2 * Copyright (c) 2014, Bryan Venteicher <bryanv@FreeBSD.org>
4 * Copyright (c) 2020, Chelsio Communications.
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
9 * 1. Redistributions of source code must retain the above copyright
10 * notice unmodified, this list of conditions, and the following
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.
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.
29 #include "opt_inet6.h"
31 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD$");
34 #include <sys/param.h>
35 #include <sys/eventhandler.h>
36 #include <sys/kernel.h>
39 #include <sys/malloc.h>
41 #include <sys/module.h>
42 #include <sys/refcount.h>
43 #include <sys/rmlock.h>
46 #include <sys/queue.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/sockio.h>
51 #include <sys/sysctl.h>
52 #include <sys/systm.h>
55 #include <net/ethernet.h>
57 #include <net/if_var.h>
58 #include <net/if_clone.h>
59 #include <net/if_dl.h>
60 #include <net/if_media.h>
61 #include <net/if_types.h>
62 #include <net/if_vxlan.h>
63 #include <net/netisr.h>
64 #include <net/route.h>
65 #include <net/route/nhop.h>
67 #include <netinet/in.h>
68 #include <netinet/in_systm.h>
69 #include <netinet/in_var.h>
70 #include <netinet/in_pcb.h>
71 #include <netinet/ip.h>
72 #include <netinet/ip6.h>
73 #include <netinet/ip_var.h>
74 #include <netinet/udp.h>
75 #include <netinet/udp_var.h>
76 #include <netinet/in_fib.h>
77 #include <netinet6/in6_fib.h>
79 #include <netinet6/ip6_var.h>
80 #include <netinet6/scope6_var.h>
83 LIST_HEAD(vxlan_softc_head, vxlan_softc);
86 SX_SYSINIT(vxlan, &vxlan_sx, "VXLAN global start/stop lock");
88 struct vxlan_socket_mc_info {
89 union vxlan_sockaddr vxlsomc_saddr;
90 union vxlan_sockaddr vxlsomc_gaddr;
96 * The maximum MTU of encapsulated ethernet frame within IPv4/UDP packet.
98 #define VXLAN_MAX_MTU (IP_MAXPACKET - \
99 60 /* Maximum IPv4 header len */ - \
100 sizeof(struct udphdr) - \
101 sizeof(struct vxlan_header) - \
102 ETHER_HDR_LEN - ETHER_CRC_LEN - ETHER_VLAN_ENCAP_LEN)
103 #define VXLAN_BASIC_IFCAPS (IFCAP_LINKSTATE | IFCAP_JUMBO_MTU)
105 #define VXLAN_SO_MC_MAX_GROUPS 32
107 #define VXLAN_SO_VNI_HASH_SHIFT 6
108 #define VXLAN_SO_VNI_HASH_SIZE (1 << VXLAN_SO_VNI_HASH_SHIFT)
109 #define VXLAN_SO_VNI_HASH(_vni) ((_vni) % VXLAN_SO_VNI_HASH_SIZE)
111 struct vxlan_socket {
112 struct socket *vxlso_sock;
113 struct rmlock vxlso_lock;
115 union vxlan_sockaddr vxlso_laddr;
116 LIST_ENTRY(vxlan_socket) vxlso_entry;
117 struct vxlan_softc_head vxlso_vni_hash[VXLAN_SO_VNI_HASH_SIZE];
118 struct vxlan_socket_mc_info vxlso_mc[VXLAN_SO_MC_MAX_GROUPS];
121 #define VXLAN_SO_RLOCK(_vso, _p) rm_rlock(&(_vso)->vxlso_lock, (_p))
122 #define VXLAN_SO_RUNLOCK(_vso, _p) rm_runlock(&(_vso)->vxlso_lock, (_p))
123 #define VXLAN_SO_WLOCK(_vso) rm_wlock(&(_vso)->vxlso_lock)
124 #define VXLAN_SO_WUNLOCK(_vso) rm_wunlock(&(_vso)->vxlso_lock)
125 #define VXLAN_SO_LOCK_ASSERT(_vso) \
126 rm_assert(&(_vso)->vxlso_lock, RA_LOCKED)
127 #define VXLAN_SO_LOCK_WASSERT(_vso) \
128 rm_assert(&(_vso)->vxlso_lock, RA_WLOCKED)
130 #define VXLAN_SO_ACQUIRE(_vso) refcount_acquire(&(_vso)->vxlso_refcnt)
131 #define VXLAN_SO_RELEASE(_vso) refcount_release(&(_vso)->vxlso_refcnt)
133 struct vxlan_ftable_entry {
134 LIST_ENTRY(vxlan_ftable_entry) vxlfe_hash;
135 uint16_t vxlfe_flags;
136 uint8_t vxlfe_mac[ETHER_ADDR_LEN];
137 union vxlan_sockaddr vxlfe_raddr;
141 #define VXLAN_FE_FLAG_DYNAMIC 0x01
142 #define VXLAN_FE_FLAG_STATIC 0x02
144 #define VXLAN_FE_IS_DYNAMIC(_fe) \
145 ((_fe)->vxlfe_flags & VXLAN_FE_FLAG_DYNAMIC)
147 #define VXLAN_SC_FTABLE_SHIFT 9
148 #define VXLAN_SC_FTABLE_SIZE (1 << VXLAN_SC_FTABLE_SHIFT)
149 #define VXLAN_SC_FTABLE_MASK (VXLAN_SC_FTABLE_SIZE - 1)
150 #define VXLAN_SC_FTABLE_HASH(_sc, _mac) \
151 (vxlan_mac_hash(_sc, _mac) % VXLAN_SC_FTABLE_SIZE)
153 LIST_HEAD(vxlan_ftable_head, vxlan_ftable_entry);
155 struct vxlan_statistics {
156 uint32_t ftable_nospace;
157 uint32_t ftable_lock_upgrade_failed;
158 counter_u64_t txcsum;
160 counter_u64_t rxcsum;
164 struct ifnet *vxl_ifp;
166 struct vxlan_socket *vxl_sock;
168 union vxlan_sockaddr vxl_src_addr;
169 union vxlan_sockaddr vxl_dst_addr;
171 #define VXLAN_FLAG_INIT 0x0001
172 #define VXLAN_FLAG_TEARDOWN 0x0002
173 #define VXLAN_FLAG_LEARN 0x0004
174 #define VXLAN_FLAG_USER_MTU 0x0008
176 uint32_t vxl_port_hash_key;
177 uint16_t vxl_min_port;
178 uint16_t vxl_max_port;
181 /* Lookup table from MAC address to forwarding entry. */
182 uint32_t vxl_ftable_cnt;
183 uint32_t vxl_ftable_max;
184 uint32_t vxl_ftable_timeout;
185 uint32_t vxl_ftable_hash_key;
186 struct vxlan_ftable_head *vxl_ftable;
188 /* Derived from vxl_dst_addr. */
189 struct vxlan_ftable_entry vxl_default_fe;
191 struct ip_moptions *vxl_im4o;
192 struct ip6_moptions *vxl_im6o;
194 struct rmlock vxl_lock;
195 volatile u_int vxl_refcnt;
198 int vxl_vso_mc_index;
199 struct vxlan_statistics vxl_stats;
200 struct sysctl_oid *vxl_sysctl_node;
201 struct sysctl_ctx_list vxl_sysctl_ctx;
202 struct callout vxl_callout;
203 struct ether_addr vxl_hwaddr;
205 struct ifnet *vxl_mc_ifp;
206 struct ifmedia vxl_media;
207 char vxl_mc_ifname[IFNAMSIZ];
208 LIST_ENTRY(vxlan_softc) vxl_entry;
209 LIST_ENTRY(vxlan_softc) vxl_ifdetach_list;
211 /* For rate limiting errors on the tx fast path. */
212 struct timeval err_time;
216 #define VXLAN_RLOCK(_sc, _p) rm_rlock(&(_sc)->vxl_lock, (_p))
217 #define VXLAN_RUNLOCK(_sc, _p) rm_runlock(&(_sc)->vxl_lock, (_p))
218 #define VXLAN_WLOCK(_sc) rm_wlock(&(_sc)->vxl_lock)
219 #define VXLAN_WUNLOCK(_sc) rm_wunlock(&(_sc)->vxl_lock)
220 #define VXLAN_LOCK_WOWNED(_sc) rm_wowned(&(_sc)->vxl_lock)
221 #define VXLAN_LOCK_ASSERT(_sc) rm_assert(&(_sc)->vxl_lock, RA_LOCKED)
222 #define VXLAN_LOCK_WASSERT(_sc) rm_assert(&(_sc)->vxl_lock, RA_WLOCKED)
223 #define VXLAN_UNLOCK(_sc, _p) do { \
224 if (VXLAN_LOCK_WOWNED(_sc)) \
225 VXLAN_WUNLOCK(_sc); \
227 VXLAN_RUNLOCK(_sc, _p); \
230 #define VXLAN_ACQUIRE(_sc) refcount_acquire(&(_sc)->vxl_refcnt)
231 #define VXLAN_RELEASE(_sc) refcount_release(&(_sc)->vxl_refcnt)
233 #define satoconstsin(sa) ((const struct sockaddr_in *)(sa))
234 #define satoconstsin6(sa) ((const struct sockaddr_in6 *)(sa))
237 struct udphdr vxlh_udp;
238 struct vxlan_header vxlh_hdr;
241 static int vxlan_ftable_addr_cmp(const uint8_t *, const uint8_t *);
242 static void vxlan_ftable_init(struct vxlan_softc *);
243 static void vxlan_ftable_fini(struct vxlan_softc *);
244 static void vxlan_ftable_flush(struct vxlan_softc *, int);
245 static void vxlan_ftable_expire(struct vxlan_softc *);
246 static int vxlan_ftable_update_locked(struct vxlan_softc *,
247 const union vxlan_sockaddr *, const uint8_t *,
248 struct rm_priotracker *);
249 static int vxlan_ftable_learn(struct vxlan_softc *,
250 const struct sockaddr *, const uint8_t *);
251 static int vxlan_ftable_sysctl_dump(SYSCTL_HANDLER_ARGS);
253 static struct vxlan_ftable_entry *
254 vxlan_ftable_entry_alloc(void);
255 static void vxlan_ftable_entry_free(struct vxlan_ftable_entry *);
256 static void vxlan_ftable_entry_init(struct vxlan_softc *,
257 struct vxlan_ftable_entry *, const uint8_t *,
258 const struct sockaddr *, uint32_t);
259 static void vxlan_ftable_entry_destroy(struct vxlan_softc *,
260 struct vxlan_ftable_entry *);
261 static int vxlan_ftable_entry_insert(struct vxlan_softc *,
262 struct vxlan_ftable_entry *);
263 static struct vxlan_ftable_entry *
264 vxlan_ftable_entry_lookup(struct vxlan_softc *,
266 static void vxlan_ftable_entry_dump(struct vxlan_ftable_entry *,
269 static struct vxlan_socket *
270 vxlan_socket_alloc(const union vxlan_sockaddr *);
271 static void vxlan_socket_destroy(struct vxlan_socket *);
272 static void vxlan_socket_release(struct vxlan_socket *);
273 static struct vxlan_socket *
274 vxlan_socket_lookup(union vxlan_sockaddr *vxlsa);
275 static void vxlan_socket_insert(struct vxlan_socket *);
276 static int vxlan_socket_init(struct vxlan_socket *, struct ifnet *);
277 static int vxlan_socket_bind(struct vxlan_socket *, struct ifnet *);
278 static int vxlan_socket_create(struct ifnet *, int,
279 const union vxlan_sockaddr *, struct vxlan_socket **);
280 static void vxlan_socket_ifdetach(struct vxlan_socket *,
281 struct ifnet *, struct vxlan_softc_head *);
283 static struct vxlan_socket *
284 vxlan_socket_mc_lookup(const union vxlan_sockaddr *);
285 static int vxlan_sockaddr_mc_info_match(
286 const struct vxlan_socket_mc_info *,
287 const union vxlan_sockaddr *,
288 const union vxlan_sockaddr *, int);
289 static int vxlan_socket_mc_join_group(struct vxlan_socket *,
290 const union vxlan_sockaddr *, const union vxlan_sockaddr *,
291 int *, union vxlan_sockaddr *);
292 static int vxlan_socket_mc_leave_group(struct vxlan_socket *,
293 const union vxlan_sockaddr *,
294 const union vxlan_sockaddr *, int);
295 static int vxlan_socket_mc_add_group(struct vxlan_socket *,
296 const union vxlan_sockaddr *, const union vxlan_sockaddr *,
298 static void vxlan_socket_mc_release_group_by_idx(struct vxlan_socket *,
301 static struct vxlan_softc *
302 vxlan_socket_lookup_softc_locked(struct vxlan_socket *,
304 static struct vxlan_softc *
305 vxlan_socket_lookup_softc(struct vxlan_socket *, uint32_t);
306 static int vxlan_socket_insert_softc(struct vxlan_socket *,
307 struct vxlan_softc *);
308 static void vxlan_socket_remove_softc(struct vxlan_socket *,
309 struct vxlan_softc *);
311 static struct ifnet *
312 vxlan_multicast_if_ref(struct vxlan_softc *, int);
313 static void vxlan_free_multicast(struct vxlan_softc *);
314 static int vxlan_setup_multicast_interface(struct vxlan_softc *);
316 static int vxlan_setup_multicast(struct vxlan_softc *);
317 static int vxlan_setup_socket(struct vxlan_softc *);
319 static void vxlan_setup_zero_checksum_port(struct vxlan_softc *);
321 static void vxlan_setup_interface_hdrlen(struct vxlan_softc *);
322 static int vxlan_valid_init_config(struct vxlan_softc *);
323 static void vxlan_init_wait(struct vxlan_softc *);
324 static void vxlan_init_complete(struct vxlan_softc *);
325 static void vxlan_init(void *);
326 static void vxlan_release(struct vxlan_softc *);
327 static void vxlan_teardown_wait(struct vxlan_softc *);
328 static void vxlan_teardown_complete(struct vxlan_softc *);
