCB! lr=%p lrh= %s" , __func__, |
2640 | so, lr_saved, solockhistory_nr(so)); |
2641 | /* NOTREACHED */ |
2642 | } else { |
2643 | LCK_MTX_ASSERT(&((struct inpcb *)so->so_pcb)->inpcb_mtx, |
2644 | LCK_MTX_ASSERT_OWNED); |
2645 | so->unlock_lr[so->next_unlock_lr] = lr_saved; |
2646 | so->next_unlock_lr = (so->next_unlock_lr + 1) % SO_LCKDBG_MAX; |
2647 | lck_mtx_unlock(lck: &((struct inpcb *)so->so_pcb)->inpcb_mtx); |
2648 | } |
2649 | return 0; |
2650 | } |
2651 | |
2652 | lck_mtx_t * |
2653 | udp_getlock(struct socket *so, int flags) |
2654 | { |
2655 | #pragma unused(flags) |
2656 | struct inpcb *inp = sotoinpcb(so); |
2657 | |
2658 | if (so->so_pcb == NULL) { |
2659 | panic("%s: so=%p NULL so_pcb lrh= %s" , __func__, |
2660 | so, solockhistory_nr(so)); |
2661 | /* NOTREACHED */ |
2662 | } |
2663 | return &inp->inpcb_mtx; |
2664 | } |
2665 | |
2666 | /* |
2667 | * UDP garbage collector callback (inpcb_timer_func_t). |
2668 | * |
2669 | * Returns > 0 to keep timer active. |
2670 | */ |
2671 | static void |
2672 | udp_gc(struct inpcbinfo *ipi) |
2673 | { |
2674 | struct inpcb *inp, *inpnxt; |
2675 | struct socket *so; |
2676 | |
2677 | if (lck_rw_try_lock_exclusive(lck: &ipi->ipi_lock) == FALSE) { |
2678 | if (udp_gc_done == TRUE) { |
2679 | udp_gc_done = FALSE; |
2680 | /* couldn't get the lock, must lock next time */ |
2681 | os_atomic_inc(&ipi->ipi_gc_req.intimer_fast, relaxed); |
2682 | return; |
2683 | } |
2684 | lck_rw_lock_exclusive(lck: &ipi->ipi_lock); |
2685 | } |
2686 | |
2687 | udp_gc_done = TRUE; |
2688 | |
2689 | for (inp = udb.lh_first; inp != NULL; inp = inpnxt) { |
2690 | inpnxt = inp->inp_list.le_next; |
2691 | |
2692 | /* |
2693 | * Skip unless it's STOPUSING; garbage collector will |
2694 | * be triggered by in_pcb_checkstate() upon setting |
2695 | * wantcnt to that value. If the PCB is already dead, |
2696 | * keep gc active to anticipate wantcnt changing. |
2697 | */ |
2698 | if (inp->inp_wantcnt != WNT_STOPUSING) { |
2699 | continue; |
2700 | } |
2701 | |
2702 | /* |
2703 | * Skip if busy, no hurry for cleanup. Keep gc active |
2704 | * and try the lock again during next round. |
2705 | */ |
2706 | if (!socket_try_lock(so: inp->inp_socket)) { |
2707 | os_atomic_inc(&ipi->ipi_gc_req.intimer_fast, relaxed); |
2708 | continue; |
2709 | } |
2710 | |
2711 | /* |
2712 | * Keep gc active unless usecount is 0. |
2713 | */ |
2714 | so = inp->inp_socket; |
2715 | if (so->so_usecount == 0) { |
2716 | if (inp->inp_state != INPCB_STATE_DEAD) { |
2717 | if (SOCK_CHECK_DOM(so, PF_INET6)) { |
2718 | in6_pcbdetach(inp); |
2719 | } else { |
2720 | in_pcbdetach(inp); |
2721 | } |
2722 | } |
2723 | in_pcbdispose(inp); |
2724 | } else { |
2725 | socket_unlock(so, refcount: 0); |
2726 | os_atomic_inc(&ipi->ipi_gc_req.intimer_fast, relaxed); |
2727 | } |
2728 | } |
2729 | lck_rw_done(lck: &ipi->ipi_lock); |
2730 | } |
2731 | |
2732 | static int |
