1 /*-
2 * Copyright (c) 2014, Bryan Venteicher <bryanv@FreeBSD.org>
3 * All rights reserved.
4 * Copyright (c) 2020, Chelsio Communications.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 * notice unmodified, this list of conditions, and the following
11 * disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
17 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
18 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
19 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
20 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
21 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
22 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
23 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
25 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26 */
27
28 #include "opt_inet.h"
29 #include "opt_inet6.h"
30
31 #include <sys/param.h>
32 #include <sys/eventhandler.h>
33 #include <sys/kernel.h>
34 #include <sys/lock.h>
35 #include <sys/hash.h>
36 #include <sys/malloc.h>
37 #include <sys/mbuf.h>
38 #include <sys/module.h>
39 #include <sys/refcount.h>
40 #include <sys/rmlock.h>
41 #include <sys/priv.h>
42 #include <sys/proc.h>
43 #include <sys/queue.h>
44 #include <sys/sbuf.h>
45 #include <sys/socket.h>
46 #include <sys/socketvar.h>
47 #include <sys/sockio.h>
48 #include <sys/sysctl.h>
49 #include <sys/systm.h>
50
51 #include <net/bpf.h>
52 #include <net/ethernet.h>
53 #include <net/if.h>
54 #include <net/if_var.h>
55 #include <net/if_private.h>
56 #include <net/if_clone.h>
57 #include <net/if_dl.h>
58 #include <net/if_media.h>
59 #include <net/if_types.h>
60 #include <net/if_vxlan.h>
61 #include <net/netisr.h>
62 #include <net/route.h>
63 #include <net/route/nhop.h>
64
65 #include <netinet/in.h>
66 #include <netinet/in_systm.h>
67 #include <netinet/in_var.h>
68 #include <netinet/in_pcb.h>
69 #include <netinet/ip.h>
70 #include <netinet/ip6.h>
71 #include <netinet/ip_var.h>
72 #include <netinet/udp.h>
73 #include <netinet/udp_var.h>
74 #include <netinet/in_fib.h>
75 #include <netinet6/in6_fib.h>
76
77 #include <netinet6/ip6_var.h>
78 #include <netinet6/scope6_var.h>
79
80 struct vxlan_softc;
81 LIST_HEAD(vxlan_softc_head, vxlan_softc);
82
83 struct sx vxlan_sx;
84 SX_SYSINIT(vxlan, &vxlan_sx, "VXLAN global start/stop lock");
85
86 struct vxlan_socket_mc_info {
87 union vxlan_sockaddr vxlsomc_saddr;
88 union vxlan_sockaddr vxlsomc_gaddr;
89 int vxlsomc_ifidx;
90 int vxlsomc_users;
91 };
92
93 /*
94 * The maximum MTU of encapsulated ethernet frame within IPv4/UDP packet.
95 */
96 #define VXLAN_MAX_MTU (IP_MAXPACKET - \
97 60 /* Maximum IPv4 header len */ - \
98 sizeof(struct udphdr) - \
99 sizeof(struct vxlan_header) - \
100 ETHER_HDR_LEN - ETHER_CRC_LEN - ETHER_VLAN_ENCAP_LEN)
101 #define VXLAN_BASIC_IFCAPS (IFCAP_LINKSTATE | IFCAP_JUMBO_MTU)
102
103 #define VXLAN_SO_MC_MAX_GROUPS 32
104
105 #define VXLAN_SO_VNI_HASH_SHIFT 6
106 #define VXLAN_SO_VNI_HASH_SIZE (1 << VXLAN_SO_VNI_HASH_SHIFT)
107 #define VXLAN_SO_VNI_HASH(_vni) ((_vni) % VXLAN_SO_VNI_HASH_SIZE)
108
109 struct vxlan_socket {
110 struct socket *vxlso_sock;
111 struct rmlock vxlso_lock;
112 u_int vxlso_refcnt;
113 union vxlan_sockaddr vxlso_laddr;
114 LIST_ENTRY(vxlan_socket) vxlso_entry;
115 struct vxlan_softc_head vxlso_vni_hash[VXLAN_SO_VNI_HASH_SIZE];
116 struct vxlan_socket_mc_info vxlso_mc[VXLAN_SO_MC_MAX_GROUPS];
117 };
118
119 #define VXLAN_SO_RLOCK(_vso, _p) rm_rlock(&(_vso)->vxlso_lock, (_p))
120 #define VXLAN_SO_RUNLOCK(_vso, _p) rm_runlock(&(_vso)->vxlso_lock, (_p))
121 #define VXLAN_SO_WLOCK(_vso) rm_wlock(&(_vso)->vxlso_lock)
122 #define VXLAN_SO_WUNLOCK(_vso) rm_wunlock(&(_vso)->vxlso_lock)
123 #define VXLAN_SO_LOCK_ASSERT(_vso) \
124 rm_assert(&(_vso)->vxlso_lock, RA_LOCKED)
125 #define VXLAN_SO_LOCK_WASSERT(_vso) \
126 rm_assert(&(_vso)->vxlso_lock, RA_WLOCKED)
127
128 #define VXLAN_SO_ACQUIRE(_vso) refcount_acquire(&(_vso)->vxlso_refcnt)
129 #define VXLAN_SO_RELEASE(_vso) refcount_release(&(_vso)->vxlso_refcnt)
130
131 struct vxlan_ftable_entry {
132 LIST_ENTRY(vxlan_ftable_entry) vxlfe_hash;
133 uint16_t vxlfe_flags;
134 uint8_t vxlfe_mac[ETHER_ADDR_LEN];
135 union vxlan_sockaddr vxlfe_raddr;
136 time_t vxlfe_expire;
137 };
138
139 #define VXLAN_FE_FLAG_DYNAMIC 0x01
140 #define VXLAN_FE_FLAG_STATIC 0x02
141
142 #define VXLAN_FE_IS_DYNAMIC(_fe) \
143 ((_fe)->vxlfe_flags & VXLAN_FE_FLAG_DYNAMIC)
144
145 #define VXLAN_SC_FTABLE_SHIFT 9
146 #define VXLAN_SC_FTABLE_SIZE (1 << VXLAN_SC_FTABLE_SHIFT)
147 #define VXLAN_SC_FTABLE_MASK (VXLAN_SC_FTABLE_SIZE - 1)
148 #define VXLAN_SC_FTABLE_HASH(_sc, _mac) \
149 (vxlan_mac_hash(_sc, _mac) % VXLAN_SC_FTABLE_SIZE)
150
151 LIST_HEAD(vxlan_ftable_head, vxlan_ftable_entry);
152
153 struct vxlan_statistics {
154 uint32_t ftable_nospace;
155 uint32_t ftable_lock_upgrade_failed;
156 counter_u64_t txcsum;
157 counter_u64_t tso;
158 counter_u64_t rxcsum;
159 };
160
161 struct vxlan_softc {
162 struct ifnet *vxl_ifp;
163 int vxl_reqcap;
164 u_int vxl_fibnum;
165 struct vxlan_socket *vxl_sock;
166 uint32_t vxl_vni;
167 union vxlan_sockaddr vxl_src_addr;
168 union vxlan_sockaddr vxl_dst_addr;
169 uint32_t vxl_flags;
170 #define VXLAN_FLAG_INIT 0x0001
171 #define VXLAN_FLAG_TEARDOWN 0x0002
172 #define VXLAN_FLAG_LEARN 0x0004
173 #define VXLAN_FLAG_USER_MTU 0x0008
174
175 uint32_t vxl_port_hash_key;
176 uint16_t vxl_min_port;
177 uint16_t vxl_max_port;
178 uint8_t vxl_ttl;
179
180 /* Lookup table from MAC address to forwarding entry. */
181 uint32_t vxl_ftable_cnt;
182 uint32_t vxl_ftable_max;
183 uint32_t vxl_ftable_timeout;
184 uint32_t vxl_ftable_hash_key;
185 struct vxlan_ftable_head *vxl_ftable;
186
187 /* Derived from vxl_dst_addr. */
188 struct vxlan_ftable_entry vxl_default_fe;
189
190 struct ip_moptions *vxl_im4o;
191 struct ip6_moptions *vxl_im6o;
192
193 struct rmlock vxl_lock;
194 volatile u_int vxl_refcnt;
195
196 int vxl_unit;
197 int vxl_vso_mc_index;
198 struct vxlan_statistics vxl_stats;
199 struct sysctl_oid *vxl_sysctl_node;
200 struct sysctl_ctx_list vxl_sysctl_ctx;
201 struct callout vxl_callout;
202 struct ether_addr vxl_hwaddr;
203 int vxl_mc_ifindex;
204 struct ifnet *vxl_mc_ifp;
205 struct ifmedia vxl_media;
206 char vxl_mc_ifname[IFNAMSIZ];
207 LIST_ENTRY(vxlan_softc) vxl_entry;
208 LIST_ENTRY(vxlan_softc) vxl_ifdetach_list;
209
210 /* For rate limiting errors on the tx fast path. */
211 struct timeval err_time;
212 int err_pps;
213 };
214
215 #define VXLAN_RLOCK(_sc, _p) rm_rlock(&(_sc)->vxl_lock, (_p))
216 #define VXLAN_RUNLOCK(_sc, _p) rm_runlock(&(_sc)->vxl_lock, (_p))
217 #define VXLAN_WLOCK(_sc) rm_wlock(&(_sc)->vxl_lock)
218 #define VXLAN_WUNLOCK(_sc) rm_wunlock(&(_sc)->vxl_lock)
219 #define VXLAN_LOCK_WOWNED(_sc) rm_wowned(&(_sc)->vxl_lock)
220 #define VXLAN_LOCK_ASSERT(_sc) rm_assert(&(_sc)->vxl_lock, RA_LOCKED)
221 #define VXLAN_LOCK_WASSERT(_sc) rm_assert(&(_sc)->vxl_lock, RA_WLOCKED)
222 #define VXLAN_UNLOCK(_sc, _p) do { \
223 if (VXLAN_LOCK_WOWNED(_sc)) \
224 VXLAN_WUNLOCK(_sc); \
225 else \
226 VXLAN_RUNLOCK(_sc, _p); \
227 } while (0)
228
229 #define VXLAN_ACQUIRE(_sc) refcount_acquire(&(_sc)->vxl_refcnt)
230 #define VXLAN_RELEASE(_sc) refcount_release(&(_sc)->vxl_refcnt)
231
232 #define satoconstsin(sa) ((const struct sockaddr_in *)(sa))
233 #define satoconstsin6(sa) ((const struct sockaddr_in6 *)(sa))
234
235 struct vxlanudphdr {
236 struct udphdr vxlh_udp;
237 struct vxlan_header vxlh_hdr;
238 } __packed;
239
240 static int vxlan_ftable_addr_cmp(const uint8_t *, const uint8_t *);
241 static void vxlan_ftable_init(struct vxlan_softc *);
242 static void vxlan_ftable_fini(struct vxlan_softc *);
243 static void vxlan_ftable_flush(struct vxlan_softc *, int);
244 static void vxlan_ftable_expire(struct vxlan_softc *);
245 static int vxlan_ftable_update_locked(struct vxlan_softc *,
246 const union vxlan_sockaddr *, const uint8_t *,
247 struct rm_priotracker *);
248 static int vxlan_ftable_learn(struct vxlan_softc *,
249 const struct sockaddr *, const uint8_t *);
250 static int vxlan_ftable_sysctl_dump(SYSCTL_HANDLER_ARGS);
251
252 static struct vxlan_ftable_entry *
253 vxlan_ftable_entry_alloc(void);
254 static void vxlan_ftable_entry_free(struct vxlan_ftable_entry *);
255 static void vxlan_ftable_entry_init(struct vxlan_softc *,
256 struct vxlan_ftable_entry *, const uint8_t *,
257 const struct sockaddr *, uint32_t);
258 static void vxlan_ftable_entry_destroy(struct vxlan_softc *,
259 struct vxlan_ftable_entry *);
260 static int vxlan_ftable_entry_insert(struct vxlan_softc *,
261 struct vxlan_ftable_entry *);
262 static struct vxlan_ftable_entry *
263 vxlan_ftable_entry_lookup(struct vxlan_softc *,
264 const uint8_t *);
265 static void vxlan_ftable_entry_dump(struct vxlan_ftable_entry *,
266 struct sbuf *);
267
268 static struct vxlan_socket *
269 vxlan_socket_alloc(const union vxlan_sockaddr *);
270 static void vxlan_socket_destroy(struct vxlan_socket *);
271 static void vxlan_socket_release(struct vxlan_socket *);
272 static struct vxlan_socket *
273 vxlan_socket_lookup(union vxlan_sockaddr *vxlsa);
274 static void vxlan_socket_insert(struct vxlan_socket *);
275 static int vxlan_socket_init(struct vxlan_socket *, struct ifnet *);
276 static int vxlan_socket_bind(struct vxlan_socket *, struct ifnet *);
277 static int vxlan_socket_create(struct ifnet *, int,
278 const union vxlan_sockaddr *, struct vxlan_socket **);
279 static void vxlan_socket_ifdetach(struct vxlan_socket *,
280 struct ifnet *, struct vxlan_softc_head *);
281
282 static struct vxlan_socket *
283 vxlan_socket_mc_lookup(const union vxlan_sockaddr *);
284 static int vxlan_sockaddr_mc_info_match(
285 const struct vxlan_socket_mc_info *,
286 const union vxlan_sockaddr *,
287 const union vxlan_sockaddr *, int);
288 static int vxlan_socket_mc_join_group(struct vxlan_socket *,
289 const union vxlan_sockaddr *, const union vxlan_sockaddr *,
290 int *, union vxlan_sockaddr *);
291 static int vxlan_socket_mc_leave_group(struct vxlan_socket *,
292 const union vxlan_sockaddr *,
293 const union vxlan_sockaddr *, int);
294 static int vxlan_socket_mc_add_group(struct vxlan_socket *,
295 const union vxlan_sockaddr *, const union vxlan_sockaddr *,
296 int, int *);
297 static void vxlan_socket_mc_release_group_by_idx(struct vxlan_socket *,
298 int);
299
300 static struct vxlan_softc *
301 vxlan_socket_lookup_softc_locked(struct vxlan_socket *,
302 uint32_t);
303 static struct vxlan_softc *
304 vxlan_socket_lookup_softc(struct vxlan_socket *, uint32_t);
305 static int vxlan_socket_insert_softc(struct vxlan_socket *,
306 struct vxlan_softc *);
307 static void vxlan_socket_remove_softc(struct vxlan_socket *,
308 struct vxlan_softc *);
309
310 static struct ifnet *
311 vxlan_multicast_if_ref(struct vxlan_softc *, int);
312 static void vxlan_free_multicast(struct vxlan_softc *);
313 static int vxlan_setup_multicast_interface(struct vxlan_softc *);
314
315 static int vxlan_setup_multicast(struct vxlan_softc *);
316 static int vxlan_setup_socket(struct vxlan_softc *);
317 #ifdef INET6
318 static void vxlan_setup_zero_checksum_port(struct vxlan_softc *);
319 #endif
320 static void vxlan_setup_interface_hdrlen(struct vxlan_softc *);
321 static int vxlan_valid_init_config(struct vxlan_softc *);
322 static void vxlan_init_wait(struct vxlan_softc *);
323 static void vxlan_init_complete(struct vxlan_softc *);
324 static void vxlan_init(void *);
325 static void vxlan_release(struct vxlan_softc *);
326 static void vxlan_teardown_wait(struct vxlan_softc *);
327 static void vxlan_teardown_complete(struct vxlan_softc *);
328 static void vxlan_teardown_locked(struct vxlan_softc *);
329 static void vxlan_teardown(struct vxlan_softc *);
330 static void vxlan_ifdetach(struct vxlan_softc *, struct ifnet *,
331 struct vxlan_softc_head *);
332 static void vxlan_timer(void *);
333
334 static int vxlan_ctrl_get_config(struct vxlan_softc *, void *);
335 static int vxlan_ctrl_set_vni(struct vxlan_softc *, void *);
336 static int vxlan_ctrl_set_local_addr(struct vxlan_softc *, void *);
337 static int vxlan_ctrl_set_remote_addr(struct vxlan_softc *, void *);
338 static int vxlan_ctrl_set_local_port(struct vxlan_softc *, void *);
339 static int vxlan_ctrl_set_remote_port(struct vxlan_softc *, void *);
340 static int vxlan_ctrl_set_port_range(struct vxlan_softc *, void *);
341 static int vxlan_ctrl_set_ftable_timeout(struct vxlan_softc *, void *);
342 static int vxlan_ctrl_set_ftable_max(struct vxlan_softc *, void *);
343 static int vxlan_ctrl_set_multicast_if(struct vxlan_softc * , void *);
344 static int vxlan_ctrl_set_ttl(struct vxlan_softc *, void *);
345 static int vxlan_ctrl_set_learn(struct vxlan_softc *, void *);
346 static int vxlan_ctrl_ftable_entry_add(struct vxlan_softc *, void *);
347 static int vxlan_ctrl_ftable_entry_rem(struct vxlan_softc *, void *);
348 static int vxlan_ctrl_flush(struct vxlan_softc *, void *);
349 static int vxlan_ioctl_drvspec(struct vxlan_softc *,
350 struct ifdrv *, int);
351 static int vxlan_ioctl_ifflags(struct vxlan_softc *);
352 static int vxlan_ioctl(struct ifnet *, u_long, caddr_t);
353
354 #if defined(INET) || defined(INET6)
355 static uint16_t vxlan_pick_source_port(struct vxlan_softc *, struct mbuf *);
356 static void vxlan_encap_header(struct vxlan_softc *, struct mbuf *,
357 int, uint16_t, uint16_t);
358 #endif
359 static int vxlan_encap4(struct vxlan_softc *,
360 const union vxlan_sockaddr *, struct mbuf *);
361 static int vxlan_encap6(struct vxlan_softc *,
362 const union vxlan_sockaddr *, struct mbuf *);
363 static int vxlan_transmit(struct ifnet *, struct mbuf *);
364 static void vxlan_qflush(struct ifnet *);
365 static bool vxlan_rcv_udp_packet(struct mbuf *, int, struct inpcb *,
366 const struct sockaddr *, void *);
367 static int vxlan_input(struct vxlan_socket *, uint32_t, struct mbuf **,
368 const struct sockaddr *);
369
370 static int vxlan_stats_alloc(struct vxlan_softc *);
371 static void vxlan_stats_free(struct vxlan_softc *);
372 static void vxlan_set_default_config(struct vxlan_softc *);
373 static int vxlan_set_user_config(struct vxlan_softc *,
374 struct ifvxlanparam *);
375 static int vxlan_set_reqcap(struct vxlan_softc *, struct ifnet *, int);
376 static void vxlan_set_hwcaps(struct vxlan_softc *);
377 static int vxlan_clone_create(struct if_clone *, char *, size_t,
378 struct ifc_data *, struct ifnet **);
379 static int vxlan_clone_destroy(struct if_clone *, struct ifnet *, uint32_t);
380
381 static uint32_t vxlan_mac_hash(struct vxlan_softc *, const uint8_t *);
382 static int vxlan_media_change(struct ifnet *);
383 static void vxlan_media_status(struct ifnet *, struct ifmediareq *);
384
385 static int vxlan_sockaddr_cmp(const union vxlan_sockaddr *,
386 const struct sockaddr *);
387 static void vxlan_sockaddr_copy(union vxlan_sockaddr *,
388 const struct sockaddr *);
389 static int vxlan_sockaddr_in_equal(const union vxlan_sockaddr *,
390 const struct sockaddr *);
391 static void vxlan_sockaddr_in_copy(union vxlan_sockaddr *,
392 const struct sockaddr *);
393 static int vxlan_sockaddr_supported(const union vxlan_sockaddr *, int);
394 static int vxlan_sockaddr_in_any(const union vxlan_sockaddr *);
395 static int vxlan_sockaddr_in_multicast(const union vxlan_sockaddr *);
396 static int vxlan_sockaddr_in6_embedscope(union vxlan_sockaddr *);
397
398 static int vxlan_can_change_config(struct vxlan_softc *);
399 static int vxlan_check_vni(uint32_t);
400 static int vxlan_check_ttl(int);
401 static int vxlan_check_ftable_timeout(uint32_t);
402 static int vxlan_check_ftable_max(uint32_t);
403
404 static void vxlan_sysctl_setup(struct vxlan_softc *);
405 static void vxlan_sysctl_destroy(struct vxlan_softc *);
406 static int vxlan_tunable_int(struct vxlan_softc *, const char *, int);
407
408 static void vxlan_ifdetach_event(void *, struct ifnet *);
409 static void vxlan_load(void);
410 static void vxlan_unload(void);
411 static int vxlan_modevent(module_t, int, void *);
412
413 static const char vxlan_name[] = "vxlan";
414 static MALLOC_DEFINE(M_VXLAN, vxlan_name,
415 "Virtual eXtensible LAN Interface");
416 static struct if_clone *vxlan_cloner;
417
418 static struct mtx vxlan_list_mtx;
419 #define VXLAN_LIST_LOCK() mtx_lock(&vxlan_list_mtx)
420 #define VXLAN_LIST_UNLOCK() mtx_unlock(&vxlan_list_mtx)
421
422 static LIST_HEAD(, vxlan_socket) vxlan_socket_list;
423
424 static eventhandler_tag vxlan_ifdetach_event_tag;
425
426 SYSCTL_DECL(_net_link);
427 SYSCTL_NODE(_net_link, OID_AUTO, vxlan, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
428 "Virtual eXtensible Local Area Network");
429
430 static int vxlan_legacy_port = 0;
431 TUNABLE_INT("net.link.vxlan.legacy_port", &vxlan_legacy_port);
432 static int vxlan_reuse_port = 0;
433 TUNABLE_INT("net.link.vxlan.reuse_port", &vxlan_reuse_port);
434
435 /*
436 * This macro controls the default upper limitation on nesting of vxlan
437 * tunnels. By default it is 3, as the overhead of IPv6 vxlan tunnel is 70
438 * bytes, this will create at most 210 bytes overhead and the most inner
439 * tunnel's MTU will be 1290 which will meet IPv6 minimum MTU size 1280.