329 static void vxlan_teardown_locked(struct vxlan_softc *);
330 static void vxlan_teardown(struct vxlan_softc *);
331 static void vxlan_ifdetach(struct vxlan_softc *, struct ifnet *,
332 struct vxlan_softc_head *);
333 static void vxlan_timer(void *);
335 static int vxlan_ctrl_get_config(struct vxlan_softc *, void *);
336 static int vxlan_ctrl_set_vni(struct vxlan_softc *, void *);
337 static int vxlan_ctrl_set_local_addr(struct vxlan_softc *, void *);
338 static int vxlan_ctrl_set_remote_addr(struct vxlan_softc *, void *);
339 static int vxlan_ctrl_set_local_port(struct vxlan_softc *, void *);
340 static int vxlan_ctrl_set_remote_port(struct vxlan_softc *, void *);
341 static int vxlan_ctrl_set_port_range(struct vxlan_softc *, void *);
342 static int vxlan_ctrl_set_ftable_timeout(struct vxlan_softc *, void *);
343 static int vxlan_ctrl_set_ftable_max(struct vxlan_softc *, void *);
344 static int vxlan_ctrl_set_multicast_if(struct vxlan_softc * , void *);
345 static int vxlan_ctrl_set_ttl(struct vxlan_softc *, void *);
346 static int vxlan_ctrl_set_learn(struct vxlan_softc *, void *);
347 static int vxlan_ctrl_ftable_entry_add(struct vxlan_softc *, void *);
348 static int vxlan_ctrl_ftable_entry_rem(struct vxlan_softc *, void *);
349 static int vxlan_ctrl_flush(struct vxlan_softc *, void *);
350 static int vxlan_ioctl_drvspec(struct vxlan_softc *,
351 struct ifdrv *, int);
352 static int vxlan_ioctl_ifflags(struct vxlan_softc *);
353 static int vxlan_ioctl(struct ifnet *, u_long, caddr_t);
355 #if defined(INET) || defined(INET6)
356 static uint16_t vxlan_pick_source_port(struct vxlan_softc *, struct mbuf *);
357 static void vxlan_encap_header(struct vxlan_softc *, struct mbuf *,
358 int, uint16_t, uint16_t);
360 static int vxlan_encap4(struct vxlan_softc *,
361 const union vxlan_sockaddr *, struct mbuf *);
362 static int vxlan_encap6(struct vxlan_softc *,
363 const union vxlan_sockaddr *, struct mbuf *);
364 static int vxlan_transmit(struct ifnet *, struct mbuf *);
365 static void vxlan_qflush(struct ifnet *);
366 static void vxlan_rcv_udp_packet(struct mbuf *, int, struct inpcb *,
367 const struct sockaddr *, void *);
368 static int vxlan_input(struct vxlan_socket *, uint32_t, struct mbuf **,
369 const struct sockaddr *);
371 static int vxlan_stats_alloc(struct vxlan_softc *);
372 static void vxlan_stats_free(struct vxlan_softc *);
373 static void vxlan_set_default_config(struct vxlan_softc *);
374 static int vxlan_set_user_config(struct vxlan_softc *,
375 struct ifvxlanparam *);
376 static int vxlan_set_reqcap(struct vxlan_softc *, struct ifnet *, int);
377 static void vxlan_set_hwcaps(struct vxlan_softc *);
378 static int vxlan_clone_create(struct if_clone *, int, caddr_t);
379 static void vxlan_clone_destroy(struct ifnet *);
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 *);
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 *);
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);
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);
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 *);
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;
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)
422 static LIST_HEAD(, vxlan_socket) vxlan_socket_list;
424 static eventhandler_tag vxlan_ifdetach_event_tag;
426 SYSCTL_DECL(_net_link);
427 SYSCTL_NODE(_net_link, OID_AUTO, vxlan, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
428 "Virtual eXtensible Local Area Network");
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);
435 /* Default maximum number of addresses in the forwarding table. */
436 #ifndef VXLAN_FTABLE_MAX
437 #define VXLAN_FTABLE_MAX 2000
440 /* Timeout (in seconds) of addresses learned in the forwarding table. */
441 #ifndef VXLAN_FTABLE_TIMEOUT
442 #define VXLAN_FTABLE_TIMEOUT (20 * 60)
446 * Maximum timeout (in seconds) of addresses learned in the forwarding
449 #ifndef VXLAN_FTABLE_MAX_TIMEOUT
450 #define VXLAN_FTABLE_MAX_TIMEOUT (60 * 60 * 24)
453 /* Number of seconds between pruning attempts of the forwarding table. */
454 #ifndef VXLAN_FTABLE_PRUNE
455 #define VXLAN_FTABLE_PRUNE (5 * 60)
458 static int vxlan_ftable_prune_period = VXLAN_FTABLE_PRUNE;
460 struct vxlan_control {
461 int (*vxlc_func)(struct vxlan_softc *, void *);
464 #define VXLAN_CTRL_FLAG_COPYIN 0x01
465 #define VXLAN_CTRL_FLAG_COPYOUT 0x02
466 #define VXLAN_CTRL_FLAG_SUSER 0x04
469 static const struct vxlan_control vxlan_control_table[] = {
470 [VXLAN_CMD_GET_CONFIG] =
471 { vxlan_ctrl_get_config, sizeof(struct ifvxlancfg),
472 VXLAN_CTRL_FLAG_COPYOUT
475 [VXLAN_CMD_SET_VNI] =
476 { vxlan_ctrl_set_vni, sizeof(struct ifvxlancmd),
477 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
480 [VXLAN_CMD_SET_LOCAL_ADDR] =
481 { vxlan_ctrl_set_local_addr, sizeof(struct ifvxlancmd),
482 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
485 [VXLAN_CMD_SET_REMOTE_ADDR] =
486 { vxlan_ctrl_set_remote_addr, sizeof(struct ifvxlancmd),
487 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
490 [VXLAN_CMD_SET_LOCAL_PORT] =
491 { vxlan_ctrl_set_local_port, sizeof(struct ifvxlancmd),
492 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
495 [VXLAN_CMD_SET_REMOTE_PORT] =
496 { vxlan_ctrl_set_remote_port, sizeof(struct ifvxlancmd),
497 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
500 [VXLAN_CMD_SET_PORT_RANGE] =
501 { vxlan_ctrl_set_port_range, sizeof(struct ifvxlancmd),
502 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
505 [VXLAN_CMD_SET_FTABLE_TIMEOUT] =
506 { vxlan_ctrl_set_ftable_timeout, sizeof(struct ifvxlancmd),
507 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
510 [VXLAN_CMD_SET_FTABLE_MAX] =
511 { vxlan_ctrl_set_ftable_max, sizeof(struct ifvxlancmd),
512 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
515 [VXLAN_CMD_SET_MULTICAST_IF] =
516 { vxlan_ctrl_set_multicast_if, sizeof(struct ifvxlancmd),
517 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
520 [VXLAN_CMD_SET_TTL] =
521 { vxlan_ctrl_set_ttl, sizeof(struct ifvxlancmd),
522 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
525 [VXLAN_CMD_SET_LEARN] =
526 { vxlan_ctrl_set_learn, sizeof(struct ifvxlancmd),
527 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
530 [VXLAN_CMD_FTABLE_ENTRY_ADD] =
531 { vxlan_ctrl_ftable_entry_add, sizeof(struct ifvxlancmd),
532 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
535 [VXLAN_CMD_FTABLE_ENTRY_REM] =
536 { vxlan_ctrl_ftable_entry_rem, sizeof(struct ifvxlancmd),
537 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
541 { vxlan_ctrl_flush, sizeof(struct ifvxlancmd),
542 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
546 static const int vxlan_control_table_size = nitems(vxlan_control_table);
549 vxlan_ftable_addr_cmp(const uint8_t *a, const uint8_t *b)
553 for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++)
554 d = ((int)a[i]) - ((int)b[i]);
560 vxlan_ftable_init(struct vxlan_softc *sc)
564 sc->vxl_ftable = malloc(sizeof(struct vxlan_ftable_head) *
565 VXLAN_SC_FTABLE_SIZE, M_VXLAN, M_ZERO | M_WAITOK);
567 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++)
568 LIST_INIT(&sc->vxl_ftable[i]);
569 sc->vxl_ftable_hash_key = arc4random();
573 vxlan_ftable_fini(struct vxlan_softc *sc)
577 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++) {
578 KASSERT(LIST_EMPTY(&sc->vxl_ftable[i]),
579 ("%s: vxlan %p ftable[%d] not empty", __func__, sc, i));
581 MPASS(sc->vxl_ftable_cnt == 0);
583 free(sc->vxl_ftable, M_VXLAN);
584 sc->vxl_ftable = NULL;
588 vxlan_ftable_flush(struct vxlan_softc *sc, int all)
590 struct vxlan_ftable_entry *fe, *tfe;
593 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++) {
594 LIST_FOREACH_SAFE(fe, &sc->vxl_ftable[i], vxlfe_hash, tfe) {
595 if (all || VXLAN_FE_IS_DYNAMIC(fe))
596 vxlan_ftable_entry_destroy(sc, fe);
602 vxlan_ftable_expire(struct vxlan_softc *sc)
604 struct vxlan_ftable_entry *fe, *tfe;
607 VXLAN_LOCK_WASSERT(sc);
609 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++) {
610 LIST_FOREACH_SAFE(fe, &sc->vxl_ftable[i], vxlfe_hash, tfe) {
611 if (VXLAN_FE_IS_DYNAMIC(fe) &&
612 time_uptime >= fe->vxlfe_expire)
613 vxlan_ftable_entry_destroy(sc, fe);
619 vxlan_ftable_update_locked(struct vxlan_softc *sc,
620 const union vxlan_sockaddr *vxlsa, const uint8_t *mac,
621 struct rm_priotracker *tracker)
623 struct vxlan_ftable_entry *fe;
626 VXLAN_LOCK_ASSERT(sc);
630 * A forwarding entry for this MAC address might already exist. If
631 * so, update it, otherwise create a new one. We may have to upgrade
632 * the lock if we have to change or create an entry.
634 fe = vxlan_ftable_entry_lookup(sc, mac);
636 fe->vxlfe_expire = time_uptime + sc->vxl_ftable_timeout;
638 if (!VXLAN_FE_IS_DYNAMIC(fe) ||
639 vxlan_sockaddr_in_equal(&fe->vxlfe_raddr, &vxlsa->sa))
641 if (!VXLAN_LOCK_WOWNED(sc)) {
642 VXLAN_RUNLOCK(sc, tracker);
644 sc->vxl_stats.ftable_lock_upgrade_failed++;
647 vxlan_sockaddr_in_copy(&fe->vxlfe_raddr, &vxlsa->sa);
651 if (!VXLAN_LOCK_WOWNED(sc)) {
652 VXLAN_RUNLOCK(sc, tracker);
654 sc->vxl_stats.ftable_lock_upgrade_failed++;
658 if (sc->vxl_ftable_cnt >= sc->vxl_ftable_max) {
659 sc->vxl_stats.ftable_nospace++;
663 fe = vxlan_ftable_entry_alloc();
667 vxlan_ftable_entry_init(sc, fe, mac, &vxlsa->sa, VXLAN_FE_FLAG_DYNAMIC);
669 /* The prior lookup failed, so the insert should not. */
670 error = vxlan_ftable_entry_insert(sc, fe);
677 vxlan_ftable_learn(struct vxlan_softc *sc, const struct sockaddr *sa,
680 struct rm_priotracker tracker;
681 union vxlan_sockaddr vxlsa;
685 * The source port may be randomly selected by the remote host, so
686 * use the port of the default destination address.
688 vxlan_sockaddr_copy(&vxlsa, sa);
689 vxlsa.in4.sin_port = sc->vxl_dst_addr.in4.sin_port;
691 if (VXLAN_SOCKADDR_IS_IPV6(&vxlsa)) {
692 error = vxlan_sockaddr_in6_embedscope(&vxlsa);
697 VXLAN_RLOCK(sc, &tracker);
698 error = vxlan_ftable_update_locked(sc, &vxlsa, mac, &tracker);
699 VXLAN_UNLOCK(sc, &tracker);
705 vxlan_ftable_sysctl_dump(SYSCTL_HANDLER_ARGS)
707 struct rm_priotracker tracker;
709 struct vxlan_softc *sc;
710 struct vxlan_ftable_entry *fe;
715 * This is mostly intended for debugging during development. It is
716 * not practical to dump an entire large table this way.
