1/*
2 * Copyright (c) 2012-2018 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
30#include <sys/systm.h>
31#include <sys/kern_control.h>
32#include <net/kpi_protocol.h>
33#include <net/kpi_interface.h>
34#include <sys/socket.h>
35#include <sys/socketvar.h>
36#include <net/if.h>
37#include <net/if_types.h>
38#include <net/bpf.h>
39#include <net/if_ipsec.h>
40#include <sys/mbuf.h>
41#include <sys/sockio.h>
42#include <netinet/in.h>
43#include <netinet/ip6.h>
44#include <netinet6/in6_var.h>
45#include <netinet6/ip6_var.h>
46#include <sys/kauth.h>
47#include <netinet6/ipsec.h>
48#include <netinet6/ipsec6.h>
49#include <netinet6/esp.h>
50#include <netinet6/esp6.h>
51#include <netinet/ip.h>
52#include <net/flowadv.h>
53#include <net/necp.h>
54#include <netkey/key.h>
55#include <net/pktap.h>
56#include <kern/zalloc.h>
57
58#define IPSEC_NEXUS 0
59
60extern int net_qos_policy_restricted;
61extern int net_qos_policy_restrict_avapps;
62
63/* Kernel Control functions */
64static errno_t ipsec_ctl_bind(kern_ctl_ref kctlref, struct sockaddr_ctl *sac,
65 void **unitinfo);
66static errno_t ipsec_ctl_connect(kern_ctl_ref kctlref, struct sockaddr_ctl *sac,
67 void **unitinfo);
68static errno_t ipsec_ctl_disconnect(kern_ctl_ref kctlref, u_int32_t unit,
69 void *unitinfo);
70static errno_t ipsec_ctl_send(kern_ctl_ref kctlref, u_int32_t unit,
71 void *unitinfo, mbuf_t m, int flags);
72static errno_t ipsec_ctl_getopt(kern_ctl_ref kctlref, u_int32_t unit, void *unitinfo,
73 int opt, void *data, size_t *len);
74static errno_t ipsec_ctl_setopt(kern_ctl_ref kctlref, u_int32_t unit, void *unitinfo,
75 int opt, void *data, size_t len);
76
77/* Network Interface functions */
78static void ipsec_start(ifnet_t interface);
79static errno_t ipsec_output(ifnet_t interface, mbuf_t data);
80static errno_t ipsec_demux(ifnet_t interface, mbuf_t data, char *frame_header,
81 protocol_family_t *protocol);
82static errno_t ipsec_add_proto(ifnet_t interface, protocol_family_t protocol,
83 const struct ifnet_demux_desc *demux_array,
84 u_int32_t demux_count);
85static errno_t ipsec_del_proto(ifnet_t interface, protocol_family_t protocol);
86static errno_t ipsec_ioctl(ifnet_t interface, u_long cmd, void *data);
87static void ipsec_detached(ifnet_t interface);
88
89/* Protocol handlers */
90static errno_t ipsec_attach_proto(ifnet_t interface, protocol_family_t proto);
91static errno_t ipsec_proto_input(ifnet_t interface, protocol_family_t protocol,
92 mbuf_t m, char *frame_header);
93static errno_t ipsec_proto_pre_output(ifnet_t interface, protocol_family_t protocol,
94 mbuf_t *packet, const struct sockaddr *dest, void *route,
95 char *frame_type, char *link_layer_dest);
96
97static kern_ctl_ref ipsec_kctlref;
98static u_int32_t ipsec_family;
99static lck_attr_t *ipsec_lck_attr;
100static lck_grp_attr_t *ipsec_lck_grp_attr;
101static lck_grp_t *ipsec_lck_grp;
102static lck_mtx_t ipsec_lock;
103
104#if IPSEC_NEXUS
105
106SYSCTL_DECL(_net_ipsec);
107SYSCTL_NODE(_net, OID_AUTO, ipsec, CTLFLAG_RW | CTLFLAG_LOCKED, 0, "IPsec");
108static int if_ipsec_verify_interface_creation = 0;
109SYSCTL_INT(_net_ipsec, OID_AUTO, verify_interface_creation, CTLFLAG_RW | CTLFLAG_LOCKED, &if_ipsec_verify_interface_creation, 0, "");
110
111#define IPSEC_IF_VERIFY(_e) if (__improbable(if_ipsec_verify_interface_creation)) { VERIFY(_e); }
112
113#define IPSEC_IF_DEFAULT_SLOT_SIZE 2048
114#define IPSEC_IF_DEFAULT_RING_SIZE 64
115#define IPSEC_IF_DEFAULT_TX_FSW_RING_SIZE 64
116#define IPSEC_IF_DEFAULT_RX_FSW_RING_SIZE 128
117#define IPSEC_IF_DEFAULT_BUF_SEG_SIZE skmem_usr_buf_seg_size
118
119#define IPSEC_IF_MIN_RING_SIZE 16
120#define IPSEC_IF_MAX_RING_SIZE 1024
121
122#define IPSEC_IF_MIN_SLOT_SIZE 1024
123#define IPSEC_IF_MAX_SLOT_SIZE 4096
124
125static int sysctl_if_ipsec_ring_size SYSCTL_HANDLER_ARGS;
126static int sysctl_if_ipsec_tx_fsw_ring_size SYSCTL_HANDLER_ARGS;
127static int sysctl_if_ipsec_rx_fsw_ring_size SYSCTL_HANDLER_ARGS;
128
129static int if_ipsec_ring_size = IPSEC_IF_DEFAULT_RING_SIZE;
130static int if_ipsec_tx_fsw_ring_size = IPSEC_IF_DEFAULT_TX_FSW_RING_SIZE;
131static int if_ipsec_rx_fsw_ring_size = IPSEC_IF_DEFAULT_RX_FSW_RING_SIZE;
132
133SYSCTL_PROC(_net_ipsec, OID_AUTO, ring_size, CTLTYPE_INT | CTLFLAG_LOCKED | CTLFLAG_RW,
134 &if_ipsec_ring_size, IPSEC_IF_DEFAULT_RING_SIZE, &sysctl_if_ipsec_ring_size, "I", "");
135SYSCTL_PROC(_net_ipsec, OID_AUTO, tx_fsw_ring_size, CTLTYPE_INT | CTLFLAG_LOCKED | CTLFLAG_RW,
136 &if_ipsec_tx_fsw_ring_size, IPSEC_IF_DEFAULT_TX_FSW_RING_SIZE, &sysctl_if_ipsec_tx_fsw_ring_size, "I", "");
137SYSCTL_PROC(_net_ipsec, OID_AUTO, rx_fsw_ring_size, CTLTYPE_INT | CTLFLAG_LOCKED | CTLFLAG_RW,
138 &if_ipsec_rx_fsw_ring_size, IPSEC_IF_DEFAULT_RX_FSW_RING_SIZE, &sysctl_if_ipsec_rx_fsw_ring_size, "I", "");
139
140static errno_t
141ipsec_register_nexus(void);
142
143typedef struct ipsec_nx {
144 uuid_t if_provider;
145 uuid_t if_instance;
146 uuid_t ms_provider;
147 uuid_t ms_instance;
148 uuid_t ms_device;
149 uuid_t ms_host;
150 uuid_t ms_agent;
151} *ipsec_nx_t;
152
153static nexus_controller_t ipsec_ncd;
154static int ipsec_ncd_refcount;
155static uuid_t ipsec_kpipe_uuid;
156
157#endif // IPSEC_NEXUS
158
159/* Control block allocated for each kernel control connection */
160struct ipsec_pcb {
161 TAILQ_ENTRY(ipsec_pcb) ipsec_chain;
162 kern_ctl_ref ipsec_ctlref;
163 ifnet_t ipsec_ifp;
164 u_int32_t ipsec_unit;
165 u_int32_t ipsec_unique_id;
166 u_int32_t ipsec_flags;
167 u_int32_t ipsec_input_frag_size;
168 bool ipsec_frag_size_set;
169 int ipsec_ext_ifdata_stats;
170 mbuf_svc_class_t ipsec_output_service_class;
171 char ipsec_if_xname[IFXNAMSIZ];
172 char ipsec_unique_name[IFXNAMSIZ];
173 // PCB lock protects state fields, like ipsec_kpipe_enabled
174 decl_lck_rw_data(, ipsec_pcb_lock);
175
176#if IPSEC_NEXUS
177 lck_mtx_t ipsec_input_chain_lock;
178 struct mbuf * ipsec_input_chain;
179 struct mbuf * ipsec_input_chain_last;
180 // Input chain lock protects the list of input mbufs
181 // The input chain lock must be taken AFTER the PCB lock if both are held
182 struct ipsec_nx ipsec_nx;
183 int ipsec_kpipe_enabled;
184 uuid_t ipsec_kpipe_uuid;
185 void * ipsec_kpipe_rxring;
186 void * ipsec_kpipe_txring;
187 kern_pbufpool_t ipsec_kpipe_pp;
188
189 kern_nexus_t ipsec_netif_nexus;
190 kern_pbufpool_t ipsec_netif_pp;
191 void * ipsec_netif_rxring;
192 void * ipsec_netif_txring;
193 uint64_t ipsec_netif_txring_size;
194
195 u_int32_t ipsec_slot_size;
196 u_int32_t ipsec_netif_ring_size;
197 u_int32_t ipsec_tx_fsw_ring_size;
198 u_int32_t ipsec_rx_fsw_ring_size;
199 bool ipsec_use_netif;
200 bool ipsec_needs_netagent;
201#endif // IPSEC_NEXUS
202};
203
204TAILQ_HEAD(ipsec_list, ipsec_pcb) ipsec_head;
205
206#define IPSEC_PCB_ZONE_MAX 32
207#define IPSEC_PCB_ZONE_NAME "net.if_ipsec"
208
209static unsigned int ipsec_pcb_size; /* size of zone element */
210static struct zone *ipsec_pcb_zone; /* zone for ipsec_pcb */
211
212#define IPSECQ_MAXLEN 256
213
214#if IPSEC_NEXUS
215static int
216sysctl_if_ipsec_ring_size SYSCTL_HANDLER_ARGS
217{
218#pragma unused(arg1, arg2)
219 int value = if_ipsec_ring_size;
220
221 int error = sysctl_handle_int(oidp, &value, 0, req);
222 if (error || !req->newptr) {
223 return (error);
224 }
225
226 if (value < IPSEC_IF_MIN_RING_SIZE ||
227 value > IPSEC_IF_MAX_RING_SIZE) {
228 return (EINVAL);
229 }
230
231 if_ipsec_ring_size = value;
232
233 return (0);
234}
235
236static int
237sysctl_if_ipsec_tx_fsw_ring_size SYSCTL_HANDLER_ARGS
238{
239#pragma unused(arg1, arg2)
240 int value = if_ipsec_tx_fsw_ring_size;
241
242 int error = sysctl_handle_int(oidp, &value, 0, req);
243 if (error || !req->newptr) {
244 return (error);
245 }
246
247 if (value < IPSEC_IF_MIN_RING_SIZE ||
248 value > IPSEC_IF_MAX_RING_SIZE) {
249 return (EINVAL);
250 }
251
252 if_ipsec_tx_fsw_ring_size = value;
253
254 return (0);
255}
256
257static int
258sysctl_if_ipsec_rx_fsw_ring_size SYSCTL_HANDLER_ARGS
259{
260#pragma unused(arg1, arg2)
261 int value = if_ipsec_rx_fsw_ring_size;
262
263 int error = sysctl_handle_int(oidp, &value, 0, req);
264 if (error || !req->newptr) {
265 return (error);
266 }
267
268 if (value < IPSEC_IF_MIN_RING_SIZE ||
269 value > IPSEC_IF_MAX_RING_SIZE) {
270 return (EINVAL);
271 }
272
273 if_ipsec_rx_fsw_ring_size = value;
274
275 return (0);
276}
277#endif // IPSEC_NEXUS
278
279errno_t
280ipsec_register_control(void)
281{
282 struct kern_ctl_reg kern_ctl;
283 errno_t result = 0;
284
285 /* Find a unique value for our interface family */
286 result = mbuf_tag_id_find(IPSEC_CONTROL_NAME, &ipsec_family);
287 if (result != 0) {
288 printf("ipsec_register_control - mbuf_tag_id_find_internal failed: %d\n", result);
289 return result;
290 }
291
292 ipsec_pcb_size = sizeof(struct ipsec_pcb);
293 ipsec_pcb_zone = zinit(ipsec_pcb_size,
294 IPSEC_PCB_ZONE_MAX * ipsec_pcb_size,
295 0, IPSEC_PCB_ZONE_NAME);
296 if (ipsec_pcb_zone == NULL) {
297 printf("ipsec_register_control - zinit(ipsec_pcb) failed");
298 return ENOMEM;
299 }
300
301#if IPSEC_NEXUS
302 ipsec_register_nexus();
303#endif // IPSEC_NEXUS
304
305 TAILQ_INIT(&ipsec_head);
306
307 bzero(&kern_ctl, sizeof(kern_ctl));
308 strlcpy(kern_ctl.ctl_name, IPSEC_CONTROL_NAME, sizeof(kern_ctl.ctl_name));
309 kern_ctl.ctl_name[sizeof(kern_ctl.ctl_name) - 1] = 0;
310 kern_ctl.ctl_flags = CTL_FLAG_PRIVILEGED; /* Require root */
311 kern_ctl.ctl_sendsize = 64 * 1024;
312 kern_ctl.ctl_recvsize = 64 * 1024;
313 kern_ctl.ctl_bind = ipsec_ctl_bind;
314 kern_ctl.ctl_connect = ipsec_ctl_connect;
315 kern_ctl.ctl_disconnect = ipsec_ctl_disconnect;
316 kern_ctl.ctl_send = ipsec_ctl_send;
317 kern_ctl.ctl_setopt = ipsec_ctl_setopt;
318 kern_ctl.ctl_getopt = ipsec_ctl_getopt;
319
320 result = ctl_register(&kern_ctl, &ipsec_kctlref);
321 if (result != 0) {
322 printf("ipsec_register_control - ctl_register failed: %d\n", result);
323 return result;
324 }
325
326 /* Register the protocol plumbers */
327 if ((result = proto_register_plumber(PF_INET, ipsec_family,
328 ipsec_attach_proto, NULL)) != 0) {
329 printf("ipsec_register_control - proto_register_plumber(PF_INET, %d) failed: %d\n",
330 ipsec_family, result);
331 ctl_deregister(ipsec_kctlref);
332 return result;
333 }
334
335 /* Register the protocol plumbers */
336 if ((result = proto_register_plumber(PF_INET6, ipsec_family,
337 ipsec_attach_proto, NULL)) != 0) {
338 proto_unregister_plumber(PF_INET, ipsec_family);
339 ctl_deregister(ipsec_kctlref);
340 printf("ipsec_register_control - proto_register_plumber(PF_INET6, %d) failed: %d\n",
341 ipsec_family, result);
342 return result;
343 }
344
345 ipsec_lck_attr = lck_attr_alloc_init();
346 ipsec_lck_grp_attr = lck_grp_attr_alloc_init();
347 ipsec_lck_grp = lck_grp_alloc_init("ipsec", ipsec_lck_grp_attr);
348 lck_mtx_init(&ipsec_lock, ipsec_lck_grp, ipsec_lck_attr);
349
350 return 0;
351}
352
353/* Helpers */
354int
355ipsec_interface_isvalid (ifnet_t interface)
356{
357 struct ipsec_pcb *pcb = NULL;
358
359 if (interface == NULL)
360 return 0;
361
362 pcb = ifnet_softc(interface);
363
364 if (pcb == NULL)
365 return 0;
366
367 /* When ctl disconnects, ipsec_unit is set to 0 */
368 if (pcb->ipsec_unit == 0)
369 return 0;
370
371 return 1;
372}
373
374#if IPSEC_NEXUS
375boolean_t
376ipsec_interface_needs_netagent(ifnet_t interface)
377{
378 struct ipsec_pcb *pcb = NULL;
379
380 if (interface == NULL) {
381 return (FALSE);
382 }
383
384 pcb = ifnet_softc(interface);
385
386 if (pcb == NULL) {
387 return (FALSE);
388 }
389
390 return (pcb->ipsec_needs_netagent == true);
391}
392#endif // IPSEC_NEXUS
393
394static errno_t
395ipsec_ifnet_set_attrs(ifnet_t ifp)
396{
397 /* Set flags and additional information. */
398 ifnet_set_mtu(ifp, 1500);
399 ifnet_set_flags(ifp, IFF_UP | IFF_MULTICAST | IFF_POINTOPOINT, 0xffff);
400
401 /* The interface must generate its own IPv6 LinkLocal address,
402 * if possible following the recommendation of RFC2472 to the 64bit interface ID
403 */
404 ifnet_set_eflags(ifp, IFEF_NOAUTOIPV6LL, IFEF_NOAUTOIPV6LL);
405
406#if !IPSEC_NEXUS
407 /* Reset the stats in case as the interface may have been recycled */
408 struct ifnet_stats_param stats;
409 bzero(&stats, sizeof(struct ifnet_stats_param));
410 ifnet_set_stat(ifp, &stats);
411#endif // !IPSEC_NEXUS
412
413 return (0);
414}
415
416#if IPSEC_NEXUS
417
418static uuid_t ipsec_nx_dom_prov;
419
420static errno_t
421ipsec_nxdp_init(__unused kern_nexus_domain_provider_t domprov)
422{
423 return 0;
424}
425
426static void
427ipsec_nxdp_fini(__unused kern_nexus_domain_provider_t domprov)
428{
429 // Ignore
430}
431
432static errno_t
433ipsec_register_nexus(void)
434{
435 const struct kern_nexus_domain_provider_init dp_init = {
436 .nxdpi_version = KERN_NEXUS_DOMAIN_PROVIDER_CURRENT_VERSION,
437 .nxdpi_flags = 0,
438 .nxdpi_init = ipsec_nxdp_init,
439 .nxdpi_fini = ipsec_nxdp_fini
440 };
441 errno_t err = 0;
442
443 /* ipsec_nxdp_init() is called before this function returns */
444 err = kern_nexus_register_domain_provider(NEXUS_TYPE_NET_IF,
445 (const uint8_t *) "com.apple.ipsec",
446 &dp_init, sizeof(dp_init),
447 &ipsec_nx_dom_prov);
448 if (err != 0) {
449 printf("%s: failed to register domain provider\n", __func__);
450 return (err);
451 }
452 return (0);
453}
454
455static errno_t
456ipsec_netif_prepare(kern_nexus_t nexus, ifnet_t ifp)
457{
458 struct ipsec_pcb *pcb = kern_nexus_get_context(nexus);
459 pcb->ipsec_netif_nexus = nexus;
460 return (ipsec_ifnet_set_attrs(ifp));
461}
462
463static errno_t
464ipsec_nexus_pre_connect(kern_nexus_provider_t nxprov,
465 proc_t p, kern_nexus_t nexus,