440 * Be careful to configure the tunnels when raising the limit. A large
441 * number of nested tunnels can introduce system crash.
442 */
443 #ifndef MAX_VXLAN_NEST
444 #define MAX_VXLAN_NEST 3
445 #endif
446 static int max_vxlan_nesting = MAX_VXLAN_NEST;
447 SYSCTL_INT(_net_link_vxlan, OID_AUTO, max_nesting, CTLFLAG_RW,
448 &max_vxlan_nesting, 0, "Max nested tunnels");
449
450 /* Default maximum number of addresses in the forwarding table. */
451 #ifndef VXLAN_FTABLE_MAX
452 #define VXLAN_FTABLE_MAX 2000
453 #endif
454
455 /* Timeout (in seconds) of addresses learned in the forwarding table. */
456 #ifndef VXLAN_FTABLE_TIMEOUT
457 #define VXLAN_FTABLE_TIMEOUT (20 * 60)
458 #endif
459
460 /*
461 * Maximum timeout (in seconds) of addresses learned in the forwarding
462 * table.
463 */
464 #ifndef VXLAN_FTABLE_MAX_TIMEOUT
465 #define VXLAN_FTABLE_MAX_TIMEOUT (60 * 60 * 24)
466 #endif
467
468 /* Number of seconds between pruning attempts of the forwarding table. */
469 #ifndef VXLAN_FTABLE_PRUNE
470 #define VXLAN_FTABLE_PRUNE (5 * 60)
471 #endif
472
473 static int vxlan_ftable_prune_period = VXLAN_FTABLE_PRUNE;
474
475 struct vxlan_control {
476 int (*vxlc_func)(struct vxlan_softc *, void *);
477 int vxlc_argsize;
478 int vxlc_flags;
479 #define VXLAN_CTRL_FLAG_COPYIN 0x01
480 #define VXLAN_CTRL_FLAG_COPYOUT 0x02
481 #define VXLAN_CTRL_FLAG_SUSER 0x04
482 };
483
484 static const struct vxlan_control vxlan_control_table[] = {
485 [VXLAN_CMD_GET_CONFIG] =
486 { vxlan_ctrl_get_config, sizeof(struct ifvxlancfg),
487 VXLAN_CTRL_FLAG_COPYOUT
488 },
489
490 [VXLAN_CMD_SET_VNI] =
491 { vxlan_ctrl_set_vni, sizeof(struct ifvxlancmd),
492 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
493 },
494
495 [VXLAN_CMD_SET_LOCAL_ADDR] =
496 { vxlan_ctrl_set_local_addr, sizeof(struct ifvxlancmd),
497 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
498 },
499
500 [VXLAN_CMD_SET_REMOTE_ADDR] =
501 { vxlan_ctrl_set_remote_addr, sizeof(struct ifvxlancmd),
502 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
503 },
504
505 [VXLAN_CMD_SET_LOCAL_PORT] =
506 { vxlan_ctrl_set_local_port, sizeof(struct ifvxlancmd),
507 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
508 },
509
510 [VXLAN_CMD_SET_REMOTE_PORT] =
511 { vxlan_ctrl_set_remote_port, sizeof(struct ifvxlancmd),
512 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
513 },
514
515 [VXLAN_CMD_SET_PORT_RANGE] =
516 { vxlan_ctrl_set_port_range, sizeof(struct ifvxlancmd),
517 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
518 },
519
520 [VXLAN_CMD_SET_FTABLE_TIMEOUT] =
521 { vxlan_ctrl_set_ftable_timeout, sizeof(struct ifvxlancmd),
522 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
523 },
524
525 [VXLAN_CMD_SET_FTABLE_MAX] =
526 { vxlan_ctrl_set_ftable_max, sizeof(struct ifvxlancmd),
527 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
528 },
529
530 [VXLAN_CMD_SET_MULTICAST_IF] =
531 { vxlan_ctrl_set_multicast_if, sizeof(struct ifvxlancmd),
532 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
533 },
534
535 [VXLAN_CMD_SET_TTL] =
536 { vxlan_ctrl_set_ttl, sizeof(struct ifvxlancmd),
537 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
538 },
539
540 [VXLAN_CMD_SET_LEARN] =
541 { vxlan_ctrl_set_learn, sizeof(struct ifvxlancmd),
542 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
543 },
544
545 [VXLAN_CMD_FTABLE_ENTRY_ADD] =
546 { vxlan_ctrl_ftable_entry_add, sizeof(struct ifvxlancmd),
547 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
548 },
549
550 [VXLAN_CMD_FTABLE_ENTRY_REM] =
551 { vxlan_ctrl_ftable_entry_rem, sizeof(struct ifvxlancmd),
552 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
553 },
554
555 [VXLAN_CMD_FLUSH] =
556 { vxlan_ctrl_flush, sizeof(struct ifvxlancmd),
557 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
558 },
559 };
560
561 static const int vxlan_control_table_size = nitems(vxlan_control_table);
562
563 static int
vxlan_ftable_addr_cmp(const uint8_t * a,const uint8_t * b)564 vxlan_ftable_addr_cmp(const uint8_t *a, const uint8_t *b)
565 {
566 int i, d;
567
568 for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++)
569 d = ((int)a[i]) - ((int)b[i]);
570
571 return (d);
572 }
573
574 static void
vxlan_ftable_init(struct vxlan_softc * sc)575 vxlan_ftable_init(struct vxlan_softc *sc)
576 {
577 int i;
578
579 sc->vxl_ftable = malloc(sizeof(struct vxlan_ftable_head) *
580 VXLAN_SC_FTABLE_SIZE, M_VXLAN, M_ZERO | M_WAITOK);
581
582 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++)
583 LIST_INIT(&sc->vxl_ftable[i]);
584 sc->vxl_ftable_hash_key = arc4random();
585 }
586
587 static void
vxlan_ftable_fini(struct vxlan_softc * sc)588 vxlan_ftable_fini(struct vxlan_softc *sc)
589 {
590 int i;
591
592 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++) {
593 KASSERT(LIST_EMPTY(&sc->vxl_ftable[i]),
594 ("%s: vxlan %p ftable[%d] not empty", __func__, sc, i));
595 }
596 MPASS(sc->vxl_ftable_cnt == 0);
597
598 free(sc->vxl_ftable, M_VXLAN);
599 sc->vxl_ftable = NULL;
600 }
601
602 static void
vxlan_ftable_flush(struct vxlan_softc * sc,int all)603 vxlan_ftable_flush(struct vxlan_softc *sc, int all)
604 {
605 struct vxlan_ftable_entry *fe, *tfe;
606 int i;
607
608 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++) {
609 LIST_FOREACH_SAFE(fe, &sc->vxl_ftable[i], vxlfe_hash, tfe) {
610 if (all || VXLAN_FE_IS_DYNAMIC(fe))
611 vxlan_ftable_entry_destroy(sc, fe);
612 }
613 }
614 }
615
616 static void
vxlan_ftable_expire(struct vxlan_softc * sc)617 vxlan_ftable_expire(struct vxlan_softc *sc)
618 {
619 struct vxlan_ftable_entry *fe, *tfe;
620 int i;
621
622 VXLAN_LOCK_WASSERT(sc);
623
624 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++) {
625 LIST_FOREACH_SAFE(fe, &sc->vxl_ftable[i], vxlfe_hash, tfe) {
626 if (VXLAN_FE_IS_DYNAMIC(fe) &&
627 time_uptime >= fe->vxlfe_expire)
628 vxlan_ftable_entry_destroy(sc, fe);
629 }
630 }
631 }
632
633 static int
vxlan_ftable_update_locked(struct vxlan_softc * sc,const union vxlan_sockaddr * vxlsa,const uint8_t * mac,struct rm_priotracker * tracker)634 vxlan_ftable_update_locked(struct vxlan_softc *sc,
635 const union vxlan_sockaddr *vxlsa, const uint8_t *mac,
636 struct rm_priotracker *tracker)
637 {
638 struct vxlan_ftable_entry *fe;
639 int error __unused;
640
641 VXLAN_LOCK_ASSERT(sc);
642
643 again:
644 /*
645 * A forwarding entry for this MAC address might already exist. If
646 * so, update it, otherwise create a new one. We may have to upgrade
647 * the lock if we have to change or create an entry.
648 */
649 fe = vxlan_ftable_entry_lookup(sc, mac);
650 if (fe != NULL) {
651 fe->vxlfe_expire = time_uptime + sc->vxl_ftable_timeout;
652
653 if (!VXLAN_FE_IS_DYNAMIC(fe) ||
654 vxlan_sockaddr_in_equal(&fe->vxlfe_raddr, &vxlsa->sa))
655 return (0);
656 if (!VXLAN_LOCK_WOWNED(sc)) {
657 VXLAN_RUNLOCK(sc, tracker);
658 VXLAN_WLOCK(sc);
659 sc->vxl_stats.ftable_lock_upgrade_failed++;
660 goto again;
661 }
662 vxlan_sockaddr_in_copy(&fe->vxlfe_raddr, &vxlsa->sa);
663 return (0);
664 }
665
666 if (!VXLAN_LOCK_WOWNED(sc)) {
667 VXLAN_RUNLOCK(sc, tracker);
668 VXLAN_WLOCK(sc);
669 sc->vxl_stats.ftable_lock_upgrade_failed++;
670 goto again;
671 }
672
673 if (sc->vxl_ftable_cnt >= sc->vxl_ftable_max) {
674 sc->vxl_stats.ftable_nospace++;
675 return (ENOSPC);
676 }
677
678 fe = vxlan_ftable_entry_alloc();
679 if (fe == NULL)
680 return (ENOMEM);
681
682 vxlan_ftable_entry_init(sc, fe, mac, &vxlsa->sa, VXLAN_FE_FLAG_DYNAMIC);
683
684 /* The prior lookup failed, so the insert should not. */
685 error = vxlan_ftable_entry_insert(sc, fe);
686 MPASS(error == 0);
687
688 return (0);
689 }
690
691 static int
vxlan_ftable_learn(struct vxlan_softc * sc,const struct sockaddr * sa,const uint8_t * mac)692 vxlan_ftable_learn(struct vxlan_softc *sc, const struct sockaddr *sa,
693 const uint8_t *mac)
694 {
695 struct rm_priotracker tracker;
696 union vxlan_sockaddr vxlsa;
697 int error;
698
699 /*
700 * The source port may be randomly selected by the remote host, so
701 * use the port of the default destination address.
702 */
703 vxlan_sockaddr_copy(&vxlsa, sa);
704 vxlsa.in4.sin_port = sc->vxl_dst_addr.in4.sin_port;
705
706 if (VXLAN_SOCKADDR_IS_IPV6(&vxlsa)) {
707 error = vxlan_sockaddr_in6_embedscope(&vxlsa);
708 if (error)
709 return (error);
710 }
711
712 VXLAN_RLOCK(sc, &tracker);
713 error = vxlan_ftable_update_locked(sc, &vxlsa, mac, &tracker);
714 VXLAN_UNLOCK(sc, &tracker);
715
716 return (error);
717 }
718
719 static int
vxlan_ftable_sysctl_dump(SYSCTL_HANDLER_ARGS)720 vxlan_ftable_sysctl_dump(SYSCTL_HANDLER_ARGS)
721 {
722 struct rm_priotracker tracker;
723 struct sbuf sb;
724 struct vxlan_softc *sc;
725 struct vxlan_ftable_entry *fe;
726 size_t size;
727 int i, error;
728
729 /*
730 * This is mostly intended for debugging during development. It is
731 * not practical to dump an entire large table this way.