720 size = PAGE_SIZE; /* Calculate later. */
722 sbuf_new(&sb, NULL, size, SBUF_FIXEDLEN);
723 sbuf_putc(&sb, '\n');
725 VXLAN_RLOCK(sc, &tracker);
726 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++) {
727 LIST_FOREACH(fe, &sc->vxl_ftable[i], vxlfe_hash) {
728 if (sbuf_error(&sb) != 0)
730 vxlan_ftable_entry_dump(fe, &sb);
733 VXLAN_RUNLOCK(sc, &tracker);
735 if (sbuf_len(&sb) == 1)
739 error = sysctl_handle_string(oidp, sbuf_data(&sb), sbuf_len(&sb), req);
745 static struct vxlan_ftable_entry *
746 vxlan_ftable_entry_alloc(void)
748 struct vxlan_ftable_entry *fe;
750 fe = malloc(sizeof(*fe), M_VXLAN, M_ZERO | M_NOWAIT);
756 vxlan_ftable_entry_free(struct vxlan_ftable_entry *fe)
763 vxlan_ftable_entry_init(struct vxlan_softc *sc, struct vxlan_ftable_entry *fe,
764 const uint8_t *mac, const struct sockaddr *sa, uint32_t flags)
767 fe->vxlfe_flags = flags;
768 fe->vxlfe_expire = time_uptime + sc->vxl_ftable_timeout;
769 memcpy(fe->vxlfe_mac, mac, ETHER_ADDR_LEN);
770 vxlan_sockaddr_copy(&fe->vxlfe_raddr, sa);
774 vxlan_ftable_entry_destroy(struct vxlan_softc *sc,
775 struct vxlan_ftable_entry *fe)
778 sc->vxl_ftable_cnt--;
779 LIST_REMOVE(fe, vxlfe_hash);
780 vxlan_ftable_entry_free(fe);
784 vxlan_ftable_entry_insert(struct vxlan_softc *sc,
785 struct vxlan_ftable_entry *fe)
787 struct vxlan_ftable_entry *lfe;
791 VXLAN_LOCK_WASSERT(sc);
792 hash = VXLAN_SC_FTABLE_HASH(sc, fe->vxlfe_mac);
794 lfe = LIST_FIRST(&sc->vxl_ftable[hash]);
796 LIST_INSERT_HEAD(&sc->vxl_ftable[hash], fe, vxlfe_hash);
801 dir = vxlan_ftable_addr_cmp(fe->vxlfe_mac, lfe->vxlfe_mac);
805 LIST_INSERT_BEFORE(lfe, fe, vxlfe_hash);
807 } else if (LIST_NEXT(lfe, vxlfe_hash) == NULL) {
808 LIST_INSERT_AFTER(lfe, fe, vxlfe_hash);
811 lfe = LIST_NEXT(lfe, vxlfe_hash);
812 } while (lfe != NULL);
815 sc->vxl_ftable_cnt++;
820 static struct vxlan_ftable_entry *
821 vxlan_ftable_entry_lookup(struct vxlan_softc *sc, const uint8_t *mac)
823 struct vxlan_ftable_entry *fe;
827 VXLAN_LOCK_ASSERT(sc);
828 hash = VXLAN_SC_FTABLE_HASH(sc, mac);
830 LIST_FOREACH(fe, &sc->vxl_ftable[hash], vxlfe_hash) {
831 dir = vxlan_ftable_addr_cmp(mac, fe->vxlfe_mac);
842 vxlan_ftable_entry_dump(struct vxlan_ftable_entry *fe, struct sbuf *sb)
845 const union vxlan_sockaddr *sa;
847 int i, len, af, width;
849 sa = &fe->vxlfe_raddr;
850 af = sa->sa.sa_family;
853 sbuf_printf(sb, "%c 0x%02X ", VXLAN_FE_IS_DYNAMIC(fe) ? 'D' : 'S',
856 for (i = 0; i < ETHER_ADDR_LEN - 1; i++)
857 sbuf_printf(sb, "%02X:", fe->vxlfe_mac[i]);
858 sbuf_printf(sb, "%02X ", fe->vxlfe_mac[i]);
861 addr = &sa->in4.sin_addr;
862 width = INET_ADDRSTRLEN - 1;
864 addr = &sa->in6.sin6_addr;
865 width = INET6_ADDRSTRLEN - 1;
867 inet_ntop(af, addr, buf, sizeof(buf));
868 sbuf_printf(sb, "%*s ", width, buf);
870 sbuf_printf(sb, "%08jd", (intmax_t)fe->vxlfe_expire);
874 /* Truncate a partial line. */
875 if (sbuf_error(sb) != 0)
876 sbuf_setpos(sb, len);
879 static struct vxlan_socket *
880 vxlan_socket_alloc(const union vxlan_sockaddr *sa)
882 struct vxlan_socket *vso;
885 vso = malloc(sizeof(*vso), M_VXLAN, M_WAITOK | M_ZERO);
886 rm_init(&vso->vxlso_lock, "vxlansorm");
887 refcount_init(&vso->vxlso_refcnt, 0);
888 for (i = 0; i < VXLAN_SO_VNI_HASH_SIZE; i++)
889 LIST_INIT(&vso->vxlso_vni_hash[i]);
890 vso->vxlso_laddr = *sa;
896 vxlan_socket_destroy(struct vxlan_socket *vso)
901 struct vxlan_socket_mc_info *mc;
903 for (i = 0; i < VXLAN_SO_MC_MAX_GROUPS; i++) {
904 mc = &vso->vxlso_mc[i];
905 KASSERT(mc->vxlsomc_gaddr.sa.sa_family == AF_UNSPEC,
906 ("%s: socket %p mc[%d] still has address",
910 for (i = 0; i < VXLAN_SO_VNI_HASH_SIZE; i++) {
911 KASSERT(LIST_EMPTY(&vso->vxlso_vni_hash[i]),
912 ("%s: socket %p vni_hash[%d] not empty",
916 so = vso->vxlso_sock;
918 vso->vxlso_sock = NULL;
922 rm_destroy(&vso->vxlso_lock);
927 vxlan_socket_release(struct vxlan_socket *vso)
932 destroy = VXLAN_SO_RELEASE(vso);
934 LIST_REMOVE(vso, vxlso_entry);
938 vxlan_socket_destroy(vso);
941 static struct vxlan_socket *
942 vxlan_socket_lookup(union vxlan_sockaddr *vxlsa)
944 struct vxlan_socket *vso;
947 LIST_FOREACH(vso, &vxlan_socket_list, vxlso_entry) {
948 if (vxlan_sockaddr_cmp(&vso->vxlso_laddr, &vxlsa->sa) == 0) {
949 VXLAN_SO_ACQUIRE(vso);
959 vxlan_socket_insert(struct vxlan_socket *vso)
963 VXLAN_SO_ACQUIRE(vso);
964 LIST_INSERT_HEAD(&vxlan_socket_list, vso, vxlso_entry);
969 vxlan_socket_init(struct vxlan_socket *vso, struct ifnet *ifp)
976 error = socreate(vso->vxlso_laddr.sa.sa_family, &vso->vxlso_sock,
977 SOCK_DGRAM, IPPROTO_UDP, td->td_ucred, td);
979 if_printf(ifp, "cannot create socket: %d\n", error);
983 error = udp_set_kernel_tunneling(vso->vxlso_sock,
984 vxlan_rcv_udp_packet, NULL, vso);
986 if_printf(ifp, "cannot set tunneling function: %d\n", error);
990 if (vxlan_reuse_port != 0) {
994 bzero(&sopt, sizeof(sopt));
995 sopt.sopt_dir = SOPT_SET;
996 sopt.sopt_level = IPPROTO_IP;
997 sopt.sopt_name = SO_REUSEPORT;
998 sopt.sopt_val = &val;
999 sopt.sopt_valsize = sizeof(val);
1000 error = sosetopt(vso->vxlso_sock, &sopt);
1003 "cannot set REUSEADDR socket opt: %d\n", error);
1012 vxlan_socket_bind(struct vxlan_socket *vso, struct ifnet *ifp)
1014 union vxlan_sockaddr laddr;
1019 laddr = vso->vxlso_laddr;
1021 error = sobind(vso->vxlso_sock, &laddr.sa, td);
1023 if (error != EADDRINUSE)
1024 if_printf(ifp, "cannot bind socket: %d\n", error);
1032 vxlan_socket_create(struct ifnet *ifp, int multicast,
1033 const union vxlan_sockaddr *saddr, struct vxlan_socket **vsop)
1035 union vxlan_sockaddr laddr;
1036 struct vxlan_socket *vso;
1042 * If this socket will be multicast, then only the local port
1043 * must be specified when binding.
1045 if (multicast != 0) {
1046 if (VXLAN_SOCKADDR_IS_IPV4(&laddr))
1047 laddr.in4.sin_addr.s_addr = INADDR_ANY;
1050 laddr.in6.sin6_addr = in6addr_any;
1054 vso = vxlan_socket_alloc(&laddr);
1058 error = vxlan_socket_init(vso, ifp);
1062 error = vxlan_socket_bind(vso, ifp);
1067 * There is a small window between the bind completing and
1068 * inserting the socket, so that a concurrent create may fail.
1069 * Let's not worry about that for now.
1071 vxlan_socket_insert(vso);
1077 vxlan_socket_destroy(vso);
1083 vxlan_socket_ifdetach(struct vxlan_socket *vso, struct ifnet *ifp,
1084 struct vxlan_softc_head *list)
1086 struct rm_priotracker tracker;
1087 struct vxlan_softc *sc;
1090 VXLAN_SO_RLOCK(vso, &tracker);
1091 for (i = 0; i < VXLAN_SO_VNI_HASH_SIZE; i++) {
1092 LIST_FOREACH(sc, &vso->vxlso_vni_hash[i], vxl_entry)
1093 vxlan_ifdetach(sc, ifp, list);
1095 VXLAN_SO_RUNLOCK(vso, &tracker);
1098 static struct vxlan_socket *
1099 vxlan_socket_mc_lookup(const union vxlan_sockaddr *vxlsa)
1101 union vxlan_sockaddr laddr;
1102 struct vxlan_socket *vso;
1106 if (VXLAN_SOCKADDR_IS_IPV4(&laddr))
1107 laddr.in4.sin_addr.s_addr = INADDR_ANY;
1110 laddr.in6.sin6_addr = in6addr_any;
1113 vso = vxlan_socket_lookup(&laddr);
1119 vxlan_sockaddr_mc_info_match(const struct vxlan_socket_mc_info *mc,
1120 const union vxlan_sockaddr *group, const union vxlan_sockaddr *local,
1124 if (!vxlan_sockaddr_in_any(local) &&
1125 !vxlan_sockaddr_in_equal(&mc->vxlsomc_saddr, &local->sa))
1127 if (!vxlan_sockaddr_in_equal(&mc->vxlsomc_gaddr, &group->sa))
1129 if (ifidx != 0 && ifidx != mc->vxlsomc_ifidx)
1136 vxlan_socket_mc_join_group(struct vxlan_socket *vso,
1137 const union vxlan_sockaddr *group, const union vxlan_sockaddr *local,
1138 int *ifidx, union vxlan_sockaddr *source)
1140 struct sockopt sopt;
1145 if (VXLAN_SOCKADDR_IS_IPV4(group)) {
1146 struct ip_mreq mreq;
1148 mreq.imr_multiaddr = group->in4.sin_addr;
1149 mreq.imr_interface = local->in4.sin_addr;
1151 bzero(&sopt, sizeof(sopt));
1152 sopt.sopt_dir = SOPT_SET;
1153 sopt.sopt_level = IPPROTO_IP;
1154 sopt.sopt_name = IP_ADD_MEMBERSHIP;
1155 sopt.sopt_val = &mreq;
1156 sopt.sopt_valsize = sizeof(mreq);
1157 error = sosetopt(vso->vxlso_sock, &sopt);
1162 * BMV: Ideally, there would be a formal way for us to get
1163 * the local interface that was selected based on the
1164 * imr_interface address. We could then update *ifidx so
1165 * vxlan_sockaddr_mc_info_match() would return a match for
1166 * later creates that explicitly set the multicast interface.
1168 * If we really need to, we can of course look in the INP's
1170 * sotoinpcb(vso->vxlso_sock)->inp_moptions->
1171 * imo_head[]->imf_inm->inm_ifp
1172 * similarly to imo_match_group().
1174 source->in4.sin_addr = local->in4.sin_addr;
1176 } else if (VXLAN_SOCKADDR_IS_IPV6(group)) {
1177 struct ipv6_mreq mreq;
1179 mreq.ipv6mr_multiaddr = group->in6.sin6_addr;
1180 mreq.ipv6mr_interface = *ifidx;
1182 bzero(&sopt, sizeof(sopt));
1183 sopt.sopt_dir = SOPT_SET;
1184 sopt.sopt_level = IPPROTO_IPV6;
1185 sopt.sopt_name = IPV6_JOIN_GROUP;
1186 sopt.sopt_val = &mreq;
1187 sopt.sopt_valsize = sizeof(mreq);
1188 error = sosetopt(vso->vxlso_sock, &sopt);
1193 * BMV: As with IPv4, we would really like to know what
1194 * interface in6p_lookup_mcast_ifp() selected.
