]> git.saurik.com Git - apple/xnu.git/blob - bsd/netinet/flow_divert.c
projects / apple / xnu.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
xnu-4570.61.1.tar.gz
[apple/xnu.git] / bsd / netinet / flow_divert.c
1 /*
2 * Copyright (c) 2012-2017 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28
29 #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 <libkern/tree.h>
46 #include <kern/locks.h>
47 #include <kern/debug.h>
48 #include <net/if_var.h>
49 #include <net/route.h>
50 #include <net/flowhash.h>
51 #include <net/ntstat.h>
52 #include <netinet/in.h>
53 #include <netinet/in_var.h>
54 #include <netinet/tcp.h>
55 #include <netinet/tcp_var.h>
56 #include <netinet/tcp_fsm.h>
57 #include <netinet/flow_divert.h>
58 #include <netinet/flow_divert_proto.h>
59 #if INET6
60 #include <netinet6/in6_pcb.h>
61 #include <netinet6/ip6protosw.h>
62 #endif /* INET6 */
63 #include <dev/random/randomdev.h>
64 #include <libkern/crypto/sha1.h>
65 #include <libkern/crypto/crypto_internal.h>
66 #include <os/log.h>
67
68 #define FLOW_DIVERT_CONNECT_STARTED 0x00000001
69 #define FLOW_DIVERT_READ_CLOSED 0x00000002
70 #define FLOW_DIVERT_WRITE_CLOSED 0x00000004
71 #define FLOW_DIVERT_TUNNEL_RD_CLOSED 0x00000008
72 #define FLOW_DIVERT_TUNNEL_WR_CLOSED 0x00000010
73 #define FLOW_DIVERT_TRANSFERRED 0x00000020
74 #define FLOW_DIVERT_HAS_HMAC 0x00000040
75
76 #define FDLOG(level, pcb, format, ...) \
77 os_log_with_type(OS_LOG_DEFAULT, flow_divert_syslog_type_to_oslog_type(level), "(%u): " format "\n", (pcb)->hash, __VA_ARGS__)
78
79 #define FDLOG0(level, pcb, msg) \
80 os_log_with_type(OS_LOG_DEFAULT, flow_divert_syslog_type_to_oslog_type(level), "(%u): " msg "\n", (pcb)->hash)
81
82 #define FDRETAIN(pcb) if ((pcb) != NULL) OSIncrementAtomic(&(pcb)->ref_count)
83 #define FDRELEASE(pcb) \
84 do { \
85 if ((pcb) != NULL && 1 == OSDecrementAtomic(&(pcb)->ref_count)) { \
86 flow_divert_pcb_destroy(pcb); \
87 } \
88 } while (0)
89
90 #define FDLOCK(pcb) lck_mtx_lock(&(pcb)->mtx)
91 #define FDUNLOCK(pcb) lck_mtx_unlock(&(pcb)->mtx)
92
93 #define FD_CTL_SENDBUFF_SIZE (128 * 1024)
94 #define FD_CTL_RCVBUFF_SIZE (128 * 1024)
95
96 #define GROUP_BIT_CTL_ENQUEUE_BLOCKED 0
97
98 #define GROUP_COUNT_MAX 32
99 #define FLOW_DIVERT_MAX_NAME_SIZE 4096
100 #define FLOW_DIVERT_MAX_KEY_SIZE 1024
101 #define FLOW_DIVERT_MAX_TRIE_MEMORY (1024 * 1024)
102
103 struct flow_divert_trie_node
104 {
105 uint16_t start;
106 uint16_t length;
107 uint16_t child_map;
108 };
109
110 #define CHILD_MAP_SIZE 256
111 #define NULL_TRIE_IDX 0xffff
112 #define TRIE_NODE(t, i) ((t)->nodes[(i)])
113 #define TRIE_CHILD(t, i, b) (((t)->child_maps + (CHILD_MAP_SIZE * TRIE_NODE(t, i).child_map))[(b)])
114 #define TRIE_BYTE(t, i) ((t)->bytes[(i)])
115
116 static struct flow_divert_pcb nil_pcb;
117
118 decl_lck_rw_data(static, g_flow_divert_group_lck);
119 static struct flow_divert_group **g_flow_divert_groups = NULL;
120 static uint32_t g_active_group_count = 0;
121
122 static lck_grp_attr_t *flow_divert_grp_attr = NULL;
123 static lck_attr_t *flow_divert_mtx_attr = NULL;
124 static lck_grp_t *flow_divert_mtx_grp = NULL;
125 static errno_t g_init_result = 0;
126
127 static kern_ctl_ref g_flow_divert_kctl_ref = NULL;
128
129 static struct protosw g_flow_divert_in_protosw;
130 static struct pr_usrreqs g_flow_divert_in_usrreqs;
131 static struct protosw g_flow_divert_in_udp_protosw;
132 static struct pr_usrreqs g_flow_divert_in_udp_usrreqs;
133 #if INET6
134 static struct ip6protosw g_flow_divert_in6_protosw;
135 static struct pr_usrreqs g_flow_divert_in6_usrreqs;
136 static struct ip6protosw g_flow_divert_in6_udp_protosw;
137 static struct pr_usrreqs g_flow_divert_in6_udp_usrreqs;
138 #endif /* INET6 */
139
140 static struct protosw *g_tcp_protosw = NULL;
141 static struct ip6protosw *g_tcp6_protosw = NULL;
142 static struct protosw *g_udp_protosw = NULL;
143 static struct ip6protosw *g_udp6_protosw = NULL;
144
145 static errno_t
146 flow_divert_dup_addr(sa_family_t family, struct sockaddr *addr, struct sockaddr **dup);
147
148 static errno_t
149 flow_divert_inp_to_sockaddr(const struct inpcb *inp, struct sockaddr **local_socket);
150
151 static boolean_t
152 flow_divert_is_sockaddr_valid(struct sockaddr *addr);
153
154 static int
155 flow_divert_append_target_endpoint_tlv(mbuf_t connect_packet, struct sockaddr *toaddr);
156
157 struct sockaddr *
158 flow_divert_get_buffered_target_address(mbuf_t buffer);
159
160 static boolean_t
161 flow_divert_has_pcb_local_address(const struct inpcb *inp);
162
163 static void
164 flow_divert_disconnect_socket(struct socket *so);
165
166 static inline uint8_t
167 flow_divert_syslog_type_to_oslog_type(int syslog_type)
168 {
169 switch (syslog_type) {
170 case LOG_ERR: return OS_LOG_TYPE_ERROR;
171 case LOG_INFO: return OS_LOG_TYPE_INFO;
172 case LOG_DEBUG: return OS_LOG_TYPE_DEBUG;
173 default: return OS_LOG_TYPE_DEFAULT;
174 }
175 }
176
177 static inline int
178 flow_divert_pcb_cmp(const struct flow_divert_pcb *pcb_a, const struct flow_divert_pcb *pcb_b)
179 {
180 return memcmp(&pcb_a->hash, &pcb_b->hash, sizeof(pcb_a->hash));
181 }
182
183 RB_PROTOTYPE(fd_pcb_tree, flow_divert_pcb, rb_link, flow_divert_pcb_cmp);
184 RB_GENERATE(fd_pcb_tree, flow_divert_pcb, rb_link, flow_divert_pcb_cmp);
185
186 static const char *
187 flow_divert_packet_type2str(uint8_t packet_type)
188 {
189 switch (packet_type) {
190 case FLOW_DIVERT_PKT_CONNECT:
191 return "connect";
192 case FLOW_DIVERT_PKT_CONNECT_RESULT:
193 return "connect result";
194 case FLOW_DIVERT_PKT_DATA:
195 return "data";
196 case FLOW_DIVERT_PKT_CLOSE:
197 return "close";
198 case FLOW_DIVERT_PKT_READ_NOTIFY:
199 return "read notification";
200 case FLOW_DIVERT_PKT_PROPERTIES_UPDATE:
201 return "properties update";
202 case FLOW_DIVERT_PKT_APP_MAP_CREATE:
203 return "app map create";
204 default:
205 return "unknown";
206 }
207 }
208
209 static struct flow_divert_pcb *
210 flow_divert_pcb_lookup(uint32_t hash, struct flow_divert_group *group)
211 {
212 struct flow_divert_pcb key_item;
213 struct flow_divert_pcb *fd_cb = NULL;
214
215 key_item.hash = hash;
216
217 lck_rw_lock_shared(&group->lck);
218 fd_cb = RB_FIND(fd_pcb_tree, &group->pcb_tree, &key_item);
219 FDRETAIN(fd_cb);
220 lck_rw_done(&group->lck);
221
222 return fd_cb;
223 }
224
225 static errno_t
226 flow_divert_pcb_insert(struct flow_divert_pcb *fd_cb, uint32_t ctl_unit)
227 {
228 errno_t error = 0;
229 struct flow_divert_pcb *exist = NULL;
230 struct flow_divert_group *group;
231 static uint32_t g_nextkey = 1;
232 static uint32_t g_hash_seed = 0;
233 int try_count = 0;
234
235 if (ctl_unit == 0 || ctl_unit >= GROUP_COUNT_MAX) {
236 return EINVAL;
237 }
238
239 socket_unlock(fd_cb->so, 0);
240 lck_rw_lock_shared(&g_flow_divert_group_lck);
241
242 if (g_flow_divert_groups == NULL || g_active_group_count == 0) {
243 FDLOG0(LOG_ERR, &nil_pcb, "No active groups, flow divert cannot be used for this socket");
244 error = ENETUNREACH;
245 goto done;
246 }
247
248 group = g_flow_divert_groups[ctl_unit];
249 if (group == NULL) {
250 FDLOG(LOG_ERR, &nil_pcb, "Group for control unit %u is NULL, flow divert cannot be used for this socket", ctl_unit);
251 error = ENETUNREACH;
252 goto done;
253 }
254
255 socket_lock(fd_cb->so, 0);
256
257 do {
258 uint32_t key[2];
259 uint32_t idx;
260
261 key[0] = g_nextkey++;
262 key[1] = RandomULong();
263
264 if (g_hash_seed == 0) {
265 g_hash_seed = RandomULong();
266 }
267
268 fd_cb->hash = net_flowhash(key, sizeof(key), g_hash_seed);
269
270 for (idx = 1; idx < GROUP_COUNT_MAX; idx++) {
271 struct flow_divert_group *curr_group = g_flow_divert_groups[idx];
272 if (curr_group != NULL && curr_group != group) {
273 lck_rw_lock_shared(&curr_group->lck);
274 exist = RB_FIND(fd_pcb_tree, &curr_group->pcb_tree, fd_cb);
275 lck_rw_done(&curr_group->lck);
276 if (exist != NULL) {
277 break;
278 }
279 }
280 }
281
282 if (exist == NULL) {
283 lck_rw_lock_exclusive(&group->lck);
284 exist = RB_INSERT(fd_pcb_tree, &group->pcb_tree, fd_cb);
285 lck_rw_done(&group->lck);
286 }
287 } while (exist != NULL && try_count++ < 3);
288
289 if (exist == NULL) {
290 fd_cb->group = group;
291 FDRETAIN(fd_cb); /* The group now has a reference */
292 } else {
293 fd_cb->hash = 0;
294 error = EEXIST;
295 }
296
297 socket_unlock(fd_cb->so, 0);
298
299 done:
300 lck_rw_done(&g_flow_divert_group_lck);
301 socket_lock(fd_cb->so, 0);
302
303 return error;
304 }
305
306 static struct flow_divert_pcb *
307 flow_divert_pcb_create(socket_t so)
308 {
309 struct flow_divert_pcb *new_pcb = NULL;
310
311 MALLOC_ZONE(new_pcb, struct flow_divert_pcb *, sizeof(*new_pcb), M_FLOW_DIVERT_PCB, M_WAITOK);
312 if (new_pcb == NULL) {
313 FDLOG0(LOG_ERR, &nil_pcb, "failed to allocate a pcb");
314 return NULL;
315 }
316
317 memset(new_pcb, 0, sizeof(*new_pcb));
318
319 lck_mtx_init(&new_pcb->mtx, flow_divert_mtx_grp, flow_divert_mtx_attr);
320 new_pcb->so = so;
321 new_pcb->log_level = nil_pcb.log_level;
322
323 FDRETAIN(new_pcb); /* Represents the socket's reference */
324
325 return new_pcb;
326 }
327
328 static void
329 flow_divert_pcb_destroy(struct flow_divert_pcb *fd_cb)
330 {
331 FDLOG(LOG_INFO, fd_cb, "Destroying, app tx %u, app rx %u, tunnel tx %u, tunnel rx %u",
332 fd_cb->bytes_written_by_app, fd_cb->bytes_read_by_app, fd_cb->bytes_sent, fd_cb->bytes_received);
333
334 if (fd_cb->local_address != NULL) {
335 FREE(fd_cb->local_address, M_SONAME);
336 }
337 if (fd_cb->remote_address != NULL) {
338 FREE(fd_cb->remote_address, M_SONAME);
339 }
340 if (fd_cb->connect_token != NULL) {
341 mbuf_freem(fd_cb->connect_token);
342 }
343 if (fd_cb->connect_packet != NULL) {
344 mbuf_freem(fd_cb->connect_packet);
345 }
346 if (fd_cb->app_data != NULL) {
347 FREE(fd_cb->app_data, M_TEMP);
348 }
349 FREE_ZONE(fd_cb, sizeof(*fd_cb), M_FLOW_DIVERT_PCB);
350 }
351
352 static void
353 flow_divert_pcb_remove(struct flow_divert_pcb *fd_cb)
354 {
355 if (fd_cb->group != NULL) {
356 struct flow_divert_group *group = fd_cb->group;
357 lck_rw_lock_exclusive(&group->lck);
358 FDLOG(LOG_INFO, fd_cb, "Removing from group %d, ref count = %d", group->ctl_unit, fd_cb->ref_count);
359 RB_REMOVE(fd_pcb_tree, &group->pcb_tree, fd_cb);
360 fd_cb->group = NULL;
361 FDRELEASE(fd_cb); /* Release the group's reference */
362 lck_rw_done(&group->lck);
363 }
364 }
365
366 static int
367 flow_divert_packet_init(struct flow_divert_pcb *fd_cb, uint8_t packet_type, mbuf_t *packet)
368 {
369 struct flow_divert_packet_header hdr;
370 int error = 0;
371
372 error = mbuf_gethdr(MBUF_DONTWAIT, MBUF_TYPE_HEADER, packet);
373 if (error) {
374 FDLOG(LOG_ERR, fd_cb, "failed to allocate the header mbuf: %d", error);
375 return error;
376 }
377
378 hdr.packet_type = packet_type;
379 hdr.conn_id = htonl(fd_cb->hash);
380
381 /* Lay down the header */
382 error = mbuf_copyback(*packet, 0, sizeof(hdr), &hdr, MBUF_DONTWAIT);
383 if (error) {
384 FDLOG(LOG_ERR, fd_cb, "mbuf_copyback(hdr) failed: %d", error);
385 mbuf_freem(*packet);
386 *packet = NULL;
387 return error;
388 }
389
390 return 0;
391 }
392
393 static int
394 flow_divert_packet_append_tlv(mbuf_t packet, uint8_t type, uint32_t length, const void *value)
395 {
396 uint32_t net_length = htonl(length);
397 int error = 0;
398
399 error = mbuf_copyback(packet, mbuf_pkthdr_len(packet), sizeof(type), &type, MBUF_DONTWAIT);
400 if (error) {
401 FDLOG(LOG_ERR, &nil_pcb, "failed to append the type (%d)", type);
402 return error;
403 }
404
405 error = mbuf_copyback(packet, mbuf_pkthdr_len(packet), sizeof(net_length), &net_length, MBUF_DONTWAIT);
406 if (error) {
407 FDLOG(LOG_ERR, &nil_pcb, "failed to append the length (%u)", length);
408 return error;
409 }
410
411 error = mbuf_copyback(packet, mbuf_pkthdr_len(packet), length, value, MBUF_DONTWAIT);
412 if (error) {
413 FDLOG0(LOG_ERR, &nil_pcb, "failed to append the value");
414 return error;
415 }
416
417 return error;
418 }
419
420 static int
421 flow_divert_packet_find_tlv(mbuf_t packet, int offset, uint8_t type, int *err, int next)
422 {
423 size_t cursor = offset;
424 int error = 0;
425 uint32_t curr_length;
426 uint8_t curr_type;
427
428 *err = 0;
429
430 do {
431 if (!next) {
432 error = mbuf_copydata(packet, cursor, sizeof(curr_type), &curr_type);
433 if (error) {
434 *err = ENOENT;
435 return -1;
436 }
437 } else {
438 next = 0;
439 curr_type = FLOW_DIVERT_TLV_NIL;
440 }
441
442 if (curr_type != type) {
443 cursor += sizeof(curr_type);
444 error = mbuf_copydata(packet, cursor, sizeof(curr_length), &curr_length);
445 if (error) {
446 *err = error;
447 return -1;
448 }
449
450 cursor += (sizeof(curr_length) + ntohl(curr_length));
451 }
452 } while (curr_type != type);
453
454 return cursor;
455 }
456
457 static int
458 flow_divert_packet_get_tlv(mbuf_t packet, int offset, uint8_t type, size_t buff_len, void *buff, uint32_t *val_size)
459 {
460 int error = 0;
461 uint32_t length;
462 int tlv_offset;
463
464 tlv_offset = flow_divert_packet_find_tlv(packet, offset, type, &error, 0);
465 if (tlv_offset < 0) {
466 return error;
467 }
468
469 error = mbuf_copydata(packet, tlv_offset + sizeof(type), sizeof(length), &length);
470 if (error) {
471 return error;
472 }
473
474 length = ntohl(length);
475
476 if (val_size != NULL) {
477 *val_size = length;
478 }
479
480 if (buff != NULL && buff_len > 0) {
481 size_t to_copy = (length < buff_len) ? length : buff_len;
482 error = mbuf_copydata(packet, tlv_offset + sizeof(type) + sizeof(length), to_copy, buff);
483 if (error) {
484 return error;
485 }
486 }
487
488 return 0;
489 }
490
491 static int
492 flow_divert_packet_compute_hmac(mbuf_t packet, struct flow_divert_group *group, uint8_t *hmac)
493 {
494 mbuf_t curr_mbuf = packet;
495
496 if (g_crypto_funcs == NULL || group->token_key == NULL) {
497 return ENOPROTOOPT;
498 }
499
500 cchmac_di_decl(g_crypto_funcs->ccsha1_di, hmac_ctx);
501 g_crypto_funcs->cchmac_init_fn(g_crypto_funcs->ccsha1_di, hmac_ctx, group->token_key_size, group->token_key);
502
503 while (curr_mbuf != NULL) {
504 g_crypto_funcs->cchmac_update_fn(g_crypto_funcs->ccsha1_di, hmac_ctx, mbuf_len(curr_mbuf), mbuf_data(curr_mbuf));
505 curr_mbuf = mbuf_next(curr_mbuf);
506 }
507
508 g_crypto_funcs->cchmac_final_fn(g_crypto_funcs->ccsha1_di, hmac_ctx, hmac);
509
510 return 0;
511 }
512
513 static int
514 flow_divert_packet_verify_hmac(mbuf_t packet, uint32_t ctl_unit)
515 {
516 int error = 0;
517 struct flow_divert_group *group = NULL;
518 int hmac_offset;
519 uint8_t packet_hmac[SHA_DIGEST_LENGTH];
520 uint8_t computed_hmac[SHA_DIGEST_LENGTH];
521 mbuf_t tail;
522
523 lck_rw_lock_shared(&g_flow_divert_group_lck);
524
525 if (g_flow_divert_groups != NULL && g_active_group_count > 0) {
526 group = g_flow_divert_groups[ctl_unit];
527 }
528
529 if (group == NULL) {
530 lck_rw_done(&g_flow_divert_group_lck);
531 return ENOPROTOOPT;
532 }
533
534 lck_rw_lock_shared(&group->lck);
535
536 if (group->token_key == NULL) {
537 error = ENOPROTOOPT;
538 goto done;
539 }
540