466 nexus_port_t nexus_port, kern_channel_t channel, void **ch_ctx)
467{
468#pragma unused(nxprov, p)
469#pragma unused(nexus, nexus_port, channel, ch_ctx)
470 return (0);
471}
472
473static errno_t
474ipsec_nexus_connected(kern_nexus_provider_t nxprov, kern_nexus_t nexus,
475 kern_channel_t channel)
476{
477#pragma unused(nxprov, channel)
478 struct ipsec_pcb *pcb = kern_nexus_get_context(nexus);
479 boolean_t ok = ifnet_is_attached(pcb->ipsec_ifp, 1);
480 return (ok ? 0 : ENXIO);
481}
482
483static void
484ipsec_nexus_pre_disconnect(kern_nexus_provider_t nxprov, kern_nexus_t nexus,
485 kern_channel_t channel)
486{
487#pragma unused(nxprov, nexus, channel)
488}
489
490static void
491ipsec_netif_pre_disconnect(kern_nexus_provider_t nxprov, kern_nexus_t nexus,
492 kern_channel_t channel)
493{
494#pragma unused(nxprov, nexus, channel)
495}
496
497static void
498ipsec_nexus_disconnected(kern_nexus_provider_t nxprov, kern_nexus_t nexus,
499 kern_channel_t channel)
500{
501#pragma unused(nxprov, channel)
502 struct ipsec_pcb *pcb = kern_nexus_get_context(nexus);
503 if (pcb->ipsec_netif_nexus == nexus) {
504 pcb->ipsec_netif_nexus = NULL;
505 }
506 ifnet_decr_iorefcnt(pcb->ipsec_ifp);
507}
508
509static errno_t
510ipsec_kpipe_ring_init(kern_nexus_provider_t nxprov, kern_nexus_t nexus,
511 kern_channel_t channel, kern_channel_ring_t ring, boolean_t is_tx_ring,
512 void **ring_ctx)
513{
514#pragma unused(nxprov)
515#pragma unused(channel)
516#pragma unused(ring_ctx)
517 struct ipsec_pcb *pcb = kern_nexus_get_context(nexus);
518 if (!is_tx_ring) {
519 VERIFY(pcb->ipsec_kpipe_rxring == NULL);
520 pcb->ipsec_kpipe_rxring = ring;
521 } else {
522 VERIFY(pcb->ipsec_kpipe_txring == NULL);
523 pcb->ipsec_kpipe_txring = ring;
524 }
525 return 0;
526}
527
528static void
529ipsec_kpipe_ring_fini(kern_nexus_provider_t nxprov, kern_nexus_t nexus,
530 kern_channel_ring_t ring)
531{
532#pragma unused(nxprov)
533 struct ipsec_pcb *pcb = kern_nexus_get_context(nexus);
534 if (pcb->ipsec_kpipe_rxring == ring) {
535 pcb->ipsec_kpipe_rxring = NULL;
536 } else if (pcb->ipsec_kpipe_txring == ring) {
537 pcb->ipsec_kpipe_txring = NULL;
538 }
539}
540
541static errno_t
542ipsec_kpipe_sync_tx(kern_nexus_provider_t nxprov, kern_nexus_t nexus,
543 kern_channel_ring_t tx_ring, uint32_t flags)
544{
545#pragma unused(nxprov)
546#pragma unused(flags)
547 struct ipsec_pcb *pcb = kern_nexus_get_context(nexus);
548
549 lck_rw_lock_shared(&pcb->ipsec_pcb_lock);
550 int channel_enabled = pcb->ipsec_kpipe_enabled;
551 if (!channel_enabled) {
552 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
553 return 0;
554 }
555
556 kern_channel_slot_t tx_slot = kern_channel_get_next_slot(tx_ring, NULL, NULL);
557 if (tx_slot == NULL) {
558 // Nothing to write, bail
559 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
560 return 0;
561 }
562
563 // Signal the netif ring to read
564 kern_channel_ring_t rx_ring = pcb->ipsec_netif_rxring;
565 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
566
567 if (rx_ring != NULL) {
568 kern_channel_notify(rx_ring, 0);
569 }
570 return 0;
571}
572
573static mbuf_t
574ipsec_encrypt_mbuf(ifnet_t interface,
575 mbuf_t data)
576{
577 struct ipsec_output_state ipsec_state;
578 int error = 0;
579 uint32_t af;
580
581 // Make sure this packet isn't looping through the interface
582 if (necp_get_last_interface_index_from_packet(data) == interface->if_index) {
583 error = -1;
584 goto ipsec_output_err;
585 }
586
587 // Mark the interface so NECP can evaluate tunnel policy
588 necp_mark_packet_from_interface(data, interface);
589
590 struct ip *ip = mtod(data, struct ip *);
591 u_int ip_version = ip->ip_v;
592
593 switch (ip_version) {
594 case 4: {
595 af = AF_INET;
596
597 memset(&ipsec_state, 0, sizeof(ipsec_state));
598 ipsec_state.m = data;
599 ipsec_state.dst = (struct sockaddr *)&ip->ip_dst;
600 memset(&ipsec_state.ro, 0, sizeof(ipsec_state.ro));
601
602 error = ipsec4_interface_output(&ipsec_state, interface);
603 if (error == 0 && ipsec_state.tunneled == 6) {
604 // Tunneled in IPv6 - packet is gone
605 // TODO: Don't lose mbuf
606 data = NULL;
607 goto done;
608 }
609
610 data = ipsec_state.m;
611 if (error || data == NULL) {
612 if (error) {
613 printf("ipsec_encrypt_mbuf: ipsec4_output error %d\n", error);
614 }
615 goto ipsec_output_err;
616 }
617 goto done;
618 }
619 case 6: {
620 af = AF_INET6;
621
622 data = ipsec6_splithdr(data);
623 if (data == NULL) {
624 printf("ipsec_encrypt_mbuf: ipsec6_splithdr returned NULL\n");
625 goto ipsec_output_err;
626 }
627
628 struct ip6_hdr *ip6 = mtod(data, struct ip6_hdr *);
629
630 memset(&ipsec_state, 0, sizeof(ipsec_state));
631 ipsec_state.m = data;
632 ipsec_state.dst = (struct sockaddr *)&ip6->ip6_dst;
633 memset(&ipsec_state.ro, 0, sizeof(ipsec_state.ro));
634
635 error = ipsec6_interface_output(&ipsec_state, interface, &ip6->ip6_nxt, ipsec_state.m);
636 if (error == 0 && ipsec_state.tunneled == 4) {
637 // Tunneled in IPv4 - packet is gone
638 // TODO: Don't lose mbuf
639 data = NULL;
640 goto done;
641 }
642 data = ipsec_state.m;
643 if (error || data == NULL) {
644 if (error) {
645 printf("ipsec_encrypt_mbuf: ipsec6_output error %d\n", error);
646 }
647 goto ipsec_output_err;
648 }
649 goto done;
650 }
651 default: {
652 printf("ipsec_encrypt_mbuf: Received unknown packet version %d\n", ip_version);
653 error = -1;
654 goto ipsec_output_err;
655 }
656 }
657
658done:
659 return data;
660
661ipsec_output_err:
662 if (data) {
663 mbuf_freem(data);
664 }
665 return NULL;
666}
667
668static errno_t
669ipsec_kpipe_sync_rx(kern_nexus_provider_t nxprov, kern_nexus_t nexus,
670 kern_channel_ring_t rx_ring, uint32_t flags)
671{
672#pragma unused(nxprov)
673#pragma unused(flags)
674 struct ipsec_pcb *pcb = kern_nexus_get_context(nexus);
675 struct kern_channel_ring_stat_increment rx_ring_stats;
676
677 lck_rw_lock_shared(&pcb->ipsec_pcb_lock);
678
679 int channel_enabled = pcb->ipsec_kpipe_enabled;
680 if (!channel_enabled) {
681 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
682 return 0;
683 }
684
685 // Reclaim user-released slots
686 (void) kern_channel_reclaim(rx_ring);
687
688 uint32_t avail = kern_channel_available_slot_count(rx_ring);
689 if (avail == 0) {
690 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
691 return 0;
692 }
693
694 kern_channel_ring_t tx_ring = pcb->ipsec_netif_txring;
695 if (tx_ring == NULL) {
696 // Net-If TX ring not set up yet, nothing to read
697 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
698 return 0;
699 }
700
701 struct netif_stats *nifs = &NX_NETIF_PRIVATE(pcb->ipsec_netif_nexus)->nif_stats;
702
703 // Unlock ipsec before entering ring
704 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
705
706 (void)kr_enter(tx_ring, TRUE);
707
708 // Lock again after entering and validate
709 lck_rw_lock_shared(&pcb->ipsec_pcb_lock);
710 if (tx_ring != pcb->ipsec_netif_txring) {
711 // Ring no longer valid
712 // Unlock first, then exit ring
713 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
714 kr_exit(tx_ring);
715 return 0;
716 }
717
718
719 struct kern_channel_ring_stat_increment tx_ring_stats;
720 bzero(&tx_ring_stats, sizeof(tx_ring_stats));
721 kern_channel_slot_t tx_pslot = NULL;
722 kern_channel_slot_t tx_slot = kern_channel_get_next_slot(tx_ring, NULL, NULL);
723 if (tx_slot == NULL) {
724 // Nothing to read, don't bother signalling
725 // Unlock first, then exit ring
726 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
727 kr_exit(tx_ring);
728 return 0;
729 }
730
731 struct kern_pbufpool *rx_pp = rx_ring->ckr_pp;
732 VERIFY(rx_pp != NULL);
733 bzero(&rx_ring_stats, sizeof(rx_ring_stats));
734 kern_channel_slot_t rx_pslot = NULL;
735 kern_channel_slot_t rx_slot = kern_channel_get_next_slot(rx_ring, NULL, NULL);
736
737 while (rx_slot != NULL && tx_slot != NULL) {
738 size_t length = 0;
739 mbuf_t data = NULL;
740 errno_t error = 0;
741
742 // Allocate rx packet
743 kern_packet_t rx_ph = 0;
744 error = kern_pbufpool_alloc_nosleep(rx_pp, 1, &rx_ph);
745 if (__improbable(error != 0)) {
746 printf("ipsec_kpipe_sync_rx %s: failed to allocate packet\n",
747 pcb->ipsec_ifp->if_xname);
748 break;
749 }
750
751 kern_packet_t tx_ph = kern_channel_slot_get_packet(tx_ring, tx_slot);
752
753 // Advance TX ring
754 tx_pslot = tx_slot;
755 tx_slot = kern_channel_get_next_slot(tx_ring, tx_slot, NULL);
756
757 if (tx_ph == 0) {
758 kern_pbufpool_free(rx_pp, rx_ph);
759 continue;
760 }
761
762 kern_buflet_t tx_buf = kern_packet_get_next_buflet(tx_ph, NULL);
763 VERIFY(tx_buf != NULL);
764 uint8_t *tx_baddr = kern_buflet_get_object_address(tx_buf);
765 VERIFY(tx_baddr != NULL);
766 tx_baddr += kern_buflet_get_data_offset(tx_buf);
767
768 bpf_tap_packet_out(pcb->ipsec_ifp, DLT_RAW, tx_ph, NULL, 0);
769
770 length = MIN(kern_packet_get_data_length(tx_ph),
771 pcb->ipsec_slot_size);
772
773 // Increment TX stats
774 tx_ring_stats.kcrsi_slots_transferred++;
775 tx_ring_stats.kcrsi_bytes_transferred += length;
776
777 if (length > 0) {
778 error = mbuf_gethdr(MBUF_DONTWAIT, MBUF_TYPE_HEADER, &data);
779 if (error == 0) {
780 error = mbuf_copyback(data, 0, length, tx_baddr, MBUF_DONTWAIT);
781 if (error == 0) {
782 // Encrypt and send packet
783 data = ipsec_encrypt_mbuf(pcb->ipsec_ifp, data);
784 } else {
785 printf("ipsec_kpipe_sync_rx %s - mbuf_copyback(%zu) error %d\n", pcb->ipsec_ifp->if_xname, length, error);
786 STATS_INC(nifs, NETIF_STATS_NOMEM_MBUF);
787 STATS_INC(nifs, NETIF_STATS_DROPPED);
788 mbuf_freem(data);
789 data = NULL;
790 }
791 } else {
792 printf("ipsec_kpipe_sync_rx %s - mbuf_gethdr error %d\n", pcb->ipsec_ifp->if_xname, error);
793 STATS_INC(nifs, NETIF_STATS_NOMEM_MBUF);
794 STATS_INC(nifs, NETIF_STATS_DROPPED);
795 }
796 } else {
797 printf("ipsec_kpipe_sync_rx %s - 0 length packet\n", pcb->ipsec_ifp->if_xname);
798 STATS_INC(nifs, NETIF_STATS_BADLEN);
799 STATS_INC(nifs, NETIF_STATS_DROPPED);
800 }
801
802 if (data == NULL) {
803 printf("ipsec_kpipe_sync_rx %s: no encrypted packet to send\n", pcb->ipsec_ifp->if_xname);
804 kern_pbufpool_free(rx_pp, rx_ph);
805 break;
806 }
807
808 length = mbuf_pkthdr_len(data);
809 if (length > rx_pp->pp_buflet_size) {
810 // Flush data
811 mbuf_freem(data);
812 kern_pbufpool_free(rx_pp, rx_ph);
813 printf("ipsec_kpipe_sync_rx %s: encrypted packet length %zu > %u\n",
814 pcb->ipsec_ifp->if_xname, length, rx_pp->pp_buflet_size);
815 continue;
816 }
817
818 // Fillout rx packet
819 kern_buflet_t rx_buf = kern_packet_get_next_buflet(rx_ph, NULL);
820 VERIFY(rx_buf != NULL);
821 void *rx_baddr = kern_buflet_get_object_address(rx_buf);
822 VERIFY(rx_baddr != NULL);
823
824 // Copy-in data from mbuf to buflet
825 mbuf_copydata(data, 0, length, (void *)rx_baddr);
826 kern_packet_clear_flow_uuid(rx_ph); // Zero flow id
827
828 // Finalize and attach the packet
829 error = kern_buflet_set_data_offset(rx_buf, 0);
830 VERIFY(error == 0);
831 error = kern_buflet_set_data_length(rx_buf, length);
832 VERIFY(error == 0);
833 error = kern_packet_finalize(rx_ph);
834 VERIFY(error == 0);
835 error = kern_channel_slot_attach_packet(rx_ring, rx_slot, rx_ph);
836 VERIFY(error == 0);
837
838 STATS_INC(nifs, NETIF_STATS_TXPKTS);
839 STATS_INC(nifs, NETIF_STATS_TXCOPY_DIRECT);
840
841 rx_ring_stats.kcrsi_slots_transferred++;
842 rx_ring_stats.kcrsi_bytes_transferred += length;
843
844 if (!pcb->ipsec_ext_ifdata_stats) {
845 ifnet_stat_increment_out(pcb->ipsec_ifp, 1, length, 0);
846 }
847
848 mbuf_freem(data);
849
850 rx_pslot = rx_slot;
851 rx_slot = kern_channel_get_next_slot(rx_ring, rx_slot, NULL);
852 }
853
854 if (rx_pslot) {
855 kern_channel_advance_slot(rx_ring, rx_pslot);
856 kern_channel_increment_ring_net_stats(rx_ring, pcb->ipsec_ifp, &rx_ring_stats);
857 }
858
859 if (tx_pslot) {
860 kern_channel_advance_slot(tx_ring, tx_pslot);
861 kern_channel_increment_ring_net_stats(tx_ring, pcb->ipsec_ifp, &tx_ring_stats);
862 (void)kern_channel_reclaim(tx_ring);
863 }
864
865 /* always reenable output */
866 errno_t error = ifnet_enable_output(pcb->ipsec_ifp);
867 if (error != 0) {
868 printf("ipsec_kpipe_sync_rx: ifnet_enable_output returned error %d\n", error);
869 }
870
871 // Unlock first, then exit ring
872 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
873
874 if (tx_pslot != NULL) {
875 kern_channel_notify(tx_ring, 0);
876 }
877 kr_exit(tx_ring);
878
879 return 0;
880}
881
882static errno_t
883ipsec_netif_ring_init(kern_nexus_provider_t nxprov, kern_nexus_t nexus,
884 kern_channel_t channel, kern_channel_ring_t ring, boolean_t is_tx_ring,
885 void **ring_ctx)
886{
887#pragma unused(nxprov)
888#pragma unused(channel)
889#pragma unused(ring_ctx)
890 struct ipsec_pcb *pcb = kern_nexus_get_context(nexus);
891 if (!is_tx_ring) {
892 VERIFY(pcb->ipsec_netif_rxring == NULL);
893 pcb->ipsec_netif_rxring = ring;
894 } else {
895 VERIFY(pcb->ipsec_netif_txring == NULL);
896 pcb->ipsec_netif_txring = ring;
897 }
898 return 0;
899}
900
901static void
902ipsec_netif_ring_fini(kern_nexus_provider_t nxprov, kern_nexus_t nexus,
903 kern_channel_ring_t ring)
904{
905#pragma unused(nxprov)
906 struct ipsec_pcb *pcb = kern_nexus_get_context(nexus);
907 if (pcb->ipsec_netif_rxring == ring) {
908 pcb->ipsec_netif_rxring = NULL;
909 } else if (pcb->ipsec_netif_txring == ring) {
910 pcb->ipsec_netif_txring = NULL;
911 }
912}
913
914static bool
915ipsec_netif_check_policy(mbuf_t data)
916{
917 necp_kernel_policy_result necp_result = 0;
918 necp_kernel_policy_result_parameter necp_result_parameter = {};
919 uint32_t necp_matched_policy_id = 0;
920
921 // This packet has been marked with IP level policy, do not mark again.
