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;
167 struct vxlan_socket *vxl_sock;
169 union vxlan_sockaddr vxl_src_addr;
170 union vxlan_sockaddr vxl_dst_addr;
172 #define VXLAN_FLAG_INIT 0x0001
173 #define VXLAN_FLAG_TEARDOWN 0x0002
174 #define VXLAN_FLAG_LEARN 0x0004
175 #define VXLAN_FLAG_USER_MTU 0x0008
177 uint32_t vxl_port_hash_key;
178 uint16_t vxl_min_port;
179 uint16_t vxl_max_port;
182 /* Lookup table from MAC address to forwarding entry. */
183 uint32_t vxl_ftable_cnt;
184 uint32_t vxl_ftable_max;
185 uint32_t vxl_ftable_timeout;
186 uint32_t vxl_ftable_hash_key;
187 struct vxlan_ftable_head *vxl_ftable;
189 /* Derived from vxl_dst_addr. */
190 struct vxlan_ftable_entry vxl_default_fe;
192 struct ip_moptions *vxl_im4o;
193 struct ip6_moptions *vxl_im6o;
195 struct rmlock vxl_lock;
196 volatile u_int vxl_refcnt;
199 int vxl_vso_mc_index;
200 struct vxlan_statistics vxl_stats;
201 struct sysctl_oid *vxl_sysctl_node;
202 struct sysctl_ctx_list vxl_sysctl_ctx;
203 struct callout vxl_callout;
204 struct ether_addr vxl_hwaddr;
206 struct ifnet *vxl_mc_ifp;
207 struct ifmedia vxl_media;
208 char vxl_mc_ifname[IFNAMSIZ];
209 LIST_ENTRY(vxlan_softc) vxl_entry;
210 LIST_ENTRY(vxlan_softc) vxl_ifdetach_list;
212 /* For rate limiting errors on the tx fast path. */
213 struct timeval err_time;
217 #define VXLAN_RLOCK(_sc, _p) rm_rlock(&(_sc)->vxl_lock, (_p))
218 #define VXLAN_RUNLOCK(_sc, _p) rm_runlock(&(_sc)->vxl_lock, (_p))
219 #define VXLAN_WLOCK(_sc) rm_wlock(&(_sc)->vxl_lock)
220 #define VXLAN_WUNLOCK(_sc) rm_wunlock(&(_sc)->vxl_lock)
221 #define VXLAN_LOCK_WOWNED(_sc) rm_wowned(&(_sc)->vxl_lock)
222 #define VXLAN_LOCK_ASSERT(_sc) rm_assert(&(_sc)->vxl_lock, RA_LOCKED)
223 #define VXLAN_LOCK_WASSERT(_sc) rm_assert(&(_sc)->vxl_lock, RA_WLOCKED)
224 #define VXLAN_UNLOCK(_sc, _p) do { \
225 if (VXLAN_LOCK_WOWNED(_sc)) \
226 VXLAN_WUNLOCK(_sc); \
228 VXLAN_RUNLOCK(_sc, _p); \
231 #define VXLAN_ACQUIRE(_sc) refcount_acquire(&(_sc)->vxl_refcnt)
232 #define VXLAN_RELEASE(_sc) refcount_release(&(_sc)->vxl_refcnt)
234 #define satoconstsin(sa) ((const struct sockaddr_in *)(sa))
235 #define satoconstsin6(sa) ((const struct sockaddr_in6 *)(sa))
238 struct udphdr vxlh_udp;
239 struct vxlan_header vxlh_hdr;
242 static int vxlan_ftable_addr_cmp(const uint8_t *, const uint8_t *);
243 static void vxlan_ftable_init(struct vxlan_softc *);
244 static void vxlan_ftable_fini(struct vxlan_softc *);
245 static void vxlan_ftable_flush(struct vxlan_softc *, int);
246 static void vxlan_ftable_expire(struct vxlan_softc *);
247 static int vxlan_ftable_update_locked(struct vxlan_softc *,
248 const union vxlan_sockaddr *, const uint8_t *,
249 struct rm_priotracker *);
250 static int vxlan_ftable_learn(struct vxlan_softc *,
251 const struct sockaddr *, const uint8_t *);
252 static int vxlan_ftable_sysctl_dump(SYSCTL_HANDLER_ARGS);
254 static struct vxlan_ftable_entry *
255 vxlan_ftable_entry_alloc(void);
256 static void vxlan_ftable_entry_free(struct vxlan_ftable_entry *);
257 static void vxlan_ftable_entry_init(struct vxlan_softc *,
258 struct vxlan_ftable_entry *, const uint8_t *,
259 const struct sockaddr *, uint32_t);
260 static void vxlan_ftable_entry_destroy(struct vxlan_softc *,
261 struct vxlan_ftable_entry *);
262 static int vxlan_ftable_entry_insert(struct vxlan_softc *,
263 struct vxlan_ftable_entry *);
264 static struct vxlan_ftable_entry *
265 vxlan_ftable_entry_lookup(struct vxlan_softc *,
267 static void vxlan_ftable_entry_dump(struct vxlan_ftable_entry *,
270 static struct vxlan_socket *
271 vxlan_socket_alloc(const union vxlan_sockaddr *);
272 static void vxlan_socket_destroy(struct vxlan_socket *);
273 static void vxlan_socket_release(struct vxlan_socket *);
274 static struct vxlan_socket *
275 vxlan_socket_lookup(union vxlan_sockaddr *vxlsa);
276 static void vxlan_socket_insert(struct vxlan_socket *);
277 static int vxlan_socket_init(struct vxlan_socket *, struct ifnet *);
278 static int vxlan_socket_bind(struct vxlan_socket *, struct ifnet *);
279 static int vxlan_socket_create(struct ifnet *, int,
280 const union vxlan_sockaddr *, struct vxlan_socket **);
281 static void vxlan_socket_ifdetach(struct vxlan_socket *,
282 struct ifnet *, struct vxlan_softc_head *);
284 static struct vxlan_socket *
285 vxlan_socket_mc_lookup(const union vxlan_sockaddr *);
286 static int vxlan_sockaddr_mc_info_match(
287 const struct vxlan_socket_mc_info *,
288 const union vxlan_sockaddr *,
289 const union vxlan_sockaddr *, int);
290 static int vxlan_socket_mc_join_group(struct vxlan_socket *,
291 const union vxlan_sockaddr *, const union vxlan_sockaddr *,
292 int *, union vxlan_sockaddr *);
293 static int vxlan_socket_mc_leave_group(struct vxlan_socket *,
294 const union vxlan_sockaddr *,
295 const union vxlan_sockaddr *, int);
296 static int vxlan_socket_mc_add_group(struct vxlan_socket *,
297 const union vxlan_sockaddr *, const union vxlan_sockaddr *,
299 static void vxlan_socket_mc_release_group_by_idx(struct vxlan_socket *,
302 static struct vxlan_softc *
303 vxlan_socket_lookup_softc_locked(struct vxlan_socket *,
305 static struct vxlan_softc *
306 vxlan_socket_lookup_softc(struct vxlan_socket *, uint32_t);
307 static int vxlan_socket_insert_softc(struct vxlan_socket *,
308 struct vxlan_softc *);
309 static void vxlan_socket_remove_softc(struct vxlan_socket *,
310 struct vxlan_softc *);
312 static struct ifnet *
313 vxlan_multicast_if_ref(struct vxlan_softc *, int);
314 static void vxlan_free_multicast(struct vxlan_softc *);
315 static int vxlan_setup_multicast_interface(struct vxlan_softc *);
317 static int vxlan_setup_multicast(struct vxlan_softc *);
318 static int vxlan_setup_socket(struct vxlan_softc *);
320 static void vxlan_setup_zero_checksum_port(struct vxlan_softc *);
322 static void vxlan_setup_interface_hdrlen(struct vxlan_softc *);
323 static int vxlan_valid_init_config(struct vxlan_softc *);
324 static void vxlan_init_wait(struct vxlan_softc *);
325 static void vxlan_init_complete(struct vxlan_softc *);
326 static void vxlan_init(void *);
327 static void vxlan_release(struct vxlan_softc *);
328 static void vxlan_teardown_wait(struct vxlan_softc *);
329 static void vxlan_teardown_complete(struct vxlan_softc *);
330 static void vxlan_teardown_locked(struct vxlan_softc *);
331 static void vxlan_teardown(struct vxlan_softc *);
332 static void vxlan_ifdetach(struct vxlan_softc *, struct ifnet *,
333 struct vxlan_softc_head *);
334 static void vxlan_timer(void *);
336 static int vxlan_ctrl_get_config(struct vxlan_softc *, void *);
337 static int vxlan_ctrl_set_vni(struct vxlan_softc *, void *);
338 static int vxlan_ctrl_set_local_addr(struct vxlan_softc *, void *);
339 static int vxlan_ctrl_set_remote_addr(struct vxlan_softc *, void *);
340 static int vxlan_ctrl_set_local_port(struct vxlan_softc *, void *);
341 static int vxlan_ctrl_set_remote_port(struct vxlan_softc *, void *);
342 static int vxlan_ctrl_set_port_range(struct vxlan_softc *, void *);
343 static int vxlan_ctrl_set_ftable_timeout(struct vxlan_softc *, void *);
344 static int vxlan_ctrl_set_ftable_max(struct vxlan_softc *, void *);
345 static int vxlan_ctrl_set_multicast_if(struct vxlan_softc * , void *);
346 static int vxlan_ctrl_set_ttl(struct vxlan_softc *, void *);
347 static int vxlan_ctrl_set_learn(struct vxlan_softc *, void *);
348 static int vxlan_ctrl_ftable_entry_add(struct vxlan_softc *, void *);
349 static int vxlan_ctrl_ftable_entry_rem(struct vxlan_softc *, void *);
350 static int vxlan_ctrl_flush(struct vxlan_softc *, void *);
351 static int vxlan_ioctl_drvspec(struct vxlan_softc *,
352 struct ifdrv *, int);
353 static int vxlan_ioctl_ifflags(struct vxlan_softc *);
354 static int vxlan_ioctl(struct ifnet *, u_long, caddr_t);
356 #if defined(INET) || defined(INET6)
357 static uint16_t vxlan_pick_source_port(struct vxlan_softc *, struct mbuf *);
358 static void vxlan_encap_header(struct vxlan_softc *, struct mbuf *,
359 int, uint16_t, uint16_t);
361 static int vxlan_encap4(struct vxlan_softc *,
362 const union vxlan_sockaddr *, struct mbuf *);
363 static int vxlan_encap6(struct vxlan_softc *,
364 const union vxlan_sockaddr *, struct mbuf *);
365 static int vxlan_transmit(struct ifnet *, struct mbuf *);
366 static void vxlan_qflush(struct ifnet *);
367 static void vxlan_rcv_udp_packet(struct mbuf *, int, struct inpcb *,
368 const struct sockaddr *, void *);
369 static int vxlan_input(struct vxlan_socket *, uint32_t, struct mbuf **,
370 const struct sockaddr *);
372 static int vxlan_stats_alloc(struct vxlan_softc *);
373 static void vxlan_stats_free(struct vxlan_softc *);
374 static void vxlan_set_default_config(struct vxlan_softc *);
375 static int vxlan_set_user_config(struct vxlan_softc *,
376 struct ifvxlanparam *);
377 static int vxlan_set_reqcap(struct vxlan_softc *, struct ifnet *, int);
378 static void vxlan_set_hwcaps(struct vxlan_softc *);
379 static int vxlan_clone_create(struct if_clone *, int, caddr_t);
380 static void vxlan_clone_destroy(struct ifnet *);
382 static uint32_t vxlan_mac_hash(struct vxlan_softc *, const uint8_t *);
383 static int vxlan_media_change(struct ifnet *);
384 static void vxlan_media_status(struct ifnet *, struct ifmediareq *);
386 static int vxlan_sockaddr_cmp(const union vxlan_sockaddr *,
387 const struct sockaddr *);
388 static void vxlan_sockaddr_copy(union vxlan_sockaddr *,
389 const struct sockaddr *);
390 static int vxlan_sockaddr_in_equal(const union vxlan_sockaddr *,
391 const struct sockaddr *);
392 static void vxlan_sockaddr_in_copy(union vxlan_sockaddr *,
393 const struct sockaddr *);
394 static int vxlan_sockaddr_supported(const union vxlan_sockaddr *, int);
395 static int vxlan_sockaddr_in_any(const union vxlan_sockaddr *);
396 static int vxlan_sockaddr_in_multicast(const union vxlan_sockaddr *);
397 static int vxlan_sockaddr_in6_embedscope(union vxlan_sockaddr *);
399 static int vxlan_can_change_config(struct vxlan_softc *);
400 static int vxlan_check_vni(uint32_t);
401 static int vxlan_check_ttl(int);
402 static int vxlan_check_ftable_timeout(uint32_t);
403 static int vxlan_check_ftable_max(uint32_t);
405 static void vxlan_sysctl_setup(struct vxlan_softc *);
406 static void vxlan_sysctl_destroy(struct vxlan_softc *);
407 static int vxlan_tunable_int(struct vxlan_softc *, const char *, int);
409 static void vxlan_ifdetach_event(void *, struct ifnet *);
410 static void vxlan_load(void);
411 static void vxlan_unload(void);
412 static int vxlan_modevent(module_t, int, void *);
414 static const char vxlan_name[] = "vxlan";
415 static MALLOC_DEFINE(M_VXLAN, vxlan_name,
416 "Virtual eXtensible LAN Interface");
417 static struct if_clone *vxlan_cloner;
419 static struct mtx vxlan_list_mtx;
420 #define VXLAN_LIST_LOCK() mtx_lock(&vxlan_list_mtx)
421 #define VXLAN_LIST_UNLOCK() mtx_unlock(&vxlan_list_mtx)
423 static LIST_HEAD(, vxlan_socket) vxlan_socket_list;