541 hmac_offset = flow_divert_packet_find_tlv(packet, 0, FLOW_DIVERT_TLV_HMAC, &error, 0);
542 if (hmac_offset < 0) {
543 goto done;
544 }
545
546 error = flow_divert_packet_get_tlv(packet, hmac_offset, FLOW_DIVERT_TLV_HMAC, sizeof(packet_hmac), packet_hmac, NULL);
547 if (error) {
548 goto done;
549 }
550
551 /* Chop off the HMAC TLV */
552 error = mbuf_split(packet, hmac_offset, MBUF_WAITOK, &tail);
553 if (error) {
554 goto done;
555 }
556
557 mbuf_free(tail);
558
559 error = flow_divert_packet_compute_hmac(packet, group, computed_hmac);
560 if (error) {
561 goto done;
562 }
563
564 if (memcmp(packet_hmac, computed_hmac, sizeof(packet_hmac))) {
565 FDLOG0(LOG_WARNING, &nil_pcb, "HMAC in token does not match computed HMAC");
566 error = EINVAL;
567 goto done;
568 }
569
570 done:
571 lck_rw_done(&group->lck);
572 lck_rw_done(&g_flow_divert_group_lck);
573 return error;
574 }
575
576 static void
577 flow_divert_add_data_statistics(struct flow_divert_pcb *fd_cb, int data_len, Boolean send)
578 {
579 struct inpcb *inp = NULL;
580 struct ifnet *ifp = NULL;
581 Boolean cell = FALSE;
582 Boolean wifi = FALSE;
583 Boolean wired = FALSE;
584
585 inp = sotoinpcb(fd_cb->so);
586 if (inp == NULL) {
587 return;
588 }
589
590 ifp = inp->inp_last_outifp;
591 if (ifp != NULL) {
592 cell = IFNET_IS_CELLULAR(ifp);
593 wifi = (!cell && IFNET_IS_WIFI(ifp));
594 wired = (!wifi && IFNET_IS_WIRED(ifp));
595 }
596
597 if (send) {
598 INP_ADD_STAT(inp, cell, wifi, wired, txpackets, 1);
599 INP_ADD_STAT(inp, cell, wifi, wired, txbytes, data_len);
600 } else {
601 INP_ADD_STAT(inp, cell, wifi, wired, rxpackets, 1);
602 INP_ADD_STAT(inp, cell, wifi, wired, rxbytes, data_len);
603 }
604 inp_set_activity_bitmap(inp);
605 }
606
607 static errno_t
608 flow_divert_check_no_cellular(struct flow_divert_pcb *fd_cb)
609 {
610 struct inpcb *inp = NULL;
611
612 inp = sotoinpcb(fd_cb->so);
613 if (inp && INP_NO_CELLULAR(inp) && inp->inp_last_outifp &&
614 IFNET_IS_CELLULAR(inp->inp_last_outifp))
615 return EHOSTUNREACH;
616
617 return 0;
618 }
619
620 static errno_t
621 flow_divert_check_no_expensive(struct flow_divert_pcb *fd_cb)
622 {
623 struct inpcb *inp = NULL;
624
625 inp = sotoinpcb(fd_cb->so);
626 if (inp && INP_NO_EXPENSIVE(inp) && inp->inp_last_outifp &&
627 IFNET_IS_EXPENSIVE(inp->inp_last_outifp))
628 return EHOSTUNREACH;
629
630 return 0;
631 }
632
633 static void
634 flow_divert_update_closed_state(struct flow_divert_pcb *fd_cb, int how, Boolean tunnel)
635 {
636 if (how != SHUT_RD) {
637 fd_cb->flags |= FLOW_DIVERT_WRITE_CLOSED;
638 if (tunnel || !(fd_cb->flags & FLOW_DIVERT_CONNECT_STARTED)) {
639 fd_cb->flags |= FLOW_DIVERT_TUNNEL_WR_CLOSED;
640 /* If the tunnel is not accepting writes any more, then flush the send buffer */
641 sbflush(&fd_cb->so->so_snd);
642 }
643 }
644 if (how != SHUT_WR) {
645 fd_cb->flags |= FLOW_DIVERT_READ_CLOSED;
646 if (tunnel || !(fd_cb->flags & FLOW_DIVERT_CONNECT_STARTED)) {
647 fd_cb->flags |= FLOW_DIVERT_TUNNEL_RD_CLOSED;
648 }
649 }
650 }
651
652 static uint16_t
653 trie_node_alloc(struct flow_divert_trie *trie)
654 {
655 if (trie->nodes_free_next < trie->nodes_count) {
656 uint16_t node_idx = trie->nodes_free_next++;
657 TRIE_NODE(trie, node_idx).child_map = NULL_TRIE_IDX;
658 return node_idx;
659 } else {
660 return NULL_TRIE_IDX;
661 }
662 }
663
664 static uint16_t
665 trie_child_map_alloc(struct flow_divert_trie *trie)
666 {
667 if (trie->child_maps_free_next < trie->child_maps_count) {
668 return trie->child_maps_free_next++;
669 } else {
670 return NULL_TRIE_IDX;
671 }
672 }
673
674 static uint16_t
675 trie_bytes_move(struct flow_divert_trie *trie, uint16_t bytes_idx, size_t bytes_size)
676 {
677 uint16_t start = trie->bytes_free_next;
678 if (start + bytes_size <= trie->bytes_count) {
679 if (start != bytes_idx) {
680 memmove(&TRIE_BYTE(trie, start), &TRIE_BYTE(trie, bytes_idx), bytes_size);
681 }
682 trie->bytes_free_next += bytes_size;
683 return start;
684 } else {
685 return NULL_TRIE_IDX;
686 }
687 }
688
689 static uint16_t
690 flow_divert_trie_insert(struct flow_divert_trie *trie, uint16_t string_start, size_t string_len)
691 {
692 uint16_t current = trie->root;
693 uint16_t child = trie->root;
694 uint16_t string_end = string_start + string_len;
695 uint16_t string_idx = string_start;
696 uint16_t string_remainder = string_len;
697
698 while (child != NULL_TRIE_IDX) {
699 uint16_t parent = current;
700 uint16_t node_idx;
701 uint16_t current_end;
702
703 current = child;
704 child = NULL_TRIE_IDX;
705
706 current_end = TRIE_NODE(trie, current).start + TRIE_NODE(trie, current).length;
707
708 for (node_idx = TRIE_NODE(trie, current).start;
709 node_idx < current_end &&
710 string_idx < string_end &&
711 TRIE_BYTE(trie, node_idx) == TRIE_BYTE(trie, string_idx);
712 node_idx++, string_idx++);
713
714 string_remainder = string_end - string_idx;
715
716 if (node_idx < (TRIE_NODE(trie, current).start + TRIE_NODE(trie, current).length)) {
717 /*
718 * We did not reach the end of the current node's string.
719 * We need to split the current node into two:
720 * 1. A new node that contains the prefix of the node that matches
721 * the prefix of the string being inserted.
722 * 2. The current node modified to point to the remainder
723 * of the current node's string.
724 */
725 uint16_t prefix = trie_node_alloc(trie);
726 if (prefix == NULL_TRIE_IDX) {
727 FDLOG0(LOG_ERR, &nil_pcb, "Ran out of trie nodes while splitting an existing node");
728 return NULL_TRIE_IDX;
729 }
730
731 /*
732 * Prefix points to the portion of the current nodes's string that has matched
733 * the input string thus far.
734 */
735 TRIE_NODE(trie, prefix).start = TRIE_NODE(trie, current).start;
736 TRIE_NODE(trie, prefix).length = (node_idx - TRIE_NODE(trie, current).start);
737
738 /*
739 * Prefix has the current node as the child corresponding to the first byte
740 * after the split.
741 */
742 TRIE_NODE(trie, prefix).child_map = trie_child_map_alloc(trie);
743 if (TRIE_NODE(trie, prefix).child_map == NULL_TRIE_IDX) {
744 FDLOG0(LOG_ERR, &nil_pcb, "Ran out of child maps while splitting an existing node");
745 return NULL_TRIE_IDX;
746 }
747 TRIE_CHILD(trie, prefix, TRIE_BYTE(trie, node_idx)) = current;
748
749 /* Parent has the prefix as the child correspoding to the first byte in the prefix */
750 TRIE_CHILD(trie, parent, TRIE_BYTE(trie, TRIE_NODE(trie, prefix).start)) = prefix;
751
752 /* Current node is adjusted to point to the remainder */
753 TRIE_NODE(trie, current).start = node_idx;
754 TRIE_NODE(trie, current).length -= TRIE_NODE(trie, prefix).length;
755
756 /* We want to insert the new leaf (if any) as a child of the prefix */
757 current = prefix;
758 }
759
760 if (string_remainder > 0) {
761 /*
762 * We still have bytes in the string that have not been matched yet.
763 * If the current node has children, iterate to the child corresponding
764 * to the next byte in the string.
765 */
766 if (TRIE_NODE(trie, current).child_map != NULL_TRIE_IDX) {
767 child = TRIE_CHILD(trie, current, TRIE_BYTE(trie, string_idx));
768 }
769 }
770 } /* while (child != NULL_TRIE_IDX) */
771
772 if (string_remainder > 0) {
773 /* Add a new leaf containing the remainder of the string */
774 uint16_t leaf = trie_node_alloc(trie);
775 if (leaf == NULL_TRIE_IDX) {
776 FDLOG0(LOG_ERR, &nil_pcb, "Ran out of trie nodes while inserting a new leaf");
777 return NULL_TRIE_IDX;
778 }
779
780 TRIE_NODE(trie, leaf).start = trie_bytes_move(trie, string_idx, string_remainder);
781 if (TRIE_NODE(trie, leaf).start == NULL_TRIE_IDX) {
782 FDLOG0(LOG_ERR, &nil_pcb, "Ran out of bytes while inserting a new leaf");
783 return NULL_TRIE_IDX;
784 }
785 TRIE_NODE(trie, leaf).length = string_remainder;
786
787 /* Set the new leaf as the child of the current node */
788 if (TRIE_NODE(trie, current).child_map == NULL_TRIE_IDX) {
789 TRIE_NODE(trie, current).child_map = trie_child_map_alloc(trie);
790 if (TRIE_NODE(trie, current).child_map == NULL_TRIE_IDX) {
791 FDLOG0(LOG_ERR, &nil_pcb, "Ran out of child maps while inserting a new leaf");
792 return NULL_TRIE_IDX;
793 }
794 }
795 TRIE_CHILD(trie, current, TRIE_BYTE(trie, TRIE_NODE(trie, leaf).start)) = leaf;
796 current = leaf;
797 } /* else duplicate or this string is a prefix of one of the existing strings */
798
799 return current;
800 }
801
802 #define APPLE_WEBCLIP_ID_PREFIX "com.apple.webapp"
803 static uint16_t
804 flow_divert_trie_search(struct flow_divert_trie *trie, uint8_t *string_bytes)
805 {
806 uint16_t current = trie->root;
807 uint16_t string_idx = 0;
808
809 while (current != NULL_TRIE_IDX) {
810 uint16_t next = NULL_TRIE_IDX;
811 uint16_t node_end = TRIE_NODE(trie, current).start + TRIE_NODE(trie, current).length;
812 uint16_t node_idx;
813
814 for (node_idx = TRIE_NODE(trie, current).start;
815 node_idx < node_end && string_bytes[string_idx] != '\0' && string_bytes[string_idx] == TRIE_BYTE(trie, node_idx);
816 node_idx++, string_idx++);
817
818 if (node_idx == node_end) {
819 if (string_bytes[string_idx] == '\0') {
820 return current; /* Got an exact match */
821 } else if (string_idx == strlen(APPLE_WEBCLIP_ID_PREFIX) &&
822 0 == strncmp((const char *)string_bytes, APPLE_WEBCLIP_ID_PREFIX, string_idx)) {
823 string_bytes[string_idx] = '\0';
824 return current; /* Got an apple webclip id prefix match */
825 } else if (TRIE_NODE(trie, current).child_map != NULL_TRIE_IDX) {
826 next = TRIE_CHILD(trie, current, string_bytes[string_idx]);
827 }
828 }
829 current = next;
830 }
831
832 return NULL_TRIE_IDX;
833 }
834
835 struct uuid_search_info {
836 uuid_t target_uuid;
837 char *found_signing_id;
838 boolean_t found_multiple_signing_ids;
839 proc_t found_proc;
840 };
841
842 static int
843 flow_divert_find_proc_by_uuid_callout(proc_t p, void *arg)
844 {
845 struct uuid_search_info *info = (struct uuid_search_info *)arg;
846 int result = PROC_RETURNED_DONE; /* By default, we didn't find the process */
847
848 if (info->found_signing_id != NULL) {
849 if (!info->found_multiple_signing_ids) {
850 /* All processes that were found had the same signing identifier, so just claim this first one and be done. */
851 info->found_proc = p;
852 result = PROC_CLAIMED_DONE;
853 } else {
854 uuid_string_t uuid_str;
855 uuid_unparse(info->target_uuid, uuid_str);
856 FDLOG(LOG_WARNING, &nil_pcb, "Found multiple processes with UUID %s with different signing identifiers", uuid_str);
857 }
858 FREE(info->found_signing_id, M_TEMP);
859 info->found_signing_id = NULL;
860 }
861
862 if (result == PROC_RETURNED_DONE) {
863 uuid_string_t uuid_str;
864 uuid_unparse(info->target_uuid, uuid_str);
865 FDLOG(LOG_WARNING, &nil_pcb, "Failed to find a process with UUID %s", uuid_str);
866 }
867
868 return result;
869 }
870
871 static int
872 flow_divert_find_proc_by_uuid_filter(proc_t p, void *arg)
873 {
874 struct uuid_search_info *info = (struct uuid_search_info *)arg;
875 int include = 0;
876
877 if (info->found_multiple_signing_ids) {
878 return include;
879 }
880
881 include = (uuid_compare(p->p_uuid, info->target_uuid) == 0);
882 if (include) {
883 const char *signing_id = cs_identity_get(p);
884 if (signing_id != NULL) {
885 FDLOG(LOG_INFO, &nil_pcb, "Found process %d with signing identifier %s", p->p_pid, signing_id);
886 size_t signing_id_size = strlen(signing_id) + 1;
887 if (info->found_signing_id == NULL) {
888 MALLOC(info->found_signing_id, char *, signing_id_size, M_TEMP, M_WAITOK);
889 memcpy(info->found_signing_id, signing_id, signing_id_size);
890 } else if (memcmp(signing_id, info->found_signing_id, signing_id_size)) {
891 info->found_multiple_signing_ids = TRUE;
892 }
893 } else {
894 info->found_multiple_signing_ids = TRUE;
895 }
896 include = !info->found_multiple_signing_ids;
897 }
898
899 return include;
900 }
901
902 static proc_t
903 flow_divert_find_proc_by_uuid(uuid_t uuid)
904 {
905 struct uuid_search_info info;
906
907 if (LOG_INFO <= nil_pcb.log_level) {
908 uuid_string_t uuid_str;
909 uuid_unparse(uuid, uuid_str);
910 FDLOG(LOG_INFO, &nil_pcb, "Looking for process with UUID %s", uuid_str);
911 }
912
913 memset(&info, 0, sizeof(info));
914 info.found_proc = PROC_NULL;
915 uuid_copy(info.target_uuid, uuid);
916
917 proc_iterate(PROC_ALLPROCLIST, flow_divert_find_proc_by_uuid_callout, &info, flow_divert_find_proc_by_uuid_filter, &info);
918
919 return info.found_proc;
920 }
921
922 static int
923 flow_divert_get_src_proc(struct socket *so, proc_t *proc)
924 {
925 int release = 0;
926
927 if (so->so_flags & SOF_DELEGATED) {
928 if ((*proc)->p_pid != so->e_pid) {
929 *proc = proc_find(so->e_pid);
930 release = 1;
931 } else if (uuid_compare((*proc)->p_uuid, so->e_uuid)) {
932 *proc = flow_divert_find_proc_by_uuid(so->e_uuid);
933 release = 1;
934 }
935 } else if (*proc == PROC_NULL) {
936 *proc = current_proc();
937 }
938
939 if (*proc != PROC_NULL) {
940 if ((*proc)->p_pid == 0) {
941 if (release) {
942 proc_rele(*proc);
943 }
944 release = 0;
945 *proc = PROC_NULL;
946 }
947 }
948
949 return release;
950 }
951
952 static int
953 flow_divert_send_packet(struct flow_divert_pcb *fd_cb, mbuf_t packet, Boolean enqueue)
954 {
955 int error;
956
957 if (fd_cb->group == NULL) {
958 fd_cb->so->so_error = ECONNABORTED;
959 flow_divert_disconnect_socket(fd_cb->so);
960 return ECONNABORTED;
961 }
962
963 lck_rw_lock_shared(&fd_cb->group->lck);
964
965 if (MBUFQ_EMPTY(&fd_cb->group->send_queue)) {
966 error = ctl_enqueuembuf(g_flow_divert_kctl_ref, fd_cb->group->ctl_unit, packet, CTL_DATA_EOR);
967 } else {
968 error = ENOBUFS;
969 }
970
971 if (error == ENOBUFS) {
972 if (enqueue) {
973 if (!lck_rw_lock_shared_to_exclusive(&fd_cb->group->lck)) {
974 lck_rw_lock_exclusive(&fd_cb->group->lck);
975 }
976 MBUFQ_ENQUEUE(&fd_cb->group->send_queue, packet);
977 error = 0;
978 }
979 OSTestAndSet(GROUP_BIT_CTL_ENQUEUE_BLOCKED, &fd_cb->group->atomic_bits);
980 }
981
982 lck_rw_done(&fd_cb->group->lck);
983
984 return error;
985 }
986
987 static int
988 flow_divert_create_connect_packet(struct flow_divert_pcb *fd_cb, struct sockaddr *to, struct socket *so, proc_t p, mbuf_t *out_connect_packet)
989 {
990 int error = 0;
991 int flow_type = 0;
992 char *signing_id = NULL;
993 int free_signing_id = 0;
994 mbuf_t connect_packet = NULL;
995 proc_t src_proc = p;
996 int release_proc = 0;
997
998 error = flow_divert_packet_init(fd_cb, FLOW_DIVERT_PKT_CONNECT, &connect_packet);
999 if (error) {
1000 goto done;
1001 }
1002
1003 error = EPERM;
1004
1005 if (fd_cb->connect_token != NULL && (fd_cb->flags & FLOW_DIVERT_HAS_HMAC)) {
1006 uint32_t sid_size = 0;
1007 int find_error = flow_divert_packet_get_tlv(fd_cb->connect_token, 0, FLOW_DIVERT_TLV_SIGNING_ID, 0, NULL, &sid_size);
1008 if (find_error == 0 && sid_size > 0) {
1009 MALLOC(signing_id, char *, sid_size + 1, M_TEMP, M_WAITOK | M_ZERO);
1010 if (signing_id != NULL) {
1011 flow_divert_packet_get_tlv(fd_cb->connect_token, 0, FLOW_DIVERT_TLV_SIGNING_ID, sid_size, signing_id, NULL);
1012 FDLOG(LOG_INFO, fd_cb, "Got %s from token", signing_id);