922 if (data && data->m_pkthdr.necp_mtag.necp_policy_id >= NECP_KERNEL_POLICY_ID_FIRST_VALID_IP) {
923 return (true);
924 }
925
926 size_t length = mbuf_pkthdr_len(data);
927 if (length < sizeof(struct ip)) {
928 return (false);
929 }
930
931 struct ip *ip = mtod(data, struct ip *);
932 u_int ip_version = ip->ip_v;
933 switch (ip_version) {
934 case 4: {
935 necp_matched_policy_id = necp_ip_output_find_policy_match(data, 0, NULL,
936 &necp_result, &necp_result_parameter);
937 break;
938 }
939 case 6: {
940 necp_matched_policy_id = necp_ip6_output_find_policy_match(data, 0, NULL,
941 &necp_result, &necp_result_parameter);
942 break;
943 }
944 default: {
945 return (false);
946 }
947 }
948
949 if (necp_result == NECP_KERNEL_POLICY_RESULT_DROP ||
950 necp_result == NECP_KERNEL_POLICY_RESULT_SOCKET_DIVERT) {
951 /* Drop and flow divert packets should be blocked at the IP layer */
952 return (false);
953 }
954
955 necp_mark_packet_from_ip(data, necp_matched_policy_id);
956 return (true);
957}
958
959static errno_t
960ipsec_netif_sync_tx(kern_nexus_provider_t nxprov, kern_nexus_t nexus,
961 kern_channel_ring_t tx_ring, uint32_t flags)
962{
963#pragma unused(nxprov)
964#pragma unused(flags)
965 struct ipsec_pcb *pcb = kern_nexus_get_context(nexus);
966
967 struct netif_stats *nifs = &NX_NETIF_PRIVATE(nexus)->nif_stats;
968
969 lck_rw_lock_shared(&pcb->ipsec_pcb_lock);
970
971 struct kern_channel_ring_stat_increment tx_ring_stats;
972 bzero(&tx_ring_stats, sizeof(tx_ring_stats));
973 kern_channel_slot_t tx_pslot = NULL;
974 kern_channel_slot_t tx_slot = kern_channel_get_next_slot(tx_ring, NULL, NULL);
975
976 STATS_INC(nifs, NETIF_STATS_TXSYNC);
977
978 if (tx_slot == NULL) {
979 // Nothing to write, don't bother signalling
980 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
981 return 0;
982 }
983
984 if (pcb->ipsec_kpipe_enabled) {
985 kern_channel_ring_t rx_ring = pcb->ipsec_kpipe_rxring;
986 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
987
988 // Signal the kernel pipe ring to read
989 if (rx_ring != NULL) {
990 kern_channel_notify(rx_ring, 0);
991 }
992 return 0;
993 }
994
995 // If we're here, we're injecting into the BSD stack
996 while (tx_slot != NULL) {
997 size_t length = 0;
998 mbuf_t data = NULL;
999
1000 kern_packet_t tx_ph = kern_channel_slot_get_packet(tx_ring, tx_slot);
1001
1002 // Advance TX ring
1003 tx_pslot = tx_slot;
1004 tx_slot = kern_channel_get_next_slot(tx_ring, tx_slot, NULL);
1005
1006 if (tx_ph == 0) {
1007 continue;
1008 }
1009
1010 kern_buflet_t tx_buf = kern_packet_get_next_buflet(tx_ph, NULL);
1011 VERIFY(tx_buf != NULL);
1012 uint8_t *tx_baddr = kern_buflet_get_object_address(tx_buf);
1013 VERIFY(tx_baddr != 0);
1014 tx_baddr += kern_buflet_get_data_offset(tx_buf);
1015
1016 bpf_tap_packet_out(pcb->ipsec_ifp, DLT_RAW, tx_ph, NULL, 0);
1017
1018 length = MIN(kern_packet_get_data_length(tx_ph),
1019 pcb->ipsec_slot_size);
1020
1021 if (length > 0) {
1022 errno_t error = mbuf_gethdr(MBUF_DONTWAIT, MBUF_TYPE_HEADER, &data);
1023 if (error == 0) {
1024 error = mbuf_copyback(data, 0, length, tx_baddr, MBUF_DONTWAIT);
1025 if (error == 0) {
1026 // Mark packet from policy
1027 uint32_t policy_id = kern_packet_get_policy_id(tx_ph);
1028 necp_mark_packet_from_ip(data, policy_id);
1029
1030 // Check policy with NECP
1031 if (!ipsec_netif_check_policy(data)) {
1032 printf("ipsec_netif_sync_tx %s - failed policy check\n", pcb->ipsec_ifp->if_xname);
1033 STATS_INC(nifs, NETIF_STATS_DROPPED);
1034 mbuf_freem(data);
1035 data = NULL;
1036 } else {
1037 // Send through encryption
1038 error = ipsec_output(pcb->ipsec_ifp, data);
1039 if (error != 0) {
1040 printf("ipsec_netif_sync_tx %s - ipsec_output error %d\n", pcb->ipsec_ifp->if_xname, error);
1041 }
1042 }
1043 } else {
1044 printf("ipsec_netif_sync_tx %s - mbuf_copyback(%zu) error %d\n", pcb->ipsec_ifp->if_xname, length, error);
1045 STATS_INC(nifs, NETIF_STATS_NOMEM_MBUF);
1046 STATS_INC(nifs, NETIF_STATS_DROPPED);
1047 mbuf_freem(data);
1048 data = NULL;
1049 }
1050 } else {
1051 printf("ipsec_netif_sync_tx %s - mbuf_gethdr error %d\n", pcb->ipsec_ifp->if_xname, error);
1052 STATS_INC(nifs, NETIF_STATS_NOMEM_MBUF);
1053 STATS_INC(nifs, NETIF_STATS_DROPPED);
1054 }
1055 } else {
1056 printf("ipsec_netif_sync_tx %s - 0 length packet\n", pcb->ipsec_ifp->if_xname);
1057 STATS_INC(nifs, NETIF_STATS_BADLEN);
1058 STATS_INC(nifs, NETIF_STATS_DROPPED);
1059 }
1060
1061 if (data == NULL) {
1062 printf("ipsec_netif_sync_tx %s: no encrypted packet to send\n", pcb->ipsec_ifp->if_xname);
1063 break;
1064 }
1065
1066 STATS_INC(nifs, NETIF_STATS_TXPKTS);
1067 STATS_INC(nifs, NETIF_STATS_TXCOPY_MBUF);
1068
1069 tx_ring_stats.kcrsi_slots_transferred++;
1070 tx_ring_stats.kcrsi_bytes_transferred += length;
1071 }
1072
1073 if (tx_pslot) {
1074 kern_channel_advance_slot(tx_ring, tx_pslot);
1075 kern_channel_increment_ring_net_stats(tx_ring, pcb->ipsec_ifp, &tx_ring_stats);
1076 (void)kern_channel_reclaim(tx_ring);
1077 }
1078
1079 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
1080
1081 return 0;
1082}
1083
1084static errno_t
1085ipsec_netif_tx_doorbell(kern_nexus_provider_t nxprov, kern_nexus_t nexus,
1086 kern_channel_ring_t ring, __unused uint32_t flags)
1087{
1088#pragma unused(nxprov)
1089 struct ipsec_pcb *pcb = kern_nexus_get_context(nexus);
1090 boolean_t more = false;
1091 errno_t rc = 0;
1092
1093 /*
1094 * Refill and sync the ring; we may be racing against another thread doing
1095 * an RX sync that also wants to do kr_enter(), and so use the blocking
1096 * variant here.
1097 */
1098 rc = kern_channel_tx_refill_canblock(ring, UINT32_MAX, UINT32_MAX, true, &more);
1099 if (rc != 0 && rc != EAGAIN && rc != EBUSY) {
1100 printf("%s, tx refill failed %d\n", __func__, rc);
1101 }
1102
1103 (void) kr_enter(ring, TRUE);
1104 lck_rw_lock_shared(&pcb->ipsec_pcb_lock);
1105
1106 if (pcb->ipsec_kpipe_enabled) {
1107 uint32_t tx_available = kern_channel_available_slot_count(ring);
1108 if (pcb->ipsec_netif_txring_size > 0 &&
1109 tx_available >= pcb->ipsec_netif_txring_size - 1) {
1110 // No room left in tx ring, disable output for now
1111 errno_t error = ifnet_disable_output(pcb->ipsec_ifp);
1112 if (error != 0) {
1113 printf("ipsec_netif_tx_doorbell: ifnet_disable_output returned error %d\n", error);
1114 }
1115 }
1116 }
1117
1118 if (pcb->ipsec_kpipe_enabled) {
1119 kern_channel_ring_t rx_ring = pcb->ipsec_kpipe_rxring;
1120
1121 // Unlock while calling notify
1122 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
1123 // Signal the kernel pipe ring to read
1124 if (rx_ring != NULL) {
1125 kern_channel_notify(rx_ring, 0);
1126 }
1127 } else {
1128 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
1129 }
1130
1131 kr_exit(ring);
1132
1133 return (0);
1134}
1135
1136static errno_t
1137ipsec_netif_sync_rx(kern_nexus_provider_t nxprov, kern_nexus_t nexus,
1138 kern_channel_ring_t rx_ring, uint32_t flags)
1139{
1140#pragma unused(nxprov)
1141#pragma unused(flags)
1142 struct ipsec_pcb *pcb = kern_nexus_get_context(nexus);
1143 struct kern_channel_ring_stat_increment rx_ring_stats;
1144
1145 struct netif_stats *nifs = &NX_NETIF_PRIVATE(nexus)->nif_stats;
1146
1147 lck_rw_lock_shared(&pcb->ipsec_pcb_lock);
1148
1149 // Reclaim user-released slots
1150 (void) kern_channel_reclaim(rx_ring);
1151
1152 STATS_INC(nifs, NETIF_STATS_RXSYNC);
1153
1154 uint32_t avail = kern_channel_available_slot_count(rx_ring);
1155 if (avail == 0) {
1156 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
1157 return 0;
1158 }
1159
1160 struct kern_pbufpool *rx_pp = rx_ring->ckr_pp;
1161 VERIFY(rx_pp != NULL);
1162 bzero(&rx_ring_stats, sizeof(rx_ring_stats));
1163 kern_channel_slot_t rx_pslot = NULL;
1164 kern_channel_slot_t rx_slot = kern_channel_get_next_slot(rx_ring, NULL, NULL);
1165
1166 while (rx_slot != NULL) {
1167 // Check for a waiting packet
1168 lck_mtx_lock(&pcb->ipsec_input_chain_lock);
1169 mbuf_t data = pcb->ipsec_input_chain;
1170 if (data == NULL) {
1171 lck_mtx_unlock(&pcb->ipsec_input_chain_lock);
1172 break;
1173 }
1174
1175 // Allocate rx packet
1176 kern_packet_t rx_ph = 0;
1177 errno_t error = kern_pbufpool_alloc_nosleep(rx_pp, 1, &rx_ph);
1178 if (__improbable(error != 0)) {
1179 STATS_INC(nifs, NETIF_STATS_NOMEM_PKT);
1180 STATS_INC(nifs, NETIF_STATS_DROPPED);
1181 lck_mtx_unlock(&pcb->ipsec_input_chain_lock);
1182 break;
1183 }
1184
1185 // Advance waiting packets
1186 pcb->ipsec_input_chain = data->m_nextpkt;
1187 data->m_nextpkt = NULL;
1188 if (pcb->ipsec_input_chain == NULL) {
1189 pcb->ipsec_input_chain_last = NULL;
1190 }
1191 lck_mtx_unlock(&pcb->ipsec_input_chain_lock);
1192
1193 size_t length = mbuf_pkthdr_len(data);
1194
1195 if (length < sizeof(struct ip)) {
1196 // Flush data
1197 mbuf_freem(data);
1198 kern_pbufpool_free(rx_pp, rx_ph);
1199 STATS_INC(nifs, NETIF_STATS_BADLEN);
1200 STATS_INC(nifs, NETIF_STATS_DROPPED);
1201 printf("ipsec_netif_sync_rx %s: legacy decrypted packet length cannot hold IP %zu < %zu\n",
1202 pcb->ipsec_ifp->if_xname, length, sizeof(struct ip));
1203 continue;
1204 }
1205
1206 uint32_t af = 0;
1207 struct ip *ip = mtod(data, struct ip *);
1208 u_int ip_version = ip->ip_v;
1209 switch (ip_version) {
1210 case 4: {
1211 af = AF_INET;
1212 break;
1213 }
1214 case 6: {
1215 af = AF_INET6;
1216 break;
1217 }
1218 default: {
1219 printf("ipsec_netif_sync_rx %s: legacy unknown ip version %u\n",
1220 pcb->ipsec_ifp->if_xname, ip_version);
1221 break;
1222 }
1223 }
1224
1225 if (length > rx_pp->pp_buflet_size ||
1226 (pcb->ipsec_frag_size_set && length > pcb->ipsec_input_frag_size)) {
1227
1228 // We need to fragment to send up into the netif
1229
1230 u_int32_t fragment_mtu = rx_pp->pp_buflet_size;
1231 if (pcb->ipsec_frag_size_set &&
1232 pcb->ipsec_input_frag_size < rx_pp->pp_buflet_size) {
1233 fragment_mtu = pcb->ipsec_input_frag_size;
1234 }
1235
1236 mbuf_t fragment_chain = NULL;
1237 switch (af) {
1238 case AF_INET: {
1239 // ip_fragment expects the length in host order
1240 ip->ip_len = ntohs(ip->ip_len);
1241
1242 // ip_fragment will modify the original data, don't free
1243 int fragment_error = ip_fragment(data, pcb->ipsec_ifp, fragment_mtu, TRUE);
1244 if (fragment_error == 0 && data != NULL) {
1245 fragment_chain = data;
1246 } else {
1247 STATS_INC(nifs, NETIF_STATS_BADLEN);
1248 STATS_INC(nifs, NETIF_STATS_DROPPED);
1249 printf("ipsec_netif_sync_rx %s: failed to fragment IPv4 packet of length %zu (%d)\n",
1250 pcb->ipsec_ifp->if_xname, length, fragment_error);
1251 }
1252 break;
1253 }
1254 case AF_INET6: {
1255 if (length < sizeof(struct ip6_hdr)) {
1256 mbuf_freem(data);
1257 STATS_INC(nifs, NETIF_STATS_BADLEN);
1258 STATS_INC(nifs, NETIF_STATS_DROPPED);
1259 printf("ipsec_netif_sync_rx %s: failed to fragment IPv6 packet of length %zu < %zu\n",
1260 pcb->ipsec_ifp->if_xname, length, sizeof(struct ip6_hdr));
1261 } else {
1262
1263 // ip6_do_fragmentation will free the original data on success only
1264 struct ip6_hdr *ip6 = mtod(data, struct ip6_hdr *);
1265 struct ip6_exthdrs exthdrs;
1266 memset(&exthdrs, 0, sizeof(exthdrs));
1267
1268 int fragment_error = ip6_do_fragmentation(&data, 0, pcb->ipsec_ifp, sizeof(struct ip6_hdr),
1269 ip6, &exthdrs, fragment_mtu, ip6->ip6_nxt);
1270 if (fragment_error == 0 && data != NULL) {
1271 fragment_chain = data;
1272 } else {
1273 mbuf_freem(data);
1274 STATS_INC(nifs, NETIF_STATS_BADLEN);
1275 STATS_INC(nifs, NETIF_STATS_DROPPED);
1276 printf("ipsec_netif_sync_rx %s: failed to fragment IPv6 packet of length %zu (%d)\n",
1277 pcb->ipsec_ifp->if_xname, length, fragment_error);
1278 }
1279 }
1280 break;
1281 }
1282 default: {
1283 // Cannot fragment unknown families
1284 mbuf_freem(data);
1285 STATS_INC(nifs, NETIF_STATS_BADLEN);
1286 STATS_INC(nifs, NETIF_STATS_DROPPED);
1287 printf("ipsec_netif_sync_rx %s: uknown legacy decrypted packet length %zu > %u\n",
1288 pcb->ipsec_ifp->if_xname, length, rx_pp->pp_buflet_size);
1289 break;
1290 }
1291 }
1292
1293 if (fragment_chain != NULL) {
1294 // Add fragments to chain before continuing
1295 lck_mtx_lock(&pcb->ipsec_input_chain_lock);
1296 if (pcb->ipsec_input_chain != NULL) {
1297 pcb->ipsec_input_chain_last->m_nextpkt = fragment_chain;
1298 } else {
1299 pcb->ipsec_input_chain = fragment_chain;
1300 }
1301 while (fragment_chain->m_nextpkt) {
1302 VERIFY(fragment_chain != fragment_chain->m_nextpkt);
1303 fragment_chain = fragment_chain->m_nextpkt;
1304 }
1305 pcb->ipsec_input_chain_last = fragment_chain;
1306 lck_mtx_unlock(&pcb->ipsec_input_chain_lock);
1307 }
1308
1309 // Make sure to free unused rx packet
1310 kern_pbufpool_free(rx_pp, rx_ph);
1311
1312 continue;
1313 }
1314
1315 mbuf_pkthdr_setrcvif(data, pcb->ipsec_ifp);
1316
1317 // Fillout rx packet
1318 kern_buflet_t rx_buf = kern_packet_get_next_buflet(rx_ph, NULL);
1319 VERIFY(rx_buf != NULL);
1320 void *rx_baddr = kern_buflet_get_object_address(rx_buf);
1321 VERIFY(rx_baddr != NULL);
1322
1323 // Copy-in data from mbuf to buflet
1324 mbuf_copydata(data, 0, length, (void *)rx_baddr);
1325 kern_packet_clear_flow_uuid(rx_ph); // Zero flow id
1326
1327 // Finalize and attach the packet
1328 error = kern_buflet_set_data_offset(rx_buf, 0);
1329 VERIFY(error == 0);
1330 error = kern_buflet_set_data_length(rx_buf, length);
1331 VERIFY(error == 0);
1332 error = kern_packet_set_link_header_offset(rx_ph, 0);
1333 VERIFY(error == 0);
1334 error = kern_packet_set_network_header_offset(rx_ph, 0);
1335 VERIFY(error == 0);
1336 error = kern_packet_finalize(rx_ph);
1337 VERIFY(error == 0);
1338 error = kern_channel_slot_attach_packet(rx_ring, rx_slot, rx_ph);
1339 VERIFY(error == 0);
1340
1341 STATS_INC(nifs, NETIF_STATS_RXPKTS);
1342 STATS_INC(nifs, NETIF_STATS_RXCOPY_MBUF);
1343 bpf_tap_packet_in(pcb->ipsec_ifp, DLT_RAW, rx_ph, NULL, 0);
1344
1345 rx_ring_stats.kcrsi_slots_transferred++;
1346 rx_ring_stats.kcrsi_bytes_transferred += length;
1347
1348 if (!pcb->ipsec_ext_ifdata_stats) {
1349 ifnet_stat_increment_in(pcb->ipsec_ifp, 1, length, 0);
1350 }
1351
1352 mbuf_freem(data);
1353
1354 // Advance ring
1355 rx_pslot = rx_slot;
1356 rx_slot = kern_channel_get_next_slot(rx_ring, rx_slot, NULL);
1357 }
1358
1359 struct kern_channel_ring_stat_increment tx_ring_stats;