425 static eventhandler_tag vxlan_ifdetach_event_tag;
427 SYSCTL_DECL(_net_link);
428 SYSCTL_NODE(_net_link, OID_AUTO, vxlan, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
429 "Virtual eXtensible Local Area Network");
431 static int vxlan_legacy_port = 0;
432 TUNABLE_INT("net.link.vxlan.legacy_port", &vxlan_legacy_port);
433 static int vxlan_reuse_port = 0;
434 TUNABLE_INT("net.link.vxlan.reuse_port", &vxlan_reuse_port);
436 /* Default maximum number of addresses in the forwarding table. */
437 #ifndef VXLAN_FTABLE_MAX
438 #define VXLAN_FTABLE_MAX 2000
441 /* Timeout (in seconds) of addresses learned in the forwarding table. */
442 #ifndef VXLAN_FTABLE_TIMEOUT
443 #define VXLAN_FTABLE_TIMEOUT (20 * 60)
447 * Maximum timeout (in seconds) of addresses learned in the forwarding
450 #ifndef VXLAN_FTABLE_MAX_TIMEOUT
451 #define VXLAN_FTABLE_MAX_TIMEOUT (60 * 60 * 24)
454 /* Number of seconds between pruning attempts of the forwarding table. */
455 #ifndef VXLAN_FTABLE_PRUNE
456 #define VXLAN_FTABLE_PRUNE (5 * 60)
459 static int vxlan_ftable_prune_period = VXLAN_FTABLE_PRUNE;
461 struct vxlan_control {
462 int (*vxlc_func)(struct vxlan_softc *, void *);
465 #define VXLAN_CTRL_FLAG_COPYIN 0x01
466 #define VXLAN_CTRL_FLAG_COPYOUT 0x02
467 #define VXLAN_CTRL_FLAG_SUSER 0x04
470 static const struct vxlan_control vxlan_control_table[] = {
471 [VXLAN_CMD_GET_CONFIG] =
472 { vxlan_ctrl_get_config, sizeof(struct ifvxlancfg),
473 VXLAN_CTRL_FLAG_COPYOUT
476 [VXLAN_CMD_SET_VNI] =
477 { vxlan_ctrl_set_vni, sizeof(struct ifvxlancmd),
478 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
481 [VXLAN_CMD_SET_LOCAL_ADDR] =
482 { vxlan_ctrl_set_local_addr, sizeof(struct ifvxlancmd),
483 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
486 [VXLAN_CMD_SET_REMOTE_ADDR] =
487 { vxlan_ctrl_set_remote_addr, sizeof(struct ifvxlancmd),
488 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
491 [VXLAN_CMD_SET_LOCAL_PORT] =
492 { vxlan_ctrl_set_local_port, sizeof(struct ifvxlancmd),
493 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
496 [VXLAN_CMD_SET_REMOTE_PORT] =
497 { vxlan_ctrl_set_remote_port, sizeof(struct ifvxlancmd),
498 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
501 [VXLAN_CMD_SET_PORT_RANGE] =
502 { vxlan_ctrl_set_port_range, sizeof(struct ifvxlancmd),
503 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
506 [VXLAN_CMD_SET_FTABLE_TIMEOUT] =
507 { vxlan_ctrl_set_ftable_timeout, sizeof(struct ifvxlancmd),
508 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
511 [VXLAN_CMD_SET_FTABLE_MAX] =
512 { vxlan_ctrl_set_ftable_max, sizeof(struct ifvxlancmd),
513 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
516 [VXLAN_CMD_SET_MULTICAST_IF] =
517 { vxlan_ctrl_set_multicast_if, sizeof(struct ifvxlancmd),
518 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
521 [VXLAN_CMD_SET_TTL] =
522 { vxlan_ctrl_set_ttl, sizeof(struct ifvxlancmd),
523 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
526 [VXLAN_CMD_SET_LEARN] =
527 { vxlan_ctrl_set_learn, sizeof(struct ifvxlancmd),
528 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
531 [VXLAN_CMD_FTABLE_ENTRY_ADD] =
532 { vxlan_ctrl_ftable_entry_add, sizeof(struct ifvxlancmd),
533 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
536 [VXLAN_CMD_FTABLE_ENTRY_REM] =
537 { vxlan_ctrl_ftable_entry_rem, sizeof(struct ifvxlancmd),
538 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
542 { vxlan_ctrl_flush, sizeof(struct ifvxlancmd),
543 VXLAN_CTRL_FLAG_COPYIN | VXLAN_CTRL_FLAG_SUSER,
547 static const int vxlan_control_table_size = nitems(vxlan_control_table);
550 vxlan_ftable_addr_cmp(const uint8_t *a, const uint8_t *b)
554 for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++)
555 d = ((int)a[i]) - ((int)b[i]);
561 vxlan_ftable_init(struct vxlan_softc *sc)
565 sc->vxl_ftable = malloc(sizeof(struct vxlan_ftable_head) *
566 VXLAN_SC_FTABLE_SIZE, M_VXLAN, M_ZERO | M_WAITOK);
568 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++)
569 LIST_INIT(&sc->vxl_ftable[i]);
570 sc->vxl_ftable_hash_key = arc4random();
574 vxlan_ftable_fini(struct vxlan_softc *sc)
578 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++) {
579 KASSERT(LIST_EMPTY(&sc->vxl_ftable[i]),
580 ("%s: vxlan %p ftable[%d] not empty", __func__, sc, i));
582 MPASS(sc->vxl_ftable_cnt == 0);
584 free(sc->vxl_ftable, M_VXLAN);
585 sc->vxl_ftable = NULL;
589 vxlan_ftable_flush(struct vxlan_softc *sc, int all)
591 struct vxlan_ftable_entry *fe, *tfe;
594 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++) {
595 LIST_FOREACH_SAFE(fe, &sc->vxl_ftable[i], vxlfe_hash, tfe) {
596 if (all || VXLAN_FE_IS_DYNAMIC(fe))
597 vxlan_ftable_entry_destroy(sc, fe);
603 vxlan_ftable_expire(struct vxlan_softc *sc)
605 struct vxlan_ftable_entry *fe, *tfe;
608 VXLAN_LOCK_WASSERT(sc);
610 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++) {
611 LIST_FOREACH_SAFE(fe, &sc->vxl_ftable[i], vxlfe_hash, tfe) {
612 if (VXLAN_FE_IS_DYNAMIC(fe) &&
613 time_uptime >= fe->vxlfe_expire)
614 vxlan_ftable_entry_destroy(sc, fe);
620 vxlan_ftable_update_locked(struct vxlan_softc *sc,
621 const union vxlan_sockaddr *vxlsa, const uint8_t *mac,
622 struct rm_priotracker *tracker)
624 struct vxlan_ftable_entry *fe;
627 VXLAN_LOCK_ASSERT(sc);
631 * A forwarding entry for this MAC address might already exist. If
632 * so, update it, otherwise create a new one. We may have to upgrade
633 * the lock if we have to change or create an entry.
635 fe = vxlan_ftable_entry_lookup(sc, mac);
637 fe->vxlfe_expire = time_uptime + sc->vxl_ftable_timeout;
639 if (!VXLAN_FE_IS_DYNAMIC(fe) ||
640 vxlan_sockaddr_in_equal(&fe->vxlfe_raddr, &vxlsa->sa))
642 if (!VXLAN_LOCK_WOWNED(sc)) {
643 VXLAN_RUNLOCK(sc, tracker);
645 sc->vxl_stats.ftable_lock_upgrade_failed++;
648 vxlan_sockaddr_in_copy(&fe->vxlfe_raddr, &vxlsa->sa);
652 if (!VXLAN_LOCK_WOWNED(sc)) {
653 VXLAN_RUNLOCK(sc, tracker);
655 sc->vxl_stats.ftable_lock_upgrade_failed++;
659 if (sc->vxl_ftable_cnt >= sc->vxl_ftable_max) {
660 sc->vxl_stats.ftable_nospace++;
664 fe = vxlan_ftable_entry_alloc();
668 vxlan_ftable_entry_init(sc, fe, mac, &vxlsa->sa, VXLAN_FE_FLAG_DYNAMIC);
670 /* The prior lookup failed, so the insert should not. */
671 error = vxlan_ftable_entry_insert(sc, fe);
678 vxlan_ftable_learn(struct vxlan_softc *sc, const struct sockaddr *sa,
681 struct rm_priotracker tracker;
682 union vxlan_sockaddr vxlsa;
686 * The source port may be randomly selected by the remote host, so
687 * use the port of the default destination address.
689 vxlan_sockaddr_copy(&vxlsa, sa);
690 vxlsa.in4.sin_port = sc->vxl_dst_addr.in4.sin_port;
692 if (VXLAN_SOCKADDR_IS_IPV6(&vxlsa)) {
693 error = vxlan_sockaddr_in6_embedscope(&vxlsa);
698 VXLAN_RLOCK(sc, &tracker);
699 error = vxlan_ftable_update_locked(sc, &vxlsa, mac, &tracker);
700 VXLAN_UNLOCK(sc, &tracker);
706 vxlan_ftable_sysctl_dump(SYSCTL_HANDLER_ARGS)
708 struct rm_priotracker tracker;
710 struct vxlan_softc *sc;
711 struct vxlan_ftable_entry *fe;
716 * This is mostly intended for debugging during development. It is
717 * not practical to dump an entire large table this way.
721 size = PAGE_SIZE; /* Calculate later. */
723 sbuf_new(&sb, NULL, size, SBUF_FIXEDLEN);
724 sbuf_putc(&sb, '\n');
726 VXLAN_RLOCK(sc, &tracker);
727 for (i = 0; i < VXLAN_SC_FTABLE_SIZE; i++) {
728 LIST_FOREACH(fe, &sc->vxl_ftable[i], vxlfe_hash) {
729 if (sbuf_error(&sb) != 0)
731 vxlan_ftable_entry_dump(fe, &sb);
734 VXLAN_RUNLOCK(sc, &tracker);
736 if (sbuf_len(&sb) == 1)
740 error = sysctl_handle_string(oidp, sbuf_data(&sb), sbuf_len(&sb), req);
746 static struct vxlan_ftable_entry *
747 vxlan_ftable_entry_alloc(void)
749 struct vxlan_ftable_entry *fe;
751 fe = malloc(sizeof(*fe), M_VXLAN, M_ZERO | M_NOWAIT);
757 vxlan_ftable_entry_free(struct vxlan_ftable_entry *fe)
764 vxlan_ftable_entry_init(struct vxlan_softc *sc, struct vxlan_ftable_entry *fe,
765 const uint8_t *mac, const struct sockaddr *sa, uint32_t flags)
768 fe->vxlfe_flags = flags;
769 fe->vxlfe_expire = time_uptime + sc->vxl_ftable_timeout;
770 memcpy(fe->vxlfe_mac, mac, ETHER_ADDR_LEN);
771 vxlan_sockaddr_copy(&fe->vxlfe_raddr, sa);
775 vxlan_ftable_entry_destroy(struct vxlan_softc *sc,
776 struct vxlan_ftable_entry *fe)
779 sc->vxl_ftable_cnt--;
780 LIST_REMOVE(fe, vxlfe_hash);
781 vxlan_ftable_entry_free(fe);
785 vxlan_ftable_entry_insert(struct vxlan_softc *sc,
786 struct vxlan_ftable_entry *fe)
788 struct vxlan_ftable_entry *lfe;
792 VXLAN_LOCK_WASSERT(sc);
793 hash = VXLAN_SC_FTABLE_HASH(sc, fe->vxlfe_mac);
795 lfe = LIST_FIRST(&sc->vxl_ftable[hash]);
797 LIST_INSERT_HEAD(&sc->vxl_ftable[hash], fe, vxlfe_hash);
802 dir = vxlan_ftable_addr_cmp(fe->vxlfe_mac, lfe->vxlfe_mac);
806 LIST_INSERT_BEFORE(lfe, fe, vxlfe_hash);
808 } else if (LIST_NEXT(lfe, vxlfe_hash) == NULL) {
809 LIST_INSERT_AFTER(lfe, fe, vxlfe_hash);
812 lfe = LIST_NEXT(lfe, vxlfe_hash);
813 } while (lfe != NULL);
816 sc->vxl_ftable_cnt++;
821 static struct vxlan_ftable_entry *
822 vxlan_ftable_entry_lookup(struct vxlan_softc *sc, const uint8_t *mac)
824 struct vxlan_ftable_entry *fe;
828 VXLAN_LOCK_ASSERT(sc);
829 hash = VXLAN_SC_FTABLE_HASH(sc, mac);
831 LIST_FOREACH(fe, &sc->vxl_ftable[hash], vxlfe_hash) {
832 dir = vxlan_ftable_addr_cmp(mac, fe->vxlfe_mac);
843 vxlan_ftable_entry_dump(struct vxlan_ftable_entry *fe, struct sbuf *sb)
846 const union vxlan_sockaddr *sa;
848 int i, len, af, width;
850 sa = &fe->vxlfe_raddr;
851 af = sa->sa.sa_family;
854 sbuf_printf(sb, "%c 0x%02X ", VXLAN_FE_IS_DYNAMIC(fe) ? 'D' : 'S',
857 for (i = 0; i < ETHER_ADDR_LEN - 1; i++)
858 sbuf_printf(sb, "%02X:", fe->vxlfe_mac[i]);
859 sbuf_printf(sb, "%02X ", fe->vxlfe_mac[i]);
862 addr = &sa->in4.sin_addr;
863 width = INET_ADDRSTRLEN - 1;
865 addr = &sa->in6.sin6_addr;
866 width = INET6_ADDRSTRLEN - 1;
868 inet_ntop(af, addr, buf, sizeof(buf));
869 sbuf_printf(sb, "%*s ", width, buf);
871 sbuf_printf(sb, "%08jd", (intmax_t)fe->vxlfe_expire);
875 /* Truncate a partial line. */
876 if (sbuf_error(sb) != 0)
877 sbuf_setpos(sb, len);
880 static struct vxlan_socket *
881 vxlan_socket_alloc(const union vxlan_sockaddr *sa)
883 struct vxlan_socket *vso;
886 vso = malloc(sizeof(*vso), M_VXLAN, M_WAITOK | M_ZERO);
887 rm_init(&vso->vxlso_lock, "vxlansorm");
888 refcount_init(&vso->vxlso_refcnt, 0);
889 for (i = 0; i < VXLAN_SO_VNI_HASH_SIZE; i++)
890 LIST_INIT(&vso->vxlso_vni_hash[i]);
891 vso->vxlso_laddr = *sa;
897 vxlan_socket_destroy(struct vxlan_socket *vso)
902 struct vxlan_socket_mc_info *mc;
904 for (i = 0; i < VXLAN_SO_MC_MAX_GROUPS; i++) {
905 mc = &vso->vxlso_mc[i];
906 KASSERT(mc->vxlsomc_gaddr.sa.sa_family == AF_UNSPEC,
907 ("%s: socket %p mc[%d] still has address",
911 for (i = 0; i < VXLAN_SO_VNI_HASH_SIZE; i++) {
912 KASSERT(LIST_EMPTY(&vso->vxlso_vni_hash[i]),
913 ("%s: socket %p vni_hash[%d] not empty",
917 so = vso->vxlso_sock;
919 vso->vxlso_sock = NULL;