732 */
733
734 sc = arg1;
735 size = PAGE_SIZE; /* Calculate later. */
736
737 sbuf_new(&sb, NULL, size, SBUF_FIXEDLEN);
738 sbuf_putc(&sb, '\n');
739
740 VXLAN_RLOCK(sc, &tracker);
741 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++) {
742 LIST_FOREACH(fe, &sc->vxl_ftable[i], vxlfe_hash) {
743 if (sbuf_error(&sb) != 0)
744 break;
745 vxlan_ftable_entry_dump(fe, &sb);
746 }
747 }
748 VXLAN_RUNLOCK(sc, &tracker);
749
750 if (sbuf_len(&sb) == 1)
751 sbuf_setpos(&sb, 0);
752
753 sbuf_finish(&sb);
754 error = sysctl_handle_string(oidp, sbuf_data(&sb), sbuf_len(&sb), req);
755 sbuf_delete(&sb);
756
757 return (error);
758 }
759
760 static struct vxlan_ftable_entry *
vxlan_ftable_entry_alloc(void)761 vxlan_ftable_entry_alloc(void)
762 {
763 struct vxlan_ftable_entry *fe;
764
765 fe = malloc(sizeof(*fe), M_VXLAN, M_ZERO | M_NOWAIT);
766
767 return (fe);
768 }
769
770 static void
vxlan_ftable_entry_free(struct vxlan_ftable_entry * fe)771 vxlan_ftable_entry_free(struct vxlan_ftable_entry *fe)
772 {
773
774 free(fe, M_VXLAN);
775 }
776
777 static void
vxlan_ftable_entry_init(struct vxlan_softc * sc,struct vxlan_ftable_entry * fe,const uint8_t * mac,const struct sockaddr * sa,uint32_t flags)778 vxlan_ftable_entry_init(struct vxlan_softc *sc, struct vxlan_ftable_entry *fe,
779 const uint8_t *mac, const struct sockaddr *sa, uint32_t flags)
780 {
781
782 fe->vxlfe_flags = flags;
783 fe->vxlfe_expire = time_uptime + sc->vxl_ftable_timeout;
784 memcpy(fe->vxlfe_mac, mac, ETHER_ADDR_LEN);
785 vxlan_sockaddr_copy(&fe->vxlfe_raddr, sa);
786 }
787
788 static void
vxlan_ftable_entry_destroy(struct vxlan_softc * sc,struct vxlan_ftable_entry * fe)789 vxlan_ftable_entry_destroy(struct vxlan_softc *sc,
790 struct vxlan_ftable_entry *fe)
791 {
792
793 sc->vxl_ftable_cnt--;
794 LIST_REMOVE(fe, vxlfe_hash);
795 vxlan_ftable_entry_free(fe);
796 }
797
798 static int
vxlan_ftable_entry_insert(struct vxlan_softc * sc,struct vxlan_ftable_entry * fe)799 vxlan_ftable_entry_insert(struct vxlan_softc *sc,
800 struct vxlan_ftable_entry *fe)
801 {
802 struct vxlan_ftable_entry *lfe;
803 uint32_t hash;
804 int dir;
805
806 VXLAN_LOCK_WASSERT(sc);
807 hash = VXLAN_SC_FTABLE_HASH(sc, fe->vxlfe_mac);
808
809 lfe = LIST_FIRST(&sc->vxl_ftable[hash]);
810 if (lfe == NULL) {
811 LIST_INSERT_HEAD(&sc->vxl_ftable[hash], fe, vxlfe_hash);
812 goto out;
813 }
814
815 do {
816 dir = vxlan_ftable_addr_cmp(fe->vxlfe_mac, lfe->vxlfe_mac);
817 if (dir == 0)
818 return (EEXIST);
819 if (dir > 0) {
820 LIST_INSERT_BEFORE(lfe, fe, vxlfe_hash);
821 goto out;
822 } else if (LIST_NEXT(lfe, vxlfe_hash) == NULL) {
823 LIST_INSERT_AFTER(lfe, fe, vxlfe_hash);
824 goto out;
825 } else
826 lfe = LIST_NEXT(lfe, vxlfe_hash);
827 } while (lfe != NULL);
828
829 out:
830 sc->vxl_ftable_cnt++;
831
832 return (0);
833 }
834
835 static struct vxlan_ftable_entry *
vxlan_ftable_entry_lookup(struct vxlan_softc * sc,const uint8_t * mac)836 vxlan_ftable_entry_lookup(struct vxlan_softc *sc, const uint8_t *mac)
837 {
838 struct vxlan_ftable_entry *fe;
839 uint32_t hash;
840 int dir;
841
842 VXLAN_LOCK_ASSERT(sc);
843 hash = VXLAN_SC_FTABLE_HASH(sc, mac);
844
845 LIST_FOREACH(fe, &sc->vxl_ftable[hash], vxlfe_hash) {
846 dir = vxlan_ftable_addr_cmp(mac, fe->vxlfe_mac);
847 if (dir == 0)
848 return (fe);
849 if (dir > 0)
850 break;
851 }
852
853 return (NULL);
854 }
855
856 static void
vxlan_ftable_entry_dump(struct vxlan_ftable_entry * fe,struct sbuf * sb)857 vxlan_ftable_entry_dump(struct vxlan_ftable_entry *fe, struct sbuf *sb)
858 {
859 char buf[64];
860 const union vxlan_sockaddr *sa;
861 const void *addr;
862 int i, len, af, width;
863
864 sa = &fe->vxlfe_raddr;
865 af = sa->sa.sa_family;
866 len = sbuf_len(sb);
867
868 sbuf_printf(sb, "%c 0x%02X ", VXLAN_FE_IS_DYNAMIC(fe) ? 'D' : 'S',
869 fe->vxlfe_flags);
870
871 for (i = 0; i < ETHER_ADDR_LEN - 1; i++)
872 sbuf_printf(sb, "%02X:", fe->vxlfe_mac[i]);
873 sbuf_printf(sb, "%02X ", fe->vxlfe_mac[i]);
874
875 if (af == AF_INET) {
876 addr = &sa->in4.sin_addr;
877 width = INET_ADDRSTRLEN - 1;
878 } else {
879 addr = &sa->in6.sin6_addr;
880 width = INET6_ADDRSTRLEN - 1;
881 }
882 inet_ntop(af, addr, buf, sizeof(buf));
883 sbuf_printf(sb, "%*s ", width, buf);
884
885 sbuf_printf(sb, "%08jd", (intmax_t)fe->vxlfe_expire);
886
887 sbuf_putc(sb, '\n');
888
889 /* Truncate a partial line. */
890 if (sbuf_error(sb) != 0)
891 sbuf_setpos(sb, len);
892 }
893
894 static struct vxlan_socket *
vxlan_socket_alloc(const union vxlan_sockaddr * sa)895 vxlan_socket_alloc(const union vxlan_sockaddr *sa)
896 {
897 struct vxlan_socket *vso;
898 int i;
899
900 vso = malloc(sizeof(*vso), M_VXLAN, M_WAITOK | M_ZERO);
901 rm_init(&vso->vxlso_lock, "vxlansorm");
902 refcount_init(&vso->vxlso_refcnt, 0);
903 for (i = 0; i < VXLAN_SO_VNI_HASH_SIZE; i++)
904 LIST_INIT(&vso->vxlso_vni_hash[i]);
905 vso->vxlso_laddr = *sa;
906
907 return (vso);
908 }
909
910 static void
vxlan_socket_destroy(struct vxlan_socket * vso)911 vxlan_socket_destroy(struct vxlan_socket *vso)
912 {
913 struct socket *so;
914 #ifdef INVARIANTS
915 int i;
916 struct vxlan_socket_mc_info *mc;
917
918 for (i = 0; i < VXLAN_SO_MC_MAX_GROUPS; i++) {
919 mc = &vso->vxlso_mc[i];
920 KASSERT(mc->vxlsomc_gaddr.sa.sa_family == AF_UNSPEC,
921 ("%s: socket %p mc[%d] still has address",
922 __func__, vso, i));
923 }
924
925 for (i = 0; i < VXLAN_SO_VNI_HASH_SIZE; i++) {
926 KASSERT(LIST_EMPTY(&vso->vxlso_vni_hash[i]),
927 ("%s: socket %p vni_hash[%d] not empty",
928 __func__, vso, i));
929 }
930 #endif
931 so = vso->vxlso_sock;
932 if (so != NULL) {
933 vso->vxlso_sock = NULL;
934 soclose(so);
935 }
936
937 rm_destroy(&vso->vxlso_lock);
938 free(vso, M_VXLAN);
939 }
940
941 static void
vxlan_socket_release(struct vxlan_socket * vso)942 vxlan_socket_release(struct vxlan_socket *vso)
943 {
944 int destroy;
945
946 VXLAN_LIST_LOCK();
947 destroy = VXLAN_SO_RELEASE(vso);
948 if (destroy != 0)
949 LIST_REMOVE(vso, vxlso_entry);
950 VXLAN_LIST_UNLOCK();
951
952 if (destroy != 0)
953 vxlan_socket_destroy(vso);
954 }
955
956 static struct vxlan_socket *
vxlan_socket_lookup(union vxlan_sockaddr * vxlsa)957 vxlan_socket_lookup(union vxlan_sockaddr *vxlsa)
958 {
959 struct vxlan_socket *vso;
960
961 VXLAN_LIST_LOCK();
962 LIST_FOREACH(vso, &vxlan_socket_list, vxlso_entry) {
963 if (vxlan_sockaddr_cmp(&vso->vxlso_laddr, &vxlsa->sa) == 0) {
964 VXLAN_SO_ACQUIRE(vso);
965 break;
966 }
967 }
968 VXLAN_LIST_UNLOCK();
969
970 return (vso);
971 }
972
973 static void
vxlan_socket_insert(struct vxlan_socket * vso)974 vxlan_socket_insert(struct vxlan_socket *vso)
975 {
976
977 VXLAN_LIST_LOCK();
978 VXLAN_SO_ACQUIRE(vso);
979 LIST_INSERT_HEAD(&vxlan_socket_list, vso, vxlso_entry);
980 VXLAN_LIST_UNLOCK();
981 }
982
983 static int
vxlan_socket_init(struct vxlan_socket * vso,struct ifnet * ifp)984 vxlan_socket_init(struct vxlan_socket *vso, struct ifnet *ifp)
985 {
986 struct thread *td;
987 int error;
988
989 td = curthread;
990
991 error = socreate(vso->vxlso_laddr.sa.sa_family, &vso->vxlso_sock,
992 SOCK_DGRAM, IPPROTO_UDP, td->td_ucred, td);
993 if (error) {
994 if_printf(ifp, "cannot create socket: %d\n", error);
995 return (error);
996 }
997
998 error = udp_set_kernel_tunneling(vso->vxlso_sock,
999 vxlan_rcv_udp_packet, NULL, vso);
1000 if (error) {
1001 if_printf(ifp, "cannot set tunneling function: %d\n", error);
1002 return (error);
1003 }
1004
1005 if (vxlan_reuse_port != 0) {
1006 struct sockopt sopt;
1007 int val = 1;
1008
1009 bzero(&sopt, sizeof(sopt));
1010 sopt.sopt_dir = SOPT_SET;
1011 sopt.sopt_level = IPPROTO_IP;
1012 sopt.sopt_name = SO_REUSEPORT;
1013 sopt.sopt_val = &val;
1014 sopt.sopt_valsize = sizeof(val);
1015 error = sosetopt(vso->vxlso_sock, &sopt);
1016 if (error) {
1017 if_printf(ifp,
1018 "cannot set REUSEADDR socket opt: %d\n", error);
1019 return (error);
1020 }
1021 }
1022
1023 return (0);
1024 }
1025
1026 static int
vxlan_socket_bind(struct vxlan_socket * vso,struct ifnet * ifp)1027 vxlan_socket_bind(struct vxlan_socket *vso, struct ifnet *ifp)
1028 {
1029 union vxlan_sockaddr laddr;
1030 struct thread *td;
1031 int error;
1032
1033 td = curthread;
1034 laddr = vso->vxlso_laddr;
1035
1036 error = sobind(vso->vxlso_sock, &laddr.sa, td);
1037 if (error) {
1038 if (error != EADDRINUSE)
1039 if_printf(ifp, "cannot bind socket: %d\n", error);
1040 return (error);
1041 }
1042
1043 return (0);
1044 }
1045
1046 static int
vxlan_socket_create(struct ifnet * ifp,int multicast,const union vxlan_sockaddr * saddr,struct vxlan_socket ** vsop)1047 vxlan_socket_create(struct ifnet *ifp, int multicast,
1048 const union vxlan_sockaddr *saddr, struct vxlan_socket **vsop)
1049 {
1050 union vxlan_sockaddr laddr;
1051 struct vxlan_socket *vso;
1052 int error;
1053
1054 laddr = *saddr;
1055
1056 /*
1057 * If this socket will be multicast, then only the local port
1058 * must be specified when binding.
1059 */
1060 if (multicast != 0) {
1061 if (VXLAN_SOCKADDR_IS_IPV4(&laddr))
1062 laddr.in4.sin_addr.s_addr = INADDR_ANY;
1063 #ifdef INET6
1064 else
1065 laddr.in6.sin6_addr = in6addr_any;
1066 #endif
1067 }
1068
1069 vso = vxlan_socket_alloc(&laddr);
1070 if (vso == NULL)
1071 return (ENOMEM);
1072
1073 error = vxlan_socket_init(vso, ifp);
1074 if (error)
1075 goto fail;
1076
1077 error = vxlan_socket_bind(vso, ifp);
1078 if (error)
1079 goto fail;
1080
1081 /*
1082 * There is a small window between the bind completing and
1083 * inserting the socket, so that a concurrent create may fail.
1084 * Let's not worry about that for now.
1085 */
1086 vxlan_socket_insert(vso);
1087 *vsop = vso;
1088
1089 return (0);
1090
1091 fail:
1092 vxlan_socket_destroy(vso);
1093
1094 return (error);
1095 }
1096
1097 static void
vxlan_socket_ifdetach(struct vxlan_socket * vso,struct ifnet * ifp,struct vxlan_softc_head * list)1098 vxlan_socket_ifdetach(struct vxlan_socket *vso, struct ifnet *ifp,
1099 struct vxlan_softc_head *list)
1100 {
1101 struct rm_priotracker tracker;
1102 struct vxlan_softc *sc;
1103 int i;
1104
1105 VXLAN_SO_RLOCK(vso, &tracker);
1106 for (i = 0; i < VXLAN_SO_VNI_HASH_SIZE; i++) {
1107 LIST_FOREACH(sc, &vso->vxlso_vni_hash[i], vxl_entry)
1108 vxlan_ifdetach(sc, ifp, list);
1109 }
1110 VXLAN_SO_RUNLOCK(vso, &tracker);
1111 }
1112
1113 static struct vxlan_socket *
vxlan_socket_mc_lookup(const union vxlan_sockaddr * vxlsa)1114 vxlan_socket_mc_lookup(const union vxlan_sockaddr *vxlsa)
1115 {
1116 union vxlan_sockaddr laddr;
1117 struct vxlan_socket *vso;
1118
1119 laddr = *vxlsa;
1120
1121 if (VXLAN_SOCKADDR_IS_IPV4(&laddr))
1122 laddr.in4.sin_addr.s_addr = INADDR_ANY;
1123 #ifdef INET6
1124 else
1125 laddr.in6.sin6_addr = in6addr_any;
1126 #endif
1127
1128 vso = vxlan_socket_lookup(&laddr);
1129
1130 return (vso);
1131 }
1132
1133 static int
vxlan_sockaddr_mc_info_match(const struct vxlan_socket_mc_info * mc,const union vxlan_sockaddr * group,const union vxlan_sockaddr * local,int ifidx)1134 vxlan_sockaddr_mc_info_match(const struct vxlan_socket_mc_info *mc,
1135 const union vxlan_sockaddr *group, const union vxlan_sockaddr *local,
1136 int ifidx)
1137 {
1138
1139 if (!vxlan_sockaddr_in_any(local) &&
1140 !vxlan_sockaddr_in_equal(&mc->vxlsomc_saddr, &local->sa))
1141 return (0);
1142 if (!vxlan_sockaddr_in_equal(&mc->vxlsomc_gaddr, &group->sa))
1143 return (0);
1144 if (ifidx != 0 && ifidx != mc->vxlsomc_ifidx)
1145 return (0);
1146
1147 return (1);
1148 }
1149
1150 static int
vxlan_socket_mc_join_group(struct vxlan_socket * vso,const union vxlan_sockaddr * group,const union vxlan_sockaddr * local,int * ifidx,union vxlan_sockaddr * source)1151 vxlan_socket_mc_join_group(struct vxlan_socket *vso,
1152 const union vxlan_sockaddr *group, const union vxlan_sockaddr *local,
1153 int *ifidx, union vxlan_sockaddr *source)
1154 {
1155 struct sockopt sopt;
1156 int error;
1157
1158 *source = *local;
1159
1160 if (VXLAN_SOCKADDR_IS_IPV4(group)) {
1161 struct ip_mreq mreq;
1162
1163 mreq.imr_multiaddr = group->in4.sin_addr;
1164 mreq.imr_interface = local->in4.sin_addr;
1165
1166 bzero(&sopt, sizeof(sopt));
1167 sopt.sopt_dir = SOPT_SET;
1168 sopt.sopt_level = IPPROTO_IP;
1169 sopt.sopt_name = IP_ADD_MEMBERSHIP;
1170 sopt.sopt_val = &mreq;
1171 sopt.sopt_valsize = sizeof(mreq);
1172 error = sosetopt(vso->vxlso_sock, &sopt);
1173 if (error)
1174 return (error);
1175
1176 /*
1177 * BMV: Ideally, there would be a formal way for us to get
1178 * the local interface that was selected based on the
1179 * imr_interface address. We could then update *ifidx so
1180 * vxlan_sockaddr_mc_info_match() would return a match for
1181 * later creates that explicitly set the multicast interface.
1182 *
1183 * If we really need to, we can of course look in the INP's
1184 * membership list:
1185 * sotoinpcb(vso->vxlso_sock)->inp_moptions->
1186 * imo_head[]->imf_inm->inm_ifp
1187 * similarly to imo_match_group().
1188 */
1189 source->in4.sin_addr = local->in4.sin_addr;
1190
1191 } else if (VXLAN_SOCKADDR_IS_IPV6(group)) {
1192 struct ipv6_mreq mreq;
1193
1194 mreq.ipv6mr_multiaddr = group->in6.sin6_addr;
1195 mreq.ipv6mr_interface = *ifidx;
1196
1197 bzero(&sopt, sizeof(sopt));
1198 sopt.sopt_dir = SOPT_SET;
1199 sopt.sopt_level = IPPROTO_IPV6;
1200 sopt.sopt_name = IPV6_JOIN_GROUP;
1201 sopt.sopt_val = &mreq;
1202 sopt.sopt_valsize = sizeof(mreq);
1203 error = sosetopt(vso->vxlso_sock, &sopt);
1204 if (error)
1205 return (error);
1206
1207 /*
1208 * BMV: As with IPv4, we would really like to know what
1209 * interface in6p_lookup_mcast_ifp() selected.
1210 */
1211 } else
1212 error = EAFNOSUPPORT;
1213
1214 return (error);
1215 }
1216
1217 static int
vxlan_socket_mc_leave_group(struct vxlan_socket * vso,const union vxlan_sockaddr * group,const union vxlan_sockaddr * source,int ifidx)1218 vxlan_socket_mc_leave_group(struct vxlan_socket *vso,
1219 const union vxlan_sockaddr *group, const union vxlan_sockaddr *source,
1220 int ifidx)
1221 {
1222 struct sockopt sopt;
1223 int error;
1224
1225 bzero(&sopt, sizeof(sopt));
1226 sopt.sopt_dir = SOPT_SET;
1227
1228 if (VXLAN_SOCKADDR_IS_IPV4(group)) {
1229 struct ip_mreq mreq;
1230
1231 mreq.imr_multiaddr = group->in4.sin_addr;
1232 mreq.imr_interface = source->in4.sin_addr;
1233
1234 sopt.sopt_level = IPPROTO_IP;
1235 sopt.sopt_name = IP_DROP_MEMBERSHIP;
1236 sopt.sopt_val = &mreq;
1237 sopt.sopt_valsize = sizeof(mreq);
1238 error = sosetopt(vso->vxlso_sock, &sopt);
1239
1240 } else if (VXLAN_SOCKADDR_IS_IPV6(group)) {
1241 struct ipv6_mreq mreq;
1242
1243 mreq.ipv6mr_multiaddr = group->in6.sin6_addr;
1244 mreq.ipv6mr_interface = ifidx;
1245
1246 sopt.sopt_level = IPPROTO_IPV6;
1247 sopt.sopt_name = IPV6_LEAVE_GROUP;
1248 sopt.sopt_val = &mreq;
1249 sopt.sopt_valsize = sizeof(mreq);
1250 error = sosetopt(vso->vxlso_sock, &sopt);
1251
1252 } else
1253 error = EAFNOSUPPORT;
1254
1255 return (error);
1256 }
1257
1258 static int
vxlan_socket_mc_add_group(struct vxlan_socket * vso,const union vxlan_sockaddr * group,const union vxlan_sockaddr * local,int ifidx,int * idx)1259 vxlan_socket_mc_add_group(struct vxlan_socket *vso,
1260 const union vxlan_sockaddr *group, const union vxlan_sockaddr *local,
1261 int ifidx, int *idx)
1262 {
1263 union vxlan_sockaddr source;
1264 struct vxlan_socket_mc_info *mc;
1265 int i, empty, error;
1266
1267 /*
1268 * Within a socket, the same multicast group may be used by multiple
1269 * interfaces, each with a different network identifier. But a socket
1270 * may only join a multicast group once, so keep track of the users
1271 * here.
1272 */
1273
1274 VXLAN_SO_WLOCK(vso);
1275 for (empty = 0, i = 0; i < VXLAN_SO_MC_MAX_GROUPS; i++) {
1276 mc = &vso->vxlso_mc[i];
1277
1278 if (mc->vxlsomc_gaddr.sa.sa_family == AF_UNSPEC) {
1279 empty++;
1280 continue;
1281 }
1282
1283 if (vxlan_sockaddr_mc_info_match(mc, group, local, ifidx))
1284 goto out;
1285 }
1286 VXLAN_SO_WUNLOCK(vso);
1287
1288 if (empty == 0)
1289 return (ENOSPC);
1290
1291 error = vxlan_socket_mc_join_group(vso, group, local, &ifidx, &source);
1292 if (error)
1293 return (error);
1294
1295 VXLAN_SO_WLOCK(vso);
1296 for (i = 0; i < VXLAN_SO_MC_MAX_GROUPS; i++) {
1297 mc = &vso->vxlso_mc[i];
1298
1299 if (mc->vxlsomc_gaddr.sa.sa_family == AF_UNSPEC) {
1300 vxlan_sockaddr_copy(&mc->vxlsomc_gaddr, &group->sa);
1301 vxlan_sockaddr_copy(&mc->vxlsomc_saddr, &source.sa);
1302 mc->vxlsomc_ifidx = ifidx;
1303 goto out;
1304 }
1305 }
1306 VXLAN_SO_WUNLOCK(vso);
1307
1308 error = vxlan_socket_mc_leave_group(vso, group, &source, ifidx);
1309 MPASS(error == 0);
1310
1311 return (ENOSPC);
1312
1313 out:
1314 mc->vxlsomc_users++;
1315 VXLAN_SO_WUNLOCK(vso);
1316
1317 *idx = i;
1318
1319 return (0);
1320 }
1321
1322 static void
vxlan_socket_mc_release_group_by_idx(struct vxlan_socket * vso,int idx)1323 vxlan_socket_mc_release_group_by_idx(struct vxlan_socket *vso, int idx)
1324 {
1325 union vxlan_sockaddr group, source;
1326 struct vxlan_socket_mc_info *mc;
1327 int ifidx, leave;
1328
1329 KASSERT(idx >= 0 && idx < VXLAN_SO_MC_MAX_GROUPS,
1330 ("%s: vso %p idx %d out of bounds", __func__, vso, idx));
1331
1332 leave = 0;
1333 mc = &vso->vxlso_mc[idx];
1334
1335 VXLAN_SO_WLOCK(vso);
1336 mc->vxlsomc_users--;
1337 if (mc->vxlsomc_users == 0) {
1338 group = mc->vxlsomc_gaddr;
1339 source = mc->vxlsomc_saddr;
1340 ifidx = mc->vxlsomc_ifidx;
1341 bzero(mc, sizeof(*mc));
1342 leave = 1;
1343 }
1344 VXLAN_SO_WUNLOCK(vso);
1345
1346 if (leave != 0) {
1347 /*
1348 * Our socket's membership in this group may have already
1349 * been removed if we joined through an interface that's
1350 * been detached.