1197 error = EAFNOSUPPORT;
1203 vxlan_socket_mc_leave_group(struct vxlan_socket *vso,
1204 const union vxlan_sockaddr *group, const union vxlan_sockaddr *source,
1207 struct sockopt sopt;
1210 bzero(&sopt, sizeof(sopt));
1211 sopt.sopt_dir = SOPT_SET;
1213 if (VXLAN_SOCKADDR_IS_IPV4(group)) {
1214 struct ip_mreq mreq;
1216 mreq.imr_multiaddr = group->in4.sin_addr;
1217 mreq.imr_interface = source->in4.sin_addr;
1219 sopt.sopt_level = IPPROTO_IP;
1220 sopt.sopt_name = IP_DROP_MEMBERSHIP;
1221 sopt.sopt_val = &mreq;
1222 sopt.sopt_valsize = sizeof(mreq);
1223 error = sosetopt(vso->vxlso_sock, &sopt);
1225 } else if (VXLAN_SOCKADDR_IS_IPV6(group)) {
1226 struct ipv6_mreq mreq;
1228 mreq.ipv6mr_multiaddr = group->in6.sin6_addr;
1229 mreq.ipv6mr_interface = ifidx;
1231 sopt.sopt_level = IPPROTO_IPV6;
1232 sopt.sopt_name = IPV6_LEAVE_GROUP;
1233 sopt.sopt_val = &mreq;
1234 sopt.sopt_valsize = sizeof(mreq);
1235 error = sosetopt(vso->vxlso_sock, &sopt);
1238 error = EAFNOSUPPORT;
1244 vxlan_socket_mc_add_group(struct vxlan_socket *vso,
1245 const union vxlan_sockaddr *group, const union vxlan_sockaddr *local,
1246 int ifidx, int *idx)
1248 union vxlan_sockaddr source;
1249 struct vxlan_socket_mc_info *mc;
1250 int i, empty, error;
1253 * Within a socket, the same multicast group may be used by multiple
1254 * interfaces, each with a different network identifier. But a socket
1255 * may only join a multicast group once, so keep track of the users
1259 VXLAN_SO_WLOCK(vso);
1260 for (empty = 0, i = 0; i < VXLAN_SO_MC_MAX_GROUPS; i++) {
1261 mc = &vso->vxlso_mc[i];
1263 if (mc->vxlsomc_gaddr.sa.sa_family == AF_UNSPEC) {
1268 if (vxlan_sockaddr_mc_info_match(mc, group, local, ifidx))
1271 VXLAN_SO_WUNLOCK(vso);
1276 error = vxlan_socket_mc_join_group(vso, group, local, &ifidx, &source);
1280 VXLAN_SO_WLOCK(vso);
1281 for (i = 0; i < VXLAN_SO_MC_MAX_GROUPS; i++) {
1282 mc = &vso->vxlso_mc[i];
1284 if (mc->vxlsomc_gaddr.sa.sa_family == AF_UNSPEC) {
1285 vxlan_sockaddr_copy(&mc->vxlsomc_gaddr, &group->sa);
1286 vxlan_sockaddr_copy(&mc->vxlsomc_saddr, &source.sa);
1287 mc->vxlsomc_ifidx = ifidx;
1291 VXLAN_SO_WUNLOCK(vso);
1293 error = vxlan_socket_mc_leave_group(vso, group, &source, ifidx);
1299 mc->vxlsomc_users++;
1300 VXLAN_SO_WUNLOCK(vso);
1308 vxlan_socket_mc_release_group_by_idx(struct vxlan_socket *vso, int idx)
1310 union vxlan_sockaddr group, source;
1311 struct vxlan_socket_mc_info *mc;
1314 KASSERT(idx >= 0 && idx < VXLAN_SO_MC_MAX_GROUPS,
1315 ("%s: vso %p idx %d out of bounds", __func__, vso, idx));
1318 mc = &vso->vxlso_mc[idx];
1320 VXLAN_SO_WLOCK(vso);
1321 mc->vxlsomc_users--;
1322 if (mc->vxlsomc_users == 0) {
1323 group = mc->vxlsomc_gaddr;
1324 source = mc->vxlsomc_saddr;
1325 ifidx = mc->vxlsomc_ifidx;
1326 bzero(mc, sizeof(*mc));
1329 VXLAN_SO_WUNLOCK(vso);
1333 * Our socket's membership in this group may have already
1334 * been removed if we joined through an interface that's
1337 vxlan_socket_mc_leave_group(vso, &group, &source, ifidx);
1341 static struct vxlan_softc *
1342 vxlan_socket_lookup_softc_locked(struct vxlan_socket *vso, uint32_t vni)
1344 struct vxlan_softc *sc;
1347 VXLAN_SO_LOCK_ASSERT(vso);
1348 hash = VXLAN_SO_VNI_HASH(vni);
1350 LIST_FOREACH(sc, &vso->vxlso_vni_hash[hash], vxl_entry) {
1351 if (sc->vxl_vni == vni) {
1360 static struct vxlan_softc *
1361 vxlan_socket_lookup_softc(struct vxlan_socket *vso, uint32_t vni)
1363 struct rm_priotracker tracker;
1364 struct vxlan_softc *sc;
1366 VXLAN_SO_RLOCK(vso, &tracker);
1367 sc = vxlan_socket_lookup_softc_locked(vso, vni);
1368 VXLAN_SO_RUNLOCK(vso, &tracker);
1374 vxlan_socket_insert_softc(struct vxlan_socket *vso, struct vxlan_softc *sc)
1376 struct vxlan_softc *tsc;
1380 hash = VXLAN_SO_VNI_HASH(vni);
1382 VXLAN_SO_WLOCK(vso);
1383 tsc = vxlan_socket_lookup_softc_locked(vso, vni);
1385 VXLAN_SO_WUNLOCK(vso);
1391 LIST_INSERT_HEAD(&vso->vxlso_vni_hash[hash], sc, vxl_entry);
1392 VXLAN_SO_WUNLOCK(vso);
1398 vxlan_socket_remove_softc(struct vxlan_socket *vso, struct vxlan_softc *sc)
1401 VXLAN_SO_WLOCK(vso);
1402 LIST_REMOVE(sc, vxl_entry);
1403 VXLAN_SO_WUNLOCK(vso);
1408 static struct ifnet *
1409 vxlan_multicast_if_ref(struct vxlan_softc *sc, int ipv4)
1413 VXLAN_LOCK_ASSERT(sc);
1415 if (ipv4 && sc->vxl_im4o != NULL)
1416 ifp = sc->vxl_im4o->imo_multicast_ifp;
1417 else if (!ipv4 && sc->vxl_im6o != NULL)
1418 ifp = sc->vxl_im6o->im6o_multicast_ifp;
1429 vxlan_free_multicast(struct vxlan_softc *sc)
1432 if (sc->vxl_mc_ifp != NULL) {
1433 if_rele(sc->vxl_mc_ifp);
1434 sc->vxl_mc_ifp = NULL;
1435 sc->vxl_mc_ifindex = 0;
1438 if (sc->vxl_im4o != NULL) {
1439 free(sc->vxl_im4o, M_VXLAN);
1440 sc->vxl_im4o = NULL;
1443 if (sc->vxl_im6o != NULL) {
1444 free(sc->vxl_im6o, M_VXLAN);
1445 sc->vxl_im6o = NULL;
1450 vxlan_setup_multicast_interface(struct vxlan_softc *sc)
1454 ifp = ifunit_ref(sc->vxl_mc_ifname);
1456 if_printf(sc->vxl_ifp, "multicast interface %s does "
1457 "not exist\n", sc->vxl_mc_ifname);
1461 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
1462 if_printf(sc->vxl_ifp, "interface %s does not support "
1463 "multicast\n", sc->vxl_mc_ifname);
1468 sc->vxl_mc_ifp = ifp;
1469 sc->vxl_mc_ifindex = ifp->if_index;
1475 vxlan_setup_multicast(struct vxlan_softc *sc)
1477 const union vxlan_sockaddr *group;
1480 group = &sc->vxl_dst_addr;
1483 if (sc->vxl_mc_ifname[0] != '\0') {
1484 error = vxlan_setup_multicast_interface(sc);
1490 * Initialize an multicast options structure that is sufficiently
1491 * populated for use in the respective IP output routine. This
1492 * structure is typically stored in the socket, but our sockets
1493 * may be shared among multiple interfaces.
1495 if (VXLAN_SOCKADDR_IS_IPV4(group)) {
1496 sc->vxl_im4o = malloc(sizeof(struct ip_moptions), M_VXLAN,
1498 sc->vxl_im4o->imo_multicast_ifp = sc->vxl_mc_ifp;
1499 sc->vxl_im4o->imo_multicast_ttl = sc->vxl_ttl;
1500 sc->vxl_im4o->imo_multicast_vif = -1;
1501 } else if (VXLAN_SOCKADDR_IS_IPV6(group)) {
1502 sc->vxl_im6o = malloc(sizeof(struct ip6_moptions), M_VXLAN,
1504 sc->vxl_im6o->im6o_multicast_ifp = sc->vxl_mc_ifp;
1505 sc->vxl_im6o->im6o_multicast_hlim = sc->vxl_ttl;
1512 vxlan_setup_socket(struct vxlan_softc *sc)
1514 struct vxlan_socket *vso;
1516 union vxlan_sockaddr *saddr, *daddr;
1517 int multicast, error;
1521 saddr = &sc->vxl_src_addr;
1522 daddr = &sc->vxl_dst_addr;
1524 multicast = vxlan_sockaddr_in_multicast(daddr);
1525 MPASS(multicast != -1);
1526 sc->vxl_vso_mc_index = -1;
1529 * Try to create the socket. If that fails, attempt to use an
1532 error = vxlan_socket_create(ifp, multicast, saddr, &vso);
1535 vso = vxlan_socket_mc_lookup(saddr);
1537 vso = vxlan_socket_lookup(saddr);
1540 if_printf(ifp, "cannot create socket (error: %d), "
1541 "and no existing socket found\n", error);
1546 if (multicast != 0) {
1547 error = vxlan_setup_multicast(sc);
1551 error = vxlan_socket_mc_add_group(vso, daddr, saddr,
1552 sc->vxl_mc_ifindex, &sc->vxl_vso_mc_index);
1558 error = vxlan_socket_insert_softc(vso, sc);
1560 sc->vxl_sock = NULL;
1561 if_printf(ifp, "network identifier %d already exists in "
1562 "this socket\n", sc->vxl_vni);
1570 if (sc->vxl_vso_mc_index != -1) {
1571 vxlan_socket_mc_release_group_by_idx(vso,
1572 sc->vxl_vso_mc_index);
1573 sc->vxl_vso_mc_index = -1;
1576 vxlan_free_multicast(sc);
1577 vxlan_socket_release(vso);
1585 vxlan_setup_zero_checksum_port(struct vxlan_softc *sc)
1588 if (!VXLAN_SOCKADDR_IS_IPV6(&sc->vxl_src_addr))
1591 MPASS(sc->vxl_src_addr.in6.sin6_port != 0);
1592 MPASS(sc->vxl_dst_addr.in6.sin6_port != 0);
1594 if (sc->vxl_src_addr.in6.sin6_port != sc->vxl_dst_addr.in6.sin6_port) {
1595 if_printf(sc->vxl_ifp, "port %d in src address does not match "
1596 "port %d in dst address, rfc6935_port (%d) not updated.\n",
1597 ntohs(sc->vxl_src_addr.in6.sin6_port),
1598 ntohs(sc->vxl_dst_addr.in6.sin6_port),
1599 V_zero_checksum_port);
1603 if (V_zero_checksum_port != 0) {
1604 if (V_zero_checksum_port !=
1605 ntohs(sc->vxl_src_addr.in6.sin6_port)) {
1606 if_printf(sc->vxl_ifp, "rfc6935_port is already set to "
1607 "%d, cannot set it to %d.\n", V_zero_checksum_port,
1608 ntohs(sc->vxl_src_addr.in6.sin6_port));
1613 V_zero_checksum_port = ntohs(sc->vxl_src_addr.in6.sin6_port);
1614 if_printf(sc->vxl_ifp, "rfc6935_port set to %d\n",
1615 V_zero_checksum_port);
1620 vxlan_setup_interface_hdrlen(struct vxlan_softc *sc)
1624 VXLAN_LOCK_WASSERT(sc);
1627 ifp->if_hdrlen = ETHER_HDR_LEN + sizeof(struct vxlanudphdr);
1629 if (VXLAN_SOCKADDR_IS_IPV4(&sc->vxl_dst_addr) != 0)
1630 ifp->if_hdrlen += sizeof(struct ip);
1631 else if (VXLAN_SOCKADDR_IS_IPV6(&sc->vxl_dst_addr) != 0)
1632 ifp->if_hdrlen += sizeof(struct ip6_hdr);
1634 if ((sc->vxl_flags & VXLAN_FLAG_USER_MTU) == 0)
1635 ifp->if_mtu = ETHERMTU - ifp->if_hdrlen;
1639 vxlan_valid_init_config(struct vxlan_softc *sc)
1643 if (vxlan_check_vni(sc->vxl_vni) != 0) {
1644 reason = "invalid virtual network identifier specified";
1648 if (vxlan_sockaddr_supported(&sc->vxl_src_addr, 1) == 0) {
1649 reason = "source address type is not supported";
1653 if (vxlan_sockaddr_supported(&sc->vxl_dst_addr, 0) == 0) {
1654 reason = "destination address type is not supported";
1658 if (vxlan_sockaddr_in_any(&sc->vxl_dst_addr) != 0) {
1659 reason = "no valid destination address specified";
1663 if (vxlan_sockaddr_in_multicast(&sc->vxl_dst_addr) == 0 &&
1664 sc->vxl_mc_ifname[0] != '\0') {
1665 reason = "can only specify interface with a group address";
1669 if (vxlan_sockaddr_in_any(&sc->vxl_src_addr) == 0) {
1670 if (VXLAN_SOCKADDR_IS_IPV4(&sc->vxl_src_addr) ^
1671 VXLAN_SOCKADDR_IS_IPV4(&sc->vxl_dst_addr)) {
1672 reason = "source and destination address must both "
1673 "be either IPv4 or IPv6";
1678 if (sc->vxl_src_addr.in4.sin_port == 0) {
1679 reason = "local port not specified";
1683 if (sc->vxl_dst_addr.in4.sin_port == 0) {
1684 reason = "remote port not specified";
1691 if_printf(sc->vxl_ifp, "cannot initialize interface: %s\n", reason);
1696 vxlan_init_wait(struct vxlan_softc *sc)
1699 VXLAN_LOCK_WASSERT(sc);
1700 while (sc->vxl_flags & VXLAN_FLAG_INIT)
1701 rm_sleep(sc, &sc->vxl_lock, 0, "vxlint", hz);
1705 vxlan_init_complete(struct vxlan_softc *sc)
1709 sc->vxl_flags &= ~VXLAN_FLAG_INIT;
1715 vxlan_init(void *xsc)
1717 static const uint8_t empty_mac[ETHER_ADDR_LEN];
1718 struct vxlan_softc *sc;
1724 sx_xlock(&vxlan_sx);
1726 if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
1728 sx_xunlock(&vxlan_sx);
1731 sc->vxl_flags |= VXLAN_FLAG_INIT;
1734 if (vxlan_valid_init_config(sc) != 0)
1737 if (vxlan_setup_socket(sc) != 0)
1741 vxlan_setup_zero_checksum_port(sc);
1744 /* Initialize the default forwarding entry. */
1745 vxlan_ftable_entry_init(sc, &sc->vxl_default_fe, empty_mac,