923 rm_destroy(&vso->vxlso_lock);
928 vxlan_socket_release(struct vxlan_socket *vso)
933 destroy = VXLAN_SO_RELEASE(vso);
935 LIST_REMOVE(vso, vxlso_entry);
939 vxlan_socket_destroy(vso);
942 static struct vxlan_socket *
943 vxlan_socket_lookup(union vxlan_sockaddr *vxlsa)
945 struct vxlan_socket *vso;
948 LIST_FOREACH(vso, &vxlan_socket_list, vxlso_entry) {
949 if (vxlan_sockaddr_cmp(&vso->vxlso_laddr, &vxlsa->sa) == 0) {
950 VXLAN_SO_ACQUIRE(vso);
960 vxlan_socket_insert(struct vxlan_socket *vso)
964 VXLAN_SO_ACQUIRE(vso);
965 LIST_INSERT_HEAD(&vxlan_socket_list, vso, vxlso_entry);
970 vxlan_socket_init(struct vxlan_socket *vso, struct ifnet *ifp)
977 error = socreate(vso->vxlso_laddr.sa.sa_family, &vso->vxlso_sock,
978 SOCK_DGRAM, IPPROTO_UDP, td->td_ucred, td);
980 if_printf(ifp, "cannot create socket: %d\n", error);
984 error = udp_set_kernel_tunneling(vso->vxlso_sock,
985 vxlan_rcv_udp_packet, NULL, vso);
987 if_printf(ifp, "cannot set tunneling function: %d\n", error);
991 if (vxlan_reuse_port != 0) {
995 bzero(&sopt, sizeof(sopt));
996 sopt.sopt_dir = SOPT_SET;
997 sopt.sopt_level = IPPROTO_IP;
998 sopt.sopt_name = SO_REUSEPORT;
999 sopt.sopt_val = &val;
1000 sopt.sopt_valsize = sizeof(val);
1001 error = sosetopt(vso->vxlso_sock, &sopt);
1004 "cannot set REUSEADDR socket opt: %d\n", error);
1013 vxlan_socket_bind(struct vxlan_socket *vso, struct ifnet *ifp)
1015 union vxlan_sockaddr laddr;
1020 laddr = vso->vxlso_laddr;
1022 error = sobind(vso->vxlso_sock, &laddr.sa, td);
1024 if (error != EADDRINUSE)
1025 if_printf(ifp, "cannot bind socket: %d\n", error);
1033 vxlan_socket_create(struct ifnet *ifp, int multicast,
1034 const union vxlan_sockaddr *saddr, struct vxlan_socket **vsop)
1036 union vxlan_sockaddr laddr;
1037 struct vxlan_socket *vso;
1043 * If this socket will be multicast, then only the local port
1044 * must be specified when binding.
1046 if (multicast != 0) {
1047 if (VXLAN_SOCKADDR_IS_IPV4(&laddr))
1048 laddr.in4.sin_addr.s_addr = INADDR_ANY;
1051 laddr.in6.sin6_addr = in6addr_any;
1055 vso = vxlan_socket_alloc(&laddr);
1059 error = vxlan_socket_init(vso, ifp);
1063 error = vxlan_socket_bind(vso, ifp);
1068 * There is a small window between the bind completing and
1069 * inserting the socket, so that a concurrent create may fail.
1070 * Let's not worry about that for now.
1072 vxlan_socket_insert(vso);
1078 vxlan_socket_destroy(vso);
1084 vxlan_socket_ifdetach(struct vxlan_socket *vso, struct ifnet *ifp,
1085 struct vxlan_softc_head *list)
1087 struct rm_priotracker tracker;
1088 struct vxlan_softc *sc;
1091 VXLAN_SO_RLOCK(vso, &tracker);
1092 for (i = 0; i < VXLAN_SO_VNI_HASH_SIZE; i++) {
1093 LIST_FOREACH(sc, &vso->vxlso_vni_hash[i], vxl_entry)
1094 vxlan_ifdetach(sc, ifp, list);
1096 VXLAN_SO_RUNLOCK(vso, &tracker);
1099 static struct vxlan_socket *
1100 vxlan_socket_mc_lookup(const union vxlan_sockaddr *vxlsa)
1102 union vxlan_sockaddr laddr;
1103 struct vxlan_socket *vso;
1107 if (VXLAN_SOCKADDR_IS_IPV4(&laddr))
1108 laddr.in4.sin_addr.s_addr = INADDR_ANY;
1111 laddr.in6.sin6_addr = in6addr_any;
1114 vso = vxlan_socket_lookup(&laddr);
1120 vxlan_sockaddr_mc_info_match(const struct vxlan_socket_mc_info *mc,
1121 const union vxlan_sockaddr *group, const union vxlan_sockaddr *local,
1125 if (!vxlan_sockaddr_in_any(local) &&
1126 !vxlan_sockaddr_in_equal(&mc->vxlsomc_saddr, &local->sa))
1128 if (!vxlan_sockaddr_in_equal(&mc->vxlsomc_gaddr, &group->sa))
1130 if (ifidx != 0 && ifidx != mc->vxlsomc_ifidx)
1137 vxlan_socket_mc_join_group(struct vxlan_socket *vso,
1138 const union vxlan_sockaddr *group, const union vxlan_sockaddr *local,
1139 int *ifidx, union vxlan_sockaddr *source)
1141 struct sockopt sopt;
1146 if (VXLAN_SOCKADDR_IS_IPV4(group)) {
1147 struct ip_mreq mreq;
1149 mreq.imr_multiaddr = group->in4.sin_addr;
1150 mreq.imr_interface = local->in4.sin_addr;
1152 bzero(&sopt, sizeof(sopt));
1153 sopt.sopt_dir = SOPT_SET;
1154 sopt.sopt_level = IPPROTO_IP;
1155 sopt.sopt_name = IP_ADD_MEMBERSHIP;
1156 sopt.sopt_val = &mreq;
1157 sopt.sopt_valsize = sizeof(mreq);
1158 error = sosetopt(vso->vxlso_sock, &sopt);
1163 * BMV: Ideally, there would be a formal way for us to get
1164 * the local interface that was selected based on the
1165 * imr_interface address. We could then update *ifidx so
1166 * vxlan_sockaddr_mc_info_match() would return a match for
1167 * later creates that explicitly set the multicast interface.
1169 * If we really need to, we can of course look in the INP's
1171 * sotoinpcb(vso->vxlso_sock)->inp_moptions->
1172 * imo_head[]->imf_inm->inm_ifp
1173 * similarly to imo_match_group().
1175 source->in4.sin_addr = local->in4.sin_addr;
1177 } else if (VXLAN_SOCKADDR_IS_IPV6(group)) {
1178 struct ipv6_mreq mreq;
1180 mreq.ipv6mr_multiaddr = group->in6.sin6_addr;
1181 mreq.ipv6mr_interface = *ifidx;
1183 bzero(&sopt, sizeof(sopt));
1184 sopt.sopt_dir = SOPT_SET;
1185 sopt.sopt_level = IPPROTO_IPV6;
1186 sopt.sopt_name = IPV6_JOIN_GROUP;
1187 sopt.sopt_val = &mreq;
1188 sopt.sopt_valsize = sizeof(mreq);
1189 error = sosetopt(vso->vxlso_sock, &sopt);
1194 * BMV: As with IPv4, we would really like to know what
1195 * interface in6p_lookup_mcast_ifp() selected.
1198 error = EAFNOSUPPORT;
1204 vxlan_socket_mc_leave_group(struct vxlan_socket *vso,
1205 const union vxlan_sockaddr *group, const union vxlan_sockaddr *source,
1208 struct sockopt sopt;
1211 bzero(&sopt, sizeof(sopt));
1212 sopt.sopt_dir = SOPT_SET;
1214 if (VXLAN_SOCKADDR_IS_IPV4(group)) {
1215 struct ip_mreq mreq;
1217 mreq.imr_multiaddr = group->in4.sin_addr;
1218 mreq.imr_interface = source->in4.sin_addr;
1220 sopt.sopt_level = IPPROTO_IP;
1221 sopt.sopt_name = IP_DROP_MEMBERSHIP;
1222 sopt.sopt_val = &mreq;
1223 sopt.sopt_valsize = sizeof(mreq);
1224 error = sosetopt(vso->vxlso_sock, &sopt);
1226 } else if (VXLAN_SOCKADDR_IS_IPV6(group)) {
1227 struct ipv6_mreq mreq;
1229 mreq.ipv6mr_multiaddr = group->in6.sin6_addr;
1230 mreq.ipv6mr_interface = ifidx;
1232 sopt.sopt_level = IPPROTO_IPV6;
1233 sopt.sopt_name = IPV6_LEAVE_GROUP;
1234 sopt.sopt_val = &mreq;
1235 sopt.sopt_valsize = sizeof(mreq);
1236 error = sosetopt(vso->vxlso_sock, &sopt);
1239 error = EAFNOSUPPORT;
1245 vxlan_socket_mc_add_group(struct vxlan_socket *vso,
1246 const union vxlan_sockaddr *group, const union vxlan_sockaddr *local,
1247 int ifidx, int *idx)
1249 union vxlan_sockaddr source;
1250 struct vxlan_socket_mc_info *mc;
1251 int i, empty, error;
1254 * Within a socket, the same multicast group may be used by multiple
1255 * interfaces, each with a different network identifier. But a socket
1256 * may only join a multicast group once, so keep track of the users
1260 VXLAN_SO_WLOCK(vso);
1261 for (empty = 0, i = 0; i < VXLAN_SO_MC_MAX_GROUPS; i++) {
1262 mc = &vso->vxlso_mc[i];
1264 if (mc->vxlsomc_gaddr.sa.sa_family == AF_UNSPEC) {
1269 if (vxlan_sockaddr_mc_info_match(mc, group, local, ifidx))
1272 VXLAN_SO_WUNLOCK(vso);
1277 error = vxlan_socket_mc_join_group(vso, group, local, &ifidx, &source);
1281 VXLAN_SO_WLOCK(vso);
1282 for (i = 0; i < VXLAN_SO_MC_MAX_GROUPS; i++) {
1283 mc = &vso->vxlso_mc[i];
1285 if (mc->vxlsomc_gaddr.sa.sa_family == AF_UNSPEC) {
1286 vxlan_sockaddr_copy(&mc->vxlsomc_gaddr, &group->sa);
1287 vxlan_sockaddr_copy(&mc->vxlsomc_saddr, &source.sa);
1288 mc->vxlsomc_ifidx = ifidx;
1292 VXLAN_SO_WUNLOCK(vso);
1294 error = vxlan_socket_mc_leave_group(vso, group, &source, ifidx);
1300 mc->vxlsomc_users++;
1301 VXLAN_SO_WUNLOCK(vso);
1309 vxlan_socket_mc_release_group_by_idx(struct vxlan_socket *vso, int idx)
1311 union vxlan_sockaddr group, source;
1312 struct vxlan_socket_mc_info *mc;
1315 KASSERT(idx >= 0 && idx < VXLAN_SO_MC_MAX_GROUPS,
1316 ("%s: vso %p idx %d out of bounds", __func__, vso, idx));
1319 mc = &vso->vxlso_mc[idx];
1321 VXLAN_SO_WLOCK(vso);
1322 mc->vxlsomc_users--;
1323 if (mc->vxlsomc_users == 0) {
1324 group = mc->vxlsomc_gaddr;
1325 source = mc->vxlsomc_saddr;
1326 ifidx = mc->vxlsomc_ifidx;
1327 bzero(mc, sizeof(*mc));
1330 VXLAN_SO_WUNLOCK(vso);
1334 * Our socket's membership in this group may have already
1335 * been removed if we joined through an interface that's
1338 vxlan_socket_mc_leave_group(vso, &group, &source, ifidx);
1342 static struct vxlan_softc *
1343 vxlan_socket_lookup_softc_locked(struct vxlan_socket *vso, uint32_t vni)
1345 struct vxlan_softc *sc;
1348 VXLAN_SO_LOCK_ASSERT(vso);
1349 hash = VXLAN_SO_VNI_HASH(vni);
1351 LIST_FOREACH(sc, &vso->vxlso_vni_hash[hash], vxl_entry) {
1352 if (sc->vxl_vni == vni) {
1361 static struct vxlan_softc *
1362 vxlan_socket_lookup_softc(struct vxlan_socket *vso, uint32_t vni)
1364 struct rm_priotracker tracker;
1365 struct vxlan_softc *sc;
1367 VXLAN_SO_RLOCK(vso, &tracker);
1368 sc = vxlan_socket_lookup_softc_locked(vso, vni);
1369 VXLAN_SO_RUNLOCK(vso, &tracker);
1375 vxlan_socket_insert_softc(struct vxlan_socket *vso, struct vxlan_softc *sc)
1377 struct vxlan_softc *tsc;
1381 hash = VXLAN_SO_VNI_HASH(vni);
1383 VXLAN_SO_WLOCK(vso);
1384 tsc = vxlan_socket_lookup_softc_locked(vso, vni);
1386 VXLAN_SO_WUNLOCK(vso);
1392 LIST_INSERT_HEAD(&vso->vxlso_vni_hash[hash], sc, vxl_entry);
1393 VXLAN_SO_WUNLOCK(vso);
1399 vxlan_socket_remove_softc(struct vxlan_socket *vso, struct vxlan_softc *sc)
1402 VXLAN_SO_WLOCK(vso);
1403 LIST_REMOVE(sc, vxl_entry);
1404 VXLAN_SO_WUNLOCK(vso);
1409 static struct ifnet *
1410 vxlan_multicast_if_ref(struct vxlan_softc *sc, int ipv4)
1414 VXLAN_LOCK_ASSERT(sc);
1416 if (ipv4 && sc->vxl_im4o != NULL)
1417 ifp = sc->vxl_im4o->imo_multicast_ifp;
1418 else if (!ipv4 && sc->vxl_im6o != NULL)
1419 ifp = sc->vxl_im6o->im6o_multicast_ifp;
1430 vxlan_free_multicast(struct vxlan_softc *sc)
1433 if (sc->vxl_mc_ifp != NULL) {
1434 if_rele(sc->vxl_mc_ifp);
1435 sc->vxl_mc_ifp = NULL;
1436 sc->vxl_mc_ifindex = 0;
1439 if (sc->vxl_im4o != NULL) {
1440 free(sc->vxl_im4o, M_VXLAN);
1441 sc->vxl_im4o = NULL;
1444 if (sc->vxl_im6o != NULL) {
1445 free(sc->vxl_im6o, M_VXLAN);
1446 sc->vxl_im6o = NULL;
1451 vxlan_setup_multicast_interface(struct vxlan_softc *sc)
1455 ifp = ifunit_ref(sc->vxl_mc_ifname);
1457 if_printf(sc->vxl_ifp, "multicast interface %s does "
1458 "not exist\n", sc->vxl_mc_ifname);
1462 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
1463 if_printf(sc->vxl_ifp, "interface %s does not support "
1464 "multicast\n", sc->vxl_mc_ifname);
1469 sc->vxl_mc_ifp = ifp;
1470 sc->vxl_mc_ifindex = ifp->if_index;
1476 vxlan_setup_multicast(struct vxlan_softc *sc)
1478 const union vxlan_sockaddr *group;
1481 group = &sc->vxl_dst_addr;
1484 if (sc->vxl_mc_ifname[0] != '\0') {
1485 error = vxlan_setup_multicast_interface(sc);
1491 * Initialize an multicast options structure that is sufficiently
1492 * populated for use in the respective IP output routine. This
1493 * structure is typically stored in the socket, but our sockets
1494 * may be shared among multiple interfaces.