1351 */
1352 vxlan_socket_mc_leave_group(vso, &group, &source, ifidx);
1353 }
1354 }
1355
1356 static struct vxlan_softc *
vxlan_socket_lookup_softc_locked(struct vxlan_socket * vso,uint32_t vni)1357 vxlan_socket_lookup_softc_locked(struct vxlan_socket *vso, uint32_t vni)
1358 {
1359 struct vxlan_softc *sc;
1360 uint32_t hash;
1361
1362 VXLAN_SO_LOCK_ASSERT(vso);
1363 hash = VXLAN_SO_VNI_HASH(vni);
1364
1365 LIST_FOREACH(sc, &vso->vxlso_vni_hash[hash], vxl_entry) {
1366 if (sc->vxl_vni == vni) {
1367 VXLAN_ACQUIRE(sc);
1368 break;
1369 }
1370 }
1371
1372 return (sc);
1373 }
1374
1375 static struct vxlan_softc *
vxlan_socket_lookup_softc(struct vxlan_socket * vso,uint32_t vni)1376 vxlan_socket_lookup_softc(struct vxlan_socket *vso, uint32_t vni)
1377 {
1378 struct rm_priotracker tracker;
1379 struct vxlan_softc *sc;
1380
1381 VXLAN_SO_RLOCK(vso, &tracker);
1382 sc = vxlan_socket_lookup_softc_locked(vso, vni);
1383 VXLAN_SO_RUNLOCK(vso, &tracker);
1384
1385 return (sc);
1386 }
1387
1388 static int
vxlan_socket_insert_softc(struct vxlan_socket * vso,struct vxlan_softc * sc)1389 vxlan_socket_insert_softc(struct vxlan_socket *vso, struct vxlan_softc *sc)
1390 {
1391 struct vxlan_softc *tsc;
1392 uint32_t vni, hash;
1393
1394 vni = sc->vxl_vni;
1395 hash = VXLAN_SO_VNI_HASH(vni);
1396
1397 VXLAN_SO_WLOCK(vso);
1398 tsc = vxlan_socket_lookup_softc_locked(vso, vni);
1399 if (tsc != NULL) {
1400 VXLAN_SO_WUNLOCK(vso);
1401 vxlan_release(tsc);
1402 return (EEXIST);
1403 }
1404
1405 VXLAN_ACQUIRE(sc);
1406 LIST_INSERT_HEAD(&vso->vxlso_vni_hash[hash], sc, vxl_entry);
1407 VXLAN_SO_WUNLOCK(vso);
1408
1409 return (0);
1410 }
1411
1412 static void
vxlan_socket_remove_softc(struct vxlan_socket * vso,struct vxlan_softc * sc)1413 vxlan_socket_remove_softc(struct vxlan_socket *vso, struct vxlan_softc *sc)
1414 {
1415
1416 VXLAN_SO_WLOCK(vso);
1417 LIST_REMOVE(sc, vxl_entry);
1418 VXLAN_SO_WUNLOCK(vso);
1419
1420 vxlan_release(sc);
1421 }
1422
1423 static struct ifnet *
vxlan_multicast_if_ref(struct vxlan_softc * sc,int ipv4)1424 vxlan_multicast_if_ref(struct vxlan_softc *sc, int ipv4)
1425 {
1426 struct ifnet *ifp;
1427
1428 VXLAN_LOCK_ASSERT(sc);
1429
1430 if (ipv4 && sc->vxl_im4o != NULL)
1431 ifp = sc->vxl_im4o->imo_multicast_ifp;
1432 else if (!ipv4 && sc->vxl_im6o != NULL)
1433 ifp = sc->vxl_im6o->im6o_multicast_ifp;
1434 else
1435 ifp = NULL;
1436
1437 if (ifp != NULL)
1438 if_ref(ifp);
1439
1440 return (ifp);
1441 }
1442
1443 static void
vxlan_free_multicast(struct vxlan_softc * sc)1444 vxlan_free_multicast(struct vxlan_softc *sc)
1445 {
1446
1447 if (sc->vxl_mc_ifp != NULL) {
1448 if_rele(sc->vxl_mc_ifp);
1449 sc->vxl_mc_ifp = NULL;
1450 sc->vxl_mc_ifindex = 0;
1451 }
1452
1453 if (sc->vxl_im4o != NULL) {
1454 free(sc->vxl_im4o, M_VXLAN);
1455 sc->vxl_im4o = NULL;
1456 }
1457
1458 if (sc->vxl_im6o != NULL) {
1459 free(sc->vxl_im6o, M_VXLAN);
1460 sc->vxl_im6o = NULL;
1461 }
1462 }
1463
1464 static int
vxlan_setup_multicast_interface(struct vxlan_softc * sc)1465 vxlan_setup_multicast_interface(struct vxlan_softc *sc)
1466 {
1467 struct ifnet *ifp;
1468
1469 ifp = ifunit_ref(sc->vxl_mc_ifname);
1470 if (ifp == NULL) {
1471 if_printf(sc->vxl_ifp, "multicast interface %s does "
1472 "not exist\n", sc->vxl_mc_ifname);
1473 return (ENOENT);
1474 }
1475
1476 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
1477 if_printf(sc->vxl_ifp, "interface %s does not support "
1478 "multicast\n", sc->vxl_mc_ifname);
1479 if_rele(ifp);
1480 return (ENOTSUP);
1481 }
1482
1483 sc->vxl_mc_ifp = ifp;
1484 sc->vxl_mc_ifindex = ifp->if_index;
1485
1486 return (0);
1487 }
1488
1489 static int
vxlan_setup_multicast(struct vxlan_softc * sc)1490 vxlan_setup_multicast(struct vxlan_softc *sc)
1491 {
1492 const union vxlan_sockaddr *group;
1493 int error;
1494
1495 group = &sc->vxl_dst_addr;
1496 error = 0;
1497
1498 if (sc->vxl_mc_ifname[0] != '\0') {
1499 error = vxlan_setup_multicast_interface(sc);
1500 if (error)
1501 return (error);
1502 }
1503
1504 /*
1505 * Initialize an multicast options structure that is sufficiently
1506 * populated for use in the respective IP output routine. This
1507 * structure is typically stored in the socket, but our sockets
1508 * may be shared among multiple interfaces.
1509 */
1510 if (VXLAN_SOCKADDR_IS_IPV4(group)) {
1511 sc->vxl_im4o = malloc(sizeof(struct ip_moptions), M_VXLAN,
1512 M_ZERO | M_WAITOK);
1513 sc->vxl_im4o->imo_multicast_ifp = sc->vxl_mc_ifp;
1514 sc->vxl_im4o->imo_multicast_ttl = sc->vxl_ttl;
1515 sc->vxl_im4o->imo_multicast_vif = -1;
1516 } else if (VXLAN_SOCKADDR_IS_IPV6(group)) {
1517 sc->vxl_im6o = malloc(sizeof(struct ip6_moptions), M_VXLAN,
1518 M_ZERO | M_WAITOK);
1519 sc->vxl_im6o->im6o_multicast_ifp = sc->vxl_mc_ifp;
1520 sc->vxl_im6o->im6o_multicast_hlim = sc->vxl_ttl;
1521 }
1522
1523 return (error);
1524 }
1525
1526 static int
vxlan_setup_socket(struct vxlan_softc * sc)1527 vxlan_setup_socket(struct vxlan_softc *sc)
1528 {
1529 struct vxlan_socket *vso;
1530 struct ifnet *ifp;
1531 union vxlan_sockaddr *saddr, *daddr;
1532 int multicast, error;
1533
1534 vso = NULL;
1535 ifp = sc->vxl_ifp;
1536 saddr = &sc->vxl_src_addr;
1537 daddr = &sc->vxl_dst_addr;
1538
1539 multicast = vxlan_sockaddr_in_multicast(daddr);
1540 MPASS(multicast != -1);
1541 sc->vxl_vso_mc_index = -1;
1542
1543 /*
1544 * Try to create the socket. If that fails, attempt to use an
1545 * existing socket.
1546 */
1547 error = vxlan_socket_create(ifp, multicast, saddr, &vso);
1548 if (error) {
1549 if (multicast != 0)
1550 vso = vxlan_socket_mc_lookup(saddr);
1551 else
1552 vso = vxlan_socket_lookup(saddr);
1553
1554 if (vso == NULL) {
1555 if_printf(ifp, "cannot create socket (error: %d), "
1556 "and no existing socket found\n", error);
1557 goto out;
1558 }
1559 }
1560
1561 if (multicast != 0) {
1562 error = vxlan_setup_multicast(sc);
1563 if (error)
1564 goto out;
1565
1566 error = vxlan_socket_mc_add_group(vso, daddr, saddr,
1567 sc->vxl_mc_ifindex, &sc->vxl_vso_mc_index);
1568 if (error)
1569 goto out;
1570 }
1571
1572 sc->vxl_sock = vso;
1573 error = vxlan_socket_insert_softc(vso, sc);
1574 if (error) {
1575 sc->vxl_sock = NULL;
1576 if_printf(ifp, "network identifier %d already exists in "
1577 "this socket\n", sc->vxl_vni);
1578 goto out;
1579 }
1580
1581 return (0);
1582
1583 out:
1584 if (vso != NULL) {
1585 if (sc->vxl_vso_mc_index != -1) {
1586 vxlan_socket_mc_release_group_by_idx(vso,
1587 sc->vxl_vso_mc_index);
1588 sc->vxl_vso_mc_index = -1;
1589 }
1590 if (multicast != 0)
1591 vxlan_free_multicast(sc);
1592 vxlan_socket_release(vso);
1593 }
1594
1595 return (error);
1596 }
1597
1598 #ifdef INET6
1599 static void
vxlan_setup_zero_checksum_port(struct vxlan_softc * sc)1600 vxlan_setup_zero_checksum_port(struct vxlan_softc *sc)
1601 {
1602
1603 if (!VXLAN_SOCKADDR_IS_IPV6(&sc->vxl_src_addr))
1604 return;
1605
1606 MPASS(sc->vxl_src_addr.in6.sin6_port != 0);
1607 MPASS(sc->vxl_dst_addr.in6.sin6_port != 0);
1608
1609 if (sc->vxl_src_addr.in6.sin6_port != sc->vxl_dst_addr.in6.sin6_port) {
1610 if_printf(sc->vxl_ifp, "port %d in src address does not match "
1611 "port %d in dst address, rfc6935_port (%d) not updated.\n",
1612 ntohs(sc->vxl_src_addr.in6.sin6_port),
1613 ntohs(sc->vxl_dst_addr.in6.sin6_port),
1614 V_zero_checksum_port);
1615 return;
1616 }
1617
1618 if (V_zero_checksum_port != 0) {
1619 if (V_zero_checksum_port !=
1620 ntohs(sc->vxl_src_addr.in6.sin6_port)) {
1621 if_printf(sc->vxl_ifp, "rfc6935_port is already set to "
1622 "%d, cannot set it to %d.\n", V_zero_checksum_port,
1623 ntohs(sc->vxl_src_addr.in6.sin6_port));
1624 }
1625 return;
1626 }
1627
1628 V_zero_checksum_port = ntohs(sc->vxl_src_addr.in6.sin6_port);
1629 if_printf(sc->vxl_ifp, "rfc6935_port set to %d\n",
1630 V_zero_checksum_port);
1631 }
1632 #endif
1633
1634 static void
vxlan_setup_interface_hdrlen(struct vxlan_softc * sc)1635 vxlan_setup_interface_hdrlen(struct vxlan_softc *sc)
1636 {
1637 struct ifnet *ifp;
1638
1639 VXLAN_LOCK_WASSERT(sc);
1640
1641 ifp = sc->vxl_ifp;
1642 ifp->if_hdrlen = ETHER_HDR_LEN + sizeof(struct vxlanudphdr);
1643
1644 if (VXLAN_SOCKADDR_IS_IPV4(&sc->vxl_dst_addr) != 0)
1645 ifp->if_hdrlen += sizeof(struct ip);
1646 else if (VXLAN_SOCKADDR_IS_IPV6(&sc->vxl_dst_addr) != 0)
1647 ifp->if_hdrlen += sizeof(struct ip6_hdr);
1648
1649 if ((sc->vxl_flags & VXLAN_FLAG_USER_MTU) == 0)
1650 ifp->if_mtu = ETHERMTU - ifp->if_hdrlen;
1651 }
1652
1653 static int
vxlan_valid_init_config(struct vxlan_softc * sc)1654 vxlan_valid_init_config(struct vxlan_softc *sc)
1655 {
1656 const char *reason;
1657
1658 if (vxlan_check_vni(sc->vxl_vni) != 0) {
1659 reason = "invalid virtual network identifier specified";
1660 goto fail;
1661 }
1662
1663 if (vxlan_sockaddr_supported(&sc->vxl_src_addr, 1) == 0) {
1664 reason = "source address type is not supported";
1665 goto fail;
1666 }
1667
1668 if (vxlan_sockaddr_supported(&sc->vxl_dst_addr, 0) == 0) {
1669 reason = "destination address type is not supported";
1670 goto fail;
1671 }
1672
1673 if (vxlan_sockaddr_in_any(&sc->vxl_dst_addr) != 0) {
1674 reason = "no valid destination address specified";
1675 goto fail;
1676 }
1677
1678 if (vxlan_sockaddr_in_multicast(&sc->vxl_dst_addr) == 0 &&
1679 sc->vxl_mc_ifname[0] != '\0') {
1680 reason = "can only specify interface with a group address";
1681 goto fail;
1682 }
1683
1684 if (vxlan_sockaddr_in_any(&sc->vxl_src_addr) == 0) {
1685 if (VXLAN_SOCKADDR_IS_IPV4(&sc->vxl_src_addr) ^
1686 VXLAN_SOCKADDR_IS_IPV4(&sc->vxl_dst_addr)) {
1687 reason = "source and destination address must both "
1688 "be either IPv4 or IPv6";
1689 goto fail;
1690 }
1691 }
1692
1693 if (sc->vxl_src_addr.in4.sin_port == 0) {
1694 reason = "local port not specified";
1695 goto fail;
1696 }
1697
1698 if (sc->vxl_dst_addr.in4.sin_port == 0) {
1699 reason = "remote port not specified";
1700 goto fail;
1701 }
1702
1703 return (0);
1704
1705 fail:
1706 if_printf(sc->vxl_ifp, "cannot initialize interface: %s\n", reason);
1707 return (EINVAL);
1708 }
1709
1710 static void
vxlan_init_wait(struct vxlan_softc * sc)1711 vxlan_init_wait(struct vxlan_softc *sc)
1712 {
1713
1714 VXLAN_LOCK_WASSERT(sc);
1715 while (sc->vxl_flags & VXLAN_FLAG_INIT)
1716 rm_sleep(sc, &sc->vxl_lock, 0, "vxlint", hz);
1717 }
1718
1719 static void
vxlan_init_complete(struct vxlan_softc * sc)1720 vxlan_init_complete(struct vxlan_softc *sc)
1721 {
1722
1723 VXLAN_WLOCK(sc);
1724 sc->vxl_flags &= ~VXLAN_FLAG_INIT;
1725 wakeup(sc);
1726 VXLAN_WUNLOCK(sc);
1727 }
1728
1729 static void
vxlan_init(void * xsc)1730 vxlan_init(void *xsc)
1731 {
1732 static const uint8_t empty_mac[ETHER_ADDR_LEN];
1733 struct vxlan_softc *sc;
1734 struct ifnet *ifp;
1735
1736 sc = xsc;
1737 ifp = sc->vxl_ifp;
1738
1739 sx_xlock(&vxlan_sx);
1740 VXLAN_WLOCK(sc);
1741 if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
1742 VXLAN_WUNLOCK(sc);
1743 sx_xunlock(&vxlan_sx);
1744 return;
1745 }
1746 sc->vxl_flags |= VXLAN_FLAG_INIT;
1747 VXLAN_WUNLOCK(sc);
1748
1749 if (vxlan_valid_init_config(sc) != 0)
1750 goto out;
1751
1752 if (vxlan_setup_socket(sc) != 0)
1753 goto out;
1754
1755 #ifdef INET6
1756 vxlan_setup_zero_checksum_port(sc);
1757 #endif
1758
1759 /* Initialize the default forwarding entry. */
1760 vxlan_ftable_entry_init(sc, &sc->vxl_default_fe, empty_mac,
1761 &sc->vxl_dst_addr.sa, VXLAN_FE_FLAG_STATIC);
1762
1763 VXLAN_WLOCK(sc);
1764 ifp->if_drv_flags |= IFF_DRV_RUNNING;
1765 callout_reset(&sc->vxl_callout, vxlan_ftable_prune_period * hz,
1766 vxlan_timer, sc);
1767 VXLAN_WUNLOCK(sc);
1768
1769 if_link_state_change(ifp, LINK_STATE_UP);
1770
1771 EVENTHANDLER_INVOKE(vxlan_start, ifp, sc->vxl_src_addr.in4.sin_family,
1772 ntohs(sc->vxl_src_addr.in4.sin_port));
1773 out:
1774 vxlan_init_complete(sc);
1775 sx_xunlock(&vxlan_sx);
1776 }
1777
1778 static void
vxlan_release(struct vxlan_softc * sc)1779 vxlan_release(struct vxlan_softc *sc)
1780 {
1781
1782 /*
1783 * The softc may be destroyed as soon as we release our reference,
1784 * so we cannot serialize the wakeup with the softc lock. We use a
1785 * timeout in our sleeps so a missed wakeup is unfortunate but not
1786 * fatal.