1013 free_signing_id = 1;
1014 }
1015 }
1016 }
1017
1018 socket_unlock(so, 0);
1019
1020 if (signing_id == NULL) {
1021 release_proc = flow_divert_get_src_proc(so, &src_proc);
1022 if (src_proc != PROC_NULL) {
1023 proc_lock(src_proc);
1024 if (src_proc->p_csflags & (CS_VALID|CS_DEBUGGED)) {
1025 const char * cs_id;
1026 cs_id = cs_identity_get(src_proc);
1027 signing_id = __DECONST(char *, cs_id);
1028 } else {
1029 FDLOG0(LOG_WARNING, fd_cb, "Signature is invalid");
1030 }
1031 } else {
1032 FDLOG0(LOG_WARNING, fd_cb, "Failed to determine the current proc");
1033 }
1034 } else {
1035 src_proc = PROC_NULL;
1036 }
1037
1038 if (signing_id != NULL) {
1039 uint16_t result = NULL_TRIE_IDX;
1040 lck_rw_lock_shared(&fd_cb->group->lck);
1041 if (fd_cb->group->flags & FLOW_DIVERT_GROUP_FLAG_NO_APP_MAP) {
1042 result = 1;
1043 } else {
1044 result = flow_divert_trie_search(&fd_cb->group->signing_id_trie, (uint8_t *)signing_id);
1045 }
1046 lck_rw_done(&fd_cb->group->lck);
1047 if (result != NULL_TRIE_IDX) {
1048 error = 0;
1049 FDLOG(LOG_INFO, fd_cb, "%s matched", signing_id);
1050
1051 error = flow_divert_packet_append_tlv(connect_packet, FLOW_DIVERT_TLV_SIGNING_ID, strlen(signing_id), signing_id);
1052 if (error == 0) {
1053 if (src_proc != PROC_NULL) {
1054 unsigned char cdhash[SHA1_RESULTLEN];
1055 error = proc_getcdhash(src_proc, cdhash);
1056 if (error == 0) {
1057 error = flow_divert_packet_append_tlv(connect_packet, FLOW_DIVERT_TLV_CDHASH, sizeof(cdhash), cdhash);
1058 if (error) {
1059 FDLOG(LOG_ERR, fd_cb, "failed to append the cdhash: %d", error);
1060 }
1061 } else {
1062 FDLOG(LOG_ERR, fd_cb, "failed to get the cdhash: %d", error);
1063 }
1064 }
1065 } else {
1066 FDLOG(LOG_ERR, fd_cb, "failed to append the signing ID: %d", error);
1067 }
1068 } else {
1069 FDLOG(LOG_WARNING, fd_cb, "%s did not match", signing_id);
1070 }
1071 } else {
1072 FDLOG0(LOG_WARNING, fd_cb, "Failed to get the code signing identity");
1073 if (fd_cb->group->flags & FLOW_DIVERT_GROUP_FLAG_NO_APP_MAP) {
1074 error = 0;
1075 }
1076 }
1077
1078 if (src_proc != PROC_NULL) {
1079 proc_unlock(src_proc);
1080 if (release_proc) {
1081 proc_rele(src_proc);
1082 }
1083 }
1084 socket_lock(so, 0);
1085
1086 if (free_signing_id) {
1087 FREE(signing_id, M_TEMP);
1088 }
1089
1090 if (error) {
1091 goto done;
1092 }
1093
1094 error = flow_divert_packet_append_tlv(connect_packet,
1095 FLOW_DIVERT_TLV_TRAFFIC_CLASS,
1096 sizeof(fd_cb->so->so_traffic_class),
1097 &fd_cb->so->so_traffic_class);
1098 if (error) {
1099 goto done;
1100 }
1101
1102 if (SOCK_TYPE(fd_cb->so) == SOCK_STREAM) {
1103 flow_type = FLOW_DIVERT_FLOW_TYPE_TCP;
1104 } else if (SOCK_TYPE(fd_cb->so) == SOCK_DGRAM) {
1105 flow_type = FLOW_DIVERT_FLOW_TYPE_UDP;
1106 } else {
1107 error = EINVAL;
1108 goto done;
1109 }
1110 error = flow_divert_packet_append_tlv(connect_packet,
1111 FLOW_DIVERT_TLV_FLOW_TYPE,
1112 sizeof(flow_type),
1113 &flow_type);
1114
1115 if (error) {
1116 goto done;
1117 }
1118
1119 if (fd_cb->so->so_flags & SOF_DELEGATED) {
1120 error = flow_divert_packet_append_tlv(connect_packet,
1121 FLOW_DIVERT_TLV_PID,
1122 sizeof(fd_cb->so->e_pid),
1123 &fd_cb->so->e_pid);
1124 if (error) {
1125 goto done;
1126 }
1127
1128 error = flow_divert_packet_append_tlv(connect_packet,
1129 FLOW_DIVERT_TLV_UUID,
1130 sizeof(fd_cb->so->e_uuid),
1131 &fd_cb->so->e_uuid);
1132 if (error) {
1133 goto done;
1134 }
1135 } else {
1136 error = flow_divert_packet_append_tlv(connect_packet,
1137 FLOW_DIVERT_TLV_PID,
1138 sizeof(fd_cb->so->e_pid),
1139 &fd_cb->so->last_pid);
1140 if (error) {
1141 goto done;
1142 }
1143
1144 error = flow_divert_packet_append_tlv(connect_packet,
1145 FLOW_DIVERT_TLV_UUID,
1146 sizeof(fd_cb->so->e_uuid),
1147 &fd_cb->so->last_uuid);
1148 if (error) {
1149 goto done;
1150 }
1151 }
1152
1153 if (fd_cb->connect_token != NULL) {
1154 unsigned int token_len = m_length(fd_cb->connect_token);
1155 mbuf_concatenate(connect_packet, fd_cb->connect_token);
1156 mbuf_pkthdr_adjustlen(connect_packet, token_len);
1157 fd_cb->connect_token = NULL;
1158 } else {
1159 uint32_t ctl_unit = htonl(fd_cb->control_group_unit);
1160
1161 error = flow_divert_packet_append_tlv(connect_packet, FLOW_DIVERT_TLV_CTL_UNIT, sizeof(ctl_unit), &ctl_unit);
1162 if (error) {
1163 goto done;
1164 }
1165
1166 error = flow_divert_append_target_endpoint_tlv(connect_packet, to);
1167 if (error) {
1168 goto done;
1169 }
1170 }
1171
1172 if (fd_cb->local_address != NULL) {
1173 error = EALREADY;
1174 goto done;
1175 } else {
1176 struct inpcb *inp = sotoinpcb(so);
1177 if (flow_divert_has_pcb_local_address(inp)) {
1178 error = flow_divert_inp_to_sockaddr(inp, &fd_cb->local_address);
1179 if (error) {
1180 FDLOG0(LOG_ERR, fd_cb, "failed to get the local socket address.");
1181 goto done;
1182 }
1183 }
1184 }
1185
1186 if (fd_cb->local_address != NULL) {
1187 /* socket is bound. */
1188 error = flow_divert_packet_append_tlv(connect_packet, FLOW_DIVERT_TLV_LOCAL_ADDR,
1189 sizeof(struct sockaddr_storage), fd_cb->local_address);
1190 if (error) {
1191 goto done;
1192 }
1193 }
1194
1195 if (so->so_flags1 & SOF1_DATA_IDEMPOTENT) {
1196 uint32_t flags = FLOW_DIVERT_TOKEN_FLAG_TFO;
1197 error = flow_divert_packet_append_tlv(connect_packet, FLOW_DIVERT_TLV_FLAGS, sizeof(flags), &flags);
1198 if (error) {
1199 goto done;
1200 }
1201 }
1202
1203 done:
1204 if (!error) {
1205 *out_connect_packet = connect_packet;
1206 } else if (connect_packet != NULL) {
1207 mbuf_freem(connect_packet);
1208 }
1209
1210 return error;
1211 }
1212
1213 static int
1214 flow_divert_send_connect_result(struct flow_divert_pcb *fd_cb)
1215 {
1216 int error = 0;
1217 mbuf_t packet = NULL;
1218 int rbuff_space = 0;
1219
1220 error = flow_divert_packet_init(fd_cb, FLOW_DIVERT_PKT_CONNECT_RESULT, &packet);
1221 if (error) {
1222 FDLOG(LOG_ERR, fd_cb, "failed to create a connect result packet: %d", error);
1223 goto done;
1224 }
1225
1226 rbuff_space = fd_cb->so->so_rcv.sb_hiwat;
1227 if (rbuff_space < 0) {
1228 rbuff_space = 0;
1229 }
1230 rbuff_space = htonl(rbuff_space);
1231 error = flow_divert_packet_append_tlv(packet,
1232 FLOW_DIVERT_TLV_SPACE_AVAILABLE,
1233 sizeof(rbuff_space),
1234 &rbuff_space);
1235 if (error) {
1236 goto done;
1237 }
1238
1239 error = flow_divert_send_packet(fd_cb, packet, TRUE);
1240 if (error) {
1241 goto done;
1242 }
1243
1244 done:
1245 if (error && packet != NULL) {
1246 mbuf_freem(packet);
1247 }
1248
1249 return error;
1250 }
1251
1252 static int
1253 flow_divert_send_close(struct flow_divert_pcb *fd_cb, int how)
1254 {
1255 int error = 0;
1256 mbuf_t packet = NULL;
1257 uint32_t zero = 0;
1258
1259 error = flow_divert_packet_init(fd_cb, FLOW_DIVERT_PKT_CLOSE, &packet);
1260 if (error) {
1261 FDLOG(LOG_ERR, fd_cb, "failed to create a close packet: %d", error);
1262 goto done;
1263 }
1264
1265 error = flow_divert_packet_append_tlv(packet, FLOW_DIVERT_TLV_ERROR_CODE, sizeof(zero), &zero);
1266 if (error) {
1267 FDLOG(LOG_ERR, fd_cb, "failed to add the error code TLV: %d", error);
1268 goto done;
1269 }
1270
1271 how = htonl(how);
1272 error = flow_divert_packet_append_tlv(packet, FLOW_DIVERT_TLV_HOW, sizeof(how), &how);
1273 if (error) {
1274 FDLOG(LOG_ERR, fd_cb, "failed to add the how flag: %d", error);
1275 goto done;
1276 }
1277
1278 error = flow_divert_send_packet(fd_cb, packet, TRUE);
1279 if (error) {
1280 goto done;
1281 }
1282
1283 done:
1284 if (error && packet != NULL) {
1285 mbuf_free(packet);
1286 }
1287
1288 return error;
1289 }
1290
1291 static int
1292 flow_divert_tunnel_how_closed(struct flow_divert_pcb *fd_cb)
1293 {
1294 if ((fd_cb->flags & (FLOW_DIVERT_TUNNEL_RD_CLOSED|FLOW_DIVERT_TUNNEL_WR_CLOSED)) ==
1295 (FLOW_DIVERT_TUNNEL_RD_CLOSED|FLOW_DIVERT_TUNNEL_WR_CLOSED))
1296 {
1297 return SHUT_RDWR;
1298 } else if (fd_cb->flags & FLOW_DIVERT_TUNNEL_RD_CLOSED) {
1299 return SHUT_RD;
1300 } else if (fd_cb->flags & FLOW_DIVERT_TUNNEL_WR_CLOSED) {
1301 return SHUT_WR;
1302 }
1303
1304 return -1;
1305 }
1306
1307 /*
1308 * Determine what close messages if any need to be sent to the tunnel. Returns TRUE if the tunnel is closed for both reads and
1309 * writes. Returns FALSE otherwise.
1310 */
1311 static void
1312 flow_divert_send_close_if_needed(struct flow_divert_pcb *fd_cb)
1313 {
1314 int how = -1;
1315
1316 /* Do not send any close messages if there is still data in the send buffer */
1317 if (fd_cb->so->so_snd.sb_cc == 0) {
1318 if ((fd_cb->flags & (FLOW_DIVERT_READ_CLOSED|FLOW_DIVERT_TUNNEL_RD_CLOSED)) == FLOW_DIVERT_READ_CLOSED) {
1319 /* Socket closed reads, but tunnel did not. Tell tunnel to close reads */
1320 how = SHUT_RD;
1321 }
1322 if ((fd_cb->flags & (FLOW_DIVERT_WRITE_CLOSED|FLOW_DIVERT_TUNNEL_WR_CLOSED)) == FLOW_DIVERT_WRITE_CLOSED) {
1323 /* Socket closed writes, but tunnel did not. Tell tunnel to close writes */
1324 if (how == SHUT_RD) {
1325 how = SHUT_RDWR;
1326 } else {
1327 how = SHUT_WR;
1328 }
1329 }
1330 }
1331
1332 if (how != -1) {
1333 FDLOG(LOG_INFO, fd_cb, "sending close, how = %d", how);
1334 if (flow_divert_send_close(fd_cb, how) != ENOBUFS) {
1335 /* Successfully sent the close packet. Record the ways in which the tunnel has been closed */
1336 if (how != SHUT_RD) {
1337 fd_cb->flags |= FLOW_DIVERT_TUNNEL_WR_CLOSED;
1338 }
1339 if (how != SHUT_WR) {
1340 fd_cb->flags |= FLOW_DIVERT_TUNNEL_RD_CLOSED;
1341 }
1342 }
1343 }
1344
1345 if (flow_divert_tunnel_how_closed(fd_cb) == SHUT_RDWR) {
1346 flow_divert_disconnect_socket(fd_cb->so);
1347 }
1348 }
1349
1350 static errno_t
1351 flow_divert_send_data_packet(struct flow_divert_pcb *fd_cb, mbuf_t data, size_t data_len, struct sockaddr *toaddr, Boolean force)
1352 {
1353 mbuf_t packet;
1354 mbuf_t last;
1355 int error = 0;
1356
1357 error = flow_divert_packet_init(fd_cb, FLOW_DIVERT_PKT_DATA, &packet);
1358 if (error) {
1359 FDLOG(LOG_ERR, fd_cb, "flow_divert_packet_init failed: %d", error);
1360 return error;
1361 }
1362
1363 if (toaddr != NULL) {
1364 error = flow_divert_append_target_endpoint_tlv(packet, toaddr);
1365 if (error) {
1366 FDLOG(LOG_ERR, fd_cb, "flow_divert_append_target_endpoint_tlv() failed: %d", error);
1367 return error;
1368 }
1369 }
1370
1371 if (data_len > 0 && data != NULL) {
1372 last = m_last(packet);
1373 mbuf_setnext(last, data);
1374 mbuf_pkthdr_adjustlen(packet, data_len);
1375 }
1376 error = flow_divert_send_packet(fd_cb, packet, force);
1377
1378 if (error) {
1379 mbuf_setnext(last, NULL);
1380 mbuf_freem(packet);
1381 } else {
1382 fd_cb->bytes_sent += data_len;
1383 flow_divert_add_data_statistics(fd_cb, data_len, TRUE);
1384 }
1385
1386 return error;
1387 }
1388
1389 static void
1390 flow_divert_send_buffered_data(struct flow_divert_pcb *fd_cb, Boolean force)
1391 {
1392 size_t to_send;
1393 size_t sent = 0;
1394 int error = 0;
1395 mbuf_t buffer;
1396
1397 to_send = fd_cb->so->so_snd.sb_cc;
1398 buffer = fd_cb->so->so_snd.sb_mb;
1399
1400 if (buffer == NULL && to_send > 0) {
1401 FDLOG(LOG_ERR, fd_cb, "Send buffer is NULL, but size is supposed to be %lu", to_send);
1402 return;
1403 }
1404
1405 /* Ignore the send window if force is enabled */
1406 if (!force && (to_send > fd_cb->send_window)) {
1407 to_send = fd_cb->send_window;
1408 }
1409
1410 if (SOCK_TYPE(fd_cb->so) == SOCK_STREAM) {
1411 while (sent < to_send) {
1412 mbuf_t data;
1413 size_t data_len;
1414
1415 data_len = to_send - sent;
1416 if (data_len > FLOW_DIVERT_CHUNK_SIZE) {
1417 data_len = FLOW_DIVERT_CHUNK_SIZE;
1418 }
1419
1420 error = mbuf_copym(buffer, sent, data_len, MBUF_DONTWAIT, &data);
1421 if (error) {
1422 FDLOG(LOG_ERR, fd_cb, "mbuf_copym failed: %d", error);
1423 break;
1424 }
1425
1426 error = flow_divert_send_data_packet(fd_cb, data, data_len, NULL, force);
1427 if (error) {
1428 mbuf_freem(data);
1429 break;
1430 }
1431
1432 sent += data_len;
1433 }
1434 sbdrop(&fd_cb->so->so_snd, sent);
1435 sowwakeup(fd_cb->so);
1436 } else if (SOCK_TYPE(fd_cb->so) == SOCK_DGRAM) {
1437 mbuf_t data;
1438 mbuf_t m;
1439 size_t data_len;
1440
1441 while(buffer) {
1442 struct sockaddr *toaddr = flow_divert_get_buffered_target_address(buffer);
1443
1444 m = buffer;
1445 if (toaddr != NULL) {
1446 /* look for data in the chain */
1447 do {
1448 m = m->m_next;
1449 if (m != NULL && m->m_type == MT_DATA) {
1450 break;
1451 }
1452 } while(m);
1453 if (m == NULL) {
1454 /* unexpected */
1455 FDLOG0(LOG_ERR, fd_cb, "failed to find type MT_DATA in the mbuf chain.");
1456 goto move_on;
1457 }
1458 }
1459 data_len = mbuf_pkthdr_len(m);
1460 if (data_len > 0) {
1461 FDLOG(LOG_DEBUG, fd_cb, "mbuf_copym() data_len = %lu", data_len);
1462 error = mbuf_copym(m, 0, data_len, MBUF_DONTWAIT, &data);
1463 if (error) {
1464 FDLOG(LOG_ERR, fd_cb, "mbuf_copym failed: %d", error);
1465 break;
1466 }
1467 } else {
1468 data = NULL;
1469 }
1470 error = flow_divert_send_data_packet(fd_cb, data, data_len, toaddr, force);
1471 if (error) {
1472 mbuf_freem(data);
1473 break;
1474 }
1475 sent += data_len;
1476 move_on:
1477 buffer = buffer->m_nextpkt;
1478 (void) sbdroprecord(&(fd_cb->so->so_snd));
1479 }
1480 }
1481
1482 if (sent > 0) {
1483 FDLOG(LOG_DEBUG, fd_cb, "sent %lu bytes of buffered data", sent);
1484 if (fd_cb->send_window >= sent) {
1485 fd_cb->send_window -= sent;
1486 } else {
1487 fd_cb->send_window = 0;
1488 }
1489 }
1490 }
1491
1492 static int
1493 flow_divert_send_app_data(struct flow_divert_pcb *fd_cb, mbuf_t data, struct sockaddr *toaddr)
1494 {
1495 size_t to_send = mbuf_pkthdr_len(data);
1496 int error = 0;
1497
1498 if (to_send > fd_cb->send_window) {
1499 to_send = fd_cb->send_window;
1500 }
1501
1502 if (fd_cb->so->so_snd.sb_cc > 0) {
1503 to_send = 0; /* If the send buffer is non-empty, then we can't send anything */
1504 }
1505
1506 if (SOCK_TYPE(fd_cb->so) == SOCK_STREAM) {
1507 size_t sent = 0;
1508 mbuf_t remaining_data = data;
1509 mbuf_t pkt_data = NULL;
1510 while (sent < to_send && remaining_data != NULL) {
1511 size_t pkt_data_len;
1512
1513 pkt_data = remaining_data;
1514
1515 if ((to_send - sent) > FLOW_DIVERT_CHUNK_SIZE) {
1516 pkt_data_len = FLOW_DIVERT_CHUNK_SIZE;
1517 } else {
1518 pkt_data_len = to_send - sent;
1519 }
1520
1521 if (pkt_data_len < mbuf_pkthdr_len(pkt_data)) {
1522 error = mbuf_split(pkt_data, pkt_data_len, MBUF_DONTWAIT, &remaining_data);
1523 if (error) {
1524 FDLOG(LOG_ERR, fd_cb, "mbuf_split failed: %d", error);
1525 pkt_data = NULL;
1526 break;
1527 }
1528 } else {
1529 remaining_data = NULL;
1530 }
1531
1532 error = flow_divert_send_data_packet(fd_cb, pkt_data, pkt_data_len, NULL, FALSE);
1533
1534 if (error) {
1535 break;
1536 }
1537
1538 pkt_data = NULL;
1539 sent += pkt_data_len;
1540 }
1541
1542 fd_cb->send_window -= sent;
1543
1544 error = 0;
1545
1546 if (pkt_data != NULL) {
1547 if (sbspace(&fd_cb->so->so_snd) > 0) {
1548 if (!sbappendstream(&fd_cb->so->so_snd, pkt_data)) {
1549 FDLOG(LOG_ERR, fd_cb, "sbappendstream failed with pkt_data, send buffer size = %u, send_window = %u\n",
1550 fd_cb->so->so_snd.sb_cc, fd_cb->send_window);