1496 if (VXLAN_SOCKADDR_IS_IPV4(group)) {
1497 sc->vxl_im4o = malloc(sizeof(struct ip_moptions), M_VXLAN,
1499 sc->vxl_im4o->imo_multicast_ifp = sc->vxl_mc_ifp;
1500 sc->vxl_im4o->imo_multicast_ttl = sc->vxl_ttl;
1501 sc->vxl_im4o->imo_multicast_vif = -1;
1502 } else if (VXLAN_SOCKADDR_IS_IPV6(group)) {
1503 sc->vxl_im6o = malloc(sizeof(struct ip6_moptions), M_VXLAN,
1505 sc->vxl_im6o->im6o_multicast_ifp = sc->vxl_mc_ifp;
1506 sc->vxl_im6o->im6o_multicast_hlim = sc->vxl_ttl;
1513 vxlan_setup_socket(struct vxlan_softc *sc)
1515 struct vxlan_socket *vso;
1517 union vxlan_sockaddr *saddr, *daddr;
1518 int multicast, error;
1522 saddr = &sc->vxl_src_addr;
1523 daddr = &sc->vxl_dst_addr;
1525 multicast = vxlan_sockaddr_in_multicast(daddr);
1526 MPASS(multicast != -1);
1527 sc->vxl_vso_mc_index = -1;
1530 * Try to create the socket. If that fails, attempt to use an
1533 error = vxlan_socket_create(ifp, multicast, saddr, &vso);
1536 vso = vxlan_socket_mc_lookup(saddr);
1538 vso = vxlan_socket_lookup(saddr);
1541 if_printf(ifp, "cannot create socket (error: %d), "
1542 "and no existing socket found\n", error);
1547 if (multicast != 0) {
1548 error = vxlan_setup_multicast(sc);
1552 error = vxlan_socket_mc_add_group(vso, daddr, saddr,
1553 sc->vxl_mc_ifindex, &sc->vxl_vso_mc_index);
1559 error = vxlan_socket_insert_softc(vso, sc);
1561 sc->vxl_sock = NULL;
1562 if_printf(ifp, "network identifier %d already exists in "
1563 "this socket\n", sc->vxl_vni);
1571 if (sc->vxl_vso_mc_index != -1) {
1572 vxlan_socket_mc_release_group_by_idx(vso,
1573 sc->vxl_vso_mc_index);
1574 sc->vxl_vso_mc_index = -1;
1577 vxlan_free_multicast(sc);
1578 vxlan_socket_release(vso);
1586 vxlan_setup_zero_checksum_port(struct vxlan_softc *sc)
1589 if (!VXLAN_SOCKADDR_IS_IPV6(&sc->vxl_src_addr))
1592 MPASS(sc->vxl_src_addr.in6.sin6_port != 0);
1593 MPASS(sc->vxl_dst_addr.in6.sin6_port != 0);
1595 if (sc->vxl_src_addr.in6.sin6_port != sc->vxl_dst_addr.in6.sin6_port) {
1596 if_printf(sc->vxl_ifp, "port %d in src address does not match "
1597 "port %d in dst address, rfc6935_port (%d) not updated.\n",
1598 ntohs(sc->vxl_src_addr.in6.sin6_port),
1599 ntohs(sc->vxl_dst_addr.in6.sin6_port),
1600 V_zero_checksum_port);
1604 if (V_zero_checksum_port != 0) {
1605 if (V_zero_checksum_port !=
1606 ntohs(sc->vxl_src_addr.in6.sin6_port)) {
1607 if_printf(sc->vxl_ifp, "rfc6935_port is already set to "
1608 "%d, cannot set it to %d.\n", V_zero_checksum_port,
1609 ntohs(sc->vxl_src_addr.in6.sin6_port));
1614 V_zero_checksum_port = ntohs(sc->vxl_src_addr.in6.sin6_port);
1615 if_printf(sc->vxl_ifp, "rfc6935_port set to %d\n",
1616 V_zero_checksum_port);
1621 vxlan_setup_interface_hdrlen(struct vxlan_softc *sc)
1625 VXLAN_LOCK_WASSERT(sc);
1628 ifp->if_hdrlen = ETHER_HDR_LEN + sizeof(struct vxlanudphdr);
1630 if (VXLAN_SOCKADDR_IS_IPV4(&sc->vxl_dst_addr) != 0)
1631 ifp->if_hdrlen += sizeof(struct ip);
1632 else if (VXLAN_SOCKADDR_IS_IPV6(&sc->vxl_dst_addr) != 0)
1633 ifp->if_hdrlen += sizeof(struct ip6_hdr);
1635 if ((sc->vxl_flags & VXLAN_FLAG_USER_MTU) == 0)
1636 ifp->if_mtu = ETHERMTU - ifp->if_hdrlen;
1640 vxlan_valid_init_config(struct vxlan_softc *sc)
1644 if (vxlan_check_vni(sc->vxl_vni) != 0) {
1645 reason = "invalid virtual network identifier specified";
1649 if (vxlan_sockaddr_supported(&sc->vxl_src_addr, 1) == 0) {
1650 reason = "source address type is not supported";
1654 if (vxlan_sockaddr_supported(&sc->vxl_dst_addr, 0) == 0) {
1655 reason = "destination address type is not supported";
1659 if (vxlan_sockaddr_in_any(&sc->vxl_dst_addr) != 0) {
1660 reason = "no valid destination address specified";
1664 if (vxlan_sockaddr_in_multicast(&sc->vxl_dst_addr) == 0 &&
1665 sc->vxl_mc_ifname[0] != '\0') {
1666 reason = "can only specify interface with a group address";
1670 if (vxlan_sockaddr_in_any(&sc->vxl_src_addr) == 0) {
1671 if (VXLAN_SOCKADDR_IS_IPV4(&sc->vxl_src_addr) ^
1672 VXLAN_SOCKADDR_IS_IPV4(&sc->vxl_dst_addr)) {
1673 reason = "source and destination address must both "
1674 "be either IPv4 or IPv6";
1679 if (sc->vxl_src_addr.in4.sin_port == 0) {
1680 reason = "local port not specified";
1684 if (sc->vxl_dst_addr.in4.sin_port == 0) {
1685 reason = "remote port not specified";
1692 if_printf(sc->vxl_ifp, "cannot initialize interface: %s\n", reason);
1697 vxlan_init_wait(struct vxlan_softc *sc)
1700 VXLAN_LOCK_WASSERT(sc);
1701 while (sc->vxl_flags & VXLAN_FLAG_INIT)
1702 rm_sleep(sc, &sc->vxl_lock, 0, "vxlint", hz);
1706 vxlan_init_complete(struct vxlan_softc *sc)
1710 sc->vxl_flags &= ~VXLAN_FLAG_INIT;
1716 vxlan_init(void *xsc)
1718 static const uint8_t empty_mac[ETHER_ADDR_LEN];
1719 struct vxlan_softc *sc;
1725 sx_xlock(&vxlan_sx);
1727 if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
1729 sx_xunlock(&vxlan_sx);
1732 sc->vxl_flags |= VXLAN_FLAG_INIT;
1735 if (vxlan_valid_init_config(sc) != 0)
1738 if (vxlan_setup_socket(sc) != 0)
1742 vxlan_setup_zero_checksum_port(sc);
1745 /* Initialize the default forwarding entry. */
1746 vxlan_ftable_entry_init(sc, &sc->vxl_default_fe, empty_mac,
1747 &sc->vxl_dst_addr.sa, VXLAN_FE_FLAG_STATIC);
1750 ifp->if_drv_flags |= IFF_DRV_RUNNING;
1751 callout_reset(&sc->vxl_callout, vxlan_ftable_prune_period * hz,
1755 if_link_state_change(ifp, LINK_STATE_UP);
1757 EVENTHANDLER_INVOKE(vxlan_start, ifp, sc->vxl_src_addr.in4.sin_family,
1758 ntohs(sc->vxl_src_addr.in4.sin_port));
1760 vxlan_init_complete(sc);
1761 sx_xunlock(&vxlan_sx);
1765 vxlan_release(struct vxlan_softc *sc)
1769 * The softc may be destroyed as soon as we release our reference,
1770 * so we cannot serialize the wakeup with the softc lock. We use a
1771 * timeout in our sleeps so a missed wakeup is unfortunate but not
1774 if (VXLAN_RELEASE(sc) != 0)
1779 vxlan_teardown_wait(struct vxlan_softc *sc)
1782 VXLAN_LOCK_WASSERT(sc);
1783 while (sc->vxl_flags & VXLAN_FLAG_TEARDOWN)
1784 rm_sleep(sc, &sc->vxl_lock, 0, "vxltrn", hz);
1788 vxlan_teardown_complete(struct vxlan_softc *sc)
1792 sc->vxl_flags &= ~VXLAN_FLAG_TEARDOWN;
1798 vxlan_teardown_locked(struct vxlan_softc *sc)
1801 struct vxlan_socket *vso;
1803 sx_assert(&vxlan_sx, SA_XLOCKED);
1804 VXLAN_LOCK_WASSERT(sc);
1805 MPASS(sc->vxl_flags & VXLAN_FLAG_TEARDOWN);
1808 ifp->if_flags &= ~IFF_UP;
1809 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1810 callout_stop(&sc->vxl_callout);
1812 sc->vxl_sock = NULL;
1815 if_link_state_change(ifp, LINK_STATE_DOWN);
1816 EVENTHANDLER_INVOKE(vxlan_stop, ifp, sc->vxl_src_addr.in4.sin_family,
1817 ntohs(sc->vxl_src_addr.in4.sin_port));
1820 vxlan_socket_remove_softc(vso, sc);
1822 if (sc->vxl_vso_mc_index != -1) {
1823 vxlan_socket_mc_release_group_by_idx(vso,
1824 sc->vxl_vso_mc_index);
1825 sc->vxl_vso_mc_index = -1;
1830 while (sc->vxl_refcnt != 0)
1831 rm_sleep(sc, &sc->vxl_lock, 0, "vxldrn", hz);
1834 callout_drain(&sc->vxl_callout);
1836 vxlan_free_multicast(sc);
1838 vxlan_socket_release(vso);
1840 vxlan_teardown_complete(sc);
1844 vxlan_teardown(struct vxlan_softc *sc)
1847 sx_xlock(&vxlan_sx);
1849 if (sc->vxl_flags & VXLAN_FLAG_TEARDOWN) {
1850 vxlan_teardown_wait(sc);
1852 sx_xunlock(&vxlan_sx);
1856 sc->vxl_flags |= VXLAN_FLAG_TEARDOWN;
1857 vxlan_teardown_locked(sc);
1858 sx_xunlock(&vxlan_sx);
1862 vxlan_ifdetach(struct vxlan_softc *sc, struct ifnet *ifp,
1863 struct vxlan_softc_head *list)
1868 if (sc->vxl_mc_ifp != ifp)
1870 if (sc->vxl_flags & VXLAN_FLAG_TEARDOWN)
1873 sc->vxl_flags |= VXLAN_FLAG_TEARDOWN;
1874 LIST_INSERT_HEAD(list, sc, vxl_ifdetach_list);
1881 vxlan_timer(void *xsc)
1883 struct vxlan_softc *sc;
1886 VXLAN_LOCK_WASSERT(sc);
1888 vxlan_ftable_expire(sc);
1889 callout_schedule(&sc->vxl_callout, vxlan_ftable_prune_period * hz);
1893 vxlan_ioctl_ifflags(struct vxlan_softc *sc)
1899 if (ifp->if_flags & IFF_UP) {
1900 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1903 if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1911 vxlan_ctrl_get_config(struct vxlan_softc *sc, void *arg)
1913 struct rm_priotracker tracker;
1914 struct ifvxlancfg *cfg;