2733 | udp_getstat SYSCTL_HANDLER_ARGS |
2734 | { |
2735 | #pragma unused(oidp, arg1, arg2) |
2736 | if (req->oldptr == USER_ADDR_NULL) { |
2737 | req->oldlen = (size_t)sizeof(struct udpstat); |
2738 | } |
2739 | |
2740 | return SYSCTL_OUT(req, &udpstat, MIN(sizeof(udpstat), req->oldlen)); |
2741 | } |
2742 | |
2743 | void |
2744 | udp_in_cksum_stats(u_int32_t len) |
2745 | { |
2746 | udpstat.udps_rcv_swcsum++; |
2747 | udpstat.udps_rcv_swcsum_bytes += len; |
2748 | } |
2749 | |
2750 | void |
2751 | udp_out_cksum_stats(u_int32_t len) |
2752 | { |
2753 | udpstat.udps_snd_swcsum++; |
2754 | udpstat.udps_snd_swcsum_bytes += len; |
2755 | } |
2756 | |
2757 | void |
2758 | udp_in6_cksum_stats(u_int32_t len) |
2759 | { |
2760 | udpstat.udps_rcv6_swcsum++; |
2761 | udpstat.udps_rcv6_swcsum_bytes += len; |
2762 | } |
2763 | |
2764 | void |
2765 | udp_out6_cksum_stats(u_int32_t len) |
2766 | { |
2767 | udpstat.udps_snd6_swcsum++; |
2768 | udpstat.udps_snd6_swcsum_bytes += len; |
2769 | } |
2770 | |
2771 | /* |
2772 | * Checksum extended UDP header and data. |
2773 | */ |
2774 | static int |
2775 | udp_input_checksum(struct mbuf *m, struct udphdr *uh, int off, int ulen) |
2776 | { |
2777 | struct ifnet *ifp = m->m_pkthdr.rcvif; |
2778 | struct ip *ip = mtod(m, struct ip *); |
2779 | struct ipovly *ipov = (struct ipovly *)ip; |
2780 | |
2781 | if (uh->uh_sum == 0) { |
2782 | udpstat.udps_nosum++; |
2783 | return 0; |
2784 | } |
2785 | |
2786 | /* ip_stripoptions() must have been called before we get here */ |
2787 | ASSERT((ip->ip_hl << 2) == sizeof(*ip)); |
2788 | |
2789 | if ((hwcksum_rx || (ifp->if_flags & IFF_LOOPBACK) || |
2790 | (m->m_pkthdr.pkt_flags & PKTF_LOOP)) && |
2791 | (m->m_pkthdr.csum_flags & CSUM_DATA_VALID)) { |
2792 | if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) { |
2793 | uh->uh_sum = m->m_pkthdr.csum_rx_val; |
2794 | } else { |
2795 | uint32_t sum = m->m_pkthdr.csum_rx_val; |
2796 | uint32_t start = m->m_pkthdr.csum_rx_start; |
2797 | int32_t trailer = (m_pktlen(m) - (off + ulen)); |
2798 | |
2799 | /* |
2800 | * Perform 1's complement adjustment of octets |
2801 | * that got included/excluded in the hardware- |
2802 | * calculated checksum value. Ignore cases |
2803 | * where the value already includes the entire |
2804 | * IP header span, as the sum for those octets |
2805 | * would already be 0 by the time we get here; |
2806 | * IP has already performed its header checksum |
2807 | * checks. If we do need to adjust, restore |
2808 | * the original fields in the IP header when |
2809 | * computing the adjustment value. Also take |
2810 | * care of any trailing bytes and subtract out |
2811 | * their partial sum. |
2812 | */ |
2813 | ASSERT(trailer >= 0); |
2814 | if ((m->m_pkthdr.csum_flags & CSUM_PARTIAL) && |
2815 | ((start != 0 && start != off) || trailer != 0)) { |
2816 | uint32_t swbytes = (uint32_t)trailer; |
2817 | |