1551 }
1552 } else {
1553 error = ENOBUFS;
1554 }
1555 }
1556
1557 if (remaining_data != NULL) {
1558 if (sbspace(&fd_cb->so->so_snd) > 0) {
1559 if (!sbappendstream(&fd_cb->so->so_snd, remaining_data)) {
1560 FDLOG(LOG_ERR, fd_cb, "sbappendstream failed with remaining_data, send buffer size = %u, send_window = %u\n",
1561 fd_cb->so->so_snd.sb_cc, fd_cb->send_window);
1562 }
1563 } else {
1564 error = ENOBUFS;
1565 }
1566 }
1567 } else if (SOCK_TYPE(fd_cb->so) == SOCK_DGRAM) {
1568 if (to_send || mbuf_pkthdr_len(data) == 0) {
1569 error = flow_divert_send_data_packet(fd_cb, data, to_send, toaddr, FALSE);
1570 if (error) {
1571 FDLOG(LOG_ERR, fd_cb, "flow_divert_send_data_packet failed. send data size = %lu", to_send);
1572 } else {
1573 fd_cb->send_window -= to_send;
1574 }
1575 } else {
1576 /* buffer it */
1577 if (sbspace(&fd_cb->so->so_snd) >= (int)mbuf_pkthdr_len(data)) {
1578 if (toaddr != NULL) {
1579 if (!sbappendaddr(&fd_cb->so->so_snd, toaddr, data, NULL, &error)) {
1580 FDLOG(LOG_ERR, fd_cb,
1581 "sbappendaddr failed. send buffer size = %u, send_window = %u, error = %d\n",
1582 fd_cb->so->so_snd.sb_cc, fd_cb->send_window, error);
1583 }
1584 } else {
1585 if (!sbappendrecord(&fd_cb->so->so_snd, data)) {
1586 FDLOG(LOG_ERR, fd_cb,
1587 "sbappendrecord failed. send buffer size = %u, send_window = %u, error = %d\n",
1588 fd_cb->so->so_snd.sb_cc, fd_cb->send_window, error);
1589 }
1590 }
1591 } else {
1592 error = ENOBUFS;
1593 }
1594 }
1595 }
1596
1597 return error;
1598 }
1599
1600 static int
1601 flow_divert_send_read_notification(struct flow_divert_pcb *fd_cb, uint32_t read_count)
1602 {
1603 int error = 0;
1604 mbuf_t packet = NULL;
1605 uint32_t net_read_count = htonl(read_count);
1606
1607 error = flow_divert_packet_init(fd_cb, FLOW_DIVERT_PKT_READ_NOTIFY, &packet);
1608 if (error) {
1609 FDLOG(LOG_ERR, fd_cb, "failed to create a read notification packet: %d", error);
1610 goto done;
1611 }
1612
1613 error = flow_divert_packet_append_tlv(packet, FLOW_DIVERT_TLV_READ_COUNT, sizeof(net_read_count), &net_read_count);
1614 if (error) {
1615 FDLOG(LOG_ERR, fd_cb, "failed to add the read count: %d", error);
1616 goto done;
1617 }
1618
1619 error = flow_divert_send_packet(fd_cb, packet, TRUE);
1620 if (error) {
1621 goto done;
1622 }
1623
1624 done:
1625 if (error && packet != NULL) {
1626 mbuf_free(packet);
1627 }
1628
1629 return error;
1630 }
1631
1632 static int
1633 flow_divert_send_traffic_class_update(struct flow_divert_pcb *fd_cb, int traffic_class)
1634 {
1635 int error = 0;
1636 mbuf_t packet = NULL;
1637
1638 error = flow_divert_packet_init(fd_cb, FLOW_DIVERT_PKT_PROPERTIES_UPDATE, &packet);
1639 if (error) {
1640 FDLOG(LOG_ERR, fd_cb, "failed to create a properties update packet: %d", error);
1641 goto done;
1642 }
1643
1644 error = flow_divert_packet_append_tlv(packet, FLOW_DIVERT_TLV_TRAFFIC_CLASS, sizeof(traffic_class), &traffic_class);
1645 if (error) {
1646 FDLOG(LOG_ERR, fd_cb, "failed to add the traffic class: %d", error);
1647 goto done;
1648 }
1649
1650 error = flow_divert_send_packet(fd_cb, packet, TRUE);
1651 if (error) {
1652 goto done;
1653 }
1654
1655 done:
1656 if (error && packet != NULL) {
1657 mbuf_free(packet);
1658 }
1659
1660 return error;
1661 }
1662
1663 static void
1664 flow_divert_handle_connect_result(struct flow_divert_pcb *fd_cb, mbuf_t packet, int offset)
1665 {
1666 uint32_t connect_error;
1667 uint32_t ctl_unit = 0;
1668 int error = 0;
1669 struct flow_divert_group *grp = NULL;
1670 struct sockaddr_storage local_address;
1671 int out_if_index = 0;
1672 struct sockaddr_storage remote_address;
1673 uint32_t send_window;
1674 uint32_t app_data_length = 0;
1675
1676 memset(&local_address, 0, sizeof(local_address));
1677 memset(&remote_address, 0, sizeof(remote_address));
1678
1679 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_ERROR_CODE, sizeof(connect_error), &connect_error, NULL);
1680 if (error) {
1681 FDLOG(LOG_ERR, fd_cb, "failed to get the connect result: %d", error);
1682 return;
1683 }
1684
1685 FDLOG(LOG_INFO, fd_cb, "received connect result %u", connect_error);
1686
1687 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_SPACE_AVAILABLE, sizeof(send_window), &send_window, NULL);
1688 if (error) {
1689 FDLOG(LOG_ERR, fd_cb, "failed to get the send window: %d", error);
1690 return;
1691 }
1692
1693 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_CTL_UNIT, sizeof(ctl_unit), &ctl_unit, NULL);
1694 if (error) {
1695 FDLOG0(LOG_INFO, fd_cb, "No control unit provided in the connect result");
1696 }
1697
1698 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_LOCAL_ADDR, sizeof(local_address), &local_address, NULL);
1699 if (error) {
1700 FDLOG0(LOG_INFO, fd_cb, "No local address provided");
1701 }
1702
1703 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_REMOTE_ADDR, sizeof(remote_address), &remote_address, NULL);
1704 if (error) {
1705 FDLOG0(LOG_INFO, fd_cb, "No remote address provided");
1706 }
1707
1708 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_OUT_IF_INDEX, sizeof(out_if_index), &out_if_index, NULL);
1709 if (error) {
1710 FDLOG0(LOG_INFO, fd_cb, "No output if index provided");
1711 }
1712
1713 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_APP_DATA, 0, NULL, &app_data_length);
1714 if (error) {
1715 FDLOG0(LOG_INFO, fd_cb, "No application data provided in connect result");
1716 }
1717
1718 error = 0;
1719 connect_error = ntohl(connect_error);
1720 ctl_unit = ntohl(ctl_unit);
1721
1722 lck_rw_lock_shared(&g_flow_divert_group_lck);
1723
1724 if (connect_error == 0 && ctl_unit > 0) {
1725 if (ctl_unit >= GROUP_COUNT_MAX) {
1726 FDLOG(LOG_ERR, fd_cb, "Connect result contains an invalid control unit: %u", ctl_unit);
1727 error = EINVAL;
1728 } else if (g_flow_divert_groups == NULL || g_active_group_count == 0) {
1729 FDLOG0(LOG_ERR, fd_cb, "No active groups, dropping connection");
1730 error = EINVAL;
1731 } else {
1732 grp = g_flow_divert_groups[ctl_unit];
1733 if (grp == NULL) {
1734 error = ECONNRESET;
1735 }
1736 }
1737 }
1738
1739 FDLOCK(fd_cb);
1740 if (fd_cb->so != NULL) {
1741 struct inpcb *inp = NULL;
1742 struct ifnet *ifp = NULL;
1743 struct flow_divert_group *old_group;
1744
1745 socket_lock(fd_cb->so, 0);
1746
1747 if (!(fd_cb->so->so_state & SS_ISCONNECTING)) {
1748 goto done;
1749 }
1750
1751 inp = sotoinpcb(fd_cb->so);
1752
1753 if (connect_error || error) {
1754 goto set_socket_state;
1755 }
1756
1757 if (local_address.ss_family == 0 && fd_cb->local_address == NULL) {
1758 error = EINVAL;
1759 goto set_socket_state;
1760 }
1761 if (local_address.ss_family != 0 && fd_cb->local_address == NULL) {
1762 if (local_address.ss_len > sizeof(local_address)) {
1763 local_address.ss_len = sizeof(local_address);
1764 }
1765 fd_cb->local_address = dup_sockaddr((struct sockaddr *)&local_address, 1);
1766 }
1767
1768 if (remote_address.ss_family != 0) {
1769 if (remote_address.ss_len > sizeof(remote_address)) {
1770 remote_address.ss_len = sizeof(remote_address);
1771 }
1772 fd_cb->remote_address = dup_sockaddr((struct sockaddr *)&remote_address, 1);
1773 } else {
1774 error = EINVAL;
1775 goto set_socket_state;
1776 }
1777
1778 if (app_data_length > 0) {
1779 uint8_t *app_data = NULL;
1780 MALLOC(app_data, uint8_t *, app_data_length, M_TEMP, M_WAITOK);
1781 if (app_data != NULL) {
1782 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_APP_DATA, app_data_length, app_data, NULL);
1783 if (error == 0) {
1784 FDLOG(LOG_INFO, fd_cb, "Got %u bytes of app data from the connect result", app_data_length);
1785 if (fd_cb->app_data != NULL) {
1786 FREE(fd_cb->app_data, M_TEMP);
1787 }
1788 fd_cb->app_data = app_data;
1789 fd_cb->app_data_length = app_data_length;
1790 } else {
1791 FDLOG(LOG_ERR, fd_cb, "Failed to copy %u bytes of application data from the connect result packet", app_data_length);
1792 FREE(app_data, M_TEMP);
1793 }
1794 } else {
1795 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);
1796 }
1797 }
1798
1799 ifnet_head_lock_shared();
1800 if (out_if_index > 0 && out_if_index <= if_index) {
1801 ifp = ifindex2ifnet[out_if_index];
1802 }
1803
1804 if (ifp != NULL) {
1805 inp->inp_last_outifp = ifp;
1806 } else {
1807 error = EINVAL;
1808 }
1809 ifnet_head_done();
1810
1811 if (error) {
1812 goto set_socket_state;
1813 }
1814
1815 if (fd_cb->group == NULL) {
1816 error = EINVAL;
1817 goto set_socket_state;
1818 }
1819
1820 if (grp != NULL) {
1821 old_group = fd_cb->group;
1822
1823 lck_rw_lock_exclusive(&old_group->lck);
1824 lck_rw_lock_exclusive(&grp->lck);
1825
1826 RB_REMOVE(fd_pcb_tree, &old_group->pcb_tree, fd_cb);
1827 if (RB_INSERT(fd_pcb_tree, &grp->pcb_tree, fd_cb) != NULL) {
1828 panic("group with unit %u already contains a connection with hash %u", grp->ctl_unit, fd_cb->hash);
1829 }
1830
1831 fd_cb->group = grp;
1832
1833 lck_rw_done(&grp->lck);
1834 lck_rw_done(&old_group->lck);
1835 }
1836
1837 fd_cb->send_window = ntohl(send_window);
1838
1839 set_socket_state:
1840 if (!connect_error && !error) {
1841 FDLOG0(LOG_INFO, fd_cb, "sending connect result");
1842 error = flow_divert_send_connect_result(fd_cb);
1843 }
1844
1845 if (connect_error || error) {
1846 if (!connect_error) {
1847 flow_divert_update_closed_state(fd_cb, SHUT_RDWR, FALSE);
1848 fd_cb->so->so_error = error;
1849 flow_divert_send_close_if_needed(fd_cb);
1850 } else {
1851 flow_divert_update_closed_state(fd_cb, SHUT_RDWR, TRUE);
1852 fd_cb->so->so_error = connect_error;
1853 }
1854 flow_divert_disconnect_socket(fd_cb->so);
1855 } else {
1856 flow_divert_send_buffered_data(fd_cb, FALSE);
1857 soisconnected(fd_cb->so);
1858 }
1859
1860 done:
1861 socket_unlock(fd_cb->so, 0);
1862 }
1863 FDUNLOCK(fd_cb);
1864
1865 lck_rw_done(&g_flow_divert_group_lck);
1866 }
1867
1868 static void
1869 flow_divert_handle_close(struct flow_divert_pcb *fd_cb, mbuf_t packet, int offset)
1870 {
1871 uint32_t close_error;
1872 int error = 0;
1873 int how;
1874
1875 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_ERROR_CODE, sizeof(close_error), &close_error, NULL);
1876 if (error) {
1877 FDLOG(LOG_ERR, fd_cb, "failed to get the close error: %d", error);
1878 return;
1879 }
1880
1881 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_HOW, sizeof(how), &how, NULL);
1882 if (error) {
1883 FDLOG(LOG_ERR, fd_cb, "failed to get the close how flag: %d", error);
1884 return;
1885 }
1886
1887 how = ntohl(how);
1888
1889 FDLOG(LOG_INFO, fd_cb, "close received, how = %d", how);
1890
1891 FDLOCK(fd_cb);
1892 if (fd_cb->so != NULL) {
1893 socket_lock(fd_cb->so, 0);
1894
1895 fd_cb->so->so_error = ntohl(close_error);
1896
1897 flow_divert_update_closed_state(fd_cb, how, TRUE);
1898
1899 how = flow_divert_tunnel_how_closed(fd_cb);
1900 if (how == SHUT_RDWR) {
1901 flow_divert_disconnect_socket(fd_cb->so);
1902 } else if (how == SHUT_RD) {
1903 socantrcvmore(fd_cb->so);
1904 } else if (how == SHUT_WR) {
1905 socantsendmore(fd_cb->so);
1906 }
1907
1908 socket_unlock(fd_cb->so, 0);
1909 }
1910 FDUNLOCK(fd_cb);
1911 }
1912
1913 static mbuf_t
1914 flow_divert_get_control_mbuf(struct flow_divert_pcb *fd_cb)
1915 {
1916 struct inpcb *inp = sotoinpcb(fd_cb->so);
1917 if (inp->inp_vflag & INP_IPV4 && inp->inp_flags & INP_RECVDSTADDR) {
1918 struct sockaddr_in *sin = (struct sockaddr_in *)(void *)fd_cb->local_address;
1919
1920 return sbcreatecontrol((caddr_t) &sin->sin_addr, sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1921 } else if (inp->inp_vflag & INP_IPV6 && (inp->inp_flags & IN6P_PKTINFO) != 0) {
1922 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)(void *)fd_cb->local_address;
1923 struct in6_pktinfo pi6;
1924
1925 bcopy(&sin6->sin6_addr, &pi6.ipi6_addr, sizeof (struct in6_addr));
1926 pi6.ipi6_ifindex = 0;
1927 return sbcreatecontrol((caddr_t)&pi6, sizeof (struct in6_pktinfo), IPV6_PKTINFO, IPPROTO_IPV6);
1928 }
1929 return (NULL);
1930 }
1931
1932 static void
1933 flow_divert_handle_data(struct flow_divert_pcb *fd_cb, mbuf_t packet, size_t offset)
1934 {
1935 FDLOCK(fd_cb);
1936 if (fd_cb->so != NULL) {
1937 int error = 0;
1938 mbuf_t data = NULL;
1939 size_t data_size;
1940 struct sockaddr_storage remote_address;
1941 boolean_t got_remote_sa = FALSE;
1942
1943 socket_lock(fd_cb->so, 0);
1944
1945 if (SOCK_TYPE(fd_cb->so) == SOCK_DGRAM) {
1946 uint32_t val_size = 0;
1947
1948 /* check if we got remote address with data */
1949 memset(&remote_address, 0, sizeof(remote_address));
1950 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_REMOTE_ADDR, sizeof(remote_address), &remote_address, &val_size);
1951 if (error || val_size > sizeof(remote_address)) {
1952 FDLOG0(LOG_INFO, fd_cb, "No remote address provided");
1953 error = 0;
1954 } else {
1955 /* validate the address */
1956 if (flow_divert_is_sockaddr_valid((struct sockaddr *)&remote_address)) {
1957 got_remote_sa = TRUE;
1958 }
1959 offset += (sizeof(uint8_t) + sizeof(uint32_t) + val_size);
1960 }
1961 }
1962
1963 data_size = (mbuf_pkthdr_len(packet) - offset);
1964
1965 FDLOG(LOG_DEBUG, fd_cb, "received %lu bytes of data", data_size);
1966
1967 error = mbuf_split(packet, offset, MBUF_DONTWAIT, &data);
1968 if (error || data == NULL) {
1969 FDLOG(LOG_ERR, fd_cb, "mbuf_split failed: %d", error);
1970 } else {
1971 if (flow_divert_check_no_cellular(fd_cb) ||
1972 flow_divert_check_no_expensive(fd_cb))
1973 {
1974 flow_divert_update_closed_state(fd_cb, SHUT_RDWR, TRUE);
1975 flow_divert_send_close(fd_cb, SHUT_RDWR);
1976 flow_divert_disconnect_socket(fd_cb->so);
1977 } else if (!(fd_cb->so->so_state & SS_CANTRCVMORE)) {
1978 if (SOCK_TYPE(fd_cb->so) == SOCK_STREAM) {
1979 if (sbappendstream(&fd_cb->so->so_rcv, data)) {
1980 fd_cb->bytes_received += data_size;
1981 flow_divert_add_data_statistics(fd_cb, data_size, FALSE);
1982 fd_cb->sb_size = fd_cb->so->so_rcv.sb_cc;
1983 sorwakeup(fd_cb->so);
1984 data = NULL;
1985 } else {
1986 FDLOG0(LOG_ERR, fd_cb, "received data, but appendstream failed");
1987 }
1988 } else if (SOCK_TYPE(fd_cb->so) == SOCK_DGRAM) {
1989 struct sockaddr *append_sa;
1990 mbuf_t mctl;
1991
1992 if (got_remote_sa == TRUE) {
1993 error = flow_divert_dup_addr(fd_cb->so->so_proto->pr_domain->dom_family,
1994 (struct sockaddr *)&remote_address, &append_sa);
1995 } else {
1996 error = flow_divert_dup_addr(fd_cb->so->so_proto->pr_domain->dom_family,
1997 fd_cb->remote_address, &append_sa);
1998 }
1999 if (error) {
2000 FDLOG0(LOG_ERR, fd_cb, "failed to dup the socket address.");
2001 }
2002
2003 mctl = flow_divert_get_control_mbuf(fd_cb);
2004 if (sbappendaddr(&fd_cb->so->so_rcv, append_sa, data, mctl, NULL)) {
2005 fd_cb->bytes_received += data_size;
2006 flow_divert_add_data_statistics(fd_cb, data_size, FALSE);
2007 fd_cb->sb_size = fd_cb->so->so_rcv.sb_cc;
2008 sorwakeup(fd_cb->so);
2009 data = NULL;
2010 } else {
2011 FDLOG0(LOG_ERR, fd_cb, "received data, but sbappendaddr failed");
2012 }
2013 if (!error) {
2014 FREE(append_sa, M_TEMP);
2015 }
2016 }
2017 }
2018 }
2019 socket_unlock(fd_cb->so, 0);
2020