1917 bzero(cfg, sizeof(*cfg));
1919 VXLAN_RLOCK(sc, &tracker);
1920 cfg->vxlc_vni = sc->vxl_vni;
1921 memcpy(&cfg->vxlc_local_sa, &sc->vxl_src_addr,
1922 sizeof(union vxlan_sockaddr));
1923 memcpy(&cfg->vxlc_remote_sa, &sc->vxl_dst_addr,
1924 sizeof(union vxlan_sockaddr));
1925 cfg->vxlc_mc_ifindex = sc->vxl_mc_ifindex;
1926 cfg->vxlc_ftable_cnt = sc->vxl_ftable_cnt;
1927 cfg->vxlc_ftable_max = sc->vxl_ftable_max;
1928 cfg->vxlc_ftable_timeout = sc->vxl_ftable_timeout;
1929 cfg->vxlc_port_min = sc->vxl_min_port;
1930 cfg->vxlc_port_max = sc->vxl_max_port;
1931 cfg->vxlc_learn = (sc->vxl_flags & VXLAN_FLAG_LEARN) != 0;
1932 cfg->vxlc_ttl = sc->vxl_ttl;
1933 VXLAN_RUNLOCK(sc, &tracker);
1936 if (VXLAN_SOCKADDR_IS_IPV6(&cfg->vxlc_local_sa))
1937 sa6_recoverscope(&cfg->vxlc_local_sa.in6);
1938 if (VXLAN_SOCKADDR_IS_IPV6(&cfg->vxlc_remote_sa))
1939 sa6_recoverscope(&cfg->vxlc_remote_sa.in6);
1946 vxlan_ctrl_set_vni(struct vxlan_softc *sc, void *arg)
1948 struct ifvxlancmd *cmd;
1953 if (vxlan_check_vni(cmd->vxlcmd_vni) != 0)
1957 if (vxlan_can_change_config(sc)) {
1958 sc->vxl_vni = cmd->vxlcmd_vni;
1968 vxlan_ctrl_set_local_addr(struct vxlan_softc *sc, void *arg)
1970 struct ifvxlancmd *cmd;
1971 union vxlan_sockaddr *vxlsa;
1975 vxlsa = &cmd->vxlcmd_sa;
1977 if (!VXLAN_SOCKADDR_IS_IPV46(vxlsa))
1979 if (vxlan_sockaddr_in_multicast(vxlsa) != 0)
1981 if (VXLAN_SOCKADDR_IS_IPV6(vxlsa)) {
1982 error = vxlan_sockaddr_in6_embedscope(vxlsa);
1988 if (vxlan_can_change_config(sc)) {
1989 vxlan_sockaddr_in_copy(&sc->vxl_src_addr, &vxlsa->sa);
1990 vxlan_set_hwcaps(sc);
2000 vxlan_ctrl_set_remote_addr(struct vxlan_softc *sc, void *arg)
2002 struct ifvxlancmd *cmd;
2003 union vxlan_sockaddr *vxlsa;
2007 vxlsa = &cmd->vxlcmd_sa;
2009 if (!VXLAN_SOCKADDR_IS_IPV46(vxlsa))
2011 if (VXLAN_SOCKADDR_IS_IPV6(vxlsa)) {
2012 error = vxlan_sockaddr_in6_embedscope(vxlsa);
2018 if (vxlan_can_change_config(sc)) {
2019 vxlan_sockaddr_in_copy(&sc->vxl_dst_addr, &vxlsa->sa);
2020 vxlan_setup_interface_hdrlen(sc);
2030 vxlan_ctrl_set_local_port(struct vxlan_softc *sc, void *arg)
2032 struct ifvxlancmd *cmd;
2037 if (cmd->vxlcmd_port == 0)
2041 if (vxlan_can_change_config(sc)) {
2042 sc->vxl_src_addr.in4.sin_port = htons(cmd->vxlcmd_port);
2052 vxlan_ctrl_set_remote_port(struct vxlan_softc *sc, void *arg)
2054 struct ifvxlancmd *cmd;
2059 if (cmd->vxlcmd_port == 0)
2063 if (vxlan_can_change_config(sc)) {
2064 sc->vxl_dst_addr.in4.sin_port = htons(cmd->vxlcmd_port);
2074 vxlan_ctrl_set_port_range(struct vxlan_softc *sc, void *arg)
2076 struct ifvxlancmd *cmd;
2081 min = cmd->vxlcmd_port_min;
2082 max = cmd->vxlcmd_port_max;
2088 if (vxlan_can_change_config(sc)) {
2089 sc->vxl_min_port = min;
2090 sc->vxl_max_port = max;
2100 vxlan_ctrl_set_ftable_timeout(struct vxlan_softc *sc, void *arg)
2102 struct ifvxlancmd *cmd;
2108 if (vxlan_check_ftable_timeout(cmd->vxlcmd_ftable_timeout) == 0) {
2109 sc->vxl_ftable_timeout = cmd->vxlcmd_ftable_timeout;
2119 vxlan_ctrl_set_ftable_max(struct vxlan_softc *sc, void *arg)
2121 struct ifvxlancmd *cmd;
2127 if (vxlan_check_ftable_max(cmd->vxlcmd_ftable_max) == 0) {
2128 sc->vxl_ftable_max = cmd->vxlcmd_ftable_max;
2138 vxlan_ctrl_set_multicast_if(struct vxlan_softc * sc, void *arg)
2140 struct ifvxlancmd *cmd;
2146 if (vxlan_can_change_config(sc)) {
2147 strlcpy(sc->vxl_mc_ifname, cmd->vxlcmd_ifname, IFNAMSIZ);
2148 vxlan_set_hwcaps(sc);
2158 vxlan_ctrl_set_ttl(struct vxlan_softc *sc, void *arg)
2160 struct ifvxlancmd *cmd;
2166 if (vxlan_check_ttl(cmd->vxlcmd_ttl) == 0) {
2167 sc->vxl_ttl = cmd->vxlcmd_ttl;
2168 if (sc->vxl_im4o != NULL)
2169 sc->vxl_im4o->imo_multicast_ttl = sc->vxl_ttl;
2170 if (sc->vxl_im6o != NULL)
2171 sc->vxl_im6o->im6o_multicast_hlim = sc->vxl_ttl;
2181 vxlan_ctrl_set_learn(struct vxlan_softc *sc, void *arg)
2183 struct ifvxlancmd *cmd;
2188 if (cmd->vxlcmd_flags & VXLAN_CMD_FLAG_LEARN)
2189 sc->vxl_flags |= VXLAN_FLAG_LEARN;
2191 sc->vxl_flags &= ~VXLAN_FLAG_LEARN;
2198 vxlan_ctrl_ftable_entry_add(struct vxlan_softc *sc, void *arg)
2200 union vxlan_sockaddr vxlsa;
2201 struct ifvxlancmd *cmd;
2202 struct vxlan_ftable_entry *fe;
2206 vxlsa = cmd->vxlcmd_sa;
2208 if (!VXLAN_SOCKADDR_IS_IPV46(&vxlsa))
2210 if (vxlan_sockaddr_in_any(&vxlsa) != 0)
2212 if (vxlan_sockaddr_in_multicast(&vxlsa) != 0)
2214 /* BMV: We could support both IPv4 and IPv6 later. */
2215 if (vxlsa.sa.sa_family != sc->vxl_dst_addr.sa.sa_family)
2216 return (EAFNOSUPPORT);
2218 if (VXLAN_SOCKADDR_IS_IPV6(&vxlsa)) {
2219 error = vxlan_sockaddr_in6_embedscope(&vxlsa);
2224 fe = vxlan_ftable_entry_alloc();
2228 if (vxlsa.in4.sin_port == 0)
2229 vxlsa.in4.sin_port = sc->vxl_dst_addr.in4.sin_port;
2231 vxlan_ftable_entry_init(sc, fe, cmd->vxlcmd_mac, &vxlsa.sa,
2232 VXLAN_FE_FLAG_STATIC);
2235 error = vxlan_ftable_entry_insert(sc, fe);
2239 vxlan_ftable_entry_free(fe);
2245 vxlan_ctrl_ftable_entry_rem(struct vxlan_softc *sc, void *arg)
2247 struct ifvxlancmd *cmd;
2248 struct vxlan_ftable_entry *fe;
2254 fe = vxlan_ftable_entry_lookup(sc, cmd->vxlcmd_mac);
2256 vxlan_ftable_entry_destroy(sc, fe);
2266 vxlan_ctrl_flush(struct vxlan_softc *sc, void *arg)
2268 struct ifvxlancmd *cmd;
2272 all = cmd->vxlcmd_flags & VXLAN_CMD_FLAG_FLUSH_ALL;
2275 vxlan_ftable_flush(sc, all);
2282 vxlan_ioctl_drvspec(struct vxlan_softc *sc, struct ifdrv *ifd, int get)
2284 const struct vxlan_control *vc;
2286 struct ifvxlancfg cfg;
2287 struct ifvxlancmd cmd;
2291 if (ifd->ifd_cmd >= vxlan_control_table_size)
2294 bzero(&args, sizeof(args));
2295 vc = &vxlan_control_table[ifd->ifd_cmd];
2296 out = (vc->vxlc_flags & VXLAN_CTRL_FLAG_COPYOUT) != 0;
2298 if ((get != 0 && out == 0) || (get == 0 && out != 0))
2301 if (vc->vxlc_flags & VXLAN_CTRL_FLAG_SUSER) {
2302 error = priv_check(curthread, PRIV_NET_VXLAN);
2307 if (ifd->ifd_len != vc->vxlc_argsize ||
2308 ifd->ifd_len > sizeof(args))
2311 if (vc->vxlc_flags & VXLAN_CTRL_FLAG_COPYIN) {
2312 error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
2317 error = vc->vxlc_func(sc, &args);
2321 if (vc->vxlc_flags & VXLAN_CTRL_FLAG_COPYOUT) {
2322 error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
2331 vxlan_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
2333 struct rm_priotracker tracker;
2334 struct vxlan_softc *sc;
2340 ifr = (struct ifreq *) data;
2341 ifd = (struct ifdrv *) data;
2352 error = vxlan_ioctl_drvspec(sc, ifd, cmd == SIOCGDRVSPEC);
2356 error = vxlan_ioctl_ifflags(sc);
2361 error = ifmedia_ioctl(ifp, ifr, &sc->vxl_media, cmd);
2365 if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > VXLAN_MAX_MTU) {
2369 ifp->if_mtu = ifr->ifr_mtu;
2370 sc->vxl_flags |= VXLAN_FLAG_USER_MTU;
2377 error = vxlan_set_reqcap(sc, ifp, ifr->ifr_reqcap);
2379 vxlan_set_hwcaps(sc);
2384 VXLAN_RLOCK(sc, &tracker);
2385 ifr->ifr_fib = sc->vxl_fibnum;
2386 VXLAN_RUNLOCK(sc, &tracker);
2390 if ((error = priv_check(curthread, PRIV_NET_VXLAN)) != 0)
2393 if (ifr->ifr_fib >= rt_numfibs)
2397 sc->vxl_fibnum = ifr->ifr_fib;
2403 error = ether_ioctl(ifp, cmd, data);
2410 #if defined(INET) || defined(INET6)
2412 vxlan_pick_source_port(struct vxlan_softc *sc, struct mbuf *m)
2417 range = sc->vxl_max_port - sc->vxl_min_port + 1;
2419 if (M_HASHTYPE_ISHASH(m))
2420 hash = m->m_pkthdr.flowid;
2422 hash = jenkins_hash(m->m_data, ETHER_HDR_LEN,
2423 sc->vxl_port_hash_key);
2425 return (sc->vxl_min_port + (hash % range));
2429 vxlan_encap_header(struct vxlan_softc *sc, struct mbuf *m, int ipoff,
2430 uint16_t srcport, uint16_t dstport)
2432 struct vxlanudphdr *hdr;
2433 struct udphdr *udph;
2434 struct vxlan_header *vxh;
2437 len = m->m_pkthdr.len - ipoff;
2438 MPASS(len >= sizeof(struct vxlanudphdr));
2439 hdr = mtodo(m, ipoff);
2441 udph = &hdr->vxlh_udp;
2442 udph->uh_sport = srcport;
2443 udph->uh_dport = dstport;
2444 udph->uh_ulen = htons(len);
2447 vxh = &hdr->vxlh_hdr;
2448 vxh->vxlh_flags = htonl(VXLAN_HDR_FLAGS_VALID_VNI);
2449 vxh->vxlh_vni = htonl(sc->vxl_vni << VXLAN_HDR_VNI_SHIFT);
2454 * Return the CSUM_INNER_* equivalent of CSUM_* caps.