2818 | if (start < off) { |
2819 | ip->ip_len += sizeof(*ip); |
2820 | #if BYTE_ORDER != BIG_ENDIAN |
2821 | HTONS(ip->ip_len); |
2822 | HTONS(ip->ip_off); |
2823 | #endif /* BYTE_ORDER != BIG_ENDIAN */ |
2824 | } |
2825 | /* callee folds in sum */ |
2826 | sum = m_adj_sum16(m, start, off, ulen, sum); |
2827 | if (off > start) { |
2828 | swbytes += (off - start); |
2829 | } else { |
2830 | swbytes += (start - off); |
2831 | } |
2832 | |
2833 | if (start < off) { |
2834 | #if BYTE_ORDER != BIG_ENDIAN |
2835 | NTOHS(ip->ip_off); |
2836 | NTOHS(ip->ip_len); |
2837 | #endif /* BYTE_ORDER != BIG_ENDIAN */ |
2838 | ip->ip_len -= sizeof(*ip); |
2839 | } |
2840 | |
2841 | if (swbytes != 0) { |
2842 | udp_in_cksum_stats(len: swbytes); |
2843 | } |
2844 | if (trailer != 0) { |
2845 | m_adj(m, -trailer); |
2846 | } |
2847 | } |
2848 | |
2849 | /* callee folds in sum */ |
2850 | uh->uh_sum = in_pseudo(ip->ip_src.s_addr, |
2851 | ip->ip_dst.s_addr, sum + htonl(ulen + IPPROTO_UDP)); |
2852 | } |
2853 | uh->uh_sum ^= 0xffff; |
2854 | } else { |
2855 | uint16_t ip_sum; |
2856 | char b[9]; |
2857 | |
2858 | bcopy(src: ipov->ih_x1, dst: b, n: sizeof(ipov->ih_x1)); |
2859 | bzero(s: ipov->ih_x1, n: sizeof(ipov->ih_x1)); |
2860 | ip_sum = ipov->ih_len; |
2861 | ipov->ih_len = uh->uh_ulen; |
2862 | uh->uh_sum = in_cksum(m, ulen + sizeof(struct ip)); |
2863 | bcopy(src: b, dst: ipov->ih_x1, n: sizeof(ipov->ih_x1)); |
2864 | ipov->ih_len = ip_sum; |
2865 | |
2866 | udp_in_cksum_stats(len: ulen); |
2867 | } |
2868 | |
2869 | if (uh->uh_sum != 0) { |
2870 | udpstat.udps_badsum++; |
2871 | IF_UDP_STATINC(ifp, badchksum); |
2872 | return -1; |
2873 | } |
2874 | |
2875 | return 0; |
2876 | } |
2877 | |
2878 | void |
2879 | udp_fill_keepalive_offload_frames(ifnet_t ifp, |
2880 | struct ifnet_keepalive_offload_frame *frames_array, |
2881 | u_int32_t frames_array_count, size_t frame_data_offset, |
2882 | u_int32_t *used_frames_count) |
2883 | { |
2884 | struct inpcb *inp; |
2885 | inp_gen_t gencnt; |
2886 | u_int32_t frame_index = *used_frames_count; |
2887 | |
2888 | if (ifp == NULL || frames_array == NULL || |
2889 | frames_array_count == 0 || |
2890 | frame_index >= frames_array_count || |
2891 | frame_data_offset >= IFNET_KEEPALIVE_OFFLOAD_FRAME_DATA_SIZE) { |
2892 | return; |
2893 | } |
2894 | |
2895 | lck_rw_lock_shared(lck: &udbinfo.ipi_lock); |
2896 | gencnt = udbinfo.ipi_gencnt; |
2897 | LIST_FOREACH(inp, udbinfo.ipi_listhead, inp_list) { |
2898 | struct socket *so; |
2899 | u_int8_t *data; |
2900 | struct ifnet_keepalive_offload_frame *frame; |
2901 | struct mbuf *m = NULL; |
2902 | |
2903 | if (frame_index >= frames_array_count) { |
2904 | break; |
2905 | } |
2906 | |
2907 | if (inp->inp_gencnt > gencnt || |
2908 | inp->inp_state == INPCB_STATE_DEAD) { |
2909 | continue; |
2910 | } |
2911 | |
2912 | if ((so = inp->inp_socket) == NULL || |
2913 | (so->so_state & SS_DEFUNCT)) { |