2021 if (data != NULL) {
2022 mbuf_freem(data);
2023 }
2024 }
2025 FDUNLOCK(fd_cb);
2026 }
2027
2028 static void
2029 flow_divert_handle_read_notification(struct flow_divert_pcb *fd_cb, mbuf_t packet, int offset)
2030 {
2031 uint32_t read_count;
2032 int error = 0;
2033
2034 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_READ_COUNT, sizeof(read_count), &read_count, NULL);
2035 if (error) {
2036 FDLOG(LOG_ERR, fd_cb, "failed to get the read count: %d", error);
2037 return;
2038 }
2039
2040 FDLOG(LOG_DEBUG, fd_cb, "received a read notification for %u bytes", ntohl(read_count));
2041
2042 FDLOCK(fd_cb);
2043 if (fd_cb->so != NULL) {
2044 socket_lock(fd_cb->so, 0);
2045 fd_cb->send_window += ntohl(read_count);
2046 flow_divert_send_buffered_data(fd_cb, FALSE);
2047 socket_unlock(fd_cb->so, 0);
2048 }
2049 FDUNLOCK(fd_cb);
2050 }
2051
2052 static void
2053 flow_divert_handle_group_init(struct flow_divert_group *group, mbuf_t packet, int offset)
2054 {
2055 int error = 0;
2056 uint32_t key_size = 0;
2057 int log_level;
2058 uint32_t flags = 0;
2059
2060 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_TOKEN_KEY, 0, NULL, &key_size);
2061 if (error) {
2062 FDLOG(LOG_ERR, &nil_pcb, "failed to get the key size: %d", error);
2063 return;
2064 }
2065
2066 if (key_size == 0 || key_size > FLOW_DIVERT_MAX_KEY_SIZE) {
2067 FDLOG(LOG_ERR, &nil_pcb, "Invalid key size: %u", key_size);
2068 return;
2069 }
2070
2071 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_LOG_LEVEL, sizeof(log_level), &log_level, NULL);
2072 if (!error) {
2073 nil_pcb.log_level = log_level;
2074 }
2075
2076 lck_rw_lock_exclusive(&group->lck);
2077
2078 MALLOC(group->token_key, uint8_t *, key_size, M_TEMP, M_WAITOK);
2079 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_TOKEN_KEY, key_size, group->token_key, NULL);
2080 if (error) {
2081 FDLOG(LOG_ERR, &nil_pcb, "failed to get the token key: %d", error);
2082 FREE(group->token_key, M_TEMP);
2083 group->token_key = NULL;
2084 lck_rw_done(&group->lck);
2085 return;
2086 }
2087
2088 group->token_key_size = key_size;
2089
2090 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_FLAGS, sizeof(flags), &flags, NULL);
2091 if (!error) {
2092 group->flags = flags;
2093 }
2094
2095 lck_rw_done(&group->lck);
2096 }
2097
2098 static void
2099 flow_divert_handle_properties_update(struct flow_divert_pcb *fd_cb, mbuf_t packet, int offset)
2100 {
2101 int error = 0;
2102 struct sockaddr_storage local_address;
2103 int out_if_index = 0;
2104 struct sockaddr_storage remote_address;
2105 uint32_t app_data_length = 0;
2106
2107 FDLOG0(LOG_INFO, fd_cb, "received a properties update");
2108
2109 memset(&local_address, 0, sizeof(local_address));
2110 memset(&remote_address, 0, sizeof(remote_address));
2111
2112 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_LOCAL_ADDR, sizeof(local_address), &local_address, NULL);
2113 if (error) {
2114 FDLOG0(LOG_INFO, fd_cb, "No local address provided in properties update");
2115 }
2116
2117 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_REMOTE_ADDR, sizeof(remote_address), &remote_address, NULL);
2118 if (error) {
2119 FDLOG0(LOG_INFO, fd_cb, "No remote address provided in properties update");
2120 }
2121
2122 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_OUT_IF_INDEX, sizeof(out_if_index), &out_if_index, NULL);
2123 if (error) {
2124 FDLOG0(LOG_INFO, fd_cb, "No output if index provided in properties update");
2125 }
2126
2127 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_APP_DATA, 0, NULL, &app_data_length);
2128 if (error) {
2129 FDLOG0(LOG_INFO, fd_cb, "No application data provided in properties update");
2130 }
2131
2132 FDLOCK(fd_cb);
2133 if (fd_cb->so != NULL) {
2134 socket_lock(fd_cb->so, 0);
2135
2136 if (local_address.ss_family != 0) {
2137 if (local_address.ss_len > sizeof(local_address)) {
2138 local_address.ss_len = sizeof(local_address);
2139 }
2140 if (fd_cb->local_address != NULL) {
2141 FREE(fd_cb->local_address, M_SONAME);
2142 fd_cb->local_address = NULL;
2143 }
2144 fd_cb->local_address = dup_sockaddr((struct sockaddr *)&local_address, 1);
2145 }
2146
2147 if (remote_address.ss_family != 0) {
2148 if (remote_address.ss_len > sizeof(remote_address)) {
2149 remote_address.ss_len = sizeof(remote_address);
2150 }
2151 if (fd_cb->remote_address != NULL) {
2152 FREE(fd_cb->remote_address, M_SONAME);
2153 fd_cb->remote_address = NULL;
2154 }
2155 fd_cb->remote_address = dup_sockaddr((struct sockaddr *)&remote_address, 1);
2156 }
2157
2158 if (out_if_index > 0) {
2159 struct inpcb *inp = NULL;
2160 struct ifnet *ifp = NULL;
2161
2162 inp = sotoinpcb(fd_cb->so);
2163
2164 ifnet_head_lock_shared();
2165 if (out_if_index <= if_index) {
2166 ifp = ifindex2ifnet[out_if_index];
2167 }
2168
2169 if (ifp != NULL) {
2170 inp->inp_last_outifp = ifp;
2171 }
2172 ifnet_head_done();
2173 }
2174
2175 if (app_data_length > 0) {
2176 uint8_t *app_data = NULL;
2177 MALLOC(app_data, uint8_t *, app_data_length, M_TEMP, M_WAITOK);
2178 if (app_data != NULL) {
2179 error = flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_APP_DATA, app_data_length, app_data, NULL);
2180 if (error == 0) {
2181 if (fd_cb->app_data != NULL) {
2182 FREE(fd_cb->app_data, M_TEMP);
2183 }
2184 fd_cb->app_data = app_data;
2185 fd_cb->app_data_length = app_data_length;
2186 } else {
2187 FDLOG(LOG_ERR, fd_cb, "Failed to copy %u bytes of application data from the properties update packet", app_data_length);
2188 FREE(app_data, M_TEMP);
2189 }
2190 } else {
2191 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);
2192 }
2193 }
2194
2195 socket_unlock(fd_cb->so, 0);
2196 }
2197 FDUNLOCK(fd_cb);
2198 }
2199
2200 static void
2201 flow_divert_handle_app_map_create(struct flow_divert_group *group, mbuf_t packet, int offset)
2202 {
2203 size_t bytes_mem_size;
2204 size_t child_maps_mem_size;
2205 int cursor;
2206 int error = 0;
2207 struct flow_divert_trie new_trie;
2208 int insert_error = 0;
2209 size_t nodes_mem_size;
2210 int prefix_count = 0;
2211 int signing_id_count = 0;
2212 size_t trie_memory_size = 0;
2213
2214 lck_rw_lock_exclusive(&group->lck);
2215
2216 /* Re-set the current trie */
2217 if (group->signing_id_trie.memory != NULL) {
2218 FREE(group->signing_id_trie.memory, M_TEMP);
2219 }
2220 memset(&group->signing_id_trie, 0, sizeof(group->signing_id_trie));
2221 group->signing_id_trie.root = NULL_TRIE_IDX;
2222
2223 memset(&new_trie, 0, sizeof(new_trie));
2224
2225 /* Get the number of shared prefixes in the new set of signing ID strings */
2226 flow_divert_packet_get_tlv(packet, offset, FLOW_DIVERT_TLV_PREFIX_COUNT, sizeof(prefix_count), &prefix_count, NULL);
2227
2228 if (prefix_count < 0) {
2229 lck_rw_done(&group->lck);
2230 return;
2231 }
2232
2233 /* Compute the number of signing IDs and the total amount of bytes needed to store them */
2234 for (cursor = flow_divert_packet_find_tlv(packet, offset, FLOW_DIVERT_TLV_SIGNING_ID, &error, 0);
2235 cursor >= 0;
2236 cursor = flow_divert_packet_find_tlv(packet, cursor, FLOW_DIVERT_TLV_SIGNING_ID, &error, 1))
2237 {
2238 uint32_t sid_size = 0;
2239 flow_divert_packet_get_tlv(packet, cursor, FLOW_DIVERT_TLV_SIGNING_ID, 0, NULL, &sid_size);
2240 new_trie.bytes_count += sid_size;
2241 signing_id_count++;
2242 }
2243
2244 if (signing_id_count == 0) {
2245 lck_rw_done(&group->lck);
2246 return;
2247 }
2248
2249 new_trie.nodes_count = (prefix_count + signing_id_count + 1); /* + 1 for the root node */
2250 new_trie.child_maps_count = (prefix_count + 1); /* + 1 for the root node */
2251
2252 FDLOG(LOG_INFO, &nil_pcb, "Nodes count = %lu, child maps count = %lu, bytes_count = %lu",
2253 new_trie.nodes_count, new_trie.child_maps_count, new_trie.bytes_count);
2254
2255 nodes_mem_size = (sizeof(*new_trie.nodes) * new_trie.nodes_count);
2256 child_maps_mem_size = (sizeof(*new_trie.child_maps) * CHILD_MAP_SIZE * new_trie.child_maps_count);
2257 bytes_mem_size = (sizeof(*new_trie.bytes) * new_trie.bytes_count);
2258
2259 trie_memory_size = nodes_mem_size + child_maps_mem_size + bytes_mem_size;
2260 if (trie_memory_size > FLOW_DIVERT_MAX_TRIE_MEMORY) {
2261 FDLOG(LOG_ERR, &nil_pcb, "Trie memory size (%lu) is too big (maximum is %u)", trie_memory_size, FLOW_DIVERT_MAX_TRIE_MEMORY);
2262 lck_rw_done(&group->lck);
2263 return;
2264 }
2265
2266 MALLOC(new_trie.memory, void *, trie_memory_size, M_TEMP, M_WAITOK);
2267 if (new_trie.memory == NULL) {
2268 FDLOG(LOG_ERR, &nil_pcb, "Failed to allocate %lu bytes of memory for the signing ID trie",
2269 nodes_mem_size + child_maps_mem_size + bytes_mem_size);
2270 lck_rw_done(&group->lck);
2271 return;
2272 }
2273
2274 /* Initialize the free lists */
2275 new_trie.nodes = (struct flow_divert_trie_node *)new_trie.memory;
2276 new_trie.nodes_free_next = 0;
2277 memset(new_trie.nodes, 0, nodes_mem_size);
2278
2279 new_trie.child_maps = (uint16_t *)(void *)((uint8_t *)new_trie.memory + nodes_mem_size);
2280 new_trie.child_maps_free_next = 0;
2281 memset(new_trie.child_maps, 0xff, child_maps_mem_size);
2282
2283 new_trie.bytes = (uint8_t *)(void *)((uint8_t *)new_trie.memory + nodes_mem_size + child_maps_mem_size);
2284 new_trie.bytes_free_next = 0;
2285
2286 /* The root is an empty node */
2287 new_trie.root = trie_node_alloc(&new_trie);
2288
2289 /* Add each signing ID to the trie */
2290 for (cursor = flow_divert_packet_find_tlv(packet, offset, FLOW_DIVERT_TLV_SIGNING_ID, &error, 0);
2291 cursor >= 0;
2292 cursor = flow_divert_packet_find_tlv(packet, cursor, FLOW_DIVERT_TLV_SIGNING_ID, &error, 1))
2293 {
2294 uint32_t sid_size = 0;
2295 flow_divert_packet_get_tlv(packet, cursor, FLOW_DIVERT_TLV_SIGNING_ID, 0, NULL, &sid_size);
2296 if (new_trie.bytes_free_next + sid_size <= new_trie.bytes_count) {
2297 uint16_t new_node_idx;
2298 flow_divert_packet_get_tlv(packet, cursor, FLOW_DIVERT_TLV_SIGNING_ID, sid_size, &TRIE_BYTE(&new_trie, new_trie.bytes_free_next), NULL);
2299 new_node_idx = flow_divert_trie_insert(&new_trie, new_trie.bytes_free_next, sid_size);
2300 if (new_node_idx == NULL_TRIE_IDX) {
2301 insert_error = EINVAL;
2302 break;
2303 }
2304 } else {
2305 FDLOG0(LOG_ERR, &nil_pcb, "No place to put signing ID for insertion");
2306 insert_error = ENOBUFS;
2307 break;
2308 }
2309 }
2310
2311 if (!insert_error) {
2312 group->signing_id_trie = new_trie;
2313 } else {
2314 FREE(new_trie.memory, M_TEMP);
2315 }
2316
2317 lck_rw_done(&group->lck);
2318 }
2319
2320 static int
2321 flow_divert_input(mbuf_t packet, struct flow_divert_group *group)
2322 {
2323 struct flow_divert_packet_header hdr;
2324 int error = 0;
2325 struct flow_divert_pcb *fd_cb;
2326
2327 if (mbuf_pkthdr_len(packet) < sizeof(hdr)) {
2328 FDLOG(LOG_ERR, &nil_pcb, "got a bad packet, length (%lu) < sizeof hdr (%lu)", mbuf_pkthdr_len(packet), sizeof(hdr));
2329 error = EINVAL;
2330 goto done;
2331 }
2332
2333 if (mbuf_pkthdr_len(packet) > FD_CTL_RCVBUFF_SIZE) {
2334 FDLOG(LOG_ERR, &nil_pcb, "got a bad packet, length (%lu) > %d", mbuf_pkthdr_len(packet), FD_CTL_RCVBUFF_SIZE);
2335 error = EINVAL;
2336 goto done;
2337 }
2338
2339 error = mbuf_copydata(packet, 0, sizeof(hdr), &hdr);
2340 if (error) {
2341 FDLOG(LOG_ERR, &nil_pcb, "mbuf_copydata failed for the header: %d", error);
2342 error = ENOBUFS;
2343 goto done;
2344 }
2345
2346 hdr.conn_id = ntohl(hdr.conn_id);
2347
2348 if (hdr.conn_id == 0) {
2349 switch (hdr.packet_type) {
2350 case FLOW_DIVERT_PKT_GROUP_INIT:
2351 flow_divert_handle_group_init(group, packet, sizeof(hdr));
2352 break;
2353 case FLOW_DIVERT_PKT_APP_MAP_CREATE:
2354 flow_divert_handle_app_map_create(group, packet, sizeof(hdr));
2355 break;
2356 default:
2357 FDLOG(LOG_WARNING, &nil_pcb, "got an unknown message type: %d", hdr.packet_type);
2358 break;
2359 }
2360 goto done;
2361 }
2362
2363 fd_cb = flow_divert_pcb_lookup(hdr.conn_id, group); /* This retains the PCB */
2364 if (fd_cb == NULL) {
2365 if (hdr.packet_type != FLOW_DIVERT_PKT_CLOSE && hdr.packet_type != FLOW_DIVERT_PKT_READ_NOTIFY) {
2366 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);
2367 }
2368 goto done;
2369 }
2370
2371 switch (hdr.packet_type) {
2372 case FLOW_DIVERT_PKT_CONNECT_RESULT:
2373 flow_divert_handle_connect_result(fd_cb, packet, sizeof(hdr));
2374 break;
2375 case FLOW_DIVERT_PKT_CLOSE:
2376 flow_divert_handle_close(fd_cb, packet, sizeof(hdr));
2377 break;
2378 case FLOW_DIVERT_PKT_DATA:
2379 flow_divert_handle_data(fd_cb, packet, sizeof(hdr));
2380 break;
2381 case FLOW_DIVERT_PKT_READ_NOTIFY:
2382 flow_divert_handle_read_notification(fd_cb, packet, sizeof(hdr));
2383 break;
2384 case FLOW_DIVERT_PKT_PROPERTIES_UPDATE:
2385 flow_divert_handle_properties_update(fd_cb, packet, sizeof(hdr));
2386 break;
2387 default:
2388 FDLOG(LOG_WARNING, fd_cb, "got an unknown message type: %d", hdr.packet_type);
2389 break;
2390 }
2391
2392 FDRELEASE(fd_cb);
2393
2394 done:
2395 mbuf_freem(packet);
2396 return error;
2397 }
2398
2399 static void
2400 flow_divert_close_all(struct flow_divert_group *group)
2401 {
2402 struct flow_divert_pcb *fd_cb;
2403 SLIST_HEAD(, flow_divert_pcb) tmp_list;
2404
2405 SLIST_INIT(&tmp_list);
2406
2407 lck_rw_lock_exclusive(&group->lck);
2408
2409 MBUFQ_DRAIN(&group->send_queue);
2410
2411 RB_FOREACH(fd_cb, fd_pcb_tree, &group->pcb_tree) {
2412 FDRETAIN(fd_cb);
2413 SLIST_INSERT_HEAD(&tmp_list, fd_cb, tmp_list_entry);
2414 }
2415
2416 lck_rw_done(&group->lck);
2417
2418 while (!SLIST_EMPTY(&tmp_list)) {
2419 fd_cb = SLIST_FIRST(&tmp_list);
2420 FDLOCK(fd_cb);
2421 SLIST_REMOVE_HEAD(&tmp_list, tmp_list_entry);
2422 if (fd_cb->so != NULL) {
2423 socket_lock(fd_cb->so, 0);
2424 flow_divert_pcb_remove(fd_cb);
2425 flow_divert_update_closed_state(fd_cb, SHUT_RDWR, TRUE);
2426 fd_cb->so->so_error = ECONNABORTED;
2427 flow_divert_disconnect_socket(fd_cb->so);
2428 socket_unlock(fd_cb->so, 0);
2429 }
2430 FDUNLOCK(fd_cb);
2431 FDRELEASE(fd_cb);
2432 }
2433 }
2434
2435 void
2436 flow_divert_detach(struct socket *so)
2437 {
2438 struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
2439
2440 VERIFY((so->so_flags & SOF_FLOW_DIVERT) && so->so_fd_pcb != NULL);
2441
2442 so->so_flags &= ~SOF_FLOW_DIVERT;
2443 so->so_fd_pcb = NULL;
2444
2445 FDLOG(LOG_INFO, fd_cb, "Detaching, ref count = %d", fd_cb->ref_count);
2446
2447 if (fd_cb->group != NULL) {
2448 /* Last-ditch effort to send any buffered data */
2449 flow_divert_send_buffered_data(fd_cb, TRUE);
2450
2451 flow_divert_update_closed_state(fd_cb, SHUT_RDWR, FALSE);
2452 flow_divert_send_close_if_needed(fd_cb);
2453 /* Remove from the group */
2454 flow_divert_pcb_remove(fd_cb);
2455 }
2456
2457 socket_unlock(so, 0);
2458 FDLOCK(fd_cb);
2459 fd_cb->so = NULL;
2460 FDUNLOCK(fd_cb);
2461 socket_lock(so, 0);
2462
2463 FDRELEASE(fd_cb); /* Release the socket's reference */
2464 }
2465
2466 static int
2467 flow_divert_close(struct socket *so)
2468 {
2469 struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
2470
2471 VERIFY((so->so_flags & SOF_FLOW_DIVERT) && so->so_fd_pcb != NULL);
2472
2473 FDLOG0(LOG_INFO, fd_cb, "Closing");
