xref: /xnu-11215.61.5/bsd/netinet/flow_divert.c (revision 4f1223e81cd707a65cc109d0b8ad6653699da3c4)
1 /*
2  * Copyright (c) 2012-2024 Apple Inc. All rights reserved.
3  *
4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5  *
6  * This file contains Original Code and/or Modifications of Original Code
7  * as defined in and that are subject to the Apple Public Source License
8  * Version 2.0 (the 'License'). You may not use this file except in
9  * compliance with the License. The rights granted to you under the License
10  * may not be used to create, or enable the creation or redistribution of,
11  * unlawful or unlicensed copies of an Apple operating system, or to
12  * circumvent, violate, or enable the circumvention or violation of, any
13  * terms of an Apple operating system software license agreement.
14  *
15  * Please obtain a copy of the License at
16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
17  *
18  * The Original Code and all software distributed under the License are
19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23  * Please see the License for the specific language governing rights and
24  * limitations under the License.
25  *
26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27  */
28 
29 #include <string.h>
30 #include <sys/types.h>
31 #include <sys/syslog.h>
32 #include <sys/queue.h>
33 #include <sys/malloc.h>
34 #include <sys/socket.h>
35 #include <sys/kpi_mbuf.h>
36 #include <sys/mbuf.h>
37 #include <sys/domain.h>
38 #include <sys/protosw.h>
39 #include <sys/socketvar.h>
40 #include <sys/kernel.h>
41 #include <sys/systm.h>
42 #include <sys/kern_control.h>
43 #include <sys/ubc.h>
44 #include <sys/codesign.h>
45 #include <sys/file_internal.h>
46 #include <sys/kauth.h>
47 #include <libkern/tree.h>
48 #include <kern/locks.h>
49 #include <kern/debug.h>
50 #include <kern/task.h>
51 #include <mach/task_info.h>
52 #include <net/if_var.h>
53 #include <net/route.h>
54 #include <net/flowhash.h>
55 #include <net/ntstat.h>
56 #include <net/content_filter.h>
57 #include <net/necp.h>
58 #include <netinet/in.h>
59 #include <netinet/in_var.h>
60 #include <netinet/tcp.h>
61 #include <netinet/tcp_var.h>
62 #include <netinet/tcp_fsm.h>
63 #include <netinet/flow_divert.h>
64 #include <netinet/flow_divert_proto.h>
65 #include <netinet6/in6_pcb.h>
66 #include <netinet6/ip6protosw.h>
67 #include <dev/random/randomdev.h>
68 #include <libkern/crypto/sha1.h>
69 #include <libkern/crypto/crypto_internal.h>
70 #include <os/log.h>
71 #include <corecrypto/cc.h>
72 #include <net/sockaddr_utils.h>
73 #if CONTENT_FILTER
74 #include <net/content_filter.h>
75 #endif /* CONTENT_FILTER */
76 
77 #define FLOW_DIVERT_CONNECT_STARTED             0x00000001
78 #define FLOW_DIVERT_READ_CLOSED                 0x00000002
79 #define FLOW_DIVERT_WRITE_CLOSED                0x00000004
80 #define FLOW_DIVERT_TUNNEL_RD_CLOSED            0x00000008
81 #define FLOW_DIVERT_TUNNEL_WR_CLOSED            0x00000010
82 #define FLOW_DIVERT_HAS_HMAC                    0x00000040
83 #define FLOW_DIVERT_NOTIFY_ON_RECEIVED          0x00000080
84 #define FLOW_DIVERT_IMPLICIT_CONNECT            0x00000100
85 #define FLOW_DIVERT_DID_SET_LOCAL_ADDR          0x00000200
86 #define FLOW_DIVERT_HAS_TOKEN                   0x00000400
87 #define FLOW_DIVERT_SHOULD_SET_LOCAL_ADDR       0x00000800
88 #define FLOW_DIVERT_FLOW_IS_TRANSPARENT         0x00001000
89 
90 #define FDLOG(level, pcb, format, ...) \
91 	os_log_with_type(OS_LOG_DEFAULT, flow_divert_syslog_type_to_oslog_type(level), "(%u): " format "\n", (pcb)->hash, __VA_ARGS__)
92 
93 #define FDLOG0(level, pcb, msg) \
94 	os_log_with_type(OS_LOG_DEFAULT, flow_divert_syslog_type_to_oslog_type(level), "(%u): " msg "\n", (pcb)->hash)
95 
96 #define FDRETAIN(pcb)                   if ((pcb) != NULL) OSIncrementAtomic(&(pcb)->ref_count)
97 #define FDRELEASE(pcb)                                                                                                          \
98 	do {                                                                                                                                    \
99 	        if ((pcb) != NULL && 1 == OSDecrementAtomic(&(pcb)->ref_count)) {       \
100 	                flow_divert_pcb_destroy(pcb);                                                                   \
101 	        }                                                                                                                                       \
102 	} while (0)
103 
104 #define FDGRP_RETAIN(grp)  if ((grp) != NULL) OSIncrementAtomic(&(grp)->ref_count)
105 #define FDGRP_RELEASE(grp) if ((grp) != NULL && 1 == OSDecrementAtomic(&(grp)->ref_count)) flow_divert_group_destroy(grp)
106 
107 #define FDLOCK(pcb)                             lck_mtx_lock(&(pcb)->mtx)
108 #define FDUNLOCK(pcb)                           lck_mtx_unlock(&(pcb)->mtx)
109 
110 #define FD_CTL_SENDBUFF_SIZE                    (128 * 1024)
111 
112 #define GROUP_BIT_CTL_ENQUEUE_BLOCKED           0
113 
114 #define GROUP_COUNT_MAX                         31
115 #define FLOW_DIVERT_MAX_NAME_SIZE               4096
116 #define FLOW_DIVERT_MAX_KEY_SIZE                1024
117 #define FLOW_DIVERT_MAX_TRIE_MEMORY             (1024 * 1024)
118 
119 #define CHILD_MAP_SIZE                  256
120 #define NULL_TRIE_IDX                   0xffff
121 #define TRIE_NODE(t, i)                 ((t)->nodes[(i)])
122 #define TRIE_CHILD(t, i, b)             (((t)->child_maps + (CHILD_MAP_SIZE * TRIE_NODE(t, i).child_map))[(b)])
123 #define TRIE_BYTE(t, i)                 ((t)->bytes[(i)])
124 
125 #define SO_IS_DIVERTED(s) (((s)->so_flags & SOF_FLOW_DIVERT) && (s)->so_fd_pcb != NULL)
126 
127 static struct flow_divert_pcb           nil_pcb;
128 
129 static LCK_ATTR_DECLARE(flow_divert_mtx_attr, 0, 0);
130 static LCK_GRP_DECLARE(flow_divert_mtx_grp, FLOW_DIVERT_CONTROL_NAME);
131 static LCK_RW_DECLARE_ATTR(g_flow_divert_group_lck, &flow_divert_mtx_grp,
132     &flow_divert_mtx_attr);
133 
134 static TAILQ_HEAD(_flow_divert_group_list, flow_divert_group) g_flow_divert_in_process_group_list;
135 
136 static struct flow_divert_group         **g_flow_divert_groups __indexable = NULL;
137 static uint32_t                         g_active_group_count    = 0;
138 
139 static  errno_t                         g_init_result           = 0;
140 
141 static  kern_ctl_ref                    g_flow_divert_kctl_ref  = NULL;
142 
143 static struct protosw                   g_flow_divert_in_protosw;
144 static struct pr_usrreqs                g_flow_divert_in_usrreqs;
145 static struct protosw                   g_flow_divert_in_udp_protosw;
146 static struct pr_usrreqs                g_flow_divert_in_udp_usrreqs;
147 static struct ip6protosw                g_flow_divert_in6_protosw;
148 static struct pr_usrreqs                g_flow_divert_in6_usrreqs;
149 static struct ip6protosw                g_flow_divert_in6_udp_protosw;
150 static struct pr_usrreqs                g_flow_divert_in6_udp_usrreqs;
151 
152 static struct protosw                   *g_tcp_protosw          = NULL;
153 static struct ip6protosw                *g_tcp6_protosw         = NULL;
154 static struct protosw                   *g_udp_protosw          = NULL;
155 static struct ip6protosw                *g_udp6_protosw         = NULL;
156 
157 static KALLOC_TYPE_DEFINE(flow_divert_group_zone, struct flow_divert_group,
158     NET_KT_DEFAULT);
159 static KALLOC_TYPE_DEFINE(flow_divert_pcb_zone, struct flow_divert_pcb,
160     NET_KT_DEFAULT);
161 
162 static errno_t
163 flow_divert_dup_addr(sa_family_t family, struct sockaddr *addr, struct sockaddr **dup);
164 
165 static boolean_t
166 flow_divert_is_sockaddr_valid(struct sockaddr *addr);
167 
168 static int
169 flow_divert_append_target_endpoint_tlv(mbuf_ref_t connect_packet, struct sockaddr *toaddr);
170 
171 struct sockaddr *
172 flow_divert_get_buffered_target_address(mbuf_ref_t buffer);
173 
174 static void
175 flow_divert_disconnect_socket(struct socket *so, bool is_connected, bool delay_if_needed);
176 
177 static void flow_divert_group_destroy(struct flow_divert_group *group);
178 
179 static inline uint8_t
flow_divert_syslog_type_to_oslog_type(int syslog_type)180 flow_divert_syslog_type_to_oslog_type(int syslog_type)
181 {
182 	switch (syslog_type) {
183 	case LOG_ERR: return OS_LOG_TYPE_ERROR;
184 	case LOG_INFO: return OS_LOG_TYPE_INFO;
185 	case LOG_DEBUG: return OS_LOG_TYPE_DEBUG;
186 	default: return OS_LOG_TYPE_DEFAULT;
187 	}
188 }
189 
190 static inline int
flow_divert_pcb_cmp(const struct flow_divert_pcb * pcb_a,const struct flow_divert_pcb * pcb_b)191 flow_divert_pcb_cmp(const struct flow_divert_pcb *pcb_a, const struct flow_divert_pcb *pcb_b)
192 {
193 	return memcmp(&pcb_a->hash, &pcb_b->hash, sizeof(pcb_a->hash));
194 }
195 
196 RB_PROTOTYPE(fd_pcb_tree, flow_divert_pcb, rb_link, flow_divert_pcb_cmp);
197 RB_GENERATE(fd_pcb_tree, flow_divert_pcb, rb_link, flow_divert_pcb_cmp);
198 
199 static const char *
flow_divert_packet_type2str(uint8_t packet_type)200 flow_divert_packet_type2str(uint8_t packet_type)
201 {
202 	switch (packet_type) {
203 	case FLOW_DIVERT_PKT_CONNECT:
204 		return "connect";
205 	case FLOW_DIVERT_PKT_CONNECT_RESULT:
206 		return "connect result";
207 	case FLOW_DIVERT_PKT_DATA:
208 		return "data";
209 	case FLOW_DIVERT_PKT_CLOSE:
210 		return "close";
211 	case FLOW_DIVERT_PKT_READ_NOTIFY:
212 		return "read notification";
213 	case FLOW_DIVERT_PKT_PROPERTIES_UPDATE:
214 		return "properties update";
215 	case FLOW_DIVERT_PKT_APP_MAP_CREATE:
216 		return "app map create";
217 	default:
218 		return "unknown";
219 	}
220 }
221 
222 static struct flow_divert_pcb *
flow_divert_pcb_lookup(uint32_t hash,struct flow_divert_group * group)223 flow_divert_pcb_lookup(uint32_t hash, struct flow_divert_group *group)
224 {
225 	struct flow_divert_pcb  key_item;
226 	struct flow_divert_pcb  *fd_cb          = NULL;
227 
228 	key_item.hash = hash;
229 
230 	lck_rw_lock_shared(&group->lck);
231 	fd_cb = RB_FIND(fd_pcb_tree, &group->pcb_tree, &key_item);
232 	FDRETAIN(fd_cb);
233 	lck_rw_done(&group->lck);
234 
235 	return fd_cb;
236 }
237 
238 static struct flow_divert_group *
flow_divert_group_lookup(uint32_t ctl_unit,struct flow_divert_pcb * fd_cb)239 flow_divert_group_lookup(uint32_t ctl_unit, struct flow_divert_pcb *fd_cb)
240 {
241 	struct flow_divert_group *group = NULL;
242 	lck_rw_lock_shared(&g_flow_divert_group_lck);
243 	if (g_active_group_count == 0) {
244 		if (fd_cb != NULL) {
245 			FDLOG0(LOG_ERR, fd_cb, "No active groups, flow divert cannot be used for this socket");
246 		}
247 	} else if (ctl_unit == 0 || (ctl_unit >= GROUP_COUNT_MAX && ctl_unit < FLOW_DIVERT_IN_PROCESS_UNIT_MIN)) {
248 		FDLOG(LOG_ERR, fd_cb, "Cannot lookup group with invalid control unit (%u)", ctl_unit);
249 	} else if (ctl_unit < FLOW_DIVERT_IN_PROCESS_UNIT_MIN) {
250 		if (g_flow_divert_groups == NULL) {
251 			if (fd_cb != NULL) {
252 				FDLOG0(LOG_ERR, fd_cb, "No active non-in-process groups, flow divert cannot be used for this socket");
253 			}
254 		} else {
255 			group = g_flow_divert_groups[ctl_unit];
256 			if (group == NULL) {
257 				if (fd_cb != NULL) {
258 					FDLOG(LOG_ERR, fd_cb, "Group for control unit %u is NULL, flow divert cannot be used for this socket", ctl_unit);
259 				}
260 			} else {
261 				FDGRP_RETAIN(group);
262 			}
263 		}
264 	} else {
265 		if (TAILQ_EMPTY(&g_flow_divert_in_process_group_list)) {
266 			if (fd_cb != NULL) {
267 				FDLOG0(LOG_ERR, fd_cb, "No active in-process groups, flow divert cannot be used for this socket");
268 			}
269 		} else {
270 			struct flow_divert_group *group_cursor = NULL;
271 			TAILQ_FOREACH(group_cursor, &g_flow_divert_in_process_group_list, chain) {
272 				if (group_cursor->ctl_unit == ctl_unit) {
273 					group = group_cursor;
274 					break;
275 				}
276 			}
277 			if (group == NULL) {
278 				if (fd_cb != NULL) {
279 					FDLOG(LOG_ERR, fd_cb, "Group for control unit %u not found, flow divert cannot be used for this socket", ctl_unit);
280 				}
281 			} else if (fd_cb != NULL &&
282 			    (fd_cb->so == NULL ||
283 			    group_cursor->in_process_pid != fd_cb->so->last_pid)) {
284 				FDLOG(LOG_ERR, fd_cb, "Cannot access group for control unit %u, mismatched PID (%u != %u)",
285 				    ctl_unit, group_cursor->in_process_pid, fd_cb->so ? fd_cb->so->last_pid : 0);
286 				group = NULL;
287 			} else {
288 				FDGRP_RETAIN(group);
289 			}
290 		}
291 	}
292 	lck_rw_done(&g_flow_divert_group_lck);
293 	return group;
294 }
295 
296 static errno_t
flow_divert_pcb_insert(struct flow_divert_pcb * fd_cb,struct flow_divert_group * group)297 flow_divert_pcb_insert(struct flow_divert_pcb *fd_cb, struct flow_divert_group *group)
298 {
299 	int error = 0;
300 	lck_rw_lock_exclusive(&group->lck);
301 	if (!(group->flags & FLOW_DIVERT_GROUP_FLAG_DEFUNCT)) {
302 		if (NULL == RB_INSERT(fd_pcb_tree, &group->pcb_tree, fd_cb)) {
303 			fd_cb->group = group;
304 			fd_cb->control_group_unit = group->ctl_unit;
305 			FDRETAIN(fd_cb); /* The group now has a reference */
306 		} else {
307 			FDLOG(LOG_ERR, fd_cb, "Group %u already contains a PCB with hash %u", group->ctl_unit, fd_cb->hash);
308 			error = EEXIST;
309 		}
310 	} else {
311 		FDLOG(LOG_ERR, fd_cb, "Group %u is defunct, cannot insert", group->ctl_unit);
312 		error = ENOENT;
313 	}
314 	lck_rw_done(&group->lck);
315 	return error;
316 }
317 
318 static errno_t
flow_divert_add_to_group(struct flow_divert_pcb * fd_cb,uint32_t ctl_unit)319 flow_divert_add_to_group(struct flow_divert_pcb *fd_cb, uint32_t ctl_unit)
320 {
321 	errno_t error = 0;
322 	struct flow_divert_group *group = NULL;
323 	static uint32_t g_nextkey = 1;
324 	static uint32_t g_hash_seed = 0;
325 	int try_count = 0;
326 
327 	group = flow_divert_group_lookup(ctl_unit, fd_cb);
328 	if (group == NULL) {
329 		return ENOENT;
330 	}
331 
332 	do {
333 		uint32_t key[2];
334 		uint32_t idx;
335 
336 		key[0] = g_nextkey++;
337 		key[1] = RandomULong();
338 
339 		if (g_hash_seed == 0) {
340 			g_hash_seed = RandomULong();
341 		}
342 
343 		error = 0;
344 		fd_cb->hash = net_flowhash(key, sizeof(key), g_hash_seed);
345 
346 		for (idx = 1; idx < GROUP_COUNT_MAX && error == 0; idx++) {
347 			if (idx == ctl_unit) {
348 				continue;
349 			}
350 			struct flow_divert_group *curr_group = flow_divert_group_lookup(idx, NULL);
351 			if (curr_group != NULL) {
352 				lck_rw_lock_shared(&curr_group->lck);
353 				if (NULL != RB_FIND(fd_pcb_tree, &curr_group->pcb_tree, fd_cb)) {
354 					error = EEXIST;
355 				}
356 				lck_rw_done(&curr_group->lck);
357 				FDGRP_RELEASE(curr_group);
358 			}
359 		}
360 
361 		if (error == 0) {
362 			error = flow_divert_pcb_insert(fd_cb, group);
363 		}
364 	} while (error == EEXIST && try_count++ < 3);
365 
366 	if (error == EEXIST) {
367 		FDLOG0(LOG_ERR, fd_cb, "Failed to create a unique hash");
368 		fd_cb->hash = 0;
369 	}
370 
371 	FDGRP_RELEASE(group);
372 	return error;
373 }
374 
375 static struct flow_divert_pcb *
flow_divert_pcb_create(socket_t so)376 flow_divert_pcb_create(socket_t so)
377 {
378 	struct flow_divert_pcb  *new_pcb = NULL;
379 
380 	new_pcb = zalloc_flags(flow_divert_pcb_zone, Z_WAITOK | Z_ZERO);
381 	lck_mtx_init(&new_pcb->mtx, &flow_divert_mtx_grp, &flow_divert_mtx_attr);
382 	new_pcb->so = so;
383 	new_pcb->log_level = nil_pcb.log_level;
384 
385 	FDRETAIN(new_pcb);      /* Represents the socket's reference */
386 
387 	return new_pcb;
388 }
389 
390 static void
flow_divert_pcb_destroy(struct flow_divert_pcb * fd_cb)391 flow_divert_pcb_destroy(struct flow_divert_pcb *fd_cb)
392 {
393 	FDLOG(LOG_INFO, fd_cb, "Destroying, app tx %llu, tunnel tx %llu, tunnel rx %llu",
394 	    fd_cb->bytes_written_by_app, fd_cb->bytes_sent, fd_cb->bytes_received);
395 
396 	if (fd_cb->connect_token != NULL) {
397 		mbuf_freem(fd_cb->connect_token);
398 	}
399 	if (fd_cb->connect_packet != NULL) {
400 		mbuf_freem(fd_cb->connect_packet);
401 	}
402 	if (fd_cb->app_data != NULL) {
403 		kfree_data_sized_by(fd_cb->app_data, fd_cb->app_data_length);
404 	}
405 	if (fd_cb->original_remote_endpoint != NULL) {
406 		free_sockaddr(fd_cb->original_remote_endpoint);
407 	}
408 	zfree(flow_divert_pcb_zone, fd_cb);
409 }
410 
411 static void
flow_divert_pcb_remove(struct flow_divert_pcb * fd_cb)412 flow_divert_pcb_remove(struct flow_divert_pcb *fd_cb)
413 {
414 	if (fd_cb->group != NULL) {
415 		struct flow_divert_group *group = fd_cb->group;
416 		lck_rw_lock_exclusive(&group->lck);
417 		FDLOG(LOG_INFO, fd_cb, "Removing from group %d, ref count = %d", group->ctl_unit, fd_cb->ref_count);
418 		RB_REMOVE(fd_pcb_tree, &group->pcb_tree, fd_cb);
419 		fd_cb->group = NULL;
420 		FDRELEASE(fd_cb);                               /* Release the group's reference */
421 		lck_rw_done(&group->lck);
422 	}
423 }
424 
425 static int
flow_divert_packet_init(struct flow_divert_pcb * fd_cb,uint8_t packet_type,mbuf_ref_t * packet)426 flow_divert_packet_init(struct flow_divert_pcb *fd_cb, uint8_t packet_type, mbuf_ref_t *packet)
427 {
428 	struct flow_divert_packet_header        hdr;
429 	int                                     error           = 0;
430 
431 	error = mbuf_gethdr(MBUF_DONTWAIT, MBUF_TYPE_HEADER, packet);
432 	if (error) {
433 		FDLOG(LOG_ERR, fd_cb, "failed to allocate the header mbuf: %d", error);
434 		return error;
435 	}
436 
437 	hdr.packet_type = packet_type;
438 	hdr.conn_id = htonl(fd_cb->hash);
439 
440 	/* Lay down the header */
441 	error = mbuf_copyback(*packet, 0, sizeof(hdr), &hdr, MBUF_DONTWAIT);
442 	if (error) {
443 		FDLOG(LOG_ERR, fd_cb, "mbuf_copyback(hdr) failed: %d", error);
444 		mbuf_freem(*packet);
445 		*packet = NULL;
446 		return error;
447 	}
448 
449 	return 0;
450 }
451 
452 static int
flow_divert_packet_append_tlv(mbuf_ref_t packet,uint8_t type,uint32_t length,const void __sized_by (length)* value)453 flow_divert_packet_append_tlv(mbuf_ref_t packet, uint8_t type, uint32_t length, const void __sized_by(length) *value)
454 {
455 	uint32_t        net_length      = htonl(length);
456 	int                     error           = 0;
457 
458 	error = mbuf_copyback(packet, mbuf_pkthdr_len(packet), sizeof(type), &type, MBUF_DONTWAIT);
459 	if (error) {
460 		FDLOG(LOG_ERR, &nil_pcb, "failed to append the type (%d)", type);
461 		return error;
462 	}
463 
464 	error = mbuf_copyback(packet, mbuf_pkthdr_len(packet), sizeof(net_length), &net_length, MBUF_DONTWAIT);
465 	if (error) {
466 		FDLOG(LOG_ERR, &nil_pcb, "failed to append the length (%u)", length);
467 		return error;
468 	}
469 
470 	error = mbuf_copyback(packet, mbuf_pkthdr_len(packet), length, value, MBUF_DONTWAIT);
471 	if (error) {
472 		FDLOG0(LOG_ERR, &nil_pcb, "failed to append the value");
473 		return error;
474 	}
475 
476 	return error;
477 }
478 
479 static int
flow_divert_packet_find_tlv(mbuf_ref_t packet,int offset,uint8_t type,int * err,int next)480 flow_divert_packet_find_tlv(mbuf_ref_t packet, int offset, uint8_t type, int *err, int next)
481 {
482 	size_t      cursor      = offset;
483 	int         error       = 0;
484 	uint32_t    curr_length = 0;
485 	uint8_t     curr_type   = 0;
486 
487 	*err = 0;
488 
489 	do {
490 		if (!next) {
491 			error = mbuf_copydata(packet, cursor, sizeof(curr_type), &curr_type);
492 			if (error) {
493 				*err = ENOENT;
494 				return -1;
495 			}
496 		} else {
497 			next = 0;
498 			curr_type = FLOW_DIVERT_TLV_NIL;
499 		}
500 
501 		if (curr_type != type) {
502 			cursor += sizeof(curr_type);
503 			error = mbuf_copydata(packet, cursor, sizeof(curr_length), &curr_length);
504 			if (error) {
505 				*err = error;
506 				return -1;
507 			}
508 
509 			cursor += (sizeof(curr_length) + ntohl(curr_length));
510 		}
511 	} while (curr_type != type);
512 
513 	return (int)cursor;
514 }
515 
516 static int