2914 | continue; |
2915 | } |
2916 | /* |
2917 | * check for keepalive offload flag without socket |
2918 | * lock to avoid a deadlock |
2919 | */ |
2920 | if (!(inp->inp_flags2 & INP2_KEEPALIVE_OFFLOAD)) { |
2921 | continue; |
2922 | } |
2923 | |
2924 | udp_lock(so, refcount: 1, debug: 0); |
2925 | if (!(inp->inp_vflag & (INP_IPV4 | INP_IPV6))) { |
2926 | udp_unlock(so, refcount: 1, debug: 0); |
2927 | continue; |
2928 | } |
2929 | if ((inp->inp_vflag & INP_IPV4) && |
2930 | (inp->inp_laddr.s_addr == INADDR_ANY || |
2931 | inp->inp_faddr.s_addr == INADDR_ANY)) { |
2932 | udp_unlock(so, refcount: 1, debug: 0); |
2933 | continue; |
2934 | } |
2935 | if ((inp->inp_vflag & INP_IPV6) && |
2936 | (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) || |
2937 | IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))) { |
2938 | udp_unlock(so, refcount: 1, debug: 0); |
2939 | continue; |
2940 | } |
2941 | if (inp->inp_lport == 0 || inp->inp_fport == 0) { |
2942 | udp_unlock(so, refcount: 1, debug: 0); |
2943 | continue; |
2944 | } |
2945 | if (inp->inp_last_outifp == NULL || |
2946 | inp->inp_last_outifp->if_index != ifp->if_index) { |
2947 | udp_unlock(so, refcount: 1, debug: 0); |
2948 | continue; |
2949 | } |
2950 | if ((inp->inp_vflag & INP_IPV4)) { |
2951 | if ((frame_data_offset + sizeof(struct udpiphdr) + |
2952 | inp->inp_keepalive_datalen) > |
2953 | IFNET_KEEPALIVE_OFFLOAD_FRAME_DATA_SIZE) { |
2954 | udp_unlock(so, refcount: 1, debug: 0); |
2955 | continue; |
2956 | } |
2957 | if ((sizeof(struct udpiphdr) + |
2958 | inp->inp_keepalive_datalen) > _MHLEN) { |
2959 | udp_unlock(so, refcount: 1, debug: 0); |
2960 | continue; |
2961 | } |
2962 | } else { |
2963 | if ((frame_data_offset + sizeof(struct ip6_hdr) + |
2964 | sizeof(struct udphdr) + |
2965 | inp->inp_keepalive_datalen) > |
2966 | IFNET_KEEPALIVE_OFFLOAD_FRAME_DATA_SIZE) { |
2967 | udp_unlock(so, refcount: 1, debug: 0); |
2968 | continue; |
2969 | } |
2970 | if ((sizeof(struct ip6_hdr) + sizeof(struct udphdr) + |
2971 | inp->inp_keepalive_datalen) > _MHLEN) { |
2972 | udp_unlock(so, refcount: 1, debug: 0); |
2973 | continue; |
2974 | } |
2975 | } |
2976 | MGETHDR(m, M_WAIT, MT_HEADER); |
2977 | if (m == NULL) { |
2978 | udp_unlock(so, refcount: 1, debug: 0); |
2979 | continue; |
2980 | } |
2981 | /* |
2982 | * This inp has all the information that is needed to |
2983 | * generate an offload frame. |
2984 | */ |
2985 | if (inp->inp_vflag & INP_IPV4) { |
2986 | struct ip *ip; |
2987 | struct udphdr *udp; |
2988 | |
2989 | frame = &frames_array[frame_index]; |
2990 | frame->length = (uint8_t)(frame_data_offset + |
2991 | sizeof(struct udpiphdr) + |
2992 | inp->inp_keepalive_datalen); |
2993 | frame->ether_type = |
2994 | IFNET_KEEPALIVE_OFFLOAD_FRAME_ETHERTYPE_IPV4; |
2995 | frame->interval = inp->inp_keepalive_interval; |
2996 | switch (inp->inp_keepalive_type) { |