1787 */
1788 if (VXLAN_RELEASE(sc) != 0)
1789 wakeup(sc);
1790 }
1791
1792 static void
vxlan_teardown_wait(struct vxlan_softc * sc)1793 vxlan_teardown_wait(struct vxlan_softc *sc)
1794 {
1795
1796 VXLAN_LOCK_WASSERT(sc);
1797 while (sc->vxl_flags & VXLAN_FLAG_TEARDOWN)
1798 rm_sleep(sc, &sc->vxl_lock, 0, "vxltrn", hz);
1799 }
1800
1801 static void
vxlan_teardown_complete(struct vxlan_softc * sc)1802 vxlan_teardown_complete(struct vxlan_softc *sc)
1803 {
1804
1805 VXLAN_WLOCK(sc);
1806 sc->vxl_flags &= ~VXLAN_FLAG_TEARDOWN;
1807 wakeup(sc);
1808 VXLAN_WUNLOCK(sc);
1809 }
1810
1811 static void
vxlan_teardown_locked(struct vxlan_softc * sc)1812 vxlan_teardown_locked(struct vxlan_softc *sc)
1813 {
1814 struct ifnet *ifp;
1815 struct vxlan_socket *vso;
1816
1817 sx_assert(&vxlan_sx, SA_XLOCKED);
1818 VXLAN_LOCK_WASSERT(sc);
1819 MPASS(sc->vxl_flags & VXLAN_FLAG_TEARDOWN);
1820
1821 ifp = sc->vxl_ifp;
1822 ifp->if_flags &= ~IFF_UP;
1823 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1824 callout_stop(&sc->vxl_callout);
1825 vso = sc->vxl_sock;
1826 sc->vxl_sock = NULL;
1827
1828 VXLAN_WUNLOCK(sc);
1829 if_link_state_change(ifp, LINK_STATE_DOWN);
1830 EVENTHANDLER_INVOKE(vxlan_stop, ifp, sc->vxl_src_addr.in4.sin_family,
1831 ntohs(sc->vxl_src_addr.in4.sin_port));
1832
1833 if (vso != NULL) {
1834 vxlan_socket_remove_softc(vso, sc);
1835
1836 if (sc->vxl_vso_mc_index != -1) {
1837 vxlan_socket_mc_release_group_by_idx(vso,
1838 sc->vxl_vso_mc_index);
1839 sc->vxl_vso_mc_index = -1;
1840 }
1841 }
1842
1843 VXLAN_WLOCK(sc);
1844 while (sc->vxl_refcnt != 0)
1845 rm_sleep(sc, &sc->vxl_lock, 0, "vxldrn", hz);
1846 VXLAN_WUNLOCK(sc);
1847
1848 callout_drain(&sc->vxl_callout);
1849
1850 vxlan_free_multicast(sc);
1851 if (vso != NULL)
1852 vxlan_socket_release(vso);
1853
1854 vxlan_teardown_complete(sc);
1855 }
1856
1857 static void
vxlan_teardown(struct vxlan_softc * sc)1858 vxlan_teardown(struct vxlan_softc *sc)
1859 {
1860
1861 sx_xlock(&vxlan_sx);
1862 VXLAN_WLOCK(sc);
1863 if (sc->vxl_flags & VXLAN_FLAG_TEARDOWN) {
1864 vxlan_teardown_wait(sc);
1865 VXLAN_WUNLOCK(sc);
1866 sx_xunlock(&vxlan_sx);
1867 return;
1868 }
1869
1870 sc->vxl_flags |= VXLAN_FLAG_TEARDOWN;
1871 vxlan_teardown_locked(sc);
1872 sx_xunlock(&vxlan_sx);
1873 }
1874
1875 static void
vxlan_ifdetach(struct vxlan_softc * sc,struct ifnet * ifp,struct vxlan_softc_head * list)1876 vxlan_ifdetach(struct vxlan_softc *sc, struct ifnet *ifp,
1877 struct vxlan_softc_head *list)
1878 {
1879
1880 VXLAN_WLOCK(sc);
1881
1882 if (sc->vxl_mc_ifp != ifp)
1883 goto out;
1884 if (sc->vxl_flags & VXLAN_FLAG_TEARDOWN)
1885 goto out;
1886
1887 sc->vxl_flags |= VXLAN_FLAG_TEARDOWN;
1888 LIST_INSERT_HEAD(list, sc, vxl_ifdetach_list);
1889
1890 out:
1891 VXLAN_WUNLOCK(sc);
1892 }
1893
1894 static void
vxlan_timer(void * xsc)1895 vxlan_timer(void *xsc)
1896 {
1897 struct vxlan_softc *sc;
1898
1899 sc = xsc;
1900 VXLAN_LOCK_WASSERT(sc);
1901
1902 vxlan_ftable_expire(sc);
1903 callout_schedule(&sc->vxl_callout, vxlan_ftable_prune_period * hz);
1904 }
1905
1906 static int
vxlan_ioctl_ifflags(struct vxlan_softc * sc)1907 vxlan_ioctl_ifflags(struct vxlan_softc *sc)
1908 {
1909 struct ifnet *ifp;
1910
1911 ifp = sc->vxl_ifp;
1912
1913 if (ifp->if_flags & IFF_UP) {
1914 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1915 vxlan_init(sc);
1916 } else {
1917 if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1918 vxlan_teardown(sc);
1919 }
1920
1921 return (0);
1922 }
1923
1924 static int
vxlan_ctrl_get_config(struct vxlan_softc * sc,void * arg)1925 vxlan_ctrl_get_config(struct vxlan_softc *sc, void *arg)
1926 {
1927 struct rm_priotracker tracker;
1928 struct ifvxlancfg *cfg;
1929
1930 cfg = arg;
1931 bzero(cfg, sizeof(*cfg));
1932
1933 VXLAN_RLOCK(sc, &tracker);
1934 cfg->vxlc_vni = sc->vxl_vni;
1935 memcpy(&cfg->vxlc_local_sa, &sc->vxl_src_addr,
1936 sizeof(union vxlan_sockaddr));
1937 memcpy(&cfg->vxlc_remote_sa, &sc->vxl_dst_addr,
1938 sizeof(union vxlan_sockaddr));
1939 cfg->vxlc_mc_ifindex = sc->vxl_mc_ifindex;
1940 cfg->vxlc_ftable_cnt = sc->vxl_ftable_cnt;
1941 cfg->vxlc_ftable_max = sc->vxl_ftable_max;
1942 cfg->vxlc_ftable_timeout = sc->vxl_ftable_timeout;
1943 cfg->vxlc_port_min = sc->vxl_min_port;
1944 cfg->vxlc_port_max = sc->vxl_max_port;
1945 cfg->vxlc_learn = (sc->vxl_flags & VXLAN_FLAG_LEARN) != 0;
1946 cfg->vxlc_ttl = sc->vxl_ttl;
1947 VXLAN_RUNLOCK(sc, &tracker);
1948
1949 #ifdef INET6
1950 if (VXLAN_SOCKADDR_IS_IPV6(&cfg->vxlc_local_sa))
1951 sa6_recoverscope(&cfg->vxlc_local_sa.in6);
1952 if (VXLAN_SOCKADDR_IS_IPV6(&cfg->vxlc_remote_sa))
1953 sa6_recoverscope(&cfg->vxlc_remote_sa.in6);
1954 #endif
1955
1956 return (0);
1957 }
1958
1959 static int
vxlan_ctrl_set_vni(struct vxlan_softc * sc,void * arg)1960 vxlan_ctrl_set_vni(struct vxlan_softc *sc, void *arg)
1961 {
1962 struct ifvxlancmd *cmd;
1963 int error;
1964
1965 cmd = arg;
1966
1967 if (vxlan_check_vni(cmd->vxlcmd_vni) != 0)
1968 return (EINVAL);
1969
1970 VXLAN_WLOCK(sc);
1971 if (vxlan_can_change_config(sc)) {
1972 sc->vxl_vni = cmd->vxlcmd_vni;
1973 error = 0;
1974 } else
1975 error = EBUSY;
1976 VXLAN_WUNLOCK(sc);
1977
1978 return (error);
1979 }
1980
1981 static int
vxlan_ctrl_set_local_addr(struct vxlan_softc * sc,void * arg)1982 vxlan_ctrl_set_local_addr(struct vxlan_softc *sc, void *arg)
1983 {
1984 struct ifvxlancmd *cmd;
1985 union vxlan_sockaddr *vxlsa;
1986 int error;
1987
1988 cmd = arg;
1989 vxlsa = &cmd->vxlcmd_sa;
1990
1991 if (!VXLAN_SOCKADDR_IS_IPV46(vxlsa))
1992 return (EINVAL);
1993 if (vxlan_sockaddr_in_multicast(vxlsa) != 0)
1994 return (EINVAL);
1995 if (VXLAN_SOCKADDR_IS_IPV6(vxlsa)) {
1996 error = vxlan_sockaddr_in6_embedscope(vxlsa);
1997 if (error)
1998 return (error);
1999 }
2000
2001 VXLAN_WLOCK(sc);
2002 if (vxlan_can_change_config(sc)) {
2003 vxlan_sockaddr_in_copy(&sc->vxl_src_addr, &vxlsa->sa);
2004 vxlan_set_hwcaps(sc);
2005 error = 0;
2006 } else
2007 error = EBUSY;
2008 VXLAN_WUNLOCK(sc);
2009
2010 return (error);
2011 }
2012
2013 static int
vxlan_ctrl_set_remote_addr(struct vxlan_softc * sc,void * arg)2014 vxlan_ctrl_set_remote_addr(struct vxlan_softc *sc, void *arg)
2015 {
2016 struct ifvxlancmd *cmd;
2017 union vxlan_sockaddr *vxlsa;
2018 int error;
2019
2020 cmd = arg;
2021 vxlsa = &cmd->vxlcmd_sa;
2022
2023 if (!VXLAN_SOCKADDR_IS_IPV46(vxlsa))
2024 return (EINVAL);
2025 if (VXLAN_SOCKADDR_IS_IPV6(vxlsa)) {
2026 error = vxlan_sockaddr_in6_embedscope(vxlsa);
2027 if (error)
2028 return (error);
2029 }
2030
2031 VXLAN_WLOCK(sc);
2032 if (vxlan_can_change_config(sc)) {
2033 vxlan_sockaddr_in_copy(&sc->vxl_dst_addr, &vxlsa->sa);
2034 vxlan_setup_interface_hdrlen(sc);
2035 error = 0;
2036 } else
2037 error = EBUSY;
2038 VXLAN_WUNLOCK(sc);
2039
2040 return (error);
2041 }
2042
2043 static int
vxlan_ctrl_set_local_port(struct vxlan_softc * sc,void * arg)2044 vxlan_ctrl_set_local_port(struct vxlan_softc *sc, void *arg)
2045 {
2046 struct ifvxlancmd *cmd;
2047 int error;
2048
2049 cmd = arg;
2050
2051 if (cmd->vxlcmd_port == 0)
2052 return (EINVAL);
2053
2054 VXLAN_WLOCK(sc);
2055 if (vxlan_can_change_config(sc)) {
2056 sc->vxl_src_addr.in4.sin_port = htons(cmd->vxlcmd_port);
2057 error = 0;
2058 } else
2059 error = EBUSY;
2060 VXLAN_WUNLOCK(sc);
2061
2062 return (error);
2063 }
2064
2065 static int
vxlan_ctrl_set_remote_port(struct vxlan_softc * sc,void * arg)2066 vxlan_ctrl_set_remote_port(struct vxlan_softc *sc, void *arg)
2067 {
2068 struct ifvxlancmd *cmd;
2069 int error;
2070
2071 cmd = arg;
2072
2073 if (cmd->vxlcmd_port == 0)
2074 return (EINVAL);
2075
2076 VXLAN_WLOCK(sc);
2077 if (vxlan_can_change_config(sc)) {
2078 sc->vxl_dst_addr.in4.sin_port = htons(cmd->vxlcmd_port);
2079 error = 0;
2080 } else
2081 error = EBUSY;
2082 VXLAN_WUNLOCK(sc);
2083
2084 return (error);
2085 }
2086
2087 static int
vxlan_ctrl_set_port_range(struct vxlan_softc * sc,void * arg)2088 vxlan_ctrl_set_port_range(struct vxlan_softc *sc, void *arg)
2089 {
2090 struct ifvxlancmd *cmd;
2091 uint16_t min, max;
2092 int error;
2093
2094 cmd = arg;
2095 min = cmd->vxlcmd_port_min;
2096 max = cmd->vxlcmd_port_max;
2097
2098 if (max < min)
2099 return (EINVAL);
2100
2101 VXLAN_WLOCK(sc);
2102 if (vxlan_can_change_config(sc)) {
2103 sc->vxl_min_port = min;
2104 sc->vxl_max_port = max;
2105 error = 0;
2106 } else
2107 error = EBUSY;
2108 VXLAN_WUNLOCK(sc);
2109
2110 return (error);
2111 }
2112
2113 static int
vxlan_ctrl_set_ftable_timeout(struct vxlan_softc * sc,void * arg)2114 vxlan_ctrl_set_ftable_timeout(struct vxlan_softc *sc, void *arg)
2115 {
2116 struct ifvxlancmd *cmd;
2117 int error;
2118
2119 cmd = arg;
2120
2121 VXLAN_WLOCK(sc);
2122 if (vxlan_check_ftable_timeout(cmd->vxlcmd_ftable_timeout) == 0) {
2123 sc->vxl_ftable_timeout = cmd->vxlcmd_ftable_timeout;
2124 error = 0;
2125 } else
2126 error = EINVAL;
2127 VXLAN_WUNLOCK(sc);
2128
2129 return (error);
2130 }
2131
2132 static int
vxlan_ctrl_set_ftable_max(struct vxlan_softc * sc,void * arg)2133 vxlan_ctrl_set_ftable_max(struct vxlan_softc *sc, void *arg)
2134 {
2135 struct ifvxlancmd *cmd;
2136 int error;
2137
2138 cmd = arg;
2139
2140 VXLAN_WLOCK(sc);
2141 if (vxlan_check_ftable_max(cmd->vxlcmd_ftable_max) == 0) {
2142 sc->vxl_ftable_max = cmd->vxlcmd_ftable_max;
2143 error = 0;
2144 } else
2145 error = EINVAL;
2146 VXLAN_WUNLOCK(sc);
2147
2148 return (error);
2149 }
2150
2151 static int
vxlan_ctrl_set_multicast_if(struct vxlan_softc * sc,void * arg)2152 vxlan_ctrl_set_multicast_if(struct vxlan_softc * sc, void *arg)
2153 {
2154 struct ifvxlancmd *cmd;
2155 int error;
2156
2157 cmd = arg;
2158
2159 VXLAN_WLOCK(sc);
2160 if (vxlan_can_change_config(sc)) {
2161 strlcpy(sc->vxl_mc_ifname, cmd->vxlcmd_ifname, IFNAMSIZ);
2162 vxlan_set_hwcaps(sc);
2163 error = 0;
2164 } else
2165 error = EBUSY;
2166 VXLAN_WUNLOCK(sc);
2167
2168 return (error);
2169 }
2170
2171 static int
vxlan_ctrl_set_ttl(struct vxlan_softc * sc,void * arg)2172 vxlan_ctrl_set_ttl(struct vxlan_softc *sc, void *arg)
2173 {
2174 struct ifvxlancmd *cmd;
2175 int error;
2176
2177 cmd = arg;
2178
2179 VXLAN_WLOCK(sc);
2180 if (vxlan_check_ttl(cmd->vxlcmd_ttl) == 0) {
2181 sc->vxl_ttl = cmd->vxlcmd_ttl;
2182 if (sc->vxl_im4o != NULL)
2183 sc->vxl_im4o->imo_multicast_ttl = sc->vxl_ttl;
2184 if (sc->vxl_im6o != NULL)
2185 sc->vxl_im6o->im6o_multicast_hlim = sc->vxl_ttl;
2186 error = 0;
2187 } else
2188 error = EINVAL;
2189 VXLAN_WUNLOCK(sc);
2190
2191 return (error);
2192 }
2193
2194 static int
vxlan_ctrl_set_learn(struct vxlan_softc * sc,void * arg)2195 vxlan_ctrl_set_learn(struct vxlan_softc *sc, void *arg)
2196 {
2197 struct ifvxlancmd *cmd;
2198
2199 cmd = arg;
2200
2201 VXLAN_WLOCK(sc);
2202 if (cmd->vxlcmd_flags & VXLAN_CMD_FLAG_LEARN)
2203 sc->vxl_flags |= VXLAN_FLAG_LEARN;
2204 else
2205 sc->vxl_flags &= ~VXLAN_FLAG_LEARN;
2206 VXLAN_WUNLOCK(sc);
2207
2208 return (0);
2209 }
2210
2211 static int
vxlan_ctrl_ftable_entry_add(struct vxlan_softc * sc,void * arg)2212 vxlan_ctrl_ftable_entry_add(struct vxlan_softc *sc, void *arg)
2213 {
2214 union vxlan_sockaddr vxlsa;
2215 struct ifvxlancmd *cmd;
2216 struct vxlan_ftable_entry *fe;
2217 int error;
2218
2219 cmd = arg;
2220 vxlsa = cmd->vxlcmd_sa;
2221
2222 if (!VXLAN_SOCKADDR_IS_IPV46(&vxlsa))
2223 return (EINVAL);
2224 if (vxlan_sockaddr_in_any(&vxlsa) != 0)
2225 return (EINVAL);
2226 if (vxlan_sockaddr_in_multicast(&vxlsa) != 0)
2227 return (EINVAL);
2228 /* BMV: We could support both IPv4 and IPv6 later. */
2229 if (vxlsa.sa.sa_family != sc->vxl_dst_addr.sa.sa_family)
2230 return (EAFNOSUPPORT);
2231
2232 if (VXLAN_SOCKADDR_IS_IPV6(&vxlsa)) {
2233 error = vxlan_sockaddr_in6_embedscope(&vxlsa);
2234 if (error)
2235 return (error);
2236 }
2237
2238 fe = vxlan_ftable_entry_alloc();
2239 if (fe == NULL)
2240 return (ENOMEM);
2241
2242 if (vxlsa.in4.sin_port == 0)
2243 vxlsa.in4.sin_port = sc->vxl_dst_addr.in4.sin_port;
2244
2245 vxlan_ftable_entry_init(sc, fe, cmd->vxlcmd_mac, &vxlsa.sa,
2246 VXLAN_FE_FLAG_STATIC);
2247
2248 VXLAN_WLOCK(sc);
2249 error = vxlan_ftable_entry_insert(sc, fe);
2250 VXLAN_WUNLOCK(sc);
2251
2252 if (error)
2253 vxlan_ftable_entry_free(fe);
2254
2255 return (error);
2256 }
2257
2258 static int
vxlan_ctrl_ftable_entry_rem(struct vxlan_softc * sc,void * arg)2259 vxlan_ctrl_ftable_entry_rem(struct vxlan_softc *sc, void *arg)
2260 {
2261 struct ifvxlancmd *cmd;
2262 struct vxlan_ftable_entry *fe;
2263 int error;
2264
2265 cmd = arg;
2266
2267 VXLAN_WLOCK(sc);
2268 fe = vxlan_ftable_entry_lookup(sc, cmd->vxlcmd_mac);
2269 if (fe != NULL) {
2270 vxlan_ftable_entry_destroy(sc, fe);
2271 error = 0;
2272 } else
2273 error = ENOENT;
2274 VXLAN_WUNLOCK(sc);
2275
2276 return (error);
2277 }
2278
2279 static int
vxlan_ctrl_flush(struct vxlan_softc * sc,void * arg)2280 vxlan_ctrl_flush(struct vxlan_softc *sc, void *arg)
2281 {
2282 struct ifvxlancmd *cmd;
2283 int all;
2284
2285 cmd = arg;
2286 all = cmd->vxlcmd_flags & VXLAN_CMD_FLAG_FLUSH_ALL;
2287
2288 VXLAN_WLOCK(sc);
2289 vxlan_ftable_flush(sc, all);
2290 VXLAN_WUNLOCK(sc);
2291
2292 return (0);
2293 }
2294
2295 static int
vxlan_ioctl_drvspec(struct vxlan_softc * sc,struct ifdrv * ifd,int get)2296 vxlan_ioctl_drvspec(struct vxlan_softc *sc, struct ifdrv *ifd, int get)
2297 {
2298 const struct vxlan_control *vc;
2299 union {
2300 struct ifvxlancfg cfg;
2301 struct ifvxlancmd cmd;
2302 } args;
2303 int out, error;
2304
2305 if (ifd->ifd_cmd >= vxlan_control_table_size)
2306 return (EINVAL);
2307
2308 bzero(&args, sizeof(args));
2309 vc = &vxlan_control_table[ifd->ifd_cmd];
2310 out = (vc->vxlc_flags & VXLAN_CTRL_FLAG_COPYOUT) != 0;
2311
2312 if ((get != 0 && out == 0) || (get == 0 && out != 0))
2313 return (EINVAL);
2314
2315 if (vc->vxlc_flags & VXLAN_CTRL_FLAG_SUSER) {
2316 error = priv_check(curthread, PRIV_NET_VXLAN);
2317 if (error)
2318 return (error);
2319 }
2320
2321 if (ifd->ifd_len != vc->vxlc_argsize ||
2322 ifd->ifd_len > sizeof(args))
2323 return (EINVAL);
2324
2325 if (vc->vxlc_flags & VXLAN_CTRL_FLAG_COPYIN) {
2326 error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
2327 if (error)
2328 return (error);
2329 }
2330
2331 error = vc->vxlc_func(sc, &args);
2332 if (error)
2333 return (error);
2334
2335 if (vc->vxlc_flags & VXLAN_CTRL_FLAG_COPYOUT) {
2336 error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
2337 if (error)
2338 return (error);
2339 }
2340
2341 return (0);
2342 }
2343
2344 static int
vxlan_ioctl(struct ifnet * ifp,u_long cmd,caddr_t data)2345 vxlan_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
2346 {
2347 struct rm_priotracker tracker;
2348 struct vxlan_softc *sc;
2349 struct ifreq *ifr;
2350 struct ifdrv *ifd;
2351 int error;
2352
2353 sc = ifp->if_softc;
2354 ifr = (struct ifreq *) data;
2355 ifd = (struct ifdrv *) data;
2356
2357 error = 0;
2358
2359 switch (cmd) {
2360 case SIOCADDMULTI:
2361 case SIOCDELMULTI:
2362 break;
2363
2364 case SIOCGDRVSPEC:
2365 case SIOCSDRVSPEC:
2366 error = vxlan_ioctl_drvspec(sc, ifd, cmd == SIOCGDRVSPEC);
2367 break;
2368
2369 case SIOCSIFFLAGS:
2370 error = vxlan_ioctl_ifflags(sc);
2371 break;
2372
2373 case SIOCSIFMEDIA:
2374 case SIOCGIFMEDIA:
2375 error = ifmedia_ioctl(ifp, ifr, &sc->vxl_media, cmd);
2376 break;
2377
2378 case SIOCSIFMTU:
2379 if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > VXLAN_MAX_MTU) {
2380 error = EINVAL;
2381 } else {
2382 VXLAN_WLOCK(sc);
2383 ifp->if_mtu = ifr->ifr_mtu;
2384 sc->vxl_flags |= VXLAN_FLAG_USER_MTU;
2385 VXLAN_WUNLOCK(sc);
2386 }
2387 break;
2388
2389 case SIOCSIFCAP:
2390 VXLAN_WLOCK(sc);
2391 error = vxlan_set_reqcap(sc, ifp, ifr->ifr_reqcap);
2392 if (error == 0)
2393 vxlan_set_hwcaps(sc);
2394 VXLAN_WUNLOCK(sc);
2395 break;
2396
2397 case SIOCGTUNFIB:
2398 VXLAN_RLOCK(sc, &tracker);
2399 ifr->ifr_fib = sc->vxl_fibnum;
2400 VXLAN_RUNLOCK(sc, &tracker);
2401 break;
2402
2403 case SIOCSTUNFIB:
2404 if ((error = priv_check(curthread, PRIV_NET_VXLAN)) != 0)
2405 break;
2406
2407 if (ifr->ifr_fib >= rt_numfibs)
2408 error = EINVAL;
2409 else {
2410 VXLAN_WLOCK(sc);
2411 sc->vxl_fibnum = ifr->ifr_fib;
2412 VXLAN_WUNLOCK(sc);
2413 }
2414 break;
2415
2416 default:
2417 error = ether_ioctl(ifp, cmd, data);
2418 break;
2419 }
2420
2421 return (error);
2422 }
2423
2424 #if defined(INET) || defined(INET6)
2425 static uint16_t
vxlan_pick_source_port(struct vxlan_softc * sc,struct mbuf * m)2426 vxlan_pick_source_port(struct vxlan_softc *sc, struct mbuf *m)
2427 {
2428 int range;
2429 uint32_t hash;
2430
2431 range = sc->vxl_max_port - sc->vxl_min_port + 1;
2432
2433 if (M_HASHTYPE_ISHASH(m))
2434 hash = m->m_pkthdr.flowid;
2435 else
2436 hash = jenkins_hash(m->m_data, ETHER_HDR_LEN,
2437 sc->vxl_port_hash_key);
2438
2439 return (sc->vxl_min_port + (hash % range));
2440 }
2441
2442 static void
vxlan_encap_header(struct vxlan_softc * sc,struct mbuf * m,int ipoff,uint16_t srcport,uint16_t dstport)2443 vxlan_encap_header(struct vxlan_softc *sc, struct mbuf *m, int ipoff,
2444 uint16_t srcport, uint16_t dstport)
2445 {
2446 struct vxlanudphdr *hdr;
2447 struct udphdr *udph;
2448 struct vxlan_header *vxh;
2449 int len;
2450
2451 len = m->m_pkthdr.len - ipoff;
2452 MPASS(len >= sizeof(struct vxlanudphdr));
2453 hdr = mtodo(m, ipoff);
2454
2455 udph = &hdr->vxlh_udp;
2456 udph->uh_sport = srcport;
2457 udph->uh_dport = dstport;
2458 udph->uh_ulen = htons(len);
2459 udph->uh_sum = 0;
2460
2461 vxh = &hdr->vxlh_hdr;
2462 vxh->vxlh_flags = htonl(VXLAN_HDR_FLAGS_VALID_VNI);
2463 vxh->vxlh_vni = htonl(sc->vxl_vni << VXLAN_HDR_VNI_SHIFT);
2464 }
2465 #endif
2466
2467 #if defined(INET6) || defined(INET)
2468 /*
2469 * Return the CSUM_INNER_* equivalent of CSUM_* caps.