flow_divert_packet_get_tlv(mbuf_ref_t packet,int offset,uint8_t type,size_t buff_len,void * buff __sized_by (buff_len),uint32_t * val_size)517 flow_divert_packet_get_tlv(mbuf_ref_t packet, int offset, uint8_t type, size_t buff_len, void *buff __sized_by(buff_len), uint32_t *val_size)
518 {
519 	int         error      = 0;
520 	uint32_t    length     = 0;
521 	int         tlv_offset = 0;
522 
523 	tlv_offset = flow_divert_packet_find_tlv(packet, offset, type, &error, 0);
524 	if (tlv_offset < 0) {
525 		return error;
526 	}
527 
528 	error = mbuf_copydata(packet, tlv_offset + sizeof(type), sizeof(length), &length);
529 	if (error) {
530 		return error;
531 	}
532 
533 	length = ntohl(length);
534 
535 	uint32_t data_offset = tlv_offset + sizeof(type) + sizeof(length);
536 
537 	if (length > (mbuf_pkthdr_len(packet) - data_offset)) {
538 		FDLOG(LOG_ERR, &nil_pcb, "Length of %u TLV (%u) is larger than remaining packet data (%lu)", type, length, (mbuf_pkthdr_len(packet) - data_offset));
539 		return EINVAL;
540 	}
541 
542 	if (val_size != NULL) {
543 		*val_size = length;
544 	}
545 
546 	if (buff != NULL && buff_len > 0) {
547 		memset(buff, 0, buff_len);
548 		size_t to_copy = (length < buff_len) ? length : buff_len;
549 		error = mbuf_copydata(packet, data_offset, to_copy, buff);
550 		if (error) {
551 			return error;
552 		}
553 	}
554 
555 	return 0;
556 }
557 
558 static int
flow_divert_packet_compute_hmac(mbuf_ref_t packet,struct flow_divert_group * group,uint8_t * hmac)559 flow_divert_packet_compute_hmac(mbuf_ref_t packet, struct flow_divert_group *group, uint8_t *hmac)
560 {
561 	mbuf_ref_t  curr_mbuf       = packet;
562 
563 	if (g_crypto_funcs == NULL || group->token_key == NULL) {
564 		return ENOPROTOOPT;
565 	}
566 
567 	cchmac_di_decl(g_crypto_funcs->ccsha1_di, hmac_ctx);
568 	g_crypto_funcs->cchmac_init_fn(g_crypto_funcs->ccsha1_di, hmac_ctx, group->token_key_size, group->token_key);
569 
570 	while (curr_mbuf != NULL) {
571 		g_crypto_funcs->cchmac_update_fn(g_crypto_funcs->ccsha1_di, hmac_ctx, mbuf_len(curr_mbuf), mbuf_data(curr_mbuf));
572 		curr_mbuf = mbuf_next(curr_mbuf);
573 	}
574 
575 	g_crypto_funcs->cchmac_final_fn(g_crypto_funcs->ccsha1_di, hmac_ctx, hmac);
576 
577 	return 0;
578 }
579 
580 static int
flow_divert_packet_verify_hmac(mbuf_ref_t packet,uint32_t ctl_unit)581 flow_divert_packet_verify_hmac(mbuf_ref_t packet, uint32_t ctl_unit)
582 {
583 	int error = 0;
584 	struct flow_divert_group *group = NULL;
585 	int hmac_offset;
586 	uint8_t packet_hmac[SHA_DIGEST_LENGTH];
587 	uint8_t computed_hmac[SHA_DIGEST_LENGTH];
588 	mbuf_ref_t tail;
589 
590 	group = flow_divert_group_lookup(ctl_unit, NULL);
591 	if (group == NULL) {
592 		FDLOG(LOG_ERR, &nil_pcb, "Failed to lookup group for control unit %u", ctl_unit);
593 		return ENOPROTOOPT;
594 	}
595 
596 	lck_rw_lock_shared(&group->lck);
597 
598 	if (group->token_key == NULL) {
599 		error = ENOPROTOOPT;
600 		goto done;
601 	}
602 
603 	hmac_offset = flow_divert_packet_find_tlv(packet, 0, FLOW_DIVERT_TLV_HMAC, &error, 0);
604 	if (hmac_offset < 0) {
605 		goto done;
606 	}
607 
608 	error = flow_divert_packet_get_tlv(packet, hmac_offset, FLOW_DIVERT_TLV_HMAC, sizeof(packet_hmac), packet_hmac, NULL);
609 	if (error) {
610 		goto done;
611 	}
612 
613 	/* Chop off the HMAC TLV */
614 	error = mbuf_split(packet, hmac_offset, MBUF_WAITOK, &tail);
615 	if (error) {
616 		goto done;
617 	}
618 
619 	mbuf_free(tail);
620 
621 	error = flow_divert_packet_compute_hmac(packet, group, computed_hmac);
622 	if (error) {
623 		goto done;
624 	}
625 
626 	if (cc_cmp_safe(sizeof(packet_hmac), packet_hmac, computed_hmac)) {
627 		FDLOG0(LOG_WARNING, &nil_pcb, "HMAC in token does not match computed HMAC");
628 		error = EINVAL;
629 		goto done;
630 	}
631 
632 done:
633 	if (group != NULL) {
634 		lck_rw_done(&group->lck);
635 		FDGRP_RELEASE(group);
636 	}
637 	return error;
638 }
639 
640 static void
flow_divert_add_data_statistics(struct flow_divert_pcb * fd_cb,size_t data_len,Boolean send)641 flow_divert_add_data_statistics(struct flow_divert_pcb *fd_cb, size_t data_len, Boolean send)
642 {
643 	struct inpcb *inp = NULL;
644 	struct ifnet *ifp = NULL;
645 	stats_functional_type ifnet_count_type = stats_functional_type_none;
646 
647 	inp = sotoinpcb(fd_cb->so);
648 	if (inp == NULL) {
649 		return;
650 	}
651 
652 	if (inp->inp_vflag & INP_IPV4) {
653 		ifp = inp->inp_last_outifp;
654 	} else if (inp->inp_vflag & INP_IPV6) {
655 		ifp = inp->in6p_last_outifp;
656 	}
657 	if (ifp != NULL) {
658 		ifnet_count_type = IFNET_COUNT_TYPE(ifp);
659 	}
660 
661 	if (send) {
662 		INP_ADD_STAT(inp, ifnet_count_type, txpackets, 1);
663 		INP_ADD_STAT(inp, ifnet_count_type, txbytes, data_len);
664 	} else {
665 		INP_ADD_STAT(inp, ifnet_count_type, rxpackets, 1);
666 		INP_ADD_STAT(inp, ifnet_count_type, rxbytes, data_len);
667 	}
668 	inp_set_activity_bitmap(inp);
669 }
670 
671 static errno_t
flow_divert_check_no_cellular(struct flow_divert_pcb * fd_cb)672 flow_divert_check_no_cellular(struct flow_divert_pcb *fd_cb)
673 {
674 	struct inpcb *inp = sotoinpcb(fd_cb->so);
675 	if (INP_NO_CELLULAR(inp)) {
676 		struct ifnet *ifp = NULL;
677 		if (inp->inp_vflag & INP_IPV4) {
678 			ifp = inp->inp_last_outifp;
679 		} else if (inp->inp_vflag & INP_IPV6) {
680 			ifp = inp->in6p_last_outifp;
681 		}
682 		if (ifp != NULL && IFNET_IS_CELLULAR(ifp)) {
683 			FDLOG0(LOG_ERR, fd_cb, "Cellular is denied");
684 			return EHOSTUNREACH;
685 		}
686 	}
687 	return 0;
688 }
689 
690 static errno_t
flow_divert_check_no_expensive(struct flow_divert_pcb * fd_cb)691 flow_divert_check_no_expensive(struct flow_divert_pcb *fd_cb)
692 {
693 	struct inpcb *inp = sotoinpcb(fd_cb->so);
694 	if (INP_NO_EXPENSIVE(inp)) {
695 		struct ifnet *ifp = NULL;
696 		if (inp->inp_vflag & INP_IPV4) {
697 			ifp = inp->inp_last_outifp;
698 		} else if (inp->inp_vflag & INP_IPV6) {
699 			ifp = inp->in6p_last_outifp;
700 		}
701 		if (ifp != NULL && IFNET_IS_EXPENSIVE(ifp)) {
702 			FDLOG0(LOG_ERR, fd_cb, "Expensive is denied");
703 			return EHOSTUNREACH;
704 		}
705 	}
706 	return 0;
707 }
708 
709 static errno_t
flow_divert_check_no_constrained(struct flow_divert_pcb * fd_cb)710 flow_divert_check_no_constrained(struct flow_divert_pcb *fd_cb)
711 {
712 	struct inpcb *inp = sotoinpcb(fd_cb->so);
713 	if (INP_NO_CONSTRAINED(inp)) {
714 		struct ifnet *ifp = NULL;
715 		if (inp->inp_vflag & INP_IPV4) {
716 			ifp = inp->inp_last_outifp;
717 		} else if (inp->inp_vflag & INP_IPV6) {
718 			ifp = inp->in6p_last_outifp;
719 		}
720 		if (ifp != NULL && IFNET_IS_CONSTRAINED(ifp)) {
721 			FDLOG0(LOG_ERR, fd_cb, "Constrained is denied");
722 			return EHOSTUNREACH;
723 		}
724 	}
725 	return 0;
726 }
727 
728 static void
flow_divert_update_closed_state(struct flow_divert_pcb * fd_cb,int how,bool tunnel,bool flush_snd)729 flow_divert_update_closed_state(struct flow_divert_pcb *fd_cb, int how, bool tunnel, bool flush_snd)
730 {
731 	if (how != SHUT_RD) {
732 		fd_cb->flags |= FLOW_DIVERT_WRITE_CLOSED;
733 		if (tunnel || !(fd_cb->flags & FLOW_DIVERT_CONNECT_STARTED)) {
734 			fd_cb->flags |= FLOW_DIVERT_TUNNEL_WR_CLOSED;
735 			if (flush_snd) {
736 				/* If the tunnel is not accepting writes any more, then flush the send buffer */
737 				sbflush(&fd_cb->so->so_snd);
738 			}
739 		}
740 	}
741 	if (how != SHUT_WR) {
742 		fd_cb->flags |= FLOW_DIVERT_READ_CLOSED;
743 		if (tunnel || !(fd_cb->flags & FLOW_DIVERT_CONNECT_STARTED)) {
744 			fd_cb->flags |= FLOW_DIVERT_TUNNEL_RD_CLOSED;
745 		}
746 	}
747 }
748 
749 static uint16_t
trie_node_alloc(struct flow_divert_trie * trie)750 trie_node_alloc(struct flow_divert_trie *trie)
751 {
752 	if (trie->nodes_free_next < trie->nodes_count) {
753 		uint16_t node_idx = trie->nodes_free_next++;
754 		TRIE_NODE(trie, node_idx).child_map = NULL_TRIE_IDX;
755 		return node_idx;
756 	} else {
757 		return NULL_TRIE_IDX;
758 	}
759 }
760 
761 static uint16_t
trie_child_map_alloc(struct flow_divert_trie * trie)762 trie_child_map_alloc(struct flow_divert_trie *trie)
763 {
764 	if (trie->child_maps_free_next < trie->child_maps_count) {
765 		return trie->child_maps_free_next++;
766 	} else {
767 		return NULL_TRIE_IDX;
768 	}
769 }
770 
771 static uint16_t
trie_bytes_move(struct flow_divert_trie * trie,uint16_t bytes_idx,size_t bytes_size)772 trie_bytes_move(struct flow_divert_trie *trie, uint16_t bytes_idx, size_t bytes_size)
773 {
774 	uint16_t start = trie->bytes_free_next;
775 	if (start + bytes_size <= trie->bytes_count) {
776 		if (start != bytes_idx) {
777 			memmove(&TRIE_BYTE(trie, start), &TRIE_BYTE(trie, bytes_idx), bytes_size);
778 		}
779 		trie->bytes_free_next += bytes_size;
780 		return start;
781 	} else {
782 		return NULL_TRIE_IDX;
783 	}
784 }
785 
786 static uint16_t
flow_divert_trie_insert(struct flow_divert_trie * trie,uint16_t string_start,size_t string_len)787 flow_divert_trie_insert(struct flow_divert_trie *trie, uint16_t string_start, size_t string_len)
788 {
789 	uint16_t current = trie->root;
790 	uint16_t child = trie->root;
791 	uint16_t string_end = string_start + (uint16_t)string_len;
792 	uint16_t string_idx = string_start;
793 	uint16_t string_remainder = (uint16_t)string_len;
794 
795 	while (child != NULL_TRIE_IDX) {
796 		uint16_t parent = current;
797 		uint16_t node_idx;
798 		uint16_t current_end;
799 
800 		current = child;
801 		child = NULL_TRIE_IDX;
802 
803 		current_end = TRIE_NODE(trie, current).start + TRIE_NODE(trie, current).length;
804 
805 		for (node_idx = TRIE_NODE(trie, current).start;
806 		    node_idx < current_end &&
807 		    string_idx < string_end &&
808 		    TRIE_BYTE(trie, node_idx) == TRIE_BYTE(trie, string_idx);
809 		    node_idx++, string_idx++) {
810 			;
811 		}
812 
813 		string_remainder = string_end - string_idx;
814 
815 		if (node_idx < (TRIE_NODE(trie, current).start + TRIE_NODE(trie, current).length)) {
816 			/*
817 			 * We did not reach the end of the current node's string.
818 			 * We need to split the current node into two:
819 			 *   1. A new node that contains the prefix of the node that matches
820 			 *      the prefix of the string being inserted.
821 			 *   2. The current node modified to point to the remainder
822 			 *      of the current node's string.
823 			 */
824 			uint16_t prefix = trie_node_alloc(trie);
825 			if (prefix == NULL_TRIE_IDX) {
826 				FDLOG0(LOG_ERR, &nil_pcb, "Ran out of trie nodes while splitting an existing node");
827 				return NULL_TRIE_IDX;
828 			}
829 
830 			/*
831 			 * Prefix points to the portion of the current nodes's string that has matched
832 			 * the input string thus far.
833 			 */
834 			TRIE_NODE(trie, prefix).start = TRIE_NODE(trie, current).start;
835 			TRIE_NODE(trie, prefix).length = (node_idx - TRIE_NODE(trie, current).start);
836 
837 			/*
838 			 * Prefix has the current node as the child corresponding to the first byte
839 			 * after the split.
840 			 */
841 			TRIE_NODE(trie, prefix).child_map = trie_child_map_alloc(trie);
842 			if (TRIE_NODE(trie, prefix).child_map == NULL_TRIE_IDX) {
843 				FDLOG0(LOG_ERR, &nil_pcb, "Ran out of child maps while splitting an existing node");
844 				return NULL_TRIE_IDX;
845 			}
846 			TRIE_CHILD(trie, prefix, TRIE_BYTE(trie, node_idx)) = current;
847 
848 			/* Parent has the prefix as the child correspoding to the first byte in the prefix */
849 			TRIE_CHILD(trie, parent, TRIE_BYTE(trie, TRIE_NODE(trie, prefix).start)) = prefix;
850 
851 			/* Current node is adjusted to point to the remainder */
852 			TRIE_NODE(trie, current).start = node_idx;
853 			TRIE_NODE(trie, current).length -= TRIE_NODE(trie, prefix).length;
854 
855 			/* We want to insert the new leaf (if any) as a child of the prefix */
856 			current = prefix;
857 		}
858 
859 		if (string_remainder > 0) {
860 			/*
861 			 * We still have bytes in the string that have not been matched yet.
862 			 * If the current node has children, iterate to the child corresponding
863 			 * to the next byte in the string.
864 			 */
865 			if (TRIE_NODE(trie, current).child_map != NULL_TRIE_IDX) {
866 				child = TRIE_CHILD(trie, current, TRIE_BYTE(trie, string_idx));
867 			}
868 		}
869 	} /* while (child != NULL_TRIE_IDX) */
870 
871 	if (string_remainder > 0) {
872 		/* Add a new leaf containing the remainder of the string */
873 		uint16_t leaf = trie_node_alloc(trie);
874 		if (leaf == NULL_TRIE_IDX) {
875 			FDLOG0(LOG_ERR, &nil_pcb, "Ran out of trie nodes while inserting a new leaf");
876 			return NULL_TRIE_IDX;
877 		}
878 
879 		TRIE_NODE(trie, leaf).start = trie_bytes_move(trie, string_idx, string_remainder);
880 		if (TRIE_NODE(trie, leaf).start == NULL_TRIE_IDX) {
881 			FDLOG0(LOG_ERR, &nil_pcb, "Ran out of bytes while inserting a new leaf");
882 			return NULL_TRIE_IDX;
883 		}
884 		TRIE_NODE(trie, leaf).length = string_remainder;
885 
886 		/* Set the new leaf as the child of the current node */
887 		if (TRIE_NODE(trie, current).child_map == NULL_TRIE_IDX) {
888 			TRIE_NODE(trie, current).child_map = trie_child_map_alloc(trie);
889 			if (TRIE_NODE(trie, current).child_map == NULL_TRIE_IDX) {
890 				FDLOG0(LOG_ERR, &nil_pcb, "Ran out of child maps while inserting a new leaf");
891 				return NULL_TRIE_IDX;
892 			}
893 		}
894 		TRIE_CHILD(trie, current, TRIE_BYTE(trie, TRIE_NODE(trie, leaf).start)) = leaf;
895 		current = leaf;
896 	} /* else duplicate or this string is a prefix of one of the existing strings */
897 
898 	return current;
899 }
900 
901 #define APPLE_WEBCLIP_ID_PREFIX "com.apple.webapp"
902 static uint16_t
flow_divert_trie_search(struct flow_divert_trie * trie,const uint8_t * string_bytes __sized_by (string_bytes_count),__unused size_t string_bytes_count)903 flow_divert_trie_search(struct flow_divert_trie *trie, const uint8_t *string_bytes __sized_by(string_bytes_count), __unused size_t string_bytes_count)
904 {
905 	uint16_t current = trie->root;
906 	uint16_t string_idx = 0;
907 
908 	while (current != NULL_TRIE_IDX) {
909 		uint16_t next = NULL_TRIE_IDX;
910 		uint16_t node_end = TRIE_NODE(trie, current).start + TRIE_NODE(trie, current).length;
911 		uint16_t node_idx;
912 
913 		for (node_idx = TRIE_NODE(trie, current).start;
914 		    node_idx < node_end && string_bytes[string_idx] != '\0' && string_bytes[string_idx] == TRIE_BYTE(trie, node_idx);
915 		    node_idx++, string_idx++) {
916 			;
917 		}
918 
919 		if (node_idx == node_end) {
920 			if (string_bytes[string_idx] == '\0') {
921 				return current; /* Got an exact match */
922 			} else if (string_idx == strlen(APPLE_WEBCLIP_ID_PREFIX) &&
923 			    0 == strlcmp((const char *)string_bytes, APPLE_WEBCLIP_ID_PREFIX, string_idx)) {
924 				return current; /* Got an apple webclip id prefix match */
925 			} else if (TRIE_NODE(trie, current).child_map != NULL_TRIE_IDX) {
926 				next = TRIE_CHILD(trie, current, string_bytes[string_idx]);
927 			}
928 		}
929 		current = next;
930 	}
931 
932 	return NULL_TRIE_IDX;
933 }
934 
935 struct uuid_search_info {
936 	uuid_t      target_uuid;
937 	char        *found_signing_id __sized_by(found_signing_id_size);
938 	boolean_t   found_multiple_signing_ids;
939 	proc_t      found_proc;
940 	size_t      found_signing_id_size;
941 };
942 
943 static int
flow_divert_find_proc_by_uuid_callout(proc_t p,void * arg)944 flow_divert_find_proc_by_uuid_callout(proc_t p, void *arg)
945 {
946 	struct uuid_search_info *info = (struct uuid_search_info *)arg;
947 	int result = PROC_RETURNED_DONE; /* By default, we didn't find the process */
948 
949 	if (info->found_signing_id != NULL) {
950 		if (!info->found_multiple_signing_ids) {
951 			/* All processes that were found had the same signing identifier, so just claim this first one and be done. */
952 			info->found_proc = p;
953 			result = PROC_CLAIMED_DONE;
954 		} else {
955 			uuid_string_t uuid_str;
956 			uuid_unparse(info->target_uuid, uuid_str);
957 			FDLOG(LOG_WARNING, &nil_pcb, "Found multiple processes with UUID %s with different signing identifiers", uuid_str);
958 		}
959 		kfree_data_sized_by(info->found_signing_id, info->found_signing_id_size);
960 	}
961 
962 	if (result == PROC_RETURNED_DONE) {
963 		uuid_string_t uuid_str;
964 		uuid_unparse(info->target_uuid, uuid_str);
965 		FDLOG(LOG_WARNING, &nil_pcb, "Failed to find a process with UUID %s", uuid_str);
966 	}
967 
968 	return result;
969 }
970 
971 static int
flow_divert_find_proc_by_uuid_filter(proc_t p,void * arg)972 flow_divert_find_proc_by_uuid_filter(proc_t p, void *arg)
973 {
974 	struct uuid_search_info *info = (struct uuid_search_info *)arg;
975 	int include = 0;
976 
977 	if (info->found_multiple_signing_ids) {
978 		return include;
979 	}
980 
981 	const unsigned char * p_uuid = proc_executableuuid_addr(p);
982 	include = (uuid_compare(p_uuid, info->target_uuid) == 0);
983 	if (include) {
984 		const char *signing_id __null_terminated = cs_identity_get(p);
985 		if (signing_id != NULL) {
986 			FDLOG(LOG_INFO, &nil_pcb, "Found process %d with signing identifier %s", proc_getpid(p), signing_id);
987 			size_t signing_id_size = strlen(signing_id) + 1;
988 			if (info->found_signing_id == NULL) {
989 				info->found_signing_id = kalloc_data(signing_id_size, Z_WAITOK);
990 				info->found_signing_id_size = signing_id_size;
991 				strlcpy(info->found_signing_id, signing_id, signing_id_size);
992 			} else if (strlcmp(info->found_signing_id, signing_id, info->found_signing_id_size)) {
993 				info->found_multiple_signing_ids = TRUE;
994 			}
995 		} else {
996 			info->found_multiple_signing_ids = TRUE;
997 		}
998 		include = !info->found_multiple_signing_ids;
999 	}
1000 
1001 	return include;
1002 }
1003 
1004 static proc_t
flow_divert_find_proc_by_uuid(uuid_t uuid)1005 flow_divert_find_proc_by_uuid(uuid_t uuid)
1006 {
1007 	struct uuid_search_info info;
1008 
1009 	if (LOG_INFO <= nil_pcb.log_level) {
1010 		uuid_string_t uuid_str;
1011 		uuid_unparse(uuid, uuid_str);
1012 		FDLOG(LOG_INFO, &nil_pcb, "Looking for process with UUID %s", uuid_str);
1013 	}
1014 
1015 	memset(&info, 0, sizeof(info));
1016 	info.found_proc = PROC_NULL;
1017 	uuid_copy(info.target_uuid, uuid);
1018 
1019 	proc_iterate(PROC_ALLPROCLIST, flow_divert_find_proc_by_uuid_callout, &info, flow_divert_find_proc_by_uuid_filter, &info);
1020 
1021 	return info.found_proc;
1022 }
1023 
1024 static int
flow_divert_add_proc_info(struct flow_divert_pcb * fd_cb,proc_t proc,const char * signing_id __null_terminated,mbuf_ref_t connect_packet,bool is_effective)1025 flow_divert_add_proc_info(struct flow_divert_pcb *fd_cb, proc_t proc, const char *signing_id __null_terminated, mbuf_ref_t connect_packet, bool is_effective)
1026 {
1027 	int error = 0;
1028 	uint8_t *cdhash = NULL;
1029 	audit_token_t audit_token = {};
1030 	const char *proc_cs_id __null_terminated = signing_id;
1031 
1032 	proc_lock(proc);
1033 
1034 	if (proc_cs_id == NULL) {
1035 		if (proc_getcsflags(proc) & (CS_VALID | CS_DEBUGGED)) {
1036 			proc_cs_id = cs_identity_get(proc);
1037 		} else {
1038 			FDLOG0(LOG_ERR, fd_cb, "Signature of proc is invalid");
1039 		}
1040 	}
1041 
1042 	if (is_effective) {
1043 		lck_rw_lock_shared(&fd_cb->group->lck);
1044 		if (!(fd_cb->group->flags & FLOW_DIVERT_GROUP_FLAG_NO_APP_MAP)) {
1045 			if (proc_cs_id != NULL) {
1046 				size_t proc_cs_id_size = strlen(proc_cs_id) + 1;
1047 				uint16_t result = flow_divert_trie_search(&fd_cb->group->signing_id_trie, (const uint8_t *)__unsafe_null_terminated_to_indexable(proc_cs_id), proc_cs_id_size);
1048 				if (result == NULL_TRIE_IDX) {
1049 					FDLOG(LOG_WARNING, fd_cb, "%s did not match", proc_cs_id);
1050 					error = EPERM;
1051 				} else {
1052 					FDLOG(LOG_INFO, fd_cb, "%s matched", proc_cs_id);
1053 				}
1054 			} else {
1055 				error = EPERM;
1056 			}
1057 		}
1058 		lck_rw_done(&fd_cb->group->lck);
1059 	}
1060 
1061 	if (error != 0) {
1062 		goto done;
1063 	}
1064 
1065 	/*
1066 	 * If signing_id is not NULL then it came from the flow divert token and will be added
1067 	 * as part of the token, so there is no need to add it here.