1746 &sc->vxl_dst_addr.sa, VXLAN_FE_FLAG_STATIC);
1749 ifp->if_drv_flags |= IFF_DRV_RUNNING;
1750 callout_reset(&sc->vxl_callout, vxlan_ftable_prune_period * hz,
1754 if_link_state_change(ifp, LINK_STATE_UP);
1756 EVENTHANDLER_INVOKE(vxlan_start, ifp, sc->vxl_src_addr.in4.sin_family,
1757 ntohs(sc->vxl_src_addr.in4.sin_port));
1759 vxlan_init_complete(sc);
1760 sx_xunlock(&vxlan_sx);
1764 vxlan_release(struct vxlan_softc *sc)
1768 * The softc may be destroyed as soon as we release our reference,
1769 * so we cannot serialize the wakeup with the softc lock. We use a
1770 * timeout in our sleeps so a missed wakeup is unfortunate but not
1773 if (VXLAN_RELEASE(sc) != 0)
1778 vxlan_teardown_wait(struct vxlan_softc *sc)
1781 VXLAN_LOCK_WASSERT(sc);
1782 while (sc->vxl_flags & VXLAN_FLAG_TEARDOWN)
1783 rm_sleep(sc, &sc->vxl_lock, 0, "vxltrn", hz);
1787 vxlan_teardown_complete(struct vxlan_softc *sc)
1791 sc->vxl_flags &= ~VXLAN_FLAG_TEARDOWN;
1797 vxlan_teardown_locked(struct vxlan_softc *sc)
1800 struct vxlan_socket *vso;
1802 sx_assert(&vxlan_sx, SA_XLOCKED);
1803 VXLAN_LOCK_WASSERT(sc);
1804 MPASS(sc->vxl_flags & VXLAN_FLAG_TEARDOWN);
1807 ifp->if_flags &= ~IFF_UP;
1808 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1809 callout_stop(&sc->vxl_callout);
1811 sc->vxl_sock = NULL;
1814 if_link_state_change(ifp, LINK_STATE_DOWN);
1815 EVENTHANDLER_INVOKE(vxlan_stop, ifp, sc->vxl_src_addr.in4.sin_family,
1816 ntohs(sc->vxl_src_addr.in4.sin_port));
1819 vxlan_socket_remove_softc(vso, sc);
1821 if (sc->vxl_vso_mc_index != -1) {
1822 vxlan_socket_mc_release_group_by_idx(vso,
1823 sc->vxl_vso_mc_index);
1824 sc->vxl_vso_mc_index = -1;
1829 while (sc->vxl_refcnt != 0)
1830 rm_sleep(sc, &sc->vxl_lock, 0, "vxldrn", hz);
1833 callout_drain(&sc->vxl_callout);
1835 vxlan_free_multicast(sc);
1837 vxlan_socket_release(vso);
1839 vxlan_teardown_complete(sc);
1843 vxlan_teardown(struct vxlan_softc *sc)
1846 sx_xlock(&vxlan_sx);
1848 if (sc->vxl_flags & VXLAN_FLAG_TEARDOWN) {
1849 vxlan_teardown_wait(sc);
1851 sx_xunlock(&vxlan_sx);
1855 sc->vxl_flags |= VXLAN_FLAG_TEARDOWN;
1856 vxlan_teardown_locked(sc);
1857 sx_xunlock(&vxlan_sx);
1861 vxlan_ifdetach(struct vxlan_softc *sc, struct ifnet *ifp,
1862 struct vxlan_softc_head *list)
1867 if (sc->vxl_mc_ifp != ifp)
1869 if (sc->vxl_flags & VXLAN_FLAG_TEARDOWN)
1872 sc->vxl_flags |= VXLAN_FLAG_TEARDOWN;
1873 LIST_INSERT_HEAD(list, sc, vxl_ifdetach_list);
1880 vxlan_timer(void *xsc)
1882 struct vxlan_softc *sc;
1885 VXLAN_LOCK_WASSERT(sc);
1887 vxlan_ftable_expire(sc);
1888 callout_schedule(&sc->vxl_callout, vxlan_ftable_prune_period * hz);
1892 vxlan_ioctl_ifflags(struct vxlan_softc *sc)
1898 if (ifp->if_flags & IFF_UP) {
1899 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1902 if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1910 vxlan_ctrl_get_config(struct vxlan_softc *sc, void *arg)
1912 struct rm_priotracker tracker;
1913 struct ifvxlancfg *cfg;
1916 bzero(cfg, sizeof(*cfg));
1918 VXLAN_RLOCK(sc, &tracker);
1919 cfg->vxlc_vni = sc->vxl_vni;
1920 memcpy(&cfg->vxlc_local_sa, &sc->vxl_src_addr,
1921 sizeof(union vxlan_sockaddr));
1922 memcpy(&cfg->vxlc_remote_sa, &sc->vxl_dst_addr,
1923 sizeof(union vxlan_sockaddr));
1924 cfg->vxlc_mc_ifindex = sc->vxl_mc_ifindex;
1925 cfg->vxlc_ftable_cnt = sc->vxl_ftable_cnt;
1926 cfg->vxlc_ftable_max = sc->vxl_ftable_max;
1927 cfg->vxlc_ftable_timeout = sc->vxl_ftable_timeout;
1928 cfg->vxlc_port_min = sc->vxl_min_port;
1929 cfg->vxlc_port_max = sc->vxl_max_port;
1930 cfg->vxlc_learn = (sc->vxl_flags & VXLAN_FLAG_LEARN) != 0;
1931 cfg->vxlc_ttl = sc->vxl_ttl;
1932 VXLAN_RUNLOCK(sc, &tracker);
1935 if (VXLAN_SOCKADDR_IS_IPV6(&cfg->vxlc_local_sa))
1936 sa6_recoverscope(&cfg->vxlc_local_sa.in6);
1937 if (VXLAN_SOCKADDR_IS_IPV6(&cfg->vxlc_remote_sa))
1938 sa6_recoverscope(&cfg->vxlc_remote_sa.in6);
1945 vxlan_ctrl_set_vni(struct vxlan_softc *sc, void *arg)
1947 struct ifvxlancmd *cmd;
1952 if (vxlan_check_vni(cmd->vxlcmd_vni) != 0)
1956 if (vxlan_can_change_config(sc)) {
1957 sc->vxl_vni = cmd->vxlcmd_vni;
1967 vxlan_ctrl_set_local_addr(struct vxlan_softc *sc, void *arg)
1969 struct ifvxlancmd *cmd;
1970 union vxlan_sockaddr *vxlsa;
1974 vxlsa = &cmd->vxlcmd_sa;
1976 if (!VXLAN_SOCKADDR_IS_IPV46(vxlsa))
1978 if (vxlan_sockaddr_in_multicast(vxlsa) != 0)
1980 if (VXLAN_SOCKADDR_IS_IPV6(vxlsa)) {
1981 error = vxlan_sockaddr_in6_embedscope(vxlsa);
1987 if (vxlan_can_change_config(sc)) {
1988 vxlan_sockaddr_in_copy(&sc->vxl_src_addr, &vxlsa->sa);
1989 vxlan_set_hwcaps(sc);
1999 vxlan_ctrl_set_remote_addr(struct vxlan_softc *sc, void *arg)
2001 struct ifvxlancmd *cmd;
2002 union vxlan_sockaddr *vxlsa;
2006 vxlsa = &cmd->vxlcmd_sa;
2008 if (!VXLAN_SOCKADDR_IS_IPV46(vxlsa))
2010 if (VXLAN_SOCKADDR_IS_IPV6(vxlsa)) {
2011 error = vxlan_sockaddr_in6_embedscope(vxlsa);
2017 if (vxlan_can_change_config(sc)) {
2018 vxlan_sockaddr_in_copy(&sc->vxl_dst_addr, &vxlsa->sa);
2019 vxlan_setup_interface_hdrlen(sc);
2029 vxlan_ctrl_set_local_port(struct vxlan_softc *sc, void *arg)
2031 struct ifvxlancmd *cmd;
2036 if (cmd->vxlcmd_port == 0)
2040 if (vxlan_can_change_config(sc)) {
2041 sc->vxl_src_addr.in4.sin_port = htons(cmd->vxlcmd_port);
2051 vxlan_ctrl_set_remote_port(struct vxlan_softc *sc, void *arg)
2053 struct ifvxlancmd *cmd;
2058 if (cmd->vxlcmd_port == 0)
2062 if (vxlan_can_change_config(sc)) {
2063 sc->vxl_dst_addr.in4.sin_port = htons(cmd->vxlcmd_port);
2073 vxlan_ctrl_set_port_range(struct vxlan_softc *sc, void *arg)
2075 struct ifvxlancmd *cmd;
2080 min = cmd->vxlcmd_port_min;
2081 max = cmd->vxlcmd_port_max;
2087 if (vxlan_can_change_config(sc)) {
2088 sc->vxl_min_port = min;
2089 sc->vxl_max_port = max;
2099 vxlan_ctrl_set_ftable_timeout(struct vxlan_softc *sc, void *arg)
2101 struct ifvxlancmd *cmd;
2107 if (vxlan_check_ftable_timeout(cmd->vxlcmd_ftable_timeout) == 0) {
2108 sc->vxl_ftable_timeout = cmd->vxlcmd_ftable_timeout;
2118 vxlan_ctrl_set_ftable_max(struct vxlan_softc *sc, void *arg)
2120 struct ifvxlancmd *cmd;
2126 if (vxlan_check_ftable_max(cmd->vxlcmd_ftable_max) == 0) {
2127 sc->vxl_ftable_max = cmd->vxlcmd_ftable_max;
2137 vxlan_ctrl_set_multicast_if(struct vxlan_softc * sc, void *arg)
2139 struct ifvxlancmd *cmd;
2145 if (vxlan_can_change_config(sc)) {
2146 strlcpy(sc->vxl_mc_ifname, cmd->vxlcmd_ifname, IFNAMSIZ);
2147 vxlan_set_hwcaps(sc);
2157 vxlan_ctrl_set_ttl(struct vxlan_softc *sc, void *arg)
2159 struct ifvxlancmd *cmd;
2165 if (vxlan_check_ttl(cmd->vxlcmd_ttl) == 0) {
2166 sc->vxl_ttl = cmd->vxlcmd_ttl;
2167 if (sc->vxl_im4o != NULL)
2168 sc->vxl_im4o->imo_multicast_ttl = sc->vxl_ttl;
2169 if (sc->vxl_im6o != NULL)
2170 sc->vxl_im6o->im6o_multicast_hlim = sc->vxl_ttl;
2180 vxlan_ctrl_set_learn(struct vxlan_softc *sc, void *arg)
2182 struct ifvxlancmd *cmd;
2187 if (cmd->vxlcmd_flags & VXLAN_CMD_FLAG_LEARN)
2188 sc->vxl_flags |= VXLAN_FLAG_LEARN;
2190 sc->vxl_flags &= ~VXLAN_FLAG_LEARN;
2197 vxlan_ctrl_ftable_entry_add(struct vxlan_softc *sc, void *arg)
2199 union vxlan_sockaddr vxlsa;
2200 struct ifvxlancmd *cmd;
2201 struct vxlan_ftable_entry *fe;
2205 vxlsa = cmd->vxlcmd_sa;
2207 if (!VXLAN_SOCKADDR_IS_IPV46(&vxlsa))
2209 if (vxlan_sockaddr_in_any(&vxlsa) != 0)
2211 if (vxlan_sockaddr_in_multicast(&vxlsa) != 0)
2213 /* BMV: We could support both IPv4 and IPv6 later. */
2214 if (vxlsa.sa.sa_family != sc->vxl_dst_addr.sa.sa_family)
2215 return (EAFNOSUPPORT);
2217 if (VXLAN_SOCKADDR_IS_IPV6(&vxlsa)) {
2218 error = vxlan_sockaddr_in6_embedscope(&vxlsa);
2223 fe = vxlan_ftable_entry_alloc();
2227 if (vxlsa.in4.sin_port == 0)
2228 vxlsa.in4.sin_port = sc->vxl_dst_addr.in4.sin_port;
2230 vxlan_ftable_entry_init(sc, fe, cmd->vxlcmd_mac, &vxlsa.sa,
2231 VXLAN_FE_FLAG_STATIC);
2234 error = vxlan_ftable_entry_insert(sc, fe);
2238 vxlan_ftable_entry_free(fe);
2244 vxlan_ctrl_ftable_entry_rem(struct vxlan_softc *sc, void *arg)
2246 struct ifvxlancmd *cmd;
2247 struct vxlan_ftable_entry *fe;
2253 fe = vxlan_ftable_entry_lookup(sc, cmd->vxlcmd_mac);
2255 vxlan_ftable_entry_destroy(sc, fe);
2265 vxlan_ctrl_flush(struct vxlan_softc *sc, void *arg)
2267 struct ifvxlancmd *cmd;
2271 all = cmd->vxlcmd_flags & VXLAN_CMD_FLAG_FLUSH_ALL;
2274 vxlan_ftable_flush(sc, all);
2281 vxlan_ioctl_drvspec(struct vxlan_softc *sc, struct ifdrv *ifd, int get)
2283 const struct vxlan_control *vc;
2285 struct ifvxlancfg cfg;
2286 struct ifvxlancmd cmd;
2290 if (ifd->ifd_cmd >= vxlan_control_table_size)
2293 bzero(&args, sizeof(args));
2294 vc = &vxlan_control_table[ifd->ifd_cmd];
2295 out = (vc->vxlc_flags & VXLAN_CTRL_FLAG_COPYOUT) != 0;
2297 if ((get != 0 && out == 0) || (get == 0 && out != 0))
2300 if (vc->vxlc_flags & VXLAN_CTRL_FLAG_SUSER) {
2301 error = priv_check(curthread, PRIV_NET_VXLAN);
2306 if (ifd->ifd_len != vc->vxlc_argsize ||
2307 ifd->ifd_len > sizeof(args))
2310 if (vc->vxlc_flags & VXLAN_CTRL_FLAG_COPYIN) {
2311 error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
2316 error = vc->vxlc_func(sc, &args);
2320 if (vc->vxlc_flags & VXLAN_CTRL_FLAG_COPYOUT) {
2321 error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
2330 vxlan_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
2332 struct vxlan_softc *sc;
2338 ifr = (struct ifreq *) data;
2339 ifd = (struct ifdrv *) data;
2350 error = vxlan_ioctl_drvspec(sc, ifd, cmd == SIOCGDRVSPEC);
2354 error = vxlan_ioctl_ifflags(sc);
2359 error = ifmedia_ioctl(ifp, ifr, &sc->vxl_media, cmd);
2363 if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > VXLAN_MAX_MTU) {
2367 ifp->if_mtu = ifr->ifr_mtu;
2368 sc->vxl_flags |= VXLAN_FLAG_USER_MTU;
2375 error = vxlan_set_reqcap(sc, ifp, ifr->ifr_reqcap);
2377 vxlan_set_hwcaps(sc);
2382 error = ether_ioctl(ifp, cmd, data);
2389 #if defined(INET) || defined(INET6)
2391 vxlan_pick_source_port(struct vxlan_softc *sc, struct mbuf *m)
2396 range = sc->vxl_max_port - sc->vxl_min_port + 1;
2398 if (M_HASHTYPE_ISHASH(m))
2399 hash = m->m_pkthdr.flowid;
2401 hash = jenkins_hash(m->m_data, ETHER_HDR_LEN,
2402 sc->vxl_port_hash_key);
2404 return (sc->vxl_min_port + (hash % range));
2408 vxlan_encap_header(struct vxlan_softc *sc, struct mbuf *m, int ipoff,
2409 uint16_t srcport, uint16_t dstport)
2411 struct vxlanudphdr *hdr;
2412 struct udphdr *udph;
2413 struct vxlan_header *vxh;
2416 len = m->m_pkthdr.len - ipoff;
2417 MPASS(len >= sizeof(struct vxlanudphdr));
2418 hdr = mtodo(m, ipoff);
2420 udph = &hdr->vxlh_udp;
2421 udph->uh_sport = srcport;
2422 udph->uh_dport = dstport;
2423 udph->uh_ulen = htons(len);
2426 vxh = &hdr->vxlh_hdr;
2427 vxh->vxlh_flags = htonl(VXLAN_HDR_FLAGS_VALID_VNI);
2428 vxh->vxlh_vni = htonl(sc->vxl_vni << VXLAN_HDR_VNI_SHIFT);
2432 #if defined(INET6) || defined(INET)
2434 * Return the CSUM_INNER_* equivalent of CSUM_* caps.