2470 */
2471 static uint32_t
csum_flags_to_inner_flags(uint32_t csum_flags_in,const uint32_t encap)2472 csum_flags_to_inner_flags(uint32_t csum_flags_in, const uint32_t encap)
2473 {
2474 uint32_t csum_flags = encap;
2475 const uint32_t v4 = CSUM_IP | CSUM_IP_UDP | CSUM_IP_TCP;
2476
2477 /*
2478 * csum_flags can request either v4 or v6 offload but not both.
2479 * tcp_output always sets CSUM_TSO (both CSUM_IP_TSO and CSUM_IP6_TSO)
2480 * so those bits are no good to detect the IP version. Other bits are
2481 * always set with CSUM_TSO and we use those to figure out the IP
2482 * version.
2483 */
2484 if (csum_flags_in & v4) {
2485 if (csum_flags_in & CSUM_IP)
2486 csum_flags |= CSUM_INNER_IP;
2487 if (csum_flags_in & CSUM_IP_UDP)
2488 csum_flags |= CSUM_INNER_IP_UDP;
2489 if (csum_flags_in & CSUM_IP_TCP)
2490 csum_flags |= CSUM_INNER_IP_TCP;
2491 if (csum_flags_in & CSUM_IP_TSO)
2492 csum_flags |= CSUM_INNER_IP_TSO;
2493 } else {
2494 #ifdef INVARIANTS
2495 const uint32_t v6 = CSUM_IP6_UDP | CSUM_IP6_TCP;
2496
2497 MPASS((csum_flags_in & v6) != 0);
2498 #endif
2499 if (csum_flags_in & CSUM_IP6_UDP)
2500 csum_flags |= CSUM_INNER_IP6_UDP;
2501 if (csum_flags_in & CSUM_IP6_TCP)
2502 csum_flags |= CSUM_INNER_IP6_TCP;
2503 if (csum_flags_in & CSUM_IP6_TSO)
2504 csum_flags |= CSUM_INNER_IP6_TSO;
2505 }
2506
2507 return (csum_flags);
2508 }
2509 #endif
2510
2511 static int
vxlan_encap4(struct vxlan_softc * sc,const union vxlan_sockaddr * fvxlsa,struct mbuf * m)2512 vxlan_encap4(struct vxlan_softc *sc, const union vxlan_sockaddr *fvxlsa,
2513 struct mbuf *m)
2514 {
2515 #ifdef INET
2516 struct ifnet *ifp;
2517 struct ip *ip;
2518 struct in_addr srcaddr, dstaddr;
2519 uint16_t srcport, dstport;
2520 int plen, mcast, error;
2521 struct route route, *ro;
2522 struct sockaddr_in *sin;
2523 uint32_t csum_flags;
2524
2525 NET_EPOCH_ASSERT();
2526
2527 ifp = sc->vxl_ifp;
2528 srcaddr = sc->vxl_src_addr.in4.sin_addr;
2529 srcport = vxlan_pick_source_port(sc, m);
2530 dstaddr = fvxlsa->in4.sin_addr;
2531 dstport = fvxlsa->in4.sin_port;
2532
2533 plen = m->m_pkthdr.len;
2534 M_PREPEND(m, sizeof(struct ip) + sizeof(struct vxlanudphdr),
2535 M_NOWAIT);
2536 if (m == NULL) {
2537 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2538 return (ENOBUFS);
2539 }
2540
2541 ip = mtod(m, struct ip *);
2542 ip->ip_tos = 0;
2543 ip->ip_len = htons(m->m_pkthdr.len);
2544 ip->ip_off = 0;
2545 ip->ip_ttl = sc->vxl_ttl;
2546 ip->ip_p = IPPROTO_UDP;
2547 ip->ip_sum = 0;
2548 ip->ip_src = srcaddr;
2549 ip->ip_dst = dstaddr;
2550
2551 vxlan_encap_header(sc, m, sizeof(struct ip), srcport, dstport);
2552
2553 mcast = (m->m_flags & (M_MCAST | M_BCAST)) ? 1 : 0;
2554 m->m_flags &= ~(M_MCAST | M_BCAST);
2555
2556 m->m_pkthdr.csum_flags &= CSUM_FLAGS_TX;
2557 if (m->m_pkthdr.csum_flags != 0) {
2558 /*
2559 * HW checksum (L3 and/or L4) or TSO has been requested. Look
2560 * up the ifnet for the outbound route and verify that the
2561 * outbound ifnet can perform the requested operation on the
2562 * inner frame.
2563 */
2564 bzero(&route, sizeof(route));
2565 ro = &route;
2566 sin = (struct sockaddr_in *)&ro->ro_dst;
2567 sin->sin_family = AF_INET;
2568 sin->sin_len = sizeof(*sin);
2569 sin->sin_addr = ip->ip_dst;
2570 ro->ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE,
2571 0);
2572 if (ro->ro_nh == NULL) {
2573 m_freem(m);
2574 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2575 return (EHOSTUNREACH);
2576 }
2577
2578 csum_flags = csum_flags_to_inner_flags(m->m_pkthdr.csum_flags,
2579 CSUM_ENCAP_VXLAN);
2580 if ((csum_flags & ro->ro_nh->nh_ifp->if_hwassist) !=
2581 csum_flags) {
2582 if (ppsratecheck(&sc->err_time, &sc->err_pps, 1)) {
2583 const struct ifnet *nh_ifp = ro->ro_nh->nh_ifp;
2584
2585 if_printf(ifp, "interface %s is missing hwcaps "
2586 "0x%08x, csum_flags 0x%08x -> 0x%08x, "
2587 "hwassist 0x%08x\n", nh_ifp->if_xname,
2588 csum_flags & ~(uint32_t)nh_ifp->if_hwassist,
2589 m->m_pkthdr.csum_flags, csum_flags,
2590 (uint32_t)nh_ifp->if_hwassist);
2591 }
2592 m_freem(m);
2593 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2594 return (ENXIO);
2595 }
2596 m->m_pkthdr.csum_flags = csum_flags;
2597 if (csum_flags &
2598 (CSUM_INNER_IP | CSUM_INNER_IP_UDP | CSUM_INNER_IP6_UDP |
2599 CSUM_INNER_IP_TCP | CSUM_INNER_IP6_TCP)) {
2600 counter_u64_add(sc->vxl_stats.txcsum, 1);
2601 if (csum_flags & CSUM_INNER_TSO)
2602 counter_u64_add(sc->vxl_stats.tso, 1);
2603 }
2604 } else
2605 ro = NULL;
2606 error = ip_output(m, NULL, ro, 0, sc->vxl_im4o, NULL);
2607 if (error == 0) {
2608 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
2609 if_inc_counter(ifp, IFCOUNTER_OBYTES, plen);
2610 if (mcast != 0)
2611 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
2612 } else
2613 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2614
2615 return (error);
2616 #else
2617 m_freem(m);
2618 return (ENOTSUP);
2619 #endif
2620 }
2621
2622 static int
vxlan_encap6(struct vxlan_softc * sc,const union vxlan_sockaddr * fvxlsa,struct mbuf * m)2623 vxlan_encap6(struct vxlan_softc *sc, const union vxlan_sockaddr *fvxlsa,
2624 struct mbuf *m)
2625 {
2626 #ifdef INET6
2627 struct ifnet *ifp;
2628 struct ip6_hdr *ip6;
2629 const struct in6_addr *srcaddr, *dstaddr;
2630 uint16_t srcport, dstport;
2631 int plen, mcast, error;
2632 struct route_in6 route, *ro;
2633 struct sockaddr_in6 *sin6;
2634 uint32_t csum_flags;
2635
2636 NET_EPOCH_ASSERT();
2637
2638 ifp = sc->vxl_ifp;
2639 srcaddr = &sc->vxl_src_addr.in6.sin6_addr;
2640 srcport = vxlan_pick_source_port(sc, m);
2641 dstaddr = &fvxlsa->in6.sin6_addr;
2642 dstport = fvxlsa->in6.sin6_port;
2643
2644 plen = m->m_pkthdr.len;
2645 M_PREPEND(m, sizeof(struct ip6_hdr) + sizeof(struct vxlanudphdr),
2646 M_NOWAIT);
2647 if (m == NULL) {
2648 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2649 return (ENOBUFS);
2650 }
2651
2652 ip6 = mtod(m, struct ip6_hdr *);
2653 ip6->ip6_flow = 0; /* BMV: Keep in forwarding entry? */
2654 ip6->ip6_vfc = IPV6_VERSION;
2655 ip6->ip6_plen = 0;
2656 ip6->ip6_nxt = IPPROTO_UDP;
2657 ip6->ip6_hlim = sc->vxl_ttl;
2658 ip6->ip6_src = *srcaddr;
2659 ip6->ip6_dst = *dstaddr;
2660
2661 vxlan_encap_header(sc, m, sizeof(struct ip6_hdr), srcport, dstport);
2662
2663 mcast = (m->m_flags & (M_MCAST | M_BCAST)) ? 1 : 0;
2664 m->m_flags &= ~(M_MCAST | M_BCAST);
2665
2666 ro = NULL;
2667 m->m_pkthdr.csum_flags &= CSUM_FLAGS_TX;
2668 if (m->m_pkthdr.csum_flags != 0) {
2669 /*
2670 * HW checksum (L3 and/or L4) or TSO has been requested. Look
2671 * up the ifnet for the outbound route and verify that the
2672 * outbound ifnet can perform the requested operation on the
2673 * inner frame.
2674 */
2675 bzero(&route, sizeof(route));
2676 ro = &route;
2677 sin6 = (struct sockaddr_in6 *)&ro->ro_dst;
2678 sin6->sin6_family = AF_INET6;
2679 sin6->sin6_len = sizeof(*sin6);
2680 sin6->sin6_addr = ip6->ip6_dst;
2681 ro->ro_nh = fib6_lookup(M_GETFIB(m), &ip6->ip6_dst, 0,
2682 NHR_NONE, 0);
2683 if (ro->ro_nh == NULL) {
2684 m_freem(m);
2685 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2686 return (EHOSTUNREACH);
2687 }
2688
2689 csum_flags = csum_flags_to_inner_flags(m->m_pkthdr.csum_flags,
2690 CSUM_ENCAP_VXLAN);
2691 if ((csum_flags & ro->ro_nh->nh_ifp->if_hwassist) !=
2692 csum_flags) {
2693 if (ppsratecheck(&sc->err_time, &sc->err_pps, 1)) {
2694 const struct ifnet *nh_ifp = ro->ro_nh->nh_ifp;
2695
2696 if_printf(ifp, "interface %s is missing hwcaps "
2697 "0x%08x, csum_flags 0x%08x -> 0x%08x, "
2698 "hwassist 0x%08x\n", nh_ifp->if_xname,
2699 csum_flags & ~(uint32_t)nh_ifp->if_hwassist,
2700 m->m_pkthdr.csum_flags, csum_flags,
2701 (uint32_t)nh_ifp->if_hwassist);
2702 }
2703 m_freem(m);
2704 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2705 return (ENXIO);
2706 }
2707 m->m_pkthdr.csum_flags = csum_flags;
2708 if (csum_flags &
2709 (CSUM_INNER_IP | CSUM_INNER_IP_UDP | CSUM_INNER_IP6_UDP |
2710 CSUM_INNER_IP_TCP | CSUM_INNER_IP6_TCP)) {
2711 counter_u64_add(sc->vxl_stats.txcsum, 1);
2712 if (csum_flags & CSUM_INNER_TSO)
2713 counter_u64_add(sc->vxl_stats.tso, 1);
2714 }
2715 } else if (ntohs(dstport) != V_zero_checksum_port) {
2716 struct udphdr *hdr = mtodo(m, sizeof(struct ip6_hdr));
2717
2718 hdr->uh_sum = in6_cksum_pseudo(ip6,
2719 m->m_pkthdr.len - sizeof(struct ip6_hdr), IPPROTO_UDP, 0);
2720 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
2721 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
2722 }
2723 error = ip6_output(m, NULL, ro, 0, sc->vxl_im6o, NULL, NULL);
2724 if (error == 0) {
2725 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
2726 if_inc_counter(ifp, IFCOUNTER_OBYTES, plen);
2727 if (mcast != 0)
2728 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
2729 } else
2730 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2731
2732 return (error);
2733 #else
2734 m_freem(m);
2735 return (ENOTSUP);
2736 #endif
2737 }
2738
2739 #define MTAG_VXLAN_LOOP 0x7876706c /* vxlp */
2740 static int
vxlan_transmit(struct ifnet * ifp,struct mbuf * m)2741 vxlan_transmit(struct ifnet *ifp, struct mbuf *m)
2742 {
2743 struct rm_priotracker tracker;
2744 union vxlan_sockaddr vxlsa;
2745 struct vxlan_softc *sc;
2746 struct vxlan_ftable_entry *fe;
2747 struct ifnet *mcifp;
2748 struct ether_header *eh;
2749 int ipv4, error;
2750
2751 sc = ifp->if_softc;
2752 eh = mtod(m, struct ether_header *);
2753 fe = NULL;
2754 mcifp = NULL;
2755
2756 ETHER_BPF_MTAP(ifp, m);
2757
2758 VXLAN_RLOCK(sc, &tracker);
2759 M_SETFIB(m, sc->vxl_fibnum);
2760 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2761 VXLAN_RUNLOCK(sc, &tracker);
2762 m_freem(m);
2763 return (ENETDOWN);
2764 }
2765 if (__predict_false(if_tunnel_check_nesting(ifp, m, MTAG_VXLAN_LOOP,
2766 max_vxlan_nesting) != 0)) {
2767 VXLAN_RUNLOCK(sc, &tracker);
2768 m_freem(m);
2769 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2770 return (ELOOP);
2771 }
2772
2773 if ((m->m_flags & (M_BCAST | M_MCAST)) == 0)
2774 fe = vxlan_ftable_entry_lookup(sc, eh->ether_dhost);
2775 if (fe == NULL)
2776 fe = &sc->vxl_default_fe;
2777 vxlan_sockaddr_copy(&vxlsa, &fe->vxlfe_raddr.sa);
2778
2779 ipv4 = VXLAN_SOCKADDR_IS_IPV4(&vxlsa) != 0;
2780 if (vxlan_sockaddr_in_multicast(&vxlsa) != 0)
2781 mcifp = vxlan_multicast_if_ref(sc, ipv4);
2782
2783 VXLAN_ACQUIRE(sc);
2784 VXLAN_RUNLOCK(sc, &tracker);
2785
2786 if (ipv4 != 0)
2787 error = vxlan_encap4(sc, &vxlsa, m);
2788 else
2789 error = vxlan_encap6(sc, &vxlsa, m);
2790
2791 vxlan_release(sc);
2792 if (mcifp != NULL)
2793 if_rele(mcifp);
2794
2795 return (error);
2796 }
2797
2798 static void
vxlan_qflush(struct ifnet * ifp __unused)2799 vxlan_qflush(struct ifnet *ifp __unused)
2800 {
2801 }
2802
2803 static bool
vxlan_rcv_udp_packet(struct mbuf * m,int offset,struct inpcb * inpcb,const struct sockaddr * srcsa,void * xvso)2804 vxlan_rcv_udp_packet(struct mbuf *m, int offset, struct inpcb *inpcb,
2805 const struct sockaddr *srcsa, void *xvso)
2806 {
2807 struct vxlan_socket *vso;
2808 struct vxlan_header *vxh, vxlanhdr;
2809 uint32_t vni;
2810 int error __unused;
2811
2812 M_ASSERTPKTHDR(m);
2813 vso = xvso;
2814 offset += sizeof(struct udphdr);
2815
2816 if (m->m_pkthdr.len < offset + sizeof(struct vxlan_header))
2817 goto out;
2818
2819 if (__predict_false(m->m_len < offset + sizeof(struct vxlan_header))) {
2820 m_copydata(m, offset, sizeof(struct vxlan_header),
2821 (caddr_t) &vxlanhdr);
2822 vxh = &vxlanhdr;
2823 } else
2824 vxh = mtodo(m, offset);
2825
2826 /*
2827 * Drop if there is a reserved bit set in either the flags or VNI
2828 * fields of the header. This goes against the specification, but
2829 * a bit set may indicate an unsupported new feature. This matches
2830 * the behavior of the Linux implementation.