1068 	 */
1069 	if (signing_id == NULL && proc_cs_id != NULL) {
1070 		error = flow_divert_packet_append_tlv(connect_packet,
1071 		    (is_effective ? FLOW_DIVERT_TLV_SIGNING_ID : FLOW_DIVERT_TLV_APP_REAL_SIGNING_ID),
1072 		    (uint32_t)strlen(proc_cs_id),
1073 		    __terminated_by_to_indexable(proc_cs_id));
1074 		if (error != 0) {
1075 			FDLOG(LOG_ERR, fd_cb, "failed to append the signing ID: %d", error);
1076 			goto done;
1077 		}
1078 	}
1079 
1080 	cdhash = cs_get_cdhash(proc);
1081 	if (cdhash != NULL) {
1082 		error = flow_divert_packet_append_tlv(connect_packet,
1083 		    (is_effective ? FLOW_DIVERT_TLV_CDHASH : FLOW_DIVERT_TLV_APP_REAL_CDHASH),
1084 		    SHA1_RESULTLEN,
1085 		    cdhash);
1086 		if (error) {
1087 			FDLOG(LOG_ERR, fd_cb, "failed to append the cdhash: %d", error);
1088 			goto done;
1089 		}
1090 	} else {
1091 		FDLOG0(LOG_ERR, fd_cb, "failed to get the cdhash");
1092 	}
1093 
1094 	task_t task __single = proc_task(proc);
1095 	if (task != TASK_NULL) {
1096 		mach_msg_type_number_t count = TASK_AUDIT_TOKEN_COUNT;
1097 		kern_return_t rc = task_info(task, TASK_AUDIT_TOKEN, (task_info_t)&audit_token, &count);
1098 		if (rc == KERN_SUCCESS) {
1099 			int append_error = flow_divert_packet_append_tlv(connect_packet,
1100 			    (is_effective ? FLOW_DIVERT_TLV_APP_AUDIT_TOKEN : FLOW_DIVERT_TLV_APP_REAL_AUDIT_TOKEN),
1101 			    sizeof(audit_token_t),
1102 			    &audit_token);
1103 			if (append_error) {
1104 				FDLOG(LOG_ERR, fd_cb, "failed to append app audit token: %d", append_error);
1105 			}
1106 		}
1107 	}
1108 
1109 done:
1110 	proc_unlock(proc);
1111 
1112 	return error;
1113 }
1114 
1115 static int
flow_divert_add_all_proc_info(struct flow_divert_pcb * fd_cb,struct socket * so,proc_t proc,const char * signing_id __null_terminated,mbuf_ref_t connect_packet)1116 flow_divert_add_all_proc_info(struct flow_divert_pcb *fd_cb, struct socket *so, proc_t proc, const char *signing_id __null_terminated, mbuf_ref_t connect_packet)
1117 {
1118 	int error = 0;
1119 	proc_t effective_proc = PROC_NULL;
1120 	proc_t responsible_proc = PROC_NULL;
1121 	proc_t real_proc = proc_find(so->last_pid);
1122 	bool release_real_proc = true;
1123 
1124 	proc_t src_proc = PROC_NULL;
1125 	proc_t real_src_proc = PROC_NULL;
1126 
1127 	if (real_proc == PROC_NULL) {
1128 		FDLOG(LOG_ERR, fd_cb, "failed to find the real proc record for %d", so->last_pid);
1129 		release_real_proc = false;
1130 		real_proc = proc;
1131 		if (real_proc == PROC_NULL) {
1132 			real_proc = current_proc();
1133 		}
1134 	}
1135 
1136 	if (so->so_flags & SOF_DELEGATED) {
1137 		if (proc_getpid(real_proc) != so->e_pid) {
1138 			effective_proc = proc_find(so->e_pid);
1139 		} else {
1140 			const unsigned char * real_proc_uuid = proc_executableuuid_addr(real_proc);
1141 			if (uuid_compare(real_proc_uuid, so->e_uuid)) {
1142 				effective_proc = flow_divert_find_proc_by_uuid(so->e_uuid);
1143 			}
1144 		}
1145 	}
1146 
1147 #if defined(XNU_TARGET_OS_OSX)
1148 	lck_rw_lock_shared(&fd_cb->group->lck);
1149 	if (!(fd_cb->group->flags & FLOW_DIVERT_GROUP_FLAG_NO_APP_MAP)) {
1150 		if (so->so_rpid > 0) {
1151 			responsible_proc = proc_find(so->so_rpid);
1152 		}
1153 	}
1154 	lck_rw_done(&fd_cb->group->lck);
1155 #endif
1156 
1157 	real_src_proc = real_proc;
1158 
1159 	if (responsible_proc != PROC_NULL) {
1160 		src_proc = responsible_proc;
1161 		if (effective_proc != NULL) {
1162 			real_src_proc = effective_proc;
1163 		}
1164 	} else if (effective_proc != PROC_NULL) {
1165 		src_proc = effective_proc;
1166 	} else {
1167 		src_proc = real_proc;
1168 	}
1169 
1170 	error = flow_divert_add_proc_info(fd_cb, src_proc, signing_id, connect_packet, true);
1171 	if (error != 0) {
1172 		goto done;
1173 	}
1174 
1175 	if (real_src_proc != NULL && real_src_proc != src_proc) {
1176 		error = flow_divert_add_proc_info(fd_cb, real_src_proc, NULL, connect_packet, false);
1177 		if (error != 0) {
1178 			goto done;
1179 		}
1180 	}
1181 
1182 done:
1183 	if (responsible_proc != PROC_NULL) {
1184 		proc_rele(responsible_proc);
1185 	}
1186 
1187 	if (effective_proc != PROC_NULL) {
1188 		proc_rele(effective_proc);
1189 	}
1190 
1191 	if (real_proc != PROC_NULL && release_real_proc) {
1192 		proc_rele(real_proc);
1193 	}
1194 
1195 	return error;
1196 }
1197 
1198 static int
flow_divert_send_packet(struct flow_divert_pcb * fd_cb,mbuf_ref_t packet)1199 flow_divert_send_packet(struct flow_divert_pcb *fd_cb, mbuf_ref_t packet)
1200 {
1201 	int             error;
1202 
1203 	if (fd_cb->group == NULL) {
1204 		FDLOG0(LOG_INFO, fd_cb, "no provider, cannot send packet");
1205 		flow_divert_update_closed_state(fd_cb, SHUT_RDWR, true, false);
1206 		flow_divert_disconnect_socket(fd_cb->so, !(fd_cb->flags & FLOW_DIVERT_IMPLICIT_CONNECT), false);
1207 		if (SOCK_TYPE(fd_cb->so) == SOCK_STREAM) {
1208 			error = ECONNABORTED;
1209 		} else {
1210 			error = EHOSTUNREACH;
1211 		}
1212 		fd_cb->so->so_error = (uint16_t)error;
1213 		return error;
1214 	}
1215 
1216 	lck_rw_lock_shared(&fd_cb->group->lck);
1217 
1218 	if (MBUFQ_EMPTY(&fd_cb->group->send_queue)) {
1219 		error = ctl_enqueuembuf(g_flow_divert_kctl_ref, fd_cb->group->ctl_unit, packet, CTL_DATA_EOR);
1220 		if (error) {
1221 			FDLOG(LOG_NOTICE, &nil_pcb, "flow_divert_send_packet: ctl_enqueuembuf returned an error: %d", error);
1222 		}
1223 	} else {
1224 		error = ENOBUFS;
1225 	}
1226 
1227 	if (error == ENOBUFS) {
1228 		if (!lck_rw_lock_shared_to_exclusive(&fd_cb->group->lck)) {
1229 			lck_rw_lock_exclusive(&fd_cb->group->lck);
1230 		}
1231 		MBUFQ_ENQUEUE(&fd_cb->group->send_queue, packet);
1232 		error = 0;
1233 		OSTestAndSet(GROUP_BIT_CTL_ENQUEUE_BLOCKED, &fd_cb->group->atomic_bits);
1234 	}
1235 
1236 	lck_rw_done(&fd_cb->group->lck);
1237 
1238 	return error;
1239 }
1240 
1241 static void
flow_divert_append_domain_name(char * domain_name __null_terminated,void * ctx)1242 flow_divert_append_domain_name(char *domain_name __null_terminated, void *ctx)
1243 {
1244 	mbuf_ref_t packet = (mbuf_ref_t)ctx;
1245 	size_t domain_name_length = 0;
1246 
1247 	if (packet == NULL || domain_name == NULL) {
1248 		return;
1249 	}
1250 
1251 	domain_name_length = strlen(domain_name);
1252 	if (domain_name_length > 0 && domain_name_length < FLOW_DIVERT_MAX_NAME_SIZE) {
1253 		int error = flow_divert_packet_append_tlv(packet, FLOW_DIVERT_TLV_TARGET_HOSTNAME, (uint32_t)domain_name_length, __terminated_by_to_indexable(domain_name));
1254 		if (error) {
1255 			FDLOG(LOG_ERR, &nil_pcb, "Failed to append %s: %d", domain_name, error);
1256 		}
1257 	}
1258 }
1259 
1260 static int
flow_divert_create_connect_packet(struct flow_divert_pcb * fd_cb,struct sockaddr * to,struct socket * so,proc_t p,mbuf_ref_t * out_connect_packet)1261 flow_divert_create_connect_packet(struct flow_divert_pcb *fd_cb, struct sockaddr *to, struct socket *so, proc_t p, mbuf_ref_t *out_connect_packet)
1262 {
1263 	int                     error           = 0;
1264 	int                     flow_type       = 0;
1265 	char *                  signing_id __indexable = NULL;
1266 	uint32_t                sid_size        = 0;
1267 	mbuf_ref_t              connect_packet  = NULL;
1268 	cfil_sock_id_t          cfil_sock_id    = CFIL_SOCK_ID_NONE;
1269 	const void              *cfil_id        = NULL;
1270 	size_t                  cfil_id_size    = 0;
1271 	struct inpcb            *inp            = sotoinpcb(so);
1272 	struct ifnet            *ifp            = NULL;
1273 	uint32_t                flags           = 0;
1274 
1275 	error = flow_divert_packet_init(fd_cb, FLOW_DIVERT_PKT_CONNECT, &connect_packet);
1276 	if (error) {
1277 		goto done;
1278 	}
1279 
1280 	if (fd_cb->connect_token != NULL && (fd_cb->flags & FLOW_DIVERT_HAS_HMAC)) {
1281 		int find_error = flow_divert_packet_get_tlv(fd_cb->connect_token, 0, FLOW_DIVERT_TLV_SIGNING_ID, 0, NULL, &sid_size);
1282 		if (find_error == 0 && sid_size > 0) {
1283 			signing_id = kalloc_data(sid_size + 1, Z_WAITOK | Z_ZERO);
1284 			if (signing_id != NULL) {
1285 				flow_divert_packet_get_tlv(fd_cb->connect_token, 0, FLOW_DIVERT_TLV_SIGNING_ID, sid_size, signing_id, NULL);
1286 				FDLOG(LOG_INFO, fd_cb, "Got %s from token", signing_id);
1287 			}
1288 		}
1289 	}
1290 
1291 	// TODO: remove ternary operator after rdar://121487109 is fixed
1292 	error = flow_divert_add_all_proc_info(fd_cb, so, p, NULL == signing_id ? NULL : __unsafe_null_terminated_from_indexable(signing_id), connect_packet);
1293 
1294 	if (signing_id != NULL) {
1295 		kfree_data(signing_id, sid_size + 1);
1296 	}
1297 
1298 	if (error) {
1299 		FDLOG(LOG_ERR, fd_cb, "Failed to add source proc info: %d", error);
1300 		goto done;
1301 	}
1302 
1303 	error = flow_divert_packet_append_tlv(connect_packet,
1304 	    FLOW_DIVERT_TLV_TRAFFIC_CLASS,
1305 	    sizeof(fd_cb->so->so_traffic_class),
1306 	    &fd_cb->so->so_traffic_class);
1307 	if (error) {
1308 		goto done;
1309 	}
1310 
1311 	if (SOCK_TYPE(fd_cb->so) == SOCK_STREAM) {
1312 		flow_type = FLOW_DIVERT_FLOW_TYPE_TCP;
1313 	} else if (SOCK_TYPE(fd_cb->so) == SOCK_DGRAM) {
1314 		flow_type = FLOW_DIVERT_FLOW_TYPE_UDP;
1315 	} else {
1316 		error = EINVAL;
1317 		goto done;
1318 	}
1319 	error = flow_divert_packet_append_tlv(connect_packet,
1320 	    FLOW_DIVERT_TLV_FLOW_TYPE,
1321 	    sizeof(flow_type),
1322 	    &flow_type);
1323 
1324 	if (error) {
1325 		goto done;
1326 	}
1327 
1328 	if (fd_cb->connect_token != NULL) {
1329 		unsigned int token_len = m_length(fd_cb->connect_token);
1330 		mbuf_concatenate(connect_packet, fd_cb->connect_token);
1331 		mbuf_pkthdr_adjustlen(connect_packet, token_len);
1332 		fd_cb->connect_token = NULL;
1333 	} else {
1334 		error = flow_divert_append_target_endpoint_tlv(connect_packet, to);
1335 		if (error) {
1336 			goto done;
1337 		}
1338 
1339 		necp_with_inp_domain_name(so, connect_packet, flow_divert_append_domain_name);
1340 	}
1341 
1342 	if (fd_cb->local_endpoint.sa.sa_family == AF_INET || fd_cb->local_endpoint.sa.sa_family == AF_INET6) {
1343 		error = flow_divert_packet_append_tlv(connect_packet, FLOW_DIVERT_TLV_LOCAL_ADDR, fd_cb->local_endpoint.sa.sa_len, SA_BYTES(&(fd_cb->local_endpoint.sa)));
1344 		if (error) {
1345 			goto done;
1346 		}
1347 	}
1348 
1349 	if (inp->inp_vflag & INP_IPV4) {
1350 		ifp = inp->inp_last_outifp;
1351 	} else if (inp->inp_vflag & INP_IPV6) {
1352 		ifp = inp->in6p_last_outifp;
1353 	}
1354 	if ((inp->inp_flags & INP_BOUND_IF) ||
1355 	    ((inp->inp_vflag & INP_IPV6) && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) ||
1356 	    ((inp->inp_vflag & INP_IPV4) && inp->inp_laddr.s_addr != INADDR_ANY)) {
1357 		flags |= FLOW_DIVERT_TOKEN_FLAG_BOUND;
1358 		if (ifp == NULL) {
1359 			ifp = inp->inp_boundifp;
1360 		}
1361 	}
1362 	if (ifp != NULL) {
1363 		uint32_t flow_if_index = ifp->if_index;
1364 		error = flow_divert_packet_append_tlv(connect_packet, FLOW_DIVERT_TLV_OUT_IF_INDEX,
1365 		    sizeof(flow_if_index), &flow_if_index);
1366 		if (error) {
1367 			goto done;
1368 		}
1369 	}
1370 
1371 	if (so->so_flags1 & SOF1_DATA_IDEMPOTENT) {
1372 		flags |= FLOW_DIVERT_TOKEN_FLAG_TFO;
1373 	}
1374 
1375 	if (flags != 0) {
1376 		error = flow_divert_packet_append_tlv(connect_packet, FLOW_DIVERT_TLV_FLAGS, sizeof(flags), &flags);
1377 		if (error) {
1378 			goto done;
1379 		}
1380 	}
1381 
1382 	if (SOCK_TYPE(so) == SOCK_DGRAM) {
1383 		cfil_sock_id = cfil_sock_id_from_datagram_socket(so, NULL, to);
1384 	} else {
1385 		cfil_sock_id = cfil_sock_id_from_socket(so);
1386 	}
1387 
1388 	if (cfil_sock_id != CFIL_SOCK_ID_NONE) {
1389 		cfil_id = &cfil_sock_id;
1390 		cfil_id_size = sizeof(cfil_sock_id);
1391 	} else if (so->so_flags1 & SOF1_CONTENT_FILTER_SKIP) {
1392 		cfil_id = &inp->necp_client_uuid;
1393 		cfil_id_size = sizeof(inp->necp_client_uuid);
1394 	}
1395 
1396 	if (cfil_id != NULL && cfil_id_size > 0 && cfil_id_size <= sizeof(uuid_t)) {
1397 		error = flow_divert_packet_append_tlv(connect_packet, FLOW_DIVERT_TLV_CFIL_ID, (uint32_t)cfil_id_size, cfil_id);
1398 		if (error) {
1399 			goto done;
1400 		}
1401 	}
1402 
1403 done:
1404 	if (!error) {
1405 		*out_connect_packet = connect_packet;
1406 	} else if (connect_packet != NULL) {
1407 		mbuf_freem(connect_packet);
1408 	}
1409 
1410 	return error;
1411 }
1412 
1413 static int
flow_divert_send_connect_packet(struct flow_divert_pcb * fd_cb)1414 flow_divert_send_connect_packet(struct flow_divert_pcb *fd_cb)
1415 {
1416 	int             error                   = 0;
1417 	mbuf_ref_t      connect_packet          = fd_cb->connect_packet;
1418 	mbuf_ref_t      saved_connect_packet    = NULL;
1419 
1420 	if (connect_packet != NULL) {
1421 		error = mbuf_copym(connect_packet, 0, mbuf_pkthdr_len(connect_packet), MBUF_DONTWAIT, &saved_connect_packet);
1422 		if (error) {
1423 			FDLOG0(LOG_ERR, fd_cb, "Failed to copy the connect packet");
1424 			goto done;
1425 		}
1426 
1427 		error = flow_divert_send_packet(fd_cb, connect_packet);
1428 		if (error) {
1429 			goto done;
1430 		}
1431 
1432 		fd_cb->connect_packet = saved_connect_packet;
1433 		saved_connect_packet = NULL;
1434 	} else {
1435 		error = ENOENT;
1436 	}
1437 done:
1438 	if (saved_connect_packet != NULL) {
1439 		mbuf_freem(saved_connect_packet);
1440 	}
1441 
1442 	return error;
1443 }
1444 
1445 static int
flow_divert_send_connect_result(struct flow_divert_pcb * fd_cb)1446 flow_divert_send_connect_result(struct flow_divert_pcb *fd_cb)
1447 {
1448 	int             error       = 0;
1449 	mbuf_ref_t      packet      = NULL;
1450 	int             rbuff_space = 0;
1451 
1452 	error = flow_divert_packet_init(fd_cb, FLOW_DIVERT_PKT_CONNECT_RESULT, &packet);
1453 	if (error) {
1454 		FDLOG(LOG_ERR, fd_cb, "failed to create a connect result packet: %d", error);
1455 		goto done;
1456 	}
1457 
1458 	rbuff_space = fd_cb->so->so_rcv.sb_hiwat;
1459 	if (rbuff_space < 0) {
1460 		rbuff_space = 0;
1461 	}
1462 	rbuff_space = htonl(rbuff_space);
1463 	error = flow_divert_packet_append_tlv(packet,
1464 	    FLOW_DIVERT_TLV_SPACE_AVAILABLE,
1465 	    sizeof(rbuff_space),
1466 	    &rbuff_space);
1467 	if (error) {
1468 		goto done;
1469 	}
1470 
1471 	if (fd_cb->local_endpoint.sa.sa_family == AF_INET || fd_cb->local_endpoint.sa.sa_family == AF_INET6) {
1472 		error = flow_divert_packet_append_tlv(packet, FLOW_DIVERT_TLV_LOCAL_ADDR, fd_cb->local_endpoint.sa.sa_len, SA_BYTES(&(fd_cb->local_endpoint.sa)));
1473 		if (error) {
1474 			goto done;
1475 		}
1476 	}
1477 
1478 	error = flow_divert_send_packet(fd_cb, packet);
1479 	if (error) {
1480 		goto done;
1481 	}
1482 
1483 done:
1484 	if (error && packet != NULL) {
1485 		mbuf_freem(packet);
1486 	}
1487 
1488 	return error;
1489 }
1490 
1491 static int
flow_divert_send_close(struct flow_divert_pcb * fd_cb,int how)1492 flow_divert_send_close(struct flow_divert_pcb *fd_cb, int how)
1493 {
1494 	int         error   = 0;
1495 	mbuf_ref_t  packet  = NULL;
1496 	uint32_t    zero    = 0;
1497 
1498 	error = flow_divert_packet_init(fd_cb, FLOW_DIVERT_PKT_CLOSE, &packet);
1499 	if (error) {
1500 		FDLOG(LOG_ERR, fd_cb, "failed to create a close packet: %d", error);
1501 		goto done;
1502 	}
1503 
1504 	error = flow_divert_packet_append_tlv(packet, FLOW_DIVERT_TLV_ERROR_CODE, sizeof(zero), &zero);
1505 	if (error) {
1506 		FDLOG(LOG_ERR, fd_cb, "failed to add the error code TLV: %d", error);
1507 		goto done;
1508 	}
1509 
1510 	how = htonl(how);
1511 	error = flow_divert_packet_append_tlv(packet, FLOW_DIVERT_TLV_HOW, sizeof(how), &how);
1512 	if (error) {
1513 		FDLOG(LOG_ERR, fd_cb, "failed to add the how flag: %d", error);
1514 		goto done;
1515 	}
1516 
1517 	error = flow_divert_send_packet(fd_cb, packet);
1518 	if (error) {
1519 		goto done;
1520 	}
1521 
1522 done:
1523 	if (error && packet != NULL) {
1524 		mbuf_free(packet);
1525 	}
1526 
1527 	return error;
1528 }
1529 
1530 static int
flow_divert_tunnel_how_closed(struct flow_divert_pcb * fd_cb)1531 flow_divert_tunnel_how_closed(struct flow_divert_pcb *fd_cb)
1532 {
1533 	if ((fd_cb->flags & (FLOW_DIVERT_TUNNEL_RD_CLOSED | FLOW_DIVERT_TUNNEL_WR_CLOSED)) ==
1534 	    (FLOW_DIVERT_TUNNEL_RD_CLOSED | FLOW_DIVERT_TUNNEL_WR_CLOSED)) {
1535 		return SHUT_RDWR;
1536 	} else if (fd_cb->flags & FLOW_DIVERT_TUNNEL_RD_CLOSED) {
1537 		return SHUT_RD;
1538 	} else if (fd_cb->flags & FLOW_DIVERT_TUNNEL_WR_CLOSED) {
1539 		return SHUT_WR;
1540 	}
1541 
1542 	return -1;
1543 }
1544 
1545 /*
1546  * Determine what close messages if any need to be sent to the tunnel. Returns TRUE if the tunnel is closed for both reads and
1547  * writes. Returns FALSE otherwise.
1548  */
1549 static void
flow_divert_send_close_if_needed(struct flow_divert_pcb * fd_cb)1550 flow_divert_send_close_if_needed(struct flow_divert_pcb *fd_cb)
1551 {
1552 	int             how             = -1;
1553 
1554 	/* Do not send any close messages if there is still data in the send buffer */
1555 	if (fd_cb->so->so_snd.sb_cc == 0) {
1556 		if ((fd_cb->flags & (FLOW_DIVERT_READ_CLOSED | FLOW_DIVERT_TUNNEL_RD_CLOSED)) == FLOW_DIVERT_READ_CLOSED) {
1557 			/* Socket closed reads, but tunnel did not. Tell tunnel to close reads */
1558 			how = SHUT_RD;
1559 		}
1560 		if ((fd_cb->flags & (FLOW_DIVERT_WRITE_CLOSED | FLOW_DIVERT_TUNNEL_WR_CLOSED)) == FLOW_DIVERT_WRITE_CLOSED) {
1561 			/* Socket closed writes, but tunnel did not. Tell tunnel to close writes */
1562 			if (how == SHUT_RD) {
1563 				how = SHUT_RDWR;
1564 			} else {
1565 				how = SHUT_WR;
1566 			}
1567 		}
1568 	}
1569 
1570 	if (how != -1) {
1571 		FDLOG(LOG_INFO, fd_cb, "sending close, how = %d", how);
1572 		if (flow_divert_send_close(fd_cb, how) != ENOBUFS) {
1573 			/* Successfully sent the close packet. Record the ways in which the tunnel has been closed */
1574 			if (how != SHUT_RD) {
1575 				fd_cb->flags |= FLOW_DIVERT_TUNNEL_WR_CLOSED;
1576 			}
1577 			if (how != SHUT_WR) {
1578 				fd_cb->flags |= FLOW_DIVERT_TUNNEL_RD_CLOSED;
1579 			}
1580 		}
1581 	}
1582 
1583 	if (flow_divert_tunnel_how_closed(fd_cb) == SHUT_RDWR) {
1584 		flow_divert_disconnect_socket(fd_cb->so, !(fd_cb->flags & FLOW_DIVERT_IMPLICIT_CONNECT), false);
1585 	}
1586 }
1587 
1588 static errno_t
flow_divert_send_data_packet(struct flow_divert_pcb * fd_cb,mbuf_ref_t data,size_t data_len)1589 flow_divert_send_data_packet(struct flow_divert_pcb *fd_cb, mbuf_ref_t data, size_t data_len)
1590 {
1591 	mbuf_ref_t  packet = NULL;
1592 	mbuf_ref_t  last   = NULL;
1593 	int         error  = 0;
1594 
1595 	error = flow_divert_packet_init(fd_cb, FLOW_DIVERT_PKT_DATA, &packet);
1596 	if (error || packet == NULL) {
1597 		FDLOG(LOG_ERR, fd_cb, "flow_divert_packet_init failed: %d", error);
1598 		goto done;
1599 	}
1600 
1601 	if (data_len > 0 && data_len <= INT_MAX && data != NULL) {
1602 		last = m_last(packet);
1603 		mbuf_setnext(last, data);
1604 		mbuf_pkthdr_adjustlen(packet, (int)data_len);
1605 	} else {
1606 		data_len = 0;
1607 	}
1608 	error = flow_divert_send_packet(fd_cb, packet);
1609 	if (error == 0 && data_len > 0) {
1610 		fd_cb->bytes_sent += data_len;
1611 		flow_divert_add_data_statistics(fd_cb, data_len, TRUE);
1612 	}
1613 
1614 done:
1615 	if (error) {
1616 		if (last != NULL) {
1617 			mbuf_setnext(last, NULL);
1618 		}
1619 		if (packet != NULL) {
1620 			mbuf_freem(packet);
1621 		}
1622 	}
1623 
1624 	return error;
1625 }
1626 
1627 static errno_t
flow_divert_send_datagram_packet(struct flow_divert_pcb * fd_cb,mbuf_ref_t data,size_t data_len,struct sockaddr * toaddr,Boolean is_fragment,size_t datagram_size)1628 flow_divert_send_datagram_packet(struct flow_divert_pcb *fd_cb, mbuf_ref_t data, size_t data_len, struct sockaddr *toaddr, Boolean is_fragment, size_t datagram_size)
1629 {
1630 	mbuf_ref_t  packet = NULL;
1631 	mbuf_ref_t  last   = NULL;
1632 	int         error  = 0;
1633 
1634 	error = flow_divert_packet_init(fd_cb, FLOW_DIVERT_PKT_DATA, &packet);
1635 	if (error || packet == NULL) {
1636 		FDLOG(LOG_ERR, fd_cb, "flow_divert_packet_init failed: %d", error);
1637 		goto done;
1638 	}
1639 
1640 	if (toaddr != NULL) {
1641 		error = flow_divert_append_target_endpoint_tlv(packet, toaddr);
1642 		if (error) {
1643 			FDLOG(LOG_ERR, fd_cb, "flow_divert_append_target_endpoint_tlv() failed: %d", error);
1644 			goto done;
1645 		}
1646 	}
1647 	if (is_fragment) {
1648 		error = flow_divert_packet_append_tlv(packet, FLOW_DIVERT_TLV_IS_FRAGMENT, sizeof(is_fragment), &is_fragment);
1649 		if (error) {
1650 			FDLOG(LOG_ERR, fd_cb, "flow_divert_packet_append_tlv(FLOW_DIVERT_TLV_IS_FRAGMENT) failed: %d", error);
1651 			goto done;
1652 		}
1653 	}
1654 
1655 	error = flow_divert_packet_append_tlv(packet, FLOW_DIVERT_TLV_DATAGRAM_SIZE, sizeof(datagram_size), &datagram_size);
1656 	if (error) {
1657 		FDLOG(LOG_ERR, fd_cb, "flow_divert_packet_append_tlv(FLOW_DIVERT_TLV_DATAGRAM_SIZE) failed: %d", error);
1658 		goto done;
1659 	}
1660 
1661 	if (data_len > 0 && data_len <= INT_MAX && data != NULL) {
1662 		last = m_last(packet);
1663 		mbuf_setnext(last, data);
1664 		mbuf_pkthdr_adjustlen(packet, (int)data_len);
1665 	} else {
1666 		data_len = 0;
1667 	}
1668 	error = flow_divert_send_packet(fd_cb, packet);
1669 	if (error == 0 && data_len > 0) {
1670 		fd_cb->bytes_sent += data_len;
1671 		flow_divert_add_data_statistics(fd_cb, data_len, TRUE);
1672 	}
1673 
1674 done:
1675 	if (error) {
1676 		if (last != NULL) {
1677 			mbuf_setnext(last, NULL);
1678 		}
1679 		if (packet != NULL) {
1680 			mbuf_freem(packet);
1681 		}
1682 	}
1683 
1684 	return error;
1685 }
1686 
1687 static errno_t
flow_divert_send_fragmented_datagram(struct flow_divert_pcb * fd_cb,mbuf_ref_t datagram,size_t datagram_len,struct sockaddr * toaddr)1688 flow_divert_send_fragmented_datagram(struct flow_divert_pcb *fd_cb, mbuf_ref_t datagram, size_t datagram_len, struct sockaddr *toaddr)
1689 {
1690 	mbuf_ref_t  next_data       = datagram;
1691 	size_t      remaining_len   = datagram_len;
1692 	mbuf_ref_t  remaining_data  = NULL;
1693 	int         error           = 0;
1694 	bool        first           = true;
1695 
1696 	while (remaining_len > 0 && next_data != NULL) {
1697 		size_t to_send = remaining_len;
1698 		remaining_data = NULL;
1699 
1700 		if (to_send > FLOW_DIVERT_CHUNK_SIZE) {
1701 			to_send = FLOW_DIVERT_CHUNK_SIZE;
1702 			error = mbuf_split(next_data, to_send, MBUF_DONTWAIT, &remaining_data);
1703 			if (error) {
1704 				break;
1705 			}
1706 		}
1707 
1708 		error = flow_divert_send_datagram_packet(fd_cb, next_data, to_send, (first ? toaddr : NULL), TRUE, (first ? datagram_len : 0));
1709 		if (error) {
1710 			break;
1711 		}
1712 
1713 		first = false;
1714 		remaining_len -= to_send;
1715 		next_data = remaining_data;
1716 	}
1717 