1360 bzero(&tx_ring_stats, sizeof(tx_ring_stats));
1361 kern_channel_ring_t tx_ring = pcb->ipsec_kpipe_txring;
1362 kern_channel_slot_t tx_pslot = NULL;
1363 kern_channel_slot_t tx_slot = NULL;
1364 if (tx_ring == NULL) {
1365 // Net-If TX ring not set up yet, nothing to read
1366 goto done;
1367 }
1368
1369
1370 // Unlock ipsec before entering ring
1371 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
1372
1373 (void)kr_enter(tx_ring, TRUE);
1374
1375 // Lock again after entering and validate
1376 lck_rw_lock_shared(&pcb->ipsec_pcb_lock);
1377
1378 if (tx_ring != pcb->ipsec_kpipe_txring) {
1379 goto done;
1380 }
1381
1382 tx_slot = kern_channel_get_next_slot(tx_ring, NULL, NULL);
1383 if (tx_slot == NULL) {
1384 // Nothing to read, don't bother signalling
1385 goto done;
1386 }
1387
1388 while (rx_slot != NULL && tx_slot != NULL) {
1389 size_t length = 0;
1390 mbuf_t data = NULL;
1391 errno_t error = 0;
1392 uint32_t af;
1393
1394 // Allocate rx packet
1395 kern_packet_t rx_ph = 0;
1396 error = kern_pbufpool_alloc_nosleep(rx_pp, 1, &rx_ph);
1397 if (__improbable(error != 0)) {
1398 STATS_INC(nifs, NETIF_STATS_NOMEM_PKT);
1399 STATS_INC(nifs, NETIF_STATS_DROPPED);
1400 break;
1401 }
1402
1403 kern_packet_t tx_ph = kern_channel_slot_get_packet(tx_ring, tx_slot);
1404
1405 // Advance TX ring
1406 tx_pslot = tx_slot;
1407 tx_slot = kern_channel_get_next_slot(tx_ring, tx_slot, NULL);
1408
1409 if (tx_ph == 0) {
1410 kern_pbufpool_free(rx_pp, rx_ph);
1411 continue;
1412 }
1413
1414 kern_buflet_t tx_buf = kern_packet_get_next_buflet(tx_ph, NULL);
1415 VERIFY(tx_buf != NULL);
1416 uint8_t *tx_baddr = kern_buflet_get_object_address(tx_buf);
1417 VERIFY(tx_baddr != 0);
1418 tx_baddr += kern_buflet_get_data_offset(tx_buf);
1419
1420 length = MIN(kern_packet_get_data_length(tx_ph),
1421 pcb->ipsec_slot_size);
1422
1423 // Increment TX stats
1424 tx_ring_stats.kcrsi_slots_transferred++;
1425 tx_ring_stats.kcrsi_bytes_transferred += length;
1426
1427 if (length >= sizeof(struct ip)) {
1428 error = mbuf_gethdr(MBUF_DONTWAIT, MBUF_TYPE_HEADER, &data);
1429 if (error == 0) {
1430 error = mbuf_copyback(data, 0, length, tx_baddr, MBUF_DONTWAIT);
1431 if (error == 0) {
1432 struct ip *ip = mtod(data, struct ip *);
1433 u_int ip_version = ip->ip_v;
1434 switch (ip_version) {
1435 case 4: {
1436 af = AF_INET;
1437 ip->ip_len = ntohs(ip->ip_len) - sizeof(struct ip);
1438 ip->ip_off = ntohs(ip->ip_off);
1439
1440 if (length < ip->ip_len) {
1441 printf("ipsec_netif_sync_rx %s: IPv4 packet length too short (%zu < %u)\n",
1442 pcb->ipsec_ifp->if_xname, length, ip->ip_len);
1443 STATS_INC(nifs, NETIF_STATS_BADLEN);
1444 STATS_INC(nifs, NETIF_STATS_DROPPED);
1445 mbuf_freem(data);
1446 data = NULL;
1447 } else {
1448 data = esp4_input_extended(data, sizeof(struct ip), pcb->ipsec_ifp);
1449 }
1450 break;
1451 }
1452 case 6: {
1453 if (length < sizeof(struct ip6_hdr)) {
1454 printf("ipsec_netif_sync_rx %s: IPv6 packet length too short for header %zu\n",
1455 pcb->ipsec_ifp->if_xname, length);
1456 STATS_INC(nifs, NETIF_STATS_BADLEN);
1457 STATS_INC(nifs, NETIF_STATS_DROPPED);
1458 mbuf_freem(data);
1459 data = NULL;
1460 } else {
1461 af = AF_INET6;
1462 struct ip6_hdr *ip6 = mtod(data, struct ip6_hdr *);
1463 const size_t ip6_len = sizeof(*ip6) + ntohs(ip6->ip6_plen);
1464 if (length < ip6_len) {
1465 printf("ipsec_netif_sync_rx %s: IPv6 packet length too short (%zu < %zu)\n",
1466 pcb->ipsec_ifp->if_xname, length, ip6_len);
1467 STATS_INC(nifs, NETIF_STATS_BADLEN);
1468 STATS_INC(nifs, NETIF_STATS_DROPPED);
1469 mbuf_freem(data);
1470 data = NULL;
1471 } else {
1472 int offset = sizeof(struct ip6_hdr);
1473 esp6_input_extended(&data, &offset, ip6->ip6_nxt, pcb->ipsec_ifp);
1474 }
1475 }
1476 break;
1477 }
1478 default: {
1479 printf("ipsec_netif_sync_rx %s: unknown ip version %u\n",
1480 pcb->ipsec_ifp->if_xname, ip_version);
1481 STATS_INC(nifs, NETIF_STATS_DROPPED);
1482 mbuf_freem(data);
1483 data = NULL;
1484 break;
1485 }
1486 }
1487 } else {
1488 printf("ipsec_netif_sync_rx %s - mbuf_copyback(%zu) error %d\n", pcb->ipsec_ifp->if_xname, length, error);
1489 STATS_INC(nifs, NETIF_STATS_NOMEM_MBUF);
1490 STATS_INC(nifs, NETIF_STATS_DROPPED);
1491 mbuf_freem(data);
1492 data = NULL;
1493 }
1494 } else {
1495 printf("ipsec_netif_sync_rx %s - mbuf_gethdr error %d\n", pcb->ipsec_ifp->if_xname, error);
1496 STATS_INC(nifs, NETIF_STATS_NOMEM_MBUF);
1497 STATS_INC(nifs, NETIF_STATS_DROPPED);
1498 }
1499 } else {
1500 printf("ipsec_netif_sync_rx %s - bad packet length %zu\n", pcb->ipsec_ifp->if_xname, length);
1501 STATS_INC(nifs, NETIF_STATS_BADLEN);
1502 STATS_INC(nifs, NETIF_STATS_DROPPED);
1503 }
1504
1505 if (data == NULL) {
1506 // Failed to get decrypted data data
1507 kern_pbufpool_free(rx_pp, rx_ph);
1508 continue;
1509 }
1510
1511 length = mbuf_pkthdr_len(data);
1512 if (length > rx_pp->pp_buflet_size) {
1513 // Flush data
1514 mbuf_freem(data);
1515 kern_pbufpool_free(rx_pp, rx_ph);
1516 STATS_INC(nifs, NETIF_STATS_BADLEN);
1517 STATS_INC(nifs, NETIF_STATS_DROPPED);
1518 printf("ipsec_netif_sync_rx %s: decrypted packet length %zu > %u\n",
1519 pcb->ipsec_ifp->if_xname, length, rx_pp->pp_buflet_size);
1520 continue;
1521 }
1522
1523 mbuf_pkthdr_setrcvif(data, pcb->ipsec_ifp);
1524
1525 // Fillout rx packet
1526 kern_buflet_t rx_buf = kern_packet_get_next_buflet(rx_ph, NULL);
1527 VERIFY(rx_buf != NULL);
1528 void *rx_baddr = kern_buflet_get_object_address(rx_buf);
1529 VERIFY(rx_baddr != NULL);
1530
1531 // Copy-in data from mbuf to buflet
1532 mbuf_copydata(data, 0, length, (void *)rx_baddr);
1533 kern_packet_clear_flow_uuid(rx_ph); // Zero flow id
1534
1535 // Finalize and attach the packet
1536 error = kern_buflet_set_data_offset(rx_buf, 0);
1537 VERIFY(error == 0);
1538 error = kern_buflet_set_data_length(rx_buf, length);
1539 VERIFY(error == 0);
1540 error = kern_packet_set_link_header_offset(rx_ph, 0);
1541 VERIFY(error == 0);
1542 error = kern_packet_set_network_header_offset(rx_ph, 0);
1543 VERIFY(error == 0);
1544 error = kern_packet_finalize(rx_ph);
1545 VERIFY(error == 0);
1546 error = kern_channel_slot_attach_packet(rx_ring, rx_slot, rx_ph);
1547 VERIFY(error == 0);
1548
1549 STATS_INC(nifs, NETIF_STATS_RXPKTS);
1550 STATS_INC(nifs, NETIF_STATS_RXCOPY_DIRECT);
1551 bpf_tap_packet_in(pcb->ipsec_ifp, DLT_RAW, rx_ph, NULL, 0);
1552
1553 rx_ring_stats.kcrsi_slots_transferred++;
1554 rx_ring_stats.kcrsi_bytes_transferred += length;
1555
1556 if (!pcb->ipsec_ext_ifdata_stats) {
1557 ifnet_stat_increment_in(pcb->ipsec_ifp, 1, length, 0);
1558 }
1559
1560 mbuf_freem(data);
1561
1562 rx_pslot = rx_slot;
1563 rx_slot = kern_channel_get_next_slot(rx_ring, rx_slot, NULL);
1564 }
1565
1566done:
1567 if (rx_pslot) {
1568 kern_channel_advance_slot(rx_ring, rx_pslot);
1569 kern_channel_increment_ring_net_stats(rx_ring, pcb->ipsec_ifp, &rx_ring_stats);
1570 }
1571
1572 if (tx_pslot) {
1573 kern_channel_advance_slot(tx_ring, tx_pslot);
1574 kern_channel_increment_ring_net_stats(tx_ring, pcb->ipsec_ifp, &tx_ring_stats);
1575 (void)kern_channel_reclaim(tx_ring);
1576 }
1577
1578 // Unlock first, then exit ring
1579 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
1580 if (tx_ring != NULL) {
1581 if (tx_pslot != NULL) {
1582 kern_channel_notify(tx_ring, 0);
1583 }
1584 kr_exit(tx_ring);
1585 }
1586
1587 return 0;
1588}
1589
1590static errno_t
1591ipsec_nexus_ifattach(struct ipsec_pcb *pcb,
1592 struct ifnet_init_eparams *init_params,
1593 struct ifnet **ifp)
1594{
1595 errno_t err;
1596 nexus_controller_t controller = kern_nexus_shared_controller();
1597 struct kern_nexus_net_init net_init;
1598 struct kern_pbufpool_init pp_init;
1599
1600 nexus_name_t provider_name;
1601 snprintf((char *)provider_name, sizeof(provider_name),
1602 "com.apple.netif.%s", pcb->ipsec_if_xname);
1603
1604 struct kern_nexus_provider_init prov_init = {
1605 .nxpi_version = KERN_NEXUS_DOMAIN_PROVIDER_CURRENT_VERSION,
1606 .nxpi_flags = NXPIF_VIRTUAL_DEVICE,
1607 .nxpi_pre_connect = ipsec_nexus_pre_connect,
1608 .nxpi_connected = ipsec_nexus_connected,
1609 .nxpi_pre_disconnect = ipsec_netif_pre_disconnect,
1610 .nxpi_disconnected = ipsec_nexus_disconnected,
1611 .nxpi_ring_init = ipsec_netif_ring_init,
1612 .nxpi_ring_fini = ipsec_netif_ring_fini,
1613 .nxpi_slot_init = NULL,
1614 .nxpi_slot_fini = NULL,
1615 .nxpi_sync_tx = ipsec_netif_sync_tx,
1616 .nxpi_sync_rx = ipsec_netif_sync_rx,
1617 .nxpi_tx_doorbell = ipsec_netif_tx_doorbell,
1618 };
1619
1620 nexus_attr_t nxa = NULL;
1621 err = kern_nexus_attr_create(&nxa);
1622 IPSEC_IF_VERIFY(err == 0);
1623 if (err != 0) {
1624 printf("%s: kern_nexus_attr_create failed: %d\n",
1625 __func__, err);
1626 goto failed;
1627 }
1628
1629 uint64_t slot_buffer_size = pcb->ipsec_slot_size;
1630 err = kern_nexus_attr_set(nxa, NEXUS_ATTR_SLOT_BUF_SIZE, slot_buffer_size);
1631 VERIFY(err == 0);
1632
1633 // Reset ring size for netif nexus to limit memory usage
1634 uint64_t ring_size = pcb->ipsec_netif_ring_size;
1635 err = kern_nexus_attr_set(nxa, NEXUS_ATTR_TX_SLOTS, ring_size);
1636 VERIFY(err == 0);
1637 err = kern_nexus_attr_set(nxa, NEXUS_ATTR_RX_SLOTS, ring_size);
1638 VERIFY(err == 0);
1639
1640 pcb->ipsec_netif_txring_size = ring_size;
1641
1642 bzero(&pp_init, sizeof (pp_init));
1643 pp_init.kbi_version = KERN_PBUFPOOL_CURRENT_VERSION;
1644 pp_init.kbi_packets = pcb->ipsec_netif_ring_size * 2;
1645 pp_init.kbi_bufsize = pcb->ipsec_slot_size;
1646 pp_init.kbi_buf_seg_size = IPSEC_IF_DEFAULT_BUF_SEG_SIZE;
1647 pp_init.kbi_max_frags = 1;
1648 (void) snprintf((char *)pp_init.kbi_name, sizeof (pp_init.kbi_name),
1649 "%s", provider_name);
1650
1651 err = kern_pbufpool_create(&pp_init, &pp_init, &pcb->ipsec_netif_pp, NULL);
1652 if (err != 0) {
1653 printf("%s pbufbool create failed, error %d\n", __func__, err);
1654 goto failed;
1655 }
1656
1657 err = kern_nexus_controller_register_provider(controller,
1658 ipsec_nx_dom_prov,
1659 provider_name,
1660 &prov_init,
1661 sizeof(prov_init),
1662 nxa,
1663 &pcb->ipsec_nx.if_provider);
1664 IPSEC_IF_VERIFY(err == 0);
1665 if (err != 0) {
1666 printf("%s register provider failed, error %d\n",
1667 __func__, err);
1668 goto failed;
1669 }
1670
1671 bzero(&net_init, sizeof(net_init));
1672 net_init.nxneti_version = KERN_NEXUS_NET_CURRENT_VERSION;
1673 net_init.nxneti_flags = 0;
1674 net_init.nxneti_eparams = init_params;
1675 net_init.nxneti_lladdr = NULL;
1676 net_init.nxneti_prepare = ipsec_netif_prepare;
1677 net_init.nxneti_tx_pbufpool = pcb->ipsec_netif_pp;
1678 err = kern_nexus_controller_alloc_net_provider_instance(controller,
1679 pcb->ipsec_nx.if_provider,
1680 pcb,
1681 &pcb->ipsec_nx.if_instance,
1682 &net_init,
1683 ifp);
1684 IPSEC_IF_VERIFY(err == 0);
1685 if (err != 0) {
1686 printf("%s alloc_net_provider_instance failed, %d\n",
1687 __func__, err);
1688 kern_nexus_controller_deregister_provider(controller,
1689 pcb->ipsec_nx.if_provider);
1690 uuid_clear(pcb->ipsec_nx.if_provider);
1691 goto failed;
1692 }
1693
1694failed:
1695 if (nxa) {
1696 kern_nexus_attr_destroy(nxa);
1697 }
1698 if (err && pcb->ipsec_netif_pp != NULL) {
1699 kern_pbufpool_destroy(pcb->ipsec_netif_pp);
1700 pcb->ipsec_netif_pp = NULL;
1701 }
1702 return (err);
1703}
1704
1705static void
1706ipsec_detach_provider_and_instance(uuid_t provider, uuid_t instance)
1707{
1708 nexus_controller_t controller = kern_nexus_shared_controller();
1709 errno_t err;
1710
1711 if (!uuid_is_null(instance)) {
1712 err = kern_nexus_controller_free_provider_instance(controller,
1713 instance);
1714 if (err != 0) {
1715 printf("%s free_provider_instance failed %d\n",
1716 __func__, err);
1717 }
1718 uuid_clear(instance);
1719 }
1720 if (!uuid_is_null(provider)) {
1721 err = kern_nexus_controller_deregister_provider(controller,
1722 provider);
1723 if (err != 0) {
1724 printf("%s deregister_provider %d\n", __func__, err);
1725 }
1726 uuid_clear(provider);
1727 }
1728 return;
1729}
1730
1731static void
1732ipsec_nexus_detach(struct ipsec_pcb *pcb)
1733{
1734 ipsec_nx_t nx = &pcb->ipsec_nx;
1735 nexus_controller_t controller = kern_nexus_shared_controller();
1736 errno_t err;
1737
1738 if (!uuid_is_null(nx->ms_host)) {
1739 err = kern_nexus_ifdetach(controller,
1740 nx->ms_instance,
1741 nx->ms_host);
1742 if (err != 0) {
1743 printf("%s: kern_nexus_ifdetach ms host failed %d\n",
1744 __func__, err);
1745 }
1746 }
1747
1748 if (!uuid_is_null(nx->ms_device)) {
1749 err = kern_nexus_ifdetach(controller,
1750 nx->ms_instance,
1751 nx->ms_device);
1752 if (err != 0) {
1753 printf("%s: kern_nexus_ifdetach ms device failed %d\n",
1754 __func__, err);
1755 }
1756 }
1757
1758 ipsec_detach_provider_and_instance(nx->if_provider,
1759 nx->if_instance);
1760 ipsec_detach_provider_and_instance(nx->ms_provider,
1761 nx->ms_instance);
1762
1763 if (pcb->ipsec_netif_pp != NULL) {
1764 kern_pbufpool_destroy(pcb->ipsec_netif_pp);
1765 pcb->ipsec_netif_pp = NULL;
1766
1767 }
1768 memset(nx, 0, sizeof(*nx));
1769}
1770
1771static errno_t
1772ipsec_create_fs_provider_and_instance(struct ipsec_pcb *pcb,
1773 uint32_t subtype, const char *type_name,
1774 const char *ifname,
1775 uuid_t *provider, uuid_t *instance)
1776{
1777 nexus_attr_t attr = NULL;
1778 nexus_controller_t controller = kern_nexus_shared_controller();
1779 uuid_t dom_prov;
1780 errno_t err;
1781 struct kern_nexus_init init;
1782 nexus_name_t provider_name;
1783
1784 err = kern_nexus_get_builtin_domain_provider(NEXUS_TYPE_FLOW_SWITCH,
1785 &dom_prov);
1786 IPSEC_IF_VERIFY(err == 0);
1787 if (err != 0) {
1788 printf("%s can't get %s provider, error %d\n",
1789 __func__, type_name, err);
1790 goto failed;
1791 }
1792
1793 err = kern_nexus_attr_create(&attr);
1794 IPSEC_IF_VERIFY(err == 0);
1795 if (err != 0) {
1796 printf("%s: kern_nexus_attr_create failed: %d\n",
1797 __func__, err);
1798 goto failed;
1799 }
1800
1801 err = kern_nexus_attr_set(attr, NEXUS_ATTR_EXTENSIONS, subtype);
1802 VERIFY(err == 0);
1803
1804 uint64_t slot_buffer_size = pcb->ipsec_slot_size;
1805 err = kern_nexus_attr_set(attr, NEXUS_ATTR_SLOT_BUF_SIZE, slot_buffer_size);
1806 VERIFY(err == 0);
1807
1808 // Reset ring size for flowswitch nexus to limit memory usage. Larger RX than netif.