2831 */
2832 if (vxh->vxlh_flags != htonl(VXLAN_HDR_FLAGS_VALID_VNI) ||
2833 vxh->vxlh_vni & ~VXLAN_VNI_MASK)
2834 goto out;
2835
2836 vni = ntohl(vxh->vxlh_vni) >> VXLAN_HDR_VNI_SHIFT;
2837
2838 /* Adjust to the start of the inner Ethernet frame. */
2839 m_adj_decap(m, offset + sizeof(struct vxlan_header));
2840
2841 error = vxlan_input(vso, vni, &m, srcsa);
2842 MPASS(error != 0 || m == NULL);
2843
2844 out:
2845 if (m != NULL)
2846 m_freem(m);
2847
2848 return (true);
2849 }
2850
2851 static int
vxlan_input(struct vxlan_socket * vso,uint32_t vni,struct mbuf ** m0,const struct sockaddr * sa)2852 vxlan_input(struct vxlan_socket *vso, uint32_t vni, struct mbuf **m0,
2853 const struct sockaddr *sa)
2854 {
2855 struct vxlan_softc *sc;
2856 struct ifnet *ifp;
2857 struct mbuf *m;
2858 struct ether_header *eh;
2859 int error;
2860
2861 m = *m0;
2862
2863 if (m->m_pkthdr.len < ETHER_HDR_LEN)
2864 return (EINVAL);
2865
2866 sc = vxlan_socket_lookup_softc(vso, vni);
2867 if (sc == NULL)
2868 return (ENOENT);
2869
2870 ifp = sc->vxl_ifp;
2871 if (m->m_len < ETHER_HDR_LEN &&
2872 (m = m_pullup(m, ETHER_HDR_LEN)) == NULL) {
2873 *m0 = NULL;
2874 error = ENOBUFS;
2875 goto out;
2876 }
2877 eh = mtod(m, struct ether_header *);
2878
2879 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2880 error = ENETDOWN;
2881 goto out;
2882 } else if (ifp == m->m_pkthdr.rcvif) {
2883 /* XXX Does not catch more complex loops. */
2884 error = EDEADLK;
2885 goto out;
2886 }
2887
2888 if (sc->vxl_flags & VXLAN_FLAG_LEARN)
2889 vxlan_ftable_learn(sc, sa, eh->ether_shost);
2890
2891 m_clrprotoflags(m);
2892 m->m_pkthdr.rcvif = ifp;
2893 M_SETFIB(m, ifp->if_fib);
2894 if (((ifp->if_capenable & IFCAP_RXCSUM &&
2895 m->m_pkthdr.csum_flags & CSUM_INNER_L3_CALC) ||
2896 (ifp->if_capenable & IFCAP_RXCSUM_IPV6 &&
2897 !(m->m_pkthdr.csum_flags & CSUM_INNER_L3_CALC)))) {
2898 uint32_t csum_flags = 0;
2899
2900 if (m->m_pkthdr.csum_flags & CSUM_INNER_L3_CALC)
2901 csum_flags |= CSUM_L3_CALC;
2902 if (m->m_pkthdr.csum_flags & CSUM_INNER_L3_VALID)
2903 csum_flags |= CSUM_L3_VALID;
2904 if (m->m_pkthdr.csum_flags & CSUM_INNER_L4_CALC)
2905 csum_flags |= CSUM_L4_CALC;
2906 if (m->m_pkthdr.csum_flags & CSUM_INNER_L4_VALID)
2907 csum_flags |= CSUM_L4_VALID;
2908 m->m_pkthdr.csum_flags = csum_flags;
2909 counter_u64_add(sc->vxl_stats.rxcsum, 1);
2910 } else {
2911 /* clear everything */
2912 m->m_pkthdr.csum_flags = 0;
2913 m->m_pkthdr.csum_data = 0;
2914 }
2915
2916 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2917 (*ifp->if_input)(ifp, m);
2918 *m0 = NULL;
2919 error = 0;
2920
2921 out:
2922 vxlan_release(sc);
2923 return (error);
2924 }
2925
2926 static int
vxlan_stats_alloc(struct vxlan_softc * sc)2927 vxlan_stats_alloc(struct vxlan_softc *sc)
2928 {
2929 struct vxlan_statistics *stats = &sc->vxl_stats;
2930
2931 stats->txcsum = counter_u64_alloc(M_WAITOK);
2932 if (stats->txcsum == NULL)
2933 goto failed;
2934
2935 stats->tso = counter_u64_alloc(M_WAITOK);
2936 if (stats->tso == NULL)
2937 goto failed;
2938
2939 stats->rxcsum = counter_u64_alloc(M_WAITOK);
2940 if (stats->rxcsum == NULL)
2941 goto failed;
2942
2943 return (0);
2944 failed:
2945 vxlan_stats_free(sc);
2946 return (ENOMEM);
2947 }
2948
2949 static void
vxlan_stats_free(struct vxlan_softc * sc)2950 vxlan_stats_free(struct vxlan_softc *sc)
2951 {
2952 struct vxlan_statistics *stats = &sc->vxl_stats;
2953
2954 if (stats->txcsum != NULL) {
2955 counter_u64_free(stats->txcsum);
2956 stats->txcsum = NULL;
2957 }
2958 if (stats->tso != NULL) {
2959 counter_u64_free(stats->tso);
2960 stats->tso = NULL;
2961 }
2962 if (stats->rxcsum != NULL) {
2963 counter_u64_free(stats->rxcsum);
2964 stats->rxcsum = NULL;
2965 }
2966 }
2967
2968 static void
vxlan_set_default_config(struct vxlan_softc * sc)2969 vxlan_set_default_config(struct vxlan_softc *sc)
2970 {
2971
2972 sc->vxl_flags |= VXLAN_FLAG_LEARN;
2973
2974 sc->vxl_vni = VXLAN_VNI_MAX;
2975 sc->vxl_ttl = IPDEFTTL;
2976
2977 if (!vxlan_tunable_int(sc, "legacy_port", vxlan_legacy_port)) {
2978 sc->vxl_src_addr.in4.sin_port = htons(VXLAN_PORT);
2979 sc->vxl_dst_addr.in4.sin_port = htons(VXLAN_PORT);
2980 } else {
2981 sc->vxl_src_addr.in4.sin_port = htons(VXLAN_LEGACY_PORT);
2982 sc->vxl_dst_addr.in4.sin_port = htons(VXLAN_LEGACY_PORT);
2983 }
2984
2985 sc->vxl_min_port = V_ipport_firstauto;
2986 sc->vxl_max_port = V_ipport_lastauto;
2987
2988 sc->vxl_ftable_max = VXLAN_FTABLE_MAX;
2989 sc->vxl_ftable_timeout = VXLAN_FTABLE_TIMEOUT;
2990 }
2991
2992 static int
vxlan_set_user_config(struct vxlan_softc * sc,struct ifvxlanparam * vxlp)2993 vxlan_set_user_config(struct vxlan_softc *sc, struct ifvxlanparam *vxlp)
2994 {
2995
2996 #ifndef INET
2997 if (vxlp->vxlp_with & (VXLAN_PARAM_WITH_LOCAL_ADDR4 |
2998 VXLAN_PARAM_WITH_REMOTE_ADDR4))
2999 return (EAFNOSUPPORT);
3000 #endif
3001
3002 #ifndef INET6
3003 if (vxlp->vxlp_with & (VXLAN_PARAM_WITH_LOCAL_ADDR6 |
3004 VXLAN_PARAM_WITH_REMOTE_ADDR6))
3005 return (EAFNOSUPPORT);
3006 #else
3007 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LOCAL_ADDR6) {
3008 int error = vxlan_sockaddr_in6_embedscope(&vxlp->vxlp_local_sa);
3009 if (error)
3010 return (error);
3011 }
3012 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_REMOTE_ADDR6) {
3013 int error = vxlan_sockaddr_in6_embedscope(
3014 &vxlp->vxlp_remote_sa);
3015 if (error)
3016 return (error);
3017 }
3018 #endif
3019
3020 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_VNI) {
3021 if (vxlan_check_vni(vxlp->vxlp_vni) == 0)
3022 sc->vxl_vni = vxlp->vxlp_vni;
3023 }
3024
3025 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LOCAL_ADDR4) {
3026 sc->vxl_src_addr.in4.sin_len = sizeof(struct sockaddr_in);
3027 sc->vxl_src_addr.in4.sin_family = AF_INET;
3028 sc->vxl_src_addr.in4.sin_addr =
3029 vxlp->vxlp_local_sa.in4.sin_addr;
3030 } else if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LOCAL_ADDR6) {
3031 sc->vxl_src_addr.in6.sin6_len = sizeof(struct sockaddr_in6);
3032 sc->vxl_src_addr.in6.sin6_family = AF_INET6;
3033 sc->vxl_src_addr.in6.sin6_addr =
3034 vxlp->vxlp_local_sa.in6.sin6_addr;
3035 }
3036
3037 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_REMOTE_ADDR4) {
3038 sc->vxl_dst_addr.in4.sin_len = sizeof(struct sockaddr_in);
3039 sc->vxl_dst_addr.in4.sin_family = AF_INET;
3040 sc->vxl_dst_addr.in4.sin_addr =
3041 vxlp->vxlp_remote_sa.in4.sin_addr;
3042 } else if (vxlp->vxlp_with & VXLAN_PARAM_WITH_REMOTE_ADDR6) {
3043 sc->vxl_dst_addr.in6.sin6_len = sizeof(struct sockaddr_in6);
3044 sc->vxl_dst_addr.in6.sin6_family = AF_INET6;
3045 sc->vxl_dst_addr.in6.sin6_addr =
3046 vxlp->vxlp_remote_sa.in6.sin6_addr;
3047 }
3048
3049 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LOCAL_PORT)
3050 sc->vxl_src_addr.in4.sin_port = htons(vxlp->vxlp_local_port);
3051 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_REMOTE_PORT)
3052 sc->vxl_dst_addr.in4.sin_port = htons(vxlp->vxlp_remote_port);
3053
3054 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_PORT_RANGE) {
3055 if (vxlp->vxlp_min_port <= vxlp->vxlp_max_port) {
3056 sc->vxl_min_port = vxlp->vxlp_min_port;
3057 sc->vxl_max_port = vxlp->vxlp_max_port;
3058 }
3059 }
3060
3061 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_MULTICAST_IF)
3062 strlcpy(sc->vxl_mc_ifname, vxlp->vxlp_mc_ifname, IFNAMSIZ);
3063
3064 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_FTABLE_TIMEOUT) {
3065 if (vxlan_check_ftable_timeout(vxlp->vxlp_ftable_timeout) == 0)
3066 sc->vxl_ftable_timeout = vxlp->vxlp_ftable_timeout;
3067 }
3068
3069 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_FTABLE_MAX) {
3070 if (vxlan_check_ftable_max(vxlp->vxlp_ftable_max) == 0)
3071 sc->vxl_ftable_max = vxlp->vxlp_ftable_max;
3072 }
3073
3074 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_TTL) {
3075 if (vxlan_check_ttl(vxlp->vxlp_ttl) == 0)
3076 sc->vxl_ttl = vxlp->vxlp_ttl;
3077 }
3078
3079 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LEARN) {
3080 if (vxlp->vxlp_learn == 0)
3081 sc->vxl_flags &= ~VXLAN_FLAG_LEARN;
3082 }
3083
3084 return (0);
3085 }
3086
3087 static int
vxlan_set_reqcap(struct vxlan_softc * sc,struct ifnet * ifp,int reqcap)3088 vxlan_set_reqcap(struct vxlan_softc *sc, struct ifnet *ifp, int reqcap)
3089 {
3090 int mask = reqcap ^ ifp->if_capenable;
3091
3092 /* Disable TSO if tx checksums are disabled. */
3093 if (mask & IFCAP_TXCSUM && !(reqcap & IFCAP_TXCSUM) &&
3094 reqcap & IFCAP_TSO4) {
3095 reqcap &= ~IFCAP_TSO4;
3096 if_printf(ifp, "tso4 disabled due to -txcsum.\n");
3097 }
3098 if (mask & IFCAP_TXCSUM_IPV6 && !(reqcap & IFCAP_TXCSUM_IPV6) &&
3099 reqcap & IFCAP_TSO6) {
3100 reqcap &= ~IFCAP_TSO6;
3101 if_printf(ifp, "tso6 disabled due to -txcsum6.\n");
3102 }
3103
3104 /* Do not enable TSO if tx checksums are disabled. */
3105 if (mask & IFCAP_TSO4 && reqcap & IFCAP_TSO4 &&
3106 !(reqcap & IFCAP_TXCSUM)) {
3107 if_printf(ifp, "enable txcsum first.\n");
3108 return (EAGAIN);
3109 }
3110 if (mask & IFCAP_TSO6 && reqcap & IFCAP_TSO6 &&
3111 !(reqcap & IFCAP_TXCSUM_IPV6)) {
3112 if_printf(ifp, "enable txcsum6 first.\n");
3113 return (EAGAIN);
3114 }
3115
3116 sc->vxl_reqcap = reqcap;
3117 return (0);
3118 }
3119
3120 /*
3121 * A VXLAN interface inherits the capabilities of the vxlandev or the interface
3122 * hosting the vxlanlocal address.