1718 	if (error) {
1719 		if (next_data != NULL) {
1720 			mbuf_freem(next_data);
1721 		}
1722 		if (remaining_data != NULL) {
1723 			mbuf_freem(remaining_data);
1724 		}
1725 	}
1726 	return error;
1727 }
1728 
1729 static void
flow_divert_send_buffered_data(struct flow_divert_pcb * fd_cb,Boolean force)1730 flow_divert_send_buffered_data(struct flow_divert_pcb *fd_cb, Boolean force)
1731 {
1732 	size_t      to_send;
1733 	size_t      sent    = 0;
1734 	int         error   = 0;
1735 	mbuf_ref_t  buffer;
1736 
1737 	to_send = fd_cb->so->so_snd.sb_cc;
1738 	buffer = fd_cb->so->so_snd.sb_mb;
1739 
1740 	if (buffer == NULL && to_send > 0) {
1741 		FDLOG(LOG_ERR, fd_cb, "Send buffer is NULL, but size is supposed to be %lu", to_send);
1742 		return;
1743 	}
1744 
1745 	/* Ignore the send window if force is enabled */
1746 	if (!force && (to_send > fd_cb->send_window)) {
1747 		to_send = fd_cb->send_window;
1748 	}
1749 
1750 	if (SOCK_TYPE(fd_cb->so) == SOCK_STREAM) {
1751 		while (sent < to_send) {
1752 			mbuf_ref_t  data;
1753 			size_t      data_len;
1754 
1755 			data_len = to_send - sent;
1756 			if (data_len > FLOW_DIVERT_CHUNK_SIZE) {
1757 				data_len = FLOW_DIVERT_CHUNK_SIZE;
1758 			}
1759 
1760 			error = mbuf_copym(buffer, sent, data_len, MBUF_DONTWAIT, &data);
1761 			if (error) {
1762 				FDLOG(LOG_ERR, fd_cb, "mbuf_copym failed: %d", error);
1763 				break;
1764 			}
1765 
1766 			error = flow_divert_send_data_packet(fd_cb, data, data_len);
1767 			if (error) {
1768 				if (data != NULL) {
1769 					mbuf_freem(data);
1770 				}
1771 				break;
1772 			}
1773 
1774 			sent += data_len;
1775 		}
1776 		sbdrop(&fd_cb->so->so_snd, (int)sent);
1777 		sowwakeup(fd_cb->so);
1778 	} else if (SOCK_TYPE(fd_cb->so) == SOCK_DGRAM) {
1779 		mbuf_ref_t  data;
1780 		mbuf_ref_t  m;
1781 		size_t      data_len;
1782 
1783 		while (buffer) {
1784 			struct sockaddr *toaddr = flow_divert_get_buffered_target_address(buffer);
1785 
1786 			m = buffer;
1787 			if (toaddr != NULL) {
1788 				/* look for data in the chain */
1789 				do {
1790 					m = m->m_next;
1791 					if (m != NULL && m->m_type == MT_DATA) {
1792 						break;
1793 					}
1794 				} while (m);
1795 				if (m == NULL) {
1796 					/* unexpected */
1797 					FDLOG0(LOG_ERR, fd_cb, "failed to find type MT_DATA in the mbuf chain.");
1798 					goto move_on;
1799 				}
1800 			}
1801 			data_len = mbuf_pkthdr_len(m);
1802 			if (data_len > 0) {
1803 				FDLOG(LOG_DEBUG, fd_cb, "mbuf_copym() data_len = %lu", data_len);
1804 				error = mbuf_copym(m, 0, data_len, MBUF_DONTWAIT, &data);
1805 				if (error) {
1806 					FDLOG(LOG_ERR, fd_cb, "mbuf_copym failed: %d", error);
1807 					break;
1808 				}
1809 			} else {
1810 				data = NULL;
1811 			}
1812 			if (data_len <= FLOW_DIVERT_CHUNK_SIZE) {
1813 				error = flow_divert_send_datagram_packet(fd_cb, data, data_len, toaddr, FALSE, 0);
1814 			} else {
1815 				error = flow_divert_send_fragmented_datagram(fd_cb, data, data_len, toaddr);
1816 				data = NULL;
1817 			}
1818 			if (error) {
1819 				if (data != NULL) {
1820 					mbuf_freem(data);
1821 				}
1822 				break;
1823 			}
1824 			sent += data_len;
1825 move_on:
1826 			buffer = buffer->m_nextpkt;
1827 			(void) sbdroprecord(&(fd_cb->so->so_snd));
1828 		}
1829 	}
1830 
1831 	if (sent > 0) {
1832 		FDLOG(LOG_DEBUG, fd_cb, "sent %lu bytes of buffered data", sent);
1833 		if (fd_cb->send_window >= sent) {
1834 			fd_cb->send_window -= sent;
1835 		} else {
1836 			fd_cb->send_window = 0;
1837 		}
1838 	}
1839 }
1840 
1841 static int
flow_divert_send_app_data(struct flow_divert_pcb * fd_cb,mbuf_ref_t data,size_t data_size,struct sockaddr * toaddr)1842 flow_divert_send_app_data(struct flow_divert_pcb *fd_cb, mbuf_ref_t data, size_t data_size, struct sockaddr *toaddr)
1843 {
1844 	size_t to_send = data_size;
1845 	int error = 0;
1846 
1847 	if (to_send > fd_cb->send_window) {
1848 		to_send = fd_cb->send_window;
1849 	}
1850 
1851 	if (fd_cb->so->so_snd.sb_cc > 0) {
1852 		to_send = 0;    /* If the send buffer is non-empty, then we can't send anything */
1853 	}
1854 
1855 	if (SOCK_TYPE(fd_cb->so) == SOCK_STREAM) {
1856 		size_t sent = 0;
1857 		mbuf_ref_t remaining_data = data;
1858 		size_t remaining_size = data_size;
1859 		mbuf_ref_t pkt_data = NULL;
1860 		while (sent < to_send && remaining_data != NULL && remaining_size > 0) {
1861 			size_t  pkt_data_len;
1862 
1863 			pkt_data = remaining_data;
1864 
1865 			if ((to_send - sent) > FLOW_DIVERT_CHUNK_SIZE) {
1866 				pkt_data_len = FLOW_DIVERT_CHUNK_SIZE;
1867 			} else {
1868 				pkt_data_len = to_send - sent;
1869 			}
1870 
1871 			if (pkt_data_len < remaining_size) {
1872 				error = mbuf_split(pkt_data, pkt_data_len, MBUF_DONTWAIT, &remaining_data);
1873 				if (error) {
1874 					FDLOG(LOG_ERR, fd_cb, "mbuf_split failed: %d", error);
1875 					pkt_data = NULL;
1876 					break;
1877 				}
1878 				remaining_size -= pkt_data_len;
1879 			} else {
1880 				remaining_data = NULL;
1881 				remaining_size = 0;
1882 			}
1883 
1884 			error = flow_divert_send_data_packet(fd_cb, pkt_data, pkt_data_len);
1885 			if (error) {
1886 				break;
1887 			}
1888 
1889 			pkt_data = NULL;
1890 			sent += pkt_data_len;
1891 		}
1892 
1893 		if (fd_cb->send_window >= sent) {
1894 			fd_cb->send_window -= sent;
1895 		} else {
1896 			fd_cb->send_window = 0;
1897 		}
1898 
1899 		error = 0;
1900 
1901 		if (pkt_data != NULL) {
1902 			if (sbspace(&fd_cb->so->so_snd) > 0) {
1903 				if (!sbappendstream(&fd_cb->so->so_snd, pkt_data)) {
1904 					FDLOG(LOG_ERR, fd_cb, "sbappendstream failed with pkt_data, send buffer size = %u, send_window = %u\n",
1905 					    fd_cb->so->so_snd.sb_cc, fd_cb->send_window);
1906 				}
1907 			} else {
1908 				mbuf_freem(pkt_data);
1909 				error = ENOBUFS;
1910 			}
1911 		}
1912 
1913 		if (remaining_data != NULL) {
1914 			if (sbspace(&fd_cb->so->so_snd) > 0) {
1915 				if (!sbappendstream(&fd_cb->so->so_snd, remaining_data)) {
1916 					FDLOG(LOG_ERR, fd_cb, "sbappendstream failed with remaining_data, send buffer size = %u, send_window = %u\n",
1917 					    fd_cb->so->so_snd.sb_cc, fd_cb->send_window);
1918 				}
1919 			} else {
1920 				mbuf_freem(remaining_data);
1921 				error = ENOBUFS;
1922 			}
1923 		}
1924 	} else if (SOCK_TYPE(fd_cb->so) == SOCK_DGRAM) {
1925 		int send_dgram_error = 0;
1926 		if (to_send || data_size == 0) {
1927 			if (data_size <= FLOW_DIVERT_CHUNK_SIZE) {
1928 				send_dgram_error = flow_divert_send_datagram_packet(fd_cb, data, data_size, toaddr, FALSE, 0);
1929 			} else {
1930 				send_dgram_error = flow_divert_send_fragmented_datagram(fd_cb, data, data_size, toaddr);
1931 				data = NULL;
1932 			}
1933 			if (send_dgram_error) {
1934 				FDLOG(LOG_NOTICE, fd_cb, "flow_divert_send_datagram_packet failed with error %d, send data size = %lu", send_dgram_error, data_size);
1935 			} else {
1936 				if (data_size >= fd_cb->send_window) {
1937 					fd_cb->send_window = 0;
1938 				} else {
1939 					fd_cb->send_window -= data_size;
1940 				}
1941 				data = NULL;
1942 			}
1943 		}
1944 
1945 		if (data != NULL) {
1946 			/* buffer it */
1947 			if (sbspace(&fd_cb->so->so_snd) > 0) {
1948 				if (toaddr != NULL) {
1949 					int append_error = 0;
1950 					if (!sbappendaddr(&fd_cb->so->so_snd, toaddr, data, NULL, &append_error)) {
1951 						FDLOG(LOG_ERR, fd_cb,
1952 						    "sbappendaddr failed. send buffer size = %u, send_window = %u, error = %d",
1953 						    fd_cb->so->so_snd.sb_cc, fd_cb->send_window, append_error);
1954 					}
1955 				} else {
1956 					if (!sbappendrecord(&fd_cb->so->so_snd, data)) {
1957 						FDLOG(LOG_ERR, fd_cb,
1958 						    "sbappendrecord failed. send buffer size = %u, send_window = %u",
1959 						    fd_cb->so->so_snd.sb_cc, fd_cb->send_window);
1960 					}
1961 				}
1962 			} else {
1963 				FDLOG(LOG_ERR, fd_cb, "flow_divert_send_datagram_packet failed with error %d, send data size = %lu, dropping the datagram", error, data_size);
1964 				mbuf_freem(data);
1965 			}
1966 		}
1967 	}
1968 
1969 	return error;
1970 }
1971 
1972 static int
flow_divert_send_read_notification(struct flow_divert_pcb * fd_cb)1973 flow_divert_send_read_notification(struct flow_divert_pcb *fd_cb)
1974 {
1975 	int         error  = 0;
1976 	mbuf_ref_t  packet = NULL;
1977 
1978 	error = flow_divert_packet_init(fd_cb, FLOW_DIVERT_PKT_READ_NOTIFY, &packet);
1979 	if (error) {
1980 		FDLOG(LOG_ERR, fd_cb, "failed to create a read notification packet: %d", error);
1981 		goto done;
1982 	}
1983 
1984 	error = flow_divert_send_packet(fd_cb, packet);
1985 	if (error) {
1986 		goto done;
1987 	}
1988 
1989 done:
1990 	if (error && packet != NULL) {
1991 		mbuf_free(packet);
1992 	}
1993 
1994 	return error;
1995 }
1996 
1997 static int
flow_divert_send_traffic_class_update(struct flow_divert_pcb * fd_cb,int traffic_class)1998 flow_divert_send_traffic_class_update(struct flow_divert_pcb *fd_cb, int traffic_class)
1999 {
2000 	int         error  = 0;
2001 	mbuf_ref_t  packet = NULL;
2002 
2003 	error = flow_divert_packet_init(fd_cb, FLOW_DIVERT_PKT_PROPERTIES_UPDATE, &packet);
2004 	if (error) {
2005 		FDLOG(LOG_ERR, fd_cb, "failed to create a properties update packet: %d", error);
2006 		goto done;
2007 	}
2008 
2009 	error = flow_divert_packet_append_tlv(packet, FLOW_DIVERT_TLV_TRAFFIC_CLASS, sizeof(traffic_class), &traffic_class);
2010 	if (error) {
2011 		FDLOG(LOG_ERR, fd_cb, "failed to add the traffic class: %d", error);
2012 		goto done;
2013 	}
2014 
2015 	error = flow_divert_send_packet(fd_cb, packet);
2016 	if (error) {
2017 		goto done;
2018 	}
2019 
2020 done:
2021 	if (error && packet != NULL) {
2022 		mbuf_free(packet);
2023 	}
2024 
2025 	return error;
2026 }
2027 
2028 static void
flow_divert_set_local_endpoint(struct flow_divert_pcb * fd_cb,struct sockaddr * local_endpoint)2029 flow_divert_set_local_endpoint(struct flow_divert_pcb *fd_cb, struct sockaddr *local_endpoint)
2030 {
2031 	struct inpcb *inp = sotoinpcb(fd_cb->so);
2032 
2033 	if (local_endpoint->sa_family == AF_INET6) {
2034 		if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) && (fd_cb->flags & FLOW_DIVERT_SHOULD_SET_LOCAL_ADDR)) {
2035 			fd_cb->flags |= FLOW_DIVERT_DID_SET_LOCAL_ADDR;
2036 			inp->in6p_laddr = (satosin6(local_endpoint))->sin6_addr;
2037 			inp->inp_lifscope = (satosin6(local_endpoint))->sin6_scope_id;
2038 			in6_verify_ifscope(&inp->in6p_laddr, inp->inp_lifscope);
2039 		}
2040 		if (inp->inp_lport == 0) {
2041 			inp->inp_lport = (satosin6(local_endpoint))->sin6_port;
2042 		}
2043 	} else if (local_endpoint->sa_family == AF_INET) {
2044 		if (inp->inp_laddr.s_addr == INADDR_ANY && (fd_cb->flags & FLOW_DIVERT_SHOULD_SET_LOCAL_ADDR)) {
2045 			fd_cb->flags |= FLOW_DIVERT_DID_SET_LOCAL_ADDR;
2046 			inp->inp_laddr = (satosin(local_endpoint))->sin_addr;
2047 		}
2048 		if (inp->inp_lport == 0) {
2049 			inp->inp_lport = (satosin(local_endpoint))->sin_port;
2050 		}
2051 	}
2052 }
2053 
2054 static void
flow_divert_set_remote_endpoint(struct flow_divert_pcb * fd_cb,struct sockaddr * remote_endpoint)2055 flow_divert_set_remote_endpoint(struct flow_divert_pcb *fd_cb, struct sockaddr *remote_endpoint)
2056 {
2057 	struct inpcb *inp = sotoinpcb(fd_cb->so);
2058 
2059 	if (remote_endpoint->sa_family == AF_INET6) {
2060 		if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
2061 			inp->in6p_faddr = (satosin6(remote_endpoint))->sin6_addr;
2062 			inp->inp_fifscope = (satosin6(remote_endpoint))->sin6_scope_id;
2063 			in6_verify_ifscope(&inp->in6p_faddr, inp->inp_fifscope);
2064 		}
2065 		if (inp->inp_fport == 0) {
2066 			inp->inp_fport = (satosin6(remote_endpoint))->sin6_port;
2067 		}
2068 	} else if (remote_endpoint->sa_family == AF_INET) {
2069 		if (inp->inp_faddr.s_addr == INADDR_ANY) {
2070 			inp->inp_faddr = (satosin(remote_endpoint))->sin_addr;
2071 		}
2072 		if (inp->inp_fport == 0) {
2073 			inp->inp_fport = (satosin(remote_endpoint))->sin_port;
2074 		}
2075 	}
2076 }
2077 
2078 static uint32_t
flow_divert_derive_kernel_control_unit(pid_t pid,uint32_t * ctl_unit,uint32_t * aggregate_unit,bool * is_aggregate)2079 flow_divert_derive_kernel_control_unit(pid_t pid, uint32_t *ctl_unit, uint32_t *aggregate_unit, bool *is_aggregate)
2080 {
2081 	uint32_t result = *ctl_unit;
2082 
2083 	// There are two models supported for deriving control units:
2084 	// 1. A series of flow divert units that allow "transparently" failing
2085 	//    over to the next unit. For this model, the aggregate_unit contains list
2086 	//    of all control units (between 1 and 30) masked over each other.
2087 	// 2. An indication that in-process flow divert should be preferred, with
2088 	//    an out of process flow divert to fail over to. For this model, the
2089 	//    ctl_unit is FLOW_DIVERT_IN_PROCESS_UNIT. In this case, that unit
2090 	//    is returned first, with the unpacked aggregate unit returned as a
2091 	//    fallback.
2092 	*is_aggregate = false;
2093 	if (*ctl_unit == FLOW_DIVERT_IN_PROCESS_UNIT) {
2094 		bool found_unit = false;
2095 		if (pid != 0) {
2096 			// Look for an in-process group that is already open, and use that unit
2097 			struct flow_divert_group *group = NULL;
2098 			TAILQ_FOREACH(group, &g_flow_divert_in_process_group_list, chain) {
2099 				if (group->in_process_pid == pid) {
2100 					// Found an in-process group for our same PID, use it
2101 					found_unit = true;
2102 					result = group->ctl_unit;
2103 					break;
2104 				}
2105 			}
2106 
2107 			// If an in-process group isn't open yet, send a signal up through NECP to request one
2108 			if (!found_unit) {
2109 				necp_client_request_in_process_flow_divert(pid);
2110 			}
2111 		}
2112 
2113 		// If a unit was found, return it
2114 		if (found_unit) {
2115 			if (aggregate_unit != NULL && *aggregate_unit != 0) {
2116 				*is_aggregate = true;
2117 			}
2118 			// The next time around, the aggregate unit values will be picked up
2119 			*ctl_unit = 0;
2120 			return result;
2121 		}
2122 
2123 		// If no unit was found, fall through and clear out the ctl_unit
2124 		result = 0;
2125 		*ctl_unit = 0;
2126 	}
2127 
2128 	if (aggregate_unit != NULL && *aggregate_unit != 0) {
2129 		uint32_t counter;
2130 		struct flow_divert_group *lower_order_group = NULL;
2131 
2132 		for (counter = 0; counter < (GROUP_COUNT_MAX - 1); counter++) {
2133 			if ((*aggregate_unit) & (1 << counter)) {
2134 				struct flow_divert_group *group = NULL;
2135 				group = flow_divert_group_lookup(counter + 1, NULL);
2136 
2137 				if (group != NULL) {
2138 					if (lower_order_group == NULL) {
2139 						lower_order_group = group;
2140 					} else if ((group->order < lower_order_group->order)) {
2141 						lower_order_group = group;
2142 					}
2143 				}
2144 			}
2145 		}
2146 
2147 		if (lower_order_group != NULL) {
2148 			*aggregate_unit &= ~(1 << (lower_order_group->ctl_unit - 1));
2149 			*is_aggregate = true;
2150 			return lower_order_group->ctl_unit;
2151 		} else {
2152 			*ctl_unit = 0;
2153 			return result;
2154 		}
2155 	} else {
2156 		*ctl_unit = 0;
2157 		return result;
2158 	}
2159 }
2160 
2161 static int
flow_divert_try_next_group(struct flow_divert_pcb * fd_cb)2162 flow_divert_try_next_group(struct flow_divert_pcb *fd_cb)
2163 {
2164 	int error = 0;
2165 	uint32_t policy_control_unit = fd_cb->policy_control_unit;
2166 
2167 	flow_divert_pcb_remove(fd_cb);
2168 
2169 	do {
2170 		struct flow_divert_group *next_group = NULL;
2171 		bool is_aggregate = false;
2172 		uint32_t next_ctl_unit = flow_divert_derive_kernel_control_unit(0, &policy_control_unit, &(fd_cb->aggregate_unit), &is_aggregate);
2173 
2174 		if (fd_cb->control_group_unit == next_ctl_unit) {
2175 			FDLOG0(LOG_NOTICE, fd_cb, "Next control unit is the same as the current control unit, disabling flow divert");
2176 			error = EALREADY;
2177 			break;
2178 		}
2179 
2180 		if (next_ctl_unit == 0 || next_ctl_unit >= GROUP_COUNT_MAX) {
2181 			FDLOG0(LOG_NOTICE, fd_cb, "No more valid control units, disabling flow divert");
2182 			error = ENOENT;
2183 			break;
2184 		}
2185 
2186 		next_group = flow_divert_group_lookup(next_ctl_unit, fd_cb);
2187 		if (next_group == NULL) {
2188 			FDLOG(LOG_NOTICE, fd_cb, "Group for control unit %u does not exist", next_ctl_unit);
2189 			continue;
2190 		}
2191 
2192 		FDLOG(LOG_NOTICE, fd_cb, "Moving from %u to %u", fd_cb->control_group_unit, next_ctl_unit);
2193 
2194 		error = flow_divert_pcb_insert(fd_cb, next_group);
2195 		if (error == 0) {
2196 			if (is_aggregate) {
2197 				fd_cb->flags |= FLOW_DIVERT_FLOW_IS_TRANSPARENT;
2198 			} else {
2199 				fd_cb->flags &= ~FLOW_DIVERT_FLOW_IS_TRANSPARENT;
2200 			}
2201 		}
2202 		FDGRP_RELEASE(next_group);
2203 	} while (fd_cb->group == NULL);
2204 
2205 	if (fd_cb->group == NULL) {
2206 		return error ? error : ENOENT;
2207 	}
2208 
2209 	error = flow_divert_send_connect_packet(fd_cb);
2210 	if (error) {
2211 		FDLOG(LOG_NOTICE, fd_cb, "Failed to send the connect packet to %u, disabling flow divert", fd_cb->control_group_unit);
2212 		flow_divert_pcb_remove(fd_cb);
2213 		error = ENOENT;
2214 	}
2215 
2216 	return error;
2217 }
2218 
2219 static void
flow_divert_disable(struct flow_divert_pcb * fd_cb)2220 flow_divert_disable(struct flow_divert_pcb *fd_cb)
2221 {
2222 	struct socket *so = NULL;
2223 	mbuf_ref_t buffer;
2224 	int error = 0;
2225 	proc_t last_proc = NULL;
2226 	struct sockaddr *remote_endpoint = fd_cb->original_remote_endpoint;
2227 	bool do_connect = !(fd_cb->flags & FLOW_DIVERT_IMPLICIT_CONNECT);
2228 	struct inpcb *inp = NULL;
2229 
2230 	so = fd_cb->so;
2231 	if (so == NULL) {
2232 		goto done;
2233 	}
2234 
2235 	FDLOG0(LOG_NOTICE, fd_cb, "Skipped all flow divert services, disabling flow divert");
2236 
2237 	/* Restore the IP state */
2238 	inp = sotoinpcb(so);
2239 	inp->inp_vflag = fd_cb->original_vflag;
2240 	inp->inp_faddr.s_addr = INADDR_ANY;
2241 	inp->inp_fport = 0;
2242 	memset(&(inp->in6p_faddr), 0, sizeof(inp->in6p_faddr));
2243 	inp->inp_fifscope = IFSCOPE_NONE;
2244 	inp->in6p_fport = 0;
2245 	/* If flow divert set the local address, clear it out */
2246 	if (fd_cb->flags & FLOW_DIVERT_DID_SET_LOCAL_ADDR) {
2247 		inp->inp_laddr.s_addr = INADDR_ANY;
2248 		memset(&(inp->in6p_laddr), 0, sizeof(inp->in6p_laddr));
2249 		inp->inp_lifscope = IFSCOPE_NONE;
2250 	}
2251 	inp->inp_last_outifp = fd_cb->original_last_outifp;
2252 	inp->in6p_last_outifp = fd_cb->original_last_outifp6;
2253 
2254 	/* Dis-associate the socket */
2255 	so->so_flags &= ~SOF_FLOW_DIVERT;
2256 	so->so_flags1 |= SOF1_FLOW_DIVERT_SKIP;
2257 	so->so_fd_pcb = NULL;
2258 	fd_cb->so = NULL;
2259 
2260 	FDRELEASE(fd_cb); /* Release the socket's reference */
2261 
2262 	/* Revert back to the original protocol */
2263 	so->so_proto = pffindproto(SOCK_DOM(so), SOCK_PROTO(so), SOCK_TYPE(so));
2264 
2265 	/* Reset the socket state to avoid confusing NECP */
2266 	so->so_state &= ~(SS_ISCONNECTING | SS_ISCONNECTED);
2267 
2268 	last_proc = proc_find(so->last_pid);
2269 
2270 	if (do_connect) {
2271 		/* Connect using the original protocol */
2272 		error = (*so->so_proto->pr_usrreqs->pru_connect)(so, remote_endpoint, (last_proc != NULL ? last_proc : current_proc()));
2273 		if (error) {
2274 			FDLOG(LOG_ERR, fd_cb, "Failed to connect using the socket's original protocol: %d", error);
2275 			goto done;
2276 		}
2277 	}
2278 
2279 	buffer = so->so_snd.sb_mb;
2280 	if (buffer == NULL) {
2281 		/* No buffered data, done */
2282 		goto done;
2283 	}
2284 
2285 	/* Send any buffered data using the original protocol */
2286 	if (SOCK_TYPE(so) == SOCK_STREAM) {
2287 		mbuf_ref_t  data_to_send = NULL;
2288 		size_t      data_len     = so->so_snd.sb_cc;
2289 
2290 		error = mbuf_copym(buffer, 0, data_len, MBUF_DONTWAIT, &data_to_send);
2291 		if (error) {
2292 			FDLOG0(LOG_ERR, fd_cb, "Failed to copy the mbuf chain in the socket's send buffer");
2293 			goto done;
2294 		}
2295 
2296 		sbflush(&so->so_snd);
2297 
2298 		if (data_to_send->m_flags & M_PKTHDR) {
2299 			mbuf_pkthdr_setlen(data_to_send, data_len);
2300 		}
2301 
2302 		error = (*so->so_proto->pr_usrreqs->pru_send)(so,
2303 		    0,
2304 		    data_to_send,
2305 		    NULL,
2306 		    NULL,
2307 		    (last_proc != NULL ? last_proc : current_proc()));
2308 
2309 		if (error && error != EWOULDBLOCK) {
2310 			FDLOG(LOG_ERR, fd_cb, "Failed to send queued TCP data using the socket's original protocol: %d", error);
2311 		} else {
2312 			error = 0;
2313 		}
2314 	} else if (SOCK_TYPE(so) == SOCK_DGRAM) {
2315 		struct sockbuf *sb = &so->so_snd;
2316 		MBUFQ_HEAD(send_queue_head) send_queue;
2317 		MBUFQ_INIT(&send_queue);
2318 
2319 		/* Flush the send buffer, moving all records to a temporary queue */
2320 		while (sb->sb_mb != NULL) {
2321 			mbuf_ref_t record = sb->sb_mb;
2322 			mbuf_ref_t m = record;
2323 			sb->sb_mb = sb->sb_mb->m_nextpkt;
2324 			while (m != NULL) {
2325 				sbfree(sb, m);
2326 				m = m->m_next;
2327 			}
2328 			record->m_nextpkt = NULL;
2329 			MBUFQ_ENQUEUE(&send_queue, record);
2330 		}
2331 		SB_EMPTY_FIXUP(sb);
2332 
2333 		while (!MBUFQ_EMPTY(&send_queue)) {
2334 			mbuf_ref_t next_record = MBUFQ_FIRST(&send_queue);
2335 			mbuf_ref_t addr = NULL;
2336 			mbuf_ref_t control = NULL;
2337 			mbuf_ref_t last_control = NULL;
2338 			mbuf_ref_t data = NULL;
2339 			mbuf_ref_t m = next_record;
2340 			struct sockaddr *to_endpoint = NULL;
2341 
2342 			MBUFQ_DEQUEUE(&send_queue, next_record);
2343 
2344 			while (m != NULL) {
2345 				if (m->m_type == MT_SONAME) {
2346 					addr = m;
2347 				} else if (m->m_type == MT_CONTROL) {
2348 					if (control == NULL) {
2349 						control = m;
2350 					}
2351 					last_control = m;
2352 				} else if (m->m_type == MT_DATA) {
2353 					data = m;
2354 					break;
2355 				}
2356 				m = m->m_next;
2357 			}
2358 
2359 			if (addr != NULL && !do_connect) {
2360 				to_endpoint = flow_divert_get_buffered_target_address(addr);
2361 				if (to_endpoint == NULL) {
2362 					FDLOG0(LOG_NOTICE, fd_cb, "Failed to get the remote address from the buffer");
2363 				}
2364 			}
2365 
2366 			if (data == NULL) {
2367 				FDLOG0(LOG_ERR, fd_cb, "Buffered record does not contain any data");
2368 				mbuf_freem(next_record);
2369 				continue;
2370 			}
2371 
2372 			if (!(data->m_flags & M_PKTHDR)) {
2373 				FDLOG0(LOG_ERR, fd_cb, "Buffered data does not have a packet header");
2374 				mbuf_freem(next_record);
2375 				continue;
2376 			}
2377 
2378 			if (addr != NULL) {
2379 				addr->m_next = NULL;
2380 			}
2381 
2382 			if (last_control != NULL) {
2383 				last_control->m_next = NULL;
2384 			}
2385 
2386 			error = (*so->so_proto->pr_usrreqs->pru_send)(so,
2387 			    0,