2457 csum_flags_to_inner_flags(uint32_t csum_flags_in, const uint32_t encap)
2459 uint32_t csum_flags = encap;
2460 const uint32_t v4 = CSUM_IP | CSUM_IP_UDP | CSUM_IP_TCP;
2463 * csum_flags can request either v4 or v6 offload but not both.
2464 * tcp_output always sets CSUM_TSO (both CSUM_IP_TSO and CSUM_IP6_TSO)
2465 * so those bits are no good to detect the IP version. Other bits are
2466 * always set with CSUM_TSO and we use those to figure out the IP
2469 if (csum_flags_in & v4) {
2470 if (csum_flags_in & CSUM_IP)
2471 csum_flags |= CSUM_INNER_IP;
2472 if (csum_flags_in & CSUM_IP_UDP)
2473 csum_flags |= CSUM_INNER_IP_UDP;
2474 if (csum_flags_in & CSUM_IP_TCP)
2475 csum_flags |= CSUM_INNER_IP_TCP;
2476 if (csum_flags_in & CSUM_IP_TSO)
2477 csum_flags |= CSUM_INNER_IP_TSO;
2480 const uint32_t v6 = CSUM_IP6_UDP | CSUM_IP6_TCP;
2482 MPASS((csum_flags_in & v6) != 0);
2484 if (csum_flags_in & CSUM_IP6_UDP)
2485 csum_flags |= CSUM_INNER_IP6_UDP;
2486 if (csum_flags_in & CSUM_IP6_TCP)
2487 csum_flags |= CSUM_INNER_IP6_TCP;
2488 if (csum_flags_in & CSUM_IP6_TSO)
2489 csum_flags |= CSUM_INNER_IP6_TSO;
2492 return (csum_flags);
2496 vxlan_encap4(struct vxlan_softc *sc, const union vxlan_sockaddr *fvxlsa,
2502 struct in_addr srcaddr, dstaddr;
2503 uint16_t srcport, dstport;
2504 int plen, mcast, error;
2505 struct route route, *ro;
2506 struct sockaddr_in *sin;
2507 uint32_t csum_flags;
2512 srcaddr = sc->vxl_src_addr.in4.sin_addr;
2513 srcport = vxlan_pick_source_port(sc, m);
2514 dstaddr = fvxlsa->in4.sin_addr;
2515 dstport = fvxlsa->in4.sin_port;
2517 plen = m->m_pkthdr.len;
2518 M_PREPEND(m, sizeof(struct ip) + sizeof(struct vxlanudphdr),
2521 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2525 ip = mtod(m, struct ip *);
2527 ip->ip_len = htons(m->m_pkthdr.len);
2529 ip->ip_ttl = sc->vxl_ttl;
2530 ip->ip_p = IPPROTO_UDP;
2532 ip->ip_src = srcaddr;
2533 ip->ip_dst = dstaddr;
2535 vxlan_encap_header(sc, m, sizeof(struct ip), srcport, dstport);
2537 mcast = (m->m_flags & (M_MCAST | M_BCAST)) ? 1 : 0;
2538 m->m_flags &= ~(M_MCAST | M_BCAST);
2540 m->m_pkthdr.csum_flags &= CSUM_FLAGS_TX;
2541 if (m->m_pkthdr.csum_flags != 0) {
2543 * HW checksum (L3 and/or L4) or TSO has been requested. Look
2544 * up the ifnet for the outbound route and verify that the
2545 * outbound ifnet can perform the requested operation on the
2548 bzero(&route, sizeof(route));
2550 sin = (struct sockaddr_in *)&ro->ro_dst;
2551 sin->sin_family = AF_INET;
2552 sin->sin_len = sizeof(*sin);
2553 sin->sin_addr = ip->ip_dst;
2554 ro->ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE,
2556 if (ro->ro_nh == NULL) {
2558 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2559 return (EHOSTUNREACH);
2562 csum_flags = csum_flags_to_inner_flags(m->m_pkthdr.csum_flags,
2564 if ((csum_flags & ro->ro_nh->nh_ifp->if_hwassist) !=
2566 if (ppsratecheck(&sc->err_time, &sc->err_pps, 1)) {
2567 const struct ifnet *nh_ifp = ro->ro_nh->nh_ifp;
2569 if_printf(ifp, "interface %s is missing hwcaps "
2570 "0x%08x, csum_flags 0x%08x -> 0x%08x, "
2571 "hwassist 0x%08x\n", nh_ifp->if_xname,
2572 csum_flags & ~(uint32_t)nh_ifp->if_hwassist,
2573 m->m_pkthdr.csum_flags, csum_flags,
2574 (uint32_t)nh_ifp->if_hwassist);
2577 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2580 m->m_pkthdr.csum_flags = csum_flags;
2582 (CSUM_INNER_IP | CSUM_INNER_IP_UDP | CSUM_INNER_IP6_UDP |
2583 CSUM_INNER_IP_TCP | CSUM_INNER_IP6_TCP)) {
2584 counter_u64_add(sc->vxl_stats.txcsum, 1);
2585 if (csum_flags & CSUM_INNER_TSO)
2586 counter_u64_add(sc->vxl_stats.tso, 1);
2590 error = ip_output(m, NULL, ro, 0, sc->vxl_im4o, NULL);
2592 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
2593 if_inc_counter(ifp, IFCOUNTER_OBYTES, plen);
2595 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
2597 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2607 vxlan_encap6(struct vxlan_softc *sc, const union vxlan_sockaddr *fvxlsa,
2612 struct ip6_hdr *ip6;
2613 const struct in6_addr *srcaddr, *dstaddr;
2614 uint16_t srcport, dstport;
2615 int plen, mcast, error;
2616 struct route_in6 route, *ro;
2617 struct sockaddr_in6 *sin6;
2618 uint32_t csum_flags;
2623 srcaddr = &sc->vxl_src_addr.in6.sin6_addr;
2624 srcport = vxlan_pick_source_port(sc, m);
2625 dstaddr = &fvxlsa->in6.sin6_addr;
2626 dstport = fvxlsa->in6.sin6_port;
2628 plen = m->m_pkthdr.len;
2629 M_PREPEND(m, sizeof(struct ip6_hdr) + sizeof(struct vxlanudphdr),
2632 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2636 ip6 = mtod(m, struct ip6_hdr *);
2637 ip6->ip6_flow = 0; /* BMV: Keep in forwarding entry? */
2638 ip6->ip6_vfc = IPV6_VERSION;
2640 ip6->ip6_nxt = IPPROTO_UDP;
2641 ip6->ip6_hlim = sc->vxl_ttl;
2642 ip6->ip6_src = *srcaddr;
2643 ip6->ip6_dst = *dstaddr;
2645 vxlan_encap_header(sc, m, sizeof(struct ip6_hdr), srcport, dstport);
2647 mcast = (m->m_flags & (M_MCAST | M_BCAST)) ? 1 : 0;
2648 m->m_flags &= ~(M_MCAST | M_BCAST);
2651 m->m_pkthdr.csum_flags &= CSUM_FLAGS_TX;
2652 if (m->m_pkthdr.csum_flags != 0) {
2654 * HW checksum (L3 and/or L4) or TSO has been requested. Look
2655 * up the ifnet for the outbound route and verify that the
2656 * outbound ifnet can perform the requested operation on the
2659 bzero(&route, sizeof(route));
2661 sin6 = (struct sockaddr_in6 *)&ro->ro_dst;
2662 sin6->sin6_family = AF_INET6;
2663 sin6->sin6_len = sizeof(*sin6);
2664 sin6->sin6_addr = ip6->ip6_dst;
2665 ro->ro_nh = fib6_lookup(M_GETFIB(m), &ip6->ip6_dst, 0,
2667 if (ro->ro_nh == NULL) {
2669 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2670 return (EHOSTUNREACH);
2673 csum_flags = csum_flags_to_inner_flags(m->m_pkthdr.csum_flags,
2675 if ((csum_flags & ro->ro_nh->nh_ifp->if_hwassist) !=
2677 if (ppsratecheck(&sc->err_time, &sc->err_pps, 1)) {
2678 const struct ifnet *nh_ifp = ro->ro_nh->nh_ifp;
2680 if_printf(ifp, "interface %s is missing hwcaps "
2681 "0x%08x, csum_flags 0x%08x -> 0x%08x, "
2682 "hwassist 0x%08x\n", nh_ifp->if_xname,
2683 csum_flags & ~(uint32_t)nh_ifp->if_hwassist,
2684 m->m_pkthdr.csum_flags, csum_flags,
2685 (uint32_t)nh_ifp->if_hwassist);
2688 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2691 m->m_pkthdr.csum_flags = csum_flags;
2693 (CSUM_INNER_IP | CSUM_INNER_IP_UDP | CSUM_INNER_IP6_UDP |
2694 CSUM_INNER_IP_TCP | CSUM_INNER_IP6_TCP)) {
2695 counter_u64_add(sc->vxl_stats.txcsum, 1);
2696 if (csum_flags & CSUM_INNER_TSO)
2697 counter_u64_add(sc->vxl_stats.tso, 1);
2699 } else if (ntohs(dstport) != V_zero_checksum_port) {
2700 struct udphdr *hdr = mtodo(m, sizeof(struct ip6_hdr));
2702 hdr->uh_sum = in6_cksum_pseudo(ip6,
2703 m->m_pkthdr.len - sizeof(struct ip6_hdr), IPPROTO_UDP, 0);
2704 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
2705 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
2707 error = ip6_output(m, NULL, ro, 0, sc->vxl_im6o, NULL, NULL);
2709 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
2710 if_inc_counter(ifp, IFCOUNTER_OBYTES, plen);
2712 if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
2714 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
2724 vxlan_transmit(struct ifnet *ifp, struct mbuf *m)
2726 struct rm_priotracker tracker;
2727 union vxlan_sockaddr vxlsa;
2728 struct vxlan_softc *sc;
2729 struct vxlan_ftable_entry *fe;
2730 struct ifnet *mcifp;
2731 struct ether_header *eh;
2735 eh = mtod(m, struct ether_header *);
2739 ETHER_BPF_MTAP(ifp, m);
2741 VXLAN_RLOCK(sc, &tracker);
2742 M_SETFIB(m, sc->vxl_fibnum);
2743 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2744 VXLAN_RUNLOCK(sc, &tracker);
2749 if ((m->m_flags & (M_BCAST | M_MCAST)) == 0)
2750 fe = vxlan_ftable_entry_lookup(sc, eh->ether_dhost);
2752 fe = &sc->vxl_default_fe;
2753 vxlan_sockaddr_copy(&vxlsa, &fe->vxlfe_raddr.sa);
2755 ipv4 = VXLAN_SOCKADDR_IS_IPV4(&vxlsa) != 0;
2756 if (vxlan_sockaddr_in_multicast(&vxlsa) != 0)
2757 mcifp = vxlan_multicast_if_ref(sc, ipv4);
2760 VXLAN_RUNLOCK(sc, &tracker);
2763 error = vxlan_encap4(sc, &vxlsa, m);
2765 error = vxlan_encap6(sc, &vxlsa, m);
2775 vxlan_qflush(struct ifnet *ifp __unused)
2780 vxlan_rcv_udp_packet(struct mbuf *m, int offset, struct inpcb *inpcb,
2781 const struct sockaddr *srcsa, void *xvso)
2783 struct vxlan_socket *vso;
2784 struct vxlan_header *vxh, vxlanhdr;
2790 offset += sizeof(struct udphdr);
2792 if (m->m_pkthdr.len < offset + sizeof(struct vxlan_header))
2795 if (__predict_false(m->m_len < offset + sizeof(struct vxlan_header))) {
2796 m_copydata(m, offset, sizeof(struct vxlan_header),
2797 (caddr_t) &vxlanhdr);
2800 vxh = mtodo(m, offset);
2803 * Drop if there is a reserved bit set in either the flags or VNI
2804 * fields of the header. This goes against the specification, but
2805 * a bit set may indicate an unsupported new feature. This matches
2806 * the behavior of the Linux implementation.