2474
2475 if (SOCK_TYPE(so) == SOCK_STREAM) {
2476 soisdisconnecting(so);
2477 sbflush(&so->so_rcv);
2478 }
2479
2480 flow_divert_send_buffered_data(fd_cb, TRUE);
2481 flow_divert_update_closed_state(fd_cb, SHUT_RDWR, FALSE);
2482 flow_divert_send_close_if_needed(fd_cb);
2483
2484 /* Remove from the group */
2485 flow_divert_pcb_remove(fd_cb);
2486
2487 return 0;
2488 }
2489
2490 static int
2491 flow_divert_disconnectx(struct socket *so, sae_associd_t aid,
2492 sae_connid_t cid __unused)
2493 {
2494 if (aid != SAE_ASSOCID_ANY && aid != SAE_ASSOCID_ALL) {
2495 return (EINVAL);
2496 }
2497
2498 return (flow_divert_close(so));
2499 }
2500
2501 static int
2502 flow_divert_shutdown(struct socket *so)
2503 {
2504 struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
2505
2506 VERIFY((so->so_flags & SOF_FLOW_DIVERT) && so->so_fd_pcb != NULL);
2507
2508 FDLOG0(LOG_INFO, fd_cb, "Can't send more");
2509
2510 socantsendmore(so);
2511
2512 flow_divert_update_closed_state(fd_cb, SHUT_WR, FALSE);
2513 flow_divert_send_close_if_needed(fd_cb);
2514
2515 return 0;
2516 }
2517
2518 static int
2519 flow_divert_rcvd(struct socket *so, int flags __unused)
2520 {
2521 struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
2522 uint32_t latest_sb_size;
2523 uint32_t read_count;
2524
2525 VERIFY((so->so_flags & SOF_FLOW_DIVERT) && so->so_fd_pcb != NULL);
2526
2527 latest_sb_size = fd_cb->so->so_rcv.sb_cc;
2528
2529 if (fd_cb->sb_size < latest_sb_size) {
2530 panic("flow divert rcvd event handler (%u): saved rcv buffer size (%u) is less than latest rcv buffer size (%u)",
2531 fd_cb->hash, fd_cb->sb_size, latest_sb_size);
2532 }
2533
2534 read_count = fd_cb->sb_size - latest_sb_size;
2535
2536 FDLOG(LOG_DEBUG, fd_cb, "app read %u bytes", read_count);
2537
2538 if (read_count > 0 && flow_divert_send_read_notification(fd_cb, read_count) == 0) {
2539 fd_cb->bytes_read_by_app += read_count;
2540 fd_cb->sb_size = latest_sb_size;
2541 }
2542
2543 return 0;
2544 }
2545
2546 static int
2547 flow_divert_append_target_endpoint_tlv(mbuf_t connect_packet, struct sockaddr *toaddr)
2548 {
2549 int error = 0;
2550 int port = 0;
2551
2552 error = flow_divert_packet_append_tlv(connect_packet, FLOW_DIVERT_TLV_TARGET_ADDRESS, toaddr->sa_len, toaddr);
2553 if (error) {
2554 goto done;
2555 }
2556
2557 if (toaddr->sa_family == AF_INET) {
2558 port = ntohs((satosin(toaddr))->sin_port);
2559 }
2560 #if INET6
2561 else {
2562 port = ntohs((satosin6(toaddr))->sin6_port);
2563 }
2564 #endif
2565
2566 error = flow_divert_packet_append_tlv(connect_packet, FLOW_DIVERT_TLV_TARGET_PORT, sizeof(port), &port);
2567 if (error) {
2568 goto done;
2569 }
2570
2571 done:
2572 return error;
2573 }
2574
2575 struct sockaddr *
2576 flow_divert_get_buffered_target_address(mbuf_t buffer)
2577 {
2578 if (buffer != NULL && buffer->m_type == MT_SONAME) {
2579 struct sockaddr *toaddr = mtod(buffer, struct sockaddr *);
2580 if (toaddr != NULL && flow_divert_is_sockaddr_valid(toaddr)) {
2581 return toaddr;
2582 }
2583 }
2584 return NULL;
2585 }
2586
2587 static boolean_t
2588 flow_divert_is_sockaddr_valid(struct sockaddr *addr)
2589 {
2590 switch(addr->sa_family)
2591 {
2592 case AF_INET:
2593 if (addr->sa_len != sizeof(struct sockaddr_in)) {
2594 return FALSE;
2595 }
2596 break;
2597 #if INET6
2598 case AF_INET6:
2599 if (addr->sa_len != sizeof(struct sockaddr_in6)) {
2600 return FALSE;
2601 }
2602 break;
2603 #endif /* INET6 */
2604 default:
2605 return FALSE;
2606 }
2607 return TRUE;
2608 }
2609
2610 static errno_t
2611 flow_divert_inp_to_sockaddr(const struct inpcb *inp, struct sockaddr **local_socket)
2612 {
2613 int error = 0;
2614 union sockaddr_in_4_6 sin46;
2615
2616 bzero(&sin46, sizeof(sin46));
2617 if (inp->inp_vflag & INP_IPV4) {
2618 struct sockaddr_in *sin = &sin46.sin;
2619
2620 sin->sin_family = AF_INET;
2621 sin->sin_len = sizeof(*sin);
2622 sin->sin_port = inp->inp_lport;
2623 sin->sin_addr = inp->inp_laddr;
2624 } else if (inp->inp_vflag & INP_IPV6) {
2625 struct sockaddr_in6 *sin6 = &sin46.sin6;
2626
2627 sin6->sin6_len = sizeof(*sin6);
2628 sin6->sin6_family = AF_INET6;
2629 sin6->sin6_port = inp->inp_lport;
2630 sin6->sin6_addr = inp->in6p_laddr;
2631 }
2632 *local_socket = dup_sockaddr((struct sockaddr *)&sin46, 1);
2633 if (*local_socket == NULL) {
2634 error = ENOBUFS;
2635 }
2636 return (error);
2637 }
2638
2639 static boolean_t
2640 flow_divert_has_pcb_local_address(const struct inpcb *inp)
2641 {
2642 return (inp->inp_lport != 0
2643 && (inp->inp_laddr.s_addr != INADDR_ANY || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)));
2644 }
2645
2646 static errno_t
2647 flow_divert_dup_addr(sa_family_t family, struct sockaddr *addr,
2648 struct sockaddr **dup)
2649 {
2650 int error = 0;
2651 struct sockaddr *result;
2652 struct sockaddr_storage ss;
2653
2654 if (addr != NULL) {
2655 result = addr;
2656 } else {
2657 memset(&ss, 0, sizeof(ss));
2658 ss.ss_family = family;
2659 if (ss.ss_family == AF_INET) {
2660 ss.ss_len = sizeof(struct sockaddr_in);
2661 }
2662 #if INET6
2663 else if (ss.ss_family == AF_INET6) {
2664 ss.ss_len = sizeof(struct sockaddr_in6);
2665 }
2666 #endif /* INET6 */
2667 else {
2668 error = EINVAL;
2669 }
2670 result = (struct sockaddr *)&ss;
2671 }
2672
2673 if (!error) {
2674 *dup = dup_sockaddr(result, 1);
2675 if (*dup == NULL) {
2676 error = ENOBUFS;
2677 }
2678 }
2679
2680 return error;
2681 }
2682
2683 static void
2684 flow_divert_disconnect_socket(struct socket *so)
2685 {
2686 soisdisconnected(so);
2687 if (SOCK_TYPE(so) == SOCK_DGRAM) {
2688 struct inpcb *inp = NULL;
2689
2690 inp = sotoinpcb(so);
2691 if (inp != NULL) {
2692 #if INET6
2693 if (SOCK_CHECK_DOM(so, PF_INET6))
2694 in6_pcbdetach(inp);
2695 else
2696 #endif /* INET6 */
2697 in_pcbdetach(inp);
2698 }
2699 }
2700 }
2701
2702 static errno_t
2703 flow_divert_getpeername(struct socket *so, struct sockaddr **sa)
2704 {
2705 struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
2706
2707 VERIFY((so->so_flags & SOF_FLOW_DIVERT) && so->so_fd_pcb != NULL);
2708
2709 return flow_divert_dup_addr(so->so_proto->pr_domain->dom_family,
2710 fd_cb->remote_address,
2711 sa);
2712 }
2713
2714 static errno_t
2715 flow_divert_getsockaddr(struct socket *so, struct sockaddr **sa)
2716 {
2717 struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
2718
2719 VERIFY((so->so_flags & SOF_FLOW_DIVERT) && so->so_fd_pcb != NULL);
2720
2721 return flow_divert_dup_addr(so->so_proto->pr_domain->dom_family,
2722 fd_cb->local_address,
2723 sa);
2724 }
2725
2726 static errno_t
2727 flow_divert_ctloutput(struct socket *so, struct sockopt *sopt)
2728 {
2729 struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
2730
2731 VERIFY((so->so_flags & SOF_FLOW_DIVERT) && so->so_fd_pcb != NULL);
2732
2733 if (sopt->sopt_name == SO_TRAFFIC_CLASS) {
2734 if (sopt->sopt_dir == SOPT_SET && fd_cb->flags & FLOW_DIVERT_CONNECT_STARTED) {
2735 flow_divert_send_traffic_class_update(fd_cb, so->so_traffic_class);
2736 }
2737 }
2738
2739 if (SOCK_DOM(so) == PF_INET) {
2740 return g_tcp_protosw->pr_ctloutput(so, sopt);
2741 }
2742 #if INET6
2743 else if (SOCK_DOM(so) == PF_INET6) {
2744 return g_tcp6_protosw->pr_ctloutput(so, sopt);
2745 }
2746 #endif
2747 return 0;
2748 }
2749
2750 errno_t
2751 flow_divert_connect_out(struct socket *so, struct sockaddr *to, proc_t p)
2752 {
2753 struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
2754 int error = 0;
2755 struct inpcb *inp = sotoinpcb(so);
2756 struct sockaddr_in *sinp;
2757 mbuf_t connect_packet = NULL;
2758 int do_send = 1;
2759
2760 VERIFY((so->so_flags & SOF_FLOW_DIVERT) && so->so_fd_pcb != NULL);
2761
2762 if (fd_cb->group == NULL) {
2763 error = ENETUNREACH;
2764 goto done;
2765 }
2766
2767 if (inp == NULL) {
2768 error = EINVAL;
2769 goto done;
2770 } else if (inp->inp_state == INPCB_STATE_DEAD) {
2771 if (so->so_error) {
2772 error = so->so_error;
2773 so->so_error = 0;
2774 } else {
2775 error = EINVAL;
2776 }
2777 goto done;
2778 }
2779
2780 if ((fd_cb->flags & FLOW_DIVERT_CONNECT_STARTED) && !(fd_cb->flags & FLOW_DIVERT_TRANSFERRED)) {
2781 error = EALREADY;
2782 goto done;
2783 }
2784
2785 if (fd_cb->flags & FLOW_DIVERT_TRANSFERRED) {
2786 FDLOG0(LOG_INFO, fd_cb, "fully transferred");
2787 fd_cb->flags &= ~FLOW_DIVERT_TRANSFERRED;
2788 if (fd_cb->remote_address != NULL) {
2789 soisconnected(fd_cb->so);
2790 goto done;
2791 }
2792 }
2793
2794 FDLOG0(LOG_INFO, fd_cb, "Connecting");
2795
2796 if (fd_cb->connect_packet == NULL) {
2797 if (to == NULL) {
2798 FDLOG0(LOG_ERR, fd_cb, "No destination address available when creating connect packet");
2799 error = EINVAL;
2800 goto done;
2801 }
2802
2803 sinp = (struct sockaddr_in *)(void *)to;
2804 if (sinp->sin_family == AF_INET && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
2805 error = EAFNOSUPPORT;
2806 goto done;
2807 }
2808
2809 error = flow_divert_create_connect_packet(fd_cb, to, so, p, &connect_packet);
2810 if (error) {
2811 goto done;
2812 }
2813
2814 if (so->so_flags1 & SOF1_PRECONNECT_DATA) {
2815 FDLOG0(LOG_INFO, fd_cb, "Delaying sending the connect packet until send or receive");
2816 do_send = 0;
2817 }
2818 } else {
2819 FDLOG0(LOG_INFO, fd_cb, "Sending saved connect packet");
2820 connect_packet = fd_cb->connect_packet;
2821 fd_cb->connect_packet = NULL;
2822 }
2823
2824 if (do_send) {
2825 error = flow_divert_send_packet(fd_cb, connect_packet, TRUE);
2826 if (error) {
2827 goto done;
2828 }
2829
2830 fd_cb->flags |= FLOW_DIVERT_CONNECT_STARTED;
2831 } else {
2832 fd_cb->connect_packet = connect_packet;
2833 connect_packet = NULL;
2834 }
2835
2836 soisconnecting(so);
2837
2838 done:
2839 if (error && connect_packet != NULL) {
2840 mbuf_freem(connect_packet);
2841 }
2842 return error;
2843 }
2844
2845 static int
2846 flow_divert_connectx_out_common(struct socket *so, struct sockaddr *dst,
2847 struct proc *p, sae_connid_t *pcid, struct uio *auio, user_ssize_t *bytes_written)
2848 {
2849 struct inpcb *inp = sotoinpcb(so);
2850 int error;
2851
2852 if (inp == NULL) {
2853 return (EINVAL);
2854 }
2855
2856 VERIFY(dst != NULL);
2857
2858 error = flow_divert_connect_out(so, dst, p);
2859
2860 if (error != 0) {
2861 return error;
2862 }
2863
2864 /* if there is data, send it */
2865 if (auio != NULL) {
2866 user_ssize_t datalen = 0;
2867
2868 socket_unlock(so, 0);
2869
2870 VERIFY(bytes_written != NULL);
2871
2872 datalen = uio_resid(auio);
2873 error = so->so_proto->pr_usrreqs->pru_sosend(so, NULL, (uio_t)auio, NULL, NULL, 0);
2874 socket_lock(so, 0);
2875
2876 if (error == 0 || error == EWOULDBLOCK) {
2877 *bytes_written = datalen - uio_resid(auio);
2878 }
2879
2880 /*
2881 * sosend returns EWOULDBLOCK if it's a non-blocking
2882 * socket or a timeout occured (this allows to return
2883 * the amount of queued data through sendit()).
2884 *
2885 * However, connectx() returns EINPROGRESS in case of a
2886 * blocking socket. So we change the return value here.
2887 */
2888 if (error == EWOULDBLOCK) {
2889 error = EINPROGRESS;
2890 }
2891 }
2892
2893 if (error == 0 && pcid != NULL) {
2894 *pcid = 1; /* there is only 1 connection for a TCP */
2895 }
2896
2897 return (error);
2898 }
2899
2900 static int
2901 flow_divert_connectx_out(struct socket *so, struct sockaddr *src __unused,
2902 struct sockaddr *dst, struct proc *p, uint32_t ifscope __unused,
2903 sae_associd_t aid __unused, sae_connid_t *pcid, uint32_t flags __unused, void *arg __unused,
2904 uint32_t arglen __unused, struct uio *uio, user_ssize_t *bytes_written)
2905 {
2906 return (flow_divert_connectx_out_common(so, dst, p, pcid, uio, bytes_written));
2907 }
2908
2909 #if INET6
2910 static int
2911 flow_divert_connectx6_out(struct socket *so, struct sockaddr *src __unused,
2912 struct sockaddr *dst, struct proc *p, uint32_t ifscope __unused,
2913 sae_associd_t aid __unused, sae_connid_t *pcid, uint32_t flags __unused, void *arg __unused,
2914 uint32_t arglen __unused, struct uio *uio, user_ssize_t *bytes_written)
2915 {
2916 return (flow_divert_connectx_out_common(so, dst, p, pcid, uio, bytes_written));
2917 }
2918 #endif /* INET6 */
2919
2920 static int
2921 flow_divert_getconninfo(struct socket *so, sae_connid_t cid, uint32_t *flags,
2922 uint32_t *ifindex, int32_t *soerror, user_addr_t src, socklen_t *src_len,
2923 user_addr_t dst, socklen_t *dst_len, uint32_t *aux_type,
2924 user_addr_t aux_data __unused, uint32_t *aux_len)
2925 {
2926 int error = 0;
2927 struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
2928 struct ifnet *ifp = NULL;
2929 struct inpcb *inp = sotoinpcb(so);
2930
2931 VERIFY((so->so_flags & SOF_FLOW_DIVERT));
2932
2933 if (so->so_fd_pcb == NULL || inp == NULL) {
2934 error = EINVAL;
2935 goto out;
2936 }
2937
2938 if (cid != SAE_CONNID_ANY && cid != SAE_CONNID_ALL && cid != 1) {
2939 error = EINVAL;
2940 goto out;
2941 }
2942
2943 ifp = inp->inp_last_outifp;
2944 *ifindex = ((ifp != NULL) ? ifp->if_index : 0);
2945 *soerror = so->so_error;
2946 *flags = 0;
2947
2948 if (so->so_state & SS_ISCONNECTED) {
2949 *flags |= (CIF_CONNECTED | CIF_PREFERRED);
2950 }
2951
2952 if (fd_cb->local_address == NULL) {
2953 struct sockaddr_in sin;
2954 bzero(&sin, sizeof(sin));
2955 sin.sin_len = sizeof(sin);
2956 sin.sin_family = AF_INET;
2957 *src_len = sin.sin_len;
2958 if (src != USER_ADDR_NULL) {
2959 error = copyout(&sin, src, sin.sin_len);
2960 if (error != 0) {
2961 goto out;
2962 }
2963 }
2964 } else {
2965 *src_len = fd_cb->local_address->sa_len;
2966 if (src != USER_ADDR_NULL) {
2967 error = copyout(fd_cb->local_address, src, fd_cb->local_address->sa_len);
2968 if (error != 0) {
2969 goto out;
2970 }
2971 }
2972 }
2973
2974 if (fd_cb->remote_address == NULL) {
2975 struct sockaddr_in sin;
2976 bzero(&sin, sizeof(sin));
2977 sin.sin_len = sizeof(sin);
2978 sin.sin_family = AF_INET;
2979 *dst_len = sin.sin_len;
2980 if (dst != USER_ADDR_NULL) {
2981 error = copyout(&sin, dst, sin.sin_len);
2982 if (error != 0) {
2983 goto out;
2984 }
2985 }
2986 } else {
2987 *dst_len = fd_cb->remote_address->sa_len;
2988 if (dst != USER_ADDR_NULL) {
2989 error = copyout(fd_cb->remote_address, dst, fd_cb->remote_address->sa_len);
2990 if (error != 0) {
2991 goto out;
2992 }
2993 }
2994 }
2995
2996 *aux_type = 0;
2997 *aux_len = 0;
2998
2999 out:
3000 return error;
3001 }
3002
3003 static int
3004 flow_divert_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp __unused, struct proc *p __unused)
3005 {
3006 int error = 0;
3007
3008 switch (cmd) {
3009 case SIOCGCONNINFO32: {
3010 struct so_cinforeq32 cifr;
3011 bcopy(data, &cifr, sizeof (cifr));
3012 error = flow_divert_getconninfo(so, cifr.scir_cid, &cifr.scir_flags,
3013 &cifr.scir_ifindex, &cifr.scir_error, cifr.scir_src,
3014 &cifr.scir_src_len, cifr.scir_dst, &cifr.scir_dst_len,
3015 &cifr.scir_aux_type, cifr.scir_aux_data,
3016 &cifr.scir_aux_len);
3017 if (error == 0) {
3018 bcopy(&cifr, data, sizeof (cifr));
3019 }
3020 break;
3021 }
3022
3023 case SIOCGCONNINFO64: {
3024 struct so_cinforeq64 cifr;