3123 */
3124 static void
vxlan_set_hwcaps(struct vxlan_softc * sc)3125 vxlan_set_hwcaps(struct vxlan_softc *sc)
3126 {
3127 struct epoch_tracker et;
3128 struct ifnet *p;
3129 struct ifaddr *ifa;
3130 u_long hwa;
3131 int cap, ena;
3132 bool rel;
3133 struct ifnet *ifp = sc->vxl_ifp;
3134
3135 /* reset caps */
3136 ifp->if_capabilities &= VXLAN_BASIC_IFCAPS;
3137 ifp->if_capenable &= VXLAN_BASIC_IFCAPS;
3138 ifp->if_hwassist = 0;
3139
3140 NET_EPOCH_ENTER(et);
3141 CURVNET_SET(ifp->if_vnet);
3142
3143 rel = false;
3144 p = NULL;
3145 if (sc->vxl_mc_ifname[0] != '\0') {
3146 rel = true;
3147 p = ifunit_ref(sc->vxl_mc_ifname);
3148 } else if (vxlan_sockaddr_in_any(&sc->vxl_src_addr) == 0) {
3149 if (sc->vxl_src_addr.sa.sa_family == AF_INET) {
3150 struct sockaddr_in in4 = sc->vxl_src_addr.in4;
3151
3152 in4.sin_port = 0;
3153 ifa = ifa_ifwithaddr((struct sockaddr *)&in4);
3154 if (ifa != NULL)
3155 p = ifa->ifa_ifp;
3156 } else if (sc->vxl_src_addr.sa.sa_family == AF_INET6) {
3157 struct sockaddr_in6 in6 = sc->vxl_src_addr.in6;
3158
3159 in6.sin6_port = 0;
3160 ifa = ifa_ifwithaddr((struct sockaddr *)&in6);
3161 if (ifa != NULL)
3162 p = ifa->ifa_ifp;
3163 }
3164 }
3165 if (p == NULL)
3166 goto done;
3167
3168 cap = ena = hwa = 0;
3169
3170 /* checksum offload */
3171 if (p->if_capabilities & IFCAP_VXLAN_HWCSUM)
3172 cap |= p->if_capabilities & (IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6);
3173 if (p->if_capenable & IFCAP_VXLAN_HWCSUM) {
3174 ena |= sc->vxl_reqcap & p->if_capenable &
3175 (IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6);
3176 if (ena & IFCAP_TXCSUM) {
3177 if (p->if_hwassist & CSUM_INNER_IP)
3178 hwa |= CSUM_IP;
3179 if (p->if_hwassist & CSUM_INNER_IP_UDP)
3180 hwa |= CSUM_IP_UDP;
3181 if (p->if_hwassist & CSUM_INNER_IP_TCP)
3182 hwa |= CSUM_IP_TCP;
3183 }
3184 if (ena & IFCAP_TXCSUM_IPV6) {
3185 if (p->if_hwassist & CSUM_INNER_IP6_UDP)
3186 hwa |= CSUM_IP6_UDP;
3187 if (p->if_hwassist & CSUM_INNER_IP6_TCP)
3188 hwa |= CSUM_IP6_TCP;
3189 }
3190 }
3191
3192 /* hardware TSO */
3193 if (p->if_capabilities & IFCAP_VXLAN_HWTSO) {
3194 cap |= p->if_capabilities & IFCAP_TSO;
3195 if (p->if_hw_tsomax > IP_MAXPACKET - ifp->if_hdrlen)
3196 ifp->if_hw_tsomax = IP_MAXPACKET - ifp->if_hdrlen;
3197 else
3198 ifp->if_hw_tsomax = p->if_hw_tsomax;
3199 /* XXX: tsomaxsegcount decrement is cxgbe specific */
3200 ifp->if_hw_tsomaxsegcount = p->if_hw_tsomaxsegcount - 1;
3201 ifp->if_hw_tsomaxsegsize = p->if_hw_tsomaxsegsize;
3202 }
3203 if (p->if_capenable & IFCAP_VXLAN_HWTSO) {
3204 ena |= sc->vxl_reqcap & p->if_capenable & IFCAP_TSO;
3205 if (ena & IFCAP_TSO) {
3206 if (p->if_hwassist & CSUM_INNER_IP_TSO)
3207 hwa |= CSUM_IP_TSO;
3208 if (p->if_hwassist & CSUM_INNER_IP6_TSO)
3209 hwa |= CSUM_IP6_TSO;
3210 }
3211 }
3212
3213 ifp->if_capabilities |= cap;
3214 ifp->if_capenable |= ena;
3215 ifp->if_hwassist |= hwa;
3216 if (rel)
3217 if_rele(p);
3218 done:
3219 CURVNET_RESTORE();
3220 NET_EPOCH_EXIT(et);
3221 }
3222
3223 static int
vxlan_clone_create(struct if_clone * ifc,char * name,size_t len,struct ifc_data * ifd,struct ifnet ** ifpp)3224 vxlan_clone_create(struct if_clone *ifc, char *name, size_t len,
3225 struct ifc_data *ifd, struct ifnet **ifpp)
3226 {
3227 struct vxlan_softc *sc;
3228 struct ifnet *ifp;
3229 struct ifvxlanparam vxlp;
3230 int error;
3231
3232 sc = malloc(sizeof(struct vxlan_softc), M_VXLAN, M_WAITOK | M_ZERO);
3233 sc->vxl_unit = ifd->unit;
3234 sc->vxl_fibnum = curthread->td_proc->p_fibnum;
3235 vxlan_set_default_config(sc);
3236 error = vxlan_stats_alloc(sc);
3237 if (error != 0)
3238 goto fail;
3239
3240 if (ifd->params != NULL) {
3241 error = ifc_copyin(ifd, &vxlp, sizeof(vxlp));
3242 if (error)
3243 goto fail;
3244
3245 error = vxlan_set_user_config(sc, &vxlp);
3246 if (error)
3247 goto fail;
3248 }
3249
3250 ifp = if_alloc(IFT_ETHER);
3251 if (ifp == NULL) {
3252 error = ENOSPC;
3253 goto fail;
3254 }
3255
3256 sc->vxl_ifp = ifp;
3257 rm_init(&sc->vxl_lock, "vxlanrm");
3258 callout_init_rw(&sc->vxl_callout, &sc->vxl_lock, 0);
3259 sc->vxl_port_hash_key = arc4random();
3260 vxlan_ftable_init(sc);
3261
3262 vxlan_sysctl_setup(sc);
3263
3264 ifp->if_softc = sc;
3265 if_initname(ifp, vxlan_name, ifd->unit);
3266 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
3267 ifp->if_init = vxlan_init;
3268 ifp->if_ioctl = vxlan_ioctl;
3269 ifp->if_transmit = vxlan_transmit;
3270 ifp->if_qflush = vxlan_qflush;
3271 ifp->if_capabilities = VXLAN_BASIC_IFCAPS;
3272 ifp->if_capenable = VXLAN_BASIC_IFCAPS;
3273 sc->vxl_reqcap = -1;
3274 vxlan_set_hwcaps(sc);
3275
3276 ifmedia_init(&sc->vxl_media, 0, vxlan_media_change, vxlan_media_status);
3277 ifmedia_add(&sc->vxl_media, IFM_ETHER | IFM_AUTO, 0, NULL);
3278 ifmedia_set(&sc->vxl_media, IFM_ETHER | IFM_AUTO);
3279
3280 ether_gen_addr(ifp, &sc->vxl_hwaddr);
3281 ether_ifattach(ifp, sc->vxl_hwaddr.octet);
3282
3283 ifp->if_baudrate = 0;
3284
3285 VXLAN_WLOCK(sc);
3286 vxlan_setup_interface_hdrlen(sc);
3287 VXLAN_WUNLOCK(sc);
3288 *ifpp = ifp;
3289
3290 return (0);
3291
3292 fail:
3293 free(sc, M_VXLAN);
3294 return (error);
3295 }
3296
3297 static int
vxlan_clone_destroy(struct if_clone * ifc,struct ifnet * ifp,uint32_t flags)3298 vxlan_clone_destroy(struct if_clone *ifc, struct ifnet *ifp, uint32_t flags)
3299 {
3300 struct vxlan_softc *sc;
3301
3302 sc = ifp->if_softc;
3303
3304 vxlan_teardown(sc);
3305
3306 vxlan_ftable_flush(sc, 1);
3307
3308 ether_ifdetach(ifp);
3309 if_free(ifp);
3310 ifmedia_removeall(&sc->vxl_media);
3311
3312 vxlan_ftable_fini(sc);
3313
3314 vxlan_sysctl_destroy(sc);
3315 rm_destroy(&sc->vxl_lock);
3316 vxlan_stats_free(sc);
3317 free(sc, M_VXLAN);
3318
3319 return (0);
3320 }
3321
3322 /* BMV: Taken from if_bridge. */
3323 static uint32_t
vxlan_mac_hash(struct vxlan_softc * sc,const uint8_t * addr)3324 vxlan_mac_hash(struct vxlan_softc *sc, const uint8_t *addr)
3325 {
3326 uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->vxl_ftable_hash_key;
3327
3328 b += addr[5] << 8;
3329 b += addr[4];
3330 a += addr[3] << 24;
3331 a += addr[2] << 16;
3332 a += addr[1] << 8;
3333 a += addr[0];
3334
3335 /*
3336 * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3337 * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3338 */
3339 #define mix(a, b, c) \
3340 do { \
3341 a -= b; a -= c; a ^= (c >> 13); \
3342 b -= c; b -= a; b ^= (a << 8); \
3343 c -= a; c -= b; c ^= (b >> 13); \
3344 a -= b; a -= c; a ^= (c >> 12); \
3345 b -= c; b -= a; b ^= (a << 16); \
3346 c -= a; c -= b; c ^= (b >> 5); \
3347 a -= b; a -= c; a ^= (c >> 3); \
3348 b -= c; b -= a; b ^= (a << 10); \
3349 c -= a; c -= b; c ^= (b >> 15); \
3350 } while (0)
3351
3352 mix(a, b, c);
3353
3354 #undef mix
3355
3356 return (c);
3357 }
3358
3359 static int
vxlan_media_change(struct ifnet * ifp)3360 vxlan_media_change(struct ifnet *ifp)
3361 {
3362
3363 /* Ignore. */
3364 return (0);
3365 }
3366
3367 static void
vxlan_media_status(struct ifnet * ifp,struct ifmediareq * ifmr)3368 vxlan_media_status(struct ifnet *ifp, struct ifmediareq *ifmr)
3369 {
3370
3371 ifmr->ifm_status = IFM_ACTIVE | IFM_AVALID;
3372 ifmr->ifm_active = IFM_ETHER | IFM_FDX;
3373 }
3374
3375 static int
vxlan_sockaddr_cmp(const union vxlan_sockaddr * vxladdr,const struct sockaddr * sa)3376 vxlan_sockaddr_cmp(const union vxlan_sockaddr *vxladdr,
3377 const struct sockaddr *sa)
3378 {
3379
3380 return (bcmp(&vxladdr->sa, sa, vxladdr->sa.sa_len));
3381 }
3382
3383 static void
vxlan_sockaddr_copy(union vxlan_sockaddr * vxladdr,const struct sockaddr * sa)3384 vxlan_sockaddr_copy(union vxlan_sockaddr *vxladdr,
3385 const struct sockaddr *sa)
3386 {
3387
3388 MPASS(sa->sa_family == AF_INET || sa->sa_family == AF_INET6);
3389 bzero(vxladdr, sizeof(*vxladdr));
3390
3391 if (sa->sa_family == AF_INET) {
3392 vxladdr->in4 = *satoconstsin(sa);
3393 vxladdr->in4.sin_len = sizeof(struct sockaddr_in);
3394 } else if (sa->sa_family == AF_INET6) {
3395 vxladdr->in6 = *satoconstsin6(sa);
3396 vxladdr->in6.sin6_len = sizeof(struct sockaddr_in6);
3397 }
3398 }
3399
3400 static int
vxlan_sockaddr_in_equal(const union vxlan_sockaddr * vxladdr,const struct sockaddr * sa)3401 vxlan_sockaddr_in_equal(const union vxlan_sockaddr *vxladdr,
3402 const struct sockaddr *sa)
3403 {
3404 int equal;
3405
3406 if (sa->sa_family == AF_INET) {
3407 const struct in_addr *in4 = &satoconstsin(sa)->sin_addr;
3408 equal = in4->s_addr == vxladdr->in4.sin_addr.s_addr;
3409 } else if (sa->sa_family == AF_INET6) {
3410 const struct in6_addr *in6 = &satoconstsin6(sa)->sin6_addr;
3411 equal = IN6_ARE_ADDR_EQUAL(in6, &vxladdr->in6.sin6_addr);
3412 } else
3413 equal = 0;
3414
3415 return (equal);
3416 }
3417
3418 static void
vxlan_sockaddr_in_copy(union vxlan_sockaddr * vxladdr,const struct sockaddr * sa)3419 vxlan_sockaddr_in_copy(union vxlan_sockaddr *vxladdr,
3420 const struct sockaddr *sa)
3421 {
3422
3423 MPASS(sa->sa_family == AF_INET || sa->sa_family == AF_INET6);
3424
3425 if (sa->sa_family == AF_INET) {
3426 const struct in_addr *in4 = &satoconstsin(sa)->sin_addr;
3427 vxladdr->in4.sin_family = AF_INET;
3428 vxladdr->in4.sin_len = sizeof(struct sockaddr_in);
3429 vxladdr->in4.sin_addr = *in4;
3430 } else if (sa->sa_family == AF_INET6) {
3431 const struct in6_addr *in6 = &satoconstsin6(sa)->sin6_addr;
3432 vxladdr->in6.sin6_family = AF_INET6;
3433 vxladdr->in6.sin6_len = sizeof(struct sockaddr_in6);
3434 vxladdr->in6.sin6_addr = *in6;
3435 }
3436 }
3437
3438 static int
vxlan_sockaddr_supported(const union vxlan_sockaddr * vxladdr,int unspec)3439 vxlan_sockaddr_supported(const union vxlan_sockaddr *vxladdr, int unspec)
3440 {
3441 const struct sockaddr *sa;
3442 int supported;
3443
3444 sa = &vxladdr->sa;
3445 supported = 0;
3446
3447 if (sa->sa_family == AF_UNSPEC && unspec != 0) {
3448 supported = 1;
3449 } else if (sa->sa_family == AF_INET) {
3450 #ifdef INET
3451 supported = 1;
3452 #endif
3453 } else if (sa->sa_family == AF_INET6) {
3454 #ifdef INET6
3455 supported = 1;
3456 #endif
3457 }
3458
3459 return (supported);
3460 }
3461
3462 static int
vxlan_sockaddr_in_any(const union vxlan_sockaddr * vxladdr)3463 vxlan_sockaddr_in_any(const union vxlan_sockaddr *vxladdr)
3464 {
3465 const struct sockaddr *sa;
3466 int any;
3467
3468 sa = &vxladdr->sa;
3469
3470 if (sa->sa_family == AF_INET) {
3471 const struct in_addr *in4 = &satoconstsin(sa)->sin_addr;
3472 any = in4->s_addr == INADDR_ANY;
3473 } else if (sa->sa_family == AF_INET6) {
3474 const struct in6_addr *in6 = &satoconstsin6(sa)->sin6_addr;
3475 any = IN6_IS_ADDR_UNSPECIFIED(in6);
3476 } else
3477 any = -1;
3478
3479 return (any);
3480 }
3481
3482 static int
vxlan_sockaddr_in_multicast(const union vxlan_sockaddr * vxladdr)3483 vxlan_sockaddr_in_multicast(const union vxlan_sockaddr *vxladdr)
3484 {
3485 const struct sockaddr *sa;
3486 int mc;
3487
3488 sa = &vxladdr->sa;
3489
3490 if (sa->sa_family == AF_INET) {
3491 const struct in_addr *in4 = &satoconstsin(sa)->sin_addr;
3492 mc = IN_MULTICAST(ntohl(in4->s_addr));
3493 } else if (sa->sa_family == AF_INET6) {
3494 const struct in6_addr *in6 = &satoconstsin6(sa)->sin6_addr;
3495 mc = IN6_IS_ADDR_MULTICAST(in6);
3496 } else
3497 mc = -1;
3498
3499 return (mc);
3500 }
3501
3502 static int
vxlan_sockaddr_in6_embedscope(union vxlan_sockaddr * vxladdr)3503 vxlan_sockaddr_in6_embedscope(union vxlan_sockaddr *vxladdr)
3504 {
3505 int error;
3506
3507 MPASS(VXLAN_SOCKADDR_IS_IPV6(vxladdr));
3508 #ifdef INET6
3509 error = sa6_embedscope(&vxladdr->in6, V_ip6_use_defzone);
3510 #else
3511 error = EAFNOSUPPORT;
3512 #endif
3513
3514 return (error);
3515 }
3516
3517 static int
vxlan_can_change_config(struct vxlan_softc * sc)3518 vxlan_can_change_config(struct vxlan_softc *sc)
3519 {
3520 struct ifnet *ifp;
3521
3522 ifp = sc->vxl_ifp;
3523 VXLAN_LOCK_ASSERT(sc);
3524
3525 if (ifp->if_drv_flags & IFF_DRV_RUNNING)
3526 return (0);
3527 if (sc->vxl_flags & (VXLAN_FLAG_INIT | VXLAN_FLAG_TEARDOWN))
3528 return (0);
3529
3530 return (1);
3531 }
3532
3533 static int
vxlan_check_vni(uint32_t vni)3534 vxlan_check_vni(uint32_t vni)
3535 {
3536
3537 return (vni >= VXLAN_VNI_MAX);
3538 }
3539
3540 static int
vxlan_check_ttl(int ttl)3541 vxlan_check_ttl(int ttl)
3542 {
3543
3544 return (ttl > MAXTTL);
3545 }
3546
3547 static int
vxlan_check_ftable_timeout(uint32_t timeout)3548 vxlan_check_ftable_timeout(uint32_t timeout)
3549 {
3550
3551 return (timeout > VXLAN_FTABLE_MAX_TIMEOUT);
3552 }
3553
3554 static int
vxlan_check_ftable_max(uint32_t max)3555 vxlan_check_ftable_max(uint32_t max)
3556 {
3557
3558 return (max > VXLAN_FTABLE_MAX);
3559 }
3560
3561 static void
vxlan_sysctl_setup(struct vxlan_softc * sc)3562 vxlan_sysctl_setup(struct vxlan_softc *sc)
3563 {
3564 struct sysctl_ctx_list *ctx;
3565 struct sysctl_oid *node;
3566 struct vxlan_statistics *stats;
3567 char namebuf[8];
3568
3569 ctx = &sc->vxl_sysctl_ctx;
3570 stats = &sc->vxl_stats;
3571 snprintf(namebuf, sizeof(namebuf), "%d", sc->vxl_unit);
3572
3573 sysctl_ctx_init(ctx);
3574 sc->vxl_sysctl_node = SYSCTL_ADD_NODE(ctx,
3575 SYSCTL_STATIC_CHILDREN(_net_link_vxlan), OID_AUTO, namebuf,
3576 CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "");
3577
3578 node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(sc->vxl_sysctl_node),
3579 OID_AUTO, "ftable", CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "");
3580 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "count",
3581 CTLFLAG_RD, &sc->vxl_ftable_cnt, 0,
3582 "Number of entries in forwarding table");
3583 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "max",
3584 CTLFLAG_RD, &sc->vxl_ftable_max, 0,
3585 "Maximum number of entries allowed in forwarding table");
3586 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "timeout",
3587 CTLFLAG_RD, &sc->vxl_ftable_timeout, 0,
3588 "Number of seconds between prunes of the forwarding table");
3589 SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "dump",
3590 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE | CTLFLAG_SKIP,
3591 sc, 0, vxlan_ftable_sysctl_dump, "A",
3592 "Dump the forwarding table entries");
3593
3594 node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(sc->vxl_sysctl_node),
3595 OID_AUTO, "stats", CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "");
3596 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO,
3597 "ftable_nospace", CTLFLAG_RD, &stats->ftable_nospace, 0,
3598 "Fowarding table reached maximum entries");
3599 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO,
3600 "ftable_lock_upgrade_failed", CTLFLAG_RD,
3601 &stats->ftable_lock_upgrade_failed, 0,
3602 "Forwarding table update required lock upgrade");
3603
3604 SYSCTL_ADD_COUNTER_U64(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "txcsum",
3605 CTLFLAG_RD, &stats->txcsum,
3606 "# of times hardware assisted with tx checksum");
3607 SYSCTL_ADD_COUNTER_U64(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "tso",
3608 CTLFLAG_RD, &stats->tso, "# of times hardware assisted with TSO");
3609 SYSCTL_ADD_COUNTER_U64(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "rxcsum",
3610 CTLFLAG_RD, &stats->rxcsum,
3611 "# of times hardware assisted with rx checksum");
3612 }
3613
3614 static void
vxlan_sysctl_destroy(struct vxlan_softc * sc)3615 vxlan_sysctl_destroy(struct vxlan_softc *sc)
3616 {
3617
3618 sysctl_ctx_free(&sc->vxl_sysctl_ctx);
3619 sc->vxl_sysctl_node = NULL;
3620 }
3621
3622 static int
vxlan_tunable_int(struct vxlan_softc * sc,const char * knob,int def)3623 vxlan_tunable_int(struct vxlan_softc *sc, const char *knob, int def)
3624 {
3625 char path[64];
3626
3627 snprintf(path, sizeof(path), "net.link.vxlan.%d.%s",
3628 sc->vxl_unit, knob);
3629 TUNABLE_INT_FETCH(path, &def);
3630
3631 return (def);
3632 }
3633
3634 static void
vxlan_ifdetach_event(void * arg __unused,struct ifnet * ifp)3635 vxlan_ifdetach_event(void *arg __unused, struct ifnet *ifp)
3636 {
3637 struct vxlan_softc_head list;
3638 struct vxlan_socket *vso;
3639 struct vxlan_softc *sc, *tsc;
3640
3641 LIST_INIT(&list);
3642
3643 if (ifp->if_flags & IFF_RENAMING)
3644 return;
3645 if ((ifp->if_flags & IFF_MULTICAST) == 0)
3646 return;
3647
3648 VXLAN_LIST_LOCK();
3649 LIST_FOREACH(vso, &vxlan_socket_list, vxlso_entry)
3650 vxlan_socket_ifdetach(vso, ifp, &list);
3651 VXLAN_LIST_UNLOCK();
3652
3653 LIST_FOREACH_SAFE(sc, &list, vxl_ifdetach_list, tsc) {
3654 LIST_REMOVE(sc, vxl_ifdetach_list);
3655
3656 sx_xlock(&vxlan_sx);
3657 VXLAN_WLOCK(sc);
3658 if (sc->vxl_flags & VXLAN_FLAG_INIT)
3659 vxlan_init_wait(sc);
3660 vxlan_teardown_locked(sc);
3661 sx_xunlock(&vxlan_sx);
3662 }
3663 }
3664
3665 static void
vxlan_load(void)3666 vxlan_load(void)
3667 {
3668
3669 mtx_init(&vxlan_list_mtx, "vxlan list", NULL, MTX_DEF);
3670 LIST_INIT(&vxlan_socket_list);
3671 vxlan_ifdetach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
3672 vxlan_ifdetach_event, NULL, EVENTHANDLER_PRI_ANY);
3673
3674 struct if_clone_addreq req = {
3675 .create_f = vxlan_clone_create,
3676 .destroy_f = vxlan_clone_destroy,
3677 .flags = IFC_F_AUTOUNIT,
3678 };
3679 vxlan_cloner = ifc_attach_cloner(vxlan_name, &req);
3680 }
3681
3682 static void
vxlan_unload(void)3683 vxlan_unload(void)
3684 {
3685
3686 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
3687 vxlan_ifdetach_event_tag);
3688 ifc_detach_cloner(vxlan_cloner);
3689 mtx_destroy(&vxlan_list_mtx);
3690 MPASS(LIST_EMPTY(&vxlan_socket_list));
3691 }
3692
3693 static int
vxlan_modevent(module_t mod,int type,void * unused)3694 vxlan_modevent(module_t mod, int type, void *unused)
3695 {
3696 int error;
3697
3698 error = 0;
3699
3700 switch (type) {
3701 case MOD_LOAD:
3702 vxlan_load();
3703 break;
3704 case MOD_UNLOAD:
3705 vxlan_unload();
3706 break;
3707 default:
3708 error = ENOTSUP;
3709 break;
3710 }
3711
3712 return (error);
3713 }
3714
3715 static moduledata_t vxlan_mod = {
3716 "if_vxlan",
3717 vxlan_modevent,
3718 0
3719 };
3720
3721 DECLARE_MODULE(if_vxlan, vxlan_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
3722 MODULE_VERSION(if_vxlan, 1);
3723