2808 if (vxh->vxlh_flags != htonl(VXLAN_HDR_FLAGS_VALID_VNI) ||
2809 vxh->vxlh_vni & ~VXLAN_VNI_MASK)
2812 vni = ntohl(vxh->vxlh_vni) >> VXLAN_HDR_VNI_SHIFT;
2814 /* Adjust to the start of the inner Ethernet frame. */
2815 m_adj_decap(m, offset + sizeof(struct vxlan_header));
2817 error = vxlan_input(vso, vni, &m, srcsa);
2818 MPASS(error != 0 || m == NULL);
2826 vxlan_input(struct vxlan_socket *vso, uint32_t vni, struct mbuf **m0,
2827 const struct sockaddr *sa)
2829 struct vxlan_softc *sc;
2832 struct ether_header *eh;
2837 if (m->m_pkthdr.len < ETHER_HDR_LEN)
2840 sc = vxlan_socket_lookup_softc(vso, vni);
2845 if (m->m_len < ETHER_HDR_LEN &&
2846 (m = m_pullup(m, ETHER_HDR_LEN)) == NULL) {
2851 eh = mtod(m, struct ether_header *);
2853 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2856 } else if (ifp == m->m_pkthdr.rcvif) {
2857 /* XXX Does not catch more complex loops. */
2862 if (sc->vxl_flags & VXLAN_FLAG_LEARN)
2863 vxlan_ftable_learn(sc, sa, eh->ether_shost);
2866 m->m_pkthdr.rcvif = ifp;
2867 M_SETFIB(m, ifp->if_fib);
2868 if (((ifp->if_capenable & IFCAP_RXCSUM &&
2869 m->m_pkthdr.csum_flags & CSUM_INNER_L3_CALC) ||
2870 (ifp->if_capenable & IFCAP_RXCSUM_IPV6 &&
2871 !(m->m_pkthdr.csum_flags & CSUM_INNER_L3_CALC)))) {
2872 uint32_t csum_flags = 0;
2874 if (m->m_pkthdr.csum_flags & CSUM_INNER_L3_CALC)
2875 csum_flags |= CSUM_L3_CALC;
2876 if (m->m_pkthdr.csum_flags & CSUM_INNER_L3_VALID)
2877 csum_flags |= CSUM_L3_VALID;
2878 if (m->m_pkthdr.csum_flags & CSUM_INNER_L4_CALC)
2879 csum_flags |= CSUM_L4_CALC;
2880 if (m->m_pkthdr.csum_flags & CSUM_INNER_L4_VALID)
2881 csum_flags |= CSUM_L4_VALID;
2882 m->m_pkthdr.csum_flags = csum_flags;
2883 counter_u64_add(sc->vxl_stats.rxcsum, 1);
2885 /* clear everything */
2886 m->m_pkthdr.csum_flags = 0;
2887 m->m_pkthdr.csum_data = 0;
2890 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2891 (*ifp->if_input)(ifp, m);
2901 vxlan_stats_alloc(struct vxlan_softc *sc)
2903 struct vxlan_statistics *stats = &sc->vxl_stats;
2905 stats->txcsum = counter_u64_alloc(M_WAITOK);
2906 if (stats->txcsum == NULL)
2909 stats->tso = counter_u64_alloc(M_WAITOK);
2910 if (stats->tso == NULL)
2913 stats->rxcsum = counter_u64_alloc(M_WAITOK);
2914 if (stats->rxcsum == NULL)
2919 vxlan_stats_free(sc);
2924 vxlan_stats_free(struct vxlan_softc *sc)
2926 struct vxlan_statistics *stats = &sc->vxl_stats;
2928 if (stats->txcsum != NULL) {
2929 counter_u64_free(stats->txcsum);
2930 stats->txcsum = NULL;
2932 if (stats->tso != NULL) {
2933 counter_u64_free(stats->tso);
2936 if (stats->rxcsum != NULL) {
2937 counter_u64_free(stats->rxcsum);
2938 stats->rxcsum = NULL;
2943 vxlan_set_default_config(struct vxlan_softc *sc)
2946 sc->vxl_flags |= VXLAN_FLAG_LEARN;
2948 sc->vxl_vni = VXLAN_VNI_MAX;
2949 sc->vxl_ttl = IPDEFTTL;
2951 if (!vxlan_tunable_int(sc, "legacy_port", vxlan_legacy_port)) {
2952 sc->vxl_src_addr.in4.sin_port = htons(VXLAN_PORT);
2953 sc->vxl_dst_addr.in4.sin_port = htons(VXLAN_PORT);
2955 sc->vxl_src_addr.in4.sin_port = htons(VXLAN_LEGACY_PORT);
2956 sc->vxl_dst_addr.in4.sin_port = htons(VXLAN_LEGACY_PORT);
2959 sc->vxl_min_port = V_ipport_firstauto;
2960 sc->vxl_max_port = V_ipport_lastauto;
2962 sc->vxl_ftable_max = VXLAN_FTABLE_MAX;
2963 sc->vxl_ftable_timeout = VXLAN_FTABLE_TIMEOUT;
2967 vxlan_set_user_config(struct vxlan_softc *sc, struct ifvxlanparam *vxlp)
2971 if (vxlp->vxlp_with & (VXLAN_PARAM_WITH_LOCAL_ADDR4 |
2972 VXLAN_PARAM_WITH_REMOTE_ADDR4))
2973 return (EAFNOSUPPORT);
2977 if (vxlp->vxlp_with & (VXLAN_PARAM_WITH_LOCAL_ADDR6 |
2978 VXLAN_PARAM_WITH_REMOTE_ADDR6))
2979 return (EAFNOSUPPORT);
2981 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LOCAL_ADDR6) {
2982 int error = vxlan_sockaddr_in6_embedscope(&vxlp->vxlp_local_sa);
2986 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_REMOTE_ADDR6) {
2987 int error = vxlan_sockaddr_in6_embedscope(
2988 &vxlp->vxlp_remote_sa);
2994 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_VNI) {
2995 if (vxlan_check_vni(vxlp->vxlp_vni) == 0)
2996 sc->vxl_vni = vxlp->vxlp_vni;
2999 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LOCAL_ADDR4) {
3000 sc->vxl_src_addr.in4.sin_len = sizeof(struct sockaddr_in);
3001 sc->vxl_src_addr.in4.sin_family = AF_INET;
3002 sc->vxl_src_addr.in4.sin_addr =
3003 vxlp->vxlp_local_sa.in4.sin_addr;
3004 } else if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LOCAL_ADDR6) {
3005 sc->vxl_src_addr.in6.sin6_len = sizeof(struct sockaddr_in6);
3006 sc->vxl_src_addr.in6.sin6_family = AF_INET6;
3007 sc->vxl_src_addr.in6.sin6_addr =
3008 vxlp->vxlp_local_sa.in6.sin6_addr;
3011 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_REMOTE_ADDR4) {
3012 sc->vxl_dst_addr.in4.sin_len = sizeof(struct sockaddr_in);
3013 sc->vxl_dst_addr.in4.sin_family = AF_INET;
3014 sc->vxl_dst_addr.in4.sin_addr =
3015 vxlp->vxlp_remote_sa.in4.sin_addr;
3016 } else if (vxlp->vxlp_with & VXLAN_PARAM_WITH_REMOTE_ADDR6) {
3017 sc->vxl_dst_addr.in6.sin6_len = sizeof(struct sockaddr_in6);
3018 sc->vxl_dst_addr.in6.sin6_family = AF_INET6;
3019 sc->vxl_dst_addr.in6.sin6_addr =
3020 vxlp->vxlp_remote_sa.in6.sin6_addr;
3023 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LOCAL_PORT)
3024 sc->vxl_src_addr.in4.sin_port = htons(vxlp->vxlp_local_port);
3025 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_REMOTE_PORT)
3026 sc->vxl_dst_addr.in4.sin_port = htons(vxlp->vxlp_remote_port);
3028 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_PORT_RANGE) {
3029 if (vxlp->vxlp_min_port <= vxlp->vxlp_max_port) {
3030 sc->vxl_min_port = vxlp->vxlp_min_port;
3031 sc->vxl_max_port = vxlp->vxlp_max_port;
3035 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_MULTICAST_IF)
3036 strlcpy(sc->vxl_mc_ifname, vxlp->vxlp_mc_ifname, IFNAMSIZ);
3038 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_FTABLE_TIMEOUT) {
3039 if (vxlan_check_ftable_timeout(vxlp->vxlp_ftable_timeout) == 0)
3040 sc->vxl_ftable_timeout = vxlp->vxlp_ftable_timeout;
3043 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_FTABLE_MAX) {
3044 if (vxlan_check_ftable_max(vxlp->vxlp_ftable_max) == 0)
3045 sc->vxl_ftable_max = vxlp->vxlp_ftable_max;
3048 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_TTL) {
3049 if (vxlan_check_ttl(vxlp->vxlp_ttl) == 0)
3050 sc->vxl_ttl = vxlp->vxlp_ttl;
3053 if (vxlp->vxlp_with & VXLAN_PARAM_WITH_LEARN) {
3054 if (vxlp->vxlp_learn == 0)
3055 sc->vxl_flags &= ~VXLAN_FLAG_LEARN;
3062 vxlan_set_reqcap(struct vxlan_softc *sc, struct ifnet *ifp, int reqcap)
3064 int mask = reqcap ^ ifp->if_capenable;
3066 /* Disable TSO if tx checksums are disabled. */
3067 if (mask & IFCAP_TXCSUM && !(reqcap & IFCAP_TXCSUM) &&
3068 reqcap & IFCAP_TSO4) {
3069 reqcap &= ~IFCAP_TSO4;
3070 if_printf(ifp, "tso4 disabled due to -txcsum.\n");
3072 if (mask & IFCAP_TXCSUM_IPV6 && !(reqcap & IFCAP_TXCSUM_IPV6) &&
3073 reqcap & IFCAP_TSO6) {
3074 reqcap &= ~IFCAP_TSO6;
3075 if_printf(ifp, "tso6 disabled due to -txcsum6.\n");
3078 /* Do not enable TSO if tx checksums are disabled. */
3079 if (mask & IFCAP_TSO4 && reqcap & IFCAP_TSO4 &&
3080 !(reqcap & IFCAP_TXCSUM)) {
3081 if_printf(ifp, "enable txcsum first.\n");
3084 if (mask & IFCAP_TSO6 && reqcap & IFCAP_TSO6 &&
3085 !(reqcap & IFCAP_TXCSUM_IPV6)) {
3086 if_printf(ifp, "enable txcsum6 first.\n");
3090 sc->vxl_reqcap = reqcap;
3095 * A VXLAN interface inherits the capabilities of the vxlandev or the interface
3096 * hosting the vxlanlocal address.