2388 			    data,
2389 			    to_endpoint,
2390 			    control,
2391 			    (last_proc != NULL ? last_proc : current_proc()));
2392 
2393 			if (addr != NULL) {
2394 				mbuf_freem(addr);
2395 			}
2396 
2397 			if (error) {
2398 				FDLOG(LOG_ERR, fd_cb, "Failed to send queued UDP data using the socket's original protocol: %d", error);
2399 			}
2400 		}
2401 	}
2402 done:
2403 	if (last_proc != NULL) {
2404 		proc_rele(last_proc);
2405 	}
2406 
2407 	if (error && so != NULL) {
2408 		so->so_error = (uint16_t)error;
2409 		flow_divert_disconnect_socket(so, do_connect, false);
2410 	}
2411 }
2412 
2413 static void
flow_divert_scope(struct flow_divert_pcb * fd_cb,int out_if_index,bool derive_new_address)2414 flow_divert_scope(struct flow_divert_pcb *fd_cb, int out_if_index, bool derive_new_address)
2415 {
2416 	struct socket           *so             = NULL;
2417 	struct inpcb            *inp            = NULL;
2418 	struct ifnet            *current_ifp    = NULL;
2419 	struct ifnet * __single new_ifp         = NULL;
2420 	int                     error           = 0;
2421 
2422 	so = fd_cb->so;
2423 	if (so == NULL) {
2424 		return;
2425 	}
2426 
2427 	inp = sotoinpcb(so);
2428 
2429 	if (out_if_index <= 0) {
2430 		return;
2431 	}
2432 
2433 	if (inp->inp_vflag & INP_IPV6) {
2434 		current_ifp = inp->in6p_last_outifp;
2435 	} else {
2436 		current_ifp = inp->inp_last_outifp;
2437 	}
2438 
2439 	if (current_ifp != NULL) {
2440 		if (current_ifp->if_index == out_if_index) {
2441 			/* No change */
2442 			return;
2443 		}
2444 
2445 		/* Scope the socket to the given interface */
2446 		error = inp_bindif(inp, out_if_index, &new_ifp);
2447 		if (error != 0) {
2448 			FDLOG(LOG_ERR, fd_cb, "failed to scope to %d because inp_bindif returned %d", out_if_index, error);
2449 			return;
2450 		}
2451 
2452 		if (derive_new_address && fd_cb->original_remote_endpoint != NULL) {
2453 			/* Get the appropriate address for the given interface */
2454 			if (inp->inp_vflag & INP_IPV6) {
2455 				inp->in6p_laddr = sa6_any.sin6_addr;
2456 				error = in6_pcbladdr(inp, fd_cb->original_remote_endpoint, &(fd_cb->local_endpoint.sin6.sin6_addr), NULL);
2457 			} else {
2458 				inp->inp_laddr.s_addr = INADDR_ANY;
2459 				error = in_pcbladdr(inp, fd_cb->original_remote_endpoint, &(fd_cb->local_endpoint.sin.sin_addr), IFSCOPE_NONE, NULL, 0);
2460 			}
2461 
2462 			if (error != 0) {
2463 				FDLOG(LOG_WARNING, fd_cb, "failed to derive a new local address from %d because in_pcbladdr returned %d", out_if_index, error);
2464 			}
2465 		}
2466 	} else {
2467 		ifnet_head_lock_shared();
2468 		if (IF_INDEX_IN_RANGE(out_if_index)) {
2469 			new_ifp = ifindex2ifnet[out_if_index];
2470 		}
2471 		ifnet_head_done();
2472 	}
2473 
2474 	/* Update the "last interface" of the socket */
2475 	if (new_ifp != NULL) {
2476 		if (inp->inp_vflag & INP_IPV6) {
2477 			inp->in6p_last_outifp = new_ifp;
2478 		} else {
2479 			inp->inp_last_outifp = new_ifp;
2480 		}
2481 
2482 #if SKYWALK
2483 		if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
2484 			netns_set_ifnet(&inp->inp_netns_token, new_ifp);
2485 		}
2486 #endif /* SKYWALK */
2487 	}
2488 }
2489 
2490 static void
flow_divert_handle_connect_result(struct flow_divert_pcb * fd_cb,mbuf_ref_t packet,int offset)2491 flow_divert_handle_connect_result(struct flow_divert_pcb *fd_cb, mbuf_ref_t packet, int offset)
2492 {
2493 	uint32_t              connect_error   = 0;
2494 	uint32_t              ctl_unit        = 0;
2495 	int                   error           = 0;
2496 	union sockaddr_in_4_6 local_endpoint  = {};
2497 	union sockaddr_in_4_6 remote_endpoint = {};
2498 	int                   out_if_index    = 0;
2499 	uint32_t              send_window     = 0;
2500 	uint32_t              app_data_length = 0;
2501 
2502 	memset(&local_endpoint, 0, sizeof(local_endpoint));
2503 	memset(&remote_endpoint, 0, sizeof(remote_endpoint));
2504 
2505 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_ERROR_CODE, sizeof(connect_error), &connect_error, NULL);
2506 	if (error) {
2507 		FDLOG(LOG_ERR, fd_cb, "failed to get the connect result: %d", error);
2508 		return;
2509 	}
2510 
2511 	connect_error = ntohl(connect_error);
2512 	FDLOG(LOG_INFO, fd_cb, "received connect result %u", connect_error);
2513 
2514 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_SPACE_AVAILABLE, sizeof(send_window), &send_window, NULL);
2515 	if (error) {
2516 		FDLOG(LOG_ERR, fd_cb, "failed to get the send window: %d", error);
2517 		return;
2518 	}
2519 
2520 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_CTL_UNIT, sizeof(ctl_unit), &ctl_unit, NULL);
2521 	if (error) {
2522 		FDLOG0(LOG_INFO, fd_cb, "No control unit provided in the connect result");
2523 	}
2524 
2525 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_LOCAL_ADDR, sizeof(local_endpoint), &(local_endpoint.sin6), NULL);
2526 	if (error) {
2527 		FDLOG0(LOG_INFO, fd_cb, "No local address provided");
2528 	}
2529 
2530 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_REMOTE_ADDR, sizeof(remote_endpoint), &(remote_endpoint.sin6), NULL);
2531 	if (error) {
2532 		FDLOG0(LOG_INFO, fd_cb, "No remote address provided");
2533 	}
2534 
2535 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_OUT_IF_INDEX, sizeof(out_if_index), &out_if_index, NULL);
2536 	if (error) {
2537 		FDLOG0(LOG_INFO, fd_cb, "No output if index provided");
2538 	}
2539 
2540 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_APP_DATA, 0, NULL, &app_data_length);
2541 	if (error) {
2542 		FDLOG0(LOG_INFO, fd_cb, "No application data provided in connect result");
2543 	}
2544 
2545 	error = 0;
2546 
2547 	FDLOCK(fd_cb);
2548 	if (fd_cb->so != NULL) {
2549 		struct inpcb *inp = NULL;
2550 		struct socket *so = fd_cb->so;
2551 		bool local_address_is_valid = false;
2552 
2553 		socket_lock(so, 1);
2554 
2555 		if (!(so->so_flags & SOF_FLOW_DIVERT)) {
2556 			FDLOG0(LOG_NOTICE, fd_cb, "socket is not attached any more, ignoring connect result");
2557 			goto done;
2558 		}
2559 
2560 		if (SOCK_TYPE(so) == SOCK_STREAM && !(so->so_state & SS_ISCONNECTING)) {
2561 			FDLOG0(LOG_ERR, fd_cb, "TCP socket is not in the connecting state, ignoring connect result");
2562 			goto done;
2563 		}
2564 
2565 		inp = sotoinpcb(so);
2566 
2567 		if (connect_error || error) {
2568 			goto set_socket_state;
2569 		}
2570 
2571 		if (flow_divert_is_sockaddr_valid(SA(&local_endpoint))) {
2572 			if (local_endpoint.sa.sa_family == AF_INET) {
2573 				local_endpoint.sa.sa_len = sizeof(struct sockaddr_in);
2574 				if ((inp->inp_vflag & INP_IPV4) && local_endpoint.sin.sin_addr.s_addr != INADDR_ANY) {
2575 					local_address_is_valid = true;
2576 					fd_cb->local_endpoint = local_endpoint;
2577 					inp->inp_laddr.s_addr = INADDR_ANY;
2578 				} else {
2579 					fd_cb->local_endpoint.sin.sin_port = local_endpoint.sin.sin_port;
2580 				}
2581 			} else if (local_endpoint.sa.sa_family == AF_INET6) {
2582 				local_endpoint.sa.sa_len = sizeof(struct sockaddr_in6);
2583 				if ((inp->inp_vflag & INP_IPV6) && !IN6_IS_ADDR_UNSPECIFIED(&local_endpoint.sin6.sin6_addr)) {
2584 					local_address_is_valid = true;
2585 					fd_cb->local_endpoint = local_endpoint;
2586 					inp->in6p_laddr = sa6_any.sin6_addr;
2587 				} else {
2588 					fd_cb->local_endpoint.sin6.sin6_port = local_endpoint.sin6.sin6_port;
2589 				}
2590 			}
2591 		}
2592 
2593 		flow_divert_scope(fd_cb, out_if_index, !local_address_is_valid);
2594 		flow_divert_set_local_endpoint(fd_cb, SA(&fd_cb->local_endpoint));
2595 
2596 		if (flow_divert_is_sockaddr_valid(SA(&remote_endpoint)) && SOCK_TYPE(so) == SOCK_STREAM) {
2597 			if (remote_endpoint.sa.sa_family == AF_INET) {
2598 				remote_endpoint.sa.sa_len = sizeof(struct sockaddr_in);
2599 			} else if (remote_endpoint.sa.sa_family == AF_INET6) {
2600 				remote_endpoint.sa.sa_len = sizeof(struct sockaddr_in6);
2601 			}
2602 			flow_divert_set_remote_endpoint(fd_cb, SA(&remote_endpoint));
2603 		}
2604 
2605 		if (app_data_length > 0) {
2606 			uint8_t * app_data = NULL;
2607 			app_data = kalloc_data(app_data_length, Z_WAITOK);
2608 			if (app_data != NULL) {
2609 				error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_APP_DATA, app_data_length, app_data, NULL);
2610 				if (error == 0) {
2611 					FDLOG(LOG_INFO, fd_cb, "Got %u bytes of app data from the connect result", app_data_length);
2612 					if (fd_cb->app_data != NULL) {
2613 						kfree_data_sized_by(fd_cb->app_data, fd_cb->app_data_length);
2614 					}
2615 					fd_cb->app_data = app_data;
2616 					fd_cb->app_data_length = app_data_length;
2617 				} else {
2618 					FDLOG(LOG_ERR, fd_cb, "Failed to copy %u bytes of application data from the connect result packet", app_data_length);
2619 					kfree_data(app_data, app_data_length);
2620 				}
2621 			} else {
2622 				FDLOG(LOG_ERR, fd_cb, "Failed to allocate a buffer of size %u to hold the application data from the connect result", app_data_length);
2623 			}
2624 		}
2625 
2626 		if (error) {
2627 			goto set_socket_state;
2628 		}
2629 
2630 		if (fd_cb->group == NULL) {
2631 			error = EINVAL;
2632 			goto set_socket_state;
2633 		}
2634 
2635 		ctl_unit = ntohl(ctl_unit);
2636 		if (ctl_unit > 0) {
2637 			int insert_error = 0;
2638 			struct flow_divert_group *grp = NULL;
2639 
2640 			if (ctl_unit >= GROUP_COUNT_MAX) {
2641 				FDLOG(LOG_ERR, fd_cb, "Connect result contains an invalid control unit: %u", ctl_unit);
2642 				error = EINVAL;
2643 				goto set_socket_state;
2644 			}
2645 
2646 			grp = flow_divert_group_lookup(ctl_unit, fd_cb);
2647 			if (grp == NULL) {
2648 				error = ECONNRESET;
2649 				goto set_socket_state;
2650 			}
2651 
2652 			flow_divert_pcb_remove(fd_cb);
2653 			insert_error = flow_divert_pcb_insert(fd_cb, grp);
2654 			FDGRP_RELEASE(grp);
2655 
2656 			if (insert_error != 0) {
2657 				error = ECONNRESET;
2658 				goto set_socket_state;
2659 			}
2660 		}
2661 
2662 		fd_cb->send_window = ntohl(send_window);
2663 
2664 set_socket_state:
2665 		if (!connect_error && !error) {
2666 			FDLOG0(LOG_INFO, fd_cb, "sending connect result");
2667 			error = flow_divert_send_connect_result(fd_cb);
2668 		}
2669 
2670 		if (connect_error || error) {
2671 			if (connect_error && fd_cb->control_group_unit != fd_cb->policy_control_unit) {
2672 				error = flow_divert_try_next_group(fd_cb);
2673 				if (error && fd_cb->policy_control_unit == 0) {
2674 					flow_divert_disable(fd_cb);
2675 					goto done;
2676 				} else if (error == 0) {
2677 					goto done;
2678 				}
2679 			}
2680 
2681 			if (!connect_error) {
2682 				flow_divert_update_closed_state(fd_cb, SHUT_RDWR, false, true);
2683 				so->so_error = (uint16_t)error;
2684 				flow_divert_send_close_if_needed(fd_cb);
2685 			} else {
2686 				flow_divert_update_closed_state(fd_cb, SHUT_RDWR, true, true);
2687 				so->so_error = (uint16_t)connect_error;
2688 			}
2689 			flow_divert_disconnect_socket(so, !(fd_cb->flags & FLOW_DIVERT_IMPLICIT_CONNECT), false);
2690 		} else {
2691 #if NECP
2692 			/* Update NECP client with connected five-tuple */
2693 			if (!uuid_is_null(inp->necp_client_uuid)) {
2694 				socket_unlock(so, 0);
2695 				necp_client_assign_from_socket(so->last_pid, inp->necp_client_uuid, inp);
2696 				socket_lock(so, 0);
2697 				if (!(so->so_flags & SOF_FLOW_DIVERT)) {
2698 					/* The socket was closed while it was unlocked */
2699 					goto done;
2700 				}
2701 			}
2702 #endif /* NECP */
2703 
2704 			flow_divert_send_buffered_data(fd_cb, FALSE);
2705 			soisconnected(so);
2706 		}
2707 
2708 		/* We don't need the connect packet any more */
2709 		if (fd_cb->connect_packet != NULL) {
2710 			mbuf_freem(fd_cb->connect_packet);
2711 			fd_cb->connect_packet = NULL;
2712 		}
2713 
2714 		/* We don't need the original remote endpoint any more */
2715 		free_sockaddr(fd_cb->original_remote_endpoint);
2716 done:
2717 		socket_unlock(so, 1);
2718 	}
2719 	FDUNLOCK(fd_cb);
2720 }
2721 
2722 static void
flow_divert_handle_close(struct flow_divert_pcb * fd_cb,mbuf_ref_t packet,int offset)2723 flow_divert_handle_close(struct flow_divert_pcb *fd_cb, mbuf_ref_t packet, int offset)
2724 {
2725 	uint32_t        close_error                     = 0;
2726 	int                     error                   = 0;
2727 	int                     how                     = 0;
2728 
2729 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_ERROR_CODE, sizeof(close_error), &close_error, NULL);
2730 	if (error) {
2731 		FDLOG(LOG_ERR, fd_cb, "failed to get the close error: %d", error);
2732 		return;
2733 	}
2734 
2735 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_HOW, sizeof(how), &how, NULL);
2736 	if (error) {
2737 		FDLOG(LOG_ERR, fd_cb, "failed to get the close how flag: %d", error);
2738 		return;
2739 	}
2740 
2741 	how = ntohl(how);
2742 
2743 	FDLOG(LOG_INFO, fd_cb, "close received, how = %d", how);
2744 
2745 	FDLOCK(fd_cb);
2746 	if (fd_cb->so != NULL) {
2747 		bool is_connected = (SOCK_TYPE(fd_cb->so) == SOCK_STREAM || !(fd_cb->flags & FLOW_DIVERT_IMPLICIT_CONNECT));
2748 		socket_lock(fd_cb->so, 0);
2749 
2750 		if (!(fd_cb->so->so_flags & SOF_FLOW_DIVERT)) {
2751 			FDLOG0(LOG_NOTICE, fd_cb, "socket is not attached any more, ignoring close from provider");
2752 			goto done;
2753 		}
2754 
2755 		fd_cb->so->so_error = (uint16_t)ntohl(close_error);
2756 
2757 		flow_divert_update_closed_state(fd_cb, how, true, true);
2758 
2759 		/* Only do this for stream flows because "shutdown by peer" doesn't make sense for datagram flows */
2760 		how = flow_divert_tunnel_how_closed(fd_cb);
2761 		if (how == SHUT_RDWR) {
2762 			flow_divert_disconnect_socket(fd_cb->so, is_connected, true);
2763 		} else if (how == SHUT_RD && is_connected) {
2764 			socantrcvmore(fd_cb->so);
2765 		} else if (how == SHUT_WR && is_connected) {
2766 			socantsendmore(fd_cb->so);
2767 		}
2768 done:
2769 		socket_unlock(fd_cb->so, 0);
2770 	}
2771 	FDUNLOCK(fd_cb);
2772 }
2773 
2774 static mbuf_ref_t
flow_divert_create_control_mbuf(struct flow_divert_pcb * fd_cb)2775 flow_divert_create_control_mbuf(struct flow_divert_pcb *fd_cb)
2776 {
2777 	struct inpcb *inp = sotoinpcb(fd_cb->so);
2778 	bool need_recvdstaddr = false;
2779 	/* Socket flow tracking needs to see the local address */
2780 	need_recvdstaddr = SOFLOW_ENABLED(inp->inp_socket);
2781 	if ((inp->inp_vflag & INP_IPV4) &&
2782 	    fd_cb->local_endpoint.sa.sa_family == AF_INET &&
2783 	    ((inp->inp_flags & INP_RECVDSTADDR) || need_recvdstaddr)) {
2784 		return sbcreatecontrol((caddr_t)&(fd_cb->local_endpoint.sin.sin_addr), sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2785 	} else if ((inp->inp_vflag & INP_IPV6) &&
2786 	    fd_cb->local_endpoint.sa.sa_family == AF_INET6 &&
2787 	    ((inp->inp_flags & IN6P_PKTINFO) || need_recvdstaddr)) {
2788 		struct in6_pktinfo pi6;
2789 		memset(&pi6, 0, sizeof(pi6));
2790 		pi6.ipi6_addr = fd_cb->local_endpoint.sin6.sin6_addr;
2791 
2792 		return sbcreatecontrol((caddr_t)&pi6, sizeof(pi6), IPV6_PKTINFO, IPPROTO_IPV6);
2793 	}
2794 	return NULL;
2795 }
2796 
2797 static int
flow_divert_handle_data(struct flow_divert_pcb * fd_cb,mbuf_ref_t packet,size_t offset)2798 flow_divert_handle_data(struct flow_divert_pcb *fd_cb, mbuf_ref_t packet, size_t offset)
2799 {
2800 	int error = 0;
2801 
2802 	FDLOCK(fd_cb);
2803 	if (fd_cb->so != NULL) {
2804 		mbuf_ref_t data = NULL;
2805 		size_t  data_size;
2806 		struct sockaddr_storage remote_address;
2807 		boolean_t got_remote_sa = FALSE;
2808 		boolean_t appended = FALSE;
2809 		boolean_t append_success = FALSE;
2810 
2811 		socket_lock(fd_cb->so, 0);
2812 
2813 		if (!(fd_cb->so->so_flags & SOF_FLOW_DIVERT)) {
2814 			FDLOG0(LOG_NOTICE, fd_cb, "socket is not attached any more, ignoring inbound data");
2815 			goto done;
2816 		}
2817 
2818 		if (sbspace(&fd_cb->so->so_rcv) == 0) {
2819 			error = ENOBUFS;
2820 			fd_cb->flags |= FLOW_DIVERT_NOTIFY_ON_RECEIVED;
2821 			FDLOG0(LOG_INFO, fd_cb, "Receive buffer is full, will send read notification when app reads some data");
2822 			goto done;
2823 		}
2824 
2825 		if (SOCK_TYPE(fd_cb->so) == SOCK_DGRAM) {
2826 			uint32_t val_size = 0;
2827 
2828 			/* check if we got remote address with data */
2829 			memset(&remote_address, 0, sizeof(remote_address));
2830 			error = flow_divert_packet_get_tlv(packet, (int)offset, FLOW_DIVERT_TLV_REMOTE_ADDR, sizeof(remote_address), &remote_address, &val_size);
2831 			if (error || val_size > sizeof(remote_address)) {
2832 				FDLOG0(LOG_INFO, fd_cb, "No remote address provided");
2833 				error = 0;
2834 			} else {
2835 				if (remote_address.ss_len > sizeof(remote_address)) {
2836 					remote_address.ss_len = sizeof(remote_address);
2837 				}
2838 				/* validate the address */
2839 				if (flow_divert_is_sockaddr_valid((struct sockaddr *)&remote_address)) {
2840 					got_remote_sa = TRUE;
2841 				} else {
2842 					FDLOG0(LOG_INFO, fd_cb, "Remote address is invalid");
2843 				}
2844 				offset += (sizeof(uint8_t) + sizeof(uint32_t) + val_size);
2845 			}
2846 		}
2847 
2848 		data_size = (mbuf_pkthdr_len(packet) - offset);
2849 
2850 		if (fd_cb->so->so_state & SS_CANTRCVMORE) {
2851 			FDLOG(LOG_NOTICE, fd_cb, "app cannot receive any more data, dropping %lu bytes of data", data_size);
2852 			goto done;
2853 		}
2854 
2855 		if (SOCK_TYPE(fd_cb->so) != SOCK_STREAM && SOCK_TYPE(fd_cb->so) != SOCK_DGRAM) {
2856 			FDLOG(LOG_ERR, fd_cb, "socket has an unsupported type: %d", SOCK_TYPE(fd_cb->so));
2857 			goto done;
2858 		}
2859 
2860 		FDLOG(LOG_DEBUG, fd_cb, "received %lu bytes of data", data_size);
2861 
2862 		error = mbuf_split(packet, offset, MBUF_DONTWAIT, &data);
2863 		if (error || data == NULL) {
2864 			FDLOG(LOG_ERR, fd_cb, "mbuf_split failed: %d", error);
2865 			goto done;
2866 		}
2867 
2868 		if (SOCK_TYPE(fd_cb->so) == SOCK_STREAM) {
2869 			appended = (sbappendstream(&fd_cb->so->so_rcv, data) != 0);
2870 			append_success = TRUE;
2871 		} else {
2872 			struct sockaddr * __single append_sa = NULL;
2873 			mbuf_ref_t mctl;
2874 
2875 			if (got_remote_sa == TRUE) {
2876 				error = flow_divert_dup_addr(remote_address.ss_family, (struct sockaddr *)&remote_address, &append_sa);
2877 			} else {
2878 				if (SOCK_CHECK_DOM(fd_cb->so, AF_INET6)) {
2879 					error = in6_mapped_peeraddr(fd_cb->so, &append_sa);
2880 				} else {
2881 					error = in_getpeeraddr(fd_cb->so, &append_sa);
2882 				}
2883 			}
2884 			if (error) {
2885 				FDLOG0(LOG_ERR, fd_cb, "failed to dup the socket address.");
2886 			}
2887 
2888 			mctl = flow_divert_create_control_mbuf(fd_cb);
2889 			int append_error = 0;
2890 			appended = sbappendaddr(&fd_cb->so->so_rcv, append_sa, data, mctl, &append_error);
2891 			if (appended || append_error == 0) {
2892 				append_success = TRUE;
2893 			} else {
2894 				FDLOG(LOG_ERR, fd_cb, "failed to append %lu bytes of data: %d", data_size, append_error);
2895 			}
2896 
2897 			free_sockaddr(append_sa);
2898 		}
2899 
2900 		if (append_success) {
2901 			fd_cb->bytes_received += data_size;
2902 			flow_divert_add_data_statistics(fd_cb, data_size, FALSE);
2903 		}
2904 
2905 		if (appended) {
2906 			sorwakeup(fd_cb->so);
2907 		}
2908 done:
2909 		socket_unlock(fd_cb->so, 0);
2910 	}
2911 	FDUNLOCK(fd_cb);
2912 
2913 	return error;
2914 }
2915 
2916 static void
flow_divert_handle_read_notification(struct flow_divert_pcb * fd_cb,mbuf_ref_t packet,int offset)2917 flow_divert_handle_read_notification(struct flow_divert_pcb *fd_cb, mbuf_ref_t packet, int offset)
2918 {
2919 	uint32_t        read_count              = 0;
2920 	int             error                   = 0;
2921 
2922 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_READ_COUNT, sizeof(read_count), &read_count, NULL);
2923 	if (error) {
2924 		FDLOG(LOG_ERR, fd_cb, "failed to get the read count: %d", error);
2925 		return;
2926 	}
2927 
2928 	FDLOG(LOG_DEBUG, fd_cb, "received a read notification for %u bytes", ntohl(read_count));
2929 
2930 	FDLOCK(fd_cb);
2931 	if (fd_cb->so != NULL) {
2932 		socket_lock(fd_cb->so, 0);
2933 
2934 		if (!(fd_cb->so->so_flags & SOF_FLOW_DIVERT)) {
2935 			FDLOG0(LOG_NOTICE, fd_cb, "socket is not attached any more, ignoring read notification");
2936 			goto done;
2937 		}
2938 
2939 		fd_cb->send_window += ntohl(read_count);
2940 		flow_divert_send_buffered_data(fd_cb, FALSE);
2941 done:
2942 		socket_unlock(fd_cb->so, 0);
2943 	}
2944 	FDUNLOCK(fd_cb);
2945 }
2946 
2947 static void
flow_divert_handle_group_init(struct flow_divert_group * group,mbuf_ref_t packet,int offset)2948 flow_divert_handle_group_init(struct flow_divert_group *group, mbuf_ref_t packet, int offset)
2949 {
2950 	int error         = 0;
2951 	uint32_t key_size = 0;
2952 	int log_level     = 0;
2953 	uint32_t flags    = 0;
2954 	int32_t order     = FLOW_DIVERT_ORDER_LAST;
2955 
2956 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_TOKEN_KEY, 0, NULL, &key_size);
2957 	if (error) {
2958 		FDLOG(LOG_ERR, &nil_pcb, "failed to get the key size: %d", error);
2959 		return;
2960 	}
2961 
2962 	if (key_size == 0 || key_size > FLOW_DIVERT_MAX_KEY_SIZE) {
2963 		FDLOG(LOG_ERR, &nil_pcb, "Invalid key size: %u", key_size);
2964 		return;
2965 	}
2966 
2967 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_LOG_LEVEL, sizeof(log_level), &log_level, NULL);
2968 	if (!error) {
2969 		nil_pcb.log_level = (uint8_t)log_level;
2970 	}
2971 
2972 	lck_rw_lock_exclusive(&group->lck);
2973 
2974 	if (group->flags & FLOW_DIVERT_GROUP_FLAG_DEFUNCT) {
2975 		FDLOG(LOG_ERR, &nil_pcb, "Skipping (re)initialization of defunct group %u", group->ctl_unit);
2976 		lck_rw_done(&group->lck);
2977 		return;
2978 	}
2979 
2980 	if (group->token_key != NULL) {
2981 		kfree_data_sized_by(group->token_key, group->token_key_size);
2982 	}
2983 
2984 	group->token_key = kalloc_data(key_size, Z_WAITOK);
2985 	group->token_key_size = key_size;
2986 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_TOKEN_KEY, key_size, group->token_key, NULL);
2987 	if (error) {
2988 		FDLOG(LOG_ERR, &nil_pcb, "failed to get the token key: %d", error);
2989 		kfree_data_sized_by(group->token_key, group->token_key_size);
2990 		lck_rw_done(&group->lck);
2991 		return;
2992 	}
2993 
2994 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_FLAGS, sizeof(flags), &flags, NULL);
2995 	if (!error) {
2996 		group->flags = flags;
2997 	}
2998 
2999 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_ORDER, sizeof(order), &order, NULL);
3000 	if (!error) {
3001 		FDLOG(LOG_INFO, &nil_pcb, "group %u order is %u", group->ctl_unit, order);
3002 		group->order = order;
3003 	}
3004 
3005 	lck_rw_done(&group->lck);
3006 }
3007 
3008 static void