1809 uint64_t tx_ring_size = pcb->ipsec_tx_fsw_ring_size;
1810 err = kern_nexus_attr_set(attr, NEXUS_ATTR_TX_SLOTS, tx_ring_size);
1811 VERIFY(err == 0);
1812 uint64_t rx_ring_size = pcb->ipsec_rx_fsw_ring_size;
1813 err = kern_nexus_attr_set(attr, NEXUS_ATTR_RX_SLOTS, rx_ring_size);
1814 VERIFY(err == 0);
1815
1816 snprintf((char *)provider_name, sizeof(provider_name),
1817 "com.apple.%s.%s", type_name, ifname);
1818 err = kern_nexus_controller_register_provider(controller,
1819 dom_prov,
1820 provider_name,
1821 NULL,
1822 0,
1823 attr,
1824 provider);
1825 kern_nexus_attr_destroy(attr);
1826 attr = NULL;
1827 IPSEC_IF_VERIFY(err == 0);
1828 if (err != 0) {
1829 printf("%s register %s provider failed, error %d\n",
1830 __func__, type_name, err);
1831 goto failed;
1832 }
1833 bzero(&init, sizeof (init));
1834 init.nxi_version = KERN_NEXUS_CURRENT_VERSION;
1835 err = kern_nexus_controller_alloc_provider_instance(controller,
1836 *provider,
1837 NULL,
1838 instance, &init);
1839 IPSEC_IF_VERIFY(err == 0);
1840 if (err != 0) {
1841 printf("%s alloc_provider_instance %s failed, %d\n",
1842 __func__, type_name, err);
1843 kern_nexus_controller_deregister_provider(controller,
1844 *provider);
1845 uuid_clear(*provider);
1846 }
1847failed:
1848 return (err);
1849}
1850
1851static errno_t
1852ipsec_multistack_attach(struct ipsec_pcb *pcb)
1853{
1854 nexus_controller_t controller = kern_nexus_shared_controller();
1855 errno_t err = 0;
1856 ipsec_nx_t nx = &pcb->ipsec_nx;
1857
1858 // Allocate multistack flowswitch
1859 err = ipsec_create_fs_provider_and_instance(pcb,
1860 NEXUS_EXTENSION_FSW_TYPE_MULTISTACK,
1861 "multistack",
1862 pcb->ipsec_ifp->if_xname,
1863 &nx->ms_provider,
1864 &nx->ms_instance);
1865 if (err != 0) {
1866 printf("%s: failed to create bridge provider and instance\n",
1867 __func__);
1868 goto failed;
1869 }
1870
1871 // Attach multistack to device port
1872 err = kern_nexus_ifattach(controller, nx->ms_instance,
1873 NULL, nx->if_instance,
1874 FALSE, &nx->ms_device);
1875 if (err != 0) {
1876 printf("%s kern_nexus_ifattach ms device %d\n", __func__, err);
1877 goto failed;
1878 }
1879
1880 // Attach multistack to host port
1881 err = kern_nexus_ifattach(controller, nx->ms_instance,
1882 NULL, nx->if_instance,
1883 TRUE, &nx->ms_host);
1884 if (err != 0) {
1885 printf("%s kern_nexus_ifattach ms host %d\n", __func__, err);
1886 goto failed;
1887 }
1888
1889 // Extract the agent UUID and save for later
1890 struct kern_nexus *multistack_nx = nx_find(nx->ms_instance, false);
1891 if (multistack_nx != NULL) {
1892 struct nx_flowswitch *flowswitch = NX_FSW_PRIVATE(multistack_nx);
1893 if (flowswitch != NULL) {
1894 FSW_RLOCK(flowswitch);
1895 struct fsw_ms_context *ms_context = (struct fsw_ms_context *)flowswitch->fsw_ops_private;
1896 if (ms_context != NULL) {
1897 uuid_copy(nx->ms_agent, ms_context->mc_agent_uuid);
1898 } else {
1899 printf("ipsec_multistack_attach - fsw_ms_context is NULL\n");
1900 }
1901 FSW_UNLOCK(flowswitch);
1902 } else {
1903 printf("ipsec_multistack_attach - flowswitch is NULL\n");
1904 }
1905 nx_release(multistack_nx);
1906 } else {
1907 printf("ipsec_multistack_attach - unable to find multistack nexus\n");
1908 }
1909
1910 return (0);
1911
1912failed:
1913 ipsec_nexus_detach(pcb);
1914
1915 errno_t detach_error = 0;
1916 if ((detach_error = ifnet_detach(pcb->ipsec_ifp)) != 0) {
1917 panic("ipsec_multistack_attach - ifnet_detach failed: %d\n", detach_error);
1918 /* NOT REACHED */
1919 }
1920
1921 return (err);
1922}
1923
1924#pragma mark Kernel Pipe Nexus
1925
1926static errno_t
1927ipsec_register_kernel_pipe_nexus(void)
1928{
1929 nexus_attr_t nxa = NULL;
1930 errno_t result;
1931
1932 lck_mtx_lock(&ipsec_lock);
1933 if (ipsec_ncd_refcount++) {
1934 lck_mtx_unlock(&ipsec_lock);
1935 return 0;
1936 }
1937
1938 result = kern_nexus_controller_create(&ipsec_ncd);
1939 if (result) {
1940 printf("%s: kern_nexus_controller_create failed: %d\n",
1941 __FUNCTION__, result);
1942 goto done;
1943 }
1944
1945 uuid_t dom_prov;
1946 result = kern_nexus_get_builtin_domain_provider(
1947 NEXUS_TYPE_KERNEL_PIPE, &dom_prov);
1948 if (result) {
1949 printf("%s: kern_nexus_get_builtin_domain_provider failed: %d\n",
1950 __FUNCTION__, result);
1951 goto done;
1952 }
1953
1954 struct kern_nexus_provider_init prov_init = {
1955 .nxpi_version = KERN_NEXUS_DOMAIN_PROVIDER_CURRENT_VERSION,
1956 .nxpi_flags = NXPIF_VIRTUAL_DEVICE,
1957 .nxpi_pre_connect = ipsec_nexus_pre_connect,
1958 .nxpi_connected = ipsec_nexus_connected,
1959 .nxpi_pre_disconnect = ipsec_nexus_pre_disconnect,
1960 .nxpi_disconnected = ipsec_nexus_disconnected,
1961 .nxpi_ring_init = ipsec_kpipe_ring_init,
1962 .nxpi_ring_fini = ipsec_kpipe_ring_fini,
1963 .nxpi_slot_init = NULL,
1964 .nxpi_slot_fini = NULL,
1965 .nxpi_sync_tx = ipsec_kpipe_sync_tx,
1966 .nxpi_sync_rx = ipsec_kpipe_sync_rx,
1967 .nxpi_tx_doorbell = NULL,
1968 };
1969
1970 result = kern_nexus_attr_create(&nxa);
1971 if (result) {
1972 printf("%s: kern_nexus_attr_create failed: %d\n",
1973 __FUNCTION__, result);
1974 goto done;
1975 }
1976
1977 uint64_t slot_buffer_size = IPSEC_IF_DEFAULT_SLOT_SIZE;
1978 result = kern_nexus_attr_set(nxa, NEXUS_ATTR_SLOT_BUF_SIZE, slot_buffer_size);
1979 VERIFY(result == 0);
1980
1981 // Reset ring size for kernel pipe nexus to limit memory usage
1982 uint64_t ring_size = if_ipsec_ring_size;
1983 result = kern_nexus_attr_set(nxa, NEXUS_ATTR_TX_SLOTS, ring_size);
1984 VERIFY(result == 0);
1985 result = kern_nexus_attr_set(nxa, NEXUS_ATTR_RX_SLOTS, ring_size);
1986 VERIFY(result == 0);
1987
1988 result = kern_nexus_controller_register_provider(ipsec_ncd,
1989 dom_prov,
1990 (const uint8_t *)"com.apple.nexus.ipsec.kpipe",
1991 &prov_init,
1992 sizeof(prov_init),
1993 nxa,
1994 &ipsec_kpipe_uuid);
1995 if (result) {
1996 printf("%s: kern_nexus_controller_register_provider failed: %d\n",
1997 __FUNCTION__, result);
1998 goto done;
1999 }
2000
2001done:
2002 if (nxa) {
2003 kern_nexus_attr_destroy(nxa);
2004 }
2005
2006 if (result) {
2007 if (ipsec_ncd) {
2008 kern_nexus_controller_destroy(ipsec_ncd);
2009 ipsec_ncd = NULL;
2010 }
2011 ipsec_ncd_refcount = 0;
2012 }
2013
2014 lck_mtx_unlock(&ipsec_lock);
2015
2016 return result;
2017}
2018
2019static void
2020ipsec_unregister_kernel_pipe_nexus(void)
2021{
2022 lck_mtx_lock(&ipsec_lock);
2023
2024 VERIFY(ipsec_ncd_refcount > 0);
2025
2026 if (--ipsec_ncd_refcount == 0) {
2027 kern_nexus_controller_destroy(ipsec_ncd);
2028 ipsec_ncd = NULL;
2029 }
2030
2031 lck_mtx_unlock(&ipsec_lock);
2032}
2033
2034// For use by socket option, not internally
2035static errno_t
2036ipsec_disable_channel(struct ipsec_pcb *pcb)
2037{
2038 errno_t result;
2039 int enabled;
2040 uuid_t uuid;
2041
2042 lck_rw_lock_exclusive(&pcb->ipsec_pcb_lock);
2043
2044 enabled = pcb->ipsec_kpipe_enabled;
2045 uuid_copy(uuid, pcb->ipsec_kpipe_uuid);
2046
2047 VERIFY(uuid_is_null(pcb->ipsec_kpipe_uuid) == !enabled);
2048
2049 pcb->ipsec_kpipe_enabled = 0;
2050 uuid_clear(pcb->ipsec_kpipe_uuid);
2051
2052 lck_rw_unlock_exclusive(&pcb->ipsec_pcb_lock);
2053
2054 if (enabled) {
2055 result = kern_nexus_controller_free_provider_instance(ipsec_ncd, uuid);
2056 } else {
2057 result = ENXIO;
2058 }
2059
2060 if (!result) {
2061 if (pcb->ipsec_kpipe_pp != NULL) {
2062 kern_pbufpool_destroy(pcb->ipsec_kpipe_pp);
2063 pcb->ipsec_kpipe_pp = NULL;
2064 }
2065 ipsec_unregister_kernel_pipe_nexus();
2066 }
2067
2068 return result;
2069}
2070
2071static errno_t
2072ipsec_enable_channel(struct ipsec_pcb *pcb, struct proc *proc)
2073{
2074 struct kern_nexus_init init;
2075 struct kern_pbufpool_init pp_init;
2076 errno_t result;
2077
2078 kauth_cred_t cred = kauth_cred_get();
2079 result = priv_check_cred(cred, PRIV_SKYWALK_REGISTER_KERNEL_PIPE, 0);
2080 if (result) {
2081 return result;
2082 }
2083
2084 result = ipsec_register_kernel_pipe_nexus();
2085 if (result) {
2086 return result;
2087 }
2088
2089 VERIFY(ipsec_ncd);
2090
2091 lck_rw_lock_exclusive(&pcb->ipsec_pcb_lock);
2092
2093 /* ipsec driver doesn't support channels without a netif */
2094 if (!pcb->ipsec_use_netif) {
2095 result = EOPNOTSUPP;
2096 goto done;
2097 }
2098
2099 if (pcb->ipsec_kpipe_enabled) {
2100 result = EEXIST; // return success instead?