2437 csum_flags_to_inner_flags(uint32_t csum_flags_in, const uint32_t encap)
2439 uint32_t csum_flags = encap;
2440 const uint32_t v4 = CSUM_IP | CSUM_IP_UDP | CSUM_IP_TCP;
2443 * csum_flags can request either v4 or v6 offload but not both.
2444 * tcp_output always sets CSUM_TSO (both CSUM_IP_TSO and CSUM_IP6_TSO)
2445 * so those bits are no good to detect the IP version. Other bits are
2446 * always set with CSUM_TSO and we use those to figure out the IP
2449 if (csum_flags_in & v4) {
2450 if (csum_flags_in & CSUM_IP)
2451 csum_flags |= CSUM_INNER_IP;
2452 if (csum_flags_in & CSUM_IP_UDP)
2453 csum_flags |= CSUM_INNER_IP_UDP;
2454 if (csum_flags_in & CSUM_IP_TCP)
2455 csum_flags |= CSUM_INNER_IP_TCP;
2456 if (csum_flags_in & CSUM_IP_TSO)
2457 csum_flags |= CSUM_INNER_IP_TSO;
2460 const uint32_t v6 = CSUM_IP6_UDP | CSUM_IP6_TCP;
2462 MPASS((csum_flags_in & v6) != 0);
2464 if (csum_flags_in & CSUM_IP6_UDP)
2465 csum_flags |= CSUM_INNER_IP6_UDP;
2466 if (csum_flags_in & CSUM_IP6_TCP)
2467 csum_flags |= CSUM_INNER_IP6_TCP;
2468 if (csum_flags_in & CSUM_IP6_TSO)
2469 csum_flags |= CSUM_INNER_IP6_TSO;
2472 return (csum_flags);
2477 vxlan_encap4(struct vxlan_softc *sc, const union vxlan_sockaddr *fvxlsa,
2483 struct in_addr srcaddr, dstaddr;
2484 uint16_t srcport, dstport;
2485 int len, mcast, error;
2486 struct route route, *ro;
2487 struct sockaddr_in *sin;
2488 uint32_t csum_flags;
2493 srcaddr = sc->vxl_src_addr.in4.sin_addr;
2494 srcport = vxlan_pick_source_port(sc, m);
2495 dstaddr = fvxlsa->in4.sin_addr;
2496 dstport = fvxlsa->in4.sin_port;
2498 M_PREPEND(m, sizeof(struct ip) + sizeof(struct vxlanudphdr),
2501 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2505 len = m->m_pkthdr.len;
2507 ip = mtod(m, struct ip *);
2509 ip->ip_len = htons(len);
2511 ip->ip_ttl = sc->vxl_ttl;
2512 ip->ip_p = IPPROTO_UDP;
2514 ip->ip_src = srcaddr;
2515 ip->ip_dst = dstaddr;
2517 vxlan_encap_header(sc, m, sizeof(struct ip), srcport, dstport);
2519 mcast = (m->m_flags & (M_MCAST | M_BCAST)) ? 1 : 0;
2520 m->m_flags &= ~(M_MCAST | M_BCAST);
2522 m->m_pkthdr.csum_flags &= CSUM_FLAGS_TX;
2523 if (m->m_pkthdr.csum_flags != 0) {
2525 * HW checksum (L3 and/or L4) or TSO has been requested. Look
2526 * up the ifnet for the outbound route and verify that the
2527 * outbound ifnet can perform the requested operation on the
2530 bzero(&route, sizeof(route));
2532 sin = (struct sockaddr_in *)&ro->ro_dst;
2533 sin->sin_family = AF_INET;
2534 sin->sin_len = sizeof(*sin);
2535 sin->sin_addr = ip->ip_dst;
2536 ro->ro_nh = fib4_lookup(RT_DEFAULT_FIB, ip->ip_dst, 0, NHR_NONE,
2538 if (ro->ro_nh == NULL) {
2540 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2541 return (EHOSTUNREACH);
2544 csum_flags = csum_flags_to_inner_flags(m->m_pkthdr.csum_flags,
2546 if ((csum_flags & ro->ro_nh->nh_ifp->if_hwassist) !=
2548 if (ppsratecheck(&sc->err_time, &sc->err_pps, 1)) {
2549 const struct ifnet *nh_ifp = ro->ro_nh->nh_ifp;
2551 if_printf(ifp, "interface %s is missing hwcaps "
2552 "0x%08x, csum_flags 0x%08x -> 0x%08x, "
2553 "hwassist 0x%08x\n", nh_ifp->if_xname,
2554 csum_flags & ~(uint32_t)nh_ifp->if_hwassist,
2555 m->m_pkthdr.csum_flags, csum_flags,
2556 (uint32_t)nh_ifp->if_hwassist);
2559 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2562 m->m_pkthdr.csum_flags = csum_flags;
2564 (CSUM_INNER_IP | CSUM_INNER_IP_UDP | CSUM_INNER_IP6_UDP |
2565 CSUM_INNER_IP_TCP | CSUM_INNER_IP6_TCP)) {
2566 counter_u64_add(sc->vxl_stats.txcsum, 1);
2567 if (csum_flags & CSUM_INNER_TSO)
2568 counter_u64_add(sc->vxl_stats.tso, 1);
2572 error = ip_output(m, NULL, ro, 0, sc->vxl_im4o, NULL);
2574 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
2575 if_inc_counter(ifp, IFCOUNTER_OBYTES, len);
2577 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
2579 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2589 vxlan_encap6(struct vxlan_softc *sc, const union vxlan_sockaddr *fvxlsa,
2594 struct ip6_hdr *ip6;
2595 const struct in6_addr *srcaddr, *dstaddr;
2596 uint16_t srcport, dstport;
2597 int len, mcast, error;
2598 struct route_in6 route, *ro;
2599 struct sockaddr_in6 *sin6;
2600 uint32_t csum_flags;
2605 srcaddr = &sc->vxl_src_addr.in6.sin6_addr;
2606 srcport = vxlan_pick_source_port(sc, m);
2607 dstaddr = &fvxlsa->in6.sin6_addr;
2608 dstport = fvxlsa->in6.sin6_port;
2610 M_PREPEND(m, sizeof(struct ip6_hdr) + sizeof(struct vxlanudphdr),
2613 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2617 len = m->m_pkthdr.len;
2619 ip6 = mtod(m, struct ip6_hdr *);
2620 ip6->ip6_flow = 0; /* BMV: Keep in forwarding entry? */
2621 ip6->ip6_vfc = IPV6_VERSION;
2623 ip6->ip6_nxt = IPPROTO_UDP;
2624 ip6->ip6_hlim = sc->vxl_ttl;
2625 ip6->ip6_src = *srcaddr;
2626 ip6->ip6_dst = *dstaddr;
2628 vxlan_encap_header(sc, m, sizeof(struct ip6_hdr), srcport, dstport);
2630 mcast = (m->m_flags & (M_MCAST | M_BCAST)) ? 1 : 0;
2631 m->m_flags &= ~(M_MCAST | M_BCAST);
2634 m->m_pkthdr.csum_flags &= CSUM_FLAGS_TX;
2635 if (m->m_pkthdr.csum_flags != 0) {
2637 * HW checksum (L3 and/or L4) or TSO has been requested. Look
2638 * up the ifnet for the outbound route and verify that the
2639 * outbound ifnet can perform the requested operation on the
2642 bzero(&route, sizeof(route));
2644 sin6 = (struct sockaddr_in6 *)&ro->ro_dst;
2645 sin6->sin6_family = AF_INET6;
2646 sin6->sin6_len = sizeof(*sin6);
2647 sin6->sin6_addr = ip6->ip6_dst;
2648 ro->ro_nh = fib6_lookup(RT_DEFAULT_FIB, &ip6->ip6_dst, 0,
2650 if (ro->ro_nh == NULL) {
2652 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2653 return (EHOSTUNREACH);
2656 csum_flags = csum_flags_to_inner_flags(m->m_pkthdr.csum_flags,
2658 if ((csum_flags & ro->ro_nh->nh_ifp->if_hwassist) !=
2660 if (ppsratecheck(&sc->err_time, &sc->err_pps, 1)) {
2661 const struct ifnet *nh_ifp = ro->ro_nh->nh_ifp;
2663 if_printf(ifp, "interface %s is missing hwcaps "
2664 "0x%08x, csum_flags 0x%08x -> 0x%08x, "
2665 "hwassist 0x%08x\n", nh_ifp->if_xname,
2666 csum_flags & ~(uint32_t)nh_ifp->if_hwassist,
2667 m->m_pkthdr.csum_flags, csum_flags,
2668 (uint32_t)nh_ifp->if_hwassist);
2671 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2674 m->m_pkthdr.csum_flags = csum_flags;
2676 (CSUM_INNER_IP | CSUM_INNER_IP_UDP | CSUM_INNER_IP6_UDP |
2677 CSUM_INNER_IP_TCP | CSUM_INNER_IP6_TCP)) {
2678 counter_u64_add(sc->vxl_stats.txcsum, 1);
2679 if (csum_flags & CSUM_INNER_TSO)
2680 counter_u64_add(sc->vxl_stats.tso, 1);
2682 } else if (ntohs(dstport) != V_zero_checksum_port) {
2683 struct udphdr *hdr = mtodo(m, sizeof(struct ip6_hdr));
2685 hdr->uh_sum = in6_cksum_pseudo(ip6,
2686 m->m_pkthdr.len - sizeof(struct ip6_hdr), IPPROTO_UDP, 0);
2687 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
2688 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
2690 error = ip6_output(m, NULL, ro, 0, sc->vxl_im6o, NULL, NULL);
2692 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
2693 if_inc_counter(ifp, IFCOUNTER_OBYTES, len);
2695 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
2697 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2707 vxlan_transmit(struct ifnet *ifp, struct mbuf *m)
2709 struct rm_priotracker tracker;
2710 union vxlan_sockaddr vxlsa;
2711 struct vxlan_softc *sc;
2712 struct vxlan_ftable_entry *fe;
2713 struct ifnet *mcifp;
2714 struct ether_header *eh;
2718 eh = mtod(m, struct ether_header *);
2722 ETHER_BPF_MTAP(ifp, m);
2724 VXLAN_RLOCK(sc, &tracker);
2725 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2726 VXLAN_RUNLOCK(sc, &tracker);
2731 if ((m->m_flags & (M_BCAST | M_MCAST)) == 0)
2732 fe = vxlan_ftable_entry_lookup(sc, eh->ether_dhost);
2734 fe = &sc->vxl_default_fe;
2735 vxlan_sockaddr_copy(&vxlsa, &fe->vxlfe_raddr.sa);
2737 ipv4 = VXLAN_SOCKADDR_IS_IPV4(&vxlsa) != 0;
2738 if (vxlan_sockaddr_in_multicast(&vxlsa) != 0)
2739 mcifp = vxlan_multicast_if_ref(sc, ipv4);
2742 VXLAN_RUNLOCK(sc, &tracker);
2745 error = vxlan_encap4(sc, &vxlsa, m);
2747 error = vxlan_encap6(sc, &vxlsa, m);
2757 vxlan_qflush(struct ifnet *ifp __unused)
2762 vxlan_rcv_udp_packet(struct mbuf *m, int offset, struct inpcb *inpcb,
2763 const struct sockaddr *srcsa, void *xvso)
2765 struct vxlan_socket *vso;
2766 struct vxlan_header *vxh, vxlanhdr;
2772 offset += sizeof(struct udphdr);
2774 if (m->m_pkthdr.len < offset + sizeof(struct vxlan_header))
2777 if (__predict_false(m->m_len < offset + sizeof(struct vxlan_header))) {
2778 m_copydata(m, offset, sizeof(struct vxlan_header),
2779 (caddr_t) &vxlanhdr);
2782 vxh = mtodo(m, offset);
2785 * Drop if there is a reserved bit set in either the flags or VNI
2786 * fields of the header. This goes against the specification, but
2787 * a bit set may indicate an unsupported new feature. This matches
2788 * the behavior of the Linux implementation.