flow_divert_handle_properties_update(struct flow_divert_pcb * fd_cb,mbuf_ref_t packet,int offset)3009 flow_divert_handle_properties_update(struct flow_divert_pcb *fd_cb, mbuf_ref_t packet, int offset)
3010 {
3011 	int         error           = 0;
3012 	int         out_if_index    = 0;
3013 	uint32_t    app_data_length = 0;
3014 
3015 	FDLOG0(LOG_INFO, fd_cb, "received a properties update");
3016 
3017 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_OUT_IF_INDEX, sizeof(out_if_index), &out_if_index, NULL);
3018 	if (error) {
3019 		FDLOG0(LOG_INFO, fd_cb, "No output if index provided in properties update");
3020 	}
3021 
3022 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_APP_DATA, 0, NULL, &app_data_length);
3023 	if (error) {
3024 		FDLOG0(LOG_INFO, fd_cb, "No application data provided in properties update");
3025 	}
3026 
3027 	FDLOCK(fd_cb);
3028 	if (fd_cb->so != NULL) {
3029 		socket_lock(fd_cb->so, 0);
3030 
3031 		if (!(fd_cb->so->so_flags & SOF_FLOW_DIVERT)) {
3032 			FDLOG0(LOG_NOTICE, fd_cb, "socket is not attached any more, ignoring properties update");
3033 			goto done;
3034 		}
3035 
3036 		if (out_if_index > 0) {
3037 			flow_divert_scope(fd_cb, out_if_index, true);
3038 			flow_divert_set_local_endpoint(fd_cb, SA(&fd_cb->local_endpoint));
3039 		}
3040 
3041 		if (app_data_length > 0) {
3042 			uint8_t * app_data __indexable = NULL;
3043 			app_data = kalloc_data(app_data_length, Z_WAITOK);
3044 			if (app_data != NULL) {
3045 				error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_APP_DATA, app_data_length, app_data, NULL);
3046 				if (error == 0) {
3047 					if (fd_cb->app_data != NULL) {
3048 						kfree_data_sized_by(fd_cb->app_data, fd_cb->app_data_length);
3049 					}
3050 					fd_cb->app_data = app_data;
3051 					fd_cb->app_data_length = app_data_length;
3052 				} else {
3053 					FDLOG(LOG_ERR, fd_cb, "Failed to copy %u bytes of application data from the properties update packet", app_data_length);
3054 					kfree_data(app_data, app_data_length);
3055 				}
3056 			} else {
3057 				FDLOG(LOG_ERR, fd_cb, "Failed to allocate a buffer of size %u to hold the application data from the properties update", app_data_length);
3058 			}
3059 		}
3060 done:
3061 		socket_unlock(fd_cb->so, 0);
3062 	}
3063 	FDUNLOCK(fd_cb);
3064 }
3065 
3066 static void
flow_divert_handle_app_map_create(struct flow_divert_group * group,mbuf_ref_t packet,int offset)3067 flow_divert_handle_app_map_create(struct flow_divert_group *group, mbuf_ref_t packet, int offset)
3068 {
3069 	size_t                  bytes_mem_size      = 0;
3070 	size_t                  child_maps_mem_size = 0;
3071 	size_t                  nodes_mem_size      = 0;
3072 	size_t                  trie_memory_size    = 0;
3073 	int                     cursor              = 0;
3074 	int                     error               = 0;
3075 	struct flow_divert_trie new_trie;
3076 	int                     insert_error        = 0;
3077 	int                     prefix_count        = -1;
3078 	int                     signing_id_count    = 0;
3079 	size_t                  bytes_count         = 0;
3080 	size_t                  nodes_count         = 0;
3081 	size_t                  maps_count          = 0;
3082 
3083 	lck_rw_lock_exclusive(&group->lck);
3084 
3085 	/* Re-set the current trie */
3086 	if (group->signing_id_trie.memory != NULL) {
3087 		kfree_data_sized_by(group->signing_id_trie.memory, group->signing_id_trie.memory_size);
3088 	}
3089 	memset(&group->signing_id_trie, 0, sizeof(group->signing_id_trie));
3090 	group->signing_id_trie.root = NULL_TRIE_IDX;
3091 
3092 	memset(&new_trie, 0, sizeof(new_trie));
3093 
3094 	/* Get the number of shared prefixes in the new set of signing ID strings */
3095 	error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_PREFIX_COUNT, sizeof(prefix_count), &prefix_count, NULL);
3096 
3097 	if (prefix_count < 0 || error) {
3098 		FDLOG(LOG_ERR, &nil_pcb, "Invalid prefix count (%d) or an error occurred while reading the prefix count: %d", prefix_count, error);
3099 		lck_rw_done(&group->lck);
3100 		return;
3101 	}
3102 
3103 	/* Compute the number of signing IDs and the total amount of bytes needed to store them */
3104 	for (cursor = flow_divert_packet_find_tlv(packet, offset, FLOW_DIVERT_TLV_SIGNING_ID, &error, 0);
3105 	    cursor >= 0;
3106 	    cursor = flow_divert_packet_find_tlv(packet, cursor, FLOW_DIVERT_TLV_SIGNING_ID, &error, 1)) {
3107 		uint32_t sid_size = 0;
3108 		error = flow_divert_packet_get_tlv(packet, cursor, FLOW_DIVERT_TLV_SIGNING_ID, 0, NULL, &sid_size);
3109 		if (error || sid_size == 0) {
3110 			FDLOG(LOG_ERR, &nil_pcb, "Failed to get the length of the signing identifier at offset %d: %d", cursor, error);
3111 			signing_id_count = 0;
3112 			break;
3113 		}
3114 		if (os_add_overflow(bytes_count, sid_size, &bytes_count)) {
3115 			FDLOG0(LOG_ERR, &nil_pcb, "Overflow while incrementing number of bytes");
3116 			signing_id_count = 0;
3117 			break;
3118 		}
3119 		signing_id_count++;
3120 	}
3121 
3122 	if (signing_id_count == 0) {
3123 		lck_rw_done(&group->lck);
3124 		FDLOG0(LOG_NOTICE, &nil_pcb, "No signing identifiers");
3125 		return;
3126 	}
3127 
3128 	if (os_add3_overflow(prefix_count, signing_id_count, 1, &nodes_count)) { /* + 1 for the root node */
3129 		lck_rw_done(&group->lck);
3130 		FDLOG0(LOG_ERR, &nil_pcb, "Overflow while computing the number of nodes");
3131 		return;
3132 	}
3133 
3134 	if (os_add_overflow(prefix_count, 1, &maps_count)) { /* + 1 for the root node */
3135 		lck_rw_done(&group->lck);
3136 		FDLOG0(LOG_ERR, &nil_pcb, "Overflow while computing the number of maps");
3137 		return;
3138 	}
3139 
3140 	if (bytes_count > UINT16_MAX || nodes_count > UINT16_MAX || maps_count > UINT16_MAX) {
3141 		lck_rw_done(&group->lck);
3142 		FDLOG(LOG_NOTICE, &nil_pcb, "Invalid bytes count (%lu), nodes count (%lu) or maps count (%lu)", bytes_count, nodes_count, maps_count);
3143 		return;
3144 	}
3145 
3146 	FDLOG(LOG_INFO, &nil_pcb, "Nodes count = %lu, child maps count = %lu, bytes_count = %lu",
3147 	    nodes_count, maps_count, bytes_count);
3148 
3149 	if (os_mul_overflow(sizeof(*new_trie.nodes), (size_t)nodes_count, &nodes_mem_size) ||
3150 	    os_mul3_overflow(sizeof(*new_trie.child_maps), CHILD_MAP_SIZE, (size_t)maps_count, &child_maps_mem_size) ||
3151 	    os_mul_overflow(sizeof(*new_trie.bytes), (size_t)bytes_count, &bytes_mem_size) ||
3152 	    os_add3_overflow(nodes_mem_size, child_maps_mem_size, bytes_mem_size, &trie_memory_size)) {
3153 		FDLOG0(LOG_ERR, &nil_pcb, "Overflow while computing trie memory sizes");
3154 		lck_rw_done(&group->lck);
3155 		return;
3156 	}
3157 
3158 	if (trie_memory_size > FLOW_DIVERT_MAX_TRIE_MEMORY) {
3159 		FDLOG(LOG_ERR, &nil_pcb, "Trie memory size (%lu) is too big (maximum is %u)", trie_memory_size, FLOW_DIVERT_MAX_TRIE_MEMORY);
3160 		lck_rw_done(&group->lck);
3161 		return;
3162 	}
3163 
3164 	new_trie.memory = kalloc_data(trie_memory_size, Z_WAITOK);
3165 	new_trie.memory_size = trie_memory_size;
3166 	if (new_trie.memory == NULL) {
3167 		FDLOG(LOG_ERR, &nil_pcb, "Failed to allocate %lu bytes of memory for the signing ID trie",
3168 		    nodes_mem_size + child_maps_mem_size + bytes_mem_size);
3169 		lck_rw_done(&group->lck);
3170 		return;
3171 	}
3172 
3173 	/* Initialize the free lists */
3174 	new_trie.nodes = (struct flow_divert_trie_node *)new_trie.memory;
3175 	new_trie.nodes_count = (uint16_t)nodes_count;
3176 
3177 	new_trie.nodes_free_next = 0;
3178 	memset(new_trie.nodes, 0, nodes_mem_size);
3179 
3180 	new_trie.child_maps = (uint16_t *)(void *)((uint8_t *)new_trie.memory + nodes_mem_size);
3181 	new_trie.child_maps_count = (uint16_t)maps_count;
3182 	new_trie.child_maps_size = child_maps_mem_size;
3183 
3184 	new_trie.child_maps_free_next = 0;
3185 	memset(new_trie.child_maps, 0xff, child_maps_mem_size);
3186 
3187 	new_trie.bytes = (uint8_t *)(void *)((uint8_t *)new_trie.memory + nodes_mem_size + child_maps_mem_size);
3188 	new_trie.bytes_count = (uint16_t)bytes_count;
3189 
3190 	new_trie.bytes_free_next = 0;
3191 	memset(new_trie.bytes, 0, bytes_mem_size);
3192 
3193 	/* The root is an empty node */
3194 	new_trie.root = trie_node_alloc(&new_trie);
3195 
3196 	/* Add each signing ID to the trie */
3197 	for (cursor = flow_divert_packet_find_tlv(packet, offset, FLOW_DIVERT_TLV_SIGNING_ID, &error, 0);
3198 	    cursor >= 0;
3199 	    cursor = flow_divert_packet_find_tlv(packet, cursor, FLOW_DIVERT_TLV_SIGNING_ID, &error, 1)) {
3200 		uint32_t sid_size = 0;
3201 		error = flow_divert_packet_get_tlv(packet, cursor, FLOW_DIVERT_TLV_SIGNING_ID, 0, NULL, &sid_size);
3202 		if (error || sid_size == 0) {
3203 			FDLOG(LOG_ERR, &nil_pcb, "Failed to get the length of the signing identifier at offset %d while building: %d", cursor, error);
3204 			insert_error = EINVAL;
3205 			break;
3206 		}
3207 		if (sid_size <= UINT16_MAX && new_trie.bytes_free_next + (uint16_t)sid_size <= new_trie.bytes_count) {
3208 			uint16_t new_node_idx;
3209 			error = flow_divert_packet_get_tlv(packet, cursor, FLOW_DIVERT_TLV_SIGNING_ID, sid_size, &TRIE_BYTE(&new_trie, new_trie.bytes_free_next), NULL);
3210 			if (error) {
3211 				FDLOG(LOG_ERR, &nil_pcb, "Failed to read the signing identifier at offset %d: %d", cursor, error);
3212 				insert_error = EINVAL;
3213 				break;
3214 			}
3215 			new_node_idx = flow_divert_trie_insert(&new_trie, new_trie.bytes_free_next, sid_size);
3216 			if (new_node_idx == NULL_TRIE_IDX) {
3217 				insert_error = EINVAL;
3218 				break;
3219 			}
3220 		} else {
3221 			FDLOG0(LOG_ERR, &nil_pcb, "No place to put signing ID for insertion");
3222 			insert_error = ENOBUFS;
3223 			break;
3224 		}
3225 	}
3226 
3227 	if (!insert_error) {
3228 		group->signing_id_trie = new_trie;
3229 	} else {
3230 		kfree_data_sized_by(new_trie.memory, new_trie.memory_size);
3231 	}
3232 
3233 	lck_rw_done(&group->lck);
3234 }
3235 
3236 static void
flow_divert_handle_flow_states_request(struct flow_divert_group * group)3237 flow_divert_handle_flow_states_request(struct flow_divert_group *group)
3238 {
3239 	struct flow_divert_pcb *fd_cb;
3240 	mbuf_ref_t packet = NULL;
3241 	SLIST_HEAD(, flow_divert_pcb) tmp_list;
3242 	int error = 0;
3243 	uint32_t ctl_unit = 0;
3244 
3245 	SLIST_INIT(&tmp_list);
3246 
3247 	error = flow_divert_packet_init(&nil_pcb, FLOW_DIVERT_PKT_FLOW_STATES, &packet);
3248 	if (error || packet == NULL) {
3249 		FDLOG(LOG_ERR, &nil_pcb, "flow_divert_packet_init failed: %d, cannot send flow states", error);
3250 		return;
3251 	}
3252 
3253 	lck_rw_lock_shared(&group->lck);
3254 
3255 	if (!MBUFQ_EMPTY(&group->send_queue)) {
3256 		FDLOG0(LOG_WARNING, &nil_pcb, "flow_divert_handle_flow_states_request: group send queue is not empty");
3257 	}
3258 
3259 	ctl_unit = group->ctl_unit;
3260 
3261 	RB_FOREACH(fd_cb, fd_pcb_tree, &group->pcb_tree) {
3262 		FDRETAIN(fd_cb);
3263 		SLIST_INSERT_HEAD(&tmp_list, fd_cb, tmp_list_entry);
3264 	}
3265 
3266 	lck_rw_done(&group->lck);
3267 
3268 	SLIST_FOREACH(fd_cb, &tmp_list, tmp_list_entry) {
3269 		FDLOCK(fd_cb);
3270 		if (fd_cb->so != NULL) {
3271 			struct flow_divert_flow_state state = {};
3272 			socket_lock(fd_cb->so, 0);
3273 
3274 			state.conn_id = fd_cb->hash;
3275 			state.bytes_written_by_app = fd_cb->bytes_written_by_app;
3276 			state.bytes_sent = fd_cb->bytes_sent;
3277 			state.bytes_received = fd_cb->bytes_received;
3278 			state.send_window = fd_cb->send_window;
3279 			state.send_buffer_bytes = fd_cb->so->so_snd.sb_cc;
3280 
3281 			error = flow_divert_packet_append_tlv(packet, FLOW_DIVERT_TLV_FLOW_STATE, sizeof(state), &state);
3282 			if (error) {
3283 				FDLOG(LOG_ERR, fd_cb, "Failed to add a flow state: %d", error);
3284 			}
3285 
3286 			socket_unlock(fd_cb->so, 0);
3287 		}
3288 		FDUNLOCK(fd_cb);
3289 		FDRELEASE(fd_cb);
3290 	}
3291 
3292 	error = ctl_enqueuembuf(g_flow_divert_kctl_ref, ctl_unit, packet, CTL_DATA_EOR);
3293 	if (error) {
3294 		FDLOG(LOG_NOTICE, &nil_pcb, "flow_divert_handle_flow_states_request: ctl_enqueuembuf returned an error: %d", error);
3295 		mbuf_freem(packet);
3296 	}
3297 }
3298 
3299 static int
flow_divert_input(mbuf_ref_t packet,struct flow_divert_group * group)3300 flow_divert_input(mbuf_ref_t packet, struct flow_divert_group *group)
3301 {
3302 	struct flow_divert_packet_header    hdr;
3303 	int                                 error  = 0;
3304 	struct flow_divert_pcb              *fd_cb;
3305 
3306 	if (mbuf_pkthdr_len(packet) < sizeof(hdr)) {
3307 		FDLOG(LOG_ERR, &nil_pcb, "got a bad packet, length (%lu) < sizeof hdr (%lu)", mbuf_pkthdr_len(packet), sizeof(hdr));
3308 		error = EINVAL;
3309 		goto done;
3310 	}
3311 
3312 	error = mbuf_copydata(packet, 0, sizeof(hdr), &hdr);
3313 	if (error) {
3314 		FDLOG(LOG_ERR, &nil_pcb, "mbuf_copydata failed for the header: %d", error);
3315 		error = ENOBUFS;
3316 		goto done;
3317 	}
3318 
3319 	hdr.conn_id = ntohl(hdr.conn_id);
3320 
3321 	if (hdr.conn_id == 0) {
3322 		switch (hdr.packet_type) {
3323 		case FLOW_DIVERT_PKT_GROUP_INIT:
3324 			flow_divert_handle_group_init(group, packet, sizeof(hdr));
3325 			break;
3326 		case FLOW_DIVERT_PKT_APP_MAP_CREATE:
3327 			flow_divert_handle_app_map_create(group, packet, sizeof(hdr));
3328 			break;
3329 		case FLOW_DIVERT_PKT_FLOW_STATES_REQUEST:
3330 			flow_divert_handle_flow_states_request(group);
3331 			break;
3332 		default:
3333 			FDLOG(LOG_WARNING, &nil_pcb, "got an unknown message type: %d", hdr.packet_type);
3334 			break;
3335 		}
3336 		goto done;
3337 	}
3338 
3339 	fd_cb = flow_divert_pcb_lookup(hdr.conn_id, group);             /* This retains the PCB */
3340 	if (fd_cb == NULL) {
3341 		if (hdr.packet_type != FLOW_DIVERT_PKT_CLOSE && hdr.packet_type != FLOW_DIVERT_PKT_READ_NOTIFY) {
3342 			FDLOG(LOG_NOTICE, &nil_pcb, "got a %s message from group %d for an unknown pcb: %u", flow_divert_packet_type2str(hdr.packet_type), group->ctl_unit, hdr.conn_id);
3343 		}
3344 		goto done;
3345 	}
3346 
3347 	switch (hdr.packet_type) {
3348 	case FLOW_DIVERT_PKT_CONNECT_RESULT:
3349 		flow_divert_handle_connect_result(fd_cb, packet, sizeof(hdr));
3350 		break;
3351 	case FLOW_DIVERT_PKT_CLOSE:
3352 		flow_divert_handle_close(fd_cb, packet, sizeof(hdr));
3353 		break;
3354 	case FLOW_DIVERT_PKT_DATA:
3355 		error = flow_divert_handle_data(fd_cb, packet, sizeof(hdr));
3356 		break;
3357 	case FLOW_DIVERT_PKT_READ_NOTIFY:
3358 		flow_divert_handle_read_notification(fd_cb, packet, sizeof(hdr));
3359 		break;
3360 	case FLOW_DIVERT_PKT_PROPERTIES_UPDATE:
3361 		flow_divert_handle_properties_update(fd_cb, packet, sizeof(hdr));
3362 		break;
3363 	default:
3364 		FDLOG(LOG_WARNING, fd_cb, "got an unknown message type: %d", hdr.packet_type);
3365 		break;
3366 	}
3367 
3368 	FDRELEASE(fd_cb);
3369 
3370 done:
3371 	mbuf_freem(packet);
3372 	return error;
3373 }
3374 
3375 static void
flow_divert_close_all(struct flow_divert_group * group)3376 flow_divert_close_all(struct flow_divert_group *group)
3377 {
3378 	struct flow_divert_pcb                  *fd_cb;
3379 	SLIST_HEAD(, flow_divert_pcb)   tmp_list;
3380 
3381 	SLIST_INIT(&tmp_list);
3382 
3383 	lck_rw_lock_exclusive(&group->lck);
3384 
3385 	MBUFQ_DRAIN(&group->send_queue);
3386 
3387 	RB_FOREACH(fd_cb, fd_pcb_tree, &group->pcb_tree) {
3388 		FDRETAIN(fd_cb);
3389 		SLIST_INSERT_HEAD(&tmp_list, fd_cb, tmp_list_entry);
3390 	}
3391 
3392 	group->flags |= FLOW_DIVERT_GROUP_FLAG_DEFUNCT;
3393 
3394 	lck_rw_done(&group->lck);
3395 
3396 	while (!SLIST_EMPTY(&tmp_list)) {
3397 		fd_cb = SLIST_FIRST(&tmp_list);
3398 		FDLOCK(fd_cb);
3399 		SLIST_REMOVE_HEAD(&tmp_list, tmp_list_entry);
3400 		if (fd_cb->so != NULL) {
3401 			socket_lock(fd_cb->so, 0);
3402 			flow_divert_pcb_remove(fd_cb);
3403 			flow_divert_update_closed_state(fd_cb, SHUT_RDWR, true, true);
3404 			fd_cb->so->so_error = ECONNABORTED;
3405 			flow_divert_disconnect_socket(fd_cb->so, !(fd_cb->flags & FLOW_DIVERT_IMPLICIT_CONNECT), false);
3406 			socket_unlock(fd_cb->so, 0);
3407 		}
3408 		FDUNLOCK(fd_cb);
3409 		FDRELEASE(fd_cb);
3410 	}
3411 }
3412 
3413 void
flow_divert_detach(struct socket * so)3414 flow_divert_detach(struct socket *so)
3415 {
3416 	struct flow_divert_pcb  *fd_cb          = so->so_fd_pcb;
3417 
3418 	if (!SO_IS_DIVERTED(so)) {
3419 		return;
3420 	}
3421 
3422 	so->so_flags &= ~SOF_FLOW_DIVERT;
3423 	so->so_fd_pcb = NULL;
3424 
3425 	FDLOG(LOG_INFO, fd_cb, "Detaching, ref count = %d", fd_cb->ref_count);
3426 
3427 	if (fd_cb->group != NULL) {
3428 		/* Last-ditch effort to send any buffered data */
3429 		flow_divert_send_buffered_data(fd_cb, TRUE);
3430 
3431 		flow_divert_update_closed_state(fd_cb, SHUT_RDWR, false, true);
3432 		flow_divert_send_close_if_needed(fd_cb);
3433 		/* Remove from the group */
3434 		flow_divert_pcb_remove(fd_cb);
3435 	}
3436 
3437 	socket_unlock(so, 0);
3438 	FDLOCK(fd_cb);
3439 	fd_cb->so = NULL;
3440 	FDUNLOCK(fd_cb);
3441 	socket_lock(so, 0);
3442 
3443 	FDRELEASE(fd_cb);       /* Release the socket's reference */
3444 }
3445 
3446 static int
flow_divert_close(struct socket * so)3447 flow_divert_close(struct socket *so)
3448 {
3449 	struct flow_divert_pcb  *fd_cb          = so->so_fd_pcb;
3450 
3451 	if (!SO_IS_DIVERTED(so)) {
3452 		return EINVAL;
3453 	}
3454 
3455 	FDLOG0(LOG_INFO, fd_cb, "Closing");
3456 
3457 	if (SOCK_TYPE(so) == SOCK_STREAM) {
3458 		soisdisconnecting(so);
3459 		sbflush(&so->so_rcv);
3460 	}
3461 
3462 	flow_divert_send_buffered_data(fd_cb, TRUE);
3463 	flow_divert_update_closed_state(fd_cb, SHUT_RDWR, false, true);
3464 	flow_divert_send_close_if_needed(fd_cb);
3465 
3466 	/* Remove from the group */
3467 	flow_divert_pcb_remove(fd_cb);
3468 
3469 	return 0;
3470 }
3471 
3472 static int
flow_divert_disconnectx(struct socket * so,sae_associd_t aid,sae_connid_t cid __unused)3473 flow_divert_disconnectx(struct socket *so, sae_associd_t aid,
3474     sae_connid_t cid __unused)
3475 {
3476 	if (aid != SAE_ASSOCID_ANY && aid != SAE_ASSOCID_ALL) {
3477 		return EINVAL;
3478 	}
3479 
3480 	return flow_divert_close(so);
3481 }
3482 
3483 static int
flow_divert_shutdown(struct socket * so)3484 flow_divert_shutdown(struct socket *so)
3485 {
3486 	struct flow_divert_pcb  *fd_cb          = so->so_fd_pcb;
3487 
3488 	if (!SO_IS_DIVERTED(so)) {
3489 		return EINVAL;
3490 	}
3491 
3492 	FDLOG0(LOG_INFO, fd_cb, "Can't send more");
3493 
3494 	socantsendmore(so);
3495 
3496 	flow_divert_update_closed_state(fd_cb, SHUT_WR, false, true);
3497 	flow_divert_send_close_if_needed(fd_cb);
3498 
3499 	return 0;
3500 }
3501 
3502 static int
flow_divert_rcvd(struct socket * so,int flags __unused)3503 flow_divert_rcvd(struct socket *so, int flags __unused)
3504 {
3505 	struct flow_divert_pcb  *fd_cb = so->so_fd_pcb;
3506 	int space = 0;
3507 
3508 	if (!SO_IS_DIVERTED(so)) {
3509 		return EINVAL;
3510 	}
3511 
3512 	space = sbspace(&so->so_rcv);
3513 	FDLOG(LOG_DEBUG, fd_cb, "app read bytes, space = %d", space);
3514 	if ((fd_cb->flags & FLOW_DIVERT_NOTIFY_ON_RECEIVED) &&
3515 	    (space > 0) &&
3516 	    flow_divert_send_read_notification(fd_cb) == 0) {
3517 		FDLOG0(LOG_INFO, fd_cb, "Sent a read notification");
3518 		fd_cb->flags &= ~FLOW_DIVERT_NOTIFY_ON_RECEIVED;
3519 	}
3520 
3521 	return 0;
3522 }
3523 
3524 static int
flow_divert_append_target_endpoint_tlv(mbuf_ref_t connect_packet,struct sockaddr * toaddr)3525 flow_divert_append_target_endpoint_tlv(mbuf_ref_t connect_packet, struct sockaddr *toaddr)
3526 {
3527 	int error = 0;
3528 	int port  = 0;
3529 
3530 	if (!flow_divert_is_sockaddr_valid(toaddr)) {
3531 		FDLOG(LOG_ERR, &nil_pcb, "Invalid target address, family = %u, length = %u", toaddr->sa_family, toaddr->sa_len);
3532 		error = EINVAL;
3533 		goto done;
3534 	}
3535 
3536 	error = flow_divert_packet_append_tlv(connect_packet, FLOW_DIVERT_TLV_TARGET_ADDRESS, toaddr->sa_len, SA_BYTES(toaddr));
3537 	if (error) {
3538 		goto done;
3539 	}
3540 
3541 	if (toaddr->sa_family == AF_INET) {
3542 		port = ntohs((satosin(toaddr))->sin_port);
3543 	} else {
3544 		port = ntohs((satosin6(toaddr))->sin6_port);
3545 	}
3546 
3547 	error = flow_divert_packet_append_tlv(connect_packet, FLOW_DIVERT_TLV_TARGET_PORT, sizeof(port), &port);
3548 	if (error) {
3549 		goto done;
3550 	}
3551 
3552 done:
3553 	return error;
3554 }
3555 
3556 struct sockaddr *
flow_divert_get_buffered_target_address(mbuf_ref_t buffer)3557 flow_divert_get_buffered_target_address(mbuf_ref_t buffer)
3558 {
3559 	if (buffer != NULL && buffer->m_type == MT_SONAME) {
3560 		struct sockaddr *toaddr = mtod(buffer, struct sockaddr *);
3561 		if (toaddr != NULL && flow_divert_is_sockaddr_valid(toaddr)) {
3562 			return toaddr;
3563 		}
3564 	}
3565 	return NULL;
3566 }
3567 
3568 static boolean_t
flow_divert_is_sockaddr_valid(struct sockaddr * addr)3569 flow_divert_is_sockaddr_valid(struct sockaddr *addr)
3570 {
3571 	switch (addr->sa_family) {
3572 	case AF_INET:
3573 		if (addr->sa_len < sizeof(struct sockaddr_in)) {
3574 			return FALSE;
3575 		}
3576 		break;
3577 	case AF_INET6:
3578 		if (addr->sa_len < sizeof(struct sockaddr_in6)) {
3579 			return FALSE;
3580 		}
3581 		break;
3582 	default:
3583 		return FALSE;
3584 	}
3585 	return TRUE;
3586 }
3587 
3588 static errno_t
flow_divert_dup_addr(sa_family_t family,struct sockaddr * addr,struct sockaddr ** dup)3589 flow_divert_dup_addr(sa_family_t family, struct sockaddr *addr,
3590     struct sockaddr **dup)
3591 {
3592 	int                                             error           = 0;
3593 	struct sockaddr                 *result;
3594 	struct sockaddr_storage ss;
3595 
3596 	if (addr != NULL) {
3597 		result = addr;
3598 	} else {
3599 		memset(&ss, 0, sizeof(ss));
3600 		ss.ss_family = family;
3601 		if (ss.ss_family == AF_INET) {
3602 			ss.ss_len = sizeof(struct sockaddr_in);
3603 		} else if (ss.ss_family == AF_INET6) {
3604 			ss.ss_len = sizeof(struct sockaddr_in6);
3605 		} else {
3606 			error = EINVAL;
3607 		}
3608 		result = (struct sockaddr *)&ss;
3609 	}
3610 
3611 	if (!error) {
3612 		*dup = dup_sockaddr(result, 1);
3613 		if (*dup == NULL) {
3614 			error = ENOBUFS;
3615 		}
3616 	}
3617 
3618 	return error;
3619 }
3620 
3621 static void
flow_divert_disconnect_socket(struct socket * so,bool is_connected,bool delay_if_needed)3622 flow_divert_disconnect_socket(struct socket *so, bool is_connected, bool delay_if_needed)
3623 {
3624 	if (SOCK_TYPE(so) == SOCK_STREAM || is_connected) {
3625 		soisdisconnected(so);
3626 	}
3627 	if (SOCK_TYPE(so) == SOCK_DGRAM) {
3628 		struct inpcb *inp = sotoinpcb(so);
3629 		if (inp != NULL && !(so->so_flags & SOF_PCBCLEARING)) {
3630 			/*
3631 			 * Let NetworkStatistics know this PCB is going away
3632 			 * before we detach it.