2101 goto done;
2102 }
2103
2104 bzero(&pp_init, sizeof (pp_init));
2105 pp_init.kbi_version = KERN_PBUFPOOL_CURRENT_VERSION;
2106 pp_init.kbi_packets = pcb->ipsec_netif_ring_size * 2;
2107 pp_init.kbi_bufsize = pcb->ipsec_slot_size;
2108 pp_init.kbi_buf_seg_size = IPSEC_IF_DEFAULT_BUF_SEG_SIZE;
2109 pp_init.kbi_max_frags = 1;
2110 pp_init.kbi_flags |= KBIF_QUANTUM;
2111 (void) snprintf((char *)pp_init.kbi_name, sizeof (pp_init.kbi_name),
2112 "com.apple.kpipe.%s", pcb->ipsec_if_xname);
2113
2114 result = kern_pbufpool_create(&pp_init, &pp_init, &pcb->ipsec_kpipe_pp,
2115 NULL);
2116 if (result != 0) {
2117 printf("%s pbufbool create failed, error %d\n", __func__, result);
2118 goto done;
2119 }
2120
2121 VERIFY(uuid_is_null(pcb->ipsec_kpipe_uuid));
2122 bzero(&init, sizeof (init));
2123 init.nxi_version = KERN_NEXUS_CURRENT_VERSION;
2124 init.nxi_tx_pbufpool = pcb->ipsec_kpipe_pp;
2125 result = kern_nexus_controller_alloc_provider_instance(ipsec_ncd,
2126 ipsec_kpipe_uuid, pcb, &pcb->ipsec_kpipe_uuid, &init);
2127 if (result) {
2128 goto done;
2129 }
2130
2131 nexus_port_t port = NEXUS_PORT_KERNEL_PIPE_CLIENT;
2132 result = kern_nexus_controller_bind_provider_instance(ipsec_ncd,
2133 pcb->ipsec_kpipe_uuid, &port,
2134 proc_pid(proc), NULL, NULL, 0, NEXUS_BIND_PID);
2135 if (result) {
2136 kern_nexus_controller_free_provider_instance(ipsec_ncd,
2137 pcb->ipsec_kpipe_uuid);
2138 uuid_clear(pcb->ipsec_kpipe_uuid);
2139 goto done;
2140 }
2141
2142 pcb->ipsec_kpipe_enabled = 1;
2143
2144done:
2145 lck_rw_unlock_exclusive(&pcb->ipsec_pcb_lock);
2146
2147 if (result) {
2148 if (pcb->ipsec_kpipe_pp != NULL) {
2149 kern_pbufpool_destroy(pcb->ipsec_kpipe_pp);
2150 pcb->ipsec_kpipe_pp = NULL;
2151 }
2152 ipsec_unregister_kernel_pipe_nexus();
2153 }
2154
2155 return result;
2156}
2157
2158#endif // IPSEC_NEXUS
2159
2160
2161/* Kernel control functions */
2162
2163static inline void
2164ipsec_free_pcb(struct ipsec_pcb *pcb, bool in_list)
2165{
2166#if IPSEC_NEXUS
2167 mbuf_freem_list(pcb->ipsec_input_chain);
2168 lck_mtx_destroy(&pcb->ipsec_input_chain_lock, ipsec_lck_grp);
2169#endif // IPSEC_NEXUS
2170 lck_rw_destroy(&pcb->ipsec_pcb_lock, ipsec_lck_grp);
2171 if (in_list) {
2172 lck_mtx_lock(&ipsec_lock);
2173 TAILQ_REMOVE(&ipsec_head, pcb, ipsec_chain);
2174 lck_mtx_unlock(&ipsec_lock);
2175 }
2176 zfree(ipsec_pcb_zone, pcb);
2177}
2178
2179static errno_t
2180ipsec_ctl_bind(kern_ctl_ref kctlref,
2181 struct sockaddr_ctl *sac,
2182 void **unitinfo)
2183{
2184 struct ipsec_pcb *pcb = zalloc(ipsec_pcb_zone);
2185 memset(pcb, 0, sizeof(*pcb));
2186
2187 /* Setup the protocol control block */
2188 *unitinfo = pcb;
2189 pcb->ipsec_ctlref = kctlref;
2190 pcb->ipsec_unit = sac->sc_unit;
2191 pcb->ipsec_output_service_class = MBUF_SC_OAM;
2192
2193#if IPSEC_NEXUS
2194 pcb->ipsec_use_netif = false;
2195 pcb->ipsec_slot_size = IPSEC_IF_DEFAULT_SLOT_SIZE;
2196 pcb->ipsec_netif_ring_size = IPSEC_IF_DEFAULT_RING_SIZE;
2197 pcb->ipsec_tx_fsw_ring_size = IPSEC_IF_DEFAULT_TX_FSW_RING_SIZE;
2198 pcb->ipsec_rx_fsw_ring_size = IPSEC_IF_DEFAULT_RX_FSW_RING_SIZE;
2199#endif // IPSEC_NEXUS
2200
2201 lck_rw_init(&pcb->ipsec_pcb_lock, ipsec_lck_grp, ipsec_lck_attr);
2202#if IPSEC_NEXUS
2203 lck_mtx_init(&pcb->ipsec_input_chain_lock, ipsec_lck_grp, ipsec_lck_attr);
2204#endif // IPSEC_NEXUS
2205
2206 return (0);
2207}
2208
2209static errno_t
2210ipsec_ctl_connect(kern_ctl_ref kctlref,
2211 struct sockaddr_ctl *sac,
2212 void **unitinfo)
2213{
2214 struct ifnet_init_eparams ipsec_init = {};
2215 errno_t result = 0;
2216
2217 if (*unitinfo == NULL) {
2218 (void)ipsec_ctl_bind(kctlref, sac, unitinfo);
2219 }
2220
2221 struct ipsec_pcb *pcb = *unitinfo;
2222
2223 lck_mtx_lock(&ipsec_lock);
2224
2225 /* Find some open interface id */
2226 u_int32_t chosen_unique_id = 1;
2227 struct ipsec_pcb *next_pcb = TAILQ_LAST(&ipsec_head, ipsec_list);
2228 if (next_pcb != NULL) {
2229 /* List was not empty, add one to the last item */
2230 chosen_unique_id = next_pcb->ipsec_unique_id + 1;
2231 next_pcb = NULL;
2232
2233 /*
2234 * If this wrapped the id number, start looking at
2235 * the front of the list for an unused id.
2236 */
2237 if (chosen_unique_id == 0) {
2238 /* Find the next unused ID */
2239 chosen_unique_id = 1;
2240 TAILQ_FOREACH(next_pcb, &ipsec_head, ipsec_chain) {
2241 if (next_pcb->ipsec_unique_id > chosen_unique_id) {
2242 /* We found a gap */
2243 break;
2244 }
2245
2246 chosen_unique_id = next_pcb->ipsec_unique_id + 1;
2247 }
2248 }
2249 }
2250
2251 pcb->ipsec_unique_id = chosen_unique_id;
2252
2253 if (next_pcb != NULL) {
2254 TAILQ_INSERT_BEFORE(next_pcb, pcb, ipsec_chain);
2255 } else {
2256 TAILQ_INSERT_TAIL(&ipsec_head, pcb, ipsec_chain);
2257 }
2258 lck_mtx_unlock(&ipsec_lock);
2259
2260 snprintf(pcb->ipsec_if_xname, sizeof(pcb->ipsec_if_xname), "ipsec%d", pcb->ipsec_unit - 1);
2261 snprintf(pcb->ipsec_unique_name, sizeof(pcb->ipsec_unique_name), "ipsecid%d", pcb->ipsec_unique_id - 1);
2262 printf("ipsec_ctl_connect: creating interface %s (id %s)\n", pcb->ipsec_if_xname, pcb->ipsec_unique_name);
2263
2264 /* Create the interface */
2265 bzero(&ipsec_init, sizeof(ipsec_init));
2266 ipsec_init.ver = IFNET_INIT_CURRENT_VERSION;
2267 ipsec_init.len = sizeof (ipsec_init);
2268
2269#if IPSEC_NEXUS
2270 if (pcb->ipsec_use_netif) {
2271 ipsec_init.flags = (IFNET_INIT_SKYWALK_NATIVE | IFNET_INIT_NX_NOAUTO);
2272 } else
2273#endif // IPSEC_NEXUS
2274 {
2275 ipsec_init.flags = IFNET_INIT_NX_NOAUTO;
2276 ipsec_init.start = ipsec_start;
2277 }
2278 ipsec_init.name = "ipsec";
2279 ipsec_init.unit = pcb->ipsec_unit - 1;
2280 ipsec_init.uniqueid = pcb->ipsec_unique_name;
2281 ipsec_init.uniqueid_len = strlen(pcb->ipsec_unique_name);
2282 ipsec_init.family = ipsec_family;
2283 ipsec_init.subfamily = IFNET_SUBFAMILY_IPSEC;
2284 ipsec_init.type = IFT_OTHER;
2285 ipsec_init.demux = ipsec_demux;
2286 ipsec_init.add_proto = ipsec_add_proto;
2287 ipsec_init.del_proto = ipsec_del_proto;
2288 ipsec_init.softc = pcb;
2289 ipsec_init.ioctl = ipsec_ioctl;
2290 ipsec_init.detach = ipsec_detached;
2291
2292#if IPSEC_NEXUS
2293 if (pcb->ipsec_use_netif) {
2294 result = ipsec_nexus_ifattach(pcb, &ipsec_init, &pcb->ipsec_ifp);
2295 if (result != 0) {
2296 printf("ipsec_ctl_connect - ipsec_nexus_ifattach failed: %d\n", result);
2297 ipsec_free_pcb(pcb, true);
2298 *unitinfo = NULL;
2299 return result;
2300 }
2301
2302 result = ipsec_multistack_attach(pcb);
2303 if (result != 0) {
2304 printf("ipsec_ctl_connect - ipsec_multistack_attach failed: %d\n", result);
2305 *unitinfo = NULL;
2306 return result;
2307 }
2308
2309 /* Attach to bpf */
2310 bpfattach(pcb->ipsec_ifp, DLT_RAW, 0);
2311 } else
2312#endif // IPSEC_NEXUS
2313 {
2314 result = ifnet_allocate_extended(&ipsec_init, &pcb->ipsec_ifp);
2315 if (result != 0) {
2316 printf("ipsec_ctl_connect - ifnet_allocate failed: %d\n", result);
2317 ipsec_free_pcb(pcb, true);
2318 *unitinfo = NULL;
2319 return result;
2320 }
2321 ipsec_ifnet_set_attrs(pcb->ipsec_ifp);
2322
2323 /* Attach the interface */
2324 result = ifnet_attach(pcb->ipsec_ifp, NULL);
2325 if (result != 0) {
2326 printf("ipsec_ctl_connect - ifnet_attach failed: %d\n", result);
2327 ifnet_release(pcb->ipsec_ifp);
2328 ipsec_free_pcb(pcb, true);
2329 *unitinfo = NULL;
2330 return (result);
2331 }
2332
2333 /* Attach to bpf */
2334 bpfattach(pcb->ipsec_ifp, DLT_NULL, 0);
2335 }
2336
2337 /* The interfaces resoures allocated, mark it as running */
2338 ifnet_set_flags(pcb->ipsec_ifp, IFF_RUNNING, IFF_RUNNING);
2339
2340 return (0);
2341}
2342
2343static errno_t
2344ipsec_detach_ip(ifnet_t interface,
2345 protocol_family_t protocol,
2346 socket_t pf_socket)
2347{
2348 errno_t result = EPROTONOSUPPORT;
2349
2350 /* Attempt a detach */
2351 if (protocol == PF_INET) {
2352 struct ifreq ifr;
2353
2354 bzero(&ifr, sizeof(ifr));
2355 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "%s%d",
2356 ifnet_name(interface), ifnet_unit(interface));
2357
2358 result = sock_ioctl(pf_socket, SIOCPROTODETACH, &ifr);
2359 }
2360 else if (protocol == PF_INET6) {
2361 struct in6_ifreq ifr6;
2362
2363 bzero(&ifr6, sizeof(ifr6));
2364 snprintf(ifr6.ifr_name, sizeof(ifr6.ifr_name), "%s%d",
2365 ifnet_name(interface), ifnet_unit(interface));
2366
2367 result = sock_ioctl(pf_socket, SIOCPROTODETACH_IN6, &ifr6);
2368 }
2369
2370 return result;
2371}
2372
2373static void
2374ipsec_remove_address(ifnet_t interface,
2375 protocol_family_t protocol,
2376 ifaddr_t address,
2377 socket_t pf_socket)
2378{
2379 errno_t result = 0;
2380
2381 /* Attempt a detach */
2382 if (protocol == PF_INET) {
2383 struct ifreq ifr;
2384
2385 bzero(&ifr, sizeof(ifr));
2386 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "%s%d",
2387 ifnet_name(interface), ifnet_unit(interface));
2388 result = ifaddr_address(address, &ifr.ifr_addr, sizeof(ifr.ifr_addr));
2389 if (result != 0) {
2390 printf("ipsec_remove_address - ifaddr_address failed: %d", result);
2391 }
2392 else {
2393 result = sock_ioctl(pf_socket, SIOCDIFADDR, &ifr);
2394 if (result != 0) {
2395 printf("ipsec_remove_address - SIOCDIFADDR failed: %d", result);
2396 }
2397 }
2398 }
2399 else if (protocol == PF_INET6) {
2400 struct in6_ifreq ifr6;
2401
2402 bzero(&ifr6, sizeof(ifr6));
2403 snprintf(ifr6.ifr_name, sizeof(ifr6.ifr_name), "%s%d",
2404 ifnet_name(interface), ifnet_unit(interface));
2405 result = ifaddr_address(address, (struct sockaddr*)&ifr6.ifr_addr,
2406 sizeof(ifr6.ifr_addr));
2407 if (result != 0) {
2408 printf("ipsec_remove_address - ifaddr_address failed (v6): %d",
2409 result);
2410 }
2411 else {
2412 result = sock_ioctl(pf_socket, SIOCDIFADDR_IN6, &ifr6);
2413 if (result != 0) {
2414 printf("ipsec_remove_address - SIOCDIFADDR_IN6 failed: %d",
2415 result);
2416 }
2417 }
2418 }
2419}
2420
2421static void
2422ipsec_cleanup_family(ifnet_t interface,
2423 protocol_family_t protocol)
2424{
2425 errno_t result = 0;
2426 socket_t pf_socket = NULL;
2427 ifaddr_t *addresses = NULL;
2428 int i;
2429
2430 if (protocol != PF_INET && protocol != PF_INET6) {
2431 printf("ipsec_cleanup_family - invalid protocol family %d\n", protocol);
2432 return;
2433 }
2434
2435 /* Create a socket for removing addresses and detaching the protocol */
2436 result = sock_socket(protocol, SOCK_DGRAM, 0, NULL, NULL, &pf_socket);
2437 if (result != 0) {
2438 if (result != EAFNOSUPPORT)
2439 printf("ipsec_cleanup_family - failed to create %s socket: %d\n",
2440 protocol == PF_INET ? "IP" : "IPv6", result);
2441 goto cleanup;
2442 }
2443
2444 /* always set SS_PRIV, we want to close and detach regardless */
2445 sock_setpriv(pf_socket, 1);
2446
2447 result = ipsec_detach_ip(interface, protocol, pf_socket);
2448 if (result == 0 || result == ENXIO) {
2449 /* We are done! We either detached or weren't attached. */
2450 goto cleanup;
2451 }
2452 else if (result != EBUSY) {
2453 /* Uh, not really sure what happened here... */
2454 printf("ipsec_cleanup_family - ipsec_detach_ip failed: %d\n", result);
2455 goto cleanup;
2456 }
2457
2458 /*
2459 * At this point, we received an EBUSY error. This means there are
2460 * addresses attached. We should detach them and then try again.
2461 */
2462 result = ifnet_get_address_list_family(interface, &addresses, protocol);
2463 if (result != 0) {
2464 printf("fnet_get_address_list_family(%s%d, 0xblah, %s) - failed: %d\n",
2465 ifnet_name(interface), ifnet_unit(interface),
2466 protocol == PF_INET ? "PF_INET" : "PF_INET6", result);
2467 goto cleanup;
2468 }
2469
2470 for (i = 0; addresses[i] != 0; i++) {
2471 ipsec_remove_address(interface, protocol, addresses[i], pf_socket);
2472 }
2473 ifnet_free_address_list(addresses);
2474 addresses = NULL;
2475
2476 /*
2477 * The addresses should be gone, we should try the remove again.
2478 */
2479 result = ipsec_detach_ip(interface, protocol, pf_socket);
2480 if (result != 0 && result != ENXIO) {
2481 printf("ipsec_cleanup_family - ipsec_detach_ip failed: %d\n", result);
2482 }
2483
2484cleanup:
2485 if (pf_socket != NULL)
2486 sock_close(pf_socket);
2487
2488 if (addresses != NULL)
2489 ifnet_free_address_list(addresses);
2490}
2491
2492static errno_t
2493ipsec_ctl_disconnect(__unused kern_ctl_ref kctlref,
2494 __unused u_int32_t unit,
2495 void *unitinfo)
2496{
2497 struct ipsec_pcb *pcb = unitinfo;
2498 ifnet_t ifp = NULL;
2499 errno_t result = 0;
2500
2501 if (pcb == NULL) {
2502 return EINVAL;
2503 }
2504
2505#if IPSEC_NEXUS
2506 // Tell the nexus to stop all rings
2507 if (pcb->ipsec_netif_nexus != NULL) {
2508 kern_nexus_stop(pcb->ipsec_netif_nexus);
2509 }
2510#endif // IPSEC_NEXUS
2511
2512 lck_rw_lock_exclusive(&pcb->ipsec_pcb_lock);
2513
2514#if IPSEC_NEXUS
2515 uuid_t kpipe_uuid;
2516 uuid_copy(kpipe_uuid, pcb->ipsec_kpipe_uuid);
2517 uuid_clear(pcb->ipsec_kpipe_uuid);
2518 pcb->ipsec_kpipe_enabled = FALSE;
2519#endif // IPSEC_NEXUS
2520
2521 pcb->ipsec_ctlref = NULL;
2522
2523 ifp = pcb->ipsec_ifp;
2524 if (ifp != NULL) {
2525#if IPSEC_NEXUS
2526 if (pcb->ipsec_netif_nexus != NULL) {
2527 /*
2528 * Quiesce the interface and flush any pending outbound packets.
2529 */
2530 if_down(ifp);
2531
2532 /* Increment refcnt, but detach interface */
2533 ifnet_incr_iorefcnt(ifp);
2534 if ((result = ifnet_detach(ifp)) != 0) {
2535 panic("ipsec_ctl_disconnect - ifnet_detach failed: %d\n", result);
2536 /* NOT REACHED */
2537 }
2538
2539 /*
2540 * We want to do everything in our power to ensure that the interface
2541 * really goes away when the socket is closed. We must remove IP/IPv6
2542 * addresses and detach the protocols. Finally, we can remove and
2543 * release the interface.
2544 */
2545 key_delsp_for_ipsec_if(ifp);
2546
2547 ipsec_cleanup_family(ifp, AF_INET);
2548 ipsec_cleanup_family(ifp, AF_INET6);
2549
2550 lck_rw_unlock_exclusive(&pcb->ipsec_pcb_lock);
2551
2552 if (!uuid_is_null(kpipe_uuid)) {
2553 if (kern_nexus_controller_free_provider_instance(ipsec_ncd, kpipe_uuid) == 0) {
2554 if (pcb->ipsec_kpipe_pp != NULL) {
2555 kern_pbufpool_destroy(pcb->ipsec_kpipe_pp);
2556 pcb->ipsec_kpipe_pp = NULL;
2557 }
2558 ipsec_unregister_kernel_pipe_nexus();
2559 }
2560 }
2561 ipsec_nexus_detach(pcb);
2562
2563 /* Decrement refcnt to finish detaching and freeing */
2564 ifnet_decr_iorefcnt(ifp);
2565 } else
2566#endif // IPSEC_NEXUS
2567 {
2568 lck_rw_unlock_exclusive(&pcb->ipsec_pcb_lock);
2569
2570#if IPSEC_NEXUS
2571 if (!uuid_is_null(kpipe_uuid)) {
2572 if (kern_nexus_controller_free_provider_instance(ipsec_ncd, kpipe_uuid) == 0) {
2573 if (pcb->ipsec_kpipe_pp != NULL) {
2574 kern_pbufpool_destroy(pcb->ipsec_kpipe_pp);
2575 pcb->ipsec_kpipe_pp = NULL;
2576 }
2577 ipsec_unregister_kernel_pipe_nexus();
2578 }
2579 }
2580#endif // IPSEC_NEXUS
2581
2582 /*
2583 * We want to do everything in our power to ensure that the interface
2584 * really goes away when the socket is closed. We must remove IP/IPv6
2585 * addresses and detach the protocols. Finally, we can remove and
2586 * release the interface.