2997 | case UDP_KEEPALIVE_OFFLOAD_TYPE_AIRPLAY: |
2998 | frame->type = |
2999 | IFNET_KEEPALIVE_OFFLOAD_FRAME_AIRPLAY; |
3000 | break; |
3001 | default: |
3002 | break; |
3003 | } |
3004 | data = mtod(m, u_int8_t *); |
3005 | bzero(s: data, n: sizeof(struct udpiphdr)); |
3006 | ip = (__typeof__(ip))(void *)data; |
3007 | udp = (__typeof__(udp))(void *) (data + |
3008 | sizeof(struct ip)); |
3009 | m->m_len = sizeof(struct udpiphdr); |
3010 | data = data + sizeof(struct udpiphdr); |
3011 | if (inp->inp_keepalive_datalen > 0 && |
3012 | inp->inp_keepalive_data != NULL) { |
3013 | bcopy(src: inp->inp_keepalive_data, dst: data, |
3014 | n: inp->inp_keepalive_datalen); |
3015 | m->m_len += inp->inp_keepalive_datalen; |
3016 | } |
3017 | m->m_pkthdr.len = m->m_len; |
3018 | |
3019 | ip->ip_v = IPVERSION; |
3020 | ip->ip_hl = (sizeof(struct ip) >> 2); |
3021 | ip->ip_p = IPPROTO_UDP; |
3022 | ip->ip_len = htons(sizeof(struct udpiphdr) + |
3023 | (u_short)inp->inp_keepalive_datalen); |
3024 | ip->ip_ttl = inp->inp_ip_ttl; |
3025 | ip->ip_tos |= (inp->inp_ip_tos & ~IPTOS_ECN_MASK); |
3026 | ip->ip_src = inp->inp_laddr; |
3027 | ip->ip_dst = inp->inp_faddr; |
3028 | ip->ip_sum = in_cksum_hdr_opt(ip); |
3029 | |
3030 | udp->uh_sport = inp->inp_lport; |
3031 | udp->uh_dport = inp->inp_fport; |
3032 | udp->uh_ulen = htons(sizeof(struct udphdr) + |
3033 | (u_short)inp->inp_keepalive_datalen); |
3034 | |
3035 | if (!(inp->inp_flags & INP_UDP_NOCKSUM)) { |
3036 | udp->uh_sum = in_pseudo(ip->ip_src.s_addr, |
3037 | ip->ip_dst.s_addr, |
3038 | htons(sizeof(struct udphdr) + |
3039 | (u_short)inp->inp_keepalive_datalen + |
3040 | IPPROTO_UDP)); |
3041 | m->m_pkthdr.csum_flags = |
3042 | (CSUM_UDP | CSUM_ZERO_INVERT); |
3043 | m->m_pkthdr.csum_data = offsetof(struct udphdr, |
3044 | uh_sum); |
3045 | } |
3046 | m->m_pkthdr.pkt_proto = IPPROTO_UDP; |
3047 | in_delayed_cksum(m); |
3048 | bcopy(src: m_mtod_current(m), dst: frame->data + frame_data_offset, |
3049 | n: m->m_len); |
3050 | } else { |
3051 | struct ip6_hdr *ip6; |
3052 | struct udphdr *udp6; |
3053 | |
3054 | VERIFY(inp->inp_vflag & INP_IPV6); |
3055 | frame = &frames_array[frame_index]; |
3056 | frame->length = (uint8_t)(frame_data_offset + |
3057 | sizeof(struct ip6_hdr) + |
3058 | sizeof(struct udphdr) + |
3059 | inp->inp_keepalive_datalen); |
3060 | frame->ether_type = |
3061 | IFNET_KEEPALIVE_OFFLOAD_FRAME_ETHERTYPE_IPV6; |
3062 | frame->interval = inp->inp_keepalive_interval; |
3063 | switch (inp->inp_keepalive_type) { |
3064 | case UDP_KEEPALIVE_OFFLOAD_TYPE_AIRPLAY: |
3065 | frame->type = |
3066 | IFNET_KEEPALIVE_OFFLOAD_FRAME_AIRPLAY; |
3067 | break; |
3068 | default: |
3069 | break; |
3070 | } |
3071 | data = mtod(m, u_int8_t *); |
3072 | bzero(s: data, n: sizeof(struct ip6_hdr) + sizeof(struct udphdr)); |
3073 | ip6 = (__typeof__(ip6))(void *)data; |
3074 | udp6 = (__typeof__(udp6))(void *)(data + |