3633 			 */
3634 			if (nstat_collect && (SOCK_PROTO(so) == IPPROTO_TCP || SOCK_PROTO(so) == IPPROTO_UDP)) {
3635 				nstat_pcb_detach(inp);
3636 			}
3637 
3638 			if (SOCK_DOM(so) == PF_INET6) {
3639 				ROUTE_RELEASE(&inp->in6p_route);
3640 			} else {
3641 				ROUTE_RELEASE(&inp->inp_route);
3642 			}
3643 			if (delay_if_needed) {
3644 				(void) cfil_sock_is_dead(so);
3645 			} else {
3646 				inp->inp_state = INPCB_STATE_DEAD;
3647 				inpcb_gc_sched(inp->inp_pcbinfo, INPCB_TIMER_FAST);
3648 			}
3649 			/* makes sure we're not called twice from so_close */
3650 			so->so_flags |= SOF_PCBCLEARING;
3651 		}
3652 	}
3653 }
3654 
3655 static errno_t
flow_divert_ctloutput(struct socket * so,struct sockopt * sopt)3656 flow_divert_ctloutput(struct socket *so, struct sockopt *sopt)
3657 {
3658 	struct flow_divert_pcb  *fd_cb  = so->so_fd_pcb;
3659 
3660 	if (!SO_IS_DIVERTED(so)) {
3661 		return EINVAL;
3662 	}
3663 
3664 	if (sopt->sopt_name == SO_TRAFFIC_CLASS) {
3665 		if (sopt->sopt_dir == SOPT_SET && fd_cb->flags & FLOW_DIVERT_CONNECT_STARTED) {
3666 			flow_divert_send_traffic_class_update(fd_cb, so->so_traffic_class);
3667 		}
3668 	}
3669 
3670 	if (SOCK_DOM(so) == PF_INET) {
3671 		return g_tcp_protosw->pr_ctloutput(so, sopt);
3672 	} else if (SOCK_DOM(so) == PF_INET6) {
3673 		return g_tcp6_protosw->pr_ctloutput(so, sopt);
3674 	}
3675 	return 0;
3676 }
3677 
3678 static errno_t
flow_divert_connect_out_internal(struct socket * so,struct sockaddr * to,proc_t p,bool implicit)3679 flow_divert_connect_out_internal(struct socket *so, struct sockaddr *to, proc_t p, bool implicit)
3680 {
3681 	struct flow_divert_pcb  *fd_cb          = so->so_fd_pcb;
3682 	int                     error           = 0;
3683 	struct inpcb            *inp            = sotoinpcb(so);
3684 	struct sockaddr_in      *sinp;
3685 	mbuf_ref_t              connect_packet  = NULL;
3686 	int                     do_send         = 1;
3687 
3688 	if (!SO_IS_DIVERTED(so)) {
3689 		return EINVAL;
3690 	}
3691 
3692 	if (fd_cb->group == NULL) {
3693 		error = ENETUNREACH;
3694 		goto done;
3695 	}
3696 
3697 	if (inp == NULL) {
3698 		error = EINVAL;
3699 		goto done;
3700 	} else if (inp->inp_state == INPCB_STATE_DEAD) {
3701 		if (so->so_error) {
3702 			error = so->so_error;
3703 			so->so_error = 0;
3704 		} else {
3705 			error = EINVAL;
3706 		}
3707 		goto done;
3708 	}
3709 
3710 	if (fd_cb->flags & FLOW_DIVERT_CONNECT_STARTED) {
3711 		error = EALREADY;
3712 		goto done;
3713 	}
3714 
3715 	FDLOG0(LOG_INFO, fd_cb, "Connecting");
3716 
3717 	if (fd_cb->connect_packet == NULL) {
3718 		struct sockaddr_in sin = {};
3719 		struct ifnet * __single ifp = NULL;
3720 
3721 		if (to == NULL) {
3722 			FDLOG0(LOG_ERR, fd_cb, "No destination address available when creating connect packet");
3723 			error = EINVAL;
3724 			goto done;
3725 		}
3726 
3727 		if (!flow_divert_is_sockaddr_valid(to)) {
3728 			FDLOG0(LOG_ERR, fd_cb, "Destination address is not valid when creating connect packet");
3729 			error = EINVAL;
3730 			goto done;
3731 		}
3732 
3733 		fd_cb->original_remote_endpoint = dup_sockaddr(to, 0);
3734 		if (fd_cb->original_remote_endpoint == NULL) {
3735 			FDLOG0(LOG_ERR, fd_cb, "Failed to dup the remote endpoint");
3736 			error = ENOMEM;
3737 			goto done;
3738 		}
3739 		fd_cb->original_vflag = inp->inp_vflag;
3740 		fd_cb->original_last_outifp = inp->inp_last_outifp;
3741 		fd_cb->original_last_outifp6 = inp->in6p_last_outifp;
3742 
3743 		sinp = (struct sockaddr_in *)(void *)to;
3744 		if (sinp->sin_family == AF_INET && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
3745 			error = EAFNOSUPPORT;
3746 			goto done;
3747 		}
3748 
3749 		if (to->sa_family == AF_INET6 && !(inp->inp_flags & IN6P_IPV6_V6ONLY)) {
3750 			struct sockaddr_in6 sin6 = {};
3751 			sin6.sin6_family = AF_INET6;
3752 			sin6.sin6_len = sizeof(struct sockaddr_in6);
3753 			sin6.sin6_port = satosin6(to)->sin6_port;
3754 			sin6.sin6_addr = satosin6(to)->sin6_addr;
3755 			if (IN6_IS_ADDR_V4MAPPED(&(sin6.sin6_addr))) {
3756 				in6_sin6_2_sin(&sin, &sin6);
3757 				to = (struct sockaddr *)&sin;
3758 			}
3759 		}
3760 
3761 		if (to->sa_family == AF_INET6) {
3762 			struct sockaddr_in6 *to6 = satosin6(to);
3763 
3764 			inp->inp_vflag &= ~INP_IPV4;
3765 			inp->inp_vflag |= INP_IPV6;
3766 			fd_cb->local_endpoint.sin6.sin6_len = sizeof(struct sockaddr_in6);
3767 			fd_cb->local_endpoint.sin6.sin6_family = AF_INET6;
3768 			fd_cb->local_endpoint.sin6.sin6_port = inp->inp_lport;
3769 			error = in6_pcbladdr(inp, to, &(fd_cb->local_endpoint.sin6.sin6_addr), &ifp);
3770 			if (error) {
3771 				FDLOG(LOG_WARNING, fd_cb, "failed to get a local IPv6 address: %d", error);
3772 				if (!(fd_cb->flags & FLOW_DIVERT_FLOW_IS_TRANSPARENT) || IN6_IS_ADDR_UNSPECIFIED(&(satosin6(to)->sin6_addr))) {
3773 					error = 0;
3774 				} else {
3775 					goto done;
3776 				}
3777 			}
3778 			if (ifp != NULL) {
3779 				inp->in6p_last_outifp = ifp;
3780 				ifnet_release(ifp);
3781 			}
3782 
3783 			if (IN6_IS_SCOPE_EMBED(&(fd_cb->local_endpoint.sin6.sin6_addr)) &&
3784 			    in6_embedded_scope &&
3785 			    fd_cb->local_endpoint.sin6.sin6_addr.s6_addr16[1] != 0) {
3786 				fd_cb->local_endpoint.sin6.sin6_scope_id = ntohs(fd_cb->local_endpoint.sin6.sin6_addr.s6_addr16[1]);
3787 				fd_cb->local_endpoint.sin6.sin6_addr.s6_addr16[1] = 0;
3788 			}
3789 
3790 			if (IN6_IS_SCOPE_EMBED(&(to6->sin6_addr)) &&
3791 			    in6_embedded_scope &&
3792 			    to6->sin6_addr.s6_addr16[1] != 0) {
3793 				to6->sin6_scope_id = ntohs(to6->sin6_addr.s6_addr16[1]);
3794 				to6->sin6_addr.s6_addr16[1] = 0;
3795 			}
3796 		} else if (to->sa_family == AF_INET) {
3797 			inp->inp_vflag |= INP_IPV4;
3798 			inp->inp_vflag &= ~INP_IPV6;
3799 			fd_cb->local_endpoint.sin.sin_len = sizeof(struct sockaddr_in);
3800 			fd_cb->local_endpoint.sin.sin_family = AF_INET;
3801 			fd_cb->local_endpoint.sin.sin_port = inp->inp_lport;
3802 			error = in_pcbladdr(inp, to, &(fd_cb->local_endpoint.sin.sin_addr), IFSCOPE_NONE, &ifp, 0);
3803 			if (error) {
3804 				FDLOG(LOG_WARNING, fd_cb, "failed to get a local IPv4 address: %d", error);
3805 				if (!(fd_cb->flags & FLOW_DIVERT_FLOW_IS_TRANSPARENT) || satosin(to)->sin_addr.s_addr == INADDR_ANY) {
3806 					error = 0;
3807 				} else {
3808 					goto done;
3809 				}
3810 			}
3811 			if (ifp != NULL) {
3812 				inp->inp_last_outifp = ifp;
3813 				ifnet_release(ifp);
3814 			}
3815 		} else {
3816 			FDLOG(LOG_WARNING, fd_cb, "target address has an unsupported family: %d", to->sa_family);
3817 		}
3818 
3819 		error = flow_divert_check_no_cellular(fd_cb) ||
3820 		    flow_divert_check_no_expensive(fd_cb) ||
3821 		    flow_divert_check_no_constrained(fd_cb);
3822 		if (error) {
3823 			goto done;
3824 		}
3825 
3826 		if (SOCK_TYPE(so) == SOCK_STREAM || /* TCP or */
3827 		    !implicit || /* connect() was called or */
3828 		    ((inp->inp_vflag & INP_IPV6) && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) || /* local address is not un-specified */
3829 		    ((inp->inp_vflag & INP_IPV4) && inp->inp_laddr.s_addr != INADDR_ANY)) {
3830 			fd_cb->flags |= FLOW_DIVERT_SHOULD_SET_LOCAL_ADDR;
3831 		}
3832 
3833 		error = flow_divert_create_connect_packet(fd_cb, to, so, p, &connect_packet);
3834 		if (error) {
3835 			goto done;
3836 		}
3837 
3838 		if (!implicit || SOCK_TYPE(so) == SOCK_STREAM) {
3839 			flow_divert_set_remote_endpoint(fd_cb, to);
3840 			flow_divert_set_local_endpoint(fd_cb, SA(&fd_cb->local_endpoint));
3841 		}
3842 
3843 		if (implicit) {
3844 			fd_cb->flags |= FLOW_DIVERT_IMPLICIT_CONNECT;
3845 		}
3846 
3847 		if (so->so_flags1 & SOF1_PRECONNECT_DATA) {
3848 			FDLOG0(LOG_INFO, fd_cb, "Delaying sending the connect packet until send or receive");
3849 			do_send = 0;
3850 		}
3851 
3852 		fd_cb->connect_packet = connect_packet;
3853 		connect_packet = NULL;
3854 	} else {
3855 		FDLOG0(LOG_INFO, fd_cb, "Sending saved connect packet");
3856 	}
3857 
3858 	if (do_send) {
3859 		error = flow_divert_send_connect_packet(fd_cb);
3860 		if (error) {
3861 			goto done;
3862 		}
3863 
3864 		fd_cb->flags |= FLOW_DIVERT_CONNECT_STARTED;
3865 	}
3866 
3867 	if (SOCK_TYPE(so) == SOCK_DGRAM && !(fd_cb->flags & FLOW_DIVERT_HAS_TOKEN)) {
3868 		soisconnected(so);
3869 	} else {
3870 		soisconnecting(so);
3871 	}
3872 
3873 done:
3874 	return error;
3875 }
3876 
3877 errno_t
flow_divert_connect_out(struct socket * so,struct sockaddr * to,proc_t p)3878 flow_divert_connect_out(struct socket *so, struct sockaddr *to, proc_t p)
3879 {
3880 #if CONTENT_FILTER
3881 	if (SOCK_TYPE(so) == SOCK_STREAM && !(so->so_flags & SOF_CONTENT_FILTER)) {
3882 		int error = cfil_sock_attach(so, NULL, to, CFS_CONNECTION_DIR_OUT);
3883 		if (error != 0) {
3884 			struct flow_divert_pcb  *fd_cb  = so->so_fd_pcb;
3885 			FDLOG(LOG_ERR, fd_cb, "Failed to attach cfil: %d", error);
3886 			return error;
3887 		}
3888 	}
3889 #endif /* CONTENT_FILTER */
3890 
3891 	return flow_divert_connect_out_internal(so, to, p, false);
3892 }
3893 
3894 static int
flow_divert_connectx_out_common(struct socket * so,struct sockaddr * dst,struct proc * p,uint32_t ifscope,sae_connid_t * pcid,struct uio * auio,user_ssize_t * bytes_written)3895 flow_divert_connectx_out_common(struct socket *so, struct sockaddr *dst,
3896     struct proc *p, uint32_t ifscope, sae_connid_t *pcid, struct uio *auio, user_ssize_t *bytes_written)
3897 {
3898 	struct inpcb *inp = sotoinpcb(so);
3899 	int error;
3900 
3901 	if (inp == NULL) {
3902 		return EINVAL;
3903 	}
3904 
3905 	VERIFY(dst != NULL);
3906 
3907 #if CONTENT_FILTER && NECP
3908 	struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
3909 	if (fd_cb != NULL && (fd_cb->flags & FLOW_DIVERT_HAS_TOKEN) &&
3910 	    SOCK_TYPE(so) == SOCK_STREAM && !(so->so_flags & SOF_CONTENT_FILTER)) {
3911 		inp_update_necp_policy(sotoinpcb(so), NULL, dst, 0);
3912 	}
3913 #endif /* CONTENT_FILTER */
3914 
3915 	/* bind socket to the specified interface, if requested */
3916 	if (ifscope != IFSCOPE_NONE &&
3917 	    (error = inp_bindif(inp, ifscope, NULL)) != 0) {
3918 		return error;
3919 	}
3920 
3921 	error = flow_divert_connect_out(so, dst, p);
3922 
3923 	if (error != 0) {
3924 		return error;
3925 	}
3926 
3927 	/* if there is data, send it */
3928 	if (auio != NULL) {
3929 		user_ssize_t datalen = 0;
3930 
3931 		socket_unlock(so, 0);
3932 
3933 		VERIFY(bytes_written != NULL);
3934 
3935 		datalen = uio_resid(auio);
3936 		error = so->so_proto->pr_usrreqs->pru_sosend(so, NULL, (uio_t)auio, NULL, NULL, 0);
3937 		socket_lock(so, 0);
3938 
3939 		if (error == 0 || error == EWOULDBLOCK) {
3940 			*bytes_written = datalen - uio_resid(auio);
3941 		}
3942 
3943 		/*
3944 		 * sosend returns EWOULDBLOCK if it's a non-blocking
3945 		 * socket or a timeout occured (this allows to return
3946 		 * the amount of queued data through sendit()).
3947 		 *
3948 		 * However, connectx() returns EINPROGRESS in case of a
3949 		 * blocking socket. So we change the return value here.
3950 		 */
3951 		if (error == EWOULDBLOCK) {
3952 			error = EINPROGRESS;
3953 		}
3954 	}
3955 
3956 	if (error == 0 && pcid != NULL) {
3957 		*pcid = 1;      /* there is only 1 connection for a TCP */
3958 	}
3959 
3960 	return error;
3961 }
3962 
3963 static int
flow_divert_connectx_out(struct socket * so,struct sockaddr * src __unused,struct sockaddr * dst,struct proc * p,uint32_t ifscope,sae_associd_t aid __unused,sae_connid_t * pcid,uint32_t flags __unused,void * arg __unused,uint32_t arglen __unused,struct uio * uio,user_ssize_t * bytes_written)3964 flow_divert_connectx_out(struct socket *so, struct sockaddr *src __unused,
3965     struct sockaddr *dst, struct proc *p, uint32_t ifscope,
3966     sae_associd_t aid __unused, sae_connid_t *pcid, uint32_t flags __unused, void *arg __unused,
3967     uint32_t arglen __unused, struct uio *uio, user_ssize_t *bytes_written)
3968 {
3969 	return flow_divert_connectx_out_common(so, dst, p, ifscope, pcid, uio, bytes_written);
3970 }
3971 
3972 static int
flow_divert_connectx6_out(struct socket * so,struct sockaddr * src __unused,struct sockaddr * dst,struct proc * p,uint32_t ifscope,sae_associd_t aid __unused,sae_connid_t * pcid,uint32_t flags __unused,void * arg __unused,uint32_t arglen __unused,struct uio * uio,user_ssize_t * bytes_written)3973 flow_divert_connectx6_out(struct socket *so, struct sockaddr *src __unused,
3974     struct sockaddr *dst, struct proc *p, uint32_t ifscope,
3975     sae_associd_t aid __unused, sae_connid_t *pcid, uint32_t flags __unused, void *arg __unused,
3976     uint32_t arglen __unused, struct uio *uio, user_ssize_t *bytes_written)
3977 {
3978 	return flow_divert_connectx_out_common(so, dst, p, ifscope, pcid, uio, bytes_written);
3979 }
3980 
3981 static errno_t
flow_divert_data_out(struct socket * so,int flags,mbuf_ref_t data,struct sockaddr * to,mbuf_ref_t control,struct proc * p)3982 flow_divert_data_out(struct socket *so, int flags, mbuf_ref_t data, struct sockaddr *to, mbuf_ref_t control, struct proc *p)
3983 {
3984 	struct flow_divert_pcb  *fd_cb  = so->so_fd_pcb;
3985 	int                     error   = 0;
3986 	struct inpcb            *inp;
3987 #if CONTENT_FILTER
3988 	struct m_tag *cfil_tag = NULL;
3989 #endif
3990 
3991 	if (!SO_IS_DIVERTED(so)) {
3992 		return EINVAL;
3993 	}
3994 
3995 	inp = sotoinpcb(so);
3996 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) {
3997 		error = ECONNRESET;
3998 		goto done;
3999 	}
4000 
4001 	if ((fd_cb->flags & FLOW_DIVERT_TUNNEL_WR_CLOSED) && SOCK_TYPE(so) == SOCK_DGRAM) {
4002 		/* The provider considers this datagram flow to be closed, so no data can be sent */
4003 		FDLOG0(LOG_INFO, fd_cb, "provider is no longer accepting writes, cannot send data");
4004 		error = EHOSTUNREACH;
4005 		goto done;
4006 	}
4007 
4008 #if CONTENT_FILTER
4009 	/*
4010 	 * If the socket is subject to a UDP Content Filter and no remote address is passed in,
4011 	 * retrieve the CFIL saved remote address from the mbuf and use it.
4012 	 */
4013 	if (to == NULL && CFIL_DGRAM_FILTERED(so)) {
4014 		struct sockaddr * __single cfil_faddr = NULL;
4015 		cfil_tag = cfil_dgram_get_socket_state(data, NULL, NULL, &cfil_faddr, NULL);
4016 		if (cfil_tag) {
4017 			to = (struct sockaddr *)(void *)cfil_faddr;
4018 		}
4019 		FDLOG(LOG_INFO, fd_cb, "Using remote address from CFIL saved state: %p", to);
4020 	}
4021 #endif
4022 
4023 	/* Implicit connect */
4024 	if (!(fd_cb->flags & FLOW_DIVERT_CONNECT_STARTED)) {
4025 		FDLOG0(LOG_INFO, fd_cb, "implicit connect");
4026 
4027 		error = flow_divert_connect_out_internal(so, to, p, true);
4028 		if (error) {
4029 			goto done;
4030 		}
4031 	} else {
4032 		error = flow_divert_check_no_cellular(fd_cb) ||
4033 		    flow_divert_check_no_expensive(fd_cb) ||
4034 		    flow_divert_check_no_constrained(fd_cb);
4035 		if (error) {
4036 			goto done;
4037 		}
4038 	}
4039 
4040 	if (data != NULL) {
4041 		size_t data_size = 0;
4042 		if (mbuf_flags(data) & M_PKTHDR) {
4043 			data_size = mbuf_pkthdr_len(data);
4044 		} else {
4045 			for (mbuf_t blob = data; blob != NULL; blob = mbuf_next(blob)) {
4046 				data_size += mbuf_len(blob);
4047 			}
4048 		}
4049 
4050 		FDLOG(LOG_DEBUG, fd_cb, "app wrote %lu bytes", data_size);
4051 		fd_cb->bytes_written_by_app += data_size;
4052 
4053 		error = flow_divert_send_app_data(fd_cb, data, data_size, to);
4054 
4055 		data = NULL;
4056 
4057 		if (error) {
4058 			goto done;
4059 		}
4060 	}
4061 
4062 	if (flags & PRUS_EOF) {
4063 		flow_divert_shutdown(so);
4064 	}
4065 
4066 done:
4067 	if (data) {
4068 		mbuf_freem(data);
4069 	}
4070 	if (control) {
4071 		mbuf_free(control);
4072 	}
4073 #if CONTENT_FILTER
4074 	if (cfil_tag) {
4075 		m_tag_free(cfil_tag);
4076 	}
4077 #endif
4078 
4079 	return error;
4080 }
4081 
4082 static int
flow_divert_preconnect(struct socket * so)4083 flow_divert_preconnect(struct socket *so)
4084 {
4085 	int error = 0;
4086 	struct flow_divert_pcb  *fd_cb  = so->so_fd_pcb;
4087 
4088 	if (!SO_IS_DIVERTED(so)) {
4089 		return EINVAL;
4090 	}
4091 
4092 	if (!(fd_cb->flags & FLOW_DIVERT_CONNECT_STARTED)) {
4093 		FDLOG0(LOG_INFO, fd_cb, "Pre-connect read: sending saved connect packet");
4094 		error = flow_divert_send_connect_packet(so->so_fd_pcb);
4095 		if (error) {
4096 			return error;
4097 		}
4098 
4099 		fd_cb->flags |= FLOW_DIVERT_CONNECT_STARTED;
4100 	}
4101 
4102 	soclearfastopen(so);
4103 
4104 	return error;
4105 }
4106 
4107 static void
flow_divert_set_protosw(struct socket * so)4108 flow_divert_set_protosw(struct socket *so)
4109 {
4110 	if (SOCK_DOM(so) == PF_INET) {
4111 		so->so_proto = &g_flow_divert_in_protosw;
4112 	} else {
4113 		so->so_proto = (struct protosw *)&g_flow_divert_in6_protosw;
4114 	}
4115 }
4116 
4117 static void
flow_divert_set_udp_protosw(struct socket * so)4118 flow_divert_set_udp_protosw(struct socket *so)
4119 {
4120 	if (SOCK_DOM(so) == PF_INET) {
4121 		so->so_proto = &g_flow_divert_in_udp_protosw;
4122 	} else {
4123 		so->so_proto = (struct protosw *)&g_flow_divert_in6_udp_protosw;
4124 	}
4125 }
4126 
4127 errno_t
flow_divert_implicit_data_out(struct socket * so,int flags,mbuf_ref_t data,struct sockaddr * to,mbuf_ref_t control,struct proc * p)4128 flow_divert_implicit_data_out(struct socket *so, int flags, mbuf_ref_t data, struct sockaddr *to, mbuf_ref_t control, struct proc *p)
4129 {
4130 	struct flow_divert_pcb  *fd_cb  = so->so_fd_pcb;
4131 	struct inpcb *inp;
4132 	int error = 0;
4133 
4134 	inp = sotoinpcb(so);
4135 	if (inp == NULL) {
4136 		return EINVAL;
4137 	}
4138 
4139 	if (fd_cb == NULL) {
4140 		error = flow_divert_pcb_init(so);
4141 		fd_cb  = so->so_fd_pcb;
4142 		if (error != 0 || fd_cb == NULL) {
4143 			goto done;
4144 		}
4145 	}
4146 	return flow_divert_data_out(so, flags, data, to, control, p);
4147 
4148 done:
4149 	if (data) {
4150 		mbuf_freem(data);
4151 	}
4152 	if (control) {
4153 		mbuf_free(control);
4154 	}
4155 
4156 	return error;
4157 }
4158 
4159 static errno_t
flow_divert_pcb_init_internal(struct socket * so,uint32_t ctl_unit,uint32_t aggregate_unit)4160 flow_divert_pcb_init_internal(struct socket *so, uint32_t ctl_unit, uint32_t aggregate_unit)
4161 {
4162 	errno_t error = 0;
4163 	struct flow_divert_pcb *fd_cb = NULL;
4164 	uint32_t agg_unit = aggregate_unit;
4165 	uint32_t policy_control_unit = ctl_unit;
4166 	bool is_aggregate = false;
4167 
4168 	if (so->so_flags & SOF_FLOW_DIVERT) {
4169 		return EALREADY;
4170 	}
4171 
4172 	fd_cb = flow_divert_pcb_create(so);
4173 	if (fd_cb == NULL) {
4174 		return ENOMEM;
4175 	}
4176 
4177 	do {
4178 		uint32_t group_unit = flow_divert_derive_kernel_control_unit(so->last_pid, &policy_control_unit, &agg_unit, &is_aggregate);
4179 		if (group_unit == 0 || (group_unit >= GROUP_COUNT_MAX && group_unit < FLOW_DIVERT_IN_PROCESS_UNIT_MIN)) {
4180 			FDLOG0(LOG_ERR, fd_cb, "No valid group is available, cannot init flow divert");
4181 			error = EINVAL;
4182 			break;
4183 		}
4184 
4185 		error = flow_divert_add_to_group(fd_cb, group_unit);
4186 		if (error == 0) {
4187 			so->so_fd_pcb = fd_cb;
4188 			so->so_flags |= SOF_FLOW_DIVERT;
4189 			fd_cb->control_group_unit = group_unit;
4190 			fd_cb->policy_control_unit = ctl_unit;
4191 			fd_cb->aggregate_unit = agg_unit;
4192 			if (is_aggregate) {
4193 				fd_cb->flags |= FLOW_DIVERT_FLOW_IS_TRANSPARENT;
4194 			} else {
4195 				fd_cb->flags &= ~FLOW_DIVERT_FLOW_IS_TRANSPARENT;
4196 			}
4197 
4198 			if (SOCK_TYPE(so) == SOCK_STREAM) {
4199 				flow_divert_set_protosw(so);
4200 			} else if (SOCK_TYPE(so) == SOCK_DGRAM) {
4201 				flow_divert_set_udp_protosw(so);
4202 			}
4203 
4204 			FDLOG0(LOG_INFO, fd_cb, "Created");
4205 		} else if (error != ENOENT) {
4206 			FDLOG(LOG_ERR, fd_cb, "pcb insert failed: %d", error);
4207 		}
4208 	} while (error == ENOENT);
4209 
4210 	if (error != 0) {
4211 		FDRELEASE(fd_cb);
4212 	}
4213 
4214 	return error;
4215 }
4216 
4217 errno_t
flow_divert_pcb_init(struct socket * so)4218 flow_divert_pcb_init(struct socket *so)
4219 {
4220 	struct inpcb *inp = sotoinpcb(so);
4221 	uint32_t aggregate_units = 0;
4222 	uint32_t ctl_unit = necp_socket_get_flow_divert_control_unit(inp, &aggregate_units);
4223 	return flow_divert_pcb_init_internal(so, ctl_unit, aggregate_units);
4224 }
4225 
4226 errno_t
flow_divert_token_set(struct socket * so,struct sockopt * sopt)4227 flow_divert_token_set(struct socket *so, struct sockopt *sopt)
4228 {
4229 	uint32_t        ctl_unit        = 0;
4230 	uint32_t        key_unit        = 0;
4231 	uint32_t        aggregate_unit  = 0;
4232 	int             error           = 0;
4233 	int             hmac_error      = 0;
4234 	mbuf_ref_t      token           = NULL;
4235 
4236 	if (so->so_flags & SOF_FLOW_DIVERT) {
4237 		error = EALREADY;
4238 		goto done;
4239 	}
4240 
4241 	if (g_init_result) {
4242 		FDLOG(LOG_ERR, &nil_pcb, "flow_divert_init failed (%d), cannot use flow divert", g_init_result);
4243 		error = ENOPROTOOPT;
4244 		goto done;
4245 	}
4246 
4247 	if ((SOCK_TYPE(so) != SOCK_STREAM && SOCK_TYPE(so) != SOCK_DGRAM) ||
4248 	    (SOCK_PROTO(so) != IPPROTO_TCP && SOCK_PROTO(so) != IPPROTO_UDP) ||
4249 	    (SOCK_DOM(so) != PF_INET && SOCK_DOM(so) != PF_INET6)) {
4250 		error = EINVAL;
4251 		goto done;
4252 	} else {
4253 		if (SOCK_TYPE(so) == SOCK_STREAM && SOCK_PROTO(so) == IPPROTO_TCP) {
4254 			struct tcpcb *tp = sototcpcb(so);
4255 			if (tp == NULL || tp->t_state != TCPS_CLOSED) {
4256 				error = EINVAL;
4257 				goto done;
4258 			}
4259 		}
4260 	}
4261 
4262 	error = soopt_getm(sopt, &token);
4263 	if (error) {
4264 		token = NULL;
4265 		goto done;
4266 	}
4267 
4268 	error = soopt_mcopyin(sopt, token);
4269 	if (error) {
4270 		token = NULL;
4271 		goto done;
4272 	}
4273 
4274 	error = flow_divert_packet_get_tlv(token, 0, FLOW_DIVERT_TLV_KEY_UNIT, sizeof(key_unit), (void *)&key_unit, NULL);
4275 	if (!error) {
4276 		key_unit = ntohl(key_unit);
4277 		if (key_unit >= GROUP_COUNT_MAX) {
4278 			key_unit = 0;
4279 		}
4280 	} else if (error != ENOENT) {
4281 		FDLOG(LOG_ERR, &nil_pcb, "Failed to get the key unit from the token: %d", error);
4282 		goto done;
4283 	} else {