2587 */
2588 key_delsp_for_ipsec_if(ifp);
2589
2590 ipsec_cleanup_family(ifp, AF_INET);
2591 ipsec_cleanup_family(ifp, AF_INET6);
2592
2593 /*
2594 * Detach now; ipsec_detach() will be called asynchronously once
2595 * the I/O reference count drops to 0. There we will invoke
2596 * ifnet_release().
2597 */
2598 if ((result = ifnet_detach(ifp)) != 0) {
2599 printf("ipsec_ctl_disconnect - ifnet_detach failed: %d\n", result);
2600 }
2601 }
2602 } else {
2603 // Bound, but not connected
2604 lck_rw_unlock_exclusive(&pcb->ipsec_pcb_lock);
2605 ipsec_free_pcb(pcb, false);
2606 }
2607
2608 return 0;
2609}
2610
2611static errno_t
2612ipsec_ctl_send(__unused kern_ctl_ref kctlref,
2613 __unused u_int32_t unit,
2614 __unused void *unitinfo,
2615 mbuf_t m,
2616 __unused int flags)
2617{
2618 /* Receive messages from the control socket. Currently unused. */
2619 mbuf_freem(m);
2620 return 0;
2621}
2622
2623static errno_t
2624ipsec_ctl_setopt(__unused kern_ctl_ref kctlref,
2625 __unused u_int32_t unit,
2626 void *unitinfo,
2627 int opt,
2628 void *data,
2629 size_t len)
2630{
2631 struct ipsec_pcb *pcb = unitinfo;
2632 errno_t result = 0;
2633
2634 /* check for privileges for privileged options */
2635 switch (opt) {
2636 case IPSEC_OPT_FLAGS:
2637 case IPSEC_OPT_EXT_IFDATA_STATS:
2638 case IPSEC_OPT_SET_DELEGATE_INTERFACE:
2639 case IPSEC_OPT_OUTPUT_TRAFFIC_CLASS:
2640 if (kauth_cred_issuser(kauth_cred_get()) == 0) {
2641 return EPERM;
2642 }
2643 break;
2644 }
2645
2646 switch (opt) {
2647 case IPSEC_OPT_FLAGS:
2648 if (len != sizeof(u_int32_t)) {
2649 result = EMSGSIZE;
2650 } else {
2651 pcb->ipsec_flags = *(u_int32_t *)data;
2652 }
2653 break;
2654
2655 case IPSEC_OPT_EXT_IFDATA_STATS:
2656 if (len != sizeof(int)) {
2657 result = EMSGSIZE;
2658 break;
2659 }
2660 if (pcb->ipsec_ifp == NULL) {
2661 // Only can set after connecting
2662 result = EINVAL;
2663 break;
2664 }
2665 pcb->ipsec_ext_ifdata_stats = (*(int *)data) ? 1 : 0;
2666 break;
2667
2668 case IPSEC_OPT_INC_IFDATA_STATS_IN:
2669 case IPSEC_OPT_INC_IFDATA_STATS_OUT: {
2670 struct ipsec_stats_param *utsp = (struct ipsec_stats_param *)data;
2671
2672 if (utsp == NULL || len < sizeof(struct ipsec_stats_param)) {
2673 result = EINVAL;
2674 break;
2675 }
2676 if (pcb->ipsec_ifp == NULL) {
2677 // Only can set after connecting
2678 result = EINVAL;
2679 break;
2680 }
2681 if (!pcb->ipsec_ext_ifdata_stats) {
2682 result = EINVAL;
2683 break;
2684 }
2685 if (opt == IPSEC_OPT_INC_IFDATA_STATS_IN)
2686 ifnet_stat_increment_in(pcb->ipsec_ifp, utsp->utsp_packets,
2687 utsp->utsp_bytes, utsp->utsp_errors);
2688 else
2689 ifnet_stat_increment_out(pcb->ipsec_ifp, utsp->utsp_packets,
2690 utsp->utsp_bytes, utsp->utsp_errors);
2691 break;
2692 }
2693
2694 case IPSEC_OPT_SET_DELEGATE_INTERFACE: {
2695 ifnet_t del_ifp = NULL;
2696 char name[IFNAMSIZ];
2697
2698 if (len > IFNAMSIZ - 1) {
2699 result = EMSGSIZE;
2700 break;
2701 }
2702 if (pcb->ipsec_ifp == NULL) {
2703 // Only can set after connecting
2704 result = EINVAL;
2705 break;
2706 }
2707 if (len != 0) { /* if len==0, del_ifp will be NULL causing the delegate to be removed */
2708 bcopy(data, name, len);
2709 name[len] = 0;
2710 result = ifnet_find_by_name(name, &del_ifp);
2711 }
2712 if (result == 0) {
2713 printf("%s IPSEC_OPT_SET_DELEGATE_INTERFACE %s to %s\n",
2714 __func__, pcb->ipsec_ifp->if_xname,
2715 del_ifp ? del_ifp->if_xname : "NULL");
2716
2717 result = ifnet_set_delegate(pcb->ipsec_ifp, del_ifp);
2718 if (del_ifp)
2719 ifnet_release(del_ifp);
2720 }
2721 break;
2722 }
2723
2724 case IPSEC_OPT_OUTPUT_TRAFFIC_CLASS: {
2725 if (len != sizeof(int)) {
2726 result = EMSGSIZE;
2727 break;
2728 }
2729 if (pcb->ipsec_ifp == NULL) {
2730 // Only can set after connecting
2731 result = EINVAL;
2732 break;
2733 }
2734 mbuf_svc_class_t output_service_class = so_tc2msc(*(int *)data);
2735 if (output_service_class == MBUF_SC_UNSPEC) {
2736 pcb->ipsec_output_service_class = MBUF_SC_OAM;
2737 } else {
2738 pcb->ipsec_output_service_class = output_service_class;
2739 }
2740 printf("%s IPSEC_OPT_OUTPUT_TRAFFIC_CLASS %s svc %d\n",
2741 __func__, pcb->ipsec_ifp->if_xname,
2742 pcb->ipsec_output_service_class);
2743 break;
2744 }
2745
2746#if IPSEC_NEXUS
2747 case IPSEC_OPT_ENABLE_CHANNEL: {
2748 if (len != sizeof(int)) {
2749 result = EMSGSIZE;
2750 break;
2751 }
2752 if (pcb->ipsec_ifp == NULL) {
2753 // Only can set after connecting
2754 result = EINVAL;
2755 break;
2756 }
2757 if (*(int *)data) {
2758 result = ipsec_enable_channel(pcb, current_proc());
2759 } else {
2760 result = ipsec_disable_channel(pcb);
2761 }
2762 break;
2763 }
2764
2765 case IPSEC_OPT_ENABLE_FLOWSWITCH: {
2766 if (len != sizeof(int)) {
2767 result = EMSGSIZE;
2768 break;
2769 }
2770 if (pcb->ipsec_ifp == NULL) {
2771 // Only can set after connecting
2772 result = EINVAL;
2773 break;
2774 }
2775 if (!if_is_netagent_enabled()) {
2776 result = ENOTSUP;
2777 break;
2778 }
2779 if (uuid_is_null(pcb->ipsec_nx.ms_agent)) {
2780 result = ENOENT;
2781 break;
2782 }
2783
2784 if (*(int *)data) {
2785 if_add_netagent(pcb->ipsec_ifp, pcb->ipsec_nx.ms_agent);
2786 pcb->ipsec_needs_netagent = true;
2787 } else {
2788 pcb->ipsec_needs_netagent = false;
2789 if_delete_netagent(pcb->ipsec_ifp, pcb->ipsec_nx.ms_agent);
2790 }
2791 break;
2792 }
2793
2794 case IPSEC_OPT_INPUT_FRAG_SIZE: {
2795 if (len != sizeof(u_int32_t)) {
2796 result = EMSGSIZE;
2797 break;
2798 }
2799 u_int32_t input_frag_size = *(u_int32_t *)data;
2800 if (input_frag_size <= sizeof(struct ip6_hdr)) {
2801 pcb->ipsec_frag_size_set = FALSE;
2802 pcb->ipsec_input_frag_size = 0;
2803 } else {
2804 printf("SET FRAG SIZE TO %u\n", input_frag_size);
2805 pcb->ipsec_frag_size_set = TRUE;
2806 pcb->ipsec_input_frag_size = input_frag_size;
2807 }
2808 break;
2809 }
2810 case IPSEC_OPT_ENABLE_NETIF: {
2811 if (len != sizeof(int)) {
2812 result = EMSGSIZE;
2813 break;
2814 }
2815 if (pcb->ipsec_ifp != NULL) {
2816 // Only can set before connecting
2817 result = EINVAL;
2818 break;
2819 }
2820 lck_rw_lock_exclusive(&pcb->ipsec_pcb_lock);
2821 pcb->ipsec_use_netif = !!(*(int *)data);
2822 lck_rw_unlock_exclusive(&pcb->ipsec_pcb_lock);
2823 break;
2824 }
2825 case IPSEC_OPT_SLOT_SIZE: {
2826 if (len != sizeof(u_int32_t)) {
2827 result = EMSGSIZE;
2828 break;
2829 }
2830 if (pcb->ipsec_ifp != NULL) {
2831 // Only can set before connecting
2832 result = EINVAL;
2833 break;
2834 }
2835 u_int32_t slot_size = *(u_int32_t *)data;
2836 if (slot_size < IPSEC_IF_MIN_SLOT_SIZE ||
2837 slot_size > IPSEC_IF_MAX_SLOT_SIZE) {
2838 return (EINVAL);
2839 }
2840 pcb->ipsec_slot_size = slot_size;
2841 break;
2842 }
2843 case IPSEC_OPT_NETIF_RING_SIZE: {
2844 if (len != sizeof(u_int32_t)) {
2845 result = EMSGSIZE;
2846 break;
2847 }
2848 if (pcb->ipsec_ifp != NULL) {
2849 // Only can set before connecting
2850 result = EINVAL;
2851 break;
2852 }
2853 u_int32_t ring_size = *(u_int32_t *)data;
2854 if (ring_size < IPSEC_IF_MIN_RING_SIZE ||
2855 ring_size > IPSEC_IF_MAX_RING_SIZE) {
2856 return (EINVAL);
2857 }
2858 pcb->ipsec_netif_ring_size = ring_size;
2859 break;
2860 }
2861 case IPSEC_OPT_TX_FSW_RING_SIZE: {
2862 if (len != sizeof(u_int32_t)) {
2863 result = EMSGSIZE;
2864 break;
2865 }
2866 if (pcb->ipsec_ifp != NULL) {
2867 // Only can set before connecting
2868 result = EINVAL;
2869 break;
2870 }
2871 u_int32_t ring_size = *(u_int32_t *)data;
2872 if (ring_size < IPSEC_IF_MIN_RING_SIZE ||
2873 ring_size > IPSEC_IF_MAX_RING_SIZE) {
2874 return (EINVAL);
2875 }
2876 pcb->ipsec_tx_fsw_ring_size = ring_size;
2877 break;
2878 }
2879 case IPSEC_OPT_RX_FSW_RING_SIZE: {
2880 if (len != sizeof(u_int32_t)) {
2881 result = EMSGSIZE;
2882 break;
2883 }
2884 if (pcb->ipsec_ifp != NULL) {
2885 // Only can set before connecting
2886 result = EINVAL;
2887 break;
2888 }
2889 u_int32_t ring_size = *(u_int32_t *)data;
2890 if (ring_size < IPSEC_IF_MIN_RING_SIZE ||
2891 ring_size > IPSEC_IF_MAX_RING_SIZE) {
2892 return (EINVAL);
2893 }
2894 pcb->ipsec_rx_fsw_ring_size = ring_size;
2895 break;
2896 }
2897
2898#endif // IPSEC_NEXUS
2899
2900 default:
2901 result = ENOPROTOOPT;
2902 break;
2903 }
2904
2905 return result;
2906}
2907
2908static errno_t
2909ipsec_ctl_getopt(__unused kern_ctl_ref kctlref,
2910 __unused u_int32_t unit,
2911 void *unitinfo,
2912 int opt,
2913 void *data,
2914 size_t *len)
2915{
2916 struct ipsec_pcb *pcb = unitinfo;
2917 errno_t result = 0;
2918
2919 switch (opt) {
2920 case IPSEC_OPT_FLAGS: {
2921 if (*len != sizeof(u_int32_t)) {
2922 result = EMSGSIZE;
2923 } else {
2924 *(u_int32_t *)data = pcb->ipsec_flags;
2925 }
2926 break;
2927 }
2928
2929 case IPSEC_OPT_EXT_IFDATA_STATS: {
2930 if (*len != sizeof(int)) {
2931 result = EMSGSIZE;
2932 } else {
2933 *(int *)data = (pcb->ipsec_ext_ifdata_stats) ? 1 : 0;
2934 }
2935 break;
2936 }
2937
2938 case IPSEC_OPT_IFNAME: {
2939 if (*len < MIN(strlen(pcb->ipsec_if_xname) + 1, sizeof(pcb->ipsec_if_xname))) {
2940 result = EMSGSIZE;
2941 } else {
2942 if (pcb->ipsec_ifp == NULL) {
2943 // Only can get after connecting
2944 result = EINVAL;
2945 break;
2946 }
2947 *len = snprintf(data, *len, "%s", pcb->ipsec_if_xname) + 1;
2948 }
2949 break;
2950 }
2951
2952 case IPSEC_OPT_OUTPUT_TRAFFIC_CLASS: {
2953 if (*len != sizeof(int)) {
2954 result = EMSGSIZE;
2955 } else {
2956 *(int *)data = so_svc2tc(pcb->ipsec_output_service_class);
2957 }
2958 break;
2959 }
2960
2961#if IPSEC_NEXUS
2962
2963 case IPSEC_OPT_ENABLE_CHANNEL: {
2964 if (*len != sizeof(int)) {
2965 result = EMSGSIZE;
2966 } else {
2967 lck_rw_lock_shared(&pcb->ipsec_pcb_lock);
2968 *(int *)data = pcb->ipsec_kpipe_enabled;
2969 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
2970 }
2971 break;
2972 }
2973
2974 case IPSEC_OPT_ENABLE_FLOWSWITCH: {
2975 if (*len != sizeof(int)) {
2976 result = EMSGSIZE;
2977 } else {
2978 *(int *)data = if_check_netagent(pcb->ipsec_ifp, pcb->ipsec_nx.ms_agent);
2979 }
2980 break;
2981 }
2982
2983 case IPSEC_OPT_ENABLE_NETIF: {
2984 if (*len != sizeof(int)) {
2985 result = EMSGSIZE;
2986 } else {
2987 lck_rw_lock_shared(&pcb->ipsec_pcb_lock);
2988 *(int *)data = !!pcb->ipsec_use_netif;
2989 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
2990 }
2991 break;
2992 }
2993
2994 case IPSEC_OPT_GET_CHANNEL_UUID: {
2995 lck_rw_lock_shared(&pcb->ipsec_pcb_lock);
2996 if (uuid_is_null(pcb->ipsec_kpipe_uuid)) {
2997 result = ENXIO;
2998 } else if (*len != sizeof(uuid_t)) {
2999 result = EMSGSIZE;
3000 } else {
3001 uuid_copy(data, pcb->ipsec_kpipe_uuid);
3002 }
3003 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
3004 break;
3005 }
3006
3007 case IPSEC_OPT_INPUT_FRAG_SIZE: {
3008 if (*len != sizeof(u_int32_t)) {
3009 result = EMSGSIZE;
3010 } else {
3011 *(u_int32_t *)data = pcb->ipsec_input_frag_size;
3012 }
3013 break;
3014 }
3015 case IPSEC_OPT_SLOT_SIZE: {
3016 if (*len != sizeof(u_int32_t)) {
3017 result = EMSGSIZE;
3018 } else {
3019 *(u_int32_t *)data = pcb->ipsec_slot_size;
3020 }
3021 break;
3022 }
3023 case IPSEC_OPT_NETIF_RING_SIZE: {
3024 if (*len != sizeof(u_int32_t)) {
3025 result = EMSGSIZE;
3026 } else {
3027 *(u_int32_t *)data = pcb->ipsec_netif_ring_size;
3028 }
3029 break;
3030 }
3031 case IPSEC_OPT_TX_FSW_RING_SIZE: {
3032 if (*len != sizeof(u_int32_t)) {
3033 result = EMSGSIZE;
3034 } else {
3035 *(u_int32_t *)data = pcb->ipsec_tx_fsw_ring_size;
3036 }
3037 break;
3038 }
3039 case IPSEC_OPT_RX_FSW_RING_SIZE: {
3040 if (*len != sizeof(u_int32_t)) {
3041 result = EMSGSIZE;
3042 } else {
3043 *(u_int32_t *)data = pcb->ipsec_rx_fsw_ring_size;
3044 }
3045 break;
3046 }
3047
3048#endif // IPSEC_NEXUS
3049
3050 default: {
3051 result = ENOPROTOOPT;
3052 break;
3053 }
3054 }
3055
3056 return result;
3057}
3058
3059/* Network Interface functions */
3060static errno_t
3061ipsec_output(ifnet_t interface,
3062 mbuf_t data)
3063{
3064 struct ipsec_pcb *pcb = ifnet_softc(interface);
3065 struct ipsec_output_state ipsec_state;
3066 struct route ro;
3067 struct route_in6 ro6;
3068 int length;
3069 struct ip *ip;
3070 struct ip6_hdr *ip6;
3071 struct ip_out_args ipoa;
3072 struct ip6_out_args ip6oa;
3073 int error = 0;
3074 u_int ip_version = 0;
3075 int flags = 0;
3076 struct flowadv *adv = NULL;
3077
3078 // Make sure this packet isn't looping through the interface
3079 if (necp_get_last_interface_index_from_packet(data) == interface->if_index) {
3080 error = EINVAL;
3081 goto ipsec_output_err;
3082 }
3083
3084 // Mark the interface so NECP can evaluate tunnel policy
3085 necp_mark_packet_from_interface(data, interface);
3086
3087 ip = mtod(data, struct ip *);