2790 if (vxh->vxlh_flags != htonl(VXLAN_HDR_FLAGS_VALID_VNI) ||
2791 vxh->vxlh_vni & ~VXLAN_VNI_MASK)
2794 vni = ntohl(vxh->vxlh_vni) >> VXLAN_HDR_VNI_SHIFT;
2796 /* Adjust to the start of the inner Ethernet frame. */
2797 m_adj_decap(m, offset + sizeof(struct vxlan_header));
2799 error = vxlan_input(vso, vni, &m, srcsa);
2800 MPASS(error != 0 || m == NULL);
2808 vxlan_input(struct vxlan_socket *vso, uint32_t vni, struct mbuf **m0,
2809 const struct sockaddr *sa)
2811 struct vxlan_softc *sc;
2814 struct ether_header *eh;
2819 if (m->m_pkthdr.len < ETHER_HDR_LEN)
2822 sc = vxlan_socket_lookup_softc(vso, vni);
2827 eh = mtod(m, struct ether_header *);
2829 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2832 } else if (ifp == m->m_pkthdr.rcvif) {
2833 /* XXX Does not catch more complex loops. */
2838 if (sc->vxl_flags & VXLAN_FLAG_LEARN)
2839 vxlan_ftable_learn(sc, sa, eh->ether_shost);
2842 m->m_pkthdr.rcvif = ifp;
2843 M_SETFIB(m, ifp->if_fib);
2844 if (((ifp->if_capenable & IFCAP_RXCSUM &&
2845 m->m_pkthdr.csum_flags & CSUM_INNER_L3_CALC) ||
2846 (ifp->if_capenable & IFCAP_RXCSUM_IPV6 &&
2847 !(m->m_pkthdr.csum_flags & CSUM_INNER_L3_CALC)))) {
2848 uint32_t csum_flags = 0;
2850 if (m->m_pkthdr.csum_flags & CSUM_INNER_L3_CALC)
2851 csum_flags |= CSUM_L3_CALC;
2852 if (m->m_pkthdr.csum_flags & CSUM_INNER_L3_VALID)
2853 csum_flags |= CSUM_L3_VALID;
2854 if (m->m_pkthdr.csum_flags & CSUM_INNER_L4_CALC)
2855 csum_flags |= CSUM_L4_CALC;
2856 if (m->m_pkthdr.csum_flags & CSUM_INNER_L4_VALID)
2857 csum_flags |= CSUM_L4_VALID;
2858 m->m_pkthdr.csum_flags = csum_flags;
2859 counter_u64_add(sc->vxl_stats.rxcsum, 1);
2861 /* clear everything */
2862 m->m_pkthdr.csum_flags = 0;
2863 m->m_pkthdr.csum_data = 0;
2866 (*ifp->if_input)(ifp, m);
2876 vxlan_stats_alloc(struct vxlan_softc *sc)
2878 struct vxlan_statistics *stats = &sc->vxl_stats;
2880 stats->txcsum = counter_u64_alloc(M_WAITOK);
2881 if (stats->txcsum == NULL)
2884 stats->tso = counter_u64_alloc(M_WAITOK);
2885 if (stats->tso == NULL)
2888 stats->rxcsum = counter_u64_alloc(M_WAITOK);
2889 if (stats->rxcsum == NULL)
2894 vxlan_stats_free(sc);
2899 vxlan_stats_free(struct vxlan_softc *sc)
2901 struct vxlan_statistics *stats = &sc->vxl_stats;
2903 if (stats->txcsum != NULL) {
2904 counter_u64_free(stats->txcsum);
2905 stats->txcsum = NULL;
2907 if (stats->tso != NULL) {
2908 counter_u64_free(stats->tso);
2911 if (stats->rxcsum != NULL) {
2912 counter_u64_free(stats->rxcsum);
2913 stats->rxcsum = NULL;
2918 vxlan_set_default_config(struct vxlan_softc *sc)
2921 sc->vxl_flags |= VXLAN_FLAG_LEARN;
2923 sc->vxl_vni = VXLAN_VNI_MAX;
2924 sc->vxl_ttl = IPDEFTTL;
2926 if (!vxlan_tunable_int(sc, "legacy_port", vxlan_legacy_port)) {
2927 sc->vxl_src_addr.in4.sin_port = htons(VXLAN_PORT);
2928 sc->vxl_dst_addr.in4.sin_port = htons(VXLAN_PORT);
2930 sc->vxl_src_addr.in4.sin_port = htons(VXLAN_LEGACY_PORT);
2931 sc->vxl_dst_addr.in4.sin_port = htons(VXLAN_LEGACY_PORT);
2934 sc->vxl_min_port = V_ipport_firstauto;
2935 sc->vxl_max_port = V_ipport_lastauto;
2937 sc->vxl_ftable_max = VXLAN_FTABLE_MAX;
2938 sc->vxl_ftable_timeout = VXLAN_FTABLE_TIMEOUT;
2942 vxlan_set_user_config(struct vxlan_softc *sc, struct ifvxlanparam *vxlp)
2946 if (vxlp->vxlp_with & (VXLAN_PARAM_WITH_LOCAL_ADDR4 |
2947 VXLAN_PARAM_WITH_REMOTE_ADDR4))
2948 return (EAFNOSUPPORT);
2952 if (vxlp->vxlp_with & (VXLAN_PARAM_WITH_LOCAL_ADDR6 |
2953 VXLAN_PARAM_WITH_REMOTE_ADDR6))
2954 return (EAFNOSUPPORT);
2956 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LOCAL_ADDR6) {
2957 int error = vxlan_sockaddr_in6_embedscope(&vxlp->vxlp_local_sa);
2961 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_REMOTE_ADDR6) {
2962 int error = vxlan_sockaddr_in6_embedscope(
2963 &vxlp->vxlp_remote_sa);
2969 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_VNI) {
2970 if (vxlan_check_vni(vxlp->vxlp_vni) == 0)
2971 sc->vxl_vni = vxlp->vxlp_vni;
2974 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LOCAL_ADDR4) {
2975 sc->vxl_src_addr.in4.sin_len = sizeof(struct sockaddr_in);
2976 sc->vxl_src_addr.in4.sin_family = AF_INET;
2977 sc->vxl_src_addr.in4.sin_addr =
2978 vxlp->vxlp_local_sa.in4.sin_addr;
2979 } else if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LOCAL_ADDR6) {
2980 sc->vxl_src_addr.in6.sin6_len = sizeof(struct sockaddr_in6);
2981 sc->vxl_src_addr.in6.sin6_family = AF_INET6;
2982 sc->vxl_src_addr.in6.sin6_addr =
2983 vxlp->vxlp_local_sa.in6.sin6_addr;
2986 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_REMOTE_ADDR4) {
2987 sc->vxl_dst_addr.in4.sin_len = sizeof(struct sockaddr_in);
2988 sc->vxl_dst_addr.in4.sin_family = AF_INET;
2989 sc->vxl_dst_addr.in4.sin_addr =
2990 vxlp->vxlp_remote_sa.in4.sin_addr;
2991 } else if (vxlp->vxlp_with & VXLAN_PARAM_WITH_REMOTE_ADDR6) {
2992 sc->vxl_dst_addr.in6.sin6_len = sizeof(struct sockaddr_in6);
2993 sc->vxl_dst_addr.in6.sin6_family = AF_INET6;
2994 sc->vxl_dst_addr.in6.sin6_addr =
2995 vxlp->vxlp_remote_sa.in6.sin6_addr;
2998 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LOCAL_PORT)
2999 sc->vxl_src_addr.in4.sin_port = htons(vxlp->vxlp_local_port);
3000 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_REMOTE_PORT)
3001 sc->vxl_dst_addr.in4.sin_port = htons(vxlp->vxlp_remote_port);
3003 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_PORT_RANGE) {
3004 if (vxlp->vxlp_min_port <= vxlp->vxlp_max_port) {
3005 sc->vxl_min_port = vxlp->vxlp_min_port;
3006 sc->vxl_max_port = vxlp->vxlp_max_port;
3010 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_MULTICAST_IF)
3011 strlcpy(sc->vxl_mc_ifname, vxlp->vxlp_mc_ifname, IFNAMSIZ);
3013 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_FTABLE_TIMEOUT) {
3014 if (vxlan_check_ftable_timeout(vxlp->vxlp_ftable_timeout) == 0)
3015 sc->vxl_ftable_timeout = vxlp->vxlp_ftable_timeout;
3018 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_FTABLE_MAX) {
3019 if (vxlan_check_ftable_max(vxlp->vxlp_ftable_max) == 0)
3020 sc->vxl_ftable_max = vxlp->vxlp_ftable_max;
3023 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_TTL) {
3024 if (vxlan_check_ttl(vxlp->vxlp_ttl) == 0)
3025 sc->vxl_ttl = vxlp->vxlp_ttl;
3028 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LEARN) {
3029 if (vxlp->vxlp_learn == 0)
3030 sc->vxl_flags &= ~VXLAN_FLAG_LEARN;
3037 vxlan_set_reqcap(struct vxlan_softc *sc, struct ifnet *ifp, int reqcap)
3039 int mask = reqcap ^ ifp->if_capenable;
3041 /* Disable TSO if tx checksums are disabled. */
3042 if (mask & IFCAP_TXCSUM && !(reqcap & IFCAP_TXCSUM) &&
3043 reqcap & IFCAP_TSO4) {
3044 reqcap &= ~IFCAP_TSO4;
3045 if_printf(ifp, "tso4 disabled due to -txcsum.\n");
3047 if (mask & IFCAP_TXCSUM_IPV6 && !(reqcap & IFCAP_TXCSUM_IPV6) &&
3048 reqcap & IFCAP_TSO6) {
3049 reqcap &= ~IFCAP_TSO6;
3050 if_printf(ifp, "tso6 disabled due to -txcsum6.\n");
3053 /* Do not enable TSO if tx checksums are disabled. */
3054 if (mask & IFCAP_TSO4 && reqcap & IFCAP_TSO4 &&
3055 !(reqcap & IFCAP_TXCSUM)) {
3056 if_printf(ifp, "enable txcsum first.\n");
3059 if (mask & IFCAP_TSO6 && reqcap & IFCAP_TSO6 &&
3060 !(reqcap & IFCAP_TXCSUM_IPV6)) {
3061 if_printf(ifp, "enable txcsum6 first.\n");
3065 sc->vxl_reqcap = reqcap;
3070 * A VXLAN interface inherits the capabilities of the vxlandev or the interface
3071 * hosting the vxlanlocal address.