3075 | sizeof(struct ip6_hdr)); |
3076 | m->m_len = sizeof(struct ip6_hdr) + |
3077 | sizeof(struct udphdr); |
3078 | data = data + (sizeof(struct ip6_hdr) + |
3079 | sizeof(struct udphdr)); |
3080 | if (inp->inp_keepalive_datalen > 0 && |
3081 | inp->inp_keepalive_data != NULL) { |
3082 | bcopy(src: inp->inp_keepalive_data, dst: data, |
3083 | n: inp->inp_keepalive_datalen); |
3084 | m->m_len += inp->inp_keepalive_datalen; |
3085 | } |
3086 | m->m_pkthdr.len = m->m_len; |
3087 | ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK; |
3088 | ip6->ip6_flow = ip6->ip6_flow & ~IPV6_FLOW_ECN_MASK; |
3089 | ip6->ip6_vfc &= ~IPV6_VERSION_MASK; |
3090 | ip6->ip6_vfc |= IPV6_VERSION; |
3091 | ip6->ip6_nxt = IPPROTO_UDP; |
3092 | ip6->ip6_hlim = (uint8_t)ip6_defhlim; |
3093 | ip6->ip6_plen = htons(sizeof(struct udphdr) + |
3094 | (u_short)inp->inp_keepalive_datalen); |
3095 | ip6->ip6_src = inp->in6p_laddr; |
3096 | if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) { |
3097 | ip6->ip6_src.s6_addr16[1] = 0; |
3098 | } |
3099 | |
3100 | ip6->ip6_dst = inp->in6p_faddr; |
3101 | if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) { |
3102 | ip6->ip6_dst.s6_addr16[1] = 0; |
3103 | } |
3104 | |
3105 | udp6->uh_sport = inp->in6p_lport; |
3106 | udp6->uh_dport = inp->in6p_fport; |
3107 | udp6->uh_ulen = htons(sizeof(struct udphdr) + |
3108 | (u_short)inp->inp_keepalive_datalen); |
3109 | if (!(inp->inp_flags & INP_UDP_NOCKSUM)) { |
3110 | udp6->uh_sum = in6_pseudo(&ip6->ip6_src, |
3111 | &ip6->ip6_dst, |
3112 | htonl(sizeof(struct udphdr) + |
3113 | (u_short)inp->inp_keepalive_datalen + |
3114 | IPPROTO_UDP)); |
3115 | m->m_pkthdr.csum_flags = |
3116 | (CSUM_UDPIPV6 | CSUM_ZERO_INVERT); |
3117 | m->m_pkthdr.csum_data = offsetof(struct udphdr, |
3118 | uh_sum); |
3119 | } |
3120 | m->m_pkthdr.pkt_proto = IPPROTO_UDP; |
3121 | in6_delayed_cksum(m); |
3122 | bcopy(src: m_mtod_current(m), dst: frame->data + frame_data_offset, n: m->m_len); |
3123 | } |
3124 | if (m != NULL) { |
3125 | m_freem(m); |
3126 | m = NULL; |
3127 | } |
3128 | frame_index++; |
3129 | udp_unlock(so, refcount: 1, debug: 0); |
3130 | } |
3131 | lck_rw_done(lck: &udbinfo.ipi_lock); |
3132 | *used_frames_count = frame_index; |
3133 | } |
3134 | |
3135 | int |
3136 | udp_defunct(struct socket *so) |
3137 | { |
3138 | struct ip_moptions *imo; |
3139 | struct inpcb *inp; |
3140 | |
3141 | inp = sotoinpcb(so); |
3142 | if (inp == NULL) { |
3143 | return EINVAL; |
3144 | } |
3145 | |
3146 | imo = inp->inp_moptions; |
3147 | if (imo != NULL) { |
3148 | struct proc *p = current_proc(); |
3149 | |
3150 | SODEFUNCTLOG("%s[%d, %s]: defuncting so 0x%llu drop multicast memberships" , |
3151 | __func__, proc_pid(p), proc_best_name(p), |
3152 | so->so_gencnt); |
3153 | |
3154 | inp->inp_moptions = NULL; |
3155 | |
3156 | IMO_REMREF(imo); |
3157 | } |
3158 | |
3159 | return 0; |
3160 | } |
3161 | |