3099 vxlan_set_hwcaps(struct vxlan_softc *sc)
3101 struct epoch_tracker et;
3107 struct ifnet *ifp = sc->vxl_ifp;
3110 ifp->if_capabilities &= VXLAN_BASIC_IFCAPS;
3111 ifp->if_capenable &= VXLAN_BASIC_IFCAPS;
3112 ifp->if_hwassist = 0;
3114 NET_EPOCH_ENTER(et);
3115 CURVNET_SET(ifp->if_vnet);
3119 if (sc->vxl_mc_ifname[0] != '\0') {
3121 p = ifunit_ref(sc->vxl_mc_ifname);
3122 } else if (vxlan_sockaddr_in_any(&sc->vxl_src_addr) == 0) {
3123 if (sc->vxl_src_addr.sa.sa_family == AF_INET) {
3124 struct sockaddr_in in4 = sc->vxl_src_addr.in4;
3127 ifa = ifa_ifwithaddr((struct sockaddr *)&in4);
3130 } else if (sc->vxl_src_addr.sa.sa_family == AF_INET6) {
3131 struct sockaddr_in6 in6 = sc->vxl_src_addr.in6;
3134 ifa = ifa_ifwithaddr((struct sockaddr *)&in6);
3142 cap = ena = hwa = 0;
3144 /* checksum offload */
3145 if (p->if_capabilities & IFCAP_VXLAN_HWCSUM)
3146 cap |= p->if_capabilities & (IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6);
3147 if (p->if_capenable & IFCAP_VXLAN_HWCSUM) {
3148 ena |= sc->vxl_reqcap & p->if_capenable &
3149 (IFCAP_HWCSUM | IFCAP_HWCSUM_IPV6);
3150 if (ena & IFCAP_TXCSUM) {
3151 if (p->if_hwassist & CSUM_INNER_IP)
3153 if (p->if_hwassist & CSUM_INNER_IP_UDP)
3155 if (p->if_hwassist & CSUM_INNER_IP_TCP)
3158 if (ena & IFCAP_TXCSUM_IPV6) {
3159 if (p->if_hwassist & CSUM_INNER_IP6_UDP)
3160 hwa |= CSUM_IP6_UDP;
3161 if (p->if_hwassist & CSUM_INNER_IP6_TCP)
3162 hwa |= CSUM_IP6_TCP;
3167 if (p->if_capabilities & IFCAP_VXLAN_HWTSO) {
3168 cap |= p->if_capabilities & IFCAP_TSO;
3169 if (p->if_hw_tsomax > IP_MAXPACKET - ifp->if_hdrlen)
3170 ifp->if_hw_tsomax = IP_MAXPACKET - ifp->if_hdrlen;
3172 ifp->if_hw_tsomax = p->if_hw_tsomax;
3173 /* XXX: tsomaxsegcount decrement is cxgbe specific */
3174 ifp->if_hw_tsomaxsegcount = p->if_hw_tsomaxsegcount - 1;
3175 ifp->if_hw_tsomaxsegsize = p->if_hw_tsomaxsegsize;
3177 if (p->if_capenable & IFCAP_VXLAN_HWTSO) {
3178 ena |= sc->vxl_reqcap & p->if_capenable & IFCAP_TSO;
3179 if (ena & IFCAP_TSO) {
3180 if (p->if_hwassist & CSUM_INNER_IP_TSO)
3182 if (p->if_hwassist & CSUM_INNER_IP6_TSO)
3183 hwa |= CSUM_IP6_TSO;
3187 ifp->if_capabilities |= cap;
3188 ifp->if_capenable |= ena;
3189 ifp->if_hwassist |= hwa;
3198 vxlan_clone_create(struct if_clone *ifc, int unit, caddr_t params)
3200 struct vxlan_softc *sc;
3202 struct ifvxlanparam vxlp;
3205 sc = malloc(sizeof(struct vxlan_softc), M_VXLAN, M_WAITOK | M_ZERO);
3206 sc->vxl_unit = unit;
3207 sc->vxl_fibnum = curthread->td_proc->p_fibnum;
3208 vxlan_set_default_config(sc);
3209 error = vxlan_stats_alloc(sc);
3214 error = copyin(params, &vxlp, sizeof(vxlp));
3218 error = vxlan_set_user_config(sc, &vxlp);
3223 ifp = if_alloc(IFT_ETHER);
3230 rm_init(&sc->vxl_lock, "vxlanrm");
3231 callout_init_rw(&sc->vxl_callout, &sc->vxl_lock, 0);
3232 sc->vxl_port_hash_key = arc4random();
3233 vxlan_ftable_init(sc);
3235 vxlan_sysctl_setup(sc);
3238 if_initname(ifp, vxlan_name, unit);
3239 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
3240 ifp->if_init = vxlan_init;
3241 ifp->if_ioctl = vxlan_ioctl;
3242 ifp->if_transmit = vxlan_transmit;
3243 ifp->if_qflush = vxlan_qflush;
3244 ifp->if_capabilities = VXLAN_BASIC_IFCAPS;
3245 ifp->if_capenable = VXLAN_BASIC_IFCAPS;
3246 sc->vxl_reqcap = -1;
3247 vxlan_set_hwcaps(sc);
3249 ifmedia_init(&sc->vxl_media, 0, vxlan_media_change, vxlan_media_status);
3250 ifmedia_add(&sc->vxl_media, IFM_ETHER | IFM_AUTO, 0, NULL);
3251 ifmedia_set(&sc->vxl_media, IFM_ETHER | IFM_AUTO);
3253 ether_gen_addr(ifp, &sc->vxl_hwaddr);
3254 ether_ifattach(ifp, sc->vxl_hwaddr.octet);
3256 ifp->if_baudrate = 0;
3259 vxlan_setup_interface_hdrlen(sc);
3270 vxlan_clone_destroy(struct ifnet *ifp)
3272 struct vxlan_softc *sc;
3278 vxlan_ftable_flush(sc, 1);
3280 ether_ifdetach(ifp);
3282 ifmedia_removeall(&sc->vxl_media);
3284 vxlan_ftable_fini(sc);
3286 vxlan_sysctl_destroy(sc);
3287 rm_destroy(&sc->vxl_lock);
3288 vxlan_stats_free(sc);
3292 /* BMV: Taken from if_bridge. */
3294 vxlan_mac_hash(struct vxlan_softc *sc, const uint8_t *addr)
3296 uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->vxl_ftable_hash_key;
3306 * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3307 * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3309 #define mix(a, b, c) \
3311 a -= b; a -= c; a ^= (c >> 13); \
3312 b -= c; b -= a; b ^= (a << 8); \
3313 c -= a; c -= b; c ^= (b >> 13); \
3314 a -= b; a -= c; a ^= (c >> 12); \
3315 b -= c; b -= a; b ^= (a << 16); \
3316 c -= a; c -= b; c ^= (b >> 5); \
3317 a -= b; a -= c; a ^= (c >> 3); \
3318 b -= c; b -= a; b ^= (a << 10); \
3319 c -= a; c -= b; c ^= (b >> 15); \
3330 vxlan_media_change(struct ifnet *ifp)
3338 vxlan_media_status(struct ifnet *ifp, struct ifmediareq *ifmr)
3341 ifmr->ifm_status = IFM_ACTIVE | IFM_AVALID;
3342 ifmr->ifm_active = IFM_ETHER | IFM_FDX;
3346 vxlan_sockaddr_cmp(const union vxlan_sockaddr *vxladdr,
3347 const struct sockaddr *sa)
3350 return (bcmp(&vxladdr->sa, sa, vxladdr->sa.sa_len));
3354 vxlan_sockaddr_copy(union vxlan_sockaddr *vxladdr,
3355 const struct sockaddr *sa)
3358 MPASS(sa->sa_family == AF_INET || sa->sa_family == AF_INET6);
3359 bzero(vxladdr, sizeof(*vxladdr));
3361 if (sa->sa_family == AF_INET) {
3362 vxladdr->in4 = *satoconstsin(sa);
3363 vxladdr->in4.sin_len = sizeof(struct sockaddr_in);
3364 } else if (sa->sa_family == AF_INET6) {
3365 vxladdr->in6 = *satoconstsin6(sa);
3366 vxladdr->in6.sin6_len = sizeof(struct sockaddr_in6);
3371 vxlan_sockaddr_in_equal(const union vxlan_sockaddr *vxladdr,
3372 const struct sockaddr *sa)
3376 if (sa->sa_family == AF_INET) {
3377 const struct in_addr *in4 = &satoconstsin(sa)->sin_addr;
3378 equal = in4->s_addr == vxladdr->in4.sin_addr.s_addr;
3379 } else if (sa->sa_family == AF_INET6) {
3380 const struct in6_addr *in6 = &satoconstsin6(sa)->sin6_addr;
3381 equal = IN6_ARE_ADDR_EQUAL(in6, &vxladdr->in6.sin6_addr);
3389 vxlan_sockaddr_in_copy(union vxlan_sockaddr *vxladdr,
3390 const struct sockaddr *sa)
3393 MPASS(sa->sa_family == AF_INET || sa->sa_family == AF_INET6);
3395 if (sa->sa_family == AF_INET) {
3396 const struct in_addr *in4 = &satoconstsin(sa)->sin_addr;
3397 vxladdr->in4.sin_family = AF_INET;
3398 vxladdr->in4.sin_len = sizeof(struct sockaddr_in);
3399 vxladdr->in4.sin_addr = *in4;
3400 } else if (sa->sa_family == AF_INET6) {
3401 const struct in6_addr *in6 = &satoconstsin6(sa)->sin6_addr;
3402 vxladdr->in6.sin6_family = AF_INET6;
3403 vxladdr->in6.sin6_len = sizeof(struct sockaddr_in6);
3404 vxladdr->in6.sin6_addr = *in6;
3409 vxlan_sockaddr_supported(const union vxlan_sockaddr *vxladdr, int unspec)
3411 const struct sockaddr *sa;
3417 if (sa->sa_family == AF_UNSPEC && unspec != 0) {
3419 } else if (sa->sa_family == AF_INET) {
3423 } else if (sa->sa_family == AF_INET6) {
3433 vxlan_sockaddr_in_any(const union vxlan_sockaddr *vxladdr)
3435 const struct sockaddr *sa;
3440 if (sa->sa_family == AF_INET) {
3441 const struct in_addr *in4 = &satoconstsin(sa)->sin_addr;
3442 any = in4->s_addr == INADDR_ANY;
3443 } else if (sa->sa_family == AF_INET6) {
3444 const struct in6_addr *in6 = &satoconstsin6(sa)->sin6_addr;
3445 any = IN6_IS_ADDR_UNSPECIFIED(in6);
3453 vxlan_sockaddr_in_multicast(const union vxlan_sockaddr *vxladdr)
3455 const struct sockaddr *sa;
3460 if (sa->sa_family == AF_INET) {
3461 const struct in_addr *in4 = &satoconstsin(sa)->sin_addr;
3462 mc = IN_MULTICAST(ntohl(in4->s_addr));
3463 } else if (sa->sa_family == AF_INET6) {
3464 const struct in6_addr *in6 = &satoconstsin6(sa)->sin6_addr;
3465 mc = IN6_IS_ADDR_MULTICAST(in6);
3473 vxlan_sockaddr_in6_embedscope(union vxlan_sockaddr *vxladdr)
3477 MPASS(VXLAN_SOCKADDR_IS_IPV6(vxladdr));
3479 error = sa6_embedscope(&vxladdr->in6, V_ip6_use_defzone);
3481 error = EAFNOSUPPORT;
3488 vxlan_can_change_config(struct vxlan_softc *sc)
3493 VXLAN_LOCK_ASSERT(sc);
3495 if (ifp->if_drv_flags & IFF_DRV_RUNNING)
3497 if (sc->vxl_flags & (VXLAN_FLAG_INIT | VXLAN_FLAG_TEARDOWN))
3504 vxlan_check_vni(uint32_t vni)
3507 return (vni >= VXLAN_VNI_MAX);
3511 vxlan_check_ttl(int ttl)
3514 return (ttl > MAXTTL);
3518 vxlan_check_ftable_timeout(uint32_t timeout)
3521 return (timeout > VXLAN_FTABLE_MAX_TIMEOUT);
3525 vxlan_check_ftable_max(uint32_t max)
3528 return (max > VXLAN_FTABLE_MAX);
3532 vxlan_sysctl_setup(struct vxlan_softc *sc)
3534 struct sysctl_ctx_list *ctx;
3535 struct sysctl_oid *node;
3536 struct vxlan_statistics *stats;
3539 ctx = &sc->vxl_sysctl_ctx;
3540 stats = &sc->vxl_stats;
3541 snprintf(namebuf, sizeof(namebuf), "%d", sc->vxl_unit);
3543 sysctl_ctx_init(ctx);
3544 sc->vxl_sysctl_node = SYSCTL_ADD_NODE(ctx,
3545 SYSCTL_STATIC_CHILDREN(_net_link_vxlan), OID_AUTO, namebuf,
3546 CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "");
3548 node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(sc->vxl_sysctl_node),
3549 OID_AUTO, "ftable", CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "");
3550 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "count",
3551 CTLFLAG_RD, &sc->vxl_ftable_cnt, 0,
3552 "Number of entries in forwarding table");
3553 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "max",
3554 CTLFLAG_RD, &sc->vxl_ftable_max, 0,
3555 "Maximum number of entries allowed in forwarding table");
3556 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "timeout",
3557 CTLFLAG_RD, &sc->vxl_ftable_timeout, 0,
3558 "Number of seconds between prunes of the forwarding table");
3559 SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "dump",
3560 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE | CTLFLAG_SKIP,
3561 sc, 0, vxlan_ftable_sysctl_dump, "A",
3562 "Dump the forwarding table entries");
3564 node = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(sc->vxl_sysctl_node),
3565 OID_AUTO, "stats", CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "");
3566 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO,
3567 "ftable_nospace", CTLFLAG_RD, &stats->ftable_nospace, 0,
3568 "Fowarding table reached maximum entries");
3569 SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(node), OID_AUTO,
3570 "ftable_lock_upgrade_failed", CTLFLAG_RD,
3571 &stats->ftable_lock_upgrade_failed, 0,
3572 "Forwarding table update required lock upgrade");
3574 SYSCTL_ADD_COUNTER_U64(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "txcsum",
3575 CTLFLAG_RD, &stats->txcsum,
3576 "# of times hardware assisted with tx checksum");
3577 SYSCTL_ADD_COUNTER_U64(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "tso",
3578 CTLFLAG_RD, &stats->tso, "# of times hardware assisted with TSO");
3579 SYSCTL_ADD_COUNTER_U64(ctx, SYSCTL_CHILDREN(node), OID_AUTO, "rxcsum",
3580 CTLFLAG_RD, &stats->rxcsum,
3581 "# of times hardware assisted with rx checksum");
3585 vxlan_sysctl_destroy(struct vxlan_softc *sc)
3588 sysctl_ctx_free(&sc->vxl_sysctl_ctx);
3589 sc->vxl_sysctl_node = NULL;
3593 vxlan_tunable_int(struct vxlan_softc *sc, const char *knob, int def)
3597 snprintf(path, sizeof(path), "net.link.vxlan.%d.%s",
3598 sc->vxl_unit, knob);
3599 TUNABLE_INT_FETCH(path, &def);
3605 vxlan_ifdetach_event(void *arg __unused, struct ifnet *ifp)
3607 struct vxlan_softc_head list;
3608 struct vxlan_socket *vso;
3609 struct vxlan_softc *sc, *tsc;
3613 if (ifp->if_flags & IFF_RENAMING)
3615 if ((ifp->if_flags & IFF_MULTICAST) == 0)
3619 LIST_FOREACH(vso, &vxlan_socket_list, vxlso_entry)
3620 vxlan_socket_ifdetach(vso, ifp, &list);
3621 VXLAN_LIST_UNLOCK();
3623 LIST_FOREACH_SAFE(sc, &list, vxl_ifdetach_list, tsc) {
3624 LIST_REMOVE(sc, vxl_ifdetach_list);
3626 sx_xlock(&vxlan_sx);
3628 if (sc->vxl_flags & VXLAN_FLAG_INIT)
3629 vxlan_init_wait(sc);
3630 vxlan_teardown_locked(sc);
3631 sx_xunlock(&vxlan_sx);
3639 mtx_init(&vxlan_list_mtx, "vxlan list", NULL, MTX_DEF);
3640 LIST_INIT(&vxlan_socket_list);
3641 vxlan_ifdetach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
3642 vxlan_ifdetach_event, NULL, EVENTHANDLER_PRI_ANY);
3643 vxlan_cloner = if_clone_simple(vxlan_name, vxlan_clone_create,
3644 vxlan_clone_destroy, 0);
3651 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
3652 vxlan_ifdetach_event_tag);
3653 if_clone_detach(vxlan_cloner);
3654 mtx_destroy(&vxlan_list_mtx);
3655 MPASS(LIST_EMPTY(&vxlan_socket_list));
3659 vxlan_modevent(module_t mod, int type, void *unused)
3680 static moduledata_t vxlan_mod = {
3686 DECLARE_MODULE(if_vxlan, vxlan_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
3687 MODULE_VERSION(if_vxlan, 1);