3074 vxlan_set_hwcaps(struct vxlan_softc *sc)
3076 struct epoch_tracker et;
3082 struct ifnet *ifp = sc->vxl_ifp;
3085 ifp->if_capabilities &= VXLAN_BASIC_IFCAPS;
3086 ifp->if_capenable &= VXLAN_BASIC_IFCAPS;
3087 ifp->if_hwassist = 0;
3089 NET_EPOCH_ENTER(et);
3090 CURVNET_SET(ifp->if_vnet);
3094 if (sc->vxl_mc_ifname[0] != '\0') {
3096 p = ifunit_ref(sc->vxl_mc_ifname);
3097 } else if (vxlan_sockaddr_in_any(&sc->vxl_src_addr) == 0) {
3098 if (sc->vxl_src_addr.sa.sa_family == AF_INET) {
3099 struct sockaddr_in in4 = sc->vxl_src_addr.in4;
3102 ifa = ifa_ifwithaddr((struct sockaddr *)&in4);
3105 } else if (sc->vxl_src_addr.sa.sa_family == AF_INET6) {
3106 struct sockaddr_in6 in6 = sc->vxl_src_addr.in6;
3109 ifa = ifa_ifwithaddr((struct sockaddr *)&in6);
3117 cap = ena = hwa = 0;
3119 /* checksum offload */
3120 if (p->if_capabilities & IFCAP_VXLAN_HWCSUM)
3121 cap |= p->if_capabilities & (IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6);
3122 if (p->if_capenable & IFCAP_VXLAN_HWCSUM) {
3123 ena |= sc->vxl_reqcap & p->if_capenable &
3124 (IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6);
3125 if (ena & IFCAP_TXCSUM) {
3126 if (p->if_hwassist & CSUM_INNER_IP)
3128 if (p->if_hwassist & CSUM_INNER_IP_UDP)
3130 if (p->if_hwassist & CSUM_INNER_IP_TCP)
3133 if (ena & IFCAP_TXCSUM_IPV6) {
3134 if (p->if_hwassist & CSUM_INNER_IP6_UDP)
3135 hwa |= CSUM_IP6_UDP;
3136 if (p->if_hwassist & CSUM_INNER_IP6_TCP)
3137 hwa |= CSUM_IP6_TCP;
3142 if (p->if_capabilities & IFCAP_VXLAN_HWTSO) {
3143 cap |= p->if_capabilities & IFCAP_TSO;
3144 if (p->if_hw_tsomax > IP_MAXPACKET - ifp->if_hdrlen)
3145 ifp->if_hw_tsomax = IP_MAXPACKET - ifp->if_hdrlen;
3147 ifp->if_hw_tsomax = p->if_hw_tsomax;
3148 /* XXX: tsomaxsegcount decrement is cxgbe specific */
3149 ifp->if_hw_tsomaxsegcount = p->if_hw_tsomaxsegcount - 1;
3150 ifp->if_hw_tsomaxsegsize = p->if_hw_tsomaxsegsize;
3152 if (p->if_capenable & IFCAP_VXLAN_HWTSO) {
3153 ena |= sc->vxl_reqcap & p->if_capenable & IFCAP_TSO;
3154 if (ena & IFCAP_TSO) {
3155 if (p->if_hwassist & CSUM_INNER_IP_TSO)
3157 if (p->if_hwassist & CSUM_INNER_IP6_TSO)
3158 hwa |= CSUM_IP6_TSO;
3162 ifp->if_capabilities |= cap;
3163 ifp->if_capenable |= ena;
3164 ifp->if_hwassist |= hwa;
3173 vxlan_clone_create(struct if_clone *ifc, int unit, caddr_t params)
3175 struct vxlan_softc *sc;
3177 struct ifvxlanparam vxlp;
3180 sc = malloc(sizeof(struct vxlan_softc), M_VXLAN, M_WAITOK | M_ZERO);
3181 sc->vxl_unit = unit;
3182 vxlan_set_default_config(sc);
3183 error = vxlan_stats_alloc(sc);
3188 error = copyin(params, &vxlp, sizeof(vxlp));
3192 error = vxlan_set_user_config(sc, &vxlp);
3197 ifp = if_alloc(IFT_ETHER);
3204 rm_init(&sc->vxl_lock, "vxlanrm");
3205 callout_init_rw(&sc->vxl_callout, &sc->vxl_lock, 0);
3206 sc->vxl_port_hash_key = arc4random();
3207 vxlan_ftable_init(sc);
3209 vxlan_sysctl_setup(sc);
3212 if_initname(ifp, vxlan_name, unit);
3213 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
3214 ifp->if_init = vxlan_init;
3215 ifp->if_ioctl = vxlan_ioctl;
3216 ifp->if_transmit = vxlan_transmit;
3217 ifp->if_qflush = vxlan_qflush;
3218 ifp->if_capabilities = VXLAN_BASIC_IFCAPS;
3219 ifp->if_capenable = VXLAN_BASIC_IFCAPS;
3220 sc->vxl_reqcap = -1;
3221 vxlan_set_hwcaps(sc);
3223 ifmedia_init(&sc->vxl_media, 0, vxlan_media_change, vxlan_media_status);
3224 ifmedia_add(&sc->vxl_media, IFM_ETHER | IFM_AUTO, 0, NULL);
3225 ifmedia_set(&sc->vxl_media, IFM_ETHER | IFM_AUTO);
3227 ether_gen_addr(ifp, &sc->vxl_hwaddr);
3228 ether_ifattach(ifp, sc->vxl_hwaddr.octet);
3230 ifp->if_baudrate = 0;
3233 vxlan_setup_interface_hdrlen(sc);
3244 vxlan_clone_destroy(struct ifnet *ifp)
3246 struct vxlan_softc *sc;
3252 vxlan_ftable_flush(sc, 1);
3254 ether_ifdetach(ifp);
3256 ifmedia_removeall(&sc->vxl_media);
3258 vxlan_ftable_fini(sc);
3260 vxlan_sysctl_destroy(sc);
3261 rm_destroy(&sc->vxl_lock);
3262 vxlan_stats_free(sc);
3266 /* BMV: Taken from if_bridge. */
3268 vxlan_mac_hash(struct vxlan_softc *sc, const uint8_t *addr)
3270 uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->vxl_ftable_hash_key;
3280 * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3281 * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3283 #define mix(a, b, c) \
3285 a -= b; a -= c; a ^= (c >> 13); \
3286 b -= c; b -= a; b ^= (a << 8); \
3287 c -= a; c -= b; c ^= (b >> 13); \
3288 a -= b; a -= c; a ^= (c >> 12); \
3289 b -= c; b -= a; b ^= (a << 16); \
3290 c -= a; c -= b; c ^= (b >> 5); \
3291 a -= b; a -= c; a ^= (c >> 3); \
3292 b -= c; b -= a; b ^= (a << 10); \
3293 c -= a; c -= b; c ^= (b >> 15); \
3304 vxlan_media_change(struct ifnet *ifp)
3312 vxlan_media_status(struct ifnet *ifp, struct ifmediareq *ifmr)
3315 ifmr->ifm_status = IFM_ACTIVE | IFM_AVALID;
3316 ifmr->ifm_active = IFM_ETHER | IFM_FDX;
3320 vxlan_sockaddr_cmp(const union vxlan_sockaddr *vxladdr,
3321 const struct sockaddr *sa)
3324 return (bcmp(&vxladdr->sa, sa, vxladdr->sa.sa_len));
3328 vxlan_sockaddr_copy(union vxlan_sockaddr *vxladdr,
3329 const struct sockaddr *sa)
3332 MPASS(sa->sa_family == AF_INET || sa->sa_family == AF_INET6);
3333 bzero(vxladdr, sizeof(*vxladdr));
3335 if (sa->sa_family == AF_INET) {
3336 vxladdr->in4 = *satoconstsin(sa);
3337 vxladdr->in4.sin_len = sizeof(struct sockaddr_in);
3338 } else if (sa->sa_family == AF_INET6) {
3339 vxladdr->in6 = *satoconstsin6(sa);
3340 vxladdr->in6.sin6_len = sizeof(struct sockaddr_in6);
3345 vxlan_sockaddr_in_equal(const union vxlan_sockaddr *vxladdr,
3346 const struct sockaddr *sa)
3350 if (sa->sa_family == AF_INET) {
3351 const struct in_addr *in4 = &satoconstsin(sa)->sin_addr;
3352 equal = in4->s_addr == vxladdr->in4.sin_addr.s_addr;
3353 } else if (sa->sa_family == AF_INET6) {
3354 const struct in6_addr *in6 = &satoconstsin6(sa)->sin6_addr;
3355 equal = IN6_ARE_ADDR_EQUAL(in6, &vxladdr->in6.sin6_addr);
3363 vxlan_sockaddr_in_copy(union vxlan_sockaddr *vxladdr,
3364 const struct sockaddr *sa)
3367 MPASS(sa->sa_family == AF_INET || sa->sa_family == AF_INET6);
3369 if (sa->sa_family == AF_INET) {
3370 const struct in_addr *in4 = &satoconstsin(sa)->sin_addr;
3371 vxladdr->in4.sin_family = AF_INET;
3372 vxladdr->in4.sin_len = sizeof(struct sockaddr_in);
3373 vxladdr->in4.sin_addr = *in4;
3374 } else if (sa->sa_family == AF_INET6) {
3375 const struct in6_addr *in6 = &satoconstsin6(sa)->sin6_addr;
3376 vxladdr->in6.sin6_family = AF_INET6;
3377 vxladdr->in6.sin6_len = sizeof(struct sockaddr_in6);
3378 vxladdr->in6.sin6_addr = *in6;
3383 vxlan_sockaddr_supported(const union vxlan_sockaddr *vxladdr, int unspec)
3385 const struct sockaddr *sa;
3391 if (sa->sa_family == AF_UNSPEC && unspec != 0) {
3393 } else if (sa->sa_family == AF_INET) {
3397 } else if (sa->sa_family == AF_INET6) {
3407 vxlan_sockaddr_in_any(const union vxlan_sockaddr *vxladdr)
3409 const struct sockaddr *sa;
3414 if (sa->sa_family == AF_INET) {
3415 const struct in_addr *in4 = &satoconstsin(sa)->sin_addr;
3416 any = in4->s_addr == INADDR_ANY;
3417 } else if (sa->sa_family == AF_INET6) {
3418 const struct in6_addr *in6 = &satoconstsin6(sa)->sin6_addr;
3419 any = IN6_IS_ADDR_UNSPECIFIED(in6);
3427 vxlan_sockaddr_in_multicast(const union vxlan_sockaddr *vxladdr)
3429 const struct sockaddr *sa;
3434 if (sa->sa_family == AF_INET) {
3435 const struct in_addr *in4 = &satoconstsin(sa)->sin_addr;
3436 mc = IN_MULTICAST(ntohl(in4->s_addr));
3437 } else if (sa->sa_family == AF_INET6) {
3438 const struct in6_addr *in6 = &satoconstsin6(sa)->sin6_addr;
3439 mc = IN6_IS_ADDR_MULTICAST(in6);
3447 vxlan_sockaddr_in6_embedscope(union vxlan_sockaddr *vxladdr)
3451 MPASS(VXLAN_SOCKADDR_IS_IPV6(vxladdr));
3453 error = sa6_embedscope(&vxladdr->in6, V_ip6_use_defzone);
3455 error = EAFNOSUPPORT;
3462 vxlan_can_change_config(struct vxlan_softc *sc)
3467 VXLAN_LOCK_ASSERT(sc);
3469 if (ifp->if_drv_flags & IFF_DRV_RUNNING)
3471 if (sc->vxl_flags & (VXLAN_FLAG_INIT | VXLAN_FLAG_TEARDOWN))
3478 vxlan_check_vni(uint32_t vni)
3481 return (vni >= VXLAN_VNI_MAX);
3485 vxlan_check_ttl(int ttl)
3488 return (ttl > MAXTTL);
3492 vxlan_check_ftable_timeout(uint32_t timeout)
3495 return (timeout > VXLAN_FTABLE_MAX_TIMEOUT);
3499 vxlan_check_ftable_max(uint32_t max)
3502 return (max > VXLAN_FTABLE_MAX);
3506 vxlan_sysctl_setup(struct vxlan_softc *sc)
3508 struct sysctl_ctx_list *ctx;
3509 struct sysctl_oid *node;
3510 struct vxlan_statistics *stats;
3513 ctx = &sc->vxl_sysctl_ctx;
3514 stats = &sc->vxl_stats;
3515 snprintf(namebuf, sizeof(namebuf), "%d", sc->vxl_unit);
3517 sysctl_ctx_init(ctx);
3518 sc->vxl_sysctl_node = SYSCTL_ADD_NODE(ctx,
3519 SYSCTL_STATIC_CHILDREN(_net_link_vxlan), OID_AUTO, namebuf,
3520 CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "");
3522 node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(sc->vxl_sysctl_node),
3523 OID_AUTO, "ftable", CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "");
3524 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "count",
3525 CTLFLAG_RD, &sc->vxl_ftable_cnt, 0,
3526 "Number of entries in fowarding table");
3527 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "max",
3528 CTLFLAG_RD, &sc->vxl_ftable_max, 0,
3529 "Maximum number of entries allowed in fowarding table");
3530 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "timeout",
3531 CTLFLAG_RD, &sc->vxl_ftable_timeout, 0,
3532 "Number of seconds between prunes of the forwarding table");
3533 SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "dump",
3534 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE | CTLFLAG_SKIP,
3535 sc, 0, vxlan_ftable_sysctl_dump, "A",
3536 "Dump the forwarding table entries");
3538 node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(sc->vxl_sysctl_node),
3539 OID_AUTO, "stats", CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "");
3540 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO,
3541 "ftable_nospace", CTLFLAG_RD, &stats->ftable_nospace, 0,
3542 "Fowarding table reached maximum entries");
3543 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO,
3544 "ftable_lock_upgrade_failed", CTLFLAG_RD,
3545 &stats->ftable_lock_upgrade_failed, 0,
3546 "Forwarding table update required lock upgrade");
3548 SYSCTL_ADD_COUNTER_U64(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "txcsum",
3549 CTLFLAG_RD, &stats->txcsum,
3550 "# of times hardware assisted with tx checksum");
3551 SYSCTL_ADD_COUNTER_U64(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "tso",
3552 CTLFLAG_RD, &stats->tso, "# of times hardware assisted with TSO");
3553 SYSCTL_ADD_COUNTER_U64(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "rxcsum",
3554 CTLFLAG_RD, &stats->rxcsum,
3555 "# of times hardware assisted with rx checksum");
3559 vxlan_sysctl_destroy(struct vxlan_softc *sc)
3562 sysctl_ctx_free(&sc->vxl_sysctl_ctx);
3563 sc->vxl_sysctl_node = NULL;
3567 vxlan_tunable_int(struct vxlan_softc *sc, const char *knob, int def)
3571 snprintf(path, sizeof(path), "net.link.vxlan.%d.%s",
3572 sc->vxl_unit, knob);
3573 TUNABLE_INT_FETCH(path, &def);
3579 vxlan_ifdetach_event(void *arg __unused, struct ifnet *ifp)
3581 struct vxlan_softc_head list;
3582 struct vxlan_socket *vso;
3583 struct vxlan_softc *sc, *tsc;
3587 if (ifp->if_flags & IFF_RENAMING)
3589 if ((ifp->if_flags & IFF_MULTICAST) == 0)
3593 LIST_FOREACH(vso, &vxlan_socket_list, vxlso_entry)
3594 vxlan_socket_ifdetach(vso, ifp, &list);
3595 VXLAN_LIST_UNLOCK();
3597 LIST_FOREACH_SAFE(sc, &list, vxl_ifdetach_list, tsc) {
3598 LIST_REMOVE(sc, vxl_ifdetach_list);
3600 sx_xlock(&vxlan_sx);
3602 if (sc->vxl_flags & VXLAN_FLAG_INIT)
3603 vxlan_init_wait(sc);
3604 vxlan_teardown_locked(sc);
3605 sx_xunlock(&vxlan_sx);
3613 mtx_init(&vxlan_list_mtx, "vxlan list", NULL, MTX_DEF);
3614 LIST_INIT(&vxlan_socket_list);
3615 vxlan_ifdetach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
3616 vxlan_ifdetach_event, NULL, EVENTHANDLER_PRI_ANY);
3617 vxlan_cloner = if_clone_simple(vxlan_name, vxlan_clone_create,
3618 vxlan_clone_destroy, 0);
3625 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
3626 vxlan_ifdetach_event_tag);
3627 if_clone_detach(vxlan_cloner);
3628 mtx_destroy(&vxlan_list_mtx);
3629 MPASS(LIST_EMPTY(&vxlan_socket_list));
3633 vxlan_modevent(module_t mod, int type, void *unused)
3654 static moduledata_t vxlan_mod = {
3660 DECLARE_MODULE(if_vxlan, vxlan_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
3661 MODULE_VERSION(if_vxlan, 1);