4284 		key_unit = 0;
4285 	}
4286 
4287 	error = flow_divert_packet_get_tlv(token, 0, FLOW_DIVERT_TLV_CTL_UNIT, sizeof(ctl_unit), (void *)&ctl_unit, NULL);
4288 	if (error) {
4289 		FDLOG(LOG_ERR, &nil_pcb, "Failed to get the control socket unit from the token: %d", error);
4290 		goto done;
4291 	}
4292 
4293 	error = flow_divert_packet_get_tlv(token, 0, FLOW_DIVERT_TLV_AGGREGATE_UNIT, sizeof(aggregate_unit), (void *)&aggregate_unit, NULL);
4294 	if (error && error != ENOENT) {
4295 		FDLOG(LOG_ERR, &nil_pcb, "Failed to get the aggregate unit from the token: %d", error);
4296 		goto done;
4297 	}
4298 
4299 	/* A valid kernel control unit is required */
4300 	ctl_unit = ntohl(ctl_unit);
4301 	aggregate_unit = ntohl(aggregate_unit);
4302 
4303 	if (ctl_unit > 0 && ctl_unit < GROUP_COUNT_MAX) {
4304 		hmac_error = flow_divert_packet_verify_hmac(token, (key_unit != 0 ? key_unit : ctl_unit));
4305 		if (hmac_error && hmac_error != ENOENT) {
4306 			FDLOG(LOG_ERR, &nil_pcb, "HMAC verfication failed: %d", hmac_error);
4307 			error = hmac_error;
4308 			goto done;
4309 		}
4310 	}
4311 
4312 	error = flow_divert_pcb_init_internal(so, ctl_unit, aggregate_unit);
4313 	if (error == 0) {
4314 		struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
4315 		int log_level = LOG_NOTICE;
4316 
4317 		error = flow_divert_packet_get_tlv(token, 0, FLOW_DIVERT_TLV_LOG_LEVEL, sizeof(log_level), &log_level, NULL);
4318 		if (error == 0) {
4319 			fd_cb->log_level = (uint8_t)log_level;
4320 		}
4321 		error = 0;
4322 
4323 		fd_cb->connect_token = token;
4324 		token = NULL;
4325 
4326 		fd_cb->flags |= FLOW_DIVERT_HAS_TOKEN;
4327 	}
4328 
4329 	if (hmac_error == 0) {
4330 		struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
4331 		if (fd_cb != NULL) {
4332 			fd_cb->flags |= FLOW_DIVERT_HAS_HMAC;
4333 		}
4334 	}
4335 
4336 done:
4337 	if (token != NULL) {
4338 		mbuf_freem(token);
4339 	}
4340 
4341 	return error;
4342 }
4343 
4344 errno_t
flow_divert_token_get(struct socket * so,struct sockopt * sopt)4345 flow_divert_token_get(struct socket *so, struct sockopt *sopt)
4346 {
4347 	uint32_t                    ctl_unit;
4348 	int                         error                   = 0;
4349 	uint8_t                     hmac[SHA_DIGEST_LENGTH];
4350 	struct flow_divert_pcb      *fd_cb                  = so->so_fd_pcb;
4351 	mbuf_ref_t                  token                   = NULL;
4352 	struct flow_divert_group    *control_group          = NULL;
4353 
4354 	if (!SO_IS_DIVERTED(so)) {
4355 		error = EINVAL;
4356 		goto done;
4357 	}
4358 
4359 	if (fd_cb->group == NULL) {
4360 		error = EINVAL;
4361 		goto done;
4362 	}
4363 
4364 	error = mbuf_gethdr(MBUF_DONTWAIT, MBUF_TYPE_HEADER, &token);
4365 	if (error) {
4366 		FDLOG(LOG_ERR, fd_cb, "failed to allocate the header mbuf: %d", error);
4367 		goto done;
4368 	}
4369 
4370 	ctl_unit = htonl(fd_cb->group->ctl_unit);
4371 
4372 	error = flow_divert_packet_append_tlv(token, FLOW_DIVERT_TLV_CTL_UNIT, sizeof(ctl_unit), &ctl_unit);
4373 	if (error) {
4374 		goto done;
4375 	}
4376 
4377 	error = flow_divert_packet_append_tlv(token, FLOW_DIVERT_TLV_FLOW_ID, sizeof(fd_cb->hash), &fd_cb->hash);
4378 	if (error) {
4379 		goto done;
4380 	}
4381 
4382 	if (fd_cb->app_data != NULL) {
4383 		error = flow_divert_packet_append_tlv(token, FLOW_DIVERT_TLV_APP_DATA, (uint32_t)fd_cb->app_data_length, fd_cb->app_data);
4384 		if (error) {
4385 			goto done;
4386 		}
4387 	}
4388 
4389 	control_group = flow_divert_group_lookup(fd_cb->control_group_unit, fd_cb);
4390 	if (control_group != NULL) {
4391 		lck_rw_lock_shared(&control_group->lck);
4392 		ctl_unit = htonl(control_group->ctl_unit);
4393 		error = flow_divert_packet_append_tlv(token, FLOW_DIVERT_TLV_KEY_UNIT, sizeof(ctl_unit), &ctl_unit);
4394 		if (!error) {
4395 			error = flow_divert_packet_compute_hmac(token, control_group, hmac);
4396 		}
4397 		lck_rw_done(&control_group->lck);
4398 		FDGRP_RELEASE(control_group);
4399 	} else {
4400 		error = ENOPROTOOPT;
4401 	}
4402 
4403 	if (error) {
4404 		goto done;
4405 	}
4406 
4407 	error = flow_divert_packet_append_tlv(token, FLOW_DIVERT_TLV_HMAC, sizeof(hmac), hmac);
4408 	if (error) {
4409 		goto done;
4410 	}
4411 
4412 	if (sopt->sopt_val == USER_ADDR_NULL) {
4413 		/* If the caller passed NULL to getsockopt, just set the size of the token and return */
4414 		sopt->sopt_valsize = mbuf_pkthdr_len(token);
4415 		goto done;
4416 	}
4417 
4418 	error = soopt_mcopyout(sopt, token);
4419 	if (error) {
4420 		token = NULL;   /* For some reason, soopt_mcopyout() frees the mbuf if it fails */
4421 		goto done;
4422 	}
4423 
4424 done:
4425 	if (token != NULL) {
4426 		mbuf_freem(token);
4427 	}
4428 
4429 	return error;
4430 }
4431 
4432 void
flow_divert_group_destroy(struct flow_divert_group * group)4433 flow_divert_group_destroy(struct flow_divert_group *group)
4434 {
4435 	lck_rw_lock_exclusive(&group->lck);
4436 
4437 	FDLOG(LOG_NOTICE, &nil_pcb, "Destroying group %u", group->ctl_unit);
4438 
4439 	if (group->token_key != NULL) {
4440 		memset(group->token_key, 0, group->token_key_size);
4441 		kfree_data_sized_by(group->token_key, group->token_key_size);
4442 	}
4443 
4444 	/* Re-set the current trie */
4445 	if (group->signing_id_trie.memory != NULL) {
4446 		kfree_data_sized_by(group->signing_id_trie.memory, group->signing_id_trie.memory_size);
4447 	}
4448 	memset(&group->signing_id_trie, 0, sizeof(group->signing_id_trie));
4449 	group->signing_id_trie.root = NULL_TRIE_IDX;
4450 
4451 	lck_rw_done(&group->lck);
4452 
4453 	zfree(flow_divert_group_zone, group);
4454 }
4455 
4456 static struct flow_divert_group *
flow_divert_allocate_group(u_int32_t unit,pid_t pid)4457 flow_divert_allocate_group(u_int32_t unit, pid_t pid)
4458 {
4459 	struct flow_divert_group *new_group = NULL;
4460 	new_group = zalloc_flags(flow_divert_group_zone, Z_WAITOK | Z_ZERO);
4461 	lck_rw_init(&new_group->lck, &flow_divert_mtx_grp, &flow_divert_mtx_attr);
4462 	RB_INIT(&new_group->pcb_tree);
4463 	new_group->ctl_unit = unit;
4464 	new_group->in_process_pid = pid;
4465 	MBUFQ_INIT(&new_group->send_queue);
4466 	new_group->signing_id_trie.root = NULL_TRIE_IDX;
4467 	new_group->ref_count = 1;
4468 	new_group->order = FLOW_DIVERT_ORDER_LAST;
4469 	return new_group;
4470 }
4471 
4472 static errno_t
flow_divert_kctl_setup(u_int32_t * unit,void ** unitinfo)4473 flow_divert_kctl_setup(u_int32_t *unit, void **unitinfo)
4474 {
4475 	if (unit == NULL || unitinfo == NULL) {
4476 		return EINVAL;
4477 	}
4478 
4479 	struct flow_divert_group *new_group = NULL;
4480 	errno_t error = 0;
4481 	lck_rw_lock_shared(&g_flow_divert_group_lck);
4482 	if (*unit == FLOW_DIVERT_IN_PROCESS_UNIT) {
4483 		// Return next unused in-process unit
4484 		u_int32_t unit_cursor = FLOW_DIVERT_IN_PROCESS_UNIT_MIN;
4485 		struct flow_divert_group *group_next = NULL;
4486 		TAILQ_FOREACH(group_next, &g_flow_divert_in_process_group_list, chain) {
4487 			if (group_next->ctl_unit > unit_cursor) {
4488 				// Found a gap, lets fill it in
4489 				break;
4490 			}
4491 			unit_cursor = group_next->ctl_unit + 1;
4492 			if (unit_cursor == FLOW_DIVERT_IN_PROCESS_UNIT_MAX) {
4493 				break;
4494 			}
4495 		}
4496 		if (unit_cursor == FLOW_DIVERT_IN_PROCESS_UNIT_MAX) {
4497 			error = EBUSY;
4498 		} else {
4499 			*unit = unit_cursor;
4500 			new_group = flow_divert_allocate_group(*unit, proc_pid(current_proc()));
4501 			if (group_next != NULL) {
4502 				TAILQ_INSERT_BEFORE(group_next, new_group, chain);
4503 			} else {
4504 				TAILQ_INSERT_TAIL(&g_flow_divert_in_process_group_list, new_group, chain);
4505 			}
4506 			g_active_group_count++;
4507 		}
4508 	} else {
4509 		if (kauth_cred_issuser(kauth_cred_get()) == 0) {
4510 			error = EPERM;
4511 		} else {
4512 			if (g_flow_divert_groups == NULL) {
4513 				g_flow_divert_groups = kalloc_type(struct flow_divert_group *,
4514 				    GROUP_COUNT_MAX, Z_WAITOK | Z_ZERO | Z_NOFAIL);
4515 			}
4516 
4517 			// Return next unused group unit
4518 			bool found_unused_unit = false;
4519 			u_int32_t unit_cursor;
4520 			for (unit_cursor = 1; unit_cursor < GROUP_COUNT_MAX; unit_cursor++) {
4521 				struct flow_divert_group *group = g_flow_divert_groups[unit_cursor];
4522 				if (group == NULL) {
4523 					// Open slot, assign this one
4524 					*unit = unit_cursor;
4525 					new_group = flow_divert_allocate_group(*unit, 0);
4526 					g_flow_divert_groups[*unit] = new_group;
4527 					found_unused_unit = true;
4528 					g_active_group_count++;
4529 					break;
4530 				}
4531 			}
4532 			if (!found_unused_unit) {
4533 				error = EBUSY;
4534 			}
4535 		}
4536 	}
4537 	lck_rw_done(&g_flow_divert_group_lck);
4538 
4539 	*unitinfo = new_group;
4540 
4541 	return error;
4542 }
4543 
4544 static errno_t
flow_divert_kctl_connect(kern_ctl_ref kctlref __unused,struct sockaddr_ctl * sac,void ** unitinfo)4545 flow_divert_kctl_connect(kern_ctl_ref kctlref __unused, struct sockaddr_ctl *sac, void **unitinfo)
4546 {
4547 	if (unitinfo == NULL) {
4548 		return EINVAL;
4549 	}
4550 
4551 	// Just validate. The group will already have been allocated.
4552 	struct flow_divert_group *group = (struct flow_divert_group *)*unitinfo;
4553 	if (group == NULL || sac->sc_unit != group->ctl_unit) {
4554 		FDLOG(LOG_ERR, &nil_pcb, "Flow divert connect fail, unit mismatch %u != %u",
4555 		    sac->sc_unit, group ? group->ctl_unit : 0);
4556 		return EINVAL;
4557 	}
4558 
4559 	return 0;
4560 }
4561 
4562 static errno_t
flow_divert_kctl_disconnect(kern_ctl_ref kctlref __unused,uint32_t unit,void * unitinfo)4563 flow_divert_kctl_disconnect(kern_ctl_ref kctlref __unused, uint32_t unit, void *unitinfo)
4564 {
4565 	struct flow_divert_group    *group  = NULL;
4566 	errno_t                     error   = 0;
4567 
4568 	if (unitinfo == NULL) {
4569 		return 0;
4570 	}
4571 
4572 	FDLOG(LOG_INFO, &nil_pcb, "disconnecting group %d", unit);
4573 
4574 	lck_rw_lock_exclusive(&g_flow_divert_group_lck);
4575 
4576 	if (g_active_group_count == 0) {
4577 		panic("flow divert group %u is disconnecting, but no groups are active (active count = %u)",
4578 		    unit, g_active_group_count);
4579 	}
4580 
4581 	if (unit < FLOW_DIVERT_IN_PROCESS_UNIT_MIN) {
4582 		if (unit >= GROUP_COUNT_MAX) {
4583 			return EINVAL;
4584 		}
4585 
4586 		if (g_flow_divert_groups == NULL) {
4587 			panic("flow divert group %u is disconnecting, but groups array is NULL",
4588 			    unit);
4589 		}
4590 		group = g_flow_divert_groups[unit];
4591 
4592 		if (group != (struct flow_divert_group *)unitinfo) {
4593 			panic("group with unit %d (%p) != unit info (%p)", unit, group, unitinfo);
4594 		}
4595 
4596 		g_flow_divert_groups[unit] = NULL;
4597 	} else {
4598 		group = (struct flow_divert_group *)unitinfo;
4599 		if (TAILQ_EMPTY(&g_flow_divert_in_process_group_list)) {
4600 			panic("flow divert group %u is disconnecting, but in-process group list is empty",
4601 			    unit);
4602 		}
4603 
4604 		TAILQ_REMOVE(&g_flow_divert_in_process_group_list, group, chain);
4605 	}
4606 
4607 	g_active_group_count--;
4608 
4609 	if (g_active_group_count == 0) {
4610 		kfree_type(struct flow_divert_group *,
4611 		    GROUP_COUNT_MAX, g_flow_divert_groups);
4612 		g_flow_divert_groups = NULL;
4613 	}
4614 
4615 	lck_rw_done(&g_flow_divert_group_lck);
4616 
4617 	if (group != NULL) {
4618 		flow_divert_close_all(group);
4619 		FDGRP_RELEASE(group);
4620 	} else {
4621 		error = EINVAL;
4622 	}
4623 
4624 	return error;
4625 }
4626 
4627 static errno_t
flow_divert_kctl_send(__unused kern_ctl_ref kctlref,uint32_t unit,__unused void * unitinfo,mbuf_ref_t m,__unused int flags)4628 flow_divert_kctl_send(__unused kern_ctl_ref kctlref, uint32_t unit, __unused void *unitinfo, mbuf_ref_t m, __unused int flags)
4629 {
4630 	errno_t error = 0;
4631 	struct flow_divert_group *group = flow_divert_group_lookup(unit, NULL);
4632 	if (group != NULL) {
4633 		error = flow_divert_input(m, group);
4634 		FDGRP_RELEASE(group);
4635 	} else {
4636 		error = ENOENT;
4637 	}
4638 	return error;
4639 }
4640 
4641 static void
flow_divert_kctl_rcvd(__unused kern_ctl_ref kctlref,uint32_t unit,__unused void * unitinfo,__unused int flags)4642 flow_divert_kctl_rcvd(__unused kern_ctl_ref kctlref, uint32_t unit, __unused void *unitinfo, __unused int flags)
4643 {
4644 	struct flow_divert_group *group = flow_divert_group_lookup(unit, NULL);
4645 	if (group == NULL) {
4646 		return;
4647 	}
4648 
4649 	if (!OSTestAndClear(GROUP_BIT_CTL_ENQUEUE_BLOCKED, &group->atomic_bits)) {
4650 		struct flow_divert_pcb                  *fd_cb;
4651 		SLIST_HEAD(, flow_divert_pcb)   tmp_list;
4652 
4653 		lck_rw_lock_exclusive(&group->lck);
4654 
4655 		while (!MBUFQ_EMPTY(&group->send_queue)) {
4656 			mbuf_ref_t next_packet;
4657 			FDLOG0(LOG_DEBUG, &nil_pcb, "trying ctl_enqueuembuf again");
4658 			next_packet = MBUFQ_FIRST(&group->send_queue);
4659 			int error = ctl_enqueuembuf(g_flow_divert_kctl_ref, group->ctl_unit, next_packet, CTL_DATA_EOR);
4660 			if (error) {
4661 				FDLOG(LOG_NOTICE, &nil_pcb, "flow_divert_kctl_rcvd: ctl_enqueuembuf returned an error: %d", error);
4662 				OSTestAndSet(GROUP_BIT_CTL_ENQUEUE_BLOCKED, &group->atomic_bits);
4663 				lck_rw_done(&group->lck);
4664 				return;
4665 			}
4666 			MBUFQ_DEQUEUE(&group->send_queue, next_packet);
4667 		}
4668 
4669 		SLIST_INIT(&tmp_list);
4670 
4671 		RB_FOREACH(fd_cb, fd_pcb_tree, &group->pcb_tree) {
4672 			FDRETAIN(fd_cb);
4673 			SLIST_INSERT_HEAD(&tmp_list, fd_cb, tmp_list_entry);
4674 		}
4675 
4676 		lck_rw_done(&group->lck);
4677 
4678 		SLIST_FOREACH(fd_cb, &tmp_list, tmp_list_entry) {
4679 			FDLOCK(fd_cb);
4680 			if (fd_cb->so != NULL) {
4681 				socket_lock(fd_cb->so, 0);
4682 				if (fd_cb->group != NULL) {
4683 					flow_divert_send_buffered_data(fd_cb, FALSE);
4684 				}
4685 				socket_unlock(fd_cb->so, 0);
4686 			}
4687 			FDUNLOCK(fd_cb);
4688 			FDRELEASE(fd_cb);
4689 		}
4690 	}
4691 
4692 	FDGRP_RELEASE(group);
4693 }
4694 
4695 static int
flow_divert_kctl_init(void)4696 flow_divert_kctl_init(void)
4697 {
4698 	struct kern_ctl_reg     ctl_reg;
4699 	int                     result;
4700 
4701 	memset(&ctl_reg, 0, sizeof(ctl_reg));
4702 
4703 	strlcpy(ctl_reg.ctl_name, FLOW_DIVERT_CONTROL_NAME, sizeof(ctl_reg.ctl_name));
4704 	ctl_reg.ctl_name[sizeof(ctl_reg.ctl_name) - 1] = '\0';
4705 
4706 	// Do not restrict to privileged processes. flow_divert_kctl_setup checks
4707 	// permissions separately.
4708 	ctl_reg.ctl_flags = CTL_FLAG_REG_EXTENDED | CTL_FLAG_REG_SETUP;
4709 	ctl_reg.ctl_sendsize = FD_CTL_SENDBUFF_SIZE;
4710 
4711 	ctl_reg.ctl_connect = flow_divert_kctl_connect;
4712 	ctl_reg.ctl_disconnect = flow_divert_kctl_disconnect;
4713 	ctl_reg.ctl_send = flow_divert_kctl_send;
4714 	ctl_reg.ctl_rcvd = flow_divert_kctl_rcvd;
4715 	ctl_reg.ctl_setup = flow_divert_kctl_setup;
4716 
4717 	result = ctl_register(&ctl_reg, &g_flow_divert_kctl_ref);
4718 
4719 	if (result) {
4720 		FDLOG(LOG_ERR, &nil_pcb, "flow_divert_kctl_init - ctl_register failed: %d\n", result);
4721 		return result;
4722 	}
4723 
4724 	return 0;
4725 }
4726 
4727 void
flow_divert_init(void)4728 flow_divert_init(void)
4729 {
4730 	memset(&nil_pcb, 0, sizeof(nil_pcb));
4731 	nil_pcb.log_level = LOG_NOTICE;
4732 
4733 	g_tcp_protosw = pffindproto(AF_INET, IPPROTO_TCP, SOCK_STREAM);
4734 
4735 	VERIFY(g_tcp_protosw != NULL);
4736 
4737 	memcpy(&g_flow_divert_in_protosw, g_tcp_protosw, sizeof(g_flow_divert_in_protosw));
4738 	memcpy(&g_flow_divert_in_usrreqs, g_tcp_protosw->pr_usrreqs, sizeof(g_flow_divert_in_usrreqs));
4739 
4740 	g_flow_divert_in_usrreqs.pru_connect = flow_divert_connect_out;
4741 	g_flow_divert_in_usrreqs.pru_connectx = flow_divert_connectx_out;
4742 	g_flow_divert_in_usrreqs.pru_disconnect = flow_divert_close;
4743 	g_flow_divert_in_usrreqs.pru_disconnectx = flow_divert_disconnectx;
4744 	g_flow_divert_in_usrreqs.pru_rcvd = flow_divert_rcvd;
4745 	g_flow_divert_in_usrreqs.pru_send = flow_divert_data_out;
4746 	g_flow_divert_in_usrreqs.pru_shutdown = flow_divert_shutdown;
4747 	g_flow_divert_in_usrreqs.pru_preconnect = flow_divert_preconnect;
4748 
4749 	g_flow_divert_in_protosw.pr_usrreqs = &g_flow_divert_in_usrreqs;
4750 	g_flow_divert_in_protosw.pr_ctloutput = flow_divert_ctloutput;
4751 
4752 	/*
4753 	 * Socket filters shouldn't attach/detach to/from this protosw
4754 	 * since pr_protosw is to be used instead, which points to the
4755 	 * real protocol; if they do, it is a bug and we should panic.
4756 	 */
4757 	g_flow_divert_in_protosw.pr_filter_head.tqh_first =
4758 	    __unsafe_forge_single(struct socket_filter *, (uintptr_t)0xdeadbeefdeadbeef);
4759 	g_flow_divert_in_protosw.pr_filter_head.tqh_last =
4760 	    __unsafe_forge_single(struct socket_filter **, (uintptr_t)0xdeadbeefdeadbeef);
4761 
4762 	/* UDP */
4763 	g_udp_protosw = pffindproto(AF_INET, IPPROTO_UDP, SOCK_DGRAM);
4764 	VERIFY(g_udp_protosw != NULL);
4765 
4766 	memcpy(&g_flow_divert_in_udp_protosw, g_udp_protosw, sizeof(g_flow_divert_in_udp_protosw));
4767 	memcpy(&g_flow_divert_in_udp_usrreqs, g_udp_protosw->pr_usrreqs, sizeof(g_flow_divert_in_udp_usrreqs));
4768 
4769 	g_flow_divert_in_udp_usrreqs.pru_connect = flow_divert_connect_out;
4770 	g_flow_divert_in_udp_usrreqs.pru_connectx = flow_divert_connectx_out;
4771 	g_flow_divert_in_udp_usrreqs.pru_disconnect = flow_divert_close;
4772 	g_flow_divert_in_udp_usrreqs.pru_disconnectx = flow_divert_disconnectx;
4773 	g_flow_divert_in_udp_usrreqs.pru_rcvd = flow_divert_rcvd;
4774 	g_flow_divert_in_udp_usrreqs.pru_send = flow_divert_data_out;
4775 	g_flow_divert_in_udp_usrreqs.pru_shutdown = flow_divert_shutdown;
4776 	g_flow_divert_in_udp_usrreqs.pru_sosend_list = pru_sosend_list_notsupp;
4777 	g_flow_divert_in_udp_usrreqs.pru_preconnect = flow_divert_preconnect;
4778 
4779 	g_flow_divert_in_udp_protosw.pr_usrreqs = &g_flow_divert_in_usrreqs;
4780 	g_flow_divert_in_udp_protosw.pr_ctloutput = flow_divert_ctloutput;
4781 
4782 	/*
4783 	 * Socket filters shouldn't attach/detach to/from this protosw
4784 	 * since pr_protosw is to be used instead, which points to the
4785 	 * real protocol; if they do, it is a bug and we should panic.
4786 	 */
4787 	g_flow_divert_in_udp_protosw.pr_filter_head.tqh_first =
4788 	    __unsafe_forge_single(struct socket_filter *, (uintptr_t)0xdeadbeefdeadbeef);
4789 	g_flow_divert_in_udp_protosw.pr_filter_head.tqh_last =
4790 	    __unsafe_forge_single(struct socket_filter **, (uintptr_t)0xdeadbeefdeadbeef);
4791 
4792 	g_tcp6_protosw = (struct ip6protosw *)pffindproto(AF_INET6, IPPROTO_TCP, SOCK_STREAM);
4793 
4794 	VERIFY(g_tcp6_protosw != NULL);
4795 
4796 	memcpy(&g_flow_divert_in6_protosw, g_tcp6_protosw, sizeof(g_flow_divert_in6_protosw));
4797 	memcpy(&g_flow_divert_in6_usrreqs, g_tcp6_protosw->pr_usrreqs, sizeof(g_flow_divert_in6_usrreqs));
4798 
4799 	g_flow_divert_in6_usrreqs.pru_connect = flow_divert_connect_out;
4800 	g_flow_divert_in6_usrreqs.pru_connectx = flow_divert_connectx6_out;
4801 	g_flow_divert_in6_usrreqs.pru_disconnect = flow_divert_close;
4802 	g_flow_divert_in6_usrreqs.pru_disconnectx = flow_divert_disconnectx;
4803 	g_flow_divert_in6_usrreqs.pru_rcvd = flow_divert_rcvd;
4804 	g_flow_divert_in6_usrreqs.pru_send = flow_divert_data_out;
4805 	g_flow_divert_in6_usrreqs.pru_shutdown = flow_divert_shutdown;
4806 	g_flow_divert_in6_usrreqs.pru_preconnect = flow_divert_preconnect;
4807 
4808 	g_flow_divert_in6_protosw.pr_usrreqs = &g_flow_divert_in6_usrreqs;
4809 	g_flow_divert_in6_protosw.pr_ctloutput = flow_divert_ctloutput;
4810 	/*
4811 	 * Socket filters shouldn't attach/detach to/from this protosw
4812 	 * since pr_protosw is to be used instead, which points to the
4813 	 * real protocol; if they do, it is a bug and we should panic.
4814 	 */
4815 	g_flow_divert_in6_protosw.pr_filter_head.tqh_first =
4816 	    __unsafe_forge_single(struct socket_filter *, (uintptr_t)0xdeadbeefdeadbeef);
4817 	g_flow_divert_in6_protosw.pr_filter_head.tqh_last =
4818 	    __unsafe_forge_single(struct socket_filter **, (uintptr_t)0xdeadbeefdeadbeef);
4819 
4820 	/* UDP6 */
4821 	g_udp6_protosw = (struct ip6protosw *)pffindproto(AF_INET6, IPPROTO_UDP, SOCK_DGRAM);
4822 
4823 	VERIFY(g_udp6_protosw != NULL);
4824 
4825 	memcpy(&g_flow_divert_in6_udp_protosw, g_udp6_protosw, sizeof(g_flow_divert_in6_udp_protosw));
4826 	memcpy(&g_flow_divert_in6_udp_usrreqs, g_udp6_protosw->pr_usrreqs, sizeof(g_flow_divert_in6_udp_usrreqs));
4827 
4828 	g_flow_divert_in6_udp_usrreqs.pru_connect = flow_divert_connect_out;
4829 	g_flow_divert_in6_udp_usrreqs.pru_connectx = flow_divert_connectx6_out;
4830 	g_flow_divert_in6_udp_usrreqs.pru_disconnect = flow_divert_close;
4831 	g_flow_divert_in6_udp_usrreqs.pru_disconnectx = flow_divert_disconnectx;
4832 	g_flow_divert_in6_udp_usrreqs.pru_rcvd = flow_divert_rcvd;
4833 	g_flow_divert_in6_udp_usrreqs.pru_send = flow_divert_data_out;
4834 	g_flow_divert_in6_udp_usrreqs.pru_shutdown = flow_divert_shutdown;
4835 	g_flow_divert_in6_udp_usrreqs.pru_sosend_list = pru_sosend_list_notsupp;
4836 	g_flow_divert_in6_udp_usrreqs.pru_preconnect = flow_divert_preconnect;
4837 
4838 	g_flow_divert_in6_udp_protosw.pr_usrreqs = &g_flow_divert_in6_udp_usrreqs;
4839 	g_flow_divert_in6_udp_protosw.pr_ctloutput = flow_divert_ctloutput;
4840 	/*
4841 	 * Socket filters shouldn't attach/detach to/from this protosw
4842 	 * since pr_protosw is to be used instead, which points to the
4843 	 * real protocol; if they do, it is a bug and we should panic.
4844 	 */
4845 	g_flow_divert_in6_udp_protosw.pr_filter_head.tqh_first =
4846 	    __unsafe_forge_single(struct socket_filter *, (uintptr_t)0xdeadbeefdeadbeef);
4847 	g_flow_divert_in6_udp_protosw.pr_filter_head.tqh_last =
4848 	    __unsafe_forge_single(struct socket_filter **, (uintptr_t)0xdeadbeefdeadbeef);
4849 
4850 	TAILQ_INIT(&g_flow_divert_in_process_group_list);
4851 
4852 	g_init_result = flow_divert_kctl_init();
4853 	if (g_init_result) {
4854 		goto done;
4855 	}
4856 
4857 done:
4858 	if (g_init_result != 0) {
4859 		if (g_flow_divert_kctl_ref != NULL) {
4860 			ctl_deregister(g_flow_divert_kctl_ref);
4861 			g_flow_divert_kctl_ref = NULL;
4862 		}
4863 	}
4864 }
4865