3025 bcopy(data, &cifr, sizeof (cifr));
3026 error = flow_divert_getconninfo(so, cifr.scir_cid, &cifr.scir_flags,
3027 &cifr.scir_ifindex, &cifr.scir_error, cifr.scir_src,
3028 &cifr.scir_src_len, cifr.scir_dst, &cifr.scir_dst_len,
3029 &cifr.scir_aux_type, cifr.scir_aux_data,
3030 &cifr.scir_aux_len);
3031 if (error == 0) {
3032 bcopy(&cifr, data, sizeof (cifr));
3033 }
3034 break;
3035 }
3036
3037 default:
3038 error = EOPNOTSUPP;
3039 }
3040
3041 return error;
3042 }
3043
3044 static int
3045 flow_divert_in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp, struct proc *p)
3046 {
3047 int error = flow_divert_control(so, cmd, data, ifp, p);
3048
3049 if (error == EOPNOTSUPP) {
3050 error = in_control(so, cmd, data, ifp, p);
3051 }
3052
3053 return error;
3054 }
3055
3056 static int
3057 flow_divert_in6_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp, struct proc *p)
3058 {
3059 int error = flow_divert_control(so, cmd, data, ifp, p);
3060
3061 if (error == EOPNOTSUPP) {
3062 error = in6_control(so, cmd, data, ifp, p);
3063 }
3064
3065 return error;
3066 }
3067
3068 static errno_t
3069 flow_divert_data_out(struct socket *so, int flags, mbuf_t data, struct sockaddr *to, mbuf_t control, struct proc *p)
3070 {
3071 struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
3072 int error = 0;
3073 struct inpcb *inp;
3074
3075 VERIFY((so->so_flags & SOF_FLOW_DIVERT) && so->so_fd_pcb != NULL);
3076
3077 inp = sotoinpcb(so);
3078 if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) {
3079 error = ECONNRESET;
3080 goto done;
3081 }
3082
3083 if (control && mbuf_len(control) > 0) {
3084 error = EINVAL;
3085 goto done;
3086 }
3087
3088 if (flags & MSG_OOB) {
3089 error = EINVAL;
3090 goto done; /* We don't support OOB data */
3091 }
3092
3093 error = flow_divert_check_no_cellular(fd_cb) ||
3094 flow_divert_check_no_expensive(fd_cb);
3095 if (error) {
3096 goto done;
3097 }
3098
3099 /* Implicit connect */
3100 if (!(fd_cb->flags & FLOW_DIVERT_CONNECT_STARTED)) {
3101 FDLOG0(LOG_INFO, fd_cb, "implicit connect");
3102 error = flow_divert_connect_out(so, to, p);
3103 if (error) {
3104 goto done;
3105 }
3106
3107 if (so->so_flags1 & SOF1_DATA_IDEMPOTENT) {
3108 /* Open up the send window so that the data will get sent right away */
3109 fd_cb->send_window = mbuf_pkthdr_len(data);
3110 }
3111 }
3112
3113 FDLOG(LOG_DEBUG, fd_cb, "app wrote %lu bytes", mbuf_pkthdr_len(data));
3114
3115 fd_cb->bytes_written_by_app += mbuf_pkthdr_len(data);
3116 error = flow_divert_send_app_data(fd_cb, data, to);
3117 if (error) {
3118 goto done;
3119 }
3120
3121 data = NULL;
3122
3123 if (flags & PRUS_EOF) {
3124 flow_divert_shutdown(so);
3125 }
3126
3127 done:
3128 if (data) {
3129 mbuf_freem(data);
3130 }
3131 if (control) {
3132 mbuf_free(control);
3133 }
3134 return error;
3135 }
3136
3137 static int
3138 flow_divert_preconnect(struct socket *so)
3139 {
3140 struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
3141 int error = 0;
3142
3143 if (!(fd_cb->flags & FLOW_DIVERT_CONNECT_STARTED) && fd_cb->connect_packet != NULL) {
3144 FDLOG0(LOG_INFO, fd_cb, "Pre-connect read: sending saved connect packet");
3145 mbuf_t connect_packet = fd_cb->connect_packet;
3146 fd_cb->connect_packet = NULL;
3147
3148 error = flow_divert_send_packet(fd_cb, connect_packet, TRUE);
3149 if (error) {
3150 mbuf_freem(connect_packet);
3151 }
3152
3153 fd_cb->flags |= FLOW_DIVERT_CONNECT_STARTED;
3154 }
3155
3156 soclearfastopen(so);
3157
3158 return error;
3159 }
3160
3161 static void
3162 flow_divert_set_protosw(struct socket *so)
3163 {
3164 so->so_flags |= SOF_FLOW_DIVERT;
3165 if (SOCK_DOM(so) == PF_INET) {
3166 so->so_proto = &g_flow_divert_in_protosw;
3167 }
3168 #if INET6
3169 else {
3170 so->so_proto = (struct protosw *)&g_flow_divert_in6_protosw;
3171 }
3172 #endif /* INET6 */
3173 }
3174
3175 static void
3176 flow_divert_set_udp_protosw(struct socket *so)
3177 {
3178 so->so_flags |= SOF_FLOW_DIVERT;
3179 if (SOCK_DOM(so) == PF_INET) {
3180 so->so_proto = &g_flow_divert_in_udp_protosw;
3181 }
3182 #if INET6
3183 else {
3184 so->so_proto = (struct protosw *)&g_flow_divert_in6_udp_protosw;
3185 }
3186 #endif /* INET6 */
3187 }
3188
3189 static errno_t
3190 flow_divert_attach(struct socket *so, uint32_t flow_id, uint32_t ctl_unit)
3191 {
3192 int error = 0;
3193 struct flow_divert_pcb *fd_cb = NULL;
3194 struct ifnet *ifp = NULL;
3195 struct inpcb *inp = NULL;
3196 struct socket *old_so;
3197 mbuf_t recv_data = NULL;
3198
3199 socket_unlock(so, 0);
3200
3201 FDLOG(LOG_INFO, &nil_pcb, "Attaching socket to flow %u", flow_id);
3202
3203 /* Find the flow divert control block */
3204 lck_rw_lock_shared(&g_flow_divert_group_lck);
3205 if (g_flow_divert_groups != NULL && g_active_group_count > 0) {
3206 struct flow_divert_group *group = g_flow_divert_groups[ctl_unit];
3207 if (group != NULL) {
3208 fd_cb = flow_divert_pcb_lookup(flow_id, group);
3209 }
3210 }
3211 lck_rw_done(&g_flow_divert_group_lck);
3212
3213 if (fd_cb == NULL) {
3214 error = ENOENT;
3215 goto done;
3216 }
3217
3218 FDLOCK(fd_cb);
3219
3220 /* Dis-associate the flow divert control block from its current socket */
3221 old_so = fd_cb->so;
3222
3223 inp = sotoinpcb(old_so);
3224
3225 VERIFY(inp != NULL);
3226
3227 socket_lock(old_so, 0);
3228 flow_divert_disconnect_socket(old_so);
3229 old_so->so_flags &= ~SOF_FLOW_DIVERT;
3230 old_so->so_fd_pcb = NULL;
3231 if (SOCK_TYPE(old_so) == SOCK_STREAM) {
3232 old_so->so_proto = pffindproto(SOCK_DOM(old_so), IPPROTO_TCP, SOCK_STREAM);
3233 } else if (SOCK_TYPE(old_so) == SOCK_DGRAM) {
3234 old_so->so_proto = pffindproto(SOCK_DOM(old_so), IPPROTO_UDP, SOCK_DGRAM);
3235 }
3236 fd_cb->so = NULL;
3237 /* Save the output interface */
3238 ifp = inp->inp_last_outifp;
3239 if (old_so->so_rcv.sb_cc > 0) {
3240 error = mbuf_dup(old_so->so_rcv.sb_mb, MBUF_DONTWAIT, &recv_data);
3241 sbflush(&old_so->so_rcv);
3242 }
3243 socket_unlock(old_so, 0);
3244
3245 /* Associate the new socket with the flow divert control block */
3246 socket_lock(so, 0);
3247 so->so_fd_pcb = fd_cb;
3248 inp = sotoinpcb(so);
3249 inp->inp_last_outifp = ifp;
3250 if (recv_data != NULL) {
3251 if (sbappendstream(&so->so_rcv, recv_data)) {
3252 sorwakeup(so);
3253 }
3254 }
3255 flow_divert_set_protosw(so);
3256 socket_unlock(so, 0);
3257
3258 fd_cb->so = so;
3259 fd_cb->flags |= FLOW_DIVERT_TRANSFERRED;
3260
3261 FDUNLOCK(fd_cb);
3262
3263 done:
3264 socket_lock(so, 0);
3265
3266 if (fd_cb != NULL) {
3267 FDRELEASE(fd_cb); /* Release the reference obtained via flow_divert_pcb_lookup */
3268 }
3269
3270 return error;
3271 }
3272
3273 errno_t
3274 flow_divert_implicit_data_out(struct socket *so, int flags, mbuf_t data, struct sockaddr *to, mbuf_t control, struct proc *p)
3275 {
3276 struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
3277 struct inpcb *inp;
3278 int error = 0;
3279
3280 inp = sotoinpcb(so);
3281 if (inp == NULL) {
3282 return (EINVAL);
3283 }
3284
3285 if (fd_cb == NULL) {
3286 uint32_t fd_ctl_unit = necp_socket_get_flow_divert_control_unit(inp);
3287 if (fd_ctl_unit > 0) {
3288 error = flow_divert_pcb_init(so, fd_ctl_unit);
3289 fd_cb = so->so_fd_pcb;
3290 if (error != 0 || fd_cb == NULL) {
3291 goto done;
3292 }
3293 } else {
3294 error = ENETDOWN;
3295 goto done;
3296 }
3297 }
3298 return flow_divert_data_out(so, flags, data, to, control, p);
3299
3300 done:
3301 if (data) {
3302 mbuf_freem(data);
3303 }
3304 if (control) {
3305 mbuf_free(control);
3306 }
3307
3308 return error;
3309 }
3310
3311 errno_t
3312 flow_divert_pcb_init(struct socket *so, uint32_t ctl_unit)
3313 {
3314 errno_t error = 0;
3315 struct flow_divert_pcb *fd_cb;
3316
3317 if (so->so_flags & SOF_FLOW_DIVERT) {
3318 return EALREADY;
3319 }
3320
3321 fd_cb = flow_divert_pcb_create(so);
3322 if (fd_cb != NULL) {
3323 error = flow_divert_pcb_insert(fd_cb, ctl_unit);
3324 if (error) {
3325 FDLOG(LOG_ERR, fd_cb, "pcb insert failed: %d", error);
3326 FDRELEASE(fd_cb);
3327 } else {
3328 fd_cb->control_group_unit = ctl_unit;
3329 so->so_fd_pcb = fd_cb;
3330
3331 if (SOCK_TYPE(so) == SOCK_STREAM) {
3332 flow_divert_set_protosw(so);
3333 } else if (SOCK_TYPE(so) == SOCK_DGRAM) {
3334 flow_divert_set_udp_protosw(so);
3335 }
3336
3337 FDLOG0(LOG_INFO, fd_cb, "Created");
3338 }
3339 } else {
3340 error = ENOMEM;
3341 }
3342
3343 return error;
3344 }
3345
3346 errno_t
3347 flow_divert_token_set(struct socket *so, struct sockopt *sopt)
3348 {
3349 uint32_t ctl_unit = 0;
3350 uint32_t key_unit = 0;
3351 uint32_t flow_id = 0;
3352 int error = 0;
3353 int hmac_error = 0;
3354 mbuf_t token = NULL;
3355
3356 if (so->so_flags & SOF_FLOW_DIVERT) {
3357 error = EALREADY;
3358 goto done;
3359 }
3360
3361 if (g_init_result) {
3362 FDLOG(LOG_ERR, &nil_pcb, "flow_divert_init failed (%d), cannot use flow divert", g_init_result);
3363 error = ENOPROTOOPT;
3364 goto done;
3365 }
3366
3367 if ((SOCK_TYPE(so) != SOCK_STREAM && SOCK_TYPE(so) != SOCK_DGRAM) ||
3368 (SOCK_PROTO(so) != IPPROTO_TCP && SOCK_PROTO(so) != IPPROTO_UDP) ||
3369 (SOCK_DOM(so) != PF_INET
3370 #if INET6
3371 && SOCK_DOM(so) != PF_INET6
3372 #endif
3373 ))
3374 {
3375 error = EINVAL;
3376 goto done;
3377 } else {
3378 if (SOCK_TYPE(so) == SOCK_STREAM && SOCK_PROTO(so) == IPPROTO_TCP) {
3379 struct tcpcb *tp = sototcpcb(so);
3380 if (tp == NULL || tp->t_state != TCPS_CLOSED) {
3381 error = EINVAL;
3382 goto done;
3383 }
3384 }
3385 }
3386
3387 error = soopt_getm(sopt, &token);
3388 if (error) {
3389 token = NULL;
3390 goto done;
3391 }
3392
3393 error = soopt_mcopyin(sopt, token);
3394 if (error) {
3395 token = NULL;
3396 goto done;
3397 }
3398
3399 error = flow_divert_packet_get_tlv(token, 0, FLOW_DIVERT_TLV_KEY_UNIT, sizeof(key_unit), (void *)&key_unit, NULL);
3400 if (!error) {
3401 key_unit = ntohl(key_unit);
3402 if (key_unit >= GROUP_COUNT_MAX) {
3403 key_unit = 0;
3404 }
3405 } else if (error != ENOENT) {
3406 FDLOG(LOG_ERR, &nil_pcb, "Failed to get the key unit from the token: %d", error);
3407 goto done;
3408 } else {
3409 key_unit = 0;
3410 }
3411
3412 error = flow_divert_packet_get_tlv(token, 0, FLOW_DIVERT_TLV_CTL_UNIT, sizeof(ctl_unit), (void *)&ctl_unit, NULL);
3413 if (error) {
3414 FDLOG(LOG_ERR, &nil_pcb, "Failed to get the control socket unit from the token: %d", error);
3415 goto done;
3416 }
3417
3418 /* A valid kernel control unit is required */
3419 ctl_unit = ntohl(ctl_unit);
3420 if (ctl_unit == 0 || ctl_unit >= GROUP_COUNT_MAX) {
3421 FDLOG(LOG_ERR, &nil_pcb, "Got an invalid control socket unit: %u", ctl_unit);
3422 error = EINVAL;
3423 goto done;
3424 }
3425
3426 socket_unlock(so, 0);
3427 hmac_error = flow_divert_packet_verify_hmac(token, (key_unit != 0 ? key_unit : ctl_unit));
3428 socket_lock(so, 0);
3429
3430 if (hmac_error && hmac_error != ENOENT) {
3431 FDLOG(LOG_ERR, &nil_pcb, "HMAC verfication failed: %d", hmac_error);
3432 error = hmac_error;
3433 goto done;
3434 }
3435
3436 error = flow_divert_packet_get_tlv(token, 0, FLOW_DIVERT_TLV_FLOW_ID, sizeof(flow_id), (void *)&flow_id, NULL);
3437 if (error && error != ENOENT) {
3438 FDLOG(LOG_ERR, &nil_pcb, "Failed to get the flow ID from the token: %d", error);
3439 goto done;
3440 }
3441
3442 if (flow_id == 0) {
3443 error = flow_divert_pcb_init(so, ctl_unit);
3444 if (error == 0) {
3445 struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
3446 int log_level = LOG_NOTICE;
3447
3448 error = flow_divert_packet_get_tlv(token, 0, FLOW_DIVERT_TLV_LOG_LEVEL,
3449 sizeof(log_level), &log_level, NULL);
3450 if (error == 0) {
3451 fd_cb->log_level = log_level;
3452 }
3453 error = 0;
3454
3455 fd_cb->connect_token = token;
3456 token = NULL;
3457 }
3458 } else {
3459 error = flow_divert_attach(so, flow_id, ctl_unit);
3460 }
3461
3462 if (hmac_error == 0) {
3463 struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
3464 if (fd_cb != NULL) {
3465 fd_cb->flags |= FLOW_DIVERT_HAS_HMAC;
3466 }
3467 }
3468
3469 done:
3470 if (token != NULL) {
3471 mbuf_freem(token);
3472 }
3473
3474 return error;
3475 }
3476
3477 errno_t
3478 flow_divert_token_get(struct socket *so, struct sockopt *sopt)
3479 {
3480 uint32_t ctl_unit;
3481 int error = 0;
3482 uint8_t hmac[SHA_DIGEST_LENGTH];
3483 struct flow_divert_pcb *fd_cb = so->so_fd_pcb;
3484 mbuf_t token = NULL;
3485 struct flow_divert_group *control_group = NULL;
3486
3487 if (!(so->so_flags & SOF_FLOW_DIVERT)) {
3488 error = EINVAL;
3489 goto done;
3490 }
3491
3492 VERIFY((so->so_flags & SOF_FLOW_DIVERT) && so->so_fd_pcb != NULL);
3493
3494 if (fd_cb->group == NULL) {
3495 error = EINVAL;
3496 goto done;
3497 }
3498
3499 error = mbuf_gethdr(MBUF_DONTWAIT, MBUF_TYPE_HEADER, &token);
3500 if (error) {
3501 FDLOG(LOG_ERR, fd_cb, "failed to allocate the header mbuf: %d", error);
3502 goto done;
3503 }
3504
3505 ctl_unit = htonl(fd_cb->group->ctl_unit);
3506
3507 error = flow_divert_packet_append_tlv(token, FLOW_DIVERT_TLV_CTL_UNIT, sizeof(ctl_unit), &ctl_unit);
3508 if (error) {
3509 goto done;
3510 }
3511
3512 error = flow_divert_packet_append_tlv(token, FLOW_DIVERT_TLV_FLOW_ID, sizeof(fd_cb->hash), &fd_cb->hash);
3513 if (error) {
3514 goto done;
3515 }
3516
3517 if (fd_cb->app_data != NULL) {
3518 error = flow_divert_packet_append_tlv(token, FLOW_DIVERT_TLV_APP_DATA, fd_cb->app_data_length, fd_cb->app_data);
3519 if (error) {
3520 goto done;
3521 }
3522 }
3523
3524 socket_unlock(so, 0);
3525 lck_rw_lock_shared(&g_flow_divert_group_lck);
3526
3527 if (g_flow_divert_groups != NULL && g_active_group_count > 0 &&
3528 fd_cb->control_group_unit > 0 && fd_cb->control_group_unit < GROUP_COUNT_MAX)
3529 {
3530 control_group = g_flow_divert_groups[fd_cb->control_group_unit];
3531 }
3532
3533 if (control_group != NULL) {
3534 lck_rw_lock_shared(&control_group->lck);
3535 ctl_unit = htonl(control_group->ctl_unit);
3536 error = flow_divert_packet_append_tlv(token, FLOW_DIVERT_TLV_KEY_UNIT, sizeof(ctl_unit), &ctl_unit);
3537 if (!error) {
3538 error = flow_divert_packet_compute_hmac(token, control_group, hmac);
3539 }
3540 lck_rw_done(&control_group->lck);
3541 } else {
3542 error = ENOPROTOOPT;
3543 }
3544
3545 lck_rw_done(&g_flow_divert_group_lck);
3546 socket_lock(so, 0);
3547
3548 if (error) {
3549 goto done;
3550 }
3551
3552 error = flow_divert_packet_append_tlv(token, FLOW_DIVERT_TLV_HMAC, sizeof(hmac), hmac);
3553 if (error) {
3554 goto done;
3555 }
3556
3557 if (sopt->sopt_val == USER_ADDR_NULL) {
3558 /* If the caller passed NULL to getsockopt, just set the size of the token and return */
3559 sopt->sopt_valsize = mbuf_pkthdr_len(token);
3560 goto done;
3561 }
3562
3563 error = soopt_mcopyout(sopt, token);
3564 if (error) {
3565 token = NULL; /* For some reason, soopt_mcopyout() frees the mbuf if it fails */
3566 goto done;
3567 }
3568
3569 done:
3570 if (token != NULL) {
3571 mbuf_freem(token);
3572 }
3573
3574 return error;
3575 }
3576
3577 static errno_t
3578 flow_divert_kctl_connect(kern_ctl_ref kctlref __unused, struct sockaddr_ctl *sac, void **unitinfo)
3579 {