3088 ip_version = ip->ip_v;
3089
3090 switch (ip_version) {
3091 case 4: {
3092#if IPSEC_NEXUS
3093 if (!pcb->ipsec_use_netif)
3094#endif // IPSEC_NEXUS
3095 {
3096 int af = AF_INET;
3097 bpf_tap_out(pcb->ipsec_ifp, DLT_NULL, data, &af, sizeof(af));
3098 }
3099
3100 /* Apply encryption */
3101 memset(&ipsec_state, 0, sizeof(ipsec_state));
3102 ipsec_state.m = data;
3103 ipsec_state.dst = (struct sockaddr *)&ip->ip_dst;
3104 memset(&ipsec_state.ro, 0, sizeof(ipsec_state.ro));
3105
3106 error = ipsec4_interface_output(&ipsec_state, interface);
3107 /* Tunneled in IPv6 - packet is gone */
3108 if (error == 0 && ipsec_state.tunneled == 6) {
3109 goto done;
3110 }
3111
3112 data = ipsec_state.m;
3113 if (error || data == NULL) {
3114 if (error) {
3115 printf("ipsec_output: ipsec4_output error %d.\n", error);
3116 }
3117 goto ipsec_output_err;
3118 }
3119
3120 /* Set traffic class, set flow */
3121 m_set_service_class(data, pcb->ipsec_output_service_class);
3122 data->m_pkthdr.pkt_flowsrc = FLOWSRC_IFNET;
3123 data->m_pkthdr.pkt_flowid = interface->if_flowhash;
3124 data->m_pkthdr.pkt_proto = ip->ip_p;
3125 data->m_pkthdr.pkt_flags = (PKTF_FLOW_ID | PKTF_FLOW_ADV | PKTF_FLOW_LOCALSRC);
3126
3127 /* Flip endian-ness for ip_output */
3128 ip = mtod(data, struct ip *);
3129 NTOHS(ip->ip_len);
3130 NTOHS(ip->ip_off);
3131
3132 /* Increment statistics */
3133 length = mbuf_pkthdr_len(data);
3134 ifnet_stat_increment_out(interface, 1, length, 0);
3135
3136 /* Send to ip_output */
3137 memset(&ro, 0, sizeof(ro));
3138
3139 flags = (IP_OUTARGS | /* Passing out args to specify interface */
3140 IP_NOIPSEC); /* To ensure the packet doesn't go through ipsec twice */
3141
3142 memset(&ipoa, 0, sizeof(ipoa));
3143 ipoa.ipoa_flowadv.code = 0;
3144 ipoa.ipoa_flags = IPOAF_SELECT_SRCIF | IPOAF_BOUND_SRCADDR;
3145 if (ipsec_state.outgoing_if) {
3146 ipoa.ipoa_boundif = ipsec_state.outgoing_if;
3147 ipoa.ipoa_flags |= IPOAF_BOUND_IF;
3148 }
3149 ipsec_set_ipoa_for_interface(pcb->ipsec_ifp, &ipoa);
3150
3151 adv = &ipoa.ipoa_flowadv;
3152
3153 (void)ip_output(data, NULL, &ro, flags, NULL, &ipoa);
3154 data = NULL;
3155
3156 if (adv->code == FADV_FLOW_CONTROLLED || adv->code == FADV_SUSPENDED) {
3157 error = ENOBUFS;
3158 ifnet_disable_output(interface);
3159 }
3160
3161 goto done;
3162 }
3163 case 6: {
3164#if IPSEC_NEXUS
3165 if (!pcb->ipsec_use_netif)
3166#endif // IPSEC_NEXUS
3167 {
3168 int af = AF_INET6;
3169 bpf_tap_out(pcb->ipsec_ifp, DLT_NULL, data, &af, sizeof(af));
3170 }
3171
3172 data = ipsec6_splithdr(data);
3173 if (data == NULL) {
3174 printf("ipsec_output: ipsec6_splithdr returned NULL\n");
3175 goto ipsec_output_err;
3176 }
3177
3178 ip6 = mtod(data, struct ip6_hdr *);
3179
3180 memset(&ipsec_state, 0, sizeof(ipsec_state));
3181 ipsec_state.m = data;
3182 ipsec_state.dst = (struct sockaddr *)&ip6->ip6_dst;
3183 memset(&ipsec_state.ro, 0, sizeof(ipsec_state.ro));
3184
3185 error = ipsec6_interface_output(&ipsec_state, interface, &ip6->ip6_nxt, ipsec_state.m);
3186 if (error == 0 && ipsec_state.tunneled == 4) { /* tunneled in IPv4 - packet is gone */
3187 goto done;
3188 }
3189 data = ipsec_state.m;
3190 if (error || data == NULL) {
3191 if (error) {
3192 printf("ipsec_output: ipsec6_output error %d\n", error);
3193 }
3194 goto ipsec_output_err;
3195 }
3196
3197 /* Set traffic class, set flow */
3198 m_set_service_class(data, pcb->ipsec_output_service_class);
3199 data->m_pkthdr.pkt_flowsrc = FLOWSRC_IFNET;
3200 data->m_pkthdr.pkt_flowid = interface->if_flowhash;
3201 data->m_pkthdr.pkt_proto = ip6->ip6_nxt;
3202 data->m_pkthdr.pkt_flags = (PKTF_FLOW_ID | PKTF_FLOW_ADV | PKTF_FLOW_LOCALSRC);
3203
3204 /* Increment statistics */
3205 length = mbuf_pkthdr_len(data);
3206 ifnet_stat_increment_out(interface, 1, length, 0);
3207
3208 /* Send to ip6_output */
3209 memset(&ro6, 0, sizeof(ro6));
3210
3211 flags = IPV6_OUTARGS;
3212
3213 memset(&ip6oa, 0, sizeof(ip6oa));
3214 ip6oa.ip6oa_flowadv.code = 0;
3215 ip6oa.ip6oa_flags = IP6OAF_SELECT_SRCIF | IP6OAF_BOUND_SRCADDR;
3216 if (ipsec_state.outgoing_if) {
3217 ip6oa.ip6oa_boundif = ipsec_state.outgoing_if;
3218 ip6oa.ip6oa_flags |= IP6OAF_BOUND_IF;
3219 }
3220 ipsec_set_ip6oa_for_interface(pcb->ipsec_ifp, &ip6oa);
3221
3222 adv = &ip6oa.ip6oa_flowadv;
3223
3224 (void) ip6_output(data, NULL, &ro6, flags, NULL, NULL, &ip6oa);
3225 data = NULL;
3226
3227 if (adv->code == FADV_FLOW_CONTROLLED || adv->code == FADV_SUSPENDED) {
3228 error = ENOBUFS;
3229 ifnet_disable_output(interface);
3230 }
3231
3232 goto done;
3233 }
3234 default: {
3235 printf("ipsec_output: Received unknown packet version %d.\n", ip_version);
3236 error = EINVAL;
3237 goto ipsec_output_err;
3238 }
3239 }
3240
3241done:
3242 return error;
3243
3244ipsec_output_err:
3245 if (data)
3246 mbuf_freem(data);
3247 goto done;
3248}
3249
3250static void
3251ipsec_start(ifnet_t interface)
3252{
3253 mbuf_t data;
3254 struct ipsec_pcb *pcb = ifnet_softc(interface);
3255
3256 VERIFY(pcb != NULL);
3257 for (;;) {
3258 if (ifnet_dequeue(interface, &data) != 0)
3259 break;
3260 if (ipsec_output(interface, data) != 0)
3261 break;
3262 }
3263}
3264
3265/* Network Interface functions */
3266static errno_t
3267ipsec_demux(__unused ifnet_t interface,
3268 mbuf_t data,
3269 __unused char *frame_header,
3270 protocol_family_t *protocol)
3271{
3272 struct ip *ip;
3273 u_int ip_version;
3274
3275 while (data != NULL && mbuf_len(data) < 1) {
3276 data = mbuf_next(data);
3277 }
3278
3279 if (data == NULL)
3280 return ENOENT;
3281
3282 ip = mtod(data, struct ip *);
3283 ip_version = ip->ip_v;
3284
3285 switch(ip_version) {
3286 case 4:
3287 *protocol = PF_INET;
3288 return 0;
3289 case 6:
3290 *protocol = PF_INET6;
3291 return 0;
3292 default:
3293 break;
3294 }
3295
3296 return 0;
3297}
3298
3299static errno_t
3300ipsec_add_proto(__unused ifnet_t interface,
3301 protocol_family_t protocol,
3302 __unused const struct ifnet_demux_desc *demux_array,
3303 __unused u_int32_t demux_count)
3304{
3305 switch(protocol) {
3306 case PF_INET:
3307 return 0;
3308 case PF_INET6:
3309 return 0;
3310 default:
3311 break;
3312 }
3313
3314 return ENOPROTOOPT;
3315}
3316
3317static errno_t
3318ipsec_del_proto(__unused ifnet_t interface,
3319 __unused protocol_family_t protocol)
3320{
3321 return 0;
3322}
3323
3324static errno_t
3325ipsec_ioctl(ifnet_t interface,
3326 u_long command,
3327 void *data)
3328{
3329#if IPSEC_NEXUS
3330 struct ipsec_pcb *pcb = ifnet_softc(interface);
3331#endif
3332 errno_t result = 0;
3333
3334 switch(command) {
3335 case SIOCSIFMTU: {
3336#if IPSEC_NEXUS
3337 if (pcb->ipsec_use_netif) {
3338 // Make sure we can fit packets in the channel buffers
3339 if (((uint64_t)((struct ifreq*)data)->ifr_mtu) > pcb->ipsec_slot_size) {
3340 result = EINVAL;
3341 } else {
3342 ifnet_set_mtu(interface, (uint32_t)((struct ifreq*)data)->ifr_mtu);
3343 }
3344 } else
3345#endif // IPSEC_NEXUS
3346 {
3347 ifnet_set_mtu(interface, ((struct ifreq*)data)->ifr_mtu);
3348 }
3349 break;
3350 }
3351
3352 case SIOCSIFFLAGS:
3353 /* ifioctl() takes care of it */
3354 break;
3355
3356 default:
3357 result = EOPNOTSUPP;
3358 }
3359
3360 return result;
3361}
3362
3363static void
3364ipsec_detached(ifnet_t interface)
3365{
3366 struct ipsec_pcb *pcb = ifnet_softc(interface);
3367 (void)ifnet_release(interface);
3368 ipsec_free_pcb(pcb, true);
3369}
3370
3371/* Protocol Handlers */
3372
3373static errno_t
3374ipsec_proto_input(ifnet_t interface,
3375 protocol_family_t protocol,
3376 mbuf_t m,
3377 __unused char *frame_header)
3378{
3379 mbuf_pkthdr_setrcvif(m, interface);
3380
3381#if IPSEC_NEXUS
3382 struct ipsec_pcb *pcb = ifnet_softc(interface);
3383 if (!pcb->ipsec_use_netif)
3384#endif // IPSEC_NEXUS
3385 {
3386 uint32_t af = 0;
3387 struct ip *ip = mtod(m, struct ip *);
3388 if (ip->ip_v == 4) {
3389 af = AF_INET;
3390 } else if (ip->ip_v == 6) {
3391 af = AF_INET6;
3392 }
3393 bpf_tap_in(interface, DLT_NULL, m, &af, sizeof(af));
3394 pktap_input(interface, protocol, m, NULL);
3395 }
3396
3397 int32_t pktlen = m->m_pkthdr.len;
3398 if (proto_input(protocol, m) != 0) {
3399 ifnet_stat_increment_in(interface, 0, 0, 1);
3400 m_freem(m);
3401 } else {
3402 ifnet_stat_increment_in(interface, 1, pktlen, 0);
3403 }
3404
3405 return 0;
3406}
3407
3408static errno_t
3409ipsec_proto_pre_output(__unused ifnet_t interface,
3410 protocol_family_t protocol,
3411 __unused mbuf_t *packet,
3412 __unused const struct sockaddr *dest,
3413 __unused void *route,
3414 __unused char *frame_type,
3415 __unused char *link_layer_dest)
3416{
3417
3418 *(protocol_family_t *)(void *)frame_type = protocol;
3419 return 0;
3420}
3421
3422static errno_t
3423ipsec_attach_proto(ifnet_t interface,
3424 protocol_family_t protocol)
3425{
3426 struct ifnet_attach_proto_param proto;
3427 errno_t result;
3428
3429 bzero(&proto, sizeof(proto));
3430 proto.input = ipsec_proto_input;
3431 proto.pre_output = ipsec_proto_pre_output;
3432
3433 result = ifnet_attach_protocol(interface, protocol, &proto);
3434 if (result != 0 && result != EEXIST) {
3435 printf("ipsec_attach_inet - ifnet_attach_protocol %d failed: %d\n",
3436 protocol, result);
3437 }
3438
3439 return result;
3440}
3441
3442errno_t
3443ipsec_inject_inbound_packet(ifnet_t interface,
3444 mbuf_t packet)
3445{
3446#if IPSEC_NEXUS
3447 struct ipsec_pcb *pcb = ifnet_softc(interface);
3448
3449 if (pcb->ipsec_use_netif) {
3450 lck_rw_lock_shared(&pcb->ipsec_pcb_lock);
3451
3452 lck_mtx_lock(&pcb->ipsec_input_chain_lock);
3453 if (pcb->ipsec_input_chain != NULL) {
3454 pcb->ipsec_input_chain_last->m_nextpkt = packet;
3455 } else {
3456 pcb->ipsec_input_chain = packet;
3457 }
3458 while (packet->m_nextpkt) {
3459 VERIFY(packet != packet->m_nextpkt);
3460 packet = packet->m_nextpkt;
3461 }
3462 pcb->ipsec_input_chain_last = packet;
3463 lck_mtx_unlock(&pcb->ipsec_input_chain_lock);
3464
3465 kern_channel_ring_t rx_ring = pcb->ipsec_netif_rxring;
3466 lck_rw_unlock_shared(&pcb->ipsec_pcb_lock);
3467
3468 if (rx_ring != NULL) {
3469 kern_channel_notify(rx_ring, 0);
3470 }
3471
3472 return (0);
3473 } else
3474#endif // IPSEC_NEXUS
3475 {
3476 errno_t error;
3477 protocol_family_t protocol;
3478 if ((error = ipsec_demux(interface, packet, NULL, &protocol)) != 0) {
3479 return error;
3480 }
3481
3482 return ipsec_proto_input(interface, protocol, packet, NULL);
3483 }
3484}
3485
3486void
3487ipsec_set_pkthdr_for_interface(ifnet_t interface, mbuf_t packet, int family)
3488{
3489 if (packet != NULL && interface != NULL) {
3490 struct ipsec_pcb *pcb = ifnet_softc(interface);
3491 if (pcb != NULL) {
3492 /* Set traffic class, set flow */
3493 m_set_service_class(packet, pcb->ipsec_output_service_class);
3494 packet->m_pkthdr.pkt_flowsrc = FLOWSRC_IFNET;
3495 packet->m_pkthdr.pkt_flowid = interface->if_flowhash;
3496 if (family == AF_INET) {
3497 struct ip *ip = mtod(packet, struct ip *);
3498 packet->m_pkthdr.pkt_proto = ip->ip_p;
3499 } else if (family == AF_INET6) {
3500 struct ip6_hdr *ip6 = mtod(packet, struct ip6_hdr *);
3501 packet->m_pkthdr.pkt_proto = ip6->ip6_nxt;
3502 }
3503 packet->m_pkthdr.pkt_flags = (PKTF_FLOW_ID | PKTF_FLOW_ADV | PKTF_FLOW_LOCALSRC);
3504 }
3505 }
3506}
3507
3508void
3509ipsec_set_ipoa_for_interface(ifnet_t interface, struct ip_out_args *ipoa)
3510{
3511 struct ipsec_pcb *pcb;
3512
3513 if (interface == NULL || ipoa == NULL)
3514 return;
3515 pcb = ifnet_softc(interface);
3516
3517 if (net_qos_policy_restricted == 0) {
3518 ipoa->ipoa_flags |= IPOAF_QOSMARKING_ALLOWED;
3519 ipoa->ipoa_sotc = so_svc2tc(pcb->ipsec_output_service_class);
3520 } else if (pcb->ipsec_output_service_class != MBUF_SC_VO ||
3521 net_qos_policy_restrict_avapps != 0) {
3522 ipoa->ipoa_flags &= ~IPOAF_QOSMARKING_ALLOWED;
3523 } else {
3524 ipoa->ipoa_flags |= IP6OAF_QOSMARKING_ALLOWED;
3525 ipoa->ipoa_sotc = SO_TC_VO;
3526 }
3527}
3528
3529void
3530ipsec_set_ip6oa_for_interface(ifnet_t interface, struct ip6_out_args *ip6oa)
3531{
3532 struct ipsec_pcb *pcb;
3533
3534 if (interface == NULL || ip6oa == NULL)
3535 return;
3536 pcb = ifnet_softc(interface);
3537
3538 if (net_qos_policy_restricted == 0) {
3539 ip6oa->ip6oa_flags |= IPOAF_QOSMARKING_ALLOWED;
3540 ip6oa->ip6oa_sotc = so_svc2tc(pcb->ipsec_output_service_class);
3541 } else if (pcb->ipsec_output_service_class != MBUF_SC_VO ||
3542 net_qos_policy_restrict_avapps != 0) {
3543 ip6oa->ip6oa_flags &= ~IPOAF_QOSMARKING_ALLOWED;
3544 } else {
3545 ip6oa->ip6oa_flags |= IP6OAF_QOSMARKING_ALLOWED;
3546 ip6oa->ip6oa_sotc = SO_TC_VO;
3547 }
3548}
3549