3580 struct flow_divert_group *new_group = NULL;
3581 int error = 0;
3582
3583 if (sac->sc_unit >= GROUP_COUNT_MAX) {
3584 error = EINVAL;
3585 goto done;
3586 }
3587
3588 *unitinfo = NULL;
3589
3590 MALLOC_ZONE(new_group, struct flow_divert_group *, sizeof(*new_group), M_FLOW_DIVERT_GROUP, M_WAITOK);
3591 if (new_group == NULL) {
3592 error = ENOBUFS;
3593 goto done;
3594 }
3595
3596 memset(new_group, 0, sizeof(*new_group));
3597
3598 lck_rw_init(&new_group->lck, flow_divert_mtx_grp, flow_divert_mtx_attr);
3599 RB_INIT(&new_group->pcb_tree);
3600 new_group->ctl_unit = sac->sc_unit;
3601 MBUFQ_INIT(&new_group->send_queue);
3602 new_group->signing_id_trie.root = NULL_TRIE_IDX;
3603
3604 lck_rw_lock_exclusive(&g_flow_divert_group_lck);
3605
3606 if (g_flow_divert_groups == NULL) {
3607 MALLOC(g_flow_divert_groups,
3608 struct flow_divert_group **,
3609 GROUP_COUNT_MAX * sizeof(struct flow_divert_group *),
3610 M_TEMP,
3611 M_WAITOK | M_ZERO);
3612 }
3613
3614 if (g_flow_divert_groups == NULL) {
3615 error = ENOBUFS;
3616 } else if (g_flow_divert_groups[sac->sc_unit] != NULL) {
3617 error = EALREADY;
3618 } else {
3619 g_flow_divert_groups[sac->sc_unit] = new_group;
3620 g_active_group_count++;
3621 }
3622
3623 lck_rw_done(&g_flow_divert_group_lck);
3624
3625 *unitinfo = new_group;
3626
3627 done:
3628 if (error != 0 && new_group != NULL) {
3629 FREE_ZONE(new_group, sizeof(*new_group), M_FLOW_DIVERT_GROUP);
3630 }
3631 return error;
3632 }
3633
3634 static errno_t
3635 flow_divert_kctl_disconnect(kern_ctl_ref kctlref __unused, uint32_t unit, void *unitinfo)
3636 {
3637 struct flow_divert_group *group = NULL;
3638 errno_t error = 0;
3639
3640 if (unit >= GROUP_COUNT_MAX) {
3641 return EINVAL;
3642 }
3643
3644 FDLOG(LOG_INFO, &nil_pcb, "disconnecting group %d", unit);
3645
3646 lck_rw_lock_exclusive(&g_flow_divert_group_lck);
3647
3648 if (g_flow_divert_groups == NULL || g_active_group_count == 0) {
3649 panic("flow divert group %u is disconnecting, but no groups are active (groups = %p, active count = %u", unit,
3650 g_flow_divert_groups, g_active_group_count);
3651 }
3652
3653 group = g_flow_divert_groups[unit];
3654
3655 if (group != (struct flow_divert_group *)unitinfo) {
3656 panic("group with unit %d (%p) != unit info (%p)", unit, group, unitinfo);
3657 }
3658
3659 if (group != NULL) {
3660 flow_divert_close_all(group);
3661 if (group->token_key != NULL) {
3662 memset(group->token_key, 0, group->token_key_size);
3663 FREE(group->token_key, M_TEMP);
3664 group->token_key = NULL;
3665 group->token_key_size = 0;
3666 }
3667
3668 /* Re-set the current trie */
3669 if (group->signing_id_trie.memory != NULL) {
3670 FREE(group->signing_id_trie.memory, M_TEMP);
3671 }
3672 memset(&group->signing_id_trie, 0, sizeof(group->signing_id_trie));
3673 group->signing_id_trie.root = NULL_TRIE_IDX;
3674
3675 FREE_ZONE(group, sizeof(*group), M_FLOW_DIVERT_GROUP);
3676 g_flow_divert_groups[unit] = NULL;
3677 g_active_group_count--;
3678 } else {
3679 error = EINVAL;
3680 }
3681
3682 if (g_active_group_count == 0) {
3683 FREE(g_flow_divert_groups, M_TEMP);
3684 g_flow_divert_groups = NULL;
3685 }
3686
3687 lck_rw_done(&g_flow_divert_group_lck);
3688
3689 return error;
3690 }
3691
3692 static errno_t
3693 flow_divert_kctl_send(kern_ctl_ref kctlref __unused, uint32_t unit __unused, void *unitinfo, mbuf_t m, int flags __unused)
3694 {
3695 return flow_divert_input(m, (struct flow_divert_group *)unitinfo);
3696 }
3697
3698 static void
3699 flow_divert_kctl_rcvd(kern_ctl_ref kctlref __unused, uint32_t unit __unused, void *unitinfo, int flags __unused)
3700 {
3701 struct flow_divert_group *group = (struct flow_divert_group *)unitinfo;
3702
3703 if (!OSTestAndClear(GROUP_BIT_CTL_ENQUEUE_BLOCKED, &group->atomic_bits)) {
3704 struct flow_divert_pcb *fd_cb;
3705 SLIST_HEAD(, flow_divert_pcb) tmp_list;
3706
3707 lck_rw_lock_shared(&g_flow_divert_group_lck);
3708 lck_rw_lock_exclusive(&group->lck);
3709
3710 while (!MBUFQ_EMPTY(&group->send_queue)) {
3711 mbuf_t next_packet;
3712 FDLOG0(LOG_DEBUG, &nil_pcb, "trying ctl_enqueuembuf again");
3713 next_packet = MBUFQ_FIRST(&group->send_queue);
3714 int error = ctl_enqueuembuf(g_flow_divert_kctl_ref, group->ctl_unit, next_packet, CTL_DATA_EOR);
3715 if (error) {
3716 FDLOG(LOG_DEBUG, &nil_pcb, "ctl_enqueuembuf returned an error: %d", error);
3717 OSTestAndSet(GROUP_BIT_CTL_ENQUEUE_BLOCKED, &group->atomic_bits);
3718 lck_rw_done(&group->lck);
3719 lck_rw_done(&g_flow_divert_group_lck);
3720 return;
3721 }
3722 MBUFQ_DEQUEUE(&group->send_queue, next_packet);
3723 }
3724
3725 SLIST_INIT(&tmp_list);
3726
3727 RB_FOREACH(fd_cb, fd_pcb_tree, &group->pcb_tree) {
3728 FDRETAIN(fd_cb);
3729 SLIST_INSERT_HEAD(&tmp_list, fd_cb, tmp_list_entry);
3730 }
3731
3732 lck_rw_done(&group->lck);
3733
3734 SLIST_FOREACH(fd_cb, &tmp_list, tmp_list_entry) {
3735 FDLOCK(fd_cb);
3736 if (fd_cb->so != NULL) {
3737 socket_lock(fd_cb->so, 0);
3738 if (fd_cb->group != NULL) {
3739 flow_divert_send_buffered_data(fd_cb, FALSE);
3740 }
3741 socket_unlock(fd_cb->so, 0);
3742 }
3743 FDUNLOCK(fd_cb);
3744 FDRELEASE(fd_cb);
3745 }
3746
3747 lck_rw_done(&g_flow_divert_group_lck);
3748 }
3749 }
3750
3751 static int
3752 flow_divert_kctl_init(void)
3753 {
3754 struct kern_ctl_reg ctl_reg;
3755 int result;
3756
3757 memset(&ctl_reg, 0, sizeof(ctl_reg));
3758
3759 strlcpy(ctl_reg.ctl_name, FLOW_DIVERT_CONTROL_NAME, sizeof(ctl_reg.ctl_name));
3760 ctl_reg.ctl_name[sizeof(ctl_reg.ctl_name)-1] = '\0';
3761 ctl_reg.ctl_flags = CTL_FLAG_PRIVILEGED | CTL_FLAG_REG_EXTENDED;
3762 ctl_reg.ctl_sendsize = FD_CTL_SENDBUFF_SIZE;
3763 ctl_reg.ctl_recvsize = FD_CTL_RCVBUFF_SIZE;
3764
3765 ctl_reg.ctl_connect = flow_divert_kctl_connect;
3766 ctl_reg.ctl_disconnect = flow_divert_kctl_disconnect;
3767 ctl_reg.ctl_send = flow_divert_kctl_send;
3768 ctl_reg.ctl_rcvd = flow_divert_kctl_rcvd;
3769
3770 result = ctl_register(&ctl_reg, &g_flow_divert_kctl_ref);
3771
3772 if (result) {
3773 FDLOG(LOG_ERR, &nil_pcb, "flow_divert_kctl_init - ctl_register failed: %d\n", result);
3774 return result;
3775 }
3776
3777 return 0;
3778 }
3779
3780 void
3781 flow_divert_init(void)
3782 {
3783 memset(&nil_pcb, 0, sizeof(nil_pcb));
3784 nil_pcb.log_level = LOG_NOTICE;
3785
3786 g_tcp_protosw = pffindproto(AF_INET, IPPROTO_TCP, SOCK_STREAM);
3787
3788 VERIFY(g_tcp_protosw != NULL);
3789
3790 memcpy(&g_flow_divert_in_protosw, g_tcp_protosw, sizeof(g_flow_divert_in_protosw));
3791 memcpy(&g_flow_divert_in_usrreqs, g_tcp_protosw->pr_usrreqs, sizeof(g_flow_divert_in_usrreqs));
3792
3793 g_flow_divert_in_usrreqs.pru_connect = flow_divert_connect_out;
3794 g_flow_divert_in_usrreqs.pru_connectx = flow_divert_connectx_out;
3795 g_flow_divert_in_usrreqs.pru_control = flow_divert_in_control;
3796 g_flow_divert_in_usrreqs.pru_disconnect = flow_divert_close;
3797 g_flow_divert_in_usrreqs.pru_disconnectx = flow_divert_disconnectx;
3798 g_flow_divert_in_usrreqs.pru_peeraddr = flow_divert_getpeername;
3799 g_flow_divert_in_usrreqs.pru_rcvd = flow_divert_rcvd;
3800 g_flow_divert_in_usrreqs.pru_send = flow_divert_data_out;
3801 g_flow_divert_in_usrreqs.pru_shutdown = flow_divert_shutdown;
3802 g_flow_divert_in_usrreqs.pru_sockaddr = flow_divert_getsockaddr;
3803 g_flow_divert_in_usrreqs.pru_preconnect = flow_divert_preconnect;
3804
3805 g_flow_divert_in_protosw.pr_usrreqs = &g_flow_divert_in_usrreqs;
3806 g_flow_divert_in_protosw.pr_ctloutput = flow_divert_ctloutput;
3807
3808 /*
3809 * Socket filters shouldn't attach/detach to/from this protosw
3810 * since pr_protosw is to be used instead, which points to the
3811 * real protocol; if they do, it is a bug and we should panic.
3812 */
3813 g_flow_divert_in_protosw.pr_filter_head.tqh_first =
3814 (struct socket_filter *)(uintptr_t)0xdeadbeefdeadbeef;
3815 g_flow_divert_in_protosw.pr_filter_head.tqh_last =
3816 (struct socket_filter **)(uintptr_t)0xdeadbeefdeadbeef;
3817
3818 /* UDP */
3819 g_udp_protosw = pffindproto(AF_INET, IPPROTO_UDP, SOCK_DGRAM);
3820 VERIFY(g_udp_protosw != NULL);
3821
3822 memcpy(&g_flow_divert_in_udp_protosw, g_udp_protosw, sizeof(g_flow_divert_in_udp_protosw));
3823 memcpy(&g_flow_divert_in_udp_usrreqs, g_udp_protosw->pr_usrreqs, sizeof(g_flow_divert_in_udp_usrreqs));
3824
3825 g_flow_divert_in_udp_usrreqs.pru_connect = flow_divert_connect_out;
3826 g_flow_divert_in_udp_usrreqs.pru_connectx = flow_divert_connectx_out;
3827 g_flow_divert_in_udp_usrreqs.pru_control = flow_divert_in_control;
3828 g_flow_divert_in_udp_usrreqs.pru_disconnect = flow_divert_close;
3829 g_flow_divert_in_udp_usrreqs.pru_disconnectx = flow_divert_disconnectx;
3830 g_flow_divert_in_udp_usrreqs.pru_peeraddr = flow_divert_getpeername;
3831 g_flow_divert_in_udp_usrreqs.pru_rcvd = flow_divert_rcvd;
3832 g_flow_divert_in_udp_usrreqs.pru_send = flow_divert_data_out;
3833 g_flow_divert_in_udp_usrreqs.pru_shutdown = flow_divert_shutdown;
3834 g_flow_divert_in_udp_usrreqs.pru_sockaddr = flow_divert_getsockaddr;
3835 g_flow_divert_in_udp_usrreqs.pru_sosend_list = pru_sosend_list_notsupp;
3836 g_flow_divert_in_udp_usrreqs.pru_soreceive_list = pru_soreceive_list_notsupp;
3837 g_flow_divert_in_udp_usrreqs.pru_preconnect = flow_divert_preconnect;
3838
3839 g_flow_divert_in_udp_protosw.pr_usrreqs = &g_flow_divert_in_usrreqs;
3840 g_flow_divert_in_udp_protosw.pr_ctloutput = flow_divert_ctloutput;
3841
3842 /*
3843 * Socket filters shouldn't attach/detach to/from this protosw
3844 * since pr_protosw is to be used instead, which points to the
3845 * real protocol; if they do, it is a bug and we should panic.
3846 */
3847 g_flow_divert_in_udp_protosw.pr_filter_head.tqh_first =
3848 (struct socket_filter *)(uintptr_t)0xdeadbeefdeadbeef;
3849 g_flow_divert_in_udp_protosw.pr_filter_head.tqh_last =
3850 (struct socket_filter **)(uintptr_t)0xdeadbeefdeadbeef;
3851
3852 #if INET6
3853 g_tcp6_protosw = (struct ip6protosw *)pffindproto(AF_INET6, IPPROTO_TCP, SOCK_STREAM);
3854
3855 VERIFY(g_tcp6_protosw != NULL);
3856
3857 memcpy(&g_flow_divert_in6_protosw, g_tcp6_protosw, sizeof(g_flow_divert_in6_protosw));
3858 memcpy(&g_flow_divert_in6_usrreqs, g_tcp6_protosw->pr_usrreqs, sizeof(g_flow_divert_in6_usrreqs));
3859
3860 g_flow_divert_in6_usrreqs.pru_connect = flow_divert_connect_out;
3861 g_flow_divert_in6_usrreqs.pru_connectx = flow_divert_connectx6_out;
3862 g_flow_divert_in6_usrreqs.pru_control = flow_divert_in6_control;
3863 g_flow_divert_in6_usrreqs.pru_disconnect = flow_divert_close;
3864 g_flow_divert_in6_usrreqs.pru_disconnectx = flow_divert_disconnectx;
3865 g_flow_divert_in6_usrreqs.pru_peeraddr = flow_divert_getpeername;
3866 g_flow_divert_in6_usrreqs.pru_rcvd = flow_divert_rcvd;
3867 g_flow_divert_in6_usrreqs.pru_send = flow_divert_data_out;
3868 g_flow_divert_in6_usrreqs.pru_shutdown = flow_divert_shutdown;
3869 g_flow_divert_in6_usrreqs.pru_sockaddr = flow_divert_getsockaddr;
3870 g_flow_divert_in6_usrreqs.pru_preconnect = flow_divert_preconnect;
3871
3872 g_flow_divert_in6_protosw.pr_usrreqs = &g_flow_divert_in6_usrreqs;
3873 g_flow_divert_in6_protosw.pr_ctloutput = flow_divert_ctloutput;
3874 /*
3875 * Socket filters shouldn't attach/detach to/from this protosw
3876 * since pr_protosw is to be used instead, which points to the
3877 * real protocol; if they do, it is a bug and we should panic.
3878 */
3879 g_flow_divert_in6_protosw.pr_filter_head.tqh_first =
3880 (struct socket_filter *)(uintptr_t)0xdeadbeefdeadbeef;
3881 g_flow_divert_in6_protosw.pr_filter_head.tqh_last =
3882 (struct socket_filter **)(uintptr_t)0xdeadbeefdeadbeef;
3883
3884 /* UDP6 */
3885 g_udp6_protosw = (struct ip6protosw *)pffindproto(AF_INET6, IPPROTO_UDP, SOCK_DGRAM);
3886
3887 VERIFY(g_udp6_protosw != NULL);
3888
3889 memcpy(&g_flow_divert_in6_udp_protosw, g_udp6_protosw, sizeof(g_flow_divert_in6_udp_protosw));
3890 memcpy(&g_flow_divert_in6_udp_usrreqs, g_udp6_protosw->pr_usrreqs, sizeof(g_flow_divert_in6_udp_usrreqs));
3891
3892 g_flow_divert_in6_udp_usrreqs.pru_connect = flow_divert_connect_out;
3893 g_flow_divert_in6_udp_usrreqs.pru_connectx = flow_divert_connectx6_out;
3894 g_flow_divert_in6_udp_usrreqs.pru_control = flow_divert_in6_control;
3895 g_flow_divert_in6_udp_usrreqs.pru_disconnect = flow_divert_close;
3896 g_flow_divert_in6_udp_usrreqs.pru_disconnectx = flow_divert_disconnectx;
3897 g_flow_divert_in6_udp_usrreqs.pru_peeraddr = flow_divert_getpeername;
3898 g_flow_divert_in6_udp_usrreqs.pru_rcvd = flow_divert_rcvd;
3899 g_flow_divert_in6_udp_usrreqs.pru_send = flow_divert_data_out;
3900 g_flow_divert_in6_udp_usrreqs.pru_shutdown = flow_divert_shutdown;
3901 g_flow_divert_in6_udp_usrreqs.pru_sockaddr = flow_divert_getsockaddr;
3902 g_flow_divert_in6_udp_usrreqs.pru_sosend_list = pru_sosend_list_notsupp;
3903 g_flow_divert_in6_udp_usrreqs.pru_soreceive_list = pru_soreceive_list_notsupp;
3904 g_flow_divert_in6_udp_usrreqs.pru_preconnect = flow_divert_preconnect;
3905
3906 g_flow_divert_in6_udp_protosw.pr_usrreqs = &g_flow_divert_in6_udp_usrreqs;
3907 g_flow_divert_in6_udp_protosw.pr_ctloutput = flow_divert_ctloutput;
3908 /*
3909 * Socket filters shouldn't attach/detach to/from this protosw
3910 * since pr_protosw is to be used instead, which points to the
3911 * real protocol; if they do, it is a bug and we should panic.
3912 */
3913 g_flow_divert_in6_udp_protosw.pr_filter_head.tqh_first =
3914 (struct socket_filter *)(uintptr_t)0xdeadbeefdeadbeef;
3915 g_flow_divert_in6_udp_protosw.pr_filter_head.tqh_last =
3916 (struct socket_filter **)(uintptr_t)0xdeadbeefdeadbeef;
3917 #endif /* INET6 */
3918
3919 flow_divert_grp_attr = lck_grp_attr_alloc_init();
3920 if (flow_divert_grp_attr == NULL) {
3921 FDLOG0(LOG_ERR, &nil_pcb, "lck_grp_attr_alloc_init failed");
3922 g_init_result = ENOMEM;
3923 goto done;
3924 }
3925
3926 flow_divert_mtx_grp = lck_grp_alloc_init(FLOW_DIVERT_CONTROL_NAME, flow_divert_grp_attr);
3927 if (flow_divert_mtx_grp == NULL) {
3928 FDLOG0(LOG_ERR, &nil_pcb, "lck_grp_alloc_init failed");
3929 g_init_result = ENOMEM;
3930 goto done;
3931 }
3932
3933 flow_divert_mtx_attr = lck_attr_alloc_init();
3934 if (flow_divert_mtx_attr == NULL) {
3935 FDLOG0(LOG_ERR, &nil_pcb, "lck_attr_alloc_init failed");
3936 g_init_result = ENOMEM;
3937 goto done;
3938 }
3939
3940 g_init_result = flow_divert_kctl_init();
3941 if (g_init_result) {
3942 goto done;
3943 }
3944
3945 lck_rw_init(&g_flow_divert_group_lck, flow_divert_mtx_grp, flow_divert_mtx_attr);
3946
3947 done:
3948 if (g_init_result != 0) {
3949 if (flow_divert_mtx_attr != NULL) {
3950 lck_attr_free(flow_divert_mtx_attr);
3951 flow_divert_mtx_attr = NULL;
3952 }
3953 if (flow_divert_mtx_grp != NULL) {
3954 lck_grp_free(flow_divert_mtx_grp);
3955 flow_divert_mtx_grp = NULL;
3956 }
3957 if (flow_divert_grp_attr != NULL) {
3958 lck_grp_attr_free(flow_divert_grp_attr);
3959 flow_divert_grp_attr = NULL;
3960 }
3961
3962 if (g_flow_divert_kctl_ref != NULL) {
3963 ctl_deregister(g_flow_divert_kctl_ref);
3964 g_flow_divert_kctl_ref = NULL;
3965 }
3966 }
3967 }
mirror of XNU